diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,208433 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 29.999740338425585, + "eval_steps": 500, + "global_step": 346590, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 1.0000000000000002e-06, + "loss": 10.6393, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 2.0000000000000003e-06, + "loss": 10.6367, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 3e-06, + "loss": 10.6305, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 4.000000000000001e-06, + "loss": 10.6293, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 5e-06, + "loss": 10.6242, + "step": 50 + }, + { + "epoch": 0.01, + "learning_rate": 6e-06, + "loss": 10.6158, + "step": 60 + }, + { + "epoch": 0.01, + "learning_rate": 7.000000000000001e-06, + "loss": 10.6073, + "step": 70 + }, + { + "epoch": 0.01, + "learning_rate": 8.000000000000001e-06, + "loss": 10.5993, + "step": 80 + }, + { + "epoch": 0.01, + "learning_rate": 9e-06, + "loss": 10.5906, + "step": 90 + }, + { + "epoch": 0.01, + "learning_rate": 1e-05, + "loss": 10.5711, + "step": 100 + }, + { + "epoch": 0.01, + "learning_rate": 1.1000000000000001e-05, + "loss": 10.5495, + "step": 110 + }, + { + "epoch": 0.01, + "learning_rate": 1.2e-05, + "loss": 10.5289, + "step": 120 + }, + { + "epoch": 0.01, + "learning_rate": 1.3000000000000001e-05, + "loss": 10.4863, + "step": 130 + }, + { + "epoch": 0.01, + "learning_rate": 1.4000000000000001e-05, + "loss": 10.4132, + "step": 140 + }, + { + "epoch": 0.01, + "learning_rate": 1.5e-05, + "loss": 10.2532, + "step": 150 + }, + { + "epoch": 0.01, + "learning_rate": 1.6000000000000003e-05, + "loss": 9.9767, + "step": 160 + }, + { + "epoch": 0.01, + "learning_rate": 1.7000000000000003e-05, + "loss": 9.707, + "step": 170 + }, + { + "epoch": 0.02, + "learning_rate": 1.8e-05, + "loss": 9.3463, + "step": 180 + }, + { + "epoch": 0.02, + "learning_rate": 1.9e-05, + "loss": 9.0405, + "step": 190 + }, + { + "epoch": 0.02, + "learning_rate": 2e-05, + "loss": 8.6773, + "step": 200 + }, + { + "epoch": 0.02, + "learning_rate": 2.1e-05, + "loss": 8.3523, + "step": 210 + }, + { + "epoch": 0.02, + "learning_rate": 2.2000000000000003e-05, + "loss": 8.0151, + "step": 220 + }, + { + "epoch": 0.02, + "learning_rate": 2.3000000000000003e-05, + "loss": 7.6284, + "step": 230 + }, + { + "epoch": 0.02, + "learning_rate": 2.4e-05, + "loss": 7.2949, + "step": 240 + }, + { + "epoch": 0.02, + "learning_rate": 2.5e-05, + "loss": 6.9695, + "step": 250 + }, + { + "epoch": 0.02, + "learning_rate": 2.6000000000000002e-05, + "loss": 6.6407, + "step": 260 + }, + { + "epoch": 0.02, + "learning_rate": 2.7000000000000002e-05, + "loss": 6.3143, + "step": 270 + }, + { + "epoch": 0.02, + "learning_rate": 2.8000000000000003e-05, + "loss": 6.0219, + "step": 280 + }, + { + "epoch": 0.03, + "learning_rate": 2.9e-05, + "loss": 5.8095, + "step": 290 + }, + { + "epoch": 0.03, + "learning_rate": 3e-05, + "loss": 5.6021, + "step": 300 + }, + { + "epoch": 0.03, + "learning_rate": 3.1e-05, + "loss": 5.4454, + "step": 310 + }, + { + "epoch": 0.03, + "learning_rate": 3.2000000000000005e-05, + "loss": 5.418, + "step": 320 + }, + { + "epoch": 0.03, + "learning_rate": 3.3e-05, + "loss": 5.2787, + "step": 330 + }, + { + "epoch": 0.03, + "learning_rate": 3.4000000000000007e-05, + "loss": 5.2464, + "step": 340 + }, + { + "epoch": 0.03, + "learning_rate": 3.5e-05, + "loss": 5.1882, + "step": 350 + }, + { + "epoch": 0.03, + "learning_rate": 3.6e-05, + "loss": 5.2197, + "step": 360 + }, + { + "epoch": 0.03, + "learning_rate": 3.7e-05, + "loss": 5.1366, + "step": 370 + }, + { + "epoch": 0.03, + "learning_rate": 3.8e-05, + "loss": 5.184, + "step": 380 + }, + { + "epoch": 0.03, + "learning_rate": 3.9000000000000006e-05, + "loss": 5.1055, + "step": 390 + }, + { + "epoch": 0.03, + "learning_rate": 4e-05, + "loss": 5.0505, + "step": 400 + }, + { + "epoch": 0.04, + "learning_rate": 4.1e-05, + "loss": 5.0544, + "step": 410 + }, + { + "epoch": 0.04, + "learning_rate": 4.2e-05, + "loss": 5.0745, + "step": 420 + }, + { + "epoch": 0.04, + "learning_rate": 4.3e-05, + "loss": 5.021, + "step": 430 + }, + { + "epoch": 0.04, + "learning_rate": 4.4000000000000006e-05, + "loss": 4.9369, + "step": 440 + }, + { + "epoch": 0.04, + "learning_rate": 4.5e-05, + "loss": 4.9577, + "step": 450 + }, + { + "epoch": 0.04, + "learning_rate": 4.600000000000001e-05, + "loss": 4.8924, + "step": 460 + }, + { + "epoch": 0.04, + "learning_rate": 4.7e-05, + "loss": 4.7812, + "step": 470 + }, + { + "epoch": 0.04, + "learning_rate": 4.8e-05, + "loss": 4.7724, + "step": 480 + }, + { + "epoch": 0.04, + "learning_rate": 4.9e-05, + "loss": 4.7002, + "step": 490 + }, + { + "epoch": 0.04, + "learning_rate": 5e-05, + "loss": 4.6201, + "step": 500 + }, + { + "epoch": 0.04, + "learning_rate": 4.99991336740882e-05, + "loss": 4.59, + "step": 510 + }, + { + "epoch": 0.05, + "learning_rate": 4.9998267348176385e-05, + "loss": 4.485, + "step": 520 + }, + { + "epoch": 0.05, + "learning_rate": 4.999740102226458e-05, + "loss": 4.3719, + "step": 530 + }, + { + "epoch": 0.05, + "learning_rate": 4.9996534696352773e-05, + "loss": 4.2912, + "step": 540 + }, + { + "epoch": 0.05, + "learning_rate": 4.999566837044096e-05, + "loss": 4.1822, + "step": 550 + }, + { + "epoch": 0.05, + "learning_rate": 4.9994802044529156e-05, + "loss": 4.0741, + "step": 560 + }, + { + "epoch": 0.05, + "learning_rate": 4.999393571861734e-05, + "loss": 3.9622, + "step": 570 + }, + { + "epoch": 0.05, + "learning_rate": 4.999306939270554e-05, + "loss": 3.9085, + "step": 580 + }, + { + "epoch": 0.05, + "learning_rate": 4.999220306679373e-05, + "loss": 3.7959, + "step": 590 + }, + { + "epoch": 0.05, + "learning_rate": 4.999133674088192e-05, + "loss": 3.7415, + "step": 600 + }, + { + "epoch": 0.05, + "learning_rate": 4.9990470414970114e-05, + "loss": 3.6459, + "step": 610 + }, + { + "epoch": 0.05, + "learning_rate": 4.998960408905831e-05, + "loss": 3.5916, + "step": 620 + }, + { + "epoch": 0.05, + "learning_rate": 4.9988737763146496e-05, + "loss": 3.5063, + "step": 630 + }, + { + "epoch": 0.06, + "learning_rate": 4.998787143723469e-05, + "loss": 3.4583, + "step": 640 + }, + { + "epoch": 0.06, + "learning_rate": 4.9987005111322885e-05, + "loss": 3.4009, + "step": 650 + }, + { + "epoch": 0.06, + "learning_rate": 4.998613878541107e-05, + "loss": 3.3349, + "step": 660 + }, + { + "epoch": 0.06, + "learning_rate": 4.998527245949927e-05, + "loss": 3.2288, + "step": 670 + }, + { + "epoch": 0.06, + "learning_rate": 4.9984406133587455e-05, + "loss": 3.2067, + "step": 680 + }, + { + "epoch": 0.06, + "learning_rate": 4.998353980767565e-05, + "loss": 3.2189, + "step": 690 + }, + { + "epoch": 0.06, + "learning_rate": 4.9982673481763844e-05, + "loss": 3.2221, + "step": 700 + }, + { + "epoch": 0.06, + "learning_rate": 4.998180715585203e-05, + "loss": 3.2056, + "step": 710 + }, + { + "epoch": 0.06, + "learning_rate": 4.9980940829940226e-05, + "loss": 3.1658, + "step": 720 + }, + { + "epoch": 0.06, + "learning_rate": 4.998007450402842e-05, + "loss": 3.1112, + "step": 730 + }, + { + "epoch": 0.06, + "learning_rate": 4.997920817811661e-05, + "loss": 3.1642, + "step": 740 + }, + { + "epoch": 0.06, + "learning_rate": 4.99783418522048e-05, + "loss": 3.0606, + "step": 750 + }, + { + "epoch": 0.07, + "learning_rate": 4.9977475526293e-05, + "loss": 3.0609, + "step": 760 + }, + { + "epoch": 0.07, + "learning_rate": 4.9976609200381185e-05, + "loss": 2.9935, + "step": 770 + }, + { + "epoch": 0.07, + "learning_rate": 4.997574287446938e-05, + "loss": 3.0259, + "step": 780 + }, + { + "epoch": 0.07, + "learning_rate": 4.997487654855757e-05, + "loss": 3.0359, + "step": 790 + }, + { + "epoch": 0.07, + "learning_rate": 4.997401022264576e-05, + "loss": 3.0148, + "step": 800 + }, + { + "epoch": 0.07, + "learning_rate": 4.9973143896733956e-05, + "loss": 2.9766, + "step": 810 + }, + { + "epoch": 0.07, + "learning_rate": 4.9972277570822144e-05, + "loss": 2.9121, + "step": 820 + }, + { + "epoch": 0.07, + "learning_rate": 4.997141124491034e-05, + "loss": 2.8921, + "step": 830 + }, + { + "epoch": 0.07, + "learning_rate": 4.997054491899853e-05, + "loss": 2.8877, + "step": 840 + }, + { + "epoch": 0.07, + "learning_rate": 4.996967859308672e-05, + "loss": 2.9138, + "step": 850 + }, + { + "epoch": 0.07, + "learning_rate": 4.9968812267174915e-05, + "loss": 2.9481, + "step": 860 + }, + { + "epoch": 0.08, + "learning_rate": 4.996794594126311e-05, + "loss": 2.9451, + "step": 870 + }, + { + "epoch": 0.08, + "learning_rate": 4.99670796153513e-05, + "loss": 2.8693, + "step": 880 + }, + { + "epoch": 0.08, + "learning_rate": 4.996621328943949e-05, + "loss": 2.9261, + "step": 890 + }, + { + "epoch": 0.08, + "learning_rate": 4.996534696352768e-05, + "loss": 2.9141, + "step": 900 + }, + { + "epoch": 0.08, + "learning_rate": 4.996448063761587e-05, + "loss": 2.9081, + "step": 910 + }, + { + "epoch": 0.08, + "learning_rate": 4.996361431170407e-05, + "loss": 2.9177, + "step": 920 + }, + { + "epoch": 0.08, + "learning_rate": 4.9962747985792255e-05, + "loss": 2.9061, + "step": 930 + }, + { + "epoch": 0.08, + "learning_rate": 4.996188165988045e-05, + "loss": 2.8218, + "step": 940 + }, + { + "epoch": 0.08, + "learning_rate": 4.9961015333968644e-05, + "loss": 2.7609, + "step": 950 + }, + { + "epoch": 0.08, + "learning_rate": 4.996014900805683e-05, + "loss": 2.9357, + "step": 960 + }, + { + "epoch": 0.08, + "learning_rate": 4.9959282682145026e-05, + "loss": 2.8294, + "step": 970 + }, + { + "epoch": 0.08, + "learning_rate": 4.9958416356233214e-05, + "loss": 2.8549, + "step": 980 + }, + { + "epoch": 0.09, + "learning_rate": 4.995755003032141e-05, + "loss": 2.8172, + "step": 990 + }, + { + "epoch": 0.09, + "learning_rate": 4.99566837044096e-05, + "loss": 2.7918, + "step": 1000 + }, + { + "epoch": 0.09, + "learning_rate": 4.995581737849779e-05, + "loss": 2.8, + "step": 1010 + }, + { + "epoch": 0.09, + "learning_rate": 4.9954951052585985e-05, + "loss": 2.7964, + "step": 1020 + }, + { + "epoch": 0.09, + "learning_rate": 4.995408472667418e-05, + "loss": 2.8003, + "step": 1030 + }, + { + "epoch": 0.09, + "learning_rate": 4.995321840076237e-05, + "loss": 2.7529, + "step": 1040 + }, + { + "epoch": 0.09, + "learning_rate": 4.995235207485056e-05, + "loss": 2.7552, + "step": 1050 + }, + { + "epoch": 0.09, + "learning_rate": 4.9951485748938756e-05, + "loss": 2.7623, + "step": 1060 + }, + { + "epoch": 0.09, + "learning_rate": 4.9950619423026944e-05, + "loss": 2.8309, + "step": 1070 + }, + { + "epoch": 0.09, + "learning_rate": 4.994975309711514e-05, + "loss": 2.7898, + "step": 1080 + }, + { + "epoch": 0.09, + "learning_rate": 4.9948886771203326e-05, + "loss": 2.7839, + "step": 1090 + }, + { + "epoch": 0.1, + "learning_rate": 4.994802044529152e-05, + "loss": 2.7784, + "step": 1100 + }, + { + "epoch": 0.1, + "learning_rate": 4.9947154119379715e-05, + "loss": 2.713, + "step": 1110 + }, + { + "epoch": 0.1, + "learning_rate": 4.99462877934679e-05, + "loss": 2.7634, + "step": 1120 + }, + { + "epoch": 0.1, + "learning_rate": 4.99454214675561e-05, + "loss": 2.7343, + "step": 1130 + }, + { + "epoch": 0.1, + "learning_rate": 4.994455514164429e-05, + "loss": 2.7261, + "step": 1140 + }, + { + "epoch": 0.1, + "learning_rate": 4.994368881573248e-05, + "loss": 2.76, + "step": 1150 + }, + { + "epoch": 0.1, + "learning_rate": 4.9942822489820674e-05, + "loss": 2.7718, + "step": 1160 + }, + { + "epoch": 0.1, + "learning_rate": 4.994195616390887e-05, + "loss": 2.746, + "step": 1170 + }, + { + "epoch": 0.1, + "learning_rate": 4.9941089837997056e-05, + "loss": 2.753, + "step": 1180 + }, + { + "epoch": 0.1, + "learning_rate": 4.994022351208525e-05, + "loss": 2.6916, + "step": 1190 + }, + { + "epoch": 0.1, + "learning_rate": 4.993935718617344e-05, + "loss": 2.7351, + "step": 1200 + }, + { + "epoch": 0.1, + "learning_rate": 4.993849086026163e-05, + "loss": 2.7632, + "step": 1210 + }, + { + "epoch": 0.11, + "learning_rate": 4.993762453434983e-05, + "loss": 2.8038, + "step": 1220 + }, + { + "epoch": 0.11, + "learning_rate": 4.9936758208438014e-05, + "loss": 2.7188, + "step": 1230 + }, + { + "epoch": 0.11, + "learning_rate": 4.993589188252621e-05, + "loss": 2.7334, + "step": 1240 + }, + { + "epoch": 0.11, + "learning_rate": 4.9935025556614403e-05, + "loss": 2.7318, + "step": 1250 + }, + { + "epoch": 0.11, + "learning_rate": 4.993415923070259e-05, + "loss": 2.7105, + "step": 1260 + }, + { + "epoch": 0.11, + "learning_rate": 4.9933292904790786e-05, + "loss": 2.695, + "step": 1270 + }, + { + "epoch": 0.11, + "learning_rate": 4.993242657887898e-05, + "loss": 2.7341, + "step": 1280 + }, + { + "epoch": 0.11, + "learning_rate": 4.993156025296717e-05, + "loss": 2.6949, + "step": 1290 + }, + { + "epoch": 0.11, + "learning_rate": 4.993069392705536e-05, + "loss": 2.715, + "step": 1300 + }, + { + "epoch": 0.11, + "learning_rate": 4.992982760114355e-05, + "loss": 2.6706, + "step": 1310 + }, + { + "epoch": 0.11, + "learning_rate": 4.9928961275231744e-05, + "loss": 2.6982, + "step": 1320 + }, + { + "epoch": 0.12, + "learning_rate": 4.992809494931994e-05, + "loss": 2.6565, + "step": 1330 + }, + { + "epoch": 0.12, + "learning_rate": 4.9927228623408126e-05, + "loss": 2.7191, + "step": 1340 + }, + { + "epoch": 0.12, + "learning_rate": 4.992636229749632e-05, + "loss": 2.6837, + "step": 1350 + }, + { + "epoch": 0.12, + "learning_rate": 4.9925495971584515e-05, + "loss": 2.7297, + "step": 1360 + }, + { + "epoch": 0.12, + "learning_rate": 4.99246296456727e-05, + "loss": 2.677, + "step": 1370 + }, + { + "epoch": 0.12, + "learning_rate": 4.99237633197609e-05, + "loss": 2.6336, + "step": 1380 + }, + { + "epoch": 0.12, + "learning_rate": 4.992289699384909e-05, + "loss": 2.6785, + "step": 1390 + }, + { + "epoch": 0.12, + "learning_rate": 4.992203066793728e-05, + "loss": 2.6466, + "step": 1400 + }, + { + "epoch": 0.12, + "learning_rate": 4.9921164342025474e-05, + "loss": 2.7148, + "step": 1410 + }, + { + "epoch": 0.12, + "learning_rate": 4.992029801611366e-05, + "loss": 2.7681, + "step": 1420 + }, + { + "epoch": 0.12, + "learning_rate": 4.9919431690201856e-05, + "loss": 2.675, + "step": 1430 + }, + { + "epoch": 0.12, + "learning_rate": 4.991856536429005e-05, + "loss": 2.6498, + "step": 1440 + }, + { + "epoch": 0.13, + "learning_rate": 4.991769903837824e-05, + "loss": 2.6845, + "step": 1450 + }, + { + "epoch": 0.13, + "learning_rate": 4.991683271246643e-05, + "loss": 2.7006, + "step": 1460 + }, + { + "epoch": 0.13, + "learning_rate": 4.991596638655463e-05, + "loss": 2.6537, + "step": 1470 + }, + { + "epoch": 0.13, + "learning_rate": 4.9915100060642815e-05, + "loss": 2.594, + "step": 1480 + }, + { + "epoch": 0.13, + "learning_rate": 4.991423373473101e-05, + "loss": 2.6712, + "step": 1490 + }, + { + "epoch": 0.13, + "learning_rate": 4.9913367408819204e-05, + "loss": 2.6429, + "step": 1500 + }, + { + "epoch": 0.13, + "learning_rate": 4.991250108290739e-05, + "loss": 2.6856, + "step": 1510 + }, + { + "epoch": 0.13, + "learning_rate": 4.9911634756995586e-05, + "loss": 2.6683, + "step": 1520 + }, + { + "epoch": 0.13, + "learning_rate": 4.9910768431083774e-05, + "loss": 2.6141, + "step": 1530 + }, + { + "epoch": 0.13, + "learning_rate": 4.990990210517197e-05, + "loss": 2.6605, + "step": 1540 + }, + { + "epoch": 0.13, + "learning_rate": 4.990903577926016e-05, + "loss": 2.6187, + "step": 1550 + }, + { + "epoch": 0.14, + "learning_rate": 4.990816945334835e-05, + "loss": 2.6334, + "step": 1560 + }, + { + "epoch": 0.14, + "learning_rate": 4.9907303127436545e-05, + "loss": 2.6158, + "step": 1570 + }, + { + "epoch": 0.14, + "learning_rate": 4.990643680152474e-05, + "loss": 2.6909, + "step": 1580 + }, + { + "epoch": 0.14, + "learning_rate": 4.990557047561293e-05, + "loss": 2.6403, + "step": 1590 + }, + { + "epoch": 0.14, + "learning_rate": 4.990470414970112e-05, + "loss": 2.6368, + "step": 1600 + }, + { + "epoch": 0.14, + "learning_rate": 4.9903837823789316e-05, + "loss": 2.6126, + "step": 1610 + }, + { + "epoch": 0.14, + "learning_rate": 4.99029714978775e-05, + "loss": 2.594, + "step": 1620 + }, + { + "epoch": 0.14, + "learning_rate": 4.99021051719657e-05, + "loss": 2.5908, + "step": 1630 + }, + { + "epoch": 0.14, + "learning_rate": 4.9901238846053885e-05, + "loss": 2.6545, + "step": 1640 + }, + { + "epoch": 0.14, + "learning_rate": 4.990037252014208e-05, + "loss": 2.5523, + "step": 1650 + }, + { + "epoch": 0.14, + "learning_rate": 4.9899506194230274e-05, + "loss": 2.6141, + "step": 1660 + }, + { + "epoch": 0.14, + "learning_rate": 4.989863986831846e-05, + "loss": 2.5962, + "step": 1670 + }, + { + "epoch": 0.15, + "learning_rate": 4.9897773542406656e-05, + "loss": 2.6104, + "step": 1680 + }, + { + "epoch": 0.15, + "learning_rate": 4.989690721649485e-05, + "loss": 2.5786, + "step": 1690 + }, + { + "epoch": 0.15, + "learning_rate": 4.989604089058304e-05, + "loss": 2.5823, + "step": 1700 + }, + { + "epoch": 0.15, + "learning_rate": 4.989517456467123e-05, + "loss": 2.6165, + "step": 1710 + }, + { + "epoch": 0.15, + "learning_rate": 4.989430823875942e-05, + "loss": 2.6198, + "step": 1720 + }, + { + "epoch": 0.15, + "learning_rate": 4.9893441912847615e-05, + "loss": 2.6019, + "step": 1730 + }, + { + "epoch": 0.15, + "learning_rate": 4.989257558693581e-05, + "loss": 2.5878, + "step": 1740 + }, + { + "epoch": 0.15, + "learning_rate": 4.9891709261024e-05, + "loss": 2.6259, + "step": 1750 + }, + { + "epoch": 0.15, + "learning_rate": 4.989084293511219e-05, + "loss": 2.6292, + "step": 1760 + }, + { + "epoch": 0.15, + "learning_rate": 4.9889976609200386e-05, + "loss": 2.5931, + "step": 1770 + }, + { + "epoch": 0.15, + "learning_rate": 4.9889110283288574e-05, + "loss": 2.5983, + "step": 1780 + }, + { + "epoch": 0.15, + "learning_rate": 4.988824395737677e-05, + "loss": 2.5529, + "step": 1790 + }, + { + "epoch": 0.16, + "learning_rate": 4.988737763146496e-05, + "loss": 2.5633, + "step": 1800 + }, + { + "epoch": 0.16, + "learning_rate": 4.988651130555315e-05, + "loss": 2.583, + "step": 1810 + }, + { + "epoch": 0.16, + "learning_rate": 4.9885644979641345e-05, + "loss": 2.5693, + "step": 1820 + }, + { + "epoch": 0.16, + "learning_rate": 4.988477865372953e-05, + "loss": 2.5778, + "step": 1830 + }, + { + "epoch": 0.16, + "learning_rate": 4.988391232781773e-05, + "loss": 2.6049, + "step": 1840 + }, + { + "epoch": 0.16, + "learning_rate": 4.988304600190592e-05, + "loss": 2.5446, + "step": 1850 + }, + { + "epoch": 0.16, + "learning_rate": 4.988217967599411e-05, + "loss": 2.5778, + "step": 1860 + }, + { + "epoch": 0.16, + "learning_rate": 4.9881313350082304e-05, + "loss": 2.6339, + "step": 1870 + }, + { + "epoch": 0.16, + "learning_rate": 4.98804470241705e-05, + "loss": 2.6104, + "step": 1880 + }, + { + "epoch": 0.16, + "learning_rate": 4.9879580698258686e-05, + "loss": 2.5984, + "step": 1890 + }, + { + "epoch": 0.16, + "learning_rate": 4.987871437234688e-05, + "loss": 2.5663, + "step": 1900 + }, + { + "epoch": 0.17, + "learning_rate": 4.9877848046435075e-05, + "loss": 2.5594, + "step": 1910 + }, + { + "epoch": 0.17, + "learning_rate": 4.987698172052326e-05, + "loss": 2.5702, + "step": 1920 + }, + { + "epoch": 0.17, + "learning_rate": 4.987611539461146e-05, + "loss": 2.5638, + "step": 1930 + }, + { + "epoch": 0.17, + "learning_rate": 4.9875249068699644e-05, + "loss": 2.634, + "step": 1940 + }, + { + "epoch": 0.17, + "learning_rate": 4.987438274278784e-05, + "loss": 2.5681, + "step": 1950 + }, + { + "epoch": 0.17, + "learning_rate": 4.987351641687603e-05, + "loss": 2.5299, + "step": 1960 + }, + { + "epoch": 0.17, + "learning_rate": 4.987265009096422e-05, + "loss": 2.6505, + "step": 1970 + }, + { + "epoch": 0.17, + "learning_rate": 4.9871783765052416e-05, + "loss": 2.527, + "step": 1980 + }, + { + "epoch": 0.17, + "learning_rate": 4.987091743914061e-05, + "loss": 2.5518, + "step": 1990 + }, + { + "epoch": 0.17, + "learning_rate": 4.98700511132288e-05, + "loss": 2.6166, + "step": 2000 + }, + { + "epoch": 0.17, + "learning_rate": 4.986918478731699e-05, + "loss": 2.5745, + "step": 2010 + }, + { + "epoch": 0.17, + "learning_rate": 4.9868318461405187e-05, + "loss": 2.493, + "step": 2020 + }, + { + "epoch": 0.18, + "learning_rate": 4.9867452135493374e-05, + "loss": 2.5437, + "step": 2030 + }, + { + "epoch": 0.18, + "learning_rate": 4.986658580958157e-05, + "loss": 2.497, + "step": 2040 + }, + { + "epoch": 0.18, + "learning_rate": 4.9865719483669756e-05, + "loss": 2.532, + "step": 2050 + }, + { + "epoch": 0.18, + "learning_rate": 4.986485315775795e-05, + "loss": 2.4897, + "step": 2060 + }, + { + "epoch": 0.18, + "learning_rate": 4.9863986831846145e-05, + "loss": 2.5676, + "step": 2070 + }, + { + "epoch": 0.18, + "learning_rate": 4.986312050593433e-05, + "loss": 2.509, + "step": 2080 + }, + { + "epoch": 0.18, + "learning_rate": 4.986225418002253e-05, + "loss": 2.5856, + "step": 2090 + }, + { + "epoch": 0.18, + "learning_rate": 4.986138785411072e-05, + "loss": 2.5146, + "step": 2100 + }, + { + "epoch": 0.18, + "learning_rate": 4.986052152819891e-05, + "loss": 2.512, + "step": 2110 + }, + { + "epoch": 0.18, + "learning_rate": 4.9859655202287104e-05, + "loss": 2.5378, + "step": 2120 + }, + { + "epoch": 0.18, + "learning_rate": 4.98587888763753e-05, + "loss": 2.5464, + "step": 2130 + }, + { + "epoch": 0.19, + "learning_rate": 4.9857922550463486e-05, + "loss": 2.5681, + "step": 2140 + }, + { + "epoch": 0.19, + "learning_rate": 4.985705622455168e-05, + "loss": 2.5316, + "step": 2150 + }, + { + "epoch": 0.19, + "learning_rate": 4.985618989863987e-05, + "loss": 2.563, + "step": 2160 + }, + { + "epoch": 0.19, + "learning_rate": 4.985532357272806e-05, + "loss": 2.531, + "step": 2170 + }, + { + "epoch": 0.19, + "learning_rate": 4.985445724681626e-05, + "loss": 2.5813, + "step": 2180 + }, + { + "epoch": 0.19, + "learning_rate": 4.9853590920904445e-05, + "loss": 2.5307, + "step": 2190 + }, + { + "epoch": 0.19, + "learning_rate": 4.985272459499264e-05, + "loss": 2.4998, + "step": 2200 + }, + { + "epoch": 0.19, + "learning_rate": 4.9851858269080834e-05, + "loss": 2.5171, + "step": 2210 + }, + { + "epoch": 0.19, + "learning_rate": 4.985099194316902e-05, + "loss": 2.5069, + "step": 2220 + }, + { + "epoch": 0.19, + "learning_rate": 4.9850125617257216e-05, + "loss": 2.5234, + "step": 2230 + }, + { + "epoch": 0.19, + "learning_rate": 4.984925929134541e-05, + "loss": 2.5411, + "step": 2240 + }, + { + "epoch": 0.19, + "learning_rate": 4.98483929654336e-05, + "loss": 2.517, + "step": 2250 + }, + { + "epoch": 0.2, + "learning_rate": 4.984752663952179e-05, + "loss": 2.5184, + "step": 2260 + }, + { + "epoch": 0.2, + "learning_rate": 4.984666031360998e-05, + "loss": 2.5385, + "step": 2270 + }, + { + "epoch": 0.2, + "learning_rate": 4.9845793987698175e-05, + "loss": 2.4877, + "step": 2280 + }, + { + "epoch": 0.2, + "learning_rate": 4.984492766178637e-05, + "loss": 2.553, + "step": 2290 + }, + { + "epoch": 0.2, + "learning_rate": 4.984406133587456e-05, + "loss": 2.4522, + "step": 2300 + }, + { + "epoch": 0.2, + "learning_rate": 4.984319500996275e-05, + "loss": 2.523, + "step": 2310 + }, + { + "epoch": 0.2, + "learning_rate": 4.9842328684050946e-05, + "loss": 2.5529, + "step": 2320 + }, + { + "epoch": 0.2, + "learning_rate": 4.984146235813913e-05, + "loss": 2.4951, + "step": 2330 + }, + { + "epoch": 0.2, + "learning_rate": 4.984059603222733e-05, + "loss": 2.5212, + "step": 2340 + }, + { + "epoch": 0.2, + "learning_rate": 4.983972970631552e-05, + "loss": 2.4968, + "step": 2350 + }, + { + "epoch": 0.2, + "learning_rate": 4.983886338040371e-05, + "loss": 2.4742, + "step": 2360 + }, + { + "epoch": 0.21, + "learning_rate": 4.9837997054491904e-05, + "loss": 2.4706, + "step": 2370 + }, + { + "epoch": 0.21, + "learning_rate": 4.983713072858009e-05, + "loss": 2.5046, + "step": 2380 + }, + { + "epoch": 0.21, + "learning_rate": 4.9836264402668286e-05, + "loss": 2.5174, + "step": 2390 + }, + { + "epoch": 0.21, + "learning_rate": 4.983539807675648e-05, + "loss": 2.5399, + "step": 2400 + }, + { + "epoch": 0.21, + "learning_rate": 4.983453175084467e-05, + "loss": 2.5119, + "step": 2410 + }, + { + "epoch": 0.21, + "learning_rate": 4.983366542493286e-05, + "loss": 2.4468, + "step": 2420 + }, + { + "epoch": 0.21, + "learning_rate": 4.983279909902106e-05, + "loss": 2.4954, + "step": 2430 + }, + { + "epoch": 0.21, + "learning_rate": 4.9831932773109245e-05, + "loss": 2.4443, + "step": 2440 + }, + { + "epoch": 0.21, + "learning_rate": 4.983106644719744e-05, + "loss": 2.4747, + "step": 2450 + }, + { + "epoch": 0.21, + "learning_rate": 4.983020012128563e-05, + "loss": 2.4706, + "step": 2460 + }, + { + "epoch": 0.21, + "learning_rate": 4.982933379537382e-05, + "loss": 2.424, + "step": 2470 + }, + { + "epoch": 0.21, + "learning_rate": 4.9828467469462016e-05, + "loss": 2.4891, + "step": 2480 + }, + { + "epoch": 0.22, + "learning_rate": 4.9827601143550204e-05, + "loss": 2.4643, + "step": 2490 + }, + { + "epoch": 0.22, + "learning_rate": 4.98267348176384e-05, + "loss": 2.4903, + "step": 2500 + }, + { + "epoch": 0.22, + "learning_rate": 4.982586849172659e-05, + "loss": 2.4793, + "step": 2510 + }, + { + "epoch": 0.22, + "learning_rate": 4.982500216581478e-05, + "loss": 2.4704, + "step": 2520 + }, + { + "epoch": 0.22, + "learning_rate": 4.9824135839902975e-05, + "loss": 2.4896, + "step": 2530 + }, + { + "epoch": 0.22, + "learning_rate": 4.982326951399117e-05, + "loss": 2.4971, + "step": 2540 + }, + { + "epoch": 0.22, + "learning_rate": 4.982240318807936e-05, + "loss": 2.5232, + "step": 2550 + }, + { + "epoch": 0.22, + "learning_rate": 4.982153686216755e-05, + "loss": 2.4904, + "step": 2560 + }, + { + "epoch": 0.22, + "learning_rate": 4.982067053625574e-05, + "loss": 2.495, + "step": 2570 + }, + { + "epoch": 0.22, + "learning_rate": 4.9819804210343934e-05, + "loss": 2.4947, + "step": 2580 + }, + { + "epoch": 0.22, + "learning_rate": 4.981893788443213e-05, + "loss": 2.5046, + "step": 2590 + }, + { + "epoch": 0.23, + "learning_rate": 4.9818071558520316e-05, + "loss": 2.4807, + "step": 2600 + }, + { + "epoch": 0.23, + "learning_rate": 4.981720523260851e-05, + "loss": 2.4955, + "step": 2610 + }, + { + "epoch": 0.23, + "learning_rate": 4.9816338906696705e-05, + "loss": 2.512, + "step": 2620 + }, + { + "epoch": 0.23, + "learning_rate": 4.981547258078489e-05, + "loss": 2.4981, + "step": 2630 + }, + { + "epoch": 0.23, + "learning_rate": 4.981460625487309e-05, + "loss": 2.4541, + "step": 2640 + }, + { + "epoch": 0.23, + "learning_rate": 4.981373992896128e-05, + "loss": 2.5117, + "step": 2650 + }, + { + "epoch": 0.23, + "learning_rate": 4.981287360304947e-05, + "loss": 2.4435, + "step": 2660 + }, + { + "epoch": 0.23, + "learning_rate": 4.981200727713766e-05, + "loss": 2.4846, + "step": 2670 + }, + { + "epoch": 0.23, + "learning_rate": 4.981114095122585e-05, + "loss": 2.4477, + "step": 2680 + }, + { + "epoch": 0.23, + "learning_rate": 4.9810274625314045e-05, + "loss": 2.4756, + "step": 2690 + }, + { + "epoch": 0.23, + "learning_rate": 4.980940829940224e-05, + "loss": 2.4202, + "step": 2700 + }, + { + "epoch": 0.23, + "learning_rate": 4.980854197349043e-05, + "loss": 2.4439, + "step": 2710 + }, + { + "epoch": 0.24, + "learning_rate": 4.980767564757862e-05, + "loss": 2.4767, + "step": 2720 + }, + { + "epoch": 0.24, + "learning_rate": 4.9806809321666817e-05, + "loss": 2.4367, + "step": 2730 + }, + { + "epoch": 0.24, + "learning_rate": 4.9805942995755004e-05, + "loss": 2.4481, + "step": 2740 + }, + { + "epoch": 0.24, + "learning_rate": 4.98050766698432e-05, + "loss": 2.4895, + "step": 2750 + }, + { + "epoch": 0.24, + "learning_rate": 4.980421034393139e-05, + "loss": 2.4848, + "step": 2760 + }, + { + "epoch": 0.24, + "learning_rate": 4.980334401801958e-05, + "loss": 2.4777, + "step": 2770 + }, + { + "epoch": 0.24, + "learning_rate": 4.9802477692107775e-05, + "loss": 2.4349, + "step": 2780 + }, + { + "epoch": 0.24, + "learning_rate": 4.980161136619596e-05, + "loss": 2.4862, + "step": 2790 + }, + { + "epoch": 0.24, + "learning_rate": 4.980074504028416e-05, + "loss": 2.4276, + "step": 2800 + }, + { + "epoch": 0.24, + "learning_rate": 4.979987871437235e-05, + "loss": 2.422, + "step": 2810 + }, + { + "epoch": 0.24, + "learning_rate": 4.979901238846054e-05, + "loss": 2.4325, + "step": 2820 + }, + { + "epoch": 0.24, + "learning_rate": 4.9798146062548734e-05, + "loss": 2.4723, + "step": 2830 + }, + { + "epoch": 0.25, + "learning_rate": 4.979727973663693e-05, + "loss": 2.5144, + "step": 2840 + }, + { + "epoch": 0.25, + "learning_rate": 4.9796413410725116e-05, + "loss": 2.3736, + "step": 2850 + }, + { + "epoch": 0.25, + "learning_rate": 4.979554708481331e-05, + "loss": 2.4835, + "step": 2860 + }, + { + "epoch": 0.25, + "learning_rate": 4.9794680758901505e-05, + "loss": 2.4329, + "step": 2870 + }, + { + "epoch": 0.25, + "learning_rate": 4.979381443298969e-05, + "loss": 2.4754, + "step": 2880 + }, + { + "epoch": 0.25, + "learning_rate": 4.979294810707789e-05, + "loss": 2.4429, + "step": 2890 + }, + { + "epoch": 0.25, + "learning_rate": 4.9792081781166075e-05, + "loss": 2.4739, + "step": 2900 + }, + { + "epoch": 0.25, + "learning_rate": 4.979121545525427e-05, + "loss": 2.4997, + "step": 2910 + }, + { + "epoch": 0.25, + "learning_rate": 4.9790349129342464e-05, + "loss": 2.426, + "step": 2920 + }, + { + "epoch": 0.25, + "learning_rate": 4.978948280343065e-05, + "loss": 2.4717, + "step": 2930 + }, + { + "epoch": 0.25, + "learning_rate": 4.9788616477518846e-05, + "loss": 2.4797, + "step": 2940 + }, + { + "epoch": 0.26, + "learning_rate": 4.978775015160704e-05, + "loss": 2.452, + "step": 2950 + }, + { + "epoch": 0.26, + "learning_rate": 4.978688382569523e-05, + "loss": 2.4478, + "step": 2960 + }, + { + "epoch": 0.26, + "learning_rate": 4.978601749978342e-05, + "loss": 2.4543, + "step": 2970 + }, + { + "epoch": 0.26, + "learning_rate": 4.978515117387162e-05, + "loss": 2.4168, + "step": 2980 + }, + { + "epoch": 0.26, + "learning_rate": 4.9784284847959805e-05, + "loss": 2.4185, + "step": 2990 + }, + { + "epoch": 0.26, + "learning_rate": 4.9783418522048e-05, + "loss": 2.3988, + "step": 3000 + }, + { + "epoch": 0.26, + "learning_rate": 4.978255219613619e-05, + "loss": 2.3971, + "step": 3010 + }, + { + "epoch": 0.26, + "learning_rate": 4.978168587022438e-05, + "loss": 2.4663, + "step": 3020 + }, + { + "epoch": 0.26, + "learning_rate": 4.9780819544312576e-05, + "loss": 2.387, + "step": 3030 + }, + { + "epoch": 0.26, + "learning_rate": 4.977995321840076e-05, + "loss": 2.4633, + "step": 3040 + }, + { + "epoch": 0.26, + "learning_rate": 4.977908689248896e-05, + "loss": 2.3963, + "step": 3050 + }, + { + "epoch": 0.26, + "learning_rate": 4.977822056657715e-05, + "loss": 2.4138, + "step": 3060 + }, + { + "epoch": 0.27, + "learning_rate": 4.977735424066534e-05, + "loss": 2.4274, + "step": 3070 + }, + { + "epoch": 0.27, + "learning_rate": 4.9776487914753534e-05, + "loss": 2.4128, + "step": 3080 + }, + { + "epoch": 0.27, + "learning_rate": 4.977562158884173e-05, + "loss": 2.4151, + "step": 3090 + }, + { + "epoch": 0.27, + "learning_rate": 4.9774755262929916e-05, + "loss": 2.4822, + "step": 3100 + }, + { + "epoch": 0.27, + "learning_rate": 4.977388893701811e-05, + "loss": 2.4119, + "step": 3110 + }, + { + "epoch": 0.27, + "learning_rate": 4.97730226111063e-05, + "loss": 2.3869, + "step": 3120 + }, + { + "epoch": 0.27, + "learning_rate": 4.977215628519449e-05, + "loss": 2.409, + "step": 3130 + }, + { + "epoch": 0.27, + "learning_rate": 4.977128995928269e-05, + "loss": 2.4093, + "step": 3140 + }, + { + "epoch": 0.27, + "learning_rate": 4.9770423633370875e-05, + "loss": 2.4425, + "step": 3150 + }, + { + "epoch": 0.27, + "learning_rate": 4.976955730745907e-05, + "loss": 2.4818, + "step": 3160 + }, + { + "epoch": 0.27, + "learning_rate": 4.9768690981547264e-05, + "loss": 2.4215, + "step": 3170 + }, + { + "epoch": 0.28, + "learning_rate": 4.976782465563545e-05, + "loss": 2.3911, + "step": 3180 + }, + { + "epoch": 0.28, + "learning_rate": 4.9766958329723646e-05, + "loss": 2.4641, + "step": 3190 + }, + { + "epoch": 0.28, + "learning_rate": 4.976609200381184e-05, + "loss": 2.3663, + "step": 3200 + }, + { + "epoch": 0.28, + "learning_rate": 4.976522567790003e-05, + "loss": 2.4394, + "step": 3210 + }, + { + "epoch": 0.28, + "learning_rate": 4.976435935198822e-05, + "loss": 2.3915, + "step": 3220 + }, + { + "epoch": 0.28, + "learning_rate": 4.976349302607641e-05, + "loss": 2.4412, + "step": 3230 + }, + { + "epoch": 0.28, + "learning_rate": 4.9762626700164605e-05, + "loss": 2.4505, + "step": 3240 + }, + { + "epoch": 0.28, + "learning_rate": 4.97617603742528e-05, + "loss": 2.3272, + "step": 3250 + }, + { + "epoch": 0.28, + "learning_rate": 4.976089404834099e-05, + "loss": 2.4048, + "step": 3260 + }, + { + "epoch": 0.28, + "learning_rate": 4.976002772242918e-05, + "loss": 2.4466, + "step": 3270 + }, + { + "epoch": 0.28, + "learning_rate": 4.9759161396517376e-05, + "loss": 2.4097, + "step": 3280 + }, + { + "epoch": 0.28, + "learning_rate": 4.9758295070605564e-05, + "loss": 2.4045, + "step": 3290 + }, + { + "epoch": 0.29, + "learning_rate": 4.975742874469376e-05, + "loss": 2.434, + "step": 3300 + }, + { + "epoch": 0.29, + "learning_rate": 4.9756562418781946e-05, + "loss": 2.4466, + "step": 3310 + }, + { + "epoch": 0.29, + "learning_rate": 4.975569609287014e-05, + "loss": 2.3577, + "step": 3320 + }, + { + "epoch": 0.29, + "learning_rate": 4.9754829766958335e-05, + "loss": 2.4346, + "step": 3330 + }, + { + "epoch": 0.29, + "learning_rate": 4.975396344104652e-05, + "loss": 2.4307, + "step": 3340 + }, + { + "epoch": 0.29, + "learning_rate": 4.975309711513472e-05, + "loss": 2.3629, + "step": 3350 + }, + { + "epoch": 0.29, + "learning_rate": 4.975223078922291e-05, + "loss": 2.4561, + "step": 3360 + }, + { + "epoch": 0.29, + "learning_rate": 4.97513644633111e-05, + "loss": 2.3675, + "step": 3370 + }, + { + "epoch": 0.29, + "learning_rate": 4.975049813739929e-05, + "loss": 2.4197, + "step": 3380 + }, + { + "epoch": 0.29, + "learning_rate": 4.974963181148749e-05, + "loss": 2.3637, + "step": 3390 + }, + { + "epoch": 0.29, + "learning_rate": 4.9748765485575675e-05, + "loss": 2.437, + "step": 3400 + }, + { + "epoch": 0.3, + "learning_rate": 4.974789915966387e-05, + "loss": 2.4176, + "step": 3410 + }, + { + "epoch": 0.3, + "learning_rate": 4.974703283375206e-05, + "loss": 2.3827, + "step": 3420 + }, + { + "epoch": 0.3, + "learning_rate": 4.974616650784025e-05, + "loss": 2.4425, + "step": 3430 + }, + { + "epoch": 0.3, + "learning_rate": 4.9745300181928446e-05, + "loss": 2.4008, + "step": 3440 + }, + { + "epoch": 0.3, + "learning_rate": 4.9744433856016634e-05, + "loss": 2.385, + "step": 3450 + }, + { + "epoch": 0.3, + "learning_rate": 4.974356753010483e-05, + "loss": 2.3698, + "step": 3460 + }, + { + "epoch": 0.3, + "learning_rate": 4.974270120419302e-05, + "loss": 2.3909, + "step": 3470 + }, + { + "epoch": 0.3, + "learning_rate": 4.974183487828121e-05, + "loss": 2.3893, + "step": 3480 + }, + { + "epoch": 0.3, + "learning_rate": 4.9740968552369405e-05, + "loss": 2.3851, + "step": 3490 + }, + { + "epoch": 0.3, + "learning_rate": 4.97401022264576e-05, + "loss": 2.3484, + "step": 3500 + }, + { + "epoch": 0.3, + "learning_rate": 4.973923590054579e-05, + "loss": 2.3615, + "step": 3510 + }, + { + "epoch": 0.3, + "learning_rate": 4.973836957463398e-05, + "loss": 2.4109, + "step": 3520 + }, + { + "epoch": 0.31, + "learning_rate": 4.973750324872217e-05, + "loss": 2.4054, + "step": 3530 + }, + { + "epoch": 0.31, + "learning_rate": 4.9736636922810364e-05, + "loss": 2.4492, + "step": 3540 + }, + { + "epoch": 0.31, + "learning_rate": 4.973577059689856e-05, + "loss": 2.409, + "step": 3550 + }, + { + "epoch": 0.31, + "learning_rate": 4.9734904270986746e-05, + "loss": 2.4141, + "step": 3560 + }, + { + "epoch": 0.31, + "learning_rate": 4.973403794507494e-05, + "loss": 2.3679, + "step": 3570 + }, + { + "epoch": 0.31, + "learning_rate": 4.9733171619163135e-05, + "loss": 2.3691, + "step": 3580 + }, + { + "epoch": 0.31, + "learning_rate": 4.973230529325132e-05, + "loss": 2.3782, + "step": 3590 + }, + { + "epoch": 0.31, + "learning_rate": 4.973143896733952e-05, + "loss": 2.3977, + "step": 3600 + }, + { + "epoch": 0.31, + "learning_rate": 4.973057264142771e-05, + "loss": 2.3582, + "step": 3610 + }, + { + "epoch": 0.31, + "learning_rate": 4.97297063155159e-05, + "loss": 2.3515, + "step": 3620 + }, + { + "epoch": 0.31, + "learning_rate": 4.9728839989604094e-05, + "loss": 2.3335, + "step": 3630 + }, + { + "epoch": 0.32, + "learning_rate": 4.972797366369228e-05, + "loss": 2.3897, + "step": 3640 + }, + { + "epoch": 0.32, + "learning_rate": 4.9727107337780476e-05, + "loss": 2.3143, + "step": 3650 + }, + { + "epoch": 0.32, + "learning_rate": 4.972624101186867e-05, + "loss": 2.399, + "step": 3660 + }, + { + "epoch": 0.32, + "learning_rate": 4.972537468595686e-05, + "loss": 2.3255, + "step": 3670 + }, + { + "epoch": 0.32, + "learning_rate": 4.972450836004505e-05, + "loss": 2.381, + "step": 3680 + }, + { + "epoch": 0.32, + "learning_rate": 4.972364203413325e-05, + "loss": 2.372, + "step": 3690 + }, + { + "epoch": 0.32, + "learning_rate": 4.9722775708221434e-05, + "loss": 2.3555, + "step": 3700 + }, + { + "epoch": 0.32, + "learning_rate": 4.972190938230963e-05, + "loss": 2.3744, + "step": 3710 + }, + { + "epoch": 0.32, + "learning_rate": 4.9721043056397823e-05, + "loss": 2.3448, + "step": 3720 + }, + { + "epoch": 0.32, + "learning_rate": 4.972017673048601e-05, + "loss": 2.4149, + "step": 3730 + }, + { + "epoch": 0.32, + "learning_rate": 4.9719310404574206e-05, + "loss": 2.3565, + "step": 3740 + }, + { + "epoch": 0.32, + "learning_rate": 4.971844407866239e-05, + "loss": 2.4101, + "step": 3750 + }, + { + "epoch": 0.33, + "learning_rate": 4.971757775275059e-05, + "loss": 2.4075, + "step": 3760 + }, + { + "epoch": 0.33, + "learning_rate": 4.971671142683878e-05, + "loss": 2.3764, + "step": 3770 + }, + { + "epoch": 0.33, + "learning_rate": 4.971584510092697e-05, + "loss": 2.3964, + "step": 3780 + }, + { + "epoch": 0.33, + "learning_rate": 4.9714978775015164e-05, + "loss": 2.4057, + "step": 3790 + }, + { + "epoch": 0.33, + "learning_rate": 4.971411244910336e-05, + "loss": 2.3979, + "step": 3800 + }, + { + "epoch": 0.33, + "learning_rate": 4.9713246123191546e-05, + "loss": 2.3848, + "step": 3810 + }, + { + "epoch": 0.33, + "learning_rate": 4.971237979727974e-05, + "loss": 2.3551, + "step": 3820 + }, + { + "epoch": 0.33, + "learning_rate": 4.9711513471367935e-05, + "loss": 2.3641, + "step": 3830 + }, + { + "epoch": 0.33, + "learning_rate": 4.971064714545612e-05, + "loss": 2.3426, + "step": 3840 + }, + { + "epoch": 0.33, + "learning_rate": 4.970978081954432e-05, + "loss": 2.3298, + "step": 3850 + }, + { + "epoch": 0.33, + "learning_rate": 4.9708914493632505e-05, + "loss": 2.3962, + "step": 3860 + }, + { + "epoch": 0.33, + "learning_rate": 4.97080481677207e-05, + "loss": 2.3085, + "step": 3870 + }, + { + "epoch": 0.34, + "learning_rate": 4.9707181841808894e-05, + "loss": 2.3599, + "step": 3880 + }, + { + "epoch": 0.34, + "learning_rate": 4.970631551589708e-05, + "loss": 2.3471, + "step": 3890 + }, + { + "epoch": 0.34, + "learning_rate": 4.9705449189985276e-05, + "loss": 2.3598, + "step": 3900 + }, + { + "epoch": 0.34, + "learning_rate": 4.970458286407347e-05, + "loss": 2.3478, + "step": 3910 + }, + { + "epoch": 0.34, + "learning_rate": 4.970371653816166e-05, + "loss": 2.291, + "step": 3920 + }, + { + "epoch": 0.34, + "learning_rate": 4.970285021224985e-05, + "loss": 2.3581, + "step": 3930 + }, + { + "epoch": 0.34, + "learning_rate": 4.970198388633805e-05, + "loss": 2.3618, + "step": 3940 + }, + { + "epoch": 0.34, + "learning_rate": 4.9701117560426235e-05, + "loss": 2.3703, + "step": 3950 + }, + { + "epoch": 0.34, + "learning_rate": 4.970025123451443e-05, + "loss": 2.2955, + "step": 3960 + }, + { + "epoch": 0.34, + "learning_rate": 4.969938490860262e-05, + "loss": 2.3776, + "step": 3970 + }, + { + "epoch": 0.34, + "learning_rate": 4.969851858269081e-05, + "loss": 2.3808, + "step": 3980 + }, + { + "epoch": 0.35, + "learning_rate": 4.9697652256779006e-05, + "loss": 2.3654, + "step": 3990 + }, + { + "epoch": 0.35, + "learning_rate": 4.9696785930867194e-05, + "loss": 2.3838, + "step": 4000 + }, + { + "epoch": 0.35, + "learning_rate": 4.969591960495539e-05, + "loss": 2.3383, + "step": 4010 + }, + { + "epoch": 0.35, + "learning_rate": 4.969505327904358e-05, + "loss": 2.3143, + "step": 4020 + }, + { + "epoch": 0.35, + "learning_rate": 4.969418695313177e-05, + "loss": 2.2962, + "step": 4030 + }, + { + "epoch": 0.35, + "learning_rate": 4.9693320627219965e-05, + "loss": 2.3014, + "step": 4040 + }, + { + "epoch": 0.35, + "learning_rate": 4.969245430130815e-05, + "loss": 2.3471, + "step": 4050 + }, + { + "epoch": 0.35, + "learning_rate": 4.969158797539635e-05, + "loss": 2.4061, + "step": 4060 + }, + { + "epoch": 0.35, + "learning_rate": 4.969072164948454e-05, + "loss": 2.3454, + "step": 4070 + }, + { + "epoch": 0.35, + "learning_rate": 4.968985532357273e-05, + "loss": 2.392, + "step": 4080 + }, + { + "epoch": 0.35, + "learning_rate": 4.968898899766092e-05, + "loss": 2.332, + "step": 4090 + }, + { + "epoch": 0.35, + "learning_rate": 4.968812267174912e-05, + "loss": 2.3938, + "step": 4100 + }, + { + "epoch": 0.36, + "learning_rate": 4.9687256345837305e-05, + "loss": 2.3241, + "step": 4110 + }, + { + "epoch": 0.36, + "learning_rate": 4.96863900199255e-05, + "loss": 2.3353, + "step": 4120 + }, + { + "epoch": 0.36, + "learning_rate": 4.9685523694013694e-05, + "loss": 2.2991, + "step": 4130 + }, + { + "epoch": 0.36, + "learning_rate": 4.968465736810188e-05, + "loss": 2.3037, + "step": 4140 + }, + { + "epoch": 0.36, + "learning_rate": 4.9683791042190076e-05, + "loss": 2.3217, + "step": 4150 + }, + { + "epoch": 0.36, + "learning_rate": 4.9682924716278264e-05, + "loss": 2.2762, + "step": 4160 + }, + { + "epoch": 0.36, + "learning_rate": 4.968205839036646e-05, + "loss": 2.2996, + "step": 4170 + }, + { + "epoch": 0.36, + "learning_rate": 4.968119206445465e-05, + "loss": 2.3604, + "step": 4180 + }, + { + "epoch": 0.36, + "learning_rate": 4.968032573854284e-05, + "loss": 2.2868, + "step": 4190 + }, + { + "epoch": 0.36, + "learning_rate": 4.9679459412631035e-05, + "loss": 2.3219, + "step": 4200 + }, + { + "epoch": 0.36, + "learning_rate": 4.967859308671923e-05, + "loss": 2.3624, + "step": 4210 + }, + { + "epoch": 0.37, + "learning_rate": 4.967772676080742e-05, + "loss": 2.3168, + "step": 4220 + }, + { + "epoch": 0.37, + "learning_rate": 4.967686043489561e-05, + "loss": 2.3273, + "step": 4230 + }, + { + "epoch": 0.37, + "learning_rate": 4.9675994108983806e-05, + "loss": 2.319, + "step": 4240 + }, + { + "epoch": 0.37, + "learning_rate": 4.9675127783071994e-05, + "loss": 2.2812, + "step": 4250 + }, + { + "epoch": 0.37, + "learning_rate": 4.967426145716019e-05, + "loss": 2.3013, + "step": 4260 + }, + { + "epoch": 0.37, + "learning_rate": 4.9673395131248376e-05, + "loss": 2.375, + "step": 4270 + }, + { + "epoch": 0.37, + "learning_rate": 4.967252880533657e-05, + "loss": 2.3685, + "step": 4280 + }, + { + "epoch": 0.37, + "learning_rate": 4.9671662479424765e-05, + "loss": 2.2761, + "step": 4290 + }, + { + "epoch": 0.37, + "learning_rate": 4.967079615351295e-05, + "loss": 2.3534, + "step": 4300 + }, + { + "epoch": 0.37, + "learning_rate": 4.966992982760115e-05, + "loss": 2.317, + "step": 4310 + }, + { + "epoch": 0.37, + "learning_rate": 4.966906350168934e-05, + "loss": 2.3313, + "step": 4320 + }, + { + "epoch": 0.37, + "learning_rate": 4.966819717577753e-05, + "loss": 2.2767, + "step": 4330 + }, + { + "epoch": 0.38, + "learning_rate": 4.9667330849865724e-05, + "loss": 2.3688, + "step": 4340 + }, + { + "epoch": 0.38, + "learning_rate": 4.966646452395392e-05, + "loss": 2.3625, + "step": 4350 + }, + { + "epoch": 0.38, + "learning_rate": 4.9665598198042106e-05, + "loss": 2.3437, + "step": 4360 + }, + { + "epoch": 0.38, + "learning_rate": 4.96647318721303e-05, + "loss": 2.2875, + "step": 4370 + }, + { + "epoch": 0.38, + "learning_rate": 4.966386554621849e-05, + "loss": 2.3714, + "step": 4380 + }, + { + "epoch": 0.38, + "learning_rate": 4.966299922030668e-05, + "loss": 2.3162, + "step": 4390 + }, + { + "epoch": 0.38, + "learning_rate": 4.966213289439488e-05, + "loss": 2.3814, + "step": 4400 + }, + { + "epoch": 0.38, + "learning_rate": 4.9661266568483064e-05, + "loss": 2.3263, + "step": 4410 + }, + { + "epoch": 0.38, + "learning_rate": 4.966040024257126e-05, + "loss": 2.2933, + "step": 4420 + }, + { + "epoch": 0.38, + "learning_rate": 4.965953391665945e-05, + "loss": 2.346, + "step": 4430 + }, + { + "epoch": 0.38, + "learning_rate": 4.965866759074764e-05, + "loss": 2.3152, + "step": 4440 + }, + { + "epoch": 0.39, + "learning_rate": 4.9657801264835836e-05, + "loss": 2.2478, + "step": 4450 + }, + { + "epoch": 0.39, + "learning_rate": 4.965693493892403e-05, + "loss": 2.3499, + "step": 4460 + }, + { + "epoch": 0.39, + "learning_rate": 4.965606861301222e-05, + "loss": 2.3274, + "step": 4470 + }, + { + "epoch": 0.39, + "learning_rate": 4.965520228710041e-05, + "loss": 2.3144, + "step": 4480 + }, + { + "epoch": 0.39, + "learning_rate": 4.96543359611886e-05, + "loss": 2.3183, + "step": 4490 + }, + { + "epoch": 0.39, + "learning_rate": 4.9653469635276794e-05, + "loss": 2.3199, + "step": 4500 + }, + { + "epoch": 0.39, + "learning_rate": 4.965260330936499e-05, + "loss": 2.3876, + "step": 4510 + }, + { + "epoch": 0.39, + "learning_rate": 4.9651736983453176e-05, + "loss": 2.2911, + "step": 4520 + }, + { + "epoch": 0.39, + "learning_rate": 4.965087065754137e-05, + "loss": 2.3148, + "step": 4530 + }, + { + "epoch": 0.39, + "learning_rate": 4.9650004331629565e-05, + "loss": 2.3024, + "step": 4540 + }, + { + "epoch": 0.39, + "learning_rate": 4.964913800571775e-05, + "loss": 2.2935, + "step": 4550 + }, + { + "epoch": 0.39, + "learning_rate": 4.964827167980595e-05, + "loss": 2.3328, + "step": 4560 + }, + { + "epoch": 0.4, + "learning_rate": 4.964740535389414e-05, + "loss": 2.3379, + "step": 4570 + }, + { + "epoch": 0.4, + "learning_rate": 4.964653902798233e-05, + "loss": 2.2988, + "step": 4580 + }, + { + "epoch": 0.4, + "learning_rate": 4.9645672702070524e-05, + "loss": 2.314, + "step": 4590 + }, + { + "epoch": 0.4, + "learning_rate": 4.964480637615871e-05, + "loss": 2.3192, + "step": 4600 + }, + { + "epoch": 0.4, + "learning_rate": 4.9643940050246906e-05, + "loss": 2.3272, + "step": 4610 + }, + { + "epoch": 0.4, + "learning_rate": 4.96430737243351e-05, + "loss": 2.4031, + "step": 4620 + }, + { + "epoch": 0.4, + "learning_rate": 4.964220739842329e-05, + "loss": 2.3032, + "step": 4630 + }, + { + "epoch": 0.4, + "learning_rate": 4.964134107251148e-05, + "loss": 2.3589, + "step": 4640 + }, + { + "epoch": 0.4, + "learning_rate": 4.964047474659968e-05, + "loss": 2.2819, + "step": 4650 + }, + { + "epoch": 0.4, + "learning_rate": 4.9639608420687865e-05, + "loss": 2.3306, + "step": 4660 + }, + { + "epoch": 0.4, + "learning_rate": 4.963874209477606e-05, + "loss": 2.2807, + "step": 4670 + }, + { + "epoch": 0.41, + "learning_rate": 4.9637875768864254e-05, + "loss": 2.2712, + "step": 4680 + }, + { + "epoch": 0.41, + "learning_rate": 4.963700944295244e-05, + "loss": 2.3311, + "step": 4690 + }, + { + "epoch": 0.41, + "learning_rate": 4.9636143117040636e-05, + "loss": 2.2921, + "step": 4700 + }, + { + "epoch": 0.41, + "learning_rate": 4.9635276791128824e-05, + "loss": 2.2539, + "step": 4710 + }, + { + "epoch": 0.41, + "learning_rate": 4.963441046521702e-05, + "loss": 2.3421, + "step": 4720 + }, + { + "epoch": 0.41, + "learning_rate": 4.963354413930521e-05, + "loss": 2.3043, + "step": 4730 + }, + { + "epoch": 0.41, + "learning_rate": 4.96326778133934e-05, + "loss": 2.3051, + "step": 4740 + }, + { + "epoch": 0.41, + "learning_rate": 4.9631811487481595e-05, + "loss": 2.295, + "step": 4750 + }, + { + "epoch": 0.41, + "learning_rate": 4.963094516156979e-05, + "loss": 2.283, + "step": 4760 + }, + { + "epoch": 0.41, + "learning_rate": 4.963007883565798e-05, + "loss": 2.3308, + "step": 4770 + }, + { + "epoch": 0.41, + "learning_rate": 4.962921250974617e-05, + "loss": 2.2952, + "step": 4780 + }, + { + "epoch": 0.41, + "learning_rate": 4.962834618383436e-05, + "loss": 2.3224, + "step": 4790 + }, + { + "epoch": 0.42, + "learning_rate": 4.962747985792255e-05, + "loss": 2.2727, + "step": 4800 + }, + { + "epoch": 0.42, + "learning_rate": 4.962661353201075e-05, + "loss": 2.2692, + "step": 4810 + }, + { + "epoch": 0.42, + "learning_rate": 4.9625747206098935e-05, + "loss": 2.3142, + "step": 4820 + }, + { + "epoch": 0.42, + "learning_rate": 4.962488088018713e-05, + "loss": 2.3268, + "step": 4830 + }, + { + "epoch": 0.42, + "learning_rate": 4.9624014554275324e-05, + "loss": 2.3102, + "step": 4840 + }, + { + "epoch": 0.42, + "learning_rate": 4.962314822836351e-05, + "loss": 2.312, + "step": 4850 + }, + { + "epoch": 0.42, + "learning_rate": 4.9622281902451706e-05, + "loss": 2.2912, + "step": 4860 + }, + { + "epoch": 0.42, + "learning_rate": 4.96214155765399e-05, + "loss": 2.282, + "step": 4870 + }, + { + "epoch": 0.42, + "learning_rate": 4.962054925062809e-05, + "loss": 2.3295, + "step": 4880 + }, + { + "epoch": 0.42, + "learning_rate": 4.961968292471628e-05, + "loss": 2.2345, + "step": 4890 + }, + { + "epoch": 0.42, + "learning_rate": 4.961881659880447e-05, + "loss": 2.2939, + "step": 4900 + }, + { + "epoch": 0.42, + "learning_rate": 4.9617950272892665e-05, + "loss": 2.3342, + "step": 4910 + }, + { + "epoch": 0.43, + "learning_rate": 4.961708394698086e-05, + "loss": 2.2679, + "step": 4920 + }, + { + "epoch": 0.43, + "learning_rate": 4.961621762106905e-05, + "loss": 2.2784, + "step": 4930 + }, + { + "epoch": 0.43, + "learning_rate": 4.961535129515724e-05, + "loss": 2.2988, + "step": 4940 + }, + { + "epoch": 0.43, + "learning_rate": 4.9614484969245436e-05, + "loss": 2.298, + "step": 4950 + }, + { + "epoch": 0.43, + "learning_rate": 4.9613618643333624e-05, + "loss": 2.2963, + "step": 4960 + }, + { + "epoch": 0.43, + "learning_rate": 4.961275231742182e-05, + "loss": 2.3016, + "step": 4970 + }, + { + "epoch": 0.43, + "learning_rate": 4.961188599151001e-05, + "loss": 2.2691, + "step": 4980 + }, + { + "epoch": 0.43, + "learning_rate": 4.96110196655982e-05, + "loss": 2.2753, + "step": 4990 + }, + { + "epoch": 0.43, + "learning_rate": 4.9610153339686395e-05, + "loss": 2.2902, + "step": 5000 + }, + { + "epoch": 0.43, + "learning_rate": 4.960928701377458e-05, + "loss": 2.3323, + "step": 5010 + }, + { + "epoch": 0.43, + "learning_rate": 4.960842068786278e-05, + "loss": 2.2588, + "step": 5020 + }, + { + "epoch": 0.44, + "learning_rate": 4.960755436195097e-05, + "loss": 2.314, + "step": 5030 + }, + { + "epoch": 0.44, + "learning_rate": 4.960668803603916e-05, + "loss": 2.3212, + "step": 5040 + }, + { + "epoch": 0.44, + "learning_rate": 4.9605821710127354e-05, + "loss": 2.2837, + "step": 5050 + }, + { + "epoch": 0.44, + "learning_rate": 4.960495538421555e-05, + "loss": 2.2768, + "step": 5060 + }, + { + "epoch": 0.44, + "learning_rate": 4.9604089058303736e-05, + "loss": 2.2729, + "step": 5070 + }, + { + "epoch": 0.44, + "learning_rate": 4.960322273239193e-05, + "loss": 2.2875, + "step": 5080 + }, + { + "epoch": 0.44, + "learning_rate": 4.9602356406480125e-05, + "loss": 2.2956, + "step": 5090 + }, + { + "epoch": 0.44, + "learning_rate": 4.960149008056831e-05, + "loss": 2.307, + "step": 5100 + }, + { + "epoch": 0.44, + "learning_rate": 4.960062375465651e-05, + "loss": 2.274, + "step": 5110 + }, + { + "epoch": 0.44, + "learning_rate": 4.9599757428744694e-05, + "loss": 2.2966, + "step": 5120 + }, + { + "epoch": 0.44, + "learning_rate": 4.959889110283289e-05, + "loss": 2.2768, + "step": 5130 + }, + { + "epoch": 0.44, + "learning_rate": 4.959802477692108e-05, + "loss": 2.2533, + "step": 5140 + }, + { + "epoch": 0.45, + "learning_rate": 4.959715845100927e-05, + "loss": 2.333, + "step": 5150 + }, + { + "epoch": 0.45, + "learning_rate": 4.9596292125097465e-05, + "loss": 2.2751, + "step": 5160 + }, + { + "epoch": 0.45, + "learning_rate": 4.959542579918566e-05, + "loss": 2.3074, + "step": 5170 + }, + { + "epoch": 0.45, + "learning_rate": 4.959455947327385e-05, + "loss": 2.2848, + "step": 5180 + }, + { + "epoch": 0.45, + "learning_rate": 4.959369314736204e-05, + "loss": 2.3097, + "step": 5190 + }, + { + "epoch": 0.45, + "learning_rate": 4.9592826821450237e-05, + "loss": 2.3033, + "step": 5200 + }, + { + "epoch": 0.45, + "learning_rate": 4.9591960495538424e-05, + "loss": 2.2968, + "step": 5210 + }, + { + "epoch": 0.45, + "learning_rate": 4.959109416962662e-05, + "loss": 2.2965, + "step": 5220 + }, + { + "epoch": 0.45, + "learning_rate": 4.9590227843714806e-05, + "loss": 2.2634, + "step": 5230 + }, + { + "epoch": 0.45, + "learning_rate": 4.9589361517803e-05, + "loss": 2.3115, + "step": 5240 + }, + { + "epoch": 0.45, + "learning_rate": 4.9588495191891195e-05, + "loss": 2.3141, + "step": 5250 + }, + { + "epoch": 0.46, + "learning_rate": 4.958762886597938e-05, + "loss": 2.3068, + "step": 5260 + }, + { + "epoch": 0.46, + "learning_rate": 4.958676254006758e-05, + "loss": 2.3336, + "step": 5270 + }, + { + "epoch": 0.46, + "learning_rate": 4.958589621415577e-05, + "loss": 2.2673, + "step": 5280 + }, + { + "epoch": 0.46, + "learning_rate": 4.958502988824396e-05, + "loss": 2.309, + "step": 5290 + }, + { + "epoch": 0.46, + "learning_rate": 4.9584163562332154e-05, + "loss": 2.2836, + "step": 5300 + }, + { + "epoch": 0.46, + "learning_rate": 4.958329723642035e-05, + "loss": 2.2727, + "step": 5310 + }, + { + "epoch": 0.46, + "learning_rate": 4.9582430910508536e-05, + "loss": 2.229, + "step": 5320 + }, + { + "epoch": 0.46, + "learning_rate": 4.958156458459673e-05, + "loss": 2.2866, + "step": 5330 + }, + { + "epoch": 0.46, + "learning_rate": 4.958069825868492e-05, + "loss": 2.2512, + "step": 5340 + }, + { + "epoch": 0.46, + "learning_rate": 4.957983193277311e-05, + "loss": 2.3244, + "step": 5350 + }, + { + "epoch": 0.46, + "learning_rate": 4.957896560686131e-05, + "loss": 2.2794, + "step": 5360 + }, + { + "epoch": 0.46, + "learning_rate": 4.9578099280949495e-05, + "loss": 2.3462, + "step": 5370 + }, + { + "epoch": 0.47, + "learning_rate": 4.957723295503769e-05, + "loss": 2.2781, + "step": 5380 + }, + { + "epoch": 0.47, + "learning_rate": 4.9576366629125884e-05, + "loss": 2.2826, + "step": 5390 + }, + { + "epoch": 0.47, + "learning_rate": 4.957550030321407e-05, + "loss": 2.3146, + "step": 5400 + }, + { + "epoch": 0.47, + "learning_rate": 4.9574633977302266e-05, + "loss": 2.2286, + "step": 5410 + }, + { + "epoch": 0.47, + "learning_rate": 4.957376765139046e-05, + "loss": 2.2516, + "step": 5420 + }, + { + "epoch": 0.47, + "learning_rate": 4.957290132547865e-05, + "loss": 2.2539, + "step": 5430 + }, + { + "epoch": 0.47, + "learning_rate": 4.957203499956684e-05, + "loss": 2.2887, + "step": 5440 + }, + { + "epoch": 0.47, + "learning_rate": 4.957116867365503e-05, + "loss": 2.2688, + "step": 5450 + }, + { + "epoch": 0.47, + "learning_rate": 4.9570302347743225e-05, + "loss": 2.2903, + "step": 5460 + }, + { + "epoch": 0.47, + "learning_rate": 4.956943602183142e-05, + "loss": 2.2997, + "step": 5470 + }, + { + "epoch": 0.47, + "learning_rate": 4.956856969591961e-05, + "loss": 2.3128, + "step": 5480 + }, + { + "epoch": 0.48, + "learning_rate": 4.95677033700078e-05, + "loss": 2.2187, + "step": 5490 + }, + { + "epoch": 0.48, + "learning_rate": 4.9566837044095996e-05, + "loss": 2.2705, + "step": 5500 + }, + { + "epoch": 0.48, + "learning_rate": 4.956597071818418e-05, + "loss": 2.2581, + "step": 5510 + }, + { + "epoch": 0.48, + "learning_rate": 4.956510439227238e-05, + "loss": 2.3095, + "step": 5520 + }, + { + "epoch": 0.48, + "learning_rate": 4.956423806636057e-05, + "loss": 2.32, + "step": 5530 + }, + { + "epoch": 0.48, + "learning_rate": 4.956337174044876e-05, + "loss": 2.3058, + "step": 5540 + }, + { + "epoch": 0.48, + "learning_rate": 4.9562505414536954e-05, + "loss": 2.2414, + "step": 5550 + }, + { + "epoch": 0.48, + "learning_rate": 4.956163908862514e-05, + "loss": 2.2563, + "step": 5560 + }, + { + "epoch": 0.48, + "learning_rate": 4.9560772762713336e-05, + "loss": 2.2594, + "step": 5570 + }, + { + "epoch": 0.48, + "learning_rate": 4.955990643680153e-05, + "loss": 2.3087, + "step": 5580 + }, + { + "epoch": 0.48, + "learning_rate": 4.955904011088972e-05, + "loss": 2.2789, + "step": 5590 + }, + { + "epoch": 0.48, + "learning_rate": 4.955817378497791e-05, + "loss": 2.2912, + "step": 5600 + }, + { + "epoch": 0.49, + "learning_rate": 4.955730745906611e-05, + "loss": 2.2623, + "step": 5610 + }, + { + "epoch": 0.49, + "learning_rate": 4.9556441133154295e-05, + "loss": 2.3012, + "step": 5620 + }, + { + "epoch": 0.49, + "learning_rate": 4.955557480724249e-05, + "loss": 2.2358, + "step": 5630 + }, + { + "epoch": 0.49, + "learning_rate": 4.955470848133068e-05, + "loss": 2.2447, + "step": 5640 + }, + { + "epoch": 0.49, + "learning_rate": 4.955384215541887e-05, + "loss": 2.22, + "step": 5650 + }, + { + "epoch": 0.49, + "learning_rate": 4.9552975829507066e-05, + "loss": 2.1912, + "step": 5660 + }, + { + "epoch": 0.49, + "learning_rate": 4.9552109503595254e-05, + "loss": 2.2737, + "step": 5670 + }, + { + "epoch": 0.49, + "learning_rate": 4.955124317768345e-05, + "loss": 2.2521, + "step": 5680 + }, + { + "epoch": 0.49, + "learning_rate": 4.955037685177164e-05, + "loss": 2.3273, + "step": 5690 + }, + { + "epoch": 0.49, + "learning_rate": 4.954951052585983e-05, + "loss": 2.3011, + "step": 5700 + }, + { + "epoch": 0.49, + "learning_rate": 4.9548644199948025e-05, + "loss": 2.1979, + "step": 5710 + }, + { + "epoch": 0.5, + "learning_rate": 4.954777787403622e-05, + "loss": 2.272, + "step": 5720 + }, + { + "epoch": 0.5, + "learning_rate": 4.954691154812441e-05, + "loss": 2.2727, + "step": 5730 + }, + { + "epoch": 0.5, + "learning_rate": 4.95460452222126e-05, + "loss": 2.2788, + "step": 5740 + }, + { + "epoch": 0.5, + "learning_rate": 4.954517889630079e-05, + "loss": 2.2381, + "step": 5750 + }, + { + "epoch": 0.5, + "learning_rate": 4.9544312570388984e-05, + "loss": 2.2667, + "step": 5760 + }, + { + "epoch": 0.5, + "learning_rate": 4.954344624447718e-05, + "loss": 2.2922, + "step": 5770 + }, + { + "epoch": 0.5, + "learning_rate": 4.9542579918565366e-05, + "loss": 2.284, + "step": 5780 + }, + { + "epoch": 0.5, + "learning_rate": 4.954171359265356e-05, + "loss": 2.2752, + "step": 5790 + }, + { + "epoch": 0.5, + "learning_rate": 4.9540847266741755e-05, + "loss": 2.2328, + "step": 5800 + }, + { + "epoch": 0.5, + "learning_rate": 4.953998094082994e-05, + "loss": 2.2386, + "step": 5810 + }, + { + "epoch": 0.5, + "learning_rate": 4.953911461491814e-05, + "loss": 2.2549, + "step": 5820 + }, + { + "epoch": 0.5, + "learning_rate": 4.953824828900633e-05, + "loss": 2.2564, + "step": 5830 + }, + { + "epoch": 0.51, + "learning_rate": 4.953738196309452e-05, + "loss": 2.2207, + "step": 5840 + }, + { + "epoch": 0.51, + "learning_rate": 4.953651563718271e-05, + "loss": 2.2892, + "step": 5850 + }, + { + "epoch": 0.51, + "learning_rate": 4.95356493112709e-05, + "loss": 2.3315, + "step": 5860 + }, + { + "epoch": 0.51, + "learning_rate": 4.9534782985359095e-05, + "loss": 2.2355, + "step": 5870 + }, + { + "epoch": 0.51, + "learning_rate": 4.953391665944729e-05, + "loss": 2.2556, + "step": 5880 + }, + { + "epoch": 0.51, + "learning_rate": 4.953305033353548e-05, + "loss": 2.2799, + "step": 5890 + }, + { + "epoch": 0.51, + "learning_rate": 4.953218400762367e-05, + "loss": 2.267, + "step": 5900 + }, + { + "epoch": 0.51, + "learning_rate": 4.9531317681711866e-05, + "loss": 2.2678, + "step": 5910 + }, + { + "epoch": 0.51, + "learning_rate": 4.9530451355800054e-05, + "loss": 2.2109, + "step": 5920 + }, + { + "epoch": 0.51, + "learning_rate": 4.952958502988825e-05, + "loss": 2.2757, + "step": 5930 + }, + { + "epoch": 0.51, + "learning_rate": 4.952871870397644e-05, + "loss": 2.2887, + "step": 5940 + }, + { + "epoch": 0.51, + "learning_rate": 4.952785237806463e-05, + "loss": 2.2191, + "step": 5950 + }, + { + "epoch": 0.52, + "learning_rate": 4.9526986052152825e-05, + "loss": 2.2937, + "step": 5960 + }, + { + "epoch": 0.52, + "learning_rate": 4.952611972624101e-05, + "loss": 2.1973, + "step": 5970 + }, + { + "epoch": 0.52, + "learning_rate": 4.952525340032921e-05, + "loss": 2.2639, + "step": 5980 + }, + { + "epoch": 0.52, + "learning_rate": 4.95243870744174e-05, + "loss": 2.2486, + "step": 5990 + }, + { + "epoch": 0.52, + "learning_rate": 4.952352074850559e-05, + "loss": 2.2596, + "step": 6000 + }, + { + "epoch": 0.52, + "learning_rate": 4.9522654422593784e-05, + "loss": 2.2913, + "step": 6010 + }, + { + "epoch": 0.52, + "learning_rate": 4.952178809668198e-05, + "loss": 2.2705, + "step": 6020 + }, + { + "epoch": 0.52, + "learning_rate": 4.9520921770770166e-05, + "loss": 2.2466, + "step": 6030 + }, + { + "epoch": 0.52, + "learning_rate": 4.952005544485836e-05, + "loss": 2.2395, + "step": 6040 + }, + { + "epoch": 0.52, + "learning_rate": 4.9519189118946555e-05, + "loss": 2.2281, + "step": 6050 + }, + { + "epoch": 0.52, + "learning_rate": 4.951832279303474e-05, + "loss": 2.2555, + "step": 6060 + }, + { + "epoch": 0.53, + "learning_rate": 4.951745646712294e-05, + "loss": 2.2776, + "step": 6070 + }, + { + "epoch": 0.53, + "learning_rate": 4.9516590141211125e-05, + "loss": 2.2947, + "step": 6080 + }, + { + "epoch": 0.53, + "learning_rate": 4.951572381529932e-05, + "loss": 2.284, + "step": 6090 + }, + { + "epoch": 0.53, + "learning_rate": 4.9514857489387514e-05, + "loss": 2.2139, + "step": 6100 + }, + { + "epoch": 0.53, + "learning_rate": 4.95139911634757e-05, + "loss": 2.2271, + "step": 6110 + }, + { + "epoch": 0.53, + "learning_rate": 4.9513124837563896e-05, + "loss": 2.2503, + "step": 6120 + }, + { + "epoch": 0.53, + "learning_rate": 4.951225851165209e-05, + "loss": 2.2053, + "step": 6130 + }, + { + "epoch": 0.53, + "learning_rate": 4.951139218574028e-05, + "loss": 2.2722, + "step": 6140 + }, + { + "epoch": 0.53, + "learning_rate": 4.951052585982847e-05, + "loss": 2.1922, + "step": 6150 + }, + { + "epoch": 0.53, + "learning_rate": 4.950965953391667e-05, + "loss": 2.2794, + "step": 6160 + }, + { + "epoch": 0.53, + "learning_rate": 4.9508793208004854e-05, + "loss": 2.2459, + "step": 6170 + }, + { + "epoch": 0.53, + "learning_rate": 4.950792688209305e-05, + "loss": 2.2593, + "step": 6180 + }, + { + "epoch": 0.54, + "learning_rate": 4.950706055618124e-05, + "loss": 2.2703, + "step": 6190 + }, + { + "epoch": 0.54, + "learning_rate": 4.950619423026943e-05, + "loss": 2.2504, + "step": 6200 + }, + { + "epoch": 0.54, + "learning_rate": 4.9505327904357626e-05, + "loss": 2.2236, + "step": 6210 + }, + { + "epoch": 0.54, + "learning_rate": 4.950446157844581e-05, + "loss": 2.2405, + "step": 6220 + }, + { + "epoch": 0.54, + "learning_rate": 4.950359525253401e-05, + "loss": 2.2334, + "step": 6230 + }, + { + "epoch": 0.54, + "learning_rate": 4.95027289266222e-05, + "loss": 2.2158, + "step": 6240 + }, + { + "epoch": 0.54, + "learning_rate": 4.950186260071039e-05, + "loss": 2.2603, + "step": 6250 + }, + { + "epoch": 0.54, + "learning_rate": 4.9500996274798584e-05, + "loss": 2.2587, + "step": 6260 + }, + { + "epoch": 0.54, + "learning_rate": 4.950012994888678e-05, + "loss": 2.2401, + "step": 6270 + }, + { + "epoch": 0.54, + "learning_rate": 4.9499263622974966e-05, + "loss": 2.2649, + "step": 6280 + }, + { + "epoch": 0.54, + "learning_rate": 4.949839729706316e-05, + "loss": 2.2337, + "step": 6290 + }, + { + "epoch": 0.55, + "learning_rate": 4.949753097115135e-05, + "loss": 2.2101, + "step": 6300 + }, + { + "epoch": 0.55, + "learning_rate": 4.949666464523954e-05, + "loss": 2.2834, + "step": 6310 + }, + { + "epoch": 0.55, + "learning_rate": 4.949579831932774e-05, + "loss": 2.2317, + "step": 6320 + }, + { + "epoch": 0.55, + "learning_rate": 4.9494931993415925e-05, + "loss": 2.3015, + "step": 6330 + }, + { + "epoch": 0.55, + "learning_rate": 4.949406566750412e-05, + "loss": 2.3359, + "step": 6340 + }, + { + "epoch": 0.55, + "learning_rate": 4.9493199341592314e-05, + "loss": 2.2304, + "step": 6350 + }, + { + "epoch": 0.55, + "learning_rate": 4.94923330156805e-05, + "loss": 2.2465, + "step": 6360 + }, + { + "epoch": 0.55, + "learning_rate": 4.9491466689768696e-05, + "loss": 2.1416, + "step": 6370 + }, + { + "epoch": 0.55, + "learning_rate": 4.9490600363856884e-05, + "loss": 2.2222, + "step": 6380 + }, + { + "epoch": 0.55, + "learning_rate": 4.948973403794508e-05, + "loss": 2.1918, + "step": 6390 + }, + { + "epoch": 0.55, + "learning_rate": 4.948886771203327e-05, + "loss": 2.2449, + "step": 6400 + }, + { + "epoch": 0.55, + "learning_rate": 4.948800138612146e-05, + "loss": 2.2279, + "step": 6410 + }, + { + "epoch": 0.56, + "learning_rate": 4.9487135060209655e-05, + "loss": 2.1566, + "step": 6420 + }, + { + "epoch": 0.56, + "learning_rate": 4.948626873429785e-05, + "loss": 2.2265, + "step": 6430 + }, + { + "epoch": 0.56, + "learning_rate": 4.948540240838604e-05, + "loss": 2.2518, + "step": 6440 + }, + { + "epoch": 0.56, + "learning_rate": 4.948453608247423e-05, + "loss": 2.2513, + "step": 6450 + }, + { + "epoch": 0.56, + "learning_rate": 4.9483669756562426e-05, + "loss": 2.2633, + "step": 6460 + }, + { + "epoch": 0.56, + "learning_rate": 4.9482803430650614e-05, + "loss": 2.2608, + "step": 6470 + }, + { + "epoch": 0.56, + "learning_rate": 4.948193710473881e-05, + "loss": 2.215, + "step": 6480 + }, + { + "epoch": 0.56, + "learning_rate": 4.9481070778826996e-05, + "loss": 2.28, + "step": 6490 + }, + { + "epoch": 0.56, + "learning_rate": 4.948020445291519e-05, + "loss": 2.2495, + "step": 6500 + }, + { + "epoch": 0.56, + "learning_rate": 4.9479338127003385e-05, + "loss": 2.2718, + "step": 6510 + }, + { + "epoch": 0.56, + "learning_rate": 4.947847180109157e-05, + "loss": 2.195, + "step": 6520 + }, + { + "epoch": 0.57, + "learning_rate": 4.947760547517977e-05, + "loss": 2.2256, + "step": 6530 + }, + { + "epoch": 0.57, + "learning_rate": 4.947673914926796e-05, + "loss": 2.2529, + "step": 6540 + }, + { + "epoch": 0.57, + "learning_rate": 4.947587282335615e-05, + "loss": 2.1819, + "step": 6550 + }, + { + "epoch": 0.57, + "learning_rate": 4.947500649744434e-05, + "loss": 2.2303, + "step": 6560 + }, + { + "epoch": 0.57, + "learning_rate": 4.947414017153254e-05, + "loss": 2.2382, + "step": 6570 + }, + { + "epoch": 0.57, + "learning_rate": 4.9473273845620725e-05, + "loss": 2.2445, + "step": 6580 + }, + { + "epoch": 0.57, + "learning_rate": 4.947240751970892e-05, + "loss": 2.2163, + "step": 6590 + }, + { + "epoch": 0.57, + "learning_rate": 4.947154119379711e-05, + "loss": 2.2119, + "step": 6600 + }, + { + "epoch": 0.57, + "learning_rate": 4.94706748678853e-05, + "loss": 2.2173, + "step": 6610 + }, + { + "epoch": 0.57, + "learning_rate": 4.9469808541973496e-05, + "loss": 2.1994, + "step": 6620 + }, + { + "epoch": 0.57, + "learning_rate": 4.9468942216061684e-05, + "loss": 2.1978, + "step": 6630 + }, + { + "epoch": 0.57, + "learning_rate": 4.946807589014988e-05, + "loss": 2.2227, + "step": 6640 + }, + { + "epoch": 0.58, + "learning_rate": 4.946720956423807e-05, + "loss": 2.2307, + "step": 6650 + }, + { + "epoch": 0.58, + "learning_rate": 4.946634323832626e-05, + "loss": 2.294, + "step": 6660 + }, + { + "epoch": 0.58, + "learning_rate": 4.9465476912414455e-05, + "loss": 2.2103, + "step": 6670 + }, + { + "epoch": 0.58, + "learning_rate": 4.946461058650265e-05, + "loss": 2.2487, + "step": 6680 + }, + { + "epoch": 0.58, + "learning_rate": 4.946374426059084e-05, + "loss": 2.2102, + "step": 6690 + }, + { + "epoch": 0.58, + "learning_rate": 4.946287793467903e-05, + "loss": 2.2618, + "step": 6700 + }, + { + "epoch": 0.58, + "learning_rate": 4.946201160876722e-05, + "loss": 2.255, + "step": 6710 + }, + { + "epoch": 0.58, + "learning_rate": 4.9461145282855414e-05, + "loss": 2.2228, + "step": 6720 + }, + { + "epoch": 0.58, + "learning_rate": 4.946027895694361e-05, + "loss": 2.2583, + "step": 6730 + }, + { + "epoch": 0.58, + "learning_rate": 4.9459412631031796e-05, + "loss": 2.1909, + "step": 6740 + }, + { + "epoch": 0.58, + "learning_rate": 4.945854630511999e-05, + "loss": 2.2014, + "step": 6750 + }, + { + "epoch": 0.59, + "learning_rate": 4.9457679979208185e-05, + "loss": 2.2416, + "step": 6760 + }, + { + "epoch": 0.59, + "learning_rate": 4.945681365329637e-05, + "loss": 2.1858, + "step": 6770 + }, + { + "epoch": 0.59, + "learning_rate": 4.945594732738457e-05, + "loss": 2.2625, + "step": 6780 + }, + { + "epoch": 0.59, + "learning_rate": 4.945508100147276e-05, + "loss": 2.2381, + "step": 6790 + }, + { + "epoch": 0.59, + "learning_rate": 4.945421467556095e-05, + "loss": 2.238, + "step": 6800 + }, + { + "epoch": 0.59, + "learning_rate": 4.9453348349649144e-05, + "loss": 2.2282, + "step": 6810 + }, + { + "epoch": 0.59, + "learning_rate": 4.945248202373733e-05, + "loss": 2.1805, + "step": 6820 + }, + { + "epoch": 0.59, + "learning_rate": 4.9451615697825526e-05, + "loss": 2.2536, + "step": 6830 + }, + { + "epoch": 0.59, + "learning_rate": 4.945074937191372e-05, + "loss": 2.2653, + "step": 6840 + }, + { + "epoch": 0.59, + "learning_rate": 4.944988304600191e-05, + "loss": 2.1914, + "step": 6850 + }, + { + "epoch": 0.59, + "learning_rate": 4.94490167200901e-05, + "loss": 2.2483, + "step": 6860 + }, + { + "epoch": 0.59, + "learning_rate": 4.94481503941783e-05, + "loss": 2.2422, + "step": 6870 + }, + { + "epoch": 0.6, + "learning_rate": 4.9447284068266484e-05, + "loss": 2.2205, + "step": 6880 + }, + { + "epoch": 0.6, + "learning_rate": 4.944641774235468e-05, + "loss": 2.2306, + "step": 6890 + }, + { + "epoch": 0.6, + "learning_rate": 4.944555141644287e-05, + "loss": 2.1867, + "step": 6900 + }, + { + "epoch": 0.6, + "learning_rate": 4.944468509053106e-05, + "loss": 2.2123, + "step": 6910 + }, + { + "epoch": 0.6, + "learning_rate": 4.9443818764619255e-05, + "loss": 2.1928, + "step": 6920 + }, + { + "epoch": 0.6, + "learning_rate": 4.944295243870744e-05, + "loss": 2.1865, + "step": 6930 + }, + { + "epoch": 0.6, + "learning_rate": 4.944208611279564e-05, + "loss": 2.2532, + "step": 6940 + }, + { + "epoch": 0.6, + "learning_rate": 4.944121978688383e-05, + "loss": 2.2181, + "step": 6950 + }, + { + "epoch": 0.6, + "learning_rate": 4.944035346097202e-05, + "loss": 2.1703, + "step": 6960 + }, + { + "epoch": 0.6, + "learning_rate": 4.9439487135060214e-05, + "loss": 2.2706, + "step": 6970 + }, + { + "epoch": 0.6, + "learning_rate": 4.943862080914841e-05, + "loss": 2.1748, + "step": 6980 + }, + { + "epoch": 0.61, + "learning_rate": 4.9437754483236596e-05, + "loss": 2.238, + "step": 6990 + }, + { + "epoch": 0.61, + "learning_rate": 4.943688815732479e-05, + "loss": 2.2281, + "step": 7000 + }, + { + "epoch": 0.61, + "learning_rate": 4.9436021831412985e-05, + "loss": 2.2177, + "step": 7010 + }, + { + "epoch": 0.61, + "learning_rate": 4.943515550550117e-05, + "loss": 2.2432, + "step": 7020 + }, + { + "epoch": 0.61, + "learning_rate": 4.943428917958937e-05, + "loss": 2.1397, + "step": 7030 + }, + { + "epoch": 0.61, + "learning_rate": 4.9433422853677555e-05, + "loss": 2.1845, + "step": 7040 + }, + { + "epoch": 0.61, + "learning_rate": 4.943255652776575e-05, + "loss": 2.1939, + "step": 7050 + }, + { + "epoch": 0.61, + "learning_rate": 4.9431690201853944e-05, + "loss": 2.23, + "step": 7060 + }, + { + "epoch": 0.61, + "learning_rate": 4.943082387594213e-05, + "loss": 2.2237, + "step": 7070 + }, + { + "epoch": 0.61, + "learning_rate": 4.9429957550030326e-05, + "loss": 2.1867, + "step": 7080 + }, + { + "epoch": 0.61, + "learning_rate": 4.942909122411852e-05, + "loss": 2.1896, + "step": 7090 + }, + { + "epoch": 0.61, + "learning_rate": 4.942822489820671e-05, + "loss": 2.1928, + "step": 7100 + }, + { + "epoch": 0.62, + "learning_rate": 4.94273585722949e-05, + "loss": 2.2066, + "step": 7110 + }, + { + "epoch": 0.62, + "learning_rate": 4.942649224638309e-05, + "loss": 2.2383, + "step": 7120 + }, + { + "epoch": 0.62, + "learning_rate": 4.9425625920471285e-05, + "loss": 2.245, + "step": 7130 + }, + { + "epoch": 0.62, + "learning_rate": 4.942475959455948e-05, + "loss": 2.206, + "step": 7140 + }, + { + "epoch": 0.62, + "learning_rate": 4.942389326864767e-05, + "loss": 2.1855, + "step": 7150 + }, + { + "epoch": 0.62, + "learning_rate": 4.942302694273586e-05, + "loss": 2.2099, + "step": 7160 + }, + { + "epoch": 0.62, + "learning_rate": 4.9422160616824056e-05, + "loss": 2.269, + "step": 7170 + }, + { + "epoch": 0.62, + "learning_rate": 4.9421294290912244e-05, + "loss": 2.1779, + "step": 7180 + }, + { + "epoch": 0.62, + "learning_rate": 4.942042796500044e-05, + "loss": 2.2201, + "step": 7190 + }, + { + "epoch": 0.62, + "learning_rate": 4.941956163908863e-05, + "loss": 2.1621, + "step": 7200 + }, + { + "epoch": 0.62, + "learning_rate": 4.941869531317682e-05, + "loss": 2.2727, + "step": 7210 + }, + { + "epoch": 0.62, + "learning_rate": 4.9417828987265015e-05, + "loss": 2.238, + "step": 7220 + }, + { + "epoch": 0.63, + "learning_rate": 4.94169626613532e-05, + "loss": 2.2096, + "step": 7230 + }, + { + "epoch": 0.63, + "learning_rate": 4.94160963354414e-05, + "loss": 2.2475, + "step": 7240 + }, + { + "epoch": 0.63, + "learning_rate": 4.941523000952959e-05, + "loss": 2.2254, + "step": 7250 + }, + { + "epoch": 0.63, + "learning_rate": 4.941436368361778e-05, + "loss": 2.2408, + "step": 7260 + }, + { + "epoch": 0.63, + "learning_rate": 4.941349735770597e-05, + "loss": 2.1603, + "step": 7270 + }, + { + "epoch": 0.63, + "learning_rate": 4.941263103179417e-05, + "loss": 2.2558, + "step": 7280 + }, + { + "epoch": 0.63, + "learning_rate": 4.9411764705882355e-05, + "loss": 2.159, + "step": 7290 + }, + { + "epoch": 0.63, + "learning_rate": 4.941089837997055e-05, + "loss": 2.21, + "step": 7300 + }, + { + "epoch": 0.63, + "learning_rate": 4.9410032054058744e-05, + "loss": 2.2032, + "step": 7310 + }, + { + "epoch": 0.63, + "learning_rate": 4.940916572814693e-05, + "loss": 2.153, + "step": 7320 + }, + { + "epoch": 0.63, + "learning_rate": 4.9408299402235126e-05, + "loss": 2.2196, + "step": 7330 + }, + { + "epoch": 0.64, + "learning_rate": 4.9407433076323314e-05, + "loss": 2.1558, + "step": 7340 + }, + { + "epoch": 0.64, + "learning_rate": 4.940656675041151e-05, + "loss": 2.2292, + "step": 7350 + }, + { + "epoch": 0.64, + "learning_rate": 4.94057004244997e-05, + "loss": 2.1793, + "step": 7360 + }, + { + "epoch": 0.64, + "learning_rate": 4.940483409858789e-05, + "loss": 2.2298, + "step": 7370 + }, + { + "epoch": 0.64, + "learning_rate": 4.9403967772676085e-05, + "loss": 2.1614, + "step": 7380 + }, + { + "epoch": 0.64, + "learning_rate": 4.940310144676428e-05, + "loss": 2.1779, + "step": 7390 + }, + { + "epoch": 0.64, + "learning_rate": 4.940223512085247e-05, + "loss": 2.2109, + "step": 7400 + }, + { + "epoch": 0.64, + "learning_rate": 4.940136879494066e-05, + "loss": 2.1952, + "step": 7410 + }, + { + "epoch": 0.64, + "learning_rate": 4.9400502469028856e-05, + "loss": 2.1814, + "step": 7420 + }, + { + "epoch": 0.64, + "learning_rate": 4.9399636143117044e-05, + "loss": 2.1628, + "step": 7430 + }, + { + "epoch": 0.64, + "learning_rate": 4.939876981720524e-05, + "loss": 2.255, + "step": 7440 + }, + { + "epoch": 0.64, + "learning_rate": 4.9397903491293426e-05, + "loss": 2.1915, + "step": 7450 + }, + { + "epoch": 0.65, + "learning_rate": 4.939703716538162e-05, + "loss": 2.172, + "step": 7460 + }, + { + "epoch": 0.65, + "learning_rate": 4.9396170839469815e-05, + "loss": 2.19, + "step": 7470 + }, + { + "epoch": 0.65, + "learning_rate": 4.9395304513558e-05, + "loss": 2.1539, + "step": 7480 + }, + { + "epoch": 0.65, + "learning_rate": 4.93944381876462e-05, + "loss": 2.1542, + "step": 7490 + }, + { + "epoch": 0.65, + "learning_rate": 4.939357186173439e-05, + "loss": 2.1376, + "step": 7500 + }, + { + "epoch": 0.65, + "learning_rate": 4.939270553582258e-05, + "loss": 2.2499, + "step": 7510 + }, + { + "epoch": 0.65, + "learning_rate": 4.9391839209910774e-05, + "loss": 2.171, + "step": 7520 + }, + { + "epoch": 0.65, + "learning_rate": 4.939097288399897e-05, + "loss": 2.1904, + "step": 7530 + }, + { + "epoch": 0.65, + "learning_rate": 4.9390106558087156e-05, + "loss": 2.2239, + "step": 7540 + }, + { + "epoch": 0.65, + "learning_rate": 4.938924023217535e-05, + "loss": 2.1338, + "step": 7550 + }, + { + "epoch": 0.65, + "learning_rate": 4.938837390626354e-05, + "loss": 2.1908, + "step": 7560 + }, + { + "epoch": 0.66, + "learning_rate": 4.938750758035173e-05, + "loss": 2.1784, + "step": 7570 + }, + { + "epoch": 0.66, + "learning_rate": 4.938664125443993e-05, + "loss": 2.211, + "step": 7580 + }, + { + "epoch": 0.66, + "learning_rate": 4.9385774928528114e-05, + "loss": 2.1746, + "step": 7590 + }, + { + "epoch": 0.66, + "learning_rate": 4.938490860261631e-05, + "loss": 2.206, + "step": 7600 + }, + { + "epoch": 0.66, + "learning_rate": 4.93840422767045e-05, + "loss": 2.2478, + "step": 7610 + }, + { + "epoch": 0.66, + "learning_rate": 4.938317595079269e-05, + "loss": 2.1423, + "step": 7620 + }, + { + "epoch": 0.66, + "learning_rate": 4.9382309624880885e-05, + "loss": 2.1936, + "step": 7630 + }, + { + "epoch": 0.66, + "learning_rate": 4.938144329896908e-05, + "loss": 2.1584, + "step": 7640 + }, + { + "epoch": 0.66, + "learning_rate": 4.938057697305727e-05, + "loss": 2.2436, + "step": 7650 + }, + { + "epoch": 0.66, + "learning_rate": 4.937971064714546e-05, + "loss": 2.2275, + "step": 7660 + }, + { + "epoch": 0.66, + "learning_rate": 4.937884432123365e-05, + "loss": 2.1809, + "step": 7670 + }, + { + "epoch": 0.66, + "learning_rate": 4.9377977995321844e-05, + "loss": 2.182, + "step": 7680 + }, + { + "epoch": 0.67, + "learning_rate": 4.937711166941004e-05, + "loss": 2.2244, + "step": 7690 + }, + { + "epoch": 0.67, + "learning_rate": 4.9376245343498226e-05, + "loss": 2.148, + "step": 7700 + }, + { + "epoch": 0.67, + "learning_rate": 4.937537901758642e-05, + "loss": 2.2137, + "step": 7710 + }, + { + "epoch": 0.67, + "learning_rate": 4.9374512691674615e-05, + "loss": 2.1629, + "step": 7720 + }, + { + "epoch": 0.67, + "learning_rate": 4.93736463657628e-05, + "loss": 2.2835, + "step": 7730 + }, + { + "epoch": 0.67, + "learning_rate": 4.9372780039851e-05, + "loss": 2.1119, + "step": 7740 + }, + { + "epoch": 0.67, + "learning_rate": 4.937191371393919e-05, + "loss": 2.1394, + "step": 7750 + }, + { + "epoch": 0.67, + "learning_rate": 4.937104738802738e-05, + "loss": 2.1792, + "step": 7760 + }, + { + "epoch": 0.67, + "learning_rate": 4.9370181062115574e-05, + "loss": 2.2111, + "step": 7770 + }, + { + "epoch": 0.67, + "learning_rate": 4.936931473620376e-05, + "loss": 2.1638, + "step": 7780 + }, + { + "epoch": 0.67, + "learning_rate": 4.9368448410291956e-05, + "loss": 2.1214, + "step": 7790 + }, + { + "epoch": 0.68, + "learning_rate": 4.936758208438015e-05, + "loss": 2.2214, + "step": 7800 + }, + { + "epoch": 0.68, + "learning_rate": 4.936671575846834e-05, + "loss": 2.1706, + "step": 7810 + }, + { + "epoch": 0.68, + "learning_rate": 4.936584943255653e-05, + "loss": 2.1294, + "step": 7820 + }, + { + "epoch": 0.68, + "learning_rate": 4.936498310664473e-05, + "loss": 2.1427, + "step": 7830 + }, + { + "epoch": 0.68, + "learning_rate": 4.9364116780732915e-05, + "loss": 2.2309, + "step": 7840 + }, + { + "epoch": 0.68, + "learning_rate": 4.936325045482111e-05, + "loss": 2.1653, + "step": 7850 + }, + { + "epoch": 0.68, + "learning_rate": 4.9362384128909304e-05, + "loss": 2.2499, + "step": 7860 + }, + { + "epoch": 0.68, + "learning_rate": 4.936151780299749e-05, + "loss": 2.142, + "step": 7870 + }, + { + "epoch": 0.68, + "learning_rate": 4.9360651477085686e-05, + "loss": 2.1746, + "step": 7880 + }, + { + "epoch": 0.68, + "learning_rate": 4.9359785151173873e-05, + "loss": 2.1595, + "step": 7890 + }, + { + "epoch": 0.68, + "learning_rate": 4.935891882526207e-05, + "loss": 2.1672, + "step": 7900 + }, + { + "epoch": 0.68, + "learning_rate": 4.935805249935026e-05, + "loss": 2.138, + "step": 7910 + }, + { + "epoch": 0.69, + "learning_rate": 4.935718617343845e-05, + "loss": 2.1742, + "step": 7920 + }, + { + "epoch": 0.69, + "learning_rate": 4.9356319847526645e-05, + "loss": 2.1909, + "step": 7930 + }, + { + "epoch": 0.69, + "learning_rate": 4.935545352161484e-05, + "loss": 2.1134, + "step": 7940 + }, + { + "epoch": 0.69, + "learning_rate": 4.935458719570303e-05, + "loss": 2.1687, + "step": 7950 + }, + { + "epoch": 0.69, + "learning_rate": 4.935372086979122e-05, + "loss": 2.1948, + "step": 7960 + }, + { + "epoch": 0.69, + "learning_rate": 4.935285454387941e-05, + "loss": 2.1578, + "step": 7970 + }, + { + "epoch": 0.69, + "learning_rate": 4.93519882179676e-05, + "loss": 2.1228, + "step": 7980 + }, + { + "epoch": 0.69, + "learning_rate": 4.93511218920558e-05, + "loss": 2.1574, + "step": 7990 + }, + { + "epoch": 0.69, + "learning_rate": 4.9350255566143985e-05, + "loss": 2.1485, + "step": 8000 + }, + { + "epoch": 0.69, + "learning_rate": 4.934938924023218e-05, + "loss": 2.2306, + "step": 8010 + }, + { + "epoch": 0.69, + "learning_rate": 4.9348522914320374e-05, + "loss": 2.1998, + "step": 8020 + }, + { + "epoch": 0.7, + "learning_rate": 4.934765658840856e-05, + "loss": 2.2279, + "step": 8030 + }, + { + "epoch": 0.7, + "learning_rate": 4.9346790262496756e-05, + "loss": 2.1342, + "step": 8040 + }, + { + "epoch": 0.7, + "learning_rate": 4.934592393658495e-05, + "loss": 2.1859, + "step": 8050 + }, + { + "epoch": 0.7, + "learning_rate": 4.934505761067314e-05, + "loss": 2.1435, + "step": 8060 + }, + { + "epoch": 0.7, + "learning_rate": 4.934419128476133e-05, + "loss": 2.1464, + "step": 8070 + }, + { + "epoch": 0.7, + "learning_rate": 4.934332495884952e-05, + "loss": 2.1988, + "step": 8080 + }, + { + "epoch": 0.7, + "learning_rate": 4.9342458632937715e-05, + "loss": 2.1819, + "step": 8090 + }, + { + "epoch": 0.7, + "learning_rate": 4.934159230702591e-05, + "loss": 2.104, + "step": 8100 + }, + { + "epoch": 0.7, + "learning_rate": 4.93407259811141e-05, + "loss": 2.1573, + "step": 8110 + }, + { + "epoch": 0.7, + "learning_rate": 4.933985965520229e-05, + "loss": 2.1404, + "step": 8120 + }, + { + "epoch": 0.7, + "learning_rate": 4.9338993329290486e-05, + "loss": 2.1336, + "step": 8130 + }, + { + "epoch": 0.7, + "learning_rate": 4.9338127003378674e-05, + "loss": 2.171, + "step": 8140 + }, + { + "epoch": 0.71, + "learning_rate": 4.933726067746687e-05, + "loss": 2.1479, + "step": 8150 + }, + { + "epoch": 0.71, + "learning_rate": 4.933639435155506e-05, + "loss": 2.1635, + "step": 8160 + }, + { + "epoch": 0.71, + "learning_rate": 4.933552802564325e-05, + "loss": 2.1951, + "step": 8170 + }, + { + "epoch": 0.71, + "learning_rate": 4.9334661699731445e-05, + "loss": 2.2091, + "step": 8180 + }, + { + "epoch": 0.71, + "learning_rate": 4.933379537381963e-05, + "loss": 2.173, + "step": 8190 + }, + { + "epoch": 0.71, + "learning_rate": 4.933292904790783e-05, + "loss": 2.1151, + "step": 8200 + }, + { + "epoch": 0.71, + "learning_rate": 4.933206272199602e-05, + "loss": 2.1769, + "step": 8210 + }, + { + "epoch": 0.71, + "learning_rate": 4.933119639608421e-05, + "loss": 2.1197, + "step": 8220 + }, + { + "epoch": 0.71, + "learning_rate": 4.9330330070172404e-05, + "loss": 2.1654, + "step": 8230 + }, + { + "epoch": 0.71, + "learning_rate": 4.93294637442606e-05, + "loss": 2.1925, + "step": 8240 + }, + { + "epoch": 0.71, + "learning_rate": 4.9328597418348786e-05, + "loss": 2.1864, + "step": 8250 + }, + { + "epoch": 0.71, + "learning_rate": 4.932773109243698e-05, + "loss": 2.1682, + "step": 8260 + }, + { + "epoch": 0.72, + "learning_rate": 4.9326864766525175e-05, + "loss": 2.1366, + "step": 8270 + }, + { + "epoch": 0.72, + "learning_rate": 4.932599844061336e-05, + "loss": 2.2343, + "step": 8280 + }, + { + "epoch": 0.72, + "learning_rate": 4.932513211470156e-05, + "loss": 2.1916, + "step": 8290 + }, + { + "epoch": 0.72, + "learning_rate": 4.9324265788789744e-05, + "loss": 2.1482, + "step": 8300 + }, + { + "epoch": 0.72, + "learning_rate": 4.932339946287794e-05, + "loss": 2.1533, + "step": 8310 + }, + { + "epoch": 0.72, + "learning_rate": 4.932253313696613e-05, + "loss": 2.1323, + "step": 8320 + }, + { + "epoch": 0.72, + "learning_rate": 4.932166681105432e-05, + "loss": 2.201, + "step": 8330 + }, + { + "epoch": 0.72, + "learning_rate": 4.9320800485142515e-05, + "loss": 2.1422, + "step": 8340 + }, + { + "epoch": 0.72, + "learning_rate": 4.931993415923071e-05, + "loss": 2.1627, + "step": 8350 + }, + { + "epoch": 0.72, + "learning_rate": 4.93190678333189e-05, + "loss": 2.1373, + "step": 8360 + }, + { + "epoch": 0.72, + "learning_rate": 4.931820150740709e-05, + "loss": 2.2303, + "step": 8370 + }, + { + "epoch": 0.73, + "learning_rate": 4.9317335181495286e-05, + "loss": 2.1791, + "step": 8380 + }, + { + "epoch": 0.73, + "learning_rate": 4.9316468855583474e-05, + "loss": 2.1342, + "step": 8390 + }, + { + "epoch": 0.73, + "learning_rate": 4.931560252967167e-05, + "loss": 2.1738, + "step": 8400 + }, + { + "epoch": 0.73, + "learning_rate": 4.9314736203759856e-05, + "loss": 2.1333, + "step": 8410 + }, + { + "epoch": 0.73, + "learning_rate": 4.931386987784805e-05, + "loss": 2.2049, + "step": 8420 + }, + { + "epoch": 0.73, + "learning_rate": 4.9313003551936245e-05, + "loss": 2.1691, + "step": 8430 + }, + { + "epoch": 0.73, + "learning_rate": 4.931213722602443e-05, + "loss": 2.1832, + "step": 8440 + }, + { + "epoch": 0.73, + "learning_rate": 4.931127090011263e-05, + "loss": 2.2123, + "step": 8450 + }, + { + "epoch": 0.73, + "learning_rate": 4.931040457420082e-05, + "loss": 2.1307, + "step": 8460 + }, + { + "epoch": 0.73, + "learning_rate": 4.930953824828901e-05, + "loss": 2.1608, + "step": 8470 + }, + { + "epoch": 0.73, + "learning_rate": 4.9308671922377204e-05, + "loss": 2.145, + "step": 8480 + }, + { + "epoch": 0.73, + "learning_rate": 4.93078055964654e-05, + "loss": 2.108, + "step": 8490 + }, + { + "epoch": 0.74, + "learning_rate": 4.9306939270553586e-05, + "loss": 2.1585, + "step": 8500 + }, + { + "epoch": 0.74, + "learning_rate": 4.930607294464178e-05, + "loss": 2.181, + "step": 8510 + }, + { + "epoch": 0.74, + "learning_rate": 4.930520661872997e-05, + "loss": 2.1725, + "step": 8520 + }, + { + "epoch": 0.74, + "learning_rate": 4.930434029281816e-05, + "loss": 2.1952, + "step": 8530 + }, + { + "epoch": 0.74, + "learning_rate": 4.930347396690636e-05, + "loss": 2.1596, + "step": 8540 + }, + { + "epoch": 0.74, + "learning_rate": 4.9302607640994545e-05, + "loss": 2.1618, + "step": 8550 + }, + { + "epoch": 0.74, + "learning_rate": 4.930174131508274e-05, + "loss": 2.1516, + "step": 8560 + }, + { + "epoch": 0.74, + "learning_rate": 4.9300874989170934e-05, + "loss": 2.1884, + "step": 8570 + }, + { + "epoch": 0.74, + "learning_rate": 4.930000866325912e-05, + "loss": 2.1319, + "step": 8580 + }, + { + "epoch": 0.74, + "learning_rate": 4.9299142337347316e-05, + "loss": 2.217, + "step": 8590 + }, + { + "epoch": 0.74, + "learning_rate": 4.929827601143551e-05, + "loss": 2.1509, + "step": 8600 + }, + { + "epoch": 0.75, + "learning_rate": 4.92974096855237e-05, + "loss": 2.1178, + "step": 8610 + }, + { + "epoch": 0.75, + "learning_rate": 4.929654335961189e-05, + "loss": 2.1439, + "step": 8620 + }, + { + "epoch": 0.75, + "learning_rate": 4.929567703370008e-05, + "loss": 2.1989, + "step": 8630 + }, + { + "epoch": 0.75, + "learning_rate": 4.9294810707788274e-05, + "loss": 2.1595, + "step": 8640 + }, + { + "epoch": 0.75, + "learning_rate": 4.929394438187647e-05, + "loss": 2.1414, + "step": 8650 + }, + { + "epoch": 0.75, + "learning_rate": 4.929307805596466e-05, + "loss": 2.0917, + "step": 8660 + }, + { + "epoch": 0.75, + "learning_rate": 4.929221173005285e-05, + "loss": 2.1484, + "step": 8670 + }, + { + "epoch": 0.75, + "learning_rate": 4.9291345404141046e-05, + "loss": 2.1668, + "step": 8680 + }, + { + "epoch": 0.75, + "learning_rate": 4.929047907822923e-05, + "loss": 2.1618, + "step": 8690 + }, + { + "epoch": 0.75, + "learning_rate": 4.928961275231743e-05, + "loss": 2.1801, + "step": 8700 + }, + { + "epoch": 0.75, + "learning_rate": 4.9288746426405615e-05, + "loss": 2.1296, + "step": 8710 + }, + { + "epoch": 0.75, + "learning_rate": 4.928788010049381e-05, + "loss": 2.121, + "step": 8720 + }, + { + "epoch": 0.76, + "learning_rate": 4.9287013774582004e-05, + "loss": 2.145, + "step": 8730 + }, + { + "epoch": 0.76, + "learning_rate": 4.928614744867019e-05, + "loss": 2.1122, + "step": 8740 + }, + { + "epoch": 0.76, + "learning_rate": 4.9285281122758386e-05, + "loss": 2.1914, + "step": 8750 + }, + { + "epoch": 0.76, + "learning_rate": 4.928441479684658e-05, + "loss": 2.1296, + "step": 8760 + }, + { + "epoch": 0.76, + "learning_rate": 4.928354847093477e-05, + "loss": 2.1735, + "step": 8770 + }, + { + "epoch": 0.76, + "learning_rate": 4.928268214502296e-05, + "loss": 2.1048, + "step": 8780 + }, + { + "epoch": 0.76, + "learning_rate": 4.928181581911116e-05, + "loss": 2.1473, + "step": 8790 + }, + { + "epoch": 0.76, + "learning_rate": 4.9280949493199345e-05, + "loss": 2.1151, + "step": 8800 + }, + { + "epoch": 0.76, + "learning_rate": 4.928008316728754e-05, + "loss": 2.1368, + "step": 8810 + }, + { + "epoch": 0.76, + "learning_rate": 4.927921684137573e-05, + "loss": 2.1458, + "step": 8820 + }, + { + "epoch": 0.76, + "learning_rate": 4.927835051546392e-05, + "loss": 2.1575, + "step": 8830 + }, + { + "epoch": 0.77, + "learning_rate": 4.9277484189552116e-05, + "loss": 2.184, + "step": 8840 + }, + { + "epoch": 0.77, + "learning_rate": 4.9276617863640304e-05, + "loss": 2.1405, + "step": 8850 + }, + { + "epoch": 0.77, + "learning_rate": 4.92757515377285e-05, + "loss": 2.1268, + "step": 8860 + }, + { + "epoch": 0.77, + "learning_rate": 4.927488521181669e-05, + "loss": 2.1598, + "step": 8870 + }, + { + "epoch": 0.77, + "learning_rate": 4.927401888590488e-05, + "loss": 2.1166, + "step": 8880 + }, + { + "epoch": 0.77, + "learning_rate": 4.9273152559993075e-05, + "loss": 2.1523, + "step": 8890 + }, + { + "epoch": 0.77, + "learning_rate": 4.927228623408127e-05, + "loss": 2.1185, + "step": 8900 + }, + { + "epoch": 0.77, + "learning_rate": 4.927141990816946e-05, + "loss": 2.1211, + "step": 8910 + }, + { + "epoch": 0.77, + "learning_rate": 4.927055358225765e-05, + "loss": 2.1545, + "step": 8920 + }, + { + "epoch": 0.77, + "learning_rate": 4.926968725634584e-05, + "loss": 2.1194, + "step": 8930 + }, + { + "epoch": 0.77, + "learning_rate": 4.9268820930434034e-05, + "loss": 2.1018, + "step": 8940 + }, + { + "epoch": 0.77, + "learning_rate": 4.926795460452223e-05, + "loss": 2.1651, + "step": 8950 + }, + { + "epoch": 0.78, + "learning_rate": 4.9267088278610416e-05, + "loss": 2.1248, + "step": 8960 + }, + { + "epoch": 0.78, + "learning_rate": 4.926622195269861e-05, + "loss": 2.1467, + "step": 8970 + }, + { + "epoch": 0.78, + "learning_rate": 4.9265355626786805e-05, + "loss": 2.1842, + "step": 8980 + }, + { + "epoch": 0.78, + "learning_rate": 4.926448930087499e-05, + "loss": 2.0957, + "step": 8990 + }, + { + "epoch": 0.78, + "learning_rate": 4.926362297496319e-05, + "loss": 2.1533, + "step": 9000 + }, + { + "epoch": 0.78, + "learning_rate": 4.926275664905138e-05, + "loss": 2.142, + "step": 9010 + }, + { + "epoch": 0.78, + "learning_rate": 4.926189032313957e-05, + "loss": 2.1405, + "step": 9020 + }, + { + "epoch": 0.78, + "learning_rate": 4.926102399722776e-05, + "loss": 2.1062, + "step": 9030 + }, + { + "epoch": 0.78, + "learning_rate": 4.926015767131595e-05, + "loss": 2.1364, + "step": 9040 + }, + { + "epoch": 0.78, + "learning_rate": 4.9259291345404145e-05, + "loss": 2.2067, + "step": 9050 + }, + { + "epoch": 0.78, + "learning_rate": 4.925842501949234e-05, + "loss": 2.161, + "step": 9060 + }, + { + "epoch": 0.79, + "learning_rate": 4.925755869358053e-05, + "loss": 2.1555, + "step": 9070 + }, + { + "epoch": 0.79, + "learning_rate": 4.925669236766872e-05, + "loss": 2.1674, + "step": 9080 + }, + { + "epoch": 0.79, + "learning_rate": 4.9255826041756916e-05, + "loss": 2.126, + "step": 9090 + }, + { + "epoch": 0.79, + "learning_rate": 4.9254959715845104e-05, + "loss": 2.1495, + "step": 9100 + }, + { + "epoch": 0.79, + "learning_rate": 4.92540933899333e-05, + "loss": 2.1736, + "step": 9110 + }, + { + "epoch": 0.79, + "learning_rate": 4.925322706402149e-05, + "loss": 2.1375, + "step": 9120 + }, + { + "epoch": 0.79, + "learning_rate": 4.925236073810968e-05, + "loss": 2.1252, + "step": 9130 + }, + { + "epoch": 0.79, + "learning_rate": 4.9251494412197875e-05, + "loss": 2.1027, + "step": 9140 + }, + { + "epoch": 0.79, + "learning_rate": 4.925062808628606e-05, + "loss": 2.1217, + "step": 9150 + }, + { + "epoch": 0.79, + "learning_rate": 4.924976176037426e-05, + "loss": 2.1208, + "step": 9160 + }, + { + "epoch": 0.79, + "learning_rate": 4.924889543446245e-05, + "loss": 2.1537, + "step": 9170 + }, + { + "epoch": 0.79, + "learning_rate": 4.924802910855064e-05, + "loss": 2.129, + "step": 9180 + }, + { + "epoch": 0.8, + "learning_rate": 4.9247162782638834e-05, + "loss": 2.1908, + "step": 9190 + }, + { + "epoch": 0.8, + "learning_rate": 4.924629645672703e-05, + "loss": 2.1443, + "step": 9200 + }, + { + "epoch": 0.8, + "learning_rate": 4.9245430130815216e-05, + "loss": 2.1125, + "step": 9210 + }, + { + "epoch": 0.8, + "learning_rate": 4.924456380490341e-05, + "loss": 2.1476, + "step": 9220 + }, + { + "epoch": 0.8, + "learning_rate": 4.9243697478991605e-05, + "loss": 2.0907, + "step": 9230 + }, + { + "epoch": 0.8, + "learning_rate": 4.924283115307979e-05, + "loss": 2.1868, + "step": 9240 + }, + { + "epoch": 0.8, + "learning_rate": 4.924196482716799e-05, + "loss": 2.171, + "step": 9250 + }, + { + "epoch": 0.8, + "learning_rate": 4.9241098501256175e-05, + "loss": 2.1531, + "step": 9260 + }, + { + "epoch": 0.8, + "learning_rate": 4.924023217534437e-05, + "loss": 2.1694, + "step": 9270 + }, + { + "epoch": 0.8, + "learning_rate": 4.9239365849432564e-05, + "loss": 2.0858, + "step": 9280 + }, + { + "epoch": 0.8, + "learning_rate": 4.923849952352075e-05, + "loss": 2.1501, + "step": 9290 + }, + { + "epoch": 0.8, + "learning_rate": 4.9237633197608946e-05, + "loss": 2.1219, + "step": 9300 + }, + { + "epoch": 0.81, + "learning_rate": 4.923676687169714e-05, + "loss": 2.1209, + "step": 9310 + }, + { + "epoch": 0.81, + "learning_rate": 4.923590054578533e-05, + "loss": 2.1582, + "step": 9320 + }, + { + "epoch": 0.81, + "learning_rate": 4.923503421987352e-05, + "loss": 2.1377, + "step": 9330 + }, + { + "epoch": 0.81, + "learning_rate": 4.923416789396172e-05, + "loss": 2.1419, + "step": 9340 + }, + { + "epoch": 0.81, + "learning_rate": 4.9233301568049904e-05, + "loss": 2.1599, + "step": 9350 + }, + { + "epoch": 0.81, + "learning_rate": 4.92324352421381e-05, + "loss": 2.1394, + "step": 9360 + }, + { + "epoch": 0.81, + "learning_rate": 4.9231568916226287e-05, + "loss": 2.1048, + "step": 9370 + }, + { + "epoch": 0.81, + "learning_rate": 4.923070259031448e-05, + "loss": 2.1419, + "step": 9380 + }, + { + "epoch": 0.81, + "learning_rate": 4.9229836264402675e-05, + "loss": 2.1776, + "step": 9390 + }, + { + "epoch": 0.81, + "learning_rate": 4.922896993849086e-05, + "loss": 2.1031, + "step": 9400 + }, + { + "epoch": 0.81, + "learning_rate": 4.922810361257906e-05, + "loss": 2.1067, + "step": 9410 + }, + { + "epoch": 0.82, + "learning_rate": 4.922723728666725e-05, + "loss": 2.1205, + "step": 9420 + }, + { + "epoch": 0.82, + "learning_rate": 4.922637096075544e-05, + "loss": 2.1335, + "step": 9430 + }, + { + "epoch": 0.82, + "learning_rate": 4.9225504634843634e-05, + "loss": 2.161, + "step": 9440 + }, + { + "epoch": 0.82, + "learning_rate": 4.922463830893182e-05, + "loss": 2.1614, + "step": 9450 + }, + { + "epoch": 0.82, + "learning_rate": 4.9223771983020016e-05, + "loss": 2.1215, + "step": 9460 + }, + { + "epoch": 0.82, + "learning_rate": 4.922290565710821e-05, + "loss": 2.0966, + "step": 9470 + }, + { + "epoch": 0.82, + "learning_rate": 4.92220393311964e-05, + "loss": 2.1464, + "step": 9480 + }, + { + "epoch": 0.82, + "learning_rate": 4.922117300528459e-05, + "loss": 2.1708, + "step": 9490 + }, + { + "epoch": 0.82, + "learning_rate": 4.922030667937279e-05, + "loss": 2.1253, + "step": 9500 + }, + { + "epoch": 0.82, + "learning_rate": 4.9219440353460975e-05, + "loss": 2.1501, + "step": 9510 + }, + { + "epoch": 0.82, + "learning_rate": 4.921857402754916e-05, + "loss": 2.1402, + "step": 9520 + }, + { + "epoch": 0.82, + "learning_rate": 4.921770770163736e-05, + "loss": 2.1114, + "step": 9530 + }, + { + "epoch": 0.83, + "learning_rate": 4.9216841375725545e-05, + "loss": 2.1493, + "step": 9540 + }, + { + "epoch": 0.83, + "learning_rate": 4.921597504981374e-05, + "loss": 2.1304, + "step": 9550 + }, + { + "epoch": 0.83, + "learning_rate": 4.9215108723901934e-05, + "loss": 2.1352, + "step": 9560 + }, + { + "epoch": 0.83, + "learning_rate": 4.921424239799012e-05, + "loss": 2.1365, + "step": 9570 + }, + { + "epoch": 0.83, + "learning_rate": 4.9213376072078316e-05, + "loss": 2.1464, + "step": 9580 + }, + { + "epoch": 0.83, + "learning_rate": 4.921250974616651e-05, + "loss": 2.1256, + "step": 9590 + }, + { + "epoch": 0.83, + "learning_rate": 4.92116434202547e-05, + "loss": 2.1539, + "step": 9600 + }, + { + "epoch": 0.83, + "learning_rate": 4.921077709434289e-05, + "loss": 2.1006, + "step": 9610 + }, + { + "epoch": 0.83, + "learning_rate": 4.920991076843108e-05, + "loss": 2.1623, + "step": 9620 + }, + { + "epoch": 0.83, + "learning_rate": 4.9209044442519275e-05, + "loss": 2.1287, + "step": 9630 + }, + { + "epoch": 0.83, + "learning_rate": 4.920817811660747e-05, + "loss": 2.1132, + "step": 9640 + }, + { + "epoch": 0.84, + "learning_rate": 4.920731179069566e-05, + "loss": 2.149, + "step": 9650 + }, + { + "epoch": 0.84, + "learning_rate": 4.920644546478385e-05, + "loss": 2.1436, + "step": 9660 + }, + { + "epoch": 0.84, + "learning_rate": 4.9205579138872046e-05, + "loss": 2.1416, + "step": 9670 + }, + { + "epoch": 0.84, + "learning_rate": 4.920471281296023e-05, + "loss": 2.1182, + "step": 9680 + }, + { + "epoch": 0.84, + "learning_rate": 4.920384648704843e-05, + "loss": 2.0704, + "step": 9690 + }, + { + "epoch": 0.84, + "learning_rate": 4.920298016113662e-05, + "loss": 2.1326, + "step": 9700 + }, + { + "epoch": 0.84, + "learning_rate": 4.920211383522481e-05, + "loss": 2.1321, + "step": 9710 + }, + { + "epoch": 0.84, + "learning_rate": 4.9201247509313004e-05, + "loss": 2.1029, + "step": 9720 + }, + { + "epoch": 0.84, + "learning_rate": 4.920038118340119e-05, + "loss": 2.0968, + "step": 9730 + }, + { + "epoch": 0.84, + "learning_rate": 4.9199514857489386e-05, + "loss": 2.1123, + "step": 9740 + }, + { + "epoch": 0.84, + "learning_rate": 4.919864853157758e-05, + "loss": 2.1068, + "step": 9750 + }, + { + "epoch": 0.84, + "learning_rate": 4.919778220566577e-05, + "loss": 2.1389, + "step": 9760 + }, + { + "epoch": 0.85, + "learning_rate": 4.919691587975396e-05, + "loss": 2.134, + "step": 9770 + }, + { + "epoch": 0.85, + "learning_rate": 4.919604955384216e-05, + "loss": 2.109, + "step": 9780 + }, + { + "epoch": 0.85, + "learning_rate": 4.9195183227930345e-05, + "loss": 2.1562, + "step": 9790 + }, + { + "epoch": 0.85, + "learning_rate": 4.919431690201854e-05, + "loss": 2.1118, + "step": 9800 + }, + { + "epoch": 0.85, + "learning_rate": 4.9193450576106734e-05, + "loss": 2.1769, + "step": 9810 + }, + { + "epoch": 0.85, + "learning_rate": 4.919258425019492e-05, + "loss": 2.1488, + "step": 9820 + }, + { + "epoch": 0.85, + "learning_rate": 4.9191717924283116e-05, + "loss": 2.153, + "step": 9830 + }, + { + "epoch": 0.85, + "learning_rate": 4.9190851598371304e-05, + "loss": 2.1003, + "step": 9840 + }, + { + "epoch": 0.85, + "learning_rate": 4.91899852724595e-05, + "loss": 2.1536, + "step": 9850 + }, + { + "epoch": 0.85, + "learning_rate": 4.918911894654769e-05, + "loss": 2.1626, + "step": 9860 + }, + { + "epoch": 0.85, + "learning_rate": 4.918825262063588e-05, + "loss": 2.1275, + "step": 9870 + }, + { + "epoch": 0.86, + "learning_rate": 4.9187386294724075e-05, + "loss": 2.1373, + "step": 9880 + }, + { + "epoch": 0.86, + "learning_rate": 4.918651996881227e-05, + "loss": 2.1383, + "step": 9890 + }, + { + "epoch": 0.86, + "learning_rate": 4.918565364290046e-05, + "loss": 2.1069, + "step": 9900 + }, + { + "epoch": 0.86, + "learning_rate": 4.918478731698865e-05, + "loss": 2.097, + "step": 9910 + }, + { + "epoch": 0.86, + "learning_rate": 4.9183920991076846e-05, + "loss": 2.1143, + "step": 9920 + }, + { + "epoch": 0.86, + "learning_rate": 4.9183054665165034e-05, + "loss": 2.1371, + "step": 9930 + }, + { + "epoch": 0.86, + "learning_rate": 4.918218833925323e-05, + "loss": 2.1477, + "step": 9940 + }, + { + "epoch": 0.86, + "learning_rate": 4.9181322013341416e-05, + "loss": 2.1549, + "step": 9950 + }, + { + "epoch": 0.86, + "learning_rate": 4.918045568742961e-05, + "loss": 2.1487, + "step": 9960 + }, + { + "epoch": 0.86, + "learning_rate": 4.9179589361517805e-05, + "loss": 2.1256, + "step": 9970 + }, + { + "epoch": 0.86, + "learning_rate": 4.917872303560599e-05, + "loss": 2.0673, + "step": 9980 + }, + { + "epoch": 0.86, + "learning_rate": 4.917785670969419e-05, + "loss": 2.1366, + "step": 9990 + }, + { + "epoch": 0.87, + "learning_rate": 4.917699038378238e-05, + "loss": 2.1082, + "step": 10000 + }, + { + "epoch": 0.87, + "learning_rate": 4.917612405787057e-05, + "loss": 2.1297, + "step": 10010 + }, + { + "epoch": 0.87, + "learning_rate": 4.9175257731958763e-05, + "loss": 2.0966, + "step": 10020 + }, + { + "epoch": 0.87, + "learning_rate": 4.917439140604696e-05, + "loss": 2.1337, + "step": 10030 + }, + { + "epoch": 0.87, + "learning_rate": 4.9173525080135146e-05, + "loss": 2.0973, + "step": 10040 + }, + { + "epoch": 0.87, + "learning_rate": 4.917265875422334e-05, + "loss": 2.0612, + "step": 10050 + }, + { + "epoch": 0.87, + "learning_rate": 4.917179242831153e-05, + "loss": 2.172, + "step": 10060 + }, + { + "epoch": 0.87, + "learning_rate": 4.917092610239972e-05, + "loss": 2.1728, + "step": 10070 + }, + { + "epoch": 0.87, + "learning_rate": 4.9170059776487917e-05, + "loss": 2.1581, + "step": 10080 + }, + { + "epoch": 0.87, + "learning_rate": 4.9169193450576104e-05, + "loss": 2.1296, + "step": 10090 + }, + { + "epoch": 0.87, + "learning_rate": 4.91683271246643e-05, + "loss": 2.1054, + "step": 10100 + }, + { + "epoch": 0.88, + "learning_rate": 4.916746079875249e-05, + "loss": 2.1082, + "step": 10110 + }, + { + "epoch": 0.88, + "learning_rate": 4.916659447284068e-05, + "loss": 2.1499, + "step": 10120 + }, + { + "epoch": 0.88, + "learning_rate": 4.9165728146928875e-05, + "loss": 2.1425, + "step": 10130 + }, + { + "epoch": 0.88, + "learning_rate": 4.916486182101707e-05, + "loss": 2.1449, + "step": 10140 + }, + { + "epoch": 0.88, + "learning_rate": 4.916399549510526e-05, + "loss": 2.1008, + "step": 10150 + }, + { + "epoch": 0.88, + "learning_rate": 4.916312916919345e-05, + "loss": 2.1486, + "step": 10160 + }, + { + "epoch": 0.88, + "learning_rate": 4.916226284328164e-05, + "loss": 2.0942, + "step": 10170 + }, + { + "epoch": 0.88, + "learning_rate": 4.9161396517369834e-05, + "loss": 2.0979, + "step": 10180 + }, + { + "epoch": 0.88, + "learning_rate": 4.916053019145803e-05, + "loss": 2.1571, + "step": 10190 + }, + { + "epoch": 0.88, + "learning_rate": 4.9159663865546216e-05, + "loss": 2.157, + "step": 10200 + }, + { + "epoch": 0.88, + "learning_rate": 4.915879753963441e-05, + "loss": 2.1217, + "step": 10210 + }, + { + "epoch": 0.88, + "learning_rate": 4.9157931213722605e-05, + "loss": 2.1387, + "step": 10220 + }, + { + "epoch": 0.89, + "learning_rate": 4.915706488781079e-05, + "loss": 2.14, + "step": 10230 + }, + { + "epoch": 0.89, + "learning_rate": 4.915619856189899e-05, + "loss": 2.1568, + "step": 10240 + }, + { + "epoch": 0.89, + "learning_rate": 4.9155332235987175e-05, + "loss": 2.0931, + "step": 10250 + }, + { + "epoch": 0.89, + "learning_rate": 4.915446591007537e-05, + "loss": 2.1637, + "step": 10260 + }, + { + "epoch": 0.89, + "learning_rate": 4.9153599584163564e-05, + "loss": 2.1342, + "step": 10270 + }, + { + "epoch": 0.89, + "learning_rate": 4.915273325825175e-05, + "loss": 2.1382, + "step": 10280 + }, + { + "epoch": 0.89, + "learning_rate": 4.9151866932339946e-05, + "loss": 2.1124, + "step": 10290 + }, + { + "epoch": 0.89, + "learning_rate": 4.915100060642814e-05, + "loss": 2.0749, + "step": 10300 + }, + { + "epoch": 0.89, + "learning_rate": 4.915013428051633e-05, + "loss": 2.1077, + "step": 10310 + }, + { + "epoch": 0.89, + "learning_rate": 4.914926795460452e-05, + "loss": 2.1093, + "step": 10320 + }, + { + "epoch": 0.89, + "learning_rate": 4.914840162869272e-05, + "loss": 2.1295, + "step": 10330 + }, + { + "epoch": 0.89, + "learning_rate": 4.9147535302780905e-05, + "loss": 2.1297, + "step": 10340 + }, + { + "epoch": 0.9, + "learning_rate": 4.91466689768691e-05, + "loss": 2.1048, + "step": 10350 + }, + { + "epoch": 0.9, + "learning_rate": 4.914580265095729e-05, + "loss": 2.1245, + "step": 10360 + }, + { + "epoch": 0.9, + "learning_rate": 4.914493632504548e-05, + "loss": 2.0768, + "step": 10370 + }, + { + "epoch": 0.9, + "learning_rate": 4.9144069999133676e-05, + "loss": 2.0722, + "step": 10380 + }, + { + "epoch": 0.9, + "learning_rate": 4.914320367322186e-05, + "loss": 2.1455, + "step": 10390 + }, + { + "epoch": 0.9, + "learning_rate": 4.914233734731006e-05, + "loss": 2.1108, + "step": 10400 + }, + { + "epoch": 0.9, + "learning_rate": 4.914147102139825e-05, + "loss": 2.0821, + "step": 10410 + }, + { + "epoch": 0.9, + "learning_rate": 4.914060469548644e-05, + "loss": 2.0995, + "step": 10420 + }, + { + "epoch": 0.9, + "learning_rate": 4.9139738369574634e-05, + "loss": 2.1056, + "step": 10430 + }, + { + "epoch": 0.9, + "learning_rate": 4.913887204366283e-05, + "loss": 2.1212, + "step": 10440 + }, + { + "epoch": 0.9, + "learning_rate": 4.9138005717751016e-05, + "loss": 2.1449, + "step": 10450 + }, + { + "epoch": 0.91, + "learning_rate": 4.913713939183921e-05, + "loss": 2.1029, + "step": 10460 + }, + { + "epoch": 0.91, + "learning_rate": 4.91362730659274e-05, + "loss": 2.0612, + "step": 10470 + }, + { + "epoch": 0.91, + "learning_rate": 4.913540674001559e-05, + "loss": 2.0924, + "step": 10480 + }, + { + "epoch": 0.91, + "learning_rate": 4.913454041410379e-05, + "loss": 2.125, + "step": 10490 + }, + { + "epoch": 0.91, + "learning_rate": 4.9133674088191975e-05, + "loss": 2.0519, + "step": 10500 + }, + { + "epoch": 0.91, + "learning_rate": 4.913280776228017e-05, + "loss": 2.1768, + "step": 10510 + }, + { + "epoch": 0.91, + "learning_rate": 4.9131941436368364e-05, + "loss": 2.1037, + "step": 10520 + }, + { + "epoch": 0.91, + "learning_rate": 4.913107511045655e-05, + "loss": 2.1255, + "step": 10530 + }, + { + "epoch": 0.91, + "learning_rate": 4.9130208784544746e-05, + "loss": 2.1717, + "step": 10540 + }, + { + "epoch": 0.91, + "learning_rate": 4.912934245863294e-05, + "loss": 2.0841, + "step": 10550 + }, + { + "epoch": 0.91, + "learning_rate": 4.912847613272113e-05, + "loss": 2.0945, + "step": 10560 + }, + { + "epoch": 0.91, + "learning_rate": 4.912760980680932e-05, + "loss": 2.072, + "step": 10570 + }, + { + "epoch": 0.92, + "learning_rate": 4.912674348089751e-05, + "loss": 2.0745, + "step": 10580 + }, + { + "epoch": 0.92, + "learning_rate": 4.9125877154985705e-05, + "loss": 2.1109, + "step": 10590 + }, + { + "epoch": 0.92, + "learning_rate": 4.91250108290739e-05, + "loss": 2.0667, + "step": 10600 + }, + { + "epoch": 0.92, + "learning_rate": 4.912414450316209e-05, + "loss": 2.0917, + "step": 10610 + }, + { + "epoch": 0.92, + "learning_rate": 4.912327817725028e-05, + "loss": 2.1162, + "step": 10620 + }, + { + "epoch": 0.92, + "learning_rate": 4.9122411851338476e-05, + "loss": 2.0851, + "step": 10630 + }, + { + "epoch": 0.92, + "learning_rate": 4.9121545525426664e-05, + "loss": 2.1146, + "step": 10640 + }, + { + "epoch": 0.92, + "learning_rate": 4.912067919951486e-05, + "loss": 2.1387, + "step": 10650 + }, + { + "epoch": 0.92, + "learning_rate": 4.911981287360305e-05, + "loss": 2.1038, + "step": 10660 + }, + { + "epoch": 0.92, + "learning_rate": 4.911894654769124e-05, + "loss": 2.0983, + "step": 10670 + }, + { + "epoch": 0.92, + "learning_rate": 4.9118080221779435e-05, + "loss": 2.0501, + "step": 10680 + }, + { + "epoch": 0.93, + "learning_rate": 4.911721389586762e-05, + "loss": 2.1357, + "step": 10690 + }, + { + "epoch": 0.93, + "learning_rate": 4.911634756995582e-05, + "loss": 2.086, + "step": 10700 + }, + { + "epoch": 0.93, + "learning_rate": 4.911548124404401e-05, + "loss": 2.0925, + "step": 10710 + }, + { + "epoch": 0.93, + "learning_rate": 4.91146149181322e-05, + "loss": 2.0582, + "step": 10720 + }, + { + "epoch": 0.93, + "learning_rate": 4.911374859222039e-05, + "loss": 2.1196, + "step": 10730 + }, + { + "epoch": 0.93, + "learning_rate": 4.911288226630859e-05, + "loss": 2.1452, + "step": 10740 + }, + { + "epoch": 0.93, + "learning_rate": 4.9112015940396776e-05, + "loss": 2.098, + "step": 10750 + }, + { + "epoch": 0.93, + "learning_rate": 4.911114961448497e-05, + "loss": 2.097, + "step": 10760 + }, + { + "epoch": 0.93, + "learning_rate": 4.9110283288573164e-05, + "loss": 2.1383, + "step": 10770 + }, + { + "epoch": 0.93, + "learning_rate": 4.910941696266135e-05, + "loss": 2.0703, + "step": 10780 + }, + { + "epoch": 0.93, + "learning_rate": 4.9108550636749547e-05, + "loss": 2.1026, + "step": 10790 + }, + { + "epoch": 0.93, + "learning_rate": 4.9107684310837734e-05, + "loss": 2.0948, + "step": 10800 + }, + { + "epoch": 0.94, + "learning_rate": 4.910681798492593e-05, + "loss": 2.1316, + "step": 10810 + }, + { + "epoch": 0.94, + "learning_rate": 4.910595165901412e-05, + "loss": 2.048, + "step": 10820 + }, + { + "epoch": 0.94, + "learning_rate": 4.910508533310231e-05, + "loss": 2.0599, + "step": 10830 + }, + { + "epoch": 0.94, + "learning_rate": 4.9104219007190505e-05, + "loss": 2.1244, + "step": 10840 + }, + { + "epoch": 0.94, + "learning_rate": 4.91033526812787e-05, + "loss": 2.1168, + "step": 10850 + }, + { + "epoch": 0.94, + "learning_rate": 4.910248635536689e-05, + "loss": 2.104, + "step": 10860 + }, + { + "epoch": 0.94, + "learning_rate": 4.910162002945508e-05, + "loss": 2.1105, + "step": 10870 + }, + { + "epoch": 0.94, + "learning_rate": 4.9100753703543276e-05, + "loss": 2.0875, + "step": 10880 + }, + { + "epoch": 0.94, + "learning_rate": 4.9099887377631464e-05, + "loss": 2.1195, + "step": 10890 + }, + { + "epoch": 0.94, + "learning_rate": 4.909902105171966e-05, + "loss": 2.0825, + "step": 10900 + }, + { + "epoch": 0.94, + "learning_rate": 4.9098154725807846e-05, + "loss": 2.0674, + "step": 10910 + }, + { + "epoch": 0.95, + "learning_rate": 4.909728839989604e-05, + "loss": 2.0612, + "step": 10920 + }, + { + "epoch": 0.95, + "learning_rate": 4.9096422073984235e-05, + "loss": 2.0954, + "step": 10930 + }, + { + "epoch": 0.95, + "learning_rate": 4.909555574807242e-05, + "loss": 2.0908, + "step": 10940 + }, + { + "epoch": 0.95, + "learning_rate": 4.909468942216062e-05, + "loss": 2.1391, + "step": 10950 + }, + { + "epoch": 0.95, + "learning_rate": 4.909382309624881e-05, + "loss": 2.0499, + "step": 10960 + }, + { + "epoch": 0.95, + "learning_rate": 4.9092956770337e-05, + "loss": 2.0742, + "step": 10970 + }, + { + "epoch": 0.95, + "learning_rate": 4.9092090444425194e-05, + "loss": 2.0932, + "step": 10980 + }, + { + "epoch": 0.95, + "learning_rate": 4.909122411851338e-05, + "loss": 2.117, + "step": 10990 + }, + { + "epoch": 0.95, + "learning_rate": 4.9090357792601576e-05, + "loss": 2.0845, + "step": 11000 + }, + { + "epoch": 0.95, + "learning_rate": 4.908949146668977e-05, + "loss": 2.1285, + "step": 11010 + }, + { + "epoch": 0.95, + "learning_rate": 4.908862514077796e-05, + "loss": 2.0853, + "step": 11020 + }, + { + "epoch": 0.95, + "learning_rate": 4.908775881486615e-05, + "loss": 2.0573, + "step": 11030 + }, + { + "epoch": 0.96, + "learning_rate": 4.908689248895435e-05, + "loss": 2.0942, + "step": 11040 + }, + { + "epoch": 0.96, + "learning_rate": 4.9086026163042535e-05, + "loss": 2.1458, + "step": 11050 + }, + { + "epoch": 0.96, + "learning_rate": 4.908515983713073e-05, + "loss": 2.1288, + "step": 11060 + }, + { + "epoch": 0.96, + "learning_rate": 4.9084293511218923e-05, + "loss": 2.098, + "step": 11070 + }, + { + "epoch": 0.96, + "learning_rate": 4.908342718530711e-05, + "loss": 2.1029, + "step": 11080 + }, + { + "epoch": 0.96, + "learning_rate": 4.9082560859395306e-05, + "loss": 2.0683, + "step": 11090 + }, + { + "epoch": 0.96, + "learning_rate": 4.908169453348349e-05, + "loss": 2.0882, + "step": 11100 + }, + { + "epoch": 0.96, + "learning_rate": 4.908082820757169e-05, + "loss": 2.1452, + "step": 11110 + }, + { + "epoch": 0.96, + "learning_rate": 4.907996188165988e-05, + "loss": 2.1113, + "step": 11120 + }, + { + "epoch": 0.96, + "learning_rate": 4.907909555574807e-05, + "loss": 2.0975, + "step": 11130 + }, + { + "epoch": 0.96, + "learning_rate": 4.9078229229836264e-05, + "loss": 2.1373, + "step": 11140 + }, + { + "epoch": 0.97, + "learning_rate": 4.907736290392446e-05, + "loss": 2.1128, + "step": 11150 + }, + { + "epoch": 0.97, + "learning_rate": 4.9076496578012646e-05, + "loss": 2.0888, + "step": 11160 + }, + { + "epoch": 0.97, + "learning_rate": 4.907563025210084e-05, + "loss": 2.078, + "step": 11170 + }, + { + "epoch": 0.97, + "learning_rate": 4.9074763926189035e-05, + "loss": 2.0355, + "step": 11180 + }, + { + "epoch": 0.97, + "learning_rate": 4.907389760027722e-05, + "loss": 2.1222, + "step": 11190 + }, + { + "epoch": 0.97, + "learning_rate": 4.907303127436542e-05, + "loss": 2.0852, + "step": 11200 + }, + { + "epoch": 0.97, + "learning_rate": 4.9072164948453605e-05, + "loss": 2.1092, + "step": 11210 + }, + { + "epoch": 0.97, + "learning_rate": 4.90712986225418e-05, + "loss": 2.116, + "step": 11220 + }, + { + "epoch": 0.97, + "learning_rate": 4.9070432296629994e-05, + "loss": 2.0696, + "step": 11230 + }, + { + "epoch": 0.97, + "learning_rate": 4.906956597071818e-05, + "loss": 2.1294, + "step": 11240 + }, + { + "epoch": 0.97, + "learning_rate": 4.9068699644806376e-05, + "loss": 2.053, + "step": 11250 + }, + { + "epoch": 0.97, + "learning_rate": 4.906783331889457e-05, + "loss": 2.0644, + "step": 11260 + }, + { + "epoch": 0.98, + "learning_rate": 4.906696699298276e-05, + "loss": 2.0727, + "step": 11270 + }, + { + "epoch": 0.98, + "learning_rate": 4.906610066707095e-05, + "loss": 2.0846, + "step": 11280 + }, + { + "epoch": 0.98, + "learning_rate": 4.906523434115915e-05, + "loss": 2.1143, + "step": 11290 + }, + { + "epoch": 0.98, + "learning_rate": 4.9064368015247335e-05, + "loss": 2.0861, + "step": 11300 + }, + { + "epoch": 0.98, + "learning_rate": 4.906350168933553e-05, + "loss": 2.0885, + "step": 11310 + }, + { + "epoch": 0.98, + "learning_rate": 4.906263536342372e-05, + "loss": 2.113, + "step": 11320 + }, + { + "epoch": 0.98, + "learning_rate": 4.906176903751191e-05, + "loss": 2.0806, + "step": 11330 + }, + { + "epoch": 0.98, + "learning_rate": 4.9060902711600106e-05, + "loss": 2.0965, + "step": 11340 + }, + { + "epoch": 0.98, + "learning_rate": 4.9060036385688294e-05, + "loss": 2.0828, + "step": 11350 + }, + { + "epoch": 0.98, + "learning_rate": 4.905917005977649e-05, + "loss": 2.1019, + "step": 11360 + }, + { + "epoch": 0.98, + "learning_rate": 4.905830373386468e-05, + "loss": 2.09, + "step": 11370 + }, + { + "epoch": 0.98, + "learning_rate": 4.905743740795287e-05, + "loss": 2.1044, + "step": 11380 + }, + { + "epoch": 0.99, + "learning_rate": 4.9056571082041065e-05, + "loss": 2.1266, + "step": 11390 + }, + { + "epoch": 0.99, + "learning_rate": 4.905570475612926e-05, + "loss": 2.0778, + "step": 11400 + }, + { + "epoch": 0.99, + "learning_rate": 4.905483843021745e-05, + "loss": 2.1026, + "step": 11410 + }, + { + "epoch": 0.99, + "learning_rate": 4.905397210430564e-05, + "loss": 2.0957, + "step": 11420 + }, + { + "epoch": 0.99, + "learning_rate": 4.905310577839383e-05, + "loss": 2.1224, + "step": 11430 + }, + { + "epoch": 0.99, + "learning_rate": 4.905223945248202e-05, + "loss": 2.1144, + "step": 11440 + }, + { + "epoch": 0.99, + "learning_rate": 4.905137312657022e-05, + "loss": 2.0836, + "step": 11450 + }, + { + "epoch": 0.99, + "learning_rate": 4.9050506800658405e-05, + "loss": 2.0948, + "step": 11460 + }, + { + "epoch": 0.99, + "learning_rate": 4.90496404747466e-05, + "loss": 2.05, + "step": 11470 + }, + { + "epoch": 0.99, + "learning_rate": 4.9048774148834794e-05, + "loss": 2.0855, + "step": 11480 + }, + { + "epoch": 0.99, + "learning_rate": 4.904790782292298e-05, + "loss": 2.0895, + "step": 11490 + }, + { + "epoch": 1.0, + "learning_rate": 4.9047041497011177e-05, + "loss": 2.0853, + "step": 11500 + }, + { + "epoch": 1.0, + "learning_rate": 4.904617517109937e-05, + "loss": 2.0674, + "step": 11510 + }, + { + "epoch": 1.0, + "learning_rate": 4.904530884518756e-05, + "loss": 2.0194, + "step": 11520 + }, + { + "epoch": 1.0, + "learning_rate": 4.904444251927575e-05, + "loss": 2.0981, + "step": 11530 + }, + { + "epoch": 1.0, + "learning_rate": 4.904357619336394e-05, + "loss": 2.057, + "step": 11540 + }, + { + "epoch": 1.0, + "learning_rate": 4.9042709867452135e-05, + "loss": 2.099, + "step": 11550 + }, + { + "epoch": 1.0, + "eval_Bleu_1": 0.13794541108454786, + "eval_Bleu_2": 6.687677499867283e-11, + "eval_Bleu_3": 5.582483535547225e-14, + "eval_Bleu_4": 1.7052818968240687e-15, + "eval_ROUGE_L": 0.27926804952729073, + "eval_cer": 0.9959283807015268, + "eval_em": 0, + "eval_f1": 0.027541209841899523, + "eval_loss": 1.9851326942443848, + "eval_runtime": 1417.7337, + "eval_samples_per_second": 3.622, + "eval_steps_per_second": 3.622, + "eval_wer": 0.9855632632021548, + "step": 11553 + }, + { + "epoch": 1.0, + "learning_rate": 4.904184354154033e-05, + "loss": 2.118, + "step": 11560 + }, + { + "epoch": 1.0, + "learning_rate": 4.904097721562852e-05, + "loss": 2.0728, + "step": 11570 + }, + { + "epoch": 1.0, + "learning_rate": 4.904011088971671e-05, + "loss": 2.109, + "step": 11580 + }, + { + "epoch": 1.0, + "learning_rate": 4.9039244563804906e-05, + "loss": 2.1471, + "step": 11590 + }, + { + "epoch": 1.0, + "learning_rate": 4.9038378237893094e-05, + "loss": 2.0988, + "step": 11600 + }, + { + "epoch": 1.0, + "learning_rate": 4.903751191198129e-05, + "loss": 2.0546, + "step": 11610 + }, + { + "epoch": 1.01, + "learning_rate": 4.903664558606948e-05, + "loss": 2.0132, + "step": 11620 + }, + { + "epoch": 1.01, + "learning_rate": 4.903577926015767e-05, + "loss": 2.0773, + "step": 11630 + }, + { + "epoch": 1.01, + "learning_rate": 4.9034912934245865e-05, + "loss": 2.0629, + "step": 11640 + }, + { + "epoch": 1.01, + "learning_rate": 4.903404660833405e-05, + "loss": 2.0992, + "step": 11650 + }, + { + "epoch": 1.01, + "learning_rate": 4.903318028242225e-05, + "loss": 2.0647, + "step": 11660 + }, + { + "epoch": 1.01, + "learning_rate": 4.903231395651044e-05, + "loss": 2.0621, + "step": 11670 + }, + { + "epoch": 1.01, + "learning_rate": 4.903144763059863e-05, + "loss": 2.0709, + "step": 11680 + }, + { + "epoch": 1.01, + "learning_rate": 4.9030581304686824e-05, + "loss": 2.0632, + "step": 11690 + }, + { + "epoch": 1.01, + "learning_rate": 4.902971497877502e-05, + "loss": 2.1068, + "step": 11700 + }, + { + "epoch": 1.01, + "learning_rate": 4.9028848652863206e-05, + "loss": 2.0779, + "step": 11710 + }, + { + "epoch": 1.01, + "learning_rate": 4.90279823269514e-05, + "loss": 2.0132, + "step": 11720 + }, + { + "epoch": 1.02, + "learning_rate": 4.9027116001039595e-05, + "loss": 2.1012, + "step": 11730 + }, + { + "epoch": 1.02, + "learning_rate": 4.902624967512778e-05, + "loss": 2.0975, + "step": 11740 + }, + { + "epoch": 1.02, + "learning_rate": 4.902538334921598e-05, + "loss": 2.0371, + "step": 11750 + }, + { + "epoch": 1.02, + "learning_rate": 4.9024517023304165e-05, + "loss": 2.0794, + "step": 11760 + }, + { + "epoch": 1.02, + "learning_rate": 4.902365069739236e-05, + "loss": 2.1078, + "step": 11770 + }, + { + "epoch": 1.02, + "learning_rate": 4.9022784371480553e-05, + "loss": 2.1558, + "step": 11780 + }, + { + "epoch": 1.02, + "learning_rate": 4.902191804556874e-05, + "loss": 2.1011, + "step": 11790 + }, + { + "epoch": 1.02, + "learning_rate": 4.9021051719656936e-05, + "loss": 2.0681, + "step": 11800 + }, + { + "epoch": 1.02, + "learning_rate": 4.902018539374513e-05, + "loss": 2.0386, + "step": 11810 + }, + { + "epoch": 1.02, + "learning_rate": 4.901931906783332e-05, + "loss": 2.1028, + "step": 11820 + }, + { + "epoch": 1.02, + "learning_rate": 4.901845274192151e-05, + "loss": 2.0318, + "step": 11830 + }, + { + "epoch": 1.02, + "learning_rate": 4.90175864160097e-05, + "loss": 2.0697, + "step": 11840 + }, + { + "epoch": 1.03, + "learning_rate": 4.9016720090097894e-05, + "loss": 2.1136, + "step": 11850 + }, + { + "epoch": 1.03, + "learning_rate": 4.901585376418609e-05, + "loss": 2.095, + "step": 11860 + }, + { + "epoch": 1.03, + "learning_rate": 4.9014987438274276e-05, + "loss": 2.0757, + "step": 11870 + }, + { + "epoch": 1.03, + "learning_rate": 4.901412111236247e-05, + "loss": 2.1089, + "step": 11880 + }, + { + "epoch": 1.03, + "learning_rate": 4.9013254786450665e-05, + "loss": 2.0527, + "step": 11890 + }, + { + "epoch": 1.03, + "learning_rate": 4.901238846053885e-05, + "loss": 2.0975, + "step": 11900 + }, + { + "epoch": 1.03, + "learning_rate": 4.901152213462705e-05, + "loss": 2.0821, + "step": 11910 + }, + { + "epoch": 1.03, + "learning_rate": 4.901065580871524e-05, + "loss": 2.0948, + "step": 11920 + }, + { + "epoch": 1.03, + "learning_rate": 4.900978948280343e-05, + "loss": 2.0855, + "step": 11930 + }, + { + "epoch": 1.03, + "learning_rate": 4.9008923156891624e-05, + "loss": 2.0488, + "step": 11940 + }, + { + "epoch": 1.03, + "learning_rate": 4.900805683097981e-05, + "loss": 2.0334, + "step": 11950 + }, + { + "epoch": 1.04, + "learning_rate": 4.9007190505068006e-05, + "loss": 2.0411, + "step": 11960 + }, + { + "epoch": 1.04, + "learning_rate": 4.90063241791562e-05, + "loss": 2.0756, + "step": 11970 + }, + { + "epoch": 1.04, + "learning_rate": 4.900545785324439e-05, + "loss": 2.0689, + "step": 11980 + }, + { + "epoch": 1.04, + "learning_rate": 4.900459152733258e-05, + "loss": 2.0961, + "step": 11990 + }, + { + "epoch": 1.04, + "learning_rate": 4.900372520142078e-05, + "loss": 2.0239, + "step": 12000 + }, + { + "epoch": 1.04, + "learning_rate": 4.9002858875508965e-05, + "loss": 2.0907, + "step": 12010 + }, + { + "epoch": 1.04, + "learning_rate": 4.900199254959716e-05, + "loss": 2.0109, + "step": 12020 + }, + { + "epoch": 1.04, + "learning_rate": 4.9001126223685354e-05, + "loss": 2.0641, + "step": 12030 + }, + { + "epoch": 1.04, + "learning_rate": 4.900025989777354e-05, + "loss": 2.1287, + "step": 12040 + }, + { + "epoch": 1.04, + "learning_rate": 4.8999393571861736e-05, + "loss": 2.083, + "step": 12050 + }, + { + "epoch": 1.04, + "learning_rate": 4.8998527245949924e-05, + "loss": 2.0787, + "step": 12060 + }, + { + "epoch": 1.04, + "learning_rate": 4.899766092003812e-05, + "loss": 2.0221, + "step": 12070 + }, + { + "epoch": 1.05, + "learning_rate": 4.899679459412631e-05, + "loss": 2.1449, + "step": 12080 + }, + { + "epoch": 1.05, + "learning_rate": 4.89959282682145e-05, + "loss": 2.0498, + "step": 12090 + }, + { + "epoch": 1.05, + "learning_rate": 4.8995061942302695e-05, + "loss": 2.0303, + "step": 12100 + }, + { + "epoch": 1.05, + "learning_rate": 4.899419561639089e-05, + "loss": 2.0155, + "step": 12110 + }, + { + "epoch": 1.05, + "learning_rate": 4.899332929047908e-05, + "loss": 2.08, + "step": 12120 + }, + { + "epoch": 1.05, + "learning_rate": 4.899246296456727e-05, + "loss": 2.0774, + "step": 12130 + }, + { + "epoch": 1.05, + "learning_rate": 4.8991596638655466e-05, + "loss": 2.103, + "step": 12140 + }, + { + "epoch": 1.05, + "learning_rate": 4.899073031274365e-05, + "loss": 2.0297, + "step": 12150 + }, + { + "epoch": 1.05, + "learning_rate": 4.898986398683185e-05, + "loss": 2.0679, + "step": 12160 + }, + { + "epoch": 1.05, + "learning_rate": 4.8988997660920035e-05, + "loss": 2.0797, + "step": 12170 + }, + { + "epoch": 1.05, + "learning_rate": 4.898813133500823e-05, + "loss": 2.0835, + "step": 12180 + }, + { + "epoch": 1.06, + "learning_rate": 4.8987265009096424e-05, + "loss": 2.0932, + "step": 12190 + }, + { + "epoch": 1.06, + "learning_rate": 4.898639868318461e-05, + "loss": 2.0083, + "step": 12200 + }, + { + "epoch": 1.06, + "learning_rate": 4.8985532357272806e-05, + "loss": 2.1096, + "step": 12210 + }, + { + "epoch": 1.06, + "learning_rate": 4.8984666031361e-05, + "loss": 2.1036, + "step": 12220 + }, + { + "epoch": 1.06, + "learning_rate": 4.898379970544919e-05, + "loss": 2.0251, + "step": 12230 + }, + { + "epoch": 1.06, + "learning_rate": 4.898293337953738e-05, + "loss": 2.0474, + "step": 12240 + }, + { + "epoch": 1.06, + "learning_rate": 4.898206705362558e-05, + "loss": 2.0316, + "step": 12250 + }, + { + "epoch": 1.06, + "learning_rate": 4.8981200727713765e-05, + "loss": 2.0998, + "step": 12260 + }, + { + "epoch": 1.06, + "learning_rate": 4.898033440180196e-05, + "loss": 2.0901, + "step": 12270 + }, + { + "epoch": 1.06, + "learning_rate": 4.897946807589015e-05, + "loss": 2.09, + "step": 12280 + }, + { + "epoch": 1.06, + "learning_rate": 4.897860174997834e-05, + "loss": 2.0338, + "step": 12290 + }, + { + "epoch": 1.06, + "learning_rate": 4.8977735424066536e-05, + "loss": 2.0974, + "step": 12300 + }, + { + "epoch": 1.07, + "learning_rate": 4.8976869098154724e-05, + "loss": 2.0565, + "step": 12310 + }, + { + "epoch": 1.07, + "learning_rate": 4.897600277224292e-05, + "loss": 2.0608, + "step": 12320 + }, + { + "epoch": 1.07, + "learning_rate": 4.897513644633111e-05, + "loss": 2.0201, + "step": 12330 + }, + { + "epoch": 1.07, + "learning_rate": 4.89742701204193e-05, + "loss": 2.0957, + "step": 12340 + }, + { + "epoch": 1.07, + "learning_rate": 4.8973403794507495e-05, + "loss": 2.0823, + "step": 12350 + }, + { + "epoch": 1.07, + "learning_rate": 4.897253746859569e-05, + "loss": 2.0745, + "step": 12360 + }, + { + "epoch": 1.07, + "learning_rate": 4.897167114268388e-05, + "loss": 2.0391, + "step": 12370 + }, + { + "epoch": 1.07, + "learning_rate": 4.897080481677207e-05, + "loss": 2.0375, + "step": 12380 + }, + { + "epoch": 1.07, + "learning_rate": 4.896993849086026e-05, + "loss": 2.0255, + "step": 12390 + }, + { + "epoch": 1.07, + "learning_rate": 4.8969072164948454e-05, + "loss": 1.9923, + "step": 12400 + }, + { + "epoch": 1.07, + "learning_rate": 4.896820583903665e-05, + "loss": 2.0161, + "step": 12410 + }, + { + "epoch": 1.07, + "learning_rate": 4.8967339513124836e-05, + "loss": 2.0296, + "step": 12420 + }, + { + "epoch": 1.08, + "learning_rate": 4.896647318721303e-05, + "loss": 2.0512, + "step": 12430 + }, + { + "epoch": 1.08, + "learning_rate": 4.8965606861301225e-05, + "loss": 2.0825, + "step": 12440 + }, + { + "epoch": 1.08, + "learning_rate": 4.896474053538941e-05, + "loss": 1.9812, + "step": 12450 + }, + { + "epoch": 1.08, + "learning_rate": 4.896387420947761e-05, + "loss": 2.0344, + "step": 12460 + }, + { + "epoch": 1.08, + "learning_rate": 4.89630078835658e-05, + "loss": 2.0459, + "step": 12470 + }, + { + "epoch": 1.08, + "learning_rate": 4.896214155765399e-05, + "loss": 2.0454, + "step": 12480 + }, + { + "epoch": 1.08, + "learning_rate": 4.8961275231742183e-05, + "loss": 2.095, + "step": 12490 + }, + { + "epoch": 1.08, + "learning_rate": 4.896040890583037e-05, + "loss": 1.9971, + "step": 12500 + }, + { + "epoch": 1.08, + "learning_rate": 4.8959542579918566e-05, + "loss": 1.9793, + "step": 12510 + }, + { + "epoch": 1.08, + "learning_rate": 4.895867625400676e-05, + "loss": 2.074, + "step": 12520 + }, + { + "epoch": 1.08, + "learning_rate": 4.895780992809495e-05, + "loss": 2.0302, + "step": 12530 + }, + { + "epoch": 1.09, + "learning_rate": 4.895694360218314e-05, + "loss": 2.0651, + "step": 12540 + }, + { + "epoch": 1.09, + "learning_rate": 4.8956077276271337e-05, + "loss": 2.0587, + "step": 12550 + }, + { + "epoch": 1.09, + "learning_rate": 4.8955210950359524e-05, + "loss": 2.032, + "step": 12560 + }, + { + "epoch": 1.09, + "learning_rate": 4.895434462444772e-05, + "loss": 2.0652, + "step": 12570 + }, + { + "epoch": 1.09, + "learning_rate": 4.8953478298535906e-05, + "loss": 2.0371, + "step": 12580 + }, + { + "epoch": 1.09, + "learning_rate": 4.89526119726241e-05, + "loss": 2.0368, + "step": 12590 + }, + { + "epoch": 1.09, + "learning_rate": 4.8951745646712295e-05, + "loss": 2.0626, + "step": 12600 + }, + { + "epoch": 1.09, + "learning_rate": 4.895087932080048e-05, + "loss": 2.0167, + "step": 12610 + }, + { + "epoch": 1.09, + "learning_rate": 4.895001299488868e-05, + "loss": 2.0193, + "step": 12620 + }, + { + "epoch": 1.09, + "learning_rate": 4.894914666897687e-05, + "loss": 2.0085, + "step": 12630 + }, + { + "epoch": 1.09, + "learning_rate": 4.894828034306506e-05, + "loss": 2.0327, + "step": 12640 + }, + { + "epoch": 1.09, + "learning_rate": 4.8947414017153254e-05, + "loss": 2.0319, + "step": 12650 + }, + { + "epoch": 1.1, + "learning_rate": 4.894654769124145e-05, + "loss": 2.0512, + "step": 12660 + }, + { + "epoch": 1.1, + "learning_rate": 4.8945681365329636e-05, + "loss": 2.0432, + "step": 12670 + }, + { + "epoch": 1.1, + "learning_rate": 4.894481503941783e-05, + "loss": 2.0298, + "step": 12680 + }, + { + "epoch": 1.1, + "learning_rate": 4.894394871350602e-05, + "loss": 2.0422, + "step": 12690 + }, + { + "epoch": 1.1, + "learning_rate": 4.894308238759421e-05, + "loss": 2.0218, + "step": 12700 + }, + { + "epoch": 1.1, + "learning_rate": 4.894221606168241e-05, + "loss": 2.0756, + "step": 12710 + }, + { + "epoch": 1.1, + "learning_rate": 4.8941349735770595e-05, + "loss": 2.0413, + "step": 12720 + }, + { + "epoch": 1.1, + "learning_rate": 4.894048340985879e-05, + "loss": 2.0578, + "step": 12730 + }, + { + "epoch": 1.1, + "learning_rate": 4.8939617083946984e-05, + "loss": 2.1019, + "step": 12740 + }, + { + "epoch": 1.1, + "learning_rate": 4.893875075803517e-05, + "loss": 2.0659, + "step": 12750 + }, + { + "epoch": 1.1, + "learning_rate": 4.8937884432123366e-05, + "loss": 2.0172, + "step": 12760 + }, + { + "epoch": 1.11, + "learning_rate": 4.893701810621156e-05, + "loss": 2.0102, + "step": 12770 + }, + { + "epoch": 1.11, + "learning_rate": 4.893615178029975e-05, + "loss": 2.0439, + "step": 12780 + }, + { + "epoch": 1.11, + "learning_rate": 4.893528545438794e-05, + "loss": 2.0352, + "step": 12790 + }, + { + "epoch": 1.11, + "learning_rate": 4.893441912847613e-05, + "loss": 2.0197, + "step": 12800 + }, + { + "epoch": 1.11, + "learning_rate": 4.8933552802564325e-05, + "loss": 2.0541, + "step": 12810 + }, + { + "epoch": 1.11, + "learning_rate": 4.893268647665252e-05, + "loss": 2.0628, + "step": 12820 + }, + { + "epoch": 1.11, + "learning_rate": 4.893182015074071e-05, + "loss": 2.0166, + "step": 12830 + }, + { + "epoch": 1.11, + "learning_rate": 4.89309538248289e-05, + "loss": 2.0021, + "step": 12840 + }, + { + "epoch": 1.11, + "learning_rate": 4.8930087498917096e-05, + "loss": 2.0423, + "step": 12850 + }, + { + "epoch": 1.11, + "learning_rate": 4.892922117300528e-05, + "loss": 2.0251, + "step": 12860 + }, + { + "epoch": 1.11, + "learning_rate": 4.892835484709348e-05, + "loss": 2.0376, + "step": 12870 + }, + { + "epoch": 1.11, + "learning_rate": 4.892748852118167e-05, + "loss": 2.0485, + "step": 12880 + }, + { + "epoch": 1.12, + "learning_rate": 4.892662219526986e-05, + "loss": 1.9797, + "step": 12890 + }, + { + "epoch": 1.12, + "learning_rate": 4.8925755869358054e-05, + "loss": 2.0611, + "step": 12900 + }, + { + "epoch": 1.12, + "learning_rate": 4.892488954344624e-05, + "loss": 2.0984, + "step": 12910 + }, + { + "epoch": 1.12, + "learning_rate": 4.8924023217534436e-05, + "loss": 2.0133, + "step": 12920 + }, + { + "epoch": 1.12, + "learning_rate": 4.892315689162263e-05, + "loss": 1.977, + "step": 12930 + }, + { + "epoch": 1.12, + "learning_rate": 4.892229056571082e-05, + "loss": 2.0708, + "step": 12940 + }, + { + "epoch": 1.12, + "learning_rate": 4.892142423979901e-05, + "loss": 2.0804, + "step": 12950 + }, + { + "epoch": 1.12, + "learning_rate": 4.892055791388721e-05, + "loss": 2.0231, + "step": 12960 + }, + { + "epoch": 1.12, + "learning_rate": 4.8919691587975395e-05, + "loss": 2.0002, + "step": 12970 + }, + { + "epoch": 1.12, + "learning_rate": 4.891882526206359e-05, + "loss": 2.0253, + "step": 12980 + }, + { + "epoch": 1.12, + "learning_rate": 4.8917958936151784e-05, + "loss": 2.0856, + "step": 12990 + }, + { + "epoch": 1.13, + "learning_rate": 4.891709261023997e-05, + "loss": 2.0332, + "step": 13000 + }, + { + "epoch": 1.13, + "learning_rate": 4.8916226284328166e-05, + "loss": 2.0469, + "step": 13010 + }, + { + "epoch": 1.13, + "learning_rate": 4.8915359958416354e-05, + "loss": 2.0104, + "step": 13020 + }, + { + "epoch": 1.13, + "learning_rate": 4.891449363250455e-05, + "loss": 2.0612, + "step": 13030 + }, + { + "epoch": 1.13, + "learning_rate": 4.891362730659274e-05, + "loss": 2.0649, + "step": 13040 + }, + { + "epoch": 1.13, + "learning_rate": 4.891276098068093e-05, + "loss": 2.0286, + "step": 13050 + }, + { + "epoch": 1.13, + "learning_rate": 4.8911894654769125e-05, + "loss": 2.065, + "step": 13060 + }, + { + "epoch": 1.13, + "learning_rate": 4.891102832885732e-05, + "loss": 2.0558, + "step": 13070 + }, + { + "epoch": 1.13, + "learning_rate": 4.891016200294551e-05, + "loss": 2.0311, + "step": 13080 + }, + { + "epoch": 1.13, + "learning_rate": 4.89092956770337e-05, + "loss": 2.033, + "step": 13090 + }, + { + "epoch": 1.13, + "learning_rate": 4.8908429351121896e-05, + "loss": 2.046, + "step": 13100 + }, + { + "epoch": 1.13, + "learning_rate": 4.8907563025210084e-05, + "loss": 2.0672, + "step": 13110 + }, + { + "epoch": 1.14, + "learning_rate": 4.890669669929828e-05, + "loss": 2.0072, + "step": 13120 + }, + { + "epoch": 1.14, + "learning_rate": 4.8905830373386466e-05, + "loss": 2.0627, + "step": 13130 + }, + { + "epoch": 1.14, + "learning_rate": 4.890496404747466e-05, + "loss": 2.0579, + "step": 13140 + }, + { + "epoch": 1.14, + "learning_rate": 4.8904097721562855e-05, + "loss": 2.0269, + "step": 13150 + }, + { + "epoch": 1.14, + "learning_rate": 4.890323139565104e-05, + "loss": 2.0282, + "step": 13160 + }, + { + "epoch": 1.14, + "learning_rate": 4.890236506973924e-05, + "loss": 2.0574, + "step": 13170 + }, + { + "epoch": 1.14, + "learning_rate": 4.890149874382743e-05, + "loss": 2.0225, + "step": 13180 + }, + { + "epoch": 1.14, + "learning_rate": 4.890063241791562e-05, + "loss": 2.036, + "step": 13190 + }, + { + "epoch": 1.14, + "learning_rate": 4.889976609200381e-05, + "loss": 2.0767, + "step": 13200 + }, + { + "epoch": 1.14, + "learning_rate": 4.889889976609201e-05, + "loss": 2.0611, + "step": 13210 + }, + { + "epoch": 1.14, + "learning_rate": 4.8898033440180195e-05, + "loss": 2.0374, + "step": 13220 + }, + { + "epoch": 1.15, + "learning_rate": 4.889716711426839e-05, + "loss": 2.0292, + "step": 13230 + }, + { + "epoch": 1.15, + "learning_rate": 4.889630078835658e-05, + "loss": 2.0141, + "step": 13240 + }, + { + "epoch": 1.15, + "learning_rate": 4.889543446244477e-05, + "loss": 2.055, + "step": 13250 + }, + { + "epoch": 1.15, + "learning_rate": 4.8894568136532967e-05, + "loss": 2.0202, + "step": 13260 + }, + { + "epoch": 1.15, + "learning_rate": 4.8893701810621154e-05, + "loss": 2.0807, + "step": 13270 + }, + { + "epoch": 1.15, + "learning_rate": 4.889283548470935e-05, + "loss": 2.008, + "step": 13280 + }, + { + "epoch": 1.15, + "learning_rate": 4.889196915879754e-05, + "loss": 2.0303, + "step": 13290 + }, + { + "epoch": 1.15, + "learning_rate": 4.889110283288573e-05, + "loss": 2.0585, + "step": 13300 + }, + { + "epoch": 1.15, + "learning_rate": 4.8890236506973925e-05, + "loss": 2.0939, + "step": 13310 + }, + { + "epoch": 1.15, + "learning_rate": 4.888937018106211e-05, + "loss": 2.0771, + "step": 13320 + }, + { + "epoch": 1.15, + "learning_rate": 4.888850385515031e-05, + "loss": 2.0234, + "step": 13330 + }, + { + "epoch": 1.15, + "learning_rate": 4.88876375292385e-05, + "loss": 2.0042, + "step": 13340 + }, + { + "epoch": 1.16, + "learning_rate": 4.888677120332669e-05, + "loss": 1.9995, + "step": 13350 + }, + { + "epoch": 1.16, + "learning_rate": 4.8885904877414884e-05, + "loss": 1.9961, + "step": 13360 + }, + { + "epoch": 1.16, + "learning_rate": 4.888503855150308e-05, + "loss": 1.9957, + "step": 13370 + }, + { + "epoch": 1.16, + "learning_rate": 4.8884172225591266e-05, + "loss": 2.0394, + "step": 13380 + }, + { + "epoch": 1.16, + "learning_rate": 4.888330589967946e-05, + "loss": 1.9716, + "step": 13390 + }, + { + "epoch": 1.16, + "learning_rate": 4.8882439573767655e-05, + "loss": 2.0219, + "step": 13400 + }, + { + "epoch": 1.16, + "learning_rate": 4.888157324785584e-05, + "loss": 2.0428, + "step": 13410 + }, + { + "epoch": 1.16, + "learning_rate": 4.888070692194404e-05, + "loss": 2.0336, + "step": 13420 + }, + { + "epoch": 1.16, + "learning_rate": 4.8879840596032225e-05, + "loss": 2.0286, + "step": 13430 + }, + { + "epoch": 1.16, + "learning_rate": 4.887897427012042e-05, + "loss": 1.9868, + "step": 13440 + }, + { + "epoch": 1.16, + "learning_rate": 4.8878107944208614e-05, + "loss": 2.014, + "step": 13450 + }, + { + "epoch": 1.17, + "learning_rate": 4.88772416182968e-05, + "loss": 2.029, + "step": 13460 + }, + { + "epoch": 1.17, + "learning_rate": 4.8876375292384996e-05, + "loss": 2.0052, + "step": 13470 + }, + { + "epoch": 1.17, + "learning_rate": 4.887550896647319e-05, + "loss": 2.008, + "step": 13480 + }, + { + "epoch": 1.17, + "learning_rate": 4.887464264056138e-05, + "loss": 2.0454, + "step": 13490 + }, + { + "epoch": 1.17, + "learning_rate": 4.887377631464957e-05, + "loss": 2.04, + "step": 13500 + }, + { + "epoch": 1.17, + "learning_rate": 4.887290998873777e-05, + "loss": 2.0621, + "step": 13510 + }, + { + "epoch": 1.17, + "learning_rate": 4.8872043662825955e-05, + "loss": 2.0656, + "step": 13520 + }, + { + "epoch": 1.17, + "learning_rate": 4.887117733691415e-05, + "loss": 2.0518, + "step": 13530 + }, + { + "epoch": 1.17, + "learning_rate": 4.887031101100234e-05, + "loss": 2.0049, + "step": 13540 + }, + { + "epoch": 1.17, + "learning_rate": 4.886944468509053e-05, + "loss": 2.0564, + "step": 13550 + }, + { + "epoch": 1.17, + "learning_rate": 4.8868578359178726e-05, + "loss": 2.0133, + "step": 13560 + }, + { + "epoch": 1.17, + "learning_rate": 4.886771203326691e-05, + "loss": 1.9917, + "step": 13570 + }, + { + "epoch": 1.18, + "learning_rate": 4.886684570735511e-05, + "loss": 2.055, + "step": 13580 + }, + { + "epoch": 1.18, + "learning_rate": 4.88659793814433e-05, + "loss": 2.0066, + "step": 13590 + }, + { + "epoch": 1.18, + "learning_rate": 4.886511305553149e-05, + "loss": 2.063, + "step": 13600 + }, + { + "epoch": 1.18, + "learning_rate": 4.8864246729619684e-05, + "loss": 2.0215, + "step": 13610 + }, + { + "epoch": 1.18, + "learning_rate": 4.886338040370788e-05, + "loss": 2.1001, + "step": 13620 + }, + { + "epoch": 1.18, + "learning_rate": 4.8862514077796066e-05, + "loss": 2.0174, + "step": 13630 + }, + { + "epoch": 1.18, + "learning_rate": 4.886164775188426e-05, + "loss": 2.034, + "step": 13640 + }, + { + "epoch": 1.18, + "learning_rate": 4.886078142597245e-05, + "loss": 2.0651, + "step": 13650 + }, + { + "epoch": 1.18, + "learning_rate": 4.885991510006064e-05, + "loss": 2.0302, + "step": 13660 + }, + { + "epoch": 1.18, + "learning_rate": 4.885904877414884e-05, + "loss": 2.0391, + "step": 13670 + }, + { + "epoch": 1.18, + "learning_rate": 4.8858182448237025e-05, + "loss": 2.0004, + "step": 13680 + }, + { + "epoch": 1.18, + "learning_rate": 4.885731612232522e-05, + "loss": 2.028, + "step": 13690 + }, + { + "epoch": 1.19, + "learning_rate": 4.8856449796413414e-05, + "loss": 2.0642, + "step": 13700 + }, + { + "epoch": 1.19, + "learning_rate": 4.88555834705016e-05, + "loss": 2.0329, + "step": 13710 + }, + { + "epoch": 1.19, + "learning_rate": 4.8854717144589796e-05, + "loss": 2.1365, + "step": 13720 + }, + { + "epoch": 1.19, + "learning_rate": 4.885385081867799e-05, + "loss": 2.0151, + "step": 13730 + }, + { + "epoch": 1.19, + "learning_rate": 4.885298449276618e-05, + "loss": 2.0448, + "step": 13740 + }, + { + "epoch": 1.19, + "learning_rate": 4.885211816685437e-05, + "loss": 2.0655, + "step": 13750 + }, + { + "epoch": 1.19, + "learning_rate": 4.885125184094256e-05, + "loss": 2.0685, + "step": 13760 + }, + { + "epoch": 1.19, + "learning_rate": 4.8850385515030755e-05, + "loss": 2.0295, + "step": 13770 + }, + { + "epoch": 1.19, + "learning_rate": 4.884951918911895e-05, + "loss": 1.9972, + "step": 13780 + }, + { + "epoch": 1.19, + "learning_rate": 4.884865286320714e-05, + "loss": 2.0466, + "step": 13790 + }, + { + "epoch": 1.19, + "learning_rate": 4.884778653729533e-05, + "loss": 2.0341, + "step": 13800 + }, + { + "epoch": 1.2, + "learning_rate": 4.8846920211383526e-05, + "loss": 2.0308, + "step": 13810 + }, + { + "epoch": 1.2, + "learning_rate": 4.8846053885471714e-05, + "loss": 2.0596, + "step": 13820 + }, + { + "epoch": 1.2, + "learning_rate": 4.884518755955991e-05, + "loss": 2.0851, + "step": 13830 + }, + { + "epoch": 1.2, + "learning_rate": 4.88443212336481e-05, + "loss": 2.0544, + "step": 13840 + }, + { + "epoch": 1.2, + "learning_rate": 4.884345490773629e-05, + "loss": 2.06, + "step": 13850 + }, + { + "epoch": 1.2, + "learning_rate": 4.8842588581824485e-05, + "loss": 2.0792, + "step": 13860 + }, + { + "epoch": 1.2, + "learning_rate": 4.884172225591267e-05, + "loss": 2.0431, + "step": 13870 + }, + { + "epoch": 1.2, + "learning_rate": 4.884085593000087e-05, + "loss": 2.0524, + "step": 13880 + }, + { + "epoch": 1.2, + "learning_rate": 4.883998960408906e-05, + "loss": 1.9996, + "step": 13890 + }, + { + "epoch": 1.2, + "learning_rate": 4.883912327817725e-05, + "loss": 2.0394, + "step": 13900 + }, + { + "epoch": 1.2, + "learning_rate": 4.883825695226544e-05, + "loss": 2.0051, + "step": 13910 + }, + { + "epoch": 1.2, + "learning_rate": 4.883739062635364e-05, + "loss": 2.0759, + "step": 13920 + }, + { + "epoch": 1.21, + "learning_rate": 4.8836524300441825e-05, + "loss": 1.9633, + "step": 13930 + }, + { + "epoch": 1.21, + "learning_rate": 4.883565797453002e-05, + "loss": 2.0236, + "step": 13940 + }, + { + "epoch": 1.21, + "learning_rate": 4.8834791648618214e-05, + "loss": 1.9494, + "step": 13950 + }, + { + "epoch": 1.21, + "learning_rate": 4.88339253227064e-05, + "loss": 2.0077, + "step": 13960 + }, + { + "epoch": 1.21, + "learning_rate": 4.8833058996794597e-05, + "loss": 1.9941, + "step": 13970 + }, + { + "epoch": 1.21, + "learning_rate": 4.8832192670882784e-05, + "loss": 2.0692, + "step": 13980 + }, + { + "epoch": 1.21, + "learning_rate": 4.883132634497098e-05, + "loss": 1.9869, + "step": 13990 + }, + { + "epoch": 1.21, + "learning_rate": 4.883046001905917e-05, + "loss": 2.0266, + "step": 14000 + }, + { + "epoch": 1.21, + "learning_rate": 4.882959369314736e-05, + "loss": 2.0181, + "step": 14010 + }, + { + "epoch": 1.21, + "learning_rate": 4.8828727367235555e-05, + "loss": 2.0119, + "step": 14020 + }, + { + "epoch": 1.21, + "learning_rate": 4.882786104132375e-05, + "loss": 2.0349, + "step": 14030 + }, + { + "epoch": 1.22, + "learning_rate": 4.882699471541194e-05, + "loss": 1.9728, + "step": 14040 + }, + { + "epoch": 1.22, + "learning_rate": 4.882612838950013e-05, + "loss": 2.0251, + "step": 14050 + }, + { + "epoch": 1.22, + "learning_rate": 4.8825262063588326e-05, + "loss": 2.0319, + "step": 14060 + }, + { + "epoch": 1.22, + "learning_rate": 4.8824395737676514e-05, + "loss": 2.0349, + "step": 14070 + }, + { + "epoch": 1.22, + "learning_rate": 4.882352941176471e-05, + "loss": 1.9857, + "step": 14080 + }, + { + "epoch": 1.22, + "learning_rate": 4.8822663085852896e-05, + "loss": 1.9817, + "step": 14090 + }, + { + "epoch": 1.22, + "learning_rate": 4.882179675994109e-05, + "loss": 2.0684, + "step": 14100 + }, + { + "epoch": 1.22, + "learning_rate": 4.8820930434029285e-05, + "loss": 1.979, + "step": 14110 + }, + { + "epoch": 1.22, + "learning_rate": 4.882006410811747e-05, + "loss": 1.9511, + "step": 14120 + }, + { + "epoch": 1.22, + "learning_rate": 4.881919778220567e-05, + "loss": 2.0477, + "step": 14130 + }, + { + "epoch": 1.22, + "learning_rate": 4.881833145629386e-05, + "loss": 2.0493, + "step": 14140 + }, + { + "epoch": 1.22, + "learning_rate": 4.881746513038205e-05, + "loss": 2.0246, + "step": 14150 + }, + { + "epoch": 1.23, + "learning_rate": 4.8816598804470244e-05, + "loss": 1.9658, + "step": 14160 + }, + { + "epoch": 1.23, + "learning_rate": 4.881573247855843e-05, + "loss": 2.0072, + "step": 14170 + }, + { + "epoch": 1.23, + "learning_rate": 4.8814866152646626e-05, + "loss": 1.9807, + "step": 14180 + }, + { + "epoch": 1.23, + "learning_rate": 4.881399982673482e-05, + "loss": 2.0599, + "step": 14190 + }, + { + "epoch": 1.23, + "learning_rate": 4.881313350082301e-05, + "loss": 1.9986, + "step": 14200 + }, + { + "epoch": 1.23, + "learning_rate": 4.88122671749112e-05, + "loss": 2.0428, + "step": 14210 + }, + { + "epoch": 1.23, + "learning_rate": 4.88114008489994e-05, + "loss": 2.0459, + "step": 14220 + }, + { + "epoch": 1.23, + "learning_rate": 4.8810534523087585e-05, + "loss": 1.9918, + "step": 14230 + }, + { + "epoch": 1.23, + "learning_rate": 4.880966819717578e-05, + "loss": 1.9943, + "step": 14240 + }, + { + "epoch": 1.23, + "learning_rate": 4.8808801871263973e-05, + "loss": 2.0334, + "step": 14250 + }, + { + "epoch": 1.23, + "learning_rate": 4.880793554535216e-05, + "loss": 2.0127, + "step": 14260 + }, + { + "epoch": 1.24, + "learning_rate": 4.8807069219440356e-05, + "loss": 2.0474, + "step": 14270 + }, + { + "epoch": 1.24, + "learning_rate": 4.880620289352854e-05, + "loss": 2.0251, + "step": 14280 + }, + { + "epoch": 1.24, + "learning_rate": 4.880533656761674e-05, + "loss": 2.0195, + "step": 14290 + }, + { + "epoch": 1.24, + "learning_rate": 4.880447024170493e-05, + "loss": 1.982, + "step": 14300 + }, + { + "epoch": 1.24, + "learning_rate": 4.880360391579312e-05, + "loss": 2.0282, + "step": 14310 + }, + { + "epoch": 1.24, + "learning_rate": 4.8802737589881314e-05, + "loss": 1.9872, + "step": 14320 + }, + { + "epoch": 1.24, + "learning_rate": 4.880187126396951e-05, + "loss": 1.984, + "step": 14330 + }, + { + "epoch": 1.24, + "learning_rate": 4.8801004938057696e-05, + "loss": 2.0676, + "step": 14340 + }, + { + "epoch": 1.24, + "learning_rate": 4.880013861214589e-05, + "loss": 2.0237, + "step": 14350 + }, + { + "epoch": 1.24, + "learning_rate": 4.8799272286234085e-05, + "loss": 1.991, + "step": 14360 + }, + { + "epoch": 1.24, + "learning_rate": 4.879840596032227e-05, + "loss": 2.0141, + "step": 14370 + }, + { + "epoch": 1.24, + "learning_rate": 4.879753963441047e-05, + "loss": 1.9932, + "step": 14380 + }, + { + "epoch": 1.25, + "learning_rate": 4.8796673308498655e-05, + "loss": 2.0358, + "step": 14390 + }, + { + "epoch": 1.25, + "learning_rate": 4.879580698258685e-05, + "loss": 2.092, + "step": 14400 + }, + { + "epoch": 1.25, + "learning_rate": 4.8794940656675044e-05, + "loss": 1.9713, + "step": 14410 + }, + { + "epoch": 1.25, + "learning_rate": 4.879407433076323e-05, + "loss": 2.0419, + "step": 14420 + }, + { + "epoch": 1.25, + "learning_rate": 4.8793208004851426e-05, + "loss": 2.0145, + "step": 14430 + }, + { + "epoch": 1.25, + "learning_rate": 4.879234167893962e-05, + "loss": 2.0113, + "step": 14440 + }, + { + "epoch": 1.25, + "learning_rate": 4.879147535302781e-05, + "loss": 2.035, + "step": 14450 + }, + { + "epoch": 1.25, + "learning_rate": 4.8790609027116e-05, + "loss": 2.035, + "step": 14460 + }, + { + "epoch": 1.25, + "learning_rate": 4.87897427012042e-05, + "loss": 2.0033, + "step": 14470 + }, + { + "epoch": 1.25, + "learning_rate": 4.8788876375292385e-05, + "loss": 2.1178, + "step": 14480 + }, + { + "epoch": 1.25, + "learning_rate": 4.878801004938058e-05, + "loss": 1.9992, + "step": 14490 + }, + { + "epoch": 1.26, + "learning_rate": 4.878714372346877e-05, + "loss": 1.9979, + "step": 14500 + }, + { + "epoch": 1.26, + "learning_rate": 4.878627739755696e-05, + "loss": 2.0372, + "step": 14510 + }, + { + "epoch": 1.26, + "learning_rate": 4.8785411071645156e-05, + "loss": 2.0492, + "step": 14520 + }, + { + "epoch": 1.26, + "learning_rate": 4.8784544745733344e-05, + "loss": 1.9342, + "step": 14530 + }, + { + "epoch": 1.26, + "learning_rate": 4.878367841982154e-05, + "loss": 1.9957, + "step": 14540 + }, + { + "epoch": 1.26, + "learning_rate": 4.878281209390973e-05, + "loss": 2.0088, + "step": 14550 + }, + { + "epoch": 1.26, + "learning_rate": 4.878194576799792e-05, + "loss": 2.0211, + "step": 14560 + }, + { + "epoch": 1.26, + "learning_rate": 4.8781079442086115e-05, + "loss": 2.0255, + "step": 14570 + }, + { + "epoch": 1.26, + "learning_rate": 4.878021311617431e-05, + "loss": 1.9808, + "step": 14580 + }, + { + "epoch": 1.26, + "learning_rate": 4.87793467902625e-05, + "loss": 1.9624, + "step": 14590 + }, + { + "epoch": 1.26, + "learning_rate": 4.877848046435069e-05, + "loss": 2.0375, + "step": 14600 + }, + { + "epoch": 1.26, + "learning_rate": 4.877761413843888e-05, + "loss": 2.0243, + "step": 14610 + }, + { + "epoch": 1.27, + "learning_rate": 4.877674781252707e-05, + "loss": 2.06, + "step": 14620 + }, + { + "epoch": 1.27, + "learning_rate": 4.877588148661527e-05, + "loss": 2.0062, + "step": 14630 + }, + { + "epoch": 1.27, + "learning_rate": 4.8775015160703455e-05, + "loss": 1.9915, + "step": 14640 + }, + { + "epoch": 1.27, + "learning_rate": 4.877414883479165e-05, + "loss": 2.0586, + "step": 14650 + }, + { + "epoch": 1.27, + "learning_rate": 4.8773282508879844e-05, + "loss": 2.0504, + "step": 14660 + }, + { + "epoch": 1.27, + "learning_rate": 4.877241618296803e-05, + "loss": 1.9886, + "step": 14670 + }, + { + "epoch": 1.27, + "learning_rate": 4.8771549857056226e-05, + "loss": 1.9997, + "step": 14680 + }, + { + "epoch": 1.27, + "learning_rate": 4.877068353114442e-05, + "loss": 2.0427, + "step": 14690 + }, + { + "epoch": 1.27, + "learning_rate": 4.876981720523261e-05, + "loss": 1.9859, + "step": 14700 + }, + { + "epoch": 1.27, + "learning_rate": 4.87689508793208e-05, + "loss": 1.9956, + "step": 14710 + }, + { + "epoch": 1.27, + "learning_rate": 4.876808455340899e-05, + "loss": 2.0068, + "step": 14720 + }, + { + "epoch": 1.27, + "learning_rate": 4.8767218227497185e-05, + "loss": 2.0298, + "step": 14730 + }, + { + "epoch": 1.28, + "learning_rate": 4.876635190158538e-05, + "loss": 2.0176, + "step": 14740 + }, + { + "epoch": 1.28, + "learning_rate": 4.876548557567357e-05, + "loss": 2.0242, + "step": 14750 + }, + { + "epoch": 1.28, + "learning_rate": 4.876461924976176e-05, + "loss": 1.9967, + "step": 14760 + }, + { + "epoch": 1.28, + "learning_rate": 4.8763752923849956e-05, + "loss": 2.017, + "step": 14770 + }, + { + "epoch": 1.28, + "learning_rate": 4.8762886597938144e-05, + "loss": 1.9884, + "step": 14780 + }, + { + "epoch": 1.28, + "learning_rate": 4.876202027202634e-05, + "loss": 2.0258, + "step": 14790 + }, + { + "epoch": 1.28, + "learning_rate": 4.876115394611453e-05, + "loss": 2.0481, + "step": 14800 + }, + { + "epoch": 1.28, + "learning_rate": 4.876028762020272e-05, + "loss": 2.0222, + "step": 14810 + }, + { + "epoch": 1.28, + "learning_rate": 4.8759421294290915e-05, + "loss": 2.0506, + "step": 14820 + }, + { + "epoch": 1.28, + "learning_rate": 4.87585549683791e-05, + "loss": 2.0306, + "step": 14830 + }, + { + "epoch": 1.28, + "learning_rate": 4.87576886424673e-05, + "loss": 2.0352, + "step": 14840 + }, + { + "epoch": 1.29, + "learning_rate": 4.875682231655549e-05, + "loss": 2.0432, + "step": 14850 + }, + { + "epoch": 1.29, + "learning_rate": 4.875595599064368e-05, + "loss": 1.9976, + "step": 14860 + }, + { + "epoch": 1.29, + "learning_rate": 4.8755089664731874e-05, + "loss": 1.9899, + "step": 14870 + }, + { + "epoch": 1.29, + "learning_rate": 4.875422333882007e-05, + "loss": 2.0808, + "step": 14880 + }, + { + "epoch": 1.29, + "learning_rate": 4.8753357012908256e-05, + "loss": 2.0804, + "step": 14890 + }, + { + "epoch": 1.29, + "learning_rate": 4.875249068699645e-05, + "loss": 2.0079, + "step": 14900 + }, + { + "epoch": 1.29, + "learning_rate": 4.875162436108464e-05, + "loss": 2.0158, + "step": 14910 + }, + { + "epoch": 1.29, + "learning_rate": 4.875075803517283e-05, + "loss": 1.9506, + "step": 14920 + }, + { + "epoch": 1.29, + "learning_rate": 4.874989170926103e-05, + "loss": 2.0209, + "step": 14930 + }, + { + "epoch": 1.29, + "learning_rate": 4.8749025383349214e-05, + "loss": 2.003, + "step": 14940 + }, + { + "epoch": 1.29, + "learning_rate": 4.874815905743741e-05, + "loss": 2.036, + "step": 14950 + }, + { + "epoch": 1.29, + "learning_rate": 4.8747292731525603e-05, + "loss": 1.9894, + "step": 14960 + }, + { + "epoch": 1.3, + "learning_rate": 4.874642640561379e-05, + "loss": 1.9669, + "step": 14970 + }, + { + "epoch": 1.3, + "learning_rate": 4.8745560079701986e-05, + "loss": 2.0018, + "step": 14980 + }, + { + "epoch": 1.3, + "learning_rate": 4.874469375379018e-05, + "loss": 2.035, + "step": 14990 + }, + { + "epoch": 1.3, + "learning_rate": 4.874382742787837e-05, + "loss": 2.024, + "step": 15000 + }, + { + "epoch": 1.3, + "learning_rate": 4.874296110196656e-05, + "loss": 2.0145, + "step": 15010 + }, + { + "epoch": 1.3, + "learning_rate": 4.874209477605475e-05, + "loss": 2.068, + "step": 15020 + }, + { + "epoch": 1.3, + "learning_rate": 4.8741228450142944e-05, + "loss": 2.0391, + "step": 15030 + }, + { + "epoch": 1.3, + "learning_rate": 4.874036212423114e-05, + "loss": 1.9781, + "step": 15040 + }, + { + "epoch": 1.3, + "learning_rate": 4.8739495798319326e-05, + "loss": 1.9973, + "step": 15050 + }, + { + "epoch": 1.3, + "learning_rate": 4.873862947240752e-05, + "loss": 2.0454, + "step": 15060 + }, + { + "epoch": 1.3, + "learning_rate": 4.8737763146495715e-05, + "loss": 2.0513, + "step": 15070 + }, + { + "epoch": 1.31, + "learning_rate": 4.87368968205839e-05, + "loss": 1.9976, + "step": 15080 + }, + { + "epoch": 1.31, + "learning_rate": 4.87360304946721e-05, + "loss": 2.0513, + "step": 15090 + }, + { + "epoch": 1.31, + "learning_rate": 4.873516416876029e-05, + "loss": 1.9863, + "step": 15100 + }, + { + "epoch": 1.31, + "learning_rate": 4.873429784284848e-05, + "loss": 1.977, + "step": 15110 + }, + { + "epoch": 1.31, + "learning_rate": 4.8733431516936674e-05, + "loss": 2.0324, + "step": 15120 + }, + { + "epoch": 1.31, + "learning_rate": 4.873256519102486e-05, + "loss": 1.9707, + "step": 15130 + }, + { + "epoch": 1.31, + "learning_rate": 4.8731698865113056e-05, + "loss": 2.0203, + "step": 15140 + }, + { + "epoch": 1.31, + "learning_rate": 4.873083253920125e-05, + "loss": 1.9756, + "step": 15150 + }, + { + "epoch": 1.31, + "learning_rate": 4.872996621328944e-05, + "loss": 1.9719, + "step": 15160 + }, + { + "epoch": 1.31, + "learning_rate": 4.872909988737763e-05, + "loss": 2.0198, + "step": 15170 + }, + { + "epoch": 1.31, + "learning_rate": 4.872823356146583e-05, + "loss": 1.9773, + "step": 15180 + }, + { + "epoch": 1.31, + "learning_rate": 4.8727367235554015e-05, + "loss": 1.9642, + "step": 15190 + }, + { + "epoch": 1.32, + "learning_rate": 4.872650090964221e-05, + "loss": 2.0206, + "step": 15200 + }, + { + "epoch": 1.32, + "learning_rate": 4.8725634583730404e-05, + "loss": 1.9629, + "step": 15210 + }, + { + "epoch": 1.32, + "learning_rate": 4.872476825781859e-05, + "loss": 1.9697, + "step": 15220 + }, + { + "epoch": 1.32, + "learning_rate": 4.8723901931906786e-05, + "loss": 2.0361, + "step": 15230 + }, + { + "epoch": 1.32, + "learning_rate": 4.8723035605994974e-05, + "loss": 2.0376, + "step": 15240 + }, + { + "epoch": 1.32, + "learning_rate": 4.872216928008317e-05, + "loss": 2.0331, + "step": 15250 + }, + { + "epoch": 1.32, + "learning_rate": 4.872130295417136e-05, + "loss": 2.0488, + "step": 15260 + }, + { + "epoch": 1.32, + "learning_rate": 4.872043662825955e-05, + "loss": 2.0117, + "step": 15270 + }, + { + "epoch": 1.32, + "learning_rate": 4.8719570302347745e-05, + "loss": 1.9777, + "step": 15280 + }, + { + "epoch": 1.32, + "learning_rate": 4.871870397643594e-05, + "loss": 1.9955, + "step": 15290 + }, + { + "epoch": 1.32, + "learning_rate": 4.871783765052413e-05, + "loss": 2.0158, + "step": 15300 + }, + { + "epoch": 1.33, + "learning_rate": 4.871697132461232e-05, + "loss": 2.0011, + "step": 15310 + }, + { + "epoch": 1.33, + "learning_rate": 4.8716104998700516e-05, + "loss": 2.0206, + "step": 15320 + }, + { + "epoch": 1.33, + "learning_rate": 4.87152386727887e-05, + "loss": 2.024, + "step": 15330 + }, + { + "epoch": 1.33, + "learning_rate": 4.87143723468769e-05, + "loss": 1.9884, + "step": 15340 + }, + { + "epoch": 1.33, + "learning_rate": 4.8713506020965085e-05, + "loss": 1.993, + "step": 15350 + }, + { + "epoch": 1.33, + "learning_rate": 4.871263969505328e-05, + "loss": 2.0493, + "step": 15360 + }, + { + "epoch": 1.33, + "learning_rate": 4.8711773369141474e-05, + "loss": 2.0066, + "step": 15370 + }, + { + "epoch": 1.33, + "learning_rate": 4.871090704322966e-05, + "loss": 2.0591, + "step": 15380 + }, + { + "epoch": 1.33, + "learning_rate": 4.8710040717317856e-05, + "loss": 2.0308, + "step": 15390 + }, + { + "epoch": 1.33, + "learning_rate": 4.870917439140605e-05, + "loss": 1.9888, + "step": 15400 + }, + { + "epoch": 1.33, + "learning_rate": 4.870830806549424e-05, + "loss": 2.0236, + "step": 15410 + }, + { + "epoch": 1.33, + "learning_rate": 4.870744173958243e-05, + "loss": 2.0184, + "step": 15420 + }, + { + "epoch": 1.34, + "learning_rate": 4.870657541367063e-05, + "loss": 2.0451, + "step": 15430 + }, + { + "epoch": 1.34, + "learning_rate": 4.8705709087758815e-05, + "loss": 2.0109, + "step": 15440 + }, + { + "epoch": 1.34, + "learning_rate": 4.870484276184701e-05, + "loss": 2.0011, + "step": 15450 + }, + { + "epoch": 1.34, + "learning_rate": 4.87039764359352e-05, + "loss": 2.0132, + "step": 15460 + }, + { + "epoch": 1.34, + "learning_rate": 4.870311011002339e-05, + "loss": 2.0356, + "step": 15470 + }, + { + "epoch": 1.34, + "learning_rate": 4.8702243784111586e-05, + "loss": 1.9902, + "step": 15480 + }, + { + "epoch": 1.34, + "learning_rate": 4.8701377458199774e-05, + "loss": 1.9209, + "step": 15490 + }, + { + "epoch": 1.34, + "learning_rate": 4.870051113228797e-05, + "loss": 2.0204, + "step": 15500 + }, + { + "epoch": 1.34, + "learning_rate": 4.869964480637616e-05, + "loss": 1.9732, + "step": 15510 + }, + { + "epoch": 1.34, + "learning_rate": 4.869877848046435e-05, + "loss": 2.0281, + "step": 15520 + }, + { + "epoch": 1.34, + "learning_rate": 4.8697912154552545e-05, + "loss": 2.0187, + "step": 15530 + }, + { + "epoch": 1.35, + "learning_rate": 4.869704582864074e-05, + "loss": 1.9911, + "step": 15540 + }, + { + "epoch": 1.35, + "learning_rate": 4.869617950272893e-05, + "loss": 2.0127, + "step": 15550 + }, + { + "epoch": 1.35, + "learning_rate": 4.869531317681712e-05, + "loss": 1.9891, + "step": 15560 + }, + { + "epoch": 1.35, + "learning_rate": 4.869444685090531e-05, + "loss": 1.9267, + "step": 15570 + }, + { + "epoch": 1.35, + "learning_rate": 4.8693580524993504e-05, + "loss": 1.9918, + "step": 15580 + }, + { + "epoch": 1.35, + "learning_rate": 4.86927141990817e-05, + "loss": 2.017, + "step": 15590 + }, + { + "epoch": 1.35, + "learning_rate": 4.8691847873169886e-05, + "loss": 2.0177, + "step": 15600 + }, + { + "epoch": 1.35, + "learning_rate": 4.869098154725808e-05, + "loss": 2.0073, + "step": 15610 + }, + { + "epoch": 1.35, + "learning_rate": 4.8690115221346275e-05, + "loss": 2.0298, + "step": 15620 + }, + { + "epoch": 1.35, + "learning_rate": 4.868924889543446e-05, + "loss": 1.9333, + "step": 15630 + }, + { + "epoch": 1.35, + "learning_rate": 4.868838256952266e-05, + "loss": 2.0084, + "step": 15640 + }, + { + "epoch": 1.35, + "learning_rate": 4.8687516243610844e-05, + "loss": 1.9834, + "step": 15650 + }, + { + "epoch": 1.36, + "learning_rate": 4.868664991769904e-05, + "loss": 1.9603, + "step": 15660 + }, + { + "epoch": 1.36, + "learning_rate": 4.868578359178723e-05, + "loss": 2.0199, + "step": 15670 + }, + { + "epoch": 1.36, + "learning_rate": 4.868491726587542e-05, + "loss": 2.0143, + "step": 15680 + }, + { + "epoch": 1.36, + "learning_rate": 4.8684050939963615e-05, + "loss": 1.9819, + "step": 15690 + }, + { + "epoch": 1.36, + "learning_rate": 4.868318461405181e-05, + "loss": 1.9989, + "step": 15700 + }, + { + "epoch": 1.36, + "learning_rate": 4.868231828814e-05, + "loss": 2.0353, + "step": 15710 + }, + { + "epoch": 1.36, + "learning_rate": 4.868145196222819e-05, + "loss": 1.9757, + "step": 15720 + }, + { + "epoch": 1.36, + "learning_rate": 4.8680585636316387e-05, + "loss": 2.0295, + "step": 15730 + }, + { + "epoch": 1.36, + "learning_rate": 4.8679719310404574e-05, + "loss": 1.9815, + "step": 15740 + }, + { + "epoch": 1.36, + "learning_rate": 4.867885298449277e-05, + "loss": 1.9877, + "step": 15750 + }, + { + "epoch": 1.36, + "learning_rate": 4.8677986658580956e-05, + "loss": 2.0352, + "step": 15760 + }, + { + "epoch": 1.36, + "learning_rate": 4.867712033266915e-05, + "loss": 2.0329, + "step": 15770 + }, + { + "epoch": 1.37, + "learning_rate": 4.8676254006757345e-05, + "loss": 2.0434, + "step": 15780 + }, + { + "epoch": 1.37, + "learning_rate": 4.867538768084553e-05, + "loss": 1.9952, + "step": 15790 + }, + { + "epoch": 1.37, + "learning_rate": 4.867452135493373e-05, + "loss": 2.0083, + "step": 15800 + }, + { + "epoch": 1.37, + "learning_rate": 4.867365502902192e-05, + "loss": 2.0209, + "step": 15810 + }, + { + "epoch": 1.37, + "learning_rate": 4.867278870311011e-05, + "loss": 2.0048, + "step": 15820 + }, + { + "epoch": 1.37, + "learning_rate": 4.8671922377198304e-05, + "loss": 2.0075, + "step": 15830 + }, + { + "epoch": 1.37, + "learning_rate": 4.86710560512865e-05, + "loss": 1.9657, + "step": 15840 + }, + { + "epoch": 1.37, + "learning_rate": 4.8670189725374686e-05, + "loss": 2.0008, + "step": 15850 + }, + { + "epoch": 1.37, + "learning_rate": 4.866932339946288e-05, + "loss": 2.015, + "step": 15860 + }, + { + "epoch": 1.37, + "learning_rate": 4.866845707355107e-05, + "loss": 2.0154, + "step": 15870 + }, + { + "epoch": 1.37, + "learning_rate": 4.866759074763926e-05, + "loss": 2.0279, + "step": 15880 + }, + { + "epoch": 1.38, + "learning_rate": 4.866672442172746e-05, + "loss": 1.9837, + "step": 15890 + }, + { + "epoch": 1.38, + "learning_rate": 4.8665858095815645e-05, + "loss": 2.0027, + "step": 15900 + }, + { + "epoch": 1.38, + "learning_rate": 4.866499176990384e-05, + "loss": 1.9427, + "step": 15910 + }, + { + "epoch": 1.38, + "learning_rate": 4.8664125443992034e-05, + "loss": 1.9674, + "step": 15920 + }, + { + "epoch": 1.38, + "learning_rate": 4.866325911808022e-05, + "loss": 1.9777, + "step": 15930 + }, + { + "epoch": 1.38, + "learning_rate": 4.8662392792168416e-05, + "loss": 1.9345, + "step": 15940 + }, + { + "epoch": 1.38, + "learning_rate": 4.866152646625661e-05, + "loss": 1.9734, + "step": 15950 + }, + { + "epoch": 1.38, + "learning_rate": 4.86606601403448e-05, + "loss": 2.0369, + "step": 15960 + }, + { + "epoch": 1.38, + "learning_rate": 4.865979381443299e-05, + "loss": 2.0092, + "step": 15970 + }, + { + "epoch": 1.38, + "learning_rate": 4.865892748852118e-05, + "loss": 1.9641, + "step": 15980 + }, + { + "epoch": 1.38, + "learning_rate": 4.8658061162609375e-05, + "loss": 1.9703, + "step": 15990 + }, + { + "epoch": 1.38, + "learning_rate": 4.865719483669757e-05, + "loss": 1.9985, + "step": 16000 + }, + { + "epoch": 1.39, + "learning_rate": 4.865632851078576e-05, + "loss": 1.9834, + "step": 16010 + }, + { + "epoch": 1.39, + "learning_rate": 4.865546218487395e-05, + "loss": 1.9837, + "step": 16020 + }, + { + "epoch": 1.39, + "learning_rate": 4.8654595858962146e-05, + "loss": 1.9874, + "step": 16030 + }, + { + "epoch": 1.39, + "learning_rate": 4.865372953305033e-05, + "loss": 2.0187, + "step": 16040 + }, + { + "epoch": 1.39, + "learning_rate": 4.865286320713853e-05, + "loss": 1.9938, + "step": 16050 + }, + { + "epoch": 1.39, + "learning_rate": 4.865199688122672e-05, + "loss": 1.9413, + "step": 16060 + }, + { + "epoch": 1.39, + "learning_rate": 4.865113055531491e-05, + "loss": 1.9598, + "step": 16070 + }, + { + "epoch": 1.39, + "learning_rate": 4.8650264229403104e-05, + "loss": 1.9763, + "step": 16080 + }, + { + "epoch": 1.39, + "learning_rate": 4.864939790349129e-05, + "loss": 2.0202, + "step": 16090 + }, + { + "epoch": 1.39, + "learning_rate": 4.8648531577579486e-05, + "loss": 1.9957, + "step": 16100 + }, + { + "epoch": 1.39, + "learning_rate": 4.864766525166768e-05, + "loss": 1.9933, + "step": 16110 + }, + { + "epoch": 1.4, + "learning_rate": 4.864679892575587e-05, + "loss": 1.9994, + "step": 16120 + }, + { + "epoch": 1.4, + "learning_rate": 4.864593259984406e-05, + "loss": 1.9797, + "step": 16130 + }, + { + "epoch": 1.4, + "learning_rate": 4.864506627393226e-05, + "loss": 1.9774, + "step": 16140 + }, + { + "epoch": 1.4, + "learning_rate": 4.8644199948020445e-05, + "loss": 1.9934, + "step": 16150 + }, + { + "epoch": 1.4, + "learning_rate": 4.864333362210864e-05, + "loss": 1.9292, + "step": 16160 + }, + { + "epoch": 1.4, + "learning_rate": 4.8642467296196834e-05, + "loss": 1.9824, + "step": 16170 + }, + { + "epoch": 1.4, + "learning_rate": 4.864160097028502e-05, + "loss": 2.0086, + "step": 16180 + }, + { + "epoch": 1.4, + "learning_rate": 4.8640734644373216e-05, + "loss": 1.9543, + "step": 16190 + }, + { + "epoch": 1.4, + "learning_rate": 4.8639868318461404e-05, + "loss": 2.0102, + "step": 16200 + }, + { + "epoch": 1.4, + "learning_rate": 4.86390019925496e-05, + "loss": 1.9002, + "step": 16210 + }, + { + "epoch": 1.4, + "learning_rate": 4.863813566663779e-05, + "loss": 2.0133, + "step": 16220 + }, + { + "epoch": 1.4, + "learning_rate": 4.863726934072598e-05, + "loss": 2.0579, + "step": 16230 + }, + { + "epoch": 1.41, + "learning_rate": 4.8636403014814175e-05, + "loss": 1.9374, + "step": 16240 + }, + { + "epoch": 1.41, + "learning_rate": 4.863553668890237e-05, + "loss": 1.991, + "step": 16250 + }, + { + "epoch": 1.41, + "learning_rate": 4.863467036299056e-05, + "loss": 2.0021, + "step": 16260 + }, + { + "epoch": 1.41, + "learning_rate": 4.863380403707875e-05, + "loss": 2.025, + "step": 16270 + }, + { + "epoch": 1.41, + "learning_rate": 4.8632937711166946e-05, + "loss": 2.018, + "step": 16280 + }, + { + "epoch": 1.41, + "learning_rate": 4.8632071385255134e-05, + "loss": 2.0153, + "step": 16290 + }, + { + "epoch": 1.41, + "learning_rate": 4.863120505934333e-05, + "loss": 2.1, + "step": 16300 + }, + { + "epoch": 1.41, + "learning_rate": 4.8630338733431516e-05, + "loss": 1.977, + "step": 16310 + }, + { + "epoch": 1.41, + "learning_rate": 4.862947240751971e-05, + "loss": 1.9338, + "step": 16320 + }, + { + "epoch": 1.41, + "learning_rate": 4.8628606081607905e-05, + "loss": 1.9833, + "step": 16330 + }, + { + "epoch": 1.41, + "learning_rate": 4.862773975569609e-05, + "loss": 1.9759, + "step": 16340 + }, + { + "epoch": 1.42, + "learning_rate": 4.862687342978429e-05, + "loss": 1.9863, + "step": 16350 + }, + { + "epoch": 1.42, + "learning_rate": 4.862600710387248e-05, + "loss": 2.0186, + "step": 16360 + }, + { + "epoch": 1.42, + "learning_rate": 4.862514077796067e-05, + "loss": 2.0137, + "step": 16370 + }, + { + "epoch": 1.42, + "learning_rate": 4.862427445204886e-05, + "loss": 1.9699, + "step": 16380 + }, + { + "epoch": 1.42, + "learning_rate": 4.862340812613706e-05, + "loss": 1.9819, + "step": 16390 + }, + { + "epoch": 1.42, + "learning_rate": 4.8622541800225245e-05, + "loss": 1.9546, + "step": 16400 + }, + { + "epoch": 1.42, + "learning_rate": 4.862167547431344e-05, + "loss": 1.9704, + "step": 16410 + }, + { + "epoch": 1.42, + "learning_rate": 4.862080914840163e-05, + "loss": 1.9778, + "step": 16420 + }, + { + "epoch": 1.42, + "learning_rate": 4.861994282248982e-05, + "loss": 1.9769, + "step": 16430 + }, + { + "epoch": 1.42, + "learning_rate": 4.8619076496578017e-05, + "loss": 1.9774, + "step": 16440 + }, + { + "epoch": 1.42, + "learning_rate": 4.8618210170666204e-05, + "loss": 2.0004, + "step": 16450 + }, + { + "epoch": 1.42, + "learning_rate": 4.86173438447544e-05, + "loss": 2.0392, + "step": 16460 + }, + { + "epoch": 1.43, + "learning_rate": 4.861647751884259e-05, + "loss": 2.0044, + "step": 16470 + }, + { + "epoch": 1.43, + "learning_rate": 4.861561119293078e-05, + "loss": 1.9718, + "step": 16480 + }, + { + "epoch": 1.43, + "learning_rate": 4.8614744867018975e-05, + "loss": 1.9657, + "step": 16490 + }, + { + "epoch": 1.43, + "learning_rate": 4.861387854110716e-05, + "loss": 1.9834, + "step": 16500 + }, + { + "epoch": 1.43, + "learning_rate": 4.861301221519536e-05, + "loss": 1.978, + "step": 16510 + }, + { + "epoch": 1.43, + "learning_rate": 4.861214588928355e-05, + "loss": 2.0191, + "step": 16520 + }, + { + "epoch": 1.43, + "learning_rate": 4.861127956337174e-05, + "loss": 1.9267, + "step": 16530 + }, + { + "epoch": 1.43, + "learning_rate": 4.8610413237459934e-05, + "loss": 1.9552, + "step": 16540 + }, + { + "epoch": 1.43, + "learning_rate": 4.860954691154813e-05, + "loss": 2.0027, + "step": 16550 + }, + { + "epoch": 1.43, + "learning_rate": 4.8608680585636316e-05, + "loss": 2.0028, + "step": 16560 + }, + { + "epoch": 1.43, + "learning_rate": 4.860781425972451e-05, + "loss": 1.9356, + "step": 16570 + }, + { + "epoch": 1.44, + "learning_rate": 4.8606947933812705e-05, + "loss": 1.9806, + "step": 16580 + }, + { + "epoch": 1.44, + "learning_rate": 4.860608160790089e-05, + "loss": 1.9356, + "step": 16590 + }, + { + "epoch": 1.44, + "learning_rate": 4.860521528198909e-05, + "loss": 1.9438, + "step": 16600 + }, + { + "epoch": 1.44, + "learning_rate": 4.8604348956077275e-05, + "loss": 1.9829, + "step": 16610 + }, + { + "epoch": 1.44, + "learning_rate": 4.860348263016547e-05, + "loss": 1.956, + "step": 16620 + }, + { + "epoch": 1.44, + "learning_rate": 4.8602616304253664e-05, + "loss": 2.0103, + "step": 16630 + }, + { + "epoch": 1.44, + "learning_rate": 4.860174997834185e-05, + "loss": 1.9472, + "step": 16640 + }, + { + "epoch": 1.44, + "learning_rate": 4.8600883652430046e-05, + "loss": 1.9693, + "step": 16650 + }, + { + "epoch": 1.44, + "learning_rate": 4.860001732651824e-05, + "loss": 2.0022, + "step": 16660 + }, + { + "epoch": 1.44, + "learning_rate": 4.859915100060643e-05, + "loss": 1.958, + "step": 16670 + }, + { + "epoch": 1.44, + "learning_rate": 4.859828467469462e-05, + "loss": 1.9672, + "step": 16680 + }, + { + "epoch": 1.44, + "learning_rate": 4.859741834878282e-05, + "loss": 1.9802, + "step": 16690 + }, + { + "epoch": 1.45, + "learning_rate": 4.8596552022871005e-05, + "loss": 1.9389, + "step": 16700 + }, + { + "epoch": 1.45, + "learning_rate": 4.85956856969592e-05, + "loss": 1.9528, + "step": 16710 + }, + { + "epoch": 1.45, + "learning_rate": 4.859481937104739e-05, + "loss": 1.9824, + "step": 16720 + }, + { + "epoch": 1.45, + "learning_rate": 4.859395304513558e-05, + "loss": 1.9973, + "step": 16730 + }, + { + "epoch": 1.45, + "learning_rate": 4.8593086719223776e-05, + "loss": 1.984, + "step": 16740 + }, + { + "epoch": 1.45, + "learning_rate": 4.859222039331196e-05, + "loss": 1.9813, + "step": 16750 + }, + { + "epoch": 1.45, + "learning_rate": 4.859135406740016e-05, + "loss": 2.0005, + "step": 16760 + }, + { + "epoch": 1.45, + "learning_rate": 4.859048774148835e-05, + "loss": 1.9758, + "step": 16770 + }, + { + "epoch": 1.45, + "learning_rate": 4.858962141557654e-05, + "loss": 1.9168, + "step": 16780 + }, + { + "epoch": 1.45, + "learning_rate": 4.8588755089664734e-05, + "loss": 1.9882, + "step": 16790 + }, + { + "epoch": 1.45, + "learning_rate": 4.858788876375293e-05, + "loss": 1.9567, + "step": 16800 + }, + { + "epoch": 1.45, + "learning_rate": 4.8587022437841116e-05, + "loss": 1.9892, + "step": 16810 + }, + { + "epoch": 1.46, + "learning_rate": 4.858615611192931e-05, + "loss": 1.9668, + "step": 16820 + }, + { + "epoch": 1.46, + "learning_rate": 4.85852897860175e-05, + "loss": 1.9689, + "step": 16830 + }, + { + "epoch": 1.46, + "learning_rate": 4.858442346010569e-05, + "loss": 1.9653, + "step": 16840 + }, + { + "epoch": 1.46, + "learning_rate": 4.858355713419389e-05, + "loss": 2.0257, + "step": 16850 + }, + { + "epoch": 1.46, + "learning_rate": 4.8582690808282075e-05, + "loss": 1.9732, + "step": 16860 + }, + { + "epoch": 1.46, + "learning_rate": 4.858182448237027e-05, + "loss": 2.0045, + "step": 16870 + }, + { + "epoch": 1.46, + "learning_rate": 4.8580958156458464e-05, + "loss": 2.016, + "step": 16880 + }, + { + "epoch": 1.46, + "learning_rate": 4.858009183054665e-05, + "loss": 1.9878, + "step": 16890 + }, + { + "epoch": 1.46, + "learning_rate": 4.8579225504634846e-05, + "loss": 1.9725, + "step": 16900 + }, + { + "epoch": 1.46, + "learning_rate": 4.857835917872304e-05, + "loss": 1.9542, + "step": 16910 + }, + { + "epoch": 1.46, + "learning_rate": 4.857749285281123e-05, + "loss": 1.9806, + "step": 16920 + }, + { + "epoch": 1.47, + "learning_rate": 4.857662652689942e-05, + "loss": 1.9901, + "step": 16930 + }, + { + "epoch": 1.47, + "learning_rate": 4.857576020098761e-05, + "loss": 1.9764, + "step": 16940 + }, + { + "epoch": 1.47, + "learning_rate": 4.8574893875075805e-05, + "loss": 1.9472, + "step": 16950 + }, + { + "epoch": 1.47, + "learning_rate": 4.8574027549164e-05, + "loss": 1.9631, + "step": 16960 + }, + { + "epoch": 1.47, + "learning_rate": 4.857316122325219e-05, + "loss": 2.0215, + "step": 16970 + }, + { + "epoch": 1.47, + "learning_rate": 4.857229489734038e-05, + "loss": 1.9526, + "step": 16980 + }, + { + "epoch": 1.47, + "learning_rate": 4.8571428571428576e-05, + "loss": 1.9726, + "step": 16990 + }, + { + "epoch": 1.47, + "learning_rate": 4.8570562245516764e-05, + "loss": 2.0058, + "step": 17000 + }, + { + "epoch": 1.47, + "learning_rate": 4.856969591960496e-05, + "loss": 1.9582, + "step": 17010 + }, + { + "epoch": 1.47, + "learning_rate": 4.856882959369315e-05, + "loss": 2.0093, + "step": 17020 + }, + { + "epoch": 1.47, + "learning_rate": 4.856796326778134e-05, + "loss": 1.962, + "step": 17030 + }, + { + "epoch": 1.47, + "learning_rate": 4.8567096941869535e-05, + "loss": 2.0057, + "step": 17040 + }, + { + "epoch": 1.48, + "learning_rate": 4.856623061595772e-05, + "loss": 1.9529, + "step": 17050 + }, + { + "epoch": 1.48, + "learning_rate": 4.856536429004592e-05, + "loss": 1.9861, + "step": 17060 + }, + { + "epoch": 1.48, + "learning_rate": 4.856449796413411e-05, + "loss": 1.9842, + "step": 17070 + }, + { + "epoch": 1.48, + "learning_rate": 4.85636316382223e-05, + "loss": 1.9607, + "step": 17080 + }, + { + "epoch": 1.48, + "learning_rate": 4.856276531231049e-05, + "loss": 2.026, + "step": 17090 + }, + { + "epoch": 1.48, + "learning_rate": 4.856189898639869e-05, + "loss": 1.9463, + "step": 17100 + }, + { + "epoch": 1.48, + "learning_rate": 4.8561032660486875e-05, + "loss": 1.9972, + "step": 17110 + }, + { + "epoch": 1.48, + "learning_rate": 4.856016633457507e-05, + "loss": 1.996, + "step": 17120 + }, + { + "epoch": 1.48, + "learning_rate": 4.8559300008663264e-05, + "loss": 2.01, + "step": 17130 + }, + { + "epoch": 1.48, + "learning_rate": 4.855843368275145e-05, + "loss": 1.95, + "step": 17140 + }, + { + "epoch": 1.48, + "learning_rate": 4.8557567356839646e-05, + "loss": 1.991, + "step": 17150 + }, + { + "epoch": 1.49, + "learning_rate": 4.8556701030927834e-05, + "loss": 1.9707, + "step": 17160 + }, + { + "epoch": 1.49, + "learning_rate": 4.855583470501603e-05, + "loss": 1.931, + "step": 17170 + }, + { + "epoch": 1.49, + "learning_rate": 4.855496837910422e-05, + "loss": 1.9346, + "step": 17180 + }, + { + "epoch": 1.49, + "learning_rate": 4.855410205319241e-05, + "loss": 2.021, + "step": 17190 + }, + { + "epoch": 1.49, + "learning_rate": 4.8553235727280605e-05, + "loss": 1.9194, + "step": 17200 + }, + { + "epoch": 1.49, + "learning_rate": 4.85523694013688e-05, + "loss": 1.9816, + "step": 17210 + }, + { + "epoch": 1.49, + "learning_rate": 4.855150307545699e-05, + "loss": 1.9781, + "step": 17220 + }, + { + "epoch": 1.49, + "learning_rate": 4.855063674954518e-05, + "loss": 2.0309, + "step": 17230 + }, + { + "epoch": 1.49, + "learning_rate": 4.854977042363337e-05, + "loss": 1.9733, + "step": 17240 + }, + { + "epoch": 1.49, + "learning_rate": 4.8548904097721564e-05, + "loss": 2.0206, + "step": 17250 + }, + { + "epoch": 1.49, + "learning_rate": 4.854803777180976e-05, + "loss": 1.9719, + "step": 17260 + }, + { + "epoch": 1.49, + "learning_rate": 4.8547171445897946e-05, + "loss": 1.997, + "step": 17270 + }, + { + "epoch": 1.5, + "learning_rate": 4.854630511998614e-05, + "loss": 1.9527, + "step": 17280 + }, + { + "epoch": 1.5, + "learning_rate": 4.8545438794074335e-05, + "loss": 1.9333, + "step": 17290 + }, + { + "epoch": 1.5, + "learning_rate": 4.854457246816252e-05, + "loss": 2.0358, + "step": 17300 + }, + { + "epoch": 1.5, + "learning_rate": 4.854370614225072e-05, + "loss": 1.9785, + "step": 17310 + }, + { + "epoch": 1.5, + "learning_rate": 4.854283981633891e-05, + "loss": 1.9962, + "step": 17320 + }, + { + "epoch": 1.5, + "learning_rate": 4.85419734904271e-05, + "loss": 1.9589, + "step": 17330 + }, + { + "epoch": 1.5, + "learning_rate": 4.8541107164515294e-05, + "loss": 1.951, + "step": 17340 + }, + { + "epoch": 1.5, + "learning_rate": 4.854024083860348e-05, + "loss": 1.9612, + "step": 17350 + }, + { + "epoch": 1.5, + "learning_rate": 4.8539374512691676e-05, + "loss": 1.965, + "step": 17360 + }, + { + "epoch": 1.5, + "learning_rate": 4.853850818677987e-05, + "loss": 1.9009, + "step": 17370 + }, + { + "epoch": 1.5, + "learning_rate": 4.853764186086806e-05, + "loss": 1.9299, + "step": 17380 + }, + { + "epoch": 1.51, + "learning_rate": 4.853677553495625e-05, + "loss": 1.9468, + "step": 17390 + }, + { + "epoch": 1.51, + "learning_rate": 4.853590920904445e-05, + "loss": 1.9913, + "step": 17400 + }, + { + "epoch": 1.51, + "learning_rate": 4.8535042883132634e-05, + "loss": 1.9911, + "step": 17410 + }, + { + "epoch": 1.51, + "learning_rate": 4.853417655722083e-05, + "loss": 2.021, + "step": 17420 + }, + { + "epoch": 1.51, + "learning_rate": 4.8533310231309023e-05, + "loss": 1.9132, + "step": 17430 + }, + { + "epoch": 1.51, + "learning_rate": 4.853244390539721e-05, + "loss": 1.969, + "step": 17440 + }, + { + "epoch": 1.51, + "learning_rate": 4.8531577579485406e-05, + "loss": 1.9827, + "step": 17450 + }, + { + "epoch": 1.51, + "learning_rate": 4.853071125357359e-05, + "loss": 1.9847, + "step": 17460 + }, + { + "epoch": 1.51, + "learning_rate": 4.852984492766179e-05, + "loss": 1.9069, + "step": 17470 + }, + { + "epoch": 1.51, + "learning_rate": 4.852897860174998e-05, + "loss": 1.9849, + "step": 17480 + }, + { + "epoch": 1.51, + "learning_rate": 4.852811227583817e-05, + "loss": 1.9882, + "step": 17490 + }, + { + "epoch": 1.51, + "learning_rate": 4.8527245949926364e-05, + "loss": 1.9606, + "step": 17500 + }, + { + "epoch": 1.52, + "learning_rate": 4.852637962401456e-05, + "loss": 1.9281, + "step": 17510 + }, + { + "epoch": 1.52, + "learning_rate": 4.8525513298102746e-05, + "loss": 1.9048, + "step": 17520 + }, + { + "epoch": 1.52, + "learning_rate": 4.852464697219094e-05, + "loss": 2.0329, + "step": 17530 + }, + { + "epoch": 1.52, + "learning_rate": 4.8523780646279135e-05, + "loss": 1.9524, + "step": 17540 + }, + { + "epoch": 1.52, + "learning_rate": 4.852291432036732e-05, + "loss": 1.9482, + "step": 17550 + }, + { + "epoch": 1.52, + "learning_rate": 4.852204799445552e-05, + "loss": 1.9792, + "step": 17560 + }, + { + "epoch": 1.52, + "learning_rate": 4.8521181668543705e-05, + "loss": 1.9512, + "step": 17570 + }, + { + "epoch": 1.52, + "learning_rate": 4.85203153426319e-05, + "loss": 2.0054, + "step": 17580 + }, + { + "epoch": 1.52, + "learning_rate": 4.8519449016720094e-05, + "loss": 1.9649, + "step": 17590 + }, + { + "epoch": 1.52, + "learning_rate": 4.851858269080828e-05, + "loss": 1.9663, + "step": 17600 + }, + { + "epoch": 1.52, + "learning_rate": 4.8517716364896476e-05, + "loss": 1.9669, + "step": 17610 + }, + { + "epoch": 1.53, + "learning_rate": 4.851685003898467e-05, + "loss": 1.9572, + "step": 17620 + }, + { + "epoch": 1.53, + "learning_rate": 4.851598371307286e-05, + "loss": 1.9304, + "step": 17630 + }, + { + "epoch": 1.53, + "learning_rate": 4.851511738716105e-05, + "loss": 1.952, + "step": 17640 + }, + { + "epoch": 1.53, + "learning_rate": 4.851425106124925e-05, + "loss": 1.9378, + "step": 17650 + }, + { + "epoch": 1.53, + "learning_rate": 4.8513384735337435e-05, + "loss": 1.9548, + "step": 17660 + }, + { + "epoch": 1.53, + "learning_rate": 4.851251840942563e-05, + "loss": 1.9459, + "step": 17670 + }, + { + "epoch": 1.53, + "learning_rate": 4.851165208351382e-05, + "loss": 1.9673, + "step": 17680 + }, + { + "epoch": 1.53, + "learning_rate": 4.851078575760201e-05, + "loss": 1.9969, + "step": 17690 + }, + { + "epoch": 1.53, + "learning_rate": 4.8509919431690206e-05, + "loss": 1.9691, + "step": 17700 + }, + { + "epoch": 1.53, + "learning_rate": 4.8509053105778394e-05, + "loss": 1.9483, + "step": 17710 + }, + { + "epoch": 1.53, + "learning_rate": 4.850818677986659e-05, + "loss": 1.9767, + "step": 17720 + }, + { + "epoch": 1.53, + "learning_rate": 4.850732045395478e-05, + "loss": 1.9659, + "step": 17730 + }, + { + "epoch": 1.54, + "learning_rate": 4.850645412804297e-05, + "loss": 1.9886, + "step": 17740 + }, + { + "epoch": 1.54, + "learning_rate": 4.8505587802131165e-05, + "loss": 1.9475, + "step": 17750 + }, + { + "epoch": 1.54, + "learning_rate": 4.850472147621936e-05, + "loss": 1.9976, + "step": 17760 + }, + { + "epoch": 1.54, + "learning_rate": 4.850385515030755e-05, + "loss": 1.9661, + "step": 17770 + }, + { + "epoch": 1.54, + "learning_rate": 4.850298882439574e-05, + "loss": 1.9585, + "step": 17780 + }, + { + "epoch": 1.54, + "learning_rate": 4.850212249848393e-05, + "loss": 1.9869, + "step": 17790 + }, + { + "epoch": 1.54, + "learning_rate": 4.850125617257212e-05, + "loss": 1.9833, + "step": 17800 + }, + { + "epoch": 1.54, + "learning_rate": 4.850038984666032e-05, + "loss": 1.9721, + "step": 17810 + }, + { + "epoch": 1.54, + "learning_rate": 4.8499523520748505e-05, + "loss": 1.942, + "step": 17820 + }, + { + "epoch": 1.54, + "learning_rate": 4.84986571948367e-05, + "loss": 1.9355, + "step": 17830 + }, + { + "epoch": 1.54, + "learning_rate": 4.8497790868924894e-05, + "loss": 1.9679, + "step": 17840 + }, + { + "epoch": 1.54, + "learning_rate": 4.849692454301308e-05, + "loss": 2.0172, + "step": 17850 + }, + { + "epoch": 1.55, + "learning_rate": 4.8496058217101276e-05, + "loss": 1.9947, + "step": 17860 + }, + { + "epoch": 1.55, + "learning_rate": 4.849519189118947e-05, + "loss": 1.9428, + "step": 17870 + }, + { + "epoch": 1.55, + "learning_rate": 4.849432556527766e-05, + "loss": 1.8822, + "step": 17880 + }, + { + "epoch": 1.55, + "learning_rate": 4.849345923936585e-05, + "loss": 1.9996, + "step": 17890 + }, + { + "epoch": 1.55, + "learning_rate": 4.849259291345404e-05, + "loss": 1.982, + "step": 17900 + }, + { + "epoch": 1.55, + "learning_rate": 4.8491726587542235e-05, + "loss": 1.9832, + "step": 17910 + }, + { + "epoch": 1.55, + "learning_rate": 4.849086026163043e-05, + "loss": 1.9757, + "step": 17920 + }, + { + "epoch": 1.55, + "learning_rate": 4.848999393571862e-05, + "loss": 1.9959, + "step": 17930 + }, + { + "epoch": 1.55, + "learning_rate": 4.848912760980681e-05, + "loss": 2.0008, + "step": 17940 + }, + { + "epoch": 1.55, + "learning_rate": 4.8488261283895006e-05, + "loss": 1.9759, + "step": 17950 + }, + { + "epoch": 1.55, + "learning_rate": 4.8487394957983194e-05, + "loss": 1.997, + "step": 17960 + }, + { + "epoch": 1.56, + "learning_rate": 4.848652863207139e-05, + "loss": 2.0377, + "step": 17970 + }, + { + "epoch": 1.56, + "learning_rate": 4.8485662306159576e-05, + "loss": 1.9709, + "step": 17980 + }, + { + "epoch": 1.56, + "learning_rate": 4.848479598024777e-05, + "loss": 1.9721, + "step": 17990 + }, + { + "epoch": 1.56, + "learning_rate": 4.8483929654335965e-05, + "loss": 1.9053, + "step": 18000 + }, + { + "epoch": 1.56, + "learning_rate": 4.848306332842415e-05, + "loss": 1.994, + "step": 18010 + }, + { + "epoch": 1.56, + "learning_rate": 4.848219700251235e-05, + "loss": 2.012, + "step": 18020 + }, + { + "epoch": 1.56, + "learning_rate": 4.848133067660054e-05, + "loss": 1.925, + "step": 18030 + }, + { + "epoch": 1.56, + "learning_rate": 4.848046435068873e-05, + "loss": 1.9914, + "step": 18040 + }, + { + "epoch": 1.56, + "learning_rate": 4.8479598024776924e-05, + "loss": 1.9283, + "step": 18050 + }, + { + "epoch": 1.56, + "learning_rate": 4.847873169886512e-05, + "loss": 1.9944, + "step": 18060 + }, + { + "epoch": 1.56, + "learning_rate": 4.8477865372953306e-05, + "loss": 1.9641, + "step": 18070 + }, + { + "epoch": 1.56, + "learning_rate": 4.84769990470415e-05, + "loss": 1.9482, + "step": 18080 + }, + { + "epoch": 1.57, + "learning_rate": 4.847613272112969e-05, + "loss": 1.9229, + "step": 18090 + }, + { + "epoch": 1.57, + "learning_rate": 4.847526639521788e-05, + "loss": 1.9688, + "step": 18100 + }, + { + "epoch": 1.57, + "learning_rate": 4.847440006930608e-05, + "loss": 1.9313, + "step": 18110 + }, + { + "epoch": 1.57, + "learning_rate": 4.8473533743394264e-05, + "loss": 1.9683, + "step": 18120 + }, + { + "epoch": 1.57, + "learning_rate": 4.847266741748246e-05, + "loss": 1.9772, + "step": 18130 + }, + { + "epoch": 1.57, + "learning_rate": 4.847180109157065e-05, + "loss": 1.9306, + "step": 18140 + }, + { + "epoch": 1.57, + "learning_rate": 4.847093476565884e-05, + "loss": 1.9657, + "step": 18150 + }, + { + "epoch": 1.57, + "learning_rate": 4.8470068439747035e-05, + "loss": 1.9467, + "step": 18160 + }, + { + "epoch": 1.57, + "learning_rate": 4.846920211383523e-05, + "loss": 1.9895, + "step": 18170 + }, + { + "epoch": 1.57, + "learning_rate": 4.846833578792342e-05, + "loss": 1.9732, + "step": 18180 + }, + { + "epoch": 1.57, + "learning_rate": 4.846746946201161e-05, + "loss": 1.9771, + "step": 18190 + }, + { + "epoch": 1.58, + "learning_rate": 4.84666031360998e-05, + "loss": 1.9731, + "step": 18200 + }, + { + "epoch": 1.58, + "learning_rate": 4.8465736810187994e-05, + "loss": 1.959, + "step": 18210 + }, + { + "epoch": 1.58, + "learning_rate": 4.846487048427619e-05, + "loss": 1.9822, + "step": 18220 + }, + { + "epoch": 1.58, + "learning_rate": 4.8464004158364376e-05, + "loss": 1.9942, + "step": 18230 + }, + { + "epoch": 1.58, + "learning_rate": 4.846313783245257e-05, + "loss": 1.9115, + "step": 18240 + }, + { + "epoch": 1.58, + "learning_rate": 4.8462271506540765e-05, + "loss": 1.9831, + "step": 18250 + }, + { + "epoch": 1.58, + "learning_rate": 4.846140518062895e-05, + "loss": 1.9691, + "step": 18260 + }, + { + "epoch": 1.58, + "learning_rate": 4.846053885471715e-05, + "loss": 1.9316, + "step": 18270 + }, + { + "epoch": 1.58, + "learning_rate": 4.845967252880534e-05, + "loss": 1.936, + "step": 18280 + }, + { + "epoch": 1.58, + "learning_rate": 4.845880620289353e-05, + "loss": 1.945, + "step": 18290 + }, + { + "epoch": 1.58, + "learning_rate": 4.8457939876981724e-05, + "loss": 1.9465, + "step": 18300 + }, + { + "epoch": 1.58, + "learning_rate": 4.845707355106991e-05, + "loss": 1.9518, + "step": 18310 + }, + { + "epoch": 1.59, + "learning_rate": 4.8456207225158106e-05, + "loss": 1.938, + "step": 18320 + }, + { + "epoch": 1.59, + "learning_rate": 4.84553408992463e-05, + "loss": 1.9786, + "step": 18330 + }, + { + "epoch": 1.59, + "learning_rate": 4.845447457333449e-05, + "loss": 1.9934, + "step": 18340 + }, + { + "epoch": 1.59, + "learning_rate": 4.845360824742268e-05, + "loss": 1.9674, + "step": 18350 + }, + { + "epoch": 1.59, + "learning_rate": 4.845274192151088e-05, + "loss": 1.9768, + "step": 18360 + }, + { + "epoch": 1.59, + "learning_rate": 4.8451875595599065e-05, + "loss": 1.9625, + "step": 18370 + }, + { + "epoch": 1.59, + "learning_rate": 4.845100926968726e-05, + "loss": 1.917, + "step": 18380 + }, + { + "epoch": 1.59, + "learning_rate": 4.8450142943775454e-05, + "loss": 1.9367, + "step": 18390 + }, + { + "epoch": 1.59, + "learning_rate": 4.844927661786364e-05, + "loss": 1.974, + "step": 18400 + }, + { + "epoch": 1.59, + "learning_rate": 4.8448410291951836e-05, + "loss": 1.9432, + "step": 18410 + }, + { + "epoch": 1.59, + "learning_rate": 4.8447543966040024e-05, + "loss": 1.9172, + "step": 18420 + }, + { + "epoch": 1.6, + "learning_rate": 4.844667764012822e-05, + "loss": 1.9661, + "step": 18430 + }, + { + "epoch": 1.6, + "learning_rate": 4.844581131421641e-05, + "loss": 1.9596, + "step": 18440 + }, + { + "epoch": 1.6, + "learning_rate": 4.84449449883046e-05, + "loss": 1.9119, + "step": 18450 + }, + { + "epoch": 1.6, + "learning_rate": 4.8444078662392795e-05, + "loss": 1.9929, + "step": 18460 + }, + { + "epoch": 1.6, + "learning_rate": 4.844321233648099e-05, + "loss": 1.9096, + "step": 18470 + }, + { + "epoch": 1.6, + "learning_rate": 4.844234601056918e-05, + "loss": 1.9527, + "step": 18480 + }, + { + "epoch": 1.6, + "learning_rate": 4.844147968465737e-05, + "loss": 1.9277, + "step": 18490 + }, + { + "epoch": 1.6, + "learning_rate": 4.8440613358745566e-05, + "loss": 1.9498, + "step": 18500 + }, + { + "epoch": 1.6, + "learning_rate": 4.843974703283375e-05, + "loss": 1.9082, + "step": 18510 + }, + { + "epoch": 1.6, + "learning_rate": 4.843888070692195e-05, + "loss": 1.9614, + "step": 18520 + }, + { + "epoch": 1.6, + "learning_rate": 4.8438014381010135e-05, + "loss": 1.9414, + "step": 18530 + }, + { + "epoch": 1.6, + "learning_rate": 4.843714805509833e-05, + "loss": 1.9833, + "step": 18540 + }, + { + "epoch": 1.61, + "learning_rate": 4.8436281729186524e-05, + "loss": 1.9758, + "step": 18550 + }, + { + "epoch": 1.61, + "learning_rate": 4.843541540327471e-05, + "loss": 1.9906, + "step": 18560 + }, + { + "epoch": 1.61, + "learning_rate": 4.8434549077362906e-05, + "loss": 1.9326, + "step": 18570 + }, + { + "epoch": 1.61, + "learning_rate": 4.84336827514511e-05, + "loss": 1.9923, + "step": 18580 + }, + { + "epoch": 1.61, + "learning_rate": 4.843281642553929e-05, + "loss": 1.9404, + "step": 18590 + }, + { + "epoch": 1.61, + "learning_rate": 4.843195009962748e-05, + "loss": 2.0039, + "step": 18600 + }, + { + "epoch": 1.61, + "learning_rate": 4.843108377371568e-05, + "loss": 1.973, + "step": 18610 + }, + { + "epoch": 1.61, + "learning_rate": 4.8430217447803865e-05, + "loss": 1.9677, + "step": 18620 + }, + { + "epoch": 1.61, + "learning_rate": 4.842935112189206e-05, + "loss": 1.931, + "step": 18630 + }, + { + "epoch": 1.61, + "learning_rate": 4.842848479598025e-05, + "loss": 1.9141, + "step": 18640 + }, + { + "epoch": 1.61, + "learning_rate": 4.842761847006844e-05, + "loss": 1.9578, + "step": 18650 + }, + { + "epoch": 1.62, + "learning_rate": 4.8426752144156636e-05, + "loss": 1.9363, + "step": 18660 + }, + { + "epoch": 1.62, + "learning_rate": 4.8425885818244824e-05, + "loss": 1.9417, + "step": 18670 + }, + { + "epoch": 1.62, + "learning_rate": 4.842501949233302e-05, + "loss": 1.9688, + "step": 18680 + }, + { + "epoch": 1.62, + "learning_rate": 4.842415316642121e-05, + "loss": 1.9174, + "step": 18690 + }, + { + "epoch": 1.62, + "learning_rate": 4.84232868405094e-05, + "loss": 1.9244, + "step": 18700 + }, + { + "epoch": 1.62, + "learning_rate": 4.8422420514597595e-05, + "loss": 1.9787, + "step": 18710 + }, + { + "epoch": 1.62, + "learning_rate": 4.842155418868579e-05, + "loss": 1.9266, + "step": 18720 + }, + { + "epoch": 1.62, + "learning_rate": 4.842068786277398e-05, + "loss": 1.9267, + "step": 18730 + }, + { + "epoch": 1.62, + "learning_rate": 4.841982153686217e-05, + "loss": 1.9711, + "step": 18740 + }, + { + "epoch": 1.62, + "learning_rate": 4.841895521095036e-05, + "loss": 1.9725, + "step": 18750 + }, + { + "epoch": 1.62, + "learning_rate": 4.8418088885038554e-05, + "loss": 1.9178, + "step": 18760 + }, + { + "epoch": 1.62, + "learning_rate": 4.841722255912675e-05, + "loss": 1.964, + "step": 18770 + }, + { + "epoch": 1.63, + "learning_rate": 4.8416356233214936e-05, + "loss": 1.9063, + "step": 18780 + }, + { + "epoch": 1.63, + "learning_rate": 4.841548990730313e-05, + "loss": 1.9317, + "step": 18790 + }, + { + "epoch": 1.63, + "learning_rate": 4.8414623581391325e-05, + "loss": 1.9589, + "step": 18800 + }, + { + "epoch": 1.63, + "learning_rate": 4.841375725547951e-05, + "loss": 1.9551, + "step": 18810 + }, + { + "epoch": 1.63, + "learning_rate": 4.841289092956771e-05, + "loss": 1.9773, + "step": 18820 + }, + { + "epoch": 1.63, + "learning_rate": 4.8412024603655894e-05, + "loss": 1.9612, + "step": 18830 + }, + { + "epoch": 1.63, + "learning_rate": 4.841115827774409e-05, + "loss": 1.9807, + "step": 18840 + }, + { + "epoch": 1.63, + "learning_rate": 4.841029195183228e-05, + "loss": 1.9892, + "step": 18850 + }, + { + "epoch": 1.63, + "learning_rate": 4.840942562592047e-05, + "loss": 1.9692, + "step": 18860 + }, + { + "epoch": 1.63, + "learning_rate": 4.8408559300008665e-05, + "loss": 1.9779, + "step": 18870 + }, + { + "epoch": 1.63, + "learning_rate": 4.840769297409686e-05, + "loss": 1.9366, + "step": 18880 + }, + { + "epoch": 1.64, + "learning_rate": 4.840682664818505e-05, + "loss": 1.9284, + "step": 18890 + }, + { + "epoch": 1.64, + "learning_rate": 4.840596032227324e-05, + "loss": 1.9522, + "step": 18900 + }, + { + "epoch": 1.64, + "learning_rate": 4.8405093996361437e-05, + "loss": 1.9419, + "step": 18910 + }, + { + "epoch": 1.64, + "learning_rate": 4.8404227670449624e-05, + "loss": 1.9354, + "step": 18920 + }, + { + "epoch": 1.64, + "learning_rate": 4.840336134453782e-05, + "loss": 1.9734, + "step": 18930 + }, + { + "epoch": 1.64, + "learning_rate": 4.8402495018626006e-05, + "loss": 1.9733, + "step": 18940 + }, + { + "epoch": 1.64, + "learning_rate": 4.84016286927142e-05, + "loss": 1.9435, + "step": 18950 + }, + { + "epoch": 1.64, + "learning_rate": 4.8400762366802395e-05, + "loss": 1.9254, + "step": 18960 + }, + { + "epoch": 1.64, + "learning_rate": 4.839989604089058e-05, + "loss": 1.9002, + "step": 18970 + }, + { + "epoch": 1.64, + "learning_rate": 4.839902971497878e-05, + "loss": 1.9685, + "step": 18980 + }, + { + "epoch": 1.64, + "learning_rate": 4.839816338906697e-05, + "loss": 1.9508, + "step": 18990 + }, + { + "epoch": 1.64, + "learning_rate": 4.839729706315516e-05, + "loss": 1.984, + "step": 19000 + }, + { + "epoch": 1.65, + "learning_rate": 4.8396430737243354e-05, + "loss": 1.9235, + "step": 19010 + }, + { + "epoch": 1.65, + "learning_rate": 4.839556441133155e-05, + "loss": 1.9479, + "step": 19020 + }, + { + "epoch": 1.65, + "learning_rate": 4.8394698085419736e-05, + "loss": 1.9291, + "step": 19030 + }, + { + "epoch": 1.65, + "learning_rate": 4.839383175950793e-05, + "loss": 1.9326, + "step": 19040 + }, + { + "epoch": 1.65, + "learning_rate": 4.839296543359612e-05, + "loss": 1.9737, + "step": 19050 + }, + { + "epoch": 1.65, + "learning_rate": 4.839209910768431e-05, + "loss": 1.9324, + "step": 19060 + }, + { + "epoch": 1.65, + "learning_rate": 4.839123278177251e-05, + "loss": 1.9626, + "step": 19070 + }, + { + "epoch": 1.65, + "learning_rate": 4.8390366455860695e-05, + "loss": 1.8732, + "step": 19080 + }, + { + "epoch": 1.65, + "learning_rate": 4.838950012994889e-05, + "loss": 1.9361, + "step": 19090 + }, + { + "epoch": 1.65, + "learning_rate": 4.8388633804037084e-05, + "loss": 1.9814, + "step": 19100 + }, + { + "epoch": 1.65, + "learning_rate": 4.838776747812527e-05, + "loss": 1.9545, + "step": 19110 + }, + { + "epoch": 1.65, + "learning_rate": 4.8386901152213466e-05, + "loss": 1.9614, + "step": 19120 + }, + { + "epoch": 1.66, + "learning_rate": 4.838603482630166e-05, + "loss": 1.9474, + "step": 19130 + }, + { + "epoch": 1.66, + "learning_rate": 4.838516850038985e-05, + "loss": 1.9478, + "step": 19140 + }, + { + "epoch": 1.66, + "learning_rate": 4.838430217447804e-05, + "loss": 1.9539, + "step": 19150 + }, + { + "epoch": 1.66, + "learning_rate": 4.838343584856623e-05, + "loss": 1.8978, + "step": 19160 + }, + { + "epoch": 1.66, + "learning_rate": 4.8382569522654425e-05, + "loss": 1.9824, + "step": 19170 + }, + { + "epoch": 1.66, + "learning_rate": 4.838170319674262e-05, + "loss": 1.9764, + "step": 19180 + }, + { + "epoch": 1.66, + "learning_rate": 4.838083687083081e-05, + "loss": 1.9728, + "step": 19190 + }, + { + "epoch": 1.66, + "learning_rate": 4.8379970544919e-05, + "loss": 1.9865, + "step": 19200 + }, + { + "epoch": 1.66, + "learning_rate": 4.8379104219007196e-05, + "loss": 1.9697, + "step": 19210 + }, + { + "epoch": 1.66, + "learning_rate": 4.837823789309538e-05, + "loss": 1.9295, + "step": 19220 + }, + { + "epoch": 1.66, + "learning_rate": 4.837737156718358e-05, + "loss": 1.9192, + "step": 19230 + }, + { + "epoch": 1.67, + "learning_rate": 4.837650524127177e-05, + "loss": 1.9466, + "step": 19240 + }, + { + "epoch": 1.67, + "learning_rate": 4.837563891535996e-05, + "loss": 1.9095, + "step": 19250 + }, + { + "epoch": 1.67, + "learning_rate": 4.8374772589448154e-05, + "loss": 1.8942, + "step": 19260 + }, + { + "epoch": 1.67, + "learning_rate": 4.837390626353634e-05, + "loss": 1.9542, + "step": 19270 + }, + { + "epoch": 1.67, + "learning_rate": 4.8373039937624536e-05, + "loss": 1.9754, + "step": 19280 + }, + { + "epoch": 1.67, + "learning_rate": 4.837217361171273e-05, + "loss": 1.9585, + "step": 19290 + }, + { + "epoch": 1.67, + "learning_rate": 4.837130728580092e-05, + "loss": 1.952, + "step": 19300 + }, + { + "epoch": 1.67, + "learning_rate": 4.837044095988911e-05, + "loss": 1.9241, + "step": 19310 + }, + { + "epoch": 1.67, + "learning_rate": 4.836957463397731e-05, + "loss": 1.9341, + "step": 19320 + }, + { + "epoch": 1.67, + "learning_rate": 4.8368708308065495e-05, + "loss": 1.9017, + "step": 19330 + }, + { + "epoch": 1.67, + "learning_rate": 4.836784198215369e-05, + "loss": 1.942, + "step": 19340 + }, + { + "epoch": 1.67, + "learning_rate": 4.8366975656241884e-05, + "loss": 1.92, + "step": 19350 + }, + { + "epoch": 1.68, + "learning_rate": 4.836610933033007e-05, + "loss": 1.9267, + "step": 19360 + }, + { + "epoch": 1.68, + "learning_rate": 4.8365243004418266e-05, + "loss": 1.9445, + "step": 19370 + }, + { + "epoch": 1.68, + "learning_rate": 4.8364376678506454e-05, + "loss": 1.98, + "step": 19380 + }, + { + "epoch": 1.68, + "learning_rate": 4.836351035259465e-05, + "loss": 1.945, + "step": 19390 + }, + { + "epoch": 1.68, + "learning_rate": 4.836264402668284e-05, + "loss": 1.9138, + "step": 19400 + }, + { + "epoch": 1.68, + "learning_rate": 4.836177770077103e-05, + "loss": 1.9655, + "step": 19410 + }, + { + "epoch": 1.68, + "learning_rate": 4.8360911374859225e-05, + "loss": 1.952, + "step": 19420 + }, + { + "epoch": 1.68, + "learning_rate": 4.836004504894742e-05, + "loss": 1.9153, + "step": 19430 + }, + { + "epoch": 1.68, + "learning_rate": 4.835917872303561e-05, + "loss": 1.9299, + "step": 19440 + }, + { + "epoch": 1.68, + "learning_rate": 4.83583123971238e-05, + "loss": 1.9602, + "step": 19450 + }, + { + "epoch": 1.68, + "learning_rate": 4.8357446071211996e-05, + "loss": 1.9315, + "step": 19460 + }, + { + "epoch": 1.69, + "learning_rate": 4.8356579745300184e-05, + "loss": 1.9419, + "step": 19470 + }, + { + "epoch": 1.69, + "learning_rate": 4.835571341938838e-05, + "loss": 1.8913, + "step": 19480 + }, + { + "epoch": 1.69, + "learning_rate": 4.8354847093476566e-05, + "loss": 1.9942, + "step": 19490 + }, + { + "epoch": 1.69, + "learning_rate": 4.835398076756476e-05, + "loss": 1.9396, + "step": 19500 + }, + { + "epoch": 1.69, + "learning_rate": 4.8353114441652955e-05, + "loss": 1.9969, + "step": 19510 + }, + { + "epoch": 1.69, + "learning_rate": 4.835224811574114e-05, + "loss": 1.9528, + "step": 19520 + }, + { + "epoch": 1.69, + "learning_rate": 4.835138178982934e-05, + "loss": 1.8909, + "step": 19530 + }, + { + "epoch": 1.69, + "learning_rate": 4.835051546391753e-05, + "loss": 1.9308, + "step": 19540 + }, + { + "epoch": 1.69, + "learning_rate": 4.834964913800572e-05, + "loss": 1.9685, + "step": 19550 + }, + { + "epoch": 1.69, + "learning_rate": 4.834878281209391e-05, + "loss": 1.9718, + "step": 19560 + }, + { + "epoch": 1.69, + "learning_rate": 4.83479164861821e-05, + "loss": 1.9222, + "step": 19570 + }, + { + "epoch": 1.69, + "learning_rate": 4.8347050160270295e-05, + "loss": 1.9353, + "step": 19580 + }, + { + "epoch": 1.7, + "learning_rate": 4.834618383435849e-05, + "loss": 1.9603, + "step": 19590 + }, + { + "epoch": 1.7, + "learning_rate": 4.834531750844668e-05, + "loss": 1.9031, + "step": 19600 + }, + { + "epoch": 1.7, + "learning_rate": 4.834445118253487e-05, + "loss": 1.9881, + "step": 19610 + }, + { + "epoch": 1.7, + "learning_rate": 4.8343584856623066e-05, + "loss": 1.9052, + "step": 19620 + }, + { + "epoch": 1.7, + "learning_rate": 4.8342718530711254e-05, + "loss": 1.9034, + "step": 19630 + }, + { + "epoch": 1.7, + "learning_rate": 4.834185220479945e-05, + "loss": 1.909, + "step": 19640 + }, + { + "epoch": 1.7, + "learning_rate": 4.834098587888764e-05, + "loss": 1.9372, + "step": 19650 + }, + { + "epoch": 1.7, + "learning_rate": 4.834011955297583e-05, + "loss": 1.9229, + "step": 19660 + }, + { + "epoch": 1.7, + "learning_rate": 4.8339253227064025e-05, + "loss": 1.9451, + "step": 19670 + }, + { + "epoch": 1.7, + "learning_rate": 4.833838690115221e-05, + "loss": 1.9657, + "step": 19680 + }, + { + "epoch": 1.7, + "learning_rate": 4.833752057524041e-05, + "loss": 1.9584, + "step": 19690 + }, + { + "epoch": 1.71, + "learning_rate": 4.83366542493286e-05, + "loss": 1.903, + "step": 19700 + }, + { + "epoch": 1.71, + "learning_rate": 4.833578792341679e-05, + "loss": 1.8805, + "step": 19710 + }, + { + "epoch": 1.71, + "learning_rate": 4.8334921597504984e-05, + "loss": 1.9246, + "step": 19720 + }, + { + "epoch": 1.71, + "learning_rate": 4.833405527159318e-05, + "loss": 1.8687, + "step": 19730 + }, + { + "epoch": 1.71, + "learning_rate": 4.8333188945681366e-05, + "loss": 2.0207, + "step": 19740 + }, + { + "epoch": 1.71, + "learning_rate": 4.833232261976956e-05, + "loss": 1.9823, + "step": 19750 + }, + { + "epoch": 1.71, + "learning_rate": 4.8331456293857755e-05, + "loss": 1.9278, + "step": 19760 + }, + { + "epoch": 1.71, + "learning_rate": 4.833058996794594e-05, + "loss": 1.9883, + "step": 19770 + }, + { + "epoch": 1.71, + "learning_rate": 4.832972364203414e-05, + "loss": 1.9205, + "step": 19780 + }, + { + "epoch": 1.71, + "learning_rate": 4.8328857316122325e-05, + "loss": 1.9187, + "step": 19790 + }, + { + "epoch": 1.71, + "learning_rate": 4.832799099021052e-05, + "loss": 1.9514, + "step": 19800 + }, + { + "epoch": 1.71, + "learning_rate": 4.8327124664298714e-05, + "loss": 1.8675, + "step": 19810 + }, + { + "epoch": 1.72, + "learning_rate": 4.83262583383869e-05, + "loss": 2.0201, + "step": 19820 + }, + { + "epoch": 1.72, + "learning_rate": 4.8325392012475096e-05, + "loss": 1.9298, + "step": 19830 + }, + { + "epoch": 1.72, + "learning_rate": 4.832452568656329e-05, + "loss": 1.9174, + "step": 19840 + }, + { + "epoch": 1.72, + "learning_rate": 4.832365936065148e-05, + "loss": 1.9353, + "step": 19850 + }, + { + "epoch": 1.72, + "learning_rate": 4.832279303473967e-05, + "loss": 1.9242, + "step": 19860 + }, + { + "epoch": 1.72, + "learning_rate": 4.832192670882787e-05, + "loss": 1.9358, + "step": 19870 + }, + { + "epoch": 1.72, + "learning_rate": 4.8321060382916054e-05, + "loss": 1.91, + "step": 19880 + }, + { + "epoch": 1.72, + "learning_rate": 4.832019405700425e-05, + "loss": 1.9192, + "step": 19890 + }, + { + "epoch": 1.72, + "learning_rate": 4.831932773109244e-05, + "loss": 1.9292, + "step": 19900 + }, + { + "epoch": 1.72, + "learning_rate": 4.831846140518063e-05, + "loss": 1.9101, + "step": 19910 + }, + { + "epoch": 1.72, + "learning_rate": 4.8317595079268826e-05, + "loss": 1.952, + "step": 19920 + }, + { + "epoch": 1.73, + "learning_rate": 4.831672875335701e-05, + "loss": 1.9141, + "step": 19930 + }, + { + "epoch": 1.73, + "learning_rate": 4.831586242744521e-05, + "loss": 1.9546, + "step": 19940 + }, + { + "epoch": 1.73, + "learning_rate": 4.83149961015334e-05, + "loss": 1.9311, + "step": 19950 + }, + { + "epoch": 1.73, + "learning_rate": 4.831412977562159e-05, + "loss": 1.9431, + "step": 19960 + }, + { + "epoch": 1.73, + "learning_rate": 4.8313263449709784e-05, + "loss": 1.9626, + "step": 19970 + }, + { + "epoch": 1.73, + "learning_rate": 4.831239712379798e-05, + "loss": 1.8988, + "step": 19980 + }, + { + "epoch": 1.73, + "learning_rate": 4.8311530797886166e-05, + "loss": 1.9303, + "step": 19990 + }, + { + "epoch": 1.73, + "learning_rate": 4.831066447197436e-05, + "loss": 1.9159, + "step": 20000 + }, + { + "epoch": 1.73, + "learning_rate": 4.830979814606255e-05, + "loss": 1.9486, + "step": 20010 + }, + { + "epoch": 1.73, + "learning_rate": 4.830893182015074e-05, + "loss": 1.9616, + "step": 20020 + }, + { + "epoch": 1.73, + "learning_rate": 4.830806549423894e-05, + "loss": 1.9383, + "step": 20030 + }, + { + "epoch": 1.73, + "learning_rate": 4.8307199168327125e-05, + "loss": 1.915, + "step": 20040 + }, + { + "epoch": 1.74, + "learning_rate": 4.830633284241532e-05, + "loss": 1.919, + "step": 20050 + }, + { + "epoch": 1.74, + "learning_rate": 4.8305466516503514e-05, + "loss": 1.8995, + "step": 20060 + }, + { + "epoch": 1.74, + "learning_rate": 4.83046001905917e-05, + "loss": 1.9711, + "step": 20070 + }, + { + "epoch": 1.74, + "learning_rate": 4.8303733864679896e-05, + "loss": 1.9701, + "step": 20080 + }, + { + "epoch": 1.74, + "learning_rate": 4.830286753876809e-05, + "loss": 1.9804, + "step": 20090 + }, + { + "epoch": 1.74, + "learning_rate": 4.830200121285628e-05, + "loss": 1.9544, + "step": 20100 + }, + { + "epoch": 1.74, + "learning_rate": 4.830113488694447e-05, + "loss": 1.9618, + "step": 20110 + }, + { + "epoch": 1.74, + "learning_rate": 4.830026856103266e-05, + "loss": 1.9141, + "step": 20120 + }, + { + "epoch": 1.74, + "learning_rate": 4.8299402235120855e-05, + "loss": 1.9082, + "step": 20130 + }, + { + "epoch": 1.74, + "learning_rate": 4.829853590920905e-05, + "loss": 1.9733, + "step": 20140 + }, + { + "epoch": 1.74, + "learning_rate": 4.829766958329724e-05, + "loss": 1.9429, + "step": 20150 + }, + { + "epoch": 1.74, + "learning_rate": 4.829680325738543e-05, + "loss": 1.9514, + "step": 20160 + }, + { + "epoch": 1.75, + "learning_rate": 4.8295936931473626e-05, + "loss": 1.9571, + "step": 20170 + }, + { + "epoch": 1.75, + "learning_rate": 4.8295070605561814e-05, + "loss": 1.8946, + "step": 20180 + }, + { + "epoch": 1.75, + "learning_rate": 4.829420427965001e-05, + "loss": 1.878, + "step": 20190 + }, + { + "epoch": 1.75, + "learning_rate": 4.82933379537382e-05, + "loss": 1.8834, + "step": 20200 + }, + { + "epoch": 1.75, + "learning_rate": 4.829247162782639e-05, + "loss": 1.9763, + "step": 20210 + }, + { + "epoch": 1.75, + "learning_rate": 4.8291605301914585e-05, + "loss": 1.9309, + "step": 20220 + }, + { + "epoch": 1.75, + "learning_rate": 4.829073897600277e-05, + "loss": 1.912, + "step": 20230 + }, + { + "epoch": 1.75, + "learning_rate": 4.828987265009097e-05, + "loss": 1.8821, + "step": 20240 + }, + { + "epoch": 1.75, + "learning_rate": 4.828900632417916e-05, + "loss": 1.8951, + "step": 20250 + }, + { + "epoch": 1.75, + "learning_rate": 4.828813999826735e-05, + "loss": 1.9655, + "step": 20260 + }, + { + "epoch": 1.75, + "learning_rate": 4.828727367235554e-05, + "loss": 1.9412, + "step": 20270 + }, + { + "epoch": 1.76, + "learning_rate": 4.828640734644374e-05, + "loss": 1.9143, + "step": 20280 + }, + { + "epoch": 1.76, + "learning_rate": 4.8285541020531925e-05, + "loss": 1.9416, + "step": 20290 + }, + { + "epoch": 1.76, + "learning_rate": 4.828467469462012e-05, + "loss": 1.8963, + "step": 20300 + }, + { + "epoch": 1.76, + "learning_rate": 4.828380836870831e-05, + "loss": 1.9345, + "step": 20310 + }, + { + "epoch": 1.76, + "learning_rate": 4.82829420427965e-05, + "loss": 1.9061, + "step": 20320 + }, + { + "epoch": 1.76, + "learning_rate": 4.8282075716884696e-05, + "loss": 1.9988, + "step": 20330 + }, + { + "epoch": 1.76, + "learning_rate": 4.8281209390972884e-05, + "loss": 1.9097, + "step": 20340 + }, + { + "epoch": 1.76, + "learning_rate": 4.828034306506108e-05, + "loss": 1.9001, + "step": 20350 + }, + { + "epoch": 1.76, + "learning_rate": 4.827947673914927e-05, + "loss": 1.9717, + "step": 20360 + }, + { + "epoch": 1.76, + "learning_rate": 4.827861041323746e-05, + "loss": 1.9775, + "step": 20370 + }, + { + "epoch": 1.76, + "learning_rate": 4.8277744087325655e-05, + "loss": 1.916, + "step": 20380 + }, + { + "epoch": 1.76, + "learning_rate": 4.827687776141385e-05, + "loss": 1.8972, + "step": 20390 + }, + { + "epoch": 1.77, + "learning_rate": 4.827601143550204e-05, + "loss": 1.965, + "step": 20400 + }, + { + "epoch": 1.77, + "learning_rate": 4.827514510959023e-05, + "loss": 1.9164, + "step": 20410 + }, + { + "epoch": 1.77, + "learning_rate": 4.827427878367842e-05, + "loss": 1.8916, + "step": 20420 + }, + { + "epoch": 1.77, + "learning_rate": 4.8273412457766614e-05, + "loss": 1.8463, + "step": 20430 + }, + { + "epoch": 1.77, + "learning_rate": 4.827254613185481e-05, + "loss": 1.9289, + "step": 20440 + }, + { + "epoch": 1.77, + "learning_rate": 4.8271679805942996e-05, + "loss": 1.877, + "step": 20450 + }, + { + "epoch": 1.77, + "learning_rate": 4.827081348003119e-05, + "loss": 1.8716, + "step": 20460 + }, + { + "epoch": 1.77, + "learning_rate": 4.8269947154119385e-05, + "loss": 1.9538, + "step": 20470 + }, + { + "epoch": 1.77, + "learning_rate": 4.826908082820757e-05, + "loss": 1.8846, + "step": 20480 + }, + { + "epoch": 1.77, + "learning_rate": 4.826821450229577e-05, + "loss": 1.943, + "step": 20490 + }, + { + "epoch": 1.77, + "learning_rate": 4.826734817638396e-05, + "loss": 1.951, + "step": 20500 + }, + { + "epoch": 1.78, + "learning_rate": 4.826648185047215e-05, + "loss": 1.9241, + "step": 20510 + }, + { + "epoch": 1.78, + "learning_rate": 4.8265615524560344e-05, + "loss": 1.9574, + "step": 20520 + }, + { + "epoch": 1.78, + "learning_rate": 4.826474919864853e-05, + "loss": 1.9405, + "step": 20530 + }, + { + "epoch": 1.78, + "learning_rate": 4.8263882872736726e-05, + "loss": 1.9282, + "step": 20540 + }, + { + "epoch": 1.78, + "learning_rate": 4.826301654682492e-05, + "loss": 1.9335, + "step": 20550 + }, + { + "epoch": 1.78, + "learning_rate": 4.826215022091311e-05, + "loss": 1.9676, + "step": 20560 + }, + { + "epoch": 1.78, + "learning_rate": 4.82612838950013e-05, + "loss": 1.9114, + "step": 20570 + }, + { + "epoch": 1.78, + "learning_rate": 4.82604175690895e-05, + "loss": 1.949, + "step": 20580 + }, + { + "epoch": 1.78, + "learning_rate": 4.8259551243177684e-05, + "loss": 1.9156, + "step": 20590 + }, + { + "epoch": 1.78, + "learning_rate": 4.825868491726588e-05, + "loss": 1.9121, + "step": 20600 + }, + { + "epoch": 1.78, + "learning_rate": 4.825781859135407e-05, + "loss": 1.9371, + "step": 20610 + }, + { + "epoch": 1.78, + "learning_rate": 4.825695226544226e-05, + "loss": 1.9302, + "step": 20620 + }, + { + "epoch": 1.79, + "learning_rate": 4.8256085939530455e-05, + "loss": 1.91, + "step": 20630 + }, + { + "epoch": 1.79, + "learning_rate": 4.825521961361864e-05, + "loss": 1.9482, + "step": 20640 + }, + { + "epoch": 1.79, + "learning_rate": 4.825435328770684e-05, + "loss": 1.8984, + "step": 20650 + }, + { + "epoch": 1.79, + "learning_rate": 4.825348696179503e-05, + "loss": 1.904, + "step": 20660 + }, + { + "epoch": 1.79, + "learning_rate": 4.825262063588322e-05, + "loss": 1.9322, + "step": 20670 + }, + { + "epoch": 1.79, + "learning_rate": 4.8251754309971414e-05, + "loss": 1.9359, + "step": 20680 + }, + { + "epoch": 1.79, + "learning_rate": 4.825088798405961e-05, + "loss": 1.9453, + "step": 20690 + }, + { + "epoch": 1.79, + "learning_rate": 4.8250021658147796e-05, + "loss": 1.9927, + "step": 20700 + }, + { + "epoch": 1.79, + "learning_rate": 4.824915533223599e-05, + "loss": 1.9536, + "step": 20710 + }, + { + "epoch": 1.79, + "learning_rate": 4.8248289006324185e-05, + "loss": 1.936, + "step": 20720 + }, + { + "epoch": 1.79, + "learning_rate": 4.824742268041237e-05, + "loss": 1.8961, + "step": 20730 + }, + { + "epoch": 1.8, + "learning_rate": 4.824655635450057e-05, + "loss": 1.9205, + "step": 20740 + }, + { + "epoch": 1.8, + "learning_rate": 4.8245690028588755e-05, + "loss": 1.9433, + "step": 20750 + }, + { + "epoch": 1.8, + "learning_rate": 4.824482370267695e-05, + "loss": 1.9495, + "step": 20760 + }, + { + "epoch": 1.8, + "learning_rate": 4.8243957376765144e-05, + "loss": 1.9154, + "step": 20770 + }, + { + "epoch": 1.8, + "learning_rate": 4.824309105085333e-05, + "loss": 1.9793, + "step": 20780 + }, + { + "epoch": 1.8, + "learning_rate": 4.8242224724941526e-05, + "loss": 1.9089, + "step": 20790 + }, + { + "epoch": 1.8, + "learning_rate": 4.824135839902972e-05, + "loss": 1.93, + "step": 20800 + }, + { + "epoch": 1.8, + "learning_rate": 4.824049207311791e-05, + "loss": 1.9722, + "step": 20810 + }, + { + "epoch": 1.8, + "learning_rate": 4.82396257472061e-05, + "loss": 1.888, + "step": 20820 + }, + { + "epoch": 1.8, + "learning_rate": 4.82387594212943e-05, + "loss": 1.864, + "step": 20830 + }, + { + "epoch": 1.8, + "learning_rate": 4.8237893095382485e-05, + "loss": 1.8685, + "step": 20840 + }, + { + "epoch": 1.8, + "learning_rate": 4.823702676947068e-05, + "loss": 1.862, + "step": 20850 + }, + { + "epoch": 1.81, + "learning_rate": 4.823616044355887e-05, + "loss": 1.9062, + "step": 20860 + }, + { + "epoch": 1.81, + "learning_rate": 4.823529411764706e-05, + "loss": 1.9574, + "step": 20870 + }, + { + "epoch": 1.81, + "learning_rate": 4.8234427791735256e-05, + "loss": 1.9285, + "step": 20880 + }, + { + "epoch": 1.81, + "learning_rate": 4.8233561465823444e-05, + "loss": 1.932, + "step": 20890 + }, + { + "epoch": 1.81, + "learning_rate": 4.823269513991164e-05, + "loss": 1.9359, + "step": 20900 + }, + { + "epoch": 1.81, + "learning_rate": 4.823182881399983e-05, + "loss": 1.9184, + "step": 20910 + }, + { + "epoch": 1.81, + "learning_rate": 4.823096248808802e-05, + "loss": 1.8881, + "step": 20920 + }, + { + "epoch": 1.81, + "learning_rate": 4.8230096162176215e-05, + "loss": 1.951, + "step": 20930 + }, + { + "epoch": 1.81, + "learning_rate": 4.822922983626441e-05, + "loss": 1.9775, + "step": 20940 + }, + { + "epoch": 1.81, + "learning_rate": 4.82283635103526e-05, + "loss": 1.9241, + "step": 20950 + }, + { + "epoch": 1.81, + "learning_rate": 4.822749718444079e-05, + "loss": 1.9722, + "step": 20960 + }, + { + "epoch": 1.82, + "learning_rate": 4.822663085852898e-05, + "loss": 1.941, + "step": 20970 + }, + { + "epoch": 1.82, + "learning_rate": 4.822576453261717e-05, + "loss": 1.9105, + "step": 20980 + }, + { + "epoch": 1.82, + "learning_rate": 4.822489820670537e-05, + "loss": 1.8963, + "step": 20990 + }, + { + "epoch": 1.82, + "learning_rate": 4.8224031880793555e-05, + "loss": 1.9108, + "step": 21000 + }, + { + "epoch": 1.82, + "learning_rate": 4.822316555488175e-05, + "loss": 1.952, + "step": 21010 + }, + { + "epoch": 1.82, + "learning_rate": 4.8222299228969944e-05, + "loss": 1.8752, + "step": 21020 + }, + { + "epoch": 1.82, + "learning_rate": 4.822143290305813e-05, + "loss": 1.9202, + "step": 21030 + }, + { + "epoch": 1.82, + "learning_rate": 4.8220566577146326e-05, + "loss": 1.8668, + "step": 21040 + }, + { + "epoch": 1.82, + "learning_rate": 4.821970025123452e-05, + "loss": 1.937, + "step": 21050 + }, + { + "epoch": 1.82, + "learning_rate": 4.821883392532271e-05, + "loss": 1.8933, + "step": 21060 + }, + { + "epoch": 1.82, + "learning_rate": 4.82179675994109e-05, + "loss": 1.9364, + "step": 21070 + }, + { + "epoch": 1.82, + "learning_rate": 4.821710127349909e-05, + "loss": 1.8783, + "step": 21080 + }, + { + "epoch": 1.83, + "learning_rate": 4.8216234947587285e-05, + "loss": 1.942, + "step": 21090 + }, + { + "epoch": 1.83, + "learning_rate": 4.821536862167548e-05, + "loss": 1.946, + "step": 21100 + }, + { + "epoch": 1.83, + "learning_rate": 4.821450229576367e-05, + "loss": 1.9014, + "step": 21110 + }, + { + "epoch": 1.83, + "learning_rate": 4.821363596985186e-05, + "loss": 1.9171, + "step": 21120 + }, + { + "epoch": 1.83, + "learning_rate": 4.8212769643940056e-05, + "loss": 1.8947, + "step": 21130 + }, + { + "epoch": 1.83, + "learning_rate": 4.8211903318028244e-05, + "loss": 1.9401, + "step": 21140 + }, + { + "epoch": 1.83, + "learning_rate": 4.821103699211644e-05, + "loss": 1.9373, + "step": 21150 + }, + { + "epoch": 1.83, + "learning_rate": 4.8210170666204626e-05, + "loss": 1.9604, + "step": 21160 + }, + { + "epoch": 1.83, + "learning_rate": 4.820930434029282e-05, + "loss": 1.9297, + "step": 21170 + }, + { + "epoch": 1.83, + "learning_rate": 4.8208438014381015e-05, + "loss": 1.9085, + "step": 21180 + }, + { + "epoch": 1.83, + "learning_rate": 4.82075716884692e-05, + "loss": 1.9115, + "step": 21190 + }, + { + "epoch": 1.83, + "learning_rate": 4.82067053625574e-05, + "loss": 1.9179, + "step": 21200 + }, + { + "epoch": 1.84, + "learning_rate": 4.820583903664559e-05, + "loss": 2.0041, + "step": 21210 + }, + { + "epoch": 1.84, + "learning_rate": 4.820497271073378e-05, + "loss": 1.9406, + "step": 21220 + }, + { + "epoch": 1.84, + "learning_rate": 4.8204106384821974e-05, + "loss": 1.9166, + "step": 21230 + }, + { + "epoch": 1.84, + "learning_rate": 4.820324005891017e-05, + "loss": 1.8891, + "step": 21240 + }, + { + "epoch": 1.84, + "learning_rate": 4.8202373732998356e-05, + "loss": 1.9371, + "step": 21250 + }, + { + "epoch": 1.84, + "learning_rate": 4.820150740708655e-05, + "loss": 1.9362, + "step": 21260 + }, + { + "epoch": 1.84, + "learning_rate": 4.820064108117474e-05, + "loss": 1.9282, + "step": 21270 + }, + { + "epoch": 1.84, + "learning_rate": 4.819977475526293e-05, + "loss": 1.9029, + "step": 21280 + }, + { + "epoch": 1.84, + "learning_rate": 4.819890842935113e-05, + "loss": 1.9044, + "step": 21290 + }, + { + "epoch": 1.84, + "learning_rate": 4.8198042103439314e-05, + "loss": 1.9724, + "step": 21300 + }, + { + "epoch": 1.84, + "learning_rate": 4.819717577752751e-05, + "loss": 1.8677, + "step": 21310 + }, + { + "epoch": 1.85, + "learning_rate": 4.81963094516157e-05, + "loss": 1.9323, + "step": 21320 + }, + { + "epoch": 1.85, + "learning_rate": 4.819544312570389e-05, + "loss": 1.8683, + "step": 21330 + }, + { + "epoch": 1.85, + "learning_rate": 4.8194576799792085e-05, + "loss": 1.8965, + "step": 21340 + }, + { + "epoch": 1.85, + "learning_rate": 4.819371047388028e-05, + "loss": 1.9232, + "step": 21350 + }, + { + "epoch": 1.85, + "learning_rate": 4.819284414796847e-05, + "loss": 1.8713, + "step": 21360 + }, + { + "epoch": 1.85, + "learning_rate": 4.819197782205666e-05, + "loss": 1.92, + "step": 21370 + }, + { + "epoch": 1.85, + "learning_rate": 4.819111149614485e-05, + "loss": 1.9024, + "step": 21380 + }, + { + "epoch": 1.85, + "learning_rate": 4.8190245170233044e-05, + "loss": 1.9259, + "step": 21390 + }, + { + "epoch": 1.85, + "learning_rate": 4.818937884432124e-05, + "loss": 1.898, + "step": 21400 + }, + { + "epoch": 1.85, + "learning_rate": 4.8188512518409426e-05, + "loss": 1.8849, + "step": 21410 + }, + { + "epoch": 1.85, + "learning_rate": 4.818764619249762e-05, + "loss": 1.9251, + "step": 21420 + }, + { + "epoch": 1.85, + "learning_rate": 4.8186779866585815e-05, + "loss": 1.9074, + "step": 21430 + }, + { + "epoch": 1.86, + "learning_rate": 4.8185913540674e-05, + "loss": 1.9455, + "step": 21440 + }, + { + "epoch": 1.86, + "learning_rate": 4.81850472147622e-05, + "loss": 1.8941, + "step": 21450 + }, + { + "epoch": 1.86, + "learning_rate": 4.818418088885039e-05, + "loss": 1.8898, + "step": 21460 + }, + { + "epoch": 1.86, + "learning_rate": 4.818331456293858e-05, + "loss": 1.8987, + "step": 21470 + }, + { + "epoch": 1.86, + "learning_rate": 4.8182448237026774e-05, + "loss": 1.9447, + "step": 21480 + }, + { + "epoch": 1.86, + "learning_rate": 4.818158191111496e-05, + "loss": 1.9361, + "step": 21490 + }, + { + "epoch": 1.86, + "learning_rate": 4.8180715585203156e-05, + "loss": 1.8852, + "step": 21500 + }, + { + "epoch": 1.86, + "learning_rate": 4.817984925929135e-05, + "loss": 1.8928, + "step": 21510 + }, + { + "epoch": 1.86, + "learning_rate": 4.817898293337954e-05, + "loss": 1.8824, + "step": 21520 + }, + { + "epoch": 1.86, + "learning_rate": 4.817811660746773e-05, + "loss": 1.8756, + "step": 21530 + }, + { + "epoch": 1.86, + "learning_rate": 4.817725028155593e-05, + "loss": 1.9185, + "step": 21540 + }, + { + "epoch": 1.87, + "learning_rate": 4.8176383955644115e-05, + "loss": 1.9153, + "step": 21550 + }, + { + "epoch": 1.87, + "learning_rate": 4.817551762973231e-05, + "loss": 1.9046, + "step": 21560 + }, + { + "epoch": 1.87, + "learning_rate": 4.8174651303820504e-05, + "loss": 1.9181, + "step": 21570 + }, + { + "epoch": 1.87, + "learning_rate": 4.817378497790869e-05, + "loss": 1.9011, + "step": 21580 + }, + { + "epoch": 1.87, + "learning_rate": 4.8172918651996886e-05, + "loss": 1.8724, + "step": 21590 + }, + { + "epoch": 1.87, + "learning_rate": 4.8172052326085073e-05, + "loss": 1.9127, + "step": 21600 + }, + { + "epoch": 1.87, + "learning_rate": 4.817118600017327e-05, + "loss": 1.8927, + "step": 21610 + }, + { + "epoch": 1.87, + "learning_rate": 4.817031967426146e-05, + "loss": 1.9814, + "step": 21620 + }, + { + "epoch": 1.87, + "learning_rate": 4.816945334834965e-05, + "loss": 1.9092, + "step": 21630 + }, + { + "epoch": 1.87, + "learning_rate": 4.8168587022437845e-05, + "loss": 1.8949, + "step": 21640 + }, + { + "epoch": 1.87, + "learning_rate": 4.816772069652604e-05, + "loss": 1.8846, + "step": 21650 + }, + { + "epoch": 1.87, + "learning_rate": 4.816685437061423e-05, + "loss": 1.8944, + "step": 21660 + }, + { + "epoch": 1.88, + "learning_rate": 4.816598804470242e-05, + "loss": 1.9458, + "step": 21670 + }, + { + "epoch": 1.88, + "learning_rate": 4.8165121718790616e-05, + "loss": 1.9466, + "step": 21680 + }, + { + "epoch": 1.88, + "learning_rate": 4.81642553928788e-05, + "loss": 1.9038, + "step": 21690 + }, + { + "epoch": 1.88, + "learning_rate": 4.8163389066967e-05, + "loss": 1.9361, + "step": 21700 + }, + { + "epoch": 1.88, + "learning_rate": 4.8162522741055185e-05, + "loss": 1.9077, + "step": 21710 + }, + { + "epoch": 1.88, + "learning_rate": 4.816165641514338e-05, + "loss": 1.8997, + "step": 21720 + }, + { + "epoch": 1.88, + "learning_rate": 4.8160790089231574e-05, + "loss": 1.8395, + "step": 21730 + }, + { + "epoch": 1.88, + "learning_rate": 4.815992376331976e-05, + "loss": 1.8707, + "step": 21740 + }, + { + "epoch": 1.88, + "learning_rate": 4.8159057437407956e-05, + "loss": 1.8757, + "step": 21750 + }, + { + "epoch": 1.88, + "learning_rate": 4.815819111149615e-05, + "loss": 1.9209, + "step": 21760 + }, + { + "epoch": 1.88, + "learning_rate": 4.815732478558434e-05, + "loss": 1.923, + "step": 21770 + }, + { + "epoch": 1.89, + "learning_rate": 4.815645845967253e-05, + "loss": 1.9123, + "step": 21780 + }, + { + "epoch": 1.89, + "learning_rate": 4.815559213376073e-05, + "loss": 1.9545, + "step": 21790 + }, + { + "epoch": 1.89, + "learning_rate": 4.8154725807848915e-05, + "loss": 1.9034, + "step": 21800 + }, + { + "epoch": 1.89, + "learning_rate": 4.815385948193711e-05, + "loss": 1.9223, + "step": 21810 + }, + { + "epoch": 1.89, + "learning_rate": 4.81529931560253e-05, + "loss": 1.8946, + "step": 21820 + }, + { + "epoch": 1.89, + "learning_rate": 4.815212683011349e-05, + "loss": 1.9779, + "step": 21830 + }, + { + "epoch": 1.89, + "learning_rate": 4.8151260504201686e-05, + "loss": 1.9188, + "step": 21840 + }, + { + "epoch": 1.89, + "learning_rate": 4.8150394178289874e-05, + "loss": 1.9052, + "step": 21850 + }, + { + "epoch": 1.89, + "learning_rate": 4.814952785237807e-05, + "loss": 1.833, + "step": 21860 + }, + { + "epoch": 1.89, + "learning_rate": 4.814866152646626e-05, + "loss": 1.8839, + "step": 21870 + }, + { + "epoch": 1.89, + "learning_rate": 4.814779520055445e-05, + "loss": 1.9023, + "step": 21880 + }, + { + "epoch": 1.89, + "learning_rate": 4.8146928874642645e-05, + "loss": 1.8833, + "step": 21890 + }, + { + "epoch": 1.9, + "learning_rate": 4.814606254873083e-05, + "loss": 1.9067, + "step": 21900 + }, + { + "epoch": 1.9, + "learning_rate": 4.814519622281903e-05, + "loss": 1.8445, + "step": 21910 + }, + { + "epoch": 1.9, + "learning_rate": 4.814432989690722e-05, + "loss": 1.9482, + "step": 21920 + }, + { + "epoch": 1.9, + "learning_rate": 4.814346357099541e-05, + "loss": 1.8806, + "step": 21930 + }, + { + "epoch": 1.9, + "learning_rate": 4.8142597245083604e-05, + "loss": 1.9039, + "step": 21940 + }, + { + "epoch": 1.9, + "learning_rate": 4.81417309191718e-05, + "loss": 1.9382, + "step": 21950 + }, + { + "epoch": 1.9, + "learning_rate": 4.8140864593259986e-05, + "loss": 1.95, + "step": 21960 + }, + { + "epoch": 1.9, + "learning_rate": 4.813999826734818e-05, + "loss": 1.8655, + "step": 21970 + }, + { + "epoch": 1.9, + "learning_rate": 4.8139131941436375e-05, + "loss": 1.9763, + "step": 21980 + }, + { + "epoch": 1.9, + "learning_rate": 4.813826561552456e-05, + "loss": 1.9016, + "step": 21990 + }, + { + "epoch": 1.9, + "learning_rate": 4.813739928961276e-05, + "loss": 1.8802, + "step": 22000 + }, + { + "epoch": 1.91, + "learning_rate": 4.8136532963700944e-05, + "loss": 1.8835, + "step": 22010 + }, + { + "epoch": 1.91, + "learning_rate": 4.813566663778914e-05, + "loss": 1.9511, + "step": 22020 + }, + { + "epoch": 1.91, + "learning_rate": 4.813480031187733e-05, + "loss": 1.925, + "step": 22030 + }, + { + "epoch": 1.91, + "learning_rate": 4.813393398596552e-05, + "loss": 1.8966, + "step": 22040 + }, + { + "epoch": 1.91, + "learning_rate": 4.8133067660053715e-05, + "loss": 1.887, + "step": 22050 + }, + { + "epoch": 1.91, + "learning_rate": 4.813220133414191e-05, + "loss": 1.9234, + "step": 22060 + }, + { + "epoch": 1.91, + "learning_rate": 4.81313350082301e-05, + "loss": 1.8813, + "step": 22070 + }, + { + "epoch": 1.91, + "learning_rate": 4.813046868231829e-05, + "loss": 1.9236, + "step": 22080 + }, + { + "epoch": 1.91, + "learning_rate": 4.8129602356406486e-05, + "loss": 1.9132, + "step": 22090 + }, + { + "epoch": 1.91, + "learning_rate": 4.8128736030494674e-05, + "loss": 1.8998, + "step": 22100 + }, + { + "epoch": 1.91, + "learning_rate": 4.812786970458287e-05, + "loss": 1.9202, + "step": 22110 + }, + { + "epoch": 1.91, + "learning_rate": 4.8127003378671056e-05, + "loss": 1.9419, + "step": 22120 + }, + { + "epoch": 1.92, + "learning_rate": 4.812613705275925e-05, + "loss": 1.886, + "step": 22130 + }, + { + "epoch": 1.92, + "learning_rate": 4.8125270726847445e-05, + "loss": 1.9351, + "step": 22140 + }, + { + "epoch": 1.92, + "learning_rate": 4.812440440093563e-05, + "loss": 1.9313, + "step": 22150 + }, + { + "epoch": 1.92, + "learning_rate": 4.812353807502383e-05, + "loss": 1.8668, + "step": 22160 + }, + { + "epoch": 1.92, + "learning_rate": 4.812267174911202e-05, + "loss": 1.9076, + "step": 22170 + }, + { + "epoch": 1.92, + "learning_rate": 4.812180542320021e-05, + "loss": 1.9329, + "step": 22180 + }, + { + "epoch": 1.92, + "learning_rate": 4.8120939097288404e-05, + "loss": 1.9271, + "step": 22190 + }, + { + "epoch": 1.92, + "learning_rate": 4.81200727713766e-05, + "loss": 1.8613, + "step": 22200 + }, + { + "epoch": 1.92, + "learning_rate": 4.8119206445464786e-05, + "loss": 1.9215, + "step": 22210 + }, + { + "epoch": 1.92, + "learning_rate": 4.811834011955298e-05, + "loss": 1.8789, + "step": 22220 + }, + { + "epoch": 1.92, + "learning_rate": 4.811747379364117e-05, + "loss": 1.9547, + "step": 22230 + }, + { + "epoch": 1.92, + "learning_rate": 4.811660746772936e-05, + "loss": 1.8768, + "step": 22240 + }, + { + "epoch": 1.93, + "learning_rate": 4.811574114181756e-05, + "loss": 1.8924, + "step": 22250 + }, + { + "epoch": 1.93, + "learning_rate": 4.8114874815905745e-05, + "loss": 1.8971, + "step": 22260 + }, + { + "epoch": 1.93, + "learning_rate": 4.811400848999394e-05, + "loss": 1.948, + "step": 22270 + }, + { + "epoch": 1.93, + "learning_rate": 4.8113142164082134e-05, + "loss": 1.9107, + "step": 22280 + }, + { + "epoch": 1.93, + "learning_rate": 4.811227583817032e-05, + "loss": 1.9337, + "step": 22290 + }, + { + "epoch": 1.93, + "learning_rate": 4.8111409512258516e-05, + "loss": 1.8786, + "step": 22300 + }, + { + "epoch": 1.93, + "learning_rate": 4.811054318634671e-05, + "loss": 1.7759, + "step": 22310 + }, + { + "epoch": 1.93, + "learning_rate": 4.81096768604349e-05, + "loss": 1.8902, + "step": 22320 + }, + { + "epoch": 1.93, + "learning_rate": 4.810881053452309e-05, + "loss": 1.9142, + "step": 22330 + }, + { + "epoch": 1.93, + "learning_rate": 4.810794420861128e-05, + "loss": 1.9162, + "step": 22340 + }, + { + "epoch": 1.93, + "learning_rate": 4.8107077882699474e-05, + "loss": 1.9187, + "step": 22350 + }, + { + "epoch": 1.94, + "learning_rate": 4.810621155678767e-05, + "loss": 1.9076, + "step": 22360 + }, + { + "epoch": 1.94, + "learning_rate": 4.8105345230875857e-05, + "loss": 1.869, + "step": 22370 + }, + { + "epoch": 1.94, + "learning_rate": 4.810447890496405e-05, + "loss": 1.8996, + "step": 22380 + }, + { + "epoch": 1.94, + "learning_rate": 4.8103612579052246e-05, + "loss": 1.8962, + "step": 22390 + }, + { + "epoch": 1.94, + "learning_rate": 4.810274625314043e-05, + "loss": 1.9662, + "step": 22400 + }, + { + "epoch": 1.94, + "learning_rate": 4.810187992722863e-05, + "loss": 1.8901, + "step": 22410 + }, + { + "epoch": 1.94, + "learning_rate": 4.810101360131682e-05, + "loss": 1.9419, + "step": 22420 + }, + { + "epoch": 1.94, + "learning_rate": 4.810014727540501e-05, + "loss": 1.9393, + "step": 22430 + }, + { + "epoch": 1.94, + "learning_rate": 4.8099280949493204e-05, + "loss": 1.889, + "step": 22440 + }, + { + "epoch": 1.94, + "learning_rate": 4.809841462358139e-05, + "loss": 1.9347, + "step": 22450 + }, + { + "epoch": 1.94, + "learning_rate": 4.8097548297669586e-05, + "loss": 1.8927, + "step": 22460 + }, + { + "epoch": 1.94, + "learning_rate": 4.809668197175778e-05, + "loss": 1.9038, + "step": 22470 + }, + { + "epoch": 1.95, + "learning_rate": 4.809581564584597e-05, + "loss": 1.9116, + "step": 22480 + }, + { + "epoch": 1.95, + "learning_rate": 4.809494931993416e-05, + "loss": 1.9095, + "step": 22490 + }, + { + "epoch": 1.95, + "learning_rate": 4.809408299402236e-05, + "loss": 1.8713, + "step": 22500 + }, + { + "epoch": 1.95, + "learning_rate": 4.8093216668110545e-05, + "loss": 1.9105, + "step": 22510 + }, + { + "epoch": 1.95, + "learning_rate": 4.809235034219874e-05, + "loss": 1.8684, + "step": 22520 + }, + { + "epoch": 1.95, + "learning_rate": 4.8091484016286934e-05, + "loss": 1.9547, + "step": 22530 + }, + { + "epoch": 1.95, + "learning_rate": 4.809061769037512e-05, + "loss": 1.8942, + "step": 22540 + }, + { + "epoch": 1.95, + "learning_rate": 4.8089751364463316e-05, + "loss": 1.9416, + "step": 22550 + }, + { + "epoch": 1.95, + "learning_rate": 4.8088885038551504e-05, + "loss": 1.8473, + "step": 22560 + }, + { + "epoch": 1.95, + "learning_rate": 4.80880187126397e-05, + "loss": 1.9233, + "step": 22570 + }, + { + "epoch": 1.95, + "learning_rate": 4.808715238672789e-05, + "loss": 1.8986, + "step": 22580 + }, + { + "epoch": 1.96, + "learning_rate": 4.808628606081608e-05, + "loss": 1.8729, + "step": 22590 + }, + { + "epoch": 1.96, + "learning_rate": 4.8085419734904275e-05, + "loss": 1.8886, + "step": 22600 + }, + { + "epoch": 1.96, + "learning_rate": 4.808455340899247e-05, + "loss": 1.9503, + "step": 22610 + }, + { + "epoch": 1.96, + "learning_rate": 4.808368708308066e-05, + "loss": 1.9175, + "step": 22620 + }, + { + "epoch": 1.96, + "learning_rate": 4.808282075716885e-05, + "loss": 1.895, + "step": 22630 + }, + { + "epoch": 1.96, + "learning_rate": 4.8081954431257046e-05, + "loss": 1.8761, + "step": 22640 + }, + { + "epoch": 1.96, + "learning_rate": 4.8081088105345234e-05, + "loss": 1.8742, + "step": 22650 + }, + { + "epoch": 1.96, + "learning_rate": 4.808022177943343e-05, + "loss": 1.8799, + "step": 22660 + }, + { + "epoch": 1.96, + "learning_rate": 4.8079355453521616e-05, + "loss": 1.937, + "step": 22670 + }, + { + "epoch": 1.96, + "learning_rate": 4.807848912760981e-05, + "loss": 1.8575, + "step": 22680 + }, + { + "epoch": 1.96, + "learning_rate": 4.8077622801698005e-05, + "loss": 1.8747, + "step": 22690 + }, + { + "epoch": 1.96, + "learning_rate": 4.807675647578619e-05, + "loss": 1.9347, + "step": 22700 + }, + { + "epoch": 1.97, + "learning_rate": 4.807589014987439e-05, + "loss": 1.8716, + "step": 22710 + }, + { + "epoch": 1.97, + "learning_rate": 4.807502382396258e-05, + "loss": 1.9225, + "step": 22720 + }, + { + "epoch": 1.97, + "learning_rate": 4.807415749805077e-05, + "loss": 1.8568, + "step": 22730 + }, + { + "epoch": 1.97, + "learning_rate": 4.807329117213896e-05, + "loss": 1.901, + "step": 22740 + }, + { + "epoch": 1.97, + "learning_rate": 4.807242484622715e-05, + "loss": 1.8799, + "step": 22750 + }, + { + "epoch": 1.97, + "learning_rate": 4.8071558520315345e-05, + "loss": 1.9426, + "step": 22760 + }, + { + "epoch": 1.97, + "learning_rate": 4.807069219440354e-05, + "loss": 1.8674, + "step": 22770 + }, + { + "epoch": 1.97, + "learning_rate": 4.806982586849173e-05, + "loss": 1.9099, + "step": 22780 + }, + { + "epoch": 1.97, + "learning_rate": 4.806895954257992e-05, + "loss": 1.9103, + "step": 22790 + }, + { + "epoch": 1.97, + "learning_rate": 4.8068093216668116e-05, + "loss": 1.8566, + "step": 22800 + }, + { + "epoch": 1.97, + "learning_rate": 4.8067226890756304e-05, + "loss": 1.9247, + "step": 22810 + }, + { + "epoch": 1.98, + "learning_rate": 4.80663605648445e-05, + "loss": 1.8922, + "step": 22820 + }, + { + "epoch": 1.98, + "learning_rate": 4.806549423893269e-05, + "loss": 1.9305, + "step": 22830 + }, + { + "epoch": 1.98, + "learning_rate": 4.806462791302088e-05, + "loss": 1.8735, + "step": 22840 + }, + { + "epoch": 1.98, + "learning_rate": 4.8063761587109075e-05, + "loss": 1.8955, + "step": 22850 + }, + { + "epoch": 1.98, + "learning_rate": 4.806289526119726e-05, + "loss": 1.8936, + "step": 22860 + }, + { + "epoch": 1.98, + "learning_rate": 4.806202893528546e-05, + "loss": 1.9201, + "step": 22870 + }, + { + "epoch": 1.98, + "learning_rate": 4.806116260937365e-05, + "loss": 1.868, + "step": 22880 + }, + { + "epoch": 1.98, + "learning_rate": 4.806029628346184e-05, + "loss": 1.9564, + "step": 22890 + }, + { + "epoch": 1.98, + "learning_rate": 4.8059429957550034e-05, + "loss": 1.9399, + "step": 22900 + }, + { + "epoch": 1.98, + "learning_rate": 4.805856363163823e-05, + "loss": 1.8774, + "step": 22910 + }, + { + "epoch": 1.98, + "learning_rate": 4.8057697305726416e-05, + "loss": 1.917, + "step": 22920 + }, + { + "epoch": 1.98, + "learning_rate": 4.805683097981461e-05, + "loss": 1.9723, + "step": 22930 + }, + { + "epoch": 1.99, + "learning_rate": 4.8055964653902805e-05, + "loss": 1.8639, + "step": 22940 + }, + { + "epoch": 1.99, + "learning_rate": 4.805509832799099e-05, + "loss": 1.894, + "step": 22950 + }, + { + "epoch": 1.99, + "learning_rate": 4.805423200207919e-05, + "loss": 1.8626, + "step": 22960 + }, + { + "epoch": 1.99, + "learning_rate": 4.8053365676167375e-05, + "loss": 1.9128, + "step": 22970 + }, + { + "epoch": 1.99, + "learning_rate": 4.805249935025557e-05, + "loss": 1.8663, + "step": 22980 + }, + { + "epoch": 1.99, + "learning_rate": 4.8051633024343764e-05, + "loss": 1.8934, + "step": 22990 + }, + { + "epoch": 1.99, + "learning_rate": 4.805076669843195e-05, + "loss": 1.8555, + "step": 23000 + }, + { + "epoch": 1.99, + "learning_rate": 4.8049900372520146e-05, + "loss": 1.903, + "step": 23010 + }, + { + "epoch": 1.99, + "learning_rate": 4.804903404660834e-05, + "loss": 1.8617, + "step": 23020 + }, + { + "epoch": 1.99, + "learning_rate": 4.804816772069653e-05, + "loss": 1.8195, + "step": 23030 + }, + { + "epoch": 1.99, + "learning_rate": 4.804730139478472e-05, + "loss": 1.9416, + "step": 23040 + }, + { + "epoch": 2.0, + "learning_rate": 4.804643506887292e-05, + "loss": 1.8234, + "step": 23050 + }, + { + "epoch": 2.0, + "learning_rate": 4.8045568742961104e-05, + "loss": 1.8917, + "step": 23060 + }, + { + "epoch": 2.0, + "learning_rate": 4.80447024170493e-05, + "loss": 1.8667, + "step": 23070 + }, + { + "epoch": 2.0, + "learning_rate": 4.8043836091137487e-05, + "loss": 1.8599, + "step": 23080 + }, + { + "epoch": 2.0, + "learning_rate": 4.804296976522568e-05, + "loss": 1.87, + "step": 23090 + }, + { + "epoch": 2.0, + "learning_rate": 4.8042103439313875e-05, + "loss": 1.9316, + "step": 23100 + }, + { + "epoch": 2.0, + "eval_Bleu_1": 0.012378275929677636, + "eval_Bleu_2": 1.3169589296668432e-11, + "eval_Bleu_3": 1.3783271388657373e-14, + "eval_Bleu_4": 4.549921752983273e-16, + "eval_ROUGE_L": 0.03495490090149432, + "eval_cer": 0.9929244896780421, + "eval_em": 0, + "eval_f1": 0.059610855566011124, + "eval_loss": 1.7868506908416748, + "eval_runtime": 1935.234, + "eval_samples_per_second": 2.653, + "eval_steps_per_second": 2.653, + "eval_wer": 0.9730054485222841, + "step": 23107 + }, + { + "epoch": 2.0, + "learning_rate": 4.804123711340206e-05, + "loss": 1.8756, + "step": 23110 + }, + { + "epoch": 2.0, + "learning_rate": 4.804037078749026e-05, + "loss": 1.8417, + "step": 23120 + }, + { + "epoch": 2.0, + "learning_rate": 4.803950446157845e-05, + "loss": 1.8636, + "step": 23130 + }, + { + "epoch": 2.0, + "learning_rate": 4.803863813566664e-05, + "loss": 1.8875, + "step": 23140 + }, + { + "epoch": 2.0, + "learning_rate": 4.8037771809754834e-05, + "loss": 1.8452, + "step": 23150 + }, + { + "epoch": 2.0, + "learning_rate": 4.803690548384303e-05, + "loss": 1.9503, + "step": 23160 + }, + { + "epoch": 2.01, + "learning_rate": 4.8036039157931216e-05, + "loss": 1.9037, + "step": 23170 + }, + { + "epoch": 2.01, + "learning_rate": 4.803517283201941e-05, + "loss": 1.9001, + "step": 23180 + }, + { + "epoch": 2.01, + "learning_rate": 4.80343065061076e-05, + "loss": 1.8355, + "step": 23190 + }, + { + "epoch": 2.01, + "learning_rate": 4.803344018019579e-05, + "loss": 1.8729, + "step": 23200 + }, + { + "epoch": 2.01, + "learning_rate": 4.803257385428399e-05, + "loss": 1.7877, + "step": 23210 + }, + { + "epoch": 2.01, + "learning_rate": 4.8031707528372175e-05, + "loss": 1.8853, + "step": 23220 + }, + { + "epoch": 2.01, + "learning_rate": 4.803084120246037e-05, + "loss": 1.8542, + "step": 23230 + }, + { + "epoch": 2.01, + "learning_rate": 4.8029974876548564e-05, + "loss": 1.8752, + "step": 23240 + }, + { + "epoch": 2.01, + "learning_rate": 4.802910855063675e-05, + "loss": 1.8879, + "step": 23250 + }, + { + "epoch": 2.01, + "learning_rate": 4.8028242224724946e-05, + "loss": 1.8651, + "step": 23260 + }, + { + "epoch": 2.01, + "learning_rate": 4.802737589881314e-05, + "loss": 1.8526, + "step": 23270 + }, + { + "epoch": 2.01, + "learning_rate": 4.802650957290133e-05, + "loss": 1.8738, + "step": 23280 + }, + { + "epoch": 2.02, + "learning_rate": 4.802564324698952e-05, + "loss": 1.9095, + "step": 23290 + }, + { + "epoch": 2.02, + "learning_rate": 4.802477692107771e-05, + "loss": 1.845, + "step": 23300 + }, + { + "epoch": 2.02, + "learning_rate": 4.8023910595165905e-05, + "loss": 1.9177, + "step": 23310 + }, + { + "epoch": 2.02, + "learning_rate": 4.80230442692541e-05, + "loss": 1.9341, + "step": 23320 + }, + { + "epoch": 2.02, + "learning_rate": 4.802217794334229e-05, + "loss": 1.8666, + "step": 23330 + }, + { + "epoch": 2.02, + "learning_rate": 4.802131161743048e-05, + "loss": 1.7993, + "step": 23340 + }, + { + "epoch": 2.02, + "learning_rate": 4.8020445291518676e-05, + "loss": 1.9133, + "step": 23350 + }, + { + "epoch": 2.02, + "learning_rate": 4.8019578965606864e-05, + "loss": 1.8599, + "step": 23360 + }, + { + "epoch": 2.02, + "learning_rate": 4.801871263969506e-05, + "loss": 1.9444, + "step": 23370 + }, + { + "epoch": 2.02, + "learning_rate": 4.801784631378325e-05, + "loss": 1.8979, + "step": 23380 + }, + { + "epoch": 2.02, + "learning_rate": 4.801697998787144e-05, + "loss": 1.8218, + "step": 23390 + }, + { + "epoch": 2.03, + "learning_rate": 4.8016113661959635e-05, + "loss": 1.8968, + "step": 23400 + }, + { + "epoch": 2.03, + "learning_rate": 4.801524733604782e-05, + "loss": 1.9142, + "step": 23410 + }, + { + "epoch": 2.03, + "learning_rate": 4.801438101013602e-05, + "loss": 1.8197, + "step": 23420 + }, + { + "epoch": 2.03, + "learning_rate": 4.801351468422421e-05, + "loss": 1.8669, + "step": 23430 + }, + { + "epoch": 2.03, + "learning_rate": 4.80126483583124e-05, + "loss": 1.8778, + "step": 23440 + }, + { + "epoch": 2.03, + "learning_rate": 4.801178203240059e-05, + "loss": 1.8791, + "step": 23450 + }, + { + "epoch": 2.03, + "learning_rate": 4.801091570648879e-05, + "loss": 1.8941, + "step": 23460 + }, + { + "epoch": 2.03, + "learning_rate": 4.8010049380576975e-05, + "loss": 1.8396, + "step": 23470 + }, + { + "epoch": 2.03, + "learning_rate": 4.800918305466517e-05, + "loss": 1.8969, + "step": 23480 + }, + { + "epoch": 2.03, + "learning_rate": 4.800831672875336e-05, + "loss": 1.8792, + "step": 23490 + }, + { + "epoch": 2.03, + "learning_rate": 4.800745040284155e-05, + "loss": 1.8768, + "step": 23500 + }, + { + "epoch": 2.03, + "learning_rate": 4.8006584076929746e-05, + "loss": 1.7999, + "step": 23510 + }, + { + "epoch": 2.04, + "learning_rate": 4.8005717751017934e-05, + "loss": 1.9331, + "step": 23520 + }, + { + "epoch": 2.04, + "learning_rate": 4.800485142510613e-05, + "loss": 1.857, + "step": 23530 + }, + { + "epoch": 2.04, + "learning_rate": 4.800398509919432e-05, + "loss": 1.8443, + "step": 23540 + }, + { + "epoch": 2.04, + "learning_rate": 4.800311877328251e-05, + "loss": 1.9244, + "step": 23550 + }, + { + "epoch": 2.04, + "learning_rate": 4.8002252447370705e-05, + "loss": 1.8123, + "step": 23560 + }, + { + "epoch": 2.04, + "learning_rate": 4.80013861214589e-05, + "loss": 1.9334, + "step": 23570 + }, + { + "epoch": 2.04, + "learning_rate": 4.800051979554709e-05, + "loss": 1.8773, + "step": 23580 + }, + { + "epoch": 2.04, + "learning_rate": 4.799965346963528e-05, + "loss": 1.8747, + "step": 23590 + }, + { + "epoch": 2.04, + "learning_rate": 4.799878714372347e-05, + "loss": 1.8605, + "step": 23600 + }, + { + "epoch": 2.04, + "learning_rate": 4.7997920817811664e-05, + "loss": 1.8996, + "step": 23610 + }, + { + "epoch": 2.04, + "learning_rate": 4.799705449189986e-05, + "loss": 1.8707, + "step": 23620 + }, + { + "epoch": 2.05, + "learning_rate": 4.7996188165988046e-05, + "loss": 1.8539, + "step": 23630 + }, + { + "epoch": 2.05, + "learning_rate": 4.799532184007624e-05, + "loss": 1.8697, + "step": 23640 + }, + { + "epoch": 2.05, + "learning_rate": 4.7994455514164435e-05, + "loss": 1.8892, + "step": 23650 + }, + { + "epoch": 2.05, + "learning_rate": 4.799358918825262e-05, + "loss": 1.8795, + "step": 23660 + }, + { + "epoch": 2.05, + "learning_rate": 4.799272286234082e-05, + "loss": 1.8137, + "step": 23670 + }, + { + "epoch": 2.05, + "learning_rate": 4.799185653642901e-05, + "loss": 1.8754, + "step": 23680 + }, + { + "epoch": 2.05, + "learning_rate": 4.79909902105172e-05, + "loss": 1.9183, + "step": 23690 + }, + { + "epoch": 2.05, + "learning_rate": 4.7990123884605394e-05, + "loss": 1.8638, + "step": 23700 + }, + { + "epoch": 2.05, + "learning_rate": 4.798925755869358e-05, + "loss": 1.9179, + "step": 23710 + }, + { + "epoch": 2.05, + "learning_rate": 4.7988391232781776e-05, + "loss": 1.8768, + "step": 23720 + }, + { + "epoch": 2.05, + "learning_rate": 4.798752490686997e-05, + "loss": 1.8879, + "step": 23730 + }, + { + "epoch": 2.05, + "learning_rate": 4.798665858095816e-05, + "loss": 1.8516, + "step": 23740 + }, + { + "epoch": 2.06, + "learning_rate": 4.798579225504635e-05, + "loss": 1.8805, + "step": 23750 + }, + { + "epoch": 2.06, + "learning_rate": 4.798492592913455e-05, + "loss": 1.8656, + "step": 23760 + }, + { + "epoch": 2.06, + "learning_rate": 4.7984059603222734e-05, + "loss": 1.8839, + "step": 23770 + }, + { + "epoch": 2.06, + "learning_rate": 4.798319327731093e-05, + "loss": 1.9173, + "step": 23780 + }, + { + "epoch": 2.06, + "learning_rate": 4.798232695139912e-05, + "loss": 1.8641, + "step": 23790 + }, + { + "epoch": 2.06, + "learning_rate": 4.798146062548731e-05, + "loss": 1.851, + "step": 23800 + }, + { + "epoch": 2.06, + "learning_rate": 4.7980594299575505e-05, + "loss": 1.9339, + "step": 23810 + }, + { + "epoch": 2.06, + "learning_rate": 4.797972797366369e-05, + "loss": 1.8125, + "step": 23820 + }, + { + "epoch": 2.06, + "learning_rate": 4.797886164775189e-05, + "loss": 1.8488, + "step": 23830 + }, + { + "epoch": 2.06, + "learning_rate": 4.797799532184008e-05, + "loss": 1.8346, + "step": 23840 + }, + { + "epoch": 2.06, + "learning_rate": 4.797712899592827e-05, + "loss": 1.8589, + "step": 23850 + }, + { + "epoch": 2.07, + "learning_rate": 4.7976262670016464e-05, + "loss": 1.897, + "step": 23860 + }, + { + "epoch": 2.07, + "learning_rate": 4.797539634410466e-05, + "loss": 1.8968, + "step": 23870 + }, + { + "epoch": 2.07, + "learning_rate": 4.7974530018192846e-05, + "loss": 1.9002, + "step": 23880 + }, + { + "epoch": 2.07, + "learning_rate": 4.797366369228104e-05, + "loss": 1.9094, + "step": 23890 + }, + { + "epoch": 2.07, + "learning_rate": 4.7972797366369235e-05, + "loss": 1.8681, + "step": 23900 + }, + { + "epoch": 2.07, + "learning_rate": 4.797193104045742e-05, + "loss": 1.9064, + "step": 23910 + }, + { + "epoch": 2.07, + "learning_rate": 4.797106471454562e-05, + "loss": 1.935, + "step": 23920 + }, + { + "epoch": 2.07, + "learning_rate": 4.7970198388633805e-05, + "loss": 1.864, + "step": 23930 + }, + { + "epoch": 2.07, + "learning_rate": 4.7969332062722e-05, + "loss": 1.8493, + "step": 23940 + }, + { + "epoch": 2.07, + "learning_rate": 4.7968465736810194e-05, + "loss": 1.8477, + "step": 23950 + }, + { + "epoch": 2.07, + "learning_rate": 4.796759941089838e-05, + "loss": 1.8711, + "step": 23960 + }, + { + "epoch": 2.07, + "learning_rate": 4.7966733084986576e-05, + "loss": 1.838, + "step": 23970 + }, + { + "epoch": 2.08, + "learning_rate": 4.796586675907477e-05, + "loss": 1.8443, + "step": 23980 + }, + { + "epoch": 2.08, + "learning_rate": 4.796500043316296e-05, + "loss": 1.912, + "step": 23990 + }, + { + "epoch": 2.08, + "learning_rate": 4.796413410725115e-05, + "loss": 1.8401, + "step": 24000 + }, + { + "epoch": 2.08, + "learning_rate": 4.796326778133935e-05, + "loss": 1.8694, + "step": 24010 + }, + { + "epoch": 2.08, + "learning_rate": 4.7962401455427535e-05, + "loss": 1.9219, + "step": 24020 + }, + { + "epoch": 2.08, + "learning_rate": 4.796153512951573e-05, + "loss": 1.8819, + "step": 24030 + }, + { + "epoch": 2.08, + "learning_rate": 4.796066880360392e-05, + "loss": 1.8632, + "step": 24040 + }, + { + "epoch": 2.08, + "learning_rate": 4.795980247769211e-05, + "loss": 1.848, + "step": 24050 + }, + { + "epoch": 2.08, + "learning_rate": 4.7958936151780306e-05, + "loss": 1.8959, + "step": 24060 + }, + { + "epoch": 2.08, + "learning_rate": 4.7958069825868493e-05, + "loss": 1.895, + "step": 24070 + }, + { + "epoch": 2.08, + "learning_rate": 4.795720349995669e-05, + "loss": 1.8611, + "step": 24080 + }, + { + "epoch": 2.09, + "learning_rate": 4.795633717404488e-05, + "loss": 1.8763, + "step": 24090 + }, + { + "epoch": 2.09, + "learning_rate": 4.795547084813307e-05, + "loss": 1.8626, + "step": 24100 + }, + { + "epoch": 2.09, + "learning_rate": 4.7954604522221265e-05, + "loss": 1.8669, + "step": 24110 + }, + { + "epoch": 2.09, + "learning_rate": 4.795373819630946e-05, + "loss": 1.8246, + "step": 24120 + }, + { + "epoch": 2.09, + "learning_rate": 4.795287187039765e-05, + "loss": 1.9042, + "step": 24130 + }, + { + "epoch": 2.09, + "learning_rate": 4.795200554448584e-05, + "loss": 1.896, + "step": 24140 + }, + { + "epoch": 2.09, + "learning_rate": 4.795113921857403e-05, + "loss": 1.8926, + "step": 24150 + }, + { + "epoch": 2.09, + "learning_rate": 4.795027289266222e-05, + "loss": 1.8018, + "step": 24160 + }, + { + "epoch": 2.09, + "learning_rate": 4.794940656675042e-05, + "loss": 1.8776, + "step": 24170 + }, + { + "epoch": 2.09, + "learning_rate": 4.7948540240838605e-05, + "loss": 1.8367, + "step": 24180 + }, + { + "epoch": 2.09, + "learning_rate": 4.79476739149268e-05, + "loss": 1.8641, + "step": 24190 + }, + { + "epoch": 2.09, + "learning_rate": 4.7946807589014994e-05, + "loss": 1.8373, + "step": 24200 + }, + { + "epoch": 2.1, + "learning_rate": 4.794594126310318e-05, + "loss": 1.8488, + "step": 24210 + }, + { + "epoch": 2.1, + "learning_rate": 4.7945074937191376e-05, + "loss": 1.8548, + "step": 24220 + }, + { + "epoch": 2.1, + "learning_rate": 4.7944208611279564e-05, + "loss": 1.8725, + "step": 24230 + }, + { + "epoch": 2.1, + "learning_rate": 4.794334228536776e-05, + "loss": 1.8992, + "step": 24240 + }, + { + "epoch": 2.1, + "learning_rate": 4.794247595945595e-05, + "loss": 1.8772, + "step": 24250 + }, + { + "epoch": 2.1, + "learning_rate": 4.794160963354414e-05, + "loss": 1.8283, + "step": 24260 + }, + { + "epoch": 2.1, + "learning_rate": 4.7940743307632335e-05, + "loss": 1.871, + "step": 24270 + }, + { + "epoch": 2.1, + "learning_rate": 4.793987698172053e-05, + "loss": 1.8577, + "step": 24280 + }, + { + "epoch": 2.1, + "learning_rate": 4.793901065580872e-05, + "loss": 1.7975, + "step": 24290 + }, + { + "epoch": 2.1, + "learning_rate": 4.793814432989691e-05, + "loss": 1.8709, + "step": 24300 + }, + { + "epoch": 2.1, + "learning_rate": 4.7937278003985106e-05, + "loss": 1.8859, + "step": 24310 + }, + { + "epoch": 2.1, + "learning_rate": 4.7936411678073294e-05, + "loss": 1.8634, + "step": 24320 + }, + { + "epoch": 2.11, + "learning_rate": 4.793554535216149e-05, + "loss": 1.8486, + "step": 24330 + }, + { + "epoch": 2.11, + "learning_rate": 4.7934679026249676e-05, + "loss": 1.9177, + "step": 24340 + }, + { + "epoch": 2.11, + "learning_rate": 4.793381270033787e-05, + "loss": 1.8683, + "step": 24350 + }, + { + "epoch": 2.11, + "learning_rate": 4.7932946374426065e-05, + "loss": 1.8207, + "step": 24360 + }, + { + "epoch": 2.11, + "learning_rate": 4.793208004851425e-05, + "loss": 1.8203, + "step": 24370 + }, + { + "epoch": 2.11, + "learning_rate": 4.793121372260245e-05, + "loss": 1.905, + "step": 24380 + }, + { + "epoch": 2.11, + "learning_rate": 4.793034739669064e-05, + "loss": 1.8776, + "step": 24390 + }, + { + "epoch": 2.11, + "learning_rate": 4.792948107077883e-05, + "loss": 1.8507, + "step": 24400 + }, + { + "epoch": 2.11, + "learning_rate": 4.7928614744867024e-05, + "loss": 1.8754, + "step": 24410 + }, + { + "epoch": 2.11, + "learning_rate": 4.792774841895522e-05, + "loss": 1.8744, + "step": 24420 + }, + { + "epoch": 2.11, + "learning_rate": 4.7926882093043406e-05, + "loss": 1.8631, + "step": 24430 + }, + { + "epoch": 2.12, + "learning_rate": 4.79260157671316e-05, + "loss": 1.8432, + "step": 24440 + }, + { + "epoch": 2.12, + "learning_rate": 4.792514944121979e-05, + "loss": 1.8962, + "step": 24450 + }, + { + "epoch": 2.12, + "learning_rate": 4.792428311530798e-05, + "loss": 1.857, + "step": 24460 + }, + { + "epoch": 2.12, + "learning_rate": 4.792341678939618e-05, + "loss": 1.8786, + "step": 24470 + }, + { + "epoch": 2.12, + "learning_rate": 4.7922550463484364e-05, + "loss": 1.8171, + "step": 24480 + }, + { + "epoch": 2.12, + "learning_rate": 4.792168413757256e-05, + "loss": 1.8605, + "step": 24490 + }, + { + "epoch": 2.12, + "learning_rate": 4.792081781166075e-05, + "loss": 1.8914, + "step": 24500 + }, + { + "epoch": 2.12, + "learning_rate": 4.791995148574894e-05, + "loss": 1.9042, + "step": 24510 + }, + { + "epoch": 2.12, + "learning_rate": 4.7919085159837135e-05, + "loss": 1.8427, + "step": 24520 + }, + { + "epoch": 2.12, + "learning_rate": 4.791821883392533e-05, + "loss": 1.8553, + "step": 24530 + }, + { + "epoch": 2.12, + "learning_rate": 4.791735250801352e-05, + "loss": 1.8639, + "step": 24540 + }, + { + "epoch": 2.12, + "learning_rate": 4.791648618210171e-05, + "loss": 1.8676, + "step": 24550 + }, + { + "epoch": 2.13, + "learning_rate": 4.79156198561899e-05, + "loss": 1.814, + "step": 24560 + }, + { + "epoch": 2.13, + "learning_rate": 4.7914753530278094e-05, + "loss": 1.8307, + "step": 24570 + }, + { + "epoch": 2.13, + "learning_rate": 4.791388720436629e-05, + "loss": 1.8205, + "step": 24580 + }, + { + "epoch": 2.13, + "learning_rate": 4.7913020878454476e-05, + "loss": 1.8371, + "step": 24590 + }, + { + "epoch": 2.13, + "learning_rate": 4.791215455254267e-05, + "loss": 1.8522, + "step": 24600 + }, + { + "epoch": 2.13, + "learning_rate": 4.7911288226630865e-05, + "loss": 1.8749, + "step": 24610 + }, + { + "epoch": 2.13, + "learning_rate": 4.791042190071905e-05, + "loss": 1.8222, + "step": 24620 + }, + { + "epoch": 2.13, + "learning_rate": 4.790955557480725e-05, + "loss": 1.8271, + "step": 24630 + }, + { + "epoch": 2.13, + "learning_rate": 4.790868924889544e-05, + "loss": 1.9041, + "step": 24640 + }, + { + "epoch": 2.13, + "learning_rate": 4.790782292298363e-05, + "loss": 1.8505, + "step": 24650 + }, + { + "epoch": 2.13, + "learning_rate": 4.7906956597071824e-05, + "loss": 1.8726, + "step": 24660 + }, + { + "epoch": 2.14, + "learning_rate": 4.790609027116001e-05, + "loss": 1.8355, + "step": 24670 + }, + { + "epoch": 2.14, + "learning_rate": 4.7905223945248206e-05, + "loss": 1.8182, + "step": 24680 + }, + { + "epoch": 2.14, + "learning_rate": 4.79043576193364e-05, + "loss": 1.8814, + "step": 24690 + }, + { + "epoch": 2.14, + "learning_rate": 4.790349129342459e-05, + "loss": 1.8741, + "step": 24700 + }, + { + "epoch": 2.14, + "learning_rate": 4.790262496751278e-05, + "loss": 1.8308, + "step": 24710 + }, + { + "epoch": 2.14, + "learning_rate": 4.790175864160098e-05, + "loss": 1.8367, + "step": 24720 + }, + { + "epoch": 2.14, + "learning_rate": 4.7900892315689165e-05, + "loss": 1.8039, + "step": 24730 + }, + { + "epoch": 2.14, + "learning_rate": 4.790002598977736e-05, + "loss": 1.8105, + "step": 24740 + }, + { + "epoch": 2.14, + "learning_rate": 4.7899159663865554e-05, + "loss": 1.8526, + "step": 24750 + }, + { + "epoch": 2.14, + "learning_rate": 4.789829333795374e-05, + "loss": 1.8869, + "step": 24760 + }, + { + "epoch": 2.14, + "learning_rate": 4.7897427012041936e-05, + "loss": 1.8677, + "step": 24770 + }, + { + "epoch": 2.14, + "learning_rate": 4.7896560686130123e-05, + "loss": 1.88, + "step": 24780 + }, + { + "epoch": 2.15, + "learning_rate": 4.789569436021832e-05, + "loss": 1.9045, + "step": 24790 + }, + { + "epoch": 2.15, + "learning_rate": 4.789482803430651e-05, + "loss": 1.8815, + "step": 24800 + }, + { + "epoch": 2.15, + "learning_rate": 4.78939617083947e-05, + "loss": 1.8814, + "step": 24810 + }, + { + "epoch": 2.15, + "learning_rate": 4.7893095382482894e-05, + "loss": 1.8472, + "step": 24820 + }, + { + "epoch": 2.15, + "learning_rate": 4.789222905657109e-05, + "loss": 1.844, + "step": 24830 + }, + { + "epoch": 2.15, + "learning_rate": 4.7891362730659277e-05, + "loss": 1.8647, + "step": 24840 + }, + { + "epoch": 2.15, + "learning_rate": 4.789049640474747e-05, + "loss": 1.831, + "step": 24850 + }, + { + "epoch": 2.15, + "learning_rate": 4.7889630078835666e-05, + "loss": 1.8405, + "step": 24860 + }, + { + "epoch": 2.15, + "learning_rate": 4.788876375292385e-05, + "loss": 1.8139, + "step": 24870 + }, + { + "epoch": 2.15, + "learning_rate": 4.788789742701205e-05, + "loss": 1.8563, + "step": 24880 + }, + { + "epoch": 2.15, + "learning_rate": 4.7887031101100235e-05, + "loss": 1.8713, + "step": 24890 + }, + { + "epoch": 2.16, + "learning_rate": 4.788616477518843e-05, + "loss": 1.8581, + "step": 24900 + }, + { + "epoch": 2.16, + "learning_rate": 4.7885298449276624e-05, + "loss": 1.8384, + "step": 24910 + }, + { + "epoch": 2.16, + "learning_rate": 4.788443212336481e-05, + "loss": 1.8303, + "step": 24920 + }, + { + "epoch": 2.16, + "learning_rate": 4.7883565797453006e-05, + "loss": 1.8677, + "step": 24930 + }, + { + "epoch": 2.16, + "learning_rate": 4.78826994715412e-05, + "loss": 1.8365, + "step": 24940 + }, + { + "epoch": 2.16, + "learning_rate": 4.788183314562939e-05, + "loss": 1.8393, + "step": 24950 + }, + { + "epoch": 2.16, + "learning_rate": 4.788096681971758e-05, + "loss": 1.8275, + "step": 24960 + }, + { + "epoch": 2.16, + "learning_rate": 4.788010049380578e-05, + "loss": 1.8471, + "step": 24970 + }, + { + "epoch": 2.16, + "learning_rate": 4.7879234167893965e-05, + "loss": 1.8414, + "step": 24980 + }, + { + "epoch": 2.16, + "learning_rate": 4.787836784198216e-05, + "loss": 1.8611, + "step": 24990 + }, + { + "epoch": 2.16, + "learning_rate": 4.787750151607035e-05, + "loss": 1.8438, + "step": 25000 + }, + { + "epoch": 2.16, + "learning_rate": 4.787663519015854e-05, + "loss": 1.8334, + "step": 25010 + }, + { + "epoch": 2.17, + "learning_rate": 4.7875768864246736e-05, + "loss": 1.8502, + "step": 25020 + }, + { + "epoch": 2.17, + "learning_rate": 4.7874902538334924e-05, + "loss": 1.8682, + "step": 25030 + }, + { + "epoch": 2.17, + "learning_rate": 4.787403621242312e-05, + "loss": 1.8563, + "step": 25040 + }, + { + "epoch": 2.17, + "learning_rate": 4.787316988651131e-05, + "loss": 1.7687, + "step": 25050 + }, + { + "epoch": 2.17, + "learning_rate": 4.78723035605995e-05, + "loss": 1.8535, + "step": 25060 + }, + { + "epoch": 2.17, + "learning_rate": 4.7871437234687695e-05, + "loss": 1.8585, + "step": 25070 + }, + { + "epoch": 2.17, + "learning_rate": 4.787057090877588e-05, + "loss": 1.8687, + "step": 25080 + }, + { + "epoch": 2.17, + "learning_rate": 4.786970458286408e-05, + "loss": 1.927, + "step": 25090 + }, + { + "epoch": 2.17, + "learning_rate": 4.786883825695227e-05, + "loss": 1.9182, + "step": 25100 + }, + { + "epoch": 2.17, + "learning_rate": 4.786797193104046e-05, + "loss": 1.869, + "step": 25110 + }, + { + "epoch": 2.17, + "learning_rate": 4.7867105605128654e-05, + "loss": 1.8936, + "step": 25120 + }, + { + "epoch": 2.18, + "learning_rate": 4.786623927921685e-05, + "loss": 1.8016, + "step": 25130 + }, + { + "epoch": 2.18, + "learning_rate": 4.7865372953305036e-05, + "loss": 1.8465, + "step": 25140 + }, + { + "epoch": 2.18, + "learning_rate": 4.786450662739323e-05, + "loss": 1.8222, + "step": 25150 + }, + { + "epoch": 2.18, + "learning_rate": 4.7863640301481425e-05, + "loss": 1.8672, + "step": 25160 + }, + { + "epoch": 2.18, + "learning_rate": 4.786277397556961e-05, + "loss": 1.8337, + "step": 25170 + }, + { + "epoch": 2.18, + "learning_rate": 4.786190764965781e-05, + "loss": 1.8317, + "step": 25180 + }, + { + "epoch": 2.18, + "learning_rate": 4.7861041323745994e-05, + "loss": 1.8305, + "step": 25190 + }, + { + "epoch": 2.18, + "learning_rate": 4.786017499783419e-05, + "loss": 1.8939, + "step": 25200 + }, + { + "epoch": 2.18, + "learning_rate": 4.785930867192238e-05, + "loss": 1.8639, + "step": 25210 + }, + { + "epoch": 2.18, + "learning_rate": 4.785844234601057e-05, + "loss": 1.8429, + "step": 25220 + }, + { + "epoch": 2.18, + "learning_rate": 4.7857576020098765e-05, + "loss": 1.8055, + "step": 25230 + }, + { + "epoch": 2.18, + "learning_rate": 4.785670969418696e-05, + "loss": 1.8837, + "step": 25240 + }, + { + "epoch": 2.19, + "learning_rate": 4.785584336827515e-05, + "loss": 1.8841, + "step": 25250 + }, + { + "epoch": 2.19, + "learning_rate": 4.785497704236334e-05, + "loss": 1.7934, + "step": 25260 + }, + { + "epoch": 2.19, + "learning_rate": 4.7854110716451536e-05, + "loss": 1.8818, + "step": 25270 + }, + { + "epoch": 2.19, + "learning_rate": 4.7853244390539724e-05, + "loss": 1.8299, + "step": 25280 + }, + { + "epoch": 2.19, + "learning_rate": 4.785237806462792e-05, + "loss": 1.851, + "step": 25290 + }, + { + "epoch": 2.19, + "learning_rate": 4.7851511738716106e-05, + "loss": 1.8705, + "step": 25300 + }, + { + "epoch": 2.19, + "learning_rate": 4.78506454128043e-05, + "loss": 1.8597, + "step": 25310 + }, + { + "epoch": 2.19, + "learning_rate": 4.7849779086892495e-05, + "loss": 1.9211, + "step": 25320 + }, + { + "epoch": 2.19, + "learning_rate": 4.784891276098068e-05, + "loss": 1.8395, + "step": 25330 + }, + { + "epoch": 2.19, + "learning_rate": 4.784804643506888e-05, + "loss": 1.8687, + "step": 25340 + }, + { + "epoch": 2.19, + "learning_rate": 4.784718010915707e-05, + "loss": 1.792, + "step": 25350 + }, + { + "epoch": 2.2, + "learning_rate": 4.784631378324526e-05, + "loss": 1.8295, + "step": 25360 + }, + { + "epoch": 2.2, + "learning_rate": 4.7845447457333454e-05, + "loss": 1.9022, + "step": 25370 + }, + { + "epoch": 2.2, + "learning_rate": 4.784458113142165e-05, + "loss": 1.8672, + "step": 25380 + }, + { + "epoch": 2.2, + "learning_rate": 4.7843714805509836e-05, + "loss": 1.8649, + "step": 25390 + }, + { + "epoch": 2.2, + "learning_rate": 4.784284847959803e-05, + "loss": 1.8363, + "step": 25400 + }, + { + "epoch": 2.2, + "learning_rate": 4.784198215368622e-05, + "loss": 1.808, + "step": 25410 + }, + { + "epoch": 2.2, + "learning_rate": 4.784111582777441e-05, + "loss": 1.8599, + "step": 25420 + }, + { + "epoch": 2.2, + "learning_rate": 4.784024950186261e-05, + "loss": 1.8034, + "step": 25430 + }, + { + "epoch": 2.2, + "learning_rate": 4.7839383175950795e-05, + "loss": 1.8214, + "step": 25440 + }, + { + "epoch": 2.2, + "learning_rate": 4.783851685003899e-05, + "loss": 1.8714, + "step": 25450 + }, + { + "epoch": 2.2, + "learning_rate": 4.7837650524127184e-05, + "loss": 1.8762, + "step": 25460 + }, + { + "epoch": 2.2, + "learning_rate": 4.783678419821537e-05, + "loss": 1.8353, + "step": 25470 + }, + { + "epoch": 2.21, + "learning_rate": 4.7835917872303566e-05, + "loss": 1.8885, + "step": 25480 + }, + { + "epoch": 2.21, + "learning_rate": 4.783505154639176e-05, + "loss": 1.9149, + "step": 25490 + }, + { + "epoch": 2.21, + "learning_rate": 4.783418522047995e-05, + "loss": 1.8554, + "step": 25500 + }, + { + "epoch": 2.21, + "learning_rate": 4.783331889456814e-05, + "loss": 1.8834, + "step": 25510 + }, + { + "epoch": 2.21, + "learning_rate": 4.783245256865633e-05, + "loss": 1.8518, + "step": 25520 + }, + { + "epoch": 2.21, + "learning_rate": 4.7831586242744524e-05, + "loss": 1.7905, + "step": 25530 + }, + { + "epoch": 2.21, + "learning_rate": 4.783071991683272e-05, + "loss": 1.856, + "step": 25540 + }, + { + "epoch": 2.21, + "learning_rate": 4.7829853590920907e-05, + "loss": 1.8881, + "step": 25550 + }, + { + "epoch": 2.21, + "learning_rate": 4.78289872650091e-05, + "loss": 1.8746, + "step": 25560 + }, + { + "epoch": 2.21, + "learning_rate": 4.7828120939097295e-05, + "loss": 1.8834, + "step": 25570 + }, + { + "epoch": 2.21, + "learning_rate": 4.782725461318548e-05, + "loss": 1.8459, + "step": 25580 + }, + { + "epoch": 2.21, + "learning_rate": 4.782638828727368e-05, + "loss": 1.854, + "step": 25590 + }, + { + "epoch": 2.22, + "learning_rate": 4.782552196136187e-05, + "loss": 1.8135, + "step": 25600 + }, + { + "epoch": 2.22, + "learning_rate": 4.782465563545006e-05, + "loss": 1.8064, + "step": 25610 + }, + { + "epoch": 2.22, + "learning_rate": 4.7823789309538254e-05, + "loss": 1.832, + "step": 25620 + }, + { + "epoch": 2.22, + "learning_rate": 4.782292298362644e-05, + "loss": 1.8387, + "step": 25630 + }, + { + "epoch": 2.22, + "learning_rate": 4.7822056657714636e-05, + "loss": 1.8951, + "step": 25640 + }, + { + "epoch": 2.22, + "learning_rate": 4.782119033180283e-05, + "loss": 1.8809, + "step": 25650 + }, + { + "epoch": 2.22, + "learning_rate": 4.782032400589102e-05, + "loss": 1.8988, + "step": 25660 + }, + { + "epoch": 2.22, + "learning_rate": 4.781945767997921e-05, + "loss": 1.8756, + "step": 25670 + }, + { + "epoch": 2.22, + "learning_rate": 4.781859135406741e-05, + "loss": 1.8847, + "step": 25680 + }, + { + "epoch": 2.22, + "learning_rate": 4.7817725028155595e-05, + "loss": 1.8823, + "step": 25690 + }, + { + "epoch": 2.22, + "learning_rate": 4.781685870224379e-05, + "loss": 1.8593, + "step": 25700 + }, + { + "epoch": 2.23, + "learning_rate": 4.7815992376331984e-05, + "loss": 1.8415, + "step": 25710 + }, + { + "epoch": 2.23, + "learning_rate": 4.781512605042017e-05, + "loss": 1.8175, + "step": 25720 + }, + { + "epoch": 2.23, + "learning_rate": 4.7814259724508366e-05, + "loss": 1.8921, + "step": 25730 + }, + { + "epoch": 2.23, + "learning_rate": 4.7813393398596554e-05, + "loss": 1.858, + "step": 25740 + }, + { + "epoch": 2.23, + "learning_rate": 4.781252707268475e-05, + "loss": 1.78, + "step": 25750 + }, + { + "epoch": 2.23, + "learning_rate": 4.781166074677294e-05, + "loss": 1.8412, + "step": 25760 + }, + { + "epoch": 2.23, + "learning_rate": 4.781079442086113e-05, + "loss": 1.8485, + "step": 25770 + }, + { + "epoch": 2.23, + "learning_rate": 4.7809928094949325e-05, + "loss": 1.856, + "step": 25780 + }, + { + "epoch": 2.23, + "learning_rate": 4.780906176903752e-05, + "loss": 1.8708, + "step": 25790 + }, + { + "epoch": 2.23, + "learning_rate": 4.780819544312571e-05, + "loss": 1.8414, + "step": 25800 + }, + { + "epoch": 2.23, + "learning_rate": 4.78073291172139e-05, + "loss": 1.8399, + "step": 25810 + }, + { + "epoch": 2.23, + "learning_rate": 4.780646279130209e-05, + "loss": 1.9306, + "step": 25820 + }, + { + "epoch": 2.24, + "learning_rate": 4.7805596465390283e-05, + "loss": 1.8031, + "step": 25830 + }, + { + "epoch": 2.24, + "learning_rate": 4.780473013947848e-05, + "loss": 1.8022, + "step": 25840 + }, + { + "epoch": 2.24, + "learning_rate": 4.7803863813566666e-05, + "loss": 1.789, + "step": 25850 + }, + { + "epoch": 2.24, + "learning_rate": 4.780299748765486e-05, + "loss": 1.843, + "step": 25860 + }, + { + "epoch": 2.24, + "learning_rate": 4.7802131161743055e-05, + "loss": 1.8497, + "step": 25870 + }, + { + "epoch": 2.24, + "learning_rate": 4.780126483583124e-05, + "loss": 1.7757, + "step": 25880 + }, + { + "epoch": 2.24, + "learning_rate": 4.780039850991944e-05, + "loss": 1.8144, + "step": 25890 + }, + { + "epoch": 2.24, + "learning_rate": 4.779953218400763e-05, + "loss": 1.8735, + "step": 25900 + }, + { + "epoch": 2.24, + "learning_rate": 4.779866585809582e-05, + "loss": 1.8234, + "step": 25910 + }, + { + "epoch": 2.24, + "learning_rate": 4.779779953218401e-05, + "loss": 1.8558, + "step": 25920 + }, + { + "epoch": 2.24, + "learning_rate": 4.77969332062722e-05, + "loss": 1.8742, + "step": 25930 + }, + { + "epoch": 2.25, + "learning_rate": 4.7796066880360395e-05, + "loss": 1.8544, + "step": 25940 + }, + { + "epoch": 2.25, + "learning_rate": 4.779520055444859e-05, + "loss": 1.8706, + "step": 25950 + }, + { + "epoch": 2.25, + "learning_rate": 4.779433422853678e-05, + "loss": 1.8335, + "step": 25960 + }, + { + "epoch": 2.25, + "learning_rate": 4.779346790262497e-05, + "loss": 1.8541, + "step": 25970 + }, + { + "epoch": 2.25, + "learning_rate": 4.7792601576713166e-05, + "loss": 1.8532, + "step": 25980 + }, + { + "epoch": 2.25, + "learning_rate": 4.7791735250801354e-05, + "loss": 1.7665, + "step": 25990 + }, + { + "epoch": 2.25, + "learning_rate": 4.779086892488955e-05, + "loss": 1.8639, + "step": 26000 + }, + { + "epoch": 2.25, + "learning_rate": 4.779000259897774e-05, + "loss": 1.7888, + "step": 26010 + }, + { + "epoch": 2.25, + "learning_rate": 4.778913627306593e-05, + "loss": 1.8663, + "step": 26020 + }, + { + "epoch": 2.25, + "learning_rate": 4.7788269947154125e-05, + "loss": 1.8507, + "step": 26030 + }, + { + "epoch": 2.25, + "learning_rate": 4.778740362124231e-05, + "loss": 1.8521, + "step": 26040 + }, + { + "epoch": 2.25, + "learning_rate": 4.778653729533051e-05, + "loss": 1.8198, + "step": 26050 + }, + { + "epoch": 2.26, + "learning_rate": 4.77856709694187e-05, + "loss": 1.8077, + "step": 26060 + }, + { + "epoch": 2.26, + "learning_rate": 4.778480464350689e-05, + "loss": 1.7929, + "step": 26070 + }, + { + "epoch": 2.26, + "learning_rate": 4.7783938317595084e-05, + "loss": 1.8279, + "step": 26080 + }, + { + "epoch": 2.26, + "learning_rate": 4.778307199168328e-05, + "loss": 1.8262, + "step": 26090 + }, + { + "epoch": 2.26, + "learning_rate": 4.7782205665771466e-05, + "loss": 1.8058, + "step": 26100 + }, + { + "epoch": 2.26, + "learning_rate": 4.778133933985966e-05, + "loss": 1.8401, + "step": 26110 + }, + { + "epoch": 2.26, + "learning_rate": 4.7780473013947855e-05, + "loss": 1.8319, + "step": 26120 + }, + { + "epoch": 2.26, + "learning_rate": 4.777960668803604e-05, + "loss": 1.8269, + "step": 26130 + }, + { + "epoch": 2.26, + "learning_rate": 4.777874036212424e-05, + "loss": 1.9092, + "step": 26140 + }, + { + "epoch": 2.26, + "learning_rate": 4.7777874036212425e-05, + "loss": 1.8387, + "step": 26150 + }, + { + "epoch": 2.26, + "learning_rate": 4.777700771030062e-05, + "loss": 1.8254, + "step": 26160 + }, + { + "epoch": 2.27, + "learning_rate": 4.7776141384388814e-05, + "loss": 1.8527, + "step": 26170 + }, + { + "epoch": 2.27, + "learning_rate": 4.7775275058477e-05, + "loss": 1.8596, + "step": 26180 + }, + { + "epoch": 2.27, + "learning_rate": 4.7774408732565196e-05, + "loss": 1.8286, + "step": 26190 + }, + { + "epoch": 2.27, + "learning_rate": 4.777354240665339e-05, + "loss": 1.8204, + "step": 26200 + }, + { + "epoch": 2.27, + "learning_rate": 4.777267608074158e-05, + "loss": 1.8299, + "step": 26210 + }, + { + "epoch": 2.27, + "learning_rate": 4.777180975482977e-05, + "loss": 1.8365, + "step": 26220 + }, + { + "epoch": 2.27, + "learning_rate": 4.777094342891797e-05, + "loss": 1.8156, + "step": 26230 + }, + { + "epoch": 2.27, + "learning_rate": 4.7770077103006154e-05, + "loss": 1.8024, + "step": 26240 + }, + { + "epoch": 2.27, + "learning_rate": 4.776921077709435e-05, + "loss": 1.8833, + "step": 26250 + }, + { + "epoch": 2.27, + "learning_rate": 4.7768344451182537e-05, + "loss": 1.8325, + "step": 26260 + }, + { + "epoch": 2.27, + "learning_rate": 4.776747812527073e-05, + "loss": 1.8564, + "step": 26270 + }, + { + "epoch": 2.27, + "learning_rate": 4.7766611799358925e-05, + "loss": 1.8399, + "step": 26280 + }, + { + "epoch": 2.28, + "learning_rate": 4.776574547344711e-05, + "loss": 1.8526, + "step": 26290 + }, + { + "epoch": 2.28, + "learning_rate": 4.776487914753531e-05, + "loss": 1.8378, + "step": 26300 + }, + { + "epoch": 2.28, + "learning_rate": 4.77640128216235e-05, + "loss": 1.8563, + "step": 26310 + }, + { + "epoch": 2.28, + "learning_rate": 4.776314649571169e-05, + "loss": 1.8125, + "step": 26320 + }, + { + "epoch": 2.28, + "learning_rate": 4.7762280169799884e-05, + "loss": 1.7974, + "step": 26330 + }, + { + "epoch": 2.28, + "learning_rate": 4.776141384388808e-05, + "loss": 1.8337, + "step": 26340 + }, + { + "epoch": 2.28, + "learning_rate": 4.7760547517976266e-05, + "loss": 1.8795, + "step": 26350 + }, + { + "epoch": 2.28, + "learning_rate": 4.775968119206446e-05, + "loss": 1.8692, + "step": 26360 + }, + { + "epoch": 2.28, + "learning_rate": 4.775881486615265e-05, + "loss": 1.8584, + "step": 26370 + }, + { + "epoch": 2.28, + "learning_rate": 4.775794854024084e-05, + "loss": 1.8361, + "step": 26380 + }, + { + "epoch": 2.28, + "learning_rate": 4.775708221432904e-05, + "loss": 1.8766, + "step": 26390 + }, + { + "epoch": 2.29, + "learning_rate": 4.7756215888417225e-05, + "loss": 1.8352, + "step": 26400 + }, + { + "epoch": 2.29, + "learning_rate": 4.775534956250542e-05, + "loss": 1.8515, + "step": 26410 + }, + { + "epoch": 2.29, + "learning_rate": 4.7754483236593614e-05, + "loss": 1.8167, + "step": 26420 + }, + { + "epoch": 2.29, + "learning_rate": 4.77536169106818e-05, + "loss": 1.8147, + "step": 26430 + }, + { + "epoch": 2.29, + "learning_rate": 4.7752750584769996e-05, + "loss": 1.8047, + "step": 26440 + }, + { + "epoch": 2.29, + "learning_rate": 4.775188425885819e-05, + "loss": 1.8041, + "step": 26450 + }, + { + "epoch": 2.29, + "learning_rate": 4.775101793294638e-05, + "loss": 1.8909, + "step": 26460 + }, + { + "epoch": 2.29, + "learning_rate": 4.775015160703457e-05, + "loss": 1.8762, + "step": 26470 + }, + { + "epoch": 2.29, + "learning_rate": 4.774928528112276e-05, + "loss": 1.8311, + "step": 26480 + }, + { + "epoch": 2.29, + "learning_rate": 4.7748418955210955e-05, + "loss": 1.8865, + "step": 26490 + }, + { + "epoch": 2.29, + "learning_rate": 4.774755262929915e-05, + "loss": 1.8138, + "step": 26500 + }, + { + "epoch": 2.29, + "learning_rate": 4.774668630338734e-05, + "loss": 1.8526, + "step": 26510 + }, + { + "epoch": 2.3, + "learning_rate": 4.774581997747553e-05, + "loss": 1.8826, + "step": 26520 + }, + { + "epoch": 2.3, + "learning_rate": 4.7744953651563726e-05, + "loss": 1.8767, + "step": 26530 + }, + { + "epoch": 2.3, + "learning_rate": 4.7744087325651913e-05, + "loss": 1.8628, + "step": 26540 + }, + { + "epoch": 2.3, + "learning_rate": 4.774322099974011e-05, + "loss": 1.8291, + "step": 26550 + }, + { + "epoch": 2.3, + "learning_rate": 4.7742354673828296e-05, + "loss": 1.8574, + "step": 26560 + }, + { + "epoch": 2.3, + "learning_rate": 4.774148834791649e-05, + "loss": 1.8543, + "step": 26570 + }, + { + "epoch": 2.3, + "learning_rate": 4.7740622022004685e-05, + "loss": 1.878, + "step": 26580 + }, + { + "epoch": 2.3, + "learning_rate": 4.773975569609287e-05, + "loss": 1.8094, + "step": 26590 + }, + { + "epoch": 2.3, + "learning_rate": 4.773888937018107e-05, + "loss": 1.9027, + "step": 26600 + }, + { + "epoch": 2.3, + "learning_rate": 4.773802304426926e-05, + "loss": 1.8294, + "step": 26610 + }, + { + "epoch": 2.3, + "learning_rate": 4.773715671835745e-05, + "loss": 1.8405, + "step": 26620 + }, + { + "epoch": 2.3, + "learning_rate": 4.773629039244564e-05, + "loss": 1.8208, + "step": 26630 + }, + { + "epoch": 2.31, + "learning_rate": 4.773542406653384e-05, + "loss": 1.8351, + "step": 26640 + }, + { + "epoch": 2.31, + "learning_rate": 4.7734557740622025e-05, + "loss": 1.8828, + "step": 26650 + }, + { + "epoch": 2.31, + "learning_rate": 4.773369141471022e-05, + "loss": 1.8256, + "step": 26660 + }, + { + "epoch": 2.31, + "learning_rate": 4.773282508879841e-05, + "loss": 1.8383, + "step": 26670 + }, + { + "epoch": 2.31, + "learning_rate": 4.77319587628866e-05, + "loss": 1.822, + "step": 26680 + }, + { + "epoch": 2.31, + "learning_rate": 4.7731092436974796e-05, + "loss": 1.8563, + "step": 26690 + }, + { + "epoch": 2.31, + "learning_rate": 4.7730226111062984e-05, + "loss": 1.8201, + "step": 26700 + }, + { + "epoch": 2.31, + "learning_rate": 4.772935978515118e-05, + "loss": 1.8522, + "step": 26710 + }, + { + "epoch": 2.31, + "learning_rate": 4.772849345923937e-05, + "loss": 1.8283, + "step": 26720 + }, + { + "epoch": 2.31, + "learning_rate": 4.772762713332756e-05, + "loss": 1.8611, + "step": 26730 + }, + { + "epoch": 2.31, + "learning_rate": 4.7726760807415755e-05, + "loss": 1.8092, + "step": 26740 + }, + { + "epoch": 2.32, + "learning_rate": 4.772589448150395e-05, + "loss": 1.7941, + "step": 26750 + }, + { + "epoch": 2.32, + "learning_rate": 4.772502815559214e-05, + "loss": 1.8344, + "step": 26760 + }, + { + "epoch": 2.32, + "learning_rate": 4.772416182968033e-05, + "loss": 1.7973, + "step": 26770 + }, + { + "epoch": 2.32, + "learning_rate": 4.772329550376852e-05, + "loss": 1.8472, + "step": 26780 + }, + { + "epoch": 2.32, + "learning_rate": 4.7722429177856714e-05, + "loss": 1.8636, + "step": 26790 + }, + { + "epoch": 2.32, + "learning_rate": 4.772156285194491e-05, + "loss": 1.8142, + "step": 26800 + }, + { + "epoch": 2.32, + "learning_rate": 4.7720696526033096e-05, + "loss": 1.8336, + "step": 26810 + }, + { + "epoch": 2.32, + "learning_rate": 4.771983020012129e-05, + "loss": 1.7773, + "step": 26820 + }, + { + "epoch": 2.32, + "learning_rate": 4.7718963874209485e-05, + "loss": 1.8302, + "step": 26830 + }, + { + "epoch": 2.32, + "learning_rate": 4.771809754829767e-05, + "loss": 1.8358, + "step": 26840 + }, + { + "epoch": 2.32, + "learning_rate": 4.771723122238587e-05, + "loss": 1.8085, + "step": 26850 + }, + { + "epoch": 2.32, + "learning_rate": 4.771636489647406e-05, + "loss": 1.8163, + "step": 26860 + }, + { + "epoch": 2.33, + "learning_rate": 4.771549857056225e-05, + "loss": 1.8047, + "step": 26870 + }, + { + "epoch": 2.33, + "learning_rate": 4.7714632244650444e-05, + "loss": 1.8261, + "step": 26880 + }, + { + "epoch": 2.33, + "learning_rate": 4.771376591873863e-05, + "loss": 1.8518, + "step": 26890 + }, + { + "epoch": 2.33, + "learning_rate": 4.7712899592826826e-05, + "loss": 1.8225, + "step": 26900 + }, + { + "epoch": 2.33, + "learning_rate": 4.771203326691502e-05, + "loss": 1.7968, + "step": 26910 + }, + { + "epoch": 2.33, + "learning_rate": 4.771116694100321e-05, + "loss": 1.8396, + "step": 26920 + }, + { + "epoch": 2.33, + "learning_rate": 4.77103006150914e-05, + "loss": 1.7964, + "step": 26930 + }, + { + "epoch": 2.33, + "learning_rate": 4.77094342891796e-05, + "loss": 1.8741, + "step": 26940 + }, + { + "epoch": 2.33, + "learning_rate": 4.7708567963267784e-05, + "loss": 1.818, + "step": 26950 + }, + { + "epoch": 2.33, + "learning_rate": 4.770770163735598e-05, + "loss": 1.8743, + "step": 26960 + }, + { + "epoch": 2.33, + "learning_rate": 4.770683531144417e-05, + "loss": 1.8866, + "step": 26970 + }, + { + "epoch": 2.34, + "learning_rate": 4.770596898553236e-05, + "loss": 1.7701, + "step": 26980 + }, + { + "epoch": 2.34, + "learning_rate": 4.7705102659620555e-05, + "loss": 1.8062, + "step": 26990 + }, + { + "epoch": 2.34, + "learning_rate": 4.770423633370874e-05, + "loss": 1.8401, + "step": 27000 + }, + { + "epoch": 2.34, + "learning_rate": 4.770337000779694e-05, + "loss": 1.8424, + "step": 27010 + }, + { + "epoch": 2.34, + "learning_rate": 4.770250368188513e-05, + "loss": 1.8402, + "step": 27020 + }, + { + "epoch": 2.34, + "learning_rate": 4.770163735597332e-05, + "loss": 1.808, + "step": 27030 + }, + { + "epoch": 2.34, + "learning_rate": 4.7700771030061514e-05, + "loss": 1.7926, + "step": 27040 + }, + { + "epoch": 2.34, + "learning_rate": 4.769990470414971e-05, + "loss": 1.8892, + "step": 27050 + }, + { + "epoch": 2.34, + "learning_rate": 4.7699038378237896e-05, + "loss": 1.8755, + "step": 27060 + }, + { + "epoch": 2.34, + "learning_rate": 4.769817205232609e-05, + "loss": 1.813, + "step": 27070 + }, + { + "epoch": 2.34, + "learning_rate": 4.7697305726414285e-05, + "loss": 1.8419, + "step": 27080 + }, + { + "epoch": 2.34, + "learning_rate": 4.769643940050247e-05, + "loss": 1.8039, + "step": 27090 + }, + { + "epoch": 2.35, + "learning_rate": 4.769557307459067e-05, + "loss": 1.845, + "step": 27100 + }, + { + "epoch": 2.35, + "learning_rate": 4.7694706748678855e-05, + "loss": 1.7816, + "step": 27110 + }, + { + "epoch": 2.35, + "learning_rate": 4.769384042276705e-05, + "loss": 1.8726, + "step": 27120 + }, + { + "epoch": 2.35, + "learning_rate": 4.7692974096855244e-05, + "loss": 1.8625, + "step": 27130 + }, + { + "epoch": 2.35, + "learning_rate": 4.769210777094343e-05, + "loss": 1.8059, + "step": 27140 + }, + { + "epoch": 2.35, + "learning_rate": 4.7691241445031626e-05, + "loss": 1.8406, + "step": 27150 + }, + { + "epoch": 2.35, + "learning_rate": 4.769037511911982e-05, + "loss": 1.8082, + "step": 27160 + }, + { + "epoch": 2.35, + "learning_rate": 4.768950879320801e-05, + "loss": 1.8209, + "step": 27170 + }, + { + "epoch": 2.35, + "learning_rate": 4.76886424672962e-05, + "loss": 1.8217, + "step": 27180 + }, + { + "epoch": 2.35, + "learning_rate": 4.76877761413844e-05, + "loss": 1.8475, + "step": 27190 + }, + { + "epoch": 2.35, + "learning_rate": 4.7686909815472585e-05, + "loss": 1.8565, + "step": 27200 + }, + { + "epoch": 2.36, + "learning_rate": 4.768604348956078e-05, + "loss": 1.7556, + "step": 27210 + }, + { + "epoch": 2.36, + "learning_rate": 4.768517716364897e-05, + "loss": 1.8556, + "step": 27220 + }, + { + "epoch": 2.36, + "learning_rate": 4.768431083773716e-05, + "loss": 1.7926, + "step": 27230 + }, + { + "epoch": 2.36, + "learning_rate": 4.7683444511825356e-05, + "loss": 1.7836, + "step": 27240 + }, + { + "epoch": 2.36, + "learning_rate": 4.7682578185913543e-05, + "loss": 1.8004, + "step": 27250 + }, + { + "epoch": 2.36, + "learning_rate": 4.768171186000174e-05, + "loss": 1.7803, + "step": 27260 + }, + { + "epoch": 2.36, + "learning_rate": 4.768084553408993e-05, + "loss": 1.8301, + "step": 27270 + }, + { + "epoch": 2.36, + "learning_rate": 4.767997920817812e-05, + "loss": 1.8374, + "step": 27280 + }, + { + "epoch": 2.36, + "learning_rate": 4.7679112882266314e-05, + "loss": 1.8223, + "step": 27290 + }, + { + "epoch": 2.36, + "learning_rate": 4.767824655635451e-05, + "loss": 1.8327, + "step": 27300 + }, + { + "epoch": 2.36, + "learning_rate": 4.7677380230442697e-05, + "loss": 1.8322, + "step": 27310 + }, + { + "epoch": 2.36, + "learning_rate": 4.767651390453089e-05, + "loss": 1.8896, + "step": 27320 + }, + { + "epoch": 2.37, + "learning_rate": 4.767564757861908e-05, + "loss": 1.8211, + "step": 27330 + }, + { + "epoch": 2.37, + "learning_rate": 4.767478125270727e-05, + "loss": 1.8094, + "step": 27340 + }, + { + "epoch": 2.37, + "learning_rate": 4.767391492679547e-05, + "loss": 1.784, + "step": 27350 + }, + { + "epoch": 2.37, + "learning_rate": 4.7673048600883655e-05, + "loss": 1.8386, + "step": 27360 + }, + { + "epoch": 2.37, + "learning_rate": 4.767218227497185e-05, + "loss": 1.8316, + "step": 27370 + }, + { + "epoch": 2.37, + "learning_rate": 4.7671315949060044e-05, + "loss": 1.7705, + "step": 27380 + }, + { + "epoch": 2.37, + "learning_rate": 4.767044962314823e-05, + "loss": 1.8072, + "step": 27390 + }, + { + "epoch": 2.37, + "learning_rate": 4.7669583297236426e-05, + "loss": 1.8129, + "step": 27400 + }, + { + "epoch": 2.37, + "learning_rate": 4.7668716971324614e-05, + "loss": 1.798, + "step": 27410 + }, + { + "epoch": 2.37, + "learning_rate": 4.766785064541281e-05, + "loss": 1.7905, + "step": 27420 + }, + { + "epoch": 2.37, + "learning_rate": 4.7666984319501e-05, + "loss": 1.8349, + "step": 27430 + }, + { + "epoch": 2.38, + "learning_rate": 4.766611799358919e-05, + "loss": 1.7519, + "step": 27440 + }, + { + "epoch": 2.38, + "learning_rate": 4.7665251667677385e-05, + "loss": 1.8001, + "step": 27450 + }, + { + "epoch": 2.38, + "learning_rate": 4.766438534176558e-05, + "loss": 1.852, + "step": 27460 + }, + { + "epoch": 2.38, + "learning_rate": 4.766351901585377e-05, + "loss": 1.8288, + "step": 27470 + }, + { + "epoch": 2.38, + "learning_rate": 4.766265268994196e-05, + "loss": 1.8146, + "step": 27480 + }, + { + "epoch": 2.38, + "learning_rate": 4.7661786364030156e-05, + "loss": 1.8611, + "step": 27490 + }, + { + "epoch": 2.38, + "learning_rate": 4.7660920038118344e-05, + "loss": 1.8021, + "step": 27500 + }, + { + "epoch": 2.38, + "learning_rate": 4.766005371220654e-05, + "loss": 1.8501, + "step": 27510 + }, + { + "epoch": 2.38, + "learning_rate": 4.7659187386294726e-05, + "loss": 1.8001, + "step": 27520 + }, + { + "epoch": 2.38, + "learning_rate": 4.765832106038292e-05, + "loss": 1.828, + "step": 27530 + }, + { + "epoch": 2.38, + "learning_rate": 4.7657454734471115e-05, + "loss": 1.7767, + "step": 27540 + }, + { + "epoch": 2.38, + "learning_rate": 4.76565884085593e-05, + "loss": 1.8889, + "step": 27550 + }, + { + "epoch": 2.39, + "learning_rate": 4.765572208264749e-05, + "loss": 1.842, + "step": 27560 + }, + { + "epoch": 2.39, + "learning_rate": 4.7654855756735685e-05, + "loss": 1.7697, + "step": 27570 + }, + { + "epoch": 2.39, + "learning_rate": 4.765398943082387e-05, + "loss": 1.8696, + "step": 27580 + }, + { + "epoch": 2.39, + "learning_rate": 4.765312310491207e-05, + "loss": 1.7888, + "step": 27590 + }, + { + "epoch": 2.39, + "learning_rate": 4.765225677900026e-05, + "loss": 1.8309, + "step": 27600 + }, + { + "epoch": 2.39, + "learning_rate": 4.765139045308845e-05, + "loss": 1.7841, + "step": 27610 + }, + { + "epoch": 2.39, + "learning_rate": 4.765052412717664e-05, + "loss": 1.8027, + "step": 27620 + }, + { + "epoch": 2.39, + "learning_rate": 4.764965780126484e-05, + "loss": 1.7722, + "step": 27630 + }, + { + "epoch": 2.39, + "learning_rate": 4.7648791475353025e-05, + "loss": 1.7986, + "step": 27640 + }, + { + "epoch": 2.39, + "learning_rate": 4.764792514944122e-05, + "loss": 1.7954, + "step": 27650 + }, + { + "epoch": 2.39, + "learning_rate": 4.7647058823529414e-05, + "loss": 1.8308, + "step": 27660 + }, + { + "epoch": 2.39, + "learning_rate": 4.76461924976176e-05, + "loss": 1.8502, + "step": 27670 + }, + { + "epoch": 2.4, + "learning_rate": 4.7645326171705796e-05, + "loss": 1.8073, + "step": 27680 + }, + { + "epoch": 2.4, + "learning_rate": 4.7644459845793984e-05, + "loss": 1.8622, + "step": 27690 + }, + { + "epoch": 2.4, + "learning_rate": 4.764359351988218e-05, + "loss": 1.8295, + "step": 27700 + }, + { + "epoch": 2.4, + "learning_rate": 4.764272719397037e-05, + "loss": 1.8969, + "step": 27710 + }, + { + "epoch": 2.4, + "learning_rate": 4.764186086805856e-05, + "loss": 1.8324, + "step": 27720 + }, + { + "epoch": 2.4, + "learning_rate": 4.7640994542146755e-05, + "loss": 1.8245, + "step": 27730 + }, + { + "epoch": 2.4, + "learning_rate": 4.764012821623495e-05, + "loss": 1.7847, + "step": 27740 + }, + { + "epoch": 2.4, + "learning_rate": 4.763926189032314e-05, + "loss": 1.7971, + "step": 27750 + }, + { + "epoch": 2.4, + "learning_rate": 4.763839556441133e-05, + "loss": 1.7692, + "step": 27760 + }, + { + "epoch": 2.4, + "learning_rate": 4.7637529238499526e-05, + "loss": 1.8094, + "step": 27770 + }, + { + "epoch": 2.4, + "learning_rate": 4.7636662912587714e-05, + "loss": 1.8398, + "step": 27780 + }, + { + "epoch": 2.41, + "learning_rate": 4.763579658667591e-05, + "loss": 1.8118, + "step": 27790 + }, + { + "epoch": 2.41, + "learning_rate": 4.7634930260764096e-05, + "loss": 1.8141, + "step": 27800 + }, + { + "epoch": 2.41, + "learning_rate": 4.763406393485229e-05, + "loss": 1.8356, + "step": 27810 + }, + { + "epoch": 2.41, + "learning_rate": 4.7633197608940485e-05, + "loss": 1.7939, + "step": 27820 + }, + { + "epoch": 2.41, + "learning_rate": 4.763233128302867e-05, + "loss": 1.8315, + "step": 27830 + }, + { + "epoch": 2.41, + "learning_rate": 4.763146495711687e-05, + "loss": 1.9217, + "step": 27840 + }, + { + "epoch": 2.41, + "learning_rate": 4.763059863120506e-05, + "loss": 1.8292, + "step": 27850 + }, + { + "epoch": 2.41, + "learning_rate": 4.762973230529325e-05, + "loss": 1.8178, + "step": 27860 + }, + { + "epoch": 2.41, + "learning_rate": 4.7628865979381444e-05, + "loss": 1.8628, + "step": 27870 + }, + { + "epoch": 2.41, + "learning_rate": 4.762799965346964e-05, + "loss": 1.7947, + "step": 27880 + }, + { + "epoch": 2.41, + "learning_rate": 4.7627133327557826e-05, + "loss": 1.8073, + "step": 27890 + }, + { + "epoch": 2.41, + "learning_rate": 4.762626700164602e-05, + "loss": 1.8632, + "step": 27900 + }, + { + "epoch": 2.42, + "learning_rate": 4.762540067573421e-05, + "loss": 1.8335, + "step": 27910 + }, + { + "epoch": 2.42, + "learning_rate": 4.76245343498224e-05, + "loss": 1.7683, + "step": 27920 + }, + { + "epoch": 2.42, + "learning_rate": 4.76236680239106e-05, + "loss": 1.8242, + "step": 27930 + }, + { + "epoch": 2.42, + "learning_rate": 4.7622801697998785e-05, + "loss": 1.8789, + "step": 27940 + }, + { + "epoch": 2.42, + "learning_rate": 4.762193537208698e-05, + "loss": 1.8206, + "step": 27950 + }, + { + "epoch": 2.42, + "learning_rate": 4.7621069046175173e-05, + "loss": 1.8097, + "step": 27960 + }, + { + "epoch": 2.42, + "learning_rate": 4.762020272026336e-05, + "loss": 1.8447, + "step": 27970 + }, + { + "epoch": 2.42, + "learning_rate": 4.7619336394351556e-05, + "loss": 1.8763, + "step": 27980 + }, + { + "epoch": 2.42, + "learning_rate": 4.761847006843975e-05, + "loss": 1.7905, + "step": 27990 + }, + { + "epoch": 2.42, + "learning_rate": 4.761760374252794e-05, + "loss": 1.8643, + "step": 28000 + }, + { + "epoch": 2.42, + "learning_rate": 4.761673741661613e-05, + "loss": 1.7701, + "step": 28010 + }, + { + "epoch": 2.43, + "learning_rate": 4.761587109070432e-05, + "loss": 1.7891, + "step": 28020 + }, + { + "epoch": 2.43, + "learning_rate": 4.7615004764792514e-05, + "loss": 1.8136, + "step": 28030 + }, + { + "epoch": 2.43, + "learning_rate": 4.761413843888071e-05, + "loss": 1.7873, + "step": 28040 + }, + { + "epoch": 2.43, + "learning_rate": 4.7613272112968896e-05, + "loss": 1.8961, + "step": 28050 + }, + { + "epoch": 2.43, + "learning_rate": 4.761240578705709e-05, + "loss": 1.798, + "step": 28060 + }, + { + "epoch": 2.43, + "learning_rate": 4.7611539461145285e-05, + "loss": 1.784, + "step": 28070 + }, + { + "epoch": 2.43, + "learning_rate": 4.761067313523347e-05, + "loss": 1.7923, + "step": 28080 + }, + { + "epoch": 2.43, + "learning_rate": 4.760980680932167e-05, + "loss": 1.7627, + "step": 28090 + }, + { + "epoch": 2.43, + "learning_rate": 4.7608940483409855e-05, + "loss": 1.7749, + "step": 28100 + }, + { + "epoch": 2.43, + "learning_rate": 4.760807415749805e-05, + "loss": 1.8077, + "step": 28110 + }, + { + "epoch": 2.43, + "learning_rate": 4.7607207831586244e-05, + "loss": 1.8379, + "step": 28120 + }, + { + "epoch": 2.43, + "learning_rate": 4.760634150567443e-05, + "loss": 1.8444, + "step": 28130 + }, + { + "epoch": 2.44, + "learning_rate": 4.7605475179762626e-05, + "loss": 1.8024, + "step": 28140 + }, + { + "epoch": 2.44, + "learning_rate": 4.760460885385082e-05, + "loss": 1.817, + "step": 28150 + }, + { + "epoch": 2.44, + "learning_rate": 4.760374252793901e-05, + "loss": 1.7257, + "step": 28160 + }, + { + "epoch": 2.44, + "learning_rate": 4.76028762020272e-05, + "loss": 1.7817, + "step": 28170 + }, + { + "epoch": 2.44, + "learning_rate": 4.76020098761154e-05, + "loss": 1.8057, + "step": 28180 + }, + { + "epoch": 2.44, + "learning_rate": 4.7601143550203585e-05, + "loss": 1.8024, + "step": 28190 + }, + { + "epoch": 2.44, + "learning_rate": 4.760027722429178e-05, + "loss": 1.7994, + "step": 28200 + }, + { + "epoch": 2.44, + "learning_rate": 4.759941089837997e-05, + "loss": 1.817, + "step": 28210 + }, + { + "epoch": 2.44, + "learning_rate": 4.759854457246816e-05, + "loss": 1.7507, + "step": 28220 + }, + { + "epoch": 2.44, + "learning_rate": 4.7597678246556356e-05, + "loss": 1.794, + "step": 28230 + }, + { + "epoch": 2.44, + "learning_rate": 4.7596811920644544e-05, + "loss": 1.8233, + "step": 28240 + }, + { + "epoch": 2.45, + "learning_rate": 4.759594559473274e-05, + "loss": 1.7567, + "step": 28250 + }, + { + "epoch": 2.45, + "learning_rate": 4.759507926882093e-05, + "loss": 1.8199, + "step": 28260 + }, + { + "epoch": 2.45, + "learning_rate": 4.759421294290912e-05, + "loss": 1.8457, + "step": 28270 + }, + { + "epoch": 2.45, + "learning_rate": 4.7593346616997315e-05, + "loss": 1.8399, + "step": 28280 + }, + { + "epoch": 2.45, + "learning_rate": 4.759248029108551e-05, + "loss": 1.798, + "step": 28290 + }, + { + "epoch": 2.45, + "learning_rate": 4.75916139651737e-05, + "loss": 1.8591, + "step": 28300 + }, + { + "epoch": 2.45, + "learning_rate": 4.759074763926189e-05, + "loss": 1.8179, + "step": 28310 + }, + { + "epoch": 2.45, + "learning_rate": 4.758988131335008e-05, + "loss": 1.8632, + "step": 28320 + }, + { + "epoch": 2.45, + "learning_rate": 4.758901498743827e-05, + "loss": 1.8325, + "step": 28330 + }, + { + "epoch": 2.45, + "learning_rate": 4.758814866152647e-05, + "loss": 1.7917, + "step": 28340 + }, + { + "epoch": 2.45, + "learning_rate": 4.7587282335614655e-05, + "loss": 1.8292, + "step": 28350 + }, + { + "epoch": 2.45, + "learning_rate": 4.758641600970285e-05, + "loss": 1.8127, + "step": 28360 + }, + { + "epoch": 2.46, + "learning_rate": 4.7585549683791044e-05, + "loss": 1.8453, + "step": 28370 + }, + { + "epoch": 2.46, + "learning_rate": 4.758468335787923e-05, + "loss": 1.8333, + "step": 28380 + }, + { + "epoch": 2.46, + "learning_rate": 4.7583817031967426e-05, + "loss": 1.7283, + "step": 28390 + }, + { + "epoch": 2.46, + "learning_rate": 4.758295070605562e-05, + "loss": 1.8755, + "step": 28400 + }, + { + "epoch": 2.46, + "learning_rate": 4.758208438014381e-05, + "loss": 1.8285, + "step": 28410 + }, + { + "epoch": 2.46, + "learning_rate": 4.7581218054232e-05, + "loss": 1.8258, + "step": 28420 + }, + { + "epoch": 2.46, + "learning_rate": 4.758035172832019e-05, + "loss": 1.782, + "step": 28430 + }, + { + "epoch": 2.46, + "learning_rate": 4.7579485402408385e-05, + "loss": 1.8467, + "step": 28440 + }, + { + "epoch": 2.46, + "learning_rate": 4.757861907649658e-05, + "loss": 1.8218, + "step": 28450 + }, + { + "epoch": 2.46, + "learning_rate": 4.757775275058477e-05, + "loss": 1.7868, + "step": 28460 + }, + { + "epoch": 2.46, + "learning_rate": 4.757688642467296e-05, + "loss": 1.8128, + "step": 28470 + }, + { + "epoch": 2.47, + "learning_rate": 4.7576020098761156e-05, + "loss": 1.7688, + "step": 28480 + }, + { + "epoch": 2.47, + "learning_rate": 4.7575153772849344e-05, + "loss": 1.7833, + "step": 28490 + }, + { + "epoch": 2.47, + "learning_rate": 4.757428744693754e-05, + "loss": 1.8373, + "step": 28500 + }, + { + "epoch": 2.47, + "learning_rate": 4.757342112102573e-05, + "loss": 1.8256, + "step": 28510 + }, + { + "epoch": 2.47, + "learning_rate": 4.757255479511392e-05, + "loss": 1.7877, + "step": 28520 + }, + { + "epoch": 2.47, + "learning_rate": 4.7571688469202115e-05, + "loss": 1.7994, + "step": 28530 + }, + { + "epoch": 2.47, + "learning_rate": 4.75708221432903e-05, + "loss": 1.8136, + "step": 28540 + }, + { + "epoch": 2.47, + "learning_rate": 4.75699558173785e-05, + "loss": 1.792, + "step": 28550 + }, + { + "epoch": 2.47, + "learning_rate": 4.756908949146669e-05, + "loss": 1.8722, + "step": 28560 + }, + { + "epoch": 2.47, + "learning_rate": 4.756822316555488e-05, + "loss": 1.7857, + "step": 28570 + }, + { + "epoch": 2.47, + "learning_rate": 4.7567356839643074e-05, + "loss": 1.8368, + "step": 28580 + }, + { + "epoch": 2.47, + "learning_rate": 4.756649051373127e-05, + "loss": 1.822, + "step": 28590 + }, + { + "epoch": 2.48, + "learning_rate": 4.7565624187819456e-05, + "loss": 1.8083, + "step": 28600 + }, + { + "epoch": 2.48, + "learning_rate": 4.756475786190765e-05, + "loss": 1.8085, + "step": 28610 + }, + { + "epoch": 2.48, + "learning_rate": 4.7563891535995845e-05, + "loss": 1.8225, + "step": 28620 + }, + { + "epoch": 2.48, + "learning_rate": 4.756302521008403e-05, + "loss": 1.8393, + "step": 28630 + }, + { + "epoch": 2.48, + "learning_rate": 4.756215888417223e-05, + "loss": 1.8154, + "step": 28640 + }, + { + "epoch": 2.48, + "learning_rate": 4.7561292558260414e-05, + "loss": 1.7805, + "step": 28650 + }, + { + "epoch": 2.48, + "learning_rate": 4.756042623234861e-05, + "loss": 1.7594, + "step": 28660 + }, + { + "epoch": 2.48, + "learning_rate": 4.75595599064368e-05, + "loss": 1.8118, + "step": 28670 + }, + { + "epoch": 2.48, + "learning_rate": 4.755869358052499e-05, + "loss": 1.8034, + "step": 28680 + }, + { + "epoch": 2.48, + "learning_rate": 4.7557827254613186e-05, + "loss": 1.7891, + "step": 28690 + }, + { + "epoch": 2.48, + "learning_rate": 4.755696092870138e-05, + "loss": 1.7925, + "step": 28700 + }, + { + "epoch": 2.48, + "learning_rate": 4.755609460278957e-05, + "loss": 1.7145, + "step": 28710 + }, + { + "epoch": 2.49, + "learning_rate": 4.755522827687776e-05, + "loss": 1.7577, + "step": 28720 + }, + { + "epoch": 2.49, + "learning_rate": 4.7554361950965957e-05, + "loss": 1.7659, + "step": 28730 + }, + { + "epoch": 2.49, + "learning_rate": 4.7553495625054144e-05, + "loss": 1.8783, + "step": 28740 + }, + { + "epoch": 2.49, + "learning_rate": 4.755262929914234e-05, + "loss": 1.7803, + "step": 28750 + }, + { + "epoch": 2.49, + "learning_rate": 4.7551762973230526e-05, + "loss": 1.7608, + "step": 28760 + }, + { + "epoch": 2.49, + "learning_rate": 4.755089664731872e-05, + "loss": 1.8191, + "step": 28770 + }, + { + "epoch": 2.49, + "learning_rate": 4.7550030321406915e-05, + "loss": 1.7505, + "step": 28780 + }, + { + "epoch": 2.49, + "learning_rate": 4.75491639954951e-05, + "loss": 1.8532, + "step": 28790 + }, + { + "epoch": 2.49, + "learning_rate": 4.75482976695833e-05, + "loss": 1.8198, + "step": 28800 + }, + { + "epoch": 2.49, + "learning_rate": 4.754743134367149e-05, + "loss": 1.785, + "step": 28810 + }, + { + "epoch": 2.49, + "learning_rate": 4.754656501775968e-05, + "loss": 1.776, + "step": 28820 + }, + { + "epoch": 2.5, + "learning_rate": 4.7545698691847874e-05, + "loss": 1.7872, + "step": 28830 + }, + { + "epoch": 2.5, + "learning_rate": 4.754483236593607e-05, + "loss": 1.807, + "step": 28840 + }, + { + "epoch": 2.5, + "learning_rate": 4.7543966040024256e-05, + "loss": 1.8227, + "step": 28850 + }, + { + "epoch": 2.5, + "learning_rate": 4.754309971411245e-05, + "loss": 1.7936, + "step": 28860 + }, + { + "epoch": 2.5, + "learning_rate": 4.754223338820064e-05, + "loss": 1.8503, + "step": 28870 + }, + { + "epoch": 2.5, + "learning_rate": 4.754136706228883e-05, + "loss": 1.7661, + "step": 28880 + }, + { + "epoch": 2.5, + "learning_rate": 4.754050073637703e-05, + "loss": 1.8037, + "step": 28890 + }, + { + "epoch": 2.5, + "learning_rate": 4.7539634410465215e-05, + "loss": 1.8519, + "step": 28900 + }, + { + "epoch": 2.5, + "learning_rate": 4.753876808455341e-05, + "loss": 1.8062, + "step": 28910 + }, + { + "epoch": 2.5, + "learning_rate": 4.7537901758641604e-05, + "loss": 1.8181, + "step": 28920 + }, + { + "epoch": 2.5, + "learning_rate": 4.753703543272979e-05, + "loss": 1.8417, + "step": 28930 + }, + { + "epoch": 2.5, + "learning_rate": 4.7536169106817986e-05, + "loss": 1.8225, + "step": 28940 + }, + { + "epoch": 2.51, + "learning_rate": 4.7535302780906174e-05, + "loss": 1.7885, + "step": 28950 + }, + { + "epoch": 2.51, + "learning_rate": 4.753443645499437e-05, + "loss": 1.7423, + "step": 28960 + }, + { + "epoch": 2.51, + "learning_rate": 4.753357012908256e-05, + "loss": 1.7695, + "step": 28970 + }, + { + "epoch": 2.51, + "learning_rate": 4.753270380317075e-05, + "loss": 1.7958, + "step": 28980 + }, + { + "epoch": 2.51, + "learning_rate": 4.7531837477258945e-05, + "loss": 1.8365, + "step": 28990 + }, + { + "epoch": 2.51, + "learning_rate": 4.753097115134714e-05, + "loss": 1.7601, + "step": 29000 + }, + { + "epoch": 2.51, + "learning_rate": 4.753010482543533e-05, + "loss": 1.8029, + "step": 29010 + }, + { + "epoch": 2.51, + "learning_rate": 4.752923849952352e-05, + "loss": 1.7718, + "step": 29020 + }, + { + "epoch": 2.51, + "learning_rate": 4.7528372173611716e-05, + "loss": 1.7726, + "step": 29030 + }, + { + "epoch": 2.51, + "learning_rate": 4.75275058476999e-05, + "loss": 1.8421, + "step": 29040 + }, + { + "epoch": 2.51, + "learning_rate": 4.75266395217881e-05, + "loss": 1.7749, + "step": 29050 + }, + { + "epoch": 2.52, + "learning_rate": 4.7525773195876285e-05, + "loss": 1.7754, + "step": 29060 + }, + { + "epoch": 2.52, + "learning_rate": 4.752490686996448e-05, + "loss": 1.8395, + "step": 29070 + }, + { + "epoch": 2.52, + "learning_rate": 4.7524040544052674e-05, + "loss": 1.7723, + "step": 29080 + }, + { + "epoch": 2.52, + "learning_rate": 4.752317421814086e-05, + "loss": 1.7926, + "step": 29090 + }, + { + "epoch": 2.52, + "learning_rate": 4.7522307892229056e-05, + "loss": 1.7761, + "step": 29100 + }, + { + "epoch": 2.52, + "learning_rate": 4.752144156631725e-05, + "loss": 1.8098, + "step": 29110 + }, + { + "epoch": 2.52, + "learning_rate": 4.752057524040544e-05, + "loss": 1.7723, + "step": 29120 + }, + { + "epoch": 2.52, + "learning_rate": 4.751970891449363e-05, + "loss": 1.7956, + "step": 29130 + }, + { + "epoch": 2.52, + "learning_rate": 4.751884258858183e-05, + "loss": 1.8383, + "step": 29140 + }, + { + "epoch": 2.52, + "learning_rate": 4.7517976262670015e-05, + "loss": 1.8114, + "step": 29150 + }, + { + "epoch": 2.52, + "learning_rate": 4.751710993675821e-05, + "loss": 1.8138, + "step": 29160 + }, + { + "epoch": 2.52, + "learning_rate": 4.75162436108464e-05, + "loss": 1.7812, + "step": 29170 + }, + { + "epoch": 2.53, + "learning_rate": 4.751537728493459e-05, + "loss": 1.8227, + "step": 29180 + }, + { + "epoch": 2.53, + "learning_rate": 4.7514510959022786e-05, + "loss": 1.8051, + "step": 29190 + }, + { + "epoch": 2.53, + "learning_rate": 4.7513644633110974e-05, + "loss": 1.7835, + "step": 29200 + }, + { + "epoch": 2.53, + "learning_rate": 4.751277830719917e-05, + "loss": 1.7982, + "step": 29210 + }, + { + "epoch": 2.53, + "learning_rate": 4.751191198128736e-05, + "loss": 1.8063, + "step": 29220 + }, + { + "epoch": 2.53, + "learning_rate": 4.751104565537555e-05, + "loss": 1.7717, + "step": 29230 + }, + { + "epoch": 2.53, + "learning_rate": 4.7510179329463745e-05, + "loss": 1.804, + "step": 29240 + }, + { + "epoch": 2.53, + "learning_rate": 4.750931300355194e-05, + "loss": 1.7804, + "step": 29250 + }, + { + "epoch": 2.53, + "learning_rate": 4.750844667764013e-05, + "loss": 1.7957, + "step": 29260 + }, + { + "epoch": 2.53, + "learning_rate": 4.750758035172832e-05, + "loss": 1.753, + "step": 29270 + }, + { + "epoch": 2.53, + "learning_rate": 4.750671402581651e-05, + "loss": 1.8668, + "step": 29280 + }, + { + "epoch": 2.54, + "learning_rate": 4.7505847699904704e-05, + "loss": 1.8266, + "step": 29290 + }, + { + "epoch": 2.54, + "learning_rate": 4.75049813739929e-05, + "loss": 1.8176, + "step": 29300 + }, + { + "epoch": 2.54, + "learning_rate": 4.7504115048081086e-05, + "loss": 1.8104, + "step": 29310 + }, + { + "epoch": 2.54, + "learning_rate": 4.750324872216928e-05, + "loss": 1.7564, + "step": 29320 + }, + { + "epoch": 2.54, + "learning_rate": 4.7502382396257475e-05, + "loss": 1.7625, + "step": 29330 + }, + { + "epoch": 2.54, + "learning_rate": 4.750151607034566e-05, + "loss": 1.7417, + "step": 29340 + }, + { + "epoch": 2.54, + "learning_rate": 4.750064974443386e-05, + "loss": 1.7333, + "step": 29350 + }, + { + "epoch": 2.54, + "learning_rate": 4.749978341852205e-05, + "loss": 1.8017, + "step": 29360 + }, + { + "epoch": 2.54, + "learning_rate": 4.749891709261024e-05, + "loss": 1.7886, + "step": 29370 + }, + { + "epoch": 2.54, + "learning_rate": 4.749805076669843e-05, + "loss": 1.791, + "step": 29380 + }, + { + "epoch": 2.54, + "learning_rate": 4.749718444078662e-05, + "loss": 1.8713, + "step": 29390 + }, + { + "epoch": 2.54, + "learning_rate": 4.7496318114874815e-05, + "loss": 1.7409, + "step": 29400 + }, + { + "epoch": 2.55, + "learning_rate": 4.749545178896301e-05, + "loss": 1.7595, + "step": 29410 + }, + { + "epoch": 2.55, + "learning_rate": 4.74945854630512e-05, + "loss": 1.8071, + "step": 29420 + }, + { + "epoch": 2.55, + "learning_rate": 4.749371913713939e-05, + "loss": 1.7821, + "step": 29430 + }, + { + "epoch": 2.55, + "learning_rate": 4.7492852811227587e-05, + "loss": 1.7387, + "step": 29440 + }, + { + "epoch": 2.55, + "learning_rate": 4.7491986485315774e-05, + "loss": 1.7863, + "step": 29450 + }, + { + "epoch": 2.55, + "learning_rate": 4.749112015940397e-05, + "loss": 1.7402, + "step": 29460 + }, + { + "epoch": 2.55, + "learning_rate": 4.749025383349216e-05, + "loss": 1.77, + "step": 29470 + }, + { + "epoch": 2.55, + "learning_rate": 4.748938750758035e-05, + "loss": 1.7581, + "step": 29480 + }, + { + "epoch": 2.55, + "learning_rate": 4.7488521181668545e-05, + "loss": 1.8278, + "step": 29490 + }, + { + "epoch": 2.55, + "learning_rate": 4.748765485575673e-05, + "loss": 1.83, + "step": 29500 + }, + { + "epoch": 2.55, + "learning_rate": 4.748678852984493e-05, + "loss": 1.8163, + "step": 29510 + }, + { + "epoch": 2.56, + "learning_rate": 4.748592220393312e-05, + "loss": 1.7913, + "step": 29520 + }, + { + "epoch": 2.56, + "learning_rate": 4.748505587802131e-05, + "loss": 1.8188, + "step": 29530 + }, + { + "epoch": 2.56, + "learning_rate": 4.7484189552109504e-05, + "loss": 1.7991, + "step": 29540 + }, + { + "epoch": 2.56, + "learning_rate": 4.74833232261977e-05, + "loss": 1.8578, + "step": 29550 + }, + { + "epoch": 2.56, + "learning_rate": 4.7482456900285886e-05, + "loss": 1.8003, + "step": 29560 + }, + { + "epoch": 2.56, + "learning_rate": 4.748159057437408e-05, + "loss": 1.7945, + "step": 29570 + }, + { + "epoch": 2.56, + "learning_rate": 4.7480724248462275e-05, + "loss": 1.7854, + "step": 29580 + }, + { + "epoch": 2.56, + "learning_rate": 4.747985792255046e-05, + "loss": 1.788, + "step": 29590 + }, + { + "epoch": 2.56, + "learning_rate": 4.747899159663866e-05, + "loss": 1.7616, + "step": 29600 + }, + { + "epoch": 2.56, + "learning_rate": 4.7478125270726845e-05, + "loss": 1.7917, + "step": 29610 + }, + { + "epoch": 2.56, + "learning_rate": 4.747725894481504e-05, + "loss": 1.7195, + "step": 29620 + }, + { + "epoch": 2.56, + "learning_rate": 4.7476392618903234e-05, + "loss": 1.8157, + "step": 29630 + }, + { + "epoch": 2.57, + "learning_rate": 4.747552629299142e-05, + "loss": 1.7974, + "step": 29640 + }, + { + "epoch": 2.57, + "learning_rate": 4.7474659967079616e-05, + "loss": 1.7461, + "step": 29650 + }, + { + "epoch": 2.57, + "learning_rate": 4.747379364116781e-05, + "loss": 1.8365, + "step": 29660 + }, + { + "epoch": 2.57, + "learning_rate": 4.7472927315256e-05, + "loss": 1.8093, + "step": 29670 + }, + { + "epoch": 2.57, + "learning_rate": 4.747206098934419e-05, + "loss": 1.7598, + "step": 29680 + }, + { + "epoch": 2.57, + "learning_rate": 4.747119466343238e-05, + "loss": 1.8488, + "step": 29690 + }, + { + "epoch": 2.57, + "learning_rate": 4.7470328337520575e-05, + "loss": 1.7789, + "step": 29700 + }, + { + "epoch": 2.57, + "learning_rate": 4.746946201160877e-05, + "loss": 1.7681, + "step": 29710 + }, + { + "epoch": 2.57, + "learning_rate": 4.746859568569696e-05, + "loss": 1.8017, + "step": 29720 + }, + { + "epoch": 2.57, + "learning_rate": 4.746772935978515e-05, + "loss": 1.7955, + "step": 29730 + }, + { + "epoch": 2.57, + "learning_rate": 4.7466863033873346e-05, + "loss": 1.7744, + "step": 29740 + }, + { + "epoch": 2.57, + "learning_rate": 4.746599670796153e-05, + "loss": 1.7948, + "step": 29750 + }, + { + "epoch": 2.58, + "learning_rate": 4.746513038204973e-05, + "loss": 1.8156, + "step": 29760 + }, + { + "epoch": 2.58, + "learning_rate": 4.746426405613792e-05, + "loss": 1.8251, + "step": 29770 + }, + { + "epoch": 2.58, + "learning_rate": 4.746339773022611e-05, + "loss": 1.7586, + "step": 29780 + }, + { + "epoch": 2.58, + "learning_rate": 4.7462531404314304e-05, + "loss": 1.7931, + "step": 29790 + }, + { + "epoch": 2.58, + "learning_rate": 4.746166507840249e-05, + "loss": 1.7627, + "step": 29800 + }, + { + "epoch": 2.58, + "learning_rate": 4.7460798752490686e-05, + "loss": 1.7807, + "step": 29810 + }, + { + "epoch": 2.58, + "learning_rate": 4.745993242657888e-05, + "loss": 1.7941, + "step": 29820 + }, + { + "epoch": 2.58, + "learning_rate": 4.745906610066707e-05, + "loss": 1.7776, + "step": 29830 + }, + { + "epoch": 2.58, + "learning_rate": 4.745819977475526e-05, + "loss": 1.8302, + "step": 29840 + }, + { + "epoch": 2.58, + "learning_rate": 4.745733344884346e-05, + "loss": 1.811, + "step": 29850 + }, + { + "epoch": 2.58, + "learning_rate": 4.7456467122931645e-05, + "loss": 1.6828, + "step": 29860 + }, + { + "epoch": 2.59, + "learning_rate": 4.745560079701984e-05, + "loss": 1.7394, + "step": 29870 + }, + { + "epoch": 2.59, + "learning_rate": 4.7454734471108034e-05, + "loss": 1.7354, + "step": 29880 + }, + { + "epoch": 2.59, + "learning_rate": 4.745386814519622e-05, + "loss": 1.7989, + "step": 29890 + }, + { + "epoch": 2.59, + "learning_rate": 4.7453001819284416e-05, + "loss": 1.8217, + "step": 29900 + }, + { + "epoch": 2.59, + "learning_rate": 4.7452135493372604e-05, + "loss": 1.7427, + "step": 29910 + }, + { + "epoch": 2.59, + "learning_rate": 4.74512691674608e-05, + "loss": 1.8235, + "step": 29920 + }, + { + "epoch": 2.59, + "learning_rate": 4.745040284154899e-05, + "loss": 1.8618, + "step": 29930 + }, + { + "epoch": 2.59, + "learning_rate": 4.744953651563718e-05, + "loss": 1.815, + "step": 29940 + }, + { + "epoch": 2.59, + "learning_rate": 4.7448670189725375e-05, + "loss": 1.8151, + "step": 29950 + }, + { + "epoch": 2.59, + "learning_rate": 4.744780386381357e-05, + "loss": 1.8058, + "step": 29960 + }, + { + "epoch": 2.59, + "learning_rate": 4.744693753790176e-05, + "loss": 1.7385, + "step": 29970 + }, + { + "epoch": 2.59, + "learning_rate": 4.744607121198995e-05, + "loss": 1.806, + "step": 29980 + }, + { + "epoch": 2.6, + "learning_rate": 4.7445204886078146e-05, + "loss": 1.7787, + "step": 29990 + }, + { + "epoch": 2.6, + "learning_rate": 4.7444338560166334e-05, + "loss": 1.8409, + "step": 30000 + }, + { + "epoch": 2.6, + "learning_rate": 4.744347223425453e-05, + "loss": 1.7893, + "step": 30010 + }, + { + "epoch": 2.6, + "learning_rate": 4.7442605908342716e-05, + "loss": 1.7438, + "step": 30020 + }, + { + "epoch": 2.6, + "learning_rate": 4.744173958243091e-05, + "loss": 1.7646, + "step": 30030 + }, + { + "epoch": 2.6, + "learning_rate": 4.7440873256519105e-05, + "loss": 1.7859, + "step": 30040 + }, + { + "epoch": 2.6, + "learning_rate": 4.744000693060729e-05, + "loss": 1.8098, + "step": 30050 + }, + { + "epoch": 2.6, + "learning_rate": 4.743914060469549e-05, + "loss": 1.7829, + "step": 30060 + }, + { + "epoch": 2.6, + "learning_rate": 4.743827427878368e-05, + "loss": 1.8016, + "step": 30070 + }, + { + "epoch": 2.6, + "learning_rate": 4.743740795287187e-05, + "loss": 1.7658, + "step": 30080 + }, + { + "epoch": 2.6, + "learning_rate": 4.743654162696006e-05, + "loss": 1.7743, + "step": 30090 + }, + { + "epoch": 2.61, + "learning_rate": 4.743567530104826e-05, + "loss": 1.7979, + "step": 30100 + }, + { + "epoch": 2.61, + "learning_rate": 4.7434808975136445e-05, + "loss": 1.8104, + "step": 30110 + }, + { + "epoch": 2.61, + "learning_rate": 4.743394264922464e-05, + "loss": 1.7668, + "step": 30120 + }, + { + "epoch": 2.61, + "learning_rate": 4.743307632331283e-05, + "loss": 1.7327, + "step": 30130 + }, + { + "epoch": 2.61, + "learning_rate": 4.743220999740102e-05, + "loss": 1.8103, + "step": 30140 + }, + { + "epoch": 2.61, + "learning_rate": 4.7431343671489216e-05, + "loss": 1.757, + "step": 30150 + }, + { + "epoch": 2.61, + "learning_rate": 4.7430477345577404e-05, + "loss": 1.7612, + "step": 30160 + }, + { + "epoch": 2.61, + "learning_rate": 4.74296110196656e-05, + "loss": 1.738, + "step": 30170 + }, + { + "epoch": 2.61, + "learning_rate": 4.742874469375379e-05, + "loss": 1.7568, + "step": 30180 + }, + { + "epoch": 2.61, + "learning_rate": 4.742787836784198e-05, + "loss": 1.8054, + "step": 30190 + }, + { + "epoch": 2.61, + "learning_rate": 4.7427012041930175e-05, + "loss": 1.784, + "step": 30200 + }, + { + "epoch": 2.61, + "learning_rate": 4.742614571601837e-05, + "loss": 1.7996, + "step": 30210 + }, + { + "epoch": 2.62, + "learning_rate": 4.742527939010656e-05, + "loss": 1.8625, + "step": 30220 + }, + { + "epoch": 2.62, + "learning_rate": 4.742441306419475e-05, + "loss": 1.7895, + "step": 30230 + }, + { + "epoch": 2.62, + "learning_rate": 4.742354673828294e-05, + "loss": 1.8327, + "step": 30240 + }, + { + "epoch": 2.62, + "learning_rate": 4.7422680412371134e-05, + "loss": 1.8245, + "step": 30250 + }, + { + "epoch": 2.62, + "learning_rate": 4.742181408645933e-05, + "loss": 1.7545, + "step": 30260 + }, + { + "epoch": 2.62, + "learning_rate": 4.7420947760547516e-05, + "loss": 1.8241, + "step": 30270 + }, + { + "epoch": 2.62, + "learning_rate": 4.742008143463571e-05, + "loss": 1.8235, + "step": 30280 + }, + { + "epoch": 2.62, + "learning_rate": 4.7419215108723905e-05, + "loss": 1.8383, + "step": 30290 + }, + { + "epoch": 2.62, + "learning_rate": 4.741834878281209e-05, + "loss": 1.7939, + "step": 30300 + }, + { + "epoch": 2.62, + "learning_rate": 4.741748245690029e-05, + "loss": 1.8132, + "step": 30310 + }, + { + "epoch": 2.62, + "learning_rate": 4.741661613098848e-05, + "loss": 1.8463, + "step": 30320 + }, + { + "epoch": 2.63, + "learning_rate": 4.741574980507667e-05, + "loss": 1.7578, + "step": 30330 + }, + { + "epoch": 2.63, + "learning_rate": 4.7414883479164864e-05, + "loss": 1.8023, + "step": 30340 + }, + { + "epoch": 2.63, + "learning_rate": 4.741401715325305e-05, + "loss": 1.8046, + "step": 30350 + }, + { + "epoch": 2.63, + "learning_rate": 4.7413150827341246e-05, + "loss": 1.7932, + "step": 30360 + }, + { + "epoch": 2.63, + "learning_rate": 4.741228450142944e-05, + "loss": 1.7789, + "step": 30370 + }, + { + "epoch": 2.63, + "learning_rate": 4.741141817551763e-05, + "loss": 1.7907, + "step": 30380 + }, + { + "epoch": 2.63, + "learning_rate": 4.741055184960582e-05, + "loss": 1.7832, + "step": 30390 + }, + { + "epoch": 2.63, + "learning_rate": 4.740968552369402e-05, + "loss": 1.8371, + "step": 30400 + }, + { + "epoch": 2.63, + "learning_rate": 4.7408819197782205e-05, + "loss": 1.7548, + "step": 30410 + }, + { + "epoch": 2.63, + "learning_rate": 4.74079528718704e-05, + "loss": 1.799, + "step": 30420 + }, + { + "epoch": 2.63, + "learning_rate": 4.740708654595859e-05, + "loss": 1.7804, + "step": 30430 + }, + { + "epoch": 2.63, + "learning_rate": 4.740622022004678e-05, + "loss": 1.8043, + "step": 30440 + }, + { + "epoch": 2.64, + "learning_rate": 4.7405353894134976e-05, + "loss": 1.7605, + "step": 30450 + }, + { + "epoch": 2.64, + "learning_rate": 4.740448756822316e-05, + "loss": 1.7771, + "step": 30460 + }, + { + "epoch": 2.64, + "learning_rate": 4.740362124231136e-05, + "loss": 1.8217, + "step": 30470 + }, + { + "epoch": 2.64, + "learning_rate": 4.740275491639955e-05, + "loss": 1.8134, + "step": 30480 + }, + { + "epoch": 2.64, + "learning_rate": 4.740188859048774e-05, + "loss": 1.7582, + "step": 30490 + }, + { + "epoch": 2.64, + "learning_rate": 4.7401022264575934e-05, + "loss": 1.7542, + "step": 30500 + }, + { + "epoch": 2.64, + "learning_rate": 4.740015593866413e-05, + "loss": 1.796, + "step": 30510 + }, + { + "epoch": 2.64, + "learning_rate": 4.7399289612752316e-05, + "loss": 1.7934, + "step": 30520 + }, + { + "epoch": 2.64, + "learning_rate": 4.739842328684051e-05, + "loss": 1.7948, + "step": 30530 + }, + { + "epoch": 2.64, + "learning_rate": 4.73975569609287e-05, + "loss": 1.7796, + "step": 30540 + }, + { + "epoch": 2.64, + "learning_rate": 4.739669063501689e-05, + "loss": 1.7806, + "step": 30550 + }, + { + "epoch": 2.65, + "learning_rate": 4.739582430910509e-05, + "loss": 1.7436, + "step": 30560 + }, + { + "epoch": 2.65, + "learning_rate": 4.7394957983193275e-05, + "loss": 1.7654, + "step": 30570 + }, + { + "epoch": 2.65, + "learning_rate": 4.739409165728147e-05, + "loss": 1.7766, + "step": 30580 + }, + { + "epoch": 2.65, + "learning_rate": 4.7393225331369664e-05, + "loss": 1.7514, + "step": 30590 + }, + { + "epoch": 2.65, + "learning_rate": 4.739235900545785e-05, + "loss": 1.8014, + "step": 30600 + }, + { + "epoch": 2.65, + "learning_rate": 4.7391492679546046e-05, + "loss": 1.7538, + "step": 30610 + }, + { + "epoch": 2.65, + "learning_rate": 4.739062635363424e-05, + "loss": 1.7617, + "step": 30620 + }, + { + "epoch": 2.65, + "learning_rate": 4.738976002772243e-05, + "loss": 1.7916, + "step": 30630 + }, + { + "epoch": 2.65, + "learning_rate": 4.738889370181062e-05, + "loss": 1.7703, + "step": 30640 + }, + { + "epoch": 2.65, + "learning_rate": 4.738802737589881e-05, + "loss": 1.7522, + "step": 30650 + }, + { + "epoch": 2.65, + "learning_rate": 4.7387161049987005e-05, + "loss": 1.7235, + "step": 30660 + }, + { + "epoch": 2.65, + "learning_rate": 4.73862947240752e-05, + "loss": 1.7897, + "step": 30670 + }, + { + "epoch": 2.66, + "learning_rate": 4.738542839816339e-05, + "loss": 1.7684, + "step": 30680 + }, + { + "epoch": 2.66, + "learning_rate": 4.738456207225158e-05, + "loss": 1.7728, + "step": 30690 + }, + { + "epoch": 2.66, + "learning_rate": 4.7383695746339776e-05, + "loss": 1.7711, + "step": 30700 + }, + { + "epoch": 2.66, + "learning_rate": 4.7382829420427964e-05, + "loss": 1.7731, + "step": 30710 + }, + { + "epoch": 2.66, + "learning_rate": 4.738196309451616e-05, + "loss": 1.7559, + "step": 30720 + }, + { + "epoch": 2.66, + "learning_rate": 4.738109676860435e-05, + "loss": 1.7948, + "step": 30730 + }, + { + "epoch": 2.66, + "learning_rate": 4.738023044269254e-05, + "loss": 1.8031, + "step": 30740 + }, + { + "epoch": 2.66, + "learning_rate": 4.7379364116780735e-05, + "loss": 1.8144, + "step": 30750 + }, + { + "epoch": 2.66, + "learning_rate": 4.737849779086892e-05, + "loss": 1.877, + "step": 30760 + }, + { + "epoch": 2.66, + "learning_rate": 4.737763146495712e-05, + "loss": 1.783, + "step": 30770 + }, + { + "epoch": 2.66, + "learning_rate": 4.737676513904531e-05, + "loss": 1.7881, + "step": 30780 + }, + { + "epoch": 2.66, + "learning_rate": 4.73758988131335e-05, + "loss": 1.7284, + "step": 30790 + }, + { + "epoch": 2.67, + "learning_rate": 4.737503248722169e-05, + "loss": 1.7427, + "step": 30800 + }, + { + "epoch": 2.67, + "learning_rate": 4.737416616130989e-05, + "loss": 1.7057, + "step": 30810 + }, + { + "epoch": 2.67, + "learning_rate": 4.7373299835398075e-05, + "loss": 1.8176, + "step": 30820 + }, + { + "epoch": 2.67, + "learning_rate": 4.737243350948627e-05, + "loss": 1.7333, + "step": 30830 + }, + { + "epoch": 2.67, + "learning_rate": 4.7371567183574464e-05, + "loss": 1.8425, + "step": 30840 + }, + { + "epoch": 2.67, + "learning_rate": 4.737070085766265e-05, + "loss": 1.8008, + "step": 30850 + }, + { + "epoch": 2.67, + "learning_rate": 4.7369834531750846e-05, + "loss": 1.8449, + "step": 30860 + }, + { + "epoch": 2.67, + "learning_rate": 4.7368968205839034e-05, + "loss": 1.7543, + "step": 30870 + }, + { + "epoch": 2.67, + "learning_rate": 4.736810187992723e-05, + "loss": 1.7367, + "step": 30880 + }, + { + "epoch": 2.67, + "learning_rate": 4.736723555401542e-05, + "loss": 1.7571, + "step": 30890 + }, + { + "epoch": 2.67, + "learning_rate": 4.736636922810361e-05, + "loss": 1.8455, + "step": 30900 + }, + { + "epoch": 2.68, + "learning_rate": 4.7365502902191805e-05, + "loss": 1.7911, + "step": 30910 + }, + { + "epoch": 2.68, + "learning_rate": 4.736463657628e-05, + "loss": 1.7245, + "step": 30920 + }, + { + "epoch": 2.68, + "learning_rate": 4.736377025036819e-05, + "loss": 1.7927, + "step": 30930 + }, + { + "epoch": 2.68, + "learning_rate": 4.736290392445638e-05, + "loss": 1.7952, + "step": 30940 + }, + { + "epoch": 2.68, + "learning_rate": 4.7362037598544576e-05, + "loss": 1.7516, + "step": 30950 + }, + { + "epoch": 2.68, + "learning_rate": 4.7361171272632764e-05, + "loss": 1.7585, + "step": 30960 + }, + { + "epoch": 2.68, + "learning_rate": 4.736030494672096e-05, + "loss": 1.7387, + "step": 30970 + }, + { + "epoch": 2.68, + "learning_rate": 4.7359438620809146e-05, + "loss": 1.732, + "step": 30980 + }, + { + "epoch": 2.68, + "learning_rate": 4.735857229489734e-05, + "loss": 1.7685, + "step": 30990 + }, + { + "epoch": 2.68, + "learning_rate": 4.7357705968985535e-05, + "loss": 1.7164, + "step": 31000 + }, + { + "epoch": 2.68, + "learning_rate": 4.735683964307372e-05, + "loss": 1.7397, + "step": 31010 + }, + { + "epoch": 2.68, + "learning_rate": 4.735597331716192e-05, + "loss": 1.7853, + "step": 31020 + }, + { + "epoch": 2.69, + "learning_rate": 4.735510699125011e-05, + "loss": 1.7408, + "step": 31030 + }, + { + "epoch": 2.69, + "learning_rate": 4.73542406653383e-05, + "loss": 1.7931, + "step": 31040 + }, + { + "epoch": 2.69, + "learning_rate": 4.7353374339426494e-05, + "loss": 1.809, + "step": 31050 + }, + { + "epoch": 2.69, + "learning_rate": 4.735250801351469e-05, + "loss": 1.7946, + "step": 31060 + }, + { + "epoch": 2.69, + "learning_rate": 4.7351641687602876e-05, + "loss": 1.7857, + "step": 31070 + }, + { + "epoch": 2.69, + "learning_rate": 4.735077536169107e-05, + "loss": 1.7924, + "step": 31080 + }, + { + "epoch": 2.69, + "learning_rate": 4.734990903577926e-05, + "loss": 1.7672, + "step": 31090 + }, + { + "epoch": 2.69, + "learning_rate": 4.734904270986745e-05, + "loss": 1.7557, + "step": 31100 + }, + { + "epoch": 2.69, + "learning_rate": 4.734817638395565e-05, + "loss": 1.7473, + "step": 31110 + }, + { + "epoch": 2.69, + "learning_rate": 4.7347310058043834e-05, + "loss": 1.834, + "step": 31120 + }, + { + "epoch": 2.69, + "learning_rate": 4.734644373213203e-05, + "loss": 1.7109, + "step": 31130 + }, + { + "epoch": 2.7, + "learning_rate": 4.734557740622022e-05, + "loss": 1.7541, + "step": 31140 + }, + { + "epoch": 2.7, + "learning_rate": 4.734471108030841e-05, + "loss": 1.795, + "step": 31150 + }, + { + "epoch": 2.7, + "learning_rate": 4.7343844754396606e-05, + "loss": 1.7887, + "step": 31160 + }, + { + "epoch": 2.7, + "learning_rate": 4.73429784284848e-05, + "loss": 1.7989, + "step": 31170 + }, + { + "epoch": 2.7, + "learning_rate": 4.734211210257299e-05, + "loss": 1.7151, + "step": 31180 + }, + { + "epoch": 2.7, + "learning_rate": 4.734124577666118e-05, + "loss": 1.7449, + "step": 31190 + }, + { + "epoch": 2.7, + "learning_rate": 4.734037945074937e-05, + "loss": 1.7637, + "step": 31200 + }, + { + "epoch": 2.7, + "learning_rate": 4.7339513124837564e-05, + "loss": 1.7257, + "step": 31210 + }, + { + "epoch": 2.7, + "learning_rate": 4.733864679892576e-05, + "loss": 1.8039, + "step": 31220 + }, + { + "epoch": 2.7, + "learning_rate": 4.7337780473013946e-05, + "loss": 1.7611, + "step": 31230 + }, + { + "epoch": 2.7, + "learning_rate": 4.733691414710214e-05, + "loss": 1.7709, + "step": 31240 + }, + { + "epoch": 2.7, + "learning_rate": 4.7336047821190335e-05, + "loss": 1.7507, + "step": 31250 + }, + { + "epoch": 2.71, + "learning_rate": 4.733518149527852e-05, + "loss": 1.727, + "step": 31260 + }, + { + "epoch": 2.71, + "learning_rate": 4.733431516936672e-05, + "loss": 1.7732, + "step": 31270 + }, + { + "epoch": 2.71, + "learning_rate": 4.7333448843454905e-05, + "loss": 1.8156, + "step": 31280 + }, + { + "epoch": 2.71, + "learning_rate": 4.73325825175431e-05, + "loss": 1.7555, + "step": 31290 + }, + { + "epoch": 2.71, + "learning_rate": 4.7331716191631294e-05, + "loss": 1.786, + "step": 31300 + }, + { + "epoch": 2.71, + "learning_rate": 4.733084986571948e-05, + "loss": 1.7582, + "step": 31310 + }, + { + "epoch": 2.71, + "learning_rate": 4.7329983539807676e-05, + "loss": 1.7862, + "step": 31320 + }, + { + "epoch": 2.71, + "learning_rate": 4.732911721389587e-05, + "loss": 1.7363, + "step": 31330 + }, + { + "epoch": 2.71, + "learning_rate": 4.732825088798406e-05, + "loss": 1.8148, + "step": 31340 + }, + { + "epoch": 2.71, + "learning_rate": 4.732738456207225e-05, + "loss": 1.7977, + "step": 31350 + }, + { + "epoch": 2.71, + "learning_rate": 4.732651823616045e-05, + "loss": 1.8238, + "step": 31360 + }, + { + "epoch": 2.72, + "learning_rate": 4.7325651910248635e-05, + "loss": 1.6812, + "step": 31370 + }, + { + "epoch": 2.72, + "learning_rate": 4.732478558433683e-05, + "loss": 1.7694, + "step": 31380 + }, + { + "epoch": 2.72, + "learning_rate": 4.732391925842502e-05, + "loss": 1.7639, + "step": 31390 + }, + { + "epoch": 2.72, + "learning_rate": 4.732305293251321e-05, + "loss": 1.7471, + "step": 31400 + }, + { + "epoch": 2.72, + "learning_rate": 4.7322186606601406e-05, + "loss": 1.7777, + "step": 31410 + }, + { + "epoch": 2.72, + "learning_rate": 4.7321320280689594e-05, + "loss": 1.83, + "step": 31420 + }, + { + "epoch": 2.72, + "learning_rate": 4.732045395477779e-05, + "loss": 1.7323, + "step": 31430 + }, + { + "epoch": 2.72, + "learning_rate": 4.731958762886598e-05, + "loss": 1.7304, + "step": 31440 + }, + { + "epoch": 2.72, + "learning_rate": 4.731872130295417e-05, + "loss": 1.8042, + "step": 31450 + }, + { + "epoch": 2.72, + "learning_rate": 4.7317854977042365e-05, + "loss": 1.7569, + "step": 31460 + }, + { + "epoch": 2.72, + "learning_rate": 4.731698865113056e-05, + "loss": 1.7996, + "step": 31470 + }, + { + "epoch": 2.72, + "learning_rate": 4.731612232521875e-05, + "loss": 1.6771, + "step": 31480 + }, + { + "epoch": 2.73, + "learning_rate": 4.731525599930694e-05, + "loss": 1.7787, + "step": 31490 + }, + { + "epoch": 2.73, + "learning_rate": 4.731438967339513e-05, + "loss": 1.7566, + "step": 31500 + }, + { + "epoch": 2.73, + "learning_rate": 4.731352334748332e-05, + "loss": 1.7235, + "step": 31510 + }, + { + "epoch": 2.73, + "learning_rate": 4.731265702157152e-05, + "loss": 1.7893, + "step": 31520 + }, + { + "epoch": 2.73, + "learning_rate": 4.7311790695659705e-05, + "loss": 1.8034, + "step": 31530 + }, + { + "epoch": 2.73, + "learning_rate": 4.73109243697479e-05, + "loss": 1.7044, + "step": 31540 + }, + { + "epoch": 2.73, + "learning_rate": 4.7310058043836094e-05, + "loss": 1.6761, + "step": 31550 + }, + { + "epoch": 2.73, + "learning_rate": 4.730919171792428e-05, + "loss": 1.759, + "step": 31560 + }, + { + "epoch": 2.73, + "learning_rate": 4.7308325392012476e-05, + "loss": 1.7857, + "step": 31570 + }, + { + "epoch": 2.73, + "learning_rate": 4.730745906610067e-05, + "loss": 1.7669, + "step": 31580 + }, + { + "epoch": 2.73, + "learning_rate": 4.730659274018886e-05, + "loss": 1.7544, + "step": 31590 + }, + { + "epoch": 2.74, + "learning_rate": 4.730572641427705e-05, + "loss": 1.7569, + "step": 31600 + }, + { + "epoch": 2.74, + "learning_rate": 4.730486008836524e-05, + "loss": 1.7412, + "step": 31610 + }, + { + "epoch": 2.74, + "learning_rate": 4.7303993762453435e-05, + "loss": 1.763, + "step": 31620 + }, + { + "epoch": 2.74, + "learning_rate": 4.730312743654163e-05, + "loss": 1.7267, + "step": 31630 + }, + { + "epoch": 2.74, + "learning_rate": 4.730226111062982e-05, + "loss": 1.7685, + "step": 31640 + }, + { + "epoch": 2.74, + "learning_rate": 4.730139478471801e-05, + "loss": 1.833, + "step": 31650 + }, + { + "epoch": 2.74, + "learning_rate": 4.7300528458806206e-05, + "loss": 1.7362, + "step": 31660 + }, + { + "epoch": 2.74, + "learning_rate": 4.7299662132894394e-05, + "loss": 1.7367, + "step": 31670 + }, + { + "epoch": 2.74, + "learning_rate": 4.729879580698259e-05, + "loss": 1.7332, + "step": 31680 + }, + { + "epoch": 2.74, + "learning_rate": 4.729792948107078e-05, + "loss": 1.7489, + "step": 31690 + }, + { + "epoch": 2.74, + "learning_rate": 4.729706315515897e-05, + "loss": 1.7206, + "step": 31700 + }, + { + "epoch": 2.74, + "learning_rate": 4.7296196829247165e-05, + "loss": 1.7787, + "step": 31710 + }, + { + "epoch": 2.75, + "learning_rate": 4.729533050333535e-05, + "loss": 1.809, + "step": 31720 + }, + { + "epoch": 2.75, + "learning_rate": 4.729446417742355e-05, + "loss": 1.8575, + "step": 31730 + }, + { + "epoch": 2.75, + "learning_rate": 4.729359785151174e-05, + "loss": 1.7884, + "step": 31740 + }, + { + "epoch": 2.75, + "learning_rate": 4.729273152559993e-05, + "loss": 1.7014, + "step": 31750 + }, + { + "epoch": 2.75, + "learning_rate": 4.7291865199688124e-05, + "loss": 1.7535, + "step": 31760 + }, + { + "epoch": 2.75, + "learning_rate": 4.729099887377632e-05, + "loss": 1.7443, + "step": 31770 + }, + { + "epoch": 2.75, + "learning_rate": 4.7290132547864506e-05, + "loss": 1.7366, + "step": 31780 + }, + { + "epoch": 2.75, + "learning_rate": 4.72892662219527e-05, + "loss": 1.7731, + "step": 31790 + }, + { + "epoch": 2.75, + "learning_rate": 4.7288399896040895e-05, + "loss": 1.8082, + "step": 31800 + }, + { + "epoch": 2.75, + "learning_rate": 4.728753357012908e-05, + "loss": 1.7731, + "step": 31810 + }, + { + "epoch": 2.75, + "learning_rate": 4.728666724421728e-05, + "loss": 1.7311, + "step": 31820 + }, + { + "epoch": 2.76, + "learning_rate": 4.7285800918305464e-05, + "loss": 1.749, + "step": 31830 + }, + { + "epoch": 2.76, + "learning_rate": 4.728493459239366e-05, + "loss": 1.8075, + "step": 31840 + }, + { + "epoch": 2.76, + "learning_rate": 4.728406826648185e-05, + "loss": 1.8278, + "step": 31850 + }, + { + "epoch": 2.76, + "learning_rate": 4.728320194057004e-05, + "loss": 1.7298, + "step": 31860 + }, + { + "epoch": 2.76, + "learning_rate": 4.7282335614658235e-05, + "loss": 1.7267, + "step": 31870 + }, + { + "epoch": 2.76, + "learning_rate": 4.728146928874643e-05, + "loss": 1.6903, + "step": 31880 + }, + { + "epoch": 2.76, + "learning_rate": 4.728060296283462e-05, + "loss": 1.764, + "step": 31890 + }, + { + "epoch": 2.76, + "learning_rate": 4.727973663692281e-05, + "loss": 1.7862, + "step": 31900 + }, + { + "epoch": 2.76, + "learning_rate": 4.7278870311011007e-05, + "loss": 1.8199, + "step": 31910 + }, + { + "epoch": 2.76, + "learning_rate": 4.7278003985099194e-05, + "loss": 1.8046, + "step": 31920 + }, + { + "epoch": 2.76, + "learning_rate": 4.727713765918739e-05, + "loss": 1.758, + "step": 31930 + }, + { + "epoch": 2.76, + "learning_rate": 4.7276271333275576e-05, + "loss": 1.7945, + "step": 31940 + }, + { + "epoch": 2.77, + "learning_rate": 4.727540500736377e-05, + "loss": 1.8339, + "step": 31950 + }, + { + "epoch": 2.77, + "learning_rate": 4.7274538681451965e-05, + "loss": 1.7977, + "step": 31960 + }, + { + "epoch": 2.77, + "learning_rate": 4.727367235554015e-05, + "loss": 1.7337, + "step": 31970 + }, + { + "epoch": 2.77, + "learning_rate": 4.727280602962835e-05, + "loss": 1.7679, + "step": 31980 + }, + { + "epoch": 2.77, + "learning_rate": 4.727193970371654e-05, + "loss": 1.7813, + "step": 31990 + }, + { + "epoch": 2.77, + "learning_rate": 4.727107337780473e-05, + "loss": 1.774, + "step": 32000 + }, + { + "epoch": 2.77, + "learning_rate": 4.7270207051892924e-05, + "loss": 1.7528, + "step": 32010 + }, + { + "epoch": 2.77, + "learning_rate": 4.726934072598111e-05, + "loss": 1.7359, + "step": 32020 + }, + { + "epoch": 2.77, + "learning_rate": 4.7268474400069306e-05, + "loss": 1.7392, + "step": 32030 + }, + { + "epoch": 2.77, + "learning_rate": 4.72676080741575e-05, + "loss": 1.7981, + "step": 32040 + }, + { + "epoch": 2.77, + "learning_rate": 4.726674174824569e-05, + "loss": 1.7986, + "step": 32050 + }, + { + "epoch": 2.77, + "learning_rate": 4.726587542233388e-05, + "loss": 1.7634, + "step": 32060 + }, + { + "epoch": 2.78, + "learning_rate": 4.726500909642208e-05, + "loss": 1.8306, + "step": 32070 + }, + { + "epoch": 2.78, + "learning_rate": 4.7264142770510265e-05, + "loss": 1.7291, + "step": 32080 + }, + { + "epoch": 2.78, + "learning_rate": 4.726327644459846e-05, + "loss": 1.7604, + "step": 32090 + }, + { + "epoch": 2.78, + "learning_rate": 4.7262410118686654e-05, + "loss": 1.7726, + "step": 32100 + }, + { + "epoch": 2.78, + "learning_rate": 4.726154379277484e-05, + "loss": 1.7746, + "step": 32110 + }, + { + "epoch": 2.78, + "learning_rate": 4.7260677466863036e-05, + "loss": 1.7328, + "step": 32120 + }, + { + "epoch": 2.78, + "learning_rate": 4.7259811140951223e-05, + "loss": 1.7167, + "step": 32130 + }, + { + "epoch": 2.78, + "learning_rate": 4.725894481503942e-05, + "loss": 1.8005, + "step": 32140 + }, + { + "epoch": 2.78, + "learning_rate": 4.725807848912761e-05, + "loss": 1.74, + "step": 32150 + }, + { + "epoch": 2.78, + "learning_rate": 4.72572121632158e-05, + "loss": 1.7419, + "step": 32160 + }, + { + "epoch": 2.78, + "learning_rate": 4.7256345837303995e-05, + "loss": 1.7355, + "step": 32170 + }, + { + "epoch": 2.79, + "learning_rate": 4.725547951139219e-05, + "loss": 1.6942, + "step": 32180 + }, + { + "epoch": 2.79, + "learning_rate": 4.725461318548038e-05, + "loss": 1.7783, + "step": 32190 + }, + { + "epoch": 2.79, + "learning_rate": 4.725374685956857e-05, + "loss": 1.6996, + "step": 32200 + }, + { + "epoch": 2.79, + "learning_rate": 4.7252880533656766e-05, + "loss": 1.7861, + "step": 32210 + }, + { + "epoch": 2.79, + "learning_rate": 4.725201420774495e-05, + "loss": 1.769, + "step": 32220 + }, + { + "epoch": 2.79, + "learning_rate": 4.725114788183315e-05, + "loss": 1.8018, + "step": 32230 + }, + { + "epoch": 2.79, + "learning_rate": 4.7250281555921335e-05, + "loss": 1.776, + "step": 32240 + }, + { + "epoch": 2.79, + "learning_rate": 4.724941523000953e-05, + "loss": 1.7606, + "step": 32250 + }, + { + "epoch": 2.79, + "learning_rate": 4.7248548904097724e-05, + "loss": 1.7827, + "step": 32260 + }, + { + "epoch": 2.79, + "learning_rate": 4.724768257818591e-05, + "loss": 1.7455, + "step": 32270 + }, + { + "epoch": 2.79, + "learning_rate": 4.7246816252274106e-05, + "loss": 1.7467, + "step": 32280 + }, + { + "epoch": 2.79, + "learning_rate": 4.72459499263623e-05, + "loss": 1.7564, + "step": 32290 + }, + { + "epoch": 2.8, + "learning_rate": 4.724508360045049e-05, + "loss": 1.7918, + "step": 32300 + }, + { + "epoch": 2.8, + "learning_rate": 4.724421727453868e-05, + "loss": 1.764, + "step": 32310 + }, + { + "epoch": 2.8, + "learning_rate": 4.724335094862688e-05, + "loss": 1.7381, + "step": 32320 + }, + { + "epoch": 2.8, + "learning_rate": 4.7242484622715065e-05, + "loss": 1.6731, + "step": 32330 + }, + { + "epoch": 2.8, + "learning_rate": 4.724161829680326e-05, + "loss": 1.7687, + "step": 32340 + }, + { + "epoch": 2.8, + "learning_rate": 4.724075197089145e-05, + "loss": 1.7561, + "step": 32350 + }, + { + "epoch": 2.8, + "learning_rate": 4.723988564497964e-05, + "loss": 1.7789, + "step": 32360 + }, + { + "epoch": 2.8, + "learning_rate": 4.7239019319067836e-05, + "loss": 1.7475, + "step": 32370 + }, + { + "epoch": 2.8, + "learning_rate": 4.7238152993156024e-05, + "loss": 1.7475, + "step": 32380 + }, + { + "epoch": 2.8, + "learning_rate": 4.723728666724422e-05, + "loss": 1.803, + "step": 32390 + }, + { + "epoch": 2.8, + "learning_rate": 4.723642034133241e-05, + "loss": 1.802, + "step": 32400 + }, + { + "epoch": 2.81, + "learning_rate": 4.72355540154206e-05, + "loss": 1.8043, + "step": 32410 + }, + { + "epoch": 2.81, + "learning_rate": 4.7234687689508795e-05, + "loss": 1.7371, + "step": 32420 + }, + { + "epoch": 2.81, + "learning_rate": 4.723382136359699e-05, + "loss": 1.7836, + "step": 32430 + }, + { + "epoch": 2.81, + "learning_rate": 4.723295503768518e-05, + "loss": 1.752, + "step": 32440 + }, + { + "epoch": 2.81, + "learning_rate": 4.723208871177337e-05, + "loss": 1.7508, + "step": 32450 + }, + { + "epoch": 2.81, + "learning_rate": 4.723122238586156e-05, + "loss": 1.7234, + "step": 32460 + }, + { + "epoch": 2.81, + "learning_rate": 4.7230356059949754e-05, + "loss": 1.7644, + "step": 32470 + }, + { + "epoch": 2.81, + "learning_rate": 4.722948973403795e-05, + "loss": 1.7764, + "step": 32480 + }, + { + "epoch": 2.81, + "learning_rate": 4.7228623408126136e-05, + "loss": 1.7477, + "step": 32490 + }, + { + "epoch": 2.81, + "learning_rate": 4.722775708221433e-05, + "loss": 1.7871, + "step": 32500 + }, + { + "epoch": 2.81, + "learning_rate": 4.7226890756302525e-05, + "loss": 1.7135, + "step": 32510 + }, + { + "epoch": 2.81, + "learning_rate": 4.722602443039071e-05, + "loss": 1.6944, + "step": 32520 + }, + { + "epoch": 2.82, + "learning_rate": 4.722515810447891e-05, + "loss": 1.731, + "step": 32530 + }, + { + "epoch": 2.82, + "learning_rate": 4.72242917785671e-05, + "loss": 1.7606, + "step": 32540 + }, + { + "epoch": 2.82, + "learning_rate": 4.722342545265529e-05, + "loss": 1.7774, + "step": 32550 + }, + { + "epoch": 2.82, + "learning_rate": 4.722255912674348e-05, + "loss": 1.7685, + "step": 32560 + }, + { + "epoch": 2.82, + "learning_rate": 4.722169280083167e-05, + "loss": 1.7203, + "step": 32570 + }, + { + "epoch": 2.82, + "learning_rate": 4.7220826474919865e-05, + "loss": 1.7029, + "step": 32580 + }, + { + "epoch": 2.82, + "learning_rate": 4.721996014900806e-05, + "loss": 1.7277, + "step": 32590 + }, + { + "epoch": 2.82, + "learning_rate": 4.721909382309625e-05, + "loss": 1.7642, + "step": 32600 + }, + { + "epoch": 2.82, + "learning_rate": 4.721822749718444e-05, + "loss": 1.8038, + "step": 32610 + }, + { + "epoch": 2.82, + "learning_rate": 4.7217361171272636e-05, + "loss": 1.7839, + "step": 32620 + }, + { + "epoch": 2.82, + "learning_rate": 4.7216494845360824e-05, + "loss": 1.7055, + "step": 32630 + }, + { + "epoch": 2.83, + "learning_rate": 4.721562851944902e-05, + "loss": 1.8281, + "step": 32640 + }, + { + "epoch": 2.83, + "learning_rate": 4.721476219353721e-05, + "loss": 1.7796, + "step": 32650 + }, + { + "epoch": 2.83, + "learning_rate": 4.72138958676254e-05, + "loss": 1.816, + "step": 32660 + }, + { + "epoch": 2.83, + "learning_rate": 4.7213029541713595e-05, + "loss": 1.6922, + "step": 32670 + }, + { + "epoch": 2.83, + "learning_rate": 4.721216321580178e-05, + "loss": 1.7788, + "step": 32680 + }, + { + "epoch": 2.83, + "learning_rate": 4.721129688988998e-05, + "loss": 1.7503, + "step": 32690 + }, + { + "epoch": 2.83, + "learning_rate": 4.721043056397817e-05, + "loss": 1.7413, + "step": 32700 + }, + { + "epoch": 2.83, + "learning_rate": 4.720956423806636e-05, + "loss": 1.7754, + "step": 32710 + }, + { + "epoch": 2.83, + "learning_rate": 4.7208697912154554e-05, + "loss": 1.7368, + "step": 32720 + }, + { + "epoch": 2.83, + "learning_rate": 4.720783158624275e-05, + "loss": 1.7949, + "step": 32730 + }, + { + "epoch": 2.83, + "learning_rate": 4.7206965260330936e-05, + "loss": 1.7669, + "step": 32740 + }, + { + "epoch": 2.83, + "learning_rate": 4.720609893441913e-05, + "loss": 1.8131, + "step": 32750 + }, + { + "epoch": 2.84, + "learning_rate": 4.720523260850732e-05, + "loss": 1.7584, + "step": 32760 + }, + { + "epoch": 2.84, + "learning_rate": 4.720436628259551e-05, + "loss": 1.7896, + "step": 32770 + }, + { + "epoch": 2.84, + "learning_rate": 4.720349995668371e-05, + "loss": 1.7583, + "step": 32780 + }, + { + "epoch": 2.84, + "learning_rate": 4.7202633630771895e-05, + "loss": 1.7796, + "step": 32790 + }, + { + "epoch": 2.84, + "learning_rate": 4.720176730486009e-05, + "loss": 1.7873, + "step": 32800 + }, + { + "epoch": 2.84, + "learning_rate": 4.7200900978948284e-05, + "loss": 1.7968, + "step": 32810 + }, + { + "epoch": 2.84, + "learning_rate": 4.720003465303647e-05, + "loss": 1.7382, + "step": 32820 + }, + { + "epoch": 2.84, + "learning_rate": 4.7199168327124666e-05, + "loss": 1.7156, + "step": 32830 + }, + { + "epoch": 2.84, + "learning_rate": 4.719830200121286e-05, + "loss": 1.7998, + "step": 32840 + }, + { + "epoch": 2.84, + "learning_rate": 4.719743567530105e-05, + "loss": 1.778, + "step": 32850 + }, + { + "epoch": 2.84, + "learning_rate": 4.719656934938924e-05, + "loss": 1.691, + "step": 32860 + }, + { + "epoch": 2.85, + "learning_rate": 4.719570302347743e-05, + "loss": 1.8193, + "step": 32870 + }, + { + "epoch": 2.85, + "learning_rate": 4.7194836697565625e-05, + "loss": 1.7707, + "step": 32880 + }, + { + "epoch": 2.85, + "learning_rate": 4.719397037165382e-05, + "loss": 1.7137, + "step": 32890 + }, + { + "epoch": 2.85, + "learning_rate": 4.719310404574201e-05, + "loss": 1.7293, + "step": 32900 + }, + { + "epoch": 2.85, + "learning_rate": 4.71922377198302e-05, + "loss": 1.7541, + "step": 32910 + }, + { + "epoch": 2.85, + "learning_rate": 4.7191371393918396e-05, + "loss": 1.712, + "step": 32920 + }, + { + "epoch": 2.85, + "learning_rate": 4.719050506800658e-05, + "loss": 1.7559, + "step": 32930 + }, + { + "epoch": 2.85, + "learning_rate": 4.718963874209478e-05, + "loss": 1.7358, + "step": 32940 + }, + { + "epoch": 2.85, + "learning_rate": 4.718877241618297e-05, + "loss": 1.7977, + "step": 32950 + }, + { + "epoch": 2.85, + "learning_rate": 4.718790609027116e-05, + "loss": 1.7234, + "step": 32960 + }, + { + "epoch": 2.85, + "learning_rate": 4.7187039764359354e-05, + "loss": 1.8105, + "step": 32970 + }, + { + "epoch": 2.85, + "learning_rate": 4.718617343844754e-05, + "loss": 1.8067, + "step": 32980 + }, + { + "epoch": 2.86, + "learning_rate": 4.7185307112535736e-05, + "loss": 1.8065, + "step": 32990 + }, + { + "epoch": 2.86, + "learning_rate": 4.718444078662393e-05, + "loss": 1.813, + "step": 33000 + }, + { + "epoch": 2.86, + "learning_rate": 4.718357446071212e-05, + "loss": 1.755, + "step": 33010 + }, + { + "epoch": 2.86, + "learning_rate": 4.718270813480031e-05, + "loss": 1.7871, + "step": 33020 + }, + { + "epoch": 2.86, + "learning_rate": 4.718184180888851e-05, + "loss": 1.7188, + "step": 33030 + }, + { + "epoch": 2.86, + "learning_rate": 4.7180975482976695e-05, + "loss": 1.7152, + "step": 33040 + }, + { + "epoch": 2.86, + "learning_rate": 4.718010915706489e-05, + "loss": 1.7757, + "step": 33050 + }, + { + "epoch": 2.86, + "learning_rate": 4.7179242831153084e-05, + "loss": 1.8081, + "step": 33060 + }, + { + "epoch": 2.86, + "learning_rate": 4.717837650524127e-05, + "loss": 1.6804, + "step": 33070 + }, + { + "epoch": 2.86, + "learning_rate": 4.7177510179329466e-05, + "loss": 1.7391, + "step": 33080 + }, + { + "epoch": 2.86, + "learning_rate": 4.7176643853417654e-05, + "loss": 1.7305, + "step": 33090 + }, + { + "epoch": 2.86, + "learning_rate": 4.717577752750585e-05, + "loss": 1.8009, + "step": 33100 + }, + { + "epoch": 2.87, + "learning_rate": 4.717491120159404e-05, + "loss": 1.7622, + "step": 33110 + }, + { + "epoch": 2.87, + "learning_rate": 4.717404487568223e-05, + "loss": 1.745, + "step": 33120 + }, + { + "epoch": 2.87, + "learning_rate": 4.7173178549770425e-05, + "loss": 1.7819, + "step": 33130 + }, + { + "epoch": 2.87, + "learning_rate": 4.717231222385862e-05, + "loss": 1.6973, + "step": 33140 + }, + { + "epoch": 2.87, + "learning_rate": 4.717144589794681e-05, + "loss": 1.783, + "step": 33150 + }, + { + "epoch": 2.87, + "learning_rate": 4.7170579572035e-05, + "loss": 1.7514, + "step": 33160 + }, + { + "epoch": 2.87, + "learning_rate": 4.7169713246123196e-05, + "loss": 1.7909, + "step": 33170 + }, + { + "epoch": 2.87, + "learning_rate": 4.7168846920211384e-05, + "loss": 1.7618, + "step": 33180 + }, + { + "epoch": 2.87, + "learning_rate": 4.716798059429958e-05, + "loss": 1.7482, + "step": 33190 + }, + { + "epoch": 2.87, + "learning_rate": 4.7167114268387766e-05, + "loss": 1.7694, + "step": 33200 + }, + { + "epoch": 2.87, + "learning_rate": 4.716624794247596e-05, + "loss": 1.7357, + "step": 33210 + }, + { + "epoch": 2.88, + "learning_rate": 4.7165381616564155e-05, + "loss": 1.7644, + "step": 33220 + }, + { + "epoch": 2.88, + "learning_rate": 4.716451529065234e-05, + "loss": 1.8028, + "step": 33230 + }, + { + "epoch": 2.88, + "learning_rate": 4.716364896474054e-05, + "loss": 1.7483, + "step": 33240 + }, + { + "epoch": 2.88, + "learning_rate": 4.716278263882873e-05, + "loss": 1.6772, + "step": 33250 + }, + { + "epoch": 2.88, + "learning_rate": 4.716191631291692e-05, + "loss": 1.743, + "step": 33260 + }, + { + "epoch": 2.88, + "learning_rate": 4.716104998700511e-05, + "loss": 1.7354, + "step": 33270 + }, + { + "epoch": 2.88, + "learning_rate": 4.716018366109331e-05, + "loss": 1.7689, + "step": 33280 + }, + { + "epoch": 2.88, + "learning_rate": 4.7159317335181495e-05, + "loss": 1.7604, + "step": 33290 + }, + { + "epoch": 2.88, + "learning_rate": 4.715845100926969e-05, + "loss": 1.7192, + "step": 33300 + }, + { + "epoch": 2.88, + "learning_rate": 4.715758468335788e-05, + "loss": 1.7391, + "step": 33310 + }, + { + "epoch": 2.88, + "learning_rate": 4.715671835744607e-05, + "loss": 1.7439, + "step": 33320 + }, + { + "epoch": 2.88, + "learning_rate": 4.7155852031534266e-05, + "loss": 1.7693, + "step": 33330 + }, + { + "epoch": 2.89, + "learning_rate": 4.7154985705622454e-05, + "loss": 1.7366, + "step": 33340 + }, + { + "epoch": 2.89, + "learning_rate": 4.715411937971065e-05, + "loss": 1.7776, + "step": 33350 + }, + { + "epoch": 2.89, + "learning_rate": 4.715325305379884e-05, + "loss": 1.7752, + "step": 33360 + }, + { + "epoch": 2.89, + "learning_rate": 4.715238672788703e-05, + "loss": 1.6863, + "step": 33370 + }, + { + "epoch": 2.89, + "learning_rate": 4.7151520401975225e-05, + "loss": 1.7448, + "step": 33380 + }, + { + "epoch": 2.89, + "learning_rate": 4.715065407606342e-05, + "loss": 1.7476, + "step": 33390 + }, + { + "epoch": 2.89, + "learning_rate": 4.714978775015161e-05, + "loss": 1.7097, + "step": 33400 + }, + { + "epoch": 2.89, + "learning_rate": 4.71489214242398e-05, + "loss": 1.7417, + "step": 33410 + }, + { + "epoch": 2.89, + "learning_rate": 4.714805509832799e-05, + "loss": 1.8071, + "step": 33420 + }, + { + "epoch": 2.89, + "learning_rate": 4.7147188772416184e-05, + "loss": 1.7391, + "step": 33430 + }, + { + "epoch": 2.89, + "learning_rate": 4.714632244650438e-05, + "loss": 1.7094, + "step": 33440 + }, + { + "epoch": 2.9, + "learning_rate": 4.7145456120592566e-05, + "loss": 1.7899, + "step": 33450 + }, + { + "epoch": 2.9, + "learning_rate": 4.714458979468076e-05, + "loss": 1.7706, + "step": 33460 + }, + { + "epoch": 2.9, + "learning_rate": 4.7143723468768955e-05, + "loss": 1.7656, + "step": 33470 + }, + { + "epoch": 2.9, + "learning_rate": 4.714285714285714e-05, + "loss": 1.7684, + "step": 33480 + }, + { + "epoch": 2.9, + "learning_rate": 4.714199081694534e-05, + "loss": 1.7806, + "step": 33490 + }, + { + "epoch": 2.9, + "learning_rate": 4.714112449103353e-05, + "loss": 1.7501, + "step": 33500 + }, + { + "epoch": 2.9, + "learning_rate": 4.714025816512172e-05, + "loss": 1.7607, + "step": 33510 + }, + { + "epoch": 2.9, + "learning_rate": 4.7139391839209914e-05, + "loss": 1.799, + "step": 33520 + }, + { + "epoch": 2.9, + "learning_rate": 4.71385255132981e-05, + "loss": 1.8033, + "step": 33530 + }, + { + "epoch": 2.9, + "learning_rate": 4.7137659187386296e-05, + "loss": 1.7685, + "step": 33540 + }, + { + "epoch": 2.9, + "learning_rate": 4.713679286147449e-05, + "loss": 1.7566, + "step": 33550 + }, + { + "epoch": 2.9, + "learning_rate": 4.713592653556268e-05, + "loss": 1.749, + "step": 33560 + }, + { + "epoch": 2.91, + "learning_rate": 4.713506020965087e-05, + "loss": 1.742, + "step": 33570 + }, + { + "epoch": 2.91, + "learning_rate": 4.713419388373907e-05, + "loss": 1.7827, + "step": 33580 + }, + { + "epoch": 2.91, + "learning_rate": 4.7133327557827254e-05, + "loss": 1.7495, + "step": 33590 + }, + { + "epoch": 2.91, + "learning_rate": 4.713246123191545e-05, + "loss": 1.7713, + "step": 33600 + }, + { + "epoch": 2.91, + "learning_rate": 4.7131594906003637e-05, + "loss": 1.738, + "step": 33610 + }, + { + "epoch": 2.91, + "learning_rate": 4.713072858009183e-05, + "loss": 1.7424, + "step": 33620 + }, + { + "epoch": 2.91, + "learning_rate": 4.7129862254180026e-05, + "loss": 1.7721, + "step": 33630 + }, + { + "epoch": 2.91, + "learning_rate": 4.712899592826821e-05, + "loss": 1.7272, + "step": 33640 + }, + { + "epoch": 2.91, + "learning_rate": 4.712812960235641e-05, + "loss": 1.7455, + "step": 33650 + }, + { + "epoch": 2.91, + "learning_rate": 4.71272632764446e-05, + "loss": 1.715, + "step": 33660 + }, + { + "epoch": 2.91, + "learning_rate": 4.712639695053279e-05, + "loss": 1.7506, + "step": 33670 + }, + { + "epoch": 2.92, + "learning_rate": 4.7125530624620984e-05, + "loss": 1.7632, + "step": 33680 + }, + { + "epoch": 2.92, + "learning_rate": 4.712466429870918e-05, + "loss": 1.672, + "step": 33690 + }, + { + "epoch": 2.92, + "learning_rate": 4.7123797972797366e-05, + "loss": 1.763, + "step": 33700 + }, + { + "epoch": 2.92, + "learning_rate": 4.712293164688556e-05, + "loss": 1.7602, + "step": 33710 + }, + { + "epoch": 2.92, + "learning_rate": 4.712206532097375e-05, + "loss": 1.7304, + "step": 33720 + }, + { + "epoch": 2.92, + "learning_rate": 4.712119899506194e-05, + "loss": 1.7432, + "step": 33730 + }, + { + "epoch": 2.92, + "learning_rate": 4.712033266915014e-05, + "loss": 1.7663, + "step": 33740 + }, + { + "epoch": 2.92, + "learning_rate": 4.7119466343238325e-05, + "loss": 1.6765, + "step": 33750 + }, + { + "epoch": 2.92, + "learning_rate": 4.711860001732652e-05, + "loss": 1.781, + "step": 33760 + }, + { + "epoch": 2.92, + "learning_rate": 4.7117733691414714e-05, + "loss": 1.7316, + "step": 33770 + }, + { + "epoch": 2.92, + "learning_rate": 4.71168673655029e-05, + "loss": 1.7175, + "step": 33780 + }, + { + "epoch": 2.92, + "learning_rate": 4.7116001039591096e-05, + "loss": 1.6766, + "step": 33790 + }, + { + "epoch": 2.93, + "learning_rate": 4.711513471367929e-05, + "loss": 1.8253, + "step": 33800 + }, + { + "epoch": 2.93, + "learning_rate": 4.711426838776748e-05, + "loss": 1.7249, + "step": 33810 + }, + { + "epoch": 2.93, + "learning_rate": 4.711340206185567e-05, + "loss": 1.7053, + "step": 33820 + }, + { + "epoch": 2.93, + "learning_rate": 4.711253573594386e-05, + "loss": 1.7833, + "step": 33830 + }, + { + "epoch": 2.93, + "learning_rate": 4.7111669410032055e-05, + "loss": 1.7179, + "step": 33840 + }, + { + "epoch": 2.93, + "learning_rate": 4.711080308412025e-05, + "loss": 1.7977, + "step": 33850 + }, + { + "epoch": 2.93, + "learning_rate": 4.710993675820844e-05, + "loss": 1.7359, + "step": 33860 + }, + { + "epoch": 2.93, + "learning_rate": 4.710907043229663e-05, + "loss": 1.7297, + "step": 33870 + }, + { + "epoch": 2.93, + "learning_rate": 4.7108204106384826e-05, + "loss": 1.7186, + "step": 33880 + }, + { + "epoch": 2.93, + "learning_rate": 4.7107337780473014e-05, + "loss": 1.6933, + "step": 33890 + }, + { + "epoch": 2.93, + "learning_rate": 4.710647145456121e-05, + "loss": 1.7637, + "step": 33900 + }, + { + "epoch": 2.94, + "learning_rate": 4.71056051286494e-05, + "loss": 1.7061, + "step": 33910 + }, + { + "epoch": 2.94, + "learning_rate": 4.710473880273759e-05, + "loss": 1.6606, + "step": 33920 + }, + { + "epoch": 2.94, + "learning_rate": 4.7103872476825785e-05, + "loss": 1.706, + "step": 33930 + }, + { + "epoch": 2.94, + "learning_rate": 4.710300615091397e-05, + "loss": 1.7666, + "step": 33940 + }, + { + "epoch": 2.94, + "learning_rate": 4.710213982500217e-05, + "loss": 1.7654, + "step": 33950 + }, + { + "epoch": 2.94, + "learning_rate": 4.710127349909036e-05, + "loss": 1.6912, + "step": 33960 + }, + { + "epoch": 2.94, + "learning_rate": 4.710040717317855e-05, + "loss": 1.7954, + "step": 33970 + }, + { + "epoch": 2.94, + "learning_rate": 4.709954084726674e-05, + "loss": 1.744, + "step": 33980 + }, + { + "epoch": 2.94, + "learning_rate": 4.709867452135494e-05, + "loss": 1.6663, + "step": 33990 + }, + { + "epoch": 2.94, + "learning_rate": 4.7097808195443125e-05, + "loss": 1.7041, + "step": 34000 + }, + { + "epoch": 2.94, + "learning_rate": 4.709694186953132e-05, + "loss": 1.7682, + "step": 34010 + }, + { + "epoch": 2.94, + "learning_rate": 4.7096075543619514e-05, + "loss": 1.7504, + "step": 34020 + }, + { + "epoch": 2.95, + "learning_rate": 4.70952092177077e-05, + "loss": 1.696, + "step": 34030 + }, + { + "epoch": 2.95, + "learning_rate": 4.7094342891795896e-05, + "loss": 1.7383, + "step": 34040 + }, + { + "epoch": 2.95, + "learning_rate": 4.7093476565884084e-05, + "loss": 1.7588, + "step": 34050 + }, + { + "epoch": 2.95, + "learning_rate": 4.709261023997228e-05, + "loss": 1.7383, + "step": 34060 + }, + { + "epoch": 2.95, + "learning_rate": 4.709174391406047e-05, + "loss": 1.7661, + "step": 34070 + }, + { + "epoch": 2.95, + "learning_rate": 4.709087758814866e-05, + "loss": 1.7192, + "step": 34080 + }, + { + "epoch": 2.95, + "learning_rate": 4.7090011262236855e-05, + "loss": 1.6903, + "step": 34090 + }, + { + "epoch": 2.95, + "learning_rate": 4.708914493632505e-05, + "loss": 1.7794, + "step": 34100 + }, + { + "epoch": 2.95, + "learning_rate": 4.708827861041324e-05, + "loss": 1.7518, + "step": 34110 + }, + { + "epoch": 2.95, + "learning_rate": 4.708741228450143e-05, + "loss": 1.7679, + "step": 34120 + }, + { + "epoch": 2.95, + "learning_rate": 4.7086545958589626e-05, + "loss": 1.7923, + "step": 34130 + }, + { + "epoch": 2.95, + "learning_rate": 4.7085679632677814e-05, + "loss": 1.7648, + "step": 34140 + }, + { + "epoch": 2.96, + "learning_rate": 4.708481330676601e-05, + "loss": 1.6437, + "step": 34150 + }, + { + "epoch": 2.96, + "learning_rate": 4.7083946980854196e-05, + "loss": 1.805, + "step": 34160 + }, + { + "epoch": 2.96, + "learning_rate": 4.708308065494239e-05, + "loss": 1.6728, + "step": 34170 + }, + { + "epoch": 2.96, + "learning_rate": 4.7082214329030585e-05, + "loss": 1.7802, + "step": 34180 + }, + { + "epoch": 2.96, + "learning_rate": 4.708134800311877e-05, + "loss": 1.7532, + "step": 34190 + }, + { + "epoch": 2.96, + "learning_rate": 4.708048167720697e-05, + "loss": 1.718, + "step": 34200 + }, + { + "epoch": 2.96, + "learning_rate": 4.707961535129516e-05, + "loss": 1.731, + "step": 34210 + }, + { + "epoch": 2.96, + "learning_rate": 4.707874902538335e-05, + "loss": 1.6992, + "step": 34220 + }, + { + "epoch": 2.96, + "learning_rate": 4.7077882699471544e-05, + "loss": 1.6695, + "step": 34230 + }, + { + "epoch": 2.96, + "learning_rate": 4.707701637355974e-05, + "loss": 1.6865, + "step": 34240 + }, + { + "epoch": 2.96, + "learning_rate": 4.7076150047647926e-05, + "loss": 1.6562, + "step": 34250 + }, + { + "epoch": 2.97, + "learning_rate": 4.707528372173612e-05, + "loss": 1.8144, + "step": 34260 + }, + { + "epoch": 2.97, + "learning_rate": 4.707441739582431e-05, + "loss": 1.7107, + "step": 34270 + }, + { + "epoch": 2.97, + "learning_rate": 4.70735510699125e-05, + "loss": 1.7139, + "step": 34280 + }, + { + "epoch": 2.97, + "learning_rate": 4.70726847440007e-05, + "loss": 1.7673, + "step": 34290 + }, + { + "epoch": 2.97, + "learning_rate": 4.7071818418088884e-05, + "loss": 1.6999, + "step": 34300 + }, + { + "epoch": 2.97, + "learning_rate": 4.707095209217708e-05, + "loss": 1.7137, + "step": 34310 + }, + { + "epoch": 2.97, + "learning_rate": 4.707008576626527e-05, + "loss": 1.7185, + "step": 34320 + }, + { + "epoch": 2.97, + "learning_rate": 4.706921944035346e-05, + "loss": 1.7073, + "step": 34330 + }, + { + "epoch": 2.97, + "learning_rate": 4.7068353114441655e-05, + "loss": 1.7265, + "step": 34340 + }, + { + "epoch": 2.97, + "learning_rate": 4.706748678852984e-05, + "loss": 1.7236, + "step": 34350 + }, + { + "epoch": 2.97, + "learning_rate": 4.706662046261804e-05, + "loss": 1.7168, + "step": 34360 + }, + { + "epoch": 2.97, + "learning_rate": 4.706575413670623e-05, + "loss": 1.6792, + "step": 34370 + }, + { + "epoch": 2.98, + "learning_rate": 4.706488781079442e-05, + "loss": 1.7391, + "step": 34380 + }, + { + "epoch": 2.98, + "learning_rate": 4.7064021484882614e-05, + "loss": 1.7551, + "step": 34390 + }, + { + "epoch": 2.98, + "learning_rate": 4.706315515897081e-05, + "loss": 1.7312, + "step": 34400 + }, + { + "epoch": 2.98, + "learning_rate": 4.7062288833058996e-05, + "loss": 1.7605, + "step": 34410 + }, + { + "epoch": 2.98, + "learning_rate": 4.706142250714719e-05, + "loss": 1.7338, + "step": 34420 + }, + { + "epoch": 2.98, + "learning_rate": 4.7060556181235385e-05, + "loss": 1.7476, + "step": 34430 + }, + { + "epoch": 2.98, + "learning_rate": 4.705968985532357e-05, + "loss": 1.742, + "step": 34440 + }, + { + "epoch": 2.98, + "learning_rate": 4.705882352941177e-05, + "loss": 1.7175, + "step": 34450 + }, + { + "epoch": 2.98, + "learning_rate": 4.7057957203499955e-05, + "loss": 1.8225, + "step": 34460 + }, + { + "epoch": 2.98, + "learning_rate": 4.705709087758815e-05, + "loss": 1.7722, + "step": 34470 + }, + { + "epoch": 2.98, + "learning_rate": 4.7056224551676344e-05, + "loss": 1.7013, + "step": 34480 + }, + { + "epoch": 2.99, + "learning_rate": 4.705535822576453e-05, + "loss": 1.7639, + "step": 34490 + }, + { + "epoch": 2.99, + "learning_rate": 4.7054491899852726e-05, + "loss": 1.7415, + "step": 34500 + }, + { + "epoch": 2.99, + "learning_rate": 4.705362557394092e-05, + "loss": 1.7658, + "step": 34510 + }, + { + "epoch": 2.99, + "learning_rate": 4.705275924802911e-05, + "loss": 1.7094, + "step": 34520 + }, + { + "epoch": 2.99, + "learning_rate": 4.70518929221173e-05, + "loss": 1.6797, + "step": 34530 + }, + { + "epoch": 2.99, + "learning_rate": 4.70510265962055e-05, + "loss": 1.7834, + "step": 34540 + }, + { + "epoch": 2.99, + "learning_rate": 4.7050160270293685e-05, + "loss": 1.6782, + "step": 34550 + }, + { + "epoch": 2.99, + "learning_rate": 4.704929394438188e-05, + "loss": 1.747, + "step": 34560 + }, + { + "epoch": 2.99, + "learning_rate": 4.704842761847007e-05, + "loss": 1.7608, + "step": 34570 + }, + { + "epoch": 2.99, + "learning_rate": 4.704756129255826e-05, + "loss": 1.7, + "step": 34580 + }, + { + "epoch": 2.99, + "learning_rate": 4.7046694966646456e-05, + "loss": 1.7846, + "step": 34590 + }, + { + "epoch": 2.99, + "learning_rate": 4.7045828640734643e-05, + "loss": 1.7798, + "step": 34600 + }, + { + "epoch": 3.0, + "learning_rate": 4.704496231482284e-05, + "loss": 1.7215, + "step": 34610 + }, + { + "epoch": 3.0, + "learning_rate": 4.704409598891103e-05, + "loss": 1.7791, + "step": 34620 + }, + { + "epoch": 3.0, + "learning_rate": 4.704322966299922e-05, + "loss": 1.7554, + "step": 34630 + }, + { + "epoch": 3.0, + "learning_rate": 4.7042363337087415e-05, + "loss": 1.7551, + "step": 34640 + }, + { + "epoch": 3.0, + "learning_rate": 4.704149701117561e-05, + "loss": 1.7305, + "step": 34650 + }, + { + "epoch": 3.0, + "learning_rate": 4.70406306852638e-05, + "loss": 1.7399, + "step": 34660 + }, + { + "epoch": 3.0, + "eval_Bleu_1": 0.04409251055220211, + "eval_Bleu_2": 2.906372811931328e-11, + "eval_Bleu_3": 2.6182107580256e-14, + "eval_Bleu_4": 8.088627322656944e-16, + "eval_ROUGE_L": 0.11364360552892155, + "eval_cer": 0.9948227517820936, + "eval_em": 0, + "eval_f1": 0.04617902565586727, + "eval_loss": 1.638124942779541, + "eval_runtime": 1844.0203, + "eval_samples_per_second": 2.785, + "eval_steps_per_second": 2.785, + "eval_wer": 0.9887159115499172, + "step": 34660 + }, + { + "epoch": 3.0, + "learning_rate": 4.703976435935199e-05, + "loss": 1.7019, + "step": 34670 + }, + { + "epoch": 3.0, + "learning_rate": 4.703889803344018e-05, + "loss": 1.6865, + "step": 34680 + }, + { + "epoch": 3.0, + "learning_rate": 4.703803170752837e-05, + "loss": 1.7456, + "step": 34690 + }, + { + "epoch": 3.0, + "learning_rate": 4.703716538161657e-05, + "loss": 1.6829, + "step": 34700 + }, + { + "epoch": 3.0, + "learning_rate": 4.7036299055704755e-05, + "loss": 1.7912, + "step": 34710 + }, + { + "epoch": 3.01, + "learning_rate": 4.703543272979295e-05, + "loss": 1.7651, + "step": 34720 + }, + { + "epoch": 3.01, + "learning_rate": 4.7034566403881144e-05, + "loss": 1.7256, + "step": 34730 + }, + { + "epoch": 3.01, + "learning_rate": 4.703370007796933e-05, + "loss": 1.731, + "step": 34740 + }, + { + "epoch": 3.01, + "learning_rate": 4.7032833752057526e-05, + "loss": 1.7277, + "step": 34750 + }, + { + "epoch": 3.01, + "learning_rate": 4.703196742614572e-05, + "loss": 1.7101, + "step": 34760 + }, + { + "epoch": 3.01, + "learning_rate": 4.703110110023391e-05, + "loss": 1.6529, + "step": 34770 + }, + { + "epoch": 3.01, + "learning_rate": 4.70302347743221e-05, + "loss": 1.7291, + "step": 34780 + }, + { + "epoch": 3.01, + "learning_rate": 4.702936844841029e-05, + "loss": 1.7198, + "step": 34790 + }, + { + "epoch": 3.01, + "learning_rate": 4.7028502122498485e-05, + "loss": 1.6916, + "step": 34800 + }, + { + "epoch": 3.01, + "learning_rate": 4.702763579658668e-05, + "loss": 1.7626, + "step": 34810 + }, + { + "epoch": 3.01, + "learning_rate": 4.702676947067487e-05, + "loss": 1.7374, + "step": 34820 + }, + { + "epoch": 3.01, + "learning_rate": 4.702590314476306e-05, + "loss": 1.7784, + "step": 34830 + }, + { + "epoch": 3.02, + "learning_rate": 4.7025036818851256e-05, + "loss": 1.7076, + "step": 34840 + }, + { + "epoch": 3.02, + "learning_rate": 4.7024170492939444e-05, + "loss": 1.6528, + "step": 34850 + }, + { + "epoch": 3.02, + "learning_rate": 4.702330416702764e-05, + "loss": 1.7299, + "step": 34860 + }, + { + "epoch": 3.02, + "learning_rate": 4.702243784111583e-05, + "loss": 1.6761, + "step": 34870 + }, + { + "epoch": 3.02, + "learning_rate": 4.702157151520402e-05, + "loss": 1.727, + "step": 34880 + }, + { + "epoch": 3.02, + "learning_rate": 4.7020705189292215e-05, + "loss": 1.677, + "step": 34890 + }, + { + "epoch": 3.02, + "learning_rate": 4.70198388633804e-05, + "loss": 1.6925, + "step": 34900 + }, + { + "epoch": 3.02, + "learning_rate": 4.70189725374686e-05, + "loss": 1.704, + "step": 34910 + }, + { + "epoch": 3.02, + "learning_rate": 4.701810621155679e-05, + "loss": 1.7357, + "step": 34920 + }, + { + "epoch": 3.02, + "learning_rate": 4.701723988564498e-05, + "loss": 1.7081, + "step": 34930 + }, + { + "epoch": 3.02, + "learning_rate": 4.7016373559733174e-05, + "loss": 1.707, + "step": 34940 + }, + { + "epoch": 3.03, + "learning_rate": 4.701550723382137e-05, + "loss": 1.7414, + "step": 34950 + }, + { + "epoch": 3.03, + "learning_rate": 4.7014640907909556e-05, + "loss": 1.7443, + "step": 34960 + }, + { + "epoch": 3.03, + "learning_rate": 4.701377458199775e-05, + "loss": 1.6854, + "step": 34970 + }, + { + "epoch": 3.03, + "learning_rate": 4.7012908256085945e-05, + "loss": 1.677, + "step": 34980 + }, + { + "epoch": 3.03, + "learning_rate": 4.701204193017413e-05, + "loss": 1.7077, + "step": 34990 + }, + { + "epoch": 3.03, + "learning_rate": 4.701117560426233e-05, + "loss": 1.7112, + "step": 35000 + }, + { + "epoch": 3.03, + "learning_rate": 4.7010309278350514e-05, + "loss": 1.6785, + "step": 35010 + }, + { + "epoch": 3.03, + "learning_rate": 4.700944295243871e-05, + "loss": 1.6884, + "step": 35020 + }, + { + "epoch": 3.03, + "learning_rate": 4.70085766265269e-05, + "loss": 1.7042, + "step": 35030 + }, + { + "epoch": 3.03, + "learning_rate": 4.700771030061509e-05, + "loss": 1.7024, + "step": 35040 + }, + { + "epoch": 3.03, + "learning_rate": 4.7006843974703285e-05, + "loss": 1.708, + "step": 35050 + }, + { + "epoch": 3.03, + "learning_rate": 4.700597764879148e-05, + "loss": 1.7347, + "step": 35060 + }, + { + "epoch": 3.04, + "learning_rate": 4.700511132287967e-05, + "loss": 1.6932, + "step": 35070 + }, + { + "epoch": 3.04, + "learning_rate": 4.700424499696786e-05, + "loss": 1.6942, + "step": 35080 + }, + { + "epoch": 3.04, + "learning_rate": 4.700337867105605e-05, + "loss": 1.6659, + "step": 35090 + }, + { + "epoch": 3.04, + "learning_rate": 4.7002512345144244e-05, + "loss": 1.6591, + "step": 35100 + }, + { + "epoch": 3.04, + "learning_rate": 4.700164601923244e-05, + "loss": 1.7329, + "step": 35110 + }, + { + "epoch": 3.04, + "learning_rate": 4.7000779693320626e-05, + "loss": 1.7193, + "step": 35120 + }, + { + "epoch": 3.04, + "learning_rate": 4.699991336740882e-05, + "loss": 1.6917, + "step": 35130 + }, + { + "epoch": 3.04, + "learning_rate": 4.6999047041497015e-05, + "loss": 1.735, + "step": 35140 + }, + { + "epoch": 3.04, + "learning_rate": 4.69981807155852e-05, + "loss": 1.747, + "step": 35150 + }, + { + "epoch": 3.04, + "learning_rate": 4.69973143896734e-05, + "loss": 1.7079, + "step": 35160 + }, + { + "epoch": 3.04, + "learning_rate": 4.699644806376159e-05, + "loss": 1.7743, + "step": 35170 + }, + { + "epoch": 3.04, + "learning_rate": 4.699558173784978e-05, + "loss": 1.7577, + "step": 35180 + }, + { + "epoch": 3.05, + "learning_rate": 4.6994715411937974e-05, + "loss": 1.7164, + "step": 35190 + }, + { + "epoch": 3.05, + "learning_rate": 4.699384908602616e-05, + "loss": 1.7565, + "step": 35200 + }, + { + "epoch": 3.05, + "learning_rate": 4.6992982760114356e-05, + "loss": 1.7074, + "step": 35210 + }, + { + "epoch": 3.05, + "learning_rate": 4.699211643420255e-05, + "loss": 1.7883, + "step": 35220 + }, + { + "epoch": 3.05, + "learning_rate": 4.699125010829074e-05, + "loss": 1.7592, + "step": 35230 + }, + { + "epoch": 3.05, + "learning_rate": 4.699038378237893e-05, + "loss": 1.7633, + "step": 35240 + }, + { + "epoch": 3.05, + "learning_rate": 4.698951745646713e-05, + "loss": 1.7632, + "step": 35250 + }, + { + "epoch": 3.05, + "learning_rate": 4.6988651130555315e-05, + "loss": 1.7329, + "step": 35260 + }, + { + "epoch": 3.05, + "learning_rate": 4.698778480464351e-05, + "loss": 1.7874, + "step": 35270 + }, + { + "epoch": 3.05, + "learning_rate": 4.6986918478731704e-05, + "loss": 1.6742, + "step": 35280 + }, + { + "epoch": 3.05, + "learning_rate": 4.698605215281989e-05, + "loss": 1.6995, + "step": 35290 + }, + { + "epoch": 3.06, + "learning_rate": 4.6985185826908086e-05, + "loss": 1.6732, + "step": 35300 + }, + { + "epoch": 3.06, + "learning_rate": 4.6984319500996273e-05, + "loss": 1.7642, + "step": 35310 + }, + { + "epoch": 3.06, + "learning_rate": 4.698345317508447e-05, + "loss": 1.7278, + "step": 35320 + }, + { + "epoch": 3.06, + "learning_rate": 4.698258684917266e-05, + "loss": 1.6628, + "step": 35330 + }, + { + "epoch": 3.06, + "learning_rate": 4.698172052326085e-05, + "loss": 1.7079, + "step": 35340 + }, + { + "epoch": 3.06, + "learning_rate": 4.6980854197349045e-05, + "loss": 1.8157, + "step": 35350 + }, + { + "epoch": 3.06, + "learning_rate": 4.697998787143724e-05, + "loss": 1.7647, + "step": 35360 + }, + { + "epoch": 3.06, + "learning_rate": 4.697912154552543e-05, + "loss": 1.7124, + "step": 35370 + }, + { + "epoch": 3.06, + "learning_rate": 4.697825521961362e-05, + "loss": 1.7724, + "step": 35380 + }, + { + "epoch": 3.06, + "learning_rate": 4.6977388893701816e-05, + "loss": 1.7214, + "step": 35390 + }, + { + "epoch": 3.06, + "learning_rate": 4.697652256779e-05, + "loss": 1.6808, + "step": 35400 + }, + { + "epoch": 3.06, + "learning_rate": 4.69756562418782e-05, + "loss": 1.7407, + "step": 35410 + }, + { + "epoch": 3.07, + "learning_rate": 4.6974789915966385e-05, + "loss": 1.739, + "step": 35420 + }, + { + "epoch": 3.07, + "learning_rate": 4.697392359005458e-05, + "loss": 1.6924, + "step": 35430 + }, + { + "epoch": 3.07, + "learning_rate": 4.6973057264142774e-05, + "loss": 1.7251, + "step": 35440 + }, + { + "epoch": 3.07, + "learning_rate": 4.697219093823096e-05, + "loss": 1.7114, + "step": 35450 + }, + { + "epoch": 3.07, + "learning_rate": 4.6971324612319156e-05, + "loss": 1.7121, + "step": 35460 + }, + { + "epoch": 3.07, + "learning_rate": 4.697045828640735e-05, + "loss": 1.7277, + "step": 35470 + }, + { + "epoch": 3.07, + "learning_rate": 4.696959196049554e-05, + "loss": 1.7446, + "step": 35480 + }, + { + "epoch": 3.07, + "learning_rate": 4.696872563458373e-05, + "loss": 1.7642, + "step": 35490 + }, + { + "epoch": 3.07, + "learning_rate": 4.696785930867193e-05, + "loss": 1.7361, + "step": 35500 + }, + { + "epoch": 3.07, + "learning_rate": 4.6966992982760115e-05, + "loss": 1.7297, + "step": 35510 + }, + { + "epoch": 3.07, + "learning_rate": 4.696612665684831e-05, + "loss": 1.7346, + "step": 35520 + }, + { + "epoch": 3.08, + "learning_rate": 4.69652603309365e-05, + "loss": 1.7105, + "step": 35530 + }, + { + "epoch": 3.08, + "learning_rate": 4.696439400502469e-05, + "loss": 1.7224, + "step": 35540 + }, + { + "epoch": 3.08, + "learning_rate": 4.6963527679112886e-05, + "loss": 1.7073, + "step": 35550 + }, + { + "epoch": 3.08, + "learning_rate": 4.6962661353201074e-05, + "loss": 1.746, + "step": 35560 + }, + { + "epoch": 3.08, + "learning_rate": 4.696179502728927e-05, + "loss": 1.7053, + "step": 35570 + }, + { + "epoch": 3.08, + "learning_rate": 4.696092870137746e-05, + "loss": 1.7227, + "step": 35580 + }, + { + "epoch": 3.08, + "learning_rate": 4.696006237546565e-05, + "loss": 1.7169, + "step": 35590 + }, + { + "epoch": 3.08, + "learning_rate": 4.6959196049553845e-05, + "loss": 1.7319, + "step": 35600 + }, + { + "epoch": 3.08, + "learning_rate": 4.695832972364204e-05, + "loss": 1.64, + "step": 35610 + }, + { + "epoch": 3.08, + "learning_rate": 4.695746339773023e-05, + "loss": 1.7359, + "step": 35620 + }, + { + "epoch": 3.08, + "learning_rate": 4.695659707181842e-05, + "loss": 1.6731, + "step": 35630 + }, + { + "epoch": 3.08, + "learning_rate": 4.695573074590661e-05, + "loss": 1.7207, + "step": 35640 + }, + { + "epoch": 3.09, + "learning_rate": 4.6954864419994804e-05, + "loss": 1.6938, + "step": 35650 + }, + { + "epoch": 3.09, + "learning_rate": 4.6953998094083e-05, + "loss": 1.7356, + "step": 35660 + }, + { + "epoch": 3.09, + "learning_rate": 4.6953131768171186e-05, + "loss": 1.6683, + "step": 35670 + }, + { + "epoch": 3.09, + "learning_rate": 4.695226544225938e-05, + "loss": 1.6939, + "step": 35680 + }, + { + "epoch": 3.09, + "learning_rate": 4.6951399116347575e-05, + "loss": 1.7235, + "step": 35690 + }, + { + "epoch": 3.09, + "learning_rate": 4.695053279043576e-05, + "loss": 1.7159, + "step": 35700 + }, + { + "epoch": 3.09, + "learning_rate": 4.694966646452396e-05, + "loss": 1.6958, + "step": 35710 + }, + { + "epoch": 3.09, + "learning_rate": 4.694880013861215e-05, + "loss": 1.7138, + "step": 35720 + }, + { + "epoch": 3.09, + "learning_rate": 4.694793381270034e-05, + "loss": 1.7361, + "step": 35730 + }, + { + "epoch": 3.09, + "learning_rate": 4.694706748678853e-05, + "loss": 1.6621, + "step": 35740 + }, + { + "epoch": 3.09, + "learning_rate": 4.694620116087672e-05, + "loss": 1.7431, + "step": 35750 + }, + { + "epoch": 3.1, + "learning_rate": 4.6945334834964915e-05, + "loss": 1.6597, + "step": 35760 + }, + { + "epoch": 3.1, + "learning_rate": 4.694446850905311e-05, + "loss": 1.6848, + "step": 35770 + }, + { + "epoch": 3.1, + "learning_rate": 4.69436021831413e-05, + "loss": 1.7173, + "step": 35780 + }, + { + "epoch": 3.1, + "learning_rate": 4.694273585722949e-05, + "loss": 1.7088, + "step": 35790 + }, + { + "epoch": 3.1, + "learning_rate": 4.6941869531317686e-05, + "loss": 1.7153, + "step": 35800 + }, + { + "epoch": 3.1, + "learning_rate": 4.6941003205405874e-05, + "loss": 1.7071, + "step": 35810 + }, + { + "epoch": 3.1, + "learning_rate": 4.694013687949407e-05, + "loss": 1.7378, + "step": 35820 + }, + { + "epoch": 3.1, + "learning_rate": 4.693927055358226e-05, + "loss": 1.6801, + "step": 35830 + }, + { + "epoch": 3.1, + "learning_rate": 4.693840422767045e-05, + "loss": 1.7232, + "step": 35840 + }, + { + "epoch": 3.1, + "learning_rate": 4.6937537901758645e-05, + "loss": 1.7232, + "step": 35850 + }, + { + "epoch": 3.1, + "learning_rate": 4.693667157584683e-05, + "loss": 1.6924, + "step": 35860 + }, + { + "epoch": 3.1, + "learning_rate": 4.693580524993503e-05, + "loss": 1.6591, + "step": 35870 + }, + { + "epoch": 3.11, + "learning_rate": 4.693493892402322e-05, + "loss": 1.7201, + "step": 35880 + }, + { + "epoch": 3.11, + "learning_rate": 4.693407259811141e-05, + "loss": 1.7081, + "step": 35890 + }, + { + "epoch": 3.11, + "learning_rate": 4.6933206272199604e-05, + "loss": 1.7686, + "step": 35900 + }, + { + "epoch": 3.11, + "learning_rate": 4.69323399462878e-05, + "loss": 1.6938, + "step": 35910 + }, + { + "epoch": 3.11, + "learning_rate": 4.6931473620375986e-05, + "loss": 1.7546, + "step": 35920 + }, + { + "epoch": 3.11, + "learning_rate": 4.693060729446418e-05, + "loss": 1.7358, + "step": 35930 + }, + { + "epoch": 3.11, + "learning_rate": 4.692974096855237e-05, + "loss": 1.7204, + "step": 35940 + }, + { + "epoch": 3.11, + "learning_rate": 4.692887464264056e-05, + "loss": 1.7009, + "step": 35950 + }, + { + "epoch": 3.11, + "learning_rate": 4.692800831672876e-05, + "loss": 1.6957, + "step": 35960 + }, + { + "epoch": 3.11, + "learning_rate": 4.6927141990816945e-05, + "loss": 1.6937, + "step": 35970 + }, + { + "epoch": 3.11, + "learning_rate": 4.692627566490514e-05, + "loss": 1.7319, + "step": 35980 + }, + { + "epoch": 3.12, + "learning_rate": 4.6925409338993334e-05, + "loss": 1.6549, + "step": 35990 + }, + { + "epoch": 3.12, + "learning_rate": 4.692454301308152e-05, + "loss": 1.6717, + "step": 36000 + }, + { + "epoch": 3.12, + "learning_rate": 4.6923676687169716e-05, + "loss": 1.7159, + "step": 36010 + }, + { + "epoch": 3.12, + "learning_rate": 4.692281036125791e-05, + "loss": 1.7148, + "step": 36020 + }, + { + "epoch": 3.12, + "learning_rate": 4.69219440353461e-05, + "loss": 1.6684, + "step": 36030 + }, + { + "epoch": 3.12, + "learning_rate": 4.692107770943429e-05, + "loss": 1.6809, + "step": 36040 + }, + { + "epoch": 3.12, + "learning_rate": 4.692021138352248e-05, + "loss": 1.7209, + "step": 36050 + }, + { + "epoch": 3.12, + "learning_rate": 4.6919345057610674e-05, + "loss": 1.7686, + "step": 36060 + }, + { + "epoch": 3.12, + "learning_rate": 4.691847873169887e-05, + "loss": 1.7349, + "step": 36070 + }, + { + "epoch": 3.12, + "learning_rate": 4.6917612405787057e-05, + "loss": 1.6569, + "step": 36080 + }, + { + "epoch": 3.12, + "learning_rate": 4.691674607987525e-05, + "loss": 1.7703, + "step": 36090 + }, + { + "epoch": 3.12, + "learning_rate": 4.6915879753963446e-05, + "loss": 1.6743, + "step": 36100 + }, + { + "epoch": 3.13, + "learning_rate": 4.691501342805163e-05, + "loss": 1.7002, + "step": 36110 + }, + { + "epoch": 3.13, + "learning_rate": 4.691414710213983e-05, + "loss": 1.6871, + "step": 36120 + }, + { + "epoch": 3.13, + "learning_rate": 4.691328077622802e-05, + "loss": 1.7206, + "step": 36130 + }, + { + "epoch": 3.13, + "learning_rate": 4.691241445031621e-05, + "loss": 1.7248, + "step": 36140 + }, + { + "epoch": 3.13, + "learning_rate": 4.6911548124404404e-05, + "loss": 1.6194, + "step": 36150 + }, + { + "epoch": 3.13, + "learning_rate": 4.691068179849259e-05, + "loss": 1.7092, + "step": 36160 + }, + { + "epoch": 3.13, + "learning_rate": 4.6909815472580786e-05, + "loss": 1.7352, + "step": 36170 + }, + { + "epoch": 3.13, + "learning_rate": 4.690894914666898e-05, + "loss": 1.6726, + "step": 36180 + }, + { + "epoch": 3.13, + "learning_rate": 4.690808282075717e-05, + "loss": 1.6632, + "step": 36190 + }, + { + "epoch": 3.13, + "learning_rate": 4.690721649484536e-05, + "loss": 1.7175, + "step": 36200 + }, + { + "epoch": 3.13, + "learning_rate": 4.690635016893356e-05, + "loss": 1.6917, + "step": 36210 + }, + { + "epoch": 3.13, + "learning_rate": 4.6905483843021745e-05, + "loss": 1.6991, + "step": 36220 + }, + { + "epoch": 3.14, + "learning_rate": 4.690461751710994e-05, + "loss": 1.7429, + "step": 36230 + }, + { + "epoch": 3.14, + "learning_rate": 4.6903751191198134e-05, + "loss": 1.6769, + "step": 36240 + }, + { + "epoch": 3.14, + "learning_rate": 4.690288486528632e-05, + "loss": 1.707, + "step": 36250 + }, + { + "epoch": 3.14, + "learning_rate": 4.6902018539374516e-05, + "loss": 1.7438, + "step": 36260 + }, + { + "epoch": 3.14, + "learning_rate": 4.6901152213462704e-05, + "loss": 1.7133, + "step": 36270 + }, + { + "epoch": 3.14, + "learning_rate": 4.69002858875509e-05, + "loss": 1.7216, + "step": 36280 + }, + { + "epoch": 3.14, + "learning_rate": 4.689941956163909e-05, + "loss": 1.7191, + "step": 36290 + }, + { + "epoch": 3.14, + "learning_rate": 4.689855323572728e-05, + "loss": 1.679, + "step": 36300 + }, + { + "epoch": 3.14, + "learning_rate": 4.6897686909815475e-05, + "loss": 1.6975, + "step": 36310 + }, + { + "epoch": 3.14, + "learning_rate": 4.689682058390367e-05, + "loss": 1.6636, + "step": 36320 + }, + { + "epoch": 3.14, + "learning_rate": 4.689595425799186e-05, + "loss": 1.6723, + "step": 36330 + }, + { + "epoch": 3.15, + "learning_rate": 4.689508793208005e-05, + "loss": 1.764, + "step": 36340 + }, + { + "epoch": 3.15, + "learning_rate": 4.6894221606168246e-05, + "loss": 1.7798, + "step": 36350 + }, + { + "epoch": 3.15, + "learning_rate": 4.6893355280256434e-05, + "loss": 1.7079, + "step": 36360 + }, + { + "epoch": 3.15, + "learning_rate": 4.689248895434463e-05, + "loss": 1.7615, + "step": 36370 + }, + { + "epoch": 3.15, + "learning_rate": 4.6891622628432816e-05, + "loss": 1.7303, + "step": 36380 + }, + { + "epoch": 3.15, + "learning_rate": 4.689075630252101e-05, + "loss": 1.6911, + "step": 36390 + }, + { + "epoch": 3.15, + "learning_rate": 4.6889889976609205e-05, + "loss": 1.7034, + "step": 36400 + }, + { + "epoch": 3.15, + "learning_rate": 4.688902365069739e-05, + "loss": 1.6911, + "step": 36410 + }, + { + "epoch": 3.15, + "learning_rate": 4.688815732478559e-05, + "loss": 1.7377, + "step": 36420 + }, + { + "epoch": 3.15, + "learning_rate": 4.688729099887378e-05, + "loss": 1.6974, + "step": 36430 + }, + { + "epoch": 3.15, + "learning_rate": 4.688642467296197e-05, + "loss": 1.7391, + "step": 36440 + }, + { + "epoch": 3.15, + "learning_rate": 4.688555834705016e-05, + "loss": 1.6753, + "step": 36450 + }, + { + "epoch": 3.16, + "learning_rate": 4.688469202113836e-05, + "loss": 1.6959, + "step": 36460 + }, + { + "epoch": 3.16, + "learning_rate": 4.6883825695226545e-05, + "loss": 1.7234, + "step": 36470 + }, + { + "epoch": 3.16, + "learning_rate": 4.688295936931474e-05, + "loss": 1.7122, + "step": 36480 + }, + { + "epoch": 3.16, + "learning_rate": 4.688209304340293e-05, + "loss": 1.7273, + "step": 36490 + }, + { + "epoch": 3.16, + "learning_rate": 4.688122671749112e-05, + "loss": 1.7352, + "step": 36500 + }, + { + "epoch": 3.16, + "learning_rate": 4.6880360391579316e-05, + "loss": 1.6846, + "step": 36510 + }, + { + "epoch": 3.16, + "learning_rate": 4.6879494065667504e-05, + "loss": 1.5989, + "step": 36520 + }, + { + "epoch": 3.16, + "learning_rate": 4.68786277397557e-05, + "loss": 1.7291, + "step": 36530 + }, + { + "epoch": 3.16, + "learning_rate": 4.687776141384389e-05, + "loss": 1.7084, + "step": 36540 + }, + { + "epoch": 3.16, + "learning_rate": 4.687689508793208e-05, + "loss": 1.6931, + "step": 36550 + }, + { + "epoch": 3.16, + "learning_rate": 4.6876028762020275e-05, + "loss": 1.7041, + "step": 36560 + }, + { + "epoch": 3.17, + "learning_rate": 4.687516243610847e-05, + "loss": 1.768, + "step": 36570 + }, + { + "epoch": 3.17, + "learning_rate": 4.687429611019666e-05, + "loss": 1.7172, + "step": 36580 + }, + { + "epoch": 3.17, + "learning_rate": 4.687342978428485e-05, + "loss": 1.6937, + "step": 36590 + }, + { + "epoch": 3.17, + "learning_rate": 4.687256345837304e-05, + "loss": 1.6725, + "step": 36600 + }, + { + "epoch": 3.17, + "learning_rate": 4.6871697132461234e-05, + "loss": 1.7553, + "step": 36610 + }, + { + "epoch": 3.17, + "learning_rate": 4.687083080654943e-05, + "loss": 1.7433, + "step": 36620 + }, + { + "epoch": 3.17, + "learning_rate": 4.6869964480637616e-05, + "loss": 1.7825, + "step": 36630 + }, + { + "epoch": 3.17, + "learning_rate": 4.686909815472581e-05, + "loss": 1.6774, + "step": 36640 + }, + { + "epoch": 3.17, + "learning_rate": 4.6868231828814005e-05, + "loss": 1.7393, + "step": 36650 + }, + { + "epoch": 3.17, + "learning_rate": 4.686736550290219e-05, + "loss": 1.7189, + "step": 36660 + }, + { + "epoch": 3.17, + "learning_rate": 4.686649917699039e-05, + "loss": 1.778, + "step": 36670 + }, + { + "epoch": 3.17, + "learning_rate": 4.6865632851078575e-05, + "loss": 1.7112, + "step": 36680 + }, + { + "epoch": 3.18, + "learning_rate": 4.686476652516677e-05, + "loss": 1.7583, + "step": 36690 + }, + { + "epoch": 3.18, + "learning_rate": 4.6863900199254964e-05, + "loss": 1.6875, + "step": 36700 + }, + { + "epoch": 3.18, + "learning_rate": 4.686303387334315e-05, + "loss": 1.6864, + "step": 36710 + }, + { + "epoch": 3.18, + "learning_rate": 4.6862167547431346e-05, + "loss": 1.6874, + "step": 36720 + }, + { + "epoch": 3.18, + "learning_rate": 4.686130122151954e-05, + "loss": 1.7, + "step": 36730 + }, + { + "epoch": 3.18, + "learning_rate": 4.686043489560773e-05, + "loss": 1.7327, + "step": 36740 + }, + { + "epoch": 3.18, + "learning_rate": 4.685956856969592e-05, + "loss": 1.6314, + "step": 36750 + }, + { + "epoch": 3.18, + "learning_rate": 4.685870224378412e-05, + "loss": 1.7124, + "step": 36760 + }, + { + "epoch": 3.18, + "learning_rate": 4.6857835917872304e-05, + "loss": 1.7555, + "step": 36770 + }, + { + "epoch": 3.18, + "learning_rate": 4.68569695919605e-05, + "loss": 1.7018, + "step": 36780 + }, + { + "epoch": 3.18, + "learning_rate": 4.6856103266048687e-05, + "loss": 1.7042, + "step": 36790 + }, + { + "epoch": 3.19, + "learning_rate": 4.685523694013688e-05, + "loss": 1.6426, + "step": 36800 + }, + { + "epoch": 3.19, + "learning_rate": 4.6854370614225075e-05, + "loss": 1.6939, + "step": 36810 + }, + { + "epoch": 3.19, + "learning_rate": 4.685350428831326e-05, + "loss": 1.6973, + "step": 36820 + }, + { + "epoch": 3.19, + "learning_rate": 4.685263796240146e-05, + "loss": 1.6716, + "step": 36830 + }, + { + "epoch": 3.19, + "learning_rate": 4.685177163648965e-05, + "loss": 1.7538, + "step": 36840 + }, + { + "epoch": 3.19, + "learning_rate": 4.685090531057784e-05, + "loss": 1.6517, + "step": 36850 + }, + { + "epoch": 3.19, + "learning_rate": 4.6850038984666034e-05, + "loss": 1.7689, + "step": 36860 + }, + { + "epoch": 3.19, + "learning_rate": 4.684917265875423e-05, + "loss": 1.6976, + "step": 36870 + }, + { + "epoch": 3.19, + "learning_rate": 4.6848306332842416e-05, + "loss": 1.7225, + "step": 36880 + }, + { + "epoch": 3.19, + "learning_rate": 4.684744000693061e-05, + "loss": 1.7953, + "step": 36890 + }, + { + "epoch": 3.19, + "learning_rate": 4.68465736810188e-05, + "loss": 1.7032, + "step": 36900 + }, + { + "epoch": 3.19, + "learning_rate": 4.684570735510699e-05, + "loss": 1.7467, + "step": 36910 + }, + { + "epoch": 3.2, + "learning_rate": 4.684484102919519e-05, + "loss": 1.6386, + "step": 36920 + }, + { + "epoch": 3.2, + "learning_rate": 4.6843974703283375e-05, + "loss": 1.6535, + "step": 36930 + }, + { + "epoch": 3.2, + "learning_rate": 4.684310837737157e-05, + "loss": 1.6826, + "step": 36940 + }, + { + "epoch": 3.2, + "learning_rate": 4.6842242051459764e-05, + "loss": 1.7423, + "step": 36950 + }, + { + "epoch": 3.2, + "learning_rate": 4.684137572554795e-05, + "loss": 1.72, + "step": 36960 + }, + { + "epoch": 3.2, + "learning_rate": 4.6840509399636146e-05, + "loss": 1.6801, + "step": 36970 + }, + { + "epoch": 3.2, + "learning_rate": 4.683964307372434e-05, + "loss": 1.7252, + "step": 36980 + }, + { + "epoch": 3.2, + "learning_rate": 4.683877674781253e-05, + "loss": 1.7012, + "step": 36990 + }, + { + "epoch": 3.2, + "learning_rate": 4.683791042190072e-05, + "loss": 1.6718, + "step": 37000 + }, + { + "epoch": 3.2, + "learning_rate": 4.683704409598891e-05, + "loss": 1.7061, + "step": 37010 + }, + { + "epoch": 3.2, + "learning_rate": 4.6836177770077105e-05, + "loss": 1.7072, + "step": 37020 + }, + { + "epoch": 3.21, + "learning_rate": 4.68353114441653e-05, + "loss": 1.7083, + "step": 37030 + }, + { + "epoch": 3.21, + "learning_rate": 4.683444511825349e-05, + "loss": 1.7028, + "step": 37040 + }, + { + "epoch": 3.21, + "learning_rate": 4.683357879234168e-05, + "loss": 1.7216, + "step": 37050 + }, + { + "epoch": 3.21, + "learning_rate": 4.6832712466429876e-05, + "loss": 1.7377, + "step": 37060 + }, + { + "epoch": 3.21, + "learning_rate": 4.6831846140518063e-05, + "loss": 1.7487, + "step": 37070 + }, + { + "epoch": 3.21, + "learning_rate": 4.683097981460626e-05, + "loss": 1.7197, + "step": 37080 + }, + { + "epoch": 3.21, + "learning_rate": 4.683011348869445e-05, + "loss": 1.7971, + "step": 37090 + }, + { + "epoch": 3.21, + "learning_rate": 4.682924716278264e-05, + "loss": 1.699, + "step": 37100 + }, + { + "epoch": 3.21, + "learning_rate": 4.6828380836870835e-05, + "loss": 1.6946, + "step": 37110 + }, + { + "epoch": 3.21, + "learning_rate": 4.682751451095902e-05, + "loss": 1.6705, + "step": 37120 + }, + { + "epoch": 3.21, + "learning_rate": 4.682664818504722e-05, + "loss": 1.6808, + "step": 37130 + }, + { + "epoch": 3.21, + "learning_rate": 4.682578185913541e-05, + "loss": 1.7352, + "step": 37140 + }, + { + "epoch": 3.22, + "learning_rate": 4.68249155332236e-05, + "loss": 1.7623, + "step": 37150 + }, + { + "epoch": 3.22, + "learning_rate": 4.682404920731179e-05, + "loss": 1.6887, + "step": 37160 + }, + { + "epoch": 3.22, + "learning_rate": 4.682318288139999e-05, + "loss": 1.7255, + "step": 37170 + }, + { + "epoch": 3.22, + "learning_rate": 4.6822316555488175e-05, + "loss": 1.6924, + "step": 37180 + }, + { + "epoch": 3.22, + "learning_rate": 4.682145022957637e-05, + "loss": 1.7439, + "step": 37190 + }, + { + "epoch": 3.22, + "learning_rate": 4.6820583903664564e-05, + "loss": 1.6825, + "step": 37200 + }, + { + "epoch": 3.22, + "learning_rate": 4.681971757775275e-05, + "loss": 1.761, + "step": 37210 + }, + { + "epoch": 3.22, + "learning_rate": 4.6818851251840946e-05, + "loss": 1.7248, + "step": 37220 + }, + { + "epoch": 3.22, + "learning_rate": 4.6817984925929134e-05, + "loss": 1.6683, + "step": 37230 + }, + { + "epoch": 3.22, + "learning_rate": 4.681711860001733e-05, + "loss": 1.7331, + "step": 37240 + }, + { + "epoch": 3.22, + "learning_rate": 4.681625227410552e-05, + "loss": 1.7505, + "step": 37250 + }, + { + "epoch": 3.22, + "learning_rate": 4.681538594819371e-05, + "loss": 1.7135, + "step": 37260 + }, + { + "epoch": 3.23, + "learning_rate": 4.6814519622281905e-05, + "loss": 1.7095, + "step": 37270 + }, + { + "epoch": 3.23, + "learning_rate": 4.68136532963701e-05, + "loss": 1.6968, + "step": 37280 + }, + { + "epoch": 3.23, + "learning_rate": 4.681278697045829e-05, + "loss": 1.6998, + "step": 37290 + }, + { + "epoch": 3.23, + "learning_rate": 4.681192064454648e-05, + "loss": 1.6821, + "step": 37300 + }, + { + "epoch": 3.23, + "learning_rate": 4.6811054318634676e-05, + "loss": 1.7077, + "step": 37310 + }, + { + "epoch": 3.23, + "learning_rate": 4.6810187992722864e-05, + "loss": 1.7854, + "step": 37320 + }, + { + "epoch": 3.23, + "learning_rate": 4.680932166681106e-05, + "loss": 1.7087, + "step": 37330 + }, + { + "epoch": 3.23, + "learning_rate": 4.6808455340899246e-05, + "loss": 1.7426, + "step": 37340 + }, + { + "epoch": 3.23, + "learning_rate": 4.680758901498744e-05, + "loss": 1.6894, + "step": 37350 + }, + { + "epoch": 3.23, + "learning_rate": 4.6806722689075635e-05, + "loss": 1.7434, + "step": 37360 + }, + { + "epoch": 3.23, + "learning_rate": 4.680585636316382e-05, + "loss": 1.7517, + "step": 37370 + }, + { + "epoch": 3.24, + "learning_rate": 4.680499003725202e-05, + "loss": 1.6705, + "step": 37380 + }, + { + "epoch": 3.24, + "learning_rate": 4.680412371134021e-05, + "loss": 1.719, + "step": 37390 + }, + { + "epoch": 3.24, + "learning_rate": 4.68032573854284e-05, + "loss": 1.7219, + "step": 37400 + }, + { + "epoch": 3.24, + "learning_rate": 4.6802391059516594e-05, + "loss": 1.6941, + "step": 37410 + }, + { + "epoch": 3.24, + "learning_rate": 4.680152473360478e-05, + "loss": 1.6688, + "step": 37420 + }, + { + "epoch": 3.24, + "learning_rate": 4.6800658407692976e-05, + "loss": 1.7506, + "step": 37430 + }, + { + "epoch": 3.24, + "learning_rate": 4.679979208178117e-05, + "loss": 1.7141, + "step": 37440 + }, + { + "epoch": 3.24, + "learning_rate": 4.679892575586936e-05, + "loss": 1.6987, + "step": 37450 + }, + { + "epoch": 3.24, + "learning_rate": 4.679805942995755e-05, + "loss": 1.6999, + "step": 37460 + }, + { + "epoch": 3.24, + "learning_rate": 4.679719310404575e-05, + "loss": 1.7233, + "step": 37470 + }, + { + "epoch": 3.24, + "learning_rate": 4.6796326778133934e-05, + "loss": 1.685, + "step": 37480 + }, + { + "epoch": 3.24, + "learning_rate": 4.679546045222213e-05, + "loss": 1.74, + "step": 37490 + }, + { + "epoch": 3.25, + "learning_rate": 4.679459412631032e-05, + "loss": 1.7133, + "step": 37500 + }, + { + "epoch": 3.25, + "learning_rate": 4.679372780039851e-05, + "loss": 1.6815, + "step": 37510 + }, + { + "epoch": 3.25, + "learning_rate": 4.6792861474486705e-05, + "loss": 1.7269, + "step": 37520 + }, + { + "epoch": 3.25, + "learning_rate": 4.679199514857489e-05, + "loss": 1.7003, + "step": 37530 + }, + { + "epoch": 3.25, + "learning_rate": 4.679112882266309e-05, + "loss": 1.7044, + "step": 37540 + }, + { + "epoch": 3.25, + "learning_rate": 4.679026249675128e-05, + "loss": 1.6939, + "step": 37550 + }, + { + "epoch": 3.25, + "learning_rate": 4.678939617083947e-05, + "loss": 1.6814, + "step": 37560 + }, + { + "epoch": 3.25, + "learning_rate": 4.6788529844927664e-05, + "loss": 1.6397, + "step": 37570 + }, + { + "epoch": 3.25, + "learning_rate": 4.678766351901586e-05, + "loss": 1.7293, + "step": 37580 + }, + { + "epoch": 3.25, + "learning_rate": 4.6786797193104046e-05, + "loss": 1.7362, + "step": 37590 + }, + { + "epoch": 3.25, + "learning_rate": 4.678593086719224e-05, + "loss": 1.6755, + "step": 37600 + }, + { + "epoch": 3.26, + "learning_rate": 4.6785064541280435e-05, + "loss": 1.7407, + "step": 37610 + }, + { + "epoch": 3.26, + "learning_rate": 4.678419821536862e-05, + "loss": 1.6845, + "step": 37620 + }, + { + "epoch": 3.26, + "learning_rate": 4.678333188945682e-05, + "loss": 1.7451, + "step": 37630 + }, + { + "epoch": 3.26, + "learning_rate": 4.6782465563545005e-05, + "loss": 1.7031, + "step": 37640 + }, + { + "epoch": 3.26, + "learning_rate": 4.67815992376332e-05, + "loss": 1.6949, + "step": 37650 + }, + { + "epoch": 3.26, + "learning_rate": 4.6780732911721394e-05, + "loss": 1.7496, + "step": 37660 + }, + { + "epoch": 3.26, + "learning_rate": 4.677986658580958e-05, + "loss": 1.7754, + "step": 37670 + }, + { + "epoch": 3.26, + "learning_rate": 4.6779000259897776e-05, + "loss": 1.6981, + "step": 37680 + }, + { + "epoch": 3.26, + "learning_rate": 4.677813393398597e-05, + "loss": 1.6844, + "step": 37690 + }, + { + "epoch": 3.26, + "learning_rate": 4.677726760807416e-05, + "loss": 1.6983, + "step": 37700 + }, + { + "epoch": 3.26, + "learning_rate": 4.677640128216235e-05, + "loss": 1.7141, + "step": 37710 + }, + { + "epoch": 3.26, + "learning_rate": 4.677553495625055e-05, + "loss": 1.7063, + "step": 37720 + }, + { + "epoch": 3.27, + "learning_rate": 4.6774668630338735e-05, + "loss": 1.7088, + "step": 37730 + }, + { + "epoch": 3.27, + "learning_rate": 4.677380230442693e-05, + "loss": 1.6424, + "step": 37740 + }, + { + "epoch": 3.27, + "learning_rate": 4.677293597851512e-05, + "loss": 1.7124, + "step": 37750 + }, + { + "epoch": 3.27, + "learning_rate": 4.677206965260331e-05, + "loss": 1.7062, + "step": 37760 + }, + { + "epoch": 3.27, + "learning_rate": 4.6771203326691506e-05, + "loss": 1.7839, + "step": 37770 + }, + { + "epoch": 3.27, + "learning_rate": 4.6770337000779693e-05, + "loss": 1.6683, + "step": 37780 + }, + { + "epoch": 3.27, + "learning_rate": 4.676947067486789e-05, + "loss": 1.7443, + "step": 37790 + }, + { + "epoch": 3.27, + "learning_rate": 4.676860434895608e-05, + "loss": 1.7276, + "step": 37800 + }, + { + "epoch": 3.27, + "learning_rate": 4.676773802304427e-05, + "loss": 1.6538, + "step": 37810 + }, + { + "epoch": 3.27, + "learning_rate": 4.6766871697132465e-05, + "loss": 1.6574, + "step": 37820 + }, + { + "epoch": 3.27, + "learning_rate": 4.676600537122066e-05, + "loss": 1.7479, + "step": 37830 + }, + { + "epoch": 3.28, + "learning_rate": 4.676513904530885e-05, + "loss": 1.6855, + "step": 37840 + }, + { + "epoch": 3.28, + "learning_rate": 4.676427271939704e-05, + "loss": 1.6814, + "step": 37850 + }, + { + "epoch": 3.28, + "learning_rate": 4.676340639348523e-05, + "loss": 1.6781, + "step": 37860 + }, + { + "epoch": 3.28, + "learning_rate": 4.676254006757342e-05, + "loss": 1.6991, + "step": 37870 + }, + { + "epoch": 3.28, + "learning_rate": 4.676167374166162e-05, + "loss": 1.6587, + "step": 37880 + }, + { + "epoch": 3.28, + "learning_rate": 4.6760807415749805e-05, + "loss": 1.7159, + "step": 37890 + }, + { + "epoch": 3.28, + "learning_rate": 4.6759941089838e-05, + "loss": 1.7272, + "step": 37900 + }, + { + "epoch": 3.28, + "learning_rate": 4.6759074763926194e-05, + "loss": 1.7167, + "step": 37910 + }, + { + "epoch": 3.28, + "learning_rate": 4.675820843801438e-05, + "loss": 1.6979, + "step": 37920 + }, + { + "epoch": 3.28, + "learning_rate": 4.6757342112102576e-05, + "loss": 1.7153, + "step": 37930 + }, + { + "epoch": 3.28, + "learning_rate": 4.675647578619077e-05, + "loss": 1.6509, + "step": 37940 + }, + { + "epoch": 3.28, + "learning_rate": 4.675560946027896e-05, + "loss": 1.7047, + "step": 37950 + }, + { + "epoch": 3.29, + "learning_rate": 4.675474313436715e-05, + "loss": 1.725, + "step": 37960 + }, + { + "epoch": 3.29, + "learning_rate": 4.675387680845534e-05, + "loss": 1.7236, + "step": 37970 + }, + { + "epoch": 3.29, + "learning_rate": 4.6753010482543535e-05, + "loss": 1.7455, + "step": 37980 + }, + { + "epoch": 3.29, + "learning_rate": 4.675214415663173e-05, + "loss": 1.6908, + "step": 37990 + }, + { + "epoch": 3.29, + "learning_rate": 4.675127783071992e-05, + "loss": 1.6642, + "step": 38000 + }, + { + "epoch": 3.29, + "learning_rate": 4.675041150480811e-05, + "loss": 1.7455, + "step": 38010 + }, + { + "epoch": 3.29, + "learning_rate": 4.6749545178896306e-05, + "loss": 1.6817, + "step": 38020 + }, + { + "epoch": 3.29, + "learning_rate": 4.6748678852984494e-05, + "loss": 1.7259, + "step": 38030 + }, + { + "epoch": 3.29, + "learning_rate": 4.674781252707269e-05, + "loss": 1.7133, + "step": 38040 + }, + { + "epoch": 3.29, + "learning_rate": 4.674694620116088e-05, + "loss": 1.7643, + "step": 38050 + }, + { + "epoch": 3.29, + "learning_rate": 4.674607987524907e-05, + "loss": 1.7311, + "step": 38060 + }, + { + "epoch": 3.3, + "learning_rate": 4.6745213549337265e-05, + "loss": 1.6968, + "step": 38070 + }, + { + "epoch": 3.3, + "learning_rate": 4.674434722342545e-05, + "loss": 1.6473, + "step": 38080 + }, + { + "epoch": 3.3, + "learning_rate": 4.674348089751365e-05, + "loss": 1.6848, + "step": 38090 + }, + { + "epoch": 3.3, + "learning_rate": 4.674261457160184e-05, + "loss": 1.6912, + "step": 38100 + }, + { + "epoch": 3.3, + "learning_rate": 4.674174824569003e-05, + "loss": 1.7342, + "step": 38110 + }, + { + "epoch": 3.3, + "learning_rate": 4.6740881919778224e-05, + "loss": 1.7007, + "step": 38120 + }, + { + "epoch": 3.3, + "learning_rate": 4.674001559386642e-05, + "loss": 1.75, + "step": 38130 + }, + { + "epoch": 3.3, + "learning_rate": 4.6739149267954606e-05, + "loss": 1.7452, + "step": 38140 + }, + { + "epoch": 3.3, + "learning_rate": 4.67382829420428e-05, + "loss": 1.6899, + "step": 38150 + }, + { + "epoch": 3.3, + "learning_rate": 4.6737416616130995e-05, + "loss": 1.724, + "step": 38160 + }, + { + "epoch": 3.3, + "learning_rate": 4.673655029021918e-05, + "loss": 1.754, + "step": 38170 + }, + { + "epoch": 3.3, + "learning_rate": 4.673568396430738e-05, + "loss": 1.6868, + "step": 38180 + }, + { + "epoch": 3.31, + "learning_rate": 4.6734817638395564e-05, + "loss": 1.7598, + "step": 38190 + }, + { + "epoch": 3.31, + "learning_rate": 4.673395131248376e-05, + "loss": 1.7123, + "step": 38200 + }, + { + "epoch": 3.31, + "learning_rate": 4.673308498657195e-05, + "loss": 1.7152, + "step": 38210 + }, + { + "epoch": 3.31, + "learning_rate": 4.673221866066014e-05, + "loss": 1.7085, + "step": 38220 + }, + { + "epoch": 3.31, + "learning_rate": 4.6731352334748335e-05, + "loss": 1.6746, + "step": 38230 + }, + { + "epoch": 3.31, + "learning_rate": 4.673048600883653e-05, + "loss": 1.7431, + "step": 38240 + }, + { + "epoch": 3.31, + "learning_rate": 4.672961968292472e-05, + "loss": 1.704, + "step": 38250 + }, + { + "epoch": 3.31, + "learning_rate": 4.672875335701291e-05, + "loss": 1.6944, + "step": 38260 + }, + { + "epoch": 3.31, + "learning_rate": 4.67278870311011e-05, + "loss": 1.6736, + "step": 38270 + }, + { + "epoch": 3.31, + "learning_rate": 4.6727020705189294e-05, + "loss": 1.7126, + "step": 38280 + }, + { + "epoch": 3.31, + "learning_rate": 4.672615437927749e-05, + "loss": 1.6612, + "step": 38290 + }, + { + "epoch": 3.32, + "learning_rate": 4.6725288053365676e-05, + "loss": 1.6222, + "step": 38300 + }, + { + "epoch": 3.32, + "learning_rate": 4.672442172745387e-05, + "loss": 1.6671, + "step": 38310 + }, + { + "epoch": 3.32, + "learning_rate": 4.6723555401542065e-05, + "loss": 1.7025, + "step": 38320 + }, + { + "epoch": 3.32, + "learning_rate": 4.672268907563025e-05, + "loss": 1.6739, + "step": 38330 + }, + { + "epoch": 3.32, + "learning_rate": 4.672182274971845e-05, + "loss": 1.6784, + "step": 38340 + }, + { + "epoch": 3.32, + "learning_rate": 4.672095642380664e-05, + "loss": 1.7409, + "step": 38350 + }, + { + "epoch": 3.32, + "learning_rate": 4.672009009789483e-05, + "loss": 1.7186, + "step": 38360 + }, + { + "epoch": 3.32, + "learning_rate": 4.6719223771983024e-05, + "loss": 1.6957, + "step": 38370 + }, + { + "epoch": 3.32, + "learning_rate": 4.671835744607121e-05, + "loss": 1.6861, + "step": 38380 + }, + { + "epoch": 3.32, + "learning_rate": 4.6717491120159406e-05, + "loss": 1.6705, + "step": 38390 + }, + { + "epoch": 3.32, + "learning_rate": 4.67166247942476e-05, + "loss": 1.7743, + "step": 38400 + }, + { + "epoch": 3.32, + "learning_rate": 4.671575846833579e-05, + "loss": 1.7647, + "step": 38410 + }, + { + "epoch": 3.33, + "learning_rate": 4.671489214242398e-05, + "loss": 1.7401, + "step": 38420 + }, + { + "epoch": 3.33, + "learning_rate": 4.671402581651218e-05, + "loss": 1.6574, + "step": 38430 + }, + { + "epoch": 3.33, + "learning_rate": 4.6713159490600365e-05, + "loss": 1.7103, + "step": 38440 + }, + { + "epoch": 3.33, + "learning_rate": 4.671229316468856e-05, + "loss": 1.7838, + "step": 38450 + }, + { + "epoch": 3.33, + "learning_rate": 4.6711426838776754e-05, + "loss": 1.6772, + "step": 38460 + }, + { + "epoch": 3.33, + "learning_rate": 4.671056051286494e-05, + "loss": 1.7178, + "step": 38470 + }, + { + "epoch": 3.33, + "learning_rate": 4.6709694186953136e-05, + "loss": 1.6424, + "step": 38480 + }, + { + "epoch": 3.33, + "learning_rate": 4.6708827861041323e-05, + "loss": 1.7073, + "step": 38490 + }, + { + "epoch": 3.33, + "learning_rate": 4.670796153512952e-05, + "loss": 1.7181, + "step": 38500 + }, + { + "epoch": 3.33, + "learning_rate": 4.670709520921771e-05, + "loss": 1.6786, + "step": 38510 + }, + { + "epoch": 3.33, + "learning_rate": 4.67062288833059e-05, + "loss": 1.7454, + "step": 38520 + }, + { + "epoch": 3.33, + "learning_rate": 4.6705362557394094e-05, + "loss": 1.6996, + "step": 38530 + }, + { + "epoch": 3.34, + "learning_rate": 4.670449623148229e-05, + "loss": 1.7268, + "step": 38540 + }, + { + "epoch": 3.34, + "learning_rate": 4.6703629905570477e-05, + "loss": 1.6172, + "step": 38550 + }, + { + "epoch": 3.34, + "learning_rate": 4.670276357965867e-05, + "loss": 1.7116, + "step": 38560 + }, + { + "epoch": 3.34, + "learning_rate": 4.6701897253746866e-05, + "loss": 1.6651, + "step": 38570 + }, + { + "epoch": 3.34, + "learning_rate": 4.670103092783505e-05, + "loss": 1.6686, + "step": 38580 + }, + { + "epoch": 3.34, + "learning_rate": 4.670016460192325e-05, + "loss": 1.6947, + "step": 38590 + }, + { + "epoch": 3.34, + "learning_rate": 4.6699298276011435e-05, + "loss": 1.6422, + "step": 38600 + }, + { + "epoch": 3.34, + "learning_rate": 4.669843195009963e-05, + "loss": 1.643, + "step": 38610 + }, + { + "epoch": 3.34, + "learning_rate": 4.6697565624187824e-05, + "loss": 1.7452, + "step": 38620 + }, + { + "epoch": 3.34, + "learning_rate": 4.669669929827601e-05, + "loss": 1.6989, + "step": 38630 + }, + { + "epoch": 3.34, + "learning_rate": 4.6695832972364206e-05, + "loss": 1.7125, + "step": 38640 + }, + { + "epoch": 3.35, + "learning_rate": 4.66949666464524e-05, + "loss": 1.6701, + "step": 38650 + }, + { + "epoch": 3.35, + "learning_rate": 4.669410032054059e-05, + "loss": 1.7701, + "step": 38660 + }, + { + "epoch": 3.35, + "learning_rate": 4.669323399462878e-05, + "loss": 1.7231, + "step": 38670 + }, + { + "epoch": 3.35, + "learning_rate": 4.669236766871698e-05, + "loss": 1.7142, + "step": 38680 + }, + { + "epoch": 3.35, + "learning_rate": 4.6691501342805165e-05, + "loss": 1.7141, + "step": 38690 + }, + { + "epoch": 3.35, + "learning_rate": 4.669063501689336e-05, + "loss": 1.673, + "step": 38700 + }, + { + "epoch": 3.35, + "learning_rate": 4.668976869098155e-05, + "loss": 1.7104, + "step": 38710 + }, + { + "epoch": 3.35, + "learning_rate": 4.668890236506974e-05, + "loss": 1.7028, + "step": 38720 + }, + { + "epoch": 3.35, + "learning_rate": 4.6688036039157936e-05, + "loss": 1.708, + "step": 38730 + }, + { + "epoch": 3.35, + "learning_rate": 4.6687169713246124e-05, + "loss": 1.7545, + "step": 38740 + }, + { + "epoch": 3.35, + "learning_rate": 4.668630338733432e-05, + "loss": 1.6793, + "step": 38750 + }, + { + "epoch": 3.35, + "learning_rate": 4.668543706142251e-05, + "loss": 1.7016, + "step": 38760 + }, + { + "epoch": 3.36, + "learning_rate": 4.66845707355107e-05, + "loss": 1.6904, + "step": 38770 + }, + { + "epoch": 3.36, + "learning_rate": 4.6683704409598895e-05, + "loss": 1.7201, + "step": 38780 + }, + { + "epoch": 3.36, + "learning_rate": 4.668283808368709e-05, + "loss": 1.6612, + "step": 38790 + }, + { + "epoch": 3.36, + "learning_rate": 4.668197175777528e-05, + "loss": 1.7429, + "step": 38800 + }, + { + "epoch": 3.36, + "learning_rate": 4.668110543186347e-05, + "loss": 1.7198, + "step": 38810 + }, + { + "epoch": 3.36, + "learning_rate": 4.668023910595166e-05, + "loss": 1.664, + "step": 38820 + }, + { + "epoch": 3.36, + "learning_rate": 4.6679372780039854e-05, + "loss": 1.6581, + "step": 38830 + }, + { + "epoch": 3.36, + "learning_rate": 4.667850645412805e-05, + "loss": 1.6704, + "step": 38840 + }, + { + "epoch": 3.36, + "learning_rate": 4.6677640128216236e-05, + "loss": 1.7358, + "step": 38850 + }, + { + "epoch": 3.36, + "learning_rate": 4.667677380230443e-05, + "loss": 1.7535, + "step": 38860 + }, + { + "epoch": 3.36, + "learning_rate": 4.6675907476392625e-05, + "loss": 1.6881, + "step": 38870 + }, + { + "epoch": 3.37, + "learning_rate": 4.667504115048081e-05, + "loss": 1.6256, + "step": 38880 + }, + { + "epoch": 3.37, + "learning_rate": 4.667417482456901e-05, + "loss": 1.7312, + "step": 38890 + }, + { + "epoch": 3.37, + "learning_rate": 4.66733084986572e-05, + "loss": 1.6519, + "step": 38900 + }, + { + "epoch": 3.37, + "learning_rate": 4.667244217274539e-05, + "loss": 1.7165, + "step": 38910 + }, + { + "epoch": 3.37, + "learning_rate": 4.667157584683358e-05, + "loss": 1.6585, + "step": 38920 + }, + { + "epoch": 3.37, + "learning_rate": 4.667070952092177e-05, + "loss": 1.7398, + "step": 38930 + }, + { + "epoch": 3.37, + "learning_rate": 4.6669843195009965e-05, + "loss": 1.7213, + "step": 38940 + }, + { + "epoch": 3.37, + "learning_rate": 4.666897686909816e-05, + "loss": 1.6634, + "step": 38950 + }, + { + "epoch": 3.37, + "learning_rate": 4.666811054318635e-05, + "loss": 1.6785, + "step": 38960 + }, + { + "epoch": 3.37, + "learning_rate": 4.666724421727454e-05, + "loss": 1.662, + "step": 38970 + }, + { + "epoch": 3.37, + "learning_rate": 4.6666377891362736e-05, + "loss": 1.7006, + "step": 38980 + }, + { + "epoch": 3.37, + "learning_rate": 4.6665511565450924e-05, + "loss": 1.6905, + "step": 38990 + }, + { + "epoch": 3.38, + "learning_rate": 4.666464523953912e-05, + "loss": 1.6367, + "step": 39000 + }, + { + "epoch": 3.38, + "learning_rate": 4.6663778913627306e-05, + "loss": 1.6801, + "step": 39010 + }, + { + "epoch": 3.38, + "learning_rate": 4.66629125877155e-05, + "loss": 1.691, + "step": 39020 + }, + { + "epoch": 3.38, + "learning_rate": 4.6662046261803695e-05, + "loss": 1.7018, + "step": 39030 + }, + { + "epoch": 3.38, + "learning_rate": 4.666117993589188e-05, + "loss": 1.6567, + "step": 39040 + }, + { + "epoch": 3.38, + "learning_rate": 4.666031360998008e-05, + "loss": 1.6724, + "step": 39050 + }, + { + "epoch": 3.38, + "learning_rate": 4.665944728406827e-05, + "loss": 1.7004, + "step": 39060 + }, + { + "epoch": 3.38, + "learning_rate": 4.665858095815646e-05, + "loss": 1.6982, + "step": 39070 + }, + { + "epoch": 3.38, + "learning_rate": 4.6657714632244654e-05, + "loss": 1.7323, + "step": 39080 + }, + { + "epoch": 3.38, + "learning_rate": 4.665684830633285e-05, + "loss": 1.7493, + "step": 39090 + }, + { + "epoch": 3.38, + "learning_rate": 4.6655981980421036e-05, + "loss": 1.7363, + "step": 39100 + }, + { + "epoch": 3.39, + "learning_rate": 4.665511565450923e-05, + "loss": 1.6526, + "step": 39110 + }, + { + "epoch": 3.39, + "learning_rate": 4.665424932859742e-05, + "loss": 1.6934, + "step": 39120 + }, + { + "epoch": 3.39, + "learning_rate": 4.665338300268561e-05, + "loss": 1.7276, + "step": 39130 + }, + { + "epoch": 3.39, + "learning_rate": 4.665251667677381e-05, + "loss": 1.6966, + "step": 39140 + }, + { + "epoch": 3.39, + "learning_rate": 4.6651650350861995e-05, + "loss": 1.7237, + "step": 39150 + }, + { + "epoch": 3.39, + "learning_rate": 4.665078402495019e-05, + "loss": 1.7506, + "step": 39160 + }, + { + "epoch": 3.39, + "learning_rate": 4.6649917699038384e-05, + "loss": 1.6644, + "step": 39170 + }, + { + "epoch": 3.39, + "learning_rate": 4.664905137312657e-05, + "loss": 1.6972, + "step": 39180 + }, + { + "epoch": 3.39, + "learning_rate": 4.6648185047214766e-05, + "loss": 1.693, + "step": 39190 + }, + { + "epoch": 3.39, + "learning_rate": 4.664731872130296e-05, + "loss": 1.6672, + "step": 39200 + }, + { + "epoch": 3.39, + "learning_rate": 4.664645239539115e-05, + "loss": 1.7395, + "step": 39210 + }, + { + "epoch": 3.39, + "learning_rate": 4.664558606947934e-05, + "loss": 1.7135, + "step": 39220 + }, + { + "epoch": 3.4, + "learning_rate": 4.664471974356753e-05, + "loss": 1.6724, + "step": 39230 + }, + { + "epoch": 3.4, + "learning_rate": 4.6643853417655724e-05, + "loss": 1.678, + "step": 39240 + }, + { + "epoch": 3.4, + "learning_rate": 4.664298709174392e-05, + "loss": 1.7023, + "step": 39250 + }, + { + "epoch": 3.4, + "learning_rate": 4.6642120765832107e-05, + "loss": 1.5717, + "step": 39260 + }, + { + "epoch": 3.4, + "learning_rate": 4.66412544399203e-05, + "loss": 1.7006, + "step": 39270 + }, + { + "epoch": 3.4, + "learning_rate": 4.6640388114008495e-05, + "loss": 1.6854, + "step": 39280 + }, + { + "epoch": 3.4, + "learning_rate": 4.663952178809668e-05, + "loss": 1.6776, + "step": 39290 + }, + { + "epoch": 3.4, + "learning_rate": 4.663865546218488e-05, + "loss": 1.7412, + "step": 39300 + }, + { + "epoch": 3.4, + "learning_rate": 4.663778913627307e-05, + "loss": 1.7501, + "step": 39310 + }, + { + "epoch": 3.4, + "learning_rate": 4.663692281036126e-05, + "loss": 1.6731, + "step": 39320 + }, + { + "epoch": 3.4, + "learning_rate": 4.6636056484449454e-05, + "loss": 1.6539, + "step": 39330 + }, + { + "epoch": 3.41, + "learning_rate": 4.663519015853764e-05, + "loss": 1.7113, + "step": 39340 + }, + { + "epoch": 3.41, + "learning_rate": 4.6634323832625836e-05, + "loss": 1.7032, + "step": 39350 + }, + { + "epoch": 3.41, + "learning_rate": 4.663345750671403e-05, + "loss": 1.6867, + "step": 39360 + }, + { + "epoch": 3.41, + "learning_rate": 4.663259118080222e-05, + "loss": 1.6332, + "step": 39370 + }, + { + "epoch": 3.41, + "learning_rate": 4.663172485489041e-05, + "loss": 1.7436, + "step": 39380 + }, + { + "epoch": 3.41, + "learning_rate": 4.663085852897861e-05, + "loss": 1.6678, + "step": 39390 + }, + { + "epoch": 3.41, + "learning_rate": 4.6629992203066795e-05, + "loss": 1.6744, + "step": 39400 + }, + { + "epoch": 3.41, + "learning_rate": 4.662912587715499e-05, + "loss": 1.7334, + "step": 39410 + }, + { + "epoch": 3.41, + "learning_rate": 4.6628259551243184e-05, + "loss": 1.7249, + "step": 39420 + }, + { + "epoch": 3.41, + "learning_rate": 4.662739322533137e-05, + "loss": 1.6676, + "step": 39430 + }, + { + "epoch": 3.41, + "learning_rate": 4.6626526899419566e-05, + "loss": 1.6704, + "step": 39440 + }, + { + "epoch": 3.41, + "learning_rate": 4.6625660573507754e-05, + "loss": 1.7333, + "step": 39450 + }, + { + "epoch": 3.42, + "learning_rate": 4.662479424759595e-05, + "loss": 1.7014, + "step": 39460 + }, + { + "epoch": 3.42, + "learning_rate": 4.662392792168414e-05, + "loss": 1.712, + "step": 39470 + }, + { + "epoch": 3.42, + "learning_rate": 4.662306159577233e-05, + "loss": 1.7416, + "step": 39480 + }, + { + "epoch": 3.42, + "learning_rate": 4.6622195269860525e-05, + "loss": 1.7045, + "step": 39490 + }, + { + "epoch": 3.42, + "learning_rate": 4.662132894394872e-05, + "loss": 1.6211, + "step": 39500 + }, + { + "epoch": 3.42, + "learning_rate": 4.662046261803691e-05, + "loss": 1.7097, + "step": 39510 + }, + { + "epoch": 3.42, + "learning_rate": 4.66195962921251e-05, + "loss": 1.64, + "step": 39520 + }, + { + "epoch": 3.42, + "learning_rate": 4.6618729966213296e-05, + "loss": 1.6788, + "step": 39530 + }, + { + "epoch": 3.42, + "learning_rate": 4.6617863640301483e-05, + "loss": 1.7402, + "step": 39540 + }, + { + "epoch": 3.42, + "learning_rate": 4.661699731438968e-05, + "loss": 1.6891, + "step": 39550 + }, + { + "epoch": 3.42, + "learning_rate": 4.6616130988477866e-05, + "loss": 1.7004, + "step": 39560 + }, + { + "epoch": 3.42, + "learning_rate": 4.661526466256606e-05, + "loss": 1.6367, + "step": 39570 + }, + { + "epoch": 3.43, + "learning_rate": 4.6614398336654255e-05, + "loss": 1.6486, + "step": 39580 + }, + { + "epoch": 3.43, + "learning_rate": 4.661353201074244e-05, + "loss": 1.6962, + "step": 39590 + }, + { + "epoch": 3.43, + "learning_rate": 4.661266568483064e-05, + "loss": 1.6453, + "step": 39600 + }, + { + "epoch": 3.43, + "learning_rate": 4.661179935891883e-05, + "loss": 1.6599, + "step": 39610 + }, + { + "epoch": 3.43, + "learning_rate": 4.661093303300702e-05, + "loss": 1.6218, + "step": 39620 + }, + { + "epoch": 3.43, + "learning_rate": 4.661006670709521e-05, + "loss": 1.6991, + "step": 39630 + }, + { + "epoch": 3.43, + "learning_rate": 4.660920038118341e-05, + "loss": 1.641, + "step": 39640 + }, + { + "epoch": 3.43, + "learning_rate": 4.6608334055271595e-05, + "loss": 1.6481, + "step": 39650 + }, + { + "epoch": 3.43, + "learning_rate": 4.660746772935979e-05, + "loss": 1.6952, + "step": 39660 + }, + { + "epoch": 3.43, + "learning_rate": 4.660660140344798e-05, + "loss": 1.6485, + "step": 39670 + }, + { + "epoch": 3.43, + "learning_rate": 4.660573507753617e-05, + "loss": 1.6512, + "step": 39680 + }, + { + "epoch": 3.44, + "learning_rate": 4.6604868751624366e-05, + "loss": 1.697, + "step": 39690 + }, + { + "epoch": 3.44, + "learning_rate": 4.6604002425712554e-05, + "loss": 1.6717, + "step": 39700 + }, + { + "epoch": 3.44, + "learning_rate": 4.660313609980075e-05, + "loss": 1.7088, + "step": 39710 + }, + { + "epoch": 3.44, + "learning_rate": 4.660226977388894e-05, + "loss": 1.771, + "step": 39720 + }, + { + "epoch": 3.44, + "learning_rate": 4.660140344797713e-05, + "loss": 1.7175, + "step": 39730 + }, + { + "epoch": 3.44, + "learning_rate": 4.6600537122065325e-05, + "loss": 1.6849, + "step": 39740 + }, + { + "epoch": 3.44, + "learning_rate": 4.659967079615351e-05, + "loss": 1.741, + "step": 39750 + }, + { + "epoch": 3.44, + "learning_rate": 4.659880447024171e-05, + "loss": 1.7438, + "step": 39760 + }, + { + "epoch": 3.44, + "learning_rate": 4.65979381443299e-05, + "loss": 1.7011, + "step": 39770 + }, + { + "epoch": 3.44, + "learning_rate": 4.659707181841809e-05, + "loss": 1.6365, + "step": 39780 + }, + { + "epoch": 3.44, + "learning_rate": 4.6596205492506284e-05, + "loss": 1.7231, + "step": 39790 + }, + { + "epoch": 3.44, + "learning_rate": 4.659533916659448e-05, + "loss": 1.6576, + "step": 39800 + }, + { + "epoch": 3.45, + "learning_rate": 4.6594472840682666e-05, + "loss": 1.7285, + "step": 39810 + }, + { + "epoch": 3.45, + "learning_rate": 4.659360651477086e-05, + "loss": 1.7084, + "step": 39820 + }, + { + "epoch": 3.45, + "learning_rate": 4.6592740188859055e-05, + "loss": 1.7023, + "step": 39830 + }, + { + "epoch": 3.45, + "learning_rate": 4.659187386294724e-05, + "loss": 1.7146, + "step": 39840 + }, + { + "epoch": 3.45, + "learning_rate": 4.659100753703544e-05, + "loss": 1.661, + "step": 39850 + }, + { + "epoch": 3.45, + "learning_rate": 4.6590141211123625e-05, + "loss": 1.7413, + "step": 39860 + }, + { + "epoch": 3.45, + "learning_rate": 4.658927488521182e-05, + "loss": 1.58, + "step": 39870 + }, + { + "epoch": 3.45, + "learning_rate": 4.6588408559300014e-05, + "loss": 1.6716, + "step": 39880 + }, + { + "epoch": 3.45, + "learning_rate": 4.65875422333882e-05, + "loss": 1.6867, + "step": 39890 + }, + { + "epoch": 3.45, + "learning_rate": 4.6586675907476396e-05, + "loss": 1.6113, + "step": 39900 + }, + { + "epoch": 3.45, + "learning_rate": 4.658580958156459e-05, + "loss": 1.6852, + "step": 39910 + }, + { + "epoch": 3.46, + "learning_rate": 4.658494325565278e-05, + "loss": 1.6838, + "step": 39920 + }, + { + "epoch": 3.46, + "learning_rate": 4.658407692974097e-05, + "loss": 1.6842, + "step": 39930 + }, + { + "epoch": 3.46, + "learning_rate": 4.658321060382917e-05, + "loss": 1.6874, + "step": 39940 + }, + { + "epoch": 3.46, + "learning_rate": 4.6582344277917354e-05, + "loss": 1.6738, + "step": 39950 + }, + { + "epoch": 3.46, + "learning_rate": 4.658147795200555e-05, + "loss": 1.6144, + "step": 39960 + }, + { + "epoch": 3.46, + "learning_rate": 4.6580611626093737e-05, + "loss": 1.5969, + "step": 39970 + }, + { + "epoch": 3.46, + "learning_rate": 4.657974530018193e-05, + "loss": 1.6522, + "step": 39980 + }, + { + "epoch": 3.46, + "learning_rate": 4.6578878974270125e-05, + "loss": 1.6829, + "step": 39990 + }, + { + "epoch": 3.46, + "learning_rate": 4.657801264835831e-05, + "loss": 1.6641, + "step": 40000 + }, + { + "epoch": 3.46, + "learning_rate": 4.657714632244651e-05, + "loss": 1.7015, + "step": 40010 + }, + { + "epoch": 3.46, + "learning_rate": 4.65762799965347e-05, + "loss": 1.6822, + "step": 40020 + }, + { + "epoch": 3.46, + "learning_rate": 4.657541367062289e-05, + "loss": 1.6607, + "step": 40030 + }, + { + "epoch": 3.47, + "learning_rate": 4.6574547344711084e-05, + "loss": 1.6789, + "step": 40040 + }, + { + "epoch": 3.47, + "learning_rate": 4.657368101879928e-05, + "loss": 1.64, + "step": 40050 + }, + { + "epoch": 3.47, + "learning_rate": 4.6572814692887466e-05, + "loss": 1.6726, + "step": 40060 + }, + { + "epoch": 3.47, + "learning_rate": 4.657194836697566e-05, + "loss": 1.6364, + "step": 40070 + }, + { + "epoch": 3.47, + "learning_rate": 4.657108204106385e-05, + "loss": 1.6997, + "step": 40080 + }, + { + "epoch": 3.47, + "learning_rate": 4.657021571515204e-05, + "loss": 1.6608, + "step": 40090 + }, + { + "epoch": 3.47, + "learning_rate": 4.656934938924024e-05, + "loss": 1.7392, + "step": 40100 + }, + { + "epoch": 3.47, + "learning_rate": 4.6568483063328425e-05, + "loss": 1.7255, + "step": 40110 + }, + { + "epoch": 3.47, + "learning_rate": 4.656761673741662e-05, + "loss": 1.6764, + "step": 40120 + }, + { + "epoch": 3.47, + "learning_rate": 4.6566750411504814e-05, + "loss": 1.6744, + "step": 40130 + }, + { + "epoch": 3.47, + "learning_rate": 4.6565884085593e-05, + "loss": 1.6625, + "step": 40140 + }, + { + "epoch": 3.48, + "learning_rate": 4.6565017759681196e-05, + "loss": 1.7119, + "step": 40150 + }, + { + "epoch": 3.48, + "learning_rate": 4.656415143376939e-05, + "loss": 1.6832, + "step": 40160 + }, + { + "epoch": 3.48, + "learning_rate": 4.656328510785758e-05, + "loss": 1.5883, + "step": 40170 + }, + { + "epoch": 3.48, + "learning_rate": 4.656241878194577e-05, + "loss": 1.718, + "step": 40180 + }, + { + "epoch": 3.48, + "learning_rate": 4.656155245603396e-05, + "loss": 1.6891, + "step": 40190 + }, + { + "epoch": 3.48, + "learning_rate": 4.6560686130122155e-05, + "loss": 1.6833, + "step": 40200 + }, + { + "epoch": 3.48, + "learning_rate": 4.655981980421035e-05, + "loss": 1.7323, + "step": 40210 + }, + { + "epoch": 3.48, + "learning_rate": 4.655895347829854e-05, + "loss": 1.7568, + "step": 40220 + }, + { + "epoch": 3.48, + "learning_rate": 4.655808715238673e-05, + "loss": 1.6458, + "step": 40230 + }, + { + "epoch": 3.48, + "learning_rate": 4.6557220826474926e-05, + "loss": 1.6461, + "step": 40240 + }, + { + "epoch": 3.48, + "learning_rate": 4.6556354500563113e-05, + "loss": 1.7246, + "step": 40250 + }, + { + "epoch": 3.48, + "learning_rate": 4.655548817465131e-05, + "loss": 1.7393, + "step": 40260 + }, + { + "epoch": 3.49, + "learning_rate": 4.65546218487395e-05, + "loss": 1.7093, + "step": 40270 + }, + { + "epoch": 3.49, + "learning_rate": 4.655375552282769e-05, + "loss": 1.6793, + "step": 40280 + }, + { + "epoch": 3.49, + "learning_rate": 4.6552889196915884e-05, + "loss": 1.6811, + "step": 40290 + }, + { + "epoch": 3.49, + "learning_rate": 4.655202287100407e-05, + "loss": 1.6738, + "step": 40300 + }, + { + "epoch": 3.49, + "learning_rate": 4.655115654509227e-05, + "loss": 1.679, + "step": 40310 + }, + { + "epoch": 3.49, + "learning_rate": 4.655029021918046e-05, + "loss": 1.6763, + "step": 40320 + }, + { + "epoch": 3.49, + "learning_rate": 4.654942389326865e-05, + "loss": 1.6817, + "step": 40330 + }, + { + "epoch": 3.49, + "learning_rate": 4.654855756735684e-05, + "loss": 1.7155, + "step": 40340 + }, + { + "epoch": 3.49, + "learning_rate": 4.654769124144504e-05, + "loss": 1.6447, + "step": 40350 + }, + { + "epoch": 3.49, + "learning_rate": 4.6546824915533225e-05, + "loss": 1.7442, + "step": 40360 + }, + { + "epoch": 3.49, + "learning_rate": 4.654595858962142e-05, + "loss": 1.696, + "step": 40370 + }, + { + "epoch": 3.5, + "learning_rate": 4.6545092263709614e-05, + "loss": 1.6173, + "step": 40380 + }, + { + "epoch": 3.5, + "learning_rate": 4.65442259377978e-05, + "loss": 1.7027, + "step": 40390 + }, + { + "epoch": 3.5, + "learning_rate": 4.6543359611885996e-05, + "loss": 1.6284, + "step": 40400 + }, + { + "epoch": 3.5, + "learning_rate": 4.6542493285974184e-05, + "loss": 1.8004, + "step": 40410 + }, + { + "epoch": 3.5, + "learning_rate": 4.654162696006238e-05, + "loss": 1.6823, + "step": 40420 + }, + { + "epoch": 3.5, + "learning_rate": 4.654076063415057e-05, + "loss": 1.7103, + "step": 40430 + }, + { + "epoch": 3.5, + "learning_rate": 4.653989430823876e-05, + "loss": 1.6577, + "step": 40440 + }, + { + "epoch": 3.5, + "learning_rate": 4.6539027982326955e-05, + "loss": 1.7229, + "step": 40450 + }, + { + "epoch": 3.5, + "learning_rate": 4.653816165641515e-05, + "loss": 1.7289, + "step": 40460 + }, + { + "epoch": 3.5, + "learning_rate": 4.653729533050334e-05, + "loss": 1.7506, + "step": 40470 + }, + { + "epoch": 3.5, + "learning_rate": 4.653642900459153e-05, + "loss": 1.6801, + "step": 40480 + }, + { + "epoch": 3.5, + "learning_rate": 4.6535562678679726e-05, + "loss": 1.6568, + "step": 40490 + }, + { + "epoch": 3.51, + "learning_rate": 4.6534696352767914e-05, + "loss": 1.6957, + "step": 40500 + }, + { + "epoch": 3.51, + "learning_rate": 4.653383002685611e-05, + "loss": 1.6796, + "step": 40510 + }, + { + "epoch": 3.51, + "learning_rate": 4.6532963700944296e-05, + "loss": 1.6155, + "step": 40520 + }, + { + "epoch": 3.51, + "learning_rate": 4.653209737503249e-05, + "loss": 1.5823, + "step": 40530 + }, + { + "epoch": 3.51, + "learning_rate": 4.6531231049120685e-05, + "loss": 1.7205, + "step": 40540 + }, + { + "epoch": 3.51, + "learning_rate": 4.653036472320887e-05, + "loss": 1.6725, + "step": 40550 + }, + { + "epoch": 3.51, + "learning_rate": 4.652949839729707e-05, + "loss": 1.6698, + "step": 40560 + }, + { + "epoch": 3.51, + "learning_rate": 4.652863207138526e-05, + "loss": 1.6927, + "step": 40570 + }, + { + "epoch": 3.51, + "learning_rate": 4.652776574547345e-05, + "loss": 1.6242, + "step": 40580 + }, + { + "epoch": 3.51, + "learning_rate": 4.6526899419561644e-05, + "loss": 1.6311, + "step": 40590 + }, + { + "epoch": 3.51, + "learning_rate": 4.652603309364983e-05, + "loss": 1.6181, + "step": 40600 + }, + { + "epoch": 3.51, + "learning_rate": 4.6525166767738026e-05, + "loss": 1.6704, + "step": 40610 + }, + { + "epoch": 3.52, + "learning_rate": 4.652430044182622e-05, + "loss": 1.6919, + "step": 40620 + }, + { + "epoch": 3.52, + "learning_rate": 4.652343411591441e-05, + "loss": 1.6799, + "step": 40630 + }, + { + "epoch": 3.52, + "learning_rate": 4.65225677900026e-05, + "loss": 1.693, + "step": 40640 + }, + { + "epoch": 3.52, + "learning_rate": 4.65217014640908e-05, + "loss": 1.7336, + "step": 40650 + }, + { + "epoch": 3.52, + "learning_rate": 4.6520835138178984e-05, + "loss": 1.7256, + "step": 40660 + }, + { + "epoch": 3.52, + "learning_rate": 4.651996881226718e-05, + "loss": 1.6686, + "step": 40670 + }, + { + "epoch": 3.52, + "learning_rate": 4.651910248635537e-05, + "loss": 1.6189, + "step": 40680 + }, + { + "epoch": 3.52, + "learning_rate": 4.651823616044356e-05, + "loss": 1.6733, + "step": 40690 + }, + { + "epoch": 3.52, + "learning_rate": 4.6517369834531755e-05, + "loss": 1.6359, + "step": 40700 + }, + { + "epoch": 3.52, + "learning_rate": 4.651650350861994e-05, + "loss": 1.7783, + "step": 40710 + }, + { + "epoch": 3.52, + "learning_rate": 4.651563718270814e-05, + "loss": 1.6806, + "step": 40720 + }, + { + "epoch": 3.53, + "learning_rate": 4.651477085679633e-05, + "loss": 1.6319, + "step": 40730 + }, + { + "epoch": 3.53, + "learning_rate": 4.651390453088452e-05, + "loss": 1.6621, + "step": 40740 + }, + { + "epoch": 3.53, + "learning_rate": 4.6513038204972714e-05, + "loss": 1.6157, + "step": 40750 + }, + { + "epoch": 3.53, + "learning_rate": 4.651217187906091e-05, + "loss": 1.6828, + "step": 40760 + }, + { + "epoch": 3.53, + "learning_rate": 4.6511305553149096e-05, + "loss": 1.6947, + "step": 40770 + }, + { + "epoch": 3.53, + "learning_rate": 4.651043922723729e-05, + "loss": 1.7464, + "step": 40780 + }, + { + "epoch": 3.53, + "learning_rate": 4.6509572901325485e-05, + "loss": 1.7139, + "step": 40790 + }, + { + "epoch": 3.53, + "learning_rate": 4.650870657541367e-05, + "loss": 1.6404, + "step": 40800 + }, + { + "epoch": 3.53, + "learning_rate": 4.650784024950187e-05, + "loss": 1.6856, + "step": 40810 + }, + { + "epoch": 3.53, + "learning_rate": 4.6506973923590055e-05, + "loss": 1.7247, + "step": 40820 + }, + { + "epoch": 3.53, + "learning_rate": 4.650610759767825e-05, + "loss": 1.6559, + "step": 40830 + }, + { + "epoch": 3.53, + "learning_rate": 4.6505241271766444e-05, + "loss": 1.7483, + "step": 40840 + }, + { + "epoch": 3.54, + "learning_rate": 4.650437494585463e-05, + "loss": 1.7408, + "step": 40850 + }, + { + "epoch": 3.54, + "learning_rate": 4.6503508619942826e-05, + "loss": 1.69, + "step": 40860 + }, + { + "epoch": 3.54, + "learning_rate": 4.650264229403102e-05, + "loss": 1.6885, + "step": 40870 + }, + { + "epoch": 3.54, + "learning_rate": 4.650177596811921e-05, + "loss": 1.6776, + "step": 40880 + }, + { + "epoch": 3.54, + "learning_rate": 4.65009096422074e-05, + "loss": 1.6537, + "step": 40890 + }, + { + "epoch": 3.54, + "learning_rate": 4.65000433162956e-05, + "loss": 1.6481, + "step": 40900 + }, + { + "epoch": 3.54, + "learning_rate": 4.6499176990383785e-05, + "loss": 1.7272, + "step": 40910 + }, + { + "epoch": 3.54, + "learning_rate": 4.649831066447198e-05, + "loss": 1.719, + "step": 40920 + }, + { + "epoch": 3.54, + "learning_rate": 4.649744433856017e-05, + "loss": 1.7333, + "step": 40930 + }, + { + "epoch": 3.54, + "learning_rate": 4.649657801264836e-05, + "loss": 1.6127, + "step": 40940 + }, + { + "epoch": 3.54, + "learning_rate": 4.6495711686736556e-05, + "loss": 1.7156, + "step": 40950 + }, + { + "epoch": 3.55, + "learning_rate": 4.6494845360824743e-05, + "loss": 1.6713, + "step": 40960 + }, + { + "epoch": 3.55, + "learning_rate": 4.649397903491294e-05, + "loss": 1.6585, + "step": 40970 + }, + { + "epoch": 3.55, + "learning_rate": 4.649311270900113e-05, + "loss": 1.6446, + "step": 40980 + }, + { + "epoch": 3.55, + "learning_rate": 4.649224638308932e-05, + "loss": 1.7197, + "step": 40990 + }, + { + "epoch": 3.55, + "learning_rate": 4.6491380057177514e-05, + "loss": 1.7401, + "step": 41000 + }, + { + "epoch": 3.55, + "learning_rate": 4.649051373126571e-05, + "loss": 1.6896, + "step": 41010 + }, + { + "epoch": 3.55, + "learning_rate": 4.6489647405353897e-05, + "loss": 1.6572, + "step": 41020 + }, + { + "epoch": 3.55, + "learning_rate": 4.648878107944209e-05, + "loss": 1.7013, + "step": 41030 + }, + { + "epoch": 3.55, + "learning_rate": 4.648791475353028e-05, + "loss": 1.6715, + "step": 41040 + }, + { + "epoch": 3.55, + "learning_rate": 4.648704842761847e-05, + "loss": 1.7237, + "step": 41050 + }, + { + "epoch": 3.55, + "learning_rate": 4.648618210170667e-05, + "loss": 1.7304, + "step": 41060 + }, + { + "epoch": 3.55, + "learning_rate": 4.6485315775794855e-05, + "loss": 1.682, + "step": 41070 + }, + { + "epoch": 3.56, + "learning_rate": 4.648444944988305e-05, + "loss": 1.6266, + "step": 41080 + }, + { + "epoch": 3.56, + "learning_rate": 4.6483583123971244e-05, + "loss": 1.6977, + "step": 41090 + }, + { + "epoch": 3.56, + "learning_rate": 4.648271679805943e-05, + "loss": 1.6779, + "step": 41100 + }, + { + "epoch": 3.56, + "learning_rate": 4.6481850472147626e-05, + "loss": 1.7475, + "step": 41110 + }, + { + "epoch": 3.56, + "learning_rate": 4.648098414623582e-05, + "loss": 1.6269, + "step": 41120 + }, + { + "epoch": 3.56, + "learning_rate": 4.648011782032401e-05, + "loss": 1.6523, + "step": 41130 + }, + { + "epoch": 3.56, + "learning_rate": 4.64792514944122e-05, + "loss": 1.6947, + "step": 41140 + }, + { + "epoch": 3.56, + "learning_rate": 4.647838516850039e-05, + "loss": 1.7031, + "step": 41150 + }, + { + "epoch": 3.56, + "learning_rate": 4.6477518842588585e-05, + "loss": 1.7246, + "step": 41160 + }, + { + "epoch": 3.56, + "learning_rate": 4.647665251667678e-05, + "loss": 1.6731, + "step": 41170 + }, + { + "epoch": 3.56, + "learning_rate": 4.647578619076497e-05, + "loss": 1.6545, + "step": 41180 + }, + { + "epoch": 3.57, + "learning_rate": 4.647491986485316e-05, + "loss": 1.6801, + "step": 41190 + }, + { + "epoch": 3.57, + "learning_rate": 4.6474053538941356e-05, + "loss": 1.6661, + "step": 41200 + }, + { + "epoch": 3.57, + "learning_rate": 4.6473187213029544e-05, + "loss": 1.6536, + "step": 41210 + }, + { + "epoch": 3.57, + "learning_rate": 4.647232088711774e-05, + "loss": 1.6783, + "step": 41220 + }, + { + "epoch": 3.57, + "learning_rate": 4.647145456120593e-05, + "loss": 1.6832, + "step": 41230 + }, + { + "epoch": 3.57, + "learning_rate": 4.647058823529412e-05, + "loss": 1.6889, + "step": 41240 + }, + { + "epoch": 3.57, + "learning_rate": 4.6469721909382315e-05, + "loss": 1.641, + "step": 41250 + }, + { + "epoch": 3.57, + "learning_rate": 4.64688555834705e-05, + "loss": 1.6217, + "step": 41260 + }, + { + "epoch": 3.57, + "learning_rate": 4.64679892575587e-05, + "loss": 1.6125, + "step": 41270 + }, + { + "epoch": 3.57, + "learning_rate": 4.646712293164689e-05, + "loss": 1.7408, + "step": 41280 + }, + { + "epoch": 3.57, + "learning_rate": 4.646625660573508e-05, + "loss": 1.6835, + "step": 41290 + }, + { + "epoch": 3.57, + "learning_rate": 4.6465390279823274e-05, + "loss": 1.6904, + "step": 41300 + }, + { + "epoch": 3.58, + "learning_rate": 4.646452395391147e-05, + "loss": 1.7128, + "step": 41310 + }, + { + "epoch": 3.58, + "learning_rate": 4.6463657627999656e-05, + "loss": 1.6303, + "step": 41320 + }, + { + "epoch": 3.58, + "learning_rate": 4.646279130208785e-05, + "loss": 1.6627, + "step": 41330 + }, + { + "epoch": 3.58, + "learning_rate": 4.646192497617604e-05, + "loss": 1.6683, + "step": 41340 + }, + { + "epoch": 3.58, + "learning_rate": 4.646105865026423e-05, + "loss": 1.6578, + "step": 41350 + }, + { + "epoch": 3.58, + "learning_rate": 4.646019232435243e-05, + "loss": 1.684, + "step": 41360 + }, + { + "epoch": 3.58, + "learning_rate": 4.6459325998440614e-05, + "loss": 1.6965, + "step": 41370 + }, + { + "epoch": 3.58, + "learning_rate": 4.645845967252881e-05, + "loss": 1.647, + "step": 41380 + }, + { + "epoch": 3.58, + "learning_rate": 4.6457593346617e-05, + "loss": 1.6753, + "step": 41390 + }, + { + "epoch": 3.58, + "learning_rate": 4.645672702070519e-05, + "loss": 1.7132, + "step": 41400 + }, + { + "epoch": 3.58, + "learning_rate": 4.6455860694793385e-05, + "loss": 1.617, + "step": 41410 + }, + { + "epoch": 3.59, + "learning_rate": 4.645499436888158e-05, + "loss": 1.6688, + "step": 41420 + }, + { + "epoch": 3.59, + "learning_rate": 4.645412804296977e-05, + "loss": 1.6576, + "step": 41430 + }, + { + "epoch": 3.59, + "learning_rate": 4.645326171705796e-05, + "loss": 1.6684, + "step": 41440 + }, + { + "epoch": 3.59, + "learning_rate": 4.645239539114615e-05, + "loss": 1.6927, + "step": 41450 + }, + { + "epoch": 3.59, + "learning_rate": 4.6451529065234344e-05, + "loss": 1.779, + "step": 41460 + }, + { + "epoch": 3.59, + "learning_rate": 4.645066273932254e-05, + "loss": 1.6653, + "step": 41470 + }, + { + "epoch": 3.59, + "learning_rate": 4.6449796413410726e-05, + "loss": 1.6844, + "step": 41480 + }, + { + "epoch": 3.59, + "learning_rate": 4.644893008749892e-05, + "loss": 1.651, + "step": 41490 + }, + { + "epoch": 3.59, + "learning_rate": 4.6448063761587115e-05, + "loss": 1.5945, + "step": 41500 + }, + { + "epoch": 3.59, + "learning_rate": 4.64471974356753e-05, + "loss": 1.725, + "step": 41510 + }, + { + "epoch": 3.59, + "learning_rate": 4.64463311097635e-05, + "loss": 1.6857, + "step": 41520 + }, + { + "epoch": 3.59, + "learning_rate": 4.644546478385169e-05, + "loss": 1.6995, + "step": 41530 + }, + { + "epoch": 3.6, + "learning_rate": 4.644459845793988e-05, + "loss": 1.7181, + "step": 41540 + }, + { + "epoch": 3.6, + "learning_rate": 4.6443732132028074e-05, + "loss": 1.6672, + "step": 41550 + }, + { + "epoch": 3.6, + "learning_rate": 4.644286580611626e-05, + "loss": 1.6812, + "step": 41560 + }, + { + "epoch": 3.6, + "learning_rate": 4.6441999480204456e-05, + "loss": 1.6204, + "step": 41570 + }, + { + "epoch": 3.6, + "learning_rate": 4.644113315429265e-05, + "loss": 1.7433, + "step": 41580 + }, + { + "epoch": 3.6, + "learning_rate": 4.644026682838084e-05, + "loss": 1.7231, + "step": 41590 + }, + { + "epoch": 3.6, + "learning_rate": 4.643940050246903e-05, + "loss": 1.6677, + "step": 41600 + }, + { + "epoch": 3.6, + "learning_rate": 4.643853417655723e-05, + "loss": 1.6857, + "step": 41610 + }, + { + "epoch": 3.6, + "learning_rate": 4.6437667850645415e-05, + "loss": 1.7195, + "step": 41620 + }, + { + "epoch": 3.6, + "learning_rate": 4.643680152473361e-05, + "loss": 1.6994, + "step": 41630 + }, + { + "epoch": 3.6, + "learning_rate": 4.6435935198821804e-05, + "loss": 1.7056, + "step": 41640 + }, + { + "epoch": 3.6, + "learning_rate": 4.643506887290999e-05, + "loss": 1.7304, + "step": 41650 + }, + { + "epoch": 3.61, + "learning_rate": 4.6434202546998186e-05, + "loss": 1.6562, + "step": 41660 + }, + { + "epoch": 3.61, + "learning_rate": 4.6433336221086373e-05, + "loss": 1.6965, + "step": 41670 + }, + { + "epoch": 3.61, + "learning_rate": 4.643246989517457e-05, + "loss": 1.6807, + "step": 41680 + }, + { + "epoch": 3.61, + "learning_rate": 4.643160356926276e-05, + "loss": 1.6653, + "step": 41690 + }, + { + "epoch": 3.61, + "learning_rate": 4.643073724335095e-05, + "loss": 1.6303, + "step": 41700 + }, + { + "epoch": 3.61, + "learning_rate": 4.6429870917439144e-05, + "loss": 1.676, + "step": 41710 + }, + { + "epoch": 3.61, + "learning_rate": 4.642900459152734e-05, + "loss": 1.6969, + "step": 41720 + }, + { + "epoch": 3.61, + "learning_rate": 4.6428138265615527e-05, + "loss": 1.6614, + "step": 41730 + }, + { + "epoch": 3.61, + "learning_rate": 4.642727193970372e-05, + "loss": 1.6977, + "step": 41740 + }, + { + "epoch": 3.61, + "learning_rate": 4.6426405613791915e-05, + "loss": 1.7515, + "step": 41750 + }, + { + "epoch": 3.61, + "learning_rate": 4.64255392878801e-05, + "loss": 1.7146, + "step": 41760 + }, + { + "epoch": 3.62, + "learning_rate": 4.64246729619683e-05, + "loss": 1.686, + "step": 41770 + }, + { + "epoch": 3.62, + "learning_rate": 4.6423806636056485e-05, + "loss": 1.6549, + "step": 41780 + }, + { + "epoch": 3.62, + "learning_rate": 4.642294031014468e-05, + "loss": 1.6732, + "step": 41790 + }, + { + "epoch": 3.62, + "learning_rate": 4.6422073984232874e-05, + "loss": 1.6881, + "step": 41800 + }, + { + "epoch": 3.62, + "learning_rate": 4.642120765832106e-05, + "loss": 1.691, + "step": 41810 + }, + { + "epoch": 3.62, + "learning_rate": 4.6420341332409256e-05, + "loss": 1.7169, + "step": 41820 + }, + { + "epoch": 3.62, + "learning_rate": 4.641947500649745e-05, + "loss": 1.749, + "step": 41830 + }, + { + "epoch": 3.62, + "learning_rate": 4.641860868058564e-05, + "loss": 1.615, + "step": 41840 + }, + { + "epoch": 3.62, + "learning_rate": 4.641774235467383e-05, + "loss": 1.6311, + "step": 41850 + }, + { + "epoch": 3.62, + "learning_rate": 4.641687602876203e-05, + "loss": 1.6627, + "step": 41860 + }, + { + "epoch": 3.62, + "learning_rate": 4.6416009702850215e-05, + "loss": 1.759, + "step": 41870 + }, + { + "epoch": 3.62, + "learning_rate": 4.641514337693841e-05, + "loss": 1.64, + "step": 41880 + }, + { + "epoch": 3.63, + "learning_rate": 4.64142770510266e-05, + "loss": 1.6969, + "step": 41890 + }, + { + "epoch": 3.63, + "learning_rate": 4.641341072511479e-05, + "loss": 1.6794, + "step": 41900 + }, + { + "epoch": 3.63, + "learning_rate": 4.6412544399202986e-05, + "loss": 1.7508, + "step": 41910 + }, + { + "epoch": 3.63, + "learning_rate": 4.6411678073291174e-05, + "loss": 1.6752, + "step": 41920 + }, + { + "epoch": 3.63, + "learning_rate": 4.641081174737937e-05, + "loss": 1.6925, + "step": 41930 + }, + { + "epoch": 3.63, + "learning_rate": 4.640994542146756e-05, + "loss": 1.6811, + "step": 41940 + }, + { + "epoch": 3.63, + "learning_rate": 4.640907909555575e-05, + "loss": 1.6718, + "step": 41950 + }, + { + "epoch": 3.63, + "learning_rate": 4.6408212769643945e-05, + "loss": 1.7292, + "step": 41960 + }, + { + "epoch": 3.63, + "learning_rate": 4.640734644373214e-05, + "loss": 1.6481, + "step": 41970 + }, + { + "epoch": 3.63, + "learning_rate": 4.640648011782033e-05, + "loss": 1.7046, + "step": 41980 + }, + { + "epoch": 3.63, + "learning_rate": 4.640561379190852e-05, + "loss": 1.7176, + "step": 41990 + }, + { + "epoch": 3.64, + "learning_rate": 4.640474746599671e-05, + "loss": 1.703, + "step": 42000 + }, + { + "epoch": 3.64, + "learning_rate": 4.6403881140084903e-05, + "loss": 1.6378, + "step": 42010 + }, + { + "epoch": 3.64, + "learning_rate": 4.64030148141731e-05, + "loss": 1.6558, + "step": 42020 + }, + { + "epoch": 3.64, + "learning_rate": 4.6402148488261286e-05, + "loss": 1.734, + "step": 42030 + }, + { + "epoch": 3.64, + "learning_rate": 4.640128216234948e-05, + "loss": 1.6844, + "step": 42040 + }, + { + "epoch": 3.64, + "learning_rate": 4.6400415836437675e-05, + "loss": 1.7115, + "step": 42050 + }, + { + "epoch": 3.64, + "learning_rate": 4.639954951052586e-05, + "loss": 1.6445, + "step": 42060 + }, + { + "epoch": 3.64, + "learning_rate": 4.639868318461406e-05, + "loss": 1.6863, + "step": 42070 + }, + { + "epoch": 3.64, + "learning_rate": 4.6397816858702244e-05, + "loss": 1.6508, + "step": 42080 + }, + { + "epoch": 3.64, + "learning_rate": 4.639695053279044e-05, + "loss": 1.6117, + "step": 42090 + }, + { + "epoch": 3.64, + "learning_rate": 4.639608420687863e-05, + "loss": 1.7006, + "step": 42100 + }, + { + "epoch": 3.64, + "learning_rate": 4.639521788096682e-05, + "loss": 1.6488, + "step": 42110 + }, + { + "epoch": 3.65, + "learning_rate": 4.6394351555055015e-05, + "loss": 1.6493, + "step": 42120 + }, + { + "epoch": 3.65, + "learning_rate": 4.639348522914321e-05, + "loss": 1.7503, + "step": 42130 + }, + { + "epoch": 3.65, + "learning_rate": 4.63926189032314e-05, + "loss": 1.7204, + "step": 42140 + }, + { + "epoch": 3.65, + "learning_rate": 4.639175257731959e-05, + "loss": 1.5847, + "step": 42150 + }, + { + "epoch": 3.65, + "learning_rate": 4.6390886251407786e-05, + "loss": 1.7117, + "step": 42160 + }, + { + "epoch": 3.65, + "learning_rate": 4.6390019925495974e-05, + "loss": 1.6306, + "step": 42170 + }, + { + "epoch": 3.65, + "learning_rate": 4.638915359958417e-05, + "loss": 1.6042, + "step": 42180 + }, + { + "epoch": 3.65, + "learning_rate": 4.6388287273672356e-05, + "loss": 1.7012, + "step": 42190 + }, + { + "epoch": 3.65, + "learning_rate": 4.638742094776055e-05, + "loss": 1.7114, + "step": 42200 + }, + { + "epoch": 3.65, + "learning_rate": 4.6386554621848745e-05, + "loss": 1.7126, + "step": 42210 + }, + { + "epoch": 3.65, + "learning_rate": 4.638568829593693e-05, + "loss": 1.6303, + "step": 42220 + }, + { + "epoch": 3.66, + "learning_rate": 4.638482197002513e-05, + "loss": 1.6086, + "step": 42230 + }, + { + "epoch": 3.66, + "learning_rate": 4.638395564411332e-05, + "loss": 1.6467, + "step": 42240 + }, + { + "epoch": 3.66, + "learning_rate": 4.638308931820151e-05, + "loss": 1.5869, + "step": 42250 + }, + { + "epoch": 3.66, + "learning_rate": 4.6382222992289704e-05, + "loss": 1.6299, + "step": 42260 + }, + { + "epoch": 3.66, + "learning_rate": 4.63813566663779e-05, + "loss": 1.6644, + "step": 42270 + }, + { + "epoch": 3.66, + "learning_rate": 4.6380490340466086e-05, + "loss": 1.62, + "step": 42280 + }, + { + "epoch": 3.66, + "learning_rate": 4.637962401455428e-05, + "loss": 1.7533, + "step": 42290 + }, + { + "epoch": 3.66, + "learning_rate": 4.637875768864247e-05, + "loss": 1.6756, + "step": 42300 + }, + { + "epoch": 3.66, + "learning_rate": 4.637789136273066e-05, + "loss": 1.7047, + "step": 42310 + }, + { + "epoch": 3.66, + "learning_rate": 4.637702503681886e-05, + "loss": 1.7085, + "step": 42320 + }, + { + "epoch": 3.66, + "learning_rate": 4.6376158710907045e-05, + "loss": 1.7042, + "step": 42330 + }, + { + "epoch": 3.66, + "learning_rate": 4.637529238499524e-05, + "loss": 1.6941, + "step": 42340 + }, + { + "epoch": 3.67, + "learning_rate": 4.6374426059083434e-05, + "loss": 1.6149, + "step": 42350 + }, + { + "epoch": 3.67, + "learning_rate": 4.637355973317162e-05, + "loss": 1.6805, + "step": 42360 + }, + { + "epoch": 3.67, + "learning_rate": 4.6372693407259816e-05, + "loss": 1.6723, + "step": 42370 + }, + { + "epoch": 3.67, + "learning_rate": 4.637182708134801e-05, + "loss": 1.6733, + "step": 42380 + }, + { + "epoch": 3.67, + "learning_rate": 4.63709607554362e-05, + "loss": 1.6622, + "step": 42390 + }, + { + "epoch": 3.67, + "learning_rate": 4.637009442952439e-05, + "loss": 1.6091, + "step": 42400 + }, + { + "epoch": 3.67, + "learning_rate": 4.636922810361258e-05, + "loss": 1.6746, + "step": 42410 + }, + { + "epoch": 3.67, + "learning_rate": 4.6368361777700774e-05, + "loss": 1.7003, + "step": 42420 + }, + { + "epoch": 3.67, + "learning_rate": 4.636749545178897e-05, + "loss": 1.6495, + "step": 42430 + }, + { + "epoch": 3.67, + "learning_rate": 4.6366629125877157e-05, + "loss": 1.7077, + "step": 42440 + }, + { + "epoch": 3.67, + "learning_rate": 4.636576279996535e-05, + "loss": 1.7226, + "step": 42450 + }, + { + "epoch": 3.68, + "learning_rate": 4.6364896474053545e-05, + "loss": 1.678, + "step": 42460 + }, + { + "epoch": 3.68, + "learning_rate": 4.636403014814173e-05, + "loss": 1.6459, + "step": 42470 + }, + { + "epoch": 3.68, + "learning_rate": 4.636316382222993e-05, + "loss": 1.7231, + "step": 42480 + }, + { + "epoch": 3.68, + "learning_rate": 4.636229749631812e-05, + "loss": 1.6936, + "step": 42490 + }, + { + "epoch": 3.68, + "learning_rate": 4.636143117040631e-05, + "loss": 1.6824, + "step": 42500 + }, + { + "epoch": 3.68, + "learning_rate": 4.6360564844494504e-05, + "loss": 1.6497, + "step": 42510 + }, + { + "epoch": 3.68, + "learning_rate": 4.635969851858269e-05, + "loss": 1.7063, + "step": 42520 + }, + { + "epoch": 3.68, + "learning_rate": 4.6358832192670886e-05, + "loss": 1.6258, + "step": 42530 + }, + { + "epoch": 3.68, + "learning_rate": 4.635796586675908e-05, + "loss": 1.6837, + "step": 42540 + }, + { + "epoch": 3.68, + "learning_rate": 4.635709954084727e-05, + "loss": 1.6527, + "step": 42550 + }, + { + "epoch": 3.68, + "learning_rate": 4.635623321493546e-05, + "loss": 1.5958, + "step": 42560 + }, + { + "epoch": 3.68, + "learning_rate": 4.635536688902366e-05, + "loss": 1.7079, + "step": 42570 + }, + { + "epoch": 3.69, + "learning_rate": 4.6354500563111845e-05, + "loss": 1.7395, + "step": 42580 + }, + { + "epoch": 3.69, + "learning_rate": 4.635363423720004e-05, + "loss": 1.7206, + "step": 42590 + }, + { + "epoch": 3.69, + "learning_rate": 4.6352767911288234e-05, + "loss": 1.671, + "step": 42600 + }, + { + "epoch": 3.69, + "learning_rate": 4.635190158537642e-05, + "loss": 1.7283, + "step": 42610 + }, + { + "epoch": 3.69, + "learning_rate": 4.6351035259464616e-05, + "loss": 1.7238, + "step": 42620 + }, + { + "epoch": 3.69, + "learning_rate": 4.6350168933552804e-05, + "loss": 1.6875, + "step": 42630 + }, + { + "epoch": 3.69, + "learning_rate": 4.6349302607641e-05, + "loss": 1.6921, + "step": 42640 + }, + { + "epoch": 3.69, + "learning_rate": 4.634843628172919e-05, + "loss": 1.6878, + "step": 42650 + }, + { + "epoch": 3.69, + "learning_rate": 4.634756995581738e-05, + "loss": 1.6963, + "step": 42660 + }, + { + "epoch": 3.69, + "learning_rate": 4.6346703629905575e-05, + "loss": 1.6846, + "step": 42670 + }, + { + "epoch": 3.69, + "learning_rate": 4.634583730399377e-05, + "loss": 1.7043, + "step": 42680 + }, + { + "epoch": 3.69, + "learning_rate": 4.634497097808196e-05, + "loss": 1.5418, + "step": 42690 + }, + { + "epoch": 3.7, + "learning_rate": 4.634410465217015e-05, + "loss": 1.6626, + "step": 42700 + }, + { + "epoch": 3.7, + "learning_rate": 4.6343238326258346e-05, + "loss": 1.6767, + "step": 42710 + }, + { + "epoch": 3.7, + "learning_rate": 4.6342372000346533e-05, + "loss": 1.6943, + "step": 42720 + }, + { + "epoch": 3.7, + "learning_rate": 4.634150567443473e-05, + "loss": 1.6603, + "step": 42730 + }, + { + "epoch": 3.7, + "learning_rate": 4.6340639348522916e-05, + "loss": 1.7074, + "step": 42740 + }, + { + "epoch": 3.7, + "learning_rate": 4.633977302261111e-05, + "loss": 1.6993, + "step": 42750 + }, + { + "epoch": 3.7, + "learning_rate": 4.6338906696699304e-05, + "loss": 1.6546, + "step": 42760 + }, + { + "epoch": 3.7, + "learning_rate": 4.633804037078749e-05, + "loss": 1.6504, + "step": 42770 + }, + { + "epoch": 3.7, + "learning_rate": 4.633717404487569e-05, + "loss": 1.6617, + "step": 42780 + }, + { + "epoch": 3.7, + "learning_rate": 4.633630771896388e-05, + "loss": 1.6386, + "step": 42790 + }, + { + "epoch": 3.7, + "learning_rate": 4.633544139305207e-05, + "loss": 1.6808, + "step": 42800 + }, + { + "epoch": 3.71, + "learning_rate": 4.633457506714026e-05, + "loss": 1.6351, + "step": 42810 + }, + { + "epoch": 3.71, + "learning_rate": 4.633370874122846e-05, + "loss": 1.6614, + "step": 42820 + }, + { + "epoch": 3.71, + "learning_rate": 4.6332842415316645e-05, + "loss": 1.6611, + "step": 42830 + }, + { + "epoch": 3.71, + "learning_rate": 4.633197608940484e-05, + "loss": 1.6916, + "step": 42840 + }, + { + "epoch": 3.71, + "learning_rate": 4.633110976349303e-05, + "loss": 1.6963, + "step": 42850 + }, + { + "epoch": 3.71, + "learning_rate": 4.633024343758122e-05, + "loss": 1.6986, + "step": 42860 + }, + { + "epoch": 3.71, + "learning_rate": 4.6329377111669416e-05, + "loss": 1.6766, + "step": 42870 + }, + { + "epoch": 3.71, + "learning_rate": 4.6328510785757604e-05, + "loss": 1.6818, + "step": 42880 + }, + { + "epoch": 3.71, + "learning_rate": 4.63276444598458e-05, + "loss": 1.7187, + "step": 42890 + }, + { + "epoch": 3.71, + "learning_rate": 4.632677813393399e-05, + "loss": 1.6043, + "step": 42900 + }, + { + "epoch": 3.71, + "learning_rate": 4.632591180802218e-05, + "loss": 1.6445, + "step": 42910 + }, + { + "epoch": 3.71, + "learning_rate": 4.6325045482110375e-05, + "loss": 1.7699, + "step": 42920 + }, + { + "epoch": 3.72, + "learning_rate": 4.632417915619856e-05, + "loss": 1.654, + "step": 42930 + }, + { + "epoch": 3.72, + "learning_rate": 4.632331283028676e-05, + "loss": 1.6089, + "step": 42940 + }, + { + "epoch": 3.72, + "learning_rate": 4.632244650437495e-05, + "loss": 1.6511, + "step": 42950 + }, + { + "epoch": 3.72, + "learning_rate": 4.632158017846314e-05, + "loss": 1.6734, + "step": 42960 + }, + { + "epoch": 3.72, + "learning_rate": 4.6320713852551334e-05, + "loss": 1.7239, + "step": 42970 + }, + { + "epoch": 3.72, + "learning_rate": 4.631984752663953e-05, + "loss": 1.6421, + "step": 42980 + }, + { + "epoch": 3.72, + "learning_rate": 4.6318981200727716e-05, + "loss": 1.6537, + "step": 42990 + }, + { + "epoch": 3.72, + "learning_rate": 4.631811487481591e-05, + "loss": 1.6993, + "step": 43000 + }, + { + "epoch": 3.72, + "learning_rate": 4.6317248548904105e-05, + "loss": 1.6443, + "step": 43010 + }, + { + "epoch": 3.72, + "learning_rate": 4.631638222299229e-05, + "loss": 1.6221, + "step": 43020 + }, + { + "epoch": 3.72, + "learning_rate": 4.631551589708049e-05, + "loss": 1.6819, + "step": 43030 + }, + { + "epoch": 3.73, + "learning_rate": 4.6314649571168675e-05, + "loss": 1.6034, + "step": 43040 + }, + { + "epoch": 3.73, + "learning_rate": 4.631378324525687e-05, + "loss": 1.6573, + "step": 43050 + }, + { + "epoch": 3.73, + "learning_rate": 4.6312916919345064e-05, + "loss": 1.6979, + "step": 43060 + }, + { + "epoch": 3.73, + "learning_rate": 4.631205059343325e-05, + "loss": 1.6841, + "step": 43070 + }, + { + "epoch": 3.73, + "learning_rate": 4.6311184267521446e-05, + "loss": 1.7238, + "step": 43080 + }, + { + "epoch": 3.73, + "learning_rate": 4.631031794160964e-05, + "loss": 1.7255, + "step": 43090 + }, + { + "epoch": 3.73, + "learning_rate": 4.630945161569783e-05, + "loss": 1.6939, + "step": 43100 + }, + { + "epoch": 3.73, + "learning_rate": 4.630858528978602e-05, + "loss": 1.6863, + "step": 43110 + }, + { + "epoch": 3.73, + "learning_rate": 4.630771896387422e-05, + "loss": 1.6685, + "step": 43120 + }, + { + "epoch": 3.73, + "learning_rate": 4.6306852637962404e-05, + "loss": 1.6371, + "step": 43130 + }, + { + "epoch": 3.73, + "learning_rate": 4.63059863120506e-05, + "loss": 1.6028, + "step": 43140 + }, + { + "epoch": 3.73, + "learning_rate": 4.6305119986138787e-05, + "loss": 1.745, + "step": 43150 + }, + { + "epoch": 3.74, + "learning_rate": 4.630425366022698e-05, + "loss": 1.7228, + "step": 43160 + }, + { + "epoch": 3.74, + "learning_rate": 4.6303387334315175e-05, + "loss": 1.6705, + "step": 43170 + }, + { + "epoch": 3.74, + "learning_rate": 4.630252100840336e-05, + "loss": 1.6071, + "step": 43180 + }, + { + "epoch": 3.74, + "learning_rate": 4.630165468249156e-05, + "loss": 1.6804, + "step": 43190 + }, + { + "epoch": 3.74, + "learning_rate": 4.630078835657975e-05, + "loss": 1.6805, + "step": 43200 + }, + { + "epoch": 3.74, + "learning_rate": 4.629992203066794e-05, + "loss": 1.6896, + "step": 43210 + }, + { + "epoch": 3.74, + "learning_rate": 4.6299055704756134e-05, + "loss": 1.7458, + "step": 43220 + }, + { + "epoch": 3.74, + "learning_rate": 4.629818937884433e-05, + "loss": 1.6455, + "step": 43230 + }, + { + "epoch": 3.74, + "learning_rate": 4.6297323052932516e-05, + "loss": 1.6787, + "step": 43240 + }, + { + "epoch": 3.74, + "learning_rate": 4.629645672702071e-05, + "loss": 1.7373, + "step": 43250 + }, + { + "epoch": 3.74, + "learning_rate": 4.62955904011089e-05, + "loss": 1.6832, + "step": 43260 + }, + { + "epoch": 3.75, + "learning_rate": 4.629472407519709e-05, + "loss": 1.6524, + "step": 43270 + }, + { + "epoch": 3.75, + "learning_rate": 4.629385774928529e-05, + "loss": 1.6743, + "step": 43280 + }, + { + "epoch": 3.75, + "learning_rate": 4.6292991423373475e-05, + "loss": 1.6637, + "step": 43290 + }, + { + "epoch": 3.75, + "learning_rate": 4.629212509746167e-05, + "loss": 1.6631, + "step": 43300 + }, + { + "epoch": 3.75, + "learning_rate": 4.6291258771549864e-05, + "loss": 1.6559, + "step": 43310 + }, + { + "epoch": 3.75, + "learning_rate": 4.629039244563805e-05, + "loss": 1.6437, + "step": 43320 + }, + { + "epoch": 3.75, + "learning_rate": 4.6289526119726246e-05, + "loss": 1.6503, + "step": 43330 + }, + { + "epoch": 3.75, + "learning_rate": 4.628865979381444e-05, + "loss": 1.7466, + "step": 43340 + }, + { + "epoch": 3.75, + "learning_rate": 4.628779346790263e-05, + "loss": 1.6652, + "step": 43350 + }, + { + "epoch": 3.75, + "learning_rate": 4.628692714199082e-05, + "loss": 1.5839, + "step": 43360 + }, + { + "epoch": 3.75, + "learning_rate": 4.628606081607901e-05, + "loss": 1.6745, + "step": 43370 + }, + { + "epoch": 3.75, + "learning_rate": 4.6285194490167205e-05, + "loss": 1.7319, + "step": 43380 + }, + { + "epoch": 3.76, + "learning_rate": 4.62843281642554e-05, + "loss": 1.6066, + "step": 43390 + }, + { + "epoch": 3.76, + "learning_rate": 4.628346183834359e-05, + "loss": 1.6674, + "step": 43400 + }, + { + "epoch": 3.76, + "learning_rate": 4.628259551243178e-05, + "loss": 1.6465, + "step": 43410 + }, + { + "epoch": 3.76, + "learning_rate": 4.6281729186519976e-05, + "loss": 1.6771, + "step": 43420 + }, + { + "epoch": 3.76, + "learning_rate": 4.6280862860608163e-05, + "loss": 1.6754, + "step": 43430 + }, + { + "epoch": 3.76, + "learning_rate": 4.627999653469636e-05, + "loss": 1.6374, + "step": 43440 + }, + { + "epoch": 3.76, + "learning_rate": 4.627913020878455e-05, + "loss": 1.6536, + "step": 43450 + }, + { + "epoch": 3.76, + "learning_rate": 4.627826388287274e-05, + "loss": 1.6693, + "step": 43460 + }, + { + "epoch": 3.76, + "learning_rate": 4.6277397556960934e-05, + "loss": 1.7048, + "step": 43470 + }, + { + "epoch": 3.76, + "learning_rate": 4.627653123104912e-05, + "loss": 1.6604, + "step": 43480 + }, + { + "epoch": 3.76, + "learning_rate": 4.6275664905137317e-05, + "loss": 1.7004, + "step": 43490 + }, + { + "epoch": 3.77, + "learning_rate": 4.627479857922551e-05, + "loss": 1.7135, + "step": 43500 + }, + { + "epoch": 3.77, + "learning_rate": 4.62739322533137e-05, + "loss": 1.5975, + "step": 43510 + }, + { + "epoch": 3.77, + "learning_rate": 4.627306592740189e-05, + "loss": 1.6777, + "step": 43520 + }, + { + "epoch": 3.77, + "learning_rate": 4.627219960149009e-05, + "loss": 1.6841, + "step": 43530 + }, + { + "epoch": 3.77, + "learning_rate": 4.6271333275578275e-05, + "loss": 1.6955, + "step": 43540 + }, + { + "epoch": 3.77, + "learning_rate": 4.627046694966647e-05, + "loss": 1.7227, + "step": 43550 + }, + { + "epoch": 3.77, + "learning_rate": 4.6269600623754664e-05, + "loss": 1.6466, + "step": 43560 + }, + { + "epoch": 3.77, + "learning_rate": 4.626873429784285e-05, + "loss": 1.7154, + "step": 43570 + }, + { + "epoch": 3.77, + "learning_rate": 4.6267867971931046e-05, + "loss": 1.6567, + "step": 43580 + }, + { + "epoch": 3.77, + "learning_rate": 4.6267001646019234e-05, + "loss": 1.6865, + "step": 43590 + }, + { + "epoch": 3.77, + "learning_rate": 4.626613532010743e-05, + "loss": 1.6724, + "step": 43600 + }, + { + "epoch": 3.77, + "learning_rate": 4.626526899419562e-05, + "loss": 1.684, + "step": 43610 + }, + { + "epoch": 3.78, + "learning_rate": 4.626440266828381e-05, + "loss": 1.6952, + "step": 43620 + }, + { + "epoch": 3.78, + "learning_rate": 4.6263536342372005e-05, + "loss": 1.6822, + "step": 43630 + }, + { + "epoch": 3.78, + "learning_rate": 4.62626700164602e-05, + "loss": 1.6919, + "step": 43640 + }, + { + "epoch": 3.78, + "learning_rate": 4.626180369054839e-05, + "loss": 1.6771, + "step": 43650 + }, + { + "epoch": 3.78, + "learning_rate": 4.626093736463658e-05, + "loss": 1.6686, + "step": 43660 + }, + { + "epoch": 3.78, + "learning_rate": 4.626007103872477e-05, + "loss": 1.6406, + "step": 43670 + }, + { + "epoch": 3.78, + "learning_rate": 4.6259204712812964e-05, + "loss": 1.6078, + "step": 43680 + }, + { + "epoch": 3.78, + "learning_rate": 4.625833838690116e-05, + "loss": 1.6791, + "step": 43690 + }, + { + "epoch": 3.78, + "learning_rate": 4.6257472060989346e-05, + "loss": 1.6271, + "step": 43700 + }, + { + "epoch": 3.78, + "learning_rate": 4.625660573507754e-05, + "loss": 1.684, + "step": 43710 + }, + { + "epoch": 3.78, + "learning_rate": 4.6255739409165735e-05, + "loss": 1.6985, + "step": 43720 + }, + { + "epoch": 3.79, + "learning_rate": 4.625487308325392e-05, + "loss": 1.6547, + "step": 43730 + }, + { + "epoch": 3.79, + "learning_rate": 4.625400675734212e-05, + "loss": 1.6973, + "step": 43740 + }, + { + "epoch": 3.79, + "learning_rate": 4.625314043143031e-05, + "loss": 1.6969, + "step": 43750 + }, + { + "epoch": 3.79, + "learning_rate": 4.62522741055185e-05, + "loss": 1.7043, + "step": 43760 + }, + { + "epoch": 3.79, + "learning_rate": 4.6251407779606694e-05, + "loss": 1.6409, + "step": 43770 + }, + { + "epoch": 3.79, + "learning_rate": 4.625054145369488e-05, + "loss": 1.6619, + "step": 43780 + }, + { + "epoch": 3.79, + "learning_rate": 4.6249675127783076e-05, + "loss": 1.6223, + "step": 43790 + }, + { + "epoch": 3.79, + "learning_rate": 4.624880880187127e-05, + "loss": 1.6204, + "step": 43800 + }, + { + "epoch": 3.79, + "learning_rate": 4.624794247595946e-05, + "loss": 1.5914, + "step": 43810 + }, + { + "epoch": 3.79, + "learning_rate": 4.624707615004765e-05, + "loss": 1.6924, + "step": 43820 + }, + { + "epoch": 3.79, + "learning_rate": 4.624620982413585e-05, + "loss": 1.5907, + "step": 43830 + }, + { + "epoch": 3.79, + "learning_rate": 4.6245343498224034e-05, + "loss": 1.6596, + "step": 43840 + }, + { + "epoch": 3.8, + "learning_rate": 4.624447717231223e-05, + "loss": 1.6931, + "step": 43850 + }, + { + "epoch": 3.8, + "learning_rate": 4.624361084640042e-05, + "loss": 1.6326, + "step": 43860 + }, + { + "epoch": 3.8, + "learning_rate": 4.624274452048861e-05, + "loss": 1.6705, + "step": 43870 + }, + { + "epoch": 3.8, + "learning_rate": 4.6241878194576805e-05, + "loss": 1.642, + "step": 43880 + }, + { + "epoch": 3.8, + "learning_rate": 4.624101186866499e-05, + "loss": 1.6707, + "step": 43890 + }, + { + "epoch": 3.8, + "learning_rate": 4.624014554275319e-05, + "loss": 1.5931, + "step": 43900 + }, + { + "epoch": 3.8, + "learning_rate": 4.623927921684138e-05, + "loss": 1.6851, + "step": 43910 + }, + { + "epoch": 3.8, + "learning_rate": 4.623841289092957e-05, + "loss": 1.6779, + "step": 43920 + }, + { + "epoch": 3.8, + "learning_rate": 4.6237546565017764e-05, + "loss": 1.6518, + "step": 43930 + }, + { + "epoch": 3.8, + "learning_rate": 4.623668023910596e-05, + "loss": 1.7154, + "step": 43940 + }, + { + "epoch": 3.8, + "learning_rate": 4.6235813913194146e-05, + "loss": 1.633, + "step": 43950 + }, + { + "epoch": 3.8, + "learning_rate": 4.623494758728234e-05, + "loss": 1.6811, + "step": 43960 + }, + { + "epoch": 3.81, + "learning_rate": 4.6234081261370535e-05, + "loss": 1.6895, + "step": 43970 + }, + { + "epoch": 3.81, + "learning_rate": 4.623321493545872e-05, + "loss": 1.678, + "step": 43980 + }, + { + "epoch": 3.81, + "learning_rate": 4.623234860954692e-05, + "loss": 1.6565, + "step": 43990 + }, + { + "epoch": 3.81, + "learning_rate": 4.6231482283635105e-05, + "loss": 1.619, + "step": 44000 + }, + { + "epoch": 3.81, + "learning_rate": 4.62306159577233e-05, + "loss": 1.6515, + "step": 44010 + }, + { + "epoch": 3.81, + "learning_rate": 4.6229749631811494e-05, + "loss": 1.7012, + "step": 44020 + }, + { + "epoch": 3.81, + "learning_rate": 4.622888330589968e-05, + "loss": 1.7164, + "step": 44030 + }, + { + "epoch": 3.81, + "learning_rate": 4.6228016979987876e-05, + "loss": 1.7012, + "step": 44040 + }, + { + "epoch": 3.81, + "learning_rate": 4.622715065407607e-05, + "loss": 1.6628, + "step": 44050 + }, + { + "epoch": 3.81, + "learning_rate": 4.622628432816426e-05, + "loss": 1.6282, + "step": 44060 + }, + { + "epoch": 3.81, + "learning_rate": 4.622541800225245e-05, + "loss": 1.7004, + "step": 44070 + }, + { + "epoch": 3.82, + "learning_rate": 4.622455167634065e-05, + "loss": 1.6676, + "step": 44080 + }, + { + "epoch": 3.82, + "learning_rate": 4.6223685350428835e-05, + "loss": 1.642, + "step": 44090 + }, + { + "epoch": 3.82, + "learning_rate": 4.622281902451703e-05, + "loss": 1.6525, + "step": 44100 + }, + { + "epoch": 3.82, + "learning_rate": 4.622195269860522e-05, + "loss": 1.6976, + "step": 44110 + }, + { + "epoch": 3.82, + "learning_rate": 4.622108637269341e-05, + "loss": 1.6276, + "step": 44120 + }, + { + "epoch": 3.82, + "learning_rate": 4.6220220046781606e-05, + "loss": 1.6532, + "step": 44130 + }, + { + "epoch": 3.82, + "learning_rate": 4.6219353720869793e-05, + "loss": 1.6532, + "step": 44140 + }, + { + "epoch": 3.82, + "learning_rate": 4.621848739495799e-05, + "loss": 1.6604, + "step": 44150 + }, + { + "epoch": 3.82, + "learning_rate": 4.621762106904618e-05, + "loss": 1.6254, + "step": 44160 + }, + { + "epoch": 3.82, + "learning_rate": 4.621675474313437e-05, + "loss": 1.6495, + "step": 44170 + }, + { + "epoch": 3.82, + "learning_rate": 4.6215888417222564e-05, + "loss": 1.7184, + "step": 44180 + }, + { + "epoch": 3.82, + "learning_rate": 4.621502209131076e-05, + "loss": 1.6333, + "step": 44190 + }, + { + "epoch": 3.83, + "learning_rate": 4.6214155765398947e-05, + "loss": 1.6954, + "step": 44200 + }, + { + "epoch": 3.83, + "learning_rate": 4.621328943948714e-05, + "loss": 1.6775, + "step": 44210 + }, + { + "epoch": 3.83, + "learning_rate": 4.621242311357533e-05, + "loss": 1.6851, + "step": 44220 + }, + { + "epoch": 3.83, + "learning_rate": 4.621155678766352e-05, + "loss": 1.6052, + "step": 44230 + }, + { + "epoch": 3.83, + "learning_rate": 4.621069046175172e-05, + "loss": 1.652, + "step": 44240 + }, + { + "epoch": 3.83, + "learning_rate": 4.6209824135839905e-05, + "loss": 1.6622, + "step": 44250 + }, + { + "epoch": 3.83, + "learning_rate": 4.62089578099281e-05, + "loss": 1.5957, + "step": 44260 + }, + { + "epoch": 3.83, + "learning_rate": 4.6208091484016294e-05, + "loss": 1.6714, + "step": 44270 + }, + { + "epoch": 3.83, + "learning_rate": 4.620722515810448e-05, + "loss": 1.7122, + "step": 44280 + }, + { + "epoch": 3.83, + "learning_rate": 4.6206358832192676e-05, + "loss": 1.6359, + "step": 44290 + }, + { + "epoch": 3.83, + "learning_rate": 4.620549250628087e-05, + "loss": 1.7636, + "step": 44300 + }, + { + "epoch": 3.84, + "learning_rate": 4.620462618036906e-05, + "loss": 1.6845, + "step": 44310 + }, + { + "epoch": 3.84, + "learning_rate": 4.620375985445725e-05, + "loss": 1.6476, + "step": 44320 + }, + { + "epoch": 3.84, + "learning_rate": 4.620289352854544e-05, + "loss": 1.6046, + "step": 44330 + }, + { + "epoch": 3.84, + "learning_rate": 4.6202027202633635e-05, + "loss": 1.6449, + "step": 44340 + }, + { + "epoch": 3.84, + "learning_rate": 4.620116087672183e-05, + "loss": 1.726, + "step": 44350 + }, + { + "epoch": 3.84, + "learning_rate": 4.620029455081002e-05, + "loss": 1.7094, + "step": 44360 + }, + { + "epoch": 3.84, + "learning_rate": 4.619942822489821e-05, + "loss": 1.7109, + "step": 44370 + }, + { + "epoch": 3.84, + "learning_rate": 4.6198561898986406e-05, + "loss": 1.7083, + "step": 44380 + }, + { + "epoch": 3.84, + "learning_rate": 4.6197695573074594e-05, + "loss": 1.6948, + "step": 44390 + }, + { + "epoch": 3.84, + "learning_rate": 4.619682924716279e-05, + "loss": 1.6611, + "step": 44400 + }, + { + "epoch": 3.84, + "learning_rate": 4.6195962921250976e-05, + "loss": 1.6032, + "step": 44410 + }, + { + "epoch": 3.84, + "learning_rate": 4.619509659533917e-05, + "loss": 1.5748, + "step": 44420 + }, + { + "epoch": 3.85, + "learning_rate": 4.6194230269427365e-05, + "loss": 1.7095, + "step": 44430 + }, + { + "epoch": 3.85, + "learning_rate": 4.619336394351555e-05, + "loss": 1.6572, + "step": 44440 + }, + { + "epoch": 3.85, + "learning_rate": 4.619249761760375e-05, + "loss": 1.6977, + "step": 44450 + }, + { + "epoch": 3.85, + "learning_rate": 4.619163129169194e-05, + "loss": 1.7325, + "step": 44460 + }, + { + "epoch": 3.85, + "learning_rate": 4.619076496578013e-05, + "loss": 1.6736, + "step": 44470 + }, + { + "epoch": 3.85, + "learning_rate": 4.6189898639868323e-05, + "loss": 1.6517, + "step": 44480 + }, + { + "epoch": 3.85, + "learning_rate": 4.618903231395652e-05, + "loss": 1.669, + "step": 44490 + }, + { + "epoch": 3.85, + "learning_rate": 4.6188165988044706e-05, + "loss": 1.6203, + "step": 44500 + }, + { + "epoch": 3.85, + "learning_rate": 4.61872996621329e-05, + "loss": 1.6749, + "step": 44510 + }, + { + "epoch": 3.85, + "learning_rate": 4.618643333622109e-05, + "loss": 1.6, + "step": 44520 + }, + { + "epoch": 3.85, + "learning_rate": 4.618556701030928e-05, + "loss": 1.6192, + "step": 44530 + }, + { + "epoch": 3.86, + "learning_rate": 4.618470068439748e-05, + "loss": 1.5406, + "step": 44540 + }, + { + "epoch": 3.86, + "learning_rate": 4.6183834358485664e-05, + "loss": 1.6304, + "step": 44550 + }, + { + "epoch": 3.86, + "learning_rate": 4.618296803257386e-05, + "loss": 1.6611, + "step": 44560 + }, + { + "epoch": 3.86, + "learning_rate": 4.618210170666205e-05, + "loss": 1.6733, + "step": 44570 + }, + { + "epoch": 3.86, + "learning_rate": 4.618123538075024e-05, + "loss": 1.7126, + "step": 44580 + }, + { + "epoch": 3.86, + "learning_rate": 4.6180369054838435e-05, + "loss": 1.7227, + "step": 44590 + }, + { + "epoch": 3.86, + "learning_rate": 4.617950272892663e-05, + "loss": 1.6395, + "step": 44600 + }, + { + "epoch": 3.86, + "learning_rate": 4.617863640301482e-05, + "loss": 1.7129, + "step": 44610 + }, + { + "epoch": 3.86, + "learning_rate": 4.617777007710301e-05, + "loss": 1.6647, + "step": 44620 + }, + { + "epoch": 3.86, + "learning_rate": 4.61769037511912e-05, + "loss": 1.7583, + "step": 44630 + }, + { + "epoch": 3.86, + "learning_rate": 4.6176037425279394e-05, + "loss": 1.6575, + "step": 44640 + }, + { + "epoch": 3.86, + "learning_rate": 4.617517109936759e-05, + "loss": 1.6293, + "step": 44650 + }, + { + "epoch": 3.87, + "learning_rate": 4.6174304773455776e-05, + "loss": 1.726, + "step": 44660 + }, + { + "epoch": 3.87, + "learning_rate": 4.617343844754397e-05, + "loss": 1.6576, + "step": 44670 + }, + { + "epoch": 3.87, + "learning_rate": 4.6172572121632165e-05, + "loss": 1.6383, + "step": 44680 + }, + { + "epoch": 3.87, + "learning_rate": 4.617170579572035e-05, + "loss": 1.6685, + "step": 44690 + }, + { + "epoch": 3.87, + "learning_rate": 4.617083946980855e-05, + "loss": 1.6502, + "step": 44700 + }, + { + "epoch": 3.87, + "learning_rate": 4.616997314389674e-05, + "loss": 1.6538, + "step": 44710 + }, + { + "epoch": 3.87, + "learning_rate": 4.616910681798493e-05, + "loss": 1.7122, + "step": 44720 + }, + { + "epoch": 3.87, + "learning_rate": 4.6168240492073124e-05, + "loss": 1.635, + "step": 44730 + }, + { + "epoch": 3.87, + "learning_rate": 4.616737416616131e-05, + "loss": 1.6655, + "step": 44740 + }, + { + "epoch": 3.87, + "learning_rate": 4.6166507840249506e-05, + "loss": 1.7254, + "step": 44750 + }, + { + "epoch": 3.87, + "learning_rate": 4.61656415143377e-05, + "loss": 1.6435, + "step": 44760 + }, + { + "epoch": 3.88, + "learning_rate": 4.616477518842589e-05, + "loss": 1.7038, + "step": 44770 + }, + { + "epoch": 3.88, + "learning_rate": 4.616390886251408e-05, + "loss": 1.6156, + "step": 44780 + }, + { + "epoch": 3.88, + "learning_rate": 4.616304253660228e-05, + "loss": 1.7181, + "step": 44790 + }, + { + "epoch": 3.88, + "learning_rate": 4.6162176210690465e-05, + "loss": 1.6145, + "step": 44800 + }, + { + "epoch": 3.88, + "learning_rate": 4.616130988477866e-05, + "loss": 1.6498, + "step": 44810 + }, + { + "epoch": 3.88, + "learning_rate": 4.6160443558866854e-05, + "loss": 1.676, + "step": 44820 + }, + { + "epoch": 3.88, + "learning_rate": 4.615957723295504e-05, + "loss": 1.7423, + "step": 44830 + }, + { + "epoch": 3.88, + "learning_rate": 4.6158710907043236e-05, + "loss": 1.6798, + "step": 44840 + }, + { + "epoch": 3.88, + "learning_rate": 4.615784458113142e-05, + "loss": 1.7076, + "step": 44850 + }, + { + "epoch": 3.88, + "learning_rate": 4.615697825521962e-05, + "loss": 1.7568, + "step": 44860 + }, + { + "epoch": 3.88, + "learning_rate": 4.615611192930781e-05, + "loss": 1.6232, + "step": 44870 + }, + { + "epoch": 3.88, + "learning_rate": 4.6155245603396e-05, + "loss": 1.6288, + "step": 44880 + }, + { + "epoch": 3.89, + "learning_rate": 4.6154379277484194e-05, + "loss": 1.6786, + "step": 44890 + }, + { + "epoch": 3.89, + "learning_rate": 4.615351295157239e-05, + "loss": 1.6567, + "step": 44900 + }, + { + "epoch": 3.89, + "learning_rate": 4.6152646625660577e-05, + "loss": 1.6633, + "step": 44910 + }, + { + "epoch": 3.89, + "learning_rate": 4.615178029974877e-05, + "loss": 1.6808, + "step": 44920 + }, + { + "epoch": 3.89, + "learning_rate": 4.6150913973836965e-05, + "loss": 1.6403, + "step": 44930 + }, + { + "epoch": 3.89, + "learning_rate": 4.615004764792515e-05, + "loss": 1.7268, + "step": 44940 + }, + { + "epoch": 3.89, + "learning_rate": 4.614918132201335e-05, + "loss": 1.6549, + "step": 44950 + }, + { + "epoch": 3.89, + "learning_rate": 4.6148314996101535e-05, + "loss": 1.6058, + "step": 44960 + }, + { + "epoch": 3.89, + "learning_rate": 4.614744867018973e-05, + "loss": 1.6684, + "step": 44970 + }, + { + "epoch": 3.89, + "learning_rate": 4.6146582344277924e-05, + "loss": 1.6272, + "step": 44980 + }, + { + "epoch": 3.89, + "learning_rate": 4.614571601836611e-05, + "loss": 1.6456, + "step": 44990 + }, + { + "epoch": 3.89, + "learning_rate": 4.6144849692454306e-05, + "loss": 1.6837, + "step": 45000 + }, + { + "epoch": 3.9, + "learning_rate": 4.61439833665425e-05, + "loss": 1.628, + "step": 45010 + }, + { + "epoch": 3.9, + "learning_rate": 4.614311704063069e-05, + "loss": 1.7044, + "step": 45020 + }, + { + "epoch": 3.9, + "learning_rate": 4.614225071471888e-05, + "loss": 1.6831, + "step": 45030 + }, + { + "epoch": 3.9, + "learning_rate": 4.614138438880708e-05, + "loss": 1.7006, + "step": 45040 + }, + { + "epoch": 3.9, + "learning_rate": 4.6140518062895265e-05, + "loss": 1.6627, + "step": 45050 + }, + { + "epoch": 3.9, + "learning_rate": 4.613965173698346e-05, + "loss": 1.677, + "step": 45060 + }, + { + "epoch": 3.9, + "learning_rate": 4.613878541107165e-05, + "loss": 1.6158, + "step": 45070 + }, + { + "epoch": 3.9, + "learning_rate": 4.613791908515984e-05, + "loss": 1.7145, + "step": 45080 + }, + { + "epoch": 3.9, + "learning_rate": 4.6137052759248036e-05, + "loss": 1.5868, + "step": 45090 + }, + { + "epoch": 3.9, + "learning_rate": 4.6136186433336224e-05, + "loss": 1.6488, + "step": 45100 + }, + { + "epoch": 3.9, + "learning_rate": 4.613532010742442e-05, + "loss": 1.6947, + "step": 45110 + }, + { + "epoch": 3.91, + "learning_rate": 4.613445378151261e-05, + "loss": 1.6534, + "step": 45120 + }, + { + "epoch": 3.91, + "learning_rate": 4.61335874556008e-05, + "loss": 1.6704, + "step": 45130 + }, + { + "epoch": 3.91, + "learning_rate": 4.6132721129688995e-05, + "loss": 1.7401, + "step": 45140 + }, + { + "epoch": 3.91, + "learning_rate": 4.613185480377719e-05, + "loss": 1.7156, + "step": 45150 + }, + { + "epoch": 3.91, + "learning_rate": 4.613098847786538e-05, + "loss": 1.6357, + "step": 45160 + }, + { + "epoch": 3.91, + "learning_rate": 4.613012215195357e-05, + "loss": 1.6797, + "step": 45170 + }, + { + "epoch": 3.91, + "learning_rate": 4.612925582604176e-05, + "loss": 1.6366, + "step": 45180 + }, + { + "epoch": 3.91, + "learning_rate": 4.6128389500129953e-05, + "loss": 1.6986, + "step": 45190 + }, + { + "epoch": 3.91, + "learning_rate": 4.612752317421815e-05, + "loss": 1.6611, + "step": 45200 + }, + { + "epoch": 3.91, + "learning_rate": 4.6126656848306336e-05, + "loss": 1.7031, + "step": 45210 + }, + { + "epoch": 3.91, + "learning_rate": 4.612579052239453e-05, + "loss": 1.7194, + "step": 45220 + }, + { + "epoch": 3.91, + "learning_rate": 4.6124924196482724e-05, + "loss": 1.606, + "step": 45230 + }, + { + "epoch": 3.92, + "learning_rate": 4.612405787057091e-05, + "loss": 1.6974, + "step": 45240 + }, + { + "epoch": 3.92, + "learning_rate": 4.612319154465911e-05, + "loss": 1.7054, + "step": 45250 + }, + { + "epoch": 3.92, + "learning_rate": 4.6122325218747294e-05, + "loss": 1.6623, + "step": 45260 + }, + { + "epoch": 3.92, + "learning_rate": 4.612145889283549e-05, + "loss": 1.7032, + "step": 45270 + }, + { + "epoch": 3.92, + "learning_rate": 4.612059256692368e-05, + "loss": 1.6142, + "step": 45280 + }, + { + "epoch": 3.92, + "learning_rate": 4.611972624101187e-05, + "loss": 1.6876, + "step": 45290 + }, + { + "epoch": 3.92, + "learning_rate": 4.6118859915100065e-05, + "loss": 1.6281, + "step": 45300 + }, + { + "epoch": 3.92, + "learning_rate": 4.611799358918826e-05, + "loss": 1.6397, + "step": 45310 + }, + { + "epoch": 3.92, + "learning_rate": 4.611712726327645e-05, + "loss": 1.6185, + "step": 45320 + }, + { + "epoch": 3.92, + "learning_rate": 4.611626093736464e-05, + "loss": 1.6772, + "step": 45330 + }, + { + "epoch": 3.92, + "learning_rate": 4.6115394611452836e-05, + "loss": 1.6254, + "step": 45340 + }, + { + "epoch": 3.93, + "learning_rate": 4.6114528285541024e-05, + "loss": 1.6541, + "step": 45350 + }, + { + "epoch": 3.93, + "learning_rate": 4.611366195962922e-05, + "loss": 1.6268, + "step": 45360 + }, + { + "epoch": 3.93, + "learning_rate": 4.6112795633717406e-05, + "loss": 1.6174, + "step": 45370 + }, + { + "epoch": 3.93, + "learning_rate": 4.61119293078056e-05, + "loss": 1.7057, + "step": 45380 + }, + { + "epoch": 3.93, + "learning_rate": 4.6111062981893795e-05, + "loss": 1.7418, + "step": 45390 + }, + { + "epoch": 3.93, + "learning_rate": 4.611019665598198e-05, + "loss": 1.644, + "step": 45400 + }, + { + "epoch": 3.93, + "learning_rate": 4.610933033007018e-05, + "loss": 1.6947, + "step": 45410 + }, + { + "epoch": 3.93, + "learning_rate": 4.610846400415837e-05, + "loss": 1.6206, + "step": 45420 + }, + { + "epoch": 3.93, + "learning_rate": 4.610759767824656e-05, + "loss": 1.6813, + "step": 45430 + }, + { + "epoch": 3.93, + "learning_rate": 4.6106731352334754e-05, + "loss": 1.6097, + "step": 45440 + }, + { + "epoch": 3.93, + "learning_rate": 4.610586502642295e-05, + "loss": 1.6689, + "step": 45450 + }, + { + "epoch": 3.93, + "learning_rate": 4.6104998700511136e-05, + "loss": 1.6806, + "step": 45460 + }, + { + "epoch": 3.94, + "learning_rate": 4.610413237459933e-05, + "loss": 1.651, + "step": 45470 + }, + { + "epoch": 3.94, + "learning_rate": 4.610326604868752e-05, + "loss": 1.6613, + "step": 45480 + }, + { + "epoch": 3.94, + "learning_rate": 4.610239972277571e-05, + "loss": 1.6677, + "step": 45490 + }, + { + "epoch": 3.94, + "learning_rate": 4.610153339686391e-05, + "loss": 1.5957, + "step": 45500 + }, + { + "epoch": 3.94, + "learning_rate": 4.6100667070952095e-05, + "loss": 1.6939, + "step": 45510 + }, + { + "epoch": 3.94, + "learning_rate": 4.609980074504029e-05, + "loss": 1.6857, + "step": 45520 + }, + { + "epoch": 3.94, + "learning_rate": 4.6098934419128484e-05, + "loss": 1.6574, + "step": 45530 + }, + { + "epoch": 3.94, + "learning_rate": 4.609806809321667e-05, + "loss": 1.7037, + "step": 45540 + }, + { + "epoch": 3.94, + "learning_rate": 4.6097201767304866e-05, + "loss": 1.6412, + "step": 45550 + }, + { + "epoch": 3.94, + "learning_rate": 4.609633544139306e-05, + "loss": 1.7094, + "step": 45560 + }, + { + "epoch": 3.94, + "learning_rate": 4.609546911548125e-05, + "loss": 1.6687, + "step": 45570 + }, + { + "epoch": 3.95, + "learning_rate": 4.609460278956944e-05, + "loss": 1.648, + "step": 45580 + }, + { + "epoch": 3.95, + "learning_rate": 4.609373646365763e-05, + "loss": 1.7115, + "step": 45590 + }, + { + "epoch": 3.95, + "learning_rate": 4.609287013774582e-05, + "loss": 1.688, + "step": 45600 + }, + { + "epoch": 3.95, + "learning_rate": 4.609200381183401e-05, + "loss": 1.6218, + "step": 45610 + }, + { + "epoch": 3.95, + "learning_rate": 4.6091137485922207e-05, + "loss": 1.6761, + "step": 45620 + }, + { + "epoch": 3.95, + "learning_rate": 4.6090271160010394e-05, + "loss": 1.6092, + "step": 45630 + }, + { + "epoch": 3.95, + "learning_rate": 4.608940483409859e-05, + "loss": 1.7005, + "step": 45640 + }, + { + "epoch": 3.95, + "learning_rate": 4.6088538508186776e-05, + "loss": 1.6427, + "step": 45650 + }, + { + "epoch": 3.95, + "learning_rate": 4.608767218227497e-05, + "loss": 1.668, + "step": 45660 + }, + { + "epoch": 3.95, + "learning_rate": 4.6086805856363165e-05, + "loss": 1.5896, + "step": 45670 + }, + { + "epoch": 3.95, + "learning_rate": 4.608593953045135e-05, + "loss": 1.66, + "step": 45680 + }, + { + "epoch": 3.95, + "learning_rate": 4.608507320453955e-05, + "loss": 1.6872, + "step": 45690 + }, + { + "epoch": 3.96, + "learning_rate": 4.608420687862774e-05, + "loss": 1.6758, + "step": 45700 + }, + { + "epoch": 3.96, + "learning_rate": 4.608334055271593e-05, + "loss": 1.6359, + "step": 45710 + }, + { + "epoch": 3.96, + "learning_rate": 4.6082474226804124e-05, + "loss": 1.6363, + "step": 45720 + }, + { + "epoch": 3.96, + "learning_rate": 4.608160790089232e-05, + "loss": 1.6439, + "step": 45730 + }, + { + "epoch": 3.96, + "learning_rate": 4.6080741574980506e-05, + "loss": 1.6322, + "step": 45740 + }, + { + "epoch": 3.96, + "learning_rate": 4.60798752490687e-05, + "loss": 1.6268, + "step": 45750 + }, + { + "epoch": 3.96, + "learning_rate": 4.607900892315689e-05, + "loss": 1.6675, + "step": 45760 + }, + { + "epoch": 3.96, + "learning_rate": 4.607814259724508e-05, + "loss": 1.6341, + "step": 45770 + }, + { + "epoch": 3.96, + "learning_rate": 4.607727627133328e-05, + "loss": 1.7116, + "step": 45780 + }, + { + "epoch": 3.96, + "learning_rate": 4.6076409945421465e-05, + "loss": 1.6659, + "step": 45790 + }, + { + "epoch": 3.96, + "learning_rate": 4.607554361950966e-05, + "loss": 1.6848, + "step": 45800 + }, + { + "epoch": 3.97, + "learning_rate": 4.6074677293597854e-05, + "loss": 1.6197, + "step": 45810 + }, + { + "epoch": 3.97, + "learning_rate": 4.607381096768604e-05, + "loss": 1.6597, + "step": 45820 + }, + { + "epoch": 3.97, + "learning_rate": 4.6072944641774236e-05, + "loss": 1.6258, + "step": 45830 + }, + { + "epoch": 3.97, + "learning_rate": 4.607207831586243e-05, + "loss": 1.6278, + "step": 45840 + }, + { + "epoch": 3.97, + "learning_rate": 4.607121198995062e-05, + "loss": 1.6584, + "step": 45850 + }, + { + "epoch": 3.97, + "learning_rate": 4.607034566403881e-05, + "loss": 1.6485, + "step": 45860 + }, + { + "epoch": 3.97, + "learning_rate": 4.6069479338127e-05, + "loss": 1.6586, + "step": 45870 + }, + { + "epoch": 3.97, + "learning_rate": 4.6068613012215195e-05, + "loss": 1.6282, + "step": 45880 + }, + { + "epoch": 3.97, + "learning_rate": 4.606774668630339e-05, + "loss": 1.6329, + "step": 45890 + }, + { + "epoch": 3.97, + "learning_rate": 4.606688036039158e-05, + "loss": 1.6605, + "step": 45900 + }, + { + "epoch": 3.97, + "learning_rate": 4.606601403447977e-05, + "loss": 1.7024, + "step": 45910 + }, + { + "epoch": 3.97, + "learning_rate": 4.6065147708567966e-05, + "loss": 1.6581, + "step": 45920 + }, + { + "epoch": 3.98, + "learning_rate": 4.606428138265615e-05, + "loss": 1.7258, + "step": 45930 + }, + { + "epoch": 3.98, + "learning_rate": 4.606341505674435e-05, + "loss": 1.6493, + "step": 45940 + }, + { + "epoch": 3.98, + "learning_rate": 4.6062548730832535e-05, + "loss": 1.6837, + "step": 45950 + }, + { + "epoch": 3.98, + "learning_rate": 4.606168240492073e-05, + "loss": 1.6805, + "step": 45960 + }, + { + "epoch": 3.98, + "learning_rate": 4.6060816079008924e-05, + "loss": 1.6497, + "step": 45970 + }, + { + "epoch": 3.98, + "learning_rate": 4.605994975309711e-05, + "loss": 1.6834, + "step": 45980 + }, + { + "epoch": 3.98, + "learning_rate": 4.6059083427185306e-05, + "loss": 1.6278, + "step": 45990 + }, + { + "epoch": 3.98, + "learning_rate": 4.60582171012735e-05, + "loss": 1.6207, + "step": 46000 + }, + { + "epoch": 3.98, + "learning_rate": 4.605735077536169e-05, + "loss": 1.6562, + "step": 46010 + }, + { + "epoch": 3.98, + "learning_rate": 4.605648444944988e-05, + "loss": 1.6339, + "step": 46020 + }, + { + "epoch": 3.98, + "learning_rate": 4.605561812353808e-05, + "loss": 1.641, + "step": 46030 + }, + { + "epoch": 3.98, + "learning_rate": 4.6054751797626265e-05, + "loss": 1.7175, + "step": 46040 + }, + { + "epoch": 3.99, + "learning_rate": 4.605388547171446e-05, + "loss": 1.7388, + "step": 46050 + }, + { + "epoch": 3.99, + "learning_rate": 4.605301914580265e-05, + "loss": 1.6448, + "step": 46060 + }, + { + "epoch": 3.99, + "learning_rate": 4.605215281989084e-05, + "loss": 1.6708, + "step": 46070 + }, + { + "epoch": 3.99, + "learning_rate": 4.6051286493979036e-05, + "loss": 1.6662, + "step": 46080 + }, + { + "epoch": 3.99, + "learning_rate": 4.6050420168067224e-05, + "loss": 1.7313, + "step": 46090 + }, + { + "epoch": 3.99, + "learning_rate": 4.604955384215542e-05, + "loss": 1.6376, + "step": 46100 + }, + { + "epoch": 3.99, + "learning_rate": 4.604868751624361e-05, + "loss": 1.587, + "step": 46110 + }, + { + "epoch": 3.99, + "learning_rate": 4.60478211903318e-05, + "loss": 1.72, + "step": 46120 + }, + { + "epoch": 3.99, + "learning_rate": 4.6046954864419995e-05, + "loss": 1.6294, + "step": 46130 + }, + { + "epoch": 3.99, + "learning_rate": 4.604608853850819e-05, + "loss": 1.6469, + "step": 46140 + }, + { + "epoch": 3.99, + "learning_rate": 4.604522221259638e-05, + "loss": 1.5715, + "step": 46150 + }, + { + "epoch": 4.0, + "learning_rate": 4.604435588668457e-05, + "loss": 1.6083, + "step": 46160 + }, + { + "epoch": 4.0, + "learning_rate": 4.604348956077276e-05, + "loss": 1.6373, + "step": 46170 + }, + { + "epoch": 4.0, + "learning_rate": 4.6042623234860954e-05, + "loss": 1.6462, + "step": 46180 + }, + { + "epoch": 4.0, + "learning_rate": 4.604175690894915e-05, + "loss": 1.6499, + "step": 46190 + }, + { + "epoch": 4.0, + "learning_rate": 4.6040890583037336e-05, + "loss": 1.6362, + "step": 46200 + }, + { + "epoch": 4.0, + "learning_rate": 4.604002425712553e-05, + "loss": 1.5959, + "step": 46210 + }, + { + "epoch": 4.0, + "eval_Bleu_1": 0.030889443774753944, + "eval_Bleu_2": 2.7341329729647925e-11, + "eval_Bleu_3": 2.7438870414530536e-14, + "eval_Bleu_4": 9.03141447600487e-16, + "eval_ROUGE_L": 0.06838356651444995, + "eval_cer": 0.995642939895962, + "eval_em": 0, + "eval_f1": 0.04987044066817675, + "eval_loss": 1.5620030164718628, + "eval_runtime": 1689.9924, + "eval_samples_per_second": 3.038, + "eval_steps_per_second": 3.038, + "eval_wer": 0.9859573442456251, + "step": 46214 + }, + { + "epoch": 4.0, + "learning_rate": 4.6039157931213725e-05, + "loss": 1.5924, + "step": 46220 + }, + { + "epoch": 4.0, + "learning_rate": 4.603829160530191e-05, + "loss": 1.593, + "step": 46230 + }, + { + "epoch": 4.0, + "learning_rate": 4.603742527939011e-05, + "loss": 1.6403, + "step": 46240 + }, + { + "epoch": 4.0, + "learning_rate": 4.60365589534783e-05, + "loss": 1.6703, + "step": 46250 + }, + { + "epoch": 4.0, + "learning_rate": 4.603569262756649e-05, + "loss": 1.6185, + "step": 46260 + }, + { + "epoch": 4.0, + "learning_rate": 4.603482630165468e-05, + "loss": 1.646, + "step": 46270 + }, + { + "epoch": 4.01, + "learning_rate": 4.603395997574287e-05, + "loss": 1.6246, + "step": 46280 + }, + { + "epoch": 4.01, + "learning_rate": 4.6033093649831065e-05, + "loss": 1.659, + "step": 46290 + }, + { + "epoch": 4.01, + "learning_rate": 4.603222732391926e-05, + "loss": 1.6526, + "step": 46300 + }, + { + "epoch": 4.01, + "learning_rate": 4.603136099800745e-05, + "loss": 1.6955, + "step": 46310 + }, + { + "epoch": 4.01, + "learning_rate": 4.603049467209564e-05, + "loss": 1.6348, + "step": 46320 + }, + { + "epoch": 4.01, + "learning_rate": 4.6029628346183836e-05, + "loss": 1.6516, + "step": 46330 + }, + { + "epoch": 4.01, + "learning_rate": 4.6028762020272024e-05, + "loss": 1.6177, + "step": 46340 + }, + { + "epoch": 4.01, + "learning_rate": 4.602789569436022e-05, + "loss": 1.6432, + "step": 46350 + }, + { + "epoch": 4.01, + "learning_rate": 4.602702936844841e-05, + "loss": 1.6726, + "step": 46360 + }, + { + "epoch": 4.01, + "learning_rate": 4.60261630425366e-05, + "loss": 1.6734, + "step": 46370 + }, + { + "epoch": 4.01, + "learning_rate": 4.6025296716624795e-05, + "loss": 1.5634, + "step": 46380 + }, + { + "epoch": 4.02, + "learning_rate": 4.602443039071298e-05, + "loss": 1.599, + "step": 46390 + }, + { + "epoch": 4.02, + "learning_rate": 4.602356406480118e-05, + "loss": 1.6562, + "step": 46400 + }, + { + "epoch": 4.02, + "learning_rate": 4.602269773888937e-05, + "loss": 1.6385, + "step": 46410 + }, + { + "epoch": 4.02, + "learning_rate": 4.602183141297756e-05, + "loss": 1.644, + "step": 46420 + }, + { + "epoch": 4.02, + "learning_rate": 4.6020965087065754e-05, + "loss": 1.6158, + "step": 46430 + }, + { + "epoch": 4.02, + "learning_rate": 4.602009876115395e-05, + "loss": 1.6262, + "step": 46440 + }, + { + "epoch": 4.02, + "learning_rate": 4.6019232435242136e-05, + "loss": 1.6225, + "step": 46450 + }, + { + "epoch": 4.02, + "learning_rate": 4.601836610933033e-05, + "loss": 1.5937, + "step": 46460 + }, + { + "epoch": 4.02, + "learning_rate": 4.6017499783418525e-05, + "loss": 1.6664, + "step": 46470 + }, + { + "epoch": 4.02, + "learning_rate": 4.601663345750671e-05, + "loss": 1.6632, + "step": 46480 + }, + { + "epoch": 4.02, + "learning_rate": 4.601576713159491e-05, + "loss": 1.607, + "step": 46490 + }, + { + "epoch": 4.02, + "learning_rate": 4.6014900805683095e-05, + "loss": 1.6376, + "step": 46500 + }, + { + "epoch": 4.03, + "learning_rate": 4.601403447977129e-05, + "loss": 1.6941, + "step": 46510 + }, + { + "epoch": 4.03, + "learning_rate": 4.6013168153859484e-05, + "loss": 1.6718, + "step": 46520 + }, + { + "epoch": 4.03, + "learning_rate": 4.601230182794767e-05, + "loss": 1.646, + "step": 46530 + }, + { + "epoch": 4.03, + "learning_rate": 4.6011435502035866e-05, + "loss": 1.6501, + "step": 46540 + }, + { + "epoch": 4.03, + "learning_rate": 4.601056917612406e-05, + "loss": 1.6421, + "step": 46550 + }, + { + "epoch": 4.03, + "learning_rate": 4.600970285021225e-05, + "loss": 1.6136, + "step": 46560 + }, + { + "epoch": 4.03, + "learning_rate": 4.600883652430044e-05, + "loss": 1.6045, + "step": 46570 + }, + { + "epoch": 4.03, + "learning_rate": 4.600797019838864e-05, + "loss": 1.6888, + "step": 46580 + }, + { + "epoch": 4.03, + "learning_rate": 4.6007103872476824e-05, + "loss": 1.6091, + "step": 46590 + }, + { + "epoch": 4.03, + "learning_rate": 4.600623754656502e-05, + "loss": 1.5829, + "step": 46600 + }, + { + "epoch": 4.03, + "learning_rate": 4.600537122065321e-05, + "loss": 1.5835, + "step": 46610 + }, + { + "epoch": 4.04, + "learning_rate": 4.60045048947414e-05, + "loss": 1.637, + "step": 46620 + }, + { + "epoch": 4.04, + "learning_rate": 4.6003638568829596e-05, + "loss": 1.695, + "step": 46630 + }, + { + "epoch": 4.04, + "learning_rate": 4.600277224291778e-05, + "loss": 1.6278, + "step": 46640 + }, + { + "epoch": 4.04, + "learning_rate": 4.600190591700598e-05, + "loss": 1.5937, + "step": 46650 + }, + { + "epoch": 4.04, + "learning_rate": 4.600103959109417e-05, + "loss": 1.5256, + "step": 46660 + }, + { + "epoch": 4.04, + "learning_rate": 4.600017326518236e-05, + "loss": 1.62, + "step": 46670 + }, + { + "epoch": 4.04, + "learning_rate": 4.5999306939270554e-05, + "loss": 1.6217, + "step": 46680 + }, + { + "epoch": 4.04, + "learning_rate": 4.599844061335875e-05, + "loss": 1.5856, + "step": 46690 + }, + { + "epoch": 4.04, + "learning_rate": 4.5997574287446936e-05, + "loss": 1.6574, + "step": 46700 + }, + { + "epoch": 4.04, + "learning_rate": 4.599670796153513e-05, + "loss": 1.5948, + "step": 46710 + }, + { + "epoch": 4.04, + "learning_rate": 4.599584163562332e-05, + "loss": 1.6362, + "step": 46720 + }, + { + "epoch": 4.04, + "learning_rate": 4.599497530971151e-05, + "loss": 1.645, + "step": 46730 + }, + { + "epoch": 4.05, + "learning_rate": 4.599410898379971e-05, + "loss": 1.658, + "step": 46740 + }, + { + "epoch": 4.05, + "learning_rate": 4.5993242657887895e-05, + "loss": 1.6146, + "step": 46750 + }, + { + "epoch": 4.05, + "learning_rate": 4.599237633197609e-05, + "loss": 1.6551, + "step": 46760 + }, + { + "epoch": 4.05, + "learning_rate": 4.5991510006064284e-05, + "loss": 1.6255, + "step": 46770 + }, + { + "epoch": 4.05, + "learning_rate": 4.599064368015247e-05, + "loss": 1.5892, + "step": 46780 + }, + { + "epoch": 4.05, + "learning_rate": 4.5989777354240666e-05, + "loss": 1.6948, + "step": 46790 + }, + { + "epoch": 4.05, + "learning_rate": 4.5988911028328854e-05, + "loss": 1.666, + "step": 46800 + }, + { + "epoch": 4.05, + "learning_rate": 4.598804470241705e-05, + "loss": 1.5978, + "step": 46810 + }, + { + "epoch": 4.05, + "learning_rate": 4.598717837650524e-05, + "loss": 1.6343, + "step": 46820 + }, + { + "epoch": 4.05, + "learning_rate": 4.598631205059343e-05, + "loss": 1.6279, + "step": 46830 + }, + { + "epoch": 4.05, + "learning_rate": 4.5985445724681625e-05, + "loss": 1.5273, + "step": 46840 + }, + { + "epoch": 4.06, + "learning_rate": 4.598457939876982e-05, + "loss": 1.6973, + "step": 46850 + }, + { + "epoch": 4.06, + "learning_rate": 4.598371307285801e-05, + "loss": 1.69, + "step": 46860 + }, + { + "epoch": 4.06, + "learning_rate": 4.59828467469462e-05, + "loss": 1.6576, + "step": 46870 + }, + { + "epoch": 4.06, + "learning_rate": 4.5981980421034396e-05, + "loss": 1.6143, + "step": 46880 + }, + { + "epoch": 4.06, + "learning_rate": 4.5981114095122584e-05, + "loss": 1.6699, + "step": 46890 + }, + { + "epoch": 4.06, + "learning_rate": 4.598024776921078e-05, + "loss": 1.6238, + "step": 46900 + }, + { + "epoch": 4.06, + "learning_rate": 4.5979381443298966e-05, + "loss": 1.6351, + "step": 46910 + }, + { + "epoch": 4.06, + "learning_rate": 4.597851511738716e-05, + "loss": 1.6368, + "step": 46920 + }, + { + "epoch": 4.06, + "learning_rate": 4.5977648791475355e-05, + "loss": 1.6423, + "step": 46930 + }, + { + "epoch": 4.06, + "learning_rate": 4.597678246556354e-05, + "loss": 1.6101, + "step": 46940 + }, + { + "epoch": 4.06, + "learning_rate": 4.597591613965174e-05, + "loss": 1.6364, + "step": 46950 + }, + { + "epoch": 4.06, + "learning_rate": 4.597504981373993e-05, + "loss": 1.6079, + "step": 46960 + }, + { + "epoch": 4.07, + "learning_rate": 4.597418348782812e-05, + "loss": 1.5641, + "step": 46970 + }, + { + "epoch": 4.07, + "learning_rate": 4.597331716191631e-05, + "loss": 1.6948, + "step": 46980 + }, + { + "epoch": 4.07, + "learning_rate": 4.597245083600451e-05, + "loss": 1.6592, + "step": 46990 + }, + { + "epoch": 4.07, + "learning_rate": 4.5971584510092695e-05, + "loss": 1.635, + "step": 47000 + }, + { + "epoch": 4.07, + "learning_rate": 4.597071818418089e-05, + "loss": 1.6275, + "step": 47010 + }, + { + "epoch": 4.07, + "learning_rate": 4.596985185826908e-05, + "loss": 1.6664, + "step": 47020 + }, + { + "epoch": 4.07, + "learning_rate": 4.596898553235727e-05, + "loss": 1.5605, + "step": 47030 + }, + { + "epoch": 4.07, + "learning_rate": 4.5968119206445466e-05, + "loss": 1.6369, + "step": 47040 + }, + { + "epoch": 4.07, + "learning_rate": 4.5967252880533654e-05, + "loss": 1.6587, + "step": 47050 + }, + { + "epoch": 4.07, + "learning_rate": 4.596638655462185e-05, + "loss": 1.6126, + "step": 47060 + }, + { + "epoch": 4.07, + "learning_rate": 4.596552022871004e-05, + "loss": 1.6349, + "step": 47070 + }, + { + "epoch": 4.07, + "learning_rate": 4.596465390279823e-05, + "loss": 1.6163, + "step": 47080 + }, + { + "epoch": 4.08, + "learning_rate": 4.5963787576886425e-05, + "loss": 1.634, + "step": 47090 + }, + { + "epoch": 4.08, + "learning_rate": 4.596292125097462e-05, + "loss": 1.6098, + "step": 47100 + }, + { + "epoch": 4.08, + "learning_rate": 4.596205492506281e-05, + "loss": 1.6516, + "step": 47110 + }, + { + "epoch": 4.08, + "learning_rate": 4.5961188599151e-05, + "loss": 1.6157, + "step": 47120 + }, + { + "epoch": 4.08, + "learning_rate": 4.596032227323919e-05, + "loss": 1.6294, + "step": 47130 + }, + { + "epoch": 4.08, + "learning_rate": 4.5959455947327384e-05, + "loss": 1.6139, + "step": 47140 + }, + { + "epoch": 4.08, + "learning_rate": 4.595858962141558e-05, + "loss": 1.6113, + "step": 47150 + }, + { + "epoch": 4.08, + "learning_rate": 4.5957723295503766e-05, + "loss": 1.6932, + "step": 47160 + }, + { + "epoch": 4.08, + "learning_rate": 4.595685696959196e-05, + "loss": 1.5566, + "step": 47170 + }, + { + "epoch": 4.08, + "learning_rate": 4.5955990643680155e-05, + "loss": 1.624, + "step": 47180 + }, + { + "epoch": 4.08, + "learning_rate": 4.595512431776834e-05, + "loss": 1.6194, + "step": 47190 + }, + { + "epoch": 4.09, + "learning_rate": 4.595425799185654e-05, + "loss": 1.5984, + "step": 47200 + }, + { + "epoch": 4.09, + "learning_rate": 4.595339166594473e-05, + "loss": 1.6333, + "step": 47210 + }, + { + "epoch": 4.09, + "learning_rate": 4.595252534003292e-05, + "loss": 1.5886, + "step": 47220 + }, + { + "epoch": 4.09, + "learning_rate": 4.5951659014121114e-05, + "loss": 1.6075, + "step": 47230 + }, + { + "epoch": 4.09, + "learning_rate": 4.59507926882093e-05, + "loss": 1.6557, + "step": 47240 + }, + { + "epoch": 4.09, + "learning_rate": 4.5949926362297496e-05, + "loss": 1.6187, + "step": 47250 + }, + { + "epoch": 4.09, + "learning_rate": 4.594906003638569e-05, + "loss": 1.6191, + "step": 47260 + }, + { + "epoch": 4.09, + "learning_rate": 4.594819371047388e-05, + "loss": 1.5553, + "step": 47270 + }, + { + "epoch": 4.09, + "learning_rate": 4.594732738456207e-05, + "loss": 1.582, + "step": 47280 + }, + { + "epoch": 4.09, + "learning_rate": 4.594646105865027e-05, + "loss": 1.5726, + "step": 47290 + }, + { + "epoch": 4.09, + "learning_rate": 4.5945594732738454e-05, + "loss": 1.6286, + "step": 47300 + }, + { + "epoch": 4.09, + "learning_rate": 4.594472840682665e-05, + "loss": 1.6027, + "step": 47310 + }, + { + "epoch": 4.1, + "learning_rate": 4.594386208091484e-05, + "loss": 1.5878, + "step": 47320 + }, + { + "epoch": 4.1, + "learning_rate": 4.594299575500303e-05, + "loss": 1.6748, + "step": 47330 + }, + { + "epoch": 4.1, + "learning_rate": 4.5942129429091226e-05, + "loss": 1.6068, + "step": 47340 + }, + { + "epoch": 4.1, + "learning_rate": 4.594126310317941e-05, + "loss": 1.6151, + "step": 47350 + }, + { + "epoch": 4.1, + "learning_rate": 4.594039677726761e-05, + "loss": 1.6231, + "step": 47360 + }, + { + "epoch": 4.1, + "learning_rate": 4.59395304513558e-05, + "loss": 1.6125, + "step": 47370 + }, + { + "epoch": 4.1, + "learning_rate": 4.593866412544399e-05, + "loss": 1.6055, + "step": 47380 + }, + { + "epoch": 4.1, + "learning_rate": 4.5937797799532184e-05, + "loss": 1.6266, + "step": 47390 + }, + { + "epoch": 4.1, + "learning_rate": 4.593693147362038e-05, + "loss": 1.6711, + "step": 47400 + }, + { + "epoch": 4.1, + "learning_rate": 4.5936065147708566e-05, + "loss": 1.7121, + "step": 47410 + }, + { + "epoch": 4.1, + "learning_rate": 4.593519882179676e-05, + "loss": 1.7072, + "step": 47420 + }, + { + "epoch": 4.11, + "learning_rate": 4.5934332495884955e-05, + "loss": 1.6655, + "step": 47430 + }, + { + "epoch": 4.11, + "learning_rate": 4.593346616997314e-05, + "loss": 1.693, + "step": 47440 + }, + { + "epoch": 4.11, + "learning_rate": 4.593259984406134e-05, + "loss": 1.6442, + "step": 47450 + }, + { + "epoch": 4.11, + "learning_rate": 4.5931733518149525e-05, + "loss": 1.6125, + "step": 47460 + }, + { + "epoch": 4.11, + "learning_rate": 4.593086719223772e-05, + "loss": 1.583, + "step": 47470 + }, + { + "epoch": 4.11, + "learning_rate": 4.5930000866325914e-05, + "loss": 1.5854, + "step": 47480 + }, + { + "epoch": 4.11, + "learning_rate": 4.59291345404141e-05, + "loss": 1.6407, + "step": 47490 + }, + { + "epoch": 4.11, + "learning_rate": 4.5928268214502296e-05, + "loss": 1.7023, + "step": 47500 + }, + { + "epoch": 4.11, + "learning_rate": 4.592740188859049e-05, + "loss": 1.7215, + "step": 47510 + }, + { + "epoch": 4.11, + "learning_rate": 4.592653556267868e-05, + "loss": 1.6265, + "step": 47520 + }, + { + "epoch": 4.11, + "learning_rate": 4.592566923676687e-05, + "loss": 1.5621, + "step": 47530 + }, + { + "epoch": 4.11, + "learning_rate": 4.592480291085506e-05, + "loss": 1.6275, + "step": 47540 + }, + { + "epoch": 4.12, + "learning_rate": 4.5923936584943255e-05, + "loss": 1.633, + "step": 47550 + }, + { + "epoch": 4.12, + "learning_rate": 4.592307025903145e-05, + "loss": 1.6267, + "step": 47560 + }, + { + "epoch": 4.12, + "learning_rate": 4.592220393311964e-05, + "loss": 1.5673, + "step": 47570 + }, + { + "epoch": 4.12, + "learning_rate": 4.592133760720783e-05, + "loss": 1.5892, + "step": 47580 + }, + { + "epoch": 4.12, + "learning_rate": 4.5920471281296026e-05, + "loss": 1.6246, + "step": 47590 + }, + { + "epoch": 4.12, + "learning_rate": 4.5919604955384214e-05, + "loss": 1.6213, + "step": 47600 + }, + { + "epoch": 4.12, + "learning_rate": 4.591873862947241e-05, + "loss": 1.678, + "step": 47610 + }, + { + "epoch": 4.12, + "learning_rate": 4.59178723035606e-05, + "loss": 1.5358, + "step": 47620 + }, + { + "epoch": 4.12, + "learning_rate": 4.591700597764879e-05, + "loss": 1.6175, + "step": 47630 + }, + { + "epoch": 4.12, + "learning_rate": 4.5916139651736985e-05, + "loss": 1.7044, + "step": 47640 + }, + { + "epoch": 4.12, + "learning_rate": 4.591527332582517e-05, + "loss": 1.6236, + "step": 47650 + }, + { + "epoch": 4.13, + "learning_rate": 4.591440699991337e-05, + "loss": 1.6489, + "step": 47660 + }, + { + "epoch": 4.13, + "learning_rate": 4.591354067400156e-05, + "loss": 1.5698, + "step": 47670 + }, + { + "epoch": 4.13, + "learning_rate": 4.591267434808975e-05, + "loss": 1.6106, + "step": 47680 + }, + { + "epoch": 4.13, + "learning_rate": 4.591180802217794e-05, + "loss": 1.66, + "step": 47690 + }, + { + "epoch": 4.13, + "learning_rate": 4.591094169626614e-05, + "loss": 1.6262, + "step": 47700 + }, + { + "epoch": 4.13, + "learning_rate": 4.5910075370354325e-05, + "loss": 1.6785, + "step": 47710 + }, + { + "epoch": 4.13, + "learning_rate": 4.590920904444252e-05, + "loss": 1.6136, + "step": 47720 + }, + { + "epoch": 4.13, + "learning_rate": 4.5908342718530714e-05, + "loss": 1.6279, + "step": 47730 + }, + { + "epoch": 4.13, + "learning_rate": 4.59074763926189e-05, + "loss": 1.595, + "step": 47740 + }, + { + "epoch": 4.13, + "learning_rate": 4.5906610066707096e-05, + "loss": 1.6449, + "step": 47750 + }, + { + "epoch": 4.13, + "learning_rate": 4.5905743740795284e-05, + "loss": 1.6444, + "step": 47760 + }, + { + "epoch": 4.13, + "learning_rate": 4.590487741488348e-05, + "loss": 1.6828, + "step": 47770 + }, + { + "epoch": 4.14, + "learning_rate": 4.590401108897167e-05, + "loss": 1.6771, + "step": 47780 + }, + { + "epoch": 4.14, + "learning_rate": 4.590314476305986e-05, + "loss": 1.5905, + "step": 47790 + }, + { + "epoch": 4.14, + "learning_rate": 4.5902278437148055e-05, + "loss": 1.6389, + "step": 47800 + }, + { + "epoch": 4.14, + "learning_rate": 4.590141211123625e-05, + "loss": 1.6619, + "step": 47810 + }, + { + "epoch": 4.14, + "learning_rate": 4.590054578532444e-05, + "loss": 1.6178, + "step": 47820 + }, + { + "epoch": 4.14, + "learning_rate": 4.589967945941263e-05, + "loss": 1.703, + "step": 47830 + }, + { + "epoch": 4.14, + "learning_rate": 4.5898813133500826e-05, + "loss": 1.5999, + "step": 47840 + }, + { + "epoch": 4.14, + "learning_rate": 4.5897946807589014e-05, + "loss": 1.6209, + "step": 47850 + }, + { + "epoch": 4.14, + "learning_rate": 4.589708048167721e-05, + "loss": 1.6507, + "step": 47860 + }, + { + "epoch": 4.14, + "learning_rate": 4.5896214155765396e-05, + "loss": 1.5733, + "step": 47870 + }, + { + "epoch": 4.14, + "learning_rate": 4.589534782985359e-05, + "loss": 1.5855, + "step": 47880 + }, + { + "epoch": 4.15, + "learning_rate": 4.5894481503941785e-05, + "loss": 1.6037, + "step": 47890 + }, + { + "epoch": 4.15, + "learning_rate": 4.589361517802997e-05, + "loss": 1.6225, + "step": 47900 + }, + { + "epoch": 4.15, + "learning_rate": 4.589274885211817e-05, + "loss": 1.6511, + "step": 47910 + }, + { + "epoch": 4.15, + "learning_rate": 4.589188252620636e-05, + "loss": 1.5512, + "step": 47920 + }, + { + "epoch": 4.15, + "learning_rate": 4.589101620029455e-05, + "loss": 1.6139, + "step": 47930 + }, + { + "epoch": 4.15, + "learning_rate": 4.5890149874382744e-05, + "loss": 1.7047, + "step": 47940 + }, + { + "epoch": 4.15, + "learning_rate": 4.588928354847094e-05, + "loss": 1.709, + "step": 47950 + }, + { + "epoch": 4.15, + "learning_rate": 4.5888417222559126e-05, + "loss": 1.5918, + "step": 47960 + }, + { + "epoch": 4.15, + "learning_rate": 4.588755089664732e-05, + "loss": 1.627, + "step": 47970 + }, + { + "epoch": 4.15, + "learning_rate": 4.588668457073551e-05, + "loss": 1.6067, + "step": 47980 + }, + { + "epoch": 4.15, + "learning_rate": 4.58858182448237e-05, + "loss": 1.5814, + "step": 47990 + }, + { + "epoch": 4.15, + "learning_rate": 4.58849519189119e-05, + "loss": 1.686, + "step": 48000 + }, + { + "epoch": 4.16, + "learning_rate": 4.5884085593000084e-05, + "loss": 1.6873, + "step": 48010 + }, + { + "epoch": 4.16, + "learning_rate": 4.588321926708828e-05, + "loss": 1.5802, + "step": 48020 + }, + { + "epoch": 4.16, + "learning_rate": 4.588235294117647e-05, + "loss": 1.6599, + "step": 48030 + }, + { + "epoch": 4.16, + "learning_rate": 4.588148661526466e-05, + "loss": 1.6519, + "step": 48040 + }, + { + "epoch": 4.16, + "learning_rate": 4.5880620289352855e-05, + "loss": 1.5669, + "step": 48050 + }, + { + "epoch": 4.16, + "learning_rate": 4.587975396344105e-05, + "loss": 1.6127, + "step": 48060 + }, + { + "epoch": 4.16, + "learning_rate": 4.587888763752924e-05, + "loss": 1.5937, + "step": 48070 + }, + { + "epoch": 4.16, + "learning_rate": 4.587802131161743e-05, + "loss": 1.6642, + "step": 48080 + }, + { + "epoch": 4.16, + "learning_rate": 4.587715498570562e-05, + "loss": 1.5642, + "step": 48090 + }, + { + "epoch": 4.16, + "learning_rate": 4.5876288659793814e-05, + "loss": 1.6214, + "step": 48100 + }, + { + "epoch": 4.16, + "learning_rate": 4.587542233388201e-05, + "loss": 1.6444, + "step": 48110 + }, + { + "epoch": 4.16, + "learning_rate": 4.5874556007970196e-05, + "loss": 1.6365, + "step": 48120 + }, + { + "epoch": 4.17, + "learning_rate": 4.587368968205839e-05, + "loss": 1.6306, + "step": 48130 + }, + { + "epoch": 4.17, + "learning_rate": 4.5872823356146585e-05, + "loss": 1.6463, + "step": 48140 + }, + { + "epoch": 4.17, + "learning_rate": 4.587195703023477e-05, + "loss": 1.6856, + "step": 48150 + }, + { + "epoch": 4.17, + "learning_rate": 4.587109070432297e-05, + "loss": 1.6671, + "step": 48160 + }, + { + "epoch": 4.17, + "learning_rate": 4.587022437841116e-05, + "loss": 1.5575, + "step": 48170 + }, + { + "epoch": 4.17, + "learning_rate": 4.586935805249935e-05, + "loss": 1.5803, + "step": 48180 + }, + { + "epoch": 4.17, + "learning_rate": 4.5868491726587544e-05, + "loss": 1.5891, + "step": 48190 + }, + { + "epoch": 4.17, + "learning_rate": 4.586762540067573e-05, + "loss": 1.644, + "step": 48200 + }, + { + "epoch": 4.17, + "learning_rate": 4.5866759074763926e-05, + "loss": 1.6483, + "step": 48210 + }, + { + "epoch": 4.17, + "learning_rate": 4.586589274885212e-05, + "loss": 1.6714, + "step": 48220 + }, + { + "epoch": 4.17, + "learning_rate": 4.586502642294031e-05, + "loss": 1.6337, + "step": 48230 + }, + { + "epoch": 4.18, + "learning_rate": 4.58641600970285e-05, + "loss": 1.604, + "step": 48240 + }, + { + "epoch": 4.18, + "learning_rate": 4.58632937711167e-05, + "loss": 1.5644, + "step": 48250 + }, + { + "epoch": 4.18, + "learning_rate": 4.5862427445204885e-05, + "loss": 1.5846, + "step": 48260 + }, + { + "epoch": 4.18, + "learning_rate": 4.586156111929308e-05, + "loss": 1.6195, + "step": 48270 + }, + { + "epoch": 4.18, + "learning_rate": 4.586069479338127e-05, + "loss": 1.6139, + "step": 48280 + }, + { + "epoch": 4.18, + "learning_rate": 4.585982846746946e-05, + "loss": 1.5638, + "step": 48290 + }, + { + "epoch": 4.18, + "learning_rate": 4.5858962141557656e-05, + "loss": 1.6033, + "step": 48300 + }, + { + "epoch": 4.18, + "learning_rate": 4.5858095815645843e-05, + "loss": 1.6214, + "step": 48310 + }, + { + "epoch": 4.18, + "learning_rate": 4.585722948973404e-05, + "loss": 1.626, + "step": 48320 + }, + { + "epoch": 4.18, + "learning_rate": 4.585636316382223e-05, + "loss": 1.6821, + "step": 48330 + }, + { + "epoch": 4.18, + "learning_rate": 4.585549683791042e-05, + "loss": 1.6617, + "step": 48340 + }, + { + "epoch": 4.18, + "learning_rate": 4.5854630511998615e-05, + "loss": 1.604, + "step": 48350 + }, + { + "epoch": 4.19, + "learning_rate": 4.585376418608681e-05, + "loss": 1.6011, + "step": 48360 + }, + { + "epoch": 4.19, + "learning_rate": 4.5852897860175e-05, + "loss": 1.6539, + "step": 48370 + }, + { + "epoch": 4.19, + "learning_rate": 4.585203153426319e-05, + "loss": 1.6134, + "step": 48380 + }, + { + "epoch": 4.19, + "learning_rate": 4.585116520835138e-05, + "loss": 1.6051, + "step": 48390 + }, + { + "epoch": 4.19, + "learning_rate": 4.585029888243957e-05, + "loss": 1.6551, + "step": 48400 + }, + { + "epoch": 4.19, + "learning_rate": 4.584943255652777e-05, + "loss": 1.6328, + "step": 48410 + }, + { + "epoch": 4.19, + "learning_rate": 4.5848566230615955e-05, + "loss": 1.5837, + "step": 48420 + }, + { + "epoch": 4.19, + "learning_rate": 4.584769990470415e-05, + "loss": 1.5786, + "step": 48430 + }, + { + "epoch": 4.19, + "learning_rate": 4.5846833578792344e-05, + "loss": 1.5759, + "step": 48440 + }, + { + "epoch": 4.19, + "learning_rate": 4.584596725288053e-05, + "loss": 1.646, + "step": 48450 + }, + { + "epoch": 4.19, + "learning_rate": 4.5845100926968726e-05, + "loss": 1.6382, + "step": 48460 + }, + { + "epoch": 4.2, + "learning_rate": 4.584423460105692e-05, + "loss": 1.6474, + "step": 48470 + }, + { + "epoch": 4.2, + "learning_rate": 4.584336827514511e-05, + "loss": 1.6074, + "step": 48480 + }, + { + "epoch": 4.2, + "learning_rate": 4.58425019492333e-05, + "loss": 1.6015, + "step": 48490 + }, + { + "epoch": 4.2, + "learning_rate": 4.584163562332149e-05, + "loss": 1.6308, + "step": 48500 + }, + { + "epoch": 4.2, + "learning_rate": 4.5840769297409685e-05, + "loss": 1.6126, + "step": 48510 + }, + { + "epoch": 4.2, + "learning_rate": 4.583990297149788e-05, + "loss": 1.6376, + "step": 48520 + }, + { + "epoch": 4.2, + "learning_rate": 4.583903664558607e-05, + "loss": 1.6142, + "step": 48530 + }, + { + "epoch": 4.2, + "learning_rate": 4.583817031967426e-05, + "loss": 1.6783, + "step": 48540 + }, + { + "epoch": 4.2, + "learning_rate": 4.5837303993762456e-05, + "loss": 1.6028, + "step": 48550 + }, + { + "epoch": 4.2, + "learning_rate": 4.5836437667850644e-05, + "loss": 1.6171, + "step": 48560 + }, + { + "epoch": 4.2, + "learning_rate": 4.583557134193884e-05, + "loss": 1.6516, + "step": 48570 + }, + { + "epoch": 4.2, + "learning_rate": 4.583470501602703e-05, + "loss": 1.5963, + "step": 48580 + }, + { + "epoch": 4.21, + "learning_rate": 4.583383869011522e-05, + "loss": 1.6485, + "step": 48590 + }, + { + "epoch": 4.21, + "learning_rate": 4.5832972364203415e-05, + "loss": 1.6293, + "step": 48600 + }, + { + "epoch": 4.21, + "learning_rate": 4.58321060382916e-05, + "loss": 1.5898, + "step": 48610 + }, + { + "epoch": 4.21, + "learning_rate": 4.58312397123798e-05, + "loss": 1.6805, + "step": 48620 + }, + { + "epoch": 4.21, + "learning_rate": 4.583037338646799e-05, + "loss": 1.6185, + "step": 48630 + }, + { + "epoch": 4.21, + "learning_rate": 4.582950706055618e-05, + "loss": 1.6286, + "step": 48640 + }, + { + "epoch": 4.21, + "learning_rate": 4.5828640734644374e-05, + "loss": 1.7031, + "step": 48650 + }, + { + "epoch": 4.21, + "learning_rate": 4.582777440873257e-05, + "loss": 1.597, + "step": 48660 + }, + { + "epoch": 4.21, + "learning_rate": 4.5826908082820756e-05, + "loss": 1.6702, + "step": 48670 + }, + { + "epoch": 4.21, + "learning_rate": 4.582604175690895e-05, + "loss": 1.6461, + "step": 48680 + }, + { + "epoch": 4.21, + "learning_rate": 4.5825175430997145e-05, + "loss": 1.6634, + "step": 48690 + }, + { + "epoch": 4.22, + "learning_rate": 4.582430910508533e-05, + "loss": 1.5984, + "step": 48700 + }, + { + "epoch": 4.22, + "learning_rate": 4.582344277917353e-05, + "loss": 1.5363, + "step": 48710 + }, + { + "epoch": 4.22, + "learning_rate": 4.5822576453261714e-05, + "loss": 1.6006, + "step": 48720 + }, + { + "epoch": 4.22, + "learning_rate": 4.582171012734991e-05, + "loss": 1.5786, + "step": 48730 + }, + { + "epoch": 4.22, + "learning_rate": 4.58208438014381e-05, + "loss": 1.5722, + "step": 48740 + }, + { + "epoch": 4.22, + "learning_rate": 4.581997747552629e-05, + "loss": 1.6533, + "step": 48750 + }, + { + "epoch": 4.22, + "learning_rate": 4.5819111149614485e-05, + "loss": 1.6137, + "step": 48760 + }, + { + "epoch": 4.22, + "learning_rate": 4.581824482370268e-05, + "loss": 1.6954, + "step": 48770 + }, + { + "epoch": 4.22, + "learning_rate": 4.581737849779087e-05, + "loss": 1.5383, + "step": 48780 + }, + { + "epoch": 4.22, + "learning_rate": 4.581651217187906e-05, + "loss": 1.6077, + "step": 48790 + }, + { + "epoch": 4.22, + "learning_rate": 4.5815645845967256e-05, + "loss": 1.6872, + "step": 48800 + }, + { + "epoch": 4.22, + "learning_rate": 4.5814779520055444e-05, + "loss": 1.5945, + "step": 48810 + }, + { + "epoch": 4.23, + "learning_rate": 4.581391319414364e-05, + "loss": 1.6184, + "step": 48820 + }, + { + "epoch": 4.23, + "learning_rate": 4.5813046868231826e-05, + "loss": 1.6176, + "step": 48830 + }, + { + "epoch": 4.23, + "learning_rate": 4.581218054232002e-05, + "loss": 1.5795, + "step": 48840 + }, + { + "epoch": 4.23, + "learning_rate": 4.5811314216408215e-05, + "loss": 1.5871, + "step": 48850 + }, + { + "epoch": 4.23, + "learning_rate": 4.58104478904964e-05, + "loss": 1.6596, + "step": 48860 + }, + { + "epoch": 4.23, + "learning_rate": 4.58095815645846e-05, + "loss": 1.6028, + "step": 48870 + }, + { + "epoch": 4.23, + "learning_rate": 4.580871523867279e-05, + "loss": 1.5897, + "step": 48880 + }, + { + "epoch": 4.23, + "learning_rate": 4.580784891276098e-05, + "loss": 1.6593, + "step": 48890 + }, + { + "epoch": 4.23, + "learning_rate": 4.5806982586849174e-05, + "loss": 1.6273, + "step": 48900 + }, + { + "epoch": 4.23, + "learning_rate": 4.580611626093737e-05, + "loss": 1.5854, + "step": 48910 + }, + { + "epoch": 4.23, + "learning_rate": 4.5805249935025556e-05, + "loss": 1.641, + "step": 48920 + }, + { + "epoch": 4.24, + "learning_rate": 4.580438360911375e-05, + "loss": 1.5832, + "step": 48930 + }, + { + "epoch": 4.24, + "learning_rate": 4.580351728320194e-05, + "loss": 1.5961, + "step": 48940 + }, + { + "epoch": 4.24, + "learning_rate": 4.580265095729013e-05, + "loss": 1.649, + "step": 48950 + }, + { + "epoch": 4.24, + "learning_rate": 4.580178463137833e-05, + "loss": 1.6304, + "step": 48960 + }, + { + "epoch": 4.24, + "learning_rate": 4.5800918305466515e-05, + "loss": 1.5805, + "step": 48970 + }, + { + "epoch": 4.24, + "learning_rate": 4.580005197955471e-05, + "loss": 1.6194, + "step": 48980 + }, + { + "epoch": 4.24, + "learning_rate": 4.5799185653642904e-05, + "loss": 1.5984, + "step": 48990 + }, + { + "epoch": 4.24, + "learning_rate": 4.579831932773109e-05, + "loss": 1.6469, + "step": 49000 + }, + { + "epoch": 4.24, + "learning_rate": 4.5797453001819286e-05, + "loss": 1.5812, + "step": 49010 + }, + { + "epoch": 4.24, + "learning_rate": 4.579658667590748e-05, + "loss": 1.623, + "step": 49020 + }, + { + "epoch": 4.24, + "learning_rate": 4.579572034999567e-05, + "loss": 1.6885, + "step": 49030 + }, + { + "epoch": 4.24, + "learning_rate": 4.579485402408386e-05, + "loss": 1.5795, + "step": 49040 + }, + { + "epoch": 4.25, + "learning_rate": 4.579398769817205e-05, + "loss": 1.5628, + "step": 49050 + }, + { + "epoch": 4.25, + "learning_rate": 4.5793121372260244e-05, + "loss": 1.5999, + "step": 49060 + }, + { + "epoch": 4.25, + "learning_rate": 4.579225504634844e-05, + "loss": 1.6633, + "step": 49070 + }, + { + "epoch": 4.25, + "learning_rate": 4.579138872043663e-05, + "loss": 1.6378, + "step": 49080 + }, + { + "epoch": 4.25, + "learning_rate": 4.579052239452482e-05, + "loss": 1.5911, + "step": 49090 + }, + { + "epoch": 4.25, + "learning_rate": 4.5789656068613016e-05, + "loss": 1.6117, + "step": 49100 + }, + { + "epoch": 4.25, + "learning_rate": 4.57887897427012e-05, + "loss": 1.6305, + "step": 49110 + }, + { + "epoch": 4.25, + "learning_rate": 4.57879234167894e-05, + "loss": 1.5133, + "step": 49120 + }, + { + "epoch": 4.25, + "learning_rate": 4.5787057090877585e-05, + "loss": 1.6723, + "step": 49130 + }, + { + "epoch": 4.25, + "learning_rate": 4.578619076496578e-05, + "loss": 1.6012, + "step": 49140 + }, + { + "epoch": 4.25, + "learning_rate": 4.5785324439053974e-05, + "loss": 1.5709, + "step": 49150 + }, + { + "epoch": 4.25, + "learning_rate": 4.578445811314216e-05, + "loss": 1.6404, + "step": 49160 + }, + { + "epoch": 4.26, + "learning_rate": 4.5783591787230356e-05, + "loss": 1.6025, + "step": 49170 + }, + { + "epoch": 4.26, + "learning_rate": 4.578272546131855e-05, + "loss": 1.6447, + "step": 49180 + }, + { + "epoch": 4.26, + "learning_rate": 4.578185913540674e-05, + "loss": 1.6072, + "step": 49190 + }, + { + "epoch": 4.26, + "learning_rate": 4.578099280949493e-05, + "loss": 1.5633, + "step": 49200 + }, + { + "epoch": 4.26, + "learning_rate": 4.578012648358313e-05, + "loss": 1.6355, + "step": 49210 + }, + { + "epoch": 4.26, + "learning_rate": 4.5779260157671315e-05, + "loss": 1.5949, + "step": 49220 + }, + { + "epoch": 4.26, + "learning_rate": 4.577839383175951e-05, + "loss": 1.5437, + "step": 49230 + }, + { + "epoch": 4.26, + "learning_rate": 4.57775275058477e-05, + "loss": 1.6058, + "step": 49240 + }, + { + "epoch": 4.26, + "learning_rate": 4.577666117993589e-05, + "loss": 1.6462, + "step": 49250 + }, + { + "epoch": 4.26, + "learning_rate": 4.5775794854024086e-05, + "loss": 1.7178, + "step": 49260 + }, + { + "epoch": 4.26, + "learning_rate": 4.5774928528112274e-05, + "loss": 1.6445, + "step": 49270 + }, + { + "epoch": 4.27, + "learning_rate": 4.577406220220047e-05, + "loss": 1.6298, + "step": 49280 + }, + { + "epoch": 4.27, + "learning_rate": 4.577319587628866e-05, + "loss": 1.6627, + "step": 49290 + }, + { + "epoch": 4.27, + "learning_rate": 4.577232955037685e-05, + "loss": 1.689, + "step": 49300 + }, + { + "epoch": 4.27, + "learning_rate": 4.5771463224465045e-05, + "loss": 1.5874, + "step": 49310 + }, + { + "epoch": 4.27, + "learning_rate": 4.577059689855324e-05, + "loss": 1.5809, + "step": 49320 + }, + { + "epoch": 4.27, + "learning_rate": 4.576973057264143e-05, + "loss": 1.6216, + "step": 49330 + }, + { + "epoch": 4.27, + "learning_rate": 4.576886424672962e-05, + "loss": 1.6235, + "step": 49340 + }, + { + "epoch": 4.27, + "learning_rate": 4.576799792081781e-05, + "loss": 1.6382, + "step": 49350 + }, + { + "epoch": 4.27, + "learning_rate": 4.5767131594906004e-05, + "loss": 1.6507, + "step": 49360 + }, + { + "epoch": 4.27, + "learning_rate": 4.57662652689942e-05, + "loss": 1.6017, + "step": 49370 + }, + { + "epoch": 4.27, + "learning_rate": 4.5765398943082386e-05, + "loss": 1.6537, + "step": 49380 + }, + { + "epoch": 4.27, + "learning_rate": 4.576453261717058e-05, + "loss": 1.5961, + "step": 49390 + }, + { + "epoch": 4.28, + "learning_rate": 4.5763666291258775e-05, + "loss": 1.5983, + "step": 49400 + }, + { + "epoch": 4.28, + "learning_rate": 4.576279996534696e-05, + "loss": 1.5728, + "step": 49410 + }, + { + "epoch": 4.28, + "learning_rate": 4.576193363943516e-05, + "loss": 1.6291, + "step": 49420 + }, + { + "epoch": 4.28, + "learning_rate": 4.576106731352335e-05, + "loss": 1.6355, + "step": 49430 + }, + { + "epoch": 4.28, + "learning_rate": 4.576020098761154e-05, + "loss": 1.5729, + "step": 49440 + }, + { + "epoch": 4.28, + "learning_rate": 4.575933466169973e-05, + "loss": 1.695, + "step": 49450 + }, + { + "epoch": 4.28, + "learning_rate": 4.575846833578792e-05, + "loss": 1.6119, + "step": 49460 + }, + { + "epoch": 4.28, + "learning_rate": 4.5757602009876115e-05, + "loss": 1.6204, + "step": 49470 + }, + { + "epoch": 4.28, + "learning_rate": 4.575673568396431e-05, + "loss": 1.5853, + "step": 49480 + }, + { + "epoch": 4.28, + "learning_rate": 4.57558693580525e-05, + "loss": 1.5972, + "step": 49490 + }, + { + "epoch": 4.28, + "learning_rate": 4.575500303214069e-05, + "loss": 1.6409, + "step": 49500 + }, + { + "epoch": 4.29, + "learning_rate": 4.5754136706228886e-05, + "loss": 1.6467, + "step": 49510 + }, + { + "epoch": 4.29, + "learning_rate": 4.5753270380317074e-05, + "loss": 1.5646, + "step": 49520 + }, + { + "epoch": 4.29, + "learning_rate": 4.575240405440527e-05, + "loss": 1.5697, + "step": 49530 + }, + { + "epoch": 4.29, + "learning_rate": 4.575153772849346e-05, + "loss": 1.648, + "step": 49540 + }, + { + "epoch": 4.29, + "learning_rate": 4.575067140258165e-05, + "loss": 1.5942, + "step": 49550 + }, + { + "epoch": 4.29, + "learning_rate": 4.5749805076669845e-05, + "loss": 1.6065, + "step": 49560 + }, + { + "epoch": 4.29, + "learning_rate": 4.574893875075803e-05, + "loss": 1.632, + "step": 49570 + }, + { + "epoch": 4.29, + "learning_rate": 4.574807242484623e-05, + "loss": 1.6496, + "step": 49580 + }, + { + "epoch": 4.29, + "learning_rate": 4.574720609893442e-05, + "loss": 1.64, + "step": 49590 + }, + { + "epoch": 4.29, + "learning_rate": 4.574633977302261e-05, + "loss": 1.6761, + "step": 49600 + }, + { + "epoch": 4.29, + "learning_rate": 4.5745473447110804e-05, + "loss": 1.6226, + "step": 49610 + }, + { + "epoch": 4.29, + "learning_rate": 4.5744607121199e-05, + "loss": 1.7076, + "step": 49620 + }, + { + "epoch": 4.3, + "learning_rate": 4.5743740795287186e-05, + "loss": 1.6621, + "step": 49630 + }, + { + "epoch": 4.3, + "learning_rate": 4.574287446937538e-05, + "loss": 1.6416, + "step": 49640 + }, + { + "epoch": 4.3, + "learning_rate": 4.5742008143463575e-05, + "loss": 1.623, + "step": 49650 + }, + { + "epoch": 4.3, + "learning_rate": 4.574114181755176e-05, + "loss": 1.6491, + "step": 49660 + }, + { + "epoch": 4.3, + "learning_rate": 4.574027549163996e-05, + "loss": 1.6588, + "step": 49670 + }, + { + "epoch": 4.3, + "learning_rate": 4.5739409165728145e-05, + "loss": 1.5786, + "step": 49680 + }, + { + "epoch": 4.3, + "learning_rate": 4.573854283981634e-05, + "loss": 1.612, + "step": 49690 + }, + { + "epoch": 4.3, + "learning_rate": 4.5737676513904534e-05, + "loss": 1.6461, + "step": 49700 + }, + { + "epoch": 4.3, + "learning_rate": 4.573681018799272e-05, + "loss": 1.6131, + "step": 49710 + }, + { + "epoch": 4.3, + "learning_rate": 4.5735943862080916e-05, + "loss": 1.5786, + "step": 49720 + }, + { + "epoch": 4.3, + "learning_rate": 4.573507753616911e-05, + "loss": 1.5864, + "step": 49730 + }, + { + "epoch": 4.31, + "learning_rate": 4.57342112102573e-05, + "loss": 1.6288, + "step": 49740 + }, + { + "epoch": 4.31, + "learning_rate": 4.573334488434549e-05, + "loss": 1.567, + "step": 49750 + }, + { + "epoch": 4.31, + "learning_rate": 4.573247855843369e-05, + "loss": 1.6051, + "step": 49760 + }, + { + "epoch": 4.31, + "learning_rate": 4.5731612232521874e-05, + "loss": 1.5519, + "step": 49770 + }, + { + "epoch": 4.31, + "learning_rate": 4.573074590661007e-05, + "loss": 1.5629, + "step": 49780 + }, + { + "epoch": 4.31, + "learning_rate": 4.5729879580698257e-05, + "loss": 1.6259, + "step": 49790 + }, + { + "epoch": 4.31, + "learning_rate": 4.572901325478645e-05, + "loss": 1.5794, + "step": 49800 + }, + { + "epoch": 4.31, + "learning_rate": 4.5728146928874646e-05, + "loss": 1.6817, + "step": 49810 + }, + { + "epoch": 4.31, + "learning_rate": 4.572728060296283e-05, + "loss": 1.6486, + "step": 49820 + }, + { + "epoch": 4.31, + "learning_rate": 4.572641427705103e-05, + "loss": 1.6146, + "step": 49830 + }, + { + "epoch": 4.31, + "learning_rate": 4.572554795113922e-05, + "loss": 1.6192, + "step": 49840 + }, + { + "epoch": 4.31, + "learning_rate": 4.572468162522741e-05, + "loss": 1.6349, + "step": 49850 + }, + { + "epoch": 4.32, + "learning_rate": 4.5723815299315604e-05, + "loss": 1.6064, + "step": 49860 + }, + { + "epoch": 4.32, + "learning_rate": 4.572294897340379e-05, + "loss": 1.6576, + "step": 49870 + }, + { + "epoch": 4.32, + "learning_rate": 4.5722082647491986e-05, + "loss": 1.627, + "step": 49880 + }, + { + "epoch": 4.32, + "learning_rate": 4.572121632158018e-05, + "loss": 1.6054, + "step": 49890 + }, + { + "epoch": 4.32, + "learning_rate": 4.572034999566837e-05, + "loss": 1.5631, + "step": 49900 + }, + { + "epoch": 4.32, + "learning_rate": 4.571948366975656e-05, + "loss": 1.5866, + "step": 49910 + }, + { + "epoch": 4.32, + "learning_rate": 4.571861734384476e-05, + "loss": 1.6537, + "step": 49920 + }, + { + "epoch": 4.32, + "learning_rate": 4.5717751017932945e-05, + "loss": 1.6012, + "step": 49930 + }, + { + "epoch": 4.32, + "learning_rate": 4.571688469202114e-05, + "loss": 1.5529, + "step": 49940 + }, + { + "epoch": 4.32, + "learning_rate": 4.5716018366109334e-05, + "loss": 1.609, + "step": 49950 + }, + { + "epoch": 4.32, + "learning_rate": 4.571515204019752e-05, + "loss": 1.5723, + "step": 49960 + }, + { + "epoch": 4.33, + "learning_rate": 4.5714285714285716e-05, + "loss": 1.6694, + "step": 49970 + }, + { + "epoch": 4.33, + "learning_rate": 4.5713419388373904e-05, + "loss": 1.646, + "step": 49980 + }, + { + "epoch": 4.33, + "learning_rate": 4.57125530624621e-05, + "loss": 1.6455, + "step": 49990 + }, + { + "epoch": 4.33, + "learning_rate": 4.571168673655029e-05, + "loss": 1.6489, + "step": 50000 + }, + { + "epoch": 4.33, + "learning_rate": 4.571082041063848e-05, + "loss": 1.5444, + "step": 50010 + }, + { + "epoch": 4.33, + "learning_rate": 4.5709954084726675e-05, + "loss": 1.6392, + "step": 50020 + }, + { + "epoch": 4.33, + "learning_rate": 4.570908775881487e-05, + "loss": 1.5868, + "step": 50030 + }, + { + "epoch": 4.33, + "learning_rate": 4.570822143290306e-05, + "loss": 1.62, + "step": 50040 + }, + { + "epoch": 4.33, + "learning_rate": 4.570735510699125e-05, + "loss": 1.6698, + "step": 50050 + }, + { + "epoch": 4.33, + "learning_rate": 4.5706488781079446e-05, + "loss": 1.6045, + "step": 50060 + }, + { + "epoch": 4.33, + "learning_rate": 4.5705622455167634e-05, + "loss": 1.6251, + "step": 50070 + }, + { + "epoch": 4.33, + "learning_rate": 4.570475612925583e-05, + "loss": 1.6268, + "step": 50080 + }, + { + "epoch": 4.34, + "learning_rate": 4.5703889803344016e-05, + "loss": 1.6167, + "step": 50090 + }, + { + "epoch": 4.34, + "learning_rate": 4.570302347743221e-05, + "loss": 1.7152, + "step": 50100 + }, + { + "epoch": 4.34, + "learning_rate": 4.5702157151520405e-05, + "loss": 1.6197, + "step": 50110 + }, + { + "epoch": 4.34, + "learning_rate": 4.570129082560859e-05, + "loss": 1.6414, + "step": 50120 + }, + { + "epoch": 4.34, + "learning_rate": 4.570042449969679e-05, + "loss": 1.6288, + "step": 50130 + }, + { + "epoch": 4.34, + "learning_rate": 4.569955817378498e-05, + "loss": 1.6569, + "step": 50140 + }, + { + "epoch": 4.34, + "learning_rate": 4.569869184787317e-05, + "loss": 1.5943, + "step": 50150 + }, + { + "epoch": 4.34, + "learning_rate": 4.569782552196136e-05, + "loss": 1.591, + "step": 50160 + }, + { + "epoch": 4.34, + "learning_rate": 4.569695919604956e-05, + "loss": 1.7166, + "step": 50170 + }, + { + "epoch": 4.34, + "learning_rate": 4.5696092870137745e-05, + "loss": 1.6343, + "step": 50180 + }, + { + "epoch": 4.34, + "learning_rate": 4.569522654422594e-05, + "loss": 1.5601, + "step": 50190 + }, + { + "epoch": 4.35, + "learning_rate": 4.569436021831413e-05, + "loss": 1.7027, + "step": 50200 + }, + { + "epoch": 4.35, + "learning_rate": 4.569349389240232e-05, + "loss": 1.667, + "step": 50210 + }, + { + "epoch": 4.35, + "learning_rate": 4.5692627566490516e-05, + "loss": 1.6196, + "step": 50220 + }, + { + "epoch": 4.35, + "learning_rate": 4.5691761240578704e-05, + "loss": 1.5808, + "step": 50230 + }, + { + "epoch": 4.35, + "learning_rate": 4.56908949146669e-05, + "loss": 1.6005, + "step": 50240 + }, + { + "epoch": 4.35, + "learning_rate": 4.569002858875509e-05, + "loss": 1.5913, + "step": 50250 + }, + { + "epoch": 4.35, + "learning_rate": 4.568916226284328e-05, + "loss": 1.631, + "step": 50260 + }, + { + "epoch": 4.35, + "learning_rate": 4.5688295936931475e-05, + "loss": 1.5961, + "step": 50270 + }, + { + "epoch": 4.35, + "learning_rate": 4.568742961101967e-05, + "loss": 1.5798, + "step": 50280 + }, + { + "epoch": 4.35, + "learning_rate": 4.568656328510786e-05, + "loss": 1.6514, + "step": 50290 + }, + { + "epoch": 4.35, + "learning_rate": 4.568569695919605e-05, + "loss": 1.6015, + "step": 50300 + }, + { + "epoch": 4.35, + "learning_rate": 4.568483063328424e-05, + "loss": 1.6015, + "step": 50310 + }, + { + "epoch": 4.36, + "learning_rate": 4.5683964307372434e-05, + "loss": 1.6348, + "step": 50320 + }, + { + "epoch": 4.36, + "learning_rate": 4.568309798146063e-05, + "loss": 1.5834, + "step": 50330 + }, + { + "epoch": 4.36, + "learning_rate": 4.5682231655548816e-05, + "loss": 1.632, + "step": 50340 + }, + { + "epoch": 4.36, + "learning_rate": 4.568136532963701e-05, + "loss": 1.6972, + "step": 50350 + }, + { + "epoch": 4.36, + "learning_rate": 4.5680499003725205e-05, + "loss": 1.7158, + "step": 50360 + }, + { + "epoch": 4.36, + "learning_rate": 4.567963267781339e-05, + "loss": 1.6463, + "step": 50370 + }, + { + "epoch": 4.36, + "learning_rate": 4.567876635190159e-05, + "loss": 1.5576, + "step": 50380 + }, + { + "epoch": 4.36, + "learning_rate": 4.567790002598978e-05, + "loss": 1.6202, + "step": 50390 + }, + { + "epoch": 4.36, + "learning_rate": 4.567703370007797e-05, + "loss": 1.6498, + "step": 50400 + }, + { + "epoch": 4.36, + "learning_rate": 4.5676167374166164e-05, + "loss": 1.605, + "step": 50410 + }, + { + "epoch": 4.36, + "learning_rate": 4.567530104825435e-05, + "loss": 1.5735, + "step": 50420 + }, + { + "epoch": 4.36, + "learning_rate": 4.5674434722342546e-05, + "loss": 1.6914, + "step": 50430 + }, + { + "epoch": 4.37, + "learning_rate": 4.567356839643074e-05, + "loss": 1.6304, + "step": 50440 + }, + { + "epoch": 4.37, + "learning_rate": 4.567270207051893e-05, + "loss": 1.6351, + "step": 50450 + }, + { + "epoch": 4.37, + "learning_rate": 4.567183574460712e-05, + "loss": 1.6471, + "step": 50460 + }, + { + "epoch": 4.37, + "learning_rate": 4.567096941869532e-05, + "loss": 1.5962, + "step": 50470 + }, + { + "epoch": 4.37, + "learning_rate": 4.5670103092783504e-05, + "loss": 1.6911, + "step": 50480 + }, + { + "epoch": 4.37, + "learning_rate": 4.56692367668717e-05, + "loss": 1.6321, + "step": 50490 + }, + { + "epoch": 4.37, + "learning_rate": 4.566837044095989e-05, + "loss": 1.5797, + "step": 50500 + }, + { + "epoch": 4.37, + "learning_rate": 4.566750411504808e-05, + "loss": 1.629, + "step": 50510 + }, + { + "epoch": 4.37, + "learning_rate": 4.5666637789136275e-05, + "loss": 1.576, + "step": 50520 + }, + { + "epoch": 4.37, + "learning_rate": 4.566577146322446e-05, + "loss": 1.6482, + "step": 50530 + }, + { + "epoch": 4.37, + "learning_rate": 4.566490513731266e-05, + "loss": 1.6941, + "step": 50540 + }, + { + "epoch": 4.38, + "learning_rate": 4.566403881140085e-05, + "loss": 1.5959, + "step": 50550 + }, + { + "epoch": 4.38, + "learning_rate": 4.566317248548904e-05, + "loss": 1.5953, + "step": 50560 + }, + { + "epoch": 4.38, + "learning_rate": 4.5662306159577234e-05, + "loss": 1.6083, + "step": 50570 + }, + { + "epoch": 4.38, + "learning_rate": 4.566143983366543e-05, + "loss": 1.6326, + "step": 50580 + }, + { + "epoch": 4.38, + "learning_rate": 4.5660573507753616e-05, + "loss": 1.6318, + "step": 50590 + }, + { + "epoch": 4.38, + "learning_rate": 4.565970718184181e-05, + "loss": 1.6089, + "step": 50600 + }, + { + "epoch": 4.38, + "learning_rate": 4.565884085593e-05, + "loss": 1.6612, + "step": 50610 + }, + { + "epoch": 4.38, + "learning_rate": 4.565797453001819e-05, + "loss": 1.6443, + "step": 50620 + }, + { + "epoch": 4.38, + "learning_rate": 4.565710820410639e-05, + "loss": 1.6186, + "step": 50630 + }, + { + "epoch": 4.38, + "learning_rate": 4.5656241878194575e-05, + "loss": 1.5943, + "step": 50640 + }, + { + "epoch": 4.38, + "learning_rate": 4.565537555228277e-05, + "loss": 1.5716, + "step": 50650 + }, + { + "epoch": 4.38, + "learning_rate": 4.5654509226370964e-05, + "loss": 1.6212, + "step": 50660 + }, + { + "epoch": 4.39, + "learning_rate": 4.565364290045915e-05, + "loss": 1.6557, + "step": 50670 + }, + { + "epoch": 4.39, + "learning_rate": 4.5652776574547346e-05, + "loss": 1.5951, + "step": 50680 + }, + { + "epoch": 4.39, + "learning_rate": 4.565191024863554e-05, + "loss": 1.6031, + "step": 50690 + }, + { + "epoch": 4.39, + "learning_rate": 4.565104392272373e-05, + "loss": 1.6461, + "step": 50700 + }, + { + "epoch": 4.39, + "learning_rate": 4.565017759681192e-05, + "loss": 1.5866, + "step": 50710 + }, + { + "epoch": 4.39, + "learning_rate": 4.564931127090011e-05, + "loss": 1.6159, + "step": 50720 + }, + { + "epoch": 4.39, + "learning_rate": 4.5648444944988305e-05, + "loss": 1.618, + "step": 50730 + }, + { + "epoch": 4.39, + "learning_rate": 4.56475786190765e-05, + "loss": 1.6333, + "step": 50740 + }, + { + "epoch": 4.39, + "learning_rate": 4.564671229316469e-05, + "loss": 1.6908, + "step": 50750 + }, + { + "epoch": 4.39, + "learning_rate": 4.564584596725288e-05, + "loss": 1.6682, + "step": 50760 + }, + { + "epoch": 4.39, + "learning_rate": 4.5644979641341076e-05, + "loss": 1.5611, + "step": 50770 + }, + { + "epoch": 4.4, + "learning_rate": 4.5644113315429263e-05, + "loss": 1.5872, + "step": 50780 + }, + { + "epoch": 4.4, + "learning_rate": 4.564324698951746e-05, + "loss": 1.6525, + "step": 50790 + }, + { + "epoch": 4.4, + "learning_rate": 4.564238066360565e-05, + "loss": 1.5701, + "step": 50800 + }, + { + "epoch": 4.4, + "learning_rate": 4.564151433769384e-05, + "loss": 1.6165, + "step": 50810 + }, + { + "epoch": 4.4, + "learning_rate": 4.5640648011782035e-05, + "loss": 1.6447, + "step": 50820 + }, + { + "epoch": 4.4, + "learning_rate": 4.563978168587022e-05, + "loss": 1.5604, + "step": 50830 + }, + { + "epoch": 4.4, + "learning_rate": 4.563891535995842e-05, + "loss": 1.57, + "step": 50840 + }, + { + "epoch": 4.4, + "learning_rate": 4.563804903404661e-05, + "loss": 1.6684, + "step": 50850 + }, + { + "epoch": 4.4, + "learning_rate": 4.56371827081348e-05, + "loss": 1.6324, + "step": 50860 + }, + { + "epoch": 4.4, + "learning_rate": 4.563631638222299e-05, + "loss": 1.6345, + "step": 50870 + }, + { + "epoch": 4.4, + "learning_rate": 4.563545005631119e-05, + "loss": 1.6145, + "step": 50880 + }, + { + "epoch": 4.4, + "learning_rate": 4.5634583730399375e-05, + "loss": 1.5755, + "step": 50890 + }, + { + "epoch": 4.41, + "learning_rate": 4.563371740448757e-05, + "loss": 1.5949, + "step": 50900 + }, + { + "epoch": 4.41, + "learning_rate": 4.5632851078575764e-05, + "loss": 1.6212, + "step": 50910 + }, + { + "epoch": 4.41, + "learning_rate": 4.563198475266395e-05, + "loss": 1.5166, + "step": 50920 + }, + { + "epoch": 4.41, + "learning_rate": 4.5631118426752146e-05, + "loss": 1.6337, + "step": 50930 + }, + { + "epoch": 4.41, + "learning_rate": 4.5630252100840334e-05, + "loss": 1.6098, + "step": 50940 + }, + { + "epoch": 4.41, + "learning_rate": 4.562938577492853e-05, + "loss": 1.6614, + "step": 50950 + }, + { + "epoch": 4.41, + "learning_rate": 4.562851944901672e-05, + "loss": 1.5822, + "step": 50960 + }, + { + "epoch": 4.41, + "learning_rate": 4.562765312310491e-05, + "loss": 1.6245, + "step": 50970 + }, + { + "epoch": 4.41, + "learning_rate": 4.5626786797193105e-05, + "loss": 1.5995, + "step": 50980 + }, + { + "epoch": 4.41, + "learning_rate": 4.56259204712813e-05, + "loss": 1.6827, + "step": 50990 + }, + { + "epoch": 4.41, + "learning_rate": 4.562505414536949e-05, + "loss": 1.6035, + "step": 51000 + }, + { + "epoch": 4.42, + "learning_rate": 4.562418781945768e-05, + "loss": 1.5925, + "step": 51010 + }, + { + "epoch": 4.42, + "learning_rate": 4.5623321493545876e-05, + "loss": 1.5804, + "step": 51020 + }, + { + "epoch": 4.42, + "learning_rate": 4.5622455167634064e-05, + "loss": 1.6447, + "step": 51030 + }, + { + "epoch": 4.42, + "learning_rate": 4.562158884172226e-05, + "loss": 1.6162, + "step": 51040 + }, + { + "epoch": 4.42, + "learning_rate": 4.5620722515810446e-05, + "loss": 1.5993, + "step": 51050 + }, + { + "epoch": 4.42, + "learning_rate": 4.561985618989864e-05, + "loss": 1.5789, + "step": 51060 + }, + { + "epoch": 4.42, + "learning_rate": 4.5618989863986835e-05, + "loss": 1.6795, + "step": 51070 + }, + { + "epoch": 4.42, + "learning_rate": 4.561812353807502e-05, + "loss": 1.6242, + "step": 51080 + }, + { + "epoch": 4.42, + "learning_rate": 4.561725721216322e-05, + "loss": 1.6451, + "step": 51090 + }, + { + "epoch": 4.42, + "learning_rate": 4.561639088625141e-05, + "loss": 1.6124, + "step": 51100 + }, + { + "epoch": 4.42, + "learning_rate": 4.56155245603396e-05, + "loss": 1.6631, + "step": 51110 + }, + { + "epoch": 4.42, + "learning_rate": 4.5614658234427794e-05, + "loss": 1.6308, + "step": 51120 + }, + { + "epoch": 4.43, + "learning_rate": 4.561379190851599e-05, + "loss": 1.549, + "step": 51130 + }, + { + "epoch": 4.43, + "learning_rate": 4.5612925582604176e-05, + "loss": 1.6485, + "step": 51140 + }, + { + "epoch": 4.43, + "learning_rate": 4.561205925669237e-05, + "loss": 1.5884, + "step": 51150 + }, + { + "epoch": 4.43, + "learning_rate": 4.561119293078056e-05, + "loss": 1.6052, + "step": 51160 + }, + { + "epoch": 4.43, + "learning_rate": 4.561032660486875e-05, + "loss": 1.613, + "step": 51170 + }, + { + "epoch": 4.43, + "learning_rate": 4.560946027895695e-05, + "loss": 1.5743, + "step": 51180 + }, + { + "epoch": 4.43, + "learning_rate": 4.5608593953045134e-05, + "loss": 1.6629, + "step": 51190 + }, + { + "epoch": 4.43, + "learning_rate": 4.560772762713333e-05, + "loss": 1.5938, + "step": 51200 + }, + { + "epoch": 4.43, + "learning_rate": 4.560686130122152e-05, + "loss": 1.5716, + "step": 51210 + }, + { + "epoch": 4.43, + "learning_rate": 4.560599497530971e-05, + "loss": 1.6907, + "step": 51220 + }, + { + "epoch": 4.43, + "learning_rate": 4.5605128649397905e-05, + "loss": 1.5952, + "step": 51230 + }, + { + "epoch": 4.44, + "learning_rate": 4.56042623234861e-05, + "loss": 1.5508, + "step": 51240 + }, + { + "epoch": 4.44, + "learning_rate": 4.560339599757429e-05, + "loss": 1.5759, + "step": 51250 + }, + { + "epoch": 4.44, + "learning_rate": 4.560252967166248e-05, + "loss": 1.6769, + "step": 51260 + }, + { + "epoch": 4.44, + "learning_rate": 4.560166334575067e-05, + "loss": 1.6051, + "step": 51270 + }, + { + "epoch": 4.44, + "learning_rate": 4.5600797019838864e-05, + "loss": 1.5714, + "step": 51280 + }, + { + "epoch": 4.44, + "learning_rate": 4.559993069392706e-05, + "loss": 1.6195, + "step": 51290 + }, + { + "epoch": 4.44, + "learning_rate": 4.5599064368015246e-05, + "loss": 1.6496, + "step": 51300 + }, + { + "epoch": 4.44, + "learning_rate": 4.559819804210344e-05, + "loss": 1.5096, + "step": 51310 + }, + { + "epoch": 4.44, + "learning_rate": 4.5597331716191635e-05, + "loss": 1.6298, + "step": 51320 + }, + { + "epoch": 4.44, + "learning_rate": 4.559646539027982e-05, + "loss": 1.659, + "step": 51330 + }, + { + "epoch": 4.44, + "learning_rate": 4.559559906436802e-05, + "loss": 1.5842, + "step": 51340 + }, + { + "epoch": 4.44, + "learning_rate": 4.559473273845621e-05, + "loss": 1.6322, + "step": 51350 + }, + { + "epoch": 4.45, + "learning_rate": 4.55938664125444e-05, + "loss": 1.6262, + "step": 51360 + }, + { + "epoch": 4.45, + "learning_rate": 4.5593000086632594e-05, + "loss": 1.643, + "step": 51370 + }, + { + "epoch": 4.45, + "learning_rate": 4.559213376072078e-05, + "loss": 1.5587, + "step": 51380 + }, + { + "epoch": 4.45, + "learning_rate": 4.5591267434808976e-05, + "loss": 1.5556, + "step": 51390 + }, + { + "epoch": 4.45, + "learning_rate": 4.559040110889717e-05, + "loss": 1.6807, + "step": 51400 + }, + { + "epoch": 4.45, + "learning_rate": 4.558953478298536e-05, + "loss": 1.6239, + "step": 51410 + }, + { + "epoch": 4.45, + "learning_rate": 4.558866845707355e-05, + "loss": 1.6351, + "step": 51420 + }, + { + "epoch": 4.45, + "learning_rate": 4.558780213116175e-05, + "loss": 1.6321, + "step": 51430 + }, + { + "epoch": 4.45, + "learning_rate": 4.5586935805249935e-05, + "loss": 1.5861, + "step": 51440 + }, + { + "epoch": 4.45, + "learning_rate": 4.558606947933813e-05, + "loss": 1.6377, + "step": 51450 + }, + { + "epoch": 4.45, + "learning_rate": 4.558520315342632e-05, + "loss": 1.6035, + "step": 51460 + }, + { + "epoch": 4.45, + "learning_rate": 4.558433682751451e-05, + "loss": 1.548, + "step": 51470 + }, + { + "epoch": 4.46, + "learning_rate": 4.5583470501602706e-05, + "loss": 1.6794, + "step": 51480 + }, + { + "epoch": 4.46, + "learning_rate": 4.5582604175690893e-05, + "loss": 1.6039, + "step": 51490 + }, + { + "epoch": 4.46, + "learning_rate": 4.558173784977909e-05, + "loss": 1.5489, + "step": 51500 + }, + { + "epoch": 4.46, + "learning_rate": 4.558087152386728e-05, + "loss": 1.568, + "step": 51510 + }, + { + "epoch": 4.46, + "learning_rate": 4.558000519795547e-05, + "loss": 1.6205, + "step": 51520 + }, + { + "epoch": 4.46, + "learning_rate": 4.5579138872043664e-05, + "loss": 1.5996, + "step": 51530 + }, + { + "epoch": 4.46, + "learning_rate": 4.557827254613186e-05, + "loss": 1.6603, + "step": 51540 + }, + { + "epoch": 4.46, + "learning_rate": 4.557740622022005e-05, + "loss": 1.5623, + "step": 51550 + }, + { + "epoch": 4.46, + "learning_rate": 4.557653989430824e-05, + "loss": 1.5837, + "step": 51560 + }, + { + "epoch": 4.46, + "learning_rate": 4.557567356839643e-05, + "loss": 1.5657, + "step": 51570 + }, + { + "epoch": 4.46, + "learning_rate": 4.557480724248462e-05, + "loss": 1.636, + "step": 51580 + }, + { + "epoch": 4.47, + "learning_rate": 4.557394091657282e-05, + "loss": 1.5956, + "step": 51590 + }, + { + "epoch": 4.47, + "learning_rate": 4.5573074590661005e-05, + "loss": 1.6553, + "step": 51600 + }, + { + "epoch": 4.47, + "learning_rate": 4.55722082647492e-05, + "loss": 1.7242, + "step": 51610 + }, + { + "epoch": 4.47, + "learning_rate": 4.5571341938837394e-05, + "loss": 1.5438, + "step": 51620 + }, + { + "epoch": 4.47, + "learning_rate": 4.557047561292558e-05, + "loss": 1.6659, + "step": 51630 + }, + { + "epoch": 4.47, + "learning_rate": 4.5569609287013776e-05, + "loss": 1.5773, + "step": 51640 + }, + { + "epoch": 4.47, + "learning_rate": 4.556874296110197e-05, + "loss": 1.5913, + "step": 51650 + }, + { + "epoch": 4.47, + "learning_rate": 4.556787663519016e-05, + "loss": 1.5853, + "step": 51660 + }, + { + "epoch": 4.47, + "learning_rate": 4.556701030927835e-05, + "loss": 1.7056, + "step": 51670 + }, + { + "epoch": 4.47, + "learning_rate": 4.556614398336654e-05, + "loss": 1.6211, + "step": 51680 + }, + { + "epoch": 4.47, + "learning_rate": 4.5565277657454735e-05, + "loss": 1.6114, + "step": 51690 + }, + { + "epoch": 4.47, + "learning_rate": 4.556441133154293e-05, + "loss": 1.6079, + "step": 51700 + }, + { + "epoch": 4.48, + "learning_rate": 4.556354500563112e-05, + "loss": 1.6013, + "step": 51710 + }, + { + "epoch": 4.48, + "learning_rate": 4.556267867971931e-05, + "loss": 1.5731, + "step": 51720 + }, + { + "epoch": 4.48, + "learning_rate": 4.5561812353807506e-05, + "loss": 1.6414, + "step": 51730 + }, + { + "epoch": 4.48, + "learning_rate": 4.5560946027895694e-05, + "loss": 1.6498, + "step": 51740 + }, + { + "epoch": 4.48, + "learning_rate": 4.556007970198389e-05, + "loss": 1.6304, + "step": 51750 + }, + { + "epoch": 4.48, + "learning_rate": 4.555921337607208e-05, + "loss": 1.6408, + "step": 51760 + }, + { + "epoch": 4.48, + "learning_rate": 4.555834705016027e-05, + "loss": 1.632, + "step": 51770 + }, + { + "epoch": 4.48, + "learning_rate": 4.5557480724248465e-05, + "loss": 1.5842, + "step": 51780 + }, + { + "epoch": 4.48, + "learning_rate": 4.555661439833665e-05, + "loss": 1.6322, + "step": 51790 + }, + { + "epoch": 4.48, + "learning_rate": 4.555574807242485e-05, + "loss": 1.6016, + "step": 51800 + }, + { + "epoch": 4.48, + "learning_rate": 4.555488174651304e-05, + "loss": 1.6307, + "step": 51810 + }, + { + "epoch": 4.49, + "learning_rate": 4.555401542060123e-05, + "loss": 1.5675, + "step": 51820 + }, + { + "epoch": 4.49, + "learning_rate": 4.5553149094689424e-05, + "loss": 1.6513, + "step": 51830 + }, + { + "epoch": 4.49, + "learning_rate": 4.555228276877762e-05, + "loss": 1.7233, + "step": 51840 + }, + { + "epoch": 4.49, + "learning_rate": 4.5551416442865806e-05, + "loss": 1.5806, + "step": 51850 + }, + { + "epoch": 4.49, + "learning_rate": 4.5550550116954e-05, + "loss": 1.6239, + "step": 51860 + }, + { + "epoch": 4.49, + "learning_rate": 4.5549683791042195e-05, + "loss": 1.5874, + "step": 51870 + }, + { + "epoch": 4.49, + "learning_rate": 4.554881746513038e-05, + "loss": 1.5978, + "step": 51880 + }, + { + "epoch": 4.49, + "learning_rate": 4.554795113921858e-05, + "loss": 1.5279, + "step": 51890 + }, + { + "epoch": 4.49, + "learning_rate": 4.5547084813306764e-05, + "loss": 1.6576, + "step": 51900 + }, + { + "epoch": 4.49, + "learning_rate": 4.554621848739496e-05, + "loss": 1.615, + "step": 51910 + }, + { + "epoch": 4.49, + "learning_rate": 4.554535216148315e-05, + "loss": 1.6181, + "step": 51920 + }, + { + "epoch": 4.49, + "learning_rate": 4.554448583557134e-05, + "loss": 1.6518, + "step": 51930 + }, + { + "epoch": 4.5, + "learning_rate": 4.5543619509659535e-05, + "loss": 1.6342, + "step": 51940 + }, + { + "epoch": 4.5, + "learning_rate": 4.554275318374773e-05, + "loss": 1.5464, + "step": 51950 + }, + { + "epoch": 4.5, + "learning_rate": 4.554188685783592e-05, + "loss": 1.5939, + "step": 51960 + }, + { + "epoch": 4.5, + "learning_rate": 4.554102053192411e-05, + "loss": 1.7434, + "step": 51970 + }, + { + "epoch": 4.5, + "learning_rate": 4.5540154206012306e-05, + "loss": 1.5588, + "step": 51980 + }, + { + "epoch": 4.5, + "learning_rate": 4.5539287880100494e-05, + "loss": 1.5915, + "step": 51990 + }, + { + "epoch": 4.5, + "learning_rate": 4.553842155418869e-05, + "loss": 1.5949, + "step": 52000 + }, + { + "epoch": 4.5, + "learning_rate": 4.5537555228276876e-05, + "loss": 1.549, + "step": 52010 + }, + { + "epoch": 4.5, + "learning_rate": 4.553668890236507e-05, + "loss": 1.6497, + "step": 52020 + }, + { + "epoch": 4.5, + "learning_rate": 4.5535822576453265e-05, + "loss": 1.6868, + "step": 52030 + }, + { + "epoch": 4.5, + "learning_rate": 4.553495625054145e-05, + "loss": 1.6436, + "step": 52040 + }, + { + "epoch": 4.51, + "learning_rate": 4.553408992462965e-05, + "loss": 1.6375, + "step": 52050 + }, + { + "epoch": 4.51, + "learning_rate": 4.553322359871784e-05, + "loss": 1.6146, + "step": 52060 + }, + { + "epoch": 4.51, + "learning_rate": 4.553235727280603e-05, + "loss": 1.6208, + "step": 52070 + }, + { + "epoch": 4.51, + "learning_rate": 4.5531490946894224e-05, + "loss": 1.572, + "step": 52080 + }, + { + "epoch": 4.51, + "learning_rate": 4.553062462098242e-05, + "loss": 1.623, + "step": 52090 + }, + { + "epoch": 4.51, + "learning_rate": 4.5529758295070606e-05, + "loss": 1.5925, + "step": 52100 + }, + { + "epoch": 4.51, + "learning_rate": 4.55288919691588e-05, + "loss": 1.5732, + "step": 52110 + }, + { + "epoch": 4.51, + "learning_rate": 4.552802564324699e-05, + "loss": 1.599, + "step": 52120 + }, + { + "epoch": 4.51, + "learning_rate": 4.552715931733518e-05, + "loss": 1.5761, + "step": 52130 + }, + { + "epoch": 4.51, + "learning_rate": 4.552629299142338e-05, + "loss": 1.5081, + "step": 52140 + }, + { + "epoch": 4.51, + "learning_rate": 4.5525426665511565e-05, + "loss": 1.627, + "step": 52150 + }, + { + "epoch": 4.51, + "learning_rate": 4.552456033959976e-05, + "loss": 1.5747, + "step": 52160 + }, + { + "epoch": 4.52, + "learning_rate": 4.5523694013687954e-05, + "loss": 1.6336, + "step": 52170 + }, + { + "epoch": 4.52, + "learning_rate": 4.552282768777614e-05, + "loss": 1.6252, + "step": 52180 + }, + { + "epoch": 4.52, + "learning_rate": 4.5521961361864336e-05, + "loss": 1.5879, + "step": 52190 + }, + { + "epoch": 4.52, + "learning_rate": 4.5521095035952523e-05, + "loss": 1.6062, + "step": 52200 + }, + { + "epoch": 4.52, + "learning_rate": 4.552022871004072e-05, + "loss": 1.5844, + "step": 52210 + }, + { + "epoch": 4.52, + "learning_rate": 4.551936238412891e-05, + "loss": 1.5712, + "step": 52220 + }, + { + "epoch": 4.52, + "learning_rate": 4.55184960582171e-05, + "loss": 1.6471, + "step": 52230 + }, + { + "epoch": 4.52, + "learning_rate": 4.5517629732305294e-05, + "loss": 1.6335, + "step": 52240 + }, + { + "epoch": 4.52, + "learning_rate": 4.551676340639349e-05, + "loss": 1.6465, + "step": 52250 + }, + { + "epoch": 4.52, + "learning_rate": 4.5515897080481677e-05, + "loss": 1.6761, + "step": 52260 + }, + { + "epoch": 4.52, + "learning_rate": 4.551503075456987e-05, + "loss": 1.6286, + "step": 52270 + }, + { + "epoch": 4.53, + "learning_rate": 4.5514164428658065e-05, + "loss": 1.6502, + "step": 52280 + }, + { + "epoch": 4.53, + "learning_rate": 4.551329810274625e-05, + "loss": 1.6291, + "step": 52290 + }, + { + "epoch": 4.53, + "learning_rate": 4.551243177683445e-05, + "loss": 1.6323, + "step": 52300 + }, + { + "epoch": 4.53, + "learning_rate": 4.5511565450922635e-05, + "loss": 1.6564, + "step": 52310 + }, + { + "epoch": 4.53, + "learning_rate": 4.551069912501083e-05, + "loss": 1.6483, + "step": 52320 + }, + { + "epoch": 4.53, + "learning_rate": 4.5509832799099024e-05, + "loss": 1.5985, + "step": 52330 + }, + { + "epoch": 4.53, + "learning_rate": 4.550896647318721e-05, + "loss": 1.6664, + "step": 52340 + }, + { + "epoch": 4.53, + "learning_rate": 4.5508100147275406e-05, + "loss": 1.6133, + "step": 52350 + }, + { + "epoch": 4.53, + "learning_rate": 4.55072338213636e-05, + "loss": 1.629, + "step": 52360 + }, + { + "epoch": 4.53, + "learning_rate": 4.550636749545179e-05, + "loss": 1.6234, + "step": 52370 + }, + { + "epoch": 4.53, + "learning_rate": 4.550550116953998e-05, + "loss": 1.5807, + "step": 52380 + }, + { + "epoch": 4.53, + "learning_rate": 4.550463484362818e-05, + "loss": 1.611, + "step": 52390 + }, + { + "epoch": 4.54, + "learning_rate": 4.5503768517716365e-05, + "loss": 1.6636, + "step": 52400 + }, + { + "epoch": 4.54, + "learning_rate": 4.550290219180456e-05, + "loss": 1.6668, + "step": 52410 + }, + { + "epoch": 4.54, + "learning_rate": 4.550203586589275e-05, + "loss": 1.626, + "step": 52420 + }, + { + "epoch": 4.54, + "learning_rate": 4.550116953998094e-05, + "loss": 1.5695, + "step": 52430 + }, + { + "epoch": 4.54, + "learning_rate": 4.5500303214069136e-05, + "loss": 1.6233, + "step": 52440 + }, + { + "epoch": 4.54, + "learning_rate": 4.5499436888157324e-05, + "loss": 1.5352, + "step": 52450 + }, + { + "epoch": 4.54, + "learning_rate": 4.549857056224552e-05, + "loss": 1.5984, + "step": 52460 + }, + { + "epoch": 4.54, + "learning_rate": 4.549770423633371e-05, + "loss": 1.6365, + "step": 52470 + }, + { + "epoch": 4.54, + "learning_rate": 4.54968379104219e-05, + "loss": 1.613, + "step": 52480 + }, + { + "epoch": 4.54, + "learning_rate": 4.5495971584510095e-05, + "loss": 1.6378, + "step": 52490 + }, + { + "epoch": 4.54, + "learning_rate": 4.549510525859829e-05, + "loss": 1.5854, + "step": 52500 + }, + { + "epoch": 4.54, + "learning_rate": 4.549423893268648e-05, + "loss": 1.6089, + "step": 52510 + }, + { + "epoch": 4.55, + "learning_rate": 4.549337260677467e-05, + "loss": 1.7132, + "step": 52520 + }, + { + "epoch": 4.55, + "learning_rate": 4.549250628086286e-05, + "loss": 1.6233, + "step": 52530 + }, + { + "epoch": 4.55, + "learning_rate": 4.5491639954951054e-05, + "loss": 1.5558, + "step": 52540 + }, + { + "epoch": 4.55, + "learning_rate": 4.549077362903925e-05, + "loss": 1.6042, + "step": 52550 + }, + { + "epoch": 4.55, + "learning_rate": 4.5489907303127436e-05, + "loss": 1.6869, + "step": 52560 + }, + { + "epoch": 4.55, + "learning_rate": 4.548904097721563e-05, + "loss": 1.6906, + "step": 52570 + }, + { + "epoch": 4.55, + "learning_rate": 4.5488174651303825e-05, + "loss": 1.6225, + "step": 52580 + }, + { + "epoch": 4.55, + "learning_rate": 4.548730832539201e-05, + "loss": 1.5663, + "step": 52590 + }, + { + "epoch": 4.55, + "learning_rate": 4.548644199948021e-05, + "loss": 1.6757, + "step": 52600 + }, + { + "epoch": 4.55, + "learning_rate": 4.54855756735684e-05, + "loss": 1.6657, + "step": 52610 + }, + { + "epoch": 4.55, + "learning_rate": 4.548470934765659e-05, + "loss": 1.6495, + "step": 52620 + }, + { + "epoch": 4.56, + "learning_rate": 4.548384302174478e-05, + "loss": 1.5499, + "step": 52630 + }, + { + "epoch": 4.56, + "learning_rate": 4.548297669583297e-05, + "loss": 1.631, + "step": 52640 + }, + { + "epoch": 4.56, + "learning_rate": 4.5482110369921165e-05, + "loss": 1.6215, + "step": 52650 + }, + { + "epoch": 4.56, + "learning_rate": 4.548124404400936e-05, + "loss": 1.6103, + "step": 52660 + }, + { + "epoch": 4.56, + "learning_rate": 4.548037771809755e-05, + "loss": 1.6122, + "step": 52670 + }, + { + "epoch": 4.56, + "learning_rate": 4.547951139218574e-05, + "loss": 1.5948, + "step": 52680 + }, + { + "epoch": 4.56, + "learning_rate": 4.5478645066273936e-05, + "loss": 1.6046, + "step": 52690 + }, + { + "epoch": 4.56, + "learning_rate": 4.5477778740362124e-05, + "loss": 1.6301, + "step": 52700 + }, + { + "epoch": 4.56, + "learning_rate": 4.547691241445032e-05, + "loss": 1.6333, + "step": 52710 + }, + { + "epoch": 4.56, + "learning_rate": 4.547604608853851e-05, + "loss": 1.5762, + "step": 52720 + }, + { + "epoch": 4.56, + "learning_rate": 4.54751797626267e-05, + "loss": 1.6009, + "step": 52730 + }, + { + "epoch": 4.56, + "learning_rate": 4.5474313436714895e-05, + "loss": 1.6314, + "step": 52740 + }, + { + "epoch": 4.57, + "learning_rate": 4.547344711080308e-05, + "loss": 1.5559, + "step": 52750 + }, + { + "epoch": 4.57, + "learning_rate": 4.547258078489128e-05, + "loss": 1.586, + "step": 52760 + }, + { + "epoch": 4.57, + "learning_rate": 4.547171445897947e-05, + "loss": 1.6605, + "step": 52770 + }, + { + "epoch": 4.57, + "learning_rate": 4.547084813306766e-05, + "loss": 1.6188, + "step": 52780 + }, + { + "epoch": 4.57, + "learning_rate": 4.5469981807155854e-05, + "loss": 1.6364, + "step": 52790 + }, + { + "epoch": 4.57, + "learning_rate": 4.546911548124405e-05, + "loss": 1.5895, + "step": 52800 + }, + { + "epoch": 4.57, + "learning_rate": 4.5468249155332236e-05, + "loss": 1.5899, + "step": 52810 + }, + { + "epoch": 4.57, + "learning_rate": 4.546738282942043e-05, + "loss": 1.6147, + "step": 52820 + }, + { + "epoch": 4.57, + "learning_rate": 4.5466516503508625e-05, + "loss": 1.6218, + "step": 52830 + }, + { + "epoch": 4.57, + "learning_rate": 4.546565017759681e-05, + "loss": 1.6039, + "step": 52840 + }, + { + "epoch": 4.57, + "learning_rate": 4.546478385168501e-05, + "loss": 1.5526, + "step": 52850 + }, + { + "epoch": 4.58, + "learning_rate": 4.5463917525773195e-05, + "loss": 1.5388, + "step": 52860 + }, + { + "epoch": 4.58, + "learning_rate": 4.546305119986139e-05, + "loss": 1.553, + "step": 52870 + }, + { + "epoch": 4.58, + "learning_rate": 4.5462184873949584e-05, + "loss": 1.5932, + "step": 52880 + }, + { + "epoch": 4.58, + "learning_rate": 4.546131854803777e-05, + "loss": 1.6103, + "step": 52890 + }, + { + "epoch": 4.58, + "learning_rate": 4.5460452222125966e-05, + "loss": 1.6742, + "step": 52900 + }, + { + "epoch": 4.58, + "learning_rate": 4.545958589621416e-05, + "loss": 1.5971, + "step": 52910 + }, + { + "epoch": 4.58, + "learning_rate": 4.545871957030235e-05, + "loss": 1.6145, + "step": 52920 + }, + { + "epoch": 4.58, + "learning_rate": 4.545785324439054e-05, + "loss": 1.6266, + "step": 52930 + }, + { + "epoch": 4.58, + "learning_rate": 4.545698691847873e-05, + "loss": 1.6338, + "step": 52940 + }, + { + "epoch": 4.58, + "learning_rate": 4.5456120592566924e-05, + "loss": 1.6529, + "step": 52950 + }, + { + "epoch": 4.58, + "learning_rate": 4.545525426665512e-05, + "loss": 1.65, + "step": 52960 + }, + { + "epoch": 4.58, + "learning_rate": 4.5454387940743307e-05, + "loss": 1.6508, + "step": 52970 + }, + { + "epoch": 4.59, + "learning_rate": 4.54535216148315e-05, + "loss": 1.5761, + "step": 52980 + }, + { + "epoch": 4.59, + "learning_rate": 4.5452655288919695e-05, + "loss": 1.6139, + "step": 52990 + }, + { + "epoch": 4.59, + "learning_rate": 4.545178896300788e-05, + "loss": 1.6141, + "step": 53000 + }, + { + "epoch": 4.59, + "learning_rate": 4.545092263709608e-05, + "loss": 1.5464, + "step": 53010 + }, + { + "epoch": 4.59, + "learning_rate": 4.545005631118427e-05, + "loss": 1.5481, + "step": 53020 + }, + { + "epoch": 4.59, + "learning_rate": 4.544918998527246e-05, + "loss": 1.6866, + "step": 53030 + }, + { + "epoch": 4.59, + "learning_rate": 4.5448323659360654e-05, + "loss": 1.6031, + "step": 53040 + }, + { + "epoch": 4.59, + "learning_rate": 4.544745733344884e-05, + "loss": 1.6077, + "step": 53050 + }, + { + "epoch": 4.59, + "learning_rate": 4.5446591007537036e-05, + "loss": 1.5953, + "step": 53060 + }, + { + "epoch": 4.59, + "learning_rate": 4.544572468162523e-05, + "loss": 1.5979, + "step": 53070 + }, + { + "epoch": 4.59, + "learning_rate": 4.544485835571342e-05, + "loss": 1.5923, + "step": 53080 + }, + { + "epoch": 4.6, + "learning_rate": 4.544399202980161e-05, + "loss": 1.6637, + "step": 53090 + }, + { + "epoch": 4.6, + "learning_rate": 4.544312570388981e-05, + "loss": 1.6624, + "step": 53100 + }, + { + "epoch": 4.6, + "learning_rate": 4.5442259377977995e-05, + "loss": 1.5984, + "step": 53110 + }, + { + "epoch": 4.6, + "learning_rate": 4.544139305206619e-05, + "loss": 1.6956, + "step": 53120 + }, + { + "epoch": 4.6, + "learning_rate": 4.5440526726154384e-05, + "loss": 1.584, + "step": 53130 + }, + { + "epoch": 4.6, + "learning_rate": 4.543966040024257e-05, + "loss": 1.6289, + "step": 53140 + }, + { + "epoch": 4.6, + "learning_rate": 4.5438794074330766e-05, + "loss": 1.6034, + "step": 53150 + }, + { + "epoch": 4.6, + "learning_rate": 4.5437927748418954e-05, + "loss": 1.6417, + "step": 53160 + }, + { + "epoch": 4.6, + "learning_rate": 4.543706142250715e-05, + "loss": 1.6337, + "step": 53170 + }, + { + "epoch": 4.6, + "learning_rate": 4.543619509659534e-05, + "loss": 1.6207, + "step": 53180 + }, + { + "epoch": 4.6, + "learning_rate": 4.543532877068353e-05, + "loss": 1.6288, + "step": 53190 + }, + { + "epoch": 4.6, + "learning_rate": 4.5434462444771725e-05, + "loss": 1.5786, + "step": 53200 + }, + { + "epoch": 4.61, + "learning_rate": 4.543359611885992e-05, + "loss": 1.6237, + "step": 53210 + }, + { + "epoch": 4.61, + "learning_rate": 4.543272979294811e-05, + "loss": 1.6151, + "step": 53220 + }, + { + "epoch": 4.61, + "learning_rate": 4.54318634670363e-05, + "loss": 1.5589, + "step": 53230 + }, + { + "epoch": 4.61, + "learning_rate": 4.5430997141124496e-05, + "loss": 1.6237, + "step": 53240 + }, + { + "epoch": 4.61, + "learning_rate": 4.5430130815212683e-05, + "loss": 1.5991, + "step": 53250 + }, + { + "epoch": 4.61, + "learning_rate": 4.542926448930088e-05, + "loss": 1.6579, + "step": 53260 + }, + { + "epoch": 4.61, + "learning_rate": 4.5428398163389066e-05, + "loss": 1.615, + "step": 53270 + }, + { + "epoch": 4.61, + "learning_rate": 4.542753183747726e-05, + "loss": 1.6082, + "step": 53280 + }, + { + "epoch": 4.61, + "learning_rate": 4.5426665511565455e-05, + "loss": 1.6451, + "step": 53290 + }, + { + "epoch": 4.61, + "learning_rate": 4.542579918565364e-05, + "loss": 1.6209, + "step": 53300 + }, + { + "epoch": 4.61, + "learning_rate": 4.542493285974184e-05, + "loss": 1.7011, + "step": 53310 + }, + { + "epoch": 4.62, + "learning_rate": 4.542406653383003e-05, + "loss": 1.6056, + "step": 53320 + }, + { + "epoch": 4.62, + "learning_rate": 4.542320020791822e-05, + "loss": 1.6065, + "step": 53330 + }, + { + "epoch": 4.62, + "learning_rate": 4.542233388200641e-05, + "loss": 1.6683, + "step": 53340 + }, + { + "epoch": 4.62, + "learning_rate": 4.542146755609461e-05, + "loss": 1.6718, + "step": 53350 + }, + { + "epoch": 4.62, + "learning_rate": 4.5420601230182795e-05, + "loss": 1.6716, + "step": 53360 + }, + { + "epoch": 4.62, + "learning_rate": 4.541973490427099e-05, + "loss": 1.5844, + "step": 53370 + }, + { + "epoch": 4.62, + "learning_rate": 4.541886857835918e-05, + "loss": 1.6443, + "step": 53380 + }, + { + "epoch": 4.62, + "learning_rate": 4.541800225244737e-05, + "loss": 1.6201, + "step": 53390 + }, + { + "epoch": 4.62, + "learning_rate": 4.5417135926535566e-05, + "loss": 1.6176, + "step": 53400 + }, + { + "epoch": 4.62, + "learning_rate": 4.5416269600623754e-05, + "loss": 1.6852, + "step": 53410 + }, + { + "epoch": 4.62, + "learning_rate": 4.541540327471195e-05, + "loss": 1.6086, + "step": 53420 + }, + { + "epoch": 4.62, + "learning_rate": 4.541453694880014e-05, + "loss": 1.6114, + "step": 53430 + }, + { + "epoch": 4.63, + "learning_rate": 4.541367062288833e-05, + "loss": 1.6143, + "step": 53440 + }, + { + "epoch": 4.63, + "learning_rate": 4.5412804296976525e-05, + "loss": 1.5905, + "step": 53450 + }, + { + "epoch": 4.63, + "learning_rate": 4.541193797106472e-05, + "loss": 1.6252, + "step": 53460 + }, + { + "epoch": 4.63, + "learning_rate": 4.541107164515291e-05, + "loss": 1.5463, + "step": 53470 + }, + { + "epoch": 4.63, + "learning_rate": 4.54102053192411e-05, + "loss": 1.6175, + "step": 53480 + }, + { + "epoch": 4.63, + "learning_rate": 4.540933899332929e-05, + "loss": 1.5564, + "step": 53490 + }, + { + "epoch": 4.63, + "learning_rate": 4.5408472667417484e-05, + "loss": 1.6098, + "step": 53500 + }, + { + "epoch": 4.63, + "learning_rate": 4.540760634150568e-05, + "loss": 1.5797, + "step": 53510 + }, + { + "epoch": 4.63, + "learning_rate": 4.5406740015593866e-05, + "loss": 1.6132, + "step": 53520 + }, + { + "epoch": 4.63, + "learning_rate": 4.540587368968206e-05, + "loss": 1.6167, + "step": 53530 + }, + { + "epoch": 4.63, + "learning_rate": 4.5405007363770255e-05, + "loss": 1.6351, + "step": 53540 + }, + { + "epoch": 4.63, + "learning_rate": 4.540414103785844e-05, + "loss": 1.5729, + "step": 53550 + }, + { + "epoch": 4.64, + "learning_rate": 4.540327471194664e-05, + "loss": 1.6273, + "step": 53560 + }, + { + "epoch": 4.64, + "learning_rate": 4.540240838603483e-05, + "loss": 1.6315, + "step": 53570 + }, + { + "epoch": 4.64, + "learning_rate": 4.540154206012302e-05, + "loss": 1.609, + "step": 53580 + }, + { + "epoch": 4.64, + "learning_rate": 4.5400675734211214e-05, + "loss": 1.5742, + "step": 53590 + }, + { + "epoch": 4.64, + "learning_rate": 4.53998094082994e-05, + "loss": 1.5514, + "step": 53600 + }, + { + "epoch": 4.64, + "learning_rate": 4.5398943082387596e-05, + "loss": 1.6113, + "step": 53610 + }, + { + "epoch": 4.64, + "learning_rate": 4.539807675647579e-05, + "loss": 1.616, + "step": 53620 + }, + { + "epoch": 4.64, + "learning_rate": 4.539721043056398e-05, + "loss": 1.5687, + "step": 53630 + }, + { + "epoch": 4.64, + "learning_rate": 4.539634410465217e-05, + "loss": 1.5822, + "step": 53640 + }, + { + "epoch": 4.64, + "learning_rate": 4.539547777874037e-05, + "loss": 1.5926, + "step": 53650 + }, + { + "epoch": 4.64, + "learning_rate": 4.5394611452828554e-05, + "loss": 1.6518, + "step": 53660 + }, + { + "epoch": 4.65, + "learning_rate": 4.539374512691675e-05, + "loss": 1.5719, + "step": 53670 + }, + { + "epoch": 4.65, + "learning_rate": 4.539287880100494e-05, + "loss": 1.7094, + "step": 53680 + }, + { + "epoch": 4.65, + "learning_rate": 4.539201247509313e-05, + "loss": 1.584, + "step": 53690 + }, + { + "epoch": 4.65, + "learning_rate": 4.5391146149181325e-05, + "loss": 1.683, + "step": 53700 + }, + { + "epoch": 4.65, + "learning_rate": 4.539027982326951e-05, + "loss": 1.6176, + "step": 53710 + }, + { + "epoch": 4.65, + "learning_rate": 4.538941349735771e-05, + "loss": 1.572, + "step": 53720 + }, + { + "epoch": 4.65, + "learning_rate": 4.53885471714459e-05, + "loss": 1.6749, + "step": 53730 + }, + { + "epoch": 4.65, + "learning_rate": 4.538768084553409e-05, + "loss": 1.6346, + "step": 53740 + }, + { + "epoch": 4.65, + "learning_rate": 4.5386814519622284e-05, + "loss": 1.5518, + "step": 53750 + }, + { + "epoch": 4.65, + "learning_rate": 4.538594819371048e-05, + "loss": 1.6314, + "step": 53760 + }, + { + "epoch": 4.65, + "learning_rate": 4.5385081867798666e-05, + "loss": 1.5901, + "step": 53770 + }, + { + "epoch": 4.65, + "learning_rate": 4.538421554188686e-05, + "loss": 1.6242, + "step": 53780 + }, + { + "epoch": 4.66, + "learning_rate": 4.538334921597505e-05, + "loss": 1.5855, + "step": 53790 + }, + { + "epoch": 4.66, + "learning_rate": 4.538248289006324e-05, + "loss": 1.6144, + "step": 53800 + }, + { + "epoch": 4.66, + "learning_rate": 4.538161656415144e-05, + "loss": 1.6159, + "step": 53810 + }, + { + "epoch": 4.66, + "learning_rate": 4.5380750238239625e-05, + "loss": 1.5737, + "step": 53820 + }, + { + "epoch": 4.66, + "learning_rate": 4.537988391232782e-05, + "loss": 1.5327, + "step": 53830 + }, + { + "epoch": 4.66, + "learning_rate": 4.5379017586416014e-05, + "loss": 1.6103, + "step": 53840 + }, + { + "epoch": 4.66, + "learning_rate": 4.53781512605042e-05, + "loss": 1.6388, + "step": 53850 + }, + { + "epoch": 4.66, + "learning_rate": 4.5377284934592396e-05, + "loss": 1.615, + "step": 53860 + }, + { + "epoch": 4.66, + "learning_rate": 4.537641860868059e-05, + "loss": 1.6522, + "step": 53870 + }, + { + "epoch": 4.66, + "learning_rate": 4.537555228276878e-05, + "loss": 1.5945, + "step": 53880 + }, + { + "epoch": 4.66, + "learning_rate": 4.537468595685697e-05, + "loss": 1.6469, + "step": 53890 + }, + { + "epoch": 4.67, + "learning_rate": 4.537381963094516e-05, + "loss": 1.5323, + "step": 53900 + }, + { + "epoch": 4.67, + "learning_rate": 4.5372953305033355e-05, + "loss": 1.6228, + "step": 53910 + }, + { + "epoch": 4.67, + "learning_rate": 4.537208697912155e-05, + "loss": 1.6543, + "step": 53920 + }, + { + "epoch": 4.67, + "learning_rate": 4.537122065320974e-05, + "loss": 1.6494, + "step": 53930 + }, + { + "epoch": 4.67, + "learning_rate": 4.537035432729793e-05, + "loss": 1.6413, + "step": 53940 + }, + { + "epoch": 4.67, + "learning_rate": 4.5369488001386126e-05, + "loss": 1.5684, + "step": 53950 + }, + { + "epoch": 4.67, + "learning_rate": 4.5368621675474313e-05, + "loss": 1.5508, + "step": 53960 + }, + { + "epoch": 4.67, + "learning_rate": 4.536775534956251e-05, + "loss": 1.5202, + "step": 53970 + }, + { + "epoch": 4.67, + "learning_rate": 4.53668890236507e-05, + "loss": 1.5998, + "step": 53980 + }, + { + "epoch": 4.67, + "learning_rate": 4.536602269773889e-05, + "loss": 1.5695, + "step": 53990 + }, + { + "epoch": 4.67, + "learning_rate": 4.5365156371827084e-05, + "loss": 1.5727, + "step": 54000 + }, + { + "epoch": 4.67, + "learning_rate": 4.536429004591527e-05, + "loss": 1.5253, + "step": 54010 + }, + { + "epoch": 4.68, + "learning_rate": 4.536342372000347e-05, + "loss": 1.5383, + "step": 54020 + }, + { + "epoch": 4.68, + "learning_rate": 4.536255739409166e-05, + "loss": 1.5602, + "step": 54030 + }, + { + "epoch": 4.68, + "learning_rate": 4.536169106817985e-05, + "loss": 1.5845, + "step": 54040 + }, + { + "epoch": 4.68, + "learning_rate": 4.536082474226804e-05, + "loss": 1.5693, + "step": 54050 + }, + { + "epoch": 4.68, + "learning_rate": 4.535995841635624e-05, + "loss": 1.5748, + "step": 54060 + }, + { + "epoch": 4.68, + "learning_rate": 4.5359092090444425e-05, + "loss": 1.6307, + "step": 54070 + }, + { + "epoch": 4.68, + "learning_rate": 4.535822576453262e-05, + "loss": 1.6009, + "step": 54080 + }, + { + "epoch": 4.68, + "learning_rate": 4.5357359438620814e-05, + "loss": 1.5994, + "step": 54090 + }, + { + "epoch": 4.68, + "learning_rate": 4.5356493112709e-05, + "loss": 1.6046, + "step": 54100 + }, + { + "epoch": 4.68, + "learning_rate": 4.5355626786797196e-05, + "loss": 1.6193, + "step": 54110 + }, + { + "epoch": 4.68, + "learning_rate": 4.5354760460885384e-05, + "loss": 1.6218, + "step": 54120 + }, + { + "epoch": 4.69, + "learning_rate": 4.535389413497358e-05, + "loss": 1.5906, + "step": 54130 + }, + { + "epoch": 4.69, + "learning_rate": 4.535302780906177e-05, + "loss": 1.606, + "step": 54140 + }, + { + "epoch": 4.69, + "learning_rate": 4.535216148314996e-05, + "loss": 1.5563, + "step": 54150 + }, + { + "epoch": 4.69, + "learning_rate": 4.5351295157238155e-05, + "loss": 1.558, + "step": 54160 + }, + { + "epoch": 4.69, + "learning_rate": 4.535042883132635e-05, + "loss": 1.6173, + "step": 54170 + }, + { + "epoch": 4.69, + "learning_rate": 4.534956250541454e-05, + "loss": 1.6146, + "step": 54180 + }, + { + "epoch": 4.69, + "learning_rate": 4.534869617950273e-05, + "loss": 1.7119, + "step": 54190 + }, + { + "epoch": 4.69, + "learning_rate": 4.5347829853590926e-05, + "loss": 1.5849, + "step": 54200 + }, + { + "epoch": 4.69, + "learning_rate": 4.5346963527679114e-05, + "loss": 1.614, + "step": 54210 + }, + { + "epoch": 4.69, + "learning_rate": 4.534609720176731e-05, + "loss": 1.6158, + "step": 54220 + }, + { + "epoch": 4.69, + "learning_rate": 4.5345230875855496e-05, + "loss": 1.6398, + "step": 54230 + }, + { + "epoch": 4.69, + "learning_rate": 4.534436454994369e-05, + "loss": 1.5721, + "step": 54240 + }, + { + "epoch": 4.7, + "learning_rate": 4.5343498224031885e-05, + "loss": 1.679, + "step": 54250 + }, + { + "epoch": 4.7, + "learning_rate": 4.534263189812007e-05, + "loss": 1.6276, + "step": 54260 + }, + { + "epoch": 4.7, + "learning_rate": 4.534176557220827e-05, + "loss": 1.6268, + "step": 54270 + }, + { + "epoch": 4.7, + "learning_rate": 4.534089924629646e-05, + "loss": 1.5961, + "step": 54280 + }, + { + "epoch": 4.7, + "learning_rate": 4.534003292038465e-05, + "loss": 1.5629, + "step": 54290 + }, + { + "epoch": 4.7, + "learning_rate": 4.5339166594472844e-05, + "loss": 1.6221, + "step": 54300 + }, + { + "epoch": 4.7, + "learning_rate": 4.533830026856104e-05, + "loss": 1.5718, + "step": 54310 + }, + { + "epoch": 4.7, + "learning_rate": 4.5337433942649226e-05, + "loss": 1.5814, + "step": 54320 + }, + { + "epoch": 4.7, + "learning_rate": 4.533656761673742e-05, + "loss": 1.5386, + "step": 54330 + }, + { + "epoch": 4.7, + "learning_rate": 4.533570129082561e-05, + "loss": 1.5897, + "step": 54340 + }, + { + "epoch": 4.7, + "learning_rate": 4.53348349649138e-05, + "loss": 1.6295, + "step": 54350 + }, + { + "epoch": 4.71, + "learning_rate": 4.5333968639002e-05, + "loss": 1.6031, + "step": 54360 + }, + { + "epoch": 4.71, + "learning_rate": 4.5333102313090184e-05, + "loss": 1.6147, + "step": 54370 + }, + { + "epoch": 4.71, + "learning_rate": 4.533223598717838e-05, + "loss": 1.6318, + "step": 54380 + }, + { + "epoch": 4.71, + "learning_rate": 4.533136966126657e-05, + "loss": 1.6969, + "step": 54390 + }, + { + "epoch": 4.71, + "learning_rate": 4.533050333535476e-05, + "loss": 1.6125, + "step": 54400 + }, + { + "epoch": 4.71, + "learning_rate": 4.5329637009442955e-05, + "loss": 1.5904, + "step": 54410 + }, + { + "epoch": 4.71, + "learning_rate": 4.532877068353115e-05, + "loss": 1.6269, + "step": 54420 + }, + { + "epoch": 4.71, + "learning_rate": 4.532790435761934e-05, + "loss": 1.6211, + "step": 54430 + }, + { + "epoch": 4.71, + "learning_rate": 4.532703803170753e-05, + "loss": 1.6103, + "step": 54440 + }, + { + "epoch": 4.71, + "learning_rate": 4.532617170579572e-05, + "loss": 1.6376, + "step": 54450 + }, + { + "epoch": 4.71, + "learning_rate": 4.5325305379883914e-05, + "loss": 1.5626, + "step": 54460 + }, + { + "epoch": 4.71, + "learning_rate": 4.532443905397211e-05, + "loss": 1.6066, + "step": 54470 + }, + { + "epoch": 4.72, + "learning_rate": 4.5323572728060296e-05, + "loss": 1.5765, + "step": 54480 + }, + { + "epoch": 4.72, + "learning_rate": 4.532270640214849e-05, + "loss": 1.6802, + "step": 54490 + }, + { + "epoch": 4.72, + "learning_rate": 4.5321840076236685e-05, + "loss": 1.5933, + "step": 54500 + }, + { + "epoch": 4.72, + "learning_rate": 4.532097375032487e-05, + "loss": 1.6476, + "step": 54510 + }, + { + "epoch": 4.72, + "learning_rate": 4.532010742441307e-05, + "loss": 1.6043, + "step": 54520 + }, + { + "epoch": 4.72, + "learning_rate": 4.5319241098501255e-05, + "loss": 1.6205, + "step": 54530 + }, + { + "epoch": 4.72, + "learning_rate": 4.531837477258945e-05, + "loss": 1.5968, + "step": 54540 + }, + { + "epoch": 4.72, + "learning_rate": 4.5317508446677644e-05, + "loss": 1.662, + "step": 54550 + }, + { + "epoch": 4.72, + "learning_rate": 4.531664212076583e-05, + "loss": 1.6056, + "step": 54560 + }, + { + "epoch": 4.72, + "learning_rate": 4.5315775794854026e-05, + "loss": 1.5633, + "step": 54570 + }, + { + "epoch": 4.72, + "learning_rate": 4.531490946894222e-05, + "loss": 1.5687, + "step": 54580 + }, + { + "epoch": 4.72, + "learning_rate": 4.531404314303041e-05, + "loss": 1.6441, + "step": 54590 + }, + { + "epoch": 4.73, + "learning_rate": 4.53131768171186e-05, + "loss": 1.578, + "step": 54600 + }, + { + "epoch": 4.73, + "learning_rate": 4.53123104912068e-05, + "loss": 1.5733, + "step": 54610 + }, + { + "epoch": 4.73, + "learning_rate": 4.5311444165294985e-05, + "loss": 1.5515, + "step": 54620 + }, + { + "epoch": 4.73, + "learning_rate": 4.531057783938318e-05, + "loss": 1.6227, + "step": 54630 + }, + { + "epoch": 4.73, + "learning_rate": 4.530971151347137e-05, + "loss": 1.5894, + "step": 54640 + }, + { + "epoch": 4.73, + "learning_rate": 4.530884518755956e-05, + "loss": 1.642, + "step": 54650 + }, + { + "epoch": 4.73, + "learning_rate": 4.5307978861647756e-05, + "loss": 1.5993, + "step": 54660 + }, + { + "epoch": 4.73, + "learning_rate": 4.5307112535735943e-05, + "loss": 1.6436, + "step": 54670 + }, + { + "epoch": 4.73, + "learning_rate": 4.530624620982414e-05, + "loss": 1.5881, + "step": 54680 + }, + { + "epoch": 4.73, + "learning_rate": 4.530537988391233e-05, + "loss": 1.6625, + "step": 54690 + }, + { + "epoch": 4.73, + "learning_rate": 4.530451355800052e-05, + "loss": 1.6023, + "step": 54700 + }, + { + "epoch": 4.74, + "learning_rate": 4.5303647232088714e-05, + "loss": 1.6369, + "step": 54710 + }, + { + "epoch": 4.74, + "learning_rate": 4.530278090617691e-05, + "loss": 1.6161, + "step": 54720 + }, + { + "epoch": 4.74, + "learning_rate": 4.5301914580265097e-05, + "loss": 1.5836, + "step": 54730 + }, + { + "epoch": 4.74, + "learning_rate": 4.530104825435329e-05, + "loss": 1.5963, + "step": 54740 + }, + { + "epoch": 4.74, + "learning_rate": 4.530018192844148e-05, + "loss": 1.6665, + "step": 54750 + }, + { + "epoch": 4.74, + "learning_rate": 4.529931560252967e-05, + "loss": 1.6413, + "step": 54760 + }, + { + "epoch": 4.74, + "learning_rate": 4.529844927661787e-05, + "loss": 1.6254, + "step": 54770 + }, + { + "epoch": 4.74, + "learning_rate": 4.5297582950706055e-05, + "loss": 1.5843, + "step": 54780 + }, + { + "epoch": 4.74, + "learning_rate": 4.529671662479425e-05, + "loss": 1.5761, + "step": 54790 + }, + { + "epoch": 4.74, + "learning_rate": 4.5295850298882444e-05, + "loss": 1.5671, + "step": 54800 + }, + { + "epoch": 4.74, + "learning_rate": 4.529498397297063e-05, + "loss": 1.5664, + "step": 54810 + }, + { + "epoch": 4.74, + "learning_rate": 4.5294117647058826e-05, + "loss": 1.5698, + "step": 54820 + }, + { + "epoch": 4.75, + "learning_rate": 4.529325132114702e-05, + "loss": 1.5564, + "step": 54830 + }, + { + "epoch": 4.75, + "learning_rate": 4.529238499523521e-05, + "loss": 1.6487, + "step": 54840 + }, + { + "epoch": 4.75, + "learning_rate": 4.52915186693234e-05, + "loss": 1.6041, + "step": 54850 + }, + { + "epoch": 4.75, + "learning_rate": 4.529065234341159e-05, + "loss": 1.646, + "step": 54860 + }, + { + "epoch": 4.75, + "learning_rate": 4.5289786017499785e-05, + "loss": 1.6038, + "step": 54870 + }, + { + "epoch": 4.75, + "learning_rate": 4.528891969158798e-05, + "loss": 1.5928, + "step": 54880 + }, + { + "epoch": 4.75, + "learning_rate": 4.528805336567617e-05, + "loss": 1.6459, + "step": 54890 + }, + { + "epoch": 4.75, + "learning_rate": 4.528718703976436e-05, + "loss": 1.6215, + "step": 54900 + }, + { + "epoch": 4.75, + "learning_rate": 4.5286320713852556e-05, + "loss": 1.5815, + "step": 54910 + }, + { + "epoch": 4.75, + "learning_rate": 4.5285454387940744e-05, + "loss": 1.6006, + "step": 54920 + }, + { + "epoch": 4.75, + "learning_rate": 4.528458806202894e-05, + "loss": 1.5773, + "step": 54930 + }, + { + "epoch": 4.76, + "learning_rate": 4.528372173611713e-05, + "loss": 1.6148, + "step": 54940 + }, + { + "epoch": 4.76, + "learning_rate": 4.528285541020532e-05, + "loss": 1.6679, + "step": 54950 + }, + { + "epoch": 4.76, + "learning_rate": 4.5281989084293515e-05, + "loss": 1.6383, + "step": 54960 + }, + { + "epoch": 4.76, + "learning_rate": 4.52811227583817e-05, + "loss": 1.5742, + "step": 54970 + }, + { + "epoch": 4.76, + "learning_rate": 4.52802564324699e-05, + "loss": 1.556, + "step": 54980 + }, + { + "epoch": 4.76, + "learning_rate": 4.527939010655809e-05, + "loss": 1.646, + "step": 54990 + }, + { + "epoch": 4.76, + "learning_rate": 4.527852378064628e-05, + "loss": 1.6178, + "step": 55000 + }, + { + "epoch": 4.76, + "learning_rate": 4.5277657454734474e-05, + "loss": 1.6119, + "step": 55010 + }, + { + "epoch": 4.76, + "learning_rate": 4.527679112882267e-05, + "loss": 1.6088, + "step": 55020 + }, + { + "epoch": 4.76, + "learning_rate": 4.5275924802910856e-05, + "loss": 1.5767, + "step": 55030 + }, + { + "epoch": 4.76, + "learning_rate": 4.527505847699905e-05, + "loss": 1.5623, + "step": 55040 + }, + { + "epoch": 4.76, + "learning_rate": 4.5274192151087245e-05, + "loss": 1.5695, + "step": 55050 + }, + { + "epoch": 4.77, + "learning_rate": 4.527332582517543e-05, + "loss": 1.6084, + "step": 55060 + }, + { + "epoch": 4.77, + "learning_rate": 4.527245949926363e-05, + "loss": 1.6017, + "step": 55070 + }, + { + "epoch": 4.77, + "learning_rate": 4.5271593173351814e-05, + "loss": 1.5792, + "step": 55080 + }, + { + "epoch": 4.77, + "learning_rate": 4.527072684744001e-05, + "loss": 1.6674, + "step": 55090 + }, + { + "epoch": 4.77, + "learning_rate": 4.52698605215282e-05, + "loss": 1.5614, + "step": 55100 + }, + { + "epoch": 4.77, + "learning_rate": 4.526899419561639e-05, + "loss": 1.648, + "step": 55110 + }, + { + "epoch": 4.77, + "learning_rate": 4.5268127869704585e-05, + "loss": 1.6931, + "step": 55120 + }, + { + "epoch": 4.77, + "learning_rate": 4.526726154379278e-05, + "loss": 1.5481, + "step": 55130 + }, + { + "epoch": 4.77, + "learning_rate": 4.526639521788097e-05, + "loss": 1.6406, + "step": 55140 + }, + { + "epoch": 4.77, + "learning_rate": 4.526552889196916e-05, + "loss": 1.5836, + "step": 55150 + }, + { + "epoch": 4.77, + "learning_rate": 4.5264662566057356e-05, + "loss": 1.5933, + "step": 55160 + }, + { + "epoch": 4.78, + "learning_rate": 4.5263796240145544e-05, + "loss": 1.5645, + "step": 55170 + }, + { + "epoch": 4.78, + "learning_rate": 4.526292991423374e-05, + "loss": 1.6216, + "step": 55180 + }, + { + "epoch": 4.78, + "learning_rate": 4.5262063588321926e-05, + "loss": 1.6471, + "step": 55190 + }, + { + "epoch": 4.78, + "learning_rate": 4.526119726241012e-05, + "loss": 1.5571, + "step": 55200 + }, + { + "epoch": 4.78, + "learning_rate": 4.5260330936498315e-05, + "loss": 1.6048, + "step": 55210 + }, + { + "epoch": 4.78, + "learning_rate": 4.52594646105865e-05, + "loss": 1.6008, + "step": 55220 + }, + { + "epoch": 4.78, + "learning_rate": 4.52585982846747e-05, + "loss": 1.5962, + "step": 55230 + }, + { + "epoch": 4.78, + "learning_rate": 4.525773195876289e-05, + "loss": 1.6565, + "step": 55240 + }, + { + "epoch": 4.78, + "learning_rate": 4.525686563285108e-05, + "loss": 1.6402, + "step": 55250 + }, + { + "epoch": 4.78, + "learning_rate": 4.5255999306939274e-05, + "loss": 1.5493, + "step": 55260 + }, + { + "epoch": 4.78, + "learning_rate": 4.525513298102746e-05, + "loss": 1.5543, + "step": 55270 + }, + { + "epoch": 4.78, + "learning_rate": 4.5254266655115656e-05, + "loss": 1.6118, + "step": 55280 + }, + { + "epoch": 4.79, + "learning_rate": 4.525340032920385e-05, + "loss": 1.5964, + "step": 55290 + }, + { + "epoch": 4.79, + "learning_rate": 4.525253400329204e-05, + "loss": 1.531, + "step": 55300 + }, + { + "epoch": 4.79, + "learning_rate": 4.525166767738023e-05, + "loss": 1.6179, + "step": 55310 + }, + { + "epoch": 4.79, + "learning_rate": 4.525080135146843e-05, + "loss": 1.5836, + "step": 55320 + }, + { + "epoch": 4.79, + "learning_rate": 4.5249935025556615e-05, + "loss": 1.5671, + "step": 55330 + }, + { + "epoch": 4.79, + "learning_rate": 4.524906869964481e-05, + "loss": 1.5408, + "step": 55340 + }, + { + "epoch": 4.79, + "learning_rate": 4.5248202373733004e-05, + "loss": 1.5522, + "step": 55350 + }, + { + "epoch": 4.79, + "learning_rate": 4.524733604782119e-05, + "loss": 1.64, + "step": 55360 + }, + { + "epoch": 4.79, + "learning_rate": 4.5246469721909386e-05, + "loss": 1.6261, + "step": 55370 + }, + { + "epoch": 4.79, + "learning_rate": 4.5245603395997573e-05, + "loss": 1.6253, + "step": 55380 + }, + { + "epoch": 4.79, + "learning_rate": 4.524473707008577e-05, + "loss": 1.58, + "step": 55390 + }, + { + "epoch": 4.8, + "learning_rate": 4.524387074417396e-05, + "loss": 1.6143, + "step": 55400 + }, + { + "epoch": 4.8, + "learning_rate": 4.524300441826215e-05, + "loss": 1.6264, + "step": 55410 + }, + { + "epoch": 4.8, + "learning_rate": 4.5242138092350344e-05, + "loss": 1.6351, + "step": 55420 + }, + { + "epoch": 4.8, + "learning_rate": 4.524127176643854e-05, + "loss": 1.6978, + "step": 55430 + }, + { + "epoch": 4.8, + "learning_rate": 4.5240405440526727e-05, + "loss": 1.6667, + "step": 55440 + }, + { + "epoch": 4.8, + "learning_rate": 4.523953911461492e-05, + "loss": 1.5347, + "step": 55450 + }, + { + "epoch": 4.8, + "learning_rate": 4.5238672788703115e-05, + "loss": 1.6233, + "step": 55460 + }, + { + "epoch": 4.8, + "learning_rate": 4.52378064627913e-05, + "loss": 1.578, + "step": 55470 + }, + { + "epoch": 4.8, + "learning_rate": 4.52369401368795e-05, + "loss": 1.6105, + "step": 55480 + }, + { + "epoch": 4.8, + "learning_rate": 4.5236073810967685e-05, + "loss": 1.5694, + "step": 55490 + }, + { + "epoch": 4.8, + "learning_rate": 4.523520748505588e-05, + "loss": 1.5986, + "step": 55500 + }, + { + "epoch": 4.8, + "learning_rate": 4.5234341159144074e-05, + "loss": 1.5801, + "step": 55510 + }, + { + "epoch": 4.81, + "learning_rate": 4.523347483323226e-05, + "loss": 1.5956, + "step": 55520 + }, + { + "epoch": 4.81, + "learning_rate": 4.5232608507320456e-05, + "loss": 1.5714, + "step": 55530 + }, + { + "epoch": 4.81, + "learning_rate": 4.523174218140865e-05, + "loss": 1.6064, + "step": 55540 + }, + { + "epoch": 4.81, + "learning_rate": 4.523087585549684e-05, + "loss": 1.6102, + "step": 55550 + }, + { + "epoch": 4.81, + "learning_rate": 4.523000952958503e-05, + "loss": 1.5747, + "step": 55560 + }, + { + "epoch": 4.81, + "learning_rate": 4.522914320367323e-05, + "loss": 1.6243, + "step": 55570 + }, + { + "epoch": 4.81, + "learning_rate": 4.5228276877761415e-05, + "loss": 1.5789, + "step": 55580 + }, + { + "epoch": 4.81, + "learning_rate": 4.522741055184961e-05, + "loss": 1.6259, + "step": 55590 + }, + { + "epoch": 4.81, + "learning_rate": 4.52265442259378e-05, + "loss": 1.5982, + "step": 55600 + }, + { + "epoch": 4.81, + "learning_rate": 4.522567790002599e-05, + "loss": 1.5733, + "step": 55610 + }, + { + "epoch": 4.81, + "learning_rate": 4.5224811574114186e-05, + "loss": 1.6055, + "step": 55620 + }, + { + "epoch": 4.81, + "learning_rate": 4.5223945248202374e-05, + "loss": 1.578, + "step": 55630 + }, + { + "epoch": 4.82, + "learning_rate": 4.522307892229057e-05, + "loss": 1.6194, + "step": 55640 + }, + { + "epoch": 4.82, + "learning_rate": 4.522221259637876e-05, + "loss": 1.6063, + "step": 55650 + }, + { + "epoch": 4.82, + "learning_rate": 4.522134627046695e-05, + "loss": 1.5394, + "step": 55660 + }, + { + "epoch": 4.82, + "learning_rate": 4.5220479944555145e-05, + "loss": 1.6064, + "step": 55670 + }, + { + "epoch": 4.82, + "learning_rate": 4.521961361864334e-05, + "loss": 1.6175, + "step": 55680 + }, + { + "epoch": 4.82, + "learning_rate": 4.521874729273153e-05, + "loss": 1.5821, + "step": 55690 + }, + { + "epoch": 4.82, + "learning_rate": 4.521788096681972e-05, + "loss": 1.5944, + "step": 55700 + }, + { + "epoch": 4.82, + "learning_rate": 4.521701464090791e-05, + "loss": 1.6753, + "step": 55710 + }, + { + "epoch": 4.82, + "learning_rate": 4.5216148314996103e-05, + "loss": 1.5987, + "step": 55720 + }, + { + "epoch": 4.82, + "learning_rate": 4.52152819890843e-05, + "loss": 1.5639, + "step": 55730 + }, + { + "epoch": 4.82, + "learning_rate": 4.5214415663172486e-05, + "loss": 1.57, + "step": 55740 + }, + { + "epoch": 4.83, + "learning_rate": 4.521354933726068e-05, + "loss": 1.5808, + "step": 55750 + }, + { + "epoch": 4.83, + "learning_rate": 4.5212683011348875e-05, + "loss": 1.5829, + "step": 55760 + }, + { + "epoch": 4.83, + "learning_rate": 4.521181668543706e-05, + "loss": 1.6096, + "step": 55770 + }, + { + "epoch": 4.83, + "learning_rate": 4.521095035952526e-05, + "loss": 1.5643, + "step": 55780 + }, + { + "epoch": 4.83, + "learning_rate": 4.521008403361345e-05, + "loss": 1.6477, + "step": 55790 + }, + { + "epoch": 4.83, + "learning_rate": 4.520921770770164e-05, + "loss": 1.5868, + "step": 55800 + }, + { + "epoch": 4.83, + "learning_rate": 4.520835138178983e-05, + "loss": 1.5829, + "step": 55810 + }, + { + "epoch": 4.83, + "learning_rate": 4.520748505587802e-05, + "loss": 1.6648, + "step": 55820 + }, + { + "epoch": 4.83, + "learning_rate": 4.5206618729966215e-05, + "loss": 1.6356, + "step": 55830 + }, + { + "epoch": 4.83, + "learning_rate": 4.520575240405441e-05, + "loss": 1.6579, + "step": 55840 + }, + { + "epoch": 4.83, + "learning_rate": 4.52048860781426e-05, + "loss": 1.5338, + "step": 55850 + }, + { + "epoch": 4.83, + "learning_rate": 4.520401975223079e-05, + "loss": 1.5846, + "step": 55860 + }, + { + "epoch": 4.84, + "learning_rate": 4.5203153426318986e-05, + "loss": 1.5478, + "step": 55870 + }, + { + "epoch": 4.84, + "learning_rate": 4.5202287100407174e-05, + "loss": 1.6013, + "step": 55880 + }, + { + "epoch": 4.84, + "learning_rate": 4.520142077449537e-05, + "loss": 1.5953, + "step": 55890 + }, + { + "epoch": 4.84, + "learning_rate": 4.520055444858356e-05, + "loss": 1.5431, + "step": 55900 + }, + { + "epoch": 4.84, + "learning_rate": 4.519968812267175e-05, + "loss": 1.5345, + "step": 55910 + }, + { + "epoch": 4.84, + "learning_rate": 4.5198821796759945e-05, + "loss": 1.5718, + "step": 55920 + }, + { + "epoch": 4.84, + "learning_rate": 4.519795547084813e-05, + "loss": 1.6058, + "step": 55930 + }, + { + "epoch": 4.84, + "learning_rate": 4.519708914493633e-05, + "loss": 1.5053, + "step": 55940 + }, + { + "epoch": 4.84, + "learning_rate": 4.519622281902452e-05, + "loss": 1.6927, + "step": 55950 + }, + { + "epoch": 4.84, + "learning_rate": 4.519535649311271e-05, + "loss": 1.5881, + "step": 55960 + }, + { + "epoch": 4.84, + "learning_rate": 4.5194490167200904e-05, + "loss": 1.5815, + "step": 55970 + }, + { + "epoch": 4.85, + "learning_rate": 4.51936238412891e-05, + "loss": 1.5856, + "step": 55980 + }, + { + "epoch": 4.85, + "learning_rate": 4.5192757515377286e-05, + "loss": 1.6678, + "step": 55990 + }, + { + "epoch": 4.85, + "learning_rate": 4.519189118946548e-05, + "loss": 1.5816, + "step": 56000 + }, + { + "epoch": 4.85, + "learning_rate": 4.5191024863553675e-05, + "loss": 1.6121, + "step": 56010 + }, + { + "epoch": 4.85, + "learning_rate": 4.519015853764186e-05, + "loss": 1.6054, + "step": 56020 + }, + { + "epoch": 4.85, + "learning_rate": 4.518929221173006e-05, + "loss": 1.5724, + "step": 56030 + }, + { + "epoch": 4.85, + "learning_rate": 4.5188425885818245e-05, + "loss": 1.6726, + "step": 56040 + }, + { + "epoch": 4.85, + "learning_rate": 4.518755955990644e-05, + "loss": 1.6597, + "step": 56050 + }, + { + "epoch": 4.85, + "learning_rate": 4.5186693233994634e-05, + "loss": 1.587, + "step": 56060 + }, + { + "epoch": 4.85, + "learning_rate": 4.518582690808282e-05, + "loss": 1.5791, + "step": 56070 + }, + { + "epoch": 4.85, + "learning_rate": 4.5184960582171016e-05, + "loss": 1.6202, + "step": 56080 + }, + { + "epoch": 4.85, + "learning_rate": 4.518409425625921e-05, + "loss": 1.6471, + "step": 56090 + }, + { + "epoch": 4.86, + "learning_rate": 4.51832279303474e-05, + "loss": 1.6046, + "step": 56100 + }, + { + "epoch": 4.86, + "learning_rate": 4.518236160443559e-05, + "loss": 1.5168, + "step": 56110 + }, + { + "epoch": 4.86, + "learning_rate": 4.518149527852378e-05, + "loss": 1.6257, + "step": 56120 + }, + { + "epoch": 4.86, + "learning_rate": 4.5180628952611974e-05, + "loss": 1.5975, + "step": 56130 + }, + { + "epoch": 4.86, + "learning_rate": 4.517976262670017e-05, + "loss": 1.5487, + "step": 56140 + }, + { + "epoch": 4.86, + "learning_rate": 4.5178896300788357e-05, + "loss": 1.524, + "step": 56150 + }, + { + "epoch": 4.86, + "learning_rate": 4.517802997487655e-05, + "loss": 1.5839, + "step": 56160 + }, + { + "epoch": 4.86, + "learning_rate": 4.5177163648964745e-05, + "loss": 1.5807, + "step": 56170 + }, + { + "epoch": 4.86, + "learning_rate": 4.517629732305293e-05, + "loss": 1.6016, + "step": 56180 + }, + { + "epoch": 4.86, + "learning_rate": 4.517543099714113e-05, + "loss": 1.5737, + "step": 56190 + }, + { + "epoch": 4.86, + "learning_rate": 4.517456467122932e-05, + "loss": 1.5198, + "step": 56200 + }, + { + "epoch": 4.87, + "learning_rate": 4.517369834531751e-05, + "loss": 1.6147, + "step": 56210 + }, + { + "epoch": 4.87, + "learning_rate": 4.5172832019405704e-05, + "loss": 1.6152, + "step": 56220 + }, + { + "epoch": 4.87, + "learning_rate": 4.517196569349389e-05, + "loss": 1.5704, + "step": 56230 + }, + { + "epoch": 4.87, + "learning_rate": 4.5171099367582086e-05, + "loss": 1.6295, + "step": 56240 + }, + { + "epoch": 4.87, + "learning_rate": 4.517023304167028e-05, + "loss": 1.6175, + "step": 56250 + }, + { + "epoch": 4.87, + "learning_rate": 4.516936671575847e-05, + "loss": 1.6365, + "step": 56260 + }, + { + "epoch": 4.87, + "learning_rate": 4.516850038984666e-05, + "loss": 1.5789, + "step": 56270 + }, + { + "epoch": 4.87, + "learning_rate": 4.516763406393486e-05, + "loss": 1.5807, + "step": 56280 + }, + { + "epoch": 4.87, + "learning_rate": 4.5166767738023045e-05, + "loss": 1.6457, + "step": 56290 + }, + { + "epoch": 4.87, + "learning_rate": 4.516590141211124e-05, + "loss": 1.5867, + "step": 56300 + }, + { + "epoch": 4.87, + "learning_rate": 4.5165035086199434e-05, + "loss": 1.6175, + "step": 56310 + }, + { + "epoch": 4.87, + "learning_rate": 4.516416876028762e-05, + "loss": 1.7045, + "step": 56320 + }, + { + "epoch": 4.88, + "learning_rate": 4.5163302434375816e-05, + "loss": 1.6025, + "step": 56330 + }, + { + "epoch": 4.88, + "learning_rate": 4.5162436108464004e-05, + "loss": 1.5907, + "step": 56340 + }, + { + "epoch": 4.88, + "learning_rate": 4.51615697825522e-05, + "loss": 1.622, + "step": 56350 + }, + { + "epoch": 4.88, + "learning_rate": 4.516070345664039e-05, + "loss": 1.557, + "step": 56360 + }, + { + "epoch": 4.88, + "learning_rate": 4.515983713072858e-05, + "loss": 1.6342, + "step": 56370 + }, + { + "epoch": 4.88, + "learning_rate": 4.5158970804816775e-05, + "loss": 1.6219, + "step": 56380 + }, + { + "epoch": 4.88, + "learning_rate": 4.515810447890497e-05, + "loss": 1.5468, + "step": 56390 + }, + { + "epoch": 4.88, + "learning_rate": 4.515723815299316e-05, + "loss": 1.6443, + "step": 56400 + }, + { + "epoch": 4.88, + "learning_rate": 4.515637182708135e-05, + "loss": 1.5884, + "step": 56410 + }, + { + "epoch": 4.88, + "learning_rate": 4.5155505501169546e-05, + "loss": 1.6486, + "step": 56420 + }, + { + "epoch": 4.88, + "learning_rate": 4.5154639175257733e-05, + "loss": 1.608, + "step": 56430 + }, + { + "epoch": 4.89, + "learning_rate": 4.515377284934593e-05, + "loss": 1.5854, + "step": 56440 + }, + { + "epoch": 4.89, + "learning_rate": 4.5152906523434116e-05, + "loss": 1.5451, + "step": 56450 + }, + { + "epoch": 4.89, + "learning_rate": 4.515204019752231e-05, + "loss": 1.6156, + "step": 56460 + }, + { + "epoch": 4.89, + "learning_rate": 4.5151173871610504e-05, + "loss": 1.5626, + "step": 56470 + }, + { + "epoch": 4.89, + "learning_rate": 4.515030754569869e-05, + "loss": 1.5664, + "step": 56480 + }, + { + "epoch": 4.89, + "learning_rate": 4.514944121978689e-05, + "loss": 1.6089, + "step": 56490 + }, + { + "epoch": 4.89, + "learning_rate": 4.514857489387508e-05, + "loss": 1.4955, + "step": 56500 + }, + { + "epoch": 4.89, + "learning_rate": 4.514770856796327e-05, + "loss": 1.6224, + "step": 56510 + }, + { + "epoch": 4.89, + "learning_rate": 4.514684224205146e-05, + "loss": 1.5714, + "step": 56520 + }, + { + "epoch": 4.89, + "learning_rate": 4.514597591613966e-05, + "loss": 1.6585, + "step": 56530 + }, + { + "epoch": 4.89, + "learning_rate": 4.5145109590227845e-05, + "loss": 1.6423, + "step": 56540 + }, + { + "epoch": 4.89, + "learning_rate": 4.514424326431604e-05, + "loss": 1.6408, + "step": 56550 + }, + { + "epoch": 4.9, + "learning_rate": 4.514337693840423e-05, + "loss": 1.5691, + "step": 56560 + }, + { + "epoch": 4.9, + "learning_rate": 4.514251061249242e-05, + "loss": 1.6024, + "step": 56570 + }, + { + "epoch": 4.9, + "learning_rate": 4.5141644286580616e-05, + "loss": 1.557, + "step": 56580 + }, + { + "epoch": 4.9, + "learning_rate": 4.5140777960668804e-05, + "loss": 1.6031, + "step": 56590 + }, + { + "epoch": 4.9, + "learning_rate": 4.5139911634757e-05, + "loss": 1.5848, + "step": 56600 + }, + { + "epoch": 4.9, + "learning_rate": 4.513904530884519e-05, + "loss": 1.6192, + "step": 56610 + }, + { + "epoch": 4.9, + "learning_rate": 4.513817898293338e-05, + "loss": 1.6212, + "step": 56620 + }, + { + "epoch": 4.9, + "learning_rate": 4.5137312657021575e-05, + "loss": 1.6545, + "step": 56630 + }, + { + "epoch": 4.9, + "learning_rate": 4.513644633110977e-05, + "loss": 1.5833, + "step": 56640 + }, + { + "epoch": 4.9, + "learning_rate": 4.513558000519796e-05, + "loss": 1.6527, + "step": 56650 + }, + { + "epoch": 4.9, + "learning_rate": 4.513471367928615e-05, + "loss": 1.5806, + "step": 56660 + }, + { + "epoch": 4.91, + "learning_rate": 4.513384735337434e-05, + "loss": 1.5878, + "step": 56670 + }, + { + "epoch": 4.91, + "learning_rate": 4.5132981027462534e-05, + "loss": 1.6942, + "step": 56680 + }, + { + "epoch": 4.91, + "learning_rate": 4.513211470155073e-05, + "loss": 1.5809, + "step": 56690 + }, + { + "epoch": 4.91, + "learning_rate": 4.5131248375638916e-05, + "loss": 1.657, + "step": 56700 + }, + { + "epoch": 4.91, + "learning_rate": 4.513038204972711e-05, + "loss": 1.6405, + "step": 56710 + }, + { + "epoch": 4.91, + "learning_rate": 4.5129515723815305e-05, + "loss": 1.5241, + "step": 56720 + }, + { + "epoch": 4.91, + "learning_rate": 4.512864939790349e-05, + "loss": 1.5167, + "step": 56730 + }, + { + "epoch": 4.91, + "learning_rate": 4.512778307199169e-05, + "loss": 1.6257, + "step": 56740 + }, + { + "epoch": 4.91, + "learning_rate": 4.512691674607988e-05, + "loss": 1.6163, + "step": 56750 + }, + { + "epoch": 4.91, + "learning_rate": 4.512605042016807e-05, + "loss": 1.584, + "step": 56760 + }, + { + "epoch": 4.91, + "learning_rate": 4.5125184094256264e-05, + "loss": 1.5864, + "step": 56770 + }, + { + "epoch": 4.91, + "learning_rate": 4.512431776834445e-05, + "loss": 1.581, + "step": 56780 + }, + { + "epoch": 4.92, + "learning_rate": 4.5123451442432646e-05, + "loss": 1.6354, + "step": 56790 + }, + { + "epoch": 4.92, + "learning_rate": 4.512258511652084e-05, + "loss": 1.6401, + "step": 56800 + }, + { + "epoch": 4.92, + "learning_rate": 4.512171879060903e-05, + "loss": 1.6918, + "step": 56810 + }, + { + "epoch": 4.92, + "learning_rate": 4.512085246469722e-05, + "loss": 1.6251, + "step": 56820 + }, + { + "epoch": 4.92, + "learning_rate": 4.511998613878542e-05, + "loss": 1.609, + "step": 56830 + }, + { + "epoch": 4.92, + "learning_rate": 4.5119119812873604e-05, + "loss": 1.6353, + "step": 56840 + }, + { + "epoch": 4.92, + "learning_rate": 4.51182534869618e-05, + "loss": 1.6131, + "step": 56850 + }, + { + "epoch": 4.92, + "learning_rate": 4.5117387161049987e-05, + "loss": 1.5842, + "step": 56860 + }, + { + "epoch": 4.92, + "learning_rate": 4.511652083513818e-05, + "loss": 1.5836, + "step": 56870 + }, + { + "epoch": 4.92, + "learning_rate": 4.5115654509226375e-05, + "loss": 1.6036, + "step": 56880 + }, + { + "epoch": 4.92, + "learning_rate": 4.511478818331456e-05, + "loss": 1.609, + "step": 56890 + }, + { + "epoch": 4.92, + "learning_rate": 4.511392185740276e-05, + "loss": 1.5611, + "step": 56900 + }, + { + "epoch": 4.93, + "learning_rate": 4.511305553149095e-05, + "loss": 1.6253, + "step": 56910 + }, + { + "epoch": 4.93, + "learning_rate": 4.511218920557914e-05, + "loss": 1.6299, + "step": 56920 + }, + { + "epoch": 4.93, + "learning_rate": 4.5111322879667334e-05, + "loss": 1.5935, + "step": 56930 + }, + { + "epoch": 4.93, + "learning_rate": 4.511045655375553e-05, + "loss": 1.6107, + "step": 56940 + }, + { + "epoch": 4.93, + "learning_rate": 4.5109590227843716e-05, + "loss": 1.6185, + "step": 56950 + }, + { + "epoch": 4.93, + "learning_rate": 4.510872390193191e-05, + "loss": 1.6077, + "step": 56960 + }, + { + "epoch": 4.93, + "learning_rate": 4.51078575760201e-05, + "loss": 1.5152, + "step": 56970 + }, + { + "epoch": 4.93, + "learning_rate": 4.510699125010829e-05, + "loss": 1.6271, + "step": 56980 + }, + { + "epoch": 4.93, + "learning_rate": 4.510612492419649e-05, + "loss": 1.5669, + "step": 56990 + }, + { + "epoch": 4.93, + "learning_rate": 4.5105258598284675e-05, + "loss": 1.6076, + "step": 57000 + }, + { + "epoch": 4.93, + "learning_rate": 4.510439227237287e-05, + "loss": 1.5727, + "step": 57010 + }, + { + "epoch": 4.94, + "learning_rate": 4.5103525946461064e-05, + "loss": 1.5563, + "step": 57020 + }, + { + "epoch": 4.94, + "learning_rate": 4.510265962054925e-05, + "loss": 1.5633, + "step": 57030 + }, + { + "epoch": 4.94, + "learning_rate": 4.5101793294637446e-05, + "loss": 1.5864, + "step": 57040 + }, + { + "epoch": 4.94, + "learning_rate": 4.510092696872564e-05, + "loss": 1.6254, + "step": 57050 + }, + { + "epoch": 4.94, + "learning_rate": 4.510006064281383e-05, + "loss": 1.603, + "step": 57060 + }, + { + "epoch": 4.94, + "learning_rate": 4.509919431690202e-05, + "loss": 1.6248, + "step": 57070 + }, + { + "epoch": 4.94, + "learning_rate": 4.509832799099021e-05, + "loss": 1.6255, + "step": 57080 + }, + { + "epoch": 4.94, + "learning_rate": 4.5097461665078405e-05, + "loss": 1.6698, + "step": 57090 + }, + { + "epoch": 4.94, + "learning_rate": 4.50965953391666e-05, + "loss": 1.5809, + "step": 57100 + }, + { + "epoch": 4.94, + "learning_rate": 4.509572901325479e-05, + "loss": 1.5509, + "step": 57110 + }, + { + "epoch": 4.94, + "learning_rate": 4.509486268734298e-05, + "loss": 1.581, + "step": 57120 + }, + { + "epoch": 4.94, + "learning_rate": 4.5093996361431176e-05, + "loss": 1.6253, + "step": 57130 + }, + { + "epoch": 4.95, + "learning_rate": 4.5093130035519363e-05, + "loss": 1.5346, + "step": 57140 + }, + { + "epoch": 4.95, + "learning_rate": 4.509226370960756e-05, + "loss": 1.5665, + "step": 57150 + }, + { + "epoch": 4.95, + "learning_rate": 4.509139738369575e-05, + "loss": 1.5716, + "step": 57160 + }, + { + "epoch": 4.95, + "learning_rate": 4.509053105778394e-05, + "loss": 1.6353, + "step": 57170 + }, + { + "epoch": 4.95, + "learning_rate": 4.5089664731872134e-05, + "loss": 1.5379, + "step": 57180 + }, + { + "epoch": 4.95, + "learning_rate": 4.508879840596032e-05, + "loss": 1.6087, + "step": 57190 + }, + { + "epoch": 4.95, + "learning_rate": 4.5087932080048517e-05, + "loss": 1.623, + "step": 57200 + }, + { + "epoch": 4.95, + "learning_rate": 4.508706575413671e-05, + "loss": 1.5124, + "step": 57210 + }, + { + "epoch": 4.95, + "learning_rate": 4.50861994282249e-05, + "loss": 1.5961, + "step": 57220 + }, + { + "epoch": 4.95, + "learning_rate": 4.508533310231309e-05, + "loss": 1.6397, + "step": 57230 + }, + { + "epoch": 4.95, + "learning_rate": 4.508446677640129e-05, + "loss": 1.6404, + "step": 57240 + }, + { + "epoch": 4.96, + "learning_rate": 4.5083600450489475e-05, + "loss": 1.6003, + "step": 57250 + }, + { + "epoch": 4.96, + "learning_rate": 4.508273412457767e-05, + "loss": 1.5676, + "step": 57260 + }, + { + "epoch": 4.96, + "learning_rate": 4.5081867798665864e-05, + "loss": 1.6385, + "step": 57270 + }, + { + "epoch": 4.96, + "learning_rate": 4.508100147275405e-05, + "loss": 1.5905, + "step": 57280 + }, + { + "epoch": 4.96, + "learning_rate": 4.5080135146842246e-05, + "loss": 1.574, + "step": 57290 + }, + { + "epoch": 4.96, + "learning_rate": 4.5079268820930434e-05, + "loss": 1.5759, + "step": 57300 + }, + { + "epoch": 4.96, + "learning_rate": 4.507840249501863e-05, + "loss": 1.6594, + "step": 57310 + }, + { + "epoch": 4.96, + "learning_rate": 4.507753616910682e-05, + "loss": 1.6321, + "step": 57320 + }, + { + "epoch": 4.96, + "learning_rate": 4.507666984319501e-05, + "loss": 1.5443, + "step": 57330 + }, + { + "epoch": 4.96, + "learning_rate": 4.5075803517283205e-05, + "loss": 1.588, + "step": 57340 + }, + { + "epoch": 4.96, + "learning_rate": 4.50749371913714e-05, + "loss": 1.6074, + "step": 57350 + }, + { + "epoch": 4.96, + "learning_rate": 4.507407086545959e-05, + "loss": 1.615, + "step": 57360 + }, + { + "epoch": 4.97, + "learning_rate": 4.507320453954778e-05, + "loss": 1.6017, + "step": 57370 + }, + { + "epoch": 4.97, + "learning_rate": 4.5072338213635976e-05, + "loss": 1.6143, + "step": 57380 + }, + { + "epoch": 4.97, + "learning_rate": 4.5071471887724164e-05, + "loss": 1.65, + "step": 57390 + }, + { + "epoch": 4.97, + "learning_rate": 4.507060556181236e-05, + "loss": 1.626, + "step": 57400 + }, + { + "epoch": 4.97, + "learning_rate": 4.5069739235900546e-05, + "loss": 1.6411, + "step": 57410 + }, + { + "epoch": 4.97, + "learning_rate": 4.506887290998874e-05, + "loss": 1.6045, + "step": 57420 + }, + { + "epoch": 4.97, + "learning_rate": 4.5068006584076935e-05, + "loss": 1.5826, + "step": 57430 + }, + { + "epoch": 4.97, + "learning_rate": 4.506714025816512e-05, + "loss": 1.6573, + "step": 57440 + }, + { + "epoch": 4.97, + "learning_rate": 4.506627393225332e-05, + "loss": 1.5106, + "step": 57450 + }, + { + "epoch": 4.97, + "learning_rate": 4.506540760634151e-05, + "loss": 1.5941, + "step": 57460 + }, + { + "epoch": 4.97, + "learning_rate": 4.50645412804297e-05, + "loss": 1.5532, + "step": 57470 + }, + { + "epoch": 4.98, + "learning_rate": 4.5063674954517894e-05, + "loss": 1.5957, + "step": 57480 + }, + { + "epoch": 4.98, + "learning_rate": 4.506280862860609e-05, + "loss": 1.6007, + "step": 57490 + }, + { + "epoch": 4.98, + "learning_rate": 4.5061942302694276e-05, + "loss": 1.5359, + "step": 57500 + }, + { + "epoch": 4.98, + "learning_rate": 4.506107597678247e-05, + "loss": 1.5751, + "step": 57510 + }, + { + "epoch": 4.98, + "learning_rate": 4.506020965087066e-05, + "loss": 1.6255, + "step": 57520 + }, + { + "epoch": 4.98, + "learning_rate": 4.505934332495885e-05, + "loss": 1.6042, + "step": 57530 + }, + { + "epoch": 4.98, + "learning_rate": 4.505847699904705e-05, + "loss": 1.6029, + "step": 57540 + }, + { + "epoch": 4.98, + "learning_rate": 4.5057610673135234e-05, + "loss": 1.5884, + "step": 57550 + }, + { + "epoch": 4.98, + "learning_rate": 4.505674434722343e-05, + "loss": 1.6355, + "step": 57560 + }, + { + "epoch": 4.98, + "learning_rate": 4.505587802131162e-05, + "loss": 1.6166, + "step": 57570 + }, + { + "epoch": 4.98, + "learning_rate": 4.505501169539981e-05, + "loss": 1.5999, + "step": 57580 + }, + { + "epoch": 4.98, + "learning_rate": 4.5054145369488005e-05, + "loss": 1.5463, + "step": 57590 + }, + { + "epoch": 4.99, + "learning_rate": 4.505327904357619e-05, + "loss": 1.5445, + "step": 57600 + }, + { + "epoch": 4.99, + "learning_rate": 4.505241271766439e-05, + "loss": 1.6948, + "step": 57610 + }, + { + "epoch": 4.99, + "learning_rate": 4.505154639175258e-05, + "loss": 1.6051, + "step": 57620 + }, + { + "epoch": 4.99, + "learning_rate": 4.505068006584077e-05, + "loss": 1.6082, + "step": 57630 + }, + { + "epoch": 4.99, + "learning_rate": 4.5049813739928964e-05, + "loss": 1.6032, + "step": 57640 + }, + { + "epoch": 4.99, + "learning_rate": 4.504894741401716e-05, + "loss": 1.5936, + "step": 57650 + }, + { + "epoch": 4.99, + "learning_rate": 4.5048081088105346e-05, + "loss": 1.6261, + "step": 57660 + }, + { + "epoch": 4.99, + "learning_rate": 4.504721476219354e-05, + "loss": 1.6012, + "step": 57670 + }, + { + "epoch": 4.99, + "learning_rate": 4.5046348436281735e-05, + "loss": 1.5531, + "step": 57680 + }, + { + "epoch": 4.99, + "learning_rate": 4.504548211036992e-05, + "loss": 1.5943, + "step": 57690 + }, + { + "epoch": 4.99, + "learning_rate": 4.504461578445812e-05, + "loss": 1.605, + "step": 57700 + }, + { + "epoch": 5.0, + "learning_rate": 4.5043749458546305e-05, + "loss": 1.5885, + "step": 57710 + }, + { + "epoch": 5.0, + "learning_rate": 4.50428831326345e-05, + "loss": 1.5607, + "step": 57720 + }, + { + "epoch": 5.0, + "learning_rate": 4.5042016806722694e-05, + "loss": 1.6047, + "step": 57730 + }, + { + "epoch": 5.0, + "learning_rate": 4.504115048081088e-05, + "loss": 1.6171, + "step": 57740 + }, + { + "epoch": 5.0, + "learning_rate": 4.5040284154899076e-05, + "loss": 1.5782, + "step": 57750 + }, + { + "epoch": 5.0, + "learning_rate": 4.503941782898727e-05, + "loss": 1.5901, + "step": 57760 + }, + { + "epoch": 5.0, + "eval_Bleu_1": 0.037539281973144704, + "eval_Bleu_2": 2.815083546033997e-11, + "eval_Bleu_3": 2.657276169229133e-14, + "eval_Bleu_4": 8.433037705828722e-16, + "eval_ROUGE_L": 0.08968639151675512, + "eval_cer": 0.9948382955883373, + "eval_em": 0, + "eval_f1": 0.053294394616487786, + "eval_loss": 1.5039069652557373, + "eval_runtime": 1731.5347, + "eval_samples_per_second": 2.966, + "eval_steps_per_second": 2.966, + "eval_wer": 0.9822522787208552, + "step": 57767 + }, + { + "epoch": 5.0, + "learning_rate": 4.503855150307546e-05, + "loss": 1.6626, + "step": 57770 + }, + { + "epoch": 5.0, + "learning_rate": 4.503768517716365e-05, + "loss": 1.5821, + "step": 57780 + }, + { + "epoch": 5.0, + "learning_rate": 4.503681885125185e-05, + "loss": 1.5834, + "step": 57790 + }, + { + "epoch": 5.0, + "learning_rate": 4.5035952525340035e-05, + "loss": 1.4865, + "step": 57800 + }, + { + "epoch": 5.0, + "learning_rate": 4.503508619942823e-05, + "loss": 1.5915, + "step": 57810 + }, + { + "epoch": 5.0, + "learning_rate": 4.503421987351642e-05, + "loss": 1.5225, + "step": 57820 + }, + { + "epoch": 5.01, + "learning_rate": 4.503335354760461e-05, + "loss": 1.5484, + "step": 57830 + }, + { + "epoch": 5.01, + "learning_rate": 4.5032487221692806e-05, + "loss": 1.5869, + "step": 57840 + }, + { + "epoch": 5.01, + "learning_rate": 4.5031620895780993e-05, + "loss": 1.48, + "step": 57850 + }, + { + "epoch": 5.01, + "learning_rate": 4.503075456986919e-05, + "loss": 1.5703, + "step": 57860 + }, + { + "epoch": 5.01, + "learning_rate": 4.502988824395738e-05, + "loss": 1.5485, + "step": 57870 + }, + { + "epoch": 5.01, + "learning_rate": 4.502902191804557e-05, + "loss": 1.5832, + "step": 57880 + }, + { + "epoch": 5.01, + "learning_rate": 4.5028155592133764e-05, + "loss": 1.6302, + "step": 57890 + }, + { + "epoch": 5.01, + "learning_rate": 4.502728926622196e-05, + "loss": 1.5693, + "step": 57900 + }, + { + "epoch": 5.01, + "learning_rate": 4.5026422940310147e-05, + "loss": 1.5662, + "step": 57910 + }, + { + "epoch": 5.01, + "learning_rate": 4.502555661439834e-05, + "loss": 1.559, + "step": 57920 + }, + { + "epoch": 5.01, + "learning_rate": 4.502469028848653e-05, + "loss": 1.5938, + "step": 57930 + }, + { + "epoch": 5.01, + "learning_rate": 4.502382396257472e-05, + "loss": 1.5152, + "step": 57940 + }, + { + "epoch": 5.02, + "learning_rate": 4.502295763666292e-05, + "loss": 1.505, + "step": 57950 + }, + { + "epoch": 5.02, + "learning_rate": 4.5022091310751105e-05, + "loss": 1.5009, + "step": 57960 + }, + { + "epoch": 5.02, + "learning_rate": 4.50212249848393e-05, + "loss": 1.5449, + "step": 57970 + }, + { + "epoch": 5.02, + "learning_rate": 4.5020358658927494e-05, + "loss": 1.6155, + "step": 57980 + }, + { + "epoch": 5.02, + "learning_rate": 4.501949233301568e-05, + "loss": 1.5475, + "step": 57990 + }, + { + "epoch": 5.02, + "learning_rate": 4.5018626007103876e-05, + "loss": 1.5346, + "step": 58000 + }, + { + "epoch": 5.02, + "learning_rate": 4.501775968119207e-05, + "loss": 1.6501, + "step": 58010 + }, + { + "epoch": 5.02, + "learning_rate": 4.501689335528026e-05, + "loss": 1.6275, + "step": 58020 + }, + { + "epoch": 5.02, + "learning_rate": 4.501602702936845e-05, + "loss": 1.5436, + "step": 58030 + }, + { + "epoch": 5.02, + "learning_rate": 4.501516070345664e-05, + "loss": 1.5987, + "step": 58040 + }, + { + "epoch": 5.02, + "learning_rate": 4.5014294377544835e-05, + "loss": 1.5711, + "step": 58050 + }, + { + "epoch": 5.03, + "learning_rate": 4.501342805163303e-05, + "loss": 1.6284, + "step": 58060 + }, + { + "epoch": 5.03, + "learning_rate": 4.501256172572122e-05, + "loss": 1.5239, + "step": 58070 + }, + { + "epoch": 5.03, + "learning_rate": 4.501169539980941e-05, + "loss": 1.6012, + "step": 58080 + }, + { + "epoch": 5.03, + "learning_rate": 4.5010829073897606e-05, + "loss": 1.6047, + "step": 58090 + }, + { + "epoch": 5.03, + "learning_rate": 4.5009962747985794e-05, + "loss": 1.6458, + "step": 58100 + }, + { + "epoch": 5.03, + "learning_rate": 4.500909642207399e-05, + "loss": 1.6237, + "step": 58110 + }, + { + "epoch": 5.03, + "learning_rate": 4.500823009616218e-05, + "loss": 1.5873, + "step": 58120 + }, + { + "epoch": 5.03, + "learning_rate": 4.500736377025037e-05, + "loss": 1.6118, + "step": 58130 + }, + { + "epoch": 5.03, + "learning_rate": 4.5006497444338565e-05, + "loss": 1.5032, + "step": 58140 + }, + { + "epoch": 5.03, + "learning_rate": 4.500563111842675e-05, + "loss": 1.5924, + "step": 58150 + }, + { + "epoch": 5.03, + "learning_rate": 4.500476479251495e-05, + "loss": 1.512, + "step": 58160 + }, + { + "epoch": 5.03, + "learning_rate": 4.500389846660314e-05, + "loss": 1.5306, + "step": 58170 + }, + { + "epoch": 5.04, + "learning_rate": 4.500303214069133e-05, + "loss": 1.556, + "step": 58180 + }, + { + "epoch": 5.04, + "learning_rate": 4.5002165814779523e-05, + "loss": 1.5545, + "step": 58190 + }, + { + "epoch": 5.04, + "learning_rate": 4.500129948886772e-05, + "loss": 1.6006, + "step": 58200 + }, + { + "epoch": 5.04, + "learning_rate": 4.5000433162955906e-05, + "loss": 1.619, + "step": 58210 + }, + { + "epoch": 5.04, + "learning_rate": 4.49995668370441e-05, + "loss": 1.6076, + "step": 58220 + }, + { + "epoch": 5.04, + "learning_rate": 4.4998700511132295e-05, + "loss": 1.5482, + "step": 58230 + }, + { + "epoch": 5.04, + "learning_rate": 4.499783418522048e-05, + "loss": 1.608, + "step": 58240 + }, + { + "epoch": 5.04, + "learning_rate": 4.499696785930868e-05, + "loss": 1.5534, + "step": 58250 + }, + { + "epoch": 5.04, + "learning_rate": 4.4996101533396864e-05, + "loss": 1.5289, + "step": 58260 + }, + { + "epoch": 5.04, + "learning_rate": 4.499523520748506e-05, + "loss": 1.5569, + "step": 58270 + }, + { + "epoch": 5.04, + "learning_rate": 4.499436888157325e-05, + "loss": 1.5575, + "step": 58280 + }, + { + "epoch": 5.05, + "learning_rate": 4.499350255566144e-05, + "loss": 1.5164, + "step": 58290 + }, + { + "epoch": 5.05, + "learning_rate": 4.4992636229749635e-05, + "loss": 1.5811, + "step": 58300 + }, + { + "epoch": 5.05, + "learning_rate": 4.499176990383783e-05, + "loss": 1.5532, + "step": 58310 + }, + { + "epoch": 5.05, + "learning_rate": 4.499090357792602e-05, + "loss": 1.6388, + "step": 58320 + }, + { + "epoch": 5.05, + "learning_rate": 4.499003725201421e-05, + "loss": 1.5722, + "step": 58330 + }, + { + "epoch": 5.05, + "learning_rate": 4.4989170926102406e-05, + "loss": 1.5926, + "step": 58340 + }, + { + "epoch": 5.05, + "learning_rate": 4.4988304600190594e-05, + "loss": 1.5532, + "step": 58350 + }, + { + "epoch": 5.05, + "learning_rate": 4.498743827427879e-05, + "loss": 1.495, + "step": 58360 + }, + { + "epoch": 5.05, + "learning_rate": 4.4986571948366976e-05, + "loss": 1.5739, + "step": 58370 + }, + { + "epoch": 5.05, + "learning_rate": 4.498570562245517e-05, + "loss": 1.4825, + "step": 58380 + }, + { + "epoch": 5.05, + "learning_rate": 4.4984839296543365e-05, + "loss": 1.5703, + "step": 58390 + }, + { + "epoch": 5.05, + "learning_rate": 4.498397297063155e-05, + "loss": 1.602, + "step": 58400 + }, + { + "epoch": 5.06, + "learning_rate": 4.498310664471975e-05, + "loss": 1.445, + "step": 58410 + }, + { + "epoch": 5.06, + "learning_rate": 4.498224031880794e-05, + "loss": 1.5559, + "step": 58420 + }, + { + "epoch": 5.06, + "learning_rate": 4.498137399289613e-05, + "loss": 1.5937, + "step": 58430 + }, + { + "epoch": 5.06, + "learning_rate": 4.4980507666984324e-05, + "loss": 1.534, + "step": 58440 + }, + { + "epoch": 5.06, + "learning_rate": 4.497964134107251e-05, + "loss": 1.5841, + "step": 58450 + }, + { + "epoch": 5.06, + "learning_rate": 4.4978775015160706e-05, + "loss": 1.537, + "step": 58460 + }, + { + "epoch": 5.06, + "learning_rate": 4.49779086892489e-05, + "loss": 1.5442, + "step": 58470 + }, + { + "epoch": 5.06, + "learning_rate": 4.497704236333709e-05, + "loss": 1.5945, + "step": 58480 + }, + { + "epoch": 5.06, + "learning_rate": 4.497617603742528e-05, + "loss": 1.5182, + "step": 58490 + }, + { + "epoch": 5.06, + "learning_rate": 4.497530971151348e-05, + "loss": 1.5144, + "step": 58500 + }, + { + "epoch": 5.06, + "learning_rate": 4.4974443385601665e-05, + "loss": 1.5574, + "step": 58510 + }, + { + "epoch": 5.07, + "learning_rate": 4.497357705968986e-05, + "loss": 1.5839, + "step": 58520 + }, + { + "epoch": 5.07, + "learning_rate": 4.4972710733778054e-05, + "loss": 1.5767, + "step": 58530 + }, + { + "epoch": 5.07, + "learning_rate": 4.497184440786624e-05, + "loss": 1.5219, + "step": 58540 + }, + { + "epoch": 5.07, + "learning_rate": 4.4970978081954436e-05, + "loss": 1.5732, + "step": 58550 + }, + { + "epoch": 5.07, + "learning_rate": 4.497011175604262e-05, + "loss": 1.6229, + "step": 58560 + }, + { + "epoch": 5.07, + "learning_rate": 4.496924543013082e-05, + "loss": 1.5742, + "step": 58570 + }, + { + "epoch": 5.07, + "learning_rate": 4.496837910421901e-05, + "loss": 1.6521, + "step": 58580 + }, + { + "epoch": 5.07, + "learning_rate": 4.49675127783072e-05, + "loss": 1.587, + "step": 58590 + }, + { + "epoch": 5.07, + "learning_rate": 4.4966646452395394e-05, + "loss": 1.6305, + "step": 58600 + }, + { + "epoch": 5.07, + "learning_rate": 4.496578012648359e-05, + "loss": 1.5503, + "step": 58610 + }, + { + "epoch": 5.07, + "learning_rate": 4.4964913800571777e-05, + "loss": 1.5552, + "step": 58620 + }, + { + "epoch": 5.07, + "learning_rate": 4.496404747465997e-05, + "loss": 1.5162, + "step": 58630 + }, + { + "epoch": 5.08, + "learning_rate": 4.4963181148748165e-05, + "loss": 1.6133, + "step": 58640 + }, + { + "epoch": 5.08, + "learning_rate": 4.496231482283635e-05, + "loss": 1.5952, + "step": 58650 + }, + { + "epoch": 5.08, + "learning_rate": 4.496144849692455e-05, + "loss": 1.5508, + "step": 58660 + }, + { + "epoch": 5.08, + "learning_rate": 4.4960582171012735e-05, + "loss": 1.6274, + "step": 58670 + }, + { + "epoch": 5.08, + "learning_rate": 4.495971584510093e-05, + "loss": 1.5062, + "step": 58680 + }, + { + "epoch": 5.08, + "learning_rate": 4.4958849519189124e-05, + "loss": 1.5188, + "step": 58690 + }, + { + "epoch": 5.08, + "learning_rate": 4.495798319327731e-05, + "loss": 1.5515, + "step": 58700 + }, + { + "epoch": 5.08, + "learning_rate": 4.4957116867365506e-05, + "loss": 1.5033, + "step": 58710 + }, + { + "epoch": 5.08, + "learning_rate": 4.49562505414537e-05, + "loss": 1.5381, + "step": 58720 + }, + { + "epoch": 5.08, + "learning_rate": 4.495538421554189e-05, + "loss": 1.5926, + "step": 58730 + }, + { + "epoch": 5.08, + "learning_rate": 4.495451788963008e-05, + "loss": 1.604, + "step": 58740 + }, + { + "epoch": 5.09, + "learning_rate": 4.495365156371828e-05, + "loss": 1.5784, + "step": 58750 + }, + { + "epoch": 5.09, + "learning_rate": 4.4952785237806465e-05, + "loss": 1.5488, + "step": 58760 + }, + { + "epoch": 5.09, + "learning_rate": 4.495191891189466e-05, + "loss": 1.5625, + "step": 58770 + }, + { + "epoch": 5.09, + "learning_rate": 4.495105258598285e-05, + "loss": 1.5829, + "step": 58780 + }, + { + "epoch": 5.09, + "learning_rate": 4.495018626007104e-05, + "loss": 1.5523, + "step": 58790 + }, + { + "epoch": 5.09, + "learning_rate": 4.4949319934159236e-05, + "loss": 1.6208, + "step": 58800 + }, + { + "epoch": 5.09, + "learning_rate": 4.4948453608247424e-05, + "loss": 1.5733, + "step": 58810 + }, + { + "epoch": 5.09, + "learning_rate": 4.494758728233562e-05, + "loss": 1.5651, + "step": 58820 + }, + { + "epoch": 5.09, + "learning_rate": 4.494672095642381e-05, + "loss": 1.5608, + "step": 58830 + }, + { + "epoch": 5.09, + "learning_rate": 4.4945854630512e-05, + "loss": 1.6415, + "step": 58840 + }, + { + "epoch": 5.09, + "learning_rate": 4.4944988304600195e-05, + "loss": 1.5802, + "step": 58850 + }, + { + "epoch": 5.09, + "learning_rate": 4.494412197868839e-05, + "loss": 1.5583, + "step": 58860 + }, + { + "epoch": 5.1, + "learning_rate": 4.494325565277658e-05, + "loss": 1.5537, + "step": 58870 + }, + { + "epoch": 5.1, + "learning_rate": 4.494238932686477e-05, + "loss": 1.5797, + "step": 58880 + }, + { + "epoch": 5.1, + "learning_rate": 4.494152300095296e-05, + "loss": 1.5578, + "step": 58890 + }, + { + "epoch": 5.1, + "learning_rate": 4.4940656675041153e-05, + "loss": 1.5046, + "step": 58900 + }, + { + "epoch": 5.1, + "learning_rate": 4.493979034912935e-05, + "loss": 1.5609, + "step": 58910 + }, + { + "epoch": 5.1, + "learning_rate": 4.4938924023217536e-05, + "loss": 1.5689, + "step": 58920 + }, + { + "epoch": 5.1, + "learning_rate": 4.493805769730573e-05, + "loss": 1.5751, + "step": 58930 + }, + { + "epoch": 5.1, + "learning_rate": 4.4937191371393924e-05, + "loss": 1.5619, + "step": 58940 + }, + { + "epoch": 5.1, + "learning_rate": 4.493632504548211e-05, + "loss": 1.6122, + "step": 58950 + }, + { + "epoch": 5.1, + "learning_rate": 4.493545871957031e-05, + "loss": 1.528, + "step": 58960 + }, + { + "epoch": 5.1, + "learning_rate": 4.49345923936585e-05, + "loss": 1.5385, + "step": 58970 + }, + { + "epoch": 5.1, + "learning_rate": 4.493372606774669e-05, + "loss": 1.5954, + "step": 58980 + }, + { + "epoch": 5.11, + "learning_rate": 4.493285974183488e-05, + "loss": 1.5832, + "step": 58990 + }, + { + "epoch": 5.11, + "learning_rate": 4.493199341592307e-05, + "loss": 1.58, + "step": 59000 + }, + { + "epoch": 5.11, + "learning_rate": 4.4931127090011265e-05, + "loss": 1.5648, + "step": 59010 + }, + { + "epoch": 5.11, + "learning_rate": 4.493026076409946e-05, + "loss": 1.5713, + "step": 59020 + }, + { + "epoch": 5.11, + "learning_rate": 4.492939443818765e-05, + "loss": 1.5045, + "step": 59030 + }, + { + "epoch": 5.11, + "learning_rate": 4.492852811227584e-05, + "loss": 1.5854, + "step": 59040 + }, + { + "epoch": 5.11, + "learning_rate": 4.4927661786364036e-05, + "loss": 1.582, + "step": 59050 + }, + { + "epoch": 5.11, + "learning_rate": 4.4926795460452224e-05, + "loss": 1.5026, + "step": 59060 + }, + { + "epoch": 5.11, + "learning_rate": 4.492592913454042e-05, + "loss": 1.5452, + "step": 59070 + }, + { + "epoch": 5.11, + "learning_rate": 4.492506280862861e-05, + "loss": 1.5484, + "step": 59080 + }, + { + "epoch": 5.11, + "learning_rate": 4.49241964827168e-05, + "loss": 1.5551, + "step": 59090 + }, + { + "epoch": 5.12, + "learning_rate": 4.4923330156804995e-05, + "loss": 1.6183, + "step": 59100 + }, + { + "epoch": 5.12, + "learning_rate": 4.492246383089318e-05, + "loss": 1.5532, + "step": 59110 + }, + { + "epoch": 5.12, + "learning_rate": 4.492159750498138e-05, + "loss": 1.5432, + "step": 59120 + }, + { + "epoch": 5.12, + "learning_rate": 4.492073117906957e-05, + "loss": 1.5109, + "step": 59130 + }, + { + "epoch": 5.12, + "learning_rate": 4.491986485315776e-05, + "loss": 1.5973, + "step": 59140 + }, + { + "epoch": 5.12, + "learning_rate": 4.4918998527245954e-05, + "loss": 1.5351, + "step": 59150 + }, + { + "epoch": 5.12, + "learning_rate": 4.491813220133415e-05, + "loss": 1.5893, + "step": 59160 + }, + { + "epoch": 5.12, + "learning_rate": 4.4917265875422336e-05, + "loss": 1.5954, + "step": 59170 + }, + { + "epoch": 5.12, + "learning_rate": 4.491639954951053e-05, + "loss": 1.5425, + "step": 59180 + }, + { + "epoch": 5.12, + "learning_rate": 4.491553322359872e-05, + "loss": 1.5476, + "step": 59190 + }, + { + "epoch": 5.12, + "learning_rate": 4.491466689768691e-05, + "loss": 1.5309, + "step": 59200 + }, + { + "epoch": 5.12, + "learning_rate": 4.491380057177511e-05, + "loss": 1.6077, + "step": 59210 + }, + { + "epoch": 5.13, + "learning_rate": 4.4912934245863295e-05, + "loss": 1.6189, + "step": 59220 + }, + { + "epoch": 5.13, + "learning_rate": 4.491206791995149e-05, + "loss": 1.586, + "step": 59230 + }, + { + "epoch": 5.13, + "learning_rate": 4.4911201594039684e-05, + "loss": 1.5922, + "step": 59240 + }, + { + "epoch": 5.13, + "learning_rate": 4.491033526812787e-05, + "loss": 1.5988, + "step": 59250 + }, + { + "epoch": 5.13, + "learning_rate": 4.4909468942216066e-05, + "loss": 1.5735, + "step": 59260 + }, + { + "epoch": 5.13, + "learning_rate": 4.490860261630426e-05, + "loss": 1.5972, + "step": 59270 + }, + { + "epoch": 5.13, + "learning_rate": 4.490773629039245e-05, + "loss": 1.589, + "step": 59280 + }, + { + "epoch": 5.13, + "learning_rate": 4.490686996448064e-05, + "loss": 1.5876, + "step": 59290 + }, + { + "epoch": 5.13, + "learning_rate": 4.490600363856883e-05, + "loss": 1.621, + "step": 59300 + }, + { + "epoch": 5.13, + "learning_rate": 4.4905137312657024e-05, + "loss": 1.6096, + "step": 59310 + }, + { + "epoch": 5.13, + "learning_rate": 4.490427098674522e-05, + "loss": 1.5638, + "step": 59320 + }, + { + "epoch": 5.14, + "learning_rate": 4.4903404660833407e-05, + "loss": 1.5291, + "step": 59330 + }, + { + "epoch": 5.14, + "learning_rate": 4.49025383349216e-05, + "loss": 1.5382, + "step": 59340 + }, + { + "epoch": 5.14, + "learning_rate": 4.4901672009009795e-05, + "loss": 1.5856, + "step": 59350 + }, + { + "epoch": 5.14, + "learning_rate": 4.490080568309798e-05, + "loss": 1.5006, + "step": 59360 + }, + { + "epoch": 5.14, + "learning_rate": 4.489993935718618e-05, + "loss": 1.5638, + "step": 59370 + }, + { + "epoch": 5.14, + "learning_rate": 4.489907303127437e-05, + "loss": 1.5467, + "step": 59380 + }, + { + "epoch": 5.14, + "learning_rate": 4.489820670536256e-05, + "loss": 1.5709, + "step": 59390 + }, + { + "epoch": 5.14, + "learning_rate": 4.4897340379450754e-05, + "loss": 1.5362, + "step": 59400 + }, + { + "epoch": 5.14, + "learning_rate": 4.489647405353894e-05, + "loss": 1.6382, + "step": 59410 + }, + { + "epoch": 5.14, + "learning_rate": 4.4895607727627136e-05, + "loss": 1.5602, + "step": 59420 + }, + { + "epoch": 5.14, + "learning_rate": 4.489474140171533e-05, + "loss": 1.529, + "step": 59430 + }, + { + "epoch": 5.14, + "learning_rate": 4.489387507580352e-05, + "loss": 1.5691, + "step": 59440 + }, + { + "epoch": 5.15, + "learning_rate": 4.489300874989171e-05, + "loss": 1.5891, + "step": 59450 + }, + { + "epoch": 5.15, + "learning_rate": 4.489214242397991e-05, + "loss": 1.5526, + "step": 59460 + }, + { + "epoch": 5.15, + "learning_rate": 4.4891276098068095e-05, + "loss": 1.5774, + "step": 59470 + }, + { + "epoch": 5.15, + "learning_rate": 4.489040977215629e-05, + "loss": 1.5931, + "step": 59480 + }, + { + "epoch": 5.15, + "learning_rate": 4.4889543446244484e-05, + "loss": 1.5267, + "step": 59490 + }, + { + "epoch": 5.15, + "learning_rate": 4.488867712033267e-05, + "loss": 1.6064, + "step": 59500 + }, + { + "epoch": 5.15, + "learning_rate": 4.4887810794420866e-05, + "loss": 1.5462, + "step": 59510 + }, + { + "epoch": 5.15, + "learning_rate": 4.4886944468509054e-05, + "loss": 1.6881, + "step": 59520 + }, + { + "epoch": 5.15, + "learning_rate": 4.488607814259725e-05, + "loss": 1.6251, + "step": 59530 + }, + { + "epoch": 5.15, + "learning_rate": 4.488521181668544e-05, + "loss": 1.5966, + "step": 59540 + }, + { + "epoch": 5.15, + "learning_rate": 4.488434549077363e-05, + "loss": 1.4842, + "step": 59550 + }, + { + "epoch": 5.16, + "learning_rate": 4.4883479164861825e-05, + "loss": 1.6544, + "step": 59560 + }, + { + "epoch": 5.16, + "learning_rate": 4.488261283895002e-05, + "loss": 1.5861, + "step": 59570 + }, + { + "epoch": 5.16, + "learning_rate": 4.488174651303821e-05, + "loss": 1.529, + "step": 59580 + }, + { + "epoch": 5.16, + "learning_rate": 4.48808801871264e-05, + "loss": 1.5399, + "step": 59590 + }, + { + "epoch": 5.16, + "learning_rate": 4.4880013861214596e-05, + "loss": 1.547, + "step": 59600 + }, + { + "epoch": 5.16, + "learning_rate": 4.4879147535302783e-05, + "loss": 1.6588, + "step": 59610 + }, + { + "epoch": 5.16, + "learning_rate": 4.487828120939098e-05, + "loss": 1.5848, + "step": 59620 + }, + { + "epoch": 5.16, + "learning_rate": 4.4877414883479166e-05, + "loss": 1.611, + "step": 59630 + }, + { + "epoch": 5.16, + "learning_rate": 4.487654855756736e-05, + "loss": 1.5764, + "step": 59640 + }, + { + "epoch": 5.16, + "learning_rate": 4.4875682231655554e-05, + "loss": 1.5582, + "step": 59650 + }, + { + "epoch": 5.16, + "learning_rate": 4.487481590574374e-05, + "loss": 1.5409, + "step": 59660 + }, + { + "epoch": 5.16, + "learning_rate": 4.4873949579831937e-05, + "loss": 1.5023, + "step": 59670 + }, + { + "epoch": 5.17, + "learning_rate": 4.487308325392013e-05, + "loss": 1.5267, + "step": 59680 + }, + { + "epoch": 5.17, + "learning_rate": 4.487221692800832e-05, + "loss": 1.5111, + "step": 59690 + }, + { + "epoch": 5.17, + "learning_rate": 4.487135060209651e-05, + "loss": 1.5451, + "step": 59700 + }, + { + "epoch": 5.17, + "learning_rate": 4.487048427618471e-05, + "loss": 1.5413, + "step": 59710 + }, + { + "epoch": 5.17, + "learning_rate": 4.4869617950272895e-05, + "loss": 1.5477, + "step": 59720 + }, + { + "epoch": 5.17, + "learning_rate": 4.486875162436109e-05, + "loss": 1.5752, + "step": 59730 + }, + { + "epoch": 5.17, + "learning_rate": 4.486788529844928e-05, + "loss": 1.5537, + "step": 59740 + }, + { + "epoch": 5.17, + "learning_rate": 4.486701897253747e-05, + "loss": 1.5565, + "step": 59750 + }, + { + "epoch": 5.17, + "learning_rate": 4.4866152646625666e-05, + "loss": 1.5413, + "step": 59760 + }, + { + "epoch": 5.17, + "learning_rate": 4.4865286320713854e-05, + "loss": 1.5701, + "step": 59770 + }, + { + "epoch": 5.17, + "learning_rate": 4.486441999480205e-05, + "loss": 1.5581, + "step": 59780 + }, + { + "epoch": 5.18, + "learning_rate": 4.486355366889024e-05, + "loss": 1.5379, + "step": 59790 + }, + { + "epoch": 5.18, + "learning_rate": 4.486268734297843e-05, + "loss": 1.5174, + "step": 59800 + }, + { + "epoch": 5.18, + "learning_rate": 4.4861821017066625e-05, + "loss": 1.4967, + "step": 59810 + }, + { + "epoch": 5.18, + "learning_rate": 4.486095469115482e-05, + "loss": 1.6126, + "step": 59820 + }, + { + "epoch": 5.18, + "learning_rate": 4.486008836524301e-05, + "loss": 1.5597, + "step": 59830 + }, + { + "epoch": 5.18, + "learning_rate": 4.48592220393312e-05, + "loss": 1.5853, + "step": 59840 + }, + { + "epoch": 5.18, + "learning_rate": 4.485835571341939e-05, + "loss": 1.5422, + "step": 59850 + }, + { + "epoch": 5.18, + "learning_rate": 4.4857489387507584e-05, + "loss": 1.5849, + "step": 59860 + }, + { + "epoch": 5.18, + "learning_rate": 4.485662306159578e-05, + "loss": 1.5393, + "step": 59870 + }, + { + "epoch": 5.18, + "learning_rate": 4.4855756735683966e-05, + "loss": 1.5216, + "step": 59880 + }, + { + "epoch": 5.18, + "learning_rate": 4.485489040977216e-05, + "loss": 1.5792, + "step": 59890 + }, + { + "epoch": 5.18, + "learning_rate": 4.4854024083860355e-05, + "loss": 1.5748, + "step": 59900 + }, + { + "epoch": 5.19, + "learning_rate": 4.485315775794854e-05, + "loss": 1.5706, + "step": 59910 + }, + { + "epoch": 5.19, + "learning_rate": 4.485229143203674e-05, + "loss": 1.6139, + "step": 59920 + }, + { + "epoch": 5.19, + "learning_rate": 4.4851425106124925e-05, + "loss": 1.5349, + "step": 59930 + }, + { + "epoch": 5.19, + "learning_rate": 4.485055878021312e-05, + "loss": 1.6146, + "step": 59940 + }, + { + "epoch": 5.19, + "learning_rate": 4.4849692454301314e-05, + "loss": 1.5642, + "step": 59950 + }, + { + "epoch": 5.19, + "learning_rate": 4.48488261283895e-05, + "loss": 1.5754, + "step": 59960 + }, + { + "epoch": 5.19, + "learning_rate": 4.4847959802477696e-05, + "loss": 1.6197, + "step": 59970 + }, + { + "epoch": 5.19, + "learning_rate": 4.484709347656589e-05, + "loss": 1.5673, + "step": 59980 + }, + { + "epoch": 5.19, + "learning_rate": 4.484622715065408e-05, + "loss": 1.5581, + "step": 59990 + }, + { + "epoch": 5.19, + "learning_rate": 4.484536082474227e-05, + "loss": 1.5033, + "step": 60000 + }, + { + "epoch": 5.19, + "learning_rate": 4.484449449883047e-05, + "loss": 1.5928, + "step": 60010 + }, + { + "epoch": 5.19, + "learning_rate": 4.4843628172918654e-05, + "loss": 1.5747, + "step": 60020 + }, + { + "epoch": 5.2, + "learning_rate": 4.484276184700685e-05, + "loss": 1.571, + "step": 60030 + }, + { + "epoch": 5.2, + "learning_rate": 4.4841895521095036e-05, + "loss": 1.5362, + "step": 60040 + }, + { + "epoch": 5.2, + "learning_rate": 4.484102919518323e-05, + "loss": 1.5163, + "step": 60050 + }, + { + "epoch": 5.2, + "learning_rate": 4.4840162869271425e-05, + "loss": 1.5919, + "step": 60060 + }, + { + "epoch": 5.2, + "learning_rate": 4.483929654335961e-05, + "loss": 1.5248, + "step": 60070 + }, + { + "epoch": 5.2, + "learning_rate": 4.483843021744781e-05, + "loss": 1.5636, + "step": 60080 + }, + { + "epoch": 5.2, + "learning_rate": 4.4837563891536e-05, + "loss": 1.5772, + "step": 60090 + }, + { + "epoch": 5.2, + "learning_rate": 4.483669756562419e-05, + "loss": 1.5564, + "step": 60100 + }, + { + "epoch": 5.2, + "learning_rate": 4.4835831239712384e-05, + "loss": 1.6043, + "step": 60110 + }, + { + "epoch": 5.2, + "learning_rate": 4.483496491380058e-05, + "loss": 1.5392, + "step": 60120 + }, + { + "epoch": 5.2, + "learning_rate": 4.4834098587888766e-05, + "loss": 1.5, + "step": 60130 + }, + { + "epoch": 5.21, + "learning_rate": 4.483323226197696e-05, + "loss": 1.6161, + "step": 60140 + }, + { + "epoch": 5.21, + "learning_rate": 4.483236593606515e-05, + "loss": 1.4674, + "step": 60150 + }, + { + "epoch": 5.21, + "learning_rate": 4.483149961015334e-05, + "loss": 1.4859, + "step": 60160 + }, + { + "epoch": 5.21, + "learning_rate": 4.483063328424154e-05, + "loss": 1.5417, + "step": 60170 + }, + { + "epoch": 5.21, + "learning_rate": 4.4829766958329725e-05, + "loss": 1.5938, + "step": 60180 + }, + { + "epoch": 5.21, + "learning_rate": 4.482890063241792e-05, + "loss": 1.5893, + "step": 60190 + }, + { + "epoch": 5.21, + "learning_rate": 4.4828034306506114e-05, + "loss": 1.5119, + "step": 60200 + }, + { + "epoch": 5.21, + "learning_rate": 4.48271679805943e-05, + "loss": 1.5623, + "step": 60210 + }, + { + "epoch": 5.21, + "learning_rate": 4.4826301654682496e-05, + "loss": 1.5683, + "step": 60220 + }, + { + "epoch": 5.21, + "learning_rate": 4.482543532877069e-05, + "loss": 1.6003, + "step": 60230 + }, + { + "epoch": 5.21, + "learning_rate": 4.482456900285888e-05, + "loss": 1.55, + "step": 60240 + }, + { + "epoch": 5.21, + "learning_rate": 4.482370267694707e-05, + "loss": 1.6658, + "step": 60250 + }, + { + "epoch": 5.22, + "learning_rate": 4.482283635103526e-05, + "loss": 1.5152, + "step": 60260 + }, + { + "epoch": 5.22, + "learning_rate": 4.4821970025123455e-05, + "loss": 1.5884, + "step": 60270 + }, + { + "epoch": 5.22, + "learning_rate": 4.482110369921165e-05, + "loss": 1.6096, + "step": 60280 + }, + { + "epoch": 5.22, + "learning_rate": 4.482023737329984e-05, + "loss": 1.5676, + "step": 60290 + }, + { + "epoch": 5.22, + "learning_rate": 4.481937104738803e-05, + "loss": 1.5827, + "step": 60300 + }, + { + "epoch": 5.22, + "learning_rate": 4.4818504721476226e-05, + "loss": 1.5147, + "step": 60310 + }, + { + "epoch": 5.22, + "learning_rate": 4.4817638395564413e-05, + "loss": 1.5807, + "step": 60320 + }, + { + "epoch": 5.22, + "learning_rate": 4.481677206965261e-05, + "loss": 1.5297, + "step": 60330 + }, + { + "epoch": 5.22, + "learning_rate": 4.48159057437408e-05, + "loss": 1.5291, + "step": 60340 + }, + { + "epoch": 5.22, + "learning_rate": 4.481503941782899e-05, + "loss": 1.585, + "step": 60350 + }, + { + "epoch": 5.22, + "learning_rate": 4.4814173091917184e-05, + "loss": 1.5266, + "step": 60360 + }, + { + "epoch": 5.23, + "learning_rate": 4.481330676600537e-05, + "loss": 1.5964, + "step": 60370 + }, + { + "epoch": 5.23, + "learning_rate": 4.4812440440093567e-05, + "loss": 1.5598, + "step": 60380 + }, + { + "epoch": 5.23, + "learning_rate": 4.481157411418176e-05, + "loss": 1.576, + "step": 60390 + }, + { + "epoch": 5.23, + "learning_rate": 4.481070778826995e-05, + "loss": 1.6032, + "step": 60400 + }, + { + "epoch": 5.23, + "learning_rate": 4.480984146235814e-05, + "loss": 1.5843, + "step": 60410 + }, + { + "epoch": 5.23, + "learning_rate": 4.480897513644634e-05, + "loss": 1.4697, + "step": 60420 + }, + { + "epoch": 5.23, + "learning_rate": 4.4808108810534525e-05, + "loss": 1.5192, + "step": 60430 + }, + { + "epoch": 5.23, + "learning_rate": 4.480724248462272e-05, + "loss": 1.4962, + "step": 60440 + }, + { + "epoch": 5.23, + "learning_rate": 4.4806376158710914e-05, + "loss": 1.5776, + "step": 60450 + }, + { + "epoch": 5.23, + "learning_rate": 4.48055098327991e-05, + "loss": 1.5594, + "step": 60460 + }, + { + "epoch": 5.23, + "learning_rate": 4.4804643506887296e-05, + "loss": 1.5078, + "step": 60470 + }, + { + "epoch": 5.23, + "learning_rate": 4.4803777180975484e-05, + "loss": 1.5434, + "step": 60480 + }, + { + "epoch": 5.24, + "learning_rate": 4.480291085506368e-05, + "loss": 1.6019, + "step": 60490 + }, + { + "epoch": 5.24, + "learning_rate": 4.480204452915187e-05, + "loss": 1.5622, + "step": 60500 + }, + { + "epoch": 5.24, + "learning_rate": 4.480117820324006e-05, + "loss": 1.5526, + "step": 60510 + }, + { + "epoch": 5.24, + "learning_rate": 4.4800311877328255e-05, + "loss": 1.543, + "step": 60520 + }, + { + "epoch": 5.24, + "learning_rate": 4.479944555141645e-05, + "loss": 1.5773, + "step": 60530 + }, + { + "epoch": 5.24, + "learning_rate": 4.479857922550464e-05, + "loss": 1.5453, + "step": 60540 + }, + { + "epoch": 5.24, + "learning_rate": 4.479771289959283e-05, + "loss": 1.6064, + "step": 60550 + }, + { + "epoch": 5.24, + "learning_rate": 4.4796846573681026e-05, + "loss": 1.5835, + "step": 60560 + }, + { + "epoch": 5.24, + "learning_rate": 4.4795980247769214e-05, + "loss": 1.5858, + "step": 60570 + }, + { + "epoch": 5.24, + "learning_rate": 4.479511392185741e-05, + "loss": 1.5562, + "step": 60580 + }, + { + "epoch": 5.24, + "learning_rate": 4.4794247595945596e-05, + "loss": 1.596, + "step": 60590 + }, + { + "epoch": 5.25, + "learning_rate": 4.479338127003379e-05, + "loss": 1.5795, + "step": 60600 + }, + { + "epoch": 5.25, + "learning_rate": 4.4792514944121985e-05, + "loss": 1.559, + "step": 60610 + }, + { + "epoch": 5.25, + "learning_rate": 4.479164861821017e-05, + "loss": 1.5817, + "step": 60620 + }, + { + "epoch": 5.25, + "learning_rate": 4.479078229229837e-05, + "loss": 1.5492, + "step": 60630 + }, + { + "epoch": 5.25, + "learning_rate": 4.478991596638656e-05, + "loss": 1.529, + "step": 60640 + }, + { + "epoch": 5.25, + "learning_rate": 4.478904964047475e-05, + "loss": 1.5022, + "step": 60650 + }, + { + "epoch": 5.25, + "learning_rate": 4.4788183314562943e-05, + "loss": 1.5898, + "step": 60660 + }, + { + "epoch": 5.25, + "learning_rate": 4.478731698865114e-05, + "loss": 1.5172, + "step": 60670 + }, + { + "epoch": 5.25, + "learning_rate": 4.4786450662739326e-05, + "loss": 1.5839, + "step": 60680 + }, + { + "epoch": 5.25, + "learning_rate": 4.478558433682752e-05, + "loss": 1.5871, + "step": 60690 + }, + { + "epoch": 5.25, + "learning_rate": 4.478471801091571e-05, + "loss": 1.5918, + "step": 60700 + }, + { + "epoch": 5.25, + "learning_rate": 4.47838516850039e-05, + "loss": 1.5503, + "step": 60710 + }, + { + "epoch": 5.26, + "learning_rate": 4.47829853590921e-05, + "loss": 1.6242, + "step": 60720 + }, + { + "epoch": 5.26, + "learning_rate": 4.4782119033180284e-05, + "loss": 1.6039, + "step": 60730 + }, + { + "epoch": 5.26, + "learning_rate": 4.478125270726848e-05, + "loss": 1.5112, + "step": 60740 + }, + { + "epoch": 5.26, + "learning_rate": 4.478038638135667e-05, + "loss": 1.5319, + "step": 60750 + }, + { + "epoch": 5.26, + "learning_rate": 4.477952005544486e-05, + "loss": 1.5392, + "step": 60760 + }, + { + "epoch": 5.26, + "learning_rate": 4.4778653729533055e-05, + "loss": 1.5545, + "step": 60770 + }, + { + "epoch": 5.26, + "learning_rate": 4.477778740362124e-05, + "loss": 1.5831, + "step": 60780 + }, + { + "epoch": 5.26, + "learning_rate": 4.477692107770944e-05, + "loss": 1.5682, + "step": 60790 + }, + { + "epoch": 5.26, + "learning_rate": 4.477605475179763e-05, + "loss": 1.5887, + "step": 60800 + }, + { + "epoch": 5.26, + "learning_rate": 4.477518842588582e-05, + "loss": 1.5228, + "step": 60810 + }, + { + "epoch": 5.26, + "learning_rate": 4.4774322099974014e-05, + "loss": 1.5552, + "step": 60820 + }, + { + "epoch": 5.27, + "learning_rate": 4.477345577406221e-05, + "loss": 1.5401, + "step": 60830 + }, + { + "epoch": 5.27, + "learning_rate": 4.4772589448150396e-05, + "loss": 1.5904, + "step": 60840 + }, + { + "epoch": 5.27, + "learning_rate": 4.477172312223859e-05, + "loss": 1.6307, + "step": 60850 + }, + { + "epoch": 5.27, + "learning_rate": 4.4770856796326785e-05, + "loss": 1.5312, + "step": 60860 + }, + { + "epoch": 5.27, + "learning_rate": 4.476999047041497e-05, + "loss": 1.5865, + "step": 60870 + }, + { + "epoch": 5.27, + "learning_rate": 4.476912414450317e-05, + "loss": 1.6029, + "step": 60880 + }, + { + "epoch": 5.27, + "learning_rate": 4.4768257818591355e-05, + "loss": 1.4893, + "step": 60890 + }, + { + "epoch": 5.27, + "learning_rate": 4.476739149267955e-05, + "loss": 1.593, + "step": 60900 + }, + { + "epoch": 5.27, + "learning_rate": 4.4766525166767744e-05, + "loss": 1.6075, + "step": 60910 + }, + { + "epoch": 5.27, + "learning_rate": 4.476565884085593e-05, + "loss": 1.5428, + "step": 60920 + }, + { + "epoch": 5.27, + "learning_rate": 4.4764792514944126e-05, + "loss": 1.6031, + "step": 60930 + }, + { + "epoch": 5.27, + "learning_rate": 4.476392618903232e-05, + "loss": 1.573, + "step": 60940 + }, + { + "epoch": 5.28, + "learning_rate": 4.476305986312051e-05, + "loss": 1.5467, + "step": 60950 + }, + { + "epoch": 5.28, + "learning_rate": 4.47621935372087e-05, + "loss": 1.5448, + "step": 60960 + }, + { + "epoch": 5.28, + "learning_rate": 4.47613272112969e-05, + "loss": 1.4674, + "step": 60970 + }, + { + "epoch": 5.28, + "learning_rate": 4.4760460885385085e-05, + "loss": 1.5678, + "step": 60980 + }, + { + "epoch": 5.28, + "learning_rate": 4.475959455947328e-05, + "loss": 1.6032, + "step": 60990 + }, + { + "epoch": 5.28, + "learning_rate": 4.475872823356147e-05, + "loss": 1.5571, + "step": 61000 + }, + { + "epoch": 5.28, + "learning_rate": 4.475786190764966e-05, + "loss": 1.6151, + "step": 61010 + }, + { + "epoch": 5.28, + "learning_rate": 4.4756995581737856e-05, + "loss": 1.5989, + "step": 61020 + }, + { + "epoch": 5.28, + "learning_rate": 4.475612925582604e-05, + "loss": 1.5722, + "step": 61030 + }, + { + "epoch": 5.28, + "learning_rate": 4.475526292991424e-05, + "loss": 1.5504, + "step": 61040 + }, + { + "epoch": 5.28, + "learning_rate": 4.475439660400243e-05, + "loss": 1.5352, + "step": 61050 + }, + { + "epoch": 5.28, + "learning_rate": 4.475353027809062e-05, + "loss": 1.6042, + "step": 61060 + }, + { + "epoch": 5.29, + "learning_rate": 4.4752663952178814e-05, + "loss": 1.6167, + "step": 61070 + }, + { + "epoch": 5.29, + "learning_rate": 4.475179762626701e-05, + "loss": 1.5418, + "step": 61080 + }, + { + "epoch": 5.29, + "learning_rate": 4.4750931300355197e-05, + "loss": 1.6286, + "step": 61090 + }, + { + "epoch": 5.29, + "learning_rate": 4.475006497444339e-05, + "loss": 1.5186, + "step": 61100 + }, + { + "epoch": 5.29, + "learning_rate": 4.474919864853158e-05, + "loss": 1.5307, + "step": 61110 + }, + { + "epoch": 5.29, + "learning_rate": 4.474833232261977e-05, + "loss": 1.5477, + "step": 61120 + }, + { + "epoch": 5.29, + "learning_rate": 4.474746599670797e-05, + "loss": 1.5334, + "step": 61130 + }, + { + "epoch": 5.29, + "learning_rate": 4.4746599670796155e-05, + "loss": 1.6062, + "step": 61140 + }, + { + "epoch": 5.29, + "learning_rate": 4.474573334488435e-05, + "loss": 1.5609, + "step": 61150 + }, + { + "epoch": 5.29, + "learning_rate": 4.4744867018972544e-05, + "loss": 1.6026, + "step": 61160 + }, + { + "epoch": 5.29, + "learning_rate": 4.474400069306073e-05, + "loss": 1.5488, + "step": 61170 + }, + { + "epoch": 5.3, + "learning_rate": 4.4743134367148926e-05, + "loss": 1.6501, + "step": 61180 + }, + { + "epoch": 5.3, + "learning_rate": 4.474226804123712e-05, + "loss": 1.5454, + "step": 61190 + }, + { + "epoch": 5.3, + "learning_rate": 4.474140171532531e-05, + "loss": 1.5819, + "step": 61200 + }, + { + "epoch": 5.3, + "learning_rate": 4.47405353894135e-05, + "loss": 1.562, + "step": 61210 + }, + { + "epoch": 5.3, + "learning_rate": 4.473966906350169e-05, + "loss": 1.5894, + "step": 61220 + }, + { + "epoch": 5.3, + "learning_rate": 4.4738802737589885e-05, + "loss": 1.5668, + "step": 61230 + }, + { + "epoch": 5.3, + "learning_rate": 4.473793641167808e-05, + "loss": 1.5938, + "step": 61240 + }, + { + "epoch": 5.3, + "learning_rate": 4.473707008576627e-05, + "loss": 1.5691, + "step": 61250 + }, + { + "epoch": 5.3, + "learning_rate": 4.473620375985446e-05, + "loss": 1.5477, + "step": 61260 + }, + { + "epoch": 5.3, + "learning_rate": 4.4735337433942656e-05, + "loss": 1.5251, + "step": 61270 + }, + { + "epoch": 5.3, + "learning_rate": 4.4734471108030844e-05, + "loss": 1.5753, + "step": 61280 + }, + { + "epoch": 5.3, + "learning_rate": 4.473360478211904e-05, + "loss": 1.5974, + "step": 61290 + }, + { + "epoch": 5.31, + "learning_rate": 4.473273845620723e-05, + "loss": 1.5182, + "step": 61300 + }, + { + "epoch": 5.31, + "learning_rate": 4.473187213029542e-05, + "loss": 1.5339, + "step": 61310 + }, + { + "epoch": 5.31, + "learning_rate": 4.4731005804383615e-05, + "loss": 1.5283, + "step": 61320 + }, + { + "epoch": 5.31, + "learning_rate": 4.47301394784718e-05, + "loss": 1.602, + "step": 61330 + }, + { + "epoch": 5.31, + "learning_rate": 4.472927315256e-05, + "loss": 1.5549, + "step": 61340 + }, + { + "epoch": 5.31, + "learning_rate": 4.472840682664819e-05, + "loss": 1.5193, + "step": 61350 + }, + { + "epoch": 5.31, + "learning_rate": 4.472754050073638e-05, + "loss": 1.604, + "step": 61360 + }, + { + "epoch": 5.31, + "learning_rate": 4.4726674174824573e-05, + "loss": 1.53, + "step": 61370 + }, + { + "epoch": 5.31, + "learning_rate": 4.472580784891277e-05, + "loss": 1.5684, + "step": 61380 + }, + { + "epoch": 5.31, + "learning_rate": 4.4724941523000956e-05, + "loss": 1.5173, + "step": 61390 + }, + { + "epoch": 5.31, + "learning_rate": 4.472407519708915e-05, + "loss": 1.5521, + "step": 61400 + }, + { + "epoch": 5.32, + "learning_rate": 4.4723208871177344e-05, + "loss": 1.5651, + "step": 61410 + }, + { + "epoch": 5.32, + "learning_rate": 4.472234254526553e-05, + "loss": 1.515, + "step": 61420 + }, + { + "epoch": 5.32, + "learning_rate": 4.472147621935373e-05, + "loss": 1.5265, + "step": 61430 + }, + { + "epoch": 5.32, + "learning_rate": 4.4720609893441914e-05, + "loss": 1.5389, + "step": 61440 + }, + { + "epoch": 5.32, + "learning_rate": 4.471974356753011e-05, + "loss": 1.6207, + "step": 61450 + }, + { + "epoch": 5.32, + "learning_rate": 4.47188772416183e-05, + "loss": 1.5495, + "step": 61460 + }, + { + "epoch": 5.32, + "learning_rate": 4.471801091570649e-05, + "loss": 1.573, + "step": 61470 + }, + { + "epoch": 5.32, + "learning_rate": 4.4717144589794685e-05, + "loss": 1.4855, + "step": 61480 + }, + { + "epoch": 5.32, + "learning_rate": 4.471627826388288e-05, + "loss": 1.5793, + "step": 61490 + }, + { + "epoch": 5.32, + "learning_rate": 4.471541193797107e-05, + "loss": 1.5523, + "step": 61500 + }, + { + "epoch": 5.32, + "learning_rate": 4.471454561205926e-05, + "loss": 1.5336, + "step": 61510 + }, + { + "epoch": 5.32, + "learning_rate": 4.471367928614745e-05, + "loss": 1.5507, + "step": 61520 + }, + { + "epoch": 5.33, + "learning_rate": 4.4712812960235644e-05, + "loss": 1.4433, + "step": 61530 + }, + { + "epoch": 5.33, + "learning_rate": 4.471194663432384e-05, + "loss": 1.5435, + "step": 61540 + }, + { + "epoch": 5.33, + "learning_rate": 4.4711080308412026e-05, + "loss": 1.5124, + "step": 61550 + }, + { + "epoch": 5.33, + "learning_rate": 4.471021398250022e-05, + "loss": 1.6554, + "step": 61560 + }, + { + "epoch": 5.33, + "learning_rate": 4.4709347656588415e-05, + "loss": 1.5081, + "step": 61570 + }, + { + "epoch": 5.33, + "learning_rate": 4.47084813306766e-05, + "loss": 1.5839, + "step": 61580 + }, + { + "epoch": 5.33, + "learning_rate": 4.47076150047648e-05, + "loss": 1.6133, + "step": 61590 + }, + { + "epoch": 5.33, + "learning_rate": 4.470674867885299e-05, + "loss": 1.5545, + "step": 61600 + }, + { + "epoch": 5.33, + "learning_rate": 4.470588235294118e-05, + "loss": 1.5219, + "step": 61610 + }, + { + "epoch": 5.33, + "learning_rate": 4.4705016027029374e-05, + "loss": 1.6047, + "step": 61620 + }, + { + "epoch": 5.33, + "learning_rate": 4.470414970111756e-05, + "loss": 1.5598, + "step": 61630 + }, + { + "epoch": 5.34, + "learning_rate": 4.4703283375205756e-05, + "loss": 1.5416, + "step": 61640 + }, + { + "epoch": 5.34, + "learning_rate": 4.470241704929395e-05, + "loss": 1.6083, + "step": 61650 + }, + { + "epoch": 5.34, + "learning_rate": 4.470155072338214e-05, + "loss": 1.5719, + "step": 61660 + }, + { + "epoch": 5.34, + "learning_rate": 4.470068439747033e-05, + "loss": 1.5062, + "step": 61670 + }, + { + "epoch": 5.34, + "learning_rate": 4.469981807155853e-05, + "loss": 1.6344, + "step": 61680 + }, + { + "epoch": 5.34, + "learning_rate": 4.4698951745646715e-05, + "loss": 1.5096, + "step": 61690 + }, + { + "epoch": 5.34, + "learning_rate": 4.469808541973491e-05, + "loss": 1.5699, + "step": 61700 + }, + { + "epoch": 5.34, + "learning_rate": 4.4697219093823104e-05, + "loss": 1.6253, + "step": 61710 + }, + { + "epoch": 5.34, + "learning_rate": 4.469635276791129e-05, + "loss": 1.583, + "step": 61720 + }, + { + "epoch": 5.34, + "learning_rate": 4.4695486441999486e-05, + "loss": 1.5523, + "step": 61730 + }, + { + "epoch": 5.34, + "learning_rate": 4.469462011608767e-05, + "loss": 1.5275, + "step": 61740 + }, + { + "epoch": 5.34, + "learning_rate": 4.469375379017587e-05, + "loss": 1.5103, + "step": 61750 + }, + { + "epoch": 5.35, + "learning_rate": 4.469288746426406e-05, + "loss": 1.528, + "step": 61760 + }, + { + "epoch": 5.35, + "learning_rate": 4.469202113835225e-05, + "loss": 1.5632, + "step": 61770 + }, + { + "epoch": 5.35, + "learning_rate": 4.4691154812440444e-05, + "loss": 1.5311, + "step": 61780 + }, + { + "epoch": 5.35, + "learning_rate": 4.469028848652864e-05, + "loss": 1.5533, + "step": 61790 + }, + { + "epoch": 5.35, + "learning_rate": 4.4689422160616827e-05, + "loss": 1.5074, + "step": 61800 + }, + { + "epoch": 5.35, + "learning_rate": 4.468855583470502e-05, + "loss": 1.5874, + "step": 61810 + }, + { + "epoch": 5.35, + "learning_rate": 4.4687689508793215e-05, + "loss": 1.5532, + "step": 61820 + }, + { + "epoch": 5.35, + "learning_rate": 4.46868231828814e-05, + "loss": 1.5927, + "step": 61830 + }, + { + "epoch": 5.35, + "learning_rate": 4.46859568569696e-05, + "loss": 1.6081, + "step": 61840 + }, + { + "epoch": 5.35, + "learning_rate": 4.4685090531057785e-05, + "loss": 1.5692, + "step": 61850 + }, + { + "epoch": 5.35, + "learning_rate": 4.468422420514598e-05, + "loss": 1.5191, + "step": 61860 + }, + { + "epoch": 5.36, + "learning_rate": 4.4683357879234174e-05, + "loss": 1.5797, + "step": 61870 + }, + { + "epoch": 5.36, + "learning_rate": 4.468249155332236e-05, + "loss": 1.5428, + "step": 61880 + }, + { + "epoch": 5.36, + "learning_rate": 4.4681625227410556e-05, + "loss": 1.5774, + "step": 61890 + }, + { + "epoch": 5.36, + "learning_rate": 4.468075890149875e-05, + "loss": 1.5505, + "step": 61900 + }, + { + "epoch": 5.36, + "learning_rate": 4.467989257558694e-05, + "loss": 1.5558, + "step": 61910 + }, + { + "epoch": 5.36, + "learning_rate": 4.467902624967513e-05, + "loss": 1.5705, + "step": 61920 + }, + { + "epoch": 5.36, + "learning_rate": 4.467815992376333e-05, + "loss": 1.5739, + "step": 61930 + }, + { + "epoch": 5.36, + "learning_rate": 4.4677293597851515e-05, + "loss": 1.5398, + "step": 61940 + }, + { + "epoch": 5.36, + "learning_rate": 4.467642727193971e-05, + "loss": 1.551, + "step": 61950 + }, + { + "epoch": 5.36, + "learning_rate": 4.46755609460279e-05, + "loss": 1.549, + "step": 61960 + }, + { + "epoch": 5.36, + "learning_rate": 4.467469462011609e-05, + "loss": 1.5224, + "step": 61970 + }, + { + "epoch": 5.36, + "learning_rate": 4.4673828294204286e-05, + "loss": 1.4895, + "step": 61980 + }, + { + "epoch": 5.37, + "learning_rate": 4.4672961968292474e-05, + "loss": 1.5539, + "step": 61990 + }, + { + "epoch": 5.37, + "learning_rate": 4.467209564238067e-05, + "loss": 1.5332, + "step": 62000 + }, + { + "epoch": 5.37, + "learning_rate": 4.467122931646886e-05, + "loss": 1.5608, + "step": 62010 + }, + { + "epoch": 5.37, + "learning_rate": 4.467036299055705e-05, + "loss": 1.5178, + "step": 62020 + }, + { + "epoch": 5.37, + "learning_rate": 4.4669496664645245e-05, + "loss": 1.5528, + "step": 62030 + }, + { + "epoch": 5.37, + "learning_rate": 4.466863033873344e-05, + "loss": 1.5871, + "step": 62040 + }, + { + "epoch": 5.37, + "learning_rate": 4.466776401282163e-05, + "loss": 1.5302, + "step": 62050 + }, + { + "epoch": 5.37, + "learning_rate": 4.466689768690982e-05, + "loss": 1.501, + "step": 62060 + }, + { + "epoch": 5.37, + "learning_rate": 4.466603136099801e-05, + "loss": 1.5308, + "step": 62070 + }, + { + "epoch": 5.37, + "learning_rate": 4.4665165035086203e-05, + "loss": 1.5272, + "step": 62080 + }, + { + "epoch": 5.37, + "learning_rate": 4.46642987091744e-05, + "loss": 1.5006, + "step": 62090 + }, + { + "epoch": 5.37, + "learning_rate": 4.4663432383262586e-05, + "loss": 1.5218, + "step": 62100 + }, + { + "epoch": 5.38, + "learning_rate": 4.466256605735078e-05, + "loss": 1.5819, + "step": 62110 + }, + { + "epoch": 5.38, + "learning_rate": 4.4661699731438974e-05, + "loss": 1.5907, + "step": 62120 + }, + { + "epoch": 5.38, + "learning_rate": 4.466083340552716e-05, + "loss": 1.6058, + "step": 62130 + }, + { + "epoch": 5.38, + "learning_rate": 4.4659967079615357e-05, + "loss": 1.4799, + "step": 62140 + }, + { + "epoch": 5.38, + "learning_rate": 4.465910075370355e-05, + "loss": 1.5239, + "step": 62150 + }, + { + "epoch": 5.38, + "learning_rate": 4.465823442779174e-05, + "loss": 1.6072, + "step": 62160 + }, + { + "epoch": 5.38, + "learning_rate": 4.465736810187993e-05, + "loss": 1.5772, + "step": 62170 + }, + { + "epoch": 5.38, + "learning_rate": 4.465650177596812e-05, + "loss": 1.5425, + "step": 62180 + }, + { + "epoch": 5.38, + "learning_rate": 4.4655635450056315e-05, + "loss": 1.5477, + "step": 62190 + }, + { + "epoch": 5.38, + "learning_rate": 4.465476912414451e-05, + "loss": 1.5582, + "step": 62200 + }, + { + "epoch": 5.38, + "learning_rate": 4.46539027982327e-05, + "loss": 1.5868, + "step": 62210 + }, + { + "epoch": 5.39, + "learning_rate": 4.465303647232089e-05, + "loss": 1.5556, + "step": 62220 + }, + { + "epoch": 5.39, + "learning_rate": 4.4652170146409086e-05, + "loss": 1.5964, + "step": 62230 + }, + { + "epoch": 5.39, + "learning_rate": 4.4651303820497274e-05, + "loss": 1.5543, + "step": 62240 + }, + { + "epoch": 5.39, + "learning_rate": 4.465043749458547e-05, + "loss": 1.5645, + "step": 62250 + }, + { + "epoch": 5.39, + "learning_rate": 4.4649571168673656e-05, + "loss": 1.5322, + "step": 62260 + }, + { + "epoch": 5.39, + "learning_rate": 4.464870484276185e-05, + "loss": 1.5735, + "step": 62270 + }, + { + "epoch": 5.39, + "learning_rate": 4.4647838516850045e-05, + "loss": 1.5813, + "step": 62280 + }, + { + "epoch": 5.39, + "learning_rate": 4.464697219093823e-05, + "loss": 1.605, + "step": 62290 + }, + { + "epoch": 5.39, + "learning_rate": 4.464610586502643e-05, + "loss": 1.5185, + "step": 62300 + }, + { + "epoch": 5.39, + "learning_rate": 4.464523953911462e-05, + "loss": 1.5599, + "step": 62310 + }, + { + "epoch": 5.39, + "learning_rate": 4.464437321320281e-05, + "loss": 1.5712, + "step": 62320 + }, + { + "epoch": 5.39, + "learning_rate": 4.4643506887291004e-05, + "loss": 1.5261, + "step": 62330 + }, + { + "epoch": 5.4, + "learning_rate": 4.46426405613792e-05, + "loss": 1.5264, + "step": 62340 + }, + { + "epoch": 5.4, + "learning_rate": 4.4641774235467386e-05, + "loss": 1.5522, + "step": 62350 + }, + { + "epoch": 5.4, + "learning_rate": 4.464090790955558e-05, + "loss": 1.6009, + "step": 62360 + }, + { + "epoch": 5.4, + "learning_rate": 4.464004158364377e-05, + "loss": 1.5824, + "step": 62370 + }, + { + "epoch": 5.4, + "learning_rate": 4.463917525773196e-05, + "loss": 1.5693, + "step": 62380 + }, + { + "epoch": 5.4, + "learning_rate": 4.463830893182016e-05, + "loss": 1.5159, + "step": 62390 + }, + { + "epoch": 5.4, + "learning_rate": 4.4637442605908345e-05, + "loss": 1.5322, + "step": 62400 + }, + { + "epoch": 5.4, + "learning_rate": 4.463657627999654e-05, + "loss": 1.5839, + "step": 62410 + }, + { + "epoch": 5.4, + "learning_rate": 4.4635709954084734e-05, + "loss": 1.4983, + "step": 62420 + }, + { + "epoch": 5.4, + "learning_rate": 4.463484362817292e-05, + "loss": 1.4627, + "step": 62430 + }, + { + "epoch": 5.4, + "learning_rate": 4.4633977302261116e-05, + "loss": 1.5804, + "step": 62440 + }, + { + "epoch": 5.41, + "learning_rate": 4.463311097634931e-05, + "loss": 1.5826, + "step": 62450 + }, + { + "epoch": 5.41, + "learning_rate": 4.46322446504375e-05, + "loss": 1.5013, + "step": 62460 + }, + { + "epoch": 5.41, + "learning_rate": 4.463137832452569e-05, + "loss": 1.4713, + "step": 62470 + }, + { + "epoch": 5.41, + "learning_rate": 4.463051199861388e-05, + "loss": 1.5617, + "step": 62480 + }, + { + "epoch": 5.41, + "learning_rate": 4.4629645672702074e-05, + "loss": 1.6195, + "step": 62490 + }, + { + "epoch": 5.41, + "learning_rate": 4.462877934679027e-05, + "loss": 1.5573, + "step": 62500 + }, + { + "epoch": 5.41, + "learning_rate": 4.4627913020878456e-05, + "loss": 1.5836, + "step": 62510 + }, + { + "epoch": 5.41, + "learning_rate": 4.462704669496665e-05, + "loss": 1.523, + "step": 62520 + }, + { + "epoch": 5.41, + "learning_rate": 4.4626180369054845e-05, + "loss": 1.5242, + "step": 62530 + }, + { + "epoch": 5.41, + "learning_rate": 4.462531404314303e-05, + "loss": 1.5881, + "step": 62540 + }, + { + "epoch": 5.41, + "learning_rate": 4.462444771723123e-05, + "loss": 1.5494, + "step": 62550 + }, + { + "epoch": 5.41, + "learning_rate": 4.462358139131942e-05, + "loss": 1.597, + "step": 62560 + }, + { + "epoch": 5.42, + "learning_rate": 4.462271506540761e-05, + "loss": 1.5636, + "step": 62570 + }, + { + "epoch": 5.42, + "learning_rate": 4.4621848739495804e-05, + "loss": 1.6106, + "step": 62580 + }, + { + "epoch": 5.42, + "learning_rate": 4.462098241358399e-05, + "loss": 1.5106, + "step": 62590 + }, + { + "epoch": 5.42, + "learning_rate": 4.4620116087672186e-05, + "loss": 1.5655, + "step": 62600 + }, + { + "epoch": 5.42, + "learning_rate": 4.461924976176038e-05, + "loss": 1.5786, + "step": 62610 + }, + { + "epoch": 5.42, + "learning_rate": 4.461838343584857e-05, + "loss": 1.4942, + "step": 62620 + }, + { + "epoch": 5.42, + "learning_rate": 4.461751710993676e-05, + "loss": 1.5593, + "step": 62630 + }, + { + "epoch": 5.42, + "learning_rate": 4.461665078402496e-05, + "loss": 1.6356, + "step": 62640 + }, + { + "epoch": 5.42, + "learning_rate": 4.4615784458113145e-05, + "loss": 1.5772, + "step": 62650 + }, + { + "epoch": 5.42, + "learning_rate": 4.461491813220134e-05, + "loss": 1.5918, + "step": 62660 + }, + { + "epoch": 5.42, + "learning_rate": 4.4614051806289534e-05, + "loss": 1.5458, + "step": 62670 + }, + { + "epoch": 5.43, + "learning_rate": 4.461318548037772e-05, + "loss": 1.5192, + "step": 62680 + }, + { + "epoch": 5.43, + "learning_rate": 4.4612319154465916e-05, + "loss": 1.6354, + "step": 62690 + }, + { + "epoch": 5.43, + "learning_rate": 4.4611452828554104e-05, + "loss": 1.5819, + "step": 62700 + }, + { + "epoch": 5.43, + "learning_rate": 4.46105865026423e-05, + "loss": 1.5683, + "step": 62710 + }, + { + "epoch": 5.43, + "learning_rate": 4.460972017673049e-05, + "loss": 1.5683, + "step": 62720 + }, + { + "epoch": 5.43, + "learning_rate": 4.460885385081868e-05, + "loss": 1.5387, + "step": 62730 + }, + { + "epoch": 5.43, + "learning_rate": 4.4607987524906875e-05, + "loss": 1.5565, + "step": 62740 + }, + { + "epoch": 5.43, + "learning_rate": 4.460712119899507e-05, + "loss": 1.5275, + "step": 62750 + }, + { + "epoch": 5.43, + "learning_rate": 4.460625487308326e-05, + "loss": 1.5919, + "step": 62760 + }, + { + "epoch": 5.43, + "learning_rate": 4.460538854717145e-05, + "loss": 1.6543, + "step": 62770 + }, + { + "epoch": 5.43, + "learning_rate": 4.4604522221259646e-05, + "loss": 1.5279, + "step": 62780 + }, + { + "epoch": 5.43, + "learning_rate": 4.4603655895347833e-05, + "loss": 1.5789, + "step": 62790 + }, + { + "epoch": 5.44, + "learning_rate": 4.460278956943603e-05, + "loss": 1.6213, + "step": 62800 + }, + { + "epoch": 5.44, + "learning_rate": 4.4601923243524216e-05, + "loss": 1.6327, + "step": 62810 + }, + { + "epoch": 5.44, + "learning_rate": 4.460105691761241e-05, + "loss": 1.5469, + "step": 62820 + }, + { + "epoch": 5.44, + "learning_rate": 4.4600190591700604e-05, + "loss": 1.5314, + "step": 62830 + }, + { + "epoch": 5.44, + "learning_rate": 4.459932426578879e-05, + "loss": 1.5927, + "step": 62840 + }, + { + "epoch": 5.44, + "learning_rate": 4.4598457939876987e-05, + "loss": 1.5529, + "step": 62850 + }, + { + "epoch": 5.44, + "learning_rate": 4.459759161396518e-05, + "loss": 1.5182, + "step": 62860 + }, + { + "epoch": 5.44, + "learning_rate": 4.459672528805337e-05, + "loss": 1.581, + "step": 62870 + }, + { + "epoch": 5.44, + "learning_rate": 4.459585896214156e-05, + "loss": 1.5671, + "step": 62880 + }, + { + "epoch": 5.44, + "learning_rate": 4.459499263622976e-05, + "loss": 1.5526, + "step": 62890 + }, + { + "epoch": 5.44, + "learning_rate": 4.4594126310317945e-05, + "loss": 1.5913, + "step": 62900 + }, + { + "epoch": 5.45, + "learning_rate": 4.459325998440614e-05, + "loss": 1.509, + "step": 62910 + }, + { + "epoch": 5.45, + "learning_rate": 4.459239365849433e-05, + "loss": 1.4949, + "step": 62920 + }, + { + "epoch": 5.45, + "learning_rate": 4.459152733258252e-05, + "loss": 1.556, + "step": 62930 + }, + { + "epoch": 5.45, + "learning_rate": 4.4590661006670716e-05, + "loss": 1.5884, + "step": 62940 + }, + { + "epoch": 5.45, + "learning_rate": 4.4589794680758904e-05, + "loss": 1.5372, + "step": 62950 + }, + { + "epoch": 5.45, + "learning_rate": 4.45889283548471e-05, + "loss": 1.5603, + "step": 62960 + }, + { + "epoch": 5.45, + "learning_rate": 4.458806202893529e-05, + "loss": 1.5682, + "step": 62970 + }, + { + "epoch": 5.45, + "learning_rate": 4.458719570302348e-05, + "loss": 1.6139, + "step": 62980 + }, + { + "epoch": 5.45, + "learning_rate": 4.4586329377111675e-05, + "loss": 1.5893, + "step": 62990 + }, + { + "epoch": 5.45, + "learning_rate": 4.458546305119987e-05, + "loss": 1.5901, + "step": 63000 + }, + { + "epoch": 5.45, + "learning_rate": 4.458459672528806e-05, + "loss": 1.5901, + "step": 63010 + }, + { + "epoch": 5.45, + "learning_rate": 4.458373039937625e-05, + "loss": 1.5097, + "step": 63020 + }, + { + "epoch": 5.46, + "learning_rate": 4.458286407346444e-05, + "loss": 1.5429, + "step": 63030 + }, + { + "epoch": 5.46, + "learning_rate": 4.4581997747552634e-05, + "loss": 1.567, + "step": 63040 + }, + { + "epoch": 5.46, + "learning_rate": 4.458113142164083e-05, + "loss": 1.5875, + "step": 63050 + }, + { + "epoch": 5.46, + "learning_rate": 4.4580265095729016e-05, + "loss": 1.5931, + "step": 63060 + }, + { + "epoch": 5.46, + "learning_rate": 4.457939876981721e-05, + "loss": 1.5133, + "step": 63070 + }, + { + "epoch": 5.46, + "learning_rate": 4.4578532443905405e-05, + "loss": 1.5191, + "step": 63080 + }, + { + "epoch": 5.46, + "learning_rate": 4.457766611799359e-05, + "loss": 1.5733, + "step": 63090 + }, + { + "epoch": 5.46, + "learning_rate": 4.457679979208179e-05, + "loss": 1.5452, + "step": 63100 + }, + { + "epoch": 5.46, + "learning_rate": 4.4575933466169975e-05, + "loss": 1.5595, + "step": 63110 + }, + { + "epoch": 5.46, + "learning_rate": 4.457506714025817e-05, + "loss": 1.5276, + "step": 63120 + }, + { + "epoch": 5.46, + "learning_rate": 4.4574200814346363e-05, + "loss": 1.5605, + "step": 63130 + }, + { + "epoch": 5.47, + "learning_rate": 4.457333448843455e-05, + "loss": 1.5243, + "step": 63140 + }, + { + "epoch": 5.47, + "learning_rate": 4.4572468162522746e-05, + "loss": 1.4584, + "step": 63150 + }, + { + "epoch": 5.47, + "learning_rate": 4.457160183661094e-05, + "loss": 1.4881, + "step": 63160 + }, + { + "epoch": 5.47, + "learning_rate": 4.457073551069913e-05, + "loss": 1.5277, + "step": 63170 + }, + { + "epoch": 5.47, + "learning_rate": 4.456986918478732e-05, + "loss": 1.5992, + "step": 63180 + }, + { + "epoch": 5.47, + "learning_rate": 4.456900285887552e-05, + "loss": 1.4904, + "step": 63190 + }, + { + "epoch": 5.47, + "learning_rate": 4.4568136532963704e-05, + "loss": 1.6294, + "step": 63200 + }, + { + "epoch": 5.47, + "learning_rate": 4.45672702070519e-05, + "loss": 1.5371, + "step": 63210 + }, + { + "epoch": 5.47, + "learning_rate": 4.4566403881140086e-05, + "loss": 1.5594, + "step": 63220 + }, + { + "epoch": 5.47, + "learning_rate": 4.456553755522828e-05, + "loss": 1.6069, + "step": 63230 + }, + { + "epoch": 5.47, + "learning_rate": 4.4564671229316475e-05, + "loss": 1.5852, + "step": 63240 + }, + { + "epoch": 5.47, + "learning_rate": 4.456380490340466e-05, + "loss": 1.5584, + "step": 63250 + }, + { + "epoch": 5.48, + "learning_rate": 4.456293857749286e-05, + "loss": 1.5561, + "step": 63260 + }, + { + "epoch": 5.48, + "learning_rate": 4.456207225158105e-05, + "loss": 1.5586, + "step": 63270 + }, + { + "epoch": 5.48, + "learning_rate": 4.456120592566924e-05, + "loss": 1.6291, + "step": 63280 + }, + { + "epoch": 5.48, + "learning_rate": 4.4560339599757434e-05, + "loss": 1.5261, + "step": 63290 + }, + { + "epoch": 5.48, + "learning_rate": 4.455947327384563e-05, + "loss": 1.5609, + "step": 63300 + }, + { + "epoch": 5.48, + "learning_rate": 4.4558606947933816e-05, + "loss": 1.5939, + "step": 63310 + }, + { + "epoch": 5.48, + "learning_rate": 4.455774062202201e-05, + "loss": 1.5264, + "step": 63320 + }, + { + "epoch": 5.48, + "learning_rate": 4.45568742961102e-05, + "loss": 1.6158, + "step": 63330 + }, + { + "epoch": 5.48, + "learning_rate": 4.455600797019839e-05, + "loss": 1.5987, + "step": 63340 + }, + { + "epoch": 5.48, + "learning_rate": 4.455514164428659e-05, + "loss": 1.5448, + "step": 63350 + }, + { + "epoch": 5.48, + "learning_rate": 4.4554275318374775e-05, + "loss": 1.5337, + "step": 63360 + }, + { + "epoch": 5.48, + "learning_rate": 4.455340899246297e-05, + "loss": 1.5652, + "step": 63370 + }, + { + "epoch": 5.49, + "learning_rate": 4.4552542666551164e-05, + "loss": 1.5748, + "step": 63380 + }, + { + "epoch": 5.49, + "learning_rate": 4.455167634063935e-05, + "loss": 1.6104, + "step": 63390 + }, + { + "epoch": 5.49, + "learning_rate": 4.4550810014727546e-05, + "loss": 1.4497, + "step": 63400 + }, + { + "epoch": 5.49, + "learning_rate": 4.454994368881574e-05, + "loss": 1.5665, + "step": 63410 + }, + { + "epoch": 5.49, + "learning_rate": 4.454907736290393e-05, + "loss": 1.5861, + "step": 63420 + }, + { + "epoch": 5.49, + "learning_rate": 4.454821103699212e-05, + "loss": 1.577, + "step": 63430 + }, + { + "epoch": 5.49, + "learning_rate": 4.454734471108031e-05, + "loss": 1.5982, + "step": 63440 + }, + { + "epoch": 5.49, + "learning_rate": 4.4546478385168505e-05, + "loss": 1.5531, + "step": 63450 + }, + { + "epoch": 5.49, + "learning_rate": 4.45456120592567e-05, + "loss": 1.561, + "step": 63460 + }, + { + "epoch": 5.49, + "learning_rate": 4.454474573334489e-05, + "loss": 1.5451, + "step": 63470 + }, + { + "epoch": 5.49, + "learning_rate": 4.454387940743308e-05, + "loss": 1.5364, + "step": 63480 + }, + { + "epoch": 5.5, + "learning_rate": 4.4543013081521276e-05, + "loss": 1.5313, + "step": 63490 + }, + { + "epoch": 5.5, + "learning_rate": 4.454214675560946e-05, + "loss": 1.5138, + "step": 63500 + }, + { + "epoch": 5.5, + "learning_rate": 4.454128042969766e-05, + "loss": 1.5496, + "step": 63510 + }, + { + "epoch": 5.5, + "learning_rate": 4.454041410378585e-05, + "loss": 1.4909, + "step": 63520 + }, + { + "epoch": 5.5, + "learning_rate": 4.453954777787404e-05, + "loss": 1.5644, + "step": 63530 + }, + { + "epoch": 5.5, + "learning_rate": 4.4538681451962234e-05, + "loss": 1.5243, + "step": 63540 + }, + { + "epoch": 5.5, + "learning_rate": 4.453781512605042e-05, + "loss": 1.5692, + "step": 63550 + }, + { + "epoch": 5.5, + "learning_rate": 4.4536948800138617e-05, + "loss": 1.4918, + "step": 63560 + }, + { + "epoch": 5.5, + "learning_rate": 4.453608247422681e-05, + "loss": 1.543, + "step": 63570 + }, + { + "epoch": 5.5, + "learning_rate": 4.4535216148315e-05, + "loss": 1.5952, + "step": 63580 + }, + { + "epoch": 5.5, + "learning_rate": 4.453434982240319e-05, + "loss": 1.545, + "step": 63590 + }, + { + "epoch": 5.5, + "learning_rate": 4.453348349649139e-05, + "loss": 1.5965, + "step": 63600 + }, + { + "epoch": 5.51, + "learning_rate": 4.4532617170579575e-05, + "loss": 1.5484, + "step": 63610 + }, + { + "epoch": 5.51, + "learning_rate": 4.453175084466777e-05, + "loss": 1.5445, + "step": 63620 + }, + { + "epoch": 5.51, + "learning_rate": 4.453088451875596e-05, + "loss": 1.5631, + "step": 63630 + }, + { + "epoch": 5.51, + "learning_rate": 4.4530018192844145e-05, + "loss": 1.5765, + "step": 63640 + }, + { + "epoch": 5.51, + "learning_rate": 4.452915186693234e-05, + "loss": 1.5158, + "step": 63650 + }, + { + "epoch": 5.51, + "learning_rate": 4.4528285541020534e-05, + "loss": 1.5383, + "step": 63660 + }, + { + "epoch": 5.51, + "learning_rate": 4.452741921510872e-05, + "loss": 1.5954, + "step": 63670 + }, + { + "epoch": 5.51, + "learning_rate": 4.4526552889196916e-05, + "loss": 1.5515, + "step": 63680 + }, + { + "epoch": 5.51, + "learning_rate": 4.452568656328511e-05, + "loss": 1.5803, + "step": 63690 + }, + { + "epoch": 5.51, + "learning_rate": 4.45248202373733e-05, + "loss": 1.5086, + "step": 63700 + }, + { + "epoch": 5.51, + "learning_rate": 4.452395391146149e-05, + "loss": 1.5779, + "step": 63710 + }, + { + "epoch": 5.52, + "learning_rate": 4.452308758554968e-05, + "loss": 1.5417, + "step": 63720 + }, + { + "epoch": 5.52, + "learning_rate": 4.4522221259637875e-05, + "loss": 1.531, + "step": 63730 + }, + { + "epoch": 5.52, + "learning_rate": 4.452135493372607e-05, + "loss": 1.5551, + "step": 63740 + }, + { + "epoch": 5.52, + "learning_rate": 4.452048860781426e-05, + "loss": 1.6174, + "step": 63750 + }, + { + "epoch": 5.52, + "learning_rate": 4.451962228190245e-05, + "loss": 1.6204, + "step": 63760 + }, + { + "epoch": 5.52, + "learning_rate": 4.4518755955990646e-05, + "loss": 1.5917, + "step": 63770 + }, + { + "epoch": 5.52, + "learning_rate": 4.4517889630078834e-05, + "loss": 1.5587, + "step": 63780 + }, + { + "epoch": 5.52, + "learning_rate": 4.451702330416703e-05, + "loss": 1.5768, + "step": 63790 + }, + { + "epoch": 5.52, + "learning_rate": 4.451615697825522e-05, + "loss": 1.5197, + "step": 63800 + }, + { + "epoch": 5.52, + "learning_rate": 4.451529065234341e-05, + "loss": 1.5618, + "step": 63810 + }, + { + "epoch": 5.52, + "learning_rate": 4.4514424326431605e-05, + "loss": 1.6265, + "step": 63820 + }, + { + "epoch": 5.52, + "learning_rate": 4.451355800051979e-05, + "loss": 1.5626, + "step": 63830 + }, + { + "epoch": 5.53, + "learning_rate": 4.451269167460799e-05, + "loss": 1.5322, + "step": 63840 + }, + { + "epoch": 5.53, + "learning_rate": 4.451182534869618e-05, + "loss": 1.5699, + "step": 63850 + }, + { + "epoch": 5.53, + "learning_rate": 4.451095902278437e-05, + "loss": 1.5924, + "step": 63860 + }, + { + "epoch": 5.53, + "learning_rate": 4.451009269687256e-05, + "loss": 1.5315, + "step": 63870 + }, + { + "epoch": 5.53, + "learning_rate": 4.450922637096076e-05, + "loss": 1.6239, + "step": 63880 + }, + { + "epoch": 5.53, + "learning_rate": 4.4508360045048945e-05, + "loss": 1.5822, + "step": 63890 + }, + { + "epoch": 5.53, + "learning_rate": 4.450749371913714e-05, + "loss": 1.5034, + "step": 63900 + }, + { + "epoch": 5.53, + "learning_rate": 4.450662739322533e-05, + "loss": 1.5712, + "step": 63910 + }, + { + "epoch": 5.53, + "learning_rate": 4.450576106731352e-05, + "loss": 1.5388, + "step": 63920 + }, + { + "epoch": 5.53, + "learning_rate": 4.4504894741401716e-05, + "loss": 1.5255, + "step": 63930 + }, + { + "epoch": 5.53, + "learning_rate": 4.4504028415489904e-05, + "loss": 1.5034, + "step": 63940 + }, + { + "epoch": 5.54, + "learning_rate": 4.45031620895781e-05, + "loss": 1.5177, + "step": 63950 + }, + { + "epoch": 5.54, + "learning_rate": 4.450229576366629e-05, + "loss": 1.5356, + "step": 63960 + }, + { + "epoch": 5.54, + "learning_rate": 4.450142943775448e-05, + "loss": 1.5238, + "step": 63970 + }, + { + "epoch": 5.54, + "learning_rate": 4.4500563111842675e-05, + "loss": 1.548, + "step": 63980 + }, + { + "epoch": 5.54, + "learning_rate": 4.449969678593087e-05, + "loss": 1.562, + "step": 63990 + }, + { + "epoch": 5.54, + "learning_rate": 4.449883046001906e-05, + "loss": 1.5477, + "step": 64000 + }, + { + "epoch": 5.54, + "learning_rate": 4.449796413410725e-05, + "loss": 1.5975, + "step": 64010 + }, + { + "epoch": 5.54, + "learning_rate": 4.449709780819544e-05, + "loss": 1.5565, + "step": 64020 + }, + { + "epoch": 5.54, + "learning_rate": 4.4496231482283634e-05, + "loss": 1.5239, + "step": 64030 + }, + { + "epoch": 5.54, + "learning_rate": 4.449536515637183e-05, + "loss": 1.5536, + "step": 64040 + }, + { + "epoch": 5.54, + "learning_rate": 4.4494498830460016e-05, + "loss": 1.5909, + "step": 64050 + }, + { + "epoch": 5.54, + "learning_rate": 4.449363250454821e-05, + "loss": 1.5213, + "step": 64060 + }, + { + "epoch": 5.55, + "learning_rate": 4.4492766178636405e-05, + "loss": 1.5802, + "step": 64070 + }, + { + "epoch": 5.55, + "learning_rate": 4.449189985272459e-05, + "loss": 1.5031, + "step": 64080 + }, + { + "epoch": 5.55, + "learning_rate": 4.449103352681279e-05, + "loss": 1.6024, + "step": 64090 + }, + { + "epoch": 5.55, + "learning_rate": 4.449016720090098e-05, + "loss": 1.6076, + "step": 64100 + }, + { + "epoch": 5.55, + "learning_rate": 4.448930087498917e-05, + "loss": 1.4989, + "step": 64110 + }, + { + "epoch": 5.55, + "learning_rate": 4.4488434549077364e-05, + "loss": 1.5494, + "step": 64120 + }, + { + "epoch": 5.55, + "learning_rate": 4.448756822316555e-05, + "loss": 1.5721, + "step": 64130 + }, + { + "epoch": 5.55, + "learning_rate": 4.4486701897253746e-05, + "loss": 1.5441, + "step": 64140 + }, + { + "epoch": 5.55, + "learning_rate": 4.448583557134194e-05, + "loss": 1.5973, + "step": 64150 + }, + { + "epoch": 5.55, + "learning_rate": 4.448496924543013e-05, + "loss": 1.5276, + "step": 64160 + }, + { + "epoch": 5.55, + "learning_rate": 4.448410291951832e-05, + "loss": 1.5718, + "step": 64170 + }, + { + "epoch": 5.56, + "learning_rate": 4.448323659360652e-05, + "loss": 1.5145, + "step": 64180 + }, + { + "epoch": 5.56, + "learning_rate": 4.4482370267694704e-05, + "loss": 1.5255, + "step": 64190 + }, + { + "epoch": 5.56, + "learning_rate": 4.44815039417829e-05, + "loss": 1.5494, + "step": 64200 + }, + { + "epoch": 5.56, + "learning_rate": 4.448063761587109e-05, + "loss": 1.5577, + "step": 64210 + }, + { + "epoch": 5.56, + "learning_rate": 4.447977128995928e-05, + "loss": 1.5118, + "step": 64220 + }, + { + "epoch": 5.56, + "learning_rate": 4.4478904964047475e-05, + "loss": 1.5906, + "step": 64230 + }, + { + "epoch": 5.56, + "learning_rate": 4.447803863813566e-05, + "loss": 1.515, + "step": 64240 + }, + { + "epoch": 5.56, + "learning_rate": 4.447717231222386e-05, + "loss": 1.5755, + "step": 64250 + }, + { + "epoch": 5.56, + "learning_rate": 4.447630598631205e-05, + "loss": 1.5163, + "step": 64260 + }, + { + "epoch": 5.56, + "learning_rate": 4.447543966040024e-05, + "loss": 1.581, + "step": 64270 + }, + { + "epoch": 5.56, + "learning_rate": 4.4474573334488434e-05, + "loss": 1.5224, + "step": 64280 + }, + { + "epoch": 5.56, + "learning_rate": 4.447370700857663e-05, + "loss": 1.504, + "step": 64290 + }, + { + "epoch": 5.57, + "learning_rate": 4.4472840682664816e-05, + "loss": 1.5563, + "step": 64300 + }, + { + "epoch": 5.57, + "learning_rate": 4.447197435675301e-05, + "loss": 1.5509, + "step": 64310 + }, + { + "epoch": 5.57, + "learning_rate": 4.4471108030841205e-05, + "loss": 1.5916, + "step": 64320 + }, + { + "epoch": 5.57, + "learning_rate": 4.447024170492939e-05, + "loss": 1.5248, + "step": 64330 + }, + { + "epoch": 5.57, + "learning_rate": 4.446937537901759e-05, + "loss": 1.6144, + "step": 64340 + }, + { + "epoch": 5.57, + "learning_rate": 4.4468509053105775e-05, + "loss": 1.5698, + "step": 64350 + }, + { + "epoch": 5.57, + "learning_rate": 4.446764272719397e-05, + "loss": 1.5598, + "step": 64360 + }, + { + "epoch": 5.57, + "learning_rate": 4.4466776401282164e-05, + "loss": 1.5765, + "step": 64370 + }, + { + "epoch": 5.57, + "learning_rate": 4.446591007537035e-05, + "loss": 1.5663, + "step": 64380 + }, + { + "epoch": 5.57, + "learning_rate": 4.4465043749458546e-05, + "loss": 1.5651, + "step": 64390 + }, + { + "epoch": 5.57, + "learning_rate": 4.446417742354674e-05, + "loss": 1.5284, + "step": 64400 + }, + { + "epoch": 5.57, + "learning_rate": 4.446331109763493e-05, + "loss": 1.6324, + "step": 64410 + }, + { + "epoch": 5.58, + "learning_rate": 4.446244477172312e-05, + "loss": 1.5083, + "step": 64420 + }, + { + "epoch": 5.58, + "learning_rate": 4.446157844581132e-05, + "loss": 1.562, + "step": 64430 + }, + { + "epoch": 5.58, + "learning_rate": 4.4460712119899505e-05, + "loss": 1.5771, + "step": 64440 + }, + { + "epoch": 5.58, + "learning_rate": 4.44598457939877e-05, + "loss": 1.5968, + "step": 64450 + }, + { + "epoch": 5.58, + "learning_rate": 4.445897946807589e-05, + "loss": 1.5575, + "step": 64460 + }, + { + "epoch": 5.58, + "learning_rate": 4.445811314216408e-05, + "loss": 1.5765, + "step": 64470 + }, + { + "epoch": 5.58, + "learning_rate": 4.4457246816252276e-05, + "loss": 1.5756, + "step": 64480 + }, + { + "epoch": 5.58, + "learning_rate": 4.4456380490340463e-05, + "loss": 1.5293, + "step": 64490 + }, + { + "epoch": 5.58, + "learning_rate": 4.445551416442866e-05, + "loss": 1.5521, + "step": 64500 + }, + { + "epoch": 5.58, + "learning_rate": 4.445464783851685e-05, + "loss": 1.5712, + "step": 64510 + }, + { + "epoch": 5.58, + "learning_rate": 4.445378151260504e-05, + "loss": 1.5544, + "step": 64520 + }, + { + "epoch": 5.59, + "learning_rate": 4.4452915186693235e-05, + "loss": 1.4934, + "step": 64530 + }, + { + "epoch": 5.59, + "learning_rate": 4.445204886078143e-05, + "loss": 1.5601, + "step": 64540 + }, + { + "epoch": 5.59, + "learning_rate": 4.445118253486962e-05, + "loss": 1.5763, + "step": 64550 + }, + { + "epoch": 5.59, + "learning_rate": 4.445031620895781e-05, + "loss": 1.5909, + "step": 64560 + }, + { + "epoch": 5.59, + "learning_rate": 4.4449449883046e-05, + "loss": 1.5922, + "step": 64570 + }, + { + "epoch": 5.59, + "learning_rate": 4.444858355713419e-05, + "loss": 1.5886, + "step": 64580 + }, + { + "epoch": 5.59, + "learning_rate": 4.444771723122239e-05, + "loss": 1.5679, + "step": 64590 + }, + { + "epoch": 5.59, + "learning_rate": 4.4446850905310575e-05, + "loss": 1.5429, + "step": 64600 + }, + { + "epoch": 5.59, + "learning_rate": 4.444598457939877e-05, + "loss": 1.5424, + "step": 64610 + }, + { + "epoch": 5.59, + "learning_rate": 4.4445118253486964e-05, + "loss": 1.6322, + "step": 64620 + }, + { + "epoch": 5.59, + "learning_rate": 4.444425192757515e-05, + "loss": 1.5752, + "step": 64630 + }, + { + "epoch": 5.59, + "learning_rate": 4.4443385601663346e-05, + "loss": 1.6093, + "step": 64640 + }, + { + "epoch": 5.6, + "learning_rate": 4.4442519275751534e-05, + "loss": 1.5531, + "step": 64650 + }, + { + "epoch": 5.6, + "learning_rate": 4.444165294983973e-05, + "loss": 1.5416, + "step": 64660 + }, + { + "epoch": 5.6, + "learning_rate": 4.444078662392792e-05, + "loss": 1.5451, + "step": 64670 + }, + { + "epoch": 5.6, + "learning_rate": 4.443992029801611e-05, + "loss": 1.5199, + "step": 64680 + }, + { + "epoch": 5.6, + "learning_rate": 4.4439053972104305e-05, + "loss": 1.6143, + "step": 64690 + }, + { + "epoch": 5.6, + "learning_rate": 4.44381876461925e-05, + "loss": 1.5217, + "step": 64700 + }, + { + "epoch": 5.6, + "learning_rate": 4.443732132028069e-05, + "loss": 1.5394, + "step": 64710 + }, + { + "epoch": 5.6, + "learning_rate": 4.443645499436888e-05, + "loss": 1.5266, + "step": 64720 + }, + { + "epoch": 5.6, + "learning_rate": 4.4435588668457076e-05, + "loss": 1.5718, + "step": 64730 + }, + { + "epoch": 5.6, + "learning_rate": 4.4434722342545264e-05, + "loss": 1.5896, + "step": 64740 + }, + { + "epoch": 5.6, + "learning_rate": 4.443385601663346e-05, + "loss": 1.5072, + "step": 64750 + }, + { + "epoch": 5.61, + "learning_rate": 4.4432989690721646e-05, + "loss": 1.579, + "step": 64760 + }, + { + "epoch": 5.61, + "learning_rate": 4.443212336480984e-05, + "loss": 1.5535, + "step": 64770 + }, + { + "epoch": 5.61, + "learning_rate": 4.4431257038898035e-05, + "loss": 1.5449, + "step": 64780 + }, + { + "epoch": 5.61, + "learning_rate": 4.443039071298622e-05, + "loss": 1.5385, + "step": 64790 + }, + { + "epoch": 5.61, + "learning_rate": 4.442952438707442e-05, + "loss": 1.5818, + "step": 64800 + }, + { + "epoch": 5.61, + "learning_rate": 4.442865806116261e-05, + "loss": 1.5381, + "step": 64810 + }, + { + "epoch": 5.61, + "learning_rate": 4.44277917352508e-05, + "loss": 1.5127, + "step": 64820 + }, + { + "epoch": 5.61, + "learning_rate": 4.4426925409338994e-05, + "loss": 1.5508, + "step": 64830 + }, + { + "epoch": 5.61, + "learning_rate": 4.442605908342719e-05, + "loss": 1.57, + "step": 64840 + }, + { + "epoch": 5.61, + "learning_rate": 4.4425192757515376e-05, + "loss": 1.5961, + "step": 64850 + }, + { + "epoch": 5.61, + "learning_rate": 4.442432643160357e-05, + "loss": 1.5943, + "step": 64860 + }, + { + "epoch": 5.61, + "learning_rate": 4.442346010569176e-05, + "loss": 1.5404, + "step": 64870 + }, + { + "epoch": 5.62, + "learning_rate": 4.442259377977995e-05, + "loss": 1.4883, + "step": 64880 + }, + { + "epoch": 5.62, + "learning_rate": 4.442172745386815e-05, + "loss": 1.5634, + "step": 64890 + }, + { + "epoch": 5.62, + "learning_rate": 4.4420861127956334e-05, + "loss": 1.4819, + "step": 64900 + }, + { + "epoch": 5.62, + "learning_rate": 4.441999480204453e-05, + "loss": 1.5499, + "step": 64910 + }, + { + "epoch": 5.62, + "learning_rate": 4.441912847613272e-05, + "loss": 1.5891, + "step": 64920 + }, + { + "epoch": 5.62, + "learning_rate": 4.441826215022091e-05, + "loss": 1.5814, + "step": 64930 + }, + { + "epoch": 5.62, + "learning_rate": 4.4417395824309105e-05, + "loss": 1.5943, + "step": 64940 + }, + { + "epoch": 5.62, + "learning_rate": 4.44165294983973e-05, + "loss": 1.5446, + "step": 64950 + }, + { + "epoch": 5.62, + "learning_rate": 4.441566317248549e-05, + "loss": 1.5363, + "step": 64960 + }, + { + "epoch": 5.62, + "learning_rate": 4.441479684657368e-05, + "loss": 1.5301, + "step": 64970 + }, + { + "epoch": 5.62, + "learning_rate": 4.441393052066187e-05, + "loss": 1.5736, + "step": 64980 + }, + { + "epoch": 5.63, + "learning_rate": 4.4413064194750064e-05, + "loss": 1.5007, + "step": 64990 + }, + { + "epoch": 5.63, + "learning_rate": 4.441219786883826e-05, + "loss": 1.5278, + "step": 65000 + }, + { + "epoch": 5.63, + "learning_rate": 4.4411331542926446e-05, + "loss": 1.488, + "step": 65010 + }, + { + "epoch": 5.63, + "learning_rate": 4.441046521701464e-05, + "loss": 1.5534, + "step": 65020 + }, + { + "epoch": 5.63, + "learning_rate": 4.4409598891102835e-05, + "loss": 1.5426, + "step": 65030 + }, + { + "epoch": 5.63, + "learning_rate": 4.440873256519102e-05, + "loss": 1.5524, + "step": 65040 + }, + { + "epoch": 5.63, + "learning_rate": 4.440786623927922e-05, + "loss": 1.5439, + "step": 65050 + }, + { + "epoch": 5.63, + "learning_rate": 4.440699991336741e-05, + "loss": 1.4766, + "step": 65060 + }, + { + "epoch": 5.63, + "learning_rate": 4.44061335874556e-05, + "loss": 1.5101, + "step": 65070 + }, + { + "epoch": 5.63, + "learning_rate": 4.4405267261543794e-05, + "loss": 1.4258, + "step": 65080 + }, + { + "epoch": 5.63, + "learning_rate": 4.440440093563198e-05, + "loss": 1.5317, + "step": 65090 + }, + { + "epoch": 5.63, + "learning_rate": 4.4403534609720176e-05, + "loss": 1.5614, + "step": 65100 + }, + { + "epoch": 5.64, + "learning_rate": 4.440266828380837e-05, + "loss": 1.5654, + "step": 65110 + }, + { + "epoch": 5.64, + "learning_rate": 4.440180195789656e-05, + "loss": 1.4537, + "step": 65120 + }, + { + "epoch": 5.64, + "learning_rate": 4.440093563198475e-05, + "loss": 1.5465, + "step": 65130 + }, + { + "epoch": 5.64, + "learning_rate": 4.440006930607295e-05, + "loss": 1.5132, + "step": 65140 + }, + { + "epoch": 5.64, + "learning_rate": 4.4399202980161135e-05, + "loss": 1.5763, + "step": 65150 + }, + { + "epoch": 5.64, + "learning_rate": 4.439833665424933e-05, + "loss": 1.5142, + "step": 65160 + }, + { + "epoch": 5.64, + "learning_rate": 4.4397470328337524e-05, + "loss": 1.5516, + "step": 65170 + }, + { + "epoch": 5.64, + "learning_rate": 4.439660400242571e-05, + "loss": 1.5476, + "step": 65180 + }, + { + "epoch": 5.64, + "learning_rate": 4.4395737676513906e-05, + "loss": 1.5768, + "step": 65190 + }, + { + "epoch": 5.64, + "learning_rate": 4.4394871350602093e-05, + "loss": 1.5353, + "step": 65200 + }, + { + "epoch": 5.64, + "learning_rate": 4.439400502469029e-05, + "loss": 1.4618, + "step": 65210 + }, + { + "epoch": 5.65, + "learning_rate": 4.439313869877848e-05, + "loss": 1.485, + "step": 65220 + }, + { + "epoch": 5.65, + "learning_rate": 4.439227237286667e-05, + "loss": 1.5894, + "step": 65230 + }, + { + "epoch": 5.65, + "learning_rate": 4.4391406046954864e-05, + "loss": 1.5311, + "step": 65240 + }, + { + "epoch": 5.65, + "learning_rate": 4.439053972104306e-05, + "loss": 1.5409, + "step": 65250 + }, + { + "epoch": 5.65, + "learning_rate": 4.438967339513125e-05, + "loss": 1.5247, + "step": 65260 + }, + { + "epoch": 5.65, + "learning_rate": 4.438880706921944e-05, + "loss": 1.5313, + "step": 65270 + }, + { + "epoch": 5.65, + "learning_rate": 4.4387940743307636e-05, + "loss": 1.5399, + "step": 65280 + }, + { + "epoch": 5.65, + "learning_rate": 4.438707441739582e-05, + "loss": 1.5211, + "step": 65290 + }, + { + "epoch": 5.65, + "learning_rate": 4.438620809148402e-05, + "loss": 1.5151, + "step": 65300 + }, + { + "epoch": 5.65, + "learning_rate": 4.4385341765572205e-05, + "loss": 1.4591, + "step": 65310 + }, + { + "epoch": 5.65, + "learning_rate": 4.43844754396604e-05, + "loss": 1.5619, + "step": 65320 + }, + { + "epoch": 5.65, + "learning_rate": 4.4383609113748594e-05, + "loss": 1.5564, + "step": 65330 + }, + { + "epoch": 5.66, + "learning_rate": 4.438274278783678e-05, + "loss": 1.5391, + "step": 65340 + }, + { + "epoch": 5.66, + "learning_rate": 4.4381876461924976e-05, + "loss": 1.5265, + "step": 65350 + }, + { + "epoch": 5.66, + "learning_rate": 4.438101013601317e-05, + "loss": 1.5993, + "step": 65360 + }, + { + "epoch": 5.66, + "learning_rate": 4.438014381010136e-05, + "loss": 1.5417, + "step": 65370 + }, + { + "epoch": 5.66, + "learning_rate": 4.437927748418955e-05, + "loss": 1.513, + "step": 65380 + }, + { + "epoch": 5.66, + "learning_rate": 4.437841115827774e-05, + "loss": 1.5543, + "step": 65390 + }, + { + "epoch": 5.66, + "learning_rate": 4.4377544832365935e-05, + "loss": 1.5505, + "step": 65400 + }, + { + "epoch": 5.66, + "learning_rate": 4.437667850645413e-05, + "loss": 1.5176, + "step": 65410 + }, + { + "epoch": 5.66, + "learning_rate": 4.437581218054232e-05, + "loss": 1.5853, + "step": 65420 + }, + { + "epoch": 5.66, + "learning_rate": 4.437494585463051e-05, + "loss": 1.4989, + "step": 65430 + }, + { + "epoch": 5.66, + "learning_rate": 4.4374079528718706e-05, + "loss": 1.5281, + "step": 65440 + }, + { + "epoch": 5.66, + "learning_rate": 4.4373213202806894e-05, + "loss": 1.566, + "step": 65450 + }, + { + "epoch": 5.67, + "learning_rate": 4.437234687689509e-05, + "loss": 1.6109, + "step": 65460 + }, + { + "epoch": 5.67, + "learning_rate": 4.437148055098328e-05, + "loss": 1.5981, + "step": 65470 + }, + { + "epoch": 5.67, + "learning_rate": 4.437061422507147e-05, + "loss": 1.5026, + "step": 65480 + }, + { + "epoch": 5.67, + "learning_rate": 4.4369747899159665e-05, + "loss": 1.5425, + "step": 65490 + }, + { + "epoch": 5.67, + "learning_rate": 4.436888157324785e-05, + "loss": 1.5593, + "step": 65500 + }, + { + "epoch": 5.67, + "learning_rate": 4.436801524733605e-05, + "loss": 1.5526, + "step": 65510 + }, + { + "epoch": 5.67, + "learning_rate": 4.436714892142424e-05, + "loss": 1.6096, + "step": 65520 + }, + { + "epoch": 5.67, + "learning_rate": 4.436628259551243e-05, + "loss": 1.5836, + "step": 65530 + }, + { + "epoch": 5.67, + "learning_rate": 4.4365416269600624e-05, + "loss": 1.6035, + "step": 65540 + }, + { + "epoch": 5.67, + "learning_rate": 4.436454994368882e-05, + "loss": 1.4925, + "step": 65550 + }, + { + "epoch": 5.67, + "learning_rate": 4.4363683617777006e-05, + "loss": 1.5598, + "step": 65560 + }, + { + "epoch": 5.68, + "learning_rate": 4.43628172918652e-05, + "loss": 1.5179, + "step": 65570 + }, + { + "epoch": 5.68, + "learning_rate": 4.4361950965953395e-05, + "loss": 1.5308, + "step": 65580 + }, + { + "epoch": 5.68, + "learning_rate": 4.436108464004158e-05, + "loss": 1.6135, + "step": 65590 + }, + { + "epoch": 5.68, + "learning_rate": 4.436021831412978e-05, + "loss": 1.5346, + "step": 65600 + }, + { + "epoch": 5.68, + "learning_rate": 4.4359351988217964e-05, + "loss": 1.6072, + "step": 65610 + }, + { + "epoch": 5.68, + "learning_rate": 4.435848566230616e-05, + "loss": 1.5229, + "step": 65620 + }, + { + "epoch": 5.68, + "learning_rate": 4.435761933639435e-05, + "loss": 1.582, + "step": 65630 + }, + { + "epoch": 5.68, + "learning_rate": 4.435675301048254e-05, + "loss": 1.6303, + "step": 65640 + }, + { + "epoch": 5.68, + "learning_rate": 4.4355886684570735e-05, + "loss": 1.5973, + "step": 65650 + }, + { + "epoch": 5.68, + "learning_rate": 4.435502035865893e-05, + "loss": 1.5806, + "step": 65660 + }, + { + "epoch": 5.68, + "learning_rate": 4.435415403274712e-05, + "loss": 1.6283, + "step": 65670 + }, + { + "epoch": 5.68, + "learning_rate": 4.435328770683531e-05, + "loss": 1.5394, + "step": 65680 + }, + { + "epoch": 5.69, + "learning_rate": 4.4352421380923506e-05, + "loss": 1.5371, + "step": 65690 + }, + { + "epoch": 5.69, + "learning_rate": 4.4351555055011694e-05, + "loss": 1.5173, + "step": 65700 + }, + { + "epoch": 5.69, + "learning_rate": 4.435068872909989e-05, + "loss": 1.4663, + "step": 65710 + }, + { + "epoch": 5.69, + "learning_rate": 4.4349822403188076e-05, + "loss": 1.5325, + "step": 65720 + }, + { + "epoch": 5.69, + "learning_rate": 4.434895607727627e-05, + "loss": 1.4967, + "step": 65730 + }, + { + "epoch": 5.69, + "learning_rate": 4.4348089751364465e-05, + "loss": 1.5654, + "step": 65740 + }, + { + "epoch": 5.69, + "learning_rate": 4.434722342545265e-05, + "loss": 1.5444, + "step": 65750 + }, + { + "epoch": 5.69, + "learning_rate": 4.434635709954085e-05, + "loss": 1.5262, + "step": 65760 + }, + { + "epoch": 5.69, + "learning_rate": 4.434549077362904e-05, + "loss": 1.588, + "step": 65770 + }, + { + "epoch": 5.69, + "learning_rate": 4.434462444771723e-05, + "loss": 1.5407, + "step": 65780 + }, + { + "epoch": 5.69, + "learning_rate": 4.4343758121805424e-05, + "loss": 1.5767, + "step": 65790 + }, + { + "epoch": 5.7, + "learning_rate": 4.434289179589362e-05, + "loss": 1.5582, + "step": 65800 + }, + { + "epoch": 5.7, + "learning_rate": 4.4342025469981806e-05, + "loss": 1.4809, + "step": 65810 + }, + { + "epoch": 5.7, + "learning_rate": 4.434115914407e-05, + "loss": 1.4628, + "step": 65820 + }, + { + "epoch": 5.7, + "learning_rate": 4.434029281815819e-05, + "loss": 1.5007, + "step": 65830 + }, + { + "epoch": 5.7, + "learning_rate": 4.433942649224638e-05, + "loss": 1.5107, + "step": 65840 + }, + { + "epoch": 5.7, + "learning_rate": 4.433856016633458e-05, + "loss": 1.5654, + "step": 65850 + }, + { + "epoch": 5.7, + "learning_rate": 4.4337693840422765e-05, + "loss": 1.5401, + "step": 65860 + }, + { + "epoch": 5.7, + "learning_rate": 4.433682751451096e-05, + "loss": 1.5024, + "step": 65870 + }, + { + "epoch": 5.7, + "learning_rate": 4.4335961188599154e-05, + "loss": 1.5503, + "step": 65880 + }, + { + "epoch": 5.7, + "learning_rate": 4.433509486268734e-05, + "loss": 1.5031, + "step": 65890 + }, + { + "epoch": 5.7, + "learning_rate": 4.4334228536775536e-05, + "loss": 1.4848, + "step": 65900 + }, + { + "epoch": 5.7, + "learning_rate": 4.433336221086373e-05, + "loss": 1.6006, + "step": 65910 + }, + { + "epoch": 5.71, + "learning_rate": 4.433249588495192e-05, + "loss": 1.6113, + "step": 65920 + }, + { + "epoch": 5.71, + "learning_rate": 4.433162955904011e-05, + "loss": 1.5451, + "step": 65930 + }, + { + "epoch": 5.71, + "learning_rate": 4.43307632331283e-05, + "loss": 1.5332, + "step": 65940 + }, + { + "epoch": 5.71, + "learning_rate": 4.4329896907216494e-05, + "loss": 1.5649, + "step": 65950 + }, + { + "epoch": 5.71, + "learning_rate": 4.432903058130469e-05, + "loss": 1.5279, + "step": 65960 + }, + { + "epoch": 5.71, + "learning_rate": 4.4328164255392877e-05, + "loss": 1.4569, + "step": 65970 + }, + { + "epoch": 5.71, + "learning_rate": 4.432729792948107e-05, + "loss": 1.5681, + "step": 65980 + }, + { + "epoch": 5.71, + "learning_rate": 4.4326431603569265e-05, + "loss": 1.5259, + "step": 65990 + }, + { + "epoch": 5.71, + "learning_rate": 4.432556527765745e-05, + "loss": 1.5858, + "step": 66000 + }, + { + "epoch": 5.71, + "learning_rate": 4.432469895174565e-05, + "loss": 1.5225, + "step": 66010 + }, + { + "epoch": 5.71, + "learning_rate": 4.432383262583384e-05, + "loss": 1.5579, + "step": 66020 + }, + { + "epoch": 5.72, + "learning_rate": 4.432296629992203e-05, + "loss": 1.5495, + "step": 66030 + }, + { + "epoch": 5.72, + "learning_rate": 4.4322099974010224e-05, + "loss": 1.5581, + "step": 66040 + }, + { + "epoch": 5.72, + "learning_rate": 4.432123364809841e-05, + "loss": 1.4654, + "step": 66050 + }, + { + "epoch": 5.72, + "learning_rate": 4.4320367322186606e-05, + "loss": 1.6195, + "step": 66060 + }, + { + "epoch": 5.72, + "learning_rate": 4.43195009962748e-05, + "loss": 1.5624, + "step": 66070 + }, + { + "epoch": 5.72, + "learning_rate": 4.431863467036299e-05, + "loss": 1.5028, + "step": 66080 + }, + { + "epoch": 5.72, + "learning_rate": 4.431776834445118e-05, + "loss": 1.5071, + "step": 66090 + }, + { + "epoch": 5.72, + "learning_rate": 4.431690201853938e-05, + "loss": 1.5466, + "step": 66100 + }, + { + "epoch": 5.72, + "learning_rate": 4.4316035692627565e-05, + "loss": 1.5185, + "step": 66110 + }, + { + "epoch": 5.72, + "learning_rate": 4.431516936671576e-05, + "loss": 1.5282, + "step": 66120 + }, + { + "epoch": 5.72, + "learning_rate": 4.4314303040803954e-05, + "loss": 1.5141, + "step": 66130 + }, + { + "epoch": 5.72, + "learning_rate": 4.431343671489214e-05, + "loss": 1.563, + "step": 66140 + }, + { + "epoch": 5.73, + "learning_rate": 4.4312570388980336e-05, + "loss": 1.5656, + "step": 66150 + }, + { + "epoch": 5.73, + "learning_rate": 4.4311704063068524e-05, + "loss": 1.5252, + "step": 66160 + }, + { + "epoch": 5.73, + "learning_rate": 4.431083773715672e-05, + "loss": 1.558, + "step": 66170 + }, + { + "epoch": 5.73, + "learning_rate": 4.430997141124491e-05, + "loss": 1.5432, + "step": 66180 + }, + { + "epoch": 5.73, + "learning_rate": 4.43091050853331e-05, + "loss": 1.5502, + "step": 66190 + }, + { + "epoch": 5.73, + "learning_rate": 4.4308238759421295e-05, + "loss": 1.6504, + "step": 66200 + }, + { + "epoch": 5.73, + "learning_rate": 4.430737243350949e-05, + "loss": 1.5178, + "step": 66210 + }, + { + "epoch": 5.73, + "learning_rate": 4.430650610759768e-05, + "loss": 1.5048, + "step": 66220 + }, + { + "epoch": 5.73, + "learning_rate": 4.430563978168587e-05, + "loss": 1.5529, + "step": 66230 + }, + { + "epoch": 5.73, + "learning_rate": 4.430477345577406e-05, + "loss": 1.55, + "step": 66240 + }, + { + "epoch": 5.73, + "learning_rate": 4.4303907129862254e-05, + "loss": 1.5668, + "step": 66250 + }, + { + "epoch": 5.74, + "learning_rate": 4.430304080395045e-05, + "loss": 1.627, + "step": 66260 + }, + { + "epoch": 5.74, + "learning_rate": 4.4302174478038636e-05, + "loss": 1.604, + "step": 66270 + }, + { + "epoch": 5.74, + "learning_rate": 4.430130815212683e-05, + "loss": 1.545, + "step": 66280 + }, + { + "epoch": 5.74, + "learning_rate": 4.4300441826215025e-05, + "loss": 1.5141, + "step": 66290 + }, + { + "epoch": 5.74, + "learning_rate": 4.429957550030321e-05, + "loss": 1.5602, + "step": 66300 + }, + { + "epoch": 5.74, + "learning_rate": 4.429870917439141e-05, + "loss": 1.5731, + "step": 66310 + }, + { + "epoch": 5.74, + "learning_rate": 4.42978428484796e-05, + "loss": 1.5004, + "step": 66320 + }, + { + "epoch": 5.74, + "learning_rate": 4.429697652256779e-05, + "loss": 1.4646, + "step": 66330 + }, + { + "epoch": 5.74, + "learning_rate": 4.429611019665598e-05, + "loss": 1.5461, + "step": 66340 + }, + { + "epoch": 5.74, + "learning_rate": 4.429524387074417e-05, + "loss": 1.5439, + "step": 66350 + }, + { + "epoch": 5.74, + "learning_rate": 4.4294377544832365e-05, + "loss": 1.5379, + "step": 66360 + }, + { + "epoch": 5.74, + "learning_rate": 4.429351121892056e-05, + "loss": 1.5711, + "step": 66370 + }, + { + "epoch": 5.75, + "learning_rate": 4.429264489300875e-05, + "loss": 1.5187, + "step": 66380 + }, + { + "epoch": 5.75, + "learning_rate": 4.429177856709694e-05, + "loss": 1.5954, + "step": 66390 + }, + { + "epoch": 5.75, + "learning_rate": 4.4290912241185136e-05, + "loss": 1.5541, + "step": 66400 + }, + { + "epoch": 5.75, + "learning_rate": 4.4290045915273324e-05, + "loss": 1.5398, + "step": 66410 + }, + { + "epoch": 5.75, + "learning_rate": 4.428917958936152e-05, + "loss": 1.5938, + "step": 66420 + }, + { + "epoch": 5.75, + "learning_rate": 4.428831326344971e-05, + "loss": 1.4931, + "step": 66430 + }, + { + "epoch": 5.75, + "learning_rate": 4.42874469375379e-05, + "loss": 1.5264, + "step": 66440 + }, + { + "epoch": 5.75, + "learning_rate": 4.4286580611626095e-05, + "loss": 1.4543, + "step": 66450 + }, + { + "epoch": 5.75, + "learning_rate": 4.428571428571428e-05, + "loss": 1.5845, + "step": 66460 + }, + { + "epoch": 5.75, + "learning_rate": 4.428484795980248e-05, + "loss": 1.5633, + "step": 66470 + }, + { + "epoch": 5.75, + "learning_rate": 4.428398163389067e-05, + "loss": 1.5204, + "step": 66480 + }, + { + "epoch": 5.75, + "learning_rate": 4.428311530797886e-05, + "loss": 1.5595, + "step": 66490 + }, + { + "epoch": 5.76, + "learning_rate": 4.4282248982067054e-05, + "loss": 1.4364, + "step": 66500 + }, + { + "epoch": 5.76, + "learning_rate": 4.428138265615525e-05, + "loss": 1.5282, + "step": 66510 + }, + { + "epoch": 5.76, + "learning_rate": 4.4280516330243436e-05, + "loss": 1.5022, + "step": 66520 + }, + { + "epoch": 5.76, + "learning_rate": 4.427965000433163e-05, + "loss": 1.5771, + "step": 66530 + }, + { + "epoch": 5.76, + "learning_rate": 4.4278783678419825e-05, + "loss": 1.5013, + "step": 66540 + }, + { + "epoch": 5.76, + "learning_rate": 4.427791735250801e-05, + "loss": 1.5756, + "step": 66550 + }, + { + "epoch": 5.76, + "learning_rate": 4.427705102659621e-05, + "loss": 1.6159, + "step": 66560 + }, + { + "epoch": 5.76, + "learning_rate": 4.4276184700684395e-05, + "loss": 1.5485, + "step": 66570 + }, + { + "epoch": 5.76, + "learning_rate": 4.427531837477259e-05, + "loss": 1.5607, + "step": 66580 + }, + { + "epoch": 5.76, + "learning_rate": 4.4274452048860784e-05, + "loss": 1.4982, + "step": 66590 + }, + { + "epoch": 5.76, + "learning_rate": 4.427358572294897e-05, + "loss": 1.5929, + "step": 66600 + }, + { + "epoch": 5.77, + "learning_rate": 4.4272719397037166e-05, + "loss": 1.5805, + "step": 66610 + }, + { + "epoch": 5.77, + "learning_rate": 4.427185307112536e-05, + "loss": 1.5139, + "step": 66620 + }, + { + "epoch": 5.77, + "learning_rate": 4.427098674521355e-05, + "loss": 1.4825, + "step": 66630 + }, + { + "epoch": 5.77, + "learning_rate": 4.427012041930174e-05, + "loss": 1.5437, + "step": 66640 + }, + { + "epoch": 5.77, + "learning_rate": 4.426925409338994e-05, + "loss": 1.5161, + "step": 66650 + }, + { + "epoch": 5.77, + "learning_rate": 4.4268387767478124e-05, + "loss": 1.4862, + "step": 66660 + }, + { + "epoch": 5.77, + "learning_rate": 4.426752144156632e-05, + "loss": 1.535, + "step": 66670 + }, + { + "epoch": 5.77, + "learning_rate": 4.4266655115654507e-05, + "loss": 1.5754, + "step": 66680 + }, + { + "epoch": 5.77, + "learning_rate": 4.42657887897427e-05, + "loss": 1.5661, + "step": 66690 + }, + { + "epoch": 5.77, + "learning_rate": 4.4264922463830895e-05, + "loss": 1.5302, + "step": 66700 + }, + { + "epoch": 5.77, + "learning_rate": 4.426405613791908e-05, + "loss": 1.5891, + "step": 66710 + }, + { + "epoch": 5.77, + "learning_rate": 4.426318981200728e-05, + "loss": 1.4852, + "step": 66720 + }, + { + "epoch": 5.78, + "learning_rate": 4.426232348609547e-05, + "loss": 1.5108, + "step": 66730 + }, + { + "epoch": 5.78, + "learning_rate": 4.426145716018366e-05, + "loss": 1.5773, + "step": 66740 + }, + { + "epoch": 5.78, + "learning_rate": 4.4260590834271854e-05, + "loss": 1.5616, + "step": 66750 + }, + { + "epoch": 5.78, + "learning_rate": 4.425972450836005e-05, + "loss": 1.5288, + "step": 66760 + }, + { + "epoch": 5.78, + "learning_rate": 4.4258858182448236e-05, + "loss": 1.5362, + "step": 66770 + }, + { + "epoch": 5.78, + "learning_rate": 4.425799185653643e-05, + "loss": 1.5574, + "step": 66780 + }, + { + "epoch": 5.78, + "learning_rate": 4.425712553062462e-05, + "loss": 1.5928, + "step": 66790 + }, + { + "epoch": 5.78, + "learning_rate": 4.425625920471281e-05, + "loss": 1.6307, + "step": 66800 + }, + { + "epoch": 5.78, + "learning_rate": 4.425539287880101e-05, + "loss": 1.577, + "step": 66810 + }, + { + "epoch": 5.78, + "learning_rate": 4.4254526552889195e-05, + "loss": 1.5287, + "step": 66820 + }, + { + "epoch": 5.78, + "learning_rate": 4.425366022697739e-05, + "loss": 1.533, + "step": 66830 + }, + { + "epoch": 5.79, + "learning_rate": 4.4252793901065584e-05, + "loss": 1.4959, + "step": 66840 + }, + { + "epoch": 5.79, + "learning_rate": 4.425192757515377e-05, + "loss": 1.4885, + "step": 66850 + }, + { + "epoch": 5.79, + "learning_rate": 4.4251061249241966e-05, + "loss": 1.5196, + "step": 66860 + }, + { + "epoch": 5.79, + "learning_rate": 4.425019492333016e-05, + "loss": 1.5181, + "step": 66870 + }, + { + "epoch": 5.79, + "learning_rate": 4.424932859741835e-05, + "loss": 1.5584, + "step": 66880 + }, + { + "epoch": 5.79, + "learning_rate": 4.424846227150654e-05, + "loss": 1.4421, + "step": 66890 + }, + { + "epoch": 5.79, + "learning_rate": 4.424759594559473e-05, + "loss": 1.5332, + "step": 66900 + }, + { + "epoch": 5.79, + "learning_rate": 4.4246729619682925e-05, + "loss": 1.5228, + "step": 66910 + }, + { + "epoch": 5.79, + "learning_rate": 4.424586329377112e-05, + "loss": 1.6008, + "step": 66920 + }, + { + "epoch": 5.79, + "learning_rate": 4.424499696785931e-05, + "loss": 1.5783, + "step": 66930 + }, + { + "epoch": 5.79, + "learning_rate": 4.42441306419475e-05, + "loss": 1.5576, + "step": 66940 + }, + { + "epoch": 5.79, + "learning_rate": 4.4243264316035696e-05, + "loss": 1.5357, + "step": 66950 + }, + { + "epoch": 5.8, + "learning_rate": 4.4242397990123883e-05, + "loss": 1.5988, + "step": 66960 + }, + { + "epoch": 5.8, + "learning_rate": 4.424153166421208e-05, + "loss": 1.5721, + "step": 66970 + }, + { + "epoch": 5.8, + "learning_rate": 4.4240665338300266e-05, + "loss": 1.5604, + "step": 66980 + }, + { + "epoch": 5.8, + "learning_rate": 4.423979901238846e-05, + "loss": 1.544, + "step": 66990 + }, + { + "epoch": 5.8, + "learning_rate": 4.4238932686476655e-05, + "loss": 1.5693, + "step": 67000 + }, + { + "epoch": 5.8, + "learning_rate": 4.423806636056484e-05, + "loss": 1.4771, + "step": 67010 + }, + { + "epoch": 5.8, + "learning_rate": 4.423720003465304e-05, + "loss": 1.4921, + "step": 67020 + }, + { + "epoch": 5.8, + "learning_rate": 4.423633370874123e-05, + "loss": 1.5525, + "step": 67030 + }, + { + "epoch": 5.8, + "learning_rate": 4.423546738282942e-05, + "loss": 1.5445, + "step": 67040 + }, + { + "epoch": 5.8, + "learning_rate": 4.423460105691761e-05, + "loss": 1.5607, + "step": 67050 + }, + { + "epoch": 5.8, + "learning_rate": 4.423373473100581e-05, + "loss": 1.5074, + "step": 67060 + }, + { + "epoch": 5.81, + "learning_rate": 4.4232868405093995e-05, + "loss": 1.5669, + "step": 67070 + }, + { + "epoch": 5.81, + "learning_rate": 4.423200207918219e-05, + "loss": 1.5658, + "step": 67080 + }, + { + "epoch": 5.81, + "learning_rate": 4.423113575327038e-05, + "loss": 1.5364, + "step": 67090 + }, + { + "epoch": 5.81, + "learning_rate": 4.423026942735857e-05, + "loss": 1.5377, + "step": 67100 + }, + { + "epoch": 5.81, + "learning_rate": 4.4229403101446766e-05, + "loss": 1.5931, + "step": 67110 + }, + { + "epoch": 5.81, + "learning_rate": 4.4228536775534954e-05, + "loss": 1.5147, + "step": 67120 + }, + { + "epoch": 5.81, + "learning_rate": 4.422767044962315e-05, + "loss": 1.5456, + "step": 67130 + }, + { + "epoch": 5.81, + "learning_rate": 4.422680412371134e-05, + "loss": 1.483, + "step": 67140 + }, + { + "epoch": 5.81, + "learning_rate": 4.422593779779953e-05, + "loss": 1.5396, + "step": 67150 + }, + { + "epoch": 5.81, + "learning_rate": 4.4225071471887725e-05, + "loss": 1.5261, + "step": 67160 + }, + { + "epoch": 5.81, + "learning_rate": 4.422420514597592e-05, + "loss": 1.5456, + "step": 67170 + }, + { + "epoch": 5.81, + "learning_rate": 4.422333882006411e-05, + "loss": 1.4933, + "step": 67180 + }, + { + "epoch": 5.82, + "learning_rate": 4.42224724941523e-05, + "loss": 1.5005, + "step": 67190 + }, + { + "epoch": 5.82, + "learning_rate": 4.422160616824049e-05, + "loss": 1.517, + "step": 67200 + }, + { + "epoch": 5.82, + "learning_rate": 4.4220739842328684e-05, + "loss": 1.5293, + "step": 67210 + }, + { + "epoch": 5.82, + "learning_rate": 4.421987351641688e-05, + "loss": 1.5988, + "step": 67220 + }, + { + "epoch": 5.82, + "learning_rate": 4.4219007190505066e-05, + "loss": 1.5876, + "step": 67230 + }, + { + "epoch": 5.82, + "learning_rate": 4.421814086459326e-05, + "loss": 1.5178, + "step": 67240 + }, + { + "epoch": 5.82, + "learning_rate": 4.4217274538681455e-05, + "loss": 1.4867, + "step": 67250 + }, + { + "epoch": 5.82, + "learning_rate": 4.421640821276964e-05, + "loss": 1.6012, + "step": 67260 + }, + { + "epoch": 5.82, + "learning_rate": 4.421554188685784e-05, + "loss": 1.5629, + "step": 67270 + }, + { + "epoch": 5.82, + "learning_rate": 4.421467556094603e-05, + "loss": 1.5485, + "step": 67280 + }, + { + "epoch": 5.82, + "learning_rate": 4.421380923503422e-05, + "loss": 1.5204, + "step": 67290 + }, + { + "epoch": 5.83, + "learning_rate": 4.4212942909122414e-05, + "loss": 1.4831, + "step": 67300 + }, + { + "epoch": 5.83, + "learning_rate": 4.42120765832106e-05, + "loss": 1.5646, + "step": 67310 + }, + { + "epoch": 5.83, + "learning_rate": 4.4211210257298796e-05, + "loss": 1.4927, + "step": 67320 + }, + { + "epoch": 5.83, + "learning_rate": 4.421034393138699e-05, + "loss": 1.5861, + "step": 67330 + }, + { + "epoch": 5.83, + "learning_rate": 4.420947760547518e-05, + "loss": 1.5643, + "step": 67340 + }, + { + "epoch": 5.83, + "learning_rate": 4.420861127956337e-05, + "loss": 1.5369, + "step": 67350 + }, + { + "epoch": 5.83, + "learning_rate": 4.420774495365157e-05, + "loss": 1.4921, + "step": 67360 + }, + { + "epoch": 5.83, + "learning_rate": 4.4206878627739754e-05, + "loss": 1.4943, + "step": 67370 + }, + { + "epoch": 5.83, + "learning_rate": 4.420601230182795e-05, + "loss": 1.5669, + "step": 67380 + }, + { + "epoch": 5.83, + "learning_rate": 4.420514597591614e-05, + "loss": 1.5436, + "step": 67390 + }, + { + "epoch": 5.83, + "learning_rate": 4.420427965000433e-05, + "loss": 1.5249, + "step": 67400 + }, + { + "epoch": 5.83, + "learning_rate": 4.4203413324092525e-05, + "loss": 1.5564, + "step": 67410 + }, + { + "epoch": 5.84, + "learning_rate": 4.420254699818071e-05, + "loss": 1.5228, + "step": 67420 + }, + { + "epoch": 5.84, + "learning_rate": 4.420168067226891e-05, + "loss": 1.5906, + "step": 67430 + }, + { + "epoch": 5.84, + "learning_rate": 4.42008143463571e-05, + "loss": 1.5725, + "step": 67440 + }, + { + "epoch": 5.84, + "learning_rate": 4.419994802044529e-05, + "loss": 1.5529, + "step": 67450 + }, + { + "epoch": 5.84, + "learning_rate": 4.4199081694533484e-05, + "loss": 1.5634, + "step": 67460 + }, + { + "epoch": 5.84, + "learning_rate": 4.419821536862168e-05, + "loss": 1.5863, + "step": 67470 + }, + { + "epoch": 5.84, + "learning_rate": 4.4197349042709866e-05, + "loss": 1.4672, + "step": 67480 + }, + { + "epoch": 5.84, + "learning_rate": 4.419648271679806e-05, + "loss": 1.4776, + "step": 67490 + }, + { + "epoch": 5.84, + "learning_rate": 4.4195616390886255e-05, + "loss": 1.5764, + "step": 67500 + }, + { + "epoch": 5.84, + "learning_rate": 4.419475006497444e-05, + "loss": 1.5262, + "step": 67510 + }, + { + "epoch": 5.84, + "learning_rate": 4.419388373906264e-05, + "loss": 1.5057, + "step": 67520 + }, + { + "epoch": 5.84, + "learning_rate": 4.4193017413150825e-05, + "loss": 1.5059, + "step": 67530 + }, + { + "epoch": 5.85, + "learning_rate": 4.419215108723902e-05, + "loss": 1.5401, + "step": 67540 + }, + { + "epoch": 5.85, + "learning_rate": 4.4191284761327214e-05, + "loss": 1.5407, + "step": 67550 + }, + { + "epoch": 5.85, + "learning_rate": 4.41904184354154e-05, + "loss": 1.4941, + "step": 67560 + }, + { + "epoch": 5.85, + "learning_rate": 4.4189552109503596e-05, + "loss": 1.5304, + "step": 67570 + }, + { + "epoch": 5.85, + "learning_rate": 4.418868578359179e-05, + "loss": 1.5442, + "step": 67580 + }, + { + "epoch": 5.85, + "learning_rate": 4.418781945767998e-05, + "loss": 1.5309, + "step": 67590 + }, + { + "epoch": 5.85, + "learning_rate": 4.418695313176817e-05, + "loss": 1.6047, + "step": 67600 + }, + { + "epoch": 5.85, + "learning_rate": 4.418608680585637e-05, + "loss": 1.5324, + "step": 67610 + }, + { + "epoch": 5.85, + "learning_rate": 4.4185220479944555e-05, + "loss": 1.5355, + "step": 67620 + }, + { + "epoch": 5.85, + "learning_rate": 4.418435415403275e-05, + "loss": 1.522, + "step": 67630 + }, + { + "epoch": 5.85, + "learning_rate": 4.418348782812094e-05, + "loss": 1.5981, + "step": 67640 + }, + { + "epoch": 5.86, + "learning_rate": 4.418262150220913e-05, + "loss": 1.5915, + "step": 67650 + }, + { + "epoch": 5.86, + "learning_rate": 4.4181755176297326e-05, + "loss": 1.6214, + "step": 67660 + }, + { + "epoch": 5.86, + "learning_rate": 4.4180888850385513e-05, + "loss": 1.5936, + "step": 67670 + }, + { + "epoch": 5.86, + "learning_rate": 4.418002252447371e-05, + "loss": 1.5507, + "step": 67680 + }, + { + "epoch": 5.86, + "learning_rate": 4.41791561985619e-05, + "loss": 1.568, + "step": 67690 + }, + { + "epoch": 5.86, + "learning_rate": 4.417828987265009e-05, + "loss": 1.5503, + "step": 67700 + }, + { + "epoch": 5.86, + "learning_rate": 4.4177423546738284e-05, + "loss": 1.5128, + "step": 67710 + }, + { + "epoch": 5.86, + "learning_rate": 4.417655722082647e-05, + "loss": 1.5376, + "step": 67720 + }, + { + "epoch": 5.86, + "learning_rate": 4.4175690894914667e-05, + "loss": 1.5762, + "step": 67730 + }, + { + "epoch": 5.86, + "learning_rate": 4.417482456900286e-05, + "loss": 1.4717, + "step": 67740 + }, + { + "epoch": 5.86, + "learning_rate": 4.417395824309105e-05, + "loss": 1.5592, + "step": 67750 + }, + { + "epoch": 5.86, + "learning_rate": 4.417309191717924e-05, + "loss": 1.5825, + "step": 67760 + }, + { + "epoch": 5.87, + "learning_rate": 4.417222559126744e-05, + "loss": 1.5244, + "step": 67770 + }, + { + "epoch": 5.87, + "learning_rate": 4.4171359265355625e-05, + "loss": 1.5354, + "step": 67780 + }, + { + "epoch": 5.87, + "learning_rate": 4.417049293944382e-05, + "loss": 1.5481, + "step": 67790 + }, + { + "epoch": 5.87, + "learning_rate": 4.4169626613532014e-05, + "loss": 1.5604, + "step": 67800 + }, + { + "epoch": 5.87, + "learning_rate": 4.41687602876202e-05, + "loss": 1.5308, + "step": 67810 + }, + { + "epoch": 5.87, + "learning_rate": 4.4167893961708396e-05, + "loss": 1.5568, + "step": 67820 + }, + { + "epoch": 5.87, + "learning_rate": 4.4167027635796584e-05, + "loss": 1.5382, + "step": 67830 + }, + { + "epoch": 5.87, + "learning_rate": 4.416616130988478e-05, + "loss": 1.5115, + "step": 67840 + }, + { + "epoch": 5.87, + "learning_rate": 4.416529498397297e-05, + "loss": 1.5287, + "step": 67850 + }, + { + "epoch": 5.87, + "learning_rate": 4.416442865806116e-05, + "loss": 1.4553, + "step": 67860 + }, + { + "epoch": 5.87, + "learning_rate": 4.4163562332149355e-05, + "loss": 1.5387, + "step": 67870 + }, + { + "epoch": 5.88, + "learning_rate": 4.416269600623755e-05, + "loss": 1.5301, + "step": 67880 + }, + { + "epoch": 5.88, + "learning_rate": 4.416182968032574e-05, + "loss": 1.5661, + "step": 67890 + }, + { + "epoch": 5.88, + "learning_rate": 4.416096335441393e-05, + "loss": 1.5635, + "step": 67900 + }, + { + "epoch": 5.88, + "learning_rate": 4.4160097028502126e-05, + "loss": 1.5099, + "step": 67910 + }, + { + "epoch": 5.88, + "learning_rate": 4.4159230702590314e-05, + "loss": 1.4902, + "step": 67920 + }, + { + "epoch": 5.88, + "learning_rate": 4.415836437667851e-05, + "loss": 1.6326, + "step": 67930 + }, + { + "epoch": 5.88, + "learning_rate": 4.4157498050766696e-05, + "loss": 1.5993, + "step": 67940 + }, + { + "epoch": 5.88, + "learning_rate": 4.415663172485489e-05, + "loss": 1.5567, + "step": 67950 + }, + { + "epoch": 5.88, + "learning_rate": 4.4155765398943085e-05, + "loss": 1.5459, + "step": 67960 + }, + { + "epoch": 5.88, + "learning_rate": 4.415489907303127e-05, + "loss": 1.47, + "step": 67970 + }, + { + "epoch": 5.88, + "learning_rate": 4.415403274711947e-05, + "loss": 1.513, + "step": 67980 + }, + { + "epoch": 5.88, + "learning_rate": 4.415316642120766e-05, + "loss": 1.6176, + "step": 67990 + }, + { + "epoch": 5.89, + "learning_rate": 4.415230009529585e-05, + "loss": 1.6354, + "step": 68000 + }, + { + "epoch": 5.89, + "learning_rate": 4.4151433769384044e-05, + "loss": 1.5573, + "step": 68010 + }, + { + "epoch": 5.89, + "learning_rate": 4.415056744347224e-05, + "loss": 1.5522, + "step": 68020 + }, + { + "epoch": 5.89, + "learning_rate": 4.4149701117560426e-05, + "loss": 1.5741, + "step": 68030 + }, + { + "epoch": 5.89, + "learning_rate": 4.414883479164862e-05, + "loss": 1.5193, + "step": 68040 + }, + { + "epoch": 5.89, + "learning_rate": 4.414796846573681e-05, + "loss": 1.5178, + "step": 68050 + }, + { + "epoch": 5.89, + "learning_rate": 4.4147102139825e-05, + "loss": 1.5611, + "step": 68060 + }, + { + "epoch": 5.89, + "learning_rate": 4.41462358139132e-05, + "loss": 1.5606, + "step": 68070 + }, + { + "epoch": 5.89, + "learning_rate": 4.4145369488001384e-05, + "loss": 1.6116, + "step": 68080 + }, + { + "epoch": 5.89, + "learning_rate": 4.414450316208958e-05, + "loss": 1.5638, + "step": 68090 + }, + { + "epoch": 5.89, + "learning_rate": 4.414363683617777e-05, + "loss": 1.5537, + "step": 68100 + }, + { + "epoch": 5.9, + "learning_rate": 4.414277051026596e-05, + "loss": 1.4841, + "step": 68110 + }, + { + "epoch": 5.9, + "learning_rate": 4.4141904184354155e-05, + "loss": 1.6242, + "step": 68120 + }, + { + "epoch": 5.9, + "learning_rate": 4.414103785844235e-05, + "loss": 1.5336, + "step": 68130 + }, + { + "epoch": 5.9, + "learning_rate": 4.414017153253054e-05, + "loss": 1.581, + "step": 68140 + }, + { + "epoch": 5.9, + "learning_rate": 4.413930520661873e-05, + "loss": 1.4979, + "step": 68150 + }, + { + "epoch": 5.9, + "learning_rate": 4.413843888070692e-05, + "loss": 1.5383, + "step": 68160 + }, + { + "epoch": 5.9, + "learning_rate": 4.4137572554795114e-05, + "loss": 1.5712, + "step": 68170 + }, + { + "epoch": 5.9, + "learning_rate": 4.413670622888331e-05, + "loss": 1.5664, + "step": 68180 + }, + { + "epoch": 5.9, + "learning_rate": 4.4135839902971496e-05, + "loss": 1.5563, + "step": 68190 + }, + { + "epoch": 5.9, + "learning_rate": 4.413497357705969e-05, + "loss": 1.5121, + "step": 68200 + }, + { + "epoch": 5.9, + "learning_rate": 4.4134107251147885e-05, + "loss": 1.5525, + "step": 68210 + }, + { + "epoch": 5.9, + "learning_rate": 4.413324092523607e-05, + "loss": 1.5156, + "step": 68220 + }, + { + "epoch": 5.91, + "learning_rate": 4.413237459932427e-05, + "loss": 1.5124, + "step": 68230 + }, + { + "epoch": 5.91, + "learning_rate": 4.413150827341246e-05, + "loss": 1.5526, + "step": 68240 + }, + { + "epoch": 5.91, + "learning_rate": 4.413064194750065e-05, + "loss": 1.5066, + "step": 68250 + }, + { + "epoch": 5.91, + "learning_rate": 4.4129775621588844e-05, + "loss": 1.5237, + "step": 68260 + }, + { + "epoch": 5.91, + "learning_rate": 4.412890929567703e-05, + "loss": 1.5609, + "step": 68270 + }, + { + "epoch": 5.91, + "learning_rate": 4.4128042969765226e-05, + "loss": 1.5298, + "step": 68280 + }, + { + "epoch": 5.91, + "learning_rate": 4.412717664385342e-05, + "loss": 1.5394, + "step": 68290 + }, + { + "epoch": 5.91, + "learning_rate": 4.412631031794161e-05, + "loss": 1.5558, + "step": 68300 + }, + { + "epoch": 5.91, + "learning_rate": 4.41254439920298e-05, + "loss": 1.4944, + "step": 68310 + }, + { + "epoch": 5.91, + "learning_rate": 4.4124577666118e-05, + "loss": 1.552, + "step": 68320 + }, + { + "epoch": 5.91, + "learning_rate": 4.4123711340206185e-05, + "loss": 1.5696, + "step": 68330 + }, + { + "epoch": 5.92, + "learning_rate": 4.412284501429438e-05, + "loss": 1.4963, + "step": 68340 + }, + { + "epoch": 5.92, + "learning_rate": 4.4121978688382574e-05, + "loss": 1.5121, + "step": 68350 + }, + { + "epoch": 5.92, + "learning_rate": 4.412111236247076e-05, + "loss": 1.4842, + "step": 68360 + }, + { + "epoch": 5.92, + "learning_rate": 4.4120246036558956e-05, + "loss": 1.5093, + "step": 68370 + }, + { + "epoch": 5.92, + "learning_rate": 4.4119379710647143e-05, + "loss": 1.5167, + "step": 68380 + }, + { + "epoch": 5.92, + "learning_rate": 4.411851338473534e-05, + "loss": 1.5914, + "step": 68390 + }, + { + "epoch": 5.92, + "learning_rate": 4.411764705882353e-05, + "loss": 1.549, + "step": 68400 + }, + { + "epoch": 5.92, + "learning_rate": 4.411678073291172e-05, + "loss": 1.5256, + "step": 68410 + }, + { + "epoch": 5.92, + "learning_rate": 4.4115914406999914e-05, + "loss": 1.577, + "step": 68420 + }, + { + "epoch": 5.92, + "learning_rate": 4.411504808108811e-05, + "loss": 1.5871, + "step": 68430 + }, + { + "epoch": 5.92, + "learning_rate": 4.4114181755176297e-05, + "loss": 1.5709, + "step": 68440 + }, + { + "epoch": 5.92, + "learning_rate": 4.411331542926449e-05, + "loss": 1.4881, + "step": 68450 + }, + { + "epoch": 5.93, + "learning_rate": 4.4112449103352685e-05, + "loss": 1.5172, + "step": 68460 + }, + { + "epoch": 5.93, + "learning_rate": 4.411158277744087e-05, + "loss": 1.519, + "step": 68470 + }, + { + "epoch": 5.93, + "learning_rate": 4.411071645152907e-05, + "loss": 1.5889, + "step": 68480 + }, + { + "epoch": 5.93, + "learning_rate": 4.4109850125617255e-05, + "loss": 1.5352, + "step": 68490 + }, + { + "epoch": 5.93, + "learning_rate": 4.410898379970545e-05, + "loss": 1.5144, + "step": 68500 + }, + { + "epoch": 5.93, + "learning_rate": 4.4108117473793644e-05, + "loss": 1.4626, + "step": 68510 + }, + { + "epoch": 5.93, + "learning_rate": 4.410725114788183e-05, + "loss": 1.5441, + "step": 68520 + }, + { + "epoch": 5.93, + "learning_rate": 4.4106384821970026e-05, + "loss": 1.4455, + "step": 68530 + }, + { + "epoch": 5.93, + "learning_rate": 4.410551849605822e-05, + "loss": 1.5156, + "step": 68540 + }, + { + "epoch": 5.93, + "learning_rate": 4.410465217014641e-05, + "loss": 1.5821, + "step": 68550 + }, + { + "epoch": 5.93, + "learning_rate": 4.41037858442346e-05, + "loss": 1.4639, + "step": 68560 + }, + { + "epoch": 5.93, + "learning_rate": 4.410291951832279e-05, + "loss": 1.5526, + "step": 68570 + }, + { + "epoch": 5.94, + "learning_rate": 4.4102053192410985e-05, + "loss": 1.4958, + "step": 68580 + }, + { + "epoch": 5.94, + "learning_rate": 4.410118686649918e-05, + "loss": 1.5502, + "step": 68590 + }, + { + "epoch": 5.94, + "learning_rate": 4.410032054058737e-05, + "loss": 1.5544, + "step": 68600 + }, + { + "epoch": 5.94, + "learning_rate": 4.409945421467556e-05, + "loss": 1.5506, + "step": 68610 + }, + { + "epoch": 5.94, + "learning_rate": 4.4098587888763756e-05, + "loss": 1.5799, + "step": 68620 + }, + { + "epoch": 5.94, + "learning_rate": 4.4097721562851944e-05, + "loss": 1.5217, + "step": 68630 + }, + { + "epoch": 5.94, + "learning_rate": 4.409685523694014e-05, + "loss": 1.5315, + "step": 68640 + }, + { + "epoch": 5.94, + "learning_rate": 4.409598891102833e-05, + "loss": 1.6646, + "step": 68650 + }, + { + "epoch": 5.94, + "learning_rate": 4.409512258511652e-05, + "loss": 1.5224, + "step": 68660 + }, + { + "epoch": 5.94, + "learning_rate": 4.4094256259204715e-05, + "loss": 1.5552, + "step": 68670 + }, + { + "epoch": 5.94, + "learning_rate": 4.40933899332929e-05, + "loss": 1.5364, + "step": 68680 + }, + { + "epoch": 5.95, + "learning_rate": 4.40925236073811e-05, + "loss": 1.5433, + "step": 68690 + }, + { + "epoch": 5.95, + "learning_rate": 4.409165728146929e-05, + "loss": 1.5222, + "step": 68700 + }, + { + "epoch": 5.95, + "learning_rate": 4.409079095555748e-05, + "loss": 1.5401, + "step": 68710 + }, + { + "epoch": 5.95, + "learning_rate": 4.4089924629645674e-05, + "loss": 1.487, + "step": 68720 + }, + { + "epoch": 5.95, + "learning_rate": 4.408905830373387e-05, + "loss": 1.5298, + "step": 68730 + }, + { + "epoch": 5.95, + "learning_rate": 4.4088191977822056e-05, + "loss": 1.5118, + "step": 68740 + }, + { + "epoch": 5.95, + "learning_rate": 4.408732565191025e-05, + "loss": 1.5113, + "step": 68750 + }, + { + "epoch": 5.95, + "learning_rate": 4.4086459325998445e-05, + "loss": 1.6108, + "step": 68760 + }, + { + "epoch": 5.95, + "learning_rate": 4.408559300008663e-05, + "loss": 1.5927, + "step": 68770 + }, + { + "epoch": 5.95, + "learning_rate": 4.408472667417483e-05, + "loss": 1.5788, + "step": 68780 + }, + { + "epoch": 5.95, + "learning_rate": 4.4083860348263014e-05, + "loss": 1.5128, + "step": 68790 + }, + { + "epoch": 5.95, + "learning_rate": 4.408299402235121e-05, + "loss": 1.5742, + "step": 68800 + }, + { + "epoch": 5.96, + "learning_rate": 4.40821276964394e-05, + "loss": 1.4841, + "step": 68810 + }, + { + "epoch": 5.96, + "learning_rate": 4.408126137052759e-05, + "loss": 1.5502, + "step": 68820 + }, + { + "epoch": 5.96, + "learning_rate": 4.4080395044615785e-05, + "loss": 1.4879, + "step": 68830 + }, + { + "epoch": 5.96, + "learning_rate": 4.407952871870398e-05, + "loss": 1.5238, + "step": 68840 + }, + { + "epoch": 5.96, + "learning_rate": 4.407866239279217e-05, + "loss": 1.5037, + "step": 68850 + }, + { + "epoch": 5.96, + "learning_rate": 4.407779606688036e-05, + "loss": 1.5617, + "step": 68860 + }, + { + "epoch": 5.96, + "learning_rate": 4.4076929740968556e-05, + "loss": 1.5556, + "step": 68870 + }, + { + "epoch": 5.96, + "learning_rate": 4.4076063415056744e-05, + "loss": 1.5094, + "step": 68880 + }, + { + "epoch": 5.96, + "learning_rate": 4.407519708914494e-05, + "loss": 1.5315, + "step": 68890 + }, + { + "epoch": 5.96, + "learning_rate": 4.4074330763233126e-05, + "loss": 1.5532, + "step": 68900 + }, + { + "epoch": 5.96, + "learning_rate": 4.407346443732132e-05, + "loss": 1.4881, + "step": 68910 + }, + { + "epoch": 5.97, + "learning_rate": 4.4072598111409515e-05, + "loss": 1.4828, + "step": 68920 + }, + { + "epoch": 5.97, + "learning_rate": 4.40717317854977e-05, + "loss": 1.5616, + "step": 68930 + }, + { + "epoch": 5.97, + "learning_rate": 4.40708654595859e-05, + "loss": 1.5419, + "step": 68940 + }, + { + "epoch": 5.97, + "learning_rate": 4.406999913367409e-05, + "loss": 1.5545, + "step": 68950 + }, + { + "epoch": 5.97, + "learning_rate": 4.406913280776228e-05, + "loss": 1.4738, + "step": 68960 + }, + { + "epoch": 5.97, + "learning_rate": 4.4068266481850474e-05, + "loss": 1.5309, + "step": 68970 + }, + { + "epoch": 5.97, + "learning_rate": 4.406740015593867e-05, + "loss": 1.5922, + "step": 68980 + }, + { + "epoch": 5.97, + "learning_rate": 4.4066533830026856e-05, + "loss": 1.4871, + "step": 68990 + }, + { + "epoch": 5.97, + "learning_rate": 4.406566750411505e-05, + "loss": 1.5565, + "step": 69000 + }, + { + "epoch": 5.97, + "learning_rate": 4.406480117820324e-05, + "loss": 1.5557, + "step": 69010 + }, + { + "epoch": 5.97, + "learning_rate": 4.406393485229143e-05, + "loss": 1.5042, + "step": 69020 + }, + { + "epoch": 5.97, + "learning_rate": 4.406306852637963e-05, + "loss": 1.488, + "step": 69030 + }, + { + "epoch": 5.98, + "learning_rate": 4.4062202200467815e-05, + "loss": 1.513, + "step": 69040 + }, + { + "epoch": 5.98, + "learning_rate": 4.406133587455601e-05, + "loss": 1.5454, + "step": 69050 + }, + { + "epoch": 5.98, + "learning_rate": 4.4060469548644204e-05, + "loss": 1.4523, + "step": 69060 + }, + { + "epoch": 5.98, + "learning_rate": 4.405960322273239e-05, + "loss": 1.606, + "step": 69070 + }, + { + "epoch": 5.98, + "learning_rate": 4.4058736896820586e-05, + "loss": 1.503, + "step": 69080 + }, + { + "epoch": 5.98, + "learning_rate": 4.405787057090878e-05, + "loss": 1.5584, + "step": 69090 + }, + { + "epoch": 5.98, + "learning_rate": 4.405700424499697e-05, + "loss": 1.538, + "step": 69100 + }, + { + "epoch": 5.98, + "learning_rate": 4.405613791908516e-05, + "loss": 1.5719, + "step": 69110 + }, + { + "epoch": 5.98, + "learning_rate": 4.405527159317335e-05, + "loss": 1.5414, + "step": 69120 + }, + { + "epoch": 5.98, + "learning_rate": 4.4054405267261544e-05, + "loss": 1.5268, + "step": 69130 + }, + { + "epoch": 5.98, + "learning_rate": 4.405353894134974e-05, + "loss": 1.4903, + "step": 69140 + }, + { + "epoch": 5.99, + "learning_rate": 4.4052672615437927e-05, + "loss": 1.5311, + "step": 69150 + }, + { + "epoch": 5.99, + "learning_rate": 4.405180628952612e-05, + "loss": 1.5589, + "step": 69160 + }, + { + "epoch": 5.99, + "learning_rate": 4.4050939963614315e-05, + "loss": 1.5455, + "step": 69170 + }, + { + "epoch": 5.99, + "learning_rate": 4.40500736377025e-05, + "loss": 1.568, + "step": 69180 + }, + { + "epoch": 5.99, + "learning_rate": 4.40492073117907e-05, + "loss": 1.4618, + "step": 69190 + }, + { + "epoch": 5.99, + "learning_rate": 4.404834098587889e-05, + "loss": 1.587, + "step": 69200 + }, + { + "epoch": 5.99, + "learning_rate": 4.404747465996708e-05, + "loss": 1.5834, + "step": 69210 + }, + { + "epoch": 5.99, + "learning_rate": 4.4046608334055274e-05, + "loss": 1.571, + "step": 69220 + }, + { + "epoch": 5.99, + "learning_rate": 4.404574200814346e-05, + "loss": 1.5446, + "step": 69230 + }, + { + "epoch": 5.99, + "learning_rate": 4.4044875682231656e-05, + "loss": 1.4922, + "step": 69240 + }, + { + "epoch": 5.99, + "learning_rate": 4.404400935631985e-05, + "loss": 1.5391, + "step": 69250 + }, + { + "epoch": 5.99, + "learning_rate": 4.404314303040804e-05, + "loss": 1.6258, + "step": 69260 + }, + { + "epoch": 6.0, + "learning_rate": 4.404227670449623e-05, + "loss": 1.5352, + "step": 69270 + }, + { + "epoch": 6.0, + "learning_rate": 4.404141037858443e-05, + "loss": 1.5335, + "step": 69280 + }, + { + "epoch": 6.0, + "learning_rate": 4.4040544052672615e-05, + "loss": 1.4779, + "step": 69290 + }, + { + "epoch": 6.0, + "learning_rate": 4.403967772676081e-05, + "loss": 1.5198, + "step": 69300 + }, + { + "epoch": 6.0, + "learning_rate": 4.4038811400849e-05, + "loss": 1.5057, + "step": 69310 + }, + { + "epoch": 6.0, + "learning_rate": 4.403794507493719e-05, + "loss": 1.4822, + "step": 69320 + }, + { + "epoch": 6.0, + "eval_Bleu_1": 0.05626603052750424, + "eval_Bleu_2": 3.33403878750568e-11, + "eval_Bleu_3": 2.901914864430052e-14, + "eval_Bleu_4": 8.821378807975252e-16, + "eval_ROUGE_L": 0.12609436326063833, + "eval_cer": 0.9945371848092703, + "eval_em": 0, + "eval_f1": 0.06365302598881438, + "eval_loss": 1.4506486654281616, + "eval_runtime": 1655.5073, + "eval_samples_per_second": 3.102, + "eval_steps_per_second": 3.102, + "eval_wer": 0.9774986761339945, + "step": 69321 + }, + { + "epoch": 6.0, + "learning_rate": 4.4037078749025386e-05, + "loss": 1.5335, + "step": 69330 + }, + { + "epoch": 6.0, + "learning_rate": 4.4036212423113574e-05, + "loss": 1.4785, + "step": 69340 + }, + { + "epoch": 6.0, + "learning_rate": 4.403534609720177e-05, + "loss": 1.5581, + "step": 69350 + }, + { + "epoch": 6.0, + "learning_rate": 4.403447977128996e-05, + "loss": 1.4988, + "step": 69360 + }, + { + "epoch": 6.0, + "learning_rate": 4.403361344537815e-05, + "loss": 1.5158, + "step": 69370 + }, + { + "epoch": 6.01, + "learning_rate": 4.4032747119466345e-05, + "loss": 1.5411, + "step": 69380 + }, + { + "epoch": 6.01, + "learning_rate": 4.403188079355454e-05, + "loss": 1.5225, + "step": 69390 + }, + { + "epoch": 6.01, + "learning_rate": 4.403101446764273e-05, + "loss": 1.5206, + "step": 69400 + }, + { + "epoch": 6.01, + "learning_rate": 4.403014814173092e-05, + "loss": 1.5683, + "step": 69410 + }, + { + "epoch": 6.01, + "learning_rate": 4.402928181581911e-05, + "loss": 1.5218, + "step": 69420 + }, + { + "epoch": 6.01, + "learning_rate": 4.4028415489907303e-05, + "loss": 1.4808, + "step": 69430 + }, + { + "epoch": 6.01, + "learning_rate": 4.40275491639955e-05, + "loss": 1.5106, + "step": 69440 + }, + { + "epoch": 6.01, + "learning_rate": 4.4026682838083686e-05, + "loss": 1.5364, + "step": 69450 + }, + { + "epoch": 6.01, + "learning_rate": 4.402581651217188e-05, + "loss": 1.5641, + "step": 69460 + }, + { + "epoch": 6.01, + "learning_rate": 4.4024950186260075e-05, + "loss": 1.5149, + "step": 69470 + }, + { + "epoch": 6.01, + "learning_rate": 4.402408386034826e-05, + "loss": 1.5105, + "step": 69480 + }, + { + "epoch": 6.01, + "learning_rate": 4.402321753443646e-05, + "loss": 1.4601, + "step": 69490 + }, + { + "epoch": 6.02, + "learning_rate": 4.402235120852465e-05, + "loss": 1.4513, + "step": 69500 + }, + { + "epoch": 6.02, + "learning_rate": 4.402148488261284e-05, + "loss": 1.5722, + "step": 69510 + }, + { + "epoch": 6.02, + "learning_rate": 4.402061855670103e-05, + "loss": 1.5197, + "step": 69520 + }, + { + "epoch": 6.02, + "learning_rate": 4.401975223078922e-05, + "loss": 1.5344, + "step": 69530 + }, + { + "epoch": 6.02, + "learning_rate": 4.4018885904877415e-05, + "loss": 1.487, + "step": 69540 + }, + { + "epoch": 6.02, + "learning_rate": 4.401801957896561e-05, + "loss": 1.4894, + "step": 69550 + }, + { + "epoch": 6.02, + "learning_rate": 4.40171532530538e-05, + "loss": 1.5016, + "step": 69560 + }, + { + "epoch": 6.02, + "learning_rate": 4.401628692714199e-05, + "loss": 1.4801, + "step": 69570 + }, + { + "epoch": 6.02, + "learning_rate": 4.4015420601230186e-05, + "loss": 1.5658, + "step": 69580 + }, + { + "epoch": 6.02, + "learning_rate": 4.4014554275318374e-05, + "loss": 1.4964, + "step": 69590 + }, + { + "epoch": 6.02, + "learning_rate": 4.401368794940657e-05, + "loss": 1.4933, + "step": 69600 + }, + { + "epoch": 6.03, + "learning_rate": 4.401282162349476e-05, + "loss": 1.5574, + "step": 69610 + }, + { + "epoch": 6.03, + "learning_rate": 4.401195529758295e-05, + "loss": 1.5492, + "step": 69620 + }, + { + "epoch": 6.03, + "learning_rate": 4.4011088971671145e-05, + "loss": 1.5245, + "step": 69630 + }, + { + "epoch": 6.03, + "learning_rate": 4.401022264575933e-05, + "loss": 1.5415, + "step": 69640 + }, + { + "epoch": 6.03, + "learning_rate": 4.400935631984753e-05, + "loss": 1.499, + "step": 69650 + }, + { + "epoch": 6.03, + "learning_rate": 4.400848999393572e-05, + "loss": 1.5254, + "step": 69660 + }, + { + "epoch": 6.03, + "learning_rate": 4.400762366802391e-05, + "loss": 1.4793, + "step": 69670 + }, + { + "epoch": 6.03, + "learning_rate": 4.4006757342112104e-05, + "loss": 1.5006, + "step": 69680 + }, + { + "epoch": 6.03, + "learning_rate": 4.40058910162003e-05, + "loss": 1.5419, + "step": 69690 + }, + { + "epoch": 6.03, + "learning_rate": 4.4005024690288486e-05, + "loss": 1.4638, + "step": 69700 + }, + { + "epoch": 6.03, + "learning_rate": 4.400415836437668e-05, + "loss": 1.4719, + "step": 69710 + }, + { + "epoch": 6.03, + "learning_rate": 4.4003292038464875e-05, + "loss": 1.4975, + "step": 69720 + }, + { + "epoch": 6.04, + "learning_rate": 4.400242571255306e-05, + "loss": 1.5594, + "step": 69730 + }, + { + "epoch": 6.04, + "learning_rate": 4.400155938664126e-05, + "loss": 1.4932, + "step": 69740 + }, + { + "epoch": 6.04, + "learning_rate": 4.4000693060729445e-05, + "loss": 1.4677, + "step": 69750 + }, + { + "epoch": 6.04, + "learning_rate": 4.399982673481764e-05, + "loss": 1.4208, + "step": 69760 + }, + { + "epoch": 6.04, + "learning_rate": 4.3998960408905834e-05, + "loss": 1.496, + "step": 69770 + }, + { + "epoch": 6.04, + "learning_rate": 4.399809408299402e-05, + "loss": 1.4463, + "step": 69780 + }, + { + "epoch": 6.04, + "learning_rate": 4.3997227757082216e-05, + "loss": 1.5067, + "step": 69790 + }, + { + "epoch": 6.04, + "learning_rate": 4.399636143117041e-05, + "loss": 1.5127, + "step": 69800 + }, + { + "epoch": 6.04, + "learning_rate": 4.39954951052586e-05, + "loss": 1.4702, + "step": 69810 + }, + { + "epoch": 6.04, + "learning_rate": 4.399462877934679e-05, + "loss": 1.486, + "step": 69820 + }, + { + "epoch": 6.04, + "learning_rate": 4.399376245343499e-05, + "loss": 1.522, + "step": 69830 + }, + { + "epoch": 6.05, + "learning_rate": 4.3992896127523174e-05, + "loss": 1.5423, + "step": 69840 + }, + { + "epoch": 6.05, + "learning_rate": 4.399202980161137e-05, + "loss": 1.4439, + "step": 69850 + }, + { + "epoch": 6.05, + "learning_rate": 4.3991163475699557e-05, + "loss": 1.5652, + "step": 69860 + }, + { + "epoch": 6.05, + "learning_rate": 4.399029714978775e-05, + "loss": 1.5482, + "step": 69870 + }, + { + "epoch": 6.05, + "learning_rate": 4.3989430823875945e-05, + "loss": 1.4807, + "step": 69880 + }, + { + "epoch": 6.05, + "learning_rate": 4.398856449796413e-05, + "loss": 1.5347, + "step": 69890 + }, + { + "epoch": 6.05, + "learning_rate": 4.398769817205233e-05, + "loss": 1.5396, + "step": 69900 + }, + { + "epoch": 6.05, + "learning_rate": 4.398683184614052e-05, + "loss": 1.5255, + "step": 69910 + }, + { + "epoch": 6.05, + "learning_rate": 4.398596552022871e-05, + "loss": 1.4857, + "step": 69920 + }, + { + "epoch": 6.05, + "learning_rate": 4.3985099194316904e-05, + "loss": 1.4891, + "step": 69930 + }, + { + "epoch": 6.05, + "learning_rate": 4.39842328684051e-05, + "loss": 1.4741, + "step": 69940 + }, + { + "epoch": 6.05, + "learning_rate": 4.3983366542493286e-05, + "loss": 1.4888, + "step": 69950 + }, + { + "epoch": 6.06, + "learning_rate": 4.398250021658148e-05, + "loss": 1.5741, + "step": 69960 + }, + { + "epoch": 6.06, + "learning_rate": 4.398163389066967e-05, + "loss": 1.5495, + "step": 69970 + }, + { + "epoch": 6.06, + "learning_rate": 4.398076756475786e-05, + "loss": 1.506, + "step": 69980 + }, + { + "epoch": 6.06, + "learning_rate": 4.397990123884606e-05, + "loss": 1.5024, + "step": 69990 + }, + { + "epoch": 6.06, + "learning_rate": 4.3979034912934245e-05, + "loss": 1.5184, + "step": 70000 + }, + { + "epoch": 6.06, + "learning_rate": 4.397816858702244e-05, + "loss": 1.5179, + "step": 70010 + }, + { + "epoch": 6.06, + "learning_rate": 4.3977302261110634e-05, + "loss": 1.5166, + "step": 70020 + }, + { + "epoch": 6.06, + "learning_rate": 4.397643593519882e-05, + "loss": 1.512, + "step": 70030 + }, + { + "epoch": 6.06, + "learning_rate": 4.3975569609287016e-05, + "loss": 1.519, + "step": 70040 + }, + { + "epoch": 6.06, + "learning_rate": 4.3974703283375204e-05, + "loss": 1.4663, + "step": 70050 + }, + { + "epoch": 6.06, + "learning_rate": 4.39738369574634e-05, + "loss": 1.5228, + "step": 70060 + }, + { + "epoch": 6.07, + "learning_rate": 4.397297063155159e-05, + "loss": 1.5002, + "step": 70070 + }, + { + "epoch": 6.07, + "learning_rate": 4.397210430563978e-05, + "loss": 1.4728, + "step": 70080 + }, + { + "epoch": 6.07, + "learning_rate": 4.3971237979727975e-05, + "loss": 1.5616, + "step": 70090 + }, + { + "epoch": 6.07, + "learning_rate": 4.397037165381617e-05, + "loss": 1.5093, + "step": 70100 + }, + { + "epoch": 6.07, + "learning_rate": 4.396950532790436e-05, + "loss": 1.5179, + "step": 70110 + }, + { + "epoch": 6.07, + "learning_rate": 4.396863900199255e-05, + "loss": 1.5401, + "step": 70120 + }, + { + "epoch": 6.07, + "learning_rate": 4.3967772676080746e-05, + "loss": 1.4529, + "step": 70130 + }, + { + "epoch": 6.07, + "learning_rate": 4.3966906350168933e-05, + "loss": 1.4928, + "step": 70140 + }, + { + "epoch": 6.07, + "learning_rate": 4.396604002425713e-05, + "loss": 1.5444, + "step": 70150 + }, + { + "epoch": 6.07, + "learning_rate": 4.3965173698345316e-05, + "loss": 1.533, + "step": 70160 + }, + { + "epoch": 6.07, + "learning_rate": 4.396430737243351e-05, + "loss": 1.4674, + "step": 70170 + }, + { + "epoch": 6.07, + "learning_rate": 4.3963441046521704e-05, + "loss": 1.5495, + "step": 70180 + }, + { + "epoch": 6.08, + "learning_rate": 4.396257472060989e-05, + "loss": 1.4775, + "step": 70190 + }, + { + "epoch": 6.08, + "learning_rate": 4.3961708394698087e-05, + "loss": 1.4696, + "step": 70200 + }, + { + "epoch": 6.08, + "learning_rate": 4.396084206878628e-05, + "loss": 1.5194, + "step": 70210 + }, + { + "epoch": 6.08, + "learning_rate": 4.395997574287447e-05, + "loss": 1.5236, + "step": 70220 + }, + { + "epoch": 6.08, + "learning_rate": 4.395910941696266e-05, + "loss": 1.4891, + "step": 70230 + }, + { + "epoch": 6.08, + "learning_rate": 4.395824309105086e-05, + "loss": 1.5103, + "step": 70240 + }, + { + "epoch": 6.08, + "learning_rate": 4.3957376765139045e-05, + "loss": 1.4618, + "step": 70250 + }, + { + "epoch": 6.08, + "learning_rate": 4.395651043922724e-05, + "loss": 1.5405, + "step": 70260 + }, + { + "epoch": 6.08, + "learning_rate": 4.395564411331543e-05, + "loss": 1.5421, + "step": 70270 + }, + { + "epoch": 6.08, + "learning_rate": 4.395477778740362e-05, + "loss": 1.4557, + "step": 70280 + }, + { + "epoch": 6.08, + "learning_rate": 4.3953911461491816e-05, + "loss": 1.4548, + "step": 70290 + }, + { + "epoch": 6.08, + "learning_rate": 4.3953045135580004e-05, + "loss": 1.5061, + "step": 70300 + }, + { + "epoch": 6.09, + "learning_rate": 4.39521788096682e-05, + "loss": 1.4792, + "step": 70310 + }, + { + "epoch": 6.09, + "learning_rate": 4.395131248375639e-05, + "loss": 1.4456, + "step": 70320 + }, + { + "epoch": 6.09, + "learning_rate": 4.395044615784458e-05, + "loss": 1.5348, + "step": 70330 + }, + { + "epoch": 6.09, + "learning_rate": 4.3949579831932775e-05, + "loss": 1.5201, + "step": 70340 + }, + { + "epoch": 6.09, + "learning_rate": 4.394871350602097e-05, + "loss": 1.5111, + "step": 70350 + }, + { + "epoch": 6.09, + "learning_rate": 4.394784718010916e-05, + "loss": 1.5016, + "step": 70360 + }, + { + "epoch": 6.09, + "learning_rate": 4.394698085419735e-05, + "loss": 1.4524, + "step": 70370 + }, + { + "epoch": 6.09, + "learning_rate": 4.394611452828554e-05, + "loss": 1.4678, + "step": 70380 + }, + { + "epoch": 6.09, + "learning_rate": 4.3945248202373734e-05, + "loss": 1.4724, + "step": 70390 + }, + { + "epoch": 6.09, + "learning_rate": 4.394438187646193e-05, + "loss": 1.4747, + "step": 70400 + }, + { + "epoch": 6.09, + "learning_rate": 4.3943515550550116e-05, + "loss": 1.4823, + "step": 70410 + }, + { + "epoch": 6.1, + "learning_rate": 4.394264922463831e-05, + "loss": 1.4759, + "step": 70420 + }, + { + "epoch": 6.1, + "learning_rate": 4.3941782898726505e-05, + "loss": 1.5114, + "step": 70430 + }, + { + "epoch": 6.1, + "learning_rate": 4.394091657281469e-05, + "loss": 1.4858, + "step": 70440 + }, + { + "epoch": 6.1, + "learning_rate": 4.394005024690289e-05, + "loss": 1.4476, + "step": 70450 + }, + { + "epoch": 6.1, + "learning_rate": 4.393918392099108e-05, + "loss": 1.5216, + "step": 70460 + }, + { + "epoch": 6.1, + "learning_rate": 4.393831759507927e-05, + "loss": 1.5429, + "step": 70470 + }, + { + "epoch": 6.1, + "learning_rate": 4.3937451269167464e-05, + "loss": 1.4795, + "step": 70480 + }, + { + "epoch": 6.1, + "learning_rate": 4.393658494325565e-05, + "loss": 1.4728, + "step": 70490 + }, + { + "epoch": 6.1, + "learning_rate": 4.3935718617343846e-05, + "loss": 1.5497, + "step": 70500 + }, + { + "epoch": 6.1, + "learning_rate": 4.393485229143204e-05, + "loss": 1.5618, + "step": 70510 + }, + { + "epoch": 6.1, + "learning_rate": 4.393398596552023e-05, + "loss": 1.5796, + "step": 70520 + }, + { + "epoch": 6.1, + "learning_rate": 4.393311963960842e-05, + "loss": 1.5778, + "step": 70530 + }, + { + "epoch": 6.11, + "learning_rate": 4.393225331369662e-05, + "loss": 1.5505, + "step": 70540 + }, + { + "epoch": 6.11, + "learning_rate": 4.3931386987784804e-05, + "loss": 1.5536, + "step": 70550 + }, + { + "epoch": 6.11, + "learning_rate": 4.3930520661873e-05, + "loss": 1.5348, + "step": 70560 + }, + { + "epoch": 6.11, + "learning_rate": 4.392965433596119e-05, + "loss": 1.4702, + "step": 70570 + }, + { + "epoch": 6.11, + "learning_rate": 4.392878801004938e-05, + "loss": 1.4761, + "step": 70580 + }, + { + "epoch": 6.11, + "learning_rate": 4.3927921684137575e-05, + "loss": 1.4631, + "step": 70590 + }, + { + "epoch": 6.11, + "learning_rate": 4.392705535822576e-05, + "loss": 1.5581, + "step": 70600 + }, + { + "epoch": 6.11, + "learning_rate": 4.392618903231396e-05, + "loss": 1.5878, + "step": 70610 + }, + { + "epoch": 6.11, + "learning_rate": 4.392532270640215e-05, + "loss": 1.5347, + "step": 70620 + }, + { + "epoch": 6.11, + "learning_rate": 4.392445638049034e-05, + "loss": 1.4767, + "step": 70630 + }, + { + "epoch": 6.11, + "learning_rate": 4.3923590054578534e-05, + "loss": 1.5032, + "step": 70640 + }, + { + "epoch": 6.12, + "learning_rate": 4.392272372866673e-05, + "loss": 1.4898, + "step": 70650 + }, + { + "epoch": 6.12, + "learning_rate": 4.3921857402754916e-05, + "loss": 1.4874, + "step": 70660 + }, + { + "epoch": 6.12, + "learning_rate": 4.392099107684311e-05, + "loss": 1.4628, + "step": 70670 + }, + { + "epoch": 6.12, + "learning_rate": 4.3920124750931305e-05, + "loss": 1.4874, + "step": 70680 + }, + { + "epoch": 6.12, + "learning_rate": 4.391925842501949e-05, + "loss": 1.473, + "step": 70690 + }, + { + "epoch": 6.12, + "learning_rate": 4.391839209910769e-05, + "loss": 1.5351, + "step": 70700 + }, + { + "epoch": 6.12, + "learning_rate": 4.3917525773195875e-05, + "loss": 1.5348, + "step": 70710 + }, + { + "epoch": 6.12, + "learning_rate": 4.391665944728407e-05, + "loss": 1.4994, + "step": 70720 + }, + { + "epoch": 6.12, + "learning_rate": 4.3915793121372264e-05, + "loss": 1.4621, + "step": 70730 + }, + { + "epoch": 6.12, + "learning_rate": 4.391492679546045e-05, + "loss": 1.5273, + "step": 70740 + }, + { + "epoch": 6.12, + "learning_rate": 4.3914060469548646e-05, + "loss": 1.555, + "step": 70750 + }, + { + "epoch": 6.12, + "learning_rate": 4.391319414363684e-05, + "loss": 1.4917, + "step": 70760 + }, + { + "epoch": 6.13, + "learning_rate": 4.391232781772503e-05, + "loss": 1.4841, + "step": 70770 + }, + { + "epoch": 6.13, + "learning_rate": 4.391146149181322e-05, + "loss": 1.4783, + "step": 70780 + }, + { + "epoch": 6.13, + "learning_rate": 4.391059516590142e-05, + "loss": 1.5315, + "step": 70790 + }, + { + "epoch": 6.13, + "learning_rate": 4.3909728839989605e-05, + "loss": 1.4959, + "step": 70800 + }, + { + "epoch": 6.13, + "learning_rate": 4.39088625140778e-05, + "loss": 1.5608, + "step": 70810 + }, + { + "epoch": 6.13, + "learning_rate": 4.390799618816599e-05, + "loss": 1.4994, + "step": 70820 + }, + { + "epoch": 6.13, + "learning_rate": 4.390712986225418e-05, + "loss": 1.5426, + "step": 70830 + }, + { + "epoch": 6.13, + "learning_rate": 4.3906263536342376e-05, + "loss": 1.4951, + "step": 70840 + }, + { + "epoch": 6.13, + "learning_rate": 4.3905397210430563e-05, + "loss": 1.4875, + "step": 70850 + }, + { + "epoch": 6.13, + "learning_rate": 4.390453088451876e-05, + "loss": 1.5442, + "step": 70860 + }, + { + "epoch": 6.13, + "learning_rate": 4.390366455860695e-05, + "loss": 1.5449, + "step": 70870 + }, + { + "epoch": 6.14, + "learning_rate": 4.390279823269514e-05, + "loss": 1.4981, + "step": 70880 + }, + { + "epoch": 6.14, + "learning_rate": 4.3901931906783334e-05, + "loss": 1.5341, + "step": 70890 + }, + { + "epoch": 6.14, + "learning_rate": 4.390106558087152e-05, + "loss": 1.474, + "step": 70900 + }, + { + "epoch": 6.14, + "learning_rate": 4.3900199254959717e-05, + "loss": 1.5266, + "step": 70910 + }, + { + "epoch": 6.14, + "learning_rate": 4.389933292904791e-05, + "loss": 1.5583, + "step": 70920 + }, + { + "epoch": 6.14, + "learning_rate": 4.38984666031361e-05, + "loss": 1.5124, + "step": 70930 + }, + { + "epoch": 6.14, + "learning_rate": 4.389760027722429e-05, + "loss": 1.5325, + "step": 70940 + }, + { + "epoch": 6.14, + "learning_rate": 4.389673395131249e-05, + "loss": 1.4991, + "step": 70950 + }, + { + "epoch": 6.14, + "learning_rate": 4.3895867625400675e-05, + "loss": 1.5235, + "step": 70960 + }, + { + "epoch": 6.14, + "learning_rate": 4.389500129948887e-05, + "loss": 1.5085, + "step": 70970 + }, + { + "epoch": 6.14, + "learning_rate": 4.3894134973577064e-05, + "loss": 1.4055, + "step": 70980 + }, + { + "epoch": 6.14, + "learning_rate": 4.389326864766525e-05, + "loss": 1.481, + "step": 70990 + }, + { + "epoch": 6.15, + "learning_rate": 4.3892402321753446e-05, + "loss": 1.5346, + "step": 71000 + }, + { + "epoch": 6.15, + "learning_rate": 4.3891535995841634e-05, + "loss": 1.4523, + "step": 71010 + }, + { + "epoch": 6.15, + "learning_rate": 4.389066966992983e-05, + "loss": 1.4867, + "step": 71020 + }, + { + "epoch": 6.15, + "learning_rate": 4.388980334401802e-05, + "loss": 1.4665, + "step": 71030 + }, + { + "epoch": 6.15, + "learning_rate": 4.388893701810621e-05, + "loss": 1.5562, + "step": 71040 + }, + { + "epoch": 6.15, + "learning_rate": 4.3888070692194405e-05, + "loss": 1.5604, + "step": 71050 + }, + { + "epoch": 6.15, + "learning_rate": 4.38872043662826e-05, + "loss": 1.5254, + "step": 71060 + }, + { + "epoch": 6.15, + "learning_rate": 4.388633804037079e-05, + "loss": 1.4973, + "step": 71070 + }, + { + "epoch": 6.15, + "learning_rate": 4.388547171445898e-05, + "loss": 1.4876, + "step": 71080 + }, + { + "epoch": 6.15, + "learning_rate": 4.3884605388547176e-05, + "loss": 1.5092, + "step": 71090 + }, + { + "epoch": 6.15, + "learning_rate": 4.3883739062635364e-05, + "loss": 1.4888, + "step": 71100 + }, + { + "epoch": 6.16, + "learning_rate": 4.388287273672356e-05, + "loss": 1.5817, + "step": 71110 + }, + { + "epoch": 6.16, + "learning_rate": 4.3882006410811746e-05, + "loss": 1.4908, + "step": 71120 + }, + { + "epoch": 6.16, + "learning_rate": 4.388114008489994e-05, + "loss": 1.488, + "step": 71130 + }, + { + "epoch": 6.16, + "learning_rate": 4.3880273758988135e-05, + "loss": 1.5163, + "step": 71140 + }, + { + "epoch": 6.16, + "learning_rate": 4.387940743307632e-05, + "loss": 1.5112, + "step": 71150 + }, + { + "epoch": 6.16, + "learning_rate": 4.387854110716452e-05, + "loss": 1.4642, + "step": 71160 + }, + { + "epoch": 6.16, + "learning_rate": 4.387767478125271e-05, + "loss": 1.4612, + "step": 71170 + }, + { + "epoch": 6.16, + "learning_rate": 4.38768084553409e-05, + "loss": 1.5465, + "step": 71180 + }, + { + "epoch": 6.16, + "learning_rate": 4.3875942129429093e-05, + "loss": 1.4824, + "step": 71190 + }, + { + "epoch": 6.16, + "learning_rate": 4.387507580351729e-05, + "loss": 1.4783, + "step": 71200 + }, + { + "epoch": 6.16, + "learning_rate": 4.3874209477605476e-05, + "loss": 1.5558, + "step": 71210 + }, + { + "epoch": 6.16, + "learning_rate": 4.387334315169367e-05, + "loss": 1.4965, + "step": 71220 + }, + { + "epoch": 6.17, + "learning_rate": 4.387247682578186e-05, + "loss": 1.5193, + "step": 71230 + }, + { + "epoch": 6.17, + "learning_rate": 4.387161049987005e-05, + "loss": 1.5255, + "step": 71240 + }, + { + "epoch": 6.17, + "learning_rate": 4.387074417395825e-05, + "loss": 1.5407, + "step": 71250 + }, + { + "epoch": 6.17, + "learning_rate": 4.3869877848046434e-05, + "loss": 1.5612, + "step": 71260 + }, + { + "epoch": 6.17, + "learning_rate": 4.386901152213463e-05, + "loss": 1.4719, + "step": 71270 + }, + { + "epoch": 6.17, + "learning_rate": 4.386814519622282e-05, + "loss": 1.4616, + "step": 71280 + }, + { + "epoch": 6.17, + "learning_rate": 4.386727887031101e-05, + "loss": 1.4772, + "step": 71290 + }, + { + "epoch": 6.17, + "learning_rate": 4.3866412544399205e-05, + "loss": 1.4939, + "step": 71300 + }, + { + "epoch": 6.17, + "learning_rate": 4.38655462184874e-05, + "loss": 1.521, + "step": 71310 + }, + { + "epoch": 6.17, + "learning_rate": 4.386467989257559e-05, + "loss": 1.5639, + "step": 71320 + }, + { + "epoch": 6.17, + "learning_rate": 4.386381356666378e-05, + "loss": 1.5037, + "step": 71330 + }, + { + "epoch": 6.18, + "learning_rate": 4.386294724075197e-05, + "loss": 1.4579, + "step": 71340 + }, + { + "epoch": 6.18, + "learning_rate": 4.3862080914840164e-05, + "loss": 1.4701, + "step": 71350 + }, + { + "epoch": 6.18, + "learning_rate": 4.386121458892836e-05, + "loss": 1.4571, + "step": 71360 + }, + { + "epoch": 6.18, + "learning_rate": 4.3860348263016546e-05, + "loss": 1.527, + "step": 71370 + }, + { + "epoch": 6.18, + "learning_rate": 4.385948193710474e-05, + "loss": 1.4744, + "step": 71380 + }, + { + "epoch": 6.18, + "learning_rate": 4.3858615611192935e-05, + "loss": 1.4568, + "step": 71390 + }, + { + "epoch": 6.18, + "learning_rate": 4.385774928528112e-05, + "loss": 1.5162, + "step": 71400 + }, + { + "epoch": 6.18, + "learning_rate": 4.385688295936932e-05, + "loss": 1.4754, + "step": 71410 + }, + { + "epoch": 6.18, + "learning_rate": 4.385601663345751e-05, + "loss": 1.49, + "step": 71420 + }, + { + "epoch": 6.18, + "learning_rate": 4.38551503075457e-05, + "loss": 1.5426, + "step": 71430 + }, + { + "epoch": 6.18, + "learning_rate": 4.3854283981633894e-05, + "loss": 1.5328, + "step": 71440 + }, + { + "epoch": 6.18, + "learning_rate": 4.385341765572208e-05, + "loss": 1.5619, + "step": 71450 + }, + { + "epoch": 6.19, + "learning_rate": 4.3852551329810276e-05, + "loss": 1.4542, + "step": 71460 + }, + { + "epoch": 6.19, + "learning_rate": 4.385168500389847e-05, + "loss": 1.4798, + "step": 71470 + }, + { + "epoch": 6.19, + "learning_rate": 4.385081867798666e-05, + "loss": 1.5349, + "step": 71480 + }, + { + "epoch": 6.19, + "learning_rate": 4.384995235207485e-05, + "loss": 1.5039, + "step": 71490 + }, + { + "epoch": 6.19, + "learning_rate": 4.384908602616305e-05, + "loss": 1.5289, + "step": 71500 + }, + { + "epoch": 6.19, + "learning_rate": 4.3848219700251235e-05, + "loss": 1.4954, + "step": 71510 + }, + { + "epoch": 6.19, + "learning_rate": 4.384735337433943e-05, + "loss": 1.4767, + "step": 71520 + }, + { + "epoch": 6.19, + "learning_rate": 4.3846487048427624e-05, + "loss": 1.4608, + "step": 71530 + }, + { + "epoch": 6.19, + "learning_rate": 4.384562072251581e-05, + "loss": 1.5003, + "step": 71540 + }, + { + "epoch": 6.19, + "learning_rate": 4.3844754396604006e-05, + "loss": 1.5021, + "step": 71550 + }, + { + "epoch": 6.19, + "learning_rate": 4.384388807069219e-05, + "loss": 1.4847, + "step": 71560 + }, + { + "epoch": 6.19, + "learning_rate": 4.384302174478039e-05, + "loss": 1.5455, + "step": 71570 + }, + { + "epoch": 6.2, + "learning_rate": 4.384215541886858e-05, + "loss": 1.5193, + "step": 71580 + }, + { + "epoch": 6.2, + "learning_rate": 4.384128909295677e-05, + "loss": 1.4547, + "step": 71590 + }, + { + "epoch": 6.2, + "learning_rate": 4.3840422767044964e-05, + "loss": 1.5395, + "step": 71600 + }, + { + "epoch": 6.2, + "learning_rate": 4.383955644113316e-05, + "loss": 1.4684, + "step": 71610 + }, + { + "epoch": 6.2, + "learning_rate": 4.3838690115221347e-05, + "loss": 1.505, + "step": 71620 + }, + { + "epoch": 6.2, + "learning_rate": 4.383782378930954e-05, + "loss": 1.5054, + "step": 71630 + }, + { + "epoch": 6.2, + "learning_rate": 4.383695746339773e-05, + "loss": 1.5366, + "step": 71640 + }, + { + "epoch": 6.2, + "learning_rate": 4.383609113748592e-05, + "loss": 1.5264, + "step": 71650 + }, + { + "epoch": 6.2, + "learning_rate": 4.383522481157412e-05, + "loss": 1.4556, + "step": 71660 + }, + { + "epoch": 6.2, + "learning_rate": 4.3834358485662305e-05, + "loss": 1.533, + "step": 71670 + }, + { + "epoch": 6.2, + "learning_rate": 4.38334921597505e-05, + "loss": 1.5088, + "step": 71680 + }, + { + "epoch": 6.21, + "learning_rate": 4.3832625833838694e-05, + "loss": 1.4723, + "step": 71690 + }, + { + "epoch": 6.21, + "learning_rate": 4.383175950792688e-05, + "loss": 1.5579, + "step": 71700 + }, + { + "epoch": 6.21, + "learning_rate": 4.3830893182015076e-05, + "loss": 1.496, + "step": 71710 + }, + { + "epoch": 6.21, + "learning_rate": 4.383002685610327e-05, + "loss": 1.5597, + "step": 71720 + }, + { + "epoch": 6.21, + "learning_rate": 4.382916053019146e-05, + "loss": 1.5363, + "step": 71730 + }, + { + "epoch": 6.21, + "learning_rate": 4.382829420427965e-05, + "loss": 1.4748, + "step": 71740 + }, + { + "epoch": 6.21, + "learning_rate": 4.382742787836784e-05, + "loss": 1.5695, + "step": 71750 + }, + { + "epoch": 6.21, + "learning_rate": 4.3826561552456035e-05, + "loss": 1.5079, + "step": 71760 + }, + { + "epoch": 6.21, + "learning_rate": 4.382569522654423e-05, + "loss": 1.5351, + "step": 71770 + }, + { + "epoch": 6.21, + "learning_rate": 4.382482890063242e-05, + "loss": 1.5252, + "step": 71780 + }, + { + "epoch": 6.21, + "learning_rate": 4.382396257472061e-05, + "loss": 1.5589, + "step": 71790 + }, + { + "epoch": 6.21, + "learning_rate": 4.3823096248808806e-05, + "loss": 1.4951, + "step": 71800 + }, + { + "epoch": 6.22, + "learning_rate": 4.3822229922896994e-05, + "loss": 1.4385, + "step": 71810 + }, + { + "epoch": 6.22, + "learning_rate": 4.382136359698519e-05, + "loss": 1.4358, + "step": 71820 + }, + { + "epoch": 6.22, + "learning_rate": 4.382049727107338e-05, + "loss": 1.4501, + "step": 71830 + }, + { + "epoch": 6.22, + "learning_rate": 4.381963094516157e-05, + "loss": 1.5021, + "step": 71840 + }, + { + "epoch": 6.22, + "learning_rate": 4.3818764619249765e-05, + "loss": 1.4668, + "step": 71850 + }, + { + "epoch": 6.22, + "learning_rate": 4.381789829333795e-05, + "loss": 1.5389, + "step": 71860 + }, + { + "epoch": 6.22, + "learning_rate": 4.381703196742615e-05, + "loss": 1.5299, + "step": 71870 + }, + { + "epoch": 6.22, + "learning_rate": 4.381616564151434e-05, + "loss": 1.4935, + "step": 71880 + }, + { + "epoch": 6.22, + "learning_rate": 4.381529931560253e-05, + "loss": 1.4423, + "step": 71890 + }, + { + "epoch": 6.22, + "learning_rate": 4.3814432989690723e-05, + "loss": 1.4822, + "step": 71900 + }, + { + "epoch": 6.22, + "learning_rate": 4.381356666377892e-05, + "loss": 1.5518, + "step": 71910 + }, + { + "epoch": 6.23, + "learning_rate": 4.3812700337867106e-05, + "loss": 1.4918, + "step": 71920 + }, + { + "epoch": 6.23, + "learning_rate": 4.38118340119553e-05, + "loss": 1.4965, + "step": 71930 + }, + { + "epoch": 6.23, + "learning_rate": 4.3810967686043495e-05, + "loss": 1.4776, + "step": 71940 + }, + { + "epoch": 6.23, + "learning_rate": 4.381010136013168e-05, + "loss": 1.4356, + "step": 71950 + }, + { + "epoch": 6.23, + "learning_rate": 4.380923503421988e-05, + "loss": 1.5176, + "step": 71960 + }, + { + "epoch": 6.23, + "learning_rate": 4.3808368708308064e-05, + "loss": 1.4977, + "step": 71970 + }, + { + "epoch": 6.23, + "learning_rate": 4.380750238239626e-05, + "loss": 1.4761, + "step": 71980 + }, + { + "epoch": 6.23, + "learning_rate": 4.380663605648445e-05, + "loss": 1.5402, + "step": 71990 + }, + { + "epoch": 6.23, + "learning_rate": 4.380576973057264e-05, + "loss": 1.461, + "step": 72000 + }, + { + "epoch": 6.23, + "learning_rate": 4.3804903404660835e-05, + "loss": 1.5141, + "step": 72010 + }, + { + "epoch": 6.23, + "learning_rate": 4.380403707874903e-05, + "loss": 1.5219, + "step": 72020 + }, + { + "epoch": 6.23, + "learning_rate": 4.380317075283722e-05, + "loss": 1.5069, + "step": 72030 + }, + { + "epoch": 6.24, + "learning_rate": 4.380230442692541e-05, + "loss": 1.4512, + "step": 72040 + }, + { + "epoch": 6.24, + "learning_rate": 4.3801438101013606e-05, + "loss": 1.4859, + "step": 72050 + }, + { + "epoch": 6.24, + "learning_rate": 4.3800571775101794e-05, + "loss": 1.5444, + "step": 72060 + }, + { + "epoch": 6.24, + "learning_rate": 4.379970544918999e-05, + "loss": 1.4942, + "step": 72070 + }, + { + "epoch": 6.24, + "learning_rate": 4.3798839123278176e-05, + "loss": 1.4502, + "step": 72080 + }, + { + "epoch": 6.24, + "learning_rate": 4.379797279736637e-05, + "loss": 1.4711, + "step": 72090 + }, + { + "epoch": 6.24, + "learning_rate": 4.3797106471454565e-05, + "loss": 1.5208, + "step": 72100 + }, + { + "epoch": 6.24, + "learning_rate": 4.379624014554275e-05, + "loss": 1.466, + "step": 72110 + }, + { + "epoch": 6.24, + "learning_rate": 4.379537381963095e-05, + "loss": 1.4867, + "step": 72120 + }, + { + "epoch": 6.24, + "learning_rate": 4.379450749371914e-05, + "loss": 1.5686, + "step": 72130 + }, + { + "epoch": 6.24, + "learning_rate": 4.379364116780733e-05, + "loss": 1.4614, + "step": 72140 + }, + { + "epoch": 6.25, + "learning_rate": 4.3792774841895524e-05, + "loss": 1.4954, + "step": 72150 + }, + { + "epoch": 6.25, + "learning_rate": 4.379190851598372e-05, + "loss": 1.4366, + "step": 72160 + }, + { + "epoch": 6.25, + "learning_rate": 4.3791042190071906e-05, + "loss": 1.5367, + "step": 72170 + }, + { + "epoch": 6.25, + "learning_rate": 4.37901758641601e-05, + "loss": 1.4832, + "step": 72180 + }, + { + "epoch": 6.25, + "learning_rate": 4.378930953824829e-05, + "loss": 1.4773, + "step": 72190 + }, + { + "epoch": 6.25, + "learning_rate": 4.378844321233648e-05, + "loss": 1.4872, + "step": 72200 + }, + { + "epoch": 6.25, + "learning_rate": 4.378757688642468e-05, + "loss": 1.5011, + "step": 72210 + }, + { + "epoch": 6.25, + "learning_rate": 4.3786710560512865e-05, + "loss": 1.4432, + "step": 72220 + }, + { + "epoch": 6.25, + "learning_rate": 4.378584423460106e-05, + "loss": 1.4881, + "step": 72230 + }, + { + "epoch": 6.25, + "learning_rate": 4.3784977908689254e-05, + "loss": 1.4968, + "step": 72240 + }, + { + "epoch": 6.25, + "learning_rate": 4.378411158277744e-05, + "loss": 1.4846, + "step": 72250 + }, + { + "epoch": 6.25, + "learning_rate": 4.3783245256865636e-05, + "loss": 1.4937, + "step": 72260 + }, + { + "epoch": 6.26, + "learning_rate": 4.378237893095383e-05, + "loss": 1.5308, + "step": 72270 + }, + { + "epoch": 6.26, + "learning_rate": 4.378151260504202e-05, + "loss": 1.4925, + "step": 72280 + }, + { + "epoch": 6.26, + "learning_rate": 4.378064627913021e-05, + "loss": 1.494, + "step": 72290 + }, + { + "epoch": 6.26, + "learning_rate": 4.37797799532184e-05, + "loss": 1.4521, + "step": 72300 + }, + { + "epoch": 6.26, + "learning_rate": 4.3778913627306594e-05, + "loss": 1.5201, + "step": 72310 + }, + { + "epoch": 6.26, + "learning_rate": 4.377804730139479e-05, + "loss": 1.465, + "step": 72320 + }, + { + "epoch": 6.26, + "learning_rate": 4.3777180975482977e-05, + "loss": 1.4221, + "step": 72330 + }, + { + "epoch": 6.26, + "learning_rate": 4.377631464957117e-05, + "loss": 1.4327, + "step": 72340 + }, + { + "epoch": 6.26, + "learning_rate": 4.3775448323659365e-05, + "loss": 1.5517, + "step": 72350 + }, + { + "epoch": 6.26, + "learning_rate": 4.377458199774755e-05, + "loss": 1.5905, + "step": 72360 + }, + { + "epoch": 6.26, + "learning_rate": 4.377371567183575e-05, + "loss": 1.5197, + "step": 72370 + }, + { + "epoch": 6.27, + "learning_rate": 4.3772849345923935e-05, + "loss": 1.511, + "step": 72380 + }, + { + "epoch": 6.27, + "learning_rate": 4.377198302001213e-05, + "loss": 1.5215, + "step": 72390 + }, + { + "epoch": 6.27, + "learning_rate": 4.3771116694100324e-05, + "loss": 1.5885, + "step": 72400 + }, + { + "epoch": 6.27, + "learning_rate": 4.377025036818851e-05, + "loss": 1.4886, + "step": 72410 + }, + { + "epoch": 6.27, + "learning_rate": 4.3769384042276706e-05, + "loss": 1.4808, + "step": 72420 + }, + { + "epoch": 6.27, + "learning_rate": 4.37685177163649e-05, + "loss": 1.4677, + "step": 72430 + }, + { + "epoch": 6.27, + "learning_rate": 4.376765139045309e-05, + "loss": 1.5166, + "step": 72440 + }, + { + "epoch": 6.27, + "learning_rate": 4.376678506454128e-05, + "loss": 1.5019, + "step": 72450 + }, + { + "epoch": 6.27, + "learning_rate": 4.376591873862948e-05, + "loss": 1.5491, + "step": 72460 + }, + { + "epoch": 6.27, + "learning_rate": 4.3765052412717665e-05, + "loss": 1.4884, + "step": 72470 + }, + { + "epoch": 6.27, + "learning_rate": 4.376418608680586e-05, + "loss": 1.5057, + "step": 72480 + }, + { + "epoch": 6.27, + "learning_rate": 4.376331976089405e-05, + "loss": 1.5185, + "step": 72490 + }, + { + "epoch": 6.28, + "learning_rate": 4.376245343498224e-05, + "loss": 1.474, + "step": 72500 + }, + { + "epoch": 6.28, + "learning_rate": 4.3761587109070436e-05, + "loss": 1.4298, + "step": 72510 + }, + { + "epoch": 6.28, + "learning_rate": 4.3760720783158624e-05, + "loss": 1.4594, + "step": 72520 + }, + { + "epoch": 6.28, + "learning_rate": 4.375985445724682e-05, + "loss": 1.5329, + "step": 72530 + }, + { + "epoch": 6.28, + "learning_rate": 4.375898813133501e-05, + "loss": 1.4944, + "step": 72540 + }, + { + "epoch": 6.28, + "learning_rate": 4.37581218054232e-05, + "loss": 1.5245, + "step": 72550 + }, + { + "epoch": 6.28, + "learning_rate": 4.3757255479511395e-05, + "loss": 1.5209, + "step": 72560 + }, + { + "epoch": 6.28, + "learning_rate": 4.375638915359959e-05, + "loss": 1.5083, + "step": 72570 + }, + { + "epoch": 6.28, + "learning_rate": 4.375552282768778e-05, + "loss": 1.4413, + "step": 72580 + }, + { + "epoch": 6.28, + "learning_rate": 4.375465650177597e-05, + "loss": 1.4893, + "step": 72590 + }, + { + "epoch": 6.28, + "learning_rate": 4.375379017586416e-05, + "loss": 1.5259, + "step": 72600 + }, + { + "epoch": 6.28, + "learning_rate": 4.3752923849952353e-05, + "loss": 1.4881, + "step": 72610 + }, + { + "epoch": 6.29, + "learning_rate": 4.375205752404055e-05, + "loss": 1.4366, + "step": 72620 + }, + { + "epoch": 6.29, + "learning_rate": 4.3751191198128736e-05, + "loss": 1.5241, + "step": 72630 + }, + { + "epoch": 6.29, + "learning_rate": 4.375032487221693e-05, + "loss": 1.4893, + "step": 72640 + }, + { + "epoch": 6.29, + "learning_rate": 4.3749458546305124e-05, + "loss": 1.4734, + "step": 72650 + }, + { + "epoch": 6.29, + "learning_rate": 4.374859222039331e-05, + "loss": 1.4943, + "step": 72660 + }, + { + "epoch": 6.29, + "learning_rate": 4.3747725894481507e-05, + "loss": 1.4968, + "step": 72670 + }, + { + "epoch": 6.29, + "learning_rate": 4.37468595685697e-05, + "loss": 1.5028, + "step": 72680 + }, + { + "epoch": 6.29, + "learning_rate": 4.374599324265789e-05, + "loss": 1.5171, + "step": 72690 + }, + { + "epoch": 6.29, + "learning_rate": 4.374512691674608e-05, + "loss": 1.5427, + "step": 72700 + }, + { + "epoch": 6.29, + "learning_rate": 4.374426059083427e-05, + "loss": 1.5409, + "step": 72710 + }, + { + "epoch": 6.29, + "learning_rate": 4.3743394264922465e-05, + "loss": 1.5833, + "step": 72720 + }, + { + "epoch": 6.3, + "learning_rate": 4.374252793901066e-05, + "loss": 1.5173, + "step": 72730 + }, + { + "epoch": 6.3, + "learning_rate": 4.374166161309885e-05, + "loss": 1.5266, + "step": 72740 + }, + { + "epoch": 6.3, + "learning_rate": 4.374079528718704e-05, + "loss": 1.5234, + "step": 72750 + }, + { + "epoch": 6.3, + "learning_rate": 4.3739928961275236e-05, + "loss": 1.5089, + "step": 72760 + }, + { + "epoch": 6.3, + "learning_rate": 4.3739062635363424e-05, + "loss": 1.5113, + "step": 72770 + }, + { + "epoch": 6.3, + "learning_rate": 4.373819630945162e-05, + "loss": 1.5442, + "step": 72780 + }, + { + "epoch": 6.3, + "learning_rate": 4.373732998353981e-05, + "loss": 1.4574, + "step": 72790 + }, + { + "epoch": 6.3, + "learning_rate": 4.3736463657628e-05, + "loss": 1.5011, + "step": 72800 + }, + { + "epoch": 6.3, + "learning_rate": 4.3735597331716195e-05, + "loss": 1.5254, + "step": 72810 + }, + { + "epoch": 6.3, + "learning_rate": 4.373473100580438e-05, + "loss": 1.4311, + "step": 72820 + }, + { + "epoch": 6.3, + "learning_rate": 4.373386467989258e-05, + "loss": 1.5305, + "step": 72830 + }, + { + "epoch": 6.3, + "learning_rate": 4.373299835398077e-05, + "loss": 1.4647, + "step": 72840 + }, + { + "epoch": 6.31, + "learning_rate": 4.373213202806896e-05, + "loss": 1.47, + "step": 72850 + }, + { + "epoch": 6.31, + "learning_rate": 4.3731265702157154e-05, + "loss": 1.4349, + "step": 72860 + }, + { + "epoch": 6.31, + "learning_rate": 4.373039937624535e-05, + "loss": 1.4384, + "step": 72870 + }, + { + "epoch": 6.31, + "learning_rate": 4.3729533050333536e-05, + "loss": 1.457, + "step": 72880 + }, + { + "epoch": 6.31, + "learning_rate": 4.372866672442173e-05, + "loss": 1.5097, + "step": 72890 + }, + { + "epoch": 6.31, + "learning_rate": 4.3727800398509925e-05, + "loss": 1.4779, + "step": 72900 + }, + { + "epoch": 6.31, + "learning_rate": 4.372693407259811e-05, + "loss": 1.5303, + "step": 72910 + }, + { + "epoch": 6.31, + "learning_rate": 4.372606774668631e-05, + "loss": 1.5255, + "step": 72920 + }, + { + "epoch": 6.31, + "learning_rate": 4.3725201420774495e-05, + "loss": 1.5159, + "step": 72930 + }, + { + "epoch": 6.31, + "learning_rate": 4.372433509486269e-05, + "loss": 1.5031, + "step": 72940 + }, + { + "epoch": 6.31, + "learning_rate": 4.3723468768950884e-05, + "loss": 1.5043, + "step": 72950 + }, + { + "epoch": 6.32, + "learning_rate": 4.372260244303907e-05, + "loss": 1.4631, + "step": 72960 + }, + { + "epoch": 6.32, + "learning_rate": 4.3721736117127266e-05, + "loss": 1.5286, + "step": 72970 + }, + { + "epoch": 6.32, + "learning_rate": 4.372086979121546e-05, + "loss": 1.5156, + "step": 72980 + }, + { + "epoch": 6.32, + "learning_rate": 4.372000346530365e-05, + "loss": 1.5287, + "step": 72990 + }, + { + "epoch": 6.32, + "learning_rate": 4.371913713939184e-05, + "loss": 1.4642, + "step": 73000 + }, + { + "epoch": 6.32, + "learning_rate": 4.371827081348004e-05, + "loss": 1.4447, + "step": 73010 + }, + { + "epoch": 6.32, + "learning_rate": 4.3717404487568224e-05, + "loss": 1.4909, + "step": 73020 + }, + { + "epoch": 6.32, + "learning_rate": 4.371653816165642e-05, + "loss": 1.4991, + "step": 73030 + }, + { + "epoch": 6.32, + "learning_rate": 4.3715671835744606e-05, + "loss": 1.4779, + "step": 73040 + }, + { + "epoch": 6.32, + "learning_rate": 4.37148055098328e-05, + "loss": 1.4441, + "step": 73050 + }, + { + "epoch": 6.32, + "learning_rate": 4.3713939183920995e-05, + "loss": 1.4699, + "step": 73060 + }, + { + "epoch": 6.32, + "learning_rate": 4.371307285800918e-05, + "loss": 1.499, + "step": 73070 + }, + { + "epoch": 6.33, + "learning_rate": 4.371220653209738e-05, + "loss": 1.5933, + "step": 73080 + }, + { + "epoch": 6.33, + "learning_rate": 4.371134020618557e-05, + "loss": 1.4957, + "step": 73090 + }, + { + "epoch": 6.33, + "learning_rate": 4.371047388027376e-05, + "loss": 1.522, + "step": 73100 + }, + { + "epoch": 6.33, + "learning_rate": 4.3709607554361954e-05, + "loss": 1.4787, + "step": 73110 + }, + { + "epoch": 6.33, + "learning_rate": 4.370874122845015e-05, + "loss": 1.4824, + "step": 73120 + }, + { + "epoch": 6.33, + "learning_rate": 4.3707874902538336e-05, + "loss": 1.475, + "step": 73130 + }, + { + "epoch": 6.33, + "learning_rate": 4.370700857662653e-05, + "loss": 1.5157, + "step": 73140 + }, + { + "epoch": 6.33, + "learning_rate": 4.370614225071472e-05, + "loss": 1.5236, + "step": 73150 + }, + { + "epoch": 6.33, + "learning_rate": 4.370527592480291e-05, + "loss": 1.4666, + "step": 73160 + }, + { + "epoch": 6.33, + "learning_rate": 4.370440959889111e-05, + "loss": 1.5218, + "step": 73170 + }, + { + "epoch": 6.33, + "learning_rate": 4.3703543272979295e-05, + "loss": 1.5467, + "step": 73180 + }, + { + "epoch": 6.34, + "learning_rate": 4.370267694706749e-05, + "loss": 1.4766, + "step": 73190 + }, + { + "epoch": 6.34, + "learning_rate": 4.3701810621155684e-05, + "loss": 1.5785, + "step": 73200 + }, + { + "epoch": 6.34, + "learning_rate": 4.370094429524387e-05, + "loss": 1.5478, + "step": 73210 + }, + { + "epoch": 6.34, + "learning_rate": 4.3700077969332066e-05, + "loss": 1.5258, + "step": 73220 + }, + { + "epoch": 6.34, + "learning_rate": 4.3699211643420254e-05, + "loss": 1.4893, + "step": 73230 + }, + { + "epoch": 6.34, + "learning_rate": 4.369834531750845e-05, + "loss": 1.4953, + "step": 73240 + }, + { + "epoch": 6.34, + "learning_rate": 4.369747899159664e-05, + "loss": 1.4854, + "step": 73250 + }, + { + "epoch": 6.34, + "learning_rate": 4.369661266568483e-05, + "loss": 1.4565, + "step": 73260 + }, + { + "epoch": 6.34, + "learning_rate": 4.3695746339773025e-05, + "loss": 1.574, + "step": 73270 + }, + { + "epoch": 6.34, + "learning_rate": 4.369488001386122e-05, + "loss": 1.541, + "step": 73280 + }, + { + "epoch": 6.34, + "learning_rate": 4.369401368794941e-05, + "loss": 1.5, + "step": 73290 + }, + { + "epoch": 6.34, + "learning_rate": 4.36931473620376e-05, + "loss": 1.5241, + "step": 73300 + }, + { + "epoch": 6.35, + "learning_rate": 4.3692281036125796e-05, + "loss": 1.5813, + "step": 73310 + }, + { + "epoch": 6.35, + "learning_rate": 4.3691414710213983e-05, + "loss": 1.5122, + "step": 73320 + }, + { + "epoch": 6.35, + "learning_rate": 4.369054838430218e-05, + "loss": 1.4256, + "step": 73330 + }, + { + "epoch": 6.35, + "learning_rate": 4.3689682058390366e-05, + "loss": 1.4912, + "step": 73340 + }, + { + "epoch": 6.35, + "learning_rate": 4.368881573247856e-05, + "loss": 1.4938, + "step": 73350 + }, + { + "epoch": 6.35, + "learning_rate": 4.3687949406566754e-05, + "loss": 1.4433, + "step": 73360 + }, + { + "epoch": 6.35, + "learning_rate": 4.368708308065494e-05, + "loss": 1.5452, + "step": 73370 + }, + { + "epoch": 6.35, + "learning_rate": 4.3686216754743137e-05, + "loss": 1.4192, + "step": 73380 + }, + { + "epoch": 6.35, + "learning_rate": 4.368535042883133e-05, + "loss": 1.5146, + "step": 73390 + }, + { + "epoch": 6.35, + "learning_rate": 4.368448410291952e-05, + "loss": 1.5035, + "step": 73400 + }, + { + "epoch": 6.35, + "learning_rate": 4.368361777700771e-05, + "loss": 1.5025, + "step": 73410 + }, + { + "epoch": 6.36, + "learning_rate": 4.368275145109591e-05, + "loss": 1.5031, + "step": 73420 + }, + { + "epoch": 6.36, + "learning_rate": 4.3681885125184095e-05, + "loss": 1.5094, + "step": 73430 + }, + { + "epoch": 6.36, + "learning_rate": 4.368101879927229e-05, + "loss": 1.4721, + "step": 73440 + }, + { + "epoch": 6.36, + "learning_rate": 4.368015247336048e-05, + "loss": 1.5549, + "step": 73450 + }, + { + "epoch": 6.36, + "learning_rate": 4.367928614744867e-05, + "loss": 1.5621, + "step": 73460 + }, + { + "epoch": 6.36, + "learning_rate": 4.3678419821536866e-05, + "loss": 1.5555, + "step": 73470 + }, + { + "epoch": 6.36, + "learning_rate": 4.3677553495625054e-05, + "loss": 1.4821, + "step": 73480 + }, + { + "epoch": 6.36, + "learning_rate": 4.367668716971325e-05, + "loss": 1.5044, + "step": 73490 + }, + { + "epoch": 6.36, + "learning_rate": 4.367582084380144e-05, + "loss": 1.5058, + "step": 73500 + }, + { + "epoch": 6.36, + "learning_rate": 4.367495451788963e-05, + "loss": 1.5278, + "step": 73510 + }, + { + "epoch": 6.36, + "learning_rate": 4.3674088191977825e-05, + "loss": 1.4407, + "step": 73520 + }, + { + "epoch": 6.36, + "learning_rate": 4.367322186606602e-05, + "loss": 1.5646, + "step": 73530 + }, + { + "epoch": 6.37, + "learning_rate": 4.367235554015421e-05, + "loss": 1.5123, + "step": 73540 + }, + { + "epoch": 6.37, + "learning_rate": 4.36714892142424e-05, + "loss": 1.5345, + "step": 73550 + }, + { + "epoch": 6.37, + "learning_rate": 4.367062288833059e-05, + "loss": 1.535, + "step": 73560 + }, + { + "epoch": 6.37, + "learning_rate": 4.3669756562418784e-05, + "loss": 1.499, + "step": 73570 + }, + { + "epoch": 6.37, + "learning_rate": 4.366889023650698e-05, + "loss": 1.4954, + "step": 73580 + }, + { + "epoch": 6.37, + "learning_rate": 4.3668023910595166e-05, + "loss": 1.5329, + "step": 73590 + }, + { + "epoch": 6.37, + "learning_rate": 4.366715758468336e-05, + "loss": 1.4953, + "step": 73600 + }, + { + "epoch": 6.37, + "learning_rate": 4.3666291258771555e-05, + "loss": 1.4814, + "step": 73610 + }, + { + "epoch": 6.37, + "learning_rate": 4.366542493285974e-05, + "loss": 1.4727, + "step": 73620 + }, + { + "epoch": 6.37, + "learning_rate": 4.366455860694794e-05, + "loss": 1.5032, + "step": 73630 + }, + { + "epoch": 6.37, + "learning_rate": 4.366369228103613e-05, + "loss": 1.5921, + "step": 73640 + }, + { + "epoch": 6.37, + "learning_rate": 4.366282595512432e-05, + "loss": 1.5517, + "step": 73650 + }, + { + "epoch": 6.38, + "learning_rate": 4.3661959629212513e-05, + "loss": 1.4082, + "step": 73660 + }, + { + "epoch": 6.38, + "learning_rate": 4.36610933033007e-05, + "loss": 1.5028, + "step": 73670 + }, + { + "epoch": 6.38, + "learning_rate": 4.3660226977388896e-05, + "loss": 1.5116, + "step": 73680 + }, + { + "epoch": 6.38, + "learning_rate": 4.365936065147709e-05, + "loss": 1.4956, + "step": 73690 + }, + { + "epoch": 6.38, + "learning_rate": 4.365849432556528e-05, + "loss": 1.517, + "step": 73700 + }, + { + "epoch": 6.38, + "learning_rate": 4.365762799965347e-05, + "loss": 1.4742, + "step": 73710 + }, + { + "epoch": 6.38, + "learning_rate": 4.365676167374167e-05, + "loss": 1.5331, + "step": 73720 + }, + { + "epoch": 6.38, + "learning_rate": 4.3655895347829854e-05, + "loss": 1.506, + "step": 73730 + }, + { + "epoch": 6.38, + "learning_rate": 4.365502902191805e-05, + "loss": 1.4912, + "step": 73740 + }, + { + "epoch": 6.38, + "learning_rate": 4.365416269600624e-05, + "loss": 1.4578, + "step": 73750 + }, + { + "epoch": 6.38, + "learning_rate": 4.365329637009443e-05, + "loss": 1.4845, + "step": 73760 + }, + { + "epoch": 6.39, + "learning_rate": 4.3652430044182625e-05, + "loss": 1.5235, + "step": 73770 + }, + { + "epoch": 6.39, + "learning_rate": 4.365156371827081e-05, + "loss": 1.4953, + "step": 73780 + }, + { + "epoch": 6.39, + "learning_rate": 4.365069739235901e-05, + "loss": 1.5055, + "step": 73790 + }, + { + "epoch": 6.39, + "learning_rate": 4.36498310664472e-05, + "loss": 1.4963, + "step": 73800 + }, + { + "epoch": 6.39, + "learning_rate": 4.364896474053539e-05, + "loss": 1.4985, + "step": 73810 + }, + { + "epoch": 6.39, + "learning_rate": 4.3648098414623584e-05, + "loss": 1.4893, + "step": 73820 + }, + { + "epoch": 6.39, + "learning_rate": 4.364723208871178e-05, + "loss": 1.4959, + "step": 73830 + }, + { + "epoch": 6.39, + "learning_rate": 4.3646365762799966e-05, + "loss": 1.4996, + "step": 73840 + }, + { + "epoch": 6.39, + "learning_rate": 4.364549943688816e-05, + "loss": 1.5539, + "step": 73850 + }, + { + "epoch": 6.39, + "learning_rate": 4.3644633110976355e-05, + "loss": 1.5514, + "step": 73860 + }, + { + "epoch": 6.39, + "learning_rate": 4.364376678506454e-05, + "loss": 1.4571, + "step": 73870 + }, + { + "epoch": 6.39, + "learning_rate": 4.364290045915274e-05, + "loss": 1.5037, + "step": 73880 + }, + { + "epoch": 6.4, + "learning_rate": 4.3642034133240925e-05, + "loss": 1.4915, + "step": 73890 + }, + { + "epoch": 6.4, + "learning_rate": 4.364116780732912e-05, + "loss": 1.4662, + "step": 73900 + }, + { + "epoch": 6.4, + "learning_rate": 4.3640301481417314e-05, + "loss": 1.4676, + "step": 73910 + }, + { + "epoch": 6.4, + "learning_rate": 4.36394351555055e-05, + "loss": 1.5238, + "step": 73920 + }, + { + "epoch": 6.4, + "learning_rate": 4.3638568829593696e-05, + "loss": 1.4516, + "step": 73930 + }, + { + "epoch": 6.4, + "learning_rate": 4.363770250368189e-05, + "loss": 1.4639, + "step": 73940 + }, + { + "epoch": 6.4, + "learning_rate": 4.363683617777008e-05, + "loss": 1.4991, + "step": 73950 + }, + { + "epoch": 6.4, + "learning_rate": 4.363596985185827e-05, + "loss": 1.5245, + "step": 73960 + }, + { + "epoch": 6.4, + "learning_rate": 4.363510352594646e-05, + "loss": 1.4809, + "step": 73970 + }, + { + "epoch": 6.4, + "learning_rate": 4.3634237200034655e-05, + "loss": 1.5022, + "step": 73980 + }, + { + "epoch": 6.4, + "learning_rate": 4.363337087412285e-05, + "loss": 1.5155, + "step": 73990 + }, + { + "epoch": 6.41, + "learning_rate": 4.363250454821104e-05, + "loss": 1.4746, + "step": 74000 + }, + { + "epoch": 6.41, + "learning_rate": 4.363163822229923e-05, + "loss": 1.4907, + "step": 74010 + }, + { + "epoch": 6.41, + "learning_rate": 4.3630771896387426e-05, + "loss": 1.4188, + "step": 74020 + }, + { + "epoch": 6.41, + "learning_rate": 4.362990557047561e-05, + "loss": 1.4816, + "step": 74030 + }, + { + "epoch": 6.41, + "learning_rate": 4.362903924456381e-05, + "loss": 1.4846, + "step": 74040 + }, + { + "epoch": 6.41, + "learning_rate": 4.3628172918652e-05, + "loss": 1.5411, + "step": 74050 + }, + { + "epoch": 6.41, + "learning_rate": 4.362730659274019e-05, + "loss": 1.4935, + "step": 74060 + }, + { + "epoch": 6.41, + "learning_rate": 4.3626440266828384e-05, + "loss": 1.5002, + "step": 74070 + }, + { + "epoch": 6.41, + "learning_rate": 4.362557394091657e-05, + "loss": 1.4963, + "step": 74080 + }, + { + "epoch": 6.41, + "learning_rate": 4.3624707615004767e-05, + "loss": 1.4895, + "step": 74090 + }, + { + "epoch": 6.41, + "learning_rate": 4.362384128909296e-05, + "loss": 1.5432, + "step": 74100 + }, + { + "epoch": 6.41, + "learning_rate": 4.362297496318115e-05, + "loss": 1.4759, + "step": 74110 + }, + { + "epoch": 6.42, + "learning_rate": 4.362210863726934e-05, + "loss": 1.4809, + "step": 74120 + }, + { + "epoch": 6.42, + "learning_rate": 4.362124231135754e-05, + "loss": 1.4946, + "step": 74130 + }, + { + "epoch": 6.42, + "learning_rate": 4.3620375985445725e-05, + "loss": 1.4699, + "step": 74140 + }, + { + "epoch": 6.42, + "learning_rate": 4.361950965953392e-05, + "loss": 1.496, + "step": 74150 + }, + { + "epoch": 6.42, + "learning_rate": 4.3618643333622114e-05, + "loss": 1.467, + "step": 74160 + }, + { + "epoch": 6.42, + "learning_rate": 4.36177770077103e-05, + "loss": 1.4884, + "step": 74170 + }, + { + "epoch": 6.42, + "learning_rate": 4.3616910681798496e-05, + "loss": 1.5523, + "step": 74180 + }, + { + "epoch": 6.42, + "learning_rate": 4.3616044355886684e-05, + "loss": 1.5561, + "step": 74190 + }, + { + "epoch": 6.42, + "learning_rate": 4.361517802997488e-05, + "loss": 1.463, + "step": 74200 + }, + { + "epoch": 6.42, + "learning_rate": 4.361431170406307e-05, + "loss": 1.5183, + "step": 74210 + }, + { + "epoch": 6.42, + "learning_rate": 4.361344537815126e-05, + "loss": 1.5425, + "step": 74220 + }, + { + "epoch": 6.43, + "learning_rate": 4.3612579052239455e-05, + "loss": 1.5017, + "step": 74230 + }, + { + "epoch": 6.43, + "learning_rate": 4.361171272632765e-05, + "loss": 1.4104, + "step": 74240 + }, + { + "epoch": 6.43, + "learning_rate": 4.361084640041584e-05, + "loss": 1.5077, + "step": 74250 + }, + { + "epoch": 6.43, + "learning_rate": 4.360998007450403e-05, + "loss": 1.4598, + "step": 74260 + }, + { + "epoch": 6.43, + "learning_rate": 4.3609113748592226e-05, + "loss": 1.4967, + "step": 74270 + }, + { + "epoch": 6.43, + "learning_rate": 4.3608247422680414e-05, + "loss": 1.4766, + "step": 74280 + }, + { + "epoch": 6.43, + "learning_rate": 4.360738109676861e-05, + "loss": 1.5108, + "step": 74290 + }, + { + "epoch": 6.43, + "learning_rate": 4.3606514770856796e-05, + "loss": 1.476, + "step": 74300 + }, + { + "epoch": 6.43, + "learning_rate": 4.360564844494499e-05, + "loss": 1.4384, + "step": 74310 + }, + { + "epoch": 6.43, + "learning_rate": 4.3604782119033185e-05, + "loss": 1.4997, + "step": 74320 + }, + { + "epoch": 6.43, + "learning_rate": 4.360391579312137e-05, + "loss": 1.5638, + "step": 74330 + }, + { + "epoch": 6.43, + "learning_rate": 4.360304946720957e-05, + "loss": 1.4488, + "step": 74340 + }, + { + "epoch": 6.44, + "learning_rate": 4.360218314129776e-05, + "loss": 1.4198, + "step": 74350 + }, + { + "epoch": 6.44, + "learning_rate": 4.360131681538595e-05, + "loss": 1.507, + "step": 74360 + }, + { + "epoch": 6.44, + "learning_rate": 4.3600450489474143e-05, + "loss": 1.5631, + "step": 74370 + }, + { + "epoch": 6.44, + "learning_rate": 4.359958416356234e-05, + "loss": 1.4368, + "step": 74380 + }, + { + "epoch": 6.44, + "learning_rate": 4.3598717837650526e-05, + "loss": 1.4874, + "step": 74390 + }, + { + "epoch": 6.44, + "learning_rate": 4.359785151173872e-05, + "loss": 1.5387, + "step": 74400 + }, + { + "epoch": 6.44, + "learning_rate": 4.359698518582691e-05, + "loss": 1.4131, + "step": 74410 + }, + { + "epoch": 6.44, + "learning_rate": 4.35961188599151e-05, + "loss": 1.4936, + "step": 74420 + }, + { + "epoch": 6.44, + "learning_rate": 4.35952525340033e-05, + "loss": 1.5236, + "step": 74430 + }, + { + "epoch": 6.44, + "learning_rate": 4.3594386208091484e-05, + "loss": 1.4642, + "step": 74440 + }, + { + "epoch": 6.44, + "learning_rate": 4.359351988217968e-05, + "loss": 1.5007, + "step": 74450 + }, + { + "epoch": 6.45, + "learning_rate": 4.359265355626787e-05, + "loss": 1.4806, + "step": 74460 + }, + { + "epoch": 6.45, + "learning_rate": 4.359178723035606e-05, + "loss": 1.5255, + "step": 74470 + }, + { + "epoch": 6.45, + "learning_rate": 4.3590920904444255e-05, + "loss": 1.4772, + "step": 74480 + }, + { + "epoch": 6.45, + "learning_rate": 4.359005457853245e-05, + "loss": 1.4502, + "step": 74490 + }, + { + "epoch": 6.45, + "learning_rate": 4.358918825262064e-05, + "loss": 1.4727, + "step": 74500 + }, + { + "epoch": 6.45, + "learning_rate": 4.358832192670883e-05, + "loss": 1.5358, + "step": 74510 + }, + { + "epoch": 6.45, + "learning_rate": 4.358745560079702e-05, + "loss": 1.4827, + "step": 74520 + }, + { + "epoch": 6.45, + "learning_rate": 4.3586589274885214e-05, + "loss": 1.4918, + "step": 74530 + }, + { + "epoch": 6.45, + "learning_rate": 4.358572294897341e-05, + "loss": 1.5309, + "step": 74540 + }, + { + "epoch": 6.45, + "learning_rate": 4.3584856623061596e-05, + "loss": 1.477, + "step": 74550 + }, + { + "epoch": 6.45, + "learning_rate": 4.358399029714979e-05, + "loss": 1.4964, + "step": 74560 + }, + { + "epoch": 6.45, + "learning_rate": 4.3583123971237985e-05, + "loss": 1.445, + "step": 74570 + }, + { + "epoch": 6.46, + "learning_rate": 4.358225764532617e-05, + "loss": 1.516, + "step": 74580 + }, + { + "epoch": 6.46, + "learning_rate": 4.358139131941437e-05, + "loss": 1.5032, + "step": 74590 + }, + { + "epoch": 6.46, + "learning_rate": 4.358052499350256e-05, + "loss": 1.4815, + "step": 74600 + }, + { + "epoch": 6.46, + "learning_rate": 4.357965866759075e-05, + "loss": 1.4319, + "step": 74610 + }, + { + "epoch": 6.46, + "learning_rate": 4.3578792341678944e-05, + "loss": 1.462, + "step": 74620 + }, + { + "epoch": 6.46, + "learning_rate": 4.357792601576713e-05, + "loss": 1.4945, + "step": 74630 + }, + { + "epoch": 6.46, + "learning_rate": 4.3577059689855326e-05, + "loss": 1.5235, + "step": 74640 + }, + { + "epoch": 6.46, + "learning_rate": 4.357619336394352e-05, + "loss": 1.5266, + "step": 74650 + }, + { + "epoch": 6.46, + "learning_rate": 4.357532703803171e-05, + "loss": 1.4059, + "step": 74660 + }, + { + "epoch": 6.46, + "learning_rate": 4.35744607121199e-05, + "loss": 1.491, + "step": 74670 + }, + { + "epoch": 6.46, + "learning_rate": 4.35735943862081e-05, + "loss": 1.4924, + "step": 74680 + }, + { + "epoch": 6.46, + "learning_rate": 4.3572728060296285e-05, + "loss": 1.4894, + "step": 74690 + }, + { + "epoch": 6.47, + "learning_rate": 4.357186173438448e-05, + "loss": 1.5077, + "step": 74700 + }, + { + "epoch": 6.47, + "learning_rate": 4.357099540847267e-05, + "loss": 1.5996, + "step": 74710 + }, + { + "epoch": 6.47, + "learning_rate": 4.357012908256086e-05, + "loss": 1.4655, + "step": 74720 + }, + { + "epoch": 6.47, + "learning_rate": 4.3569262756649056e-05, + "loss": 1.5401, + "step": 74730 + }, + { + "epoch": 6.47, + "learning_rate": 4.356839643073724e-05, + "loss": 1.4816, + "step": 74740 + }, + { + "epoch": 6.47, + "learning_rate": 4.356753010482544e-05, + "loss": 1.431, + "step": 74750 + }, + { + "epoch": 6.47, + "learning_rate": 4.356666377891363e-05, + "loss": 1.4511, + "step": 74760 + }, + { + "epoch": 6.47, + "learning_rate": 4.356579745300182e-05, + "loss": 1.5685, + "step": 74770 + }, + { + "epoch": 6.47, + "learning_rate": 4.3564931127090014e-05, + "loss": 1.5397, + "step": 74780 + }, + { + "epoch": 6.47, + "learning_rate": 4.356406480117821e-05, + "loss": 1.5027, + "step": 74790 + }, + { + "epoch": 6.47, + "learning_rate": 4.3563198475266397e-05, + "loss": 1.499, + "step": 74800 + }, + { + "epoch": 6.48, + "learning_rate": 4.356233214935459e-05, + "loss": 1.4875, + "step": 74810 + }, + { + "epoch": 6.48, + "learning_rate": 4.356146582344278e-05, + "loss": 1.4969, + "step": 74820 + }, + { + "epoch": 6.48, + "learning_rate": 4.356059949753097e-05, + "loss": 1.4905, + "step": 74830 + }, + { + "epoch": 6.48, + "learning_rate": 4.355973317161917e-05, + "loss": 1.5576, + "step": 74840 + }, + { + "epoch": 6.48, + "learning_rate": 4.3558866845707355e-05, + "loss": 1.4696, + "step": 74850 + }, + { + "epoch": 6.48, + "learning_rate": 4.355800051979555e-05, + "loss": 1.5294, + "step": 74860 + }, + { + "epoch": 6.48, + "learning_rate": 4.3557134193883744e-05, + "loss": 1.521, + "step": 74870 + }, + { + "epoch": 6.48, + "learning_rate": 4.355626786797193e-05, + "loss": 1.4762, + "step": 74880 + }, + { + "epoch": 6.48, + "learning_rate": 4.3555401542060126e-05, + "loss": 1.4691, + "step": 74890 + }, + { + "epoch": 6.48, + "learning_rate": 4.355453521614832e-05, + "loss": 1.4811, + "step": 74900 + }, + { + "epoch": 6.48, + "learning_rate": 4.355366889023651e-05, + "loss": 1.5372, + "step": 74910 + }, + { + "epoch": 6.48, + "learning_rate": 4.35528025643247e-05, + "loss": 1.4902, + "step": 74920 + }, + { + "epoch": 6.49, + "learning_rate": 4.355193623841289e-05, + "loss": 1.4724, + "step": 74930 + }, + { + "epoch": 6.49, + "learning_rate": 4.3551069912501085e-05, + "loss": 1.6308, + "step": 74940 + }, + { + "epoch": 6.49, + "learning_rate": 4.355020358658928e-05, + "loss": 1.4488, + "step": 74950 + }, + { + "epoch": 6.49, + "learning_rate": 4.354933726067747e-05, + "loss": 1.5015, + "step": 74960 + }, + { + "epoch": 6.49, + "learning_rate": 4.354847093476566e-05, + "loss": 1.4767, + "step": 74970 + }, + { + "epoch": 6.49, + "learning_rate": 4.3547604608853856e-05, + "loss": 1.4929, + "step": 74980 + }, + { + "epoch": 6.49, + "learning_rate": 4.3546738282942044e-05, + "loss": 1.4312, + "step": 74990 + }, + { + "epoch": 6.49, + "learning_rate": 4.354587195703024e-05, + "loss": 1.4545, + "step": 75000 + }, + { + "epoch": 6.49, + "learning_rate": 4.354500563111843e-05, + "loss": 1.511, + "step": 75010 + }, + { + "epoch": 6.49, + "learning_rate": 4.354413930520662e-05, + "loss": 1.5174, + "step": 75020 + }, + { + "epoch": 6.49, + "learning_rate": 4.3543272979294815e-05, + "loss": 1.5241, + "step": 75030 + }, + { + "epoch": 6.5, + "learning_rate": 4.3542406653383e-05, + "loss": 1.5098, + "step": 75040 + }, + { + "epoch": 6.5, + "learning_rate": 4.35415403274712e-05, + "loss": 1.4732, + "step": 75050 + }, + { + "epoch": 6.5, + "learning_rate": 4.354067400155939e-05, + "loss": 1.4393, + "step": 75060 + }, + { + "epoch": 6.5, + "learning_rate": 4.353980767564758e-05, + "loss": 1.567, + "step": 75070 + }, + { + "epoch": 6.5, + "learning_rate": 4.3538941349735773e-05, + "loss": 1.4874, + "step": 75080 + }, + { + "epoch": 6.5, + "learning_rate": 4.353807502382397e-05, + "loss": 1.4634, + "step": 75090 + }, + { + "epoch": 6.5, + "learning_rate": 4.3537208697912156e-05, + "loss": 1.4384, + "step": 75100 + }, + { + "epoch": 6.5, + "learning_rate": 4.353634237200035e-05, + "loss": 1.4459, + "step": 75110 + }, + { + "epoch": 6.5, + "learning_rate": 4.3535476046088544e-05, + "loss": 1.5137, + "step": 75120 + }, + { + "epoch": 6.5, + "learning_rate": 4.353460972017673e-05, + "loss": 1.5764, + "step": 75130 + }, + { + "epoch": 6.5, + "learning_rate": 4.3533743394264927e-05, + "loss": 1.5627, + "step": 75140 + }, + { + "epoch": 6.5, + "learning_rate": 4.3532877068353114e-05, + "loss": 1.5181, + "step": 75150 + }, + { + "epoch": 6.51, + "learning_rate": 4.353201074244131e-05, + "loss": 1.4896, + "step": 75160 + }, + { + "epoch": 6.51, + "learning_rate": 4.35311444165295e-05, + "loss": 1.4956, + "step": 75170 + }, + { + "epoch": 6.51, + "learning_rate": 4.353027809061769e-05, + "loss": 1.4822, + "step": 75180 + }, + { + "epoch": 6.51, + "learning_rate": 4.3529411764705885e-05, + "loss": 1.4572, + "step": 75190 + }, + { + "epoch": 6.51, + "learning_rate": 4.352854543879408e-05, + "loss": 1.4752, + "step": 75200 + }, + { + "epoch": 6.51, + "learning_rate": 4.352767911288227e-05, + "loss": 1.4495, + "step": 75210 + }, + { + "epoch": 6.51, + "learning_rate": 4.352681278697046e-05, + "loss": 1.471, + "step": 75220 + }, + { + "epoch": 6.51, + "learning_rate": 4.3525946461058656e-05, + "loss": 1.4913, + "step": 75230 + }, + { + "epoch": 6.51, + "learning_rate": 4.3525080135146844e-05, + "loss": 1.4463, + "step": 75240 + }, + { + "epoch": 6.51, + "learning_rate": 4.352421380923504e-05, + "loss": 1.4135, + "step": 75250 + }, + { + "epoch": 6.51, + "learning_rate": 4.3523347483323226e-05, + "loss": 1.5063, + "step": 75260 + }, + { + "epoch": 6.52, + "learning_rate": 4.352248115741142e-05, + "loss": 1.5038, + "step": 75270 + }, + { + "epoch": 6.52, + "learning_rate": 4.3521614831499615e-05, + "loss": 1.4661, + "step": 75280 + }, + { + "epoch": 6.52, + "learning_rate": 4.35207485055878e-05, + "loss": 1.4767, + "step": 75290 + }, + { + "epoch": 6.52, + "learning_rate": 4.3519882179676e-05, + "loss": 1.5189, + "step": 75300 + }, + { + "epoch": 6.52, + "learning_rate": 4.351901585376419e-05, + "loss": 1.4353, + "step": 75310 + }, + { + "epoch": 6.52, + "learning_rate": 4.351814952785238e-05, + "loss": 1.4929, + "step": 75320 + }, + { + "epoch": 6.52, + "learning_rate": 4.3517283201940574e-05, + "loss": 1.4786, + "step": 75330 + }, + { + "epoch": 6.52, + "learning_rate": 4.351641687602877e-05, + "loss": 1.5036, + "step": 75340 + }, + { + "epoch": 6.52, + "learning_rate": 4.3515550550116956e-05, + "loss": 1.5298, + "step": 75350 + }, + { + "epoch": 6.52, + "learning_rate": 4.351468422420515e-05, + "loss": 1.5249, + "step": 75360 + }, + { + "epoch": 6.52, + "learning_rate": 4.351381789829334e-05, + "loss": 1.4754, + "step": 75370 + }, + { + "epoch": 6.52, + "learning_rate": 4.351295157238153e-05, + "loss": 1.5777, + "step": 75380 + }, + { + "epoch": 6.53, + "learning_rate": 4.351208524646973e-05, + "loss": 1.4805, + "step": 75390 + }, + { + "epoch": 6.53, + "learning_rate": 4.3511218920557915e-05, + "loss": 1.5346, + "step": 75400 + }, + { + "epoch": 6.53, + "learning_rate": 4.351035259464611e-05, + "loss": 1.5164, + "step": 75410 + }, + { + "epoch": 6.53, + "learning_rate": 4.3509486268734304e-05, + "loss": 1.5463, + "step": 75420 + }, + { + "epoch": 6.53, + "learning_rate": 4.350861994282249e-05, + "loss": 1.4998, + "step": 75430 + }, + { + "epoch": 6.53, + "learning_rate": 4.3507753616910686e-05, + "loss": 1.5138, + "step": 75440 + }, + { + "epoch": 6.53, + "learning_rate": 4.350688729099888e-05, + "loss": 1.5095, + "step": 75450 + }, + { + "epoch": 6.53, + "learning_rate": 4.350602096508707e-05, + "loss": 1.4597, + "step": 75460 + }, + { + "epoch": 6.53, + "learning_rate": 4.350515463917526e-05, + "loss": 1.5037, + "step": 75470 + }, + { + "epoch": 6.53, + "learning_rate": 4.350428831326345e-05, + "loss": 1.5037, + "step": 75480 + }, + { + "epoch": 6.53, + "learning_rate": 4.3503421987351644e-05, + "loss": 1.4611, + "step": 75490 + }, + { + "epoch": 6.54, + "learning_rate": 4.350255566143984e-05, + "loss": 1.5429, + "step": 75500 + }, + { + "epoch": 6.54, + "learning_rate": 4.3501689335528026e-05, + "loss": 1.5691, + "step": 75510 + }, + { + "epoch": 6.54, + "learning_rate": 4.350082300961622e-05, + "loss": 1.5021, + "step": 75520 + }, + { + "epoch": 6.54, + "learning_rate": 4.3499956683704415e-05, + "loss": 1.4486, + "step": 75530 + }, + { + "epoch": 6.54, + "learning_rate": 4.34990903577926e-05, + "loss": 1.5103, + "step": 75540 + }, + { + "epoch": 6.54, + "learning_rate": 4.34982240318808e-05, + "loss": 1.4316, + "step": 75550 + }, + { + "epoch": 6.54, + "learning_rate": 4.3497357705968985e-05, + "loss": 1.4305, + "step": 75560 + }, + { + "epoch": 6.54, + "learning_rate": 4.349649138005718e-05, + "loss": 1.5172, + "step": 75570 + }, + { + "epoch": 6.54, + "learning_rate": 4.3495625054145374e-05, + "loss": 1.516, + "step": 75580 + }, + { + "epoch": 6.54, + "learning_rate": 4.349475872823356e-05, + "loss": 1.4893, + "step": 75590 + }, + { + "epoch": 6.54, + "learning_rate": 4.3493892402321756e-05, + "loss": 1.5361, + "step": 75600 + }, + { + "epoch": 6.54, + "learning_rate": 4.349302607640995e-05, + "loss": 1.4341, + "step": 75610 + }, + { + "epoch": 6.55, + "learning_rate": 4.349215975049814e-05, + "loss": 1.6145, + "step": 75620 + }, + { + "epoch": 6.55, + "learning_rate": 4.349129342458633e-05, + "loss": 1.5251, + "step": 75630 + }, + { + "epoch": 6.55, + "learning_rate": 4.349042709867453e-05, + "loss": 1.4502, + "step": 75640 + }, + { + "epoch": 6.55, + "learning_rate": 4.3489560772762715e-05, + "loss": 1.4625, + "step": 75650 + }, + { + "epoch": 6.55, + "learning_rate": 4.348869444685091e-05, + "loss": 1.5344, + "step": 75660 + }, + { + "epoch": 6.55, + "learning_rate": 4.34878281209391e-05, + "loss": 1.599, + "step": 75670 + }, + { + "epoch": 6.55, + "learning_rate": 4.348696179502729e-05, + "loss": 1.4687, + "step": 75680 + }, + { + "epoch": 6.55, + "learning_rate": 4.3486095469115486e-05, + "loss": 1.5248, + "step": 75690 + }, + { + "epoch": 6.55, + "learning_rate": 4.3485229143203674e-05, + "loss": 1.5092, + "step": 75700 + }, + { + "epoch": 6.55, + "learning_rate": 4.348436281729187e-05, + "loss": 1.546, + "step": 75710 + }, + { + "epoch": 6.55, + "learning_rate": 4.348349649138006e-05, + "loss": 1.529, + "step": 75720 + }, + { + "epoch": 6.55, + "learning_rate": 4.348263016546825e-05, + "loss": 1.4663, + "step": 75730 + }, + { + "epoch": 6.56, + "learning_rate": 4.3481763839556445e-05, + "loss": 1.4872, + "step": 75740 + }, + { + "epoch": 6.56, + "learning_rate": 4.348089751364464e-05, + "loss": 1.4951, + "step": 75750 + }, + { + "epoch": 6.56, + "learning_rate": 4.348003118773283e-05, + "loss": 1.5068, + "step": 75760 + }, + { + "epoch": 6.56, + "learning_rate": 4.347916486182102e-05, + "loss": 1.5055, + "step": 75770 + }, + { + "epoch": 6.56, + "learning_rate": 4.347829853590921e-05, + "loss": 1.4747, + "step": 75780 + }, + { + "epoch": 6.56, + "learning_rate": 4.3477432209997403e-05, + "loss": 1.4754, + "step": 75790 + }, + { + "epoch": 6.56, + "learning_rate": 4.34765658840856e-05, + "loss": 1.4841, + "step": 75800 + }, + { + "epoch": 6.56, + "learning_rate": 4.3475699558173786e-05, + "loss": 1.5444, + "step": 75810 + }, + { + "epoch": 6.56, + "learning_rate": 4.347483323226198e-05, + "loss": 1.4257, + "step": 75820 + }, + { + "epoch": 6.56, + "learning_rate": 4.3473966906350174e-05, + "loss": 1.4683, + "step": 75830 + }, + { + "epoch": 6.56, + "learning_rate": 4.347310058043836e-05, + "loss": 1.4938, + "step": 75840 + }, + { + "epoch": 6.57, + "learning_rate": 4.3472234254526557e-05, + "loss": 1.4771, + "step": 75850 + }, + { + "epoch": 6.57, + "learning_rate": 4.347136792861475e-05, + "loss": 1.496, + "step": 75860 + }, + { + "epoch": 6.57, + "learning_rate": 4.347050160270294e-05, + "loss": 1.5097, + "step": 75870 + }, + { + "epoch": 6.57, + "learning_rate": 4.346963527679113e-05, + "loss": 1.51, + "step": 75880 + }, + { + "epoch": 6.57, + "learning_rate": 4.346876895087932e-05, + "loss": 1.5139, + "step": 75890 + }, + { + "epoch": 6.57, + "learning_rate": 4.3467902624967515e-05, + "loss": 1.5034, + "step": 75900 + }, + { + "epoch": 6.57, + "learning_rate": 4.346703629905571e-05, + "loss": 1.4446, + "step": 75910 + }, + { + "epoch": 6.57, + "learning_rate": 4.34661699731439e-05, + "loss": 1.4712, + "step": 75920 + }, + { + "epoch": 6.57, + "learning_rate": 4.346530364723209e-05, + "loss": 1.5387, + "step": 75930 + }, + { + "epoch": 6.57, + "learning_rate": 4.3464437321320286e-05, + "loss": 1.473, + "step": 75940 + }, + { + "epoch": 6.57, + "learning_rate": 4.3463570995408474e-05, + "loss": 1.4136, + "step": 75950 + }, + { + "epoch": 6.57, + "learning_rate": 4.346270466949667e-05, + "loss": 1.4433, + "step": 75960 + }, + { + "epoch": 6.58, + "learning_rate": 4.346183834358486e-05, + "loss": 1.4248, + "step": 75970 + }, + { + "epoch": 6.58, + "learning_rate": 4.346097201767305e-05, + "loss": 1.4579, + "step": 75980 + }, + { + "epoch": 6.58, + "learning_rate": 4.3460105691761245e-05, + "loss": 1.4716, + "step": 75990 + }, + { + "epoch": 6.58, + "learning_rate": 4.345923936584943e-05, + "loss": 1.5588, + "step": 76000 + }, + { + "epoch": 6.58, + "learning_rate": 4.345837303993763e-05, + "loss": 1.4872, + "step": 76010 + }, + { + "epoch": 6.58, + "learning_rate": 4.345750671402582e-05, + "loss": 1.4848, + "step": 76020 + }, + { + "epoch": 6.58, + "learning_rate": 4.345664038811401e-05, + "loss": 1.5115, + "step": 76030 + }, + { + "epoch": 6.58, + "learning_rate": 4.3455774062202204e-05, + "loss": 1.5145, + "step": 76040 + }, + { + "epoch": 6.58, + "learning_rate": 4.34549077362904e-05, + "loss": 1.4926, + "step": 76050 + }, + { + "epoch": 6.58, + "learning_rate": 4.3454041410378586e-05, + "loss": 1.5376, + "step": 76060 + }, + { + "epoch": 6.58, + "learning_rate": 4.345317508446678e-05, + "loss": 1.5296, + "step": 76070 + }, + { + "epoch": 6.59, + "learning_rate": 4.3452308758554975e-05, + "loss": 1.4757, + "step": 76080 + }, + { + "epoch": 6.59, + "learning_rate": 4.345144243264316e-05, + "loss": 1.4876, + "step": 76090 + }, + { + "epoch": 6.59, + "learning_rate": 4.345057610673136e-05, + "loss": 1.4714, + "step": 76100 + }, + { + "epoch": 6.59, + "learning_rate": 4.3449709780819545e-05, + "loss": 1.4629, + "step": 76110 + }, + { + "epoch": 6.59, + "learning_rate": 4.344884345490774e-05, + "loss": 1.4118, + "step": 76120 + }, + { + "epoch": 6.59, + "learning_rate": 4.3447977128995933e-05, + "loss": 1.4914, + "step": 76130 + }, + { + "epoch": 6.59, + "learning_rate": 4.344711080308412e-05, + "loss": 1.5161, + "step": 76140 + }, + { + "epoch": 6.59, + "learning_rate": 4.3446244477172316e-05, + "loss": 1.5273, + "step": 76150 + }, + { + "epoch": 6.59, + "learning_rate": 4.344537815126051e-05, + "loss": 1.4756, + "step": 76160 + }, + { + "epoch": 6.59, + "learning_rate": 4.34445118253487e-05, + "loss": 1.4502, + "step": 76170 + }, + { + "epoch": 6.59, + "learning_rate": 4.344364549943689e-05, + "loss": 1.4773, + "step": 76180 + }, + { + "epoch": 6.59, + "learning_rate": 4.344277917352509e-05, + "loss": 1.5477, + "step": 76190 + }, + { + "epoch": 6.6, + "learning_rate": 4.3441912847613274e-05, + "loss": 1.4996, + "step": 76200 + }, + { + "epoch": 6.6, + "learning_rate": 4.344104652170147e-05, + "loss": 1.5444, + "step": 76210 + }, + { + "epoch": 6.6, + "learning_rate": 4.3440180195789656e-05, + "loss": 1.5295, + "step": 76220 + }, + { + "epoch": 6.6, + "learning_rate": 4.343931386987785e-05, + "loss": 1.512, + "step": 76230 + }, + { + "epoch": 6.6, + "learning_rate": 4.3438447543966045e-05, + "loss": 1.4933, + "step": 76240 + }, + { + "epoch": 6.6, + "learning_rate": 4.343758121805423e-05, + "loss": 1.4767, + "step": 76250 + }, + { + "epoch": 6.6, + "learning_rate": 4.343671489214243e-05, + "loss": 1.4679, + "step": 76260 + }, + { + "epoch": 6.6, + "learning_rate": 4.343584856623062e-05, + "loss": 1.5793, + "step": 76270 + }, + { + "epoch": 6.6, + "learning_rate": 4.343498224031881e-05, + "loss": 1.5084, + "step": 76280 + }, + { + "epoch": 6.6, + "learning_rate": 4.3434115914407004e-05, + "loss": 1.5446, + "step": 76290 + }, + { + "epoch": 6.6, + "learning_rate": 4.343324958849519e-05, + "loss": 1.4781, + "step": 76300 + }, + { + "epoch": 6.61, + "learning_rate": 4.3432383262583386e-05, + "loss": 1.507, + "step": 76310 + }, + { + "epoch": 6.61, + "learning_rate": 4.343151693667158e-05, + "loss": 1.4703, + "step": 76320 + }, + { + "epoch": 6.61, + "learning_rate": 4.343065061075977e-05, + "loss": 1.4428, + "step": 76330 + }, + { + "epoch": 6.61, + "learning_rate": 4.342978428484796e-05, + "loss": 1.4503, + "step": 76340 + }, + { + "epoch": 6.61, + "learning_rate": 4.342891795893616e-05, + "loss": 1.513, + "step": 76350 + }, + { + "epoch": 6.61, + "learning_rate": 4.3428051633024345e-05, + "loss": 1.5138, + "step": 76360 + }, + { + "epoch": 6.61, + "learning_rate": 4.342718530711254e-05, + "loss": 1.515, + "step": 76370 + }, + { + "epoch": 6.61, + "learning_rate": 4.3426318981200734e-05, + "loss": 1.4943, + "step": 76380 + }, + { + "epoch": 6.61, + "learning_rate": 4.342545265528892e-05, + "loss": 1.4568, + "step": 76390 + }, + { + "epoch": 6.61, + "learning_rate": 4.3424586329377116e-05, + "loss": 1.4986, + "step": 76400 + }, + { + "epoch": 6.61, + "learning_rate": 4.3423720003465304e-05, + "loss": 1.5543, + "step": 76410 + }, + { + "epoch": 6.61, + "learning_rate": 4.34228536775535e-05, + "loss": 1.5517, + "step": 76420 + }, + { + "epoch": 6.62, + "learning_rate": 4.342198735164169e-05, + "loss": 1.4449, + "step": 76430 + }, + { + "epoch": 6.62, + "learning_rate": 4.342112102572988e-05, + "loss": 1.5699, + "step": 76440 + }, + { + "epoch": 6.62, + "learning_rate": 4.3420254699818075e-05, + "loss": 1.5373, + "step": 76450 + }, + { + "epoch": 6.62, + "learning_rate": 4.341938837390627e-05, + "loss": 1.5094, + "step": 76460 + }, + { + "epoch": 6.62, + "learning_rate": 4.341852204799446e-05, + "loss": 1.508, + "step": 76470 + }, + { + "epoch": 6.62, + "learning_rate": 4.341765572208265e-05, + "loss": 1.5438, + "step": 76480 + }, + { + "epoch": 6.62, + "learning_rate": 4.3416789396170846e-05, + "loss": 1.5011, + "step": 76490 + }, + { + "epoch": 6.62, + "learning_rate": 4.341592307025903e-05, + "loss": 1.5073, + "step": 76500 + }, + { + "epoch": 6.62, + "learning_rate": 4.341505674434723e-05, + "loss": 1.5487, + "step": 76510 + }, + { + "epoch": 6.62, + "learning_rate": 4.3414190418435416e-05, + "loss": 1.4938, + "step": 76520 + }, + { + "epoch": 6.62, + "learning_rate": 4.341332409252361e-05, + "loss": 1.498, + "step": 76530 + }, + { + "epoch": 6.63, + "learning_rate": 4.3412457766611804e-05, + "loss": 1.4287, + "step": 76540 + }, + { + "epoch": 6.63, + "learning_rate": 4.341159144069999e-05, + "loss": 1.4831, + "step": 76550 + }, + { + "epoch": 6.63, + "learning_rate": 4.3410725114788187e-05, + "loss": 1.5106, + "step": 76560 + }, + { + "epoch": 6.63, + "learning_rate": 4.340985878887638e-05, + "loss": 1.4878, + "step": 76570 + }, + { + "epoch": 6.63, + "learning_rate": 4.340899246296457e-05, + "loss": 1.4896, + "step": 76580 + }, + { + "epoch": 6.63, + "learning_rate": 4.340812613705276e-05, + "loss": 1.405, + "step": 76590 + }, + { + "epoch": 6.63, + "learning_rate": 4.340725981114096e-05, + "loss": 1.4979, + "step": 76600 + }, + { + "epoch": 6.63, + "learning_rate": 4.3406393485229145e-05, + "loss": 1.4853, + "step": 76610 + }, + { + "epoch": 6.63, + "learning_rate": 4.340552715931734e-05, + "loss": 1.4578, + "step": 76620 + }, + { + "epoch": 6.63, + "learning_rate": 4.340466083340553e-05, + "loss": 1.4965, + "step": 76630 + }, + { + "epoch": 6.63, + "learning_rate": 4.340379450749372e-05, + "loss": 1.4872, + "step": 76640 + }, + { + "epoch": 6.63, + "learning_rate": 4.3402928181581916e-05, + "loss": 1.4947, + "step": 76650 + }, + { + "epoch": 6.64, + "learning_rate": 4.3402061855670104e-05, + "loss": 1.51, + "step": 76660 + }, + { + "epoch": 6.64, + "learning_rate": 4.34011955297583e-05, + "loss": 1.4758, + "step": 76670 + }, + { + "epoch": 6.64, + "learning_rate": 4.340032920384649e-05, + "loss": 1.4695, + "step": 76680 + }, + { + "epoch": 6.64, + "learning_rate": 4.339946287793468e-05, + "loss": 1.5246, + "step": 76690 + }, + { + "epoch": 6.64, + "learning_rate": 4.3398596552022875e-05, + "loss": 1.4082, + "step": 76700 + }, + { + "epoch": 6.64, + "learning_rate": 4.339773022611107e-05, + "loss": 1.4757, + "step": 76710 + }, + { + "epoch": 6.64, + "learning_rate": 4.339686390019926e-05, + "loss": 1.5118, + "step": 76720 + }, + { + "epoch": 6.64, + "learning_rate": 4.339599757428745e-05, + "loss": 1.4657, + "step": 76730 + }, + { + "epoch": 6.64, + "learning_rate": 4.339513124837564e-05, + "loss": 1.4626, + "step": 76740 + }, + { + "epoch": 6.64, + "learning_rate": 4.3394264922463834e-05, + "loss": 1.4698, + "step": 76750 + }, + { + "epoch": 6.64, + "learning_rate": 4.339339859655203e-05, + "loss": 1.4847, + "step": 76760 + }, + { + "epoch": 6.64, + "learning_rate": 4.3392532270640216e-05, + "loss": 1.4818, + "step": 76770 + }, + { + "epoch": 6.65, + "learning_rate": 4.339166594472841e-05, + "loss": 1.5188, + "step": 76780 + }, + { + "epoch": 6.65, + "learning_rate": 4.3390799618816605e-05, + "loss": 1.4796, + "step": 76790 + }, + { + "epoch": 6.65, + "learning_rate": 4.338993329290479e-05, + "loss": 1.5734, + "step": 76800 + }, + { + "epoch": 6.65, + "learning_rate": 4.338906696699299e-05, + "loss": 1.4887, + "step": 76810 + }, + { + "epoch": 6.65, + "learning_rate": 4.338820064108118e-05, + "loss": 1.4531, + "step": 76820 + }, + { + "epoch": 6.65, + "learning_rate": 4.338733431516937e-05, + "loss": 1.5569, + "step": 76830 + }, + { + "epoch": 6.65, + "learning_rate": 4.3386467989257563e-05, + "loss": 1.468, + "step": 76840 + }, + { + "epoch": 6.65, + "learning_rate": 4.338560166334575e-05, + "loss": 1.5161, + "step": 76850 + }, + { + "epoch": 6.65, + "learning_rate": 4.3384735337433946e-05, + "loss": 1.4824, + "step": 76860 + }, + { + "epoch": 6.65, + "learning_rate": 4.338386901152214e-05, + "loss": 1.4982, + "step": 76870 + }, + { + "epoch": 6.65, + "learning_rate": 4.338300268561033e-05, + "loss": 1.4697, + "step": 76880 + }, + { + "epoch": 6.66, + "learning_rate": 4.338213635969852e-05, + "loss": 1.4792, + "step": 76890 + }, + { + "epoch": 6.66, + "learning_rate": 4.338127003378672e-05, + "loss": 1.4999, + "step": 76900 + }, + { + "epoch": 6.66, + "learning_rate": 4.3380403707874904e-05, + "loss": 1.4876, + "step": 76910 + }, + { + "epoch": 6.66, + "learning_rate": 4.33795373819631e-05, + "loss": 1.4444, + "step": 76920 + }, + { + "epoch": 6.66, + "learning_rate": 4.337867105605129e-05, + "loss": 1.3999, + "step": 76930 + }, + { + "epoch": 6.66, + "learning_rate": 4.337780473013948e-05, + "loss": 1.4929, + "step": 76940 + }, + { + "epoch": 6.66, + "learning_rate": 4.3376938404227675e-05, + "loss": 1.5, + "step": 76950 + }, + { + "epoch": 6.66, + "learning_rate": 4.337607207831586e-05, + "loss": 1.4876, + "step": 76960 + }, + { + "epoch": 6.66, + "learning_rate": 4.337520575240406e-05, + "loss": 1.5338, + "step": 76970 + }, + { + "epoch": 6.66, + "learning_rate": 4.337433942649225e-05, + "loss": 1.4928, + "step": 76980 + }, + { + "epoch": 6.66, + "learning_rate": 4.337347310058044e-05, + "loss": 1.4777, + "step": 76990 + }, + { + "epoch": 6.66, + "learning_rate": 4.3372606774668634e-05, + "loss": 1.4624, + "step": 77000 + }, + { + "epoch": 6.67, + "learning_rate": 4.337174044875683e-05, + "loss": 1.4659, + "step": 77010 + }, + { + "epoch": 6.67, + "learning_rate": 4.3370874122845016e-05, + "loss": 1.5217, + "step": 77020 + }, + { + "epoch": 6.67, + "learning_rate": 4.337000779693321e-05, + "loss": 1.5384, + "step": 77030 + }, + { + "epoch": 6.67, + "learning_rate": 4.33691414710214e-05, + "loss": 1.5322, + "step": 77040 + }, + { + "epoch": 6.67, + "learning_rate": 4.336827514510959e-05, + "loss": 1.5015, + "step": 77050 + }, + { + "epoch": 6.67, + "learning_rate": 4.336740881919779e-05, + "loss": 1.4105, + "step": 77060 + }, + { + "epoch": 6.67, + "learning_rate": 4.3366542493285975e-05, + "loss": 1.4263, + "step": 77070 + }, + { + "epoch": 6.67, + "learning_rate": 4.336567616737417e-05, + "loss": 1.4343, + "step": 77080 + }, + { + "epoch": 6.67, + "learning_rate": 4.3364809841462364e-05, + "loss": 1.4761, + "step": 77090 + }, + { + "epoch": 6.67, + "learning_rate": 4.336394351555055e-05, + "loss": 1.4368, + "step": 77100 + }, + { + "epoch": 6.67, + "learning_rate": 4.3363077189638746e-05, + "loss": 1.4601, + "step": 77110 + }, + { + "epoch": 6.68, + "learning_rate": 4.336221086372694e-05, + "loss": 1.3672, + "step": 77120 + }, + { + "epoch": 6.68, + "learning_rate": 4.336134453781513e-05, + "loss": 1.4713, + "step": 77130 + }, + { + "epoch": 6.68, + "learning_rate": 4.336047821190332e-05, + "loss": 1.4228, + "step": 77140 + }, + { + "epoch": 6.68, + "learning_rate": 4.335961188599151e-05, + "loss": 1.4385, + "step": 77150 + }, + { + "epoch": 6.68, + "learning_rate": 4.3358745560079705e-05, + "loss": 1.4506, + "step": 77160 + }, + { + "epoch": 6.68, + "learning_rate": 4.33578792341679e-05, + "loss": 1.5202, + "step": 77170 + }, + { + "epoch": 6.68, + "learning_rate": 4.335701290825609e-05, + "loss": 1.4464, + "step": 77180 + }, + { + "epoch": 6.68, + "learning_rate": 4.335614658234428e-05, + "loss": 1.4878, + "step": 77190 + }, + { + "epoch": 6.68, + "learning_rate": 4.3355280256432476e-05, + "loss": 1.481, + "step": 77200 + }, + { + "epoch": 6.68, + "learning_rate": 4.335441393052066e-05, + "loss": 1.4931, + "step": 77210 + }, + { + "epoch": 6.68, + "learning_rate": 4.335354760460886e-05, + "loss": 1.5438, + "step": 77220 + }, + { + "epoch": 6.68, + "learning_rate": 4.335268127869705e-05, + "loss": 1.4834, + "step": 77230 + }, + { + "epoch": 6.69, + "learning_rate": 4.335181495278524e-05, + "loss": 1.4881, + "step": 77240 + }, + { + "epoch": 6.69, + "learning_rate": 4.3350948626873434e-05, + "loss": 1.4211, + "step": 77250 + }, + { + "epoch": 6.69, + "learning_rate": 4.335008230096162e-05, + "loss": 1.4922, + "step": 77260 + }, + { + "epoch": 6.69, + "learning_rate": 4.3349215975049817e-05, + "loss": 1.4361, + "step": 77270 + }, + { + "epoch": 6.69, + "learning_rate": 4.334834964913801e-05, + "loss": 1.5162, + "step": 77280 + }, + { + "epoch": 6.69, + "learning_rate": 4.33474833232262e-05, + "loss": 1.5252, + "step": 77290 + }, + { + "epoch": 6.69, + "learning_rate": 4.334661699731439e-05, + "loss": 1.5328, + "step": 77300 + }, + { + "epoch": 6.69, + "learning_rate": 4.334575067140259e-05, + "loss": 1.4699, + "step": 77310 + }, + { + "epoch": 6.69, + "learning_rate": 4.3344884345490775e-05, + "loss": 1.5342, + "step": 77320 + }, + { + "epoch": 6.69, + "learning_rate": 4.334401801957897e-05, + "loss": 1.4641, + "step": 77330 + }, + { + "epoch": 6.69, + "learning_rate": 4.3343151693667164e-05, + "loss": 1.4983, + "step": 77340 + }, + { + "epoch": 6.7, + "learning_rate": 4.334228536775535e-05, + "loss": 1.4906, + "step": 77350 + }, + { + "epoch": 6.7, + "learning_rate": 4.3341419041843546e-05, + "loss": 1.5265, + "step": 77360 + }, + { + "epoch": 6.7, + "learning_rate": 4.3340552715931734e-05, + "loss": 1.5446, + "step": 77370 + }, + { + "epoch": 6.7, + "learning_rate": 4.333968639001993e-05, + "loss": 1.4927, + "step": 77380 + }, + { + "epoch": 6.7, + "learning_rate": 4.333882006410812e-05, + "loss": 1.443, + "step": 77390 + }, + { + "epoch": 6.7, + "learning_rate": 4.333795373819631e-05, + "loss": 1.485, + "step": 77400 + }, + { + "epoch": 6.7, + "learning_rate": 4.3337087412284505e-05, + "loss": 1.4501, + "step": 77410 + }, + { + "epoch": 6.7, + "learning_rate": 4.33362210863727e-05, + "loss": 1.479, + "step": 77420 + }, + { + "epoch": 6.7, + "learning_rate": 4.333535476046089e-05, + "loss": 1.4815, + "step": 77430 + }, + { + "epoch": 6.7, + "learning_rate": 4.333448843454908e-05, + "loss": 1.3892, + "step": 77440 + }, + { + "epoch": 6.7, + "learning_rate": 4.3333622108637276e-05, + "loss": 1.5012, + "step": 77450 + }, + { + "epoch": 6.7, + "learning_rate": 4.3332755782725464e-05, + "loss": 1.4657, + "step": 77460 + }, + { + "epoch": 6.71, + "learning_rate": 4.333188945681366e-05, + "loss": 1.5102, + "step": 77470 + }, + { + "epoch": 6.71, + "learning_rate": 4.3331023130901846e-05, + "loss": 1.4824, + "step": 77480 + }, + { + "epoch": 6.71, + "learning_rate": 4.333015680499004e-05, + "loss": 1.5351, + "step": 77490 + }, + { + "epoch": 6.71, + "learning_rate": 4.3329290479078235e-05, + "loss": 1.5119, + "step": 77500 + }, + { + "epoch": 6.71, + "learning_rate": 4.332842415316642e-05, + "loss": 1.5513, + "step": 77510 + }, + { + "epoch": 6.71, + "learning_rate": 4.332755782725462e-05, + "loss": 1.4626, + "step": 77520 + }, + { + "epoch": 6.71, + "learning_rate": 4.332669150134281e-05, + "loss": 1.4994, + "step": 77530 + }, + { + "epoch": 6.71, + "learning_rate": 4.3325825175431e-05, + "loss": 1.4688, + "step": 77540 + }, + { + "epoch": 6.71, + "learning_rate": 4.3324958849519193e-05, + "loss": 1.52, + "step": 77550 + }, + { + "epoch": 6.71, + "learning_rate": 4.332409252360739e-05, + "loss": 1.4607, + "step": 77560 + }, + { + "epoch": 6.71, + "learning_rate": 4.3323226197695576e-05, + "loss": 1.5088, + "step": 77570 + }, + { + "epoch": 6.72, + "learning_rate": 4.332235987178377e-05, + "loss": 1.4468, + "step": 77580 + }, + { + "epoch": 6.72, + "learning_rate": 4.332149354587196e-05, + "loss": 1.5138, + "step": 77590 + }, + { + "epoch": 6.72, + "learning_rate": 4.332062721996015e-05, + "loss": 1.5354, + "step": 77600 + }, + { + "epoch": 6.72, + "learning_rate": 4.3319760894048347e-05, + "loss": 1.4859, + "step": 77610 + }, + { + "epoch": 6.72, + "learning_rate": 4.3318894568136534e-05, + "loss": 1.5257, + "step": 77620 + }, + { + "epoch": 6.72, + "learning_rate": 4.331802824222473e-05, + "loss": 1.4459, + "step": 77630 + }, + { + "epoch": 6.72, + "learning_rate": 4.331716191631292e-05, + "loss": 1.5506, + "step": 77640 + }, + { + "epoch": 6.72, + "learning_rate": 4.331629559040111e-05, + "loss": 1.4862, + "step": 77650 + }, + { + "epoch": 6.72, + "learning_rate": 4.3315429264489305e-05, + "loss": 1.5387, + "step": 77660 + }, + { + "epoch": 6.72, + "learning_rate": 4.33145629385775e-05, + "loss": 1.4326, + "step": 77670 + }, + { + "epoch": 6.72, + "learning_rate": 4.331369661266569e-05, + "loss": 1.4649, + "step": 77680 + }, + { + "epoch": 6.72, + "learning_rate": 4.331283028675388e-05, + "loss": 1.4987, + "step": 77690 + }, + { + "epoch": 6.73, + "learning_rate": 4.331196396084207e-05, + "loss": 1.4861, + "step": 77700 + }, + { + "epoch": 6.73, + "learning_rate": 4.3311097634930264e-05, + "loss": 1.4331, + "step": 77710 + }, + { + "epoch": 6.73, + "learning_rate": 4.331023130901846e-05, + "loss": 1.4666, + "step": 77720 + }, + { + "epoch": 6.73, + "learning_rate": 4.3309364983106646e-05, + "loss": 1.492, + "step": 77730 + }, + { + "epoch": 6.73, + "learning_rate": 4.330849865719484e-05, + "loss": 1.4637, + "step": 77740 + }, + { + "epoch": 6.73, + "learning_rate": 4.3307632331283035e-05, + "loss": 1.5004, + "step": 77750 + }, + { + "epoch": 6.73, + "learning_rate": 4.330676600537122e-05, + "loss": 1.5408, + "step": 77760 + }, + { + "epoch": 6.73, + "learning_rate": 4.330589967945942e-05, + "loss": 1.4562, + "step": 77770 + }, + { + "epoch": 6.73, + "learning_rate": 4.330503335354761e-05, + "loss": 1.5458, + "step": 77780 + }, + { + "epoch": 6.73, + "learning_rate": 4.33041670276358e-05, + "loss": 1.4916, + "step": 77790 + }, + { + "epoch": 6.73, + "learning_rate": 4.3303300701723994e-05, + "loss": 1.5113, + "step": 77800 + }, + { + "epoch": 6.74, + "learning_rate": 4.330243437581218e-05, + "loss": 1.5027, + "step": 77810 + }, + { + "epoch": 6.74, + "learning_rate": 4.3301568049900376e-05, + "loss": 1.4877, + "step": 77820 + }, + { + "epoch": 6.74, + "learning_rate": 4.330070172398857e-05, + "loss": 1.4912, + "step": 77830 + }, + { + "epoch": 6.74, + "learning_rate": 4.329983539807676e-05, + "loss": 1.4896, + "step": 77840 + }, + { + "epoch": 6.74, + "learning_rate": 4.329896907216495e-05, + "loss": 1.4944, + "step": 77850 + }, + { + "epoch": 6.74, + "learning_rate": 4.329810274625315e-05, + "loss": 1.5078, + "step": 77860 + }, + { + "epoch": 6.74, + "learning_rate": 4.3297236420341335e-05, + "loss": 1.5013, + "step": 77870 + }, + { + "epoch": 6.74, + "learning_rate": 4.329637009442953e-05, + "loss": 1.4989, + "step": 77880 + }, + { + "epoch": 6.74, + "learning_rate": 4.329550376851772e-05, + "loss": 1.4198, + "step": 77890 + }, + { + "epoch": 6.74, + "learning_rate": 4.329463744260591e-05, + "loss": 1.4893, + "step": 77900 + }, + { + "epoch": 6.74, + "learning_rate": 4.3293771116694106e-05, + "loss": 1.4773, + "step": 77910 + }, + { + "epoch": 6.74, + "learning_rate": 4.329290479078229e-05, + "loss": 1.4336, + "step": 77920 + }, + { + "epoch": 6.75, + "learning_rate": 4.329203846487049e-05, + "loss": 1.412, + "step": 77930 + }, + { + "epoch": 6.75, + "learning_rate": 4.329117213895868e-05, + "loss": 1.5005, + "step": 77940 + }, + { + "epoch": 6.75, + "learning_rate": 4.329030581304687e-05, + "loss": 1.4859, + "step": 77950 + }, + { + "epoch": 6.75, + "learning_rate": 4.3289439487135064e-05, + "loss": 1.5151, + "step": 77960 + }, + { + "epoch": 6.75, + "learning_rate": 4.328857316122326e-05, + "loss": 1.4655, + "step": 77970 + }, + { + "epoch": 6.75, + "learning_rate": 4.3287706835311446e-05, + "loss": 1.4749, + "step": 77980 + }, + { + "epoch": 6.75, + "learning_rate": 4.328684050939964e-05, + "loss": 1.4886, + "step": 77990 + }, + { + "epoch": 6.75, + "learning_rate": 4.328597418348783e-05, + "loss": 1.5253, + "step": 78000 + }, + { + "epoch": 6.75, + "learning_rate": 4.328510785757602e-05, + "loss": 1.4598, + "step": 78010 + }, + { + "epoch": 6.75, + "learning_rate": 4.328424153166422e-05, + "loss": 1.4776, + "step": 78020 + }, + { + "epoch": 6.75, + "learning_rate": 4.3283375205752405e-05, + "loss": 1.4474, + "step": 78030 + }, + { + "epoch": 6.75, + "learning_rate": 4.32825088798406e-05, + "loss": 1.5221, + "step": 78040 + }, + { + "epoch": 6.76, + "learning_rate": 4.3281642553928794e-05, + "loss": 1.5189, + "step": 78050 + }, + { + "epoch": 6.76, + "learning_rate": 4.328077622801698e-05, + "loss": 1.5672, + "step": 78060 + }, + { + "epoch": 6.76, + "learning_rate": 4.3279909902105176e-05, + "loss": 1.4019, + "step": 78070 + }, + { + "epoch": 6.76, + "learning_rate": 4.327904357619337e-05, + "loss": 1.4575, + "step": 78080 + }, + { + "epoch": 6.76, + "learning_rate": 4.327817725028156e-05, + "loss": 1.5336, + "step": 78090 + }, + { + "epoch": 6.76, + "learning_rate": 4.327731092436975e-05, + "loss": 1.495, + "step": 78100 + }, + { + "epoch": 6.76, + "learning_rate": 4.327644459845794e-05, + "loss": 1.501, + "step": 78110 + }, + { + "epoch": 6.76, + "learning_rate": 4.3275578272546135e-05, + "loss": 1.4832, + "step": 78120 + }, + { + "epoch": 6.76, + "learning_rate": 4.327471194663433e-05, + "loss": 1.4703, + "step": 78130 + }, + { + "epoch": 6.76, + "learning_rate": 4.327384562072252e-05, + "loss": 1.4558, + "step": 78140 + }, + { + "epoch": 6.76, + "learning_rate": 4.327297929481071e-05, + "loss": 1.4395, + "step": 78150 + }, + { + "epoch": 6.77, + "learning_rate": 4.3272112968898906e-05, + "loss": 1.449, + "step": 78160 + }, + { + "epoch": 6.77, + "learning_rate": 4.3271246642987094e-05, + "loss": 1.4596, + "step": 78170 + }, + { + "epoch": 6.77, + "learning_rate": 4.327038031707529e-05, + "loss": 1.5021, + "step": 78180 + }, + { + "epoch": 6.77, + "learning_rate": 4.326951399116348e-05, + "loss": 1.5213, + "step": 78190 + }, + { + "epoch": 6.77, + "learning_rate": 4.326864766525167e-05, + "loss": 1.4923, + "step": 78200 + }, + { + "epoch": 6.77, + "learning_rate": 4.3267781339339865e-05, + "loss": 1.4818, + "step": 78210 + }, + { + "epoch": 6.77, + "learning_rate": 4.326691501342805e-05, + "loss": 1.5072, + "step": 78220 + }, + { + "epoch": 6.77, + "learning_rate": 4.326604868751625e-05, + "loss": 1.4785, + "step": 78230 + }, + { + "epoch": 6.77, + "learning_rate": 4.326518236160444e-05, + "loss": 1.4865, + "step": 78240 + }, + { + "epoch": 6.77, + "learning_rate": 4.326431603569263e-05, + "loss": 1.4839, + "step": 78250 + }, + { + "epoch": 6.77, + "learning_rate": 4.3263449709780823e-05, + "loss": 1.4293, + "step": 78260 + }, + { + "epoch": 6.77, + "learning_rate": 4.326258338386902e-05, + "loss": 1.4642, + "step": 78270 + }, + { + "epoch": 6.78, + "learning_rate": 4.3261717057957206e-05, + "loss": 1.4889, + "step": 78280 + }, + { + "epoch": 6.78, + "learning_rate": 4.32608507320454e-05, + "loss": 1.5282, + "step": 78290 + }, + { + "epoch": 6.78, + "learning_rate": 4.3259984406133594e-05, + "loss": 1.4636, + "step": 78300 + }, + { + "epoch": 6.78, + "learning_rate": 4.325911808022178e-05, + "loss": 1.4415, + "step": 78310 + }, + { + "epoch": 6.78, + "learning_rate": 4.3258251754309977e-05, + "loss": 1.5118, + "step": 78320 + }, + { + "epoch": 6.78, + "learning_rate": 4.3257385428398164e-05, + "loss": 1.4859, + "step": 78330 + }, + { + "epoch": 6.78, + "learning_rate": 4.325651910248636e-05, + "loss": 1.469, + "step": 78340 + }, + { + "epoch": 6.78, + "learning_rate": 4.325565277657455e-05, + "loss": 1.5662, + "step": 78350 + }, + { + "epoch": 6.78, + "learning_rate": 4.325478645066274e-05, + "loss": 1.4654, + "step": 78360 + }, + { + "epoch": 6.78, + "learning_rate": 4.3253920124750935e-05, + "loss": 1.4305, + "step": 78370 + }, + { + "epoch": 6.78, + "learning_rate": 4.325305379883913e-05, + "loss": 1.4674, + "step": 78380 + }, + { + "epoch": 6.79, + "learning_rate": 4.325218747292732e-05, + "loss": 1.4688, + "step": 78390 + }, + { + "epoch": 6.79, + "learning_rate": 4.325132114701551e-05, + "loss": 1.4547, + "step": 78400 + }, + { + "epoch": 6.79, + "learning_rate": 4.3250454821103706e-05, + "loss": 1.4327, + "step": 78410 + }, + { + "epoch": 6.79, + "learning_rate": 4.3249588495191894e-05, + "loss": 1.4632, + "step": 78420 + }, + { + "epoch": 6.79, + "learning_rate": 4.324872216928009e-05, + "loss": 1.5009, + "step": 78430 + }, + { + "epoch": 6.79, + "learning_rate": 4.3247855843368276e-05, + "loss": 1.3994, + "step": 78440 + }, + { + "epoch": 6.79, + "learning_rate": 4.324698951745647e-05, + "loss": 1.5245, + "step": 78450 + }, + { + "epoch": 6.79, + "learning_rate": 4.3246123191544665e-05, + "loss": 1.4428, + "step": 78460 + }, + { + "epoch": 6.79, + "learning_rate": 4.324525686563285e-05, + "loss": 1.5122, + "step": 78470 + }, + { + "epoch": 6.79, + "learning_rate": 4.324439053972105e-05, + "loss": 1.4913, + "step": 78480 + }, + { + "epoch": 6.79, + "learning_rate": 4.324352421380924e-05, + "loss": 1.472, + "step": 78490 + }, + { + "epoch": 6.79, + "learning_rate": 4.324265788789743e-05, + "loss": 1.5398, + "step": 78500 + }, + { + "epoch": 6.8, + "learning_rate": 4.3241791561985624e-05, + "loss": 1.485, + "step": 78510 + }, + { + "epoch": 6.8, + "learning_rate": 4.324092523607382e-05, + "loss": 1.4915, + "step": 78520 + }, + { + "epoch": 6.8, + "learning_rate": 4.3240058910162006e-05, + "loss": 1.5483, + "step": 78530 + }, + { + "epoch": 6.8, + "learning_rate": 4.32391925842502e-05, + "loss": 1.56, + "step": 78540 + }, + { + "epoch": 6.8, + "learning_rate": 4.323832625833839e-05, + "loss": 1.5015, + "step": 78550 + }, + { + "epoch": 6.8, + "learning_rate": 4.323745993242658e-05, + "loss": 1.4125, + "step": 78560 + }, + { + "epoch": 6.8, + "learning_rate": 4.323659360651478e-05, + "loss": 1.4998, + "step": 78570 + }, + { + "epoch": 6.8, + "learning_rate": 4.3235727280602965e-05, + "loss": 1.4839, + "step": 78580 + }, + { + "epoch": 6.8, + "learning_rate": 4.323486095469116e-05, + "loss": 1.4627, + "step": 78590 + }, + { + "epoch": 6.8, + "learning_rate": 4.3233994628779353e-05, + "loss": 1.4542, + "step": 78600 + }, + { + "epoch": 6.8, + "learning_rate": 4.323312830286754e-05, + "loss": 1.4897, + "step": 78610 + }, + { + "epoch": 6.81, + "learning_rate": 4.3232261976955736e-05, + "loss": 1.4954, + "step": 78620 + }, + { + "epoch": 6.81, + "learning_rate": 4.323139565104392e-05, + "loss": 1.4825, + "step": 78630 + }, + { + "epoch": 6.81, + "learning_rate": 4.323052932513212e-05, + "loss": 1.4359, + "step": 78640 + }, + { + "epoch": 6.81, + "learning_rate": 4.322966299922031e-05, + "loss": 1.4658, + "step": 78650 + }, + { + "epoch": 6.81, + "learning_rate": 4.32287966733085e-05, + "loss": 1.4662, + "step": 78660 + }, + { + "epoch": 6.81, + "learning_rate": 4.3227930347396694e-05, + "loss": 1.4481, + "step": 78670 + }, + { + "epoch": 6.81, + "learning_rate": 4.322706402148489e-05, + "loss": 1.5521, + "step": 78680 + }, + { + "epoch": 6.81, + "learning_rate": 4.3226197695573076e-05, + "loss": 1.4667, + "step": 78690 + }, + { + "epoch": 6.81, + "learning_rate": 4.322533136966127e-05, + "loss": 1.4775, + "step": 78700 + }, + { + "epoch": 6.81, + "learning_rate": 4.3224465043749465e-05, + "loss": 1.4778, + "step": 78710 + }, + { + "epoch": 6.81, + "learning_rate": 4.322359871783765e-05, + "loss": 1.4991, + "step": 78720 + }, + { + "epoch": 6.81, + "learning_rate": 4.322273239192585e-05, + "loss": 1.4617, + "step": 78730 + }, + { + "epoch": 6.82, + "learning_rate": 4.3221866066014035e-05, + "loss": 1.494, + "step": 78740 + }, + { + "epoch": 6.82, + "learning_rate": 4.322099974010223e-05, + "loss": 1.4263, + "step": 78750 + }, + { + "epoch": 6.82, + "learning_rate": 4.3220133414190424e-05, + "loss": 1.4903, + "step": 78760 + }, + { + "epoch": 6.82, + "learning_rate": 4.321926708827861e-05, + "loss": 1.4572, + "step": 78770 + }, + { + "epoch": 6.82, + "learning_rate": 4.3218400762366806e-05, + "loss": 1.4699, + "step": 78780 + }, + { + "epoch": 6.82, + "learning_rate": 4.3217534436455e-05, + "loss": 1.5352, + "step": 78790 + }, + { + "epoch": 6.82, + "learning_rate": 4.321666811054319e-05, + "loss": 1.4584, + "step": 78800 + }, + { + "epoch": 6.82, + "learning_rate": 4.321580178463138e-05, + "loss": 1.5135, + "step": 78810 + }, + { + "epoch": 6.82, + "learning_rate": 4.321493545871958e-05, + "loss": 1.5297, + "step": 78820 + }, + { + "epoch": 6.82, + "learning_rate": 4.3214069132807765e-05, + "loss": 1.425, + "step": 78830 + }, + { + "epoch": 6.82, + "learning_rate": 4.321320280689596e-05, + "loss": 1.4854, + "step": 78840 + }, + { + "epoch": 6.83, + "learning_rate": 4.321233648098415e-05, + "loss": 1.4385, + "step": 78850 + }, + { + "epoch": 6.83, + "learning_rate": 4.321147015507234e-05, + "loss": 1.4753, + "step": 78860 + }, + { + "epoch": 6.83, + "learning_rate": 4.3210603829160536e-05, + "loss": 1.4698, + "step": 78870 + }, + { + "epoch": 6.83, + "learning_rate": 4.3209737503248724e-05, + "loss": 1.4672, + "step": 78880 + }, + { + "epoch": 6.83, + "learning_rate": 4.320887117733692e-05, + "loss": 1.474, + "step": 78890 + }, + { + "epoch": 6.83, + "learning_rate": 4.320800485142511e-05, + "loss": 1.4994, + "step": 78900 + }, + { + "epoch": 6.83, + "learning_rate": 4.32071385255133e-05, + "loss": 1.4982, + "step": 78910 + }, + { + "epoch": 6.83, + "learning_rate": 4.3206272199601495e-05, + "loss": 1.4838, + "step": 78920 + }, + { + "epoch": 6.83, + "learning_rate": 4.320540587368969e-05, + "loss": 1.5384, + "step": 78930 + }, + { + "epoch": 6.83, + "learning_rate": 4.320453954777788e-05, + "loss": 1.5061, + "step": 78940 + }, + { + "epoch": 6.83, + "learning_rate": 4.320367322186607e-05, + "loss": 1.4923, + "step": 78950 + }, + { + "epoch": 6.83, + "learning_rate": 4.320280689595426e-05, + "loss": 1.4181, + "step": 78960 + }, + { + "epoch": 6.84, + "learning_rate": 4.320194057004245e-05, + "loss": 1.4676, + "step": 78970 + }, + { + "epoch": 6.84, + "learning_rate": 4.320107424413065e-05, + "loss": 1.4308, + "step": 78980 + }, + { + "epoch": 6.84, + "learning_rate": 4.3200207918218836e-05, + "loss": 1.4509, + "step": 78990 + }, + { + "epoch": 6.84, + "learning_rate": 4.319934159230703e-05, + "loss": 1.4191, + "step": 79000 + }, + { + "epoch": 6.84, + "learning_rate": 4.3198475266395224e-05, + "loss": 1.4802, + "step": 79010 + }, + { + "epoch": 6.84, + "learning_rate": 4.319760894048341e-05, + "loss": 1.4109, + "step": 79020 + }, + { + "epoch": 6.84, + "learning_rate": 4.3196742614571607e-05, + "loss": 1.4979, + "step": 79030 + }, + { + "epoch": 6.84, + "learning_rate": 4.31958762886598e-05, + "loss": 1.408, + "step": 79040 + }, + { + "epoch": 6.84, + "learning_rate": 4.319500996274799e-05, + "loss": 1.4934, + "step": 79050 + }, + { + "epoch": 6.84, + "learning_rate": 4.319414363683618e-05, + "loss": 1.5053, + "step": 79060 + }, + { + "epoch": 6.84, + "learning_rate": 4.319327731092437e-05, + "loss": 1.481, + "step": 79070 + }, + { + "epoch": 6.84, + "learning_rate": 4.3192410985012565e-05, + "loss": 1.4488, + "step": 79080 + }, + { + "epoch": 6.85, + "learning_rate": 4.319154465910076e-05, + "loss": 1.5286, + "step": 79090 + }, + { + "epoch": 6.85, + "learning_rate": 4.319067833318895e-05, + "loss": 1.4592, + "step": 79100 + }, + { + "epoch": 6.85, + "learning_rate": 4.318981200727714e-05, + "loss": 1.4845, + "step": 79110 + }, + { + "epoch": 6.85, + "learning_rate": 4.3188945681365336e-05, + "loss": 1.5334, + "step": 79120 + }, + { + "epoch": 6.85, + "learning_rate": 4.3188079355453524e-05, + "loss": 1.4579, + "step": 79130 + }, + { + "epoch": 6.85, + "learning_rate": 4.318721302954172e-05, + "loss": 1.5402, + "step": 79140 + }, + { + "epoch": 6.85, + "learning_rate": 4.318634670362991e-05, + "loss": 1.5133, + "step": 79150 + }, + { + "epoch": 6.85, + "learning_rate": 4.31854803777181e-05, + "loss": 1.5107, + "step": 79160 + }, + { + "epoch": 6.85, + "learning_rate": 4.3184614051806295e-05, + "loss": 1.4614, + "step": 79170 + }, + { + "epoch": 6.85, + "learning_rate": 4.318374772589448e-05, + "loss": 1.4703, + "step": 79180 + }, + { + "epoch": 6.85, + "learning_rate": 4.318288139998268e-05, + "loss": 1.5468, + "step": 79190 + }, + { + "epoch": 6.86, + "learning_rate": 4.318201507407087e-05, + "loss": 1.4725, + "step": 79200 + }, + { + "epoch": 6.86, + "learning_rate": 4.318114874815906e-05, + "loss": 1.4432, + "step": 79210 + }, + { + "epoch": 6.86, + "learning_rate": 4.3180282422247254e-05, + "loss": 1.447, + "step": 79220 + }, + { + "epoch": 6.86, + "learning_rate": 4.317941609633545e-05, + "loss": 1.47, + "step": 79230 + }, + { + "epoch": 6.86, + "learning_rate": 4.3178549770423636e-05, + "loss": 1.5133, + "step": 79240 + }, + { + "epoch": 6.86, + "learning_rate": 4.317768344451183e-05, + "loss": 1.4232, + "step": 79250 + }, + { + "epoch": 6.86, + "learning_rate": 4.3176817118600025e-05, + "loss": 1.4255, + "step": 79260 + }, + { + "epoch": 6.86, + "learning_rate": 4.317595079268821e-05, + "loss": 1.4398, + "step": 79270 + }, + { + "epoch": 6.86, + "learning_rate": 4.317508446677641e-05, + "loss": 1.4972, + "step": 79280 + }, + { + "epoch": 6.86, + "learning_rate": 4.3174218140864595e-05, + "loss": 1.457, + "step": 79290 + }, + { + "epoch": 6.86, + "learning_rate": 4.317335181495279e-05, + "loss": 1.4426, + "step": 79300 + }, + { + "epoch": 6.86, + "learning_rate": 4.3172485489040983e-05, + "loss": 1.4181, + "step": 79310 + }, + { + "epoch": 6.87, + "learning_rate": 4.317161916312917e-05, + "loss": 1.4546, + "step": 79320 + }, + { + "epoch": 6.87, + "learning_rate": 4.3170752837217366e-05, + "loss": 1.4344, + "step": 79330 + }, + { + "epoch": 6.87, + "learning_rate": 4.316988651130556e-05, + "loss": 1.49, + "step": 79340 + }, + { + "epoch": 6.87, + "learning_rate": 4.316902018539375e-05, + "loss": 1.5265, + "step": 79350 + }, + { + "epoch": 6.87, + "learning_rate": 4.316815385948194e-05, + "loss": 1.5083, + "step": 79360 + }, + { + "epoch": 6.87, + "learning_rate": 4.316728753357013e-05, + "loss": 1.506, + "step": 79370 + }, + { + "epoch": 6.87, + "learning_rate": 4.3166421207658324e-05, + "loss": 1.4811, + "step": 79380 + }, + { + "epoch": 6.87, + "learning_rate": 4.316555488174652e-05, + "loss": 1.4649, + "step": 79390 + }, + { + "epoch": 6.87, + "learning_rate": 4.3164688555834706e-05, + "loss": 1.5239, + "step": 79400 + }, + { + "epoch": 6.87, + "learning_rate": 4.31638222299229e-05, + "loss": 1.4383, + "step": 79410 + }, + { + "epoch": 6.87, + "learning_rate": 4.3162955904011095e-05, + "loss": 1.5939, + "step": 79420 + }, + { + "epoch": 6.88, + "learning_rate": 4.316208957809928e-05, + "loss": 1.5055, + "step": 79430 + }, + { + "epoch": 6.88, + "learning_rate": 4.316122325218748e-05, + "loss": 1.4578, + "step": 79440 + }, + { + "epoch": 6.88, + "learning_rate": 4.316035692627567e-05, + "loss": 1.5347, + "step": 79450 + }, + { + "epoch": 6.88, + "learning_rate": 4.315949060036386e-05, + "loss": 1.4285, + "step": 79460 + }, + { + "epoch": 6.88, + "learning_rate": 4.3158624274452054e-05, + "loss": 1.4797, + "step": 79470 + }, + { + "epoch": 6.88, + "learning_rate": 4.315775794854024e-05, + "loss": 1.5362, + "step": 79480 + }, + { + "epoch": 6.88, + "learning_rate": 4.3156891622628436e-05, + "loss": 1.4301, + "step": 79490 + }, + { + "epoch": 6.88, + "learning_rate": 4.315602529671663e-05, + "loss": 1.5057, + "step": 79500 + }, + { + "epoch": 6.88, + "learning_rate": 4.315515897080482e-05, + "loss": 1.435, + "step": 79510 + }, + { + "epoch": 6.88, + "learning_rate": 4.315429264489301e-05, + "loss": 1.534, + "step": 79520 + }, + { + "epoch": 6.88, + "learning_rate": 4.315342631898121e-05, + "loss": 1.4891, + "step": 79530 + }, + { + "epoch": 6.88, + "learning_rate": 4.3152559993069395e-05, + "loss": 1.4817, + "step": 79540 + }, + { + "epoch": 6.89, + "learning_rate": 4.315169366715759e-05, + "loss": 1.4439, + "step": 79550 + }, + { + "epoch": 6.89, + "learning_rate": 4.3150827341245784e-05, + "loss": 1.4478, + "step": 79560 + }, + { + "epoch": 6.89, + "learning_rate": 4.314996101533397e-05, + "loss": 1.4394, + "step": 79570 + }, + { + "epoch": 6.89, + "learning_rate": 4.3149094689422166e-05, + "loss": 1.5017, + "step": 79580 + }, + { + "epoch": 6.89, + "learning_rate": 4.3148228363510354e-05, + "loss": 1.4335, + "step": 79590 + }, + { + "epoch": 6.89, + "learning_rate": 4.314736203759855e-05, + "loss": 1.4246, + "step": 79600 + }, + { + "epoch": 6.89, + "learning_rate": 4.314649571168674e-05, + "loss": 1.4508, + "step": 79610 + }, + { + "epoch": 6.89, + "learning_rate": 4.314562938577493e-05, + "loss": 1.4736, + "step": 79620 + }, + { + "epoch": 6.89, + "learning_rate": 4.3144763059863125e-05, + "loss": 1.455, + "step": 79630 + }, + { + "epoch": 6.89, + "learning_rate": 4.314389673395132e-05, + "loss": 1.5747, + "step": 79640 + }, + { + "epoch": 6.89, + "learning_rate": 4.314303040803951e-05, + "loss": 1.4694, + "step": 79650 + }, + { + "epoch": 6.9, + "learning_rate": 4.31421640821277e-05, + "loss": 1.5412, + "step": 79660 + }, + { + "epoch": 6.9, + "learning_rate": 4.3141297756215896e-05, + "loss": 1.4708, + "step": 79670 + }, + { + "epoch": 6.9, + "learning_rate": 4.314043143030408e-05, + "loss": 1.4151, + "step": 79680 + }, + { + "epoch": 6.9, + "learning_rate": 4.313956510439228e-05, + "loss": 1.4552, + "step": 79690 + }, + { + "epoch": 6.9, + "learning_rate": 4.3138698778480465e-05, + "loss": 1.5288, + "step": 79700 + }, + { + "epoch": 6.9, + "learning_rate": 4.313783245256866e-05, + "loss": 1.4617, + "step": 79710 + }, + { + "epoch": 6.9, + "learning_rate": 4.3136966126656854e-05, + "loss": 1.4969, + "step": 79720 + }, + { + "epoch": 6.9, + "learning_rate": 4.313609980074504e-05, + "loss": 1.5428, + "step": 79730 + }, + { + "epoch": 6.9, + "learning_rate": 4.3135233474833237e-05, + "loss": 1.4505, + "step": 79740 + }, + { + "epoch": 6.9, + "learning_rate": 4.313436714892143e-05, + "loss": 1.541, + "step": 79750 + }, + { + "epoch": 6.9, + "learning_rate": 4.313350082300962e-05, + "loss": 1.4468, + "step": 79760 + }, + { + "epoch": 6.9, + "learning_rate": 4.313263449709781e-05, + "loss": 1.507, + "step": 79770 + }, + { + "epoch": 6.91, + "learning_rate": 4.313176817118601e-05, + "loss": 1.4866, + "step": 79780 + }, + { + "epoch": 6.91, + "learning_rate": 4.3130901845274195e-05, + "loss": 1.5379, + "step": 79790 + }, + { + "epoch": 6.91, + "learning_rate": 4.313003551936239e-05, + "loss": 1.4892, + "step": 79800 + }, + { + "epoch": 6.91, + "learning_rate": 4.312916919345058e-05, + "loss": 1.5332, + "step": 79810 + }, + { + "epoch": 6.91, + "learning_rate": 4.312830286753877e-05, + "loss": 1.4034, + "step": 79820 + }, + { + "epoch": 6.91, + "learning_rate": 4.3127436541626966e-05, + "loss": 1.437, + "step": 79830 + }, + { + "epoch": 6.91, + "learning_rate": 4.3126570215715154e-05, + "loss": 1.4793, + "step": 79840 + }, + { + "epoch": 6.91, + "learning_rate": 4.312570388980335e-05, + "loss": 1.4841, + "step": 79850 + }, + { + "epoch": 6.91, + "learning_rate": 4.312483756389154e-05, + "loss": 1.4373, + "step": 79860 + }, + { + "epoch": 6.91, + "learning_rate": 4.312397123797973e-05, + "loss": 1.4882, + "step": 79870 + }, + { + "epoch": 6.91, + "learning_rate": 4.3123104912067925e-05, + "loss": 1.4513, + "step": 79880 + }, + { + "epoch": 6.92, + "learning_rate": 4.312223858615612e-05, + "loss": 1.517, + "step": 79890 + }, + { + "epoch": 6.92, + "learning_rate": 4.312137226024431e-05, + "loss": 1.4924, + "step": 79900 + }, + { + "epoch": 6.92, + "learning_rate": 4.31205059343325e-05, + "loss": 1.5151, + "step": 79910 + }, + { + "epoch": 6.92, + "learning_rate": 4.311963960842069e-05, + "loss": 1.558, + "step": 79920 + }, + { + "epoch": 6.92, + "learning_rate": 4.3118773282508884e-05, + "loss": 1.4839, + "step": 79930 + }, + { + "epoch": 6.92, + "learning_rate": 4.311790695659708e-05, + "loss": 1.4553, + "step": 79940 + }, + { + "epoch": 6.92, + "learning_rate": 4.3117040630685266e-05, + "loss": 1.5313, + "step": 79950 + }, + { + "epoch": 6.92, + "learning_rate": 4.311617430477346e-05, + "loss": 1.469, + "step": 79960 + }, + { + "epoch": 6.92, + "learning_rate": 4.3115307978861655e-05, + "loss": 1.4685, + "step": 79970 + }, + { + "epoch": 6.92, + "learning_rate": 4.311444165294984e-05, + "loss": 1.4841, + "step": 79980 + }, + { + "epoch": 6.92, + "learning_rate": 4.311357532703804e-05, + "loss": 1.4494, + "step": 79990 + }, + { + "epoch": 6.92, + "learning_rate": 4.311270900112623e-05, + "loss": 1.5044, + "step": 80000 + }, + { + "epoch": 6.93, + "learning_rate": 4.311184267521442e-05, + "loss": 1.4735, + "step": 80010 + }, + { + "epoch": 6.93, + "learning_rate": 4.3110976349302613e-05, + "loss": 1.4985, + "step": 80020 + }, + { + "epoch": 6.93, + "learning_rate": 4.31101100233908e-05, + "loss": 1.4795, + "step": 80030 + }, + { + "epoch": 6.93, + "learning_rate": 4.3109243697478996e-05, + "loss": 1.4824, + "step": 80040 + }, + { + "epoch": 6.93, + "learning_rate": 4.310837737156719e-05, + "loss": 1.4768, + "step": 80050 + }, + { + "epoch": 6.93, + "learning_rate": 4.310751104565538e-05, + "loss": 1.5064, + "step": 80060 + }, + { + "epoch": 6.93, + "learning_rate": 4.310664471974357e-05, + "loss": 1.3771, + "step": 80070 + }, + { + "epoch": 6.93, + "learning_rate": 4.3105778393831767e-05, + "loss": 1.4683, + "step": 80080 + }, + { + "epoch": 6.93, + "learning_rate": 4.3104912067919954e-05, + "loss": 1.5248, + "step": 80090 + }, + { + "epoch": 6.93, + "learning_rate": 4.310404574200815e-05, + "loss": 1.4513, + "step": 80100 + }, + { + "epoch": 6.93, + "learning_rate": 4.310317941609634e-05, + "loss": 1.457, + "step": 80110 + }, + { + "epoch": 6.93, + "learning_rate": 4.310231309018453e-05, + "loss": 1.4671, + "step": 80120 + }, + { + "epoch": 6.94, + "learning_rate": 4.3101446764272725e-05, + "loss": 1.4027, + "step": 80130 + }, + { + "epoch": 6.94, + "learning_rate": 4.310058043836091e-05, + "loss": 1.4943, + "step": 80140 + }, + { + "epoch": 6.94, + "learning_rate": 4.309971411244911e-05, + "loss": 1.428, + "step": 80150 + }, + { + "epoch": 6.94, + "learning_rate": 4.30988477865373e-05, + "loss": 1.5485, + "step": 80160 + }, + { + "epoch": 6.94, + "learning_rate": 4.309798146062549e-05, + "loss": 1.4891, + "step": 80170 + }, + { + "epoch": 6.94, + "learning_rate": 4.3097115134713684e-05, + "loss": 1.5052, + "step": 80180 + }, + { + "epoch": 6.94, + "learning_rate": 4.309624880880188e-05, + "loss": 1.5222, + "step": 80190 + }, + { + "epoch": 6.94, + "learning_rate": 4.3095382482890066e-05, + "loss": 1.4872, + "step": 80200 + }, + { + "epoch": 6.94, + "learning_rate": 4.309451615697826e-05, + "loss": 1.4545, + "step": 80210 + }, + { + "epoch": 6.94, + "learning_rate": 4.309364983106645e-05, + "loss": 1.4408, + "step": 80220 + }, + { + "epoch": 6.94, + "learning_rate": 4.309278350515464e-05, + "loss": 1.5122, + "step": 80230 + }, + { + "epoch": 6.95, + "learning_rate": 4.309191717924284e-05, + "loss": 1.4345, + "step": 80240 + }, + { + "epoch": 6.95, + "learning_rate": 4.3091050853331025e-05, + "loss": 1.4061, + "step": 80250 + }, + { + "epoch": 6.95, + "learning_rate": 4.309018452741922e-05, + "loss": 1.4782, + "step": 80260 + }, + { + "epoch": 6.95, + "learning_rate": 4.3089318201507414e-05, + "loss": 1.5164, + "step": 80270 + }, + { + "epoch": 6.95, + "learning_rate": 4.30884518755956e-05, + "loss": 1.3757, + "step": 80280 + }, + { + "epoch": 6.95, + "learning_rate": 4.3087585549683796e-05, + "loss": 1.4817, + "step": 80290 + }, + { + "epoch": 6.95, + "learning_rate": 4.308671922377199e-05, + "loss": 1.5177, + "step": 80300 + }, + { + "epoch": 6.95, + "learning_rate": 4.308585289786018e-05, + "loss": 1.4253, + "step": 80310 + }, + { + "epoch": 6.95, + "learning_rate": 4.308498657194837e-05, + "loss": 1.4396, + "step": 80320 + }, + { + "epoch": 6.95, + "learning_rate": 4.308412024603656e-05, + "loss": 1.5242, + "step": 80330 + }, + { + "epoch": 6.95, + "learning_rate": 4.3083253920124755e-05, + "loss": 1.5067, + "step": 80340 + }, + { + "epoch": 6.95, + "learning_rate": 4.308238759421295e-05, + "loss": 1.4685, + "step": 80350 + }, + { + "epoch": 6.96, + "learning_rate": 4.308152126830114e-05, + "loss": 1.5064, + "step": 80360 + }, + { + "epoch": 6.96, + "learning_rate": 4.308065494238933e-05, + "loss": 1.43, + "step": 80370 + }, + { + "epoch": 6.96, + "learning_rate": 4.3079788616477526e-05, + "loss": 1.5, + "step": 80380 + }, + { + "epoch": 6.96, + "learning_rate": 4.307892229056571e-05, + "loss": 1.4666, + "step": 80390 + }, + { + "epoch": 6.96, + "learning_rate": 4.307805596465391e-05, + "loss": 1.407, + "step": 80400 + }, + { + "epoch": 6.96, + "learning_rate": 4.30771896387421e-05, + "loss": 1.5568, + "step": 80410 + }, + { + "epoch": 6.96, + "learning_rate": 4.307632331283029e-05, + "loss": 1.5175, + "step": 80420 + }, + { + "epoch": 6.96, + "learning_rate": 4.3075456986918484e-05, + "loss": 1.4412, + "step": 80430 + }, + { + "epoch": 6.96, + "learning_rate": 4.307459066100667e-05, + "loss": 1.4441, + "step": 80440 + }, + { + "epoch": 6.96, + "learning_rate": 4.3073724335094866e-05, + "loss": 1.4969, + "step": 80450 + }, + { + "epoch": 6.96, + "learning_rate": 4.307285800918306e-05, + "loss": 1.4724, + "step": 80460 + }, + { + "epoch": 6.97, + "learning_rate": 4.307199168327125e-05, + "loss": 1.4679, + "step": 80470 + }, + { + "epoch": 6.97, + "learning_rate": 4.307112535735944e-05, + "loss": 1.5363, + "step": 80480 + }, + { + "epoch": 6.97, + "learning_rate": 4.307025903144764e-05, + "loss": 1.5046, + "step": 80490 + }, + { + "epoch": 6.97, + "learning_rate": 4.3069392705535825e-05, + "loss": 1.5499, + "step": 80500 + }, + { + "epoch": 6.97, + "learning_rate": 4.306852637962402e-05, + "loss": 1.5132, + "step": 80510 + }, + { + "epoch": 6.97, + "learning_rate": 4.3067660053712214e-05, + "loss": 1.5343, + "step": 80520 + }, + { + "epoch": 6.97, + "learning_rate": 4.30667937278004e-05, + "loss": 1.4553, + "step": 80530 + }, + { + "epoch": 6.97, + "learning_rate": 4.3065927401888596e-05, + "loss": 1.486, + "step": 80540 + }, + { + "epoch": 6.97, + "learning_rate": 4.3065061075976784e-05, + "loss": 1.4612, + "step": 80550 + }, + { + "epoch": 6.97, + "learning_rate": 4.306419475006498e-05, + "loss": 1.4554, + "step": 80560 + }, + { + "epoch": 6.97, + "learning_rate": 4.306332842415317e-05, + "loss": 1.458, + "step": 80570 + }, + { + "epoch": 6.97, + "learning_rate": 4.306246209824136e-05, + "loss": 1.4316, + "step": 80580 + }, + { + "epoch": 6.98, + "learning_rate": 4.3061595772329555e-05, + "loss": 1.5207, + "step": 80590 + }, + { + "epoch": 6.98, + "learning_rate": 4.306072944641775e-05, + "loss": 1.3912, + "step": 80600 + }, + { + "epoch": 6.98, + "learning_rate": 4.305986312050594e-05, + "loss": 1.4191, + "step": 80610 + }, + { + "epoch": 6.98, + "learning_rate": 4.305899679459413e-05, + "loss": 1.4832, + "step": 80620 + }, + { + "epoch": 6.98, + "learning_rate": 4.3058130468682326e-05, + "loss": 1.5033, + "step": 80630 + }, + { + "epoch": 6.98, + "learning_rate": 4.3057264142770514e-05, + "loss": 1.4749, + "step": 80640 + }, + { + "epoch": 6.98, + "learning_rate": 4.305639781685871e-05, + "loss": 1.4888, + "step": 80650 + }, + { + "epoch": 6.98, + "learning_rate": 4.3055531490946896e-05, + "loss": 1.4556, + "step": 80660 + }, + { + "epoch": 6.98, + "learning_rate": 4.305466516503509e-05, + "loss": 1.5472, + "step": 80670 + }, + { + "epoch": 6.98, + "learning_rate": 4.3053798839123285e-05, + "loss": 1.4693, + "step": 80680 + }, + { + "epoch": 6.98, + "learning_rate": 4.305293251321147e-05, + "loss": 1.4381, + "step": 80690 + }, + { + "epoch": 6.99, + "learning_rate": 4.305206618729967e-05, + "loss": 1.4392, + "step": 80700 + }, + { + "epoch": 6.99, + "learning_rate": 4.305119986138786e-05, + "loss": 1.5178, + "step": 80710 + }, + { + "epoch": 6.99, + "learning_rate": 4.305033353547605e-05, + "loss": 1.494, + "step": 80720 + }, + { + "epoch": 6.99, + "learning_rate": 4.3049467209564243e-05, + "loss": 1.4688, + "step": 80730 + }, + { + "epoch": 6.99, + "learning_rate": 4.304860088365244e-05, + "loss": 1.4877, + "step": 80740 + }, + { + "epoch": 6.99, + "learning_rate": 4.3047734557740626e-05, + "loss": 1.4931, + "step": 80750 + }, + { + "epoch": 6.99, + "learning_rate": 4.304686823182882e-05, + "loss": 1.4343, + "step": 80760 + }, + { + "epoch": 6.99, + "learning_rate": 4.304600190591701e-05, + "loss": 1.5367, + "step": 80770 + }, + { + "epoch": 6.99, + "learning_rate": 4.30451355800052e-05, + "loss": 1.4419, + "step": 80780 + }, + { + "epoch": 6.99, + "learning_rate": 4.3044269254093397e-05, + "loss": 1.4597, + "step": 80790 + }, + { + "epoch": 6.99, + "learning_rate": 4.3043402928181584e-05, + "loss": 1.4928, + "step": 80800 + }, + { + "epoch": 6.99, + "learning_rate": 4.304253660226978e-05, + "loss": 1.468, + "step": 80810 + }, + { + "epoch": 7.0, + "learning_rate": 4.304167027635797e-05, + "loss": 1.4163, + "step": 80820 + }, + { + "epoch": 7.0, + "learning_rate": 4.304080395044616e-05, + "loss": 1.5027, + "step": 80830 + }, + { + "epoch": 7.0, + "learning_rate": 4.3039937624534355e-05, + "loss": 1.4999, + "step": 80840 + }, + { + "epoch": 7.0, + "learning_rate": 4.303907129862255e-05, + "loss": 1.4486, + "step": 80850 + }, + { + "epoch": 7.0, + "learning_rate": 4.303820497271074e-05, + "loss": 1.5094, + "step": 80860 + }, + { + "epoch": 7.0, + "learning_rate": 4.303733864679893e-05, + "loss": 1.4961, + "step": 80870 + }, + { + "epoch": 7.0, + "eval_Bleu_1": 0.03272335057859943, + "eval_Bleu_2": 2.569430921621577e-11, + "eval_Bleu_3": 2.4581065785855107e-14, + "eval_Bleu_4": 7.838960994399909e-16, + "eval_ROUGE_L": 0.07573568130231072, + "eval_cer": 0.9940909312160583, + "eval_em": 0, + "eval_f1": 0.06406934454212156, + "eval_loss": 1.3894063234329224, + "eval_runtime": 2340.0216, + "eval_samples_per_second": 2.194, + "eval_steps_per_second": 2.194, + "eval_wer": 0.9814500423109335, + "step": 80871 + }, + { + "epoch": 7.0, + "learning_rate": 4.303647232088712e-05, + "loss": 1.5219, + "step": 80880 + }, + { + "epoch": 7.0, + "learning_rate": 4.3035605994975314e-05, + "loss": 1.4231, + "step": 80890 + }, + { + "epoch": 7.0, + "learning_rate": 4.303473966906351e-05, + "loss": 1.4496, + "step": 80900 + }, + { + "epoch": 7.0, + "learning_rate": 4.3033873343151696e-05, + "loss": 1.4744, + "step": 80910 + }, + { + "epoch": 7.0, + "learning_rate": 4.303300701723989e-05, + "loss": 1.4604, + "step": 80920 + }, + { + "epoch": 7.01, + "learning_rate": 4.3032140691328085e-05, + "loss": 1.5, + "step": 80930 + }, + { + "epoch": 7.01, + "learning_rate": 4.303127436541627e-05, + "loss": 1.4907, + "step": 80940 + }, + { + "epoch": 7.01, + "learning_rate": 4.303040803950447e-05, + "loss": 1.4639, + "step": 80950 + }, + { + "epoch": 7.01, + "learning_rate": 4.3029541713592655e-05, + "loss": 1.4595, + "step": 80960 + }, + { + "epoch": 7.01, + "learning_rate": 4.302867538768085e-05, + "loss": 1.4212, + "step": 80970 + }, + { + "epoch": 7.01, + "learning_rate": 4.3027809061769044e-05, + "loss": 1.4538, + "step": 80980 + }, + { + "epoch": 7.01, + "learning_rate": 4.302694273585723e-05, + "loss": 1.454, + "step": 80990 + }, + { + "epoch": 7.01, + "learning_rate": 4.3026076409945426e-05, + "loss": 1.4174, + "step": 81000 + }, + { + "epoch": 7.01, + "learning_rate": 4.302521008403362e-05, + "loss": 1.4118, + "step": 81010 + }, + { + "epoch": 7.01, + "learning_rate": 4.302434375812181e-05, + "loss": 1.4626, + "step": 81020 + }, + { + "epoch": 7.01, + "learning_rate": 4.302347743221e-05, + "loss": 1.4702, + "step": 81030 + }, + { + "epoch": 7.01, + "learning_rate": 4.30226111062982e-05, + "loss": 1.4462, + "step": 81040 + }, + { + "epoch": 7.02, + "learning_rate": 4.3021744780386385e-05, + "loss": 1.4791, + "step": 81050 + }, + { + "epoch": 7.02, + "learning_rate": 4.302087845447458e-05, + "loss": 1.4303, + "step": 81060 + }, + { + "epoch": 7.02, + "learning_rate": 4.302001212856277e-05, + "loss": 1.4519, + "step": 81070 + }, + { + "epoch": 7.02, + "learning_rate": 4.301914580265096e-05, + "loss": 1.4557, + "step": 81080 + }, + { + "epoch": 7.02, + "learning_rate": 4.3018279476739156e-05, + "loss": 1.4667, + "step": 81090 + }, + { + "epoch": 7.02, + "learning_rate": 4.301741315082734e-05, + "loss": 1.4256, + "step": 81100 + }, + { + "epoch": 7.02, + "learning_rate": 4.301654682491554e-05, + "loss": 1.4286, + "step": 81110 + }, + { + "epoch": 7.02, + "learning_rate": 4.301568049900373e-05, + "loss": 1.5137, + "step": 81120 + }, + { + "epoch": 7.02, + "learning_rate": 4.301481417309192e-05, + "loss": 1.4949, + "step": 81130 + }, + { + "epoch": 7.02, + "learning_rate": 4.3013947847180114e-05, + "loss": 1.4087, + "step": 81140 + }, + { + "epoch": 7.02, + "learning_rate": 4.301308152126831e-05, + "loss": 1.4129, + "step": 81150 + }, + { + "epoch": 7.02, + "learning_rate": 4.3012215195356496e-05, + "loss": 1.4402, + "step": 81160 + }, + { + "epoch": 7.03, + "learning_rate": 4.301134886944469e-05, + "loss": 1.4651, + "step": 81170 + }, + { + "epoch": 7.03, + "learning_rate": 4.301048254353288e-05, + "loss": 1.419, + "step": 81180 + }, + { + "epoch": 7.03, + "learning_rate": 4.300961621762107e-05, + "loss": 1.4268, + "step": 81190 + }, + { + "epoch": 7.03, + "learning_rate": 4.300874989170927e-05, + "loss": 1.5014, + "step": 81200 + }, + { + "epoch": 7.03, + "learning_rate": 4.3007883565797455e-05, + "loss": 1.4471, + "step": 81210 + }, + { + "epoch": 7.03, + "learning_rate": 4.300701723988565e-05, + "loss": 1.4924, + "step": 81220 + }, + { + "epoch": 7.03, + "learning_rate": 4.3006150913973844e-05, + "loss": 1.5303, + "step": 81230 + }, + { + "epoch": 7.03, + "learning_rate": 4.300528458806203e-05, + "loss": 1.4352, + "step": 81240 + }, + { + "epoch": 7.03, + "learning_rate": 4.3004418262150226e-05, + "loss": 1.4365, + "step": 81250 + }, + { + "epoch": 7.03, + "learning_rate": 4.300355193623842e-05, + "loss": 1.4745, + "step": 81260 + }, + { + "epoch": 7.03, + "learning_rate": 4.300268561032661e-05, + "loss": 1.4208, + "step": 81270 + }, + { + "epoch": 7.04, + "learning_rate": 4.30018192844148e-05, + "loss": 1.3759, + "step": 81280 + }, + { + "epoch": 7.04, + "learning_rate": 4.300095295850299e-05, + "loss": 1.4695, + "step": 81290 + }, + { + "epoch": 7.04, + "learning_rate": 4.3000086632591185e-05, + "loss": 1.4863, + "step": 81300 + }, + { + "epoch": 7.04, + "learning_rate": 4.299922030667938e-05, + "loss": 1.4637, + "step": 81310 + }, + { + "epoch": 7.04, + "learning_rate": 4.299835398076757e-05, + "loss": 1.4664, + "step": 81320 + }, + { + "epoch": 7.04, + "learning_rate": 4.299748765485576e-05, + "loss": 1.4364, + "step": 81330 + }, + { + "epoch": 7.04, + "learning_rate": 4.2996621328943956e-05, + "loss": 1.503, + "step": 81340 + }, + { + "epoch": 7.04, + "learning_rate": 4.2995755003032144e-05, + "loss": 1.4565, + "step": 81350 + }, + { + "epoch": 7.04, + "learning_rate": 4.299488867712034e-05, + "loss": 1.4366, + "step": 81360 + }, + { + "epoch": 7.04, + "learning_rate": 4.299402235120853e-05, + "loss": 1.4736, + "step": 81370 + }, + { + "epoch": 7.04, + "learning_rate": 4.299315602529672e-05, + "loss": 1.4609, + "step": 81380 + }, + { + "epoch": 7.04, + "learning_rate": 4.2992289699384915e-05, + "loss": 1.4764, + "step": 81390 + }, + { + "epoch": 7.05, + "learning_rate": 4.29914233734731e-05, + "loss": 1.5457, + "step": 81400 + }, + { + "epoch": 7.05, + "learning_rate": 4.29905570475613e-05, + "loss": 1.4519, + "step": 81410 + }, + { + "epoch": 7.05, + "learning_rate": 4.298969072164949e-05, + "loss": 1.4641, + "step": 81420 + }, + { + "epoch": 7.05, + "learning_rate": 4.298882439573768e-05, + "loss": 1.445, + "step": 81430 + }, + { + "epoch": 7.05, + "learning_rate": 4.298795806982587e-05, + "loss": 1.4447, + "step": 81440 + }, + { + "epoch": 7.05, + "learning_rate": 4.298709174391407e-05, + "loss": 1.4339, + "step": 81450 + }, + { + "epoch": 7.05, + "learning_rate": 4.2986225418002256e-05, + "loss": 1.4251, + "step": 81460 + }, + { + "epoch": 7.05, + "learning_rate": 4.298535909209045e-05, + "loss": 1.5436, + "step": 81470 + }, + { + "epoch": 7.05, + "learning_rate": 4.2984492766178644e-05, + "loss": 1.5086, + "step": 81480 + }, + { + "epoch": 7.05, + "learning_rate": 4.298362644026683e-05, + "loss": 1.4229, + "step": 81490 + }, + { + "epoch": 7.05, + "learning_rate": 4.2982760114355027e-05, + "loss": 1.3874, + "step": 81500 + }, + { + "epoch": 7.06, + "learning_rate": 4.2981893788443214e-05, + "loss": 1.5413, + "step": 81510 + }, + { + "epoch": 7.06, + "learning_rate": 4.298102746253141e-05, + "loss": 1.4674, + "step": 81520 + }, + { + "epoch": 7.06, + "learning_rate": 4.29801611366196e-05, + "loss": 1.4612, + "step": 81530 + }, + { + "epoch": 7.06, + "learning_rate": 4.297929481070779e-05, + "loss": 1.4758, + "step": 81540 + }, + { + "epoch": 7.06, + "learning_rate": 4.2978428484795985e-05, + "loss": 1.4417, + "step": 81550 + }, + { + "epoch": 7.06, + "learning_rate": 4.297756215888418e-05, + "loss": 1.4873, + "step": 81560 + }, + { + "epoch": 7.06, + "learning_rate": 4.297669583297237e-05, + "loss": 1.5075, + "step": 81570 + }, + { + "epoch": 7.06, + "learning_rate": 4.297582950706056e-05, + "loss": 1.463, + "step": 81580 + }, + { + "epoch": 7.06, + "learning_rate": 4.2974963181148756e-05, + "loss": 1.4676, + "step": 81590 + }, + { + "epoch": 7.06, + "learning_rate": 4.2974096855236944e-05, + "loss": 1.4271, + "step": 81600 + }, + { + "epoch": 7.06, + "learning_rate": 4.297323052932514e-05, + "loss": 1.4511, + "step": 81610 + }, + { + "epoch": 7.06, + "learning_rate": 4.2972364203413326e-05, + "loss": 1.5005, + "step": 81620 + }, + { + "epoch": 7.07, + "learning_rate": 4.297149787750152e-05, + "loss": 1.4423, + "step": 81630 + }, + { + "epoch": 7.07, + "learning_rate": 4.2970631551589715e-05, + "loss": 1.4627, + "step": 81640 + }, + { + "epoch": 7.07, + "learning_rate": 4.29697652256779e-05, + "loss": 1.4822, + "step": 81650 + }, + { + "epoch": 7.07, + "learning_rate": 4.29688988997661e-05, + "loss": 1.4855, + "step": 81660 + }, + { + "epoch": 7.07, + "learning_rate": 4.2968032573854285e-05, + "loss": 1.4176, + "step": 81670 + }, + { + "epoch": 7.07, + "learning_rate": 4.296716624794247e-05, + "loss": 1.4017, + "step": 81680 + }, + { + "epoch": 7.07, + "learning_rate": 4.296629992203067e-05, + "loss": 1.4417, + "step": 81690 + }, + { + "epoch": 7.07, + "learning_rate": 4.296543359611886e-05, + "loss": 1.4442, + "step": 81700 + }, + { + "epoch": 7.07, + "learning_rate": 4.296456727020705e-05, + "loss": 1.3974, + "step": 81710 + }, + { + "epoch": 7.07, + "learning_rate": 4.2963700944295244e-05, + "loss": 1.4971, + "step": 81720 + }, + { + "epoch": 7.07, + "learning_rate": 4.296283461838344e-05, + "loss": 1.5105, + "step": 81730 + }, + { + "epoch": 7.08, + "learning_rate": 4.2961968292471626e-05, + "loss": 1.4286, + "step": 81740 + }, + { + "epoch": 7.08, + "learning_rate": 4.296110196655982e-05, + "loss": 1.4706, + "step": 81750 + }, + { + "epoch": 7.08, + "learning_rate": 4.296023564064801e-05, + "loss": 1.466, + "step": 81760 + }, + { + "epoch": 7.08, + "learning_rate": 4.29593693147362e-05, + "loss": 1.3431, + "step": 81770 + }, + { + "epoch": 7.08, + "learning_rate": 4.29585029888244e-05, + "loss": 1.4808, + "step": 81780 + }, + { + "epoch": 7.08, + "learning_rate": 4.2957636662912584e-05, + "loss": 1.4257, + "step": 81790 + }, + { + "epoch": 7.08, + "learning_rate": 4.295677033700078e-05, + "loss": 1.4912, + "step": 81800 + }, + { + "epoch": 7.08, + "learning_rate": 4.295590401108897e-05, + "loss": 1.4615, + "step": 81810 + }, + { + "epoch": 7.08, + "learning_rate": 4.295503768517716e-05, + "loss": 1.4637, + "step": 81820 + }, + { + "epoch": 7.08, + "learning_rate": 4.2954171359265355e-05, + "loss": 1.4454, + "step": 81830 + }, + { + "epoch": 7.08, + "learning_rate": 4.295330503335355e-05, + "loss": 1.4456, + "step": 81840 + }, + { + "epoch": 7.08, + "learning_rate": 4.295243870744174e-05, + "loss": 1.4969, + "step": 81850 + }, + { + "epoch": 7.09, + "learning_rate": 4.295157238152993e-05, + "loss": 1.4547, + "step": 81860 + }, + { + "epoch": 7.09, + "learning_rate": 4.295070605561812e-05, + "loss": 1.4923, + "step": 81870 + }, + { + "epoch": 7.09, + "learning_rate": 4.2949839729706314e-05, + "loss": 1.4367, + "step": 81880 + }, + { + "epoch": 7.09, + "learning_rate": 4.294897340379451e-05, + "loss": 1.4603, + "step": 81890 + }, + { + "epoch": 7.09, + "learning_rate": 4.2948107077882696e-05, + "loss": 1.4036, + "step": 81900 + }, + { + "epoch": 7.09, + "learning_rate": 4.294724075197089e-05, + "loss": 1.4594, + "step": 81910 + }, + { + "epoch": 7.09, + "learning_rate": 4.2946374426059085e-05, + "loss": 1.3607, + "step": 81920 + }, + { + "epoch": 7.09, + "learning_rate": 4.294550810014727e-05, + "loss": 1.3979, + "step": 81930 + }, + { + "epoch": 7.09, + "learning_rate": 4.294464177423547e-05, + "loss": 1.4902, + "step": 81940 + }, + { + "epoch": 7.09, + "learning_rate": 4.294377544832366e-05, + "loss": 1.3714, + "step": 81950 + }, + { + "epoch": 7.09, + "learning_rate": 4.294290912241185e-05, + "loss": 1.4707, + "step": 81960 + }, + { + "epoch": 7.1, + "learning_rate": 4.2942042796500044e-05, + "loss": 1.434, + "step": 81970 + }, + { + "epoch": 7.1, + "learning_rate": 4.294117647058823e-05, + "loss": 1.4857, + "step": 81980 + }, + { + "epoch": 7.1, + "learning_rate": 4.2940310144676426e-05, + "loss": 1.4514, + "step": 81990 + }, + { + "epoch": 7.1, + "learning_rate": 4.293944381876462e-05, + "loss": 1.524, + "step": 82000 + }, + { + "epoch": 7.1, + "learning_rate": 4.293857749285281e-05, + "loss": 1.4519, + "step": 82010 + }, + { + "epoch": 7.1, + "learning_rate": 4.2937711166941e-05, + "loss": 1.4462, + "step": 82020 + }, + { + "epoch": 7.1, + "learning_rate": 4.29368448410292e-05, + "loss": 1.4066, + "step": 82030 + }, + { + "epoch": 7.1, + "learning_rate": 4.2935978515117385e-05, + "loss": 1.4301, + "step": 82040 + }, + { + "epoch": 7.1, + "learning_rate": 4.293511218920558e-05, + "loss": 1.4765, + "step": 82050 + }, + { + "epoch": 7.1, + "learning_rate": 4.2934245863293774e-05, + "loss": 1.4558, + "step": 82060 + }, + { + "epoch": 7.1, + "learning_rate": 4.293337953738196e-05, + "loss": 1.4594, + "step": 82070 + }, + { + "epoch": 7.1, + "learning_rate": 4.2932513211470156e-05, + "loss": 1.425, + "step": 82080 + }, + { + "epoch": 7.11, + "learning_rate": 4.2931646885558343e-05, + "loss": 1.3797, + "step": 82090 + }, + { + "epoch": 7.11, + "learning_rate": 4.293078055964654e-05, + "loss": 1.4786, + "step": 82100 + }, + { + "epoch": 7.11, + "learning_rate": 4.292991423373473e-05, + "loss": 1.458, + "step": 82110 + }, + { + "epoch": 7.11, + "learning_rate": 4.292904790782292e-05, + "loss": 1.3989, + "step": 82120 + }, + { + "epoch": 7.11, + "learning_rate": 4.2928181581911114e-05, + "loss": 1.467, + "step": 82130 + }, + { + "epoch": 7.11, + "learning_rate": 4.292731525599931e-05, + "loss": 1.4359, + "step": 82140 + }, + { + "epoch": 7.11, + "learning_rate": 4.2926448930087497e-05, + "loss": 1.4305, + "step": 82150 + }, + { + "epoch": 7.11, + "learning_rate": 4.292558260417569e-05, + "loss": 1.4424, + "step": 82160 + }, + { + "epoch": 7.11, + "learning_rate": 4.2924716278263885e-05, + "loss": 1.4373, + "step": 82170 + }, + { + "epoch": 7.11, + "learning_rate": 4.292384995235207e-05, + "loss": 1.4359, + "step": 82180 + }, + { + "epoch": 7.11, + "learning_rate": 4.292298362644027e-05, + "loss": 1.4485, + "step": 82190 + }, + { + "epoch": 7.11, + "learning_rate": 4.2922117300528455e-05, + "loss": 1.3877, + "step": 82200 + }, + { + "epoch": 7.12, + "learning_rate": 4.292125097461665e-05, + "loss": 1.4355, + "step": 82210 + }, + { + "epoch": 7.12, + "learning_rate": 4.2920384648704844e-05, + "loss": 1.33, + "step": 82220 + }, + { + "epoch": 7.12, + "learning_rate": 4.291951832279303e-05, + "loss": 1.4347, + "step": 82230 + }, + { + "epoch": 7.12, + "learning_rate": 4.2918651996881226e-05, + "loss": 1.4656, + "step": 82240 + }, + { + "epoch": 7.12, + "learning_rate": 4.291778567096942e-05, + "loss": 1.4466, + "step": 82250 + }, + { + "epoch": 7.12, + "learning_rate": 4.291691934505761e-05, + "loss": 1.541, + "step": 82260 + }, + { + "epoch": 7.12, + "learning_rate": 4.29160530191458e-05, + "loss": 1.4236, + "step": 82270 + }, + { + "epoch": 7.12, + "learning_rate": 4.2915186693234e-05, + "loss": 1.4591, + "step": 82280 + }, + { + "epoch": 7.12, + "learning_rate": 4.2914320367322185e-05, + "loss": 1.4086, + "step": 82290 + }, + { + "epoch": 7.12, + "learning_rate": 4.291345404141038e-05, + "loss": 1.4592, + "step": 82300 + }, + { + "epoch": 7.12, + "learning_rate": 4.291258771549857e-05, + "loss": 1.4478, + "step": 82310 + }, + { + "epoch": 7.13, + "learning_rate": 4.291172138958676e-05, + "loss": 1.4842, + "step": 82320 + }, + { + "epoch": 7.13, + "learning_rate": 4.2910855063674956e-05, + "loss": 1.4459, + "step": 82330 + }, + { + "epoch": 7.13, + "learning_rate": 4.2909988737763144e-05, + "loss": 1.4587, + "step": 82340 + }, + { + "epoch": 7.13, + "learning_rate": 4.290912241185134e-05, + "loss": 1.4288, + "step": 82350 + }, + { + "epoch": 7.13, + "learning_rate": 4.290825608593953e-05, + "loss": 1.4582, + "step": 82360 + }, + { + "epoch": 7.13, + "learning_rate": 4.290738976002772e-05, + "loss": 1.4832, + "step": 82370 + }, + { + "epoch": 7.13, + "learning_rate": 4.2906523434115915e-05, + "loss": 1.4908, + "step": 82380 + }, + { + "epoch": 7.13, + "learning_rate": 4.290565710820411e-05, + "loss": 1.4679, + "step": 82390 + }, + { + "epoch": 7.13, + "learning_rate": 4.29047907822923e-05, + "loss": 1.4382, + "step": 82400 + }, + { + "epoch": 7.13, + "learning_rate": 4.290392445638049e-05, + "loss": 1.412, + "step": 82410 + }, + { + "epoch": 7.13, + "learning_rate": 4.290305813046868e-05, + "loss": 1.4361, + "step": 82420 + }, + { + "epoch": 7.13, + "learning_rate": 4.2902191804556873e-05, + "loss": 1.4487, + "step": 82430 + }, + { + "epoch": 7.14, + "learning_rate": 4.290132547864507e-05, + "loss": 1.444, + "step": 82440 + }, + { + "epoch": 7.14, + "learning_rate": 4.2900459152733256e-05, + "loss": 1.4185, + "step": 82450 + }, + { + "epoch": 7.14, + "learning_rate": 4.289959282682145e-05, + "loss": 1.4209, + "step": 82460 + }, + { + "epoch": 7.14, + "learning_rate": 4.2898726500909645e-05, + "loss": 1.4712, + "step": 82470 + }, + { + "epoch": 7.14, + "learning_rate": 4.289786017499783e-05, + "loss": 1.4636, + "step": 82480 + }, + { + "epoch": 7.14, + "learning_rate": 4.289699384908603e-05, + "loss": 1.4474, + "step": 82490 + }, + { + "epoch": 7.14, + "learning_rate": 4.2896127523174214e-05, + "loss": 1.4569, + "step": 82500 + }, + { + "epoch": 7.14, + "learning_rate": 4.289526119726241e-05, + "loss": 1.4453, + "step": 82510 + }, + { + "epoch": 7.14, + "learning_rate": 4.28943948713506e-05, + "loss": 1.3925, + "step": 82520 + }, + { + "epoch": 7.14, + "learning_rate": 4.289352854543879e-05, + "loss": 1.3954, + "step": 82530 + }, + { + "epoch": 7.14, + "learning_rate": 4.2892662219526985e-05, + "loss": 1.4858, + "step": 82540 + }, + { + "epoch": 7.15, + "learning_rate": 4.289179589361518e-05, + "loss": 1.4284, + "step": 82550 + }, + { + "epoch": 7.15, + "learning_rate": 4.289092956770337e-05, + "loss": 1.3547, + "step": 82560 + }, + { + "epoch": 7.15, + "learning_rate": 4.289006324179156e-05, + "loss": 1.4406, + "step": 82570 + }, + { + "epoch": 7.15, + "learning_rate": 4.2889196915879756e-05, + "loss": 1.4505, + "step": 82580 + }, + { + "epoch": 7.15, + "learning_rate": 4.2888330589967944e-05, + "loss": 1.493, + "step": 82590 + }, + { + "epoch": 7.15, + "learning_rate": 4.288746426405614e-05, + "loss": 1.4353, + "step": 82600 + }, + { + "epoch": 7.15, + "learning_rate": 4.2886597938144326e-05, + "loss": 1.482, + "step": 82610 + }, + { + "epoch": 7.15, + "learning_rate": 4.288573161223252e-05, + "loss": 1.476, + "step": 82620 + }, + { + "epoch": 7.15, + "learning_rate": 4.2884865286320715e-05, + "loss": 1.3797, + "step": 82630 + }, + { + "epoch": 7.15, + "learning_rate": 4.28839989604089e-05, + "loss": 1.4582, + "step": 82640 + }, + { + "epoch": 7.15, + "learning_rate": 4.28831326344971e-05, + "loss": 1.4121, + "step": 82650 + }, + { + "epoch": 7.15, + "learning_rate": 4.288226630858529e-05, + "loss": 1.5598, + "step": 82660 + }, + { + "epoch": 7.16, + "learning_rate": 4.288139998267348e-05, + "loss": 1.4133, + "step": 82670 + }, + { + "epoch": 7.16, + "learning_rate": 4.2880533656761674e-05, + "loss": 1.4775, + "step": 82680 + }, + { + "epoch": 7.16, + "learning_rate": 4.287966733084987e-05, + "loss": 1.4637, + "step": 82690 + }, + { + "epoch": 7.16, + "learning_rate": 4.2878801004938056e-05, + "loss": 1.3892, + "step": 82700 + }, + { + "epoch": 7.16, + "learning_rate": 4.287793467902625e-05, + "loss": 1.4629, + "step": 82710 + }, + { + "epoch": 7.16, + "learning_rate": 4.287706835311444e-05, + "loss": 1.4685, + "step": 82720 + }, + { + "epoch": 7.16, + "learning_rate": 4.287620202720263e-05, + "loss": 1.433, + "step": 82730 + }, + { + "epoch": 7.16, + "learning_rate": 4.287533570129083e-05, + "loss": 1.4188, + "step": 82740 + }, + { + "epoch": 7.16, + "learning_rate": 4.2874469375379015e-05, + "loss": 1.3691, + "step": 82750 + }, + { + "epoch": 7.16, + "learning_rate": 4.287360304946721e-05, + "loss": 1.4901, + "step": 82760 + }, + { + "epoch": 7.16, + "learning_rate": 4.2872736723555404e-05, + "loss": 1.4718, + "step": 82770 + }, + { + "epoch": 7.17, + "learning_rate": 4.287187039764359e-05, + "loss": 1.4397, + "step": 82780 + }, + { + "epoch": 7.17, + "learning_rate": 4.2871004071731786e-05, + "loss": 1.4862, + "step": 82790 + }, + { + "epoch": 7.17, + "learning_rate": 4.287013774581998e-05, + "loss": 1.4533, + "step": 82800 + }, + { + "epoch": 7.17, + "learning_rate": 4.286927141990817e-05, + "loss": 1.4304, + "step": 82810 + }, + { + "epoch": 7.17, + "learning_rate": 4.286840509399636e-05, + "loss": 1.4487, + "step": 82820 + }, + { + "epoch": 7.17, + "learning_rate": 4.286753876808455e-05, + "loss": 1.4, + "step": 82830 + }, + { + "epoch": 7.17, + "learning_rate": 4.2866672442172744e-05, + "loss": 1.3668, + "step": 82840 + }, + { + "epoch": 7.17, + "learning_rate": 4.286580611626094e-05, + "loss": 1.379, + "step": 82850 + }, + { + "epoch": 7.17, + "learning_rate": 4.2864939790349127e-05, + "loss": 1.4846, + "step": 82860 + }, + { + "epoch": 7.17, + "learning_rate": 4.286407346443732e-05, + "loss": 1.4033, + "step": 82870 + }, + { + "epoch": 7.17, + "learning_rate": 4.2863207138525515e-05, + "loss": 1.4833, + "step": 82880 + }, + { + "epoch": 7.17, + "learning_rate": 4.28623408126137e-05, + "loss": 1.3608, + "step": 82890 + }, + { + "epoch": 7.18, + "learning_rate": 4.28614744867019e-05, + "loss": 1.4459, + "step": 82900 + }, + { + "epoch": 7.18, + "learning_rate": 4.286060816079009e-05, + "loss": 1.4025, + "step": 82910 + }, + { + "epoch": 7.18, + "learning_rate": 4.285974183487828e-05, + "loss": 1.4361, + "step": 82920 + }, + { + "epoch": 7.18, + "learning_rate": 4.2858875508966474e-05, + "loss": 1.4769, + "step": 82930 + }, + { + "epoch": 7.18, + "learning_rate": 4.285800918305466e-05, + "loss": 1.5024, + "step": 82940 + }, + { + "epoch": 7.18, + "learning_rate": 4.2857142857142856e-05, + "loss": 1.4245, + "step": 82950 + }, + { + "epoch": 7.18, + "learning_rate": 4.285627653123105e-05, + "loss": 1.4351, + "step": 82960 + }, + { + "epoch": 7.18, + "learning_rate": 4.285541020531924e-05, + "loss": 1.403, + "step": 82970 + }, + { + "epoch": 7.18, + "learning_rate": 4.285454387940743e-05, + "loss": 1.3481, + "step": 82980 + }, + { + "epoch": 7.18, + "learning_rate": 4.285367755349563e-05, + "loss": 1.4423, + "step": 82990 + }, + { + "epoch": 7.18, + "learning_rate": 4.2852811227583815e-05, + "loss": 1.4453, + "step": 83000 + }, + { + "epoch": 7.19, + "learning_rate": 4.285194490167201e-05, + "loss": 1.4663, + "step": 83010 + }, + { + "epoch": 7.19, + "learning_rate": 4.2851078575760204e-05, + "loss": 1.3842, + "step": 83020 + }, + { + "epoch": 7.19, + "learning_rate": 4.285021224984839e-05, + "loss": 1.3979, + "step": 83030 + }, + { + "epoch": 7.19, + "learning_rate": 4.2849345923936586e-05, + "loss": 1.3928, + "step": 83040 + }, + { + "epoch": 7.19, + "learning_rate": 4.2848479598024774e-05, + "loss": 1.492, + "step": 83050 + }, + { + "epoch": 7.19, + "learning_rate": 4.284761327211297e-05, + "loss": 1.4374, + "step": 83060 + }, + { + "epoch": 7.19, + "learning_rate": 4.284674694620116e-05, + "loss": 1.4789, + "step": 83070 + }, + { + "epoch": 7.19, + "learning_rate": 4.284588062028935e-05, + "loss": 1.3951, + "step": 83080 + }, + { + "epoch": 7.19, + "learning_rate": 4.2845014294377545e-05, + "loss": 1.5298, + "step": 83090 + }, + { + "epoch": 7.19, + "learning_rate": 4.284414796846574e-05, + "loss": 1.4394, + "step": 83100 + }, + { + "epoch": 7.19, + "learning_rate": 4.284328164255393e-05, + "loss": 1.5265, + "step": 83110 + }, + { + "epoch": 7.19, + "learning_rate": 4.284241531664212e-05, + "loss": 1.4178, + "step": 83120 + }, + { + "epoch": 7.2, + "learning_rate": 4.2841548990730316e-05, + "loss": 1.4463, + "step": 83130 + }, + { + "epoch": 7.2, + "learning_rate": 4.2840682664818503e-05, + "loss": 1.3606, + "step": 83140 + }, + { + "epoch": 7.2, + "learning_rate": 4.28398163389067e-05, + "loss": 1.4787, + "step": 83150 + }, + { + "epoch": 7.2, + "learning_rate": 4.2838950012994886e-05, + "loss": 1.4122, + "step": 83160 + }, + { + "epoch": 7.2, + "learning_rate": 4.283808368708308e-05, + "loss": 1.4197, + "step": 83170 + }, + { + "epoch": 7.2, + "learning_rate": 4.2837217361171274e-05, + "loss": 1.4426, + "step": 83180 + }, + { + "epoch": 7.2, + "learning_rate": 4.283635103525946e-05, + "loss": 1.4315, + "step": 83190 + }, + { + "epoch": 7.2, + "learning_rate": 4.283548470934766e-05, + "loss": 1.4654, + "step": 83200 + }, + { + "epoch": 7.2, + "learning_rate": 4.283461838343585e-05, + "loss": 1.463, + "step": 83210 + }, + { + "epoch": 7.2, + "learning_rate": 4.283375205752404e-05, + "loss": 1.4272, + "step": 83220 + }, + { + "epoch": 7.2, + "learning_rate": 4.283288573161223e-05, + "loss": 1.4107, + "step": 83230 + }, + { + "epoch": 7.21, + "learning_rate": 4.283201940570042e-05, + "loss": 1.5056, + "step": 83240 + }, + { + "epoch": 7.21, + "learning_rate": 4.2831153079788615e-05, + "loss": 1.4292, + "step": 83250 + }, + { + "epoch": 7.21, + "learning_rate": 4.283028675387681e-05, + "loss": 1.4149, + "step": 83260 + }, + { + "epoch": 7.21, + "learning_rate": 4.2829420427965e-05, + "loss": 1.4294, + "step": 83270 + }, + { + "epoch": 7.21, + "learning_rate": 4.282855410205319e-05, + "loss": 1.4083, + "step": 83280 + }, + { + "epoch": 7.21, + "learning_rate": 4.2827687776141386e-05, + "loss": 1.4023, + "step": 83290 + }, + { + "epoch": 7.21, + "learning_rate": 4.2826821450229574e-05, + "loss": 1.4303, + "step": 83300 + }, + { + "epoch": 7.21, + "learning_rate": 4.282595512431777e-05, + "loss": 1.4045, + "step": 83310 + }, + { + "epoch": 7.21, + "learning_rate": 4.282508879840596e-05, + "loss": 1.3919, + "step": 83320 + }, + { + "epoch": 7.21, + "learning_rate": 4.282422247249415e-05, + "loss": 1.4502, + "step": 83330 + }, + { + "epoch": 7.21, + "learning_rate": 4.2823356146582345e-05, + "loss": 1.4901, + "step": 83340 + }, + { + "epoch": 7.21, + "learning_rate": 4.282248982067053e-05, + "loss": 1.4293, + "step": 83350 + }, + { + "epoch": 7.22, + "learning_rate": 4.282162349475873e-05, + "loss": 1.4446, + "step": 83360 + }, + { + "epoch": 7.22, + "learning_rate": 4.282075716884692e-05, + "loss": 1.4484, + "step": 83370 + }, + { + "epoch": 7.22, + "learning_rate": 4.281989084293511e-05, + "loss": 1.4308, + "step": 83380 + }, + { + "epoch": 7.22, + "learning_rate": 4.2819024517023304e-05, + "loss": 1.4405, + "step": 83390 + }, + { + "epoch": 7.22, + "learning_rate": 4.28181581911115e-05, + "loss": 1.4316, + "step": 83400 + }, + { + "epoch": 7.22, + "learning_rate": 4.2817291865199686e-05, + "loss": 1.4515, + "step": 83410 + }, + { + "epoch": 7.22, + "learning_rate": 4.281642553928788e-05, + "loss": 1.387, + "step": 83420 + }, + { + "epoch": 7.22, + "learning_rate": 4.2815559213376075e-05, + "loss": 1.4324, + "step": 83430 + }, + { + "epoch": 7.22, + "learning_rate": 4.281469288746426e-05, + "loss": 1.4203, + "step": 83440 + }, + { + "epoch": 7.22, + "learning_rate": 4.281382656155246e-05, + "loss": 1.3872, + "step": 83450 + }, + { + "epoch": 7.22, + "learning_rate": 4.2812960235640645e-05, + "loss": 1.3888, + "step": 83460 + }, + { + "epoch": 7.22, + "learning_rate": 4.281209390972884e-05, + "loss": 1.4494, + "step": 83470 + }, + { + "epoch": 7.23, + "learning_rate": 4.2811227583817034e-05, + "loss": 1.402, + "step": 83480 + }, + { + "epoch": 7.23, + "learning_rate": 4.281036125790522e-05, + "loss": 1.4082, + "step": 83490 + }, + { + "epoch": 7.23, + "learning_rate": 4.2809494931993416e-05, + "loss": 1.4363, + "step": 83500 + }, + { + "epoch": 7.23, + "learning_rate": 4.280862860608161e-05, + "loss": 1.5009, + "step": 83510 + }, + { + "epoch": 7.23, + "learning_rate": 4.28077622801698e-05, + "loss": 1.4744, + "step": 83520 + }, + { + "epoch": 7.23, + "learning_rate": 4.280689595425799e-05, + "loss": 1.4479, + "step": 83530 + }, + { + "epoch": 7.23, + "learning_rate": 4.280602962834619e-05, + "loss": 1.4666, + "step": 83540 + }, + { + "epoch": 7.23, + "learning_rate": 4.2805163302434374e-05, + "loss": 1.4827, + "step": 83550 + }, + { + "epoch": 7.23, + "learning_rate": 4.280429697652257e-05, + "loss": 1.3629, + "step": 83560 + }, + { + "epoch": 7.23, + "learning_rate": 4.2803430650610757e-05, + "loss": 1.4109, + "step": 83570 + }, + { + "epoch": 7.23, + "learning_rate": 4.280256432469895e-05, + "loss": 1.3995, + "step": 83580 + }, + { + "epoch": 7.24, + "learning_rate": 4.2801697998787145e-05, + "loss": 1.4846, + "step": 83590 + }, + { + "epoch": 7.24, + "learning_rate": 4.280083167287533e-05, + "loss": 1.4012, + "step": 83600 + }, + { + "epoch": 7.24, + "learning_rate": 4.279996534696353e-05, + "loss": 1.425, + "step": 83610 + }, + { + "epoch": 7.24, + "learning_rate": 4.279909902105172e-05, + "loss": 1.4785, + "step": 83620 + }, + { + "epoch": 7.24, + "learning_rate": 4.279823269513991e-05, + "loss": 1.4137, + "step": 83630 + }, + { + "epoch": 7.24, + "learning_rate": 4.2797366369228104e-05, + "loss": 1.4982, + "step": 83640 + }, + { + "epoch": 7.24, + "learning_rate": 4.27965000433163e-05, + "loss": 1.4945, + "step": 83650 + }, + { + "epoch": 7.24, + "learning_rate": 4.2795633717404486e-05, + "loss": 1.454, + "step": 83660 + }, + { + "epoch": 7.24, + "learning_rate": 4.279476739149268e-05, + "loss": 1.4266, + "step": 83670 + }, + { + "epoch": 7.24, + "learning_rate": 4.279390106558087e-05, + "loss": 1.39, + "step": 83680 + }, + { + "epoch": 7.24, + "learning_rate": 4.279303473966906e-05, + "loss": 1.4882, + "step": 83690 + }, + { + "epoch": 7.24, + "learning_rate": 4.279216841375726e-05, + "loss": 1.4533, + "step": 83700 + }, + { + "epoch": 7.25, + "learning_rate": 4.2791302087845445e-05, + "loss": 1.468, + "step": 83710 + }, + { + "epoch": 7.25, + "learning_rate": 4.279043576193364e-05, + "loss": 1.3985, + "step": 83720 + }, + { + "epoch": 7.25, + "learning_rate": 4.2789569436021834e-05, + "loss": 1.469, + "step": 83730 + }, + { + "epoch": 7.25, + "learning_rate": 4.278870311011002e-05, + "loss": 1.4521, + "step": 83740 + }, + { + "epoch": 7.25, + "learning_rate": 4.2787836784198216e-05, + "loss": 1.374, + "step": 83750 + }, + { + "epoch": 7.25, + "learning_rate": 4.278697045828641e-05, + "loss": 1.4648, + "step": 83760 + }, + { + "epoch": 7.25, + "learning_rate": 4.27861041323746e-05, + "loss": 1.4565, + "step": 83770 + }, + { + "epoch": 7.25, + "learning_rate": 4.278523780646279e-05, + "loss": 1.4408, + "step": 83780 + }, + { + "epoch": 7.25, + "learning_rate": 4.278437148055098e-05, + "loss": 1.3761, + "step": 83790 + }, + { + "epoch": 7.25, + "learning_rate": 4.2783505154639175e-05, + "loss": 1.4187, + "step": 83800 + }, + { + "epoch": 7.25, + "learning_rate": 4.278263882872737e-05, + "loss": 1.4722, + "step": 83810 + }, + { + "epoch": 7.26, + "learning_rate": 4.278177250281556e-05, + "loss": 1.4402, + "step": 83820 + }, + { + "epoch": 7.26, + "learning_rate": 4.278090617690375e-05, + "loss": 1.4411, + "step": 83830 + }, + { + "epoch": 7.26, + "learning_rate": 4.2780039850991946e-05, + "loss": 1.4255, + "step": 83840 + }, + { + "epoch": 7.26, + "learning_rate": 4.2779173525080133e-05, + "loss": 1.4079, + "step": 83850 + }, + { + "epoch": 7.26, + "learning_rate": 4.277830719916833e-05, + "loss": 1.4843, + "step": 83860 + }, + { + "epoch": 7.26, + "learning_rate": 4.277744087325652e-05, + "loss": 1.4435, + "step": 83870 + }, + { + "epoch": 7.26, + "learning_rate": 4.277657454734471e-05, + "loss": 1.4812, + "step": 83880 + }, + { + "epoch": 7.26, + "learning_rate": 4.2775708221432904e-05, + "loss": 1.455, + "step": 83890 + }, + { + "epoch": 7.26, + "learning_rate": 4.277484189552109e-05, + "loss": 1.4423, + "step": 83900 + }, + { + "epoch": 7.26, + "learning_rate": 4.2773975569609287e-05, + "loss": 1.4682, + "step": 83910 + }, + { + "epoch": 7.26, + "learning_rate": 4.277310924369748e-05, + "loss": 1.4556, + "step": 83920 + }, + { + "epoch": 7.26, + "learning_rate": 4.277224291778567e-05, + "loss": 1.4752, + "step": 83930 + }, + { + "epoch": 7.27, + "learning_rate": 4.277137659187386e-05, + "loss": 1.443, + "step": 83940 + }, + { + "epoch": 7.27, + "learning_rate": 4.277051026596206e-05, + "loss": 1.4521, + "step": 83950 + }, + { + "epoch": 7.27, + "learning_rate": 4.2769643940050245e-05, + "loss": 1.4405, + "step": 83960 + }, + { + "epoch": 7.27, + "learning_rate": 4.276877761413844e-05, + "loss": 1.4072, + "step": 83970 + }, + { + "epoch": 7.27, + "learning_rate": 4.2767911288226634e-05, + "loss": 1.4267, + "step": 83980 + }, + { + "epoch": 7.27, + "learning_rate": 4.276704496231482e-05, + "loss": 1.4803, + "step": 83990 + }, + { + "epoch": 7.27, + "learning_rate": 4.2766178636403016e-05, + "loss": 1.4427, + "step": 84000 + }, + { + "epoch": 7.27, + "learning_rate": 4.2765312310491204e-05, + "loss": 1.4136, + "step": 84010 + }, + { + "epoch": 7.27, + "learning_rate": 4.27644459845794e-05, + "loss": 1.3922, + "step": 84020 + }, + { + "epoch": 7.27, + "learning_rate": 4.276357965866759e-05, + "loss": 1.3678, + "step": 84030 + }, + { + "epoch": 7.27, + "learning_rate": 4.276271333275578e-05, + "loss": 1.4162, + "step": 84040 + }, + { + "epoch": 7.28, + "learning_rate": 4.2761847006843975e-05, + "loss": 1.4539, + "step": 84050 + }, + { + "epoch": 7.28, + "learning_rate": 4.276098068093217e-05, + "loss": 1.4712, + "step": 84060 + }, + { + "epoch": 7.28, + "learning_rate": 4.276011435502036e-05, + "loss": 1.4261, + "step": 84070 + }, + { + "epoch": 7.28, + "learning_rate": 4.275924802910855e-05, + "loss": 1.4486, + "step": 84080 + }, + { + "epoch": 7.28, + "learning_rate": 4.275838170319674e-05, + "loss": 1.4576, + "step": 84090 + }, + { + "epoch": 7.28, + "learning_rate": 4.2757515377284934e-05, + "loss": 1.3611, + "step": 84100 + }, + { + "epoch": 7.28, + "learning_rate": 4.275664905137313e-05, + "loss": 1.4429, + "step": 84110 + }, + { + "epoch": 7.28, + "learning_rate": 4.2755782725461316e-05, + "loss": 1.425, + "step": 84120 + }, + { + "epoch": 7.28, + "learning_rate": 4.275491639954951e-05, + "loss": 1.3768, + "step": 84130 + }, + { + "epoch": 7.28, + "learning_rate": 4.2754050073637705e-05, + "loss": 1.4388, + "step": 84140 + }, + { + "epoch": 7.28, + "learning_rate": 4.275318374772589e-05, + "loss": 1.4578, + "step": 84150 + }, + { + "epoch": 7.28, + "learning_rate": 4.275231742181409e-05, + "loss": 1.3767, + "step": 84160 + }, + { + "epoch": 7.29, + "learning_rate": 4.275145109590228e-05, + "loss": 1.4577, + "step": 84170 + }, + { + "epoch": 7.29, + "learning_rate": 4.275058476999047e-05, + "loss": 1.5067, + "step": 84180 + }, + { + "epoch": 7.29, + "learning_rate": 4.2749718444078664e-05, + "loss": 1.4239, + "step": 84190 + }, + { + "epoch": 7.29, + "learning_rate": 4.274885211816685e-05, + "loss": 1.3891, + "step": 84200 + }, + { + "epoch": 7.29, + "learning_rate": 4.2747985792255046e-05, + "loss": 1.4432, + "step": 84210 + }, + { + "epoch": 7.29, + "learning_rate": 4.274711946634324e-05, + "loss": 1.4629, + "step": 84220 + }, + { + "epoch": 7.29, + "learning_rate": 4.274625314043143e-05, + "loss": 1.3538, + "step": 84230 + }, + { + "epoch": 7.29, + "learning_rate": 4.274538681451962e-05, + "loss": 1.4348, + "step": 84240 + }, + { + "epoch": 7.29, + "learning_rate": 4.274452048860782e-05, + "loss": 1.4258, + "step": 84250 + }, + { + "epoch": 7.29, + "learning_rate": 4.2743654162696004e-05, + "loss": 1.4417, + "step": 84260 + }, + { + "epoch": 7.29, + "learning_rate": 4.27427878367842e-05, + "loss": 1.4851, + "step": 84270 + }, + { + "epoch": 7.3, + "learning_rate": 4.274192151087239e-05, + "loss": 1.4628, + "step": 84280 + }, + { + "epoch": 7.3, + "learning_rate": 4.274105518496058e-05, + "loss": 1.3953, + "step": 84290 + }, + { + "epoch": 7.3, + "learning_rate": 4.2740188859048775e-05, + "loss": 1.4471, + "step": 84300 + }, + { + "epoch": 7.3, + "learning_rate": 4.273932253313696e-05, + "loss": 1.4651, + "step": 84310 + }, + { + "epoch": 7.3, + "learning_rate": 4.273845620722516e-05, + "loss": 1.4071, + "step": 84320 + }, + { + "epoch": 7.3, + "learning_rate": 4.273758988131335e-05, + "loss": 1.428, + "step": 84330 + }, + { + "epoch": 7.3, + "learning_rate": 4.273672355540154e-05, + "loss": 1.4521, + "step": 84340 + }, + { + "epoch": 7.3, + "learning_rate": 4.2735857229489734e-05, + "loss": 1.4322, + "step": 84350 + }, + { + "epoch": 7.3, + "learning_rate": 4.273499090357793e-05, + "loss": 1.4269, + "step": 84360 + }, + { + "epoch": 7.3, + "learning_rate": 4.2734124577666116e-05, + "loss": 1.4195, + "step": 84370 + }, + { + "epoch": 7.3, + "learning_rate": 4.273325825175431e-05, + "loss": 1.4429, + "step": 84380 + }, + { + "epoch": 7.3, + "learning_rate": 4.2732391925842505e-05, + "loss": 1.4795, + "step": 84390 + }, + { + "epoch": 7.31, + "learning_rate": 4.273152559993069e-05, + "loss": 1.4467, + "step": 84400 + }, + { + "epoch": 7.31, + "learning_rate": 4.273065927401889e-05, + "loss": 1.4805, + "step": 84410 + }, + { + "epoch": 7.31, + "learning_rate": 4.2729792948107075e-05, + "loss": 1.4494, + "step": 84420 + }, + { + "epoch": 7.31, + "learning_rate": 4.272892662219527e-05, + "loss": 1.501, + "step": 84430 + }, + { + "epoch": 7.31, + "learning_rate": 4.2728060296283464e-05, + "loss": 1.4194, + "step": 84440 + }, + { + "epoch": 7.31, + "learning_rate": 4.272719397037165e-05, + "loss": 1.475, + "step": 84450 + }, + { + "epoch": 7.31, + "learning_rate": 4.2726327644459846e-05, + "loss": 1.4287, + "step": 84460 + }, + { + "epoch": 7.31, + "learning_rate": 4.272546131854804e-05, + "loss": 1.4884, + "step": 84470 + }, + { + "epoch": 7.31, + "learning_rate": 4.272459499263623e-05, + "loss": 1.4495, + "step": 84480 + }, + { + "epoch": 7.31, + "learning_rate": 4.272372866672442e-05, + "loss": 1.4351, + "step": 84490 + }, + { + "epoch": 7.31, + "learning_rate": 4.272286234081262e-05, + "loss": 1.4452, + "step": 84500 + }, + { + "epoch": 7.31, + "learning_rate": 4.2721996014900805e-05, + "loss": 1.4985, + "step": 84510 + }, + { + "epoch": 7.32, + "learning_rate": 4.2721129688989e-05, + "loss": 1.3935, + "step": 84520 + }, + { + "epoch": 7.32, + "learning_rate": 4.272026336307719e-05, + "loss": 1.409, + "step": 84530 + }, + { + "epoch": 7.32, + "learning_rate": 4.271939703716538e-05, + "loss": 1.4169, + "step": 84540 + }, + { + "epoch": 7.32, + "learning_rate": 4.2718530711253576e-05, + "loss": 1.455, + "step": 84550 + }, + { + "epoch": 7.32, + "learning_rate": 4.2717664385341763e-05, + "loss": 1.4248, + "step": 84560 + }, + { + "epoch": 7.32, + "learning_rate": 4.271679805942996e-05, + "loss": 1.3814, + "step": 84570 + }, + { + "epoch": 7.32, + "learning_rate": 4.271593173351815e-05, + "loss": 1.4271, + "step": 84580 + }, + { + "epoch": 7.32, + "learning_rate": 4.271506540760634e-05, + "loss": 1.4598, + "step": 84590 + }, + { + "epoch": 7.32, + "learning_rate": 4.2714199081694534e-05, + "loss": 1.3909, + "step": 84600 + }, + { + "epoch": 7.32, + "learning_rate": 4.271333275578273e-05, + "loss": 1.4768, + "step": 84610 + }, + { + "epoch": 7.32, + "learning_rate": 4.2712466429870917e-05, + "loss": 1.386, + "step": 84620 + }, + { + "epoch": 7.33, + "learning_rate": 4.271160010395911e-05, + "loss": 1.4069, + "step": 84630 + }, + { + "epoch": 7.33, + "learning_rate": 4.27107337780473e-05, + "loss": 1.4331, + "step": 84640 + }, + { + "epoch": 7.33, + "learning_rate": 4.270986745213549e-05, + "loss": 1.4178, + "step": 84650 + }, + { + "epoch": 7.33, + "learning_rate": 4.270900112622369e-05, + "loss": 1.46, + "step": 84660 + }, + { + "epoch": 7.33, + "learning_rate": 4.2708134800311875e-05, + "loss": 1.3809, + "step": 84670 + }, + { + "epoch": 7.33, + "learning_rate": 4.270726847440007e-05, + "loss": 1.4651, + "step": 84680 + }, + { + "epoch": 7.33, + "learning_rate": 4.2706402148488264e-05, + "loss": 1.3919, + "step": 84690 + }, + { + "epoch": 7.33, + "learning_rate": 4.270553582257645e-05, + "loss": 1.3746, + "step": 84700 + }, + { + "epoch": 7.33, + "learning_rate": 4.2704669496664646e-05, + "loss": 1.4164, + "step": 84710 + }, + { + "epoch": 7.33, + "learning_rate": 4.270380317075284e-05, + "loss": 1.4601, + "step": 84720 + }, + { + "epoch": 7.33, + "learning_rate": 4.270293684484103e-05, + "loss": 1.4662, + "step": 84730 + }, + { + "epoch": 7.33, + "learning_rate": 4.270207051892922e-05, + "loss": 1.4632, + "step": 84740 + }, + { + "epoch": 7.34, + "learning_rate": 4.270120419301741e-05, + "loss": 1.4316, + "step": 84750 + }, + { + "epoch": 7.34, + "learning_rate": 4.2700337867105605e-05, + "loss": 1.4291, + "step": 84760 + }, + { + "epoch": 7.34, + "learning_rate": 4.26994715411938e-05, + "loss": 1.4308, + "step": 84770 + }, + { + "epoch": 7.34, + "learning_rate": 4.269860521528199e-05, + "loss": 1.4048, + "step": 84780 + }, + { + "epoch": 7.34, + "learning_rate": 4.269773888937018e-05, + "loss": 1.438, + "step": 84790 + }, + { + "epoch": 7.34, + "learning_rate": 4.2696872563458376e-05, + "loss": 1.4643, + "step": 84800 + }, + { + "epoch": 7.34, + "learning_rate": 4.2696006237546564e-05, + "loss": 1.4158, + "step": 84810 + }, + { + "epoch": 7.34, + "learning_rate": 4.269513991163476e-05, + "loss": 1.43, + "step": 84820 + }, + { + "epoch": 7.34, + "learning_rate": 4.2694273585722946e-05, + "loss": 1.3664, + "step": 84830 + }, + { + "epoch": 7.34, + "learning_rate": 4.269340725981114e-05, + "loss": 1.4682, + "step": 84840 + }, + { + "epoch": 7.34, + "learning_rate": 4.2692540933899335e-05, + "loss": 1.4453, + "step": 84850 + }, + { + "epoch": 7.35, + "learning_rate": 4.269167460798752e-05, + "loss": 1.3884, + "step": 84860 + }, + { + "epoch": 7.35, + "learning_rate": 4.269080828207572e-05, + "loss": 1.3656, + "step": 84870 + }, + { + "epoch": 7.35, + "learning_rate": 4.268994195616391e-05, + "loss": 1.4897, + "step": 84880 + }, + { + "epoch": 7.35, + "learning_rate": 4.26890756302521e-05, + "loss": 1.4109, + "step": 84890 + }, + { + "epoch": 7.35, + "learning_rate": 4.2688209304340293e-05, + "loss": 1.4101, + "step": 84900 + }, + { + "epoch": 7.35, + "learning_rate": 4.268734297842849e-05, + "loss": 1.4539, + "step": 84910 + }, + { + "epoch": 7.35, + "learning_rate": 4.2686476652516676e-05, + "loss": 1.4376, + "step": 84920 + }, + { + "epoch": 7.35, + "learning_rate": 4.268561032660487e-05, + "loss": 1.4439, + "step": 84930 + }, + { + "epoch": 7.35, + "learning_rate": 4.268474400069306e-05, + "loss": 1.4715, + "step": 84940 + }, + { + "epoch": 7.35, + "learning_rate": 4.268387767478125e-05, + "loss": 1.4532, + "step": 84950 + }, + { + "epoch": 7.35, + "learning_rate": 4.268301134886945e-05, + "loss": 1.4517, + "step": 84960 + }, + { + "epoch": 7.35, + "learning_rate": 4.2682145022957634e-05, + "loss": 1.443, + "step": 84970 + }, + { + "epoch": 7.36, + "learning_rate": 4.268127869704583e-05, + "loss": 1.4295, + "step": 84980 + }, + { + "epoch": 7.36, + "learning_rate": 4.268041237113402e-05, + "loss": 1.4451, + "step": 84990 + }, + { + "epoch": 7.36, + "learning_rate": 4.267954604522221e-05, + "loss": 1.4256, + "step": 85000 + }, + { + "epoch": 7.36, + "learning_rate": 4.2678679719310405e-05, + "loss": 1.4975, + "step": 85010 + }, + { + "epoch": 7.36, + "learning_rate": 4.26778133933986e-05, + "loss": 1.4742, + "step": 85020 + }, + { + "epoch": 7.36, + "learning_rate": 4.267694706748679e-05, + "loss": 1.4284, + "step": 85030 + }, + { + "epoch": 7.36, + "learning_rate": 4.267608074157498e-05, + "loss": 1.422, + "step": 85040 + }, + { + "epoch": 7.36, + "learning_rate": 4.267521441566317e-05, + "loss": 1.4571, + "step": 85050 + }, + { + "epoch": 7.36, + "learning_rate": 4.2674348089751364e-05, + "loss": 1.4629, + "step": 85060 + }, + { + "epoch": 7.36, + "learning_rate": 4.267348176383956e-05, + "loss": 1.4579, + "step": 85070 + }, + { + "epoch": 7.36, + "learning_rate": 4.2672615437927746e-05, + "loss": 1.4871, + "step": 85080 + }, + { + "epoch": 7.37, + "learning_rate": 4.267174911201594e-05, + "loss": 1.4589, + "step": 85090 + }, + { + "epoch": 7.37, + "learning_rate": 4.2670882786104135e-05, + "loss": 1.369, + "step": 85100 + }, + { + "epoch": 7.37, + "learning_rate": 4.267001646019232e-05, + "loss": 1.4047, + "step": 85110 + }, + { + "epoch": 7.37, + "learning_rate": 4.266915013428052e-05, + "loss": 1.4753, + "step": 85120 + }, + { + "epoch": 7.37, + "learning_rate": 4.266828380836871e-05, + "loss": 1.4103, + "step": 85130 + }, + { + "epoch": 7.37, + "learning_rate": 4.26674174824569e-05, + "loss": 1.4583, + "step": 85140 + }, + { + "epoch": 7.37, + "learning_rate": 4.2666551156545094e-05, + "loss": 1.4324, + "step": 85150 + }, + { + "epoch": 7.37, + "learning_rate": 4.266568483063328e-05, + "loss": 1.4101, + "step": 85160 + }, + { + "epoch": 7.37, + "learning_rate": 4.2664818504721476e-05, + "loss": 1.4001, + "step": 85170 + }, + { + "epoch": 7.37, + "learning_rate": 4.266395217880967e-05, + "loss": 1.4268, + "step": 85180 + }, + { + "epoch": 7.37, + "learning_rate": 4.266308585289786e-05, + "loss": 1.4201, + "step": 85190 + }, + { + "epoch": 7.37, + "learning_rate": 4.266221952698605e-05, + "loss": 1.3616, + "step": 85200 + }, + { + "epoch": 7.38, + "learning_rate": 4.266135320107425e-05, + "loss": 1.3868, + "step": 85210 + }, + { + "epoch": 7.38, + "learning_rate": 4.2660486875162435e-05, + "loss": 1.4072, + "step": 85220 + }, + { + "epoch": 7.38, + "learning_rate": 4.265962054925063e-05, + "loss": 1.5022, + "step": 85230 + }, + { + "epoch": 7.38, + "learning_rate": 4.2658754223338824e-05, + "loss": 1.4153, + "step": 85240 + }, + { + "epoch": 7.38, + "learning_rate": 4.265788789742701e-05, + "loss": 1.476, + "step": 85250 + }, + { + "epoch": 7.38, + "learning_rate": 4.2657021571515206e-05, + "loss": 1.4716, + "step": 85260 + }, + { + "epoch": 7.38, + "learning_rate": 4.265615524560339e-05, + "loss": 1.3896, + "step": 85270 + }, + { + "epoch": 7.38, + "learning_rate": 4.265528891969159e-05, + "loss": 1.4412, + "step": 85280 + }, + { + "epoch": 7.38, + "learning_rate": 4.265442259377978e-05, + "loss": 1.3722, + "step": 85290 + }, + { + "epoch": 7.38, + "learning_rate": 4.265355626786797e-05, + "loss": 1.4479, + "step": 85300 + }, + { + "epoch": 7.38, + "learning_rate": 4.2652689941956164e-05, + "loss": 1.3743, + "step": 85310 + }, + { + "epoch": 7.39, + "learning_rate": 4.265182361604436e-05, + "loss": 1.3623, + "step": 85320 + }, + { + "epoch": 7.39, + "learning_rate": 4.2650957290132547e-05, + "loss": 1.3769, + "step": 85330 + }, + { + "epoch": 7.39, + "learning_rate": 4.265009096422074e-05, + "loss": 1.4143, + "step": 85340 + }, + { + "epoch": 7.39, + "learning_rate": 4.2649224638308935e-05, + "loss": 1.3968, + "step": 85350 + }, + { + "epoch": 7.39, + "learning_rate": 4.264835831239712e-05, + "loss": 1.3813, + "step": 85360 + }, + { + "epoch": 7.39, + "learning_rate": 4.264749198648532e-05, + "loss": 1.4032, + "step": 85370 + }, + { + "epoch": 7.39, + "learning_rate": 4.2646625660573505e-05, + "loss": 1.4054, + "step": 85380 + }, + { + "epoch": 7.39, + "learning_rate": 4.26457593346617e-05, + "loss": 1.5063, + "step": 85390 + }, + { + "epoch": 7.39, + "learning_rate": 4.2644893008749894e-05, + "loss": 1.4558, + "step": 85400 + }, + { + "epoch": 7.39, + "learning_rate": 4.264402668283808e-05, + "loss": 1.4439, + "step": 85410 + }, + { + "epoch": 7.39, + "learning_rate": 4.2643160356926276e-05, + "loss": 1.4304, + "step": 85420 + }, + { + "epoch": 7.39, + "learning_rate": 4.264229403101447e-05, + "loss": 1.4201, + "step": 85430 + }, + { + "epoch": 7.4, + "learning_rate": 4.264142770510266e-05, + "loss": 1.377, + "step": 85440 + }, + { + "epoch": 7.4, + "learning_rate": 4.264056137919085e-05, + "loss": 1.37, + "step": 85450 + }, + { + "epoch": 7.4, + "learning_rate": 4.263969505327905e-05, + "loss": 1.4475, + "step": 85460 + }, + { + "epoch": 7.4, + "learning_rate": 4.2638828727367235e-05, + "loss": 1.4231, + "step": 85470 + }, + { + "epoch": 7.4, + "learning_rate": 4.263796240145543e-05, + "loss": 1.4678, + "step": 85480 + }, + { + "epoch": 7.4, + "learning_rate": 4.263709607554362e-05, + "loss": 1.4536, + "step": 85490 + }, + { + "epoch": 7.4, + "learning_rate": 4.263622974963181e-05, + "loss": 1.5014, + "step": 85500 + }, + { + "epoch": 7.4, + "learning_rate": 4.2635363423720006e-05, + "loss": 1.4944, + "step": 85510 + }, + { + "epoch": 7.4, + "learning_rate": 4.2634497097808194e-05, + "loss": 1.4723, + "step": 85520 + }, + { + "epoch": 7.4, + "learning_rate": 4.263363077189639e-05, + "loss": 1.4202, + "step": 85530 + }, + { + "epoch": 7.4, + "learning_rate": 4.263276444598458e-05, + "loss": 1.4615, + "step": 85540 + }, + { + "epoch": 7.4, + "learning_rate": 4.263189812007277e-05, + "loss": 1.3624, + "step": 85550 + }, + { + "epoch": 7.41, + "learning_rate": 4.2631031794160965e-05, + "loss": 1.4826, + "step": 85560 + }, + { + "epoch": 7.41, + "learning_rate": 4.263016546824915e-05, + "loss": 1.3757, + "step": 85570 + }, + { + "epoch": 7.41, + "learning_rate": 4.262929914233735e-05, + "loss": 1.5685, + "step": 85580 + }, + { + "epoch": 7.41, + "learning_rate": 4.262843281642554e-05, + "loss": 1.4049, + "step": 85590 + }, + { + "epoch": 7.41, + "learning_rate": 4.262756649051373e-05, + "loss": 1.3774, + "step": 85600 + }, + { + "epoch": 7.41, + "learning_rate": 4.2626700164601923e-05, + "loss": 1.4683, + "step": 85610 + }, + { + "epoch": 7.41, + "learning_rate": 4.262583383869012e-05, + "loss": 1.4137, + "step": 85620 + }, + { + "epoch": 7.41, + "learning_rate": 4.2624967512778306e-05, + "loss": 1.4482, + "step": 85630 + }, + { + "epoch": 7.41, + "learning_rate": 4.26241011868665e-05, + "loss": 1.4433, + "step": 85640 + }, + { + "epoch": 7.41, + "learning_rate": 4.2623234860954694e-05, + "loss": 1.4536, + "step": 85650 + }, + { + "epoch": 7.41, + "learning_rate": 4.262236853504288e-05, + "loss": 1.4377, + "step": 85660 + }, + { + "epoch": 7.42, + "learning_rate": 4.262150220913108e-05, + "loss": 1.4181, + "step": 85670 + }, + { + "epoch": 7.42, + "learning_rate": 4.2620635883219264e-05, + "loss": 1.4218, + "step": 85680 + }, + { + "epoch": 7.42, + "learning_rate": 4.261976955730746e-05, + "loss": 1.4279, + "step": 85690 + }, + { + "epoch": 7.42, + "learning_rate": 4.261890323139565e-05, + "loss": 1.4148, + "step": 85700 + }, + { + "epoch": 7.42, + "learning_rate": 4.261803690548384e-05, + "loss": 1.4577, + "step": 85710 + }, + { + "epoch": 7.42, + "learning_rate": 4.2617170579572035e-05, + "loss": 1.4797, + "step": 85720 + }, + { + "epoch": 7.42, + "learning_rate": 4.261630425366023e-05, + "loss": 1.4336, + "step": 85730 + }, + { + "epoch": 7.42, + "learning_rate": 4.261543792774842e-05, + "loss": 1.4284, + "step": 85740 + }, + { + "epoch": 7.42, + "learning_rate": 4.261457160183661e-05, + "loss": 1.3982, + "step": 85750 + }, + { + "epoch": 7.42, + "learning_rate": 4.2613705275924806e-05, + "loss": 1.3937, + "step": 85760 + }, + { + "epoch": 7.42, + "learning_rate": 4.2612838950012994e-05, + "loss": 1.4182, + "step": 85770 + }, + { + "epoch": 7.42, + "learning_rate": 4.261197262410119e-05, + "loss": 1.3292, + "step": 85780 + }, + { + "epoch": 7.43, + "learning_rate": 4.2611106298189376e-05, + "loss": 1.4198, + "step": 85790 + }, + { + "epoch": 7.43, + "learning_rate": 4.261023997227757e-05, + "loss": 1.4763, + "step": 85800 + }, + { + "epoch": 7.43, + "learning_rate": 4.2609373646365765e-05, + "loss": 1.4206, + "step": 85810 + }, + { + "epoch": 7.43, + "learning_rate": 4.260850732045395e-05, + "loss": 1.4504, + "step": 85820 + }, + { + "epoch": 7.43, + "learning_rate": 4.260764099454215e-05, + "loss": 1.4021, + "step": 85830 + }, + { + "epoch": 7.43, + "learning_rate": 4.260677466863034e-05, + "loss": 1.4828, + "step": 85840 + }, + { + "epoch": 7.43, + "learning_rate": 4.260590834271853e-05, + "loss": 1.4576, + "step": 85850 + }, + { + "epoch": 7.43, + "learning_rate": 4.2605042016806724e-05, + "loss": 1.4994, + "step": 85860 + }, + { + "epoch": 7.43, + "learning_rate": 4.260417569089492e-05, + "loss": 1.4056, + "step": 85870 + }, + { + "epoch": 7.43, + "learning_rate": 4.2603309364983106e-05, + "loss": 1.4164, + "step": 85880 + }, + { + "epoch": 7.43, + "learning_rate": 4.26024430390713e-05, + "loss": 1.419, + "step": 85890 + }, + { + "epoch": 7.44, + "learning_rate": 4.260157671315949e-05, + "loss": 1.3892, + "step": 85900 + }, + { + "epoch": 7.44, + "learning_rate": 4.260071038724768e-05, + "loss": 1.4719, + "step": 85910 + }, + { + "epoch": 7.44, + "learning_rate": 4.259984406133588e-05, + "loss": 1.3986, + "step": 85920 + }, + { + "epoch": 7.44, + "learning_rate": 4.2598977735424065e-05, + "loss": 1.3593, + "step": 85930 + }, + { + "epoch": 7.44, + "learning_rate": 4.259811140951226e-05, + "loss": 1.3827, + "step": 85940 + }, + { + "epoch": 7.44, + "learning_rate": 4.2597245083600454e-05, + "loss": 1.4219, + "step": 85950 + }, + { + "epoch": 7.44, + "learning_rate": 4.259637875768864e-05, + "loss": 1.3678, + "step": 85960 + }, + { + "epoch": 7.44, + "learning_rate": 4.2595512431776836e-05, + "loss": 1.4045, + "step": 85970 + }, + { + "epoch": 7.44, + "learning_rate": 4.259464610586503e-05, + "loss": 1.3954, + "step": 85980 + }, + { + "epoch": 7.44, + "learning_rate": 4.259377977995322e-05, + "loss": 1.3831, + "step": 85990 + }, + { + "epoch": 7.44, + "learning_rate": 4.259291345404141e-05, + "loss": 1.4656, + "step": 86000 + }, + { + "epoch": 7.44, + "learning_rate": 4.25920471281296e-05, + "loss": 1.409, + "step": 86010 + }, + { + "epoch": 7.45, + "learning_rate": 4.2591180802217794e-05, + "loss": 1.499, + "step": 86020 + }, + { + "epoch": 7.45, + "learning_rate": 4.259031447630599e-05, + "loss": 1.4447, + "step": 86030 + }, + { + "epoch": 7.45, + "learning_rate": 4.2589448150394177e-05, + "loss": 1.412, + "step": 86040 + }, + { + "epoch": 7.45, + "learning_rate": 4.258858182448237e-05, + "loss": 1.4227, + "step": 86050 + }, + { + "epoch": 7.45, + "learning_rate": 4.2587715498570565e-05, + "loss": 1.4038, + "step": 86060 + }, + { + "epoch": 7.45, + "learning_rate": 4.258684917265875e-05, + "loss": 1.4801, + "step": 86070 + }, + { + "epoch": 7.45, + "learning_rate": 4.258598284674695e-05, + "loss": 1.4278, + "step": 86080 + }, + { + "epoch": 7.45, + "learning_rate": 4.258511652083514e-05, + "loss": 1.4535, + "step": 86090 + }, + { + "epoch": 7.45, + "learning_rate": 4.258425019492333e-05, + "loss": 1.4447, + "step": 86100 + }, + { + "epoch": 7.45, + "learning_rate": 4.2583383869011524e-05, + "loss": 1.3931, + "step": 86110 + }, + { + "epoch": 7.45, + "learning_rate": 4.258251754309971e-05, + "loss": 1.4891, + "step": 86120 + }, + { + "epoch": 7.46, + "learning_rate": 4.2581651217187906e-05, + "loss": 1.4572, + "step": 86130 + }, + { + "epoch": 7.46, + "learning_rate": 4.25807848912761e-05, + "loss": 1.3771, + "step": 86140 + }, + { + "epoch": 7.46, + "learning_rate": 4.257991856536429e-05, + "loss": 1.373, + "step": 86150 + }, + { + "epoch": 7.46, + "learning_rate": 4.257905223945248e-05, + "loss": 1.4847, + "step": 86160 + }, + { + "epoch": 7.46, + "learning_rate": 4.257818591354068e-05, + "loss": 1.4316, + "step": 86170 + }, + { + "epoch": 7.46, + "learning_rate": 4.2577319587628865e-05, + "loss": 1.4767, + "step": 86180 + }, + { + "epoch": 7.46, + "learning_rate": 4.257645326171706e-05, + "loss": 1.4305, + "step": 86190 + }, + { + "epoch": 7.46, + "learning_rate": 4.2575586935805254e-05, + "loss": 1.4254, + "step": 86200 + }, + { + "epoch": 7.46, + "learning_rate": 4.257472060989344e-05, + "loss": 1.4419, + "step": 86210 + }, + { + "epoch": 7.46, + "learning_rate": 4.2573854283981636e-05, + "loss": 1.4381, + "step": 86220 + }, + { + "epoch": 7.46, + "learning_rate": 4.2572987958069824e-05, + "loss": 1.4136, + "step": 86230 + }, + { + "epoch": 7.46, + "learning_rate": 4.257212163215802e-05, + "loss": 1.4469, + "step": 86240 + }, + { + "epoch": 7.47, + "learning_rate": 4.257125530624621e-05, + "loss": 1.4651, + "step": 86250 + }, + { + "epoch": 7.47, + "learning_rate": 4.25703889803344e-05, + "loss": 1.4773, + "step": 86260 + }, + { + "epoch": 7.47, + "learning_rate": 4.2569522654422595e-05, + "loss": 1.4522, + "step": 86270 + }, + { + "epoch": 7.47, + "learning_rate": 4.256865632851079e-05, + "loss": 1.4668, + "step": 86280 + }, + { + "epoch": 7.47, + "learning_rate": 4.256779000259898e-05, + "loss": 1.397, + "step": 86290 + }, + { + "epoch": 7.47, + "learning_rate": 4.256692367668717e-05, + "loss": 1.3849, + "step": 86300 + }, + { + "epoch": 7.47, + "learning_rate": 4.2566057350775366e-05, + "loss": 1.4434, + "step": 86310 + }, + { + "epoch": 7.47, + "learning_rate": 4.2565191024863553e-05, + "loss": 1.4023, + "step": 86320 + }, + { + "epoch": 7.47, + "learning_rate": 4.256432469895175e-05, + "loss": 1.4522, + "step": 86330 + }, + { + "epoch": 7.47, + "learning_rate": 4.2563458373039936e-05, + "loss": 1.4256, + "step": 86340 + }, + { + "epoch": 7.47, + "learning_rate": 4.256259204712813e-05, + "loss": 1.4487, + "step": 86350 + }, + { + "epoch": 7.48, + "learning_rate": 4.2561725721216324e-05, + "loss": 1.4048, + "step": 86360 + }, + { + "epoch": 7.48, + "learning_rate": 4.256085939530451e-05, + "loss": 1.4066, + "step": 86370 + }, + { + "epoch": 7.48, + "learning_rate": 4.2559993069392707e-05, + "loss": 1.3406, + "step": 86380 + }, + { + "epoch": 7.48, + "learning_rate": 4.25591267434809e-05, + "loss": 1.3606, + "step": 86390 + }, + { + "epoch": 7.48, + "learning_rate": 4.255826041756909e-05, + "loss": 1.46, + "step": 86400 + }, + { + "epoch": 7.48, + "learning_rate": 4.255739409165728e-05, + "loss": 1.4048, + "step": 86410 + }, + { + "epoch": 7.48, + "learning_rate": 4.255652776574547e-05, + "loss": 1.3758, + "step": 86420 + }, + { + "epoch": 7.48, + "learning_rate": 4.2555661439833665e-05, + "loss": 1.4576, + "step": 86430 + }, + { + "epoch": 7.48, + "learning_rate": 4.255479511392186e-05, + "loss": 1.459, + "step": 86440 + }, + { + "epoch": 7.48, + "learning_rate": 4.255392878801005e-05, + "loss": 1.4827, + "step": 86450 + }, + { + "epoch": 7.48, + "learning_rate": 4.255306246209824e-05, + "loss": 1.4665, + "step": 86460 + }, + { + "epoch": 7.48, + "learning_rate": 4.2552196136186436e-05, + "loss": 1.4033, + "step": 86470 + }, + { + "epoch": 7.49, + "learning_rate": 4.2551329810274624e-05, + "loss": 1.4417, + "step": 86480 + }, + { + "epoch": 7.49, + "learning_rate": 4.255046348436282e-05, + "loss": 1.4413, + "step": 86490 + }, + { + "epoch": 7.49, + "learning_rate": 4.254959715845101e-05, + "loss": 1.5044, + "step": 86500 + }, + { + "epoch": 7.49, + "learning_rate": 4.25487308325392e-05, + "loss": 1.4454, + "step": 86510 + }, + { + "epoch": 7.49, + "learning_rate": 4.2547864506627395e-05, + "loss": 1.391, + "step": 86520 + }, + { + "epoch": 7.49, + "learning_rate": 4.254699818071558e-05, + "loss": 1.4273, + "step": 86530 + }, + { + "epoch": 7.49, + "learning_rate": 4.254613185480378e-05, + "loss": 1.4075, + "step": 86540 + }, + { + "epoch": 7.49, + "learning_rate": 4.254526552889197e-05, + "loss": 1.3972, + "step": 86550 + }, + { + "epoch": 7.49, + "learning_rate": 4.254439920298016e-05, + "loss": 1.4185, + "step": 86560 + }, + { + "epoch": 7.49, + "learning_rate": 4.2543532877068354e-05, + "loss": 1.3881, + "step": 86570 + }, + { + "epoch": 7.49, + "learning_rate": 4.254266655115655e-05, + "loss": 1.4493, + "step": 86580 + }, + { + "epoch": 7.49, + "learning_rate": 4.2541800225244736e-05, + "loss": 1.4551, + "step": 86590 + }, + { + "epoch": 7.5, + "learning_rate": 4.254093389933293e-05, + "loss": 1.3536, + "step": 86600 + }, + { + "epoch": 7.5, + "learning_rate": 4.2540067573421125e-05, + "loss": 1.4217, + "step": 86610 + }, + { + "epoch": 7.5, + "learning_rate": 4.253920124750931e-05, + "loss": 1.4141, + "step": 86620 + }, + { + "epoch": 7.5, + "learning_rate": 4.253833492159751e-05, + "loss": 1.3951, + "step": 86630 + }, + { + "epoch": 7.5, + "learning_rate": 4.2537468595685695e-05, + "loss": 1.4546, + "step": 86640 + }, + { + "epoch": 7.5, + "learning_rate": 4.253660226977389e-05, + "loss": 1.3546, + "step": 86650 + }, + { + "epoch": 7.5, + "learning_rate": 4.2535735943862084e-05, + "loss": 1.3966, + "step": 86660 + }, + { + "epoch": 7.5, + "learning_rate": 4.253486961795027e-05, + "loss": 1.3847, + "step": 86670 + }, + { + "epoch": 7.5, + "learning_rate": 4.2534003292038466e-05, + "loss": 1.3953, + "step": 86680 + }, + { + "epoch": 7.5, + "learning_rate": 4.253313696612666e-05, + "loss": 1.4771, + "step": 86690 + }, + { + "epoch": 7.5, + "learning_rate": 4.253227064021485e-05, + "loss": 1.4068, + "step": 86700 + }, + { + "epoch": 7.51, + "learning_rate": 4.253140431430304e-05, + "loss": 1.4021, + "step": 86710 + }, + { + "epoch": 7.51, + "learning_rate": 4.253053798839124e-05, + "loss": 1.3872, + "step": 86720 + }, + { + "epoch": 7.51, + "learning_rate": 4.2529671662479424e-05, + "loss": 1.364, + "step": 86730 + }, + { + "epoch": 7.51, + "learning_rate": 4.252880533656762e-05, + "loss": 1.4195, + "step": 86740 + }, + { + "epoch": 7.51, + "learning_rate": 4.2527939010655806e-05, + "loss": 1.4418, + "step": 86750 + }, + { + "epoch": 7.51, + "learning_rate": 4.2527072684744e-05, + "loss": 1.4945, + "step": 86760 + }, + { + "epoch": 7.51, + "learning_rate": 4.2526206358832195e-05, + "loss": 1.3867, + "step": 86770 + }, + { + "epoch": 7.51, + "learning_rate": 4.252534003292038e-05, + "loss": 1.4045, + "step": 86780 + }, + { + "epoch": 7.51, + "learning_rate": 4.252447370700858e-05, + "loss": 1.4779, + "step": 86790 + }, + { + "epoch": 7.51, + "learning_rate": 4.252360738109677e-05, + "loss": 1.4744, + "step": 86800 + }, + { + "epoch": 7.51, + "learning_rate": 4.252274105518496e-05, + "loss": 1.421, + "step": 86810 + }, + { + "epoch": 7.51, + "learning_rate": 4.2521874729273154e-05, + "loss": 1.4285, + "step": 86820 + }, + { + "epoch": 7.52, + "learning_rate": 4.252100840336135e-05, + "loss": 1.3853, + "step": 86830 + }, + { + "epoch": 7.52, + "learning_rate": 4.2520142077449536e-05, + "loss": 1.3987, + "step": 86840 + }, + { + "epoch": 7.52, + "learning_rate": 4.251927575153773e-05, + "loss": 1.5138, + "step": 86850 + }, + { + "epoch": 7.52, + "learning_rate": 4.251840942562592e-05, + "loss": 1.4471, + "step": 86860 + }, + { + "epoch": 7.52, + "learning_rate": 4.251754309971411e-05, + "loss": 1.4645, + "step": 86870 + }, + { + "epoch": 7.52, + "learning_rate": 4.251667677380231e-05, + "loss": 1.4449, + "step": 86880 + }, + { + "epoch": 7.52, + "learning_rate": 4.2515810447890495e-05, + "loss": 1.4359, + "step": 86890 + }, + { + "epoch": 7.52, + "learning_rate": 4.251494412197869e-05, + "loss": 1.378, + "step": 86900 + }, + { + "epoch": 7.52, + "learning_rate": 4.2514077796066884e-05, + "loss": 1.4584, + "step": 86910 + }, + { + "epoch": 7.52, + "learning_rate": 4.251321147015507e-05, + "loss": 1.3984, + "step": 86920 + }, + { + "epoch": 7.52, + "learning_rate": 4.2512345144243266e-05, + "loss": 1.3514, + "step": 86930 + }, + { + "epoch": 7.53, + "learning_rate": 4.251147881833146e-05, + "loss": 1.4521, + "step": 86940 + }, + { + "epoch": 7.53, + "learning_rate": 4.251061249241965e-05, + "loss": 1.4392, + "step": 86950 + }, + { + "epoch": 7.53, + "learning_rate": 4.250974616650784e-05, + "loss": 1.4237, + "step": 86960 + }, + { + "epoch": 7.53, + "learning_rate": 4.250887984059603e-05, + "loss": 1.4293, + "step": 86970 + }, + { + "epoch": 7.53, + "learning_rate": 4.2508013514684225e-05, + "loss": 1.3892, + "step": 86980 + }, + { + "epoch": 7.53, + "learning_rate": 4.250714718877242e-05, + "loss": 1.4372, + "step": 86990 + }, + { + "epoch": 7.53, + "learning_rate": 4.250628086286061e-05, + "loss": 1.4157, + "step": 87000 + }, + { + "epoch": 7.53, + "learning_rate": 4.25054145369488e-05, + "loss": 1.4297, + "step": 87010 + }, + { + "epoch": 7.53, + "learning_rate": 4.2504548211036996e-05, + "loss": 1.4453, + "step": 87020 + }, + { + "epoch": 7.53, + "learning_rate": 4.2503681885125183e-05, + "loss": 1.4476, + "step": 87030 + }, + { + "epoch": 7.53, + "learning_rate": 4.250281555921338e-05, + "loss": 1.5186, + "step": 87040 + }, + { + "epoch": 7.53, + "learning_rate": 4.250194923330157e-05, + "loss": 1.4363, + "step": 87050 + }, + { + "epoch": 7.54, + "learning_rate": 4.250108290738976e-05, + "loss": 1.4089, + "step": 87060 + }, + { + "epoch": 7.54, + "learning_rate": 4.2500216581477954e-05, + "loss": 1.4129, + "step": 87070 + }, + { + "epoch": 7.54, + "learning_rate": 4.249935025556614e-05, + "loss": 1.3695, + "step": 87080 + }, + { + "epoch": 7.54, + "learning_rate": 4.2498483929654337e-05, + "loss": 1.3881, + "step": 87090 + }, + { + "epoch": 7.54, + "learning_rate": 4.249761760374253e-05, + "loss": 1.4174, + "step": 87100 + }, + { + "epoch": 7.54, + "learning_rate": 4.249675127783072e-05, + "loss": 1.3971, + "step": 87110 + }, + { + "epoch": 7.54, + "learning_rate": 4.249588495191891e-05, + "loss": 1.4532, + "step": 87120 + }, + { + "epoch": 7.54, + "learning_rate": 4.249501862600711e-05, + "loss": 1.447, + "step": 87130 + }, + { + "epoch": 7.54, + "learning_rate": 4.2494152300095295e-05, + "loss": 1.4496, + "step": 87140 + }, + { + "epoch": 7.54, + "learning_rate": 4.249328597418349e-05, + "loss": 1.3951, + "step": 87150 + }, + { + "epoch": 7.54, + "learning_rate": 4.249241964827168e-05, + "loss": 1.4093, + "step": 87160 + }, + { + "epoch": 7.55, + "learning_rate": 4.249155332235987e-05, + "loss": 1.43, + "step": 87170 + }, + { + "epoch": 7.55, + "learning_rate": 4.2490686996448066e-05, + "loss": 1.3898, + "step": 87180 + }, + { + "epoch": 7.55, + "learning_rate": 4.2489820670536254e-05, + "loss": 1.4276, + "step": 87190 + }, + { + "epoch": 7.55, + "learning_rate": 4.248895434462445e-05, + "loss": 1.4515, + "step": 87200 + }, + { + "epoch": 7.55, + "learning_rate": 4.248808801871264e-05, + "loss": 1.4759, + "step": 87210 + }, + { + "epoch": 7.55, + "learning_rate": 4.248722169280083e-05, + "loss": 1.414, + "step": 87220 + }, + { + "epoch": 7.55, + "learning_rate": 4.2486355366889025e-05, + "loss": 1.4305, + "step": 87230 + }, + { + "epoch": 7.55, + "learning_rate": 4.248548904097722e-05, + "loss": 1.4206, + "step": 87240 + }, + { + "epoch": 7.55, + "learning_rate": 4.248462271506541e-05, + "loss": 1.3517, + "step": 87250 + }, + { + "epoch": 7.55, + "learning_rate": 4.24837563891536e-05, + "loss": 1.4246, + "step": 87260 + }, + { + "epoch": 7.55, + "learning_rate": 4.248289006324179e-05, + "loss": 1.4401, + "step": 87270 + }, + { + "epoch": 7.55, + "learning_rate": 4.2482023737329984e-05, + "loss": 1.4565, + "step": 87280 + }, + { + "epoch": 7.56, + "learning_rate": 4.248115741141818e-05, + "loss": 1.4808, + "step": 87290 + }, + { + "epoch": 7.56, + "learning_rate": 4.2480291085506366e-05, + "loss": 1.4774, + "step": 87300 + }, + { + "epoch": 7.56, + "learning_rate": 4.247942475959456e-05, + "loss": 1.414, + "step": 87310 + }, + { + "epoch": 7.56, + "learning_rate": 4.2478558433682755e-05, + "loss": 1.4128, + "step": 87320 + }, + { + "epoch": 7.56, + "learning_rate": 4.247769210777094e-05, + "loss": 1.4196, + "step": 87330 + }, + { + "epoch": 7.56, + "learning_rate": 4.247682578185914e-05, + "loss": 1.4075, + "step": 87340 + }, + { + "epoch": 7.56, + "learning_rate": 4.247595945594733e-05, + "loss": 1.4109, + "step": 87350 + }, + { + "epoch": 7.56, + "learning_rate": 4.247509313003552e-05, + "loss": 1.4372, + "step": 87360 + }, + { + "epoch": 7.56, + "learning_rate": 4.2474226804123713e-05, + "loss": 1.4369, + "step": 87370 + }, + { + "epoch": 7.56, + "learning_rate": 4.24733604782119e-05, + "loss": 1.4295, + "step": 87380 + }, + { + "epoch": 7.56, + "learning_rate": 4.2472494152300096e-05, + "loss": 1.4367, + "step": 87390 + }, + { + "epoch": 7.57, + "learning_rate": 4.247162782638829e-05, + "loss": 1.4425, + "step": 87400 + }, + { + "epoch": 7.57, + "learning_rate": 4.247076150047648e-05, + "loss": 1.451, + "step": 87410 + }, + { + "epoch": 7.57, + "learning_rate": 4.246989517456467e-05, + "loss": 1.4169, + "step": 87420 + }, + { + "epoch": 7.57, + "learning_rate": 4.246902884865287e-05, + "loss": 1.4457, + "step": 87430 + }, + { + "epoch": 7.57, + "learning_rate": 4.2468162522741054e-05, + "loss": 1.4475, + "step": 87440 + }, + { + "epoch": 7.57, + "learning_rate": 4.246729619682925e-05, + "loss": 1.4251, + "step": 87450 + }, + { + "epoch": 7.57, + "learning_rate": 4.246642987091744e-05, + "loss": 1.3732, + "step": 87460 + }, + { + "epoch": 7.57, + "learning_rate": 4.246556354500563e-05, + "loss": 1.3435, + "step": 87470 + }, + { + "epoch": 7.57, + "learning_rate": 4.2464697219093825e-05, + "loss": 1.3778, + "step": 87480 + }, + { + "epoch": 7.57, + "learning_rate": 4.246383089318201e-05, + "loss": 1.4257, + "step": 87490 + }, + { + "epoch": 7.57, + "learning_rate": 4.246296456727021e-05, + "loss": 1.4181, + "step": 87500 + }, + { + "epoch": 7.57, + "learning_rate": 4.24620982413584e-05, + "loss": 1.4109, + "step": 87510 + }, + { + "epoch": 7.58, + "learning_rate": 4.246123191544659e-05, + "loss": 1.4087, + "step": 87520 + }, + { + "epoch": 7.58, + "learning_rate": 4.2460365589534784e-05, + "loss": 1.4504, + "step": 87530 + }, + { + "epoch": 7.58, + "learning_rate": 4.245949926362298e-05, + "loss": 1.46, + "step": 87540 + }, + { + "epoch": 7.58, + "learning_rate": 4.2458632937711166e-05, + "loss": 1.3833, + "step": 87550 + }, + { + "epoch": 7.58, + "learning_rate": 4.245776661179936e-05, + "loss": 1.4451, + "step": 87560 + }, + { + "epoch": 7.58, + "learning_rate": 4.2456900285887555e-05, + "loss": 1.3893, + "step": 87570 + }, + { + "epoch": 7.58, + "learning_rate": 4.245603395997574e-05, + "loss": 1.3581, + "step": 87580 + }, + { + "epoch": 7.58, + "learning_rate": 4.245516763406394e-05, + "loss": 1.438, + "step": 87590 + }, + { + "epoch": 7.58, + "learning_rate": 4.2454301308152125e-05, + "loss": 1.376, + "step": 87600 + }, + { + "epoch": 7.58, + "learning_rate": 4.245343498224032e-05, + "loss": 1.4157, + "step": 87610 + }, + { + "epoch": 7.58, + "learning_rate": 4.2452568656328514e-05, + "loss": 1.429, + "step": 87620 + }, + { + "epoch": 7.58, + "learning_rate": 4.24517023304167e-05, + "loss": 1.4631, + "step": 87630 + }, + { + "epoch": 7.59, + "learning_rate": 4.2450836004504896e-05, + "loss": 1.4221, + "step": 87640 + }, + { + "epoch": 7.59, + "learning_rate": 4.244996967859309e-05, + "loss": 1.5141, + "step": 87650 + }, + { + "epoch": 7.59, + "learning_rate": 4.244910335268128e-05, + "loss": 1.4205, + "step": 87660 + }, + { + "epoch": 7.59, + "learning_rate": 4.244823702676947e-05, + "loss": 1.4101, + "step": 87670 + }, + { + "epoch": 7.59, + "learning_rate": 4.244737070085767e-05, + "loss": 1.4175, + "step": 87680 + }, + { + "epoch": 7.59, + "learning_rate": 4.2446504374945855e-05, + "loss": 1.3983, + "step": 87690 + }, + { + "epoch": 7.59, + "learning_rate": 4.244563804903405e-05, + "loss": 1.5201, + "step": 87700 + }, + { + "epoch": 7.59, + "learning_rate": 4.244477172312224e-05, + "loss": 1.4135, + "step": 87710 + }, + { + "epoch": 7.59, + "learning_rate": 4.244390539721043e-05, + "loss": 1.4435, + "step": 87720 + }, + { + "epoch": 7.59, + "learning_rate": 4.2443039071298626e-05, + "loss": 1.4661, + "step": 87730 + }, + { + "epoch": 7.59, + "learning_rate": 4.244217274538681e-05, + "loss": 1.3817, + "step": 87740 + }, + { + "epoch": 7.6, + "learning_rate": 4.244130641947501e-05, + "loss": 1.4855, + "step": 87750 + }, + { + "epoch": 7.6, + "learning_rate": 4.24404400935632e-05, + "loss": 1.4262, + "step": 87760 + }, + { + "epoch": 7.6, + "learning_rate": 4.243957376765139e-05, + "loss": 1.3796, + "step": 87770 + }, + { + "epoch": 7.6, + "learning_rate": 4.2438707441739584e-05, + "loss": 1.4512, + "step": 87780 + }, + { + "epoch": 7.6, + "learning_rate": 4.243784111582778e-05, + "loss": 1.3933, + "step": 87790 + }, + { + "epoch": 7.6, + "learning_rate": 4.2436974789915967e-05, + "loss": 1.4011, + "step": 87800 + }, + { + "epoch": 7.6, + "learning_rate": 4.243610846400416e-05, + "loss": 1.4333, + "step": 87810 + }, + { + "epoch": 7.6, + "learning_rate": 4.243524213809235e-05, + "loss": 1.4783, + "step": 87820 + }, + { + "epoch": 7.6, + "learning_rate": 4.243437581218054e-05, + "loss": 1.4017, + "step": 87830 + }, + { + "epoch": 7.6, + "learning_rate": 4.243350948626874e-05, + "loss": 1.4525, + "step": 87840 + }, + { + "epoch": 7.6, + "learning_rate": 4.2432643160356925e-05, + "loss": 1.4294, + "step": 87850 + }, + { + "epoch": 7.6, + "learning_rate": 4.243177683444512e-05, + "loss": 1.3885, + "step": 87860 + }, + { + "epoch": 7.61, + "learning_rate": 4.2430910508533314e-05, + "loss": 1.4196, + "step": 87870 + }, + { + "epoch": 7.61, + "learning_rate": 4.24300441826215e-05, + "loss": 1.3804, + "step": 87880 + }, + { + "epoch": 7.61, + "learning_rate": 4.2429177856709696e-05, + "loss": 1.4328, + "step": 87890 + }, + { + "epoch": 7.61, + "learning_rate": 4.2428311530797884e-05, + "loss": 1.4056, + "step": 87900 + }, + { + "epoch": 7.61, + "learning_rate": 4.242744520488608e-05, + "loss": 1.4502, + "step": 87910 + }, + { + "epoch": 7.61, + "learning_rate": 4.242657887897427e-05, + "loss": 1.4011, + "step": 87920 + }, + { + "epoch": 7.61, + "learning_rate": 4.242571255306246e-05, + "loss": 1.4053, + "step": 87930 + }, + { + "epoch": 7.61, + "learning_rate": 4.2424846227150655e-05, + "loss": 1.3586, + "step": 87940 + }, + { + "epoch": 7.61, + "learning_rate": 4.242397990123885e-05, + "loss": 1.3781, + "step": 87950 + }, + { + "epoch": 7.61, + "learning_rate": 4.242311357532704e-05, + "loss": 1.3753, + "step": 87960 + }, + { + "epoch": 7.61, + "learning_rate": 4.242224724941523e-05, + "loss": 1.4362, + "step": 87970 + }, + { + "epoch": 7.62, + "learning_rate": 4.2421380923503426e-05, + "loss": 1.4445, + "step": 87980 + }, + { + "epoch": 7.62, + "learning_rate": 4.2420514597591614e-05, + "loss": 1.3678, + "step": 87990 + }, + { + "epoch": 7.62, + "learning_rate": 4.241964827167981e-05, + "loss": 1.3268, + "step": 88000 + }, + { + "epoch": 7.62, + "learning_rate": 4.2418781945767996e-05, + "loss": 1.4065, + "step": 88010 + }, + { + "epoch": 7.62, + "learning_rate": 4.241791561985619e-05, + "loss": 1.4656, + "step": 88020 + }, + { + "epoch": 7.62, + "learning_rate": 4.2417049293944385e-05, + "loss": 1.4247, + "step": 88030 + }, + { + "epoch": 7.62, + "learning_rate": 4.241618296803257e-05, + "loss": 1.3443, + "step": 88040 + }, + { + "epoch": 7.62, + "learning_rate": 4.241531664212077e-05, + "loss": 1.4005, + "step": 88050 + }, + { + "epoch": 7.62, + "learning_rate": 4.241445031620896e-05, + "loss": 1.3254, + "step": 88060 + }, + { + "epoch": 7.62, + "learning_rate": 4.241358399029715e-05, + "loss": 1.4308, + "step": 88070 + }, + { + "epoch": 7.62, + "learning_rate": 4.2412717664385343e-05, + "loss": 1.3874, + "step": 88080 + }, + { + "epoch": 7.62, + "learning_rate": 4.241185133847354e-05, + "loss": 1.3761, + "step": 88090 + }, + { + "epoch": 7.63, + "learning_rate": 4.2410985012561726e-05, + "loss": 1.3817, + "step": 88100 + }, + { + "epoch": 7.63, + "learning_rate": 4.241011868664992e-05, + "loss": 1.4097, + "step": 88110 + }, + { + "epoch": 7.63, + "learning_rate": 4.240925236073811e-05, + "loss": 1.3882, + "step": 88120 + }, + { + "epoch": 7.63, + "learning_rate": 4.24083860348263e-05, + "loss": 1.406, + "step": 88130 + }, + { + "epoch": 7.63, + "learning_rate": 4.24075197089145e-05, + "loss": 1.397, + "step": 88140 + }, + { + "epoch": 7.63, + "learning_rate": 4.2406653383002684e-05, + "loss": 1.4367, + "step": 88150 + }, + { + "epoch": 7.63, + "learning_rate": 4.240578705709088e-05, + "loss": 1.3659, + "step": 88160 + }, + { + "epoch": 7.63, + "learning_rate": 4.240492073117907e-05, + "loss": 1.474, + "step": 88170 + }, + { + "epoch": 7.63, + "learning_rate": 4.240405440526726e-05, + "loss": 1.4243, + "step": 88180 + }, + { + "epoch": 7.63, + "learning_rate": 4.2403188079355455e-05, + "loss": 1.4559, + "step": 88190 + }, + { + "epoch": 7.63, + "learning_rate": 4.240232175344365e-05, + "loss": 1.4481, + "step": 88200 + }, + { + "epoch": 7.64, + "learning_rate": 4.240145542753184e-05, + "loss": 1.3959, + "step": 88210 + }, + { + "epoch": 7.64, + "learning_rate": 4.240058910162003e-05, + "loss": 1.4321, + "step": 88220 + }, + { + "epoch": 7.64, + "learning_rate": 4.239972277570822e-05, + "loss": 1.3836, + "step": 88230 + }, + { + "epoch": 7.64, + "learning_rate": 4.2398856449796414e-05, + "loss": 1.3759, + "step": 88240 + }, + { + "epoch": 7.64, + "learning_rate": 4.239799012388461e-05, + "loss": 1.489, + "step": 88250 + }, + { + "epoch": 7.64, + "learning_rate": 4.2397123797972796e-05, + "loss": 1.4131, + "step": 88260 + }, + { + "epoch": 7.64, + "learning_rate": 4.239625747206099e-05, + "loss": 1.3445, + "step": 88270 + }, + { + "epoch": 7.64, + "learning_rate": 4.2395391146149185e-05, + "loss": 1.4473, + "step": 88280 + }, + { + "epoch": 7.64, + "learning_rate": 4.239452482023737e-05, + "loss": 1.3892, + "step": 88290 + }, + { + "epoch": 7.64, + "learning_rate": 4.239365849432557e-05, + "loss": 1.3955, + "step": 88300 + }, + { + "epoch": 7.64, + "learning_rate": 4.239279216841376e-05, + "loss": 1.4408, + "step": 88310 + }, + { + "epoch": 7.64, + "learning_rate": 4.239192584250195e-05, + "loss": 1.4149, + "step": 88320 + }, + { + "epoch": 7.65, + "learning_rate": 4.2391059516590144e-05, + "loss": 1.4821, + "step": 88330 + }, + { + "epoch": 7.65, + "learning_rate": 4.239019319067833e-05, + "loss": 1.4215, + "step": 88340 + }, + { + "epoch": 7.65, + "learning_rate": 4.2389326864766526e-05, + "loss": 1.443, + "step": 88350 + }, + { + "epoch": 7.65, + "learning_rate": 4.238846053885472e-05, + "loss": 1.3766, + "step": 88360 + }, + { + "epoch": 7.65, + "learning_rate": 4.238759421294291e-05, + "loss": 1.4369, + "step": 88370 + }, + { + "epoch": 7.65, + "learning_rate": 4.23867278870311e-05, + "loss": 1.4198, + "step": 88380 + }, + { + "epoch": 7.65, + "learning_rate": 4.23858615611193e-05, + "loss": 1.4153, + "step": 88390 + }, + { + "epoch": 7.65, + "learning_rate": 4.2384995235207485e-05, + "loss": 1.4054, + "step": 88400 + }, + { + "epoch": 7.65, + "learning_rate": 4.238412890929568e-05, + "loss": 1.4295, + "step": 88410 + }, + { + "epoch": 7.65, + "learning_rate": 4.2383262583383874e-05, + "loss": 1.3925, + "step": 88420 + }, + { + "epoch": 7.65, + "learning_rate": 4.238239625747206e-05, + "loss": 1.4368, + "step": 88430 + }, + { + "epoch": 7.66, + "learning_rate": 4.2381529931560256e-05, + "loss": 1.3697, + "step": 88440 + }, + { + "epoch": 7.66, + "learning_rate": 4.238066360564844e-05, + "loss": 1.3951, + "step": 88450 + }, + { + "epoch": 7.66, + "learning_rate": 4.237979727973664e-05, + "loss": 1.4365, + "step": 88460 + }, + { + "epoch": 7.66, + "learning_rate": 4.237893095382483e-05, + "loss": 1.3916, + "step": 88470 + }, + { + "epoch": 7.66, + "learning_rate": 4.237806462791302e-05, + "loss": 1.3871, + "step": 88480 + }, + { + "epoch": 7.66, + "learning_rate": 4.2377198302001214e-05, + "loss": 1.4264, + "step": 88490 + }, + { + "epoch": 7.66, + "learning_rate": 4.237633197608941e-05, + "loss": 1.3787, + "step": 88500 + }, + { + "epoch": 7.66, + "learning_rate": 4.2375465650177597e-05, + "loss": 1.4457, + "step": 88510 + }, + { + "epoch": 7.66, + "learning_rate": 4.237459932426579e-05, + "loss": 1.3667, + "step": 88520 + }, + { + "epoch": 7.66, + "learning_rate": 4.2373732998353985e-05, + "loss": 1.3911, + "step": 88530 + }, + { + "epoch": 7.66, + "learning_rate": 4.237286667244217e-05, + "loss": 1.436, + "step": 88540 + }, + { + "epoch": 7.66, + "learning_rate": 4.237200034653037e-05, + "loss": 1.37, + "step": 88550 + }, + { + "epoch": 7.67, + "learning_rate": 4.2371134020618555e-05, + "loss": 1.4677, + "step": 88560 + }, + { + "epoch": 7.67, + "learning_rate": 4.237026769470675e-05, + "loss": 1.413, + "step": 88570 + }, + { + "epoch": 7.67, + "learning_rate": 4.2369401368794944e-05, + "loss": 1.4541, + "step": 88580 + }, + { + "epoch": 7.67, + "learning_rate": 4.236853504288313e-05, + "loss": 1.3637, + "step": 88590 + }, + { + "epoch": 7.67, + "learning_rate": 4.2367668716971326e-05, + "loss": 1.4182, + "step": 88600 + }, + { + "epoch": 7.67, + "learning_rate": 4.236680239105952e-05, + "loss": 1.4172, + "step": 88610 + }, + { + "epoch": 7.67, + "learning_rate": 4.236593606514771e-05, + "loss": 1.4175, + "step": 88620 + }, + { + "epoch": 7.67, + "learning_rate": 4.23650697392359e-05, + "loss": 1.4038, + "step": 88630 + }, + { + "epoch": 7.67, + "learning_rate": 4.23642034133241e-05, + "loss": 1.4076, + "step": 88640 + }, + { + "epoch": 7.67, + "learning_rate": 4.2363337087412285e-05, + "loss": 1.3807, + "step": 88650 + }, + { + "epoch": 7.67, + "learning_rate": 4.236247076150048e-05, + "loss": 1.4547, + "step": 88660 + }, + { + "epoch": 7.67, + "learning_rate": 4.236160443558867e-05, + "loss": 1.4315, + "step": 88670 + }, + { + "epoch": 7.68, + "learning_rate": 4.236073810967686e-05, + "loss": 1.4197, + "step": 88680 + }, + { + "epoch": 7.68, + "learning_rate": 4.2359871783765056e-05, + "loss": 1.4031, + "step": 88690 + }, + { + "epoch": 7.68, + "learning_rate": 4.2359005457853244e-05, + "loss": 1.3912, + "step": 88700 + }, + { + "epoch": 7.68, + "learning_rate": 4.235813913194144e-05, + "loss": 1.3926, + "step": 88710 + }, + { + "epoch": 7.68, + "learning_rate": 4.235727280602963e-05, + "loss": 1.3782, + "step": 88720 + }, + { + "epoch": 7.68, + "learning_rate": 4.235640648011782e-05, + "loss": 1.4082, + "step": 88730 + }, + { + "epoch": 7.68, + "learning_rate": 4.2355540154206015e-05, + "loss": 1.4517, + "step": 88740 + }, + { + "epoch": 7.68, + "learning_rate": 4.23546738282942e-05, + "loss": 1.3552, + "step": 88750 + }, + { + "epoch": 7.68, + "learning_rate": 4.23538075023824e-05, + "loss": 1.4135, + "step": 88760 + }, + { + "epoch": 7.68, + "learning_rate": 4.235294117647059e-05, + "loss": 1.4186, + "step": 88770 + }, + { + "epoch": 7.68, + "learning_rate": 4.235207485055878e-05, + "loss": 1.4191, + "step": 88780 + }, + { + "epoch": 7.69, + "learning_rate": 4.2351208524646973e-05, + "loss": 1.3921, + "step": 88790 + }, + { + "epoch": 7.69, + "learning_rate": 4.235034219873517e-05, + "loss": 1.4102, + "step": 88800 + }, + { + "epoch": 7.69, + "learning_rate": 4.2349475872823356e-05, + "loss": 1.3788, + "step": 88810 + }, + { + "epoch": 7.69, + "learning_rate": 4.234860954691155e-05, + "loss": 1.4002, + "step": 88820 + }, + { + "epoch": 7.69, + "learning_rate": 4.2347743220999744e-05, + "loss": 1.4699, + "step": 88830 + }, + { + "epoch": 7.69, + "learning_rate": 4.234687689508793e-05, + "loss": 1.4244, + "step": 88840 + }, + { + "epoch": 7.69, + "learning_rate": 4.2346010569176127e-05, + "loss": 1.4388, + "step": 88850 + }, + { + "epoch": 7.69, + "learning_rate": 4.2345144243264314e-05, + "loss": 1.4091, + "step": 88860 + }, + { + "epoch": 7.69, + "learning_rate": 4.234427791735251e-05, + "loss": 1.3621, + "step": 88870 + }, + { + "epoch": 7.69, + "learning_rate": 4.23434115914407e-05, + "loss": 1.3807, + "step": 88880 + }, + { + "epoch": 7.69, + "learning_rate": 4.234254526552889e-05, + "loss": 1.4177, + "step": 88890 + }, + { + "epoch": 7.69, + "learning_rate": 4.2341678939617085e-05, + "loss": 1.4366, + "step": 88900 + }, + { + "epoch": 7.7, + "learning_rate": 4.234081261370528e-05, + "loss": 1.3736, + "step": 88910 + }, + { + "epoch": 7.7, + "learning_rate": 4.233994628779347e-05, + "loss": 1.3904, + "step": 88920 + }, + { + "epoch": 7.7, + "learning_rate": 4.233907996188166e-05, + "loss": 1.3743, + "step": 88930 + }, + { + "epoch": 7.7, + "learning_rate": 4.2338213635969856e-05, + "loss": 1.3775, + "step": 88940 + }, + { + "epoch": 7.7, + "learning_rate": 4.2337347310058044e-05, + "loss": 1.4641, + "step": 88950 + }, + { + "epoch": 7.7, + "learning_rate": 4.233648098414624e-05, + "loss": 1.467, + "step": 88960 + }, + { + "epoch": 7.7, + "learning_rate": 4.2335614658234426e-05, + "loss": 1.3941, + "step": 88970 + }, + { + "epoch": 7.7, + "learning_rate": 4.233474833232262e-05, + "loss": 1.4372, + "step": 88980 + }, + { + "epoch": 7.7, + "learning_rate": 4.2333882006410815e-05, + "loss": 1.3695, + "step": 88990 + }, + { + "epoch": 7.7, + "learning_rate": 4.2333015680499e-05, + "loss": 1.3874, + "step": 89000 + }, + { + "epoch": 7.7, + "learning_rate": 4.23321493545872e-05, + "loss": 1.399, + "step": 89010 + }, + { + "epoch": 7.71, + "learning_rate": 4.233128302867539e-05, + "loss": 1.4511, + "step": 89020 + }, + { + "epoch": 7.71, + "learning_rate": 4.233041670276358e-05, + "loss": 1.3768, + "step": 89030 + }, + { + "epoch": 7.71, + "learning_rate": 4.2329550376851774e-05, + "loss": 1.4178, + "step": 89040 + }, + { + "epoch": 7.71, + "learning_rate": 4.232868405093997e-05, + "loss": 1.3943, + "step": 89050 + }, + { + "epoch": 7.71, + "learning_rate": 4.2327817725028156e-05, + "loss": 1.3735, + "step": 89060 + }, + { + "epoch": 7.71, + "learning_rate": 4.232695139911635e-05, + "loss": 1.4345, + "step": 89070 + }, + { + "epoch": 7.71, + "learning_rate": 4.232608507320454e-05, + "loss": 1.4774, + "step": 89080 + }, + { + "epoch": 7.71, + "learning_rate": 4.232521874729273e-05, + "loss": 1.3932, + "step": 89090 + }, + { + "epoch": 7.71, + "learning_rate": 4.232435242138093e-05, + "loss": 1.3665, + "step": 89100 + }, + { + "epoch": 7.71, + "learning_rate": 4.2323486095469115e-05, + "loss": 1.395, + "step": 89110 + }, + { + "epoch": 7.71, + "learning_rate": 4.232261976955731e-05, + "loss": 1.3709, + "step": 89120 + }, + { + "epoch": 7.71, + "learning_rate": 4.2321753443645504e-05, + "loss": 1.444, + "step": 89130 + }, + { + "epoch": 7.72, + "learning_rate": 4.232088711773369e-05, + "loss": 1.4197, + "step": 89140 + }, + { + "epoch": 7.72, + "learning_rate": 4.2320020791821886e-05, + "loss": 1.4414, + "step": 89150 + }, + { + "epoch": 7.72, + "learning_rate": 4.231915446591008e-05, + "loss": 1.4891, + "step": 89160 + }, + { + "epoch": 7.72, + "learning_rate": 4.231828813999827e-05, + "loss": 1.4167, + "step": 89170 + }, + { + "epoch": 7.72, + "learning_rate": 4.231742181408646e-05, + "loss": 1.3895, + "step": 89180 + }, + { + "epoch": 7.72, + "learning_rate": 4.231655548817465e-05, + "loss": 1.4426, + "step": 89190 + }, + { + "epoch": 7.72, + "learning_rate": 4.2315689162262844e-05, + "loss": 1.4477, + "step": 89200 + }, + { + "epoch": 7.72, + "learning_rate": 4.231482283635104e-05, + "loss": 1.4221, + "step": 89210 + }, + { + "epoch": 7.72, + "learning_rate": 4.2313956510439226e-05, + "loss": 1.4538, + "step": 89220 + }, + { + "epoch": 7.72, + "learning_rate": 4.231309018452742e-05, + "loss": 1.3718, + "step": 89230 + }, + { + "epoch": 7.72, + "learning_rate": 4.2312223858615615e-05, + "loss": 1.4372, + "step": 89240 + }, + { + "epoch": 7.73, + "learning_rate": 4.23113575327038e-05, + "loss": 1.4094, + "step": 89250 + }, + { + "epoch": 7.73, + "learning_rate": 4.2310491206792e-05, + "loss": 1.4295, + "step": 89260 + }, + { + "epoch": 7.73, + "learning_rate": 4.230962488088019e-05, + "loss": 1.3931, + "step": 89270 + }, + { + "epoch": 7.73, + "learning_rate": 4.230875855496838e-05, + "loss": 1.4311, + "step": 89280 + }, + { + "epoch": 7.73, + "learning_rate": 4.2307892229056574e-05, + "loss": 1.3773, + "step": 89290 + }, + { + "epoch": 7.73, + "learning_rate": 4.230702590314476e-05, + "loss": 1.3831, + "step": 89300 + }, + { + "epoch": 7.73, + "learning_rate": 4.2306159577232956e-05, + "loss": 1.4287, + "step": 89310 + }, + { + "epoch": 7.73, + "learning_rate": 4.230529325132115e-05, + "loss": 1.3598, + "step": 89320 + }, + { + "epoch": 7.73, + "learning_rate": 4.230442692540934e-05, + "loss": 1.425, + "step": 89330 + }, + { + "epoch": 7.73, + "learning_rate": 4.230356059949753e-05, + "loss": 1.4298, + "step": 89340 + }, + { + "epoch": 7.73, + "learning_rate": 4.230269427358573e-05, + "loss": 1.3968, + "step": 89350 + }, + { + "epoch": 7.73, + "learning_rate": 4.2301827947673915e-05, + "loss": 1.4023, + "step": 89360 + }, + { + "epoch": 7.74, + "learning_rate": 4.230096162176211e-05, + "loss": 1.4233, + "step": 89370 + }, + { + "epoch": 7.74, + "learning_rate": 4.2300095295850304e-05, + "loss": 1.4972, + "step": 89380 + }, + { + "epoch": 7.74, + "learning_rate": 4.229922896993849e-05, + "loss": 1.3898, + "step": 89390 + }, + { + "epoch": 7.74, + "learning_rate": 4.2298362644026686e-05, + "loss": 1.4041, + "step": 89400 + }, + { + "epoch": 7.74, + "learning_rate": 4.2297496318114874e-05, + "loss": 1.451, + "step": 89410 + }, + { + "epoch": 7.74, + "learning_rate": 4.229662999220307e-05, + "loss": 1.4077, + "step": 89420 + }, + { + "epoch": 7.74, + "learning_rate": 4.229576366629126e-05, + "loss": 1.4291, + "step": 89430 + }, + { + "epoch": 7.74, + "learning_rate": 4.229489734037945e-05, + "loss": 1.3352, + "step": 89440 + }, + { + "epoch": 7.74, + "learning_rate": 4.2294031014467645e-05, + "loss": 1.3369, + "step": 89450 + }, + { + "epoch": 7.74, + "learning_rate": 4.229316468855584e-05, + "loss": 1.3885, + "step": 89460 + }, + { + "epoch": 7.74, + "learning_rate": 4.229229836264403e-05, + "loss": 1.4771, + "step": 89470 + }, + { + "epoch": 7.75, + "learning_rate": 4.229143203673222e-05, + "loss": 1.4132, + "step": 89480 + }, + { + "epoch": 7.75, + "learning_rate": 4.229056571082041e-05, + "loss": 1.3889, + "step": 89490 + }, + { + "epoch": 7.75, + "learning_rate": 4.2289699384908603e-05, + "loss": 1.4122, + "step": 89500 + }, + { + "epoch": 7.75, + "learning_rate": 4.22888330589968e-05, + "loss": 1.3839, + "step": 89510 + }, + { + "epoch": 7.75, + "learning_rate": 4.2287966733084986e-05, + "loss": 1.3672, + "step": 89520 + }, + { + "epoch": 7.75, + "learning_rate": 4.228710040717318e-05, + "loss": 1.4363, + "step": 89530 + }, + { + "epoch": 7.75, + "learning_rate": 4.2286234081261374e-05, + "loss": 1.386, + "step": 89540 + }, + { + "epoch": 7.75, + "learning_rate": 4.228536775534956e-05, + "loss": 1.4222, + "step": 89550 + }, + { + "epoch": 7.75, + "learning_rate": 4.2284501429437757e-05, + "loss": 1.4638, + "step": 89560 + }, + { + "epoch": 7.75, + "learning_rate": 4.228363510352595e-05, + "loss": 1.4164, + "step": 89570 + }, + { + "epoch": 7.75, + "learning_rate": 4.228276877761414e-05, + "loss": 1.3627, + "step": 89580 + }, + { + "epoch": 7.75, + "learning_rate": 4.228190245170233e-05, + "loss": 1.4221, + "step": 89590 + }, + { + "epoch": 7.76, + "learning_rate": 4.228103612579052e-05, + "loss": 1.4462, + "step": 89600 + }, + { + "epoch": 7.76, + "learning_rate": 4.2280169799878715e-05, + "loss": 1.4107, + "step": 89610 + }, + { + "epoch": 7.76, + "learning_rate": 4.227930347396691e-05, + "loss": 1.4214, + "step": 89620 + }, + { + "epoch": 7.76, + "learning_rate": 4.22784371480551e-05, + "loss": 1.3402, + "step": 89630 + }, + { + "epoch": 7.76, + "learning_rate": 4.227757082214329e-05, + "loss": 1.4459, + "step": 89640 + }, + { + "epoch": 7.76, + "learning_rate": 4.2276704496231486e-05, + "loss": 1.4345, + "step": 89650 + }, + { + "epoch": 7.76, + "learning_rate": 4.2275838170319674e-05, + "loss": 1.4317, + "step": 89660 + }, + { + "epoch": 7.76, + "learning_rate": 4.227497184440787e-05, + "loss": 1.473, + "step": 89670 + }, + { + "epoch": 7.76, + "learning_rate": 4.227410551849606e-05, + "loss": 1.389, + "step": 89680 + }, + { + "epoch": 7.76, + "learning_rate": 4.227323919258425e-05, + "loss": 1.4458, + "step": 89690 + }, + { + "epoch": 7.76, + "learning_rate": 4.2272372866672445e-05, + "loss": 1.464, + "step": 89700 + }, + { + "epoch": 7.77, + "learning_rate": 4.227150654076063e-05, + "loss": 1.4372, + "step": 89710 + }, + { + "epoch": 7.77, + "learning_rate": 4.227064021484883e-05, + "loss": 1.4573, + "step": 89720 + }, + { + "epoch": 7.77, + "learning_rate": 4.226977388893702e-05, + "loss": 1.3839, + "step": 89730 + }, + { + "epoch": 7.77, + "learning_rate": 4.226890756302521e-05, + "loss": 1.4381, + "step": 89740 + }, + { + "epoch": 7.77, + "learning_rate": 4.2268041237113404e-05, + "loss": 1.51, + "step": 89750 + }, + { + "epoch": 7.77, + "learning_rate": 4.22671749112016e-05, + "loss": 1.3883, + "step": 89760 + }, + { + "epoch": 7.77, + "learning_rate": 4.2266308585289786e-05, + "loss": 1.4254, + "step": 89770 + }, + { + "epoch": 7.77, + "learning_rate": 4.226544225937798e-05, + "loss": 1.3866, + "step": 89780 + }, + { + "epoch": 7.77, + "learning_rate": 4.2264575933466175e-05, + "loss": 1.3716, + "step": 89790 + }, + { + "epoch": 7.77, + "learning_rate": 4.226370960755436e-05, + "loss": 1.4366, + "step": 89800 + }, + { + "epoch": 7.77, + "learning_rate": 4.226284328164256e-05, + "loss": 1.3808, + "step": 89810 + }, + { + "epoch": 7.77, + "learning_rate": 4.2261976955730745e-05, + "loss": 1.4975, + "step": 89820 + }, + { + "epoch": 7.78, + "learning_rate": 4.226111062981894e-05, + "loss": 1.3999, + "step": 89830 + }, + { + "epoch": 7.78, + "learning_rate": 4.2260244303907133e-05, + "loss": 1.4171, + "step": 89840 + }, + { + "epoch": 7.78, + "learning_rate": 4.225937797799532e-05, + "loss": 1.4012, + "step": 89850 + }, + { + "epoch": 7.78, + "learning_rate": 4.2258511652083516e-05, + "loss": 1.4012, + "step": 89860 + }, + { + "epoch": 7.78, + "learning_rate": 4.225764532617171e-05, + "loss": 1.4791, + "step": 89870 + }, + { + "epoch": 7.78, + "learning_rate": 4.22567790002599e-05, + "loss": 1.3935, + "step": 89880 + }, + { + "epoch": 7.78, + "learning_rate": 4.225591267434809e-05, + "loss": 1.3808, + "step": 89890 + }, + { + "epoch": 7.78, + "learning_rate": 4.225504634843629e-05, + "loss": 1.4248, + "step": 89900 + }, + { + "epoch": 7.78, + "learning_rate": 4.2254180022524474e-05, + "loss": 1.4249, + "step": 89910 + }, + { + "epoch": 7.78, + "learning_rate": 4.225331369661267e-05, + "loss": 1.4086, + "step": 89920 + }, + { + "epoch": 7.78, + "learning_rate": 4.2252447370700856e-05, + "loss": 1.3581, + "step": 89930 + }, + { + "epoch": 7.78, + "learning_rate": 4.225158104478905e-05, + "loss": 1.3733, + "step": 89940 + }, + { + "epoch": 7.79, + "learning_rate": 4.2250714718877245e-05, + "loss": 1.4564, + "step": 89950 + }, + { + "epoch": 7.79, + "learning_rate": 4.224984839296543e-05, + "loss": 1.4174, + "step": 89960 + }, + { + "epoch": 7.79, + "learning_rate": 4.224898206705363e-05, + "loss": 1.4115, + "step": 89970 + }, + { + "epoch": 7.79, + "learning_rate": 4.224811574114182e-05, + "loss": 1.4639, + "step": 89980 + }, + { + "epoch": 7.79, + "learning_rate": 4.224724941523001e-05, + "loss": 1.4088, + "step": 89990 + }, + { + "epoch": 7.79, + "learning_rate": 4.2246383089318204e-05, + "loss": 1.3848, + "step": 90000 + }, + { + "epoch": 7.79, + "learning_rate": 4.22455167634064e-05, + "loss": 1.3899, + "step": 90010 + }, + { + "epoch": 7.79, + "learning_rate": 4.2244650437494586e-05, + "loss": 1.4048, + "step": 90020 + }, + { + "epoch": 7.79, + "learning_rate": 4.224378411158278e-05, + "loss": 1.4029, + "step": 90030 + }, + { + "epoch": 7.79, + "learning_rate": 4.224291778567097e-05, + "loss": 1.4859, + "step": 90040 + }, + { + "epoch": 7.79, + "learning_rate": 4.224205145975916e-05, + "loss": 1.424, + "step": 90050 + }, + { + "epoch": 7.8, + "learning_rate": 4.224118513384736e-05, + "loss": 1.4027, + "step": 90060 + }, + { + "epoch": 7.8, + "learning_rate": 4.2240318807935545e-05, + "loss": 1.4221, + "step": 90070 + }, + { + "epoch": 7.8, + "learning_rate": 4.223945248202374e-05, + "loss": 1.3618, + "step": 90080 + }, + { + "epoch": 7.8, + "learning_rate": 4.2238586156111934e-05, + "loss": 1.3892, + "step": 90090 + }, + { + "epoch": 7.8, + "learning_rate": 4.223771983020012e-05, + "loss": 1.4183, + "step": 90100 + }, + { + "epoch": 7.8, + "learning_rate": 4.2236853504288316e-05, + "loss": 1.4696, + "step": 90110 + }, + { + "epoch": 7.8, + "learning_rate": 4.223598717837651e-05, + "loss": 1.3234, + "step": 90120 + }, + { + "epoch": 7.8, + "learning_rate": 4.22351208524647e-05, + "loss": 1.4946, + "step": 90130 + }, + { + "epoch": 7.8, + "learning_rate": 4.223425452655289e-05, + "loss": 1.4601, + "step": 90140 + }, + { + "epoch": 7.8, + "learning_rate": 4.223338820064108e-05, + "loss": 1.3408, + "step": 90150 + }, + { + "epoch": 7.8, + "learning_rate": 4.2232521874729275e-05, + "loss": 1.3454, + "step": 90160 + }, + { + "epoch": 7.8, + "learning_rate": 4.223165554881747e-05, + "loss": 1.4255, + "step": 90170 + }, + { + "epoch": 7.81, + "learning_rate": 4.223078922290566e-05, + "loss": 1.3746, + "step": 90180 + }, + { + "epoch": 7.81, + "learning_rate": 4.222992289699385e-05, + "loss": 1.3552, + "step": 90190 + }, + { + "epoch": 7.81, + "learning_rate": 4.2229056571082046e-05, + "loss": 1.352, + "step": 90200 + }, + { + "epoch": 7.81, + "learning_rate": 4.222819024517023e-05, + "loss": 1.4713, + "step": 90210 + }, + { + "epoch": 7.81, + "learning_rate": 4.222732391925843e-05, + "loss": 1.4779, + "step": 90220 + }, + { + "epoch": 7.81, + "learning_rate": 4.2226457593346616e-05, + "loss": 1.3806, + "step": 90230 + }, + { + "epoch": 7.81, + "learning_rate": 4.222559126743481e-05, + "loss": 1.3537, + "step": 90240 + }, + { + "epoch": 7.81, + "learning_rate": 4.2224724941523004e-05, + "loss": 1.4167, + "step": 90250 + }, + { + "epoch": 7.81, + "learning_rate": 4.222385861561119e-05, + "loss": 1.4224, + "step": 90260 + }, + { + "epoch": 7.81, + "learning_rate": 4.2222992289699387e-05, + "loss": 1.3788, + "step": 90270 + }, + { + "epoch": 7.81, + "learning_rate": 4.222212596378758e-05, + "loss": 1.3662, + "step": 90280 + }, + { + "epoch": 7.82, + "learning_rate": 4.222125963787577e-05, + "loss": 1.416, + "step": 90290 + }, + { + "epoch": 7.82, + "learning_rate": 4.222039331196396e-05, + "loss": 1.4412, + "step": 90300 + }, + { + "epoch": 7.82, + "learning_rate": 4.221952698605216e-05, + "loss": 1.4364, + "step": 90310 + }, + { + "epoch": 7.82, + "learning_rate": 4.2218660660140345e-05, + "loss": 1.4356, + "step": 90320 + }, + { + "epoch": 7.82, + "learning_rate": 4.221779433422854e-05, + "loss": 1.3527, + "step": 90330 + }, + { + "epoch": 7.82, + "learning_rate": 4.221692800831673e-05, + "loss": 1.3802, + "step": 90340 + }, + { + "epoch": 7.82, + "learning_rate": 4.221606168240492e-05, + "loss": 1.5108, + "step": 90350 + }, + { + "epoch": 7.82, + "learning_rate": 4.2215195356493116e-05, + "loss": 1.3353, + "step": 90360 + }, + { + "epoch": 7.82, + "learning_rate": 4.2214329030581304e-05, + "loss": 1.3923, + "step": 90370 + }, + { + "epoch": 7.82, + "learning_rate": 4.22134627046695e-05, + "loss": 1.365, + "step": 90380 + }, + { + "epoch": 7.82, + "learning_rate": 4.221259637875769e-05, + "loss": 1.3874, + "step": 90390 + }, + { + "epoch": 7.82, + "learning_rate": 4.221173005284588e-05, + "loss": 1.3999, + "step": 90400 + }, + { + "epoch": 7.83, + "learning_rate": 4.2210863726934075e-05, + "loss": 1.3904, + "step": 90410 + }, + { + "epoch": 7.83, + "learning_rate": 4.220999740102227e-05, + "loss": 1.3351, + "step": 90420 + }, + { + "epoch": 7.83, + "learning_rate": 4.220913107511046e-05, + "loss": 1.4913, + "step": 90430 + }, + { + "epoch": 7.83, + "learning_rate": 4.220826474919865e-05, + "loss": 1.4698, + "step": 90440 + }, + { + "epoch": 7.83, + "learning_rate": 4.220739842328684e-05, + "loss": 1.387, + "step": 90450 + }, + { + "epoch": 7.83, + "learning_rate": 4.2206532097375034e-05, + "loss": 1.4149, + "step": 90460 + }, + { + "epoch": 7.83, + "learning_rate": 4.220566577146323e-05, + "loss": 1.416, + "step": 90470 + }, + { + "epoch": 7.83, + "learning_rate": 4.2204799445551416e-05, + "loss": 1.3857, + "step": 90480 + }, + { + "epoch": 7.83, + "learning_rate": 4.220393311963961e-05, + "loss": 1.4045, + "step": 90490 + }, + { + "epoch": 7.83, + "learning_rate": 4.2203066793727805e-05, + "loss": 1.4016, + "step": 90500 + }, + { + "epoch": 7.83, + "learning_rate": 4.220220046781599e-05, + "loss": 1.3896, + "step": 90510 + }, + { + "epoch": 7.84, + "learning_rate": 4.220133414190419e-05, + "loss": 1.4164, + "step": 90520 + }, + { + "epoch": 7.84, + "learning_rate": 4.220046781599238e-05, + "loss": 1.4198, + "step": 90530 + }, + { + "epoch": 7.84, + "learning_rate": 4.219960149008057e-05, + "loss": 1.3671, + "step": 90540 + }, + { + "epoch": 7.84, + "learning_rate": 4.2198735164168763e-05, + "loss": 1.4191, + "step": 90550 + }, + { + "epoch": 7.84, + "learning_rate": 4.219786883825695e-05, + "loss": 1.4197, + "step": 90560 + }, + { + "epoch": 7.84, + "learning_rate": 4.2197002512345146e-05, + "loss": 1.5482, + "step": 90570 + }, + { + "epoch": 7.84, + "learning_rate": 4.219613618643334e-05, + "loss": 1.3851, + "step": 90580 + }, + { + "epoch": 7.84, + "learning_rate": 4.219526986052153e-05, + "loss": 1.4574, + "step": 90590 + }, + { + "epoch": 7.84, + "learning_rate": 4.219440353460972e-05, + "loss": 1.3997, + "step": 90600 + }, + { + "epoch": 7.84, + "learning_rate": 4.219353720869792e-05, + "loss": 1.3684, + "step": 90610 + }, + { + "epoch": 7.84, + "learning_rate": 4.2192670882786104e-05, + "loss": 1.393, + "step": 90620 + }, + { + "epoch": 7.84, + "learning_rate": 4.21918045568743e-05, + "loss": 1.3554, + "step": 90630 + }, + { + "epoch": 7.85, + "learning_rate": 4.219093823096249e-05, + "loss": 1.437, + "step": 90640 + }, + { + "epoch": 7.85, + "learning_rate": 4.219007190505068e-05, + "loss": 1.3607, + "step": 90650 + }, + { + "epoch": 7.85, + "learning_rate": 4.2189205579138875e-05, + "loss": 1.4512, + "step": 90660 + }, + { + "epoch": 7.85, + "learning_rate": 4.218833925322706e-05, + "loss": 1.4413, + "step": 90670 + }, + { + "epoch": 7.85, + "learning_rate": 4.218747292731526e-05, + "loss": 1.3134, + "step": 90680 + }, + { + "epoch": 7.85, + "learning_rate": 4.218660660140345e-05, + "loss": 1.3435, + "step": 90690 + }, + { + "epoch": 7.85, + "learning_rate": 4.218574027549164e-05, + "loss": 1.4132, + "step": 90700 + }, + { + "epoch": 7.85, + "learning_rate": 4.2184873949579834e-05, + "loss": 1.4055, + "step": 90710 + }, + { + "epoch": 7.85, + "learning_rate": 4.218400762366803e-05, + "loss": 1.4363, + "step": 90720 + }, + { + "epoch": 7.85, + "learning_rate": 4.2183141297756216e-05, + "loss": 1.417, + "step": 90730 + }, + { + "epoch": 7.85, + "learning_rate": 4.218227497184441e-05, + "loss": 1.437, + "step": 90740 + }, + { + "epoch": 7.86, + "learning_rate": 4.2181408645932605e-05, + "loss": 1.4058, + "step": 90750 + }, + { + "epoch": 7.86, + "learning_rate": 4.218054232002079e-05, + "loss": 1.4746, + "step": 90760 + }, + { + "epoch": 7.86, + "learning_rate": 4.217967599410899e-05, + "loss": 1.3617, + "step": 90770 + }, + { + "epoch": 7.86, + "learning_rate": 4.2178809668197175e-05, + "loss": 1.4193, + "step": 90780 + }, + { + "epoch": 7.86, + "learning_rate": 4.217794334228537e-05, + "loss": 1.3526, + "step": 90790 + }, + { + "epoch": 7.86, + "learning_rate": 4.2177077016373564e-05, + "loss": 1.4022, + "step": 90800 + }, + { + "epoch": 7.86, + "learning_rate": 4.217621069046175e-05, + "loss": 1.4105, + "step": 90810 + }, + { + "epoch": 7.86, + "learning_rate": 4.2175344364549946e-05, + "loss": 1.4069, + "step": 90820 + }, + { + "epoch": 7.86, + "learning_rate": 4.217447803863814e-05, + "loss": 1.3683, + "step": 90830 + }, + { + "epoch": 7.86, + "learning_rate": 4.217361171272633e-05, + "loss": 1.3316, + "step": 90840 + }, + { + "epoch": 7.86, + "learning_rate": 4.217274538681452e-05, + "loss": 1.4113, + "step": 90850 + }, + { + "epoch": 7.86, + "learning_rate": 4.217187906090272e-05, + "loss": 1.4135, + "step": 90860 + }, + { + "epoch": 7.87, + "learning_rate": 4.2171012734990905e-05, + "loss": 1.4345, + "step": 90870 + }, + { + "epoch": 7.87, + "learning_rate": 4.21701464090791e-05, + "loss": 1.4618, + "step": 90880 + }, + { + "epoch": 7.87, + "learning_rate": 4.216928008316729e-05, + "loss": 1.384, + "step": 90890 + }, + { + "epoch": 7.87, + "learning_rate": 4.216841375725548e-05, + "loss": 1.3808, + "step": 90900 + }, + { + "epoch": 7.87, + "learning_rate": 4.2167547431343676e-05, + "loss": 1.3729, + "step": 90910 + }, + { + "epoch": 7.87, + "learning_rate": 4.216668110543186e-05, + "loss": 1.4248, + "step": 90920 + }, + { + "epoch": 7.87, + "learning_rate": 4.216581477952006e-05, + "loss": 1.4021, + "step": 90930 + }, + { + "epoch": 7.87, + "learning_rate": 4.216494845360825e-05, + "loss": 1.414, + "step": 90940 + }, + { + "epoch": 7.87, + "learning_rate": 4.216408212769644e-05, + "loss": 1.3233, + "step": 90950 + }, + { + "epoch": 7.87, + "learning_rate": 4.2163215801784634e-05, + "loss": 1.3368, + "step": 90960 + }, + { + "epoch": 7.87, + "learning_rate": 4.216234947587283e-05, + "loss": 1.4475, + "step": 90970 + }, + { + "epoch": 7.87, + "learning_rate": 4.2161483149961017e-05, + "loss": 1.4601, + "step": 90980 + }, + { + "epoch": 7.88, + "learning_rate": 4.216061682404921e-05, + "loss": 1.3904, + "step": 90990 + }, + { + "epoch": 7.88, + "learning_rate": 4.21597504981374e-05, + "loss": 1.3429, + "step": 91000 + }, + { + "epoch": 7.88, + "learning_rate": 4.215888417222559e-05, + "loss": 1.3373, + "step": 91010 + }, + { + "epoch": 7.88, + "learning_rate": 4.215801784631379e-05, + "loss": 1.4029, + "step": 91020 + }, + { + "epoch": 7.88, + "learning_rate": 4.2157151520401975e-05, + "loss": 1.3886, + "step": 91030 + }, + { + "epoch": 7.88, + "learning_rate": 4.215628519449017e-05, + "loss": 1.42, + "step": 91040 + }, + { + "epoch": 7.88, + "learning_rate": 4.2155418868578364e-05, + "loss": 1.4227, + "step": 91050 + }, + { + "epoch": 7.88, + "learning_rate": 4.215455254266655e-05, + "loss": 1.3019, + "step": 91060 + }, + { + "epoch": 7.88, + "learning_rate": 4.2153686216754746e-05, + "loss": 1.3739, + "step": 91070 + }, + { + "epoch": 7.88, + "learning_rate": 4.2152819890842934e-05, + "loss": 1.4175, + "step": 91080 + }, + { + "epoch": 7.88, + "learning_rate": 4.215195356493113e-05, + "loss": 1.4449, + "step": 91090 + }, + { + "epoch": 7.89, + "learning_rate": 4.215108723901932e-05, + "loss": 1.3815, + "step": 91100 + }, + { + "epoch": 7.89, + "learning_rate": 4.215022091310751e-05, + "loss": 1.3641, + "step": 91110 + }, + { + "epoch": 7.89, + "learning_rate": 4.2149354587195705e-05, + "loss": 1.3814, + "step": 91120 + }, + { + "epoch": 7.89, + "learning_rate": 4.21484882612839e-05, + "loss": 1.4608, + "step": 91130 + }, + { + "epoch": 7.89, + "learning_rate": 4.214762193537209e-05, + "loss": 1.3085, + "step": 91140 + }, + { + "epoch": 7.89, + "learning_rate": 4.214675560946028e-05, + "loss": 1.3797, + "step": 91150 + }, + { + "epoch": 7.89, + "learning_rate": 4.2145889283548476e-05, + "loss": 1.406, + "step": 91160 + }, + { + "epoch": 7.89, + "learning_rate": 4.2145022957636664e-05, + "loss": 1.4759, + "step": 91170 + }, + { + "epoch": 7.89, + "learning_rate": 4.214415663172486e-05, + "loss": 1.365, + "step": 91180 + }, + { + "epoch": 7.89, + "learning_rate": 4.2143290305813046e-05, + "loss": 1.383, + "step": 91190 + }, + { + "epoch": 7.89, + "learning_rate": 4.214242397990124e-05, + "loss": 1.3791, + "step": 91200 + }, + { + "epoch": 7.89, + "learning_rate": 4.2141557653989435e-05, + "loss": 1.4371, + "step": 91210 + }, + { + "epoch": 7.9, + "learning_rate": 4.214069132807762e-05, + "loss": 1.3758, + "step": 91220 + }, + { + "epoch": 7.9, + "learning_rate": 4.213982500216582e-05, + "loss": 1.3972, + "step": 91230 + }, + { + "epoch": 7.9, + "learning_rate": 4.213895867625401e-05, + "loss": 1.3504, + "step": 91240 + }, + { + "epoch": 7.9, + "learning_rate": 4.21380923503422e-05, + "loss": 1.3603, + "step": 91250 + }, + { + "epoch": 7.9, + "learning_rate": 4.2137226024430393e-05, + "loss": 1.4241, + "step": 91260 + }, + { + "epoch": 7.9, + "learning_rate": 4.213635969851859e-05, + "loss": 1.4214, + "step": 91270 + }, + { + "epoch": 7.9, + "learning_rate": 4.2135493372606776e-05, + "loss": 1.3925, + "step": 91280 + }, + { + "epoch": 7.9, + "learning_rate": 4.213462704669497e-05, + "loss": 1.4499, + "step": 91290 + }, + { + "epoch": 7.9, + "learning_rate": 4.213376072078316e-05, + "loss": 1.372, + "step": 91300 + }, + { + "epoch": 7.9, + "learning_rate": 4.213289439487135e-05, + "loss": 1.4289, + "step": 91310 + }, + { + "epoch": 7.9, + "learning_rate": 4.2132028068959547e-05, + "loss": 1.3378, + "step": 91320 + }, + { + "epoch": 7.91, + "learning_rate": 4.2131161743047734e-05, + "loss": 1.4624, + "step": 91330 + }, + { + "epoch": 7.91, + "learning_rate": 4.213029541713593e-05, + "loss": 1.3778, + "step": 91340 + }, + { + "epoch": 7.91, + "learning_rate": 4.212942909122412e-05, + "loss": 1.376, + "step": 91350 + }, + { + "epoch": 7.91, + "learning_rate": 4.212856276531231e-05, + "loss": 1.4297, + "step": 91360 + }, + { + "epoch": 7.91, + "learning_rate": 4.2127696439400505e-05, + "loss": 1.4289, + "step": 91370 + }, + { + "epoch": 7.91, + "learning_rate": 4.21268301134887e-05, + "loss": 1.4425, + "step": 91380 + }, + { + "epoch": 7.91, + "learning_rate": 4.212596378757689e-05, + "loss": 1.4465, + "step": 91390 + }, + { + "epoch": 7.91, + "learning_rate": 4.212509746166508e-05, + "loss": 1.4107, + "step": 91400 + }, + { + "epoch": 7.91, + "learning_rate": 4.212423113575327e-05, + "loss": 1.3491, + "step": 91410 + }, + { + "epoch": 7.91, + "learning_rate": 4.2123364809841464e-05, + "loss": 1.3951, + "step": 91420 + }, + { + "epoch": 7.91, + "learning_rate": 4.212249848392966e-05, + "loss": 1.4326, + "step": 91430 + }, + { + "epoch": 7.91, + "learning_rate": 4.2121632158017846e-05, + "loss": 1.4031, + "step": 91440 + }, + { + "epoch": 7.92, + "learning_rate": 4.212076583210604e-05, + "loss": 1.4654, + "step": 91450 + }, + { + "epoch": 7.92, + "learning_rate": 4.2119899506194235e-05, + "loss": 1.3853, + "step": 91460 + }, + { + "epoch": 7.92, + "learning_rate": 4.211903318028242e-05, + "loss": 1.4038, + "step": 91470 + }, + { + "epoch": 7.92, + "learning_rate": 4.211816685437062e-05, + "loss": 1.3146, + "step": 91480 + }, + { + "epoch": 7.92, + "learning_rate": 4.211730052845881e-05, + "loss": 1.4318, + "step": 91490 + }, + { + "epoch": 7.92, + "learning_rate": 4.2116434202547e-05, + "loss": 1.3573, + "step": 91500 + }, + { + "epoch": 7.92, + "learning_rate": 4.2115567876635194e-05, + "loss": 1.4023, + "step": 91510 + }, + { + "epoch": 7.92, + "learning_rate": 4.211470155072338e-05, + "loss": 1.4079, + "step": 91520 + }, + { + "epoch": 7.92, + "learning_rate": 4.2113835224811576e-05, + "loss": 1.4089, + "step": 91530 + }, + { + "epoch": 7.92, + "learning_rate": 4.211296889889977e-05, + "loss": 1.429, + "step": 91540 + }, + { + "epoch": 7.92, + "learning_rate": 4.211210257298796e-05, + "loss": 1.4087, + "step": 91550 + }, + { + "epoch": 7.93, + "learning_rate": 4.211123624707615e-05, + "loss": 1.4822, + "step": 91560 + }, + { + "epoch": 7.93, + "learning_rate": 4.211036992116435e-05, + "loss": 1.4375, + "step": 91570 + }, + { + "epoch": 7.93, + "learning_rate": 4.2109503595252535e-05, + "loss": 1.3534, + "step": 91580 + }, + { + "epoch": 7.93, + "learning_rate": 4.210863726934073e-05, + "loss": 1.3929, + "step": 91590 + }, + { + "epoch": 7.93, + "learning_rate": 4.2107770943428924e-05, + "loss": 1.3394, + "step": 91600 + }, + { + "epoch": 7.93, + "learning_rate": 4.210690461751711e-05, + "loss": 1.399, + "step": 91610 + }, + { + "epoch": 7.93, + "learning_rate": 4.2106038291605306e-05, + "loss": 1.3814, + "step": 91620 + }, + { + "epoch": 7.93, + "learning_rate": 4.210517196569349e-05, + "loss": 1.3599, + "step": 91630 + }, + { + "epoch": 7.93, + "learning_rate": 4.210430563978169e-05, + "loss": 1.4391, + "step": 91640 + }, + { + "epoch": 7.93, + "learning_rate": 4.210343931386988e-05, + "loss": 1.3618, + "step": 91650 + }, + { + "epoch": 7.93, + "learning_rate": 4.210257298795807e-05, + "loss": 1.3686, + "step": 91660 + }, + { + "epoch": 7.93, + "learning_rate": 4.2101706662046264e-05, + "loss": 1.3879, + "step": 91670 + }, + { + "epoch": 7.94, + "learning_rate": 4.210084033613446e-05, + "loss": 1.5007, + "step": 91680 + }, + { + "epoch": 7.94, + "learning_rate": 4.2099974010222646e-05, + "loss": 1.3727, + "step": 91690 + }, + { + "epoch": 7.94, + "learning_rate": 4.209910768431084e-05, + "loss": 1.3669, + "step": 91700 + }, + { + "epoch": 7.94, + "learning_rate": 4.2098241358399035e-05, + "loss": 1.4662, + "step": 91710 + }, + { + "epoch": 7.94, + "learning_rate": 4.209737503248722e-05, + "loss": 1.3598, + "step": 91720 + }, + { + "epoch": 7.94, + "learning_rate": 4.209650870657542e-05, + "loss": 1.4828, + "step": 91730 + }, + { + "epoch": 7.94, + "learning_rate": 4.2095642380663605e-05, + "loss": 1.3729, + "step": 91740 + }, + { + "epoch": 7.94, + "learning_rate": 4.20947760547518e-05, + "loss": 1.3897, + "step": 91750 + }, + { + "epoch": 7.94, + "learning_rate": 4.2093909728839994e-05, + "loss": 1.432, + "step": 91760 + }, + { + "epoch": 7.94, + "learning_rate": 4.209304340292818e-05, + "loss": 1.4236, + "step": 91770 + }, + { + "epoch": 7.94, + "learning_rate": 4.2092177077016376e-05, + "loss": 1.3978, + "step": 91780 + }, + { + "epoch": 7.95, + "learning_rate": 4.209131075110457e-05, + "loss": 1.4163, + "step": 91790 + }, + { + "epoch": 7.95, + "learning_rate": 4.209044442519276e-05, + "loss": 1.4408, + "step": 91800 + }, + { + "epoch": 7.95, + "learning_rate": 4.208957809928095e-05, + "loss": 1.3594, + "step": 91810 + }, + { + "epoch": 7.95, + "learning_rate": 4.208871177336914e-05, + "loss": 1.4214, + "step": 91820 + }, + { + "epoch": 7.95, + "learning_rate": 4.2087845447457335e-05, + "loss": 1.3582, + "step": 91830 + }, + { + "epoch": 7.95, + "learning_rate": 4.208697912154553e-05, + "loss": 1.404, + "step": 91840 + }, + { + "epoch": 7.95, + "learning_rate": 4.208611279563372e-05, + "loss": 1.451, + "step": 91850 + }, + { + "epoch": 7.95, + "learning_rate": 4.208524646972191e-05, + "loss": 1.4293, + "step": 91860 + }, + { + "epoch": 7.95, + "learning_rate": 4.2084380143810106e-05, + "loss": 1.4241, + "step": 91870 + }, + { + "epoch": 7.95, + "learning_rate": 4.2083513817898294e-05, + "loss": 1.4039, + "step": 91880 + }, + { + "epoch": 7.95, + "learning_rate": 4.208264749198649e-05, + "loss": 1.4275, + "step": 91890 + }, + { + "epoch": 7.95, + "learning_rate": 4.208178116607468e-05, + "loss": 1.4158, + "step": 91900 + }, + { + "epoch": 7.96, + "learning_rate": 4.208091484016287e-05, + "loss": 1.4182, + "step": 91910 + }, + { + "epoch": 7.96, + "learning_rate": 4.2080048514251065e-05, + "loss": 1.3972, + "step": 91920 + }, + { + "epoch": 7.96, + "learning_rate": 4.207918218833925e-05, + "loss": 1.4559, + "step": 91930 + }, + { + "epoch": 7.96, + "learning_rate": 4.207831586242745e-05, + "loss": 1.4232, + "step": 91940 + }, + { + "epoch": 7.96, + "learning_rate": 4.207744953651564e-05, + "loss": 1.4214, + "step": 91950 + }, + { + "epoch": 7.96, + "learning_rate": 4.207658321060383e-05, + "loss": 1.3728, + "step": 91960 + }, + { + "epoch": 7.96, + "learning_rate": 4.2075716884692023e-05, + "loss": 1.43, + "step": 91970 + }, + { + "epoch": 7.96, + "learning_rate": 4.207485055878022e-05, + "loss": 1.4003, + "step": 91980 + }, + { + "epoch": 7.96, + "learning_rate": 4.2073984232868406e-05, + "loss": 1.3851, + "step": 91990 + }, + { + "epoch": 7.96, + "learning_rate": 4.20731179069566e-05, + "loss": 1.4332, + "step": 92000 + }, + { + "epoch": 7.96, + "learning_rate": 4.2072251581044794e-05, + "loss": 1.3251, + "step": 92010 + }, + { + "epoch": 7.96, + "learning_rate": 4.207138525513298e-05, + "loss": 1.3663, + "step": 92020 + }, + { + "epoch": 7.97, + "learning_rate": 4.2070518929221177e-05, + "loss": 1.4145, + "step": 92030 + }, + { + "epoch": 7.97, + "learning_rate": 4.2069652603309364e-05, + "loss": 1.3938, + "step": 92040 + }, + { + "epoch": 7.97, + "learning_rate": 4.206878627739756e-05, + "loss": 1.4238, + "step": 92050 + }, + { + "epoch": 7.97, + "learning_rate": 4.206791995148575e-05, + "loss": 1.4073, + "step": 92060 + }, + { + "epoch": 7.97, + "learning_rate": 4.206705362557394e-05, + "loss": 1.3957, + "step": 92070 + }, + { + "epoch": 7.97, + "learning_rate": 4.2066187299662135e-05, + "loss": 1.3513, + "step": 92080 + }, + { + "epoch": 7.97, + "learning_rate": 4.206532097375033e-05, + "loss": 1.469, + "step": 92090 + }, + { + "epoch": 7.97, + "learning_rate": 4.206445464783852e-05, + "loss": 1.3977, + "step": 92100 + }, + { + "epoch": 7.97, + "learning_rate": 4.206358832192671e-05, + "loss": 1.3732, + "step": 92110 + }, + { + "epoch": 7.97, + "learning_rate": 4.2062721996014906e-05, + "loss": 1.3389, + "step": 92120 + }, + { + "epoch": 7.97, + "learning_rate": 4.2061855670103094e-05, + "loss": 1.3735, + "step": 92130 + }, + { + "epoch": 7.98, + "learning_rate": 4.206098934419129e-05, + "loss": 1.3731, + "step": 92140 + }, + { + "epoch": 7.98, + "learning_rate": 4.2060123018279476e-05, + "loss": 1.4851, + "step": 92150 + }, + { + "epoch": 7.98, + "learning_rate": 4.205925669236767e-05, + "loss": 1.3971, + "step": 92160 + }, + { + "epoch": 7.98, + "learning_rate": 4.2058390366455865e-05, + "loss": 1.4045, + "step": 92170 + }, + { + "epoch": 7.98, + "learning_rate": 4.205752404054405e-05, + "loss": 1.4495, + "step": 92180 + }, + { + "epoch": 7.98, + "learning_rate": 4.205665771463225e-05, + "loss": 1.3464, + "step": 92190 + }, + { + "epoch": 7.98, + "learning_rate": 4.205579138872044e-05, + "loss": 1.3424, + "step": 92200 + }, + { + "epoch": 7.98, + "learning_rate": 4.205492506280863e-05, + "loss": 1.4094, + "step": 92210 + }, + { + "epoch": 7.98, + "learning_rate": 4.2054058736896824e-05, + "loss": 1.4452, + "step": 92220 + }, + { + "epoch": 7.98, + "learning_rate": 4.205319241098502e-05, + "loss": 1.3911, + "step": 92230 + }, + { + "epoch": 7.98, + "learning_rate": 4.2052326085073206e-05, + "loss": 1.5221, + "step": 92240 + }, + { + "epoch": 7.98, + "learning_rate": 4.20514597591614e-05, + "loss": 1.4552, + "step": 92250 + }, + { + "epoch": 7.99, + "learning_rate": 4.205059343324959e-05, + "loss": 1.3293, + "step": 92260 + }, + { + "epoch": 7.99, + "learning_rate": 4.204972710733778e-05, + "loss": 1.3764, + "step": 92270 + }, + { + "epoch": 7.99, + "learning_rate": 4.204886078142598e-05, + "loss": 1.3861, + "step": 92280 + }, + { + "epoch": 7.99, + "learning_rate": 4.2047994455514165e-05, + "loss": 1.3946, + "step": 92290 + }, + { + "epoch": 7.99, + "learning_rate": 4.204712812960236e-05, + "loss": 1.4517, + "step": 92300 + }, + { + "epoch": 7.99, + "learning_rate": 4.2046261803690553e-05, + "loss": 1.4704, + "step": 92310 + }, + { + "epoch": 7.99, + "learning_rate": 4.204539547777874e-05, + "loss": 1.4174, + "step": 92320 + }, + { + "epoch": 7.99, + "learning_rate": 4.2044529151866936e-05, + "loss": 1.3704, + "step": 92330 + }, + { + "epoch": 7.99, + "learning_rate": 4.204366282595513e-05, + "loss": 1.4821, + "step": 92340 + }, + { + "epoch": 7.99, + "learning_rate": 4.204279650004332e-05, + "loss": 1.4583, + "step": 92350 + }, + { + "epoch": 7.99, + "learning_rate": 4.204193017413151e-05, + "loss": 1.4416, + "step": 92360 + }, + { + "epoch": 8.0, + "learning_rate": 4.20410638482197e-05, + "loss": 1.3788, + "step": 92370 + }, + { + "epoch": 8.0, + "learning_rate": 4.2040197522307894e-05, + "loss": 1.3929, + "step": 92380 + }, + { + "epoch": 8.0, + "learning_rate": 4.203933119639609e-05, + "loss": 1.3666, + "step": 92390 + }, + { + "epoch": 8.0, + "learning_rate": 4.2038464870484276e-05, + "loss": 1.4202, + "step": 92400 + }, + { + "epoch": 8.0, + "learning_rate": 4.203759854457247e-05, + "loss": 1.4732, + "step": 92410 + }, + { + "epoch": 8.0, + "learning_rate": 4.2036732218660665e-05, + "loss": 1.3565, + "step": 92420 + }, + { + "epoch": 8.0, + "eval_Bleu_1": 0.031380603679227875, + "eval_Bleu_2": 2.484069930055871e-11, + "eval_Bleu_3": 2.3804624460376584e-14, + "eval_Bleu_4": 7.59050961680482e-16, + "eval_ROUGE_L": 0.07114991643574073, + "eval_cer": 0.994147227046788, + "eval_em": 0, + "eval_f1": 0.07318070432492041, + "eval_loss": 1.3043513298034668, + "eval_runtime": 2018.2157, + "eval_samples_per_second": 2.544, + "eval_steps_per_second": 2.544, + "eval_wer": 0.9783114682861521, + "step": 92425 + }, + { + "epoch": 8.0, + "learning_rate": 4.203586589274885e-05, + "loss": 1.4494, + "step": 92430 + }, + { + "epoch": 8.0, + "learning_rate": 4.203499956683705e-05, + "loss": 1.3628, + "step": 92440 + }, + { + "epoch": 8.0, + "learning_rate": 4.203413324092524e-05, + "loss": 1.3928, + "step": 92450 + }, + { + "epoch": 8.0, + "learning_rate": 4.203326691501343e-05, + "loss": 1.3941, + "step": 92460 + }, + { + "epoch": 8.0, + "learning_rate": 4.2032400589101624e-05, + "loss": 1.3014, + "step": 92470 + }, + { + "epoch": 8.0, + "learning_rate": 4.203153426318981e-05, + "loss": 1.3343, + "step": 92480 + }, + { + "epoch": 8.01, + "learning_rate": 4.2030667937278006e-05, + "loss": 1.385, + "step": 92490 + }, + { + "epoch": 8.01, + "learning_rate": 4.20298016113662e-05, + "loss": 1.36, + "step": 92500 + }, + { + "epoch": 8.01, + "learning_rate": 4.202893528545439e-05, + "loss": 1.3596, + "step": 92510 + }, + { + "epoch": 8.01, + "learning_rate": 4.202806895954258e-05, + "loss": 1.3599, + "step": 92520 + }, + { + "epoch": 8.01, + "learning_rate": 4.202720263363078e-05, + "loss": 1.3682, + "step": 92530 + }, + { + "epoch": 8.01, + "learning_rate": 4.2026336307718965e-05, + "loss": 1.3843, + "step": 92540 + }, + { + "epoch": 8.01, + "learning_rate": 4.202546998180716e-05, + "loss": 1.4136, + "step": 92550 + }, + { + "epoch": 8.01, + "learning_rate": 4.202460365589535e-05, + "loss": 1.394, + "step": 92560 + }, + { + "epoch": 8.01, + "learning_rate": 4.202373732998354e-05, + "loss": 1.3394, + "step": 92570 + }, + { + "epoch": 8.01, + "learning_rate": 4.2022871004071736e-05, + "loss": 1.3502, + "step": 92580 + }, + { + "epoch": 8.01, + "learning_rate": 4.2022004678159924e-05, + "loss": 1.3375, + "step": 92590 + }, + { + "epoch": 8.02, + "learning_rate": 4.202113835224812e-05, + "loss": 1.3064, + "step": 92600 + }, + { + "epoch": 8.02, + "learning_rate": 4.202027202633631e-05, + "loss": 1.3522, + "step": 92610 + }, + { + "epoch": 8.02, + "learning_rate": 4.20194057004245e-05, + "loss": 1.3685, + "step": 92620 + }, + { + "epoch": 8.02, + "learning_rate": 4.2018539374512695e-05, + "loss": 1.4126, + "step": 92630 + }, + { + "epoch": 8.02, + "learning_rate": 4.201767304860089e-05, + "loss": 1.3639, + "step": 92640 + }, + { + "epoch": 8.02, + "learning_rate": 4.201680672268908e-05, + "loss": 1.3373, + "step": 92650 + }, + { + "epoch": 8.02, + "learning_rate": 4.201594039677727e-05, + "loss": 1.3789, + "step": 92660 + }, + { + "epoch": 8.02, + "learning_rate": 4.201507407086546e-05, + "loss": 1.3596, + "step": 92670 + }, + { + "epoch": 8.02, + "learning_rate": 4.201420774495365e-05, + "loss": 1.3315, + "step": 92680 + }, + { + "epoch": 8.02, + "learning_rate": 4.201334141904185e-05, + "loss": 1.3928, + "step": 92690 + }, + { + "epoch": 8.02, + "learning_rate": 4.2012475093130036e-05, + "loss": 1.3544, + "step": 92700 + }, + { + "epoch": 8.02, + "learning_rate": 4.201160876721823e-05, + "loss": 1.4408, + "step": 92710 + }, + { + "epoch": 8.03, + "learning_rate": 4.2010742441306424e-05, + "loss": 1.2842, + "step": 92720 + }, + { + "epoch": 8.03, + "learning_rate": 4.200987611539461e-05, + "loss": 1.3601, + "step": 92730 + }, + { + "epoch": 8.03, + "learning_rate": 4.2009009789482807e-05, + "loss": 1.3875, + "step": 92740 + }, + { + "epoch": 8.03, + "learning_rate": 4.2008143463571e-05, + "loss": 1.3544, + "step": 92750 + }, + { + "epoch": 8.03, + "learning_rate": 4.200727713765919e-05, + "loss": 1.3851, + "step": 92760 + }, + { + "epoch": 8.03, + "learning_rate": 4.200641081174738e-05, + "loss": 1.4021, + "step": 92770 + }, + { + "epoch": 8.03, + "learning_rate": 4.200554448583557e-05, + "loss": 1.3492, + "step": 92780 + }, + { + "epoch": 8.03, + "learning_rate": 4.2004678159923765e-05, + "loss": 1.3543, + "step": 92790 + }, + { + "epoch": 8.03, + "learning_rate": 4.200381183401196e-05, + "loss": 1.3751, + "step": 92800 + }, + { + "epoch": 8.03, + "learning_rate": 4.200294550810015e-05, + "loss": 1.3513, + "step": 92810 + }, + { + "epoch": 8.03, + "learning_rate": 4.200207918218834e-05, + "loss": 1.3559, + "step": 92820 + }, + { + "epoch": 8.04, + "learning_rate": 4.2001212856276536e-05, + "loss": 1.3331, + "step": 92830 + }, + { + "epoch": 8.04, + "learning_rate": 4.2000346530364724e-05, + "loss": 1.3989, + "step": 92840 + }, + { + "epoch": 8.04, + "learning_rate": 4.199948020445292e-05, + "loss": 1.3643, + "step": 92850 + }, + { + "epoch": 8.04, + "learning_rate": 4.199861387854111e-05, + "loss": 1.4135, + "step": 92860 + }, + { + "epoch": 8.04, + "learning_rate": 4.19977475526293e-05, + "loss": 1.4178, + "step": 92870 + }, + { + "epoch": 8.04, + "learning_rate": 4.1996881226717495e-05, + "loss": 1.3539, + "step": 92880 + }, + { + "epoch": 8.04, + "learning_rate": 4.199601490080568e-05, + "loss": 1.3621, + "step": 92890 + }, + { + "epoch": 8.04, + "learning_rate": 4.199514857489388e-05, + "loss": 1.4241, + "step": 92900 + }, + { + "epoch": 8.04, + "learning_rate": 4.199428224898207e-05, + "loss": 1.3172, + "step": 92910 + }, + { + "epoch": 8.04, + "learning_rate": 4.199341592307026e-05, + "loss": 1.3876, + "step": 92920 + }, + { + "epoch": 8.04, + "learning_rate": 4.1992549597158454e-05, + "loss": 1.3246, + "step": 92930 + }, + { + "epoch": 8.04, + "learning_rate": 4.199168327124665e-05, + "loss": 1.3285, + "step": 92940 + }, + { + "epoch": 8.05, + "learning_rate": 4.1990816945334836e-05, + "loss": 1.3786, + "step": 92950 + }, + { + "epoch": 8.05, + "learning_rate": 4.198995061942303e-05, + "loss": 1.3793, + "step": 92960 + }, + { + "epoch": 8.05, + "learning_rate": 4.1989084293511225e-05, + "loss": 1.4105, + "step": 92970 + }, + { + "epoch": 8.05, + "learning_rate": 4.198821796759941e-05, + "loss": 1.3469, + "step": 92980 + }, + { + "epoch": 8.05, + "learning_rate": 4.198735164168761e-05, + "loss": 1.3594, + "step": 92990 + }, + { + "epoch": 8.05, + "learning_rate": 4.1986485315775795e-05, + "loss": 1.3679, + "step": 93000 + }, + { + "epoch": 8.05, + "learning_rate": 4.198561898986399e-05, + "loss": 1.3356, + "step": 93010 + }, + { + "epoch": 8.05, + "learning_rate": 4.1984752663952183e-05, + "loss": 1.3891, + "step": 93020 + }, + { + "epoch": 8.05, + "learning_rate": 4.198388633804037e-05, + "loss": 1.3867, + "step": 93030 + }, + { + "epoch": 8.05, + "learning_rate": 4.1983020012128566e-05, + "loss": 1.3327, + "step": 93040 + }, + { + "epoch": 8.05, + "learning_rate": 4.198215368621676e-05, + "loss": 1.3373, + "step": 93050 + }, + { + "epoch": 8.05, + "learning_rate": 4.198128736030495e-05, + "loss": 1.3908, + "step": 93060 + }, + { + "epoch": 8.06, + "learning_rate": 4.198042103439314e-05, + "loss": 1.3558, + "step": 93070 + }, + { + "epoch": 8.06, + "learning_rate": 4.197955470848134e-05, + "loss": 1.3545, + "step": 93080 + }, + { + "epoch": 8.06, + "learning_rate": 4.1978688382569524e-05, + "loss": 1.3798, + "step": 93090 + }, + { + "epoch": 8.06, + "learning_rate": 4.197782205665772e-05, + "loss": 1.4037, + "step": 93100 + }, + { + "epoch": 8.06, + "learning_rate": 4.1976955730745906e-05, + "loss": 1.3804, + "step": 93110 + }, + { + "epoch": 8.06, + "learning_rate": 4.19760894048341e-05, + "loss": 1.3914, + "step": 93120 + }, + { + "epoch": 8.06, + "learning_rate": 4.1975223078922295e-05, + "loss": 1.3956, + "step": 93130 + }, + { + "epoch": 8.06, + "learning_rate": 4.197435675301048e-05, + "loss": 1.3494, + "step": 93140 + }, + { + "epoch": 8.06, + "learning_rate": 4.197349042709868e-05, + "loss": 1.3538, + "step": 93150 + }, + { + "epoch": 8.06, + "learning_rate": 4.197262410118687e-05, + "loss": 1.3167, + "step": 93160 + }, + { + "epoch": 8.06, + "learning_rate": 4.197175777527506e-05, + "loss": 1.4425, + "step": 93170 + }, + { + "epoch": 8.07, + "learning_rate": 4.1970891449363254e-05, + "loss": 1.3949, + "step": 93180 + }, + { + "epoch": 8.07, + "learning_rate": 4.197002512345145e-05, + "loss": 1.3656, + "step": 93190 + }, + { + "epoch": 8.07, + "learning_rate": 4.1969158797539636e-05, + "loss": 1.409, + "step": 93200 + }, + { + "epoch": 8.07, + "learning_rate": 4.196829247162783e-05, + "loss": 1.4572, + "step": 93210 + }, + { + "epoch": 8.07, + "learning_rate": 4.196742614571602e-05, + "loss": 1.4204, + "step": 93220 + }, + { + "epoch": 8.07, + "learning_rate": 4.196655981980421e-05, + "loss": 1.3696, + "step": 93230 + }, + { + "epoch": 8.07, + "learning_rate": 4.196569349389241e-05, + "loss": 1.3908, + "step": 93240 + }, + { + "epoch": 8.07, + "learning_rate": 4.1964827167980595e-05, + "loss": 1.363, + "step": 93250 + }, + { + "epoch": 8.07, + "learning_rate": 4.196396084206879e-05, + "loss": 1.3672, + "step": 93260 + }, + { + "epoch": 8.07, + "learning_rate": 4.1963094516156984e-05, + "loss": 1.3903, + "step": 93270 + }, + { + "epoch": 8.07, + "learning_rate": 4.196222819024517e-05, + "loss": 1.4037, + "step": 93280 + }, + { + "epoch": 8.07, + "learning_rate": 4.1961361864333366e-05, + "loss": 1.3907, + "step": 93290 + }, + { + "epoch": 8.08, + "learning_rate": 4.196049553842156e-05, + "loss": 1.4578, + "step": 93300 + }, + { + "epoch": 8.08, + "learning_rate": 4.195962921250975e-05, + "loss": 1.3542, + "step": 93310 + }, + { + "epoch": 8.08, + "learning_rate": 4.195876288659794e-05, + "loss": 1.3828, + "step": 93320 + }, + { + "epoch": 8.08, + "learning_rate": 4.195789656068613e-05, + "loss": 1.2984, + "step": 93330 + }, + { + "epoch": 8.08, + "learning_rate": 4.1957030234774325e-05, + "loss": 1.3323, + "step": 93340 + }, + { + "epoch": 8.08, + "learning_rate": 4.195616390886252e-05, + "loss": 1.3586, + "step": 93350 + }, + { + "epoch": 8.08, + "learning_rate": 4.195529758295071e-05, + "loss": 1.3915, + "step": 93360 + }, + { + "epoch": 8.08, + "learning_rate": 4.19544312570389e-05, + "loss": 1.3596, + "step": 93370 + }, + { + "epoch": 8.08, + "learning_rate": 4.1953564931127096e-05, + "loss": 1.3672, + "step": 93380 + }, + { + "epoch": 8.08, + "learning_rate": 4.195269860521528e-05, + "loss": 1.3491, + "step": 93390 + }, + { + "epoch": 8.08, + "learning_rate": 4.195183227930348e-05, + "loss": 1.3728, + "step": 93400 + }, + { + "epoch": 8.09, + "learning_rate": 4.1950965953391665e-05, + "loss": 1.319, + "step": 93410 + }, + { + "epoch": 8.09, + "learning_rate": 4.195009962747986e-05, + "loss": 1.4186, + "step": 93420 + }, + { + "epoch": 8.09, + "learning_rate": 4.1949233301568054e-05, + "loss": 1.4322, + "step": 93430 + }, + { + "epoch": 8.09, + "learning_rate": 4.194836697565624e-05, + "loss": 1.316, + "step": 93440 + }, + { + "epoch": 8.09, + "learning_rate": 4.1947500649744437e-05, + "loss": 1.3427, + "step": 93450 + }, + { + "epoch": 8.09, + "learning_rate": 4.194663432383263e-05, + "loss": 1.4014, + "step": 93460 + }, + { + "epoch": 8.09, + "learning_rate": 4.194576799792082e-05, + "loss": 1.3421, + "step": 93470 + }, + { + "epoch": 8.09, + "learning_rate": 4.194490167200901e-05, + "loss": 1.3698, + "step": 93480 + }, + { + "epoch": 8.09, + "learning_rate": 4.194403534609721e-05, + "loss": 1.4113, + "step": 93490 + }, + { + "epoch": 8.09, + "learning_rate": 4.1943169020185395e-05, + "loss": 1.3631, + "step": 93500 + }, + { + "epoch": 8.09, + "learning_rate": 4.194230269427359e-05, + "loss": 1.3459, + "step": 93510 + }, + { + "epoch": 8.09, + "learning_rate": 4.194143636836178e-05, + "loss": 1.3041, + "step": 93520 + }, + { + "epoch": 8.1, + "learning_rate": 4.194057004244997e-05, + "loss": 1.3339, + "step": 93530 + }, + { + "epoch": 8.1, + "learning_rate": 4.1939703716538166e-05, + "loss": 1.393, + "step": 93540 + }, + { + "epoch": 8.1, + "learning_rate": 4.1938837390626354e-05, + "loss": 1.424, + "step": 93550 + }, + { + "epoch": 8.1, + "learning_rate": 4.193797106471455e-05, + "loss": 1.3363, + "step": 93560 + }, + { + "epoch": 8.1, + "learning_rate": 4.193710473880274e-05, + "loss": 1.3393, + "step": 93570 + }, + { + "epoch": 8.1, + "learning_rate": 4.193623841289093e-05, + "loss": 1.401, + "step": 93580 + }, + { + "epoch": 8.1, + "learning_rate": 4.1935372086979125e-05, + "loss": 1.3852, + "step": 93590 + }, + { + "epoch": 8.1, + "learning_rate": 4.193450576106732e-05, + "loss": 1.327, + "step": 93600 + }, + { + "epoch": 8.1, + "learning_rate": 4.193363943515551e-05, + "loss": 1.3009, + "step": 93610 + }, + { + "epoch": 8.1, + "learning_rate": 4.19327731092437e-05, + "loss": 1.31, + "step": 93620 + }, + { + "epoch": 8.1, + "learning_rate": 4.193190678333189e-05, + "loss": 1.3576, + "step": 93630 + }, + { + "epoch": 8.11, + "learning_rate": 4.1931040457420084e-05, + "loss": 1.3713, + "step": 93640 + }, + { + "epoch": 8.11, + "learning_rate": 4.193017413150828e-05, + "loss": 1.386, + "step": 93650 + }, + { + "epoch": 8.11, + "learning_rate": 4.1929307805596466e-05, + "loss": 1.3924, + "step": 93660 + }, + { + "epoch": 8.11, + "learning_rate": 4.192844147968466e-05, + "loss": 1.318, + "step": 93670 + }, + { + "epoch": 8.11, + "learning_rate": 4.1927575153772855e-05, + "loss": 1.38, + "step": 93680 + }, + { + "epoch": 8.11, + "learning_rate": 4.192670882786104e-05, + "loss": 1.3777, + "step": 93690 + }, + { + "epoch": 8.11, + "learning_rate": 4.192584250194924e-05, + "loss": 1.3976, + "step": 93700 + }, + { + "epoch": 8.11, + "learning_rate": 4.192497617603743e-05, + "loss": 1.4397, + "step": 93710 + }, + { + "epoch": 8.11, + "learning_rate": 4.192410985012562e-05, + "loss": 1.3967, + "step": 93720 + }, + { + "epoch": 8.11, + "learning_rate": 4.1923243524213813e-05, + "loss": 1.4313, + "step": 93730 + }, + { + "epoch": 8.11, + "learning_rate": 4.1922377198302e-05, + "loss": 1.3074, + "step": 93740 + }, + { + "epoch": 8.11, + "learning_rate": 4.1921510872390196e-05, + "loss": 1.3753, + "step": 93750 + }, + { + "epoch": 8.12, + "learning_rate": 4.192064454647839e-05, + "loss": 1.3716, + "step": 93760 + }, + { + "epoch": 8.12, + "learning_rate": 4.191977822056658e-05, + "loss": 1.3618, + "step": 93770 + }, + { + "epoch": 8.12, + "learning_rate": 4.191891189465477e-05, + "loss": 1.3258, + "step": 93780 + }, + { + "epoch": 8.12, + "learning_rate": 4.1918045568742967e-05, + "loss": 1.3879, + "step": 93790 + }, + { + "epoch": 8.12, + "learning_rate": 4.1917179242831154e-05, + "loss": 1.3221, + "step": 93800 + }, + { + "epoch": 8.12, + "learning_rate": 4.191631291691935e-05, + "loss": 1.3589, + "step": 93810 + }, + { + "epoch": 8.12, + "learning_rate": 4.191544659100754e-05, + "loss": 1.4543, + "step": 93820 + }, + { + "epoch": 8.12, + "learning_rate": 4.191458026509573e-05, + "loss": 1.3357, + "step": 93830 + }, + { + "epoch": 8.12, + "learning_rate": 4.1913713939183925e-05, + "loss": 1.3581, + "step": 93840 + }, + { + "epoch": 8.12, + "learning_rate": 4.191284761327211e-05, + "loss": 1.4222, + "step": 93850 + }, + { + "epoch": 8.12, + "learning_rate": 4.191198128736031e-05, + "loss": 1.3297, + "step": 93860 + }, + { + "epoch": 8.13, + "learning_rate": 4.19111149614485e-05, + "loss": 1.3886, + "step": 93870 + }, + { + "epoch": 8.13, + "learning_rate": 4.191024863553669e-05, + "loss": 1.3989, + "step": 93880 + }, + { + "epoch": 8.13, + "learning_rate": 4.1909382309624884e-05, + "loss": 1.3593, + "step": 93890 + }, + { + "epoch": 8.13, + "learning_rate": 4.190851598371308e-05, + "loss": 1.4726, + "step": 93900 + }, + { + "epoch": 8.13, + "learning_rate": 4.1907649657801266e-05, + "loss": 1.354, + "step": 93910 + }, + { + "epoch": 8.13, + "learning_rate": 4.190678333188946e-05, + "loss": 1.3436, + "step": 93920 + }, + { + "epoch": 8.13, + "learning_rate": 4.1905917005977655e-05, + "loss": 1.3372, + "step": 93930 + }, + { + "epoch": 8.13, + "learning_rate": 4.190505068006584e-05, + "loss": 1.2973, + "step": 93940 + }, + { + "epoch": 8.13, + "learning_rate": 4.190418435415404e-05, + "loss": 1.3987, + "step": 93950 + }, + { + "epoch": 8.13, + "learning_rate": 4.1903318028242225e-05, + "loss": 1.3657, + "step": 93960 + }, + { + "epoch": 8.13, + "learning_rate": 4.190245170233042e-05, + "loss": 1.455, + "step": 93970 + }, + { + "epoch": 8.13, + "learning_rate": 4.1901585376418614e-05, + "loss": 1.3647, + "step": 93980 + }, + { + "epoch": 8.14, + "learning_rate": 4.19007190505068e-05, + "loss": 1.3708, + "step": 93990 + }, + { + "epoch": 8.14, + "learning_rate": 4.1899852724594996e-05, + "loss": 1.3587, + "step": 94000 + }, + { + "epoch": 8.14, + "learning_rate": 4.189898639868319e-05, + "loss": 1.3771, + "step": 94010 + }, + { + "epoch": 8.14, + "learning_rate": 4.189812007277138e-05, + "loss": 1.4412, + "step": 94020 + }, + { + "epoch": 8.14, + "learning_rate": 4.189725374685957e-05, + "loss": 1.3865, + "step": 94030 + }, + { + "epoch": 8.14, + "learning_rate": 4.189638742094777e-05, + "loss": 1.3917, + "step": 94040 + }, + { + "epoch": 8.14, + "learning_rate": 4.1895521095035955e-05, + "loss": 1.3597, + "step": 94050 + }, + { + "epoch": 8.14, + "learning_rate": 4.189465476912415e-05, + "loss": 1.3635, + "step": 94060 + }, + { + "epoch": 8.14, + "learning_rate": 4.189378844321234e-05, + "loss": 1.2788, + "step": 94070 + }, + { + "epoch": 8.14, + "learning_rate": 4.189292211730053e-05, + "loss": 1.3602, + "step": 94080 + }, + { + "epoch": 8.14, + "learning_rate": 4.1892055791388726e-05, + "loss": 1.34, + "step": 94090 + }, + { + "epoch": 8.14, + "learning_rate": 4.189118946547691e-05, + "loss": 1.3544, + "step": 94100 + }, + { + "epoch": 8.15, + "learning_rate": 4.189032313956511e-05, + "loss": 1.4005, + "step": 94110 + }, + { + "epoch": 8.15, + "learning_rate": 4.18894568136533e-05, + "loss": 1.3423, + "step": 94120 + }, + { + "epoch": 8.15, + "learning_rate": 4.188859048774149e-05, + "loss": 1.3583, + "step": 94130 + }, + { + "epoch": 8.15, + "learning_rate": 4.1887724161829684e-05, + "loss": 1.3546, + "step": 94140 + }, + { + "epoch": 8.15, + "learning_rate": 4.188685783591787e-05, + "loss": 1.4437, + "step": 94150 + }, + { + "epoch": 8.15, + "learning_rate": 4.1885991510006066e-05, + "loss": 1.3871, + "step": 94160 + }, + { + "epoch": 8.15, + "learning_rate": 4.188512518409426e-05, + "loss": 1.3675, + "step": 94170 + }, + { + "epoch": 8.15, + "learning_rate": 4.188425885818245e-05, + "loss": 1.4462, + "step": 94180 + }, + { + "epoch": 8.15, + "learning_rate": 4.188339253227064e-05, + "loss": 1.3763, + "step": 94190 + }, + { + "epoch": 8.15, + "learning_rate": 4.188252620635884e-05, + "loss": 1.464, + "step": 94200 + }, + { + "epoch": 8.15, + "learning_rate": 4.1881659880447025e-05, + "loss": 1.3526, + "step": 94210 + }, + { + "epoch": 8.16, + "learning_rate": 4.188079355453522e-05, + "loss": 1.306, + "step": 94220 + }, + { + "epoch": 8.16, + "learning_rate": 4.1879927228623414e-05, + "loss": 1.4418, + "step": 94230 + }, + { + "epoch": 8.16, + "learning_rate": 4.18790609027116e-05, + "loss": 1.2686, + "step": 94240 + }, + { + "epoch": 8.16, + "learning_rate": 4.1878194576799796e-05, + "loss": 1.3855, + "step": 94250 + }, + { + "epoch": 8.16, + "learning_rate": 4.1877328250887984e-05, + "loss": 1.3656, + "step": 94260 + }, + { + "epoch": 8.16, + "learning_rate": 4.187646192497618e-05, + "loss": 1.3609, + "step": 94270 + }, + { + "epoch": 8.16, + "learning_rate": 4.187559559906437e-05, + "loss": 1.281, + "step": 94280 + }, + { + "epoch": 8.16, + "learning_rate": 4.187472927315256e-05, + "loss": 1.3843, + "step": 94290 + }, + { + "epoch": 8.16, + "learning_rate": 4.1873862947240755e-05, + "loss": 1.3446, + "step": 94300 + }, + { + "epoch": 8.16, + "learning_rate": 4.187299662132895e-05, + "loss": 1.3485, + "step": 94310 + }, + { + "epoch": 8.16, + "learning_rate": 4.187213029541714e-05, + "loss": 1.3623, + "step": 94320 + }, + { + "epoch": 8.16, + "learning_rate": 4.187126396950533e-05, + "loss": 1.3353, + "step": 94330 + }, + { + "epoch": 8.17, + "learning_rate": 4.1870397643593526e-05, + "loss": 1.4038, + "step": 94340 + }, + { + "epoch": 8.17, + "learning_rate": 4.1869531317681714e-05, + "loss": 1.4135, + "step": 94350 + }, + { + "epoch": 8.17, + "learning_rate": 4.186866499176991e-05, + "loss": 1.307, + "step": 94360 + }, + { + "epoch": 8.17, + "learning_rate": 4.1867798665858096e-05, + "loss": 1.3632, + "step": 94370 + }, + { + "epoch": 8.17, + "learning_rate": 4.186693233994629e-05, + "loss": 1.3449, + "step": 94380 + }, + { + "epoch": 8.17, + "learning_rate": 4.1866066014034485e-05, + "loss": 1.3971, + "step": 94390 + }, + { + "epoch": 8.17, + "learning_rate": 4.186519968812267e-05, + "loss": 1.3527, + "step": 94400 + }, + { + "epoch": 8.17, + "learning_rate": 4.186433336221087e-05, + "loss": 1.3725, + "step": 94410 + }, + { + "epoch": 8.17, + "learning_rate": 4.186346703629906e-05, + "loss": 1.3718, + "step": 94420 + }, + { + "epoch": 8.17, + "learning_rate": 4.186260071038725e-05, + "loss": 1.3733, + "step": 94430 + }, + { + "epoch": 8.17, + "learning_rate": 4.1861734384475443e-05, + "loss": 1.3078, + "step": 94440 + }, + { + "epoch": 8.18, + "learning_rate": 4.186086805856364e-05, + "loss": 1.4197, + "step": 94450 + }, + { + "epoch": 8.18, + "learning_rate": 4.1860001732651826e-05, + "loss": 1.3735, + "step": 94460 + }, + { + "epoch": 8.18, + "learning_rate": 4.185913540674002e-05, + "loss": 1.3931, + "step": 94470 + }, + { + "epoch": 8.18, + "learning_rate": 4.185826908082821e-05, + "loss": 1.4092, + "step": 94480 + }, + { + "epoch": 8.18, + "learning_rate": 4.18574027549164e-05, + "loss": 1.2883, + "step": 94490 + }, + { + "epoch": 8.18, + "learning_rate": 4.1856536429004597e-05, + "loss": 1.2854, + "step": 94500 + }, + { + "epoch": 8.18, + "learning_rate": 4.1855670103092784e-05, + "loss": 1.344, + "step": 94510 + }, + { + "epoch": 8.18, + "learning_rate": 4.185480377718098e-05, + "loss": 1.3719, + "step": 94520 + }, + { + "epoch": 8.18, + "learning_rate": 4.185393745126917e-05, + "loss": 1.2691, + "step": 94530 + }, + { + "epoch": 8.18, + "learning_rate": 4.185307112535736e-05, + "loss": 1.2775, + "step": 94540 + }, + { + "epoch": 8.18, + "learning_rate": 4.1852204799445555e-05, + "loss": 1.4237, + "step": 94550 + }, + { + "epoch": 8.18, + "learning_rate": 4.185133847353375e-05, + "loss": 1.3862, + "step": 94560 + }, + { + "epoch": 8.19, + "learning_rate": 4.185047214762194e-05, + "loss": 1.3919, + "step": 94570 + }, + { + "epoch": 8.19, + "learning_rate": 4.184960582171013e-05, + "loss": 1.3143, + "step": 94580 + }, + { + "epoch": 8.19, + "learning_rate": 4.184873949579832e-05, + "loss": 1.4057, + "step": 94590 + }, + { + "epoch": 8.19, + "learning_rate": 4.1847873169886514e-05, + "loss": 1.3155, + "step": 94600 + }, + { + "epoch": 8.19, + "learning_rate": 4.184700684397471e-05, + "loss": 1.3864, + "step": 94610 + }, + { + "epoch": 8.19, + "learning_rate": 4.1846140518062896e-05, + "loss": 1.3993, + "step": 94620 + }, + { + "epoch": 8.19, + "learning_rate": 4.184527419215109e-05, + "loss": 1.3378, + "step": 94630 + }, + { + "epoch": 8.19, + "learning_rate": 4.1844407866239285e-05, + "loss": 1.3531, + "step": 94640 + }, + { + "epoch": 8.19, + "learning_rate": 4.184354154032747e-05, + "loss": 1.3743, + "step": 94650 + }, + { + "epoch": 8.19, + "learning_rate": 4.184267521441567e-05, + "loss": 1.3869, + "step": 94660 + }, + { + "epoch": 8.19, + "learning_rate": 4.184180888850386e-05, + "loss": 1.4257, + "step": 94670 + }, + { + "epoch": 8.2, + "learning_rate": 4.184094256259205e-05, + "loss": 1.3904, + "step": 94680 + }, + { + "epoch": 8.2, + "learning_rate": 4.1840076236680244e-05, + "loss": 1.4258, + "step": 94690 + }, + { + "epoch": 8.2, + "learning_rate": 4.183920991076843e-05, + "loss": 1.3751, + "step": 94700 + }, + { + "epoch": 8.2, + "learning_rate": 4.1838343584856626e-05, + "loss": 1.3344, + "step": 94710 + }, + { + "epoch": 8.2, + "learning_rate": 4.183747725894482e-05, + "loss": 1.3016, + "step": 94720 + }, + { + "epoch": 8.2, + "learning_rate": 4.183661093303301e-05, + "loss": 1.3921, + "step": 94730 + }, + { + "epoch": 8.2, + "learning_rate": 4.18357446071212e-05, + "loss": 1.3216, + "step": 94740 + }, + { + "epoch": 8.2, + "learning_rate": 4.18348782812094e-05, + "loss": 1.3455, + "step": 94750 + }, + { + "epoch": 8.2, + "learning_rate": 4.1834011955297585e-05, + "loss": 1.3545, + "step": 94760 + }, + { + "epoch": 8.2, + "learning_rate": 4.183314562938578e-05, + "loss": 1.3914, + "step": 94770 + }, + { + "epoch": 8.2, + "learning_rate": 4.1832279303473973e-05, + "loss": 1.2716, + "step": 94780 + }, + { + "epoch": 8.2, + "learning_rate": 4.183141297756216e-05, + "loss": 1.3418, + "step": 94790 + }, + { + "epoch": 8.21, + "learning_rate": 4.1830546651650356e-05, + "loss": 1.2768, + "step": 94800 + }, + { + "epoch": 8.21, + "learning_rate": 4.182968032573854e-05, + "loss": 1.3667, + "step": 94810 + }, + { + "epoch": 8.21, + "learning_rate": 4.182881399982674e-05, + "loss": 1.3424, + "step": 94820 + }, + { + "epoch": 8.21, + "learning_rate": 4.182794767391493e-05, + "loss": 1.3681, + "step": 94830 + }, + { + "epoch": 8.21, + "learning_rate": 4.182708134800312e-05, + "loss": 1.4061, + "step": 94840 + }, + { + "epoch": 8.21, + "learning_rate": 4.1826215022091314e-05, + "loss": 1.3715, + "step": 94850 + }, + { + "epoch": 8.21, + "learning_rate": 4.182534869617951e-05, + "loss": 1.2767, + "step": 94860 + }, + { + "epoch": 8.21, + "learning_rate": 4.1824482370267696e-05, + "loss": 1.3428, + "step": 94870 + }, + { + "epoch": 8.21, + "learning_rate": 4.182361604435589e-05, + "loss": 1.3924, + "step": 94880 + }, + { + "epoch": 8.21, + "learning_rate": 4.182274971844408e-05, + "loss": 1.3593, + "step": 94890 + }, + { + "epoch": 8.21, + "learning_rate": 4.182188339253227e-05, + "loss": 1.3738, + "step": 94900 + }, + { + "epoch": 8.22, + "learning_rate": 4.182101706662047e-05, + "loss": 1.3425, + "step": 94910 + }, + { + "epoch": 8.22, + "learning_rate": 4.1820150740708655e-05, + "loss": 1.3736, + "step": 94920 + }, + { + "epoch": 8.22, + "learning_rate": 4.181928441479685e-05, + "loss": 1.3616, + "step": 94930 + }, + { + "epoch": 8.22, + "learning_rate": 4.1818418088885044e-05, + "loss": 1.3277, + "step": 94940 + }, + { + "epoch": 8.22, + "learning_rate": 4.181755176297323e-05, + "loss": 1.3443, + "step": 94950 + }, + { + "epoch": 8.22, + "learning_rate": 4.1816685437061426e-05, + "loss": 1.408, + "step": 94960 + }, + { + "epoch": 8.22, + "learning_rate": 4.181581911114962e-05, + "loss": 1.3836, + "step": 94970 + }, + { + "epoch": 8.22, + "learning_rate": 4.181495278523781e-05, + "loss": 1.3278, + "step": 94980 + }, + { + "epoch": 8.22, + "learning_rate": 4.1814086459326e-05, + "loss": 1.3549, + "step": 94990 + }, + { + "epoch": 8.22, + "learning_rate": 4.181322013341419e-05, + "loss": 1.4192, + "step": 95000 + }, + { + "epoch": 8.22, + "learning_rate": 4.1812353807502385e-05, + "loss": 1.3731, + "step": 95010 + }, + { + "epoch": 8.22, + "learning_rate": 4.181148748159058e-05, + "loss": 1.2919, + "step": 95020 + }, + { + "epoch": 8.23, + "learning_rate": 4.181062115567877e-05, + "loss": 1.3499, + "step": 95030 + }, + { + "epoch": 8.23, + "learning_rate": 4.180975482976696e-05, + "loss": 1.3335, + "step": 95040 + }, + { + "epoch": 8.23, + "learning_rate": 4.1808888503855156e-05, + "loss": 1.3161, + "step": 95050 + }, + { + "epoch": 8.23, + "learning_rate": 4.1808022177943344e-05, + "loss": 1.4142, + "step": 95060 + }, + { + "epoch": 8.23, + "learning_rate": 4.180715585203154e-05, + "loss": 1.4199, + "step": 95070 + }, + { + "epoch": 8.23, + "learning_rate": 4.180628952611973e-05, + "loss": 1.3216, + "step": 95080 + }, + { + "epoch": 8.23, + "learning_rate": 4.180542320020792e-05, + "loss": 1.4098, + "step": 95090 + }, + { + "epoch": 8.23, + "learning_rate": 4.1804556874296115e-05, + "loss": 1.2868, + "step": 95100 + }, + { + "epoch": 8.23, + "learning_rate": 4.18036905483843e-05, + "loss": 1.3472, + "step": 95110 + }, + { + "epoch": 8.23, + "learning_rate": 4.18028242224725e-05, + "loss": 1.3225, + "step": 95120 + }, + { + "epoch": 8.23, + "learning_rate": 4.180195789656069e-05, + "loss": 1.323, + "step": 95130 + }, + { + "epoch": 8.23, + "learning_rate": 4.180109157064888e-05, + "loss": 1.3878, + "step": 95140 + }, + { + "epoch": 8.24, + "learning_rate": 4.180022524473707e-05, + "loss": 1.3476, + "step": 95150 + }, + { + "epoch": 8.24, + "learning_rate": 4.179935891882527e-05, + "loss": 1.3505, + "step": 95160 + }, + { + "epoch": 8.24, + "learning_rate": 4.1798492592913455e-05, + "loss": 1.3816, + "step": 95170 + }, + { + "epoch": 8.24, + "learning_rate": 4.179762626700165e-05, + "loss": 1.3566, + "step": 95180 + }, + { + "epoch": 8.24, + "learning_rate": 4.1796759941089844e-05, + "loss": 1.3439, + "step": 95190 + }, + { + "epoch": 8.24, + "learning_rate": 4.179589361517803e-05, + "loss": 1.386, + "step": 95200 + }, + { + "epoch": 8.24, + "learning_rate": 4.1795027289266227e-05, + "loss": 1.435, + "step": 95210 + }, + { + "epoch": 8.24, + "learning_rate": 4.1794160963354414e-05, + "loss": 1.3329, + "step": 95220 + }, + { + "epoch": 8.24, + "learning_rate": 4.179329463744261e-05, + "loss": 1.3584, + "step": 95230 + }, + { + "epoch": 8.24, + "learning_rate": 4.17924283115308e-05, + "loss": 1.3502, + "step": 95240 + }, + { + "epoch": 8.24, + "learning_rate": 4.179156198561899e-05, + "loss": 1.3774, + "step": 95250 + }, + { + "epoch": 8.25, + "learning_rate": 4.1790695659707185e-05, + "loss": 1.4184, + "step": 95260 + }, + { + "epoch": 8.25, + "learning_rate": 4.178982933379538e-05, + "loss": 1.3435, + "step": 95270 + }, + { + "epoch": 8.25, + "learning_rate": 4.178896300788357e-05, + "loss": 1.3552, + "step": 95280 + }, + { + "epoch": 8.25, + "learning_rate": 4.178809668197176e-05, + "loss": 1.3842, + "step": 95290 + }, + { + "epoch": 8.25, + "learning_rate": 4.1787230356059956e-05, + "loss": 1.3701, + "step": 95300 + }, + { + "epoch": 8.25, + "learning_rate": 4.1786364030148144e-05, + "loss": 1.3053, + "step": 95310 + }, + { + "epoch": 8.25, + "learning_rate": 4.178549770423634e-05, + "loss": 1.4014, + "step": 95320 + }, + { + "epoch": 8.25, + "learning_rate": 4.1784631378324526e-05, + "loss": 1.3866, + "step": 95330 + }, + { + "epoch": 8.25, + "learning_rate": 4.178376505241272e-05, + "loss": 1.3345, + "step": 95340 + }, + { + "epoch": 8.25, + "learning_rate": 4.1782898726500915e-05, + "loss": 1.3922, + "step": 95350 + }, + { + "epoch": 8.25, + "learning_rate": 4.17820324005891e-05, + "loss": 1.359, + "step": 95360 + }, + { + "epoch": 8.25, + "learning_rate": 4.17811660746773e-05, + "loss": 1.3901, + "step": 95370 + }, + { + "epoch": 8.26, + "learning_rate": 4.178029974876549e-05, + "loss": 1.3078, + "step": 95380 + }, + { + "epoch": 8.26, + "learning_rate": 4.177943342285368e-05, + "loss": 1.3889, + "step": 95390 + }, + { + "epoch": 8.26, + "learning_rate": 4.1778567096941874e-05, + "loss": 1.3689, + "step": 95400 + }, + { + "epoch": 8.26, + "learning_rate": 4.177770077103007e-05, + "loss": 1.3609, + "step": 95410 + }, + { + "epoch": 8.26, + "learning_rate": 4.1776834445118256e-05, + "loss": 1.3874, + "step": 95420 + }, + { + "epoch": 8.26, + "learning_rate": 4.177596811920645e-05, + "loss": 1.3912, + "step": 95430 + }, + { + "epoch": 8.26, + "learning_rate": 4.177510179329464e-05, + "loss": 1.3125, + "step": 95440 + }, + { + "epoch": 8.26, + "learning_rate": 4.177423546738283e-05, + "loss": 1.3777, + "step": 95450 + }, + { + "epoch": 8.26, + "learning_rate": 4.177336914147103e-05, + "loss": 1.3292, + "step": 95460 + }, + { + "epoch": 8.26, + "learning_rate": 4.1772502815559215e-05, + "loss": 1.4408, + "step": 95470 + }, + { + "epoch": 8.26, + "learning_rate": 4.177163648964741e-05, + "loss": 1.3829, + "step": 95480 + }, + { + "epoch": 8.27, + "learning_rate": 4.1770770163735603e-05, + "loss": 1.3926, + "step": 95490 + }, + { + "epoch": 8.27, + "learning_rate": 4.176990383782379e-05, + "loss": 1.3484, + "step": 95500 + }, + { + "epoch": 8.27, + "learning_rate": 4.1769037511911986e-05, + "loss": 1.3988, + "step": 95510 + }, + { + "epoch": 8.27, + "learning_rate": 4.176817118600018e-05, + "loss": 1.4035, + "step": 95520 + }, + { + "epoch": 8.27, + "learning_rate": 4.176730486008837e-05, + "loss": 1.3408, + "step": 95530 + }, + { + "epoch": 8.27, + "learning_rate": 4.176643853417656e-05, + "loss": 1.3603, + "step": 95540 + }, + { + "epoch": 8.27, + "learning_rate": 4.176557220826475e-05, + "loss": 1.3682, + "step": 95550 + }, + { + "epoch": 8.27, + "learning_rate": 4.1764705882352944e-05, + "loss": 1.3607, + "step": 95560 + }, + { + "epoch": 8.27, + "learning_rate": 4.176383955644114e-05, + "loss": 1.3149, + "step": 95570 + }, + { + "epoch": 8.27, + "learning_rate": 4.1762973230529326e-05, + "loss": 1.3547, + "step": 95580 + }, + { + "epoch": 8.27, + "learning_rate": 4.176210690461752e-05, + "loss": 1.3884, + "step": 95590 + }, + { + "epoch": 8.27, + "learning_rate": 4.1761240578705715e-05, + "loss": 1.328, + "step": 95600 + }, + { + "epoch": 8.28, + "learning_rate": 4.17603742527939e-05, + "loss": 1.2957, + "step": 95610 + }, + { + "epoch": 8.28, + "learning_rate": 4.17595079268821e-05, + "loss": 1.3699, + "step": 95620 + }, + { + "epoch": 8.28, + "learning_rate": 4.175864160097029e-05, + "loss": 1.3456, + "step": 95630 + }, + { + "epoch": 8.28, + "learning_rate": 4.175777527505848e-05, + "loss": 1.3589, + "step": 95640 + }, + { + "epoch": 8.28, + "learning_rate": 4.1756908949146674e-05, + "loss": 1.3831, + "step": 95650 + }, + { + "epoch": 8.28, + "learning_rate": 4.175604262323486e-05, + "loss": 1.3255, + "step": 95660 + }, + { + "epoch": 8.28, + "learning_rate": 4.1755176297323056e-05, + "loss": 1.3806, + "step": 95670 + }, + { + "epoch": 8.28, + "learning_rate": 4.175430997141125e-05, + "loss": 1.379, + "step": 95680 + }, + { + "epoch": 8.28, + "learning_rate": 4.175344364549944e-05, + "loss": 1.3792, + "step": 95690 + }, + { + "epoch": 8.28, + "learning_rate": 4.175257731958763e-05, + "loss": 1.3657, + "step": 95700 + }, + { + "epoch": 8.28, + "learning_rate": 4.175171099367583e-05, + "loss": 1.4237, + "step": 95710 + }, + { + "epoch": 8.29, + "learning_rate": 4.1750844667764015e-05, + "loss": 1.33, + "step": 95720 + }, + { + "epoch": 8.29, + "learning_rate": 4.174997834185221e-05, + "loss": 1.3305, + "step": 95730 + }, + { + "epoch": 8.29, + "learning_rate": 4.17491120159404e-05, + "loss": 1.3254, + "step": 95740 + }, + { + "epoch": 8.29, + "learning_rate": 4.174824569002859e-05, + "loss": 1.4064, + "step": 95750 + }, + { + "epoch": 8.29, + "learning_rate": 4.1747379364116786e-05, + "loss": 1.2766, + "step": 95760 + }, + { + "epoch": 8.29, + "learning_rate": 4.1746513038204974e-05, + "loss": 1.365, + "step": 95770 + }, + { + "epoch": 8.29, + "learning_rate": 4.174564671229317e-05, + "loss": 1.3703, + "step": 95780 + }, + { + "epoch": 8.29, + "learning_rate": 4.174478038638136e-05, + "loss": 1.3557, + "step": 95790 + }, + { + "epoch": 8.29, + "learning_rate": 4.174391406046955e-05, + "loss": 1.3737, + "step": 95800 + }, + { + "epoch": 8.29, + "learning_rate": 4.1743047734557745e-05, + "loss": 1.4121, + "step": 95810 + }, + { + "epoch": 8.29, + "learning_rate": 4.174218140864594e-05, + "loss": 1.4032, + "step": 95820 + }, + { + "epoch": 8.29, + "learning_rate": 4.174131508273413e-05, + "loss": 1.359, + "step": 95830 + }, + { + "epoch": 8.3, + "learning_rate": 4.174044875682232e-05, + "loss": 1.3468, + "step": 95840 + }, + { + "epoch": 8.3, + "learning_rate": 4.173958243091051e-05, + "loss": 1.4191, + "step": 95850 + }, + { + "epoch": 8.3, + "learning_rate": 4.17387161049987e-05, + "loss": 1.3834, + "step": 95860 + }, + { + "epoch": 8.3, + "learning_rate": 4.17378497790869e-05, + "loss": 1.3261, + "step": 95870 + }, + { + "epoch": 8.3, + "learning_rate": 4.1736983453175085e-05, + "loss": 1.3528, + "step": 95880 + }, + { + "epoch": 8.3, + "learning_rate": 4.173611712726328e-05, + "loss": 1.3429, + "step": 95890 + }, + { + "epoch": 8.3, + "learning_rate": 4.1735250801351474e-05, + "loss": 1.3529, + "step": 95900 + }, + { + "epoch": 8.3, + "learning_rate": 4.173438447543966e-05, + "loss": 1.3862, + "step": 95910 + }, + { + "epoch": 8.3, + "learning_rate": 4.1733518149527857e-05, + "loss": 1.4136, + "step": 95920 + }, + { + "epoch": 8.3, + "learning_rate": 4.173265182361605e-05, + "loss": 1.3345, + "step": 95930 + }, + { + "epoch": 8.3, + "learning_rate": 4.173178549770424e-05, + "loss": 1.3335, + "step": 95940 + }, + { + "epoch": 8.31, + "learning_rate": 4.173091917179243e-05, + "loss": 1.353, + "step": 95950 + }, + { + "epoch": 8.31, + "learning_rate": 4.173005284588062e-05, + "loss": 1.3036, + "step": 95960 + }, + { + "epoch": 8.31, + "learning_rate": 4.1729186519968815e-05, + "loss": 1.3342, + "step": 95970 + }, + { + "epoch": 8.31, + "learning_rate": 4.172832019405701e-05, + "loss": 1.39, + "step": 95980 + }, + { + "epoch": 8.31, + "learning_rate": 4.17274538681452e-05, + "loss": 1.2971, + "step": 95990 + }, + { + "epoch": 8.31, + "learning_rate": 4.172658754223339e-05, + "loss": 1.3892, + "step": 96000 + }, + { + "epoch": 8.31, + "learning_rate": 4.1725721216321586e-05, + "loss": 1.3998, + "step": 96010 + }, + { + "epoch": 8.31, + "learning_rate": 4.1724854890409774e-05, + "loss": 1.2887, + "step": 96020 + }, + { + "epoch": 8.31, + "learning_rate": 4.172398856449797e-05, + "loss": 1.4093, + "step": 96030 + }, + { + "epoch": 8.31, + "learning_rate": 4.172312223858616e-05, + "loss": 1.3901, + "step": 96040 + }, + { + "epoch": 8.31, + "learning_rate": 4.172225591267435e-05, + "loss": 1.3805, + "step": 96050 + }, + { + "epoch": 8.31, + "learning_rate": 4.1721389586762545e-05, + "loss": 1.3829, + "step": 96060 + }, + { + "epoch": 8.32, + "learning_rate": 4.172052326085073e-05, + "loss": 1.3429, + "step": 96070 + }, + { + "epoch": 8.32, + "learning_rate": 4.171965693493893e-05, + "loss": 1.3176, + "step": 96080 + }, + { + "epoch": 8.32, + "learning_rate": 4.171879060902712e-05, + "loss": 1.3748, + "step": 96090 + }, + { + "epoch": 8.32, + "learning_rate": 4.171792428311531e-05, + "loss": 1.3881, + "step": 96100 + }, + { + "epoch": 8.32, + "learning_rate": 4.1717057957203504e-05, + "loss": 1.3463, + "step": 96110 + }, + { + "epoch": 8.32, + "learning_rate": 4.17161916312917e-05, + "loss": 1.3648, + "step": 96120 + }, + { + "epoch": 8.32, + "learning_rate": 4.1715325305379886e-05, + "loss": 1.362, + "step": 96130 + }, + { + "epoch": 8.32, + "learning_rate": 4.171445897946808e-05, + "loss": 1.3899, + "step": 96140 + }, + { + "epoch": 8.32, + "learning_rate": 4.1713592653556275e-05, + "loss": 1.2724, + "step": 96150 + }, + { + "epoch": 8.32, + "learning_rate": 4.171272632764446e-05, + "loss": 1.4146, + "step": 96160 + }, + { + "epoch": 8.32, + "learning_rate": 4.171186000173266e-05, + "loss": 1.3223, + "step": 96170 + }, + { + "epoch": 8.33, + "learning_rate": 4.1710993675820845e-05, + "loss": 1.4005, + "step": 96180 + }, + { + "epoch": 8.33, + "learning_rate": 4.171012734990904e-05, + "loss": 1.4248, + "step": 96190 + }, + { + "epoch": 8.33, + "learning_rate": 4.1709261023997233e-05, + "loss": 1.2977, + "step": 96200 + }, + { + "epoch": 8.33, + "learning_rate": 4.170839469808542e-05, + "loss": 1.3542, + "step": 96210 + }, + { + "epoch": 8.33, + "learning_rate": 4.1707528372173616e-05, + "loss": 1.3764, + "step": 96220 + }, + { + "epoch": 8.33, + "learning_rate": 4.170666204626181e-05, + "loss": 1.367, + "step": 96230 + }, + { + "epoch": 8.33, + "learning_rate": 4.170579572035e-05, + "loss": 1.4346, + "step": 96240 + }, + { + "epoch": 8.33, + "learning_rate": 4.170492939443819e-05, + "loss": 1.3935, + "step": 96250 + }, + { + "epoch": 8.33, + "learning_rate": 4.1704063068526387e-05, + "loss": 1.346, + "step": 96260 + }, + { + "epoch": 8.33, + "learning_rate": 4.1703196742614574e-05, + "loss": 1.3151, + "step": 96270 + }, + { + "epoch": 8.33, + "learning_rate": 4.170233041670277e-05, + "loss": 1.3685, + "step": 96280 + }, + { + "epoch": 8.33, + "learning_rate": 4.1701464090790956e-05, + "loss": 1.3416, + "step": 96290 + }, + { + "epoch": 8.34, + "learning_rate": 4.170059776487915e-05, + "loss": 1.3876, + "step": 96300 + }, + { + "epoch": 8.34, + "learning_rate": 4.1699731438967345e-05, + "loss": 1.2612, + "step": 96310 + }, + { + "epoch": 8.34, + "learning_rate": 4.169886511305553e-05, + "loss": 1.3993, + "step": 96320 + }, + { + "epoch": 8.34, + "learning_rate": 4.169799878714373e-05, + "loss": 1.3418, + "step": 96330 + }, + { + "epoch": 8.34, + "learning_rate": 4.169713246123192e-05, + "loss": 1.3714, + "step": 96340 + }, + { + "epoch": 8.34, + "learning_rate": 4.169626613532011e-05, + "loss": 1.375, + "step": 96350 + }, + { + "epoch": 8.34, + "learning_rate": 4.1695399809408304e-05, + "loss": 1.3507, + "step": 96360 + }, + { + "epoch": 8.34, + "learning_rate": 4.16945334834965e-05, + "loss": 1.3551, + "step": 96370 + }, + { + "epoch": 8.34, + "learning_rate": 4.1693667157584686e-05, + "loss": 1.3819, + "step": 96380 + }, + { + "epoch": 8.34, + "learning_rate": 4.169280083167288e-05, + "loss": 1.3621, + "step": 96390 + }, + { + "epoch": 8.34, + "learning_rate": 4.169193450576107e-05, + "loss": 1.3418, + "step": 96400 + }, + { + "epoch": 8.34, + "learning_rate": 4.169106817984926e-05, + "loss": 1.3277, + "step": 96410 + }, + { + "epoch": 8.35, + "learning_rate": 4.169020185393746e-05, + "loss": 1.3238, + "step": 96420 + }, + { + "epoch": 8.35, + "learning_rate": 4.1689335528025645e-05, + "loss": 1.302, + "step": 96430 + }, + { + "epoch": 8.35, + "learning_rate": 4.168846920211384e-05, + "loss": 1.399, + "step": 96440 + }, + { + "epoch": 8.35, + "learning_rate": 4.1687602876202034e-05, + "loss": 1.419, + "step": 96450 + }, + { + "epoch": 8.35, + "learning_rate": 4.168673655029022e-05, + "loss": 1.4001, + "step": 96460 + }, + { + "epoch": 8.35, + "learning_rate": 4.1685870224378416e-05, + "loss": 1.3642, + "step": 96470 + }, + { + "epoch": 8.35, + "learning_rate": 4.1685003898466604e-05, + "loss": 1.3834, + "step": 96480 + }, + { + "epoch": 8.35, + "learning_rate": 4.16841375725548e-05, + "loss": 1.3859, + "step": 96490 + }, + { + "epoch": 8.35, + "learning_rate": 4.168327124664299e-05, + "loss": 1.4285, + "step": 96500 + }, + { + "epoch": 8.35, + "learning_rate": 4.168240492073118e-05, + "loss": 1.3719, + "step": 96510 + }, + { + "epoch": 8.35, + "learning_rate": 4.1681538594819375e-05, + "loss": 1.3706, + "step": 96520 + }, + { + "epoch": 8.36, + "learning_rate": 4.168067226890757e-05, + "loss": 1.4218, + "step": 96530 + }, + { + "epoch": 8.36, + "learning_rate": 4.167980594299576e-05, + "loss": 1.3127, + "step": 96540 + }, + { + "epoch": 8.36, + "learning_rate": 4.167893961708395e-05, + "loss": 1.3329, + "step": 96550 + }, + { + "epoch": 8.36, + "learning_rate": 4.1678073291172146e-05, + "loss": 1.3485, + "step": 96560 + }, + { + "epoch": 8.36, + "learning_rate": 4.167720696526033e-05, + "loss": 1.3852, + "step": 96570 + }, + { + "epoch": 8.36, + "learning_rate": 4.167634063934853e-05, + "loss": 1.4145, + "step": 96580 + }, + { + "epoch": 8.36, + "learning_rate": 4.1675474313436715e-05, + "loss": 1.3328, + "step": 96590 + }, + { + "epoch": 8.36, + "learning_rate": 4.167460798752491e-05, + "loss": 1.355, + "step": 96600 + }, + { + "epoch": 8.36, + "learning_rate": 4.1673741661613104e-05, + "loss": 1.3689, + "step": 96610 + }, + { + "epoch": 8.36, + "learning_rate": 4.167287533570129e-05, + "loss": 1.2468, + "step": 96620 + }, + { + "epoch": 8.36, + "learning_rate": 4.1672009009789486e-05, + "loss": 1.3023, + "step": 96630 + }, + { + "epoch": 8.36, + "learning_rate": 4.167114268387768e-05, + "loss": 1.3319, + "step": 96640 + }, + { + "epoch": 8.37, + "learning_rate": 4.167027635796587e-05, + "loss": 1.3648, + "step": 96650 + }, + { + "epoch": 8.37, + "learning_rate": 4.166941003205406e-05, + "loss": 1.3673, + "step": 96660 + }, + { + "epoch": 8.37, + "learning_rate": 4.166854370614226e-05, + "loss": 1.3742, + "step": 96670 + }, + { + "epoch": 8.37, + "learning_rate": 4.1667677380230445e-05, + "loss": 1.3347, + "step": 96680 + }, + { + "epoch": 8.37, + "learning_rate": 4.166681105431864e-05, + "loss": 1.3202, + "step": 96690 + }, + { + "epoch": 8.37, + "learning_rate": 4.166594472840683e-05, + "loss": 1.3564, + "step": 96700 + }, + { + "epoch": 8.37, + "learning_rate": 4.166507840249502e-05, + "loss": 1.3849, + "step": 96710 + }, + { + "epoch": 8.37, + "learning_rate": 4.1664212076583216e-05, + "loss": 1.4054, + "step": 96720 + }, + { + "epoch": 8.37, + "learning_rate": 4.1663345750671404e-05, + "loss": 1.3818, + "step": 96730 + }, + { + "epoch": 8.37, + "learning_rate": 4.16624794247596e-05, + "loss": 1.3506, + "step": 96740 + }, + { + "epoch": 8.37, + "learning_rate": 4.166161309884779e-05, + "loss": 1.2725, + "step": 96750 + }, + { + "epoch": 8.38, + "learning_rate": 4.166074677293598e-05, + "loss": 1.3858, + "step": 96760 + }, + { + "epoch": 8.38, + "learning_rate": 4.1659880447024175e-05, + "loss": 1.409, + "step": 96770 + }, + { + "epoch": 8.38, + "learning_rate": 4.165901412111237e-05, + "loss": 1.369, + "step": 96780 + }, + { + "epoch": 8.38, + "learning_rate": 4.165814779520056e-05, + "loss": 1.3564, + "step": 96790 + }, + { + "epoch": 8.38, + "learning_rate": 4.165728146928875e-05, + "loss": 1.3023, + "step": 96800 + }, + { + "epoch": 8.38, + "learning_rate": 4.165641514337694e-05, + "loss": 1.3208, + "step": 96810 + }, + { + "epoch": 8.38, + "learning_rate": 4.1655548817465134e-05, + "loss": 1.3967, + "step": 96820 + }, + { + "epoch": 8.38, + "learning_rate": 4.165468249155333e-05, + "loss": 1.3549, + "step": 96830 + }, + { + "epoch": 8.38, + "learning_rate": 4.1653816165641516e-05, + "loss": 1.3397, + "step": 96840 + }, + { + "epoch": 8.38, + "learning_rate": 4.165294983972971e-05, + "loss": 1.4126, + "step": 96850 + }, + { + "epoch": 8.38, + "learning_rate": 4.1652083513817905e-05, + "loss": 1.3465, + "step": 96860 + }, + { + "epoch": 8.38, + "learning_rate": 4.165121718790609e-05, + "loss": 1.3692, + "step": 96870 + }, + { + "epoch": 8.39, + "learning_rate": 4.165035086199429e-05, + "loss": 1.4116, + "step": 96880 + }, + { + "epoch": 8.39, + "learning_rate": 4.164948453608248e-05, + "loss": 1.3089, + "step": 96890 + }, + { + "epoch": 8.39, + "learning_rate": 4.164861821017067e-05, + "loss": 1.3438, + "step": 96900 + }, + { + "epoch": 8.39, + "learning_rate": 4.1647751884258863e-05, + "loss": 1.3683, + "step": 96910 + }, + { + "epoch": 8.39, + "learning_rate": 4.164688555834705e-05, + "loss": 1.406, + "step": 96920 + }, + { + "epoch": 8.39, + "learning_rate": 4.1646019232435246e-05, + "loss": 1.4128, + "step": 96930 + }, + { + "epoch": 8.39, + "learning_rate": 4.164515290652344e-05, + "loss": 1.2813, + "step": 96940 + }, + { + "epoch": 8.39, + "learning_rate": 4.164428658061163e-05, + "loss": 1.3754, + "step": 96950 + }, + { + "epoch": 8.39, + "learning_rate": 4.164342025469982e-05, + "loss": 1.3681, + "step": 96960 + }, + { + "epoch": 8.39, + "learning_rate": 4.1642553928788017e-05, + "loss": 1.3973, + "step": 96970 + }, + { + "epoch": 8.39, + "learning_rate": 4.1641687602876204e-05, + "loss": 1.3852, + "step": 96980 + }, + { + "epoch": 8.4, + "learning_rate": 4.16408212769644e-05, + "loss": 1.3376, + "step": 96990 + }, + { + "epoch": 8.4, + "learning_rate": 4.163995495105259e-05, + "loss": 1.4084, + "step": 97000 + }, + { + "epoch": 8.4, + "learning_rate": 4.163908862514078e-05, + "loss": 1.3841, + "step": 97010 + }, + { + "epoch": 8.4, + "learning_rate": 4.1638222299228975e-05, + "loss": 1.4384, + "step": 97020 + }, + { + "epoch": 8.4, + "learning_rate": 4.163735597331716e-05, + "loss": 1.3761, + "step": 97030 + }, + { + "epoch": 8.4, + "learning_rate": 4.163648964740536e-05, + "loss": 1.343, + "step": 97040 + }, + { + "epoch": 8.4, + "learning_rate": 4.163562332149355e-05, + "loss": 1.4321, + "step": 97050 + }, + { + "epoch": 8.4, + "learning_rate": 4.163475699558174e-05, + "loss": 1.4001, + "step": 97060 + }, + { + "epoch": 8.4, + "learning_rate": 4.1633890669669934e-05, + "loss": 1.3237, + "step": 97070 + }, + { + "epoch": 8.4, + "learning_rate": 4.163302434375813e-05, + "loss": 1.4072, + "step": 97080 + }, + { + "epoch": 8.4, + "learning_rate": 4.1632158017846316e-05, + "loss": 1.3789, + "step": 97090 + }, + { + "epoch": 8.4, + "learning_rate": 4.163129169193451e-05, + "loss": 1.3905, + "step": 97100 + }, + { + "epoch": 8.41, + "learning_rate": 4.1630425366022705e-05, + "loss": 1.357, + "step": 97110 + }, + { + "epoch": 8.41, + "learning_rate": 4.162955904011089e-05, + "loss": 1.3692, + "step": 97120 + }, + { + "epoch": 8.41, + "learning_rate": 4.162869271419909e-05, + "loss": 1.3563, + "step": 97130 + }, + { + "epoch": 8.41, + "learning_rate": 4.1627826388287275e-05, + "loss": 1.3725, + "step": 97140 + }, + { + "epoch": 8.41, + "learning_rate": 4.162696006237547e-05, + "loss": 1.3807, + "step": 97150 + }, + { + "epoch": 8.41, + "learning_rate": 4.1626093736463664e-05, + "loss": 1.3244, + "step": 97160 + }, + { + "epoch": 8.41, + "learning_rate": 4.162522741055185e-05, + "loss": 1.3037, + "step": 97170 + }, + { + "epoch": 8.41, + "learning_rate": 4.1624361084640046e-05, + "loss": 1.2793, + "step": 97180 + }, + { + "epoch": 8.41, + "learning_rate": 4.162349475872824e-05, + "loss": 1.4076, + "step": 97190 + }, + { + "epoch": 8.41, + "learning_rate": 4.162262843281643e-05, + "loss": 1.3057, + "step": 97200 + }, + { + "epoch": 8.41, + "learning_rate": 4.162176210690462e-05, + "loss": 1.3027, + "step": 97210 + }, + { + "epoch": 8.42, + "learning_rate": 4.162089578099281e-05, + "loss": 1.3856, + "step": 97220 + }, + { + "epoch": 8.42, + "learning_rate": 4.1620029455081005e-05, + "loss": 1.3598, + "step": 97230 + }, + { + "epoch": 8.42, + "learning_rate": 4.16191631291692e-05, + "loss": 1.3131, + "step": 97240 + }, + { + "epoch": 8.42, + "learning_rate": 4.161829680325739e-05, + "loss": 1.3143, + "step": 97250 + }, + { + "epoch": 8.42, + "learning_rate": 4.161743047734558e-05, + "loss": 1.3583, + "step": 97260 + }, + { + "epoch": 8.42, + "learning_rate": 4.1616564151433776e-05, + "loss": 1.3389, + "step": 97270 + }, + { + "epoch": 8.42, + "learning_rate": 4.161569782552196e-05, + "loss": 1.3782, + "step": 97280 + }, + { + "epoch": 8.42, + "learning_rate": 4.161483149961016e-05, + "loss": 1.4028, + "step": 97290 + }, + { + "epoch": 8.42, + "learning_rate": 4.161396517369835e-05, + "loss": 1.3426, + "step": 97300 + }, + { + "epoch": 8.42, + "learning_rate": 4.161309884778654e-05, + "loss": 1.3948, + "step": 97310 + }, + { + "epoch": 8.42, + "learning_rate": 4.1612232521874734e-05, + "loss": 1.3527, + "step": 97320 + }, + { + "epoch": 8.42, + "learning_rate": 4.161136619596292e-05, + "loss": 1.3252, + "step": 97330 + }, + { + "epoch": 8.43, + "learning_rate": 4.1610499870051116e-05, + "loss": 1.3368, + "step": 97340 + }, + { + "epoch": 8.43, + "learning_rate": 4.160963354413931e-05, + "loss": 1.3272, + "step": 97350 + }, + { + "epoch": 8.43, + "learning_rate": 4.16087672182275e-05, + "loss": 1.2938, + "step": 97360 + }, + { + "epoch": 8.43, + "learning_rate": 4.160790089231569e-05, + "loss": 1.3352, + "step": 97370 + }, + { + "epoch": 8.43, + "learning_rate": 4.160703456640389e-05, + "loss": 1.4159, + "step": 97380 + }, + { + "epoch": 8.43, + "learning_rate": 4.1606168240492075e-05, + "loss": 1.3324, + "step": 97390 + }, + { + "epoch": 8.43, + "learning_rate": 4.160530191458027e-05, + "loss": 1.3241, + "step": 97400 + }, + { + "epoch": 8.43, + "learning_rate": 4.1604435588668464e-05, + "loss": 1.3018, + "step": 97410 + }, + { + "epoch": 8.43, + "learning_rate": 4.160356926275665e-05, + "loss": 1.3509, + "step": 97420 + }, + { + "epoch": 8.43, + "learning_rate": 4.1602702936844846e-05, + "loss": 1.3121, + "step": 97430 + }, + { + "epoch": 8.43, + "learning_rate": 4.1601836610933034e-05, + "loss": 1.3937, + "step": 97440 + }, + { + "epoch": 8.43, + "learning_rate": 4.160097028502123e-05, + "loss": 1.3638, + "step": 97450 + }, + { + "epoch": 8.44, + "learning_rate": 4.160010395910942e-05, + "loss": 1.3456, + "step": 97460 + }, + { + "epoch": 8.44, + "learning_rate": 4.159923763319761e-05, + "loss": 1.3264, + "step": 97470 + }, + { + "epoch": 8.44, + "learning_rate": 4.1598371307285805e-05, + "loss": 1.3735, + "step": 97480 + }, + { + "epoch": 8.44, + "learning_rate": 4.1597504981374e-05, + "loss": 1.362, + "step": 97490 + }, + { + "epoch": 8.44, + "learning_rate": 4.159663865546219e-05, + "loss": 1.3551, + "step": 97500 + }, + { + "epoch": 8.44, + "learning_rate": 4.159577232955038e-05, + "loss": 1.347, + "step": 97510 + }, + { + "epoch": 8.44, + "learning_rate": 4.1594906003638576e-05, + "loss": 1.3941, + "step": 97520 + }, + { + "epoch": 8.44, + "learning_rate": 4.1594039677726764e-05, + "loss": 1.4006, + "step": 97530 + }, + { + "epoch": 8.44, + "learning_rate": 4.159317335181496e-05, + "loss": 1.3525, + "step": 97540 + }, + { + "epoch": 8.44, + "learning_rate": 4.1592307025903146e-05, + "loss": 1.3163, + "step": 97550 + }, + { + "epoch": 8.44, + "learning_rate": 4.159144069999134e-05, + "loss": 1.3527, + "step": 97560 + }, + { + "epoch": 8.45, + "learning_rate": 4.1590574374079535e-05, + "loss": 1.3397, + "step": 97570 + }, + { + "epoch": 8.45, + "learning_rate": 4.158970804816772e-05, + "loss": 1.3398, + "step": 97580 + }, + { + "epoch": 8.45, + "learning_rate": 4.158884172225592e-05, + "loss": 1.328, + "step": 97590 + }, + { + "epoch": 8.45, + "learning_rate": 4.158797539634411e-05, + "loss": 1.3704, + "step": 97600 + }, + { + "epoch": 8.45, + "learning_rate": 4.15871090704323e-05, + "loss": 1.4238, + "step": 97610 + }, + { + "epoch": 8.45, + "learning_rate": 4.158624274452049e-05, + "loss": 1.3177, + "step": 97620 + }, + { + "epoch": 8.45, + "learning_rate": 4.158537641860869e-05, + "loss": 1.3356, + "step": 97630 + }, + { + "epoch": 8.45, + "learning_rate": 4.1584510092696875e-05, + "loss": 1.3539, + "step": 97640 + }, + { + "epoch": 8.45, + "learning_rate": 4.158364376678507e-05, + "loss": 1.2972, + "step": 97650 + }, + { + "epoch": 8.45, + "learning_rate": 4.158277744087326e-05, + "loss": 1.3337, + "step": 97660 + }, + { + "epoch": 8.45, + "learning_rate": 4.158191111496145e-05, + "loss": 1.347, + "step": 97670 + }, + { + "epoch": 8.45, + "learning_rate": 4.1581044789049647e-05, + "loss": 1.3539, + "step": 97680 + }, + { + "epoch": 8.46, + "learning_rate": 4.1580178463137834e-05, + "loss": 1.3719, + "step": 97690 + }, + { + "epoch": 8.46, + "learning_rate": 4.157931213722603e-05, + "loss": 1.3287, + "step": 97700 + }, + { + "epoch": 8.46, + "learning_rate": 4.157844581131422e-05, + "loss": 1.2898, + "step": 97710 + }, + { + "epoch": 8.46, + "learning_rate": 4.157757948540241e-05, + "loss": 1.3127, + "step": 97720 + }, + { + "epoch": 8.46, + "learning_rate": 4.1576713159490605e-05, + "loss": 1.3476, + "step": 97730 + }, + { + "epoch": 8.46, + "learning_rate": 4.15758468335788e-05, + "loss": 1.3717, + "step": 97740 + }, + { + "epoch": 8.46, + "learning_rate": 4.157498050766699e-05, + "loss": 1.346, + "step": 97750 + }, + { + "epoch": 8.46, + "learning_rate": 4.157411418175518e-05, + "loss": 1.3032, + "step": 97760 + }, + { + "epoch": 8.46, + "learning_rate": 4.157324785584337e-05, + "loss": 1.3937, + "step": 97770 + }, + { + "epoch": 8.46, + "learning_rate": 4.1572381529931564e-05, + "loss": 1.3962, + "step": 97780 + }, + { + "epoch": 8.46, + "learning_rate": 4.157151520401976e-05, + "loss": 1.3429, + "step": 97790 + }, + { + "epoch": 8.47, + "learning_rate": 4.1570648878107946e-05, + "loss": 1.3845, + "step": 97800 + }, + { + "epoch": 8.47, + "learning_rate": 4.156978255219614e-05, + "loss": 1.3269, + "step": 97810 + }, + { + "epoch": 8.47, + "learning_rate": 4.1568916226284335e-05, + "loss": 1.4263, + "step": 97820 + }, + { + "epoch": 8.47, + "learning_rate": 4.156804990037252e-05, + "loss": 1.3159, + "step": 97830 + }, + { + "epoch": 8.47, + "learning_rate": 4.156718357446072e-05, + "loss": 1.3413, + "step": 97840 + }, + { + "epoch": 8.47, + "learning_rate": 4.156631724854891e-05, + "loss": 1.3785, + "step": 97850 + }, + { + "epoch": 8.47, + "learning_rate": 4.15654509226371e-05, + "loss": 1.3688, + "step": 97860 + }, + { + "epoch": 8.47, + "learning_rate": 4.1564584596725294e-05, + "loss": 1.3093, + "step": 97870 + }, + { + "epoch": 8.47, + "learning_rate": 4.156371827081348e-05, + "loss": 1.2824, + "step": 97880 + }, + { + "epoch": 8.47, + "learning_rate": 4.1562851944901676e-05, + "loss": 1.3642, + "step": 97890 + }, + { + "epoch": 8.47, + "learning_rate": 4.156198561898987e-05, + "loss": 1.3395, + "step": 97900 + }, + { + "epoch": 8.47, + "learning_rate": 4.156111929307806e-05, + "loss": 1.3173, + "step": 97910 + }, + { + "epoch": 8.48, + "learning_rate": 4.156025296716625e-05, + "loss": 1.3588, + "step": 97920 + }, + { + "epoch": 8.48, + "learning_rate": 4.155938664125445e-05, + "loss": 1.3486, + "step": 97930 + }, + { + "epoch": 8.48, + "learning_rate": 4.1558520315342635e-05, + "loss": 1.2708, + "step": 97940 + }, + { + "epoch": 8.48, + "learning_rate": 4.155765398943083e-05, + "loss": 1.3682, + "step": 97950 + }, + { + "epoch": 8.48, + "learning_rate": 4.1556787663519023e-05, + "loss": 1.3368, + "step": 97960 + }, + { + "epoch": 8.48, + "learning_rate": 4.155592133760721e-05, + "loss": 1.3468, + "step": 97970 + }, + { + "epoch": 8.48, + "learning_rate": 4.1555055011695406e-05, + "loss": 1.3862, + "step": 97980 + }, + { + "epoch": 8.48, + "learning_rate": 4.155418868578359e-05, + "loss": 1.3437, + "step": 97990 + }, + { + "epoch": 8.48, + "learning_rate": 4.155332235987179e-05, + "loss": 1.3251, + "step": 98000 + }, + { + "epoch": 8.48, + "learning_rate": 4.155245603395998e-05, + "loss": 1.3556, + "step": 98010 + }, + { + "epoch": 8.48, + "learning_rate": 4.155158970804817e-05, + "loss": 1.2968, + "step": 98020 + }, + { + "epoch": 8.49, + "learning_rate": 4.1550723382136364e-05, + "loss": 1.3617, + "step": 98030 + }, + { + "epoch": 8.49, + "learning_rate": 4.154985705622456e-05, + "loss": 1.367, + "step": 98040 + }, + { + "epoch": 8.49, + "learning_rate": 4.1548990730312746e-05, + "loss": 1.3747, + "step": 98050 + }, + { + "epoch": 8.49, + "learning_rate": 4.154812440440094e-05, + "loss": 1.4016, + "step": 98060 + }, + { + "epoch": 8.49, + "learning_rate": 4.154725807848913e-05, + "loss": 1.3503, + "step": 98070 + }, + { + "epoch": 8.49, + "learning_rate": 4.154639175257732e-05, + "loss": 1.3515, + "step": 98080 + }, + { + "epoch": 8.49, + "learning_rate": 4.154552542666552e-05, + "loss": 1.3949, + "step": 98090 + }, + { + "epoch": 8.49, + "learning_rate": 4.1544659100753705e-05, + "loss": 1.3456, + "step": 98100 + }, + { + "epoch": 8.49, + "learning_rate": 4.15437927748419e-05, + "loss": 1.4218, + "step": 98110 + }, + { + "epoch": 8.49, + "learning_rate": 4.1542926448930094e-05, + "loss": 1.2965, + "step": 98120 + }, + { + "epoch": 8.49, + "learning_rate": 4.154206012301828e-05, + "loss": 1.4028, + "step": 98130 + }, + { + "epoch": 8.49, + "learning_rate": 4.1541193797106476e-05, + "loss": 1.3302, + "step": 98140 + }, + { + "epoch": 8.5, + "learning_rate": 4.154032747119467e-05, + "loss": 1.4001, + "step": 98150 + }, + { + "epoch": 8.5, + "learning_rate": 4.153946114528286e-05, + "loss": 1.3619, + "step": 98160 + }, + { + "epoch": 8.5, + "learning_rate": 4.153859481937105e-05, + "loss": 1.375, + "step": 98170 + }, + { + "epoch": 8.5, + "learning_rate": 4.153772849345924e-05, + "loss": 1.4202, + "step": 98180 + }, + { + "epoch": 8.5, + "learning_rate": 4.1536862167547435e-05, + "loss": 1.3095, + "step": 98190 + }, + { + "epoch": 8.5, + "learning_rate": 4.153599584163563e-05, + "loss": 1.3406, + "step": 98200 + }, + { + "epoch": 8.5, + "learning_rate": 4.153512951572382e-05, + "loss": 1.3854, + "step": 98210 + }, + { + "epoch": 8.5, + "learning_rate": 4.153426318981201e-05, + "loss": 1.3207, + "step": 98220 + }, + { + "epoch": 8.5, + "learning_rate": 4.1533396863900206e-05, + "loss": 1.3134, + "step": 98230 + }, + { + "epoch": 8.5, + "learning_rate": 4.1532530537988394e-05, + "loss": 1.3861, + "step": 98240 + }, + { + "epoch": 8.5, + "learning_rate": 4.153166421207659e-05, + "loss": 1.3777, + "step": 98250 + }, + { + "epoch": 8.51, + "learning_rate": 4.153079788616478e-05, + "loss": 1.2589, + "step": 98260 + }, + { + "epoch": 8.51, + "learning_rate": 4.152993156025297e-05, + "loss": 1.3805, + "step": 98270 + }, + { + "epoch": 8.51, + "learning_rate": 4.1529065234341165e-05, + "loss": 1.3305, + "step": 98280 + }, + { + "epoch": 8.51, + "learning_rate": 4.152819890842935e-05, + "loss": 1.3263, + "step": 98290 + }, + { + "epoch": 8.51, + "learning_rate": 4.152733258251755e-05, + "loss": 1.3195, + "step": 98300 + }, + { + "epoch": 8.51, + "learning_rate": 4.152646625660574e-05, + "loss": 1.3995, + "step": 98310 + }, + { + "epoch": 8.51, + "learning_rate": 4.152559993069393e-05, + "loss": 1.3475, + "step": 98320 + }, + { + "epoch": 8.51, + "learning_rate": 4.152473360478212e-05, + "loss": 1.3345, + "step": 98330 + }, + { + "epoch": 8.51, + "learning_rate": 4.152386727887032e-05, + "loss": 1.2793, + "step": 98340 + }, + { + "epoch": 8.51, + "learning_rate": 4.1523000952958505e-05, + "loss": 1.3661, + "step": 98350 + }, + { + "epoch": 8.51, + "learning_rate": 4.15221346270467e-05, + "loss": 1.3885, + "step": 98360 + }, + { + "epoch": 8.51, + "learning_rate": 4.1521268301134894e-05, + "loss": 1.3165, + "step": 98370 + }, + { + "epoch": 8.52, + "learning_rate": 4.152040197522308e-05, + "loss": 1.2864, + "step": 98380 + }, + { + "epoch": 8.52, + "learning_rate": 4.1519535649311277e-05, + "loss": 1.4053, + "step": 98390 + }, + { + "epoch": 8.52, + "learning_rate": 4.1518669323399464e-05, + "loss": 1.3792, + "step": 98400 + }, + { + "epoch": 8.52, + "learning_rate": 4.151780299748766e-05, + "loss": 1.3728, + "step": 98410 + }, + { + "epoch": 8.52, + "learning_rate": 4.151693667157585e-05, + "loss": 1.3847, + "step": 98420 + }, + { + "epoch": 8.52, + "learning_rate": 4.151607034566404e-05, + "loss": 1.3609, + "step": 98430 + }, + { + "epoch": 8.52, + "learning_rate": 4.1515204019752235e-05, + "loss": 1.3002, + "step": 98440 + }, + { + "epoch": 8.52, + "learning_rate": 4.151433769384043e-05, + "loss": 1.3509, + "step": 98450 + }, + { + "epoch": 8.52, + "learning_rate": 4.151347136792862e-05, + "loss": 1.379, + "step": 98460 + }, + { + "epoch": 8.52, + "learning_rate": 4.151260504201681e-05, + "loss": 1.3741, + "step": 98470 + }, + { + "epoch": 8.52, + "learning_rate": 4.1511738716105006e-05, + "loss": 1.4161, + "step": 98480 + }, + { + "epoch": 8.52, + "learning_rate": 4.1510872390193194e-05, + "loss": 1.3395, + "step": 98490 + }, + { + "epoch": 8.53, + "learning_rate": 4.151000606428139e-05, + "loss": 1.345, + "step": 98500 + }, + { + "epoch": 8.53, + "learning_rate": 4.1509139738369576e-05, + "loss": 1.2805, + "step": 98510 + }, + { + "epoch": 8.53, + "learning_rate": 4.150827341245777e-05, + "loss": 1.373, + "step": 98520 + }, + { + "epoch": 8.53, + "learning_rate": 4.1507407086545965e-05, + "loss": 1.3393, + "step": 98530 + }, + { + "epoch": 8.53, + "learning_rate": 4.150654076063415e-05, + "loss": 1.3759, + "step": 98540 + }, + { + "epoch": 8.53, + "learning_rate": 4.150567443472235e-05, + "loss": 1.3516, + "step": 98550 + }, + { + "epoch": 8.53, + "learning_rate": 4.150480810881054e-05, + "loss": 1.3602, + "step": 98560 + }, + { + "epoch": 8.53, + "learning_rate": 4.150394178289873e-05, + "loss": 1.3388, + "step": 98570 + }, + { + "epoch": 8.53, + "learning_rate": 4.1503075456986924e-05, + "loss": 1.3185, + "step": 98580 + }, + { + "epoch": 8.53, + "learning_rate": 4.150220913107512e-05, + "loss": 1.3459, + "step": 98590 + }, + { + "epoch": 8.53, + "learning_rate": 4.1501342805163306e-05, + "loss": 1.3714, + "step": 98600 + }, + { + "epoch": 8.54, + "learning_rate": 4.15004764792515e-05, + "loss": 1.3646, + "step": 98610 + }, + { + "epoch": 8.54, + "learning_rate": 4.149961015333969e-05, + "loss": 1.3076, + "step": 98620 + }, + { + "epoch": 8.54, + "learning_rate": 4.149874382742788e-05, + "loss": 1.3481, + "step": 98630 + }, + { + "epoch": 8.54, + "learning_rate": 4.149787750151608e-05, + "loss": 1.3493, + "step": 98640 + }, + { + "epoch": 8.54, + "learning_rate": 4.1497011175604265e-05, + "loss": 1.3551, + "step": 98650 + }, + { + "epoch": 8.54, + "learning_rate": 4.149614484969246e-05, + "loss": 1.3763, + "step": 98660 + }, + { + "epoch": 8.54, + "learning_rate": 4.1495278523780653e-05, + "loss": 1.3531, + "step": 98670 + }, + { + "epoch": 8.54, + "learning_rate": 4.149441219786884e-05, + "loss": 1.3735, + "step": 98680 + }, + { + "epoch": 8.54, + "learning_rate": 4.1493545871957036e-05, + "loss": 1.3591, + "step": 98690 + }, + { + "epoch": 8.54, + "learning_rate": 4.149267954604523e-05, + "loss": 1.36, + "step": 98700 + }, + { + "epoch": 8.54, + "learning_rate": 4.149181322013342e-05, + "loss": 1.3627, + "step": 98710 + }, + { + "epoch": 8.54, + "learning_rate": 4.149094689422161e-05, + "loss": 1.3403, + "step": 98720 + }, + { + "epoch": 8.55, + "learning_rate": 4.14900805683098e-05, + "loss": 1.3889, + "step": 98730 + }, + { + "epoch": 8.55, + "learning_rate": 4.1489214242397994e-05, + "loss": 1.3378, + "step": 98740 + }, + { + "epoch": 8.55, + "learning_rate": 4.148834791648619e-05, + "loss": 1.3746, + "step": 98750 + }, + { + "epoch": 8.55, + "learning_rate": 4.1487481590574376e-05, + "loss": 1.4011, + "step": 98760 + }, + { + "epoch": 8.55, + "learning_rate": 4.148661526466257e-05, + "loss": 1.3931, + "step": 98770 + }, + { + "epoch": 8.55, + "learning_rate": 4.1485748938750765e-05, + "loss": 1.2802, + "step": 98780 + }, + { + "epoch": 8.55, + "learning_rate": 4.148488261283895e-05, + "loss": 1.3783, + "step": 98790 + }, + { + "epoch": 8.55, + "learning_rate": 4.148401628692715e-05, + "loss": 1.3549, + "step": 98800 + }, + { + "epoch": 8.55, + "learning_rate": 4.1483149961015335e-05, + "loss": 1.3025, + "step": 98810 + }, + { + "epoch": 8.55, + "learning_rate": 4.148228363510353e-05, + "loss": 1.3653, + "step": 98820 + }, + { + "epoch": 8.55, + "learning_rate": 4.1481417309191724e-05, + "loss": 1.3641, + "step": 98830 + }, + { + "epoch": 8.56, + "learning_rate": 4.148055098327991e-05, + "loss": 1.2966, + "step": 98840 + }, + { + "epoch": 8.56, + "learning_rate": 4.1479684657368106e-05, + "loss": 1.3365, + "step": 98850 + }, + { + "epoch": 8.56, + "learning_rate": 4.14788183314563e-05, + "loss": 1.4026, + "step": 98860 + }, + { + "epoch": 8.56, + "learning_rate": 4.147795200554449e-05, + "loss": 1.3253, + "step": 98870 + }, + { + "epoch": 8.56, + "learning_rate": 4.147708567963268e-05, + "loss": 1.3967, + "step": 98880 + }, + { + "epoch": 8.56, + "learning_rate": 4.147621935372088e-05, + "loss": 1.3532, + "step": 98890 + }, + { + "epoch": 8.56, + "learning_rate": 4.1475353027809065e-05, + "loss": 1.3172, + "step": 98900 + }, + { + "epoch": 8.56, + "learning_rate": 4.147448670189726e-05, + "loss": 1.3279, + "step": 98910 + }, + { + "epoch": 8.56, + "learning_rate": 4.147362037598545e-05, + "loss": 1.3962, + "step": 98920 + }, + { + "epoch": 8.56, + "learning_rate": 4.147275405007364e-05, + "loss": 1.402, + "step": 98930 + }, + { + "epoch": 8.56, + "learning_rate": 4.1471887724161836e-05, + "loss": 1.3704, + "step": 98940 + }, + { + "epoch": 8.56, + "learning_rate": 4.1471021398250024e-05, + "loss": 1.3464, + "step": 98950 + }, + { + "epoch": 8.57, + "learning_rate": 4.147015507233822e-05, + "loss": 1.3531, + "step": 98960 + }, + { + "epoch": 8.57, + "learning_rate": 4.146928874642641e-05, + "loss": 1.3699, + "step": 98970 + }, + { + "epoch": 8.57, + "learning_rate": 4.14684224205146e-05, + "loss": 1.3358, + "step": 98980 + }, + { + "epoch": 8.57, + "learning_rate": 4.1467556094602795e-05, + "loss": 1.3488, + "step": 98990 + }, + { + "epoch": 8.57, + "learning_rate": 4.146668976869099e-05, + "loss": 1.3897, + "step": 99000 + }, + { + "epoch": 8.57, + "learning_rate": 4.146582344277918e-05, + "loss": 1.3596, + "step": 99010 + }, + { + "epoch": 8.57, + "learning_rate": 4.146495711686737e-05, + "loss": 1.3225, + "step": 99020 + }, + { + "epoch": 8.57, + "learning_rate": 4.146409079095556e-05, + "loss": 1.4174, + "step": 99030 + }, + { + "epoch": 8.57, + "learning_rate": 4.146322446504375e-05, + "loss": 1.3337, + "step": 99040 + }, + { + "epoch": 8.57, + "learning_rate": 4.146235813913195e-05, + "loss": 1.3233, + "step": 99050 + }, + { + "epoch": 8.57, + "learning_rate": 4.1461491813220135e-05, + "loss": 1.3654, + "step": 99060 + }, + { + "epoch": 8.58, + "learning_rate": 4.146062548730833e-05, + "loss": 1.3899, + "step": 99070 + }, + { + "epoch": 8.58, + "learning_rate": 4.1459759161396524e-05, + "loss": 1.3156, + "step": 99080 + }, + { + "epoch": 8.58, + "learning_rate": 4.145889283548471e-05, + "loss": 1.3761, + "step": 99090 + }, + { + "epoch": 8.58, + "learning_rate": 4.1458026509572906e-05, + "loss": 1.3301, + "step": 99100 + }, + { + "epoch": 8.58, + "learning_rate": 4.14571601836611e-05, + "loss": 1.3176, + "step": 99110 + }, + { + "epoch": 8.58, + "learning_rate": 4.145629385774929e-05, + "loss": 1.3119, + "step": 99120 + }, + { + "epoch": 8.58, + "learning_rate": 4.145542753183748e-05, + "loss": 1.3185, + "step": 99130 + }, + { + "epoch": 8.58, + "learning_rate": 4.145456120592567e-05, + "loss": 1.3362, + "step": 99140 + }, + { + "epoch": 8.58, + "learning_rate": 4.1453694880013865e-05, + "loss": 1.338, + "step": 99150 + }, + { + "epoch": 8.58, + "learning_rate": 4.145282855410206e-05, + "loss": 1.3215, + "step": 99160 + }, + { + "epoch": 8.58, + "learning_rate": 4.145196222819025e-05, + "loss": 1.3579, + "step": 99170 + }, + { + "epoch": 8.58, + "learning_rate": 4.145109590227844e-05, + "loss": 1.304, + "step": 99180 + }, + { + "epoch": 8.59, + "learning_rate": 4.1450229576366636e-05, + "loss": 1.3345, + "step": 99190 + }, + { + "epoch": 8.59, + "learning_rate": 4.1449363250454824e-05, + "loss": 1.2928, + "step": 99200 + }, + { + "epoch": 8.59, + "learning_rate": 4.144849692454302e-05, + "loss": 1.3635, + "step": 99210 + }, + { + "epoch": 8.59, + "learning_rate": 4.144763059863121e-05, + "loss": 1.289, + "step": 99220 + }, + { + "epoch": 8.59, + "learning_rate": 4.14467642727194e-05, + "loss": 1.3672, + "step": 99230 + }, + { + "epoch": 8.59, + "learning_rate": 4.1445897946807595e-05, + "loss": 1.3926, + "step": 99240 + }, + { + "epoch": 8.59, + "learning_rate": 4.144503162089578e-05, + "loss": 1.4025, + "step": 99250 + }, + { + "epoch": 8.59, + "learning_rate": 4.144416529498398e-05, + "loss": 1.3432, + "step": 99260 + }, + { + "epoch": 8.59, + "learning_rate": 4.144329896907217e-05, + "loss": 1.3883, + "step": 99270 + }, + { + "epoch": 8.59, + "learning_rate": 4.144243264316036e-05, + "loss": 1.3252, + "step": 99280 + }, + { + "epoch": 8.59, + "learning_rate": 4.1441566317248554e-05, + "loss": 1.2476, + "step": 99290 + }, + { + "epoch": 8.6, + "learning_rate": 4.144069999133675e-05, + "loss": 1.3307, + "step": 99300 + }, + { + "epoch": 8.6, + "learning_rate": 4.1439833665424936e-05, + "loss": 1.3051, + "step": 99310 + }, + { + "epoch": 8.6, + "learning_rate": 4.143896733951313e-05, + "loss": 1.3169, + "step": 99320 + }, + { + "epoch": 8.6, + "learning_rate": 4.1438101013601325e-05, + "loss": 1.3595, + "step": 99330 + }, + { + "epoch": 8.6, + "learning_rate": 4.143723468768951e-05, + "loss": 1.403, + "step": 99340 + }, + { + "epoch": 8.6, + "learning_rate": 4.143636836177771e-05, + "loss": 1.3235, + "step": 99350 + }, + { + "epoch": 8.6, + "learning_rate": 4.1435502035865894e-05, + "loss": 1.393, + "step": 99360 + }, + { + "epoch": 8.6, + "learning_rate": 4.143463570995409e-05, + "loss": 1.3282, + "step": 99370 + }, + { + "epoch": 8.6, + "learning_rate": 4.1433769384042283e-05, + "loss": 1.3083, + "step": 99380 + }, + { + "epoch": 8.6, + "learning_rate": 4.143290305813047e-05, + "loss": 1.3251, + "step": 99390 + }, + { + "epoch": 8.6, + "learning_rate": 4.1432036732218666e-05, + "loss": 1.3313, + "step": 99400 + }, + { + "epoch": 8.6, + "learning_rate": 4.143117040630686e-05, + "loss": 1.3375, + "step": 99410 + }, + { + "epoch": 8.61, + "learning_rate": 4.143030408039505e-05, + "loss": 1.3625, + "step": 99420 + }, + { + "epoch": 8.61, + "learning_rate": 4.142943775448324e-05, + "loss": 1.3123, + "step": 99430 + }, + { + "epoch": 8.61, + "learning_rate": 4.1428571428571437e-05, + "loss": 1.3221, + "step": 99440 + }, + { + "epoch": 8.61, + "learning_rate": 4.1427705102659624e-05, + "loss": 1.3597, + "step": 99450 + }, + { + "epoch": 8.61, + "learning_rate": 4.142683877674782e-05, + "loss": 1.3846, + "step": 99460 + }, + { + "epoch": 8.61, + "learning_rate": 4.1425972450836006e-05, + "loss": 1.4208, + "step": 99470 + }, + { + "epoch": 8.61, + "learning_rate": 4.14251061249242e-05, + "loss": 1.3534, + "step": 99480 + }, + { + "epoch": 8.61, + "learning_rate": 4.1424239799012395e-05, + "loss": 1.2875, + "step": 99490 + }, + { + "epoch": 8.61, + "learning_rate": 4.142337347310058e-05, + "loss": 1.3954, + "step": 99500 + }, + { + "epoch": 8.61, + "learning_rate": 4.142250714718878e-05, + "loss": 1.3186, + "step": 99510 + }, + { + "epoch": 8.61, + "learning_rate": 4.142164082127697e-05, + "loss": 1.3875, + "step": 99520 + }, + { + "epoch": 8.61, + "learning_rate": 4.142077449536516e-05, + "loss": 1.326, + "step": 99530 + }, + { + "epoch": 8.62, + "learning_rate": 4.1419908169453354e-05, + "loss": 1.3872, + "step": 99540 + }, + { + "epoch": 8.62, + "learning_rate": 4.141904184354154e-05, + "loss": 1.3502, + "step": 99550 + }, + { + "epoch": 8.62, + "learning_rate": 4.1418175517629736e-05, + "loss": 1.3917, + "step": 99560 + }, + { + "epoch": 8.62, + "learning_rate": 4.141730919171793e-05, + "loss": 1.388, + "step": 99570 + }, + { + "epoch": 8.62, + "learning_rate": 4.141644286580612e-05, + "loss": 1.24, + "step": 99580 + }, + { + "epoch": 8.62, + "learning_rate": 4.141557653989431e-05, + "loss": 1.3814, + "step": 99590 + }, + { + "epoch": 8.62, + "learning_rate": 4.141471021398251e-05, + "loss": 1.4291, + "step": 99600 + }, + { + "epoch": 8.62, + "learning_rate": 4.1413843888070695e-05, + "loss": 1.3154, + "step": 99610 + }, + { + "epoch": 8.62, + "learning_rate": 4.141297756215889e-05, + "loss": 1.3076, + "step": 99620 + }, + { + "epoch": 8.62, + "learning_rate": 4.1412111236247084e-05, + "loss": 1.3389, + "step": 99630 + }, + { + "epoch": 8.62, + "learning_rate": 4.141124491033527e-05, + "loss": 1.2932, + "step": 99640 + }, + { + "epoch": 8.63, + "learning_rate": 4.1410378584423466e-05, + "loss": 1.345, + "step": 99650 + }, + { + "epoch": 8.63, + "learning_rate": 4.1409512258511654e-05, + "loss": 1.3049, + "step": 99660 + }, + { + "epoch": 8.63, + "learning_rate": 4.140864593259985e-05, + "loss": 1.3478, + "step": 99670 + }, + { + "epoch": 8.63, + "learning_rate": 4.140777960668804e-05, + "loss": 1.4038, + "step": 99680 + }, + { + "epoch": 8.63, + "learning_rate": 4.140691328077623e-05, + "loss": 1.32, + "step": 99690 + }, + { + "epoch": 8.63, + "learning_rate": 4.140604695486442e-05, + "loss": 1.338, + "step": 99700 + }, + { + "epoch": 8.63, + "learning_rate": 4.140518062895261e-05, + "loss": 1.3543, + "step": 99710 + }, + { + "epoch": 8.63, + "learning_rate": 4.14043143030408e-05, + "loss": 1.2749, + "step": 99720 + }, + { + "epoch": 8.63, + "learning_rate": 4.1403447977128994e-05, + "loss": 1.304, + "step": 99730 + }, + { + "epoch": 8.63, + "learning_rate": 4.140258165121719e-05, + "loss": 1.3769, + "step": 99740 + }, + { + "epoch": 8.63, + "learning_rate": 4.1401715325305377e-05, + "loss": 1.3179, + "step": 99750 + }, + { + "epoch": 8.63, + "learning_rate": 4.140084899939357e-05, + "loss": 1.3165, + "step": 99760 + }, + { + "epoch": 8.64, + "learning_rate": 4.1399982673481765e-05, + "loss": 1.3084, + "step": 99770 + }, + { + "epoch": 8.64, + "learning_rate": 4.139911634756995e-05, + "loss": 1.2804, + "step": 99780 + }, + { + "epoch": 8.64, + "learning_rate": 4.139825002165815e-05, + "loss": 1.3472, + "step": 99790 + }, + { + "epoch": 8.64, + "learning_rate": 4.139738369574634e-05, + "loss": 1.3813, + "step": 99800 + }, + { + "epoch": 8.64, + "learning_rate": 4.139651736983453e-05, + "loss": 1.3074, + "step": 99810 + }, + { + "epoch": 8.64, + "learning_rate": 4.1395651043922724e-05, + "loss": 1.3607, + "step": 99820 + }, + { + "epoch": 8.64, + "learning_rate": 4.139478471801091e-05, + "loss": 1.2911, + "step": 99830 + }, + { + "epoch": 8.64, + "learning_rate": 4.1393918392099106e-05, + "loss": 1.3471, + "step": 99840 + }, + { + "epoch": 8.64, + "learning_rate": 4.13930520661873e-05, + "loss": 1.3543, + "step": 99850 + }, + { + "epoch": 8.64, + "learning_rate": 4.139218574027549e-05, + "loss": 1.3593, + "step": 99860 + }, + { + "epoch": 8.64, + "learning_rate": 4.139131941436368e-05, + "loss": 1.3578, + "step": 99870 + }, + { + "epoch": 8.65, + "learning_rate": 4.139045308845188e-05, + "loss": 1.3282, + "step": 99880 + }, + { + "epoch": 8.65, + "learning_rate": 4.1389586762540065e-05, + "loss": 1.3067, + "step": 99890 + }, + { + "epoch": 8.65, + "learning_rate": 4.138872043662826e-05, + "loss": 1.3865, + "step": 99900 + }, + { + "epoch": 8.65, + "learning_rate": 4.1387854110716454e-05, + "loss": 1.4013, + "step": 99910 + }, + { + "epoch": 8.65, + "learning_rate": 4.138698778480464e-05, + "loss": 1.393, + "step": 99920 + }, + { + "epoch": 8.65, + "learning_rate": 4.1386121458892836e-05, + "loss": 1.3213, + "step": 99930 + }, + { + "epoch": 8.65, + "learning_rate": 4.1385255132981024e-05, + "loss": 1.2628, + "step": 99940 + }, + { + "epoch": 8.65, + "learning_rate": 4.138438880706922e-05, + "loss": 1.2643, + "step": 99950 + }, + { + "epoch": 8.65, + "learning_rate": 4.138352248115741e-05, + "loss": 1.3412, + "step": 99960 + }, + { + "epoch": 8.65, + "learning_rate": 4.13826561552456e-05, + "loss": 1.3101, + "step": 99970 + }, + { + "epoch": 8.65, + "learning_rate": 4.1381789829333795e-05, + "loss": 1.3069, + "step": 99980 + }, + { + "epoch": 8.65, + "learning_rate": 4.138092350342199e-05, + "loss": 1.3946, + "step": 99990 + }, + { + "epoch": 8.66, + "learning_rate": 4.138005717751018e-05, + "loss": 1.3349, + "step": 100000 + }, + { + "epoch": 8.66, + "learning_rate": 4.137919085159837e-05, + "loss": 1.3585, + "step": 100010 + }, + { + "epoch": 8.66, + "learning_rate": 4.1378324525686566e-05, + "loss": 1.3308, + "step": 100020 + }, + { + "epoch": 8.66, + "learning_rate": 4.1377458199774753e-05, + "loss": 1.3194, + "step": 100030 + }, + { + "epoch": 8.66, + "learning_rate": 4.137659187386295e-05, + "loss": 1.3742, + "step": 100040 + }, + { + "epoch": 8.66, + "learning_rate": 4.1375725547951136e-05, + "loss": 1.3912, + "step": 100050 + }, + { + "epoch": 8.66, + "learning_rate": 4.137485922203933e-05, + "loss": 1.3286, + "step": 100060 + }, + { + "epoch": 8.66, + "learning_rate": 4.1373992896127524e-05, + "loss": 1.3712, + "step": 100070 + }, + { + "epoch": 8.66, + "learning_rate": 4.137312657021571e-05, + "loss": 1.3407, + "step": 100080 + }, + { + "epoch": 8.66, + "learning_rate": 4.1372260244303907e-05, + "loss": 1.3606, + "step": 100090 + }, + { + "epoch": 8.66, + "learning_rate": 4.13713939183921e-05, + "loss": 1.3477, + "step": 100100 + }, + { + "epoch": 8.67, + "learning_rate": 4.137052759248029e-05, + "loss": 1.3256, + "step": 100110 + }, + { + "epoch": 8.67, + "learning_rate": 4.136966126656848e-05, + "loss": 1.3755, + "step": 100120 + }, + { + "epoch": 8.67, + "learning_rate": 4.136879494065668e-05, + "loss": 1.3161, + "step": 100130 + }, + { + "epoch": 8.67, + "learning_rate": 4.1367928614744865e-05, + "loss": 1.3108, + "step": 100140 + }, + { + "epoch": 8.67, + "learning_rate": 4.136706228883306e-05, + "loss": 1.393, + "step": 100150 + }, + { + "epoch": 8.67, + "learning_rate": 4.136619596292125e-05, + "loss": 1.3331, + "step": 100160 + }, + { + "epoch": 8.67, + "learning_rate": 4.136532963700944e-05, + "loss": 1.3172, + "step": 100170 + }, + { + "epoch": 8.67, + "learning_rate": 4.1364463311097636e-05, + "loss": 1.344, + "step": 100180 + }, + { + "epoch": 8.67, + "learning_rate": 4.1363596985185824e-05, + "loss": 1.2836, + "step": 100190 + }, + { + "epoch": 8.67, + "learning_rate": 4.136273065927402e-05, + "loss": 1.2783, + "step": 100200 + }, + { + "epoch": 8.67, + "learning_rate": 4.136186433336221e-05, + "loss": 1.3303, + "step": 100210 + }, + { + "epoch": 8.67, + "learning_rate": 4.13609980074504e-05, + "loss": 1.2631, + "step": 100220 + }, + { + "epoch": 8.68, + "learning_rate": 4.1360131681538595e-05, + "loss": 1.3433, + "step": 100230 + }, + { + "epoch": 8.68, + "learning_rate": 4.135926535562679e-05, + "loss": 1.3033, + "step": 100240 + }, + { + "epoch": 8.68, + "learning_rate": 4.135839902971498e-05, + "loss": 1.3092, + "step": 100250 + }, + { + "epoch": 8.68, + "learning_rate": 4.135753270380317e-05, + "loss": 1.326, + "step": 100260 + }, + { + "epoch": 8.68, + "learning_rate": 4.135666637789136e-05, + "loss": 1.3472, + "step": 100270 + }, + { + "epoch": 8.68, + "learning_rate": 4.1355800051979554e-05, + "loss": 1.2912, + "step": 100280 + }, + { + "epoch": 8.68, + "learning_rate": 4.135493372606775e-05, + "loss": 1.2868, + "step": 100290 + }, + { + "epoch": 8.68, + "learning_rate": 4.1354067400155936e-05, + "loss": 1.3791, + "step": 100300 + }, + { + "epoch": 8.68, + "learning_rate": 4.135320107424413e-05, + "loss": 1.4048, + "step": 100310 + }, + { + "epoch": 8.68, + "learning_rate": 4.1352334748332325e-05, + "loss": 1.3692, + "step": 100320 + }, + { + "epoch": 8.68, + "learning_rate": 4.135146842242051e-05, + "loss": 1.3697, + "step": 100330 + }, + { + "epoch": 8.69, + "learning_rate": 4.135060209650871e-05, + "loss": 1.3902, + "step": 100340 + }, + { + "epoch": 8.69, + "learning_rate": 4.1349735770596895e-05, + "loss": 1.3347, + "step": 100350 + }, + { + "epoch": 8.69, + "learning_rate": 4.134886944468509e-05, + "loss": 1.3387, + "step": 100360 + }, + { + "epoch": 8.69, + "learning_rate": 4.1348003118773284e-05, + "loss": 1.3699, + "step": 100370 + }, + { + "epoch": 8.69, + "learning_rate": 4.134713679286147e-05, + "loss": 1.3735, + "step": 100380 + }, + { + "epoch": 8.69, + "learning_rate": 4.1346270466949666e-05, + "loss": 1.3031, + "step": 100390 + }, + { + "epoch": 8.69, + "learning_rate": 4.134540414103786e-05, + "loss": 1.4274, + "step": 100400 + }, + { + "epoch": 8.69, + "learning_rate": 4.134453781512605e-05, + "loss": 1.3357, + "step": 100410 + }, + { + "epoch": 8.69, + "learning_rate": 4.134367148921424e-05, + "loss": 1.3333, + "step": 100420 + }, + { + "epoch": 8.69, + "learning_rate": 4.134280516330244e-05, + "loss": 1.3328, + "step": 100430 + }, + { + "epoch": 8.69, + "learning_rate": 4.1341938837390624e-05, + "loss": 1.363, + "step": 100440 + }, + { + "epoch": 8.69, + "learning_rate": 4.134107251147882e-05, + "loss": 1.3865, + "step": 100450 + }, + { + "epoch": 8.7, + "learning_rate": 4.1340206185567006e-05, + "loss": 1.2984, + "step": 100460 + }, + { + "epoch": 8.7, + "learning_rate": 4.13393398596552e-05, + "loss": 1.4207, + "step": 100470 + }, + { + "epoch": 8.7, + "learning_rate": 4.1338473533743395e-05, + "loss": 1.4145, + "step": 100480 + }, + { + "epoch": 8.7, + "learning_rate": 4.133760720783158e-05, + "loss": 1.3819, + "step": 100490 + }, + { + "epoch": 8.7, + "learning_rate": 4.133674088191978e-05, + "loss": 1.4084, + "step": 100500 + }, + { + "epoch": 8.7, + "learning_rate": 4.133587455600797e-05, + "loss": 1.3577, + "step": 100510 + }, + { + "epoch": 8.7, + "learning_rate": 4.133500823009616e-05, + "loss": 1.2535, + "step": 100520 + }, + { + "epoch": 8.7, + "learning_rate": 4.1334141904184354e-05, + "loss": 1.3615, + "step": 100530 + }, + { + "epoch": 8.7, + "learning_rate": 4.133327557827255e-05, + "loss": 1.3479, + "step": 100540 + }, + { + "epoch": 8.7, + "learning_rate": 4.1332409252360736e-05, + "loss": 1.4258, + "step": 100550 + }, + { + "epoch": 8.7, + "learning_rate": 4.133154292644893e-05, + "loss": 1.4037, + "step": 100560 + }, + { + "epoch": 8.7, + "learning_rate": 4.133067660053712e-05, + "loss": 1.2906, + "step": 100570 + }, + { + "epoch": 8.71, + "learning_rate": 4.132981027462531e-05, + "loss": 1.347, + "step": 100580 + }, + { + "epoch": 8.71, + "learning_rate": 4.132894394871351e-05, + "loss": 1.344, + "step": 100590 + }, + { + "epoch": 8.71, + "learning_rate": 4.1328077622801695e-05, + "loss": 1.3261, + "step": 100600 + }, + { + "epoch": 8.71, + "learning_rate": 4.132721129688989e-05, + "loss": 1.3111, + "step": 100610 + }, + { + "epoch": 8.71, + "learning_rate": 4.1326344970978084e-05, + "loss": 1.3116, + "step": 100620 + }, + { + "epoch": 8.71, + "learning_rate": 4.132547864506627e-05, + "loss": 1.388, + "step": 100630 + }, + { + "epoch": 8.71, + "learning_rate": 4.1324612319154466e-05, + "loss": 1.3448, + "step": 100640 + }, + { + "epoch": 8.71, + "learning_rate": 4.132374599324266e-05, + "loss": 1.2725, + "step": 100650 + }, + { + "epoch": 8.71, + "learning_rate": 4.132287966733085e-05, + "loss": 1.3685, + "step": 100660 + }, + { + "epoch": 8.71, + "learning_rate": 4.132201334141904e-05, + "loss": 1.3034, + "step": 100670 + }, + { + "epoch": 8.71, + "learning_rate": 4.132114701550723e-05, + "loss": 1.401, + "step": 100680 + }, + { + "epoch": 8.72, + "learning_rate": 4.1320280689595425e-05, + "loss": 1.3819, + "step": 100690 + }, + { + "epoch": 8.72, + "learning_rate": 4.131941436368362e-05, + "loss": 1.348, + "step": 100700 + }, + { + "epoch": 8.72, + "learning_rate": 4.131854803777181e-05, + "loss": 1.3618, + "step": 100710 + }, + { + "epoch": 8.72, + "learning_rate": 4.131768171186e-05, + "loss": 1.3169, + "step": 100720 + }, + { + "epoch": 8.72, + "learning_rate": 4.1316815385948196e-05, + "loss": 1.3149, + "step": 100730 + }, + { + "epoch": 8.72, + "learning_rate": 4.1315949060036383e-05, + "loss": 1.3403, + "step": 100740 + }, + { + "epoch": 8.72, + "learning_rate": 4.131508273412458e-05, + "loss": 1.3054, + "step": 100750 + }, + { + "epoch": 8.72, + "learning_rate": 4.131421640821277e-05, + "loss": 1.2825, + "step": 100760 + }, + { + "epoch": 8.72, + "learning_rate": 4.131335008230096e-05, + "loss": 1.3792, + "step": 100770 + }, + { + "epoch": 8.72, + "learning_rate": 4.1312483756389154e-05, + "loss": 1.3283, + "step": 100780 + }, + { + "epoch": 8.72, + "learning_rate": 4.131161743047734e-05, + "loss": 1.4031, + "step": 100790 + }, + { + "epoch": 8.72, + "learning_rate": 4.1310751104565537e-05, + "loss": 1.3925, + "step": 100800 + }, + { + "epoch": 8.73, + "learning_rate": 4.130988477865373e-05, + "loss": 1.3352, + "step": 100810 + }, + { + "epoch": 8.73, + "learning_rate": 4.130901845274192e-05, + "loss": 1.367, + "step": 100820 + }, + { + "epoch": 8.73, + "learning_rate": 4.130815212683011e-05, + "loss": 1.3221, + "step": 100830 + }, + { + "epoch": 8.73, + "learning_rate": 4.130728580091831e-05, + "loss": 1.3037, + "step": 100840 + }, + { + "epoch": 8.73, + "learning_rate": 4.1306419475006495e-05, + "loss": 1.3188, + "step": 100850 + }, + { + "epoch": 8.73, + "learning_rate": 4.130555314909469e-05, + "loss": 1.3322, + "step": 100860 + }, + { + "epoch": 8.73, + "learning_rate": 4.1304686823182884e-05, + "loss": 1.2871, + "step": 100870 + }, + { + "epoch": 8.73, + "learning_rate": 4.130382049727107e-05, + "loss": 1.2916, + "step": 100880 + }, + { + "epoch": 8.73, + "learning_rate": 4.1302954171359266e-05, + "loss": 1.3218, + "step": 100890 + }, + { + "epoch": 8.73, + "learning_rate": 4.1302087845447454e-05, + "loss": 1.3361, + "step": 100900 + }, + { + "epoch": 8.73, + "learning_rate": 4.130122151953565e-05, + "loss": 1.3184, + "step": 100910 + }, + { + "epoch": 8.74, + "learning_rate": 4.130035519362384e-05, + "loss": 1.273, + "step": 100920 + }, + { + "epoch": 8.74, + "learning_rate": 4.129948886771203e-05, + "loss": 1.3431, + "step": 100930 + }, + { + "epoch": 8.74, + "learning_rate": 4.1298622541800225e-05, + "loss": 1.3327, + "step": 100940 + }, + { + "epoch": 8.74, + "learning_rate": 4.129775621588842e-05, + "loss": 1.3355, + "step": 100950 + }, + { + "epoch": 8.74, + "learning_rate": 4.129688988997661e-05, + "loss": 1.2608, + "step": 100960 + }, + { + "epoch": 8.74, + "learning_rate": 4.12960235640648e-05, + "loss": 1.2849, + "step": 100970 + }, + { + "epoch": 8.74, + "learning_rate": 4.1295157238152996e-05, + "loss": 1.4125, + "step": 100980 + }, + { + "epoch": 8.74, + "learning_rate": 4.1294290912241184e-05, + "loss": 1.347, + "step": 100990 + }, + { + "epoch": 8.74, + "learning_rate": 4.129342458632938e-05, + "loss": 1.3753, + "step": 101000 + }, + { + "epoch": 8.74, + "learning_rate": 4.1292558260417566e-05, + "loss": 1.4303, + "step": 101010 + }, + { + "epoch": 8.74, + "learning_rate": 4.129169193450576e-05, + "loss": 1.3745, + "step": 101020 + }, + { + "epoch": 8.74, + "learning_rate": 4.1290825608593955e-05, + "loss": 1.3735, + "step": 101030 + }, + { + "epoch": 8.75, + "learning_rate": 4.128995928268214e-05, + "loss": 1.3487, + "step": 101040 + }, + { + "epoch": 8.75, + "learning_rate": 4.128909295677034e-05, + "loss": 1.3114, + "step": 101050 + }, + { + "epoch": 8.75, + "learning_rate": 4.128822663085853e-05, + "loss": 1.3905, + "step": 101060 + }, + { + "epoch": 8.75, + "learning_rate": 4.128736030494672e-05, + "loss": 1.2865, + "step": 101070 + }, + { + "epoch": 8.75, + "learning_rate": 4.1286493979034913e-05, + "loss": 1.3379, + "step": 101080 + }, + { + "epoch": 8.75, + "learning_rate": 4.12856276531231e-05, + "loss": 1.3008, + "step": 101090 + }, + { + "epoch": 8.75, + "learning_rate": 4.1284761327211296e-05, + "loss": 1.3611, + "step": 101100 + }, + { + "epoch": 8.75, + "learning_rate": 4.128389500129949e-05, + "loss": 1.374, + "step": 101110 + }, + { + "epoch": 8.75, + "learning_rate": 4.128302867538768e-05, + "loss": 1.2991, + "step": 101120 + }, + { + "epoch": 8.75, + "learning_rate": 4.128216234947587e-05, + "loss": 1.3169, + "step": 101130 + }, + { + "epoch": 8.75, + "learning_rate": 4.128129602356407e-05, + "loss": 1.2713, + "step": 101140 + }, + { + "epoch": 8.76, + "learning_rate": 4.1280429697652254e-05, + "loss": 1.2735, + "step": 101150 + }, + { + "epoch": 8.76, + "learning_rate": 4.127956337174045e-05, + "loss": 1.3983, + "step": 101160 + }, + { + "epoch": 8.76, + "learning_rate": 4.127869704582864e-05, + "loss": 1.2897, + "step": 101170 + }, + { + "epoch": 8.76, + "learning_rate": 4.127783071991683e-05, + "loss": 1.3197, + "step": 101180 + }, + { + "epoch": 8.76, + "learning_rate": 4.1276964394005025e-05, + "loss": 1.3219, + "step": 101190 + }, + { + "epoch": 8.76, + "learning_rate": 4.127609806809321e-05, + "loss": 1.3264, + "step": 101200 + }, + { + "epoch": 8.76, + "learning_rate": 4.127523174218141e-05, + "loss": 1.3662, + "step": 101210 + }, + { + "epoch": 8.76, + "learning_rate": 4.12743654162696e-05, + "loss": 1.3096, + "step": 101220 + }, + { + "epoch": 8.76, + "learning_rate": 4.127349909035779e-05, + "loss": 1.3731, + "step": 101230 + }, + { + "epoch": 8.76, + "learning_rate": 4.1272632764445984e-05, + "loss": 1.2871, + "step": 101240 + }, + { + "epoch": 8.76, + "learning_rate": 4.127176643853418e-05, + "loss": 1.333, + "step": 101250 + }, + { + "epoch": 8.76, + "learning_rate": 4.1270900112622366e-05, + "loss": 1.3489, + "step": 101260 + }, + { + "epoch": 8.77, + "learning_rate": 4.127003378671056e-05, + "loss": 1.3644, + "step": 101270 + }, + { + "epoch": 8.77, + "learning_rate": 4.1269167460798755e-05, + "loss": 1.3076, + "step": 101280 + }, + { + "epoch": 8.77, + "learning_rate": 4.126830113488694e-05, + "loss": 1.3905, + "step": 101290 + }, + { + "epoch": 8.77, + "learning_rate": 4.126743480897514e-05, + "loss": 1.3324, + "step": 101300 + }, + { + "epoch": 8.77, + "learning_rate": 4.1266568483063325e-05, + "loss": 1.3397, + "step": 101310 + }, + { + "epoch": 8.77, + "learning_rate": 4.126570215715152e-05, + "loss": 1.3001, + "step": 101320 + }, + { + "epoch": 8.77, + "learning_rate": 4.1264835831239714e-05, + "loss": 1.3334, + "step": 101330 + }, + { + "epoch": 8.77, + "learning_rate": 4.12639695053279e-05, + "loss": 1.3577, + "step": 101340 + }, + { + "epoch": 8.77, + "learning_rate": 4.1263103179416096e-05, + "loss": 1.3233, + "step": 101350 + }, + { + "epoch": 8.77, + "learning_rate": 4.126223685350429e-05, + "loss": 1.3109, + "step": 101360 + }, + { + "epoch": 8.77, + "learning_rate": 4.126137052759248e-05, + "loss": 1.2989, + "step": 101370 + }, + { + "epoch": 8.78, + "learning_rate": 4.126050420168067e-05, + "loss": 1.3475, + "step": 101380 + }, + { + "epoch": 8.78, + "learning_rate": 4.125963787576887e-05, + "loss": 1.3772, + "step": 101390 + }, + { + "epoch": 8.78, + "learning_rate": 4.1258771549857055e-05, + "loss": 1.4023, + "step": 101400 + }, + { + "epoch": 8.78, + "learning_rate": 4.125790522394525e-05, + "loss": 1.2729, + "step": 101410 + }, + { + "epoch": 8.78, + "learning_rate": 4.125703889803344e-05, + "loss": 1.2912, + "step": 101420 + }, + { + "epoch": 8.78, + "learning_rate": 4.125617257212163e-05, + "loss": 1.3624, + "step": 101430 + }, + { + "epoch": 8.78, + "learning_rate": 4.1255306246209826e-05, + "loss": 1.4124, + "step": 101440 + }, + { + "epoch": 8.78, + "learning_rate": 4.125443992029801e-05, + "loss": 1.3322, + "step": 101450 + }, + { + "epoch": 8.78, + "learning_rate": 4.125357359438621e-05, + "loss": 1.3659, + "step": 101460 + }, + { + "epoch": 8.78, + "learning_rate": 4.12527072684744e-05, + "loss": 1.3791, + "step": 101470 + }, + { + "epoch": 8.78, + "learning_rate": 4.125184094256259e-05, + "loss": 1.2949, + "step": 101480 + }, + { + "epoch": 8.78, + "learning_rate": 4.1250974616650784e-05, + "loss": 1.321, + "step": 101490 + }, + { + "epoch": 8.79, + "learning_rate": 4.125010829073898e-05, + "loss": 1.2957, + "step": 101500 + }, + { + "epoch": 8.79, + "learning_rate": 4.1249241964827167e-05, + "loss": 1.2899, + "step": 101510 + }, + { + "epoch": 8.79, + "learning_rate": 4.124837563891536e-05, + "loss": 1.2783, + "step": 101520 + }, + { + "epoch": 8.79, + "learning_rate": 4.124750931300355e-05, + "loss": 1.3589, + "step": 101530 + }, + { + "epoch": 8.79, + "learning_rate": 4.124664298709174e-05, + "loss": 1.3678, + "step": 101540 + }, + { + "epoch": 8.79, + "learning_rate": 4.124577666117994e-05, + "loss": 1.3832, + "step": 101550 + }, + { + "epoch": 8.79, + "learning_rate": 4.1244910335268125e-05, + "loss": 1.3052, + "step": 101560 + }, + { + "epoch": 8.79, + "learning_rate": 4.124404400935632e-05, + "loss": 1.3431, + "step": 101570 + }, + { + "epoch": 8.79, + "learning_rate": 4.1243177683444514e-05, + "loss": 1.3612, + "step": 101580 + }, + { + "epoch": 8.79, + "learning_rate": 4.12423113575327e-05, + "loss": 1.2722, + "step": 101590 + }, + { + "epoch": 8.79, + "learning_rate": 4.1241445031620896e-05, + "loss": 1.2521, + "step": 101600 + }, + { + "epoch": 8.79, + "learning_rate": 4.124057870570909e-05, + "loss": 1.3149, + "step": 101610 + }, + { + "epoch": 8.8, + "learning_rate": 4.123971237979728e-05, + "loss": 1.3214, + "step": 101620 + }, + { + "epoch": 8.8, + "learning_rate": 4.123884605388547e-05, + "loss": 1.3263, + "step": 101630 + }, + { + "epoch": 8.8, + "learning_rate": 4.123797972797366e-05, + "loss": 1.3302, + "step": 101640 + }, + { + "epoch": 8.8, + "learning_rate": 4.1237113402061855e-05, + "loss": 1.4234, + "step": 101650 + }, + { + "epoch": 8.8, + "learning_rate": 4.123624707615005e-05, + "loss": 1.3785, + "step": 101660 + }, + { + "epoch": 8.8, + "learning_rate": 4.123538075023824e-05, + "loss": 1.3365, + "step": 101670 + }, + { + "epoch": 8.8, + "learning_rate": 4.123451442432643e-05, + "loss": 1.3318, + "step": 101680 + }, + { + "epoch": 8.8, + "learning_rate": 4.1233648098414626e-05, + "loss": 1.3279, + "step": 101690 + }, + { + "epoch": 8.8, + "learning_rate": 4.1232781772502814e-05, + "loss": 1.3673, + "step": 101700 + }, + { + "epoch": 8.8, + "learning_rate": 4.123191544659101e-05, + "loss": 1.3299, + "step": 101710 + }, + { + "epoch": 8.8, + "learning_rate": 4.12310491206792e-05, + "loss": 1.3305, + "step": 101720 + }, + { + "epoch": 8.81, + "learning_rate": 4.123018279476739e-05, + "loss": 1.319, + "step": 101730 + }, + { + "epoch": 8.81, + "learning_rate": 4.1229316468855585e-05, + "loss": 1.2966, + "step": 101740 + }, + { + "epoch": 8.81, + "learning_rate": 4.122845014294377e-05, + "loss": 1.376, + "step": 101750 + }, + { + "epoch": 8.81, + "learning_rate": 4.122758381703197e-05, + "loss": 1.2742, + "step": 101760 + }, + { + "epoch": 8.81, + "learning_rate": 4.122671749112016e-05, + "loss": 1.3182, + "step": 101770 + }, + { + "epoch": 8.81, + "learning_rate": 4.122585116520835e-05, + "loss": 1.3881, + "step": 101780 + }, + { + "epoch": 8.81, + "learning_rate": 4.1224984839296543e-05, + "loss": 1.2806, + "step": 101790 + }, + { + "epoch": 8.81, + "learning_rate": 4.122411851338474e-05, + "loss": 1.3332, + "step": 101800 + }, + { + "epoch": 8.81, + "learning_rate": 4.1223252187472926e-05, + "loss": 1.3489, + "step": 101810 + }, + { + "epoch": 8.81, + "learning_rate": 4.122238586156112e-05, + "loss": 1.3531, + "step": 101820 + }, + { + "epoch": 8.81, + "learning_rate": 4.1221519535649314e-05, + "loss": 1.325, + "step": 101830 + }, + { + "epoch": 8.81, + "learning_rate": 4.12206532097375e-05, + "loss": 1.3228, + "step": 101840 + }, + { + "epoch": 8.82, + "learning_rate": 4.12197868838257e-05, + "loss": 1.2739, + "step": 101850 + }, + { + "epoch": 8.82, + "learning_rate": 4.1218920557913884e-05, + "loss": 1.3261, + "step": 101860 + }, + { + "epoch": 8.82, + "learning_rate": 4.121805423200208e-05, + "loss": 1.3217, + "step": 101870 + }, + { + "epoch": 8.82, + "learning_rate": 4.121718790609027e-05, + "loss": 1.3019, + "step": 101880 + }, + { + "epoch": 8.82, + "learning_rate": 4.121632158017846e-05, + "loss": 1.3541, + "step": 101890 + }, + { + "epoch": 8.82, + "learning_rate": 4.1215455254266655e-05, + "loss": 1.3242, + "step": 101900 + }, + { + "epoch": 8.82, + "learning_rate": 4.121458892835485e-05, + "loss": 1.3096, + "step": 101910 + }, + { + "epoch": 8.82, + "learning_rate": 4.121372260244304e-05, + "loss": 1.3327, + "step": 101920 + }, + { + "epoch": 8.82, + "learning_rate": 4.121285627653123e-05, + "loss": 1.2536, + "step": 101930 + }, + { + "epoch": 8.82, + "learning_rate": 4.121198995061942e-05, + "loss": 1.3743, + "step": 101940 + }, + { + "epoch": 8.82, + "learning_rate": 4.1211123624707614e-05, + "loss": 1.4038, + "step": 101950 + }, + { + "epoch": 8.83, + "learning_rate": 4.121025729879581e-05, + "loss": 1.3194, + "step": 101960 + }, + { + "epoch": 8.83, + "learning_rate": 4.1209390972883996e-05, + "loss": 1.3079, + "step": 101970 + }, + { + "epoch": 8.83, + "learning_rate": 4.120852464697219e-05, + "loss": 1.32, + "step": 101980 + }, + { + "epoch": 8.83, + "learning_rate": 4.1207658321060385e-05, + "loss": 1.3342, + "step": 101990 + }, + { + "epoch": 8.83, + "learning_rate": 4.120679199514857e-05, + "loss": 1.3963, + "step": 102000 + }, + { + "epoch": 8.83, + "learning_rate": 4.120592566923677e-05, + "loss": 1.2466, + "step": 102010 + }, + { + "epoch": 8.83, + "learning_rate": 4.120505934332496e-05, + "loss": 1.297, + "step": 102020 + }, + { + "epoch": 8.83, + "learning_rate": 4.120419301741315e-05, + "loss": 1.2549, + "step": 102030 + }, + { + "epoch": 8.83, + "learning_rate": 4.1203326691501344e-05, + "loss": 1.3173, + "step": 102040 + }, + { + "epoch": 8.83, + "learning_rate": 4.120246036558953e-05, + "loss": 1.3836, + "step": 102050 + }, + { + "epoch": 8.83, + "learning_rate": 4.1201594039677726e-05, + "loss": 1.3404, + "step": 102060 + }, + { + "epoch": 8.83, + "learning_rate": 4.120072771376592e-05, + "loss": 1.3496, + "step": 102070 + }, + { + "epoch": 8.84, + "learning_rate": 4.119986138785411e-05, + "loss": 1.2624, + "step": 102080 + }, + { + "epoch": 8.84, + "learning_rate": 4.11989950619423e-05, + "loss": 1.3951, + "step": 102090 + }, + { + "epoch": 8.84, + "learning_rate": 4.11981287360305e-05, + "loss": 1.3185, + "step": 102100 + }, + { + "epoch": 8.84, + "learning_rate": 4.1197262410118685e-05, + "loss": 1.3759, + "step": 102110 + }, + { + "epoch": 8.84, + "learning_rate": 4.119639608420688e-05, + "loss": 1.2882, + "step": 102120 + }, + { + "epoch": 8.84, + "learning_rate": 4.1195529758295074e-05, + "loss": 1.3682, + "step": 102130 + }, + { + "epoch": 8.84, + "learning_rate": 4.119466343238326e-05, + "loss": 1.3045, + "step": 102140 + }, + { + "epoch": 8.84, + "learning_rate": 4.1193797106471456e-05, + "loss": 1.2925, + "step": 102150 + }, + { + "epoch": 8.84, + "learning_rate": 4.119293078055964e-05, + "loss": 1.2865, + "step": 102160 + }, + { + "epoch": 8.84, + "learning_rate": 4.119206445464784e-05, + "loss": 1.2906, + "step": 102170 + }, + { + "epoch": 8.84, + "learning_rate": 4.119119812873603e-05, + "loss": 1.3436, + "step": 102180 + }, + { + "epoch": 8.85, + "learning_rate": 4.119033180282422e-05, + "loss": 1.3002, + "step": 102190 + }, + { + "epoch": 8.85, + "learning_rate": 4.1189465476912414e-05, + "loss": 1.3108, + "step": 102200 + }, + { + "epoch": 8.85, + "learning_rate": 4.118859915100061e-05, + "loss": 1.2724, + "step": 102210 + }, + { + "epoch": 8.85, + "learning_rate": 4.1187732825088797e-05, + "loss": 1.3307, + "step": 102220 + }, + { + "epoch": 8.85, + "learning_rate": 4.118686649917699e-05, + "loss": 1.3568, + "step": 102230 + }, + { + "epoch": 8.85, + "learning_rate": 4.1186000173265185e-05, + "loss": 1.2937, + "step": 102240 + }, + { + "epoch": 8.85, + "learning_rate": 4.118513384735337e-05, + "loss": 1.3492, + "step": 102250 + }, + { + "epoch": 8.85, + "learning_rate": 4.118426752144157e-05, + "loss": 1.2821, + "step": 102260 + }, + { + "epoch": 8.85, + "learning_rate": 4.1183401195529755e-05, + "loss": 1.3368, + "step": 102270 + }, + { + "epoch": 8.85, + "learning_rate": 4.118253486961795e-05, + "loss": 1.3163, + "step": 102280 + }, + { + "epoch": 8.85, + "learning_rate": 4.1181668543706144e-05, + "loss": 1.2548, + "step": 102290 + }, + { + "epoch": 8.85, + "learning_rate": 4.118080221779433e-05, + "loss": 1.3896, + "step": 102300 + }, + { + "epoch": 8.86, + "learning_rate": 4.1179935891882526e-05, + "loss": 1.2681, + "step": 102310 + }, + { + "epoch": 8.86, + "learning_rate": 4.117906956597072e-05, + "loss": 1.384, + "step": 102320 + }, + { + "epoch": 8.86, + "learning_rate": 4.117820324005891e-05, + "loss": 1.3263, + "step": 102330 + }, + { + "epoch": 8.86, + "learning_rate": 4.11773369141471e-05, + "loss": 1.3651, + "step": 102340 + }, + { + "epoch": 8.86, + "learning_rate": 4.11764705882353e-05, + "loss": 1.324, + "step": 102350 + }, + { + "epoch": 8.86, + "learning_rate": 4.1175604262323485e-05, + "loss": 1.3884, + "step": 102360 + }, + { + "epoch": 8.86, + "learning_rate": 4.117473793641168e-05, + "loss": 1.2733, + "step": 102370 + }, + { + "epoch": 8.86, + "learning_rate": 4.117387161049987e-05, + "loss": 1.3988, + "step": 102380 + }, + { + "epoch": 8.86, + "learning_rate": 4.117300528458806e-05, + "loss": 1.3811, + "step": 102390 + }, + { + "epoch": 8.86, + "learning_rate": 4.1172138958676256e-05, + "loss": 1.3493, + "step": 102400 + }, + { + "epoch": 8.86, + "learning_rate": 4.1171272632764444e-05, + "loss": 1.3992, + "step": 102410 + }, + { + "epoch": 8.87, + "learning_rate": 4.117040630685264e-05, + "loss": 1.3525, + "step": 102420 + }, + { + "epoch": 8.87, + "learning_rate": 4.116953998094083e-05, + "loss": 1.2895, + "step": 102430 + }, + { + "epoch": 8.87, + "learning_rate": 4.116867365502902e-05, + "loss": 1.3133, + "step": 102440 + }, + { + "epoch": 8.87, + "learning_rate": 4.1167807329117215e-05, + "loss": 1.344, + "step": 102450 + }, + { + "epoch": 8.87, + "learning_rate": 4.116694100320541e-05, + "loss": 1.2649, + "step": 102460 + }, + { + "epoch": 8.87, + "learning_rate": 4.11660746772936e-05, + "loss": 1.3685, + "step": 102470 + }, + { + "epoch": 8.87, + "learning_rate": 4.116520835138179e-05, + "loss": 1.3483, + "step": 102480 + }, + { + "epoch": 8.87, + "learning_rate": 4.116434202546998e-05, + "loss": 1.2881, + "step": 102490 + }, + { + "epoch": 8.87, + "learning_rate": 4.1163475699558173e-05, + "loss": 1.3538, + "step": 102500 + }, + { + "epoch": 8.87, + "learning_rate": 4.116260937364637e-05, + "loss": 1.3781, + "step": 102510 + }, + { + "epoch": 8.87, + "learning_rate": 4.1161743047734556e-05, + "loss": 1.316, + "step": 102520 + }, + { + "epoch": 8.87, + "learning_rate": 4.116087672182275e-05, + "loss": 1.3764, + "step": 102530 + }, + { + "epoch": 8.88, + "learning_rate": 4.1160010395910944e-05, + "loss": 1.3158, + "step": 102540 + }, + { + "epoch": 8.88, + "learning_rate": 4.115914406999913e-05, + "loss": 1.3139, + "step": 102550 + }, + { + "epoch": 8.88, + "learning_rate": 4.1158277744087327e-05, + "loss": 1.3218, + "step": 102560 + }, + { + "epoch": 8.88, + "learning_rate": 4.115741141817552e-05, + "loss": 1.3207, + "step": 102570 + }, + { + "epoch": 8.88, + "learning_rate": 4.115654509226371e-05, + "loss": 1.2993, + "step": 102580 + }, + { + "epoch": 8.88, + "learning_rate": 4.11556787663519e-05, + "loss": 1.2987, + "step": 102590 + }, + { + "epoch": 8.88, + "learning_rate": 4.115481244044009e-05, + "loss": 1.3283, + "step": 102600 + }, + { + "epoch": 8.88, + "learning_rate": 4.1153946114528285e-05, + "loss": 1.3435, + "step": 102610 + }, + { + "epoch": 8.88, + "learning_rate": 4.115307978861648e-05, + "loss": 1.3138, + "step": 102620 + }, + { + "epoch": 8.88, + "learning_rate": 4.115221346270467e-05, + "loss": 1.3191, + "step": 102630 + }, + { + "epoch": 8.88, + "learning_rate": 4.115134713679286e-05, + "loss": 1.2806, + "step": 102640 + }, + { + "epoch": 8.89, + "learning_rate": 4.1150480810881056e-05, + "loss": 1.3136, + "step": 102650 + }, + { + "epoch": 8.89, + "learning_rate": 4.1149614484969244e-05, + "loss": 1.3335, + "step": 102660 + }, + { + "epoch": 8.89, + "learning_rate": 4.114874815905744e-05, + "loss": 1.3617, + "step": 102670 + }, + { + "epoch": 8.89, + "learning_rate": 4.1147881833145626e-05, + "loss": 1.3253, + "step": 102680 + }, + { + "epoch": 8.89, + "learning_rate": 4.114701550723382e-05, + "loss": 1.242, + "step": 102690 + }, + { + "epoch": 8.89, + "learning_rate": 4.1146149181322015e-05, + "loss": 1.4047, + "step": 102700 + }, + { + "epoch": 8.89, + "learning_rate": 4.11452828554102e-05, + "loss": 1.2937, + "step": 102710 + }, + { + "epoch": 8.89, + "learning_rate": 4.11444165294984e-05, + "loss": 1.4228, + "step": 102720 + }, + { + "epoch": 8.89, + "learning_rate": 4.114355020358659e-05, + "loss": 1.3466, + "step": 102730 + }, + { + "epoch": 8.89, + "learning_rate": 4.114268387767478e-05, + "loss": 1.3558, + "step": 102740 + }, + { + "epoch": 8.89, + "learning_rate": 4.1141817551762974e-05, + "loss": 1.3077, + "step": 102750 + }, + { + "epoch": 8.89, + "learning_rate": 4.114095122585117e-05, + "loss": 1.3638, + "step": 102760 + }, + { + "epoch": 8.9, + "learning_rate": 4.1140084899939356e-05, + "loss": 1.2837, + "step": 102770 + }, + { + "epoch": 8.9, + "learning_rate": 4.113921857402755e-05, + "loss": 1.3438, + "step": 102780 + }, + { + "epoch": 8.9, + "learning_rate": 4.113835224811574e-05, + "loss": 1.3162, + "step": 102790 + }, + { + "epoch": 8.9, + "learning_rate": 4.113748592220393e-05, + "loss": 1.341, + "step": 102800 + }, + { + "epoch": 8.9, + "learning_rate": 4.113661959629213e-05, + "loss": 1.3427, + "step": 102810 + }, + { + "epoch": 8.9, + "learning_rate": 4.1135753270380315e-05, + "loss": 1.3681, + "step": 102820 + }, + { + "epoch": 8.9, + "learning_rate": 4.113488694446851e-05, + "loss": 1.3378, + "step": 102830 + }, + { + "epoch": 8.9, + "learning_rate": 4.1134020618556704e-05, + "loss": 1.2819, + "step": 102840 + }, + { + "epoch": 8.9, + "learning_rate": 4.113315429264489e-05, + "loss": 1.2949, + "step": 102850 + }, + { + "epoch": 8.9, + "learning_rate": 4.1132287966733086e-05, + "loss": 1.3139, + "step": 102860 + }, + { + "epoch": 8.9, + "learning_rate": 4.113142164082128e-05, + "loss": 1.191, + "step": 102870 + }, + { + "epoch": 8.9, + "learning_rate": 4.113055531490947e-05, + "loss": 1.3164, + "step": 102880 + }, + { + "epoch": 8.91, + "learning_rate": 4.112968898899766e-05, + "loss": 1.2648, + "step": 102890 + }, + { + "epoch": 8.91, + "learning_rate": 4.112882266308585e-05, + "loss": 1.3581, + "step": 102900 + }, + { + "epoch": 8.91, + "learning_rate": 4.1127956337174044e-05, + "loss": 1.3787, + "step": 102910 + }, + { + "epoch": 8.91, + "learning_rate": 4.112709001126224e-05, + "loss": 1.34, + "step": 102920 + }, + { + "epoch": 8.91, + "learning_rate": 4.1126223685350426e-05, + "loss": 1.3266, + "step": 102930 + }, + { + "epoch": 8.91, + "learning_rate": 4.112535735943862e-05, + "loss": 1.3379, + "step": 102940 + }, + { + "epoch": 8.91, + "learning_rate": 4.1124491033526815e-05, + "loss": 1.3214, + "step": 102950 + }, + { + "epoch": 8.91, + "learning_rate": 4.1123624707615e-05, + "loss": 1.2896, + "step": 102960 + }, + { + "epoch": 8.91, + "learning_rate": 4.11227583817032e-05, + "loss": 1.3537, + "step": 102970 + }, + { + "epoch": 8.91, + "learning_rate": 4.112189205579139e-05, + "loss": 1.3305, + "step": 102980 + }, + { + "epoch": 8.91, + "learning_rate": 4.112102572987958e-05, + "loss": 1.301, + "step": 102990 + }, + { + "epoch": 8.92, + "learning_rate": 4.1120159403967774e-05, + "loss": 1.3386, + "step": 103000 + }, + { + "epoch": 8.92, + "learning_rate": 4.111929307805596e-05, + "loss": 1.3454, + "step": 103010 + }, + { + "epoch": 8.92, + "learning_rate": 4.1118426752144156e-05, + "loss": 1.3685, + "step": 103020 + }, + { + "epoch": 8.92, + "learning_rate": 4.111756042623235e-05, + "loss": 1.3295, + "step": 103030 + }, + { + "epoch": 8.92, + "learning_rate": 4.111669410032054e-05, + "loss": 1.3345, + "step": 103040 + }, + { + "epoch": 8.92, + "learning_rate": 4.111582777440873e-05, + "loss": 1.3077, + "step": 103050 + }, + { + "epoch": 8.92, + "learning_rate": 4.111496144849693e-05, + "loss": 1.3028, + "step": 103060 + }, + { + "epoch": 8.92, + "learning_rate": 4.1114095122585115e-05, + "loss": 1.3198, + "step": 103070 + }, + { + "epoch": 8.92, + "learning_rate": 4.111322879667331e-05, + "loss": 1.3447, + "step": 103080 + }, + { + "epoch": 8.92, + "learning_rate": 4.1112362470761504e-05, + "loss": 1.2855, + "step": 103090 + }, + { + "epoch": 8.92, + "learning_rate": 4.111149614484969e-05, + "loss": 1.3121, + "step": 103100 + }, + { + "epoch": 8.92, + "learning_rate": 4.1110629818937886e-05, + "loss": 1.2921, + "step": 103110 + }, + { + "epoch": 8.93, + "learning_rate": 4.1109763493026074e-05, + "loss": 1.3669, + "step": 103120 + }, + { + "epoch": 8.93, + "learning_rate": 4.110889716711427e-05, + "loss": 1.3473, + "step": 103130 + }, + { + "epoch": 8.93, + "learning_rate": 4.110803084120246e-05, + "loss": 1.3499, + "step": 103140 + }, + { + "epoch": 8.93, + "learning_rate": 4.110716451529065e-05, + "loss": 1.3331, + "step": 103150 + }, + { + "epoch": 8.93, + "learning_rate": 4.1106298189378845e-05, + "loss": 1.3072, + "step": 103160 + }, + { + "epoch": 8.93, + "learning_rate": 4.110543186346704e-05, + "loss": 1.3809, + "step": 103170 + }, + { + "epoch": 8.93, + "learning_rate": 4.110456553755523e-05, + "loss": 1.2891, + "step": 103180 + }, + { + "epoch": 8.93, + "learning_rate": 4.110369921164342e-05, + "loss": 1.2317, + "step": 103190 + }, + { + "epoch": 8.93, + "learning_rate": 4.1102832885731616e-05, + "loss": 1.3019, + "step": 103200 + }, + { + "epoch": 8.93, + "learning_rate": 4.1101966559819803e-05, + "loss": 1.296, + "step": 103210 + }, + { + "epoch": 8.93, + "learning_rate": 4.1101100233908e-05, + "loss": 1.2394, + "step": 103220 + }, + { + "epoch": 8.94, + "learning_rate": 4.1100233907996186e-05, + "loss": 1.3646, + "step": 103230 + }, + { + "epoch": 8.94, + "learning_rate": 4.109936758208438e-05, + "loss": 1.3043, + "step": 103240 + }, + { + "epoch": 8.94, + "learning_rate": 4.1098501256172574e-05, + "loss": 1.3085, + "step": 103250 + }, + { + "epoch": 8.94, + "learning_rate": 4.109763493026076e-05, + "loss": 1.3153, + "step": 103260 + }, + { + "epoch": 8.94, + "learning_rate": 4.1096768604348957e-05, + "loss": 1.3047, + "step": 103270 + }, + { + "epoch": 8.94, + "learning_rate": 4.109590227843715e-05, + "loss": 1.339, + "step": 103280 + }, + { + "epoch": 8.94, + "learning_rate": 4.109503595252534e-05, + "loss": 1.2213, + "step": 103290 + }, + { + "epoch": 8.94, + "learning_rate": 4.109416962661353e-05, + "loss": 1.285, + "step": 103300 + }, + { + "epoch": 8.94, + "learning_rate": 4.109330330070173e-05, + "loss": 1.3369, + "step": 103310 + }, + { + "epoch": 8.94, + "learning_rate": 4.1092436974789915e-05, + "loss": 1.2985, + "step": 103320 + }, + { + "epoch": 8.94, + "learning_rate": 4.109157064887811e-05, + "loss": 1.4076, + "step": 103330 + }, + { + "epoch": 8.94, + "learning_rate": 4.10907043229663e-05, + "loss": 1.2966, + "step": 103340 + }, + { + "epoch": 8.95, + "learning_rate": 4.108983799705449e-05, + "loss": 1.2452, + "step": 103350 + }, + { + "epoch": 8.95, + "learning_rate": 4.1088971671142686e-05, + "loss": 1.3051, + "step": 103360 + }, + { + "epoch": 8.95, + "learning_rate": 4.1088105345230874e-05, + "loss": 1.3349, + "step": 103370 + }, + { + "epoch": 8.95, + "learning_rate": 4.108723901931907e-05, + "loss": 1.2613, + "step": 103380 + }, + { + "epoch": 8.95, + "learning_rate": 4.108637269340726e-05, + "loss": 1.3763, + "step": 103390 + }, + { + "epoch": 8.95, + "learning_rate": 4.108550636749545e-05, + "loss": 1.33, + "step": 103400 + }, + { + "epoch": 8.95, + "learning_rate": 4.1084640041583645e-05, + "loss": 1.3383, + "step": 103410 + }, + { + "epoch": 8.95, + "learning_rate": 4.108377371567184e-05, + "loss": 1.2932, + "step": 103420 + }, + { + "epoch": 8.95, + "learning_rate": 4.108290738976003e-05, + "loss": 1.348, + "step": 103430 + }, + { + "epoch": 8.95, + "learning_rate": 4.108204106384822e-05, + "loss": 1.2266, + "step": 103440 + }, + { + "epoch": 8.95, + "learning_rate": 4.108117473793641e-05, + "loss": 1.3776, + "step": 103450 + }, + { + "epoch": 8.96, + "learning_rate": 4.1080308412024604e-05, + "loss": 1.314, + "step": 103460 + }, + { + "epoch": 8.96, + "learning_rate": 4.10794420861128e-05, + "loss": 1.3022, + "step": 103470 + }, + { + "epoch": 8.96, + "learning_rate": 4.1078575760200986e-05, + "loss": 1.2953, + "step": 103480 + }, + { + "epoch": 8.96, + "learning_rate": 4.107770943428918e-05, + "loss": 1.335, + "step": 103490 + }, + { + "epoch": 8.96, + "learning_rate": 4.1076843108377375e-05, + "loss": 1.2657, + "step": 103500 + }, + { + "epoch": 8.96, + "learning_rate": 4.107597678246556e-05, + "loss": 1.337, + "step": 103510 + }, + { + "epoch": 8.96, + "learning_rate": 4.107511045655376e-05, + "loss": 1.3181, + "step": 103520 + }, + { + "epoch": 8.96, + "learning_rate": 4.1074244130641945e-05, + "loss": 1.3344, + "step": 103530 + }, + { + "epoch": 8.96, + "learning_rate": 4.107337780473014e-05, + "loss": 1.3594, + "step": 103540 + }, + { + "epoch": 8.96, + "learning_rate": 4.1072511478818333e-05, + "loss": 1.2668, + "step": 103550 + }, + { + "epoch": 8.96, + "learning_rate": 4.107164515290652e-05, + "loss": 1.2748, + "step": 103560 + }, + { + "epoch": 8.96, + "learning_rate": 4.1070778826994716e-05, + "loss": 1.3556, + "step": 103570 + }, + { + "epoch": 8.97, + "learning_rate": 4.106991250108291e-05, + "loss": 1.3152, + "step": 103580 + }, + { + "epoch": 8.97, + "learning_rate": 4.10690461751711e-05, + "loss": 1.3201, + "step": 103590 + }, + { + "epoch": 8.97, + "learning_rate": 4.106817984925929e-05, + "loss": 1.3295, + "step": 103600 + }, + { + "epoch": 8.97, + "learning_rate": 4.106731352334749e-05, + "loss": 1.3514, + "step": 103610 + }, + { + "epoch": 8.97, + "learning_rate": 4.1066447197435674e-05, + "loss": 1.3119, + "step": 103620 + }, + { + "epoch": 8.97, + "learning_rate": 4.106558087152387e-05, + "loss": 1.2956, + "step": 103630 + }, + { + "epoch": 8.97, + "learning_rate": 4.1064714545612056e-05, + "loss": 1.3567, + "step": 103640 + }, + { + "epoch": 8.97, + "learning_rate": 4.106384821970025e-05, + "loss": 1.3113, + "step": 103650 + }, + { + "epoch": 8.97, + "learning_rate": 4.1062981893788445e-05, + "loss": 1.2878, + "step": 103660 + }, + { + "epoch": 8.97, + "learning_rate": 4.106211556787663e-05, + "loss": 1.2988, + "step": 103670 + }, + { + "epoch": 8.97, + "learning_rate": 4.106124924196483e-05, + "loss": 1.3629, + "step": 103680 + }, + { + "epoch": 8.98, + "learning_rate": 4.106038291605302e-05, + "loss": 1.3894, + "step": 103690 + }, + { + "epoch": 8.98, + "learning_rate": 4.105951659014121e-05, + "loss": 1.3681, + "step": 103700 + }, + { + "epoch": 8.98, + "learning_rate": 4.1058650264229404e-05, + "loss": 1.2978, + "step": 103710 + }, + { + "epoch": 8.98, + "learning_rate": 4.10577839383176e-05, + "loss": 1.3047, + "step": 103720 + }, + { + "epoch": 8.98, + "learning_rate": 4.1056917612405786e-05, + "loss": 1.3672, + "step": 103730 + }, + { + "epoch": 8.98, + "learning_rate": 4.105605128649398e-05, + "loss": 1.3076, + "step": 103740 + }, + { + "epoch": 8.98, + "learning_rate": 4.105518496058217e-05, + "loss": 1.2568, + "step": 103750 + }, + { + "epoch": 8.98, + "learning_rate": 4.105431863467036e-05, + "loss": 1.2845, + "step": 103760 + }, + { + "epoch": 8.98, + "learning_rate": 4.105345230875856e-05, + "loss": 1.3459, + "step": 103770 + }, + { + "epoch": 8.98, + "learning_rate": 4.1052585982846745e-05, + "loss": 1.3397, + "step": 103780 + }, + { + "epoch": 8.98, + "learning_rate": 4.105171965693494e-05, + "loss": 1.4312, + "step": 103790 + }, + { + "epoch": 8.98, + "learning_rate": 4.1050853331023134e-05, + "loss": 1.316, + "step": 103800 + }, + { + "epoch": 8.99, + "learning_rate": 4.104998700511132e-05, + "loss": 1.3433, + "step": 103810 + }, + { + "epoch": 8.99, + "learning_rate": 4.1049120679199516e-05, + "loss": 1.322, + "step": 103820 + }, + { + "epoch": 8.99, + "learning_rate": 4.104825435328771e-05, + "loss": 1.3495, + "step": 103830 + }, + { + "epoch": 8.99, + "learning_rate": 4.10473880273759e-05, + "loss": 1.3259, + "step": 103840 + }, + { + "epoch": 8.99, + "learning_rate": 4.104652170146409e-05, + "loss": 1.286, + "step": 103850 + }, + { + "epoch": 8.99, + "learning_rate": 4.104565537555228e-05, + "loss": 1.2758, + "step": 103860 + }, + { + "epoch": 8.99, + "learning_rate": 4.1044789049640475e-05, + "loss": 1.264, + "step": 103870 + }, + { + "epoch": 8.99, + "learning_rate": 4.104392272372867e-05, + "loss": 1.2563, + "step": 103880 + }, + { + "epoch": 8.99, + "learning_rate": 4.104305639781686e-05, + "loss": 1.2732, + "step": 103890 + }, + { + "epoch": 8.99, + "learning_rate": 4.104219007190505e-05, + "loss": 1.2913, + "step": 103900 + }, + { + "epoch": 8.99, + "learning_rate": 4.1041323745993246e-05, + "loss": 1.3459, + "step": 103910 + }, + { + "epoch": 8.99, + "learning_rate": 4.104045742008143e-05, + "loss": 1.301, + "step": 103920 + }, + { + "epoch": 9.0, + "learning_rate": 4.103959109416963e-05, + "loss": 1.2813, + "step": 103930 + }, + { + "epoch": 9.0, + "learning_rate": 4.103872476825782e-05, + "loss": 1.3189, + "step": 103940 + }, + { + "epoch": 9.0, + "learning_rate": 4.103785844234601e-05, + "loss": 1.3077, + "step": 103950 + }, + { + "epoch": 9.0, + "learning_rate": 4.1036992116434204e-05, + "loss": 1.2721, + "step": 103960 + }, + { + "epoch": 9.0, + "learning_rate": 4.103612579052239e-05, + "loss": 1.335, + "step": 103970 + }, + { + "epoch": 9.0, + "eval_Bleu_1": 0.05036714713081225, + "eval_Bleu_2": 3.1732320535214644e-11, + "eval_Bleu_3": 2.8192892779935983e-14, + "eval_Bleu_4": 8.659487693475707e-16, + "eval_ROUGE_L": 0.11112687922846864, + "eval_cer": 0.9941402878475721, + "eval_em": 0, + "eval_f1": 0.07772100120123805, + "eval_loss": 1.2269008159637451, + "eval_runtime": 1718.1378, + "eval_samples_per_second": 2.989, + "eval_steps_per_second": 2.989, + "eval_wer": 0.9781337799585159, + "step": 103978 + }, + { + "epoch": 9.0, + "learning_rate": 4.1035259464610587e-05, + "loss": 1.2931, + "step": 103980 + }, + { + "epoch": 9.0, + "learning_rate": 4.103439313869878e-05, + "loss": 1.2922, + "step": 103990 + }, + { + "epoch": 9.0, + "learning_rate": 4.103352681278697e-05, + "loss": 1.3375, + "step": 104000 + }, + { + "epoch": 9.0, + "learning_rate": 4.103266048687516e-05, + "loss": 1.3815, + "step": 104010 + }, + { + "epoch": 9.0, + "learning_rate": 4.103179416096336e-05, + "loss": 1.2933, + "step": 104020 + }, + { + "epoch": 9.0, + "learning_rate": 4.1030927835051545e-05, + "loss": 1.3485, + "step": 104030 + }, + { + "epoch": 9.01, + "learning_rate": 4.103006150913974e-05, + "loss": 1.3028, + "step": 104040 + }, + { + "epoch": 9.01, + "learning_rate": 4.1029195183227934e-05, + "loss": 1.3199, + "step": 104050 + }, + { + "epoch": 9.01, + "learning_rate": 4.102832885731612e-05, + "loss": 1.2868, + "step": 104060 + }, + { + "epoch": 9.01, + "learning_rate": 4.1027462531404316e-05, + "loss": 1.3025, + "step": 104070 + }, + { + "epoch": 9.01, + "learning_rate": 4.1026596205492504e-05, + "loss": 1.3227, + "step": 104080 + }, + { + "epoch": 9.01, + "learning_rate": 4.10257298795807e-05, + "loss": 1.3072, + "step": 104090 + }, + { + "epoch": 9.01, + "learning_rate": 4.102486355366889e-05, + "loss": 1.2848, + "step": 104100 + }, + { + "epoch": 9.01, + "learning_rate": 4.102399722775708e-05, + "loss": 1.3098, + "step": 104110 + }, + { + "epoch": 9.01, + "learning_rate": 4.1023130901845275e-05, + "loss": 1.3124, + "step": 104120 + }, + { + "epoch": 9.01, + "learning_rate": 4.102226457593347e-05, + "loss": 1.277, + "step": 104130 + }, + { + "epoch": 9.01, + "learning_rate": 4.102139825002166e-05, + "loss": 1.3116, + "step": 104140 + }, + { + "epoch": 9.01, + "learning_rate": 4.102053192410985e-05, + "loss": 1.3259, + "step": 104150 + }, + { + "epoch": 9.02, + "learning_rate": 4.1019665598198046e-05, + "loss": 1.3454, + "step": 104160 + }, + { + "epoch": 9.02, + "learning_rate": 4.1018799272286234e-05, + "loss": 1.2171, + "step": 104170 + }, + { + "epoch": 9.02, + "learning_rate": 4.101793294637443e-05, + "loss": 1.2986, + "step": 104180 + }, + { + "epoch": 9.02, + "learning_rate": 4.1017066620462616e-05, + "loss": 1.2536, + "step": 104190 + }, + { + "epoch": 9.02, + "learning_rate": 4.101620029455081e-05, + "loss": 1.291, + "step": 104200 + }, + { + "epoch": 9.02, + "learning_rate": 4.1015333968639005e-05, + "loss": 1.3157, + "step": 104210 + }, + { + "epoch": 9.02, + "learning_rate": 4.101446764272719e-05, + "loss": 1.3494, + "step": 104220 + }, + { + "epoch": 9.02, + "learning_rate": 4.101360131681539e-05, + "loss": 1.2788, + "step": 104230 + }, + { + "epoch": 9.02, + "learning_rate": 4.101273499090358e-05, + "loss": 1.2214, + "step": 104240 + }, + { + "epoch": 9.02, + "learning_rate": 4.101186866499177e-05, + "loss": 1.2872, + "step": 104250 + }, + { + "epoch": 9.02, + "learning_rate": 4.1011002339079963e-05, + "loss": 1.3001, + "step": 104260 + }, + { + "epoch": 9.03, + "learning_rate": 4.101013601316815e-05, + "loss": 1.2283, + "step": 104270 + }, + { + "epoch": 9.03, + "learning_rate": 4.1009269687256346e-05, + "loss": 1.2173, + "step": 104280 + }, + { + "epoch": 9.03, + "learning_rate": 4.100840336134454e-05, + "loss": 1.3233, + "step": 104290 + }, + { + "epoch": 9.03, + "learning_rate": 4.100753703543273e-05, + "loss": 1.2658, + "step": 104300 + }, + { + "epoch": 9.03, + "learning_rate": 4.100667070952092e-05, + "loss": 1.2881, + "step": 104310 + }, + { + "epoch": 9.03, + "learning_rate": 4.100580438360912e-05, + "loss": 1.3435, + "step": 104320 + }, + { + "epoch": 9.03, + "learning_rate": 4.1004938057697304e-05, + "loss": 1.2429, + "step": 104330 + }, + { + "epoch": 9.03, + "learning_rate": 4.10040717317855e-05, + "loss": 1.3096, + "step": 104340 + }, + { + "epoch": 9.03, + "learning_rate": 4.100320540587369e-05, + "loss": 1.3466, + "step": 104350 + }, + { + "epoch": 9.03, + "learning_rate": 4.100233907996188e-05, + "loss": 1.3258, + "step": 104360 + }, + { + "epoch": 9.03, + "learning_rate": 4.1001472754050075e-05, + "loss": 1.3053, + "step": 104370 + }, + { + "epoch": 9.03, + "learning_rate": 4.100060642813826e-05, + "loss": 1.3467, + "step": 104380 + }, + { + "epoch": 9.04, + "learning_rate": 4.099974010222646e-05, + "loss": 1.2596, + "step": 104390 + }, + { + "epoch": 9.04, + "learning_rate": 4.099887377631465e-05, + "loss": 1.2997, + "step": 104400 + }, + { + "epoch": 9.04, + "learning_rate": 4.099800745040284e-05, + "loss": 1.3038, + "step": 104410 + }, + { + "epoch": 9.04, + "learning_rate": 4.0997141124491034e-05, + "loss": 1.2398, + "step": 104420 + }, + { + "epoch": 9.04, + "learning_rate": 4.099627479857923e-05, + "loss": 1.2636, + "step": 104430 + }, + { + "epoch": 9.04, + "learning_rate": 4.0995408472667416e-05, + "loss": 1.3241, + "step": 104440 + }, + { + "epoch": 9.04, + "learning_rate": 4.099454214675561e-05, + "loss": 1.2989, + "step": 104450 + }, + { + "epoch": 9.04, + "learning_rate": 4.0993675820843805e-05, + "loss": 1.3255, + "step": 104460 + }, + { + "epoch": 9.04, + "learning_rate": 4.099280949493199e-05, + "loss": 1.2144, + "step": 104470 + }, + { + "epoch": 9.04, + "learning_rate": 4.099194316902019e-05, + "loss": 1.3487, + "step": 104480 + }, + { + "epoch": 9.04, + "learning_rate": 4.0991076843108375e-05, + "loss": 1.2474, + "step": 104490 + }, + { + "epoch": 9.05, + "learning_rate": 4.099021051719657e-05, + "loss": 1.3248, + "step": 104500 + }, + { + "epoch": 9.05, + "learning_rate": 4.0989344191284764e-05, + "loss": 1.2935, + "step": 104510 + }, + { + "epoch": 9.05, + "learning_rate": 4.098847786537295e-05, + "loss": 1.2772, + "step": 104520 + }, + { + "epoch": 9.05, + "learning_rate": 4.0987611539461146e-05, + "loss": 1.2825, + "step": 104530 + }, + { + "epoch": 9.05, + "learning_rate": 4.098674521354934e-05, + "loss": 1.3248, + "step": 104540 + }, + { + "epoch": 9.05, + "learning_rate": 4.098587888763753e-05, + "loss": 1.2863, + "step": 104550 + }, + { + "epoch": 9.05, + "learning_rate": 4.098501256172572e-05, + "loss": 1.2104, + "step": 104560 + }, + { + "epoch": 9.05, + "learning_rate": 4.098414623581392e-05, + "loss": 1.3267, + "step": 104570 + }, + { + "epoch": 9.05, + "learning_rate": 4.0983279909902105e-05, + "loss": 1.307, + "step": 104580 + }, + { + "epoch": 9.05, + "learning_rate": 4.09824135839903e-05, + "loss": 1.3173, + "step": 104590 + }, + { + "epoch": 9.05, + "learning_rate": 4.098154725807849e-05, + "loss": 1.2967, + "step": 104600 + }, + { + "epoch": 9.05, + "learning_rate": 4.098068093216668e-05, + "loss": 1.2611, + "step": 104610 + }, + { + "epoch": 9.06, + "learning_rate": 4.0979814606254876e-05, + "loss": 1.3161, + "step": 104620 + }, + { + "epoch": 9.06, + "learning_rate": 4.097894828034306e-05, + "loss": 1.2832, + "step": 104630 + }, + { + "epoch": 9.06, + "learning_rate": 4.097808195443126e-05, + "loss": 1.3057, + "step": 104640 + }, + { + "epoch": 9.06, + "learning_rate": 4.097721562851945e-05, + "loss": 1.2831, + "step": 104650 + }, + { + "epoch": 9.06, + "learning_rate": 4.097634930260764e-05, + "loss": 1.2495, + "step": 104660 + }, + { + "epoch": 9.06, + "learning_rate": 4.0975482976695834e-05, + "loss": 1.2789, + "step": 104670 + }, + { + "epoch": 9.06, + "learning_rate": 4.097461665078403e-05, + "loss": 1.3502, + "step": 104680 + }, + { + "epoch": 9.06, + "learning_rate": 4.0973750324872217e-05, + "loss": 1.3244, + "step": 104690 + }, + { + "epoch": 9.06, + "learning_rate": 4.097288399896041e-05, + "loss": 1.2826, + "step": 104700 + }, + { + "epoch": 9.06, + "learning_rate": 4.09720176730486e-05, + "loss": 1.3628, + "step": 104710 + }, + { + "epoch": 9.06, + "learning_rate": 4.097115134713679e-05, + "loss": 1.2577, + "step": 104720 + }, + { + "epoch": 9.07, + "learning_rate": 4.097028502122499e-05, + "loss": 1.2664, + "step": 104730 + }, + { + "epoch": 9.07, + "learning_rate": 4.0969418695313175e-05, + "loss": 1.3478, + "step": 104740 + }, + { + "epoch": 9.07, + "learning_rate": 4.096855236940137e-05, + "loss": 1.3221, + "step": 104750 + }, + { + "epoch": 9.07, + "learning_rate": 4.0967686043489564e-05, + "loss": 1.3648, + "step": 104760 + }, + { + "epoch": 9.07, + "learning_rate": 4.096681971757775e-05, + "loss": 1.3092, + "step": 104770 + }, + { + "epoch": 9.07, + "learning_rate": 4.0965953391665946e-05, + "loss": 1.3442, + "step": 104780 + }, + { + "epoch": 9.07, + "learning_rate": 4.096508706575414e-05, + "loss": 1.2915, + "step": 104790 + }, + { + "epoch": 9.07, + "learning_rate": 4.096422073984233e-05, + "loss": 1.3245, + "step": 104800 + }, + { + "epoch": 9.07, + "learning_rate": 4.096335441393052e-05, + "loss": 1.3453, + "step": 104810 + }, + { + "epoch": 9.07, + "learning_rate": 4.096248808801871e-05, + "loss": 1.2893, + "step": 104820 + }, + { + "epoch": 9.07, + "learning_rate": 4.0961621762106905e-05, + "loss": 1.3812, + "step": 104830 + }, + { + "epoch": 9.07, + "learning_rate": 4.09607554361951e-05, + "loss": 1.3071, + "step": 104840 + }, + { + "epoch": 9.08, + "learning_rate": 4.095988911028329e-05, + "loss": 1.215, + "step": 104850 + }, + { + "epoch": 9.08, + "learning_rate": 4.095902278437148e-05, + "loss": 1.3421, + "step": 104860 + }, + { + "epoch": 9.08, + "learning_rate": 4.0958156458459676e-05, + "loss": 1.332, + "step": 104870 + }, + { + "epoch": 9.08, + "learning_rate": 4.0957290132547864e-05, + "loss": 1.2919, + "step": 104880 + }, + { + "epoch": 9.08, + "learning_rate": 4.095642380663606e-05, + "loss": 1.2948, + "step": 104890 + }, + { + "epoch": 9.08, + "learning_rate": 4.095555748072425e-05, + "loss": 1.2831, + "step": 104900 + }, + { + "epoch": 9.08, + "learning_rate": 4.095469115481244e-05, + "loss": 1.2891, + "step": 104910 + }, + { + "epoch": 9.08, + "learning_rate": 4.0953824828900635e-05, + "loss": 1.2481, + "step": 104920 + }, + { + "epoch": 9.08, + "learning_rate": 4.095295850298882e-05, + "loss": 1.2963, + "step": 104930 + }, + { + "epoch": 9.08, + "learning_rate": 4.095209217707702e-05, + "loss": 1.2947, + "step": 104940 + }, + { + "epoch": 9.08, + "learning_rate": 4.095122585116521e-05, + "loss": 1.3976, + "step": 104950 + }, + { + "epoch": 9.08, + "learning_rate": 4.09503595252534e-05, + "loss": 1.3356, + "step": 104960 + }, + { + "epoch": 9.09, + "learning_rate": 4.0949493199341593e-05, + "loss": 1.2392, + "step": 104970 + }, + { + "epoch": 9.09, + "learning_rate": 4.094862687342979e-05, + "loss": 1.3345, + "step": 104980 + }, + { + "epoch": 9.09, + "learning_rate": 4.0947760547517976e-05, + "loss": 1.318, + "step": 104990 + }, + { + "epoch": 9.09, + "learning_rate": 4.094689422160617e-05, + "loss": 1.3653, + "step": 105000 + }, + { + "epoch": 9.09, + "learning_rate": 4.094602789569436e-05, + "loss": 1.2515, + "step": 105010 + }, + { + "epoch": 9.09, + "learning_rate": 4.094516156978255e-05, + "loss": 1.2538, + "step": 105020 + }, + { + "epoch": 9.09, + "learning_rate": 4.0944295243870747e-05, + "loss": 1.3192, + "step": 105030 + }, + { + "epoch": 9.09, + "learning_rate": 4.0943428917958934e-05, + "loss": 1.2938, + "step": 105040 + }, + { + "epoch": 9.09, + "learning_rate": 4.094256259204713e-05, + "loss": 1.2385, + "step": 105050 + }, + { + "epoch": 9.09, + "learning_rate": 4.094169626613532e-05, + "loss": 1.3116, + "step": 105060 + }, + { + "epoch": 9.09, + "learning_rate": 4.094082994022351e-05, + "loss": 1.2853, + "step": 105070 + }, + { + "epoch": 9.1, + "learning_rate": 4.0939963614311705e-05, + "loss": 1.2517, + "step": 105080 + }, + { + "epoch": 9.1, + "learning_rate": 4.09390972883999e-05, + "loss": 1.2986, + "step": 105090 + }, + { + "epoch": 9.1, + "learning_rate": 4.093823096248809e-05, + "loss": 1.2622, + "step": 105100 + }, + { + "epoch": 9.1, + "learning_rate": 4.093736463657628e-05, + "loss": 1.3128, + "step": 105110 + }, + { + "epoch": 9.1, + "learning_rate": 4.093649831066447e-05, + "loss": 1.2602, + "step": 105120 + }, + { + "epoch": 9.1, + "learning_rate": 4.0935631984752664e-05, + "loss": 1.3508, + "step": 105130 + }, + { + "epoch": 9.1, + "learning_rate": 4.093476565884086e-05, + "loss": 1.2694, + "step": 105140 + }, + { + "epoch": 9.1, + "learning_rate": 4.0933899332929046e-05, + "loss": 1.2832, + "step": 105150 + }, + { + "epoch": 9.1, + "learning_rate": 4.093303300701724e-05, + "loss": 1.2908, + "step": 105160 + }, + { + "epoch": 9.1, + "learning_rate": 4.0932166681105435e-05, + "loss": 1.2321, + "step": 105170 + }, + { + "epoch": 9.1, + "learning_rate": 4.093130035519362e-05, + "loss": 1.3545, + "step": 105180 + }, + { + "epoch": 9.1, + "learning_rate": 4.093043402928182e-05, + "loss": 1.2873, + "step": 105190 + }, + { + "epoch": 9.11, + "learning_rate": 4.092956770337001e-05, + "loss": 1.2676, + "step": 105200 + }, + { + "epoch": 9.11, + "learning_rate": 4.09287013774582e-05, + "loss": 1.3013, + "step": 105210 + }, + { + "epoch": 9.11, + "learning_rate": 4.0927835051546394e-05, + "loss": 1.2909, + "step": 105220 + }, + { + "epoch": 9.11, + "learning_rate": 4.092696872563458e-05, + "loss": 1.2556, + "step": 105230 + }, + { + "epoch": 9.11, + "learning_rate": 4.0926102399722776e-05, + "loss": 1.32, + "step": 105240 + }, + { + "epoch": 9.11, + "learning_rate": 4.092523607381097e-05, + "loss": 1.3577, + "step": 105250 + }, + { + "epoch": 9.11, + "learning_rate": 4.092436974789916e-05, + "loss": 1.32, + "step": 105260 + }, + { + "epoch": 9.11, + "learning_rate": 4.092350342198735e-05, + "loss": 1.2572, + "step": 105270 + }, + { + "epoch": 9.11, + "learning_rate": 4.092263709607555e-05, + "loss": 1.2683, + "step": 105280 + }, + { + "epoch": 9.11, + "learning_rate": 4.0921770770163735e-05, + "loss": 1.3146, + "step": 105290 + }, + { + "epoch": 9.11, + "learning_rate": 4.092090444425193e-05, + "loss": 1.3019, + "step": 105300 + }, + { + "epoch": 9.12, + "learning_rate": 4.0920038118340124e-05, + "loss": 1.3332, + "step": 105310 + }, + { + "epoch": 9.12, + "learning_rate": 4.091917179242831e-05, + "loss": 1.3169, + "step": 105320 + }, + { + "epoch": 9.12, + "learning_rate": 4.0918305466516506e-05, + "loss": 1.2171, + "step": 105330 + }, + { + "epoch": 9.12, + "learning_rate": 4.091743914060469e-05, + "loss": 1.293, + "step": 105340 + }, + { + "epoch": 9.12, + "learning_rate": 4.091657281469289e-05, + "loss": 1.3127, + "step": 105350 + }, + { + "epoch": 9.12, + "learning_rate": 4.091570648878108e-05, + "loss": 1.3381, + "step": 105360 + }, + { + "epoch": 9.12, + "learning_rate": 4.091484016286927e-05, + "loss": 1.2853, + "step": 105370 + }, + { + "epoch": 9.12, + "learning_rate": 4.0913973836957464e-05, + "loss": 1.2301, + "step": 105380 + }, + { + "epoch": 9.12, + "learning_rate": 4.091310751104566e-05, + "loss": 1.2031, + "step": 105390 + }, + { + "epoch": 9.12, + "learning_rate": 4.0912241185133846e-05, + "loss": 1.3239, + "step": 105400 + }, + { + "epoch": 9.12, + "learning_rate": 4.091137485922204e-05, + "loss": 1.3069, + "step": 105410 + }, + { + "epoch": 9.12, + "learning_rate": 4.0910508533310235e-05, + "loss": 1.339, + "step": 105420 + }, + { + "epoch": 9.13, + "learning_rate": 4.090964220739842e-05, + "loss": 1.3261, + "step": 105430 + }, + { + "epoch": 9.13, + "learning_rate": 4.090877588148662e-05, + "loss": 1.3339, + "step": 105440 + }, + { + "epoch": 9.13, + "learning_rate": 4.0907909555574805e-05, + "loss": 1.2958, + "step": 105450 + }, + { + "epoch": 9.13, + "learning_rate": 4.0907043229663e-05, + "loss": 1.3107, + "step": 105460 + }, + { + "epoch": 9.13, + "learning_rate": 4.0906176903751194e-05, + "loss": 1.2556, + "step": 105470 + }, + { + "epoch": 9.13, + "learning_rate": 4.090531057783938e-05, + "loss": 1.3459, + "step": 105480 + }, + { + "epoch": 9.13, + "learning_rate": 4.0904444251927576e-05, + "loss": 1.3358, + "step": 105490 + }, + { + "epoch": 9.13, + "learning_rate": 4.090357792601577e-05, + "loss": 1.3315, + "step": 105500 + }, + { + "epoch": 9.13, + "learning_rate": 4.090271160010396e-05, + "loss": 1.2938, + "step": 105510 + }, + { + "epoch": 9.13, + "learning_rate": 4.090184527419215e-05, + "loss": 1.3104, + "step": 105520 + }, + { + "epoch": 9.13, + "learning_rate": 4.090097894828035e-05, + "loss": 1.3019, + "step": 105530 + }, + { + "epoch": 9.14, + "learning_rate": 4.0900112622368535e-05, + "loss": 1.2589, + "step": 105540 + }, + { + "epoch": 9.14, + "learning_rate": 4.089924629645673e-05, + "loss": 1.2552, + "step": 105550 + }, + { + "epoch": 9.14, + "learning_rate": 4.089837997054492e-05, + "loss": 1.2494, + "step": 105560 + }, + { + "epoch": 9.14, + "learning_rate": 4.089751364463311e-05, + "loss": 1.3022, + "step": 105570 + }, + { + "epoch": 9.14, + "learning_rate": 4.0896647318721306e-05, + "loss": 1.2569, + "step": 105580 + }, + { + "epoch": 9.14, + "learning_rate": 4.0895780992809494e-05, + "loss": 1.3469, + "step": 105590 + }, + { + "epoch": 9.14, + "learning_rate": 4.089491466689769e-05, + "loss": 1.3141, + "step": 105600 + }, + { + "epoch": 9.14, + "learning_rate": 4.089404834098588e-05, + "loss": 1.2577, + "step": 105610 + }, + { + "epoch": 9.14, + "learning_rate": 4.089318201507407e-05, + "loss": 1.3472, + "step": 105620 + }, + { + "epoch": 9.14, + "learning_rate": 4.0892315689162265e-05, + "loss": 1.2992, + "step": 105630 + }, + { + "epoch": 9.14, + "learning_rate": 4.089144936325046e-05, + "loss": 1.2262, + "step": 105640 + }, + { + "epoch": 9.14, + "learning_rate": 4.089058303733865e-05, + "loss": 1.2817, + "step": 105650 + }, + { + "epoch": 9.15, + "learning_rate": 4.088971671142684e-05, + "loss": 1.3146, + "step": 105660 + }, + { + "epoch": 9.15, + "learning_rate": 4.088885038551503e-05, + "loss": 1.3168, + "step": 105670 + }, + { + "epoch": 9.15, + "learning_rate": 4.0887984059603223e-05, + "loss": 1.3213, + "step": 105680 + }, + { + "epoch": 9.15, + "learning_rate": 4.088711773369142e-05, + "loss": 1.2511, + "step": 105690 + }, + { + "epoch": 9.15, + "learning_rate": 4.0886251407779606e-05, + "loss": 1.2519, + "step": 105700 + }, + { + "epoch": 9.15, + "learning_rate": 4.08853850818678e-05, + "loss": 1.3128, + "step": 105710 + }, + { + "epoch": 9.15, + "learning_rate": 4.0884518755955994e-05, + "loss": 1.3086, + "step": 105720 + }, + { + "epoch": 9.15, + "learning_rate": 4.088365243004418e-05, + "loss": 1.3387, + "step": 105730 + }, + { + "epoch": 9.15, + "learning_rate": 4.0882786104132377e-05, + "loss": 1.3064, + "step": 105740 + }, + { + "epoch": 9.15, + "learning_rate": 4.088191977822057e-05, + "loss": 1.2784, + "step": 105750 + }, + { + "epoch": 9.15, + "learning_rate": 4.088105345230876e-05, + "loss": 1.3061, + "step": 105760 + }, + { + "epoch": 9.16, + "learning_rate": 4.088018712639695e-05, + "loss": 1.3187, + "step": 105770 + }, + { + "epoch": 9.16, + "learning_rate": 4.087932080048514e-05, + "loss": 1.2814, + "step": 105780 + }, + { + "epoch": 9.16, + "learning_rate": 4.0878454474573335e-05, + "loss": 1.2297, + "step": 105790 + }, + { + "epoch": 9.16, + "learning_rate": 4.087758814866153e-05, + "loss": 1.2679, + "step": 105800 + }, + { + "epoch": 9.16, + "learning_rate": 4.087672182274972e-05, + "loss": 1.3014, + "step": 105810 + }, + { + "epoch": 9.16, + "learning_rate": 4.087585549683791e-05, + "loss": 1.2685, + "step": 105820 + }, + { + "epoch": 9.16, + "learning_rate": 4.0874989170926106e-05, + "loss": 1.2586, + "step": 105830 + }, + { + "epoch": 9.16, + "learning_rate": 4.0874122845014294e-05, + "loss": 1.2864, + "step": 105840 + }, + { + "epoch": 9.16, + "learning_rate": 4.087325651910249e-05, + "loss": 1.2685, + "step": 105850 + }, + { + "epoch": 9.16, + "learning_rate": 4.0872390193190676e-05, + "loss": 1.3786, + "step": 105860 + }, + { + "epoch": 9.16, + "learning_rate": 4.087152386727887e-05, + "loss": 1.2962, + "step": 105870 + }, + { + "epoch": 9.16, + "learning_rate": 4.0870657541367065e-05, + "loss": 1.2661, + "step": 105880 + }, + { + "epoch": 9.17, + "learning_rate": 4.086979121545525e-05, + "loss": 1.3137, + "step": 105890 + }, + { + "epoch": 9.17, + "learning_rate": 4.086892488954345e-05, + "loss": 1.2959, + "step": 105900 + }, + { + "epoch": 9.17, + "learning_rate": 4.086805856363164e-05, + "loss": 1.2639, + "step": 105910 + }, + { + "epoch": 9.17, + "learning_rate": 4.086719223771983e-05, + "loss": 1.3123, + "step": 105920 + }, + { + "epoch": 9.17, + "learning_rate": 4.0866325911808024e-05, + "loss": 1.3266, + "step": 105930 + }, + { + "epoch": 9.17, + "learning_rate": 4.086545958589622e-05, + "loss": 1.3157, + "step": 105940 + }, + { + "epoch": 9.17, + "learning_rate": 4.0864593259984406e-05, + "loss": 1.2918, + "step": 105950 + }, + { + "epoch": 9.17, + "learning_rate": 4.08637269340726e-05, + "loss": 1.2905, + "step": 105960 + }, + { + "epoch": 9.17, + "learning_rate": 4.086286060816079e-05, + "loss": 1.2807, + "step": 105970 + }, + { + "epoch": 9.17, + "learning_rate": 4.086199428224898e-05, + "loss": 1.3079, + "step": 105980 + }, + { + "epoch": 9.17, + "learning_rate": 4.086112795633718e-05, + "loss": 1.3107, + "step": 105990 + }, + { + "epoch": 9.17, + "learning_rate": 4.0860261630425365e-05, + "loss": 1.2863, + "step": 106000 + }, + { + "epoch": 9.18, + "learning_rate": 4.085939530451356e-05, + "loss": 1.2587, + "step": 106010 + }, + { + "epoch": 9.18, + "learning_rate": 4.0858528978601753e-05, + "loss": 1.2508, + "step": 106020 + }, + { + "epoch": 9.18, + "learning_rate": 4.085766265268994e-05, + "loss": 1.2707, + "step": 106030 + }, + { + "epoch": 9.18, + "learning_rate": 4.0856796326778136e-05, + "loss": 1.2812, + "step": 106040 + }, + { + "epoch": 9.18, + "learning_rate": 4.085593000086633e-05, + "loss": 1.3114, + "step": 106050 + }, + { + "epoch": 9.18, + "learning_rate": 4.085506367495452e-05, + "loss": 1.2969, + "step": 106060 + }, + { + "epoch": 9.18, + "learning_rate": 4.085419734904271e-05, + "loss": 1.2476, + "step": 106070 + }, + { + "epoch": 9.18, + "learning_rate": 4.08533310231309e-05, + "loss": 1.298, + "step": 106080 + }, + { + "epoch": 9.18, + "learning_rate": 4.0852464697219094e-05, + "loss": 1.2503, + "step": 106090 + }, + { + "epoch": 9.18, + "learning_rate": 4.085159837130729e-05, + "loss": 1.2778, + "step": 106100 + }, + { + "epoch": 9.18, + "learning_rate": 4.0850732045395476e-05, + "loss": 1.3036, + "step": 106110 + }, + { + "epoch": 9.19, + "learning_rate": 4.084986571948367e-05, + "loss": 1.336, + "step": 106120 + }, + { + "epoch": 9.19, + "learning_rate": 4.0848999393571865e-05, + "loss": 1.3643, + "step": 106130 + }, + { + "epoch": 9.19, + "learning_rate": 4.084813306766005e-05, + "loss": 1.3193, + "step": 106140 + }, + { + "epoch": 9.19, + "learning_rate": 4.084726674174825e-05, + "loss": 1.2956, + "step": 106150 + }, + { + "epoch": 9.19, + "learning_rate": 4.084640041583644e-05, + "loss": 1.3492, + "step": 106160 + }, + { + "epoch": 9.19, + "learning_rate": 4.084553408992463e-05, + "loss": 1.2745, + "step": 106170 + }, + { + "epoch": 9.19, + "learning_rate": 4.0844667764012824e-05, + "loss": 1.2499, + "step": 106180 + }, + { + "epoch": 9.19, + "learning_rate": 4.084380143810101e-05, + "loss": 1.3674, + "step": 106190 + }, + { + "epoch": 9.19, + "learning_rate": 4.0842935112189206e-05, + "loss": 1.3595, + "step": 106200 + }, + { + "epoch": 9.19, + "learning_rate": 4.08420687862774e-05, + "loss": 1.235, + "step": 106210 + }, + { + "epoch": 9.19, + "learning_rate": 4.084120246036559e-05, + "loss": 1.2592, + "step": 106220 + }, + { + "epoch": 9.19, + "learning_rate": 4.084033613445378e-05, + "loss": 1.279, + "step": 106230 + }, + { + "epoch": 9.2, + "learning_rate": 4.083946980854198e-05, + "loss": 1.3444, + "step": 106240 + }, + { + "epoch": 9.2, + "learning_rate": 4.0838603482630165e-05, + "loss": 1.2869, + "step": 106250 + }, + { + "epoch": 9.2, + "learning_rate": 4.083773715671836e-05, + "loss": 1.3247, + "step": 106260 + }, + { + "epoch": 9.2, + "learning_rate": 4.0836870830806554e-05, + "loss": 1.2793, + "step": 106270 + }, + { + "epoch": 9.2, + "learning_rate": 4.083600450489474e-05, + "loss": 1.2815, + "step": 106280 + }, + { + "epoch": 9.2, + "learning_rate": 4.0835138178982936e-05, + "loss": 1.3199, + "step": 106290 + }, + { + "epoch": 9.2, + "learning_rate": 4.0834271853071124e-05, + "loss": 1.3336, + "step": 106300 + }, + { + "epoch": 9.2, + "learning_rate": 4.083340552715932e-05, + "loss": 1.2635, + "step": 106310 + }, + { + "epoch": 9.2, + "learning_rate": 4.083253920124751e-05, + "loss": 1.3173, + "step": 106320 + }, + { + "epoch": 9.2, + "learning_rate": 4.08316728753357e-05, + "loss": 1.2464, + "step": 106330 + }, + { + "epoch": 9.2, + "learning_rate": 4.0830806549423895e-05, + "loss": 1.2775, + "step": 106340 + }, + { + "epoch": 9.21, + "learning_rate": 4.082994022351209e-05, + "loss": 1.203, + "step": 106350 + }, + { + "epoch": 9.21, + "learning_rate": 4.082907389760028e-05, + "loss": 1.2841, + "step": 106360 + }, + { + "epoch": 9.21, + "learning_rate": 4.082820757168847e-05, + "loss": 1.3617, + "step": 106370 + }, + { + "epoch": 9.21, + "learning_rate": 4.0827341245776666e-05, + "loss": 1.2287, + "step": 106380 + }, + { + "epoch": 9.21, + "learning_rate": 4.082647491986485e-05, + "loss": 1.2373, + "step": 106390 + }, + { + "epoch": 9.21, + "learning_rate": 4.082560859395305e-05, + "loss": 1.264, + "step": 106400 + }, + { + "epoch": 9.21, + "learning_rate": 4.0824742268041235e-05, + "loss": 1.2352, + "step": 106410 + }, + { + "epoch": 9.21, + "learning_rate": 4.082387594212943e-05, + "loss": 1.2916, + "step": 106420 + }, + { + "epoch": 9.21, + "learning_rate": 4.0823009616217624e-05, + "loss": 1.2966, + "step": 106430 + }, + { + "epoch": 9.21, + "learning_rate": 4.082214329030581e-05, + "loss": 1.367, + "step": 106440 + }, + { + "epoch": 9.21, + "learning_rate": 4.0821276964394007e-05, + "loss": 1.3243, + "step": 106450 + }, + { + "epoch": 9.21, + "learning_rate": 4.08204106384822e-05, + "loss": 1.3101, + "step": 106460 + }, + { + "epoch": 9.22, + "learning_rate": 4.081954431257039e-05, + "loss": 1.3212, + "step": 106470 + }, + { + "epoch": 9.22, + "learning_rate": 4.081867798665858e-05, + "loss": 1.3037, + "step": 106480 + }, + { + "epoch": 9.22, + "learning_rate": 4.081781166074678e-05, + "loss": 1.3201, + "step": 106490 + }, + { + "epoch": 9.22, + "learning_rate": 4.0816945334834965e-05, + "loss": 1.3057, + "step": 106500 + }, + { + "epoch": 9.22, + "learning_rate": 4.081607900892316e-05, + "loss": 1.305, + "step": 106510 + }, + { + "epoch": 9.22, + "learning_rate": 4.081521268301135e-05, + "loss": 1.2869, + "step": 106520 + }, + { + "epoch": 9.22, + "learning_rate": 4.081434635709954e-05, + "loss": 1.2829, + "step": 106530 + }, + { + "epoch": 9.22, + "learning_rate": 4.0813480031187736e-05, + "loss": 1.3741, + "step": 106540 + }, + { + "epoch": 9.22, + "learning_rate": 4.0812613705275924e-05, + "loss": 1.2984, + "step": 106550 + }, + { + "epoch": 9.22, + "learning_rate": 4.081174737936412e-05, + "loss": 1.2893, + "step": 106560 + }, + { + "epoch": 9.22, + "learning_rate": 4.081088105345231e-05, + "loss": 1.305, + "step": 106570 + }, + { + "epoch": 9.23, + "learning_rate": 4.08100147275405e-05, + "loss": 1.3353, + "step": 106580 + }, + { + "epoch": 9.23, + "learning_rate": 4.0809148401628695e-05, + "loss": 1.2347, + "step": 106590 + }, + { + "epoch": 9.23, + "learning_rate": 4.080828207571688e-05, + "loss": 1.2589, + "step": 106600 + }, + { + "epoch": 9.23, + "learning_rate": 4.080741574980508e-05, + "loss": 1.3172, + "step": 106610 + }, + { + "epoch": 9.23, + "learning_rate": 4.080654942389327e-05, + "loss": 1.2952, + "step": 106620 + }, + { + "epoch": 9.23, + "learning_rate": 4.080568309798146e-05, + "loss": 1.2733, + "step": 106630 + }, + { + "epoch": 9.23, + "learning_rate": 4.0804816772069654e-05, + "loss": 1.3363, + "step": 106640 + }, + { + "epoch": 9.23, + "learning_rate": 4.080395044615785e-05, + "loss": 1.2753, + "step": 106650 + }, + { + "epoch": 9.23, + "learning_rate": 4.0803084120246036e-05, + "loss": 1.1966, + "step": 106660 + }, + { + "epoch": 9.23, + "learning_rate": 4.080221779433423e-05, + "loss": 1.3799, + "step": 106670 + }, + { + "epoch": 9.23, + "learning_rate": 4.0801351468422425e-05, + "loss": 1.3065, + "step": 106680 + }, + { + "epoch": 9.23, + "learning_rate": 4.080048514251061e-05, + "loss": 1.2694, + "step": 106690 + }, + { + "epoch": 9.24, + "learning_rate": 4.079961881659881e-05, + "loss": 1.2703, + "step": 106700 + }, + { + "epoch": 9.24, + "learning_rate": 4.0798752490686995e-05, + "loss": 1.2832, + "step": 106710 + }, + { + "epoch": 9.24, + "learning_rate": 4.079788616477519e-05, + "loss": 1.317, + "step": 106720 + }, + { + "epoch": 9.24, + "learning_rate": 4.0797019838863383e-05, + "loss": 1.2808, + "step": 106730 + }, + { + "epoch": 9.24, + "learning_rate": 4.079615351295157e-05, + "loss": 1.2887, + "step": 106740 + }, + { + "epoch": 9.24, + "learning_rate": 4.0795287187039766e-05, + "loss": 1.3292, + "step": 106750 + }, + { + "epoch": 9.24, + "learning_rate": 4.079442086112796e-05, + "loss": 1.282, + "step": 106760 + }, + { + "epoch": 9.24, + "learning_rate": 4.079355453521615e-05, + "loss": 1.2376, + "step": 106770 + }, + { + "epoch": 9.24, + "learning_rate": 4.079268820930434e-05, + "loss": 1.2894, + "step": 106780 + }, + { + "epoch": 9.24, + "learning_rate": 4.0791821883392537e-05, + "loss": 1.2958, + "step": 106790 + }, + { + "epoch": 9.24, + "learning_rate": 4.0790955557480724e-05, + "loss": 1.292, + "step": 106800 + }, + { + "epoch": 9.25, + "learning_rate": 4.079008923156892e-05, + "loss": 1.3039, + "step": 106810 + }, + { + "epoch": 9.25, + "learning_rate": 4.0789222905657106e-05, + "loss": 1.3223, + "step": 106820 + }, + { + "epoch": 9.25, + "learning_rate": 4.07883565797453e-05, + "loss": 1.3025, + "step": 106830 + }, + { + "epoch": 9.25, + "learning_rate": 4.0787490253833495e-05, + "loss": 1.3296, + "step": 106840 + }, + { + "epoch": 9.25, + "learning_rate": 4.078662392792168e-05, + "loss": 1.3296, + "step": 106850 + }, + { + "epoch": 9.25, + "learning_rate": 4.078575760200988e-05, + "loss": 1.2839, + "step": 106860 + }, + { + "epoch": 9.25, + "learning_rate": 4.078489127609807e-05, + "loss": 1.2525, + "step": 106870 + }, + { + "epoch": 9.25, + "learning_rate": 4.078402495018626e-05, + "loss": 1.2013, + "step": 106880 + }, + { + "epoch": 9.25, + "learning_rate": 4.0783158624274454e-05, + "loss": 1.3055, + "step": 106890 + }, + { + "epoch": 9.25, + "learning_rate": 4.078229229836265e-05, + "loss": 1.2461, + "step": 106900 + }, + { + "epoch": 9.25, + "learning_rate": 4.0781425972450836e-05, + "loss": 1.2424, + "step": 106910 + }, + { + "epoch": 9.25, + "learning_rate": 4.078055964653903e-05, + "loss": 1.248, + "step": 106920 + }, + { + "epoch": 9.26, + "learning_rate": 4.077969332062722e-05, + "loss": 1.273, + "step": 106930 + }, + { + "epoch": 9.26, + "learning_rate": 4.077882699471541e-05, + "loss": 1.2828, + "step": 106940 + }, + { + "epoch": 9.26, + "learning_rate": 4.077796066880361e-05, + "loss": 1.2858, + "step": 106950 + }, + { + "epoch": 9.26, + "learning_rate": 4.0777094342891795e-05, + "loss": 1.3339, + "step": 106960 + }, + { + "epoch": 9.26, + "learning_rate": 4.077622801697999e-05, + "loss": 1.3115, + "step": 106970 + }, + { + "epoch": 9.26, + "learning_rate": 4.0775361691068184e-05, + "loss": 1.308, + "step": 106980 + }, + { + "epoch": 9.26, + "learning_rate": 4.077449536515637e-05, + "loss": 1.3348, + "step": 106990 + }, + { + "epoch": 9.26, + "learning_rate": 4.0773629039244566e-05, + "loss": 1.3226, + "step": 107000 + }, + { + "epoch": 9.26, + "learning_rate": 4.077276271333276e-05, + "loss": 1.2416, + "step": 107010 + }, + { + "epoch": 9.26, + "learning_rate": 4.077189638742095e-05, + "loss": 1.3296, + "step": 107020 + }, + { + "epoch": 9.26, + "learning_rate": 4.077103006150914e-05, + "loss": 1.2416, + "step": 107030 + }, + { + "epoch": 9.26, + "learning_rate": 4.077016373559733e-05, + "loss": 1.3196, + "step": 107040 + }, + { + "epoch": 9.27, + "learning_rate": 4.0769297409685525e-05, + "loss": 1.3465, + "step": 107050 + }, + { + "epoch": 9.27, + "learning_rate": 4.076843108377372e-05, + "loss": 1.2422, + "step": 107060 + }, + { + "epoch": 9.27, + "learning_rate": 4.076756475786191e-05, + "loss": 1.2778, + "step": 107070 + }, + { + "epoch": 9.27, + "learning_rate": 4.07666984319501e-05, + "loss": 1.2494, + "step": 107080 + }, + { + "epoch": 9.27, + "learning_rate": 4.0765832106038296e-05, + "loss": 1.3263, + "step": 107090 + }, + { + "epoch": 9.27, + "learning_rate": 4.076496578012648e-05, + "loss": 1.2853, + "step": 107100 + }, + { + "epoch": 9.27, + "learning_rate": 4.076409945421468e-05, + "loss": 1.2571, + "step": 107110 + }, + { + "epoch": 9.27, + "learning_rate": 4.076323312830287e-05, + "loss": 1.2815, + "step": 107120 + }, + { + "epoch": 9.27, + "learning_rate": 4.076236680239106e-05, + "loss": 1.2351, + "step": 107130 + }, + { + "epoch": 9.27, + "learning_rate": 4.0761500476479254e-05, + "loss": 1.331, + "step": 107140 + }, + { + "epoch": 9.27, + "learning_rate": 4.076063415056744e-05, + "loss": 1.2917, + "step": 107150 + }, + { + "epoch": 9.28, + "learning_rate": 4.0759767824655636e-05, + "loss": 1.2775, + "step": 107160 + }, + { + "epoch": 9.28, + "learning_rate": 4.075890149874383e-05, + "loss": 1.2631, + "step": 107170 + }, + { + "epoch": 9.28, + "learning_rate": 4.075803517283202e-05, + "loss": 1.2234, + "step": 107180 + }, + { + "epoch": 9.28, + "learning_rate": 4.075716884692021e-05, + "loss": 1.2797, + "step": 107190 + }, + { + "epoch": 9.28, + "learning_rate": 4.075630252100841e-05, + "loss": 1.2959, + "step": 107200 + }, + { + "epoch": 9.28, + "learning_rate": 4.0755436195096595e-05, + "loss": 1.277, + "step": 107210 + }, + { + "epoch": 9.28, + "learning_rate": 4.075456986918479e-05, + "loss": 1.298, + "step": 107220 + }, + { + "epoch": 9.28, + "learning_rate": 4.0753703543272984e-05, + "loss": 1.2823, + "step": 107230 + }, + { + "epoch": 9.28, + "learning_rate": 4.075283721736117e-05, + "loss": 1.3572, + "step": 107240 + }, + { + "epoch": 9.28, + "learning_rate": 4.0751970891449366e-05, + "loss": 1.2327, + "step": 107250 + }, + { + "epoch": 9.28, + "learning_rate": 4.0751104565537554e-05, + "loss": 1.2893, + "step": 107260 + }, + { + "epoch": 9.28, + "learning_rate": 4.075023823962575e-05, + "loss": 1.3017, + "step": 107270 + }, + { + "epoch": 9.29, + "learning_rate": 4.074937191371394e-05, + "loss": 1.301, + "step": 107280 + }, + { + "epoch": 9.29, + "learning_rate": 4.074850558780213e-05, + "loss": 1.2779, + "step": 107290 + }, + { + "epoch": 9.29, + "learning_rate": 4.0747639261890325e-05, + "loss": 1.301, + "step": 107300 + }, + { + "epoch": 9.29, + "learning_rate": 4.074677293597852e-05, + "loss": 1.267, + "step": 107310 + }, + { + "epoch": 9.29, + "learning_rate": 4.074590661006671e-05, + "loss": 1.3128, + "step": 107320 + }, + { + "epoch": 9.29, + "learning_rate": 4.07450402841549e-05, + "loss": 1.3443, + "step": 107330 + }, + { + "epoch": 9.29, + "learning_rate": 4.074417395824309e-05, + "loss": 1.2788, + "step": 107340 + }, + { + "epoch": 9.29, + "learning_rate": 4.0743307632331284e-05, + "loss": 1.2954, + "step": 107350 + }, + { + "epoch": 9.29, + "learning_rate": 4.074244130641948e-05, + "loss": 1.2065, + "step": 107360 + }, + { + "epoch": 9.29, + "learning_rate": 4.0741574980507666e-05, + "loss": 1.2651, + "step": 107370 + }, + { + "epoch": 9.29, + "learning_rate": 4.074070865459586e-05, + "loss": 1.2417, + "step": 107380 + }, + { + "epoch": 9.3, + "learning_rate": 4.0739842328684055e-05, + "loss": 1.2984, + "step": 107390 + }, + { + "epoch": 9.3, + "learning_rate": 4.073897600277224e-05, + "loss": 1.2662, + "step": 107400 + }, + { + "epoch": 9.3, + "learning_rate": 4.073810967686044e-05, + "loss": 1.3454, + "step": 107410 + }, + { + "epoch": 9.3, + "learning_rate": 4.073724335094863e-05, + "loss": 1.2243, + "step": 107420 + }, + { + "epoch": 9.3, + "learning_rate": 4.073637702503682e-05, + "loss": 1.2518, + "step": 107430 + }, + { + "epoch": 9.3, + "learning_rate": 4.0735510699125013e-05, + "loss": 1.3086, + "step": 107440 + }, + { + "epoch": 9.3, + "learning_rate": 4.07346443732132e-05, + "loss": 1.3269, + "step": 107450 + }, + { + "epoch": 9.3, + "learning_rate": 4.0733778047301396e-05, + "loss": 1.284, + "step": 107460 + }, + { + "epoch": 9.3, + "learning_rate": 4.073291172138959e-05, + "loss": 1.3001, + "step": 107470 + }, + { + "epoch": 9.3, + "learning_rate": 4.073204539547778e-05, + "loss": 1.3582, + "step": 107480 + }, + { + "epoch": 9.3, + "learning_rate": 4.073117906956597e-05, + "loss": 1.3356, + "step": 107490 + }, + { + "epoch": 9.3, + "learning_rate": 4.0730312743654167e-05, + "loss": 1.1985, + "step": 107500 + }, + { + "epoch": 9.31, + "learning_rate": 4.0729446417742354e-05, + "loss": 1.2711, + "step": 107510 + }, + { + "epoch": 9.31, + "learning_rate": 4.072858009183055e-05, + "loss": 1.264, + "step": 107520 + }, + { + "epoch": 9.31, + "learning_rate": 4.072771376591874e-05, + "loss": 1.2988, + "step": 107530 + }, + { + "epoch": 9.31, + "learning_rate": 4.072684744000693e-05, + "loss": 1.3346, + "step": 107540 + }, + { + "epoch": 9.31, + "learning_rate": 4.0725981114095125e-05, + "loss": 1.2683, + "step": 107550 + }, + { + "epoch": 9.31, + "learning_rate": 4.072511478818331e-05, + "loss": 1.2515, + "step": 107560 + }, + { + "epoch": 9.31, + "learning_rate": 4.072424846227151e-05, + "loss": 1.3008, + "step": 107570 + }, + { + "epoch": 9.31, + "learning_rate": 4.07233821363597e-05, + "loss": 1.2682, + "step": 107580 + }, + { + "epoch": 9.31, + "learning_rate": 4.072251581044789e-05, + "loss": 1.3286, + "step": 107590 + }, + { + "epoch": 9.31, + "learning_rate": 4.0721649484536084e-05, + "loss": 1.3095, + "step": 107600 + }, + { + "epoch": 9.31, + "learning_rate": 4.072078315862428e-05, + "loss": 1.3248, + "step": 107610 + }, + { + "epoch": 9.32, + "learning_rate": 4.0719916832712466e-05, + "loss": 1.3112, + "step": 107620 + }, + { + "epoch": 9.32, + "learning_rate": 4.071905050680066e-05, + "loss": 1.2702, + "step": 107630 + }, + { + "epoch": 9.32, + "learning_rate": 4.0718184180888855e-05, + "loss": 1.2576, + "step": 107640 + }, + { + "epoch": 9.32, + "learning_rate": 4.071731785497704e-05, + "loss": 1.3247, + "step": 107650 + }, + { + "epoch": 9.32, + "learning_rate": 4.071645152906524e-05, + "loss": 1.2963, + "step": 107660 + }, + { + "epoch": 9.32, + "learning_rate": 4.0715585203153425e-05, + "loss": 1.2654, + "step": 107670 + }, + { + "epoch": 9.32, + "learning_rate": 4.071471887724162e-05, + "loss": 1.3432, + "step": 107680 + }, + { + "epoch": 9.32, + "learning_rate": 4.0713852551329814e-05, + "loss": 1.2522, + "step": 107690 + }, + { + "epoch": 9.32, + "learning_rate": 4.0712986225418e-05, + "loss": 1.3071, + "step": 107700 + }, + { + "epoch": 9.32, + "learning_rate": 4.0712119899506196e-05, + "loss": 1.3181, + "step": 107710 + }, + { + "epoch": 9.32, + "learning_rate": 4.071125357359439e-05, + "loss": 1.3399, + "step": 107720 + }, + { + "epoch": 9.32, + "learning_rate": 4.071038724768258e-05, + "loss": 1.3081, + "step": 107730 + }, + { + "epoch": 9.33, + "learning_rate": 4.070952092177077e-05, + "loss": 1.282, + "step": 107740 + }, + { + "epoch": 9.33, + "learning_rate": 4.070865459585897e-05, + "loss": 1.2898, + "step": 107750 + }, + { + "epoch": 9.33, + "learning_rate": 4.0707788269947155e-05, + "loss": 1.2003, + "step": 107760 + }, + { + "epoch": 9.33, + "learning_rate": 4.070692194403535e-05, + "loss": 1.3217, + "step": 107770 + }, + { + "epoch": 9.33, + "learning_rate": 4.070605561812354e-05, + "loss": 1.227, + "step": 107780 + }, + { + "epoch": 9.33, + "learning_rate": 4.070518929221173e-05, + "loss": 1.331, + "step": 107790 + }, + { + "epoch": 9.33, + "learning_rate": 4.0704322966299926e-05, + "loss": 1.271, + "step": 107800 + }, + { + "epoch": 9.33, + "learning_rate": 4.070345664038811e-05, + "loss": 1.315, + "step": 107810 + }, + { + "epoch": 9.33, + "learning_rate": 4.070259031447631e-05, + "loss": 1.2944, + "step": 107820 + }, + { + "epoch": 9.33, + "learning_rate": 4.07017239885645e-05, + "loss": 1.2621, + "step": 107830 + }, + { + "epoch": 9.33, + "learning_rate": 4.070085766265269e-05, + "loss": 1.2947, + "step": 107840 + }, + { + "epoch": 9.34, + "learning_rate": 4.0699991336740884e-05, + "loss": 1.1874, + "step": 107850 + }, + { + "epoch": 9.34, + "learning_rate": 4.069912501082908e-05, + "loss": 1.2642, + "step": 107860 + }, + { + "epoch": 9.34, + "learning_rate": 4.0698258684917266e-05, + "loss": 1.2455, + "step": 107870 + }, + { + "epoch": 9.34, + "learning_rate": 4.069739235900546e-05, + "loss": 1.3027, + "step": 107880 + }, + { + "epoch": 9.34, + "learning_rate": 4.069652603309365e-05, + "loss": 1.3024, + "step": 107890 + }, + { + "epoch": 9.34, + "learning_rate": 4.069565970718184e-05, + "loss": 1.2048, + "step": 107900 + }, + { + "epoch": 9.34, + "learning_rate": 4.069479338127004e-05, + "loss": 1.3158, + "step": 107910 + }, + { + "epoch": 9.34, + "learning_rate": 4.0693927055358225e-05, + "loss": 1.3165, + "step": 107920 + }, + { + "epoch": 9.34, + "learning_rate": 4.069306072944642e-05, + "loss": 1.2695, + "step": 107930 + }, + { + "epoch": 9.34, + "learning_rate": 4.0692194403534614e-05, + "loss": 1.2457, + "step": 107940 + }, + { + "epoch": 9.34, + "learning_rate": 4.06913280776228e-05, + "loss": 1.2835, + "step": 107950 + }, + { + "epoch": 9.34, + "learning_rate": 4.0690461751710996e-05, + "loss": 1.2703, + "step": 107960 + }, + { + "epoch": 9.35, + "learning_rate": 4.068959542579919e-05, + "loss": 1.2763, + "step": 107970 + }, + { + "epoch": 9.35, + "learning_rate": 4.068872909988738e-05, + "loss": 1.3097, + "step": 107980 + }, + { + "epoch": 9.35, + "learning_rate": 4.068786277397557e-05, + "loss": 1.2836, + "step": 107990 + }, + { + "epoch": 9.35, + "learning_rate": 4.068699644806376e-05, + "loss": 1.2497, + "step": 108000 + }, + { + "epoch": 9.35, + "learning_rate": 4.0686130122151955e-05, + "loss": 1.2988, + "step": 108010 + }, + { + "epoch": 9.35, + "learning_rate": 4.068526379624015e-05, + "loss": 1.2463, + "step": 108020 + }, + { + "epoch": 9.35, + "learning_rate": 4.068439747032834e-05, + "loss": 1.3131, + "step": 108030 + }, + { + "epoch": 9.35, + "learning_rate": 4.068353114441653e-05, + "loss": 1.3001, + "step": 108040 + }, + { + "epoch": 9.35, + "learning_rate": 4.0682664818504726e-05, + "loss": 1.2432, + "step": 108050 + }, + { + "epoch": 9.35, + "learning_rate": 4.0681798492592914e-05, + "loss": 1.3558, + "step": 108060 + }, + { + "epoch": 9.35, + "learning_rate": 4.068093216668111e-05, + "loss": 1.2859, + "step": 108070 + }, + { + "epoch": 9.36, + "learning_rate": 4.06800658407693e-05, + "loss": 1.3164, + "step": 108080 + }, + { + "epoch": 9.36, + "learning_rate": 4.067919951485749e-05, + "loss": 1.318, + "step": 108090 + }, + { + "epoch": 9.36, + "learning_rate": 4.0678333188945685e-05, + "loss": 1.2374, + "step": 108100 + }, + { + "epoch": 9.36, + "learning_rate": 4.067746686303387e-05, + "loss": 1.3251, + "step": 108110 + }, + { + "epoch": 9.36, + "learning_rate": 4.067660053712207e-05, + "loss": 1.3371, + "step": 108120 + }, + { + "epoch": 9.36, + "learning_rate": 4.067573421121026e-05, + "loss": 1.2908, + "step": 108130 + }, + { + "epoch": 9.36, + "learning_rate": 4.067486788529845e-05, + "loss": 1.3247, + "step": 108140 + }, + { + "epoch": 9.36, + "learning_rate": 4.0674001559386643e-05, + "loss": 1.272, + "step": 108150 + }, + { + "epoch": 9.36, + "learning_rate": 4.067313523347484e-05, + "loss": 1.3317, + "step": 108160 + }, + { + "epoch": 9.36, + "learning_rate": 4.0672268907563026e-05, + "loss": 1.2519, + "step": 108170 + }, + { + "epoch": 9.36, + "learning_rate": 4.067140258165122e-05, + "loss": 1.1963, + "step": 108180 + }, + { + "epoch": 9.36, + "learning_rate": 4.067053625573941e-05, + "loss": 1.2642, + "step": 108190 + }, + { + "epoch": 9.37, + "learning_rate": 4.06696699298276e-05, + "loss": 1.2999, + "step": 108200 + }, + { + "epoch": 9.37, + "learning_rate": 4.0668803603915797e-05, + "loss": 1.2971, + "step": 108210 + }, + { + "epoch": 9.37, + "learning_rate": 4.0667937278003984e-05, + "loss": 1.2727, + "step": 108220 + }, + { + "epoch": 9.37, + "learning_rate": 4.066707095209218e-05, + "loss": 1.2801, + "step": 108230 + }, + { + "epoch": 9.37, + "learning_rate": 4.066620462618037e-05, + "loss": 1.2575, + "step": 108240 + }, + { + "epoch": 9.37, + "learning_rate": 4.066533830026856e-05, + "loss": 1.1453, + "step": 108250 + }, + { + "epoch": 9.37, + "learning_rate": 4.0664471974356755e-05, + "loss": 1.2898, + "step": 108260 + }, + { + "epoch": 9.37, + "learning_rate": 4.066360564844495e-05, + "loss": 1.2363, + "step": 108270 + }, + { + "epoch": 9.37, + "learning_rate": 4.066273932253314e-05, + "loss": 1.3258, + "step": 108280 + }, + { + "epoch": 9.37, + "learning_rate": 4.066187299662133e-05, + "loss": 1.2481, + "step": 108290 + }, + { + "epoch": 9.37, + "learning_rate": 4.066100667070952e-05, + "loss": 1.279, + "step": 108300 + }, + { + "epoch": 9.37, + "learning_rate": 4.0660140344797714e-05, + "loss": 1.3565, + "step": 108310 + }, + { + "epoch": 9.38, + "learning_rate": 4.065927401888591e-05, + "loss": 1.2675, + "step": 108320 + }, + { + "epoch": 9.38, + "learning_rate": 4.0658407692974096e-05, + "loss": 1.3156, + "step": 108330 + }, + { + "epoch": 9.38, + "learning_rate": 4.065754136706229e-05, + "loss": 1.3188, + "step": 108340 + }, + { + "epoch": 9.38, + "learning_rate": 4.0656675041150485e-05, + "loss": 1.2954, + "step": 108350 + }, + { + "epoch": 9.38, + "learning_rate": 4.065580871523867e-05, + "loss": 1.2581, + "step": 108360 + }, + { + "epoch": 9.38, + "learning_rate": 4.065494238932687e-05, + "loss": 1.3056, + "step": 108370 + }, + { + "epoch": 9.38, + "learning_rate": 4.065407606341506e-05, + "loss": 1.2931, + "step": 108380 + }, + { + "epoch": 9.38, + "learning_rate": 4.065320973750325e-05, + "loss": 1.3113, + "step": 108390 + }, + { + "epoch": 9.38, + "learning_rate": 4.0652343411591444e-05, + "loss": 1.298, + "step": 108400 + }, + { + "epoch": 9.38, + "learning_rate": 4.065147708567963e-05, + "loss": 1.2497, + "step": 108410 + }, + { + "epoch": 9.38, + "learning_rate": 4.0650610759767826e-05, + "loss": 1.3172, + "step": 108420 + }, + { + "epoch": 9.39, + "learning_rate": 4.064974443385602e-05, + "loss": 1.3359, + "step": 108430 + }, + { + "epoch": 9.39, + "learning_rate": 4.064887810794421e-05, + "loss": 1.2438, + "step": 108440 + }, + { + "epoch": 9.39, + "learning_rate": 4.06480117820324e-05, + "loss": 1.2986, + "step": 108450 + }, + { + "epoch": 9.39, + "learning_rate": 4.06471454561206e-05, + "loss": 1.3178, + "step": 108460 + }, + { + "epoch": 9.39, + "learning_rate": 4.0646279130208785e-05, + "loss": 1.2983, + "step": 108470 + }, + { + "epoch": 9.39, + "learning_rate": 4.064541280429698e-05, + "loss": 1.2993, + "step": 108480 + }, + { + "epoch": 9.39, + "learning_rate": 4.0644546478385173e-05, + "loss": 1.3009, + "step": 108490 + }, + { + "epoch": 9.39, + "learning_rate": 4.064368015247336e-05, + "loss": 1.2646, + "step": 108500 + }, + { + "epoch": 9.39, + "learning_rate": 4.0642813826561556e-05, + "loss": 1.3038, + "step": 108510 + }, + { + "epoch": 9.39, + "learning_rate": 4.064194750064974e-05, + "loss": 1.3052, + "step": 108520 + }, + { + "epoch": 9.39, + "learning_rate": 4.064108117473794e-05, + "loss": 1.27, + "step": 108530 + }, + { + "epoch": 9.39, + "learning_rate": 4.064021484882613e-05, + "loss": 1.3312, + "step": 108540 + }, + { + "epoch": 9.4, + "learning_rate": 4.063934852291432e-05, + "loss": 1.2956, + "step": 108550 + }, + { + "epoch": 9.4, + "learning_rate": 4.0638482197002514e-05, + "loss": 1.2215, + "step": 108560 + }, + { + "epoch": 9.4, + "learning_rate": 4.063761587109071e-05, + "loss": 1.3069, + "step": 108570 + }, + { + "epoch": 9.4, + "learning_rate": 4.0636749545178896e-05, + "loss": 1.2643, + "step": 108580 + }, + { + "epoch": 9.4, + "learning_rate": 4.063588321926709e-05, + "loss": 1.2864, + "step": 108590 + }, + { + "epoch": 9.4, + "learning_rate": 4.0635016893355285e-05, + "loss": 1.2338, + "step": 108600 + }, + { + "epoch": 9.4, + "learning_rate": 4.063415056744347e-05, + "loss": 1.2666, + "step": 108610 + }, + { + "epoch": 9.4, + "learning_rate": 4.063328424153167e-05, + "loss": 1.3248, + "step": 108620 + }, + { + "epoch": 9.4, + "learning_rate": 4.0632417915619855e-05, + "loss": 1.2611, + "step": 108630 + }, + { + "epoch": 9.4, + "learning_rate": 4.063155158970805e-05, + "loss": 1.2466, + "step": 108640 + }, + { + "epoch": 9.4, + "learning_rate": 4.0630685263796244e-05, + "loss": 1.2611, + "step": 108650 + }, + { + "epoch": 9.41, + "learning_rate": 4.062981893788443e-05, + "loss": 1.3289, + "step": 108660 + }, + { + "epoch": 9.41, + "learning_rate": 4.0628952611972626e-05, + "loss": 1.2928, + "step": 108670 + }, + { + "epoch": 9.41, + "learning_rate": 4.062808628606082e-05, + "loss": 1.3228, + "step": 108680 + }, + { + "epoch": 9.41, + "learning_rate": 4.062721996014901e-05, + "loss": 1.2889, + "step": 108690 + }, + { + "epoch": 9.41, + "learning_rate": 4.06263536342372e-05, + "loss": 1.2934, + "step": 108700 + }, + { + "epoch": 9.41, + "learning_rate": 4.06254873083254e-05, + "loss": 1.241, + "step": 108710 + }, + { + "epoch": 9.41, + "learning_rate": 4.0624620982413585e-05, + "loss": 1.2414, + "step": 108720 + }, + { + "epoch": 9.41, + "learning_rate": 4.062375465650178e-05, + "loss": 1.281, + "step": 108730 + }, + { + "epoch": 9.41, + "learning_rate": 4.062288833058997e-05, + "loss": 1.2583, + "step": 108740 + }, + { + "epoch": 9.41, + "learning_rate": 4.062202200467816e-05, + "loss": 1.2934, + "step": 108750 + }, + { + "epoch": 9.41, + "learning_rate": 4.0621155678766356e-05, + "loss": 1.2745, + "step": 108760 + }, + { + "epoch": 9.41, + "learning_rate": 4.0620289352854544e-05, + "loss": 1.3083, + "step": 108770 + }, + { + "epoch": 9.42, + "learning_rate": 4.061942302694274e-05, + "loss": 1.2899, + "step": 108780 + }, + { + "epoch": 9.42, + "learning_rate": 4.061855670103093e-05, + "loss": 1.2325, + "step": 108790 + }, + { + "epoch": 9.42, + "learning_rate": 4.061769037511912e-05, + "loss": 1.294, + "step": 108800 + }, + { + "epoch": 9.42, + "learning_rate": 4.0616824049207315e-05, + "loss": 1.2252, + "step": 108810 + }, + { + "epoch": 9.42, + "learning_rate": 4.061595772329551e-05, + "loss": 1.254, + "step": 108820 + }, + { + "epoch": 9.42, + "learning_rate": 4.06150913973837e-05, + "loss": 1.3276, + "step": 108830 + }, + { + "epoch": 9.42, + "learning_rate": 4.061422507147189e-05, + "loss": 1.3035, + "step": 108840 + }, + { + "epoch": 9.42, + "learning_rate": 4.061335874556008e-05, + "loss": 1.2832, + "step": 108850 + }, + { + "epoch": 9.42, + "learning_rate": 4.061249241964827e-05, + "loss": 1.2454, + "step": 108860 + }, + { + "epoch": 9.42, + "learning_rate": 4.061162609373647e-05, + "loss": 1.2366, + "step": 108870 + }, + { + "epoch": 9.42, + "learning_rate": 4.0610759767824655e-05, + "loss": 1.2783, + "step": 108880 + }, + { + "epoch": 9.43, + "learning_rate": 4.060989344191285e-05, + "loss": 1.3678, + "step": 108890 + }, + { + "epoch": 9.43, + "learning_rate": 4.0609027116001044e-05, + "loss": 1.2554, + "step": 108900 + }, + { + "epoch": 9.43, + "learning_rate": 4.060816079008923e-05, + "loss": 1.3032, + "step": 108910 + }, + { + "epoch": 9.43, + "learning_rate": 4.0607294464177427e-05, + "loss": 1.2887, + "step": 108920 + }, + { + "epoch": 9.43, + "learning_rate": 4.0606428138265614e-05, + "loss": 1.2387, + "step": 108930 + }, + { + "epoch": 9.43, + "learning_rate": 4.060556181235381e-05, + "loss": 1.2952, + "step": 108940 + }, + { + "epoch": 9.43, + "learning_rate": 4.0604695486442e-05, + "loss": 1.2711, + "step": 108950 + }, + { + "epoch": 9.43, + "learning_rate": 4.060382916053019e-05, + "loss": 1.2466, + "step": 108960 + }, + { + "epoch": 9.43, + "learning_rate": 4.0602962834618385e-05, + "loss": 1.28, + "step": 108970 + }, + { + "epoch": 9.43, + "learning_rate": 4.060209650870658e-05, + "loss": 1.2673, + "step": 108980 + }, + { + "epoch": 9.43, + "learning_rate": 4.060123018279477e-05, + "loss": 1.3131, + "step": 108990 + }, + { + "epoch": 9.43, + "learning_rate": 4.060036385688296e-05, + "loss": 1.2915, + "step": 109000 + }, + { + "epoch": 9.44, + "learning_rate": 4.0599497530971156e-05, + "loss": 1.2934, + "step": 109010 + }, + { + "epoch": 9.44, + "learning_rate": 4.0598631205059344e-05, + "loss": 1.3062, + "step": 109020 + }, + { + "epoch": 9.44, + "learning_rate": 4.059776487914754e-05, + "loss": 1.2771, + "step": 109030 + }, + { + "epoch": 9.44, + "learning_rate": 4.0596898553235726e-05, + "loss": 1.3115, + "step": 109040 + }, + { + "epoch": 9.44, + "learning_rate": 4.059603222732392e-05, + "loss": 1.3094, + "step": 109050 + }, + { + "epoch": 9.44, + "learning_rate": 4.0595165901412115e-05, + "loss": 1.2868, + "step": 109060 + }, + { + "epoch": 9.44, + "learning_rate": 4.05942995755003e-05, + "loss": 1.2535, + "step": 109070 + }, + { + "epoch": 9.44, + "learning_rate": 4.05934332495885e-05, + "loss": 1.2932, + "step": 109080 + }, + { + "epoch": 9.44, + "learning_rate": 4.059256692367669e-05, + "loss": 1.254, + "step": 109090 + }, + { + "epoch": 9.44, + "learning_rate": 4.059170059776488e-05, + "loss": 1.3096, + "step": 109100 + }, + { + "epoch": 9.44, + "learning_rate": 4.0590834271853074e-05, + "loss": 1.2585, + "step": 109110 + }, + { + "epoch": 9.45, + "learning_rate": 4.058996794594127e-05, + "loss": 1.3258, + "step": 109120 + }, + { + "epoch": 9.45, + "learning_rate": 4.0589101620029456e-05, + "loss": 1.2793, + "step": 109130 + }, + { + "epoch": 9.45, + "learning_rate": 4.058823529411765e-05, + "loss": 1.2927, + "step": 109140 + }, + { + "epoch": 9.45, + "learning_rate": 4.058736896820584e-05, + "loss": 1.2757, + "step": 109150 + }, + { + "epoch": 9.45, + "learning_rate": 4.058650264229403e-05, + "loss": 1.2725, + "step": 109160 + }, + { + "epoch": 9.45, + "learning_rate": 4.058563631638223e-05, + "loss": 1.2497, + "step": 109170 + }, + { + "epoch": 9.45, + "learning_rate": 4.0584769990470415e-05, + "loss": 1.3142, + "step": 109180 + }, + { + "epoch": 9.45, + "learning_rate": 4.058390366455861e-05, + "loss": 1.2987, + "step": 109190 + }, + { + "epoch": 9.45, + "learning_rate": 4.0583037338646803e-05, + "loss": 1.2613, + "step": 109200 + }, + { + "epoch": 9.45, + "learning_rate": 4.058217101273499e-05, + "loss": 1.2751, + "step": 109210 + }, + { + "epoch": 9.45, + "learning_rate": 4.0581304686823186e-05, + "loss": 1.3337, + "step": 109220 + }, + { + "epoch": 9.45, + "learning_rate": 4.058043836091138e-05, + "loss": 1.2717, + "step": 109230 + }, + { + "epoch": 9.46, + "learning_rate": 4.057957203499957e-05, + "loss": 1.2314, + "step": 109240 + }, + { + "epoch": 9.46, + "learning_rate": 4.057870570908776e-05, + "loss": 1.2969, + "step": 109250 + }, + { + "epoch": 9.46, + "learning_rate": 4.057783938317595e-05, + "loss": 1.2266, + "step": 109260 + }, + { + "epoch": 9.46, + "learning_rate": 4.0576973057264144e-05, + "loss": 1.3017, + "step": 109270 + }, + { + "epoch": 9.46, + "learning_rate": 4.057610673135234e-05, + "loss": 1.2879, + "step": 109280 + }, + { + "epoch": 9.46, + "learning_rate": 4.0575240405440526e-05, + "loss": 1.3236, + "step": 109290 + }, + { + "epoch": 9.46, + "learning_rate": 4.057437407952872e-05, + "loss": 1.2673, + "step": 109300 + }, + { + "epoch": 9.46, + "learning_rate": 4.0573507753616915e-05, + "loss": 1.2642, + "step": 109310 + }, + { + "epoch": 9.46, + "learning_rate": 4.05726414277051e-05, + "loss": 1.3231, + "step": 109320 + }, + { + "epoch": 9.46, + "learning_rate": 4.05717751017933e-05, + "loss": 1.2408, + "step": 109330 + }, + { + "epoch": 9.46, + "learning_rate": 4.057090877588149e-05, + "loss": 1.2759, + "step": 109340 + }, + { + "epoch": 9.46, + "learning_rate": 4.057004244996968e-05, + "loss": 1.2614, + "step": 109350 + }, + { + "epoch": 9.47, + "learning_rate": 4.0569176124057874e-05, + "loss": 1.3257, + "step": 109360 + }, + { + "epoch": 9.47, + "learning_rate": 4.056830979814606e-05, + "loss": 1.24, + "step": 109370 + }, + { + "epoch": 9.47, + "learning_rate": 4.0567443472234256e-05, + "loss": 1.2876, + "step": 109380 + }, + { + "epoch": 9.47, + "learning_rate": 4.056657714632245e-05, + "loss": 1.2987, + "step": 109390 + }, + { + "epoch": 9.47, + "learning_rate": 4.056571082041064e-05, + "loss": 1.2767, + "step": 109400 + }, + { + "epoch": 9.47, + "learning_rate": 4.056484449449883e-05, + "loss": 1.3232, + "step": 109410 + }, + { + "epoch": 9.47, + "learning_rate": 4.056397816858703e-05, + "loss": 1.2883, + "step": 109420 + }, + { + "epoch": 9.47, + "learning_rate": 4.0563111842675215e-05, + "loss": 1.2906, + "step": 109430 + }, + { + "epoch": 9.47, + "learning_rate": 4.056224551676341e-05, + "loss": 1.3019, + "step": 109440 + }, + { + "epoch": 9.47, + "learning_rate": 4.0561379190851604e-05, + "loss": 1.3115, + "step": 109450 + }, + { + "epoch": 9.47, + "learning_rate": 4.056051286493979e-05, + "loss": 1.2986, + "step": 109460 + }, + { + "epoch": 9.48, + "learning_rate": 4.0559646539027986e-05, + "loss": 1.2904, + "step": 109470 + }, + { + "epoch": 9.48, + "learning_rate": 4.0558780213116174e-05, + "loss": 1.2754, + "step": 109480 + }, + { + "epoch": 9.48, + "learning_rate": 4.055791388720437e-05, + "loss": 1.3, + "step": 109490 + }, + { + "epoch": 9.48, + "learning_rate": 4.055704756129256e-05, + "loss": 1.3667, + "step": 109500 + }, + { + "epoch": 9.48, + "learning_rate": 4.055618123538075e-05, + "loss": 1.2803, + "step": 109510 + }, + { + "epoch": 9.48, + "learning_rate": 4.0555314909468945e-05, + "loss": 1.2408, + "step": 109520 + }, + { + "epoch": 9.48, + "learning_rate": 4.055444858355714e-05, + "loss": 1.2294, + "step": 109530 + }, + { + "epoch": 9.48, + "learning_rate": 4.055358225764533e-05, + "loss": 1.2362, + "step": 109540 + }, + { + "epoch": 9.48, + "learning_rate": 4.055271593173352e-05, + "loss": 1.2722, + "step": 109550 + }, + { + "epoch": 9.48, + "learning_rate": 4.0551849605821716e-05, + "loss": 1.2467, + "step": 109560 + }, + { + "epoch": 9.48, + "learning_rate": 4.05509832799099e-05, + "loss": 1.2751, + "step": 109570 + }, + { + "epoch": 9.48, + "learning_rate": 4.05501169539981e-05, + "loss": 1.3052, + "step": 109580 + }, + { + "epoch": 9.49, + "learning_rate": 4.0549250628086285e-05, + "loss": 1.2271, + "step": 109590 + }, + { + "epoch": 9.49, + "learning_rate": 4.054838430217448e-05, + "loss": 1.3007, + "step": 109600 + }, + { + "epoch": 9.49, + "learning_rate": 4.0547517976262674e-05, + "loss": 1.279, + "step": 109610 + }, + { + "epoch": 9.49, + "learning_rate": 4.054665165035086e-05, + "loss": 1.2795, + "step": 109620 + }, + { + "epoch": 9.49, + "learning_rate": 4.0545785324439056e-05, + "loss": 1.245, + "step": 109630 + }, + { + "epoch": 9.49, + "learning_rate": 4.054491899852725e-05, + "loss": 1.2672, + "step": 109640 + }, + { + "epoch": 9.49, + "learning_rate": 4.054405267261544e-05, + "loss": 1.2372, + "step": 109650 + }, + { + "epoch": 9.49, + "learning_rate": 4.054318634670363e-05, + "loss": 1.2821, + "step": 109660 + }, + { + "epoch": 9.49, + "learning_rate": 4.054232002079182e-05, + "loss": 1.285, + "step": 109670 + }, + { + "epoch": 9.49, + "learning_rate": 4.0541453694880015e-05, + "loss": 1.2916, + "step": 109680 + }, + { + "epoch": 9.49, + "learning_rate": 4.054058736896821e-05, + "loss": 1.2465, + "step": 109690 + }, + { + "epoch": 9.5, + "learning_rate": 4.05397210430564e-05, + "loss": 1.283, + "step": 109700 + }, + { + "epoch": 9.5, + "learning_rate": 4.053885471714459e-05, + "loss": 1.2824, + "step": 109710 + }, + { + "epoch": 9.5, + "learning_rate": 4.0537988391232786e-05, + "loss": 1.3089, + "step": 109720 + }, + { + "epoch": 9.5, + "learning_rate": 4.0537122065320974e-05, + "loss": 1.3212, + "step": 109730 + }, + { + "epoch": 9.5, + "learning_rate": 4.053625573940917e-05, + "loss": 1.2804, + "step": 109740 + }, + { + "epoch": 9.5, + "learning_rate": 4.053538941349736e-05, + "loss": 1.3435, + "step": 109750 + }, + { + "epoch": 9.5, + "learning_rate": 4.053452308758555e-05, + "loss": 1.2858, + "step": 109760 + }, + { + "epoch": 9.5, + "learning_rate": 4.0533656761673745e-05, + "loss": 1.2351, + "step": 109770 + }, + { + "epoch": 9.5, + "learning_rate": 4.053279043576193e-05, + "loss": 1.2538, + "step": 109780 + }, + { + "epoch": 9.5, + "learning_rate": 4.053192410985013e-05, + "loss": 1.3299, + "step": 109790 + }, + { + "epoch": 9.5, + "learning_rate": 4.053105778393832e-05, + "loss": 1.2584, + "step": 109800 + }, + { + "epoch": 9.5, + "learning_rate": 4.053019145802651e-05, + "loss": 1.2955, + "step": 109810 + }, + { + "epoch": 9.51, + "learning_rate": 4.0529325132114704e-05, + "loss": 1.2879, + "step": 109820 + }, + { + "epoch": 9.51, + "learning_rate": 4.05284588062029e-05, + "loss": 1.3177, + "step": 109830 + }, + { + "epoch": 9.51, + "learning_rate": 4.0527592480291086e-05, + "loss": 1.1974, + "step": 109840 + }, + { + "epoch": 9.51, + "learning_rate": 4.052672615437928e-05, + "loss": 1.2941, + "step": 109850 + }, + { + "epoch": 9.51, + "learning_rate": 4.0525859828467475e-05, + "loss": 1.2699, + "step": 109860 + }, + { + "epoch": 9.51, + "learning_rate": 4.052499350255566e-05, + "loss": 1.1672, + "step": 109870 + }, + { + "epoch": 9.51, + "learning_rate": 4.052412717664386e-05, + "loss": 1.2575, + "step": 109880 + }, + { + "epoch": 9.51, + "learning_rate": 4.0523260850732045e-05, + "loss": 1.3099, + "step": 109890 + }, + { + "epoch": 9.51, + "learning_rate": 4.052239452482024e-05, + "loss": 1.2601, + "step": 109900 + }, + { + "epoch": 9.51, + "learning_rate": 4.0521528198908433e-05, + "loss": 1.3716, + "step": 109910 + }, + { + "epoch": 9.51, + "learning_rate": 4.052066187299662e-05, + "loss": 1.2698, + "step": 109920 + }, + { + "epoch": 9.52, + "learning_rate": 4.0519795547084816e-05, + "loss": 1.2755, + "step": 109930 + }, + { + "epoch": 9.52, + "learning_rate": 4.051892922117301e-05, + "loss": 1.2742, + "step": 109940 + }, + { + "epoch": 9.52, + "learning_rate": 4.05180628952612e-05, + "loss": 1.2863, + "step": 109950 + }, + { + "epoch": 9.52, + "learning_rate": 4.051719656934939e-05, + "loss": 1.2726, + "step": 109960 + }, + { + "epoch": 9.52, + "learning_rate": 4.0516330243437587e-05, + "loss": 1.2743, + "step": 109970 + }, + { + "epoch": 9.52, + "learning_rate": 4.0515463917525774e-05, + "loss": 1.3319, + "step": 109980 + }, + { + "epoch": 9.52, + "learning_rate": 4.051459759161397e-05, + "loss": 1.3108, + "step": 109990 + }, + { + "epoch": 9.52, + "learning_rate": 4.0513731265702156e-05, + "loss": 1.3286, + "step": 110000 + }, + { + "epoch": 9.52, + "learning_rate": 4.051286493979035e-05, + "loss": 1.299, + "step": 110010 + }, + { + "epoch": 9.52, + "learning_rate": 4.0511998613878545e-05, + "loss": 1.3028, + "step": 110020 + }, + { + "epoch": 9.52, + "learning_rate": 4.051113228796673e-05, + "loss": 1.2786, + "step": 110030 + }, + { + "epoch": 9.52, + "learning_rate": 4.051026596205493e-05, + "loss": 1.3126, + "step": 110040 + }, + { + "epoch": 9.53, + "learning_rate": 4.050939963614312e-05, + "loss": 1.2497, + "step": 110050 + }, + { + "epoch": 9.53, + "learning_rate": 4.050853331023131e-05, + "loss": 1.2877, + "step": 110060 + }, + { + "epoch": 9.53, + "learning_rate": 4.0507666984319504e-05, + "loss": 1.2515, + "step": 110070 + }, + { + "epoch": 9.53, + "learning_rate": 4.05068006584077e-05, + "loss": 1.3066, + "step": 110080 + }, + { + "epoch": 9.53, + "learning_rate": 4.0505934332495886e-05, + "loss": 1.2837, + "step": 110090 + }, + { + "epoch": 9.53, + "learning_rate": 4.050506800658408e-05, + "loss": 1.298, + "step": 110100 + }, + { + "epoch": 9.53, + "learning_rate": 4.050420168067227e-05, + "loss": 1.2957, + "step": 110110 + }, + { + "epoch": 9.53, + "learning_rate": 4.050333535476046e-05, + "loss": 1.2943, + "step": 110120 + }, + { + "epoch": 9.53, + "learning_rate": 4.050246902884866e-05, + "loss": 1.3203, + "step": 110130 + }, + { + "epoch": 9.53, + "learning_rate": 4.0501602702936845e-05, + "loss": 1.2824, + "step": 110140 + }, + { + "epoch": 9.53, + "learning_rate": 4.050073637702504e-05, + "loss": 1.2405, + "step": 110150 + }, + { + "epoch": 9.54, + "learning_rate": 4.0499870051113234e-05, + "loss": 1.2288, + "step": 110160 + }, + { + "epoch": 9.54, + "learning_rate": 4.049900372520142e-05, + "loss": 1.2776, + "step": 110170 + }, + { + "epoch": 9.54, + "learning_rate": 4.0498137399289616e-05, + "loss": 1.244, + "step": 110180 + }, + { + "epoch": 9.54, + "learning_rate": 4.049727107337781e-05, + "loss": 1.3067, + "step": 110190 + }, + { + "epoch": 9.54, + "learning_rate": 4.0496404747466e-05, + "loss": 1.2692, + "step": 110200 + }, + { + "epoch": 9.54, + "learning_rate": 4.049553842155419e-05, + "loss": 1.3286, + "step": 110210 + }, + { + "epoch": 9.54, + "learning_rate": 4.049467209564238e-05, + "loss": 1.2136, + "step": 110220 + }, + { + "epoch": 9.54, + "learning_rate": 4.0493805769730575e-05, + "loss": 1.2245, + "step": 110230 + }, + { + "epoch": 9.54, + "learning_rate": 4.049293944381877e-05, + "loss": 1.3072, + "step": 110240 + }, + { + "epoch": 9.54, + "learning_rate": 4.049207311790696e-05, + "loss": 1.2496, + "step": 110250 + }, + { + "epoch": 9.54, + "learning_rate": 4.049120679199515e-05, + "loss": 1.2458, + "step": 110260 + }, + { + "epoch": 9.54, + "learning_rate": 4.0490340466083346e-05, + "loss": 1.2779, + "step": 110270 + }, + { + "epoch": 9.55, + "learning_rate": 4.048947414017153e-05, + "loss": 1.2366, + "step": 110280 + }, + { + "epoch": 9.55, + "learning_rate": 4.048860781425973e-05, + "loss": 1.3026, + "step": 110290 + }, + { + "epoch": 9.55, + "learning_rate": 4.048774148834792e-05, + "loss": 1.2958, + "step": 110300 + }, + { + "epoch": 9.55, + "learning_rate": 4.048687516243611e-05, + "loss": 1.2467, + "step": 110310 + }, + { + "epoch": 9.55, + "learning_rate": 4.0486008836524304e-05, + "loss": 1.2322, + "step": 110320 + }, + { + "epoch": 9.55, + "learning_rate": 4.048514251061249e-05, + "loss": 1.2815, + "step": 110330 + }, + { + "epoch": 9.55, + "learning_rate": 4.0484276184700686e-05, + "loss": 1.2408, + "step": 110340 + }, + { + "epoch": 9.55, + "learning_rate": 4.048340985878888e-05, + "loss": 1.2927, + "step": 110350 + }, + { + "epoch": 9.55, + "learning_rate": 4.048254353287707e-05, + "loss": 1.271, + "step": 110360 + }, + { + "epoch": 9.55, + "learning_rate": 4.048167720696526e-05, + "loss": 1.2737, + "step": 110370 + }, + { + "epoch": 9.55, + "learning_rate": 4.048081088105346e-05, + "loss": 1.2467, + "step": 110380 + }, + { + "epoch": 9.55, + "learning_rate": 4.0479944555141645e-05, + "loss": 1.2881, + "step": 110390 + }, + { + "epoch": 9.56, + "learning_rate": 4.047907822922984e-05, + "loss": 1.2503, + "step": 110400 + }, + { + "epoch": 9.56, + "learning_rate": 4.0478211903318034e-05, + "loss": 1.2712, + "step": 110410 + }, + { + "epoch": 9.56, + "learning_rate": 4.047734557740622e-05, + "loss": 1.2453, + "step": 110420 + }, + { + "epoch": 9.56, + "learning_rate": 4.0476479251494416e-05, + "loss": 1.3018, + "step": 110430 + }, + { + "epoch": 9.56, + "learning_rate": 4.0475612925582604e-05, + "loss": 1.2661, + "step": 110440 + }, + { + "epoch": 9.56, + "learning_rate": 4.04747465996708e-05, + "loss": 1.3099, + "step": 110450 + }, + { + "epoch": 9.56, + "learning_rate": 4.047388027375899e-05, + "loss": 1.2286, + "step": 110460 + }, + { + "epoch": 9.56, + "learning_rate": 4.047301394784718e-05, + "loss": 1.3264, + "step": 110470 + }, + { + "epoch": 9.56, + "learning_rate": 4.0472147621935375e-05, + "loss": 1.2434, + "step": 110480 + }, + { + "epoch": 9.56, + "learning_rate": 4.047128129602357e-05, + "loss": 1.255, + "step": 110490 + }, + { + "epoch": 9.56, + "learning_rate": 4.047041497011176e-05, + "loss": 1.2871, + "step": 110500 + }, + { + "epoch": 9.57, + "learning_rate": 4.046954864419995e-05, + "loss": 1.244, + "step": 110510 + }, + { + "epoch": 9.57, + "learning_rate": 4.046868231828814e-05, + "loss": 1.2367, + "step": 110520 + }, + { + "epoch": 9.57, + "learning_rate": 4.0467815992376334e-05, + "loss": 1.2355, + "step": 110530 + }, + { + "epoch": 9.57, + "learning_rate": 4.046694966646453e-05, + "loss": 1.2912, + "step": 110540 + }, + { + "epoch": 9.57, + "learning_rate": 4.0466083340552716e-05, + "loss": 1.236, + "step": 110550 + }, + { + "epoch": 9.57, + "learning_rate": 4.046521701464091e-05, + "loss": 1.2793, + "step": 110560 + }, + { + "epoch": 9.57, + "learning_rate": 4.0464350688729105e-05, + "loss": 1.2027, + "step": 110570 + }, + { + "epoch": 9.57, + "learning_rate": 4.046348436281729e-05, + "loss": 1.2295, + "step": 110580 + }, + { + "epoch": 9.57, + "learning_rate": 4.046261803690549e-05, + "loss": 1.2988, + "step": 110590 + }, + { + "epoch": 9.57, + "learning_rate": 4.046175171099368e-05, + "loss": 1.3128, + "step": 110600 + }, + { + "epoch": 9.57, + "learning_rate": 4.046088538508187e-05, + "loss": 1.2614, + "step": 110610 + }, + { + "epoch": 9.57, + "learning_rate": 4.046001905917006e-05, + "loss": 1.2203, + "step": 110620 + }, + { + "epoch": 9.58, + "learning_rate": 4.045915273325825e-05, + "loss": 1.2006, + "step": 110630 + }, + { + "epoch": 9.58, + "learning_rate": 4.0458286407346446e-05, + "loss": 1.2291, + "step": 110640 + }, + { + "epoch": 9.58, + "learning_rate": 4.045742008143464e-05, + "loss": 1.3563, + "step": 110650 + }, + { + "epoch": 9.58, + "learning_rate": 4.045655375552283e-05, + "loss": 1.293, + "step": 110660 + }, + { + "epoch": 9.58, + "learning_rate": 4.045568742961102e-05, + "loss": 1.2451, + "step": 110670 + }, + { + "epoch": 9.58, + "learning_rate": 4.0454821103699217e-05, + "loss": 1.3055, + "step": 110680 + }, + { + "epoch": 9.58, + "learning_rate": 4.0453954777787404e-05, + "loss": 1.2622, + "step": 110690 + }, + { + "epoch": 9.58, + "learning_rate": 4.04530884518756e-05, + "loss": 1.2464, + "step": 110700 + }, + { + "epoch": 9.58, + "learning_rate": 4.045222212596379e-05, + "loss": 1.2568, + "step": 110710 + }, + { + "epoch": 9.58, + "learning_rate": 4.045135580005198e-05, + "loss": 1.3179, + "step": 110720 + }, + { + "epoch": 9.58, + "learning_rate": 4.0450489474140175e-05, + "loss": 1.2179, + "step": 110730 + }, + { + "epoch": 9.59, + "learning_rate": 4.044962314822836e-05, + "loss": 1.3134, + "step": 110740 + }, + { + "epoch": 9.59, + "learning_rate": 4.044875682231656e-05, + "loss": 1.2988, + "step": 110750 + }, + { + "epoch": 9.59, + "learning_rate": 4.044789049640475e-05, + "loss": 1.27, + "step": 110760 + }, + { + "epoch": 9.59, + "learning_rate": 4.044702417049294e-05, + "loss": 1.2881, + "step": 110770 + }, + { + "epoch": 9.59, + "learning_rate": 4.0446157844581134e-05, + "loss": 1.2399, + "step": 110780 + }, + { + "epoch": 9.59, + "learning_rate": 4.044529151866933e-05, + "loss": 1.3221, + "step": 110790 + }, + { + "epoch": 9.59, + "learning_rate": 4.0444425192757516e-05, + "loss": 1.1978, + "step": 110800 + }, + { + "epoch": 9.59, + "learning_rate": 4.044355886684571e-05, + "loss": 1.2629, + "step": 110810 + }, + { + "epoch": 9.59, + "learning_rate": 4.0442692540933905e-05, + "loss": 1.2841, + "step": 110820 + }, + { + "epoch": 9.59, + "learning_rate": 4.044182621502209e-05, + "loss": 1.3154, + "step": 110830 + }, + { + "epoch": 9.59, + "learning_rate": 4.044095988911029e-05, + "loss": 1.2378, + "step": 110840 + }, + { + "epoch": 9.59, + "learning_rate": 4.0440093563198475e-05, + "loss": 1.2819, + "step": 110850 + }, + { + "epoch": 9.6, + "learning_rate": 4.043922723728667e-05, + "loss": 1.2728, + "step": 110860 + }, + { + "epoch": 9.6, + "learning_rate": 4.0438360911374864e-05, + "loss": 1.2759, + "step": 110870 + }, + { + "epoch": 9.6, + "learning_rate": 4.043749458546305e-05, + "loss": 1.355, + "step": 110880 + }, + { + "epoch": 9.6, + "learning_rate": 4.0436628259551246e-05, + "loss": 1.2283, + "step": 110890 + }, + { + "epoch": 9.6, + "learning_rate": 4.043576193363944e-05, + "loss": 1.2641, + "step": 110900 + }, + { + "epoch": 9.6, + "learning_rate": 4.043489560772763e-05, + "loss": 1.223, + "step": 110910 + }, + { + "epoch": 9.6, + "learning_rate": 4.043402928181582e-05, + "loss": 1.2536, + "step": 110920 + }, + { + "epoch": 9.6, + "learning_rate": 4.043316295590402e-05, + "loss": 1.2692, + "step": 110930 + }, + { + "epoch": 9.6, + "learning_rate": 4.0432296629992205e-05, + "loss": 1.2999, + "step": 110940 + }, + { + "epoch": 9.6, + "learning_rate": 4.04314303040804e-05, + "loss": 1.2463, + "step": 110950 + }, + { + "epoch": 9.6, + "learning_rate": 4.043056397816859e-05, + "loss": 1.2863, + "step": 110960 + }, + { + "epoch": 9.61, + "learning_rate": 4.042969765225678e-05, + "loss": 1.2785, + "step": 110970 + }, + { + "epoch": 9.61, + "learning_rate": 4.0428831326344976e-05, + "loss": 1.2396, + "step": 110980 + }, + { + "epoch": 9.61, + "learning_rate": 4.042796500043316e-05, + "loss": 1.286, + "step": 110990 + }, + { + "epoch": 9.61, + "learning_rate": 4.042709867452136e-05, + "loss": 1.2624, + "step": 111000 + }, + { + "epoch": 9.61, + "learning_rate": 4.042623234860955e-05, + "loss": 1.2818, + "step": 111010 + }, + { + "epoch": 9.61, + "learning_rate": 4.042536602269774e-05, + "loss": 1.2745, + "step": 111020 + }, + { + "epoch": 9.61, + "learning_rate": 4.0424499696785934e-05, + "loss": 1.3453, + "step": 111030 + }, + { + "epoch": 9.61, + "learning_rate": 4.042363337087413e-05, + "loss": 1.217, + "step": 111040 + }, + { + "epoch": 9.61, + "learning_rate": 4.0422767044962316e-05, + "loss": 1.2983, + "step": 111050 + }, + { + "epoch": 9.61, + "learning_rate": 4.042190071905051e-05, + "loss": 1.2743, + "step": 111060 + }, + { + "epoch": 9.61, + "learning_rate": 4.04210343931387e-05, + "loss": 1.2519, + "step": 111070 + }, + { + "epoch": 9.61, + "learning_rate": 4.042016806722689e-05, + "loss": 1.2409, + "step": 111080 + }, + { + "epoch": 9.62, + "learning_rate": 4.041930174131509e-05, + "loss": 1.3002, + "step": 111090 + }, + { + "epoch": 9.62, + "learning_rate": 4.0418435415403275e-05, + "loss": 1.2104, + "step": 111100 + }, + { + "epoch": 9.62, + "learning_rate": 4.041756908949147e-05, + "loss": 1.222, + "step": 111110 + }, + { + "epoch": 9.62, + "learning_rate": 4.0416702763579664e-05, + "loss": 1.2504, + "step": 111120 + }, + { + "epoch": 9.62, + "learning_rate": 4.041583643766785e-05, + "loss": 1.2497, + "step": 111130 + }, + { + "epoch": 9.62, + "learning_rate": 4.0414970111756046e-05, + "loss": 1.3246, + "step": 111140 + }, + { + "epoch": 9.62, + "learning_rate": 4.041410378584424e-05, + "loss": 1.3501, + "step": 111150 + }, + { + "epoch": 9.62, + "learning_rate": 4.041323745993243e-05, + "loss": 1.2529, + "step": 111160 + }, + { + "epoch": 9.62, + "learning_rate": 4.041237113402062e-05, + "loss": 1.2394, + "step": 111170 + }, + { + "epoch": 9.62, + "learning_rate": 4.041150480810881e-05, + "loss": 1.2415, + "step": 111180 + }, + { + "epoch": 9.62, + "learning_rate": 4.0410638482197005e-05, + "loss": 1.2152, + "step": 111190 + }, + { + "epoch": 9.63, + "learning_rate": 4.04097721562852e-05, + "loss": 1.3315, + "step": 111200 + }, + { + "epoch": 9.63, + "learning_rate": 4.040890583037339e-05, + "loss": 1.2654, + "step": 111210 + }, + { + "epoch": 9.63, + "learning_rate": 4.040803950446158e-05, + "loss": 1.2465, + "step": 111220 + }, + { + "epoch": 9.63, + "learning_rate": 4.0407173178549776e-05, + "loss": 1.2934, + "step": 111230 + }, + { + "epoch": 9.63, + "learning_rate": 4.0406306852637964e-05, + "loss": 1.2077, + "step": 111240 + }, + { + "epoch": 9.63, + "learning_rate": 4.040544052672616e-05, + "loss": 1.2302, + "step": 111250 + }, + { + "epoch": 9.63, + "learning_rate": 4.0404574200814346e-05, + "loss": 1.2941, + "step": 111260 + }, + { + "epoch": 9.63, + "learning_rate": 4.040370787490254e-05, + "loss": 1.2561, + "step": 111270 + }, + { + "epoch": 9.63, + "learning_rate": 4.0402841548990735e-05, + "loss": 1.2968, + "step": 111280 + }, + { + "epoch": 9.63, + "learning_rate": 4.040197522307892e-05, + "loss": 1.2808, + "step": 111290 + }, + { + "epoch": 9.63, + "learning_rate": 4.040110889716712e-05, + "loss": 1.2797, + "step": 111300 + }, + { + "epoch": 9.63, + "learning_rate": 4.040024257125531e-05, + "loss": 1.2931, + "step": 111310 + }, + { + "epoch": 9.64, + "learning_rate": 4.03993762453435e-05, + "loss": 1.2613, + "step": 111320 + }, + { + "epoch": 9.64, + "learning_rate": 4.039850991943169e-05, + "loss": 1.2478, + "step": 111330 + }, + { + "epoch": 9.64, + "learning_rate": 4.039764359351989e-05, + "loss": 1.2926, + "step": 111340 + }, + { + "epoch": 9.64, + "learning_rate": 4.0396777267608075e-05, + "loss": 1.2861, + "step": 111350 + }, + { + "epoch": 9.64, + "learning_rate": 4.039591094169627e-05, + "loss": 1.2884, + "step": 111360 + }, + { + "epoch": 9.64, + "learning_rate": 4.039504461578446e-05, + "loss": 1.2948, + "step": 111370 + }, + { + "epoch": 9.64, + "learning_rate": 4.039417828987265e-05, + "loss": 1.2757, + "step": 111380 + }, + { + "epoch": 9.64, + "learning_rate": 4.0393311963960847e-05, + "loss": 1.2609, + "step": 111390 + }, + { + "epoch": 9.64, + "learning_rate": 4.0392445638049034e-05, + "loss": 1.2945, + "step": 111400 + }, + { + "epoch": 9.64, + "learning_rate": 4.039157931213723e-05, + "loss": 1.3034, + "step": 111410 + }, + { + "epoch": 9.64, + "learning_rate": 4.039071298622542e-05, + "loss": 1.2271, + "step": 111420 + }, + { + "epoch": 9.64, + "learning_rate": 4.038984666031361e-05, + "loss": 1.3023, + "step": 111430 + }, + { + "epoch": 9.65, + "learning_rate": 4.0388980334401805e-05, + "loss": 1.2908, + "step": 111440 + }, + { + "epoch": 9.65, + "learning_rate": 4.038811400849e-05, + "loss": 1.2296, + "step": 111450 + }, + { + "epoch": 9.65, + "learning_rate": 4.038724768257819e-05, + "loss": 1.3006, + "step": 111460 + }, + { + "epoch": 9.65, + "learning_rate": 4.038638135666638e-05, + "loss": 1.2876, + "step": 111470 + }, + { + "epoch": 9.65, + "learning_rate": 4.038551503075457e-05, + "loss": 1.2894, + "step": 111480 + }, + { + "epoch": 9.65, + "learning_rate": 4.0384648704842764e-05, + "loss": 1.3071, + "step": 111490 + }, + { + "epoch": 9.65, + "learning_rate": 4.038378237893096e-05, + "loss": 1.2626, + "step": 111500 + }, + { + "epoch": 9.65, + "learning_rate": 4.0382916053019146e-05, + "loss": 1.2769, + "step": 111510 + }, + { + "epoch": 9.65, + "learning_rate": 4.038204972710734e-05, + "loss": 1.3438, + "step": 111520 + }, + { + "epoch": 9.65, + "learning_rate": 4.0381183401195535e-05, + "loss": 1.26, + "step": 111530 + }, + { + "epoch": 9.65, + "learning_rate": 4.038031707528372e-05, + "loss": 1.3028, + "step": 111540 + }, + { + "epoch": 9.66, + "learning_rate": 4.037945074937192e-05, + "loss": 1.2497, + "step": 111550 + }, + { + "epoch": 9.66, + "learning_rate": 4.037858442346011e-05, + "loss": 1.3109, + "step": 111560 + }, + { + "epoch": 9.66, + "learning_rate": 4.03777180975483e-05, + "loss": 1.2728, + "step": 111570 + }, + { + "epoch": 9.66, + "learning_rate": 4.0376851771636494e-05, + "loss": 1.2635, + "step": 111580 + }, + { + "epoch": 9.66, + "learning_rate": 4.037598544572468e-05, + "loss": 1.2244, + "step": 111590 + }, + { + "epoch": 9.66, + "learning_rate": 4.0375119119812876e-05, + "loss": 1.2749, + "step": 111600 + }, + { + "epoch": 9.66, + "learning_rate": 4.037425279390107e-05, + "loss": 1.2908, + "step": 111610 + }, + { + "epoch": 9.66, + "learning_rate": 4.037338646798926e-05, + "loss": 1.2854, + "step": 111620 + }, + { + "epoch": 9.66, + "learning_rate": 4.037252014207745e-05, + "loss": 1.2443, + "step": 111630 + }, + { + "epoch": 9.66, + "learning_rate": 4.037165381616565e-05, + "loss": 1.307, + "step": 111640 + }, + { + "epoch": 9.66, + "learning_rate": 4.0370787490253835e-05, + "loss": 1.315, + "step": 111650 + }, + { + "epoch": 9.66, + "learning_rate": 4.036992116434203e-05, + "loss": 1.2848, + "step": 111660 + }, + { + "epoch": 9.67, + "learning_rate": 4.0369054838430223e-05, + "loss": 1.2747, + "step": 111670 + }, + { + "epoch": 9.67, + "learning_rate": 4.036818851251841e-05, + "loss": 1.2501, + "step": 111680 + }, + { + "epoch": 9.67, + "learning_rate": 4.0367322186606606e-05, + "loss": 1.2711, + "step": 111690 + }, + { + "epoch": 9.67, + "learning_rate": 4.036645586069479e-05, + "loss": 1.2847, + "step": 111700 + }, + { + "epoch": 9.67, + "learning_rate": 4.036558953478299e-05, + "loss": 1.2716, + "step": 111710 + }, + { + "epoch": 9.67, + "learning_rate": 4.036472320887118e-05, + "loss": 1.3133, + "step": 111720 + }, + { + "epoch": 9.67, + "learning_rate": 4.036385688295937e-05, + "loss": 1.2734, + "step": 111730 + }, + { + "epoch": 9.67, + "learning_rate": 4.0362990557047564e-05, + "loss": 1.2487, + "step": 111740 + }, + { + "epoch": 9.67, + "learning_rate": 4.036212423113576e-05, + "loss": 1.2141, + "step": 111750 + }, + { + "epoch": 9.67, + "learning_rate": 4.0361257905223946e-05, + "loss": 1.2972, + "step": 111760 + }, + { + "epoch": 9.67, + "learning_rate": 4.036039157931214e-05, + "loss": 1.1657, + "step": 111770 + }, + { + "epoch": 9.68, + "learning_rate": 4.0359525253400335e-05, + "loss": 1.232, + "step": 111780 + }, + { + "epoch": 9.68, + "learning_rate": 4.035865892748852e-05, + "loss": 1.346, + "step": 111790 + }, + { + "epoch": 9.68, + "learning_rate": 4.035779260157672e-05, + "loss": 1.249, + "step": 111800 + }, + { + "epoch": 9.68, + "learning_rate": 4.0356926275664905e-05, + "loss": 1.2848, + "step": 111810 + }, + { + "epoch": 9.68, + "learning_rate": 4.03560599497531e-05, + "loss": 1.2869, + "step": 111820 + }, + { + "epoch": 9.68, + "learning_rate": 4.0355193623841294e-05, + "loss": 1.2316, + "step": 111830 + }, + { + "epoch": 9.68, + "learning_rate": 4.035432729792948e-05, + "loss": 1.2326, + "step": 111840 + }, + { + "epoch": 9.68, + "learning_rate": 4.0353460972017676e-05, + "loss": 1.2387, + "step": 111850 + }, + { + "epoch": 9.68, + "learning_rate": 4.035259464610587e-05, + "loss": 1.2954, + "step": 111860 + }, + { + "epoch": 9.68, + "learning_rate": 4.035172832019406e-05, + "loss": 1.2436, + "step": 111870 + }, + { + "epoch": 9.68, + "learning_rate": 4.035086199428225e-05, + "loss": 1.2683, + "step": 111880 + }, + { + "epoch": 9.68, + "learning_rate": 4.034999566837045e-05, + "loss": 1.3132, + "step": 111890 + }, + { + "epoch": 9.69, + "learning_rate": 4.0349129342458635e-05, + "loss": 1.2465, + "step": 111900 + }, + { + "epoch": 9.69, + "learning_rate": 4.034826301654683e-05, + "loss": 1.3303, + "step": 111910 + }, + { + "epoch": 9.69, + "learning_rate": 4.034739669063502e-05, + "loss": 1.2758, + "step": 111920 + }, + { + "epoch": 9.69, + "learning_rate": 4.034653036472321e-05, + "loss": 1.2088, + "step": 111930 + }, + { + "epoch": 9.69, + "learning_rate": 4.0345664038811406e-05, + "loss": 1.2665, + "step": 111940 + }, + { + "epoch": 9.69, + "learning_rate": 4.0344797712899594e-05, + "loss": 1.2881, + "step": 111950 + }, + { + "epoch": 9.69, + "learning_rate": 4.034393138698779e-05, + "loss": 1.2963, + "step": 111960 + }, + { + "epoch": 9.69, + "learning_rate": 4.034306506107598e-05, + "loss": 1.2213, + "step": 111970 + }, + { + "epoch": 9.69, + "learning_rate": 4.034219873516417e-05, + "loss": 1.2779, + "step": 111980 + }, + { + "epoch": 9.69, + "learning_rate": 4.0341332409252365e-05, + "loss": 1.2149, + "step": 111990 + }, + { + "epoch": 9.69, + "learning_rate": 4.034046608334055e-05, + "loss": 1.3292, + "step": 112000 + }, + { + "epoch": 9.7, + "learning_rate": 4.033959975742875e-05, + "loss": 1.2907, + "step": 112010 + }, + { + "epoch": 9.7, + "learning_rate": 4.033873343151694e-05, + "loss": 1.1937, + "step": 112020 + }, + { + "epoch": 9.7, + "learning_rate": 4.033786710560513e-05, + "loss": 1.2997, + "step": 112030 + }, + { + "epoch": 9.7, + "learning_rate": 4.033700077969332e-05, + "loss": 1.2758, + "step": 112040 + }, + { + "epoch": 9.7, + "learning_rate": 4.033613445378152e-05, + "loss": 1.2477, + "step": 112050 + }, + { + "epoch": 9.7, + "learning_rate": 4.0335268127869705e-05, + "loss": 1.238, + "step": 112060 + }, + { + "epoch": 9.7, + "learning_rate": 4.03344018019579e-05, + "loss": 1.2675, + "step": 112070 + }, + { + "epoch": 9.7, + "learning_rate": 4.0333535476046094e-05, + "loss": 1.3453, + "step": 112080 + }, + { + "epoch": 9.7, + "learning_rate": 4.033266915013428e-05, + "loss": 1.2822, + "step": 112090 + }, + { + "epoch": 9.7, + "learning_rate": 4.0331802824222476e-05, + "loss": 1.2816, + "step": 112100 + }, + { + "epoch": 9.7, + "learning_rate": 4.0330936498310664e-05, + "loss": 1.2735, + "step": 112110 + }, + { + "epoch": 9.7, + "learning_rate": 4.033007017239886e-05, + "loss": 1.3067, + "step": 112120 + }, + { + "epoch": 9.71, + "learning_rate": 4.032920384648705e-05, + "loss": 1.2553, + "step": 112130 + }, + { + "epoch": 9.71, + "learning_rate": 4.032833752057524e-05, + "loss": 1.2892, + "step": 112140 + }, + { + "epoch": 9.71, + "learning_rate": 4.0327471194663435e-05, + "loss": 1.2097, + "step": 112150 + }, + { + "epoch": 9.71, + "learning_rate": 4.032660486875163e-05, + "loss": 1.2857, + "step": 112160 + }, + { + "epoch": 9.71, + "learning_rate": 4.032573854283982e-05, + "loss": 1.2216, + "step": 112170 + }, + { + "epoch": 9.71, + "learning_rate": 4.032487221692801e-05, + "loss": 1.304, + "step": 112180 + }, + { + "epoch": 9.71, + "learning_rate": 4.0324005891016206e-05, + "loss": 1.2912, + "step": 112190 + }, + { + "epoch": 9.71, + "learning_rate": 4.0323139565104394e-05, + "loss": 1.3107, + "step": 112200 + }, + { + "epoch": 9.71, + "learning_rate": 4.032227323919259e-05, + "loss": 1.2801, + "step": 112210 + }, + { + "epoch": 9.71, + "learning_rate": 4.0321406913280776e-05, + "loss": 1.2391, + "step": 112220 + }, + { + "epoch": 9.71, + "learning_rate": 4.032054058736897e-05, + "loss": 1.2333, + "step": 112230 + }, + { + "epoch": 9.72, + "learning_rate": 4.0319674261457165e-05, + "loss": 1.2261, + "step": 112240 + }, + { + "epoch": 9.72, + "learning_rate": 4.031880793554535e-05, + "loss": 1.2825, + "step": 112250 + }, + { + "epoch": 9.72, + "learning_rate": 4.031794160963355e-05, + "loss": 1.2636, + "step": 112260 + }, + { + "epoch": 9.72, + "learning_rate": 4.031707528372174e-05, + "loss": 1.3049, + "step": 112270 + }, + { + "epoch": 9.72, + "learning_rate": 4.031620895780993e-05, + "loss": 1.3317, + "step": 112280 + }, + { + "epoch": 9.72, + "learning_rate": 4.0315342631898124e-05, + "loss": 1.2458, + "step": 112290 + }, + { + "epoch": 9.72, + "learning_rate": 4.031447630598632e-05, + "loss": 1.2513, + "step": 112300 + }, + { + "epoch": 9.72, + "learning_rate": 4.0313609980074506e-05, + "loss": 1.2765, + "step": 112310 + }, + { + "epoch": 9.72, + "learning_rate": 4.03127436541627e-05, + "loss": 1.222, + "step": 112320 + }, + { + "epoch": 9.72, + "learning_rate": 4.031187732825089e-05, + "loss": 1.2536, + "step": 112330 + }, + { + "epoch": 9.72, + "learning_rate": 4.031101100233908e-05, + "loss": 1.2725, + "step": 112340 + }, + { + "epoch": 9.72, + "learning_rate": 4.031014467642728e-05, + "loss": 1.2631, + "step": 112350 + }, + { + "epoch": 9.73, + "learning_rate": 4.0309278350515465e-05, + "loss": 1.252, + "step": 112360 + }, + { + "epoch": 9.73, + "learning_rate": 4.030841202460366e-05, + "loss": 1.2567, + "step": 112370 + }, + { + "epoch": 9.73, + "learning_rate": 4.0307545698691853e-05, + "loss": 1.2706, + "step": 112380 + }, + { + "epoch": 9.73, + "learning_rate": 4.030667937278004e-05, + "loss": 1.2298, + "step": 112390 + }, + { + "epoch": 9.73, + "learning_rate": 4.0305813046868236e-05, + "loss": 1.2918, + "step": 112400 + }, + { + "epoch": 9.73, + "learning_rate": 4.030494672095643e-05, + "loss": 1.2794, + "step": 112410 + }, + { + "epoch": 9.73, + "learning_rate": 4.030408039504462e-05, + "loss": 1.3806, + "step": 112420 + }, + { + "epoch": 9.73, + "learning_rate": 4.030321406913281e-05, + "loss": 1.2884, + "step": 112430 + }, + { + "epoch": 9.73, + "learning_rate": 4.0302347743221e-05, + "loss": 1.2693, + "step": 112440 + }, + { + "epoch": 9.73, + "learning_rate": 4.0301481417309194e-05, + "loss": 1.2302, + "step": 112450 + }, + { + "epoch": 9.73, + "learning_rate": 4.030061509139739e-05, + "loss": 1.2943, + "step": 112460 + }, + { + "epoch": 9.73, + "learning_rate": 4.0299748765485576e-05, + "loss": 1.2666, + "step": 112470 + }, + { + "epoch": 9.74, + "learning_rate": 4.029888243957377e-05, + "loss": 1.2901, + "step": 112480 + }, + { + "epoch": 9.74, + "learning_rate": 4.0298016113661965e-05, + "loss": 1.1808, + "step": 112490 + }, + { + "epoch": 9.74, + "learning_rate": 4.029714978775015e-05, + "loss": 1.2607, + "step": 112500 + }, + { + "epoch": 9.74, + "learning_rate": 4.029628346183835e-05, + "loss": 1.2685, + "step": 112510 + }, + { + "epoch": 9.74, + "learning_rate": 4.029541713592654e-05, + "loss": 1.2978, + "step": 112520 + }, + { + "epoch": 9.74, + "learning_rate": 4.029455081001473e-05, + "loss": 1.2875, + "step": 112530 + }, + { + "epoch": 9.74, + "learning_rate": 4.0293684484102924e-05, + "loss": 1.2012, + "step": 112540 + }, + { + "epoch": 9.74, + "learning_rate": 4.029281815819111e-05, + "loss": 1.2545, + "step": 112550 + }, + { + "epoch": 9.74, + "learning_rate": 4.0291951832279306e-05, + "loss": 1.2853, + "step": 112560 + }, + { + "epoch": 9.74, + "learning_rate": 4.02910855063675e-05, + "loss": 1.253, + "step": 112570 + }, + { + "epoch": 9.74, + "learning_rate": 4.029021918045569e-05, + "loss": 1.2687, + "step": 112580 + }, + { + "epoch": 9.75, + "learning_rate": 4.028935285454388e-05, + "loss": 1.2341, + "step": 112590 + }, + { + "epoch": 9.75, + "learning_rate": 4.028848652863208e-05, + "loss": 1.3109, + "step": 112600 + }, + { + "epoch": 9.75, + "learning_rate": 4.0287620202720265e-05, + "loss": 1.2442, + "step": 112610 + }, + { + "epoch": 9.75, + "learning_rate": 4.028675387680846e-05, + "loss": 1.2886, + "step": 112620 + }, + { + "epoch": 9.75, + "learning_rate": 4.0285887550896654e-05, + "loss": 1.2474, + "step": 112630 + }, + { + "epoch": 9.75, + "learning_rate": 4.028502122498484e-05, + "loss": 1.2753, + "step": 112640 + }, + { + "epoch": 9.75, + "learning_rate": 4.0284154899073036e-05, + "loss": 1.2081, + "step": 112650 + }, + { + "epoch": 9.75, + "learning_rate": 4.0283288573161224e-05, + "loss": 1.183, + "step": 112660 + }, + { + "epoch": 9.75, + "learning_rate": 4.028242224724942e-05, + "loss": 1.2524, + "step": 112670 + }, + { + "epoch": 9.75, + "learning_rate": 4.028155592133761e-05, + "loss": 1.2022, + "step": 112680 + }, + { + "epoch": 9.75, + "learning_rate": 4.02806895954258e-05, + "loss": 1.2844, + "step": 112690 + }, + { + "epoch": 9.75, + "learning_rate": 4.0279823269513995e-05, + "loss": 1.2392, + "step": 112700 + }, + { + "epoch": 9.76, + "learning_rate": 4.027895694360219e-05, + "loss": 1.3178, + "step": 112710 + }, + { + "epoch": 9.76, + "learning_rate": 4.027809061769038e-05, + "loss": 1.2517, + "step": 112720 + }, + { + "epoch": 9.76, + "learning_rate": 4.027722429177857e-05, + "loss": 1.3081, + "step": 112730 + }, + { + "epoch": 9.76, + "learning_rate": 4.0276357965866766e-05, + "loss": 1.2348, + "step": 112740 + }, + { + "epoch": 9.76, + "learning_rate": 4.027549163995495e-05, + "loss": 1.2548, + "step": 112750 + }, + { + "epoch": 9.76, + "learning_rate": 4.027462531404315e-05, + "loss": 1.344, + "step": 112760 + }, + { + "epoch": 9.76, + "learning_rate": 4.0273758988131335e-05, + "loss": 1.2419, + "step": 112770 + }, + { + "epoch": 9.76, + "learning_rate": 4.027289266221953e-05, + "loss": 1.2647, + "step": 112780 + }, + { + "epoch": 9.76, + "learning_rate": 4.0272026336307724e-05, + "loss": 1.2886, + "step": 112790 + }, + { + "epoch": 9.76, + "learning_rate": 4.027116001039591e-05, + "loss": 1.2431, + "step": 112800 + }, + { + "epoch": 9.76, + "learning_rate": 4.0270293684484106e-05, + "loss": 1.2533, + "step": 112810 + }, + { + "epoch": 9.77, + "learning_rate": 4.02694273585723e-05, + "loss": 1.2655, + "step": 112820 + }, + { + "epoch": 9.77, + "learning_rate": 4.026856103266049e-05, + "loss": 1.2609, + "step": 112830 + }, + { + "epoch": 9.77, + "learning_rate": 4.026769470674868e-05, + "loss": 1.2556, + "step": 112840 + }, + { + "epoch": 9.77, + "learning_rate": 4.026682838083687e-05, + "loss": 1.2789, + "step": 112850 + }, + { + "epoch": 9.77, + "learning_rate": 4.0265962054925065e-05, + "loss": 1.2448, + "step": 112860 + }, + { + "epoch": 9.77, + "learning_rate": 4.026509572901326e-05, + "loss": 1.2696, + "step": 112870 + }, + { + "epoch": 9.77, + "learning_rate": 4.026422940310145e-05, + "loss": 1.3432, + "step": 112880 + }, + { + "epoch": 9.77, + "learning_rate": 4.026336307718964e-05, + "loss": 1.2485, + "step": 112890 + }, + { + "epoch": 9.77, + "learning_rate": 4.0262496751277836e-05, + "loss": 1.2723, + "step": 112900 + }, + { + "epoch": 9.77, + "learning_rate": 4.0261630425366024e-05, + "loss": 1.1905, + "step": 112910 + }, + { + "epoch": 9.77, + "learning_rate": 4.026076409945422e-05, + "loss": 1.2474, + "step": 112920 + }, + { + "epoch": 9.77, + "learning_rate": 4.025989777354241e-05, + "loss": 1.2693, + "step": 112930 + }, + { + "epoch": 9.78, + "learning_rate": 4.02590314476306e-05, + "loss": 1.2318, + "step": 112940 + }, + { + "epoch": 9.78, + "learning_rate": 4.0258165121718795e-05, + "loss": 1.1776, + "step": 112950 + }, + { + "epoch": 9.78, + "learning_rate": 4.025729879580698e-05, + "loss": 1.2757, + "step": 112960 + }, + { + "epoch": 9.78, + "learning_rate": 4.025643246989518e-05, + "loss": 1.3452, + "step": 112970 + }, + { + "epoch": 9.78, + "learning_rate": 4.025556614398337e-05, + "loss": 1.2947, + "step": 112980 + }, + { + "epoch": 9.78, + "learning_rate": 4.025469981807156e-05, + "loss": 1.2695, + "step": 112990 + }, + { + "epoch": 9.78, + "learning_rate": 4.0253833492159754e-05, + "loss": 1.2746, + "step": 113000 + }, + { + "epoch": 9.78, + "learning_rate": 4.025296716624795e-05, + "loss": 1.3292, + "step": 113010 + }, + { + "epoch": 9.78, + "learning_rate": 4.0252100840336136e-05, + "loss": 1.2556, + "step": 113020 + }, + { + "epoch": 9.78, + "learning_rate": 4.025123451442433e-05, + "loss": 1.3128, + "step": 113030 + }, + { + "epoch": 9.78, + "learning_rate": 4.0250368188512525e-05, + "loss": 1.2311, + "step": 113040 + }, + { + "epoch": 9.79, + "learning_rate": 4.024950186260071e-05, + "loss": 1.2829, + "step": 113050 + }, + { + "epoch": 9.79, + "learning_rate": 4.024863553668891e-05, + "loss": 1.2109, + "step": 113060 + }, + { + "epoch": 9.79, + "learning_rate": 4.0247769210777094e-05, + "loss": 1.1952, + "step": 113070 + }, + { + "epoch": 9.79, + "learning_rate": 4.024690288486529e-05, + "loss": 1.2866, + "step": 113080 + }, + { + "epoch": 9.79, + "learning_rate": 4.024603655895348e-05, + "loss": 1.2874, + "step": 113090 + }, + { + "epoch": 9.79, + "learning_rate": 4.024517023304167e-05, + "loss": 1.2466, + "step": 113100 + }, + { + "epoch": 9.79, + "learning_rate": 4.0244303907129866e-05, + "loss": 1.2581, + "step": 113110 + }, + { + "epoch": 9.79, + "learning_rate": 4.024343758121806e-05, + "loss": 1.2896, + "step": 113120 + }, + { + "epoch": 9.79, + "learning_rate": 4.024257125530625e-05, + "loss": 1.2818, + "step": 113130 + }, + { + "epoch": 9.79, + "learning_rate": 4.024170492939444e-05, + "loss": 1.363, + "step": 113140 + }, + { + "epoch": 9.79, + "learning_rate": 4.0240838603482637e-05, + "loss": 1.2607, + "step": 113150 + }, + { + "epoch": 9.79, + "learning_rate": 4.0239972277570824e-05, + "loss": 1.2549, + "step": 113160 + }, + { + "epoch": 9.8, + "learning_rate": 4.023910595165902e-05, + "loss": 1.242, + "step": 113170 + }, + { + "epoch": 9.8, + "learning_rate": 4.0238239625747206e-05, + "loss": 1.1357, + "step": 113180 + }, + { + "epoch": 9.8, + "learning_rate": 4.02373732998354e-05, + "loss": 1.2838, + "step": 113190 + }, + { + "epoch": 9.8, + "learning_rate": 4.0236506973923595e-05, + "loss": 1.2083, + "step": 113200 + }, + { + "epoch": 9.8, + "learning_rate": 4.023564064801178e-05, + "loss": 1.2754, + "step": 113210 + }, + { + "epoch": 9.8, + "learning_rate": 4.023477432209998e-05, + "loss": 1.2918, + "step": 113220 + }, + { + "epoch": 9.8, + "learning_rate": 4.023390799618817e-05, + "loss": 1.2405, + "step": 113230 + }, + { + "epoch": 9.8, + "learning_rate": 4.023304167027636e-05, + "loss": 1.2781, + "step": 113240 + }, + { + "epoch": 9.8, + "learning_rate": 4.0232175344364554e-05, + "loss": 1.2588, + "step": 113250 + }, + { + "epoch": 9.8, + "learning_rate": 4.023130901845275e-05, + "loss": 1.3127, + "step": 113260 + }, + { + "epoch": 9.8, + "learning_rate": 4.0230442692540936e-05, + "loss": 1.2694, + "step": 113270 + }, + { + "epoch": 9.81, + "learning_rate": 4.022957636662913e-05, + "loss": 1.2915, + "step": 113280 + }, + { + "epoch": 9.81, + "learning_rate": 4.022871004071732e-05, + "loss": 1.2385, + "step": 113290 + }, + { + "epoch": 9.81, + "learning_rate": 4.022784371480551e-05, + "loss": 1.2001, + "step": 113300 + }, + { + "epoch": 9.81, + "learning_rate": 4.022697738889371e-05, + "loss": 1.2941, + "step": 113310 + }, + { + "epoch": 9.81, + "learning_rate": 4.0226111062981895e-05, + "loss": 1.2558, + "step": 113320 + }, + { + "epoch": 9.81, + "learning_rate": 4.022524473707009e-05, + "loss": 1.1924, + "step": 113330 + }, + { + "epoch": 9.81, + "learning_rate": 4.0224378411158284e-05, + "loss": 1.2729, + "step": 113340 + }, + { + "epoch": 9.81, + "learning_rate": 4.022351208524647e-05, + "loss": 1.2622, + "step": 113350 + }, + { + "epoch": 9.81, + "learning_rate": 4.0222645759334666e-05, + "loss": 1.2739, + "step": 113360 + }, + { + "epoch": 9.81, + "learning_rate": 4.022177943342286e-05, + "loss": 1.2525, + "step": 113370 + }, + { + "epoch": 9.81, + "learning_rate": 4.022091310751105e-05, + "loss": 1.3219, + "step": 113380 + }, + { + "epoch": 9.81, + "learning_rate": 4.022004678159924e-05, + "loss": 1.3196, + "step": 113390 + }, + { + "epoch": 9.82, + "learning_rate": 4.021918045568743e-05, + "loss": 1.2908, + "step": 113400 + }, + { + "epoch": 9.82, + "learning_rate": 4.0218314129775625e-05, + "loss": 1.2646, + "step": 113410 + }, + { + "epoch": 9.82, + "learning_rate": 4.021744780386382e-05, + "loss": 1.2159, + "step": 113420 + }, + { + "epoch": 9.82, + "learning_rate": 4.021658147795201e-05, + "loss": 1.3133, + "step": 113430 + }, + { + "epoch": 9.82, + "learning_rate": 4.02157151520402e-05, + "loss": 1.226, + "step": 113440 + }, + { + "epoch": 9.82, + "learning_rate": 4.0214848826128396e-05, + "loss": 1.2987, + "step": 113450 + }, + { + "epoch": 9.82, + "learning_rate": 4.021398250021658e-05, + "loss": 1.2605, + "step": 113460 + }, + { + "epoch": 9.82, + "learning_rate": 4.021311617430478e-05, + "loss": 1.2734, + "step": 113470 + }, + { + "epoch": 9.82, + "learning_rate": 4.021224984839297e-05, + "loss": 1.2966, + "step": 113480 + }, + { + "epoch": 9.82, + "learning_rate": 4.021138352248116e-05, + "loss": 1.246, + "step": 113490 + }, + { + "epoch": 9.82, + "learning_rate": 4.0210517196569354e-05, + "loss": 1.2925, + "step": 113500 + }, + { + "epoch": 9.82, + "learning_rate": 4.020965087065754e-05, + "loss": 1.2776, + "step": 113510 + }, + { + "epoch": 9.83, + "learning_rate": 4.0208784544745736e-05, + "loss": 1.2248, + "step": 113520 + }, + { + "epoch": 9.83, + "learning_rate": 4.020791821883393e-05, + "loss": 1.2915, + "step": 113530 + }, + { + "epoch": 9.83, + "learning_rate": 4.020705189292212e-05, + "loss": 1.2435, + "step": 113540 + }, + { + "epoch": 9.83, + "learning_rate": 4.020618556701031e-05, + "loss": 1.2345, + "step": 113550 + }, + { + "epoch": 9.83, + "learning_rate": 4.020531924109851e-05, + "loss": 1.2788, + "step": 113560 + }, + { + "epoch": 9.83, + "learning_rate": 4.0204452915186695e-05, + "loss": 1.2652, + "step": 113570 + }, + { + "epoch": 9.83, + "learning_rate": 4.020358658927489e-05, + "loss": 1.2264, + "step": 113580 + }, + { + "epoch": 9.83, + "learning_rate": 4.020272026336308e-05, + "loss": 1.259, + "step": 113590 + }, + { + "epoch": 9.83, + "learning_rate": 4.020185393745127e-05, + "loss": 1.2592, + "step": 113600 + }, + { + "epoch": 9.83, + "learning_rate": 4.0200987611539466e-05, + "loss": 1.2293, + "step": 113610 + }, + { + "epoch": 9.83, + "learning_rate": 4.0200121285627654e-05, + "loss": 1.2668, + "step": 113620 + }, + { + "epoch": 9.84, + "learning_rate": 4.019925495971585e-05, + "loss": 1.2589, + "step": 113630 + }, + { + "epoch": 9.84, + "learning_rate": 4.019838863380404e-05, + "loss": 1.2904, + "step": 113640 + }, + { + "epoch": 9.84, + "learning_rate": 4.019752230789223e-05, + "loss": 1.3156, + "step": 113650 + }, + { + "epoch": 9.84, + "learning_rate": 4.0196655981980425e-05, + "loss": 1.2835, + "step": 113660 + }, + { + "epoch": 9.84, + "learning_rate": 4.019578965606862e-05, + "loss": 1.2907, + "step": 113670 + }, + { + "epoch": 9.84, + "learning_rate": 4.019492333015681e-05, + "loss": 1.304, + "step": 113680 + }, + { + "epoch": 9.84, + "learning_rate": 4.0194057004245e-05, + "loss": 1.2669, + "step": 113690 + }, + { + "epoch": 9.84, + "learning_rate": 4.019319067833319e-05, + "loss": 1.2823, + "step": 113700 + }, + { + "epoch": 9.84, + "learning_rate": 4.0192324352421384e-05, + "loss": 1.216, + "step": 113710 + }, + { + "epoch": 9.84, + "learning_rate": 4.019145802650958e-05, + "loss": 1.2515, + "step": 113720 + }, + { + "epoch": 9.84, + "learning_rate": 4.0190591700597766e-05, + "loss": 1.2288, + "step": 113730 + }, + { + "epoch": 9.84, + "learning_rate": 4.018972537468596e-05, + "loss": 1.2185, + "step": 113740 + }, + { + "epoch": 9.85, + "learning_rate": 4.0188859048774155e-05, + "loss": 1.3034, + "step": 113750 + }, + { + "epoch": 9.85, + "learning_rate": 4.018799272286234e-05, + "loss": 1.2757, + "step": 113760 + }, + { + "epoch": 9.85, + "learning_rate": 4.018712639695054e-05, + "loss": 1.2388, + "step": 113770 + }, + { + "epoch": 9.85, + "learning_rate": 4.018626007103873e-05, + "loss": 1.2813, + "step": 113780 + }, + { + "epoch": 9.85, + "learning_rate": 4.018539374512692e-05, + "loss": 1.2558, + "step": 113790 + }, + { + "epoch": 9.85, + "learning_rate": 4.018452741921511e-05, + "loss": 1.1764, + "step": 113800 + }, + { + "epoch": 9.85, + "learning_rate": 4.01836610933033e-05, + "loss": 1.265, + "step": 113810 + }, + { + "epoch": 9.85, + "learning_rate": 4.0182794767391495e-05, + "loss": 1.2543, + "step": 113820 + }, + { + "epoch": 9.85, + "learning_rate": 4.018192844147969e-05, + "loss": 1.1794, + "step": 113830 + }, + { + "epoch": 9.85, + "learning_rate": 4.018106211556788e-05, + "loss": 1.2801, + "step": 113840 + }, + { + "epoch": 9.85, + "learning_rate": 4.018019578965607e-05, + "loss": 1.2328, + "step": 113850 + }, + { + "epoch": 9.86, + "learning_rate": 4.0179329463744267e-05, + "loss": 1.2365, + "step": 113860 + }, + { + "epoch": 9.86, + "learning_rate": 4.0178463137832454e-05, + "loss": 1.2348, + "step": 113870 + }, + { + "epoch": 9.86, + "learning_rate": 4.017759681192065e-05, + "loss": 1.2505, + "step": 113880 + }, + { + "epoch": 9.86, + "learning_rate": 4.017673048600884e-05, + "loss": 1.1878, + "step": 113890 + }, + { + "epoch": 9.86, + "learning_rate": 4.017586416009703e-05, + "loss": 1.2279, + "step": 113900 + }, + { + "epoch": 9.86, + "learning_rate": 4.0174997834185225e-05, + "loss": 1.2333, + "step": 113910 + }, + { + "epoch": 9.86, + "learning_rate": 4.017413150827341e-05, + "loss": 1.2345, + "step": 113920 + }, + { + "epoch": 9.86, + "learning_rate": 4.017326518236161e-05, + "loss": 1.3052, + "step": 113930 + }, + { + "epoch": 9.86, + "learning_rate": 4.01723988564498e-05, + "loss": 1.201, + "step": 113940 + }, + { + "epoch": 9.86, + "learning_rate": 4.017153253053799e-05, + "loss": 1.26, + "step": 113950 + }, + { + "epoch": 9.86, + "learning_rate": 4.0170666204626184e-05, + "loss": 1.297, + "step": 113960 + }, + { + "epoch": 9.86, + "learning_rate": 4.016979987871438e-05, + "loss": 1.2302, + "step": 113970 + }, + { + "epoch": 9.87, + "learning_rate": 4.0168933552802566e-05, + "loss": 1.3248, + "step": 113980 + }, + { + "epoch": 9.87, + "learning_rate": 4.016806722689076e-05, + "loss": 1.2745, + "step": 113990 + }, + { + "epoch": 9.87, + "learning_rate": 4.0167200900978955e-05, + "loss": 1.2588, + "step": 114000 + }, + { + "epoch": 9.87, + "learning_rate": 4.016633457506714e-05, + "loss": 1.2271, + "step": 114010 + }, + { + "epoch": 9.87, + "learning_rate": 4.016546824915534e-05, + "loss": 1.2733, + "step": 114020 + }, + { + "epoch": 9.87, + "learning_rate": 4.0164601923243525e-05, + "loss": 1.2538, + "step": 114030 + }, + { + "epoch": 9.87, + "learning_rate": 4.016373559733172e-05, + "loss": 1.2872, + "step": 114040 + }, + { + "epoch": 9.87, + "learning_rate": 4.0162869271419914e-05, + "loss": 1.2827, + "step": 114050 + }, + { + "epoch": 9.87, + "learning_rate": 4.01620029455081e-05, + "loss": 1.1823, + "step": 114060 + }, + { + "epoch": 9.87, + "learning_rate": 4.0161136619596296e-05, + "loss": 1.3135, + "step": 114070 + }, + { + "epoch": 9.87, + "learning_rate": 4.016027029368449e-05, + "loss": 1.2774, + "step": 114080 + }, + { + "epoch": 9.88, + "learning_rate": 4.015940396777268e-05, + "loss": 1.2723, + "step": 114090 + }, + { + "epoch": 9.88, + "learning_rate": 4.015853764186087e-05, + "loss": 1.2826, + "step": 114100 + }, + { + "epoch": 9.88, + "learning_rate": 4.015767131594907e-05, + "loss": 1.2427, + "step": 114110 + }, + { + "epoch": 9.88, + "learning_rate": 4.0156804990037255e-05, + "loss": 1.2155, + "step": 114120 + }, + { + "epoch": 9.88, + "learning_rate": 4.015593866412545e-05, + "loss": 1.2768, + "step": 114130 + }, + { + "epoch": 9.88, + "learning_rate": 4.015507233821364e-05, + "loss": 1.241, + "step": 114140 + }, + { + "epoch": 9.88, + "learning_rate": 4.015420601230183e-05, + "loss": 1.2507, + "step": 114150 + }, + { + "epoch": 9.88, + "learning_rate": 4.0153339686390026e-05, + "loss": 1.2524, + "step": 114160 + }, + { + "epoch": 9.88, + "learning_rate": 4.015247336047821e-05, + "loss": 1.2178, + "step": 114170 + }, + { + "epoch": 9.88, + "learning_rate": 4.015160703456641e-05, + "loss": 1.2358, + "step": 114180 + }, + { + "epoch": 9.88, + "learning_rate": 4.01507407086546e-05, + "loss": 1.2977, + "step": 114190 + }, + { + "epoch": 9.88, + "learning_rate": 4.014987438274279e-05, + "loss": 1.2296, + "step": 114200 + }, + { + "epoch": 9.89, + "learning_rate": 4.0149008056830984e-05, + "loss": 1.2966, + "step": 114210 + }, + { + "epoch": 9.89, + "learning_rate": 4.014814173091918e-05, + "loss": 1.2728, + "step": 114220 + }, + { + "epoch": 9.89, + "learning_rate": 4.0147275405007366e-05, + "loss": 1.1613, + "step": 114230 + }, + { + "epoch": 9.89, + "learning_rate": 4.014640907909556e-05, + "loss": 1.2349, + "step": 114240 + }, + { + "epoch": 9.89, + "learning_rate": 4.014554275318375e-05, + "loss": 1.2301, + "step": 114250 + }, + { + "epoch": 9.89, + "learning_rate": 4.014467642727194e-05, + "loss": 1.2306, + "step": 114260 + }, + { + "epoch": 9.89, + "learning_rate": 4.014381010136014e-05, + "loss": 1.2916, + "step": 114270 + }, + { + "epoch": 9.89, + "learning_rate": 4.0142943775448325e-05, + "loss": 1.1786, + "step": 114280 + }, + { + "epoch": 9.89, + "learning_rate": 4.014207744953652e-05, + "loss": 1.2643, + "step": 114290 + }, + { + "epoch": 9.89, + "learning_rate": 4.0141211123624714e-05, + "loss": 1.2664, + "step": 114300 + }, + { + "epoch": 9.89, + "learning_rate": 4.01403447977129e-05, + "loss": 1.1751, + "step": 114310 + }, + { + "epoch": 9.9, + "learning_rate": 4.0139478471801096e-05, + "loss": 1.203, + "step": 114320 + }, + { + "epoch": 9.9, + "learning_rate": 4.0138612145889284e-05, + "loss": 1.2903, + "step": 114330 + }, + { + "epoch": 9.9, + "learning_rate": 4.013774581997748e-05, + "loss": 1.226, + "step": 114340 + }, + { + "epoch": 9.9, + "learning_rate": 4.013687949406567e-05, + "loss": 1.2743, + "step": 114350 + }, + { + "epoch": 9.9, + "learning_rate": 4.013601316815386e-05, + "loss": 1.2288, + "step": 114360 + }, + { + "epoch": 9.9, + "learning_rate": 4.0135146842242055e-05, + "loss": 1.2808, + "step": 114370 + }, + { + "epoch": 9.9, + "learning_rate": 4.013428051633025e-05, + "loss": 1.3053, + "step": 114380 + }, + { + "epoch": 9.9, + "learning_rate": 4.013341419041844e-05, + "loss": 1.2714, + "step": 114390 + }, + { + "epoch": 9.9, + "learning_rate": 4.013254786450663e-05, + "loss": 1.2401, + "step": 114400 + }, + { + "epoch": 9.9, + "learning_rate": 4.0131681538594826e-05, + "loss": 1.2045, + "step": 114410 + }, + { + "epoch": 9.9, + "learning_rate": 4.0130815212683014e-05, + "loss": 1.2647, + "step": 114420 + }, + { + "epoch": 9.9, + "learning_rate": 4.012994888677121e-05, + "loss": 1.2674, + "step": 114430 + }, + { + "epoch": 9.91, + "learning_rate": 4.0129082560859396e-05, + "loss": 1.2657, + "step": 114440 + }, + { + "epoch": 9.91, + "learning_rate": 4.012821623494759e-05, + "loss": 1.2726, + "step": 114450 + }, + { + "epoch": 9.91, + "learning_rate": 4.0127349909035785e-05, + "loss": 1.2945, + "step": 114460 + }, + { + "epoch": 9.91, + "learning_rate": 4.012648358312397e-05, + "loss": 1.3032, + "step": 114470 + }, + { + "epoch": 9.91, + "learning_rate": 4.012561725721217e-05, + "loss": 1.2288, + "step": 114480 + }, + { + "epoch": 9.91, + "learning_rate": 4.012475093130036e-05, + "loss": 1.344, + "step": 114490 + }, + { + "epoch": 9.91, + "learning_rate": 4.012388460538855e-05, + "loss": 1.2362, + "step": 114500 + }, + { + "epoch": 9.91, + "learning_rate": 4.012301827947674e-05, + "loss": 1.2548, + "step": 114510 + }, + { + "epoch": 9.91, + "learning_rate": 4.012215195356494e-05, + "loss": 1.2504, + "step": 114520 + }, + { + "epoch": 9.91, + "learning_rate": 4.0121285627653125e-05, + "loss": 1.2048, + "step": 114530 + }, + { + "epoch": 9.91, + "learning_rate": 4.012041930174132e-05, + "loss": 1.2795, + "step": 114540 + }, + { + "epoch": 9.92, + "learning_rate": 4.011955297582951e-05, + "loss": 1.2941, + "step": 114550 + }, + { + "epoch": 9.92, + "learning_rate": 4.01186866499177e-05, + "loss": 1.3019, + "step": 114560 + }, + { + "epoch": 9.92, + "learning_rate": 4.0117820324005896e-05, + "loss": 1.2562, + "step": 114570 + }, + { + "epoch": 9.92, + "learning_rate": 4.0116953998094084e-05, + "loss": 1.2302, + "step": 114580 + }, + { + "epoch": 9.92, + "learning_rate": 4.011608767218228e-05, + "loss": 1.2716, + "step": 114590 + }, + { + "epoch": 9.92, + "learning_rate": 4.011522134627047e-05, + "loss": 1.2124, + "step": 114600 + }, + { + "epoch": 9.92, + "learning_rate": 4.011435502035866e-05, + "loss": 1.2656, + "step": 114610 + }, + { + "epoch": 9.92, + "learning_rate": 4.0113488694446855e-05, + "loss": 1.1848, + "step": 114620 + }, + { + "epoch": 9.92, + "learning_rate": 4.011262236853505e-05, + "loss": 1.2517, + "step": 114630 + }, + { + "epoch": 9.92, + "learning_rate": 4.011175604262324e-05, + "loss": 1.2135, + "step": 114640 + }, + { + "epoch": 9.92, + "learning_rate": 4.011088971671143e-05, + "loss": 1.3107, + "step": 114650 + }, + { + "epoch": 9.92, + "learning_rate": 4.011002339079962e-05, + "loss": 1.3227, + "step": 114660 + }, + { + "epoch": 9.93, + "learning_rate": 4.0109157064887814e-05, + "loss": 1.2863, + "step": 114670 + }, + { + "epoch": 9.93, + "learning_rate": 4.010829073897601e-05, + "loss": 1.2801, + "step": 114680 + }, + { + "epoch": 9.93, + "learning_rate": 4.0107424413064196e-05, + "loss": 1.244, + "step": 114690 + }, + { + "epoch": 9.93, + "learning_rate": 4.010655808715239e-05, + "loss": 1.225, + "step": 114700 + }, + { + "epoch": 9.93, + "learning_rate": 4.0105691761240585e-05, + "loss": 1.3166, + "step": 114710 + }, + { + "epoch": 9.93, + "learning_rate": 4.010482543532877e-05, + "loss": 1.2063, + "step": 114720 + }, + { + "epoch": 9.93, + "learning_rate": 4.010395910941697e-05, + "loss": 1.2636, + "step": 114730 + }, + { + "epoch": 9.93, + "learning_rate": 4.010309278350516e-05, + "loss": 1.3063, + "step": 114740 + }, + { + "epoch": 9.93, + "learning_rate": 4.010222645759335e-05, + "loss": 1.3354, + "step": 114750 + }, + { + "epoch": 9.93, + "learning_rate": 4.0101360131681544e-05, + "loss": 1.2299, + "step": 114760 + }, + { + "epoch": 9.93, + "learning_rate": 4.010049380576973e-05, + "loss": 1.2854, + "step": 114770 + }, + { + "epoch": 9.93, + "learning_rate": 4.0099627479857926e-05, + "loss": 1.2589, + "step": 114780 + }, + { + "epoch": 9.94, + "learning_rate": 4.009876115394612e-05, + "loss": 1.2164, + "step": 114790 + }, + { + "epoch": 9.94, + "learning_rate": 4.009789482803431e-05, + "loss": 1.2278, + "step": 114800 + }, + { + "epoch": 9.94, + "learning_rate": 4.00970285021225e-05, + "loss": 1.2364, + "step": 114810 + }, + { + "epoch": 9.94, + "learning_rate": 4.00961621762107e-05, + "loss": 1.2971, + "step": 114820 + }, + { + "epoch": 9.94, + "learning_rate": 4.0095295850298885e-05, + "loss": 1.3068, + "step": 114830 + }, + { + "epoch": 9.94, + "learning_rate": 4.009442952438708e-05, + "loss": 1.2876, + "step": 114840 + }, + { + "epoch": 9.94, + "learning_rate": 4.0093563198475273e-05, + "loss": 1.2435, + "step": 114850 + }, + { + "epoch": 9.94, + "learning_rate": 4.009269687256346e-05, + "loss": 1.2489, + "step": 114860 + }, + { + "epoch": 9.94, + "learning_rate": 4.0091830546651656e-05, + "loss": 1.2993, + "step": 114870 + }, + { + "epoch": 9.94, + "learning_rate": 4.009096422073984e-05, + "loss": 1.2115, + "step": 114880 + }, + { + "epoch": 9.94, + "learning_rate": 4.009009789482804e-05, + "loss": 1.2561, + "step": 114890 + }, + { + "epoch": 9.95, + "learning_rate": 4.008923156891623e-05, + "loss": 1.2809, + "step": 114900 + }, + { + "epoch": 9.95, + "learning_rate": 4.008836524300442e-05, + "loss": 1.3143, + "step": 114910 + }, + { + "epoch": 9.95, + "learning_rate": 4.0087498917092614e-05, + "loss": 1.2427, + "step": 114920 + }, + { + "epoch": 9.95, + "learning_rate": 4.008663259118081e-05, + "loss": 1.2256, + "step": 114930 + }, + { + "epoch": 9.95, + "learning_rate": 4.0085766265268996e-05, + "loss": 1.2557, + "step": 114940 + }, + { + "epoch": 9.95, + "learning_rate": 4.008489993935719e-05, + "loss": 1.2598, + "step": 114950 + }, + { + "epoch": 9.95, + "learning_rate": 4.0084033613445385e-05, + "loss": 1.2896, + "step": 114960 + }, + { + "epoch": 9.95, + "learning_rate": 4.008316728753357e-05, + "loss": 1.2575, + "step": 114970 + }, + { + "epoch": 9.95, + "learning_rate": 4.008230096162177e-05, + "loss": 1.2313, + "step": 114980 + }, + { + "epoch": 9.95, + "learning_rate": 4.0081434635709955e-05, + "loss": 1.2622, + "step": 114990 + }, + { + "epoch": 9.95, + "learning_rate": 4.008056830979815e-05, + "loss": 1.2554, + "step": 115000 + }, + { + "epoch": 9.95, + "learning_rate": 4.0079701983886344e-05, + "loss": 1.2388, + "step": 115010 + }, + { + "epoch": 9.96, + "learning_rate": 4.007883565797453e-05, + "loss": 1.2397, + "step": 115020 + }, + { + "epoch": 9.96, + "learning_rate": 4.0077969332062726e-05, + "loss": 1.2748, + "step": 115030 + }, + { + "epoch": 9.96, + "learning_rate": 4.007710300615092e-05, + "loss": 1.255, + "step": 115040 + }, + { + "epoch": 9.96, + "learning_rate": 4.007623668023911e-05, + "loss": 1.3035, + "step": 115050 + }, + { + "epoch": 9.96, + "learning_rate": 4.00753703543273e-05, + "loss": 1.2127, + "step": 115060 + }, + { + "epoch": 9.96, + "learning_rate": 4.00745040284155e-05, + "loss": 1.1726, + "step": 115070 + }, + { + "epoch": 9.96, + "learning_rate": 4.0073637702503685e-05, + "loss": 1.2019, + "step": 115080 + }, + { + "epoch": 9.96, + "learning_rate": 4.007277137659188e-05, + "loss": 1.2509, + "step": 115090 + }, + { + "epoch": 9.96, + "learning_rate": 4.007190505068007e-05, + "loss": 1.1909, + "step": 115100 + }, + { + "epoch": 9.96, + "learning_rate": 4.007103872476826e-05, + "loss": 1.2561, + "step": 115110 + }, + { + "epoch": 9.96, + "learning_rate": 4.0070172398856456e-05, + "loss": 1.2424, + "step": 115120 + }, + { + "epoch": 9.97, + "learning_rate": 4.0069306072944644e-05, + "loss": 1.2838, + "step": 115130 + }, + { + "epoch": 9.97, + "learning_rate": 4.006843974703284e-05, + "loss": 1.2237, + "step": 115140 + }, + { + "epoch": 9.97, + "learning_rate": 4.006757342112103e-05, + "loss": 1.2797, + "step": 115150 + }, + { + "epoch": 9.97, + "learning_rate": 4.006670709520922e-05, + "loss": 1.2532, + "step": 115160 + }, + { + "epoch": 9.97, + "learning_rate": 4.0065840769297415e-05, + "loss": 1.2928, + "step": 115170 + }, + { + "epoch": 9.97, + "learning_rate": 4.00649744433856e-05, + "loss": 1.2565, + "step": 115180 + }, + { + "epoch": 9.97, + "learning_rate": 4.00641081174738e-05, + "loss": 1.3059, + "step": 115190 + }, + { + "epoch": 9.97, + "learning_rate": 4.006324179156199e-05, + "loss": 1.2035, + "step": 115200 + }, + { + "epoch": 9.97, + "learning_rate": 4.006237546565018e-05, + "loss": 1.2279, + "step": 115210 + }, + { + "epoch": 9.97, + "learning_rate": 4.006150913973837e-05, + "loss": 1.2413, + "step": 115220 + }, + { + "epoch": 9.97, + "learning_rate": 4.006064281382657e-05, + "loss": 1.2271, + "step": 115230 + }, + { + "epoch": 9.97, + "learning_rate": 4.0059776487914755e-05, + "loss": 1.2433, + "step": 115240 + }, + { + "epoch": 9.98, + "learning_rate": 4.005891016200295e-05, + "loss": 1.1823, + "step": 115250 + }, + { + "epoch": 9.98, + "learning_rate": 4.0058043836091144e-05, + "loss": 1.2149, + "step": 115260 + }, + { + "epoch": 9.98, + "learning_rate": 4.005717751017933e-05, + "loss": 1.2251, + "step": 115270 + }, + { + "epoch": 9.98, + "learning_rate": 4.0056311184267526e-05, + "loss": 1.2609, + "step": 115280 + }, + { + "epoch": 9.98, + "learning_rate": 4.0055444858355714e-05, + "loss": 1.2094, + "step": 115290 + }, + { + "epoch": 9.98, + "learning_rate": 4.005457853244391e-05, + "loss": 1.2474, + "step": 115300 + }, + { + "epoch": 9.98, + "learning_rate": 4.00537122065321e-05, + "loss": 1.2629, + "step": 115310 + }, + { + "epoch": 9.98, + "learning_rate": 4.005284588062029e-05, + "loss": 1.2185, + "step": 115320 + }, + { + "epoch": 9.98, + "learning_rate": 4.0051979554708485e-05, + "loss": 1.2382, + "step": 115330 + }, + { + "epoch": 9.98, + "learning_rate": 4.005111322879668e-05, + "loss": 1.2433, + "step": 115340 + }, + { + "epoch": 9.98, + "learning_rate": 4.005024690288487e-05, + "loss": 1.2808, + "step": 115350 + }, + { + "epoch": 9.99, + "learning_rate": 4.004938057697306e-05, + "loss": 1.2355, + "step": 115360 + }, + { + "epoch": 9.99, + "learning_rate": 4.0048514251061256e-05, + "loss": 1.2888, + "step": 115370 + }, + { + "epoch": 9.99, + "learning_rate": 4.0047647925149444e-05, + "loss": 1.3165, + "step": 115380 + }, + { + "epoch": 9.99, + "learning_rate": 4.004678159923764e-05, + "loss": 1.235, + "step": 115390 + }, + { + "epoch": 9.99, + "learning_rate": 4.0045915273325826e-05, + "loss": 1.2442, + "step": 115400 + }, + { + "epoch": 9.99, + "learning_rate": 4.004504894741402e-05, + "loss": 1.2681, + "step": 115410 + }, + { + "epoch": 9.99, + "learning_rate": 4.0044182621502215e-05, + "loss": 1.2797, + "step": 115420 + }, + { + "epoch": 9.99, + "learning_rate": 4.00433162955904e-05, + "loss": 1.2411, + "step": 115430 + }, + { + "epoch": 9.99, + "learning_rate": 4.00424499696786e-05, + "loss": 1.3164, + "step": 115440 + }, + { + "epoch": 9.99, + "learning_rate": 4.004158364376679e-05, + "loss": 1.2496, + "step": 115450 + }, + { + "epoch": 9.99, + "learning_rate": 4.004071731785498e-05, + "loss": 1.2231, + "step": 115460 + }, + { + "epoch": 9.99, + "learning_rate": 4.0039850991943174e-05, + "loss": 1.2935, + "step": 115470 + }, + { + "epoch": 10.0, + "learning_rate": 4.003898466603137e-05, + "loss": 1.2442, + "step": 115480 + }, + { + "epoch": 10.0, + "learning_rate": 4.0038118340119556e-05, + "loss": 1.3379, + "step": 115490 + }, + { + "epoch": 10.0, + "learning_rate": 4.003725201420775e-05, + "loss": 1.2905, + "step": 115500 + }, + { + "epoch": 10.0, + "learning_rate": 4.003638568829594e-05, + "loss": 1.2463, + "step": 115510 + }, + { + "epoch": 10.0, + "learning_rate": 4.003551936238413e-05, + "loss": 1.3015, + "step": 115520 + }, + { + "epoch": 10.0, + "learning_rate": 4.003465303647233e-05, + "loss": 1.2398, + "step": 115530 + }, + { + "epoch": 10.0, + "eval_Bleu_1": 0.03199315959171256, + "eval_Bleu_2": 2.5045803142306567e-11, + "eval_Bleu_3": 2.390485953862402e-14, + "eval_Bleu_4": 7.604801750101983e-16, + "eval_ROUGE_L": 0.06714386086255107, + "eval_cer": 0.9940441736300689, + "eval_em": 0, + "eval_f1": 0.08118331233164663, + "eval_loss": 1.1635690927505493, + "eval_runtime": 1673.2485, + "eval_samples_per_second": 3.069, + "eval_steps_per_second": 3.069, + "eval_wer": 0.979681955486435, + "step": 115532 + }, + { + "epoch": 10.0, + "learning_rate": 4.0033786710560514e-05, + "loss": 1.2381, + "step": 115540 + }, + { + "epoch": 10.0, + "learning_rate": 4.003292038464871e-05, + "loss": 1.2393, + "step": 115550 + }, + { + "epoch": 10.0, + "learning_rate": 4.00320540587369e-05, + "loss": 1.2102, + "step": 115560 + }, + { + "epoch": 10.0, + "learning_rate": 4.003118773282509e-05, + "loss": 1.2836, + "step": 115570 + }, + { + "epoch": 10.0, + "learning_rate": 4.0030321406913286e-05, + "loss": 1.2203, + "step": 115580 + }, + { + "epoch": 10.01, + "learning_rate": 4.002945508100148e-05, + "loss": 1.2904, + "step": 115590 + }, + { + "epoch": 10.01, + "learning_rate": 4.002858875508967e-05, + "loss": 1.2721, + "step": 115600 + }, + { + "epoch": 10.01, + "learning_rate": 4.002772242917786e-05, + "loss": 1.1836, + "step": 115610 + }, + { + "epoch": 10.01, + "learning_rate": 4.002685610326605e-05, + "loss": 1.2114, + "step": 115620 + }, + { + "epoch": 10.01, + "learning_rate": 4.0025989777354244e-05, + "loss": 1.2727, + "step": 115630 + }, + { + "epoch": 10.01, + "learning_rate": 4.002512345144244e-05, + "loss": 1.2517, + "step": 115640 + }, + { + "epoch": 10.01, + "learning_rate": 4.0024257125530626e-05, + "loss": 1.2802, + "step": 115650 + }, + { + "epoch": 10.01, + "learning_rate": 4.002339079961882e-05, + "loss": 1.1891, + "step": 115660 + }, + { + "epoch": 10.01, + "learning_rate": 4.0022524473707015e-05, + "loss": 1.1928, + "step": 115670 + }, + { + "epoch": 10.01, + "learning_rate": 4.00216581477952e-05, + "loss": 1.2454, + "step": 115680 + }, + { + "epoch": 10.01, + "learning_rate": 4.00207918218834e-05, + "loss": 1.2003, + "step": 115690 + }, + { + "epoch": 10.01, + "learning_rate": 4.001992549597159e-05, + "loss": 1.3027, + "step": 115700 + }, + { + "epoch": 10.02, + "learning_rate": 4.001905917005978e-05, + "loss": 1.2815, + "step": 115710 + }, + { + "epoch": 10.02, + "learning_rate": 4.0018192844147974e-05, + "loss": 1.2457, + "step": 115720 + }, + { + "epoch": 10.02, + "learning_rate": 4.001732651823616e-05, + "loss": 1.2084, + "step": 115730 + }, + { + "epoch": 10.02, + "learning_rate": 4.0016460192324356e-05, + "loss": 1.1796, + "step": 115740 + }, + { + "epoch": 10.02, + "learning_rate": 4.001559386641255e-05, + "loss": 1.3295, + "step": 115750 + }, + { + "epoch": 10.02, + "learning_rate": 4.001472754050074e-05, + "loss": 1.2772, + "step": 115760 + }, + { + "epoch": 10.02, + "learning_rate": 4.001386121458893e-05, + "loss": 1.2337, + "step": 115770 + }, + { + "epoch": 10.02, + "learning_rate": 4.001299488867713e-05, + "loss": 1.242, + "step": 115780 + }, + { + "epoch": 10.02, + "learning_rate": 4.0012128562765315e-05, + "loss": 1.2828, + "step": 115790 + }, + { + "epoch": 10.02, + "learning_rate": 4.001126223685351e-05, + "loss": 1.2925, + "step": 115800 + }, + { + "epoch": 10.02, + "learning_rate": 4.0010395910941704e-05, + "loss": 1.2043, + "step": 115810 + }, + { + "epoch": 10.02, + "learning_rate": 4.000952958502989e-05, + "loss": 1.2675, + "step": 115820 + }, + { + "epoch": 10.03, + "learning_rate": 4.0008663259118086e-05, + "loss": 1.2104, + "step": 115830 + }, + { + "epoch": 10.03, + "learning_rate": 4.0007796933206274e-05, + "loss": 1.2288, + "step": 115840 + }, + { + "epoch": 10.03, + "learning_rate": 4.000693060729447e-05, + "loss": 1.2508, + "step": 115850 + }, + { + "epoch": 10.03, + "learning_rate": 4.000606428138266e-05, + "loss": 1.2723, + "step": 115860 + }, + { + "epoch": 10.03, + "learning_rate": 4.000519795547085e-05, + "loss": 1.295, + "step": 115870 + }, + { + "epoch": 10.03, + "learning_rate": 4.0004331629559045e-05, + "loss": 1.2052, + "step": 115880 + }, + { + "epoch": 10.03, + "learning_rate": 4.000346530364724e-05, + "loss": 1.1923, + "step": 115890 + }, + { + "epoch": 10.03, + "learning_rate": 4.000259897773543e-05, + "loss": 1.2406, + "step": 115900 + }, + { + "epoch": 10.03, + "learning_rate": 4.000173265182362e-05, + "loss": 1.1647, + "step": 115910 + }, + { + "epoch": 10.03, + "learning_rate": 4.000086632591181e-05, + "loss": 1.2426, + "step": 115920 + }, + { + "epoch": 10.03, + "learning_rate": 4e-05, + "loss": 1.2441, + "step": 115930 + }, + { + "epoch": 10.04, + "learning_rate": 3.99991336740882e-05, + "loss": 1.2478, + "step": 115940 + }, + { + "epoch": 10.04, + "learning_rate": 3.9998267348176385e-05, + "loss": 1.2559, + "step": 115950 + }, + { + "epoch": 10.04, + "learning_rate": 3.999740102226458e-05, + "loss": 1.2143, + "step": 115960 + }, + { + "epoch": 10.04, + "learning_rate": 3.9996534696352774e-05, + "loss": 1.1956, + "step": 115970 + }, + { + "epoch": 10.04, + "learning_rate": 3.999566837044096e-05, + "loss": 1.2267, + "step": 115980 + }, + { + "epoch": 10.04, + "learning_rate": 3.9994802044529156e-05, + "loss": 1.1943, + "step": 115990 + }, + { + "epoch": 10.04, + "learning_rate": 3.999393571861735e-05, + "loss": 1.2661, + "step": 116000 + }, + { + "epoch": 10.04, + "learning_rate": 3.999306939270554e-05, + "loss": 1.306, + "step": 116010 + }, + { + "epoch": 10.04, + "learning_rate": 3.999220306679373e-05, + "loss": 1.2237, + "step": 116020 + }, + { + "epoch": 10.04, + "learning_rate": 3.999133674088192e-05, + "loss": 1.2276, + "step": 116030 + }, + { + "epoch": 10.04, + "learning_rate": 3.9990470414970115e-05, + "loss": 1.2341, + "step": 116040 + }, + { + "epoch": 10.04, + "learning_rate": 3.998960408905831e-05, + "loss": 1.201, + "step": 116050 + }, + { + "epoch": 10.05, + "learning_rate": 3.99887377631465e-05, + "loss": 1.2905, + "step": 116060 + }, + { + "epoch": 10.05, + "learning_rate": 3.998787143723469e-05, + "loss": 1.2022, + "step": 116070 + }, + { + "epoch": 10.05, + "learning_rate": 3.9987005111322886e-05, + "loss": 1.255, + "step": 116080 + }, + { + "epoch": 10.05, + "learning_rate": 3.9986138785411074e-05, + "loss": 1.2557, + "step": 116090 + }, + { + "epoch": 10.05, + "learning_rate": 3.998527245949927e-05, + "loss": 1.2132, + "step": 116100 + }, + { + "epoch": 10.05, + "learning_rate": 3.998440613358746e-05, + "loss": 1.1798, + "step": 116110 + }, + { + "epoch": 10.05, + "learning_rate": 3.998353980767565e-05, + "loss": 1.2109, + "step": 116120 + }, + { + "epoch": 10.05, + "learning_rate": 3.9982673481763845e-05, + "loss": 1.2621, + "step": 116130 + }, + { + "epoch": 10.05, + "learning_rate": 3.998180715585203e-05, + "loss": 1.2306, + "step": 116140 + }, + { + "epoch": 10.05, + "learning_rate": 3.998094082994023e-05, + "loss": 1.2989, + "step": 116150 + }, + { + "epoch": 10.05, + "learning_rate": 3.998007450402842e-05, + "loss": 1.2525, + "step": 116160 + }, + { + "epoch": 10.06, + "learning_rate": 3.997920817811661e-05, + "loss": 1.2399, + "step": 116170 + }, + { + "epoch": 10.06, + "learning_rate": 3.9978341852204804e-05, + "loss": 1.2577, + "step": 116180 + }, + { + "epoch": 10.06, + "learning_rate": 3.9977475526293e-05, + "loss": 1.3346, + "step": 116190 + }, + { + "epoch": 10.06, + "learning_rate": 3.9976609200381186e-05, + "loss": 1.289, + "step": 116200 + }, + { + "epoch": 10.06, + "learning_rate": 3.997574287446938e-05, + "loss": 1.2211, + "step": 116210 + }, + { + "epoch": 10.06, + "learning_rate": 3.9974876548557575e-05, + "loss": 1.1376, + "step": 116220 + }, + { + "epoch": 10.06, + "learning_rate": 3.997401022264576e-05, + "loss": 1.2367, + "step": 116230 + }, + { + "epoch": 10.06, + "learning_rate": 3.997314389673396e-05, + "loss": 1.2377, + "step": 116240 + }, + { + "epoch": 10.06, + "learning_rate": 3.9972277570822144e-05, + "loss": 1.2604, + "step": 116250 + }, + { + "epoch": 10.06, + "learning_rate": 3.997141124491034e-05, + "loss": 1.2835, + "step": 116260 + }, + { + "epoch": 10.06, + "learning_rate": 3.997054491899853e-05, + "loss": 1.249, + "step": 116270 + }, + { + "epoch": 10.06, + "learning_rate": 3.996967859308672e-05, + "loss": 1.2568, + "step": 116280 + }, + { + "epoch": 10.07, + "learning_rate": 3.9968812267174915e-05, + "loss": 1.2747, + "step": 116290 + }, + { + "epoch": 10.07, + "learning_rate": 3.996794594126311e-05, + "loss": 1.237, + "step": 116300 + }, + { + "epoch": 10.07, + "learning_rate": 3.99670796153513e-05, + "loss": 1.2004, + "step": 116310 + }, + { + "epoch": 10.07, + "learning_rate": 3.996621328943949e-05, + "loss": 1.2345, + "step": 116320 + }, + { + "epoch": 10.07, + "learning_rate": 3.9965346963527687e-05, + "loss": 1.2291, + "step": 116330 + }, + { + "epoch": 10.07, + "learning_rate": 3.9964480637615874e-05, + "loss": 1.2485, + "step": 116340 + }, + { + "epoch": 10.07, + "learning_rate": 3.996361431170407e-05, + "loss": 1.2592, + "step": 116350 + }, + { + "epoch": 10.07, + "learning_rate": 3.9962747985792256e-05, + "loss": 1.1729, + "step": 116360 + }, + { + "epoch": 10.07, + "learning_rate": 3.996188165988045e-05, + "loss": 1.3088, + "step": 116370 + }, + { + "epoch": 10.07, + "learning_rate": 3.9961015333968645e-05, + "loss": 1.194, + "step": 116380 + }, + { + "epoch": 10.07, + "learning_rate": 3.996014900805683e-05, + "loss": 1.1842, + "step": 116390 + }, + { + "epoch": 10.08, + "learning_rate": 3.995928268214503e-05, + "loss": 1.28, + "step": 116400 + }, + { + "epoch": 10.08, + "learning_rate": 3.995841635623322e-05, + "loss": 1.262, + "step": 116410 + }, + { + "epoch": 10.08, + "learning_rate": 3.995755003032141e-05, + "loss": 1.2124, + "step": 116420 + }, + { + "epoch": 10.08, + "learning_rate": 3.9956683704409604e-05, + "loss": 1.2597, + "step": 116430 + }, + { + "epoch": 10.08, + "learning_rate": 3.99558173784978e-05, + "loss": 1.2589, + "step": 116440 + }, + { + "epoch": 10.08, + "learning_rate": 3.9954951052585986e-05, + "loss": 1.2562, + "step": 116450 + }, + { + "epoch": 10.08, + "learning_rate": 3.995408472667418e-05, + "loss": 1.2326, + "step": 116460 + }, + { + "epoch": 10.08, + "learning_rate": 3.995321840076237e-05, + "loss": 1.194, + "step": 116470 + }, + { + "epoch": 10.08, + "learning_rate": 3.995235207485056e-05, + "loss": 1.2058, + "step": 116480 + }, + { + "epoch": 10.08, + "learning_rate": 3.995148574893876e-05, + "loss": 1.1612, + "step": 116490 + }, + { + "epoch": 10.08, + "learning_rate": 3.9950619423026945e-05, + "loss": 1.3083, + "step": 116500 + }, + { + "epoch": 10.08, + "learning_rate": 3.994975309711514e-05, + "loss": 1.2798, + "step": 116510 + }, + { + "epoch": 10.09, + "learning_rate": 3.9948886771203334e-05, + "loss": 1.2151, + "step": 116520 + }, + { + "epoch": 10.09, + "learning_rate": 3.994802044529152e-05, + "loss": 1.1771, + "step": 116530 + }, + { + "epoch": 10.09, + "learning_rate": 3.9947154119379716e-05, + "loss": 1.2379, + "step": 116540 + }, + { + "epoch": 10.09, + "learning_rate": 3.994628779346791e-05, + "loss": 1.2741, + "step": 116550 + }, + { + "epoch": 10.09, + "learning_rate": 3.99454214675561e-05, + "loss": 1.2175, + "step": 116560 + }, + { + "epoch": 10.09, + "learning_rate": 3.994455514164429e-05, + "loss": 1.2163, + "step": 116570 + }, + { + "epoch": 10.09, + "learning_rate": 3.994368881573248e-05, + "loss": 1.1995, + "step": 116580 + }, + { + "epoch": 10.09, + "learning_rate": 3.9942822489820675e-05, + "loss": 1.3193, + "step": 116590 + }, + { + "epoch": 10.09, + "learning_rate": 3.994195616390887e-05, + "loss": 1.2192, + "step": 116600 + }, + { + "epoch": 10.09, + "learning_rate": 3.994108983799706e-05, + "loss": 1.2704, + "step": 116610 + }, + { + "epoch": 10.09, + "learning_rate": 3.994022351208525e-05, + "loss": 1.2342, + "step": 116620 + }, + { + "epoch": 10.1, + "learning_rate": 3.9939357186173446e-05, + "loss": 1.2762, + "step": 116630 + }, + { + "epoch": 10.1, + "learning_rate": 3.993849086026163e-05, + "loss": 1.1585, + "step": 116640 + }, + { + "epoch": 10.1, + "learning_rate": 3.993762453434983e-05, + "loss": 1.227, + "step": 116650 + }, + { + "epoch": 10.1, + "learning_rate": 3.9936758208438015e-05, + "loss": 1.2214, + "step": 116660 + }, + { + "epoch": 10.1, + "learning_rate": 3.993589188252621e-05, + "loss": 1.221, + "step": 116670 + }, + { + "epoch": 10.1, + "learning_rate": 3.9935025556614404e-05, + "loss": 1.2353, + "step": 116680 + }, + { + "epoch": 10.1, + "learning_rate": 3.993415923070259e-05, + "loss": 1.2681, + "step": 116690 + }, + { + "epoch": 10.1, + "learning_rate": 3.9933292904790786e-05, + "loss": 1.2428, + "step": 116700 + }, + { + "epoch": 10.1, + "learning_rate": 3.993242657887898e-05, + "loss": 1.1795, + "step": 116710 + }, + { + "epoch": 10.1, + "learning_rate": 3.993156025296717e-05, + "loss": 1.1766, + "step": 116720 + }, + { + "epoch": 10.1, + "learning_rate": 3.993069392705536e-05, + "loss": 1.2994, + "step": 116730 + }, + { + "epoch": 10.1, + "learning_rate": 3.992982760114356e-05, + "loss": 1.2642, + "step": 116740 + }, + { + "epoch": 10.11, + "learning_rate": 3.9928961275231745e-05, + "loss": 1.2105, + "step": 116750 + }, + { + "epoch": 10.11, + "learning_rate": 3.992809494931994e-05, + "loss": 1.2322, + "step": 116760 + }, + { + "epoch": 10.11, + "learning_rate": 3.992722862340813e-05, + "loss": 1.1914, + "step": 116770 + }, + { + "epoch": 10.11, + "learning_rate": 3.992636229749632e-05, + "loss": 1.2458, + "step": 116780 + }, + { + "epoch": 10.11, + "learning_rate": 3.9925495971584516e-05, + "loss": 1.3025, + "step": 116790 + }, + { + "epoch": 10.11, + "learning_rate": 3.9924629645672704e-05, + "loss": 1.2095, + "step": 116800 + }, + { + "epoch": 10.11, + "learning_rate": 3.99237633197609e-05, + "loss": 1.1591, + "step": 116810 + }, + { + "epoch": 10.11, + "learning_rate": 3.992289699384909e-05, + "loss": 1.1945, + "step": 116820 + }, + { + "epoch": 10.11, + "learning_rate": 3.992203066793728e-05, + "loss": 1.249, + "step": 116830 + }, + { + "epoch": 10.11, + "learning_rate": 3.9921164342025475e-05, + "loss": 1.2124, + "step": 116840 + }, + { + "epoch": 10.11, + "learning_rate": 3.992029801611367e-05, + "loss": 1.2349, + "step": 116850 + }, + { + "epoch": 10.11, + "learning_rate": 3.991943169020186e-05, + "loss": 1.2299, + "step": 116860 + }, + { + "epoch": 10.12, + "learning_rate": 3.991856536429005e-05, + "loss": 1.2397, + "step": 116870 + }, + { + "epoch": 10.12, + "learning_rate": 3.991769903837824e-05, + "loss": 1.2627, + "step": 116880 + }, + { + "epoch": 10.12, + "learning_rate": 3.9916832712466434e-05, + "loss": 1.2996, + "step": 116890 + }, + { + "epoch": 10.12, + "learning_rate": 3.991596638655463e-05, + "loss": 1.2223, + "step": 116900 + }, + { + "epoch": 10.12, + "learning_rate": 3.9915100060642816e-05, + "loss": 1.2257, + "step": 116910 + }, + { + "epoch": 10.12, + "learning_rate": 3.991423373473101e-05, + "loss": 1.1711, + "step": 116920 + }, + { + "epoch": 10.12, + "learning_rate": 3.9913367408819205e-05, + "loss": 1.2199, + "step": 116930 + }, + { + "epoch": 10.12, + "learning_rate": 3.991250108290739e-05, + "loss": 1.1581, + "step": 116940 + }, + { + "epoch": 10.12, + "learning_rate": 3.991163475699559e-05, + "loss": 1.2155, + "step": 116950 + }, + { + "epoch": 10.12, + "learning_rate": 3.991076843108378e-05, + "loss": 1.2059, + "step": 116960 + }, + { + "epoch": 10.12, + "learning_rate": 3.990990210517197e-05, + "loss": 1.2828, + "step": 116970 + }, + { + "epoch": 10.13, + "learning_rate": 3.990903577926016e-05, + "loss": 1.2206, + "step": 116980 + }, + { + "epoch": 10.13, + "learning_rate": 3.990816945334835e-05, + "loss": 1.1983, + "step": 116990 + }, + { + "epoch": 10.13, + "learning_rate": 3.9907303127436545e-05, + "loss": 1.1931, + "step": 117000 + }, + { + "epoch": 10.13, + "learning_rate": 3.990643680152474e-05, + "loss": 1.2782, + "step": 117010 + }, + { + "epoch": 10.13, + "learning_rate": 3.990557047561293e-05, + "loss": 1.1884, + "step": 117020 + }, + { + "epoch": 10.13, + "learning_rate": 3.990470414970112e-05, + "loss": 1.1953, + "step": 117030 + }, + { + "epoch": 10.13, + "learning_rate": 3.9903837823789316e-05, + "loss": 1.1952, + "step": 117040 + }, + { + "epoch": 10.13, + "learning_rate": 3.9902971497877504e-05, + "loss": 1.1521, + "step": 117050 + }, + { + "epoch": 10.13, + "learning_rate": 3.99021051719657e-05, + "loss": 1.2449, + "step": 117060 + }, + { + "epoch": 10.13, + "learning_rate": 3.990123884605389e-05, + "loss": 1.2716, + "step": 117070 + }, + { + "epoch": 10.13, + "learning_rate": 3.990037252014208e-05, + "loss": 1.2007, + "step": 117080 + }, + { + "epoch": 10.13, + "learning_rate": 3.9899506194230275e-05, + "loss": 1.2307, + "step": 117090 + }, + { + "epoch": 10.14, + "learning_rate": 3.989863986831846e-05, + "loss": 1.2395, + "step": 117100 + }, + { + "epoch": 10.14, + "learning_rate": 3.989777354240666e-05, + "loss": 1.2455, + "step": 117110 + }, + { + "epoch": 10.14, + "learning_rate": 3.989690721649485e-05, + "loss": 1.163, + "step": 117120 + }, + { + "epoch": 10.14, + "learning_rate": 3.989604089058304e-05, + "loss": 1.2692, + "step": 117130 + }, + { + "epoch": 10.14, + "learning_rate": 3.9895174564671234e-05, + "loss": 1.2312, + "step": 117140 + }, + { + "epoch": 10.14, + "learning_rate": 3.989430823875943e-05, + "loss": 1.241, + "step": 117150 + }, + { + "epoch": 10.14, + "learning_rate": 3.9893441912847616e-05, + "loss": 1.2338, + "step": 117160 + }, + { + "epoch": 10.14, + "learning_rate": 3.989257558693581e-05, + "loss": 1.2422, + "step": 117170 + }, + { + "epoch": 10.14, + "learning_rate": 3.9891709261024005e-05, + "loss": 1.2725, + "step": 117180 + }, + { + "epoch": 10.14, + "learning_rate": 3.989084293511219e-05, + "loss": 1.2435, + "step": 117190 + }, + { + "epoch": 10.14, + "learning_rate": 3.988997660920039e-05, + "loss": 1.2668, + "step": 117200 + }, + { + "epoch": 10.15, + "learning_rate": 3.9889110283288575e-05, + "loss": 1.2309, + "step": 117210 + }, + { + "epoch": 10.15, + "learning_rate": 3.988824395737677e-05, + "loss": 1.2596, + "step": 117220 + }, + { + "epoch": 10.15, + "learning_rate": 3.9887377631464964e-05, + "loss": 1.1639, + "step": 117230 + }, + { + "epoch": 10.15, + "learning_rate": 3.988651130555315e-05, + "loss": 1.2292, + "step": 117240 + }, + { + "epoch": 10.15, + "learning_rate": 3.9885644979641346e-05, + "loss": 1.2248, + "step": 117250 + }, + { + "epoch": 10.15, + "learning_rate": 3.988477865372954e-05, + "loss": 1.2208, + "step": 117260 + }, + { + "epoch": 10.15, + "learning_rate": 3.988391232781773e-05, + "loss": 1.2297, + "step": 117270 + }, + { + "epoch": 10.15, + "learning_rate": 3.988304600190592e-05, + "loss": 1.2155, + "step": 117280 + }, + { + "epoch": 10.15, + "learning_rate": 3.988217967599412e-05, + "loss": 1.2619, + "step": 117290 + }, + { + "epoch": 10.15, + "learning_rate": 3.9881313350082305e-05, + "loss": 1.2247, + "step": 117300 + }, + { + "epoch": 10.15, + "learning_rate": 3.98804470241705e-05, + "loss": 1.236, + "step": 117310 + }, + { + "epoch": 10.15, + "learning_rate": 3.987958069825869e-05, + "loss": 1.2161, + "step": 117320 + }, + { + "epoch": 10.16, + "learning_rate": 3.987871437234688e-05, + "loss": 1.2231, + "step": 117330 + }, + { + "epoch": 10.16, + "learning_rate": 3.9877848046435076e-05, + "loss": 1.2305, + "step": 117340 + }, + { + "epoch": 10.16, + "learning_rate": 3.987698172052326e-05, + "loss": 1.2244, + "step": 117350 + }, + { + "epoch": 10.16, + "learning_rate": 3.987611539461146e-05, + "loss": 1.1634, + "step": 117360 + }, + { + "epoch": 10.16, + "learning_rate": 3.987524906869965e-05, + "loss": 1.2275, + "step": 117370 + }, + { + "epoch": 10.16, + "learning_rate": 3.987438274278784e-05, + "loss": 1.2375, + "step": 117380 + }, + { + "epoch": 10.16, + "learning_rate": 3.9873516416876034e-05, + "loss": 1.1828, + "step": 117390 + }, + { + "epoch": 10.16, + "learning_rate": 3.987265009096423e-05, + "loss": 1.2268, + "step": 117400 + }, + { + "epoch": 10.16, + "learning_rate": 3.9871783765052416e-05, + "loss": 1.1984, + "step": 117410 + }, + { + "epoch": 10.16, + "learning_rate": 3.987091743914061e-05, + "loss": 1.2806, + "step": 117420 + }, + { + "epoch": 10.16, + "learning_rate": 3.98700511132288e-05, + "loss": 1.2623, + "step": 117430 + }, + { + "epoch": 10.17, + "learning_rate": 3.986918478731699e-05, + "loss": 1.2431, + "step": 117440 + }, + { + "epoch": 10.17, + "learning_rate": 3.986831846140519e-05, + "loss": 1.1586, + "step": 117450 + }, + { + "epoch": 10.17, + "learning_rate": 3.9867452135493375e-05, + "loss": 1.1922, + "step": 117460 + }, + { + "epoch": 10.17, + "learning_rate": 3.986658580958157e-05, + "loss": 1.1957, + "step": 117470 + }, + { + "epoch": 10.17, + "learning_rate": 3.9865719483669764e-05, + "loss": 1.2821, + "step": 117480 + }, + { + "epoch": 10.17, + "learning_rate": 3.986485315775795e-05, + "loss": 1.2591, + "step": 117490 + }, + { + "epoch": 10.17, + "learning_rate": 3.9863986831846146e-05, + "loss": 1.2511, + "step": 117500 + }, + { + "epoch": 10.17, + "learning_rate": 3.9863120505934334e-05, + "loss": 1.2275, + "step": 117510 + }, + { + "epoch": 10.17, + "learning_rate": 3.986225418002253e-05, + "loss": 1.2493, + "step": 117520 + }, + { + "epoch": 10.17, + "learning_rate": 3.986138785411072e-05, + "loss": 1.1998, + "step": 117530 + }, + { + "epoch": 10.17, + "learning_rate": 3.986052152819891e-05, + "loss": 1.247, + "step": 117540 + }, + { + "epoch": 10.17, + "learning_rate": 3.9859655202287105e-05, + "loss": 1.2723, + "step": 117550 + }, + { + "epoch": 10.18, + "learning_rate": 3.98587888763753e-05, + "loss": 1.2534, + "step": 117560 + }, + { + "epoch": 10.18, + "learning_rate": 3.985792255046349e-05, + "loss": 1.2517, + "step": 117570 + }, + { + "epoch": 10.18, + "learning_rate": 3.985705622455168e-05, + "loss": 1.2338, + "step": 117580 + }, + { + "epoch": 10.18, + "learning_rate": 3.9856189898639876e-05, + "loss": 1.2582, + "step": 117590 + }, + { + "epoch": 10.18, + "learning_rate": 3.9855323572728064e-05, + "loss": 1.2458, + "step": 117600 + }, + { + "epoch": 10.18, + "learning_rate": 3.985445724681626e-05, + "loss": 1.2874, + "step": 117610 + }, + { + "epoch": 10.18, + "learning_rate": 3.9853590920904446e-05, + "loss": 1.2328, + "step": 117620 + }, + { + "epoch": 10.18, + "learning_rate": 3.985272459499264e-05, + "loss": 1.2767, + "step": 117630 + }, + { + "epoch": 10.18, + "learning_rate": 3.9851858269080835e-05, + "loss": 1.2066, + "step": 117640 + }, + { + "epoch": 10.18, + "learning_rate": 3.985099194316902e-05, + "loss": 1.2133, + "step": 117650 + }, + { + "epoch": 10.18, + "learning_rate": 3.985012561725722e-05, + "loss": 1.2289, + "step": 117660 + }, + { + "epoch": 10.19, + "learning_rate": 3.984925929134541e-05, + "loss": 1.2001, + "step": 117670 + }, + { + "epoch": 10.19, + "learning_rate": 3.98483929654336e-05, + "loss": 1.1961, + "step": 117680 + }, + { + "epoch": 10.19, + "learning_rate": 3.984752663952179e-05, + "loss": 1.1983, + "step": 117690 + }, + { + "epoch": 10.19, + "learning_rate": 3.984666031360999e-05, + "loss": 1.2591, + "step": 117700 + }, + { + "epoch": 10.19, + "learning_rate": 3.9845793987698175e-05, + "loss": 1.1757, + "step": 117710 + }, + { + "epoch": 10.19, + "learning_rate": 3.984492766178637e-05, + "loss": 1.2597, + "step": 117720 + }, + { + "epoch": 10.19, + "learning_rate": 3.984406133587456e-05, + "loss": 1.2152, + "step": 117730 + }, + { + "epoch": 10.19, + "learning_rate": 3.9843195009962745e-05, + "loss": 1.2436, + "step": 117740 + }, + { + "epoch": 10.19, + "learning_rate": 3.984232868405094e-05, + "loss": 1.2181, + "step": 117750 + }, + { + "epoch": 10.19, + "learning_rate": 3.9841462358139134e-05, + "loss": 1.206, + "step": 117760 + }, + { + "epoch": 10.19, + "learning_rate": 3.984059603222732e-05, + "loss": 1.2672, + "step": 117770 + }, + { + "epoch": 10.19, + "learning_rate": 3.9839729706315516e-05, + "loss": 1.185, + "step": 117780 + }, + { + "epoch": 10.2, + "learning_rate": 3.9838863380403704e-05, + "loss": 1.2452, + "step": 117790 + }, + { + "epoch": 10.2, + "learning_rate": 3.98379970544919e-05, + "loss": 1.2394, + "step": 117800 + }, + { + "epoch": 10.2, + "learning_rate": 3.983713072858009e-05, + "loss": 1.2514, + "step": 117810 + }, + { + "epoch": 10.2, + "learning_rate": 3.983626440266828e-05, + "loss": 1.1617, + "step": 117820 + }, + { + "epoch": 10.2, + "learning_rate": 3.9835398076756475e-05, + "loss": 1.1968, + "step": 117830 + }, + { + "epoch": 10.2, + "learning_rate": 3.983453175084467e-05, + "loss": 1.1969, + "step": 117840 + }, + { + "epoch": 10.2, + "learning_rate": 3.983366542493286e-05, + "loss": 1.2117, + "step": 117850 + }, + { + "epoch": 10.2, + "learning_rate": 3.983279909902105e-05, + "loss": 1.239, + "step": 117860 + }, + { + "epoch": 10.2, + "learning_rate": 3.9831932773109246e-05, + "loss": 1.1632, + "step": 117870 + }, + { + "epoch": 10.2, + "learning_rate": 3.9831066447197434e-05, + "loss": 1.1536, + "step": 117880 + }, + { + "epoch": 10.2, + "learning_rate": 3.983020012128563e-05, + "loss": 1.2306, + "step": 117890 + }, + { + "epoch": 10.2, + "learning_rate": 3.9829333795373816e-05, + "loss": 1.2162, + "step": 117900 + }, + { + "epoch": 10.21, + "learning_rate": 3.982846746946201e-05, + "loss": 1.2415, + "step": 117910 + }, + { + "epoch": 10.21, + "learning_rate": 3.9827601143550205e-05, + "loss": 1.2352, + "step": 117920 + }, + { + "epoch": 10.21, + "learning_rate": 3.982673481763839e-05, + "loss": 1.2768, + "step": 117930 + }, + { + "epoch": 10.21, + "learning_rate": 3.982586849172659e-05, + "loss": 1.2881, + "step": 117940 + }, + { + "epoch": 10.21, + "learning_rate": 3.982500216581478e-05, + "loss": 1.2768, + "step": 117950 + }, + { + "epoch": 10.21, + "learning_rate": 3.982413583990297e-05, + "loss": 1.2549, + "step": 117960 + }, + { + "epoch": 10.21, + "learning_rate": 3.9823269513991163e-05, + "loss": 1.26, + "step": 117970 + }, + { + "epoch": 10.21, + "learning_rate": 3.982240318807936e-05, + "loss": 1.2357, + "step": 117980 + }, + { + "epoch": 10.21, + "learning_rate": 3.9821536862167546e-05, + "loss": 1.2487, + "step": 117990 + }, + { + "epoch": 10.21, + "learning_rate": 3.982067053625574e-05, + "loss": 1.1738, + "step": 118000 + }, + { + "epoch": 10.21, + "learning_rate": 3.981980421034393e-05, + "loss": 1.2481, + "step": 118010 + }, + { + "epoch": 10.22, + "learning_rate": 3.981893788443212e-05, + "loss": 1.1448, + "step": 118020 + }, + { + "epoch": 10.22, + "learning_rate": 3.9818071558520317e-05, + "loss": 1.2104, + "step": 118030 + }, + { + "epoch": 10.22, + "learning_rate": 3.9817205232608504e-05, + "loss": 1.2057, + "step": 118040 + }, + { + "epoch": 10.22, + "learning_rate": 3.98163389066967e-05, + "loss": 1.2406, + "step": 118050 + }, + { + "epoch": 10.22, + "learning_rate": 3.981547258078489e-05, + "loss": 1.2181, + "step": 118060 + }, + { + "epoch": 10.22, + "learning_rate": 3.981460625487308e-05, + "loss": 1.1885, + "step": 118070 + }, + { + "epoch": 10.22, + "learning_rate": 3.9813739928961275e-05, + "loss": 1.1741, + "step": 118080 + }, + { + "epoch": 10.22, + "learning_rate": 3.981287360304947e-05, + "loss": 1.2478, + "step": 118090 + }, + { + "epoch": 10.22, + "learning_rate": 3.981200727713766e-05, + "loss": 1.2394, + "step": 118100 + }, + { + "epoch": 10.22, + "learning_rate": 3.981114095122585e-05, + "loss": 1.2498, + "step": 118110 + }, + { + "epoch": 10.22, + "learning_rate": 3.981027462531404e-05, + "loss": 1.2143, + "step": 118120 + }, + { + "epoch": 10.22, + "learning_rate": 3.9809408299402234e-05, + "loss": 1.1949, + "step": 118130 + }, + { + "epoch": 10.23, + "learning_rate": 3.980854197349043e-05, + "loss": 1.217, + "step": 118140 + }, + { + "epoch": 10.23, + "learning_rate": 3.9807675647578616e-05, + "loss": 1.2507, + "step": 118150 + }, + { + "epoch": 10.23, + "learning_rate": 3.980680932166681e-05, + "loss": 1.2068, + "step": 118160 + }, + { + "epoch": 10.23, + "learning_rate": 3.9805942995755005e-05, + "loss": 1.2625, + "step": 118170 + }, + { + "epoch": 10.23, + "learning_rate": 3.980507666984319e-05, + "loss": 1.2331, + "step": 118180 + }, + { + "epoch": 10.23, + "learning_rate": 3.980421034393139e-05, + "loss": 1.2632, + "step": 118190 + }, + { + "epoch": 10.23, + "learning_rate": 3.9803344018019575e-05, + "loss": 1.2144, + "step": 118200 + }, + { + "epoch": 10.23, + "learning_rate": 3.980247769210777e-05, + "loss": 1.1878, + "step": 118210 + }, + { + "epoch": 10.23, + "learning_rate": 3.9801611366195964e-05, + "loss": 1.2309, + "step": 118220 + }, + { + "epoch": 10.23, + "learning_rate": 3.980074504028415e-05, + "loss": 1.2302, + "step": 118230 + }, + { + "epoch": 10.23, + "learning_rate": 3.9799878714372346e-05, + "loss": 1.2316, + "step": 118240 + }, + { + "epoch": 10.24, + "learning_rate": 3.979901238846054e-05, + "loss": 1.2941, + "step": 118250 + }, + { + "epoch": 10.24, + "learning_rate": 3.979814606254873e-05, + "loss": 1.2391, + "step": 118260 + }, + { + "epoch": 10.24, + "learning_rate": 3.979727973663692e-05, + "loss": 1.2503, + "step": 118270 + }, + { + "epoch": 10.24, + "learning_rate": 3.979641341072512e-05, + "loss": 1.1823, + "step": 118280 + }, + { + "epoch": 10.24, + "learning_rate": 3.9795547084813305e-05, + "loss": 1.2664, + "step": 118290 + }, + { + "epoch": 10.24, + "learning_rate": 3.97946807589015e-05, + "loss": 1.1962, + "step": 118300 + }, + { + "epoch": 10.24, + "learning_rate": 3.979381443298969e-05, + "loss": 1.2855, + "step": 118310 + }, + { + "epoch": 10.24, + "learning_rate": 3.979294810707788e-05, + "loss": 1.2402, + "step": 118320 + }, + { + "epoch": 10.24, + "learning_rate": 3.9792081781166076e-05, + "loss": 1.2152, + "step": 118330 + }, + { + "epoch": 10.24, + "learning_rate": 3.979121545525426e-05, + "loss": 1.2434, + "step": 118340 + }, + { + "epoch": 10.24, + "learning_rate": 3.979034912934246e-05, + "loss": 1.1964, + "step": 118350 + }, + { + "epoch": 10.24, + "learning_rate": 3.978948280343065e-05, + "loss": 1.1974, + "step": 118360 + }, + { + "epoch": 10.25, + "learning_rate": 3.978861647751884e-05, + "loss": 1.236, + "step": 118370 + }, + { + "epoch": 10.25, + "learning_rate": 3.9787750151607034e-05, + "loss": 1.2253, + "step": 118380 + }, + { + "epoch": 10.25, + "learning_rate": 3.978688382569523e-05, + "loss": 1.1599, + "step": 118390 + }, + { + "epoch": 10.25, + "learning_rate": 3.9786017499783416e-05, + "loss": 1.2449, + "step": 118400 + }, + { + "epoch": 10.25, + "learning_rate": 3.978515117387161e-05, + "loss": 1.2208, + "step": 118410 + }, + { + "epoch": 10.25, + "learning_rate": 3.97842848479598e-05, + "loss": 1.2011, + "step": 118420 + }, + { + "epoch": 10.25, + "learning_rate": 3.978341852204799e-05, + "loss": 1.279, + "step": 118430 + }, + { + "epoch": 10.25, + "learning_rate": 3.978255219613619e-05, + "loss": 1.1611, + "step": 118440 + }, + { + "epoch": 10.25, + "learning_rate": 3.9781685870224375e-05, + "loss": 1.2352, + "step": 118450 + }, + { + "epoch": 10.25, + "learning_rate": 3.978081954431257e-05, + "loss": 1.2283, + "step": 118460 + }, + { + "epoch": 10.25, + "learning_rate": 3.9779953218400764e-05, + "loss": 1.2754, + "step": 118470 + }, + { + "epoch": 10.26, + "learning_rate": 3.977908689248895e-05, + "loss": 1.2651, + "step": 118480 + }, + { + "epoch": 10.26, + "learning_rate": 3.9778220566577146e-05, + "loss": 1.2664, + "step": 118490 + }, + { + "epoch": 10.26, + "learning_rate": 3.977735424066534e-05, + "loss": 1.1999, + "step": 118500 + }, + { + "epoch": 10.26, + "learning_rate": 3.977648791475353e-05, + "loss": 1.2447, + "step": 118510 + }, + { + "epoch": 10.26, + "learning_rate": 3.977562158884172e-05, + "loss": 1.1894, + "step": 118520 + }, + { + "epoch": 10.26, + "learning_rate": 3.977475526292991e-05, + "loss": 1.2946, + "step": 118530 + }, + { + "epoch": 10.26, + "learning_rate": 3.9773888937018105e-05, + "loss": 1.2487, + "step": 118540 + }, + { + "epoch": 10.26, + "learning_rate": 3.97730226111063e-05, + "loss": 1.2432, + "step": 118550 + }, + { + "epoch": 10.26, + "learning_rate": 3.977215628519449e-05, + "loss": 1.1566, + "step": 118560 + }, + { + "epoch": 10.26, + "learning_rate": 3.977128995928268e-05, + "loss": 1.227, + "step": 118570 + }, + { + "epoch": 10.26, + "learning_rate": 3.9770423633370876e-05, + "loss": 1.1845, + "step": 118580 + }, + { + "epoch": 10.26, + "learning_rate": 3.9769557307459064e-05, + "loss": 1.2327, + "step": 118590 + }, + { + "epoch": 10.27, + "learning_rate": 3.976869098154726e-05, + "loss": 1.1284, + "step": 118600 + }, + { + "epoch": 10.27, + "learning_rate": 3.976782465563545e-05, + "loss": 1.2699, + "step": 118610 + }, + { + "epoch": 10.27, + "learning_rate": 3.976695832972364e-05, + "loss": 1.243, + "step": 118620 + }, + { + "epoch": 10.27, + "learning_rate": 3.9766092003811835e-05, + "loss": 1.1901, + "step": 118630 + }, + { + "epoch": 10.27, + "learning_rate": 3.976522567790002e-05, + "loss": 1.2434, + "step": 118640 + }, + { + "epoch": 10.27, + "learning_rate": 3.976435935198822e-05, + "loss": 1.207, + "step": 118650 + }, + { + "epoch": 10.27, + "learning_rate": 3.976349302607641e-05, + "loss": 1.2459, + "step": 118660 + }, + { + "epoch": 10.27, + "learning_rate": 3.97626267001646e-05, + "loss": 1.2264, + "step": 118670 + }, + { + "epoch": 10.27, + "learning_rate": 3.9761760374252793e-05, + "loss": 1.2261, + "step": 118680 + }, + { + "epoch": 10.27, + "learning_rate": 3.976089404834099e-05, + "loss": 1.2065, + "step": 118690 + }, + { + "epoch": 10.27, + "learning_rate": 3.9760027722429176e-05, + "loss": 1.2575, + "step": 118700 + }, + { + "epoch": 10.28, + "learning_rate": 3.975916139651737e-05, + "loss": 1.2778, + "step": 118710 + }, + { + "epoch": 10.28, + "learning_rate": 3.9758295070605564e-05, + "loss": 1.2418, + "step": 118720 + }, + { + "epoch": 10.28, + "learning_rate": 3.975742874469375e-05, + "loss": 1.2399, + "step": 118730 + }, + { + "epoch": 10.28, + "learning_rate": 3.9756562418781947e-05, + "loss": 1.2158, + "step": 118740 + }, + { + "epoch": 10.28, + "learning_rate": 3.9755696092870134e-05, + "loss": 1.184, + "step": 118750 + }, + { + "epoch": 10.28, + "learning_rate": 3.975482976695833e-05, + "loss": 1.236, + "step": 118760 + }, + { + "epoch": 10.28, + "learning_rate": 3.975396344104652e-05, + "loss": 1.2337, + "step": 118770 + }, + { + "epoch": 10.28, + "learning_rate": 3.975309711513471e-05, + "loss": 1.2662, + "step": 118780 + }, + { + "epoch": 10.28, + "learning_rate": 3.9752230789222905e-05, + "loss": 1.236, + "step": 118790 + }, + { + "epoch": 10.28, + "learning_rate": 3.97513644633111e-05, + "loss": 1.1875, + "step": 118800 + }, + { + "epoch": 10.28, + "learning_rate": 3.975049813739929e-05, + "loss": 1.2472, + "step": 118810 + }, + { + "epoch": 10.28, + "learning_rate": 3.974963181148748e-05, + "loss": 1.1815, + "step": 118820 + }, + { + "epoch": 10.29, + "learning_rate": 3.9748765485575676e-05, + "loss": 1.2577, + "step": 118830 + }, + { + "epoch": 10.29, + "learning_rate": 3.9747899159663864e-05, + "loss": 1.2578, + "step": 118840 + }, + { + "epoch": 10.29, + "learning_rate": 3.974703283375206e-05, + "loss": 1.2461, + "step": 118850 + }, + { + "epoch": 10.29, + "learning_rate": 3.9746166507840246e-05, + "loss": 1.2005, + "step": 118860 + }, + { + "epoch": 10.29, + "learning_rate": 3.974530018192844e-05, + "loss": 1.23, + "step": 118870 + }, + { + "epoch": 10.29, + "learning_rate": 3.9744433856016635e-05, + "loss": 1.25, + "step": 118880 + }, + { + "epoch": 10.29, + "learning_rate": 3.974356753010482e-05, + "loss": 1.2095, + "step": 118890 + }, + { + "epoch": 10.29, + "learning_rate": 3.974270120419302e-05, + "loss": 1.2845, + "step": 118900 + }, + { + "epoch": 10.29, + "learning_rate": 3.974183487828121e-05, + "loss": 1.234, + "step": 118910 + }, + { + "epoch": 10.29, + "learning_rate": 3.97409685523694e-05, + "loss": 1.1905, + "step": 118920 + }, + { + "epoch": 10.29, + "learning_rate": 3.9740102226457594e-05, + "loss": 1.2381, + "step": 118930 + }, + { + "epoch": 10.29, + "learning_rate": 3.973923590054579e-05, + "loss": 1.2309, + "step": 118940 + }, + { + "epoch": 10.3, + "learning_rate": 3.9738369574633976e-05, + "loss": 1.198, + "step": 118950 + }, + { + "epoch": 10.3, + "learning_rate": 3.973750324872217e-05, + "loss": 1.2306, + "step": 118960 + }, + { + "epoch": 10.3, + "learning_rate": 3.973663692281036e-05, + "loss": 1.1503, + "step": 118970 + }, + { + "epoch": 10.3, + "learning_rate": 3.973577059689855e-05, + "loss": 1.2171, + "step": 118980 + }, + { + "epoch": 10.3, + "learning_rate": 3.973490427098675e-05, + "loss": 1.2574, + "step": 118990 + }, + { + "epoch": 10.3, + "learning_rate": 3.9734037945074935e-05, + "loss": 1.1844, + "step": 119000 + }, + { + "epoch": 10.3, + "learning_rate": 3.973317161916313e-05, + "loss": 1.2658, + "step": 119010 + }, + { + "epoch": 10.3, + "learning_rate": 3.9732305293251323e-05, + "loss": 1.235, + "step": 119020 + }, + { + "epoch": 10.3, + "learning_rate": 3.973143896733951e-05, + "loss": 1.2625, + "step": 119030 + }, + { + "epoch": 10.3, + "learning_rate": 3.9730572641427706e-05, + "loss": 1.2452, + "step": 119040 + }, + { + "epoch": 10.3, + "learning_rate": 3.972970631551589e-05, + "loss": 1.2459, + "step": 119050 + }, + { + "epoch": 10.31, + "learning_rate": 3.972883998960409e-05, + "loss": 1.204, + "step": 119060 + }, + { + "epoch": 10.31, + "learning_rate": 3.972797366369228e-05, + "loss": 1.2507, + "step": 119070 + }, + { + "epoch": 10.31, + "learning_rate": 3.972710733778047e-05, + "loss": 1.249, + "step": 119080 + }, + { + "epoch": 10.31, + "learning_rate": 3.9726241011868664e-05, + "loss": 1.2683, + "step": 119090 + }, + { + "epoch": 10.31, + "learning_rate": 3.972537468595686e-05, + "loss": 1.241, + "step": 119100 + }, + { + "epoch": 10.31, + "learning_rate": 3.9724508360045046e-05, + "loss": 1.2249, + "step": 119110 + }, + { + "epoch": 10.31, + "learning_rate": 3.972364203413324e-05, + "loss": 1.2224, + "step": 119120 + }, + { + "epoch": 10.31, + "learning_rate": 3.9722775708221435e-05, + "loss": 1.2403, + "step": 119130 + }, + { + "epoch": 10.31, + "learning_rate": 3.972190938230962e-05, + "loss": 1.2387, + "step": 119140 + }, + { + "epoch": 10.31, + "learning_rate": 3.972104305639782e-05, + "loss": 1.1943, + "step": 119150 + }, + { + "epoch": 10.31, + "learning_rate": 3.9720176730486005e-05, + "loss": 1.2207, + "step": 119160 + }, + { + "epoch": 10.31, + "learning_rate": 3.97193104045742e-05, + "loss": 1.2351, + "step": 119170 + }, + { + "epoch": 10.32, + "learning_rate": 3.9718444078662394e-05, + "loss": 1.2675, + "step": 119180 + }, + { + "epoch": 10.32, + "learning_rate": 3.971757775275058e-05, + "loss": 1.2693, + "step": 119190 + }, + { + "epoch": 10.32, + "learning_rate": 3.9716711426838776e-05, + "loss": 1.2578, + "step": 119200 + }, + { + "epoch": 10.32, + "learning_rate": 3.971584510092697e-05, + "loss": 1.2318, + "step": 119210 + }, + { + "epoch": 10.32, + "learning_rate": 3.971497877501516e-05, + "loss": 1.2238, + "step": 119220 + }, + { + "epoch": 10.32, + "learning_rate": 3.971411244910335e-05, + "loss": 1.1999, + "step": 119230 + }, + { + "epoch": 10.32, + "learning_rate": 3.971324612319155e-05, + "loss": 1.2792, + "step": 119240 + }, + { + "epoch": 10.32, + "learning_rate": 3.9712379797279735e-05, + "loss": 1.3184, + "step": 119250 + }, + { + "epoch": 10.32, + "learning_rate": 3.971151347136793e-05, + "loss": 1.2682, + "step": 119260 + }, + { + "epoch": 10.32, + "learning_rate": 3.971064714545612e-05, + "loss": 1.2115, + "step": 119270 + }, + { + "epoch": 10.32, + "learning_rate": 3.970978081954431e-05, + "loss": 1.1788, + "step": 119280 + }, + { + "epoch": 10.33, + "learning_rate": 3.9708914493632506e-05, + "loss": 1.2055, + "step": 119290 + }, + { + "epoch": 10.33, + "learning_rate": 3.9708048167720694e-05, + "loss": 1.2397, + "step": 119300 + }, + { + "epoch": 10.33, + "learning_rate": 3.970718184180889e-05, + "loss": 1.1832, + "step": 119310 + }, + { + "epoch": 10.33, + "learning_rate": 3.970631551589708e-05, + "loss": 1.1953, + "step": 119320 + }, + { + "epoch": 10.33, + "learning_rate": 3.970544918998527e-05, + "loss": 1.2077, + "step": 119330 + }, + { + "epoch": 10.33, + "learning_rate": 3.9704582864073465e-05, + "loss": 1.166, + "step": 119340 + }, + { + "epoch": 10.33, + "learning_rate": 3.970371653816166e-05, + "loss": 1.203, + "step": 119350 + }, + { + "epoch": 10.33, + "learning_rate": 3.970285021224985e-05, + "loss": 1.1989, + "step": 119360 + }, + { + "epoch": 10.33, + "learning_rate": 3.970198388633804e-05, + "loss": 1.2021, + "step": 119370 + }, + { + "epoch": 10.33, + "learning_rate": 3.970111756042623e-05, + "loss": 1.2194, + "step": 119380 + }, + { + "epoch": 10.33, + "learning_rate": 3.970025123451442e-05, + "loss": 1.2295, + "step": 119390 + }, + { + "epoch": 10.33, + "learning_rate": 3.969938490860262e-05, + "loss": 1.1707, + "step": 119400 + }, + { + "epoch": 10.34, + "learning_rate": 3.9698518582690806e-05, + "loss": 1.1883, + "step": 119410 + }, + { + "epoch": 10.34, + "learning_rate": 3.9697652256779e-05, + "loss": 1.3355, + "step": 119420 + }, + { + "epoch": 10.34, + "learning_rate": 3.9696785930867194e-05, + "loss": 1.2375, + "step": 119430 + }, + { + "epoch": 10.34, + "learning_rate": 3.969591960495538e-05, + "loss": 1.2183, + "step": 119440 + }, + { + "epoch": 10.34, + "learning_rate": 3.9695053279043577e-05, + "loss": 1.181, + "step": 119450 + }, + { + "epoch": 10.34, + "learning_rate": 3.969418695313177e-05, + "loss": 1.2569, + "step": 119460 + }, + { + "epoch": 10.34, + "learning_rate": 3.969332062721996e-05, + "loss": 1.1849, + "step": 119470 + }, + { + "epoch": 10.34, + "learning_rate": 3.969245430130815e-05, + "loss": 1.2189, + "step": 119480 + }, + { + "epoch": 10.34, + "learning_rate": 3.969158797539634e-05, + "loss": 1.1676, + "step": 119490 + }, + { + "epoch": 10.34, + "learning_rate": 3.9690721649484535e-05, + "loss": 1.1618, + "step": 119500 + }, + { + "epoch": 10.34, + "learning_rate": 3.968985532357273e-05, + "loss": 1.2276, + "step": 119510 + }, + { + "epoch": 10.35, + "learning_rate": 3.968898899766092e-05, + "loss": 1.2064, + "step": 119520 + }, + { + "epoch": 10.35, + "learning_rate": 3.968812267174911e-05, + "loss": 1.2493, + "step": 119530 + }, + { + "epoch": 10.35, + "learning_rate": 3.9687256345837306e-05, + "loss": 1.209, + "step": 119540 + }, + { + "epoch": 10.35, + "learning_rate": 3.9686390019925494e-05, + "loss": 1.179, + "step": 119550 + }, + { + "epoch": 10.35, + "learning_rate": 3.968552369401369e-05, + "loss": 1.3034, + "step": 119560 + }, + { + "epoch": 10.35, + "learning_rate": 3.968465736810188e-05, + "loss": 1.2303, + "step": 119570 + }, + { + "epoch": 10.35, + "learning_rate": 3.968379104219007e-05, + "loss": 1.2321, + "step": 119580 + }, + { + "epoch": 10.35, + "learning_rate": 3.9682924716278265e-05, + "loss": 1.1978, + "step": 119590 + }, + { + "epoch": 10.35, + "learning_rate": 3.968205839036645e-05, + "loss": 1.1264, + "step": 119600 + }, + { + "epoch": 10.35, + "learning_rate": 3.968119206445465e-05, + "loss": 1.2381, + "step": 119610 + }, + { + "epoch": 10.35, + "learning_rate": 3.968032573854284e-05, + "loss": 1.1751, + "step": 119620 + }, + { + "epoch": 10.35, + "learning_rate": 3.967945941263103e-05, + "loss": 1.1952, + "step": 119630 + }, + { + "epoch": 10.36, + "learning_rate": 3.9678593086719224e-05, + "loss": 1.1727, + "step": 119640 + }, + { + "epoch": 10.36, + "learning_rate": 3.967772676080742e-05, + "loss": 1.29, + "step": 119650 + }, + { + "epoch": 10.36, + "learning_rate": 3.9676860434895606e-05, + "loss": 1.1321, + "step": 119660 + }, + { + "epoch": 10.36, + "learning_rate": 3.96759941089838e-05, + "loss": 1.1894, + "step": 119670 + }, + { + "epoch": 10.36, + "learning_rate": 3.9675127783071995e-05, + "loss": 1.1358, + "step": 119680 + }, + { + "epoch": 10.36, + "learning_rate": 3.967426145716018e-05, + "loss": 1.185, + "step": 119690 + }, + { + "epoch": 10.36, + "learning_rate": 3.967339513124838e-05, + "loss": 1.1823, + "step": 119700 + }, + { + "epoch": 10.36, + "learning_rate": 3.9672528805336565e-05, + "loss": 1.2223, + "step": 119710 + }, + { + "epoch": 10.36, + "learning_rate": 3.967166247942476e-05, + "loss": 1.1933, + "step": 119720 + }, + { + "epoch": 10.36, + "learning_rate": 3.9670796153512953e-05, + "loss": 1.223, + "step": 119730 + }, + { + "epoch": 10.36, + "learning_rate": 3.966992982760114e-05, + "loss": 1.2459, + "step": 119740 + }, + { + "epoch": 10.37, + "learning_rate": 3.9669063501689336e-05, + "loss": 1.1928, + "step": 119750 + }, + { + "epoch": 10.37, + "learning_rate": 3.966819717577753e-05, + "loss": 1.2159, + "step": 119760 + }, + { + "epoch": 10.37, + "learning_rate": 3.966733084986572e-05, + "loss": 1.2087, + "step": 119770 + }, + { + "epoch": 10.37, + "learning_rate": 3.966646452395391e-05, + "loss": 1.227, + "step": 119780 + }, + { + "epoch": 10.37, + "learning_rate": 3.96655981980421e-05, + "loss": 1.2728, + "step": 119790 + }, + { + "epoch": 10.37, + "learning_rate": 3.9664731872130294e-05, + "loss": 1.2251, + "step": 119800 + }, + { + "epoch": 10.37, + "learning_rate": 3.966386554621849e-05, + "loss": 1.2512, + "step": 119810 + }, + { + "epoch": 10.37, + "learning_rate": 3.9662999220306676e-05, + "loss": 1.2613, + "step": 119820 + }, + { + "epoch": 10.37, + "learning_rate": 3.966213289439487e-05, + "loss": 1.2314, + "step": 119830 + }, + { + "epoch": 10.37, + "learning_rate": 3.9661266568483065e-05, + "loss": 1.2025, + "step": 119840 + }, + { + "epoch": 10.37, + "learning_rate": 3.966040024257125e-05, + "loss": 1.2171, + "step": 119850 + }, + { + "epoch": 10.37, + "learning_rate": 3.965953391665945e-05, + "loss": 1.275, + "step": 119860 + }, + { + "epoch": 10.38, + "learning_rate": 3.965866759074764e-05, + "loss": 1.2101, + "step": 119870 + }, + { + "epoch": 10.38, + "learning_rate": 3.965780126483583e-05, + "loss": 1.2509, + "step": 119880 + }, + { + "epoch": 10.38, + "learning_rate": 3.9656934938924024e-05, + "loss": 1.2445, + "step": 119890 + }, + { + "epoch": 10.38, + "learning_rate": 3.965606861301221e-05, + "loss": 1.2292, + "step": 119900 + }, + { + "epoch": 10.38, + "learning_rate": 3.9655202287100406e-05, + "loss": 1.2266, + "step": 119910 + }, + { + "epoch": 10.38, + "learning_rate": 3.96543359611886e-05, + "loss": 1.2045, + "step": 119920 + }, + { + "epoch": 10.38, + "learning_rate": 3.965346963527679e-05, + "loss": 1.2168, + "step": 119930 + }, + { + "epoch": 10.38, + "learning_rate": 3.965260330936498e-05, + "loss": 1.209, + "step": 119940 + }, + { + "epoch": 10.38, + "learning_rate": 3.965173698345318e-05, + "loss": 1.2157, + "step": 119950 + }, + { + "epoch": 10.38, + "learning_rate": 3.9650870657541365e-05, + "loss": 1.2525, + "step": 119960 + }, + { + "epoch": 10.38, + "learning_rate": 3.965000433162956e-05, + "loss": 1.223, + "step": 119970 + }, + { + "epoch": 10.38, + "learning_rate": 3.9649138005717754e-05, + "loss": 1.2019, + "step": 119980 + }, + { + "epoch": 10.39, + "learning_rate": 3.964827167980594e-05, + "loss": 1.2493, + "step": 119990 + }, + { + "epoch": 10.39, + "learning_rate": 3.9647405353894136e-05, + "loss": 1.307, + "step": 120000 + }, + { + "epoch": 10.39, + "learning_rate": 3.9646539027982324e-05, + "loss": 1.1762, + "step": 120010 + }, + { + "epoch": 10.39, + "learning_rate": 3.964567270207052e-05, + "loss": 1.2662, + "step": 120020 + }, + { + "epoch": 10.39, + "learning_rate": 3.964480637615871e-05, + "loss": 1.2393, + "step": 120030 + }, + { + "epoch": 10.39, + "learning_rate": 3.96439400502469e-05, + "loss": 1.2558, + "step": 120040 + }, + { + "epoch": 10.39, + "learning_rate": 3.9643073724335095e-05, + "loss": 1.2257, + "step": 120050 + }, + { + "epoch": 10.39, + "learning_rate": 3.964220739842329e-05, + "loss": 1.2219, + "step": 120060 + }, + { + "epoch": 10.39, + "learning_rate": 3.964134107251148e-05, + "loss": 1.2298, + "step": 120070 + }, + { + "epoch": 10.39, + "learning_rate": 3.964047474659967e-05, + "loss": 1.2017, + "step": 120080 + }, + { + "epoch": 10.39, + "learning_rate": 3.9639608420687866e-05, + "loss": 1.1954, + "step": 120090 + }, + { + "epoch": 10.4, + "learning_rate": 3.963874209477605e-05, + "loss": 1.177, + "step": 120100 + }, + { + "epoch": 10.4, + "learning_rate": 3.963787576886425e-05, + "loss": 1.1643, + "step": 120110 + }, + { + "epoch": 10.4, + "learning_rate": 3.9637009442952435e-05, + "loss": 1.246, + "step": 120120 + }, + { + "epoch": 10.4, + "learning_rate": 3.963614311704063e-05, + "loss": 1.209, + "step": 120130 + }, + { + "epoch": 10.4, + "learning_rate": 3.9635276791128824e-05, + "loss": 1.1641, + "step": 120140 + }, + { + "epoch": 10.4, + "learning_rate": 3.963441046521701e-05, + "loss": 1.2404, + "step": 120150 + }, + { + "epoch": 10.4, + "learning_rate": 3.9633544139305207e-05, + "loss": 1.2398, + "step": 120160 + }, + { + "epoch": 10.4, + "learning_rate": 3.96326778133934e-05, + "loss": 1.2063, + "step": 120170 + }, + { + "epoch": 10.4, + "learning_rate": 3.963181148748159e-05, + "loss": 1.2248, + "step": 120180 + }, + { + "epoch": 10.4, + "learning_rate": 3.963094516156978e-05, + "loss": 1.2066, + "step": 120190 + }, + { + "epoch": 10.4, + "learning_rate": 3.963007883565798e-05, + "loss": 1.2532, + "step": 120200 + }, + { + "epoch": 10.4, + "learning_rate": 3.9629212509746165e-05, + "loss": 1.195, + "step": 120210 + }, + { + "epoch": 10.41, + "learning_rate": 3.962834618383436e-05, + "loss": 1.2009, + "step": 120220 + }, + { + "epoch": 10.41, + "learning_rate": 3.962747985792255e-05, + "loss": 1.198, + "step": 120230 + }, + { + "epoch": 10.41, + "learning_rate": 3.962661353201074e-05, + "loss": 1.1665, + "step": 120240 + }, + { + "epoch": 10.41, + "learning_rate": 3.9625747206098936e-05, + "loss": 1.1933, + "step": 120250 + }, + { + "epoch": 10.41, + "learning_rate": 3.9624880880187124e-05, + "loss": 1.2284, + "step": 120260 + }, + { + "epoch": 10.41, + "learning_rate": 3.962401455427532e-05, + "loss": 1.2063, + "step": 120270 + }, + { + "epoch": 10.41, + "learning_rate": 3.962314822836351e-05, + "loss": 1.1864, + "step": 120280 + }, + { + "epoch": 10.41, + "learning_rate": 3.96222819024517e-05, + "loss": 1.1834, + "step": 120290 + }, + { + "epoch": 10.41, + "learning_rate": 3.9621415576539895e-05, + "loss": 1.2415, + "step": 120300 + }, + { + "epoch": 10.41, + "learning_rate": 3.962054925062809e-05, + "loss": 1.2641, + "step": 120310 + }, + { + "epoch": 10.41, + "learning_rate": 3.961968292471628e-05, + "loss": 1.2424, + "step": 120320 + }, + { + "epoch": 10.42, + "learning_rate": 3.961881659880447e-05, + "loss": 1.2419, + "step": 120330 + }, + { + "epoch": 10.42, + "learning_rate": 3.961795027289266e-05, + "loss": 1.2111, + "step": 120340 + }, + { + "epoch": 10.42, + "learning_rate": 3.9617083946980854e-05, + "loss": 1.2909, + "step": 120350 + }, + { + "epoch": 10.42, + "learning_rate": 3.961621762106905e-05, + "loss": 1.2259, + "step": 120360 + }, + { + "epoch": 10.42, + "learning_rate": 3.9615351295157236e-05, + "loss": 1.1969, + "step": 120370 + }, + { + "epoch": 10.42, + "learning_rate": 3.961448496924543e-05, + "loss": 1.1848, + "step": 120380 + }, + { + "epoch": 10.42, + "learning_rate": 3.9613618643333625e-05, + "loss": 1.2684, + "step": 120390 + }, + { + "epoch": 10.42, + "learning_rate": 3.961275231742181e-05, + "loss": 1.2511, + "step": 120400 + }, + { + "epoch": 10.42, + "learning_rate": 3.961188599151001e-05, + "loss": 1.2456, + "step": 120410 + }, + { + "epoch": 10.42, + "learning_rate": 3.96110196655982e-05, + "loss": 1.2397, + "step": 120420 + }, + { + "epoch": 10.42, + "learning_rate": 3.961015333968639e-05, + "loss": 1.2611, + "step": 120430 + }, + { + "epoch": 10.42, + "learning_rate": 3.9609287013774583e-05, + "loss": 1.2394, + "step": 120440 + }, + { + "epoch": 10.43, + "learning_rate": 3.960842068786277e-05, + "loss": 1.2484, + "step": 120450 + }, + { + "epoch": 10.43, + "learning_rate": 3.9607554361950966e-05, + "loss": 1.2199, + "step": 120460 + }, + { + "epoch": 10.43, + "learning_rate": 3.960668803603916e-05, + "loss": 1.2059, + "step": 120470 + }, + { + "epoch": 10.43, + "learning_rate": 3.960582171012735e-05, + "loss": 1.1954, + "step": 120480 + }, + { + "epoch": 10.43, + "learning_rate": 3.960495538421554e-05, + "loss": 1.2213, + "step": 120490 + }, + { + "epoch": 10.43, + "learning_rate": 3.9604089058303737e-05, + "loss": 1.2142, + "step": 120500 + }, + { + "epoch": 10.43, + "learning_rate": 3.9603222732391924e-05, + "loss": 1.2763, + "step": 120510 + }, + { + "epoch": 10.43, + "learning_rate": 3.960235640648012e-05, + "loss": 1.2693, + "step": 120520 + }, + { + "epoch": 10.43, + "learning_rate": 3.9601490080568306e-05, + "loss": 1.2718, + "step": 120530 + }, + { + "epoch": 10.43, + "learning_rate": 3.96006237546565e-05, + "loss": 1.1725, + "step": 120540 + }, + { + "epoch": 10.43, + "learning_rate": 3.9599757428744695e-05, + "loss": 1.1849, + "step": 120550 + }, + { + "epoch": 10.44, + "learning_rate": 3.959889110283288e-05, + "loss": 1.1994, + "step": 120560 + }, + { + "epoch": 10.44, + "learning_rate": 3.959802477692108e-05, + "loss": 1.2148, + "step": 120570 + }, + { + "epoch": 10.44, + "learning_rate": 3.959715845100927e-05, + "loss": 1.1977, + "step": 120580 + }, + { + "epoch": 10.44, + "learning_rate": 3.959629212509746e-05, + "loss": 1.1601, + "step": 120590 + }, + { + "epoch": 10.44, + "learning_rate": 3.9595425799185654e-05, + "loss": 1.2595, + "step": 120600 + }, + { + "epoch": 10.44, + "learning_rate": 3.959455947327385e-05, + "loss": 1.1624, + "step": 120610 + }, + { + "epoch": 10.44, + "learning_rate": 3.9593693147362036e-05, + "loss": 1.2197, + "step": 120620 + }, + { + "epoch": 10.44, + "learning_rate": 3.959282682145023e-05, + "loss": 1.2236, + "step": 120630 + }, + { + "epoch": 10.44, + "learning_rate": 3.959196049553842e-05, + "loss": 1.2103, + "step": 120640 + }, + { + "epoch": 10.44, + "learning_rate": 3.959109416962661e-05, + "loss": 1.1707, + "step": 120650 + }, + { + "epoch": 10.44, + "learning_rate": 3.959022784371481e-05, + "loss": 1.2248, + "step": 120660 + }, + { + "epoch": 10.44, + "learning_rate": 3.9589361517802995e-05, + "loss": 1.1977, + "step": 120670 + }, + { + "epoch": 10.45, + "learning_rate": 3.958849519189119e-05, + "loss": 1.1906, + "step": 120680 + }, + { + "epoch": 10.45, + "learning_rate": 3.9587628865979384e-05, + "loss": 1.2083, + "step": 120690 + }, + { + "epoch": 10.45, + "learning_rate": 3.958676254006757e-05, + "loss": 1.1638, + "step": 120700 + }, + { + "epoch": 10.45, + "learning_rate": 3.9585896214155766e-05, + "loss": 1.2, + "step": 120710 + }, + { + "epoch": 10.45, + "learning_rate": 3.958502988824396e-05, + "loss": 1.2555, + "step": 120720 + }, + { + "epoch": 10.45, + "learning_rate": 3.958416356233215e-05, + "loss": 1.2415, + "step": 120730 + }, + { + "epoch": 10.45, + "learning_rate": 3.958329723642034e-05, + "loss": 1.254, + "step": 120740 + }, + { + "epoch": 10.45, + "learning_rate": 3.958243091050853e-05, + "loss": 1.192, + "step": 120750 + }, + { + "epoch": 10.45, + "learning_rate": 3.9581564584596725e-05, + "loss": 1.2577, + "step": 120760 + }, + { + "epoch": 10.45, + "learning_rate": 3.958069825868492e-05, + "loss": 1.2166, + "step": 120770 + }, + { + "epoch": 10.45, + "learning_rate": 3.957983193277311e-05, + "loss": 1.2503, + "step": 120780 + }, + { + "epoch": 10.46, + "learning_rate": 3.95789656068613e-05, + "loss": 1.2456, + "step": 120790 + }, + { + "epoch": 10.46, + "learning_rate": 3.9578099280949496e-05, + "loss": 1.2127, + "step": 120800 + }, + { + "epoch": 10.46, + "learning_rate": 3.957723295503768e-05, + "loss": 1.1845, + "step": 120810 + }, + { + "epoch": 10.46, + "learning_rate": 3.957636662912588e-05, + "loss": 1.2733, + "step": 120820 + }, + { + "epoch": 10.46, + "learning_rate": 3.957550030321407e-05, + "loss": 1.2233, + "step": 120830 + }, + { + "epoch": 10.46, + "learning_rate": 3.957463397730226e-05, + "loss": 1.2231, + "step": 120840 + }, + { + "epoch": 10.46, + "learning_rate": 3.9573767651390454e-05, + "loss": 1.2348, + "step": 120850 + }, + { + "epoch": 10.46, + "learning_rate": 3.957290132547864e-05, + "loss": 1.1861, + "step": 120860 + }, + { + "epoch": 10.46, + "learning_rate": 3.9572034999566836e-05, + "loss": 1.1947, + "step": 120870 + }, + { + "epoch": 10.46, + "learning_rate": 3.957116867365503e-05, + "loss": 1.2192, + "step": 120880 + }, + { + "epoch": 10.46, + "learning_rate": 3.957030234774322e-05, + "loss": 1.2539, + "step": 120890 + }, + { + "epoch": 10.46, + "learning_rate": 3.956943602183141e-05, + "loss": 1.2339, + "step": 120900 + }, + { + "epoch": 10.47, + "learning_rate": 3.956856969591961e-05, + "loss": 1.1608, + "step": 120910 + }, + { + "epoch": 10.47, + "learning_rate": 3.9567703370007795e-05, + "loss": 1.2009, + "step": 120920 + }, + { + "epoch": 10.47, + "learning_rate": 3.956683704409599e-05, + "loss": 1.1757, + "step": 120930 + }, + { + "epoch": 10.47, + "learning_rate": 3.9565970718184184e-05, + "loss": 1.2516, + "step": 120940 + }, + { + "epoch": 10.47, + "learning_rate": 3.956510439227237e-05, + "loss": 1.2897, + "step": 120950 + }, + { + "epoch": 10.47, + "learning_rate": 3.9564238066360566e-05, + "loss": 1.243, + "step": 120960 + }, + { + "epoch": 10.47, + "learning_rate": 3.9563371740448754e-05, + "loss": 1.2174, + "step": 120970 + }, + { + "epoch": 10.47, + "learning_rate": 3.956250541453695e-05, + "loss": 1.2541, + "step": 120980 + }, + { + "epoch": 10.47, + "learning_rate": 3.956163908862514e-05, + "loss": 1.2024, + "step": 120990 + }, + { + "epoch": 10.47, + "learning_rate": 3.956077276271333e-05, + "loss": 1.2252, + "step": 121000 + }, + { + "epoch": 10.47, + "learning_rate": 3.9559906436801525e-05, + "loss": 1.2694, + "step": 121010 + }, + { + "epoch": 10.48, + "learning_rate": 3.955904011088972e-05, + "loss": 1.2726, + "step": 121020 + }, + { + "epoch": 10.48, + "learning_rate": 3.955817378497791e-05, + "loss": 1.2765, + "step": 121030 + }, + { + "epoch": 10.48, + "learning_rate": 3.95573074590661e-05, + "loss": 1.182, + "step": 121040 + }, + { + "epoch": 10.48, + "learning_rate": 3.9556441133154296e-05, + "loss": 1.3037, + "step": 121050 + }, + { + "epoch": 10.48, + "learning_rate": 3.9555574807242484e-05, + "loss": 1.2583, + "step": 121060 + }, + { + "epoch": 10.48, + "learning_rate": 3.955470848133068e-05, + "loss": 1.2023, + "step": 121070 + }, + { + "epoch": 10.48, + "learning_rate": 3.9553842155418866e-05, + "loss": 1.2442, + "step": 121080 + }, + { + "epoch": 10.48, + "learning_rate": 3.955297582950706e-05, + "loss": 1.2055, + "step": 121090 + }, + { + "epoch": 10.48, + "learning_rate": 3.9552109503595255e-05, + "loss": 1.2842, + "step": 121100 + }, + { + "epoch": 10.48, + "learning_rate": 3.955124317768344e-05, + "loss": 1.1756, + "step": 121110 + }, + { + "epoch": 10.48, + "learning_rate": 3.955037685177164e-05, + "loss": 1.2388, + "step": 121120 + }, + { + "epoch": 10.48, + "learning_rate": 3.954951052585983e-05, + "loss": 1.1583, + "step": 121130 + }, + { + "epoch": 10.49, + "learning_rate": 3.954864419994802e-05, + "loss": 1.1972, + "step": 121140 + }, + { + "epoch": 10.49, + "learning_rate": 3.9547777874036213e-05, + "loss": 1.2386, + "step": 121150 + }, + { + "epoch": 10.49, + "learning_rate": 3.954691154812441e-05, + "loss": 1.2521, + "step": 121160 + }, + { + "epoch": 10.49, + "learning_rate": 3.9546045222212596e-05, + "loss": 1.2718, + "step": 121170 + }, + { + "epoch": 10.49, + "learning_rate": 3.954517889630079e-05, + "loss": 1.1973, + "step": 121180 + }, + { + "epoch": 10.49, + "learning_rate": 3.954431257038898e-05, + "loss": 1.2138, + "step": 121190 + }, + { + "epoch": 10.49, + "learning_rate": 3.954344624447717e-05, + "loss": 1.275, + "step": 121200 + }, + { + "epoch": 10.49, + "learning_rate": 3.9542579918565367e-05, + "loss": 1.2025, + "step": 121210 + }, + { + "epoch": 10.49, + "learning_rate": 3.9541713592653554e-05, + "loss": 1.202, + "step": 121220 + }, + { + "epoch": 10.49, + "learning_rate": 3.954084726674175e-05, + "loss": 1.2365, + "step": 121230 + }, + { + "epoch": 10.49, + "learning_rate": 3.953998094082994e-05, + "loss": 1.2704, + "step": 121240 + }, + { + "epoch": 10.49, + "learning_rate": 3.953911461491813e-05, + "loss": 1.127, + "step": 121250 + }, + { + "epoch": 10.5, + "learning_rate": 3.9538248289006325e-05, + "loss": 1.2329, + "step": 121260 + }, + { + "epoch": 10.5, + "learning_rate": 3.953738196309452e-05, + "loss": 1.2306, + "step": 121270 + }, + { + "epoch": 10.5, + "learning_rate": 3.953651563718271e-05, + "loss": 1.1659, + "step": 121280 + }, + { + "epoch": 10.5, + "learning_rate": 3.95356493112709e-05, + "loss": 1.2099, + "step": 121290 + }, + { + "epoch": 10.5, + "learning_rate": 3.953478298535909e-05, + "loss": 1.1402, + "step": 121300 + }, + { + "epoch": 10.5, + "learning_rate": 3.9533916659447284e-05, + "loss": 1.2307, + "step": 121310 + }, + { + "epoch": 10.5, + "learning_rate": 3.953305033353548e-05, + "loss": 1.2003, + "step": 121320 + }, + { + "epoch": 10.5, + "learning_rate": 3.9532184007623666e-05, + "loss": 1.2111, + "step": 121330 + }, + { + "epoch": 10.5, + "learning_rate": 3.953131768171186e-05, + "loss": 1.2091, + "step": 121340 + }, + { + "epoch": 10.5, + "learning_rate": 3.9530451355800055e-05, + "loss": 1.2868, + "step": 121350 + }, + { + "epoch": 10.5, + "learning_rate": 3.952958502988824e-05, + "loss": 1.201, + "step": 121360 + }, + { + "epoch": 10.51, + "learning_rate": 3.952871870397644e-05, + "loss": 1.1283, + "step": 121370 + }, + { + "epoch": 10.51, + "learning_rate": 3.9527852378064625e-05, + "loss": 1.2005, + "step": 121380 + }, + { + "epoch": 10.51, + "learning_rate": 3.952698605215282e-05, + "loss": 1.2438, + "step": 121390 + }, + { + "epoch": 10.51, + "learning_rate": 3.9526119726241014e-05, + "loss": 1.1689, + "step": 121400 + }, + { + "epoch": 10.51, + "learning_rate": 3.95252534003292e-05, + "loss": 1.1983, + "step": 121410 + }, + { + "epoch": 10.51, + "learning_rate": 3.9524387074417396e-05, + "loss": 1.2552, + "step": 121420 + }, + { + "epoch": 10.51, + "learning_rate": 3.952352074850559e-05, + "loss": 1.2342, + "step": 121430 + }, + { + "epoch": 10.51, + "learning_rate": 3.952265442259378e-05, + "loss": 1.1615, + "step": 121440 + }, + { + "epoch": 10.51, + "learning_rate": 3.952178809668197e-05, + "loss": 1.2299, + "step": 121450 + }, + { + "epoch": 10.51, + "learning_rate": 3.952092177077017e-05, + "loss": 1.1713, + "step": 121460 + }, + { + "epoch": 10.51, + "learning_rate": 3.9520055444858355e-05, + "loss": 1.2541, + "step": 121470 + }, + { + "epoch": 10.51, + "learning_rate": 3.951918911894655e-05, + "loss": 1.2598, + "step": 121480 + }, + { + "epoch": 10.52, + "learning_rate": 3.951832279303474e-05, + "loss": 1.2002, + "step": 121490 + }, + { + "epoch": 10.52, + "learning_rate": 3.951745646712293e-05, + "loss": 1.2108, + "step": 121500 + }, + { + "epoch": 10.52, + "learning_rate": 3.9516590141211126e-05, + "loss": 1.1863, + "step": 121510 + }, + { + "epoch": 10.52, + "learning_rate": 3.951572381529931e-05, + "loss": 1.2347, + "step": 121520 + }, + { + "epoch": 10.52, + "learning_rate": 3.951485748938751e-05, + "loss": 1.2619, + "step": 121530 + }, + { + "epoch": 10.52, + "learning_rate": 3.95139911634757e-05, + "loss": 1.1765, + "step": 121540 + }, + { + "epoch": 10.52, + "learning_rate": 3.951312483756389e-05, + "loss": 1.1938, + "step": 121550 + }, + { + "epoch": 10.52, + "learning_rate": 3.9512258511652084e-05, + "loss": 1.2694, + "step": 121560 + }, + { + "epoch": 10.52, + "learning_rate": 3.951139218574028e-05, + "loss": 1.2163, + "step": 121570 + }, + { + "epoch": 10.52, + "learning_rate": 3.9510525859828466e-05, + "loss": 1.1299, + "step": 121580 + }, + { + "epoch": 10.52, + "learning_rate": 3.950965953391666e-05, + "loss": 1.1585, + "step": 121590 + }, + { + "epoch": 10.53, + "learning_rate": 3.950879320800485e-05, + "loss": 1.2748, + "step": 121600 + }, + { + "epoch": 10.53, + "learning_rate": 3.950792688209304e-05, + "loss": 1.3045, + "step": 121610 + }, + { + "epoch": 10.53, + "learning_rate": 3.950706055618124e-05, + "loss": 1.2148, + "step": 121620 + }, + { + "epoch": 10.53, + "learning_rate": 3.9506194230269425e-05, + "loss": 1.3165, + "step": 121630 + }, + { + "epoch": 10.53, + "learning_rate": 3.950532790435762e-05, + "loss": 1.1829, + "step": 121640 + }, + { + "epoch": 10.53, + "learning_rate": 3.9504461578445814e-05, + "loss": 1.1997, + "step": 121650 + }, + { + "epoch": 10.53, + "learning_rate": 3.9503595252534e-05, + "loss": 1.2357, + "step": 121660 + }, + { + "epoch": 10.53, + "learning_rate": 3.9502728926622196e-05, + "loss": 1.2601, + "step": 121670 + }, + { + "epoch": 10.53, + "learning_rate": 3.950186260071039e-05, + "loss": 1.2214, + "step": 121680 + }, + { + "epoch": 10.53, + "learning_rate": 3.950099627479858e-05, + "loss": 1.2755, + "step": 121690 + }, + { + "epoch": 10.53, + "learning_rate": 3.950012994888677e-05, + "loss": 1.217, + "step": 121700 + }, + { + "epoch": 10.53, + "learning_rate": 3.949926362297496e-05, + "loss": 1.2379, + "step": 121710 + }, + { + "epoch": 10.54, + "learning_rate": 3.9498397297063155e-05, + "loss": 1.1959, + "step": 121720 + }, + { + "epoch": 10.54, + "learning_rate": 3.949753097115135e-05, + "loss": 1.1944, + "step": 121730 + }, + { + "epoch": 10.54, + "learning_rate": 3.949666464523954e-05, + "loss": 1.2425, + "step": 121740 + }, + { + "epoch": 10.54, + "learning_rate": 3.949579831932773e-05, + "loss": 1.2624, + "step": 121750 + }, + { + "epoch": 10.54, + "learning_rate": 3.9494931993415926e-05, + "loss": 1.2703, + "step": 121760 + }, + { + "epoch": 10.54, + "learning_rate": 3.9494065667504114e-05, + "loss": 1.239, + "step": 121770 + }, + { + "epoch": 10.54, + "learning_rate": 3.949319934159231e-05, + "loss": 1.1807, + "step": 121780 + }, + { + "epoch": 10.54, + "learning_rate": 3.94923330156805e-05, + "loss": 1.1975, + "step": 121790 + }, + { + "epoch": 10.54, + "learning_rate": 3.949146668976869e-05, + "loss": 1.2154, + "step": 121800 + }, + { + "epoch": 10.54, + "learning_rate": 3.9490600363856885e-05, + "loss": 1.2069, + "step": 121810 + }, + { + "epoch": 10.54, + "learning_rate": 3.948973403794507e-05, + "loss": 1.2115, + "step": 121820 + }, + { + "epoch": 10.55, + "learning_rate": 3.948886771203327e-05, + "loss": 1.2134, + "step": 121830 + }, + { + "epoch": 10.55, + "learning_rate": 3.948800138612146e-05, + "loss": 1.2229, + "step": 121840 + }, + { + "epoch": 10.55, + "learning_rate": 3.948713506020965e-05, + "loss": 1.1942, + "step": 121850 + }, + { + "epoch": 10.55, + "learning_rate": 3.948626873429784e-05, + "loss": 1.196, + "step": 121860 + }, + { + "epoch": 10.55, + "learning_rate": 3.948540240838604e-05, + "loss": 1.2339, + "step": 121870 + }, + { + "epoch": 10.55, + "learning_rate": 3.9484536082474226e-05, + "loss": 1.2412, + "step": 121880 + }, + { + "epoch": 10.55, + "learning_rate": 3.948366975656242e-05, + "loss": 1.196, + "step": 121890 + }, + { + "epoch": 10.55, + "learning_rate": 3.9482803430650614e-05, + "loss": 1.2434, + "step": 121900 + }, + { + "epoch": 10.55, + "learning_rate": 3.94819371047388e-05, + "loss": 1.2376, + "step": 121910 + }, + { + "epoch": 10.55, + "learning_rate": 3.9481070778826997e-05, + "loss": 1.2082, + "step": 121920 + }, + { + "epoch": 10.55, + "learning_rate": 3.9480204452915184e-05, + "loss": 1.2396, + "step": 121930 + }, + { + "epoch": 10.55, + "learning_rate": 3.947933812700338e-05, + "loss": 1.1801, + "step": 121940 + }, + { + "epoch": 10.56, + "learning_rate": 3.947847180109157e-05, + "loss": 1.2331, + "step": 121950 + }, + { + "epoch": 10.56, + "learning_rate": 3.947760547517976e-05, + "loss": 1.135, + "step": 121960 + }, + { + "epoch": 10.56, + "learning_rate": 3.9476739149267955e-05, + "loss": 1.2401, + "step": 121970 + }, + { + "epoch": 10.56, + "learning_rate": 3.947587282335615e-05, + "loss": 1.1953, + "step": 121980 + }, + { + "epoch": 10.56, + "learning_rate": 3.947500649744434e-05, + "loss": 1.1854, + "step": 121990 + }, + { + "epoch": 10.56, + "learning_rate": 3.947414017153253e-05, + "loss": 1.2144, + "step": 122000 + }, + { + "epoch": 10.56, + "learning_rate": 3.9473273845620726e-05, + "loss": 1.2521, + "step": 122010 + }, + { + "epoch": 10.56, + "learning_rate": 3.9472407519708914e-05, + "loss": 1.2467, + "step": 122020 + }, + { + "epoch": 10.56, + "learning_rate": 3.947154119379711e-05, + "loss": 1.1548, + "step": 122030 + }, + { + "epoch": 10.56, + "learning_rate": 3.9470674867885296e-05, + "loss": 1.2123, + "step": 122040 + }, + { + "epoch": 10.56, + "learning_rate": 3.946980854197349e-05, + "loss": 1.2035, + "step": 122050 + }, + { + "epoch": 10.57, + "learning_rate": 3.9468942216061685e-05, + "loss": 1.2304, + "step": 122060 + }, + { + "epoch": 10.57, + "learning_rate": 3.946807589014987e-05, + "loss": 1.2031, + "step": 122070 + }, + { + "epoch": 10.57, + "learning_rate": 3.946720956423807e-05, + "loss": 1.1894, + "step": 122080 + }, + { + "epoch": 10.57, + "learning_rate": 3.946634323832626e-05, + "loss": 1.2155, + "step": 122090 + }, + { + "epoch": 10.57, + "learning_rate": 3.946547691241445e-05, + "loss": 1.185, + "step": 122100 + }, + { + "epoch": 10.57, + "learning_rate": 3.9464610586502644e-05, + "loss": 1.1835, + "step": 122110 + }, + { + "epoch": 10.57, + "learning_rate": 3.946374426059083e-05, + "loss": 1.2114, + "step": 122120 + }, + { + "epoch": 10.57, + "learning_rate": 3.9462877934679026e-05, + "loss": 1.234, + "step": 122130 + }, + { + "epoch": 10.57, + "learning_rate": 3.946201160876722e-05, + "loss": 1.198, + "step": 122140 + }, + { + "epoch": 10.57, + "learning_rate": 3.946114528285541e-05, + "loss": 1.2104, + "step": 122150 + }, + { + "epoch": 10.57, + "learning_rate": 3.94602789569436e-05, + "loss": 1.2442, + "step": 122160 + }, + { + "epoch": 10.57, + "learning_rate": 3.94594126310318e-05, + "loss": 1.2743, + "step": 122170 + }, + { + "epoch": 10.58, + "learning_rate": 3.9458546305119985e-05, + "loss": 1.2408, + "step": 122180 + }, + { + "epoch": 10.58, + "learning_rate": 3.945767997920818e-05, + "loss": 1.2681, + "step": 122190 + }, + { + "epoch": 10.58, + "learning_rate": 3.9456813653296373e-05, + "loss": 1.1927, + "step": 122200 + }, + { + "epoch": 10.58, + "learning_rate": 3.945594732738456e-05, + "loss": 1.2435, + "step": 122210 + }, + { + "epoch": 10.58, + "learning_rate": 3.9455081001472756e-05, + "loss": 1.2121, + "step": 122220 + }, + { + "epoch": 10.58, + "learning_rate": 3.945421467556094e-05, + "loss": 1.2073, + "step": 122230 + }, + { + "epoch": 10.58, + "learning_rate": 3.945334834964914e-05, + "loss": 1.241, + "step": 122240 + }, + { + "epoch": 10.58, + "learning_rate": 3.945248202373733e-05, + "loss": 1.2174, + "step": 122250 + }, + { + "epoch": 10.58, + "learning_rate": 3.945161569782552e-05, + "loss": 1.201, + "step": 122260 + }, + { + "epoch": 10.58, + "learning_rate": 3.9450749371913714e-05, + "loss": 1.2022, + "step": 122270 + }, + { + "epoch": 10.58, + "learning_rate": 3.944988304600191e-05, + "loss": 1.2525, + "step": 122280 + }, + { + "epoch": 10.58, + "learning_rate": 3.9449016720090096e-05, + "loss": 1.2085, + "step": 122290 + }, + { + "epoch": 10.59, + "learning_rate": 3.944815039417829e-05, + "loss": 1.2019, + "step": 122300 + }, + { + "epoch": 10.59, + "learning_rate": 3.9447284068266485e-05, + "loss": 1.2127, + "step": 122310 + }, + { + "epoch": 10.59, + "learning_rate": 3.944641774235467e-05, + "loss": 1.2395, + "step": 122320 + }, + { + "epoch": 10.59, + "learning_rate": 3.944555141644287e-05, + "loss": 1.239, + "step": 122330 + }, + { + "epoch": 10.59, + "learning_rate": 3.9444685090531055e-05, + "loss": 1.1953, + "step": 122340 + }, + { + "epoch": 10.59, + "learning_rate": 3.944381876461925e-05, + "loss": 1.1787, + "step": 122350 + }, + { + "epoch": 10.59, + "learning_rate": 3.9442952438707444e-05, + "loss": 1.2057, + "step": 122360 + }, + { + "epoch": 10.59, + "learning_rate": 3.944208611279563e-05, + "loss": 1.2507, + "step": 122370 + }, + { + "epoch": 10.59, + "learning_rate": 3.9441219786883826e-05, + "loss": 1.2489, + "step": 122380 + }, + { + "epoch": 10.59, + "learning_rate": 3.944035346097202e-05, + "loss": 1.2061, + "step": 122390 + }, + { + "epoch": 10.59, + "learning_rate": 3.943948713506021e-05, + "loss": 1.288, + "step": 122400 + }, + { + "epoch": 10.6, + "learning_rate": 3.94386208091484e-05, + "loss": 1.1828, + "step": 122410 + }, + { + "epoch": 10.6, + "learning_rate": 3.94377544832366e-05, + "loss": 1.1766, + "step": 122420 + }, + { + "epoch": 10.6, + "learning_rate": 3.9436888157324785e-05, + "loss": 1.2, + "step": 122430 + }, + { + "epoch": 10.6, + "learning_rate": 3.943602183141298e-05, + "loss": 1.2196, + "step": 122440 + }, + { + "epoch": 10.6, + "learning_rate": 3.943515550550117e-05, + "loss": 1.1665, + "step": 122450 + }, + { + "epoch": 10.6, + "learning_rate": 3.943428917958936e-05, + "loss": 1.1966, + "step": 122460 + }, + { + "epoch": 10.6, + "learning_rate": 3.9433422853677556e-05, + "loss": 1.2328, + "step": 122470 + }, + { + "epoch": 10.6, + "learning_rate": 3.9432556527765744e-05, + "loss": 1.2126, + "step": 122480 + }, + { + "epoch": 10.6, + "learning_rate": 3.943169020185394e-05, + "loss": 1.2054, + "step": 122490 + }, + { + "epoch": 10.6, + "learning_rate": 3.943082387594213e-05, + "loss": 1.2192, + "step": 122500 + }, + { + "epoch": 10.6, + "learning_rate": 3.942995755003032e-05, + "loss": 1.1972, + "step": 122510 + }, + { + "epoch": 10.6, + "learning_rate": 3.9429091224118515e-05, + "loss": 1.2684, + "step": 122520 + }, + { + "epoch": 10.61, + "learning_rate": 3.942822489820671e-05, + "loss": 1.1966, + "step": 122530 + }, + { + "epoch": 10.61, + "learning_rate": 3.94273585722949e-05, + "loss": 1.2339, + "step": 122540 + }, + { + "epoch": 10.61, + "learning_rate": 3.942649224638309e-05, + "loss": 1.1114, + "step": 122550 + }, + { + "epoch": 10.61, + "learning_rate": 3.942562592047128e-05, + "loss": 1.1612, + "step": 122560 + }, + { + "epoch": 10.61, + "learning_rate": 3.942475959455947e-05, + "loss": 1.2012, + "step": 122570 + }, + { + "epoch": 10.61, + "learning_rate": 3.942389326864767e-05, + "loss": 1.2528, + "step": 122580 + }, + { + "epoch": 10.61, + "learning_rate": 3.9423026942735855e-05, + "loss": 1.1889, + "step": 122590 + }, + { + "epoch": 10.61, + "learning_rate": 3.942216061682405e-05, + "loss": 1.2047, + "step": 122600 + }, + { + "epoch": 10.61, + "learning_rate": 3.9421294290912244e-05, + "loss": 1.1731, + "step": 122610 + }, + { + "epoch": 10.61, + "learning_rate": 3.942042796500043e-05, + "loss": 1.2355, + "step": 122620 + }, + { + "epoch": 10.61, + "learning_rate": 3.9419561639088627e-05, + "loss": 1.2532, + "step": 122630 + }, + { + "epoch": 10.62, + "learning_rate": 3.941869531317682e-05, + "loss": 1.2251, + "step": 122640 + }, + { + "epoch": 10.62, + "learning_rate": 3.941782898726501e-05, + "loss": 1.204, + "step": 122650 + }, + { + "epoch": 10.62, + "learning_rate": 3.94169626613532e-05, + "loss": 1.1941, + "step": 122660 + }, + { + "epoch": 10.62, + "learning_rate": 3.941609633544139e-05, + "loss": 1.178, + "step": 122670 + }, + { + "epoch": 10.62, + "learning_rate": 3.9415230009529585e-05, + "loss": 1.1679, + "step": 122680 + }, + { + "epoch": 10.62, + "learning_rate": 3.941436368361778e-05, + "loss": 1.1857, + "step": 122690 + }, + { + "epoch": 10.62, + "learning_rate": 3.941349735770597e-05, + "loss": 1.2728, + "step": 122700 + }, + { + "epoch": 10.62, + "learning_rate": 3.941263103179416e-05, + "loss": 1.2514, + "step": 122710 + }, + { + "epoch": 10.62, + "learning_rate": 3.9411764705882356e-05, + "loss": 1.1604, + "step": 122720 + }, + { + "epoch": 10.62, + "learning_rate": 3.9410898379970544e-05, + "loss": 1.1768, + "step": 122730 + }, + { + "epoch": 10.62, + "learning_rate": 3.941003205405874e-05, + "loss": 1.282, + "step": 122740 + }, + { + "epoch": 10.62, + "learning_rate": 3.940916572814693e-05, + "loss": 1.2007, + "step": 122750 + }, + { + "epoch": 10.63, + "learning_rate": 3.940829940223512e-05, + "loss": 1.263, + "step": 122760 + }, + { + "epoch": 10.63, + "learning_rate": 3.9407433076323315e-05, + "loss": 1.211, + "step": 122770 + }, + { + "epoch": 10.63, + "learning_rate": 3.94065667504115e-05, + "loss": 1.2633, + "step": 122780 + }, + { + "epoch": 10.63, + "learning_rate": 3.94057004244997e-05, + "loss": 1.1758, + "step": 122790 + }, + { + "epoch": 10.63, + "learning_rate": 3.940483409858789e-05, + "loss": 1.224, + "step": 122800 + }, + { + "epoch": 10.63, + "learning_rate": 3.940396777267608e-05, + "loss": 1.2314, + "step": 122810 + }, + { + "epoch": 10.63, + "learning_rate": 3.9403101446764274e-05, + "loss": 1.2554, + "step": 122820 + }, + { + "epoch": 10.63, + "learning_rate": 3.940223512085247e-05, + "loss": 1.2205, + "step": 122830 + }, + { + "epoch": 10.63, + "learning_rate": 3.9401368794940656e-05, + "loss": 1.2425, + "step": 122840 + }, + { + "epoch": 10.63, + "learning_rate": 3.940050246902885e-05, + "loss": 1.1855, + "step": 122850 + }, + { + "epoch": 10.63, + "learning_rate": 3.939963614311704e-05, + "loss": 1.2566, + "step": 122860 + }, + { + "epoch": 10.64, + "learning_rate": 3.939876981720523e-05, + "loss": 1.2396, + "step": 122870 + }, + { + "epoch": 10.64, + "learning_rate": 3.939790349129343e-05, + "loss": 1.2275, + "step": 122880 + }, + { + "epoch": 10.64, + "learning_rate": 3.9397037165381615e-05, + "loss": 1.1956, + "step": 122890 + }, + { + "epoch": 10.64, + "learning_rate": 3.939617083946981e-05, + "loss": 1.2063, + "step": 122900 + }, + { + "epoch": 10.64, + "learning_rate": 3.9395304513558003e-05, + "loss": 1.1746, + "step": 122910 + }, + { + "epoch": 10.64, + "learning_rate": 3.939443818764619e-05, + "loss": 1.2462, + "step": 122920 + }, + { + "epoch": 10.64, + "learning_rate": 3.9393571861734386e-05, + "loss": 1.2359, + "step": 122930 + }, + { + "epoch": 10.64, + "learning_rate": 3.939270553582258e-05, + "loss": 1.1404, + "step": 122940 + }, + { + "epoch": 10.64, + "learning_rate": 3.939183920991077e-05, + "loss": 1.2075, + "step": 122950 + }, + { + "epoch": 10.64, + "learning_rate": 3.939097288399896e-05, + "loss": 1.2031, + "step": 122960 + }, + { + "epoch": 10.64, + "learning_rate": 3.939010655808715e-05, + "loss": 1.1886, + "step": 122970 + }, + { + "epoch": 10.64, + "learning_rate": 3.9389240232175344e-05, + "loss": 1.2543, + "step": 122980 + }, + { + "epoch": 10.65, + "learning_rate": 3.938837390626354e-05, + "loss": 1.2588, + "step": 122990 + }, + { + "epoch": 10.65, + "learning_rate": 3.9387507580351726e-05, + "loss": 1.2317, + "step": 123000 + }, + { + "epoch": 10.65, + "learning_rate": 3.938664125443992e-05, + "loss": 1.2166, + "step": 123010 + }, + { + "epoch": 10.65, + "learning_rate": 3.9385774928528115e-05, + "loss": 1.217, + "step": 123020 + }, + { + "epoch": 10.65, + "learning_rate": 3.93849086026163e-05, + "loss": 1.2212, + "step": 123030 + }, + { + "epoch": 10.65, + "learning_rate": 3.93840422767045e-05, + "loss": 1.1956, + "step": 123040 + }, + { + "epoch": 10.65, + "learning_rate": 3.938317595079269e-05, + "loss": 1.1675, + "step": 123050 + }, + { + "epoch": 10.65, + "learning_rate": 3.938230962488088e-05, + "loss": 1.2538, + "step": 123060 + }, + { + "epoch": 10.65, + "learning_rate": 3.9381443298969074e-05, + "loss": 1.1908, + "step": 123070 + }, + { + "epoch": 10.65, + "learning_rate": 3.938057697305726e-05, + "loss": 1.2223, + "step": 123080 + }, + { + "epoch": 10.65, + "learning_rate": 3.9379710647145456e-05, + "loss": 1.2636, + "step": 123090 + }, + { + "epoch": 10.66, + "learning_rate": 3.937884432123365e-05, + "loss": 1.2971, + "step": 123100 + }, + { + "epoch": 10.66, + "learning_rate": 3.937797799532184e-05, + "loss": 1.2109, + "step": 123110 + }, + { + "epoch": 10.66, + "learning_rate": 3.937711166941003e-05, + "loss": 1.1865, + "step": 123120 + }, + { + "epoch": 10.66, + "learning_rate": 3.937624534349823e-05, + "loss": 1.209, + "step": 123130 + }, + { + "epoch": 10.66, + "learning_rate": 3.9375379017586415e-05, + "loss": 1.1551, + "step": 123140 + }, + { + "epoch": 10.66, + "learning_rate": 3.937451269167461e-05, + "loss": 1.2371, + "step": 123150 + }, + { + "epoch": 10.66, + "learning_rate": 3.9373646365762804e-05, + "loss": 1.1882, + "step": 123160 + }, + { + "epoch": 10.66, + "learning_rate": 3.937278003985099e-05, + "loss": 1.1933, + "step": 123170 + }, + { + "epoch": 10.66, + "learning_rate": 3.9371913713939186e-05, + "loss": 1.2702, + "step": 123180 + }, + { + "epoch": 10.66, + "learning_rate": 3.9371047388027374e-05, + "loss": 1.2425, + "step": 123190 + }, + { + "epoch": 10.66, + "learning_rate": 3.937018106211557e-05, + "loss": 1.1812, + "step": 123200 + }, + { + "epoch": 10.66, + "learning_rate": 3.936931473620376e-05, + "loss": 1.1811, + "step": 123210 + }, + { + "epoch": 10.67, + "learning_rate": 3.936844841029195e-05, + "loss": 1.2222, + "step": 123220 + }, + { + "epoch": 10.67, + "learning_rate": 3.9367582084380145e-05, + "loss": 1.1951, + "step": 123230 + }, + { + "epoch": 10.67, + "learning_rate": 3.936671575846834e-05, + "loss": 1.2247, + "step": 123240 + }, + { + "epoch": 10.67, + "learning_rate": 3.936584943255653e-05, + "loss": 1.2022, + "step": 123250 + }, + { + "epoch": 10.67, + "learning_rate": 3.936498310664472e-05, + "loss": 1.2077, + "step": 123260 + }, + { + "epoch": 10.67, + "learning_rate": 3.9364116780732916e-05, + "loss": 1.206, + "step": 123270 + }, + { + "epoch": 10.67, + "learning_rate": 3.93632504548211e-05, + "loss": 1.2243, + "step": 123280 + }, + { + "epoch": 10.67, + "learning_rate": 3.93623841289093e-05, + "loss": 1.2127, + "step": 123290 + }, + { + "epoch": 10.67, + "learning_rate": 3.9361517802997485e-05, + "loss": 1.2297, + "step": 123300 + }, + { + "epoch": 10.67, + "learning_rate": 3.936065147708568e-05, + "loss": 1.2106, + "step": 123310 + }, + { + "epoch": 10.67, + "learning_rate": 3.9359785151173874e-05, + "loss": 1.1842, + "step": 123320 + }, + { + "epoch": 10.67, + "learning_rate": 3.935891882526206e-05, + "loss": 1.2214, + "step": 123330 + }, + { + "epoch": 10.68, + "learning_rate": 3.9358052499350256e-05, + "loss": 1.2362, + "step": 123340 + }, + { + "epoch": 10.68, + "learning_rate": 3.935718617343845e-05, + "loss": 1.2301, + "step": 123350 + }, + { + "epoch": 10.68, + "learning_rate": 3.935631984752664e-05, + "loss": 1.1922, + "step": 123360 + }, + { + "epoch": 10.68, + "learning_rate": 3.935545352161483e-05, + "loss": 1.1973, + "step": 123370 + }, + { + "epoch": 10.68, + "learning_rate": 3.935458719570303e-05, + "loss": 1.2252, + "step": 123380 + }, + { + "epoch": 10.68, + "learning_rate": 3.9353720869791215e-05, + "loss": 1.2361, + "step": 123390 + }, + { + "epoch": 10.68, + "learning_rate": 3.935285454387941e-05, + "loss": 1.1587, + "step": 123400 + }, + { + "epoch": 10.68, + "learning_rate": 3.93519882179676e-05, + "loss": 1.1747, + "step": 123410 + }, + { + "epoch": 10.68, + "learning_rate": 3.935112189205579e-05, + "loss": 1.2397, + "step": 123420 + }, + { + "epoch": 10.68, + "learning_rate": 3.9350255566143986e-05, + "loss": 1.2308, + "step": 123430 + }, + { + "epoch": 10.68, + "learning_rate": 3.9349389240232174e-05, + "loss": 1.2012, + "step": 123440 + }, + { + "epoch": 10.69, + "learning_rate": 3.934852291432037e-05, + "loss": 1.2166, + "step": 123450 + }, + { + "epoch": 10.69, + "learning_rate": 3.934765658840856e-05, + "loss": 1.1684, + "step": 123460 + }, + { + "epoch": 10.69, + "learning_rate": 3.934679026249675e-05, + "loss": 1.2127, + "step": 123470 + }, + { + "epoch": 10.69, + "learning_rate": 3.9345923936584945e-05, + "loss": 1.2288, + "step": 123480 + }, + { + "epoch": 10.69, + "learning_rate": 3.934505761067314e-05, + "loss": 1.2639, + "step": 123490 + }, + { + "epoch": 10.69, + "learning_rate": 3.934419128476133e-05, + "loss": 1.2404, + "step": 123500 + }, + { + "epoch": 10.69, + "learning_rate": 3.934332495884952e-05, + "loss": 1.1895, + "step": 123510 + }, + { + "epoch": 10.69, + "learning_rate": 3.934245863293771e-05, + "loss": 1.1556, + "step": 123520 + }, + { + "epoch": 10.69, + "learning_rate": 3.9341592307025904e-05, + "loss": 1.2118, + "step": 123530 + }, + { + "epoch": 10.69, + "learning_rate": 3.93407259811141e-05, + "loss": 1.2016, + "step": 123540 + }, + { + "epoch": 10.69, + "learning_rate": 3.9339859655202286e-05, + "loss": 1.2115, + "step": 123550 + }, + { + "epoch": 10.69, + "learning_rate": 3.933899332929048e-05, + "loss": 1.2226, + "step": 123560 + }, + { + "epoch": 10.7, + "learning_rate": 3.9338127003378675e-05, + "loss": 1.2112, + "step": 123570 + }, + { + "epoch": 10.7, + "learning_rate": 3.933726067746686e-05, + "loss": 1.1796, + "step": 123580 + }, + { + "epoch": 10.7, + "learning_rate": 3.933639435155506e-05, + "loss": 1.1589, + "step": 123590 + }, + { + "epoch": 10.7, + "learning_rate": 3.933552802564325e-05, + "loss": 1.187, + "step": 123600 + }, + { + "epoch": 10.7, + "learning_rate": 3.933466169973144e-05, + "loss": 1.2378, + "step": 123610 + }, + { + "epoch": 10.7, + "learning_rate": 3.9333795373819633e-05, + "loss": 1.208, + "step": 123620 + }, + { + "epoch": 10.7, + "learning_rate": 3.933292904790782e-05, + "loss": 1.1987, + "step": 123630 + }, + { + "epoch": 10.7, + "learning_rate": 3.9332062721996016e-05, + "loss": 1.2042, + "step": 123640 + }, + { + "epoch": 10.7, + "learning_rate": 3.933119639608421e-05, + "loss": 1.2106, + "step": 123650 + }, + { + "epoch": 10.7, + "learning_rate": 3.93303300701724e-05, + "loss": 1.2108, + "step": 123660 + }, + { + "epoch": 10.7, + "learning_rate": 3.932946374426059e-05, + "loss": 1.2234, + "step": 123670 + }, + { + "epoch": 10.71, + "learning_rate": 3.9328597418348787e-05, + "loss": 1.2238, + "step": 123680 + }, + { + "epoch": 10.71, + "learning_rate": 3.9327731092436974e-05, + "loss": 1.1938, + "step": 123690 + }, + { + "epoch": 10.71, + "learning_rate": 3.932686476652517e-05, + "loss": 1.227, + "step": 123700 + }, + { + "epoch": 10.71, + "learning_rate": 3.9325998440613356e-05, + "loss": 1.1898, + "step": 123710 + }, + { + "epoch": 10.71, + "learning_rate": 3.932513211470155e-05, + "loss": 1.2363, + "step": 123720 + }, + { + "epoch": 10.71, + "learning_rate": 3.9324265788789745e-05, + "loss": 1.1863, + "step": 123730 + }, + { + "epoch": 10.71, + "learning_rate": 3.932339946287793e-05, + "loss": 1.2383, + "step": 123740 + }, + { + "epoch": 10.71, + "learning_rate": 3.932253313696613e-05, + "loss": 1.2298, + "step": 123750 + }, + { + "epoch": 10.71, + "learning_rate": 3.932166681105432e-05, + "loss": 1.229, + "step": 123760 + }, + { + "epoch": 10.71, + "learning_rate": 3.932080048514251e-05, + "loss": 1.1939, + "step": 123770 + }, + { + "epoch": 10.71, + "learning_rate": 3.9319934159230704e-05, + "loss": 1.157, + "step": 123780 + }, + { + "epoch": 10.71, + "learning_rate": 3.93190678333189e-05, + "loss": 1.2271, + "step": 123790 + }, + { + "epoch": 10.72, + "learning_rate": 3.9318201507407086e-05, + "loss": 1.1723, + "step": 123800 + }, + { + "epoch": 10.72, + "learning_rate": 3.931733518149528e-05, + "loss": 1.2637, + "step": 123810 + }, + { + "epoch": 10.72, + "learning_rate": 3.931646885558347e-05, + "loss": 1.2459, + "step": 123820 + }, + { + "epoch": 10.72, + "learning_rate": 3.931560252967166e-05, + "loss": 1.206, + "step": 123830 + }, + { + "epoch": 10.72, + "learning_rate": 3.931473620375986e-05, + "loss": 1.2103, + "step": 123840 + }, + { + "epoch": 10.72, + "learning_rate": 3.9313869877848045e-05, + "loss": 1.1841, + "step": 123850 + }, + { + "epoch": 10.72, + "learning_rate": 3.931300355193624e-05, + "loss": 1.1928, + "step": 123860 + }, + { + "epoch": 10.72, + "learning_rate": 3.9312137226024434e-05, + "loss": 1.2056, + "step": 123870 + }, + { + "epoch": 10.72, + "learning_rate": 3.931127090011262e-05, + "loss": 1.1594, + "step": 123880 + }, + { + "epoch": 10.72, + "learning_rate": 3.9310404574200816e-05, + "loss": 1.1919, + "step": 123890 + }, + { + "epoch": 10.72, + "learning_rate": 3.930953824828901e-05, + "loss": 1.1941, + "step": 123900 + }, + { + "epoch": 10.73, + "learning_rate": 3.93086719223772e-05, + "loss": 1.221, + "step": 123910 + }, + { + "epoch": 10.73, + "learning_rate": 3.930780559646539e-05, + "loss": 1.2133, + "step": 123920 + }, + { + "epoch": 10.73, + "learning_rate": 3.930693927055358e-05, + "loss": 1.216, + "step": 123930 + }, + { + "epoch": 10.73, + "learning_rate": 3.9306072944641775e-05, + "loss": 1.2395, + "step": 123940 + }, + { + "epoch": 10.73, + "learning_rate": 3.930520661872997e-05, + "loss": 1.2708, + "step": 123950 + }, + { + "epoch": 10.73, + "learning_rate": 3.930434029281816e-05, + "loss": 1.2499, + "step": 123960 + }, + { + "epoch": 10.73, + "learning_rate": 3.930347396690635e-05, + "loss": 1.1937, + "step": 123970 + }, + { + "epoch": 10.73, + "learning_rate": 3.9302607640994546e-05, + "loss": 1.2303, + "step": 123980 + }, + { + "epoch": 10.73, + "learning_rate": 3.930174131508273e-05, + "loss": 1.1811, + "step": 123990 + }, + { + "epoch": 10.73, + "learning_rate": 3.930087498917093e-05, + "loss": 1.2119, + "step": 124000 + }, + { + "epoch": 10.73, + "learning_rate": 3.930000866325912e-05, + "loss": 1.2135, + "step": 124010 + }, + { + "epoch": 10.73, + "learning_rate": 3.929914233734731e-05, + "loss": 1.2298, + "step": 124020 + }, + { + "epoch": 10.74, + "learning_rate": 3.9298276011435504e-05, + "loss": 1.1918, + "step": 124030 + }, + { + "epoch": 10.74, + "learning_rate": 3.929740968552369e-05, + "loss": 1.233, + "step": 124040 + }, + { + "epoch": 10.74, + "learning_rate": 3.9296543359611886e-05, + "loss": 1.2195, + "step": 124050 + }, + { + "epoch": 10.74, + "learning_rate": 3.929567703370008e-05, + "loss": 1.2489, + "step": 124060 + }, + { + "epoch": 10.74, + "learning_rate": 3.929481070778827e-05, + "loss": 1.2105, + "step": 124070 + }, + { + "epoch": 10.74, + "learning_rate": 3.929394438187646e-05, + "loss": 1.2349, + "step": 124080 + }, + { + "epoch": 10.74, + "learning_rate": 3.929307805596466e-05, + "loss": 1.2685, + "step": 124090 + }, + { + "epoch": 10.74, + "learning_rate": 3.9292211730052845e-05, + "loss": 1.2041, + "step": 124100 + }, + { + "epoch": 10.74, + "learning_rate": 3.929134540414104e-05, + "loss": 1.2239, + "step": 124110 + }, + { + "epoch": 10.74, + "learning_rate": 3.9290479078229234e-05, + "loss": 1.1924, + "step": 124120 + }, + { + "epoch": 10.74, + "learning_rate": 3.928961275231742e-05, + "loss": 1.1672, + "step": 124130 + }, + { + "epoch": 10.75, + "learning_rate": 3.9288746426405616e-05, + "loss": 1.2073, + "step": 124140 + }, + { + "epoch": 10.75, + "learning_rate": 3.9287880100493804e-05, + "loss": 1.2084, + "step": 124150 + }, + { + "epoch": 10.75, + "learning_rate": 3.9287013774582e-05, + "loss": 1.1866, + "step": 124160 + }, + { + "epoch": 10.75, + "learning_rate": 3.928614744867019e-05, + "loss": 1.1688, + "step": 124170 + }, + { + "epoch": 10.75, + "learning_rate": 3.928528112275838e-05, + "loss": 1.2097, + "step": 124180 + }, + { + "epoch": 10.75, + "learning_rate": 3.9284414796846575e-05, + "loss": 1.2096, + "step": 124190 + }, + { + "epoch": 10.75, + "learning_rate": 3.928354847093477e-05, + "loss": 1.1834, + "step": 124200 + }, + { + "epoch": 10.75, + "learning_rate": 3.928268214502296e-05, + "loss": 1.2074, + "step": 124210 + }, + { + "epoch": 10.75, + "learning_rate": 3.928181581911115e-05, + "loss": 1.2247, + "step": 124220 + }, + { + "epoch": 10.75, + "learning_rate": 3.9280949493199346e-05, + "loss": 1.2606, + "step": 124230 + }, + { + "epoch": 10.75, + "learning_rate": 3.9280083167287534e-05, + "loss": 1.2069, + "step": 124240 + }, + { + "epoch": 10.75, + "learning_rate": 3.927921684137573e-05, + "loss": 1.2059, + "step": 124250 + }, + { + "epoch": 10.76, + "learning_rate": 3.9278350515463916e-05, + "loss": 1.1504, + "step": 124260 + }, + { + "epoch": 10.76, + "learning_rate": 3.927748418955211e-05, + "loss": 1.16, + "step": 124270 + }, + { + "epoch": 10.76, + "learning_rate": 3.9276617863640305e-05, + "loss": 1.23, + "step": 124280 + }, + { + "epoch": 10.76, + "learning_rate": 3.927575153772849e-05, + "loss": 1.1611, + "step": 124290 + }, + { + "epoch": 10.76, + "learning_rate": 3.927488521181669e-05, + "loss": 1.1875, + "step": 124300 + }, + { + "epoch": 10.76, + "learning_rate": 3.927401888590488e-05, + "loss": 1.2249, + "step": 124310 + }, + { + "epoch": 10.76, + "learning_rate": 3.927315255999307e-05, + "loss": 1.2082, + "step": 124320 + }, + { + "epoch": 10.76, + "learning_rate": 3.927228623408126e-05, + "loss": 1.2129, + "step": 124330 + }, + { + "epoch": 10.76, + "learning_rate": 3.927141990816946e-05, + "loss": 1.2407, + "step": 124340 + }, + { + "epoch": 10.76, + "learning_rate": 3.9270553582257646e-05, + "loss": 1.1862, + "step": 124350 + }, + { + "epoch": 10.76, + "learning_rate": 3.926968725634584e-05, + "loss": 1.2072, + "step": 124360 + }, + { + "epoch": 10.76, + "learning_rate": 3.926882093043403e-05, + "loss": 1.1755, + "step": 124370 + }, + { + "epoch": 10.77, + "learning_rate": 3.926795460452222e-05, + "loss": 1.1868, + "step": 124380 + }, + { + "epoch": 10.77, + "learning_rate": 3.9267088278610417e-05, + "loss": 1.1889, + "step": 124390 + }, + { + "epoch": 10.77, + "learning_rate": 3.9266221952698604e-05, + "loss": 1.2307, + "step": 124400 + }, + { + "epoch": 10.77, + "learning_rate": 3.92653556267868e-05, + "loss": 1.1583, + "step": 124410 + }, + { + "epoch": 10.77, + "learning_rate": 3.926448930087499e-05, + "loss": 1.216, + "step": 124420 + }, + { + "epoch": 10.77, + "learning_rate": 3.926362297496318e-05, + "loss": 1.1855, + "step": 124430 + }, + { + "epoch": 10.77, + "learning_rate": 3.9262756649051375e-05, + "loss": 1.2573, + "step": 124440 + }, + { + "epoch": 10.77, + "learning_rate": 3.926189032313956e-05, + "loss": 1.2563, + "step": 124450 + }, + { + "epoch": 10.77, + "learning_rate": 3.926102399722776e-05, + "loss": 1.1845, + "step": 124460 + }, + { + "epoch": 10.77, + "learning_rate": 3.926015767131595e-05, + "loss": 1.1686, + "step": 124470 + }, + { + "epoch": 10.77, + "learning_rate": 3.925929134540414e-05, + "loss": 1.1992, + "step": 124480 + }, + { + "epoch": 10.78, + "learning_rate": 3.9258425019492334e-05, + "loss": 1.2733, + "step": 124490 + }, + { + "epoch": 10.78, + "learning_rate": 3.925755869358053e-05, + "loss": 1.2105, + "step": 124500 + }, + { + "epoch": 10.78, + "learning_rate": 3.9256692367668716e-05, + "loss": 1.1782, + "step": 124510 + }, + { + "epoch": 10.78, + "learning_rate": 3.925582604175691e-05, + "loss": 1.1614, + "step": 124520 + }, + { + "epoch": 10.78, + "learning_rate": 3.9254959715845105e-05, + "loss": 1.2174, + "step": 124530 + }, + { + "epoch": 10.78, + "learning_rate": 3.925409338993329e-05, + "loss": 1.2155, + "step": 124540 + }, + { + "epoch": 10.78, + "learning_rate": 3.925322706402149e-05, + "loss": 1.2332, + "step": 124550 + }, + { + "epoch": 10.78, + "learning_rate": 3.9252360738109675e-05, + "loss": 1.1817, + "step": 124560 + }, + { + "epoch": 10.78, + "learning_rate": 3.925149441219787e-05, + "loss": 1.2123, + "step": 124570 + }, + { + "epoch": 10.78, + "learning_rate": 3.9250628086286064e-05, + "loss": 1.1812, + "step": 124580 + }, + { + "epoch": 10.78, + "learning_rate": 3.924976176037425e-05, + "loss": 1.2173, + "step": 124590 + }, + { + "epoch": 10.78, + "learning_rate": 3.9248895434462446e-05, + "loss": 1.2012, + "step": 124600 + }, + { + "epoch": 10.79, + "learning_rate": 3.924802910855064e-05, + "loss": 1.2327, + "step": 124610 + }, + { + "epoch": 10.79, + "learning_rate": 3.924716278263883e-05, + "loss": 1.2253, + "step": 124620 + }, + { + "epoch": 10.79, + "learning_rate": 3.924629645672702e-05, + "loss": 1.2332, + "step": 124630 + }, + { + "epoch": 10.79, + "learning_rate": 3.924543013081522e-05, + "loss": 1.1849, + "step": 124640 + }, + { + "epoch": 10.79, + "learning_rate": 3.9244563804903405e-05, + "loss": 1.2277, + "step": 124650 + }, + { + "epoch": 10.79, + "learning_rate": 3.92436974789916e-05, + "loss": 1.2376, + "step": 124660 + }, + { + "epoch": 10.79, + "learning_rate": 3.924283115307979e-05, + "loss": 1.1763, + "step": 124670 + }, + { + "epoch": 10.79, + "learning_rate": 3.924196482716798e-05, + "loss": 1.1903, + "step": 124680 + }, + { + "epoch": 10.79, + "learning_rate": 3.9241098501256176e-05, + "loss": 1.1742, + "step": 124690 + }, + { + "epoch": 10.79, + "learning_rate": 3.924023217534436e-05, + "loss": 1.1973, + "step": 124700 + }, + { + "epoch": 10.79, + "learning_rate": 3.923936584943256e-05, + "loss": 1.2446, + "step": 124710 + }, + { + "epoch": 10.8, + "learning_rate": 3.923849952352075e-05, + "loss": 1.2018, + "step": 124720 + }, + { + "epoch": 10.8, + "learning_rate": 3.923763319760894e-05, + "loss": 1.2035, + "step": 124730 + }, + { + "epoch": 10.8, + "learning_rate": 3.9236766871697134e-05, + "loss": 1.23, + "step": 124740 + }, + { + "epoch": 10.8, + "learning_rate": 3.923590054578533e-05, + "loss": 1.2227, + "step": 124750 + }, + { + "epoch": 10.8, + "learning_rate": 3.9235034219873516e-05, + "loss": 1.1887, + "step": 124760 + }, + { + "epoch": 10.8, + "learning_rate": 3.923416789396171e-05, + "loss": 1.1601, + "step": 124770 + }, + { + "epoch": 10.8, + "learning_rate": 3.92333015680499e-05, + "loss": 1.1941, + "step": 124780 + }, + { + "epoch": 10.8, + "learning_rate": 3.923243524213809e-05, + "loss": 1.1812, + "step": 124790 + }, + { + "epoch": 10.8, + "learning_rate": 3.923156891622629e-05, + "loss": 1.1869, + "step": 124800 + }, + { + "epoch": 10.8, + "learning_rate": 3.9230702590314475e-05, + "loss": 1.236, + "step": 124810 + }, + { + "epoch": 10.8, + "learning_rate": 3.922983626440267e-05, + "loss": 1.2135, + "step": 124820 + }, + { + "epoch": 10.8, + "learning_rate": 3.9228969938490864e-05, + "loss": 1.2304, + "step": 124830 + }, + { + "epoch": 10.81, + "learning_rate": 3.922810361257905e-05, + "loss": 1.1666, + "step": 124840 + }, + { + "epoch": 10.81, + "learning_rate": 3.9227237286667246e-05, + "loss": 1.223, + "step": 124850 + }, + { + "epoch": 10.81, + "learning_rate": 3.922637096075544e-05, + "loss": 1.2309, + "step": 124860 + }, + { + "epoch": 10.81, + "learning_rate": 3.922550463484363e-05, + "loss": 1.2882, + "step": 124870 + }, + { + "epoch": 10.81, + "learning_rate": 3.922463830893182e-05, + "loss": 1.1847, + "step": 124880 + }, + { + "epoch": 10.81, + "learning_rate": 3.922377198302001e-05, + "loss": 1.1655, + "step": 124890 + }, + { + "epoch": 10.81, + "learning_rate": 3.9222905657108205e-05, + "loss": 1.1811, + "step": 124900 + }, + { + "epoch": 10.81, + "learning_rate": 3.92220393311964e-05, + "loss": 1.2227, + "step": 124910 + }, + { + "epoch": 10.81, + "learning_rate": 3.922117300528459e-05, + "loss": 1.2324, + "step": 124920 + }, + { + "epoch": 10.81, + "learning_rate": 3.922030667937278e-05, + "loss": 1.2558, + "step": 124930 + }, + { + "epoch": 10.81, + "learning_rate": 3.9219440353460976e-05, + "loss": 1.1311, + "step": 124940 + }, + { + "epoch": 10.82, + "learning_rate": 3.9218574027549164e-05, + "loss": 1.209, + "step": 124950 + }, + { + "epoch": 10.82, + "learning_rate": 3.921770770163736e-05, + "loss": 1.228, + "step": 124960 + }, + { + "epoch": 10.82, + "learning_rate": 3.921684137572555e-05, + "loss": 1.1923, + "step": 124970 + }, + { + "epoch": 10.82, + "learning_rate": 3.921597504981374e-05, + "loss": 1.194, + "step": 124980 + }, + { + "epoch": 10.82, + "learning_rate": 3.9215108723901935e-05, + "loss": 1.1998, + "step": 124990 + }, + { + "epoch": 10.82, + "learning_rate": 3.921424239799012e-05, + "loss": 1.2387, + "step": 125000 + }, + { + "epoch": 10.82, + "learning_rate": 3.921337607207832e-05, + "loss": 1.2176, + "step": 125010 + }, + { + "epoch": 10.82, + "learning_rate": 3.921250974616651e-05, + "loss": 1.2747, + "step": 125020 + }, + { + "epoch": 10.82, + "learning_rate": 3.92116434202547e-05, + "loss": 1.2187, + "step": 125030 + }, + { + "epoch": 10.82, + "learning_rate": 3.921077709434289e-05, + "loss": 1.2148, + "step": 125040 + }, + { + "epoch": 10.82, + "learning_rate": 3.920991076843109e-05, + "loss": 1.1467, + "step": 125050 + }, + { + "epoch": 10.82, + "learning_rate": 3.9209044442519275e-05, + "loss": 1.2232, + "step": 125060 + }, + { + "epoch": 10.83, + "learning_rate": 3.920817811660747e-05, + "loss": 1.2214, + "step": 125070 + }, + { + "epoch": 10.83, + "learning_rate": 3.9207311790695664e-05, + "loss": 1.2218, + "step": 125080 + }, + { + "epoch": 10.83, + "learning_rate": 3.920644546478385e-05, + "loss": 1.246, + "step": 125090 + }, + { + "epoch": 10.83, + "learning_rate": 3.9205579138872047e-05, + "loss": 1.1795, + "step": 125100 + }, + { + "epoch": 10.83, + "learning_rate": 3.9204712812960234e-05, + "loss": 1.1955, + "step": 125110 + }, + { + "epoch": 10.83, + "learning_rate": 3.920384648704843e-05, + "loss": 1.1985, + "step": 125120 + }, + { + "epoch": 10.83, + "learning_rate": 3.920298016113662e-05, + "loss": 1.1503, + "step": 125130 + }, + { + "epoch": 10.83, + "learning_rate": 3.920211383522481e-05, + "loss": 1.1704, + "step": 125140 + }, + { + "epoch": 10.83, + "learning_rate": 3.9201247509313005e-05, + "loss": 1.249, + "step": 125150 + }, + { + "epoch": 10.83, + "learning_rate": 3.92003811834012e-05, + "loss": 1.2375, + "step": 125160 + }, + { + "epoch": 10.83, + "learning_rate": 3.919951485748939e-05, + "loss": 1.2289, + "step": 125170 + }, + { + "epoch": 10.84, + "learning_rate": 3.919864853157758e-05, + "loss": 1.22, + "step": 125180 + }, + { + "epoch": 10.84, + "learning_rate": 3.919778220566577e-05, + "loss": 1.2391, + "step": 125190 + }, + { + "epoch": 10.84, + "learning_rate": 3.9196915879753964e-05, + "loss": 1.2768, + "step": 125200 + }, + { + "epoch": 10.84, + "learning_rate": 3.919604955384216e-05, + "loss": 1.214, + "step": 125210 + }, + { + "epoch": 10.84, + "learning_rate": 3.9195183227930346e-05, + "loss": 1.2352, + "step": 125220 + }, + { + "epoch": 10.84, + "learning_rate": 3.919431690201854e-05, + "loss": 1.1519, + "step": 125230 + }, + { + "epoch": 10.84, + "learning_rate": 3.9193450576106735e-05, + "loss": 1.1908, + "step": 125240 + }, + { + "epoch": 10.84, + "learning_rate": 3.919258425019492e-05, + "loss": 1.2078, + "step": 125250 + }, + { + "epoch": 10.84, + "learning_rate": 3.919171792428312e-05, + "loss": 1.2221, + "step": 125260 + }, + { + "epoch": 10.84, + "learning_rate": 3.919085159837131e-05, + "loss": 1.2132, + "step": 125270 + }, + { + "epoch": 10.84, + "learning_rate": 3.91899852724595e-05, + "loss": 1.1982, + "step": 125280 + }, + { + "epoch": 10.84, + "learning_rate": 3.9189118946547694e-05, + "loss": 1.2085, + "step": 125290 + }, + { + "epoch": 10.85, + "learning_rate": 3.918825262063588e-05, + "loss": 1.2164, + "step": 125300 + }, + { + "epoch": 10.85, + "learning_rate": 3.9187386294724076e-05, + "loss": 1.2187, + "step": 125310 + }, + { + "epoch": 10.85, + "learning_rate": 3.918651996881227e-05, + "loss": 1.2426, + "step": 125320 + }, + { + "epoch": 10.85, + "learning_rate": 3.918565364290046e-05, + "loss": 1.2316, + "step": 125330 + }, + { + "epoch": 10.85, + "learning_rate": 3.918478731698865e-05, + "loss": 1.1756, + "step": 125340 + }, + { + "epoch": 10.85, + "learning_rate": 3.918392099107685e-05, + "loss": 1.1954, + "step": 125350 + }, + { + "epoch": 10.85, + "learning_rate": 3.9183054665165035e-05, + "loss": 1.2102, + "step": 125360 + }, + { + "epoch": 10.85, + "learning_rate": 3.918218833925323e-05, + "loss": 1.2111, + "step": 125370 + }, + { + "epoch": 10.85, + "learning_rate": 3.9181322013341423e-05, + "loss": 1.211, + "step": 125380 + }, + { + "epoch": 10.85, + "learning_rate": 3.918045568742961e-05, + "loss": 1.2153, + "step": 125390 + }, + { + "epoch": 10.85, + "learning_rate": 3.9179589361517806e-05, + "loss": 1.1501, + "step": 125400 + }, + { + "epoch": 10.85, + "learning_rate": 3.917872303560599e-05, + "loss": 1.1926, + "step": 125410 + }, + { + "epoch": 10.86, + "learning_rate": 3.917785670969419e-05, + "loss": 1.2375, + "step": 125420 + }, + { + "epoch": 10.86, + "learning_rate": 3.917699038378238e-05, + "loss": 1.2324, + "step": 125430 + }, + { + "epoch": 10.86, + "learning_rate": 3.917612405787057e-05, + "loss": 1.2419, + "step": 125440 + }, + { + "epoch": 10.86, + "learning_rate": 3.9175257731958764e-05, + "loss": 1.1903, + "step": 125450 + }, + { + "epoch": 10.86, + "learning_rate": 3.917439140604696e-05, + "loss": 1.1823, + "step": 125460 + }, + { + "epoch": 10.86, + "learning_rate": 3.9173525080135146e-05, + "loss": 1.2268, + "step": 125470 + }, + { + "epoch": 10.86, + "learning_rate": 3.917265875422334e-05, + "loss": 1.2289, + "step": 125480 + }, + { + "epoch": 10.86, + "learning_rate": 3.9171792428311535e-05, + "loss": 1.1392, + "step": 125490 + }, + { + "epoch": 10.86, + "learning_rate": 3.917092610239972e-05, + "loss": 1.2101, + "step": 125500 + }, + { + "epoch": 10.86, + "learning_rate": 3.917005977648792e-05, + "loss": 1.1949, + "step": 125510 + }, + { + "epoch": 10.86, + "learning_rate": 3.9169193450576105e-05, + "loss": 1.1899, + "step": 125520 + }, + { + "epoch": 10.87, + "learning_rate": 3.91683271246643e-05, + "loss": 1.1431, + "step": 125530 + }, + { + "epoch": 10.87, + "learning_rate": 3.9167460798752494e-05, + "loss": 1.193, + "step": 125540 + }, + { + "epoch": 10.87, + "learning_rate": 3.916659447284068e-05, + "loss": 1.1975, + "step": 125550 + }, + { + "epoch": 10.87, + "learning_rate": 3.9165728146928876e-05, + "loss": 1.1978, + "step": 125560 + }, + { + "epoch": 10.87, + "learning_rate": 3.916486182101707e-05, + "loss": 1.1935, + "step": 125570 + }, + { + "epoch": 10.87, + "learning_rate": 3.916399549510526e-05, + "loss": 1.2044, + "step": 125580 + }, + { + "epoch": 10.87, + "learning_rate": 3.916312916919345e-05, + "loss": 1.206, + "step": 125590 + }, + { + "epoch": 10.87, + "learning_rate": 3.916226284328165e-05, + "loss": 1.1725, + "step": 125600 + }, + { + "epoch": 10.87, + "learning_rate": 3.9161396517369835e-05, + "loss": 1.159, + "step": 125610 + }, + { + "epoch": 10.87, + "learning_rate": 3.916053019145803e-05, + "loss": 1.2125, + "step": 125620 + }, + { + "epoch": 10.87, + "learning_rate": 3.915966386554622e-05, + "loss": 1.2019, + "step": 125630 + }, + { + "epoch": 10.87, + "learning_rate": 3.915879753963441e-05, + "loss": 1.2553, + "step": 125640 + }, + { + "epoch": 10.88, + "learning_rate": 3.9157931213722606e-05, + "loss": 1.1872, + "step": 125650 + }, + { + "epoch": 10.88, + "learning_rate": 3.9157064887810794e-05, + "loss": 1.2122, + "step": 125660 + }, + { + "epoch": 10.88, + "learning_rate": 3.915619856189899e-05, + "loss": 1.2063, + "step": 125670 + }, + { + "epoch": 10.88, + "learning_rate": 3.915533223598718e-05, + "loss": 1.1859, + "step": 125680 + }, + { + "epoch": 10.88, + "learning_rate": 3.915446591007537e-05, + "loss": 1.2034, + "step": 125690 + }, + { + "epoch": 10.88, + "learning_rate": 3.9153599584163565e-05, + "loss": 1.2337, + "step": 125700 + }, + { + "epoch": 10.88, + "learning_rate": 3.915273325825176e-05, + "loss": 1.1966, + "step": 125710 + }, + { + "epoch": 10.88, + "learning_rate": 3.915186693233995e-05, + "loss": 1.2623, + "step": 125720 + }, + { + "epoch": 10.88, + "learning_rate": 3.915100060642814e-05, + "loss": 1.2457, + "step": 125730 + }, + { + "epoch": 10.88, + "learning_rate": 3.915013428051633e-05, + "loss": 1.1706, + "step": 125740 + }, + { + "epoch": 10.88, + "learning_rate": 3.914926795460452e-05, + "loss": 1.1695, + "step": 125750 + }, + { + "epoch": 10.89, + "learning_rate": 3.914840162869272e-05, + "loss": 1.2121, + "step": 125760 + }, + { + "epoch": 10.89, + "learning_rate": 3.9147535302780905e-05, + "loss": 1.2205, + "step": 125770 + }, + { + "epoch": 10.89, + "learning_rate": 3.91466689768691e-05, + "loss": 1.2273, + "step": 125780 + }, + { + "epoch": 10.89, + "learning_rate": 3.9145802650957294e-05, + "loss": 1.2033, + "step": 125790 + }, + { + "epoch": 10.89, + "learning_rate": 3.914493632504548e-05, + "loss": 1.2037, + "step": 125800 + }, + { + "epoch": 10.89, + "learning_rate": 3.9144069999133676e-05, + "loss": 1.2557, + "step": 125810 + }, + { + "epoch": 10.89, + "learning_rate": 3.914320367322187e-05, + "loss": 1.2172, + "step": 125820 + }, + { + "epoch": 10.89, + "learning_rate": 3.914233734731006e-05, + "loss": 1.1714, + "step": 125830 + }, + { + "epoch": 10.89, + "learning_rate": 3.914147102139825e-05, + "loss": 1.2011, + "step": 125840 + }, + { + "epoch": 10.89, + "learning_rate": 3.914060469548644e-05, + "loss": 1.1669, + "step": 125850 + }, + { + "epoch": 10.89, + "learning_rate": 3.9139738369574635e-05, + "loss": 1.1632, + "step": 125860 + }, + { + "epoch": 10.89, + "learning_rate": 3.913887204366283e-05, + "loss": 1.2158, + "step": 125870 + }, + { + "epoch": 10.9, + "learning_rate": 3.913800571775102e-05, + "loss": 1.1935, + "step": 125880 + }, + { + "epoch": 10.9, + "learning_rate": 3.913713939183921e-05, + "loss": 1.2238, + "step": 125890 + }, + { + "epoch": 10.9, + "learning_rate": 3.9136273065927406e-05, + "loss": 1.1749, + "step": 125900 + }, + { + "epoch": 10.9, + "learning_rate": 3.9135406740015594e-05, + "loss": 1.1955, + "step": 125910 + }, + { + "epoch": 10.9, + "learning_rate": 3.913454041410379e-05, + "loss": 1.1884, + "step": 125920 + }, + { + "epoch": 10.9, + "learning_rate": 3.913367408819198e-05, + "loss": 1.2034, + "step": 125930 + }, + { + "epoch": 10.9, + "learning_rate": 3.913280776228017e-05, + "loss": 1.2171, + "step": 125940 + }, + { + "epoch": 10.9, + "learning_rate": 3.9131941436368365e-05, + "loss": 1.1271, + "step": 125950 + }, + { + "epoch": 10.9, + "learning_rate": 3.913107511045655e-05, + "loss": 1.229, + "step": 125960 + }, + { + "epoch": 10.9, + "learning_rate": 3.913020878454475e-05, + "loss": 1.1732, + "step": 125970 + }, + { + "epoch": 10.9, + "learning_rate": 3.912934245863294e-05, + "loss": 1.2392, + "step": 125980 + }, + { + "epoch": 10.91, + "learning_rate": 3.912847613272113e-05, + "loss": 1.2237, + "step": 125990 + }, + { + "epoch": 10.91, + "learning_rate": 3.9127609806809324e-05, + "loss": 1.1564, + "step": 126000 + }, + { + "epoch": 10.91, + "learning_rate": 3.912674348089752e-05, + "loss": 1.234, + "step": 126010 + }, + { + "epoch": 10.91, + "learning_rate": 3.9125877154985706e-05, + "loss": 1.1847, + "step": 126020 + }, + { + "epoch": 10.91, + "learning_rate": 3.91250108290739e-05, + "loss": 1.2486, + "step": 126030 + }, + { + "epoch": 10.91, + "learning_rate": 3.912414450316209e-05, + "loss": 1.21, + "step": 126040 + }, + { + "epoch": 10.91, + "learning_rate": 3.912327817725028e-05, + "loss": 1.177, + "step": 126050 + }, + { + "epoch": 10.91, + "learning_rate": 3.912241185133848e-05, + "loss": 1.2178, + "step": 126060 + }, + { + "epoch": 10.91, + "learning_rate": 3.9121545525426664e-05, + "loss": 1.1058, + "step": 126070 + }, + { + "epoch": 10.91, + "learning_rate": 3.912067919951486e-05, + "loss": 1.2339, + "step": 126080 + }, + { + "epoch": 10.91, + "learning_rate": 3.9119812873603053e-05, + "loss": 1.2313, + "step": 126090 + }, + { + "epoch": 10.91, + "learning_rate": 3.911894654769124e-05, + "loss": 1.1737, + "step": 126100 + }, + { + "epoch": 10.92, + "learning_rate": 3.9118080221779436e-05, + "loss": 1.193, + "step": 126110 + }, + { + "epoch": 10.92, + "learning_rate": 3.911721389586763e-05, + "loss": 1.2111, + "step": 126120 + }, + { + "epoch": 10.92, + "learning_rate": 3.911634756995582e-05, + "loss": 1.204, + "step": 126130 + }, + { + "epoch": 10.92, + "learning_rate": 3.911548124404401e-05, + "loss": 1.1138, + "step": 126140 + }, + { + "epoch": 10.92, + "learning_rate": 3.91146149181322e-05, + "loss": 1.2185, + "step": 126150 + }, + { + "epoch": 10.92, + "learning_rate": 3.9113748592220394e-05, + "loss": 1.1896, + "step": 126160 + }, + { + "epoch": 10.92, + "learning_rate": 3.911288226630859e-05, + "loss": 1.2036, + "step": 126170 + }, + { + "epoch": 10.92, + "learning_rate": 3.9112015940396776e-05, + "loss": 1.1839, + "step": 126180 + }, + { + "epoch": 10.92, + "learning_rate": 3.911114961448497e-05, + "loss": 1.1644, + "step": 126190 + }, + { + "epoch": 10.92, + "learning_rate": 3.9110283288573165e-05, + "loss": 1.2037, + "step": 126200 + }, + { + "epoch": 10.92, + "learning_rate": 3.910941696266135e-05, + "loss": 1.2383, + "step": 126210 + }, + { + "epoch": 10.93, + "learning_rate": 3.910855063674955e-05, + "loss": 1.2048, + "step": 126220 + }, + { + "epoch": 10.93, + "learning_rate": 3.910768431083774e-05, + "loss": 1.1874, + "step": 126230 + }, + { + "epoch": 10.93, + "learning_rate": 3.910681798492593e-05, + "loss": 1.2123, + "step": 126240 + }, + { + "epoch": 10.93, + "learning_rate": 3.9105951659014124e-05, + "loss": 1.1547, + "step": 126250 + }, + { + "epoch": 10.93, + "learning_rate": 3.910508533310231e-05, + "loss": 1.2255, + "step": 126260 + }, + { + "epoch": 10.93, + "learning_rate": 3.9104219007190506e-05, + "loss": 1.2091, + "step": 126270 + }, + { + "epoch": 10.93, + "learning_rate": 3.91033526812787e-05, + "loss": 1.1095, + "step": 126280 + }, + { + "epoch": 10.93, + "learning_rate": 3.910248635536689e-05, + "loss": 1.25, + "step": 126290 + }, + { + "epoch": 10.93, + "learning_rate": 3.910162002945508e-05, + "loss": 1.2475, + "step": 126300 + }, + { + "epoch": 10.93, + "learning_rate": 3.910075370354328e-05, + "loss": 1.187, + "step": 126310 + }, + { + "epoch": 10.93, + "learning_rate": 3.9099887377631465e-05, + "loss": 1.2436, + "step": 126320 + }, + { + "epoch": 10.93, + "learning_rate": 3.909902105171966e-05, + "loss": 1.2105, + "step": 126330 + }, + { + "epoch": 10.94, + "learning_rate": 3.9098154725807854e-05, + "loss": 1.1598, + "step": 126340 + }, + { + "epoch": 10.94, + "learning_rate": 3.909728839989604e-05, + "loss": 1.1824, + "step": 126350 + }, + { + "epoch": 10.94, + "learning_rate": 3.9096422073984236e-05, + "loss": 1.2061, + "step": 126360 + }, + { + "epoch": 10.94, + "learning_rate": 3.9095555748072424e-05, + "loss": 1.2504, + "step": 126370 + }, + { + "epoch": 10.94, + "learning_rate": 3.909468942216062e-05, + "loss": 1.1369, + "step": 126380 + }, + { + "epoch": 10.94, + "learning_rate": 3.909382309624881e-05, + "loss": 1.1709, + "step": 126390 + }, + { + "epoch": 10.94, + "learning_rate": 3.9092956770337e-05, + "loss": 1.2342, + "step": 126400 + }, + { + "epoch": 10.94, + "learning_rate": 3.9092090444425195e-05, + "loss": 1.2026, + "step": 126410 + }, + { + "epoch": 10.94, + "learning_rate": 3.909122411851339e-05, + "loss": 1.1879, + "step": 126420 + }, + { + "epoch": 10.94, + "learning_rate": 3.909035779260158e-05, + "loss": 1.2265, + "step": 126430 + }, + { + "epoch": 10.94, + "learning_rate": 3.908949146668977e-05, + "loss": 1.2349, + "step": 126440 + }, + { + "epoch": 10.94, + "learning_rate": 3.9088625140777966e-05, + "loss": 1.197, + "step": 126450 + }, + { + "epoch": 10.95, + "learning_rate": 3.908775881486615e-05, + "loss": 1.1769, + "step": 126460 + }, + { + "epoch": 10.95, + "learning_rate": 3.908689248895435e-05, + "loss": 1.1987, + "step": 126470 + }, + { + "epoch": 10.95, + "learning_rate": 3.9086026163042535e-05, + "loss": 1.2406, + "step": 126480 + }, + { + "epoch": 10.95, + "learning_rate": 3.908515983713073e-05, + "loss": 1.1699, + "step": 126490 + }, + { + "epoch": 10.95, + "learning_rate": 3.9084293511218924e-05, + "loss": 1.156, + "step": 126500 + }, + { + "epoch": 10.95, + "learning_rate": 3.908342718530711e-05, + "loss": 1.2283, + "step": 126510 + }, + { + "epoch": 10.95, + "learning_rate": 3.9082560859395306e-05, + "loss": 1.1976, + "step": 126520 + }, + { + "epoch": 10.95, + "learning_rate": 3.90816945334835e-05, + "loss": 1.2089, + "step": 126530 + }, + { + "epoch": 10.95, + "learning_rate": 3.908082820757169e-05, + "loss": 1.2239, + "step": 126540 + }, + { + "epoch": 10.95, + "learning_rate": 3.907996188165988e-05, + "loss": 1.2591, + "step": 126550 + }, + { + "epoch": 10.95, + "learning_rate": 3.907909555574808e-05, + "loss": 1.1896, + "step": 126560 + }, + { + "epoch": 10.96, + "learning_rate": 3.9078229229836265e-05, + "loss": 1.1979, + "step": 126570 + }, + { + "epoch": 10.96, + "learning_rate": 3.907736290392446e-05, + "loss": 1.2758, + "step": 126580 + }, + { + "epoch": 10.96, + "learning_rate": 3.907649657801265e-05, + "loss": 1.1722, + "step": 126590 + }, + { + "epoch": 10.96, + "learning_rate": 3.907563025210084e-05, + "loss": 1.174, + "step": 126600 + }, + { + "epoch": 10.96, + "learning_rate": 3.9074763926189036e-05, + "loss": 1.1254, + "step": 126610 + }, + { + "epoch": 10.96, + "learning_rate": 3.9073897600277224e-05, + "loss": 1.1872, + "step": 126620 + }, + { + "epoch": 10.96, + "learning_rate": 3.907303127436542e-05, + "loss": 1.2168, + "step": 126630 + }, + { + "epoch": 10.96, + "learning_rate": 3.907216494845361e-05, + "loss": 1.1956, + "step": 126640 + }, + { + "epoch": 10.96, + "learning_rate": 3.90712986225418e-05, + "loss": 1.2324, + "step": 126650 + }, + { + "epoch": 10.96, + "learning_rate": 3.9070432296629995e-05, + "loss": 1.2405, + "step": 126660 + }, + { + "epoch": 10.96, + "learning_rate": 3.906956597071819e-05, + "loss": 1.133, + "step": 126670 + }, + { + "epoch": 10.96, + "learning_rate": 3.906869964480638e-05, + "loss": 1.2138, + "step": 126680 + }, + { + "epoch": 10.97, + "learning_rate": 3.906783331889457e-05, + "loss": 1.1504, + "step": 126690 + }, + { + "epoch": 10.97, + "learning_rate": 3.906696699298276e-05, + "loss": 1.1549, + "step": 126700 + }, + { + "epoch": 10.97, + "learning_rate": 3.9066100667070954e-05, + "loss": 1.1856, + "step": 126710 + }, + { + "epoch": 10.97, + "learning_rate": 3.906523434115915e-05, + "loss": 1.1511, + "step": 126720 + }, + { + "epoch": 10.97, + "learning_rate": 3.9064368015247336e-05, + "loss": 1.1251, + "step": 126730 + }, + { + "epoch": 10.97, + "learning_rate": 3.906350168933553e-05, + "loss": 1.183, + "step": 126740 + }, + { + "epoch": 10.97, + "learning_rate": 3.9062635363423725e-05, + "loss": 1.146, + "step": 126750 + }, + { + "epoch": 10.97, + "learning_rate": 3.906176903751191e-05, + "loss": 1.1744, + "step": 126760 + }, + { + "epoch": 10.97, + "learning_rate": 3.906090271160011e-05, + "loss": 1.2317, + "step": 126770 + }, + { + "epoch": 10.97, + "learning_rate": 3.9060036385688294e-05, + "loss": 1.2077, + "step": 126780 + }, + { + "epoch": 10.97, + "learning_rate": 3.905917005977649e-05, + "loss": 1.1495, + "step": 126790 + }, + { + "epoch": 10.98, + "learning_rate": 3.905830373386468e-05, + "loss": 1.1973, + "step": 126800 + }, + { + "epoch": 10.98, + "learning_rate": 3.905743740795287e-05, + "loss": 1.1597, + "step": 126810 + }, + { + "epoch": 10.98, + "learning_rate": 3.9056571082041066e-05, + "loss": 1.1832, + "step": 126820 + }, + { + "epoch": 10.98, + "learning_rate": 3.905570475612926e-05, + "loss": 1.2262, + "step": 126830 + }, + { + "epoch": 10.98, + "learning_rate": 3.905483843021745e-05, + "loss": 1.1627, + "step": 126840 + }, + { + "epoch": 10.98, + "learning_rate": 3.905397210430564e-05, + "loss": 1.16, + "step": 126850 + }, + { + "epoch": 10.98, + "learning_rate": 3.9053105778393837e-05, + "loss": 1.2536, + "step": 126860 + }, + { + "epoch": 10.98, + "learning_rate": 3.9052239452482024e-05, + "loss": 1.1927, + "step": 126870 + }, + { + "epoch": 10.98, + "learning_rate": 3.905137312657022e-05, + "loss": 1.2269, + "step": 126880 + }, + { + "epoch": 10.98, + "learning_rate": 3.9050506800658406e-05, + "loss": 1.1878, + "step": 126890 + }, + { + "epoch": 10.98, + "learning_rate": 3.90496404747466e-05, + "loss": 1.1761, + "step": 126900 + }, + { + "epoch": 10.98, + "learning_rate": 3.9048774148834795e-05, + "loss": 1.2413, + "step": 126910 + }, + { + "epoch": 10.99, + "learning_rate": 3.904790782292298e-05, + "loss": 1.1975, + "step": 126920 + }, + { + "epoch": 10.99, + "learning_rate": 3.904704149701118e-05, + "loss": 1.2396, + "step": 126930 + }, + { + "epoch": 10.99, + "learning_rate": 3.904617517109937e-05, + "loss": 1.2091, + "step": 126940 + }, + { + "epoch": 10.99, + "learning_rate": 3.904530884518756e-05, + "loss": 1.212, + "step": 126950 + }, + { + "epoch": 10.99, + "learning_rate": 3.9044442519275754e-05, + "loss": 1.1862, + "step": 126960 + }, + { + "epoch": 10.99, + "learning_rate": 3.904357619336395e-05, + "loss": 1.2464, + "step": 126970 + }, + { + "epoch": 10.99, + "learning_rate": 3.9042709867452136e-05, + "loss": 1.2337, + "step": 126980 + }, + { + "epoch": 10.99, + "learning_rate": 3.904184354154033e-05, + "loss": 1.1654, + "step": 126990 + }, + { + "epoch": 10.99, + "learning_rate": 3.904097721562852e-05, + "loss": 1.2269, + "step": 127000 + }, + { + "epoch": 10.99, + "learning_rate": 3.904011088971671e-05, + "loss": 1.207, + "step": 127010 + }, + { + "epoch": 10.99, + "learning_rate": 3.903924456380491e-05, + "loss": 1.2253, + "step": 127020 + }, + { + "epoch": 11.0, + "learning_rate": 3.9038378237893095e-05, + "loss": 1.1928, + "step": 127030 + }, + { + "epoch": 11.0, + "learning_rate": 3.903751191198129e-05, + "loss": 1.1742, + "step": 127040 + }, + { + "epoch": 11.0, + "learning_rate": 3.9036645586069484e-05, + "loss": 1.1806, + "step": 127050 + }, + { + "epoch": 11.0, + "learning_rate": 3.903577926015767e-05, + "loss": 1.1095, + "step": 127060 + }, + { + "epoch": 11.0, + "learning_rate": 3.9034912934245866e-05, + "loss": 1.1388, + "step": 127070 + }, + { + "epoch": 11.0, + "learning_rate": 3.903404660833406e-05, + "loss": 1.2309, + "step": 127080 + }, + { + "epoch": 11.0, + "eval_Bleu_1": 0.03691568745175338, + "eval_Bleu_2": 2.635171648304036e-11, + "eval_Bleu_3": 2.4356639889426694e-14, + "eval_Bleu_4": 7.615430094915245e-16, + "eval_ROUGE_L": 0.0801855534905433, + "eval_cer": 0.9936494972121199, + "eval_em": 0, + "eval_f1": 0.08977498046639625, + "eval_loss": 1.1203044652938843, + "eval_runtime": 1713.6097, + "eval_samples_per_second": 2.997, + "eval_steps_per_second": 2.997, + "eval_wer": 0.9740680599073558, + "step": 127085 + }, + { + "epoch": 11.0, + "learning_rate": 3.903318028242225e-05, + "loss": 1.1793, + "step": 127090 + }, + { + "epoch": 11.0, + "learning_rate": 3.903231395651044e-05, + "loss": 1.1256, + "step": 127100 + }, + { + "epoch": 11.0, + "learning_rate": 3.903144763059863e-05, + "loss": 1.1859, + "step": 127110 + }, + { + "epoch": 11.0, + "learning_rate": 3.9030581304686825e-05, + "loss": 1.1684, + "step": 127120 + }, + { + "epoch": 11.0, + "learning_rate": 3.902971497877502e-05, + "loss": 1.2108, + "step": 127130 + }, + { + "epoch": 11.0, + "learning_rate": 3.902884865286321e-05, + "loss": 1.1567, + "step": 127140 + }, + { + "epoch": 11.01, + "learning_rate": 3.90279823269514e-05, + "loss": 1.1352, + "step": 127150 + }, + { + "epoch": 11.01, + "learning_rate": 3.9027116001039596e-05, + "loss": 1.2146, + "step": 127160 + }, + { + "epoch": 11.01, + "learning_rate": 3.902624967512778e-05, + "loss": 1.2019, + "step": 127170 + }, + { + "epoch": 11.01, + "learning_rate": 3.902538334921598e-05, + "loss": 1.1678, + "step": 127180 + }, + { + "epoch": 11.01, + "learning_rate": 3.902451702330417e-05, + "loss": 1.118, + "step": 127190 + }, + { + "epoch": 11.01, + "learning_rate": 3.902365069739236e-05, + "loss": 1.1784, + "step": 127200 + }, + { + "epoch": 11.01, + "learning_rate": 3.9022784371480554e-05, + "loss": 1.2967, + "step": 127210 + }, + { + "epoch": 11.01, + "learning_rate": 3.902191804556874e-05, + "loss": 1.1518, + "step": 127220 + }, + { + "epoch": 11.01, + "learning_rate": 3.9021051719656936e-05, + "loss": 1.2055, + "step": 127230 + }, + { + "epoch": 11.01, + "learning_rate": 3.902018539374513e-05, + "loss": 1.172, + "step": 127240 + }, + { + "epoch": 11.01, + "learning_rate": 3.901931906783332e-05, + "loss": 1.2357, + "step": 127250 + }, + { + "epoch": 11.02, + "learning_rate": 3.901845274192151e-05, + "loss": 1.2077, + "step": 127260 + }, + { + "epoch": 11.02, + "learning_rate": 3.901758641600971e-05, + "loss": 1.1609, + "step": 127270 + }, + { + "epoch": 11.02, + "learning_rate": 3.9016720090097895e-05, + "loss": 1.1509, + "step": 127280 + }, + { + "epoch": 11.02, + "learning_rate": 3.901585376418609e-05, + "loss": 1.1946, + "step": 127290 + }, + { + "epoch": 11.02, + "learning_rate": 3.9014987438274284e-05, + "loss": 1.2156, + "step": 127300 + }, + { + "epoch": 11.02, + "learning_rate": 3.901412111236247e-05, + "loss": 1.1933, + "step": 127310 + }, + { + "epoch": 11.02, + "learning_rate": 3.9013254786450666e-05, + "loss": 1.1729, + "step": 127320 + }, + { + "epoch": 11.02, + "learning_rate": 3.9012388460538854e-05, + "loss": 1.1489, + "step": 127330 + }, + { + "epoch": 11.02, + "learning_rate": 3.901152213462705e-05, + "loss": 1.1429, + "step": 127340 + }, + { + "epoch": 11.02, + "learning_rate": 3.901065580871524e-05, + "loss": 1.2344, + "step": 127350 + }, + { + "epoch": 11.02, + "learning_rate": 3.900978948280343e-05, + "loss": 1.1969, + "step": 127360 + }, + { + "epoch": 11.02, + "learning_rate": 3.9008923156891625e-05, + "loss": 1.16, + "step": 127370 + }, + { + "epoch": 11.03, + "learning_rate": 3.900805683097982e-05, + "loss": 1.1907, + "step": 127380 + }, + { + "epoch": 11.03, + "learning_rate": 3.900719050506801e-05, + "loss": 1.1324, + "step": 127390 + }, + { + "epoch": 11.03, + "learning_rate": 3.90063241791562e-05, + "loss": 1.2298, + "step": 127400 + }, + { + "epoch": 11.03, + "learning_rate": 3.9005457853244396e-05, + "loss": 1.2356, + "step": 127410 + }, + { + "epoch": 11.03, + "learning_rate": 3.9004591527332584e-05, + "loss": 1.2011, + "step": 127420 + }, + { + "epoch": 11.03, + "learning_rate": 3.900372520142078e-05, + "loss": 1.2352, + "step": 127430 + }, + { + "epoch": 11.03, + "learning_rate": 3.9002858875508966e-05, + "loss": 1.197, + "step": 127440 + }, + { + "epoch": 11.03, + "learning_rate": 3.900199254959716e-05, + "loss": 1.1953, + "step": 127450 + }, + { + "epoch": 11.03, + "learning_rate": 3.9001126223685355e-05, + "loss": 1.1315, + "step": 127460 + }, + { + "epoch": 11.03, + "learning_rate": 3.900025989777354e-05, + "loss": 1.1756, + "step": 127470 + }, + { + "epoch": 11.03, + "learning_rate": 3.899939357186174e-05, + "loss": 1.2364, + "step": 127480 + }, + { + "epoch": 11.04, + "learning_rate": 3.899852724594993e-05, + "loss": 1.1814, + "step": 127490 + }, + { + "epoch": 11.04, + "learning_rate": 3.899766092003812e-05, + "loss": 1.2331, + "step": 127500 + }, + { + "epoch": 11.04, + "learning_rate": 3.899679459412631e-05, + "loss": 1.1963, + "step": 127510 + }, + { + "epoch": 11.04, + "learning_rate": 3.89959282682145e-05, + "loss": 1.2126, + "step": 127520 + }, + { + "epoch": 11.04, + "learning_rate": 3.8995061942302695e-05, + "loss": 1.2068, + "step": 127530 + }, + { + "epoch": 11.04, + "learning_rate": 3.899419561639089e-05, + "loss": 1.2199, + "step": 127540 + }, + { + "epoch": 11.04, + "learning_rate": 3.899332929047908e-05, + "loss": 1.1628, + "step": 127550 + }, + { + "epoch": 11.04, + "learning_rate": 3.899246296456727e-05, + "loss": 1.2264, + "step": 127560 + }, + { + "epoch": 11.04, + "learning_rate": 3.8991596638655467e-05, + "loss": 1.1647, + "step": 127570 + }, + { + "epoch": 11.04, + "learning_rate": 3.8990730312743654e-05, + "loss": 1.17, + "step": 127580 + }, + { + "epoch": 11.04, + "learning_rate": 3.898986398683185e-05, + "loss": 1.1357, + "step": 127590 + }, + { + "epoch": 11.04, + "learning_rate": 3.898899766092004e-05, + "loss": 1.13, + "step": 127600 + }, + { + "epoch": 11.05, + "learning_rate": 3.898813133500823e-05, + "loss": 1.1737, + "step": 127610 + }, + { + "epoch": 11.05, + "learning_rate": 3.8987265009096425e-05, + "loss": 1.1432, + "step": 127620 + }, + { + "epoch": 11.05, + "learning_rate": 3.898639868318461e-05, + "loss": 1.1644, + "step": 127630 + }, + { + "epoch": 11.05, + "learning_rate": 3.898553235727281e-05, + "loss": 1.1407, + "step": 127640 + }, + { + "epoch": 11.05, + "learning_rate": 3.8984666031361e-05, + "loss": 1.191, + "step": 127650 + }, + { + "epoch": 11.05, + "learning_rate": 3.898379970544919e-05, + "loss": 1.225, + "step": 127660 + }, + { + "epoch": 11.05, + "learning_rate": 3.8982933379537384e-05, + "loss": 1.1679, + "step": 127670 + }, + { + "epoch": 11.05, + "learning_rate": 3.898206705362558e-05, + "loss": 1.0736, + "step": 127680 + }, + { + "epoch": 11.05, + "learning_rate": 3.8981200727713766e-05, + "loss": 1.2233, + "step": 127690 + }, + { + "epoch": 11.05, + "learning_rate": 3.898033440180196e-05, + "loss": 1.1546, + "step": 127700 + }, + { + "epoch": 11.05, + "learning_rate": 3.8979468075890155e-05, + "loss": 1.2046, + "step": 127710 + }, + { + "epoch": 11.05, + "learning_rate": 3.897860174997834e-05, + "loss": 1.2117, + "step": 127720 + }, + { + "epoch": 11.06, + "learning_rate": 3.897773542406654e-05, + "loss": 1.1765, + "step": 127730 + }, + { + "epoch": 11.06, + "learning_rate": 3.8976869098154725e-05, + "loss": 1.1839, + "step": 127740 + }, + { + "epoch": 11.06, + "learning_rate": 3.897600277224292e-05, + "loss": 1.1253, + "step": 127750 + }, + { + "epoch": 11.06, + "learning_rate": 3.8975136446331114e-05, + "loss": 1.184, + "step": 127760 + }, + { + "epoch": 11.06, + "learning_rate": 3.89742701204193e-05, + "loss": 1.1509, + "step": 127770 + }, + { + "epoch": 11.06, + "learning_rate": 3.8973403794507496e-05, + "loss": 1.1844, + "step": 127780 + }, + { + "epoch": 11.06, + "learning_rate": 3.897253746859569e-05, + "loss": 1.1581, + "step": 127790 + }, + { + "epoch": 11.06, + "learning_rate": 3.897167114268388e-05, + "loss": 1.1089, + "step": 127800 + }, + { + "epoch": 11.06, + "learning_rate": 3.897080481677207e-05, + "loss": 1.211, + "step": 127810 + }, + { + "epoch": 11.06, + "learning_rate": 3.896993849086027e-05, + "loss": 1.1798, + "step": 127820 + }, + { + "epoch": 11.06, + "learning_rate": 3.8969072164948455e-05, + "loss": 1.1309, + "step": 127830 + }, + { + "epoch": 11.07, + "learning_rate": 3.896820583903665e-05, + "loss": 1.1777, + "step": 127840 + }, + { + "epoch": 11.07, + "learning_rate": 3.896733951312484e-05, + "loss": 1.2213, + "step": 127850 + }, + { + "epoch": 11.07, + "learning_rate": 3.896647318721303e-05, + "loss": 1.1591, + "step": 127860 + }, + { + "epoch": 11.07, + "learning_rate": 3.8965606861301226e-05, + "loss": 1.209, + "step": 127870 + }, + { + "epoch": 11.07, + "learning_rate": 3.896474053538941e-05, + "loss": 1.1513, + "step": 127880 + }, + { + "epoch": 11.07, + "learning_rate": 3.896387420947761e-05, + "loss": 1.2116, + "step": 127890 + }, + { + "epoch": 11.07, + "learning_rate": 3.89630078835658e-05, + "loss": 1.2222, + "step": 127900 + }, + { + "epoch": 11.07, + "learning_rate": 3.896214155765399e-05, + "loss": 1.1708, + "step": 127910 + }, + { + "epoch": 11.07, + "learning_rate": 3.8961275231742184e-05, + "loss": 1.216, + "step": 127920 + }, + { + "epoch": 11.07, + "learning_rate": 3.896040890583038e-05, + "loss": 1.1961, + "step": 127930 + }, + { + "epoch": 11.07, + "learning_rate": 3.8959542579918566e-05, + "loss": 1.1531, + "step": 127940 + }, + { + "epoch": 11.07, + "learning_rate": 3.895867625400676e-05, + "loss": 1.1894, + "step": 127950 + }, + { + "epoch": 11.08, + "learning_rate": 3.895780992809495e-05, + "loss": 1.1243, + "step": 127960 + }, + { + "epoch": 11.08, + "learning_rate": 3.895694360218314e-05, + "loss": 1.1515, + "step": 127970 + }, + { + "epoch": 11.08, + "learning_rate": 3.895607727627134e-05, + "loss": 1.1478, + "step": 127980 + }, + { + "epoch": 11.08, + "learning_rate": 3.8955210950359525e-05, + "loss": 1.1583, + "step": 127990 + }, + { + "epoch": 11.08, + "learning_rate": 3.895434462444772e-05, + "loss": 1.1911, + "step": 128000 + }, + { + "epoch": 11.08, + "learning_rate": 3.8953478298535914e-05, + "loss": 1.2014, + "step": 128010 + }, + { + "epoch": 11.08, + "learning_rate": 3.89526119726241e-05, + "loss": 1.1626, + "step": 128020 + }, + { + "epoch": 11.08, + "learning_rate": 3.8951745646712296e-05, + "loss": 1.1514, + "step": 128030 + }, + { + "epoch": 11.08, + "learning_rate": 3.895087932080049e-05, + "loss": 1.2001, + "step": 128040 + }, + { + "epoch": 11.08, + "learning_rate": 3.895001299488868e-05, + "loss": 1.198, + "step": 128050 + }, + { + "epoch": 11.08, + "learning_rate": 3.894914666897687e-05, + "loss": 1.194, + "step": 128060 + }, + { + "epoch": 11.09, + "learning_rate": 3.894828034306506e-05, + "loss": 1.193, + "step": 128070 + }, + { + "epoch": 11.09, + "learning_rate": 3.8947414017153255e-05, + "loss": 1.1684, + "step": 128080 + }, + { + "epoch": 11.09, + "learning_rate": 3.894654769124145e-05, + "loss": 1.2225, + "step": 128090 + }, + { + "epoch": 11.09, + "learning_rate": 3.894568136532964e-05, + "loss": 1.1912, + "step": 128100 + }, + { + "epoch": 11.09, + "learning_rate": 3.894481503941783e-05, + "loss": 1.1893, + "step": 128110 + }, + { + "epoch": 11.09, + "learning_rate": 3.8943948713506026e-05, + "loss": 1.2141, + "step": 128120 + }, + { + "epoch": 11.09, + "learning_rate": 3.8943082387594214e-05, + "loss": 1.2136, + "step": 128130 + }, + { + "epoch": 11.09, + "learning_rate": 3.894221606168241e-05, + "loss": 1.2192, + "step": 128140 + }, + { + "epoch": 11.09, + "learning_rate": 3.89413497357706e-05, + "loss": 1.1975, + "step": 128150 + }, + { + "epoch": 11.09, + "learning_rate": 3.894048340985879e-05, + "loss": 1.2225, + "step": 128160 + }, + { + "epoch": 11.09, + "learning_rate": 3.8939617083946985e-05, + "loss": 1.156, + "step": 128170 + }, + { + "epoch": 11.09, + "learning_rate": 3.893875075803517e-05, + "loss": 1.1779, + "step": 128180 + }, + { + "epoch": 11.1, + "learning_rate": 3.893788443212337e-05, + "loss": 1.2155, + "step": 128190 + }, + { + "epoch": 11.1, + "learning_rate": 3.893701810621156e-05, + "loss": 1.134, + "step": 128200 + }, + { + "epoch": 11.1, + "learning_rate": 3.893615178029975e-05, + "loss": 1.1584, + "step": 128210 + }, + { + "epoch": 11.1, + "learning_rate": 3.893528545438794e-05, + "loss": 1.1759, + "step": 128220 + }, + { + "epoch": 11.1, + "learning_rate": 3.893441912847614e-05, + "loss": 1.1603, + "step": 128230 + }, + { + "epoch": 11.1, + "learning_rate": 3.8933552802564325e-05, + "loss": 1.152, + "step": 128240 + }, + { + "epoch": 11.1, + "learning_rate": 3.893268647665252e-05, + "loss": 1.21, + "step": 128250 + }, + { + "epoch": 11.1, + "learning_rate": 3.8931820150740714e-05, + "loss": 1.1452, + "step": 128260 + }, + { + "epoch": 11.1, + "learning_rate": 3.89309538248289e-05, + "loss": 1.1688, + "step": 128270 + }, + { + "epoch": 11.1, + "learning_rate": 3.8930087498917096e-05, + "loss": 1.178, + "step": 128280 + }, + { + "epoch": 11.1, + "learning_rate": 3.8929221173005284e-05, + "loss": 1.1384, + "step": 128290 + }, + { + "epoch": 11.11, + "learning_rate": 3.892835484709348e-05, + "loss": 1.1638, + "step": 128300 + }, + { + "epoch": 11.11, + "learning_rate": 3.892748852118167e-05, + "loss": 1.188, + "step": 128310 + }, + { + "epoch": 11.11, + "learning_rate": 3.892662219526986e-05, + "loss": 1.2044, + "step": 128320 + }, + { + "epoch": 11.11, + "learning_rate": 3.8925755869358055e-05, + "loss": 1.1983, + "step": 128330 + }, + { + "epoch": 11.11, + "learning_rate": 3.892488954344625e-05, + "loss": 1.169, + "step": 128340 + }, + { + "epoch": 11.11, + "learning_rate": 3.892402321753444e-05, + "loss": 1.2196, + "step": 128350 + }, + { + "epoch": 11.11, + "learning_rate": 3.892315689162263e-05, + "loss": 1.2261, + "step": 128360 + }, + { + "epoch": 11.11, + "learning_rate": 3.892229056571082e-05, + "loss": 1.2145, + "step": 128370 + }, + { + "epoch": 11.11, + "learning_rate": 3.8921424239799014e-05, + "loss": 1.2569, + "step": 128380 + }, + { + "epoch": 11.11, + "learning_rate": 3.892055791388721e-05, + "loss": 1.165, + "step": 128390 + }, + { + "epoch": 11.11, + "learning_rate": 3.8919691587975396e-05, + "loss": 1.2065, + "step": 128400 + }, + { + "epoch": 11.11, + "learning_rate": 3.891882526206359e-05, + "loss": 1.1361, + "step": 128410 + }, + { + "epoch": 11.12, + "learning_rate": 3.8917958936151785e-05, + "loss": 1.148, + "step": 128420 + }, + { + "epoch": 11.12, + "learning_rate": 3.891709261023997e-05, + "loss": 1.1522, + "step": 128430 + }, + { + "epoch": 11.12, + "learning_rate": 3.891622628432817e-05, + "loss": 1.226, + "step": 128440 + }, + { + "epoch": 11.12, + "learning_rate": 3.891535995841636e-05, + "loss": 1.1318, + "step": 128450 + }, + { + "epoch": 11.12, + "learning_rate": 3.891449363250455e-05, + "loss": 1.1832, + "step": 128460 + }, + { + "epoch": 11.12, + "learning_rate": 3.8913627306592744e-05, + "loss": 1.166, + "step": 128470 + }, + { + "epoch": 11.12, + "learning_rate": 3.891276098068093e-05, + "loss": 1.1728, + "step": 128480 + }, + { + "epoch": 11.12, + "learning_rate": 3.8911894654769126e-05, + "loss": 1.1832, + "step": 128490 + }, + { + "epoch": 11.12, + "learning_rate": 3.891102832885732e-05, + "loss": 1.1971, + "step": 128500 + }, + { + "epoch": 11.12, + "learning_rate": 3.891016200294551e-05, + "loss": 1.2111, + "step": 128510 + }, + { + "epoch": 11.12, + "learning_rate": 3.89092956770337e-05, + "loss": 1.1865, + "step": 128520 + }, + { + "epoch": 11.13, + "learning_rate": 3.89084293511219e-05, + "loss": 1.1567, + "step": 128530 + }, + { + "epoch": 11.13, + "learning_rate": 3.8907563025210084e-05, + "loss": 1.1561, + "step": 128540 + }, + { + "epoch": 11.13, + "learning_rate": 3.890669669929828e-05, + "loss": 1.1432, + "step": 128550 + }, + { + "epoch": 11.13, + "learning_rate": 3.8905830373386473e-05, + "loss": 1.1902, + "step": 128560 + }, + { + "epoch": 11.13, + "learning_rate": 3.890496404747466e-05, + "loss": 1.198, + "step": 128570 + }, + { + "epoch": 11.13, + "learning_rate": 3.8904097721562856e-05, + "loss": 1.1884, + "step": 128580 + }, + { + "epoch": 11.13, + "learning_rate": 3.890323139565104e-05, + "loss": 1.211, + "step": 128590 + }, + { + "epoch": 11.13, + "learning_rate": 3.890236506973924e-05, + "loss": 1.0881, + "step": 128600 + }, + { + "epoch": 11.13, + "learning_rate": 3.890149874382743e-05, + "loss": 1.1872, + "step": 128610 + }, + { + "epoch": 11.13, + "learning_rate": 3.890063241791562e-05, + "loss": 1.1602, + "step": 128620 + }, + { + "epoch": 11.13, + "learning_rate": 3.8899766092003814e-05, + "loss": 1.197, + "step": 128630 + }, + { + "epoch": 11.13, + "learning_rate": 3.889889976609201e-05, + "loss": 1.1848, + "step": 128640 + }, + { + "epoch": 11.14, + "learning_rate": 3.8898033440180196e-05, + "loss": 1.1108, + "step": 128650 + }, + { + "epoch": 11.14, + "learning_rate": 3.889716711426839e-05, + "loss": 1.2236, + "step": 128660 + }, + { + "epoch": 11.14, + "learning_rate": 3.8896300788356585e-05, + "loss": 1.1286, + "step": 128670 + }, + { + "epoch": 11.14, + "learning_rate": 3.889543446244477e-05, + "loss": 1.167, + "step": 128680 + }, + { + "epoch": 11.14, + "learning_rate": 3.889456813653297e-05, + "loss": 1.2205, + "step": 128690 + }, + { + "epoch": 11.14, + "learning_rate": 3.8893701810621155e-05, + "loss": 1.1247, + "step": 128700 + }, + { + "epoch": 11.14, + "learning_rate": 3.889283548470935e-05, + "loss": 1.1638, + "step": 128710 + }, + { + "epoch": 11.14, + "learning_rate": 3.8891969158797544e-05, + "loss": 1.2423, + "step": 128720 + }, + { + "epoch": 11.14, + "learning_rate": 3.889110283288573e-05, + "loss": 1.1898, + "step": 128730 + }, + { + "epoch": 11.14, + "learning_rate": 3.8890236506973926e-05, + "loss": 1.1375, + "step": 128740 + }, + { + "epoch": 11.14, + "learning_rate": 3.888937018106212e-05, + "loss": 1.1249, + "step": 128750 + }, + { + "epoch": 11.14, + "learning_rate": 3.888850385515031e-05, + "loss": 1.1069, + "step": 128760 + }, + { + "epoch": 11.15, + "learning_rate": 3.88876375292385e-05, + "loss": 1.1493, + "step": 128770 + }, + { + "epoch": 11.15, + "learning_rate": 3.88867712033267e-05, + "loss": 1.1125, + "step": 128780 + }, + { + "epoch": 11.15, + "learning_rate": 3.8885904877414885e-05, + "loss": 1.1843, + "step": 128790 + }, + { + "epoch": 11.15, + "learning_rate": 3.888503855150308e-05, + "loss": 1.2736, + "step": 128800 + }, + { + "epoch": 11.15, + "learning_rate": 3.888417222559127e-05, + "loss": 1.1871, + "step": 128810 + }, + { + "epoch": 11.15, + "learning_rate": 3.888330589967946e-05, + "loss": 1.1902, + "step": 128820 + }, + { + "epoch": 11.15, + "learning_rate": 3.8882439573767656e-05, + "loss": 1.1721, + "step": 128830 + }, + { + "epoch": 11.15, + "learning_rate": 3.8881573247855844e-05, + "loss": 1.2007, + "step": 128840 + }, + { + "epoch": 11.15, + "learning_rate": 3.888070692194404e-05, + "loss": 1.1384, + "step": 128850 + }, + { + "epoch": 11.15, + "learning_rate": 3.887984059603223e-05, + "loss": 1.2005, + "step": 128860 + }, + { + "epoch": 11.15, + "learning_rate": 3.887897427012042e-05, + "loss": 1.2049, + "step": 128870 + }, + { + "epoch": 11.16, + "learning_rate": 3.8878107944208615e-05, + "loss": 1.1938, + "step": 128880 + }, + { + "epoch": 11.16, + "learning_rate": 3.887724161829681e-05, + "loss": 1.2557, + "step": 128890 + }, + { + "epoch": 11.16, + "learning_rate": 3.8876375292385e-05, + "loss": 1.2302, + "step": 128900 + }, + { + "epoch": 11.16, + "learning_rate": 3.887550896647319e-05, + "loss": 1.1542, + "step": 128910 + }, + { + "epoch": 11.16, + "learning_rate": 3.887464264056138e-05, + "loss": 1.1854, + "step": 128920 + }, + { + "epoch": 11.16, + "learning_rate": 3.887377631464957e-05, + "loss": 1.1172, + "step": 128930 + }, + { + "epoch": 11.16, + "learning_rate": 3.887290998873777e-05, + "loss": 1.082, + "step": 128940 + }, + { + "epoch": 11.16, + "learning_rate": 3.8872043662825955e-05, + "loss": 1.1879, + "step": 128950 + }, + { + "epoch": 11.16, + "learning_rate": 3.887117733691415e-05, + "loss": 1.2246, + "step": 128960 + }, + { + "epoch": 11.16, + "learning_rate": 3.8870311011002344e-05, + "loss": 1.1812, + "step": 128970 + }, + { + "epoch": 11.16, + "learning_rate": 3.886944468509053e-05, + "loss": 1.1743, + "step": 128980 + }, + { + "epoch": 11.16, + "learning_rate": 3.8868578359178726e-05, + "loss": 1.1327, + "step": 128990 + }, + { + "epoch": 11.17, + "learning_rate": 3.886771203326692e-05, + "loss": 1.1999, + "step": 129000 + }, + { + "epoch": 11.17, + "learning_rate": 3.886684570735511e-05, + "loss": 1.1455, + "step": 129010 + }, + { + "epoch": 11.17, + "learning_rate": 3.88659793814433e-05, + "loss": 1.1985, + "step": 129020 + }, + { + "epoch": 11.17, + "learning_rate": 3.886511305553149e-05, + "loss": 1.1519, + "step": 129030 + }, + { + "epoch": 11.17, + "learning_rate": 3.8864246729619685e-05, + "loss": 1.2486, + "step": 129040 + }, + { + "epoch": 11.17, + "learning_rate": 3.886338040370788e-05, + "loss": 1.1428, + "step": 129050 + }, + { + "epoch": 11.17, + "learning_rate": 3.886251407779607e-05, + "loss": 1.1048, + "step": 129060 + }, + { + "epoch": 11.17, + "learning_rate": 3.886164775188426e-05, + "loss": 1.1353, + "step": 129070 + }, + { + "epoch": 11.17, + "learning_rate": 3.8860781425972456e-05, + "loss": 1.1877, + "step": 129080 + }, + { + "epoch": 11.17, + "learning_rate": 3.8859915100060644e-05, + "loss": 1.1738, + "step": 129090 + }, + { + "epoch": 11.17, + "learning_rate": 3.885904877414884e-05, + "loss": 1.1308, + "step": 129100 + }, + { + "epoch": 11.18, + "learning_rate": 3.8858182448237026e-05, + "loss": 1.1992, + "step": 129110 + }, + { + "epoch": 11.18, + "learning_rate": 3.885731612232522e-05, + "loss": 1.1836, + "step": 129120 + }, + { + "epoch": 11.18, + "learning_rate": 3.8856449796413415e-05, + "loss": 1.2105, + "step": 129130 + }, + { + "epoch": 11.18, + "learning_rate": 3.88555834705016e-05, + "loss": 1.1702, + "step": 129140 + }, + { + "epoch": 11.18, + "learning_rate": 3.88547171445898e-05, + "loss": 1.1825, + "step": 129150 + }, + { + "epoch": 11.18, + "learning_rate": 3.885385081867799e-05, + "loss": 1.1399, + "step": 129160 + }, + { + "epoch": 11.18, + "learning_rate": 3.885298449276618e-05, + "loss": 1.1055, + "step": 129170 + }, + { + "epoch": 11.18, + "learning_rate": 3.8852118166854374e-05, + "loss": 1.17, + "step": 129180 + }, + { + "epoch": 11.18, + "learning_rate": 3.885125184094257e-05, + "loss": 1.2319, + "step": 129190 + }, + { + "epoch": 11.18, + "learning_rate": 3.8850385515030756e-05, + "loss": 1.1502, + "step": 129200 + }, + { + "epoch": 11.18, + "learning_rate": 3.884951918911895e-05, + "loss": 1.1626, + "step": 129210 + }, + { + "epoch": 11.18, + "learning_rate": 3.884865286320714e-05, + "loss": 1.171, + "step": 129220 + }, + { + "epoch": 11.19, + "learning_rate": 3.884778653729533e-05, + "loss": 1.1836, + "step": 129230 + }, + { + "epoch": 11.19, + "learning_rate": 3.884692021138353e-05, + "loss": 1.1732, + "step": 129240 + }, + { + "epoch": 11.19, + "learning_rate": 3.8846053885471714e-05, + "loss": 1.2309, + "step": 129250 + }, + { + "epoch": 11.19, + "learning_rate": 3.884518755955991e-05, + "loss": 1.1997, + "step": 129260 + }, + { + "epoch": 11.19, + "learning_rate": 3.88443212336481e-05, + "loss": 1.1506, + "step": 129270 + }, + { + "epoch": 11.19, + "learning_rate": 3.884345490773629e-05, + "loss": 1.1937, + "step": 129280 + }, + { + "epoch": 11.19, + "learning_rate": 3.8842588581824486e-05, + "loss": 1.2269, + "step": 129290 + }, + { + "epoch": 11.19, + "learning_rate": 3.884172225591268e-05, + "loss": 1.2113, + "step": 129300 + }, + { + "epoch": 11.19, + "learning_rate": 3.884085593000087e-05, + "loss": 1.1977, + "step": 129310 + }, + { + "epoch": 11.19, + "learning_rate": 3.883998960408906e-05, + "loss": 1.134, + "step": 129320 + }, + { + "epoch": 11.19, + "learning_rate": 3.883912327817725e-05, + "loss": 1.1782, + "step": 129330 + }, + { + "epoch": 11.2, + "learning_rate": 3.8838256952265444e-05, + "loss": 1.253, + "step": 129340 + }, + { + "epoch": 11.2, + "learning_rate": 3.883739062635364e-05, + "loss": 1.0774, + "step": 129350 + }, + { + "epoch": 11.2, + "learning_rate": 3.8836524300441826e-05, + "loss": 1.135, + "step": 129360 + }, + { + "epoch": 11.2, + "learning_rate": 3.883565797453002e-05, + "loss": 1.1161, + "step": 129370 + }, + { + "epoch": 11.2, + "learning_rate": 3.8834791648618215e-05, + "loss": 1.1984, + "step": 129380 + }, + { + "epoch": 11.2, + "learning_rate": 3.88339253227064e-05, + "loss": 1.2438, + "step": 129390 + }, + { + "epoch": 11.2, + "learning_rate": 3.88330589967946e-05, + "loss": 1.1214, + "step": 129400 + }, + { + "epoch": 11.2, + "learning_rate": 3.883219267088279e-05, + "loss": 1.2257, + "step": 129410 + }, + { + "epoch": 11.2, + "learning_rate": 3.883132634497098e-05, + "loss": 1.2026, + "step": 129420 + }, + { + "epoch": 11.2, + "learning_rate": 3.8830460019059174e-05, + "loss": 1.1606, + "step": 129430 + }, + { + "epoch": 11.2, + "learning_rate": 3.882959369314736e-05, + "loss": 1.1945, + "step": 129440 + }, + { + "epoch": 11.2, + "learning_rate": 3.8828727367235556e-05, + "loss": 1.1706, + "step": 129450 + }, + { + "epoch": 11.21, + "learning_rate": 3.882786104132375e-05, + "loss": 1.1847, + "step": 129460 + }, + { + "epoch": 11.21, + "learning_rate": 3.882699471541194e-05, + "loss": 1.1477, + "step": 129470 + }, + { + "epoch": 11.21, + "learning_rate": 3.882612838950013e-05, + "loss": 1.1958, + "step": 129480 + }, + { + "epoch": 11.21, + "learning_rate": 3.882526206358833e-05, + "loss": 1.1857, + "step": 129490 + }, + { + "epoch": 11.21, + "learning_rate": 3.8824395737676515e-05, + "loss": 1.201, + "step": 129500 + }, + { + "epoch": 11.21, + "learning_rate": 3.882352941176471e-05, + "loss": 1.1895, + "step": 129510 + }, + { + "epoch": 11.21, + "learning_rate": 3.8822663085852904e-05, + "loss": 1.1375, + "step": 129520 + }, + { + "epoch": 11.21, + "learning_rate": 3.882179675994109e-05, + "loss": 1.1749, + "step": 129530 + }, + { + "epoch": 11.21, + "learning_rate": 3.8820930434029286e-05, + "loss": 1.2015, + "step": 129540 + }, + { + "epoch": 11.21, + "learning_rate": 3.8820064108117474e-05, + "loss": 1.1819, + "step": 129550 + }, + { + "epoch": 11.21, + "learning_rate": 3.881919778220567e-05, + "loss": 1.1664, + "step": 129560 + }, + { + "epoch": 11.22, + "learning_rate": 3.881833145629386e-05, + "loss": 1.1793, + "step": 129570 + }, + { + "epoch": 11.22, + "learning_rate": 3.881746513038205e-05, + "loss": 1.1953, + "step": 129580 + }, + { + "epoch": 11.22, + "learning_rate": 3.8816598804470245e-05, + "loss": 1.1177, + "step": 129590 + }, + { + "epoch": 11.22, + "learning_rate": 3.881573247855844e-05, + "loss": 1.171, + "step": 129600 + }, + { + "epoch": 11.22, + "learning_rate": 3.881486615264663e-05, + "loss": 1.1759, + "step": 129610 + }, + { + "epoch": 11.22, + "learning_rate": 3.881399982673482e-05, + "loss": 1.1964, + "step": 129620 + }, + { + "epoch": 11.22, + "learning_rate": 3.8813133500823016e-05, + "loss": 1.1753, + "step": 129630 + }, + { + "epoch": 11.22, + "learning_rate": 3.88122671749112e-05, + "loss": 1.2551, + "step": 129640 + }, + { + "epoch": 11.22, + "learning_rate": 3.88114008489994e-05, + "loss": 1.1528, + "step": 129650 + }, + { + "epoch": 11.22, + "learning_rate": 3.8810534523087585e-05, + "loss": 1.185, + "step": 129660 + }, + { + "epoch": 11.22, + "learning_rate": 3.880966819717578e-05, + "loss": 1.1473, + "step": 129670 + }, + { + "epoch": 11.22, + "learning_rate": 3.8808801871263974e-05, + "loss": 1.2023, + "step": 129680 + }, + { + "epoch": 11.23, + "learning_rate": 3.880793554535216e-05, + "loss": 1.1912, + "step": 129690 + }, + { + "epoch": 11.23, + "learning_rate": 3.8807069219440356e-05, + "loss": 1.2146, + "step": 129700 + }, + { + "epoch": 11.23, + "learning_rate": 3.880620289352855e-05, + "loss": 1.2326, + "step": 129710 + }, + { + "epoch": 11.23, + "learning_rate": 3.880533656761674e-05, + "loss": 1.208, + "step": 129720 + }, + { + "epoch": 11.23, + "learning_rate": 3.880447024170493e-05, + "loss": 1.1118, + "step": 129730 + }, + { + "epoch": 11.23, + "learning_rate": 3.880360391579313e-05, + "loss": 1.184, + "step": 129740 + }, + { + "epoch": 11.23, + "learning_rate": 3.8802737589881315e-05, + "loss": 1.1393, + "step": 129750 + }, + { + "epoch": 11.23, + "learning_rate": 3.880187126396951e-05, + "loss": 1.1479, + "step": 129760 + }, + { + "epoch": 11.23, + "learning_rate": 3.88010049380577e-05, + "loss": 1.1859, + "step": 129770 + }, + { + "epoch": 11.23, + "learning_rate": 3.880013861214589e-05, + "loss": 1.2034, + "step": 129780 + }, + { + "epoch": 11.23, + "learning_rate": 3.8799272286234086e-05, + "loss": 1.194, + "step": 129790 + }, + { + "epoch": 11.23, + "learning_rate": 3.8798405960322274e-05, + "loss": 1.1499, + "step": 129800 + }, + { + "epoch": 11.24, + "learning_rate": 3.879753963441047e-05, + "loss": 1.1779, + "step": 129810 + }, + { + "epoch": 11.24, + "learning_rate": 3.879667330849866e-05, + "loss": 1.2128, + "step": 129820 + }, + { + "epoch": 11.24, + "learning_rate": 3.879580698258685e-05, + "loss": 1.2021, + "step": 129830 + }, + { + "epoch": 11.24, + "learning_rate": 3.8794940656675045e-05, + "loss": 1.1748, + "step": 129840 + }, + { + "epoch": 11.24, + "learning_rate": 3.879407433076323e-05, + "loss": 1.1364, + "step": 129850 + }, + { + "epoch": 11.24, + "learning_rate": 3.879320800485143e-05, + "loss": 1.2175, + "step": 129860 + }, + { + "epoch": 11.24, + "learning_rate": 3.879234167893962e-05, + "loss": 1.1721, + "step": 129870 + }, + { + "epoch": 11.24, + "learning_rate": 3.879147535302781e-05, + "loss": 1.1517, + "step": 129880 + }, + { + "epoch": 11.24, + "learning_rate": 3.8790609027116004e-05, + "loss": 1.1739, + "step": 129890 + }, + { + "epoch": 11.24, + "learning_rate": 3.87897427012042e-05, + "loss": 1.1627, + "step": 129900 + }, + { + "epoch": 11.24, + "learning_rate": 3.8788876375292386e-05, + "loss": 1.1592, + "step": 129910 + }, + { + "epoch": 11.25, + "learning_rate": 3.878801004938058e-05, + "loss": 1.1686, + "step": 129920 + }, + { + "epoch": 11.25, + "learning_rate": 3.8787143723468775e-05, + "loss": 1.2027, + "step": 129930 + }, + { + "epoch": 11.25, + "learning_rate": 3.878627739755696e-05, + "loss": 1.1971, + "step": 129940 + }, + { + "epoch": 11.25, + "learning_rate": 3.878541107164516e-05, + "loss": 1.1478, + "step": 129950 + }, + { + "epoch": 11.25, + "learning_rate": 3.8784544745733344e-05, + "loss": 1.2049, + "step": 129960 + }, + { + "epoch": 11.25, + "learning_rate": 3.878367841982154e-05, + "loss": 1.2081, + "step": 129970 + }, + { + "epoch": 11.25, + "learning_rate": 3.878281209390973e-05, + "loss": 1.1588, + "step": 129980 + }, + { + "epoch": 11.25, + "learning_rate": 3.878194576799792e-05, + "loss": 1.208, + "step": 129990 + }, + { + "epoch": 11.25, + "learning_rate": 3.8781079442086115e-05, + "loss": 1.1403, + "step": 130000 + }, + { + "epoch": 11.25, + "learning_rate": 3.878021311617431e-05, + "loss": 1.1852, + "step": 130010 + }, + { + "epoch": 11.25, + "learning_rate": 3.87793467902625e-05, + "loss": 1.176, + "step": 130020 + }, + { + "epoch": 11.25, + "learning_rate": 3.877848046435069e-05, + "loss": 1.1774, + "step": 130030 + }, + { + "epoch": 11.26, + "learning_rate": 3.8777614138438887e-05, + "loss": 1.2081, + "step": 130040 + }, + { + "epoch": 11.26, + "learning_rate": 3.8776747812527074e-05, + "loss": 1.2058, + "step": 130050 + }, + { + "epoch": 11.26, + "learning_rate": 3.877588148661527e-05, + "loss": 1.2119, + "step": 130060 + }, + { + "epoch": 11.26, + "learning_rate": 3.8775015160703456e-05, + "loss": 1.1167, + "step": 130070 + }, + { + "epoch": 11.26, + "learning_rate": 3.877414883479165e-05, + "loss": 1.1487, + "step": 130080 + }, + { + "epoch": 11.26, + "learning_rate": 3.8773282508879845e-05, + "loss": 1.1335, + "step": 130090 + }, + { + "epoch": 11.26, + "learning_rate": 3.877241618296803e-05, + "loss": 1.2774, + "step": 130100 + }, + { + "epoch": 11.26, + "learning_rate": 3.877154985705623e-05, + "loss": 1.0974, + "step": 130110 + }, + { + "epoch": 11.26, + "learning_rate": 3.877068353114442e-05, + "loss": 1.1532, + "step": 130120 + }, + { + "epoch": 11.26, + "learning_rate": 3.876981720523261e-05, + "loss": 1.1964, + "step": 130130 + }, + { + "epoch": 11.26, + "learning_rate": 3.8768950879320804e-05, + "loss": 1.1801, + "step": 130140 + }, + { + "epoch": 11.27, + "learning_rate": 3.8768084553409e-05, + "loss": 1.1681, + "step": 130150 + }, + { + "epoch": 11.27, + "learning_rate": 3.8767218227497186e-05, + "loss": 1.1842, + "step": 130160 + }, + { + "epoch": 11.27, + "learning_rate": 3.876635190158538e-05, + "loss": 1.1685, + "step": 130170 + }, + { + "epoch": 11.27, + "learning_rate": 3.876548557567357e-05, + "loss": 1.1406, + "step": 130180 + }, + { + "epoch": 11.27, + "learning_rate": 3.876461924976176e-05, + "loss": 1.222, + "step": 130190 + }, + { + "epoch": 11.27, + "learning_rate": 3.876375292384996e-05, + "loss": 1.238, + "step": 130200 + }, + { + "epoch": 11.27, + "learning_rate": 3.8762886597938145e-05, + "loss": 1.1828, + "step": 130210 + }, + { + "epoch": 11.27, + "learning_rate": 3.876202027202634e-05, + "loss": 1.1051, + "step": 130220 + }, + { + "epoch": 11.27, + "learning_rate": 3.8761153946114534e-05, + "loss": 1.1773, + "step": 130230 + }, + { + "epoch": 11.27, + "learning_rate": 3.876028762020272e-05, + "loss": 1.1567, + "step": 130240 + }, + { + "epoch": 11.27, + "learning_rate": 3.8759421294290916e-05, + "loss": 1.1893, + "step": 130250 + }, + { + "epoch": 11.27, + "learning_rate": 3.875855496837911e-05, + "loss": 1.1742, + "step": 130260 + }, + { + "epoch": 11.28, + "learning_rate": 3.87576886424673e-05, + "loss": 1.1599, + "step": 130270 + }, + { + "epoch": 11.28, + "learning_rate": 3.875682231655549e-05, + "loss": 1.1819, + "step": 130280 + }, + { + "epoch": 11.28, + "learning_rate": 3.875595599064368e-05, + "loss": 1.1831, + "step": 130290 + }, + { + "epoch": 11.28, + "learning_rate": 3.8755089664731875e-05, + "loss": 1.2088, + "step": 130300 + }, + { + "epoch": 11.28, + "learning_rate": 3.875422333882007e-05, + "loss": 1.1377, + "step": 130310 + }, + { + "epoch": 11.28, + "learning_rate": 3.875335701290826e-05, + "loss": 1.1447, + "step": 130320 + }, + { + "epoch": 11.28, + "learning_rate": 3.875249068699645e-05, + "loss": 1.1924, + "step": 130330 + }, + { + "epoch": 11.28, + "learning_rate": 3.8751624361084646e-05, + "loss": 1.1667, + "step": 130340 + }, + { + "epoch": 11.28, + "learning_rate": 3.875075803517283e-05, + "loss": 1.1982, + "step": 130350 + }, + { + "epoch": 11.28, + "learning_rate": 3.874989170926103e-05, + "loss": 1.1535, + "step": 130360 + }, + { + "epoch": 11.28, + "learning_rate": 3.874902538334922e-05, + "loss": 1.0391, + "step": 130370 + }, + { + "epoch": 11.29, + "learning_rate": 3.874815905743741e-05, + "loss": 1.1273, + "step": 130380 + }, + { + "epoch": 11.29, + "learning_rate": 3.8747292731525604e-05, + "loss": 1.164, + "step": 130390 + }, + { + "epoch": 11.29, + "learning_rate": 3.874642640561379e-05, + "loss": 1.2091, + "step": 130400 + }, + { + "epoch": 11.29, + "learning_rate": 3.8745560079701986e-05, + "loss": 1.164, + "step": 130410 + }, + { + "epoch": 11.29, + "learning_rate": 3.874469375379018e-05, + "loss": 1.2409, + "step": 130420 + }, + { + "epoch": 11.29, + "learning_rate": 3.874382742787837e-05, + "loss": 1.1582, + "step": 130430 + }, + { + "epoch": 11.29, + "learning_rate": 3.874296110196656e-05, + "loss": 1.0905, + "step": 130440 + }, + { + "epoch": 11.29, + "learning_rate": 3.874209477605476e-05, + "loss": 1.1358, + "step": 130450 + }, + { + "epoch": 11.29, + "learning_rate": 3.8741228450142945e-05, + "loss": 1.1657, + "step": 130460 + }, + { + "epoch": 11.29, + "learning_rate": 3.874036212423114e-05, + "loss": 1.1794, + "step": 130470 + }, + { + "epoch": 11.29, + "learning_rate": 3.8739495798319334e-05, + "loss": 1.1778, + "step": 130480 + }, + { + "epoch": 11.29, + "learning_rate": 3.873862947240752e-05, + "loss": 1.1354, + "step": 130490 + }, + { + "epoch": 11.3, + "learning_rate": 3.8737763146495716e-05, + "loss": 1.1657, + "step": 130500 + }, + { + "epoch": 11.3, + "learning_rate": 3.8736896820583904e-05, + "loss": 1.1656, + "step": 130510 + }, + { + "epoch": 11.3, + "learning_rate": 3.87360304946721e-05, + "loss": 1.1844, + "step": 130520 + }, + { + "epoch": 11.3, + "learning_rate": 3.873516416876029e-05, + "loss": 1.107, + "step": 130530 + }, + { + "epoch": 11.3, + "learning_rate": 3.873429784284848e-05, + "loss": 1.1971, + "step": 130540 + }, + { + "epoch": 11.3, + "learning_rate": 3.8733431516936675e-05, + "loss": 1.1727, + "step": 130550 + }, + { + "epoch": 11.3, + "learning_rate": 3.873256519102487e-05, + "loss": 1.2045, + "step": 130560 + }, + { + "epoch": 11.3, + "learning_rate": 3.873169886511306e-05, + "loss": 1.2154, + "step": 130570 + }, + { + "epoch": 11.3, + "learning_rate": 3.873083253920125e-05, + "loss": 1.221, + "step": 130580 + }, + { + "epoch": 11.3, + "learning_rate": 3.8729966213289446e-05, + "loss": 1.1812, + "step": 130590 + }, + { + "epoch": 11.3, + "learning_rate": 3.8729099887377634e-05, + "loss": 1.2855, + "step": 130600 + }, + { + "epoch": 11.31, + "learning_rate": 3.872823356146583e-05, + "loss": 1.1431, + "step": 130610 + }, + { + "epoch": 11.31, + "learning_rate": 3.8727367235554016e-05, + "loss": 1.2127, + "step": 130620 + }, + { + "epoch": 11.31, + "learning_rate": 3.872650090964221e-05, + "loss": 1.133, + "step": 130630 + }, + { + "epoch": 11.31, + "learning_rate": 3.8725634583730405e-05, + "loss": 1.1677, + "step": 130640 + }, + { + "epoch": 11.31, + "learning_rate": 3.872476825781859e-05, + "loss": 1.1873, + "step": 130650 + }, + { + "epoch": 11.31, + "learning_rate": 3.872390193190679e-05, + "loss": 1.1771, + "step": 130660 + }, + { + "epoch": 11.31, + "learning_rate": 3.872303560599498e-05, + "loss": 1.1994, + "step": 130670 + }, + { + "epoch": 11.31, + "learning_rate": 3.872216928008317e-05, + "loss": 1.1178, + "step": 130680 + }, + { + "epoch": 11.31, + "learning_rate": 3.872130295417136e-05, + "loss": 1.1545, + "step": 130690 + }, + { + "epoch": 11.31, + "learning_rate": 3.872043662825955e-05, + "loss": 1.174, + "step": 130700 + }, + { + "epoch": 11.31, + "learning_rate": 3.8719570302347745e-05, + "loss": 1.1784, + "step": 130710 + }, + { + "epoch": 11.31, + "learning_rate": 3.871870397643594e-05, + "loss": 1.1928, + "step": 130720 + }, + { + "epoch": 11.32, + "learning_rate": 3.871783765052413e-05, + "loss": 1.2071, + "step": 130730 + }, + { + "epoch": 11.32, + "learning_rate": 3.871697132461232e-05, + "loss": 1.1689, + "step": 130740 + }, + { + "epoch": 11.32, + "learning_rate": 3.8716104998700516e-05, + "loss": 1.2246, + "step": 130750 + }, + { + "epoch": 11.32, + "learning_rate": 3.8715238672788704e-05, + "loss": 1.1971, + "step": 130760 + }, + { + "epoch": 11.32, + "learning_rate": 3.87143723468769e-05, + "loss": 1.188, + "step": 130770 + }, + { + "epoch": 11.32, + "learning_rate": 3.871350602096509e-05, + "loss": 1.1353, + "step": 130780 + }, + { + "epoch": 11.32, + "learning_rate": 3.871263969505328e-05, + "loss": 1.1924, + "step": 130790 + }, + { + "epoch": 11.32, + "learning_rate": 3.8711773369141475e-05, + "loss": 1.1587, + "step": 130800 + }, + { + "epoch": 11.32, + "learning_rate": 3.871090704322966e-05, + "loss": 1.2199, + "step": 130810 + }, + { + "epoch": 11.32, + "learning_rate": 3.871004071731786e-05, + "loss": 1.1756, + "step": 130820 + }, + { + "epoch": 11.32, + "learning_rate": 3.870917439140605e-05, + "loss": 1.2543, + "step": 130830 + }, + { + "epoch": 11.32, + "learning_rate": 3.870830806549424e-05, + "loss": 1.1817, + "step": 130840 + }, + { + "epoch": 11.33, + "learning_rate": 3.8707441739582434e-05, + "loss": 1.1634, + "step": 130850 + }, + { + "epoch": 11.33, + "learning_rate": 3.870657541367063e-05, + "loss": 1.1462, + "step": 130860 + }, + { + "epoch": 11.33, + "learning_rate": 3.8705709087758816e-05, + "loss": 1.1705, + "step": 130870 + }, + { + "epoch": 11.33, + "learning_rate": 3.870484276184701e-05, + "loss": 1.1791, + "step": 130880 + }, + { + "epoch": 11.33, + "learning_rate": 3.8703976435935205e-05, + "loss": 1.1741, + "step": 130890 + }, + { + "epoch": 11.33, + "learning_rate": 3.870311011002339e-05, + "loss": 1.1824, + "step": 130900 + }, + { + "epoch": 11.33, + "learning_rate": 3.870224378411159e-05, + "loss": 1.1802, + "step": 130910 + }, + { + "epoch": 11.33, + "learning_rate": 3.8701377458199775e-05, + "loss": 1.1615, + "step": 130920 + }, + { + "epoch": 11.33, + "learning_rate": 3.870051113228797e-05, + "loss": 1.1506, + "step": 130930 + }, + { + "epoch": 11.33, + "learning_rate": 3.8699644806376164e-05, + "loss": 1.1837, + "step": 130940 + }, + { + "epoch": 11.33, + "learning_rate": 3.869877848046435e-05, + "loss": 1.1609, + "step": 130950 + }, + { + "epoch": 11.34, + "learning_rate": 3.8697912154552546e-05, + "loss": 1.1769, + "step": 130960 + }, + { + "epoch": 11.34, + "learning_rate": 3.869704582864074e-05, + "loss": 1.1465, + "step": 130970 + }, + { + "epoch": 11.34, + "learning_rate": 3.869617950272893e-05, + "loss": 1.1578, + "step": 130980 + }, + { + "epoch": 11.34, + "learning_rate": 3.869531317681712e-05, + "loss": 1.1577, + "step": 130990 + }, + { + "epoch": 11.34, + "learning_rate": 3.869444685090532e-05, + "loss": 1.2106, + "step": 131000 + }, + { + "epoch": 11.34, + "learning_rate": 3.8693580524993504e-05, + "loss": 1.2049, + "step": 131010 + }, + { + "epoch": 11.34, + "learning_rate": 3.86927141990817e-05, + "loss": 1.119, + "step": 131020 + }, + { + "epoch": 11.34, + "learning_rate": 3.869184787316989e-05, + "loss": 1.1547, + "step": 131030 + }, + { + "epoch": 11.34, + "learning_rate": 3.869098154725808e-05, + "loss": 1.2114, + "step": 131040 + }, + { + "epoch": 11.34, + "learning_rate": 3.8690115221346276e-05, + "loss": 1.1957, + "step": 131050 + }, + { + "epoch": 11.34, + "learning_rate": 3.868924889543446e-05, + "loss": 1.1618, + "step": 131060 + }, + { + "epoch": 11.34, + "learning_rate": 3.868838256952266e-05, + "loss": 1.2096, + "step": 131070 + }, + { + "epoch": 11.35, + "learning_rate": 3.868751624361085e-05, + "loss": 1.1806, + "step": 131080 + }, + { + "epoch": 11.35, + "learning_rate": 3.868664991769904e-05, + "loss": 1.131, + "step": 131090 + }, + { + "epoch": 11.35, + "learning_rate": 3.8685783591787234e-05, + "loss": 1.1953, + "step": 131100 + }, + { + "epoch": 11.35, + "learning_rate": 3.868491726587543e-05, + "loss": 1.1652, + "step": 131110 + }, + { + "epoch": 11.35, + "learning_rate": 3.8684050939963616e-05, + "loss": 1.211, + "step": 131120 + }, + { + "epoch": 11.35, + "learning_rate": 3.868318461405181e-05, + "loss": 1.1737, + "step": 131130 + }, + { + "epoch": 11.35, + "learning_rate": 3.868231828814e-05, + "loss": 1.2323, + "step": 131140 + }, + { + "epoch": 11.35, + "learning_rate": 3.868145196222819e-05, + "loss": 1.1719, + "step": 131150 + }, + { + "epoch": 11.35, + "learning_rate": 3.868058563631639e-05, + "loss": 1.1958, + "step": 131160 + }, + { + "epoch": 11.35, + "learning_rate": 3.8679719310404575e-05, + "loss": 1.1121, + "step": 131170 + }, + { + "epoch": 11.35, + "learning_rate": 3.867885298449277e-05, + "loss": 1.182, + "step": 131180 + }, + { + "epoch": 11.36, + "learning_rate": 3.8677986658580964e-05, + "loss": 1.1695, + "step": 131190 + }, + { + "epoch": 11.36, + "learning_rate": 3.867712033266915e-05, + "loss": 1.1632, + "step": 131200 + }, + { + "epoch": 11.36, + "learning_rate": 3.8676254006757346e-05, + "loss": 1.1893, + "step": 131210 + }, + { + "epoch": 11.36, + "learning_rate": 3.867538768084554e-05, + "loss": 1.2211, + "step": 131220 + }, + { + "epoch": 11.36, + "learning_rate": 3.867452135493373e-05, + "loss": 1.1561, + "step": 131230 + }, + { + "epoch": 11.36, + "learning_rate": 3.867365502902192e-05, + "loss": 1.1969, + "step": 131240 + }, + { + "epoch": 11.36, + "learning_rate": 3.867278870311011e-05, + "loss": 1.1557, + "step": 131250 + }, + { + "epoch": 11.36, + "learning_rate": 3.8671922377198305e-05, + "loss": 1.1798, + "step": 131260 + }, + { + "epoch": 11.36, + "learning_rate": 3.86710560512865e-05, + "loss": 1.1651, + "step": 131270 + }, + { + "epoch": 11.36, + "learning_rate": 3.867018972537469e-05, + "loss": 1.23, + "step": 131280 + }, + { + "epoch": 11.36, + "learning_rate": 3.866932339946288e-05, + "loss": 1.1293, + "step": 131290 + }, + { + "epoch": 11.36, + "learning_rate": 3.8668457073551076e-05, + "loss": 1.171, + "step": 131300 + }, + { + "epoch": 11.37, + "learning_rate": 3.8667590747639264e-05, + "loss": 1.1619, + "step": 131310 + }, + { + "epoch": 11.37, + "learning_rate": 3.866672442172746e-05, + "loss": 1.2129, + "step": 131320 + }, + { + "epoch": 11.37, + "learning_rate": 3.866585809581565e-05, + "loss": 1.1287, + "step": 131330 + }, + { + "epoch": 11.37, + "learning_rate": 3.866499176990384e-05, + "loss": 1.1088, + "step": 131340 + }, + { + "epoch": 11.37, + "learning_rate": 3.8664125443992035e-05, + "loss": 1.1347, + "step": 131350 + }, + { + "epoch": 11.37, + "learning_rate": 3.866325911808022e-05, + "loss": 1.1793, + "step": 131360 + }, + { + "epoch": 11.37, + "learning_rate": 3.866239279216842e-05, + "loss": 1.1541, + "step": 131370 + }, + { + "epoch": 11.37, + "learning_rate": 3.866152646625661e-05, + "loss": 1.1855, + "step": 131380 + }, + { + "epoch": 11.37, + "learning_rate": 3.86606601403448e-05, + "loss": 1.253, + "step": 131390 + }, + { + "epoch": 11.37, + "learning_rate": 3.865979381443299e-05, + "loss": 1.2127, + "step": 131400 + }, + { + "epoch": 11.37, + "learning_rate": 3.865892748852119e-05, + "loss": 1.2052, + "step": 131410 + }, + { + "epoch": 11.38, + "learning_rate": 3.8658061162609375e-05, + "loss": 1.2013, + "step": 131420 + }, + { + "epoch": 11.38, + "learning_rate": 3.865719483669757e-05, + "loss": 1.1424, + "step": 131430 + }, + { + "epoch": 11.38, + "learning_rate": 3.865632851078576e-05, + "loss": 1.1639, + "step": 131440 + }, + { + "epoch": 11.38, + "learning_rate": 3.865546218487395e-05, + "loss": 1.2052, + "step": 131450 + }, + { + "epoch": 11.38, + "learning_rate": 3.8654595858962146e-05, + "loss": 1.1373, + "step": 131460 + }, + { + "epoch": 11.38, + "learning_rate": 3.8653729533050334e-05, + "loss": 1.2277, + "step": 131470 + }, + { + "epoch": 11.38, + "learning_rate": 3.865286320713853e-05, + "loss": 1.2149, + "step": 131480 + }, + { + "epoch": 11.38, + "learning_rate": 3.865199688122672e-05, + "loss": 1.2156, + "step": 131490 + }, + { + "epoch": 11.38, + "learning_rate": 3.865113055531491e-05, + "loss": 1.1782, + "step": 131500 + }, + { + "epoch": 11.38, + "learning_rate": 3.8650264229403105e-05, + "loss": 1.2056, + "step": 131510 + }, + { + "epoch": 11.38, + "learning_rate": 3.86493979034913e-05, + "loss": 1.1727, + "step": 131520 + }, + { + "epoch": 11.38, + "learning_rate": 3.864853157757949e-05, + "loss": 1.1569, + "step": 131530 + }, + { + "epoch": 11.39, + "learning_rate": 3.864766525166768e-05, + "loss": 1.1913, + "step": 131540 + }, + { + "epoch": 11.39, + "learning_rate": 3.864679892575587e-05, + "loss": 1.186, + "step": 131550 + }, + { + "epoch": 11.39, + "learning_rate": 3.8645932599844064e-05, + "loss": 1.1126, + "step": 131560 + }, + { + "epoch": 11.39, + "learning_rate": 3.864506627393226e-05, + "loss": 1.2169, + "step": 131570 + }, + { + "epoch": 11.39, + "learning_rate": 3.8644199948020446e-05, + "loss": 1.1397, + "step": 131580 + }, + { + "epoch": 11.39, + "learning_rate": 3.864333362210864e-05, + "loss": 1.1709, + "step": 131590 + }, + { + "epoch": 11.39, + "learning_rate": 3.8642467296196835e-05, + "loss": 1.1361, + "step": 131600 + }, + { + "epoch": 11.39, + "learning_rate": 3.864160097028502e-05, + "loss": 1.128, + "step": 131610 + }, + { + "epoch": 11.39, + "learning_rate": 3.864073464437322e-05, + "loss": 1.1836, + "step": 131620 + }, + { + "epoch": 11.39, + "learning_rate": 3.863986831846141e-05, + "loss": 1.226, + "step": 131630 + }, + { + "epoch": 11.39, + "learning_rate": 3.86390019925496e-05, + "loss": 1.2207, + "step": 131640 + }, + { + "epoch": 11.4, + "learning_rate": 3.8638135666637794e-05, + "loss": 1.1858, + "step": 131650 + }, + { + "epoch": 11.4, + "learning_rate": 3.863726934072598e-05, + "loss": 1.1184, + "step": 131660 + }, + { + "epoch": 11.4, + "learning_rate": 3.8636403014814176e-05, + "loss": 1.1248, + "step": 131670 + }, + { + "epoch": 11.4, + "learning_rate": 3.863553668890237e-05, + "loss": 1.1797, + "step": 131680 + }, + { + "epoch": 11.4, + "learning_rate": 3.863467036299056e-05, + "loss": 1.2206, + "step": 131690 + }, + { + "epoch": 11.4, + "learning_rate": 3.863380403707875e-05, + "loss": 1.1716, + "step": 131700 + }, + { + "epoch": 11.4, + "learning_rate": 3.863293771116695e-05, + "loss": 1.2566, + "step": 131710 + }, + { + "epoch": 11.4, + "learning_rate": 3.8632071385255134e-05, + "loss": 1.1776, + "step": 131720 + }, + { + "epoch": 11.4, + "learning_rate": 3.863120505934333e-05, + "loss": 1.1786, + "step": 131730 + }, + { + "epoch": 11.4, + "learning_rate": 3.863033873343152e-05, + "loss": 1.2007, + "step": 131740 + }, + { + "epoch": 11.4, + "learning_rate": 3.862947240751971e-05, + "loss": 1.1909, + "step": 131750 + }, + { + "epoch": 11.4, + "learning_rate": 3.8628606081607906e-05, + "loss": 1.1886, + "step": 131760 + }, + { + "epoch": 11.41, + "learning_rate": 3.862773975569609e-05, + "loss": 1.1258, + "step": 131770 + }, + { + "epoch": 11.41, + "learning_rate": 3.862687342978429e-05, + "loss": 1.146, + "step": 131780 + }, + { + "epoch": 11.41, + "learning_rate": 3.862600710387248e-05, + "loss": 1.1689, + "step": 131790 + }, + { + "epoch": 11.41, + "learning_rate": 3.862514077796067e-05, + "loss": 1.1561, + "step": 131800 + }, + { + "epoch": 11.41, + "learning_rate": 3.8624274452048864e-05, + "loss": 1.1054, + "step": 131810 + }, + { + "epoch": 11.41, + "learning_rate": 3.862340812613706e-05, + "loss": 1.1578, + "step": 131820 + }, + { + "epoch": 11.41, + "learning_rate": 3.8622541800225246e-05, + "loss": 1.1519, + "step": 131830 + }, + { + "epoch": 11.41, + "learning_rate": 3.862167547431344e-05, + "loss": 1.1295, + "step": 131840 + }, + { + "epoch": 11.41, + "learning_rate": 3.8620809148401635e-05, + "loss": 1.1652, + "step": 131850 + }, + { + "epoch": 11.41, + "learning_rate": 3.861994282248982e-05, + "loss": 1.1954, + "step": 131860 + }, + { + "epoch": 11.41, + "learning_rate": 3.861907649657802e-05, + "loss": 1.1886, + "step": 131870 + }, + { + "epoch": 11.41, + "learning_rate": 3.8618210170666205e-05, + "loss": 1.2113, + "step": 131880 + }, + { + "epoch": 11.42, + "learning_rate": 3.86173438447544e-05, + "loss": 1.1905, + "step": 131890 + }, + { + "epoch": 11.42, + "learning_rate": 3.8616477518842594e-05, + "loss": 1.1855, + "step": 131900 + }, + { + "epoch": 11.42, + "learning_rate": 3.861561119293078e-05, + "loss": 1.202, + "step": 131910 + }, + { + "epoch": 11.42, + "learning_rate": 3.8614744867018976e-05, + "loss": 1.2101, + "step": 131920 + }, + { + "epoch": 11.42, + "learning_rate": 3.861387854110717e-05, + "loss": 1.1229, + "step": 131930 + }, + { + "epoch": 11.42, + "learning_rate": 3.861301221519536e-05, + "loss": 1.1923, + "step": 131940 + }, + { + "epoch": 11.42, + "learning_rate": 3.861214588928355e-05, + "loss": 1.2215, + "step": 131950 + }, + { + "epoch": 11.42, + "learning_rate": 3.861127956337175e-05, + "loss": 1.1899, + "step": 131960 + }, + { + "epoch": 11.42, + "learning_rate": 3.8610413237459935e-05, + "loss": 1.1572, + "step": 131970 + }, + { + "epoch": 11.42, + "learning_rate": 3.860954691154813e-05, + "loss": 1.1358, + "step": 131980 + }, + { + "epoch": 11.42, + "learning_rate": 3.860868058563632e-05, + "loss": 1.1416, + "step": 131990 + }, + { + "epoch": 11.43, + "learning_rate": 3.860781425972451e-05, + "loss": 1.1822, + "step": 132000 + }, + { + "epoch": 11.43, + "learning_rate": 3.8606947933812706e-05, + "loss": 1.1508, + "step": 132010 + }, + { + "epoch": 11.43, + "learning_rate": 3.8606081607900894e-05, + "loss": 1.2351, + "step": 132020 + }, + { + "epoch": 11.43, + "learning_rate": 3.860521528198909e-05, + "loss": 1.1206, + "step": 132030 + }, + { + "epoch": 11.43, + "learning_rate": 3.860434895607728e-05, + "loss": 1.1686, + "step": 132040 + }, + { + "epoch": 11.43, + "learning_rate": 3.860348263016547e-05, + "loss": 1.2074, + "step": 132050 + }, + { + "epoch": 11.43, + "learning_rate": 3.8602616304253665e-05, + "loss": 1.1347, + "step": 132060 + }, + { + "epoch": 11.43, + "learning_rate": 3.860174997834186e-05, + "loss": 1.1131, + "step": 132070 + }, + { + "epoch": 11.43, + "learning_rate": 3.860088365243005e-05, + "loss": 1.142, + "step": 132080 + }, + { + "epoch": 11.43, + "learning_rate": 3.860001732651824e-05, + "loss": 1.2134, + "step": 132090 + }, + { + "epoch": 11.43, + "learning_rate": 3.859915100060643e-05, + "loss": 1.1915, + "step": 132100 + }, + { + "epoch": 11.43, + "learning_rate": 3.859828467469462e-05, + "loss": 1.125, + "step": 132110 + }, + { + "epoch": 11.44, + "learning_rate": 3.859741834878282e-05, + "loss": 1.1646, + "step": 132120 + }, + { + "epoch": 11.44, + "learning_rate": 3.8596552022871005e-05, + "loss": 1.1232, + "step": 132130 + }, + { + "epoch": 11.44, + "learning_rate": 3.85956856969592e-05, + "loss": 1.207, + "step": 132140 + }, + { + "epoch": 11.44, + "learning_rate": 3.8594819371047394e-05, + "loss": 1.1322, + "step": 132150 + }, + { + "epoch": 11.44, + "learning_rate": 3.859395304513558e-05, + "loss": 1.1927, + "step": 132160 + }, + { + "epoch": 11.44, + "learning_rate": 3.8593086719223776e-05, + "loss": 1.1613, + "step": 132170 + }, + { + "epoch": 11.44, + "learning_rate": 3.8592220393311964e-05, + "loss": 1.109, + "step": 132180 + }, + { + "epoch": 11.44, + "learning_rate": 3.859135406740016e-05, + "loss": 1.2043, + "step": 132190 + }, + { + "epoch": 11.44, + "learning_rate": 3.859048774148835e-05, + "loss": 1.0755, + "step": 132200 + }, + { + "epoch": 11.44, + "learning_rate": 3.858962141557654e-05, + "loss": 1.1999, + "step": 132210 + }, + { + "epoch": 11.44, + "learning_rate": 3.8588755089664735e-05, + "loss": 1.1757, + "step": 132220 + }, + { + "epoch": 11.45, + "learning_rate": 3.858788876375293e-05, + "loss": 1.1365, + "step": 132230 + }, + { + "epoch": 11.45, + "learning_rate": 3.858702243784112e-05, + "loss": 1.1948, + "step": 132240 + }, + { + "epoch": 11.45, + "learning_rate": 3.858615611192931e-05, + "loss": 1.1378, + "step": 132250 + }, + { + "epoch": 11.45, + "learning_rate": 3.8585289786017506e-05, + "loss": 1.1776, + "step": 132260 + }, + { + "epoch": 11.45, + "learning_rate": 3.8584423460105694e-05, + "loss": 1.1549, + "step": 132270 + }, + { + "epoch": 11.45, + "learning_rate": 3.858355713419389e-05, + "loss": 1.1868, + "step": 132280 + }, + { + "epoch": 11.45, + "learning_rate": 3.8582690808282076e-05, + "loss": 1.1613, + "step": 132290 + }, + { + "epoch": 11.45, + "learning_rate": 3.858182448237027e-05, + "loss": 1.1632, + "step": 132300 + }, + { + "epoch": 11.45, + "learning_rate": 3.8580958156458465e-05, + "loss": 1.2342, + "step": 132310 + }, + { + "epoch": 11.45, + "learning_rate": 3.858009183054665e-05, + "loss": 1.112, + "step": 132320 + }, + { + "epoch": 11.45, + "learning_rate": 3.857922550463485e-05, + "loss": 1.1406, + "step": 132330 + }, + { + "epoch": 11.45, + "learning_rate": 3.857835917872304e-05, + "loss": 1.2336, + "step": 132340 + }, + { + "epoch": 11.46, + "learning_rate": 3.857749285281123e-05, + "loss": 1.2136, + "step": 132350 + }, + { + "epoch": 11.46, + "learning_rate": 3.8576626526899424e-05, + "loss": 1.195, + "step": 132360 + }, + { + "epoch": 11.46, + "learning_rate": 3.857576020098762e-05, + "loss": 1.1399, + "step": 132370 + }, + { + "epoch": 11.46, + "learning_rate": 3.8574893875075806e-05, + "loss": 1.1897, + "step": 132380 + }, + { + "epoch": 11.46, + "learning_rate": 3.8574027549164e-05, + "loss": 1.1586, + "step": 132390 + }, + { + "epoch": 11.46, + "learning_rate": 3.857316122325219e-05, + "loss": 1.1173, + "step": 132400 + }, + { + "epoch": 11.46, + "learning_rate": 3.857229489734038e-05, + "loss": 1.1447, + "step": 132410 + }, + { + "epoch": 11.46, + "learning_rate": 3.857142857142858e-05, + "loss": 1.1778, + "step": 132420 + }, + { + "epoch": 11.46, + "learning_rate": 3.8570562245516764e-05, + "loss": 1.1483, + "step": 132430 + }, + { + "epoch": 11.46, + "learning_rate": 3.856969591960496e-05, + "loss": 1.2303, + "step": 132440 + }, + { + "epoch": 11.46, + "learning_rate": 3.856882959369315e-05, + "loss": 1.1745, + "step": 132450 + }, + { + "epoch": 11.47, + "learning_rate": 3.856796326778134e-05, + "loss": 1.1509, + "step": 132460 + }, + { + "epoch": 11.47, + "learning_rate": 3.8567096941869535e-05, + "loss": 1.1484, + "step": 132470 + }, + { + "epoch": 11.47, + "learning_rate": 3.856623061595773e-05, + "loss": 1.1669, + "step": 132480 + }, + { + "epoch": 11.47, + "learning_rate": 3.856536429004592e-05, + "loss": 1.133, + "step": 132490 + }, + { + "epoch": 11.47, + "learning_rate": 3.856449796413411e-05, + "loss": 1.262, + "step": 132500 + }, + { + "epoch": 11.47, + "learning_rate": 3.85636316382223e-05, + "loss": 1.1717, + "step": 132510 + }, + { + "epoch": 11.47, + "learning_rate": 3.8562765312310494e-05, + "loss": 1.1811, + "step": 132520 + }, + { + "epoch": 11.47, + "learning_rate": 3.856189898639869e-05, + "loss": 1.2062, + "step": 132530 + }, + { + "epoch": 11.47, + "learning_rate": 3.8561032660486876e-05, + "loss": 1.2119, + "step": 132540 + }, + { + "epoch": 11.47, + "learning_rate": 3.856016633457507e-05, + "loss": 1.1575, + "step": 132550 + }, + { + "epoch": 11.47, + "learning_rate": 3.8559300008663265e-05, + "loss": 1.1428, + "step": 132560 + }, + { + "epoch": 11.47, + "learning_rate": 3.855843368275145e-05, + "loss": 1.1612, + "step": 132570 + }, + { + "epoch": 11.48, + "learning_rate": 3.855756735683965e-05, + "loss": 1.1627, + "step": 132580 + }, + { + "epoch": 11.48, + "learning_rate": 3.855670103092784e-05, + "loss": 1.1688, + "step": 132590 + }, + { + "epoch": 11.48, + "learning_rate": 3.855583470501603e-05, + "loss": 1.1631, + "step": 132600 + }, + { + "epoch": 11.48, + "learning_rate": 3.8554968379104224e-05, + "loss": 1.204, + "step": 132610 + }, + { + "epoch": 11.48, + "learning_rate": 3.855410205319241e-05, + "loss": 1.1275, + "step": 132620 + }, + { + "epoch": 11.48, + "learning_rate": 3.8553235727280606e-05, + "loss": 1.1974, + "step": 132630 + }, + { + "epoch": 11.48, + "learning_rate": 3.85523694013688e-05, + "loss": 1.1532, + "step": 132640 + }, + { + "epoch": 11.48, + "learning_rate": 3.855150307545699e-05, + "loss": 1.1966, + "step": 132650 + }, + { + "epoch": 11.48, + "learning_rate": 3.855063674954518e-05, + "loss": 1.1733, + "step": 132660 + }, + { + "epoch": 11.48, + "learning_rate": 3.854977042363338e-05, + "loss": 1.0894, + "step": 132670 + }, + { + "epoch": 11.48, + "learning_rate": 3.8548904097721565e-05, + "loss": 1.161, + "step": 132680 + }, + { + "epoch": 11.49, + "learning_rate": 3.854803777180976e-05, + "loss": 1.1344, + "step": 132690 + }, + { + "epoch": 11.49, + "learning_rate": 3.8547171445897954e-05, + "loss": 1.1371, + "step": 132700 + }, + { + "epoch": 11.49, + "learning_rate": 3.854630511998614e-05, + "loss": 1.1594, + "step": 132710 + }, + { + "epoch": 11.49, + "learning_rate": 3.8545438794074336e-05, + "loss": 1.1143, + "step": 132720 + }, + { + "epoch": 11.49, + "learning_rate": 3.8544572468162523e-05, + "loss": 1.1806, + "step": 132730 + }, + { + "epoch": 11.49, + "learning_rate": 3.854370614225072e-05, + "loss": 1.1688, + "step": 132740 + }, + { + "epoch": 11.49, + "learning_rate": 3.854283981633891e-05, + "loss": 1.1543, + "step": 132750 + }, + { + "epoch": 11.49, + "learning_rate": 3.85419734904271e-05, + "loss": 1.1643, + "step": 132760 + }, + { + "epoch": 11.49, + "learning_rate": 3.8541107164515295e-05, + "loss": 1.2144, + "step": 132770 + }, + { + "epoch": 11.49, + "learning_rate": 3.854024083860349e-05, + "loss": 1.1745, + "step": 132780 + }, + { + "epoch": 11.49, + "learning_rate": 3.853937451269168e-05, + "loss": 1.1887, + "step": 132790 + }, + { + "epoch": 11.49, + "learning_rate": 3.853850818677987e-05, + "loss": 1.2414, + "step": 132800 + }, + { + "epoch": 11.5, + "learning_rate": 3.8537641860868066e-05, + "loss": 1.1992, + "step": 132810 + }, + { + "epoch": 11.5, + "learning_rate": 3.853677553495625e-05, + "loss": 1.1759, + "step": 132820 + }, + { + "epoch": 11.5, + "learning_rate": 3.853590920904445e-05, + "loss": 1.1595, + "step": 132830 + }, + { + "epoch": 11.5, + "learning_rate": 3.8535042883132635e-05, + "loss": 1.1992, + "step": 132840 + }, + { + "epoch": 11.5, + "learning_rate": 3.853417655722083e-05, + "loss": 1.1892, + "step": 132850 + }, + { + "epoch": 11.5, + "learning_rate": 3.8533310231309024e-05, + "loss": 1.1939, + "step": 132860 + }, + { + "epoch": 11.5, + "learning_rate": 3.853244390539721e-05, + "loss": 1.2424, + "step": 132870 + }, + { + "epoch": 11.5, + "learning_rate": 3.8531577579485406e-05, + "loss": 1.1598, + "step": 132880 + }, + { + "epoch": 11.5, + "learning_rate": 3.85307112535736e-05, + "loss": 1.126, + "step": 132890 + }, + { + "epoch": 11.5, + "learning_rate": 3.852984492766179e-05, + "loss": 1.1252, + "step": 132900 + }, + { + "epoch": 11.5, + "learning_rate": 3.852897860174998e-05, + "loss": 1.1098, + "step": 132910 + }, + { + "epoch": 11.5, + "learning_rate": 3.852811227583818e-05, + "loss": 1.1406, + "step": 132920 + }, + { + "epoch": 11.51, + "learning_rate": 3.8527245949926365e-05, + "loss": 1.1816, + "step": 132930 + }, + { + "epoch": 11.51, + "learning_rate": 3.852637962401456e-05, + "loss": 1.1827, + "step": 132940 + }, + { + "epoch": 11.51, + "learning_rate": 3.852551329810275e-05, + "loss": 1.1405, + "step": 132950 + }, + { + "epoch": 11.51, + "learning_rate": 3.852464697219094e-05, + "loss": 1.1428, + "step": 132960 + }, + { + "epoch": 11.51, + "learning_rate": 3.8523780646279136e-05, + "loss": 1.1623, + "step": 132970 + }, + { + "epoch": 11.51, + "learning_rate": 3.8522914320367324e-05, + "loss": 1.1187, + "step": 132980 + }, + { + "epoch": 11.51, + "learning_rate": 3.852204799445552e-05, + "loss": 1.1233, + "step": 132990 + }, + { + "epoch": 11.51, + "learning_rate": 3.852118166854371e-05, + "loss": 1.1717, + "step": 133000 + }, + { + "epoch": 11.51, + "learning_rate": 3.85203153426319e-05, + "loss": 1.1396, + "step": 133010 + }, + { + "epoch": 11.51, + "learning_rate": 3.8519449016720095e-05, + "loss": 1.1186, + "step": 133020 + }, + { + "epoch": 11.51, + "learning_rate": 3.851858269080828e-05, + "loss": 1.0996, + "step": 133030 + }, + { + "epoch": 11.52, + "learning_rate": 3.851771636489648e-05, + "loss": 1.1576, + "step": 133040 + }, + { + "epoch": 11.52, + "learning_rate": 3.851685003898467e-05, + "loss": 1.1793, + "step": 133050 + }, + { + "epoch": 11.52, + "learning_rate": 3.851598371307286e-05, + "loss": 1.1371, + "step": 133060 + }, + { + "epoch": 11.52, + "learning_rate": 3.8515117387161054e-05, + "loss": 1.2225, + "step": 133070 + }, + { + "epoch": 11.52, + "learning_rate": 3.851425106124925e-05, + "loss": 1.1681, + "step": 133080 + }, + { + "epoch": 11.52, + "learning_rate": 3.8513384735337436e-05, + "loss": 1.1448, + "step": 133090 + }, + { + "epoch": 11.52, + "learning_rate": 3.851251840942563e-05, + "loss": 1.2, + "step": 133100 + }, + { + "epoch": 11.52, + "learning_rate": 3.8511652083513825e-05, + "loss": 1.0809, + "step": 133110 + }, + { + "epoch": 11.52, + "learning_rate": 3.851078575760201e-05, + "loss": 1.1873, + "step": 133120 + }, + { + "epoch": 11.52, + "learning_rate": 3.850991943169021e-05, + "loss": 1.1482, + "step": 133130 + }, + { + "epoch": 11.52, + "learning_rate": 3.8509053105778394e-05, + "loss": 1.1652, + "step": 133140 + }, + { + "epoch": 11.52, + "learning_rate": 3.850818677986659e-05, + "loss": 1.196, + "step": 133150 + }, + { + "epoch": 11.53, + "learning_rate": 3.850732045395478e-05, + "loss": 1.163, + "step": 133160 + }, + { + "epoch": 11.53, + "learning_rate": 3.850645412804297e-05, + "loss": 1.1737, + "step": 133170 + }, + { + "epoch": 11.53, + "learning_rate": 3.8505587802131165e-05, + "loss": 1.1732, + "step": 133180 + }, + { + "epoch": 11.53, + "learning_rate": 3.850472147621936e-05, + "loss": 1.1561, + "step": 133190 + }, + { + "epoch": 11.53, + "learning_rate": 3.850385515030755e-05, + "loss": 1.1538, + "step": 133200 + }, + { + "epoch": 11.53, + "learning_rate": 3.850298882439574e-05, + "loss": 1.1949, + "step": 133210 + }, + { + "epoch": 11.53, + "learning_rate": 3.8502122498483936e-05, + "loss": 1.1627, + "step": 133220 + }, + { + "epoch": 11.53, + "learning_rate": 3.8501256172572124e-05, + "loss": 1.1367, + "step": 133230 + }, + { + "epoch": 11.53, + "learning_rate": 3.850038984666032e-05, + "loss": 1.1557, + "step": 133240 + }, + { + "epoch": 11.53, + "learning_rate": 3.8499523520748506e-05, + "loss": 1.1947, + "step": 133250 + }, + { + "epoch": 11.53, + "learning_rate": 3.84986571948367e-05, + "loss": 1.1653, + "step": 133260 + }, + { + "epoch": 11.54, + "learning_rate": 3.8497790868924895e-05, + "loss": 1.1552, + "step": 133270 + }, + { + "epoch": 11.54, + "learning_rate": 3.849692454301308e-05, + "loss": 1.1643, + "step": 133280 + }, + { + "epoch": 11.54, + "learning_rate": 3.849605821710128e-05, + "loss": 1.1956, + "step": 133290 + }, + { + "epoch": 11.54, + "learning_rate": 3.849519189118947e-05, + "loss": 1.1663, + "step": 133300 + }, + { + "epoch": 11.54, + "learning_rate": 3.849432556527766e-05, + "loss": 1.1825, + "step": 133310 + }, + { + "epoch": 11.54, + "learning_rate": 3.8493459239365854e-05, + "loss": 1.1567, + "step": 133320 + }, + { + "epoch": 11.54, + "learning_rate": 3.849259291345405e-05, + "loss": 1.1652, + "step": 133330 + }, + { + "epoch": 11.54, + "learning_rate": 3.8491726587542236e-05, + "loss": 1.2203, + "step": 133340 + }, + { + "epoch": 11.54, + "learning_rate": 3.849086026163043e-05, + "loss": 1.1278, + "step": 133350 + }, + { + "epoch": 11.54, + "learning_rate": 3.848999393571862e-05, + "loss": 1.1928, + "step": 133360 + }, + { + "epoch": 11.54, + "learning_rate": 3.848912760980681e-05, + "loss": 1.1248, + "step": 133370 + }, + { + "epoch": 11.54, + "learning_rate": 3.848826128389501e-05, + "loss": 1.1733, + "step": 133380 + }, + { + "epoch": 11.55, + "learning_rate": 3.8487394957983195e-05, + "loss": 1.1154, + "step": 133390 + }, + { + "epoch": 11.55, + "learning_rate": 3.848652863207139e-05, + "loss": 1.1704, + "step": 133400 + }, + { + "epoch": 11.55, + "learning_rate": 3.8485662306159584e-05, + "loss": 1.219, + "step": 133410 + }, + { + "epoch": 11.55, + "learning_rate": 3.848479598024777e-05, + "loss": 1.2043, + "step": 133420 + }, + { + "epoch": 11.55, + "learning_rate": 3.8483929654335966e-05, + "loss": 1.1493, + "step": 133430 + }, + { + "epoch": 11.55, + "learning_rate": 3.848306332842416e-05, + "loss": 1.2057, + "step": 133440 + }, + { + "epoch": 11.55, + "learning_rate": 3.848219700251235e-05, + "loss": 1.1795, + "step": 133450 + }, + { + "epoch": 11.55, + "learning_rate": 3.848133067660054e-05, + "loss": 1.1572, + "step": 133460 + }, + { + "epoch": 11.55, + "learning_rate": 3.848046435068873e-05, + "loss": 1.1945, + "step": 133470 + }, + { + "epoch": 11.55, + "learning_rate": 3.8479598024776924e-05, + "loss": 1.1408, + "step": 133480 + }, + { + "epoch": 11.55, + "learning_rate": 3.847873169886512e-05, + "loss": 1.1257, + "step": 133490 + }, + { + "epoch": 11.56, + "learning_rate": 3.847786537295331e-05, + "loss": 1.1865, + "step": 133500 + }, + { + "epoch": 11.56, + "learning_rate": 3.84769990470415e-05, + "loss": 1.1739, + "step": 133510 + }, + { + "epoch": 11.56, + "learning_rate": 3.8476132721129696e-05, + "loss": 1.1581, + "step": 133520 + }, + { + "epoch": 11.56, + "learning_rate": 3.847526639521788e-05, + "loss": 1.1401, + "step": 133530 + }, + { + "epoch": 11.56, + "learning_rate": 3.847440006930608e-05, + "loss": 1.1031, + "step": 133540 + }, + { + "epoch": 11.56, + "learning_rate": 3.847353374339427e-05, + "loss": 1.1317, + "step": 133550 + }, + { + "epoch": 11.56, + "learning_rate": 3.847266741748246e-05, + "loss": 1.1325, + "step": 133560 + }, + { + "epoch": 11.56, + "learning_rate": 3.8471801091570654e-05, + "loss": 1.1754, + "step": 133570 + }, + { + "epoch": 11.56, + "learning_rate": 3.847093476565884e-05, + "loss": 1.187, + "step": 133580 + }, + { + "epoch": 11.56, + "learning_rate": 3.8470068439747036e-05, + "loss": 1.1406, + "step": 133590 + }, + { + "epoch": 11.56, + "learning_rate": 3.846920211383523e-05, + "loss": 1.1165, + "step": 133600 + }, + { + "epoch": 11.56, + "learning_rate": 3.846833578792342e-05, + "loss": 1.1255, + "step": 133610 + }, + { + "epoch": 11.57, + "learning_rate": 3.846746946201161e-05, + "loss": 1.1969, + "step": 133620 + }, + { + "epoch": 11.57, + "learning_rate": 3.846660313609981e-05, + "loss": 1.1928, + "step": 133630 + }, + { + "epoch": 11.57, + "learning_rate": 3.8465736810187995e-05, + "loss": 1.2506, + "step": 133640 + }, + { + "epoch": 11.57, + "learning_rate": 3.846487048427619e-05, + "loss": 1.1053, + "step": 133650 + }, + { + "epoch": 11.57, + "learning_rate": 3.8464004158364384e-05, + "loss": 1.1516, + "step": 133660 + }, + { + "epoch": 11.57, + "learning_rate": 3.846313783245257e-05, + "loss": 1.1523, + "step": 133670 + }, + { + "epoch": 11.57, + "learning_rate": 3.8462271506540766e-05, + "loss": 1.2708, + "step": 133680 + }, + { + "epoch": 11.57, + "learning_rate": 3.8461405180628954e-05, + "loss": 1.1597, + "step": 133690 + }, + { + "epoch": 11.57, + "learning_rate": 3.846053885471715e-05, + "loss": 1.1195, + "step": 133700 + }, + { + "epoch": 11.57, + "learning_rate": 3.845967252880534e-05, + "loss": 1.1833, + "step": 133710 + }, + { + "epoch": 11.57, + "learning_rate": 3.845880620289353e-05, + "loss": 1.1177, + "step": 133720 + }, + { + "epoch": 11.58, + "learning_rate": 3.8457939876981725e-05, + "loss": 1.1912, + "step": 133730 + }, + { + "epoch": 11.58, + "learning_rate": 3.845707355106992e-05, + "loss": 1.1467, + "step": 133740 + }, + { + "epoch": 11.58, + "learning_rate": 3.845620722515811e-05, + "loss": 1.2255, + "step": 133750 + }, + { + "epoch": 11.58, + "learning_rate": 3.84553408992463e-05, + "loss": 1.166, + "step": 133760 + }, + { + "epoch": 11.58, + "learning_rate": 3.845447457333449e-05, + "loss": 1.1712, + "step": 133770 + }, + { + "epoch": 11.58, + "learning_rate": 3.8453608247422684e-05, + "loss": 1.1239, + "step": 133780 + }, + { + "epoch": 11.58, + "learning_rate": 3.845274192151088e-05, + "loss": 1.2237, + "step": 133790 + }, + { + "epoch": 11.58, + "learning_rate": 3.8451875595599066e-05, + "loss": 1.1487, + "step": 133800 + }, + { + "epoch": 11.58, + "learning_rate": 3.845100926968726e-05, + "loss": 1.2028, + "step": 133810 + }, + { + "epoch": 11.58, + "learning_rate": 3.8450142943775455e-05, + "loss": 1.1827, + "step": 133820 + }, + { + "epoch": 11.58, + "learning_rate": 3.844927661786364e-05, + "loss": 1.1708, + "step": 133830 + }, + { + "epoch": 11.58, + "learning_rate": 3.844841029195184e-05, + "loss": 1.1484, + "step": 133840 + }, + { + "epoch": 11.59, + "learning_rate": 3.844754396604003e-05, + "loss": 1.1547, + "step": 133850 + }, + { + "epoch": 11.59, + "learning_rate": 3.844667764012822e-05, + "loss": 1.2122, + "step": 133860 + }, + { + "epoch": 11.59, + "learning_rate": 3.844581131421641e-05, + "loss": 1.177, + "step": 133870 + }, + { + "epoch": 11.59, + "learning_rate": 3.84449449883046e-05, + "loss": 1.147, + "step": 133880 + }, + { + "epoch": 11.59, + "learning_rate": 3.8444078662392795e-05, + "loss": 1.146, + "step": 133890 + }, + { + "epoch": 11.59, + "learning_rate": 3.844321233648099e-05, + "loss": 1.1928, + "step": 133900 + }, + { + "epoch": 11.59, + "learning_rate": 3.844234601056918e-05, + "loss": 1.187, + "step": 133910 + }, + { + "epoch": 11.59, + "learning_rate": 3.844147968465737e-05, + "loss": 1.1625, + "step": 133920 + }, + { + "epoch": 11.59, + "learning_rate": 3.8440613358745566e-05, + "loss": 1.1179, + "step": 133930 + }, + { + "epoch": 11.59, + "learning_rate": 3.8439747032833754e-05, + "loss": 1.1801, + "step": 133940 + }, + { + "epoch": 11.59, + "learning_rate": 3.843888070692195e-05, + "loss": 1.1266, + "step": 133950 + }, + { + "epoch": 11.6, + "learning_rate": 3.843801438101014e-05, + "loss": 1.1518, + "step": 133960 + }, + { + "epoch": 11.6, + "learning_rate": 3.843714805509833e-05, + "loss": 1.1797, + "step": 133970 + }, + { + "epoch": 11.6, + "learning_rate": 3.8436281729186525e-05, + "loss": 1.242, + "step": 133980 + }, + { + "epoch": 11.6, + "learning_rate": 3.843541540327471e-05, + "loss": 1.2369, + "step": 133990 + }, + { + "epoch": 11.6, + "learning_rate": 3.843454907736291e-05, + "loss": 1.1785, + "step": 134000 + }, + { + "epoch": 11.6, + "learning_rate": 3.84336827514511e-05, + "loss": 1.1694, + "step": 134010 + }, + { + "epoch": 11.6, + "learning_rate": 3.843281642553929e-05, + "loss": 1.2229, + "step": 134020 + }, + { + "epoch": 11.6, + "learning_rate": 3.8431950099627484e-05, + "loss": 1.1554, + "step": 134030 + }, + { + "epoch": 11.6, + "learning_rate": 3.843108377371568e-05, + "loss": 1.1744, + "step": 134040 + }, + { + "epoch": 11.6, + "learning_rate": 3.8430217447803866e-05, + "loss": 1.1437, + "step": 134050 + }, + { + "epoch": 11.6, + "learning_rate": 3.842935112189206e-05, + "loss": 1.2159, + "step": 134060 + }, + { + "epoch": 11.6, + "learning_rate": 3.8428484795980255e-05, + "loss": 1.1954, + "step": 134070 + }, + { + "epoch": 11.61, + "learning_rate": 3.842761847006844e-05, + "loss": 1.2092, + "step": 134080 + }, + { + "epoch": 11.61, + "learning_rate": 3.842675214415664e-05, + "loss": 1.1804, + "step": 134090 + }, + { + "epoch": 11.61, + "learning_rate": 3.8425885818244825e-05, + "loss": 1.1552, + "step": 134100 + }, + { + "epoch": 11.61, + "learning_rate": 3.842501949233302e-05, + "loss": 1.1605, + "step": 134110 + }, + { + "epoch": 11.61, + "learning_rate": 3.8424153166421214e-05, + "loss": 1.1775, + "step": 134120 + }, + { + "epoch": 11.61, + "learning_rate": 3.84232868405094e-05, + "loss": 1.1974, + "step": 134130 + }, + { + "epoch": 11.61, + "learning_rate": 3.8422420514597596e-05, + "loss": 1.1485, + "step": 134140 + }, + { + "epoch": 11.61, + "learning_rate": 3.842155418868579e-05, + "loss": 1.1683, + "step": 134150 + }, + { + "epoch": 11.61, + "learning_rate": 3.842068786277398e-05, + "loss": 1.1716, + "step": 134160 + }, + { + "epoch": 11.61, + "learning_rate": 3.841982153686217e-05, + "loss": 1.1688, + "step": 134170 + }, + { + "epoch": 11.61, + "learning_rate": 3.841895521095037e-05, + "loss": 1.1156, + "step": 134180 + }, + { + "epoch": 11.61, + "learning_rate": 3.8418088885038554e-05, + "loss": 1.1517, + "step": 134190 + }, + { + "epoch": 11.62, + "learning_rate": 3.841722255912675e-05, + "loss": 1.1805, + "step": 134200 + }, + { + "epoch": 11.62, + "learning_rate": 3.8416356233214937e-05, + "loss": 1.187, + "step": 134210 + }, + { + "epoch": 11.62, + "learning_rate": 3.841548990730313e-05, + "loss": 1.1094, + "step": 134220 + }, + { + "epoch": 11.62, + "learning_rate": 3.8414623581391325e-05, + "loss": 1.145, + "step": 134230 + }, + { + "epoch": 11.62, + "learning_rate": 3.841375725547951e-05, + "loss": 1.1487, + "step": 134240 + }, + { + "epoch": 11.62, + "learning_rate": 3.841289092956771e-05, + "loss": 1.2317, + "step": 134250 + }, + { + "epoch": 11.62, + "learning_rate": 3.84120246036559e-05, + "loss": 1.146, + "step": 134260 + }, + { + "epoch": 11.62, + "learning_rate": 3.841115827774409e-05, + "loss": 1.1686, + "step": 134270 + }, + { + "epoch": 11.62, + "learning_rate": 3.8410291951832284e-05, + "loss": 1.1826, + "step": 134280 + }, + { + "epoch": 11.62, + "learning_rate": 3.840942562592048e-05, + "loss": 1.092, + "step": 134290 + }, + { + "epoch": 11.62, + "learning_rate": 3.8408559300008666e-05, + "loss": 1.1254, + "step": 134300 + }, + { + "epoch": 11.63, + "learning_rate": 3.840769297409686e-05, + "loss": 1.1027, + "step": 134310 + }, + { + "epoch": 11.63, + "learning_rate": 3.840682664818505e-05, + "loss": 1.13, + "step": 134320 + }, + { + "epoch": 11.63, + "learning_rate": 3.840596032227324e-05, + "loss": 1.1799, + "step": 134330 + }, + { + "epoch": 11.63, + "learning_rate": 3.840509399636144e-05, + "loss": 1.1455, + "step": 134340 + }, + { + "epoch": 11.63, + "learning_rate": 3.8404227670449625e-05, + "loss": 1.2032, + "step": 134350 + }, + { + "epoch": 11.63, + "learning_rate": 3.840336134453782e-05, + "loss": 1.1703, + "step": 134360 + }, + { + "epoch": 11.63, + "learning_rate": 3.8402495018626014e-05, + "loss": 1.1543, + "step": 134370 + }, + { + "epoch": 11.63, + "learning_rate": 3.84016286927142e-05, + "loss": 1.1384, + "step": 134380 + }, + { + "epoch": 11.63, + "learning_rate": 3.8400762366802396e-05, + "loss": 1.1297, + "step": 134390 + }, + { + "epoch": 11.63, + "learning_rate": 3.839989604089059e-05, + "loss": 1.1564, + "step": 134400 + }, + { + "epoch": 11.63, + "learning_rate": 3.839902971497878e-05, + "loss": 1.1269, + "step": 134410 + }, + { + "epoch": 11.63, + "learning_rate": 3.839816338906697e-05, + "loss": 1.2413, + "step": 134420 + }, + { + "epoch": 11.64, + "learning_rate": 3.839729706315516e-05, + "loss": 1.1745, + "step": 134430 + }, + { + "epoch": 11.64, + "learning_rate": 3.8396430737243355e-05, + "loss": 1.1858, + "step": 134440 + }, + { + "epoch": 11.64, + "learning_rate": 3.839556441133155e-05, + "loss": 1.1981, + "step": 134450 + }, + { + "epoch": 11.64, + "learning_rate": 3.839469808541974e-05, + "loss": 1.1481, + "step": 134460 + }, + { + "epoch": 11.64, + "learning_rate": 3.839383175950793e-05, + "loss": 1.2111, + "step": 134470 + }, + { + "epoch": 11.64, + "learning_rate": 3.8392965433596126e-05, + "loss": 1.1351, + "step": 134480 + }, + { + "epoch": 11.64, + "learning_rate": 3.8392099107684314e-05, + "loss": 1.1266, + "step": 134490 + }, + { + "epoch": 11.64, + "learning_rate": 3.839123278177251e-05, + "loss": 1.1866, + "step": 134500 + }, + { + "epoch": 11.64, + "learning_rate": 3.8390366455860696e-05, + "loss": 1.1118, + "step": 134510 + }, + { + "epoch": 11.64, + "learning_rate": 3.838950012994889e-05, + "loss": 1.1562, + "step": 134520 + }, + { + "epoch": 11.64, + "learning_rate": 3.8388633804037085e-05, + "loss": 1.1515, + "step": 134530 + }, + { + "epoch": 11.65, + "learning_rate": 3.838776747812527e-05, + "loss": 1.1687, + "step": 134540 + }, + { + "epoch": 11.65, + "learning_rate": 3.838690115221347e-05, + "loss": 1.2175, + "step": 134550 + }, + { + "epoch": 11.65, + "learning_rate": 3.838603482630166e-05, + "loss": 1.1456, + "step": 134560 + }, + { + "epoch": 11.65, + "learning_rate": 3.838516850038985e-05, + "loss": 1.1383, + "step": 134570 + }, + { + "epoch": 11.65, + "learning_rate": 3.838430217447804e-05, + "loss": 1.116, + "step": 134580 + }, + { + "epoch": 11.65, + "learning_rate": 3.838343584856624e-05, + "loss": 1.1634, + "step": 134590 + }, + { + "epoch": 11.65, + "learning_rate": 3.8382569522654425e-05, + "loss": 1.2345, + "step": 134600 + }, + { + "epoch": 11.65, + "learning_rate": 3.838170319674262e-05, + "loss": 1.1722, + "step": 134610 + }, + { + "epoch": 11.65, + "learning_rate": 3.838083687083081e-05, + "loss": 1.2131, + "step": 134620 + }, + { + "epoch": 11.65, + "learning_rate": 3.8379970544919e-05, + "loss": 1.1716, + "step": 134630 + }, + { + "epoch": 11.65, + "learning_rate": 3.8379104219007196e-05, + "loss": 1.1293, + "step": 134640 + }, + { + "epoch": 11.65, + "learning_rate": 3.8378237893095384e-05, + "loss": 1.1478, + "step": 134650 + }, + { + "epoch": 11.66, + "learning_rate": 3.837737156718358e-05, + "loss": 1.1517, + "step": 134660 + }, + { + "epoch": 11.66, + "learning_rate": 3.837650524127177e-05, + "loss": 1.1433, + "step": 134670 + }, + { + "epoch": 11.66, + "learning_rate": 3.837563891535996e-05, + "loss": 1.1738, + "step": 134680 + }, + { + "epoch": 11.66, + "learning_rate": 3.8374772589448155e-05, + "loss": 1.1586, + "step": 134690 + }, + { + "epoch": 11.66, + "learning_rate": 3.837390626353635e-05, + "loss": 1.1743, + "step": 134700 + }, + { + "epoch": 11.66, + "learning_rate": 3.837303993762454e-05, + "loss": 1.2005, + "step": 134710 + }, + { + "epoch": 11.66, + "learning_rate": 3.837217361171273e-05, + "loss": 1.1235, + "step": 134720 + }, + { + "epoch": 11.66, + "learning_rate": 3.837130728580092e-05, + "loss": 1.1795, + "step": 134730 + }, + { + "epoch": 11.66, + "learning_rate": 3.8370440959889114e-05, + "loss": 1.1867, + "step": 134740 + }, + { + "epoch": 11.66, + "learning_rate": 3.836957463397731e-05, + "loss": 1.0919, + "step": 134750 + }, + { + "epoch": 11.66, + "learning_rate": 3.8368708308065496e-05, + "loss": 1.147, + "step": 134760 + }, + { + "epoch": 11.67, + "learning_rate": 3.836784198215369e-05, + "loss": 1.1574, + "step": 134770 + }, + { + "epoch": 11.67, + "learning_rate": 3.8366975656241885e-05, + "loss": 1.1593, + "step": 134780 + }, + { + "epoch": 11.67, + "learning_rate": 3.836610933033007e-05, + "loss": 1.2345, + "step": 134790 + }, + { + "epoch": 11.67, + "learning_rate": 3.836524300441827e-05, + "loss": 1.1813, + "step": 134800 + }, + { + "epoch": 11.67, + "learning_rate": 3.836437667850646e-05, + "loss": 1.1809, + "step": 134810 + }, + { + "epoch": 11.67, + "learning_rate": 3.836351035259465e-05, + "loss": 1.186, + "step": 134820 + }, + { + "epoch": 11.67, + "learning_rate": 3.8362644026682844e-05, + "loss": 1.166, + "step": 134830 + }, + { + "epoch": 11.67, + "learning_rate": 3.836177770077103e-05, + "loss": 1.1124, + "step": 134840 + }, + { + "epoch": 11.67, + "learning_rate": 3.8360911374859226e-05, + "loss": 1.1417, + "step": 134850 + }, + { + "epoch": 11.67, + "learning_rate": 3.836004504894742e-05, + "loss": 1.189, + "step": 134860 + }, + { + "epoch": 11.67, + "learning_rate": 3.835917872303561e-05, + "loss": 1.1225, + "step": 134870 + }, + { + "epoch": 11.67, + "learning_rate": 3.83583123971238e-05, + "loss": 1.1698, + "step": 134880 + }, + { + "epoch": 11.68, + "learning_rate": 3.8357446071212e-05, + "loss": 1.1463, + "step": 134890 + }, + { + "epoch": 11.68, + "learning_rate": 3.8356579745300184e-05, + "loss": 1.2128, + "step": 134900 + }, + { + "epoch": 11.68, + "learning_rate": 3.835571341938838e-05, + "loss": 1.1604, + "step": 134910 + }, + { + "epoch": 11.68, + "learning_rate": 3.835484709347657e-05, + "loss": 1.1857, + "step": 134920 + }, + { + "epoch": 11.68, + "learning_rate": 3.835398076756476e-05, + "loss": 1.0651, + "step": 134930 + }, + { + "epoch": 11.68, + "learning_rate": 3.8353114441652955e-05, + "loss": 1.1176, + "step": 134940 + }, + { + "epoch": 11.68, + "learning_rate": 3.835224811574114e-05, + "loss": 1.1607, + "step": 134950 + }, + { + "epoch": 11.68, + "learning_rate": 3.835138178982934e-05, + "loss": 1.207, + "step": 134960 + }, + { + "epoch": 11.68, + "learning_rate": 3.835051546391753e-05, + "loss": 1.1802, + "step": 134970 + }, + { + "epoch": 11.68, + "learning_rate": 3.834964913800572e-05, + "loss": 1.1272, + "step": 134980 + }, + { + "epoch": 11.68, + "learning_rate": 3.8348782812093914e-05, + "loss": 1.2105, + "step": 134990 + }, + { + "epoch": 11.69, + "learning_rate": 3.834791648618211e-05, + "loss": 1.1602, + "step": 135000 + }, + { + "epoch": 11.69, + "learning_rate": 3.8347050160270296e-05, + "loss": 1.1484, + "step": 135010 + }, + { + "epoch": 11.69, + "learning_rate": 3.834618383435849e-05, + "loss": 1.2546, + "step": 135020 + }, + { + "epoch": 11.69, + "learning_rate": 3.8345317508446685e-05, + "loss": 1.1864, + "step": 135030 + }, + { + "epoch": 11.69, + "learning_rate": 3.834445118253487e-05, + "loss": 1.1801, + "step": 135040 + }, + { + "epoch": 11.69, + "learning_rate": 3.834358485662307e-05, + "loss": 1.1633, + "step": 135050 + }, + { + "epoch": 11.69, + "learning_rate": 3.8342718530711255e-05, + "loss": 1.1808, + "step": 135060 + }, + { + "epoch": 11.69, + "learning_rate": 3.834185220479945e-05, + "loss": 1.1734, + "step": 135070 + }, + { + "epoch": 11.69, + "learning_rate": 3.8340985878887644e-05, + "loss": 1.1033, + "step": 135080 + }, + { + "epoch": 11.69, + "learning_rate": 3.834011955297583e-05, + "loss": 1.1738, + "step": 135090 + }, + { + "epoch": 11.69, + "learning_rate": 3.8339253227064026e-05, + "loss": 1.1449, + "step": 135100 + }, + { + "epoch": 11.69, + "learning_rate": 3.833838690115222e-05, + "loss": 1.1355, + "step": 135110 + }, + { + "epoch": 11.7, + "learning_rate": 3.833752057524041e-05, + "loss": 1.2091, + "step": 135120 + }, + { + "epoch": 11.7, + "learning_rate": 3.83366542493286e-05, + "loss": 1.2181, + "step": 135130 + }, + { + "epoch": 11.7, + "learning_rate": 3.83357879234168e-05, + "loss": 1.1324, + "step": 135140 + }, + { + "epoch": 11.7, + "learning_rate": 3.8334921597504985e-05, + "loss": 1.1676, + "step": 135150 + }, + { + "epoch": 11.7, + "learning_rate": 3.833405527159318e-05, + "loss": 1.2143, + "step": 135160 + }, + { + "epoch": 11.7, + "learning_rate": 3.833318894568137e-05, + "loss": 1.1923, + "step": 135170 + }, + { + "epoch": 11.7, + "learning_rate": 3.833232261976956e-05, + "loss": 1.2234, + "step": 135180 + }, + { + "epoch": 11.7, + "learning_rate": 3.8331456293857756e-05, + "loss": 1.1949, + "step": 135190 + }, + { + "epoch": 11.7, + "learning_rate": 3.8330589967945943e-05, + "loss": 1.2242, + "step": 135200 + }, + { + "epoch": 11.7, + "learning_rate": 3.832972364203414e-05, + "loss": 1.1535, + "step": 135210 + }, + { + "epoch": 11.7, + "learning_rate": 3.832885731612233e-05, + "loss": 1.105, + "step": 135220 + }, + { + "epoch": 11.7, + "learning_rate": 3.832799099021052e-05, + "loss": 1.1791, + "step": 135230 + }, + { + "epoch": 11.71, + "learning_rate": 3.8327124664298715e-05, + "loss": 1.1681, + "step": 135240 + }, + { + "epoch": 11.71, + "learning_rate": 3.832625833838691e-05, + "loss": 1.1496, + "step": 135250 + }, + { + "epoch": 11.71, + "learning_rate": 3.83253920124751e-05, + "loss": 1.1497, + "step": 135260 + }, + { + "epoch": 11.71, + "learning_rate": 3.832452568656329e-05, + "loss": 1.143, + "step": 135270 + }, + { + "epoch": 11.71, + "learning_rate": 3.832365936065148e-05, + "loss": 1.0935, + "step": 135280 + }, + { + "epoch": 11.71, + "learning_rate": 3.832279303473967e-05, + "loss": 1.2048, + "step": 135290 + }, + { + "epoch": 11.71, + "learning_rate": 3.832192670882787e-05, + "loss": 1.1395, + "step": 135300 + }, + { + "epoch": 11.71, + "learning_rate": 3.8321060382916055e-05, + "loss": 1.1837, + "step": 135310 + }, + { + "epoch": 11.71, + "learning_rate": 3.832019405700425e-05, + "loss": 1.151, + "step": 135320 + }, + { + "epoch": 11.71, + "learning_rate": 3.8319327731092444e-05, + "loss": 1.1686, + "step": 135330 + }, + { + "epoch": 11.71, + "learning_rate": 3.831846140518063e-05, + "loss": 1.1854, + "step": 135340 + }, + { + "epoch": 11.72, + "learning_rate": 3.8317595079268826e-05, + "loss": 1.0916, + "step": 135350 + }, + { + "epoch": 11.72, + "learning_rate": 3.8316728753357014e-05, + "loss": 1.1219, + "step": 135360 + }, + { + "epoch": 11.72, + "learning_rate": 3.831586242744521e-05, + "loss": 1.1861, + "step": 135370 + }, + { + "epoch": 11.72, + "learning_rate": 3.83149961015334e-05, + "loss": 1.1787, + "step": 135380 + }, + { + "epoch": 11.72, + "learning_rate": 3.831412977562159e-05, + "loss": 1.1298, + "step": 135390 + }, + { + "epoch": 11.72, + "learning_rate": 3.8313263449709785e-05, + "loss": 1.1495, + "step": 135400 + }, + { + "epoch": 11.72, + "learning_rate": 3.831239712379798e-05, + "loss": 1.1965, + "step": 135410 + }, + { + "epoch": 11.72, + "learning_rate": 3.831153079788617e-05, + "loss": 1.1306, + "step": 135420 + }, + { + "epoch": 11.72, + "learning_rate": 3.831066447197436e-05, + "loss": 1.1637, + "step": 135430 + }, + { + "epoch": 11.72, + "learning_rate": 3.8309798146062556e-05, + "loss": 1.1591, + "step": 135440 + }, + { + "epoch": 11.72, + "learning_rate": 3.8308931820150744e-05, + "loss": 1.1015, + "step": 135450 + }, + { + "epoch": 11.72, + "learning_rate": 3.830806549423894e-05, + "loss": 1.1953, + "step": 135460 + }, + { + "epoch": 11.73, + "learning_rate": 3.8307199168327126e-05, + "loss": 1.2102, + "step": 135470 + }, + { + "epoch": 11.73, + "learning_rate": 3.830633284241532e-05, + "loss": 1.1302, + "step": 135480 + }, + { + "epoch": 11.73, + "learning_rate": 3.8305466516503515e-05, + "loss": 1.1785, + "step": 135490 + }, + { + "epoch": 11.73, + "learning_rate": 3.83046001905917e-05, + "loss": 1.2059, + "step": 135500 + }, + { + "epoch": 11.73, + "learning_rate": 3.83037338646799e-05, + "loss": 1.1479, + "step": 135510 + }, + { + "epoch": 11.73, + "learning_rate": 3.830286753876809e-05, + "loss": 1.1506, + "step": 135520 + }, + { + "epoch": 11.73, + "learning_rate": 3.830200121285628e-05, + "loss": 1.1661, + "step": 135530 + }, + { + "epoch": 11.73, + "learning_rate": 3.8301134886944474e-05, + "loss": 1.1584, + "step": 135540 + }, + { + "epoch": 11.73, + "learning_rate": 3.830026856103267e-05, + "loss": 1.1722, + "step": 135550 + }, + { + "epoch": 11.73, + "learning_rate": 3.8299402235120856e-05, + "loss": 1.15, + "step": 135560 + }, + { + "epoch": 11.73, + "learning_rate": 3.829853590920905e-05, + "loss": 1.158, + "step": 135570 + }, + { + "epoch": 11.74, + "learning_rate": 3.829766958329724e-05, + "loss": 1.146, + "step": 135580 + }, + { + "epoch": 11.74, + "learning_rate": 3.829680325738543e-05, + "loss": 1.1926, + "step": 135590 + }, + { + "epoch": 11.74, + "learning_rate": 3.829593693147363e-05, + "loss": 1.1439, + "step": 135600 + }, + { + "epoch": 11.74, + "learning_rate": 3.8295070605561814e-05, + "loss": 1.1724, + "step": 135610 + }, + { + "epoch": 11.74, + "learning_rate": 3.829420427965001e-05, + "loss": 1.116, + "step": 135620 + }, + { + "epoch": 11.74, + "learning_rate": 3.82933379537382e-05, + "loss": 1.1496, + "step": 135630 + }, + { + "epoch": 11.74, + "learning_rate": 3.829247162782639e-05, + "loss": 1.1138, + "step": 135640 + }, + { + "epoch": 11.74, + "learning_rate": 3.8291605301914585e-05, + "loss": 1.179, + "step": 135650 + }, + { + "epoch": 11.74, + "learning_rate": 3.829073897600278e-05, + "loss": 1.1592, + "step": 135660 + }, + { + "epoch": 11.74, + "learning_rate": 3.828987265009097e-05, + "loss": 1.1836, + "step": 135670 + }, + { + "epoch": 11.74, + "learning_rate": 3.828900632417916e-05, + "loss": 1.2043, + "step": 135680 + }, + { + "epoch": 11.74, + "learning_rate": 3.828813999826735e-05, + "loss": 1.2091, + "step": 135690 + }, + { + "epoch": 11.75, + "learning_rate": 3.8287273672355544e-05, + "loss": 1.2125, + "step": 135700 + }, + { + "epoch": 11.75, + "learning_rate": 3.828640734644374e-05, + "loss": 1.2126, + "step": 135710 + }, + { + "epoch": 11.75, + "learning_rate": 3.8285541020531926e-05, + "loss": 1.1841, + "step": 135720 + }, + { + "epoch": 11.75, + "learning_rate": 3.828467469462012e-05, + "loss": 1.1219, + "step": 135730 + }, + { + "epoch": 11.75, + "learning_rate": 3.8283808368708315e-05, + "loss": 1.1339, + "step": 135740 + }, + { + "epoch": 11.75, + "learning_rate": 3.82829420427965e-05, + "loss": 1.1523, + "step": 135750 + }, + { + "epoch": 11.75, + "learning_rate": 3.82820757168847e-05, + "loss": 1.1049, + "step": 135760 + }, + { + "epoch": 11.75, + "learning_rate": 3.8281209390972885e-05, + "loss": 1.1021, + "step": 135770 + }, + { + "epoch": 11.75, + "learning_rate": 3.828034306506107e-05, + "loss": 1.1646, + "step": 135780 + }, + { + "epoch": 11.75, + "learning_rate": 3.827947673914927e-05, + "loss": 1.1084, + "step": 135790 + }, + { + "epoch": 11.75, + "learning_rate": 3.827861041323746e-05, + "loss": 1.1419, + "step": 135800 + }, + { + "epoch": 11.76, + "learning_rate": 3.827774408732565e-05, + "loss": 1.1793, + "step": 135810 + }, + { + "epoch": 11.76, + "learning_rate": 3.8276877761413844e-05, + "loss": 1.0881, + "step": 135820 + }, + { + "epoch": 11.76, + "learning_rate": 3.827601143550204e-05, + "loss": 1.1949, + "step": 135830 + }, + { + "epoch": 11.76, + "learning_rate": 3.8275145109590226e-05, + "loss": 1.131, + "step": 135840 + }, + { + "epoch": 11.76, + "learning_rate": 3.827427878367842e-05, + "loss": 1.1066, + "step": 135850 + }, + { + "epoch": 11.76, + "learning_rate": 3.827341245776661e-05, + "loss": 1.0975, + "step": 135860 + }, + { + "epoch": 11.76, + "learning_rate": 3.82725461318548e-05, + "loss": 1.1128, + "step": 135870 + }, + { + "epoch": 11.76, + "learning_rate": 3.8271679805943e-05, + "loss": 1.1576, + "step": 135880 + }, + { + "epoch": 11.76, + "learning_rate": 3.8270813480031185e-05, + "loss": 1.1948, + "step": 135890 + }, + { + "epoch": 11.76, + "learning_rate": 3.826994715411938e-05, + "loss": 1.171, + "step": 135900 + }, + { + "epoch": 11.76, + "learning_rate": 3.8269080828207573e-05, + "loss": 1.1812, + "step": 135910 + }, + { + "epoch": 11.76, + "learning_rate": 3.826821450229576e-05, + "loss": 1.1602, + "step": 135920 + }, + { + "epoch": 11.77, + "learning_rate": 3.8267348176383956e-05, + "loss": 1.1904, + "step": 135930 + }, + { + "epoch": 11.77, + "learning_rate": 3.826648185047215e-05, + "loss": 1.1745, + "step": 135940 + }, + { + "epoch": 11.77, + "learning_rate": 3.826561552456034e-05, + "loss": 1.1603, + "step": 135950 + }, + { + "epoch": 11.77, + "learning_rate": 3.826474919864853e-05, + "loss": 1.1272, + "step": 135960 + }, + { + "epoch": 11.77, + "learning_rate": 3.826388287273672e-05, + "loss": 1.1297, + "step": 135970 + }, + { + "epoch": 11.77, + "learning_rate": 3.8263016546824914e-05, + "loss": 1.1277, + "step": 135980 + }, + { + "epoch": 11.77, + "learning_rate": 3.826215022091311e-05, + "loss": 1.1612, + "step": 135990 + }, + { + "epoch": 11.77, + "learning_rate": 3.8261283895001296e-05, + "loss": 1.1258, + "step": 136000 + }, + { + "epoch": 11.77, + "learning_rate": 3.826041756908949e-05, + "loss": 1.163, + "step": 136010 + }, + { + "epoch": 11.77, + "learning_rate": 3.8259551243177685e-05, + "loss": 1.2135, + "step": 136020 + }, + { + "epoch": 11.77, + "learning_rate": 3.825868491726587e-05, + "loss": 1.1176, + "step": 136030 + }, + { + "epoch": 11.78, + "learning_rate": 3.825781859135407e-05, + "loss": 1.1612, + "step": 136040 + }, + { + "epoch": 11.78, + "learning_rate": 3.8256952265442255e-05, + "loss": 1.2481, + "step": 136050 + }, + { + "epoch": 11.78, + "learning_rate": 3.825608593953045e-05, + "loss": 1.1741, + "step": 136060 + }, + { + "epoch": 11.78, + "learning_rate": 3.8255219613618644e-05, + "loss": 1.1607, + "step": 136070 + }, + { + "epoch": 11.78, + "learning_rate": 3.825435328770683e-05, + "loss": 1.1749, + "step": 136080 + }, + { + "epoch": 11.78, + "learning_rate": 3.8253486961795026e-05, + "loss": 1.1887, + "step": 136090 + }, + { + "epoch": 11.78, + "learning_rate": 3.825262063588322e-05, + "loss": 1.1786, + "step": 136100 + }, + { + "epoch": 11.78, + "learning_rate": 3.825175430997141e-05, + "loss": 1.1898, + "step": 136110 + }, + { + "epoch": 11.78, + "learning_rate": 3.82508879840596e-05, + "loss": 1.1527, + "step": 136120 + }, + { + "epoch": 11.78, + "learning_rate": 3.82500216581478e-05, + "loss": 1.1593, + "step": 136130 + }, + { + "epoch": 11.78, + "learning_rate": 3.8249155332235985e-05, + "loss": 1.1134, + "step": 136140 + }, + { + "epoch": 11.78, + "learning_rate": 3.824828900632418e-05, + "loss": 1.129, + "step": 136150 + }, + { + "epoch": 11.79, + "learning_rate": 3.824742268041237e-05, + "loss": 1.1326, + "step": 136160 + }, + { + "epoch": 11.79, + "learning_rate": 3.824655635450056e-05, + "loss": 1.1457, + "step": 136170 + }, + { + "epoch": 11.79, + "learning_rate": 3.8245690028588756e-05, + "loss": 1.1011, + "step": 136180 + }, + { + "epoch": 11.79, + "learning_rate": 3.8244823702676944e-05, + "loss": 1.1058, + "step": 136190 + }, + { + "epoch": 11.79, + "learning_rate": 3.824395737676514e-05, + "loss": 1.2112, + "step": 136200 + }, + { + "epoch": 11.79, + "learning_rate": 3.824309105085333e-05, + "loss": 1.1414, + "step": 136210 + }, + { + "epoch": 11.79, + "learning_rate": 3.824222472494152e-05, + "loss": 1.1779, + "step": 136220 + }, + { + "epoch": 11.79, + "learning_rate": 3.8241358399029715e-05, + "loss": 1.199, + "step": 136230 + }, + { + "epoch": 11.79, + "learning_rate": 3.824049207311791e-05, + "loss": 1.109, + "step": 136240 + }, + { + "epoch": 11.79, + "learning_rate": 3.82396257472061e-05, + "loss": 1.17, + "step": 136250 + }, + { + "epoch": 11.79, + "learning_rate": 3.823875942129429e-05, + "loss": 1.1737, + "step": 136260 + }, + { + "epoch": 11.79, + "learning_rate": 3.823789309538248e-05, + "loss": 1.1558, + "step": 136270 + }, + { + "epoch": 11.8, + "learning_rate": 3.823702676947067e-05, + "loss": 1.1908, + "step": 136280 + }, + { + "epoch": 11.8, + "learning_rate": 3.823616044355887e-05, + "loss": 1.152, + "step": 136290 + }, + { + "epoch": 11.8, + "learning_rate": 3.8235294117647055e-05, + "loss": 1.1009, + "step": 136300 + }, + { + "epoch": 11.8, + "learning_rate": 3.823442779173525e-05, + "loss": 1.1683, + "step": 136310 + }, + { + "epoch": 11.8, + "learning_rate": 3.8233561465823444e-05, + "loss": 1.1727, + "step": 136320 + }, + { + "epoch": 11.8, + "learning_rate": 3.823269513991163e-05, + "loss": 1.1177, + "step": 136330 + }, + { + "epoch": 11.8, + "learning_rate": 3.8231828813999827e-05, + "loss": 1.1962, + "step": 136340 + }, + { + "epoch": 11.8, + "learning_rate": 3.823096248808802e-05, + "loss": 1.1713, + "step": 136350 + }, + { + "epoch": 11.8, + "learning_rate": 3.823009616217621e-05, + "loss": 1.1139, + "step": 136360 + }, + { + "epoch": 11.8, + "learning_rate": 3.82292298362644e-05, + "loss": 1.1147, + "step": 136370 + }, + { + "epoch": 11.8, + "learning_rate": 3.822836351035259e-05, + "loss": 1.1341, + "step": 136380 + }, + { + "epoch": 11.81, + "learning_rate": 3.8227497184440785e-05, + "loss": 1.1799, + "step": 136390 + }, + { + "epoch": 11.81, + "learning_rate": 3.822663085852898e-05, + "loss": 1.235, + "step": 136400 + }, + { + "epoch": 11.81, + "learning_rate": 3.822576453261717e-05, + "loss": 1.208, + "step": 136410 + }, + { + "epoch": 11.81, + "learning_rate": 3.822489820670536e-05, + "loss": 1.2014, + "step": 136420 + }, + { + "epoch": 11.81, + "learning_rate": 3.8224031880793556e-05, + "loss": 1.134, + "step": 136430 + }, + { + "epoch": 11.81, + "learning_rate": 3.8223165554881744e-05, + "loss": 1.1431, + "step": 136440 + }, + { + "epoch": 11.81, + "learning_rate": 3.822229922896994e-05, + "loss": 1.1605, + "step": 136450 + }, + { + "epoch": 11.81, + "learning_rate": 3.822143290305813e-05, + "loss": 1.1234, + "step": 136460 + }, + { + "epoch": 11.81, + "learning_rate": 3.822056657714632e-05, + "loss": 1.1625, + "step": 136470 + }, + { + "epoch": 11.81, + "learning_rate": 3.8219700251234515e-05, + "loss": 1.1848, + "step": 136480 + }, + { + "epoch": 11.81, + "learning_rate": 3.82188339253227e-05, + "loss": 1.0903, + "step": 136490 + }, + { + "epoch": 11.81, + "learning_rate": 3.82179675994109e-05, + "loss": 1.1532, + "step": 136500 + }, + { + "epoch": 11.82, + "learning_rate": 3.821710127349909e-05, + "loss": 1.1726, + "step": 136510 + }, + { + "epoch": 11.82, + "learning_rate": 3.821623494758728e-05, + "loss": 1.1951, + "step": 136520 + }, + { + "epoch": 11.82, + "learning_rate": 3.8215368621675474e-05, + "loss": 1.164, + "step": 136530 + }, + { + "epoch": 11.82, + "learning_rate": 3.821450229576367e-05, + "loss": 1.1403, + "step": 136540 + }, + { + "epoch": 11.82, + "learning_rate": 3.8213635969851856e-05, + "loss": 1.1594, + "step": 136550 + }, + { + "epoch": 11.82, + "learning_rate": 3.821276964394005e-05, + "loss": 1.1652, + "step": 136560 + }, + { + "epoch": 11.82, + "learning_rate": 3.8211903318028245e-05, + "loss": 1.1899, + "step": 136570 + }, + { + "epoch": 11.82, + "learning_rate": 3.821103699211643e-05, + "loss": 1.1802, + "step": 136580 + }, + { + "epoch": 11.82, + "learning_rate": 3.821017066620463e-05, + "loss": 1.2131, + "step": 136590 + }, + { + "epoch": 11.82, + "learning_rate": 3.8209304340292815e-05, + "loss": 1.1486, + "step": 136600 + }, + { + "epoch": 11.82, + "learning_rate": 3.820843801438101e-05, + "loss": 1.0913, + "step": 136610 + }, + { + "epoch": 11.83, + "learning_rate": 3.8207571688469203e-05, + "loss": 1.1441, + "step": 136620 + }, + { + "epoch": 11.83, + "learning_rate": 3.820670536255739e-05, + "loss": 1.1766, + "step": 136630 + }, + { + "epoch": 11.83, + "learning_rate": 3.8205839036645586e-05, + "loss": 1.1655, + "step": 136640 + }, + { + "epoch": 11.83, + "learning_rate": 3.820497271073378e-05, + "loss": 1.111, + "step": 136650 + }, + { + "epoch": 11.83, + "learning_rate": 3.820410638482197e-05, + "loss": 1.2381, + "step": 136660 + }, + { + "epoch": 11.83, + "learning_rate": 3.820324005891016e-05, + "loss": 1.1955, + "step": 136670 + }, + { + "epoch": 11.83, + "learning_rate": 3.8202373732998357e-05, + "loss": 1.1437, + "step": 136680 + }, + { + "epoch": 11.83, + "learning_rate": 3.8201507407086544e-05, + "loss": 1.1942, + "step": 136690 + }, + { + "epoch": 11.83, + "learning_rate": 3.820064108117474e-05, + "loss": 1.179, + "step": 136700 + }, + { + "epoch": 11.83, + "learning_rate": 3.8199774755262926e-05, + "loss": 1.2054, + "step": 136710 + }, + { + "epoch": 11.83, + "learning_rate": 3.819890842935112e-05, + "loss": 1.2209, + "step": 136720 + }, + { + "epoch": 11.83, + "learning_rate": 3.8198042103439315e-05, + "loss": 1.2134, + "step": 136730 + }, + { + "epoch": 11.84, + "learning_rate": 3.81971757775275e-05, + "loss": 1.1734, + "step": 136740 + }, + { + "epoch": 11.84, + "learning_rate": 3.81963094516157e-05, + "loss": 1.192, + "step": 136750 + }, + { + "epoch": 11.84, + "learning_rate": 3.819544312570389e-05, + "loss": 1.2063, + "step": 136760 + }, + { + "epoch": 11.84, + "learning_rate": 3.819457679979208e-05, + "loss": 1.1866, + "step": 136770 + }, + { + "epoch": 11.84, + "learning_rate": 3.8193710473880274e-05, + "loss": 1.1487, + "step": 136780 + }, + { + "epoch": 11.84, + "learning_rate": 3.819284414796847e-05, + "loss": 1.1866, + "step": 136790 + }, + { + "epoch": 11.84, + "learning_rate": 3.8191977822056656e-05, + "loss": 1.1885, + "step": 136800 + }, + { + "epoch": 11.84, + "learning_rate": 3.819111149614485e-05, + "loss": 1.1838, + "step": 136810 + }, + { + "epoch": 11.84, + "learning_rate": 3.819024517023304e-05, + "loss": 1.1705, + "step": 136820 + }, + { + "epoch": 11.84, + "learning_rate": 3.818937884432123e-05, + "loss": 1.1409, + "step": 136830 + }, + { + "epoch": 11.84, + "learning_rate": 3.818851251840943e-05, + "loss": 1.1242, + "step": 136840 + }, + { + "epoch": 11.85, + "learning_rate": 3.8187646192497615e-05, + "loss": 1.1449, + "step": 136850 + }, + { + "epoch": 11.85, + "learning_rate": 3.818677986658581e-05, + "loss": 1.1465, + "step": 136860 + }, + { + "epoch": 11.85, + "learning_rate": 3.8185913540674004e-05, + "loss": 1.2236, + "step": 136870 + }, + { + "epoch": 11.85, + "learning_rate": 3.818504721476219e-05, + "loss": 1.1211, + "step": 136880 + }, + { + "epoch": 11.85, + "learning_rate": 3.8184180888850386e-05, + "loss": 1.1016, + "step": 136890 + }, + { + "epoch": 11.85, + "learning_rate": 3.8183314562938574e-05, + "loss": 1.1285, + "step": 136900 + }, + { + "epoch": 11.85, + "learning_rate": 3.818244823702677e-05, + "loss": 1.1913, + "step": 136910 + }, + { + "epoch": 11.85, + "learning_rate": 3.818158191111496e-05, + "loss": 1.1657, + "step": 136920 + }, + { + "epoch": 11.85, + "learning_rate": 3.818071558520315e-05, + "loss": 1.1365, + "step": 136930 + }, + { + "epoch": 11.85, + "learning_rate": 3.8179849259291345e-05, + "loss": 1.1078, + "step": 136940 + }, + { + "epoch": 11.85, + "learning_rate": 3.817898293337954e-05, + "loss": 1.1389, + "step": 136950 + }, + { + "epoch": 11.85, + "learning_rate": 3.817811660746773e-05, + "loss": 1.1991, + "step": 136960 + }, + { + "epoch": 11.86, + "learning_rate": 3.817725028155592e-05, + "loss": 1.1573, + "step": 136970 + }, + { + "epoch": 11.86, + "learning_rate": 3.8176383955644116e-05, + "loss": 1.1652, + "step": 136980 + }, + { + "epoch": 11.86, + "learning_rate": 3.81755176297323e-05, + "loss": 1.1688, + "step": 136990 + }, + { + "epoch": 11.86, + "learning_rate": 3.81746513038205e-05, + "loss": 1.1432, + "step": 137000 + }, + { + "epoch": 11.86, + "learning_rate": 3.8173784977908685e-05, + "loss": 1.1199, + "step": 137010 + }, + { + "epoch": 11.86, + "learning_rate": 3.817291865199688e-05, + "loss": 1.1986, + "step": 137020 + }, + { + "epoch": 11.86, + "learning_rate": 3.8172052326085074e-05, + "loss": 1.1661, + "step": 137030 + }, + { + "epoch": 11.86, + "learning_rate": 3.817118600017326e-05, + "loss": 1.1752, + "step": 137040 + }, + { + "epoch": 11.86, + "learning_rate": 3.8170319674261456e-05, + "loss": 1.1319, + "step": 137050 + }, + { + "epoch": 11.86, + "learning_rate": 3.816945334834965e-05, + "loss": 1.1593, + "step": 137060 + }, + { + "epoch": 11.86, + "learning_rate": 3.816858702243784e-05, + "loss": 1.1868, + "step": 137070 + }, + { + "epoch": 11.87, + "learning_rate": 3.816772069652603e-05, + "loss": 1.1434, + "step": 137080 + }, + { + "epoch": 11.87, + "learning_rate": 3.816685437061423e-05, + "loss": 1.2143, + "step": 137090 + }, + { + "epoch": 11.87, + "learning_rate": 3.8165988044702415e-05, + "loss": 1.2007, + "step": 137100 + }, + { + "epoch": 11.87, + "learning_rate": 3.816512171879061e-05, + "loss": 1.1289, + "step": 137110 + }, + { + "epoch": 11.87, + "learning_rate": 3.81642553928788e-05, + "loss": 1.2013, + "step": 137120 + }, + { + "epoch": 11.87, + "learning_rate": 3.816338906696699e-05, + "loss": 1.1054, + "step": 137130 + }, + { + "epoch": 11.87, + "learning_rate": 3.8162522741055186e-05, + "loss": 1.1668, + "step": 137140 + }, + { + "epoch": 11.87, + "learning_rate": 3.8161656415143374e-05, + "loss": 1.1381, + "step": 137150 + }, + { + "epoch": 11.87, + "learning_rate": 3.816079008923157e-05, + "loss": 1.1641, + "step": 137160 + }, + { + "epoch": 11.87, + "learning_rate": 3.815992376331976e-05, + "loss": 1.1615, + "step": 137170 + }, + { + "epoch": 11.87, + "learning_rate": 3.815905743740795e-05, + "loss": 1.178, + "step": 137180 + }, + { + "epoch": 11.87, + "learning_rate": 3.8158191111496145e-05, + "loss": 1.1483, + "step": 137190 + }, + { + "epoch": 11.88, + "learning_rate": 3.815732478558434e-05, + "loss": 1.2345, + "step": 137200 + }, + { + "epoch": 11.88, + "learning_rate": 3.815645845967253e-05, + "loss": 1.2042, + "step": 137210 + }, + { + "epoch": 11.88, + "learning_rate": 3.815559213376072e-05, + "loss": 1.2033, + "step": 137220 + }, + { + "epoch": 11.88, + "learning_rate": 3.815472580784891e-05, + "loss": 1.1681, + "step": 137230 + }, + { + "epoch": 11.88, + "learning_rate": 3.8153859481937104e-05, + "loss": 1.1819, + "step": 137240 + }, + { + "epoch": 11.88, + "learning_rate": 3.81529931560253e-05, + "loss": 1.1192, + "step": 137250 + }, + { + "epoch": 11.88, + "learning_rate": 3.8152126830113486e-05, + "loss": 1.1265, + "step": 137260 + }, + { + "epoch": 11.88, + "learning_rate": 3.815126050420168e-05, + "loss": 1.2148, + "step": 137270 + }, + { + "epoch": 11.88, + "learning_rate": 3.8150394178289875e-05, + "loss": 1.2045, + "step": 137280 + }, + { + "epoch": 11.88, + "learning_rate": 3.814952785237806e-05, + "loss": 1.2171, + "step": 137290 + }, + { + "epoch": 11.88, + "learning_rate": 3.814866152646626e-05, + "loss": 1.16, + "step": 137300 + }, + { + "epoch": 11.88, + "learning_rate": 3.814779520055445e-05, + "loss": 1.0711, + "step": 137310 + }, + { + "epoch": 11.89, + "learning_rate": 3.814692887464264e-05, + "loss": 1.1256, + "step": 137320 + }, + { + "epoch": 11.89, + "learning_rate": 3.8146062548730833e-05, + "loss": 1.1662, + "step": 137330 + }, + { + "epoch": 11.89, + "learning_rate": 3.814519622281902e-05, + "loss": 1.1449, + "step": 137340 + }, + { + "epoch": 11.89, + "learning_rate": 3.8144329896907216e-05, + "loss": 1.2176, + "step": 137350 + }, + { + "epoch": 11.89, + "learning_rate": 3.814346357099541e-05, + "loss": 1.1602, + "step": 137360 + }, + { + "epoch": 11.89, + "learning_rate": 3.81425972450836e-05, + "loss": 1.1226, + "step": 137370 + }, + { + "epoch": 11.89, + "learning_rate": 3.814173091917179e-05, + "loss": 1.1593, + "step": 137380 + }, + { + "epoch": 11.89, + "learning_rate": 3.8140864593259987e-05, + "loss": 1.1641, + "step": 137390 + }, + { + "epoch": 11.89, + "learning_rate": 3.8139998267348174e-05, + "loss": 1.14, + "step": 137400 + }, + { + "epoch": 11.89, + "learning_rate": 3.813913194143637e-05, + "loss": 1.1306, + "step": 137410 + }, + { + "epoch": 11.89, + "learning_rate": 3.813826561552456e-05, + "loss": 1.1288, + "step": 137420 + }, + { + "epoch": 11.9, + "learning_rate": 3.813739928961275e-05, + "loss": 1.1591, + "step": 137430 + }, + { + "epoch": 11.9, + "learning_rate": 3.8136532963700945e-05, + "loss": 1.1315, + "step": 137440 + }, + { + "epoch": 11.9, + "learning_rate": 3.813566663778913e-05, + "loss": 1.209, + "step": 137450 + }, + { + "epoch": 11.9, + "learning_rate": 3.813480031187733e-05, + "loss": 1.2497, + "step": 137460 + }, + { + "epoch": 11.9, + "learning_rate": 3.813393398596552e-05, + "loss": 1.1774, + "step": 137470 + }, + { + "epoch": 11.9, + "learning_rate": 3.813306766005371e-05, + "loss": 1.2047, + "step": 137480 + }, + { + "epoch": 11.9, + "learning_rate": 3.8132201334141904e-05, + "loss": 1.124, + "step": 137490 + }, + { + "epoch": 11.9, + "learning_rate": 3.81313350082301e-05, + "loss": 1.1028, + "step": 137500 + }, + { + "epoch": 11.9, + "learning_rate": 3.8130468682318286e-05, + "loss": 1.1826, + "step": 137510 + }, + { + "epoch": 11.9, + "learning_rate": 3.812960235640648e-05, + "loss": 1.2618, + "step": 137520 + }, + { + "epoch": 11.9, + "learning_rate": 3.8128736030494675e-05, + "loss": 1.1745, + "step": 137530 + }, + { + "epoch": 11.9, + "learning_rate": 3.812786970458286e-05, + "loss": 1.1403, + "step": 137540 + }, + { + "epoch": 11.91, + "learning_rate": 3.812700337867106e-05, + "loss": 1.1321, + "step": 137550 + }, + { + "epoch": 11.91, + "learning_rate": 3.8126137052759245e-05, + "loss": 1.2283, + "step": 137560 + }, + { + "epoch": 11.91, + "learning_rate": 3.812527072684744e-05, + "loss": 1.146, + "step": 137570 + }, + { + "epoch": 11.91, + "learning_rate": 3.8124404400935634e-05, + "loss": 1.1057, + "step": 137580 + }, + { + "epoch": 11.91, + "learning_rate": 3.812353807502382e-05, + "loss": 1.1913, + "step": 137590 + }, + { + "epoch": 11.91, + "learning_rate": 3.8122671749112016e-05, + "loss": 1.1942, + "step": 137600 + }, + { + "epoch": 11.91, + "learning_rate": 3.812180542320021e-05, + "loss": 1.153, + "step": 137610 + }, + { + "epoch": 11.91, + "learning_rate": 3.81209390972884e-05, + "loss": 1.1322, + "step": 137620 + }, + { + "epoch": 11.91, + "learning_rate": 3.812007277137659e-05, + "loss": 1.1933, + "step": 137630 + }, + { + "epoch": 11.91, + "learning_rate": 3.811920644546478e-05, + "loss": 1.1741, + "step": 137640 + }, + { + "epoch": 11.91, + "learning_rate": 3.8118340119552975e-05, + "loss": 1.1265, + "step": 137650 + }, + { + "epoch": 11.92, + "learning_rate": 3.811747379364117e-05, + "loss": 1.1837, + "step": 137660 + }, + { + "epoch": 11.92, + "learning_rate": 3.811660746772936e-05, + "loss": 1.1324, + "step": 137670 + }, + { + "epoch": 11.92, + "learning_rate": 3.811574114181755e-05, + "loss": 1.2123, + "step": 137680 + }, + { + "epoch": 11.92, + "learning_rate": 3.8114874815905746e-05, + "loss": 1.1836, + "step": 137690 + }, + { + "epoch": 11.92, + "learning_rate": 3.811400848999393e-05, + "loss": 1.2015, + "step": 137700 + }, + { + "epoch": 11.92, + "learning_rate": 3.811314216408213e-05, + "loss": 1.2087, + "step": 137710 + }, + { + "epoch": 11.92, + "learning_rate": 3.811227583817032e-05, + "loss": 1.144, + "step": 137720 + }, + { + "epoch": 11.92, + "learning_rate": 3.811140951225851e-05, + "loss": 1.1282, + "step": 137730 + }, + { + "epoch": 11.92, + "learning_rate": 3.8110543186346704e-05, + "loss": 1.1871, + "step": 137740 + }, + { + "epoch": 11.92, + "learning_rate": 3.810967686043489e-05, + "loss": 1.1629, + "step": 137750 + }, + { + "epoch": 11.92, + "learning_rate": 3.8108810534523086e-05, + "loss": 1.1429, + "step": 137760 + }, + { + "epoch": 11.92, + "learning_rate": 3.810794420861128e-05, + "loss": 1.1538, + "step": 137770 + }, + { + "epoch": 11.93, + "learning_rate": 3.810707788269947e-05, + "loss": 1.177, + "step": 137780 + }, + { + "epoch": 11.93, + "learning_rate": 3.810621155678766e-05, + "loss": 1.1426, + "step": 137790 + }, + { + "epoch": 11.93, + "learning_rate": 3.810534523087586e-05, + "loss": 1.1686, + "step": 137800 + }, + { + "epoch": 11.93, + "learning_rate": 3.8104478904964045e-05, + "loss": 1.2018, + "step": 137810 + }, + { + "epoch": 11.93, + "learning_rate": 3.810361257905224e-05, + "loss": 1.1809, + "step": 137820 + }, + { + "epoch": 11.93, + "learning_rate": 3.8102746253140434e-05, + "loss": 1.1328, + "step": 137830 + }, + { + "epoch": 11.93, + "learning_rate": 3.810187992722862e-05, + "loss": 1.251, + "step": 137840 + }, + { + "epoch": 11.93, + "learning_rate": 3.8101013601316816e-05, + "loss": 1.1914, + "step": 137850 + }, + { + "epoch": 11.93, + "learning_rate": 3.8100147275405004e-05, + "loss": 1.1692, + "step": 137860 + }, + { + "epoch": 11.93, + "learning_rate": 3.80992809494932e-05, + "loss": 1.1903, + "step": 137870 + }, + { + "epoch": 11.93, + "learning_rate": 3.809841462358139e-05, + "loss": 1.1254, + "step": 137880 + }, + { + "epoch": 11.94, + "learning_rate": 3.809754829766958e-05, + "loss": 1.1729, + "step": 137890 + }, + { + "epoch": 11.94, + "learning_rate": 3.8096681971757775e-05, + "loss": 1.1277, + "step": 137900 + }, + { + "epoch": 11.94, + "learning_rate": 3.809581564584597e-05, + "loss": 1.1571, + "step": 137910 + }, + { + "epoch": 11.94, + "learning_rate": 3.809494931993416e-05, + "loss": 1.1645, + "step": 137920 + }, + { + "epoch": 11.94, + "learning_rate": 3.809408299402235e-05, + "loss": 1.1425, + "step": 137930 + }, + { + "epoch": 11.94, + "learning_rate": 3.8093216668110546e-05, + "loss": 1.1009, + "step": 137940 + }, + { + "epoch": 11.94, + "learning_rate": 3.8092350342198734e-05, + "loss": 1.1845, + "step": 137950 + }, + { + "epoch": 11.94, + "learning_rate": 3.809148401628693e-05, + "loss": 1.179, + "step": 137960 + }, + { + "epoch": 11.94, + "learning_rate": 3.8090617690375116e-05, + "loss": 1.1562, + "step": 137970 + }, + { + "epoch": 11.94, + "learning_rate": 3.808975136446331e-05, + "loss": 1.2276, + "step": 137980 + }, + { + "epoch": 11.94, + "learning_rate": 3.8088885038551505e-05, + "loss": 1.1434, + "step": 137990 + }, + { + "epoch": 11.94, + "learning_rate": 3.808801871263969e-05, + "loss": 1.1609, + "step": 138000 + }, + { + "epoch": 11.95, + "learning_rate": 3.808715238672789e-05, + "loss": 1.1536, + "step": 138010 + }, + { + "epoch": 11.95, + "learning_rate": 3.808628606081608e-05, + "loss": 1.1074, + "step": 138020 + }, + { + "epoch": 11.95, + "learning_rate": 3.808541973490427e-05, + "loss": 1.1336, + "step": 138030 + }, + { + "epoch": 11.95, + "learning_rate": 3.808455340899246e-05, + "loss": 1.1936, + "step": 138040 + }, + { + "epoch": 11.95, + "learning_rate": 3.808368708308066e-05, + "loss": 1.096, + "step": 138050 + }, + { + "epoch": 11.95, + "learning_rate": 3.8082820757168846e-05, + "loss": 1.2175, + "step": 138060 + }, + { + "epoch": 11.95, + "learning_rate": 3.808195443125704e-05, + "loss": 1.1885, + "step": 138070 + }, + { + "epoch": 11.95, + "learning_rate": 3.808108810534523e-05, + "loss": 1.181, + "step": 138080 + }, + { + "epoch": 11.95, + "learning_rate": 3.808022177943342e-05, + "loss": 1.2139, + "step": 138090 + }, + { + "epoch": 11.95, + "learning_rate": 3.8079355453521617e-05, + "loss": 1.1405, + "step": 138100 + }, + { + "epoch": 11.95, + "learning_rate": 3.8078489127609804e-05, + "loss": 1.1697, + "step": 138110 + }, + { + "epoch": 11.96, + "learning_rate": 3.8077622801698e-05, + "loss": 1.2236, + "step": 138120 + }, + { + "epoch": 11.96, + "learning_rate": 3.807675647578619e-05, + "loss": 1.0969, + "step": 138130 + }, + { + "epoch": 11.96, + "learning_rate": 3.807589014987438e-05, + "loss": 1.181, + "step": 138140 + }, + { + "epoch": 11.96, + "learning_rate": 3.8075023823962575e-05, + "loss": 1.1804, + "step": 138150 + }, + { + "epoch": 11.96, + "learning_rate": 3.807415749805077e-05, + "loss": 1.2031, + "step": 138160 + }, + { + "epoch": 11.96, + "learning_rate": 3.807329117213896e-05, + "loss": 1.1634, + "step": 138170 + }, + { + "epoch": 11.96, + "learning_rate": 3.807242484622715e-05, + "loss": 1.2165, + "step": 138180 + }, + { + "epoch": 11.96, + "learning_rate": 3.807155852031534e-05, + "loss": 1.185, + "step": 138190 + }, + { + "epoch": 11.96, + "learning_rate": 3.8070692194403534e-05, + "loss": 1.1984, + "step": 138200 + }, + { + "epoch": 11.96, + "learning_rate": 3.806982586849173e-05, + "loss": 1.11, + "step": 138210 + }, + { + "epoch": 11.96, + "learning_rate": 3.8068959542579916e-05, + "loss": 1.2236, + "step": 138220 + }, + { + "epoch": 11.96, + "learning_rate": 3.806809321666811e-05, + "loss": 1.1503, + "step": 138230 + }, + { + "epoch": 11.97, + "learning_rate": 3.8067226890756305e-05, + "loss": 1.1652, + "step": 138240 + }, + { + "epoch": 11.97, + "learning_rate": 3.806636056484449e-05, + "loss": 1.1818, + "step": 138250 + }, + { + "epoch": 11.97, + "learning_rate": 3.806549423893269e-05, + "loss": 1.1303, + "step": 138260 + }, + { + "epoch": 11.97, + "learning_rate": 3.806462791302088e-05, + "loss": 1.1427, + "step": 138270 + }, + { + "epoch": 11.97, + "learning_rate": 3.806376158710907e-05, + "loss": 1.1649, + "step": 138280 + }, + { + "epoch": 11.97, + "learning_rate": 3.8062895261197264e-05, + "loss": 1.1147, + "step": 138290 + }, + { + "epoch": 11.97, + "learning_rate": 3.806202893528545e-05, + "loss": 1.1996, + "step": 138300 + }, + { + "epoch": 11.97, + "learning_rate": 3.8061162609373646e-05, + "loss": 1.1989, + "step": 138310 + }, + { + "epoch": 11.97, + "learning_rate": 3.806029628346184e-05, + "loss": 1.1553, + "step": 138320 + }, + { + "epoch": 11.97, + "learning_rate": 3.805942995755003e-05, + "loss": 1.1495, + "step": 138330 + }, + { + "epoch": 11.97, + "learning_rate": 3.805856363163822e-05, + "loss": 1.1229, + "step": 138340 + }, + { + "epoch": 11.97, + "learning_rate": 3.805769730572642e-05, + "loss": 1.1523, + "step": 138350 + }, + { + "epoch": 11.98, + "learning_rate": 3.8056830979814605e-05, + "loss": 1.1196, + "step": 138360 + }, + { + "epoch": 11.98, + "learning_rate": 3.80559646539028e-05, + "loss": 1.1484, + "step": 138370 + }, + { + "epoch": 11.98, + "learning_rate": 3.805509832799099e-05, + "loss": 1.1734, + "step": 138380 + }, + { + "epoch": 11.98, + "learning_rate": 3.805423200207918e-05, + "loss": 1.1384, + "step": 138390 + }, + { + "epoch": 11.98, + "learning_rate": 3.8053365676167376e-05, + "loss": 1.1606, + "step": 138400 + }, + { + "epoch": 11.98, + "learning_rate": 3.805249935025556e-05, + "loss": 1.0839, + "step": 138410 + }, + { + "epoch": 11.98, + "learning_rate": 3.805163302434376e-05, + "loss": 1.2148, + "step": 138420 + }, + { + "epoch": 11.98, + "learning_rate": 3.805076669843195e-05, + "loss": 1.1558, + "step": 138430 + }, + { + "epoch": 11.98, + "learning_rate": 3.804990037252014e-05, + "loss": 1.1372, + "step": 138440 + }, + { + "epoch": 11.98, + "learning_rate": 3.8049034046608334e-05, + "loss": 1.2012, + "step": 138450 + }, + { + "epoch": 11.98, + "learning_rate": 3.804816772069653e-05, + "loss": 1.1297, + "step": 138460 + }, + { + "epoch": 11.99, + "learning_rate": 3.8047301394784716e-05, + "loss": 1.1381, + "step": 138470 + }, + { + "epoch": 11.99, + "learning_rate": 3.804643506887291e-05, + "loss": 1.189, + "step": 138480 + }, + { + "epoch": 11.99, + "learning_rate": 3.80455687429611e-05, + "loss": 1.1472, + "step": 138490 + }, + { + "epoch": 11.99, + "learning_rate": 3.804470241704929e-05, + "loss": 1.1402, + "step": 138500 + }, + { + "epoch": 11.99, + "learning_rate": 3.804383609113749e-05, + "loss": 1.1914, + "step": 138510 + }, + { + "epoch": 11.99, + "learning_rate": 3.8042969765225675e-05, + "loss": 1.1596, + "step": 138520 + }, + { + "epoch": 11.99, + "learning_rate": 3.804210343931387e-05, + "loss": 1.1449, + "step": 138530 + }, + { + "epoch": 11.99, + "learning_rate": 3.8041237113402064e-05, + "loss": 1.1508, + "step": 138540 + }, + { + "epoch": 11.99, + "learning_rate": 3.804037078749025e-05, + "loss": 1.1605, + "step": 138550 + }, + { + "epoch": 11.99, + "learning_rate": 3.8039504461578446e-05, + "loss": 1.1069, + "step": 138560 + }, + { + "epoch": 11.99, + "learning_rate": 3.803863813566664e-05, + "loss": 1.1449, + "step": 138570 + }, + { + "epoch": 11.99, + "learning_rate": 3.803777180975483e-05, + "loss": 1.088, + "step": 138580 + }, + { + "epoch": 12.0, + "learning_rate": 3.803690548384302e-05, + "loss": 1.2025, + "step": 138590 + }, + { + "epoch": 12.0, + "learning_rate": 3.803603915793121e-05, + "loss": 1.1918, + "step": 138600 + }, + { + "epoch": 12.0, + "learning_rate": 3.8035172832019405e-05, + "loss": 1.129, + "step": 138610 + }, + { + "epoch": 12.0, + "learning_rate": 3.80343065061076e-05, + "loss": 1.111, + "step": 138620 + }, + { + "epoch": 12.0, + "learning_rate": 3.803344018019579e-05, + "loss": 1.199, + "step": 138630 + }, + { + "epoch": 12.0, + "eval_Bleu_1": 0.040084530853760944, + "eval_Bleu_2": 2.724151222006443e-11, + "eval_Bleu_3": 2.475222688997043e-14, + "eval_Bleu_4": 7.667546626480337e-16, + "eval_ROUGE_L": 0.08613809982070882, + "eval_cer": 0.9936453084591387, + "eval_em": 0, + "eval_f1": 0.09067596924330444, + "eval_loss": 1.0848437547683716, + "eval_runtime": 2108.4814, + "eval_samples_per_second": 2.435, + "eval_steps_per_second": 2.435, + "eval_wer": 0.9735824957645084, + "step": 138639 + }, + { + "epoch": 12.0, + "learning_rate": 3.803257385428398e-05, + "loss": 1.1381, + "step": 138640 + }, + { + "epoch": 12.0, + "learning_rate": 3.8031707528372176e-05, + "loss": 1.1236, + "step": 138650 + }, + { + "epoch": 12.0, + "learning_rate": 3.8030841202460364e-05, + "loss": 1.1234, + "step": 138660 + }, + { + "epoch": 12.0, + "learning_rate": 3.802997487654856e-05, + "loss": 1.1551, + "step": 138670 + }, + { + "epoch": 12.0, + "learning_rate": 3.802910855063675e-05, + "loss": 1.1253, + "step": 138680 + }, + { + "epoch": 12.0, + "learning_rate": 3.802824222472494e-05, + "loss": 1.1771, + "step": 138690 + }, + { + "epoch": 12.01, + "learning_rate": 3.8027375898813135e-05, + "loss": 1.1494, + "step": 138700 + }, + { + "epoch": 12.01, + "learning_rate": 3.802650957290132e-05, + "loss": 1.1605, + "step": 138710 + }, + { + "epoch": 12.01, + "learning_rate": 3.802564324698952e-05, + "loss": 1.1642, + "step": 138720 + }, + { + "epoch": 12.01, + "learning_rate": 3.802477692107771e-05, + "loss": 1.1235, + "step": 138730 + }, + { + "epoch": 12.01, + "learning_rate": 3.80239105951659e-05, + "loss": 1.1222, + "step": 138740 + }, + { + "epoch": 12.01, + "learning_rate": 3.802304426925409e-05, + "loss": 1.0874, + "step": 138750 + }, + { + "epoch": 12.01, + "learning_rate": 3.802217794334229e-05, + "loss": 1.1014, + "step": 138760 + }, + { + "epoch": 12.01, + "learning_rate": 3.8021311617430475e-05, + "loss": 1.1128, + "step": 138770 + }, + { + "epoch": 12.01, + "learning_rate": 3.802044529151867e-05, + "loss": 1.1314, + "step": 138780 + }, + { + "epoch": 12.01, + "learning_rate": 3.8019578965606864e-05, + "loss": 1.1561, + "step": 138790 + }, + { + "epoch": 12.01, + "learning_rate": 3.801871263969505e-05, + "loss": 1.1306, + "step": 138800 + }, + { + "epoch": 12.01, + "learning_rate": 3.8017846313783247e-05, + "loss": 1.1563, + "step": 138810 + }, + { + "epoch": 12.02, + "learning_rate": 3.8016979987871434e-05, + "loss": 1.1128, + "step": 138820 + }, + { + "epoch": 12.02, + "learning_rate": 3.801611366195963e-05, + "loss": 1.1245, + "step": 138830 + }, + { + "epoch": 12.02, + "learning_rate": 3.801524733604782e-05, + "loss": 1.1298, + "step": 138840 + }, + { + "epoch": 12.02, + "learning_rate": 3.801438101013601e-05, + "loss": 1.1161, + "step": 138850 + }, + { + "epoch": 12.02, + "learning_rate": 3.8013514684224205e-05, + "loss": 1.1625, + "step": 138860 + }, + { + "epoch": 12.02, + "learning_rate": 3.80126483583124e-05, + "loss": 1.1938, + "step": 138870 + }, + { + "epoch": 12.02, + "learning_rate": 3.801178203240059e-05, + "loss": 1.0942, + "step": 138880 + }, + { + "epoch": 12.02, + "learning_rate": 3.801091570648878e-05, + "loss": 1.1148, + "step": 138890 + }, + { + "epoch": 12.02, + "learning_rate": 3.8010049380576976e-05, + "loss": 1.0932, + "step": 138900 + }, + { + "epoch": 12.02, + "learning_rate": 3.8009183054665164e-05, + "loss": 1.0872, + "step": 138910 + }, + { + "epoch": 12.02, + "learning_rate": 3.800831672875336e-05, + "loss": 1.1376, + "step": 138920 + }, + { + "epoch": 12.03, + "learning_rate": 3.8007450402841546e-05, + "loss": 1.1403, + "step": 138930 + }, + { + "epoch": 12.03, + "learning_rate": 3.800658407692974e-05, + "loss": 1.2111, + "step": 138940 + }, + { + "epoch": 12.03, + "learning_rate": 3.8005717751017935e-05, + "loss": 1.1366, + "step": 138950 + }, + { + "epoch": 12.03, + "learning_rate": 3.800485142510612e-05, + "loss": 1.0942, + "step": 138960 + }, + { + "epoch": 12.03, + "learning_rate": 3.800398509919432e-05, + "loss": 1.153, + "step": 138970 + }, + { + "epoch": 12.03, + "learning_rate": 3.800311877328251e-05, + "loss": 1.1453, + "step": 138980 + }, + { + "epoch": 12.03, + "learning_rate": 3.80022524473707e-05, + "loss": 1.2232, + "step": 138990 + }, + { + "epoch": 12.03, + "learning_rate": 3.8001386121458894e-05, + "loss": 1.1199, + "step": 139000 + }, + { + "epoch": 12.03, + "learning_rate": 3.800051979554709e-05, + "loss": 1.0884, + "step": 139010 + }, + { + "epoch": 12.03, + "learning_rate": 3.7999653469635276e-05, + "loss": 1.0735, + "step": 139020 + }, + { + "epoch": 12.03, + "learning_rate": 3.799878714372347e-05, + "loss": 1.1461, + "step": 139030 + }, + { + "epoch": 12.03, + "learning_rate": 3.799792081781166e-05, + "loss": 1.1328, + "step": 139040 + }, + { + "epoch": 12.04, + "learning_rate": 3.799705449189985e-05, + "loss": 1.1201, + "step": 139050 + }, + { + "epoch": 12.04, + "learning_rate": 3.799618816598805e-05, + "loss": 1.1037, + "step": 139060 + }, + { + "epoch": 12.04, + "learning_rate": 3.7995321840076235e-05, + "loss": 1.134, + "step": 139070 + }, + { + "epoch": 12.04, + "learning_rate": 3.799445551416443e-05, + "loss": 1.1242, + "step": 139080 + }, + { + "epoch": 12.04, + "learning_rate": 3.7993589188252623e-05, + "loss": 1.1706, + "step": 139090 + }, + { + "epoch": 12.04, + "learning_rate": 3.799272286234081e-05, + "loss": 1.1677, + "step": 139100 + }, + { + "epoch": 12.04, + "learning_rate": 3.7991856536429006e-05, + "loss": 1.1533, + "step": 139110 + }, + { + "epoch": 12.04, + "learning_rate": 3.79909902105172e-05, + "loss": 1.1324, + "step": 139120 + }, + { + "epoch": 12.04, + "learning_rate": 3.799012388460539e-05, + "loss": 1.1624, + "step": 139130 + }, + { + "epoch": 12.04, + "learning_rate": 3.798925755869358e-05, + "loss": 1.1599, + "step": 139140 + }, + { + "epoch": 12.04, + "learning_rate": 3.798839123278177e-05, + "loss": 1.1226, + "step": 139150 + }, + { + "epoch": 12.05, + "learning_rate": 3.7987524906869964e-05, + "loss": 1.096, + "step": 139160 + }, + { + "epoch": 12.05, + "learning_rate": 3.798665858095816e-05, + "loss": 1.1387, + "step": 139170 + }, + { + "epoch": 12.05, + "learning_rate": 3.7985792255046346e-05, + "loss": 1.1626, + "step": 139180 + }, + { + "epoch": 12.05, + "learning_rate": 3.798492592913454e-05, + "loss": 1.1137, + "step": 139190 + }, + { + "epoch": 12.05, + "learning_rate": 3.7984059603222735e-05, + "loss": 1.1463, + "step": 139200 + }, + { + "epoch": 12.05, + "learning_rate": 3.798319327731092e-05, + "loss": 1.1523, + "step": 139210 + }, + { + "epoch": 12.05, + "learning_rate": 3.798232695139912e-05, + "loss": 1.1292, + "step": 139220 + }, + { + "epoch": 12.05, + "learning_rate": 3.7981460625487305e-05, + "loss": 1.1358, + "step": 139230 + }, + { + "epoch": 12.05, + "learning_rate": 3.79805942995755e-05, + "loss": 1.1719, + "step": 139240 + }, + { + "epoch": 12.05, + "learning_rate": 3.7979727973663694e-05, + "loss": 1.154, + "step": 139250 + }, + { + "epoch": 12.05, + "learning_rate": 3.797886164775188e-05, + "loss": 1.1401, + "step": 139260 + }, + { + "epoch": 12.05, + "learning_rate": 3.7977995321840076e-05, + "loss": 1.1681, + "step": 139270 + }, + { + "epoch": 12.06, + "learning_rate": 3.797712899592827e-05, + "loss": 1.1231, + "step": 139280 + }, + { + "epoch": 12.06, + "learning_rate": 3.797626267001646e-05, + "loss": 1.0589, + "step": 139290 + }, + { + "epoch": 12.06, + "learning_rate": 3.797539634410465e-05, + "loss": 1.1054, + "step": 139300 + }, + { + "epoch": 12.06, + "learning_rate": 3.797453001819285e-05, + "loss": 1.0708, + "step": 139310 + }, + { + "epoch": 12.06, + "learning_rate": 3.7973663692281035e-05, + "loss": 1.1865, + "step": 139320 + }, + { + "epoch": 12.06, + "learning_rate": 3.797279736636923e-05, + "loss": 1.1516, + "step": 139330 + }, + { + "epoch": 12.06, + "learning_rate": 3.797193104045742e-05, + "loss": 1.1148, + "step": 139340 + }, + { + "epoch": 12.06, + "learning_rate": 3.797106471454561e-05, + "loss": 1.093, + "step": 139350 + }, + { + "epoch": 12.06, + "learning_rate": 3.7970198388633806e-05, + "loss": 1.1751, + "step": 139360 + }, + { + "epoch": 12.06, + "learning_rate": 3.7969332062721994e-05, + "loss": 1.1393, + "step": 139370 + }, + { + "epoch": 12.06, + "learning_rate": 3.796846573681019e-05, + "loss": 1.1614, + "step": 139380 + }, + { + "epoch": 12.07, + "learning_rate": 3.796759941089838e-05, + "loss": 1.1355, + "step": 139390 + }, + { + "epoch": 12.07, + "learning_rate": 3.796673308498657e-05, + "loss": 1.171, + "step": 139400 + }, + { + "epoch": 12.07, + "learning_rate": 3.7965866759074765e-05, + "loss": 1.1235, + "step": 139410 + }, + { + "epoch": 12.07, + "learning_rate": 3.796500043316296e-05, + "loss": 1.1661, + "step": 139420 + }, + { + "epoch": 12.07, + "learning_rate": 3.796413410725115e-05, + "loss": 1.122, + "step": 139430 + }, + { + "epoch": 12.07, + "learning_rate": 3.796326778133934e-05, + "loss": 1.1573, + "step": 139440 + }, + { + "epoch": 12.07, + "learning_rate": 3.796240145542753e-05, + "loss": 1.1288, + "step": 139450 + }, + { + "epoch": 12.07, + "learning_rate": 3.796153512951572e-05, + "loss": 1.0998, + "step": 139460 + }, + { + "epoch": 12.07, + "learning_rate": 3.796066880360392e-05, + "loss": 1.145, + "step": 139470 + }, + { + "epoch": 12.07, + "learning_rate": 3.7959802477692105e-05, + "loss": 1.1416, + "step": 139480 + }, + { + "epoch": 12.07, + "learning_rate": 3.79589361517803e-05, + "loss": 1.1092, + "step": 139490 + }, + { + "epoch": 12.07, + "learning_rate": 3.7958069825868494e-05, + "loss": 1.1523, + "step": 139500 + }, + { + "epoch": 12.08, + "learning_rate": 3.795720349995668e-05, + "loss": 1.1462, + "step": 139510 + }, + { + "epoch": 12.08, + "learning_rate": 3.7956337174044876e-05, + "loss": 1.1242, + "step": 139520 + }, + { + "epoch": 12.08, + "learning_rate": 3.795547084813307e-05, + "loss": 1.1844, + "step": 139530 + }, + { + "epoch": 12.08, + "learning_rate": 3.795460452222126e-05, + "loss": 1.1303, + "step": 139540 + }, + { + "epoch": 12.08, + "learning_rate": 3.795373819630945e-05, + "loss": 1.1335, + "step": 139550 + }, + { + "epoch": 12.08, + "learning_rate": 3.795287187039764e-05, + "loss": 1.1528, + "step": 139560 + }, + { + "epoch": 12.08, + "learning_rate": 3.7952005544485835e-05, + "loss": 1.1188, + "step": 139570 + }, + { + "epoch": 12.08, + "learning_rate": 3.795113921857403e-05, + "loss": 1.1487, + "step": 139580 + }, + { + "epoch": 12.08, + "learning_rate": 3.795027289266222e-05, + "loss": 1.1575, + "step": 139590 + }, + { + "epoch": 12.08, + "learning_rate": 3.794940656675041e-05, + "loss": 1.1457, + "step": 139600 + }, + { + "epoch": 12.08, + "learning_rate": 3.7948540240838606e-05, + "loss": 1.1793, + "step": 139610 + }, + { + "epoch": 12.08, + "learning_rate": 3.7947673914926794e-05, + "loss": 1.1601, + "step": 139620 + }, + { + "epoch": 12.09, + "learning_rate": 3.794680758901499e-05, + "loss": 1.1132, + "step": 139630 + }, + { + "epoch": 12.09, + "learning_rate": 3.794594126310318e-05, + "loss": 1.1049, + "step": 139640 + }, + { + "epoch": 12.09, + "learning_rate": 3.794507493719137e-05, + "loss": 1.1449, + "step": 139650 + }, + { + "epoch": 12.09, + "learning_rate": 3.7944208611279565e-05, + "loss": 1.1728, + "step": 139660 + }, + { + "epoch": 12.09, + "learning_rate": 3.794334228536775e-05, + "loss": 1.1318, + "step": 139670 + }, + { + "epoch": 12.09, + "learning_rate": 3.794247595945595e-05, + "loss": 1.1588, + "step": 139680 + }, + { + "epoch": 12.09, + "learning_rate": 3.794160963354414e-05, + "loss": 1.1347, + "step": 139690 + }, + { + "epoch": 12.09, + "learning_rate": 3.794074330763233e-05, + "loss": 1.0958, + "step": 139700 + }, + { + "epoch": 12.09, + "learning_rate": 3.7939876981720524e-05, + "loss": 1.1448, + "step": 139710 + }, + { + "epoch": 12.09, + "learning_rate": 3.793901065580872e-05, + "loss": 1.1361, + "step": 139720 + }, + { + "epoch": 12.09, + "learning_rate": 3.7938144329896906e-05, + "loss": 1.1056, + "step": 139730 + }, + { + "epoch": 12.1, + "learning_rate": 3.79372780039851e-05, + "loss": 1.1617, + "step": 139740 + }, + { + "epoch": 12.1, + "learning_rate": 3.7936411678073295e-05, + "loss": 1.152, + "step": 139750 + }, + { + "epoch": 12.1, + "learning_rate": 3.793554535216148e-05, + "loss": 1.0851, + "step": 139760 + }, + { + "epoch": 12.1, + "learning_rate": 3.793467902624968e-05, + "loss": 1.0956, + "step": 139770 + }, + { + "epoch": 12.1, + "learning_rate": 3.7933812700337864e-05, + "loss": 1.1137, + "step": 139780 + }, + { + "epoch": 12.1, + "learning_rate": 3.793294637442606e-05, + "loss": 1.1467, + "step": 139790 + }, + { + "epoch": 12.1, + "learning_rate": 3.7932080048514253e-05, + "loss": 1.1364, + "step": 139800 + }, + { + "epoch": 12.1, + "learning_rate": 3.793121372260244e-05, + "loss": 1.1256, + "step": 139810 + }, + { + "epoch": 12.1, + "learning_rate": 3.7930347396690636e-05, + "loss": 1.1702, + "step": 139820 + }, + { + "epoch": 12.1, + "learning_rate": 3.792948107077883e-05, + "loss": 1.1358, + "step": 139830 + }, + { + "epoch": 12.1, + "learning_rate": 3.792861474486702e-05, + "loss": 1.135, + "step": 139840 + }, + { + "epoch": 12.1, + "learning_rate": 3.792774841895521e-05, + "loss": 1.1264, + "step": 139850 + }, + { + "epoch": 12.11, + "learning_rate": 3.7926882093043407e-05, + "loss": 1.1168, + "step": 139860 + }, + { + "epoch": 12.11, + "learning_rate": 3.7926015767131594e-05, + "loss": 1.1457, + "step": 139870 + }, + { + "epoch": 12.11, + "learning_rate": 3.792514944121979e-05, + "loss": 1.1439, + "step": 139880 + }, + { + "epoch": 12.11, + "learning_rate": 3.7924283115307976e-05, + "loss": 1.1302, + "step": 139890 + }, + { + "epoch": 12.11, + "learning_rate": 3.792341678939617e-05, + "loss": 1.1536, + "step": 139900 + }, + { + "epoch": 12.11, + "learning_rate": 3.7922550463484365e-05, + "loss": 1.1089, + "step": 139910 + }, + { + "epoch": 12.11, + "learning_rate": 3.792168413757255e-05, + "loss": 1.1362, + "step": 139920 + }, + { + "epoch": 12.11, + "learning_rate": 3.792081781166075e-05, + "loss": 1.1504, + "step": 139930 + }, + { + "epoch": 12.11, + "learning_rate": 3.791995148574894e-05, + "loss": 1.1493, + "step": 139940 + }, + { + "epoch": 12.11, + "learning_rate": 3.791908515983713e-05, + "loss": 1.0973, + "step": 139950 + }, + { + "epoch": 12.11, + "learning_rate": 3.7918218833925324e-05, + "loss": 1.1416, + "step": 139960 + }, + { + "epoch": 12.12, + "learning_rate": 3.791735250801351e-05, + "loss": 1.116, + "step": 139970 + }, + { + "epoch": 12.12, + "learning_rate": 3.7916486182101706e-05, + "loss": 1.1062, + "step": 139980 + }, + { + "epoch": 12.12, + "learning_rate": 3.79156198561899e-05, + "loss": 1.1436, + "step": 139990 + }, + { + "epoch": 12.12, + "learning_rate": 3.791475353027809e-05, + "loss": 1.1947, + "step": 140000 + }, + { + "epoch": 12.12, + "learning_rate": 3.791388720436628e-05, + "loss": 1.0944, + "step": 140010 + }, + { + "epoch": 12.12, + "learning_rate": 3.791302087845448e-05, + "loss": 1.209, + "step": 140020 + }, + { + "epoch": 12.12, + "learning_rate": 3.7912154552542665e-05, + "loss": 1.1632, + "step": 140030 + }, + { + "epoch": 12.12, + "learning_rate": 3.791128822663086e-05, + "loss": 1.1328, + "step": 140040 + }, + { + "epoch": 12.12, + "learning_rate": 3.7910421900719054e-05, + "loss": 1.0857, + "step": 140050 + }, + { + "epoch": 12.12, + "learning_rate": 3.790955557480724e-05, + "loss": 1.1291, + "step": 140060 + }, + { + "epoch": 12.12, + "learning_rate": 3.7908689248895436e-05, + "loss": 1.0976, + "step": 140070 + }, + { + "epoch": 12.12, + "learning_rate": 3.7907822922983624e-05, + "loss": 1.1624, + "step": 140080 + }, + { + "epoch": 12.13, + "learning_rate": 3.790695659707182e-05, + "loss": 1.0683, + "step": 140090 + }, + { + "epoch": 12.13, + "learning_rate": 3.790609027116001e-05, + "loss": 1.0586, + "step": 140100 + }, + { + "epoch": 12.13, + "learning_rate": 3.79052239452482e-05, + "loss": 1.1998, + "step": 140110 + }, + { + "epoch": 12.13, + "learning_rate": 3.7904357619336395e-05, + "loss": 1.1312, + "step": 140120 + }, + { + "epoch": 12.13, + "learning_rate": 3.790349129342459e-05, + "loss": 1.0894, + "step": 140130 + }, + { + "epoch": 12.13, + "learning_rate": 3.790262496751278e-05, + "loss": 1.1261, + "step": 140140 + }, + { + "epoch": 12.13, + "learning_rate": 3.790175864160097e-05, + "loss": 1.1544, + "step": 140150 + }, + { + "epoch": 12.13, + "learning_rate": 3.7900892315689166e-05, + "loss": 1.149, + "step": 140160 + }, + { + "epoch": 12.13, + "learning_rate": 3.790002598977735e-05, + "loss": 1.107, + "step": 140170 + }, + { + "epoch": 12.13, + "learning_rate": 3.789915966386555e-05, + "loss": 1.1216, + "step": 140180 + }, + { + "epoch": 12.13, + "learning_rate": 3.7898293337953735e-05, + "loss": 1.0824, + "step": 140190 + }, + { + "epoch": 12.14, + "learning_rate": 3.789742701204193e-05, + "loss": 1.1702, + "step": 140200 + }, + { + "epoch": 12.14, + "learning_rate": 3.7896560686130124e-05, + "loss": 1.103, + "step": 140210 + }, + { + "epoch": 12.14, + "learning_rate": 3.789569436021831e-05, + "loss": 1.1605, + "step": 140220 + }, + { + "epoch": 12.14, + "learning_rate": 3.7894828034306506e-05, + "loss": 1.1181, + "step": 140230 + }, + { + "epoch": 12.14, + "learning_rate": 3.78939617083947e-05, + "loss": 1.1638, + "step": 140240 + }, + { + "epoch": 12.14, + "learning_rate": 3.789309538248289e-05, + "loss": 1.1632, + "step": 140250 + }, + { + "epoch": 12.14, + "learning_rate": 3.789222905657108e-05, + "loss": 1.1484, + "step": 140260 + }, + { + "epoch": 12.14, + "learning_rate": 3.789136273065928e-05, + "loss": 1.1081, + "step": 140270 + }, + { + "epoch": 12.14, + "learning_rate": 3.7890496404747465e-05, + "loss": 1.0999, + "step": 140280 + }, + { + "epoch": 12.14, + "learning_rate": 3.788963007883566e-05, + "loss": 1.1296, + "step": 140290 + }, + { + "epoch": 12.14, + "learning_rate": 3.788876375292385e-05, + "loss": 1.0942, + "step": 140300 + }, + { + "epoch": 12.14, + "learning_rate": 3.788789742701204e-05, + "loss": 1.1019, + "step": 140310 + }, + { + "epoch": 12.15, + "learning_rate": 3.7887031101100236e-05, + "loss": 1.0727, + "step": 140320 + }, + { + "epoch": 12.15, + "learning_rate": 3.7886164775188424e-05, + "loss": 1.1068, + "step": 140330 + }, + { + "epoch": 12.15, + "learning_rate": 3.788529844927662e-05, + "loss": 1.1729, + "step": 140340 + }, + { + "epoch": 12.15, + "learning_rate": 3.788443212336481e-05, + "loss": 1.1435, + "step": 140350 + }, + { + "epoch": 12.15, + "learning_rate": 3.7883565797453e-05, + "loss": 1.1372, + "step": 140360 + }, + { + "epoch": 12.15, + "learning_rate": 3.7882699471541195e-05, + "loss": 1.1802, + "step": 140370 + }, + { + "epoch": 12.15, + "learning_rate": 3.788183314562939e-05, + "loss": 1.1581, + "step": 140380 + }, + { + "epoch": 12.15, + "learning_rate": 3.788096681971758e-05, + "loss": 1.1049, + "step": 140390 + }, + { + "epoch": 12.15, + "learning_rate": 3.788010049380577e-05, + "loss": 1.1932, + "step": 140400 + }, + { + "epoch": 12.15, + "learning_rate": 3.787923416789396e-05, + "loss": 1.1309, + "step": 140410 + }, + { + "epoch": 12.15, + "learning_rate": 3.7878367841982154e-05, + "loss": 1.0978, + "step": 140420 + }, + { + "epoch": 12.16, + "learning_rate": 3.787750151607035e-05, + "loss": 1.1553, + "step": 140430 + }, + { + "epoch": 12.16, + "learning_rate": 3.7876635190158536e-05, + "loss": 1.1835, + "step": 140440 + }, + { + "epoch": 12.16, + "learning_rate": 3.787576886424673e-05, + "loss": 1.148, + "step": 140450 + }, + { + "epoch": 12.16, + "learning_rate": 3.7874902538334925e-05, + "loss": 1.1734, + "step": 140460 + }, + { + "epoch": 12.16, + "learning_rate": 3.787403621242311e-05, + "loss": 1.1301, + "step": 140470 + }, + { + "epoch": 12.16, + "learning_rate": 3.787316988651131e-05, + "loss": 1.1262, + "step": 140480 + }, + { + "epoch": 12.16, + "learning_rate": 3.78723035605995e-05, + "loss": 1.0966, + "step": 140490 + }, + { + "epoch": 12.16, + "learning_rate": 3.787143723468769e-05, + "loss": 1.14, + "step": 140500 + }, + { + "epoch": 12.16, + "learning_rate": 3.787057090877588e-05, + "loss": 1.165, + "step": 140510 + }, + { + "epoch": 12.16, + "learning_rate": 3.786970458286407e-05, + "loss": 1.181, + "step": 140520 + }, + { + "epoch": 12.16, + "learning_rate": 3.7868838256952265e-05, + "loss": 1.1694, + "step": 140530 + }, + { + "epoch": 12.16, + "learning_rate": 3.786797193104046e-05, + "loss": 1.1225, + "step": 140540 + }, + { + "epoch": 12.17, + "learning_rate": 3.786710560512865e-05, + "loss": 1.1661, + "step": 140550 + }, + { + "epoch": 12.17, + "learning_rate": 3.786623927921684e-05, + "loss": 1.1724, + "step": 140560 + }, + { + "epoch": 12.17, + "learning_rate": 3.7865372953305037e-05, + "loss": 1.0977, + "step": 140570 + }, + { + "epoch": 12.17, + "learning_rate": 3.7864506627393224e-05, + "loss": 1.1129, + "step": 140580 + }, + { + "epoch": 12.17, + "learning_rate": 3.786364030148142e-05, + "loss": 1.1513, + "step": 140590 + }, + { + "epoch": 12.17, + "learning_rate": 3.786277397556961e-05, + "loss": 1.1053, + "step": 140600 + }, + { + "epoch": 12.17, + "learning_rate": 3.78619076496578e-05, + "loss": 1.103, + "step": 140610 + }, + { + "epoch": 12.17, + "learning_rate": 3.7861041323745995e-05, + "loss": 1.1385, + "step": 140620 + }, + { + "epoch": 12.17, + "learning_rate": 3.786017499783418e-05, + "loss": 1.2532, + "step": 140630 + }, + { + "epoch": 12.17, + "learning_rate": 3.785930867192238e-05, + "loss": 1.1225, + "step": 140640 + }, + { + "epoch": 12.17, + "learning_rate": 3.785844234601057e-05, + "loss": 1.1847, + "step": 140650 + }, + { + "epoch": 12.17, + "learning_rate": 3.785757602009876e-05, + "loss": 1.1238, + "step": 140660 + }, + { + "epoch": 12.18, + "learning_rate": 3.7856709694186954e-05, + "loss": 1.1248, + "step": 140670 + }, + { + "epoch": 12.18, + "learning_rate": 3.785584336827515e-05, + "loss": 1.1686, + "step": 140680 + }, + { + "epoch": 12.18, + "learning_rate": 3.7854977042363336e-05, + "loss": 1.1492, + "step": 140690 + }, + { + "epoch": 12.18, + "learning_rate": 3.785411071645153e-05, + "loss": 1.1046, + "step": 140700 + }, + { + "epoch": 12.18, + "learning_rate": 3.785324439053972e-05, + "loss": 1.1581, + "step": 140710 + }, + { + "epoch": 12.18, + "learning_rate": 3.785237806462791e-05, + "loss": 1.1722, + "step": 140720 + }, + { + "epoch": 12.18, + "learning_rate": 3.785151173871611e-05, + "loss": 1.1003, + "step": 140730 + }, + { + "epoch": 12.18, + "learning_rate": 3.7850645412804295e-05, + "loss": 1.1931, + "step": 140740 + }, + { + "epoch": 12.18, + "learning_rate": 3.784977908689249e-05, + "loss": 1.1937, + "step": 140750 + }, + { + "epoch": 12.18, + "learning_rate": 3.7848912760980684e-05, + "loss": 1.1344, + "step": 140760 + }, + { + "epoch": 12.18, + "learning_rate": 3.784804643506887e-05, + "loss": 1.0985, + "step": 140770 + }, + { + "epoch": 12.19, + "learning_rate": 3.7847180109157066e-05, + "loss": 1.197, + "step": 140780 + }, + { + "epoch": 12.19, + "learning_rate": 3.784631378324526e-05, + "loss": 1.173, + "step": 140790 + }, + { + "epoch": 12.19, + "learning_rate": 3.784544745733345e-05, + "loss": 1.1146, + "step": 140800 + }, + { + "epoch": 12.19, + "learning_rate": 3.784458113142164e-05, + "loss": 1.0892, + "step": 140810 + }, + { + "epoch": 12.19, + "learning_rate": 3.784371480550983e-05, + "loss": 1.0953, + "step": 140820 + }, + { + "epoch": 12.19, + "learning_rate": 3.7842848479598025e-05, + "loss": 1.0895, + "step": 140830 + }, + { + "epoch": 12.19, + "learning_rate": 3.784198215368622e-05, + "loss": 1.1209, + "step": 140840 + }, + { + "epoch": 12.19, + "learning_rate": 3.784111582777441e-05, + "loss": 1.0974, + "step": 140850 + }, + { + "epoch": 12.19, + "learning_rate": 3.78402495018626e-05, + "loss": 1.1476, + "step": 140860 + }, + { + "epoch": 12.19, + "learning_rate": 3.7839383175950796e-05, + "loss": 1.1261, + "step": 140870 + }, + { + "epoch": 12.19, + "learning_rate": 3.783851685003898e-05, + "loss": 1.0954, + "step": 140880 + }, + { + "epoch": 12.19, + "learning_rate": 3.783765052412718e-05, + "loss": 1.194, + "step": 140890 + }, + { + "epoch": 12.2, + "learning_rate": 3.783678419821537e-05, + "loss": 1.1572, + "step": 140900 + }, + { + "epoch": 12.2, + "learning_rate": 3.783591787230356e-05, + "loss": 1.1313, + "step": 140910 + }, + { + "epoch": 12.2, + "learning_rate": 3.7835051546391754e-05, + "loss": 1.1373, + "step": 140920 + }, + { + "epoch": 12.2, + "learning_rate": 3.783418522047994e-05, + "loss": 1.1064, + "step": 140930 + }, + { + "epoch": 12.2, + "learning_rate": 3.7833318894568136e-05, + "loss": 1.1731, + "step": 140940 + }, + { + "epoch": 12.2, + "learning_rate": 3.783245256865633e-05, + "loss": 1.1315, + "step": 140950 + }, + { + "epoch": 12.2, + "learning_rate": 3.783158624274452e-05, + "loss": 1.1506, + "step": 140960 + }, + { + "epoch": 12.2, + "learning_rate": 3.783071991683271e-05, + "loss": 1.1578, + "step": 140970 + }, + { + "epoch": 12.2, + "learning_rate": 3.782985359092091e-05, + "loss": 1.1324, + "step": 140980 + }, + { + "epoch": 12.2, + "learning_rate": 3.7828987265009095e-05, + "loss": 1.1598, + "step": 140990 + }, + { + "epoch": 12.2, + "learning_rate": 3.782812093909729e-05, + "loss": 1.0941, + "step": 141000 + }, + { + "epoch": 12.21, + "learning_rate": 3.7827254613185484e-05, + "loss": 1.1804, + "step": 141010 + }, + { + "epoch": 12.21, + "learning_rate": 3.782638828727367e-05, + "loss": 1.153, + "step": 141020 + }, + { + "epoch": 12.21, + "learning_rate": 3.7825521961361866e-05, + "loss": 1.1284, + "step": 141030 + }, + { + "epoch": 12.21, + "learning_rate": 3.7824655635450054e-05, + "loss": 1.1706, + "step": 141040 + }, + { + "epoch": 12.21, + "learning_rate": 3.782378930953825e-05, + "loss": 1.1447, + "step": 141050 + }, + { + "epoch": 12.21, + "learning_rate": 3.782292298362644e-05, + "loss": 1.1048, + "step": 141060 + }, + { + "epoch": 12.21, + "learning_rate": 3.782205665771463e-05, + "loss": 1.1234, + "step": 141070 + }, + { + "epoch": 12.21, + "learning_rate": 3.7821190331802825e-05, + "loss": 1.1466, + "step": 141080 + }, + { + "epoch": 12.21, + "learning_rate": 3.782032400589102e-05, + "loss": 1.1735, + "step": 141090 + }, + { + "epoch": 12.21, + "learning_rate": 3.781945767997921e-05, + "loss": 1.1123, + "step": 141100 + }, + { + "epoch": 12.21, + "learning_rate": 3.78185913540674e-05, + "loss": 1.1413, + "step": 141110 + }, + { + "epoch": 12.21, + "learning_rate": 3.7817725028155596e-05, + "loss": 1.1659, + "step": 141120 + }, + { + "epoch": 12.22, + "learning_rate": 3.7816858702243784e-05, + "loss": 1.1607, + "step": 141130 + }, + { + "epoch": 12.22, + "learning_rate": 3.781599237633198e-05, + "loss": 1.0798, + "step": 141140 + }, + { + "epoch": 12.22, + "learning_rate": 3.7815126050420166e-05, + "loss": 1.1214, + "step": 141150 + }, + { + "epoch": 12.22, + "learning_rate": 3.781425972450836e-05, + "loss": 1.1449, + "step": 141160 + }, + { + "epoch": 12.22, + "learning_rate": 3.7813393398596555e-05, + "loss": 1.1008, + "step": 141170 + }, + { + "epoch": 12.22, + "learning_rate": 3.781252707268474e-05, + "loss": 1.1592, + "step": 141180 + }, + { + "epoch": 12.22, + "learning_rate": 3.781166074677294e-05, + "loss": 1.1195, + "step": 141190 + }, + { + "epoch": 12.22, + "learning_rate": 3.781079442086113e-05, + "loss": 1.1039, + "step": 141200 + }, + { + "epoch": 12.22, + "learning_rate": 3.780992809494932e-05, + "loss": 1.1824, + "step": 141210 + }, + { + "epoch": 12.22, + "learning_rate": 3.780906176903751e-05, + "loss": 1.1233, + "step": 141220 + }, + { + "epoch": 12.22, + "learning_rate": 3.780819544312571e-05, + "loss": 1.1302, + "step": 141230 + }, + { + "epoch": 12.23, + "learning_rate": 3.7807329117213895e-05, + "loss": 1.0854, + "step": 141240 + }, + { + "epoch": 12.23, + "learning_rate": 3.780646279130209e-05, + "loss": 1.1409, + "step": 141250 + }, + { + "epoch": 12.23, + "learning_rate": 3.780559646539028e-05, + "loss": 1.1039, + "step": 141260 + }, + { + "epoch": 12.23, + "learning_rate": 3.780473013947847e-05, + "loss": 1.1994, + "step": 141270 + }, + { + "epoch": 12.23, + "learning_rate": 3.7803863813566667e-05, + "loss": 1.1772, + "step": 141280 + }, + { + "epoch": 12.23, + "learning_rate": 3.7802997487654854e-05, + "loss": 1.1329, + "step": 141290 + }, + { + "epoch": 12.23, + "learning_rate": 3.780213116174305e-05, + "loss": 1.1117, + "step": 141300 + }, + { + "epoch": 12.23, + "learning_rate": 3.780126483583124e-05, + "loss": 1.1132, + "step": 141310 + }, + { + "epoch": 12.23, + "learning_rate": 3.780039850991943e-05, + "loss": 1.1311, + "step": 141320 + }, + { + "epoch": 12.23, + "learning_rate": 3.7799532184007625e-05, + "loss": 1.1285, + "step": 141330 + }, + { + "epoch": 12.23, + "learning_rate": 3.779866585809582e-05, + "loss": 1.083, + "step": 141340 + }, + { + "epoch": 12.23, + "learning_rate": 3.779779953218401e-05, + "loss": 1.1635, + "step": 141350 + }, + { + "epoch": 12.24, + "learning_rate": 3.77969332062722e-05, + "loss": 1.1286, + "step": 141360 + }, + { + "epoch": 12.24, + "learning_rate": 3.779606688036039e-05, + "loss": 1.0891, + "step": 141370 + }, + { + "epoch": 12.24, + "learning_rate": 3.7795200554448584e-05, + "loss": 1.0939, + "step": 141380 + }, + { + "epoch": 12.24, + "learning_rate": 3.779433422853678e-05, + "loss": 1.1157, + "step": 141390 + }, + { + "epoch": 12.24, + "learning_rate": 3.7793467902624966e-05, + "loss": 1.1482, + "step": 141400 + }, + { + "epoch": 12.24, + "learning_rate": 3.779260157671316e-05, + "loss": 1.1606, + "step": 141410 + }, + { + "epoch": 12.24, + "learning_rate": 3.7791735250801355e-05, + "loss": 1.1512, + "step": 141420 + }, + { + "epoch": 12.24, + "learning_rate": 3.779086892488954e-05, + "loss": 1.0908, + "step": 141430 + }, + { + "epoch": 12.24, + "learning_rate": 3.779000259897774e-05, + "loss": 1.1073, + "step": 141440 + }, + { + "epoch": 12.24, + "learning_rate": 3.778913627306593e-05, + "loss": 1.1316, + "step": 141450 + }, + { + "epoch": 12.24, + "learning_rate": 3.778826994715412e-05, + "loss": 1.0958, + "step": 141460 + }, + { + "epoch": 12.25, + "learning_rate": 3.7787403621242314e-05, + "loss": 1.0653, + "step": 141470 + }, + { + "epoch": 12.25, + "learning_rate": 3.77865372953305e-05, + "loss": 1.1028, + "step": 141480 + }, + { + "epoch": 12.25, + "learning_rate": 3.7785670969418696e-05, + "loss": 1.1644, + "step": 141490 + }, + { + "epoch": 12.25, + "learning_rate": 3.778480464350689e-05, + "loss": 1.1507, + "step": 141500 + }, + { + "epoch": 12.25, + "learning_rate": 3.778393831759508e-05, + "loss": 1.1243, + "step": 141510 + }, + { + "epoch": 12.25, + "learning_rate": 3.778307199168327e-05, + "loss": 1.0904, + "step": 141520 + }, + { + "epoch": 12.25, + "learning_rate": 3.778220566577147e-05, + "loss": 1.1641, + "step": 141530 + }, + { + "epoch": 12.25, + "learning_rate": 3.7781339339859655e-05, + "loss": 1.1496, + "step": 141540 + }, + { + "epoch": 12.25, + "learning_rate": 3.778047301394785e-05, + "loss": 1.0986, + "step": 141550 + }, + { + "epoch": 12.25, + "learning_rate": 3.777960668803604e-05, + "loss": 1.1723, + "step": 141560 + }, + { + "epoch": 12.25, + "learning_rate": 3.777874036212423e-05, + "loss": 1.1255, + "step": 141570 + }, + { + "epoch": 12.25, + "learning_rate": 3.7777874036212426e-05, + "loss": 1.0844, + "step": 141580 + }, + { + "epoch": 12.26, + "learning_rate": 3.777700771030061e-05, + "loss": 1.0968, + "step": 141590 + }, + { + "epoch": 12.26, + "learning_rate": 3.777614138438881e-05, + "loss": 1.1676, + "step": 141600 + }, + { + "epoch": 12.26, + "learning_rate": 3.7775275058477e-05, + "loss": 1.1755, + "step": 141610 + }, + { + "epoch": 12.26, + "learning_rate": 3.777440873256519e-05, + "loss": 1.095, + "step": 141620 + }, + { + "epoch": 12.26, + "learning_rate": 3.7773542406653384e-05, + "loss": 1.0887, + "step": 141630 + }, + { + "epoch": 12.26, + "learning_rate": 3.777267608074158e-05, + "loss": 1.1525, + "step": 141640 + }, + { + "epoch": 12.26, + "learning_rate": 3.7771809754829766e-05, + "loss": 1.1315, + "step": 141650 + }, + { + "epoch": 12.26, + "learning_rate": 3.777094342891796e-05, + "loss": 1.1475, + "step": 141660 + }, + { + "epoch": 12.26, + "learning_rate": 3.777007710300615e-05, + "loss": 1.2014, + "step": 141670 + }, + { + "epoch": 12.26, + "learning_rate": 3.776921077709434e-05, + "loss": 1.1377, + "step": 141680 + }, + { + "epoch": 12.26, + "learning_rate": 3.776834445118254e-05, + "loss": 1.094, + "step": 141690 + }, + { + "epoch": 12.26, + "learning_rate": 3.7767478125270725e-05, + "loss": 1.1375, + "step": 141700 + }, + { + "epoch": 12.27, + "learning_rate": 3.776661179935892e-05, + "loss": 1.1448, + "step": 141710 + }, + { + "epoch": 12.27, + "learning_rate": 3.7765745473447114e-05, + "loss": 1.1578, + "step": 141720 + }, + { + "epoch": 12.27, + "learning_rate": 3.77648791475353e-05, + "loss": 1.0831, + "step": 141730 + }, + { + "epoch": 12.27, + "learning_rate": 3.7764012821623496e-05, + "loss": 1.1759, + "step": 141740 + }, + { + "epoch": 12.27, + "learning_rate": 3.776314649571169e-05, + "loss": 1.1518, + "step": 141750 + }, + { + "epoch": 12.27, + "learning_rate": 3.776228016979988e-05, + "loss": 1.1128, + "step": 141760 + }, + { + "epoch": 12.27, + "learning_rate": 3.776141384388807e-05, + "loss": 1.1411, + "step": 141770 + }, + { + "epoch": 12.27, + "learning_rate": 3.776054751797626e-05, + "loss": 1.1855, + "step": 141780 + }, + { + "epoch": 12.27, + "learning_rate": 3.7759681192064455e-05, + "loss": 1.1007, + "step": 141790 + }, + { + "epoch": 12.27, + "learning_rate": 3.775881486615265e-05, + "loss": 1.1483, + "step": 141800 + }, + { + "epoch": 12.27, + "learning_rate": 3.775794854024084e-05, + "loss": 1.1147, + "step": 141810 + }, + { + "epoch": 12.28, + "learning_rate": 3.775708221432903e-05, + "loss": 1.1189, + "step": 141820 + }, + { + "epoch": 12.28, + "learning_rate": 3.7756215888417226e-05, + "loss": 1.0889, + "step": 141830 + }, + { + "epoch": 12.28, + "learning_rate": 3.7755349562505414e-05, + "loss": 1.119, + "step": 141840 + }, + { + "epoch": 12.28, + "learning_rate": 3.775448323659361e-05, + "loss": 1.1018, + "step": 141850 + }, + { + "epoch": 12.28, + "learning_rate": 3.77536169106818e-05, + "loss": 1.1661, + "step": 141860 + }, + { + "epoch": 12.28, + "learning_rate": 3.775275058476999e-05, + "loss": 1.1789, + "step": 141870 + }, + { + "epoch": 12.28, + "learning_rate": 3.7751884258858185e-05, + "loss": 1.1385, + "step": 141880 + }, + { + "epoch": 12.28, + "learning_rate": 3.775101793294637e-05, + "loss": 1.1571, + "step": 141890 + }, + { + "epoch": 12.28, + "learning_rate": 3.775015160703457e-05, + "loss": 1.1347, + "step": 141900 + }, + { + "epoch": 12.28, + "learning_rate": 3.774928528112276e-05, + "loss": 1.1221, + "step": 141910 + }, + { + "epoch": 12.28, + "learning_rate": 3.774841895521095e-05, + "loss": 1.1042, + "step": 141920 + }, + { + "epoch": 12.28, + "learning_rate": 3.774755262929914e-05, + "loss": 1.149, + "step": 141930 + }, + { + "epoch": 12.29, + "learning_rate": 3.774668630338734e-05, + "loss": 1.0994, + "step": 141940 + }, + { + "epoch": 12.29, + "learning_rate": 3.7745819977475525e-05, + "loss": 1.1437, + "step": 141950 + }, + { + "epoch": 12.29, + "learning_rate": 3.774495365156372e-05, + "loss": 1.1355, + "step": 141960 + }, + { + "epoch": 12.29, + "learning_rate": 3.7744087325651914e-05, + "loss": 1.1804, + "step": 141970 + }, + { + "epoch": 12.29, + "learning_rate": 3.77432209997401e-05, + "loss": 1.112, + "step": 141980 + }, + { + "epoch": 12.29, + "learning_rate": 3.7742354673828296e-05, + "loss": 1.1037, + "step": 141990 + }, + { + "epoch": 12.29, + "learning_rate": 3.7741488347916484e-05, + "loss": 1.1369, + "step": 142000 + }, + { + "epoch": 12.29, + "learning_rate": 3.774062202200468e-05, + "loss": 1.0852, + "step": 142010 + }, + { + "epoch": 12.29, + "learning_rate": 3.773975569609287e-05, + "loss": 1.1159, + "step": 142020 + }, + { + "epoch": 12.29, + "learning_rate": 3.773888937018106e-05, + "loss": 1.0902, + "step": 142030 + }, + { + "epoch": 12.29, + "learning_rate": 3.7738023044269255e-05, + "loss": 1.2046, + "step": 142040 + }, + { + "epoch": 12.3, + "learning_rate": 3.773715671835745e-05, + "loss": 1.1087, + "step": 142050 + }, + { + "epoch": 12.3, + "learning_rate": 3.773629039244564e-05, + "loss": 1.1025, + "step": 142060 + }, + { + "epoch": 12.3, + "learning_rate": 3.773542406653383e-05, + "loss": 1.0884, + "step": 142070 + }, + { + "epoch": 12.3, + "learning_rate": 3.7734557740622026e-05, + "loss": 1.167, + "step": 142080 + }, + { + "epoch": 12.3, + "learning_rate": 3.7733691414710214e-05, + "loss": 1.0816, + "step": 142090 + }, + { + "epoch": 12.3, + "learning_rate": 3.773282508879841e-05, + "loss": 1.0978, + "step": 142100 + }, + { + "epoch": 12.3, + "learning_rate": 3.7731958762886596e-05, + "loss": 1.1202, + "step": 142110 + }, + { + "epoch": 12.3, + "learning_rate": 3.773109243697479e-05, + "loss": 1.1232, + "step": 142120 + }, + { + "epoch": 12.3, + "learning_rate": 3.7730226111062985e-05, + "loss": 1.0939, + "step": 142130 + }, + { + "epoch": 12.3, + "learning_rate": 3.772935978515117e-05, + "loss": 1.1121, + "step": 142140 + }, + { + "epoch": 12.3, + "learning_rate": 3.772849345923937e-05, + "loss": 1.2148, + "step": 142150 + }, + { + "epoch": 12.3, + "learning_rate": 3.772762713332756e-05, + "loss": 1.1507, + "step": 142160 + }, + { + "epoch": 12.31, + "learning_rate": 3.772676080741575e-05, + "loss": 1.1905, + "step": 142170 + }, + { + "epoch": 12.31, + "learning_rate": 3.7725894481503944e-05, + "loss": 1.1194, + "step": 142180 + }, + { + "epoch": 12.31, + "learning_rate": 3.772502815559214e-05, + "loss": 1.141, + "step": 142190 + }, + { + "epoch": 12.31, + "learning_rate": 3.7724161829680326e-05, + "loss": 1.039, + "step": 142200 + }, + { + "epoch": 12.31, + "learning_rate": 3.772329550376852e-05, + "loss": 1.1486, + "step": 142210 + }, + { + "epoch": 12.31, + "learning_rate": 3.772242917785671e-05, + "loss": 1.0702, + "step": 142220 + }, + { + "epoch": 12.31, + "learning_rate": 3.77215628519449e-05, + "loss": 1.1484, + "step": 142230 + }, + { + "epoch": 12.31, + "learning_rate": 3.77206965260331e-05, + "loss": 1.1539, + "step": 142240 + }, + { + "epoch": 12.31, + "learning_rate": 3.7719830200121284e-05, + "loss": 1.1351, + "step": 142250 + }, + { + "epoch": 12.31, + "learning_rate": 3.771896387420948e-05, + "loss": 1.0999, + "step": 142260 + }, + { + "epoch": 12.31, + "learning_rate": 3.7718097548297673e-05, + "loss": 1.1196, + "step": 142270 + }, + { + "epoch": 12.32, + "learning_rate": 3.771723122238586e-05, + "loss": 1.1417, + "step": 142280 + }, + { + "epoch": 12.32, + "learning_rate": 3.7716364896474056e-05, + "loss": 1.1166, + "step": 142290 + }, + { + "epoch": 12.32, + "learning_rate": 3.771549857056224e-05, + "loss": 1.1316, + "step": 142300 + }, + { + "epoch": 12.32, + "learning_rate": 3.771463224465044e-05, + "loss": 1.2399, + "step": 142310 + }, + { + "epoch": 12.32, + "learning_rate": 3.771376591873863e-05, + "loss": 1.1521, + "step": 142320 + }, + { + "epoch": 12.32, + "learning_rate": 3.771289959282682e-05, + "loss": 1.1412, + "step": 142330 + }, + { + "epoch": 12.32, + "learning_rate": 3.7712033266915014e-05, + "loss": 1.1589, + "step": 142340 + }, + { + "epoch": 12.32, + "learning_rate": 3.771116694100321e-05, + "loss": 1.1593, + "step": 142350 + }, + { + "epoch": 12.32, + "learning_rate": 3.7710300615091396e-05, + "loss": 1.1498, + "step": 142360 + }, + { + "epoch": 12.32, + "learning_rate": 3.770943428917959e-05, + "loss": 1.1282, + "step": 142370 + }, + { + "epoch": 12.32, + "learning_rate": 3.7708567963267785e-05, + "loss": 1.1579, + "step": 142380 + }, + { + "epoch": 12.32, + "learning_rate": 3.770770163735597e-05, + "loss": 1.0917, + "step": 142390 + }, + { + "epoch": 12.33, + "learning_rate": 3.770683531144417e-05, + "loss": 1.0831, + "step": 142400 + }, + { + "epoch": 12.33, + "learning_rate": 3.7705968985532355e-05, + "loss": 1.1781, + "step": 142410 + }, + { + "epoch": 12.33, + "learning_rate": 3.770510265962055e-05, + "loss": 1.1487, + "step": 142420 + }, + { + "epoch": 12.33, + "learning_rate": 3.7704236333708744e-05, + "loss": 1.1426, + "step": 142430 + }, + { + "epoch": 12.33, + "learning_rate": 3.770337000779693e-05, + "loss": 1.1271, + "step": 142440 + }, + { + "epoch": 12.33, + "learning_rate": 3.7702503681885126e-05, + "loss": 1.1536, + "step": 142450 + }, + { + "epoch": 12.33, + "learning_rate": 3.770163735597332e-05, + "loss": 1.1355, + "step": 142460 + }, + { + "epoch": 12.33, + "learning_rate": 3.770077103006151e-05, + "loss": 1.1337, + "step": 142470 + }, + { + "epoch": 12.33, + "learning_rate": 3.76999047041497e-05, + "loss": 1.1274, + "step": 142480 + }, + { + "epoch": 12.33, + "learning_rate": 3.76990383782379e-05, + "loss": 1.1836, + "step": 142490 + }, + { + "epoch": 12.33, + "learning_rate": 3.7698172052326085e-05, + "loss": 1.1262, + "step": 142500 + }, + { + "epoch": 12.34, + "learning_rate": 3.769730572641428e-05, + "loss": 1.1347, + "step": 142510 + }, + { + "epoch": 12.34, + "learning_rate": 3.769643940050247e-05, + "loss": 1.1288, + "step": 142520 + }, + { + "epoch": 12.34, + "learning_rate": 3.769557307459066e-05, + "loss": 1.1341, + "step": 142530 + }, + { + "epoch": 12.34, + "learning_rate": 3.7694706748678856e-05, + "loss": 1.1318, + "step": 142540 + }, + { + "epoch": 12.34, + "learning_rate": 3.7693840422767044e-05, + "loss": 1.1535, + "step": 142550 + }, + { + "epoch": 12.34, + "learning_rate": 3.769297409685524e-05, + "loss": 1.0963, + "step": 142560 + }, + { + "epoch": 12.34, + "learning_rate": 3.769210777094343e-05, + "loss": 1.1182, + "step": 142570 + }, + { + "epoch": 12.34, + "learning_rate": 3.769124144503162e-05, + "loss": 1.1505, + "step": 142580 + }, + { + "epoch": 12.34, + "learning_rate": 3.7690375119119815e-05, + "loss": 1.1563, + "step": 142590 + }, + { + "epoch": 12.34, + "learning_rate": 3.768950879320801e-05, + "loss": 1.1886, + "step": 142600 + }, + { + "epoch": 12.34, + "learning_rate": 3.76886424672962e-05, + "loss": 1.1255, + "step": 142610 + }, + { + "epoch": 12.34, + "learning_rate": 3.768777614138439e-05, + "loss": 1.137, + "step": 142620 + }, + { + "epoch": 12.35, + "learning_rate": 3.768690981547258e-05, + "loss": 1.1208, + "step": 142630 + }, + { + "epoch": 12.35, + "learning_rate": 3.768604348956077e-05, + "loss": 1.099, + "step": 142640 + }, + { + "epoch": 12.35, + "learning_rate": 3.768517716364897e-05, + "loss": 1.0317, + "step": 142650 + }, + { + "epoch": 12.35, + "learning_rate": 3.7684310837737155e-05, + "loss": 1.1932, + "step": 142660 + }, + { + "epoch": 12.35, + "learning_rate": 3.768344451182535e-05, + "loss": 1.1047, + "step": 142670 + }, + { + "epoch": 12.35, + "learning_rate": 3.7682578185913544e-05, + "loss": 1.1296, + "step": 142680 + }, + { + "epoch": 12.35, + "learning_rate": 3.768171186000173e-05, + "loss": 1.1297, + "step": 142690 + }, + { + "epoch": 12.35, + "learning_rate": 3.7680845534089926e-05, + "loss": 1.0739, + "step": 142700 + }, + { + "epoch": 12.35, + "learning_rate": 3.767997920817812e-05, + "loss": 1.1284, + "step": 142710 + }, + { + "epoch": 12.35, + "learning_rate": 3.767911288226631e-05, + "loss": 1.1541, + "step": 142720 + }, + { + "epoch": 12.35, + "learning_rate": 3.76782465563545e-05, + "loss": 1.1713, + "step": 142730 + }, + { + "epoch": 12.35, + "learning_rate": 3.767738023044269e-05, + "loss": 1.1333, + "step": 142740 + }, + { + "epoch": 12.36, + "learning_rate": 3.7676513904530885e-05, + "loss": 1.1605, + "step": 142750 + }, + { + "epoch": 12.36, + "learning_rate": 3.767564757861908e-05, + "loss": 1.1191, + "step": 142760 + }, + { + "epoch": 12.36, + "learning_rate": 3.767478125270727e-05, + "loss": 1.1105, + "step": 142770 + }, + { + "epoch": 12.36, + "learning_rate": 3.767391492679546e-05, + "loss": 1.186, + "step": 142780 + }, + { + "epoch": 12.36, + "learning_rate": 3.7673048600883656e-05, + "loss": 1.1957, + "step": 142790 + }, + { + "epoch": 12.36, + "learning_rate": 3.7672182274971844e-05, + "loss": 1.1424, + "step": 142800 + }, + { + "epoch": 12.36, + "learning_rate": 3.767131594906004e-05, + "loss": 1.1556, + "step": 142810 + }, + { + "epoch": 12.36, + "learning_rate": 3.767044962314823e-05, + "loss": 1.1335, + "step": 142820 + }, + { + "epoch": 12.36, + "learning_rate": 3.766958329723642e-05, + "loss": 1.1227, + "step": 142830 + }, + { + "epoch": 12.36, + "learning_rate": 3.7668716971324615e-05, + "loss": 1.0757, + "step": 142840 + }, + { + "epoch": 12.36, + "learning_rate": 3.76678506454128e-05, + "loss": 1.1359, + "step": 142850 + }, + { + "epoch": 12.37, + "learning_rate": 3.7666984319501e-05, + "loss": 1.1191, + "step": 142860 + }, + { + "epoch": 12.37, + "learning_rate": 3.766611799358919e-05, + "loss": 1.1558, + "step": 142870 + }, + { + "epoch": 12.37, + "learning_rate": 3.766525166767738e-05, + "loss": 1.1251, + "step": 142880 + }, + { + "epoch": 12.37, + "learning_rate": 3.7664385341765574e-05, + "loss": 1.1518, + "step": 142890 + }, + { + "epoch": 12.37, + "learning_rate": 3.766351901585377e-05, + "loss": 1.0987, + "step": 142900 + }, + { + "epoch": 12.37, + "learning_rate": 3.7662652689941956e-05, + "loss": 1.0844, + "step": 142910 + }, + { + "epoch": 12.37, + "learning_rate": 3.766178636403015e-05, + "loss": 1.1011, + "step": 142920 + }, + { + "epoch": 12.37, + "learning_rate": 3.7660920038118345e-05, + "loss": 1.1162, + "step": 142930 + }, + { + "epoch": 12.37, + "learning_rate": 3.766005371220653e-05, + "loss": 1.1249, + "step": 142940 + }, + { + "epoch": 12.37, + "learning_rate": 3.765918738629473e-05, + "loss": 1.1658, + "step": 142950 + }, + { + "epoch": 12.37, + "learning_rate": 3.7658321060382914e-05, + "loss": 1.1057, + "step": 142960 + }, + { + "epoch": 12.37, + "learning_rate": 3.765745473447111e-05, + "loss": 1.2007, + "step": 142970 + }, + { + "epoch": 12.38, + "learning_rate": 3.76565884085593e-05, + "loss": 1.0841, + "step": 142980 + }, + { + "epoch": 12.38, + "learning_rate": 3.765572208264749e-05, + "loss": 1.1457, + "step": 142990 + }, + { + "epoch": 12.38, + "learning_rate": 3.7654855756735685e-05, + "loss": 1.0763, + "step": 143000 + }, + { + "epoch": 12.38, + "learning_rate": 3.765398943082388e-05, + "loss": 1.1815, + "step": 143010 + }, + { + "epoch": 12.38, + "learning_rate": 3.765312310491207e-05, + "loss": 1.1793, + "step": 143020 + }, + { + "epoch": 12.38, + "learning_rate": 3.765225677900026e-05, + "loss": 1.1151, + "step": 143030 + }, + { + "epoch": 12.38, + "learning_rate": 3.765139045308845e-05, + "loss": 1.1464, + "step": 143040 + }, + { + "epoch": 12.38, + "learning_rate": 3.7650524127176644e-05, + "loss": 1.1004, + "step": 143050 + }, + { + "epoch": 12.38, + "learning_rate": 3.764965780126484e-05, + "loss": 1.1892, + "step": 143060 + }, + { + "epoch": 12.38, + "learning_rate": 3.7648791475353026e-05, + "loss": 1.1081, + "step": 143070 + }, + { + "epoch": 12.38, + "learning_rate": 3.764792514944122e-05, + "loss": 1.1607, + "step": 143080 + }, + { + "epoch": 12.39, + "learning_rate": 3.7647058823529415e-05, + "loss": 1.1293, + "step": 143090 + }, + { + "epoch": 12.39, + "learning_rate": 3.76461924976176e-05, + "loss": 1.1519, + "step": 143100 + }, + { + "epoch": 12.39, + "learning_rate": 3.76453261717058e-05, + "loss": 1.1383, + "step": 143110 + }, + { + "epoch": 12.39, + "learning_rate": 3.764445984579399e-05, + "loss": 1.0726, + "step": 143120 + }, + { + "epoch": 12.39, + "learning_rate": 3.764359351988218e-05, + "loss": 1.1472, + "step": 143130 + }, + { + "epoch": 12.39, + "learning_rate": 3.7642727193970374e-05, + "loss": 1.1542, + "step": 143140 + }, + { + "epoch": 12.39, + "learning_rate": 3.764186086805856e-05, + "loss": 1.1087, + "step": 143150 + }, + { + "epoch": 12.39, + "learning_rate": 3.7640994542146756e-05, + "loss": 1.1283, + "step": 143160 + }, + { + "epoch": 12.39, + "learning_rate": 3.764012821623495e-05, + "loss": 1.1114, + "step": 143170 + }, + { + "epoch": 12.39, + "learning_rate": 3.763926189032314e-05, + "loss": 1.142, + "step": 143180 + }, + { + "epoch": 12.39, + "learning_rate": 3.763839556441133e-05, + "loss": 1.144, + "step": 143190 + }, + { + "epoch": 12.39, + "learning_rate": 3.763752923849953e-05, + "loss": 1.1343, + "step": 143200 + }, + { + "epoch": 12.4, + "learning_rate": 3.7636662912587715e-05, + "loss": 1.1339, + "step": 143210 + }, + { + "epoch": 12.4, + "learning_rate": 3.763579658667591e-05, + "loss": 1.1409, + "step": 143220 + }, + { + "epoch": 12.4, + "learning_rate": 3.7634930260764104e-05, + "loss": 1.1602, + "step": 143230 + }, + { + "epoch": 12.4, + "learning_rate": 3.763406393485229e-05, + "loss": 1.0936, + "step": 143240 + }, + { + "epoch": 12.4, + "learning_rate": 3.7633197608940486e-05, + "loss": 1.2031, + "step": 143250 + }, + { + "epoch": 12.4, + "learning_rate": 3.7632331283028674e-05, + "loss": 1.0933, + "step": 143260 + }, + { + "epoch": 12.4, + "learning_rate": 3.763146495711687e-05, + "loss": 1.0751, + "step": 143270 + }, + { + "epoch": 12.4, + "learning_rate": 3.763059863120506e-05, + "loss": 1.1032, + "step": 143280 + }, + { + "epoch": 12.4, + "learning_rate": 3.762973230529325e-05, + "loss": 1.1571, + "step": 143290 + }, + { + "epoch": 12.4, + "learning_rate": 3.7628865979381445e-05, + "loss": 1.1091, + "step": 143300 + }, + { + "epoch": 12.4, + "learning_rate": 3.762799965346964e-05, + "loss": 1.1463, + "step": 143310 + }, + { + "epoch": 12.41, + "learning_rate": 3.762713332755783e-05, + "loss": 1.1146, + "step": 143320 + }, + { + "epoch": 12.41, + "learning_rate": 3.762626700164602e-05, + "loss": 1.1318, + "step": 143330 + }, + { + "epoch": 12.41, + "learning_rate": 3.7625400675734216e-05, + "loss": 1.0925, + "step": 143340 + }, + { + "epoch": 12.41, + "learning_rate": 3.76245343498224e-05, + "loss": 1.0921, + "step": 143350 + }, + { + "epoch": 12.41, + "learning_rate": 3.76236680239106e-05, + "loss": 1.1116, + "step": 143360 + }, + { + "epoch": 12.41, + "learning_rate": 3.7622801697998785e-05, + "loss": 1.1509, + "step": 143370 + }, + { + "epoch": 12.41, + "learning_rate": 3.762193537208698e-05, + "loss": 1.1258, + "step": 143380 + }, + { + "epoch": 12.41, + "learning_rate": 3.7621069046175174e-05, + "loss": 1.1653, + "step": 143390 + }, + { + "epoch": 12.41, + "learning_rate": 3.762020272026336e-05, + "loss": 1.1029, + "step": 143400 + }, + { + "epoch": 12.41, + "learning_rate": 3.7619336394351556e-05, + "loss": 1.2107, + "step": 143410 + }, + { + "epoch": 12.41, + "learning_rate": 3.761847006843975e-05, + "loss": 1.1284, + "step": 143420 + }, + { + "epoch": 12.41, + "learning_rate": 3.761760374252794e-05, + "loss": 1.0908, + "step": 143430 + }, + { + "epoch": 12.42, + "learning_rate": 3.761673741661613e-05, + "loss": 1.1627, + "step": 143440 + }, + { + "epoch": 12.42, + "learning_rate": 3.761587109070433e-05, + "loss": 1.1378, + "step": 143450 + }, + { + "epoch": 12.42, + "learning_rate": 3.7615004764792515e-05, + "loss": 1.172, + "step": 143460 + }, + { + "epoch": 12.42, + "learning_rate": 3.761413843888071e-05, + "loss": 1.1604, + "step": 143470 + }, + { + "epoch": 12.42, + "learning_rate": 3.76132721129689e-05, + "loss": 1.1224, + "step": 143480 + }, + { + "epoch": 12.42, + "learning_rate": 3.761240578705709e-05, + "loss": 1.1601, + "step": 143490 + }, + { + "epoch": 12.42, + "learning_rate": 3.7611539461145286e-05, + "loss": 1.1305, + "step": 143500 + }, + { + "epoch": 12.42, + "learning_rate": 3.7610673135233474e-05, + "loss": 1.199, + "step": 143510 + }, + { + "epoch": 12.42, + "learning_rate": 3.760980680932167e-05, + "loss": 1.0995, + "step": 143520 + }, + { + "epoch": 12.42, + "learning_rate": 3.760894048340986e-05, + "loss": 1.0545, + "step": 143530 + }, + { + "epoch": 12.42, + "learning_rate": 3.760807415749805e-05, + "loss": 1.1509, + "step": 143540 + }, + { + "epoch": 12.43, + "learning_rate": 3.7607207831586245e-05, + "loss": 1.1291, + "step": 143550 + }, + { + "epoch": 12.43, + "learning_rate": 3.760634150567444e-05, + "loss": 1.0669, + "step": 143560 + }, + { + "epoch": 12.43, + "learning_rate": 3.760547517976263e-05, + "loss": 1.1242, + "step": 143570 + }, + { + "epoch": 12.43, + "learning_rate": 3.760460885385082e-05, + "loss": 1.0923, + "step": 143580 + }, + { + "epoch": 12.43, + "learning_rate": 3.760374252793901e-05, + "loss": 1.1139, + "step": 143590 + }, + { + "epoch": 12.43, + "learning_rate": 3.7602876202027204e-05, + "loss": 1.1318, + "step": 143600 + }, + { + "epoch": 12.43, + "learning_rate": 3.76020098761154e-05, + "loss": 1.1031, + "step": 143610 + }, + { + "epoch": 12.43, + "learning_rate": 3.7601143550203586e-05, + "loss": 1.0985, + "step": 143620 + }, + { + "epoch": 12.43, + "learning_rate": 3.760027722429178e-05, + "loss": 1.1276, + "step": 143630 + }, + { + "epoch": 12.43, + "learning_rate": 3.7599410898379975e-05, + "loss": 1.1144, + "step": 143640 + }, + { + "epoch": 12.43, + "learning_rate": 3.759854457246816e-05, + "loss": 1.1515, + "step": 143650 + }, + { + "epoch": 12.43, + "learning_rate": 3.759767824655636e-05, + "loss": 1.1549, + "step": 143660 + }, + { + "epoch": 12.44, + "learning_rate": 3.759681192064455e-05, + "loss": 1.1595, + "step": 143670 + }, + { + "epoch": 12.44, + "learning_rate": 3.759594559473274e-05, + "loss": 1.1502, + "step": 143680 + }, + { + "epoch": 12.44, + "learning_rate": 3.759507926882093e-05, + "loss": 1.1083, + "step": 143690 + }, + { + "epoch": 12.44, + "learning_rate": 3.759421294290912e-05, + "loss": 1.1537, + "step": 143700 + }, + { + "epoch": 12.44, + "learning_rate": 3.7593346616997315e-05, + "loss": 1.1215, + "step": 143710 + }, + { + "epoch": 12.44, + "learning_rate": 3.759248029108551e-05, + "loss": 1.094, + "step": 143720 + }, + { + "epoch": 12.44, + "learning_rate": 3.75916139651737e-05, + "loss": 1.073, + "step": 143730 + }, + { + "epoch": 12.44, + "learning_rate": 3.759074763926189e-05, + "loss": 1.1026, + "step": 143740 + }, + { + "epoch": 12.44, + "learning_rate": 3.7589881313350087e-05, + "loss": 1.1043, + "step": 143750 + }, + { + "epoch": 12.44, + "learning_rate": 3.7589014987438274e-05, + "loss": 1.1512, + "step": 143760 + }, + { + "epoch": 12.44, + "learning_rate": 3.758814866152647e-05, + "loss": 1.1124, + "step": 143770 + }, + { + "epoch": 12.44, + "learning_rate": 3.758728233561466e-05, + "loss": 1.089, + "step": 143780 + }, + { + "epoch": 12.45, + "learning_rate": 3.758641600970285e-05, + "loss": 1.1158, + "step": 143790 + }, + { + "epoch": 12.45, + "learning_rate": 3.7585549683791045e-05, + "loss": 1.1129, + "step": 143800 + }, + { + "epoch": 12.45, + "learning_rate": 3.758468335787923e-05, + "loss": 1.1822, + "step": 143810 + }, + { + "epoch": 12.45, + "learning_rate": 3.758381703196743e-05, + "loss": 1.1027, + "step": 143820 + }, + { + "epoch": 12.45, + "learning_rate": 3.758295070605562e-05, + "loss": 1.1836, + "step": 143830 + }, + { + "epoch": 12.45, + "learning_rate": 3.758208438014381e-05, + "loss": 1.109, + "step": 143840 + }, + { + "epoch": 12.45, + "learning_rate": 3.7581218054232004e-05, + "loss": 1.1403, + "step": 143850 + }, + { + "epoch": 12.45, + "learning_rate": 3.75803517283202e-05, + "loss": 1.1026, + "step": 143860 + }, + { + "epoch": 12.45, + "learning_rate": 3.7579485402408386e-05, + "loss": 1.1597, + "step": 143870 + }, + { + "epoch": 12.45, + "learning_rate": 3.757861907649658e-05, + "loss": 1.1726, + "step": 143880 + }, + { + "epoch": 12.45, + "learning_rate": 3.757775275058477e-05, + "loss": 1.1746, + "step": 143890 + }, + { + "epoch": 12.46, + "learning_rate": 3.757688642467296e-05, + "loss": 1.1779, + "step": 143900 + }, + { + "epoch": 12.46, + "learning_rate": 3.757602009876116e-05, + "loss": 1.1309, + "step": 143910 + }, + { + "epoch": 12.46, + "learning_rate": 3.7575153772849345e-05, + "loss": 1.1312, + "step": 143920 + }, + { + "epoch": 12.46, + "learning_rate": 3.757428744693754e-05, + "loss": 1.1428, + "step": 143930 + }, + { + "epoch": 12.46, + "learning_rate": 3.7573421121025734e-05, + "loss": 1.1749, + "step": 143940 + }, + { + "epoch": 12.46, + "learning_rate": 3.757255479511392e-05, + "loss": 1.1518, + "step": 143950 + }, + { + "epoch": 12.46, + "learning_rate": 3.7571688469202116e-05, + "loss": 1.1168, + "step": 143960 + }, + { + "epoch": 12.46, + "learning_rate": 3.757082214329031e-05, + "loss": 1.171, + "step": 143970 + }, + { + "epoch": 12.46, + "learning_rate": 3.75699558173785e-05, + "loss": 1.0922, + "step": 143980 + }, + { + "epoch": 12.46, + "learning_rate": 3.756908949146669e-05, + "loss": 1.1607, + "step": 143990 + }, + { + "epoch": 12.46, + "learning_rate": 3.756822316555488e-05, + "loss": 1.0977, + "step": 144000 + }, + { + "epoch": 12.46, + "learning_rate": 3.7567356839643075e-05, + "loss": 1.0777, + "step": 144010 + }, + { + "epoch": 12.47, + "learning_rate": 3.756649051373127e-05, + "loss": 1.1156, + "step": 144020 + }, + { + "epoch": 12.47, + "learning_rate": 3.756562418781946e-05, + "loss": 1.0922, + "step": 144030 + }, + { + "epoch": 12.47, + "learning_rate": 3.756475786190765e-05, + "loss": 1.1432, + "step": 144040 + }, + { + "epoch": 12.47, + "learning_rate": 3.7563891535995846e-05, + "loss": 1.1567, + "step": 144050 + }, + { + "epoch": 12.47, + "learning_rate": 3.756302521008403e-05, + "loss": 1.1577, + "step": 144060 + }, + { + "epoch": 12.47, + "learning_rate": 3.756215888417223e-05, + "loss": 1.1953, + "step": 144070 + }, + { + "epoch": 12.47, + "learning_rate": 3.756129255826042e-05, + "loss": 1.1464, + "step": 144080 + }, + { + "epoch": 12.47, + "learning_rate": 3.756042623234861e-05, + "loss": 1.1135, + "step": 144090 + }, + { + "epoch": 12.47, + "learning_rate": 3.7559559906436804e-05, + "loss": 1.1093, + "step": 144100 + }, + { + "epoch": 12.47, + "learning_rate": 3.755869358052499e-05, + "loss": 1.1197, + "step": 144110 + }, + { + "epoch": 12.47, + "learning_rate": 3.7557827254613186e-05, + "loss": 1.0837, + "step": 144120 + }, + { + "epoch": 12.48, + "learning_rate": 3.755696092870138e-05, + "loss": 1.1284, + "step": 144130 + }, + { + "epoch": 12.48, + "learning_rate": 3.755609460278957e-05, + "loss": 1.0482, + "step": 144140 + }, + { + "epoch": 12.48, + "learning_rate": 3.755522827687776e-05, + "loss": 1.0574, + "step": 144150 + }, + { + "epoch": 12.48, + "learning_rate": 3.755436195096596e-05, + "loss": 1.1752, + "step": 144160 + }, + { + "epoch": 12.48, + "learning_rate": 3.7553495625054145e-05, + "loss": 1.1407, + "step": 144170 + }, + { + "epoch": 12.48, + "learning_rate": 3.755262929914234e-05, + "loss": 1.1007, + "step": 144180 + }, + { + "epoch": 12.48, + "learning_rate": 3.7551762973230534e-05, + "loss": 1.1238, + "step": 144190 + }, + { + "epoch": 12.48, + "learning_rate": 3.755089664731872e-05, + "loss": 1.1286, + "step": 144200 + }, + { + "epoch": 12.48, + "learning_rate": 3.7550030321406916e-05, + "loss": 1.1682, + "step": 144210 + }, + { + "epoch": 12.48, + "learning_rate": 3.7549163995495104e-05, + "loss": 1.0765, + "step": 144220 + }, + { + "epoch": 12.48, + "learning_rate": 3.75482976695833e-05, + "loss": 1.144, + "step": 144230 + }, + { + "epoch": 12.48, + "learning_rate": 3.754743134367149e-05, + "loss": 1.1263, + "step": 144240 + }, + { + "epoch": 12.49, + "learning_rate": 3.754656501775968e-05, + "loss": 1.1265, + "step": 144250 + }, + { + "epoch": 12.49, + "learning_rate": 3.7545698691847875e-05, + "loss": 1.0803, + "step": 144260 + }, + { + "epoch": 12.49, + "learning_rate": 3.754483236593607e-05, + "loss": 1.1402, + "step": 144270 + }, + { + "epoch": 12.49, + "learning_rate": 3.754396604002426e-05, + "loss": 1.1556, + "step": 144280 + }, + { + "epoch": 12.49, + "learning_rate": 3.754309971411245e-05, + "loss": 1.1313, + "step": 144290 + }, + { + "epoch": 12.49, + "learning_rate": 3.7542233388200646e-05, + "loss": 1.1309, + "step": 144300 + }, + { + "epoch": 12.49, + "learning_rate": 3.7541367062288834e-05, + "loss": 1.1915, + "step": 144310 + }, + { + "epoch": 12.49, + "learning_rate": 3.754050073637703e-05, + "loss": 1.0878, + "step": 144320 + }, + { + "epoch": 12.49, + "learning_rate": 3.7539634410465216e-05, + "loss": 1.1398, + "step": 144330 + }, + { + "epoch": 12.49, + "learning_rate": 3.753876808455341e-05, + "loss": 1.1412, + "step": 144340 + }, + { + "epoch": 12.49, + "learning_rate": 3.7537901758641605e-05, + "loss": 1.0785, + "step": 144350 + }, + { + "epoch": 12.5, + "learning_rate": 3.753703543272979e-05, + "loss": 1.1851, + "step": 144360 + }, + { + "epoch": 12.5, + "learning_rate": 3.753616910681799e-05, + "loss": 1.1162, + "step": 144370 + }, + { + "epoch": 12.5, + "learning_rate": 3.753530278090618e-05, + "loss": 1.0784, + "step": 144380 + }, + { + "epoch": 12.5, + "learning_rate": 3.753443645499437e-05, + "loss": 1.0223, + "step": 144390 + }, + { + "epoch": 12.5, + "learning_rate": 3.753357012908256e-05, + "loss": 1.1032, + "step": 144400 + }, + { + "epoch": 12.5, + "learning_rate": 3.753270380317076e-05, + "loss": 1.1046, + "step": 144410 + }, + { + "epoch": 12.5, + "learning_rate": 3.7531837477258945e-05, + "loss": 1.1317, + "step": 144420 + }, + { + "epoch": 12.5, + "learning_rate": 3.753097115134714e-05, + "loss": 1.1466, + "step": 144430 + }, + { + "epoch": 12.5, + "learning_rate": 3.753010482543533e-05, + "loss": 1.096, + "step": 144440 + }, + { + "epoch": 12.5, + "learning_rate": 3.752923849952352e-05, + "loss": 1.1278, + "step": 144450 + }, + { + "epoch": 12.5, + "learning_rate": 3.7528372173611716e-05, + "loss": 1.1149, + "step": 144460 + }, + { + "epoch": 12.5, + "learning_rate": 3.7527505847699904e-05, + "loss": 1.0973, + "step": 144470 + }, + { + "epoch": 12.51, + "learning_rate": 3.75266395217881e-05, + "loss": 1.1545, + "step": 144480 + }, + { + "epoch": 12.51, + "learning_rate": 3.752577319587629e-05, + "loss": 1.1389, + "step": 144490 + }, + { + "epoch": 12.51, + "learning_rate": 3.752490686996448e-05, + "loss": 1.1558, + "step": 144500 + }, + { + "epoch": 12.51, + "learning_rate": 3.7524040544052675e-05, + "loss": 1.1082, + "step": 144510 + }, + { + "epoch": 12.51, + "learning_rate": 3.752317421814087e-05, + "loss": 1.1226, + "step": 144520 + }, + { + "epoch": 12.51, + "learning_rate": 3.752230789222906e-05, + "loss": 1.1305, + "step": 144530 + }, + { + "epoch": 12.51, + "learning_rate": 3.752144156631725e-05, + "loss": 1.1553, + "step": 144540 + }, + { + "epoch": 12.51, + "learning_rate": 3.752057524040544e-05, + "loss": 1.1415, + "step": 144550 + }, + { + "epoch": 12.51, + "learning_rate": 3.7519708914493634e-05, + "loss": 1.0957, + "step": 144560 + }, + { + "epoch": 12.51, + "learning_rate": 3.751884258858183e-05, + "loss": 1.0813, + "step": 144570 + }, + { + "epoch": 12.51, + "learning_rate": 3.7517976262670016e-05, + "loss": 1.1177, + "step": 144580 + }, + { + "epoch": 12.52, + "learning_rate": 3.751710993675821e-05, + "loss": 1.1068, + "step": 144590 + }, + { + "epoch": 12.52, + "learning_rate": 3.7516243610846405e-05, + "loss": 1.1193, + "step": 144600 + }, + { + "epoch": 12.52, + "learning_rate": 3.751537728493459e-05, + "loss": 1.0847, + "step": 144610 + }, + { + "epoch": 12.52, + "learning_rate": 3.751451095902279e-05, + "loss": 1.1392, + "step": 144620 + }, + { + "epoch": 12.52, + "learning_rate": 3.7513644633110975e-05, + "loss": 1.1371, + "step": 144630 + }, + { + "epoch": 12.52, + "learning_rate": 3.751277830719917e-05, + "loss": 1.104, + "step": 144640 + }, + { + "epoch": 12.52, + "learning_rate": 3.7511911981287364e-05, + "loss": 1.1309, + "step": 144650 + }, + { + "epoch": 12.52, + "learning_rate": 3.751104565537555e-05, + "loss": 1.1735, + "step": 144660 + }, + { + "epoch": 12.52, + "learning_rate": 3.7510179329463746e-05, + "loss": 1.1302, + "step": 144670 + }, + { + "epoch": 12.52, + "learning_rate": 3.750931300355194e-05, + "loss": 1.1882, + "step": 144680 + }, + { + "epoch": 12.52, + "learning_rate": 3.750844667764013e-05, + "loss": 1.1119, + "step": 144690 + }, + { + "epoch": 12.52, + "learning_rate": 3.750758035172832e-05, + "loss": 1.1811, + "step": 144700 + }, + { + "epoch": 12.53, + "learning_rate": 3.750671402581652e-05, + "loss": 1.1066, + "step": 144710 + }, + { + "epoch": 12.53, + "learning_rate": 3.7505847699904704e-05, + "loss": 1.1475, + "step": 144720 + }, + { + "epoch": 12.53, + "learning_rate": 3.75049813739929e-05, + "loss": 1.1699, + "step": 144730 + }, + { + "epoch": 12.53, + "learning_rate": 3.750411504808109e-05, + "loss": 1.1201, + "step": 144740 + }, + { + "epoch": 12.53, + "learning_rate": 3.750324872216928e-05, + "loss": 1.1773, + "step": 144750 + }, + { + "epoch": 12.53, + "learning_rate": 3.7502382396257476e-05, + "loss": 1.1982, + "step": 144760 + }, + { + "epoch": 12.53, + "learning_rate": 3.750151607034566e-05, + "loss": 1.1096, + "step": 144770 + }, + { + "epoch": 12.53, + "learning_rate": 3.750064974443386e-05, + "loss": 1.1136, + "step": 144780 + }, + { + "epoch": 12.53, + "learning_rate": 3.749978341852205e-05, + "loss": 1.1001, + "step": 144790 + }, + { + "epoch": 12.53, + "learning_rate": 3.749891709261024e-05, + "loss": 1.0839, + "step": 144800 + }, + { + "epoch": 12.53, + "learning_rate": 3.7498050766698434e-05, + "loss": 1.134, + "step": 144810 + }, + { + "epoch": 12.53, + "learning_rate": 3.749718444078663e-05, + "loss": 1.1273, + "step": 144820 + }, + { + "epoch": 12.54, + "learning_rate": 3.7496318114874816e-05, + "loss": 1.146, + "step": 144830 + }, + { + "epoch": 12.54, + "learning_rate": 3.749545178896301e-05, + "loss": 1.1452, + "step": 144840 + }, + { + "epoch": 12.54, + "learning_rate": 3.74945854630512e-05, + "loss": 1.1918, + "step": 144850 + }, + { + "epoch": 12.54, + "learning_rate": 3.749371913713939e-05, + "loss": 1.0621, + "step": 144860 + }, + { + "epoch": 12.54, + "learning_rate": 3.749285281122759e-05, + "loss": 1.1824, + "step": 144870 + }, + { + "epoch": 12.54, + "learning_rate": 3.7491986485315775e-05, + "loss": 1.1738, + "step": 144880 + }, + { + "epoch": 12.54, + "learning_rate": 3.749112015940397e-05, + "loss": 1.1478, + "step": 144890 + }, + { + "epoch": 12.54, + "learning_rate": 3.7490253833492164e-05, + "loss": 1.1756, + "step": 144900 + }, + { + "epoch": 12.54, + "learning_rate": 3.748938750758035e-05, + "loss": 1.0931, + "step": 144910 + }, + { + "epoch": 12.54, + "learning_rate": 3.7488521181668546e-05, + "loss": 1.1793, + "step": 144920 + }, + { + "epoch": 12.54, + "learning_rate": 3.748765485575674e-05, + "loss": 1.1914, + "step": 144930 + }, + { + "epoch": 12.55, + "learning_rate": 3.748678852984493e-05, + "loss": 1.083, + "step": 144940 + }, + { + "epoch": 12.55, + "learning_rate": 3.748592220393312e-05, + "loss": 1.107, + "step": 144950 + }, + { + "epoch": 12.55, + "learning_rate": 3.748505587802131e-05, + "loss": 1.0843, + "step": 144960 + }, + { + "epoch": 12.55, + "learning_rate": 3.7484189552109505e-05, + "loss": 1.1669, + "step": 144970 + }, + { + "epoch": 12.55, + "learning_rate": 3.74833232261977e-05, + "loss": 1.1613, + "step": 144980 + }, + { + "epoch": 12.55, + "learning_rate": 3.748245690028589e-05, + "loss": 1.1249, + "step": 144990 + }, + { + "epoch": 12.55, + "learning_rate": 3.748159057437408e-05, + "loss": 1.0686, + "step": 145000 + }, + { + "epoch": 12.55, + "learning_rate": 3.7480724248462276e-05, + "loss": 1.1128, + "step": 145010 + }, + { + "epoch": 12.55, + "learning_rate": 3.7479857922550464e-05, + "loss": 1.1118, + "step": 145020 + }, + { + "epoch": 12.55, + "learning_rate": 3.747899159663866e-05, + "loss": 1.152, + "step": 145030 + }, + { + "epoch": 12.55, + "learning_rate": 3.747812527072685e-05, + "loss": 1.1277, + "step": 145040 + }, + { + "epoch": 12.55, + "learning_rate": 3.747725894481504e-05, + "loss": 1.1127, + "step": 145050 + }, + { + "epoch": 12.56, + "learning_rate": 3.7476392618903235e-05, + "loss": 1.0854, + "step": 145060 + }, + { + "epoch": 12.56, + "learning_rate": 3.747552629299142e-05, + "loss": 1.1081, + "step": 145070 + }, + { + "epoch": 12.56, + "learning_rate": 3.747465996707962e-05, + "loss": 1.1571, + "step": 145080 + }, + { + "epoch": 12.56, + "learning_rate": 3.747379364116781e-05, + "loss": 1.117, + "step": 145090 + }, + { + "epoch": 12.56, + "learning_rate": 3.7472927315256e-05, + "loss": 1.1775, + "step": 145100 + }, + { + "epoch": 12.56, + "learning_rate": 3.747206098934419e-05, + "loss": 1.0888, + "step": 145110 + }, + { + "epoch": 12.56, + "learning_rate": 3.747119466343239e-05, + "loss": 1.1637, + "step": 145120 + }, + { + "epoch": 12.56, + "learning_rate": 3.7470328337520575e-05, + "loss": 1.1235, + "step": 145130 + }, + { + "epoch": 12.56, + "learning_rate": 3.746946201160877e-05, + "loss": 1.1265, + "step": 145140 + }, + { + "epoch": 12.56, + "learning_rate": 3.7468595685696964e-05, + "loss": 1.1214, + "step": 145150 + }, + { + "epoch": 12.56, + "learning_rate": 3.746772935978515e-05, + "loss": 1.1328, + "step": 145160 + }, + { + "epoch": 12.57, + "learning_rate": 3.7466863033873346e-05, + "loss": 1.117, + "step": 145170 + }, + { + "epoch": 12.57, + "learning_rate": 3.7465996707961534e-05, + "loss": 1.0774, + "step": 145180 + }, + { + "epoch": 12.57, + "learning_rate": 3.746513038204973e-05, + "loss": 1.1029, + "step": 145190 + }, + { + "epoch": 12.57, + "learning_rate": 3.746426405613792e-05, + "loss": 1.1332, + "step": 145200 + }, + { + "epoch": 12.57, + "learning_rate": 3.746339773022611e-05, + "loss": 1.106, + "step": 145210 + }, + { + "epoch": 12.57, + "learning_rate": 3.7462531404314305e-05, + "loss": 1.1501, + "step": 145220 + }, + { + "epoch": 12.57, + "learning_rate": 3.74616650784025e-05, + "loss": 1.1738, + "step": 145230 + }, + { + "epoch": 12.57, + "learning_rate": 3.746079875249069e-05, + "loss": 1.0992, + "step": 145240 + }, + { + "epoch": 12.57, + "learning_rate": 3.745993242657888e-05, + "loss": 1.0929, + "step": 145250 + }, + { + "epoch": 12.57, + "learning_rate": 3.7459066100667076e-05, + "loss": 1.155, + "step": 145260 + }, + { + "epoch": 12.57, + "learning_rate": 3.7458199774755264e-05, + "loss": 1.1357, + "step": 145270 + }, + { + "epoch": 12.57, + "learning_rate": 3.745733344884346e-05, + "loss": 1.1275, + "step": 145280 + }, + { + "epoch": 12.58, + "learning_rate": 3.7456467122931646e-05, + "loss": 1.1107, + "step": 145290 + }, + { + "epoch": 12.58, + "learning_rate": 3.745560079701984e-05, + "loss": 1.1338, + "step": 145300 + }, + { + "epoch": 12.58, + "learning_rate": 3.7454734471108035e-05, + "loss": 1.1098, + "step": 145310 + }, + { + "epoch": 12.58, + "learning_rate": 3.745386814519622e-05, + "loss": 1.1333, + "step": 145320 + }, + { + "epoch": 12.58, + "learning_rate": 3.745300181928442e-05, + "loss": 1.1049, + "step": 145330 + }, + { + "epoch": 12.58, + "learning_rate": 3.745213549337261e-05, + "loss": 1.1326, + "step": 145340 + }, + { + "epoch": 12.58, + "learning_rate": 3.74512691674608e-05, + "loss": 1.0982, + "step": 145350 + }, + { + "epoch": 12.58, + "learning_rate": 3.7450402841548994e-05, + "loss": 1.1271, + "step": 145360 + }, + { + "epoch": 12.58, + "learning_rate": 3.744953651563719e-05, + "loss": 1.1328, + "step": 145370 + }, + { + "epoch": 12.58, + "learning_rate": 3.7448670189725376e-05, + "loss": 1.0977, + "step": 145380 + }, + { + "epoch": 12.58, + "learning_rate": 3.744780386381357e-05, + "loss": 1.157, + "step": 145390 + }, + { + "epoch": 12.59, + "learning_rate": 3.744693753790176e-05, + "loss": 1.1336, + "step": 145400 + }, + { + "epoch": 12.59, + "learning_rate": 3.744607121198995e-05, + "loss": 1.1696, + "step": 145410 + }, + { + "epoch": 12.59, + "learning_rate": 3.744520488607815e-05, + "loss": 1.136, + "step": 145420 + }, + { + "epoch": 12.59, + "learning_rate": 3.7444338560166334e-05, + "loss": 1.1407, + "step": 145430 + }, + { + "epoch": 12.59, + "learning_rate": 3.744347223425453e-05, + "loss": 1.1726, + "step": 145440 + }, + { + "epoch": 12.59, + "learning_rate": 3.744260590834272e-05, + "loss": 1.0919, + "step": 145450 + }, + { + "epoch": 12.59, + "learning_rate": 3.744173958243091e-05, + "loss": 1.1037, + "step": 145460 + }, + { + "epoch": 12.59, + "learning_rate": 3.7440873256519105e-05, + "loss": 1.103, + "step": 145470 + }, + { + "epoch": 12.59, + "learning_rate": 3.744000693060729e-05, + "loss": 1.1434, + "step": 145480 + }, + { + "epoch": 12.59, + "learning_rate": 3.743914060469549e-05, + "loss": 1.122, + "step": 145490 + }, + { + "epoch": 12.59, + "learning_rate": 3.743827427878368e-05, + "loss": 1.1596, + "step": 145500 + }, + { + "epoch": 12.59, + "learning_rate": 3.743740795287187e-05, + "loss": 1.1603, + "step": 145510 + }, + { + "epoch": 12.6, + "learning_rate": 3.7436541626960064e-05, + "loss": 1.1333, + "step": 145520 + }, + { + "epoch": 12.6, + "learning_rate": 3.743567530104826e-05, + "loss": 1.1009, + "step": 145530 + }, + { + "epoch": 12.6, + "learning_rate": 3.7434808975136446e-05, + "loss": 1.1634, + "step": 145540 + }, + { + "epoch": 12.6, + "learning_rate": 3.743394264922464e-05, + "loss": 1.0955, + "step": 145550 + }, + { + "epoch": 12.6, + "learning_rate": 3.7433076323312835e-05, + "loss": 1.1276, + "step": 145560 + }, + { + "epoch": 12.6, + "learning_rate": 3.743220999740102e-05, + "loss": 1.0837, + "step": 145570 + }, + { + "epoch": 12.6, + "learning_rate": 3.743134367148922e-05, + "loss": 1.1144, + "step": 145580 + }, + { + "epoch": 12.6, + "learning_rate": 3.7430477345577405e-05, + "loss": 1.115, + "step": 145590 + }, + { + "epoch": 12.6, + "learning_rate": 3.74296110196656e-05, + "loss": 1.1943, + "step": 145600 + }, + { + "epoch": 12.6, + "learning_rate": 3.7428744693753794e-05, + "loss": 1.1231, + "step": 145610 + }, + { + "epoch": 12.6, + "learning_rate": 3.742787836784198e-05, + "loss": 1.1427, + "step": 145620 + }, + { + "epoch": 12.61, + "learning_rate": 3.7427012041930176e-05, + "loss": 1.1402, + "step": 145630 + }, + { + "epoch": 12.61, + "learning_rate": 3.742614571601837e-05, + "loss": 1.0982, + "step": 145640 + }, + { + "epoch": 12.61, + "learning_rate": 3.742527939010656e-05, + "loss": 1.1472, + "step": 145650 + }, + { + "epoch": 12.61, + "learning_rate": 3.742441306419475e-05, + "loss": 1.1475, + "step": 145660 + }, + { + "epoch": 12.61, + "learning_rate": 3.742354673828295e-05, + "loss": 1.1176, + "step": 145670 + }, + { + "epoch": 12.61, + "learning_rate": 3.7422680412371135e-05, + "loss": 1.1577, + "step": 145680 + }, + { + "epoch": 12.61, + "learning_rate": 3.742181408645933e-05, + "loss": 1.1543, + "step": 145690 + }, + { + "epoch": 12.61, + "learning_rate": 3.742094776054752e-05, + "loss": 1.1117, + "step": 145700 + }, + { + "epoch": 12.61, + "learning_rate": 3.742008143463571e-05, + "loss": 1.1394, + "step": 145710 + }, + { + "epoch": 12.61, + "learning_rate": 3.7419215108723906e-05, + "loss": 1.185, + "step": 145720 + }, + { + "epoch": 12.61, + "learning_rate": 3.7418348782812094e-05, + "loss": 1.1602, + "step": 145730 + }, + { + "epoch": 12.61, + "learning_rate": 3.741748245690029e-05, + "loss": 1.108, + "step": 145740 + }, + { + "epoch": 12.62, + "learning_rate": 3.741661613098848e-05, + "loss": 1.086, + "step": 145750 + }, + { + "epoch": 12.62, + "learning_rate": 3.741574980507667e-05, + "loss": 1.15, + "step": 145760 + }, + { + "epoch": 12.62, + "learning_rate": 3.7414883479164865e-05, + "loss": 1.1457, + "step": 145770 + }, + { + "epoch": 12.62, + "learning_rate": 3.741401715325306e-05, + "loss": 1.1422, + "step": 145780 + }, + { + "epoch": 12.62, + "learning_rate": 3.741315082734125e-05, + "loss": 1.1852, + "step": 145790 + }, + { + "epoch": 12.62, + "learning_rate": 3.741228450142944e-05, + "loss": 1.0933, + "step": 145800 + }, + { + "epoch": 12.62, + "learning_rate": 3.741141817551763e-05, + "loss": 1.1422, + "step": 145810 + }, + { + "epoch": 12.62, + "learning_rate": 3.741055184960582e-05, + "loss": 1.1445, + "step": 145820 + }, + { + "epoch": 12.62, + "learning_rate": 3.740968552369402e-05, + "loss": 1.1779, + "step": 145830 + }, + { + "epoch": 12.62, + "learning_rate": 3.7408819197782205e-05, + "loss": 1.1012, + "step": 145840 + }, + { + "epoch": 12.62, + "learning_rate": 3.74079528718704e-05, + "loss": 1.1033, + "step": 145850 + }, + { + "epoch": 12.63, + "learning_rate": 3.7407086545958594e-05, + "loss": 1.0979, + "step": 145860 + }, + { + "epoch": 12.63, + "learning_rate": 3.740622022004678e-05, + "loss": 1.1923, + "step": 145870 + }, + { + "epoch": 12.63, + "learning_rate": 3.7405353894134976e-05, + "loss": 1.0254, + "step": 145880 + }, + { + "epoch": 12.63, + "learning_rate": 3.740448756822317e-05, + "loss": 1.1487, + "step": 145890 + }, + { + "epoch": 12.63, + "learning_rate": 3.740362124231136e-05, + "loss": 1.1545, + "step": 145900 + }, + { + "epoch": 12.63, + "learning_rate": 3.740275491639955e-05, + "loss": 1.1306, + "step": 145910 + }, + { + "epoch": 12.63, + "learning_rate": 3.740188859048774e-05, + "loss": 1.1253, + "step": 145920 + }, + { + "epoch": 12.63, + "learning_rate": 3.7401022264575935e-05, + "loss": 1.0699, + "step": 145930 + }, + { + "epoch": 12.63, + "learning_rate": 3.740015593866413e-05, + "loss": 1.1477, + "step": 145940 + }, + { + "epoch": 12.63, + "learning_rate": 3.739928961275232e-05, + "loss": 1.1339, + "step": 145950 + }, + { + "epoch": 12.63, + "learning_rate": 3.739842328684051e-05, + "loss": 1.0833, + "step": 145960 + }, + { + "epoch": 12.63, + "learning_rate": 3.7397556960928706e-05, + "loss": 1.1312, + "step": 145970 + }, + { + "epoch": 12.64, + "learning_rate": 3.7396690635016894e-05, + "loss": 1.1434, + "step": 145980 + }, + { + "epoch": 12.64, + "learning_rate": 3.739582430910509e-05, + "loss": 1.066, + "step": 145990 + }, + { + "epoch": 12.64, + "learning_rate": 3.739495798319328e-05, + "loss": 1.1676, + "step": 146000 + }, + { + "epoch": 12.64, + "learning_rate": 3.739409165728147e-05, + "loss": 1.1252, + "step": 146010 + }, + { + "epoch": 12.64, + "learning_rate": 3.7393225331369665e-05, + "loss": 1.1332, + "step": 146020 + }, + { + "epoch": 12.64, + "learning_rate": 3.739235900545785e-05, + "loss": 1.1497, + "step": 146030 + }, + { + "epoch": 12.64, + "learning_rate": 3.739149267954605e-05, + "loss": 1.0976, + "step": 146040 + }, + { + "epoch": 12.64, + "learning_rate": 3.739062635363424e-05, + "loss": 1.1636, + "step": 146050 + }, + { + "epoch": 12.64, + "learning_rate": 3.738976002772243e-05, + "loss": 1.1229, + "step": 146060 + }, + { + "epoch": 12.64, + "learning_rate": 3.7388893701810624e-05, + "loss": 1.0679, + "step": 146070 + }, + { + "epoch": 12.64, + "learning_rate": 3.738802737589882e-05, + "loss": 1.1779, + "step": 146080 + }, + { + "epoch": 12.64, + "learning_rate": 3.7387161049987006e-05, + "loss": 1.1421, + "step": 146090 + }, + { + "epoch": 12.65, + "learning_rate": 3.73862947240752e-05, + "loss": 1.1471, + "step": 146100 + }, + { + "epoch": 12.65, + "learning_rate": 3.7385428398163395e-05, + "loss": 1.1462, + "step": 146110 + }, + { + "epoch": 12.65, + "learning_rate": 3.738456207225158e-05, + "loss": 1.0908, + "step": 146120 + }, + { + "epoch": 12.65, + "learning_rate": 3.738369574633978e-05, + "loss": 1.0968, + "step": 146130 + }, + { + "epoch": 12.65, + "learning_rate": 3.7382829420427964e-05, + "loss": 1.1176, + "step": 146140 + }, + { + "epoch": 12.65, + "learning_rate": 3.738196309451616e-05, + "loss": 1.171, + "step": 146150 + }, + { + "epoch": 12.65, + "learning_rate": 3.738109676860435e-05, + "loss": 1.1385, + "step": 146160 + }, + { + "epoch": 12.65, + "learning_rate": 3.738023044269254e-05, + "loss": 1.1093, + "step": 146170 + }, + { + "epoch": 12.65, + "learning_rate": 3.7379364116780735e-05, + "loss": 1.136, + "step": 146180 + }, + { + "epoch": 12.65, + "learning_rate": 3.737849779086893e-05, + "loss": 1.1772, + "step": 146190 + }, + { + "epoch": 12.65, + "learning_rate": 3.737763146495712e-05, + "loss": 1.104, + "step": 146200 + }, + { + "epoch": 12.66, + "learning_rate": 3.737676513904531e-05, + "loss": 1.1442, + "step": 146210 + }, + { + "epoch": 12.66, + "learning_rate": 3.73758988131335e-05, + "loss": 1.1747, + "step": 146220 + }, + { + "epoch": 12.66, + "learning_rate": 3.7375032487221694e-05, + "loss": 1.1305, + "step": 146230 + }, + { + "epoch": 12.66, + "learning_rate": 3.737416616130989e-05, + "loss": 1.1124, + "step": 146240 + }, + { + "epoch": 12.66, + "learning_rate": 3.7373299835398076e-05, + "loss": 1.1827, + "step": 146250 + }, + { + "epoch": 12.66, + "learning_rate": 3.737243350948627e-05, + "loss": 1.1001, + "step": 146260 + }, + { + "epoch": 12.66, + "learning_rate": 3.7371567183574465e-05, + "loss": 1.0735, + "step": 146270 + }, + { + "epoch": 12.66, + "learning_rate": 3.737070085766265e-05, + "loss": 1.1359, + "step": 146280 + }, + { + "epoch": 12.66, + "learning_rate": 3.736983453175085e-05, + "loss": 1.1113, + "step": 146290 + }, + { + "epoch": 12.66, + "learning_rate": 3.736896820583904e-05, + "loss": 1.1298, + "step": 146300 + }, + { + "epoch": 12.66, + "learning_rate": 3.736810187992723e-05, + "loss": 1.1717, + "step": 146310 + }, + { + "epoch": 12.66, + "learning_rate": 3.7367235554015424e-05, + "loss": 1.139, + "step": 146320 + }, + { + "epoch": 12.67, + "learning_rate": 3.736636922810361e-05, + "loss": 1.1545, + "step": 146330 + }, + { + "epoch": 12.67, + "learning_rate": 3.7365502902191806e-05, + "loss": 1.164, + "step": 146340 + }, + { + "epoch": 12.67, + "learning_rate": 3.736463657628e-05, + "loss": 1.1417, + "step": 146350 + }, + { + "epoch": 12.67, + "learning_rate": 3.736377025036819e-05, + "loss": 1.1491, + "step": 146360 + }, + { + "epoch": 12.67, + "learning_rate": 3.736290392445638e-05, + "loss": 1.1037, + "step": 146370 + }, + { + "epoch": 12.67, + "learning_rate": 3.736203759854458e-05, + "loss": 1.1521, + "step": 146380 + }, + { + "epoch": 12.67, + "learning_rate": 3.7361171272632765e-05, + "loss": 1.0918, + "step": 146390 + }, + { + "epoch": 12.67, + "learning_rate": 3.736030494672096e-05, + "loss": 1.0965, + "step": 146400 + }, + { + "epoch": 12.67, + "learning_rate": 3.7359438620809154e-05, + "loss": 1.0482, + "step": 146410 + }, + { + "epoch": 12.67, + "learning_rate": 3.735857229489734e-05, + "loss": 1.0909, + "step": 146420 + }, + { + "epoch": 12.67, + "learning_rate": 3.7357705968985536e-05, + "loss": 1.1268, + "step": 146430 + }, + { + "epoch": 12.68, + "learning_rate": 3.7356839643073723e-05, + "loss": 1.155, + "step": 146440 + }, + { + "epoch": 12.68, + "learning_rate": 3.735597331716192e-05, + "loss": 1.0969, + "step": 146450 + }, + { + "epoch": 12.68, + "learning_rate": 3.735510699125011e-05, + "loss": 1.0687, + "step": 146460 + }, + { + "epoch": 12.68, + "learning_rate": 3.73542406653383e-05, + "loss": 1.1367, + "step": 146470 + }, + { + "epoch": 12.68, + "learning_rate": 3.7353374339426495e-05, + "loss": 1.0943, + "step": 146480 + }, + { + "epoch": 12.68, + "learning_rate": 3.735250801351469e-05, + "loss": 1.122, + "step": 146490 + }, + { + "epoch": 12.68, + "learning_rate": 3.735164168760288e-05, + "loss": 1.1386, + "step": 146500 + }, + { + "epoch": 12.68, + "learning_rate": 3.735077536169107e-05, + "loss": 1.1054, + "step": 146510 + }, + { + "epoch": 12.68, + "learning_rate": 3.7349909035779266e-05, + "loss": 1.1326, + "step": 146520 + }, + { + "epoch": 12.68, + "learning_rate": 3.734904270986745e-05, + "loss": 1.1453, + "step": 146530 + }, + { + "epoch": 12.68, + "learning_rate": 3.734817638395565e-05, + "loss": 1.16, + "step": 146540 + }, + { + "epoch": 12.68, + "learning_rate": 3.7347310058043835e-05, + "loss": 1.1197, + "step": 146550 + }, + { + "epoch": 12.69, + "learning_rate": 3.734644373213203e-05, + "loss": 1.0839, + "step": 146560 + }, + { + "epoch": 12.69, + "learning_rate": 3.7345577406220224e-05, + "loss": 1.2238, + "step": 146570 + }, + { + "epoch": 12.69, + "learning_rate": 3.734471108030841e-05, + "loss": 1.1769, + "step": 146580 + }, + { + "epoch": 12.69, + "learning_rate": 3.7343844754396606e-05, + "loss": 1.161, + "step": 146590 + }, + { + "epoch": 12.69, + "learning_rate": 3.73429784284848e-05, + "loss": 1.1391, + "step": 146600 + }, + { + "epoch": 12.69, + "learning_rate": 3.734211210257299e-05, + "loss": 1.1358, + "step": 146610 + }, + { + "epoch": 12.69, + "learning_rate": 3.734124577666118e-05, + "loss": 1.1517, + "step": 146620 + }, + { + "epoch": 12.69, + "learning_rate": 3.734037945074938e-05, + "loss": 1.1435, + "step": 146630 + }, + { + "epoch": 12.69, + "learning_rate": 3.7339513124837565e-05, + "loss": 1.1404, + "step": 146640 + }, + { + "epoch": 12.69, + "learning_rate": 3.733864679892576e-05, + "loss": 1.1496, + "step": 146650 + }, + { + "epoch": 12.69, + "learning_rate": 3.733778047301395e-05, + "loss": 1.1759, + "step": 146660 + }, + { + "epoch": 12.7, + "learning_rate": 3.733691414710214e-05, + "loss": 1.1389, + "step": 146670 + }, + { + "epoch": 12.7, + "learning_rate": 3.7336047821190336e-05, + "loss": 1.1376, + "step": 146680 + }, + { + "epoch": 12.7, + "learning_rate": 3.7335181495278524e-05, + "loss": 1.0871, + "step": 146690 + }, + { + "epoch": 12.7, + "learning_rate": 3.733431516936672e-05, + "loss": 1.1611, + "step": 146700 + }, + { + "epoch": 12.7, + "learning_rate": 3.733344884345491e-05, + "loss": 1.1183, + "step": 146710 + }, + { + "epoch": 12.7, + "learning_rate": 3.73325825175431e-05, + "loss": 1.0993, + "step": 146720 + }, + { + "epoch": 12.7, + "learning_rate": 3.7331716191631295e-05, + "loss": 1.1004, + "step": 146730 + }, + { + "epoch": 12.7, + "learning_rate": 3.733084986571949e-05, + "loss": 1.0985, + "step": 146740 + }, + { + "epoch": 12.7, + "learning_rate": 3.732998353980768e-05, + "loss": 1.1445, + "step": 146750 + }, + { + "epoch": 12.7, + "learning_rate": 3.732911721389587e-05, + "loss": 1.1075, + "step": 146760 + }, + { + "epoch": 12.7, + "learning_rate": 3.732825088798406e-05, + "loss": 1.1696, + "step": 146770 + }, + { + "epoch": 12.7, + "learning_rate": 3.7327384562072254e-05, + "loss": 1.1146, + "step": 146780 + }, + { + "epoch": 12.71, + "learning_rate": 3.732651823616045e-05, + "loss": 1.1105, + "step": 146790 + }, + { + "epoch": 12.71, + "learning_rate": 3.7325651910248636e-05, + "loss": 1.1137, + "step": 146800 + }, + { + "epoch": 12.71, + "learning_rate": 3.732478558433683e-05, + "loss": 1.0672, + "step": 146810 + }, + { + "epoch": 12.71, + "learning_rate": 3.7323919258425025e-05, + "loss": 1.1511, + "step": 146820 + }, + { + "epoch": 12.71, + "learning_rate": 3.732305293251321e-05, + "loss": 1.138, + "step": 146830 + }, + { + "epoch": 12.71, + "learning_rate": 3.732218660660141e-05, + "loss": 1.163, + "step": 146840 + }, + { + "epoch": 12.71, + "learning_rate": 3.73213202806896e-05, + "loss": 1.1428, + "step": 146850 + }, + { + "epoch": 12.71, + "learning_rate": 3.732045395477779e-05, + "loss": 1.1196, + "step": 146860 + }, + { + "epoch": 12.71, + "learning_rate": 3.731958762886598e-05, + "loss": 1.1014, + "step": 146870 + }, + { + "epoch": 12.71, + "learning_rate": 3.731872130295417e-05, + "loss": 1.0651, + "step": 146880 + }, + { + "epoch": 12.71, + "learning_rate": 3.7317854977042365e-05, + "loss": 1.1594, + "step": 146890 + }, + { + "epoch": 12.72, + "learning_rate": 3.731698865113056e-05, + "loss": 1.0964, + "step": 146900 + }, + { + "epoch": 12.72, + "learning_rate": 3.731612232521875e-05, + "loss": 1.171, + "step": 146910 + }, + { + "epoch": 12.72, + "learning_rate": 3.731525599930694e-05, + "loss": 1.1677, + "step": 146920 + }, + { + "epoch": 12.72, + "learning_rate": 3.7314389673395136e-05, + "loss": 1.1319, + "step": 146930 + }, + { + "epoch": 12.72, + "learning_rate": 3.7313523347483324e-05, + "loss": 1.1068, + "step": 146940 + }, + { + "epoch": 12.72, + "learning_rate": 3.731265702157152e-05, + "loss": 1.1424, + "step": 146950 + }, + { + "epoch": 12.72, + "learning_rate": 3.7311790695659706e-05, + "loss": 1.1421, + "step": 146960 + }, + { + "epoch": 12.72, + "learning_rate": 3.73109243697479e-05, + "loss": 1.1019, + "step": 146970 + }, + { + "epoch": 12.72, + "learning_rate": 3.7310058043836095e-05, + "loss": 1.1106, + "step": 146980 + }, + { + "epoch": 12.72, + "learning_rate": 3.730919171792428e-05, + "loss": 1.1736, + "step": 146990 + }, + { + "epoch": 12.72, + "learning_rate": 3.730832539201248e-05, + "loss": 1.0886, + "step": 147000 + }, + { + "epoch": 12.72, + "learning_rate": 3.730745906610067e-05, + "loss": 1.1777, + "step": 147010 + }, + { + "epoch": 12.73, + "learning_rate": 3.730659274018886e-05, + "loss": 1.1057, + "step": 147020 + }, + { + "epoch": 12.73, + "learning_rate": 3.7305726414277054e-05, + "loss": 1.0657, + "step": 147030 + }, + { + "epoch": 12.73, + "learning_rate": 3.730486008836525e-05, + "loss": 1.089, + "step": 147040 + }, + { + "epoch": 12.73, + "learning_rate": 3.7303993762453436e-05, + "loss": 1.1373, + "step": 147050 + }, + { + "epoch": 12.73, + "learning_rate": 3.730312743654163e-05, + "loss": 1.1598, + "step": 147060 + }, + { + "epoch": 12.73, + "learning_rate": 3.730226111062982e-05, + "loss": 1.1467, + "step": 147070 + }, + { + "epoch": 12.73, + "learning_rate": 3.730139478471801e-05, + "loss": 1.0892, + "step": 147080 + }, + { + "epoch": 12.73, + "learning_rate": 3.730052845880621e-05, + "loss": 1.0838, + "step": 147090 + }, + { + "epoch": 12.73, + "learning_rate": 3.7299662132894395e-05, + "loss": 1.1254, + "step": 147100 + }, + { + "epoch": 12.73, + "learning_rate": 3.729879580698259e-05, + "loss": 1.1077, + "step": 147110 + }, + { + "epoch": 12.73, + "learning_rate": 3.7297929481070784e-05, + "loss": 1.1316, + "step": 147120 + }, + { + "epoch": 12.73, + "learning_rate": 3.729706315515897e-05, + "loss": 1.117, + "step": 147130 + }, + { + "epoch": 12.74, + "learning_rate": 3.7296196829247166e-05, + "loss": 1.1065, + "step": 147140 + }, + { + "epoch": 12.74, + "learning_rate": 3.729533050333536e-05, + "loss": 1.1243, + "step": 147150 + }, + { + "epoch": 12.74, + "learning_rate": 3.729446417742355e-05, + "loss": 1.1485, + "step": 147160 + }, + { + "epoch": 12.74, + "learning_rate": 3.729359785151174e-05, + "loss": 1.0549, + "step": 147170 + }, + { + "epoch": 12.74, + "learning_rate": 3.729273152559993e-05, + "loss": 1.1662, + "step": 147180 + }, + { + "epoch": 12.74, + "learning_rate": 3.7291865199688124e-05, + "loss": 1.1018, + "step": 147190 + }, + { + "epoch": 12.74, + "learning_rate": 3.729099887377632e-05, + "loss": 1.1795, + "step": 147200 + }, + { + "epoch": 12.74, + "learning_rate": 3.729013254786451e-05, + "loss": 1.0841, + "step": 147210 + }, + { + "epoch": 12.74, + "learning_rate": 3.72892662219527e-05, + "loss": 1.1374, + "step": 147220 + }, + { + "epoch": 12.74, + "learning_rate": 3.7288399896040896e-05, + "loss": 1.015, + "step": 147230 + }, + { + "epoch": 12.74, + "learning_rate": 3.728753357012908e-05, + "loss": 1.1009, + "step": 147240 + }, + { + "epoch": 12.75, + "learning_rate": 3.728666724421728e-05, + "loss": 1.1101, + "step": 147250 + }, + { + "epoch": 12.75, + "learning_rate": 3.728580091830547e-05, + "loss": 1.1208, + "step": 147260 + }, + { + "epoch": 12.75, + "learning_rate": 3.728493459239366e-05, + "loss": 1.1742, + "step": 147270 + }, + { + "epoch": 12.75, + "learning_rate": 3.7284068266481854e-05, + "loss": 1.133, + "step": 147280 + }, + { + "epoch": 12.75, + "learning_rate": 3.728320194057004e-05, + "loss": 1.1799, + "step": 147290 + }, + { + "epoch": 12.75, + "learning_rate": 3.7282335614658236e-05, + "loss": 1.1519, + "step": 147300 + }, + { + "epoch": 12.75, + "learning_rate": 3.728146928874643e-05, + "loss": 1.157, + "step": 147310 + }, + { + "epoch": 12.75, + "learning_rate": 3.728060296283462e-05, + "loss": 1.1605, + "step": 147320 + }, + { + "epoch": 12.75, + "learning_rate": 3.727973663692281e-05, + "loss": 1.107, + "step": 147330 + }, + { + "epoch": 12.75, + "learning_rate": 3.727887031101101e-05, + "loss": 1.1032, + "step": 147340 + }, + { + "epoch": 12.75, + "learning_rate": 3.7278003985099195e-05, + "loss": 1.1469, + "step": 147350 + }, + { + "epoch": 12.75, + "learning_rate": 3.727713765918739e-05, + "loss": 1.1752, + "step": 147360 + }, + { + "epoch": 12.76, + "learning_rate": 3.7276271333275584e-05, + "loss": 1.0985, + "step": 147370 + }, + { + "epoch": 12.76, + "learning_rate": 3.727540500736377e-05, + "loss": 1.1766, + "step": 147380 + }, + { + "epoch": 12.76, + "learning_rate": 3.7274538681451966e-05, + "loss": 1.1522, + "step": 147390 + }, + { + "epoch": 12.76, + "learning_rate": 3.7273672355540154e-05, + "loss": 1.1023, + "step": 147400 + }, + { + "epoch": 12.76, + "learning_rate": 3.727280602962835e-05, + "loss": 1.0837, + "step": 147410 + }, + { + "epoch": 12.76, + "learning_rate": 3.727193970371654e-05, + "loss": 1.1448, + "step": 147420 + }, + { + "epoch": 12.76, + "learning_rate": 3.727107337780473e-05, + "loss": 1.0695, + "step": 147430 + }, + { + "epoch": 12.76, + "learning_rate": 3.7270207051892925e-05, + "loss": 1.1034, + "step": 147440 + }, + { + "epoch": 12.76, + "learning_rate": 3.726934072598112e-05, + "loss": 1.0642, + "step": 147450 + }, + { + "epoch": 12.76, + "learning_rate": 3.726847440006931e-05, + "loss": 1.1064, + "step": 147460 + }, + { + "epoch": 12.76, + "learning_rate": 3.72676080741575e-05, + "loss": 1.2062, + "step": 147470 + }, + { + "epoch": 12.77, + "learning_rate": 3.7266741748245696e-05, + "loss": 1.172, + "step": 147480 + }, + { + "epoch": 12.77, + "learning_rate": 3.7265875422333884e-05, + "loss": 1.0708, + "step": 147490 + }, + { + "epoch": 12.77, + "learning_rate": 3.726500909642208e-05, + "loss": 1.1348, + "step": 147500 + }, + { + "epoch": 12.77, + "learning_rate": 3.7264142770510266e-05, + "loss": 1.1385, + "step": 147510 + }, + { + "epoch": 12.77, + "learning_rate": 3.726327644459846e-05, + "loss": 1.1294, + "step": 147520 + }, + { + "epoch": 12.77, + "learning_rate": 3.7262410118686655e-05, + "loss": 1.1601, + "step": 147530 + }, + { + "epoch": 12.77, + "learning_rate": 3.726154379277484e-05, + "loss": 1.1211, + "step": 147540 + }, + { + "epoch": 12.77, + "learning_rate": 3.726067746686304e-05, + "loss": 1.1535, + "step": 147550 + }, + { + "epoch": 12.77, + "learning_rate": 3.725981114095123e-05, + "loss": 1.126, + "step": 147560 + }, + { + "epoch": 12.77, + "learning_rate": 3.725894481503942e-05, + "loss": 1.1377, + "step": 147570 + }, + { + "epoch": 12.77, + "learning_rate": 3.725807848912761e-05, + "loss": 1.0655, + "step": 147580 + }, + { + "epoch": 12.77, + "learning_rate": 3.725721216321581e-05, + "loss": 1.0936, + "step": 147590 + }, + { + "epoch": 12.78, + "learning_rate": 3.7256345837303995e-05, + "loss": 1.1571, + "step": 147600 + }, + { + "epoch": 12.78, + "learning_rate": 3.725547951139219e-05, + "loss": 1.1175, + "step": 147610 + }, + { + "epoch": 12.78, + "learning_rate": 3.725461318548038e-05, + "loss": 1.1456, + "step": 147620 + }, + { + "epoch": 12.78, + "learning_rate": 3.725374685956857e-05, + "loss": 1.1173, + "step": 147630 + }, + { + "epoch": 12.78, + "learning_rate": 3.7252880533656766e-05, + "loss": 1.1045, + "step": 147640 + }, + { + "epoch": 12.78, + "learning_rate": 3.7252014207744954e-05, + "loss": 1.1714, + "step": 147650 + }, + { + "epoch": 12.78, + "learning_rate": 3.725114788183315e-05, + "loss": 1.2242, + "step": 147660 + }, + { + "epoch": 12.78, + "learning_rate": 3.725028155592134e-05, + "loss": 1.1333, + "step": 147670 + }, + { + "epoch": 12.78, + "learning_rate": 3.724941523000953e-05, + "loss": 1.1137, + "step": 147680 + }, + { + "epoch": 12.78, + "learning_rate": 3.7248548904097725e-05, + "loss": 1.1444, + "step": 147690 + }, + { + "epoch": 12.78, + "learning_rate": 3.724768257818592e-05, + "loss": 1.1299, + "step": 147700 + }, + { + "epoch": 12.79, + "learning_rate": 3.724681625227411e-05, + "loss": 1.1176, + "step": 147710 + }, + { + "epoch": 12.79, + "learning_rate": 3.72459499263623e-05, + "loss": 1.1094, + "step": 147720 + }, + { + "epoch": 12.79, + "learning_rate": 3.724508360045049e-05, + "loss": 1.1115, + "step": 147730 + }, + { + "epoch": 12.79, + "learning_rate": 3.7244217274538684e-05, + "loss": 1.1358, + "step": 147740 + }, + { + "epoch": 12.79, + "learning_rate": 3.724335094862688e-05, + "loss": 1.1141, + "step": 147750 + }, + { + "epoch": 12.79, + "learning_rate": 3.7242484622715066e-05, + "loss": 1.1185, + "step": 147760 + }, + { + "epoch": 12.79, + "learning_rate": 3.724161829680326e-05, + "loss": 1.1662, + "step": 147770 + }, + { + "epoch": 12.79, + "learning_rate": 3.7240751970891455e-05, + "loss": 1.0878, + "step": 147780 + }, + { + "epoch": 12.79, + "learning_rate": 3.723988564497964e-05, + "loss": 1.109, + "step": 147790 + }, + { + "epoch": 12.79, + "learning_rate": 3.723901931906784e-05, + "loss": 1.1055, + "step": 147800 + }, + { + "epoch": 12.79, + "learning_rate": 3.7238152993156025e-05, + "loss": 1.0465, + "step": 147810 + }, + { + "epoch": 12.79, + "learning_rate": 3.723728666724422e-05, + "loss": 1.0861, + "step": 147820 + }, + { + "epoch": 12.8, + "learning_rate": 3.7236420341332414e-05, + "loss": 1.1108, + "step": 147830 + }, + { + "epoch": 12.8, + "learning_rate": 3.72355540154206e-05, + "loss": 1.1665, + "step": 147840 + }, + { + "epoch": 12.8, + "learning_rate": 3.7234687689508796e-05, + "loss": 1.084, + "step": 147850 + }, + { + "epoch": 12.8, + "learning_rate": 3.723382136359699e-05, + "loss": 1.082, + "step": 147860 + }, + { + "epoch": 12.8, + "learning_rate": 3.723295503768518e-05, + "loss": 1.1785, + "step": 147870 + }, + { + "epoch": 12.8, + "learning_rate": 3.723208871177337e-05, + "loss": 1.1641, + "step": 147880 + }, + { + "epoch": 12.8, + "learning_rate": 3.723122238586157e-05, + "loss": 1.1271, + "step": 147890 + }, + { + "epoch": 12.8, + "learning_rate": 3.7230356059949754e-05, + "loss": 1.16, + "step": 147900 + }, + { + "epoch": 12.8, + "learning_rate": 3.722948973403795e-05, + "loss": 1.1282, + "step": 147910 + }, + { + "epoch": 12.8, + "learning_rate": 3.7228623408126137e-05, + "loss": 1.1604, + "step": 147920 + }, + { + "epoch": 12.8, + "learning_rate": 3.722775708221433e-05, + "loss": 1.1079, + "step": 147930 + }, + { + "epoch": 12.81, + "learning_rate": 3.7226890756302525e-05, + "loss": 1.1395, + "step": 147940 + }, + { + "epoch": 12.81, + "learning_rate": 3.722602443039071e-05, + "loss": 1.086, + "step": 147950 + }, + { + "epoch": 12.81, + "learning_rate": 3.722515810447891e-05, + "loss": 1.1958, + "step": 147960 + }, + { + "epoch": 12.81, + "learning_rate": 3.72242917785671e-05, + "loss": 1.1679, + "step": 147970 + }, + { + "epoch": 12.81, + "learning_rate": 3.722342545265529e-05, + "loss": 1.1463, + "step": 147980 + }, + { + "epoch": 12.81, + "learning_rate": 3.7222559126743484e-05, + "loss": 1.1251, + "step": 147990 + }, + { + "epoch": 12.81, + "learning_rate": 3.722169280083168e-05, + "loss": 1.145, + "step": 148000 + }, + { + "epoch": 12.81, + "learning_rate": 3.7220826474919866e-05, + "loss": 1.1095, + "step": 148010 + }, + { + "epoch": 12.81, + "learning_rate": 3.721996014900806e-05, + "loss": 1.1445, + "step": 148020 + }, + { + "epoch": 12.81, + "learning_rate": 3.721909382309625e-05, + "loss": 1.1721, + "step": 148030 + }, + { + "epoch": 12.81, + "learning_rate": 3.721822749718444e-05, + "loss": 1.0796, + "step": 148040 + }, + { + "epoch": 12.81, + "learning_rate": 3.721736117127264e-05, + "loss": 1.1514, + "step": 148050 + }, + { + "epoch": 12.82, + "learning_rate": 3.7216494845360825e-05, + "loss": 1.1093, + "step": 148060 + }, + { + "epoch": 12.82, + "learning_rate": 3.721562851944902e-05, + "loss": 1.1742, + "step": 148070 + }, + { + "epoch": 12.82, + "learning_rate": 3.7214762193537214e-05, + "loss": 1.2167, + "step": 148080 + }, + { + "epoch": 12.82, + "learning_rate": 3.72138958676254e-05, + "loss": 1.133, + "step": 148090 + }, + { + "epoch": 12.82, + "learning_rate": 3.7213029541713596e-05, + "loss": 1.1822, + "step": 148100 + }, + { + "epoch": 12.82, + "learning_rate": 3.721216321580179e-05, + "loss": 1.1685, + "step": 148110 + }, + { + "epoch": 12.82, + "learning_rate": 3.721129688988998e-05, + "loss": 1.1101, + "step": 148120 + }, + { + "epoch": 12.82, + "learning_rate": 3.721043056397817e-05, + "loss": 1.1243, + "step": 148130 + }, + { + "epoch": 12.82, + "learning_rate": 3.720956423806636e-05, + "loss": 1.1555, + "step": 148140 + }, + { + "epoch": 12.82, + "learning_rate": 3.7208697912154555e-05, + "loss": 1.1128, + "step": 148150 + }, + { + "epoch": 12.82, + "learning_rate": 3.720783158624275e-05, + "loss": 1.1044, + "step": 148160 + }, + { + "epoch": 12.82, + "learning_rate": 3.720696526033094e-05, + "loss": 1.1021, + "step": 148170 + }, + { + "epoch": 12.83, + "learning_rate": 3.720609893441913e-05, + "loss": 1.1256, + "step": 148180 + }, + { + "epoch": 12.83, + "learning_rate": 3.7205232608507326e-05, + "loss": 1.1467, + "step": 148190 + }, + { + "epoch": 12.83, + "learning_rate": 3.7204366282595514e-05, + "loss": 1.1553, + "step": 148200 + }, + { + "epoch": 12.83, + "learning_rate": 3.720349995668371e-05, + "loss": 1.1427, + "step": 148210 + }, + { + "epoch": 12.83, + "learning_rate": 3.72026336307719e-05, + "loss": 1.1107, + "step": 148220 + }, + { + "epoch": 12.83, + "learning_rate": 3.720176730486009e-05, + "loss": 1.1311, + "step": 148230 + }, + { + "epoch": 12.83, + "learning_rate": 3.7200900978948285e-05, + "loss": 1.1474, + "step": 148240 + }, + { + "epoch": 12.83, + "learning_rate": 3.720003465303647e-05, + "loss": 1.127, + "step": 148250 + }, + { + "epoch": 12.83, + "learning_rate": 3.719916832712467e-05, + "loss": 1.138, + "step": 148260 + }, + { + "epoch": 12.83, + "learning_rate": 3.719830200121286e-05, + "loss": 1.135, + "step": 148270 + }, + { + "epoch": 12.83, + "learning_rate": 3.719743567530105e-05, + "loss": 1.1493, + "step": 148280 + }, + { + "epoch": 12.84, + "learning_rate": 3.719656934938924e-05, + "loss": 1.095, + "step": 148290 + }, + { + "epoch": 12.84, + "learning_rate": 3.719570302347744e-05, + "loss": 1.1071, + "step": 148300 + }, + { + "epoch": 12.84, + "learning_rate": 3.7194836697565625e-05, + "loss": 1.0952, + "step": 148310 + }, + { + "epoch": 12.84, + "learning_rate": 3.719397037165382e-05, + "loss": 1.1161, + "step": 148320 + }, + { + "epoch": 12.84, + "learning_rate": 3.7193104045742014e-05, + "loss": 1.0516, + "step": 148330 + }, + { + "epoch": 12.84, + "learning_rate": 3.71922377198302e-05, + "loss": 1.1754, + "step": 148340 + }, + { + "epoch": 12.84, + "learning_rate": 3.7191371393918396e-05, + "loss": 1.0948, + "step": 148350 + }, + { + "epoch": 12.84, + "learning_rate": 3.7190505068006584e-05, + "loss": 1.1184, + "step": 148360 + }, + { + "epoch": 12.84, + "learning_rate": 3.718963874209478e-05, + "loss": 1.098, + "step": 148370 + }, + { + "epoch": 12.84, + "learning_rate": 3.718877241618297e-05, + "loss": 1.1436, + "step": 148380 + }, + { + "epoch": 12.84, + "learning_rate": 3.718790609027116e-05, + "loss": 1.1293, + "step": 148390 + }, + { + "epoch": 12.84, + "learning_rate": 3.7187039764359355e-05, + "loss": 1.1299, + "step": 148400 + }, + { + "epoch": 12.85, + "learning_rate": 3.718617343844755e-05, + "loss": 1.1355, + "step": 148410 + }, + { + "epoch": 12.85, + "learning_rate": 3.718530711253574e-05, + "loss": 1.1117, + "step": 148420 + }, + { + "epoch": 12.85, + "learning_rate": 3.718444078662393e-05, + "loss": 1.1685, + "step": 148430 + }, + { + "epoch": 12.85, + "learning_rate": 3.7183574460712126e-05, + "loss": 1.0335, + "step": 148440 + }, + { + "epoch": 12.85, + "learning_rate": 3.7182708134800314e-05, + "loss": 1.1377, + "step": 148450 + }, + { + "epoch": 12.85, + "learning_rate": 3.718184180888851e-05, + "loss": 1.0746, + "step": 148460 + }, + { + "epoch": 12.85, + "learning_rate": 3.7180975482976696e-05, + "loss": 1.1359, + "step": 148470 + }, + { + "epoch": 12.85, + "learning_rate": 3.718010915706489e-05, + "loss": 1.005, + "step": 148480 + }, + { + "epoch": 12.85, + "learning_rate": 3.7179242831153085e-05, + "loss": 1.1118, + "step": 148490 + }, + { + "epoch": 12.85, + "learning_rate": 3.717837650524127e-05, + "loss": 1.0505, + "step": 148500 + }, + { + "epoch": 12.85, + "learning_rate": 3.717751017932947e-05, + "loss": 1.1551, + "step": 148510 + }, + { + "epoch": 12.86, + "learning_rate": 3.717664385341766e-05, + "loss": 1.1087, + "step": 148520 + }, + { + "epoch": 12.86, + "learning_rate": 3.717577752750585e-05, + "loss": 1.1434, + "step": 148530 + }, + { + "epoch": 12.86, + "learning_rate": 3.7174911201594044e-05, + "loss": 1.105, + "step": 148540 + }, + { + "epoch": 12.86, + "learning_rate": 3.717404487568223e-05, + "loss": 1.1355, + "step": 148550 + }, + { + "epoch": 12.86, + "learning_rate": 3.7173178549770426e-05, + "loss": 1.1444, + "step": 148560 + }, + { + "epoch": 12.86, + "learning_rate": 3.717231222385862e-05, + "loss": 1.0799, + "step": 148570 + }, + { + "epoch": 12.86, + "learning_rate": 3.717144589794681e-05, + "loss": 1.0656, + "step": 148580 + }, + { + "epoch": 12.86, + "learning_rate": 3.7170579572035e-05, + "loss": 1.0994, + "step": 148590 + }, + { + "epoch": 12.86, + "learning_rate": 3.71697132461232e-05, + "loss": 1.1043, + "step": 148600 + }, + { + "epoch": 12.86, + "learning_rate": 3.7168846920211384e-05, + "loss": 1.1183, + "step": 148610 + }, + { + "epoch": 12.86, + "learning_rate": 3.716798059429958e-05, + "loss": 1.0776, + "step": 148620 + }, + { + "epoch": 12.86, + "learning_rate": 3.716711426838777e-05, + "loss": 1.1648, + "step": 148630 + }, + { + "epoch": 12.87, + "learning_rate": 3.716624794247596e-05, + "loss": 1.1172, + "step": 148640 + }, + { + "epoch": 12.87, + "learning_rate": 3.7165381616564155e-05, + "loss": 1.1081, + "step": 148650 + }, + { + "epoch": 12.87, + "learning_rate": 3.716451529065234e-05, + "loss": 1.1408, + "step": 148660 + }, + { + "epoch": 12.87, + "learning_rate": 3.716364896474054e-05, + "loss": 1.1499, + "step": 148670 + }, + { + "epoch": 12.87, + "learning_rate": 3.716278263882873e-05, + "loss": 1.1701, + "step": 148680 + }, + { + "epoch": 12.87, + "learning_rate": 3.716191631291692e-05, + "loss": 1.1359, + "step": 148690 + }, + { + "epoch": 12.87, + "learning_rate": 3.7161049987005114e-05, + "loss": 1.1261, + "step": 148700 + }, + { + "epoch": 12.87, + "learning_rate": 3.716018366109331e-05, + "loss": 1.1329, + "step": 148710 + }, + { + "epoch": 12.87, + "learning_rate": 3.7159317335181496e-05, + "loss": 1.1286, + "step": 148720 + }, + { + "epoch": 12.87, + "learning_rate": 3.715845100926969e-05, + "loss": 1.1116, + "step": 148730 + }, + { + "epoch": 12.87, + "learning_rate": 3.7157584683357885e-05, + "loss": 1.095, + "step": 148740 + }, + { + "epoch": 12.88, + "learning_rate": 3.715671835744607e-05, + "loss": 1.1333, + "step": 148750 + }, + { + "epoch": 12.88, + "learning_rate": 3.715585203153427e-05, + "loss": 1.1255, + "step": 148760 + }, + { + "epoch": 12.88, + "learning_rate": 3.7154985705622455e-05, + "loss": 1.0954, + "step": 148770 + }, + { + "epoch": 12.88, + "learning_rate": 3.715411937971065e-05, + "loss": 1.1279, + "step": 148780 + }, + { + "epoch": 12.88, + "learning_rate": 3.7153253053798844e-05, + "loss": 1.0734, + "step": 148790 + }, + { + "epoch": 12.88, + "learning_rate": 3.715238672788703e-05, + "loss": 1.1101, + "step": 148800 + }, + { + "epoch": 12.88, + "learning_rate": 3.7151520401975226e-05, + "loss": 1.1364, + "step": 148810 + }, + { + "epoch": 12.88, + "learning_rate": 3.715065407606342e-05, + "loss": 1.1615, + "step": 148820 + }, + { + "epoch": 12.88, + "learning_rate": 3.714978775015161e-05, + "loss": 1.1204, + "step": 148830 + }, + { + "epoch": 12.88, + "learning_rate": 3.71489214242398e-05, + "loss": 1.0836, + "step": 148840 + }, + { + "epoch": 12.88, + "learning_rate": 3.7148055098328e-05, + "loss": 1.0953, + "step": 148850 + }, + { + "epoch": 12.88, + "learning_rate": 3.7147188772416185e-05, + "loss": 1.136, + "step": 148860 + }, + { + "epoch": 12.89, + "learning_rate": 3.714632244650438e-05, + "loss": 1.0938, + "step": 148870 + }, + { + "epoch": 12.89, + "learning_rate": 3.714545612059257e-05, + "loss": 1.098, + "step": 148880 + }, + { + "epoch": 12.89, + "learning_rate": 3.714458979468076e-05, + "loss": 1.1155, + "step": 148890 + }, + { + "epoch": 12.89, + "learning_rate": 3.7143723468768956e-05, + "loss": 1.0929, + "step": 148900 + }, + { + "epoch": 12.89, + "learning_rate": 3.7142857142857143e-05, + "loss": 1.1639, + "step": 148910 + }, + { + "epoch": 12.89, + "learning_rate": 3.714199081694534e-05, + "loss": 1.0867, + "step": 148920 + }, + { + "epoch": 12.89, + "learning_rate": 3.714112449103353e-05, + "loss": 1.1377, + "step": 148930 + }, + { + "epoch": 12.89, + "learning_rate": 3.714025816512172e-05, + "loss": 1.1409, + "step": 148940 + }, + { + "epoch": 12.89, + "learning_rate": 3.7139391839209915e-05, + "loss": 1.1008, + "step": 148950 + }, + { + "epoch": 12.89, + "learning_rate": 3.713852551329811e-05, + "loss": 1.1499, + "step": 148960 + }, + { + "epoch": 12.89, + "learning_rate": 3.71376591873863e-05, + "loss": 1.1059, + "step": 148970 + }, + { + "epoch": 12.9, + "learning_rate": 3.713679286147449e-05, + "loss": 1.1225, + "step": 148980 + }, + { + "epoch": 12.9, + "learning_rate": 3.713592653556268e-05, + "loss": 1.1596, + "step": 148990 + }, + { + "epoch": 12.9, + "learning_rate": 3.713506020965087e-05, + "loss": 1.1442, + "step": 149000 + }, + { + "epoch": 12.9, + "learning_rate": 3.713419388373907e-05, + "loss": 1.1139, + "step": 149010 + }, + { + "epoch": 12.9, + "learning_rate": 3.7133327557827255e-05, + "loss": 1.1131, + "step": 149020 + }, + { + "epoch": 12.9, + "learning_rate": 3.713246123191545e-05, + "loss": 1.1297, + "step": 149030 + }, + { + "epoch": 12.9, + "learning_rate": 3.7131594906003644e-05, + "loss": 1.1841, + "step": 149040 + }, + { + "epoch": 12.9, + "learning_rate": 3.713072858009183e-05, + "loss": 1.1661, + "step": 149050 + }, + { + "epoch": 12.9, + "learning_rate": 3.7129862254180026e-05, + "loss": 1.1203, + "step": 149060 + }, + { + "epoch": 12.9, + "learning_rate": 3.712899592826822e-05, + "loss": 1.1224, + "step": 149070 + }, + { + "epoch": 12.9, + "learning_rate": 3.712812960235641e-05, + "loss": 1.094, + "step": 149080 + }, + { + "epoch": 12.9, + "learning_rate": 3.71272632764446e-05, + "loss": 1.0867, + "step": 149090 + }, + { + "epoch": 12.91, + "learning_rate": 3.712639695053279e-05, + "loss": 1.1827, + "step": 149100 + }, + { + "epoch": 12.91, + "learning_rate": 3.7125530624620985e-05, + "loss": 1.1082, + "step": 149110 + }, + { + "epoch": 12.91, + "learning_rate": 3.712466429870918e-05, + "loss": 1.1096, + "step": 149120 + }, + { + "epoch": 12.91, + "learning_rate": 3.712379797279737e-05, + "loss": 1.1593, + "step": 149130 + }, + { + "epoch": 12.91, + "learning_rate": 3.712293164688556e-05, + "loss": 1.0988, + "step": 149140 + }, + { + "epoch": 12.91, + "learning_rate": 3.7122065320973756e-05, + "loss": 1.1506, + "step": 149150 + }, + { + "epoch": 12.91, + "learning_rate": 3.7121198995061944e-05, + "loss": 1.1526, + "step": 149160 + }, + { + "epoch": 12.91, + "learning_rate": 3.712033266915014e-05, + "loss": 1.0903, + "step": 149170 + }, + { + "epoch": 12.91, + "learning_rate": 3.711946634323833e-05, + "loss": 1.131, + "step": 149180 + }, + { + "epoch": 12.91, + "learning_rate": 3.711860001732652e-05, + "loss": 1.1454, + "step": 149190 + }, + { + "epoch": 12.91, + "learning_rate": 3.7117733691414715e-05, + "loss": 1.1154, + "step": 149200 + }, + { + "epoch": 12.91, + "learning_rate": 3.71168673655029e-05, + "loss": 1.103, + "step": 149210 + }, + { + "epoch": 12.92, + "learning_rate": 3.71160010395911e-05, + "loss": 1.0927, + "step": 149220 + }, + { + "epoch": 12.92, + "learning_rate": 3.711513471367929e-05, + "loss": 1.1229, + "step": 149230 + }, + { + "epoch": 12.92, + "learning_rate": 3.711426838776748e-05, + "loss": 1.1411, + "step": 149240 + }, + { + "epoch": 12.92, + "learning_rate": 3.7113402061855674e-05, + "loss": 1.0807, + "step": 149250 + }, + { + "epoch": 12.92, + "learning_rate": 3.711253573594387e-05, + "loss": 1.085, + "step": 149260 + }, + { + "epoch": 12.92, + "learning_rate": 3.7111669410032056e-05, + "loss": 1.1075, + "step": 149270 + }, + { + "epoch": 12.92, + "learning_rate": 3.711080308412025e-05, + "loss": 1.169, + "step": 149280 + }, + { + "epoch": 12.92, + "learning_rate": 3.710993675820844e-05, + "loss": 1.0929, + "step": 149290 + }, + { + "epoch": 12.92, + "learning_rate": 3.710907043229663e-05, + "loss": 1.1241, + "step": 149300 + }, + { + "epoch": 12.92, + "learning_rate": 3.710820410638483e-05, + "loss": 1.1649, + "step": 149310 + }, + { + "epoch": 12.92, + "learning_rate": 3.7107337780473014e-05, + "loss": 1.1061, + "step": 149320 + }, + { + "epoch": 12.93, + "learning_rate": 3.710647145456121e-05, + "loss": 1.1596, + "step": 149330 + }, + { + "epoch": 12.93, + "learning_rate": 3.71056051286494e-05, + "loss": 1.0976, + "step": 149340 + }, + { + "epoch": 12.93, + "learning_rate": 3.710473880273759e-05, + "loss": 1.1381, + "step": 149350 + }, + { + "epoch": 12.93, + "learning_rate": 3.7103872476825785e-05, + "loss": 1.1487, + "step": 149360 + }, + { + "epoch": 12.93, + "learning_rate": 3.710300615091398e-05, + "loss": 1.1301, + "step": 149370 + }, + { + "epoch": 12.93, + "learning_rate": 3.710213982500217e-05, + "loss": 1.1548, + "step": 149380 + }, + { + "epoch": 12.93, + "learning_rate": 3.710127349909036e-05, + "loss": 1.1113, + "step": 149390 + }, + { + "epoch": 12.93, + "learning_rate": 3.710040717317855e-05, + "loss": 1.1423, + "step": 149400 + }, + { + "epoch": 12.93, + "learning_rate": 3.7099540847266744e-05, + "loss": 1.1368, + "step": 149410 + }, + { + "epoch": 12.93, + "learning_rate": 3.709867452135494e-05, + "loss": 1.1287, + "step": 149420 + }, + { + "epoch": 12.93, + "learning_rate": 3.7097808195443126e-05, + "loss": 1.124, + "step": 149430 + }, + { + "epoch": 12.93, + "learning_rate": 3.709694186953132e-05, + "loss": 1.1633, + "step": 149440 + }, + { + "epoch": 12.94, + "learning_rate": 3.7096075543619515e-05, + "loss": 1.124, + "step": 149450 + }, + { + "epoch": 12.94, + "learning_rate": 3.70952092177077e-05, + "loss": 1.139, + "step": 149460 + }, + { + "epoch": 12.94, + "learning_rate": 3.70943428917959e-05, + "loss": 1.0607, + "step": 149470 + }, + { + "epoch": 12.94, + "learning_rate": 3.709347656588409e-05, + "loss": 1.132, + "step": 149480 + }, + { + "epoch": 12.94, + "learning_rate": 3.709261023997228e-05, + "loss": 1.1279, + "step": 149490 + }, + { + "epoch": 12.94, + "learning_rate": 3.7091743914060474e-05, + "loss": 1.145, + "step": 149500 + }, + { + "epoch": 12.94, + "learning_rate": 3.709087758814866e-05, + "loss": 1.1803, + "step": 149510 + }, + { + "epoch": 12.94, + "learning_rate": 3.7090011262236856e-05, + "loss": 1.07, + "step": 149520 + }, + { + "epoch": 12.94, + "learning_rate": 3.708914493632505e-05, + "loss": 1.1375, + "step": 149530 + }, + { + "epoch": 12.94, + "learning_rate": 3.708827861041324e-05, + "loss": 1.1141, + "step": 149540 + }, + { + "epoch": 12.94, + "learning_rate": 3.708741228450143e-05, + "loss": 1.0625, + "step": 149550 + }, + { + "epoch": 12.95, + "learning_rate": 3.708654595858963e-05, + "loss": 1.1466, + "step": 149560 + }, + { + "epoch": 12.95, + "learning_rate": 3.7085679632677815e-05, + "loss": 1.1331, + "step": 149570 + }, + { + "epoch": 12.95, + "learning_rate": 3.708481330676601e-05, + "loss": 1.0785, + "step": 149580 + }, + { + "epoch": 12.95, + "learning_rate": 3.7083946980854204e-05, + "loss": 1.1053, + "step": 149590 + }, + { + "epoch": 12.95, + "learning_rate": 3.708308065494239e-05, + "loss": 1.1493, + "step": 149600 + }, + { + "epoch": 12.95, + "learning_rate": 3.7082214329030586e-05, + "loss": 1.064, + "step": 149610 + }, + { + "epoch": 12.95, + "learning_rate": 3.7081348003118773e-05, + "loss": 1.1206, + "step": 149620 + }, + { + "epoch": 12.95, + "learning_rate": 3.708048167720697e-05, + "loss": 1.1289, + "step": 149630 + }, + { + "epoch": 12.95, + "learning_rate": 3.707961535129516e-05, + "loss": 1.1252, + "step": 149640 + }, + { + "epoch": 12.95, + "learning_rate": 3.707874902538335e-05, + "loss": 1.065, + "step": 149650 + }, + { + "epoch": 12.95, + "learning_rate": 3.7077882699471544e-05, + "loss": 1.1169, + "step": 149660 + }, + { + "epoch": 12.95, + "learning_rate": 3.707701637355974e-05, + "loss": 1.1456, + "step": 149670 + }, + { + "epoch": 12.96, + "learning_rate": 3.707615004764793e-05, + "loss": 1.1957, + "step": 149680 + }, + { + "epoch": 12.96, + "learning_rate": 3.707528372173612e-05, + "loss": 1.1445, + "step": 149690 + }, + { + "epoch": 12.96, + "learning_rate": 3.7074417395824316e-05, + "loss": 1.081, + "step": 149700 + }, + { + "epoch": 12.96, + "learning_rate": 3.70735510699125e-05, + "loss": 1.1634, + "step": 149710 + }, + { + "epoch": 12.96, + "learning_rate": 3.70726847440007e-05, + "loss": 1.0933, + "step": 149720 + }, + { + "epoch": 12.96, + "learning_rate": 3.7071818418088885e-05, + "loss": 1.1777, + "step": 149730 + }, + { + "epoch": 12.96, + "learning_rate": 3.707095209217708e-05, + "loss": 1.1618, + "step": 149740 + }, + { + "epoch": 12.96, + "learning_rate": 3.7070085766265274e-05, + "loss": 1.1134, + "step": 149750 + }, + { + "epoch": 12.96, + "learning_rate": 3.706921944035346e-05, + "loss": 1.1531, + "step": 149760 + }, + { + "epoch": 12.96, + "learning_rate": 3.7068353114441656e-05, + "loss": 1.118, + "step": 149770 + }, + { + "epoch": 12.96, + "learning_rate": 3.706748678852985e-05, + "loss": 1.136, + "step": 149780 + }, + { + "epoch": 12.97, + "learning_rate": 3.706662046261804e-05, + "loss": 1.0673, + "step": 149790 + }, + { + "epoch": 12.97, + "learning_rate": 3.706575413670623e-05, + "loss": 1.0959, + "step": 149800 + }, + { + "epoch": 12.97, + "learning_rate": 3.706488781079443e-05, + "loss": 1.1077, + "step": 149810 + }, + { + "epoch": 12.97, + "learning_rate": 3.7064021484882615e-05, + "loss": 1.1557, + "step": 149820 + }, + { + "epoch": 12.97, + "learning_rate": 3.706315515897081e-05, + "loss": 1.1177, + "step": 149830 + }, + { + "epoch": 12.97, + "learning_rate": 3.7062288833059e-05, + "loss": 1.1598, + "step": 149840 + }, + { + "epoch": 12.97, + "learning_rate": 3.706142250714719e-05, + "loss": 1.1212, + "step": 149850 + }, + { + "epoch": 12.97, + "learning_rate": 3.7060556181235386e-05, + "loss": 1.1536, + "step": 149860 + }, + { + "epoch": 12.97, + "learning_rate": 3.7059689855323574e-05, + "loss": 1.1422, + "step": 149870 + }, + { + "epoch": 12.97, + "learning_rate": 3.705882352941177e-05, + "loss": 1.1361, + "step": 149880 + }, + { + "epoch": 12.97, + "learning_rate": 3.705795720349996e-05, + "loss": 1.1299, + "step": 149890 + }, + { + "epoch": 12.97, + "learning_rate": 3.705709087758815e-05, + "loss": 1.0791, + "step": 149900 + }, + { + "epoch": 12.98, + "learning_rate": 3.7056224551676345e-05, + "loss": 1.1017, + "step": 149910 + }, + { + "epoch": 12.98, + "learning_rate": 3.705535822576454e-05, + "loss": 1.0802, + "step": 149920 + }, + { + "epoch": 12.98, + "learning_rate": 3.705449189985273e-05, + "loss": 1.1221, + "step": 149930 + }, + { + "epoch": 12.98, + "learning_rate": 3.705362557394092e-05, + "loss": 1.106, + "step": 149940 + }, + { + "epoch": 12.98, + "learning_rate": 3.705275924802911e-05, + "loss": 1.1334, + "step": 149950 + }, + { + "epoch": 12.98, + "learning_rate": 3.7051892922117304e-05, + "loss": 1.0963, + "step": 149960 + }, + { + "epoch": 12.98, + "learning_rate": 3.70510265962055e-05, + "loss": 1.1231, + "step": 149970 + }, + { + "epoch": 12.98, + "learning_rate": 3.7050160270293686e-05, + "loss": 1.1224, + "step": 149980 + }, + { + "epoch": 12.98, + "learning_rate": 3.704929394438188e-05, + "loss": 1.1104, + "step": 149990 + }, + { + "epoch": 12.98, + "learning_rate": 3.7048427618470075e-05, + "loss": 1.1757, + "step": 150000 + }, + { + "epoch": 12.98, + "learning_rate": 3.704756129255826e-05, + "loss": 1.0643, + "step": 150010 + }, + { + "epoch": 12.99, + "learning_rate": 3.704669496664646e-05, + "loss": 1.1421, + "step": 150020 + }, + { + "epoch": 12.99, + "learning_rate": 3.704582864073465e-05, + "loss": 1.1771, + "step": 150030 + }, + { + "epoch": 12.99, + "learning_rate": 3.704496231482284e-05, + "loss": 1.1275, + "step": 150040 + }, + { + "epoch": 12.99, + "learning_rate": 3.704409598891103e-05, + "loss": 1.074, + "step": 150050 + }, + { + "epoch": 12.99, + "learning_rate": 3.704322966299922e-05, + "loss": 1.0678, + "step": 150060 + }, + { + "epoch": 12.99, + "learning_rate": 3.7042363337087415e-05, + "loss": 1.1615, + "step": 150070 + }, + { + "epoch": 12.99, + "learning_rate": 3.704149701117561e-05, + "loss": 1.1479, + "step": 150080 + }, + { + "epoch": 12.99, + "learning_rate": 3.70406306852638e-05, + "loss": 1.0415, + "step": 150090 + }, + { + "epoch": 12.99, + "learning_rate": 3.703976435935199e-05, + "loss": 1.1411, + "step": 150100 + }, + { + "epoch": 12.99, + "learning_rate": 3.7038898033440186e-05, + "loss": 1.2009, + "step": 150110 + }, + { + "epoch": 12.99, + "learning_rate": 3.7038031707528374e-05, + "loss": 1.1433, + "step": 150120 + }, + { + "epoch": 12.99, + "learning_rate": 3.703716538161657e-05, + "loss": 1.0981, + "step": 150130 + }, + { + "epoch": 13.0, + "learning_rate": 3.7036299055704756e-05, + "loss": 1.1222, + "step": 150140 + }, + { + "epoch": 13.0, + "learning_rate": 3.703543272979295e-05, + "loss": 1.1615, + "step": 150150 + }, + { + "epoch": 13.0, + "learning_rate": 3.7034566403881145e-05, + "loss": 1.1115, + "step": 150160 + }, + { + "epoch": 13.0, + "learning_rate": 3.703370007796933e-05, + "loss": 1.1024, + "step": 150170 + }, + { + "epoch": 13.0, + "learning_rate": 3.703283375205753e-05, + "loss": 1.0824, + "step": 150180 + }, + { + "epoch": 13.0, + "learning_rate": 3.703196742614572e-05, + "loss": 1.1303, + "step": 150190 + }, + { + "epoch": 13.0, + "eval_Bleu_1": 0.03586812458581783, + "eval_Bleu_2": 2.5485107656880558e-11, + "eval_Bleu_3": 2.3478283307293683e-14, + "eval_Bleu_4": 7.316835522297573e-16, + "eval_ROUGE_L": 0.0769872528712538, + "eval_cer": 0.9932627693314546, + "eval_em": 0, + "eval_f1": 0.09460293639074517, + "eval_loss": 1.0595579147338867, + "eval_runtime": 1740.7267, + "eval_samples_per_second": 2.95, + "eval_steps_per_second": 2.95, + "eval_wer": 0.9731479510424675, + "step": 150192 + }, + { + "epoch": 13.0, + "learning_rate": 3.703110110023391e-05, + "loss": 1.1112, + "step": 150200 + }, + { + "epoch": 13.0, + "learning_rate": 3.7030234774322104e-05, + "loss": 1.0997, + "step": 150210 + }, + { + "epoch": 13.0, + "learning_rate": 3.70293684484103e-05, + "loss": 1.1024, + "step": 150220 + }, + { + "epoch": 13.0, + "learning_rate": 3.7028502122498486e-05, + "loss": 1.1021, + "step": 150230 + }, + { + "epoch": 13.0, + "learning_rate": 3.702763579658668e-05, + "loss": 0.9987, + "step": 150240 + }, + { + "epoch": 13.0, + "learning_rate": 3.702676947067487e-05, + "loss": 1.026, + "step": 150250 + }, + { + "epoch": 13.01, + "learning_rate": 3.702590314476306e-05, + "loss": 1.1309, + "step": 150260 + }, + { + "epoch": 13.01, + "learning_rate": 3.702503681885126e-05, + "loss": 1.1183, + "step": 150270 + }, + { + "epoch": 13.01, + "learning_rate": 3.7024170492939445e-05, + "loss": 1.1339, + "step": 150280 + }, + { + "epoch": 13.01, + "learning_rate": 3.702330416702764e-05, + "loss": 1.1307, + "step": 150290 + }, + { + "epoch": 13.01, + "learning_rate": 3.7022437841115834e-05, + "loss": 1.1387, + "step": 150300 + }, + { + "epoch": 13.01, + "learning_rate": 3.702157151520402e-05, + "loss": 1.1254, + "step": 150310 + }, + { + "epoch": 13.01, + "learning_rate": 3.7020705189292216e-05, + "loss": 1.1261, + "step": 150320 + }, + { + "epoch": 13.01, + "learning_rate": 3.701983886338041e-05, + "loss": 1.1265, + "step": 150330 + }, + { + "epoch": 13.01, + "learning_rate": 3.70189725374686e-05, + "loss": 1.0471, + "step": 150340 + }, + { + "epoch": 13.01, + "learning_rate": 3.701810621155679e-05, + "loss": 1.0802, + "step": 150350 + }, + { + "epoch": 13.01, + "learning_rate": 3.701723988564498e-05, + "loss": 1.0802, + "step": 150360 + }, + { + "epoch": 13.02, + "learning_rate": 3.7016373559733174e-05, + "loss": 1.1012, + "step": 150370 + }, + { + "epoch": 13.02, + "learning_rate": 3.701550723382137e-05, + "loss": 1.1365, + "step": 150380 + }, + { + "epoch": 13.02, + "learning_rate": 3.7014640907909557e-05, + "loss": 1.1028, + "step": 150390 + }, + { + "epoch": 13.02, + "learning_rate": 3.701377458199775e-05, + "loss": 1.0451, + "step": 150400 + }, + { + "epoch": 13.02, + "learning_rate": 3.7012908256085945e-05, + "loss": 1.1165, + "step": 150410 + }, + { + "epoch": 13.02, + "learning_rate": 3.701204193017413e-05, + "loss": 1.1157, + "step": 150420 + }, + { + "epoch": 13.02, + "learning_rate": 3.701117560426233e-05, + "loss": 1.1034, + "step": 150430 + }, + { + "epoch": 13.02, + "learning_rate": 3.701030927835052e-05, + "loss": 1.0553, + "step": 150440 + }, + { + "epoch": 13.02, + "learning_rate": 3.700944295243871e-05, + "loss": 1.0616, + "step": 150450 + }, + { + "epoch": 13.02, + "learning_rate": 3.7008576626526904e-05, + "loss": 1.1024, + "step": 150460 + }, + { + "epoch": 13.02, + "learning_rate": 3.700771030061509e-05, + "loss": 1.0462, + "step": 150470 + }, + { + "epoch": 13.02, + "learning_rate": 3.7006843974703286e-05, + "loss": 1.0833, + "step": 150480 + }, + { + "epoch": 13.03, + "learning_rate": 3.700597764879148e-05, + "loss": 1.0765, + "step": 150490 + }, + { + "epoch": 13.03, + "learning_rate": 3.700511132287967e-05, + "loss": 1.1213, + "step": 150500 + }, + { + "epoch": 13.03, + "learning_rate": 3.700424499696786e-05, + "loss": 1.1007, + "step": 150510 + }, + { + "epoch": 13.03, + "learning_rate": 3.700337867105606e-05, + "loss": 1.1052, + "step": 150520 + }, + { + "epoch": 13.03, + "learning_rate": 3.7002512345144245e-05, + "loss": 1.0551, + "step": 150530 + }, + { + "epoch": 13.03, + "learning_rate": 3.700164601923244e-05, + "loss": 1.1141, + "step": 150540 + }, + { + "epoch": 13.03, + "learning_rate": 3.7000779693320634e-05, + "loss": 1.1058, + "step": 150550 + }, + { + "epoch": 13.03, + "learning_rate": 3.699991336740882e-05, + "loss": 1.0803, + "step": 150560 + }, + { + "epoch": 13.03, + "learning_rate": 3.6999047041497016e-05, + "loss": 1.16, + "step": 150570 + }, + { + "epoch": 13.03, + "learning_rate": 3.6998180715585204e-05, + "loss": 1.0575, + "step": 150580 + }, + { + "epoch": 13.03, + "learning_rate": 3.69973143896734e-05, + "loss": 1.0784, + "step": 150590 + }, + { + "epoch": 13.04, + "learning_rate": 3.699644806376159e-05, + "loss": 1.1046, + "step": 150600 + }, + { + "epoch": 13.04, + "learning_rate": 3.699558173784978e-05, + "loss": 1.088, + "step": 150610 + }, + { + "epoch": 13.04, + "learning_rate": 3.6994715411937975e-05, + "loss": 1.1197, + "step": 150620 + }, + { + "epoch": 13.04, + "learning_rate": 3.699384908602617e-05, + "loss": 1.127, + "step": 150630 + }, + { + "epoch": 13.04, + "learning_rate": 3.699298276011436e-05, + "loss": 1.0455, + "step": 150640 + }, + { + "epoch": 13.04, + "learning_rate": 3.699211643420255e-05, + "loss": 1.082, + "step": 150650 + }, + { + "epoch": 13.04, + "learning_rate": 3.6991250108290746e-05, + "loss": 1.1383, + "step": 150660 + }, + { + "epoch": 13.04, + "learning_rate": 3.6990383782378934e-05, + "loss": 1.0994, + "step": 150670 + }, + { + "epoch": 13.04, + "learning_rate": 3.698951745646713e-05, + "loss": 1.0495, + "step": 150680 + }, + { + "epoch": 13.04, + "learning_rate": 3.6988651130555316e-05, + "loss": 1.0493, + "step": 150690 + }, + { + "epoch": 13.04, + "learning_rate": 3.698778480464351e-05, + "loss": 1.1167, + "step": 150700 + }, + { + "epoch": 13.04, + "learning_rate": 3.6986918478731705e-05, + "loss": 1.126, + "step": 150710 + }, + { + "epoch": 13.05, + "learning_rate": 3.698605215281989e-05, + "loss": 1.1455, + "step": 150720 + }, + { + "epoch": 13.05, + "learning_rate": 3.698518582690809e-05, + "loss": 1.1004, + "step": 150730 + }, + { + "epoch": 13.05, + "learning_rate": 3.698431950099628e-05, + "loss": 1.0652, + "step": 150740 + }, + { + "epoch": 13.05, + "learning_rate": 3.698345317508447e-05, + "loss": 1.107, + "step": 150750 + }, + { + "epoch": 13.05, + "learning_rate": 3.698258684917266e-05, + "loss": 1.0867, + "step": 150760 + }, + { + "epoch": 13.05, + "learning_rate": 3.698172052326086e-05, + "loss": 1.1066, + "step": 150770 + }, + { + "epoch": 13.05, + "learning_rate": 3.6980854197349045e-05, + "loss": 1.0918, + "step": 150780 + }, + { + "epoch": 13.05, + "learning_rate": 3.697998787143724e-05, + "loss": 1.1167, + "step": 150790 + }, + { + "epoch": 13.05, + "learning_rate": 3.697912154552543e-05, + "loss": 1.1362, + "step": 150800 + }, + { + "epoch": 13.05, + "learning_rate": 3.697825521961362e-05, + "loss": 1.0983, + "step": 150810 + }, + { + "epoch": 13.05, + "learning_rate": 3.6977388893701816e-05, + "loss": 1.0771, + "step": 150820 + }, + { + "epoch": 13.06, + "learning_rate": 3.6976522567790004e-05, + "loss": 1.0625, + "step": 150830 + }, + { + "epoch": 13.06, + "learning_rate": 3.69756562418782e-05, + "loss": 1.1022, + "step": 150840 + }, + { + "epoch": 13.06, + "learning_rate": 3.697478991596639e-05, + "loss": 1.0719, + "step": 150850 + }, + { + "epoch": 13.06, + "learning_rate": 3.697392359005458e-05, + "loss": 1.1439, + "step": 150860 + }, + { + "epoch": 13.06, + "learning_rate": 3.6973057264142775e-05, + "loss": 1.0595, + "step": 150870 + }, + { + "epoch": 13.06, + "learning_rate": 3.697219093823096e-05, + "loss": 1.1418, + "step": 150880 + }, + { + "epoch": 13.06, + "learning_rate": 3.697132461231916e-05, + "loss": 1.0689, + "step": 150890 + }, + { + "epoch": 13.06, + "learning_rate": 3.697045828640735e-05, + "loss": 1.063, + "step": 150900 + }, + { + "epoch": 13.06, + "learning_rate": 3.696959196049554e-05, + "loss": 1.1024, + "step": 150910 + }, + { + "epoch": 13.06, + "learning_rate": 3.6968725634583734e-05, + "loss": 1.1535, + "step": 150920 + }, + { + "epoch": 13.06, + "learning_rate": 3.696785930867193e-05, + "loss": 1.1144, + "step": 150930 + }, + { + "epoch": 13.06, + "learning_rate": 3.6966992982760116e-05, + "loss": 1.0792, + "step": 150940 + }, + { + "epoch": 13.07, + "learning_rate": 3.696612665684831e-05, + "loss": 1.0624, + "step": 150950 + }, + { + "epoch": 13.07, + "learning_rate": 3.6965260330936505e-05, + "loss": 1.038, + "step": 150960 + }, + { + "epoch": 13.07, + "learning_rate": 3.696439400502469e-05, + "loss": 1.0754, + "step": 150970 + }, + { + "epoch": 13.07, + "learning_rate": 3.696352767911289e-05, + "loss": 1.1315, + "step": 150980 + }, + { + "epoch": 13.07, + "learning_rate": 3.6962661353201075e-05, + "loss": 1.0648, + "step": 150990 + }, + { + "epoch": 13.07, + "learning_rate": 3.696179502728927e-05, + "loss": 1.117, + "step": 151000 + }, + { + "epoch": 13.07, + "learning_rate": 3.6960928701377464e-05, + "loss": 1.1048, + "step": 151010 + }, + { + "epoch": 13.07, + "learning_rate": 3.696006237546565e-05, + "loss": 1.1099, + "step": 151020 + }, + { + "epoch": 13.07, + "learning_rate": 3.6959196049553846e-05, + "loss": 1.0723, + "step": 151030 + }, + { + "epoch": 13.07, + "learning_rate": 3.695832972364204e-05, + "loss": 1.1173, + "step": 151040 + }, + { + "epoch": 13.07, + "learning_rate": 3.695746339773023e-05, + "loss": 1.0951, + "step": 151050 + }, + { + "epoch": 13.08, + "learning_rate": 3.695659707181842e-05, + "loss": 1.121, + "step": 151060 + }, + { + "epoch": 13.08, + "learning_rate": 3.695573074590662e-05, + "loss": 1.1505, + "step": 151070 + }, + { + "epoch": 13.08, + "learning_rate": 3.6954864419994804e-05, + "loss": 1.0229, + "step": 151080 + }, + { + "epoch": 13.08, + "learning_rate": 3.6953998094083e-05, + "loss": 1.1275, + "step": 151090 + }, + { + "epoch": 13.08, + "learning_rate": 3.6953131768171187e-05, + "loss": 1.07, + "step": 151100 + }, + { + "epoch": 13.08, + "learning_rate": 3.695226544225938e-05, + "loss": 1.1136, + "step": 151110 + }, + { + "epoch": 13.08, + "learning_rate": 3.6951399116347575e-05, + "loss": 1.0678, + "step": 151120 + }, + { + "epoch": 13.08, + "learning_rate": 3.695053279043576e-05, + "loss": 1.1711, + "step": 151130 + }, + { + "epoch": 13.08, + "learning_rate": 3.694966646452396e-05, + "loss": 1.0869, + "step": 151140 + }, + { + "epoch": 13.08, + "learning_rate": 3.694880013861215e-05, + "loss": 1.0783, + "step": 151150 + }, + { + "epoch": 13.08, + "learning_rate": 3.694793381270034e-05, + "loss": 1.0567, + "step": 151160 + }, + { + "epoch": 13.08, + "learning_rate": 3.6947067486788534e-05, + "loss": 1.1042, + "step": 151170 + }, + { + "epoch": 13.09, + "learning_rate": 3.694620116087673e-05, + "loss": 1.0855, + "step": 151180 + }, + { + "epoch": 13.09, + "learning_rate": 3.6945334834964916e-05, + "loss": 1.0674, + "step": 151190 + }, + { + "epoch": 13.09, + "learning_rate": 3.694446850905311e-05, + "loss": 1.0938, + "step": 151200 + }, + { + "epoch": 13.09, + "learning_rate": 3.69436021831413e-05, + "loss": 1.098, + "step": 151210 + }, + { + "epoch": 13.09, + "learning_rate": 3.694273585722949e-05, + "loss": 1.1001, + "step": 151220 + }, + { + "epoch": 13.09, + "learning_rate": 3.694186953131769e-05, + "loss": 1.1247, + "step": 151230 + }, + { + "epoch": 13.09, + "learning_rate": 3.6941003205405875e-05, + "loss": 1.1013, + "step": 151240 + }, + { + "epoch": 13.09, + "learning_rate": 3.694013687949407e-05, + "loss": 1.1309, + "step": 151250 + }, + { + "epoch": 13.09, + "learning_rate": 3.6939270553582264e-05, + "loss": 1.0243, + "step": 151260 + }, + { + "epoch": 13.09, + "learning_rate": 3.693840422767045e-05, + "loss": 1.0933, + "step": 151270 + }, + { + "epoch": 13.09, + "learning_rate": 3.6937537901758646e-05, + "loss": 1.0562, + "step": 151280 + }, + { + "epoch": 13.09, + "learning_rate": 3.693667157584684e-05, + "loss": 1.1086, + "step": 151290 + }, + { + "epoch": 13.1, + "learning_rate": 3.693580524993503e-05, + "loss": 1.1048, + "step": 151300 + }, + { + "epoch": 13.1, + "learning_rate": 3.693493892402322e-05, + "loss": 1.108, + "step": 151310 + }, + { + "epoch": 13.1, + "learning_rate": 3.693407259811141e-05, + "loss": 1.1174, + "step": 151320 + }, + { + "epoch": 13.1, + "learning_rate": 3.6933206272199605e-05, + "loss": 1.131, + "step": 151330 + }, + { + "epoch": 13.1, + "learning_rate": 3.69323399462878e-05, + "loss": 1.1188, + "step": 151340 + }, + { + "epoch": 13.1, + "learning_rate": 3.693147362037599e-05, + "loss": 1.0826, + "step": 151350 + }, + { + "epoch": 13.1, + "learning_rate": 3.693060729446418e-05, + "loss": 1.0669, + "step": 151360 + }, + { + "epoch": 13.1, + "learning_rate": 3.6929740968552376e-05, + "loss": 1.0933, + "step": 151370 + }, + { + "epoch": 13.1, + "learning_rate": 3.6928874642640563e-05, + "loss": 1.0338, + "step": 151380 + }, + { + "epoch": 13.1, + "learning_rate": 3.692800831672876e-05, + "loss": 1.0776, + "step": 151390 + }, + { + "epoch": 13.1, + "learning_rate": 3.692714199081695e-05, + "loss": 1.1561, + "step": 151400 + }, + { + "epoch": 13.11, + "learning_rate": 3.692627566490514e-05, + "loss": 1.1213, + "step": 151410 + }, + { + "epoch": 13.11, + "learning_rate": 3.6925409338993335e-05, + "loss": 1.0601, + "step": 151420 + }, + { + "epoch": 13.11, + "learning_rate": 3.692454301308152e-05, + "loss": 1.0665, + "step": 151430 + }, + { + "epoch": 13.11, + "learning_rate": 3.692367668716972e-05, + "loss": 1.0399, + "step": 151440 + }, + { + "epoch": 13.11, + "learning_rate": 3.692281036125791e-05, + "loss": 1.0837, + "step": 151450 + }, + { + "epoch": 13.11, + "learning_rate": 3.69219440353461e-05, + "loss": 1.0774, + "step": 151460 + }, + { + "epoch": 13.11, + "learning_rate": 3.692107770943429e-05, + "loss": 1.0652, + "step": 151470 + }, + { + "epoch": 13.11, + "learning_rate": 3.692021138352249e-05, + "loss": 1.1077, + "step": 151480 + }, + { + "epoch": 13.11, + "learning_rate": 3.6919345057610675e-05, + "loss": 1.102, + "step": 151490 + }, + { + "epoch": 13.11, + "learning_rate": 3.691847873169887e-05, + "loss": 1.1598, + "step": 151500 + }, + { + "epoch": 13.11, + "learning_rate": 3.6917612405787064e-05, + "loss": 1.0924, + "step": 151510 + }, + { + "epoch": 13.11, + "learning_rate": 3.691674607987525e-05, + "loss": 1.0571, + "step": 151520 + }, + { + "epoch": 13.12, + "learning_rate": 3.6915879753963446e-05, + "loss": 1.1066, + "step": 151530 + }, + { + "epoch": 13.12, + "learning_rate": 3.6915013428051634e-05, + "loss": 1.1063, + "step": 151540 + }, + { + "epoch": 13.12, + "learning_rate": 3.691414710213983e-05, + "loss": 1.1307, + "step": 151550 + }, + { + "epoch": 13.12, + "learning_rate": 3.691328077622802e-05, + "loss": 1.1018, + "step": 151560 + }, + { + "epoch": 13.12, + "learning_rate": 3.691241445031621e-05, + "loss": 1.1587, + "step": 151570 + }, + { + "epoch": 13.12, + "learning_rate": 3.6911548124404405e-05, + "loss": 1.1627, + "step": 151580 + }, + { + "epoch": 13.12, + "learning_rate": 3.69106817984926e-05, + "loss": 1.1171, + "step": 151590 + }, + { + "epoch": 13.12, + "learning_rate": 3.690981547258079e-05, + "loss": 1.1103, + "step": 151600 + }, + { + "epoch": 13.12, + "learning_rate": 3.690894914666898e-05, + "loss": 1.0284, + "step": 151610 + }, + { + "epoch": 13.12, + "learning_rate": 3.690808282075717e-05, + "loss": 1.0775, + "step": 151620 + }, + { + "epoch": 13.12, + "learning_rate": 3.6907216494845364e-05, + "loss": 1.1588, + "step": 151630 + }, + { + "epoch": 13.13, + "learning_rate": 3.690635016893356e-05, + "loss": 1.1227, + "step": 151640 + }, + { + "epoch": 13.13, + "learning_rate": 3.6905483843021746e-05, + "loss": 1.0924, + "step": 151650 + }, + { + "epoch": 13.13, + "learning_rate": 3.690461751710994e-05, + "loss": 1.0815, + "step": 151660 + }, + { + "epoch": 13.13, + "learning_rate": 3.6903751191198135e-05, + "loss": 1.0301, + "step": 151670 + }, + { + "epoch": 13.13, + "learning_rate": 3.690288486528632e-05, + "loss": 1.0891, + "step": 151680 + }, + { + "epoch": 13.13, + "learning_rate": 3.690201853937452e-05, + "loss": 1.0526, + "step": 151690 + }, + { + "epoch": 13.13, + "learning_rate": 3.690115221346271e-05, + "loss": 0.9817, + "step": 151700 + }, + { + "epoch": 13.13, + "learning_rate": 3.69002858875509e-05, + "loss": 1.1687, + "step": 151710 + }, + { + "epoch": 13.13, + "learning_rate": 3.6899419561639094e-05, + "loss": 1.1366, + "step": 151720 + }, + { + "epoch": 13.13, + "learning_rate": 3.689855323572728e-05, + "loss": 1.1185, + "step": 151730 + }, + { + "epoch": 13.13, + "learning_rate": 3.6897686909815476e-05, + "loss": 1.1831, + "step": 151740 + }, + { + "epoch": 13.13, + "learning_rate": 3.689682058390367e-05, + "loss": 1.1144, + "step": 151750 + }, + { + "epoch": 13.14, + "learning_rate": 3.689595425799186e-05, + "loss": 1.1482, + "step": 151760 + }, + { + "epoch": 13.14, + "learning_rate": 3.689508793208005e-05, + "loss": 1.1659, + "step": 151770 + }, + { + "epoch": 13.14, + "learning_rate": 3.689422160616825e-05, + "loss": 1.0952, + "step": 151780 + }, + { + "epoch": 13.14, + "learning_rate": 3.6893355280256434e-05, + "loss": 1.1305, + "step": 151790 + }, + { + "epoch": 13.14, + "learning_rate": 3.689248895434463e-05, + "loss": 1.1012, + "step": 151800 + }, + { + "epoch": 13.14, + "learning_rate": 3.689162262843282e-05, + "loss": 1.0718, + "step": 151810 + }, + { + "epoch": 13.14, + "learning_rate": 3.689075630252101e-05, + "loss": 1.0231, + "step": 151820 + }, + { + "epoch": 13.14, + "learning_rate": 3.6889889976609205e-05, + "loss": 1.0832, + "step": 151830 + }, + { + "epoch": 13.14, + "learning_rate": 3.688902365069739e-05, + "loss": 1.0697, + "step": 151840 + }, + { + "epoch": 13.14, + "learning_rate": 3.688815732478559e-05, + "loss": 1.1061, + "step": 151850 + }, + { + "epoch": 13.14, + "learning_rate": 3.688729099887378e-05, + "loss": 1.0748, + "step": 151860 + }, + { + "epoch": 13.15, + "learning_rate": 3.688642467296197e-05, + "loss": 1.093, + "step": 151870 + }, + { + "epoch": 13.15, + "learning_rate": 3.6885558347050164e-05, + "loss": 1.0421, + "step": 151880 + }, + { + "epoch": 13.15, + "learning_rate": 3.688469202113836e-05, + "loss": 1.0784, + "step": 151890 + }, + { + "epoch": 13.15, + "learning_rate": 3.6883825695226546e-05, + "loss": 1.1087, + "step": 151900 + }, + { + "epoch": 13.15, + "learning_rate": 3.688295936931474e-05, + "loss": 1.1351, + "step": 151910 + }, + { + "epoch": 13.15, + "learning_rate": 3.6882093043402935e-05, + "loss": 1.1223, + "step": 151920 + }, + { + "epoch": 13.15, + "learning_rate": 3.688122671749112e-05, + "loss": 1.1067, + "step": 151930 + }, + { + "epoch": 13.15, + "learning_rate": 3.688036039157932e-05, + "loss": 1.0853, + "step": 151940 + }, + { + "epoch": 13.15, + "learning_rate": 3.6879494065667505e-05, + "loss": 1.1562, + "step": 151950 + }, + { + "epoch": 13.15, + "learning_rate": 3.68786277397557e-05, + "loss": 1.0948, + "step": 151960 + }, + { + "epoch": 13.15, + "learning_rate": 3.6877761413843894e-05, + "loss": 1.0376, + "step": 151970 + }, + { + "epoch": 13.15, + "learning_rate": 3.687689508793208e-05, + "loss": 1.1626, + "step": 151980 + }, + { + "epoch": 13.16, + "learning_rate": 3.6876028762020276e-05, + "loss": 1.1836, + "step": 151990 + }, + { + "epoch": 13.16, + "learning_rate": 3.687516243610847e-05, + "loss": 1.0927, + "step": 152000 + }, + { + "epoch": 13.16, + "learning_rate": 3.687429611019666e-05, + "loss": 1.1071, + "step": 152010 + }, + { + "epoch": 13.16, + "learning_rate": 3.687342978428485e-05, + "loss": 1.1223, + "step": 152020 + }, + { + "epoch": 13.16, + "learning_rate": 3.687256345837305e-05, + "loss": 1.0958, + "step": 152030 + }, + { + "epoch": 13.16, + "learning_rate": 3.6871697132461235e-05, + "loss": 1.0856, + "step": 152040 + }, + { + "epoch": 13.16, + "learning_rate": 3.687083080654943e-05, + "loss": 1.0614, + "step": 152050 + }, + { + "epoch": 13.16, + "learning_rate": 3.686996448063762e-05, + "loss": 1.0324, + "step": 152060 + }, + { + "epoch": 13.16, + "learning_rate": 3.686909815472581e-05, + "loss": 1.063, + "step": 152070 + }, + { + "epoch": 13.16, + "learning_rate": 3.6868231828814006e-05, + "loss": 1.1363, + "step": 152080 + }, + { + "epoch": 13.16, + "learning_rate": 3.6867365502902193e-05, + "loss": 1.1035, + "step": 152090 + }, + { + "epoch": 13.17, + "learning_rate": 3.686649917699039e-05, + "loss": 1.0424, + "step": 152100 + }, + { + "epoch": 13.17, + "learning_rate": 3.686563285107858e-05, + "loss": 1.1217, + "step": 152110 + }, + { + "epoch": 13.17, + "learning_rate": 3.686476652516677e-05, + "loss": 1.0847, + "step": 152120 + }, + { + "epoch": 13.17, + "learning_rate": 3.6863900199254964e-05, + "loss": 1.081, + "step": 152130 + }, + { + "epoch": 13.17, + "learning_rate": 3.686303387334316e-05, + "loss": 1.0918, + "step": 152140 + }, + { + "epoch": 13.17, + "learning_rate": 3.6862167547431347e-05, + "loss": 1.1626, + "step": 152150 + }, + { + "epoch": 13.17, + "learning_rate": 3.686130122151954e-05, + "loss": 1.1502, + "step": 152160 + }, + { + "epoch": 13.17, + "learning_rate": 3.686043489560773e-05, + "loss": 1.1182, + "step": 152170 + }, + { + "epoch": 13.17, + "learning_rate": 3.685956856969592e-05, + "loss": 1.1033, + "step": 152180 + }, + { + "epoch": 13.17, + "learning_rate": 3.685870224378412e-05, + "loss": 1.1147, + "step": 152190 + }, + { + "epoch": 13.17, + "learning_rate": 3.6857835917872305e-05, + "loss": 1.137, + "step": 152200 + }, + { + "epoch": 13.17, + "learning_rate": 3.68569695919605e-05, + "loss": 1.0879, + "step": 152210 + }, + { + "epoch": 13.18, + "learning_rate": 3.6856103266048694e-05, + "loss": 1.0699, + "step": 152220 + }, + { + "epoch": 13.18, + "learning_rate": 3.685523694013688e-05, + "loss": 1.1003, + "step": 152230 + }, + { + "epoch": 13.18, + "learning_rate": 3.6854370614225076e-05, + "loss": 1.111, + "step": 152240 + }, + { + "epoch": 13.18, + "learning_rate": 3.685350428831327e-05, + "loss": 1.1054, + "step": 152250 + }, + { + "epoch": 13.18, + "learning_rate": 3.685263796240146e-05, + "loss": 1.098, + "step": 152260 + }, + { + "epoch": 13.18, + "learning_rate": 3.685177163648965e-05, + "loss": 1.0551, + "step": 152270 + }, + { + "epoch": 13.18, + "learning_rate": 3.685090531057784e-05, + "loss": 1.1404, + "step": 152280 + }, + { + "epoch": 13.18, + "learning_rate": 3.6850038984666035e-05, + "loss": 1.1214, + "step": 152290 + }, + { + "epoch": 13.18, + "learning_rate": 3.684917265875423e-05, + "loss": 1.0637, + "step": 152300 + }, + { + "epoch": 13.18, + "learning_rate": 3.684830633284242e-05, + "loss": 1.0891, + "step": 152310 + }, + { + "epoch": 13.18, + "learning_rate": 3.684744000693061e-05, + "loss": 1.0829, + "step": 152320 + }, + { + "epoch": 13.19, + "learning_rate": 3.6846573681018806e-05, + "loss": 1.1116, + "step": 152330 + }, + { + "epoch": 13.19, + "learning_rate": 3.6845707355106994e-05, + "loss": 1.1037, + "step": 152340 + }, + { + "epoch": 13.19, + "learning_rate": 3.684484102919519e-05, + "loss": 1.1337, + "step": 152350 + }, + { + "epoch": 13.19, + "learning_rate": 3.684397470328338e-05, + "loss": 1.0453, + "step": 152360 + }, + { + "epoch": 13.19, + "learning_rate": 3.684310837737157e-05, + "loss": 1.1548, + "step": 152370 + }, + { + "epoch": 13.19, + "learning_rate": 3.6842242051459765e-05, + "loss": 1.0775, + "step": 152380 + }, + { + "epoch": 13.19, + "learning_rate": 3.684137572554795e-05, + "loss": 1.0019, + "step": 152390 + }, + { + "epoch": 13.19, + "learning_rate": 3.684050939963615e-05, + "loss": 1.0641, + "step": 152400 + }, + { + "epoch": 13.19, + "learning_rate": 3.683964307372434e-05, + "loss": 1.1006, + "step": 152410 + }, + { + "epoch": 13.19, + "learning_rate": 3.683877674781253e-05, + "loss": 1.0585, + "step": 152420 + }, + { + "epoch": 13.19, + "learning_rate": 3.6837910421900724e-05, + "loss": 1.1075, + "step": 152430 + }, + { + "epoch": 13.19, + "learning_rate": 3.683704409598892e-05, + "loss": 1.0582, + "step": 152440 + }, + { + "epoch": 13.2, + "learning_rate": 3.6836177770077106e-05, + "loss": 1.0583, + "step": 152450 + }, + { + "epoch": 13.2, + "learning_rate": 3.68353114441653e-05, + "loss": 1.1227, + "step": 152460 + }, + { + "epoch": 13.2, + "learning_rate": 3.683444511825349e-05, + "loss": 1.0794, + "step": 152470 + }, + { + "epoch": 13.2, + "learning_rate": 3.683357879234168e-05, + "loss": 1.1797, + "step": 152480 + }, + { + "epoch": 13.2, + "learning_rate": 3.683271246642988e-05, + "loss": 1.0963, + "step": 152490 + }, + { + "epoch": 13.2, + "learning_rate": 3.6831846140518064e-05, + "loss": 1.087, + "step": 152500 + }, + { + "epoch": 13.2, + "learning_rate": 3.683097981460626e-05, + "loss": 1.0834, + "step": 152510 + }, + { + "epoch": 13.2, + "learning_rate": 3.683011348869445e-05, + "loss": 1.1273, + "step": 152520 + }, + { + "epoch": 13.2, + "learning_rate": 3.682924716278264e-05, + "loss": 1.1373, + "step": 152530 + }, + { + "epoch": 13.2, + "learning_rate": 3.6828380836870835e-05, + "loss": 1.0455, + "step": 152540 + }, + { + "epoch": 13.2, + "learning_rate": 3.682751451095903e-05, + "loss": 1.0955, + "step": 152550 + }, + { + "epoch": 13.2, + "learning_rate": 3.682664818504722e-05, + "loss": 1.1223, + "step": 152560 + }, + { + "epoch": 13.21, + "learning_rate": 3.682578185913541e-05, + "loss": 1.1181, + "step": 152570 + }, + { + "epoch": 13.21, + "learning_rate": 3.68249155332236e-05, + "loss": 1.087, + "step": 152580 + }, + { + "epoch": 13.21, + "learning_rate": 3.6824049207311794e-05, + "loss": 1.0102, + "step": 152590 + }, + { + "epoch": 13.21, + "learning_rate": 3.682318288139999e-05, + "loss": 1.0823, + "step": 152600 + }, + { + "epoch": 13.21, + "learning_rate": 3.6822316555488176e-05, + "loss": 1.1207, + "step": 152610 + }, + { + "epoch": 13.21, + "learning_rate": 3.682145022957637e-05, + "loss": 1.1016, + "step": 152620 + }, + { + "epoch": 13.21, + "learning_rate": 3.6820583903664565e-05, + "loss": 1.1055, + "step": 152630 + }, + { + "epoch": 13.21, + "learning_rate": 3.681971757775275e-05, + "loss": 1.1119, + "step": 152640 + }, + { + "epoch": 13.21, + "learning_rate": 3.681885125184095e-05, + "loss": 1.1816, + "step": 152650 + }, + { + "epoch": 13.21, + "learning_rate": 3.681798492592914e-05, + "loss": 1.0841, + "step": 152660 + }, + { + "epoch": 13.21, + "learning_rate": 3.681711860001733e-05, + "loss": 1.1156, + "step": 152670 + }, + { + "epoch": 13.22, + "learning_rate": 3.6816252274105524e-05, + "loss": 1.093, + "step": 152680 + }, + { + "epoch": 13.22, + "learning_rate": 3.681538594819371e-05, + "loss": 1.0787, + "step": 152690 + }, + { + "epoch": 13.22, + "learning_rate": 3.6814519622281906e-05, + "loss": 1.0676, + "step": 152700 + }, + { + "epoch": 13.22, + "learning_rate": 3.68136532963701e-05, + "loss": 1.0953, + "step": 152710 + }, + { + "epoch": 13.22, + "learning_rate": 3.681278697045829e-05, + "loss": 1.1024, + "step": 152720 + }, + { + "epoch": 13.22, + "learning_rate": 3.681192064454648e-05, + "loss": 1.0563, + "step": 152730 + }, + { + "epoch": 13.22, + "learning_rate": 3.681105431863468e-05, + "loss": 1.1081, + "step": 152740 + }, + { + "epoch": 13.22, + "learning_rate": 3.6810187992722865e-05, + "loss": 1.103, + "step": 152750 + }, + { + "epoch": 13.22, + "learning_rate": 3.680932166681106e-05, + "loss": 1.1061, + "step": 152760 + }, + { + "epoch": 13.22, + "learning_rate": 3.6808455340899254e-05, + "loss": 1.0896, + "step": 152770 + }, + { + "epoch": 13.22, + "learning_rate": 3.680758901498744e-05, + "loss": 1.0494, + "step": 152780 + }, + { + "epoch": 13.22, + "learning_rate": 3.6806722689075636e-05, + "loss": 1.1107, + "step": 152790 + }, + { + "epoch": 13.23, + "learning_rate": 3.6805856363163823e-05, + "loss": 1.12, + "step": 152800 + }, + { + "epoch": 13.23, + "learning_rate": 3.680499003725202e-05, + "loss": 1.1215, + "step": 152810 + }, + { + "epoch": 13.23, + "learning_rate": 3.680412371134021e-05, + "loss": 1.0307, + "step": 152820 + }, + { + "epoch": 13.23, + "learning_rate": 3.68032573854284e-05, + "loss": 1.0706, + "step": 152830 + }, + { + "epoch": 13.23, + "learning_rate": 3.6802391059516594e-05, + "loss": 1.0979, + "step": 152840 + }, + { + "epoch": 13.23, + "learning_rate": 3.680152473360479e-05, + "loss": 1.0899, + "step": 152850 + }, + { + "epoch": 13.23, + "learning_rate": 3.6800658407692977e-05, + "loss": 1.129, + "step": 152860 + }, + { + "epoch": 13.23, + "learning_rate": 3.679979208178117e-05, + "loss": 1.1004, + "step": 152870 + }, + { + "epoch": 13.23, + "learning_rate": 3.6798925755869365e-05, + "loss": 1.1138, + "step": 152880 + }, + { + "epoch": 13.23, + "learning_rate": 3.679805942995755e-05, + "loss": 1.1458, + "step": 152890 + }, + { + "epoch": 13.23, + "learning_rate": 3.679719310404575e-05, + "loss": 1.0386, + "step": 152900 + }, + { + "epoch": 13.24, + "learning_rate": 3.6796326778133935e-05, + "loss": 1.1347, + "step": 152910 + }, + { + "epoch": 13.24, + "learning_rate": 3.679546045222213e-05, + "loss": 1.1093, + "step": 152920 + }, + { + "epoch": 13.24, + "learning_rate": 3.6794594126310324e-05, + "loss": 1.1399, + "step": 152930 + }, + { + "epoch": 13.24, + "learning_rate": 3.679372780039851e-05, + "loss": 1.0313, + "step": 152940 + }, + { + "epoch": 13.24, + "learning_rate": 3.6792861474486706e-05, + "loss": 1.1353, + "step": 152950 + }, + { + "epoch": 13.24, + "learning_rate": 3.67919951485749e-05, + "loss": 1.0891, + "step": 152960 + }, + { + "epoch": 13.24, + "learning_rate": 3.679112882266309e-05, + "loss": 1.0991, + "step": 152970 + }, + { + "epoch": 13.24, + "learning_rate": 3.679026249675128e-05, + "loss": 1.1289, + "step": 152980 + }, + { + "epoch": 13.24, + "learning_rate": 3.678939617083948e-05, + "loss": 1.1001, + "step": 152990 + }, + { + "epoch": 13.24, + "learning_rate": 3.6788529844927665e-05, + "loss": 1.1338, + "step": 153000 + }, + { + "epoch": 13.24, + "learning_rate": 3.678766351901586e-05, + "loss": 1.1219, + "step": 153010 + }, + { + "epoch": 13.24, + "learning_rate": 3.678679719310405e-05, + "loss": 1.0985, + "step": 153020 + }, + { + "epoch": 13.25, + "learning_rate": 3.678593086719224e-05, + "loss": 1.1426, + "step": 153030 + }, + { + "epoch": 13.25, + "learning_rate": 3.6785064541280436e-05, + "loss": 1.0927, + "step": 153040 + }, + { + "epoch": 13.25, + "learning_rate": 3.6784198215368624e-05, + "loss": 1.1761, + "step": 153050 + }, + { + "epoch": 13.25, + "learning_rate": 3.678333188945682e-05, + "loss": 1.052, + "step": 153060 + }, + { + "epoch": 13.25, + "learning_rate": 3.678246556354501e-05, + "loss": 1.0636, + "step": 153070 + }, + { + "epoch": 13.25, + "learning_rate": 3.67815992376332e-05, + "loss": 1.0461, + "step": 153080 + }, + { + "epoch": 13.25, + "learning_rate": 3.6780732911721395e-05, + "loss": 1.058, + "step": 153090 + }, + { + "epoch": 13.25, + "learning_rate": 3.677986658580959e-05, + "loss": 1.0886, + "step": 153100 + }, + { + "epoch": 13.25, + "learning_rate": 3.677900025989778e-05, + "loss": 1.0479, + "step": 153110 + }, + { + "epoch": 13.25, + "learning_rate": 3.677813393398597e-05, + "loss": 1.1061, + "step": 153120 + }, + { + "epoch": 13.25, + "learning_rate": 3.677726760807416e-05, + "loss": 1.0913, + "step": 153130 + }, + { + "epoch": 13.26, + "learning_rate": 3.6776401282162353e-05, + "loss": 1.0858, + "step": 153140 + }, + { + "epoch": 13.26, + "learning_rate": 3.677553495625055e-05, + "loss": 1.0911, + "step": 153150 + }, + { + "epoch": 13.26, + "learning_rate": 3.6774668630338736e-05, + "loss": 1.0849, + "step": 153160 + }, + { + "epoch": 13.26, + "learning_rate": 3.677380230442693e-05, + "loss": 1.1313, + "step": 153170 + }, + { + "epoch": 13.26, + "learning_rate": 3.6772935978515125e-05, + "loss": 1.1283, + "step": 153180 + }, + { + "epoch": 13.26, + "learning_rate": 3.677206965260331e-05, + "loss": 1.0847, + "step": 153190 + }, + { + "epoch": 13.26, + "learning_rate": 3.677120332669151e-05, + "loss": 1.0891, + "step": 153200 + }, + { + "epoch": 13.26, + "learning_rate": 3.6770337000779694e-05, + "loss": 1.1266, + "step": 153210 + }, + { + "epoch": 13.26, + "learning_rate": 3.676947067486789e-05, + "loss": 1.1016, + "step": 153220 + }, + { + "epoch": 13.26, + "learning_rate": 3.676860434895608e-05, + "loss": 1.1267, + "step": 153230 + }, + { + "epoch": 13.26, + "learning_rate": 3.676773802304427e-05, + "loss": 1.1045, + "step": 153240 + }, + { + "epoch": 13.26, + "learning_rate": 3.6766871697132465e-05, + "loss": 1.1128, + "step": 153250 + }, + { + "epoch": 13.27, + "learning_rate": 3.676600537122066e-05, + "loss": 1.1022, + "step": 153260 + }, + { + "epoch": 13.27, + "learning_rate": 3.676513904530885e-05, + "loss": 1.0801, + "step": 153270 + }, + { + "epoch": 13.27, + "learning_rate": 3.676427271939704e-05, + "loss": 1.0976, + "step": 153280 + }, + { + "epoch": 13.27, + "learning_rate": 3.6763406393485236e-05, + "loss": 1.0862, + "step": 153290 + }, + { + "epoch": 13.27, + "learning_rate": 3.6762540067573424e-05, + "loss": 1.0898, + "step": 153300 + }, + { + "epoch": 13.27, + "learning_rate": 3.676167374166162e-05, + "loss": 1.1327, + "step": 153310 + }, + { + "epoch": 13.27, + "learning_rate": 3.6760807415749806e-05, + "loss": 1.0656, + "step": 153320 + }, + { + "epoch": 13.27, + "learning_rate": 3.6759941089838e-05, + "loss": 1.1101, + "step": 153330 + }, + { + "epoch": 13.27, + "learning_rate": 3.6759074763926195e-05, + "loss": 1.111, + "step": 153340 + }, + { + "epoch": 13.27, + "learning_rate": 3.675820843801438e-05, + "loss": 1.0941, + "step": 153350 + }, + { + "epoch": 13.27, + "learning_rate": 3.675734211210258e-05, + "loss": 1.0722, + "step": 153360 + }, + { + "epoch": 13.28, + "learning_rate": 3.675647578619077e-05, + "loss": 1.1438, + "step": 153370 + }, + { + "epoch": 13.28, + "learning_rate": 3.675560946027896e-05, + "loss": 1.1317, + "step": 153380 + }, + { + "epoch": 13.28, + "learning_rate": 3.6754743134367154e-05, + "loss": 1.0729, + "step": 153390 + }, + { + "epoch": 13.28, + "learning_rate": 3.675387680845535e-05, + "loss": 1.1039, + "step": 153400 + }, + { + "epoch": 13.28, + "learning_rate": 3.6753010482543536e-05, + "loss": 1.061, + "step": 153410 + }, + { + "epoch": 13.28, + "learning_rate": 3.675214415663173e-05, + "loss": 1.0877, + "step": 153420 + }, + { + "epoch": 13.28, + "learning_rate": 3.675127783071992e-05, + "loss": 1.0956, + "step": 153430 + }, + { + "epoch": 13.28, + "learning_rate": 3.675041150480811e-05, + "loss": 1.0989, + "step": 153440 + }, + { + "epoch": 13.28, + "learning_rate": 3.674954517889631e-05, + "loss": 1.1171, + "step": 153450 + }, + { + "epoch": 13.28, + "learning_rate": 3.6748678852984495e-05, + "loss": 1.1166, + "step": 153460 + }, + { + "epoch": 13.28, + "learning_rate": 3.674781252707269e-05, + "loss": 1.0842, + "step": 153470 + }, + { + "epoch": 13.28, + "learning_rate": 3.6746946201160884e-05, + "loss": 1.122, + "step": 153480 + }, + { + "epoch": 13.29, + "learning_rate": 3.674607987524907e-05, + "loss": 1.0567, + "step": 153490 + }, + { + "epoch": 13.29, + "learning_rate": 3.6745213549337266e-05, + "loss": 1.0406, + "step": 153500 + }, + { + "epoch": 13.29, + "learning_rate": 3.674434722342546e-05, + "loss": 1.1327, + "step": 153510 + }, + { + "epoch": 13.29, + "learning_rate": 3.674348089751365e-05, + "loss": 1.0628, + "step": 153520 + }, + { + "epoch": 13.29, + "learning_rate": 3.674261457160184e-05, + "loss": 1.1148, + "step": 153530 + }, + { + "epoch": 13.29, + "learning_rate": 3.674174824569003e-05, + "loss": 1.1311, + "step": 153540 + }, + { + "epoch": 13.29, + "learning_rate": 3.6740881919778224e-05, + "loss": 1.0605, + "step": 153550 + }, + { + "epoch": 13.29, + "learning_rate": 3.674001559386642e-05, + "loss": 1.1042, + "step": 153560 + }, + { + "epoch": 13.29, + "learning_rate": 3.6739149267954607e-05, + "loss": 1.0984, + "step": 153570 + }, + { + "epoch": 13.29, + "learning_rate": 3.67382829420428e-05, + "loss": 1.0417, + "step": 153580 + }, + { + "epoch": 13.29, + "learning_rate": 3.6737416616130995e-05, + "loss": 1.1263, + "step": 153590 + }, + { + "epoch": 13.29, + "learning_rate": 3.673655029021918e-05, + "loss": 1.0473, + "step": 153600 + }, + { + "epoch": 13.3, + "learning_rate": 3.673568396430738e-05, + "loss": 1.1142, + "step": 153610 + }, + { + "epoch": 13.3, + "learning_rate": 3.673481763839557e-05, + "loss": 1.0693, + "step": 153620 + }, + { + "epoch": 13.3, + "learning_rate": 3.673395131248376e-05, + "loss": 1.068, + "step": 153630 + }, + { + "epoch": 13.3, + "learning_rate": 3.6733084986571954e-05, + "loss": 1.1392, + "step": 153640 + }, + { + "epoch": 13.3, + "learning_rate": 3.673221866066014e-05, + "loss": 1.1474, + "step": 153650 + }, + { + "epoch": 13.3, + "learning_rate": 3.6731352334748336e-05, + "loss": 1.1386, + "step": 153660 + }, + { + "epoch": 13.3, + "learning_rate": 3.673048600883653e-05, + "loss": 1.0775, + "step": 153670 + }, + { + "epoch": 13.3, + "learning_rate": 3.672961968292472e-05, + "loss": 1.1061, + "step": 153680 + }, + { + "epoch": 13.3, + "learning_rate": 3.672875335701291e-05, + "loss": 1.0738, + "step": 153690 + }, + { + "epoch": 13.3, + "learning_rate": 3.672788703110111e-05, + "loss": 1.0696, + "step": 153700 + }, + { + "epoch": 13.3, + "learning_rate": 3.6727020705189295e-05, + "loss": 1.0726, + "step": 153710 + }, + { + "epoch": 13.31, + "learning_rate": 3.672615437927749e-05, + "loss": 1.1618, + "step": 153720 + }, + { + "epoch": 13.31, + "learning_rate": 3.6725288053365684e-05, + "loss": 1.1259, + "step": 153730 + }, + { + "epoch": 13.31, + "learning_rate": 3.672442172745387e-05, + "loss": 1.0781, + "step": 153740 + }, + { + "epoch": 13.31, + "learning_rate": 3.6723555401542066e-05, + "loss": 1.1079, + "step": 153750 + }, + { + "epoch": 13.31, + "learning_rate": 3.6722689075630254e-05, + "loss": 1.0409, + "step": 153760 + }, + { + "epoch": 13.31, + "learning_rate": 3.672182274971845e-05, + "loss": 1.1213, + "step": 153770 + }, + { + "epoch": 13.31, + "learning_rate": 3.672095642380664e-05, + "loss": 1.0969, + "step": 153780 + }, + { + "epoch": 13.31, + "learning_rate": 3.672009009789483e-05, + "loss": 1.0664, + "step": 153790 + }, + { + "epoch": 13.31, + "learning_rate": 3.6719223771983025e-05, + "loss": 1.0686, + "step": 153800 + }, + { + "epoch": 13.31, + "learning_rate": 3.671835744607121e-05, + "loss": 1.1434, + "step": 153810 + }, + { + "epoch": 13.31, + "learning_rate": 3.67174911201594e-05, + "loss": 1.1408, + "step": 153820 + }, + { + "epoch": 13.31, + "learning_rate": 3.6716624794247595e-05, + "loss": 1.0905, + "step": 153830 + }, + { + "epoch": 13.32, + "learning_rate": 3.671575846833579e-05, + "loss": 1.1318, + "step": 153840 + }, + { + "epoch": 13.32, + "learning_rate": 3.671489214242398e-05, + "loss": 1.1155, + "step": 153850 + }, + { + "epoch": 13.32, + "learning_rate": 3.671402581651217e-05, + "loss": 1.1121, + "step": 153860 + }, + { + "epoch": 13.32, + "learning_rate": 3.6713159490600366e-05, + "loss": 1.091, + "step": 153870 + }, + { + "epoch": 13.32, + "learning_rate": 3.671229316468855e-05, + "loss": 1.0932, + "step": 153880 + }, + { + "epoch": 13.32, + "learning_rate": 3.671142683877675e-05, + "loss": 1.1974, + "step": 153890 + }, + { + "epoch": 13.32, + "learning_rate": 3.671056051286494e-05, + "loss": 1.0885, + "step": 153900 + }, + { + "epoch": 13.32, + "learning_rate": 3.670969418695313e-05, + "loss": 1.024, + "step": 153910 + }, + { + "epoch": 13.32, + "learning_rate": 3.6708827861041324e-05, + "loss": 1.1551, + "step": 153920 + }, + { + "epoch": 13.32, + "learning_rate": 3.670796153512951e-05, + "loss": 1.0407, + "step": 153930 + }, + { + "epoch": 13.32, + "learning_rate": 3.6707095209217706e-05, + "loss": 1.1172, + "step": 153940 + }, + { + "epoch": 13.33, + "learning_rate": 3.67062288833059e-05, + "loss": 1.1188, + "step": 153950 + }, + { + "epoch": 13.33, + "learning_rate": 3.670536255739409e-05, + "loss": 1.0654, + "step": 153960 + }, + { + "epoch": 13.33, + "learning_rate": 3.670449623148228e-05, + "loss": 1.1053, + "step": 153970 + }, + { + "epoch": 13.33, + "learning_rate": 3.670362990557048e-05, + "loss": 1.0909, + "step": 153980 + }, + { + "epoch": 13.33, + "learning_rate": 3.6702763579658665e-05, + "loss": 1.0928, + "step": 153990 + }, + { + "epoch": 13.33, + "learning_rate": 3.670189725374686e-05, + "loss": 1.125, + "step": 154000 + }, + { + "epoch": 13.33, + "learning_rate": 3.670103092783505e-05, + "loss": 1.0807, + "step": 154010 + }, + { + "epoch": 13.33, + "learning_rate": 3.670016460192324e-05, + "loss": 1.0841, + "step": 154020 + }, + { + "epoch": 13.33, + "learning_rate": 3.6699298276011436e-05, + "loss": 1.1223, + "step": 154030 + }, + { + "epoch": 13.33, + "learning_rate": 3.6698431950099624e-05, + "loss": 1.1225, + "step": 154040 + }, + { + "epoch": 13.33, + "learning_rate": 3.669756562418782e-05, + "loss": 1.0778, + "step": 154050 + }, + { + "epoch": 13.33, + "learning_rate": 3.669669929827601e-05, + "loss": 1.0995, + "step": 154060 + }, + { + "epoch": 13.34, + "learning_rate": 3.66958329723642e-05, + "loss": 1.0861, + "step": 154070 + }, + { + "epoch": 13.34, + "learning_rate": 3.6694966646452395e-05, + "loss": 1.1191, + "step": 154080 + }, + { + "epoch": 13.34, + "learning_rate": 3.669410032054059e-05, + "loss": 1.0753, + "step": 154090 + }, + { + "epoch": 13.34, + "learning_rate": 3.669323399462878e-05, + "loss": 1.0821, + "step": 154100 + }, + { + "epoch": 13.34, + "learning_rate": 3.669236766871697e-05, + "loss": 1.0917, + "step": 154110 + }, + { + "epoch": 13.34, + "learning_rate": 3.669150134280516e-05, + "loss": 1.1446, + "step": 154120 + }, + { + "epoch": 13.34, + "learning_rate": 3.6690635016893354e-05, + "loss": 1.1131, + "step": 154130 + }, + { + "epoch": 13.34, + "learning_rate": 3.668976869098155e-05, + "loss": 1.1019, + "step": 154140 + }, + { + "epoch": 13.34, + "learning_rate": 3.6688902365069736e-05, + "loss": 1.1331, + "step": 154150 + }, + { + "epoch": 13.34, + "learning_rate": 3.668803603915793e-05, + "loss": 1.1066, + "step": 154160 + }, + { + "epoch": 13.34, + "learning_rate": 3.6687169713246125e-05, + "loss": 1.0967, + "step": 154170 + }, + { + "epoch": 13.35, + "learning_rate": 3.668630338733431e-05, + "loss": 1.1097, + "step": 154180 + }, + { + "epoch": 13.35, + "learning_rate": 3.668543706142251e-05, + "loss": 1.0858, + "step": 154190 + }, + { + "epoch": 13.35, + "learning_rate": 3.66845707355107e-05, + "loss": 1.1361, + "step": 154200 + }, + { + "epoch": 13.35, + "learning_rate": 3.668370440959889e-05, + "loss": 1.0693, + "step": 154210 + }, + { + "epoch": 13.35, + "learning_rate": 3.668283808368708e-05, + "loss": 1.094, + "step": 154220 + }, + { + "epoch": 13.35, + "learning_rate": 3.668197175777527e-05, + "loss": 1.0529, + "step": 154230 + }, + { + "epoch": 13.35, + "learning_rate": 3.6681105431863465e-05, + "loss": 1.0776, + "step": 154240 + }, + { + "epoch": 13.35, + "learning_rate": 3.668023910595166e-05, + "loss": 1.0585, + "step": 154250 + }, + { + "epoch": 13.35, + "learning_rate": 3.667937278003985e-05, + "loss": 1.0773, + "step": 154260 + }, + { + "epoch": 13.35, + "learning_rate": 3.667850645412804e-05, + "loss": 1.1246, + "step": 154270 + }, + { + "epoch": 13.35, + "learning_rate": 3.6677640128216237e-05, + "loss": 1.1043, + "step": 154280 + }, + { + "epoch": 13.35, + "learning_rate": 3.6676773802304424e-05, + "loss": 1.058, + "step": 154290 + }, + { + "epoch": 13.36, + "learning_rate": 3.667590747639262e-05, + "loss": 1.0663, + "step": 154300 + }, + { + "epoch": 13.36, + "learning_rate": 3.667504115048081e-05, + "loss": 1.0959, + "step": 154310 + }, + { + "epoch": 13.36, + "learning_rate": 3.6674174824569e-05, + "loss": 1.1001, + "step": 154320 + }, + { + "epoch": 13.36, + "learning_rate": 3.6673308498657195e-05, + "loss": 1.1296, + "step": 154330 + }, + { + "epoch": 13.36, + "learning_rate": 3.667244217274538e-05, + "loss": 1.1333, + "step": 154340 + }, + { + "epoch": 13.36, + "learning_rate": 3.667157584683358e-05, + "loss": 1.0802, + "step": 154350 + }, + { + "epoch": 13.36, + "learning_rate": 3.667070952092177e-05, + "loss": 1.1225, + "step": 154360 + }, + { + "epoch": 13.36, + "learning_rate": 3.666984319500996e-05, + "loss": 1.1605, + "step": 154370 + }, + { + "epoch": 13.36, + "learning_rate": 3.6668976869098154e-05, + "loss": 1.1332, + "step": 154380 + }, + { + "epoch": 13.36, + "learning_rate": 3.666811054318635e-05, + "loss": 1.0809, + "step": 154390 + }, + { + "epoch": 13.36, + "learning_rate": 3.6667244217274536e-05, + "loss": 1.0883, + "step": 154400 + }, + { + "epoch": 13.37, + "learning_rate": 3.666637789136273e-05, + "loss": 1.0946, + "step": 154410 + }, + { + "epoch": 13.37, + "learning_rate": 3.6665511565450925e-05, + "loss": 1.1376, + "step": 154420 + }, + { + "epoch": 13.37, + "learning_rate": 3.666464523953911e-05, + "loss": 1.0724, + "step": 154430 + }, + { + "epoch": 13.37, + "learning_rate": 3.666377891362731e-05, + "loss": 1.0661, + "step": 154440 + }, + { + "epoch": 13.37, + "learning_rate": 3.6662912587715495e-05, + "loss": 1.0962, + "step": 154450 + }, + { + "epoch": 13.37, + "learning_rate": 3.666204626180369e-05, + "loss": 1.1224, + "step": 154460 + }, + { + "epoch": 13.37, + "learning_rate": 3.6661179935891884e-05, + "loss": 1.0511, + "step": 154470 + }, + { + "epoch": 13.37, + "learning_rate": 3.666031360998007e-05, + "loss": 1.096, + "step": 154480 + }, + { + "epoch": 13.37, + "learning_rate": 3.6659447284068266e-05, + "loss": 1.1539, + "step": 154490 + }, + { + "epoch": 13.37, + "learning_rate": 3.665858095815646e-05, + "loss": 1.0695, + "step": 154500 + }, + { + "epoch": 13.37, + "learning_rate": 3.665771463224465e-05, + "loss": 1.1041, + "step": 154510 + }, + { + "epoch": 13.37, + "learning_rate": 3.665684830633284e-05, + "loss": 1.1399, + "step": 154520 + }, + { + "epoch": 13.38, + "learning_rate": 3.665598198042104e-05, + "loss": 1.1034, + "step": 154530 + }, + { + "epoch": 13.38, + "learning_rate": 3.6655115654509225e-05, + "loss": 1.0514, + "step": 154540 + }, + { + "epoch": 13.38, + "learning_rate": 3.665424932859742e-05, + "loss": 1.1173, + "step": 154550 + }, + { + "epoch": 13.38, + "learning_rate": 3.665338300268561e-05, + "loss": 1.0081, + "step": 154560 + }, + { + "epoch": 13.38, + "learning_rate": 3.66525166767738e-05, + "loss": 1.0774, + "step": 154570 + }, + { + "epoch": 13.38, + "learning_rate": 3.6651650350861996e-05, + "loss": 1.0852, + "step": 154580 + }, + { + "epoch": 13.38, + "learning_rate": 3.665078402495018e-05, + "loss": 1.0766, + "step": 154590 + }, + { + "epoch": 13.38, + "learning_rate": 3.664991769903838e-05, + "loss": 1.0952, + "step": 154600 + }, + { + "epoch": 13.38, + "learning_rate": 3.664905137312657e-05, + "loss": 1.0684, + "step": 154610 + }, + { + "epoch": 13.38, + "learning_rate": 3.664818504721476e-05, + "loss": 1.1818, + "step": 154620 + }, + { + "epoch": 13.38, + "learning_rate": 3.6647318721302954e-05, + "loss": 1.0391, + "step": 154630 + }, + { + "epoch": 13.38, + "learning_rate": 3.664645239539115e-05, + "loss": 1.0374, + "step": 154640 + }, + { + "epoch": 13.39, + "learning_rate": 3.6645586069479336e-05, + "loss": 1.0764, + "step": 154650 + }, + { + "epoch": 13.39, + "learning_rate": 3.664471974356753e-05, + "loss": 1.0555, + "step": 154660 + }, + { + "epoch": 13.39, + "learning_rate": 3.664385341765572e-05, + "loss": 1.1458, + "step": 154670 + }, + { + "epoch": 13.39, + "learning_rate": 3.664298709174391e-05, + "loss": 1.1428, + "step": 154680 + }, + { + "epoch": 13.39, + "learning_rate": 3.664212076583211e-05, + "loss": 1.0973, + "step": 154690 + }, + { + "epoch": 13.39, + "learning_rate": 3.6641254439920295e-05, + "loss": 1.0989, + "step": 154700 + }, + { + "epoch": 13.39, + "learning_rate": 3.664038811400849e-05, + "loss": 1.0243, + "step": 154710 + }, + { + "epoch": 13.39, + "learning_rate": 3.6639521788096684e-05, + "loss": 1.0845, + "step": 154720 + }, + { + "epoch": 13.39, + "learning_rate": 3.663865546218487e-05, + "loss": 1.1202, + "step": 154730 + }, + { + "epoch": 13.39, + "learning_rate": 3.6637789136273066e-05, + "loss": 1.0798, + "step": 154740 + }, + { + "epoch": 13.39, + "learning_rate": 3.6636922810361254e-05, + "loss": 1.0742, + "step": 154750 + }, + { + "epoch": 13.4, + "learning_rate": 3.663605648444945e-05, + "loss": 1.0486, + "step": 154760 + }, + { + "epoch": 13.4, + "learning_rate": 3.663519015853764e-05, + "loss": 1.0557, + "step": 154770 + }, + { + "epoch": 13.4, + "learning_rate": 3.663432383262583e-05, + "loss": 1.1476, + "step": 154780 + }, + { + "epoch": 13.4, + "learning_rate": 3.6633457506714025e-05, + "loss": 1.1343, + "step": 154790 + }, + { + "epoch": 13.4, + "learning_rate": 3.663259118080222e-05, + "loss": 1.0376, + "step": 154800 + }, + { + "epoch": 13.4, + "learning_rate": 3.663172485489041e-05, + "loss": 1.1325, + "step": 154810 + }, + { + "epoch": 13.4, + "learning_rate": 3.66308585289786e-05, + "loss": 1.0863, + "step": 154820 + }, + { + "epoch": 13.4, + "learning_rate": 3.6629992203066796e-05, + "loss": 1.0831, + "step": 154830 + }, + { + "epoch": 13.4, + "learning_rate": 3.6629125877154984e-05, + "loss": 1.0589, + "step": 154840 + }, + { + "epoch": 13.4, + "learning_rate": 3.662825955124318e-05, + "loss": 1.0918, + "step": 154850 + }, + { + "epoch": 13.4, + "learning_rate": 3.6627393225331366e-05, + "loss": 1.0954, + "step": 154860 + }, + { + "epoch": 13.4, + "learning_rate": 3.662652689941956e-05, + "loss": 1.084, + "step": 154870 + }, + { + "epoch": 13.41, + "learning_rate": 3.6625660573507755e-05, + "loss": 1.1492, + "step": 154880 + }, + { + "epoch": 13.41, + "learning_rate": 3.662479424759594e-05, + "loss": 1.0819, + "step": 154890 + }, + { + "epoch": 13.41, + "learning_rate": 3.662392792168414e-05, + "loss": 1.1082, + "step": 154900 + }, + { + "epoch": 13.41, + "learning_rate": 3.662306159577233e-05, + "loss": 1.0981, + "step": 154910 + }, + { + "epoch": 13.41, + "learning_rate": 3.662219526986052e-05, + "loss": 1.126, + "step": 154920 + }, + { + "epoch": 13.41, + "learning_rate": 3.662132894394871e-05, + "loss": 1.0991, + "step": 154930 + }, + { + "epoch": 13.41, + "learning_rate": 3.662046261803691e-05, + "loss": 1.0948, + "step": 154940 + }, + { + "epoch": 13.41, + "learning_rate": 3.6619596292125095e-05, + "loss": 1.1287, + "step": 154950 + }, + { + "epoch": 13.41, + "learning_rate": 3.661872996621329e-05, + "loss": 1.1499, + "step": 154960 + }, + { + "epoch": 13.41, + "learning_rate": 3.661786364030148e-05, + "loss": 1.1267, + "step": 154970 + }, + { + "epoch": 13.41, + "learning_rate": 3.661699731438967e-05, + "loss": 1.1881, + "step": 154980 + }, + { + "epoch": 13.42, + "learning_rate": 3.6616130988477866e-05, + "loss": 1.1324, + "step": 154990 + }, + { + "epoch": 13.42, + "learning_rate": 3.6615264662566054e-05, + "loss": 1.0707, + "step": 155000 + }, + { + "epoch": 13.42, + "learning_rate": 3.661439833665425e-05, + "loss": 1.0967, + "step": 155010 + }, + { + "epoch": 13.42, + "learning_rate": 3.661353201074244e-05, + "loss": 1.1254, + "step": 155020 + }, + { + "epoch": 13.42, + "learning_rate": 3.661266568483063e-05, + "loss": 1.1299, + "step": 155030 + }, + { + "epoch": 13.42, + "learning_rate": 3.6611799358918825e-05, + "loss": 1.0432, + "step": 155040 + }, + { + "epoch": 13.42, + "learning_rate": 3.661093303300702e-05, + "loss": 1.1248, + "step": 155050 + }, + { + "epoch": 13.42, + "learning_rate": 3.661006670709521e-05, + "loss": 1.1172, + "step": 155060 + }, + { + "epoch": 13.42, + "learning_rate": 3.66092003811834e-05, + "loss": 1.0844, + "step": 155070 + }, + { + "epoch": 13.42, + "learning_rate": 3.660833405527159e-05, + "loss": 1.056, + "step": 155080 + }, + { + "epoch": 13.42, + "learning_rate": 3.6607467729359784e-05, + "loss": 1.0667, + "step": 155090 + }, + { + "epoch": 13.42, + "learning_rate": 3.660660140344798e-05, + "loss": 1.1256, + "step": 155100 + }, + { + "epoch": 13.43, + "learning_rate": 3.6605735077536166e-05, + "loss": 1.1124, + "step": 155110 + }, + { + "epoch": 13.43, + "learning_rate": 3.660486875162436e-05, + "loss": 1.1148, + "step": 155120 + }, + { + "epoch": 13.43, + "learning_rate": 3.6604002425712555e-05, + "loss": 1.143, + "step": 155130 + }, + { + "epoch": 13.43, + "learning_rate": 3.660313609980074e-05, + "loss": 1.0965, + "step": 155140 + }, + { + "epoch": 13.43, + "learning_rate": 3.660226977388894e-05, + "loss": 1.0691, + "step": 155150 + }, + { + "epoch": 13.43, + "learning_rate": 3.660140344797713e-05, + "loss": 1.0529, + "step": 155160 + }, + { + "epoch": 13.43, + "learning_rate": 3.660053712206532e-05, + "loss": 1.1381, + "step": 155170 + }, + { + "epoch": 13.43, + "learning_rate": 3.6599670796153514e-05, + "loss": 1.0885, + "step": 155180 + }, + { + "epoch": 13.43, + "learning_rate": 3.65988044702417e-05, + "loss": 1.1285, + "step": 155190 + }, + { + "epoch": 13.43, + "learning_rate": 3.6597938144329896e-05, + "loss": 1.2129, + "step": 155200 + }, + { + "epoch": 13.43, + "learning_rate": 3.659707181841809e-05, + "loss": 1.1057, + "step": 155210 + }, + { + "epoch": 13.44, + "learning_rate": 3.659620549250628e-05, + "loss": 1.0646, + "step": 155220 + }, + { + "epoch": 13.44, + "learning_rate": 3.659533916659447e-05, + "loss": 1.1069, + "step": 155230 + }, + { + "epoch": 13.44, + "learning_rate": 3.659447284068267e-05, + "loss": 1.1284, + "step": 155240 + }, + { + "epoch": 13.44, + "learning_rate": 3.6593606514770855e-05, + "loss": 1.0434, + "step": 155250 + }, + { + "epoch": 13.44, + "learning_rate": 3.659274018885905e-05, + "loss": 1.0554, + "step": 155260 + }, + { + "epoch": 13.44, + "learning_rate": 3.6591873862947243e-05, + "loss": 1.1353, + "step": 155270 + }, + { + "epoch": 13.44, + "learning_rate": 3.659100753703543e-05, + "loss": 1.076, + "step": 155280 + }, + { + "epoch": 13.44, + "learning_rate": 3.6590141211123626e-05, + "loss": 1.1117, + "step": 155290 + }, + { + "epoch": 13.44, + "learning_rate": 3.658927488521181e-05, + "loss": 1.1223, + "step": 155300 + }, + { + "epoch": 13.44, + "learning_rate": 3.658840855930001e-05, + "loss": 1.0979, + "step": 155310 + }, + { + "epoch": 13.44, + "learning_rate": 3.65875422333882e-05, + "loss": 1.1227, + "step": 155320 + }, + { + "epoch": 13.44, + "learning_rate": 3.658667590747639e-05, + "loss": 1.0329, + "step": 155330 + }, + { + "epoch": 13.45, + "learning_rate": 3.6585809581564584e-05, + "loss": 1.1074, + "step": 155340 + }, + { + "epoch": 13.45, + "learning_rate": 3.658494325565278e-05, + "loss": 1.0866, + "step": 155350 + }, + { + "epoch": 13.45, + "learning_rate": 3.6584076929740966e-05, + "loss": 1.1077, + "step": 155360 + }, + { + "epoch": 13.45, + "learning_rate": 3.658321060382916e-05, + "loss": 1.0769, + "step": 155370 + }, + { + "epoch": 13.45, + "learning_rate": 3.6582344277917355e-05, + "loss": 1.1426, + "step": 155380 + }, + { + "epoch": 13.45, + "learning_rate": 3.658147795200554e-05, + "loss": 1.1212, + "step": 155390 + }, + { + "epoch": 13.45, + "learning_rate": 3.658061162609374e-05, + "loss": 1.092, + "step": 155400 + }, + { + "epoch": 13.45, + "learning_rate": 3.6579745300181925e-05, + "loss": 0.9567, + "step": 155410 + }, + { + "epoch": 13.45, + "learning_rate": 3.657887897427012e-05, + "loss": 1.0818, + "step": 155420 + }, + { + "epoch": 13.45, + "learning_rate": 3.6578012648358314e-05, + "loss": 1.102, + "step": 155430 + }, + { + "epoch": 13.45, + "learning_rate": 3.65771463224465e-05, + "loss": 1.0414, + "step": 155440 + }, + { + "epoch": 13.46, + "learning_rate": 3.6576279996534696e-05, + "loss": 1.0663, + "step": 155450 + }, + { + "epoch": 13.46, + "learning_rate": 3.657541367062289e-05, + "loss": 1.0924, + "step": 155460 + }, + { + "epoch": 13.46, + "learning_rate": 3.657454734471108e-05, + "loss": 1.1343, + "step": 155470 + }, + { + "epoch": 13.46, + "learning_rate": 3.657368101879927e-05, + "loss": 1.0418, + "step": 155480 + }, + { + "epoch": 13.46, + "learning_rate": 3.657281469288746e-05, + "loss": 1.0438, + "step": 155490 + }, + { + "epoch": 13.46, + "learning_rate": 3.6571948366975655e-05, + "loss": 1.0869, + "step": 155500 + }, + { + "epoch": 13.46, + "learning_rate": 3.657108204106385e-05, + "loss": 1.0932, + "step": 155510 + }, + { + "epoch": 13.46, + "learning_rate": 3.657021571515204e-05, + "loss": 1.1134, + "step": 155520 + }, + { + "epoch": 13.46, + "learning_rate": 3.656934938924023e-05, + "loss": 1.0974, + "step": 155530 + }, + { + "epoch": 13.46, + "learning_rate": 3.6568483063328426e-05, + "loss": 1.1065, + "step": 155540 + }, + { + "epoch": 13.46, + "learning_rate": 3.6567616737416614e-05, + "loss": 1.0289, + "step": 155550 + }, + { + "epoch": 13.46, + "learning_rate": 3.656675041150481e-05, + "loss": 1.0502, + "step": 155560 + }, + { + "epoch": 13.47, + "learning_rate": 3.6565884085593e-05, + "loss": 1.0879, + "step": 155570 + }, + { + "epoch": 13.47, + "learning_rate": 3.656501775968119e-05, + "loss": 1.0447, + "step": 155580 + }, + { + "epoch": 13.47, + "learning_rate": 3.6564151433769385e-05, + "loss": 1.1378, + "step": 155590 + }, + { + "epoch": 13.47, + "learning_rate": 3.656328510785757e-05, + "loss": 1.1434, + "step": 155600 + }, + { + "epoch": 13.47, + "learning_rate": 3.656241878194577e-05, + "loss": 1.0527, + "step": 155610 + }, + { + "epoch": 13.47, + "learning_rate": 3.656155245603396e-05, + "loss": 1.139, + "step": 155620 + }, + { + "epoch": 13.47, + "learning_rate": 3.656068613012215e-05, + "loss": 1.1015, + "step": 155630 + }, + { + "epoch": 13.47, + "learning_rate": 3.655981980421034e-05, + "loss": 1.0306, + "step": 155640 + }, + { + "epoch": 13.47, + "learning_rate": 3.655895347829854e-05, + "loss": 1.0928, + "step": 155650 + }, + { + "epoch": 13.47, + "learning_rate": 3.6558087152386725e-05, + "loss": 1.1337, + "step": 155660 + }, + { + "epoch": 13.47, + "learning_rate": 3.655722082647492e-05, + "loss": 1.1899, + "step": 155670 + }, + { + "epoch": 13.47, + "learning_rate": 3.6556354500563114e-05, + "loss": 1.147, + "step": 155680 + }, + { + "epoch": 13.48, + "learning_rate": 3.65554881746513e-05, + "loss": 1.0948, + "step": 155690 + }, + { + "epoch": 13.48, + "learning_rate": 3.6554621848739496e-05, + "loss": 1.0813, + "step": 155700 + }, + { + "epoch": 13.48, + "learning_rate": 3.6553755522827684e-05, + "loss": 1.1446, + "step": 155710 + }, + { + "epoch": 13.48, + "learning_rate": 3.655288919691588e-05, + "loss": 1.1543, + "step": 155720 + }, + { + "epoch": 13.48, + "learning_rate": 3.655202287100407e-05, + "loss": 1.1532, + "step": 155730 + }, + { + "epoch": 13.48, + "learning_rate": 3.655115654509226e-05, + "loss": 1.1266, + "step": 155740 + }, + { + "epoch": 13.48, + "learning_rate": 3.6550290219180455e-05, + "loss": 1.1641, + "step": 155750 + }, + { + "epoch": 13.48, + "learning_rate": 3.654942389326865e-05, + "loss": 1.1353, + "step": 155760 + }, + { + "epoch": 13.48, + "learning_rate": 3.654855756735684e-05, + "loss": 1.0394, + "step": 155770 + }, + { + "epoch": 13.48, + "learning_rate": 3.654769124144503e-05, + "loss": 1.0743, + "step": 155780 + }, + { + "epoch": 13.48, + "learning_rate": 3.6546824915533226e-05, + "loss": 1.1, + "step": 155790 + }, + { + "epoch": 13.49, + "learning_rate": 3.6545958589621414e-05, + "loss": 1.1102, + "step": 155800 + }, + { + "epoch": 13.49, + "learning_rate": 3.654509226370961e-05, + "loss": 1.1358, + "step": 155810 + }, + { + "epoch": 13.49, + "learning_rate": 3.6544225937797796e-05, + "loss": 1.056, + "step": 155820 + }, + { + "epoch": 13.49, + "learning_rate": 3.654335961188599e-05, + "loss": 1.1388, + "step": 155830 + }, + { + "epoch": 13.49, + "learning_rate": 3.6542493285974185e-05, + "loss": 1.1627, + "step": 155840 + }, + { + "epoch": 13.49, + "learning_rate": 3.654162696006237e-05, + "loss": 1.0576, + "step": 155850 + }, + { + "epoch": 13.49, + "learning_rate": 3.654076063415057e-05, + "loss": 1.1295, + "step": 155860 + }, + { + "epoch": 13.49, + "learning_rate": 3.653989430823876e-05, + "loss": 1.1079, + "step": 155870 + }, + { + "epoch": 13.49, + "learning_rate": 3.653902798232695e-05, + "loss": 1.0375, + "step": 155880 + }, + { + "epoch": 13.49, + "learning_rate": 3.6538161656415144e-05, + "loss": 1.116, + "step": 155890 + }, + { + "epoch": 13.49, + "learning_rate": 3.653729533050334e-05, + "loss": 1.1241, + "step": 155900 + }, + { + "epoch": 13.49, + "learning_rate": 3.6536429004591526e-05, + "loss": 1.072, + "step": 155910 + }, + { + "epoch": 13.5, + "learning_rate": 3.653556267867972e-05, + "loss": 1.0702, + "step": 155920 + }, + { + "epoch": 13.5, + "learning_rate": 3.653469635276791e-05, + "loss": 1.0944, + "step": 155930 + }, + { + "epoch": 13.5, + "learning_rate": 3.65338300268561e-05, + "loss": 1.1148, + "step": 155940 + }, + { + "epoch": 13.5, + "learning_rate": 3.65329637009443e-05, + "loss": 1.1174, + "step": 155950 + }, + { + "epoch": 13.5, + "learning_rate": 3.6532097375032484e-05, + "loss": 1.0454, + "step": 155960 + }, + { + "epoch": 13.5, + "learning_rate": 3.653123104912068e-05, + "loss": 1.0996, + "step": 155970 + }, + { + "epoch": 13.5, + "learning_rate": 3.653036472320887e-05, + "loss": 1.1072, + "step": 155980 + }, + { + "epoch": 13.5, + "learning_rate": 3.652949839729706e-05, + "loss": 1.1325, + "step": 155990 + }, + { + "epoch": 13.5, + "learning_rate": 3.6528632071385256e-05, + "loss": 1.0742, + "step": 156000 + }, + { + "epoch": 13.5, + "learning_rate": 3.652776574547345e-05, + "loss": 1.064, + "step": 156010 + }, + { + "epoch": 13.5, + "learning_rate": 3.652689941956164e-05, + "loss": 1.0982, + "step": 156020 + }, + { + "epoch": 13.51, + "learning_rate": 3.652603309364983e-05, + "loss": 1.0973, + "step": 156030 + }, + { + "epoch": 13.51, + "learning_rate": 3.652516676773802e-05, + "loss": 1.1171, + "step": 156040 + }, + { + "epoch": 13.51, + "learning_rate": 3.6524300441826214e-05, + "loss": 1.1038, + "step": 156050 + }, + { + "epoch": 13.51, + "learning_rate": 3.652343411591441e-05, + "loss": 1.1129, + "step": 156060 + }, + { + "epoch": 13.51, + "learning_rate": 3.6522567790002596e-05, + "loss": 1.0778, + "step": 156070 + }, + { + "epoch": 13.51, + "learning_rate": 3.652170146409079e-05, + "loss": 1.0801, + "step": 156080 + }, + { + "epoch": 13.51, + "learning_rate": 3.6520835138178985e-05, + "loss": 1.1238, + "step": 156090 + }, + { + "epoch": 13.51, + "learning_rate": 3.651996881226717e-05, + "loss": 1.0742, + "step": 156100 + }, + { + "epoch": 13.51, + "learning_rate": 3.651910248635537e-05, + "loss": 1.0636, + "step": 156110 + }, + { + "epoch": 13.51, + "learning_rate": 3.651823616044356e-05, + "loss": 1.0629, + "step": 156120 + }, + { + "epoch": 13.51, + "learning_rate": 3.651736983453175e-05, + "loss": 1.1396, + "step": 156130 + }, + { + "epoch": 13.51, + "learning_rate": 3.6516503508619944e-05, + "loss": 1.1379, + "step": 156140 + }, + { + "epoch": 13.52, + "learning_rate": 3.651563718270813e-05, + "loss": 1.0413, + "step": 156150 + }, + { + "epoch": 13.52, + "learning_rate": 3.6514770856796326e-05, + "loss": 1.0691, + "step": 156160 + }, + { + "epoch": 13.52, + "learning_rate": 3.651390453088452e-05, + "loss": 1.0802, + "step": 156170 + }, + { + "epoch": 13.52, + "learning_rate": 3.651303820497271e-05, + "loss": 1.1004, + "step": 156180 + }, + { + "epoch": 13.52, + "learning_rate": 3.65121718790609e-05, + "loss": 1.0793, + "step": 156190 + }, + { + "epoch": 13.52, + "learning_rate": 3.65113055531491e-05, + "loss": 1.0788, + "step": 156200 + }, + { + "epoch": 13.52, + "learning_rate": 3.6510439227237285e-05, + "loss": 1.1016, + "step": 156210 + }, + { + "epoch": 13.52, + "learning_rate": 3.650957290132548e-05, + "loss": 1.0876, + "step": 156220 + }, + { + "epoch": 13.52, + "learning_rate": 3.6508706575413674e-05, + "loss": 1.0933, + "step": 156230 + }, + { + "epoch": 13.52, + "learning_rate": 3.650784024950186e-05, + "loss": 1.0921, + "step": 156240 + }, + { + "epoch": 13.52, + "learning_rate": 3.6506973923590056e-05, + "loss": 1.0609, + "step": 156250 + }, + { + "epoch": 13.53, + "learning_rate": 3.6506107597678244e-05, + "loss": 1.0969, + "step": 156260 + }, + { + "epoch": 13.53, + "learning_rate": 3.650524127176644e-05, + "loss": 1.1258, + "step": 156270 + }, + { + "epoch": 13.53, + "learning_rate": 3.650437494585463e-05, + "loss": 1.0991, + "step": 156280 + }, + { + "epoch": 13.53, + "learning_rate": 3.650350861994282e-05, + "loss": 1.1309, + "step": 156290 + }, + { + "epoch": 13.53, + "learning_rate": 3.6502642294031015e-05, + "loss": 1.0948, + "step": 156300 + }, + { + "epoch": 13.53, + "learning_rate": 3.650177596811921e-05, + "loss": 1.0978, + "step": 156310 + }, + { + "epoch": 13.53, + "learning_rate": 3.65009096422074e-05, + "loss": 1.1029, + "step": 156320 + }, + { + "epoch": 13.53, + "learning_rate": 3.650004331629559e-05, + "loss": 1.1365, + "step": 156330 + }, + { + "epoch": 13.53, + "learning_rate": 3.649917699038378e-05, + "loss": 1.0992, + "step": 156340 + }, + { + "epoch": 13.53, + "learning_rate": 3.649831066447197e-05, + "loss": 1.1174, + "step": 156350 + }, + { + "epoch": 13.53, + "learning_rate": 3.649744433856017e-05, + "loss": 1.0941, + "step": 156360 + }, + { + "epoch": 13.53, + "learning_rate": 3.6496578012648355e-05, + "loss": 1.0858, + "step": 156370 + }, + { + "epoch": 13.54, + "learning_rate": 3.649571168673655e-05, + "loss": 1.1067, + "step": 156380 + }, + { + "epoch": 13.54, + "learning_rate": 3.6494845360824744e-05, + "loss": 1.0827, + "step": 156390 + }, + { + "epoch": 13.54, + "learning_rate": 3.649397903491293e-05, + "loss": 1.1272, + "step": 156400 + }, + { + "epoch": 13.54, + "learning_rate": 3.6493112709001126e-05, + "loss": 1.0966, + "step": 156410 + }, + { + "epoch": 13.54, + "learning_rate": 3.649224638308932e-05, + "loss": 1.0671, + "step": 156420 + }, + { + "epoch": 13.54, + "learning_rate": 3.649138005717751e-05, + "loss": 1.1566, + "step": 156430 + }, + { + "epoch": 13.54, + "learning_rate": 3.64905137312657e-05, + "loss": 1.086, + "step": 156440 + }, + { + "epoch": 13.54, + "learning_rate": 3.648964740535389e-05, + "loss": 1.0907, + "step": 156450 + }, + { + "epoch": 13.54, + "learning_rate": 3.6488781079442085e-05, + "loss": 1.149, + "step": 156460 + }, + { + "epoch": 13.54, + "learning_rate": 3.648791475353028e-05, + "loss": 1.1616, + "step": 156470 + }, + { + "epoch": 13.54, + "learning_rate": 3.648704842761847e-05, + "loss": 1.0994, + "step": 156480 + }, + { + "epoch": 13.55, + "learning_rate": 3.648618210170666e-05, + "loss": 1.0878, + "step": 156490 + }, + { + "epoch": 13.55, + "learning_rate": 3.6485315775794856e-05, + "loss": 1.0762, + "step": 156500 + }, + { + "epoch": 13.55, + "learning_rate": 3.6484449449883044e-05, + "loss": 1.0569, + "step": 156510 + }, + { + "epoch": 13.55, + "learning_rate": 3.648358312397124e-05, + "loss": 1.1185, + "step": 156520 + }, + { + "epoch": 13.55, + "learning_rate": 3.648271679805943e-05, + "loss": 1.1273, + "step": 156530 + }, + { + "epoch": 13.55, + "learning_rate": 3.648185047214762e-05, + "loss": 1.1006, + "step": 156540 + }, + { + "epoch": 13.55, + "learning_rate": 3.6480984146235815e-05, + "loss": 1.0852, + "step": 156550 + }, + { + "epoch": 13.55, + "learning_rate": 3.6480117820324e-05, + "loss": 1.063, + "step": 156560 + }, + { + "epoch": 13.55, + "learning_rate": 3.64792514944122e-05, + "loss": 1.026, + "step": 156570 + }, + { + "epoch": 13.55, + "learning_rate": 3.647838516850039e-05, + "loss": 1.0851, + "step": 156580 + }, + { + "epoch": 13.55, + "learning_rate": 3.647751884258858e-05, + "loss": 1.079, + "step": 156590 + }, + { + "epoch": 13.55, + "learning_rate": 3.6476652516676774e-05, + "loss": 1.1144, + "step": 156600 + }, + { + "epoch": 13.56, + "learning_rate": 3.647578619076497e-05, + "loss": 1.1473, + "step": 156610 + }, + { + "epoch": 13.56, + "learning_rate": 3.6474919864853156e-05, + "loss": 1.1349, + "step": 156620 + }, + { + "epoch": 13.56, + "learning_rate": 3.647405353894135e-05, + "loss": 1.14, + "step": 156630 + }, + { + "epoch": 13.56, + "learning_rate": 3.6473187213029545e-05, + "loss": 1.0658, + "step": 156640 + }, + { + "epoch": 13.56, + "learning_rate": 3.647232088711773e-05, + "loss": 1.1156, + "step": 156650 + }, + { + "epoch": 13.56, + "learning_rate": 3.647145456120593e-05, + "loss": 1.0548, + "step": 156660 + }, + { + "epoch": 13.56, + "learning_rate": 3.6470588235294114e-05, + "loss": 1.1236, + "step": 156670 + }, + { + "epoch": 13.56, + "learning_rate": 3.646972190938231e-05, + "loss": 1.1111, + "step": 156680 + }, + { + "epoch": 13.56, + "learning_rate": 3.64688555834705e-05, + "loss": 1.1518, + "step": 156690 + }, + { + "epoch": 13.56, + "learning_rate": 3.646798925755869e-05, + "loss": 1.0135, + "step": 156700 + }, + { + "epoch": 13.56, + "learning_rate": 3.6467122931646885e-05, + "loss": 1.0594, + "step": 156710 + }, + { + "epoch": 13.56, + "learning_rate": 3.646625660573508e-05, + "loss": 1.1117, + "step": 156720 + }, + { + "epoch": 13.57, + "learning_rate": 3.646539027982327e-05, + "loss": 1.1116, + "step": 156730 + }, + { + "epoch": 13.57, + "learning_rate": 3.646452395391146e-05, + "loss": 1.0883, + "step": 156740 + }, + { + "epoch": 13.57, + "learning_rate": 3.6463657627999657e-05, + "loss": 1.1223, + "step": 156750 + }, + { + "epoch": 13.57, + "learning_rate": 3.6462791302087844e-05, + "loss": 1.1051, + "step": 156760 + }, + { + "epoch": 13.57, + "learning_rate": 3.646192497617604e-05, + "loss": 1.1368, + "step": 156770 + }, + { + "epoch": 13.57, + "learning_rate": 3.6461058650264226e-05, + "loss": 1.0998, + "step": 156780 + }, + { + "epoch": 13.57, + "learning_rate": 3.646019232435242e-05, + "loss": 1.0914, + "step": 156790 + }, + { + "epoch": 13.57, + "learning_rate": 3.6459325998440615e-05, + "loss": 1.0918, + "step": 156800 + }, + { + "epoch": 13.57, + "learning_rate": 3.64584596725288e-05, + "loss": 1.1031, + "step": 156810 + }, + { + "epoch": 13.57, + "learning_rate": 3.6457593346617e-05, + "loss": 1.1252, + "step": 156820 + }, + { + "epoch": 13.57, + "learning_rate": 3.645672702070519e-05, + "loss": 1.0441, + "step": 156830 + }, + { + "epoch": 13.58, + "learning_rate": 3.645586069479338e-05, + "loss": 1.114, + "step": 156840 + }, + { + "epoch": 13.58, + "learning_rate": 3.6454994368881574e-05, + "loss": 1.093, + "step": 156850 + }, + { + "epoch": 13.58, + "learning_rate": 3.645412804296977e-05, + "loss": 1.0754, + "step": 156860 + }, + { + "epoch": 13.58, + "learning_rate": 3.6453261717057956e-05, + "loss": 1.0498, + "step": 156870 + }, + { + "epoch": 13.58, + "learning_rate": 3.645239539114615e-05, + "loss": 1.0775, + "step": 156880 + }, + { + "epoch": 13.58, + "learning_rate": 3.645152906523434e-05, + "loss": 1.0936, + "step": 156890 + }, + { + "epoch": 13.58, + "learning_rate": 3.645066273932253e-05, + "loss": 1.0736, + "step": 156900 + }, + { + "epoch": 13.58, + "learning_rate": 3.644979641341073e-05, + "loss": 1.0489, + "step": 156910 + }, + { + "epoch": 13.58, + "learning_rate": 3.6448930087498915e-05, + "loss": 1.1312, + "step": 156920 + }, + { + "epoch": 13.58, + "learning_rate": 3.644806376158711e-05, + "loss": 1.0802, + "step": 156930 + }, + { + "epoch": 13.58, + "learning_rate": 3.6447197435675304e-05, + "loss": 1.1237, + "step": 156940 + }, + { + "epoch": 13.58, + "learning_rate": 3.644633110976349e-05, + "loss": 1.0893, + "step": 156950 + }, + { + "epoch": 13.59, + "learning_rate": 3.6445464783851686e-05, + "loss": 1.0933, + "step": 156960 + }, + { + "epoch": 13.59, + "learning_rate": 3.644459845793988e-05, + "loss": 1.1208, + "step": 156970 + }, + { + "epoch": 13.59, + "learning_rate": 3.644373213202807e-05, + "loss": 1.0297, + "step": 156980 + }, + { + "epoch": 13.59, + "learning_rate": 3.644286580611626e-05, + "loss": 1.08, + "step": 156990 + }, + { + "epoch": 13.59, + "learning_rate": 3.644199948020445e-05, + "loss": 1.1228, + "step": 157000 + }, + { + "epoch": 13.59, + "learning_rate": 3.6441133154292645e-05, + "loss": 1.0875, + "step": 157010 + }, + { + "epoch": 13.59, + "learning_rate": 3.644026682838084e-05, + "loss": 1.1168, + "step": 157020 + }, + { + "epoch": 13.59, + "learning_rate": 3.643940050246903e-05, + "loss": 1.0627, + "step": 157030 + }, + { + "epoch": 13.59, + "learning_rate": 3.643853417655722e-05, + "loss": 1.0759, + "step": 157040 + }, + { + "epoch": 13.59, + "learning_rate": 3.6437667850645416e-05, + "loss": 1.1144, + "step": 157050 + }, + { + "epoch": 13.59, + "learning_rate": 3.64368015247336e-05, + "loss": 1.1029, + "step": 157060 + }, + { + "epoch": 13.6, + "learning_rate": 3.64359351988218e-05, + "loss": 1.0711, + "step": 157070 + }, + { + "epoch": 13.6, + "learning_rate": 3.6435068872909985e-05, + "loss": 1.0681, + "step": 157080 + }, + { + "epoch": 13.6, + "learning_rate": 3.643420254699818e-05, + "loss": 1.1002, + "step": 157090 + }, + { + "epoch": 13.6, + "learning_rate": 3.6433336221086374e-05, + "loss": 1.0746, + "step": 157100 + }, + { + "epoch": 13.6, + "learning_rate": 3.643246989517456e-05, + "loss": 1.0746, + "step": 157110 + }, + { + "epoch": 13.6, + "learning_rate": 3.6431603569262756e-05, + "loss": 1.1439, + "step": 157120 + }, + { + "epoch": 13.6, + "learning_rate": 3.643073724335095e-05, + "loss": 1.0387, + "step": 157130 + }, + { + "epoch": 13.6, + "learning_rate": 3.642987091743914e-05, + "loss": 1.0796, + "step": 157140 + }, + { + "epoch": 13.6, + "learning_rate": 3.642900459152733e-05, + "loss": 1.1263, + "step": 157150 + }, + { + "epoch": 13.6, + "learning_rate": 3.642813826561553e-05, + "loss": 1.0834, + "step": 157160 + }, + { + "epoch": 13.6, + "learning_rate": 3.6427271939703715e-05, + "loss": 1.047, + "step": 157170 + }, + { + "epoch": 13.6, + "learning_rate": 3.642640561379191e-05, + "loss": 1.0974, + "step": 157180 + }, + { + "epoch": 13.61, + "learning_rate": 3.64255392878801e-05, + "loss": 1.1016, + "step": 157190 + }, + { + "epoch": 13.61, + "learning_rate": 3.642467296196829e-05, + "loss": 1.0751, + "step": 157200 + }, + { + "epoch": 13.61, + "learning_rate": 3.6423806636056486e-05, + "loss": 1.1184, + "step": 157210 + }, + { + "epoch": 13.61, + "learning_rate": 3.6422940310144674e-05, + "loss": 1.1498, + "step": 157220 + }, + { + "epoch": 13.61, + "learning_rate": 3.642207398423287e-05, + "loss": 1.0885, + "step": 157230 + }, + { + "epoch": 13.61, + "learning_rate": 3.642120765832106e-05, + "loss": 1.0709, + "step": 157240 + }, + { + "epoch": 13.61, + "learning_rate": 3.642034133240925e-05, + "loss": 1.0926, + "step": 157250 + }, + { + "epoch": 13.61, + "learning_rate": 3.6419475006497445e-05, + "loss": 1.0473, + "step": 157260 + }, + { + "epoch": 13.61, + "learning_rate": 3.641860868058564e-05, + "loss": 1.0581, + "step": 157270 + }, + { + "epoch": 13.61, + "learning_rate": 3.641774235467383e-05, + "loss": 1.081, + "step": 157280 + }, + { + "epoch": 13.61, + "learning_rate": 3.641687602876202e-05, + "loss": 1.053, + "step": 157290 + }, + { + "epoch": 13.62, + "learning_rate": 3.641600970285021e-05, + "loss": 1.09, + "step": 157300 + }, + { + "epoch": 13.62, + "learning_rate": 3.6415143376938404e-05, + "loss": 1.1083, + "step": 157310 + }, + { + "epoch": 13.62, + "learning_rate": 3.64142770510266e-05, + "loss": 1.1276, + "step": 157320 + }, + { + "epoch": 13.62, + "learning_rate": 3.6413410725114786e-05, + "loss": 1.0586, + "step": 157330 + }, + { + "epoch": 13.62, + "learning_rate": 3.641254439920298e-05, + "loss": 1.0761, + "step": 157340 + }, + { + "epoch": 13.62, + "learning_rate": 3.6411678073291175e-05, + "loss": 1.0641, + "step": 157350 + }, + { + "epoch": 13.62, + "learning_rate": 3.641081174737936e-05, + "loss": 1.0653, + "step": 157360 + }, + { + "epoch": 13.62, + "learning_rate": 3.640994542146756e-05, + "loss": 1.0606, + "step": 157370 + }, + { + "epoch": 13.62, + "learning_rate": 3.640907909555575e-05, + "loss": 1.1206, + "step": 157380 + }, + { + "epoch": 13.62, + "learning_rate": 3.640821276964394e-05, + "loss": 1.1203, + "step": 157390 + }, + { + "epoch": 13.62, + "learning_rate": 3.640734644373213e-05, + "loss": 1.1125, + "step": 157400 + }, + { + "epoch": 13.62, + "learning_rate": 3.640648011782032e-05, + "loss": 1.0789, + "step": 157410 + }, + { + "epoch": 13.63, + "learning_rate": 3.6405613791908515e-05, + "loss": 1.1094, + "step": 157420 + }, + { + "epoch": 13.63, + "learning_rate": 3.640474746599671e-05, + "loss": 1.1041, + "step": 157430 + }, + { + "epoch": 13.63, + "learning_rate": 3.64038811400849e-05, + "loss": 1.0762, + "step": 157440 + }, + { + "epoch": 13.63, + "learning_rate": 3.640301481417309e-05, + "loss": 1.1632, + "step": 157450 + }, + { + "epoch": 13.63, + "learning_rate": 3.6402148488261286e-05, + "loss": 1.1017, + "step": 157460 + }, + { + "epoch": 13.63, + "learning_rate": 3.6401282162349474e-05, + "loss": 1.1239, + "step": 157470 + }, + { + "epoch": 13.63, + "learning_rate": 3.640041583643767e-05, + "loss": 1.0851, + "step": 157480 + }, + { + "epoch": 13.63, + "learning_rate": 3.639954951052586e-05, + "loss": 1.093, + "step": 157490 + }, + { + "epoch": 13.63, + "learning_rate": 3.639868318461405e-05, + "loss": 1.1288, + "step": 157500 + }, + { + "epoch": 13.63, + "learning_rate": 3.6397816858702245e-05, + "loss": 1.0104, + "step": 157510 + }, + { + "epoch": 13.63, + "learning_rate": 3.639695053279043e-05, + "loss": 1.0659, + "step": 157520 + }, + { + "epoch": 13.64, + "learning_rate": 3.639608420687863e-05, + "loss": 1.0759, + "step": 157530 + }, + { + "epoch": 13.64, + "learning_rate": 3.639521788096682e-05, + "loss": 1.1134, + "step": 157540 + }, + { + "epoch": 13.64, + "learning_rate": 3.639435155505501e-05, + "loss": 1.1292, + "step": 157550 + }, + { + "epoch": 13.64, + "learning_rate": 3.6393485229143204e-05, + "loss": 1.1531, + "step": 157560 + }, + { + "epoch": 13.64, + "learning_rate": 3.63926189032314e-05, + "loss": 1.0713, + "step": 157570 + }, + { + "epoch": 13.64, + "learning_rate": 3.6391752577319586e-05, + "loss": 1.0748, + "step": 157580 + }, + { + "epoch": 13.64, + "learning_rate": 3.639088625140778e-05, + "loss": 1.0954, + "step": 157590 + }, + { + "epoch": 13.64, + "learning_rate": 3.6390019925495975e-05, + "loss": 1.0644, + "step": 157600 + }, + { + "epoch": 13.64, + "learning_rate": 3.638915359958416e-05, + "loss": 1.1241, + "step": 157610 + }, + { + "epoch": 13.64, + "learning_rate": 3.638828727367236e-05, + "loss": 1.1583, + "step": 157620 + }, + { + "epoch": 13.64, + "learning_rate": 3.6387420947760545e-05, + "loss": 1.135, + "step": 157630 + }, + { + "epoch": 13.64, + "learning_rate": 3.638655462184874e-05, + "loss": 1.0859, + "step": 157640 + }, + { + "epoch": 13.65, + "learning_rate": 3.6385688295936934e-05, + "loss": 1.0767, + "step": 157650 + }, + { + "epoch": 13.65, + "learning_rate": 3.638482197002512e-05, + "loss": 1.0902, + "step": 157660 + }, + { + "epoch": 13.65, + "learning_rate": 3.6383955644113316e-05, + "loss": 1.1061, + "step": 157670 + }, + { + "epoch": 13.65, + "learning_rate": 3.638308931820151e-05, + "loss": 1.0716, + "step": 157680 + }, + { + "epoch": 13.65, + "learning_rate": 3.63822229922897e-05, + "loss": 0.9899, + "step": 157690 + }, + { + "epoch": 13.65, + "learning_rate": 3.638135666637789e-05, + "loss": 1.0927, + "step": 157700 + }, + { + "epoch": 13.65, + "learning_rate": 3.638049034046609e-05, + "loss": 1.0234, + "step": 157710 + }, + { + "epoch": 13.65, + "learning_rate": 3.6379624014554275e-05, + "loss": 1.0981, + "step": 157720 + }, + { + "epoch": 13.65, + "learning_rate": 3.637875768864247e-05, + "loss": 1.137, + "step": 157730 + }, + { + "epoch": 13.65, + "learning_rate": 3.637789136273066e-05, + "loss": 1.0694, + "step": 157740 + }, + { + "epoch": 13.65, + "learning_rate": 3.637702503681885e-05, + "loss": 1.1322, + "step": 157750 + }, + { + "epoch": 13.65, + "learning_rate": 3.6376158710907046e-05, + "loss": 1.0889, + "step": 157760 + }, + { + "epoch": 13.66, + "learning_rate": 3.637529238499523e-05, + "loss": 1.0535, + "step": 157770 + }, + { + "epoch": 13.66, + "learning_rate": 3.637442605908343e-05, + "loss": 1.0855, + "step": 157780 + }, + { + "epoch": 13.66, + "learning_rate": 3.637355973317162e-05, + "loss": 1.08, + "step": 157790 + }, + { + "epoch": 13.66, + "learning_rate": 3.637269340725981e-05, + "loss": 1.0388, + "step": 157800 + }, + { + "epoch": 13.66, + "learning_rate": 3.6371827081348004e-05, + "loss": 1.1389, + "step": 157810 + }, + { + "epoch": 13.66, + "learning_rate": 3.637096075543619e-05, + "loss": 1.1104, + "step": 157820 + }, + { + "epoch": 13.66, + "learning_rate": 3.6370094429524386e-05, + "loss": 1.119, + "step": 157830 + }, + { + "epoch": 13.66, + "learning_rate": 3.636922810361258e-05, + "loss": 1.1195, + "step": 157840 + }, + { + "epoch": 13.66, + "learning_rate": 3.636836177770077e-05, + "loss": 1.1115, + "step": 157850 + }, + { + "epoch": 13.66, + "learning_rate": 3.636749545178896e-05, + "loss": 1.1281, + "step": 157860 + }, + { + "epoch": 13.66, + "learning_rate": 3.636662912587716e-05, + "loss": 1.0539, + "step": 157870 + }, + { + "epoch": 13.67, + "learning_rate": 3.6365762799965345e-05, + "loss": 1.0973, + "step": 157880 + }, + { + "epoch": 13.67, + "learning_rate": 3.636489647405354e-05, + "loss": 1.0884, + "step": 157890 + }, + { + "epoch": 13.67, + "learning_rate": 3.6364030148141734e-05, + "loss": 1.0923, + "step": 157900 + }, + { + "epoch": 13.67, + "learning_rate": 3.636316382222992e-05, + "loss": 1.1333, + "step": 157910 + }, + { + "epoch": 13.67, + "learning_rate": 3.6362297496318116e-05, + "loss": 1.0642, + "step": 157920 + }, + { + "epoch": 13.67, + "learning_rate": 3.6361431170406304e-05, + "loss": 1.0979, + "step": 157930 + }, + { + "epoch": 13.67, + "learning_rate": 3.63605648444945e-05, + "loss": 1.0931, + "step": 157940 + }, + { + "epoch": 13.67, + "learning_rate": 3.635969851858269e-05, + "loss": 1.1834, + "step": 157950 + }, + { + "epoch": 13.67, + "learning_rate": 3.635883219267088e-05, + "loss": 1.1527, + "step": 157960 + }, + { + "epoch": 13.67, + "learning_rate": 3.6357965866759075e-05, + "loss": 1.0829, + "step": 157970 + }, + { + "epoch": 13.67, + "learning_rate": 3.635709954084727e-05, + "loss": 1.1094, + "step": 157980 + }, + { + "epoch": 13.67, + "learning_rate": 3.635623321493546e-05, + "loss": 1.0558, + "step": 157990 + }, + { + "epoch": 13.68, + "learning_rate": 3.635536688902365e-05, + "loss": 1.1078, + "step": 158000 + }, + { + "epoch": 13.68, + "learning_rate": 3.6354500563111846e-05, + "loss": 1.1133, + "step": 158010 + }, + { + "epoch": 13.68, + "learning_rate": 3.6353634237200034e-05, + "loss": 1.0941, + "step": 158020 + }, + { + "epoch": 13.68, + "learning_rate": 3.635276791128823e-05, + "loss": 1.1114, + "step": 158030 + }, + { + "epoch": 13.68, + "learning_rate": 3.6351901585376416e-05, + "loss": 1.0683, + "step": 158040 + }, + { + "epoch": 13.68, + "learning_rate": 3.635103525946461e-05, + "loss": 1.0887, + "step": 158050 + }, + { + "epoch": 13.68, + "learning_rate": 3.6350168933552805e-05, + "loss": 1.0324, + "step": 158060 + }, + { + "epoch": 13.68, + "learning_rate": 3.634930260764099e-05, + "loss": 1.1345, + "step": 158070 + }, + { + "epoch": 13.68, + "learning_rate": 3.634843628172919e-05, + "loss": 1.052, + "step": 158080 + }, + { + "epoch": 13.68, + "learning_rate": 3.634756995581738e-05, + "loss": 1.1384, + "step": 158090 + }, + { + "epoch": 13.68, + "learning_rate": 3.634670362990557e-05, + "loss": 1.117, + "step": 158100 + }, + { + "epoch": 13.69, + "learning_rate": 3.634583730399376e-05, + "loss": 1.0949, + "step": 158110 + }, + { + "epoch": 13.69, + "learning_rate": 3.634497097808196e-05, + "loss": 1.165, + "step": 158120 + }, + { + "epoch": 13.69, + "learning_rate": 3.6344104652170145e-05, + "loss": 1.0679, + "step": 158130 + }, + { + "epoch": 13.69, + "learning_rate": 3.634323832625834e-05, + "loss": 1.0826, + "step": 158140 + }, + { + "epoch": 13.69, + "learning_rate": 3.634237200034653e-05, + "loss": 1.1127, + "step": 158150 + }, + { + "epoch": 13.69, + "learning_rate": 3.634150567443472e-05, + "loss": 1.0024, + "step": 158160 + }, + { + "epoch": 13.69, + "learning_rate": 3.6340639348522916e-05, + "loss": 1.0627, + "step": 158170 + }, + { + "epoch": 13.69, + "learning_rate": 3.6339773022611104e-05, + "loss": 1.0959, + "step": 158180 + }, + { + "epoch": 13.69, + "learning_rate": 3.63389066966993e-05, + "loss": 1.1181, + "step": 158190 + }, + { + "epoch": 13.69, + "learning_rate": 3.633804037078749e-05, + "loss": 1.1314, + "step": 158200 + }, + { + "epoch": 13.69, + "learning_rate": 3.633717404487568e-05, + "loss": 1.1201, + "step": 158210 + }, + { + "epoch": 13.69, + "learning_rate": 3.6336307718963875e-05, + "loss": 1.0683, + "step": 158220 + }, + { + "epoch": 13.7, + "learning_rate": 3.633544139305207e-05, + "loss": 1.0888, + "step": 158230 + }, + { + "epoch": 13.7, + "learning_rate": 3.633457506714026e-05, + "loss": 1.1058, + "step": 158240 + }, + { + "epoch": 13.7, + "learning_rate": 3.633370874122845e-05, + "loss": 1.1522, + "step": 158250 + }, + { + "epoch": 13.7, + "learning_rate": 3.633284241531664e-05, + "loss": 1.1238, + "step": 158260 + }, + { + "epoch": 13.7, + "learning_rate": 3.6331976089404834e-05, + "loss": 1.034, + "step": 158270 + }, + { + "epoch": 13.7, + "learning_rate": 3.633110976349303e-05, + "loss": 1.0586, + "step": 158280 + }, + { + "epoch": 13.7, + "learning_rate": 3.6330243437581216e-05, + "loss": 1.0284, + "step": 158290 + }, + { + "epoch": 13.7, + "learning_rate": 3.632937711166941e-05, + "loss": 1.0796, + "step": 158300 + }, + { + "epoch": 13.7, + "learning_rate": 3.6328510785757605e-05, + "loss": 1.1531, + "step": 158310 + }, + { + "epoch": 13.7, + "learning_rate": 3.632764445984579e-05, + "loss": 1.0568, + "step": 158320 + }, + { + "epoch": 13.7, + "learning_rate": 3.632677813393399e-05, + "loss": 1.1289, + "step": 158330 + }, + { + "epoch": 13.71, + "learning_rate": 3.632591180802218e-05, + "loss": 1.0931, + "step": 158340 + }, + { + "epoch": 13.71, + "learning_rate": 3.632504548211037e-05, + "loss": 1.1776, + "step": 158350 + }, + { + "epoch": 13.71, + "learning_rate": 3.6324179156198564e-05, + "loss": 1.104, + "step": 158360 + }, + { + "epoch": 13.71, + "learning_rate": 3.632331283028675e-05, + "loss": 1.0879, + "step": 158370 + }, + { + "epoch": 13.71, + "learning_rate": 3.6322446504374946e-05, + "loss": 1.0435, + "step": 158380 + }, + { + "epoch": 13.71, + "learning_rate": 3.632158017846314e-05, + "loss": 1.1638, + "step": 158390 + }, + { + "epoch": 13.71, + "learning_rate": 3.632071385255133e-05, + "loss": 1.0918, + "step": 158400 + }, + { + "epoch": 13.71, + "learning_rate": 3.631984752663952e-05, + "loss": 1.0909, + "step": 158410 + }, + { + "epoch": 13.71, + "learning_rate": 3.631898120072772e-05, + "loss": 1.0527, + "step": 158420 + }, + { + "epoch": 13.71, + "learning_rate": 3.6318114874815904e-05, + "loss": 1.1335, + "step": 158430 + }, + { + "epoch": 13.71, + "learning_rate": 3.63172485489041e-05, + "loss": 1.0936, + "step": 158440 + }, + { + "epoch": 13.71, + "learning_rate": 3.631638222299229e-05, + "loss": 1.0808, + "step": 158450 + }, + { + "epoch": 13.72, + "learning_rate": 3.631551589708048e-05, + "loss": 1.0352, + "step": 158460 + }, + { + "epoch": 13.72, + "learning_rate": 3.6314649571168676e-05, + "loss": 1.1048, + "step": 158470 + }, + { + "epoch": 13.72, + "learning_rate": 3.631378324525686e-05, + "loss": 1.1349, + "step": 158480 + }, + { + "epoch": 13.72, + "learning_rate": 3.631291691934506e-05, + "loss": 1.0926, + "step": 158490 + }, + { + "epoch": 13.72, + "learning_rate": 3.631205059343325e-05, + "loss": 1.1154, + "step": 158500 + }, + { + "epoch": 13.72, + "learning_rate": 3.631118426752144e-05, + "loss": 1.0912, + "step": 158510 + }, + { + "epoch": 13.72, + "learning_rate": 3.6310317941609634e-05, + "loss": 1.099, + "step": 158520 + }, + { + "epoch": 13.72, + "learning_rate": 3.630945161569783e-05, + "loss": 1.0947, + "step": 158530 + }, + { + "epoch": 13.72, + "learning_rate": 3.6308585289786016e-05, + "loss": 1.1312, + "step": 158540 + }, + { + "epoch": 13.72, + "learning_rate": 3.630771896387421e-05, + "loss": 1.0863, + "step": 158550 + }, + { + "epoch": 13.72, + "learning_rate": 3.6306852637962405e-05, + "loss": 1.0968, + "step": 158560 + }, + { + "epoch": 13.73, + "learning_rate": 3.630598631205059e-05, + "loss": 1.1357, + "step": 158570 + }, + { + "epoch": 13.73, + "learning_rate": 3.630511998613879e-05, + "loss": 1.1511, + "step": 158580 + }, + { + "epoch": 13.73, + "learning_rate": 3.6304253660226975e-05, + "loss": 1.0836, + "step": 158590 + }, + { + "epoch": 13.73, + "learning_rate": 3.630338733431517e-05, + "loss": 1.0814, + "step": 158600 + }, + { + "epoch": 13.73, + "learning_rate": 3.6302521008403364e-05, + "loss": 1.063, + "step": 158610 + }, + { + "epoch": 13.73, + "learning_rate": 3.630165468249155e-05, + "loss": 1.1021, + "step": 158620 + }, + { + "epoch": 13.73, + "learning_rate": 3.6300788356579746e-05, + "loss": 1.1304, + "step": 158630 + }, + { + "epoch": 13.73, + "learning_rate": 3.629992203066794e-05, + "loss": 1.1003, + "step": 158640 + }, + { + "epoch": 13.73, + "learning_rate": 3.629905570475613e-05, + "loss": 1.0977, + "step": 158650 + }, + { + "epoch": 13.73, + "learning_rate": 3.629818937884432e-05, + "loss": 1.0483, + "step": 158660 + }, + { + "epoch": 13.73, + "learning_rate": 3.629732305293251e-05, + "loss": 1.1208, + "step": 158670 + }, + { + "epoch": 13.73, + "learning_rate": 3.6296456727020705e-05, + "loss": 1.0861, + "step": 158680 + }, + { + "epoch": 13.74, + "learning_rate": 3.62955904011089e-05, + "loss": 1.0833, + "step": 158690 + }, + { + "epoch": 13.74, + "learning_rate": 3.629472407519709e-05, + "loss": 1.1176, + "step": 158700 + }, + { + "epoch": 13.74, + "learning_rate": 3.629385774928528e-05, + "loss": 1.113, + "step": 158710 + }, + { + "epoch": 13.74, + "learning_rate": 3.6292991423373476e-05, + "loss": 1.098, + "step": 158720 + }, + { + "epoch": 13.74, + "learning_rate": 3.6292125097461664e-05, + "loss": 1.0735, + "step": 158730 + }, + { + "epoch": 13.74, + "learning_rate": 3.629125877154986e-05, + "loss": 1.0595, + "step": 158740 + }, + { + "epoch": 13.74, + "learning_rate": 3.629039244563805e-05, + "loss": 1.0992, + "step": 158750 + }, + { + "epoch": 13.74, + "learning_rate": 3.628952611972624e-05, + "loss": 1.0655, + "step": 158760 + }, + { + "epoch": 13.74, + "learning_rate": 3.6288659793814435e-05, + "loss": 1.0849, + "step": 158770 + }, + { + "epoch": 13.74, + "learning_rate": 3.628779346790262e-05, + "loss": 1.1825, + "step": 158780 + }, + { + "epoch": 13.74, + "learning_rate": 3.628692714199082e-05, + "loss": 1.1375, + "step": 158790 + }, + { + "epoch": 13.75, + "learning_rate": 3.628606081607901e-05, + "loss": 1.1091, + "step": 158800 + }, + { + "epoch": 13.75, + "learning_rate": 3.62851944901672e-05, + "loss": 1.055, + "step": 158810 + }, + { + "epoch": 13.75, + "learning_rate": 3.628432816425539e-05, + "loss": 1.0884, + "step": 158820 + }, + { + "epoch": 13.75, + "learning_rate": 3.628346183834359e-05, + "loss": 1.0625, + "step": 158830 + }, + { + "epoch": 13.75, + "learning_rate": 3.6282595512431775e-05, + "loss": 1.1143, + "step": 158840 + }, + { + "epoch": 13.75, + "learning_rate": 3.628172918651997e-05, + "loss": 1.0926, + "step": 158850 + }, + { + "epoch": 13.75, + "learning_rate": 3.6280862860608164e-05, + "loss": 1.1001, + "step": 158860 + }, + { + "epoch": 13.75, + "learning_rate": 3.627999653469635e-05, + "loss": 1.0863, + "step": 158870 + }, + { + "epoch": 13.75, + "learning_rate": 3.6279130208784546e-05, + "loss": 1.1078, + "step": 158880 + }, + { + "epoch": 13.75, + "learning_rate": 3.6278263882872734e-05, + "loss": 1.0818, + "step": 158890 + }, + { + "epoch": 13.75, + "learning_rate": 3.627739755696093e-05, + "loss": 1.1274, + "step": 158900 + }, + { + "epoch": 13.75, + "learning_rate": 3.627653123104912e-05, + "loss": 1.114, + "step": 158910 + }, + { + "epoch": 13.76, + "learning_rate": 3.627566490513731e-05, + "loss": 1.1075, + "step": 158920 + }, + { + "epoch": 13.76, + "learning_rate": 3.6274798579225505e-05, + "loss": 1.1246, + "step": 158930 + }, + { + "epoch": 13.76, + "learning_rate": 3.62739322533137e-05, + "loss": 1.0717, + "step": 158940 + }, + { + "epoch": 13.76, + "learning_rate": 3.627306592740189e-05, + "loss": 1.0608, + "step": 158950 + }, + { + "epoch": 13.76, + "learning_rate": 3.627219960149008e-05, + "loss": 1.0958, + "step": 158960 + }, + { + "epoch": 13.76, + "learning_rate": 3.6271333275578276e-05, + "loss": 1.0534, + "step": 158970 + }, + { + "epoch": 13.76, + "learning_rate": 3.6270466949666464e-05, + "loss": 1.0903, + "step": 158980 + }, + { + "epoch": 13.76, + "learning_rate": 3.626960062375466e-05, + "loss": 1.0714, + "step": 158990 + }, + { + "epoch": 13.76, + "learning_rate": 3.6268734297842846e-05, + "loss": 1.045, + "step": 159000 + }, + { + "epoch": 13.76, + "learning_rate": 3.626786797193104e-05, + "loss": 1.1081, + "step": 159010 + }, + { + "epoch": 13.76, + "learning_rate": 3.6267001646019235e-05, + "loss": 1.0558, + "step": 159020 + }, + { + "epoch": 13.76, + "learning_rate": 3.626613532010742e-05, + "loss": 1.041, + "step": 159030 + }, + { + "epoch": 13.77, + "learning_rate": 3.626526899419562e-05, + "loss": 1.0824, + "step": 159040 + }, + { + "epoch": 13.77, + "learning_rate": 3.626440266828381e-05, + "loss": 1.0564, + "step": 159050 + }, + { + "epoch": 13.77, + "learning_rate": 3.6263536342372e-05, + "loss": 1.1269, + "step": 159060 + }, + { + "epoch": 13.77, + "learning_rate": 3.6262670016460194e-05, + "loss": 1.0579, + "step": 159070 + }, + { + "epoch": 13.77, + "learning_rate": 3.626180369054839e-05, + "loss": 1.098, + "step": 159080 + }, + { + "epoch": 13.77, + "learning_rate": 3.6260937364636576e-05, + "loss": 1.0767, + "step": 159090 + }, + { + "epoch": 13.77, + "learning_rate": 3.626007103872477e-05, + "loss": 1.1047, + "step": 159100 + }, + { + "epoch": 13.77, + "learning_rate": 3.625920471281296e-05, + "loss": 1.0622, + "step": 159110 + }, + { + "epoch": 13.77, + "learning_rate": 3.625833838690115e-05, + "loss": 1.0662, + "step": 159120 + }, + { + "epoch": 13.77, + "learning_rate": 3.625747206098935e-05, + "loss": 1.0914, + "step": 159130 + }, + { + "epoch": 13.77, + "learning_rate": 3.6256605735077534e-05, + "loss": 1.0731, + "step": 159140 + }, + { + "epoch": 13.78, + "learning_rate": 3.625573940916573e-05, + "loss": 1.0847, + "step": 159150 + }, + { + "epoch": 13.78, + "learning_rate": 3.625487308325392e-05, + "loss": 1.0965, + "step": 159160 + }, + { + "epoch": 13.78, + "learning_rate": 3.625400675734211e-05, + "loss": 1.1126, + "step": 159170 + }, + { + "epoch": 13.78, + "learning_rate": 3.6253140431430305e-05, + "loss": 1.093, + "step": 159180 + }, + { + "epoch": 13.78, + "learning_rate": 3.62522741055185e-05, + "loss": 1.1225, + "step": 159190 + }, + { + "epoch": 13.78, + "learning_rate": 3.625140777960669e-05, + "loss": 1.0565, + "step": 159200 + }, + { + "epoch": 13.78, + "learning_rate": 3.625054145369488e-05, + "loss": 1.0798, + "step": 159210 + }, + { + "epoch": 13.78, + "learning_rate": 3.624967512778307e-05, + "loss": 1.1255, + "step": 159220 + }, + { + "epoch": 13.78, + "learning_rate": 3.6248808801871264e-05, + "loss": 1.0723, + "step": 159230 + }, + { + "epoch": 13.78, + "learning_rate": 3.624794247595946e-05, + "loss": 1.0818, + "step": 159240 + }, + { + "epoch": 13.78, + "learning_rate": 3.6247076150047646e-05, + "loss": 1.0886, + "step": 159250 + }, + { + "epoch": 13.78, + "learning_rate": 3.624620982413584e-05, + "loss": 1.0852, + "step": 159260 + }, + { + "epoch": 13.79, + "learning_rate": 3.6245343498224035e-05, + "loss": 1.0656, + "step": 159270 + }, + { + "epoch": 13.79, + "learning_rate": 3.624447717231222e-05, + "loss": 1.1272, + "step": 159280 + }, + { + "epoch": 13.79, + "learning_rate": 3.624361084640042e-05, + "loss": 1.068, + "step": 159290 + }, + { + "epoch": 13.79, + "learning_rate": 3.624274452048861e-05, + "loss": 1.0479, + "step": 159300 + }, + { + "epoch": 13.79, + "learning_rate": 3.62418781945768e-05, + "loss": 1.125, + "step": 159310 + }, + { + "epoch": 13.79, + "learning_rate": 3.6241011868664994e-05, + "loss": 1.1159, + "step": 159320 + }, + { + "epoch": 13.79, + "learning_rate": 3.624014554275318e-05, + "loss": 1.0498, + "step": 159330 + }, + { + "epoch": 13.79, + "learning_rate": 3.6239279216841376e-05, + "loss": 1.1248, + "step": 159340 + }, + { + "epoch": 13.79, + "learning_rate": 3.623841289092957e-05, + "loss": 1.0972, + "step": 159350 + }, + { + "epoch": 13.79, + "learning_rate": 3.623754656501776e-05, + "loss": 1.086, + "step": 159360 + }, + { + "epoch": 13.79, + "learning_rate": 3.623668023910595e-05, + "loss": 1.0931, + "step": 159370 + }, + { + "epoch": 13.8, + "learning_rate": 3.623581391319415e-05, + "loss": 1.0491, + "step": 159380 + }, + { + "epoch": 13.8, + "learning_rate": 3.6234947587282335e-05, + "loss": 1.1716, + "step": 159390 + }, + { + "epoch": 13.8, + "learning_rate": 3.623408126137053e-05, + "loss": 1.098, + "step": 159400 + }, + { + "epoch": 13.8, + "learning_rate": 3.623321493545872e-05, + "loss": 1.0757, + "step": 159410 + }, + { + "epoch": 13.8, + "learning_rate": 3.623234860954691e-05, + "loss": 1.1427, + "step": 159420 + }, + { + "epoch": 13.8, + "learning_rate": 3.6231482283635106e-05, + "loss": 1.0922, + "step": 159430 + }, + { + "epoch": 13.8, + "learning_rate": 3.6230615957723293e-05, + "loss": 1.0689, + "step": 159440 + }, + { + "epoch": 13.8, + "learning_rate": 3.622974963181149e-05, + "loss": 1.013, + "step": 159450 + }, + { + "epoch": 13.8, + "learning_rate": 3.622888330589968e-05, + "loss": 1.0864, + "step": 159460 + }, + { + "epoch": 13.8, + "learning_rate": 3.622801697998787e-05, + "loss": 1.1087, + "step": 159470 + }, + { + "epoch": 13.8, + "learning_rate": 3.6227150654076065e-05, + "loss": 1.0531, + "step": 159480 + }, + { + "epoch": 13.8, + "learning_rate": 3.622628432816426e-05, + "loss": 1.0699, + "step": 159490 + }, + { + "epoch": 13.81, + "learning_rate": 3.622541800225245e-05, + "loss": 1.0868, + "step": 159500 + }, + { + "epoch": 13.81, + "learning_rate": 3.622455167634064e-05, + "loss": 1.1361, + "step": 159510 + }, + { + "epoch": 13.81, + "learning_rate": 3.622368535042883e-05, + "loss": 1.0589, + "step": 159520 + }, + { + "epoch": 13.81, + "learning_rate": 3.622281902451702e-05, + "loss": 1.0663, + "step": 159530 + }, + { + "epoch": 13.81, + "learning_rate": 3.622195269860522e-05, + "loss": 1.0945, + "step": 159540 + }, + { + "epoch": 13.81, + "learning_rate": 3.6221086372693405e-05, + "loss": 1.0941, + "step": 159550 + }, + { + "epoch": 13.81, + "learning_rate": 3.62202200467816e-05, + "loss": 1.1161, + "step": 159560 + }, + { + "epoch": 13.81, + "learning_rate": 3.6219353720869794e-05, + "loss": 1.0508, + "step": 159570 + }, + { + "epoch": 13.81, + "learning_rate": 3.621848739495798e-05, + "loss": 1.0996, + "step": 159580 + }, + { + "epoch": 13.81, + "learning_rate": 3.6217621069046176e-05, + "loss": 1.1529, + "step": 159590 + }, + { + "epoch": 13.81, + "learning_rate": 3.621675474313437e-05, + "loss": 1.0886, + "step": 159600 + }, + { + "epoch": 13.82, + "learning_rate": 3.621588841722256e-05, + "loss": 1.0759, + "step": 159610 + }, + { + "epoch": 13.82, + "learning_rate": 3.621502209131075e-05, + "loss": 1.0565, + "step": 159620 + }, + { + "epoch": 13.82, + "learning_rate": 3.621415576539894e-05, + "loss": 1.0643, + "step": 159630 + }, + { + "epoch": 13.82, + "learning_rate": 3.6213289439487135e-05, + "loss": 1.1133, + "step": 159640 + }, + { + "epoch": 13.82, + "learning_rate": 3.621242311357533e-05, + "loss": 1.0213, + "step": 159650 + }, + { + "epoch": 13.82, + "learning_rate": 3.621155678766352e-05, + "loss": 1.103, + "step": 159660 + }, + { + "epoch": 13.82, + "learning_rate": 3.621069046175171e-05, + "loss": 1.0634, + "step": 159670 + }, + { + "epoch": 13.82, + "learning_rate": 3.6209824135839906e-05, + "loss": 1.0793, + "step": 159680 + }, + { + "epoch": 13.82, + "learning_rate": 3.6208957809928094e-05, + "loss": 1.129, + "step": 159690 + }, + { + "epoch": 13.82, + "learning_rate": 3.620809148401629e-05, + "loss": 1.0671, + "step": 159700 + }, + { + "epoch": 13.82, + "learning_rate": 3.620722515810448e-05, + "loss": 1.0867, + "step": 159710 + }, + { + "epoch": 13.82, + "learning_rate": 3.620635883219267e-05, + "loss": 1.11, + "step": 159720 + }, + { + "epoch": 13.83, + "learning_rate": 3.6205492506280865e-05, + "loss": 1.0789, + "step": 159730 + }, + { + "epoch": 13.83, + "learning_rate": 3.620462618036905e-05, + "loss": 1.0434, + "step": 159740 + }, + { + "epoch": 13.83, + "learning_rate": 3.620375985445725e-05, + "loss": 1.105, + "step": 159750 + }, + { + "epoch": 13.83, + "learning_rate": 3.620289352854544e-05, + "loss": 1.1171, + "step": 159760 + }, + { + "epoch": 13.83, + "learning_rate": 3.620202720263363e-05, + "loss": 1.1248, + "step": 159770 + }, + { + "epoch": 13.83, + "learning_rate": 3.6201160876721824e-05, + "loss": 1.1327, + "step": 159780 + }, + { + "epoch": 13.83, + "learning_rate": 3.620029455081002e-05, + "loss": 1.0906, + "step": 159790 + }, + { + "epoch": 13.83, + "learning_rate": 3.6199428224898206e-05, + "loss": 1.1004, + "step": 159800 + }, + { + "epoch": 13.83, + "learning_rate": 3.61985618989864e-05, + "loss": 1.1326, + "step": 159810 + }, + { + "epoch": 13.83, + "learning_rate": 3.6197695573074595e-05, + "loss": 1.0399, + "step": 159820 + }, + { + "epoch": 13.83, + "learning_rate": 3.619682924716278e-05, + "loss": 1.1373, + "step": 159830 + }, + { + "epoch": 13.84, + "learning_rate": 3.619596292125098e-05, + "loss": 1.0973, + "step": 159840 + }, + { + "epoch": 13.84, + "learning_rate": 3.6195096595339164e-05, + "loss": 1.1244, + "step": 159850 + }, + { + "epoch": 13.84, + "learning_rate": 3.619423026942736e-05, + "loss": 1.0767, + "step": 159860 + }, + { + "epoch": 13.84, + "learning_rate": 3.619336394351555e-05, + "loss": 1.0997, + "step": 159870 + }, + { + "epoch": 13.84, + "learning_rate": 3.619249761760374e-05, + "loss": 1.0766, + "step": 159880 + }, + { + "epoch": 13.84, + "learning_rate": 3.6191631291691935e-05, + "loss": 1.0643, + "step": 159890 + }, + { + "epoch": 13.84, + "learning_rate": 3.619076496578013e-05, + "loss": 1.1109, + "step": 159900 + }, + { + "epoch": 13.84, + "learning_rate": 3.618989863986832e-05, + "loss": 1.0369, + "step": 159910 + }, + { + "epoch": 13.84, + "learning_rate": 3.618903231395651e-05, + "loss": 1.1014, + "step": 159920 + }, + { + "epoch": 13.84, + "learning_rate": 3.6188165988044706e-05, + "loss": 1.1463, + "step": 159930 + }, + { + "epoch": 13.84, + "learning_rate": 3.6187299662132894e-05, + "loss": 1.1469, + "step": 159940 + }, + { + "epoch": 13.84, + "learning_rate": 3.618643333622109e-05, + "loss": 1.051, + "step": 159950 + }, + { + "epoch": 13.85, + "learning_rate": 3.6185567010309276e-05, + "loss": 1.0471, + "step": 159960 + }, + { + "epoch": 13.85, + "learning_rate": 3.618470068439747e-05, + "loss": 1.0829, + "step": 159970 + }, + { + "epoch": 13.85, + "learning_rate": 3.6183834358485665e-05, + "loss": 1.0961, + "step": 159980 + }, + { + "epoch": 13.85, + "learning_rate": 3.618296803257385e-05, + "loss": 1.0545, + "step": 159990 + }, + { + "epoch": 13.85, + "learning_rate": 3.618210170666205e-05, + "loss": 1.0944, + "step": 160000 + }, + { + "epoch": 13.85, + "learning_rate": 3.618123538075024e-05, + "loss": 1.1046, + "step": 160010 + }, + { + "epoch": 13.85, + "learning_rate": 3.618036905483843e-05, + "loss": 1.0524, + "step": 160020 + }, + { + "epoch": 13.85, + "learning_rate": 3.6179502728926624e-05, + "loss": 1.0855, + "step": 160030 + }, + { + "epoch": 13.85, + "learning_rate": 3.617863640301482e-05, + "loss": 1.0931, + "step": 160040 + }, + { + "epoch": 13.85, + "learning_rate": 3.6177770077103006e-05, + "loss": 1.0792, + "step": 160050 + }, + { + "epoch": 13.85, + "learning_rate": 3.61769037511912e-05, + "loss": 1.0954, + "step": 160060 + }, + { + "epoch": 13.85, + "learning_rate": 3.617603742527939e-05, + "loss": 1.098, + "step": 160070 + }, + { + "epoch": 13.86, + "learning_rate": 3.617517109936758e-05, + "loss": 1.126, + "step": 160080 + }, + { + "epoch": 13.86, + "learning_rate": 3.617430477345578e-05, + "loss": 1.0466, + "step": 160090 + }, + { + "epoch": 13.86, + "learning_rate": 3.6173438447543965e-05, + "loss": 1.0899, + "step": 160100 + }, + { + "epoch": 13.86, + "learning_rate": 3.617257212163216e-05, + "loss": 1.0733, + "step": 160110 + }, + { + "epoch": 13.86, + "learning_rate": 3.6171705795720354e-05, + "loss": 1.1529, + "step": 160120 + }, + { + "epoch": 13.86, + "learning_rate": 3.617083946980854e-05, + "loss": 1.1533, + "step": 160130 + }, + { + "epoch": 13.86, + "learning_rate": 3.6169973143896736e-05, + "loss": 1.1209, + "step": 160140 + }, + { + "epoch": 13.86, + "learning_rate": 3.6169106817984923e-05, + "loss": 1.108, + "step": 160150 + }, + { + "epoch": 13.86, + "learning_rate": 3.616824049207312e-05, + "loss": 1.133, + "step": 160160 + }, + { + "epoch": 13.86, + "learning_rate": 3.616737416616131e-05, + "loss": 1.0817, + "step": 160170 + }, + { + "epoch": 13.86, + "learning_rate": 3.61665078402495e-05, + "loss": 1.1288, + "step": 160180 + }, + { + "epoch": 13.87, + "learning_rate": 3.6165641514337695e-05, + "loss": 1.1137, + "step": 160190 + }, + { + "epoch": 13.87, + "learning_rate": 3.616477518842589e-05, + "loss": 1.129, + "step": 160200 + }, + { + "epoch": 13.87, + "learning_rate": 3.616390886251408e-05, + "loss": 1.0255, + "step": 160210 + }, + { + "epoch": 13.87, + "learning_rate": 3.616304253660227e-05, + "loss": 1.0521, + "step": 160220 + }, + { + "epoch": 13.87, + "learning_rate": 3.6162176210690466e-05, + "loss": 1.1237, + "step": 160230 + }, + { + "epoch": 13.87, + "learning_rate": 3.616130988477865e-05, + "loss": 1.0525, + "step": 160240 + }, + { + "epoch": 13.87, + "learning_rate": 3.616044355886685e-05, + "loss": 1.0984, + "step": 160250 + }, + { + "epoch": 13.87, + "learning_rate": 3.6159577232955035e-05, + "loss": 1.0759, + "step": 160260 + }, + { + "epoch": 13.87, + "learning_rate": 3.615871090704323e-05, + "loss": 1.1101, + "step": 160270 + }, + { + "epoch": 13.87, + "learning_rate": 3.6157844581131424e-05, + "loss": 1.1109, + "step": 160280 + }, + { + "epoch": 13.87, + "learning_rate": 3.615697825521961e-05, + "loss": 1.0316, + "step": 160290 + }, + { + "epoch": 13.87, + "learning_rate": 3.6156111929307806e-05, + "loss": 1.0711, + "step": 160300 + }, + { + "epoch": 13.88, + "learning_rate": 3.6155245603396e-05, + "loss": 1.1604, + "step": 160310 + }, + { + "epoch": 13.88, + "learning_rate": 3.615437927748419e-05, + "loss": 1.1311, + "step": 160320 + }, + { + "epoch": 13.88, + "learning_rate": 3.615351295157238e-05, + "loss": 1.1142, + "step": 160330 + }, + { + "epoch": 13.88, + "learning_rate": 3.615264662566058e-05, + "loss": 1.1074, + "step": 160340 + }, + { + "epoch": 13.88, + "learning_rate": 3.6151780299748765e-05, + "loss": 1.1086, + "step": 160350 + }, + { + "epoch": 13.88, + "learning_rate": 3.615091397383696e-05, + "loss": 1.1182, + "step": 160360 + }, + { + "epoch": 13.88, + "learning_rate": 3.615004764792515e-05, + "loss": 1.1381, + "step": 160370 + }, + { + "epoch": 13.88, + "learning_rate": 3.614918132201334e-05, + "loss": 1.0657, + "step": 160380 + }, + { + "epoch": 13.88, + "learning_rate": 3.6148314996101536e-05, + "loss": 1.076, + "step": 160390 + }, + { + "epoch": 13.88, + "learning_rate": 3.6147448670189724e-05, + "loss": 1.044, + "step": 160400 + }, + { + "epoch": 13.88, + "learning_rate": 3.614658234427792e-05, + "loss": 1.09, + "step": 160410 + }, + { + "epoch": 13.89, + "learning_rate": 3.614571601836611e-05, + "loss": 1.1003, + "step": 160420 + }, + { + "epoch": 13.89, + "learning_rate": 3.61448496924543e-05, + "loss": 1.1064, + "step": 160430 + }, + { + "epoch": 13.89, + "learning_rate": 3.6143983366542495e-05, + "loss": 1.1058, + "step": 160440 + }, + { + "epoch": 13.89, + "learning_rate": 3.614311704063069e-05, + "loss": 1.165, + "step": 160450 + }, + { + "epoch": 13.89, + "learning_rate": 3.614225071471888e-05, + "loss": 1.0421, + "step": 160460 + }, + { + "epoch": 13.89, + "learning_rate": 3.614138438880707e-05, + "loss": 1.0792, + "step": 160470 + }, + { + "epoch": 13.89, + "learning_rate": 3.614051806289526e-05, + "loss": 1.1548, + "step": 160480 + }, + { + "epoch": 13.89, + "learning_rate": 3.6139651736983454e-05, + "loss": 1.0613, + "step": 160490 + }, + { + "epoch": 13.89, + "learning_rate": 3.613878541107165e-05, + "loss": 1.1014, + "step": 160500 + }, + { + "epoch": 13.89, + "learning_rate": 3.6137919085159836e-05, + "loss": 1.1245, + "step": 160510 + }, + { + "epoch": 13.89, + "learning_rate": 3.613705275924803e-05, + "loss": 1.1598, + "step": 160520 + }, + { + "epoch": 13.89, + "learning_rate": 3.6136186433336225e-05, + "loss": 1.1056, + "step": 160530 + }, + { + "epoch": 13.9, + "learning_rate": 3.613532010742441e-05, + "loss": 1.0775, + "step": 160540 + }, + { + "epoch": 13.9, + "learning_rate": 3.613445378151261e-05, + "loss": 1.0974, + "step": 160550 + }, + { + "epoch": 13.9, + "learning_rate": 3.61335874556008e-05, + "loss": 1.0364, + "step": 160560 + }, + { + "epoch": 13.9, + "learning_rate": 3.613272112968899e-05, + "loss": 1.1682, + "step": 160570 + }, + { + "epoch": 13.9, + "learning_rate": 3.613185480377718e-05, + "loss": 1.1035, + "step": 160580 + }, + { + "epoch": 13.9, + "learning_rate": 3.613098847786537e-05, + "loss": 1.0974, + "step": 160590 + }, + { + "epoch": 13.9, + "learning_rate": 3.6130122151953565e-05, + "loss": 1.0398, + "step": 160600 + }, + { + "epoch": 13.9, + "learning_rate": 3.612925582604176e-05, + "loss": 1.0458, + "step": 160610 + }, + { + "epoch": 13.9, + "learning_rate": 3.612838950012995e-05, + "loss": 1.0361, + "step": 160620 + }, + { + "epoch": 13.9, + "learning_rate": 3.612752317421814e-05, + "loss": 1.1263, + "step": 160630 + }, + { + "epoch": 13.9, + "learning_rate": 3.6126656848306336e-05, + "loss": 1.1712, + "step": 160640 + }, + { + "epoch": 13.91, + "learning_rate": 3.6125790522394524e-05, + "loss": 1.1322, + "step": 160650 + }, + { + "epoch": 13.91, + "learning_rate": 3.612492419648272e-05, + "loss": 1.0583, + "step": 160660 + }, + { + "epoch": 13.91, + "learning_rate": 3.612405787057091e-05, + "loss": 1.1092, + "step": 160670 + }, + { + "epoch": 13.91, + "learning_rate": 3.61231915446591e-05, + "loss": 1.0572, + "step": 160680 + }, + { + "epoch": 13.91, + "learning_rate": 3.6122325218747295e-05, + "loss": 1.1251, + "step": 160690 + }, + { + "epoch": 13.91, + "learning_rate": 3.612145889283548e-05, + "loss": 1.0683, + "step": 160700 + }, + { + "epoch": 13.91, + "learning_rate": 3.612059256692368e-05, + "loss": 1.104, + "step": 160710 + }, + { + "epoch": 13.91, + "learning_rate": 3.611972624101187e-05, + "loss": 1.095, + "step": 160720 + }, + { + "epoch": 13.91, + "learning_rate": 3.611885991510006e-05, + "loss": 1.0517, + "step": 160730 + }, + { + "epoch": 13.91, + "learning_rate": 3.6117993589188254e-05, + "loss": 1.0796, + "step": 160740 + }, + { + "epoch": 13.91, + "learning_rate": 3.611712726327645e-05, + "loss": 1.0547, + "step": 160750 + }, + { + "epoch": 13.91, + "learning_rate": 3.6116260937364636e-05, + "loss": 1.0694, + "step": 160760 + }, + { + "epoch": 13.92, + "learning_rate": 3.611539461145283e-05, + "loss": 1.1048, + "step": 160770 + }, + { + "epoch": 13.92, + "learning_rate": 3.6114528285541025e-05, + "loss": 1.0527, + "step": 160780 + }, + { + "epoch": 13.92, + "learning_rate": 3.611366195962921e-05, + "loss": 1.0912, + "step": 160790 + }, + { + "epoch": 13.92, + "learning_rate": 3.611279563371741e-05, + "loss": 1.1083, + "step": 160800 + }, + { + "epoch": 13.92, + "learning_rate": 3.6111929307805595e-05, + "loss": 1.075, + "step": 160810 + }, + { + "epoch": 13.92, + "learning_rate": 3.611106298189379e-05, + "loss": 1.049, + "step": 160820 + }, + { + "epoch": 13.92, + "learning_rate": 3.6110196655981984e-05, + "loss": 1.0549, + "step": 160830 + }, + { + "epoch": 13.92, + "learning_rate": 3.610933033007017e-05, + "loss": 1.0972, + "step": 160840 + }, + { + "epoch": 13.92, + "learning_rate": 3.6108464004158366e-05, + "loss": 1.0673, + "step": 160850 + }, + { + "epoch": 13.92, + "learning_rate": 3.610759767824656e-05, + "loss": 1.1396, + "step": 160860 + }, + { + "epoch": 13.92, + "learning_rate": 3.610673135233475e-05, + "loss": 1.091, + "step": 160870 + }, + { + "epoch": 13.93, + "learning_rate": 3.610586502642294e-05, + "loss": 1.0797, + "step": 160880 + }, + { + "epoch": 13.93, + "learning_rate": 3.610499870051114e-05, + "loss": 1.0628, + "step": 160890 + }, + { + "epoch": 13.93, + "learning_rate": 3.6104132374599324e-05, + "loss": 1.1104, + "step": 160900 + }, + { + "epoch": 13.93, + "learning_rate": 3.610326604868752e-05, + "loss": 1.1002, + "step": 160910 + }, + { + "epoch": 13.93, + "learning_rate": 3.6102399722775707e-05, + "loss": 1.1645, + "step": 160920 + }, + { + "epoch": 13.93, + "learning_rate": 3.61015333968639e-05, + "loss": 1.0869, + "step": 160930 + }, + { + "epoch": 13.93, + "learning_rate": 3.6100667070952096e-05, + "loss": 1.1044, + "step": 160940 + }, + { + "epoch": 13.93, + "learning_rate": 3.609980074504028e-05, + "loss": 1.0443, + "step": 160950 + }, + { + "epoch": 13.93, + "learning_rate": 3.609893441912848e-05, + "loss": 1.126, + "step": 160960 + }, + { + "epoch": 13.93, + "learning_rate": 3.609806809321667e-05, + "loss": 1.0957, + "step": 160970 + }, + { + "epoch": 13.93, + "learning_rate": 3.609720176730486e-05, + "loss": 1.1253, + "step": 160980 + }, + { + "epoch": 13.93, + "learning_rate": 3.6096335441393054e-05, + "loss": 1.1453, + "step": 160990 + }, + { + "epoch": 13.94, + "learning_rate": 3.609546911548124e-05, + "loss": 1.0558, + "step": 161000 + }, + { + "epoch": 13.94, + "learning_rate": 3.6094602789569436e-05, + "loss": 1.0576, + "step": 161010 + }, + { + "epoch": 13.94, + "learning_rate": 3.609373646365763e-05, + "loss": 1.1046, + "step": 161020 + }, + { + "epoch": 13.94, + "learning_rate": 3.609287013774582e-05, + "loss": 1.0751, + "step": 161030 + }, + { + "epoch": 13.94, + "learning_rate": 3.609200381183401e-05, + "loss": 1.1916, + "step": 161040 + }, + { + "epoch": 13.94, + "learning_rate": 3.609113748592221e-05, + "loss": 1.0508, + "step": 161050 + }, + { + "epoch": 13.94, + "learning_rate": 3.6090271160010395e-05, + "loss": 1.0391, + "step": 161060 + }, + { + "epoch": 13.94, + "learning_rate": 3.608940483409859e-05, + "loss": 1.1545, + "step": 161070 + }, + { + "epoch": 13.94, + "learning_rate": 3.6088538508186784e-05, + "loss": 1.103, + "step": 161080 + }, + { + "epoch": 13.94, + "learning_rate": 3.608767218227497e-05, + "loss": 1.0947, + "step": 161090 + }, + { + "epoch": 13.94, + "learning_rate": 3.6086805856363166e-05, + "loss": 1.1327, + "step": 161100 + }, + { + "epoch": 13.94, + "learning_rate": 3.6085939530451354e-05, + "loss": 1.0855, + "step": 161110 + }, + { + "epoch": 13.95, + "learning_rate": 3.608507320453955e-05, + "loss": 1.102, + "step": 161120 + }, + { + "epoch": 13.95, + "learning_rate": 3.608420687862774e-05, + "loss": 1.0147, + "step": 161130 + }, + { + "epoch": 13.95, + "learning_rate": 3.608334055271593e-05, + "loss": 1.0392, + "step": 161140 + }, + { + "epoch": 13.95, + "learning_rate": 3.6082474226804125e-05, + "loss": 1.1079, + "step": 161150 + }, + { + "epoch": 13.95, + "learning_rate": 3.608160790089232e-05, + "loss": 1.1072, + "step": 161160 + }, + { + "epoch": 13.95, + "learning_rate": 3.608074157498051e-05, + "loss": 1.1262, + "step": 161170 + }, + { + "epoch": 13.95, + "learning_rate": 3.60798752490687e-05, + "loss": 1.1694, + "step": 161180 + }, + { + "epoch": 13.95, + "learning_rate": 3.6079008923156896e-05, + "loss": 1.0656, + "step": 161190 + }, + { + "epoch": 13.95, + "learning_rate": 3.6078142597245084e-05, + "loss": 1.1164, + "step": 161200 + }, + { + "epoch": 13.95, + "learning_rate": 3.607727627133328e-05, + "loss": 1.0597, + "step": 161210 + }, + { + "epoch": 13.95, + "learning_rate": 3.6076409945421466e-05, + "loss": 1.0953, + "step": 161220 + }, + { + "epoch": 13.96, + "learning_rate": 3.607554361950966e-05, + "loss": 1.0905, + "step": 161230 + }, + { + "epoch": 13.96, + "learning_rate": 3.6074677293597855e-05, + "loss": 1.0836, + "step": 161240 + }, + { + "epoch": 13.96, + "learning_rate": 3.607381096768604e-05, + "loss": 1.0894, + "step": 161250 + }, + { + "epoch": 13.96, + "learning_rate": 3.607294464177424e-05, + "loss": 1.1737, + "step": 161260 + }, + { + "epoch": 13.96, + "learning_rate": 3.607207831586243e-05, + "loss": 1.071, + "step": 161270 + }, + { + "epoch": 13.96, + "learning_rate": 3.607121198995062e-05, + "loss": 1.073, + "step": 161280 + }, + { + "epoch": 13.96, + "learning_rate": 3.607034566403881e-05, + "loss": 1.0768, + "step": 161290 + }, + { + "epoch": 13.96, + "learning_rate": 3.606947933812701e-05, + "loss": 1.0966, + "step": 161300 + }, + { + "epoch": 13.96, + "learning_rate": 3.6068613012215195e-05, + "loss": 1.0574, + "step": 161310 + }, + { + "epoch": 13.96, + "learning_rate": 3.606774668630339e-05, + "loss": 1.1321, + "step": 161320 + }, + { + "epoch": 13.96, + "learning_rate": 3.606688036039158e-05, + "loss": 1.0969, + "step": 161330 + }, + { + "epoch": 13.96, + "learning_rate": 3.606601403447977e-05, + "loss": 1.089, + "step": 161340 + }, + { + "epoch": 13.97, + "learning_rate": 3.6065147708567966e-05, + "loss": 1.1117, + "step": 161350 + }, + { + "epoch": 13.97, + "learning_rate": 3.6064281382656154e-05, + "loss": 1.127, + "step": 161360 + }, + { + "epoch": 13.97, + "learning_rate": 3.606341505674435e-05, + "loss": 1.1089, + "step": 161370 + }, + { + "epoch": 13.97, + "learning_rate": 3.606254873083254e-05, + "loss": 1.1193, + "step": 161380 + }, + { + "epoch": 13.97, + "learning_rate": 3.606168240492073e-05, + "loss": 1.1518, + "step": 161390 + }, + { + "epoch": 13.97, + "learning_rate": 3.6060816079008925e-05, + "loss": 1.0973, + "step": 161400 + }, + { + "epoch": 13.97, + "learning_rate": 3.605994975309712e-05, + "loss": 1.0737, + "step": 161410 + }, + { + "epoch": 13.97, + "learning_rate": 3.605908342718531e-05, + "loss": 1.1225, + "step": 161420 + }, + { + "epoch": 13.97, + "learning_rate": 3.60582171012735e-05, + "loss": 1.0864, + "step": 161430 + }, + { + "epoch": 13.97, + "learning_rate": 3.605735077536169e-05, + "loss": 1.0785, + "step": 161440 + }, + { + "epoch": 13.97, + "learning_rate": 3.6056484449449884e-05, + "loss": 1.117, + "step": 161450 + }, + { + "epoch": 13.98, + "learning_rate": 3.605561812353808e-05, + "loss": 1.1415, + "step": 161460 + }, + { + "epoch": 13.98, + "learning_rate": 3.6054751797626266e-05, + "loss": 1.1293, + "step": 161470 + }, + { + "epoch": 13.98, + "learning_rate": 3.605388547171446e-05, + "loss": 1.1239, + "step": 161480 + }, + { + "epoch": 13.98, + "learning_rate": 3.6053019145802655e-05, + "loss": 1.0454, + "step": 161490 + }, + { + "epoch": 13.98, + "learning_rate": 3.605215281989084e-05, + "loss": 1.0921, + "step": 161500 + }, + { + "epoch": 13.98, + "learning_rate": 3.605128649397904e-05, + "loss": 1.0464, + "step": 161510 + }, + { + "epoch": 13.98, + "learning_rate": 3.605042016806723e-05, + "loss": 1.0601, + "step": 161520 + }, + { + "epoch": 13.98, + "learning_rate": 3.604955384215542e-05, + "loss": 1.1274, + "step": 161530 + }, + { + "epoch": 13.98, + "learning_rate": 3.6048687516243614e-05, + "loss": 1.13, + "step": 161540 + }, + { + "epoch": 13.98, + "learning_rate": 3.60478211903318e-05, + "loss": 1.05, + "step": 161550 + }, + { + "epoch": 13.98, + "learning_rate": 3.6046954864419996e-05, + "loss": 1.0816, + "step": 161560 + }, + { + "epoch": 13.98, + "learning_rate": 3.604608853850819e-05, + "loss": 1.0639, + "step": 161570 + }, + { + "epoch": 13.99, + "learning_rate": 3.604522221259638e-05, + "loss": 1.0476, + "step": 161580 + }, + { + "epoch": 13.99, + "learning_rate": 3.604435588668457e-05, + "loss": 1.1025, + "step": 161590 + }, + { + "epoch": 13.99, + "learning_rate": 3.604348956077277e-05, + "loss": 1.0499, + "step": 161600 + }, + { + "epoch": 13.99, + "learning_rate": 3.6042623234860954e-05, + "loss": 1.0754, + "step": 161610 + }, + { + "epoch": 13.99, + "learning_rate": 3.604175690894915e-05, + "loss": 1.0814, + "step": 161620 + }, + { + "epoch": 13.99, + "learning_rate": 3.604089058303734e-05, + "loss": 1.0473, + "step": 161630 + }, + { + "epoch": 13.99, + "learning_rate": 3.604002425712553e-05, + "loss": 1.1154, + "step": 161640 + }, + { + "epoch": 13.99, + "learning_rate": 3.6039157931213725e-05, + "loss": 1.0326, + "step": 161650 + }, + { + "epoch": 13.99, + "learning_rate": 3.603829160530191e-05, + "loss": 1.0922, + "step": 161660 + }, + { + "epoch": 13.99, + "learning_rate": 3.603742527939011e-05, + "loss": 1.0992, + "step": 161670 + }, + { + "epoch": 13.99, + "learning_rate": 3.60365589534783e-05, + "loss": 1.0934, + "step": 161680 + }, + { + "epoch": 14.0, + "learning_rate": 3.603569262756649e-05, + "loss": 1.0656, + "step": 161690 + }, + { + "epoch": 14.0, + "learning_rate": 3.6034826301654684e-05, + "loss": 1.1609, + "step": 161700 + }, + { + "epoch": 14.0, + "learning_rate": 3.603395997574288e-05, + "loss": 1.0606, + "step": 161710 + }, + { + "epoch": 14.0, + "learning_rate": 3.6033093649831066e-05, + "loss": 1.0207, + "step": 161720 + }, + { + "epoch": 14.0, + "learning_rate": 3.603222732391926e-05, + "loss": 0.9906, + "step": 161730 + }, + { + "epoch": 14.0, + "learning_rate": 3.603136099800745e-05, + "loss": 1.1145, + "step": 161740 + }, + { + "epoch": 14.0, + "eval_Bleu_1": 0.03949698321091226, + "eval_Bleu_2": 2.6591519790538537e-11, + "eval_Bleu_3": 2.4057804341419326e-14, + "eval_Bleu_4": 7.428419811859049e-16, + "eval_ROUGE_L": 0.08397955252825481, + "eval_cer": 0.9933634760371661, + "eval_em": 0, + "eval_f1": 0.1021844003610808, + "eval_loss": 1.0377905368804932, + "eval_runtime": 2467.2303, + "eval_samples_per_second": 2.081, + "eval_steps_per_second": 2.081, + "eval_wer": 0.9709857831744987, + "step": 161746 + }, + { + "epoch": 14.0, + "learning_rate": 3.603049467209564e-05, + "loss": 1.0706, + "step": 161750 + }, + { + "epoch": 14.0, + "learning_rate": 3.602962834618384e-05, + "loss": 1.0924, + "step": 161760 + }, + { + "epoch": 14.0, + "learning_rate": 3.6028762020272025e-05, + "loss": 1.0735, + "step": 161770 + }, + { + "epoch": 14.0, + "learning_rate": 3.602789569436022e-05, + "loss": 1.063, + "step": 161780 + }, + { + "epoch": 14.0, + "learning_rate": 3.6027029368448414e-05, + "loss": 0.9854, + "step": 161790 + }, + { + "epoch": 14.0, + "learning_rate": 3.60261630425366e-05, + "loss": 1.0165, + "step": 161800 + }, + { + "epoch": 14.01, + "learning_rate": 3.6025296716624796e-05, + "loss": 1.0205, + "step": 161810 + }, + { + "epoch": 14.01, + "learning_rate": 3.602443039071299e-05, + "loss": 1.0888, + "step": 161820 + }, + { + "epoch": 14.01, + "learning_rate": 3.602356406480118e-05, + "loss": 1.0321, + "step": 161830 + }, + { + "epoch": 14.01, + "learning_rate": 3.602269773888937e-05, + "loss": 1.0733, + "step": 161840 + }, + { + "epoch": 14.01, + "learning_rate": 3.602183141297756e-05, + "loss": 1.1617, + "step": 161850 + }, + { + "epoch": 14.01, + "learning_rate": 3.6020965087065755e-05, + "loss": 1.0487, + "step": 161860 + }, + { + "epoch": 14.01, + "learning_rate": 3.602009876115395e-05, + "loss": 1.1052, + "step": 161870 + }, + { + "epoch": 14.01, + "learning_rate": 3.601923243524214e-05, + "loss": 1.0144, + "step": 161880 + }, + { + "epoch": 14.01, + "learning_rate": 3.601836610933033e-05, + "loss": 1.0664, + "step": 161890 + }, + { + "epoch": 14.01, + "learning_rate": 3.6017499783418526e-05, + "loss": 1.0669, + "step": 161900 + }, + { + "epoch": 14.01, + "learning_rate": 3.6016633457506713e-05, + "loss": 1.0693, + "step": 161910 + }, + { + "epoch": 14.02, + "learning_rate": 3.601576713159491e-05, + "loss": 1.0717, + "step": 161920 + }, + { + "epoch": 14.02, + "learning_rate": 3.60149008056831e-05, + "loss": 1.0194, + "step": 161930 + }, + { + "epoch": 14.02, + "learning_rate": 3.601403447977129e-05, + "loss": 1.0692, + "step": 161940 + }, + { + "epoch": 14.02, + "learning_rate": 3.6013168153859485e-05, + "loss": 1.0249, + "step": 161950 + }, + { + "epoch": 14.02, + "learning_rate": 3.601230182794767e-05, + "loss": 1.0732, + "step": 161960 + }, + { + "epoch": 14.02, + "learning_rate": 3.601143550203587e-05, + "loss": 1.0357, + "step": 161970 + }, + { + "epoch": 14.02, + "learning_rate": 3.601056917612406e-05, + "loss": 1.0248, + "step": 161980 + }, + { + "epoch": 14.02, + "learning_rate": 3.600970285021225e-05, + "loss": 1.0103, + "step": 161990 + }, + { + "epoch": 14.02, + "learning_rate": 3.600883652430044e-05, + "loss": 1.0815, + "step": 162000 + }, + { + "epoch": 14.02, + "learning_rate": 3.600797019838864e-05, + "loss": 1.0427, + "step": 162010 + }, + { + "epoch": 14.02, + "learning_rate": 3.6007103872476825e-05, + "loss": 1.0696, + "step": 162020 + }, + { + "epoch": 14.02, + "learning_rate": 3.600623754656502e-05, + "loss": 1.0149, + "step": 162030 + }, + { + "epoch": 14.03, + "learning_rate": 3.6005371220653214e-05, + "loss": 1.0534, + "step": 162040 + }, + { + "epoch": 14.03, + "learning_rate": 3.60045048947414e-05, + "loss": 1.0902, + "step": 162050 + }, + { + "epoch": 14.03, + "learning_rate": 3.6003638568829596e-05, + "loss": 1.0326, + "step": 162060 + }, + { + "epoch": 14.03, + "learning_rate": 3.6002772242917784e-05, + "loss": 1.1144, + "step": 162070 + }, + { + "epoch": 14.03, + "learning_rate": 3.600190591700598e-05, + "loss": 1.0601, + "step": 162080 + }, + { + "epoch": 14.03, + "learning_rate": 3.600103959109417e-05, + "loss": 1.0054, + "step": 162090 + }, + { + "epoch": 14.03, + "learning_rate": 3.600017326518236e-05, + "loss": 1.1121, + "step": 162100 + }, + { + "epoch": 14.03, + "learning_rate": 3.5999306939270555e-05, + "loss": 1.1369, + "step": 162110 + }, + { + "epoch": 14.03, + "learning_rate": 3.599844061335875e-05, + "loss": 1.0536, + "step": 162120 + }, + { + "epoch": 14.03, + "learning_rate": 3.599757428744694e-05, + "loss": 1.0681, + "step": 162130 + }, + { + "epoch": 14.03, + "learning_rate": 3.599670796153513e-05, + "loss": 1.1419, + "step": 162140 + }, + { + "epoch": 14.03, + "learning_rate": 3.5995841635623326e-05, + "loss": 1.0603, + "step": 162150 + }, + { + "epoch": 14.04, + "learning_rate": 3.5994975309711514e-05, + "loss": 1.0552, + "step": 162160 + }, + { + "epoch": 14.04, + "learning_rate": 3.599410898379971e-05, + "loss": 1.1127, + "step": 162170 + }, + { + "epoch": 14.04, + "learning_rate": 3.5993242657887896e-05, + "loss": 1.0689, + "step": 162180 + }, + { + "epoch": 14.04, + "learning_rate": 3.599237633197609e-05, + "loss": 1.1076, + "step": 162190 + }, + { + "epoch": 14.04, + "learning_rate": 3.5991510006064285e-05, + "loss": 1.0299, + "step": 162200 + }, + { + "epoch": 14.04, + "learning_rate": 3.599064368015247e-05, + "loss": 1.0249, + "step": 162210 + }, + { + "epoch": 14.04, + "learning_rate": 3.598977735424067e-05, + "loss": 1.048, + "step": 162220 + }, + { + "epoch": 14.04, + "learning_rate": 3.598891102832886e-05, + "loss": 1.0446, + "step": 162230 + }, + { + "epoch": 14.04, + "learning_rate": 3.598804470241705e-05, + "loss": 1.0703, + "step": 162240 + }, + { + "epoch": 14.04, + "learning_rate": 3.5987178376505244e-05, + "loss": 1.0969, + "step": 162250 + }, + { + "epoch": 14.04, + "learning_rate": 3.598631205059344e-05, + "loss": 1.0353, + "step": 162260 + }, + { + "epoch": 14.05, + "learning_rate": 3.5985445724681626e-05, + "loss": 1.1401, + "step": 162270 + }, + { + "epoch": 14.05, + "learning_rate": 3.598457939876982e-05, + "loss": 1.0848, + "step": 162280 + }, + { + "epoch": 14.05, + "learning_rate": 3.598371307285801e-05, + "loss": 1.0551, + "step": 162290 + }, + { + "epoch": 14.05, + "learning_rate": 3.59828467469462e-05, + "loss": 1.1242, + "step": 162300 + }, + { + "epoch": 14.05, + "learning_rate": 3.59819804210344e-05, + "loss": 1.0214, + "step": 162310 + }, + { + "epoch": 14.05, + "learning_rate": 3.5981114095122584e-05, + "loss": 1.0831, + "step": 162320 + }, + { + "epoch": 14.05, + "learning_rate": 3.598024776921078e-05, + "loss": 1.0485, + "step": 162330 + }, + { + "epoch": 14.05, + "learning_rate": 3.597938144329897e-05, + "loss": 1.071, + "step": 162340 + }, + { + "epoch": 14.05, + "learning_rate": 3.597851511738716e-05, + "loss": 1.114, + "step": 162350 + }, + { + "epoch": 14.05, + "learning_rate": 3.5977648791475355e-05, + "loss": 1.0975, + "step": 162360 + }, + { + "epoch": 14.05, + "learning_rate": 3.597678246556355e-05, + "loss": 1.0411, + "step": 162370 + }, + { + "epoch": 14.05, + "learning_rate": 3.597591613965174e-05, + "loss": 1.1005, + "step": 162380 + }, + { + "epoch": 14.06, + "learning_rate": 3.597504981373993e-05, + "loss": 1.058, + "step": 162390 + }, + { + "epoch": 14.06, + "learning_rate": 3.597418348782812e-05, + "loss": 1.0971, + "step": 162400 + }, + { + "epoch": 14.06, + "learning_rate": 3.5973317161916314e-05, + "loss": 1.1144, + "step": 162410 + }, + { + "epoch": 14.06, + "learning_rate": 3.597245083600451e-05, + "loss": 1.0945, + "step": 162420 + }, + { + "epoch": 14.06, + "learning_rate": 3.5971584510092696e-05, + "loss": 1.136, + "step": 162430 + }, + { + "epoch": 14.06, + "learning_rate": 3.597071818418089e-05, + "loss": 1.1192, + "step": 162440 + }, + { + "epoch": 14.06, + "learning_rate": 3.5969851858269085e-05, + "loss": 1.0519, + "step": 162450 + }, + { + "epoch": 14.06, + "learning_rate": 3.596898553235727e-05, + "loss": 1.0694, + "step": 162460 + }, + { + "epoch": 14.06, + "learning_rate": 3.596811920644547e-05, + "loss": 1.072, + "step": 162470 + }, + { + "epoch": 14.06, + "learning_rate": 3.5967252880533655e-05, + "loss": 1.0725, + "step": 162480 + }, + { + "epoch": 14.06, + "learning_rate": 3.596638655462185e-05, + "loss": 1.0793, + "step": 162490 + }, + { + "epoch": 14.07, + "learning_rate": 3.5965520228710044e-05, + "loss": 1.0266, + "step": 162500 + }, + { + "epoch": 14.07, + "learning_rate": 3.596465390279823e-05, + "loss": 1.1387, + "step": 162510 + }, + { + "epoch": 14.07, + "learning_rate": 3.5963787576886426e-05, + "loss": 1.0376, + "step": 162520 + }, + { + "epoch": 14.07, + "learning_rate": 3.596292125097462e-05, + "loss": 1.0217, + "step": 162530 + }, + { + "epoch": 14.07, + "learning_rate": 3.596205492506281e-05, + "loss": 1.0033, + "step": 162540 + }, + { + "epoch": 14.07, + "learning_rate": 3.5961188599151e-05, + "loss": 1.0439, + "step": 162550 + }, + { + "epoch": 14.07, + "learning_rate": 3.59603222732392e-05, + "loss": 1.053, + "step": 162560 + }, + { + "epoch": 14.07, + "learning_rate": 3.5959455947327385e-05, + "loss": 1.0617, + "step": 162570 + }, + { + "epoch": 14.07, + "learning_rate": 3.595858962141558e-05, + "loss": 1.0512, + "step": 162580 + }, + { + "epoch": 14.07, + "learning_rate": 3.595772329550377e-05, + "loss": 1.0722, + "step": 162590 + }, + { + "epoch": 14.07, + "learning_rate": 3.595685696959196e-05, + "loss": 1.0979, + "step": 162600 + }, + { + "epoch": 14.07, + "learning_rate": 3.5955990643680156e-05, + "loss": 1.0158, + "step": 162610 + }, + { + "epoch": 14.08, + "learning_rate": 3.5955124317768343e-05, + "loss": 1.0981, + "step": 162620 + }, + { + "epoch": 14.08, + "learning_rate": 3.595425799185654e-05, + "loss": 1.0469, + "step": 162630 + }, + { + "epoch": 14.08, + "learning_rate": 3.595339166594473e-05, + "loss": 1.079, + "step": 162640 + }, + { + "epoch": 14.08, + "learning_rate": 3.595252534003292e-05, + "loss": 1.1321, + "step": 162650 + }, + { + "epoch": 14.08, + "learning_rate": 3.5951659014121115e-05, + "loss": 1.0909, + "step": 162660 + }, + { + "epoch": 14.08, + "learning_rate": 3.595079268820931e-05, + "loss": 1.006, + "step": 162670 + }, + { + "epoch": 14.08, + "learning_rate": 3.59499263622975e-05, + "loss": 1.0958, + "step": 162680 + }, + { + "epoch": 14.08, + "learning_rate": 3.594906003638569e-05, + "loss": 1.0469, + "step": 162690 + }, + { + "epoch": 14.08, + "learning_rate": 3.594819371047388e-05, + "loss": 1.0884, + "step": 162700 + }, + { + "epoch": 14.08, + "learning_rate": 3.594732738456207e-05, + "loss": 1.0967, + "step": 162710 + }, + { + "epoch": 14.08, + "learning_rate": 3.594646105865027e-05, + "loss": 1.0337, + "step": 162720 + }, + { + "epoch": 14.09, + "learning_rate": 3.5945594732738455e-05, + "loss": 1.0895, + "step": 162730 + }, + { + "epoch": 14.09, + "learning_rate": 3.594472840682665e-05, + "loss": 1.1147, + "step": 162740 + }, + { + "epoch": 14.09, + "learning_rate": 3.5943862080914844e-05, + "loss": 1.0413, + "step": 162750 + }, + { + "epoch": 14.09, + "learning_rate": 3.594299575500303e-05, + "loss": 1.0587, + "step": 162760 + }, + { + "epoch": 14.09, + "learning_rate": 3.5942129429091226e-05, + "loss": 1.1038, + "step": 162770 + }, + { + "epoch": 14.09, + "learning_rate": 3.594126310317942e-05, + "loss": 1.0419, + "step": 162780 + }, + { + "epoch": 14.09, + "learning_rate": 3.594039677726761e-05, + "loss": 1.0948, + "step": 162790 + }, + { + "epoch": 14.09, + "learning_rate": 3.59395304513558e-05, + "loss": 1.0405, + "step": 162800 + }, + { + "epoch": 14.09, + "learning_rate": 3.593866412544399e-05, + "loss": 1.0692, + "step": 162810 + }, + { + "epoch": 14.09, + "learning_rate": 3.5937797799532185e-05, + "loss": 1.0589, + "step": 162820 + }, + { + "epoch": 14.09, + "learning_rate": 3.593693147362038e-05, + "loss": 1.0761, + "step": 162830 + }, + { + "epoch": 14.09, + "learning_rate": 3.593606514770857e-05, + "loss": 1.0279, + "step": 162840 + }, + { + "epoch": 14.1, + "learning_rate": 3.593519882179676e-05, + "loss": 1.0939, + "step": 162850 + }, + { + "epoch": 14.1, + "learning_rate": 3.5934332495884956e-05, + "loss": 1.016, + "step": 162860 + }, + { + "epoch": 14.1, + "learning_rate": 3.5933466169973144e-05, + "loss": 1.0865, + "step": 162870 + }, + { + "epoch": 14.1, + "learning_rate": 3.593259984406134e-05, + "loss": 1.1144, + "step": 162880 + }, + { + "epoch": 14.1, + "learning_rate": 3.593173351814953e-05, + "loss": 1.0626, + "step": 162890 + }, + { + "epoch": 14.1, + "learning_rate": 3.593086719223772e-05, + "loss": 1.024, + "step": 162900 + }, + { + "epoch": 14.1, + "learning_rate": 3.5930000866325915e-05, + "loss": 1.1408, + "step": 162910 + }, + { + "epoch": 14.1, + "learning_rate": 3.59291345404141e-05, + "loss": 1.1009, + "step": 162920 + }, + { + "epoch": 14.1, + "learning_rate": 3.59282682145023e-05, + "loss": 0.9986, + "step": 162930 + }, + { + "epoch": 14.1, + "learning_rate": 3.592740188859049e-05, + "loss": 1.0969, + "step": 162940 + }, + { + "epoch": 14.1, + "learning_rate": 3.592653556267868e-05, + "loss": 1.0622, + "step": 162950 + }, + { + "epoch": 14.11, + "learning_rate": 3.5925669236766874e-05, + "loss": 1.0952, + "step": 162960 + }, + { + "epoch": 14.11, + "learning_rate": 3.592480291085507e-05, + "loss": 1.0426, + "step": 162970 + }, + { + "epoch": 14.11, + "learning_rate": 3.5923936584943256e-05, + "loss": 1.0559, + "step": 162980 + }, + { + "epoch": 14.11, + "learning_rate": 3.592307025903145e-05, + "loss": 1.1047, + "step": 162990 + }, + { + "epoch": 14.11, + "learning_rate": 3.5922203933119645e-05, + "loss": 0.9798, + "step": 163000 + }, + { + "epoch": 14.11, + "learning_rate": 3.592133760720783e-05, + "loss": 1.0483, + "step": 163010 + }, + { + "epoch": 14.11, + "learning_rate": 3.592047128129603e-05, + "loss": 1.0864, + "step": 163020 + }, + { + "epoch": 14.11, + "learning_rate": 3.5919604955384214e-05, + "loss": 1.0818, + "step": 163030 + }, + { + "epoch": 14.11, + "learning_rate": 3.591873862947241e-05, + "loss": 1.1055, + "step": 163040 + }, + { + "epoch": 14.11, + "learning_rate": 3.59178723035606e-05, + "loss": 1.126, + "step": 163050 + }, + { + "epoch": 14.11, + "learning_rate": 3.591700597764879e-05, + "loss": 1.0505, + "step": 163060 + }, + { + "epoch": 14.11, + "learning_rate": 3.5916139651736985e-05, + "loss": 1.0563, + "step": 163070 + }, + { + "epoch": 14.12, + "learning_rate": 3.591527332582518e-05, + "loss": 1.0098, + "step": 163080 + }, + { + "epoch": 14.12, + "learning_rate": 3.591440699991337e-05, + "loss": 1.101, + "step": 163090 + }, + { + "epoch": 14.12, + "learning_rate": 3.591354067400156e-05, + "loss": 1.0812, + "step": 163100 + }, + { + "epoch": 14.12, + "learning_rate": 3.5912674348089756e-05, + "loss": 1.0778, + "step": 163110 + }, + { + "epoch": 14.12, + "learning_rate": 3.5911808022177944e-05, + "loss": 1.0371, + "step": 163120 + }, + { + "epoch": 14.12, + "learning_rate": 3.591094169626614e-05, + "loss": 1.0571, + "step": 163130 + }, + { + "epoch": 14.12, + "learning_rate": 3.5910075370354326e-05, + "loss": 1.1243, + "step": 163140 + }, + { + "epoch": 14.12, + "learning_rate": 3.590920904444252e-05, + "loss": 1.1191, + "step": 163150 + }, + { + "epoch": 14.12, + "learning_rate": 3.5908342718530715e-05, + "loss": 1.102, + "step": 163160 + }, + { + "epoch": 14.12, + "learning_rate": 3.59074763926189e-05, + "loss": 1.1192, + "step": 163170 + }, + { + "epoch": 14.12, + "learning_rate": 3.59066100667071e-05, + "loss": 1.051, + "step": 163180 + }, + { + "epoch": 14.12, + "learning_rate": 3.590574374079529e-05, + "loss": 1.1082, + "step": 163190 + }, + { + "epoch": 14.13, + "learning_rate": 3.590487741488348e-05, + "loss": 1.0531, + "step": 163200 + }, + { + "epoch": 14.13, + "learning_rate": 3.5904011088971674e-05, + "loss": 1.0801, + "step": 163210 + }, + { + "epoch": 14.13, + "learning_rate": 3.590314476305987e-05, + "loss": 1.0947, + "step": 163220 + }, + { + "epoch": 14.13, + "learning_rate": 3.5902278437148056e-05, + "loss": 1.0806, + "step": 163230 + }, + { + "epoch": 14.13, + "learning_rate": 3.590141211123625e-05, + "loss": 1.0519, + "step": 163240 + }, + { + "epoch": 14.13, + "learning_rate": 3.590054578532444e-05, + "loss": 1.0199, + "step": 163250 + }, + { + "epoch": 14.13, + "learning_rate": 3.589967945941263e-05, + "loss": 1.0662, + "step": 163260 + }, + { + "epoch": 14.13, + "learning_rate": 3.589881313350083e-05, + "loss": 1.088, + "step": 163270 + }, + { + "epoch": 14.13, + "learning_rate": 3.5897946807589015e-05, + "loss": 1.081, + "step": 163280 + }, + { + "epoch": 14.13, + "learning_rate": 3.589708048167721e-05, + "loss": 1.1095, + "step": 163290 + }, + { + "epoch": 14.13, + "learning_rate": 3.5896214155765404e-05, + "loss": 1.0094, + "step": 163300 + }, + { + "epoch": 14.14, + "learning_rate": 3.589534782985359e-05, + "loss": 1.0759, + "step": 163310 + }, + { + "epoch": 14.14, + "learning_rate": 3.5894481503941786e-05, + "loss": 1.0639, + "step": 163320 + }, + { + "epoch": 14.14, + "learning_rate": 3.5893615178029973e-05, + "loss": 1.1046, + "step": 163330 + }, + { + "epoch": 14.14, + "learning_rate": 3.589274885211817e-05, + "loss": 1.0972, + "step": 163340 + }, + { + "epoch": 14.14, + "learning_rate": 3.589188252620636e-05, + "loss": 1.0949, + "step": 163350 + }, + { + "epoch": 14.14, + "learning_rate": 3.589101620029455e-05, + "loss": 1.0562, + "step": 163360 + }, + { + "epoch": 14.14, + "learning_rate": 3.5890149874382744e-05, + "loss": 1.0419, + "step": 163370 + }, + { + "epoch": 14.14, + "learning_rate": 3.588928354847094e-05, + "loss": 1.0701, + "step": 163380 + }, + { + "epoch": 14.14, + "learning_rate": 3.5888417222559127e-05, + "loss": 1.0296, + "step": 163390 + }, + { + "epoch": 14.14, + "learning_rate": 3.588755089664732e-05, + "loss": 1.0639, + "step": 163400 + }, + { + "epoch": 14.14, + "learning_rate": 3.5886684570735516e-05, + "loss": 1.0839, + "step": 163410 + }, + { + "epoch": 14.14, + "learning_rate": 3.58858182448237e-05, + "loss": 1.04, + "step": 163420 + }, + { + "epoch": 14.15, + "learning_rate": 3.58849519189119e-05, + "loss": 1.0861, + "step": 163430 + }, + { + "epoch": 14.15, + "learning_rate": 3.5884085593000085e-05, + "loss": 1.0525, + "step": 163440 + }, + { + "epoch": 14.15, + "learning_rate": 3.588321926708828e-05, + "loss": 1.0697, + "step": 163450 + }, + { + "epoch": 14.15, + "learning_rate": 3.5882352941176474e-05, + "loss": 1.0397, + "step": 163460 + }, + { + "epoch": 14.15, + "learning_rate": 3.588148661526466e-05, + "loss": 1.0409, + "step": 163470 + }, + { + "epoch": 14.15, + "learning_rate": 3.5880620289352856e-05, + "loss": 1.0481, + "step": 163480 + }, + { + "epoch": 14.15, + "learning_rate": 3.587975396344105e-05, + "loss": 1.0505, + "step": 163490 + }, + { + "epoch": 14.15, + "learning_rate": 3.587888763752924e-05, + "loss": 1.0335, + "step": 163500 + }, + { + "epoch": 14.15, + "learning_rate": 3.587802131161743e-05, + "loss": 1.1296, + "step": 163510 + }, + { + "epoch": 14.15, + "learning_rate": 3.587715498570563e-05, + "loss": 1.0254, + "step": 163520 + }, + { + "epoch": 14.15, + "learning_rate": 3.5876288659793815e-05, + "loss": 1.0324, + "step": 163530 + }, + { + "epoch": 14.16, + "learning_rate": 3.587542233388201e-05, + "loss": 1.1136, + "step": 163540 + }, + { + "epoch": 14.16, + "learning_rate": 3.58745560079702e-05, + "loss": 1.1227, + "step": 163550 + }, + { + "epoch": 14.16, + "learning_rate": 3.587368968205839e-05, + "loss": 1.0992, + "step": 163560 + }, + { + "epoch": 14.16, + "learning_rate": 3.5872823356146586e-05, + "loss": 1.0899, + "step": 163570 + }, + { + "epoch": 14.16, + "learning_rate": 3.5871957030234774e-05, + "loss": 1.075, + "step": 163580 + }, + { + "epoch": 14.16, + "learning_rate": 3.587109070432297e-05, + "loss": 1.0625, + "step": 163590 + }, + { + "epoch": 14.16, + "learning_rate": 3.587022437841116e-05, + "loss": 1.0775, + "step": 163600 + }, + { + "epoch": 14.16, + "learning_rate": 3.586935805249935e-05, + "loss": 1.0687, + "step": 163610 + }, + { + "epoch": 14.16, + "learning_rate": 3.5868491726587545e-05, + "loss": 1.0534, + "step": 163620 + }, + { + "epoch": 14.16, + "learning_rate": 3.586762540067574e-05, + "loss": 1.1108, + "step": 163630 + }, + { + "epoch": 14.16, + "learning_rate": 3.586675907476393e-05, + "loss": 1.0892, + "step": 163640 + }, + { + "epoch": 14.16, + "learning_rate": 3.586589274885212e-05, + "loss": 1.0575, + "step": 163650 + }, + { + "epoch": 14.17, + "learning_rate": 3.586502642294031e-05, + "loss": 1.0378, + "step": 163660 + }, + { + "epoch": 14.17, + "learning_rate": 3.5864160097028504e-05, + "loss": 1.0449, + "step": 163670 + }, + { + "epoch": 14.17, + "learning_rate": 3.58632937711167e-05, + "loss": 1.0389, + "step": 163680 + }, + { + "epoch": 14.17, + "learning_rate": 3.5862427445204886e-05, + "loss": 1.0718, + "step": 163690 + }, + { + "epoch": 14.17, + "learning_rate": 3.586156111929308e-05, + "loss": 1.0337, + "step": 163700 + }, + { + "epoch": 14.17, + "learning_rate": 3.5860694793381275e-05, + "loss": 1.1249, + "step": 163710 + }, + { + "epoch": 14.17, + "learning_rate": 3.585982846746946e-05, + "loss": 1.0782, + "step": 163720 + }, + { + "epoch": 14.17, + "learning_rate": 3.585896214155766e-05, + "loss": 1.085, + "step": 163730 + }, + { + "epoch": 14.17, + "learning_rate": 3.585809581564585e-05, + "loss": 1.0647, + "step": 163740 + }, + { + "epoch": 14.17, + "learning_rate": 3.585722948973404e-05, + "loss": 1.0961, + "step": 163750 + }, + { + "epoch": 14.17, + "learning_rate": 3.585636316382223e-05, + "loss": 1.0568, + "step": 163760 + }, + { + "epoch": 14.18, + "learning_rate": 3.585549683791042e-05, + "loss": 1.0872, + "step": 163770 + }, + { + "epoch": 14.18, + "learning_rate": 3.5854630511998615e-05, + "loss": 1.0652, + "step": 163780 + }, + { + "epoch": 14.18, + "learning_rate": 3.585376418608681e-05, + "loss": 1.0787, + "step": 163790 + }, + { + "epoch": 14.18, + "learning_rate": 3.5852897860175e-05, + "loss": 1.0588, + "step": 163800 + }, + { + "epoch": 14.18, + "learning_rate": 3.585203153426319e-05, + "loss": 0.9962, + "step": 163810 + }, + { + "epoch": 14.18, + "learning_rate": 3.5851165208351386e-05, + "loss": 1.07, + "step": 163820 + }, + { + "epoch": 14.18, + "learning_rate": 3.5850298882439574e-05, + "loss": 1.0937, + "step": 163830 + }, + { + "epoch": 14.18, + "learning_rate": 3.584943255652777e-05, + "loss": 1.069, + "step": 163840 + }, + { + "epoch": 14.18, + "learning_rate": 3.584856623061596e-05, + "loss": 1.1033, + "step": 163850 + }, + { + "epoch": 14.18, + "learning_rate": 3.584769990470415e-05, + "loss": 1.0604, + "step": 163860 + }, + { + "epoch": 14.18, + "learning_rate": 3.5846833578792345e-05, + "loss": 1.0693, + "step": 163870 + }, + { + "epoch": 14.18, + "learning_rate": 3.584596725288053e-05, + "loss": 0.98, + "step": 163880 + }, + { + "epoch": 14.19, + "learning_rate": 3.584510092696873e-05, + "loss": 1.0722, + "step": 163890 + }, + { + "epoch": 14.19, + "learning_rate": 3.584423460105692e-05, + "loss": 1.1534, + "step": 163900 + }, + { + "epoch": 14.19, + "learning_rate": 3.584336827514511e-05, + "loss": 1.0718, + "step": 163910 + }, + { + "epoch": 14.19, + "learning_rate": 3.5842501949233304e-05, + "loss": 1.0712, + "step": 163920 + }, + { + "epoch": 14.19, + "learning_rate": 3.58416356233215e-05, + "loss": 1.0879, + "step": 163930 + }, + { + "epoch": 14.19, + "learning_rate": 3.5840769297409686e-05, + "loss": 1.0412, + "step": 163940 + }, + { + "epoch": 14.19, + "learning_rate": 3.583990297149788e-05, + "loss": 1.0738, + "step": 163950 + }, + { + "epoch": 14.19, + "learning_rate": 3.5839036645586075e-05, + "loss": 1.041, + "step": 163960 + }, + { + "epoch": 14.19, + "learning_rate": 3.583817031967426e-05, + "loss": 1.0606, + "step": 163970 + }, + { + "epoch": 14.19, + "learning_rate": 3.583730399376246e-05, + "loss": 1.0943, + "step": 163980 + }, + { + "epoch": 14.19, + "learning_rate": 3.5836437667850645e-05, + "loss": 1.0384, + "step": 163990 + }, + { + "epoch": 14.2, + "learning_rate": 3.583557134193884e-05, + "loss": 1.0326, + "step": 164000 + }, + { + "epoch": 14.2, + "learning_rate": 3.5834705016027034e-05, + "loss": 1.0972, + "step": 164010 + }, + { + "epoch": 14.2, + "learning_rate": 3.583383869011522e-05, + "loss": 1.0663, + "step": 164020 + }, + { + "epoch": 14.2, + "learning_rate": 3.5832972364203416e-05, + "loss": 0.9853, + "step": 164030 + }, + { + "epoch": 14.2, + "learning_rate": 3.583210603829161e-05, + "loss": 1.124, + "step": 164040 + }, + { + "epoch": 14.2, + "learning_rate": 3.58312397123798e-05, + "loss": 1.0482, + "step": 164050 + }, + { + "epoch": 14.2, + "learning_rate": 3.583037338646799e-05, + "loss": 1.1063, + "step": 164060 + }, + { + "epoch": 14.2, + "learning_rate": 3.582950706055618e-05, + "loss": 1.0989, + "step": 164070 + }, + { + "epoch": 14.2, + "learning_rate": 3.5828640734644374e-05, + "loss": 1.0344, + "step": 164080 + }, + { + "epoch": 14.2, + "learning_rate": 3.582777440873257e-05, + "loss": 1.0863, + "step": 164090 + }, + { + "epoch": 14.2, + "learning_rate": 3.5826908082820757e-05, + "loss": 1.0279, + "step": 164100 + }, + { + "epoch": 14.2, + "learning_rate": 3.582604175690895e-05, + "loss": 1.0913, + "step": 164110 + }, + { + "epoch": 14.21, + "learning_rate": 3.5825175430997145e-05, + "loss": 1.0769, + "step": 164120 + }, + { + "epoch": 14.21, + "learning_rate": 3.582430910508533e-05, + "loss": 1.032, + "step": 164130 + }, + { + "epoch": 14.21, + "learning_rate": 3.582344277917353e-05, + "loss": 1.1509, + "step": 164140 + }, + { + "epoch": 14.21, + "learning_rate": 3.582257645326172e-05, + "loss": 1.0402, + "step": 164150 + }, + { + "epoch": 14.21, + "learning_rate": 3.582171012734991e-05, + "loss": 1.0336, + "step": 164160 + }, + { + "epoch": 14.21, + "learning_rate": 3.5820843801438104e-05, + "loss": 1.0973, + "step": 164170 + }, + { + "epoch": 14.21, + "learning_rate": 3.581997747552629e-05, + "loss": 1.0667, + "step": 164180 + }, + { + "epoch": 14.21, + "learning_rate": 3.5819111149614486e-05, + "loss": 1.084, + "step": 164190 + }, + { + "epoch": 14.21, + "learning_rate": 3.581824482370268e-05, + "loss": 1.0694, + "step": 164200 + }, + { + "epoch": 14.21, + "learning_rate": 3.581737849779087e-05, + "loss": 1.1127, + "step": 164210 + }, + { + "epoch": 14.21, + "learning_rate": 3.581651217187906e-05, + "loss": 1.0795, + "step": 164220 + }, + { + "epoch": 14.21, + "learning_rate": 3.581564584596726e-05, + "loss": 1.096, + "step": 164230 + }, + { + "epoch": 14.22, + "learning_rate": 3.5814779520055445e-05, + "loss": 1.1058, + "step": 164240 + }, + { + "epoch": 14.22, + "learning_rate": 3.581391319414364e-05, + "loss": 0.9947, + "step": 164250 + }, + { + "epoch": 14.22, + "learning_rate": 3.5813046868231834e-05, + "loss": 1.0365, + "step": 164260 + }, + { + "epoch": 14.22, + "learning_rate": 3.581218054232002e-05, + "loss": 1.1168, + "step": 164270 + }, + { + "epoch": 14.22, + "learning_rate": 3.5811314216408216e-05, + "loss": 1.045, + "step": 164280 + }, + { + "epoch": 14.22, + "learning_rate": 3.5810447890496404e-05, + "loss": 1.0743, + "step": 164290 + }, + { + "epoch": 14.22, + "learning_rate": 3.58095815645846e-05, + "loss": 1.0732, + "step": 164300 + }, + { + "epoch": 14.22, + "learning_rate": 3.580871523867279e-05, + "loss": 1.0356, + "step": 164310 + }, + { + "epoch": 14.22, + "learning_rate": 3.580784891276098e-05, + "loss": 1.0261, + "step": 164320 + }, + { + "epoch": 14.22, + "learning_rate": 3.5806982586849175e-05, + "loss": 1.0243, + "step": 164330 + }, + { + "epoch": 14.22, + "learning_rate": 3.580611626093737e-05, + "loss": 1.0422, + "step": 164340 + }, + { + "epoch": 14.23, + "learning_rate": 3.580524993502556e-05, + "loss": 1.0657, + "step": 164350 + }, + { + "epoch": 14.23, + "learning_rate": 3.580438360911375e-05, + "loss": 1.0477, + "step": 164360 + }, + { + "epoch": 14.23, + "learning_rate": 3.5803517283201946e-05, + "loss": 1.0825, + "step": 164370 + }, + { + "epoch": 14.23, + "learning_rate": 3.5802650957290133e-05, + "loss": 1.0705, + "step": 164380 + }, + { + "epoch": 14.23, + "learning_rate": 3.580178463137833e-05, + "loss": 1.0561, + "step": 164390 + }, + { + "epoch": 14.23, + "learning_rate": 3.5800918305466516e-05, + "loss": 1.0289, + "step": 164400 + }, + { + "epoch": 14.23, + "learning_rate": 3.580005197955471e-05, + "loss": 1.1107, + "step": 164410 + }, + { + "epoch": 14.23, + "learning_rate": 3.5799185653642905e-05, + "loss": 1.0629, + "step": 164420 + }, + { + "epoch": 14.23, + "learning_rate": 3.579831932773109e-05, + "loss": 1.0709, + "step": 164430 + }, + { + "epoch": 14.23, + "learning_rate": 3.579745300181929e-05, + "loss": 1.0211, + "step": 164440 + }, + { + "epoch": 14.23, + "learning_rate": 3.579658667590748e-05, + "loss": 1.0654, + "step": 164450 + }, + { + "epoch": 14.23, + "learning_rate": 3.579572034999567e-05, + "loss": 1.0638, + "step": 164460 + }, + { + "epoch": 14.24, + "learning_rate": 3.579485402408386e-05, + "loss": 1.0753, + "step": 164470 + }, + { + "epoch": 14.24, + "learning_rate": 3.579398769817206e-05, + "loss": 1.0619, + "step": 164480 + }, + { + "epoch": 14.24, + "learning_rate": 3.5793121372260245e-05, + "loss": 1.0834, + "step": 164490 + }, + { + "epoch": 14.24, + "learning_rate": 3.579225504634844e-05, + "loss": 1.0922, + "step": 164500 + }, + { + "epoch": 14.24, + "learning_rate": 3.579138872043663e-05, + "loss": 1.0593, + "step": 164510 + }, + { + "epoch": 14.24, + "learning_rate": 3.579052239452482e-05, + "loss": 1.0707, + "step": 164520 + }, + { + "epoch": 14.24, + "learning_rate": 3.5789656068613016e-05, + "loss": 1.0164, + "step": 164530 + }, + { + "epoch": 14.24, + "learning_rate": 3.5788789742701204e-05, + "loss": 1.0882, + "step": 164540 + }, + { + "epoch": 14.24, + "learning_rate": 3.57879234167894e-05, + "loss": 1.0269, + "step": 164550 + }, + { + "epoch": 14.24, + "learning_rate": 3.578705709087759e-05, + "loss": 1.0711, + "step": 164560 + }, + { + "epoch": 14.24, + "learning_rate": 3.578619076496578e-05, + "loss": 1.0728, + "step": 164570 + }, + { + "epoch": 14.25, + "learning_rate": 3.5785324439053975e-05, + "loss": 1.1357, + "step": 164580 + }, + { + "epoch": 14.25, + "learning_rate": 3.578445811314217e-05, + "loss": 1.1242, + "step": 164590 + }, + { + "epoch": 14.25, + "learning_rate": 3.578359178723036e-05, + "loss": 1.1236, + "step": 164600 + }, + { + "epoch": 14.25, + "learning_rate": 3.578272546131855e-05, + "loss": 1.089, + "step": 164610 + }, + { + "epoch": 14.25, + "learning_rate": 3.578185913540674e-05, + "loss": 1.0853, + "step": 164620 + }, + { + "epoch": 14.25, + "learning_rate": 3.5780992809494934e-05, + "loss": 1.0601, + "step": 164630 + }, + { + "epoch": 14.25, + "learning_rate": 3.578012648358313e-05, + "loss": 1.0351, + "step": 164640 + }, + { + "epoch": 14.25, + "learning_rate": 3.5779260157671316e-05, + "loss": 1.0113, + "step": 164650 + }, + { + "epoch": 14.25, + "learning_rate": 3.577839383175951e-05, + "loss": 1.112, + "step": 164660 + }, + { + "epoch": 14.25, + "learning_rate": 3.5777527505847705e-05, + "loss": 1.1006, + "step": 164670 + }, + { + "epoch": 14.25, + "learning_rate": 3.577666117993589e-05, + "loss": 1.097, + "step": 164680 + }, + { + "epoch": 14.25, + "learning_rate": 3.577579485402409e-05, + "loss": 1.092, + "step": 164690 + }, + { + "epoch": 14.26, + "learning_rate": 3.577492852811228e-05, + "loss": 1.0056, + "step": 164700 + }, + { + "epoch": 14.26, + "learning_rate": 3.577406220220047e-05, + "loss": 1.0182, + "step": 164710 + }, + { + "epoch": 14.26, + "learning_rate": 3.5773195876288664e-05, + "loss": 1.0899, + "step": 164720 + }, + { + "epoch": 14.26, + "learning_rate": 3.577232955037685e-05, + "loss": 1.0781, + "step": 164730 + }, + { + "epoch": 14.26, + "learning_rate": 3.5771463224465046e-05, + "loss": 1.0615, + "step": 164740 + }, + { + "epoch": 14.26, + "learning_rate": 3.577059689855324e-05, + "loss": 1.0282, + "step": 164750 + }, + { + "epoch": 14.26, + "learning_rate": 3.576973057264143e-05, + "loss": 1.0621, + "step": 164760 + }, + { + "epoch": 14.26, + "learning_rate": 3.576886424672962e-05, + "loss": 1.1263, + "step": 164770 + }, + { + "epoch": 14.26, + "learning_rate": 3.576799792081782e-05, + "loss": 1.0899, + "step": 164780 + }, + { + "epoch": 14.26, + "learning_rate": 3.5767131594906004e-05, + "loss": 1.0373, + "step": 164790 + }, + { + "epoch": 14.26, + "learning_rate": 3.57662652689942e-05, + "loss": 1.1036, + "step": 164800 + }, + { + "epoch": 14.27, + "learning_rate": 3.5765398943082387e-05, + "loss": 1.0484, + "step": 164810 + }, + { + "epoch": 14.27, + "learning_rate": 3.576453261717058e-05, + "loss": 1.0751, + "step": 164820 + }, + { + "epoch": 14.27, + "learning_rate": 3.5763666291258775e-05, + "loss": 1.0927, + "step": 164830 + }, + { + "epoch": 14.27, + "learning_rate": 3.576279996534696e-05, + "loss": 1.1002, + "step": 164840 + }, + { + "epoch": 14.27, + "learning_rate": 3.576193363943516e-05, + "loss": 1.0355, + "step": 164850 + }, + { + "epoch": 14.27, + "learning_rate": 3.576106731352335e-05, + "loss": 1.1029, + "step": 164860 + }, + { + "epoch": 14.27, + "learning_rate": 3.576020098761154e-05, + "loss": 1.0663, + "step": 164870 + }, + { + "epoch": 14.27, + "learning_rate": 3.5759334661699734e-05, + "loss": 1.0936, + "step": 164880 + }, + { + "epoch": 14.27, + "learning_rate": 3.575846833578793e-05, + "loss": 1.0654, + "step": 164890 + }, + { + "epoch": 14.27, + "learning_rate": 3.5757602009876116e-05, + "loss": 1.0701, + "step": 164900 + }, + { + "epoch": 14.27, + "learning_rate": 3.575673568396431e-05, + "loss": 1.0731, + "step": 164910 + }, + { + "epoch": 14.27, + "learning_rate": 3.57558693580525e-05, + "loss": 1.0941, + "step": 164920 + }, + { + "epoch": 14.28, + "learning_rate": 3.575500303214069e-05, + "loss": 1.0601, + "step": 164930 + }, + { + "epoch": 14.28, + "learning_rate": 3.575413670622889e-05, + "loss": 1.0654, + "step": 164940 + }, + { + "epoch": 14.28, + "learning_rate": 3.5753270380317075e-05, + "loss": 1.1419, + "step": 164950 + }, + { + "epoch": 14.28, + "learning_rate": 3.575240405440527e-05, + "loss": 1.0663, + "step": 164960 + }, + { + "epoch": 14.28, + "learning_rate": 3.5751537728493464e-05, + "loss": 1.002, + "step": 164970 + }, + { + "epoch": 14.28, + "learning_rate": 3.575067140258165e-05, + "loss": 1.0838, + "step": 164980 + }, + { + "epoch": 14.28, + "learning_rate": 3.5749805076669846e-05, + "loss": 1.0666, + "step": 164990 + }, + { + "epoch": 14.28, + "learning_rate": 3.574893875075804e-05, + "loss": 1.1208, + "step": 165000 + }, + { + "epoch": 14.28, + "learning_rate": 3.574807242484623e-05, + "loss": 1.0691, + "step": 165010 + }, + { + "epoch": 14.28, + "learning_rate": 3.574720609893442e-05, + "loss": 1.0333, + "step": 165020 + }, + { + "epoch": 14.28, + "learning_rate": 3.574633977302261e-05, + "loss": 0.9844, + "step": 165030 + }, + { + "epoch": 14.29, + "learning_rate": 3.5745473447110805e-05, + "loss": 1.0108, + "step": 165040 + }, + { + "epoch": 14.29, + "learning_rate": 3.5744607121199e-05, + "loss": 1.0222, + "step": 165050 + }, + { + "epoch": 14.29, + "learning_rate": 3.574374079528719e-05, + "loss": 1.058, + "step": 165060 + }, + { + "epoch": 14.29, + "learning_rate": 3.574287446937538e-05, + "loss": 1.0453, + "step": 165070 + }, + { + "epoch": 14.29, + "learning_rate": 3.5742008143463576e-05, + "loss": 1.0551, + "step": 165080 + }, + { + "epoch": 14.29, + "learning_rate": 3.5741141817551763e-05, + "loss": 1.0709, + "step": 165090 + }, + { + "epoch": 14.29, + "learning_rate": 3.574027549163996e-05, + "loss": 1.0672, + "step": 165100 + }, + { + "epoch": 14.29, + "learning_rate": 3.573940916572815e-05, + "loss": 1.0511, + "step": 165110 + }, + { + "epoch": 14.29, + "learning_rate": 3.573854283981634e-05, + "loss": 1.0292, + "step": 165120 + }, + { + "epoch": 14.29, + "learning_rate": 3.5737676513904535e-05, + "loss": 1.0662, + "step": 165130 + }, + { + "epoch": 14.29, + "learning_rate": 3.573681018799272e-05, + "loss": 1.1275, + "step": 165140 + }, + { + "epoch": 14.29, + "learning_rate": 3.573594386208092e-05, + "loss": 1.0373, + "step": 165150 + }, + { + "epoch": 14.3, + "learning_rate": 3.573507753616911e-05, + "loss": 1.0961, + "step": 165160 + }, + { + "epoch": 14.3, + "learning_rate": 3.57342112102573e-05, + "loss": 1.1077, + "step": 165170 + }, + { + "epoch": 14.3, + "learning_rate": 3.573334488434549e-05, + "loss": 1.0915, + "step": 165180 + }, + { + "epoch": 14.3, + "learning_rate": 3.573247855843369e-05, + "loss": 1.0947, + "step": 165190 + }, + { + "epoch": 14.3, + "learning_rate": 3.5731612232521875e-05, + "loss": 1.0525, + "step": 165200 + }, + { + "epoch": 14.3, + "learning_rate": 3.573074590661007e-05, + "loss": 1.0422, + "step": 165210 + }, + { + "epoch": 14.3, + "learning_rate": 3.5729879580698264e-05, + "loss": 0.9753, + "step": 165220 + }, + { + "epoch": 14.3, + "learning_rate": 3.572901325478645e-05, + "loss": 1.0045, + "step": 165230 + }, + { + "epoch": 14.3, + "learning_rate": 3.5728146928874646e-05, + "loss": 1.0989, + "step": 165240 + }, + { + "epoch": 14.3, + "learning_rate": 3.5727280602962834e-05, + "loss": 1.0982, + "step": 165250 + }, + { + "epoch": 14.3, + "learning_rate": 3.572641427705103e-05, + "loss": 1.0239, + "step": 165260 + }, + { + "epoch": 14.31, + "learning_rate": 3.572554795113922e-05, + "loss": 1.1258, + "step": 165270 + }, + { + "epoch": 14.31, + "learning_rate": 3.572468162522741e-05, + "loss": 1.0809, + "step": 165280 + }, + { + "epoch": 14.31, + "learning_rate": 3.5723815299315605e-05, + "loss": 1.0433, + "step": 165290 + }, + { + "epoch": 14.31, + "learning_rate": 3.57229489734038e-05, + "loss": 1.0481, + "step": 165300 + }, + { + "epoch": 14.31, + "learning_rate": 3.572208264749199e-05, + "loss": 1.0224, + "step": 165310 + }, + { + "epoch": 14.31, + "learning_rate": 3.572121632158018e-05, + "loss": 1.0434, + "step": 165320 + }, + { + "epoch": 14.31, + "learning_rate": 3.5720349995668376e-05, + "loss": 0.9749, + "step": 165330 + }, + { + "epoch": 14.31, + "learning_rate": 3.5719483669756564e-05, + "loss": 1.099, + "step": 165340 + }, + { + "epoch": 14.31, + "learning_rate": 3.571861734384476e-05, + "loss": 1.0667, + "step": 165350 + }, + { + "epoch": 14.31, + "learning_rate": 3.5717751017932946e-05, + "loss": 1.0904, + "step": 165360 + }, + { + "epoch": 14.31, + "learning_rate": 3.571688469202114e-05, + "loss": 1.0711, + "step": 165370 + }, + { + "epoch": 14.31, + "learning_rate": 3.5716018366109335e-05, + "loss": 1.0712, + "step": 165380 + }, + { + "epoch": 14.32, + "learning_rate": 3.571515204019752e-05, + "loss": 1.0549, + "step": 165390 + }, + { + "epoch": 14.32, + "learning_rate": 3.571428571428572e-05, + "loss": 1.0579, + "step": 165400 + }, + { + "epoch": 14.32, + "learning_rate": 3.571341938837391e-05, + "loss": 1.047, + "step": 165410 + }, + { + "epoch": 14.32, + "learning_rate": 3.57125530624621e-05, + "loss": 1.041, + "step": 165420 + }, + { + "epoch": 14.32, + "learning_rate": 3.5711686736550294e-05, + "loss": 1.1109, + "step": 165430 + }, + { + "epoch": 14.32, + "learning_rate": 3.571082041063849e-05, + "loss": 1.0804, + "step": 165440 + }, + { + "epoch": 14.32, + "learning_rate": 3.5709954084726676e-05, + "loss": 1.0804, + "step": 165450 + }, + { + "epoch": 14.32, + "learning_rate": 3.570908775881487e-05, + "loss": 1.0605, + "step": 165460 + }, + { + "epoch": 14.32, + "learning_rate": 3.570822143290306e-05, + "loss": 1.0448, + "step": 165470 + }, + { + "epoch": 14.32, + "learning_rate": 3.570735510699125e-05, + "loss": 1.1261, + "step": 165480 + }, + { + "epoch": 14.32, + "learning_rate": 3.570648878107945e-05, + "loss": 1.0733, + "step": 165490 + }, + { + "epoch": 14.32, + "learning_rate": 3.5705622455167634e-05, + "loss": 1.1187, + "step": 165500 + }, + { + "epoch": 14.33, + "learning_rate": 3.570475612925583e-05, + "loss": 1.1188, + "step": 165510 + }, + { + "epoch": 14.33, + "learning_rate": 3.570388980334402e-05, + "loss": 1.0482, + "step": 165520 + }, + { + "epoch": 14.33, + "learning_rate": 3.570302347743221e-05, + "loss": 1.0554, + "step": 165530 + }, + { + "epoch": 14.33, + "learning_rate": 3.5702157151520405e-05, + "loss": 1.0862, + "step": 165540 + }, + { + "epoch": 14.33, + "learning_rate": 3.57012908256086e-05, + "loss": 1.0052, + "step": 165550 + }, + { + "epoch": 14.33, + "learning_rate": 3.570042449969679e-05, + "loss": 1.0591, + "step": 165560 + }, + { + "epoch": 14.33, + "learning_rate": 3.569955817378498e-05, + "loss": 1.1086, + "step": 165570 + }, + { + "epoch": 14.33, + "learning_rate": 3.569869184787317e-05, + "loss": 1.0991, + "step": 165580 + }, + { + "epoch": 14.33, + "learning_rate": 3.5697825521961364e-05, + "loss": 1.0828, + "step": 165590 + }, + { + "epoch": 14.33, + "learning_rate": 3.569695919604956e-05, + "loss": 1.0766, + "step": 165600 + }, + { + "epoch": 14.33, + "learning_rate": 3.5696092870137746e-05, + "loss": 1.0885, + "step": 165610 + }, + { + "epoch": 14.34, + "learning_rate": 3.569522654422594e-05, + "loss": 1.0603, + "step": 165620 + }, + { + "epoch": 14.34, + "learning_rate": 3.5694360218314135e-05, + "loss": 1.0712, + "step": 165630 + }, + { + "epoch": 14.34, + "learning_rate": 3.569349389240232e-05, + "loss": 1.0337, + "step": 165640 + }, + { + "epoch": 14.34, + "learning_rate": 3.569262756649052e-05, + "loss": 1.0354, + "step": 165650 + }, + { + "epoch": 14.34, + "learning_rate": 3.5691761240578705e-05, + "loss": 1.0629, + "step": 165660 + }, + { + "epoch": 14.34, + "learning_rate": 3.56908949146669e-05, + "loss": 1.1486, + "step": 165670 + }, + { + "epoch": 14.34, + "learning_rate": 3.5690028588755094e-05, + "loss": 1.0071, + "step": 165680 + }, + { + "epoch": 14.34, + "learning_rate": 3.568916226284328e-05, + "loss": 1.0757, + "step": 165690 + }, + { + "epoch": 14.34, + "learning_rate": 3.5688295936931476e-05, + "loss": 1.0898, + "step": 165700 + }, + { + "epoch": 14.34, + "learning_rate": 3.568742961101967e-05, + "loss": 1.0393, + "step": 165710 + }, + { + "epoch": 14.34, + "learning_rate": 3.568656328510786e-05, + "loss": 0.9938, + "step": 165720 + }, + { + "epoch": 14.34, + "learning_rate": 3.568569695919605e-05, + "loss": 1.0601, + "step": 165730 + }, + { + "epoch": 14.35, + "learning_rate": 3.568483063328425e-05, + "loss": 1.1222, + "step": 165740 + }, + { + "epoch": 14.35, + "learning_rate": 3.5683964307372435e-05, + "loss": 1.0677, + "step": 165750 + }, + { + "epoch": 14.35, + "learning_rate": 3.568309798146063e-05, + "loss": 1.0022, + "step": 165760 + }, + { + "epoch": 14.35, + "learning_rate": 3.568223165554882e-05, + "loss": 1.0949, + "step": 165770 + }, + { + "epoch": 14.35, + "learning_rate": 3.568136532963701e-05, + "loss": 1.1013, + "step": 165780 + }, + { + "epoch": 14.35, + "learning_rate": 3.5680499003725206e-05, + "loss": 1.0351, + "step": 165790 + }, + { + "epoch": 14.35, + "learning_rate": 3.5679632677813393e-05, + "loss": 1.0974, + "step": 165800 + }, + { + "epoch": 14.35, + "learning_rate": 3.567876635190159e-05, + "loss": 1.05, + "step": 165810 + }, + { + "epoch": 14.35, + "learning_rate": 3.567790002598978e-05, + "loss": 1.0512, + "step": 165820 + }, + { + "epoch": 14.35, + "learning_rate": 3.567703370007797e-05, + "loss": 1.0613, + "step": 165830 + }, + { + "epoch": 14.35, + "learning_rate": 3.5676167374166164e-05, + "loss": 1.0582, + "step": 165840 + }, + { + "epoch": 14.36, + "learning_rate": 3.567530104825436e-05, + "loss": 1.0748, + "step": 165850 + }, + { + "epoch": 14.36, + "learning_rate": 3.5674434722342547e-05, + "loss": 1.0762, + "step": 165860 + }, + { + "epoch": 14.36, + "learning_rate": 3.567356839643074e-05, + "loss": 1.0242, + "step": 165870 + }, + { + "epoch": 14.36, + "learning_rate": 3.567270207051893e-05, + "loss": 1.0434, + "step": 165880 + }, + { + "epoch": 14.36, + "learning_rate": 3.567183574460712e-05, + "loss": 1.0208, + "step": 165890 + }, + { + "epoch": 14.36, + "learning_rate": 3.567096941869532e-05, + "loss": 1.0432, + "step": 165900 + }, + { + "epoch": 14.36, + "learning_rate": 3.5670103092783505e-05, + "loss": 1.1216, + "step": 165910 + }, + { + "epoch": 14.36, + "learning_rate": 3.56692367668717e-05, + "loss": 1.0496, + "step": 165920 + }, + { + "epoch": 14.36, + "learning_rate": 3.5668370440959894e-05, + "loss": 1.0251, + "step": 165930 + }, + { + "epoch": 14.36, + "learning_rate": 3.566750411504808e-05, + "loss": 1.1154, + "step": 165940 + }, + { + "epoch": 14.36, + "learning_rate": 3.5666637789136276e-05, + "loss": 1.1039, + "step": 165950 + }, + { + "epoch": 14.36, + "learning_rate": 3.566577146322447e-05, + "loss": 1.0814, + "step": 165960 + }, + { + "epoch": 14.37, + "learning_rate": 3.566490513731266e-05, + "loss": 1.0899, + "step": 165970 + }, + { + "epoch": 14.37, + "learning_rate": 3.566403881140085e-05, + "loss": 1.0895, + "step": 165980 + }, + { + "epoch": 14.37, + "learning_rate": 3.566317248548904e-05, + "loss": 1.0521, + "step": 165990 + }, + { + "epoch": 14.37, + "learning_rate": 3.5662306159577235e-05, + "loss": 1.096, + "step": 166000 + }, + { + "epoch": 14.37, + "learning_rate": 3.566143983366543e-05, + "loss": 1.0592, + "step": 166010 + }, + { + "epoch": 14.37, + "learning_rate": 3.566057350775362e-05, + "loss": 1.0817, + "step": 166020 + }, + { + "epoch": 14.37, + "learning_rate": 3.565970718184181e-05, + "loss": 1.0262, + "step": 166030 + }, + { + "epoch": 14.37, + "learning_rate": 3.5658840855930006e-05, + "loss": 1.0005, + "step": 166040 + }, + { + "epoch": 14.37, + "learning_rate": 3.5657974530018194e-05, + "loss": 1.0176, + "step": 166050 + }, + { + "epoch": 14.37, + "learning_rate": 3.565710820410639e-05, + "loss": 1.0592, + "step": 166060 + }, + { + "epoch": 14.37, + "learning_rate": 3.565624187819458e-05, + "loss": 1.055, + "step": 166070 + }, + { + "epoch": 14.38, + "learning_rate": 3.565537555228277e-05, + "loss": 1.0605, + "step": 166080 + }, + { + "epoch": 14.38, + "learning_rate": 3.5654509226370965e-05, + "loss": 1.1162, + "step": 166090 + }, + { + "epoch": 14.38, + "learning_rate": 3.565364290045915e-05, + "loss": 1.0579, + "step": 166100 + }, + { + "epoch": 14.38, + "learning_rate": 3.565277657454735e-05, + "loss": 1.1186, + "step": 166110 + }, + { + "epoch": 14.38, + "learning_rate": 3.565191024863554e-05, + "loss": 1.1095, + "step": 166120 + }, + { + "epoch": 14.38, + "learning_rate": 3.565104392272373e-05, + "loss": 1.0921, + "step": 166130 + }, + { + "epoch": 14.38, + "learning_rate": 3.5650177596811924e-05, + "loss": 1.0937, + "step": 166140 + }, + { + "epoch": 14.38, + "learning_rate": 3.564931127090012e-05, + "loss": 1.0592, + "step": 166150 + }, + { + "epoch": 14.38, + "learning_rate": 3.5648444944988306e-05, + "loss": 1.0207, + "step": 166160 + }, + { + "epoch": 14.38, + "learning_rate": 3.56475786190765e-05, + "loss": 1.0452, + "step": 166170 + }, + { + "epoch": 14.38, + "learning_rate": 3.5646712293164695e-05, + "loss": 1.0733, + "step": 166180 + }, + { + "epoch": 14.38, + "learning_rate": 3.564584596725288e-05, + "loss": 1.074, + "step": 166190 + }, + { + "epoch": 14.39, + "learning_rate": 3.564497964134108e-05, + "loss": 1.1385, + "step": 166200 + }, + { + "epoch": 14.39, + "learning_rate": 3.5644113315429264e-05, + "loss": 1.1113, + "step": 166210 + }, + { + "epoch": 14.39, + "learning_rate": 3.564324698951746e-05, + "loss": 1.0834, + "step": 166220 + }, + { + "epoch": 14.39, + "learning_rate": 3.564238066360565e-05, + "loss": 1.12, + "step": 166230 + }, + { + "epoch": 14.39, + "learning_rate": 3.564151433769384e-05, + "loss": 1.0638, + "step": 166240 + }, + { + "epoch": 14.39, + "learning_rate": 3.5640648011782035e-05, + "loss": 1.1299, + "step": 166250 + }, + { + "epoch": 14.39, + "learning_rate": 3.563978168587023e-05, + "loss": 1.0684, + "step": 166260 + }, + { + "epoch": 14.39, + "learning_rate": 3.563891535995842e-05, + "loss": 1.0706, + "step": 166270 + }, + { + "epoch": 14.39, + "learning_rate": 3.563804903404661e-05, + "loss": 1.0118, + "step": 166280 + }, + { + "epoch": 14.39, + "learning_rate": 3.5637182708134806e-05, + "loss": 1.1589, + "step": 166290 + }, + { + "epoch": 14.39, + "learning_rate": 3.5636316382222994e-05, + "loss": 1.0375, + "step": 166300 + }, + { + "epoch": 14.4, + "learning_rate": 3.563545005631119e-05, + "loss": 1.0562, + "step": 166310 + }, + { + "epoch": 14.4, + "learning_rate": 3.5634583730399376e-05, + "loss": 1.0057, + "step": 166320 + }, + { + "epoch": 14.4, + "learning_rate": 3.563371740448757e-05, + "loss": 1.0204, + "step": 166330 + }, + { + "epoch": 14.4, + "learning_rate": 3.5632851078575765e-05, + "loss": 1.0853, + "step": 166340 + }, + { + "epoch": 14.4, + "learning_rate": 3.563198475266395e-05, + "loss": 1.0122, + "step": 166350 + }, + { + "epoch": 14.4, + "learning_rate": 3.563111842675215e-05, + "loss": 1.1016, + "step": 166360 + }, + { + "epoch": 14.4, + "learning_rate": 3.563025210084034e-05, + "loss": 1.0116, + "step": 166370 + }, + { + "epoch": 14.4, + "learning_rate": 3.562938577492853e-05, + "loss": 1.0658, + "step": 166380 + }, + { + "epoch": 14.4, + "learning_rate": 3.5628519449016724e-05, + "loss": 1.1008, + "step": 166390 + }, + { + "epoch": 14.4, + "learning_rate": 3.562765312310491e-05, + "loss": 1.0306, + "step": 166400 + }, + { + "epoch": 14.4, + "learning_rate": 3.5626786797193106e-05, + "loss": 1.0965, + "step": 166410 + }, + { + "epoch": 14.4, + "learning_rate": 3.56259204712813e-05, + "loss": 1.0507, + "step": 166420 + }, + { + "epoch": 14.41, + "learning_rate": 3.562505414536949e-05, + "loss": 1.0477, + "step": 166430 + }, + { + "epoch": 14.41, + "learning_rate": 3.562418781945768e-05, + "loss": 1.0822, + "step": 166440 + }, + { + "epoch": 14.41, + "learning_rate": 3.562332149354588e-05, + "loss": 1.1154, + "step": 166450 + }, + { + "epoch": 14.41, + "learning_rate": 3.5622455167634065e-05, + "loss": 1.0728, + "step": 166460 + }, + { + "epoch": 14.41, + "learning_rate": 3.562158884172226e-05, + "loss": 1.0874, + "step": 166470 + }, + { + "epoch": 14.41, + "learning_rate": 3.5620722515810454e-05, + "loss": 1.0949, + "step": 166480 + }, + { + "epoch": 14.41, + "learning_rate": 3.561985618989864e-05, + "loss": 1.0433, + "step": 166490 + }, + { + "epoch": 14.41, + "learning_rate": 3.5618989863986836e-05, + "loss": 1.0394, + "step": 166500 + }, + { + "epoch": 14.41, + "learning_rate": 3.5618123538075023e-05, + "loss": 1.0471, + "step": 166510 + }, + { + "epoch": 14.41, + "learning_rate": 3.561725721216322e-05, + "loss": 1.0542, + "step": 166520 + }, + { + "epoch": 14.41, + "learning_rate": 3.561639088625141e-05, + "loss": 1.0224, + "step": 166530 + }, + { + "epoch": 14.41, + "learning_rate": 3.56155245603396e-05, + "loss": 1.0794, + "step": 166540 + }, + { + "epoch": 14.42, + "learning_rate": 3.5614658234427794e-05, + "loss": 1.0375, + "step": 166550 + }, + { + "epoch": 14.42, + "learning_rate": 3.561379190851599e-05, + "loss": 1.0918, + "step": 166560 + }, + { + "epoch": 14.42, + "learning_rate": 3.5612925582604177e-05, + "loss": 1.1222, + "step": 166570 + }, + { + "epoch": 14.42, + "learning_rate": 3.561205925669237e-05, + "loss": 1.0723, + "step": 166580 + }, + { + "epoch": 14.42, + "learning_rate": 3.5611192930780565e-05, + "loss": 1.0621, + "step": 166590 + }, + { + "epoch": 14.42, + "learning_rate": 3.561032660486875e-05, + "loss": 1.0366, + "step": 166600 + }, + { + "epoch": 14.42, + "learning_rate": 3.560946027895695e-05, + "loss": 1.0134, + "step": 166610 + }, + { + "epoch": 14.42, + "learning_rate": 3.5608593953045135e-05, + "loss": 1.0361, + "step": 166620 + }, + { + "epoch": 14.42, + "learning_rate": 3.560772762713333e-05, + "loss": 1.0435, + "step": 166630 + }, + { + "epoch": 14.42, + "learning_rate": 3.5606861301221524e-05, + "loss": 1.1122, + "step": 166640 + }, + { + "epoch": 14.42, + "learning_rate": 3.560599497530971e-05, + "loss": 1.1096, + "step": 166650 + }, + { + "epoch": 14.43, + "learning_rate": 3.5605128649397906e-05, + "loss": 1.0872, + "step": 166660 + }, + { + "epoch": 14.43, + "learning_rate": 3.56042623234861e-05, + "loss": 1.0817, + "step": 166670 + }, + { + "epoch": 14.43, + "learning_rate": 3.560339599757429e-05, + "loss": 1.0998, + "step": 166680 + }, + { + "epoch": 14.43, + "learning_rate": 3.560252967166248e-05, + "loss": 1.076, + "step": 166690 + }, + { + "epoch": 14.43, + "learning_rate": 3.560166334575068e-05, + "loss": 1.0661, + "step": 166700 + }, + { + "epoch": 14.43, + "learning_rate": 3.5600797019838865e-05, + "loss": 1.0484, + "step": 166710 + }, + { + "epoch": 14.43, + "learning_rate": 3.559993069392706e-05, + "loss": 1.0678, + "step": 166720 + }, + { + "epoch": 14.43, + "learning_rate": 3.559906436801525e-05, + "loss": 1.0929, + "step": 166730 + }, + { + "epoch": 14.43, + "learning_rate": 3.559819804210344e-05, + "loss": 1.0982, + "step": 166740 + }, + { + "epoch": 14.43, + "learning_rate": 3.5597331716191636e-05, + "loss": 1.0472, + "step": 166750 + }, + { + "epoch": 14.43, + "learning_rate": 3.5596465390279824e-05, + "loss": 1.0699, + "step": 166760 + }, + { + "epoch": 14.43, + "learning_rate": 3.559559906436802e-05, + "loss": 1.0289, + "step": 166770 + }, + { + "epoch": 14.44, + "learning_rate": 3.559473273845621e-05, + "loss": 1.0851, + "step": 166780 + }, + { + "epoch": 14.44, + "learning_rate": 3.55938664125444e-05, + "loss": 1.0654, + "step": 166790 + }, + { + "epoch": 14.44, + "learning_rate": 3.5593000086632595e-05, + "loss": 1.062, + "step": 166800 + }, + { + "epoch": 14.44, + "learning_rate": 3.559213376072079e-05, + "loss": 1.0562, + "step": 166810 + }, + { + "epoch": 14.44, + "learning_rate": 3.559126743480898e-05, + "loss": 1.1096, + "step": 166820 + }, + { + "epoch": 14.44, + "learning_rate": 3.559040110889717e-05, + "loss": 1.0453, + "step": 166830 + }, + { + "epoch": 14.44, + "learning_rate": 3.558953478298536e-05, + "loss": 1.0621, + "step": 166840 + }, + { + "epoch": 14.44, + "learning_rate": 3.5588668457073553e-05, + "loss": 1.0491, + "step": 166850 + }, + { + "epoch": 14.44, + "learning_rate": 3.558780213116175e-05, + "loss": 1.1104, + "step": 166860 + }, + { + "epoch": 14.44, + "learning_rate": 3.5586935805249936e-05, + "loss": 1.0558, + "step": 166870 + }, + { + "epoch": 14.44, + "learning_rate": 3.558606947933813e-05, + "loss": 1.0262, + "step": 166880 + }, + { + "epoch": 14.45, + "learning_rate": 3.5585203153426325e-05, + "loss": 1.0993, + "step": 166890 + }, + { + "epoch": 14.45, + "learning_rate": 3.558433682751451e-05, + "loss": 1.0561, + "step": 166900 + }, + { + "epoch": 14.45, + "learning_rate": 3.558347050160271e-05, + "loss": 1.0217, + "step": 166910 + }, + { + "epoch": 14.45, + "learning_rate": 3.55826041756909e-05, + "loss": 1.0717, + "step": 166920 + }, + { + "epoch": 14.45, + "learning_rate": 3.558173784977909e-05, + "loss": 1.0517, + "step": 166930 + }, + { + "epoch": 14.45, + "learning_rate": 3.558087152386728e-05, + "loss": 1.0527, + "step": 166940 + }, + { + "epoch": 14.45, + "learning_rate": 3.558000519795547e-05, + "loss": 1.0702, + "step": 166950 + }, + { + "epoch": 14.45, + "learning_rate": 3.5579138872043665e-05, + "loss": 1.0579, + "step": 166960 + }, + { + "epoch": 14.45, + "learning_rate": 3.557827254613186e-05, + "loss": 1.0362, + "step": 166970 + }, + { + "epoch": 14.45, + "learning_rate": 3.557740622022005e-05, + "loss": 1.1048, + "step": 166980 + }, + { + "epoch": 14.45, + "learning_rate": 3.557653989430824e-05, + "loss": 1.0818, + "step": 166990 + }, + { + "epoch": 14.45, + "learning_rate": 3.5575673568396436e-05, + "loss": 1.0702, + "step": 167000 + }, + { + "epoch": 14.46, + "learning_rate": 3.5574807242484624e-05, + "loss": 1.0674, + "step": 167010 + }, + { + "epoch": 14.46, + "learning_rate": 3.557394091657282e-05, + "loss": 1.1126, + "step": 167020 + }, + { + "epoch": 14.46, + "learning_rate": 3.557307459066101e-05, + "loss": 1.0694, + "step": 167030 + }, + { + "epoch": 14.46, + "learning_rate": 3.55722082647492e-05, + "loss": 1.0742, + "step": 167040 + }, + { + "epoch": 14.46, + "learning_rate": 3.5571341938837395e-05, + "loss": 1.0635, + "step": 167050 + }, + { + "epoch": 14.46, + "learning_rate": 3.557047561292558e-05, + "loss": 1.1172, + "step": 167060 + }, + { + "epoch": 14.46, + "learning_rate": 3.556960928701378e-05, + "loss": 1.0384, + "step": 167070 + }, + { + "epoch": 14.46, + "learning_rate": 3.556874296110197e-05, + "loss": 1.0462, + "step": 167080 + }, + { + "epoch": 14.46, + "learning_rate": 3.556787663519016e-05, + "loss": 1.0922, + "step": 167090 + }, + { + "epoch": 14.46, + "learning_rate": 3.5567010309278354e-05, + "loss": 1.0412, + "step": 167100 + }, + { + "epoch": 14.46, + "learning_rate": 3.556614398336655e-05, + "loss": 1.0568, + "step": 167110 + }, + { + "epoch": 14.47, + "learning_rate": 3.5565277657454736e-05, + "loss": 1.026, + "step": 167120 + }, + { + "epoch": 14.47, + "learning_rate": 3.556441133154293e-05, + "loss": 1.0411, + "step": 167130 + }, + { + "epoch": 14.47, + "learning_rate": 3.556354500563112e-05, + "loss": 1.1016, + "step": 167140 + }, + { + "epoch": 14.47, + "learning_rate": 3.556267867971931e-05, + "loss": 0.9988, + "step": 167150 + }, + { + "epoch": 14.47, + "learning_rate": 3.556181235380751e-05, + "loss": 1.1065, + "step": 167160 + }, + { + "epoch": 14.47, + "learning_rate": 3.5560946027895695e-05, + "loss": 1.0398, + "step": 167170 + }, + { + "epoch": 14.47, + "learning_rate": 3.556007970198389e-05, + "loss": 1.1009, + "step": 167180 + }, + { + "epoch": 14.47, + "learning_rate": 3.5559213376072084e-05, + "loss": 1.091, + "step": 167190 + }, + { + "epoch": 14.47, + "learning_rate": 3.555834705016027e-05, + "loss": 1.0466, + "step": 167200 + }, + { + "epoch": 14.47, + "learning_rate": 3.5557480724248466e-05, + "loss": 1.05, + "step": 167210 + }, + { + "epoch": 14.47, + "learning_rate": 3.555661439833666e-05, + "loss": 1.0579, + "step": 167220 + }, + { + "epoch": 14.47, + "learning_rate": 3.555574807242485e-05, + "loss": 1.0911, + "step": 167230 + }, + { + "epoch": 14.48, + "learning_rate": 3.555488174651304e-05, + "loss": 1.0759, + "step": 167240 + }, + { + "epoch": 14.48, + "learning_rate": 3.555401542060123e-05, + "loss": 1.0624, + "step": 167250 + }, + { + "epoch": 14.48, + "learning_rate": 3.5553149094689424e-05, + "loss": 1.0693, + "step": 167260 + }, + { + "epoch": 14.48, + "learning_rate": 3.555228276877762e-05, + "loss": 1.0494, + "step": 167270 + }, + { + "epoch": 14.48, + "learning_rate": 3.5551416442865807e-05, + "loss": 1.0195, + "step": 167280 + }, + { + "epoch": 14.48, + "learning_rate": 3.5550550116954e-05, + "loss": 1.077, + "step": 167290 + }, + { + "epoch": 14.48, + "learning_rate": 3.5549683791042195e-05, + "loss": 1.0584, + "step": 167300 + }, + { + "epoch": 14.48, + "learning_rate": 3.554881746513038e-05, + "loss": 1.0973, + "step": 167310 + }, + { + "epoch": 14.48, + "learning_rate": 3.554795113921858e-05, + "loss": 1.1137, + "step": 167320 + }, + { + "epoch": 14.48, + "learning_rate": 3.554708481330677e-05, + "loss": 1.1002, + "step": 167330 + }, + { + "epoch": 14.48, + "learning_rate": 3.554621848739496e-05, + "loss": 1.0617, + "step": 167340 + }, + { + "epoch": 14.49, + "learning_rate": 3.5545352161483154e-05, + "loss": 1.0773, + "step": 167350 + }, + { + "epoch": 14.49, + "learning_rate": 3.554448583557134e-05, + "loss": 1.1144, + "step": 167360 + }, + { + "epoch": 14.49, + "learning_rate": 3.5543619509659536e-05, + "loss": 1.0206, + "step": 167370 + }, + { + "epoch": 14.49, + "learning_rate": 3.554275318374773e-05, + "loss": 1.0878, + "step": 167380 + }, + { + "epoch": 14.49, + "learning_rate": 3.554188685783592e-05, + "loss": 1.0984, + "step": 167390 + }, + { + "epoch": 14.49, + "learning_rate": 3.554102053192411e-05, + "loss": 0.9556, + "step": 167400 + }, + { + "epoch": 14.49, + "learning_rate": 3.554015420601231e-05, + "loss": 1.0209, + "step": 167410 + }, + { + "epoch": 14.49, + "learning_rate": 3.5539287880100495e-05, + "loss": 1.0724, + "step": 167420 + }, + { + "epoch": 14.49, + "learning_rate": 3.553842155418869e-05, + "loss": 1.0738, + "step": 167430 + }, + { + "epoch": 14.49, + "learning_rate": 3.5537555228276884e-05, + "loss": 1.0596, + "step": 167440 + }, + { + "epoch": 14.49, + "learning_rate": 3.553668890236507e-05, + "loss": 1.1523, + "step": 167450 + }, + { + "epoch": 14.49, + "learning_rate": 3.5535822576453266e-05, + "loss": 1.0796, + "step": 167460 + }, + { + "epoch": 14.5, + "learning_rate": 3.5534956250541454e-05, + "loss": 1.0911, + "step": 167470 + }, + { + "epoch": 14.5, + "learning_rate": 3.553408992462965e-05, + "loss": 1.0573, + "step": 167480 + }, + { + "epoch": 14.5, + "learning_rate": 3.553322359871784e-05, + "loss": 1.0846, + "step": 167490 + }, + { + "epoch": 14.5, + "learning_rate": 3.553235727280603e-05, + "loss": 1.021, + "step": 167500 + }, + { + "epoch": 14.5, + "learning_rate": 3.5531490946894225e-05, + "loss": 1.2039, + "step": 167510 + }, + { + "epoch": 14.5, + "learning_rate": 3.553062462098242e-05, + "loss": 1.0358, + "step": 167520 + }, + { + "epoch": 14.5, + "learning_rate": 3.552975829507061e-05, + "loss": 1.0529, + "step": 167530 + }, + { + "epoch": 14.5, + "learning_rate": 3.55288919691588e-05, + "loss": 1.1033, + "step": 167540 + }, + { + "epoch": 14.5, + "learning_rate": 3.5528025643246996e-05, + "loss": 1.0687, + "step": 167550 + }, + { + "epoch": 14.5, + "learning_rate": 3.5527159317335183e-05, + "loss": 1.0844, + "step": 167560 + }, + { + "epoch": 14.5, + "learning_rate": 3.552629299142338e-05, + "loss": 1.0486, + "step": 167570 + }, + { + "epoch": 14.5, + "learning_rate": 3.5525426665511566e-05, + "loss": 1.069, + "step": 167580 + }, + { + "epoch": 14.51, + "learning_rate": 3.552456033959976e-05, + "loss": 1.0826, + "step": 167590 + }, + { + "epoch": 14.51, + "learning_rate": 3.5523694013687954e-05, + "loss": 1.0845, + "step": 167600 + }, + { + "epoch": 14.51, + "learning_rate": 3.552282768777614e-05, + "loss": 1.1059, + "step": 167610 + }, + { + "epoch": 14.51, + "learning_rate": 3.552196136186434e-05, + "loss": 0.9737, + "step": 167620 + }, + { + "epoch": 14.51, + "learning_rate": 3.552109503595253e-05, + "loss": 1.1089, + "step": 167630 + }, + { + "epoch": 14.51, + "learning_rate": 3.552022871004072e-05, + "loss": 1.0477, + "step": 167640 + }, + { + "epoch": 14.51, + "learning_rate": 3.551936238412891e-05, + "loss": 1.0266, + "step": 167650 + }, + { + "epoch": 14.51, + "learning_rate": 3.551849605821711e-05, + "loss": 1.1139, + "step": 167660 + }, + { + "epoch": 14.51, + "learning_rate": 3.5517629732305295e-05, + "loss": 1.1286, + "step": 167670 + }, + { + "epoch": 14.51, + "learning_rate": 3.551676340639349e-05, + "loss": 1.0475, + "step": 167680 + }, + { + "epoch": 14.51, + "learning_rate": 3.551589708048168e-05, + "loss": 1.0598, + "step": 167690 + }, + { + "epoch": 14.52, + "learning_rate": 3.551503075456987e-05, + "loss": 1.0784, + "step": 167700 + }, + { + "epoch": 14.52, + "learning_rate": 3.5514164428658066e-05, + "loss": 1.044, + "step": 167710 + }, + { + "epoch": 14.52, + "learning_rate": 3.5513298102746254e-05, + "loss": 1.1178, + "step": 167720 + }, + { + "epoch": 14.52, + "learning_rate": 3.551243177683445e-05, + "loss": 1.037, + "step": 167730 + }, + { + "epoch": 14.52, + "learning_rate": 3.551156545092264e-05, + "loss": 1.0852, + "step": 167740 + }, + { + "epoch": 14.52, + "learning_rate": 3.551069912501083e-05, + "loss": 1.0988, + "step": 167750 + }, + { + "epoch": 14.52, + "learning_rate": 3.5509832799099025e-05, + "loss": 1.0968, + "step": 167760 + }, + { + "epoch": 14.52, + "learning_rate": 3.550896647318722e-05, + "loss": 1.0494, + "step": 167770 + }, + { + "epoch": 14.52, + "learning_rate": 3.550810014727541e-05, + "loss": 1.052, + "step": 167780 + }, + { + "epoch": 14.52, + "learning_rate": 3.55072338213636e-05, + "loss": 1.0344, + "step": 167790 + }, + { + "epoch": 14.52, + "learning_rate": 3.550636749545179e-05, + "loss": 1.1454, + "step": 167800 + }, + { + "epoch": 14.52, + "learning_rate": 3.5505501169539984e-05, + "loss": 1.0788, + "step": 167810 + }, + { + "epoch": 14.53, + "learning_rate": 3.550463484362818e-05, + "loss": 1.0701, + "step": 167820 + }, + { + "epoch": 14.53, + "learning_rate": 3.5503768517716366e-05, + "loss": 1.0238, + "step": 167830 + }, + { + "epoch": 14.53, + "learning_rate": 3.550290219180456e-05, + "loss": 1.0635, + "step": 167840 + }, + { + "epoch": 14.53, + "learning_rate": 3.5502035865892755e-05, + "loss": 1.154, + "step": 167850 + }, + { + "epoch": 14.53, + "learning_rate": 3.550116953998094e-05, + "loss": 1.0216, + "step": 167860 + }, + { + "epoch": 14.53, + "learning_rate": 3.550030321406914e-05, + "loss": 1.0742, + "step": 167870 + }, + { + "epoch": 14.53, + "learning_rate": 3.549943688815733e-05, + "loss": 1.0496, + "step": 167880 + }, + { + "epoch": 14.53, + "learning_rate": 3.549857056224552e-05, + "loss": 1.042, + "step": 167890 + }, + { + "epoch": 14.53, + "learning_rate": 3.5497704236333714e-05, + "loss": 1.0852, + "step": 167900 + }, + { + "epoch": 14.53, + "learning_rate": 3.54968379104219e-05, + "loss": 1.0886, + "step": 167910 + }, + { + "epoch": 14.53, + "learning_rate": 3.5495971584510096e-05, + "loss": 1.1097, + "step": 167920 + }, + { + "epoch": 14.54, + "learning_rate": 3.549510525859829e-05, + "loss": 1.1069, + "step": 167930 + }, + { + "epoch": 14.54, + "learning_rate": 3.549423893268648e-05, + "loss": 1.1015, + "step": 167940 + }, + { + "epoch": 14.54, + "learning_rate": 3.549337260677467e-05, + "loss": 1.0464, + "step": 167950 + }, + { + "epoch": 14.54, + "learning_rate": 3.549250628086287e-05, + "loss": 1.0768, + "step": 167960 + }, + { + "epoch": 14.54, + "learning_rate": 3.5491639954951054e-05, + "loss": 1.079, + "step": 167970 + }, + { + "epoch": 14.54, + "learning_rate": 3.549077362903925e-05, + "loss": 1.0869, + "step": 167980 + }, + { + "epoch": 14.54, + "learning_rate": 3.5489907303127437e-05, + "loss": 1.0892, + "step": 167990 + }, + { + "epoch": 14.54, + "learning_rate": 3.548904097721563e-05, + "loss": 1.0486, + "step": 168000 + }, + { + "epoch": 14.54, + "learning_rate": 3.5488174651303825e-05, + "loss": 1.05, + "step": 168010 + }, + { + "epoch": 14.54, + "learning_rate": 3.548730832539201e-05, + "loss": 1.0986, + "step": 168020 + }, + { + "epoch": 14.54, + "learning_rate": 3.548644199948021e-05, + "loss": 1.0424, + "step": 168030 + }, + { + "epoch": 14.54, + "learning_rate": 3.54855756735684e-05, + "loss": 1.0558, + "step": 168040 + }, + { + "epoch": 14.55, + "learning_rate": 3.548470934765659e-05, + "loss": 1.0485, + "step": 168050 + }, + { + "epoch": 14.55, + "learning_rate": 3.5483843021744784e-05, + "loss": 1.1222, + "step": 168060 + }, + { + "epoch": 14.55, + "learning_rate": 3.548297669583298e-05, + "loss": 1.069, + "step": 168070 + }, + { + "epoch": 14.55, + "learning_rate": 3.5482110369921166e-05, + "loss": 1.0557, + "step": 168080 + }, + { + "epoch": 14.55, + "learning_rate": 3.548124404400936e-05, + "loss": 1.087, + "step": 168090 + }, + { + "epoch": 14.55, + "learning_rate": 3.548037771809755e-05, + "loss": 1.0626, + "step": 168100 + }, + { + "epoch": 14.55, + "learning_rate": 3.547951139218574e-05, + "loss": 1.0952, + "step": 168110 + }, + { + "epoch": 14.55, + "learning_rate": 3.547864506627394e-05, + "loss": 1.1244, + "step": 168120 + }, + { + "epoch": 14.55, + "learning_rate": 3.5477778740362125e-05, + "loss": 1.047, + "step": 168130 + }, + { + "epoch": 14.55, + "learning_rate": 3.547691241445032e-05, + "loss": 1.0345, + "step": 168140 + }, + { + "epoch": 14.55, + "learning_rate": 3.5476046088538514e-05, + "loss": 1.0592, + "step": 168150 + }, + { + "epoch": 14.56, + "learning_rate": 3.54751797626267e-05, + "loss": 1.0495, + "step": 168160 + }, + { + "epoch": 14.56, + "learning_rate": 3.5474313436714896e-05, + "loss": 1.0263, + "step": 168170 + }, + { + "epoch": 14.56, + "learning_rate": 3.547344711080309e-05, + "loss": 1.0935, + "step": 168180 + }, + { + "epoch": 14.56, + "learning_rate": 3.547258078489128e-05, + "loss": 1.0714, + "step": 168190 + }, + { + "epoch": 14.56, + "learning_rate": 3.547171445897947e-05, + "loss": 1.0367, + "step": 168200 + }, + { + "epoch": 14.56, + "learning_rate": 3.547084813306766e-05, + "loss": 1.0893, + "step": 168210 + }, + { + "epoch": 14.56, + "learning_rate": 3.5469981807155855e-05, + "loss": 1.113, + "step": 168220 + }, + { + "epoch": 14.56, + "learning_rate": 3.546911548124405e-05, + "loss": 1.0663, + "step": 168230 + }, + { + "epoch": 14.56, + "learning_rate": 3.546824915533224e-05, + "loss": 1.105, + "step": 168240 + }, + { + "epoch": 14.56, + "learning_rate": 3.546738282942043e-05, + "loss": 1.09, + "step": 168250 + }, + { + "epoch": 14.56, + "learning_rate": 3.5466516503508626e-05, + "loss": 1.0445, + "step": 168260 + }, + { + "epoch": 14.56, + "learning_rate": 3.5465650177596813e-05, + "loss": 1.0744, + "step": 168270 + }, + { + "epoch": 14.57, + "learning_rate": 3.546478385168501e-05, + "loss": 1.1114, + "step": 168280 + }, + { + "epoch": 14.57, + "learning_rate": 3.54639175257732e-05, + "loss": 1.0413, + "step": 168290 + }, + { + "epoch": 14.57, + "learning_rate": 3.546305119986139e-05, + "loss": 1.0575, + "step": 168300 + }, + { + "epoch": 14.57, + "learning_rate": 3.5462184873949584e-05, + "loss": 1.0544, + "step": 168310 + }, + { + "epoch": 14.57, + "learning_rate": 3.546131854803777e-05, + "loss": 1.0475, + "step": 168320 + }, + { + "epoch": 14.57, + "learning_rate": 3.5460452222125967e-05, + "loss": 0.9793, + "step": 168330 + }, + { + "epoch": 14.57, + "learning_rate": 3.545958589621416e-05, + "loss": 1.058, + "step": 168340 + }, + { + "epoch": 14.57, + "learning_rate": 3.545871957030235e-05, + "loss": 1.1172, + "step": 168350 + }, + { + "epoch": 14.57, + "learning_rate": 3.545785324439054e-05, + "loss": 1.0691, + "step": 168360 + }, + { + "epoch": 14.57, + "learning_rate": 3.545698691847874e-05, + "loss": 1.019, + "step": 168370 + }, + { + "epoch": 14.57, + "learning_rate": 3.5456120592566925e-05, + "loss": 1.0661, + "step": 168380 + }, + { + "epoch": 14.58, + "learning_rate": 3.545525426665512e-05, + "loss": 1.0193, + "step": 168390 + }, + { + "epoch": 14.58, + "learning_rate": 3.5454387940743314e-05, + "loss": 1.1139, + "step": 168400 + }, + { + "epoch": 14.58, + "learning_rate": 3.54535216148315e-05, + "loss": 1.1117, + "step": 168410 + }, + { + "epoch": 14.58, + "learning_rate": 3.5452655288919696e-05, + "loss": 1.0445, + "step": 168420 + }, + { + "epoch": 14.58, + "learning_rate": 3.5451788963007884e-05, + "loss": 1.0395, + "step": 168430 + }, + { + "epoch": 14.58, + "learning_rate": 3.545092263709608e-05, + "loss": 1.0171, + "step": 168440 + }, + { + "epoch": 14.58, + "learning_rate": 3.545005631118427e-05, + "loss": 1.0936, + "step": 168450 + }, + { + "epoch": 14.58, + "learning_rate": 3.544918998527246e-05, + "loss": 1.0531, + "step": 168460 + }, + { + "epoch": 14.58, + "learning_rate": 3.5448323659360655e-05, + "loss": 1.0583, + "step": 168470 + }, + { + "epoch": 14.58, + "learning_rate": 3.544745733344885e-05, + "loss": 1.0603, + "step": 168480 + }, + { + "epoch": 14.58, + "learning_rate": 3.544659100753704e-05, + "loss": 1.1712, + "step": 168490 + }, + { + "epoch": 14.58, + "learning_rate": 3.544572468162523e-05, + "loss": 1.0606, + "step": 168500 + }, + { + "epoch": 14.59, + "learning_rate": 3.5444858355713426e-05, + "loss": 1.0921, + "step": 168510 + }, + { + "epoch": 14.59, + "learning_rate": 3.5443992029801614e-05, + "loss": 1.0367, + "step": 168520 + }, + { + "epoch": 14.59, + "learning_rate": 3.544312570388981e-05, + "loss": 1.0706, + "step": 168530 + }, + { + "epoch": 14.59, + "learning_rate": 3.5442259377977996e-05, + "loss": 1.0515, + "step": 168540 + }, + { + "epoch": 14.59, + "learning_rate": 3.544139305206619e-05, + "loss": 1.0671, + "step": 168550 + }, + { + "epoch": 14.59, + "learning_rate": 3.5440526726154385e-05, + "loss": 1.0462, + "step": 168560 + }, + { + "epoch": 14.59, + "learning_rate": 3.543966040024257e-05, + "loss": 1.0715, + "step": 168570 + }, + { + "epoch": 14.59, + "learning_rate": 3.543879407433077e-05, + "loss": 1.1294, + "step": 168580 + }, + { + "epoch": 14.59, + "learning_rate": 3.543792774841896e-05, + "loss": 1.05, + "step": 168590 + }, + { + "epoch": 14.59, + "learning_rate": 3.543706142250715e-05, + "loss": 1.1075, + "step": 168600 + }, + { + "epoch": 14.59, + "learning_rate": 3.5436195096595344e-05, + "loss": 1.1009, + "step": 168610 + }, + { + "epoch": 14.59, + "learning_rate": 3.543532877068354e-05, + "loss": 1.0588, + "step": 168620 + }, + { + "epoch": 14.6, + "learning_rate": 3.5434462444771726e-05, + "loss": 1.0706, + "step": 168630 + }, + { + "epoch": 14.6, + "learning_rate": 3.543359611885992e-05, + "loss": 1.0175, + "step": 168640 + }, + { + "epoch": 14.6, + "learning_rate": 3.543272979294811e-05, + "loss": 1.1096, + "step": 168650 + }, + { + "epoch": 14.6, + "learning_rate": 3.54318634670363e-05, + "loss": 1.0802, + "step": 168660 + }, + { + "epoch": 14.6, + "learning_rate": 3.54309971411245e-05, + "loss": 1.0993, + "step": 168670 + }, + { + "epoch": 14.6, + "learning_rate": 3.5430130815212684e-05, + "loss": 1.0905, + "step": 168680 + }, + { + "epoch": 14.6, + "learning_rate": 3.542926448930088e-05, + "loss": 1.0255, + "step": 168690 + }, + { + "epoch": 14.6, + "learning_rate": 3.542839816338907e-05, + "loss": 1.1003, + "step": 168700 + }, + { + "epoch": 14.6, + "learning_rate": 3.542753183747726e-05, + "loss": 1.0759, + "step": 168710 + }, + { + "epoch": 14.6, + "learning_rate": 3.5426665511565455e-05, + "loss": 1.0588, + "step": 168720 + }, + { + "epoch": 14.6, + "learning_rate": 3.542579918565364e-05, + "loss": 1.0895, + "step": 168730 + }, + { + "epoch": 14.61, + "learning_rate": 3.542493285974184e-05, + "loss": 1.0575, + "step": 168740 + }, + { + "epoch": 14.61, + "learning_rate": 3.542406653383003e-05, + "loss": 1.073, + "step": 168750 + }, + { + "epoch": 14.61, + "learning_rate": 3.542320020791822e-05, + "loss": 1.0023, + "step": 168760 + }, + { + "epoch": 14.61, + "learning_rate": 3.5422333882006414e-05, + "loss": 1.1161, + "step": 168770 + }, + { + "epoch": 14.61, + "learning_rate": 3.542146755609461e-05, + "loss": 1.0302, + "step": 168780 + }, + { + "epoch": 14.61, + "learning_rate": 3.5420601230182796e-05, + "loss": 1.0108, + "step": 168790 + }, + { + "epoch": 14.61, + "learning_rate": 3.541973490427099e-05, + "loss": 1.0858, + "step": 168800 + }, + { + "epoch": 14.61, + "learning_rate": 3.5418868578359185e-05, + "loss": 0.9798, + "step": 168810 + }, + { + "epoch": 14.61, + "learning_rate": 3.541800225244737e-05, + "loss": 1.0526, + "step": 168820 + }, + { + "epoch": 14.61, + "learning_rate": 3.541713592653557e-05, + "loss": 1.0845, + "step": 168830 + }, + { + "epoch": 14.61, + "learning_rate": 3.5416269600623755e-05, + "loss": 1.0749, + "step": 168840 + }, + { + "epoch": 14.61, + "learning_rate": 3.541540327471195e-05, + "loss": 1.1344, + "step": 168850 + }, + { + "epoch": 14.62, + "learning_rate": 3.5414536948800144e-05, + "loss": 1.0449, + "step": 168860 + }, + { + "epoch": 14.62, + "learning_rate": 3.541367062288833e-05, + "loss": 1.0761, + "step": 168870 + }, + { + "epoch": 14.62, + "learning_rate": 3.5412804296976526e-05, + "loss": 1.0459, + "step": 168880 + }, + { + "epoch": 14.62, + "learning_rate": 3.541193797106472e-05, + "loss": 1.0343, + "step": 168890 + }, + { + "epoch": 14.62, + "learning_rate": 3.541107164515291e-05, + "loss": 1.0293, + "step": 168900 + }, + { + "epoch": 14.62, + "learning_rate": 3.54102053192411e-05, + "loss": 1.04, + "step": 168910 + }, + { + "epoch": 14.62, + "learning_rate": 3.54093389933293e-05, + "loss": 1.0602, + "step": 168920 + }, + { + "epoch": 14.62, + "learning_rate": 3.5408472667417485e-05, + "loss": 1.085, + "step": 168930 + }, + { + "epoch": 14.62, + "learning_rate": 3.540760634150568e-05, + "loss": 1.0432, + "step": 168940 + }, + { + "epoch": 14.62, + "learning_rate": 3.540674001559387e-05, + "loss": 1.0681, + "step": 168950 + }, + { + "epoch": 14.62, + "learning_rate": 3.540587368968206e-05, + "loss": 1.1422, + "step": 168960 + }, + { + "epoch": 14.63, + "learning_rate": 3.5405007363770256e-05, + "loss": 1.0327, + "step": 168970 + }, + { + "epoch": 14.63, + "learning_rate": 3.5404141037858443e-05, + "loss": 1.0525, + "step": 168980 + }, + { + "epoch": 14.63, + "learning_rate": 3.540327471194664e-05, + "loss": 1.1512, + "step": 168990 + }, + { + "epoch": 14.63, + "learning_rate": 3.540240838603483e-05, + "loss": 1.0257, + "step": 169000 + }, + { + "epoch": 14.63, + "learning_rate": 3.540154206012302e-05, + "loss": 1.0314, + "step": 169010 + }, + { + "epoch": 14.63, + "learning_rate": 3.5400675734211214e-05, + "loss": 1.1152, + "step": 169020 + }, + { + "epoch": 14.63, + "learning_rate": 3.539980940829941e-05, + "loss": 1.1008, + "step": 169030 + }, + { + "epoch": 14.63, + "learning_rate": 3.5398943082387597e-05, + "loss": 1.0771, + "step": 169040 + }, + { + "epoch": 14.63, + "learning_rate": 3.539807675647579e-05, + "loss": 1.0789, + "step": 169050 + }, + { + "epoch": 14.63, + "learning_rate": 3.539721043056398e-05, + "loss": 1.0421, + "step": 169060 + }, + { + "epoch": 14.63, + "learning_rate": 3.539634410465217e-05, + "loss": 1.0554, + "step": 169070 + }, + { + "epoch": 14.63, + "learning_rate": 3.539547777874037e-05, + "loss": 1.0958, + "step": 169080 + }, + { + "epoch": 14.64, + "learning_rate": 3.5394611452828555e-05, + "loss": 1.1217, + "step": 169090 + }, + { + "epoch": 14.64, + "learning_rate": 3.539374512691675e-05, + "loss": 1.0888, + "step": 169100 + }, + { + "epoch": 14.64, + "learning_rate": 3.5392878801004944e-05, + "loss": 1.0731, + "step": 169110 + }, + { + "epoch": 14.64, + "learning_rate": 3.539201247509313e-05, + "loss": 1.0909, + "step": 169120 + }, + { + "epoch": 14.64, + "learning_rate": 3.5391146149181326e-05, + "loss": 1.0494, + "step": 169130 + }, + { + "epoch": 14.64, + "learning_rate": 3.539027982326952e-05, + "loss": 1.0347, + "step": 169140 + }, + { + "epoch": 14.64, + "learning_rate": 3.538941349735771e-05, + "loss": 0.9874, + "step": 169150 + }, + { + "epoch": 14.64, + "learning_rate": 3.53885471714459e-05, + "loss": 1.04, + "step": 169160 + }, + { + "epoch": 14.64, + "learning_rate": 3.538768084553409e-05, + "loss": 1.0661, + "step": 169170 + }, + { + "epoch": 14.64, + "learning_rate": 3.5386814519622285e-05, + "loss": 1.0332, + "step": 169180 + }, + { + "epoch": 14.64, + "learning_rate": 3.538594819371048e-05, + "loss": 1.011, + "step": 169190 + }, + { + "epoch": 14.65, + "learning_rate": 3.538508186779867e-05, + "loss": 1.0471, + "step": 169200 + }, + { + "epoch": 14.65, + "learning_rate": 3.538421554188686e-05, + "loss": 1.0885, + "step": 169210 + }, + { + "epoch": 14.65, + "learning_rate": 3.5383349215975056e-05, + "loss": 1.0613, + "step": 169220 + }, + { + "epoch": 14.65, + "learning_rate": 3.5382482890063244e-05, + "loss": 1.034, + "step": 169230 + }, + { + "epoch": 14.65, + "learning_rate": 3.538161656415144e-05, + "loss": 1.0639, + "step": 169240 + }, + { + "epoch": 14.65, + "learning_rate": 3.538075023823963e-05, + "loss": 1.0427, + "step": 169250 + }, + { + "epoch": 14.65, + "learning_rate": 3.537988391232782e-05, + "loss": 1.052, + "step": 169260 + }, + { + "epoch": 14.65, + "learning_rate": 3.5379017586416015e-05, + "loss": 1.0526, + "step": 169270 + }, + { + "epoch": 14.65, + "learning_rate": 3.53781512605042e-05, + "loss": 1.0648, + "step": 169280 + }, + { + "epoch": 14.65, + "learning_rate": 3.53772849345924e-05, + "loss": 1.1358, + "step": 169290 + }, + { + "epoch": 14.65, + "learning_rate": 3.537641860868059e-05, + "loss": 1.0608, + "step": 169300 + }, + { + "epoch": 14.65, + "learning_rate": 3.537555228276878e-05, + "loss": 1.106, + "step": 169310 + }, + { + "epoch": 14.66, + "learning_rate": 3.5374685956856973e-05, + "loss": 1.0668, + "step": 169320 + }, + { + "epoch": 14.66, + "learning_rate": 3.537381963094517e-05, + "loss": 1.0693, + "step": 169330 + }, + { + "epoch": 14.66, + "learning_rate": 3.5372953305033356e-05, + "loss": 1.1057, + "step": 169340 + }, + { + "epoch": 14.66, + "learning_rate": 3.537208697912155e-05, + "loss": 1.0295, + "step": 169350 + }, + { + "epoch": 14.66, + "learning_rate": 3.5371220653209745e-05, + "loss": 1.1075, + "step": 169360 + }, + { + "epoch": 14.66, + "learning_rate": 3.537035432729793e-05, + "loss": 1.0308, + "step": 169370 + }, + { + "epoch": 14.66, + "learning_rate": 3.536948800138613e-05, + "loss": 1.0735, + "step": 169380 + }, + { + "epoch": 14.66, + "learning_rate": 3.5368621675474314e-05, + "loss": 1.0045, + "step": 169390 + }, + { + "epoch": 14.66, + "learning_rate": 3.536775534956251e-05, + "loss": 1.0839, + "step": 169400 + }, + { + "epoch": 14.66, + "learning_rate": 3.53668890236507e-05, + "loss": 1.061, + "step": 169410 + }, + { + "epoch": 14.66, + "learning_rate": 3.536602269773889e-05, + "loss": 1.06, + "step": 169420 + }, + { + "epoch": 14.67, + "learning_rate": 3.5365156371827085e-05, + "loss": 1.0627, + "step": 169430 + }, + { + "epoch": 14.67, + "learning_rate": 3.536429004591528e-05, + "loss": 1.0519, + "step": 169440 + }, + { + "epoch": 14.67, + "learning_rate": 3.536342372000347e-05, + "loss": 1.0837, + "step": 169450 + }, + { + "epoch": 14.67, + "learning_rate": 3.536255739409166e-05, + "loss": 1.037, + "step": 169460 + }, + { + "epoch": 14.67, + "learning_rate": 3.536169106817985e-05, + "loss": 1.0617, + "step": 169470 + }, + { + "epoch": 14.67, + "learning_rate": 3.5360824742268044e-05, + "loss": 1.0798, + "step": 169480 + }, + { + "epoch": 14.67, + "learning_rate": 3.535995841635624e-05, + "loss": 1.0943, + "step": 169490 + }, + { + "epoch": 14.67, + "learning_rate": 3.5359092090444426e-05, + "loss": 1.0674, + "step": 169500 + }, + { + "epoch": 14.67, + "learning_rate": 3.535822576453262e-05, + "loss": 1.0768, + "step": 169510 + }, + { + "epoch": 14.67, + "learning_rate": 3.5357359438620815e-05, + "loss": 1.0869, + "step": 169520 + }, + { + "epoch": 14.67, + "learning_rate": 3.5356493112709e-05, + "loss": 1.1148, + "step": 169530 + }, + { + "epoch": 14.67, + "learning_rate": 3.53556267867972e-05, + "loss": 1.1108, + "step": 169540 + }, + { + "epoch": 14.68, + "learning_rate": 3.535476046088539e-05, + "loss": 1.0624, + "step": 169550 + }, + { + "epoch": 14.68, + "learning_rate": 3.535389413497358e-05, + "loss": 1.0643, + "step": 169560 + }, + { + "epoch": 14.68, + "learning_rate": 3.5353027809061774e-05, + "loss": 1.0291, + "step": 169570 + }, + { + "epoch": 14.68, + "learning_rate": 3.535216148314996e-05, + "loss": 1.1029, + "step": 169580 + }, + { + "epoch": 14.68, + "learning_rate": 3.5351295157238156e-05, + "loss": 1.0605, + "step": 169590 + }, + { + "epoch": 14.68, + "learning_rate": 3.535042883132635e-05, + "loss": 1.0789, + "step": 169600 + }, + { + "epoch": 14.68, + "learning_rate": 3.534956250541454e-05, + "loss": 1.0459, + "step": 169610 + }, + { + "epoch": 14.68, + "learning_rate": 3.534869617950273e-05, + "loss": 1.1029, + "step": 169620 + }, + { + "epoch": 14.68, + "learning_rate": 3.534782985359093e-05, + "loss": 1.0216, + "step": 169630 + }, + { + "epoch": 14.68, + "learning_rate": 3.5346963527679115e-05, + "loss": 1.088, + "step": 169640 + }, + { + "epoch": 14.68, + "learning_rate": 3.534609720176731e-05, + "loss": 1.0278, + "step": 169650 + }, + { + "epoch": 14.68, + "learning_rate": 3.5345230875855504e-05, + "loss": 1.0619, + "step": 169660 + }, + { + "epoch": 14.69, + "learning_rate": 3.534436454994369e-05, + "loss": 1.1119, + "step": 169670 + }, + { + "epoch": 14.69, + "learning_rate": 3.5343498224031886e-05, + "loss": 1.026, + "step": 169680 + }, + { + "epoch": 14.69, + "learning_rate": 3.534263189812007e-05, + "loss": 1.0413, + "step": 169690 + }, + { + "epoch": 14.69, + "learning_rate": 3.534176557220827e-05, + "loss": 1.1172, + "step": 169700 + }, + { + "epoch": 14.69, + "learning_rate": 3.534089924629646e-05, + "loss": 1.0533, + "step": 169710 + }, + { + "epoch": 14.69, + "learning_rate": 3.534003292038465e-05, + "loss": 1.0289, + "step": 169720 + }, + { + "epoch": 14.69, + "learning_rate": 3.5339166594472844e-05, + "loss": 1.0125, + "step": 169730 + }, + { + "epoch": 14.69, + "learning_rate": 3.533830026856104e-05, + "loss": 1.0718, + "step": 169740 + }, + { + "epoch": 14.69, + "learning_rate": 3.5337433942649227e-05, + "loss": 1.1064, + "step": 169750 + }, + { + "epoch": 14.69, + "learning_rate": 3.533656761673742e-05, + "loss": 0.9904, + "step": 169760 + }, + { + "epoch": 14.69, + "learning_rate": 3.5335701290825615e-05, + "loss": 1.0444, + "step": 169770 + }, + { + "epoch": 14.7, + "learning_rate": 3.53348349649138e-05, + "loss": 1.0999, + "step": 169780 + }, + { + "epoch": 14.7, + "learning_rate": 3.5333968639002e-05, + "loss": 1.0439, + "step": 169790 + }, + { + "epoch": 14.7, + "learning_rate": 3.5333102313090185e-05, + "loss": 1.0812, + "step": 169800 + }, + { + "epoch": 14.7, + "learning_rate": 3.533223598717838e-05, + "loss": 1.0452, + "step": 169810 + }, + { + "epoch": 14.7, + "learning_rate": 3.5331369661266574e-05, + "loss": 1.0672, + "step": 169820 + }, + { + "epoch": 14.7, + "learning_rate": 3.533050333535476e-05, + "loss": 1.0297, + "step": 169830 + }, + { + "epoch": 14.7, + "learning_rate": 3.5329637009442956e-05, + "loss": 1.1002, + "step": 169840 + }, + { + "epoch": 14.7, + "learning_rate": 3.532877068353115e-05, + "loss": 1.0021, + "step": 169850 + }, + { + "epoch": 14.7, + "learning_rate": 3.532790435761934e-05, + "loss": 1.0668, + "step": 169860 + }, + { + "epoch": 14.7, + "learning_rate": 3.532703803170753e-05, + "loss": 1.0764, + "step": 169870 + }, + { + "epoch": 14.7, + "learning_rate": 3.532617170579573e-05, + "loss": 1.0525, + "step": 169880 + }, + { + "epoch": 14.7, + "learning_rate": 3.5325305379883915e-05, + "loss": 1.1048, + "step": 169890 + }, + { + "epoch": 14.71, + "learning_rate": 3.532443905397211e-05, + "loss": 1.1304, + "step": 169900 + }, + { + "epoch": 14.71, + "learning_rate": 3.53235727280603e-05, + "loss": 1.1187, + "step": 169910 + }, + { + "epoch": 14.71, + "learning_rate": 3.532270640214849e-05, + "loss": 1.0871, + "step": 169920 + }, + { + "epoch": 14.71, + "learning_rate": 3.5321840076236686e-05, + "loss": 1.0989, + "step": 169930 + }, + { + "epoch": 14.71, + "learning_rate": 3.5320973750324874e-05, + "loss": 1.0514, + "step": 169940 + }, + { + "epoch": 14.71, + "learning_rate": 3.532010742441307e-05, + "loss": 1.1264, + "step": 169950 + }, + { + "epoch": 14.71, + "learning_rate": 3.531924109850126e-05, + "loss": 1.1274, + "step": 169960 + }, + { + "epoch": 14.71, + "learning_rate": 3.531837477258945e-05, + "loss": 1.0908, + "step": 169970 + }, + { + "epoch": 14.71, + "learning_rate": 3.5317508446677645e-05, + "loss": 1.0317, + "step": 169980 + }, + { + "epoch": 14.71, + "learning_rate": 3.531664212076584e-05, + "loss": 1.0324, + "step": 169990 + }, + { + "epoch": 14.71, + "learning_rate": 3.531577579485403e-05, + "loss": 1.1028, + "step": 170000 + }, + { + "epoch": 14.72, + "learning_rate": 3.531490946894222e-05, + "loss": 1.0005, + "step": 170010 + }, + { + "epoch": 14.72, + "learning_rate": 3.531404314303041e-05, + "loss": 1.0725, + "step": 170020 + }, + { + "epoch": 14.72, + "learning_rate": 3.5313176817118603e-05, + "loss": 1.0436, + "step": 170030 + }, + { + "epoch": 14.72, + "learning_rate": 3.53123104912068e-05, + "loss": 1.0366, + "step": 170040 + }, + { + "epoch": 14.72, + "learning_rate": 3.5311444165294986e-05, + "loss": 1.0348, + "step": 170050 + }, + { + "epoch": 14.72, + "learning_rate": 3.531057783938318e-05, + "loss": 1.1501, + "step": 170060 + }, + { + "epoch": 14.72, + "learning_rate": 3.5309711513471374e-05, + "loss": 1.1002, + "step": 170070 + }, + { + "epoch": 14.72, + "learning_rate": 3.530884518755956e-05, + "loss": 1.0188, + "step": 170080 + }, + { + "epoch": 14.72, + "learning_rate": 3.530797886164776e-05, + "loss": 1.0881, + "step": 170090 + }, + { + "epoch": 14.72, + "learning_rate": 3.530711253573595e-05, + "loss": 1.0413, + "step": 170100 + }, + { + "epoch": 14.72, + "learning_rate": 3.530624620982414e-05, + "loss": 1.0717, + "step": 170110 + }, + { + "epoch": 14.72, + "learning_rate": 3.530537988391233e-05, + "loss": 1.1147, + "step": 170120 + }, + { + "epoch": 14.73, + "learning_rate": 3.530451355800052e-05, + "loss": 1.1007, + "step": 170130 + }, + { + "epoch": 14.73, + "learning_rate": 3.5303647232088715e-05, + "loss": 1.0652, + "step": 170140 + }, + { + "epoch": 14.73, + "learning_rate": 3.530278090617691e-05, + "loss": 1.0708, + "step": 170150 + }, + { + "epoch": 14.73, + "learning_rate": 3.53019145802651e-05, + "loss": 1.0265, + "step": 170160 + }, + { + "epoch": 14.73, + "learning_rate": 3.530104825435329e-05, + "loss": 1.0214, + "step": 170170 + }, + { + "epoch": 14.73, + "learning_rate": 3.5300181928441486e-05, + "loss": 1.0811, + "step": 170180 + }, + { + "epoch": 14.73, + "learning_rate": 3.5299315602529674e-05, + "loss": 1.0665, + "step": 170190 + }, + { + "epoch": 14.73, + "learning_rate": 3.529844927661787e-05, + "loss": 1.0783, + "step": 170200 + }, + { + "epoch": 14.73, + "learning_rate": 3.529758295070606e-05, + "loss": 1.0773, + "step": 170210 + }, + { + "epoch": 14.73, + "learning_rate": 3.529671662479425e-05, + "loss": 1.0483, + "step": 170220 + }, + { + "epoch": 14.73, + "learning_rate": 3.5295850298882445e-05, + "loss": 1.098, + "step": 170230 + }, + { + "epoch": 14.74, + "learning_rate": 3.529498397297063e-05, + "loss": 1.0409, + "step": 170240 + }, + { + "epoch": 14.74, + "learning_rate": 3.529411764705883e-05, + "loss": 1.0607, + "step": 170250 + }, + { + "epoch": 14.74, + "learning_rate": 3.529325132114702e-05, + "loss": 1.0661, + "step": 170260 + }, + { + "epoch": 14.74, + "learning_rate": 3.529238499523521e-05, + "loss": 1.0288, + "step": 170270 + }, + { + "epoch": 14.74, + "learning_rate": 3.5291518669323404e-05, + "loss": 1.0693, + "step": 170280 + }, + { + "epoch": 14.74, + "learning_rate": 3.52906523434116e-05, + "loss": 1.0459, + "step": 170290 + }, + { + "epoch": 14.74, + "learning_rate": 3.5289786017499786e-05, + "loss": 1.1036, + "step": 170300 + }, + { + "epoch": 14.74, + "learning_rate": 3.528891969158798e-05, + "loss": 1.0861, + "step": 170310 + }, + { + "epoch": 14.74, + "learning_rate": 3.528805336567617e-05, + "loss": 1.0378, + "step": 170320 + }, + { + "epoch": 14.74, + "learning_rate": 3.528718703976436e-05, + "loss": 1.0396, + "step": 170330 + }, + { + "epoch": 14.74, + "learning_rate": 3.528632071385256e-05, + "loss": 1.0769, + "step": 170340 + }, + { + "epoch": 14.74, + "learning_rate": 3.5285454387940745e-05, + "loss": 1.0796, + "step": 170350 + }, + { + "epoch": 14.75, + "learning_rate": 3.528458806202894e-05, + "loss": 1.0494, + "step": 170360 + }, + { + "epoch": 14.75, + "learning_rate": 3.5283721736117134e-05, + "loss": 1.0661, + "step": 170370 + }, + { + "epoch": 14.75, + "learning_rate": 3.528285541020532e-05, + "loss": 1.0972, + "step": 170380 + }, + { + "epoch": 14.75, + "learning_rate": 3.5281989084293516e-05, + "loss": 1.0821, + "step": 170390 + }, + { + "epoch": 14.75, + "learning_rate": 3.528112275838171e-05, + "loss": 1.0134, + "step": 170400 + }, + { + "epoch": 14.75, + "learning_rate": 3.52802564324699e-05, + "loss": 1.1064, + "step": 170410 + }, + { + "epoch": 14.75, + "learning_rate": 3.527939010655809e-05, + "loss": 1.0948, + "step": 170420 + }, + { + "epoch": 14.75, + "learning_rate": 3.527852378064628e-05, + "loss": 1.1186, + "step": 170430 + }, + { + "epoch": 14.75, + "learning_rate": 3.5277657454734474e-05, + "loss": 1.1051, + "step": 170440 + }, + { + "epoch": 14.75, + "learning_rate": 3.527679112882267e-05, + "loss": 1.094, + "step": 170450 + }, + { + "epoch": 14.75, + "learning_rate": 3.5275924802910857e-05, + "loss": 1.0592, + "step": 170460 + }, + { + "epoch": 14.76, + "learning_rate": 3.527505847699905e-05, + "loss": 1.0513, + "step": 170470 + }, + { + "epoch": 14.76, + "learning_rate": 3.5274192151087245e-05, + "loss": 1.074, + "step": 170480 + }, + { + "epoch": 14.76, + "learning_rate": 3.527332582517543e-05, + "loss": 1.0158, + "step": 170490 + }, + { + "epoch": 14.76, + "learning_rate": 3.527245949926363e-05, + "loss": 1.0703, + "step": 170500 + }, + { + "epoch": 14.76, + "learning_rate": 3.527159317335182e-05, + "loss": 1.0505, + "step": 170510 + }, + { + "epoch": 14.76, + "learning_rate": 3.527072684744001e-05, + "loss": 1.0874, + "step": 170520 + }, + { + "epoch": 14.76, + "learning_rate": 3.5269860521528204e-05, + "loss": 1.1039, + "step": 170530 + }, + { + "epoch": 14.76, + "learning_rate": 3.526899419561639e-05, + "loss": 1.0244, + "step": 170540 + }, + { + "epoch": 14.76, + "learning_rate": 3.5268127869704586e-05, + "loss": 1.0423, + "step": 170550 + }, + { + "epoch": 14.76, + "learning_rate": 3.526726154379278e-05, + "loss": 1.0636, + "step": 170560 + }, + { + "epoch": 14.76, + "learning_rate": 3.526639521788097e-05, + "loss": 1.0849, + "step": 170570 + }, + { + "epoch": 14.76, + "learning_rate": 3.526552889196916e-05, + "loss": 1.0768, + "step": 170580 + }, + { + "epoch": 14.77, + "learning_rate": 3.526466256605736e-05, + "loss": 1.0787, + "step": 170590 + }, + { + "epoch": 14.77, + "learning_rate": 3.5263796240145545e-05, + "loss": 1.0805, + "step": 170600 + }, + { + "epoch": 14.77, + "learning_rate": 3.526292991423374e-05, + "loss": 1.0833, + "step": 170610 + }, + { + "epoch": 14.77, + "learning_rate": 3.5262063588321934e-05, + "loss": 1.0967, + "step": 170620 + }, + { + "epoch": 14.77, + "learning_rate": 3.526119726241012e-05, + "loss": 1.0668, + "step": 170630 + }, + { + "epoch": 14.77, + "learning_rate": 3.5260330936498316e-05, + "loss": 1.0682, + "step": 170640 + }, + { + "epoch": 14.77, + "learning_rate": 3.5259464610586504e-05, + "loss": 1.0857, + "step": 170650 + }, + { + "epoch": 14.77, + "learning_rate": 3.52585982846747e-05, + "loss": 1.0385, + "step": 170660 + }, + { + "epoch": 14.77, + "learning_rate": 3.525773195876289e-05, + "loss": 1.0952, + "step": 170670 + }, + { + "epoch": 14.77, + "learning_rate": 3.525686563285108e-05, + "loss": 1.0731, + "step": 170680 + }, + { + "epoch": 14.77, + "learning_rate": 3.5255999306939275e-05, + "loss": 1.0446, + "step": 170690 + }, + { + "epoch": 14.78, + "learning_rate": 3.525513298102747e-05, + "loss": 1.0491, + "step": 170700 + }, + { + "epoch": 14.78, + "learning_rate": 3.525426665511566e-05, + "loss": 1.062, + "step": 170710 + }, + { + "epoch": 14.78, + "learning_rate": 3.525340032920385e-05, + "loss": 1.071, + "step": 170720 + }, + { + "epoch": 14.78, + "learning_rate": 3.5252534003292046e-05, + "loss": 1.0731, + "step": 170730 + }, + { + "epoch": 14.78, + "learning_rate": 3.5251667677380233e-05, + "loss": 1.0477, + "step": 170740 + }, + { + "epoch": 14.78, + "learning_rate": 3.525080135146843e-05, + "loss": 1.0241, + "step": 170750 + }, + { + "epoch": 14.78, + "learning_rate": 3.5249935025556616e-05, + "loss": 1.0669, + "step": 170760 + }, + { + "epoch": 14.78, + "learning_rate": 3.524906869964481e-05, + "loss": 1.0606, + "step": 170770 + }, + { + "epoch": 14.78, + "learning_rate": 3.5248202373733004e-05, + "loss": 1.1362, + "step": 170780 + }, + { + "epoch": 14.78, + "learning_rate": 3.524733604782119e-05, + "loss": 1.0562, + "step": 170790 + }, + { + "epoch": 14.78, + "learning_rate": 3.5246469721909387e-05, + "loss": 1.087, + "step": 170800 + }, + { + "epoch": 14.78, + "learning_rate": 3.524560339599758e-05, + "loss": 1.0615, + "step": 170810 + }, + { + "epoch": 14.79, + "learning_rate": 3.524473707008577e-05, + "loss": 1.0955, + "step": 170820 + }, + { + "epoch": 14.79, + "learning_rate": 3.524387074417396e-05, + "loss": 1.0636, + "step": 170830 + }, + { + "epoch": 14.79, + "learning_rate": 3.524300441826216e-05, + "loss": 1.0146, + "step": 170840 + }, + { + "epoch": 14.79, + "learning_rate": 3.5242138092350345e-05, + "loss": 1.0488, + "step": 170850 + }, + { + "epoch": 14.79, + "learning_rate": 3.524127176643854e-05, + "loss": 1.0385, + "step": 170860 + }, + { + "epoch": 14.79, + "learning_rate": 3.524040544052673e-05, + "loss": 1.1119, + "step": 170870 + }, + { + "epoch": 14.79, + "learning_rate": 3.523953911461492e-05, + "loss": 1.0678, + "step": 170880 + }, + { + "epoch": 14.79, + "learning_rate": 3.5238672788703116e-05, + "loss": 1.0459, + "step": 170890 + }, + { + "epoch": 14.79, + "learning_rate": 3.5237806462791304e-05, + "loss": 0.9908, + "step": 170900 + }, + { + "epoch": 14.79, + "learning_rate": 3.52369401368795e-05, + "loss": 1.0334, + "step": 170910 + }, + { + "epoch": 14.79, + "learning_rate": 3.523607381096769e-05, + "loss": 1.0166, + "step": 170920 + }, + { + "epoch": 14.79, + "learning_rate": 3.523520748505588e-05, + "loss": 1.0272, + "step": 170930 + }, + { + "epoch": 14.8, + "learning_rate": 3.5234341159144075e-05, + "loss": 1.0825, + "step": 170940 + }, + { + "epoch": 14.8, + "learning_rate": 3.523347483323227e-05, + "loss": 1.1541, + "step": 170950 + }, + { + "epoch": 14.8, + "learning_rate": 3.523260850732046e-05, + "loss": 1.0662, + "step": 170960 + }, + { + "epoch": 14.8, + "learning_rate": 3.523174218140865e-05, + "loss": 1.073, + "step": 170970 + }, + { + "epoch": 14.8, + "learning_rate": 3.523087585549684e-05, + "loss": 1.0559, + "step": 170980 + }, + { + "epoch": 14.8, + "learning_rate": 3.5230009529585034e-05, + "loss": 1.0613, + "step": 170990 + }, + { + "epoch": 14.8, + "learning_rate": 3.522914320367323e-05, + "loss": 1.054, + "step": 171000 + }, + { + "epoch": 14.8, + "learning_rate": 3.5228276877761416e-05, + "loss": 1.0565, + "step": 171010 + }, + { + "epoch": 14.8, + "learning_rate": 3.522741055184961e-05, + "loss": 1.0442, + "step": 171020 + }, + { + "epoch": 14.8, + "learning_rate": 3.5226544225937805e-05, + "loss": 1.0029, + "step": 171030 + }, + { + "epoch": 14.8, + "learning_rate": 3.522567790002599e-05, + "loss": 1.0564, + "step": 171040 + }, + { + "epoch": 14.81, + "learning_rate": 3.522481157411419e-05, + "loss": 1.0395, + "step": 171050 + }, + { + "epoch": 14.81, + "learning_rate": 3.5223945248202375e-05, + "loss": 1.059, + "step": 171060 + }, + { + "epoch": 14.81, + "learning_rate": 3.522307892229057e-05, + "loss": 1.0758, + "step": 171070 + }, + { + "epoch": 14.81, + "learning_rate": 3.5222212596378764e-05, + "loss": 1.048, + "step": 171080 + }, + { + "epoch": 14.81, + "learning_rate": 3.522134627046695e-05, + "loss": 1.0825, + "step": 171090 + }, + { + "epoch": 14.81, + "learning_rate": 3.5220479944555146e-05, + "loss": 1.1218, + "step": 171100 + }, + { + "epoch": 14.81, + "learning_rate": 3.521961361864334e-05, + "loss": 1.0856, + "step": 171110 + }, + { + "epoch": 14.81, + "learning_rate": 3.521874729273153e-05, + "loss": 0.9965, + "step": 171120 + }, + { + "epoch": 14.81, + "learning_rate": 3.521788096681972e-05, + "loss": 1.0531, + "step": 171130 + }, + { + "epoch": 14.81, + "learning_rate": 3.521701464090792e-05, + "loss": 1.064, + "step": 171140 + }, + { + "epoch": 14.81, + "learning_rate": 3.5216148314996104e-05, + "loss": 1.0941, + "step": 171150 + }, + { + "epoch": 14.81, + "learning_rate": 3.52152819890843e-05, + "loss": 1.0106, + "step": 171160 + }, + { + "epoch": 14.82, + "learning_rate": 3.5214415663172486e-05, + "loss": 1.0914, + "step": 171170 + }, + { + "epoch": 14.82, + "learning_rate": 3.521354933726068e-05, + "loss": 1.0672, + "step": 171180 + }, + { + "epoch": 14.82, + "learning_rate": 3.5212683011348875e-05, + "loss": 1.0541, + "step": 171190 + }, + { + "epoch": 14.82, + "learning_rate": 3.521181668543706e-05, + "loss": 1.028, + "step": 171200 + }, + { + "epoch": 14.82, + "learning_rate": 3.521095035952526e-05, + "loss": 1.0661, + "step": 171210 + }, + { + "epoch": 14.82, + "learning_rate": 3.521008403361345e-05, + "loss": 1.0125, + "step": 171220 + }, + { + "epoch": 14.82, + "learning_rate": 3.520921770770164e-05, + "loss": 1.0496, + "step": 171230 + }, + { + "epoch": 14.82, + "learning_rate": 3.5208351381789834e-05, + "loss": 1.0849, + "step": 171240 + }, + { + "epoch": 14.82, + "learning_rate": 3.520748505587803e-05, + "loss": 1.0698, + "step": 171250 + }, + { + "epoch": 14.82, + "learning_rate": 3.5206618729966216e-05, + "loss": 1.0887, + "step": 171260 + }, + { + "epoch": 14.82, + "learning_rate": 3.520575240405441e-05, + "loss": 1.1235, + "step": 171270 + }, + { + "epoch": 14.83, + "learning_rate": 3.52048860781426e-05, + "loss": 1.0505, + "step": 171280 + }, + { + "epoch": 14.83, + "learning_rate": 3.520401975223079e-05, + "loss": 1.0331, + "step": 171290 + }, + { + "epoch": 14.83, + "learning_rate": 3.520315342631899e-05, + "loss": 1.0837, + "step": 171300 + }, + { + "epoch": 14.83, + "learning_rate": 3.5202287100407175e-05, + "loss": 1.0375, + "step": 171310 + }, + { + "epoch": 14.83, + "learning_rate": 3.520142077449537e-05, + "loss": 1.0149, + "step": 171320 + }, + { + "epoch": 14.83, + "learning_rate": 3.5200554448583564e-05, + "loss": 1.0301, + "step": 171330 + }, + { + "epoch": 14.83, + "learning_rate": 3.519968812267175e-05, + "loss": 1.0809, + "step": 171340 + }, + { + "epoch": 14.83, + "learning_rate": 3.5198821796759946e-05, + "loss": 0.9922, + "step": 171350 + }, + { + "epoch": 14.83, + "learning_rate": 3.519795547084814e-05, + "loss": 1.0059, + "step": 171360 + }, + { + "epoch": 14.83, + "learning_rate": 3.519708914493633e-05, + "loss": 1.0843, + "step": 171370 + }, + { + "epoch": 14.83, + "learning_rate": 3.519622281902452e-05, + "loss": 1.0687, + "step": 171380 + }, + { + "epoch": 14.83, + "learning_rate": 3.519535649311271e-05, + "loss": 1.0412, + "step": 171390 + }, + { + "epoch": 14.84, + "learning_rate": 3.5194490167200905e-05, + "loss": 1.044, + "step": 171400 + }, + { + "epoch": 14.84, + "learning_rate": 3.51936238412891e-05, + "loss": 1.0759, + "step": 171410 + }, + { + "epoch": 14.84, + "learning_rate": 3.519275751537729e-05, + "loss": 1.0485, + "step": 171420 + }, + { + "epoch": 14.84, + "learning_rate": 3.519189118946548e-05, + "loss": 1.0651, + "step": 171430 + }, + { + "epoch": 14.84, + "learning_rate": 3.5191024863553676e-05, + "loss": 1.0069, + "step": 171440 + }, + { + "epoch": 14.84, + "learning_rate": 3.5190158537641863e-05, + "loss": 1.1052, + "step": 171450 + }, + { + "epoch": 14.84, + "learning_rate": 3.518929221173006e-05, + "loss": 0.995, + "step": 171460 + }, + { + "epoch": 14.84, + "learning_rate": 3.518842588581825e-05, + "loss": 1.078, + "step": 171470 + }, + { + "epoch": 14.84, + "learning_rate": 3.518755955990644e-05, + "loss": 1.1196, + "step": 171480 + }, + { + "epoch": 14.84, + "learning_rate": 3.5186693233994634e-05, + "loss": 1.136, + "step": 171490 + }, + { + "epoch": 14.84, + "learning_rate": 3.518582690808282e-05, + "loss": 1.0354, + "step": 171500 + }, + { + "epoch": 14.85, + "learning_rate": 3.5184960582171017e-05, + "loss": 1.0472, + "step": 171510 + }, + { + "epoch": 14.85, + "learning_rate": 3.518409425625921e-05, + "loss": 1.0283, + "step": 171520 + }, + { + "epoch": 14.85, + "learning_rate": 3.51832279303474e-05, + "loss": 1.0284, + "step": 171530 + }, + { + "epoch": 14.85, + "learning_rate": 3.518236160443559e-05, + "loss": 1.0419, + "step": 171540 + }, + { + "epoch": 14.85, + "learning_rate": 3.518149527852379e-05, + "loss": 1.0432, + "step": 171550 + }, + { + "epoch": 14.85, + "learning_rate": 3.5180628952611975e-05, + "loss": 1.0365, + "step": 171560 + }, + { + "epoch": 14.85, + "learning_rate": 3.517976262670017e-05, + "loss": 1.092, + "step": 171570 + }, + { + "epoch": 14.85, + "learning_rate": 3.5178896300788364e-05, + "loss": 1.0582, + "step": 171580 + }, + { + "epoch": 14.85, + "learning_rate": 3.517802997487655e-05, + "loss": 1.0152, + "step": 171590 + }, + { + "epoch": 14.85, + "learning_rate": 3.5177163648964746e-05, + "loss": 1.0711, + "step": 171600 + }, + { + "epoch": 14.85, + "learning_rate": 3.5176297323052934e-05, + "loss": 1.0059, + "step": 171610 + }, + { + "epoch": 14.85, + "learning_rate": 3.517543099714113e-05, + "loss": 1.0589, + "step": 171620 + }, + { + "epoch": 14.86, + "learning_rate": 3.517456467122932e-05, + "loss": 1.0575, + "step": 171630 + }, + { + "epoch": 14.86, + "learning_rate": 3.517369834531751e-05, + "loss": 1.1265, + "step": 171640 + }, + { + "epoch": 14.86, + "learning_rate": 3.5172832019405705e-05, + "loss": 1.1036, + "step": 171650 + }, + { + "epoch": 14.86, + "learning_rate": 3.51719656934939e-05, + "loss": 1.1245, + "step": 171660 + }, + { + "epoch": 14.86, + "learning_rate": 3.517109936758209e-05, + "loss": 1.0584, + "step": 171670 + }, + { + "epoch": 14.86, + "learning_rate": 3.517023304167028e-05, + "loss": 1.0258, + "step": 171680 + }, + { + "epoch": 14.86, + "learning_rate": 3.5169366715758476e-05, + "loss": 1.0968, + "step": 171690 + }, + { + "epoch": 14.86, + "learning_rate": 3.5168500389846664e-05, + "loss": 1.0494, + "step": 171700 + }, + { + "epoch": 14.86, + "learning_rate": 3.516763406393486e-05, + "loss": 1.072, + "step": 171710 + }, + { + "epoch": 14.86, + "learning_rate": 3.5166767738023046e-05, + "loss": 1.0604, + "step": 171720 + }, + { + "epoch": 14.86, + "learning_rate": 3.516590141211124e-05, + "loss": 1.0862, + "step": 171730 + }, + { + "epoch": 14.87, + "learning_rate": 3.5165035086199435e-05, + "loss": 1.0471, + "step": 171740 + }, + { + "epoch": 14.87, + "learning_rate": 3.516416876028762e-05, + "loss": 1.1556, + "step": 171750 + }, + { + "epoch": 14.87, + "learning_rate": 3.516330243437582e-05, + "loss": 1.0257, + "step": 171760 + }, + { + "epoch": 14.87, + "learning_rate": 3.516243610846401e-05, + "loss": 1.0059, + "step": 171770 + }, + { + "epoch": 14.87, + "learning_rate": 3.51615697825522e-05, + "loss": 1.0709, + "step": 171780 + }, + { + "epoch": 14.87, + "learning_rate": 3.5160703456640393e-05, + "loss": 1.104, + "step": 171790 + }, + { + "epoch": 14.87, + "learning_rate": 3.515983713072858e-05, + "loss": 1.0201, + "step": 171800 + }, + { + "epoch": 14.87, + "learning_rate": 3.5158970804816776e-05, + "loss": 1.0824, + "step": 171810 + }, + { + "epoch": 14.87, + "learning_rate": 3.515810447890497e-05, + "loss": 1.0163, + "step": 171820 + }, + { + "epoch": 14.87, + "learning_rate": 3.515723815299316e-05, + "loss": 1.1093, + "step": 171830 + }, + { + "epoch": 14.87, + "learning_rate": 3.515637182708135e-05, + "loss": 1.0927, + "step": 171840 + }, + { + "epoch": 14.87, + "learning_rate": 3.515550550116954e-05, + "loss": 1.0716, + "step": 171850 + }, + { + "epoch": 14.88, + "learning_rate": 3.515463917525773e-05, + "loss": 1.0521, + "step": 171860 + }, + { + "epoch": 14.88, + "learning_rate": 3.515377284934592e-05, + "loss": 0.9918, + "step": 171870 + }, + { + "epoch": 14.88, + "learning_rate": 3.5152906523434116e-05, + "loss": 1.0845, + "step": 171880 + }, + { + "epoch": 14.88, + "learning_rate": 3.5152040197522304e-05, + "loss": 1.0911, + "step": 171890 + }, + { + "epoch": 14.88, + "learning_rate": 3.51511738716105e-05, + "loss": 1.0504, + "step": 171900 + }, + { + "epoch": 14.88, + "learning_rate": 3.515030754569869e-05, + "loss": 1.0463, + "step": 171910 + }, + { + "epoch": 14.88, + "learning_rate": 3.514944121978688e-05, + "loss": 1.157, + "step": 171920 + }, + { + "epoch": 14.88, + "learning_rate": 3.5148574893875075e-05, + "loss": 1.0869, + "step": 171930 + }, + { + "epoch": 14.88, + "learning_rate": 3.514770856796327e-05, + "loss": 1.013, + "step": 171940 + }, + { + "epoch": 14.88, + "learning_rate": 3.514684224205146e-05, + "loss": 1.0354, + "step": 171950 + }, + { + "epoch": 14.88, + "learning_rate": 3.514597591613965e-05, + "loss": 1.061, + "step": 171960 + }, + { + "epoch": 14.88, + "learning_rate": 3.514510959022784e-05, + "loss": 1.1253, + "step": 171970 + }, + { + "epoch": 14.89, + "learning_rate": 3.5144243264316034e-05, + "loss": 1.0514, + "step": 171980 + }, + { + "epoch": 14.89, + "learning_rate": 3.514337693840423e-05, + "loss": 1.0348, + "step": 171990 + }, + { + "epoch": 14.89, + "learning_rate": 3.5142510612492416e-05, + "loss": 1.061, + "step": 172000 + }, + { + "epoch": 14.89, + "learning_rate": 3.514164428658061e-05, + "loss": 1.0575, + "step": 172010 + }, + { + "epoch": 14.89, + "learning_rate": 3.5140777960668805e-05, + "loss": 1.0235, + "step": 172020 + }, + { + "epoch": 14.89, + "learning_rate": 3.513991163475699e-05, + "loss": 1.0731, + "step": 172030 + }, + { + "epoch": 14.89, + "learning_rate": 3.513904530884519e-05, + "loss": 1.0556, + "step": 172040 + }, + { + "epoch": 14.89, + "learning_rate": 3.513817898293338e-05, + "loss": 1.0573, + "step": 172050 + }, + { + "epoch": 14.89, + "learning_rate": 3.513731265702157e-05, + "loss": 1.0597, + "step": 172060 + }, + { + "epoch": 14.89, + "learning_rate": 3.5136446331109764e-05, + "loss": 1.1438, + "step": 172070 + }, + { + "epoch": 14.89, + "learning_rate": 3.513558000519795e-05, + "loss": 1.0998, + "step": 172080 + }, + { + "epoch": 14.9, + "learning_rate": 3.5134713679286146e-05, + "loss": 1.0988, + "step": 172090 + }, + { + "epoch": 14.9, + "learning_rate": 3.513384735337434e-05, + "loss": 1.1087, + "step": 172100 + }, + { + "epoch": 14.9, + "learning_rate": 3.513298102746253e-05, + "loss": 1.1201, + "step": 172110 + }, + { + "epoch": 14.9, + "learning_rate": 3.513211470155072e-05, + "loss": 1.0252, + "step": 172120 + }, + { + "epoch": 14.9, + "learning_rate": 3.513124837563892e-05, + "loss": 1.0266, + "step": 172130 + }, + { + "epoch": 14.9, + "learning_rate": 3.5130382049727104e-05, + "loss": 1.0606, + "step": 172140 + }, + { + "epoch": 14.9, + "learning_rate": 3.51295157238153e-05, + "loss": 1.0485, + "step": 172150 + }, + { + "epoch": 14.9, + "learning_rate": 3.512864939790349e-05, + "loss": 1.0667, + "step": 172160 + }, + { + "epoch": 14.9, + "learning_rate": 3.512778307199168e-05, + "loss": 1.024, + "step": 172170 + }, + { + "epoch": 14.9, + "learning_rate": 3.5126916746079876e-05, + "loss": 1.0465, + "step": 172180 + }, + { + "epoch": 14.9, + "learning_rate": 3.512605042016806e-05, + "loss": 1.0437, + "step": 172190 + }, + { + "epoch": 14.9, + "learning_rate": 3.512518409425626e-05, + "loss": 1.0932, + "step": 172200 + }, + { + "epoch": 14.91, + "learning_rate": 3.512431776834445e-05, + "loss": 1.0555, + "step": 172210 + }, + { + "epoch": 14.91, + "learning_rate": 3.512345144243264e-05, + "loss": 1.0455, + "step": 172220 + }, + { + "epoch": 14.91, + "learning_rate": 3.5122585116520834e-05, + "loss": 1.0559, + "step": 172230 + }, + { + "epoch": 14.91, + "learning_rate": 3.512171879060903e-05, + "loss": 1.058, + "step": 172240 + }, + { + "epoch": 14.91, + "learning_rate": 3.5120852464697216e-05, + "loss": 1.0794, + "step": 172250 + }, + { + "epoch": 14.91, + "learning_rate": 3.511998613878541e-05, + "loss": 1.04, + "step": 172260 + }, + { + "epoch": 14.91, + "learning_rate": 3.5119119812873605e-05, + "loss": 1.1111, + "step": 172270 + }, + { + "epoch": 14.91, + "learning_rate": 3.511825348696179e-05, + "loss": 1.1205, + "step": 172280 + }, + { + "epoch": 14.91, + "learning_rate": 3.511738716104999e-05, + "loss": 1.0411, + "step": 172290 + }, + { + "epoch": 14.91, + "learning_rate": 3.5116520835138175e-05, + "loss": 1.0742, + "step": 172300 + }, + { + "epoch": 14.91, + "learning_rate": 3.511565450922637e-05, + "loss": 1.0564, + "step": 172310 + }, + { + "epoch": 14.92, + "learning_rate": 3.5114788183314564e-05, + "loss": 1.0871, + "step": 172320 + }, + { + "epoch": 14.92, + "learning_rate": 3.511392185740275e-05, + "loss": 1.0664, + "step": 172330 + }, + { + "epoch": 14.92, + "learning_rate": 3.5113055531490946e-05, + "loss": 1.0837, + "step": 172340 + }, + { + "epoch": 14.92, + "learning_rate": 3.511218920557914e-05, + "loss": 1.0514, + "step": 172350 + }, + { + "epoch": 14.92, + "learning_rate": 3.511132287966733e-05, + "loss": 1.0747, + "step": 172360 + }, + { + "epoch": 14.92, + "learning_rate": 3.511045655375552e-05, + "loss": 1.0564, + "step": 172370 + }, + { + "epoch": 14.92, + "learning_rate": 3.510959022784372e-05, + "loss": 1.1165, + "step": 172380 + }, + { + "epoch": 14.92, + "learning_rate": 3.5108723901931905e-05, + "loss": 1.0059, + "step": 172390 + }, + { + "epoch": 14.92, + "learning_rate": 3.51078575760201e-05, + "loss": 1.0754, + "step": 172400 + }, + { + "epoch": 14.92, + "learning_rate": 3.510699125010829e-05, + "loss": 1.0772, + "step": 172410 + }, + { + "epoch": 14.92, + "learning_rate": 3.510612492419648e-05, + "loss": 1.0877, + "step": 172420 + }, + { + "epoch": 14.92, + "learning_rate": 3.5105258598284676e-05, + "loss": 1.0837, + "step": 172430 + }, + { + "epoch": 14.93, + "learning_rate": 3.5104392272372864e-05, + "loss": 1.0777, + "step": 172440 + }, + { + "epoch": 14.93, + "learning_rate": 3.510352594646106e-05, + "loss": 1.0984, + "step": 172450 + }, + { + "epoch": 14.93, + "learning_rate": 3.510265962054925e-05, + "loss": 1.0137, + "step": 172460 + }, + { + "epoch": 14.93, + "learning_rate": 3.510179329463744e-05, + "loss": 0.9775, + "step": 172470 + }, + { + "epoch": 14.93, + "learning_rate": 3.5100926968725635e-05, + "loss": 1.0781, + "step": 172480 + }, + { + "epoch": 14.93, + "learning_rate": 3.510006064281383e-05, + "loss": 1.0587, + "step": 172490 + }, + { + "epoch": 14.93, + "learning_rate": 3.509919431690202e-05, + "loss": 1.0682, + "step": 172500 + }, + { + "epoch": 14.93, + "learning_rate": 3.509832799099021e-05, + "loss": 1.0853, + "step": 172510 + }, + { + "epoch": 14.93, + "learning_rate": 3.50974616650784e-05, + "loss": 1.066, + "step": 172520 + }, + { + "epoch": 14.93, + "learning_rate": 3.509659533916659e-05, + "loss": 1.0942, + "step": 172530 + }, + { + "epoch": 14.93, + "learning_rate": 3.509572901325479e-05, + "loss": 1.0099, + "step": 172540 + }, + { + "epoch": 14.94, + "learning_rate": 3.5094862687342975e-05, + "loss": 1.0251, + "step": 172550 + }, + { + "epoch": 14.94, + "learning_rate": 3.509399636143117e-05, + "loss": 1.0845, + "step": 172560 + }, + { + "epoch": 14.94, + "learning_rate": 3.5093130035519364e-05, + "loss": 1.0578, + "step": 172570 + }, + { + "epoch": 14.94, + "learning_rate": 3.509226370960755e-05, + "loss": 1.0714, + "step": 172580 + }, + { + "epoch": 14.94, + "learning_rate": 3.5091397383695746e-05, + "loss": 1.1319, + "step": 172590 + }, + { + "epoch": 14.94, + "learning_rate": 3.5090531057783934e-05, + "loss": 1.0544, + "step": 172600 + }, + { + "epoch": 14.94, + "learning_rate": 3.508966473187213e-05, + "loss": 1.0665, + "step": 172610 + }, + { + "epoch": 14.94, + "learning_rate": 3.508879840596032e-05, + "loss": 1.0831, + "step": 172620 + }, + { + "epoch": 14.94, + "learning_rate": 3.508793208004851e-05, + "loss": 1.0978, + "step": 172630 + }, + { + "epoch": 14.94, + "learning_rate": 3.5087065754136705e-05, + "loss": 1.0941, + "step": 172640 + }, + { + "epoch": 14.94, + "learning_rate": 3.50861994282249e-05, + "loss": 0.9979, + "step": 172650 + }, + { + "epoch": 14.94, + "learning_rate": 3.508533310231309e-05, + "loss": 1.0537, + "step": 172660 + }, + { + "epoch": 14.95, + "learning_rate": 3.508446677640128e-05, + "loss": 1.0304, + "step": 172670 + }, + { + "epoch": 14.95, + "learning_rate": 3.5083600450489476e-05, + "loss": 1.0227, + "step": 172680 + }, + { + "epoch": 14.95, + "learning_rate": 3.5082734124577664e-05, + "loss": 1.0838, + "step": 172690 + }, + { + "epoch": 14.95, + "learning_rate": 3.508186779866586e-05, + "loss": 1.0619, + "step": 172700 + }, + { + "epoch": 14.95, + "learning_rate": 3.5081001472754046e-05, + "loss": 1.1156, + "step": 172710 + }, + { + "epoch": 14.95, + "learning_rate": 3.508013514684224e-05, + "loss": 1.0929, + "step": 172720 + }, + { + "epoch": 14.95, + "learning_rate": 3.5079268820930435e-05, + "loss": 1.0593, + "step": 172730 + }, + { + "epoch": 14.95, + "learning_rate": 3.507840249501862e-05, + "loss": 1.0687, + "step": 172740 + }, + { + "epoch": 14.95, + "learning_rate": 3.507753616910682e-05, + "loss": 1.0904, + "step": 172750 + }, + { + "epoch": 14.95, + "learning_rate": 3.507666984319501e-05, + "loss": 1.0991, + "step": 172760 + }, + { + "epoch": 14.95, + "learning_rate": 3.50758035172832e-05, + "loss": 1.0624, + "step": 172770 + }, + { + "epoch": 14.96, + "learning_rate": 3.5074937191371394e-05, + "loss": 1.0776, + "step": 172780 + }, + { + "epoch": 14.96, + "learning_rate": 3.507407086545959e-05, + "loss": 1.0583, + "step": 172790 + }, + { + "epoch": 14.96, + "learning_rate": 3.5073204539547776e-05, + "loss": 1.0498, + "step": 172800 + }, + { + "epoch": 14.96, + "learning_rate": 3.507233821363597e-05, + "loss": 1.0739, + "step": 172810 + }, + { + "epoch": 14.96, + "learning_rate": 3.507147188772416e-05, + "loss": 1.0215, + "step": 172820 + }, + { + "epoch": 14.96, + "learning_rate": 3.507060556181235e-05, + "loss": 1.1049, + "step": 172830 + }, + { + "epoch": 14.96, + "learning_rate": 3.506973923590055e-05, + "loss": 1.0521, + "step": 172840 + }, + { + "epoch": 14.96, + "learning_rate": 3.5068872909988734e-05, + "loss": 0.9828, + "step": 172850 + }, + { + "epoch": 14.96, + "learning_rate": 3.506800658407693e-05, + "loss": 1.0495, + "step": 172860 + }, + { + "epoch": 14.96, + "learning_rate": 3.506714025816512e-05, + "loss": 1.0347, + "step": 172870 + }, + { + "epoch": 14.96, + "learning_rate": 3.506627393225331e-05, + "loss": 1.0608, + "step": 172880 + }, + { + "epoch": 14.96, + "learning_rate": 3.5065407606341505e-05, + "loss": 1.0508, + "step": 172890 + }, + { + "epoch": 14.97, + "learning_rate": 3.50645412804297e-05, + "loss": 1.033, + "step": 172900 + }, + { + "epoch": 14.97, + "learning_rate": 3.506367495451789e-05, + "loss": 1.0759, + "step": 172910 + }, + { + "epoch": 14.97, + "learning_rate": 3.506280862860608e-05, + "loss": 1.0231, + "step": 172920 + }, + { + "epoch": 14.97, + "learning_rate": 3.506194230269427e-05, + "loss": 1.0925, + "step": 172930 + }, + { + "epoch": 14.97, + "learning_rate": 3.5061075976782464e-05, + "loss": 1.0214, + "step": 172940 + }, + { + "epoch": 14.97, + "learning_rate": 3.506020965087066e-05, + "loss": 1.0697, + "step": 172950 + }, + { + "epoch": 14.97, + "learning_rate": 3.5059343324958846e-05, + "loss": 1.0492, + "step": 172960 + }, + { + "epoch": 14.97, + "learning_rate": 3.505847699904704e-05, + "loss": 1.0609, + "step": 172970 + }, + { + "epoch": 14.97, + "learning_rate": 3.5057610673135235e-05, + "loss": 1.0012, + "step": 172980 + }, + { + "epoch": 14.97, + "learning_rate": 3.505674434722342e-05, + "loss": 1.0632, + "step": 172990 + }, + { + "epoch": 14.97, + "learning_rate": 3.505587802131162e-05, + "loss": 1.0834, + "step": 173000 + }, + { + "epoch": 14.97, + "learning_rate": 3.505501169539981e-05, + "loss": 1.1249, + "step": 173010 + }, + { + "epoch": 14.98, + "learning_rate": 3.5054145369488e-05, + "loss": 1.0691, + "step": 173020 + }, + { + "epoch": 14.98, + "learning_rate": 3.5053279043576194e-05, + "loss": 1.0447, + "step": 173030 + }, + { + "epoch": 14.98, + "learning_rate": 3.505241271766438e-05, + "loss": 0.9982, + "step": 173040 + }, + { + "epoch": 14.98, + "learning_rate": 3.5051546391752576e-05, + "loss": 1.1098, + "step": 173050 + }, + { + "epoch": 14.98, + "learning_rate": 3.505068006584077e-05, + "loss": 1.0754, + "step": 173060 + }, + { + "epoch": 14.98, + "learning_rate": 3.504981373992896e-05, + "loss": 1.0609, + "step": 173070 + }, + { + "epoch": 14.98, + "learning_rate": 3.504894741401715e-05, + "loss": 0.9893, + "step": 173080 + }, + { + "epoch": 14.98, + "learning_rate": 3.504808108810535e-05, + "loss": 1.0227, + "step": 173090 + }, + { + "epoch": 14.98, + "learning_rate": 3.5047214762193535e-05, + "loss": 1.0777, + "step": 173100 + }, + { + "epoch": 14.98, + "learning_rate": 3.504634843628173e-05, + "loss": 1.0284, + "step": 173110 + }, + { + "epoch": 14.98, + "learning_rate": 3.5045482110369924e-05, + "loss": 1.0963, + "step": 173120 + }, + { + "epoch": 14.99, + "learning_rate": 3.504461578445811e-05, + "loss": 1.0469, + "step": 173130 + }, + { + "epoch": 14.99, + "learning_rate": 3.5043749458546306e-05, + "loss": 1.0578, + "step": 173140 + }, + { + "epoch": 14.99, + "learning_rate": 3.5042883132634493e-05, + "loss": 1.1023, + "step": 173150 + }, + { + "epoch": 14.99, + "learning_rate": 3.504201680672269e-05, + "loss": 1.0097, + "step": 173160 + }, + { + "epoch": 14.99, + "learning_rate": 3.504115048081088e-05, + "loss": 1.0836, + "step": 173170 + }, + { + "epoch": 14.99, + "learning_rate": 3.504028415489907e-05, + "loss": 1.0187, + "step": 173180 + }, + { + "epoch": 14.99, + "learning_rate": 3.5039417828987265e-05, + "loss": 1.0538, + "step": 173190 + }, + { + "epoch": 14.99, + "learning_rate": 3.503855150307546e-05, + "loss": 1.0663, + "step": 173200 + }, + { + "epoch": 14.99, + "learning_rate": 3.503768517716365e-05, + "loss": 1.0991, + "step": 173210 + }, + { + "epoch": 14.99, + "learning_rate": 3.503681885125184e-05, + "loss": 1.0775, + "step": 173220 + }, + { + "epoch": 14.99, + "learning_rate": 3.5035952525340036e-05, + "loss": 1.0449, + "step": 173230 + }, + { + "epoch": 14.99, + "learning_rate": 3.503508619942822e-05, + "loss": 1.0698, + "step": 173240 + }, + { + "epoch": 15.0, + "learning_rate": 3.503421987351642e-05, + "loss": 1.0721, + "step": 173250 + }, + { + "epoch": 15.0, + "learning_rate": 3.5033353547604605e-05, + "loss": 1.0069, + "step": 173260 + }, + { + "epoch": 15.0, + "learning_rate": 3.50324872216928e-05, + "loss": 1.078, + "step": 173270 + }, + { + "epoch": 15.0, + "learning_rate": 3.5031620895780994e-05, + "loss": 1.0726, + "step": 173280 + }, + { + "epoch": 15.0, + "learning_rate": 3.503075456986918e-05, + "loss": 1.0777, + "step": 173290 + }, + { + "epoch": 15.0, + "eval_Bleu_1": 0.04296015956630623, + "eval_Bleu_2": 2.833025920129063e-11, + "eval_Bleu_3": 2.5486829544077285e-14, + "eval_Bleu_4": 7.856957601019205e-16, + "eval_ROUGE_L": 0.08922511541374702, + "eval_cer": 0.9935054646698495, + "eval_em": 0, + "eval_f1": 0.10113533885595156, + "eval_loss": 1.0194092988967896, + "eval_runtime": 1727.8757, + "eval_samples_per_second": 2.972, + "eval_steps_per_second": 2.972, + "eval_wer": 0.9720343202365894, + "step": 173299 + }, + { + "epoch": 15.0, + "learning_rate": 3.5029888243957376e-05, + "loss": 1.0544, + "step": 173300 + }, + { + "epoch": 15.0, + "learning_rate": 3.502902191804557e-05, + "loss": 1.0471, + "step": 173310 + }, + { + "epoch": 15.0, + "learning_rate": 3.502815559213376e-05, + "loss": 1.0518, + "step": 173320 + }, + { + "epoch": 15.0, + "learning_rate": 3.502728926622195e-05, + "loss": 1.0268, + "step": 173330 + }, + { + "epoch": 15.0, + "learning_rate": 3.502642294031014e-05, + "loss": 1.0221, + "step": 173340 + }, + { + "epoch": 15.0, + "learning_rate": 3.5025556614398335e-05, + "loss": 1.0463, + "step": 173350 + }, + { + "epoch": 15.01, + "learning_rate": 3.502469028848653e-05, + "loss": 1.0222, + "step": 173360 + }, + { + "epoch": 15.01, + "learning_rate": 3.502382396257472e-05, + "loss": 1.0111, + "step": 173370 + }, + { + "epoch": 15.01, + "learning_rate": 3.502295763666291e-05, + "loss": 1.0207, + "step": 173380 + }, + { + "epoch": 15.01, + "learning_rate": 3.5022091310751106e-05, + "loss": 0.999, + "step": 173390 + }, + { + "epoch": 15.01, + "learning_rate": 3.5021224984839294e-05, + "loss": 1.0096, + "step": 173400 + }, + { + "epoch": 15.01, + "learning_rate": 3.502035865892749e-05, + "loss": 1.0775, + "step": 173410 + }, + { + "epoch": 15.01, + "learning_rate": 3.501949233301568e-05, + "loss": 1.0614, + "step": 173420 + }, + { + "epoch": 15.01, + "learning_rate": 3.501862600710387e-05, + "loss": 1.1037, + "step": 173430 + }, + { + "epoch": 15.01, + "learning_rate": 3.5017759681192065e-05, + "loss": 1.0646, + "step": 173440 + }, + { + "epoch": 15.01, + "learning_rate": 3.501689335528025e-05, + "loss": 0.9855, + "step": 173450 + }, + { + "epoch": 15.01, + "learning_rate": 3.501602702936845e-05, + "loss": 1.0217, + "step": 173460 + }, + { + "epoch": 15.01, + "learning_rate": 3.501516070345664e-05, + "loss": 1.0097, + "step": 173470 + }, + { + "epoch": 15.02, + "learning_rate": 3.501429437754483e-05, + "loss": 1.0117, + "step": 173480 + }, + { + "epoch": 15.02, + "learning_rate": 3.5013428051633024e-05, + "loss": 1.0516, + "step": 173490 + }, + { + "epoch": 15.02, + "learning_rate": 3.501256172572122e-05, + "loss": 1.0483, + "step": 173500 + }, + { + "epoch": 15.02, + "learning_rate": 3.5011695399809406e-05, + "loss": 0.998, + "step": 173510 + }, + { + "epoch": 15.02, + "learning_rate": 3.50108290738976e-05, + "loss": 1.0521, + "step": 173520 + }, + { + "epoch": 15.02, + "learning_rate": 3.5009962747985795e-05, + "loss": 1.0562, + "step": 173530 + }, + { + "epoch": 15.02, + "learning_rate": 3.500909642207398e-05, + "loss": 1.0286, + "step": 173540 + }, + { + "epoch": 15.02, + "learning_rate": 3.500823009616218e-05, + "loss": 1.073, + "step": 173550 + }, + { + "epoch": 15.02, + "learning_rate": 3.5007363770250364e-05, + "loss": 1.0291, + "step": 173560 + }, + { + "epoch": 15.02, + "learning_rate": 3.500649744433856e-05, + "loss": 1.0373, + "step": 173570 + }, + { + "epoch": 15.02, + "learning_rate": 3.500563111842675e-05, + "loss": 1.0198, + "step": 173580 + }, + { + "epoch": 15.03, + "learning_rate": 3.500476479251494e-05, + "loss": 0.9428, + "step": 173590 + }, + { + "epoch": 15.03, + "learning_rate": 3.5003898466603135e-05, + "loss": 1.043, + "step": 173600 + }, + { + "epoch": 15.03, + "learning_rate": 3.500303214069133e-05, + "loss": 1.0068, + "step": 173610 + }, + { + "epoch": 15.03, + "learning_rate": 3.500216581477952e-05, + "loss": 1.0007, + "step": 173620 + }, + { + "epoch": 15.03, + "learning_rate": 3.500129948886771e-05, + "loss": 1.0274, + "step": 173630 + }, + { + "epoch": 15.03, + "learning_rate": 3.5000433162955906e-05, + "loss": 1.0552, + "step": 173640 + }, + { + "epoch": 15.03, + "learning_rate": 3.4999566837044094e-05, + "loss": 1.0922, + "step": 173650 + }, + { + "epoch": 15.03, + "learning_rate": 3.499870051113229e-05, + "loss": 1.0487, + "step": 173660 + }, + { + "epoch": 15.03, + "learning_rate": 3.4997834185220476e-05, + "loss": 1.0664, + "step": 173670 + }, + { + "epoch": 15.03, + "learning_rate": 3.499696785930867e-05, + "loss": 1.0988, + "step": 173680 + }, + { + "epoch": 15.03, + "learning_rate": 3.4996101533396865e-05, + "loss": 1.0149, + "step": 173690 + }, + { + "epoch": 15.03, + "learning_rate": 3.499523520748505e-05, + "loss": 1.0464, + "step": 173700 + }, + { + "epoch": 15.04, + "learning_rate": 3.499436888157325e-05, + "loss": 1.0234, + "step": 173710 + }, + { + "epoch": 15.04, + "learning_rate": 3.499350255566144e-05, + "loss": 1.0617, + "step": 173720 + }, + { + "epoch": 15.04, + "learning_rate": 3.499263622974963e-05, + "loss": 1.0493, + "step": 173730 + }, + { + "epoch": 15.04, + "learning_rate": 3.4991769903837824e-05, + "loss": 1.0724, + "step": 173740 + }, + { + "epoch": 15.04, + "learning_rate": 3.499090357792602e-05, + "loss": 1.0586, + "step": 173750 + }, + { + "epoch": 15.04, + "learning_rate": 3.4990037252014206e-05, + "loss": 0.984, + "step": 173760 + }, + { + "epoch": 15.04, + "learning_rate": 3.49891709261024e-05, + "loss": 1.0462, + "step": 173770 + }, + { + "epoch": 15.04, + "learning_rate": 3.498830460019059e-05, + "loss": 1.0433, + "step": 173780 + }, + { + "epoch": 15.04, + "learning_rate": 3.498743827427878e-05, + "loss": 0.9529, + "step": 173790 + }, + { + "epoch": 15.04, + "learning_rate": 3.498657194836698e-05, + "loss": 1.0637, + "step": 173800 + }, + { + "epoch": 15.04, + "learning_rate": 3.4985705622455165e-05, + "loss": 1.0894, + "step": 173810 + }, + { + "epoch": 15.05, + "learning_rate": 3.498483929654336e-05, + "loss": 1.0647, + "step": 173820 + }, + { + "epoch": 15.05, + "learning_rate": 3.4983972970631554e-05, + "loss": 1.0608, + "step": 173830 + }, + { + "epoch": 15.05, + "learning_rate": 3.498310664471974e-05, + "loss": 1.0648, + "step": 173840 + }, + { + "epoch": 15.05, + "learning_rate": 3.4982240318807936e-05, + "loss": 0.9331, + "step": 173850 + }, + { + "epoch": 15.05, + "learning_rate": 3.498137399289613e-05, + "loss": 1.0542, + "step": 173860 + }, + { + "epoch": 15.05, + "learning_rate": 3.498050766698432e-05, + "loss": 1.0638, + "step": 173870 + }, + { + "epoch": 15.05, + "learning_rate": 3.497964134107251e-05, + "loss": 0.9888, + "step": 173880 + }, + { + "epoch": 15.05, + "learning_rate": 3.49787750151607e-05, + "loss": 1.0658, + "step": 173890 + }, + { + "epoch": 15.05, + "learning_rate": 3.4977908689248894e-05, + "loss": 1.0221, + "step": 173900 + }, + { + "epoch": 15.05, + "learning_rate": 3.497704236333709e-05, + "loss": 1.0344, + "step": 173910 + }, + { + "epoch": 15.05, + "learning_rate": 3.497617603742528e-05, + "loss": 1.072, + "step": 173920 + }, + { + "epoch": 15.05, + "learning_rate": 3.497530971151347e-05, + "loss": 1.0909, + "step": 173930 + }, + { + "epoch": 15.06, + "learning_rate": 3.4974443385601666e-05, + "loss": 1.0637, + "step": 173940 + }, + { + "epoch": 15.06, + "learning_rate": 3.497357705968985e-05, + "loss": 1.0855, + "step": 173950 + }, + { + "epoch": 15.06, + "learning_rate": 3.497271073377805e-05, + "loss": 1.065, + "step": 173960 + }, + { + "epoch": 15.06, + "learning_rate": 3.497184440786624e-05, + "loss": 1.0221, + "step": 173970 + }, + { + "epoch": 15.06, + "learning_rate": 3.497097808195443e-05, + "loss": 1.0019, + "step": 173980 + }, + { + "epoch": 15.06, + "learning_rate": 3.4970111756042624e-05, + "loss": 1.0385, + "step": 173990 + }, + { + "epoch": 15.06, + "learning_rate": 3.496924543013081e-05, + "loss": 1.053, + "step": 174000 + }, + { + "epoch": 15.06, + "learning_rate": 3.4968379104219006e-05, + "loss": 1.0317, + "step": 174010 + }, + { + "epoch": 15.06, + "learning_rate": 3.49675127783072e-05, + "loss": 1.0292, + "step": 174020 + }, + { + "epoch": 15.06, + "learning_rate": 3.496664645239539e-05, + "loss": 0.9866, + "step": 174030 + }, + { + "epoch": 15.06, + "learning_rate": 3.496578012648358e-05, + "loss": 0.994, + "step": 174040 + }, + { + "epoch": 15.06, + "learning_rate": 3.496491380057178e-05, + "loss": 1.042, + "step": 174050 + }, + { + "epoch": 15.07, + "learning_rate": 3.4964047474659965e-05, + "loss": 1.0819, + "step": 174060 + }, + { + "epoch": 15.07, + "learning_rate": 3.496318114874816e-05, + "loss": 1.0367, + "step": 174070 + }, + { + "epoch": 15.07, + "learning_rate": 3.4962314822836354e-05, + "loss": 1.0713, + "step": 174080 + }, + { + "epoch": 15.07, + "learning_rate": 3.496144849692454e-05, + "loss": 1.0582, + "step": 174090 + }, + { + "epoch": 15.07, + "learning_rate": 3.4960582171012736e-05, + "loss": 1.0071, + "step": 174100 + }, + { + "epoch": 15.07, + "learning_rate": 3.4959715845100924e-05, + "loss": 1.04, + "step": 174110 + }, + { + "epoch": 15.07, + "learning_rate": 3.495884951918912e-05, + "loss": 1.0593, + "step": 174120 + }, + { + "epoch": 15.07, + "learning_rate": 3.495798319327731e-05, + "loss": 1.0377, + "step": 174130 + }, + { + "epoch": 15.07, + "learning_rate": 3.49571168673655e-05, + "loss": 0.9833, + "step": 174140 + }, + { + "epoch": 15.07, + "learning_rate": 3.4956250541453695e-05, + "loss": 0.9956, + "step": 174150 + }, + { + "epoch": 15.07, + "learning_rate": 3.495538421554189e-05, + "loss": 1.0045, + "step": 174160 + }, + { + "epoch": 15.08, + "learning_rate": 3.495451788963008e-05, + "loss": 1.0408, + "step": 174170 + }, + { + "epoch": 15.08, + "learning_rate": 3.495365156371827e-05, + "loss": 1.0989, + "step": 174180 + }, + { + "epoch": 15.08, + "learning_rate": 3.495278523780646e-05, + "loss": 1.0491, + "step": 174190 + }, + { + "epoch": 15.08, + "learning_rate": 3.4951918911894654e-05, + "loss": 0.9815, + "step": 174200 + }, + { + "epoch": 15.08, + "learning_rate": 3.495105258598285e-05, + "loss": 1.0359, + "step": 174210 + }, + { + "epoch": 15.08, + "learning_rate": 3.4950186260071036e-05, + "loss": 1.0882, + "step": 174220 + }, + { + "epoch": 15.08, + "learning_rate": 3.494931993415923e-05, + "loss": 1.083, + "step": 174230 + }, + { + "epoch": 15.08, + "learning_rate": 3.4948453608247425e-05, + "loss": 1.0476, + "step": 174240 + }, + { + "epoch": 15.08, + "learning_rate": 3.494758728233561e-05, + "loss": 1.0423, + "step": 174250 + }, + { + "epoch": 15.08, + "learning_rate": 3.494672095642381e-05, + "loss": 1.0613, + "step": 174260 + }, + { + "epoch": 15.08, + "learning_rate": 3.4945854630512e-05, + "loss": 1.0315, + "step": 174270 + }, + { + "epoch": 15.08, + "learning_rate": 3.494498830460019e-05, + "loss": 0.9992, + "step": 174280 + }, + { + "epoch": 15.09, + "learning_rate": 3.494412197868838e-05, + "loss": 1.022, + "step": 174290 + }, + { + "epoch": 15.09, + "learning_rate": 3.494325565277657e-05, + "loss": 1.0362, + "step": 174300 + }, + { + "epoch": 15.09, + "learning_rate": 3.4942389326864765e-05, + "loss": 0.9982, + "step": 174310 + }, + { + "epoch": 15.09, + "learning_rate": 3.494152300095296e-05, + "loss": 1.0673, + "step": 174320 + }, + { + "epoch": 15.09, + "learning_rate": 3.494065667504115e-05, + "loss": 1.0675, + "step": 174330 + }, + { + "epoch": 15.09, + "learning_rate": 3.493979034912934e-05, + "loss": 1.0068, + "step": 174340 + }, + { + "epoch": 15.09, + "learning_rate": 3.4938924023217536e-05, + "loss": 1.0655, + "step": 174350 + }, + { + "epoch": 15.09, + "learning_rate": 3.4938057697305724e-05, + "loss": 1.0282, + "step": 174360 + }, + { + "epoch": 15.09, + "learning_rate": 3.493719137139392e-05, + "loss": 1.0141, + "step": 174370 + }, + { + "epoch": 15.09, + "learning_rate": 3.493632504548211e-05, + "loss": 1.054, + "step": 174380 + }, + { + "epoch": 15.09, + "learning_rate": 3.49354587195703e-05, + "loss": 1.0203, + "step": 174390 + }, + { + "epoch": 15.1, + "learning_rate": 3.4934592393658495e-05, + "loss": 1.0321, + "step": 174400 + }, + { + "epoch": 15.1, + "learning_rate": 3.493372606774668e-05, + "loss": 1.0145, + "step": 174410 + }, + { + "epoch": 15.1, + "learning_rate": 3.493285974183488e-05, + "loss": 1.0102, + "step": 174420 + }, + { + "epoch": 15.1, + "learning_rate": 3.493199341592307e-05, + "loss": 1.05, + "step": 174430 + }, + { + "epoch": 15.1, + "learning_rate": 3.493112709001126e-05, + "loss": 1.0427, + "step": 174440 + }, + { + "epoch": 15.1, + "learning_rate": 3.4930260764099454e-05, + "loss": 1.0039, + "step": 174450 + }, + { + "epoch": 15.1, + "learning_rate": 3.492939443818765e-05, + "loss": 1.0731, + "step": 174460 + }, + { + "epoch": 15.1, + "learning_rate": 3.4928528112275836e-05, + "loss": 1.0113, + "step": 174470 + }, + { + "epoch": 15.1, + "learning_rate": 3.492766178636403e-05, + "loss": 0.9964, + "step": 174480 + }, + { + "epoch": 15.1, + "learning_rate": 3.4926795460452225e-05, + "loss": 1.0412, + "step": 174490 + }, + { + "epoch": 15.1, + "learning_rate": 3.492592913454041e-05, + "loss": 1.0564, + "step": 174500 + }, + { + "epoch": 15.1, + "learning_rate": 3.492506280862861e-05, + "loss": 1.0852, + "step": 174510 + }, + { + "epoch": 15.11, + "learning_rate": 3.4924196482716795e-05, + "loss": 1.0722, + "step": 174520 + }, + { + "epoch": 15.11, + "learning_rate": 3.492333015680499e-05, + "loss": 1.1005, + "step": 174530 + }, + { + "epoch": 15.11, + "learning_rate": 3.4922463830893184e-05, + "loss": 1.0606, + "step": 174540 + }, + { + "epoch": 15.11, + "learning_rate": 3.492159750498137e-05, + "loss": 1.0657, + "step": 174550 + }, + { + "epoch": 15.11, + "learning_rate": 3.4920731179069566e-05, + "loss": 1.0352, + "step": 174560 + }, + { + "epoch": 15.11, + "learning_rate": 3.491986485315776e-05, + "loss": 0.9787, + "step": 174570 + }, + { + "epoch": 15.11, + "learning_rate": 3.491899852724595e-05, + "loss": 1.0244, + "step": 174580 + }, + { + "epoch": 15.11, + "learning_rate": 3.491813220133414e-05, + "loss": 1.0418, + "step": 174590 + }, + { + "epoch": 15.11, + "learning_rate": 3.491726587542234e-05, + "loss": 1.0341, + "step": 174600 + }, + { + "epoch": 15.11, + "learning_rate": 3.4916399549510524e-05, + "loss": 1.0523, + "step": 174610 + }, + { + "epoch": 15.11, + "learning_rate": 3.491553322359872e-05, + "loss": 1.0305, + "step": 174620 + }, + { + "epoch": 15.12, + "learning_rate": 3.4914666897686907e-05, + "loss": 1.0782, + "step": 174630 + }, + { + "epoch": 15.12, + "learning_rate": 3.49138005717751e-05, + "loss": 1.0179, + "step": 174640 + }, + { + "epoch": 15.12, + "learning_rate": 3.4912934245863296e-05, + "loss": 1.0368, + "step": 174650 + }, + { + "epoch": 15.12, + "learning_rate": 3.491206791995148e-05, + "loss": 1.0513, + "step": 174660 + }, + { + "epoch": 15.12, + "learning_rate": 3.491120159403968e-05, + "loss": 0.989, + "step": 174670 + }, + { + "epoch": 15.12, + "learning_rate": 3.491033526812787e-05, + "loss": 1.0256, + "step": 174680 + }, + { + "epoch": 15.12, + "learning_rate": 3.490946894221606e-05, + "loss": 1.0547, + "step": 174690 + }, + { + "epoch": 15.12, + "learning_rate": 3.4908602616304254e-05, + "loss": 1.1048, + "step": 174700 + }, + { + "epoch": 15.12, + "learning_rate": 3.490773629039245e-05, + "loss": 1.1091, + "step": 174710 + }, + { + "epoch": 15.12, + "learning_rate": 3.4906869964480636e-05, + "loss": 1.0541, + "step": 174720 + }, + { + "epoch": 15.12, + "learning_rate": 3.490600363856883e-05, + "loss": 1.0244, + "step": 174730 + }, + { + "epoch": 15.12, + "learning_rate": 3.490513731265702e-05, + "loss": 1.0156, + "step": 174740 + }, + { + "epoch": 15.13, + "learning_rate": 3.490427098674521e-05, + "loss": 1.1213, + "step": 174750 + }, + { + "epoch": 15.13, + "learning_rate": 3.490340466083341e-05, + "loss": 1.0947, + "step": 174760 + }, + { + "epoch": 15.13, + "learning_rate": 3.4902538334921595e-05, + "loss": 1.0279, + "step": 174770 + }, + { + "epoch": 15.13, + "learning_rate": 3.490167200900979e-05, + "loss": 1.0012, + "step": 174780 + }, + { + "epoch": 15.13, + "learning_rate": 3.4900805683097984e-05, + "loss": 1.0895, + "step": 174790 + }, + { + "epoch": 15.13, + "learning_rate": 3.489993935718617e-05, + "loss": 1.0478, + "step": 174800 + }, + { + "epoch": 15.13, + "learning_rate": 3.4899073031274366e-05, + "loss": 1.0088, + "step": 174810 + }, + { + "epoch": 15.13, + "learning_rate": 3.489820670536256e-05, + "loss": 1.0357, + "step": 174820 + }, + { + "epoch": 15.13, + "learning_rate": 3.489734037945075e-05, + "loss": 1.0152, + "step": 174830 + }, + { + "epoch": 15.13, + "learning_rate": 3.489647405353894e-05, + "loss": 1.0704, + "step": 174840 + }, + { + "epoch": 15.13, + "learning_rate": 3.489560772762713e-05, + "loss": 1.0565, + "step": 174850 + }, + { + "epoch": 15.14, + "learning_rate": 3.4894741401715325e-05, + "loss": 1.0371, + "step": 174860 + }, + { + "epoch": 15.14, + "learning_rate": 3.489387507580352e-05, + "loss": 1.0261, + "step": 174870 + }, + { + "epoch": 15.14, + "learning_rate": 3.489300874989171e-05, + "loss": 1.0813, + "step": 174880 + }, + { + "epoch": 15.14, + "learning_rate": 3.48921424239799e-05, + "loss": 1.0181, + "step": 174890 + }, + { + "epoch": 15.14, + "learning_rate": 3.4891276098068096e-05, + "loss": 1.0288, + "step": 174900 + }, + { + "epoch": 15.14, + "learning_rate": 3.4890409772156284e-05, + "loss": 0.9892, + "step": 174910 + }, + { + "epoch": 15.14, + "learning_rate": 3.488954344624448e-05, + "loss": 1.0244, + "step": 174920 + }, + { + "epoch": 15.14, + "learning_rate": 3.4888677120332666e-05, + "loss": 1.0433, + "step": 174930 + }, + { + "epoch": 15.14, + "learning_rate": 3.488781079442086e-05, + "loss": 1.0304, + "step": 174940 + }, + { + "epoch": 15.14, + "learning_rate": 3.4886944468509055e-05, + "loss": 1.0053, + "step": 174950 + }, + { + "epoch": 15.14, + "learning_rate": 3.488607814259724e-05, + "loss": 1.0616, + "step": 174960 + }, + { + "epoch": 15.14, + "learning_rate": 3.488521181668544e-05, + "loss": 1.0168, + "step": 174970 + }, + { + "epoch": 15.15, + "learning_rate": 3.488434549077363e-05, + "loss": 1.0092, + "step": 174980 + }, + { + "epoch": 15.15, + "learning_rate": 3.488347916486182e-05, + "loss": 1.0235, + "step": 174990 + }, + { + "epoch": 15.15, + "learning_rate": 3.488261283895001e-05, + "loss": 1.0627, + "step": 175000 + }, + { + "epoch": 15.15, + "learning_rate": 3.488174651303821e-05, + "loss": 1.0392, + "step": 175010 + }, + { + "epoch": 15.15, + "learning_rate": 3.4880880187126395e-05, + "loss": 1.0097, + "step": 175020 + }, + { + "epoch": 15.15, + "learning_rate": 3.488001386121459e-05, + "loss": 1.0624, + "step": 175030 + }, + { + "epoch": 15.15, + "learning_rate": 3.487914753530278e-05, + "loss": 1.0302, + "step": 175040 + }, + { + "epoch": 15.15, + "learning_rate": 3.487828120939097e-05, + "loss": 1.0404, + "step": 175050 + }, + { + "epoch": 15.15, + "learning_rate": 3.4877414883479166e-05, + "loss": 1.0998, + "step": 175060 + }, + { + "epoch": 15.15, + "learning_rate": 3.4876548557567354e-05, + "loss": 0.9794, + "step": 175070 + }, + { + "epoch": 15.15, + "learning_rate": 3.487568223165555e-05, + "loss": 1.0115, + "step": 175080 + }, + { + "epoch": 15.15, + "learning_rate": 3.487481590574374e-05, + "loss": 1.0276, + "step": 175090 + }, + { + "epoch": 15.16, + "learning_rate": 3.487394957983193e-05, + "loss": 1.0721, + "step": 175100 + }, + { + "epoch": 15.16, + "learning_rate": 3.4873083253920125e-05, + "loss": 1.0458, + "step": 175110 + }, + { + "epoch": 15.16, + "learning_rate": 3.487221692800832e-05, + "loss": 1.0665, + "step": 175120 + }, + { + "epoch": 15.16, + "learning_rate": 3.487135060209651e-05, + "loss": 1.0405, + "step": 175130 + }, + { + "epoch": 15.16, + "learning_rate": 3.48704842761847e-05, + "loss": 1.0066, + "step": 175140 + }, + { + "epoch": 15.16, + "learning_rate": 3.486961795027289e-05, + "loss": 1.0772, + "step": 175150 + }, + { + "epoch": 15.16, + "learning_rate": 3.4868751624361084e-05, + "loss": 1.0182, + "step": 175160 + }, + { + "epoch": 15.16, + "learning_rate": 3.486788529844928e-05, + "loss": 1.003, + "step": 175170 + }, + { + "epoch": 15.16, + "learning_rate": 3.4867018972537466e-05, + "loss": 1.0587, + "step": 175180 + }, + { + "epoch": 15.16, + "learning_rate": 3.486615264662566e-05, + "loss": 1.0236, + "step": 175190 + }, + { + "epoch": 15.16, + "learning_rate": 3.4865286320713855e-05, + "loss": 1.0278, + "step": 175200 + }, + { + "epoch": 15.17, + "learning_rate": 3.486441999480204e-05, + "loss": 1.0443, + "step": 175210 + }, + { + "epoch": 15.17, + "learning_rate": 3.486355366889024e-05, + "loss": 1.0266, + "step": 175220 + }, + { + "epoch": 15.17, + "learning_rate": 3.486268734297843e-05, + "loss": 1.0449, + "step": 175230 + }, + { + "epoch": 15.17, + "learning_rate": 3.486182101706662e-05, + "loss": 1.0655, + "step": 175240 + }, + { + "epoch": 15.17, + "learning_rate": 3.4860954691154814e-05, + "loss": 1.0097, + "step": 175250 + }, + { + "epoch": 15.17, + "learning_rate": 3.4860088365243e-05, + "loss": 0.9919, + "step": 175260 + }, + { + "epoch": 15.17, + "learning_rate": 3.4859222039331196e-05, + "loss": 1.0539, + "step": 175270 + }, + { + "epoch": 15.17, + "learning_rate": 3.485835571341939e-05, + "loss": 1.0542, + "step": 175280 + }, + { + "epoch": 15.17, + "learning_rate": 3.485748938750758e-05, + "loss": 1.0443, + "step": 175290 + }, + { + "epoch": 15.17, + "learning_rate": 3.485662306159577e-05, + "loss": 1.1046, + "step": 175300 + }, + { + "epoch": 15.17, + "learning_rate": 3.485575673568397e-05, + "loss": 1.0332, + "step": 175310 + }, + { + "epoch": 15.17, + "learning_rate": 3.4854890409772154e-05, + "loss": 0.9818, + "step": 175320 + }, + { + "epoch": 15.18, + "learning_rate": 3.485402408386035e-05, + "loss": 1.0752, + "step": 175330 + }, + { + "epoch": 15.18, + "learning_rate": 3.485315775794854e-05, + "loss": 1.0197, + "step": 175340 + }, + { + "epoch": 15.18, + "learning_rate": 3.485229143203673e-05, + "loss": 0.9976, + "step": 175350 + }, + { + "epoch": 15.18, + "learning_rate": 3.4851425106124925e-05, + "loss": 1.034, + "step": 175360 + }, + { + "epoch": 15.18, + "learning_rate": 3.485055878021311e-05, + "loss": 1.0649, + "step": 175370 + }, + { + "epoch": 15.18, + "learning_rate": 3.484969245430131e-05, + "loss": 1.0828, + "step": 175380 + }, + { + "epoch": 15.18, + "learning_rate": 3.48488261283895e-05, + "loss": 0.9611, + "step": 175390 + }, + { + "epoch": 15.18, + "learning_rate": 3.484795980247769e-05, + "loss": 1.0971, + "step": 175400 + }, + { + "epoch": 15.18, + "learning_rate": 3.4847093476565884e-05, + "loss": 1.0369, + "step": 175410 + }, + { + "epoch": 15.18, + "learning_rate": 3.484622715065408e-05, + "loss": 1.0311, + "step": 175420 + }, + { + "epoch": 15.18, + "learning_rate": 3.4845360824742266e-05, + "loss": 1.014, + "step": 175430 + }, + { + "epoch": 15.19, + "learning_rate": 3.484449449883046e-05, + "loss": 0.9909, + "step": 175440 + }, + { + "epoch": 15.19, + "learning_rate": 3.4843628172918655e-05, + "loss": 0.9992, + "step": 175450 + }, + { + "epoch": 15.19, + "learning_rate": 3.484276184700684e-05, + "loss": 1.0074, + "step": 175460 + }, + { + "epoch": 15.19, + "learning_rate": 3.484189552109504e-05, + "loss": 1.049, + "step": 175470 + }, + { + "epoch": 15.19, + "learning_rate": 3.4841029195183225e-05, + "loss": 1.0552, + "step": 175480 + }, + { + "epoch": 15.19, + "learning_rate": 3.484016286927142e-05, + "loss": 1.0363, + "step": 175490 + }, + { + "epoch": 15.19, + "learning_rate": 3.4839296543359614e-05, + "loss": 1.0373, + "step": 175500 + }, + { + "epoch": 15.19, + "learning_rate": 3.48384302174478e-05, + "loss": 1.025, + "step": 175510 + }, + { + "epoch": 15.19, + "learning_rate": 3.4837563891535996e-05, + "loss": 1.0254, + "step": 175520 + }, + { + "epoch": 15.19, + "learning_rate": 3.483669756562419e-05, + "loss": 1.0408, + "step": 175530 + }, + { + "epoch": 15.19, + "learning_rate": 3.483583123971238e-05, + "loss": 1.0465, + "step": 175540 + }, + { + "epoch": 15.19, + "learning_rate": 3.483496491380057e-05, + "loss": 1.0656, + "step": 175550 + }, + { + "epoch": 15.2, + "learning_rate": 3.483409858788877e-05, + "loss": 1.0042, + "step": 175560 + }, + { + "epoch": 15.2, + "learning_rate": 3.4833232261976955e-05, + "loss": 1.0507, + "step": 175570 + }, + { + "epoch": 15.2, + "learning_rate": 3.483236593606515e-05, + "loss": 1.0047, + "step": 175580 + }, + { + "epoch": 15.2, + "learning_rate": 3.483149961015334e-05, + "loss": 1.0085, + "step": 175590 + }, + { + "epoch": 15.2, + "learning_rate": 3.483063328424153e-05, + "loss": 1.0238, + "step": 175600 + }, + { + "epoch": 15.2, + "learning_rate": 3.4829766958329726e-05, + "loss": 1.0435, + "step": 175610 + }, + { + "epoch": 15.2, + "learning_rate": 3.4828900632417913e-05, + "loss": 1.0626, + "step": 175620 + }, + { + "epoch": 15.2, + "learning_rate": 3.482803430650611e-05, + "loss": 1.0969, + "step": 175630 + }, + { + "epoch": 15.2, + "learning_rate": 3.48271679805943e-05, + "loss": 1.0005, + "step": 175640 + }, + { + "epoch": 15.2, + "learning_rate": 3.482630165468249e-05, + "loss": 1.055, + "step": 175650 + }, + { + "epoch": 15.2, + "learning_rate": 3.4825435328770685e-05, + "loss": 1.0383, + "step": 175660 + }, + { + "epoch": 15.21, + "learning_rate": 3.482456900285887e-05, + "loss": 1.0398, + "step": 175670 + }, + { + "epoch": 15.21, + "learning_rate": 3.482370267694707e-05, + "loss": 1.0726, + "step": 175680 + }, + { + "epoch": 15.21, + "learning_rate": 3.482283635103526e-05, + "loss": 1.0928, + "step": 175690 + }, + { + "epoch": 15.21, + "learning_rate": 3.482197002512345e-05, + "loss": 1.0302, + "step": 175700 + }, + { + "epoch": 15.21, + "learning_rate": 3.482110369921164e-05, + "loss": 1.0185, + "step": 175710 + }, + { + "epoch": 15.21, + "learning_rate": 3.482023737329984e-05, + "loss": 1.0623, + "step": 175720 + }, + { + "epoch": 15.21, + "learning_rate": 3.4819371047388025e-05, + "loss": 1.063, + "step": 175730 + }, + { + "epoch": 15.21, + "learning_rate": 3.481850472147622e-05, + "loss": 1.0336, + "step": 175740 + }, + { + "epoch": 15.21, + "learning_rate": 3.4817638395564414e-05, + "loss": 1.0951, + "step": 175750 + }, + { + "epoch": 15.21, + "learning_rate": 3.48167720696526e-05, + "loss": 1.0695, + "step": 175760 + }, + { + "epoch": 15.21, + "learning_rate": 3.4815905743740796e-05, + "loss": 1.0566, + "step": 175770 + }, + { + "epoch": 15.21, + "learning_rate": 3.4815039417828984e-05, + "loss": 1.0733, + "step": 175780 + }, + { + "epoch": 15.22, + "learning_rate": 3.481417309191718e-05, + "loss": 1.0231, + "step": 175790 + }, + { + "epoch": 15.22, + "learning_rate": 3.481330676600537e-05, + "loss": 1.0489, + "step": 175800 + }, + { + "epoch": 15.22, + "learning_rate": 3.481244044009356e-05, + "loss": 1.0006, + "step": 175810 + }, + { + "epoch": 15.22, + "learning_rate": 3.4811574114181755e-05, + "loss": 1.0403, + "step": 175820 + }, + { + "epoch": 15.22, + "learning_rate": 3.481070778826995e-05, + "loss": 1.1259, + "step": 175830 + }, + { + "epoch": 15.22, + "learning_rate": 3.480984146235814e-05, + "loss": 1.0016, + "step": 175840 + }, + { + "epoch": 15.22, + "learning_rate": 3.480897513644633e-05, + "loss": 1.0106, + "step": 175850 + }, + { + "epoch": 15.22, + "learning_rate": 3.4808108810534526e-05, + "loss": 0.9599, + "step": 175860 + }, + { + "epoch": 15.22, + "learning_rate": 3.4807242484622714e-05, + "loss": 0.9799, + "step": 175870 + }, + { + "epoch": 15.22, + "learning_rate": 3.480637615871091e-05, + "loss": 1.0967, + "step": 175880 + }, + { + "epoch": 15.22, + "learning_rate": 3.4805509832799096e-05, + "loss": 1.0655, + "step": 175890 + }, + { + "epoch": 15.23, + "learning_rate": 3.480464350688729e-05, + "loss": 0.9873, + "step": 175900 + }, + { + "epoch": 15.23, + "learning_rate": 3.4803777180975485e-05, + "loss": 1.0667, + "step": 175910 + }, + { + "epoch": 15.23, + "learning_rate": 3.480291085506367e-05, + "loss": 1.0662, + "step": 175920 + }, + { + "epoch": 15.23, + "learning_rate": 3.480204452915187e-05, + "loss": 1.0298, + "step": 175930 + }, + { + "epoch": 15.23, + "learning_rate": 3.480117820324006e-05, + "loss": 0.9879, + "step": 175940 + }, + { + "epoch": 15.23, + "learning_rate": 3.480031187732825e-05, + "loss": 1.0619, + "step": 175950 + }, + { + "epoch": 15.23, + "learning_rate": 3.4799445551416444e-05, + "loss": 1.0626, + "step": 175960 + }, + { + "epoch": 15.23, + "learning_rate": 3.479857922550464e-05, + "loss": 1.0614, + "step": 175970 + }, + { + "epoch": 15.23, + "learning_rate": 3.4797712899592826e-05, + "loss": 1.0539, + "step": 175980 + }, + { + "epoch": 15.23, + "learning_rate": 3.479684657368102e-05, + "loss": 1.0185, + "step": 175990 + }, + { + "epoch": 15.23, + "learning_rate": 3.479598024776921e-05, + "loss": 1.0482, + "step": 176000 + }, + { + "epoch": 15.23, + "learning_rate": 3.47951139218574e-05, + "loss": 0.9751, + "step": 176010 + }, + { + "epoch": 15.24, + "learning_rate": 3.47942475959456e-05, + "loss": 1.0444, + "step": 176020 + }, + { + "epoch": 15.24, + "learning_rate": 3.4793381270033784e-05, + "loss": 1.0483, + "step": 176030 + }, + { + "epoch": 15.24, + "learning_rate": 3.479251494412198e-05, + "loss": 1.0458, + "step": 176040 + }, + { + "epoch": 15.24, + "learning_rate": 3.479164861821017e-05, + "loss": 0.9846, + "step": 176050 + }, + { + "epoch": 15.24, + "learning_rate": 3.479078229229836e-05, + "loss": 1.0553, + "step": 176060 + }, + { + "epoch": 15.24, + "learning_rate": 3.4789915966386555e-05, + "loss": 1.0485, + "step": 176070 + }, + { + "epoch": 15.24, + "learning_rate": 3.478904964047475e-05, + "loss": 1.0584, + "step": 176080 + }, + { + "epoch": 15.24, + "learning_rate": 3.478818331456294e-05, + "loss": 1.067, + "step": 176090 + }, + { + "epoch": 15.24, + "learning_rate": 3.478731698865113e-05, + "loss": 1.0269, + "step": 176100 + }, + { + "epoch": 15.24, + "learning_rate": 3.478645066273932e-05, + "loss": 1.0185, + "step": 176110 + }, + { + "epoch": 15.24, + "learning_rate": 3.4785584336827514e-05, + "loss": 1.0316, + "step": 176120 + }, + { + "epoch": 15.24, + "learning_rate": 3.478471801091571e-05, + "loss": 1.033, + "step": 176130 + }, + { + "epoch": 15.25, + "learning_rate": 3.4783851685003896e-05, + "loss": 0.9982, + "step": 176140 + }, + { + "epoch": 15.25, + "learning_rate": 3.478298535909209e-05, + "loss": 1.0201, + "step": 176150 + }, + { + "epoch": 15.25, + "learning_rate": 3.4782119033180285e-05, + "loss": 1.0994, + "step": 176160 + }, + { + "epoch": 15.25, + "learning_rate": 3.478125270726847e-05, + "loss": 1.0105, + "step": 176170 + }, + { + "epoch": 15.25, + "learning_rate": 3.478038638135667e-05, + "loss": 1.0679, + "step": 176180 + }, + { + "epoch": 15.25, + "learning_rate": 3.477952005544486e-05, + "loss": 1.0891, + "step": 176190 + }, + { + "epoch": 15.25, + "learning_rate": 3.477865372953305e-05, + "loss": 1.0801, + "step": 176200 + }, + { + "epoch": 15.25, + "learning_rate": 3.4777787403621244e-05, + "loss": 1.1033, + "step": 176210 + }, + { + "epoch": 15.25, + "learning_rate": 3.477692107770943e-05, + "loss": 1.0095, + "step": 176220 + }, + { + "epoch": 15.25, + "learning_rate": 3.4776054751797626e-05, + "loss": 1.0863, + "step": 176230 + }, + { + "epoch": 15.25, + "learning_rate": 3.477518842588582e-05, + "loss": 1.0205, + "step": 176240 + }, + { + "epoch": 15.26, + "learning_rate": 3.477432209997401e-05, + "loss": 1.0378, + "step": 176250 + }, + { + "epoch": 15.26, + "learning_rate": 3.47734557740622e-05, + "loss": 1.0697, + "step": 176260 + }, + { + "epoch": 15.26, + "learning_rate": 3.47725894481504e-05, + "loss": 1.0855, + "step": 176270 + }, + { + "epoch": 15.26, + "learning_rate": 3.4771723122238585e-05, + "loss": 1.0458, + "step": 176280 + }, + { + "epoch": 15.26, + "learning_rate": 3.477085679632678e-05, + "loss": 0.9928, + "step": 176290 + }, + { + "epoch": 15.26, + "learning_rate": 3.4769990470414974e-05, + "loss": 1.0191, + "step": 176300 + }, + { + "epoch": 15.26, + "learning_rate": 3.476912414450316e-05, + "loss": 1.033, + "step": 176310 + }, + { + "epoch": 15.26, + "learning_rate": 3.4768257818591356e-05, + "loss": 1.0359, + "step": 176320 + }, + { + "epoch": 15.26, + "learning_rate": 3.4767391492679543e-05, + "loss": 1.0707, + "step": 176330 + }, + { + "epoch": 15.26, + "learning_rate": 3.476652516676774e-05, + "loss": 1.009, + "step": 176340 + }, + { + "epoch": 15.26, + "learning_rate": 3.476565884085593e-05, + "loss": 1.08, + "step": 176350 + }, + { + "epoch": 15.26, + "learning_rate": 3.476479251494412e-05, + "loss": 1.0335, + "step": 176360 + }, + { + "epoch": 15.27, + "learning_rate": 3.4763926189032314e-05, + "loss": 1.0917, + "step": 176370 + }, + { + "epoch": 15.27, + "learning_rate": 3.476305986312051e-05, + "loss": 1.0562, + "step": 176380 + }, + { + "epoch": 15.27, + "learning_rate": 3.47621935372087e-05, + "loss": 1.0585, + "step": 176390 + }, + { + "epoch": 15.27, + "learning_rate": 3.476132721129689e-05, + "loss": 1.0446, + "step": 176400 + }, + { + "epoch": 15.27, + "learning_rate": 3.4760460885385086e-05, + "loss": 1.0843, + "step": 176410 + }, + { + "epoch": 15.27, + "learning_rate": 3.475959455947327e-05, + "loss": 1.0664, + "step": 176420 + }, + { + "epoch": 15.27, + "learning_rate": 3.475872823356147e-05, + "loss": 1.0033, + "step": 176430 + }, + { + "epoch": 15.27, + "learning_rate": 3.4757861907649655e-05, + "loss": 1.0932, + "step": 176440 + }, + { + "epoch": 15.27, + "learning_rate": 3.475699558173785e-05, + "loss": 1.0733, + "step": 176450 + }, + { + "epoch": 15.27, + "learning_rate": 3.4756129255826044e-05, + "loss": 1.0581, + "step": 176460 + }, + { + "epoch": 15.27, + "learning_rate": 3.475526292991423e-05, + "loss": 1.0889, + "step": 176470 + }, + { + "epoch": 15.28, + "learning_rate": 3.4754396604002426e-05, + "loss": 1.0806, + "step": 176480 + }, + { + "epoch": 15.28, + "learning_rate": 3.475353027809062e-05, + "loss": 1.0263, + "step": 176490 + }, + { + "epoch": 15.28, + "learning_rate": 3.475266395217881e-05, + "loss": 1.0493, + "step": 176500 + }, + { + "epoch": 15.28, + "learning_rate": 3.4751797626267e-05, + "loss": 1.0628, + "step": 176510 + }, + { + "epoch": 15.28, + "learning_rate": 3.475093130035519e-05, + "loss": 1.0156, + "step": 176520 + }, + { + "epoch": 15.28, + "learning_rate": 3.4750064974443385e-05, + "loss": 0.9994, + "step": 176530 + }, + { + "epoch": 15.28, + "learning_rate": 3.474919864853158e-05, + "loss": 1.1186, + "step": 176540 + }, + { + "epoch": 15.28, + "learning_rate": 3.474833232261977e-05, + "loss": 1.0644, + "step": 176550 + }, + { + "epoch": 15.28, + "learning_rate": 3.474746599670796e-05, + "loss": 1.0556, + "step": 176560 + }, + { + "epoch": 15.28, + "learning_rate": 3.4746599670796156e-05, + "loss": 1.0531, + "step": 176570 + }, + { + "epoch": 15.28, + "learning_rate": 3.4745733344884344e-05, + "loss": 1.0183, + "step": 176580 + }, + { + "epoch": 15.28, + "learning_rate": 3.474486701897254e-05, + "loss": 1.0266, + "step": 176590 + }, + { + "epoch": 15.29, + "learning_rate": 3.474400069306073e-05, + "loss": 1.0898, + "step": 176600 + }, + { + "epoch": 15.29, + "learning_rate": 3.474313436714892e-05, + "loss": 1.0267, + "step": 176610 + }, + { + "epoch": 15.29, + "learning_rate": 3.4742268041237115e-05, + "loss": 1.0299, + "step": 176620 + }, + { + "epoch": 15.29, + "learning_rate": 3.47414017153253e-05, + "loss": 1.0724, + "step": 176630 + }, + { + "epoch": 15.29, + "learning_rate": 3.47405353894135e-05, + "loss": 0.9996, + "step": 176640 + }, + { + "epoch": 15.29, + "learning_rate": 3.473966906350169e-05, + "loss": 1.0392, + "step": 176650 + }, + { + "epoch": 15.29, + "learning_rate": 3.473880273758988e-05, + "loss": 1.0636, + "step": 176660 + }, + { + "epoch": 15.29, + "learning_rate": 3.4737936411678074e-05, + "loss": 1.0595, + "step": 176670 + }, + { + "epoch": 15.29, + "learning_rate": 3.473707008576627e-05, + "loss": 1.0062, + "step": 176680 + }, + { + "epoch": 15.29, + "learning_rate": 3.4736203759854456e-05, + "loss": 1.0367, + "step": 176690 + }, + { + "epoch": 15.29, + "learning_rate": 3.473533743394265e-05, + "loss": 1.0813, + "step": 176700 + }, + { + "epoch": 15.3, + "learning_rate": 3.4734471108030845e-05, + "loss": 1.0394, + "step": 176710 + }, + { + "epoch": 15.3, + "learning_rate": 3.473360478211903e-05, + "loss": 1.0525, + "step": 176720 + }, + { + "epoch": 15.3, + "learning_rate": 3.473273845620723e-05, + "loss": 1.045, + "step": 176730 + }, + { + "epoch": 15.3, + "learning_rate": 3.4731872130295414e-05, + "loss": 1.0534, + "step": 176740 + }, + { + "epoch": 15.3, + "learning_rate": 3.473100580438361e-05, + "loss": 1.0642, + "step": 176750 + }, + { + "epoch": 15.3, + "learning_rate": 3.47301394784718e-05, + "loss": 1.0762, + "step": 176760 + }, + { + "epoch": 15.3, + "learning_rate": 3.472927315255999e-05, + "loss": 1.0343, + "step": 176770 + }, + { + "epoch": 15.3, + "learning_rate": 3.4728406826648185e-05, + "loss": 1.076, + "step": 176780 + }, + { + "epoch": 15.3, + "learning_rate": 3.472754050073638e-05, + "loss": 1.0423, + "step": 176790 + }, + { + "epoch": 15.3, + "learning_rate": 3.472667417482457e-05, + "loss": 1.0019, + "step": 176800 + }, + { + "epoch": 15.3, + "learning_rate": 3.472580784891276e-05, + "loss": 0.9891, + "step": 176810 + }, + { + "epoch": 15.3, + "learning_rate": 3.4724941523000956e-05, + "loss": 1.0324, + "step": 176820 + }, + { + "epoch": 15.31, + "learning_rate": 3.4724075197089144e-05, + "loss": 1.0297, + "step": 176830 + }, + { + "epoch": 15.31, + "learning_rate": 3.472320887117734e-05, + "loss": 1.0376, + "step": 176840 + }, + { + "epoch": 15.31, + "learning_rate": 3.4722342545265526e-05, + "loss": 1.086, + "step": 176850 + }, + { + "epoch": 15.31, + "learning_rate": 3.472147621935372e-05, + "loss": 0.9806, + "step": 176860 + }, + { + "epoch": 15.31, + "learning_rate": 3.4720609893441915e-05, + "loss": 1.0317, + "step": 176870 + }, + { + "epoch": 15.31, + "learning_rate": 3.47197435675301e-05, + "loss": 1.0785, + "step": 176880 + }, + { + "epoch": 15.31, + "learning_rate": 3.47188772416183e-05, + "loss": 1.0834, + "step": 176890 + }, + { + "epoch": 15.31, + "learning_rate": 3.471801091570649e-05, + "loss": 0.9573, + "step": 176900 + }, + { + "epoch": 15.31, + "learning_rate": 3.471714458979468e-05, + "loss": 1.038, + "step": 176910 + }, + { + "epoch": 15.31, + "learning_rate": 3.4716278263882874e-05, + "loss": 1.1143, + "step": 176920 + }, + { + "epoch": 15.31, + "learning_rate": 3.471541193797107e-05, + "loss": 1.0329, + "step": 176930 + }, + { + "epoch": 15.32, + "learning_rate": 3.4714545612059256e-05, + "loss": 1.0432, + "step": 176940 + }, + { + "epoch": 15.32, + "learning_rate": 3.471367928614745e-05, + "loss": 1.0386, + "step": 176950 + }, + { + "epoch": 15.32, + "learning_rate": 3.471281296023564e-05, + "loss": 1.0208, + "step": 176960 + }, + { + "epoch": 15.32, + "learning_rate": 3.471194663432383e-05, + "loss": 1.0411, + "step": 176970 + }, + { + "epoch": 15.32, + "learning_rate": 3.471108030841203e-05, + "loss": 1.0257, + "step": 176980 + }, + { + "epoch": 15.32, + "learning_rate": 3.4710213982500215e-05, + "loss": 0.999, + "step": 176990 + }, + { + "epoch": 15.32, + "learning_rate": 3.470934765658841e-05, + "loss": 1.0305, + "step": 177000 + }, + { + "epoch": 15.32, + "learning_rate": 3.4708481330676604e-05, + "loss": 1.0511, + "step": 177010 + }, + { + "epoch": 15.32, + "learning_rate": 3.470761500476479e-05, + "loss": 1.0577, + "step": 177020 + }, + { + "epoch": 15.32, + "learning_rate": 3.4706748678852986e-05, + "loss": 1.0343, + "step": 177030 + }, + { + "epoch": 15.32, + "learning_rate": 3.470588235294118e-05, + "loss": 1.0296, + "step": 177040 + }, + { + "epoch": 15.32, + "learning_rate": 3.470501602702937e-05, + "loss": 1.0074, + "step": 177050 + }, + { + "epoch": 15.33, + "learning_rate": 3.470414970111756e-05, + "loss": 1.0433, + "step": 177060 + }, + { + "epoch": 15.33, + "learning_rate": 3.470328337520575e-05, + "loss": 1.0224, + "step": 177070 + }, + { + "epoch": 15.33, + "learning_rate": 3.4702417049293944e-05, + "loss": 1.0363, + "step": 177080 + }, + { + "epoch": 15.33, + "learning_rate": 3.470155072338214e-05, + "loss": 1.0466, + "step": 177090 + }, + { + "epoch": 15.33, + "learning_rate": 3.4700684397470327e-05, + "loss": 1.0599, + "step": 177100 + }, + { + "epoch": 15.33, + "learning_rate": 3.469981807155852e-05, + "loss": 1.082, + "step": 177110 + }, + { + "epoch": 15.33, + "learning_rate": 3.4698951745646716e-05, + "loss": 1.0077, + "step": 177120 + }, + { + "epoch": 15.33, + "learning_rate": 3.46980854197349e-05, + "loss": 1.0007, + "step": 177130 + }, + { + "epoch": 15.33, + "learning_rate": 3.46972190938231e-05, + "loss": 1.056, + "step": 177140 + }, + { + "epoch": 15.33, + "learning_rate": 3.469635276791129e-05, + "loss": 1.0649, + "step": 177150 + }, + { + "epoch": 15.33, + "learning_rate": 3.469548644199948e-05, + "loss": 1.0261, + "step": 177160 + }, + { + "epoch": 15.34, + "learning_rate": 3.4694620116087674e-05, + "loss": 1.0343, + "step": 177170 + }, + { + "epoch": 15.34, + "learning_rate": 3.469375379017586e-05, + "loss": 1.1108, + "step": 177180 + }, + { + "epoch": 15.34, + "learning_rate": 3.4692887464264056e-05, + "loss": 1.0284, + "step": 177190 + }, + { + "epoch": 15.34, + "learning_rate": 3.469202113835225e-05, + "loss": 1.0655, + "step": 177200 + }, + { + "epoch": 15.34, + "learning_rate": 3.469115481244044e-05, + "loss": 0.998, + "step": 177210 + }, + { + "epoch": 15.34, + "learning_rate": 3.469028848652863e-05, + "loss": 1.049, + "step": 177220 + }, + { + "epoch": 15.34, + "learning_rate": 3.468942216061683e-05, + "loss": 1.1237, + "step": 177230 + }, + { + "epoch": 15.34, + "learning_rate": 3.4688555834705015e-05, + "loss": 1.0775, + "step": 177240 + }, + { + "epoch": 15.34, + "learning_rate": 3.468768950879321e-05, + "loss": 1.0439, + "step": 177250 + }, + { + "epoch": 15.34, + "learning_rate": 3.46868231828814e-05, + "loss": 1.0697, + "step": 177260 + }, + { + "epoch": 15.34, + "learning_rate": 3.468595685696959e-05, + "loss": 1.0421, + "step": 177270 + }, + { + "epoch": 15.34, + "learning_rate": 3.4685090531057786e-05, + "loss": 1.0395, + "step": 177280 + }, + { + "epoch": 15.35, + "learning_rate": 3.4684224205145974e-05, + "loss": 1.0378, + "step": 177290 + }, + { + "epoch": 15.35, + "learning_rate": 3.468335787923417e-05, + "loss": 1.1152, + "step": 177300 + }, + { + "epoch": 15.35, + "learning_rate": 3.468249155332236e-05, + "loss": 1.0116, + "step": 177310 + }, + { + "epoch": 15.35, + "learning_rate": 3.468162522741055e-05, + "loss": 1.0393, + "step": 177320 + }, + { + "epoch": 15.35, + "learning_rate": 3.4680758901498745e-05, + "loss": 1.0739, + "step": 177330 + }, + { + "epoch": 15.35, + "learning_rate": 3.467989257558694e-05, + "loss": 1.0485, + "step": 177340 + }, + { + "epoch": 15.35, + "learning_rate": 3.467902624967513e-05, + "loss": 1.0108, + "step": 177350 + }, + { + "epoch": 15.35, + "learning_rate": 3.467815992376332e-05, + "loss": 1.0855, + "step": 177360 + }, + { + "epoch": 15.35, + "learning_rate": 3.467729359785151e-05, + "loss": 1.0194, + "step": 177370 + }, + { + "epoch": 15.35, + "learning_rate": 3.4676427271939704e-05, + "loss": 1.0545, + "step": 177380 + }, + { + "epoch": 15.35, + "learning_rate": 3.46755609460279e-05, + "loss": 1.0282, + "step": 177390 + }, + { + "epoch": 15.35, + "learning_rate": 3.4674694620116086e-05, + "loss": 1.0035, + "step": 177400 + }, + { + "epoch": 15.36, + "learning_rate": 3.467382829420428e-05, + "loss": 1.0729, + "step": 177410 + }, + { + "epoch": 15.36, + "learning_rate": 3.4672961968292475e-05, + "loss": 1.0528, + "step": 177420 + }, + { + "epoch": 15.36, + "learning_rate": 3.467209564238066e-05, + "loss": 1.0287, + "step": 177430 + }, + { + "epoch": 15.36, + "learning_rate": 3.467122931646886e-05, + "loss": 1.0636, + "step": 177440 + }, + { + "epoch": 15.36, + "learning_rate": 3.467036299055705e-05, + "loss": 1.0376, + "step": 177450 + }, + { + "epoch": 15.36, + "learning_rate": 3.466949666464524e-05, + "loss": 1.059, + "step": 177460 + }, + { + "epoch": 15.36, + "learning_rate": 3.466863033873343e-05, + "loss": 1.0464, + "step": 177470 + }, + { + "epoch": 15.36, + "learning_rate": 3.466776401282162e-05, + "loss": 1.0522, + "step": 177480 + }, + { + "epoch": 15.36, + "learning_rate": 3.4666897686909815e-05, + "loss": 1.1145, + "step": 177490 + }, + { + "epoch": 15.36, + "learning_rate": 3.466603136099801e-05, + "loss": 1.063, + "step": 177500 + }, + { + "epoch": 15.36, + "learning_rate": 3.46651650350862e-05, + "loss": 1.0343, + "step": 177510 + }, + { + "epoch": 15.37, + "learning_rate": 3.466429870917439e-05, + "loss": 1.0467, + "step": 177520 + }, + { + "epoch": 15.37, + "learning_rate": 3.4663432383262586e-05, + "loss": 0.9997, + "step": 177530 + }, + { + "epoch": 15.37, + "learning_rate": 3.4662566057350774e-05, + "loss": 1.024, + "step": 177540 + }, + { + "epoch": 15.37, + "learning_rate": 3.466169973143897e-05, + "loss": 1.0727, + "step": 177550 + }, + { + "epoch": 15.37, + "learning_rate": 3.466083340552716e-05, + "loss": 1.0211, + "step": 177560 + }, + { + "epoch": 15.37, + "learning_rate": 3.465996707961535e-05, + "loss": 1.0259, + "step": 177570 + }, + { + "epoch": 15.37, + "learning_rate": 3.4659100753703545e-05, + "loss": 1.0277, + "step": 177580 + }, + { + "epoch": 15.37, + "learning_rate": 3.465823442779173e-05, + "loss": 1.0274, + "step": 177590 + }, + { + "epoch": 15.37, + "learning_rate": 3.465736810187993e-05, + "loss": 1.0172, + "step": 177600 + }, + { + "epoch": 15.37, + "learning_rate": 3.465650177596812e-05, + "loss": 1.0625, + "step": 177610 + }, + { + "epoch": 15.37, + "learning_rate": 3.465563545005631e-05, + "loss": 1.0709, + "step": 177620 + }, + { + "epoch": 15.37, + "learning_rate": 3.4654769124144504e-05, + "loss": 0.9947, + "step": 177630 + }, + { + "epoch": 15.38, + "learning_rate": 3.46539027982327e-05, + "loss": 1.0847, + "step": 177640 + }, + { + "epoch": 15.38, + "learning_rate": 3.4653036472320886e-05, + "loss": 1.1009, + "step": 177650 + }, + { + "epoch": 15.38, + "learning_rate": 3.465217014640908e-05, + "loss": 1.0334, + "step": 177660 + }, + { + "epoch": 15.38, + "learning_rate": 3.4651303820497275e-05, + "loss": 1.0136, + "step": 177670 + }, + { + "epoch": 15.38, + "learning_rate": 3.465043749458546e-05, + "loss": 1.0689, + "step": 177680 + }, + { + "epoch": 15.38, + "learning_rate": 3.464957116867366e-05, + "loss": 1.0844, + "step": 177690 + }, + { + "epoch": 15.38, + "learning_rate": 3.4648704842761845e-05, + "loss": 1.0058, + "step": 177700 + }, + { + "epoch": 15.38, + "learning_rate": 3.464783851685004e-05, + "loss": 1.0354, + "step": 177710 + }, + { + "epoch": 15.38, + "learning_rate": 3.4646972190938234e-05, + "loss": 1.075, + "step": 177720 + }, + { + "epoch": 15.38, + "learning_rate": 3.464610586502642e-05, + "loss": 1.0333, + "step": 177730 + }, + { + "epoch": 15.38, + "learning_rate": 3.4645239539114616e-05, + "loss": 1.0282, + "step": 177740 + }, + { + "epoch": 15.39, + "learning_rate": 3.464437321320281e-05, + "loss": 1.0795, + "step": 177750 + }, + { + "epoch": 15.39, + "learning_rate": 3.4643506887291e-05, + "loss": 1.0712, + "step": 177760 + }, + { + "epoch": 15.39, + "learning_rate": 3.464264056137919e-05, + "loss": 1.0592, + "step": 177770 + }, + { + "epoch": 15.39, + "learning_rate": 3.464177423546739e-05, + "loss": 1.086, + "step": 177780 + }, + { + "epoch": 15.39, + "learning_rate": 3.4640907909555574e-05, + "loss": 1.0419, + "step": 177790 + }, + { + "epoch": 15.39, + "learning_rate": 3.464004158364377e-05, + "loss": 1.0268, + "step": 177800 + }, + { + "epoch": 15.39, + "learning_rate": 3.4639175257731957e-05, + "loss": 1.0134, + "step": 177810 + }, + { + "epoch": 15.39, + "learning_rate": 3.463830893182015e-05, + "loss": 1.0119, + "step": 177820 + }, + { + "epoch": 15.39, + "learning_rate": 3.4637442605908345e-05, + "loss": 1.0537, + "step": 177830 + }, + { + "epoch": 15.39, + "learning_rate": 3.463657627999653e-05, + "loss": 1.0274, + "step": 177840 + }, + { + "epoch": 15.39, + "learning_rate": 3.463570995408473e-05, + "loss": 1.0535, + "step": 177850 + }, + { + "epoch": 15.39, + "learning_rate": 3.463484362817292e-05, + "loss": 1.0574, + "step": 177860 + }, + { + "epoch": 15.4, + "learning_rate": 3.463397730226111e-05, + "loss": 1.0892, + "step": 177870 + }, + { + "epoch": 15.4, + "learning_rate": 3.4633110976349304e-05, + "loss": 1.0723, + "step": 177880 + }, + { + "epoch": 15.4, + "learning_rate": 3.46322446504375e-05, + "loss": 1.0697, + "step": 177890 + }, + { + "epoch": 15.4, + "learning_rate": 3.4631378324525686e-05, + "loss": 1.0095, + "step": 177900 + }, + { + "epoch": 15.4, + "learning_rate": 3.463051199861388e-05, + "loss": 1.0351, + "step": 177910 + }, + { + "epoch": 15.4, + "learning_rate": 3.462964567270207e-05, + "loss": 1.0429, + "step": 177920 + }, + { + "epoch": 15.4, + "learning_rate": 3.462877934679026e-05, + "loss": 1.0417, + "step": 177930 + }, + { + "epoch": 15.4, + "learning_rate": 3.462791302087846e-05, + "loss": 1.0509, + "step": 177940 + }, + { + "epoch": 15.4, + "learning_rate": 3.4627046694966645e-05, + "loss": 1.0328, + "step": 177950 + }, + { + "epoch": 15.4, + "learning_rate": 3.462618036905484e-05, + "loss": 1.0509, + "step": 177960 + }, + { + "epoch": 15.4, + "learning_rate": 3.4625314043143034e-05, + "loss": 1.0698, + "step": 177970 + }, + { + "epoch": 15.41, + "learning_rate": 3.462444771723122e-05, + "loss": 1.01, + "step": 177980 + }, + { + "epoch": 15.41, + "learning_rate": 3.4623581391319416e-05, + "loss": 1.0415, + "step": 177990 + }, + { + "epoch": 15.41, + "learning_rate": 3.4622715065407604e-05, + "loss": 1.0276, + "step": 178000 + }, + { + "epoch": 15.41, + "learning_rate": 3.46218487394958e-05, + "loss": 0.9975, + "step": 178010 + }, + { + "epoch": 15.41, + "learning_rate": 3.462098241358399e-05, + "loss": 1.0467, + "step": 178020 + }, + { + "epoch": 15.41, + "learning_rate": 3.462011608767218e-05, + "loss": 1.0413, + "step": 178030 + }, + { + "epoch": 15.41, + "learning_rate": 3.4619249761760375e-05, + "loss": 1.0212, + "step": 178040 + }, + { + "epoch": 15.41, + "learning_rate": 3.461838343584857e-05, + "loss": 1.075, + "step": 178050 + }, + { + "epoch": 15.41, + "learning_rate": 3.461751710993676e-05, + "loss": 1.0656, + "step": 178060 + }, + { + "epoch": 15.41, + "learning_rate": 3.461665078402495e-05, + "loss": 1.0645, + "step": 178070 + }, + { + "epoch": 15.41, + "learning_rate": 3.4615784458113146e-05, + "loss": 1.056, + "step": 178080 + }, + { + "epoch": 15.41, + "learning_rate": 3.4614918132201333e-05, + "loss": 1.0352, + "step": 178090 + }, + { + "epoch": 15.42, + "learning_rate": 3.461405180628953e-05, + "loss": 1.0378, + "step": 178100 + }, + { + "epoch": 15.42, + "learning_rate": 3.4613185480377716e-05, + "loss": 0.9843, + "step": 178110 + }, + { + "epoch": 15.42, + "learning_rate": 3.461231915446591e-05, + "loss": 1.0618, + "step": 178120 + }, + { + "epoch": 15.42, + "learning_rate": 3.4611452828554105e-05, + "loss": 1.0295, + "step": 178130 + }, + { + "epoch": 15.42, + "learning_rate": 3.461058650264229e-05, + "loss": 1.0573, + "step": 178140 + }, + { + "epoch": 15.42, + "learning_rate": 3.460972017673049e-05, + "loss": 1.0332, + "step": 178150 + }, + { + "epoch": 15.42, + "learning_rate": 3.460885385081868e-05, + "loss": 1.0315, + "step": 178160 + }, + { + "epoch": 15.42, + "learning_rate": 3.460798752490687e-05, + "loss": 1.079, + "step": 178170 + }, + { + "epoch": 15.42, + "learning_rate": 3.460712119899506e-05, + "loss": 1.0261, + "step": 178180 + }, + { + "epoch": 15.42, + "learning_rate": 3.460625487308326e-05, + "loss": 1.0309, + "step": 178190 + }, + { + "epoch": 15.42, + "learning_rate": 3.4605388547171445e-05, + "loss": 0.9724, + "step": 178200 + }, + { + "epoch": 15.43, + "learning_rate": 3.460452222125964e-05, + "loss": 0.9796, + "step": 178210 + }, + { + "epoch": 15.43, + "learning_rate": 3.460365589534783e-05, + "loss": 1.1353, + "step": 178220 + }, + { + "epoch": 15.43, + "learning_rate": 3.460278956943602e-05, + "loss": 0.9772, + "step": 178230 + }, + { + "epoch": 15.43, + "learning_rate": 3.4601923243524216e-05, + "loss": 0.9723, + "step": 178240 + }, + { + "epoch": 15.43, + "learning_rate": 3.4601056917612404e-05, + "loss": 1.0262, + "step": 178250 + }, + { + "epoch": 15.43, + "learning_rate": 3.46001905917006e-05, + "loss": 1.0424, + "step": 178260 + }, + { + "epoch": 15.43, + "learning_rate": 3.459932426578879e-05, + "loss": 1.0497, + "step": 178270 + }, + { + "epoch": 15.43, + "learning_rate": 3.459845793987698e-05, + "loss": 1.0728, + "step": 178280 + }, + { + "epoch": 15.43, + "learning_rate": 3.4597591613965175e-05, + "loss": 1.0085, + "step": 178290 + }, + { + "epoch": 15.43, + "learning_rate": 3.459672528805337e-05, + "loss": 1.0265, + "step": 178300 + }, + { + "epoch": 15.43, + "learning_rate": 3.459585896214156e-05, + "loss": 0.9856, + "step": 178310 + }, + { + "epoch": 15.43, + "learning_rate": 3.459499263622975e-05, + "loss": 1.1262, + "step": 178320 + }, + { + "epoch": 15.44, + "learning_rate": 3.459412631031794e-05, + "loss": 1.0385, + "step": 178330 + }, + { + "epoch": 15.44, + "learning_rate": 3.4593259984406134e-05, + "loss": 1.0666, + "step": 178340 + }, + { + "epoch": 15.44, + "learning_rate": 3.459239365849433e-05, + "loss": 1.0336, + "step": 178350 + }, + { + "epoch": 15.44, + "learning_rate": 3.4591527332582516e-05, + "loss": 1.0133, + "step": 178360 + }, + { + "epoch": 15.44, + "learning_rate": 3.459066100667071e-05, + "loss": 1.0073, + "step": 178370 + }, + { + "epoch": 15.44, + "learning_rate": 3.4589794680758905e-05, + "loss": 1.0121, + "step": 178380 + }, + { + "epoch": 15.44, + "learning_rate": 3.458892835484709e-05, + "loss": 1.0519, + "step": 178390 + }, + { + "epoch": 15.44, + "learning_rate": 3.458806202893529e-05, + "loss": 1.0218, + "step": 178400 + }, + { + "epoch": 15.44, + "learning_rate": 3.458719570302348e-05, + "loss": 1.0592, + "step": 178410 + }, + { + "epoch": 15.44, + "learning_rate": 3.458632937711167e-05, + "loss": 1.0051, + "step": 178420 + }, + { + "epoch": 15.44, + "learning_rate": 3.4585463051199864e-05, + "loss": 1.0741, + "step": 178430 + }, + { + "epoch": 15.44, + "learning_rate": 3.458459672528805e-05, + "loss": 0.9369, + "step": 178440 + }, + { + "epoch": 15.45, + "learning_rate": 3.4583730399376246e-05, + "loss": 1.0327, + "step": 178450 + }, + { + "epoch": 15.45, + "learning_rate": 3.458286407346444e-05, + "loss": 0.9966, + "step": 178460 + }, + { + "epoch": 15.45, + "learning_rate": 3.458199774755263e-05, + "loss": 1.076, + "step": 178470 + }, + { + "epoch": 15.45, + "learning_rate": 3.458113142164082e-05, + "loss": 1.0512, + "step": 178480 + }, + { + "epoch": 15.45, + "learning_rate": 3.458026509572902e-05, + "loss": 1.0363, + "step": 178490 + }, + { + "epoch": 15.45, + "learning_rate": 3.4579398769817204e-05, + "loss": 1.0334, + "step": 178500 + }, + { + "epoch": 15.45, + "learning_rate": 3.45785324439054e-05, + "loss": 1.0598, + "step": 178510 + }, + { + "epoch": 15.45, + "learning_rate": 3.457766611799359e-05, + "loss": 1.0288, + "step": 178520 + }, + { + "epoch": 15.45, + "learning_rate": 3.457679979208178e-05, + "loss": 1.0715, + "step": 178530 + }, + { + "epoch": 15.45, + "learning_rate": 3.4575933466169975e-05, + "loss": 1.0254, + "step": 178540 + }, + { + "epoch": 15.45, + "learning_rate": 3.457506714025816e-05, + "loss": 1.0017, + "step": 178550 + }, + { + "epoch": 15.46, + "learning_rate": 3.457420081434636e-05, + "loss": 1.0339, + "step": 178560 + }, + { + "epoch": 15.46, + "learning_rate": 3.457333448843455e-05, + "loss": 1.0066, + "step": 178570 + }, + { + "epoch": 15.46, + "learning_rate": 3.457246816252274e-05, + "loss": 1.0819, + "step": 178580 + }, + { + "epoch": 15.46, + "learning_rate": 3.4571601836610934e-05, + "loss": 1.0633, + "step": 178590 + }, + { + "epoch": 15.46, + "learning_rate": 3.457073551069913e-05, + "loss": 0.9629, + "step": 178600 + }, + { + "epoch": 15.46, + "learning_rate": 3.4569869184787316e-05, + "loss": 1.0163, + "step": 178610 + }, + { + "epoch": 15.46, + "learning_rate": 3.456900285887551e-05, + "loss": 1.0044, + "step": 178620 + }, + { + "epoch": 15.46, + "learning_rate": 3.4568136532963705e-05, + "loss": 1.0167, + "step": 178630 + }, + { + "epoch": 15.46, + "learning_rate": 3.456727020705189e-05, + "loss": 1.03, + "step": 178640 + }, + { + "epoch": 15.46, + "learning_rate": 3.456640388114009e-05, + "loss": 0.9979, + "step": 178650 + }, + { + "epoch": 15.46, + "learning_rate": 3.4565537555228275e-05, + "loss": 1.0128, + "step": 178660 + }, + { + "epoch": 15.46, + "learning_rate": 3.456467122931647e-05, + "loss": 1.0836, + "step": 178670 + }, + { + "epoch": 15.47, + "learning_rate": 3.4563804903404664e-05, + "loss": 1.1004, + "step": 178680 + }, + { + "epoch": 15.47, + "learning_rate": 3.456293857749285e-05, + "loss": 1.0887, + "step": 178690 + }, + { + "epoch": 15.47, + "learning_rate": 3.4562072251581046e-05, + "loss": 0.9793, + "step": 178700 + }, + { + "epoch": 15.47, + "learning_rate": 3.456120592566924e-05, + "loss": 1.097, + "step": 178710 + }, + { + "epoch": 15.47, + "learning_rate": 3.456033959975743e-05, + "loss": 1.026, + "step": 178720 + }, + { + "epoch": 15.47, + "learning_rate": 3.455947327384562e-05, + "loss": 1.0607, + "step": 178730 + }, + { + "epoch": 15.47, + "learning_rate": 3.455860694793382e-05, + "loss": 1.0285, + "step": 178740 + }, + { + "epoch": 15.47, + "learning_rate": 3.4557740622022005e-05, + "loss": 1.0615, + "step": 178750 + }, + { + "epoch": 15.47, + "learning_rate": 3.45568742961102e-05, + "loss": 1.0569, + "step": 178760 + }, + { + "epoch": 15.47, + "learning_rate": 3.455600797019839e-05, + "loss": 1.0846, + "step": 178770 + }, + { + "epoch": 15.47, + "learning_rate": 3.455514164428658e-05, + "loss": 1.0509, + "step": 178780 + }, + { + "epoch": 15.48, + "learning_rate": 3.4554275318374776e-05, + "loss": 1.0418, + "step": 178790 + }, + { + "epoch": 15.48, + "learning_rate": 3.4553408992462963e-05, + "loss": 1.0846, + "step": 178800 + }, + { + "epoch": 15.48, + "learning_rate": 3.455254266655116e-05, + "loss": 0.9913, + "step": 178810 + }, + { + "epoch": 15.48, + "learning_rate": 3.455167634063935e-05, + "loss": 1.0522, + "step": 178820 + }, + { + "epoch": 15.48, + "learning_rate": 3.455081001472754e-05, + "loss": 1.0535, + "step": 178830 + }, + { + "epoch": 15.48, + "learning_rate": 3.4549943688815734e-05, + "loss": 1.0427, + "step": 178840 + }, + { + "epoch": 15.48, + "learning_rate": 3.454907736290392e-05, + "loss": 1.0559, + "step": 178850 + }, + { + "epoch": 15.48, + "learning_rate": 3.454821103699212e-05, + "loss": 1.0643, + "step": 178860 + }, + { + "epoch": 15.48, + "learning_rate": 3.454734471108031e-05, + "loss": 1.0621, + "step": 178870 + }, + { + "epoch": 15.48, + "learning_rate": 3.45464783851685e-05, + "loss": 1.0474, + "step": 178880 + }, + { + "epoch": 15.48, + "learning_rate": 3.454561205925669e-05, + "loss": 1.0509, + "step": 178890 + }, + { + "epoch": 15.48, + "learning_rate": 3.454474573334489e-05, + "loss": 1.0646, + "step": 178900 + }, + { + "epoch": 15.49, + "learning_rate": 3.4543879407433075e-05, + "loss": 1.0523, + "step": 178910 + }, + { + "epoch": 15.49, + "learning_rate": 3.454301308152127e-05, + "loss": 1.0812, + "step": 178920 + }, + { + "epoch": 15.49, + "learning_rate": 3.4542146755609464e-05, + "loss": 1.0754, + "step": 178930 + }, + { + "epoch": 15.49, + "learning_rate": 3.454128042969765e-05, + "loss": 1.047, + "step": 178940 + }, + { + "epoch": 15.49, + "learning_rate": 3.4540414103785846e-05, + "loss": 1.0248, + "step": 178950 + }, + { + "epoch": 15.49, + "learning_rate": 3.4539547777874034e-05, + "loss": 1.0655, + "step": 178960 + }, + { + "epoch": 15.49, + "learning_rate": 3.453868145196223e-05, + "loss": 1.0518, + "step": 178970 + }, + { + "epoch": 15.49, + "learning_rate": 3.453781512605042e-05, + "loss": 0.9842, + "step": 178980 + }, + { + "epoch": 15.49, + "learning_rate": 3.453694880013861e-05, + "loss": 1.0378, + "step": 178990 + }, + { + "epoch": 15.49, + "learning_rate": 3.4536082474226805e-05, + "loss": 1.011, + "step": 179000 + }, + { + "epoch": 15.49, + "learning_rate": 3.4535216148315e-05, + "loss": 1.0353, + "step": 179010 + }, + { + "epoch": 15.5, + "learning_rate": 3.453434982240319e-05, + "loss": 1.0589, + "step": 179020 + }, + { + "epoch": 15.5, + "learning_rate": 3.453348349649138e-05, + "loss": 1.0138, + "step": 179030 + }, + { + "epoch": 15.5, + "learning_rate": 3.4532617170579576e-05, + "loss": 1.0545, + "step": 179040 + }, + { + "epoch": 15.5, + "learning_rate": 3.4531750844667764e-05, + "loss": 1.0373, + "step": 179050 + }, + { + "epoch": 15.5, + "learning_rate": 3.453088451875596e-05, + "loss": 1.0468, + "step": 179060 + }, + { + "epoch": 15.5, + "learning_rate": 3.4530018192844146e-05, + "loss": 0.9945, + "step": 179070 + }, + { + "epoch": 15.5, + "learning_rate": 3.452915186693234e-05, + "loss": 1.0794, + "step": 179080 + }, + { + "epoch": 15.5, + "learning_rate": 3.4528285541020535e-05, + "loss": 1.0196, + "step": 179090 + }, + { + "epoch": 15.5, + "learning_rate": 3.452741921510872e-05, + "loss": 0.9927, + "step": 179100 + }, + { + "epoch": 15.5, + "learning_rate": 3.452655288919692e-05, + "loss": 1.021, + "step": 179110 + }, + { + "epoch": 15.5, + "learning_rate": 3.452568656328511e-05, + "loss": 1.0146, + "step": 179120 + }, + { + "epoch": 15.5, + "learning_rate": 3.45248202373733e-05, + "loss": 1.0846, + "step": 179130 + }, + { + "epoch": 15.51, + "learning_rate": 3.4523953911461494e-05, + "loss": 1.0397, + "step": 179140 + }, + { + "epoch": 15.51, + "learning_rate": 3.452308758554969e-05, + "loss": 1.0564, + "step": 179150 + }, + { + "epoch": 15.51, + "learning_rate": 3.4522221259637876e-05, + "loss": 1.1132, + "step": 179160 + }, + { + "epoch": 15.51, + "learning_rate": 3.452135493372607e-05, + "loss": 1.0387, + "step": 179170 + }, + { + "epoch": 15.51, + "learning_rate": 3.452048860781426e-05, + "loss": 0.9913, + "step": 179180 + }, + { + "epoch": 15.51, + "learning_rate": 3.451962228190245e-05, + "loss": 1.0477, + "step": 179190 + }, + { + "epoch": 15.51, + "learning_rate": 3.451875595599065e-05, + "loss": 1.0265, + "step": 179200 + }, + { + "epoch": 15.51, + "learning_rate": 3.4517889630078834e-05, + "loss": 1.0811, + "step": 179210 + }, + { + "epoch": 15.51, + "learning_rate": 3.451702330416703e-05, + "loss": 0.9639, + "step": 179220 + }, + { + "epoch": 15.51, + "learning_rate": 3.451615697825522e-05, + "loss": 1.0645, + "step": 179230 + }, + { + "epoch": 15.51, + "learning_rate": 3.451529065234341e-05, + "loss": 1.0065, + "step": 179240 + }, + { + "epoch": 15.52, + "learning_rate": 3.4514424326431605e-05, + "loss": 1.0704, + "step": 179250 + }, + { + "epoch": 15.52, + "learning_rate": 3.45135580005198e-05, + "loss": 1.0017, + "step": 179260 + }, + { + "epoch": 15.52, + "learning_rate": 3.451269167460799e-05, + "loss": 1.0408, + "step": 179270 + }, + { + "epoch": 15.52, + "learning_rate": 3.451182534869618e-05, + "loss": 1.0391, + "step": 179280 + }, + { + "epoch": 15.52, + "learning_rate": 3.451095902278437e-05, + "loss": 0.9761, + "step": 179290 + }, + { + "epoch": 15.52, + "learning_rate": 3.4510092696872564e-05, + "loss": 0.9986, + "step": 179300 + }, + { + "epoch": 15.52, + "learning_rate": 3.450922637096076e-05, + "loss": 1.0862, + "step": 179310 + }, + { + "epoch": 15.52, + "learning_rate": 3.4508360045048946e-05, + "loss": 1.1175, + "step": 179320 + }, + { + "epoch": 15.52, + "learning_rate": 3.450749371913714e-05, + "loss": 1.062, + "step": 179330 + }, + { + "epoch": 15.52, + "learning_rate": 3.4506627393225335e-05, + "loss": 1.0254, + "step": 179340 + }, + { + "epoch": 15.52, + "learning_rate": 3.450576106731352e-05, + "loss": 0.9945, + "step": 179350 + }, + { + "epoch": 15.52, + "learning_rate": 3.450489474140172e-05, + "loss": 1.0478, + "step": 179360 + }, + { + "epoch": 15.53, + "learning_rate": 3.450402841548991e-05, + "loss": 1.0386, + "step": 179370 + }, + { + "epoch": 15.53, + "learning_rate": 3.45031620895781e-05, + "loss": 1.0404, + "step": 179380 + }, + { + "epoch": 15.53, + "learning_rate": 3.4502295763666294e-05, + "loss": 1.0575, + "step": 179390 + }, + { + "epoch": 15.53, + "learning_rate": 3.450142943775448e-05, + "loss": 1.1004, + "step": 179400 + }, + { + "epoch": 15.53, + "learning_rate": 3.4500563111842676e-05, + "loss": 1.0322, + "step": 179410 + }, + { + "epoch": 15.53, + "learning_rate": 3.449969678593087e-05, + "loss": 1.0839, + "step": 179420 + }, + { + "epoch": 15.53, + "learning_rate": 3.449883046001906e-05, + "loss": 1.035, + "step": 179430 + }, + { + "epoch": 15.53, + "learning_rate": 3.449796413410725e-05, + "loss": 1.0499, + "step": 179440 + }, + { + "epoch": 15.53, + "learning_rate": 3.449709780819545e-05, + "loss": 1.0492, + "step": 179450 + }, + { + "epoch": 15.53, + "learning_rate": 3.4496231482283635e-05, + "loss": 1.0503, + "step": 179460 + }, + { + "epoch": 15.53, + "learning_rate": 3.449536515637183e-05, + "loss": 1.0487, + "step": 179470 + }, + { + "epoch": 15.53, + "learning_rate": 3.4494498830460024e-05, + "loss": 1.0575, + "step": 179480 + }, + { + "epoch": 15.54, + "learning_rate": 3.449363250454821e-05, + "loss": 1.0878, + "step": 179490 + }, + { + "epoch": 15.54, + "learning_rate": 3.4492766178636406e-05, + "loss": 1.0539, + "step": 179500 + }, + { + "epoch": 15.54, + "learning_rate": 3.4491899852724593e-05, + "loss": 1.0496, + "step": 179510 + }, + { + "epoch": 15.54, + "learning_rate": 3.449103352681279e-05, + "loss": 1.0585, + "step": 179520 + }, + { + "epoch": 15.54, + "learning_rate": 3.449016720090098e-05, + "loss": 1.0257, + "step": 179530 + }, + { + "epoch": 15.54, + "learning_rate": 3.448930087498917e-05, + "loss": 1.0877, + "step": 179540 + }, + { + "epoch": 15.54, + "learning_rate": 3.4488434549077364e-05, + "loss": 1.0283, + "step": 179550 + }, + { + "epoch": 15.54, + "learning_rate": 3.448756822316556e-05, + "loss": 1.0044, + "step": 179560 + }, + { + "epoch": 15.54, + "learning_rate": 3.4486701897253747e-05, + "loss": 1.0767, + "step": 179570 + }, + { + "epoch": 15.54, + "learning_rate": 3.448583557134194e-05, + "loss": 1.0673, + "step": 179580 + }, + { + "epoch": 15.54, + "learning_rate": 3.448496924543013e-05, + "loss": 1.086, + "step": 179590 + }, + { + "epoch": 15.55, + "learning_rate": 3.448410291951832e-05, + "loss": 0.9895, + "step": 179600 + }, + { + "epoch": 15.55, + "learning_rate": 3.448323659360652e-05, + "loss": 1.0386, + "step": 179610 + }, + { + "epoch": 15.55, + "learning_rate": 3.4482370267694705e-05, + "loss": 1.0577, + "step": 179620 + }, + { + "epoch": 15.55, + "learning_rate": 3.44815039417829e-05, + "loss": 1.0296, + "step": 179630 + }, + { + "epoch": 15.55, + "learning_rate": 3.4480637615871094e-05, + "loss": 1.0427, + "step": 179640 + }, + { + "epoch": 15.55, + "learning_rate": 3.447977128995928e-05, + "loss": 0.9846, + "step": 179650 + }, + { + "epoch": 15.55, + "learning_rate": 3.4478904964047476e-05, + "loss": 1.0638, + "step": 179660 + }, + { + "epoch": 15.55, + "learning_rate": 3.447803863813567e-05, + "loss": 0.9909, + "step": 179670 + }, + { + "epoch": 15.55, + "learning_rate": 3.447717231222386e-05, + "loss": 1.1338, + "step": 179680 + }, + { + "epoch": 15.55, + "learning_rate": 3.447630598631205e-05, + "loss": 1.0373, + "step": 179690 + }, + { + "epoch": 15.55, + "learning_rate": 3.447543966040024e-05, + "loss": 1.1121, + "step": 179700 + }, + { + "epoch": 15.55, + "learning_rate": 3.4474573334488435e-05, + "loss": 1.0414, + "step": 179710 + }, + { + "epoch": 15.56, + "learning_rate": 3.447370700857663e-05, + "loss": 1.0514, + "step": 179720 + }, + { + "epoch": 15.56, + "learning_rate": 3.447284068266482e-05, + "loss": 1.0703, + "step": 179730 + }, + { + "epoch": 15.56, + "learning_rate": 3.447197435675301e-05, + "loss": 1.0303, + "step": 179740 + }, + { + "epoch": 15.56, + "learning_rate": 3.4471108030841206e-05, + "loss": 1.075, + "step": 179750 + }, + { + "epoch": 15.56, + "learning_rate": 3.4470241704929394e-05, + "loss": 1.0685, + "step": 179760 + }, + { + "epoch": 15.56, + "learning_rate": 3.446937537901759e-05, + "loss": 1.0722, + "step": 179770 + }, + { + "epoch": 15.56, + "learning_rate": 3.446850905310578e-05, + "loss": 1.0332, + "step": 179780 + }, + { + "epoch": 15.56, + "learning_rate": 3.446764272719397e-05, + "loss": 1.0612, + "step": 179790 + }, + { + "epoch": 15.56, + "learning_rate": 3.4466776401282165e-05, + "loss": 1.0611, + "step": 179800 + }, + { + "epoch": 15.56, + "learning_rate": 3.446591007537035e-05, + "loss": 1.0131, + "step": 179810 + }, + { + "epoch": 15.56, + "learning_rate": 3.446504374945855e-05, + "loss": 1.0032, + "step": 179820 + }, + { + "epoch": 15.57, + "learning_rate": 3.446417742354674e-05, + "loss": 0.9978, + "step": 179830 + }, + { + "epoch": 15.57, + "learning_rate": 3.446331109763493e-05, + "loss": 1.0211, + "step": 179840 + }, + { + "epoch": 15.57, + "learning_rate": 3.4462444771723124e-05, + "loss": 1.0516, + "step": 179850 + }, + { + "epoch": 15.57, + "learning_rate": 3.446157844581132e-05, + "loss": 1.0877, + "step": 179860 + }, + { + "epoch": 15.57, + "learning_rate": 3.4460712119899506e-05, + "loss": 1.0247, + "step": 179870 + }, + { + "epoch": 15.57, + "learning_rate": 3.44598457939877e-05, + "loss": 1.067, + "step": 179880 + }, + { + "epoch": 15.57, + "learning_rate": 3.4458979468075895e-05, + "loss": 1.0185, + "step": 179890 + }, + { + "epoch": 15.57, + "learning_rate": 3.445811314216408e-05, + "loss": 1.0489, + "step": 179900 + }, + { + "epoch": 15.57, + "learning_rate": 3.445724681625228e-05, + "loss": 1.0593, + "step": 179910 + }, + { + "epoch": 15.57, + "learning_rate": 3.4456380490340464e-05, + "loss": 1.0581, + "step": 179920 + }, + { + "epoch": 15.57, + "learning_rate": 3.445551416442866e-05, + "loss": 1.0411, + "step": 179930 + }, + { + "epoch": 15.57, + "learning_rate": 3.445464783851685e-05, + "loss": 1.0113, + "step": 179940 + }, + { + "epoch": 15.58, + "learning_rate": 3.445378151260504e-05, + "loss": 1.0624, + "step": 179950 + }, + { + "epoch": 15.58, + "learning_rate": 3.4452915186693235e-05, + "loss": 1.0537, + "step": 179960 + }, + { + "epoch": 15.58, + "learning_rate": 3.445204886078143e-05, + "loss": 1.0619, + "step": 179970 + }, + { + "epoch": 15.58, + "learning_rate": 3.445118253486962e-05, + "loss": 1.033, + "step": 179980 + }, + { + "epoch": 15.58, + "learning_rate": 3.445031620895781e-05, + "loss": 1.0193, + "step": 179990 + }, + { + "epoch": 15.58, + "learning_rate": 3.4449449883046006e-05, + "loss": 1.016, + "step": 180000 + }, + { + "epoch": 15.58, + "learning_rate": 3.4448583557134194e-05, + "loss": 1.0556, + "step": 180010 + }, + { + "epoch": 15.58, + "learning_rate": 3.444771723122239e-05, + "loss": 1.0343, + "step": 180020 + }, + { + "epoch": 15.58, + "learning_rate": 3.4446850905310576e-05, + "loss": 1.077, + "step": 180030 + }, + { + "epoch": 15.58, + "learning_rate": 3.444598457939877e-05, + "loss": 0.9956, + "step": 180040 + }, + { + "epoch": 15.58, + "learning_rate": 3.4445118253486965e-05, + "loss": 0.9952, + "step": 180050 + }, + { + "epoch": 15.59, + "learning_rate": 3.444425192757515e-05, + "loss": 0.9967, + "step": 180060 + }, + { + "epoch": 15.59, + "learning_rate": 3.444338560166335e-05, + "loss": 1.0193, + "step": 180070 + }, + { + "epoch": 15.59, + "learning_rate": 3.444251927575154e-05, + "loss": 1.1202, + "step": 180080 + }, + { + "epoch": 15.59, + "learning_rate": 3.444165294983973e-05, + "loss": 1.0568, + "step": 180090 + }, + { + "epoch": 15.59, + "learning_rate": 3.4440786623927924e-05, + "loss": 1.0422, + "step": 180100 + }, + { + "epoch": 15.59, + "learning_rate": 3.443992029801612e-05, + "loss": 1.0327, + "step": 180110 + }, + { + "epoch": 15.59, + "learning_rate": 3.4439053972104306e-05, + "loss": 1.0572, + "step": 180120 + }, + { + "epoch": 15.59, + "learning_rate": 3.44381876461925e-05, + "loss": 1.0382, + "step": 180130 + }, + { + "epoch": 15.59, + "learning_rate": 3.443732132028069e-05, + "loss": 1.0136, + "step": 180140 + }, + { + "epoch": 15.59, + "learning_rate": 3.443645499436888e-05, + "loss": 1.0309, + "step": 180150 + }, + { + "epoch": 15.59, + "learning_rate": 3.443558866845708e-05, + "loss": 0.9977, + "step": 180160 + }, + { + "epoch": 15.59, + "learning_rate": 3.4434722342545265e-05, + "loss": 1.0826, + "step": 180170 + }, + { + "epoch": 15.6, + "learning_rate": 3.443385601663346e-05, + "loss": 1.0812, + "step": 180180 + }, + { + "epoch": 15.6, + "learning_rate": 3.4432989690721654e-05, + "loss": 0.98, + "step": 180190 + }, + { + "epoch": 15.6, + "learning_rate": 3.443212336480984e-05, + "loss": 1.05, + "step": 180200 + }, + { + "epoch": 15.6, + "learning_rate": 3.4431257038898036e-05, + "loss": 1.06, + "step": 180210 + }, + { + "epoch": 15.6, + "learning_rate": 3.443039071298623e-05, + "loss": 0.9992, + "step": 180220 + }, + { + "epoch": 15.6, + "learning_rate": 3.442952438707442e-05, + "loss": 1.0745, + "step": 180230 + }, + { + "epoch": 15.6, + "learning_rate": 3.442865806116261e-05, + "loss": 1.053, + "step": 180240 + }, + { + "epoch": 15.6, + "learning_rate": 3.44277917352508e-05, + "loss": 1.0458, + "step": 180250 + }, + { + "epoch": 15.6, + "learning_rate": 3.4426925409338994e-05, + "loss": 1.0371, + "step": 180260 + }, + { + "epoch": 15.6, + "learning_rate": 3.442605908342719e-05, + "loss": 1.0308, + "step": 180270 + }, + { + "epoch": 15.6, + "learning_rate": 3.4425192757515377e-05, + "loss": 1.0588, + "step": 180280 + }, + { + "epoch": 15.61, + "learning_rate": 3.442432643160357e-05, + "loss": 1.038, + "step": 180290 + }, + { + "epoch": 15.61, + "learning_rate": 3.4423460105691765e-05, + "loss": 1.0275, + "step": 180300 + }, + { + "epoch": 15.61, + "learning_rate": 3.442259377977995e-05, + "loss": 1.0112, + "step": 180310 + }, + { + "epoch": 15.61, + "learning_rate": 3.442172745386815e-05, + "loss": 1.0191, + "step": 180320 + }, + { + "epoch": 15.61, + "learning_rate": 3.4420861127956335e-05, + "loss": 1.0059, + "step": 180330 + }, + { + "epoch": 15.61, + "learning_rate": 3.441999480204453e-05, + "loss": 1.0288, + "step": 180340 + }, + { + "epoch": 15.61, + "learning_rate": 3.4419128476132724e-05, + "loss": 1.0407, + "step": 180350 + }, + { + "epoch": 15.61, + "learning_rate": 3.441826215022091e-05, + "loss": 1.0042, + "step": 180360 + }, + { + "epoch": 15.61, + "learning_rate": 3.4417395824309106e-05, + "loss": 1.0053, + "step": 180370 + }, + { + "epoch": 15.61, + "learning_rate": 3.44165294983973e-05, + "loss": 1.034, + "step": 180380 + }, + { + "epoch": 15.61, + "learning_rate": 3.441566317248549e-05, + "loss": 1.0161, + "step": 180390 + }, + { + "epoch": 15.61, + "learning_rate": 3.441479684657368e-05, + "loss": 1.015, + "step": 180400 + }, + { + "epoch": 15.62, + "learning_rate": 3.441393052066188e-05, + "loss": 1.0596, + "step": 180410 + }, + { + "epoch": 15.62, + "learning_rate": 3.4413064194750065e-05, + "loss": 1.0992, + "step": 180420 + }, + { + "epoch": 15.62, + "learning_rate": 3.441219786883826e-05, + "loss": 1.0141, + "step": 180430 + }, + { + "epoch": 15.62, + "learning_rate": 3.441133154292645e-05, + "loss": 1.05, + "step": 180440 + }, + { + "epoch": 15.62, + "learning_rate": 3.441046521701464e-05, + "loss": 0.9847, + "step": 180450 + }, + { + "epoch": 15.62, + "learning_rate": 3.4409598891102836e-05, + "loss": 1.0592, + "step": 180460 + }, + { + "epoch": 15.62, + "learning_rate": 3.4408732565191024e-05, + "loss": 0.9746, + "step": 180470 + }, + { + "epoch": 15.62, + "learning_rate": 3.440786623927922e-05, + "loss": 1.0411, + "step": 180480 + }, + { + "epoch": 15.62, + "learning_rate": 3.440699991336741e-05, + "loss": 1.0401, + "step": 180490 + }, + { + "epoch": 15.62, + "learning_rate": 3.44061335874556e-05, + "loss": 1.0048, + "step": 180500 + }, + { + "epoch": 15.62, + "learning_rate": 3.4405267261543795e-05, + "loss": 1.003, + "step": 180510 + }, + { + "epoch": 15.62, + "learning_rate": 3.440440093563199e-05, + "loss": 1.0447, + "step": 180520 + }, + { + "epoch": 15.63, + "learning_rate": 3.440353460972018e-05, + "loss": 1.0884, + "step": 180530 + }, + { + "epoch": 15.63, + "learning_rate": 3.440266828380837e-05, + "loss": 1.0148, + "step": 180540 + }, + { + "epoch": 15.63, + "learning_rate": 3.440180195789656e-05, + "loss": 1.0691, + "step": 180550 + }, + { + "epoch": 15.63, + "learning_rate": 3.4400935631984753e-05, + "loss": 1.0353, + "step": 180560 + }, + { + "epoch": 15.63, + "learning_rate": 3.440006930607295e-05, + "loss": 1.0296, + "step": 180570 + }, + { + "epoch": 15.63, + "learning_rate": 3.4399202980161136e-05, + "loss": 1.0016, + "step": 180580 + }, + { + "epoch": 15.63, + "learning_rate": 3.439833665424933e-05, + "loss": 1.0145, + "step": 180590 + }, + { + "epoch": 15.63, + "learning_rate": 3.4397470328337525e-05, + "loss": 1.0647, + "step": 180600 + }, + { + "epoch": 15.63, + "learning_rate": 3.439660400242571e-05, + "loss": 0.9614, + "step": 180610 + }, + { + "epoch": 15.63, + "learning_rate": 3.439573767651391e-05, + "loss": 1.0153, + "step": 180620 + }, + { + "epoch": 15.63, + "learning_rate": 3.43948713506021e-05, + "loss": 1.0806, + "step": 180630 + }, + { + "epoch": 15.64, + "learning_rate": 3.439400502469029e-05, + "loss": 0.9769, + "step": 180640 + }, + { + "epoch": 15.64, + "learning_rate": 3.439313869877848e-05, + "loss": 1.0717, + "step": 180650 + }, + { + "epoch": 15.64, + "learning_rate": 3.439227237286667e-05, + "loss": 1.0963, + "step": 180660 + }, + { + "epoch": 15.64, + "learning_rate": 3.4391406046954865e-05, + "loss": 1.0467, + "step": 180670 + }, + { + "epoch": 15.64, + "learning_rate": 3.439053972104306e-05, + "loss": 1.0545, + "step": 180680 + }, + { + "epoch": 15.64, + "learning_rate": 3.438967339513125e-05, + "loss": 1.0131, + "step": 180690 + }, + { + "epoch": 15.64, + "learning_rate": 3.438880706921944e-05, + "loss": 1.0557, + "step": 180700 + }, + { + "epoch": 15.64, + "learning_rate": 3.4387940743307636e-05, + "loss": 1.0417, + "step": 180710 + }, + { + "epoch": 15.64, + "learning_rate": 3.4387074417395824e-05, + "loss": 1.11, + "step": 180720 + }, + { + "epoch": 15.64, + "learning_rate": 3.438620809148402e-05, + "loss": 1.0862, + "step": 180730 + }, + { + "epoch": 15.64, + "learning_rate": 3.438534176557221e-05, + "loss": 0.9553, + "step": 180740 + }, + { + "epoch": 15.64, + "learning_rate": 3.43844754396604e-05, + "loss": 1.0055, + "step": 180750 + }, + { + "epoch": 15.65, + "learning_rate": 3.4383609113748595e-05, + "loss": 1.0344, + "step": 180760 + }, + { + "epoch": 15.65, + "learning_rate": 3.438274278783678e-05, + "loss": 1.0555, + "step": 180770 + }, + { + "epoch": 15.65, + "learning_rate": 3.438187646192498e-05, + "loss": 1.1037, + "step": 180780 + }, + { + "epoch": 15.65, + "learning_rate": 3.438101013601317e-05, + "loss": 1.0901, + "step": 180790 + }, + { + "epoch": 15.65, + "learning_rate": 3.438014381010136e-05, + "loss": 1.0391, + "step": 180800 + }, + { + "epoch": 15.65, + "learning_rate": 3.4379277484189554e-05, + "loss": 1.0486, + "step": 180810 + }, + { + "epoch": 15.65, + "learning_rate": 3.437841115827775e-05, + "loss": 1.0687, + "step": 180820 + }, + { + "epoch": 15.65, + "learning_rate": 3.4377544832365936e-05, + "loss": 1.087, + "step": 180830 + }, + { + "epoch": 15.65, + "learning_rate": 3.437667850645413e-05, + "loss": 1.0126, + "step": 180840 + }, + { + "epoch": 15.65, + "learning_rate": 3.4375812180542325e-05, + "loss": 1.0866, + "step": 180850 + }, + { + "epoch": 15.65, + "learning_rate": 3.437494585463051e-05, + "loss": 0.9746, + "step": 180860 + }, + { + "epoch": 15.66, + "learning_rate": 3.437407952871871e-05, + "loss": 1.0638, + "step": 180870 + }, + { + "epoch": 15.66, + "learning_rate": 3.4373213202806895e-05, + "loss": 1.0431, + "step": 180880 + }, + { + "epoch": 15.66, + "learning_rate": 3.437234687689509e-05, + "loss": 0.997, + "step": 180890 + }, + { + "epoch": 15.66, + "learning_rate": 3.4371480550983284e-05, + "loss": 0.9917, + "step": 180900 + }, + { + "epoch": 15.66, + "learning_rate": 3.437061422507147e-05, + "loss": 1.0946, + "step": 180910 + }, + { + "epoch": 15.66, + "learning_rate": 3.4369747899159666e-05, + "loss": 1.0444, + "step": 180920 + }, + { + "epoch": 15.66, + "learning_rate": 3.436888157324786e-05, + "loss": 1.0157, + "step": 180930 + }, + { + "epoch": 15.66, + "learning_rate": 3.436801524733605e-05, + "loss": 0.9922, + "step": 180940 + }, + { + "epoch": 15.66, + "learning_rate": 3.436714892142424e-05, + "loss": 1.0516, + "step": 180950 + }, + { + "epoch": 15.66, + "learning_rate": 3.436628259551244e-05, + "loss": 1.0416, + "step": 180960 + }, + { + "epoch": 15.66, + "learning_rate": 3.4365416269600624e-05, + "loss": 1.0739, + "step": 180970 + }, + { + "epoch": 15.66, + "learning_rate": 3.436454994368882e-05, + "loss": 1.0823, + "step": 180980 + }, + { + "epoch": 15.67, + "learning_rate": 3.4363683617777007e-05, + "loss": 1.0341, + "step": 180990 + }, + { + "epoch": 15.67, + "learning_rate": 3.43628172918652e-05, + "loss": 1.0355, + "step": 181000 + }, + { + "epoch": 15.67, + "learning_rate": 3.4361950965953395e-05, + "loss": 1.0662, + "step": 181010 + }, + { + "epoch": 15.67, + "learning_rate": 3.436108464004158e-05, + "loss": 1.0087, + "step": 181020 + }, + { + "epoch": 15.67, + "learning_rate": 3.436021831412978e-05, + "loss": 1.0847, + "step": 181030 + }, + { + "epoch": 15.67, + "learning_rate": 3.435935198821797e-05, + "loss": 1.0159, + "step": 181040 + }, + { + "epoch": 15.67, + "learning_rate": 3.435848566230616e-05, + "loss": 1.049, + "step": 181050 + }, + { + "epoch": 15.67, + "learning_rate": 3.4357619336394354e-05, + "loss": 1.0642, + "step": 181060 + }, + { + "epoch": 15.67, + "learning_rate": 3.435675301048255e-05, + "loss": 1.0276, + "step": 181070 + }, + { + "epoch": 15.67, + "learning_rate": 3.4355886684570736e-05, + "loss": 1.0499, + "step": 181080 + }, + { + "epoch": 15.67, + "learning_rate": 3.435502035865893e-05, + "loss": 1.0438, + "step": 181090 + }, + { + "epoch": 15.68, + "learning_rate": 3.435415403274712e-05, + "loss": 1.056, + "step": 181100 + }, + { + "epoch": 15.68, + "learning_rate": 3.435328770683531e-05, + "loss": 1.0339, + "step": 181110 + }, + { + "epoch": 15.68, + "learning_rate": 3.435242138092351e-05, + "loss": 1.0657, + "step": 181120 + }, + { + "epoch": 15.68, + "learning_rate": 3.4351555055011695e-05, + "loss": 1.0299, + "step": 181130 + }, + { + "epoch": 15.68, + "learning_rate": 3.435068872909989e-05, + "loss": 1.0294, + "step": 181140 + }, + { + "epoch": 15.68, + "learning_rate": 3.4349822403188084e-05, + "loss": 1.1083, + "step": 181150 + }, + { + "epoch": 15.68, + "learning_rate": 3.434895607727627e-05, + "loss": 1.062, + "step": 181160 + }, + { + "epoch": 15.68, + "learning_rate": 3.4348089751364466e-05, + "loss": 1.0827, + "step": 181170 + }, + { + "epoch": 15.68, + "learning_rate": 3.4347223425452654e-05, + "loss": 1.0171, + "step": 181180 + }, + { + "epoch": 15.68, + "learning_rate": 3.434635709954085e-05, + "loss": 1.0079, + "step": 181190 + }, + { + "epoch": 15.68, + "learning_rate": 3.434549077362904e-05, + "loss": 1.0824, + "step": 181200 + }, + { + "epoch": 15.68, + "learning_rate": 3.434462444771723e-05, + "loss": 1.0569, + "step": 181210 + }, + { + "epoch": 15.69, + "learning_rate": 3.4343758121805425e-05, + "loss": 1.0951, + "step": 181220 + }, + { + "epoch": 15.69, + "learning_rate": 3.434289179589362e-05, + "loss": 0.9421, + "step": 181230 + }, + { + "epoch": 15.69, + "learning_rate": 3.434202546998181e-05, + "loss": 1.067, + "step": 181240 + }, + { + "epoch": 15.69, + "learning_rate": 3.434115914407e-05, + "loss": 1.0649, + "step": 181250 + }, + { + "epoch": 15.69, + "learning_rate": 3.4340292818158196e-05, + "loss": 1.0799, + "step": 181260 + }, + { + "epoch": 15.69, + "learning_rate": 3.4339426492246383e-05, + "loss": 1.0103, + "step": 181270 + }, + { + "epoch": 15.69, + "learning_rate": 3.433856016633458e-05, + "loss": 0.9675, + "step": 181280 + }, + { + "epoch": 15.69, + "learning_rate": 3.4337693840422766e-05, + "loss": 1.0538, + "step": 181290 + }, + { + "epoch": 15.69, + "learning_rate": 3.433682751451096e-05, + "loss": 1.0468, + "step": 181300 + }, + { + "epoch": 15.69, + "learning_rate": 3.4335961188599154e-05, + "loss": 1.0604, + "step": 181310 + }, + { + "epoch": 15.69, + "learning_rate": 3.433509486268734e-05, + "loss": 1.0011, + "step": 181320 + }, + { + "epoch": 15.7, + "learning_rate": 3.433422853677554e-05, + "loss": 0.9774, + "step": 181330 + }, + { + "epoch": 15.7, + "learning_rate": 3.433336221086373e-05, + "loss": 1.0597, + "step": 181340 + }, + { + "epoch": 15.7, + "learning_rate": 3.433249588495192e-05, + "loss": 1.0212, + "step": 181350 + }, + { + "epoch": 15.7, + "learning_rate": 3.433162955904011e-05, + "loss": 1.0456, + "step": 181360 + }, + { + "epoch": 15.7, + "learning_rate": 3.433076323312831e-05, + "loss": 1.0391, + "step": 181370 + }, + { + "epoch": 15.7, + "learning_rate": 3.4329896907216495e-05, + "loss": 1.0598, + "step": 181380 + }, + { + "epoch": 15.7, + "learning_rate": 3.432903058130469e-05, + "loss": 1.059, + "step": 181390 + }, + { + "epoch": 15.7, + "learning_rate": 3.432816425539288e-05, + "loss": 1.062, + "step": 181400 + }, + { + "epoch": 15.7, + "learning_rate": 3.432729792948107e-05, + "loss": 1.1162, + "step": 181410 + }, + { + "epoch": 15.7, + "learning_rate": 3.4326431603569266e-05, + "loss": 1.0506, + "step": 181420 + }, + { + "epoch": 15.7, + "learning_rate": 3.4325565277657454e-05, + "loss": 1.0616, + "step": 181430 + }, + { + "epoch": 15.7, + "learning_rate": 3.432469895174565e-05, + "loss": 0.9903, + "step": 181440 + }, + { + "epoch": 15.71, + "learning_rate": 3.432383262583384e-05, + "loss": 1.0469, + "step": 181450 + }, + { + "epoch": 15.71, + "learning_rate": 3.432296629992203e-05, + "loss": 1.0113, + "step": 181460 + }, + { + "epoch": 15.71, + "learning_rate": 3.4322099974010225e-05, + "loss": 1.089, + "step": 181470 + }, + { + "epoch": 15.71, + "learning_rate": 3.432123364809842e-05, + "loss": 1.0401, + "step": 181480 + }, + { + "epoch": 15.71, + "learning_rate": 3.432036732218661e-05, + "loss": 1.0493, + "step": 181490 + }, + { + "epoch": 15.71, + "learning_rate": 3.43195009962748e-05, + "loss": 1.0838, + "step": 181500 + }, + { + "epoch": 15.71, + "learning_rate": 3.431863467036299e-05, + "loss": 1.0295, + "step": 181510 + }, + { + "epoch": 15.71, + "learning_rate": 3.4317768344451184e-05, + "loss": 1.0751, + "step": 181520 + }, + { + "epoch": 15.71, + "learning_rate": 3.431690201853938e-05, + "loss": 1.0725, + "step": 181530 + }, + { + "epoch": 15.71, + "learning_rate": 3.4316035692627566e-05, + "loss": 1.0283, + "step": 181540 + }, + { + "epoch": 15.71, + "learning_rate": 3.431516936671576e-05, + "loss": 1.0213, + "step": 181550 + }, + { + "epoch": 15.71, + "learning_rate": 3.4314303040803955e-05, + "loss": 1.02, + "step": 181560 + }, + { + "epoch": 15.72, + "learning_rate": 3.431343671489214e-05, + "loss": 1.0697, + "step": 181570 + }, + { + "epoch": 15.72, + "learning_rate": 3.431257038898034e-05, + "loss": 1.0101, + "step": 181580 + }, + { + "epoch": 15.72, + "learning_rate": 3.431170406306853e-05, + "loss": 1.0092, + "step": 181590 + }, + { + "epoch": 15.72, + "learning_rate": 3.431083773715672e-05, + "loss": 0.9768, + "step": 181600 + }, + { + "epoch": 15.72, + "learning_rate": 3.4309971411244914e-05, + "loss": 1.0274, + "step": 181610 + }, + { + "epoch": 15.72, + "learning_rate": 3.43091050853331e-05, + "loss": 1.0373, + "step": 181620 + }, + { + "epoch": 15.72, + "learning_rate": 3.4308238759421296e-05, + "loss": 1.0114, + "step": 181630 + }, + { + "epoch": 15.72, + "learning_rate": 3.430737243350949e-05, + "loss": 1.0169, + "step": 181640 + }, + { + "epoch": 15.72, + "learning_rate": 3.430650610759768e-05, + "loss": 1.0074, + "step": 181650 + }, + { + "epoch": 15.72, + "learning_rate": 3.430563978168587e-05, + "loss": 1.0072, + "step": 181660 + }, + { + "epoch": 15.72, + "learning_rate": 3.430477345577407e-05, + "loss": 1.0725, + "step": 181670 + }, + { + "epoch": 15.73, + "learning_rate": 3.4303907129862254e-05, + "loss": 1.0486, + "step": 181680 + }, + { + "epoch": 15.73, + "learning_rate": 3.430304080395045e-05, + "loss": 1.057, + "step": 181690 + }, + { + "epoch": 15.73, + "learning_rate": 3.430217447803864e-05, + "loss": 1.0266, + "step": 181700 + }, + { + "epoch": 15.73, + "learning_rate": 3.430130815212683e-05, + "loss": 1.004, + "step": 181710 + }, + { + "epoch": 15.73, + "learning_rate": 3.4300441826215025e-05, + "loss": 1.0432, + "step": 181720 + }, + { + "epoch": 15.73, + "learning_rate": 3.429957550030321e-05, + "loss": 1.0328, + "step": 181730 + }, + { + "epoch": 15.73, + "learning_rate": 3.429870917439141e-05, + "loss": 1.0198, + "step": 181740 + }, + { + "epoch": 15.73, + "learning_rate": 3.42978428484796e-05, + "loss": 1.0489, + "step": 181750 + }, + { + "epoch": 15.73, + "learning_rate": 3.429697652256779e-05, + "loss": 1.0132, + "step": 181760 + }, + { + "epoch": 15.73, + "learning_rate": 3.4296110196655984e-05, + "loss": 1.0397, + "step": 181770 + }, + { + "epoch": 15.73, + "learning_rate": 3.429524387074418e-05, + "loss": 1.033, + "step": 181780 + }, + { + "epoch": 15.73, + "learning_rate": 3.4294377544832366e-05, + "loss": 1.0569, + "step": 181790 + }, + { + "epoch": 15.74, + "learning_rate": 3.429351121892056e-05, + "loss": 1.075, + "step": 181800 + }, + { + "epoch": 15.74, + "learning_rate": 3.4292644893008755e-05, + "loss": 1.0587, + "step": 181810 + }, + { + "epoch": 15.74, + "learning_rate": 3.429177856709694e-05, + "loss": 0.9805, + "step": 181820 + }, + { + "epoch": 15.74, + "learning_rate": 3.429091224118514e-05, + "loss": 1.0312, + "step": 181830 + }, + { + "epoch": 15.74, + "learning_rate": 3.4290045915273325e-05, + "loss": 1.0476, + "step": 181840 + }, + { + "epoch": 15.74, + "learning_rate": 3.428917958936152e-05, + "loss": 1.0743, + "step": 181850 + }, + { + "epoch": 15.74, + "learning_rate": 3.4288313263449714e-05, + "loss": 1.0343, + "step": 181860 + }, + { + "epoch": 15.74, + "learning_rate": 3.42874469375379e-05, + "loss": 1.0796, + "step": 181870 + }, + { + "epoch": 15.74, + "learning_rate": 3.4286580611626096e-05, + "loss": 1.0745, + "step": 181880 + }, + { + "epoch": 15.74, + "learning_rate": 3.428571428571429e-05, + "loss": 1.0085, + "step": 181890 + }, + { + "epoch": 15.74, + "learning_rate": 3.428484795980248e-05, + "loss": 1.0058, + "step": 181900 + }, + { + "epoch": 15.75, + "learning_rate": 3.428398163389067e-05, + "loss": 1.0153, + "step": 181910 + }, + { + "epoch": 15.75, + "learning_rate": 3.428311530797886e-05, + "loss": 1.077, + "step": 181920 + }, + { + "epoch": 15.75, + "learning_rate": 3.4282248982067055e-05, + "loss": 1.0689, + "step": 181930 + }, + { + "epoch": 15.75, + "learning_rate": 3.428138265615525e-05, + "loss": 1.115, + "step": 181940 + }, + { + "epoch": 15.75, + "learning_rate": 3.428051633024344e-05, + "loss": 1.0723, + "step": 181950 + }, + { + "epoch": 15.75, + "learning_rate": 3.427965000433163e-05, + "loss": 1.0725, + "step": 181960 + }, + { + "epoch": 15.75, + "learning_rate": 3.4278783678419826e-05, + "loss": 1.0249, + "step": 181970 + }, + { + "epoch": 15.75, + "learning_rate": 3.4277917352508013e-05, + "loss": 0.992, + "step": 181980 + }, + { + "epoch": 15.75, + "learning_rate": 3.427705102659621e-05, + "loss": 1.0109, + "step": 181990 + }, + { + "epoch": 15.75, + "learning_rate": 3.42761847006844e-05, + "loss": 1.0669, + "step": 182000 + }, + { + "epoch": 15.75, + "learning_rate": 3.427531837477259e-05, + "loss": 1.0903, + "step": 182010 + }, + { + "epoch": 15.75, + "learning_rate": 3.4274452048860784e-05, + "loss": 1.0484, + "step": 182020 + }, + { + "epoch": 15.76, + "learning_rate": 3.427358572294897e-05, + "loss": 1.0302, + "step": 182030 + }, + { + "epoch": 15.76, + "learning_rate": 3.4272719397037167e-05, + "loss": 1.0652, + "step": 182040 + }, + { + "epoch": 15.76, + "learning_rate": 3.427185307112536e-05, + "loss": 0.9954, + "step": 182050 + }, + { + "epoch": 15.76, + "learning_rate": 3.427098674521355e-05, + "loss": 1.0609, + "step": 182060 + }, + { + "epoch": 15.76, + "learning_rate": 3.427012041930174e-05, + "loss": 1.0476, + "step": 182070 + }, + { + "epoch": 15.76, + "learning_rate": 3.426925409338994e-05, + "loss": 1.0591, + "step": 182080 + }, + { + "epoch": 15.76, + "learning_rate": 3.4268387767478125e-05, + "loss": 1.0441, + "step": 182090 + }, + { + "epoch": 15.76, + "learning_rate": 3.426752144156632e-05, + "loss": 1.1235, + "step": 182100 + }, + { + "epoch": 15.76, + "learning_rate": 3.4266655115654514e-05, + "loss": 1.1295, + "step": 182110 + }, + { + "epoch": 15.76, + "learning_rate": 3.42657887897427e-05, + "loss": 1.0444, + "step": 182120 + }, + { + "epoch": 15.76, + "learning_rate": 3.4264922463830896e-05, + "loss": 1.0574, + "step": 182130 + }, + { + "epoch": 15.77, + "learning_rate": 3.4264056137919084e-05, + "loss": 1.0327, + "step": 182140 + }, + { + "epoch": 15.77, + "learning_rate": 3.426318981200728e-05, + "loss": 1.0365, + "step": 182150 + }, + { + "epoch": 15.77, + "learning_rate": 3.426232348609547e-05, + "loss": 1.06, + "step": 182160 + }, + { + "epoch": 15.77, + "learning_rate": 3.426145716018366e-05, + "loss": 1.0262, + "step": 182170 + }, + { + "epoch": 15.77, + "learning_rate": 3.4260590834271855e-05, + "loss": 1.0355, + "step": 182180 + }, + { + "epoch": 15.77, + "learning_rate": 3.425972450836005e-05, + "loss": 1.0591, + "step": 182190 + }, + { + "epoch": 15.77, + "learning_rate": 3.425885818244824e-05, + "loss": 1.0653, + "step": 182200 + }, + { + "epoch": 15.77, + "learning_rate": 3.425799185653643e-05, + "loss": 1.0537, + "step": 182210 + }, + { + "epoch": 15.77, + "learning_rate": 3.4257125530624626e-05, + "loss": 1.0492, + "step": 182220 + }, + { + "epoch": 15.77, + "learning_rate": 3.4256259204712814e-05, + "loss": 1.0623, + "step": 182230 + }, + { + "epoch": 15.77, + "learning_rate": 3.425539287880101e-05, + "loss": 0.9851, + "step": 182240 + }, + { + "epoch": 15.77, + "learning_rate": 3.4254526552889196e-05, + "loss": 1.0185, + "step": 182250 + }, + { + "epoch": 15.78, + "learning_rate": 3.425366022697739e-05, + "loss": 1.0425, + "step": 182260 + }, + { + "epoch": 15.78, + "learning_rate": 3.4252793901065585e-05, + "loss": 1.0651, + "step": 182270 + }, + { + "epoch": 15.78, + "learning_rate": 3.425192757515377e-05, + "loss": 1.0715, + "step": 182280 + }, + { + "epoch": 15.78, + "learning_rate": 3.425106124924197e-05, + "loss": 1.0507, + "step": 182290 + }, + { + "epoch": 15.78, + "learning_rate": 3.425019492333016e-05, + "loss": 1.0929, + "step": 182300 + }, + { + "epoch": 15.78, + "learning_rate": 3.424932859741835e-05, + "loss": 0.9991, + "step": 182310 + }, + { + "epoch": 15.78, + "learning_rate": 3.4248462271506544e-05, + "loss": 1.0439, + "step": 182320 + }, + { + "epoch": 15.78, + "learning_rate": 3.424759594559474e-05, + "loss": 0.9857, + "step": 182330 + }, + { + "epoch": 15.78, + "learning_rate": 3.4246729619682926e-05, + "loss": 1.04, + "step": 182340 + }, + { + "epoch": 15.78, + "learning_rate": 3.424586329377112e-05, + "loss": 1.0024, + "step": 182350 + }, + { + "epoch": 15.78, + "learning_rate": 3.424499696785931e-05, + "loss": 1.0713, + "step": 182360 + }, + { + "epoch": 15.79, + "learning_rate": 3.42441306419475e-05, + "loss": 1.0308, + "step": 182370 + }, + { + "epoch": 15.79, + "learning_rate": 3.42432643160357e-05, + "loss": 1.0429, + "step": 182380 + }, + { + "epoch": 15.79, + "learning_rate": 3.4242397990123884e-05, + "loss": 1.0352, + "step": 182390 + }, + { + "epoch": 15.79, + "learning_rate": 3.424153166421208e-05, + "loss": 1.086, + "step": 182400 + }, + { + "epoch": 15.79, + "learning_rate": 3.424066533830027e-05, + "loss": 1.0264, + "step": 182410 + }, + { + "epoch": 15.79, + "learning_rate": 3.423979901238846e-05, + "loss": 1.0416, + "step": 182420 + }, + { + "epoch": 15.79, + "learning_rate": 3.4238932686476655e-05, + "loss": 1.046, + "step": 182430 + }, + { + "epoch": 15.79, + "learning_rate": 3.423806636056485e-05, + "loss": 1.0956, + "step": 182440 + }, + { + "epoch": 15.79, + "learning_rate": 3.423720003465304e-05, + "loss": 1.0462, + "step": 182450 + }, + { + "epoch": 15.79, + "learning_rate": 3.423633370874123e-05, + "loss": 1.032, + "step": 182460 + }, + { + "epoch": 15.79, + "learning_rate": 3.423546738282942e-05, + "loss": 1.0135, + "step": 182470 + }, + { + "epoch": 15.79, + "learning_rate": 3.4234601056917614e-05, + "loss": 1.0675, + "step": 182480 + }, + { + "epoch": 15.8, + "learning_rate": 3.423373473100581e-05, + "loss": 1.0357, + "step": 182490 + }, + { + "epoch": 15.8, + "learning_rate": 3.4232868405093996e-05, + "loss": 1.0626, + "step": 182500 + }, + { + "epoch": 15.8, + "learning_rate": 3.423200207918219e-05, + "loss": 1.0123, + "step": 182510 + }, + { + "epoch": 15.8, + "learning_rate": 3.4231135753270385e-05, + "loss": 1.0227, + "step": 182520 + }, + { + "epoch": 15.8, + "learning_rate": 3.423026942735857e-05, + "loss": 1.107, + "step": 182530 + }, + { + "epoch": 15.8, + "learning_rate": 3.422940310144677e-05, + "loss": 1.0663, + "step": 182540 + }, + { + "epoch": 15.8, + "learning_rate": 3.422853677553496e-05, + "loss": 0.9415, + "step": 182550 + }, + { + "epoch": 15.8, + "learning_rate": 3.422767044962315e-05, + "loss": 1.0629, + "step": 182560 + }, + { + "epoch": 15.8, + "learning_rate": 3.4226804123711344e-05, + "loss": 1.0139, + "step": 182570 + }, + { + "epoch": 15.8, + "learning_rate": 3.422593779779953e-05, + "loss": 1.0908, + "step": 182580 + }, + { + "epoch": 15.8, + "learning_rate": 3.4225071471887726e-05, + "loss": 1.0897, + "step": 182590 + }, + { + "epoch": 15.8, + "learning_rate": 3.422420514597592e-05, + "loss": 0.9656, + "step": 182600 + }, + { + "epoch": 15.81, + "learning_rate": 3.422333882006411e-05, + "loss": 1.0381, + "step": 182610 + }, + { + "epoch": 15.81, + "learning_rate": 3.42224724941523e-05, + "loss": 1.0745, + "step": 182620 + }, + { + "epoch": 15.81, + "learning_rate": 3.42216061682405e-05, + "loss": 1.0642, + "step": 182630 + }, + { + "epoch": 15.81, + "learning_rate": 3.4220739842328685e-05, + "loss": 1.0565, + "step": 182640 + }, + { + "epoch": 15.81, + "learning_rate": 3.421987351641688e-05, + "loss": 1.0348, + "step": 182650 + }, + { + "epoch": 15.81, + "learning_rate": 3.421900719050507e-05, + "loss": 1.0524, + "step": 182660 + }, + { + "epoch": 15.81, + "learning_rate": 3.421814086459326e-05, + "loss": 0.9698, + "step": 182670 + }, + { + "epoch": 15.81, + "learning_rate": 3.4217274538681456e-05, + "loss": 1.0308, + "step": 182680 + }, + { + "epoch": 15.81, + "learning_rate": 3.4216408212769643e-05, + "loss": 1.0081, + "step": 182690 + }, + { + "epoch": 15.81, + "learning_rate": 3.421554188685784e-05, + "loss": 1.0089, + "step": 182700 + }, + { + "epoch": 15.81, + "learning_rate": 3.421467556094603e-05, + "loss": 1.0486, + "step": 182710 + }, + { + "epoch": 15.82, + "learning_rate": 3.421380923503422e-05, + "loss": 1.0132, + "step": 182720 + }, + { + "epoch": 15.82, + "learning_rate": 3.4212942909122414e-05, + "loss": 1.0764, + "step": 182730 + }, + { + "epoch": 15.82, + "learning_rate": 3.421207658321061e-05, + "loss": 1.0406, + "step": 182740 + }, + { + "epoch": 15.82, + "learning_rate": 3.4211210257298797e-05, + "loss": 1.0586, + "step": 182750 + }, + { + "epoch": 15.82, + "learning_rate": 3.421034393138699e-05, + "loss": 1.0, + "step": 182760 + }, + { + "epoch": 15.82, + "learning_rate": 3.420947760547518e-05, + "loss": 1.0183, + "step": 182770 + }, + { + "epoch": 15.82, + "learning_rate": 3.420861127956337e-05, + "loss": 1.0478, + "step": 182780 + }, + { + "epoch": 15.82, + "learning_rate": 3.420774495365157e-05, + "loss": 1.0323, + "step": 182790 + }, + { + "epoch": 15.82, + "learning_rate": 3.4206878627739755e-05, + "loss": 1.0356, + "step": 182800 + }, + { + "epoch": 15.82, + "learning_rate": 3.420601230182795e-05, + "loss": 1.0606, + "step": 182810 + }, + { + "epoch": 15.82, + "learning_rate": 3.4205145975916144e-05, + "loss": 1.0105, + "step": 182820 + }, + { + "epoch": 15.82, + "learning_rate": 3.420427965000433e-05, + "loss": 1.0832, + "step": 182830 + }, + { + "epoch": 15.83, + "learning_rate": 3.4203413324092526e-05, + "loss": 1.0388, + "step": 182840 + }, + { + "epoch": 15.83, + "learning_rate": 3.420254699818072e-05, + "loss": 1.0597, + "step": 182850 + }, + { + "epoch": 15.83, + "learning_rate": 3.420168067226891e-05, + "loss": 1.0815, + "step": 182860 + }, + { + "epoch": 15.83, + "learning_rate": 3.42008143463571e-05, + "loss": 1.0539, + "step": 182870 + }, + { + "epoch": 15.83, + "learning_rate": 3.419994802044529e-05, + "loss": 1.0374, + "step": 182880 + }, + { + "epoch": 15.83, + "learning_rate": 3.4199081694533485e-05, + "loss": 1.0395, + "step": 182890 + }, + { + "epoch": 15.83, + "learning_rate": 3.419821536862168e-05, + "loss": 1.0071, + "step": 182900 + }, + { + "epoch": 15.83, + "learning_rate": 3.419734904270987e-05, + "loss": 1.043, + "step": 182910 + }, + { + "epoch": 15.83, + "learning_rate": 3.419648271679806e-05, + "loss": 1.0451, + "step": 182920 + }, + { + "epoch": 15.83, + "learning_rate": 3.4195616390886256e-05, + "loss": 0.9769, + "step": 182930 + }, + { + "epoch": 15.83, + "learning_rate": 3.4194750064974444e-05, + "loss": 1.075, + "step": 182940 + }, + { + "epoch": 15.84, + "learning_rate": 3.419388373906264e-05, + "loss": 1.0455, + "step": 182950 + }, + { + "epoch": 15.84, + "learning_rate": 3.419301741315083e-05, + "loss": 1.0401, + "step": 182960 + }, + { + "epoch": 15.84, + "learning_rate": 3.419215108723902e-05, + "loss": 1.0458, + "step": 182970 + }, + { + "epoch": 15.84, + "learning_rate": 3.4191284761327215e-05, + "loss": 1.0412, + "step": 182980 + }, + { + "epoch": 15.84, + "learning_rate": 3.41904184354154e-05, + "loss": 1.0202, + "step": 182990 + }, + { + "epoch": 15.84, + "learning_rate": 3.41895521095036e-05, + "loss": 1.1015, + "step": 183000 + }, + { + "epoch": 15.84, + "learning_rate": 3.418868578359179e-05, + "loss": 0.9937, + "step": 183010 + }, + { + "epoch": 15.84, + "learning_rate": 3.418781945767998e-05, + "loss": 1.041, + "step": 183020 + }, + { + "epoch": 15.84, + "learning_rate": 3.4186953131768173e-05, + "loss": 1.016, + "step": 183030 + }, + { + "epoch": 15.84, + "learning_rate": 3.418608680585637e-05, + "loss": 1.0552, + "step": 183040 + }, + { + "epoch": 15.84, + "learning_rate": 3.4185220479944556e-05, + "loss": 1.0453, + "step": 183050 + }, + { + "epoch": 15.84, + "learning_rate": 3.418435415403275e-05, + "loss": 1.0559, + "step": 183060 + }, + { + "epoch": 15.85, + "learning_rate": 3.4183487828120945e-05, + "loss": 1.0261, + "step": 183070 + }, + { + "epoch": 15.85, + "learning_rate": 3.418262150220913e-05, + "loss": 0.9978, + "step": 183080 + }, + { + "epoch": 15.85, + "learning_rate": 3.418175517629733e-05, + "loss": 1.0531, + "step": 183090 + }, + { + "epoch": 15.85, + "learning_rate": 3.4180888850385514e-05, + "loss": 1.0203, + "step": 183100 + }, + { + "epoch": 15.85, + "learning_rate": 3.418002252447371e-05, + "loss": 1.1257, + "step": 183110 + }, + { + "epoch": 15.85, + "learning_rate": 3.41791561985619e-05, + "loss": 1.0471, + "step": 183120 + }, + { + "epoch": 15.85, + "learning_rate": 3.417828987265009e-05, + "loss": 1.0691, + "step": 183130 + }, + { + "epoch": 15.85, + "learning_rate": 3.4177423546738285e-05, + "loss": 1.0421, + "step": 183140 + }, + { + "epoch": 15.85, + "learning_rate": 3.417655722082648e-05, + "loss": 1.0071, + "step": 183150 + }, + { + "epoch": 15.85, + "learning_rate": 3.417569089491467e-05, + "loss": 0.987, + "step": 183160 + }, + { + "epoch": 15.85, + "learning_rate": 3.417482456900286e-05, + "loss": 1.0337, + "step": 183170 + }, + { + "epoch": 15.86, + "learning_rate": 3.4173958243091056e-05, + "loss": 1.0406, + "step": 183180 + }, + { + "epoch": 15.86, + "learning_rate": 3.4173091917179244e-05, + "loss": 1.09, + "step": 183190 + }, + { + "epoch": 15.86, + "learning_rate": 3.417222559126744e-05, + "loss": 1.0399, + "step": 183200 + }, + { + "epoch": 15.86, + "learning_rate": 3.4171359265355626e-05, + "loss": 1.0274, + "step": 183210 + }, + { + "epoch": 15.86, + "learning_rate": 3.417049293944382e-05, + "loss": 1.0585, + "step": 183220 + }, + { + "epoch": 15.86, + "learning_rate": 3.4169626613532015e-05, + "loss": 0.9714, + "step": 183230 + }, + { + "epoch": 15.86, + "learning_rate": 3.41687602876202e-05, + "loss": 1.0494, + "step": 183240 + }, + { + "epoch": 15.86, + "learning_rate": 3.41678939617084e-05, + "loss": 1.035, + "step": 183250 + }, + { + "epoch": 15.86, + "learning_rate": 3.416702763579659e-05, + "loss": 1.0187, + "step": 183260 + }, + { + "epoch": 15.86, + "learning_rate": 3.416616130988478e-05, + "loss": 1.009, + "step": 183270 + }, + { + "epoch": 15.86, + "learning_rate": 3.4165294983972974e-05, + "loss": 1.026, + "step": 183280 + }, + { + "epoch": 15.86, + "learning_rate": 3.416442865806117e-05, + "loss": 0.9955, + "step": 183290 + }, + { + "epoch": 15.87, + "learning_rate": 3.4163562332149356e-05, + "loss": 1.0792, + "step": 183300 + }, + { + "epoch": 15.87, + "learning_rate": 3.416269600623755e-05, + "loss": 1.0756, + "step": 183310 + }, + { + "epoch": 15.87, + "learning_rate": 3.416182968032574e-05, + "loss": 1.0564, + "step": 183320 + }, + { + "epoch": 15.87, + "learning_rate": 3.416096335441393e-05, + "loss": 1.058, + "step": 183330 + }, + { + "epoch": 15.87, + "learning_rate": 3.416009702850213e-05, + "loss": 1.0723, + "step": 183340 + }, + { + "epoch": 15.87, + "learning_rate": 3.4159230702590315e-05, + "loss": 1.0556, + "step": 183350 + }, + { + "epoch": 15.87, + "learning_rate": 3.415836437667851e-05, + "loss": 1.0922, + "step": 183360 + }, + { + "epoch": 15.87, + "learning_rate": 3.4157498050766704e-05, + "loss": 1.0066, + "step": 183370 + }, + { + "epoch": 15.87, + "learning_rate": 3.415663172485489e-05, + "loss": 1.0517, + "step": 183380 + }, + { + "epoch": 15.87, + "learning_rate": 3.4155765398943086e-05, + "loss": 1.024, + "step": 183390 + }, + { + "epoch": 15.87, + "learning_rate": 3.415489907303128e-05, + "loss": 1.0188, + "step": 183400 + }, + { + "epoch": 15.88, + "learning_rate": 3.415403274711947e-05, + "loss": 1.0593, + "step": 183410 + }, + { + "epoch": 15.88, + "learning_rate": 3.415316642120766e-05, + "loss": 1.0356, + "step": 183420 + }, + { + "epoch": 15.88, + "learning_rate": 3.415230009529585e-05, + "loss": 1.0293, + "step": 183430 + }, + { + "epoch": 15.88, + "learning_rate": 3.4151433769384044e-05, + "loss": 1.0518, + "step": 183440 + }, + { + "epoch": 15.88, + "learning_rate": 3.415056744347224e-05, + "loss": 0.9951, + "step": 183450 + }, + { + "epoch": 15.88, + "learning_rate": 3.4149701117560427e-05, + "loss": 1.0799, + "step": 183460 + }, + { + "epoch": 15.88, + "learning_rate": 3.414883479164862e-05, + "loss": 1.0402, + "step": 183470 + }, + { + "epoch": 15.88, + "learning_rate": 3.4147968465736815e-05, + "loss": 1.0746, + "step": 183480 + }, + { + "epoch": 15.88, + "learning_rate": 3.4147102139825e-05, + "loss": 1.0103, + "step": 183490 + }, + { + "epoch": 15.88, + "learning_rate": 3.41462358139132e-05, + "loss": 1.1337, + "step": 183500 + }, + { + "epoch": 15.88, + "learning_rate": 3.4145369488001385e-05, + "loss": 1.0282, + "step": 183510 + }, + { + "epoch": 15.88, + "learning_rate": 3.414450316208958e-05, + "loss": 1.0846, + "step": 183520 + }, + { + "epoch": 15.89, + "learning_rate": 3.4143636836177774e-05, + "loss": 0.9766, + "step": 183530 + }, + { + "epoch": 15.89, + "learning_rate": 3.414277051026596e-05, + "loss": 1.0975, + "step": 183540 + }, + { + "epoch": 15.89, + "learning_rate": 3.4141904184354156e-05, + "loss": 1.0257, + "step": 183550 + }, + { + "epoch": 15.89, + "learning_rate": 3.414103785844235e-05, + "loss": 1.0607, + "step": 183560 + }, + { + "epoch": 15.89, + "learning_rate": 3.414017153253054e-05, + "loss": 1.0179, + "step": 183570 + }, + { + "epoch": 15.89, + "learning_rate": 3.413930520661873e-05, + "loss": 1.0355, + "step": 183580 + }, + { + "epoch": 15.89, + "learning_rate": 3.413843888070693e-05, + "loss": 1.0586, + "step": 183590 + }, + { + "epoch": 15.89, + "learning_rate": 3.4137572554795115e-05, + "loss": 0.9659, + "step": 183600 + }, + { + "epoch": 15.89, + "learning_rate": 3.413670622888331e-05, + "loss": 1.0486, + "step": 183610 + }, + { + "epoch": 15.89, + "learning_rate": 3.41358399029715e-05, + "loss": 1.0435, + "step": 183620 + }, + { + "epoch": 15.89, + "learning_rate": 3.413497357705969e-05, + "loss": 1.0044, + "step": 183630 + }, + { + "epoch": 15.9, + "learning_rate": 3.4134107251147886e-05, + "loss": 1.076, + "step": 183640 + }, + { + "epoch": 15.9, + "learning_rate": 3.4133240925236074e-05, + "loss": 1.0117, + "step": 183650 + }, + { + "epoch": 15.9, + "learning_rate": 3.413237459932427e-05, + "loss": 1.0114, + "step": 183660 + }, + { + "epoch": 15.9, + "learning_rate": 3.413150827341246e-05, + "loss": 1.024, + "step": 183670 + }, + { + "epoch": 15.9, + "learning_rate": 3.413064194750065e-05, + "loss": 1.0833, + "step": 183680 + }, + { + "epoch": 15.9, + "learning_rate": 3.4129775621588845e-05, + "loss": 1.0684, + "step": 183690 + }, + { + "epoch": 15.9, + "learning_rate": 3.412890929567704e-05, + "loss": 1.0649, + "step": 183700 + }, + { + "epoch": 15.9, + "learning_rate": 3.412804296976523e-05, + "loss": 0.9816, + "step": 183710 + }, + { + "epoch": 15.9, + "learning_rate": 3.412717664385342e-05, + "loss": 1.0032, + "step": 183720 + }, + { + "epoch": 15.9, + "learning_rate": 3.412631031794161e-05, + "loss": 1.001, + "step": 183730 + }, + { + "epoch": 15.9, + "learning_rate": 3.4125443992029803e-05, + "loss": 1.059, + "step": 183740 + }, + { + "epoch": 15.9, + "learning_rate": 3.4124577666118e-05, + "loss": 1.0623, + "step": 183750 + }, + { + "epoch": 15.91, + "learning_rate": 3.4123711340206186e-05, + "loss": 1.0808, + "step": 183760 + }, + { + "epoch": 15.91, + "learning_rate": 3.412284501429438e-05, + "loss": 1.0979, + "step": 183770 + }, + { + "epoch": 15.91, + "learning_rate": 3.4121978688382574e-05, + "loss": 1.0752, + "step": 183780 + }, + { + "epoch": 15.91, + "learning_rate": 3.412111236247076e-05, + "loss": 1.0239, + "step": 183790 + }, + { + "epoch": 15.91, + "learning_rate": 3.412024603655896e-05, + "loss": 1.1025, + "step": 183800 + }, + { + "epoch": 15.91, + "learning_rate": 3.411937971064715e-05, + "loss": 1.0171, + "step": 183810 + }, + { + "epoch": 15.91, + "learning_rate": 3.411851338473534e-05, + "loss": 1.0728, + "step": 183820 + }, + { + "epoch": 15.91, + "learning_rate": 3.411764705882353e-05, + "loss": 1.0095, + "step": 183830 + }, + { + "epoch": 15.91, + "learning_rate": 3.411678073291172e-05, + "loss": 1.0248, + "step": 183840 + }, + { + "epoch": 15.91, + "learning_rate": 3.4115914406999915e-05, + "loss": 1.0377, + "step": 183850 + }, + { + "epoch": 15.91, + "learning_rate": 3.411504808108811e-05, + "loss": 1.111, + "step": 183860 + }, + { + "epoch": 15.91, + "learning_rate": 3.41141817551763e-05, + "loss": 1.0837, + "step": 183870 + }, + { + "epoch": 15.92, + "learning_rate": 3.411331542926449e-05, + "loss": 1.0931, + "step": 183880 + }, + { + "epoch": 15.92, + "learning_rate": 3.4112449103352686e-05, + "loss": 1.043, + "step": 183890 + }, + { + "epoch": 15.92, + "learning_rate": 3.4111582777440874e-05, + "loss": 1.007, + "step": 183900 + }, + { + "epoch": 15.92, + "learning_rate": 3.411071645152907e-05, + "loss": 0.9931, + "step": 183910 + }, + { + "epoch": 15.92, + "learning_rate": 3.410985012561726e-05, + "loss": 1.0204, + "step": 183920 + }, + { + "epoch": 15.92, + "learning_rate": 3.410898379970545e-05, + "loss": 0.9939, + "step": 183930 + }, + { + "epoch": 15.92, + "learning_rate": 3.4108117473793645e-05, + "loss": 1.0175, + "step": 183940 + }, + { + "epoch": 15.92, + "learning_rate": 3.410725114788183e-05, + "loss": 1.0985, + "step": 183950 + }, + { + "epoch": 15.92, + "learning_rate": 3.410638482197003e-05, + "loss": 1.023, + "step": 183960 + }, + { + "epoch": 15.92, + "learning_rate": 3.410551849605822e-05, + "loss": 1.026, + "step": 183970 + }, + { + "epoch": 15.92, + "learning_rate": 3.410465217014641e-05, + "loss": 1.0018, + "step": 183980 + }, + { + "epoch": 15.93, + "learning_rate": 3.4103785844234604e-05, + "loss": 1.0224, + "step": 183990 + }, + { + "epoch": 15.93, + "learning_rate": 3.41029195183228e-05, + "loss": 1.0386, + "step": 184000 + }, + { + "epoch": 15.93, + "learning_rate": 3.4102053192410986e-05, + "loss": 1.0419, + "step": 184010 + }, + { + "epoch": 15.93, + "learning_rate": 3.410118686649918e-05, + "loss": 1.0604, + "step": 184020 + }, + { + "epoch": 15.93, + "learning_rate": 3.4100320540587375e-05, + "loss": 1.0217, + "step": 184030 + }, + { + "epoch": 15.93, + "learning_rate": 3.409945421467556e-05, + "loss": 1.0037, + "step": 184040 + }, + { + "epoch": 15.93, + "learning_rate": 3.409858788876376e-05, + "loss": 1.0895, + "step": 184050 + }, + { + "epoch": 15.93, + "learning_rate": 3.4097721562851945e-05, + "loss": 0.9939, + "step": 184060 + }, + { + "epoch": 15.93, + "learning_rate": 3.409685523694014e-05, + "loss": 1.0536, + "step": 184070 + }, + { + "epoch": 15.93, + "learning_rate": 3.4095988911028334e-05, + "loss": 1.0698, + "step": 184080 + }, + { + "epoch": 15.93, + "learning_rate": 3.409512258511652e-05, + "loss": 1.0692, + "step": 184090 + }, + { + "epoch": 15.93, + "learning_rate": 3.4094256259204716e-05, + "loss": 1.027, + "step": 184100 + }, + { + "epoch": 15.94, + "learning_rate": 3.409338993329291e-05, + "loss": 1.0188, + "step": 184110 + }, + { + "epoch": 15.94, + "learning_rate": 3.40925236073811e-05, + "loss": 1.0551, + "step": 184120 + }, + { + "epoch": 15.94, + "learning_rate": 3.409165728146929e-05, + "loss": 1.087, + "step": 184130 + }, + { + "epoch": 15.94, + "learning_rate": 3.409079095555749e-05, + "loss": 1.077, + "step": 184140 + }, + { + "epoch": 15.94, + "learning_rate": 3.4089924629645674e-05, + "loss": 1.0382, + "step": 184150 + }, + { + "epoch": 15.94, + "learning_rate": 3.408905830373387e-05, + "loss": 1.0782, + "step": 184160 + }, + { + "epoch": 15.94, + "learning_rate": 3.4088191977822057e-05, + "loss": 0.9901, + "step": 184170 + }, + { + "epoch": 15.94, + "learning_rate": 3.408732565191025e-05, + "loss": 1.0143, + "step": 184180 + }, + { + "epoch": 15.94, + "learning_rate": 3.4086459325998445e-05, + "loss": 1.0645, + "step": 184190 + }, + { + "epoch": 15.94, + "learning_rate": 3.408559300008663e-05, + "loss": 1.012, + "step": 184200 + }, + { + "epoch": 15.94, + "learning_rate": 3.408472667417483e-05, + "loss": 1.0355, + "step": 184210 + }, + { + "epoch": 15.95, + "learning_rate": 3.408386034826302e-05, + "loss": 1.0226, + "step": 184220 + }, + { + "epoch": 15.95, + "learning_rate": 3.408299402235121e-05, + "loss": 1.0133, + "step": 184230 + }, + { + "epoch": 15.95, + "learning_rate": 3.4082127696439404e-05, + "loss": 1.0724, + "step": 184240 + }, + { + "epoch": 15.95, + "learning_rate": 3.408126137052759e-05, + "loss": 1.0517, + "step": 184250 + }, + { + "epoch": 15.95, + "learning_rate": 3.4080395044615786e-05, + "loss": 1.1096, + "step": 184260 + }, + { + "epoch": 15.95, + "learning_rate": 3.407952871870398e-05, + "loss": 1.0251, + "step": 184270 + }, + { + "epoch": 15.95, + "learning_rate": 3.407866239279217e-05, + "loss": 1.0393, + "step": 184280 + }, + { + "epoch": 15.95, + "learning_rate": 3.407779606688036e-05, + "loss": 1.0892, + "step": 184290 + }, + { + "epoch": 15.95, + "learning_rate": 3.407692974096856e-05, + "loss": 1.0696, + "step": 184300 + }, + { + "epoch": 15.95, + "learning_rate": 3.4076063415056745e-05, + "loss": 1.0795, + "step": 184310 + }, + { + "epoch": 15.95, + "learning_rate": 3.407519708914494e-05, + "loss": 1.0412, + "step": 184320 + }, + { + "epoch": 15.95, + "learning_rate": 3.4074330763233134e-05, + "loss": 1.0475, + "step": 184330 + }, + { + "epoch": 15.96, + "learning_rate": 3.407346443732132e-05, + "loss": 1.0455, + "step": 184340 + }, + { + "epoch": 15.96, + "learning_rate": 3.4072598111409516e-05, + "loss": 1.0603, + "step": 184350 + }, + { + "epoch": 15.96, + "learning_rate": 3.4071731785497704e-05, + "loss": 1.0641, + "step": 184360 + }, + { + "epoch": 15.96, + "learning_rate": 3.40708654595859e-05, + "loss": 1.0149, + "step": 184370 + }, + { + "epoch": 15.96, + "learning_rate": 3.406999913367409e-05, + "loss": 1.0662, + "step": 184380 + }, + { + "epoch": 15.96, + "learning_rate": 3.406913280776228e-05, + "loss": 0.9886, + "step": 184390 + }, + { + "epoch": 15.96, + "learning_rate": 3.4068266481850475e-05, + "loss": 0.9941, + "step": 184400 + }, + { + "epoch": 15.96, + "learning_rate": 3.406740015593867e-05, + "loss": 1.0661, + "step": 184410 + }, + { + "epoch": 15.96, + "learning_rate": 3.406653383002686e-05, + "loss": 1.1033, + "step": 184420 + }, + { + "epoch": 15.96, + "learning_rate": 3.406566750411505e-05, + "loss": 1.054, + "step": 184430 + }, + { + "epoch": 15.96, + "learning_rate": 3.4064801178203246e-05, + "loss": 0.9765, + "step": 184440 + }, + { + "epoch": 15.97, + "learning_rate": 3.4063934852291433e-05, + "loss": 1.0713, + "step": 184450 + }, + { + "epoch": 15.97, + "learning_rate": 3.406306852637963e-05, + "loss": 1.0639, + "step": 184460 + }, + { + "epoch": 15.97, + "learning_rate": 3.4062202200467816e-05, + "loss": 1.0204, + "step": 184470 + }, + { + "epoch": 15.97, + "learning_rate": 3.406133587455601e-05, + "loss": 0.996, + "step": 184480 + }, + { + "epoch": 15.97, + "learning_rate": 3.4060469548644204e-05, + "loss": 1.0205, + "step": 184490 + }, + { + "epoch": 15.97, + "learning_rate": 3.405960322273239e-05, + "loss": 1.0428, + "step": 184500 + }, + { + "epoch": 15.97, + "learning_rate": 3.4058736896820587e-05, + "loss": 1.0812, + "step": 184510 + }, + { + "epoch": 15.97, + "learning_rate": 3.405787057090878e-05, + "loss": 0.9979, + "step": 184520 + }, + { + "epoch": 15.97, + "learning_rate": 3.405700424499697e-05, + "loss": 1.0744, + "step": 184530 + }, + { + "epoch": 15.97, + "learning_rate": 3.405613791908516e-05, + "loss": 1.0688, + "step": 184540 + }, + { + "epoch": 15.97, + "learning_rate": 3.405527159317336e-05, + "loss": 1.079, + "step": 184550 + }, + { + "epoch": 15.97, + "learning_rate": 3.4054405267261545e-05, + "loss": 1.0388, + "step": 184560 + }, + { + "epoch": 15.98, + "learning_rate": 3.405353894134974e-05, + "loss": 1.0088, + "step": 184570 + }, + { + "epoch": 15.98, + "learning_rate": 3.405267261543793e-05, + "loss": 1.0316, + "step": 184580 + }, + { + "epoch": 15.98, + "learning_rate": 3.405180628952612e-05, + "loss": 1.0727, + "step": 184590 + }, + { + "epoch": 15.98, + "learning_rate": 3.4050939963614316e-05, + "loss": 1.0269, + "step": 184600 + }, + { + "epoch": 15.98, + "learning_rate": 3.4050073637702504e-05, + "loss": 1.0525, + "step": 184610 + }, + { + "epoch": 15.98, + "learning_rate": 3.40492073117907e-05, + "loss": 1.0422, + "step": 184620 + }, + { + "epoch": 15.98, + "learning_rate": 3.404834098587889e-05, + "loss": 1.0213, + "step": 184630 + }, + { + "epoch": 15.98, + "learning_rate": 3.404747465996708e-05, + "loss": 1.0392, + "step": 184640 + }, + { + "epoch": 15.98, + "learning_rate": 3.4046608334055275e-05, + "loss": 1.0559, + "step": 184650 + }, + { + "epoch": 15.98, + "learning_rate": 3.404574200814347e-05, + "loss": 1.0395, + "step": 184660 + }, + { + "epoch": 15.98, + "learning_rate": 3.404487568223166e-05, + "loss": 1.0574, + "step": 184670 + }, + { + "epoch": 15.99, + "learning_rate": 3.404400935631985e-05, + "loss": 1.0898, + "step": 184680 + }, + { + "epoch": 15.99, + "learning_rate": 3.404314303040804e-05, + "loss": 1.0296, + "step": 184690 + }, + { + "epoch": 15.99, + "learning_rate": 3.4042276704496234e-05, + "loss": 1.0532, + "step": 184700 + }, + { + "epoch": 15.99, + "learning_rate": 3.404141037858443e-05, + "loss": 1.0113, + "step": 184710 + }, + { + "epoch": 15.99, + "learning_rate": 3.4040544052672616e-05, + "loss": 1.004, + "step": 184720 + }, + { + "epoch": 15.99, + "learning_rate": 3.403967772676081e-05, + "loss": 1.0594, + "step": 184730 + }, + { + "epoch": 15.99, + "learning_rate": 3.4038811400849005e-05, + "loss": 1.0391, + "step": 184740 + }, + { + "epoch": 15.99, + "learning_rate": 3.403794507493719e-05, + "loss": 1.033, + "step": 184750 + }, + { + "epoch": 15.99, + "learning_rate": 3.403707874902539e-05, + "loss": 1.0143, + "step": 184760 + }, + { + "epoch": 15.99, + "learning_rate": 3.403621242311358e-05, + "loss": 1.0326, + "step": 184770 + }, + { + "epoch": 15.99, + "learning_rate": 3.403534609720177e-05, + "loss": 1.02, + "step": 184780 + }, + { + "epoch": 15.99, + "learning_rate": 3.4034479771289964e-05, + "loss": 1.0435, + "step": 184790 + }, + { + "epoch": 16.0, + "learning_rate": 3.403361344537815e-05, + "loss": 1.0452, + "step": 184800 + }, + { + "epoch": 16.0, + "learning_rate": 3.4032747119466346e-05, + "loss": 1.0822, + "step": 184810 + }, + { + "epoch": 16.0, + "learning_rate": 3.403188079355454e-05, + "loss": 0.9678, + "step": 184820 + }, + { + "epoch": 16.0, + "learning_rate": 3.403101446764273e-05, + "loss": 1.0756, + "step": 184830 + }, + { + "epoch": 16.0, + "learning_rate": 3.403014814173092e-05, + "loss": 1.0307, + "step": 184840 + }, + { + "epoch": 16.0, + "learning_rate": 3.402928181581912e-05, + "loss": 1.0384, + "step": 184850 + }, + { + "epoch": 16.0, + "eval_Bleu_1": 0.042099391314333774, + "eval_Bleu_2": 2.746366882084911e-11, + "eval_Bleu_3": 2.458084079017139e-14, + "eval_Bleu_4": 7.547279996436947e-16, + "eval_ROUGE_L": 0.08651493103251238, + "eval_cer": 0.993294588714041, + "eval_em": 0, + "eval_f1": 0.10694390121260682, + "eval_loss": 1.0050785541534424, + "eval_runtime": 1689.9513, + "eval_samples_per_second": 3.039, + "eval_steps_per_second": 3.039, + "eval_wer": 0.9686652791729928, + "step": 184853 + }, + { + "epoch": 16.0, + "learning_rate": 3.4028415489907304e-05, + "loss": 1.0348, + "step": 184860 + }, + { + "epoch": 16.0, + "learning_rate": 3.40275491639955e-05, + "loss": 1.0063, + "step": 184870 + }, + { + "epoch": 16.0, + "learning_rate": 3.402668283808369e-05, + "loss": 1.0662, + "step": 184880 + }, + { + "epoch": 16.0, + "learning_rate": 3.402581651217188e-05, + "loss": 0.9299, + "step": 184890 + }, + { + "epoch": 16.0, + "learning_rate": 3.4024950186260075e-05, + "loss": 1.035, + "step": 184900 + }, + { + "epoch": 16.0, + "learning_rate": 3.402408386034826e-05, + "loss": 0.9943, + "step": 184910 + }, + { + "epoch": 16.01, + "learning_rate": 3.402321753443646e-05, + "loss": 0.9921, + "step": 184920 + }, + { + "epoch": 16.01, + "learning_rate": 3.402235120852465e-05, + "loss": 0.9851, + "step": 184930 + }, + { + "epoch": 16.01, + "learning_rate": 3.402148488261284e-05, + "loss": 1.0657, + "step": 184940 + }, + { + "epoch": 16.01, + "learning_rate": 3.4020618556701034e-05, + "loss": 0.9865, + "step": 184950 + }, + { + "epoch": 16.01, + "learning_rate": 3.401975223078923e-05, + "loss": 1.0676, + "step": 184960 + }, + { + "epoch": 16.01, + "learning_rate": 3.4018885904877416e-05, + "loss": 1.023, + "step": 184970 + }, + { + "epoch": 16.01, + "learning_rate": 3.401801957896561e-05, + "loss": 1.0021, + "step": 184980 + }, + { + "epoch": 16.01, + "learning_rate": 3.40171532530538e-05, + "loss": 1.0024, + "step": 184990 + }, + { + "epoch": 16.01, + "learning_rate": 3.401628692714199e-05, + "loss": 1.0463, + "step": 185000 + }, + { + "epoch": 16.01, + "learning_rate": 3.401542060123019e-05, + "loss": 1.029, + "step": 185010 + }, + { + "epoch": 16.01, + "learning_rate": 3.4014554275318375e-05, + "loss": 1.0005, + "step": 185020 + }, + { + "epoch": 16.02, + "learning_rate": 3.401368794940657e-05, + "loss": 0.9519, + "step": 185030 + }, + { + "epoch": 16.02, + "learning_rate": 3.4012821623494764e-05, + "loss": 0.9924, + "step": 185040 + }, + { + "epoch": 16.02, + "learning_rate": 3.401195529758295e-05, + "loss": 1.0215, + "step": 185050 + }, + { + "epoch": 16.02, + "learning_rate": 3.4011088971671146e-05, + "loss": 1.0596, + "step": 185060 + }, + { + "epoch": 16.02, + "learning_rate": 3.401022264575934e-05, + "loss": 0.9967, + "step": 185070 + }, + { + "epoch": 16.02, + "learning_rate": 3.400935631984753e-05, + "loss": 0.987, + "step": 185080 + }, + { + "epoch": 16.02, + "learning_rate": 3.400848999393572e-05, + "loss": 1.0056, + "step": 185090 + }, + { + "epoch": 16.02, + "learning_rate": 3.400762366802391e-05, + "loss": 1.0114, + "step": 185100 + }, + { + "epoch": 16.02, + "learning_rate": 3.4006757342112105e-05, + "loss": 1.0495, + "step": 185110 + }, + { + "epoch": 16.02, + "learning_rate": 3.40058910162003e-05, + "loss": 1.0219, + "step": 185120 + }, + { + "epoch": 16.02, + "learning_rate": 3.400502469028849e-05, + "loss": 1.0843, + "step": 185130 + }, + { + "epoch": 16.02, + "learning_rate": 3.400415836437668e-05, + "loss": 1.0042, + "step": 185140 + }, + { + "epoch": 16.03, + "learning_rate": 3.4003292038464876e-05, + "loss": 1.034, + "step": 185150 + }, + { + "epoch": 16.03, + "learning_rate": 3.4002425712553063e-05, + "loss": 1.0283, + "step": 185160 + }, + { + "epoch": 16.03, + "learning_rate": 3.400155938664126e-05, + "loss": 1.0216, + "step": 185170 + }, + { + "epoch": 16.03, + "learning_rate": 3.400069306072945e-05, + "loss": 1.07, + "step": 185180 + }, + { + "epoch": 16.03, + "learning_rate": 3.399982673481764e-05, + "loss": 1.0503, + "step": 185190 + }, + { + "epoch": 16.03, + "learning_rate": 3.3998960408905834e-05, + "loss": 0.9972, + "step": 185200 + }, + { + "epoch": 16.03, + "learning_rate": 3.399809408299402e-05, + "loss": 1.0274, + "step": 185210 + }, + { + "epoch": 16.03, + "learning_rate": 3.3997227757082217e-05, + "loss": 1.0109, + "step": 185220 + }, + { + "epoch": 16.03, + "learning_rate": 3.399636143117041e-05, + "loss": 1.026, + "step": 185230 + }, + { + "epoch": 16.03, + "learning_rate": 3.39954951052586e-05, + "loss": 0.9897, + "step": 185240 + }, + { + "epoch": 16.03, + "learning_rate": 3.399462877934679e-05, + "loss": 0.9936, + "step": 185250 + }, + { + "epoch": 16.04, + "learning_rate": 3.399376245343499e-05, + "loss": 0.9603, + "step": 185260 + }, + { + "epoch": 16.04, + "learning_rate": 3.3992896127523175e-05, + "loss": 1.0504, + "step": 185270 + }, + { + "epoch": 16.04, + "learning_rate": 3.399202980161137e-05, + "loss": 1.0472, + "step": 185280 + }, + { + "epoch": 16.04, + "learning_rate": 3.3991163475699564e-05, + "loss": 1.0387, + "step": 185290 + }, + { + "epoch": 16.04, + "learning_rate": 3.399029714978775e-05, + "loss": 1.0296, + "step": 185300 + }, + { + "epoch": 16.04, + "learning_rate": 3.3989430823875946e-05, + "loss": 0.9909, + "step": 185310 + }, + { + "epoch": 16.04, + "learning_rate": 3.3988564497964134e-05, + "loss": 0.9922, + "step": 185320 + }, + { + "epoch": 16.04, + "learning_rate": 3.398769817205233e-05, + "loss": 1.0158, + "step": 185330 + }, + { + "epoch": 16.04, + "learning_rate": 3.398683184614052e-05, + "loss": 1.055, + "step": 185340 + }, + { + "epoch": 16.04, + "learning_rate": 3.398596552022871e-05, + "loss": 1.0131, + "step": 185350 + }, + { + "epoch": 16.04, + "learning_rate": 3.3985099194316905e-05, + "loss": 1.0448, + "step": 185360 + }, + { + "epoch": 16.04, + "learning_rate": 3.39842328684051e-05, + "loss": 1.0527, + "step": 185370 + }, + { + "epoch": 16.05, + "learning_rate": 3.398336654249329e-05, + "loss": 1.0043, + "step": 185380 + }, + { + "epoch": 16.05, + "learning_rate": 3.398250021658148e-05, + "loss": 1.0324, + "step": 185390 + }, + { + "epoch": 16.05, + "learning_rate": 3.3981633890669676e-05, + "loss": 1.0265, + "step": 185400 + }, + { + "epoch": 16.05, + "learning_rate": 3.3980767564757864e-05, + "loss": 1.0105, + "step": 185410 + }, + { + "epoch": 16.05, + "learning_rate": 3.397990123884606e-05, + "loss": 1.0155, + "step": 185420 + }, + { + "epoch": 16.05, + "learning_rate": 3.3979034912934246e-05, + "loss": 1.0548, + "step": 185430 + }, + { + "epoch": 16.05, + "learning_rate": 3.397816858702244e-05, + "loss": 0.9619, + "step": 185440 + }, + { + "epoch": 16.05, + "learning_rate": 3.3977302261110635e-05, + "loss": 1.0606, + "step": 185450 + }, + { + "epoch": 16.05, + "learning_rate": 3.397643593519882e-05, + "loss": 1.0643, + "step": 185460 + }, + { + "epoch": 16.05, + "learning_rate": 3.397556960928702e-05, + "loss": 1.0367, + "step": 185470 + }, + { + "epoch": 16.05, + "learning_rate": 3.397470328337521e-05, + "loss": 1.0424, + "step": 185480 + }, + { + "epoch": 16.06, + "learning_rate": 3.39738369574634e-05, + "loss": 1.0166, + "step": 185490 + }, + { + "epoch": 16.06, + "learning_rate": 3.3972970631551593e-05, + "loss": 1.049, + "step": 185500 + }, + { + "epoch": 16.06, + "learning_rate": 3.397210430563979e-05, + "loss": 1.0627, + "step": 185510 + }, + { + "epoch": 16.06, + "learning_rate": 3.3971237979727976e-05, + "loss": 1.1075, + "step": 185520 + }, + { + "epoch": 16.06, + "learning_rate": 3.397037165381617e-05, + "loss": 1.0075, + "step": 185530 + }, + { + "epoch": 16.06, + "learning_rate": 3.396950532790436e-05, + "loss": 1.0884, + "step": 185540 + }, + { + "epoch": 16.06, + "learning_rate": 3.396863900199255e-05, + "loss": 0.9887, + "step": 185550 + }, + { + "epoch": 16.06, + "learning_rate": 3.396777267608075e-05, + "loss": 1.0428, + "step": 185560 + }, + { + "epoch": 16.06, + "learning_rate": 3.3966906350168934e-05, + "loss": 1.0486, + "step": 185570 + }, + { + "epoch": 16.06, + "learning_rate": 3.396604002425713e-05, + "loss": 1.0481, + "step": 185580 + }, + { + "epoch": 16.06, + "learning_rate": 3.396517369834532e-05, + "loss": 1.0237, + "step": 185590 + }, + { + "epoch": 16.06, + "learning_rate": 3.396430737243351e-05, + "loss": 1.0024, + "step": 185600 + }, + { + "epoch": 16.07, + "learning_rate": 3.3963441046521705e-05, + "loss": 1.0238, + "step": 185610 + }, + { + "epoch": 16.07, + "learning_rate": 3.39625747206099e-05, + "loss": 1.0202, + "step": 185620 + }, + { + "epoch": 16.07, + "learning_rate": 3.396170839469809e-05, + "loss": 0.9898, + "step": 185630 + }, + { + "epoch": 16.07, + "learning_rate": 3.396084206878628e-05, + "loss": 1.0193, + "step": 185640 + }, + { + "epoch": 16.07, + "learning_rate": 3.395997574287447e-05, + "loss": 0.945, + "step": 185650 + }, + { + "epoch": 16.07, + "learning_rate": 3.3959109416962664e-05, + "loss": 1.0104, + "step": 185660 + }, + { + "epoch": 16.07, + "learning_rate": 3.395824309105086e-05, + "loss": 0.9534, + "step": 185670 + }, + { + "epoch": 16.07, + "learning_rate": 3.3957376765139046e-05, + "loss": 1.0023, + "step": 185680 + }, + { + "epoch": 16.07, + "learning_rate": 3.395651043922724e-05, + "loss": 1.0548, + "step": 185690 + }, + { + "epoch": 16.07, + "learning_rate": 3.3955644113315435e-05, + "loss": 0.9996, + "step": 185700 + }, + { + "epoch": 16.07, + "learning_rate": 3.395477778740362e-05, + "loss": 1.028, + "step": 185710 + }, + { + "epoch": 16.08, + "learning_rate": 3.395391146149182e-05, + "loss": 1.0144, + "step": 185720 + }, + { + "epoch": 16.08, + "learning_rate": 3.395304513558001e-05, + "loss": 1.0091, + "step": 185730 + }, + { + "epoch": 16.08, + "learning_rate": 3.39521788096682e-05, + "loss": 1.0403, + "step": 185740 + }, + { + "epoch": 16.08, + "learning_rate": 3.3951312483756394e-05, + "loss": 1.0584, + "step": 185750 + }, + { + "epoch": 16.08, + "learning_rate": 3.395044615784458e-05, + "loss": 1.0351, + "step": 185760 + }, + { + "epoch": 16.08, + "learning_rate": 3.3949579831932776e-05, + "loss": 1.0438, + "step": 185770 + }, + { + "epoch": 16.08, + "learning_rate": 3.394871350602097e-05, + "loss": 0.9658, + "step": 185780 + }, + { + "epoch": 16.08, + "learning_rate": 3.394784718010916e-05, + "loss": 1.0502, + "step": 185790 + }, + { + "epoch": 16.08, + "learning_rate": 3.394698085419735e-05, + "loss": 1.0219, + "step": 185800 + }, + { + "epoch": 16.08, + "learning_rate": 3.394611452828555e-05, + "loss": 1.0184, + "step": 185810 + }, + { + "epoch": 16.08, + "learning_rate": 3.3945248202373735e-05, + "loss": 1.0212, + "step": 185820 + }, + { + "epoch": 16.08, + "learning_rate": 3.394438187646193e-05, + "loss": 1.0599, + "step": 185830 + }, + { + "epoch": 16.09, + "learning_rate": 3.394351555055012e-05, + "loss": 1.011, + "step": 185840 + }, + { + "epoch": 16.09, + "learning_rate": 3.394264922463831e-05, + "loss": 1.0292, + "step": 185850 + }, + { + "epoch": 16.09, + "learning_rate": 3.3941782898726506e-05, + "loss": 1.0705, + "step": 185860 + }, + { + "epoch": 16.09, + "learning_rate": 3.394091657281469e-05, + "loss": 0.9854, + "step": 185870 + }, + { + "epoch": 16.09, + "learning_rate": 3.394005024690289e-05, + "loss": 1.0546, + "step": 185880 + }, + { + "epoch": 16.09, + "learning_rate": 3.393918392099108e-05, + "loss": 1.0053, + "step": 185890 + }, + { + "epoch": 16.09, + "learning_rate": 3.393831759507927e-05, + "loss": 1.0597, + "step": 185900 + }, + { + "epoch": 16.09, + "learning_rate": 3.3937451269167464e-05, + "loss": 0.9946, + "step": 185910 + }, + { + "epoch": 16.09, + "learning_rate": 3.393658494325566e-05, + "loss": 1.0227, + "step": 185920 + }, + { + "epoch": 16.09, + "learning_rate": 3.3935718617343847e-05, + "loss": 0.97, + "step": 185930 + }, + { + "epoch": 16.09, + "learning_rate": 3.393485229143204e-05, + "loss": 1.0808, + "step": 185940 + }, + { + "epoch": 16.09, + "learning_rate": 3.393398596552023e-05, + "loss": 1.0235, + "step": 185950 + }, + { + "epoch": 16.1, + "learning_rate": 3.393311963960842e-05, + "loss": 1.0142, + "step": 185960 + }, + { + "epoch": 16.1, + "learning_rate": 3.393225331369662e-05, + "loss": 1.0041, + "step": 185970 + }, + { + "epoch": 16.1, + "learning_rate": 3.3931386987784805e-05, + "loss": 1.0084, + "step": 185980 + }, + { + "epoch": 16.1, + "learning_rate": 3.3930520661873e-05, + "loss": 1.0341, + "step": 185990 + }, + { + "epoch": 16.1, + "learning_rate": 3.3929654335961194e-05, + "loss": 1.077, + "step": 186000 + }, + { + "epoch": 16.1, + "learning_rate": 3.392878801004938e-05, + "loss": 1.0331, + "step": 186010 + }, + { + "epoch": 16.1, + "learning_rate": 3.3927921684137576e-05, + "loss": 1.0406, + "step": 186020 + }, + { + "epoch": 16.1, + "learning_rate": 3.392705535822577e-05, + "loss": 1.0279, + "step": 186030 + }, + { + "epoch": 16.1, + "learning_rate": 3.392618903231396e-05, + "loss": 1.1037, + "step": 186040 + }, + { + "epoch": 16.1, + "learning_rate": 3.392532270640215e-05, + "loss": 0.993, + "step": 186050 + }, + { + "epoch": 16.1, + "learning_rate": 3.392445638049034e-05, + "loss": 1.0474, + "step": 186060 + }, + { + "epoch": 16.11, + "learning_rate": 3.3923590054578535e-05, + "loss": 1.0207, + "step": 186070 + }, + { + "epoch": 16.11, + "learning_rate": 3.392272372866673e-05, + "loss": 1.0207, + "step": 186080 + }, + { + "epoch": 16.11, + "learning_rate": 3.392185740275492e-05, + "loss": 0.9701, + "step": 186090 + }, + { + "epoch": 16.11, + "learning_rate": 3.392099107684311e-05, + "loss": 0.9813, + "step": 186100 + }, + { + "epoch": 16.11, + "learning_rate": 3.3920124750931306e-05, + "loss": 1.0797, + "step": 186110 + }, + { + "epoch": 16.11, + "learning_rate": 3.3919258425019494e-05, + "loss": 1.0017, + "step": 186120 + }, + { + "epoch": 16.11, + "learning_rate": 3.391839209910769e-05, + "loss": 1.067, + "step": 186130 + }, + { + "epoch": 16.11, + "learning_rate": 3.391752577319588e-05, + "loss": 1.0029, + "step": 186140 + }, + { + "epoch": 16.11, + "learning_rate": 3.391665944728407e-05, + "loss": 1.068, + "step": 186150 + }, + { + "epoch": 16.11, + "learning_rate": 3.3915793121372265e-05, + "loss": 1.0207, + "step": 186160 + }, + { + "epoch": 16.11, + "learning_rate": 3.391492679546045e-05, + "loss": 1.0243, + "step": 186170 + }, + { + "epoch": 16.11, + "learning_rate": 3.391406046954865e-05, + "loss": 1.0664, + "step": 186180 + }, + { + "epoch": 16.12, + "learning_rate": 3.391319414363684e-05, + "loss": 1.0817, + "step": 186190 + }, + { + "epoch": 16.12, + "learning_rate": 3.391232781772503e-05, + "loss": 1.062, + "step": 186200 + }, + { + "epoch": 16.12, + "learning_rate": 3.3911461491813223e-05, + "loss": 0.9992, + "step": 186210 + }, + { + "epoch": 16.12, + "learning_rate": 3.391059516590142e-05, + "loss": 1.0226, + "step": 186220 + }, + { + "epoch": 16.12, + "learning_rate": 3.3909728839989606e-05, + "loss": 0.9877, + "step": 186230 + }, + { + "epoch": 16.12, + "learning_rate": 3.39088625140778e-05, + "loss": 0.9834, + "step": 186240 + }, + { + "epoch": 16.12, + "learning_rate": 3.3907996188165994e-05, + "loss": 1.0288, + "step": 186250 + }, + { + "epoch": 16.12, + "learning_rate": 3.390712986225418e-05, + "loss": 0.986, + "step": 186260 + }, + { + "epoch": 16.12, + "learning_rate": 3.390626353634238e-05, + "loss": 1.0038, + "step": 186270 + }, + { + "epoch": 16.12, + "learning_rate": 3.3905397210430564e-05, + "loss": 0.9701, + "step": 186280 + }, + { + "epoch": 16.12, + "learning_rate": 3.390453088451876e-05, + "loss": 1.0527, + "step": 186290 + }, + { + "epoch": 16.13, + "learning_rate": 3.390366455860695e-05, + "loss": 0.9816, + "step": 186300 + }, + { + "epoch": 16.13, + "learning_rate": 3.390279823269514e-05, + "loss": 1.0721, + "step": 186310 + }, + { + "epoch": 16.13, + "learning_rate": 3.3901931906783335e-05, + "loss": 0.9898, + "step": 186320 + }, + { + "epoch": 16.13, + "learning_rate": 3.390106558087153e-05, + "loss": 1.0244, + "step": 186330 + }, + { + "epoch": 16.13, + "learning_rate": 3.390019925495972e-05, + "loss": 0.9782, + "step": 186340 + }, + { + "epoch": 16.13, + "learning_rate": 3.389933292904791e-05, + "loss": 1.0516, + "step": 186350 + }, + { + "epoch": 16.13, + "learning_rate": 3.3898466603136106e-05, + "loss": 1.0267, + "step": 186360 + }, + { + "epoch": 16.13, + "learning_rate": 3.3897600277224294e-05, + "loss": 1.0783, + "step": 186370 + }, + { + "epoch": 16.13, + "learning_rate": 3.389673395131249e-05, + "loss": 1.064, + "step": 186380 + }, + { + "epoch": 16.13, + "learning_rate": 3.3895867625400676e-05, + "loss": 1.0924, + "step": 186390 + }, + { + "epoch": 16.13, + "learning_rate": 3.389500129948887e-05, + "loss": 1.0463, + "step": 186400 + }, + { + "epoch": 16.13, + "learning_rate": 3.3894134973577065e-05, + "loss": 0.9993, + "step": 186410 + }, + { + "epoch": 16.14, + "learning_rate": 3.389326864766525e-05, + "loss": 1.0514, + "step": 186420 + }, + { + "epoch": 16.14, + "learning_rate": 3.389240232175345e-05, + "loss": 1.0747, + "step": 186430 + }, + { + "epoch": 16.14, + "learning_rate": 3.389153599584164e-05, + "loss": 1.0015, + "step": 186440 + }, + { + "epoch": 16.14, + "learning_rate": 3.389066966992983e-05, + "loss": 0.9545, + "step": 186450 + }, + { + "epoch": 16.14, + "learning_rate": 3.3889803344018024e-05, + "loss": 1.0204, + "step": 186460 + }, + { + "epoch": 16.14, + "learning_rate": 3.388893701810622e-05, + "loss": 1.0218, + "step": 186470 + }, + { + "epoch": 16.14, + "learning_rate": 3.3888070692194406e-05, + "loss": 1.0314, + "step": 186480 + }, + { + "epoch": 16.14, + "learning_rate": 3.38872043662826e-05, + "loss": 1.042, + "step": 186490 + }, + { + "epoch": 16.14, + "learning_rate": 3.388633804037079e-05, + "loss": 0.9968, + "step": 186500 + }, + { + "epoch": 16.14, + "learning_rate": 3.388547171445898e-05, + "loss": 0.9949, + "step": 186510 + }, + { + "epoch": 16.14, + "learning_rate": 3.388460538854718e-05, + "loss": 1.0931, + "step": 186520 + }, + { + "epoch": 16.15, + "learning_rate": 3.3883739062635365e-05, + "loss": 1.0131, + "step": 186530 + }, + { + "epoch": 16.15, + "learning_rate": 3.388287273672356e-05, + "loss": 0.9816, + "step": 186540 + }, + { + "epoch": 16.15, + "learning_rate": 3.3882006410811754e-05, + "loss": 1.0335, + "step": 186550 + }, + { + "epoch": 16.15, + "learning_rate": 3.388114008489994e-05, + "loss": 0.9782, + "step": 186560 + }, + { + "epoch": 16.15, + "learning_rate": 3.3880273758988136e-05, + "loss": 1.022, + "step": 186570 + }, + { + "epoch": 16.15, + "learning_rate": 3.387940743307632e-05, + "loss": 0.9932, + "step": 186580 + }, + { + "epoch": 16.15, + "learning_rate": 3.387854110716452e-05, + "loss": 0.9941, + "step": 186590 + }, + { + "epoch": 16.15, + "learning_rate": 3.387767478125271e-05, + "loss": 1.0058, + "step": 186600 + }, + { + "epoch": 16.15, + "learning_rate": 3.38768084553409e-05, + "loss": 1.011, + "step": 186610 + }, + { + "epoch": 16.15, + "learning_rate": 3.3875942129429094e-05, + "loss": 1.0256, + "step": 186620 + }, + { + "epoch": 16.15, + "learning_rate": 3.387507580351729e-05, + "loss": 0.9862, + "step": 186630 + }, + { + "epoch": 16.15, + "learning_rate": 3.3874209477605477e-05, + "loss": 1.0219, + "step": 186640 + }, + { + "epoch": 16.16, + "learning_rate": 3.387334315169367e-05, + "loss": 0.9602, + "step": 186650 + }, + { + "epoch": 16.16, + "learning_rate": 3.3872476825781865e-05, + "loss": 0.982, + "step": 186660 + }, + { + "epoch": 16.16, + "learning_rate": 3.387161049987005e-05, + "loss": 1.0197, + "step": 186670 + }, + { + "epoch": 16.16, + "learning_rate": 3.387074417395825e-05, + "loss": 1.0422, + "step": 186680 + }, + { + "epoch": 16.16, + "learning_rate": 3.3869877848046435e-05, + "loss": 1.0362, + "step": 186690 + }, + { + "epoch": 16.16, + "learning_rate": 3.386901152213463e-05, + "loss": 1.0407, + "step": 186700 + }, + { + "epoch": 16.16, + "learning_rate": 3.3868145196222824e-05, + "loss": 1.018, + "step": 186710 + }, + { + "epoch": 16.16, + "learning_rate": 3.386727887031101e-05, + "loss": 0.9764, + "step": 186720 + }, + { + "epoch": 16.16, + "learning_rate": 3.3866412544399206e-05, + "loss": 1.0316, + "step": 186730 + }, + { + "epoch": 16.16, + "learning_rate": 3.38655462184874e-05, + "loss": 1.0512, + "step": 186740 + }, + { + "epoch": 16.16, + "learning_rate": 3.386467989257559e-05, + "loss": 0.9885, + "step": 186750 + }, + { + "epoch": 16.17, + "learning_rate": 3.386381356666378e-05, + "loss": 1.0348, + "step": 186760 + }, + { + "epoch": 16.17, + "learning_rate": 3.386294724075198e-05, + "loss": 0.9887, + "step": 186770 + }, + { + "epoch": 16.17, + "learning_rate": 3.3862080914840165e-05, + "loss": 0.9587, + "step": 186780 + }, + { + "epoch": 16.17, + "learning_rate": 3.386121458892836e-05, + "loss": 1.0199, + "step": 186790 + }, + { + "epoch": 16.17, + "learning_rate": 3.386034826301655e-05, + "loss": 0.9997, + "step": 186800 + }, + { + "epoch": 16.17, + "learning_rate": 3.385948193710474e-05, + "loss": 1.0367, + "step": 186810 + }, + { + "epoch": 16.17, + "learning_rate": 3.3858615611192936e-05, + "loss": 1.0219, + "step": 186820 + }, + { + "epoch": 16.17, + "learning_rate": 3.3857749285281124e-05, + "loss": 1.0207, + "step": 186830 + }, + { + "epoch": 16.17, + "learning_rate": 3.385688295936932e-05, + "loss": 1.0229, + "step": 186840 + }, + { + "epoch": 16.17, + "learning_rate": 3.385601663345751e-05, + "loss": 1.0114, + "step": 186850 + }, + { + "epoch": 16.17, + "learning_rate": 3.38551503075457e-05, + "loss": 1.0282, + "step": 186860 + }, + { + "epoch": 16.17, + "learning_rate": 3.3854283981633895e-05, + "loss": 1.056, + "step": 186870 + }, + { + "epoch": 16.18, + "learning_rate": 3.385341765572209e-05, + "loss": 1.0143, + "step": 186880 + }, + { + "epoch": 16.18, + "learning_rate": 3.385255132981028e-05, + "loss": 1.0142, + "step": 186890 + }, + { + "epoch": 16.18, + "learning_rate": 3.385168500389847e-05, + "loss": 1.0087, + "step": 186900 + }, + { + "epoch": 16.18, + "learning_rate": 3.385081867798666e-05, + "loss": 1.0308, + "step": 186910 + }, + { + "epoch": 16.18, + "learning_rate": 3.3849952352074853e-05, + "loss": 0.9991, + "step": 186920 + }, + { + "epoch": 16.18, + "learning_rate": 3.384908602616305e-05, + "loss": 1.0556, + "step": 186930 + }, + { + "epoch": 16.18, + "learning_rate": 3.3848219700251236e-05, + "loss": 1.0083, + "step": 186940 + }, + { + "epoch": 16.18, + "learning_rate": 3.384735337433943e-05, + "loss": 1.0754, + "step": 186950 + }, + { + "epoch": 16.18, + "learning_rate": 3.3846487048427624e-05, + "loss": 0.976, + "step": 186960 + }, + { + "epoch": 16.18, + "learning_rate": 3.384562072251581e-05, + "loss": 0.9688, + "step": 186970 + }, + { + "epoch": 16.18, + "learning_rate": 3.3844754396604007e-05, + "loss": 1.0223, + "step": 186980 + }, + { + "epoch": 16.18, + "learning_rate": 3.38438880706922e-05, + "loss": 1.0481, + "step": 186990 + }, + { + "epoch": 16.19, + "learning_rate": 3.384302174478039e-05, + "loss": 1.0508, + "step": 187000 + }, + { + "epoch": 16.19, + "learning_rate": 3.384215541886858e-05, + "loss": 1.0463, + "step": 187010 + }, + { + "epoch": 16.19, + "learning_rate": 3.384128909295677e-05, + "loss": 1.0204, + "step": 187020 + }, + { + "epoch": 16.19, + "learning_rate": 3.3840422767044965e-05, + "loss": 1.0001, + "step": 187030 + }, + { + "epoch": 16.19, + "learning_rate": 3.383955644113316e-05, + "loss": 1.019, + "step": 187040 + }, + { + "epoch": 16.19, + "learning_rate": 3.383869011522135e-05, + "loss": 0.9576, + "step": 187050 + }, + { + "epoch": 16.19, + "learning_rate": 3.383782378930954e-05, + "loss": 1.032, + "step": 187060 + }, + { + "epoch": 16.19, + "learning_rate": 3.3836957463397736e-05, + "loss": 0.9892, + "step": 187070 + }, + { + "epoch": 16.19, + "learning_rate": 3.3836091137485924e-05, + "loss": 1.0499, + "step": 187080 + }, + { + "epoch": 16.19, + "learning_rate": 3.383522481157412e-05, + "loss": 1.1021, + "step": 187090 + }, + { + "epoch": 16.19, + "learning_rate": 3.383435848566231e-05, + "loss": 1.0324, + "step": 187100 + }, + { + "epoch": 16.2, + "learning_rate": 3.38334921597505e-05, + "loss": 1.0167, + "step": 187110 + }, + { + "epoch": 16.2, + "learning_rate": 3.3832625833838695e-05, + "loss": 1.0164, + "step": 187120 + }, + { + "epoch": 16.2, + "learning_rate": 3.383175950792688e-05, + "loss": 1.0079, + "step": 187130 + }, + { + "epoch": 16.2, + "learning_rate": 3.383089318201508e-05, + "loss": 1.0781, + "step": 187140 + }, + { + "epoch": 16.2, + "learning_rate": 3.383002685610327e-05, + "loss": 0.9834, + "step": 187150 + }, + { + "epoch": 16.2, + "learning_rate": 3.382916053019146e-05, + "loss": 1.0991, + "step": 187160 + }, + { + "epoch": 16.2, + "learning_rate": 3.3828294204279654e-05, + "loss": 1.0418, + "step": 187170 + }, + { + "epoch": 16.2, + "learning_rate": 3.382742787836785e-05, + "loss": 0.9487, + "step": 187180 + }, + { + "epoch": 16.2, + "learning_rate": 3.3826561552456036e-05, + "loss": 1.0092, + "step": 187190 + }, + { + "epoch": 16.2, + "learning_rate": 3.382569522654423e-05, + "loss": 1.037, + "step": 187200 + }, + { + "epoch": 16.2, + "learning_rate": 3.3824828900632425e-05, + "loss": 0.9432, + "step": 187210 + }, + { + "epoch": 16.2, + "learning_rate": 3.382396257472061e-05, + "loss": 0.9686, + "step": 187220 + }, + { + "epoch": 16.21, + "learning_rate": 3.382309624880881e-05, + "loss": 1.024, + "step": 187230 + }, + { + "epoch": 16.21, + "learning_rate": 3.3822229922896995e-05, + "loss": 1.0266, + "step": 187240 + }, + { + "epoch": 16.21, + "learning_rate": 3.382136359698519e-05, + "loss": 1.0399, + "step": 187250 + }, + { + "epoch": 16.21, + "learning_rate": 3.3820497271073384e-05, + "loss": 0.9791, + "step": 187260 + }, + { + "epoch": 16.21, + "learning_rate": 3.381963094516157e-05, + "loss": 0.9889, + "step": 187270 + }, + { + "epoch": 16.21, + "learning_rate": 3.3818764619249766e-05, + "loss": 1.0881, + "step": 187280 + }, + { + "epoch": 16.21, + "learning_rate": 3.381789829333796e-05, + "loss": 1.05, + "step": 187290 + }, + { + "epoch": 16.21, + "learning_rate": 3.381703196742615e-05, + "loss": 1.0408, + "step": 187300 + }, + { + "epoch": 16.21, + "learning_rate": 3.381616564151434e-05, + "loss": 1.009, + "step": 187310 + }, + { + "epoch": 16.21, + "learning_rate": 3.381529931560253e-05, + "loss": 1.0159, + "step": 187320 + }, + { + "epoch": 16.21, + "learning_rate": 3.3814432989690724e-05, + "loss": 0.9812, + "step": 187330 + }, + { + "epoch": 16.22, + "learning_rate": 3.381356666377892e-05, + "loss": 1.0368, + "step": 187340 + }, + { + "epoch": 16.22, + "learning_rate": 3.3812700337867106e-05, + "loss": 0.9733, + "step": 187350 + }, + { + "epoch": 16.22, + "learning_rate": 3.38118340119553e-05, + "loss": 0.9949, + "step": 187360 + }, + { + "epoch": 16.22, + "learning_rate": 3.3810967686043495e-05, + "loss": 1.0589, + "step": 187370 + }, + { + "epoch": 16.22, + "learning_rate": 3.381010136013168e-05, + "loss": 1.0714, + "step": 187380 + }, + { + "epoch": 16.22, + "learning_rate": 3.380923503421988e-05, + "loss": 1.0183, + "step": 187390 + }, + { + "epoch": 16.22, + "learning_rate": 3.380836870830807e-05, + "loss": 0.9966, + "step": 187400 + }, + { + "epoch": 16.22, + "learning_rate": 3.380750238239626e-05, + "loss": 1.013, + "step": 187410 + }, + { + "epoch": 16.22, + "learning_rate": 3.3806636056484454e-05, + "loss": 1.0128, + "step": 187420 + }, + { + "epoch": 16.22, + "learning_rate": 3.380576973057264e-05, + "loss": 0.9826, + "step": 187430 + }, + { + "epoch": 16.22, + "learning_rate": 3.3804903404660836e-05, + "loss": 1.003, + "step": 187440 + }, + { + "epoch": 16.22, + "learning_rate": 3.380403707874903e-05, + "loss": 0.9799, + "step": 187450 + }, + { + "epoch": 16.23, + "learning_rate": 3.380317075283722e-05, + "loss": 0.9861, + "step": 187460 + }, + { + "epoch": 16.23, + "learning_rate": 3.380230442692541e-05, + "loss": 1.067, + "step": 187470 + }, + { + "epoch": 16.23, + "learning_rate": 3.380143810101361e-05, + "loss": 0.9436, + "step": 187480 + }, + { + "epoch": 16.23, + "learning_rate": 3.3800571775101795e-05, + "loss": 1.0599, + "step": 187490 + }, + { + "epoch": 16.23, + "learning_rate": 3.379970544918999e-05, + "loss": 0.9946, + "step": 187500 + }, + { + "epoch": 16.23, + "learning_rate": 3.3798839123278184e-05, + "loss": 0.9961, + "step": 187510 + }, + { + "epoch": 16.23, + "learning_rate": 3.379797279736637e-05, + "loss": 1.0101, + "step": 187520 + }, + { + "epoch": 16.23, + "learning_rate": 3.3797106471454566e-05, + "loss": 0.9785, + "step": 187530 + }, + { + "epoch": 16.23, + "learning_rate": 3.3796240145542754e-05, + "loss": 1.026, + "step": 187540 + }, + { + "epoch": 16.23, + "learning_rate": 3.379537381963095e-05, + "loss": 1.0174, + "step": 187550 + }, + { + "epoch": 16.23, + "learning_rate": 3.379450749371914e-05, + "loss": 1.0342, + "step": 187560 + }, + { + "epoch": 16.24, + "learning_rate": 3.379364116780733e-05, + "loss": 1.0046, + "step": 187570 + }, + { + "epoch": 16.24, + "learning_rate": 3.3792774841895525e-05, + "loss": 1.0173, + "step": 187580 + }, + { + "epoch": 16.24, + "learning_rate": 3.379190851598372e-05, + "loss": 0.9999, + "step": 187590 + }, + { + "epoch": 16.24, + "learning_rate": 3.379104219007191e-05, + "loss": 0.9954, + "step": 187600 + }, + { + "epoch": 16.24, + "learning_rate": 3.37901758641601e-05, + "loss": 1.0387, + "step": 187610 + }, + { + "epoch": 16.24, + "learning_rate": 3.3789309538248296e-05, + "loss": 1.0616, + "step": 187620 + }, + { + "epoch": 16.24, + "learning_rate": 3.3788443212336483e-05, + "loss": 0.9963, + "step": 187630 + }, + { + "epoch": 16.24, + "learning_rate": 3.378757688642468e-05, + "loss": 0.9622, + "step": 187640 + }, + { + "epoch": 16.24, + "learning_rate": 3.3786710560512866e-05, + "loss": 1.0084, + "step": 187650 + }, + { + "epoch": 16.24, + "learning_rate": 3.378584423460106e-05, + "loss": 0.9808, + "step": 187660 + }, + { + "epoch": 16.24, + "learning_rate": 3.3784977908689254e-05, + "loss": 1.0709, + "step": 187670 + }, + { + "epoch": 16.24, + "learning_rate": 3.378411158277744e-05, + "loss": 1.0008, + "step": 187680 + }, + { + "epoch": 16.25, + "learning_rate": 3.3783245256865637e-05, + "loss": 1.0758, + "step": 187690 + }, + { + "epoch": 16.25, + "learning_rate": 3.378237893095383e-05, + "loss": 1.0495, + "step": 187700 + }, + { + "epoch": 16.25, + "learning_rate": 3.378151260504202e-05, + "loss": 1.0289, + "step": 187710 + }, + { + "epoch": 16.25, + "learning_rate": 3.378064627913021e-05, + "loss": 1.0062, + "step": 187720 + }, + { + "epoch": 16.25, + "learning_rate": 3.377977995321841e-05, + "loss": 1.0287, + "step": 187730 + }, + { + "epoch": 16.25, + "learning_rate": 3.3778913627306595e-05, + "loss": 1.0073, + "step": 187740 + }, + { + "epoch": 16.25, + "learning_rate": 3.377804730139479e-05, + "loss": 0.9732, + "step": 187750 + }, + { + "epoch": 16.25, + "learning_rate": 3.377718097548298e-05, + "loss": 1.0096, + "step": 187760 + }, + { + "epoch": 16.25, + "learning_rate": 3.377631464957117e-05, + "loss": 1.0457, + "step": 187770 + }, + { + "epoch": 16.25, + "learning_rate": 3.3775448323659366e-05, + "loss": 1.0065, + "step": 187780 + }, + { + "epoch": 16.25, + "learning_rate": 3.3774581997747554e-05, + "loss": 1.0261, + "step": 187790 + }, + { + "epoch": 16.26, + "learning_rate": 3.377371567183575e-05, + "loss": 0.9986, + "step": 187800 + }, + { + "epoch": 16.26, + "learning_rate": 3.377284934592394e-05, + "loss": 1.0106, + "step": 187810 + }, + { + "epoch": 16.26, + "learning_rate": 3.377198302001213e-05, + "loss": 0.9946, + "step": 187820 + }, + { + "epoch": 16.26, + "learning_rate": 3.3771116694100325e-05, + "loss": 0.992, + "step": 187830 + }, + { + "epoch": 16.26, + "learning_rate": 3.377025036818852e-05, + "loss": 1.0783, + "step": 187840 + }, + { + "epoch": 16.26, + "learning_rate": 3.376938404227671e-05, + "loss": 1.0436, + "step": 187850 + }, + { + "epoch": 16.26, + "learning_rate": 3.37685177163649e-05, + "loss": 1.0459, + "step": 187860 + }, + { + "epoch": 16.26, + "learning_rate": 3.376765139045309e-05, + "loss": 1.0254, + "step": 187870 + }, + { + "epoch": 16.26, + "learning_rate": 3.3766785064541284e-05, + "loss": 1.0102, + "step": 187880 + }, + { + "epoch": 16.26, + "learning_rate": 3.376591873862948e-05, + "loss": 0.9822, + "step": 187890 + }, + { + "epoch": 16.26, + "learning_rate": 3.3765052412717666e-05, + "loss": 1.0098, + "step": 187900 + }, + { + "epoch": 16.26, + "learning_rate": 3.376418608680586e-05, + "loss": 1.0025, + "step": 187910 + }, + { + "epoch": 16.27, + "learning_rate": 3.3763319760894055e-05, + "loss": 1.0911, + "step": 187920 + }, + { + "epoch": 16.27, + "learning_rate": 3.376245343498224e-05, + "loss": 1.03, + "step": 187930 + }, + { + "epoch": 16.27, + "learning_rate": 3.376158710907044e-05, + "loss": 0.9708, + "step": 187940 + }, + { + "epoch": 16.27, + "learning_rate": 3.376072078315863e-05, + "loss": 0.9958, + "step": 187950 + }, + { + "epoch": 16.27, + "learning_rate": 3.375985445724682e-05, + "loss": 1.0362, + "step": 187960 + }, + { + "epoch": 16.27, + "learning_rate": 3.3758988131335013e-05, + "loss": 1.0023, + "step": 187970 + }, + { + "epoch": 16.27, + "learning_rate": 3.37581218054232e-05, + "loss": 1.0241, + "step": 187980 + }, + { + "epoch": 16.27, + "learning_rate": 3.3757255479511396e-05, + "loss": 1.0078, + "step": 187990 + }, + { + "epoch": 16.27, + "learning_rate": 3.375638915359959e-05, + "loss": 1.0613, + "step": 188000 + }, + { + "epoch": 16.27, + "learning_rate": 3.375552282768778e-05, + "loss": 1.0515, + "step": 188010 + }, + { + "epoch": 16.27, + "learning_rate": 3.375465650177597e-05, + "loss": 0.9789, + "step": 188020 + }, + { + "epoch": 16.27, + "learning_rate": 3.375379017586417e-05, + "loss": 1.0165, + "step": 188030 + }, + { + "epoch": 16.28, + "learning_rate": 3.3752923849952354e-05, + "loss": 1.0686, + "step": 188040 + }, + { + "epoch": 16.28, + "learning_rate": 3.375205752404055e-05, + "loss": 0.9928, + "step": 188050 + }, + { + "epoch": 16.28, + "learning_rate": 3.375119119812874e-05, + "loss": 1.0488, + "step": 188060 + }, + { + "epoch": 16.28, + "learning_rate": 3.375032487221693e-05, + "loss": 1.0003, + "step": 188070 + }, + { + "epoch": 16.28, + "learning_rate": 3.3749458546305125e-05, + "loss": 0.9814, + "step": 188080 + }, + { + "epoch": 16.28, + "learning_rate": 3.374859222039331e-05, + "loss": 1.034, + "step": 188090 + }, + { + "epoch": 16.28, + "learning_rate": 3.374772589448151e-05, + "loss": 1.0372, + "step": 188100 + }, + { + "epoch": 16.28, + "learning_rate": 3.37468595685697e-05, + "loss": 1.0371, + "step": 188110 + }, + { + "epoch": 16.28, + "learning_rate": 3.374599324265789e-05, + "loss": 1.0202, + "step": 188120 + }, + { + "epoch": 16.28, + "learning_rate": 3.3745126916746084e-05, + "loss": 1.0033, + "step": 188130 + }, + { + "epoch": 16.28, + "learning_rate": 3.374426059083428e-05, + "loss": 1.054, + "step": 188140 + }, + { + "epoch": 16.29, + "learning_rate": 3.3743394264922466e-05, + "loss": 0.9547, + "step": 188150 + }, + { + "epoch": 16.29, + "learning_rate": 3.374252793901066e-05, + "loss": 1.0422, + "step": 188160 + }, + { + "epoch": 16.29, + "learning_rate": 3.374166161309885e-05, + "loss": 0.9627, + "step": 188170 + }, + { + "epoch": 16.29, + "learning_rate": 3.374079528718704e-05, + "loss": 0.9853, + "step": 188180 + }, + { + "epoch": 16.29, + "learning_rate": 3.373992896127524e-05, + "loss": 1.0284, + "step": 188190 + }, + { + "epoch": 16.29, + "learning_rate": 3.3739062635363425e-05, + "loss": 1.0224, + "step": 188200 + }, + { + "epoch": 16.29, + "learning_rate": 3.373819630945162e-05, + "loss": 0.9623, + "step": 188210 + }, + { + "epoch": 16.29, + "learning_rate": 3.3737329983539814e-05, + "loss": 1.0406, + "step": 188220 + }, + { + "epoch": 16.29, + "learning_rate": 3.3736463657628e-05, + "loss": 1.0534, + "step": 188230 + }, + { + "epoch": 16.29, + "learning_rate": 3.3735597331716196e-05, + "loss": 1.0284, + "step": 188240 + }, + { + "epoch": 16.29, + "learning_rate": 3.373473100580439e-05, + "loss": 1.0118, + "step": 188250 + }, + { + "epoch": 16.29, + "learning_rate": 3.373386467989258e-05, + "loss": 1.0756, + "step": 188260 + }, + { + "epoch": 16.3, + "learning_rate": 3.373299835398077e-05, + "loss": 1.0623, + "step": 188270 + }, + { + "epoch": 16.3, + "learning_rate": 3.373213202806896e-05, + "loss": 0.964, + "step": 188280 + }, + { + "epoch": 16.3, + "learning_rate": 3.3731265702157155e-05, + "loss": 1.041, + "step": 188290 + }, + { + "epoch": 16.3, + "learning_rate": 3.373039937624535e-05, + "loss": 1.0069, + "step": 188300 + }, + { + "epoch": 16.3, + "learning_rate": 3.372953305033354e-05, + "loss": 0.9968, + "step": 188310 + }, + { + "epoch": 16.3, + "learning_rate": 3.372866672442173e-05, + "loss": 1.0819, + "step": 188320 + }, + { + "epoch": 16.3, + "learning_rate": 3.3727800398509926e-05, + "loss": 1.0059, + "step": 188330 + }, + { + "epoch": 16.3, + "learning_rate": 3.372693407259811e-05, + "loss": 1.0348, + "step": 188340 + }, + { + "epoch": 16.3, + "learning_rate": 3.372606774668631e-05, + "loss": 1.0415, + "step": 188350 + }, + { + "epoch": 16.3, + "learning_rate": 3.37252014207745e-05, + "loss": 1.0321, + "step": 188360 + }, + { + "epoch": 16.3, + "learning_rate": 3.372433509486269e-05, + "loss": 0.9998, + "step": 188370 + }, + { + "epoch": 16.31, + "learning_rate": 3.3723468768950884e-05, + "loss": 0.9981, + "step": 188380 + }, + { + "epoch": 16.31, + "learning_rate": 3.372260244303907e-05, + "loss": 1.0131, + "step": 188390 + }, + { + "epoch": 16.31, + "learning_rate": 3.3721736117127267e-05, + "loss": 0.9894, + "step": 188400 + }, + { + "epoch": 16.31, + "learning_rate": 3.372086979121546e-05, + "loss": 1.0432, + "step": 188410 + }, + { + "epoch": 16.31, + "learning_rate": 3.372000346530365e-05, + "loss": 1.0326, + "step": 188420 + }, + { + "epoch": 16.31, + "learning_rate": 3.371913713939184e-05, + "loss": 1.0233, + "step": 188430 + }, + { + "epoch": 16.31, + "learning_rate": 3.371827081348004e-05, + "loss": 1.0007, + "step": 188440 + }, + { + "epoch": 16.31, + "learning_rate": 3.3717404487568225e-05, + "loss": 1.0234, + "step": 188450 + }, + { + "epoch": 16.31, + "learning_rate": 3.371653816165642e-05, + "loss": 1.0349, + "step": 188460 + }, + { + "epoch": 16.31, + "learning_rate": 3.3715671835744614e-05, + "loss": 1.0315, + "step": 188470 + }, + { + "epoch": 16.31, + "learning_rate": 3.37148055098328e-05, + "loss": 1.0084, + "step": 188480 + }, + { + "epoch": 16.31, + "learning_rate": 3.3713939183920996e-05, + "loss": 0.9963, + "step": 188490 + }, + { + "epoch": 16.32, + "learning_rate": 3.3713072858009184e-05, + "loss": 1.0496, + "step": 188500 + }, + { + "epoch": 16.32, + "learning_rate": 3.371220653209738e-05, + "loss": 1.0169, + "step": 188510 + }, + { + "epoch": 16.32, + "learning_rate": 3.371134020618557e-05, + "loss": 1.0387, + "step": 188520 + }, + { + "epoch": 16.32, + "learning_rate": 3.371047388027376e-05, + "loss": 1.0035, + "step": 188530 + }, + { + "epoch": 16.32, + "learning_rate": 3.3709607554361955e-05, + "loss": 1.0557, + "step": 188540 + }, + { + "epoch": 16.32, + "learning_rate": 3.370874122845015e-05, + "loss": 1.0069, + "step": 188550 + }, + { + "epoch": 16.32, + "learning_rate": 3.370787490253834e-05, + "loss": 1.0005, + "step": 188560 + }, + { + "epoch": 16.32, + "learning_rate": 3.370700857662653e-05, + "loss": 1.0494, + "step": 188570 + }, + { + "epoch": 16.32, + "learning_rate": 3.3706142250714726e-05, + "loss": 0.9448, + "step": 188580 + }, + { + "epoch": 16.32, + "learning_rate": 3.3705275924802914e-05, + "loss": 1.0259, + "step": 188590 + }, + { + "epoch": 16.32, + "learning_rate": 3.370440959889111e-05, + "loss": 1.0084, + "step": 188600 + }, + { + "epoch": 16.33, + "learning_rate": 3.3703543272979296e-05, + "loss": 0.9939, + "step": 188610 + }, + { + "epoch": 16.33, + "learning_rate": 3.370267694706749e-05, + "loss": 1.0327, + "step": 188620 + }, + { + "epoch": 16.33, + "learning_rate": 3.3701810621155685e-05, + "loss": 1.0852, + "step": 188630 + }, + { + "epoch": 16.33, + "learning_rate": 3.370094429524387e-05, + "loss": 1.0329, + "step": 188640 + }, + { + "epoch": 16.33, + "learning_rate": 3.370007796933207e-05, + "loss": 0.9837, + "step": 188650 + }, + { + "epoch": 16.33, + "learning_rate": 3.369921164342026e-05, + "loss": 1.0758, + "step": 188660 + }, + { + "epoch": 16.33, + "learning_rate": 3.369834531750845e-05, + "loss": 1.0414, + "step": 188670 + }, + { + "epoch": 16.33, + "learning_rate": 3.3697478991596643e-05, + "loss": 1.0179, + "step": 188680 + }, + { + "epoch": 16.33, + "learning_rate": 3.369661266568484e-05, + "loss": 1.0325, + "step": 188690 + }, + { + "epoch": 16.33, + "learning_rate": 3.3695746339773026e-05, + "loss": 0.9464, + "step": 188700 + }, + { + "epoch": 16.33, + "learning_rate": 3.369488001386122e-05, + "loss": 1.0028, + "step": 188710 + }, + { + "epoch": 16.33, + "learning_rate": 3.369401368794941e-05, + "loss": 1.007, + "step": 188720 + }, + { + "epoch": 16.34, + "learning_rate": 3.36931473620376e-05, + "loss": 1.0316, + "step": 188730 + }, + { + "epoch": 16.34, + "learning_rate": 3.36922810361258e-05, + "loss": 1.0088, + "step": 188740 + }, + { + "epoch": 16.34, + "learning_rate": 3.3691414710213984e-05, + "loss": 1.0066, + "step": 188750 + }, + { + "epoch": 16.34, + "learning_rate": 3.369054838430218e-05, + "loss": 0.9597, + "step": 188760 + }, + { + "epoch": 16.34, + "learning_rate": 3.368968205839037e-05, + "loss": 0.9827, + "step": 188770 + }, + { + "epoch": 16.34, + "learning_rate": 3.368881573247856e-05, + "loss": 0.9919, + "step": 188780 + }, + { + "epoch": 16.34, + "learning_rate": 3.3687949406566755e-05, + "loss": 1.0209, + "step": 188790 + }, + { + "epoch": 16.34, + "learning_rate": 3.368708308065495e-05, + "loss": 1.0599, + "step": 188800 + }, + { + "epoch": 16.34, + "learning_rate": 3.368621675474314e-05, + "loss": 1.0136, + "step": 188810 + }, + { + "epoch": 16.34, + "learning_rate": 3.368535042883133e-05, + "loss": 1.0093, + "step": 188820 + }, + { + "epoch": 16.34, + "learning_rate": 3.368448410291952e-05, + "loss": 1.0179, + "step": 188830 + }, + { + "epoch": 16.35, + "learning_rate": 3.3683617777007714e-05, + "loss": 0.9784, + "step": 188840 + }, + { + "epoch": 16.35, + "learning_rate": 3.368275145109591e-05, + "loss": 1.0321, + "step": 188850 + }, + { + "epoch": 16.35, + "learning_rate": 3.3681885125184096e-05, + "loss": 0.9864, + "step": 188860 + }, + { + "epoch": 16.35, + "learning_rate": 3.368101879927229e-05, + "loss": 1.0125, + "step": 188870 + }, + { + "epoch": 16.35, + "learning_rate": 3.3680152473360485e-05, + "loss": 1.0045, + "step": 188880 + }, + { + "epoch": 16.35, + "learning_rate": 3.367928614744867e-05, + "loss": 1.0501, + "step": 188890 + }, + { + "epoch": 16.35, + "learning_rate": 3.367841982153687e-05, + "loss": 1.0107, + "step": 188900 + }, + { + "epoch": 16.35, + "learning_rate": 3.3677553495625055e-05, + "loss": 1.0333, + "step": 188910 + }, + { + "epoch": 16.35, + "learning_rate": 3.367668716971325e-05, + "loss": 0.9968, + "step": 188920 + }, + { + "epoch": 16.35, + "learning_rate": 3.3675820843801444e-05, + "loss": 1.0672, + "step": 188930 + }, + { + "epoch": 16.35, + "learning_rate": 3.367495451788963e-05, + "loss": 0.9284, + "step": 188940 + }, + { + "epoch": 16.35, + "learning_rate": 3.3674088191977826e-05, + "loss": 1.0109, + "step": 188950 + }, + { + "epoch": 16.36, + "learning_rate": 3.367322186606602e-05, + "loss": 1.0435, + "step": 188960 + }, + { + "epoch": 16.36, + "learning_rate": 3.367235554015421e-05, + "loss": 1.0136, + "step": 188970 + }, + { + "epoch": 16.36, + "learning_rate": 3.36714892142424e-05, + "loss": 1.0383, + "step": 188980 + }, + { + "epoch": 16.36, + "learning_rate": 3.36706228883306e-05, + "loss": 1.028, + "step": 188990 + }, + { + "epoch": 16.36, + "learning_rate": 3.3669756562418785e-05, + "loss": 1.0505, + "step": 189000 + }, + { + "epoch": 16.36, + "learning_rate": 3.366889023650698e-05, + "loss": 0.979, + "step": 189010 + }, + { + "epoch": 16.36, + "learning_rate": 3.366802391059517e-05, + "loss": 1.0243, + "step": 189020 + }, + { + "epoch": 16.36, + "learning_rate": 3.366715758468336e-05, + "loss": 0.9782, + "step": 189030 + }, + { + "epoch": 16.36, + "learning_rate": 3.3666291258771556e-05, + "loss": 1.0372, + "step": 189040 + }, + { + "epoch": 16.36, + "learning_rate": 3.366542493285974e-05, + "loss": 0.9725, + "step": 189050 + }, + { + "epoch": 16.36, + "learning_rate": 3.366455860694794e-05, + "loss": 0.9946, + "step": 189060 + }, + { + "epoch": 16.36, + "learning_rate": 3.366369228103613e-05, + "loss": 1.0206, + "step": 189070 + }, + { + "epoch": 16.37, + "learning_rate": 3.366282595512432e-05, + "loss": 1.0465, + "step": 189080 + }, + { + "epoch": 16.37, + "learning_rate": 3.3661959629212514e-05, + "loss": 0.9695, + "step": 189090 + }, + { + "epoch": 16.37, + "learning_rate": 3.366109330330071e-05, + "loss": 1.0361, + "step": 189100 + }, + { + "epoch": 16.37, + "learning_rate": 3.3660226977388897e-05, + "loss": 1.0194, + "step": 189110 + }, + { + "epoch": 16.37, + "learning_rate": 3.365936065147709e-05, + "loss": 1.0177, + "step": 189120 + }, + { + "epoch": 16.37, + "learning_rate": 3.365849432556528e-05, + "loss": 1.0216, + "step": 189130 + }, + { + "epoch": 16.37, + "learning_rate": 3.365762799965347e-05, + "loss": 1.032, + "step": 189140 + }, + { + "epoch": 16.37, + "learning_rate": 3.365676167374167e-05, + "loss": 1.0105, + "step": 189150 + }, + { + "epoch": 16.37, + "learning_rate": 3.3655895347829855e-05, + "loss": 1.0297, + "step": 189160 + }, + { + "epoch": 16.37, + "learning_rate": 3.365502902191805e-05, + "loss": 1.0102, + "step": 189170 + }, + { + "epoch": 16.37, + "learning_rate": 3.3654162696006244e-05, + "loss": 1.0854, + "step": 189180 + }, + { + "epoch": 16.38, + "learning_rate": 3.365329637009443e-05, + "loss": 1.0354, + "step": 189190 + }, + { + "epoch": 16.38, + "learning_rate": 3.3652430044182626e-05, + "loss": 0.9848, + "step": 189200 + }, + { + "epoch": 16.38, + "learning_rate": 3.365156371827082e-05, + "loss": 1.0293, + "step": 189210 + }, + { + "epoch": 16.38, + "learning_rate": 3.365069739235901e-05, + "loss": 1.0419, + "step": 189220 + }, + { + "epoch": 16.38, + "learning_rate": 3.36498310664472e-05, + "loss": 0.9969, + "step": 189230 + }, + { + "epoch": 16.38, + "learning_rate": 3.364896474053539e-05, + "loss": 1.051, + "step": 189240 + }, + { + "epoch": 16.38, + "learning_rate": 3.3648098414623585e-05, + "loss": 1.0092, + "step": 189250 + }, + { + "epoch": 16.38, + "learning_rate": 3.364723208871178e-05, + "loss": 0.9807, + "step": 189260 + }, + { + "epoch": 16.38, + "learning_rate": 3.364636576279997e-05, + "loss": 1.059, + "step": 189270 + }, + { + "epoch": 16.38, + "learning_rate": 3.364549943688816e-05, + "loss": 0.9916, + "step": 189280 + }, + { + "epoch": 16.38, + "learning_rate": 3.3644633110976356e-05, + "loss": 1.0687, + "step": 189290 + }, + { + "epoch": 16.38, + "learning_rate": 3.3643766785064544e-05, + "loss": 0.9852, + "step": 189300 + }, + { + "epoch": 16.39, + "learning_rate": 3.364290045915274e-05, + "loss": 1.0089, + "step": 189310 + }, + { + "epoch": 16.39, + "learning_rate": 3.364203413324093e-05, + "loss": 0.9803, + "step": 189320 + }, + { + "epoch": 16.39, + "learning_rate": 3.364116780732912e-05, + "loss": 0.9531, + "step": 189330 + }, + { + "epoch": 16.39, + "learning_rate": 3.3640301481417315e-05, + "loss": 1.0094, + "step": 189340 + }, + { + "epoch": 16.39, + "learning_rate": 3.36394351555055e-05, + "loss": 1.0179, + "step": 189350 + }, + { + "epoch": 16.39, + "learning_rate": 3.36385688295937e-05, + "loss": 1.0028, + "step": 189360 + }, + { + "epoch": 16.39, + "learning_rate": 3.363770250368189e-05, + "loss": 1.0185, + "step": 189370 + }, + { + "epoch": 16.39, + "learning_rate": 3.363683617777008e-05, + "loss": 0.983, + "step": 189380 + }, + { + "epoch": 16.39, + "learning_rate": 3.3635969851858273e-05, + "loss": 1.107, + "step": 189390 + }, + { + "epoch": 16.39, + "learning_rate": 3.363510352594647e-05, + "loss": 0.9723, + "step": 189400 + }, + { + "epoch": 16.39, + "learning_rate": 3.3634237200034656e-05, + "loss": 0.9772, + "step": 189410 + }, + { + "epoch": 16.4, + "learning_rate": 3.363337087412285e-05, + "loss": 1.0161, + "step": 189420 + }, + { + "epoch": 16.4, + "learning_rate": 3.3632504548211044e-05, + "loss": 0.9783, + "step": 189430 + }, + { + "epoch": 16.4, + "learning_rate": 3.363163822229923e-05, + "loss": 1.0025, + "step": 189440 + }, + { + "epoch": 16.4, + "learning_rate": 3.3630771896387427e-05, + "loss": 1.0334, + "step": 189450 + }, + { + "epoch": 16.4, + "learning_rate": 3.3629905570475614e-05, + "loss": 1.0342, + "step": 189460 + }, + { + "epoch": 16.4, + "learning_rate": 3.362903924456381e-05, + "loss": 1.0192, + "step": 189470 + }, + { + "epoch": 16.4, + "learning_rate": 3.3628172918652e-05, + "loss": 0.9928, + "step": 189480 + }, + { + "epoch": 16.4, + "learning_rate": 3.362730659274019e-05, + "loss": 1.0453, + "step": 189490 + }, + { + "epoch": 16.4, + "learning_rate": 3.3626440266828385e-05, + "loss": 0.9768, + "step": 189500 + }, + { + "epoch": 16.4, + "learning_rate": 3.362557394091658e-05, + "loss": 1.0375, + "step": 189510 + }, + { + "epoch": 16.4, + "learning_rate": 3.362470761500477e-05, + "loss": 1.0176, + "step": 189520 + }, + { + "epoch": 16.4, + "learning_rate": 3.362384128909296e-05, + "loss": 1.018, + "step": 189530 + }, + { + "epoch": 16.41, + "learning_rate": 3.3622974963181156e-05, + "loss": 1.0245, + "step": 189540 + }, + { + "epoch": 16.41, + "learning_rate": 3.3622108637269344e-05, + "loss": 1.0215, + "step": 189550 + }, + { + "epoch": 16.41, + "learning_rate": 3.362124231135754e-05, + "loss": 1.0809, + "step": 189560 + }, + { + "epoch": 16.41, + "learning_rate": 3.3620375985445726e-05, + "loss": 1.0027, + "step": 189570 + }, + { + "epoch": 16.41, + "learning_rate": 3.361950965953392e-05, + "loss": 1.0353, + "step": 189580 + }, + { + "epoch": 16.41, + "learning_rate": 3.3618643333622115e-05, + "loss": 0.9994, + "step": 189590 + }, + { + "epoch": 16.41, + "learning_rate": 3.36177770077103e-05, + "loss": 1.027, + "step": 189600 + }, + { + "epoch": 16.41, + "learning_rate": 3.36169106817985e-05, + "loss": 1.0467, + "step": 189610 + }, + { + "epoch": 16.41, + "learning_rate": 3.361604435588669e-05, + "loss": 1.0205, + "step": 189620 + }, + { + "epoch": 16.41, + "learning_rate": 3.361517802997488e-05, + "loss": 1.0556, + "step": 189630 + }, + { + "epoch": 16.41, + "learning_rate": 3.3614311704063074e-05, + "loss": 1.0841, + "step": 189640 + }, + { + "epoch": 16.42, + "learning_rate": 3.361344537815127e-05, + "loss": 0.9892, + "step": 189650 + }, + { + "epoch": 16.42, + "learning_rate": 3.3612579052239456e-05, + "loss": 1.032, + "step": 189660 + }, + { + "epoch": 16.42, + "learning_rate": 3.361171272632765e-05, + "loss": 1.0409, + "step": 189670 + }, + { + "epoch": 16.42, + "learning_rate": 3.361084640041584e-05, + "loss": 1.0622, + "step": 189680 + }, + { + "epoch": 16.42, + "learning_rate": 3.360998007450403e-05, + "loss": 1.0184, + "step": 189690 + }, + { + "epoch": 16.42, + "learning_rate": 3.360911374859223e-05, + "loss": 1.0366, + "step": 189700 + }, + { + "epoch": 16.42, + "learning_rate": 3.3608247422680415e-05, + "loss": 1.0203, + "step": 189710 + }, + { + "epoch": 16.42, + "learning_rate": 3.360738109676861e-05, + "loss": 0.9929, + "step": 189720 + }, + { + "epoch": 16.42, + "learning_rate": 3.3606514770856804e-05, + "loss": 1.0404, + "step": 189730 + }, + { + "epoch": 16.42, + "learning_rate": 3.360564844494499e-05, + "loss": 1.0085, + "step": 189740 + }, + { + "epoch": 16.42, + "learning_rate": 3.3604782119033186e-05, + "loss": 1.0736, + "step": 189750 + }, + { + "epoch": 16.42, + "learning_rate": 3.360391579312137e-05, + "loss": 0.9877, + "step": 189760 + }, + { + "epoch": 16.43, + "learning_rate": 3.360304946720957e-05, + "loss": 1.0105, + "step": 189770 + }, + { + "epoch": 16.43, + "learning_rate": 3.360218314129776e-05, + "loss": 1.026, + "step": 189780 + }, + { + "epoch": 16.43, + "learning_rate": 3.360131681538595e-05, + "loss": 1.0219, + "step": 189790 + }, + { + "epoch": 16.43, + "learning_rate": 3.3600450489474144e-05, + "loss": 1.0353, + "step": 189800 + }, + { + "epoch": 16.43, + "learning_rate": 3.359958416356234e-05, + "loss": 1.0488, + "step": 189810 + }, + { + "epoch": 16.43, + "learning_rate": 3.3598717837650526e-05, + "loss": 1.066, + "step": 189820 + }, + { + "epoch": 16.43, + "learning_rate": 3.359785151173872e-05, + "loss": 0.9869, + "step": 189830 + }, + { + "epoch": 16.43, + "learning_rate": 3.3596985185826915e-05, + "loss": 1.0498, + "step": 189840 + }, + { + "epoch": 16.43, + "learning_rate": 3.35961188599151e-05, + "loss": 1.0299, + "step": 189850 + }, + { + "epoch": 16.43, + "learning_rate": 3.35952525340033e-05, + "loss": 1.0133, + "step": 189860 + }, + { + "epoch": 16.43, + "learning_rate": 3.3594386208091485e-05, + "loss": 0.9979, + "step": 189870 + }, + { + "epoch": 16.44, + "learning_rate": 3.359351988217967e-05, + "loss": 1.0702, + "step": 189880 + }, + { + "epoch": 16.44, + "learning_rate": 3.359265355626787e-05, + "loss": 1.0019, + "step": 189890 + }, + { + "epoch": 16.44, + "learning_rate": 3.359178723035606e-05, + "loss": 1.0097, + "step": 189900 + }, + { + "epoch": 16.44, + "learning_rate": 3.359092090444425e-05, + "loss": 1.0521, + "step": 189910 + }, + { + "epoch": 16.44, + "learning_rate": 3.3590054578532444e-05, + "loss": 0.9898, + "step": 189920 + }, + { + "epoch": 16.44, + "learning_rate": 3.358918825262063e-05, + "loss": 1.0303, + "step": 189930 + }, + { + "epoch": 16.44, + "learning_rate": 3.3588321926708826e-05, + "loss": 1.0198, + "step": 189940 + }, + { + "epoch": 16.44, + "learning_rate": 3.358745560079702e-05, + "loss": 0.9775, + "step": 189950 + }, + { + "epoch": 16.44, + "learning_rate": 3.358658927488521e-05, + "loss": 0.9957, + "step": 189960 + }, + { + "epoch": 16.44, + "learning_rate": 3.35857229489734e-05, + "loss": 1.0312, + "step": 189970 + }, + { + "epoch": 16.44, + "learning_rate": 3.35848566230616e-05, + "loss": 1.0261, + "step": 189980 + }, + { + "epoch": 16.44, + "learning_rate": 3.3583990297149785e-05, + "loss": 0.9975, + "step": 189990 + }, + { + "epoch": 16.45, + "learning_rate": 3.358312397123798e-05, + "loss": 0.9847, + "step": 190000 + }, + { + "epoch": 16.45, + "learning_rate": 3.3582257645326174e-05, + "loss": 1.0523, + "step": 190010 + }, + { + "epoch": 16.45, + "learning_rate": 3.358139131941436e-05, + "loss": 1.0065, + "step": 190020 + }, + { + "epoch": 16.45, + "learning_rate": 3.3580524993502556e-05, + "loss": 1.0363, + "step": 190030 + }, + { + "epoch": 16.45, + "learning_rate": 3.3579658667590743e-05, + "loss": 1.0169, + "step": 190040 + }, + { + "epoch": 16.45, + "learning_rate": 3.357879234167894e-05, + "loss": 1.0064, + "step": 190050 + }, + { + "epoch": 16.45, + "learning_rate": 3.357792601576713e-05, + "loss": 1.0007, + "step": 190060 + }, + { + "epoch": 16.45, + "learning_rate": 3.357705968985532e-05, + "loss": 1.035, + "step": 190070 + }, + { + "epoch": 16.45, + "learning_rate": 3.3576193363943514e-05, + "loss": 1.01, + "step": 190080 + }, + { + "epoch": 16.45, + "learning_rate": 3.357532703803171e-05, + "loss": 1.0482, + "step": 190090 + }, + { + "epoch": 16.45, + "learning_rate": 3.35744607121199e-05, + "loss": 0.9785, + "step": 190100 + }, + { + "epoch": 16.46, + "learning_rate": 3.357359438620809e-05, + "loss": 1.0391, + "step": 190110 + }, + { + "epoch": 16.46, + "learning_rate": 3.3572728060296286e-05, + "loss": 1.0285, + "step": 190120 + }, + { + "epoch": 16.46, + "learning_rate": 3.357186173438447e-05, + "loss": 0.9834, + "step": 190130 + }, + { + "epoch": 16.46, + "learning_rate": 3.357099540847267e-05, + "loss": 1.014, + "step": 190140 + }, + { + "epoch": 16.46, + "learning_rate": 3.3570129082560855e-05, + "loss": 1.0433, + "step": 190150 + }, + { + "epoch": 16.46, + "learning_rate": 3.356926275664905e-05, + "loss": 1.0053, + "step": 190160 + }, + { + "epoch": 16.46, + "learning_rate": 3.3568396430737244e-05, + "loss": 0.9764, + "step": 190170 + }, + { + "epoch": 16.46, + "learning_rate": 3.356753010482543e-05, + "loss": 1.016, + "step": 190180 + }, + { + "epoch": 16.46, + "learning_rate": 3.3566663778913626e-05, + "loss": 1.0045, + "step": 190190 + }, + { + "epoch": 16.46, + "learning_rate": 3.356579745300182e-05, + "loss": 0.9894, + "step": 190200 + }, + { + "epoch": 16.46, + "learning_rate": 3.356493112709001e-05, + "loss": 1.0119, + "step": 190210 + }, + { + "epoch": 16.46, + "learning_rate": 3.35640648011782e-05, + "loss": 0.9934, + "step": 190220 + }, + { + "epoch": 16.47, + "learning_rate": 3.35631984752664e-05, + "loss": 0.995, + "step": 190230 + }, + { + "epoch": 16.47, + "learning_rate": 3.3562332149354585e-05, + "loss": 1.0392, + "step": 190240 + }, + { + "epoch": 16.47, + "learning_rate": 3.356146582344278e-05, + "loss": 0.9926, + "step": 190250 + }, + { + "epoch": 16.47, + "learning_rate": 3.356059949753097e-05, + "loss": 1.0554, + "step": 190260 + }, + { + "epoch": 16.47, + "learning_rate": 3.355973317161916e-05, + "loss": 1.0441, + "step": 190270 + }, + { + "epoch": 16.47, + "learning_rate": 3.3558866845707356e-05, + "loss": 1.0184, + "step": 190280 + }, + { + "epoch": 16.47, + "learning_rate": 3.3558000519795544e-05, + "loss": 1.0246, + "step": 190290 + }, + { + "epoch": 16.47, + "learning_rate": 3.355713419388374e-05, + "loss": 1.0154, + "step": 190300 + }, + { + "epoch": 16.47, + "learning_rate": 3.355626786797193e-05, + "loss": 1.0826, + "step": 190310 + }, + { + "epoch": 16.47, + "learning_rate": 3.355540154206012e-05, + "loss": 1.0302, + "step": 190320 + }, + { + "epoch": 16.47, + "learning_rate": 3.3554535216148315e-05, + "loss": 1.013, + "step": 190330 + }, + { + "epoch": 16.47, + "learning_rate": 3.355366889023651e-05, + "loss": 1.035, + "step": 190340 + }, + { + "epoch": 16.48, + "learning_rate": 3.35528025643247e-05, + "loss": 1.0231, + "step": 190350 + }, + { + "epoch": 16.48, + "learning_rate": 3.355193623841289e-05, + "loss": 1.0253, + "step": 190360 + }, + { + "epoch": 16.48, + "learning_rate": 3.355106991250108e-05, + "loss": 1.0453, + "step": 190370 + }, + { + "epoch": 16.48, + "learning_rate": 3.3550203586589274e-05, + "loss": 1.0066, + "step": 190380 + }, + { + "epoch": 16.48, + "learning_rate": 3.354933726067747e-05, + "loss": 1.0778, + "step": 190390 + }, + { + "epoch": 16.48, + "learning_rate": 3.3548470934765656e-05, + "loss": 1.0419, + "step": 190400 + }, + { + "epoch": 16.48, + "learning_rate": 3.354760460885385e-05, + "loss": 1.0385, + "step": 190410 + }, + { + "epoch": 16.48, + "learning_rate": 3.3546738282942045e-05, + "loss": 1.0316, + "step": 190420 + }, + { + "epoch": 16.48, + "learning_rate": 3.354587195703023e-05, + "loss": 1.0562, + "step": 190430 + }, + { + "epoch": 16.48, + "learning_rate": 3.354500563111843e-05, + "loss": 1.0246, + "step": 190440 + }, + { + "epoch": 16.48, + "learning_rate": 3.3544139305206614e-05, + "loss": 0.9776, + "step": 190450 + }, + { + "epoch": 16.49, + "learning_rate": 3.354327297929481e-05, + "loss": 0.9947, + "step": 190460 + }, + { + "epoch": 16.49, + "learning_rate": 3.3542406653383e-05, + "loss": 1.0378, + "step": 190470 + }, + { + "epoch": 16.49, + "learning_rate": 3.354154032747119e-05, + "loss": 1.0289, + "step": 190480 + }, + { + "epoch": 16.49, + "learning_rate": 3.3540674001559385e-05, + "loss": 1.0641, + "step": 190490 + }, + { + "epoch": 16.49, + "learning_rate": 3.353980767564758e-05, + "loss": 1.0282, + "step": 190500 + }, + { + "epoch": 16.49, + "learning_rate": 3.353894134973577e-05, + "loss": 0.9873, + "step": 190510 + }, + { + "epoch": 16.49, + "learning_rate": 3.353807502382396e-05, + "loss": 1.0332, + "step": 190520 + }, + { + "epoch": 16.49, + "learning_rate": 3.3537208697912156e-05, + "loss": 0.9846, + "step": 190530 + }, + { + "epoch": 16.49, + "learning_rate": 3.3536342372000344e-05, + "loss": 0.9832, + "step": 190540 + }, + { + "epoch": 16.49, + "learning_rate": 3.353547604608854e-05, + "loss": 1.0139, + "step": 190550 + }, + { + "epoch": 16.49, + "learning_rate": 3.3534609720176726e-05, + "loss": 1.009, + "step": 190560 + }, + { + "epoch": 16.49, + "learning_rate": 3.353374339426492e-05, + "loss": 1.0412, + "step": 190570 + }, + { + "epoch": 16.5, + "learning_rate": 3.3532877068353115e-05, + "loss": 1.0268, + "step": 190580 + }, + { + "epoch": 16.5, + "learning_rate": 3.35320107424413e-05, + "loss": 0.9883, + "step": 190590 + }, + { + "epoch": 16.5, + "learning_rate": 3.35311444165295e-05, + "loss": 1.0254, + "step": 190600 + }, + { + "epoch": 16.5, + "learning_rate": 3.353027809061769e-05, + "loss": 0.9832, + "step": 190610 + }, + { + "epoch": 16.5, + "learning_rate": 3.352941176470588e-05, + "loss": 1.0093, + "step": 190620 + }, + { + "epoch": 16.5, + "learning_rate": 3.3528545438794074e-05, + "loss": 1.0511, + "step": 190630 + }, + { + "epoch": 16.5, + "learning_rate": 3.352767911288227e-05, + "loss": 1.0192, + "step": 190640 + }, + { + "epoch": 16.5, + "learning_rate": 3.3526812786970456e-05, + "loss": 1.026, + "step": 190650 + }, + { + "epoch": 16.5, + "learning_rate": 3.352594646105865e-05, + "loss": 1.012, + "step": 190660 + }, + { + "epoch": 16.5, + "learning_rate": 3.352508013514684e-05, + "loss": 1.0743, + "step": 190670 + }, + { + "epoch": 16.5, + "learning_rate": 3.352421380923503e-05, + "loss": 1.0199, + "step": 190680 + }, + { + "epoch": 16.51, + "learning_rate": 3.352334748332323e-05, + "loss": 1.0722, + "step": 190690 + }, + { + "epoch": 16.51, + "learning_rate": 3.3522481157411415e-05, + "loss": 0.9949, + "step": 190700 + }, + { + "epoch": 16.51, + "learning_rate": 3.352161483149961e-05, + "loss": 1.038, + "step": 190710 + }, + { + "epoch": 16.51, + "learning_rate": 3.3520748505587804e-05, + "loss": 1.0749, + "step": 190720 + }, + { + "epoch": 16.51, + "learning_rate": 3.351988217967599e-05, + "loss": 1.0409, + "step": 190730 + }, + { + "epoch": 16.51, + "learning_rate": 3.3519015853764186e-05, + "loss": 1.0054, + "step": 190740 + }, + { + "epoch": 16.51, + "learning_rate": 3.351814952785238e-05, + "loss": 1.0506, + "step": 190750 + }, + { + "epoch": 16.51, + "learning_rate": 3.351728320194057e-05, + "loss": 0.9846, + "step": 190760 + }, + { + "epoch": 16.51, + "learning_rate": 3.351641687602876e-05, + "loss": 1.0627, + "step": 190770 + }, + { + "epoch": 16.51, + "learning_rate": 3.351555055011695e-05, + "loss": 0.992, + "step": 190780 + }, + { + "epoch": 16.51, + "learning_rate": 3.3514684224205144e-05, + "loss": 1.0365, + "step": 190790 + }, + { + "epoch": 16.51, + "learning_rate": 3.351381789829334e-05, + "loss": 1.0318, + "step": 190800 + }, + { + "epoch": 16.52, + "learning_rate": 3.3512951572381527e-05, + "loss": 0.9582, + "step": 190810 + }, + { + "epoch": 16.52, + "learning_rate": 3.351208524646972e-05, + "loss": 1.0406, + "step": 190820 + }, + { + "epoch": 16.52, + "learning_rate": 3.3511218920557915e-05, + "loss": 1.0045, + "step": 190830 + }, + { + "epoch": 16.52, + "learning_rate": 3.35103525946461e-05, + "loss": 1.0091, + "step": 190840 + }, + { + "epoch": 16.52, + "learning_rate": 3.35094862687343e-05, + "loss": 1.0289, + "step": 190850 + }, + { + "epoch": 16.52, + "learning_rate": 3.350861994282249e-05, + "loss": 0.9906, + "step": 190860 + }, + { + "epoch": 16.52, + "learning_rate": 3.350775361691068e-05, + "loss": 1.0317, + "step": 190870 + }, + { + "epoch": 16.52, + "learning_rate": 3.3506887290998874e-05, + "loss": 0.9848, + "step": 190880 + }, + { + "epoch": 16.52, + "learning_rate": 3.350602096508706e-05, + "loss": 1.0296, + "step": 190890 + }, + { + "epoch": 16.52, + "learning_rate": 3.3505154639175256e-05, + "loss": 0.9813, + "step": 190900 + }, + { + "epoch": 16.52, + "learning_rate": 3.350428831326345e-05, + "loss": 1.0403, + "step": 190910 + }, + { + "epoch": 16.53, + "learning_rate": 3.350342198735164e-05, + "loss": 0.9963, + "step": 190920 + }, + { + "epoch": 16.53, + "learning_rate": 3.350255566143983e-05, + "loss": 0.9969, + "step": 190930 + }, + { + "epoch": 16.53, + "learning_rate": 3.350168933552803e-05, + "loss": 1.0347, + "step": 190940 + }, + { + "epoch": 16.53, + "learning_rate": 3.3500823009616215e-05, + "loss": 0.999, + "step": 190950 + }, + { + "epoch": 16.53, + "learning_rate": 3.349995668370441e-05, + "loss": 1.0028, + "step": 190960 + }, + { + "epoch": 16.53, + "learning_rate": 3.3499090357792604e-05, + "loss": 1.0625, + "step": 190970 + }, + { + "epoch": 16.53, + "learning_rate": 3.349822403188079e-05, + "loss": 1.0392, + "step": 190980 + }, + { + "epoch": 16.53, + "learning_rate": 3.3497357705968986e-05, + "loss": 0.9847, + "step": 190990 + }, + { + "epoch": 16.53, + "learning_rate": 3.3496491380057174e-05, + "loss": 1.0016, + "step": 191000 + }, + { + "epoch": 16.53, + "learning_rate": 3.349562505414537e-05, + "loss": 1.005, + "step": 191010 + }, + { + "epoch": 16.53, + "learning_rate": 3.349475872823356e-05, + "loss": 1.0321, + "step": 191020 + }, + { + "epoch": 16.53, + "learning_rate": 3.349389240232175e-05, + "loss": 0.9983, + "step": 191030 + }, + { + "epoch": 16.54, + "learning_rate": 3.3493026076409945e-05, + "loss": 0.9888, + "step": 191040 + }, + { + "epoch": 16.54, + "learning_rate": 3.349215975049814e-05, + "loss": 1.0258, + "step": 191050 + }, + { + "epoch": 16.54, + "learning_rate": 3.349129342458633e-05, + "loss": 0.9733, + "step": 191060 + }, + { + "epoch": 16.54, + "learning_rate": 3.349042709867452e-05, + "loss": 0.9816, + "step": 191070 + }, + { + "epoch": 16.54, + "learning_rate": 3.3489560772762716e-05, + "loss": 1.0449, + "step": 191080 + }, + { + "epoch": 16.54, + "learning_rate": 3.3488694446850904e-05, + "loss": 1.0246, + "step": 191090 + }, + { + "epoch": 16.54, + "learning_rate": 3.34878281209391e-05, + "loss": 0.9427, + "step": 191100 + }, + { + "epoch": 16.54, + "learning_rate": 3.3486961795027286e-05, + "loss": 0.9698, + "step": 191110 + }, + { + "epoch": 16.54, + "learning_rate": 3.348609546911548e-05, + "loss": 1.0519, + "step": 191120 + }, + { + "epoch": 16.54, + "learning_rate": 3.3485229143203675e-05, + "loss": 1.0157, + "step": 191130 + }, + { + "epoch": 16.54, + "learning_rate": 3.348436281729186e-05, + "loss": 1.0341, + "step": 191140 + }, + { + "epoch": 16.55, + "learning_rate": 3.348349649138006e-05, + "loss": 1.0102, + "step": 191150 + }, + { + "epoch": 16.55, + "learning_rate": 3.348263016546825e-05, + "loss": 1.0388, + "step": 191160 + }, + { + "epoch": 16.55, + "learning_rate": 3.348176383955644e-05, + "loss": 0.9832, + "step": 191170 + }, + { + "epoch": 16.55, + "learning_rate": 3.348089751364463e-05, + "loss": 1.0711, + "step": 191180 + }, + { + "epoch": 16.55, + "learning_rate": 3.348003118773283e-05, + "loss": 1.0215, + "step": 191190 + }, + { + "epoch": 16.55, + "learning_rate": 3.3479164861821015e-05, + "loss": 1.0436, + "step": 191200 + }, + { + "epoch": 16.55, + "learning_rate": 3.347829853590921e-05, + "loss": 1.0578, + "step": 191210 + }, + { + "epoch": 16.55, + "learning_rate": 3.34774322099974e-05, + "loss": 1.0986, + "step": 191220 + }, + { + "epoch": 16.55, + "learning_rate": 3.347656588408559e-05, + "loss": 1.0009, + "step": 191230 + }, + { + "epoch": 16.55, + "learning_rate": 3.3475699558173786e-05, + "loss": 1.0369, + "step": 191240 + }, + { + "epoch": 16.55, + "learning_rate": 3.3474833232261974e-05, + "loss": 0.9643, + "step": 191250 + }, + { + "epoch": 16.55, + "learning_rate": 3.347396690635017e-05, + "loss": 1.0292, + "step": 191260 + }, + { + "epoch": 16.56, + "learning_rate": 3.347310058043836e-05, + "loss": 1.047, + "step": 191270 + }, + { + "epoch": 16.56, + "learning_rate": 3.347223425452655e-05, + "loss": 1.0182, + "step": 191280 + }, + { + "epoch": 16.56, + "learning_rate": 3.3471367928614745e-05, + "loss": 1.0165, + "step": 191290 + }, + { + "epoch": 16.56, + "learning_rate": 3.347050160270293e-05, + "loss": 0.9901, + "step": 191300 + }, + { + "epoch": 16.56, + "learning_rate": 3.346963527679113e-05, + "loss": 0.9729, + "step": 191310 + }, + { + "epoch": 16.56, + "learning_rate": 3.346876895087932e-05, + "loss": 0.9665, + "step": 191320 + }, + { + "epoch": 16.56, + "learning_rate": 3.346790262496751e-05, + "loss": 1.0019, + "step": 191330 + }, + { + "epoch": 16.56, + "learning_rate": 3.3467036299055704e-05, + "loss": 0.9999, + "step": 191340 + }, + { + "epoch": 16.56, + "learning_rate": 3.34661699731439e-05, + "loss": 1.0433, + "step": 191350 + }, + { + "epoch": 16.56, + "learning_rate": 3.3465303647232086e-05, + "loss": 1.0172, + "step": 191360 + }, + { + "epoch": 16.56, + "learning_rate": 3.346443732132028e-05, + "loss": 1.0219, + "step": 191370 + }, + { + "epoch": 16.56, + "learning_rate": 3.3463570995408475e-05, + "loss": 0.9843, + "step": 191380 + }, + { + "epoch": 16.57, + "learning_rate": 3.346270466949666e-05, + "loss": 1.0718, + "step": 191390 + }, + { + "epoch": 16.57, + "learning_rate": 3.346183834358486e-05, + "loss": 0.9418, + "step": 191400 + }, + { + "epoch": 16.57, + "learning_rate": 3.3460972017673045e-05, + "loss": 1.0637, + "step": 191410 + }, + { + "epoch": 16.57, + "learning_rate": 3.346010569176124e-05, + "loss": 0.9934, + "step": 191420 + }, + { + "epoch": 16.57, + "learning_rate": 3.3459239365849434e-05, + "loss": 1.0385, + "step": 191430 + }, + { + "epoch": 16.57, + "learning_rate": 3.345837303993762e-05, + "loss": 1.0338, + "step": 191440 + }, + { + "epoch": 16.57, + "learning_rate": 3.3457506714025816e-05, + "loss": 0.9967, + "step": 191450 + }, + { + "epoch": 16.57, + "learning_rate": 3.345664038811401e-05, + "loss": 1.0327, + "step": 191460 + }, + { + "epoch": 16.57, + "learning_rate": 3.34557740622022e-05, + "loss": 0.962, + "step": 191470 + }, + { + "epoch": 16.57, + "learning_rate": 3.345490773629039e-05, + "loss": 0.9517, + "step": 191480 + }, + { + "epoch": 16.57, + "learning_rate": 3.345404141037859e-05, + "loss": 1.001, + "step": 191490 + }, + { + "epoch": 16.58, + "learning_rate": 3.3453175084466774e-05, + "loss": 1.0588, + "step": 191500 + }, + { + "epoch": 16.58, + "learning_rate": 3.345230875855497e-05, + "loss": 0.9851, + "step": 191510 + }, + { + "epoch": 16.58, + "learning_rate": 3.3451442432643157e-05, + "loss": 1.0297, + "step": 191520 + }, + { + "epoch": 16.58, + "learning_rate": 3.345057610673135e-05, + "loss": 1.0439, + "step": 191530 + }, + { + "epoch": 16.58, + "learning_rate": 3.3449709780819545e-05, + "loss": 0.9524, + "step": 191540 + }, + { + "epoch": 16.58, + "learning_rate": 3.344884345490773e-05, + "loss": 0.9833, + "step": 191550 + }, + { + "epoch": 16.58, + "learning_rate": 3.344797712899593e-05, + "loss": 1.0309, + "step": 191560 + }, + { + "epoch": 16.58, + "learning_rate": 3.344711080308412e-05, + "loss": 1.0517, + "step": 191570 + }, + { + "epoch": 16.58, + "learning_rate": 3.344624447717231e-05, + "loss": 1.0022, + "step": 191580 + }, + { + "epoch": 16.58, + "learning_rate": 3.3445378151260504e-05, + "loss": 1.0743, + "step": 191590 + }, + { + "epoch": 16.58, + "learning_rate": 3.34445118253487e-05, + "loss": 0.9952, + "step": 191600 + }, + { + "epoch": 16.58, + "learning_rate": 3.3443645499436886e-05, + "loss": 1.0629, + "step": 191610 + }, + { + "epoch": 16.59, + "learning_rate": 3.344277917352508e-05, + "loss": 0.9869, + "step": 191620 + }, + { + "epoch": 16.59, + "learning_rate": 3.344191284761327e-05, + "loss": 0.995, + "step": 191630 + }, + { + "epoch": 16.59, + "learning_rate": 3.344104652170146e-05, + "loss": 0.9999, + "step": 191640 + }, + { + "epoch": 16.59, + "learning_rate": 3.344018019578966e-05, + "loss": 1.0141, + "step": 191650 + }, + { + "epoch": 16.59, + "learning_rate": 3.3439313869877845e-05, + "loss": 0.9937, + "step": 191660 + }, + { + "epoch": 16.59, + "learning_rate": 3.343844754396604e-05, + "loss": 1.0239, + "step": 191670 + }, + { + "epoch": 16.59, + "learning_rate": 3.3437581218054234e-05, + "loss": 1.0308, + "step": 191680 + }, + { + "epoch": 16.59, + "learning_rate": 3.343671489214242e-05, + "loss": 0.9868, + "step": 191690 + }, + { + "epoch": 16.59, + "learning_rate": 3.3435848566230616e-05, + "loss": 1.0485, + "step": 191700 + }, + { + "epoch": 16.59, + "learning_rate": 3.343498224031881e-05, + "loss": 1.03, + "step": 191710 + }, + { + "epoch": 16.59, + "learning_rate": 3.3434115914407e-05, + "loss": 1.0139, + "step": 191720 + }, + { + "epoch": 16.6, + "learning_rate": 3.343324958849519e-05, + "loss": 0.9894, + "step": 191730 + }, + { + "epoch": 16.6, + "learning_rate": 3.343238326258338e-05, + "loss": 0.9884, + "step": 191740 + }, + { + "epoch": 16.6, + "learning_rate": 3.3431516936671575e-05, + "loss": 1.0468, + "step": 191750 + }, + { + "epoch": 16.6, + "learning_rate": 3.343065061075977e-05, + "loss": 1.0101, + "step": 191760 + }, + { + "epoch": 16.6, + "learning_rate": 3.342978428484796e-05, + "loss": 0.9891, + "step": 191770 + }, + { + "epoch": 16.6, + "learning_rate": 3.342891795893615e-05, + "loss": 0.9971, + "step": 191780 + }, + { + "epoch": 16.6, + "learning_rate": 3.3428051633024346e-05, + "loss": 1.0676, + "step": 191790 + }, + { + "epoch": 16.6, + "learning_rate": 3.3427185307112533e-05, + "loss": 1.0452, + "step": 191800 + }, + { + "epoch": 16.6, + "learning_rate": 3.342631898120073e-05, + "loss": 0.9745, + "step": 191810 + }, + { + "epoch": 16.6, + "learning_rate": 3.342545265528892e-05, + "loss": 1.0027, + "step": 191820 + }, + { + "epoch": 16.6, + "learning_rate": 3.342458632937711e-05, + "loss": 1.0426, + "step": 191830 + }, + { + "epoch": 16.6, + "learning_rate": 3.3423720003465305e-05, + "loss": 1.0334, + "step": 191840 + }, + { + "epoch": 16.61, + "learning_rate": 3.342285367755349e-05, + "loss": 1.0269, + "step": 191850 + }, + { + "epoch": 16.61, + "learning_rate": 3.342198735164169e-05, + "loss": 0.9933, + "step": 191860 + }, + { + "epoch": 16.61, + "learning_rate": 3.342112102572988e-05, + "loss": 1.1056, + "step": 191870 + }, + { + "epoch": 16.61, + "learning_rate": 3.342025469981807e-05, + "loss": 1.0545, + "step": 191880 + }, + { + "epoch": 16.61, + "learning_rate": 3.341938837390626e-05, + "loss": 1.018, + "step": 191890 + }, + { + "epoch": 16.61, + "learning_rate": 3.341852204799446e-05, + "loss": 0.987, + "step": 191900 + }, + { + "epoch": 16.61, + "learning_rate": 3.3417655722082645e-05, + "loss": 1.02, + "step": 191910 + }, + { + "epoch": 16.61, + "learning_rate": 3.341678939617084e-05, + "loss": 1.04, + "step": 191920 + }, + { + "epoch": 16.61, + "learning_rate": 3.3415923070259034e-05, + "loss": 1.0012, + "step": 191930 + }, + { + "epoch": 16.61, + "learning_rate": 3.341505674434722e-05, + "loss": 1.0676, + "step": 191940 + }, + { + "epoch": 16.61, + "learning_rate": 3.3414190418435416e-05, + "loss": 1.0646, + "step": 191950 + }, + { + "epoch": 16.62, + "learning_rate": 3.3413324092523604e-05, + "loss": 0.9996, + "step": 191960 + }, + { + "epoch": 16.62, + "learning_rate": 3.34124577666118e-05, + "loss": 1.0613, + "step": 191970 + }, + { + "epoch": 16.62, + "learning_rate": 3.341159144069999e-05, + "loss": 1.0274, + "step": 191980 + }, + { + "epoch": 16.62, + "learning_rate": 3.341072511478818e-05, + "loss": 0.9948, + "step": 191990 + }, + { + "epoch": 16.62, + "learning_rate": 3.3409858788876375e-05, + "loss": 0.9863, + "step": 192000 + }, + { + "epoch": 16.62, + "learning_rate": 3.340899246296457e-05, + "loss": 1.0256, + "step": 192010 + }, + { + "epoch": 16.62, + "learning_rate": 3.340812613705276e-05, + "loss": 0.9651, + "step": 192020 + }, + { + "epoch": 16.62, + "learning_rate": 3.340725981114095e-05, + "loss": 1.0371, + "step": 192030 + }, + { + "epoch": 16.62, + "learning_rate": 3.340639348522914e-05, + "loss": 1.0188, + "step": 192040 + }, + { + "epoch": 16.62, + "learning_rate": 3.3405527159317334e-05, + "loss": 1.0255, + "step": 192050 + }, + { + "epoch": 16.62, + "learning_rate": 3.340466083340553e-05, + "loss": 1.0009, + "step": 192060 + }, + { + "epoch": 16.62, + "learning_rate": 3.3403794507493716e-05, + "loss": 1.0142, + "step": 192070 + }, + { + "epoch": 16.63, + "learning_rate": 3.340292818158191e-05, + "loss": 1.0511, + "step": 192080 + }, + { + "epoch": 16.63, + "learning_rate": 3.3402061855670105e-05, + "loss": 0.9992, + "step": 192090 + }, + { + "epoch": 16.63, + "learning_rate": 3.340119552975829e-05, + "loss": 1.0264, + "step": 192100 + }, + { + "epoch": 16.63, + "learning_rate": 3.340032920384649e-05, + "loss": 1.0045, + "step": 192110 + }, + { + "epoch": 16.63, + "learning_rate": 3.339946287793468e-05, + "loss": 1.0471, + "step": 192120 + }, + { + "epoch": 16.63, + "learning_rate": 3.339859655202287e-05, + "loss": 1.0241, + "step": 192130 + }, + { + "epoch": 16.63, + "learning_rate": 3.3397730226111064e-05, + "loss": 1.0127, + "step": 192140 + }, + { + "epoch": 16.63, + "learning_rate": 3.339686390019925e-05, + "loss": 1.0028, + "step": 192150 + }, + { + "epoch": 16.63, + "learning_rate": 3.3395997574287446e-05, + "loss": 1.0649, + "step": 192160 + }, + { + "epoch": 16.63, + "learning_rate": 3.339513124837564e-05, + "loss": 0.9826, + "step": 192170 + }, + { + "epoch": 16.63, + "learning_rate": 3.339426492246383e-05, + "loss": 1.0236, + "step": 192180 + }, + { + "epoch": 16.64, + "learning_rate": 3.339339859655202e-05, + "loss": 1.078, + "step": 192190 + }, + { + "epoch": 16.64, + "learning_rate": 3.339253227064022e-05, + "loss": 1.023, + "step": 192200 + }, + { + "epoch": 16.64, + "learning_rate": 3.3391665944728404e-05, + "loss": 1.0425, + "step": 192210 + }, + { + "epoch": 16.64, + "learning_rate": 3.33907996188166e-05, + "loss": 1.0006, + "step": 192220 + }, + { + "epoch": 16.64, + "learning_rate": 3.338993329290479e-05, + "loss": 0.9758, + "step": 192230 + }, + { + "epoch": 16.64, + "learning_rate": 3.338906696699298e-05, + "loss": 0.9723, + "step": 192240 + }, + { + "epoch": 16.64, + "learning_rate": 3.3388200641081175e-05, + "loss": 1.0365, + "step": 192250 + }, + { + "epoch": 16.64, + "learning_rate": 3.338733431516936e-05, + "loss": 1.0512, + "step": 192260 + }, + { + "epoch": 16.64, + "learning_rate": 3.338646798925756e-05, + "loss": 1.0253, + "step": 192270 + }, + { + "epoch": 16.64, + "learning_rate": 3.338560166334575e-05, + "loss": 1.0163, + "step": 192280 + }, + { + "epoch": 16.64, + "learning_rate": 3.338473533743394e-05, + "loss": 1.0551, + "step": 192290 + }, + { + "epoch": 16.64, + "learning_rate": 3.3383869011522134e-05, + "loss": 0.9815, + "step": 192300 + }, + { + "epoch": 16.65, + "learning_rate": 3.338300268561033e-05, + "loss": 1.0078, + "step": 192310 + }, + { + "epoch": 16.65, + "learning_rate": 3.3382136359698516e-05, + "loss": 0.9994, + "step": 192320 + }, + { + "epoch": 16.65, + "learning_rate": 3.338127003378671e-05, + "loss": 1.0411, + "step": 192330 + }, + { + "epoch": 16.65, + "learning_rate": 3.3380403707874905e-05, + "loss": 1.0618, + "step": 192340 + }, + { + "epoch": 16.65, + "learning_rate": 3.337953738196309e-05, + "loss": 1.034, + "step": 192350 + }, + { + "epoch": 16.65, + "learning_rate": 3.337867105605129e-05, + "loss": 1.0165, + "step": 192360 + }, + { + "epoch": 16.65, + "learning_rate": 3.3377804730139475e-05, + "loss": 1.0643, + "step": 192370 + }, + { + "epoch": 16.65, + "learning_rate": 3.337693840422767e-05, + "loss": 1.0492, + "step": 192380 + }, + { + "epoch": 16.65, + "learning_rate": 3.3376072078315864e-05, + "loss": 1.0754, + "step": 192390 + }, + { + "epoch": 16.65, + "learning_rate": 3.337520575240405e-05, + "loss": 0.9585, + "step": 192400 + }, + { + "epoch": 16.65, + "learning_rate": 3.3374339426492246e-05, + "loss": 1.0288, + "step": 192410 + }, + { + "epoch": 16.65, + "learning_rate": 3.337347310058044e-05, + "loss": 0.9827, + "step": 192420 + }, + { + "epoch": 16.66, + "learning_rate": 3.337260677466863e-05, + "loss": 1.0198, + "step": 192430 + }, + { + "epoch": 16.66, + "learning_rate": 3.337174044875682e-05, + "loss": 1.0703, + "step": 192440 + }, + { + "epoch": 16.66, + "learning_rate": 3.337087412284502e-05, + "loss": 0.9974, + "step": 192450 + }, + { + "epoch": 16.66, + "learning_rate": 3.3370007796933205e-05, + "loss": 1.0524, + "step": 192460 + }, + { + "epoch": 16.66, + "learning_rate": 3.33691414710214e-05, + "loss": 0.9947, + "step": 192470 + }, + { + "epoch": 16.66, + "learning_rate": 3.336827514510959e-05, + "loss": 1.0676, + "step": 192480 + }, + { + "epoch": 16.66, + "learning_rate": 3.336740881919778e-05, + "loss": 1.0392, + "step": 192490 + }, + { + "epoch": 16.66, + "learning_rate": 3.3366542493285976e-05, + "loss": 1.0156, + "step": 192500 + }, + { + "epoch": 16.66, + "learning_rate": 3.3365676167374163e-05, + "loss": 0.9584, + "step": 192510 + }, + { + "epoch": 16.66, + "learning_rate": 3.336480984146236e-05, + "loss": 0.9698, + "step": 192520 + }, + { + "epoch": 16.66, + "learning_rate": 3.336394351555055e-05, + "loss": 1.0297, + "step": 192530 + }, + { + "epoch": 16.67, + "learning_rate": 3.336307718963874e-05, + "loss": 1.0202, + "step": 192540 + }, + { + "epoch": 16.67, + "learning_rate": 3.3362210863726934e-05, + "loss": 1.0497, + "step": 192550 + }, + { + "epoch": 16.67, + "learning_rate": 3.336134453781513e-05, + "loss": 1.005, + "step": 192560 + }, + { + "epoch": 16.67, + "learning_rate": 3.336047821190332e-05, + "loss": 0.9906, + "step": 192570 + }, + { + "epoch": 16.67, + "learning_rate": 3.335961188599151e-05, + "loss": 1.0111, + "step": 192580 + }, + { + "epoch": 16.67, + "learning_rate": 3.33587455600797e-05, + "loss": 1.0118, + "step": 192590 + }, + { + "epoch": 16.67, + "learning_rate": 3.335787923416789e-05, + "loss": 1.0522, + "step": 192600 + }, + { + "epoch": 16.67, + "learning_rate": 3.335701290825609e-05, + "loss": 1.0125, + "step": 192610 + }, + { + "epoch": 16.67, + "learning_rate": 3.3356146582344275e-05, + "loss": 1.0536, + "step": 192620 + }, + { + "epoch": 16.67, + "learning_rate": 3.335528025643247e-05, + "loss": 1.041, + "step": 192630 + }, + { + "epoch": 16.67, + "learning_rate": 3.3354413930520664e-05, + "loss": 1.0116, + "step": 192640 + }, + { + "epoch": 16.67, + "learning_rate": 3.335354760460885e-05, + "loss": 1.0154, + "step": 192650 + }, + { + "epoch": 16.68, + "learning_rate": 3.3352681278697046e-05, + "loss": 0.9551, + "step": 192660 + }, + { + "epoch": 16.68, + "learning_rate": 3.335181495278524e-05, + "loss": 1.0403, + "step": 192670 + }, + { + "epoch": 16.68, + "learning_rate": 3.335094862687343e-05, + "loss": 1.0221, + "step": 192680 + }, + { + "epoch": 16.68, + "learning_rate": 3.335008230096162e-05, + "loss": 1.0282, + "step": 192690 + }, + { + "epoch": 16.68, + "learning_rate": 3.334921597504981e-05, + "loss": 0.9668, + "step": 192700 + }, + { + "epoch": 16.68, + "learning_rate": 3.3348349649138005e-05, + "loss": 1.0108, + "step": 192710 + }, + { + "epoch": 16.68, + "learning_rate": 3.33474833232262e-05, + "loss": 1.0332, + "step": 192720 + }, + { + "epoch": 16.68, + "learning_rate": 3.334661699731439e-05, + "loss": 0.9955, + "step": 192730 + }, + { + "epoch": 16.68, + "learning_rate": 3.334575067140258e-05, + "loss": 1.0151, + "step": 192740 + }, + { + "epoch": 16.68, + "learning_rate": 3.3344884345490776e-05, + "loss": 1.0014, + "step": 192750 + }, + { + "epoch": 16.68, + "learning_rate": 3.3344018019578964e-05, + "loss": 0.9736, + "step": 192760 + }, + { + "epoch": 16.69, + "learning_rate": 3.334315169366716e-05, + "loss": 1.0019, + "step": 192770 + }, + { + "epoch": 16.69, + "learning_rate": 3.3342285367755346e-05, + "loss": 1.0864, + "step": 192780 + }, + { + "epoch": 16.69, + "learning_rate": 3.334141904184354e-05, + "loss": 1.0212, + "step": 192790 + }, + { + "epoch": 16.69, + "learning_rate": 3.3340552715931735e-05, + "loss": 1.0304, + "step": 192800 + }, + { + "epoch": 16.69, + "learning_rate": 3.333968639001992e-05, + "loss": 1.0905, + "step": 192810 + }, + { + "epoch": 16.69, + "learning_rate": 3.333882006410812e-05, + "loss": 1.0154, + "step": 192820 + }, + { + "epoch": 16.69, + "learning_rate": 3.333795373819631e-05, + "loss": 1.0572, + "step": 192830 + }, + { + "epoch": 16.69, + "learning_rate": 3.33370874122845e-05, + "loss": 0.9946, + "step": 192840 + }, + { + "epoch": 16.69, + "learning_rate": 3.3336221086372694e-05, + "loss": 1.0677, + "step": 192850 + }, + { + "epoch": 16.69, + "learning_rate": 3.333535476046089e-05, + "loss": 1.0145, + "step": 192860 + }, + { + "epoch": 16.69, + "learning_rate": 3.3334488434549076e-05, + "loss": 0.9737, + "step": 192870 + }, + { + "epoch": 16.69, + "learning_rate": 3.333362210863727e-05, + "loss": 1.0587, + "step": 192880 + }, + { + "epoch": 16.7, + "learning_rate": 3.333275578272546e-05, + "loss": 0.9886, + "step": 192890 + }, + { + "epoch": 16.7, + "learning_rate": 3.333188945681365e-05, + "loss": 1.0318, + "step": 192900 + }, + { + "epoch": 16.7, + "learning_rate": 3.333102313090185e-05, + "loss": 0.9695, + "step": 192910 + }, + { + "epoch": 16.7, + "learning_rate": 3.3330156804990034e-05, + "loss": 1.0473, + "step": 192920 + }, + { + "epoch": 16.7, + "learning_rate": 3.332929047907823e-05, + "loss": 1.0728, + "step": 192930 + }, + { + "epoch": 16.7, + "learning_rate": 3.332842415316642e-05, + "loss": 1.0243, + "step": 192940 + }, + { + "epoch": 16.7, + "learning_rate": 3.332755782725461e-05, + "loss": 1.0076, + "step": 192950 + }, + { + "epoch": 16.7, + "learning_rate": 3.3326691501342805e-05, + "loss": 1.0846, + "step": 192960 + }, + { + "epoch": 16.7, + "learning_rate": 3.3325825175431e-05, + "loss": 1.0088, + "step": 192970 + }, + { + "epoch": 16.7, + "learning_rate": 3.332495884951919e-05, + "loss": 1.0433, + "step": 192980 + }, + { + "epoch": 16.7, + "learning_rate": 3.332409252360738e-05, + "loss": 0.9921, + "step": 192990 + }, + { + "epoch": 16.71, + "learning_rate": 3.332322619769557e-05, + "loss": 0.9975, + "step": 193000 + }, + { + "epoch": 16.71, + "learning_rate": 3.3322359871783764e-05, + "loss": 1.0083, + "step": 193010 + }, + { + "epoch": 16.71, + "learning_rate": 3.332149354587196e-05, + "loss": 1.0071, + "step": 193020 + }, + { + "epoch": 16.71, + "learning_rate": 3.3320627219960146e-05, + "loss": 1.0472, + "step": 193030 + }, + { + "epoch": 16.71, + "learning_rate": 3.331976089404834e-05, + "loss": 1.0126, + "step": 193040 + }, + { + "epoch": 16.71, + "learning_rate": 3.3318894568136535e-05, + "loss": 0.977, + "step": 193050 + }, + { + "epoch": 16.71, + "learning_rate": 3.331802824222472e-05, + "loss": 1.022, + "step": 193060 + }, + { + "epoch": 16.71, + "learning_rate": 3.331716191631292e-05, + "loss": 1.0355, + "step": 193070 + }, + { + "epoch": 16.71, + "learning_rate": 3.331629559040111e-05, + "loss": 1.0471, + "step": 193080 + }, + { + "epoch": 16.71, + "learning_rate": 3.33154292644893e-05, + "loss": 0.9939, + "step": 193090 + }, + { + "epoch": 16.71, + "learning_rate": 3.3314562938577494e-05, + "loss": 1.0214, + "step": 193100 + }, + { + "epoch": 16.71, + "learning_rate": 3.331369661266568e-05, + "loss": 0.9908, + "step": 193110 + }, + { + "epoch": 16.72, + "learning_rate": 3.3312830286753876e-05, + "loss": 1.0527, + "step": 193120 + }, + { + "epoch": 16.72, + "learning_rate": 3.331196396084207e-05, + "loss": 1.0033, + "step": 193130 + }, + { + "epoch": 16.72, + "learning_rate": 3.331109763493026e-05, + "loss": 1.0488, + "step": 193140 + }, + { + "epoch": 16.72, + "learning_rate": 3.331023130901845e-05, + "loss": 0.9994, + "step": 193150 + }, + { + "epoch": 16.72, + "learning_rate": 3.330936498310665e-05, + "loss": 1.0542, + "step": 193160 + }, + { + "epoch": 16.72, + "learning_rate": 3.3308498657194835e-05, + "loss": 0.9658, + "step": 193170 + }, + { + "epoch": 16.72, + "learning_rate": 3.330763233128303e-05, + "loss": 0.9952, + "step": 193180 + }, + { + "epoch": 16.72, + "learning_rate": 3.3306766005371224e-05, + "loss": 1.0117, + "step": 193190 + }, + { + "epoch": 16.72, + "learning_rate": 3.330589967945941e-05, + "loss": 0.9608, + "step": 193200 + }, + { + "epoch": 16.72, + "learning_rate": 3.3305033353547606e-05, + "loss": 1.0271, + "step": 193210 + }, + { + "epoch": 16.72, + "learning_rate": 3.3304167027635793e-05, + "loss": 1.0133, + "step": 193220 + }, + { + "epoch": 16.73, + "learning_rate": 3.330330070172399e-05, + "loss": 1.0147, + "step": 193230 + }, + { + "epoch": 16.73, + "learning_rate": 3.330243437581218e-05, + "loss": 1.0225, + "step": 193240 + }, + { + "epoch": 16.73, + "learning_rate": 3.330156804990037e-05, + "loss": 1.0238, + "step": 193250 + }, + { + "epoch": 16.73, + "learning_rate": 3.3300701723988564e-05, + "loss": 0.9791, + "step": 193260 + }, + { + "epoch": 16.73, + "learning_rate": 3.329983539807676e-05, + "loss": 0.9889, + "step": 193270 + }, + { + "epoch": 16.73, + "learning_rate": 3.3298969072164947e-05, + "loss": 1.0134, + "step": 193280 + }, + { + "epoch": 16.73, + "learning_rate": 3.329810274625314e-05, + "loss": 1.0018, + "step": 193290 + }, + { + "epoch": 16.73, + "learning_rate": 3.3297236420341335e-05, + "loss": 0.9962, + "step": 193300 + }, + { + "epoch": 16.73, + "learning_rate": 3.329637009442952e-05, + "loss": 1.0116, + "step": 193310 + }, + { + "epoch": 16.73, + "learning_rate": 3.329550376851772e-05, + "loss": 1.0618, + "step": 193320 + }, + { + "epoch": 16.73, + "learning_rate": 3.3294637442605905e-05, + "loss": 1.0393, + "step": 193330 + }, + { + "epoch": 16.73, + "learning_rate": 3.32937711166941e-05, + "loss": 1.0245, + "step": 193340 + }, + { + "epoch": 16.74, + "learning_rate": 3.3292904790782294e-05, + "loss": 1.0642, + "step": 193350 + }, + { + "epoch": 16.74, + "learning_rate": 3.329203846487048e-05, + "loss": 1.0364, + "step": 193360 + }, + { + "epoch": 16.74, + "learning_rate": 3.3291172138958676e-05, + "loss": 1.0077, + "step": 193370 + }, + { + "epoch": 16.74, + "learning_rate": 3.329030581304687e-05, + "loss": 1.0581, + "step": 193380 + }, + { + "epoch": 16.74, + "learning_rate": 3.328943948713506e-05, + "loss": 1.059, + "step": 193390 + }, + { + "epoch": 16.74, + "learning_rate": 3.328857316122325e-05, + "loss": 0.9631, + "step": 193400 + }, + { + "epoch": 16.74, + "learning_rate": 3.328770683531145e-05, + "loss": 1.0423, + "step": 193410 + }, + { + "epoch": 16.74, + "learning_rate": 3.3286840509399635e-05, + "loss": 1.0145, + "step": 193420 + }, + { + "epoch": 16.74, + "learning_rate": 3.328597418348783e-05, + "loss": 1.0425, + "step": 193430 + }, + { + "epoch": 16.74, + "learning_rate": 3.328510785757602e-05, + "loss": 0.9708, + "step": 193440 + }, + { + "epoch": 16.74, + "learning_rate": 3.328424153166421e-05, + "loss": 1.0413, + "step": 193450 + }, + { + "epoch": 16.74, + "learning_rate": 3.3283375205752406e-05, + "loss": 1.0355, + "step": 193460 + }, + { + "epoch": 16.75, + "learning_rate": 3.3282508879840594e-05, + "loss": 1.0748, + "step": 193470 + }, + { + "epoch": 16.75, + "learning_rate": 3.328164255392879e-05, + "loss": 0.9796, + "step": 193480 + }, + { + "epoch": 16.75, + "learning_rate": 3.328077622801698e-05, + "loss": 1.0043, + "step": 193490 + }, + { + "epoch": 16.75, + "learning_rate": 3.327990990210517e-05, + "loss": 1.0168, + "step": 193500 + }, + { + "epoch": 16.75, + "learning_rate": 3.3279043576193365e-05, + "loss": 0.9583, + "step": 193510 + }, + { + "epoch": 16.75, + "learning_rate": 3.327817725028156e-05, + "loss": 0.9912, + "step": 193520 + }, + { + "epoch": 16.75, + "learning_rate": 3.327731092436975e-05, + "loss": 1.0034, + "step": 193530 + }, + { + "epoch": 16.75, + "learning_rate": 3.327644459845794e-05, + "loss": 1.0204, + "step": 193540 + }, + { + "epoch": 16.75, + "learning_rate": 3.327557827254613e-05, + "loss": 1.0622, + "step": 193550 + }, + { + "epoch": 16.75, + "learning_rate": 3.3274711946634324e-05, + "loss": 1.0839, + "step": 193560 + }, + { + "epoch": 16.75, + "learning_rate": 3.327384562072252e-05, + "loss": 1.0261, + "step": 193570 + }, + { + "epoch": 16.76, + "learning_rate": 3.3272979294810706e-05, + "loss": 1.0815, + "step": 193580 + }, + { + "epoch": 16.76, + "learning_rate": 3.32721129688989e-05, + "loss": 1.0062, + "step": 193590 + }, + { + "epoch": 16.76, + "learning_rate": 3.3271246642987095e-05, + "loss": 1.0055, + "step": 193600 + }, + { + "epoch": 16.76, + "learning_rate": 3.327038031707528e-05, + "loss": 0.9685, + "step": 193610 + }, + { + "epoch": 16.76, + "learning_rate": 3.326951399116348e-05, + "loss": 0.9995, + "step": 193620 + }, + { + "epoch": 16.76, + "learning_rate": 3.3268647665251664e-05, + "loss": 1.0286, + "step": 193630 + }, + { + "epoch": 16.76, + "learning_rate": 3.326778133933986e-05, + "loss": 1.0402, + "step": 193640 + }, + { + "epoch": 16.76, + "learning_rate": 3.326691501342805e-05, + "loss": 1.0459, + "step": 193650 + }, + { + "epoch": 16.76, + "learning_rate": 3.326604868751624e-05, + "loss": 1.0595, + "step": 193660 + }, + { + "epoch": 16.76, + "learning_rate": 3.3265182361604435e-05, + "loss": 1.0486, + "step": 193670 + }, + { + "epoch": 16.76, + "learning_rate": 3.326431603569263e-05, + "loss": 1.009, + "step": 193680 + }, + { + "epoch": 16.76, + "learning_rate": 3.326344970978082e-05, + "loss": 0.9989, + "step": 193690 + }, + { + "epoch": 16.77, + "learning_rate": 3.326258338386901e-05, + "loss": 0.985, + "step": 193700 + }, + { + "epoch": 16.77, + "learning_rate": 3.3261717057957206e-05, + "loss": 1.0122, + "step": 193710 + }, + { + "epoch": 16.77, + "learning_rate": 3.3260850732045394e-05, + "loss": 0.9748, + "step": 193720 + }, + { + "epoch": 16.77, + "learning_rate": 3.325998440613359e-05, + "loss": 1.0239, + "step": 193730 + }, + { + "epoch": 16.77, + "learning_rate": 3.3259118080221776e-05, + "loss": 0.9997, + "step": 193740 + }, + { + "epoch": 16.77, + "learning_rate": 3.325825175430997e-05, + "loss": 0.999, + "step": 193750 + }, + { + "epoch": 16.77, + "learning_rate": 3.3257385428398165e-05, + "loss": 0.9935, + "step": 193760 + }, + { + "epoch": 16.77, + "learning_rate": 3.325651910248635e-05, + "loss": 1.1144, + "step": 193770 + }, + { + "epoch": 16.77, + "learning_rate": 3.325565277657455e-05, + "loss": 1.0221, + "step": 193780 + }, + { + "epoch": 16.77, + "learning_rate": 3.325478645066274e-05, + "loss": 1.0285, + "step": 193790 + }, + { + "epoch": 16.77, + "learning_rate": 3.325392012475093e-05, + "loss": 1.0041, + "step": 193800 + }, + { + "epoch": 16.78, + "learning_rate": 3.3253053798839124e-05, + "loss": 0.9961, + "step": 193810 + }, + { + "epoch": 16.78, + "learning_rate": 3.325218747292732e-05, + "loss": 1.0156, + "step": 193820 + }, + { + "epoch": 16.78, + "learning_rate": 3.3251321147015506e-05, + "loss": 1.0224, + "step": 193830 + }, + { + "epoch": 16.78, + "learning_rate": 3.32504548211037e-05, + "loss": 1.0121, + "step": 193840 + }, + { + "epoch": 16.78, + "learning_rate": 3.324958849519189e-05, + "loss": 0.9979, + "step": 193850 + }, + { + "epoch": 16.78, + "learning_rate": 3.324872216928008e-05, + "loss": 1.0264, + "step": 193860 + }, + { + "epoch": 16.78, + "learning_rate": 3.324785584336828e-05, + "loss": 1.0487, + "step": 193870 + }, + { + "epoch": 16.78, + "learning_rate": 3.3246989517456465e-05, + "loss": 1.0283, + "step": 193880 + }, + { + "epoch": 16.78, + "learning_rate": 3.324612319154466e-05, + "loss": 1.023, + "step": 193890 + }, + { + "epoch": 16.78, + "learning_rate": 3.3245256865632854e-05, + "loss": 1.0136, + "step": 193900 + }, + { + "epoch": 16.78, + "learning_rate": 3.324439053972104e-05, + "loss": 1.0228, + "step": 193910 + }, + { + "epoch": 16.78, + "learning_rate": 3.3243524213809236e-05, + "loss": 0.9731, + "step": 193920 + }, + { + "epoch": 16.79, + "learning_rate": 3.324265788789743e-05, + "loss": 1.0341, + "step": 193930 + }, + { + "epoch": 16.79, + "learning_rate": 3.324179156198562e-05, + "loss": 1.0478, + "step": 193940 + }, + { + "epoch": 16.79, + "learning_rate": 3.324092523607381e-05, + "loss": 1.0273, + "step": 193950 + }, + { + "epoch": 16.79, + "learning_rate": 3.3240058910162e-05, + "loss": 1.0331, + "step": 193960 + }, + { + "epoch": 16.79, + "learning_rate": 3.3239192584250194e-05, + "loss": 1.032, + "step": 193970 + }, + { + "epoch": 16.79, + "learning_rate": 3.323832625833839e-05, + "loss": 1.0482, + "step": 193980 + }, + { + "epoch": 16.79, + "learning_rate": 3.3237459932426577e-05, + "loss": 1.0143, + "step": 193990 + }, + { + "epoch": 16.79, + "learning_rate": 3.323659360651477e-05, + "loss": 1.0101, + "step": 194000 + }, + { + "epoch": 16.79, + "learning_rate": 3.3235727280602965e-05, + "loss": 1.0411, + "step": 194010 + }, + { + "epoch": 16.79, + "learning_rate": 3.323486095469115e-05, + "loss": 1.0194, + "step": 194020 + }, + { + "epoch": 16.79, + "learning_rate": 3.323399462877935e-05, + "loss": 0.9711, + "step": 194030 + }, + { + "epoch": 16.8, + "learning_rate": 3.323312830286754e-05, + "loss": 0.9991, + "step": 194040 + }, + { + "epoch": 16.8, + "learning_rate": 3.323226197695573e-05, + "loss": 1.0592, + "step": 194050 + }, + { + "epoch": 16.8, + "learning_rate": 3.3231395651043924e-05, + "loss": 0.9788, + "step": 194060 + }, + { + "epoch": 16.8, + "learning_rate": 3.323052932513211e-05, + "loss": 0.9648, + "step": 194070 + }, + { + "epoch": 16.8, + "learning_rate": 3.3229662999220306e-05, + "loss": 1.0445, + "step": 194080 + }, + { + "epoch": 16.8, + "learning_rate": 3.32287966733085e-05, + "loss": 0.9821, + "step": 194090 + }, + { + "epoch": 16.8, + "learning_rate": 3.322793034739669e-05, + "loss": 1.0559, + "step": 194100 + }, + { + "epoch": 16.8, + "learning_rate": 3.322706402148488e-05, + "loss": 1.0096, + "step": 194110 + }, + { + "epoch": 16.8, + "learning_rate": 3.322619769557308e-05, + "loss": 1.0201, + "step": 194120 + }, + { + "epoch": 16.8, + "learning_rate": 3.3225331369661265e-05, + "loss": 1.0369, + "step": 194130 + }, + { + "epoch": 16.8, + "learning_rate": 3.322446504374946e-05, + "loss": 1.0236, + "step": 194140 + }, + { + "epoch": 16.8, + "learning_rate": 3.3223598717837654e-05, + "loss": 1.0159, + "step": 194150 + }, + { + "epoch": 16.81, + "learning_rate": 3.322273239192584e-05, + "loss": 1.0494, + "step": 194160 + }, + { + "epoch": 16.81, + "learning_rate": 3.3221866066014036e-05, + "loss": 1.0253, + "step": 194170 + }, + { + "epoch": 16.81, + "learning_rate": 3.3220999740102224e-05, + "loss": 1.019, + "step": 194180 + }, + { + "epoch": 16.81, + "learning_rate": 3.322013341419042e-05, + "loss": 0.9499, + "step": 194190 + }, + { + "epoch": 16.81, + "learning_rate": 3.321926708827861e-05, + "loss": 1.0057, + "step": 194200 + }, + { + "epoch": 16.81, + "learning_rate": 3.32184007623668e-05, + "loss": 0.9929, + "step": 194210 + }, + { + "epoch": 16.81, + "learning_rate": 3.3217534436454995e-05, + "loss": 0.9815, + "step": 194220 + }, + { + "epoch": 16.81, + "learning_rate": 3.321666811054319e-05, + "loss": 0.9967, + "step": 194230 + }, + { + "epoch": 16.81, + "learning_rate": 3.321580178463138e-05, + "loss": 1.0462, + "step": 194240 + }, + { + "epoch": 16.81, + "learning_rate": 3.321493545871957e-05, + "loss": 0.9761, + "step": 194250 + }, + { + "epoch": 16.81, + "learning_rate": 3.3214069132807766e-05, + "loss": 1.0289, + "step": 194260 + }, + { + "epoch": 16.82, + "learning_rate": 3.3213202806895953e-05, + "loss": 1.1131, + "step": 194270 + }, + { + "epoch": 16.82, + "learning_rate": 3.321233648098415e-05, + "loss": 0.9515, + "step": 194280 + }, + { + "epoch": 16.82, + "learning_rate": 3.3211470155072336e-05, + "loss": 0.9705, + "step": 194290 + }, + { + "epoch": 16.82, + "learning_rate": 3.321060382916053e-05, + "loss": 1.0198, + "step": 194300 + }, + { + "epoch": 16.82, + "learning_rate": 3.3209737503248725e-05, + "loss": 1.051, + "step": 194310 + }, + { + "epoch": 16.82, + "learning_rate": 3.320887117733691e-05, + "loss": 0.9603, + "step": 194320 + }, + { + "epoch": 16.82, + "learning_rate": 3.320800485142511e-05, + "loss": 1.0467, + "step": 194330 + }, + { + "epoch": 16.82, + "learning_rate": 3.32071385255133e-05, + "loss": 1.0248, + "step": 194340 + }, + { + "epoch": 16.82, + "learning_rate": 3.320627219960149e-05, + "loss": 1.0106, + "step": 194350 + }, + { + "epoch": 16.82, + "learning_rate": 3.320540587368968e-05, + "loss": 1.0232, + "step": 194360 + }, + { + "epoch": 16.82, + "learning_rate": 3.320453954777787e-05, + "loss": 0.903, + "step": 194370 + }, + { + "epoch": 16.82, + "learning_rate": 3.3203673221866065e-05, + "loss": 1.0498, + "step": 194380 + }, + { + "epoch": 16.83, + "learning_rate": 3.320280689595426e-05, + "loss": 1.055, + "step": 194390 + }, + { + "epoch": 16.83, + "learning_rate": 3.320194057004245e-05, + "loss": 1.0132, + "step": 194400 + }, + { + "epoch": 16.83, + "learning_rate": 3.320107424413064e-05, + "loss": 0.9807, + "step": 194410 + }, + { + "epoch": 16.83, + "learning_rate": 3.3200207918218836e-05, + "loss": 0.9915, + "step": 194420 + }, + { + "epoch": 16.83, + "learning_rate": 3.3199341592307024e-05, + "loss": 0.9894, + "step": 194430 + }, + { + "epoch": 16.83, + "learning_rate": 3.319847526639522e-05, + "loss": 1.0457, + "step": 194440 + }, + { + "epoch": 16.83, + "learning_rate": 3.319760894048341e-05, + "loss": 0.971, + "step": 194450 + }, + { + "epoch": 16.83, + "learning_rate": 3.31967426145716e-05, + "loss": 1.0401, + "step": 194460 + }, + { + "epoch": 16.83, + "learning_rate": 3.3195876288659795e-05, + "loss": 1.0431, + "step": 194470 + }, + { + "epoch": 16.83, + "learning_rate": 3.319500996274798e-05, + "loss": 1.0443, + "step": 194480 + }, + { + "epoch": 16.83, + "learning_rate": 3.319414363683618e-05, + "loss": 0.9903, + "step": 194490 + }, + { + "epoch": 16.83, + "learning_rate": 3.319327731092437e-05, + "loss": 1.0641, + "step": 194500 + }, + { + "epoch": 16.84, + "learning_rate": 3.319241098501256e-05, + "loss": 1.0315, + "step": 194510 + }, + { + "epoch": 16.84, + "learning_rate": 3.3191544659100754e-05, + "loss": 1.0522, + "step": 194520 + }, + { + "epoch": 16.84, + "learning_rate": 3.319067833318895e-05, + "loss": 1.1076, + "step": 194530 + }, + { + "epoch": 16.84, + "learning_rate": 3.3189812007277136e-05, + "loss": 0.993, + "step": 194540 + }, + { + "epoch": 16.84, + "learning_rate": 3.318894568136533e-05, + "loss": 1.0311, + "step": 194550 + }, + { + "epoch": 16.84, + "learning_rate": 3.3188079355453525e-05, + "loss": 1.0326, + "step": 194560 + }, + { + "epoch": 16.84, + "learning_rate": 3.318721302954171e-05, + "loss": 1.0364, + "step": 194570 + }, + { + "epoch": 16.84, + "learning_rate": 3.318634670362991e-05, + "loss": 1.1228, + "step": 194580 + }, + { + "epoch": 16.84, + "learning_rate": 3.3185480377718095e-05, + "loss": 1.0018, + "step": 194590 + }, + { + "epoch": 16.84, + "learning_rate": 3.318461405180629e-05, + "loss": 0.9941, + "step": 194600 + }, + { + "epoch": 16.84, + "learning_rate": 3.3183747725894484e-05, + "loss": 1.0284, + "step": 194610 + }, + { + "epoch": 16.85, + "learning_rate": 3.318288139998267e-05, + "loss": 1.001, + "step": 194620 + }, + { + "epoch": 16.85, + "learning_rate": 3.3182015074070866e-05, + "loss": 1.0054, + "step": 194630 + }, + { + "epoch": 16.85, + "learning_rate": 3.318114874815906e-05, + "loss": 1.0141, + "step": 194640 + }, + { + "epoch": 16.85, + "learning_rate": 3.318028242224725e-05, + "loss": 1.0167, + "step": 194650 + }, + { + "epoch": 16.85, + "learning_rate": 3.317941609633544e-05, + "loss": 1.0595, + "step": 194660 + }, + { + "epoch": 16.85, + "learning_rate": 3.317854977042364e-05, + "loss": 1.0872, + "step": 194670 + }, + { + "epoch": 16.85, + "learning_rate": 3.3177683444511824e-05, + "loss": 0.9707, + "step": 194680 + }, + { + "epoch": 16.85, + "learning_rate": 3.317681711860002e-05, + "loss": 0.9648, + "step": 194690 + }, + { + "epoch": 16.85, + "learning_rate": 3.3175950792688207e-05, + "loss": 1.0496, + "step": 194700 + }, + { + "epoch": 16.85, + "learning_rate": 3.31750844667764e-05, + "loss": 1.0019, + "step": 194710 + }, + { + "epoch": 16.85, + "learning_rate": 3.3174218140864595e-05, + "loss": 1.055, + "step": 194720 + }, + { + "epoch": 16.85, + "learning_rate": 3.317335181495278e-05, + "loss": 1.0153, + "step": 194730 + }, + { + "epoch": 16.86, + "learning_rate": 3.317248548904098e-05, + "loss": 1.0396, + "step": 194740 + }, + { + "epoch": 16.86, + "learning_rate": 3.317161916312917e-05, + "loss": 1.0359, + "step": 194750 + }, + { + "epoch": 16.86, + "learning_rate": 3.317075283721736e-05, + "loss": 1.0424, + "step": 194760 + }, + { + "epoch": 16.86, + "learning_rate": 3.3169886511305554e-05, + "loss": 0.9823, + "step": 194770 + }, + { + "epoch": 16.86, + "learning_rate": 3.316902018539375e-05, + "loss": 1.0281, + "step": 194780 + }, + { + "epoch": 16.86, + "learning_rate": 3.3168153859481936e-05, + "loss": 0.9909, + "step": 194790 + }, + { + "epoch": 16.86, + "learning_rate": 3.316728753357013e-05, + "loss": 1.0517, + "step": 194800 + }, + { + "epoch": 16.86, + "learning_rate": 3.316642120765832e-05, + "loss": 1.0476, + "step": 194810 + }, + { + "epoch": 16.86, + "learning_rate": 3.316555488174651e-05, + "loss": 1.0048, + "step": 194820 + }, + { + "epoch": 16.86, + "learning_rate": 3.316468855583471e-05, + "loss": 1.0814, + "step": 194830 + }, + { + "epoch": 16.86, + "learning_rate": 3.3163822229922895e-05, + "loss": 1.0175, + "step": 194840 + }, + { + "epoch": 16.87, + "learning_rate": 3.316295590401109e-05, + "loss": 1.105, + "step": 194850 + }, + { + "epoch": 16.87, + "learning_rate": 3.3162089578099284e-05, + "loss": 1.0058, + "step": 194860 + }, + { + "epoch": 16.87, + "learning_rate": 3.316122325218747e-05, + "loss": 1.0174, + "step": 194870 + }, + { + "epoch": 16.87, + "learning_rate": 3.3160356926275666e-05, + "loss": 1.0164, + "step": 194880 + }, + { + "epoch": 16.87, + "learning_rate": 3.315949060036386e-05, + "loss": 0.9554, + "step": 194890 + }, + { + "epoch": 16.87, + "learning_rate": 3.315862427445205e-05, + "loss": 1.0567, + "step": 194900 + }, + { + "epoch": 16.87, + "learning_rate": 3.315775794854024e-05, + "loss": 1.0371, + "step": 194910 + }, + { + "epoch": 16.87, + "learning_rate": 3.315689162262843e-05, + "loss": 1.0231, + "step": 194920 + }, + { + "epoch": 16.87, + "learning_rate": 3.3156025296716625e-05, + "loss": 0.9738, + "step": 194930 + }, + { + "epoch": 16.87, + "learning_rate": 3.315515897080482e-05, + "loss": 1.0195, + "step": 194940 + }, + { + "epoch": 16.87, + "learning_rate": 3.315429264489301e-05, + "loss": 1.0495, + "step": 194950 + }, + { + "epoch": 16.87, + "learning_rate": 3.31534263189812e-05, + "loss": 1.0271, + "step": 194960 + }, + { + "epoch": 16.88, + "learning_rate": 3.3152559993069396e-05, + "loss": 1.0375, + "step": 194970 + }, + { + "epoch": 16.88, + "learning_rate": 3.3151693667157583e-05, + "loss": 0.9961, + "step": 194980 + }, + { + "epoch": 16.88, + "learning_rate": 3.315082734124578e-05, + "loss": 0.9776, + "step": 194990 + }, + { + "epoch": 16.88, + "learning_rate": 3.314996101533397e-05, + "loss": 1.0009, + "step": 195000 + }, + { + "epoch": 16.88, + "learning_rate": 3.314909468942216e-05, + "loss": 0.999, + "step": 195010 + }, + { + "epoch": 16.88, + "learning_rate": 3.3148228363510354e-05, + "loss": 1.0069, + "step": 195020 + }, + { + "epoch": 16.88, + "learning_rate": 3.314736203759854e-05, + "loss": 1.0471, + "step": 195030 + }, + { + "epoch": 16.88, + "learning_rate": 3.3146495711686737e-05, + "loss": 1.0191, + "step": 195040 + }, + { + "epoch": 16.88, + "learning_rate": 3.314562938577493e-05, + "loss": 1.0148, + "step": 195050 + }, + { + "epoch": 16.88, + "learning_rate": 3.314476305986312e-05, + "loss": 1.074, + "step": 195060 + }, + { + "epoch": 16.88, + "learning_rate": 3.314389673395131e-05, + "loss": 0.968, + "step": 195070 + }, + { + "epoch": 16.89, + "learning_rate": 3.314303040803951e-05, + "loss": 1.0033, + "step": 195080 + }, + { + "epoch": 16.89, + "learning_rate": 3.3142164082127695e-05, + "loss": 1.0571, + "step": 195090 + }, + { + "epoch": 16.89, + "learning_rate": 3.314129775621589e-05, + "loss": 0.9983, + "step": 195100 + }, + { + "epoch": 16.89, + "learning_rate": 3.314043143030408e-05, + "loss": 0.9767, + "step": 195110 + }, + { + "epoch": 16.89, + "learning_rate": 3.313956510439227e-05, + "loss": 1.0357, + "step": 195120 + }, + { + "epoch": 16.89, + "learning_rate": 3.3138698778480466e-05, + "loss": 0.9937, + "step": 195130 + }, + { + "epoch": 16.89, + "learning_rate": 3.3137832452568654e-05, + "loss": 0.9796, + "step": 195140 + }, + { + "epoch": 16.89, + "learning_rate": 3.313696612665685e-05, + "loss": 0.9733, + "step": 195150 + }, + { + "epoch": 16.89, + "learning_rate": 3.313609980074504e-05, + "loss": 1.0688, + "step": 195160 + }, + { + "epoch": 16.89, + "learning_rate": 3.313523347483323e-05, + "loss": 0.9981, + "step": 195170 + }, + { + "epoch": 16.89, + "learning_rate": 3.3134367148921425e-05, + "loss": 1.045, + "step": 195180 + }, + { + "epoch": 16.89, + "learning_rate": 3.313350082300962e-05, + "loss": 1.0032, + "step": 195190 + }, + { + "epoch": 16.9, + "learning_rate": 3.313263449709781e-05, + "loss": 1.0263, + "step": 195200 + }, + { + "epoch": 16.9, + "learning_rate": 3.3131768171186e-05, + "loss": 1.0273, + "step": 195210 + }, + { + "epoch": 16.9, + "learning_rate": 3.313090184527419e-05, + "loss": 0.99, + "step": 195220 + }, + { + "epoch": 16.9, + "learning_rate": 3.3130035519362384e-05, + "loss": 1.0215, + "step": 195230 + }, + { + "epoch": 16.9, + "learning_rate": 3.312916919345058e-05, + "loss": 1.0341, + "step": 195240 + }, + { + "epoch": 16.9, + "learning_rate": 3.3128302867538766e-05, + "loss": 1.0755, + "step": 195250 + }, + { + "epoch": 16.9, + "learning_rate": 3.312743654162696e-05, + "loss": 1.0355, + "step": 195260 + }, + { + "epoch": 16.9, + "learning_rate": 3.3126570215715155e-05, + "loss": 1.0589, + "step": 195270 + }, + { + "epoch": 16.9, + "learning_rate": 3.312570388980334e-05, + "loss": 1.0313, + "step": 195280 + }, + { + "epoch": 16.9, + "learning_rate": 3.312483756389154e-05, + "loss": 1.0013, + "step": 195290 + }, + { + "epoch": 16.9, + "learning_rate": 3.312397123797973e-05, + "loss": 0.9658, + "step": 195300 + }, + { + "epoch": 16.91, + "learning_rate": 3.312310491206792e-05, + "loss": 1.014, + "step": 195310 + }, + { + "epoch": 16.91, + "learning_rate": 3.3122238586156114e-05, + "loss": 1.0096, + "step": 195320 + }, + { + "epoch": 16.91, + "learning_rate": 3.31213722602443e-05, + "loss": 1.0045, + "step": 195330 + }, + { + "epoch": 16.91, + "learning_rate": 3.3120505934332496e-05, + "loss": 0.9943, + "step": 195340 + }, + { + "epoch": 16.91, + "learning_rate": 3.311963960842069e-05, + "loss": 1.007, + "step": 195350 + }, + { + "epoch": 16.91, + "learning_rate": 3.311877328250888e-05, + "loss": 1.0003, + "step": 195360 + }, + { + "epoch": 16.91, + "learning_rate": 3.311790695659707e-05, + "loss": 1.0387, + "step": 195370 + }, + { + "epoch": 16.91, + "learning_rate": 3.311704063068527e-05, + "loss": 1.0463, + "step": 195380 + }, + { + "epoch": 16.91, + "learning_rate": 3.3116174304773454e-05, + "loss": 0.9497, + "step": 195390 + }, + { + "epoch": 16.91, + "learning_rate": 3.311530797886165e-05, + "loss": 0.9738, + "step": 195400 + }, + { + "epoch": 16.91, + "learning_rate": 3.311444165294984e-05, + "loss": 1.0558, + "step": 195410 + }, + { + "epoch": 16.91, + "learning_rate": 3.311357532703803e-05, + "loss": 0.994, + "step": 195420 + }, + { + "epoch": 16.92, + "learning_rate": 3.3112709001126225e-05, + "loss": 1.0413, + "step": 195430 + }, + { + "epoch": 16.92, + "learning_rate": 3.311184267521441e-05, + "loss": 0.9823, + "step": 195440 + }, + { + "epoch": 16.92, + "learning_rate": 3.311097634930261e-05, + "loss": 1.0665, + "step": 195450 + }, + { + "epoch": 16.92, + "learning_rate": 3.31101100233908e-05, + "loss": 1.0165, + "step": 195460 + }, + { + "epoch": 16.92, + "learning_rate": 3.310924369747899e-05, + "loss": 1.0346, + "step": 195470 + }, + { + "epoch": 16.92, + "learning_rate": 3.3108377371567184e-05, + "loss": 0.986, + "step": 195480 + }, + { + "epoch": 16.92, + "learning_rate": 3.310751104565538e-05, + "loss": 1.0218, + "step": 195490 + }, + { + "epoch": 16.92, + "learning_rate": 3.3106644719743566e-05, + "loss": 1.0441, + "step": 195500 + }, + { + "epoch": 16.92, + "learning_rate": 3.310577839383176e-05, + "loss": 1.0101, + "step": 195510 + }, + { + "epoch": 16.92, + "learning_rate": 3.3104912067919955e-05, + "loss": 0.9932, + "step": 195520 + }, + { + "epoch": 16.92, + "learning_rate": 3.310404574200814e-05, + "loss": 0.9971, + "step": 195530 + }, + { + "epoch": 16.93, + "learning_rate": 3.310317941609634e-05, + "loss": 1.0045, + "step": 195540 + }, + { + "epoch": 16.93, + "learning_rate": 3.3102313090184525e-05, + "loss": 1.088, + "step": 195550 + }, + { + "epoch": 16.93, + "learning_rate": 3.310144676427272e-05, + "loss": 1.0134, + "step": 195560 + }, + { + "epoch": 16.93, + "learning_rate": 3.3100580438360914e-05, + "loss": 0.9805, + "step": 195570 + }, + { + "epoch": 16.93, + "learning_rate": 3.30997141124491e-05, + "loss": 0.9943, + "step": 195580 + }, + { + "epoch": 16.93, + "learning_rate": 3.3098847786537296e-05, + "loss": 1.0287, + "step": 195590 + }, + { + "epoch": 16.93, + "learning_rate": 3.309798146062549e-05, + "loss": 1.0442, + "step": 195600 + }, + { + "epoch": 16.93, + "learning_rate": 3.309711513471368e-05, + "loss": 1.0467, + "step": 195610 + }, + { + "epoch": 16.93, + "learning_rate": 3.309624880880187e-05, + "loss": 1.0537, + "step": 195620 + }, + { + "epoch": 16.93, + "learning_rate": 3.309538248289007e-05, + "loss": 1.0402, + "step": 195630 + }, + { + "epoch": 16.93, + "learning_rate": 3.3094516156978255e-05, + "loss": 1.0221, + "step": 195640 + }, + { + "epoch": 16.93, + "learning_rate": 3.309364983106645e-05, + "loss": 0.9731, + "step": 195650 + }, + { + "epoch": 16.94, + "learning_rate": 3.309278350515464e-05, + "loss": 1.0546, + "step": 195660 + }, + { + "epoch": 16.94, + "learning_rate": 3.309191717924283e-05, + "loss": 1.0408, + "step": 195670 + }, + { + "epoch": 16.94, + "learning_rate": 3.3091050853331026e-05, + "loss": 1.0148, + "step": 195680 + }, + { + "epoch": 16.94, + "learning_rate": 3.3090184527419213e-05, + "loss": 0.9689, + "step": 195690 + }, + { + "epoch": 16.94, + "learning_rate": 3.308931820150741e-05, + "loss": 1.0679, + "step": 195700 + }, + { + "epoch": 16.94, + "learning_rate": 3.30884518755956e-05, + "loss": 0.9827, + "step": 195710 + }, + { + "epoch": 16.94, + "learning_rate": 3.308758554968379e-05, + "loss": 1.0247, + "step": 195720 + }, + { + "epoch": 16.94, + "learning_rate": 3.3086719223771984e-05, + "loss": 1.0019, + "step": 195730 + }, + { + "epoch": 16.94, + "learning_rate": 3.308585289786018e-05, + "loss": 1.0451, + "step": 195740 + }, + { + "epoch": 16.94, + "learning_rate": 3.3084986571948367e-05, + "loss": 1.0038, + "step": 195750 + }, + { + "epoch": 16.94, + "learning_rate": 3.308412024603656e-05, + "loss": 1.0322, + "step": 195760 + }, + { + "epoch": 16.94, + "learning_rate": 3.308325392012475e-05, + "loss": 1.0051, + "step": 195770 + }, + { + "epoch": 16.95, + "learning_rate": 3.308238759421294e-05, + "loss": 0.9859, + "step": 195780 + }, + { + "epoch": 16.95, + "learning_rate": 3.308152126830114e-05, + "loss": 1.0319, + "step": 195790 + }, + { + "epoch": 16.95, + "learning_rate": 3.3080654942389325e-05, + "loss": 0.9921, + "step": 195800 + }, + { + "epoch": 16.95, + "learning_rate": 3.307978861647752e-05, + "loss": 1.0782, + "step": 195810 + }, + { + "epoch": 16.95, + "learning_rate": 3.3078922290565714e-05, + "loss": 1.0464, + "step": 195820 + }, + { + "epoch": 16.95, + "learning_rate": 3.30780559646539e-05, + "loss": 1.0094, + "step": 195830 + }, + { + "epoch": 16.95, + "learning_rate": 3.3077189638742096e-05, + "loss": 0.9989, + "step": 195840 + }, + { + "epoch": 16.95, + "learning_rate": 3.307632331283029e-05, + "loss": 1.0674, + "step": 195850 + }, + { + "epoch": 16.95, + "learning_rate": 3.307545698691848e-05, + "loss": 0.9831, + "step": 195860 + }, + { + "epoch": 16.95, + "learning_rate": 3.307459066100667e-05, + "loss": 0.9952, + "step": 195870 + }, + { + "epoch": 16.95, + "learning_rate": 3.307372433509486e-05, + "loss": 1.0587, + "step": 195880 + }, + { + "epoch": 16.96, + "learning_rate": 3.3072858009183055e-05, + "loss": 0.9639, + "step": 195890 + }, + { + "epoch": 16.96, + "learning_rate": 3.307199168327125e-05, + "loss": 1.05, + "step": 195900 + }, + { + "epoch": 16.96, + "learning_rate": 3.307112535735944e-05, + "loss": 1.0256, + "step": 195910 + }, + { + "epoch": 16.96, + "learning_rate": 3.307025903144763e-05, + "loss": 1.016, + "step": 195920 + }, + { + "epoch": 16.96, + "learning_rate": 3.3069392705535826e-05, + "loss": 1.0353, + "step": 195930 + }, + { + "epoch": 16.96, + "learning_rate": 3.3068526379624014e-05, + "loss": 1.0718, + "step": 195940 + }, + { + "epoch": 16.96, + "learning_rate": 3.306766005371221e-05, + "loss": 1.0063, + "step": 195950 + }, + { + "epoch": 16.96, + "learning_rate": 3.3066793727800396e-05, + "loss": 0.9997, + "step": 195960 + }, + { + "epoch": 16.96, + "learning_rate": 3.306592740188859e-05, + "loss": 1.007, + "step": 195970 + }, + { + "epoch": 16.96, + "learning_rate": 3.3065061075976785e-05, + "loss": 1.0284, + "step": 195980 + }, + { + "epoch": 16.96, + "learning_rate": 3.306419475006497e-05, + "loss": 1.072, + "step": 195990 + }, + { + "epoch": 16.96, + "learning_rate": 3.306332842415317e-05, + "loss": 1.0606, + "step": 196000 + }, + { + "epoch": 16.97, + "learning_rate": 3.306246209824136e-05, + "loss": 1.0042, + "step": 196010 + }, + { + "epoch": 16.97, + "learning_rate": 3.306159577232955e-05, + "loss": 0.9949, + "step": 196020 + }, + { + "epoch": 16.97, + "learning_rate": 3.3060729446417744e-05, + "loss": 1.0916, + "step": 196030 + }, + { + "epoch": 16.97, + "learning_rate": 3.305986312050594e-05, + "loss": 1.0269, + "step": 196040 + }, + { + "epoch": 16.97, + "learning_rate": 3.3058996794594126e-05, + "loss": 1.098, + "step": 196050 + }, + { + "epoch": 16.97, + "learning_rate": 3.305813046868232e-05, + "loss": 1.0, + "step": 196060 + }, + { + "epoch": 16.97, + "learning_rate": 3.305726414277051e-05, + "loss": 1.0525, + "step": 196070 + }, + { + "epoch": 16.97, + "learning_rate": 3.30563978168587e-05, + "loss": 1.0228, + "step": 196080 + }, + { + "epoch": 16.97, + "learning_rate": 3.30555314909469e-05, + "loss": 1.028, + "step": 196090 + }, + { + "epoch": 16.97, + "learning_rate": 3.3054665165035084e-05, + "loss": 1.0266, + "step": 196100 + }, + { + "epoch": 16.97, + "learning_rate": 3.305379883912328e-05, + "loss": 1.0, + "step": 196110 + }, + { + "epoch": 16.98, + "learning_rate": 3.305293251321147e-05, + "loss": 1.0048, + "step": 196120 + }, + { + "epoch": 16.98, + "learning_rate": 3.305206618729966e-05, + "loss": 0.9409, + "step": 196130 + }, + { + "epoch": 16.98, + "learning_rate": 3.3051199861387855e-05, + "loss": 0.9463, + "step": 196140 + }, + { + "epoch": 16.98, + "learning_rate": 3.305033353547605e-05, + "loss": 0.9789, + "step": 196150 + }, + { + "epoch": 16.98, + "learning_rate": 3.304946720956424e-05, + "loss": 0.9939, + "step": 196160 + }, + { + "epoch": 16.98, + "learning_rate": 3.304860088365243e-05, + "loss": 1.0071, + "step": 196170 + }, + { + "epoch": 16.98, + "learning_rate": 3.304773455774062e-05, + "loss": 0.9861, + "step": 196180 + }, + { + "epoch": 16.98, + "learning_rate": 3.3046868231828814e-05, + "loss": 1.0078, + "step": 196190 + }, + { + "epoch": 16.98, + "learning_rate": 3.304600190591701e-05, + "loss": 0.9623, + "step": 196200 + }, + { + "epoch": 16.98, + "learning_rate": 3.3045135580005196e-05, + "loss": 1.0164, + "step": 196210 + }, + { + "epoch": 16.98, + "learning_rate": 3.304426925409339e-05, + "loss": 1.0218, + "step": 196220 + }, + { + "epoch": 16.98, + "learning_rate": 3.3043402928181585e-05, + "loss": 1.0446, + "step": 196230 + }, + { + "epoch": 16.99, + "learning_rate": 3.304253660226977e-05, + "loss": 1.0273, + "step": 196240 + }, + { + "epoch": 16.99, + "learning_rate": 3.304167027635797e-05, + "loss": 1.0086, + "step": 196250 + }, + { + "epoch": 16.99, + "learning_rate": 3.304080395044616e-05, + "loss": 1.0065, + "step": 196260 + }, + { + "epoch": 16.99, + "learning_rate": 3.303993762453435e-05, + "loss": 1.0409, + "step": 196270 + }, + { + "epoch": 16.99, + "learning_rate": 3.3039071298622544e-05, + "loss": 1.0214, + "step": 196280 + }, + { + "epoch": 16.99, + "learning_rate": 3.303820497271073e-05, + "loss": 1.0246, + "step": 196290 + }, + { + "epoch": 16.99, + "learning_rate": 3.3037338646798926e-05, + "loss": 1.0092, + "step": 196300 + }, + { + "epoch": 16.99, + "learning_rate": 3.303647232088712e-05, + "loss": 0.9685, + "step": 196310 + }, + { + "epoch": 16.99, + "learning_rate": 3.303560599497531e-05, + "loss": 0.9731, + "step": 196320 + }, + { + "epoch": 16.99, + "learning_rate": 3.30347396690635e-05, + "loss": 0.9961, + "step": 196330 + }, + { + "epoch": 16.99, + "learning_rate": 3.30338733431517e-05, + "loss": 1.0314, + "step": 196340 + }, + { + "epoch": 17.0, + "learning_rate": 3.3033007017239885e-05, + "loss": 0.9811, + "step": 196350 + }, + { + "epoch": 17.0, + "learning_rate": 3.303214069132808e-05, + "loss": 0.9931, + "step": 196360 + }, + { + "epoch": 17.0, + "learning_rate": 3.3031274365416274e-05, + "loss": 0.9693, + "step": 196370 + }, + { + "epoch": 17.0, + "learning_rate": 3.303040803950446e-05, + "loss": 1.0189, + "step": 196380 + }, + { + "epoch": 17.0, + "learning_rate": 3.3029541713592656e-05, + "loss": 1.0432, + "step": 196390 + }, + { + "epoch": 17.0, + "learning_rate": 3.3028675387680843e-05, + "loss": 1.0438, + "step": 196400 + }, + { + "epoch": 17.0, + "eval_Bleu_1": 0.04139911853551376, + "eval_Bleu_2": 2.7324523940252082e-11, + "eval_Bleu_3": 2.4561103263122834e-14, + "eval_Bleu_4": 7.560511134976592e-16, + "eval_ROUGE_L": 0.08484606082232739, + "eval_cer": 0.9933107128896735, + "eval_em": 0, + "eval_f1": 0.10970863563576468, + "eval_loss": 0.9939275979995728, + "eval_runtime": 2096.4663, + "eval_samples_per_second": 2.449, + "eval_steps_per_second": 2.449, + "eval_wer": 0.9693320502242215, + "step": 196406 + }, + { + "epoch": 17.0, + "learning_rate": 3.302780906176904e-05, + "loss": 1.0251, + "step": 196410 + }, + { + "epoch": 17.0, + "learning_rate": 3.302694273585723e-05, + "loss": 0.9778, + "step": 196420 + }, + { + "epoch": 17.0, + "learning_rate": 3.302607640994542e-05, + "loss": 0.9908, + "step": 196430 + }, + { + "epoch": 17.0, + "learning_rate": 3.3025210084033614e-05, + "loss": 1.0345, + "step": 196440 + }, + { + "epoch": 17.0, + "learning_rate": 3.302434375812181e-05, + "loss": 0.9713, + "step": 196450 + }, + { + "epoch": 17.0, + "learning_rate": 3.3023477432209997e-05, + "loss": 0.9882, + "step": 196460 + }, + { + "epoch": 17.01, + "learning_rate": 3.302261110629819e-05, + "loss": 1.0277, + "step": 196470 + }, + { + "epoch": 17.01, + "learning_rate": 3.3021744780386385e-05, + "loss": 1.0362, + "step": 196480 + }, + { + "epoch": 17.01, + "learning_rate": 3.302087845447457e-05, + "loss": 0.9511, + "step": 196490 + }, + { + "epoch": 17.01, + "learning_rate": 3.302001212856277e-05, + "loss": 1.036, + "step": 196500 + }, + { + "epoch": 17.01, + "learning_rate": 3.3019145802650955e-05, + "loss": 1.0193, + "step": 196510 + }, + { + "epoch": 17.01, + "learning_rate": 3.301827947673915e-05, + "loss": 0.9815, + "step": 196520 + }, + { + "epoch": 17.01, + "learning_rate": 3.3017413150827344e-05, + "loss": 0.9833, + "step": 196530 + }, + { + "epoch": 17.01, + "learning_rate": 3.301654682491553e-05, + "loss": 1.0088, + "step": 196540 + }, + { + "epoch": 17.01, + "learning_rate": 3.3015680499003726e-05, + "loss": 1.0183, + "step": 196550 + }, + { + "epoch": 17.01, + "learning_rate": 3.301481417309192e-05, + "loss": 0.9143, + "step": 196560 + }, + { + "epoch": 17.01, + "learning_rate": 3.301394784718011e-05, + "loss": 1.0115, + "step": 196570 + }, + { + "epoch": 17.02, + "learning_rate": 3.30130815212683e-05, + "loss": 0.9874, + "step": 196580 + }, + { + "epoch": 17.02, + "learning_rate": 3.30122151953565e-05, + "loss": 1.03, + "step": 196590 + }, + { + "epoch": 17.02, + "learning_rate": 3.3011348869444685e-05, + "loss": 1.0464, + "step": 196600 + }, + { + "epoch": 17.02, + "learning_rate": 3.301048254353288e-05, + "loss": 1.0079, + "step": 196610 + }, + { + "epoch": 17.02, + "learning_rate": 3.300961621762107e-05, + "loss": 0.9729, + "step": 196620 + }, + { + "epoch": 17.02, + "learning_rate": 3.300874989170926e-05, + "loss": 0.9619, + "step": 196630 + }, + { + "epoch": 17.02, + "learning_rate": 3.3007883565797456e-05, + "loss": 1.1019, + "step": 196640 + }, + { + "epoch": 17.02, + "learning_rate": 3.3007017239885644e-05, + "loss": 1.0224, + "step": 196650 + }, + { + "epoch": 17.02, + "learning_rate": 3.300615091397384e-05, + "loss": 1.0164, + "step": 196660 + }, + { + "epoch": 17.02, + "learning_rate": 3.300528458806203e-05, + "loss": 0.9769, + "step": 196670 + }, + { + "epoch": 17.02, + "learning_rate": 3.300441826215022e-05, + "loss": 1.0252, + "step": 196680 + }, + { + "epoch": 17.02, + "learning_rate": 3.3003551936238415e-05, + "loss": 0.986, + "step": 196690 + }, + { + "epoch": 17.03, + "learning_rate": 3.30026856103266e-05, + "loss": 1.1032, + "step": 196700 + }, + { + "epoch": 17.03, + "learning_rate": 3.30018192844148e-05, + "loss": 1.0262, + "step": 196710 + }, + { + "epoch": 17.03, + "learning_rate": 3.300095295850299e-05, + "loss": 1.065, + "step": 196720 + }, + { + "epoch": 17.03, + "learning_rate": 3.300008663259118e-05, + "loss": 0.9417, + "step": 196730 + }, + { + "epoch": 17.03, + "learning_rate": 3.2999220306679373e-05, + "loss": 0.958, + "step": 196740 + }, + { + "epoch": 17.03, + "learning_rate": 3.299835398076757e-05, + "loss": 0.9924, + "step": 196750 + }, + { + "epoch": 17.03, + "learning_rate": 3.2997487654855756e-05, + "loss": 0.9861, + "step": 196760 + }, + { + "epoch": 17.03, + "learning_rate": 3.299662132894395e-05, + "loss": 1.0279, + "step": 196770 + }, + { + "epoch": 17.03, + "learning_rate": 3.2995755003032145e-05, + "loss": 0.9786, + "step": 196780 + }, + { + "epoch": 17.03, + "learning_rate": 3.299488867712033e-05, + "loss": 0.9717, + "step": 196790 + }, + { + "epoch": 17.03, + "learning_rate": 3.299402235120853e-05, + "loss": 0.9589, + "step": 196800 + }, + { + "epoch": 17.03, + "learning_rate": 3.2993156025296714e-05, + "loss": 1.0031, + "step": 196810 + }, + { + "epoch": 17.04, + "learning_rate": 3.299228969938491e-05, + "loss": 0.9455, + "step": 196820 + }, + { + "epoch": 17.04, + "learning_rate": 3.29914233734731e-05, + "loss": 0.9697, + "step": 196830 + }, + { + "epoch": 17.04, + "learning_rate": 3.299055704756129e-05, + "loss": 0.9765, + "step": 196840 + }, + { + "epoch": 17.04, + "learning_rate": 3.2989690721649485e-05, + "loss": 0.9787, + "step": 196850 + }, + { + "epoch": 17.04, + "learning_rate": 3.298882439573768e-05, + "loss": 0.9576, + "step": 196860 + }, + { + "epoch": 17.04, + "learning_rate": 3.298795806982587e-05, + "loss": 1.0719, + "step": 196870 + }, + { + "epoch": 17.04, + "learning_rate": 3.298709174391406e-05, + "loss": 1.0028, + "step": 196880 + }, + { + "epoch": 17.04, + "learning_rate": 3.2986225418002256e-05, + "loss": 0.9908, + "step": 196890 + }, + { + "epoch": 17.04, + "learning_rate": 3.2985359092090444e-05, + "loss": 1.0015, + "step": 196900 + }, + { + "epoch": 17.04, + "learning_rate": 3.298449276617864e-05, + "loss": 1.0842, + "step": 196910 + }, + { + "epoch": 17.04, + "learning_rate": 3.2983626440266826e-05, + "loss": 0.978, + "step": 196920 + }, + { + "epoch": 17.05, + "learning_rate": 3.298276011435502e-05, + "loss": 0.9823, + "step": 196930 + }, + { + "epoch": 17.05, + "learning_rate": 3.2981893788443215e-05, + "loss": 1.0217, + "step": 196940 + }, + { + "epoch": 17.05, + "learning_rate": 3.29810274625314e-05, + "loss": 0.9497, + "step": 196950 + }, + { + "epoch": 17.05, + "learning_rate": 3.29801611366196e-05, + "loss": 1.0032, + "step": 196960 + }, + { + "epoch": 17.05, + "learning_rate": 3.297929481070779e-05, + "loss": 0.9918, + "step": 196970 + }, + { + "epoch": 17.05, + "learning_rate": 3.297842848479598e-05, + "loss": 1.0055, + "step": 196980 + }, + { + "epoch": 17.05, + "learning_rate": 3.2977562158884174e-05, + "loss": 0.9852, + "step": 196990 + }, + { + "epoch": 17.05, + "learning_rate": 3.297669583297237e-05, + "loss": 0.968, + "step": 197000 + }, + { + "epoch": 17.05, + "learning_rate": 3.2975829507060556e-05, + "loss": 1.0365, + "step": 197010 + }, + { + "epoch": 17.05, + "learning_rate": 3.297496318114875e-05, + "loss": 1.0004, + "step": 197020 + }, + { + "epoch": 17.05, + "learning_rate": 3.297409685523694e-05, + "loss": 0.9933, + "step": 197030 + }, + { + "epoch": 17.05, + "learning_rate": 3.297323052932513e-05, + "loss": 1.0675, + "step": 197040 + }, + { + "epoch": 17.06, + "learning_rate": 3.297236420341333e-05, + "loss": 0.9651, + "step": 197050 + }, + { + "epoch": 17.06, + "learning_rate": 3.2971497877501515e-05, + "loss": 0.9834, + "step": 197060 + }, + { + "epoch": 17.06, + "learning_rate": 3.297063155158971e-05, + "loss": 1.0289, + "step": 197070 + }, + { + "epoch": 17.06, + "learning_rate": 3.2969765225677904e-05, + "loss": 1.029, + "step": 197080 + }, + { + "epoch": 17.06, + "learning_rate": 3.296889889976609e-05, + "loss": 0.9422, + "step": 197090 + }, + { + "epoch": 17.06, + "learning_rate": 3.2968032573854286e-05, + "loss": 0.988, + "step": 197100 + }, + { + "epoch": 17.06, + "learning_rate": 3.296716624794248e-05, + "loss": 1.0075, + "step": 197110 + }, + { + "epoch": 17.06, + "learning_rate": 3.296629992203067e-05, + "loss": 1.0788, + "step": 197120 + }, + { + "epoch": 17.06, + "learning_rate": 3.296543359611886e-05, + "loss": 0.9874, + "step": 197130 + }, + { + "epoch": 17.06, + "learning_rate": 3.296456727020705e-05, + "loss": 0.9916, + "step": 197140 + }, + { + "epoch": 17.06, + "learning_rate": 3.2963700944295244e-05, + "loss": 0.972, + "step": 197150 + }, + { + "epoch": 17.07, + "learning_rate": 3.296283461838344e-05, + "loss": 0.971, + "step": 197160 + }, + { + "epoch": 17.07, + "learning_rate": 3.2961968292471627e-05, + "loss": 1.0369, + "step": 197170 + }, + { + "epoch": 17.07, + "learning_rate": 3.296110196655982e-05, + "loss": 1.0227, + "step": 197180 + }, + { + "epoch": 17.07, + "learning_rate": 3.2960235640648015e-05, + "loss": 1.0398, + "step": 197190 + }, + { + "epoch": 17.07, + "learning_rate": 3.29593693147362e-05, + "loss": 0.9771, + "step": 197200 + }, + { + "epoch": 17.07, + "learning_rate": 3.29585029888244e-05, + "loss": 0.9416, + "step": 197210 + }, + { + "epoch": 17.07, + "learning_rate": 3.295763666291259e-05, + "loss": 0.9911, + "step": 197220 + }, + { + "epoch": 17.07, + "learning_rate": 3.295677033700078e-05, + "loss": 0.9809, + "step": 197230 + }, + { + "epoch": 17.07, + "learning_rate": 3.2955904011088974e-05, + "loss": 1.0275, + "step": 197240 + }, + { + "epoch": 17.07, + "learning_rate": 3.295503768517716e-05, + "loss": 0.9711, + "step": 197250 + }, + { + "epoch": 17.07, + "learning_rate": 3.2954171359265356e-05, + "loss": 0.9235, + "step": 197260 + }, + { + "epoch": 17.07, + "learning_rate": 3.295330503335355e-05, + "loss": 1.0205, + "step": 197270 + }, + { + "epoch": 17.08, + "learning_rate": 3.295243870744174e-05, + "loss": 1.0115, + "step": 197280 + }, + { + "epoch": 17.08, + "learning_rate": 3.295157238152993e-05, + "loss": 0.9862, + "step": 197290 + }, + { + "epoch": 17.08, + "learning_rate": 3.295070605561813e-05, + "loss": 1.0488, + "step": 197300 + }, + { + "epoch": 17.08, + "learning_rate": 3.2949839729706315e-05, + "loss": 0.9786, + "step": 197310 + }, + { + "epoch": 17.08, + "learning_rate": 3.294897340379451e-05, + "loss": 0.9801, + "step": 197320 + }, + { + "epoch": 17.08, + "learning_rate": 3.2948107077882704e-05, + "loss": 1.008, + "step": 197330 + }, + { + "epoch": 17.08, + "learning_rate": 3.294724075197089e-05, + "loss": 0.9542, + "step": 197340 + }, + { + "epoch": 17.08, + "learning_rate": 3.2946374426059086e-05, + "loss": 1.0101, + "step": 197350 + }, + { + "epoch": 17.08, + "learning_rate": 3.2945508100147274e-05, + "loss": 1.0376, + "step": 197360 + }, + { + "epoch": 17.08, + "learning_rate": 3.294464177423547e-05, + "loss": 1.0661, + "step": 197370 + }, + { + "epoch": 17.08, + "learning_rate": 3.294377544832366e-05, + "loss": 1.0329, + "step": 197380 + }, + { + "epoch": 17.09, + "learning_rate": 3.294290912241185e-05, + "loss": 1.0295, + "step": 197390 + }, + { + "epoch": 17.09, + "learning_rate": 3.2942042796500045e-05, + "loss": 1.0407, + "step": 197400 + }, + { + "epoch": 17.09, + "learning_rate": 3.294117647058824e-05, + "loss": 0.9807, + "step": 197410 + }, + { + "epoch": 17.09, + "learning_rate": 3.294031014467643e-05, + "loss": 1.0111, + "step": 197420 + }, + { + "epoch": 17.09, + "learning_rate": 3.293944381876462e-05, + "loss": 0.9708, + "step": 197430 + }, + { + "epoch": 17.09, + "learning_rate": 3.293857749285281e-05, + "loss": 0.9913, + "step": 197440 + }, + { + "epoch": 17.09, + "learning_rate": 3.2937711166941003e-05, + "loss": 0.9878, + "step": 197450 + }, + { + "epoch": 17.09, + "learning_rate": 3.29368448410292e-05, + "loss": 1.0234, + "step": 197460 + }, + { + "epoch": 17.09, + "learning_rate": 3.2935978515117386e-05, + "loss": 0.9646, + "step": 197470 + }, + { + "epoch": 17.09, + "learning_rate": 3.293511218920558e-05, + "loss": 0.9694, + "step": 197480 + }, + { + "epoch": 17.09, + "learning_rate": 3.2934245863293774e-05, + "loss": 0.9843, + "step": 197490 + }, + { + "epoch": 17.09, + "learning_rate": 3.293337953738196e-05, + "loss": 0.9881, + "step": 197500 + }, + { + "epoch": 17.1, + "learning_rate": 3.2932513211470157e-05, + "loss": 0.9921, + "step": 197510 + }, + { + "epoch": 17.1, + "learning_rate": 3.293164688555835e-05, + "loss": 0.9935, + "step": 197520 + }, + { + "epoch": 17.1, + "learning_rate": 3.293078055964654e-05, + "loss": 1.0567, + "step": 197530 + }, + { + "epoch": 17.1, + "learning_rate": 3.292991423373473e-05, + "loss": 0.9226, + "step": 197540 + }, + { + "epoch": 17.1, + "learning_rate": 3.292904790782292e-05, + "loss": 0.9675, + "step": 197550 + }, + { + "epoch": 17.1, + "learning_rate": 3.2928181581911115e-05, + "loss": 1.0178, + "step": 197560 + }, + { + "epoch": 17.1, + "learning_rate": 3.292731525599931e-05, + "loss": 0.985, + "step": 197570 + }, + { + "epoch": 17.1, + "learning_rate": 3.29264489300875e-05, + "loss": 1.0194, + "step": 197580 + }, + { + "epoch": 17.1, + "learning_rate": 3.292558260417569e-05, + "loss": 1.0224, + "step": 197590 + }, + { + "epoch": 17.1, + "learning_rate": 3.2924716278263886e-05, + "loss": 0.9955, + "step": 197600 + }, + { + "epoch": 17.1, + "learning_rate": 3.2923849952352074e-05, + "loss": 0.9967, + "step": 197610 + }, + { + "epoch": 17.11, + "learning_rate": 3.292298362644027e-05, + "loss": 1.0423, + "step": 197620 + }, + { + "epoch": 17.11, + "learning_rate": 3.292211730052846e-05, + "loss": 0.9803, + "step": 197630 + }, + { + "epoch": 17.11, + "learning_rate": 3.292125097461665e-05, + "loss": 0.9631, + "step": 197640 + }, + { + "epoch": 17.11, + "learning_rate": 3.2920384648704845e-05, + "loss": 1.0102, + "step": 197650 + }, + { + "epoch": 17.11, + "learning_rate": 3.291951832279303e-05, + "loss": 1.0263, + "step": 197660 + }, + { + "epoch": 17.11, + "learning_rate": 3.291865199688123e-05, + "loss": 1.0324, + "step": 197670 + }, + { + "epoch": 17.11, + "learning_rate": 3.291778567096942e-05, + "loss": 1.0257, + "step": 197680 + }, + { + "epoch": 17.11, + "learning_rate": 3.291691934505761e-05, + "loss": 0.992, + "step": 197690 + }, + { + "epoch": 17.11, + "learning_rate": 3.2916053019145804e-05, + "loss": 1.018, + "step": 197700 + }, + { + "epoch": 17.11, + "learning_rate": 3.2915186693234e-05, + "loss": 1.0238, + "step": 197710 + }, + { + "epoch": 17.11, + "learning_rate": 3.2914320367322186e-05, + "loss": 1.0446, + "step": 197720 + }, + { + "epoch": 17.11, + "learning_rate": 3.291345404141038e-05, + "loss": 1.0211, + "step": 197730 + }, + { + "epoch": 17.12, + "learning_rate": 3.2912587715498575e-05, + "loss": 0.9602, + "step": 197740 + }, + { + "epoch": 17.12, + "learning_rate": 3.291172138958676e-05, + "loss": 0.9988, + "step": 197750 + }, + { + "epoch": 17.12, + "learning_rate": 3.291085506367496e-05, + "loss": 1.0148, + "step": 197760 + }, + { + "epoch": 17.12, + "learning_rate": 3.2909988737763145e-05, + "loss": 0.9974, + "step": 197770 + }, + { + "epoch": 17.12, + "learning_rate": 3.290912241185134e-05, + "loss": 0.9947, + "step": 197780 + }, + { + "epoch": 17.12, + "learning_rate": 3.2908256085939534e-05, + "loss": 1.0263, + "step": 197790 + }, + { + "epoch": 17.12, + "learning_rate": 3.290738976002772e-05, + "loss": 1.0076, + "step": 197800 + }, + { + "epoch": 17.12, + "learning_rate": 3.2906523434115916e-05, + "loss": 1.017, + "step": 197810 + }, + { + "epoch": 17.12, + "learning_rate": 3.290565710820411e-05, + "loss": 1.0229, + "step": 197820 + }, + { + "epoch": 17.12, + "learning_rate": 3.29047907822923e-05, + "loss": 1.0136, + "step": 197830 + }, + { + "epoch": 17.12, + "learning_rate": 3.290392445638049e-05, + "loss": 0.9934, + "step": 197840 + }, + { + "epoch": 17.12, + "learning_rate": 3.290305813046869e-05, + "loss": 1.0179, + "step": 197850 + }, + { + "epoch": 17.13, + "learning_rate": 3.2902191804556874e-05, + "loss": 0.985, + "step": 197860 + }, + { + "epoch": 17.13, + "learning_rate": 3.290132547864507e-05, + "loss": 0.8896, + "step": 197870 + }, + { + "epoch": 17.13, + "learning_rate": 3.2900459152733256e-05, + "loss": 0.9432, + "step": 197880 + }, + { + "epoch": 17.13, + "learning_rate": 3.289959282682145e-05, + "loss": 1.0141, + "step": 197890 + }, + { + "epoch": 17.13, + "learning_rate": 3.2898726500909645e-05, + "loss": 1.0017, + "step": 197900 + }, + { + "epoch": 17.13, + "learning_rate": 3.289786017499783e-05, + "loss": 0.9792, + "step": 197910 + }, + { + "epoch": 17.13, + "learning_rate": 3.289699384908603e-05, + "loss": 0.9849, + "step": 197920 + }, + { + "epoch": 17.13, + "learning_rate": 3.289612752317422e-05, + "loss": 1.0199, + "step": 197930 + }, + { + "epoch": 17.13, + "learning_rate": 3.289526119726241e-05, + "loss": 1.0086, + "step": 197940 + }, + { + "epoch": 17.13, + "learning_rate": 3.2894394871350604e-05, + "loss": 0.9875, + "step": 197950 + }, + { + "epoch": 17.13, + "learning_rate": 3.28935285454388e-05, + "loss": 1.002, + "step": 197960 + }, + { + "epoch": 17.14, + "learning_rate": 3.2892662219526986e-05, + "loss": 0.948, + "step": 197970 + }, + { + "epoch": 17.14, + "learning_rate": 3.289179589361518e-05, + "loss": 1.0184, + "step": 197980 + }, + { + "epoch": 17.14, + "learning_rate": 3.289092956770337e-05, + "loss": 0.9783, + "step": 197990 + }, + { + "epoch": 17.14, + "learning_rate": 3.289006324179156e-05, + "loss": 1.0113, + "step": 198000 + }, + { + "epoch": 17.14, + "learning_rate": 3.288919691587976e-05, + "loss": 1.0124, + "step": 198010 + }, + { + "epoch": 17.14, + "learning_rate": 3.2888330589967945e-05, + "loss": 0.9512, + "step": 198020 + }, + { + "epoch": 17.14, + "learning_rate": 3.288746426405614e-05, + "loss": 0.9574, + "step": 198030 + }, + { + "epoch": 17.14, + "learning_rate": 3.2886597938144334e-05, + "loss": 0.9887, + "step": 198040 + }, + { + "epoch": 17.14, + "learning_rate": 3.288573161223252e-05, + "loss": 1.0018, + "step": 198050 + }, + { + "epoch": 17.14, + "learning_rate": 3.2884865286320716e-05, + "loss": 1.0531, + "step": 198060 + }, + { + "epoch": 17.14, + "learning_rate": 3.288399896040891e-05, + "loss": 0.9897, + "step": 198070 + }, + { + "epoch": 17.14, + "learning_rate": 3.28831326344971e-05, + "loss": 1.0232, + "step": 198080 + }, + { + "epoch": 17.15, + "learning_rate": 3.288226630858529e-05, + "loss": 1.0142, + "step": 198090 + }, + { + "epoch": 17.15, + "learning_rate": 3.288139998267348e-05, + "loss": 0.9805, + "step": 198100 + }, + { + "epoch": 17.15, + "learning_rate": 3.2880533656761675e-05, + "loss": 0.9985, + "step": 198110 + }, + { + "epoch": 17.15, + "learning_rate": 3.287966733084987e-05, + "loss": 1.0207, + "step": 198120 + }, + { + "epoch": 17.15, + "learning_rate": 3.287880100493806e-05, + "loss": 1.0297, + "step": 198130 + }, + { + "epoch": 17.15, + "learning_rate": 3.287793467902625e-05, + "loss": 0.9782, + "step": 198140 + }, + { + "epoch": 17.15, + "learning_rate": 3.2877068353114446e-05, + "loss": 0.99, + "step": 198150 + }, + { + "epoch": 17.15, + "learning_rate": 3.2876202027202633e-05, + "loss": 1.0185, + "step": 198160 + }, + { + "epoch": 17.15, + "learning_rate": 3.287533570129083e-05, + "loss": 1.0414, + "step": 198170 + }, + { + "epoch": 17.15, + "learning_rate": 3.287446937537902e-05, + "loss": 1.0063, + "step": 198180 + }, + { + "epoch": 17.15, + "learning_rate": 3.287360304946721e-05, + "loss": 1.003, + "step": 198190 + }, + { + "epoch": 17.16, + "learning_rate": 3.2872736723555404e-05, + "loss": 0.9365, + "step": 198200 + }, + { + "epoch": 17.16, + "learning_rate": 3.287187039764359e-05, + "loss": 0.9472, + "step": 198210 + }, + { + "epoch": 17.16, + "learning_rate": 3.2871004071731787e-05, + "loss": 1.0339, + "step": 198220 + }, + { + "epoch": 17.16, + "learning_rate": 3.287013774581998e-05, + "loss": 1.0023, + "step": 198230 + }, + { + "epoch": 17.16, + "learning_rate": 3.286927141990817e-05, + "loss": 1.013, + "step": 198240 + }, + { + "epoch": 17.16, + "learning_rate": 3.286840509399636e-05, + "loss": 0.9864, + "step": 198250 + }, + { + "epoch": 17.16, + "learning_rate": 3.286753876808456e-05, + "loss": 0.9544, + "step": 198260 + }, + { + "epoch": 17.16, + "learning_rate": 3.2866672442172745e-05, + "loss": 1.0918, + "step": 198270 + }, + { + "epoch": 17.16, + "learning_rate": 3.286580611626094e-05, + "loss": 1.0186, + "step": 198280 + }, + { + "epoch": 17.16, + "learning_rate": 3.286493979034913e-05, + "loss": 1.0435, + "step": 198290 + }, + { + "epoch": 17.16, + "learning_rate": 3.286407346443732e-05, + "loss": 1.0767, + "step": 198300 + }, + { + "epoch": 17.16, + "learning_rate": 3.2863207138525516e-05, + "loss": 1.0071, + "step": 198310 + }, + { + "epoch": 17.17, + "learning_rate": 3.2862340812613704e-05, + "loss": 0.9415, + "step": 198320 + }, + { + "epoch": 17.17, + "learning_rate": 3.28614744867019e-05, + "loss": 0.9743, + "step": 198330 + }, + { + "epoch": 17.17, + "learning_rate": 3.286060816079009e-05, + "loss": 1.0375, + "step": 198340 + }, + { + "epoch": 17.17, + "learning_rate": 3.285974183487828e-05, + "loss": 1.0044, + "step": 198350 + }, + { + "epoch": 17.17, + "learning_rate": 3.2858875508966475e-05, + "loss": 0.9747, + "step": 198360 + }, + { + "epoch": 17.17, + "learning_rate": 3.285800918305467e-05, + "loss": 0.9761, + "step": 198370 + }, + { + "epoch": 17.17, + "learning_rate": 3.285714285714286e-05, + "loss": 0.996, + "step": 198380 + }, + { + "epoch": 17.17, + "learning_rate": 3.285627653123105e-05, + "loss": 0.9982, + "step": 198390 + }, + { + "epoch": 17.17, + "learning_rate": 3.285541020531924e-05, + "loss": 0.9403, + "step": 198400 + }, + { + "epoch": 17.17, + "learning_rate": 3.2854543879407434e-05, + "loss": 0.9784, + "step": 198410 + }, + { + "epoch": 17.17, + "learning_rate": 3.285367755349563e-05, + "loss": 1.0021, + "step": 198420 + }, + { + "epoch": 17.18, + "learning_rate": 3.2852811227583816e-05, + "loss": 0.981, + "step": 198430 + }, + { + "epoch": 17.18, + "learning_rate": 3.285194490167201e-05, + "loss": 1.0195, + "step": 198440 + }, + { + "epoch": 17.18, + "learning_rate": 3.2851078575760205e-05, + "loss": 0.9905, + "step": 198450 + }, + { + "epoch": 17.18, + "learning_rate": 3.285021224984839e-05, + "loss": 0.9607, + "step": 198460 + }, + { + "epoch": 17.18, + "learning_rate": 3.284934592393659e-05, + "loss": 0.9596, + "step": 198470 + }, + { + "epoch": 17.18, + "learning_rate": 3.284847959802478e-05, + "loss": 1.0082, + "step": 198480 + }, + { + "epoch": 17.18, + "learning_rate": 3.284761327211297e-05, + "loss": 1.0479, + "step": 198490 + }, + { + "epoch": 17.18, + "learning_rate": 3.2846746946201163e-05, + "loss": 1.0031, + "step": 198500 + }, + { + "epoch": 17.18, + "learning_rate": 3.284588062028935e-05, + "loss": 1.0438, + "step": 198510 + }, + { + "epoch": 17.18, + "learning_rate": 3.2845014294377546e-05, + "loss": 1.0199, + "step": 198520 + }, + { + "epoch": 17.18, + "learning_rate": 3.284414796846574e-05, + "loss": 0.989, + "step": 198530 + }, + { + "epoch": 17.18, + "learning_rate": 3.284328164255393e-05, + "loss": 0.9609, + "step": 198540 + }, + { + "epoch": 17.19, + "learning_rate": 3.284241531664212e-05, + "loss": 0.9869, + "step": 198550 + }, + { + "epoch": 17.19, + "learning_rate": 3.284154899073032e-05, + "loss": 1.0406, + "step": 198560 + }, + { + "epoch": 17.19, + "learning_rate": 3.2840682664818504e-05, + "loss": 0.9683, + "step": 198570 + }, + { + "epoch": 17.19, + "learning_rate": 3.28398163389067e-05, + "loss": 1.0244, + "step": 198580 + }, + { + "epoch": 17.19, + "learning_rate": 3.283895001299489e-05, + "loss": 1.0132, + "step": 198590 + }, + { + "epoch": 17.19, + "learning_rate": 3.283808368708308e-05, + "loss": 1.03, + "step": 198600 + }, + { + "epoch": 17.19, + "learning_rate": 3.2837217361171275e-05, + "loss": 1.0557, + "step": 198610 + }, + { + "epoch": 17.19, + "learning_rate": 3.283635103525946e-05, + "loss": 0.908, + "step": 198620 + }, + { + "epoch": 17.19, + "learning_rate": 3.283548470934766e-05, + "loss": 1.011, + "step": 198630 + }, + { + "epoch": 17.19, + "learning_rate": 3.283461838343585e-05, + "loss": 0.9818, + "step": 198640 + }, + { + "epoch": 17.19, + "learning_rate": 3.283375205752404e-05, + "loss": 0.9823, + "step": 198650 + }, + { + "epoch": 17.2, + "learning_rate": 3.2832885731612234e-05, + "loss": 0.9824, + "step": 198660 + }, + { + "epoch": 17.2, + "learning_rate": 3.283201940570043e-05, + "loss": 0.9549, + "step": 198670 + }, + { + "epoch": 17.2, + "learning_rate": 3.2831153079788616e-05, + "loss": 1.0399, + "step": 198680 + }, + { + "epoch": 17.2, + "learning_rate": 3.283028675387681e-05, + "loss": 1.0136, + "step": 198690 + }, + { + "epoch": 17.2, + "learning_rate": 3.2829420427965005e-05, + "loss": 1.004, + "step": 198700 + }, + { + "epoch": 17.2, + "learning_rate": 3.282855410205319e-05, + "loss": 0.983, + "step": 198710 + }, + { + "epoch": 17.2, + "learning_rate": 3.282768777614139e-05, + "loss": 0.9896, + "step": 198720 + }, + { + "epoch": 17.2, + "learning_rate": 3.2826821450229575e-05, + "loss": 0.9961, + "step": 198730 + }, + { + "epoch": 17.2, + "learning_rate": 3.282595512431777e-05, + "loss": 0.9834, + "step": 198740 + }, + { + "epoch": 17.2, + "learning_rate": 3.2825088798405964e-05, + "loss": 1.0183, + "step": 198750 + }, + { + "epoch": 17.2, + "learning_rate": 3.282422247249415e-05, + "loss": 1.0116, + "step": 198760 + }, + { + "epoch": 17.2, + "learning_rate": 3.2823356146582346e-05, + "loss": 0.9442, + "step": 198770 + }, + { + "epoch": 17.21, + "learning_rate": 3.282248982067054e-05, + "loss": 0.9746, + "step": 198780 + }, + { + "epoch": 17.21, + "learning_rate": 3.282162349475873e-05, + "loss": 0.9756, + "step": 198790 + }, + { + "epoch": 17.21, + "learning_rate": 3.282075716884692e-05, + "loss": 1.0024, + "step": 198800 + }, + { + "epoch": 17.21, + "learning_rate": 3.281989084293512e-05, + "loss": 1.0041, + "step": 198810 + }, + { + "epoch": 17.21, + "learning_rate": 3.2819024517023305e-05, + "loss": 1.0169, + "step": 198820 + }, + { + "epoch": 17.21, + "learning_rate": 3.28181581911115e-05, + "loss": 0.9831, + "step": 198830 + }, + { + "epoch": 17.21, + "learning_rate": 3.281729186519969e-05, + "loss": 1.0294, + "step": 198840 + }, + { + "epoch": 17.21, + "learning_rate": 3.281642553928788e-05, + "loss": 0.9308, + "step": 198850 + }, + { + "epoch": 17.21, + "learning_rate": 3.2815559213376076e-05, + "loss": 0.9749, + "step": 198860 + }, + { + "epoch": 17.21, + "learning_rate": 3.281469288746426e-05, + "loss": 0.9824, + "step": 198870 + }, + { + "epoch": 17.21, + "learning_rate": 3.281382656155246e-05, + "loss": 1.0348, + "step": 198880 + }, + { + "epoch": 17.21, + "learning_rate": 3.281296023564065e-05, + "loss": 1.0326, + "step": 198890 + }, + { + "epoch": 17.22, + "learning_rate": 3.281209390972884e-05, + "loss": 0.9822, + "step": 198900 + }, + { + "epoch": 17.22, + "learning_rate": 3.2811227583817034e-05, + "loss": 0.9728, + "step": 198910 + }, + { + "epoch": 17.22, + "learning_rate": 3.281036125790523e-05, + "loss": 0.9912, + "step": 198920 + }, + { + "epoch": 17.22, + "learning_rate": 3.2809494931993417e-05, + "loss": 1.0764, + "step": 198930 + }, + { + "epoch": 17.22, + "learning_rate": 3.280862860608161e-05, + "loss": 1.0241, + "step": 198940 + }, + { + "epoch": 17.22, + "learning_rate": 3.28077622801698e-05, + "loss": 1.0307, + "step": 198950 + }, + { + "epoch": 17.22, + "learning_rate": 3.280689595425799e-05, + "loss": 0.9748, + "step": 198960 + }, + { + "epoch": 17.22, + "learning_rate": 3.280602962834619e-05, + "loss": 1.0084, + "step": 198970 + }, + { + "epoch": 17.22, + "learning_rate": 3.2805163302434375e-05, + "loss": 1.0126, + "step": 198980 + }, + { + "epoch": 17.22, + "learning_rate": 3.280429697652257e-05, + "loss": 0.9626, + "step": 198990 + }, + { + "epoch": 17.22, + "learning_rate": 3.2803430650610764e-05, + "loss": 0.986, + "step": 199000 + }, + { + "epoch": 17.23, + "learning_rate": 3.280256432469895e-05, + "loss": 0.992, + "step": 199010 + }, + { + "epoch": 17.23, + "learning_rate": 3.2801697998787146e-05, + "loss": 1.0286, + "step": 199020 + }, + { + "epoch": 17.23, + "learning_rate": 3.2800831672875334e-05, + "loss": 0.956, + "step": 199030 + }, + { + "epoch": 17.23, + "learning_rate": 3.279996534696353e-05, + "loss": 1.0451, + "step": 199040 + }, + { + "epoch": 17.23, + "learning_rate": 3.279909902105172e-05, + "loss": 1.0147, + "step": 199050 + }, + { + "epoch": 17.23, + "learning_rate": 3.279823269513991e-05, + "loss": 0.9694, + "step": 199060 + }, + { + "epoch": 17.23, + "learning_rate": 3.2797366369228105e-05, + "loss": 0.9671, + "step": 199070 + }, + { + "epoch": 17.23, + "learning_rate": 3.27965000433163e-05, + "loss": 0.9852, + "step": 199080 + }, + { + "epoch": 17.23, + "learning_rate": 3.279563371740449e-05, + "loss": 0.9964, + "step": 199090 + }, + { + "epoch": 17.23, + "learning_rate": 3.279476739149268e-05, + "loss": 1.0173, + "step": 199100 + }, + { + "epoch": 17.23, + "learning_rate": 3.2793901065580876e-05, + "loss": 0.9804, + "step": 199110 + }, + { + "epoch": 17.23, + "learning_rate": 3.2793034739669064e-05, + "loss": 1.0478, + "step": 199120 + }, + { + "epoch": 17.24, + "learning_rate": 3.279216841375726e-05, + "loss": 0.9368, + "step": 199130 + }, + { + "epoch": 17.24, + "learning_rate": 3.2791302087845446e-05, + "loss": 1.0305, + "step": 199140 + }, + { + "epoch": 17.24, + "learning_rate": 3.279043576193364e-05, + "loss": 0.9861, + "step": 199150 + }, + { + "epoch": 17.24, + "learning_rate": 3.2789569436021835e-05, + "loss": 1.0151, + "step": 199160 + }, + { + "epoch": 17.24, + "learning_rate": 3.278870311011002e-05, + "loss": 1.0289, + "step": 199170 + }, + { + "epoch": 17.24, + "learning_rate": 3.278783678419822e-05, + "loss": 1.0483, + "step": 199180 + }, + { + "epoch": 17.24, + "learning_rate": 3.278697045828641e-05, + "loss": 0.991, + "step": 199190 + }, + { + "epoch": 17.24, + "learning_rate": 3.27861041323746e-05, + "loss": 1.0142, + "step": 199200 + }, + { + "epoch": 17.24, + "learning_rate": 3.2785237806462793e-05, + "loss": 1.0004, + "step": 199210 + }, + { + "epoch": 17.24, + "learning_rate": 3.278437148055099e-05, + "loss": 1.0254, + "step": 199220 + }, + { + "epoch": 17.24, + "learning_rate": 3.2783505154639176e-05, + "loss": 1.0392, + "step": 199230 + }, + { + "epoch": 17.25, + "learning_rate": 3.278263882872737e-05, + "loss": 1.0273, + "step": 199240 + }, + { + "epoch": 17.25, + "learning_rate": 3.278177250281556e-05, + "loss": 0.9567, + "step": 199250 + }, + { + "epoch": 17.25, + "learning_rate": 3.278090617690375e-05, + "loss": 1.0128, + "step": 199260 + }, + { + "epoch": 17.25, + "learning_rate": 3.278003985099195e-05, + "loss": 1.0181, + "step": 199270 + }, + { + "epoch": 17.25, + "learning_rate": 3.2779173525080134e-05, + "loss": 1.0608, + "step": 199280 + }, + { + "epoch": 17.25, + "learning_rate": 3.277830719916833e-05, + "loss": 1.0394, + "step": 199290 + }, + { + "epoch": 17.25, + "learning_rate": 3.277744087325652e-05, + "loss": 0.9887, + "step": 199300 + }, + { + "epoch": 17.25, + "learning_rate": 3.277657454734471e-05, + "loss": 1.0023, + "step": 199310 + }, + { + "epoch": 17.25, + "learning_rate": 3.2775708221432905e-05, + "loss": 0.9534, + "step": 199320 + }, + { + "epoch": 17.25, + "learning_rate": 3.27748418955211e-05, + "loss": 1.0157, + "step": 199330 + }, + { + "epoch": 17.25, + "learning_rate": 3.277397556960929e-05, + "loss": 1.0445, + "step": 199340 + }, + { + "epoch": 17.25, + "learning_rate": 3.277310924369748e-05, + "loss": 1.0101, + "step": 199350 + }, + { + "epoch": 17.26, + "learning_rate": 3.277224291778567e-05, + "loss": 0.98, + "step": 199360 + }, + { + "epoch": 17.26, + "learning_rate": 3.2771376591873864e-05, + "loss": 1.0437, + "step": 199370 + }, + { + "epoch": 17.26, + "learning_rate": 3.277051026596206e-05, + "loss": 0.9805, + "step": 199380 + }, + { + "epoch": 17.26, + "learning_rate": 3.2769643940050246e-05, + "loss": 0.9713, + "step": 199390 + }, + { + "epoch": 17.26, + "learning_rate": 3.276877761413844e-05, + "loss": 0.959, + "step": 199400 + }, + { + "epoch": 17.26, + "learning_rate": 3.2767911288226635e-05, + "loss": 0.9886, + "step": 199410 + }, + { + "epoch": 17.26, + "learning_rate": 3.276704496231482e-05, + "loss": 1.0138, + "step": 199420 + }, + { + "epoch": 17.26, + "learning_rate": 3.276617863640302e-05, + "loss": 0.9821, + "step": 199430 + }, + { + "epoch": 17.26, + "learning_rate": 3.276531231049121e-05, + "loss": 1.0354, + "step": 199440 + }, + { + "epoch": 17.26, + "learning_rate": 3.27644459845794e-05, + "loss": 0.952, + "step": 199450 + }, + { + "epoch": 17.26, + "learning_rate": 3.2763579658667594e-05, + "loss": 0.9792, + "step": 199460 + }, + { + "epoch": 17.27, + "learning_rate": 3.276271333275578e-05, + "loss": 0.9291, + "step": 199470 + }, + { + "epoch": 17.27, + "learning_rate": 3.2761847006843976e-05, + "loss": 0.9711, + "step": 199480 + }, + { + "epoch": 17.27, + "learning_rate": 3.276098068093217e-05, + "loss": 0.9524, + "step": 199490 + }, + { + "epoch": 17.27, + "learning_rate": 3.276011435502036e-05, + "loss": 0.9798, + "step": 199500 + }, + { + "epoch": 17.27, + "learning_rate": 3.275924802910855e-05, + "loss": 0.9617, + "step": 199510 + }, + { + "epoch": 17.27, + "learning_rate": 3.275838170319675e-05, + "loss": 1.0109, + "step": 199520 + }, + { + "epoch": 17.27, + "learning_rate": 3.2757515377284935e-05, + "loss": 0.9839, + "step": 199530 + }, + { + "epoch": 17.27, + "learning_rate": 3.275664905137313e-05, + "loss": 0.9862, + "step": 199540 + }, + { + "epoch": 17.27, + "learning_rate": 3.2755782725461324e-05, + "loss": 1.013, + "step": 199550 + }, + { + "epoch": 17.27, + "learning_rate": 3.275491639954951e-05, + "loss": 0.9911, + "step": 199560 + }, + { + "epoch": 17.27, + "learning_rate": 3.2754050073637706e-05, + "loss": 1.0142, + "step": 199570 + }, + { + "epoch": 17.27, + "learning_rate": 3.275318374772589e-05, + "loss": 0.9397, + "step": 199580 + }, + { + "epoch": 17.28, + "learning_rate": 3.275231742181409e-05, + "loss": 1.0207, + "step": 199590 + }, + { + "epoch": 17.28, + "learning_rate": 3.275145109590228e-05, + "loss": 1.0731, + "step": 199600 + }, + { + "epoch": 17.28, + "learning_rate": 3.275058476999047e-05, + "loss": 1.0023, + "step": 199610 + }, + { + "epoch": 17.28, + "learning_rate": 3.2749718444078664e-05, + "loss": 1.0347, + "step": 199620 + }, + { + "epoch": 17.28, + "learning_rate": 3.274885211816686e-05, + "loss": 1.0138, + "step": 199630 + }, + { + "epoch": 17.28, + "learning_rate": 3.2747985792255047e-05, + "loss": 0.9585, + "step": 199640 + }, + { + "epoch": 17.28, + "learning_rate": 3.274711946634324e-05, + "loss": 1.0314, + "step": 199650 + }, + { + "epoch": 17.28, + "learning_rate": 3.2746253140431435e-05, + "loss": 1.0222, + "step": 199660 + }, + { + "epoch": 17.28, + "learning_rate": 3.274538681451962e-05, + "loss": 0.9724, + "step": 199670 + }, + { + "epoch": 17.28, + "learning_rate": 3.274452048860782e-05, + "loss": 0.9865, + "step": 199680 + }, + { + "epoch": 17.28, + "learning_rate": 3.2743654162696005e-05, + "loss": 1.0187, + "step": 199690 + }, + { + "epoch": 17.29, + "learning_rate": 3.27427878367842e-05, + "loss": 1.0079, + "step": 199700 + }, + { + "epoch": 17.29, + "learning_rate": 3.2741921510872394e-05, + "loss": 0.9592, + "step": 199710 + }, + { + "epoch": 17.29, + "learning_rate": 3.274105518496058e-05, + "loss": 1.0294, + "step": 199720 + }, + { + "epoch": 17.29, + "learning_rate": 3.2740188859048776e-05, + "loss": 0.9957, + "step": 199730 + }, + { + "epoch": 17.29, + "learning_rate": 3.273932253313697e-05, + "loss": 1.0443, + "step": 199740 + }, + { + "epoch": 17.29, + "learning_rate": 3.273845620722516e-05, + "loss": 0.9989, + "step": 199750 + }, + { + "epoch": 17.29, + "learning_rate": 3.273758988131335e-05, + "loss": 1.0574, + "step": 199760 + }, + { + "epoch": 17.29, + "learning_rate": 3.273672355540154e-05, + "loss": 0.9632, + "step": 199770 + }, + { + "epoch": 17.29, + "learning_rate": 3.2735857229489735e-05, + "loss": 0.9942, + "step": 199780 + }, + { + "epoch": 17.29, + "learning_rate": 3.273499090357793e-05, + "loss": 0.9683, + "step": 199790 + }, + { + "epoch": 17.29, + "learning_rate": 3.273412457766612e-05, + "loss": 0.9949, + "step": 199800 + }, + { + "epoch": 17.29, + "learning_rate": 3.273325825175431e-05, + "loss": 1.063, + "step": 199810 + }, + { + "epoch": 17.3, + "learning_rate": 3.2732391925842506e-05, + "loss": 0.9381, + "step": 199820 + }, + { + "epoch": 17.3, + "learning_rate": 3.2731525599930694e-05, + "loss": 0.9503, + "step": 199830 + }, + { + "epoch": 17.3, + "learning_rate": 3.273065927401889e-05, + "loss": 1.0123, + "step": 199840 + }, + { + "epoch": 17.3, + "learning_rate": 3.272979294810708e-05, + "loss": 1.0025, + "step": 199850 + }, + { + "epoch": 17.3, + "learning_rate": 3.272892662219527e-05, + "loss": 1.0636, + "step": 199860 + }, + { + "epoch": 17.3, + "learning_rate": 3.2728060296283465e-05, + "loss": 1.0688, + "step": 199870 + }, + { + "epoch": 17.3, + "learning_rate": 3.272719397037165e-05, + "loss": 1.0547, + "step": 199880 + }, + { + "epoch": 17.3, + "learning_rate": 3.272632764445985e-05, + "loss": 1.0086, + "step": 199890 + }, + { + "epoch": 17.3, + "learning_rate": 3.272546131854804e-05, + "loss": 1.0062, + "step": 199900 + }, + { + "epoch": 17.3, + "learning_rate": 3.272459499263623e-05, + "loss": 1.0938, + "step": 199910 + }, + { + "epoch": 17.3, + "learning_rate": 3.2723728666724423e-05, + "loss": 0.9785, + "step": 199920 + }, + { + "epoch": 17.3, + "learning_rate": 3.272286234081262e-05, + "loss": 1.0428, + "step": 199930 + }, + { + "epoch": 17.31, + "learning_rate": 3.2721996014900806e-05, + "loss": 1.0802, + "step": 199940 + }, + { + "epoch": 17.31, + "learning_rate": 3.2721129688989e-05, + "loss": 0.9422, + "step": 199950 + }, + { + "epoch": 17.31, + "learning_rate": 3.2720263363077194e-05, + "loss": 1.0066, + "step": 199960 + }, + { + "epoch": 17.31, + "learning_rate": 3.271939703716538e-05, + "loss": 1.0696, + "step": 199970 + }, + { + "epoch": 17.31, + "learning_rate": 3.2718530711253577e-05, + "loss": 0.9704, + "step": 199980 + }, + { + "epoch": 17.31, + "learning_rate": 3.2717664385341764e-05, + "loss": 1.0294, + "step": 199990 + }, + { + "epoch": 17.31, + "learning_rate": 3.271679805942996e-05, + "loss": 1.0071, + "step": 200000 + }, + { + "epoch": 17.31, + "learning_rate": 3.271593173351815e-05, + "loss": 0.9752, + "step": 200010 + }, + { + "epoch": 17.31, + "learning_rate": 3.271506540760634e-05, + "loss": 1.0409, + "step": 200020 + }, + { + "epoch": 17.31, + "learning_rate": 3.2714199081694535e-05, + "loss": 1.0123, + "step": 200030 + }, + { + "epoch": 17.31, + "learning_rate": 3.271333275578273e-05, + "loss": 0.9706, + "step": 200040 + }, + { + "epoch": 17.32, + "learning_rate": 3.271246642987092e-05, + "loss": 0.9338, + "step": 200050 + }, + { + "epoch": 17.32, + "learning_rate": 3.271160010395911e-05, + "loss": 1.0454, + "step": 200060 + }, + { + "epoch": 17.32, + "learning_rate": 3.2710733778047306e-05, + "loss": 0.9774, + "step": 200070 + }, + { + "epoch": 17.32, + "learning_rate": 3.2709867452135494e-05, + "loss": 0.957, + "step": 200080 + }, + { + "epoch": 17.32, + "learning_rate": 3.270900112622369e-05, + "loss": 0.9795, + "step": 200090 + }, + { + "epoch": 17.32, + "learning_rate": 3.2708134800311876e-05, + "loss": 1.0067, + "step": 200100 + }, + { + "epoch": 17.32, + "learning_rate": 3.270726847440007e-05, + "loss": 1.0516, + "step": 200110 + }, + { + "epoch": 17.32, + "learning_rate": 3.2706402148488265e-05, + "loss": 0.9392, + "step": 200120 + }, + { + "epoch": 17.32, + "learning_rate": 3.270553582257645e-05, + "loss": 1.0164, + "step": 200130 + }, + { + "epoch": 17.32, + "learning_rate": 3.270466949666465e-05, + "loss": 1.0123, + "step": 200140 + }, + { + "epoch": 17.32, + "learning_rate": 3.270380317075284e-05, + "loss": 1.0886, + "step": 200150 + }, + { + "epoch": 17.32, + "learning_rate": 3.270293684484103e-05, + "loss": 0.9742, + "step": 200160 + }, + { + "epoch": 17.33, + "learning_rate": 3.2702070518929224e-05, + "loss": 0.9669, + "step": 200170 + }, + { + "epoch": 17.33, + "learning_rate": 3.270120419301742e-05, + "loss": 1.049, + "step": 200180 + }, + { + "epoch": 17.33, + "learning_rate": 3.2700337867105606e-05, + "loss": 1.0066, + "step": 200190 + }, + { + "epoch": 17.33, + "learning_rate": 3.26994715411938e-05, + "loss": 0.9924, + "step": 200200 + }, + { + "epoch": 17.33, + "learning_rate": 3.269860521528199e-05, + "loss": 0.9856, + "step": 200210 + }, + { + "epoch": 17.33, + "learning_rate": 3.269773888937018e-05, + "loss": 0.9677, + "step": 200220 + }, + { + "epoch": 17.33, + "learning_rate": 3.269687256345838e-05, + "loss": 1.0502, + "step": 200230 + }, + { + "epoch": 17.33, + "learning_rate": 3.2696006237546565e-05, + "loss": 1.0207, + "step": 200240 + }, + { + "epoch": 17.33, + "learning_rate": 3.269513991163476e-05, + "loss": 1.0286, + "step": 200250 + }, + { + "epoch": 17.33, + "learning_rate": 3.2694273585722954e-05, + "loss": 0.9419, + "step": 200260 + }, + { + "epoch": 17.33, + "learning_rate": 3.269340725981114e-05, + "loss": 1.0292, + "step": 200270 + }, + { + "epoch": 17.34, + "learning_rate": 3.2692540933899336e-05, + "loss": 1.0036, + "step": 200280 + }, + { + "epoch": 17.34, + "learning_rate": 3.269167460798753e-05, + "loss": 0.972, + "step": 200290 + }, + { + "epoch": 17.34, + "learning_rate": 3.269080828207572e-05, + "loss": 1.0173, + "step": 200300 + }, + { + "epoch": 17.34, + "learning_rate": 3.268994195616391e-05, + "loss": 1.0457, + "step": 200310 + }, + { + "epoch": 17.34, + "learning_rate": 3.26890756302521e-05, + "loss": 0.9317, + "step": 200320 + }, + { + "epoch": 17.34, + "learning_rate": 3.2688209304340294e-05, + "loss": 0.994, + "step": 200330 + }, + { + "epoch": 17.34, + "learning_rate": 3.268734297842849e-05, + "loss": 1.0159, + "step": 200340 + }, + { + "epoch": 17.34, + "learning_rate": 3.2686476652516676e-05, + "loss": 0.9973, + "step": 200350 + }, + { + "epoch": 17.34, + "learning_rate": 3.268561032660487e-05, + "loss": 0.9493, + "step": 200360 + }, + { + "epoch": 17.34, + "learning_rate": 3.2684744000693065e-05, + "loss": 1.0267, + "step": 200370 + }, + { + "epoch": 17.34, + "learning_rate": 3.268387767478125e-05, + "loss": 0.9488, + "step": 200380 + }, + { + "epoch": 17.34, + "learning_rate": 3.268301134886945e-05, + "loss": 1.0588, + "step": 200390 + }, + { + "epoch": 17.35, + "learning_rate": 3.268214502295764e-05, + "loss": 1.0473, + "step": 200400 + }, + { + "epoch": 17.35, + "learning_rate": 3.268127869704583e-05, + "loss": 1.0051, + "step": 200410 + }, + { + "epoch": 17.35, + "learning_rate": 3.2680412371134024e-05, + "loss": 1.0051, + "step": 200420 + }, + { + "epoch": 17.35, + "learning_rate": 3.267954604522221e-05, + "loss": 0.9791, + "step": 200430 + }, + { + "epoch": 17.35, + "learning_rate": 3.2678679719310406e-05, + "loss": 0.9915, + "step": 200440 + }, + { + "epoch": 17.35, + "learning_rate": 3.26778133933986e-05, + "loss": 1.0241, + "step": 200450 + }, + { + "epoch": 17.35, + "learning_rate": 3.267694706748679e-05, + "loss": 1.0414, + "step": 200460 + }, + { + "epoch": 17.35, + "learning_rate": 3.267608074157498e-05, + "loss": 0.9685, + "step": 200470 + }, + { + "epoch": 17.35, + "learning_rate": 3.267521441566318e-05, + "loss": 1.0059, + "step": 200480 + }, + { + "epoch": 17.35, + "learning_rate": 3.2674348089751365e-05, + "loss": 0.9547, + "step": 200490 + }, + { + "epoch": 17.35, + "learning_rate": 3.267348176383956e-05, + "loss": 0.9736, + "step": 200500 + }, + { + "epoch": 17.36, + "learning_rate": 3.2672615437927754e-05, + "loss": 1.0477, + "step": 200510 + }, + { + "epoch": 17.36, + "learning_rate": 3.267174911201594e-05, + "loss": 0.9535, + "step": 200520 + }, + { + "epoch": 17.36, + "learning_rate": 3.2670882786104136e-05, + "loss": 0.9819, + "step": 200530 + }, + { + "epoch": 17.36, + "learning_rate": 3.2670016460192324e-05, + "loss": 1.0312, + "step": 200540 + }, + { + "epoch": 17.36, + "learning_rate": 3.266915013428052e-05, + "loss": 1.0045, + "step": 200550 + }, + { + "epoch": 17.36, + "learning_rate": 3.266828380836871e-05, + "loss": 1.055, + "step": 200560 + }, + { + "epoch": 17.36, + "learning_rate": 3.26674174824569e-05, + "loss": 1.0204, + "step": 200570 + }, + { + "epoch": 17.36, + "learning_rate": 3.2666551156545095e-05, + "loss": 1.0063, + "step": 200580 + }, + { + "epoch": 17.36, + "learning_rate": 3.266568483063329e-05, + "loss": 0.96, + "step": 200590 + }, + { + "epoch": 17.36, + "learning_rate": 3.266481850472148e-05, + "loss": 1.0141, + "step": 200600 + }, + { + "epoch": 17.36, + "learning_rate": 3.266395217880967e-05, + "loss": 1.017, + "step": 200610 + }, + { + "epoch": 17.36, + "learning_rate": 3.266308585289786e-05, + "loss": 1.0085, + "step": 200620 + }, + { + "epoch": 17.37, + "learning_rate": 3.2662219526986053e-05, + "loss": 0.9765, + "step": 200630 + }, + { + "epoch": 17.37, + "learning_rate": 3.266135320107425e-05, + "loss": 1.0177, + "step": 200640 + }, + { + "epoch": 17.37, + "learning_rate": 3.2660486875162436e-05, + "loss": 1.0348, + "step": 200650 + }, + { + "epoch": 17.37, + "learning_rate": 3.265962054925063e-05, + "loss": 1.0029, + "step": 200660 + }, + { + "epoch": 17.37, + "learning_rate": 3.2658754223338824e-05, + "loss": 1.0095, + "step": 200670 + }, + { + "epoch": 17.37, + "learning_rate": 3.265788789742701e-05, + "loss": 0.9955, + "step": 200680 + }, + { + "epoch": 17.37, + "learning_rate": 3.2657021571515207e-05, + "loss": 0.9787, + "step": 200690 + }, + { + "epoch": 17.37, + "learning_rate": 3.26561552456034e-05, + "loss": 0.9724, + "step": 200700 + }, + { + "epoch": 17.37, + "learning_rate": 3.265528891969159e-05, + "loss": 1.0364, + "step": 200710 + }, + { + "epoch": 17.37, + "learning_rate": 3.265442259377978e-05, + "loss": 0.9955, + "step": 200720 + }, + { + "epoch": 17.37, + "learning_rate": 3.265355626786797e-05, + "loss": 0.9846, + "step": 200730 + }, + { + "epoch": 17.38, + "learning_rate": 3.2652689941956165e-05, + "loss": 1.0249, + "step": 200740 + }, + { + "epoch": 17.38, + "learning_rate": 3.265182361604436e-05, + "loss": 0.9859, + "step": 200750 + }, + { + "epoch": 17.38, + "learning_rate": 3.265095729013255e-05, + "loss": 1.0016, + "step": 200760 + }, + { + "epoch": 17.38, + "learning_rate": 3.265009096422074e-05, + "loss": 0.9758, + "step": 200770 + }, + { + "epoch": 17.38, + "learning_rate": 3.2649224638308936e-05, + "loss": 0.9757, + "step": 200780 + }, + { + "epoch": 17.38, + "learning_rate": 3.2648358312397124e-05, + "loss": 1.0058, + "step": 200790 + }, + { + "epoch": 17.38, + "learning_rate": 3.264749198648532e-05, + "loss": 1.0135, + "step": 200800 + }, + { + "epoch": 17.38, + "learning_rate": 3.264662566057351e-05, + "loss": 0.9726, + "step": 200810 + }, + { + "epoch": 17.38, + "learning_rate": 3.26457593346617e-05, + "loss": 0.9962, + "step": 200820 + }, + { + "epoch": 17.38, + "learning_rate": 3.2644893008749895e-05, + "loss": 1.0325, + "step": 200830 + }, + { + "epoch": 17.38, + "learning_rate": 3.264402668283808e-05, + "loss": 1.0285, + "step": 200840 + }, + { + "epoch": 17.38, + "learning_rate": 3.264316035692628e-05, + "loss": 0.9865, + "step": 200850 + }, + { + "epoch": 17.39, + "learning_rate": 3.264229403101447e-05, + "loss": 1.0336, + "step": 200860 + }, + { + "epoch": 17.39, + "learning_rate": 3.264142770510266e-05, + "loss": 1.0395, + "step": 200870 + }, + { + "epoch": 17.39, + "learning_rate": 3.2640561379190854e-05, + "loss": 0.9867, + "step": 200880 + }, + { + "epoch": 17.39, + "learning_rate": 3.263969505327905e-05, + "loss": 0.9762, + "step": 200890 + }, + { + "epoch": 17.39, + "learning_rate": 3.2638828727367236e-05, + "loss": 1.0134, + "step": 200900 + }, + { + "epoch": 17.39, + "learning_rate": 3.263796240145543e-05, + "loss": 0.9845, + "step": 200910 + }, + { + "epoch": 17.39, + "learning_rate": 3.2637096075543625e-05, + "loss": 1.0542, + "step": 200920 + }, + { + "epoch": 17.39, + "learning_rate": 3.263622974963181e-05, + "loss": 1.0114, + "step": 200930 + }, + { + "epoch": 17.39, + "learning_rate": 3.263536342372001e-05, + "loss": 0.9237, + "step": 200940 + }, + { + "epoch": 17.39, + "learning_rate": 3.2634497097808195e-05, + "loss": 1.0007, + "step": 200950 + }, + { + "epoch": 17.39, + "learning_rate": 3.263363077189639e-05, + "loss": 0.9164, + "step": 200960 + }, + { + "epoch": 17.39, + "learning_rate": 3.2632764445984583e-05, + "loss": 0.9924, + "step": 200970 + }, + { + "epoch": 17.4, + "learning_rate": 3.263189812007277e-05, + "loss": 0.9859, + "step": 200980 + }, + { + "epoch": 17.4, + "learning_rate": 3.2631031794160966e-05, + "loss": 1.0003, + "step": 200990 + }, + { + "epoch": 17.4, + "learning_rate": 3.263016546824916e-05, + "loss": 0.9782, + "step": 201000 + }, + { + "epoch": 17.4, + "learning_rate": 3.262929914233735e-05, + "loss": 1.0156, + "step": 201010 + }, + { + "epoch": 17.4, + "learning_rate": 3.262843281642554e-05, + "loss": 1.008, + "step": 201020 + }, + { + "epoch": 17.4, + "learning_rate": 3.262756649051374e-05, + "loss": 1.0151, + "step": 201030 + }, + { + "epoch": 17.4, + "learning_rate": 3.2626700164601924e-05, + "loss": 0.9644, + "step": 201040 + }, + { + "epoch": 17.4, + "learning_rate": 3.262583383869012e-05, + "loss": 1.0156, + "step": 201050 + }, + { + "epoch": 17.4, + "learning_rate": 3.2624967512778306e-05, + "loss": 0.998, + "step": 201060 + }, + { + "epoch": 17.4, + "learning_rate": 3.26241011868665e-05, + "loss": 0.9983, + "step": 201070 + }, + { + "epoch": 17.4, + "learning_rate": 3.2623234860954695e-05, + "loss": 0.9778, + "step": 201080 + }, + { + "epoch": 17.41, + "learning_rate": 3.262236853504288e-05, + "loss": 0.9941, + "step": 201090 + }, + { + "epoch": 17.41, + "learning_rate": 3.262150220913108e-05, + "loss": 1.0463, + "step": 201100 + }, + { + "epoch": 17.41, + "learning_rate": 3.262063588321927e-05, + "loss": 0.9838, + "step": 201110 + }, + { + "epoch": 17.41, + "learning_rate": 3.261976955730746e-05, + "loss": 1.0015, + "step": 201120 + }, + { + "epoch": 17.41, + "learning_rate": 3.2618903231395654e-05, + "loss": 0.9568, + "step": 201130 + }, + { + "epoch": 17.41, + "learning_rate": 3.261803690548385e-05, + "loss": 1.0028, + "step": 201140 + }, + { + "epoch": 17.41, + "learning_rate": 3.2617170579572036e-05, + "loss": 0.9795, + "step": 201150 + }, + { + "epoch": 17.41, + "learning_rate": 3.261630425366023e-05, + "loss": 1.0276, + "step": 201160 + }, + { + "epoch": 17.41, + "learning_rate": 3.261543792774842e-05, + "loss": 0.9863, + "step": 201170 + }, + { + "epoch": 17.41, + "learning_rate": 3.261457160183661e-05, + "loss": 1.0324, + "step": 201180 + }, + { + "epoch": 17.41, + "learning_rate": 3.261370527592481e-05, + "loss": 0.9829, + "step": 201190 + }, + { + "epoch": 17.41, + "learning_rate": 3.2612838950012995e-05, + "loss": 1.0261, + "step": 201200 + }, + { + "epoch": 17.42, + "learning_rate": 3.261197262410119e-05, + "loss": 0.9474, + "step": 201210 + }, + { + "epoch": 17.42, + "learning_rate": 3.2611106298189384e-05, + "loss": 0.9892, + "step": 201220 + }, + { + "epoch": 17.42, + "learning_rate": 3.261023997227757e-05, + "loss": 0.9568, + "step": 201230 + }, + { + "epoch": 17.42, + "learning_rate": 3.2609373646365766e-05, + "loss": 0.9524, + "step": 201240 + }, + { + "epoch": 17.42, + "learning_rate": 3.260850732045396e-05, + "loss": 0.9924, + "step": 201250 + }, + { + "epoch": 17.42, + "learning_rate": 3.260764099454215e-05, + "loss": 1.0131, + "step": 201260 + }, + { + "epoch": 17.42, + "learning_rate": 3.260677466863034e-05, + "loss": 0.9493, + "step": 201270 + }, + { + "epoch": 17.42, + "learning_rate": 3.260590834271853e-05, + "loss": 1.037, + "step": 201280 + }, + { + "epoch": 17.42, + "learning_rate": 3.2605042016806725e-05, + "loss": 0.94, + "step": 201290 + }, + { + "epoch": 17.42, + "learning_rate": 3.260417569089492e-05, + "loss": 0.9892, + "step": 201300 + }, + { + "epoch": 17.42, + "learning_rate": 3.260330936498311e-05, + "loss": 1.0326, + "step": 201310 + }, + { + "epoch": 17.43, + "learning_rate": 3.26024430390713e-05, + "loss": 1.0088, + "step": 201320 + }, + { + "epoch": 17.43, + "learning_rate": 3.2601576713159496e-05, + "loss": 1.0412, + "step": 201330 + }, + { + "epoch": 17.43, + "learning_rate": 3.260071038724768e-05, + "loss": 1.0279, + "step": 201340 + }, + { + "epoch": 17.43, + "learning_rate": 3.259984406133588e-05, + "loss": 0.9693, + "step": 201350 + }, + { + "epoch": 17.43, + "learning_rate": 3.2598977735424066e-05, + "loss": 0.9473, + "step": 201360 + }, + { + "epoch": 17.43, + "learning_rate": 3.259811140951226e-05, + "loss": 1.0078, + "step": 201370 + }, + { + "epoch": 17.43, + "learning_rate": 3.2597245083600454e-05, + "loss": 0.9919, + "step": 201380 + }, + { + "epoch": 17.43, + "learning_rate": 3.259637875768864e-05, + "loss": 0.9256, + "step": 201390 + }, + { + "epoch": 17.43, + "learning_rate": 3.2595512431776837e-05, + "loss": 0.9908, + "step": 201400 + }, + { + "epoch": 17.43, + "learning_rate": 3.259464610586503e-05, + "loss": 1.0446, + "step": 201410 + }, + { + "epoch": 17.43, + "learning_rate": 3.259377977995322e-05, + "loss": 0.9575, + "step": 201420 + }, + { + "epoch": 17.43, + "learning_rate": 3.259291345404141e-05, + "loss": 1.0096, + "step": 201430 + }, + { + "epoch": 17.44, + "learning_rate": 3.259204712812961e-05, + "loss": 0.986, + "step": 201440 + }, + { + "epoch": 17.44, + "learning_rate": 3.2591180802217795e-05, + "loss": 1.037, + "step": 201450 + }, + { + "epoch": 17.44, + "learning_rate": 3.259031447630599e-05, + "loss": 0.989, + "step": 201460 + }, + { + "epoch": 17.44, + "learning_rate": 3.258944815039418e-05, + "loss": 0.9832, + "step": 201470 + }, + { + "epoch": 17.44, + "learning_rate": 3.258858182448237e-05, + "loss": 0.9939, + "step": 201480 + }, + { + "epoch": 17.44, + "learning_rate": 3.2587715498570566e-05, + "loss": 0.9394, + "step": 201490 + }, + { + "epoch": 17.44, + "learning_rate": 3.2586849172658754e-05, + "loss": 0.9583, + "step": 201500 + }, + { + "epoch": 17.44, + "learning_rate": 3.258598284674695e-05, + "loss": 1.0716, + "step": 201510 + }, + { + "epoch": 17.44, + "learning_rate": 3.258511652083514e-05, + "loss": 0.962, + "step": 201520 + }, + { + "epoch": 17.44, + "learning_rate": 3.258425019492333e-05, + "loss": 1.026, + "step": 201530 + }, + { + "epoch": 17.44, + "learning_rate": 3.2583383869011525e-05, + "loss": 1.0131, + "step": 201540 + }, + { + "epoch": 17.45, + "learning_rate": 3.258251754309972e-05, + "loss": 0.9827, + "step": 201550 + }, + { + "epoch": 17.45, + "learning_rate": 3.258165121718791e-05, + "loss": 1.0132, + "step": 201560 + }, + { + "epoch": 17.45, + "learning_rate": 3.25807848912761e-05, + "loss": 0.9769, + "step": 201570 + }, + { + "epoch": 17.45, + "learning_rate": 3.257991856536429e-05, + "loss": 1.0103, + "step": 201580 + }, + { + "epoch": 17.45, + "learning_rate": 3.2579052239452484e-05, + "loss": 1.0715, + "step": 201590 + }, + { + "epoch": 17.45, + "learning_rate": 3.257818591354068e-05, + "loss": 1.0082, + "step": 201600 + }, + { + "epoch": 17.45, + "learning_rate": 3.2577319587628866e-05, + "loss": 1.045, + "step": 201610 + }, + { + "epoch": 17.45, + "learning_rate": 3.257645326171706e-05, + "loss": 0.9732, + "step": 201620 + }, + { + "epoch": 17.45, + "learning_rate": 3.2575586935805255e-05, + "loss": 1.0285, + "step": 201630 + }, + { + "epoch": 17.45, + "learning_rate": 3.257472060989344e-05, + "loss": 1.0196, + "step": 201640 + }, + { + "epoch": 17.45, + "learning_rate": 3.257385428398164e-05, + "loss": 1.0559, + "step": 201650 + }, + { + "epoch": 17.45, + "learning_rate": 3.257298795806983e-05, + "loss": 0.9969, + "step": 201660 + }, + { + "epoch": 17.46, + "learning_rate": 3.257212163215802e-05, + "loss": 1.012, + "step": 201670 + }, + { + "epoch": 17.46, + "learning_rate": 3.2571255306246213e-05, + "loss": 0.9177, + "step": 201680 + }, + { + "epoch": 17.46, + "learning_rate": 3.25703889803344e-05, + "loss": 1.0041, + "step": 201690 + }, + { + "epoch": 17.46, + "learning_rate": 3.2569522654422596e-05, + "loss": 0.9725, + "step": 201700 + }, + { + "epoch": 17.46, + "learning_rate": 3.256865632851079e-05, + "loss": 0.926, + "step": 201710 + }, + { + "epoch": 17.46, + "learning_rate": 3.256779000259898e-05, + "loss": 1.0494, + "step": 201720 + }, + { + "epoch": 17.46, + "learning_rate": 3.256692367668717e-05, + "loss": 1.0161, + "step": 201730 + }, + { + "epoch": 17.46, + "learning_rate": 3.256605735077537e-05, + "loss": 0.982, + "step": 201740 + }, + { + "epoch": 17.46, + "learning_rate": 3.2565191024863554e-05, + "loss": 0.9951, + "step": 201750 + }, + { + "epoch": 17.46, + "learning_rate": 3.256432469895175e-05, + "loss": 0.9946, + "step": 201760 + }, + { + "epoch": 17.46, + "learning_rate": 3.256345837303994e-05, + "loss": 1.0112, + "step": 201770 + }, + { + "epoch": 17.47, + "learning_rate": 3.256259204712813e-05, + "loss": 0.9952, + "step": 201780 + }, + { + "epoch": 17.47, + "learning_rate": 3.2561725721216325e-05, + "loss": 1.0057, + "step": 201790 + }, + { + "epoch": 17.47, + "learning_rate": 3.256085939530451e-05, + "loss": 0.9822, + "step": 201800 + }, + { + "epoch": 17.47, + "learning_rate": 3.255999306939271e-05, + "loss": 1.0394, + "step": 201810 + }, + { + "epoch": 17.47, + "learning_rate": 3.25591267434809e-05, + "loss": 1.0737, + "step": 201820 + }, + { + "epoch": 17.47, + "learning_rate": 3.255826041756909e-05, + "loss": 0.9972, + "step": 201830 + }, + { + "epoch": 17.47, + "learning_rate": 3.2557394091657284e-05, + "loss": 0.9903, + "step": 201840 + }, + { + "epoch": 17.47, + "learning_rate": 3.255652776574548e-05, + "loss": 0.9661, + "step": 201850 + }, + { + "epoch": 17.47, + "learning_rate": 3.2555661439833666e-05, + "loss": 1.008, + "step": 201860 + }, + { + "epoch": 17.47, + "learning_rate": 3.255479511392186e-05, + "loss": 0.9588, + "step": 201870 + }, + { + "epoch": 17.47, + "learning_rate": 3.2553928788010055e-05, + "loss": 1.0303, + "step": 201880 + }, + { + "epoch": 17.47, + "learning_rate": 3.255306246209824e-05, + "loss": 1.0445, + "step": 201890 + }, + { + "epoch": 17.48, + "learning_rate": 3.255219613618644e-05, + "loss": 0.9853, + "step": 201900 + }, + { + "epoch": 17.48, + "learning_rate": 3.2551329810274625e-05, + "loss": 1.0326, + "step": 201910 + }, + { + "epoch": 17.48, + "learning_rate": 3.255046348436282e-05, + "loss": 0.9997, + "step": 201920 + }, + { + "epoch": 17.48, + "learning_rate": 3.2549597158451014e-05, + "loss": 0.9483, + "step": 201930 + }, + { + "epoch": 17.48, + "learning_rate": 3.25487308325392e-05, + "loss": 0.981, + "step": 201940 + }, + { + "epoch": 17.48, + "learning_rate": 3.2547864506627396e-05, + "loss": 1.0166, + "step": 201950 + }, + { + "epoch": 17.48, + "learning_rate": 3.254699818071559e-05, + "loss": 1.0938, + "step": 201960 + }, + { + "epoch": 17.48, + "learning_rate": 3.254613185480378e-05, + "loss": 0.9406, + "step": 201970 + }, + { + "epoch": 17.48, + "learning_rate": 3.254526552889197e-05, + "loss": 0.9492, + "step": 201980 + }, + { + "epoch": 17.48, + "learning_rate": 3.254439920298017e-05, + "loss": 1.0143, + "step": 201990 + }, + { + "epoch": 17.48, + "learning_rate": 3.2543532877068355e-05, + "loss": 1.018, + "step": 202000 + }, + { + "epoch": 17.49, + "learning_rate": 3.254266655115655e-05, + "loss": 1.0078, + "step": 202010 + }, + { + "epoch": 17.49, + "learning_rate": 3.254180022524474e-05, + "loss": 0.951, + "step": 202020 + }, + { + "epoch": 17.49, + "learning_rate": 3.254093389933293e-05, + "loss": 1.0115, + "step": 202030 + }, + { + "epoch": 17.49, + "learning_rate": 3.2540067573421126e-05, + "loss": 0.9103, + "step": 202040 + }, + { + "epoch": 17.49, + "learning_rate": 3.253920124750931e-05, + "loss": 1.0036, + "step": 202050 + }, + { + "epoch": 17.49, + "learning_rate": 3.253833492159751e-05, + "loss": 1.0191, + "step": 202060 + }, + { + "epoch": 17.49, + "learning_rate": 3.25374685956857e-05, + "loss": 0.9933, + "step": 202070 + }, + { + "epoch": 17.49, + "learning_rate": 3.253660226977389e-05, + "loss": 1.0285, + "step": 202080 + }, + { + "epoch": 17.49, + "learning_rate": 3.2535735943862084e-05, + "loss": 0.9917, + "step": 202090 + }, + { + "epoch": 17.49, + "learning_rate": 3.253486961795027e-05, + "loss": 0.9853, + "step": 202100 + }, + { + "epoch": 17.49, + "learning_rate": 3.2534003292038467e-05, + "loss": 0.9758, + "step": 202110 + }, + { + "epoch": 17.49, + "learning_rate": 3.253313696612666e-05, + "loss": 0.9822, + "step": 202120 + }, + { + "epoch": 17.5, + "learning_rate": 3.253227064021485e-05, + "loss": 1.0231, + "step": 202130 + }, + { + "epoch": 17.5, + "learning_rate": 3.253140431430304e-05, + "loss": 0.9656, + "step": 202140 + }, + { + "epoch": 17.5, + "learning_rate": 3.253053798839124e-05, + "loss": 1.0223, + "step": 202150 + }, + { + "epoch": 17.5, + "learning_rate": 3.2529671662479425e-05, + "loss": 0.9898, + "step": 202160 + }, + { + "epoch": 17.5, + "learning_rate": 3.252880533656762e-05, + "loss": 0.9699, + "step": 202170 + }, + { + "epoch": 17.5, + "learning_rate": 3.2527939010655814e-05, + "loss": 1.0012, + "step": 202180 + }, + { + "epoch": 17.5, + "learning_rate": 3.2527072684744e-05, + "loss": 1.0584, + "step": 202190 + }, + { + "epoch": 17.5, + "learning_rate": 3.2526206358832196e-05, + "loss": 0.954, + "step": 202200 + }, + { + "epoch": 17.5, + "learning_rate": 3.2525340032920384e-05, + "loss": 0.9906, + "step": 202210 + }, + { + "epoch": 17.5, + "learning_rate": 3.252447370700858e-05, + "loss": 0.9904, + "step": 202220 + }, + { + "epoch": 17.5, + "learning_rate": 3.252360738109677e-05, + "loss": 1.0035, + "step": 202230 + }, + { + "epoch": 17.5, + "learning_rate": 3.252274105518496e-05, + "loss": 0.9535, + "step": 202240 + }, + { + "epoch": 17.51, + "learning_rate": 3.2521874729273155e-05, + "loss": 1.0229, + "step": 202250 + }, + { + "epoch": 17.51, + "learning_rate": 3.252100840336135e-05, + "loss": 1.0049, + "step": 202260 + }, + { + "epoch": 17.51, + "learning_rate": 3.252014207744954e-05, + "loss": 0.951, + "step": 202270 + }, + { + "epoch": 17.51, + "learning_rate": 3.251927575153773e-05, + "loss": 1.0386, + "step": 202280 + }, + { + "epoch": 17.51, + "learning_rate": 3.2518409425625926e-05, + "loss": 1.0006, + "step": 202290 + }, + { + "epoch": 17.51, + "learning_rate": 3.2517543099714114e-05, + "loss": 0.9713, + "step": 202300 + }, + { + "epoch": 17.51, + "learning_rate": 3.251667677380231e-05, + "loss": 1.0087, + "step": 202310 + }, + { + "epoch": 17.51, + "learning_rate": 3.2515810447890496e-05, + "loss": 0.9789, + "step": 202320 + }, + { + "epoch": 17.51, + "learning_rate": 3.251494412197869e-05, + "loss": 0.9579, + "step": 202330 + }, + { + "epoch": 17.51, + "learning_rate": 3.2514077796066885e-05, + "loss": 1.0033, + "step": 202340 + }, + { + "epoch": 17.51, + "learning_rate": 3.251321147015507e-05, + "loss": 0.9534, + "step": 202350 + }, + { + "epoch": 17.52, + "learning_rate": 3.251234514424327e-05, + "loss": 1.0198, + "step": 202360 + }, + { + "epoch": 17.52, + "learning_rate": 3.251147881833146e-05, + "loss": 0.9737, + "step": 202370 + }, + { + "epoch": 17.52, + "learning_rate": 3.251061249241965e-05, + "loss": 0.9299, + "step": 202380 + }, + { + "epoch": 17.52, + "learning_rate": 3.2509746166507843e-05, + "loss": 0.9564, + "step": 202390 + }, + { + "epoch": 17.52, + "learning_rate": 3.250887984059604e-05, + "loss": 0.96, + "step": 202400 + }, + { + "epoch": 17.52, + "learning_rate": 3.2508013514684226e-05, + "loss": 1.0001, + "step": 202410 + }, + { + "epoch": 17.52, + "learning_rate": 3.250714718877242e-05, + "loss": 1.0159, + "step": 202420 + }, + { + "epoch": 17.52, + "learning_rate": 3.250628086286061e-05, + "loss": 0.9537, + "step": 202430 + }, + { + "epoch": 17.52, + "learning_rate": 3.25054145369488e-05, + "loss": 1.0406, + "step": 202440 + }, + { + "epoch": 17.52, + "learning_rate": 3.2504548211036997e-05, + "loss": 0.9807, + "step": 202450 + }, + { + "epoch": 17.52, + "learning_rate": 3.2503681885125184e-05, + "loss": 1.0133, + "step": 202460 + }, + { + "epoch": 17.52, + "learning_rate": 3.250281555921338e-05, + "loss": 1.0291, + "step": 202470 + }, + { + "epoch": 17.53, + "learning_rate": 3.250194923330157e-05, + "loss": 1.0545, + "step": 202480 + }, + { + "epoch": 17.53, + "learning_rate": 3.250108290738976e-05, + "loss": 1.0758, + "step": 202490 + }, + { + "epoch": 17.53, + "learning_rate": 3.2500216581477955e-05, + "loss": 0.9492, + "step": 202500 + }, + { + "epoch": 17.53, + "learning_rate": 3.249935025556615e-05, + "loss": 1.0036, + "step": 202510 + }, + { + "epoch": 17.53, + "learning_rate": 3.249848392965434e-05, + "loss": 1.0057, + "step": 202520 + }, + { + "epoch": 17.53, + "learning_rate": 3.249761760374253e-05, + "loss": 0.9791, + "step": 202530 + }, + { + "epoch": 17.53, + "learning_rate": 3.249675127783072e-05, + "loss": 0.9579, + "step": 202540 + }, + { + "epoch": 17.53, + "learning_rate": 3.2495884951918914e-05, + "loss": 1.0161, + "step": 202550 + }, + { + "epoch": 17.53, + "learning_rate": 3.249501862600711e-05, + "loss": 1.0932, + "step": 202560 + }, + { + "epoch": 17.53, + "learning_rate": 3.2494152300095296e-05, + "loss": 0.997, + "step": 202570 + }, + { + "epoch": 17.53, + "learning_rate": 3.249328597418349e-05, + "loss": 1.0103, + "step": 202580 + }, + { + "epoch": 17.54, + "learning_rate": 3.2492419648271685e-05, + "loss": 0.9687, + "step": 202590 + }, + { + "epoch": 17.54, + "learning_rate": 3.249155332235987e-05, + "loss": 0.9965, + "step": 202600 + }, + { + "epoch": 17.54, + "learning_rate": 3.249068699644807e-05, + "loss": 0.9908, + "step": 202610 + }, + { + "epoch": 17.54, + "learning_rate": 3.248982067053626e-05, + "loss": 1.0552, + "step": 202620 + }, + { + "epoch": 17.54, + "learning_rate": 3.248895434462445e-05, + "loss": 1.0043, + "step": 202630 + }, + { + "epoch": 17.54, + "learning_rate": 3.2488088018712644e-05, + "loss": 0.96, + "step": 202640 + }, + { + "epoch": 17.54, + "learning_rate": 3.248722169280083e-05, + "loss": 0.9752, + "step": 202650 + }, + { + "epoch": 17.54, + "learning_rate": 3.2486355366889026e-05, + "loss": 1.0151, + "step": 202660 + }, + { + "epoch": 17.54, + "learning_rate": 3.248548904097722e-05, + "loss": 1.0136, + "step": 202670 + }, + { + "epoch": 17.54, + "learning_rate": 3.248462271506541e-05, + "loss": 0.9543, + "step": 202680 + }, + { + "epoch": 17.54, + "learning_rate": 3.24837563891536e-05, + "loss": 0.9787, + "step": 202690 + }, + { + "epoch": 17.54, + "learning_rate": 3.24828900632418e-05, + "loss": 1.0073, + "step": 202700 + }, + { + "epoch": 17.55, + "learning_rate": 3.2482023737329985e-05, + "loss": 1.0009, + "step": 202710 + }, + { + "epoch": 17.55, + "learning_rate": 3.248115741141818e-05, + "loss": 0.9744, + "step": 202720 + }, + { + "epoch": 17.55, + "learning_rate": 3.2480291085506374e-05, + "loss": 0.976, + "step": 202730 + }, + { + "epoch": 17.55, + "learning_rate": 3.247942475959456e-05, + "loss": 0.9573, + "step": 202740 + }, + { + "epoch": 17.55, + "learning_rate": 3.2478558433682756e-05, + "loss": 0.9668, + "step": 202750 + }, + { + "epoch": 17.55, + "learning_rate": 3.247769210777094e-05, + "loss": 0.9975, + "step": 202760 + }, + { + "epoch": 17.55, + "learning_rate": 3.247682578185914e-05, + "loss": 0.929, + "step": 202770 + }, + { + "epoch": 17.55, + "learning_rate": 3.247595945594733e-05, + "loss": 0.9957, + "step": 202780 + }, + { + "epoch": 17.55, + "learning_rate": 3.247509313003552e-05, + "loss": 0.9507, + "step": 202790 + }, + { + "epoch": 17.55, + "learning_rate": 3.2474226804123714e-05, + "loss": 0.9882, + "step": 202800 + }, + { + "epoch": 17.55, + "learning_rate": 3.247336047821191e-05, + "loss": 1.0048, + "step": 202810 + }, + { + "epoch": 17.56, + "learning_rate": 3.2472494152300096e-05, + "loss": 0.9743, + "step": 202820 + }, + { + "epoch": 17.56, + "learning_rate": 3.247162782638829e-05, + "loss": 1.0208, + "step": 202830 + }, + { + "epoch": 17.56, + "learning_rate": 3.2470761500476485e-05, + "loss": 1.0318, + "step": 202840 + }, + { + "epoch": 17.56, + "learning_rate": 3.246989517456467e-05, + "loss": 1.0222, + "step": 202850 + }, + { + "epoch": 17.56, + "learning_rate": 3.246902884865287e-05, + "loss": 1.0184, + "step": 202860 + }, + { + "epoch": 17.56, + "learning_rate": 3.2468162522741055e-05, + "loss": 0.979, + "step": 202870 + }, + { + "epoch": 17.56, + "learning_rate": 3.246729619682925e-05, + "loss": 0.9626, + "step": 202880 + }, + { + "epoch": 17.56, + "learning_rate": 3.2466429870917444e-05, + "loss": 1.0099, + "step": 202890 + }, + { + "epoch": 17.56, + "learning_rate": 3.246556354500563e-05, + "loss": 0.9708, + "step": 202900 + }, + { + "epoch": 17.56, + "learning_rate": 3.2464697219093826e-05, + "loss": 1.07, + "step": 202910 + }, + { + "epoch": 17.56, + "learning_rate": 3.246383089318202e-05, + "loss": 0.9706, + "step": 202920 + }, + { + "epoch": 17.56, + "learning_rate": 3.246296456727021e-05, + "loss": 0.9456, + "step": 202930 + }, + { + "epoch": 17.57, + "learning_rate": 3.24620982413584e-05, + "loss": 1.0005, + "step": 202940 + }, + { + "epoch": 17.57, + "learning_rate": 3.246123191544659e-05, + "loss": 1.0132, + "step": 202950 + }, + { + "epoch": 17.57, + "learning_rate": 3.2460365589534785e-05, + "loss": 1.0185, + "step": 202960 + }, + { + "epoch": 17.57, + "learning_rate": 3.245949926362298e-05, + "loss": 0.9363, + "step": 202970 + }, + { + "epoch": 17.57, + "learning_rate": 3.245863293771117e-05, + "loss": 0.9745, + "step": 202980 + }, + { + "epoch": 17.57, + "learning_rate": 3.245776661179936e-05, + "loss": 1.0185, + "step": 202990 + }, + { + "epoch": 17.57, + "learning_rate": 3.2456900285887556e-05, + "loss": 0.9521, + "step": 203000 + }, + { + "epoch": 17.57, + "learning_rate": 3.2456033959975744e-05, + "loss": 1.0441, + "step": 203010 + }, + { + "epoch": 17.57, + "learning_rate": 3.245516763406394e-05, + "loss": 0.9879, + "step": 203020 + }, + { + "epoch": 17.57, + "learning_rate": 3.245430130815213e-05, + "loss": 1.0242, + "step": 203030 + }, + { + "epoch": 17.57, + "learning_rate": 3.245343498224032e-05, + "loss": 1.0501, + "step": 203040 + }, + { + "epoch": 17.58, + "learning_rate": 3.2452568656328515e-05, + "loss": 0.9936, + "step": 203050 + }, + { + "epoch": 17.58, + "learning_rate": 3.24517023304167e-05, + "loss": 1.0252, + "step": 203060 + }, + { + "epoch": 17.58, + "learning_rate": 3.24508360045049e-05, + "loss": 1.0016, + "step": 203070 + }, + { + "epoch": 17.58, + "learning_rate": 3.244996967859309e-05, + "loss": 0.9488, + "step": 203080 + }, + { + "epoch": 17.58, + "learning_rate": 3.244910335268128e-05, + "loss": 1.042, + "step": 203090 + }, + { + "epoch": 17.58, + "learning_rate": 3.2448237026769473e-05, + "loss": 0.9873, + "step": 203100 + }, + { + "epoch": 17.58, + "learning_rate": 3.244737070085767e-05, + "loss": 0.9146, + "step": 203110 + }, + { + "epoch": 17.58, + "learning_rate": 3.2446504374945856e-05, + "loss": 1.0085, + "step": 203120 + }, + { + "epoch": 17.58, + "learning_rate": 3.244563804903405e-05, + "loss": 1.0515, + "step": 203130 + }, + { + "epoch": 17.58, + "learning_rate": 3.2444771723122244e-05, + "loss": 0.9975, + "step": 203140 + }, + { + "epoch": 17.58, + "learning_rate": 3.244390539721043e-05, + "loss": 1.0259, + "step": 203150 + }, + { + "epoch": 17.58, + "learning_rate": 3.2443039071298627e-05, + "loss": 0.9931, + "step": 203160 + }, + { + "epoch": 17.59, + "learning_rate": 3.2442172745386814e-05, + "loss": 0.9789, + "step": 203170 + }, + { + "epoch": 17.59, + "learning_rate": 3.244130641947501e-05, + "loss": 0.9548, + "step": 203180 + }, + { + "epoch": 17.59, + "learning_rate": 3.24404400935632e-05, + "loss": 0.9887, + "step": 203190 + }, + { + "epoch": 17.59, + "learning_rate": 3.243957376765139e-05, + "loss": 1.0171, + "step": 203200 + }, + { + "epoch": 17.59, + "learning_rate": 3.2438707441739585e-05, + "loss": 1.0074, + "step": 203210 + }, + { + "epoch": 17.59, + "learning_rate": 3.243784111582778e-05, + "loss": 0.9687, + "step": 203220 + }, + { + "epoch": 17.59, + "learning_rate": 3.243697478991597e-05, + "loss": 0.9781, + "step": 203230 + }, + { + "epoch": 17.59, + "learning_rate": 3.243610846400416e-05, + "loss": 1.0149, + "step": 203240 + }, + { + "epoch": 17.59, + "learning_rate": 3.2435242138092356e-05, + "loss": 1.0608, + "step": 203250 + }, + { + "epoch": 17.59, + "learning_rate": 3.2434375812180544e-05, + "loss": 0.9377, + "step": 203260 + }, + { + "epoch": 17.59, + "learning_rate": 3.243350948626874e-05, + "loss": 1.0102, + "step": 203270 + }, + { + "epoch": 17.59, + "learning_rate": 3.2432643160356926e-05, + "loss": 1.0533, + "step": 203280 + }, + { + "epoch": 17.6, + "learning_rate": 3.243177683444512e-05, + "loss": 1.0058, + "step": 203290 + }, + { + "epoch": 17.6, + "learning_rate": 3.2430910508533315e-05, + "loss": 0.9915, + "step": 203300 + }, + { + "epoch": 17.6, + "learning_rate": 3.24300441826215e-05, + "loss": 0.9568, + "step": 203310 + }, + { + "epoch": 17.6, + "learning_rate": 3.24291778567097e-05, + "loss": 1.0336, + "step": 203320 + }, + { + "epoch": 17.6, + "learning_rate": 3.242831153079789e-05, + "loss": 0.9646, + "step": 203330 + }, + { + "epoch": 17.6, + "learning_rate": 3.242744520488608e-05, + "loss": 1.0226, + "step": 203340 + }, + { + "epoch": 17.6, + "learning_rate": 3.2426578878974274e-05, + "loss": 1.0156, + "step": 203350 + }, + { + "epoch": 17.6, + "learning_rate": 3.242571255306247e-05, + "loss": 1.0618, + "step": 203360 + }, + { + "epoch": 17.6, + "learning_rate": 3.2424846227150656e-05, + "loss": 1.0051, + "step": 203370 + }, + { + "epoch": 17.6, + "learning_rate": 3.242397990123885e-05, + "loss": 1.0036, + "step": 203380 + }, + { + "epoch": 17.6, + "learning_rate": 3.242311357532704e-05, + "loss": 0.9772, + "step": 203390 + }, + { + "epoch": 17.61, + "learning_rate": 3.242224724941523e-05, + "loss": 1.0059, + "step": 203400 + }, + { + "epoch": 17.61, + "learning_rate": 3.242138092350343e-05, + "loss": 0.9719, + "step": 203410 + }, + { + "epoch": 17.61, + "learning_rate": 3.2420514597591615e-05, + "loss": 0.982, + "step": 203420 + }, + { + "epoch": 17.61, + "learning_rate": 3.241964827167981e-05, + "loss": 1.0537, + "step": 203430 + }, + { + "epoch": 17.61, + "learning_rate": 3.2418781945768003e-05, + "loss": 1.0188, + "step": 203440 + }, + { + "epoch": 17.61, + "learning_rate": 3.241791561985619e-05, + "loss": 1.0271, + "step": 203450 + }, + { + "epoch": 17.61, + "learning_rate": 3.2417049293944386e-05, + "loss": 0.996, + "step": 203460 + }, + { + "epoch": 17.61, + "learning_rate": 3.241618296803258e-05, + "loss": 0.9618, + "step": 203470 + }, + { + "epoch": 17.61, + "learning_rate": 3.241531664212077e-05, + "loss": 0.9312, + "step": 203480 + }, + { + "epoch": 17.61, + "learning_rate": 3.241445031620896e-05, + "loss": 0.9698, + "step": 203490 + }, + { + "epoch": 17.61, + "learning_rate": 3.241358399029715e-05, + "loss": 1.0089, + "step": 203500 + }, + { + "epoch": 17.61, + "learning_rate": 3.2412717664385344e-05, + "loss": 0.9989, + "step": 203510 + }, + { + "epoch": 17.62, + "learning_rate": 3.241185133847354e-05, + "loss": 0.966, + "step": 203520 + }, + { + "epoch": 17.62, + "learning_rate": 3.2410985012561726e-05, + "loss": 1.0143, + "step": 203530 + }, + { + "epoch": 17.62, + "learning_rate": 3.241011868664992e-05, + "loss": 1.0172, + "step": 203540 + }, + { + "epoch": 17.62, + "learning_rate": 3.2409252360738115e-05, + "loss": 0.9544, + "step": 203550 + }, + { + "epoch": 17.62, + "learning_rate": 3.24083860348263e-05, + "loss": 0.991, + "step": 203560 + }, + { + "epoch": 17.62, + "learning_rate": 3.24075197089145e-05, + "loss": 0.9896, + "step": 203570 + }, + { + "epoch": 17.62, + "learning_rate": 3.240665338300269e-05, + "loss": 1.0242, + "step": 203580 + }, + { + "epoch": 17.62, + "learning_rate": 3.240578705709088e-05, + "loss": 0.9868, + "step": 203590 + }, + { + "epoch": 17.62, + "learning_rate": 3.2404920731179074e-05, + "loss": 1.0353, + "step": 203600 + }, + { + "epoch": 17.62, + "learning_rate": 3.240405440526726e-05, + "loss": 1.0146, + "step": 203610 + }, + { + "epoch": 17.62, + "learning_rate": 3.2403188079355456e-05, + "loss": 1.0009, + "step": 203620 + }, + { + "epoch": 17.63, + "learning_rate": 3.240232175344365e-05, + "loss": 1.0491, + "step": 203630 + }, + { + "epoch": 17.63, + "learning_rate": 3.240145542753184e-05, + "loss": 0.9875, + "step": 203640 + }, + { + "epoch": 17.63, + "learning_rate": 3.240058910162003e-05, + "loss": 0.9887, + "step": 203650 + }, + { + "epoch": 17.63, + "learning_rate": 3.239972277570823e-05, + "loss": 1.0116, + "step": 203660 + }, + { + "epoch": 17.63, + "learning_rate": 3.2398856449796415e-05, + "loss": 1.0074, + "step": 203670 + }, + { + "epoch": 17.63, + "learning_rate": 3.239799012388461e-05, + "loss": 0.9719, + "step": 203680 + }, + { + "epoch": 17.63, + "learning_rate": 3.23971237979728e-05, + "loss": 0.9411, + "step": 203690 + }, + { + "epoch": 17.63, + "learning_rate": 3.239625747206099e-05, + "loss": 0.9743, + "step": 203700 + }, + { + "epoch": 17.63, + "learning_rate": 3.2395391146149186e-05, + "loss": 0.9659, + "step": 203710 + }, + { + "epoch": 17.63, + "learning_rate": 3.2394524820237374e-05, + "loss": 0.9688, + "step": 203720 + }, + { + "epoch": 17.63, + "learning_rate": 3.239365849432557e-05, + "loss": 0.9946, + "step": 203730 + }, + { + "epoch": 17.63, + "learning_rate": 3.239279216841376e-05, + "loss": 1.0415, + "step": 203740 + }, + { + "epoch": 17.64, + "learning_rate": 3.239192584250195e-05, + "loss": 0.9459, + "step": 203750 + }, + { + "epoch": 17.64, + "learning_rate": 3.2391059516590145e-05, + "loss": 1.0086, + "step": 203760 + }, + { + "epoch": 17.64, + "learning_rate": 3.239019319067834e-05, + "loss": 1.0004, + "step": 203770 + }, + { + "epoch": 17.64, + "learning_rate": 3.238932686476653e-05, + "loss": 0.9851, + "step": 203780 + }, + { + "epoch": 17.64, + "learning_rate": 3.238846053885472e-05, + "loss": 0.8984, + "step": 203790 + }, + { + "epoch": 17.64, + "learning_rate": 3.238759421294291e-05, + "loss": 0.9854, + "step": 203800 + }, + { + "epoch": 17.64, + "learning_rate": 3.23867278870311e-05, + "loss": 0.9864, + "step": 203810 + }, + { + "epoch": 17.64, + "learning_rate": 3.23858615611193e-05, + "loss": 0.9418, + "step": 203820 + }, + { + "epoch": 17.64, + "learning_rate": 3.2384995235207486e-05, + "loss": 0.9463, + "step": 203830 + }, + { + "epoch": 17.64, + "learning_rate": 3.238412890929568e-05, + "loss": 0.9837, + "step": 203840 + }, + { + "epoch": 17.64, + "learning_rate": 3.2383262583383874e-05, + "loss": 1.0085, + "step": 203850 + }, + { + "epoch": 17.65, + "learning_rate": 3.238239625747206e-05, + "loss": 0.9796, + "step": 203860 + }, + { + "epoch": 17.65, + "learning_rate": 3.2381529931560257e-05, + "loss": 1.0414, + "step": 203870 + }, + { + "epoch": 17.65, + "learning_rate": 3.238066360564845e-05, + "loss": 0.9794, + "step": 203880 + }, + { + "epoch": 17.65, + "learning_rate": 3.237979727973664e-05, + "loss": 1.0549, + "step": 203890 + }, + { + "epoch": 17.65, + "learning_rate": 3.237893095382483e-05, + "loss": 1.0183, + "step": 203900 + }, + { + "epoch": 17.65, + "learning_rate": 3.237806462791302e-05, + "loss": 0.9976, + "step": 203910 + }, + { + "epoch": 17.65, + "learning_rate": 3.2377198302001215e-05, + "loss": 0.9766, + "step": 203920 + }, + { + "epoch": 17.65, + "learning_rate": 3.237633197608941e-05, + "loss": 0.9762, + "step": 203930 + }, + { + "epoch": 17.65, + "learning_rate": 3.23754656501776e-05, + "loss": 0.9569, + "step": 203940 + }, + { + "epoch": 17.65, + "learning_rate": 3.237459932426579e-05, + "loss": 1.0332, + "step": 203950 + }, + { + "epoch": 17.65, + "learning_rate": 3.2373732998353986e-05, + "loss": 0.979, + "step": 203960 + }, + { + "epoch": 17.65, + "learning_rate": 3.2372866672442174e-05, + "loss": 1.0049, + "step": 203970 + }, + { + "epoch": 17.66, + "learning_rate": 3.237200034653037e-05, + "loss": 1.0324, + "step": 203980 + }, + { + "epoch": 17.66, + "learning_rate": 3.237113402061856e-05, + "loss": 0.9988, + "step": 203990 + }, + { + "epoch": 17.66, + "learning_rate": 3.237026769470675e-05, + "loss": 0.937, + "step": 204000 + }, + { + "epoch": 17.66, + "learning_rate": 3.2369401368794945e-05, + "loss": 0.9855, + "step": 204010 + }, + { + "epoch": 17.66, + "learning_rate": 3.236853504288313e-05, + "loss": 1.0009, + "step": 204020 + }, + { + "epoch": 17.66, + "learning_rate": 3.236766871697133e-05, + "loss": 1.0086, + "step": 204030 + }, + { + "epoch": 17.66, + "learning_rate": 3.236680239105952e-05, + "loss": 0.9912, + "step": 204040 + }, + { + "epoch": 17.66, + "learning_rate": 3.236593606514771e-05, + "loss": 1.0425, + "step": 204050 + }, + { + "epoch": 17.66, + "learning_rate": 3.2365069739235904e-05, + "loss": 1.0031, + "step": 204060 + }, + { + "epoch": 17.66, + "learning_rate": 3.23642034133241e-05, + "loss": 1.0086, + "step": 204070 + }, + { + "epoch": 17.66, + "learning_rate": 3.2363337087412286e-05, + "loss": 0.9458, + "step": 204080 + }, + { + "epoch": 17.67, + "learning_rate": 3.236247076150048e-05, + "loss": 1.0355, + "step": 204090 + }, + { + "epoch": 17.67, + "learning_rate": 3.2361604435588675e-05, + "loss": 1.0338, + "step": 204100 + }, + { + "epoch": 17.67, + "learning_rate": 3.236073810967686e-05, + "loss": 1.0207, + "step": 204110 + }, + { + "epoch": 17.67, + "learning_rate": 3.235987178376506e-05, + "loss": 0.9912, + "step": 204120 + }, + { + "epoch": 17.67, + "learning_rate": 3.2359005457853245e-05, + "loss": 1.0015, + "step": 204130 + }, + { + "epoch": 17.67, + "learning_rate": 3.235813913194144e-05, + "loss": 0.9826, + "step": 204140 + }, + { + "epoch": 17.67, + "learning_rate": 3.2357272806029633e-05, + "loss": 0.9707, + "step": 204150 + }, + { + "epoch": 17.67, + "learning_rate": 3.235640648011782e-05, + "loss": 1.0124, + "step": 204160 + }, + { + "epoch": 17.67, + "learning_rate": 3.2355540154206016e-05, + "loss": 1.0032, + "step": 204170 + }, + { + "epoch": 17.67, + "learning_rate": 3.235467382829421e-05, + "loss": 1.0605, + "step": 204180 + }, + { + "epoch": 17.67, + "learning_rate": 3.23538075023824e-05, + "loss": 0.9795, + "step": 204190 + }, + { + "epoch": 17.67, + "learning_rate": 3.235294117647059e-05, + "loss": 1.0497, + "step": 204200 + }, + { + "epoch": 17.68, + "learning_rate": 3.235207485055879e-05, + "loss": 0.9675, + "step": 204210 + }, + { + "epoch": 17.68, + "learning_rate": 3.2351208524646974e-05, + "loss": 0.9277, + "step": 204220 + }, + { + "epoch": 17.68, + "learning_rate": 3.235034219873517e-05, + "loss": 1.0368, + "step": 204230 + }, + { + "epoch": 17.68, + "learning_rate": 3.2349475872823356e-05, + "loss": 1.0245, + "step": 204240 + }, + { + "epoch": 17.68, + "learning_rate": 3.234860954691155e-05, + "loss": 0.9972, + "step": 204250 + }, + { + "epoch": 17.68, + "learning_rate": 3.2347743220999745e-05, + "loss": 0.9725, + "step": 204260 + }, + { + "epoch": 17.68, + "learning_rate": 3.234687689508793e-05, + "loss": 0.9619, + "step": 204270 + }, + { + "epoch": 17.68, + "learning_rate": 3.234601056917613e-05, + "loss": 0.9839, + "step": 204280 + }, + { + "epoch": 17.68, + "learning_rate": 3.234514424326432e-05, + "loss": 1.0178, + "step": 204290 + }, + { + "epoch": 17.68, + "learning_rate": 3.234427791735251e-05, + "loss": 1.0181, + "step": 204300 + }, + { + "epoch": 17.68, + "learning_rate": 3.2343411591440704e-05, + "loss": 1.0223, + "step": 204310 + }, + { + "epoch": 17.68, + "learning_rate": 3.23425452655289e-05, + "loss": 0.9826, + "step": 204320 + }, + { + "epoch": 17.69, + "learning_rate": 3.2341678939617086e-05, + "loss": 1.0271, + "step": 204330 + }, + { + "epoch": 17.69, + "learning_rate": 3.234081261370528e-05, + "loss": 1.0226, + "step": 204340 + }, + { + "epoch": 17.69, + "learning_rate": 3.233994628779347e-05, + "loss": 1.0271, + "step": 204350 + }, + { + "epoch": 17.69, + "learning_rate": 3.233907996188166e-05, + "loss": 0.9891, + "step": 204360 + }, + { + "epoch": 17.69, + "learning_rate": 3.233821363596986e-05, + "loss": 1.066, + "step": 204370 + }, + { + "epoch": 17.69, + "learning_rate": 3.2337347310058045e-05, + "loss": 0.9924, + "step": 204380 + }, + { + "epoch": 17.69, + "learning_rate": 3.233648098414624e-05, + "loss": 1.0041, + "step": 204390 + }, + { + "epoch": 17.69, + "learning_rate": 3.2335614658234434e-05, + "loss": 0.9579, + "step": 204400 + }, + { + "epoch": 17.69, + "learning_rate": 3.233474833232262e-05, + "loss": 1.0051, + "step": 204410 + }, + { + "epoch": 17.69, + "learning_rate": 3.2333882006410816e-05, + "loss": 1.0143, + "step": 204420 + }, + { + "epoch": 17.69, + "learning_rate": 3.2333015680499004e-05, + "loss": 0.9974, + "step": 204430 + }, + { + "epoch": 17.7, + "learning_rate": 3.23321493545872e-05, + "loss": 0.985, + "step": 204440 + }, + { + "epoch": 17.7, + "learning_rate": 3.233128302867539e-05, + "loss": 1.0536, + "step": 204450 + }, + { + "epoch": 17.7, + "learning_rate": 3.233041670276358e-05, + "loss": 0.9848, + "step": 204460 + }, + { + "epoch": 17.7, + "learning_rate": 3.2329550376851775e-05, + "loss": 1.0391, + "step": 204470 + }, + { + "epoch": 17.7, + "learning_rate": 3.232868405093997e-05, + "loss": 1.0032, + "step": 204480 + }, + { + "epoch": 17.7, + "learning_rate": 3.232781772502816e-05, + "loss": 0.9919, + "step": 204490 + }, + { + "epoch": 17.7, + "learning_rate": 3.232695139911635e-05, + "loss": 0.9821, + "step": 204500 + }, + { + "epoch": 17.7, + "learning_rate": 3.2326085073204546e-05, + "loss": 1.0001, + "step": 204510 + }, + { + "epoch": 17.7, + "learning_rate": 3.232521874729273e-05, + "loss": 1.0292, + "step": 204520 + }, + { + "epoch": 17.7, + "learning_rate": 3.232435242138093e-05, + "loss": 1.0028, + "step": 204530 + }, + { + "epoch": 17.7, + "learning_rate": 3.2323486095469115e-05, + "loss": 0.9968, + "step": 204540 + }, + { + "epoch": 17.7, + "learning_rate": 3.232261976955731e-05, + "loss": 1.0463, + "step": 204550 + }, + { + "epoch": 17.71, + "learning_rate": 3.2321753443645504e-05, + "loss": 0.9923, + "step": 204560 + }, + { + "epoch": 17.71, + "learning_rate": 3.232088711773369e-05, + "loss": 0.9711, + "step": 204570 + }, + { + "epoch": 17.71, + "learning_rate": 3.2320020791821887e-05, + "loss": 0.9807, + "step": 204580 + }, + { + "epoch": 17.71, + "learning_rate": 3.231915446591008e-05, + "loss": 0.9525, + "step": 204590 + }, + { + "epoch": 17.71, + "learning_rate": 3.231828813999827e-05, + "loss": 1.0087, + "step": 204600 + }, + { + "epoch": 17.71, + "learning_rate": 3.231742181408646e-05, + "loss": 0.9821, + "step": 204610 + }, + { + "epoch": 17.71, + "learning_rate": 3.231655548817466e-05, + "loss": 0.9828, + "step": 204620 + }, + { + "epoch": 17.71, + "learning_rate": 3.2315689162262845e-05, + "loss": 0.9849, + "step": 204630 + }, + { + "epoch": 17.71, + "learning_rate": 3.231482283635104e-05, + "loss": 1.0239, + "step": 204640 + }, + { + "epoch": 17.71, + "learning_rate": 3.231395651043923e-05, + "loss": 0.9695, + "step": 204650 + }, + { + "epoch": 17.71, + "learning_rate": 3.231309018452742e-05, + "loss": 0.9808, + "step": 204660 + }, + { + "epoch": 17.72, + "learning_rate": 3.2312223858615616e-05, + "loss": 1.0227, + "step": 204670 + }, + { + "epoch": 17.72, + "learning_rate": 3.2311357532703804e-05, + "loss": 0.9987, + "step": 204680 + }, + { + "epoch": 17.72, + "learning_rate": 3.2310491206792e-05, + "loss": 1.0062, + "step": 204690 + }, + { + "epoch": 17.72, + "learning_rate": 3.230962488088019e-05, + "loss": 1.0014, + "step": 204700 + }, + { + "epoch": 17.72, + "learning_rate": 3.230875855496838e-05, + "loss": 1.0269, + "step": 204710 + }, + { + "epoch": 17.72, + "learning_rate": 3.2307892229056575e-05, + "loss": 1.0554, + "step": 204720 + }, + { + "epoch": 17.72, + "learning_rate": 3.230702590314477e-05, + "loss": 0.9804, + "step": 204730 + }, + { + "epoch": 17.72, + "learning_rate": 3.230615957723296e-05, + "loss": 1.0019, + "step": 204740 + }, + { + "epoch": 17.72, + "learning_rate": 3.230529325132115e-05, + "loss": 1.0375, + "step": 204750 + }, + { + "epoch": 17.72, + "learning_rate": 3.230442692540934e-05, + "loss": 1.0061, + "step": 204760 + }, + { + "epoch": 17.72, + "learning_rate": 3.2303560599497534e-05, + "loss": 1.0237, + "step": 204770 + }, + { + "epoch": 17.72, + "learning_rate": 3.230269427358573e-05, + "loss": 0.9838, + "step": 204780 + }, + { + "epoch": 17.73, + "learning_rate": 3.2301827947673916e-05, + "loss": 1.0022, + "step": 204790 + }, + { + "epoch": 17.73, + "learning_rate": 3.230096162176211e-05, + "loss": 0.9822, + "step": 204800 + }, + { + "epoch": 17.73, + "learning_rate": 3.2300095295850305e-05, + "loss": 0.9537, + "step": 204810 + }, + { + "epoch": 17.73, + "learning_rate": 3.229922896993849e-05, + "loss": 1.0613, + "step": 204820 + }, + { + "epoch": 17.73, + "learning_rate": 3.229836264402669e-05, + "loss": 0.9992, + "step": 204830 + }, + { + "epoch": 17.73, + "learning_rate": 3.229749631811488e-05, + "loss": 0.9718, + "step": 204840 + }, + { + "epoch": 17.73, + "learning_rate": 3.229662999220307e-05, + "loss": 1.0249, + "step": 204850 + }, + { + "epoch": 17.73, + "learning_rate": 3.2295763666291263e-05, + "loss": 1.0161, + "step": 204860 + }, + { + "epoch": 17.73, + "learning_rate": 3.229489734037945e-05, + "loss": 0.9909, + "step": 204870 + }, + { + "epoch": 17.73, + "learning_rate": 3.2294031014467646e-05, + "loss": 0.9802, + "step": 204880 + }, + { + "epoch": 17.73, + "learning_rate": 3.229316468855584e-05, + "loss": 0.9965, + "step": 204890 + }, + { + "epoch": 17.74, + "learning_rate": 3.229229836264403e-05, + "loss": 0.9792, + "step": 204900 + }, + { + "epoch": 17.74, + "learning_rate": 3.229143203673222e-05, + "loss": 1.0033, + "step": 204910 + }, + { + "epoch": 17.74, + "learning_rate": 3.2290565710820417e-05, + "loss": 0.9495, + "step": 204920 + }, + { + "epoch": 17.74, + "learning_rate": 3.2289699384908604e-05, + "loss": 0.9595, + "step": 204930 + }, + { + "epoch": 17.74, + "learning_rate": 3.22888330589968e-05, + "loss": 0.9871, + "step": 204940 + }, + { + "epoch": 17.74, + "learning_rate": 3.228796673308499e-05, + "loss": 0.9982, + "step": 204950 + }, + { + "epoch": 17.74, + "learning_rate": 3.228710040717318e-05, + "loss": 1.0043, + "step": 204960 + }, + { + "epoch": 17.74, + "learning_rate": 3.2286234081261375e-05, + "loss": 0.9757, + "step": 204970 + }, + { + "epoch": 17.74, + "learning_rate": 3.228536775534956e-05, + "loss": 1.0027, + "step": 204980 + }, + { + "epoch": 17.74, + "learning_rate": 3.228450142943776e-05, + "loss": 0.9685, + "step": 204990 + }, + { + "epoch": 17.74, + "learning_rate": 3.228363510352595e-05, + "loss": 0.9712, + "step": 205000 + }, + { + "epoch": 17.74, + "learning_rate": 3.228276877761414e-05, + "loss": 1.0061, + "step": 205010 + }, + { + "epoch": 17.75, + "learning_rate": 3.2281902451702334e-05, + "loss": 1.0314, + "step": 205020 + }, + { + "epoch": 17.75, + "learning_rate": 3.228103612579053e-05, + "loss": 0.9502, + "step": 205030 + }, + { + "epoch": 17.75, + "learning_rate": 3.2280169799878716e-05, + "loss": 1.0063, + "step": 205040 + }, + { + "epoch": 17.75, + "learning_rate": 3.227930347396691e-05, + "loss": 0.9924, + "step": 205050 + }, + { + "epoch": 17.75, + "learning_rate": 3.2278437148055105e-05, + "loss": 0.9158, + "step": 205060 + }, + { + "epoch": 17.75, + "learning_rate": 3.227757082214329e-05, + "loss": 0.9818, + "step": 205070 + }, + { + "epoch": 17.75, + "learning_rate": 3.227670449623149e-05, + "loss": 0.9722, + "step": 205080 + }, + { + "epoch": 17.75, + "learning_rate": 3.2275838170319675e-05, + "loss": 1.0055, + "step": 205090 + }, + { + "epoch": 17.75, + "learning_rate": 3.227497184440787e-05, + "loss": 0.9913, + "step": 205100 + }, + { + "epoch": 17.75, + "learning_rate": 3.2274105518496064e-05, + "loss": 0.9424, + "step": 205110 + }, + { + "epoch": 17.75, + "learning_rate": 3.227323919258425e-05, + "loss": 0.9551, + "step": 205120 + }, + { + "epoch": 17.76, + "learning_rate": 3.2272372866672446e-05, + "loss": 0.9398, + "step": 205130 + }, + { + "epoch": 17.76, + "learning_rate": 3.227150654076064e-05, + "loss": 1.0026, + "step": 205140 + }, + { + "epoch": 17.76, + "learning_rate": 3.227064021484883e-05, + "loss": 1.0268, + "step": 205150 + }, + { + "epoch": 17.76, + "learning_rate": 3.226977388893702e-05, + "loss": 0.9696, + "step": 205160 + }, + { + "epoch": 17.76, + "learning_rate": 3.226890756302522e-05, + "loss": 1.0029, + "step": 205170 + }, + { + "epoch": 17.76, + "learning_rate": 3.2268041237113405e-05, + "loss": 0.9165, + "step": 205180 + }, + { + "epoch": 17.76, + "learning_rate": 3.22671749112016e-05, + "loss": 0.993, + "step": 205190 + }, + { + "epoch": 17.76, + "learning_rate": 3.226630858528979e-05, + "loss": 0.9515, + "step": 205200 + }, + { + "epoch": 17.76, + "learning_rate": 3.226544225937798e-05, + "loss": 1.041, + "step": 205210 + }, + { + "epoch": 17.76, + "learning_rate": 3.2264575933466176e-05, + "loss": 1.0595, + "step": 205220 + }, + { + "epoch": 17.76, + "learning_rate": 3.226370960755436e-05, + "loss": 0.9858, + "step": 205230 + }, + { + "epoch": 17.76, + "learning_rate": 3.226284328164256e-05, + "loss": 0.9878, + "step": 205240 + }, + { + "epoch": 17.77, + "learning_rate": 3.226197695573075e-05, + "loss": 1.0202, + "step": 205250 + }, + { + "epoch": 17.77, + "learning_rate": 3.226111062981894e-05, + "loss": 1.0724, + "step": 205260 + }, + { + "epoch": 17.77, + "learning_rate": 3.2260244303907134e-05, + "loss": 0.9567, + "step": 205270 + }, + { + "epoch": 17.77, + "learning_rate": 3.225937797799532e-05, + "loss": 1.014, + "step": 205280 + }, + { + "epoch": 17.77, + "learning_rate": 3.2258511652083516e-05, + "loss": 0.9613, + "step": 205290 + }, + { + "epoch": 17.77, + "learning_rate": 3.225764532617171e-05, + "loss": 1.0097, + "step": 205300 + }, + { + "epoch": 17.77, + "learning_rate": 3.22567790002599e-05, + "loss": 1.0196, + "step": 205310 + }, + { + "epoch": 17.77, + "learning_rate": 3.225591267434809e-05, + "loss": 0.9946, + "step": 205320 + }, + { + "epoch": 17.77, + "learning_rate": 3.225504634843629e-05, + "loss": 0.9758, + "step": 205330 + }, + { + "epoch": 17.77, + "learning_rate": 3.2254180022524475e-05, + "loss": 1.0189, + "step": 205340 + }, + { + "epoch": 17.77, + "learning_rate": 3.225331369661267e-05, + "loss": 0.9734, + "step": 205350 + }, + { + "epoch": 17.77, + "learning_rate": 3.2252447370700864e-05, + "loss": 0.9972, + "step": 205360 + }, + { + "epoch": 17.78, + "learning_rate": 3.225158104478905e-05, + "loss": 1.017, + "step": 205370 + }, + { + "epoch": 17.78, + "learning_rate": 3.2250714718877246e-05, + "loss": 0.9966, + "step": 205380 + }, + { + "epoch": 17.78, + "learning_rate": 3.2249848392965434e-05, + "loss": 1.0173, + "step": 205390 + }, + { + "epoch": 17.78, + "learning_rate": 3.224898206705363e-05, + "loss": 0.9437, + "step": 205400 + }, + { + "epoch": 17.78, + "learning_rate": 3.224811574114182e-05, + "loss": 0.9997, + "step": 205410 + }, + { + "epoch": 17.78, + "learning_rate": 3.224724941523001e-05, + "loss": 1.022, + "step": 205420 + }, + { + "epoch": 17.78, + "learning_rate": 3.2246383089318205e-05, + "loss": 1.0441, + "step": 205430 + }, + { + "epoch": 17.78, + "learning_rate": 3.22455167634064e-05, + "loss": 1.0537, + "step": 205440 + }, + { + "epoch": 17.78, + "learning_rate": 3.224465043749459e-05, + "loss": 1.0146, + "step": 205450 + }, + { + "epoch": 17.78, + "learning_rate": 3.224378411158278e-05, + "loss": 0.997, + "step": 205460 + }, + { + "epoch": 17.78, + "learning_rate": 3.2242917785670976e-05, + "loss": 1.0247, + "step": 205470 + }, + { + "epoch": 17.79, + "learning_rate": 3.2242051459759164e-05, + "loss": 0.9883, + "step": 205480 + }, + { + "epoch": 17.79, + "learning_rate": 3.224118513384736e-05, + "loss": 0.9736, + "step": 205490 + }, + { + "epoch": 17.79, + "learning_rate": 3.2240318807935546e-05, + "loss": 1.0445, + "step": 205500 + }, + { + "epoch": 17.79, + "learning_rate": 3.223945248202374e-05, + "loss": 1.0506, + "step": 205510 + }, + { + "epoch": 17.79, + "learning_rate": 3.2238586156111935e-05, + "loss": 1.0404, + "step": 205520 + }, + { + "epoch": 17.79, + "learning_rate": 3.223771983020012e-05, + "loss": 1.0261, + "step": 205530 + }, + { + "epoch": 17.79, + "learning_rate": 3.223685350428832e-05, + "loss": 1.0083, + "step": 205540 + }, + { + "epoch": 17.79, + "learning_rate": 3.223598717837651e-05, + "loss": 1.0142, + "step": 205550 + }, + { + "epoch": 17.79, + "learning_rate": 3.22351208524647e-05, + "loss": 0.9696, + "step": 205560 + }, + { + "epoch": 17.79, + "learning_rate": 3.2234254526552893e-05, + "loss": 1.0419, + "step": 205570 + }, + { + "epoch": 17.79, + "learning_rate": 3.223338820064109e-05, + "loss": 1.0338, + "step": 205580 + }, + { + "epoch": 17.79, + "learning_rate": 3.2232521874729276e-05, + "loss": 1.0358, + "step": 205590 + }, + { + "epoch": 17.8, + "learning_rate": 3.223165554881747e-05, + "loss": 0.9854, + "step": 205600 + }, + { + "epoch": 17.8, + "learning_rate": 3.223078922290566e-05, + "loss": 0.9422, + "step": 205610 + }, + { + "epoch": 17.8, + "learning_rate": 3.222992289699385e-05, + "loss": 1.0011, + "step": 205620 + }, + { + "epoch": 17.8, + "learning_rate": 3.2229056571082047e-05, + "loss": 1.0035, + "step": 205630 + }, + { + "epoch": 17.8, + "learning_rate": 3.2228190245170234e-05, + "loss": 0.9431, + "step": 205640 + }, + { + "epoch": 17.8, + "learning_rate": 3.222732391925843e-05, + "loss": 1.0448, + "step": 205650 + }, + { + "epoch": 17.8, + "learning_rate": 3.222645759334662e-05, + "loss": 0.9921, + "step": 205660 + }, + { + "epoch": 17.8, + "learning_rate": 3.222559126743481e-05, + "loss": 0.9812, + "step": 205670 + }, + { + "epoch": 17.8, + "learning_rate": 3.2224724941523005e-05, + "loss": 0.9996, + "step": 205680 + }, + { + "epoch": 17.8, + "learning_rate": 3.22238586156112e-05, + "loss": 1.0165, + "step": 205690 + }, + { + "epoch": 17.8, + "learning_rate": 3.222299228969939e-05, + "loss": 1.0522, + "step": 205700 + }, + { + "epoch": 17.81, + "learning_rate": 3.222212596378758e-05, + "loss": 0.9797, + "step": 205710 + }, + { + "epoch": 17.81, + "learning_rate": 3.222125963787577e-05, + "loss": 0.9702, + "step": 205720 + }, + { + "epoch": 17.81, + "learning_rate": 3.2220393311963964e-05, + "loss": 1.0362, + "step": 205730 + }, + { + "epoch": 17.81, + "learning_rate": 3.221952698605216e-05, + "loss": 1.0112, + "step": 205740 + }, + { + "epoch": 17.81, + "learning_rate": 3.2218660660140346e-05, + "loss": 1.0429, + "step": 205750 + }, + { + "epoch": 17.81, + "learning_rate": 3.221779433422854e-05, + "loss": 1.0212, + "step": 205760 + }, + { + "epoch": 17.81, + "learning_rate": 3.2216928008316735e-05, + "loss": 1.0221, + "step": 205770 + }, + { + "epoch": 17.81, + "learning_rate": 3.221606168240492e-05, + "loss": 0.9977, + "step": 205780 + }, + { + "epoch": 17.81, + "learning_rate": 3.221519535649312e-05, + "loss": 0.9483, + "step": 205790 + }, + { + "epoch": 17.81, + "learning_rate": 3.221432903058131e-05, + "loss": 1.0195, + "step": 205800 + }, + { + "epoch": 17.81, + "learning_rate": 3.22134627046695e-05, + "loss": 0.9882, + "step": 205810 + }, + { + "epoch": 17.81, + "learning_rate": 3.2212596378757694e-05, + "loss": 1.0042, + "step": 205820 + }, + { + "epoch": 17.82, + "learning_rate": 3.221173005284588e-05, + "loss": 1.024, + "step": 205830 + }, + { + "epoch": 17.82, + "learning_rate": 3.2210863726934076e-05, + "loss": 0.924, + "step": 205840 + }, + { + "epoch": 17.82, + "learning_rate": 3.220999740102227e-05, + "loss": 0.9952, + "step": 205850 + }, + { + "epoch": 17.82, + "learning_rate": 3.220913107511046e-05, + "loss": 1.0001, + "step": 205860 + }, + { + "epoch": 17.82, + "learning_rate": 3.220826474919865e-05, + "loss": 1.037, + "step": 205870 + }, + { + "epoch": 17.82, + "learning_rate": 3.220739842328685e-05, + "loss": 0.9806, + "step": 205880 + }, + { + "epoch": 17.82, + "learning_rate": 3.2206532097375035e-05, + "loss": 1.0131, + "step": 205890 + }, + { + "epoch": 17.82, + "learning_rate": 3.220566577146323e-05, + "loss": 0.9745, + "step": 205900 + }, + { + "epoch": 17.82, + "learning_rate": 3.2204799445551423e-05, + "loss": 1.0067, + "step": 205910 + }, + { + "epoch": 17.82, + "learning_rate": 3.220393311963961e-05, + "loss": 0.9864, + "step": 205920 + }, + { + "epoch": 17.82, + "learning_rate": 3.2203066793727806e-05, + "loss": 0.9749, + "step": 205930 + }, + { + "epoch": 17.83, + "learning_rate": 3.220220046781599e-05, + "loss": 1.0327, + "step": 205940 + }, + { + "epoch": 17.83, + "learning_rate": 3.220133414190419e-05, + "loss": 1.0133, + "step": 205950 + }, + { + "epoch": 17.83, + "learning_rate": 3.220046781599238e-05, + "loss": 1.0112, + "step": 205960 + }, + { + "epoch": 17.83, + "learning_rate": 3.219960149008057e-05, + "loss": 0.9801, + "step": 205970 + }, + { + "epoch": 17.83, + "learning_rate": 3.2198735164168764e-05, + "loss": 1.025, + "step": 205980 + }, + { + "epoch": 17.83, + "learning_rate": 3.219786883825696e-05, + "loss": 1.0342, + "step": 205990 + }, + { + "epoch": 17.83, + "learning_rate": 3.2197002512345146e-05, + "loss": 1.0339, + "step": 206000 + }, + { + "epoch": 17.83, + "learning_rate": 3.219613618643334e-05, + "loss": 0.9836, + "step": 206010 + }, + { + "epoch": 17.83, + "learning_rate": 3.219526986052153e-05, + "loss": 0.9924, + "step": 206020 + }, + { + "epoch": 17.83, + "learning_rate": 3.219440353460972e-05, + "loss": 0.9579, + "step": 206030 + }, + { + "epoch": 17.83, + "learning_rate": 3.219353720869792e-05, + "loss": 1.0119, + "step": 206040 + }, + { + "epoch": 17.83, + "learning_rate": 3.2192670882786105e-05, + "loss": 1.035, + "step": 206050 + }, + { + "epoch": 17.84, + "learning_rate": 3.21918045568743e-05, + "loss": 1.0548, + "step": 206060 + }, + { + "epoch": 17.84, + "learning_rate": 3.2190938230962494e-05, + "loss": 1.0141, + "step": 206070 + }, + { + "epoch": 17.84, + "learning_rate": 3.219007190505068e-05, + "loss": 0.9574, + "step": 206080 + }, + { + "epoch": 17.84, + "learning_rate": 3.2189205579138876e-05, + "loss": 1.0378, + "step": 206090 + }, + { + "epoch": 17.84, + "learning_rate": 3.218833925322707e-05, + "loss": 1.0219, + "step": 206100 + }, + { + "epoch": 17.84, + "learning_rate": 3.218747292731526e-05, + "loss": 0.9981, + "step": 206110 + }, + { + "epoch": 17.84, + "learning_rate": 3.218660660140345e-05, + "loss": 1.0237, + "step": 206120 + }, + { + "epoch": 17.84, + "learning_rate": 3.218574027549164e-05, + "loss": 0.9614, + "step": 206130 + }, + { + "epoch": 17.84, + "learning_rate": 3.2184873949579835e-05, + "loss": 1.0114, + "step": 206140 + }, + { + "epoch": 17.84, + "learning_rate": 3.218400762366803e-05, + "loss": 0.9717, + "step": 206150 + }, + { + "epoch": 17.84, + "learning_rate": 3.218314129775622e-05, + "loss": 1.0093, + "step": 206160 + }, + { + "epoch": 17.85, + "learning_rate": 3.218227497184441e-05, + "loss": 1.0019, + "step": 206170 + }, + { + "epoch": 17.85, + "learning_rate": 3.2181408645932606e-05, + "loss": 1.025, + "step": 206180 + }, + { + "epoch": 17.85, + "learning_rate": 3.2180542320020794e-05, + "loss": 0.9282, + "step": 206190 + }, + { + "epoch": 17.85, + "learning_rate": 3.217967599410899e-05, + "loss": 1.011, + "step": 206200 + }, + { + "epoch": 17.85, + "learning_rate": 3.217880966819718e-05, + "loss": 0.9575, + "step": 206210 + }, + { + "epoch": 17.85, + "learning_rate": 3.217794334228537e-05, + "loss": 0.9924, + "step": 206220 + }, + { + "epoch": 17.85, + "learning_rate": 3.2177077016373565e-05, + "loss": 0.9945, + "step": 206230 + }, + { + "epoch": 17.85, + "learning_rate": 3.217621069046175e-05, + "loss": 1.0397, + "step": 206240 + }, + { + "epoch": 17.85, + "learning_rate": 3.217534436454995e-05, + "loss": 0.9768, + "step": 206250 + }, + { + "epoch": 17.85, + "learning_rate": 3.217447803863814e-05, + "loss": 1.0163, + "step": 206260 + }, + { + "epoch": 17.85, + "learning_rate": 3.217361171272633e-05, + "loss": 0.9994, + "step": 206270 + }, + { + "epoch": 17.85, + "learning_rate": 3.217274538681452e-05, + "loss": 1.0217, + "step": 206280 + }, + { + "epoch": 17.86, + "learning_rate": 3.217187906090272e-05, + "loss": 0.9624, + "step": 206290 + }, + { + "epoch": 17.86, + "learning_rate": 3.2171012734990906e-05, + "loss": 0.9529, + "step": 206300 + }, + { + "epoch": 17.86, + "learning_rate": 3.21701464090791e-05, + "loss": 0.9995, + "step": 206310 + }, + { + "epoch": 17.86, + "learning_rate": 3.2169280083167294e-05, + "loss": 0.9866, + "step": 206320 + }, + { + "epoch": 17.86, + "learning_rate": 3.216841375725548e-05, + "loss": 0.9971, + "step": 206330 + }, + { + "epoch": 17.86, + "learning_rate": 3.2167547431343677e-05, + "loss": 0.9483, + "step": 206340 + }, + { + "epoch": 17.86, + "learning_rate": 3.2166681105431864e-05, + "loss": 0.955, + "step": 206350 + }, + { + "epoch": 17.86, + "learning_rate": 3.216581477952006e-05, + "loss": 0.9544, + "step": 206360 + }, + { + "epoch": 17.86, + "learning_rate": 3.216494845360825e-05, + "loss": 1.0025, + "step": 206370 + }, + { + "epoch": 17.86, + "learning_rate": 3.216408212769644e-05, + "loss": 0.9731, + "step": 206380 + }, + { + "epoch": 17.86, + "learning_rate": 3.2163215801784635e-05, + "loss": 0.9705, + "step": 206390 + }, + { + "epoch": 17.86, + "learning_rate": 3.216234947587283e-05, + "loss": 0.9234, + "step": 206400 + }, + { + "epoch": 17.87, + "learning_rate": 3.216148314996102e-05, + "loss": 0.9819, + "step": 206410 + }, + { + "epoch": 17.87, + "learning_rate": 3.216061682404921e-05, + "loss": 1.0105, + "step": 206420 + }, + { + "epoch": 17.87, + "learning_rate": 3.2159750498137406e-05, + "loss": 1.0055, + "step": 206430 + }, + { + "epoch": 17.87, + "learning_rate": 3.2158884172225594e-05, + "loss": 1.0399, + "step": 206440 + }, + { + "epoch": 17.87, + "learning_rate": 3.215801784631379e-05, + "loss": 1.0335, + "step": 206450 + }, + { + "epoch": 17.87, + "learning_rate": 3.2157151520401976e-05, + "loss": 1.028, + "step": 206460 + }, + { + "epoch": 17.87, + "learning_rate": 3.215628519449017e-05, + "loss": 1.0031, + "step": 206470 + }, + { + "epoch": 17.87, + "learning_rate": 3.2155418868578365e-05, + "loss": 1.0294, + "step": 206480 + }, + { + "epoch": 17.87, + "learning_rate": 3.215455254266655e-05, + "loss": 1.066, + "step": 206490 + }, + { + "epoch": 17.87, + "learning_rate": 3.215368621675475e-05, + "loss": 0.9911, + "step": 206500 + }, + { + "epoch": 17.87, + "learning_rate": 3.215281989084294e-05, + "loss": 1.0335, + "step": 206510 + }, + { + "epoch": 17.88, + "learning_rate": 3.215195356493113e-05, + "loss": 0.9798, + "step": 206520 + }, + { + "epoch": 17.88, + "learning_rate": 3.2151087239019324e-05, + "loss": 1.0142, + "step": 206530 + }, + { + "epoch": 17.88, + "learning_rate": 3.215022091310752e-05, + "loss": 1.0182, + "step": 206540 + }, + { + "epoch": 17.88, + "learning_rate": 3.2149354587195706e-05, + "loss": 0.9889, + "step": 206550 + }, + { + "epoch": 17.88, + "learning_rate": 3.21484882612839e-05, + "loss": 1.0007, + "step": 206560 + }, + { + "epoch": 17.88, + "learning_rate": 3.214762193537209e-05, + "loss": 1.0135, + "step": 206570 + }, + { + "epoch": 17.88, + "learning_rate": 3.214675560946028e-05, + "loss": 1.0037, + "step": 206580 + }, + { + "epoch": 17.88, + "learning_rate": 3.214588928354848e-05, + "loss": 0.9984, + "step": 206590 + }, + { + "epoch": 17.88, + "learning_rate": 3.2145022957636665e-05, + "loss": 1.0244, + "step": 206600 + }, + { + "epoch": 17.88, + "learning_rate": 3.214415663172486e-05, + "loss": 1.0753, + "step": 206610 + }, + { + "epoch": 17.88, + "learning_rate": 3.2143290305813053e-05, + "loss": 0.996, + "step": 206620 + }, + { + "epoch": 17.88, + "learning_rate": 3.214242397990124e-05, + "loss": 1.0149, + "step": 206630 + }, + { + "epoch": 17.89, + "learning_rate": 3.2141557653989436e-05, + "loss": 1.0562, + "step": 206640 + }, + { + "epoch": 17.89, + "learning_rate": 3.214069132807763e-05, + "loss": 0.9703, + "step": 206650 + }, + { + "epoch": 17.89, + "learning_rate": 3.213982500216582e-05, + "loss": 0.9622, + "step": 206660 + }, + { + "epoch": 17.89, + "learning_rate": 3.213895867625401e-05, + "loss": 1.0158, + "step": 206670 + }, + { + "epoch": 17.89, + "learning_rate": 3.21380923503422e-05, + "loss": 1.0289, + "step": 206680 + }, + { + "epoch": 17.89, + "learning_rate": 3.2137226024430394e-05, + "loss": 1.0517, + "step": 206690 + }, + { + "epoch": 17.89, + "learning_rate": 3.213635969851859e-05, + "loss": 0.9976, + "step": 206700 + }, + { + "epoch": 17.89, + "learning_rate": 3.2135493372606776e-05, + "loss": 1.0398, + "step": 206710 + }, + { + "epoch": 17.89, + "learning_rate": 3.213462704669497e-05, + "loss": 0.9782, + "step": 206720 + }, + { + "epoch": 17.89, + "learning_rate": 3.2133760720783165e-05, + "loss": 1.0831, + "step": 206730 + }, + { + "epoch": 17.89, + "learning_rate": 3.213289439487135e-05, + "loss": 0.9473, + "step": 206740 + }, + { + "epoch": 17.9, + "learning_rate": 3.213202806895955e-05, + "loss": 0.9781, + "step": 206750 + }, + { + "epoch": 17.9, + "learning_rate": 3.2131161743047735e-05, + "loss": 1.0356, + "step": 206760 + }, + { + "epoch": 17.9, + "learning_rate": 3.213029541713593e-05, + "loss": 1.0117, + "step": 206770 + }, + { + "epoch": 17.9, + "learning_rate": 3.2129429091224124e-05, + "loss": 1.0367, + "step": 206780 + }, + { + "epoch": 17.9, + "learning_rate": 3.212856276531231e-05, + "loss": 0.9973, + "step": 206790 + }, + { + "epoch": 17.9, + "learning_rate": 3.2127696439400506e-05, + "loss": 0.9787, + "step": 206800 + }, + { + "epoch": 17.9, + "learning_rate": 3.21268301134887e-05, + "loss": 1.0097, + "step": 206810 + }, + { + "epoch": 17.9, + "learning_rate": 3.212596378757689e-05, + "loss": 1.0121, + "step": 206820 + }, + { + "epoch": 17.9, + "learning_rate": 3.212509746166508e-05, + "loss": 1.0143, + "step": 206830 + }, + { + "epoch": 17.9, + "learning_rate": 3.212423113575328e-05, + "loss": 1.0888, + "step": 206840 + }, + { + "epoch": 17.9, + "learning_rate": 3.2123364809841465e-05, + "loss": 1.033, + "step": 206850 + }, + { + "epoch": 17.9, + "learning_rate": 3.212249848392966e-05, + "loss": 1.0305, + "step": 206860 + }, + { + "epoch": 17.91, + "learning_rate": 3.212163215801785e-05, + "loss": 0.9892, + "step": 206870 + }, + { + "epoch": 17.91, + "learning_rate": 3.212076583210604e-05, + "loss": 0.9899, + "step": 206880 + }, + { + "epoch": 17.91, + "learning_rate": 3.2119899506194236e-05, + "loss": 1.0139, + "step": 206890 + }, + { + "epoch": 17.91, + "learning_rate": 3.2119033180282424e-05, + "loss": 1.0199, + "step": 206900 + }, + { + "epoch": 17.91, + "learning_rate": 3.211816685437062e-05, + "loss": 0.973, + "step": 206910 + }, + { + "epoch": 17.91, + "learning_rate": 3.211730052845881e-05, + "loss": 0.9587, + "step": 206920 + }, + { + "epoch": 17.91, + "learning_rate": 3.2116434202547e-05, + "loss": 0.9944, + "step": 206930 + }, + { + "epoch": 17.91, + "learning_rate": 3.2115567876635195e-05, + "loss": 0.953, + "step": 206940 + }, + { + "epoch": 17.91, + "learning_rate": 3.211470155072339e-05, + "loss": 0.9406, + "step": 206950 + }, + { + "epoch": 17.91, + "learning_rate": 3.211383522481158e-05, + "loss": 1.0151, + "step": 206960 + }, + { + "epoch": 17.91, + "learning_rate": 3.211296889889977e-05, + "loss": 0.9953, + "step": 206970 + }, + { + "epoch": 17.92, + "learning_rate": 3.211210257298796e-05, + "loss": 0.9497, + "step": 206980 + }, + { + "epoch": 17.92, + "learning_rate": 3.211123624707615e-05, + "loss": 0.9835, + "step": 206990 + }, + { + "epoch": 17.92, + "learning_rate": 3.211036992116435e-05, + "loss": 0.9866, + "step": 207000 + }, + { + "epoch": 17.92, + "learning_rate": 3.2109503595252535e-05, + "loss": 1.0032, + "step": 207010 + }, + { + "epoch": 17.92, + "learning_rate": 3.210863726934073e-05, + "loss": 1.0102, + "step": 207020 + }, + { + "epoch": 17.92, + "learning_rate": 3.2107770943428924e-05, + "loss": 0.9784, + "step": 207030 + }, + { + "epoch": 17.92, + "learning_rate": 3.210690461751711e-05, + "loss": 0.9872, + "step": 207040 + }, + { + "epoch": 17.92, + "learning_rate": 3.2106038291605307e-05, + "loss": 0.9972, + "step": 207050 + }, + { + "epoch": 17.92, + "learning_rate": 3.21051719656935e-05, + "loss": 0.9533, + "step": 207060 + }, + { + "epoch": 17.92, + "learning_rate": 3.210430563978169e-05, + "loss": 1.063, + "step": 207070 + }, + { + "epoch": 17.92, + "learning_rate": 3.210343931386988e-05, + "loss": 1.0175, + "step": 207080 + }, + { + "epoch": 17.92, + "learning_rate": 3.210257298795807e-05, + "loss": 0.9815, + "step": 207090 + }, + { + "epoch": 17.93, + "learning_rate": 3.2101706662046265e-05, + "loss": 0.988, + "step": 207100 + }, + { + "epoch": 17.93, + "learning_rate": 3.210084033613446e-05, + "loss": 0.9862, + "step": 207110 + }, + { + "epoch": 17.93, + "learning_rate": 3.209997401022265e-05, + "loss": 1.0773, + "step": 207120 + }, + { + "epoch": 17.93, + "learning_rate": 3.209910768431084e-05, + "loss": 0.9788, + "step": 207130 + }, + { + "epoch": 17.93, + "learning_rate": 3.2098241358399036e-05, + "loss": 1.0632, + "step": 207140 + }, + { + "epoch": 17.93, + "learning_rate": 3.2097375032487224e-05, + "loss": 0.9913, + "step": 207150 + }, + { + "epoch": 17.93, + "learning_rate": 3.209650870657542e-05, + "loss": 1.0116, + "step": 207160 + }, + { + "epoch": 17.93, + "learning_rate": 3.209564238066361e-05, + "loss": 1.0567, + "step": 207170 + }, + { + "epoch": 17.93, + "learning_rate": 3.20947760547518e-05, + "loss": 1.0152, + "step": 207180 + }, + { + "epoch": 17.93, + "learning_rate": 3.2093909728839995e-05, + "loss": 0.9772, + "step": 207190 + }, + { + "epoch": 17.93, + "learning_rate": 3.209304340292818e-05, + "loss": 0.9544, + "step": 207200 + }, + { + "epoch": 17.94, + "learning_rate": 3.209217707701638e-05, + "loss": 1.029, + "step": 207210 + }, + { + "epoch": 17.94, + "learning_rate": 3.209131075110457e-05, + "loss": 1.0149, + "step": 207220 + }, + { + "epoch": 17.94, + "learning_rate": 3.209044442519276e-05, + "loss": 1.0503, + "step": 207230 + }, + { + "epoch": 17.94, + "learning_rate": 3.2089578099280954e-05, + "loss": 1.0037, + "step": 207240 + }, + { + "epoch": 17.94, + "learning_rate": 3.208871177336915e-05, + "loss": 0.9408, + "step": 207250 + }, + { + "epoch": 17.94, + "learning_rate": 3.2087845447457336e-05, + "loss": 0.9686, + "step": 207260 + }, + { + "epoch": 17.94, + "learning_rate": 3.208697912154553e-05, + "loss": 1.0085, + "step": 207270 + }, + { + "epoch": 17.94, + "learning_rate": 3.2086112795633725e-05, + "loss": 0.9654, + "step": 207280 + }, + { + "epoch": 17.94, + "learning_rate": 3.208524646972191e-05, + "loss": 0.9632, + "step": 207290 + }, + { + "epoch": 17.94, + "learning_rate": 3.208438014381011e-05, + "loss": 1.0144, + "step": 207300 + }, + { + "epoch": 17.94, + "learning_rate": 3.2083513817898295e-05, + "loss": 0.9717, + "step": 207310 + }, + { + "epoch": 17.94, + "learning_rate": 3.208264749198649e-05, + "loss": 1.0004, + "step": 207320 + }, + { + "epoch": 17.95, + "learning_rate": 3.2081781166074683e-05, + "loss": 0.986, + "step": 207330 + }, + { + "epoch": 17.95, + "learning_rate": 3.208091484016287e-05, + "loss": 1.1004, + "step": 207340 + }, + { + "epoch": 17.95, + "learning_rate": 3.2080048514251066e-05, + "loss": 0.9532, + "step": 207350 + }, + { + "epoch": 17.95, + "learning_rate": 3.207918218833926e-05, + "loss": 0.9832, + "step": 207360 + }, + { + "epoch": 17.95, + "learning_rate": 3.207831586242745e-05, + "loss": 1.0048, + "step": 207370 + }, + { + "epoch": 17.95, + "learning_rate": 3.207744953651564e-05, + "loss": 1.045, + "step": 207380 + }, + { + "epoch": 17.95, + "learning_rate": 3.2076583210603837e-05, + "loss": 0.9828, + "step": 207390 + }, + { + "epoch": 17.95, + "learning_rate": 3.2075716884692024e-05, + "loss": 0.9811, + "step": 207400 + }, + { + "epoch": 17.95, + "learning_rate": 3.207485055878022e-05, + "loss": 0.9983, + "step": 207410 + }, + { + "epoch": 17.95, + "learning_rate": 3.2073984232868406e-05, + "loss": 0.9964, + "step": 207420 + }, + { + "epoch": 17.95, + "learning_rate": 3.20731179069566e-05, + "loss": 0.9982, + "step": 207430 + }, + { + "epoch": 17.95, + "learning_rate": 3.2072251581044795e-05, + "loss": 0.9901, + "step": 207440 + }, + { + "epoch": 17.96, + "learning_rate": 3.207138525513298e-05, + "loss": 1.0494, + "step": 207450 + }, + { + "epoch": 17.96, + "learning_rate": 3.207051892922118e-05, + "loss": 0.9836, + "step": 207460 + }, + { + "epoch": 17.96, + "learning_rate": 3.206965260330937e-05, + "loss": 0.9634, + "step": 207470 + }, + { + "epoch": 17.96, + "learning_rate": 3.206878627739756e-05, + "loss": 1.01, + "step": 207480 + }, + { + "epoch": 17.96, + "learning_rate": 3.2067919951485754e-05, + "loss": 0.9892, + "step": 207490 + }, + { + "epoch": 17.96, + "learning_rate": 3.206705362557395e-05, + "loss": 1.0213, + "step": 207500 + }, + { + "epoch": 17.96, + "learning_rate": 3.2066187299662136e-05, + "loss": 1.0044, + "step": 207510 + }, + { + "epoch": 17.96, + "learning_rate": 3.206532097375033e-05, + "loss": 0.9993, + "step": 207520 + }, + { + "epoch": 17.96, + "learning_rate": 3.206445464783852e-05, + "loss": 1.0143, + "step": 207530 + }, + { + "epoch": 17.96, + "learning_rate": 3.206358832192671e-05, + "loss": 1.0113, + "step": 207540 + }, + { + "epoch": 17.96, + "learning_rate": 3.206272199601491e-05, + "loss": 0.964, + "step": 207550 + }, + { + "epoch": 17.97, + "learning_rate": 3.2061855670103095e-05, + "loss": 1.0223, + "step": 207560 + }, + { + "epoch": 17.97, + "learning_rate": 3.206098934419129e-05, + "loss": 1.035, + "step": 207570 + }, + { + "epoch": 17.97, + "learning_rate": 3.2060123018279484e-05, + "loss": 0.992, + "step": 207580 + }, + { + "epoch": 17.97, + "learning_rate": 3.205925669236767e-05, + "loss": 0.944, + "step": 207590 + }, + { + "epoch": 17.97, + "learning_rate": 3.2058390366455866e-05, + "loss": 0.9975, + "step": 207600 + }, + { + "epoch": 17.97, + "learning_rate": 3.2057524040544054e-05, + "loss": 0.9907, + "step": 207610 + }, + { + "epoch": 17.97, + "learning_rate": 3.205665771463225e-05, + "loss": 1.0079, + "step": 207620 + }, + { + "epoch": 17.97, + "learning_rate": 3.205579138872044e-05, + "loss": 0.9689, + "step": 207630 + }, + { + "epoch": 17.97, + "learning_rate": 3.205492506280863e-05, + "loss": 1.0173, + "step": 207640 + }, + { + "epoch": 17.97, + "learning_rate": 3.2054058736896825e-05, + "loss": 1.0474, + "step": 207650 + }, + { + "epoch": 17.97, + "learning_rate": 3.205319241098502e-05, + "loss": 1.0347, + "step": 207660 + }, + { + "epoch": 17.97, + "learning_rate": 3.205232608507321e-05, + "loss": 0.9647, + "step": 207670 + }, + { + "epoch": 17.98, + "learning_rate": 3.20514597591614e-05, + "loss": 0.9756, + "step": 207680 + }, + { + "epoch": 17.98, + "learning_rate": 3.2050593433249596e-05, + "loss": 1.0348, + "step": 207690 + }, + { + "epoch": 17.98, + "learning_rate": 3.204972710733778e-05, + "loss": 1.0331, + "step": 207700 + }, + { + "epoch": 17.98, + "learning_rate": 3.204886078142598e-05, + "loss": 1.0058, + "step": 207710 + }, + { + "epoch": 17.98, + "learning_rate": 3.2047994455514165e-05, + "loss": 1.0307, + "step": 207720 + }, + { + "epoch": 17.98, + "learning_rate": 3.204712812960236e-05, + "loss": 1.0086, + "step": 207730 + }, + { + "epoch": 17.98, + "learning_rate": 3.2046261803690554e-05, + "loss": 1.0073, + "step": 207740 + }, + { + "epoch": 17.98, + "learning_rate": 3.204539547777874e-05, + "loss": 0.9988, + "step": 207750 + }, + { + "epoch": 17.98, + "learning_rate": 3.2044529151866936e-05, + "loss": 0.9406, + "step": 207760 + }, + { + "epoch": 17.98, + "learning_rate": 3.204366282595513e-05, + "loss": 0.9497, + "step": 207770 + }, + { + "epoch": 17.98, + "learning_rate": 3.204279650004332e-05, + "loss": 1.0188, + "step": 207780 + }, + { + "epoch": 17.99, + "learning_rate": 3.204193017413151e-05, + "loss": 1.023, + "step": 207790 + }, + { + "epoch": 17.99, + "learning_rate": 3.204106384821971e-05, + "loss": 1.0017, + "step": 207800 + }, + { + "epoch": 17.99, + "learning_rate": 3.2040197522307895e-05, + "loss": 0.9716, + "step": 207810 + }, + { + "epoch": 17.99, + "learning_rate": 3.203933119639609e-05, + "loss": 1.0431, + "step": 207820 + }, + { + "epoch": 17.99, + "learning_rate": 3.203846487048428e-05, + "loss": 1.004, + "step": 207830 + }, + { + "epoch": 17.99, + "learning_rate": 3.203759854457247e-05, + "loss": 0.9971, + "step": 207840 + }, + { + "epoch": 17.99, + "learning_rate": 3.2036732218660666e-05, + "loss": 0.9669, + "step": 207850 + }, + { + "epoch": 17.99, + "learning_rate": 3.2035865892748854e-05, + "loss": 0.9725, + "step": 207860 + }, + { + "epoch": 17.99, + "learning_rate": 3.203499956683705e-05, + "loss": 0.9621, + "step": 207870 + }, + { + "epoch": 17.99, + "learning_rate": 3.203413324092524e-05, + "loss": 0.975, + "step": 207880 + }, + { + "epoch": 17.99, + "learning_rate": 3.203326691501343e-05, + "loss": 1.0053, + "step": 207890 + }, + { + "epoch": 17.99, + "learning_rate": 3.2032400589101625e-05, + "loss": 1.0027, + "step": 207900 + }, + { + "epoch": 18.0, + "learning_rate": 3.203153426318982e-05, + "loss": 1.0206, + "step": 207910 + }, + { + "epoch": 18.0, + "learning_rate": 3.2030667937278e-05, + "loss": 0.9993, + "step": 207920 + }, + { + "epoch": 18.0, + "learning_rate": 3.2029801611366195e-05, + "loss": 1.0266, + "step": 207930 + }, + { + "epoch": 18.0, + "learning_rate": 3.202893528545439e-05, + "loss": 0.9956, + "step": 207940 + }, + { + "epoch": 18.0, + "learning_rate": 3.202806895954258e-05, + "loss": 1.0584, + "step": 207950 + }, + { + "epoch": 18.0, + "learning_rate": 3.202720263363077e-05, + "loss": 1.0051, + "step": 207960 + }, + { + "epoch": 18.0, + "eval_Bleu_1": 0.046066491112573306, + "eval_Bleu_2": 2.877753898350885e-11, + "eval_Bleu_3": 2.539608345254435e-14, + "eval_Bleu_4": 7.7447145599517325e-16, + "eval_ROUGE_L": 0.09344267465094487, + "eval_cer": 0.9933173240540174, + "eval_em": 0, + "eval_f1": 0.11158915274774159, + "eval_loss": 0.981272280216217, + "eval_runtime": 2432.9985, + "eval_samples_per_second": 2.111, + "eval_steps_per_second": 2.111, + "eval_wer": 0.9688412082102563, + "step": 207960 + }, + { + "epoch": 18.0, + "learning_rate": 3.2026336307718966e-05, + "loss": 1.0212, + "step": 207970 + }, + { + "epoch": 18.0, + "learning_rate": 3.2025469981807153e-05, + "loss": 0.967, + "step": 207980 + }, + { + "epoch": 18.0, + "learning_rate": 3.202460365589535e-05, + "loss": 0.9704, + "step": 207990 + }, + { + "epoch": 18.0, + "learning_rate": 3.2023737329983536e-05, + "loss": 1.0134, + "step": 208000 + }, + { + "epoch": 18.0, + "learning_rate": 3.202287100407173e-05, + "loss": 0.9343, + "step": 208010 + }, + { + "epoch": 18.01, + "learning_rate": 3.2022004678159925e-05, + "loss": 1.0012, + "step": 208020 + }, + { + "epoch": 18.01, + "learning_rate": 3.202113835224811e-05, + "loss": 0.9817, + "step": 208030 + }, + { + "epoch": 18.01, + "learning_rate": 3.202027202633631e-05, + "loss": 0.9754, + "step": 208040 + }, + { + "epoch": 18.01, + "learning_rate": 3.20194057004245e-05, + "loss": 1.0114, + "step": 208050 + }, + { + "epoch": 18.01, + "learning_rate": 3.201853937451269e-05, + "loss": 0.9913, + "step": 208060 + }, + { + "epoch": 18.01, + "learning_rate": 3.201767304860088e-05, + "loss": 0.9971, + "step": 208070 + }, + { + "epoch": 18.01, + "learning_rate": 3.201680672268908e-05, + "loss": 0.9708, + "step": 208080 + }, + { + "epoch": 18.01, + "learning_rate": 3.2015940396777265e-05, + "loss": 0.9505, + "step": 208090 + }, + { + "epoch": 18.01, + "learning_rate": 3.201507407086546e-05, + "loss": 0.9636, + "step": 208100 + }, + { + "epoch": 18.01, + "learning_rate": 3.201420774495365e-05, + "loss": 0.9717, + "step": 208110 + }, + { + "epoch": 18.01, + "learning_rate": 3.201334141904184e-05, + "loss": 0.9609, + "step": 208120 + }, + { + "epoch": 18.01, + "learning_rate": 3.2012475093130036e-05, + "loss": 0.9524, + "step": 208130 + }, + { + "epoch": 18.02, + "learning_rate": 3.2011608767218224e-05, + "loss": 0.9801, + "step": 208140 + }, + { + "epoch": 18.02, + "learning_rate": 3.201074244130642e-05, + "loss": 0.9116, + "step": 208150 + }, + { + "epoch": 18.02, + "learning_rate": 3.200987611539461e-05, + "loss": 0.9486, + "step": 208160 + }, + { + "epoch": 18.02, + "learning_rate": 3.20090097894828e-05, + "loss": 0.9334, + "step": 208170 + }, + { + "epoch": 18.02, + "learning_rate": 3.2008143463570995e-05, + "loss": 1.0095, + "step": 208180 + }, + { + "epoch": 18.02, + "learning_rate": 3.200727713765919e-05, + "loss": 1.0266, + "step": 208190 + }, + { + "epoch": 18.02, + "learning_rate": 3.200641081174738e-05, + "loss": 0.983, + "step": 208200 + }, + { + "epoch": 18.02, + "learning_rate": 3.200554448583557e-05, + "loss": 1.0277, + "step": 208210 + }, + { + "epoch": 18.02, + "learning_rate": 3.200467815992376e-05, + "loss": 0.9997, + "step": 208220 + }, + { + "epoch": 18.02, + "learning_rate": 3.2003811834011954e-05, + "loss": 1.0095, + "step": 208230 + }, + { + "epoch": 18.02, + "learning_rate": 3.200294550810015e-05, + "loss": 0.9895, + "step": 208240 + }, + { + "epoch": 18.03, + "learning_rate": 3.2002079182188336e-05, + "loss": 1.0161, + "step": 208250 + }, + { + "epoch": 18.03, + "learning_rate": 3.200121285627653e-05, + "loss": 1.0471, + "step": 208260 + }, + { + "epoch": 18.03, + "learning_rate": 3.2000346530364725e-05, + "loss": 1.0571, + "step": 208270 + }, + { + "epoch": 18.03, + "learning_rate": 3.199948020445291e-05, + "loss": 0.9937, + "step": 208280 + }, + { + "epoch": 18.03, + "learning_rate": 3.199861387854111e-05, + "loss": 1.0212, + "step": 208290 + }, + { + "epoch": 18.03, + "learning_rate": 3.1997747552629295e-05, + "loss": 0.9808, + "step": 208300 + }, + { + "epoch": 18.03, + "learning_rate": 3.199688122671749e-05, + "loss": 1.0173, + "step": 208310 + }, + { + "epoch": 18.03, + "learning_rate": 3.1996014900805684e-05, + "loss": 1.0207, + "step": 208320 + }, + { + "epoch": 18.03, + "learning_rate": 3.199514857489387e-05, + "loss": 1.0584, + "step": 208330 + }, + { + "epoch": 18.03, + "learning_rate": 3.1994282248982066e-05, + "loss": 1.0183, + "step": 208340 + }, + { + "epoch": 18.03, + "learning_rate": 3.199341592307026e-05, + "loss": 0.963, + "step": 208350 + }, + { + "epoch": 18.03, + "learning_rate": 3.199254959715845e-05, + "loss": 0.929, + "step": 208360 + }, + { + "epoch": 18.04, + "learning_rate": 3.199168327124664e-05, + "loss": 0.9572, + "step": 208370 + }, + { + "epoch": 18.04, + "learning_rate": 3.199081694533484e-05, + "loss": 0.9635, + "step": 208380 + }, + { + "epoch": 18.04, + "learning_rate": 3.1989950619423024e-05, + "loss": 0.9867, + "step": 208390 + }, + { + "epoch": 18.04, + "learning_rate": 3.198908429351122e-05, + "loss": 1.0163, + "step": 208400 + }, + { + "epoch": 18.04, + "learning_rate": 3.1988217967599407e-05, + "loss": 0.9356, + "step": 208410 + }, + { + "epoch": 18.04, + "learning_rate": 3.19873516416876e-05, + "loss": 0.9684, + "step": 208420 + }, + { + "epoch": 18.04, + "learning_rate": 3.1986485315775795e-05, + "loss": 0.9783, + "step": 208430 + }, + { + "epoch": 18.04, + "learning_rate": 3.198561898986398e-05, + "loss": 1.0075, + "step": 208440 + }, + { + "epoch": 18.04, + "learning_rate": 3.198475266395218e-05, + "loss": 0.9592, + "step": 208450 + }, + { + "epoch": 18.04, + "learning_rate": 3.198388633804037e-05, + "loss": 0.9574, + "step": 208460 + }, + { + "epoch": 18.04, + "learning_rate": 3.198302001212856e-05, + "loss": 1.0015, + "step": 208470 + }, + { + "epoch": 18.05, + "learning_rate": 3.1982153686216754e-05, + "loss": 0.985, + "step": 208480 + }, + { + "epoch": 18.05, + "learning_rate": 3.198128736030495e-05, + "loss": 1.0155, + "step": 208490 + }, + { + "epoch": 18.05, + "learning_rate": 3.1980421034393136e-05, + "loss": 0.989, + "step": 208500 + }, + { + "epoch": 18.05, + "learning_rate": 3.197955470848133e-05, + "loss": 0.9376, + "step": 208510 + }, + { + "epoch": 18.05, + "learning_rate": 3.197868838256952e-05, + "loss": 0.9836, + "step": 208520 + }, + { + "epoch": 18.05, + "learning_rate": 3.197782205665771e-05, + "loss": 1.0322, + "step": 208530 + }, + { + "epoch": 18.05, + "learning_rate": 3.197695573074591e-05, + "loss": 1.0072, + "step": 208540 + }, + { + "epoch": 18.05, + "learning_rate": 3.1976089404834095e-05, + "loss": 0.9645, + "step": 208550 + }, + { + "epoch": 18.05, + "learning_rate": 3.197522307892229e-05, + "loss": 1.012, + "step": 208560 + }, + { + "epoch": 18.05, + "learning_rate": 3.1974356753010484e-05, + "loss": 1.0383, + "step": 208570 + }, + { + "epoch": 18.05, + "learning_rate": 3.197349042709867e-05, + "loss": 0.9955, + "step": 208580 + }, + { + "epoch": 18.05, + "learning_rate": 3.1972624101186866e-05, + "loss": 1.0426, + "step": 208590 + }, + { + "epoch": 18.06, + "learning_rate": 3.197175777527506e-05, + "loss": 0.9613, + "step": 208600 + }, + { + "epoch": 18.06, + "learning_rate": 3.197089144936325e-05, + "loss": 0.9854, + "step": 208610 + }, + { + "epoch": 18.06, + "learning_rate": 3.197002512345144e-05, + "loss": 0.9852, + "step": 208620 + }, + { + "epoch": 18.06, + "learning_rate": 3.196915879753963e-05, + "loss": 1.0048, + "step": 208630 + }, + { + "epoch": 18.06, + "learning_rate": 3.1968292471627825e-05, + "loss": 0.9585, + "step": 208640 + }, + { + "epoch": 18.06, + "learning_rate": 3.196742614571602e-05, + "loss": 0.9624, + "step": 208650 + }, + { + "epoch": 18.06, + "learning_rate": 3.196655981980421e-05, + "loss": 0.9686, + "step": 208660 + }, + { + "epoch": 18.06, + "learning_rate": 3.19656934938924e-05, + "loss": 0.9212, + "step": 208670 + }, + { + "epoch": 18.06, + "learning_rate": 3.1964827167980596e-05, + "loss": 0.9597, + "step": 208680 + }, + { + "epoch": 18.06, + "learning_rate": 3.1963960842068783e-05, + "loss": 0.9988, + "step": 208690 + }, + { + "epoch": 18.06, + "learning_rate": 3.196309451615698e-05, + "loss": 0.968, + "step": 208700 + }, + { + "epoch": 18.06, + "learning_rate": 3.196222819024517e-05, + "loss": 0.9501, + "step": 208710 + }, + { + "epoch": 18.07, + "learning_rate": 3.196136186433336e-05, + "loss": 0.9753, + "step": 208720 + }, + { + "epoch": 18.07, + "learning_rate": 3.1960495538421554e-05, + "loss": 1.0123, + "step": 208730 + }, + { + "epoch": 18.07, + "learning_rate": 3.195962921250974e-05, + "loss": 0.9438, + "step": 208740 + }, + { + "epoch": 18.07, + "learning_rate": 3.1958762886597937e-05, + "loss": 1.0006, + "step": 208750 + }, + { + "epoch": 18.07, + "learning_rate": 3.195789656068613e-05, + "loss": 0.9632, + "step": 208760 + }, + { + "epoch": 18.07, + "learning_rate": 3.195703023477432e-05, + "loss": 0.9609, + "step": 208770 + }, + { + "epoch": 18.07, + "learning_rate": 3.195616390886251e-05, + "loss": 0.9994, + "step": 208780 + }, + { + "epoch": 18.07, + "learning_rate": 3.195529758295071e-05, + "loss": 0.9835, + "step": 208790 + }, + { + "epoch": 18.07, + "learning_rate": 3.1954431257038895e-05, + "loss": 0.9904, + "step": 208800 + }, + { + "epoch": 18.07, + "learning_rate": 3.195356493112709e-05, + "loss": 1.0007, + "step": 208810 + }, + { + "epoch": 18.07, + "learning_rate": 3.1952698605215284e-05, + "loss": 0.9592, + "step": 208820 + }, + { + "epoch": 18.08, + "learning_rate": 3.195183227930347e-05, + "loss": 0.9901, + "step": 208830 + }, + { + "epoch": 18.08, + "learning_rate": 3.1950965953391666e-05, + "loss": 0.9333, + "step": 208840 + }, + { + "epoch": 18.08, + "learning_rate": 3.1950099627479854e-05, + "loss": 0.9943, + "step": 208850 + }, + { + "epoch": 18.08, + "learning_rate": 3.194923330156805e-05, + "loss": 0.9825, + "step": 208860 + }, + { + "epoch": 18.08, + "learning_rate": 3.194836697565624e-05, + "loss": 0.9249, + "step": 208870 + }, + { + "epoch": 18.08, + "learning_rate": 3.194750064974443e-05, + "loss": 0.9441, + "step": 208880 + }, + { + "epoch": 18.08, + "learning_rate": 3.1946634323832625e-05, + "loss": 0.9512, + "step": 208890 + }, + { + "epoch": 18.08, + "learning_rate": 3.194576799792082e-05, + "loss": 0.9194, + "step": 208900 + }, + { + "epoch": 18.08, + "learning_rate": 3.194490167200901e-05, + "loss": 0.9843, + "step": 208910 + }, + { + "epoch": 18.08, + "learning_rate": 3.19440353460972e-05, + "loss": 0.966, + "step": 208920 + }, + { + "epoch": 18.08, + "learning_rate": 3.1943169020185396e-05, + "loss": 0.9655, + "step": 208930 + }, + { + "epoch": 18.08, + "learning_rate": 3.1942302694273584e-05, + "loss": 1.0168, + "step": 208940 + }, + { + "epoch": 18.09, + "learning_rate": 3.194143636836178e-05, + "loss": 0.9828, + "step": 208950 + }, + { + "epoch": 18.09, + "learning_rate": 3.1940570042449966e-05, + "loss": 1.0018, + "step": 208960 + }, + { + "epoch": 18.09, + "learning_rate": 3.193970371653816e-05, + "loss": 0.9838, + "step": 208970 + }, + { + "epoch": 18.09, + "learning_rate": 3.1938837390626355e-05, + "loss": 0.9134, + "step": 208980 + }, + { + "epoch": 18.09, + "learning_rate": 3.193797106471454e-05, + "loss": 0.9976, + "step": 208990 + }, + { + "epoch": 18.09, + "learning_rate": 3.193710473880274e-05, + "loss": 0.9382, + "step": 209000 + }, + { + "epoch": 18.09, + "learning_rate": 3.193623841289093e-05, + "loss": 0.9834, + "step": 209010 + }, + { + "epoch": 18.09, + "learning_rate": 3.193537208697912e-05, + "loss": 1.0171, + "step": 209020 + }, + { + "epoch": 18.09, + "learning_rate": 3.1934505761067314e-05, + "loss": 0.9481, + "step": 209030 + }, + { + "epoch": 18.09, + "learning_rate": 3.193363943515551e-05, + "loss": 0.9453, + "step": 209040 + }, + { + "epoch": 18.09, + "learning_rate": 3.1932773109243696e-05, + "loss": 0.9654, + "step": 209050 + }, + { + "epoch": 18.1, + "learning_rate": 3.193190678333189e-05, + "loss": 0.9746, + "step": 209060 + }, + { + "epoch": 18.1, + "learning_rate": 3.193104045742008e-05, + "loss": 0.9588, + "step": 209070 + }, + { + "epoch": 18.1, + "learning_rate": 3.193017413150827e-05, + "loss": 0.9252, + "step": 209080 + }, + { + "epoch": 18.1, + "learning_rate": 3.192930780559647e-05, + "loss": 0.9758, + "step": 209090 + }, + { + "epoch": 18.1, + "learning_rate": 3.1928441479684654e-05, + "loss": 1.0258, + "step": 209100 + }, + { + "epoch": 18.1, + "learning_rate": 3.192757515377285e-05, + "loss": 1.0115, + "step": 209110 + }, + { + "epoch": 18.1, + "learning_rate": 3.192670882786104e-05, + "loss": 1.0134, + "step": 209120 + }, + { + "epoch": 18.1, + "learning_rate": 3.192584250194923e-05, + "loss": 0.9859, + "step": 209130 + }, + { + "epoch": 18.1, + "learning_rate": 3.1924976176037425e-05, + "loss": 0.9965, + "step": 209140 + }, + { + "epoch": 18.1, + "learning_rate": 3.192410985012561e-05, + "loss": 0.9943, + "step": 209150 + }, + { + "epoch": 18.1, + "learning_rate": 3.192324352421381e-05, + "loss": 1.0023, + "step": 209160 + }, + { + "epoch": 18.1, + "learning_rate": 3.1922377198302e-05, + "loss": 0.9499, + "step": 209170 + }, + { + "epoch": 18.11, + "learning_rate": 3.192151087239019e-05, + "loss": 0.9504, + "step": 209180 + }, + { + "epoch": 18.11, + "learning_rate": 3.1920644546478384e-05, + "loss": 0.9723, + "step": 209190 + }, + { + "epoch": 18.11, + "learning_rate": 3.191977822056658e-05, + "loss": 0.9497, + "step": 209200 + }, + { + "epoch": 18.11, + "learning_rate": 3.1918911894654766e-05, + "loss": 0.9365, + "step": 209210 + }, + { + "epoch": 18.11, + "learning_rate": 3.191804556874296e-05, + "loss": 0.9711, + "step": 209220 + }, + { + "epoch": 18.11, + "learning_rate": 3.1917179242831155e-05, + "loss": 0.9506, + "step": 209230 + }, + { + "epoch": 18.11, + "learning_rate": 3.191631291691934e-05, + "loss": 0.9783, + "step": 209240 + }, + { + "epoch": 18.11, + "learning_rate": 3.191544659100754e-05, + "loss": 1.0038, + "step": 209250 + }, + { + "epoch": 18.11, + "learning_rate": 3.1914580265095725e-05, + "loss": 1.0486, + "step": 209260 + }, + { + "epoch": 18.11, + "learning_rate": 3.191371393918392e-05, + "loss": 0.9782, + "step": 209270 + }, + { + "epoch": 18.11, + "learning_rate": 3.1912847613272114e-05, + "loss": 1.0455, + "step": 209280 + }, + { + "epoch": 18.12, + "learning_rate": 3.19119812873603e-05, + "loss": 0.9892, + "step": 209290 + }, + { + "epoch": 18.12, + "learning_rate": 3.1911114961448496e-05, + "loss": 0.9301, + "step": 209300 + }, + { + "epoch": 18.12, + "learning_rate": 3.191024863553669e-05, + "loss": 1.0035, + "step": 209310 + }, + { + "epoch": 18.12, + "learning_rate": 3.190938230962488e-05, + "loss": 0.9414, + "step": 209320 + }, + { + "epoch": 18.12, + "learning_rate": 3.190851598371307e-05, + "loss": 1.0078, + "step": 209330 + }, + { + "epoch": 18.12, + "learning_rate": 3.190764965780127e-05, + "loss": 1.0187, + "step": 209340 + }, + { + "epoch": 18.12, + "learning_rate": 3.1906783331889455e-05, + "loss": 1.0207, + "step": 209350 + }, + { + "epoch": 18.12, + "learning_rate": 3.190591700597765e-05, + "loss": 1.0284, + "step": 209360 + }, + { + "epoch": 18.12, + "learning_rate": 3.190505068006584e-05, + "loss": 1.0009, + "step": 209370 + }, + { + "epoch": 18.12, + "learning_rate": 3.190418435415403e-05, + "loss": 0.9636, + "step": 209380 + }, + { + "epoch": 18.12, + "learning_rate": 3.1903318028242226e-05, + "loss": 1.0088, + "step": 209390 + }, + { + "epoch": 18.12, + "learning_rate": 3.1902451702330413e-05, + "loss": 1.0162, + "step": 209400 + }, + { + "epoch": 18.13, + "learning_rate": 3.190158537641861e-05, + "loss": 0.8997, + "step": 209410 + }, + { + "epoch": 18.13, + "learning_rate": 3.19007190505068e-05, + "loss": 0.9819, + "step": 209420 + }, + { + "epoch": 18.13, + "learning_rate": 3.189985272459499e-05, + "loss": 0.9508, + "step": 209430 + }, + { + "epoch": 18.13, + "learning_rate": 3.1898986398683184e-05, + "loss": 1.0009, + "step": 209440 + }, + { + "epoch": 18.13, + "learning_rate": 3.189812007277138e-05, + "loss": 0.9584, + "step": 209450 + }, + { + "epoch": 18.13, + "learning_rate": 3.1897253746859567e-05, + "loss": 0.9499, + "step": 209460 + }, + { + "epoch": 18.13, + "learning_rate": 3.189638742094776e-05, + "loss": 1.0052, + "step": 209470 + }, + { + "epoch": 18.13, + "learning_rate": 3.189552109503595e-05, + "loss": 0.9901, + "step": 209480 + }, + { + "epoch": 18.13, + "learning_rate": 3.189465476912414e-05, + "loss": 0.9392, + "step": 209490 + }, + { + "epoch": 18.13, + "learning_rate": 3.189378844321234e-05, + "loss": 1.0166, + "step": 209500 + }, + { + "epoch": 18.13, + "learning_rate": 3.1892922117300525e-05, + "loss": 0.9654, + "step": 209510 + }, + { + "epoch": 18.14, + "learning_rate": 3.189205579138872e-05, + "loss": 0.9063, + "step": 209520 + }, + { + "epoch": 18.14, + "learning_rate": 3.1891189465476914e-05, + "loss": 0.9468, + "step": 209530 + }, + { + "epoch": 18.14, + "learning_rate": 3.18903231395651e-05, + "loss": 1.0063, + "step": 209540 + }, + { + "epoch": 18.14, + "learning_rate": 3.1889456813653296e-05, + "loss": 0.927, + "step": 209550 + }, + { + "epoch": 18.14, + "learning_rate": 3.188859048774149e-05, + "loss": 1.0384, + "step": 209560 + }, + { + "epoch": 18.14, + "learning_rate": 3.188772416182968e-05, + "loss": 0.9884, + "step": 209570 + }, + { + "epoch": 18.14, + "learning_rate": 3.188685783591787e-05, + "loss": 0.9421, + "step": 209580 + }, + { + "epoch": 18.14, + "learning_rate": 3.188599151000606e-05, + "loss": 1.0144, + "step": 209590 + }, + { + "epoch": 18.14, + "learning_rate": 3.1885125184094255e-05, + "loss": 0.9988, + "step": 209600 + }, + { + "epoch": 18.14, + "learning_rate": 3.188425885818245e-05, + "loss": 0.9756, + "step": 209610 + }, + { + "epoch": 18.14, + "learning_rate": 3.188339253227064e-05, + "loss": 0.9488, + "step": 209620 + }, + { + "epoch": 18.14, + "learning_rate": 3.188252620635883e-05, + "loss": 1.0235, + "step": 209630 + }, + { + "epoch": 18.15, + "learning_rate": 3.1881659880447026e-05, + "loss": 0.9673, + "step": 209640 + }, + { + "epoch": 18.15, + "learning_rate": 3.1880793554535214e-05, + "loss": 0.9015, + "step": 209650 + }, + { + "epoch": 18.15, + "learning_rate": 3.187992722862341e-05, + "loss": 0.9802, + "step": 209660 + }, + { + "epoch": 18.15, + "learning_rate": 3.18790609027116e-05, + "loss": 1.0033, + "step": 209670 + }, + { + "epoch": 18.15, + "learning_rate": 3.187819457679979e-05, + "loss": 0.9827, + "step": 209680 + }, + { + "epoch": 18.15, + "learning_rate": 3.1877328250887985e-05, + "loss": 0.9645, + "step": 209690 + }, + { + "epoch": 18.15, + "learning_rate": 3.187646192497617e-05, + "loss": 0.9556, + "step": 209700 + }, + { + "epoch": 18.15, + "learning_rate": 3.187559559906437e-05, + "loss": 0.9789, + "step": 209710 + }, + { + "epoch": 18.15, + "learning_rate": 3.187472927315256e-05, + "loss": 0.9372, + "step": 209720 + }, + { + "epoch": 18.15, + "learning_rate": 3.187386294724075e-05, + "loss": 0.959, + "step": 209730 + }, + { + "epoch": 18.15, + "learning_rate": 3.1872996621328943e-05, + "loss": 1.0368, + "step": 209740 + }, + { + "epoch": 18.15, + "learning_rate": 3.187213029541714e-05, + "loss": 0.8947, + "step": 209750 + }, + { + "epoch": 18.16, + "learning_rate": 3.1871263969505326e-05, + "loss": 1.0146, + "step": 209760 + }, + { + "epoch": 18.16, + "learning_rate": 3.187039764359352e-05, + "loss": 0.9579, + "step": 209770 + }, + { + "epoch": 18.16, + "learning_rate": 3.1869531317681715e-05, + "loss": 0.9638, + "step": 209780 + }, + { + "epoch": 18.16, + "learning_rate": 3.18686649917699e-05, + "loss": 1.0247, + "step": 209790 + }, + { + "epoch": 18.16, + "learning_rate": 3.18677986658581e-05, + "loss": 1.0292, + "step": 209800 + }, + { + "epoch": 18.16, + "learning_rate": 3.1866932339946284e-05, + "loss": 0.9444, + "step": 209810 + }, + { + "epoch": 18.16, + "learning_rate": 3.186606601403448e-05, + "loss": 0.9389, + "step": 209820 + }, + { + "epoch": 18.16, + "learning_rate": 3.186519968812267e-05, + "loss": 0.9973, + "step": 209830 + }, + { + "epoch": 18.16, + "learning_rate": 3.186433336221086e-05, + "loss": 0.9877, + "step": 209840 + }, + { + "epoch": 18.16, + "learning_rate": 3.1863467036299055e-05, + "loss": 1.0315, + "step": 209850 + }, + { + "epoch": 18.16, + "learning_rate": 3.186260071038725e-05, + "loss": 0.9928, + "step": 209860 + }, + { + "epoch": 18.17, + "learning_rate": 3.186173438447544e-05, + "loss": 0.9472, + "step": 209870 + }, + { + "epoch": 18.17, + "learning_rate": 3.186086805856363e-05, + "loss": 1.0192, + "step": 209880 + }, + { + "epoch": 18.17, + "learning_rate": 3.186000173265182e-05, + "loss": 1.0336, + "step": 209890 + }, + { + "epoch": 18.17, + "learning_rate": 3.1859135406740014e-05, + "loss": 0.9688, + "step": 209900 + }, + { + "epoch": 18.17, + "learning_rate": 3.185826908082821e-05, + "loss": 0.9956, + "step": 209910 + }, + { + "epoch": 18.17, + "learning_rate": 3.1857402754916396e-05, + "loss": 1.024, + "step": 209920 + }, + { + "epoch": 18.17, + "learning_rate": 3.185653642900459e-05, + "loss": 0.9949, + "step": 209930 + }, + { + "epoch": 18.17, + "learning_rate": 3.1855670103092785e-05, + "loss": 1.0312, + "step": 209940 + }, + { + "epoch": 18.17, + "learning_rate": 3.185480377718097e-05, + "loss": 1.0273, + "step": 209950 + }, + { + "epoch": 18.17, + "learning_rate": 3.185393745126917e-05, + "loss": 0.9832, + "step": 209960 + }, + { + "epoch": 18.17, + "learning_rate": 3.185307112535736e-05, + "loss": 0.921, + "step": 209970 + }, + { + "epoch": 18.17, + "learning_rate": 3.185220479944555e-05, + "loss": 0.9887, + "step": 209980 + }, + { + "epoch": 18.18, + "learning_rate": 3.1851338473533744e-05, + "loss": 0.9657, + "step": 209990 + }, + { + "epoch": 18.18, + "learning_rate": 3.185047214762193e-05, + "loss": 0.9512, + "step": 210000 + }, + { + "epoch": 18.18, + "learning_rate": 3.1849605821710126e-05, + "loss": 0.976, + "step": 210010 + }, + { + "epoch": 18.18, + "learning_rate": 3.184873949579832e-05, + "loss": 0.958, + "step": 210020 + }, + { + "epoch": 18.18, + "learning_rate": 3.184787316988651e-05, + "loss": 0.9725, + "step": 210030 + }, + { + "epoch": 18.18, + "learning_rate": 3.18470068439747e-05, + "loss": 0.9513, + "step": 210040 + }, + { + "epoch": 18.18, + "learning_rate": 3.18461405180629e-05, + "loss": 0.9628, + "step": 210050 + }, + { + "epoch": 18.18, + "learning_rate": 3.1845274192151085e-05, + "loss": 0.9146, + "step": 210060 + }, + { + "epoch": 18.18, + "learning_rate": 3.184440786623928e-05, + "loss": 0.9744, + "step": 210070 + }, + { + "epoch": 18.18, + "learning_rate": 3.1843541540327474e-05, + "loss": 1.0117, + "step": 210080 + }, + { + "epoch": 18.18, + "learning_rate": 3.184267521441566e-05, + "loss": 1.0481, + "step": 210090 + }, + { + "epoch": 18.19, + "learning_rate": 3.1841808888503856e-05, + "loss": 1.0055, + "step": 210100 + }, + { + "epoch": 18.19, + "learning_rate": 3.184094256259204e-05, + "loss": 0.9788, + "step": 210110 + }, + { + "epoch": 18.19, + "learning_rate": 3.184007623668024e-05, + "loss": 0.8607, + "step": 210120 + }, + { + "epoch": 18.19, + "learning_rate": 3.183920991076843e-05, + "loss": 0.9791, + "step": 210130 + }, + { + "epoch": 18.19, + "learning_rate": 3.183834358485662e-05, + "loss": 1.0067, + "step": 210140 + }, + { + "epoch": 18.19, + "learning_rate": 3.1837477258944814e-05, + "loss": 1.0042, + "step": 210150 + }, + { + "epoch": 18.19, + "learning_rate": 3.183661093303301e-05, + "loss": 0.8808, + "step": 210160 + }, + { + "epoch": 18.19, + "learning_rate": 3.1835744607121197e-05, + "loss": 0.9729, + "step": 210170 + }, + { + "epoch": 18.19, + "learning_rate": 3.183487828120939e-05, + "loss": 0.9656, + "step": 210180 + }, + { + "epoch": 18.19, + "learning_rate": 3.1834011955297585e-05, + "loss": 0.9867, + "step": 210190 + }, + { + "epoch": 18.19, + "learning_rate": 3.183314562938577e-05, + "loss": 1.0106, + "step": 210200 + }, + { + "epoch": 18.19, + "learning_rate": 3.183227930347397e-05, + "loss": 0.9475, + "step": 210210 + }, + { + "epoch": 18.2, + "learning_rate": 3.1831412977562155e-05, + "loss": 0.9698, + "step": 210220 + }, + { + "epoch": 18.2, + "learning_rate": 3.183054665165035e-05, + "loss": 0.9418, + "step": 210230 + }, + { + "epoch": 18.2, + "learning_rate": 3.1829680325738544e-05, + "loss": 0.9941, + "step": 210240 + }, + { + "epoch": 18.2, + "learning_rate": 3.182881399982673e-05, + "loss": 1.0119, + "step": 210250 + }, + { + "epoch": 18.2, + "learning_rate": 3.1827947673914926e-05, + "loss": 0.9546, + "step": 210260 + }, + { + "epoch": 18.2, + "learning_rate": 3.182708134800312e-05, + "loss": 0.9974, + "step": 210270 + }, + { + "epoch": 18.2, + "learning_rate": 3.182621502209131e-05, + "loss": 1.0182, + "step": 210280 + }, + { + "epoch": 18.2, + "learning_rate": 3.18253486961795e-05, + "loss": 0.9358, + "step": 210290 + }, + { + "epoch": 18.2, + "learning_rate": 3.18244823702677e-05, + "loss": 1.0069, + "step": 210300 + }, + { + "epoch": 18.2, + "learning_rate": 3.1823616044355885e-05, + "loss": 0.9771, + "step": 210310 + }, + { + "epoch": 18.2, + "learning_rate": 3.182274971844408e-05, + "loss": 0.9832, + "step": 210320 + }, + { + "epoch": 18.21, + "learning_rate": 3.182188339253227e-05, + "loss": 0.9754, + "step": 210330 + }, + { + "epoch": 18.21, + "learning_rate": 3.182101706662046e-05, + "loss": 0.9902, + "step": 210340 + }, + { + "epoch": 18.21, + "learning_rate": 3.1820150740708656e-05, + "loss": 1.0784, + "step": 210350 + }, + { + "epoch": 18.21, + "learning_rate": 3.1819284414796844e-05, + "loss": 1.002, + "step": 210360 + }, + { + "epoch": 18.21, + "learning_rate": 3.181841808888504e-05, + "loss": 0.9556, + "step": 210370 + }, + { + "epoch": 18.21, + "learning_rate": 3.181755176297323e-05, + "loss": 0.9514, + "step": 210380 + }, + { + "epoch": 18.21, + "learning_rate": 3.181668543706142e-05, + "loss": 0.9795, + "step": 210390 + }, + { + "epoch": 18.21, + "learning_rate": 3.1815819111149615e-05, + "loss": 1.0425, + "step": 210400 + }, + { + "epoch": 18.21, + "learning_rate": 3.181495278523781e-05, + "loss": 1.0069, + "step": 210410 + }, + { + "epoch": 18.21, + "learning_rate": 3.1814086459326e-05, + "loss": 1.0244, + "step": 210420 + }, + { + "epoch": 18.21, + "learning_rate": 3.181322013341419e-05, + "loss": 1.0189, + "step": 210430 + }, + { + "epoch": 18.21, + "learning_rate": 3.181235380750238e-05, + "loss": 1.0373, + "step": 210440 + }, + { + "epoch": 18.22, + "learning_rate": 3.1811487481590573e-05, + "loss": 0.9775, + "step": 210450 + }, + { + "epoch": 18.22, + "learning_rate": 3.181062115567877e-05, + "loss": 0.9911, + "step": 210460 + }, + { + "epoch": 18.22, + "learning_rate": 3.1809754829766956e-05, + "loss": 0.9409, + "step": 210470 + }, + { + "epoch": 18.22, + "learning_rate": 3.180888850385515e-05, + "loss": 0.9072, + "step": 210480 + }, + { + "epoch": 18.22, + "learning_rate": 3.1808022177943344e-05, + "loss": 0.9894, + "step": 210490 + }, + { + "epoch": 18.22, + "learning_rate": 3.180715585203153e-05, + "loss": 0.9521, + "step": 210500 + }, + { + "epoch": 18.22, + "learning_rate": 3.180628952611973e-05, + "loss": 0.9582, + "step": 210510 + }, + { + "epoch": 18.22, + "learning_rate": 3.180542320020792e-05, + "loss": 0.9918, + "step": 210520 + }, + { + "epoch": 18.22, + "learning_rate": 3.180455687429611e-05, + "loss": 0.9775, + "step": 210530 + }, + { + "epoch": 18.22, + "learning_rate": 3.18036905483843e-05, + "loss": 0.9766, + "step": 210540 + }, + { + "epoch": 18.22, + "learning_rate": 3.180282422247249e-05, + "loss": 0.9426, + "step": 210550 + }, + { + "epoch": 18.23, + "learning_rate": 3.1801957896560685e-05, + "loss": 0.9659, + "step": 210560 + }, + { + "epoch": 18.23, + "learning_rate": 3.180109157064888e-05, + "loss": 0.9766, + "step": 210570 + }, + { + "epoch": 18.23, + "learning_rate": 3.180022524473707e-05, + "loss": 1.0203, + "step": 210580 + }, + { + "epoch": 18.23, + "learning_rate": 3.179935891882526e-05, + "loss": 1.0013, + "step": 210590 + }, + { + "epoch": 18.23, + "learning_rate": 3.1798492592913456e-05, + "loss": 0.9624, + "step": 210600 + }, + { + "epoch": 18.23, + "learning_rate": 3.1797626267001644e-05, + "loss": 1.0228, + "step": 210610 + }, + { + "epoch": 18.23, + "learning_rate": 3.179675994108984e-05, + "loss": 0.9851, + "step": 210620 + }, + { + "epoch": 18.23, + "learning_rate": 3.1795893615178026e-05, + "loss": 0.949, + "step": 210630 + }, + { + "epoch": 18.23, + "learning_rate": 3.179502728926622e-05, + "loss": 0.9886, + "step": 210640 + }, + { + "epoch": 18.23, + "learning_rate": 3.1794160963354415e-05, + "loss": 0.9642, + "step": 210650 + }, + { + "epoch": 18.23, + "learning_rate": 3.17932946374426e-05, + "loss": 1.011, + "step": 210660 + }, + { + "epoch": 18.23, + "learning_rate": 3.17924283115308e-05, + "loss": 1.0275, + "step": 210670 + }, + { + "epoch": 18.24, + "learning_rate": 3.179156198561899e-05, + "loss": 0.9617, + "step": 210680 + }, + { + "epoch": 18.24, + "learning_rate": 3.179069565970718e-05, + "loss": 0.9612, + "step": 210690 + }, + { + "epoch": 18.24, + "learning_rate": 3.1789829333795374e-05, + "loss": 1.0013, + "step": 210700 + }, + { + "epoch": 18.24, + "learning_rate": 3.178896300788357e-05, + "loss": 0.9963, + "step": 210710 + }, + { + "epoch": 18.24, + "learning_rate": 3.1788096681971756e-05, + "loss": 0.9628, + "step": 210720 + }, + { + "epoch": 18.24, + "learning_rate": 3.178723035605995e-05, + "loss": 0.987, + "step": 210730 + }, + { + "epoch": 18.24, + "learning_rate": 3.178636403014814e-05, + "loss": 0.9727, + "step": 210740 + }, + { + "epoch": 18.24, + "learning_rate": 3.178549770423633e-05, + "loss": 0.9603, + "step": 210750 + }, + { + "epoch": 18.24, + "learning_rate": 3.178463137832453e-05, + "loss": 1.0119, + "step": 210760 + }, + { + "epoch": 18.24, + "learning_rate": 3.1783765052412715e-05, + "loss": 0.9594, + "step": 210770 + }, + { + "epoch": 18.24, + "learning_rate": 3.178289872650091e-05, + "loss": 0.9561, + "step": 210780 + }, + { + "epoch": 18.24, + "learning_rate": 3.1782032400589104e-05, + "loss": 0.9756, + "step": 210790 + }, + { + "epoch": 18.25, + "learning_rate": 3.178116607467729e-05, + "loss": 0.99, + "step": 210800 + }, + { + "epoch": 18.25, + "learning_rate": 3.1780299748765486e-05, + "loss": 0.9667, + "step": 210810 + }, + { + "epoch": 18.25, + "learning_rate": 3.177943342285368e-05, + "loss": 0.9199, + "step": 210820 + }, + { + "epoch": 18.25, + "learning_rate": 3.177856709694187e-05, + "loss": 0.9566, + "step": 210830 + }, + { + "epoch": 18.25, + "learning_rate": 3.177770077103006e-05, + "loss": 0.9453, + "step": 210840 + }, + { + "epoch": 18.25, + "learning_rate": 3.177683444511825e-05, + "loss": 0.9426, + "step": 210850 + }, + { + "epoch": 18.25, + "learning_rate": 3.1775968119206444e-05, + "loss": 0.9491, + "step": 210860 + }, + { + "epoch": 18.25, + "learning_rate": 3.177510179329464e-05, + "loss": 0.9601, + "step": 210870 + }, + { + "epoch": 18.25, + "learning_rate": 3.1774235467382827e-05, + "loss": 1.021, + "step": 210880 + }, + { + "epoch": 18.25, + "learning_rate": 3.177336914147102e-05, + "loss": 0.9373, + "step": 210890 + }, + { + "epoch": 18.25, + "learning_rate": 3.1772502815559215e-05, + "loss": 0.956, + "step": 210900 + }, + { + "epoch": 18.26, + "learning_rate": 3.17716364896474e-05, + "loss": 0.9604, + "step": 210910 + }, + { + "epoch": 18.26, + "learning_rate": 3.17707701637356e-05, + "loss": 1.0341, + "step": 210920 + }, + { + "epoch": 18.26, + "learning_rate": 3.176990383782379e-05, + "loss": 0.9697, + "step": 210930 + }, + { + "epoch": 18.26, + "learning_rate": 3.176903751191198e-05, + "loss": 0.9534, + "step": 210940 + }, + { + "epoch": 18.26, + "learning_rate": 3.1768171186000174e-05, + "loss": 0.9163, + "step": 210950 + }, + { + "epoch": 18.26, + "learning_rate": 3.176730486008836e-05, + "loss": 1.0043, + "step": 210960 + }, + { + "epoch": 18.26, + "learning_rate": 3.1766438534176556e-05, + "loss": 0.9847, + "step": 210970 + }, + { + "epoch": 18.26, + "learning_rate": 3.176557220826475e-05, + "loss": 1.01, + "step": 210980 + }, + { + "epoch": 18.26, + "learning_rate": 3.176470588235294e-05, + "loss": 0.9957, + "step": 210990 + }, + { + "epoch": 18.26, + "learning_rate": 3.176383955644113e-05, + "loss": 0.9879, + "step": 211000 + }, + { + "epoch": 18.26, + "learning_rate": 3.176297323052933e-05, + "loss": 0.9788, + "step": 211010 + }, + { + "epoch": 18.26, + "learning_rate": 3.1762106904617515e-05, + "loss": 1.0308, + "step": 211020 + }, + { + "epoch": 18.27, + "learning_rate": 3.176124057870571e-05, + "loss": 1.0017, + "step": 211030 + }, + { + "epoch": 18.27, + "learning_rate": 3.1760374252793904e-05, + "loss": 0.976, + "step": 211040 + }, + { + "epoch": 18.27, + "learning_rate": 3.175950792688209e-05, + "loss": 0.9417, + "step": 211050 + }, + { + "epoch": 18.27, + "learning_rate": 3.1758641600970286e-05, + "loss": 1.0083, + "step": 211060 + }, + { + "epoch": 18.27, + "learning_rate": 3.1757775275058474e-05, + "loss": 0.9822, + "step": 211070 + }, + { + "epoch": 18.27, + "learning_rate": 3.175690894914667e-05, + "loss": 0.9366, + "step": 211080 + }, + { + "epoch": 18.27, + "learning_rate": 3.175604262323486e-05, + "loss": 0.9502, + "step": 211090 + }, + { + "epoch": 18.27, + "learning_rate": 3.175517629732305e-05, + "loss": 0.9838, + "step": 211100 + }, + { + "epoch": 18.27, + "learning_rate": 3.1754309971411245e-05, + "loss": 0.9941, + "step": 211110 + }, + { + "epoch": 18.27, + "learning_rate": 3.175344364549944e-05, + "loss": 0.9494, + "step": 211120 + }, + { + "epoch": 18.27, + "learning_rate": 3.175257731958763e-05, + "loss": 1.0136, + "step": 211130 + }, + { + "epoch": 18.28, + "learning_rate": 3.175171099367582e-05, + "loss": 1.0198, + "step": 211140 + }, + { + "epoch": 18.28, + "learning_rate": 3.1750844667764016e-05, + "loss": 0.986, + "step": 211150 + }, + { + "epoch": 18.28, + "learning_rate": 3.1749978341852203e-05, + "loss": 1.0291, + "step": 211160 + }, + { + "epoch": 18.28, + "learning_rate": 3.17491120159404e-05, + "loss": 0.938, + "step": 211170 + }, + { + "epoch": 18.28, + "learning_rate": 3.1748245690028586e-05, + "loss": 1.0163, + "step": 211180 + }, + { + "epoch": 18.28, + "learning_rate": 3.174737936411678e-05, + "loss": 1.0273, + "step": 211190 + }, + { + "epoch": 18.28, + "learning_rate": 3.1746513038204974e-05, + "loss": 0.9565, + "step": 211200 + }, + { + "epoch": 18.28, + "learning_rate": 3.174564671229316e-05, + "loss": 0.9597, + "step": 211210 + }, + { + "epoch": 18.28, + "learning_rate": 3.1744780386381357e-05, + "loss": 0.9717, + "step": 211220 + }, + { + "epoch": 18.28, + "learning_rate": 3.174391406046955e-05, + "loss": 0.9883, + "step": 211230 + }, + { + "epoch": 18.28, + "learning_rate": 3.174304773455774e-05, + "loss": 0.9769, + "step": 211240 + }, + { + "epoch": 18.28, + "learning_rate": 3.174218140864593e-05, + "loss": 1.0101, + "step": 211250 + }, + { + "epoch": 18.29, + "learning_rate": 3.174131508273413e-05, + "loss": 0.9362, + "step": 211260 + }, + { + "epoch": 18.29, + "learning_rate": 3.1740448756822315e-05, + "loss": 1.0407, + "step": 211270 + }, + { + "epoch": 18.29, + "learning_rate": 3.173958243091051e-05, + "loss": 0.982, + "step": 211280 + }, + { + "epoch": 18.29, + "learning_rate": 3.17387161049987e-05, + "loss": 0.9996, + "step": 211290 + }, + { + "epoch": 18.29, + "learning_rate": 3.173784977908689e-05, + "loss": 0.9761, + "step": 211300 + }, + { + "epoch": 18.29, + "learning_rate": 3.1736983453175086e-05, + "loss": 0.9528, + "step": 211310 + }, + { + "epoch": 18.29, + "learning_rate": 3.1736117127263274e-05, + "loss": 0.9378, + "step": 211320 + }, + { + "epoch": 18.29, + "learning_rate": 3.173525080135147e-05, + "loss": 1.0502, + "step": 211330 + }, + { + "epoch": 18.29, + "learning_rate": 3.173438447543966e-05, + "loss": 1.0276, + "step": 211340 + }, + { + "epoch": 18.29, + "learning_rate": 3.173351814952785e-05, + "loss": 0.9635, + "step": 211350 + }, + { + "epoch": 18.29, + "learning_rate": 3.1732651823616045e-05, + "loss": 1.0265, + "step": 211360 + }, + { + "epoch": 18.3, + "learning_rate": 3.173178549770424e-05, + "loss": 1.0035, + "step": 211370 + }, + { + "epoch": 18.3, + "learning_rate": 3.173091917179243e-05, + "loss": 0.9787, + "step": 211380 + }, + { + "epoch": 18.3, + "learning_rate": 3.173005284588062e-05, + "loss": 0.9756, + "step": 211390 + }, + { + "epoch": 18.3, + "learning_rate": 3.172918651996881e-05, + "loss": 0.9574, + "step": 211400 + }, + { + "epoch": 18.3, + "learning_rate": 3.1728320194057004e-05, + "loss": 0.9463, + "step": 211410 + }, + { + "epoch": 18.3, + "learning_rate": 3.17274538681452e-05, + "loss": 1.0355, + "step": 211420 + }, + { + "epoch": 18.3, + "learning_rate": 3.1726587542233386e-05, + "loss": 0.9642, + "step": 211430 + }, + { + "epoch": 18.3, + "learning_rate": 3.172572121632158e-05, + "loss": 0.9502, + "step": 211440 + }, + { + "epoch": 18.3, + "learning_rate": 3.1724854890409775e-05, + "loss": 1.0197, + "step": 211450 + }, + { + "epoch": 18.3, + "learning_rate": 3.172398856449796e-05, + "loss": 1.0149, + "step": 211460 + }, + { + "epoch": 18.3, + "learning_rate": 3.172312223858616e-05, + "loss": 1.0195, + "step": 211470 + }, + { + "epoch": 18.3, + "learning_rate": 3.1722255912674345e-05, + "loss": 0.9972, + "step": 211480 + }, + { + "epoch": 18.31, + "learning_rate": 3.172138958676254e-05, + "loss": 0.925, + "step": 211490 + }, + { + "epoch": 18.31, + "learning_rate": 3.1720523260850734e-05, + "loss": 0.9613, + "step": 211500 + }, + { + "epoch": 18.31, + "learning_rate": 3.171965693493892e-05, + "loss": 0.9663, + "step": 211510 + }, + { + "epoch": 18.31, + "learning_rate": 3.1718790609027116e-05, + "loss": 0.9727, + "step": 211520 + }, + { + "epoch": 18.31, + "learning_rate": 3.171792428311531e-05, + "loss": 0.9747, + "step": 211530 + }, + { + "epoch": 18.31, + "learning_rate": 3.17170579572035e-05, + "loss": 0.9693, + "step": 211540 + }, + { + "epoch": 18.31, + "learning_rate": 3.171619163129169e-05, + "loss": 0.9794, + "step": 211550 + }, + { + "epoch": 18.31, + "learning_rate": 3.171532530537989e-05, + "loss": 0.9898, + "step": 211560 + }, + { + "epoch": 18.31, + "learning_rate": 3.1714458979468074e-05, + "loss": 1.0037, + "step": 211570 + }, + { + "epoch": 18.31, + "learning_rate": 3.171359265355627e-05, + "loss": 1.0341, + "step": 211580 + }, + { + "epoch": 18.31, + "learning_rate": 3.1712726327644456e-05, + "loss": 1.0107, + "step": 211590 + }, + { + "epoch": 18.32, + "learning_rate": 3.171186000173265e-05, + "loss": 0.9278, + "step": 211600 + }, + { + "epoch": 18.32, + "learning_rate": 3.1710993675820845e-05, + "loss": 0.9544, + "step": 211610 + }, + { + "epoch": 18.32, + "learning_rate": 3.171012734990903e-05, + "loss": 0.9956, + "step": 211620 + }, + { + "epoch": 18.32, + "learning_rate": 3.170926102399723e-05, + "loss": 0.9761, + "step": 211630 + }, + { + "epoch": 18.32, + "learning_rate": 3.170839469808542e-05, + "loss": 0.9692, + "step": 211640 + }, + { + "epoch": 18.32, + "learning_rate": 3.170752837217361e-05, + "loss": 0.9374, + "step": 211650 + }, + { + "epoch": 18.32, + "learning_rate": 3.1706662046261804e-05, + "loss": 0.9644, + "step": 211660 + }, + { + "epoch": 18.32, + "learning_rate": 3.170579572035e-05, + "loss": 0.9783, + "step": 211670 + }, + { + "epoch": 18.32, + "learning_rate": 3.1704929394438186e-05, + "loss": 0.9176, + "step": 211680 + }, + { + "epoch": 18.32, + "learning_rate": 3.170406306852638e-05, + "loss": 0.9478, + "step": 211690 + }, + { + "epoch": 18.32, + "learning_rate": 3.170319674261457e-05, + "loss": 0.9569, + "step": 211700 + }, + { + "epoch": 18.32, + "learning_rate": 3.170233041670276e-05, + "loss": 0.9761, + "step": 211710 + }, + { + "epoch": 18.33, + "learning_rate": 3.170146409079096e-05, + "loss": 1.002, + "step": 211720 + }, + { + "epoch": 18.33, + "learning_rate": 3.1700597764879145e-05, + "loss": 1.0041, + "step": 211730 + }, + { + "epoch": 18.33, + "learning_rate": 3.169973143896734e-05, + "loss": 0.9918, + "step": 211740 + }, + { + "epoch": 18.33, + "learning_rate": 3.1698865113055534e-05, + "loss": 0.9694, + "step": 211750 + }, + { + "epoch": 18.33, + "learning_rate": 3.169799878714372e-05, + "loss": 0.9978, + "step": 211760 + }, + { + "epoch": 18.33, + "learning_rate": 3.1697132461231916e-05, + "loss": 0.965, + "step": 211770 + }, + { + "epoch": 18.33, + "learning_rate": 3.169626613532011e-05, + "loss": 1.0161, + "step": 211780 + }, + { + "epoch": 18.33, + "learning_rate": 3.16953998094083e-05, + "loss": 1.0355, + "step": 211790 + }, + { + "epoch": 18.33, + "learning_rate": 3.169453348349649e-05, + "loss": 0.9592, + "step": 211800 + }, + { + "epoch": 18.33, + "learning_rate": 3.169366715758468e-05, + "loss": 1.0084, + "step": 211810 + }, + { + "epoch": 18.33, + "learning_rate": 3.1692800831672875e-05, + "loss": 0.9433, + "step": 211820 + }, + { + "epoch": 18.33, + "learning_rate": 3.169193450576107e-05, + "loss": 1.0043, + "step": 211830 + }, + { + "epoch": 18.34, + "learning_rate": 3.169106817984926e-05, + "loss": 0.9716, + "step": 211840 + }, + { + "epoch": 18.34, + "learning_rate": 3.169020185393745e-05, + "loss": 1.0022, + "step": 211850 + }, + { + "epoch": 18.34, + "learning_rate": 3.1689335528025646e-05, + "loss": 0.9945, + "step": 211860 + }, + { + "epoch": 18.34, + "learning_rate": 3.1688469202113833e-05, + "loss": 1.0213, + "step": 211870 + }, + { + "epoch": 18.34, + "learning_rate": 3.168760287620203e-05, + "loss": 0.9831, + "step": 211880 + }, + { + "epoch": 18.34, + "learning_rate": 3.168673655029022e-05, + "loss": 0.9457, + "step": 211890 + }, + { + "epoch": 18.34, + "learning_rate": 3.168587022437841e-05, + "loss": 0.9415, + "step": 211900 + }, + { + "epoch": 18.34, + "learning_rate": 3.1685003898466604e-05, + "loss": 0.9416, + "step": 211910 + }, + { + "epoch": 18.34, + "learning_rate": 3.168413757255479e-05, + "loss": 0.9724, + "step": 211920 + }, + { + "epoch": 18.34, + "learning_rate": 3.1683271246642987e-05, + "loss": 0.9363, + "step": 211930 + }, + { + "epoch": 18.34, + "learning_rate": 3.168240492073118e-05, + "loss": 0.9138, + "step": 211940 + }, + { + "epoch": 18.35, + "learning_rate": 3.168153859481937e-05, + "loss": 1.0293, + "step": 211950 + }, + { + "epoch": 18.35, + "learning_rate": 3.168067226890756e-05, + "loss": 0.9553, + "step": 211960 + }, + { + "epoch": 18.35, + "learning_rate": 3.167980594299576e-05, + "loss": 0.9539, + "step": 211970 + }, + { + "epoch": 18.35, + "learning_rate": 3.1678939617083945e-05, + "loss": 0.9228, + "step": 211980 + }, + { + "epoch": 18.35, + "learning_rate": 3.167807329117214e-05, + "loss": 0.9817, + "step": 211990 + }, + { + "epoch": 18.35, + "learning_rate": 3.1677206965260334e-05, + "loss": 0.9964, + "step": 212000 + }, + { + "epoch": 18.35, + "learning_rate": 3.167634063934852e-05, + "loss": 0.9879, + "step": 212010 + }, + { + "epoch": 18.35, + "learning_rate": 3.1675474313436716e-05, + "loss": 1.0076, + "step": 212020 + }, + { + "epoch": 18.35, + "learning_rate": 3.1674607987524904e-05, + "loss": 1.0006, + "step": 212030 + }, + { + "epoch": 18.35, + "learning_rate": 3.16737416616131e-05, + "loss": 1.0304, + "step": 212040 + }, + { + "epoch": 18.35, + "learning_rate": 3.167287533570129e-05, + "loss": 0.9749, + "step": 212050 + }, + { + "epoch": 18.35, + "learning_rate": 3.167200900978948e-05, + "loss": 1.0021, + "step": 212060 + }, + { + "epoch": 18.36, + "learning_rate": 3.1671142683877675e-05, + "loss": 0.9617, + "step": 212070 + }, + { + "epoch": 18.36, + "learning_rate": 3.167027635796587e-05, + "loss": 0.9833, + "step": 212080 + }, + { + "epoch": 18.36, + "learning_rate": 3.166941003205406e-05, + "loss": 0.9562, + "step": 212090 + }, + { + "epoch": 18.36, + "learning_rate": 3.166854370614225e-05, + "loss": 0.9743, + "step": 212100 + }, + { + "epoch": 18.36, + "learning_rate": 3.1667677380230446e-05, + "loss": 0.9877, + "step": 212110 + }, + { + "epoch": 18.36, + "learning_rate": 3.1666811054318634e-05, + "loss": 0.9115, + "step": 212120 + }, + { + "epoch": 18.36, + "learning_rate": 3.166594472840683e-05, + "loss": 1.0127, + "step": 212130 + }, + { + "epoch": 18.36, + "learning_rate": 3.1665078402495016e-05, + "loss": 1.0013, + "step": 212140 + }, + { + "epoch": 18.36, + "learning_rate": 3.166421207658321e-05, + "loss": 0.9911, + "step": 212150 + }, + { + "epoch": 18.36, + "learning_rate": 3.1663345750671405e-05, + "loss": 0.9606, + "step": 212160 + }, + { + "epoch": 18.36, + "learning_rate": 3.166247942475959e-05, + "loss": 0.9781, + "step": 212170 + }, + { + "epoch": 18.37, + "learning_rate": 3.166161309884779e-05, + "loss": 0.952, + "step": 212180 + }, + { + "epoch": 18.37, + "learning_rate": 3.166074677293598e-05, + "loss": 1.0618, + "step": 212190 + }, + { + "epoch": 18.37, + "learning_rate": 3.165988044702417e-05, + "loss": 0.9897, + "step": 212200 + }, + { + "epoch": 18.37, + "learning_rate": 3.1659014121112363e-05, + "loss": 0.9303, + "step": 212210 + }, + { + "epoch": 18.37, + "learning_rate": 3.165814779520055e-05, + "loss": 0.9982, + "step": 212220 + }, + { + "epoch": 18.37, + "learning_rate": 3.1657281469288746e-05, + "loss": 0.9848, + "step": 212230 + }, + { + "epoch": 18.37, + "learning_rate": 3.165641514337694e-05, + "loss": 0.9426, + "step": 212240 + }, + { + "epoch": 18.37, + "learning_rate": 3.165554881746513e-05, + "loss": 0.9757, + "step": 212250 + }, + { + "epoch": 18.37, + "learning_rate": 3.165468249155332e-05, + "loss": 0.9301, + "step": 212260 + }, + { + "epoch": 18.37, + "learning_rate": 3.165381616564152e-05, + "loss": 0.9701, + "step": 212270 + }, + { + "epoch": 18.37, + "learning_rate": 3.1652949839729704e-05, + "loss": 0.9662, + "step": 212280 + }, + { + "epoch": 18.37, + "learning_rate": 3.16520835138179e-05, + "loss": 0.9934, + "step": 212290 + }, + { + "epoch": 18.38, + "learning_rate": 3.165121718790609e-05, + "loss": 1.0134, + "step": 212300 + }, + { + "epoch": 18.38, + "learning_rate": 3.165035086199428e-05, + "loss": 1.0058, + "step": 212310 + }, + { + "epoch": 18.38, + "learning_rate": 3.1649484536082475e-05, + "loss": 0.99, + "step": 212320 + }, + { + "epoch": 18.38, + "learning_rate": 3.164861821017066e-05, + "loss": 0.9457, + "step": 212330 + }, + { + "epoch": 18.38, + "learning_rate": 3.164775188425886e-05, + "loss": 0.9684, + "step": 212340 + }, + { + "epoch": 18.38, + "learning_rate": 3.164688555834705e-05, + "loss": 0.9916, + "step": 212350 + }, + { + "epoch": 18.38, + "learning_rate": 3.164601923243524e-05, + "loss": 1.0018, + "step": 212360 + }, + { + "epoch": 18.38, + "learning_rate": 3.1645152906523434e-05, + "loss": 1.0034, + "step": 212370 + }, + { + "epoch": 18.38, + "learning_rate": 3.164428658061163e-05, + "loss": 0.9727, + "step": 212380 + }, + { + "epoch": 18.38, + "learning_rate": 3.1643420254699816e-05, + "loss": 0.973, + "step": 212390 + }, + { + "epoch": 18.38, + "learning_rate": 3.164255392878801e-05, + "loss": 0.9634, + "step": 212400 + }, + { + "epoch": 18.39, + "learning_rate": 3.1641687602876205e-05, + "loss": 0.97, + "step": 212410 + }, + { + "epoch": 18.39, + "learning_rate": 3.164082127696439e-05, + "loss": 0.95, + "step": 212420 + }, + { + "epoch": 18.39, + "learning_rate": 3.163995495105259e-05, + "loss": 0.9586, + "step": 212430 + }, + { + "epoch": 18.39, + "learning_rate": 3.1639088625140775e-05, + "loss": 0.9423, + "step": 212440 + }, + { + "epoch": 18.39, + "learning_rate": 3.163822229922897e-05, + "loss": 1.0165, + "step": 212450 + }, + { + "epoch": 18.39, + "learning_rate": 3.1637355973317164e-05, + "loss": 0.9678, + "step": 212460 + }, + { + "epoch": 18.39, + "learning_rate": 3.163648964740535e-05, + "loss": 0.9933, + "step": 212470 + }, + { + "epoch": 18.39, + "learning_rate": 3.1635623321493546e-05, + "loss": 0.9516, + "step": 212480 + }, + { + "epoch": 18.39, + "learning_rate": 3.163475699558174e-05, + "loss": 0.9965, + "step": 212490 + }, + { + "epoch": 18.39, + "learning_rate": 3.163389066966993e-05, + "loss": 0.9829, + "step": 212500 + }, + { + "epoch": 18.39, + "learning_rate": 3.163302434375812e-05, + "loss": 0.9556, + "step": 212510 + }, + { + "epoch": 18.39, + "learning_rate": 3.163215801784632e-05, + "loss": 1.0232, + "step": 212520 + }, + { + "epoch": 18.4, + "learning_rate": 3.1631291691934505e-05, + "loss": 1.0009, + "step": 212530 + }, + { + "epoch": 18.4, + "learning_rate": 3.16304253660227e-05, + "loss": 0.9914, + "step": 212540 + }, + { + "epoch": 18.4, + "learning_rate": 3.162955904011089e-05, + "loss": 0.9387, + "step": 212550 + }, + { + "epoch": 18.4, + "learning_rate": 3.162869271419908e-05, + "loss": 0.9855, + "step": 212560 + }, + { + "epoch": 18.4, + "learning_rate": 3.1627826388287276e-05, + "loss": 1.0115, + "step": 212570 + }, + { + "epoch": 18.4, + "learning_rate": 3.162696006237546e-05, + "loss": 0.9105, + "step": 212580 + }, + { + "epoch": 18.4, + "learning_rate": 3.162609373646366e-05, + "loss": 0.9486, + "step": 212590 + }, + { + "epoch": 18.4, + "learning_rate": 3.162522741055185e-05, + "loss": 0.9271, + "step": 212600 + }, + { + "epoch": 18.4, + "learning_rate": 3.162436108464004e-05, + "loss": 0.9709, + "step": 212610 + }, + { + "epoch": 18.4, + "learning_rate": 3.1623494758728234e-05, + "loss": 1.0181, + "step": 212620 + }, + { + "epoch": 18.4, + "learning_rate": 3.162262843281643e-05, + "loss": 0.974, + "step": 212630 + }, + { + "epoch": 18.41, + "learning_rate": 3.1621762106904617e-05, + "loss": 0.9476, + "step": 212640 + }, + { + "epoch": 18.41, + "learning_rate": 3.162089578099281e-05, + "loss": 1.009, + "step": 212650 + }, + { + "epoch": 18.41, + "learning_rate": 3.1620029455081e-05, + "loss": 1.0359, + "step": 212660 + }, + { + "epoch": 18.41, + "learning_rate": 3.161916312916919e-05, + "loss": 0.9654, + "step": 212670 + }, + { + "epoch": 18.41, + "learning_rate": 3.161829680325739e-05, + "loss": 0.9636, + "step": 212680 + }, + { + "epoch": 18.41, + "learning_rate": 3.1617430477345575e-05, + "loss": 1.0013, + "step": 212690 + }, + { + "epoch": 18.41, + "learning_rate": 3.161656415143377e-05, + "loss": 1.0269, + "step": 212700 + }, + { + "epoch": 18.41, + "learning_rate": 3.1615697825521964e-05, + "loss": 0.9601, + "step": 212710 + }, + { + "epoch": 18.41, + "learning_rate": 3.161483149961015e-05, + "loss": 0.995, + "step": 212720 + }, + { + "epoch": 18.41, + "learning_rate": 3.1613965173698346e-05, + "loss": 0.9721, + "step": 212730 + }, + { + "epoch": 18.41, + "learning_rate": 3.161309884778654e-05, + "loss": 0.9481, + "step": 212740 + }, + { + "epoch": 18.41, + "learning_rate": 3.161223252187473e-05, + "loss": 0.9834, + "step": 212750 + }, + { + "epoch": 18.42, + "learning_rate": 3.161136619596292e-05, + "loss": 1.0225, + "step": 212760 + }, + { + "epoch": 18.42, + "learning_rate": 3.161049987005111e-05, + "loss": 0.9967, + "step": 212770 + }, + { + "epoch": 18.42, + "learning_rate": 3.1609633544139305e-05, + "loss": 0.9551, + "step": 212780 + }, + { + "epoch": 18.42, + "learning_rate": 3.16087672182275e-05, + "loss": 0.9916, + "step": 212790 + }, + { + "epoch": 18.42, + "learning_rate": 3.160790089231569e-05, + "loss": 1.0261, + "step": 212800 + }, + { + "epoch": 18.42, + "learning_rate": 3.160703456640388e-05, + "loss": 1.004, + "step": 212810 + }, + { + "epoch": 18.42, + "learning_rate": 3.1606168240492076e-05, + "loss": 1.0341, + "step": 212820 + }, + { + "epoch": 18.42, + "learning_rate": 3.1605301914580264e-05, + "loss": 0.9854, + "step": 212830 + }, + { + "epoch": 18.42, + "learning_rate": 3.160443558866846e-05, + "loss": 0.9856, + "step": 212840 + }, + { + "epoch": 18.42, + "learning_rate": 3.160356926275665e-05, + "loss": 0.9587, + "step": 212850 + }, + { + "epoch": 18.42, + "learning_rate": 3.160270293684484e-05, + "loss": 0.966, + "step": 212860 + }, + { + "epoch": 18.42, + "learning_rate": 3.1601836610933035e-05, + "loss": 0.9195, + "step": 212870 + }, + { + "epoch": 18.43, + "learning_rate": 3.160097028502122e-05, + "loss": 0.9699, + "step": 212880 + }, + { + "epoch": 18.43, + "learning_rate": 3.160010395910942e-05, + "loss": 1.0094, + "step": 212890 + }, + { + "epoch": 18.43, + "learning_rate": 3.159923763319761e-05, + "loss": 1.004, + "step": 212900 + }, + { + "epoch": 18.43, + "learning_rate": 3.15983713072858e-05, + "loss": 1.0258, + "step": 212910 + }, + { + "epoch": 18.43, + "learning_rate": 3.1597504981373993e-05, + "loss": 0.9975, + "step": 212920 + }, + { + "epoch": 18.43, + "learning_rate": 3.159663865546219e-05, + "loss": 0.9582, + "step": 212930 + }, + { + "epoch": 18.43, + "learning_rate": 3.1595772329550376e-05, + "loss": 0.8683, + "step": 212940 + }, + { + "epoch": 18.43, + "learning_rate": 3.159490600363857e-05, + "loss": 0.9261, + "step": 212950 + }, + { + "epoch": 18.43, + "learning_rate": 3.159403967772676e-05, + "loss": 0.9862, + "step": 212960 + }, + { + "epoch": 18.43, + "learning_rate": 3.159317335181495e-05, + "loss": 0.9844, + "step": 212970 + }, + { + "epoch": 18.43, + "learning_rate": 3.159230702590315e-05, + "loss": 1.0085, + "step": 212980 + }, + { + "epoch": 18.44, + "learning_rate": 3.1591440699991334e-05, + "loss": 1.0007, + "step": 212990 + }, + { + "epoch": 18.44, + "learning_rate": 3.159057437407953e-05, + "loss": 0.9879, + "step": 213000 + }, + { + "epoch": 18.44, + "learning_rate": 3.158970804816772e-05, + "loss": 1.0356, + "step": 213010 + }, + { + "epoch": 18.44, + "learning_rate": 3.158884172225591e-05, + "loss": 0.9962, + "step": 213020 + }, + { + "epoch": 18.44, + "learning_rate": 3.1587975396344105e-05, + "loss": 0.9264, + "step": 213030 + }, + { + "epoch": 18.44, + "learning_rate": 3.15871090704323e-05, + "loss": 1.0138, + "step": 213040 + }, + { + "epoch": 18.44, + "learning_rate": 3.158624274452049e-05, + "loss": 0.9761, + "step": 213050 + }, + { + "epoch": 18.44, + "learning_rate": 3.158537641860868e-05, + "loss": 1.0286, + "step": 213060 + }, + { + "epoch": 18.44, + "learning_rate": 3.158451009269687e-05, + "loss": 0.963, + "step": 213070 + }, + { + "epoch": 18.44, + "learning_rate": 3.1583643766785064e-05, + "loss": 0.9941, + "step": 213080 + }, + { + "epoch": 18.44, + "learning_rate": 3.158277744087326e-05, + "loss": 0.9346, + "step": 213090 + }, + { + "epoch": 18.44, + "learning_rate": 3.1581911114961446e-05, + "loss": 0.995, + "step": 213100 + }, + { + "epoch": 18.45, + "learning_rate": 3.158104478904964e-05, + "loss": 0.9713, + "step": 213110 + }, + { + "epoch": 18.45, + "learning_rate": 3.1580178463137835e-05, + "loss": 1.0503, + "step": 213120 + }, + { + "epoch": 18.45, + "learning_rate": 3.157931213722602e-05, + "loss": 0.9366, + "step": 213130 + }, + { + "epoch": 18.45, + "learning_rate": 3.157844581131422e-05, + "loss": 0.9977, + "step": 213140 + }, + { + "epoch": 18.45, + "learning_rate": 3.157757948540241e-05, + "loss": 0.9964, + "step": 213150 + }, + { + "epoch": 18.45, + "learning_rate": 3.15767131594906e-05, + "loss": 0.9957, + "step": 213160 + }, + { + "epoch": 18.45, + "learning_rate": 3.1575846833578794e-05, + "loss": 0.9658, + "step": 213170 + }, + { + "epoch": 18.45, + "learning_rate": 3.157498050766698e-05, + "loss": 0.9875, + "step": 213180 + }, + { + "epoch": 18.45, + "learning_rate": 3.1574114181755176e-05, + "loss": 1.0174, + "step": 213190 + }, + { + "epoch": 18.45, + "learning_rate": 3.157324785584337e-05, + "loss": 0.9447, + "step": 213200 + }, + { + "epoch": 18.45, + "learning_rate": 3.157238152993156e-05, + "loss": 0.9715, + "step": 213210 + }, + { + "epoch": 18.46, + "learning_rate": 3.157151520401975e-05, + "loss": 0.9851, + "step": 213220 + }, + { + "epoch": 18.46, + "learning_rate": 3.157064887810795e-05, + "loss": 0.9772, + "step": 213230 + }, + { + "epoch": 18.46, + "learning_rate": 3.1569782552196135e-05, + "loss": 1.0234, + "step": 213240 + }, + { + "epoch": 18.46, + "learning_rate": 3.156891622628433e-05, + "loss": 0.9373, + "step": 213250 + }, + { + "epoch": 18.46, + "learning_rate": 3.1568049900372524e-05, + "loss": 1.0188, + "step": 213260 + }, + { + "epoch": 18.46, + "learning_rate": 3.156718357446071e-05, + "loss": 0.9737, + "step": 213270 + }, + { + "epoch": 18.46, + "learning_rate": 3.1566317248548906e-05, + "loss": 0.9493, + "step": 213280 + }, + { + "epoch": 18.46, + "learning_rate": 3.156545092263709e-05, + "loss": 1.0233, + "step": 213290 + }, + { + "epoch": 18.46, + "learning_rate": 3.156458459672529e-05, + "loss": 0.9926, + "step": 213300 + }, + { + "epoch": 18.46, + "learning_rate": 3.156371827081348e-05, + "loss": 0.9344, + "step": 213310 + }, + { + "epoch": 18.46, + "learning_rate": 3.156285194490167e-05, + "loss": 0.9747, + "step": 213320 + }, + { + "epoch": 18.46, + "learning_rate": 3.1561985618989864e-05, + "loss": 0.9911, + "step": 213330 + }, + { + "epoch": 18.47, + "learning_rate": 3.156111929307806e-05, + "loss": 1.0622, + "step": 213340 + }, + { + "epoch": 18.47, + "learning_rate": 3.1560252967166247e-05, + "loss": 0.9782, + "step": 213350 + }, + { + "epoch": 18.47, + "learning_rate": 3.155938664125444e-05, + "loss": 1.0186, + "step": 213360 + }, + { + "epoch": 18.47, + "learning_rate": 3.1558520315342635e-05, + "loss": 1.0022, + "step": 213370 + }, + { + "epoch": 18.47, + "learning_rate": 3.155765398943082e-05, + "loss": 0.9257, + "step": 213380 + }, + { + "epoch": 18.47, + "learning_rate": 3.155678766351902e-05, + "loss": 0.9901, + "step": 213390 + }, + { + "epoch": 18.47, + "learning_rate": 3.1555921337607205e-05, + "loss": 0.9552, + "step": 213400 + }, + { + "epoch": 18.47, + "learning_rate": 3.15550550116954e-05, + "loss": 1.0109, + "step": 213410 + }, + { + "epoch": 18.47, + "learning_rate": 3.1554188685783594e-05, + "loss": 1.0065, + "step": 213420 + }, + { + "epoch": 18.47, + "learning_rate": 3.155332235987178e-05, + "loss": 0.9726, + "step": 213430 + }, + { + "epoch": 18.47, + "learning_rate": 3.1552456033959976e-05, + "loss": 0.9538, + "step": 213440 + }, + { + "epoch": 18.48, + "learning_rate": 3.155158970804817e-05, + "loss": 0.9575, + "step": 213450 + }, + { + "epoch": 18.48, + "learning_rate": 3.155072338213636e-05, + "loss": 0.9552, + "step": 213460 + }, + { + "epoch": 18.48, + "learning_rate": 3.154985705622455e-05, + "loss": 1.0029, + "step": 213470 + }, + { + "epoch": 18.48, + "learning_rate": 3.154899073031275e-05, + "loss": 1.0001, + "step": 213480 + }, + { + "epoch": 18.48, + "learning_rate": 3.1548124404400935e-05, + "loss": 0.9807, + "step": 213490 + }, + { + "epoch": 18.48, + "learning_rate": 3.154725807848913e-05, + "loss": 0.931, + "step": 213500 + }, + { + "epoch": 18.48, + "learning_rate": 3.154639175257732e-05, + "loss": 0.9988, + "step": 213510 + }, + { + "epoch": 18.48, + "learning_rate": 3.154552542666551e-05, + "loss": 0.9755, + "step": 213520 + }, + { + "epoch": 18.48, + "learning_rate": 3.1544659100753706e-05, + "loss": 0.9717, + "step": 213530 + }, + { + "epoch": 18.48, + "learning_rate": 3.1543792774841894e-05, + "loss": 0.9435, + "step": 213540 + }, + { + "epoch": 18.48, + "learning_rate": 3.154292644893009e-05, + "loss": 0.9973, + "step": 213550 + }, + { + "epoch": 18.48, + "learning_rate": 3.154206012301828e-05, + "loss": 0.9483, + "step": 213560 + }, + { + "epoch": 18.49, + "learning_rate": 3.154119379710647e-05, + "loss": 0.9578, + "step": 213570 + }, + { + "epoch": 18.49, + "learning_rate": 3.1540327471194665e-05, + "loss": 0.8903, + "step": 213580 + }, + { + "epoch": 18.49, + "learning_rate": 3.153946114528286e-05, + "loss": 0.9426, + "step": 213590 + }, + { + "epoch": 18.49, + "learning_rate": 3.153859481937105e-05, + "loss": 0.9687, + "step": 213600 + }, + { + "epoch": 18.49, + "learning_rate": 3.153772849345924e-05, + "loss": 0.9653, + "step": 213610 + }, + { + "epoch": 18.49, + "learning_rate": 3.153686216754743e-05, + "loss": 0.9497, + "step": 213620 + }, + { + "epoch": 18.49, + "learning_rate": 3.1535995841635623e-05, + "loss": 0.9975, + "step": 213630 + }, + { + "epoch": 18.49, + "learning_rate": 3.153512951572382e-05, + "loss": 0.9756, + "step": 213640 + }, + { + "epoch": 18.49, + "learning_rate": 3.1534263189812006e-05, + "loss": 0.9663, + "step": 213650 + }, + { + "epoch": 18.49, + "learning_rate": 3.15333968639002e-05, + "loss": 1.0059, + "step": 213660 + }, + { + "epoch": 18.49, + "learning_rate": 3.1532530537988394e-05, + "loss": 0.9439, + "step": 213670 + }, + { + "epoch": 18.5, + "learning_rate": 3.153166421207658e-05, + "loss": 0.9572, + "step": 213680 + }, + { + "epoch": 18.5, + "learning_rate": 3.1530797886164777e-05, + "loss": 0.9705, + "step": 213690 + }, + { + "epoch": 18.5, + "learning_rate": 3.152993156025297e-05, + "loss": 0.9562, + "step": 213700 + }, + { + "epoch": 18.5, + "learning_rate": 3.152906523434116e-05, + "loss": 0.9791, + "step": 213710 + }, + { + "epoch": 18.5, + "learning_rate": 3.152819890842935e-05, + "loss": 0.9694, + "step": 213720 + }, + { + "epoch": 18.5, + "learning_rate": 3.152733258251754e-05, + "loss": 1.0019, + "step": 213730 + }, + { + "epoch": 18.5, + "learning_rate": 3.1526466256605735e-05, + "loss": 1.0418, + "step": 213740 + }, + { + "epoch": 18.5, + "learning_rate": 3.152559993069393e-05, + "loss": 0.9567, + "step": 213750 + }, + { + "epoch": 18.5, + "learning_rate": 3.152473360478212e-05, + "loss": 1.0075, + "step": 213760 + }, + { + "epoch": 18.5, + "learning_rate": 3.152386727887031e-05, + "loss": 0.9414, + "step": 213770 + }, + { + "epoch": 18.5, + "learning_rate": 3.1523000952958506e-05, + "loss": 0.9918, + "step": 213780 + }, + { + "epoch": 18.5, + "learning_rate": 3.1522134627046694e-05, + "loss": 0.9726, + "step": 213790 + }, + { + "epoch": 18.51, + "learning_rate": 3.152126830113489e-05, + "loss": 0.9586, + "step": 213800 + }, + { + "epoch": 18.51, + "learning_rate": 3.1520401975223076e-05, + "loss": 1.0085, + "step": 213810 + }, + { + "epoch": 18.51, + "learning_rate": 3.151953564931127e-05, + "loss": 0.9648, + "step": 213820 + }, + { + "epoch": 18.51, + "learning_rate": 3.1518669323399465e-05, + "loss": 0.9423, + "step": 213830 + }, + { + "epoch": 18.51, + "learning_rate": 3.151780299748765e-05, + "loss": 0.925, + "step": 213840 + }, + { + "epoch": 18.51, + "learning_rate": 3.151693667157585e-05, + "loss": 0.9502, + "step": 213850 + }, + { + "epoch": 18.51, + "learning_rate": 3.151607034566404e-05, + "loss": 0.9986, + "step": 213860 + }, + { + "epoch": 18.51, + "learning_rate": 3.151520401975223e-05, + "loss": 1.0471, + "step": 213870 + }, + { + "epoch": 18.51, + "learning_rate": 3.1514337693840424e-05, + "loss": 1.0678, + "step": 213880 + }, + { + "epoch": 18.51, + "learning_rate": 3.151347136792862e-05, + "loss": 0.9702, + "step": 213890 + }, + { + "epoch": 18.51, + "learning_rate": 3.1512605042016806e-05, + "loss": 0.9981, + "step": 213900 + }, + { + "epoch": 18.51, + "learning_rate": 3.1511738716105e-05, + "loss": 0.949, + "step": 213910 + }, + { + "epoch": 18.52, + "learning_rate": 3.151087239019319e-05, + "loss": 0.9859, + "step": 213920 + }, + { + "epoch": 18.52, + "learning_rate": 3.151000606428138e-05, + "loss": 0.9699, + "step": 213930 + }, + { + "epoch": 18.52, + "learning_rate": 3.150913973836958e-05, + "loss": 0.9369, + "step": 213940 + }, + { + "epoch": 18.52, + "learning_rate": 3.1508273412457765e-05, + "loss": 1.0513, + "step": 213950 + }, + { + "epoch": 18.52, + "learning_rate": 3.150740708654596e-05, + "loss": 0.9357, + "step": 213960 + }, + { + "epoch": 18.52, + "learning_rate": 3.1506540760634154e-05, + "loss": 0.9906, + "step": 213970 + }, + { + "epoch": 18.52, + "learning_rate": 3.150567443472234e-05, + "loss": 1.0057, + "step": 213980 + }, + { + "epoch": 18.52, + "learning_rate": 3.1504808108810536e-05, + "loss": 0.9733, + "step": 213990 + }, + { + "epoch": 18.52, + "learning_rate": 3.150394178289873e-05, + "loss": 0.9902, + "step": 214000 + }, + { + "epoch": 18.52, + "learning_rate": 3.150307545698692e-05, + "loss": 1.056, + "step": 214010 + }, + { + "epoch": 18.52, + "learning_rate": 3.150220913107511e-05, + "loss": 1.001, + "step": 214020 + }, + { + "epoch": 18.53, + "learning_rate": 3.15013428051633e-05, + "loss": 0.9998, + "step": 214030 + }, + { + "epoch": 18.53, + "learning_rate": 3.1500476479251494e-05, + "loss": 0.9295, + "step": 214040 + }, + { + "epoch": 18.53, + "learning_rate": 3.149961015333969e-05, + "loss": 1.0528, + "step": 214050 + }, + { + "epoch": 18.53, + "learning_rate": 3.1498743827427876e-05, + "loss": 0.9943, + "step": 214060 + }, + { + "epoch": 18.53, + "learning_rate": 3.149787750151607e-05, + "loss": 1.0141, + "step": 214070 + }, + { + "epoch": 18.53, + "learning_rate": 3.1497011175604265e-05, + "loss": 0.9942, + "step": 214080 + }, + { + "epoch": 18.53, + "learning_rate": 3.149614484969245e-05, + "loss": 0.9357, + "step": 214090 + }, + { + "epoch": 18.53, + "learning_rate": 3.149527852378065e-05, + "loss": 0.994, + "step": 214100 + }, + { + "epoch": 18.53, + "learning_rate": 3.149441219786884e-05, + "loss": 0.9778, + "step": 214110 + }, + { + "epoch": 18.53, + "learning_rate": 3.149354587195703e-05, + "loss": 0.9258, + "step": 214120 + }, + { + "epoch": 18.53, + "learning_rate": 3.1492679546045224e-05, + "loss": 1.0673, + "step": 214130 + }, + { + "epoch": 18.53, + "learning_rate": 3.149181322013341e-05, + "loss": 0.9982, + "step": 214140 + }, + { + "epoch": 18.54, + "learning_rate": 3.1490946894221606e-05, + "loss": 0.9854, + "step": 214150 + }, + { + "epoch": 18.54, + "learning_rate": 3.14900805683098e-05, + "loss": 0.9378, + "step": 214160 + }, + { + "epoch": 18.54, + "learning_rate": 3.148921424239799e-05, + "loss": 0.9862, + "step": 214170 + }, + { + "epoch": 18.54, + "learning_rate": 3.148834791648618e-05, + "loss": 1.0001, + "step": 214180 + }, + { + "epoch": 18.54, + "learning_rate": 3.148748159057438e-05, + "loss": 0.9827, + "step": 214190 + }, + { + "epoch": 18.54, + "learning_rate": 3.1486615264662565e-05, + "loss": 0.9844, + "step": 214200 + }, + { + "epoch": 18.54, + "learning_rate": 3.148574893875076e-05, + "loss": 0.9212, + "step": 214210 + }, + { + "epoch": 18.54, + "learning_rate": 3.1484882612838954e-05, + "loss": 1.0073, + "step": 214220 + }, + { + "epoch": 18.54, + "learning_rate": 3.148401628692714e-05, + "loss": 1.0113, + "step": 214230 + }, + { + "epoch": 18.54, + "learning_rate": 3.1483149961015336e-05, + "loss": 0.9921, + "step": 214240 + }, + { + "epoch": 18.54, + "learning_rate": 3.1482283635103524e-05, + "loss": 1.0088, + "step": 214250 + }, + { + "epoch": 18.55, + "learning_rate": 3.148141730919172e-05, + "loss": 1.0392, + "step": 214260 + }, + { + "epoch": 18.55, + "learning_rate": 3.148055098327991e-05, + "loss": 0.9887, + "step": 214270 + }, + { + "epoch": 18.55, + "learning_rate": 3.14796846573681e-05, + "loss": 0.9848, + "step": 214280 + }, + { + "epoch": 18.55, + "learning_rate": 3.1478818331456295e-05, + "loss": 0.9768, + "step": 214290 + }, + { + "epoch": 18.55, + "learning_rate": 3.147795200554449e-05, + "loss": 0.9762, + "step": 214300 + }, + { + "epoch": 18.55, + "learning_rate": 3.147708567963268e-05, + "loss": 0.9891, + "step": 214310 + }, + { + "epoch": 18.55, + "learning_rate": 3.147621935372087e-05, + "loss": 1.017, + "step": 214320 + }, + { + "epoch": 18.55, + "learning_rate": 3.1475353027809066e-05, + "loss": 0.9736, + "step": 214330 + }, + { + "epoch": 18.55, + "learning_rate": 3.1474486701897253e-05, + "loss": 0.9544, + "step": 214340 + }, + { + "epoch": 18.55, + "learning_rate": 3.147362037598545e-05, + "loss": 0.9694, + "step": 214350 + }, + { + "epoch": 18.55, + "learning_rate": 3.1472754050073636e-05, + "loss": 0.9451, + "step": 214360 + }, + { + "epoch": 18.55, + "learning_rate": 3.147188772416183e-05, + "loss": 1.0091, + "step": 214370 + }, + { + "epoch": 18.56, + "learning_rate": 3.1471021398250024e-05, + "loss": 0.9329, + "step": 214380 + }, + { + "epoch": 18.56, + "learning_rate": 3.147015507233821e-05, + "loss": 0.9326, + "step": 214390 + }, + { + "epoch": 18.56, + "learning_rate": 3.1469288746426407e-05, + "loss": 1.0182, + "step": 214400 + }, + { + "epoch": 18.56, + "learning_rate": 3.14684224205146e-05, + "loss": 1.0214, + "step": 214410 + }, + { + "epoch": 18.56, + "learning_rate": 3.146755609460279e-05, + "loss": 0.9849, + "step": 214420 + }, + { + "epoch": 18.56, + "learning_rate": 3.146668976869098e-05, + "loss": 0.9833, + "step": 214430 + }, + { + "epoch": 18.56, + "learning_rate": 3.146582344277918e-05, + "loss": 1.035, + "step": 214440 + }, + { + "epoch": 18.56, + "learning_rate": 3.1464957116867365e-05, + "loss": 0.9551, + "step": 214450 + }, + { + "epoch": 18.56, + "learning_rate": 3.146409079095556e-05, + "loss": 0.9673, + "step": 214460 + }, + { + "epoch": 18.56, + "learning_rate": 3.146322446504375e-05, + "loss": 0.9389, + "step": 214470 + }, + { + "epoch": 18.56, + "learning_rate": 3.146235813913194e-05, + "loss": 0.9728, + "step": 214480 + }, + { + "epoch": 18.57, + "learning_rate": 3.1461491813220136e-05, + "loss": 0.966, + "step": 214490 + }, + { + "epoch": 18.57, + "learning_rate": 3.1460625487308324e-05, + "loss": 0.9843, + "step": 214500 + }, + { + "epoch": 18.57, + "learning_rate": 3.145975916139652e-05, + "loss": 1.0017, + "step": 214510 + }, + { + "epoch": 18.57, + "learning_rate": 3.145889283548471e-05, + "loss": 1.0016, + "step": 214520 + }, + { + "epoch": 18.57, + "learning_rate": 3.14580265095729e-05, + "loss": 0.9656, + "step": 214530 + }, + { + "epoch": 18.57, + "learning_rate": 3.1457160183661095e-05, + "loss": 0.9777, + "step": 214540 + }, + { + "epoch": 18.57, + "learning_rate": 3.145629385774928e-05, + "loss": 0.9917, + "step": 214550 + }, + { + "epoch": 18.57, + "learning_rate": 3.145542753183748e-05, + "loss": 0.9859, + "step": 214560 + }, + { + "epoch": 18.57, + "learning_rate": 3.145456120592567e-05, + "loss": 1.0267, + "step": 214570 + }, + { + "epoch": 18.57, + "learning_rate": 3.145369488001386e-05, + "loss": 0.9407, + "step": 214580 + }, + { + "epoch": 18.57, + "learning_rate": 3.1452828554102054e-05, + "loss": 0.9729, + "step": 214590 + }, + { + "epoch": 18.57, + "learning_rate": 3.145196222819025e-05, + "loss": 0.9578, + "step": 214600 + }, + { + "epoch": 18.58, + "learning_rate": 3.1451095902278436e-05, + "loss": 1.0121, + "step": 214610 + }, + { + "epoch": 18.58, + "learning_rate": 3.145022957636663e-05, + "loss": 0.9826, + "step": 214620 + }, + { + "epoch": 18.58, + "learning_rate": 3.1449363250454825e-05, + "loss": 0.9987, + "step": 214630 + }, + { + "epoch": 18.58, + "learning_rate": 3.144849692454301e-05, + "loss": 0.9545, + "step": 214640 + }, + { + "epoch": 18.58, + "learning_rate": 3.144763059863121e-05, + "loss": 0.9103, + "step": 214650 + }, + { + "epoch": 18.58, + "learning_rate": 3.1446764272719395e-05, + "loss": 0.9486, + "step": 214660 + }, + { + "epoch": 18.58, + "learning_rate": 3.144589794680759e-05, + "loss": 0.9696, + "step": 214670 + }, + { + "epoch": 18.58, + "learning_rate": 3.1445031620895783e-05, + "loss": 0.9698, + "step": 214680 + }, + { + "epoch": 18.58, + "learning_rate": 3.144416529498397e-05, + "loss": 1.0256, + "step": 214690 + }, + { + "epoch": 18.58, + "learning_rate": 3.1443298969072166e-05, + "loss": 1.0358, + "step": 214700 + }, + { + "epoch": 18.58, + "learning_rate": 3.144243264316036e-05, + "loss": 1.0163, + "step": 214710 + }, + { + "epoch": 18.59, + "learning_rate": 3.144156631724855e-05, + "loss": 0.9337, + "step": 214720 + }, + { + "epoch": 18.59, + "learning_rate": 3.144069999133674e-05, + "loss": 1.0475, + "step": 214730 + }, + { + "epoch": 18.59, + "learning_rate": 3.143983366542494e-05, + "loss": 0.9694, + "step": 214740 + }, + { + "epoch": 18.59, + "learning_rate": 3.1438967339513124e-05, + "loss": 0.934, + "step": 214750 + }, + { + "epoch": 18.59, + "learning_rate": 3.143810101360132e-05, + "loss": 1.0166, + "step": 214760 + }, + { + "epoch": 18.59, + "learning_rate": 3.1437234687689506e-05, + "loss": 0.9827, + "step": 214770 + }, + { + "epoch": 18.59, + "learning_rate": 3.14363683617777e-05, + "loss": 0.9864, + "step": 214780 + }, + { + "epoch": 18.59, + "learning_rate": 3.1435502035865895e-05, + "loss": 0.9528, + "step": 214790 + }, + { + "epoch": 18.59, + "learning_rate": 3.143463570995408e-05, + "loss": 1.0519, + "step": 214800 + }, + { + "epoch": 18.59, + "learning_rate": 3.143376938404228e-05, + "loss": 0.9263, + "step": 214810 + }, + { + "epoch": 18.59, + "learning_rate": 3.143290305813047e-05, + "loss": 0.9553, + "step": 214820 + }, + { + "epoch": 18.59, + "learning_rate": 3.143203673221866e-05, + "loss": 0.9485, + "step": 214830 + }, + { + "epoch": 18.6, + "learning_rate": 3.1431170406306854e-05, + "loss": 0.9579, + "step": 214840 + }, + { + "epoch": 18.6, + "learning_rate": 3.143030408039505e-05, + "loss": 0.9546, + "step": 214850 + }, + { + "epoch": 18.6, + "learning_rate": 3.1429437754483236e-05, + "loss": 1.0394, + "step": 214860 + }, + { + "epoch": 18.6, + "learning_rate": 3.142857142857143e-05, + "loss": 1.0152, + "step": 214870 + }, + { + "epoch": 18.6, + "learning_rate": 3.142770510265962e-05, + "loss": 0.9937, + "step": 214880 + }, + { + "epoch": 18.6, + "learning_rate": 3.142683877674781e-05, + "loss": 0.9453, + "step": 214890 + }, + { + "epoch": 18.6, + "learning_rate": 3.142597245083601e-05, + "loss": 1.056, + "step": 214900 + }, + { + "epoch": 18.6, + "learning_rate": 3.1425106124924195e-05, + "loss": 0.9908, + "step": 214910 + }, + { + "epoch": 18.6, + "learning_rate": 3.142423979901239e-05, + "loss": 1.0314, + "step": 214920 + }, + { + "epoch": 18.6, + "learning_rate": 3.1423373473100584e-05, + "loss": 0.9867, + "step": 214930 + }, + { + "epoch": 18.6, + "learning_rate": 3.142250714718877e-05, + "loss": 1.0466, + "step": 214940 + }, + { + "epoch": 18.61, + "learning_rate": 3.1421640821276966e-05, + "loss": 1.0258, + "step": 214950 + }, + { + "epoch": 18.61, + "learning_rate": 3.142077449536516e-05, + "loss": 1.0191, + "step": 214960 + }, + { + "epoch": 18.61, + "learning_rate": 3.141990816945335e-05, + "loss": 0.967, + "step": 214970 + }, + { + "epoch": 18.61, + "learning_rate": 3.141904184354154e-05, + "loss": 0.944, + "step": 214980 + }, + { + "epoch": 18.61, + "learning_rate": 3.141817551762973e-05, + "loss": 0.9377, + "step": 214990 + }, + { + "epoch": 18.61, + "learning_rate": 3.1417309191717925e-05, + "loss": 0.9419, + "step": 215000 + }, + { + "epoch": 18.61, + "learning_rate": 3.141644286580612e-05, + "loss": 1.0418, + "step": 215010 + }, + { + "epoch": 18.61, + "learning_rate": 3.141557653989431e-05, + "loss": 0.9421, + "step": 215020 + }, + { + "epoch": 18.61, + "learning_rate": 3.14147102139825e-05, + "loss": 1.0544, + "step": 215030 + }, + { + "epoch": 18.61, + "learning_rate": 3.1413843888070696e-05, + "loss": 0.9538, + "step": 215040 + }, + { + "epoch": 18.61, + "learning_rate": 3.141297756215888e-05, + "loss": 1.031, + "step": 215050 + }, + { + "epoch": 18.61, + "learning_rate": 3.141211123624708e-05, + "loss": 0.9741, + "step": 215060 + }, + { + "epoch": 18.62, + "learning_rate": 3.141124491033527e-05, + "loss": 0.9877, + "step": 215070 + }, + { + "epoch": 18.62, + "learning_rate": 3.141037858442346e-05, + "loss": 0.9927, + "step": 215080 + }, + { + "epoch": 18.62, + "learning_rate": 3.1409512258511654e-05, + "loss": 0.991, + "step": 215090 + }, + { + "epoch": 18.62, + "learning_rate": 3.140864593259984e-05, + "loss": 0.9458, + "step": 215100 + }, + { + "epoch": 18.62, + "learning_rate": 3.1407779606688037e-05, + "loss": 0.9632, + "step": 215110 + }, + { + "epoch": 18.62, + "learning_rate": 3.140691328077623e-05, + "loss": 1.009, + "step": 215120 + }, + { + "epoch": 18.62, + "learning_rate": 3.140604695486442e-05, + "loss": 0.964, + "step": 215130 + }, + { + "epoch": 18.62, + "learning_rate": 3.140518062895261e-05, + "loss": 1.0297, + "step": 215140 + }, + { + "epoch": 18.62, + "learning_rate": 3.140431430304081e-05, + "loss": 0.9118, + "step": 215150 + }, + { + "epoch": 18.62, + "learning_rate": 3.1403447977128995e-05, + "loss": 0.922, + "step": 215160 + }, + { + "epoch": 18.62, + "learning_rate": 3.140258165121719e-05, + "loss": 0.9811, + "step": 215170 + }, + { + "epoch": 18.62, + "learning_rate": 3.1401715325305384e-05, + "loss": 0.9695, + "step": 215180 + }, + { + "epoch": 18.63, + "learning_rate": 3.140084899939357e-05, + "loss": 0.9979, + "step": 215190 + }, + { + "epoch": 18.63, + "learning_rate": 3.1399982673481766e-05, + "loss": 0.9826, + "step": 215200 + }, + { + "epoch": 18.63, + "learning_rate": 3.1399116347569954e-05, + "loss": 0.9629, + "step": 215210 + }, + { + "epoch": 18.63, + "learning_rate": 3.139825002165815e-05, + "loss": 1.0074, + "step": 215220 + }, + { + "epoch": 18.63, + "learning_rate": 3.139738369574634e-05, + "loss": 0.9844, + "step": 215230 + }, + { + "epoch": 18.63, + "learning_rate": 3.139651736983453e-05, + "loss": 0.9455, + "step": 215240 + }, + { + "epoch": 18.63, + "learning_rate": 3.1395651043922725e-05, + "loss": 1.0223, + "step": 215250 + }, + { + "epoch": 18.63, + "learning_rate": 3.139478471801092e-05, + "loss": 1.0461, + "step": 215260 + }, + { + "epoch": 18.63, + "learning_rate": 3.139391839209911e-05, + "loss": 0.9788, + "step": 215270 + }, + { + "epoch": 18.63, + "learning_rate": 3.13930520661873e-05, + "loss": 0.9856, + "step": 215280 + }, + { + "epoch": 18.63, + "learning_rate": 3.139218574027549e-05, + "loss": 0.9801, + "step": 215290 + }, + { + "epoch": 18.64, + "learning_rate": 3.1391319414363684e-05, + "loss": 0.9405, + "step": 215300 + }, + { + "epoch": 18.64, + "learning_rate": 3.139045308845188e-05, + "loss": 0.9706, + "step": 215310 + }, + { + "epoch": 18.64, + "learning_rate": 3.1389586762540066e-05, + "loss": 0.9739, + "step": 215320 + }, + { + "epoch": 18.64, + "learning_rate": 3.138872043662826e-05, + "loss": 0.9761, + "step": 215330 + }, + { + "epoch": 18.64, + "learning_rate": 3.1387854110716455e-05, + "loss": 0.9872, + "step": 215340 + }, + { + "epoch": 18.64, + "learning_rate": 3.138698778480464e-05, + "loss": 0.9353, + "step": 215350 + }, + { + "epoch": 18.64, + "learning_rate": 3.138612145889284e-05, + "loss": 0.9765, + "step": 215360 + }, + { + "epoch": 18.64, + "learning_rate": 3.138525513298103e-05, + "loss": 0.91, + "step": 215370 + }, + { + "epoch": 18.64, + "learning_rate": 3.138438880706922e-05, + "loss": 1.0069, + "step": 215380 + }, + { + "epoch": 18.64, + "learning_rate": 3.1383522481157413e-05, + "loss": 0.9976, + "step": 215390 + }, + { + "epoch": 18.64, + "learning_rate": 3.13826561552456e-05, + "loss": 0.9395, + "step": 215400 + }, + { + "epoch": 18.64, + "learning_rate": 3.1381789829333796e-05, + "loss": 0.9885, + "step": 215410 + }, + { + "epoch": 18.65, + "learning_rate": 3.138092350342199e-05, + "loss": 1.0033, + "step": 215420 + }, + { + "epoch": 18.65, + "learning_rate": 3.138005717751018e-05, + "loss": 0.9543, + "step": 215430 + }, + { + "epoch": 18.65, + "learning_rate": 3.137919085159837e-05, + "loss": 0.9679, + "step": 215440 + }, + { + "epoch": 18.65, + "learning_rate": 3.137832452568657e-05, + "loss": 0.974, + "step": 215450 + }, + { + "epoch": 18.65, + "learning_rate": 3.1377458199774754e-05, + "loss": 1.0238, + "step": 215460 + }, + { + "epoch": 18.65, + "learning_rate": 3.137659187386295e-05, + "loss": 1.0033, + "step": 215470 + }, + { + "epoch": 18.65, + "learning_rate": 3.137572554795114e-05, + "loss": 1.0249, + "step": 215480 + }, + { + "epoch": 18.65, + "learning_rate": 3.137485922203933e-05, + "loss": 0.9461, + "step": 215490 + }, + { + "epoch": 18.65, + "learning_rate": 3.1373992896127525e-05, + "loss": 0.9968, + "step": 215500 + }, + { + "epoch": 18.65, + "learning_rate": 3.137312657021571e-05, + "loss": 0.9586, + "step": 215510 + }, + { + "epoch": 18.65, + "learning_rate": 3.137226024430391e-05, + "loss": 0.9677, + "step": 215520 + }, + { + "epoch": 18.66, + "learning_rate": 3.13713939183921e-05, + "loss": 0.9421, + "step": 215530 + }, + { + "epoch": 18.66, + "learning_rate": 3.137052759248029e-05, + "loss": 0.9816, + "step": 215540 + }, + { + "epoch": 18.66, + "learning_rate": 3.1369661266568484e-05, + "loss": 0.9945, + "step": 215550 + }, + { + "epoch": 18.66, + "learning_rate": 3.136879494065668e-05, + "loss": 1.0183, + "step": 215560 + }, + { + "epoch": 18.66, + "learning_rate": 3.1367928614744866e-05, + "loss": 0.9867, + "step": 215570 + }, + { + "epoch": 18.66, + "learning_rate": 3.136706228883306e-05, + "loss": 1.0156, + "step": 215580 + }, + { + "epoch": 18.66, + "learning_rate": 3.1366195962921255e-05, + "loss": 0.9545, + "step": 215590 + }, + { + "epoch": 18.66, + "learning_rate": 3.136532963700944e-05, + "loss": 1.0122, + "step": 215600 + }, + { + "epoch": 18.66, + "learning_rate": 3.136446331109764e-05, + "loss": 1.0371, + "step": 215610 + }, + { + "epoch": 18.66, + "learning_rate": 3.1363596985185825e-05, + "loss": 0.9643, + "step": 215620 + }, + { + "epoch": 18.66, + "learning_rate": 3.136273065927402e-05, + "loss": 0.9882, + "step": 215630 + }, + { + "epoch": 18.66, + "learning_rate": 3.1361864333362214e-05, + "loss": 1.0012, + "step": 215640 + }, + { + "epoch": 18.67, + "learning_rate": 3.13609980074504e-05, + "loss": 0.9237, + "step": 215650 + }, + { + "epoch": 18.67, + "learning_rate": 3.1360131681538596e-05, + "loss": 0.9886, + "step": 215660 + }, + { + "epoch": 18.67, + "learning_rate": 3.135926535562679e-05, + "loss": 0.9531, + "step": 215670 + }, + { + "epoch": 18.67, + "learning_rate": 3.135839902971498e-05, + "loss": 0.9934, + "step": 215680 + }, + { + "epoch": 18.67, + "learning_rate": 3.135753270380317e-05, + "loss": 0.9837, + "step": 215690 + }, + { + "epoch": 18.67, + "learning_rate": 3.135666637789137e-05, + "loss": 1.0101, + "step": 215700 + }, + { + "epoch": 18.67, + "learning_rate": 3.1355800051979555e-05, + "loss": 1.0031, + "step": 215710 + }, + { + "epoch": 18.67, + "learning_rate": 3.135493372606775e-05, + "loss": 0.9711, + "step": 215720 + }, + { + "epoch": 18.67, + "learning_rate": 3.135406740015594e-05, + "loss": 1.0466, + "step": 215730 + }, + { + "epoch": 18.67, + "learning_rate": 3.135320107424413e-05, + "loss": 0.9673, + "step": 215740 + }, + { + "epoch": 18.67, + "learning_rate": 3.1352334748332326e-05, + "loss": 0.9445, + "step": 215750 + }, + { + "epoch": 18.68, + "learning_rate": 3.135146842242051e-05, + "loss": 1.0369, + "step": 215760 + }, + { + "epoch": 18.68, + "learning_rate": 3.135060209650871e-05, + "loss": 0.9177, + "step": 215770 + }, + { + "epoch": 18.68, + "learning_rate": 3.13497357705969e-05, + "loss": 0.9685, + "step": 215780 + }, + { + "epoch": 18.68, + "learning_rate": 3.134886944468509e-05, + "loss": 0.9304, + "step": 215790 + }, + { + "epoch": 18.68, + "learning_rate": 3.1348003118773284e-05, + "loss": 0.9265, + "step": 215800 + }, + { + "epoch": 18.68, + "learning_rate": 3.134713679286148e-05, + "loss": 1.0055, + "step": 215810 + }, + { + "epoch": 18.68, + "learning_rate": 3.1346270466949667e-05, + "loss": 0.9975, + "step": 215820 + }, + { + "epoch": 18.68, + "learning_rate": 3.134540414103786e-05, + "loss": 1.0124, + "step": 215830 + }, + { + "epoch": 18.68, + "learning_rate": 3.134453781512605e-05, + "loss": 1.0159, + "step": 215840 + }, + { + "epoch": 18.68, + "learning_rate": 3.134367148921424e-05, + "loss": 1.0024, + "step": 215850 + }, + { + "epoch": 18.68, + "learning_rate": 3.134280516330244e-05, + "loss": 0.9788, + "step": 215860 + }, + { + "epoch": 18.68, + "learning_rate": 3.1341938837390625e-05, + "loss": 1.0123, + "step": 215870 + }, + { + "epoch": 18.69, + "learning_rate": 3.134107251147882e-05, + "loss": 0.9484, + "step": 215880 + }, + { + "epoch": 18.69, + "learning_rate": 3.1340206185567014e-05, + "loss": 0.9911, + "step": 215890 + }, + { + "epoch": 18.69, + "learning_rate": 3.13393398596552e-05, + "loss": 0.9917, + "step": 215900 + }, + { + "epoch": 18.69, + "learning_rate": 3.1338473533743396e-05, + "loss": 0.9984, + "step": 215910 + }, + { + "epoch": 18.69, + "learning_rate": 3.133760720783159e-05, + "loss": 0.9365, + "step": 215920 + }, + { + "epoch": 18.69, + "learning_rate": 3.133674088191978e-05, + "loss": 0.9011, + "step": 215930 + }, + { + "epoch": 18.69, + "learning_rate": 3.133587455600797e-05, + "loss": 1.0101, + "step": 215940 + }, + { + "epoch": 18.69, + "learning_rate": 3.133500823009616e-05, + "loss": 1.0221, + "step": 215950 + }, + { + "epoch": 18.69, + "learning_rate": 3.1334141904184355e-05, + "loss": 0.9722, + "step": 215960 + }, + { + "epoch": 18.69, + "learning_rate": 3.133327557827255e-05, + "loss": 0.9966, + "step": 215970 + }, + { + "epoch": 18.69, + "learning_rate": 3.133240925236074e-05, + "loss": 0.909, + "step": 215980 + }, + { + "epoch": 18.7, + "learning_rate": 3.133154292644893e-05, + "loss": 0.9565, + "step": 215990 + }, + { + "epoch": 18.7, + "learning_rate": 3.1330676600537126e-05, + "loss": 0.9803, + "step": 216000 + }, + { + "epoch": 18.7, + "learning_rate": 3.1329810274625314e-05, + "loss": 0.9945, + "step": 216010 + }, + { + "epoch": 18.7, + "learning_rate": 3.132894394871351e-05, + "loss": 0.9083, + "step": 216020 + }, + { + "epoch": 18.7, + "learning_rate": 3.13280776228017e-05, + "loss": 0.9399, + "step": 216030 + }, + { + "epoch": 18.7, + "learning_rate": 3.132721129688989e-05, + "loss": 0.9356, + "step": 216040 + }, + { + "epoch": 18.7, + "learning_rate": 3.1326344970978085e-05, + "loss": 0.9741, + "step": 216050 + }, + { + "epoch": 18.7, + "learning_rate": 3.132547864506627e-05, + "loss": 0.9937, + "step": 216060 + }, + { + "epoch": 18.7, + "learning_rate": 3.132461231915447e-05, + "loss": 0.9916, + "step": 216070 + }, + { + "epoch": 18.7, + "learning_rate": 3.132374599324266e-05, + "loss": 0.9955, + "step": 216080 + }, + { + "epoch": 18.7, + "learning_rate": 3.132287966733085e-05, + "loss": 0.994, + "step": 216090 + }, + { + "epoch": 18.7, + "learning_rate": 3.1322013341419043e-05, + "loss": 0.9483, + "step": 216100 + }, + { + "epoch": 18.71, + "learning_rate": 3.132114701550724e-05, + "loss": 1.0474, + "step": 216110 + }, + { + "epoch": 18.71, + "learning_rate": 3.1320280689595426e-05, + "loss": 0.9535, + "step": 216120 + }, + { + "epoch": 18.71, + "learning_rate": 3.131941436368362e-05, + "loss": 0.991, + "step": 216130 + }, + { + "epoch": 18.71, + "learning_rate": 3.131854803777181e-05, + "loss": 0.9955, + "step": 216140 + }, + { + "epoch": 18.71, + "learning_rate": 3.131768171186e-05, + "loss": 0.9979, + "step": 216150 + }, + { + "epoch": 18.71, + "learning_rate": 3.1316815385948197e-05, + "loss": 1.0054, + "step": 216160 + }, + { + "epoch": 18.71, + "learning_rate": 3.1315949060036384e-05, + "loss": 0.9721, + "step": 216170 + }, + { + "epoch": 18.71, + "learning_rate": 3.131508273412458e-05, + "loss": 0.9323, + "step": 216180 + }, + { + "epoch": 18.71, + "learning_rate": 3.131421640821277e-05, + "loss": 0.9751, + "step": 216190 + }, + { + "epoch": 18.71, + "learning_rate": 3.131335008230096e-05, + "loss": 0.9998, + "step": 216200 + }, + { + "epoch": 18.71, + "learning_rate": 3.1312483756389155e-05, + "loss": 0.9645, + "step": 216210 + }, + { + "epoch": 18.71, + "learning_rate": 3.131161743047735e-05, + "loss": 0.9946, + "step": 216220 + }, + { + "epoch": 18.72, + "learning_rate": 3.131075110456554e-05, + "loss": 1.0046, + "step": 216230 + }, + { + "epoch": 18.72, + "learning_rate": 3.130988477865373e-05, + "loss": 1.0036, + "step": 216240 + }, + { + "epoch": 18.72, + "learning_rate": 3.130901845274192e-05, + "loss": 0.9794, + "step": 216250 + }, + { + "epoch": 18.72, + "learning_rate": 3.1308152126830114e-05, + "loss": 0.9379, + "step": 216260 + }, + { + "epoch": 18.72, + "learning_rate": 3.130728580091831e-05, + "loss": 0.9655, + "step": 216270 + }, + { + "epoch": 18.72, + "learning_rate": 3.1306419475006496e-05, + "loss": 0.9654, + "step": 216280 + }, + { + "epoch": 18.72, + "learning_rate": 3.130555314909469e-05, + "loss": 0.9844, + "step": 216290 + }, + { + "epoch": 18.72, + "learning_rate": 3.1304686823182885e-05, + "loss": 0.9708, + "step": 216300 + }, + { + "epoch": 18.72, + "learning_rate": 3.130382049727107e-05, + "loss": 0.9688, + "step": 216310 + }, + { + "epoch": 18.72, + "learning_rate": 3.130295417135927e-05, + "loss": 1.0263, + "step": 216320 + }, + { + "epoch": 18.72, + "learning_rate": 3.130208784544746e-05, + "loss": 0.9939, + "step": 216330 + }, + { + "epoch": 18.73, + "learning_rate": 3.130122151953565e-05, + "loss": 1.0339, + "step": 216340 + }, + { + "epoch": 18.73, + "learning_rate": 3.1300355193623844e-05, + "loss": 0.9543, + "step": 216350 + }, + { + "epoch": 18.73, + "learning_rate": 3.129948886771203e-05, + "loss": 1.0146, + "step": 216360 + }, + { + "epoch": 18.73, + "learning_rate": 3.1298622541800226e-05, + "loss": 1.0147, + "step": 216370 + }, + { + "epoch": 18.73, + "learning_rate": 3.129775621588842e-05, + "loss": 0.9979, + "step": 216380 + }, + { + "epoch": 18.73, + "learning_rate": 3.129688988997661e-05, + "loss": 1.0151, + "step": 216390 + }, + { + "epoch": 18.73, + "learning_rate": 3.12960235640648e-05, + "loss": 0.9948, + "step": 216400 + }, + { + "epoch": 18.73, + "learning_rate": 3.1295157238153e-05, + "loss": 0.9759, + "step": 216410 + }, + { + "epoch": 18.73, + "learning_rate": 3.1294290912241185e-05, + "loss": 0.9963, + "step": 216420 + }, + { + "epoch": 18.73, + "learning_rate": 3.129342458632938e-05, + "loss": 1.0149, + "step": 216430 + }, + { + "epoch": 18.73, + "learning_rate": 3.1292558260417574e-05, + "loss": 0.9288, + "step": 216440 + }, + { + "epoch": 18.73, + "learning_rate": 3.129169193450576e-05, + "loss": 1.0442, + "step": 216450 + }, + { + "epoch": 18.74, + "learning_rate": 3.1290825608593956e-05, + "loss": 0.9445, + "step": 216460 + }, + { + "epoch": 18.74, + "learning_rate": 3.128995928268214e-05, + "loss": 1.0419, + "step": 216470 + }, + { + "epoch": 18.74, + "learning_rate": 3.128909295677034e-05, + "loss": 0.9594, + "step": 216480 + }, + { + "epoch": 18.74, + "learning_rate": 3.128822663085853e-05, + "loss": 0.9969, + "step": 216490 + }, + { + "epoch": 18.74, + "learning_rate": 3.128736030494672e-05, + "loss": 1.0536, + "step": 216500 + }, + { + "epoch": 18.74, + "learning_rate": 3.1286493979034914e-05, + "loss": 1.0072, + "step": 216510 + }, + { + "epoch": 18.74, + "learning_rate": 3.128562765312311e-05, + "loss": 1.0286, + "step": 216520 + }, + { + "epoch": 18.74, + "learning_rate": 3.1284761327211296e-05, + "loss": 1.0252, + "step": 216530 + }, + { + "epoch": 18.74, + "learning_rate": 3.128389500129949e-05, + "loss": 0.9668, + "step": 216540 + }, + { + "epoch": 18.74, + "learning_rate": 3.1283028675387685e-05, + "loss": 0.9709, + "step": 216550 + }, + { + "epoch": 18.74, + "learning_rate": 3.128216234947587e-05, + "loss": 1.0001, + "step": 216560 + }, + { + "epoch": 18.75, + "learning_rate": 3.128129602356407e-05, + "loss": 0.9923, + "step": 216570 + }, + { + "epoch": 18.75, + "learning_rate": 3.1280429697652255e-05, + "loss": 0.9768, + "step": 216580 + }, + { + "epoch": 18.75, + "learning_rate": 3.127956337174045e-05, + "loss": 0.9551, + "step": 216590 + }, + { + "epoch": 18.75, + "learning_rate": 3.1278697045828644e-05, + "loss": 1.0151, + "step": 216600 + }, + { + "epoch": 18.75, + "learning_rate": 3.127783071991683e-05, + "loss": 1.0242, + "step": 216610 + }, + { + "epoch": 18.75, + "learning_rate": 3.1276964394005026e-05, + "loss": 1.0034, + "step": 216620 + }, + { + "epoch": 18.75, + "learning_rate": 3.127609806809322e-05, + "loss": 1.0216, + "step": 216630 + }, + { + "epoch": 18.75, + "learning_rate": 3.127523174218141e-05, + "loss": 0.9566, + "step": 216640 + }, + { + "epoch": 18.75, + "learning_rate": 3.12743654162696e-05, + "loss": 1.0019, + "step": 216650 + }, + { + "epoch": 18.75, + "learning_rate": 3.12734990903578e-05, + "loss": 0.9276, + "step": 216660 + }, + { + "epoch": 18.75, + "learning_rate": 3.1272632764445985e-05, + "loss": 0.9467, + "step": 216670 + }, + { + "epoch": 18.75, + "learning_rate": 3.127176643853418e-05, + "loss": 0.9858, + "step": 216680 + }, + { + "epoch": 18.76, + "learning_rate": 3.127090011262237e-05, + "loss": 0.9792, + "step": 216690 + }, + { + "epoch": 18.76, + "learning_rate": 3.127003378671056e-05, + "loss": 1.0001, + "step": 216700 + }, + { + "epoch": 18.76, + "learning_rate": 3.1269167460798756e-05, + "loss": 0.9296, + "step": 216710 + }, + { + "epoch": 18.76, + "learning_rate": 3.1268301134886944e-05, + "loss": 0.965, + "step": 216720 + }, + { + "epoch": 18.76, + "learning_rate": 3.126743480897514e-05, + "loss": 0.96, + "step": 216730 + }, + { + "epoch": 18.76, + "learning_rate": 3.126656848306333e-05, + "loss": 1.028, + "step": 216740 + }, + { + "epoch": 18.76, + "learning_rate": 3.126570215715152e-05, + "loss": 0.9944, + "step": 216750 + }, + { + "epoch": 18.76, + "learning_rate": 3.1264835831239715e-05, + "loss": 0.9705, + "step": 216760 + }, + { + "epoch": 18.76, + "learning_rate": 3.126396950532791e-05, + "loss": 1.0019, + "step": 216770 + }, + { + "epoch": 18.76, + "learning_rate": 3.12631031794161e-05, + "loss": 0.9473, + "step": 216780 + }, + { + "epoch": 18.76, + "learning_rate": 3.126223685350429e-05, + "loss": 1.0251, + "step": 216790 + }, + { + "epoch": 18.77, + "learning_rate": 3.126137052759248e-05, + "loss": 0.954, + "step": 216800 + }, + { + "epoch": 18.77, + "learning_rate": 3.1260504201680673e-05, + "loss": 0.9917, + "step": 216810 + }, + { + "epoch": 18.77, + "learning_rate": 3.125963787576887e-05, + "loss": 1.0094, + "step": 216820 + }, + { + "epoch": 18.77, + "learning_rate": 3.1258771549857056e-05, + "loss": 0.9449, + "step": 216830 + }, + { + "epoch": 18.77, + "learning_rate": 3.125790522394525e-05, + "loss": 0.9155, + "step": 216840 + }, + { + "epoch": 18.77, + "learning_rate": 3.1257038898033444e-05, + "loss": 0.9988, + "step": 216850 + }, + { + "epoch": 18.77, + "learning_rate": 3.125617257212163e-05, + "loss": 0.9978, + "step": 216860 + }, + { + "epoch": 18.77, + "learning_rate": 3.1255306246209827e-05, + "loss": 0.9901, + "step": 216870 + }, + { + "epoch": 18.77, + "learning_rate": 3.1254439920298014e-05, + "loss": 1.0519, + "step": 216880 + }, + { + "epoch": 18.77, + "learning_rate": 3.125357359438621e-05, + "loss": 1.032, + "step": 216890 + }, + { + "epoch": 18.77, + "learning_rate": 3.12527072684744e-05, + "loss": 0.9204, + "step": 216900 + }, + { + "epoch": 18.77, + "learning_rate": 3.125184094256259e-05, + "loss": 0.9655, + "step": 216910 + }, + { + "epoch": 18.78, + "learning_rate": 3.1250974616650785e-05, + "loss": 0.9762, + "step": 216920 + }, + { + "epoch": 18.78, + "learning_rate": 3.125010829073898e-05, + "loss": 0.9546, + "step": 216930 + }, + { + "epoch": 18.78, + "learning_rate": 3.124924196482717e-05, + "loss": 0.9644, + "step": 216940 + }, + { + "epoch": 18.78, + "learning_rate": 3.124837563891536e-05, + "loss": 0.9606, + "step": 216950 + }, + { + "epoch": 18.78, + "learning_rate": 3.1247509313003556e-05, + "loss": 0.9749, + "step": 216960 + }, + { + "epoch": 18.78, + "learning_rate": 3.1246642987091744e-05, + "loss": 0.9654, + "step": 216970 + }, + { + "epoch": 18.78, + "learning_rate": 3.124577666117994e-05, + "loss": 0.9883, + "step": 216980 + }, + { + "epoch": 18.78, + "learning_rate": 3.1244910335268126e-05, + "loss": 1.0156, + "step": 216990 + }, + { + "epoch": 18.78, + "learning_rate": 3.124404400935632e-05, + "loss": 0.9966, + "step": 217000 + }, + { + "epoch": 18.78, + "learning_rate": 3.1243177683444515e-05, + "loss": 0.9878, + "step": 217010 + }, + { + "epoch": 18.78, + "learning_rate": 3.12423113575327e-05, + "loss": 0.8737, + "step": 217020 + }, + { + "epoch": 18.79, + "learning_rate": 3.12414450316209e-05, + "loss": 1.0906, + "step": 217030 + }, + { + "epoch": 18.79, + "learning_rate": 3.124057870570909e-05, + "loss": 1.0359, + "step": 217040 + }, + { + "epoch": 18.79, + "learning_rate": 3.123971237979728e-05, + "loss": 0.97, + "step": 217050 + }, + { + "epoch": 18.79, + "learning_rate": 3.1238846053885474e-05, + "loss": 0.9562, + "step": 217060 + }, + { + "epoch": 18.79, + "learning_rate": 3.123797972797367e-05, + "loss": 0.9564, + "step": 217070 + }, + { + "epoch": 18.79, + "learning_rate": 3.1237113402061856e-05, + "loss": 0.9773, + "step": 217080 + }, + { + "epoch": 18.79, + "learning_rate": 3.123624707615005e-05, + "loss": 0.917, + "step": 217090 + }, + { + "epoch": 18.79, + "learning_rate": 3.123538075023824e-05, + "loss": 0.9967, + "step": 217100 + }, + { + "epoch": 18.79, + "learning_rate": 3.123451442432643e-05, + "loss": 1.0432, + "step": 217110 + }, + { + "epoch": 18.79, + "learning_rate": 3.123364809841463e-05, + "loss": 1.0168, + "step": 217120 + }, + { + "epoch": 18.79, + "learning_rate": 3.1232781772502815e-05, + "loss": 0.9981, + "step": 217130 + }, + { + "epoch": 18.79, + "learning_rate": 3.123191544659101e-05, + "loss": 0.9942, + "step": 217140 + }, + { + "epoch": 18.8, + "learning_rate": 3.1231049120679203e-05, + "loss": 0.992, + "step": 217150 + }, + { + "epoch": 18.8, + "learning_rate": 3.123018279476739e-05, + "loss": 0.9678, + "step": 217160 + }, + { + "epoch": 18.8, + "learning_rate": 3.1229316468855586e-05, + "loss": 0.988, + "step": 217170 + }, + { + "epoch": 18.8, + "learning_rate": 3.122845014294378e-05, + "loss": 1.0264, + "step": 217180 + }, + { + "epoch": 18.8, + "learning_rate": 3.122758381703197e-05, + "loss": 0.9631, + "step": 217190 + }, + { + "epoch": 18.8, + "learning_rate": 3.122671749112016e-05, + "loss": 0.9495, + "step": 217200 + }, + { + "epoch": 18.8, + "learning_rate": 3.122585116520835e-05, + "loss": 0.9853, + "step": 217210 + }, + { + "epoch": 18.8, + "learning_rate": 3.1224984839296544e-05, + "loss": 0.9682, + "step": 217220 + }, + { + "epoch": 18.8, + "learning_rate": 3.122411851338474e-05, + "loss": 0.9808, + "step": 217230 + }, + { + "epoch": 18.8, + "learning_rate": 3.1223252187472926e-05, + "loss": 0.9459, + "step": 217240 + }, + { + "epoch": 18.8, + "learning_rate": 3.122238586156112e-05, + "loss": 0.9479, + "step": 217250 + }, + { + "epoch": 18.8, + "learning_rate": 3.1221519535649315e-05, + "loss": 0.9433, + "step": 217260 + }, + { + "epoch": 18.81, + "learning_rate": 3.12206532097375e-05, + "loss": 1.0112, + "step": 217270 + }, + { + "epoch": 18.81, + "learning_rate": 3.12197868838257e-05, + "loss": 0.9346, + "step": 217280 + }, + { + "epoch": 18.81, + "learning_rate": 3.121892055791389e-05, + "loss": 0.9349, + "step": 217290 + }, + { + "epoch": 18.81, + "learning_rate": 3.121805423200208e-05, + "loss": 0.9206, + "step": 217300 + }, + { + "epoch": 18.81, + "learning_rate": 3.1217187906090274e-05, + "loss": 0.9956, + "step": 217310 + }, + { + "epoch": 18.81, + "learning_rate": 3.121632158017846e-05, + "loss": 1.0282, + "step": 217320 + }, + { + "epoch": 18.81, + "learning_rate": 3.1215455254266656e-05, + "loss": 0.9147, + "step": 217330 + }, + { + "epoch": 18.81, + "learning_rate": 3.121458892835485e-05, + "loss": 1.0015, + "step": 217340 + }, + { + "epoch": 18.81, + "learning_rate": 3.121372260244304e-05, + "loss": 0.9826, + "step": 217350 + }, + { + "epoch": 18.81, + "learning_rate": 3.121285627653123e-05, + "loss": 0.9331, + "step": 217360 + }, + { + "epoch": 18.81, + "learning_rate": 3.121198995061943e-05, + "loss": 1.0002, + "step": 217370 + }, + { + "epoch": 18.82, + "learning_rate": 3.1211123624707615e-05, + "loss": 0.9446, + "step": 217380 + }, + { + "epoch": 18.82, + "learning_rate": 3.121025729879581e-05, + "loss": 0.9979, + "step": 217390 + }, + { + "epoch": 18.82, + "learning_rate": 3.1209390972884004e-05, + "loss": 0.9053, + "step": 217400 + }, + { + "epoch": 18.82, + "learning_rate": 3.120852464697219e-05, + "loss": 1.0453, + "step": 217410 + }, + { + "epoch": 18.82, + "learning_rate": 3.1207658321060386e-05, + "loss": 0.979, + "step": 217420 + }, + { + "epoch": 18.82, + "learning_rate": 3.1206791995148574e-05, + "loss": 0.9725, + "step": 217430 + }, + { + "epoch": 18.82, + "learning_rate": 3.120592566923677e-05, + "loss": 0.944, + "step": 217440 + }, + { + "epoch": 18.82, + "learning_rate": 3.120505934332496e-05, + "loss": 0.973, + "step": 217450 + }, + { + "epoch": 18.82, + "learning_rate": 3.120419301741315e-05, + "loss": 0.9746, + "step": 217460 + }, + { + "epoch": 18.82, + "learning_rate": 3.1203326691501345e-05, + "loss": 1.0074, + "step": 217470 + }, + { + "epoch": 18.82, + "learning_rate": 3.120246036558954e-05, + "loss": 0.955, + "step": 217480 + }, + { + "epoch": 18.82, + "learning_rate": 3.120159403967773e-05, + "loss": 0.9747, + "step": 217490 + }, + { + "epoch": 18.83, + "learning_rate": 3.120072771376592e-05, + "loss": 0.9835, + "step": 217500 + }, + { + "epoch": 18.83, + "learning_rate": 3.1199861387854116e-05, + "loss": 0.974, + "step": 217510 + }, + { + "epoch": 18.83, + "learning_rate": 3.11989950619423e-05, + "loss": 0.9934, + "step": 217520 + }, + { + "epoch": 18.83, + "learning_rate": 3.11981287360305e-05, + "loss": 1.0128, + "step": 217530 + }, + { + "epoch": 18.83, + "learning_rate": 3.1197262410118686e-05, + "loss": 1.0148, + "step": 217540 + }, + { + "epoch": 18.83, + "learning_rate": 3.119639608420688e-05, + "loss": 0.9983, + "step": 217550 + }, + { + "epoch": 18.83, + "learning_rate": 3.1195529758295074e-05, + "loss": 0.8819, + "step": 217560 + }, + { + "epoch": 18.83, + "learning_rate": 3.119466343238326e-05, + "loss": 0.9871, + "step": 217570 + }, + { + "epoch": 18.83, + "learning_rate": 3.1193797106471457e-05, + "loss": 0.9576, + "step": 217580 + }, + { + "epoch": 18.83, + "learning_rate": 3.119293078055965e-05, + "loss": 0.9993, + "step": 217590 + }, + { + "epoch": 18.83, + "learning_rate": 3.119206445464784e-05, + "loss": 0.9902, + "step": 217600 + }, + { + "epoch": 18.84, + "learning_rate": 3.119119812873603e-05, + "loss": 1.0058, + "step": 217610 + }, + { + "epoch": 18.84, + "learning_rate": 3.119033180282422e-05, + "loss": 0.9989, + "step": 217620 + }, + { + "epoch": 18.84, + "learning_rate": 3.1189465476912415e-05, + "loss": 0.985, + "step": 217630 + }, + { + "epoch": 18.84, + "learning_rate": 3.118859915100061e-05, + "loss": 1.0227, + "step": 217640 + }, + { + "epoch": 18.84, + "learning_rate": 3.11877328250888e-05, + "loss": 0.9922, + "step": 217650 + }, + { + "epoch": 18.84, + "learning_rate": 3.118686649917699e-05, + "loss": 0.9984, + "step": 217660 + }, + { + "epoch": 18.84, + "learning_rate": 3.1186000173265186e-05, + "loss": 0.9818, + "step": 217670 + }, + { + "epoch": 18.84, + "learning_rate": 3.1185133847353374e-05, + "loss": 0.9304, + "step": 217680 + }, + { + "epoch": 18.84, + "learning_rate": 3.118426752144157e-05, + "loss": 0.9846, + "step": 217690 + }, + { + "epoch": 18.84, + "learning_rate": 3.118340119552976e-05, + "loss": 0.9943, + "step": 217700 + }, + { + "epoch": 18.84, + "learning_rate": 3.118253486961795e-05, + "loss": 1.0043, + "step": 217710 + }, + { + "epoch": 18.84, + "learning_rate": 3.1181668543706145e-05, + "loss": 0.9796, + "step": 217720 + }, + { + "epoch": 18.85, + "learning_rate": 3.118080221779433e-05, + "loss": 0.9478, + "step": 217730 + }, + { + "epoch": 18.85, + "learning_rate": 3.117993589188253e-05, + "loss": 1.0679, + "step": 217740 + }, + { + "epoch": 18.85, + "learning_rate": 3.117906956597072e-05, + "loss": 0.9646, + "step": 217750 + }, + { + "epoch": 18.85, + "learning_rate": 3.117820324005891e-05, + "loss": 0.9648, + "step": 217760 + }, + { + "epoch": 18.85, + "learning_rate": 3.1177336914147104e-05, + "loss": 1.0119, + "step": 217770 + }, + { + "epoch": 18.85, + "learning_rate": 3.11764705882353e-05, + "loss": 0.9572, + "step": 217780 + }, + { + "epoch": 18.85, + "learning_rate": 3.1175604262323486e-05, + "loss": 0.9312, + "step": 217790 + }, + { + "epoch": 18.85, + "learning_rate": 3.117473793641168e-05, + "loss": 0.9665, + "step": 217800 + }, + { + "epoch": 18.85, + "learning_rate": 3.1173871610499875e-05, + "loss": 0.944, + "step": 217810 + }, + { + "epoch": 18.85, + "learning_rate": 3.117300528458806e-05, + "loss": 1.0196, + "step": 217820 + }, + { + "epoch": 18.85, + "learning_rate": 3.117213895867626e-05, + "loss": 0.9933, + "step": 217830 + }, + { + "epoch": 18.86, + "learning_rate": 3.1171272632764445e-05, + "loss": 1.0104, + "step": 217840 + }, + { + "epoch": 18.86, + "learning_rate": 3.117040630685264e-05, + "loss": 0.9528, + "step": 217850 + }, + { + "epoch": 18.86, + "learning_rate": 3.1169539980940833e-05, + "loss": 0.9815, + "step": 217860 + }, + { + "epoch": 18.86, + "learning_rate": 3.116867365502902e-05, + "loss": 0.9498, + "step": 217870 + }, + { + "epoch": 18.86, + "learning_rate": 3.1167807329117216e-05, + "loss": 0.9963, + "step": 217880 + }, + { + "epoch": 18.86, + "learning_rate": 3.116694100320541e-05, + "loss": 0.9783, + "step": 217890 + }, + { + "epoch": 18.86, + "learning_rate": 3.11660746772936e-05, + "loss": 1.0201, + "step": 217900 + }, + { + "epoch": 18.86, + "learning_rate": 3.116520835138179e-05, + "loss": 0.936, + "step": 217910 + }, + { + "epoch": 18.86, + "learning_rate": 3.116434202546999e-05, + "loss": 0.9972, + "step": 217920 + }, + { + "epoch": 18.86, + "learning_rate": 3.1163475699558174e-05, + "loss": 1.0271, + "step": 217930 + }, + { + "epoch": 18.86, + "learning_rate": 3.116260937364637e-05, + "loss": 1.0093, + "step": 217940 + }, + { + "epoch": 18.86, + "learning_rate": 3.1161743047734556e-05, + "loss": 1.0055, + "step": 217950 + }, + { + "epoch": 18.87, + "learning_rate": 3.116087672182275e-05, + "loss": 0.9689, + "step": 217960 + }, + { + "epoch": 18.87, + "learning_rate": 3.1160010395910945e-05, + "loss": 0.9439, + "step": 217970 + }, + { + "epoch": 18.87, + "learning_rate": 3.115914406999913e-05, + "loss": 0.9364, + "step": 217980 + }, + { + "epoch": 18.87, + "learning_rate": 3.115827774408733e-05, + "loss": 1.0157, + "step": 217990 + }, + { + "epoch": 18.87, + "learning_rate": 3.115741141817552e-05, + "loss": 1.0109, + "step": 218000 + }, + { + "epoch": 18.87, + "learning_rate": 3.115654509226371e-05, + "loss": 1.0171, + "step": 218010 + }, + { + "epoch": 18.87, + "learning_rate": 3.1155678766351904e-05, + "loss": 0.9507, + "step": 218020 + }, + { + "epoch": 18.87, + "learning_rate": 3.11548124404401e-05, + "loss": 1.0022, + "step": 218030 + }, + { + "epoch": 18.87, + "learning_rate": 3.1153946114528286e-05, + "loss": 1.0434, + "step": 218040 + }, + { + "epoch": 18.87, + "learning_rate": 3.115307978861648e-05, + "loss": 0.9564, + "step": 218050 + }, + { + "epoch": 18.87, + "learning_rate": 3.115221346270467e-05, + "loss": 0.9898, + "step": 218060 + }, + { + "epoch": 18.88, + "learning_rate": 3.115134713679286e-05, + "loss": 0.9949, + "step": 218070 + }, + { + "epoch": 18.88, + "learning_rate": 3.115048081088106e-05, + "loss": 0.9704, + "step": 218080 + }, + { + "epoch": 18.88, + "learning_rate": 3.1149614484969245e-05, + "loss": 0.9948, + "step": 218090 + }, + { + "epoch": 18.88, + "learning_rate": 3.114874815905744e-05, + "loss": 0.9492, + "step": 218100 + }, + { + "epoch": 18.88, + "learning_rate": 3.1147881833145634e-05, + "loss": 0.9722, + "step": 218110 + }, + { + "epoch": 18.88, + "learning_rate": 3.114701550723382e-05, + "loss": 0.9407, + "step": 218120 + }, + { + "epoch": 18.88, + "learning_rate": 3.1146149181322016e-05, + "loss": 0.9635, + "step": 218130 + }, + { + "epoch": 18.88, + "learning_rate": 3.114528285541021e-05, + "loss": 0.9629, + "step": 218140 + }, + { + "epoch": 18.88, + "learning_rate": 3.11444165294984e-05, + "loss": 0.992, + "step": 218150 + }, + { + "epoch": 18.88, + "learning_rate": 3.114355020358659e-05, + "loss": 0.997, + "step": 218160 + }, + { + "epoch": 18.88, + "learning_rate": 3.114268387767478e-05, + "loss": 0.9337, + "step": 218170 + }, + { + "epoch": 18.88, + "learning_rate": 3.1141817551762975e-05, + "loss": 0.9373, + "step": 218180 + }, + { + "epoch": 18.89, + "learning_rate": 3.114095122585117e-05, + "loss": 0.9992, + "step": 218190 + }, + { + "epoch": 18.89, + "learning_rate": 3.114008489993936e-05, + "loss": 0.9775, + "step": 218200 + }, + { + "epoch": 18.89, + "learning_rate": 3.113921857402755e-05, + "loss": 0.9492, + "step": 218210 + }, + { + "epoch": 18.89, + "learning_rate": 3.1138352248115746e-05, + "loss": 0.9799, + "step": 218220 + }, + { + "epoch": 18.89, + "learning_rate": 3.113748592220393e-05, + "loss": 1.0509, + "step": 218230 + }, + { + "epoch": 18.89, + "learning_rate": 3.113661959629213e-05, + "loss": 0.979, + "step": 218240 + }, + { + "epoch": 18.89, + "learning_rate": 3.113575327038032e-05, + "loss": 0.9845, + "step": 218250 + }, + { + "epoch": 18.89, + "learning_rate": 3.113488694446851e-05, + "loss": 1.0583, + "step": 218260 + }, + { + "epoch": 18.89, + "learning_rate": 3.1134020618556704e-05, + "loss": 0.9433, + "step": 218270 + }, + { + "epoch": 18.89, + "learning_rate": 3.113315429264489e-05, + "loss": 0.9511, + "step": 218280 + }, + { + "epoch": 18.89, + "learning_rate": 3.1132287966733087e-05, + "loss": 0.9826, + "step": 218290 + }, + { + "epoch": 18.89, + "learning_rate": 3.113142164082128e-05, + "loss": 1.0273, + "step": 218300 + }, + { + "epoch": 18.9, + "learning_rate": 3.113055531490947e-05, + "loss": 1.0146, + "step": 218310 + }, + { + "epoch": 18.9, + "learning_rate": 3.112968898899766e-05, + "loss": 0.9617, + "step": 218320 + }, + { + "epoch": 18.9, + "learning_rate": 3.112882266308586e-05, + "loss": 0.9743, + "step": 218330 + }, + { + "epoch": 18.9, + "learning_rate": 3.1127956337174045e-05, + "loss": 0.9378, + "step": 218340 + }, + { + "epoch": 18.9, + "learning_rate": 3.112709001126224e-05, + "loss": 0.9643, + "step": 218350 + }, + { + "epoch": 18.9, + "learning_rate": 3.1126223685350434e-05, + "loss": 0.9643, + "step": 218360 + }, + { + "epoch": 18.9, + "learning_rate": 3.112535735943862e-05, + "loss": 0.9471, + "step": 218370 + }, + { + "epoch": 18.9, + "learning_rate": 3.1124491033526816e-05, + "loss": 0.9803, + "step": 218380 + }, + { + "epoch": 18.9, + "learning_rate": 3.1123624707615004e-05, + "loss": 0.9923, + "step": 218390 + }, + { + "epoch": 18.9, + "learning_rate": 3.11227583817032e-05, + "loss": 0.9462, + "step": 218400 + }, + { + "epoch": 18.9, + "learning_rate": 3.112189205579139e-05, + "loss": 1.0319, + "step": 218410 + }, + { + "epoch": 18.91, + "learning_rate": 3.112102572987958e-05, + "loss": 0.9773, + "step": 218420 + }, + { + "epoch": 18.91, + "learning_rate": 3.1120159403967775e-05, + "loss": 0.9622, + "step": 218430 + }, + { + "epoch": 18.91, + "learning_rate": 3.111929307805597e-05, + "loss": 0.967, + "step": 218440 + }, + { + "epoch": 18.91, + "learning_rate": 3.111842675214416e-05, + "loss": 0.9684, + "step": 218450 + }, + { + "epoch": 18.91, + "learning_rate": 3.111756042623235e-05, + "loss": 0.9806, + "step": 218460 + }, + { + "epoch": 18.91, + "learning_rate": 3.111669410032054e-05, + "loss": 0.9924, + "step": 218470 + }, + { + "epoch": 18.91, + "learning_rate": 3.1115827774408734e-05, + "loss": 1.0266, + "step": 218480 + }, + { + "epoch": 18.91, + "learning_rate": 3.111496144849693e-05, + "loss": 1.004, + "step": 218490 + }, + { + "epoch": 18.91, + "learning_rate": 3.1114095122585116e-05, + "loss": 0.9748, + "step": 218500 + }, + { + "epoch": 18.91, + "learning_rate": 3.111322879667331e-05, + "loss": 1.0116, + "step": 218510 + }, + { + "epoch": 18.91, + "learning_rate": 3.1112362470761505e-05, + "loss": 0.9869, + "step": 218520 + }, + { + "epoch": 18.91, + "learning_rate": 3.111149614484969e-05, + "loss": 1.026, + "step": 218530 + }, + { + "epoch": 18.92, + "learning_rate": 3.111062981893789e-05, + "loss": 1.033, + "step": 218540 + }, + { + "epoch": 18.92, + "learning_rate": 3.110976349302608e-05, + "loss": 0.9773, + "step": 218550 + }, + { + "epoch": 18.92, + "learning_rate": 3.110889716711427e-05, + "loss": 0.9952, + "step": 218560 + }, + { + "epoch": 18.92, + "learning_rate": 3.1108030841202463e-05, + "loss": 1.0246, + "step": 218570 + }, + { + "epoch": 18.92, + "learning_rate": 3.110716451529065e-05, + "loss": 1.0303, + "step": 218580 + }, + { + "epoch": 18.92, + "learning_rate": 3.1106298189378846e-05, + "loss": 0.9497, + "step": 218590 + }, + { + "epoch": 18.92, + "learning_rate": 3.110543186346704e-05, + "loss": 1.0246, + "step": 218600 + }, + { + "epoch": 18.92, + "learning_rate": 3.110456553755523e-05, + "loss": 0.9843, + "step": 218610 + }, + { + "epoch": 18.92, + "learning_rate": 3.110369921164342e-05, + "loss": 0.9476, + "step": 218620 + }, + { + "epoch": 18.92, + "learning_rate": 3.1102832885731617e-05, + "loss": 0.9853, + "step": 218630 + }, + { + "epoch": 18.92, + "learning_rate": 3.1101966559819804e-05, + "loss": 0.9599, + "step": 218640 + }, + { + "epoch": 18.93, + "learning_rate": 3.1101100233908e-05, + "loss": 1.0205, + "step": 218650 + }, + { + "epoch": 18.93, + "learning_rate": 3.110023390799619e-05, + "loss": 0.9796, + "step": 218660 + }, + { + "epoch": 18.93, + "learning_rate": 3.109936758208438e-05, + "loss": 1.0026, + "step": 218670 + }, + { + "epoch": 18.93, + "learning_rate": 3.1098501256172575e-05, + "loss": 0.9611, + "step": 218680 + }, + { + "epoch": 18.93, + "learning_rate": 3.109763493026076e-05, + "loss": 0.9606, + "step": 218690 + }, + { + "epoch": 18.93, + "learning_rate": 3.109676860434896e-05, + "loss": 0.9225, + "step": 218700 + }, + { + "epoch": 18.93, + "learning_rate": 3.109590227843715e-05, + "loss": 0.9014, + "step": 218710 + }, + { + "epoch": 18.93, + "learning_rate": 3.109503595252534e-05, + "loss": 0.9465, + "step": 218720 + }, + { + "epoch": 18.93, + "learning_rate": 3.1094169626613534e-05, + "loss": 0.963, + "step": 218730 + }, + { + "epoch": 18.93, + "learning_rate": 3.109330330070173e-05, + "loss": 0.9852, + "step": 218740 + }, + { + "epoch": 18.93, + "learning_rate": 3.1092436974789916e-05, + "loss": 0.9942, + "step": 218750 + }, + { + "epoch": 18.93, + "learning_rate": 3.109157064887811e-05, + "loss": 0.9368, + "step": 218760 + }, + { + "epoch": 18.94, + "learning_rate": 3.1090704322966305e-05, + "loss": 0.9429, + "step": 218770 + }, + { + "epoch": 18.94, + "learning_rate": 3.108983799705449e-05, + "loss": 0.9396, + "step": 218780 + }, + { + "epoch": 18.94, + "learning_rate": 3.108897167114269e-05, + "loss": 1.025, + "step": 218790 + }, + { + "epoch": 18.94, + "learning_rate": 3.1088105345230875e-05, + "loss": 0.9877, + "step": 218800 + }, + { + "epoch": 18.94, + "learning_rate": 3.108723901931907e-05, + "loss": 0.9573, + "step": 218810 + }, + { + "epoch": 18.94, + "learning_rate": 3.1086372693407264e-05, + "loss": 0.9523, + "step": 218820 + }, + { + "epoch": 18.94, + "learning_rate": 3.108550636749545e-05, + "loss": 0.9648, + "step": 218830 + }, + { + "epoch": 18.94, + "learning_rate": 3.1084640041583646e-05, + "loss": 0.9716, + "step": 218840 + }, + { + "epoch": 18.94, + "learning_rate": 3.108377371567184e-05, + "loss": 0.9795, + "step": 218850 + }, + { + "epoch": 18.94, + "learning_rate": 3.108290738976003e-05, + "loss": 0.9405, + "step": 218860 + }, + { + "epoch": 18.94, + "learning_rate": 3.108204106384822e-05, + "loss": 0.9515, + "step": 218870 + }, + { + "epoch": 18.95, + "learning_rate": 3.108117473793642e-05, + "loss": 1.0239, + "step": 218880 + }, + { + "epoch": 18.95, + "learning_rate": 3.1080308412024605e-05, + "loss": 1.0294, + "step": 218890 + }, + { + "epoch": 18.95, + "learning_rate": 3.10794420861128e-05, + "loss": 0.9972, + "step": 218900 + }, + { + "epoch": 18.95, + "learning_rate": 3.107857576020099e-05, + "loss": 0.971, + "step": 218910 + }, + { + "epoch": 18.95, + "learning_rate": 3.107770943428918e-05, + "loss": 0.9711, + "step": 218920 + }, + { + "epoch": 18.95, + "learning_rate": 3.1076843108377376e-05, + "loss": 0.9315, + "step": 218930 + }, + { + "epoch": 18.95, + "learning_rate": 3.107597678246556e-05, + "loss": 1.0421, + "step": 218940 + }, + { + "epoch": 18.95, + "learning_rate": 3.107511045655376e-05, + "loss": 1.0201, + "step": 218950 + }, + { + "epoch": 18.95, + "learning_rate": 3.107424413064195e-05, + "loss": 0.9792, + "step": 218960 + }, + { + "epoch": 18.95, + "learning_rate": 3.107337780473014e-05, + "loss": 0.9441, + "step": 218970 + }, + { + "epoch": 18.95, + "learning_rate": 3.1072511478818334e-05, + "loss": 0.9813, + "step": 218980 + }, + { + "epoch": 18.95, + "learning_rate": 3.107164515290653e-05, + "loss": 0.9989, + "step": 218990 + }, + { + "epoch": 18.96, + "learning_rate": 3.1070778826994716e-05, + "loss": 0.9716, + "step": 219000 + }, + { + "epoch": 18.96, + "learning_rate": 3.106991250108291e-05, + "loss": 0.9563, + "step": 219010 + }, + { + "epoch": 18.96, + "learning_rate": 3.10690461751711e-05, + "loss": 0.9444, + "step": 219020 + }, + { + "epoch": 18.96, + "learning_rate": 3.106817984925929e-05, + "loss": 0.952, + "step": 219030 + }, + { + "epoch": 18.96, + "learning_rate": 3.106731352334749e-05, + "loss": 0.983, + "step": 219040 + }, + { + "epoch": 18.96, + "learning_rate": 3.1066447197435675e-05, + "loss": 0.9561, + "step": 219050 + }, + { + "epoch": 18.96, + "learning_rate": 3.106558087152387e-05, + "loss": 1.0015, + "step": 219060 + }, + { + "epoch": 18.96, + "learning_rate": 3.1064714545612064e-05, + "loss": 1.0118, + "step": 219070 + }, + { + "epoch": 18.96, + "learning_rate": 3.106384821970025e-05, + "loss": 0.9267, + "step": 219080 + }, + { + "epoch": 18.96, + "learning_rate": 3.1062981893788446e-05, + "loss": 1.0568, + "step": 219090 + }, + { + "epoch": 18.96, + "learning_rate": 3.106211556787664e-05, + "loss": 1.0344, + "step": 219100 + }, + { + "epoch": 18.97, + "learning_rate": 3.106124924196483e-05, + "loss": 0.9763, + "step": 219110 + }, + { + "epoch": 18.97, + "learning_rate": 3.106038291605302e-05, + "loss": 0.9528, + "step": 219120 + }, + { + "epoch": 18.97, + "learning_rate": 3.105951659014121e-05, + "loss": 0.9688, + "step": 219130 + }, + { + "epoch": 18.97, + "learning_rate": 3.1058650264229405e-05, + "loss": 1.0013, + "step": 219140 + }, + { + "epoch": 18.97, + "learning_rate": 3.10577839383176e-05, + "loss": 0.9852, + "step": 219150 + }, + { + "epoch": 18.97, + "learning_rate": 3.105691761240579e-05, + "loss": 1.0029, + "step": 219160 + }, + { + "epoch": 18.97, + "learning_rate": 3.105605128649398e-05, + "loss": 0.9883, + "step": 219170 + }, + { + "epoch": 18.97, + "learning_rate": 3.1055184960582176e-05, + "loss": 0.9606, + "step": 219180 + }, + { + "epoch": 18.97, + "learning_rate": 3.1054318634670364e-05, + "loss": 0.9515, + "step": 219190 + }, + { + "epoch": 18.97, + "learning_rate": 3.105345230875856e-05, + "loss": 1.0618, + "step": 219200 + }, + { + "epoch": 18.97, + "learning_rate": 3.1052585982846746e-05, + "loss": 0.9917, + "step": 219210 + }, + { + "epoch": 18.97, + "learning_rate": 3.105171965693494e-05, + "loss": 0.9682, + "step": 219220 + }, + { + "epoch": 18.98, + "learning_rate": 3.1050853331023135e-05, + "loss": 1.0513, + "step": 219230 + }, + { + "epoch": 18.98, + "learning_rate": 3.104998700511132e-05, + "loss": 0.9807, + "step": 219240 + }, + { + "epoch": 18.98, + "learning_rate": 3.104912067919952e-05, + "loss": 0.9727, + "step": 219250 + }, + { + "epoch": 18.98, + "learning_rate": 3.104825435328771e-05, + "loss": 0.9923, + "step": 219260 + }, + { + "epoch": 18.98, + "learning_rate": 3.10473880273759e-05, + "loss": 0.9974, + "step": 219270 + }, + { + "epoch": 18.98, + "learning_rate": 3.1046521701464093e-05, + "loss": 1.0227, + "step": 219280 + }, + { + "epoch": 18.98, + "learning_rate": 3.104565537555229e-05, + "loss": 1.0158, + "step": 219290 + }, + { + "epoch": 18.98, + "learning_rate": 3.1044789049640476e-05, + "loss": 1.021, + "step": 219300 + }, + { + "epoch": 18.98, + "learning_rate": 3.104392272372867e-05, + "loss": 1.0752, + "step": 219310 + }, + { + "epoch": 18.98, + "learning_rate": 3.104305639781686e-05, + "loss": 0.98, + "step": 219320 + }, + { + "epoch": 18.98, + "learning_rate": 3.104219007190505e-05, + "loss": 0.9876, + "step": 219330 + }, + { + "epoch": 18.98, + "learning_rate": 3.1041323745993247e-05, + "loss": 1.0262, + "step": 219340 + }, + { + "epoch": 18.99, + "learning_rate": 3.1040457420081434e-05, + "loss": 1.0147, + "step": 219350 + }, + { + "epoch": 18.99, + "learning_rate": 3.103959109416963e-05, + "loss": 1.0022, + "step": 219360 + }, + { + "epoch": 18.99, + "learning_rate": 3.103872476825782e-05, + "loss": 0.9707, + "step": 219370 + }, + { + "epoch": 18.99, + "learning_rate": 3.103785844234601e-05, + "loss": 1.0034, + "step": 219380 + }, + { + "epoch": 18.99, + "learning_rate": 3.1036992116434205e-05, + "loss": 0.9787, + "step": 219390 + }, + { + "epoch": 18.99, + "learning_rate": 3.10361257905224e-05, + "loss": 0.9729, + "step": 219400 + }, + { + "epoch": 18.99, + "learning_rate": 3.103525946461059e-05, + "loss": 1.001, + "step": 219410 + }, + { + "epoch": 18.99, + "learning_rate": 3.103439313869878e-05, + "loss": 0.952, + "step": 219420 + }, + { + "epoch": 18.99, + "learning_rate": 3.103352681278697e-05, + "loss": 0.9485, + "step": 219430 + }, + { + "epoch": 18.99, + "learning_rate": 3.1032660486875164e-05, + "loss": 0.9361, + "step": 219440 + }, + { + "epoch": 18.99, + "learning_rate": 3.103179416096336e-05, + "loss": 0.9709, + "step": 219450 + }, + { + "epoch": 19.0, + "learning_rate": 3.1030927835051546e-05, + "loss": 1.0019, + "step": 219460 + }, + { + "epoch": 19.0, + "learning_rate": 3.103006150913974e-05, + "loss": 0.9178, + "step": 219470 + }, + { + "epoch": 19.0, + "learning_rate": 3.1029195183227935e-05, + "loss": 0.9953, + "step": 219480 + }, + { + "epoch": 19.0, + "learning_rate": 3.102832885731612e-05, + "loss": 1.0011, + "step": 219490 + }, + { + "epoch": 19.0, + "learning_rate": 3.102746253140432e-05, + "loss": 0.9702, + "step": 219500 + }, + { + "epoch": 19.0, + "learning_rate": 3.102659620549251e-05, + "loss": 0.9869, + "step": 219510 + }, + { + "epoch": 19.0, + "eval_Bleu_1": 0.04219876039107803, + "eval_Bleu_2": 2.736813980583177e-11, + "eval_Bleu_3": 2.4446548498168482e-14, + "eval_Bleu_4": 7.498744739589606e-16, + "eval_ROUGE_L": 0.08663949498956944, + "eval_cer": 0.9932266097951767, + "eval_em": 0, + "eval_f1": 0.11298791057523076, + "eval_loss": 0.9726815223693848, + "eval_runtime": 2495.0537, + "eval_samples_per_second": 2.058, + "eval_steps_per_second": 2.058, + "eval_wer": 0.9687356507878981, + "step": 219513 + }, + { + "epoch": 19.0, + "learning_rate": 3.10257298795807e-05, + "loss": 0.9713, + "step": 219520 + }, + { + "epoch": 19.0, + "learning_rate": 3.1024863553668894e-05, + "loss": 0.9399, + "step": 219530 + }, + { + "epoch": 19.0, + "learning_rate": 3.102399722775708e-05, + "loss": 0.9519, + "step": 219540 + }, + { + "epoch": 19.0, + "learning_rate": 3.1023130901845276e-05, + "loss": 0.9516, + "step": 219550 + }, + { + "epoch": 19.0, + "learning_rate": 3.102226457593347e-05, + "loss": 0.9786, + "step": 219560 + }, + { + "epoch": 19.0, + "learning_rate": 3.102139825002166e-05, + "loss": 0.9394, + "step": 219570 + }, + { + "epoch": 19.01, + "learning_rate": 3.102053192410985e-05, + "loss": 0.9198, + "step": 219580 + }, + { + "epoch": 19.01, + "learning_rate": 3.101966559819805e-05, + "loss": 0.9648, + "step": 219590 + }, + { + "epoch": 19.01, + "learning_rate": 3.1018799272286235e-05, + "loss": 0.9375, + "step": 219600 + }, + { + "epoch": 19.01, + "learning_rate": 3.101793294637443e-05, + "loss": 0.9884, + "step": 219610 + }, + { + "epoch": 19.01, + "learning_rate": 3.1017066620462623e-05, + "loss": 0.9799, + "step": 219620 + }, + { + "epoch": 19.01, + "learning_rate": 3.101620029455081e-05, + "loss": 0.9657, + "step": 219630 + }, + { + "epoch": 19.01, + "learning_rate": 3.1015333968639006e-05, + "loss": 0.9878, + "step": 219640 + }, + { + "epoch": 19.01, + "learning_rate": 3.101446764272719e-05, + "loss": 0.9384, + "step": 219650 + }, + { + "epoch": 19.01, + "learning_rate": 3.101360131681539e-05, + "loss": 0.9042, + "step": 219660 + }, + { + "epoch": 19.01, + "learning_rate": 3.101273499090358e-05, + "loss": 0.9575, + "step": 219670 + }, + { + "epoch": 19.01, + "learning_rate": 3.101186866499177e-05, + "loss": 0.9747, + "step": 219680 + }, + { + "epoch": 19.02, + "learning_rate": 3.1011002339079964e-05, + "loss": 0.9805, + "step": 219690 + }, + { + "epoch": 19.02, + "learning_rate": 3.101013601316816e-05, + "loss": 0.9937, + "step": 219700 + }, + { + "epoch": 19.02, + "learning_rate": 3.1009269687256346e-05, + "loss": 1.0146, + "step": 219710 + }, + { + "epoch": 19.02, + "learning_rate": 3.100840336134454e-05, + "loss": 0.8958, + "step": 219720 + }, + { + "epoch": 19.02, + "learning_rate": 3.1007537035432735e-05, + "loss": 0.8546, + "step": 219730 + }, + { + "epoch": 19.02, + "learning_rate": 3.100667070952092e-05, + "loss": 0.9947, + "step": 219740 + }, + { + "epoch": 19.02, + "learning_rate": 3.100580438360912e-05, + "loss": 0.9481, + "step": 219750 + }, + { + "epoch": 19.02, + "learning_rate": 3.1004938057697305e-05, + "loss": 0.9384, + "step": 219760 + }, + { + "epoch": 19.02, + "learning_rate": 3.10040717317855e-05, + "loss": 0.9477, + "step": 219770 + }, + { + "epoch": 19.02, + "learning_rate": 3.1003205405873694e-05, + "loss": 1.0101, + "step": 219780 + }, + { + "epoch": 19.02, + "learning_rate": 3.100233907996188e-05, + "loss": 0.8723, + "step": 219790 + }, + { + "epoch": 19.02, + "learning_rate": 3.1001472754050076e-05, + "loss": 0.9737, + "step": 219800 + }, + { + "epoch": 19.03, + "learning_rate": 3.100060642813827e-05, + "loss": 1.0255, + "step": 219810 + }, + { + "epoch": 19.03, + "learning_rate": 3.099974010222646e-05, + "loss": 0.9811, + "step": 219820 + }, + { + "epoch": 19.03, + "learning_rate": 3.099887377631465e-05, + "loss": 0.9947, + "step": 219830 + }, + { + "epoch": 19.03, + "learning_rate": 3.099800745040285e-05, + "loss": 0.9817, + "step": 219840 + }, + { + "epoch": 19.03, + "learning_rate": 3.0997141124491035e-05, + "loss": 0.9596, + "step": 219850 + }, + { + "epoch": 19.03, + "learning_rate": 3.099627479857923e-05, + "loss": 0.988, + "step": 219860 + }, + { + "epoch": 19.03, + "learning_rate": 3.099540847266742e-05, + "loss": 0.9788, + "step": 219870 + }, + { + "epoch": 19.03, + "learning_rate": 3.099454214675561e-05, + "loss": 0.9794, + "step": 219880 + }, + { + "epoch": 19.03, + "learning_rate": 3.0993675820843806e-05, + "loss": 1.0356, + "step": 219890 + }, + { + "epoch": 19.03, + "learning_rate": 3.0992809494931994e-05, + "loss": 0.923, + "step": 219900 + }, + { + "epoch": 19.03, + "learning_rate": 3.099194316902019e-05, + "loss": 0.9986, + "step": 219910 + }, + { + "epoch": 19.04, + "learning_rate": 3.099107684310838e-05, + "loss": 0.9384, + "step": 219920 + }, + { + "epoch": 19.04, + "learning_rate": 3.099021051719657e-05, + "loss": 0.9793, + "step": 219930 + }, + { + "epoch": 19.04, + "learning_rate": 3.0989344191284765e-05, + "loss": 0.9848, + "step": 219940 + }, + { + "epoch": 19.04, + "learning_rate": 3.098847786537295e-05, + "loss": 1.0043, + "step": 219950 + }, + { + "epoch": 19.04, + "learning_rate": 3.098761153946115e-05, + "loss": 0.9556, + "step": 219960 + }, + { + "epoch": 19.04, + "learning_rate": 3.098674521354934e-05, + "loss": 0.9484, + "step": 219970 + }, + { + "epoch": 19.04, + "learning_rate": 3.098587888763753e-05, + "loss": 0.9705, + "step": 219980 + }, + { + "epoch": 19.04, + "learning_rate": 3.098501256172572e-05, + "loss": 0.9223, + "step": 219990 + }, + { + "epoch": 19.04, + "learning_rate": 3.098414623581392e-05, + "loss": 0.9427, + "step": 220000 + }, + { + "epoch": 19.04, + "learning_rate": 3.0983279909902106e-05, + "loss": 0.9656, + "step": 220010 + }, + { + "epoch": 19.04, + "learning_rate": 3.09824135839903e-05, + "loss": 0.9444, + "step": 220020 + }, + { + "epoch": 19.04, + "learning_rate": 3.0981547258078494e-05, + "loss": 1.0243, + "step": 220030 + }, + { + "epoch": 19.05, + "learning_rate": 3.098068093216668e-05, + "loss": 0.9967, + "step": 220040 + }, + { + "epoch": 19.05, + "learning_rate": 3.0979814606254877e-05, + "loss": 0.9673, + "step": 220050 + }, + { + "epoch": 19.05, + "learning_rate": 3.0978948280343064e-05, + "loss": 1.0189, + "step": 220060 + }, + { + "epoch": 19.05, + "learning_rate": 3.097808195443126e-05, + "loss": 0.9547, + "step": 220070 + }, + { + "epoch": 19.05, + "learning_rate": 3.097721562851945e-05, + "loss": 0.9691, + "step": 220080 + }, + { + "epoch": 19.05, + "learning_rate": 3.097634930260764e-05, + "loss": 0.9638, + "step": 220090 + }, + { + "epoch": 19.05, + "learning_rate": 3.0975482976695835e-05, + "loss": 0.9339, + "step": 220100 + }, + { + "epoch": 19.05, + "learning_rate": 3.097461665078403e-05, + "loss": 0.9275, + "step": 220110 + }, + { + "epoch": 19.05, + "learning_rate": 3.097375032487222e-05, + "loss": 0.9877, + "step": 220120 + }, + { + "epoch": 19.05, + "learning_rate": 3.097288399896041e-05, + "loss": 0.9638, + "step": 220130 + }, + { + "epoch": 19.05, + "learning_rate": 3.0972017673048606e-05, + "loss": 0.9786, + "step": 220140 + }, + { + "epoch": 19.06, + "learning_rate": 3.0971151347136794e-05, + "loss": 0.9952, + "step": 220150 + }, + { + "epoch": 19.06, + "learning_rate": 3.097028502122499e-05, + "loss": 1.0201, + "step": 220160 + }, + { + "epoch": 19.06, + "learning_rate": 3.0969418695313176e-05, + "loss": 0.976, + "step": 220170 + }, + { + "epoch": 19.06, + "learning_rate": 3.096855236940137e-05, + "loss": 0.9724, + "step": 220180 + }, + { + "epoch": 19.06, + "learning_rate": 3.0967686043489565e-05, + "loss": 0.9844, + "step": 220190 + }, + { + "epoch": 19.06, + "learning_rate": 3.096681971757775e-05, + "loss": 0.9538, + "step": 220200 + }, + { + "epoch": 19.06, + "learning_rate": 3.096595339166595e-05, + "loss": 0.9699, + "step": 220210 + }, + { + "epoch": 19.06, + "learning_rate": 3.096508706575414e-05, + "loss": 1.0083, + "step": 220220 + }, + { + "epoch": 19.06, + "learning_rate": 3.096422073984233e-05, + "loss": 1.0084, + "step": 220230 + }, + { + "epoch": 19.06, + "learning_rate": 3.0963354413930524e-05, + "loss": 0.9737, + "step": 220240 + }, + { + "epoch": 19.06, + "learning_rate": 3.096248808801872e-05, + "loss": 0.9523, + "step": 220250 + }, + { + "epoch": 19.06, + "learning_rate": 3.0961621762106906e-05, + "loss": 0.9786, + "step": 220260 + }, + { + "epoch": 19.07, + "learning_rate": 3.09607554361951e-05, + "loss": 0.9462, + "step": 220270 + }, + { + "epoch": 19.07, + "learning_rate": 3.095988911028329e-05, + "loss": 0.9837, + "step": 220280 + }, + { + "epoch": 19.07, + "learning_rate": 3.095902278437148e-05, + "loss": 0.8981, + "step": 220290 + }, + { + "epoch": 19.07, + "learning_rate": 3.095815645845968e-05, + "loss": 0.9721, + "step": 220300 + }, + { + "epoch": 19.07, + "learning_rate": 3.0957290132547865e-05, + "loss": 0.9144, + "step": 220310 + }, + { + "epoch": 19.07, + "learning_rate": 3.095642380663606e-05, + "loss": 0.95, + "step": 220320 + }, + { + "epoch": 19.07, + "learning_rate": 3.0955557480724253e-05, + "loss": 0.9122, + "step": 220330 + }, + { + "epoch": 19.07, + "learning_rate": 3.095469115481244e-05, + "loss": 0.9966, + "step": 220340 + }, + { + "epoch": 19.07, + "learning_rate": 3.0953824828900636e-05, + "loss": 0.9435, + "step": 220350 + }, + { + "epoch": 19.07, + "learning_rate": 3.095295850298883e-05, + "loss": 0.952, + "step": 220360 + }, + { + "epoch": 19.07, + "learning_rate": 3.095209217707702e-05, + "loss": 0.9711, + "step": 220370 + }, + { + "epoch": 19.07, + "learning_rate": 3.095122585116521e-05, + "loss": 0.9913, + "step": 220380 + }, + { + "epoch": 19.08, + "learning_rate": 3.09503595252534e-05, + "loss": 0.9848, + "step": 220390 + }, + { + "epoch": 19.08, + "learning_rate": 3.0949493199341594e-05, + "loss": 0.9329, + "step": 220400 + }, + { + "epoch": 19.08, + "learning_rate": 3.094862687342979e-05, + "loss": 0.9208, + "step": 220410 + }, + { + "epoch": 19.08, + "learning_rate": 3.0947760547517976e-05, + "loss": 0.9482, + "step": 220420 + }, + { + "epoch": 19.08, + "learning_rate": 3.094689422160617e-05, + "loss": 1.0098, + "step": 220430 + }, + { + "epoch": 19.08, + "learning_rate": 3.0946027895694365e-05, + "loss": 0.9513, + "step": 220440 + }, + { + "epoch": 19.08, + "learning_rate": 3.094516156978255e-05, + "loss": 0.9175, + "step": 220450 + }, + { + "epoch": 19.08, + "learning_rate": 3.094429524387075e-05, + "loss": 1.0513, + "step": 220460 + }, + { + "epoch": 19.08, + "learning_rate": 3.094342891795894e-05, + "loss": 0.9866, + "step": 220470 + }, + { + "epoch": 19.08, + "learning_rate": 3.094256259204713e-05, + "loss": 0.9799, + "step": 220480 + }, + { + "epoch": 19.08, + "learning_rate": 3.0941696266135324e-05, + "loss": 0.9739, + "step": 220490 + }, + { + "epoch": 19.09, + "learning_rate": 3.094082994022351e-05, + "loss": 0.9677, + "step": 220500 + }, + { + "epoch": 19.09, + "learning_rate": 3.0939963614311706e-05, + "loss": 0.9938, + "step": 220510 + }, + { + "epoch": 19.09, + "learning_rate": 3.09390972883999e-05, + "loss": 1.0271, + "step": 220520 + }, + { + "epoch": 19.09, + "learning_rate": 3.093823096248809e-05, + "loss": 0.9805, + "step": 220530 + }, + { + "epoch": 19.09, + "learning_rate": 3.093736463657628e-05, + "loss": 0.9619, + "step": 220540 + }, + { + "epoch": 19.09, + "learning_rate": 3.093649831066448e-05, + "loss": 0.9991, + "step": 220550 + }, + { + "epoch": 19.09, + "learning_rate": 3.0935631984752665e-05, + "loss": 0.9756, + "step": 220560 + }, + { + "epoch": 19.09, + "learning_rate": 3.093476565884086e-05, + "loss": 1.0402, + "step": 220570 + }, + { + "epoch": 19.09, + "learning_rate": 3.0933899332929054e-05, + "loss": 0.9883, + "step": 220580 + }, + { + "epoch": 19.09, + "learning_rate": 3.093303300701724e-05, + "loss": 0.9382, + "step": 220590 + }, + { + "epoch": 19.09, + "learning_rate": 3.0932166681105436e-05, + "loss": 0.9661, + "step": 220600 + }, + { + "epoch": 19.09, + "learning_rate": 3.0931300355193624e-05, + "loss": 0.9696, + "step": 220610 + }, + { + "epoch": 19.1, + "learning_rate": 3.093043402928182e-05, + "loss": 1.0118, + "step": 220620 + }, + { + "epoch": 19.1, + "learning_rate": 3.092956770337001e-05, + "loss": 0.9611, + "step": 220630 + }, + { + "epoch": 19.1, + "learning_rate": 3.09287013774582e-05, + "loss": 0.9213, + "step": 220640 + }, + { + "epoch": 19.1, + "learning_rate": 3.0927835051546395e-05, + "loss": 1.0092, + "step": 220650 + }, + { + "epoch": 19.1, + "learning_rate": 3.092696872563459e-05, + "loss": 0.9976, + "step": 220660 + }, + { + "epoch": 19.1, + "learning_rate": 3.092610239972278e-05, + "loss": 0.9595, + "step": 220670 + }, + { + "epoch": 19.1, + "learning_rate": 3.092523607381097e-05, + "loss": 0.9774, + "step": 220680 + }, + { + "epoch": 19.1, + "learning_rate": 3.0924369747899166e-05, + "loss": 0.9617, + "step": 220690 + }, + { + "epoch": 19.1, + "learning_rate": 3.092350342198735e-05, + "loss": 0.9786, + "step": 220700 + }, + { + "epoch": 19.1, + "learning_rate": 3.092263709607555e-05, + "loss": 0.9922, + "step": 220710 + }, + { + "epoch": 19.1, + "learning_rate": 3.0921770770163735e-05, + "loss": 0.9436, + "step": 220720 + }, + { + "epoch": 19.11, + "learning_rate": 3.092090444425193e-05, + "loss": 0.9443, + "step": 220730 + }, + { + "epoch": 19.11, + "learning_rate": 3.0920038118340124e-05, + "loss": 0.9593, + "step": 220740 + }, + { + "epoch": 19.11, + "learning_rate": 3.091917179242831e-05, + "loss": 1.0172, + "step": 220750 + }, + { + "epoch": 19.11, + "learning_rate": 3.0918305466516507e-05, + "loss": 0.969, + "step": 220760 + }, + { + "epoch": 19.11, + "learning_rate": 3.09174391406047e-05, + "loss": 0.9199, + "step": 220770 + }, + { + "epoch": 19.11, + "learning_rate": 3.091657281469289e-05, + "loss": 0.961, + "step": 220780 + }, + { + "epoch": 19.11, + "learning_rate": 3.091570648878108e-05, + "loss": 0.9522, + "step": 220790 + }, + { + "epoch": 19.11, + "learning_rate": 3.091484016286927e-05, + "loss": 1.0181, + "step": 220800 + }, + { + "epoch": 19.11, + "learning_rate": 3.0913973836957465e-05, + "loss": 0.9304, + "step": 220810 + }, + { + "epoch": 19.11, + "learning_rate": 3.091310751104566e-05, + "loss": 0.9382, + "step": 220820 + }, + { + "epoch": 19.11, + "learning_rate": 3.091224118513385e-05, + "loss": 0.9695, + "step": 220830 + }, + { + "epoch": 19.11, + "learning_rate": 3.091137485922204e-05, + "loss": 0.9953, + "step": 220840 + }, + { + "epoch": 19.12, + "learning_rate": 3.0910508533310236e-05, + "loss": 0.9963, + "step": 220850 + }, + { + "epoch": 19.12, + "learning_rate": 3.0909642207398424e-05, + "loss": 0.9748, + "step": 220860 + }, + { + "epoch": 19.12, + "learning_rate": 3.090877588148662e-05, + "loss": 0.978, + "step": 220870 + }, + { + "epoch": 19.12, + "learning_rate": 3.090790955557481e-05, + "loss": 0.9071, + "step": 220880 + }, + { + "epoch": 19.12, + "learning_rate": 3.0907043229663e-05, + "loss": 0.9655, + "step": 220890 + }, + { + "epoch": 19.12, + "learning_rate": 3.0906176903751195e-05, + "loss": 0.9716, + "step": 220900 + }, + { + "epoch": 19.12, + "learning_rate": 3.090531057783938e-05, + "loss": 0.9265, + "step": 220910 + }, + { + "epoch": 19.12, + "learning_rate": 3.090444425192758e-05, + "loss": 0.9811, + "step": 220920 + }, + { + "epoch": 19.12, + "learning_rate": 3.090357792601577e-05, + "loss": 1.0087, + "step": 220930 + }, + { + "epoch": 19.12, + "learning_rate": 3.090271160010396e-05, + "loss": 0.9391, + "step": 220940 + }, + { + "epoch": 19.12, + "learning_rate": 3.0901845274192154e-05, + "loss": 0.9244, + "step": 220950 + }, + { + "epoch": 19.13, + "learning_rate": 3.090097894828035e-05, + "loss": 0.9926, + "step": 220960 + }, + { + "epoch": 19.13, + "learning_rate": 3.0900112622368536e-05, + "loss": 1.0028, + "step": 220970 + }, + { + "epoch": 19.13, + "learning_rate": 3.089924629645673e-05, + "loss": 0.9722, + "step": 220980 + }, + { + "epoch": 19.13, + "learning_rate": 3.0898379970544925e-05, + "loss": 0.9499, + "step": 220990 + }, + { + "epoch": 19.13, + "learning_rate": 3.089751364463311e-05, + "loss": 0.9521, + "step": 221000 + }, + { + "epoch": 19.13, + "learning_rate": 3.089664731872131e-05, + "loss": 0.9604, + "step": 221010 + }, + { + "epoch": 19.13, + "learning_rate": 3.0895780992809495e-05, + "loss": 0.9212, + "step": 221020 + }, + { + "epoch": 19.13, + "learning_rate": 3.089491466689769e-05, + "loss": 0.9706, + "step": 221030 + }, + { + "epoch": 19.13, + "learning_rate": 3.0894048340985883e-05, + "loss": 1.0238, + "step": 221040 + }, + { + "epoch": 19.13, + "learning_rate": 3.089318201507407e-05, + "loss": 1.0177, + "step": 221050 + }, + { + "epoch": 19.13, + "learning_rate": 3.0892315689162266e-05, + "loss": 0.9776, + "step": 221060 + }, + { + "epoch": 19.13, + "learning_rate": 3.089144936325046e-05, + "loss": 0.9055, + "step": 221070 + }, + { + "epoch": 19.14, + "learning_rate": 3.089058303733865e-05, + "loss": 0.9714, + "step": 221080 + }, + { + "epoch": 19.14, + "learning_rate": 3.088971671142684e-05, + "loss": 0.9843, + "step": 221090 + }, + { + "epoch": 19.14, + "learning_rate": 3.0888850385515037e-05, + "loss": 1.0174, + "step": 221100 + }, + { + "epoch": 19.14, + "learning_rate": 3.0887984059603224e-05, + "loss": 0.9226, + "step": 221110 + }, + { + "epoch": 19.14, + "learning_rate": 3.088711773369142e-05, + "loss": 0.9709, + "step": 221120 + }, + { + "epoch": 19.14, + "learning_rate": 3.0886251407779606e-05, + "loss": 0.9302, + "step": 221130 + }, + { + "epoch": 19.14, + "learning_rate": 3.08853850818678e-05, + "loss": 0.921, + "step": 221140 + }, + { + "epoch": 19.14, + "learning_rate": 3.0884518755955995e-05, + "loss": 0.9911, + "step": 221150 + }, + { + "epoch": 19.14, + "learning_rate": 3.088365243004418e-05, + "loss": 0.9347, + "step": 221160 + }, + { + "epoch": 19.14, + "learning_rate": 3.088278610413238e-05, + "loss": 0.9425, + "step": 221170 + }, + { + "epoch": 19.14, + "learning_rate": 3.088191977822057e-05, + "loss": 0.9375, + "step": 221180 + }, + { + "epoch": 19.15, + "learning_rate": 3.088105345230876e-05, + "loss": 0.9688, + "step": 221190 + }, + { + "epoch": 19.15, + "learning_rate": 3.0880187126396954e-05, + "loss": 0.9689, + "step": 221200 + }, + { + "epoch": 19.15, + "learning_rate": 3.087932080048515e-05, + "loss": 1.0311, + "step": 221210 + }, + { + "epoch": 19.15, + "learning_rate": 3.0878454474573336e-05, + "loss": 0.9417, + "step": 221220 + }, + { + "epoch": 19.15, + "learning_rate": 3.087758814866153e-05, + "loss": 0.9436, + "step": 221230 + }, + { + "epoch": 19.15, + "learning_rate": 3.087672182274972e-05, + "loss": 0.9273, + "step": 221240 + }, + { + "epoch": 19.15, + "learning_rate": 3.087585549683791e-05, + "loss": 0.937, + "step": 221250 + }, + { + "epoch": 19.15, + "learning_rate": 3.087498917092611e-05, + "loss": 0.9219, + "step": 221260 + }, + { + "epoch": 19.15, + "learning_rate": 3.0874122845014295e-05, + "loss": 0.9309, + "step": 221270 + }, + { + "epoch": 19.15, + "learning_rate": 3.087325651910249e-05, + "loss": 0.9799, + "step": 221280 + }, + { + "epoch": 19.15, + "learning_rate": 3.0872390193190684e-05, + "loss": 0.9203, + "step": 221290 + }, + { + "epoch": 19.15, + "learning_rate": 3.087152386727887e-05, + "loss": 0.9783, + "step": 221300 + }, + { + "epoch": 19.16, + "learning_rate": 3.0870657541367066e-05, + "loss": 0.9625, + "step": 221310 + }, + { + "epoch": 19.16, + "learning_rate": 3.086979121545526e-05, + "loss": 1.007, + "step": 221320 + }, + { + "epoch": 19.16, + "learning_rate": 3.086892488954345e-05, + "loss": 0.9545, + "step": 221330 + }, + { + "epoch": 19.16, + "learning_rate": 3.086805856363164e-05, + "loss": 0.9944, + "step": 221340 + }, + { + "epoch": 19.16, + "learning_rate": 3.086719223771983e-05, + "loss": 1.0152, + "step": 221350 + }, + { + "epoch": 19.16, + "learning_rate": 3.0866325911808025e-05, + "loss": 0.9614, + "step": 221360 + }, + { + "epoch": 19.16, + "learning_rate": 3.086545958589622e-05, + "loss": 1.0139, + "step": 221370 + }, + { + "epoch": 19.16, + "learning_rate": 3.086459325998441e-05, + "loss": 0.9488, + "step": 221380 + }, + { + "epoch": 19.16, + "learning_rate": 3.08637269340726e-05, + "loss": 0.9522, + "step": 221390 + }, + { + "epoch": 19.16, + "learning_rate": 3.0862860608160796e-05, + "loss": 0.9718, + "step": 221400 + }, + { + "epoch": 19.16, + "learning_rate": 3.086199428224898e-05, + "loss": 0.9654, + "step": 221410 + }, + { + "epoch": 19.17, + "learning_rate": 3.086112795633718e-05, + "loss": 1.0045, + "step": 221420 + }, + { + "epoch": 19.17, + "learning_rate": 3.086026163042537e-05, + "loss": 0.9685, + "step": 221430 + }, + { + "epoch": 19.17, + "learning_rate": 3.085939530451356e-05, + "loss": 0.9532, + "step": 221440 + }, + { + "epoch": 19.17, + "learning_rate": 3.0858528978601754e-05, + "loss": 0.9337, + "step": 221450 + }, + { + "epoch": 19.17, + "learning_rate": 3.085766265268994e-05, + "loss": 0.9312, + "step": 221460 + }, + { + "epoch": 19.17, + "learning_rate": 3.0856796326778136e-05, + "loss": 1.0118, + "step": 221470 + }, + { + "epoch": 19.17, + "learning_rate": 3.085593000086633e-05, + "loss": 0.9435, + "step": 221480 + }, + { + "epoch": 19.17, + "learning_rate": 3.085506367495452e-05, + "loss": 0.9881, + "step": 221490 + }, + { + "epoch": 19.17, + "learning_rate": 3.085419734904271e-05, + "loss": 0.9799, + "step": 221500 + }, + { + "epoch": 19.17, + "learning_rate": 3.085333102313091e-05, + "loss": 0.8913, + "step": 221510 + }, + { + "epoch": 19.17, + "learning_rate": 3.0852464697219095e-05, + "loss": 0.9412, + "step": 221520 + }, + { + "epoch": 19.17, + "learning_rate": 3.085159837130729e-05, + "loss": 0.9881, + "step": 221530 + }, + { + "epoch": 19.18, + "learning_rate": 3.085073204539548e-05, + "loss": 0.931, + "step": 221540 + }, + { + "epoch": 19.18, + "learning_rate": 3.084986571948367e-05, + "loss": 0.987, + "step": 221550 + }, + { + "epoch": 19.18, + "learning_rate": 3.0848999393571866e-05, + "loss": 1.0014, + "step": 221560 + }, + { + "epoch": 19.18, + "learning_rate": 3.0848133067660054e-05, + "loss": 0.9699, + "step": 221570 + }, + { + "epoch": 19.18, + "learning_rate": 3.084726674174825e-05, + "loss": 0.9715, + "step": 221580 + }, + { + "epoch": 19.18, + "learning_rate": 3.084640041583644e-05, + "loss": 0.9832, + "step": 221590 + }, + { + "epoch": 19.18, + "learning_rate": 3.084553408992463e-05, + "loss": 0.956, + "step": 221600 + }, + { + "epoch": 19.18, + "learning_rate": 3.0844667764012825e-05, + "loss": 0.9049, + "step": 221610 + }, + { + "epoch": 19.18, + "learning_rate": 3.084380143810102e-05, + "loss": 0.953, + "step": 221620 + }, + { + "epoch": 19.18, + "learning_rate": 3.084293511218921e-05, + "loss": 0.972, + "step": 221630 + }, + { + "epoch": 19.18, + "learning_rate": 3.08420687862774e-05, + "loss": 0.972, + "step": 221640 + }, + { + "epoch": 19.18, + "learning_rate": 3.084120246036559e-05, + "loss": 0.9686, + "step": 221650 + }, + { + "epoch": 19.19, + "learning_rate": 3.0840336134453784e-05, + "loss": 0.9504, + "step": 221660 + }, + { + "epoch": 19.19, + "learning_rate": 3.083946980854198e-05, + "loss": 1.0026, + "step": 221670 + }, + { + "epoch": 19.19, + "learning_rate": 3.0838603482630166e-05, + "loss": 0.9958, + "step": 221680 + }, + { + "epoch": 19.19, + "learning_rate": 3.083773715671836e-05, + "loss": 0.9704, + "step": 221690 + }, + { + "epoch": 19.19, + "learning_rate": 3.0836870830806555e-05, + "loss": 0.9596, + "step": 221700 + }, + { + "epoch": 19.19, + "learning_rate": 3.083600450489474e-05, + "loss": 0.9826, + "step": 221710 + }, + { + "epoch": 19.19, + "learning_rate": 3.083513817898294e-05, + "loss": 0.9848, + "step": 221720 + }, + { + "epoch": 19.19, + "learning_rate": 3.083427185307113e-05, + "loss": 0.9746, + "step": 221730 + }, + { + "epoch": 19.19, + "learning_rate": 3.083340552715932e-05, + "loss": 0.9938, + "step": 221740 + }, + { + "epoch": 19.19, + "learning_rate": 3.0832539201247513e-05, + "loss": 0.9926, + "step": 221750 + }, + { + "epoch": 19.19, + "learning_rate": 3.08316728753357e-05, + "loss": 0.9241, + "step": 221760 + }, + { + "epoch": 19.2, + "learning_rate": 3.0830806549423896e-05, + "loss": 0.9726, + "step": 221770 + }, + { + "epoch": 19.2, + "learning_rate": 3.082994022351209e-05, + "loss": 0.9878, + "step": 221780 + }, + { + "epoch": 19.2, + "learning_rate": 3.082907389760028e-05, + "loss": 0.971, + "step": 221790 + }, + { + "epoch": 19.2, + "learning_rate": 3.082820757168847e-05, + "loss": 0.9309, + "step": 221800 + }, + { + "epoch": 19.2, + "learning_rate": 3.0827341245776667e-05, + "loss": 1.0336, + "step": 221810 + }, + { + "epoch": 19.2, + "learning_rate": 3.0826474919864854e-05, + "loss": 0.9775, + "step": 221820 + }, + { + "epoch": 19.2, + "learning_rate": 3.082560859395305e-05, + "loss": 0.9005, + "step": 221830 + }, + { + "epoch": 19.2, + "learning_rate": 3.082474226804124e-05, + "loss": 0.9625, + "step": 221840 + }, + { + "epoch": 19.2, + "learning_rate": 3.082387594212943e-05, + "loss": 1.013, + "step": 221850 + }, + { + "epoch": 19.2, + "learning_rate": 3.0823009616217625e-05, + "loss": 0.9436, + "step": 221860 + }, + { + "epoch": 19.2, + "learning_rate": 3.082214329030581e-05, + "loss": 0.9228, + "step": 221870 + }, + { + "epoch": 19.2, + "learning_rate": 3.082127696439401e-05, + "loss": 1.0071, + "step": 221880 + }, + { + "epoch": 19.21, + "learning_rate": 3.08204106384822e-05, + "loss": 0.9127, + "step": 221890 + }, + { + "epoch": 19.21, + "learning_rate": 3.081954431257039e-05, + "loss": 0.9406, + "step": 221900 + }, + { + "epoch": 19.21, + "learning_rate": 3.0818677986658584e-05, + "loss": 0.9643, + "step": 221910 + }, + { + "epoch": 19.21, + "learning_rate": 3.081781166074678e-05, + "loss": 0.928, + "step": 221920 + }, + { + "epoch": 19.21, + "learning_rate": 3.0816945334834966e-05, + "loss": 0.9694, + "step": 221930 + }, + { + "epoch": 19.21, + "learning_rate": 3.081607900892316e-05, + "loss": 0.9455, + "step": 221940 + }, + { + "epoch": 19.21, + "learning_rate": 3.0815212683011355e-05, + "loss": 0.9345, + "step": 221950 + }, + { + "epoch": 19.21, + "learning_rate": 3.081434635709954e-05, + "loss": 0.967, + "step": 221960 + }, + { + "epoch": 19.21, + "learning_rate": 3.081348003118774e-05, + "loss": 0.9804, + "step": 221970 + }, + { + "epoch": 19.21, + "learning_rate": 3.0812613705275925e-05, + "loss": 0.9985, + "step": 221980 + }, + { + "epoch": 19.21, + "learning_rate": 3.081174737936412e-05, + "loss": 0.9807, + "step": 221990 + }, + { + "epoch": 19.22, + "learning_rate": 3.0810881053452314e-05, + "loss": 0.9579, + "step": 222000 + }, + { + "epoch": 19.22, + "learning_rate": 3.08100147275405e-05, + "loss": 0.9876, + "step": 222010 + }, + { + "epoch": 19.22, + "learning_rate": 3.0809148401628696e-05, + "loss": 0.9894, + "step": 222020 + }, + { + "epoch": 19.22, + "learning_rate": 3.080828207571689e-05, + "loss": 0.9636, + "step": 222030 + }, + { + "epoch": 19.22, + "learning_rate": 3.080741574980508e-05, + "loss": 0.9546, + "step": 222040 + }, + { + "epoch": 19.22, + "learning_rate": 3.080654942389327e-05, + "loss": 0.9809, + "step": 222050 + }, + { + "epoch": 19.22, + "learning_rate": 3.080568309798147e-05, + "loss": 0.9745, + "step": 222060 + }, + { + "epoch": 19.22, + "learning_rate": 3.0804816772069655e-05, + "loss": 0.9638, + "step": 222070 + }, + { + "epoch": 19.22, + "learning_rate": 3.080395044615785e-05, + "loss": 0.9788, + "step": 222080 + }, + { + "epoch": 19.22, + "learning_rate": 3.080308412024604e-05, + "loss": 0.9722, + "step": 222090 + }, + { + "epoch": 19.22, + "learning_rate": 3.080221779433423e-05, + "loss": 0.9137, + "step": 222100 + }, + { + "epoch": 19.22, + "learning_rate": 3.0801351468422426e-05, + "loss": 0.9688, + "step": 222110 + }, + { + "epoch": 19.23, + "learning_rate": 3.080048514251061e-05, + "loss": 0.9256, + "step": 222120 + }, + { + "epoch": 19.23, + "learning_rate": 3.079961881659881e-05, + "loss": 0.9649, + "step": 222130 + }, + { + "epoch": 19.23, + "learning_rate": 3.0798752490687e-05, + "loss": 0.9611, + "step": 222140 + }, + { + "epoch": 19.23, + "learning_rate": 3.079788616477519e-05, + "loss": 1.0217, + "step": 222150 + }, + { + "epoch": 19.23, + "learning_rate": 3.0797019838863384e-05, + "loss": 1.0096, + "step": 222160 + }, + { + "epoch": 19.23, + "learning_rate": 3.079615351295158e-05, + "loss": 0.9467, + "step": 222170 + }, + { + "epoch": 19.23, + "learning_rate": 3.0795287187039766e-05, + "loss": 0.9202, + "step": 222180 + }, + { + "epoch": 19.23, + "learning_rate": 3.079442086112796e-05, + "loss": 0.9403, + "step": 222190 + }, + { + "epoch": 19.23, + "learning_rate": 3.079355453521615e-05, + "loss": 0.9674, + "step": 222200 + }, + { + "epoch": 19.23, + "learning_rate": 3.079268820930434e-05, + "loss": 0.9575, + "step": 222210 + }, + { + "epoch": 19.23, + "learning_rate": 3.079182188339254e-05, + "loss": 0.9447, + "step": 222220 + }, + { + "epoch": 19.24, + "learning_rate": 3.0790955557480725e-05, + "loss": 0.9988, + "step": 222230 + }, + { + "epoch": 19.24, + "learning_rate": 3.079008923156892e-05, + "loss": 0.9591, + "step": 222240 + }, + { + "epoch": 19.24, + "learning_rate": 3.0789222905657114e-05, + "loss": 0.9513, + "step": 222250 + }, + { + "epoch": 19.24, + "learning_rate": 3.07883565797453e-05, + "loss": 0.9742, + "step": 222260 + }, + { + "epoch": 19.24, + "learning_rate": 3.0787490253833496e-05, + "loss": 0.9815, + "step": 222270 + }, + { + "epoch": 19.24, + "learning_rate": 3.0786623927921684e-05, + "loss": 0.9618, + "step": 222280 + }, + { + "epoch": 19.24, + "learning_rate": 3.078575760200988e-05, + "loss": 0.9976, + "step": 222290 + }, + { + "epoch": 19.24, + "learning_rate": 3.078489127609807e-05, + "loss": 0.9094, + "step": 222300 + }, + { + "epoch": 19.24, + "learning_rate": 3.078402495018626e-05, + "loss": 0.9479, + "step": 222310 + }, + { + "epoch": 19.24, + "learning_rate": 3.0783158624274455e-05, + "loss": 0.934, + "step": 222320 + }, + { + "epoch": 19.24, + "learning_rate": 3.078229229836265e-05, + "loss": 1.0044, + "step": 222330 + }, + { + "epoch": 19.24, + "learning_rate": 3.078142597245084e-05, + "loss": 0.946, + "step": 222340 + }, + { + "epoch": 19.25, + "learning_rate": 3.078055964653903e-05, + "loss": 1.0173, + "step": 222350 + }, + { + "epoch": 19.25, + "learning_rate": 3.0779693320627226e-05, + "loss": 0.9209, + "step": 222360 + }, + { + "epoch": 19.25, + "learning_rate": 3.0778826994715414e-05, + "loss": 0.9982, + "step": 222370 + }, + { + "epoch": 19.25, + "learning_rate": 3.077796066880361e-05, + "loss": 0.986, + "step": 222380 + }, + { + "epoch": 19.25, + "learning_rate": 3.0777094342891796e-05, + "loss": 0.943, + "step": 222390 + }, + { + "epoch": 19.25, + "learning_rate": 3.077622801697999e-05, + "loss": 0.9883, + "step": 222400 + }, + { + "epoch": 19.25, + "learning_rate": 3.0775361691068185e-05, + "loss": 0.9102, + "step": 222410 + }, + { + "epoch": 19.25, + "learning_rate": 3.077449536515637e-05, + "loss": 0.976, + "step": 222420 + }, + { + "epoch": 19.25, + "learning_rate": 3.077362903924457e-05, + "loss": 0.9411, + "step": 222430 + }, + { + "epoch": 19.25, + "learning_rate": 3.077276271333276e-05, + "loss": 0.9373, + "step": 222440 + }, + { + "epoch": 19.25, + "learning_rate": 3.077189638742095e-05, + "loss": 0.9509, + "step": 222450 + }, + { + "epoch": 19.26, + "learning_rate": 3.077103006150914e-05, + "loss": 0.9445, + "step": 222460 + }, + { + "epoch": 19.26, + "learning_rate": 3.077016373559734e-05, + "loss": 0.9495, + "step": 222470 + }, + { + "epoch": 19.26, + "learning_rate": 3.0769297409685525e-05, + "loss": 0.9741, + "step": 222480 + }, + { + "epoch": 19.26, + "learning_rate": 3.076843108377372e-05, + "loss": 0.9234, + "step": 222490 + }, + { + "epoch": 19.26, + "learning_rate": 3.076756475786191e-05, + "loss": 0.9388, + "step": 222500 + }, + { + "epoch": 19.26, + "learning_rate": 3.07666984319501e-05, + "loss": 0.9369, + "step": 222510 + }, + { + "epoch": 19.26, + "learning_rate": 3.0765832106038297e-05, + "loss": 0.9463, + "step": 222520 + }, + { + "epoch": 19.26, + "learning_rate": 3.0764965780126484e-05, + "loss": 1.0027, + "step": 222530 + }, + { + "epoch": 19.26, + "learning_rate": 3.076409945421468e-05, + "loss": 0.9192, + "step": 222540 + }, + { + "epoch": 19.26, + "learning_rate": 3.076323312830287e-05, + "loss": 1.0004, + "step": 222550 + }, + { + "epoch": 19.26, + "learning_rate": 3.076236680239106e-05, + "loss": 0.9549, + "step": 222560 + }, + { + "epoch": 19.26, + "learning_rate": 3.0761500476479255e-05, + "loss": 0.9245, + "step": 222570 + }, + { + "epoch": 19.27, + "learning_rate": 3.076063415056745e-05, + "loss": 0.9887, + "step": 222580 + }, + { + "epoch": 19.27, + "learning_rate": 3.075976782465564e-05, + "loss": 0.9426, + "step": 222590 + }, + { + "epoch": 19.27, + "learning_rate": 3.075890149874383e-05, + "loss": 1.0202, + "step": 222600 + }, + { + "epoch": 19.27, + "learning_rate": 3.075803517283202e-05, + "loss": 0.9995, + "step": 222610 + }, + { + "epoch": 19.27, + "learning_rate": 3.0757168846920214e-05, + "loss": 0.945, + "step": 222620 + }, + { + "epoch": 19.27, + "learning_rate": 3.075630252100841e-05, + "loss": 0.9396, + "step": 222630 + }, + { + "epoch": 19.27, + "learning_rate": 3.0755436195096596e-05, + "loss": 0.9213, + "step": 222640 + }, + { + "epoch": 19.27, + "learning_rate": 3.075456986918479e-05, + "loss": 0.9736, + "step": 222650 + }, + { + "epoch": 19.27, + "learning_rate": 3.0753703543272985e-05, + "loss": 0.959, + "step": 222660 + }, + { + "epoch": 19.27, + "learning_rate": 3.075283721736117e-05, + "loss": 0.9825, + "step": 222670 + }, + { + "epoch": 19.27, + "learning_rate": 3.075197089144937e-05, + "loss": 1.0228, + "step": 222680 + }, + { + "epoch": 19.27, + "learning_rate": 3.075110456553756e-05, + "loss": 1.0161, + "step": 222690 + }, + { + "epoch": 19.28, + "learning_rate": 3.075023823962575e-05, + "loss": 0.9282, + "step": 222700 + }, + { + "epoch": 19.28, + "learning_rate": 3.0749371913713944e-05, + "loss": 1.0006, + "step": 222710 + }, + { + "epoch": 19.28, + "learning_rate": 3.074850558780213e-05, + "loss": 0.8992, + "step": 222720 + }, + { + "epoch": 19.28, + "learning_rate": 3.0747639261890326e-05, + "loss": 0.9888, + "step": 222730 + }, + { + "epoch": 19.28, + "learning_rate": 3.074677293597852e-05, + "loss": 0.9256, + "step": 222740 + }, + { + "epoch": 19.28, + "learning_rate": 3.074590661006671e-05, + "loss": 0.9793, + "step": 222750 + }, + { + "epoch": 19.28, + "learning_rate": 3.07450402841549e-05, + "loss": 1.0102, + "step": 222760 + }, + { + "epoch": 19.28, + "learning_rate": 3.07441739582431e-05, + "loss": 0.9574, + "step": 222770 + }, + { + "epoch": 19.28, + "learning_rate": 3.0743307632331285e-05, + "loss": 0.9437, + "step": 222780 + }, + { + "epoch": 19.28, + "learning_rate": 3.074244130641948e-05, + "loss": 1.0068, + "step": 222790 + }, + { + "epoch": 19.28, + "learning_rate": 3.0741574980507673e-05, + "loss": 0.9751, + "step": 222800 + }, + { + "epoch": 19.29, + "learning_rate": 3.074070865459586e-05, + "loss": 1.0276, + "step": 222810 + }, + { + "epoch": 19.29, + "learning_rate": 3.0739842328684056e-05, + "loss": 0.9903, + "step": 222820 + }, + { + "epoch": 19.29, + "learning_rate": 3.073897600277224e-05, + "loss": 0.9295, + "step": 222830 + }, + { + "epoch": 19.29, + "learning_rate": 3.073810967686044e-05, + "loss": 0.9652, + "step": 222840 + }, + { + "epoch": 19.29, + "learning_rate": 3.073724335094863e-05, + "loss": 0.9812, + "step": 222850 + }, + { + "epoch": 19.29, + "learning_rate": 3.073637702503682e-05, + "loss": 0.9823, + "step": 222860 + }, + { + "epoch": 19.29, + "learning_rate": 3.0735510699125014e-05, + "loss": 0.945, + "step": 222870 + }, + { + "epoch": 19.29, + "learning_rate": 3.073464437321321e-05, + "loss": 0.9581, + "step": 222880 + }, + { + "epoch": 19.29, + "learning_rate": 3.0733778047301396e-05, + "loss": 0.9695, + "step": 222890 + }, + { + "epoch": 19.29, + "learning_rate": 3.073291172138959e-05, + "loss": 0.9174, + "step": 222900 + }, + { + "epoch": 19.29, + "learning_rate": 3.0732045395477785e-05, + "loss": 0.9697, + "step": 222910 + }, + { + "epoch": 19.29, + "learning_rate": 3.073117906956597e-05, + "loss": 0.969, + "step": 222920 + }, + { + "epoch": 19.3, + "learning_rate": 3.073031274365417e-05, + "loss": 0.9792, + "step": 222930 + }, + { + "epoch": 19.3, + "learning_rate": 3.0729446417742355e-05, + "loss": 0.9531, + "step": 222940 + }, + { + "epoch": 19.3, + "learning_rate": 3.072858009183055e-05, + "loss": 0.9582, + "step": 222950 + }, + { + "epoch": 19.3, + "learning_rate": 3.0727713765918744e-05, + "loss": 0.9018, + "step": 222960 + }, + { + "epoch": 19.3, + "learning_rate": 3.072684744000693e-05, + "loss": 0.9466, + "step": 222970 + }, + { + "epoch": 19.3, + "learning_rate": 3.0725981114095126e-05, + "loss": 0.9694, + "step": 222980 + }, + { + "epoch": 19.3, + "learning_rate": 3.072511478818332e-05, + "loss": 0.9419, + "step": 222990 + }, + { + "epoch": 19.3, + "learning_rate": 3.072424846227151e-05, + "loss": 0.9716, + "step": 223000 + }, + { + "epoch": 19.3, + "learning_rate": 3.07233821363597e-05, + "loss": 0.944, + "step": 223010 + }, + { + "epoch": 19.3, + "learning_rate": 3.07225158104479e-05, + "loss": 0.9724, + "step": 223020 + }, + { + "epoch": 19.3, + "learning_rate": 3.0721649484536085e-05, + "loss": 0.86, + "step": 223030 + }, + { + "epoch": 19.31, + "learning_rate": 3.072078315862428e-05, + "loss": 0.8679, + "step": 223040 + }, + { + "epoch": 19.31, + "learning_rate": 3.071991683271247e-05, + "loss": 0.9624, + "step": 223050 + }, + { + "epoch": 19.31, + "learning_rate": 3.071905050680066e-05, + "loss": 0.9818, + "step": 223060 + }, + { + "epoch": 19.31, + "learning_rate": 3.0718184180888856e-05, + "loss": 0.9652, + "step": 223070 + }, + { + "epoch": 19.31, + "learning_rate": 3.0717317854977044e-05, + "loss": 0.9304, + "step": 223080 + }, + { + "epoch": 19.31, + "learning_rate": 3.071645152906524e-05, + "loss": 0.9764, + "step": 223090 + }, + { + "epoch": 19.31, + "learning_rate": 3.071558520315343e-05, + "loss": 0.9783, + "step": 223100 + }, + { + "epoch": 19.31, + "learning_rate": 3.071471887724162e-05, + "loss": 0.9418, + "step": 223110 + }, + { + "epoch": 19.31, + "learning_rate": 3.0713852551329815e-05, + "loss": 0.9891, + "step": 223120 + }, + { + "epoch": 19.31, + "learning_rate": 3.0712986225418e-05, + "loss": 0.9371, + "step": 223130 + }, + { + "epoch": 19.31, + "learning_rate": 3.07121198995062e-05, + "loss": 0.9758, + "step": 223140 + }, + { + "epoch": 19.31, + "learning_rate": 3.071125357359439e-05, + "loss": 0.9713, + "step": 223150 + }, + { + "epoch": 19.32, + "learning_rate": 3.071038724768258e-05, + "loss": 0.9706, + "step": 223160 + }, + { + "epoch": 19.32, + "learning_rate": 3.070952092177077e-05, + "loss": 1.0123, + "step": 223170 + }, + { + "epoch": 19.32, + "learning_rate": 3.070865459585897e-05, + "loss": 0.9969, + "step": 223180 + }, + { + "epoch": 19.32, + "learning_rate": 3.0707788269947155e-05, + "loss": 0.9363, + "step": 223190 + }, + { + "epoch": 19.32, + "learning_rate": 3.070692194403535e-05, + "loss": 1.03, + "step": 223200 + }, + { + "epoch": 19.32, + "learning_rate": 3.0706055618123544e-05, + "loss": 0.963, + "step": 223210 + }, + { + "epoch": 19.32, + "learning_rate": 3.070518929221173e-05, + "loss": 0.9071, + "step": 223220 + }, + { + "epoch": 19.32, + "learning_rate": 3.0704322966299927e-05, + "loss": 0.9843, + "step": 223230 + }, + { + "epoch": 19.32, + "learning_rate": 3.0703456640388114e-05, + "loss": 0.9798, + "step": 223240 + }, + { + "epoch": 19.32, + "learning_rate": 3.070259031447631e-05, + "loss": 0.9855, + "step": 223250 + }, + { + "epoch": 19.32, + "learning_rate": 3.07017239885645e-05, + "loss": 0.9481, + "step": 223260 + }, + { + "epoch": 19.33, + "learning_rate": 3.070085766265269e-05, + "loss": 0.9322, + "step": 223270 + }, + { + "epoch": 19.33, + "learning_rate": 3.0699991336740885e-05, + "loss": 0.9746, + "step": 223280 + }, + { + "epoch": 19.33, + "learning_rate": 3.069912501082908e-05, + "loss": 0.9691, + "step": 223290 + }, + { + "epoch": 19.33, + "learning_rate": 3.069825868491727e-05, + "loss": 0.9665, + "step": 223300 + }, + { + "epoch": 19.33, + "learning_rate": 3.069739235900546e-05, + "loss": 0.9757, + "step": 223310 + }, + { + "epoch": 19.33, + "learning_rate": 3.0696526033093656e-05, + "loss": 0.9684, + "step": 223320 + }, + { + "epoch": 19.33, + "learning_rate": 3.0695659707181844e-05, + "loss": 0.9567, + "step": 223330 + }, + { + "epoch": 19.33, + "learning_rate": 3.069479338127004e-05, + "loss": 0.9742, + "step": 223340 + }, + { + "epoch": 19.33, + "learning_rate": 3.0693927055358226e-05, + "loss": 0.9938, + "step": 223350 + }, + { + "epoch": 19.33, + "learning_rate": 3.069306072944642e-05, + "loss": 0.9577, + "step": 223360 + }, + { + "epoch": 19.33, + "learning_rate": 3.0692194403534615e-05, + "loss": 0.9648, + "step": 223370 + }, + { + "epoch": 19.33, + "learning_rate": 3.06913280776228e-05, + "loss": 0.9396, + "step": 223380 + }, + { + "epoch": 19.34, + "learning_rate": 3.0690461751711e-05, + "loss": 0.96, + "step": 223390 + }, + { + "epoch": 19.34, + "learning_rate": 3.068959542579919e-05, + "loss": 1.0151, + "step": 223400 + }, + { + "epoch": 19.34, + "learning_rate": 3.068872909988738e-05, + "loss": 0.9879, + "step": 223410 + }, + { + "epoch": 19.34, + "learning_rate": 3.0687862773975574e-05, + "loss": 0.9205, + "step": 223420 + }, + { + "epoch": 19.34, + "learning_rate": 3.068699644806377e-05, + "loss": 1.0378, + "step": 223430 + }, + { + "epoch": 19.34, + "learning_rate": 3.0686130122151956e-05, + "loss": 1.011, + "step": 223440 + }, + { + "epoch": 19.34, + "learning_rate": 3.068526379624015e-05, + "loss": 0.9869, + "step": 223450 + }, + { + "epoch": 19.34, + "learning_rate": 3.068439747032834e-05, + "loss": 0.9401, + "step": 223460 + }, + { + "epoch": 19.34, + "learning_rate": 3.068353114441653e-05, + "loss": 0.9882, + "step": 223470 + }, + { + "epoch": 19.34, + "learning_rate": 3.068266481850473e-05, + "loss": 0.9826, + "step": 223480 + }, + { + "epoch": 19.34, + "learning_rate": 3.0681798492592915e-05, + "loss": 1.0181, + "step": 223490 + }, + { + "epoch": 19.35, + "learning_rate": 3.068093216668111e-05, + "loss": 0.992, + "step": 223500 + }, + { + "epoch": 19.35, + "learning_rate": 3.0680065840769303e-05, + "loss": 0.972, + "step": 223510 + }, + { + "epoch": 19.35, + "learning_rate": 3.067919951485749e-05, + "loss": 0.9563, + "step": 223520 + }, + { + "epoch": 19.35, + "learning_rate": 3.0678333188945686e-05, + "loss": 0.9502, + "step": 223530 + }, + { + "epoch": 19.35, + "learning_rate": 3.067746686303388e-05, + "loss": 0.9342, + "step": 223540 + }, + { + "epoch": 19.35, + "learning_rate": 3.067660053712207e-05, + "loss": 0.9799, + "step": 223550 + }, + { + "epoch": 19.35, + "learning_rate": 3.067573421121026e-05, + "loss": 0.965, + "step": 223560 + }, + { + "epoch": 19.35, + "learning_rate": 3.067486788529845e-05, + "loss": 0.9959, + "step": 223570 + }, + { + "epoch": 19.35, + "learning_rate": 3.0674001559386644e-05, + "loss": 0.9836, + "step": 223580 + }, + { + "epoch": 19.35, + "learning_rate": 3.067313523347484e-05, + "loss": 0.9653, + "step": 223590 + }, + { + "epoch": 19.35, + "learning_rate": 3.0672268907563026e-05, + "loss": 0.9462, + "step": 223600 + }, + { + "epoch": 19.35, + "learning_rate": 3.067140258165122e-05, + "loss": 0.9733, + "step": 223610 + }, + { + "epoch": 19.36, + "learning_rate": 3.0670536255739415e-05, + "loss": 0.9495, + "step": 223620 + }, + { + "epoch": 19.36, + "learning_rate": 3.06696699298276e-05, + "loss": 0.9753, + "step": 223630 + }, + { + "epoch": 19.36, + "learning_rate": 3.06688036039158e-05, + "loss": 0.9475, + "step": 223640 + }, + { + "epoch": 19.36, + "learning_rate": 3.066793727800399e-05, + "loss": 0.9845, + "step": 223650 + }, + { + "epoch": 19.36, + "learning_rate": 3.066707095209218e-05, + "loss": 0.9459, + "step": 223660 + }, + { + "epoch": 19.36, + "learning_rate": 3.0666204626180374e-05, + "loss": 0.9262, + "step": 223670 + }, + { + "epoch": 19.36, + "learning_rate": 3.066533830026856e-05, + "loss": 0.9827, + "step": 223680 + }, + { + "epoch": 19.36, + "learning_rate": 3.0664471974356756e-05, + "loss": 0.9683, + "step": 223690 + }, + { + "epoch": 19.36, + "learning_rate": 3.066360564844495e-05, + "loss": 0.9352, + "step": 223700 + }, + { + "epoch": 19.36, + "learning_rate": 3.066273932253314e-05, + "loss": 0.9391, + "step": 223710 + }, + { + "epoch": 19.36, + "learning_rate": 3.066187299662133e-05, + "loss": 0.9227, + "step": 223720 + }, + { + "epoch": 19.36, + "learning_rate": 3.066100667070953e-05, + "loss": 0.9208, + "step": 223730 + }, + { + "epoch": 19.37, + "learning_rate": 3.0660140344797715e-05, + "loss": 0.9024, + "step": 223740 + }, + { + "epoch": 19.37, + "learning_rate": 3.065927401888591e-05, + "loss": 0.9776, + "step": 223750 + }, + { + "epoch": 19.37, + "learning_rate": 3.0658407692974104e-05, + "loss": 0.9436, + "step": 223760 + }, + { + "epoch": 19.37, + "learning_rate": 3.065754136706229e-05, + "loss": 0.9726, + "step": 223770 + }, + { + "epoch": 19.37, + "learning_rate": 3.0656675041150486e-05, + "loss": 0.9249, + "step": 223780 + }, + { + "epoch": 19.37, + "learning_rate": 3.0655808715238674e-05, + "loss": 0.958, + "step": 223790 + }, + { + "epoch": 19.37, + "learning_rate": 3.065494238932687e-05, + "loss": 0.9334, + "step": 223800 + }, + { + "epoch": 19.37, + "learning_rate": 3.065407606341506e-05, + "loss": 0.9723, + "step": 223810 + }, + { + "epoch": 19.37, + "learning_rate": 3.065320973750325e-05, + "loss": 1.027, + "step": 223820 + }, + { + "epoch": 19.37, + "learning_rate": 3.0652343411591445e-05, + "loss": 0.9879, + "step": 223830 + }, + { + "epoch": 19.37, + "learning_rate": 3.065147708567964e-05, + "loss": 0.9955, + "step": 223840 + }, + { + "epoch": 19.38, + "learning_rate": 3.065061075976783e-05, + "loss": 1.0247, + "step": 223850 + }, + { + "epoch": 19.38, + "learning_rate": 3.064974443385602e-05, + "loss": 0.9543, + "step": 223860 + }, + { + "epoch": 19.38, + "learning_rate": 3.064887810794421e-05, + "loss": 0.9715, + "step": 223870 + }, + { + "epoch": 19.38, + "learning_rate": 3.06480117820324e-05, + "loss": 0.9467, + "step": 223880 + }, + { + "epoch": 19.38, + "learning_rate": 3.06471454561206e-05, + "loss": 0.9715, + "step": 223890 + }, + { + "epoch": 19.38, + "learning_rate": 3.0646279130208785e-05, + "loss": 0.9593, + "step": 223900 + }, + { + "epoch": 19.38, + "learning_rate": 3.064541280429698e-05, + "loss": 1.0172, + "step": 223910 + }, + { + "epoch": 19.38, + "learning_rate": 3.0644546478385174e-05, + "loss": 0.9476, + "step": 223920 + }, + { + "epoch": 19.38, + "learning_rate": 3.064368015247336e-05, + "loss": 0.976, + "step": 223930 + }, + { + "epoch": 19.38, + "learning_rate": 3.0642813826561556e-05, + "loss": 0.964, + "step": 223940 + }, + { + "epoch": 19.38, + "learning_rate": 3.064194750064975e-05, + "loss": 0.9775, + "step": 223950 + }, + { + "epoch": 19.38, + "learning_rate": 3.064108117473794e-05, + "loss": 1.0039, + "step": 223960 + }, + { + "epoch": 19.39, + "learning_rate": 3.064021484882613e-05, + "loss": 0.9307, + "step": 223970 + }, + { + "epoch": 19.39, + "learning_rate": 3.063934852291432e-05, + "loss": 0.9419, + "step": 223980 + }, + { + "epoch": 19.39, + "learning_rate": 3.0638482197002515e-05, + "loss": 0.9398, + "step": 223990 + }, + { + "epoch": 19.39, + "learning_rate": 3.063761587109071e-05, + "loss": 0.924, + "step": 224000 + }, + { + "epoch": 19.39, + "learning_rate": 3.06367495451789e-05, + "loss": 0.9831, + "step": 224010 + }, + { + "epoch": 19.39, + "learning_rate": 3.063588321926709e-05, + "loss": 0.9493, + "step": 224020 + }, + { + "epoch": 19.39, + "learning_rate": 3.0635016893355286e-05, + "loss": 0.9637, + "step": 224030 + }, + { + "epoch": 19.39, + "learning_rate": 3.0634150567443474e-05, + "loss": 0.9737, + "step": 224040 + }, + { + "epoch": 19.39, + "learning_rate": 3.063328424153167e-05, + "loss": 0.9929, + "step": 224050 + }, + { + "epoch": 19.39, + "learning_rate": 3.063241791561986e-05, + "loss": 0.9721, + "step": 224060 + }, + { + "epoch": 19.39, + "learning_rate": 3.063155158970805e-05, + "loss": 0.9823, + "step": 224070 + }, + { + "epoch": 19.4, + "learning_rate": 3.0630685263796245e-05, + "loss": 0.9712, + "step": 224080 + }, + { + "epoch": 19.4, + "learning_rate": 3.062981893788443e-05, + "loss": 1.0028, + "step": 224090 + }, + { + "epoch": 19.4, + "learning_rate": 3.062895261197263e-05, + "loss": 0.9065, + "step": 224100 + }, + { + "epoch": 19.4, + "learning_rate": 3.062808628606082e-05, + "loss": 0.9725, + "step": 224110 + }, + { + "epoch": 19.4, + "learning_rate": 3.062721996014901e-05, + "loss": 0.9511, + "step": 224120 + }, + { + "epoch": 19.4, + "learning_rate": 3.0626353634237204e-05, + "loss": 0.9379, + "step": 224130 + }, + { + "epoch": 19.4, + "learning_rate": 3.06254873083254e-05, + "loss": 1.0441, + "step": 224140 + }, + { + "epoch": 19.4, + "learning_rate": 3.0624620982413586e-05, + "loss": 0.9603, + "step": 224150 + }, + { + "epoch": 19.4, + "learning_rate": 3.062375465650178e-05, + "loss": 0.9744, + "step": 224160 + }, + { + "epoch": 19.4, + "learning_rate": 3.0622888330589975e-05, + "loss": 0.9328, + "step": 224170 + }, + { + "epoch": 19.4, + "learning_rate": 3.062202200467816e-05, + "loss": 0.9852, + "step": 224180 + }, + { + "epoch": 19.4, + "learning_rate": 3.062115567876636e-05, + "loss": 0.9505, + "step": 224190 + }, + { + "epoch": 19.41, + "learning_rate": 3.0620289352854544e-05, + "loss": 0.9863, + "step": 224200 + }, + { + "epoch": 19.41, + "learning_rate": 3.061942302694274e-05, + "loss": 0.9723, + "step": 224210 + }, + { + "epoch": 19.41, + "learning_rate": 3.0618556701030933e-05, + "loss": 0.971, + "step": 224220 + }, + { + "epoch": 19.41, + "learning_rate": 3.061769037511912e-05, + "loss": 0.9656, + "step": 224230 + }, + { + "epoch": 19.41, + "learning_rate": 3.0616824049207316e-05, + "loss": 0.9618, + "step": 224240 + }, + { + "epoch": 19.41, + "learning_rate": 3.061595772329551e-05, + "loss": 0.9253, + "step": 224250 + }, + { + "epoch": 19.41, + "learning_rate": 3.06150913973837e-05, + "loss": 0.8889, + "step": 224260 + }, + { + "epoch": 19.41, + "learning_rate": 3.061422507147189e-05, + "loss": 0.9172, + "step": 224270 + }, + { + "epoch": 19.41, + "learning_rate": 3.0613358745560087e-05, + "loss": 0.977, + "step": 224280 + }, + { + "epoch": 19.41, + "learning_rate": 3.0612492419648274e-05, + "loss": 0.9799, + "step": 224290 + }, + { + "epoch": 19.41, + "learning_rate": 3.061162609373647e-05, + "loss": 0.9508, + "step": 224300 + }, + { + "epoch": 19.42, + "learning_rate": 3.0610759767824656e-05, + "loss": 0.929, + "step": 224310 + }, + { + "epoch": 19.42, + "learning_rate": 3.060989344191285e-05, + "loss": 1.0121, + "step": 224320 + }, + { + "epoch": 19.42, + "learning_rate": 3.0609027116001045e-05, + "loss": 0.9518, + "step": 224330 + }, + { + "epoch": 19.42, + "learning_rate": 3.060816079008923e-05, + "loss": 0.9484, + "step": 224340 + }, + { + "epoch": 19.42, + "learning_rate": 3.060729446417743e-05, + "loss": 0.9502, + "step": 224350 + }, + { + "epoch": 19.42, + "learning_rate": 3.060642813826562e-05, + "loss": 0.9688, + "step": 224360 + }, + { + "epoch": 19.42, + "learning_rate": 3.060556181235381e-05, + "loss": 0.9578, + "step": 224370 + }, + { + "epoch": 19.42, + "learning_rate": 3.0604695486442004e-05, + "loss": 0.9746, + "step": 224380 + }, + { + "epoch": 19.42, + "learning_rate": 3.06038291605302e-05, + "loss": 0.9765, + "step": 224390 + }, + { + "epoch": 19.42, + "learning_rate": 3.0602962834618386e-05, + "loss": 1.0269, + "step": 224400 + }, + { + "epoch": 19.42, + "learning_rate": 3.060209650870658e-05, + "loss": 0.939, + "step": 224410 + }, + { + "epoch": 19.42, + "learning_rate": 3.060123018279477e-05, + "loss": 0.9728, + "step": 224420 + }, + { + "epoch": 19.43, + "learning_rate": 3.060036385688296e-05, + "loss": 0.9794, + "step": 224430 + }, + { + "epoch": 19.43, + "learning_rate": 3.059949753097116e-05, + "loss": 0.9792, + "step": 224440 + }, + { + "epoch": 19.43, + "learning_rate": 3.0598631205059345e-05, + "loss": 0.9586, + "step": 224450 + }, + { + "epoch": 19.43, + "learning_rate": 3.059776487914754e-05, + "loss": 0.9499, + "step": 224460 + }, + { + "epoch": 19.43, + "learning_rate": 3.0596898553235734e-05, + "loss": 1.0007, + "step": 224470 + }, + { + "epoch": 19.43, + "learning_rate": 3.059603222732392e-05, + "loss": 0.9496, + "step": 224480 + }, + { + "epoch": 19.43, + "learning_rate": 3.0595165901412116e-05, + "loss": 0.9855, + "step": 224490 + }, + { + "epoch": 19.43, + "learning_rate": 3.059429957550031e-05, + "loss": 0.9911, + "step": 224500 + }, + { + "epoch": 19.43, + "learning_rate": 3.05934332495885e-05, + "loss": 0.9517, + "step": 224510 + }, + { + "epoch": 19.43, + "learning_rate": 3.059256692367669e-05, + "loss": 1.0007, + "step": 224520 + }, + { + "epoch": 19.43, + "learning_rate": 3.059170059776488e-05, + "loss": 0.9866, + "step": 224530 + }, + { + "epoch": 19.44, + "learning_rate": 3.0590834271853075e-05, + "loss": 1.004, + "step": 224540 + }, + { + "epoch": 19.44, + "learning_rate": 3.058996794594127e-05, + "loss": 0.9712, + "step": 224550 + }, + { + "epoch": 19.44, + "learning_rate": 3.058910162002946e-05, + "loss": 0.9506, + "step": 224560 + }, + { + "epoch": 19.44, + "learning_rate": 3.058823529411765e-05, + "loss": 1.0025, + "step": 224570 + }, + { + "epoch": 19.44, + "learning_rate": 3.0587368968205846e-05, + "loss": 0.9562, + "step": 224580 + }, + { + "epoch": 19.44, + "learning_rate": 3.058650264229403e-05, + "loss": 0.8881, + "step": 224590 + }, + { + "epoch": 19.44, + "learning_rate": 3.058563631638223e-05, + "loss": 0.9664, + "step": 224600 + }, + { + "epoch": 19.44, + "learning_rate": 3.0584769990470415e-05, + "loss": 0.9649, + "step": 224610 + }, + { + "epoch": 19.44, + "learning_rate": 3.058390366455861e-05, + "loss": 0.925, + "step": 224620 + }, + { + "epoch": 19.44, + "learning_rate": 3.0583037338646804e-05, + "loss": 0.9637, + "step": 224630 + }, + { + "epoch": 19.44, + "learning_rate": 3.058217101273499e-05, + "loss": 0.9699, + "step": 224640 + }, + { + "epoch": 19.44, + "learning_rate": 3.0581304686823186e-05, + "loss": 1.0017, + "step": 224650 + }, + { + "epoch": 19.45, + "learning_rate": 3.058043836091138e-05, + "loss": 0.9673, + "step": 224660 + }, + { + "epoch": 19.45, + "learning_rate": 3.057957203499957e-05, + "loss": 0.9302, + "step": 224670 + }, + { + "epoch": 19.45, + "learning_rate": 3.057870570908776e-05, + "loss": 0.9869, + "step": 224680 + }, + { + "epoch": 19.45, + "learning_rate": 3.057783938317596e-05, + "loss": 0.9528, + "step": 224690 + }, + { + "epoch": 19.45, + "learning_rate": 3.0576973057264145e-05, + "loss": 0.9463, + "step": 224700 + }, + { + "epoch": 19.45, + "learning_rate": 3.057610673135234e-05, + "loss": 0.931, + "step": 224710 + }, + { + "epoch": 19.45, + "learning_rate": 3.057524040544053e-05, + "loss": 0.9781, + "step": 224720 + }, + { + "epoch": 19.45, + "learning_rate": 3.057437407952872e-05, + "loss": 0.9869, + "step": 224730 + }, + { + "epoch": 19.45, + "learning_rate": 3.0573507753616916e-05, + "loss": 0.9753, + "step": 224740 + }, + { + "epoch": 19.45, + "learning_rate": 3.0572641427705104e-05, + "loss": 0.9349, + "step": 224750 + }, + { + "epoch": 19.45, + "learning_rate": 3.05717751017933e-05, + "loss": 0.9868, + "step": 224760 + }, + { + "epoch": 19.45, + "learning_rate": 3.057090877588149e-05, + "loss": 0.8805, + "step": 224770 + }, + { + "epoch": 19.46, + "learning_rate": 3.057004244996968e-05, + "loss": 0.9082, + "step": 224780 + }, + { + "epoch": 19.46, + "learning_rate": 3.0569176124057875e-05, + "loss": 0.9643, + "step": 224790 + }, + { + "epoch": 19.46, + "learning_rate": 3.056830979814607e-05, + "loss": 0.9484, + "step": 224800 + }, + { + "epoch": 19.46, + "learning_rate": 3.056744347223426e-05, + "loss": 0.9821, + "step": 224810 + }, + { + "epoch": 19.46, + "learning_rate": 3.056657714632245e-05, + "loss": 0.9635, + "step": 224820 + }, + { + "epoch": 19.46, + "learning_rate": 3.056571082041064e-05, + "loss": 0.8968, + "step": 224830 + }, + { + "epoch": 19.46, + "learning_rate": 3.0564844494498834e-05, + "loss": 0.959, + "step": 224840 + }, + { + "epoch": 19.46, + "learning_rate": 3.056397816858703e-05, + "loss": 0.9611, + "step": 224850 + }, + { + "epoch": 19.46, + "learning_rate": 3.0563111842675216e-05, + "loss": 0.9545, + "step": 224860 + }, + { + "epoch": 19.46, + "learning_rate": 3.056224551676341e-05, + "loss": 0.9756, + "step": 224870 + }, + { + "epoch": 19.46, + "learning_rate": 3.0561379190851605e-05, + "loss": 0.9211, + "step": 224880 + }, + { + "epoch": 19.47, + "learning_rate": 3.056051286493979e-05, + "loss": 0.9916, + "step": 224890 + }, + { + "epoch": 19.47, + "learning_rate": 3.055964653902799e-05, + "loss": 0.96, + "step": 224900 + }, + { + "epoch": 19.47, + "learning_rate": 3.055878021311618e-05, + "loss": 1.0325, + "step": 224910 + }, + { + "epoch": 19.47, + "learning_rate": 3.055791388720437e-05, + "loss": 0.9623, + "step": 224920 + }, + { + "epoch": 19.47, + "learning_rate": 3.055704756129256e-05, + "loss": 0.9593, + "step": 224930 + }, + { + "epoch": 19.47, + "learning_rate": 3.055618123538075e-05, + "loss": 1.0202, + "step": 224940 + }, + { + "epoch": 19.47, + "learning_rate": 3.0555314909468945e-05, + "loss": 0.9183, + "step": 224950 + }, + { + "epoch": 19.47, + "learning_rate": 3.055444858355714e-05, + "loss": 0.9675, + "step": 224960 + }, + { + "epoch": 19.47, + "learning_rate": 3.055358225764533e-05, + "loss": 1.0039, + "step": 224970 + }, + { + "epoch": 19.47, + "learning_rate": 3.055271593173352e-05, + "loss": 0.9594, + "step": 224980 + }, + { + "epoch": 19.47, + "learning_rate": 3.0551849605821717e-05, + "loss": 0.9374, + "step": 224990 + }, + { + "epoch": 19.47, + "learning_rate": 3.0550983279909904e-05, + "loss": 0.9691, + "step": 225000 + }, + { + "epoch": 19.48, + "learning_rate": 3.05501169539981e-05, + "loss": 0.9704, + "step": 225010 + }, + { + "epoch": 19.48, + "learning_rate": 3.054925062808629e-05, + "loss": 1.0056, + "step": 225020 + }, + { + "epoch": 19.48, + "learning_rate": 3.054838430217448e-05, + "loss": 0.971, + "step": 225030 + }, + { + "epoch": 19.48, + "learning_rate": 3.0547517976262675e-05, + "loss": 0.9753, + "step": 225040 + }, + { + "epoch": 19.48, + "learning_rate": 3.054665165035086e-05, + "loss": 0.9085, + "step": 225050 + }, + { + "epoch": 19.48, + "learning_rate": 3.054578532443906e-05, + "loss": 0.8765, + "step": 225060 + }, + { + "epoch": 19.48, + "learning_rate": 3.054491899852725e-05, + "loss": 0.9589, + "step": 225070 + }, + { + "epoch": 19.48, + "learning_rate": 3.054405267261544e-05, + "loss": 0.9669, + "step": 225080 + }, + { + "epoch": 19.48, + "learning_rate": 3.0543186346703634e-05, + "loss": 0.8978, + "step": 225090 + }, + { + "epoch": 19.48, + "learning_rate": 3.054232002079183e-05, + "loss": 0.9761, + "step": 225100 + }, + { + "epoch": 19.48, + "learning_rate": 3.0541453694880016e-05, + "loss": 0.9628, + "step": 225110 + }, + { + "epoch": 19.49, + "learning_rate": 3.054058736896821e-05, + "loss": 0.9312, + "step": 225120 + }, + { + "epoch": 19.49, + "learning_rate": 3.0539721043056405e-05, + "loss": 0.9539, + "step": 225130 + }, + { + "epoch": 19.49, + "learning_rate": 3.053885471714459e-05, + "loss": 1.0143, + "step": 225140 + }, + { + "epoch": 19.49, + "learning_rate": 3.053798839123279e-05, + "loss": 0.9467, + "step": 225150 + }, + { + "epoch": 19.49, + "learning_rate": 3.0537122065320975e-05, + "loss": 0.9591, + "step": 225160 + }, + { + "epoch": 19.49, + "learning_rate": 3.053625573940917e-05, + "loss": 0.9568, + "step": 225170 + }, + { + "epoch": 19.49, + "learning_rate": 3.0535389413497364e-05, + "loss": 0.9265, + "step": 225180 + }, + { + "epoch": 19.49, + "learning_rate": 3.053452308758555e-05, + "loss": 1.0312, + "step": 225190 + }, + { + "epoch": 19.49, + "learning_rate": 3.0533656761673746e-05, + "loss": 0.9548, + "step": 225200 + }, + { + "epoch": 19.49, + "learning_rate": 3.053279043576194e-05, + "loss": 0.935, + "step": 225210 + }, + { + "epoch": 19.49, + "learning_rate": 3.053192410985013e-05, + "loss": 0.9887, + "step": 225220 + }, + { + "epoch": 19.49, + "learning_rate": 3.053105778393832e-05, + "loss": 0.9557, + "step": 225230 + }, + { + "epoch": 19.5, + "learning_rate": 3.053019145802652e-05, + "loss": 0.9664, + "step": 225240 + }, + { + "epoch": 19.5, + "learning_rate": 3.0529325132114705e-05, + "loss": 0.9911, + "step": 225250 + }, + { + "epoch": 19.5, + "learning_rate": 3.05284588062029e-05, + "loss": 0.9656, + "step": 225260 + }, + { + "epoch": 19.5, + "learning_rate": 3.052759248029109e-05, + "loss": 0.9241, + "step": 225270 + }, + { + "epoch": 19.5, + "learning_rate": 3.052672615437928e-05, + "loss": 0.9813, + "step": 225280 + }, + { + "epoch": 19.5, + "learning_rate": 3.0525859828467476e-05, + "loss": 0.9605, + "step": 225290 + }, + { + "epoch": 19.5, + "learning_rate": 3.052499350255566e-05, + "loss": 0.9618, + "step": 225300 + }, + { + "epoch": 19.5, + "learning_rate": 3.052412717664386e-05, + "loss": 0.9336, + "step": 225310 + }, + { + "epoch": 19.5, + "learning_rate": 3.052326085073205e-05, + "loss": 0.9449, + "step": 225320 + }, + { + "epoch": 19.5, + "learning_rate": 3.052239452482024e-05, + "loss": 0.9244, + "step": 225330 + }, + { + "epoch": 19.5, + "learning_rate": 3.0521528198908434e-05, + "loss": 0.909, + "step": 225340 + }, + { + "epoch": 19.51, + "learning_rate": 3.052066187299663e-05, + "loss": 0.9645, + "step": 225350 + }, + { + "epoch": 19.51, + "learning_rate": 3.0519795547084816e-05, + "loss": 0.9204, + "step": 225360 + }, + { + "epoch": 19.51, + "learning_rate": 3.051892922117301e-05, + "loss": 0.9897, + "step": 225370 + }, + { + "epoch": 19.51, + "learning_rate": 3.05180628952612e-05, + "loss": 0.9037, + "step": 225380 + }, + { + "epoch": 19.51, + "learning_rate": 3.0517196569349393e-05, + "loss": 0.9533, + "step": 225390 + }, + { + "epoch": 19.51, + "learning_rate": 3.0516330243437584e-05, + "loss": 0.9477, + "step": 225400 + }, + { + "epoch": 19.51, + "learning_rate": 3.051546391752578e-05, + "loss": 0.972, + "step": 225410 + }, + { + "epoch": 19.51, + "learning_rate": 3.051459759161397e-05, + "loss": 0.95, + "step": 225420 + }, + { + "epoch": 19.51, + "learning_rate": 3.051373126570216e-05, + "loss": 1.0138, + "step": 225430 + }, + { + "epoch": 19.51, + "learning_rate": 3.0512864939790352e-05, + "loss": 1.0045, + "step": 225440 + }, + { + "epoch": 19.51, + "learning_rate": 3.0511998613878546e-05, + "loss": 0.9107, + "step": 225450 + }, + { + "epoch": 19.51, + "learning_rate": 3.0511132287966737e-05, + "loss": 0.9338, + "step": 225460 + }, + { + "epoch": 19.52, + "learning_rate": 3.0510265962054928e-05, + "loss": 0.9808, + "step": 225470 + }, + { + "epoch": 19.52, + "learning_rate": 3.0509399636143123e-05, + "loss": 0.9712, + "step": 225480 + }, + { + "epoch": 19.52, + "learning_rate": 3.0508533310231314e-05, + "loss": 1.058, + "step": 225490 + }, + { + "epoch": 19.52, + "learning_rate": 3.0507666984319505e-05, + "loss": 0.9878, + "step": 225500 + }, + { + "epoch": 19.52, + "learning_rate": 3.0506800658407696e-05, + "loss": 0.9179, + "step": 225510 + }, + { + "epoch": 19.52, + "learning_rate": 3.050593433249589e-05, + "loss": 0.9234, + "step": 225520 + }, + { + "epoch": 19.52, + "learning_rate": 3.050506800658408e-05, + "loss": 0.9646, + "step": 225530 + }, + { + "epoch": 19.52, + "learning_rate": 3.0504201680672273e-05, + "loss": 0.9639, + "step": 225540 + }, + { + "epoch": 19.52, + "learning_rate": 3.0503335354760464e-05, + "loss": 0.9738, + "step": 225550 + }, + { + "epoch": 19.52, + "learning_rate": 3.0502469028848658e-05, + "loss": 0.8915, + "step": 225560 + }, + { + "epoch": 19.52, + "learning_rate": 3.050160270293685e-05, + "loss": 0.9106, + "step": 225570 + }, + { + "epoch": 19.53, + "learning_rate": 3.050073637702504e-05, + "loss": 1.008, + "step": 225580 + }, + { + "epoch": 19.53, + "learning_rate": 3.049987005111323e-05, + "loss": 0.9695, + "step": 225590 + }, + { + "epoch": 19.53, + "learning_rate": 3.0499003725201426e-05, + "loss": 0.9455, + "step": 225600 + }, + { + "epoch": 19.53, + "learning_rate": 3.0498137399289617e-05, + "loss": 0.9613, + "step": 225610 + }, + { + "epoch": 19.53, + "learning_rate": 3.0497271073377808e-05, + "loss": 0.9797, + "step": 225620 + }, + { + "epoch": 19.53, + "learning_rate": 3.0496404747466002e-05, + "loss": 0.983, + "step": 225630 + }, + { + "epoch": 19.53, + "learning_rate": 3.0495538421554193e-05, + "loss": 0.9547, + "step": 225640 + }, + { + "epoch": 19.53, + "learning_rate": 3.0494672095642384e-05, + "loss": 0.9253, + "step": 225650 + }, + { + "epoch": 19.53, + "learning_rate": 3.0493805769730575e-05, + "loss": 0.9635, + "step": 225660 + }, + { + "epoch": 19.53, + "learning_rate": 3.049293944381877e-05, + "loss": 0.9136, + "step": 225670 + }, + { + "epoch": 19.53, + "learning_rate": 3.049207311790696e-05, + "loss": 1.0106, + "step": 225680 + }, + { + "epoch": 19.53, + "learning_rate": 3.0491206791995152e-05, + "loss": 0.982, + "step": 225690 + }, + { + "epoch": 19.54, + "learning_rate": 3.0490340466083343e-05, + "loss": 0.9155, + "step": 225700 + }, + { + "epoch": 19.54, + "learning_rate": 3.0489474140171538e-05, + "loss": 0.9656, + "step": 225710 + }, + { + "epoch": 19.54, + "learning_rate": 3.048860781425973e-05, + "loss": 0.9996, + "step": 225720 + }, + { + "epoch": 19.54, + "learning_rate": 3.048774148834792e-05, + "loss": 0.928, + "step": 225730 + }, + { + "epoch": 19.54, + "learning_rate": 3.0486875162436114e-05, + "loss": 0.9271, + "step": 225740 + }, + { + "epoch": 19.54, + "learning_rate": 3.0486008836524305e-05, + "loss": 0.9837, + "step": 225750 + }, + { + "epoch": 19.54, + "learning_rate": 3.0485142510612496e-05, + "loss": 0.9917, + "step": 225760 + }, + { + "epoch": 19.54, + "learning_rate": 3.0484276184700687e-05, + "loss": 0.9451, + "step": 225770 + }, + { + "epoch": 19.54, + "learning_rate": 3.0483409858788882e-05, + "loss": 1.0076, + "step": 225780 + }, + { + "epoch": 19.54, + "learning_rate": 3.0482543532877073e-05, + "loss": 0.953, + "step": 225790 + }, + { + "epoch": 19.54, + "learning_rate": 3.0481677206965264e-05, + "loss": 0.9352, + "step": 225800 + }, + { + "epoch": 19.54, + "learning_rate": 3.0480810881053455e-05, + "loss": 0.9305, + "step": 225810 + }, + { + "epoch": 19.55, + "learning_rate": 3.047994455514165e-05, + "loss": 0.9994, + "step": 225820 + }, + { + "epoch": 19.55, + "learning_rate": 3.047907822922984e-05, + "loss": 0.9381, + "step": 225830 + }, + { + "epoch": 19.55, + "learning_rate": 3.047821190331803e-05, + "loss": 0.926, + "step": 225840 + }, + { + "epoch": 19.55, + "learning_rate": 3.0477345577406226e-05, + "loss": 0.9508, + "step": 225850 + }, + { + "epoch": 19.55, + "learning_rate": 3.0476479251494417e-05, + "loss": 0.9501, + "step": 225860 + }, + { + "epoch": 19.55, + "learning_rate": 3.0475612925582608e-05, + "loss": 0.957, + "step": 225870 + }, + { + "epoch": 19.55, + "learning_rate": 3.04747465996708e-05, + "loss": 1.0155, + "step": 225880 + }, + { + "epoch": 19.55, + "learning_rate": 3.0473880273758994e-05, + "loss": 0.9238, + "step": 225890 + }, + { + "epoch": 19.55, + "learning_rate": 3.0473013947847185e-05, + "loss": 1.0324, + "step": 225900 + }, + { + "epoch": 19.55, + "learning_rate": 3.0472147621935376e-05, + "loss": 0.911, + "step": 225910 + }, + { + "epoch": 19.55, + "learning_rate": 3.0471281296023567e-05, + "loss": 0.959, + "step": 225920 + }, + { + "epoch": 19.56, + "learning_rate": 3.047041497011176e-05, + "loss": 0.9619, + "step": 225930 + }, + { + "epoch": 19.56, + "learning_rate": 3.0469548644199952e-05, + "loss": 0.9702, + "step": 225940 + }, + { + "epoch": 19.56, + "learning_rate": 3.046868231828814e-05, + "loss": 1.0319, + "step": 225950 + }, + { + "epoch": 19.56, + "learning_rate": 3.046781599237633e-05, + "loss": 1.0025, + "step": 225960 + }, + { + "epoch": 19.56, + "learning_rate": 3.0466949666464522e-05, + "loss": 0.9388, + "step": 225970 + }, + { + "epoch": 19.56, + "learning_rate": 3.0466083340552713e-05, + "loss": 0.9584, + "step": 225980 + }, + { + "epoch": 19.56, + "learning_rate": 3.0465217014640908e-05, + "loss": 0.945, + "step": 225990 + }, + { + "epoch": 19.56, + "learning_rate": 3.04643506887291e-05, + "loss": 1.0276, + "step": 226000 + }, + { + "epoch": 19.56, + "learning_rate": 3.046348436281729e-05, + "loss": 0.9813, + "step": 226010 + }, + { + "epoch": 19.56, + "learning_rate": 3.046261803690548e-05, + "loss": 1.0686, + "step": 226020 + }, + { + "epoch": 19.56, + "learning_rate": 3.0461751710993675e-05, + "loss": 1.0052, + "step": 226030 + }, + { + "epoch": 19.56, + "learning_rate": 3.0460885385081866e-05, + "loss": 0.9935, + "step": 226040 + }, + { + "epoch": 19.57, + "learning_rate": 3.0460019059170057e-05, + "loss": 0.9145, + "step": 226050 + }, + { + "epoch": 19.57, + "learning_rate": 3.0459152733258252e-05, + "loss": 0.9563, + "step": 226060 + }, + { + "epoch": 19.57, + "learning_rate": 3.0458286407346443e-05, + "loss": 0.9322, + "step": 226070 + }, + { + "epoch": 19.57, + "learning_rate": 3.0457420081434634e-05, + "loss": 0.9875, + "step": 226080 + }, + { + "epoch": 19.57, + "learning_rate": 3.0456553755522825e-05, + "loss": 0.9898, + "step": 226090 + }, + { + "epoch": 19.57, + "learning_rate": 3.045568742961102e-05, + "loss": 0.9702, + "step": 226100 + }, + { + "epoch": 19.57, + "learning_rate": 3.045482110369921e-05, + "loss": 0.984, + "step": 226110 + }, + { + "epoch": 19.57, + "learning_rate": 3.0453954777787402e-05, + "loss": 0.9992, + "step": 226120 + }, + { + "epoch": 19.57, + "learning_rate": 3.0453088451875593e-05, + "loss": 0.9577, + "step": 226130 + }, + { + "epoch": 19.57, + "learning_rate": 3.0452222125963787e-05, + "loss": 0.9711, + "step": 226140 + }, + { + "epoch": 19.57, + "learning_rate": 3.0451355800051978e-05, + "loss": 0.9545, + "step": 226150 + }, + { + "epoch": 19.58, + "learning_rate": 3.045048947414017e-05, + "loss": 0.9813, + "step": 226160 + }, + { + "epoch": 19.58, + "learning_rate": 3.0449623148228364e-05, + "loss": 1.0071, + "step": 226170 + }, + { + "epoch": 19.58, + "learning_rate": 3.0448756822316555e-05, + "loss": 0.9632, + "step": 226180 + }, + { + "epoch": 19.58, + "learning_rate": 3.0447890496404746e-05, + "loss": 0.9874, + "step": 226190 + }, + { + "epoch": 19.58, + "learning_rate": 3.0447024170492937e-05, + "loss": 0.9776, + "step": 226200 + }, + { + "epoch": 19.58, + "learning_rate": 3.044615784458113e-05, + "loss": 0.9722, + "step": 226210 + }, + { + "epoch": 19.58, + "learning_rate": 3.0445291518669323e-05, + "loss": 0.967, + "step": 226220 + }, + { + "epoch": 19.58, + "learning_rate": 3.0444425192757514e-05, + "loss": 0.9939, + "step": 226230 + }, + { + "epoch": 19.58, + "learning_rate": 3.0443558866845705e-05, + "loss": 0.9628, + "step": 226240 + }, + { + "epoch": 19.58, + "learning_rate": 3.04426925409339e-05, + "loss": 0.9691, + "step": 226250 + }, + { + "epoch": 19.58, + "learning_rate": 3.044182621502209e-05, + "loss": 0.9681, + "step": 226260 + }, + { + "epoch": 19.58, + "learning_rate": 3.044095988911028e-05, + "loss": 1.0104, + "step": 226270 + }, + { + "epoch": 19.59, + "learning_rate": 3.0440093563198476e-05, + "loss": 0.9074, + "step": 226280 + }, + { + "epoch": 19.59, + "learning_rate": 3.0439227237286667e-05, + "loss": 0.9324, + "step": 226290 + }, + { + "epoch": 19.59, + "learning_rate": 3.0438360911374858e-05, + "loss": 0.9523, + "step": 226300 + }, + { + "epoch": 19.59, + "learning_rate": 3.043749458546305e-05, + "loss": 0.8892, + "step": 226310 + }, + { + "epoch": 19.59, + "learning_rate": 3.0436628259551243e-05, + "loss": 0.9573, + "step": 226320 + }, + { + "epoch": 19.59, + "learning_rate": 3.0435761933639434e-05, + "loss": 1.025, + "step": 226330 + }, + { + "epoch": 19.59, + "learning_rate": 3.0434895607727625e-05, + "loss": 0.9743, + "step": 226340 + }, + { + "epoch": 19.59, + "learning_rate": 3.0434029281815817e-05, + "loss": 0.9614, + "step": 226350 + }, + { + "epoch": 19.59, + "learning_rate": 3.043316295590401e-05, + "loss": 0.8942, + "step": 226360 + }, + { + "epoch": 19.59, + "learning_rate": 3.0432296629992202e-05, + "loss": 0.9352, + "step": 226370 + }, + { + "epoch": 19.59, + "learning_rate": 3.0431430304080393e-05, + "loss": 0.9915, + "step": 226380 + }, + { + "epoch": 19.6, + "learning_rate": 3.0430563978168584e-05, + "loss": 0.9795, + "step": 226390 + }, + { + "epoch": 19.6, + "learning_rate": 3.042969765225678e-05, + "loss": 0.941, + "step": 226400 + }, + { + "epoch": 19.6, + "learning_rate": 3.042883132634497e-05, + "loss": 0.9705, + "step": 226410 + }, + { + "epoch": 19.6, + "learning_rate": 3.042796500043316e-05, + "loss": 1.0002, + "step": 226420 + }, + { + "epoch": 19.6, + "learning_rate": 3.0427098674521355e-05, + "loss": 0.9891, + "step": 226430 + }, + { + "epoch": 19.6, + "learning_rate": 3.0426232348609546e-05, + "loss": 0.9424, + "step": 226440 + }, + { + "epoch": 19.6, + "learning_rate": 3.0425366022697737e-05, + "loss": 1.0315, + "step": 226450 + }, + { + "epoch": 19.6, + "learning_rate": 3.042449969678593e-05, + "loss": 0.9558, + "step": 226460 + }, + { + "epoch": 19.6, + "learning_rate": 3.0423633370874123e-05, + "loss": 0.9435, + "step": 226470 + }, + { + "epoch": 19.6, + "learning_rate": 3.0422767044962314e-05, + "loss": 0.935, + "step": 226480 + }, + { + "epoch": 19.6, + "learning_rate": 3.0421900719050505e-05, + "loss": 1.0216, + "step": 226490 + }, + { + "epoch": 19.6, + "learning_rate": 3.0421034393138696e-05, + "loss": 0.9588, + "step": 226500 + }, + { + "epoch": 19.61, + "learning_rate": 3.042016806722689e-05, + "loss": 0.9718, + "step": 226510 + }, + { + "epoch": 19.61, + "learning_rate": 3.041930174131508e-05, + "loss": 0.9557, + "step": 226520 + }, + { + "epoch": 19.61, + "learning_rate": 3.0418435415403273e-05, + "loss": 0.9768, + "step": 226530 + }, + { + "epoch": 19.61, + "learning_rate": 3.0417569089491467e-05, + "loss": 0.9507, + "step": 226540 + }, + { + "epoch": 19.61, + "learning_rate": 3.0416702763579658e-05, + "loss": 0.9193, + "step": 226550 + }, + { + "epoch": 19.61, + "learning_rate": 3.041583643766785e-05, + "loss": 0.9338, + "step": 226560 + }, + { + "epoch": 19.61, + "learning_rate": 3.041497011175604e-05, + "loss": 0.9553, + "step": 226570 + }, + { + "epoch": 19.61, + "learning_rate": 3.0414103785844235e-05, + "loss": 0.9555, + "step": 226580 + }, + { + "epoch": 19.61, + "learning_rate": 3.0413237459932426e-05, + "loss": 1.0061, + "step": 226590 + }, + { + "epoch": 19.61, + "learning_rate": 3.0412371134020617e-05, + "loss": 0.9866, + "step": 226600 + }, + { + "epoch": 19.61, + "learning_rate": 3.0411504808108808e-05, + "loss": 0.9922, + "step": 226610 + }, + { + "epoch": 19.62, + "learning_rate": 3.0410638482197002e-05, + "loss": 0.9658, + "step": 226620 + }, + { + "epoch": 19.62, + "learning_rate": 3.0409772156285193e-05, + "loss": 0.9384, + "step": 226630 + }, + { + "epoch": 19.62, + "learning_rate": 3.0408905830373385e-05, + "loss": 0.9267, + "step": 226640 + }, + { + "epoch": 19.62, + "learning_rate": 3.040803950446158e-05, + "loss": 0.965, + "step": 226650 + }, + { + "epoch": 19.62, + "learning_rate": 3.040717317854977e-05, + "loss": 0.9248, + "step": 226660 + }, + { + "epoch": 19.62, + "learning_rate": 3.040630685263796e-05, + "loss": 1.0028, + "step": 226670 + }, + { + "epoch": 19.62, + "learning_rate": 3.0405440526726152e-05, + "loss": 0.9959, + "step": 226680 + }, + { + "epoch": 19.62, + "learning_rate": 3.0404574200814347e-05, + "loss": 0.987, + "step": 226690 + }, + { + "epoch": 19.62, + "learning_rate": 3.0403707874902538e-05, + "loss": 0.9734, + "step": 226700 + }, + { + "epoch": 19.62, + "learning_rate": 3.040284154899073e-05, + "loss": 0.9514, + "step": 226710 + }, + { + "epoch": 19.62, + "learning_rate": 3.040197522307892e-05, + "loss": 0.9046, + "step": 226720 + }, + { + "epoch": 19.62, + "learning_rate": 3.0401108897167114e-05, + "loss": 0.9865, + "step": 226730 + }, + { + "epoch": 19.63, + "learning_rate": 3.0400242571255305e-05, + "loss": 0.9664, + "step": 226740 + }, + { + "epoch": 19.63, + "learning_rate": 3.0399376245343496e-05, + "loss": 0.9873, + "step": 226750 + }, + { + "epoch": 19.63, + "learning_rate": 3.0398509919431687e-05, + "loss": 0.9381, + "step": 226760 + }, + { + "epoch": 19.63, + "learning_rate": 3.0397643593519882e-05, + "loss": 0.9511, + "step": 226770 + }, + { + "epoch": 19.63, + "learning_rate": 3.0396777267608073e-05, + "loss": 0.9527, + "step": 226780 + }, + { + "epoch": 19.63, + "learning_rate": 3.0395910941696264e-05, + "loss": 0.9682, + "step": 226790 + }, + { + "epoch": 19.63, + "learning_rate": 3.039504461578446e-05, + "loss": 0.9989, + "step": 226800 + }, + { + "epoch": 19.63, + "learning_rate": 3.039417828987265e-05, + "loss": 0.9923, + "step": 226810 + }, + { + "epoch": 19.63, + "learning_rate": 3.039331196396084e-05, + "loss": 0.9115, + "step": 226820 + }, + { + "epoch": 19.63, + "learning_rate": 3.039244563804903e-05, + "loss": 0.9303, + "step": 226830 + }, + { + "epoch": 19.63, + "learning_rate": 3.0391579312137226e-05, + "loss": 0.9849, + "step": 226840 + }, + { + "epoch": 19.64, + "learning_rate": 3.0390712986225417e-05, + "loss": 0.9705, + "step": 226850 + }, + { + "epoch": 19.64, + "learning_rate": 3.0389846660313608e-05, + "loss": 0.9757, + "step": 226860 + }, + { + "epoch": 19.64, + "learning_rate": 3.03889803344018e-05, + "loss": 0.9773, + "step": 226870 + }, + { + "epoch": 19.64, + "learning_rate": 3.0388114008489994e-05, + "loss": 0.9336, + "step": 226880 + }, + { + "epoch": 19.64, + "learning_rate": 3.0387247682578185e-05, + "loss": 0.9379, + "step": 226890 + }, + { + "epoch": 19.64, + "learning_rate": 3.0386381356666376e-05, + "loss": 0.9824, + "step": 226900 + }, + { + "epoch": 19.64, + "learning_rate": 3.038551503075457e-05, + "loss": 0.9675, + "step": 226910 + }, + { + "epoch": 19.64, + "learning_rate": 3.038464870484276e-05, + "loss": 0.9282, + "step": 226920 + }, + { + "epoch": 19.64, + "learning_rate": 3.0383782378930953e-05, + "loss": 0.9354, + "step": 226930 + }, + { + "epoch": 19.64, + "learning_rate": 3.0382916053019144e-05, + "loss": 0.9823, + "step": 226940 + }, + { + "epoch": 19.64, + "learning_rate": 3.0382049727107338e-05, + "loss": 0.9514, + "step": 226950 + }, + { + "epoch": 19.64, + "learning_rate": 3.038118340119553e-05, + "loss": 0.9947, + "step": 226960 + }, + { + "epoch": 19.65, + "learning_rate": 3.038031707528372e-05, + "loss": 0.9815, + "step": 226970 + }, + { + "epoch": 19.65, + "learning_rate": 3.037945074937191e-05, + "loss": 0.966, + "step": 226980 + }, + { + "epoch": 19.65, + "learning_rate": 3.0378584423460106e-05, + "loss": 0.929, + "step": 226990 + }, + { + "epoch": 19.65, + "learning_rate": 3.0377718097548297e-05, + "loss": 0.9993, + "step": 227000 + }, + { + "epoch": 19.65, + "learning_rate": 3.0376851771636488e-05, + "loss": 0.9828, + "step": 227010 + }, + { + "epoch": 19.65, + "learning_rate": 3.0375985445724682e-05, + "loss": 0.925, + "step": 227020 + }, + { + "epoch": 19.65, + "learning_rate": 3.0375119119812873e-05, + "loss": 0.9239, + "step": 227030 + }, + { + "epoch": 19.65, + "learning_rate": 3.0374252793901064e-05, + "loss": 0.9157, + "step": 227040 + }, + { + "epoch": 19.65, + "learning_rate": 3.0373386467989255e-05, + "loss": 1.02, + "step": 227050 + }, + { + "epoch": 19.65, + "learning_rate": 3.037252014207745e-05, + "loss": 0.9557, + "step": 227060 + }, + { + "epoch": 19.65, + "learning_rate": 3.037165381616564e-05, + "loss": 0.9514, + "step": 227070 + }, + { + "epoch": 19.65, + "learning_rate": 3.0370787490253832e-05, + "loss": 0.9235, + "step": 227080 + }, + { + "epoch": 19.66, + "learning_rate": 3.0369921164342023e-05, + "loss": 0.9864, + "step": 227090 + }, + { + "epoch": 19.66, + "learning_rate": 3.0369054838430218e-05, + "loss": 0.9029, + "step": 227100 + }, + { + "epoch": 19.66, + "learning_rate": 3.036818851251841e-05, + "loss": 0.9725, + "step": 227110 + }, + { + "epoch": 19.66, + "learning_rate": 3.03673221866066e-05, + "loss": 0.9883, + "step": 227120 + }, + { + "epoch": 19.66, + "learning_rate": 3.036645586069479e-05, + "loss": 0.9896, + "step": 227130 + }, + { + "epoch": 19.66, + "learning_rate": 3.0365589534782985e-05, + "loss": 0.9863, + "step": 227140 + }, + { + "epoch": 19.66, + "learning_rate": 3.0364723208871176e-05, + "loss": 1.0091, + "step": 227150 + }, + { + "epoch": 19.66, + "learning_rate": 3.0363856882959367e-05, + "loss": 0.97, + "step": 227160 + }, + { + "epoch": 19.66, + "learning_rate": 3.0362990557047562e-05, + "loss": 0.9776, + "step": 227170 + }, + { + "epoch": 19.66, + "learning_rate": 3.0362124231135753e-05, + "loss": 0.955, + "step": 227180 + }, + { + "epoch": 19.66, + "learning_rate": 3.0361257905223944e-05, + "loss": 0.9724, + "step": 227190 + }, + { + "epoch": 19.67, + "learning_rate": 3.0360391579312135e-05, + "loss": 0.9993, + "step": 227200 + }, + { + "epoch": 19.67, + "learning_rate": 3.035952525340033e-05, + "loss": 0.9641, + "step": 227210 + }, + { + "epoch": 19.67, + "learning_rate": 3.035865892748852e-05, + "loss": 0.9534, + "step": 227220 + }, + { + "epoch": 19.67, + "learning_rate": 3.035779260157671e-05, + "loss": 0.9565, + "step": 227230 + }, + { + "epoch": 19.67, + "learning_rate": 3.0356926275664903e-05, + "loss": 0.9029, + "step": 227240 + }, + { + "epoch": 19.67, + "learning_rate": 3.0356059949753097e-05, + "loss": 0.98, + "step": 227250 + }, + { + "epoch": 19.67, + "learning_rate": 3.0355193623841288e-05, + "loss": 0.9342, + "step": 227260 + }, + { + "epoch": 19.67, + "learning_rate": 3.035432729792948e-05, + "loss": 0.9974, + "step": 227270 + }, + { + "epoch": 19.67, + "learning_rate": 3.0353460972017674e-05, + "loss": 0.9574, + "step": 227280 + }, + { + "epoch": 19.67, + "learning_rate": 3.0352594646105865e-05, + "loss": 0.9794, + "step": 227290 + }, + { + "epoch": 19.67, + "learning_rate": 3.0351728320194056e-05, + "loss": 0.9553, + "step": 227300 + }, + { + "epoch": 19.67, + "learning_rate": 3.0350861994282247e-05, + "loss": 0.9522, + "step": 227310 + }, + { + "epoch": 19.68, + "learning_rate": 3.034999566837044e-05, + "loss": 1.0155, + "step": 227320 + }, + { + "epoch": 19.68, + "learning_rate": 3.0349129342458632e-05, + "loss": 0.9903, + "step": 227330 + }, + { + "epoch": 19.68, + "learning_rate": 3.0348263016546823e-05, + "loss": 0.9376, + "step": 227340 + }, + { + "epoch": 19.68, + "learning_rate": 3.0347396690635014e-05, + "loss": 0.9866, + "step": 227350 + }, + { + "epoch": 19.68, + "learning_rate": 3.034653036472321e-05, + "loss": 0.901, + "step": 227360 + }, + { + "epoch": 19.68, + "learning_rate": 3.03456640388114e-05, + "loss": 0.9886, + "step": 227370 + }, + { + "epoch": 19.68, + "learning_rate": 3.034479771289959e-05, + "loss": 0.9465, + "step": 227380 + }, + { + "epoch": 19.68, + "learning_rate": 3.0343931386987786e-05, + "loss": 0.9452, + "step": 227390 + }, + { + "epoch": 19.68, + "learning_rate": 3.0343065061075977e-05, + "loss": 1.0019, + "step": 227400 + }, + { + "epoch": 19.68, + "learning_rate": 3.0342198735164168e-05, + "loss": 1.0136, + "step": 227410 + }, + { + "epoch": 19.68, + "learning_rate": 3.034133240925236e-05, + "loss": 0.9133, + "step": 227420 + }, + { + "epoch": 19.69, + "learning_rate": 3.0340466083340553e-05, + "loss": 0.9538, + "step": 227430 + }, + { + "epoch": 19.69, + "learning_rate": 3.0339599757428744e-05, + "loss": 1.0017, + "step": 227440 + }, + { + "epoch": 19.69, + "learning_rate": 3.0338733431516935e-05, + "loss": 0.9618, + "step": 227450 + }, + { + "epoch": 19.69, + "learning_rate": 3.0337867105605126e-05, + "loss": 0.985, + "step": 227460 + }, + { + "epoch": 19.69, + "learning_rate": 3.033700077969332e-05, + "loss": 0.982, + "step": 227470 + }, + { + "epoch": 19.69, + "learning_rate": 3.0336134453781512e-05, + "loss": 0.9791, + "step": 227480 + }, + { + "epoch": 19.69, + "learning_rate": 3.0335268127869703e-05, + "loss": 0.9517, + "step": 227490 + }, + { + "epoch": 19.69, + "learning_rate": 3.0334401801957897e-05, + "loss": 0.9436, + "step": 227500 + }, + { + "epoch": 19.69, + "learning_rate": 3.033353547604609e-05, + "loss": 0.9891, + "step": 227510 + }, + { + "epoch": 19.69, + "learning_rate": 3.033266915013428e-05, + "loss": 1.0102, + "step": 227520 + }, + { + "epoch": 19.69, + "learning_rate": 3.033180282422247e-05, + "loss": 0.9288, + "step": 227530 + }, + { + "epoch": 19.69, + "learning_rate": 3.0330936498310665e-05, + "loss": 0.9695, + "step": 227540 + }, + { + "epoch": 19.7, + "learning_rate": 3.0330070172398856e-05, + "loss": 0.945, + "step": 227550 + }, + { + "epoch": 19.7, + "learning_rate": 3.0329203846487047e-05, + "loss": 0.9293, + "step": 227560 + }, + { + "epoch": 19.7, + "learning_rate": 3.0328337520575238e-05, + "loss": 0.9392, + "step": 227570 + }, + { + "epoch": 19.7, + "learning_rate": 3.0327471194663433e-05, + "loss": 1.0205, + "step": 227580 + }, + { + "epoch": 19.7, + "learning_rate": 3.0326604868751624e-05, + "loss": 0.9743, + "step": 227590 + }, + { + "epoch": 19.7, + "learning_rate": 3.0325738542839815e-05, + "loss": 0.961, + "step": 227600 + }, + { + "epoch": 19.7, + "learning_rate": 3.0324872216928006e-05, + "loss": 0.9686, + "step": 227610 + }, + { + "epoch": 19.7, + "learning_rate": 3.03240058910162e-05, + "loss": 0.9507, + "step": 227620 + }, + { + "epoch": 19.7, + "learning_rate": 3.032313956510439e-05, + "loss": 0.9772, + "step": 227630 + }, + { + "epoch": 19.7, + "learning_rate": 3.0322273239192582e-05, + "loss": 0.92, + "step": 227640 + }, + { + "epoch": 19.7, + "learning_rate": 3.0321406913280777e-05, + "loss": 0.9775, + "step": 227650 + }, + { + "epoch": 19.71, + "learning_rate": 3.0320540587368968e-05, + "loss": 0.9144, + "step": 227660 + }, + { + "epoch": 19.71, + "learning_rate": 3.031967426145716e-05, + "loss": 1.0039, + "step": 227670 + }, + { + "epoch": 19.71, + "learning_rate": 3.031880793554535e-05, + "loss": 0.9736, + "step": 227680 + }, + { + "epoch": 19.71, + "learning_rate": 3.0317941609633545e-05, + "loss": 0.97, + "step": 227690 + }, + { + "epoch": 19.71, + "learning_rate": 3.0317075283721736e-05, + "loss": 0.9757, + "step": 227700 + }, + { + "epoch": 19.71, + "learning_rate": 3.0316208957809927e-05, + "loss": 0.9861, + "step": 227710 + }, + { + "epoch": 19.71, + "learning_rate": 3.0315342631898118e-05, + "loss": 0.9106, + "step": 227720 + }, + { + "epoch": 19.71, + "learning_rate": 3.0314476305986312e-05, + "loss": 1.0089, + "step": 227730 + }, + { + "epoch": 19.71, + "learning_rate": 3.0313609980074503e-05, + "loss": 0.9567, + "step": 227740 + }, + { + "epoch": 19.71, + "learning_rate": 3.0312743654162694e-05, + "loss": 0.9439, + "step": 227750 + }, + { + "epoch": 19.71, + "learning_rate": 3.031187732825089e-05, + "loss": 0.9724, + "step": 227760 + }, + { + "epoch": 19.71, + "learning_rate": 3.031101100233908e-05, + "loss": 0.9337, + "step": 227770 + }, + { + "epoch": 19.72, + "learning_rate": 3.031014467642727e-05, + "loss": 0.9521, + "step": 227780 + }, + { + "epoch": 19.72, + "learning_rate": 3.0309278350515462e-05, + "loss": 0.9753, + "step": 227790 + }, + { + "epoch": 19.72, + "learning_rate": 3.0308412024603656e-05, + "loss": 0.9576, + "step": 227800 + }, + { + "epoch": 19.72, + "learning_rate": 3.0307545698691848e-05, + "loss": 0.9311, + "step": 227810 + }, + { + "epoch": 19.72, + "learning_rate": 3.030667937278004e-05, + "loss": 0.9949, + "step": 227820 + }, + { + "epoch": 19.72, + "learning_rate": 3.030581304686823e-05, + "loss": 1.0177, + "step": 227830 + }, + { + "epoch": 19.72, + "learning_rate": 3.0304946720956424e-05, + "loss": 1.0029, + "step": 227840 + }, + { + "epoch": 19.72, + "learning_rate": 3.0304080395044615e-05, + "loss": 0.9669, + "step": 227850 + }, + { + "epoch": 19.72, + "learning_rate": 3.0303214069132806e-05, + "loss": 1.0153, + "step": 227860 + }, + { + "epoch": 19.72, + "learning_rate": 3.0302347743221e-05, + "loss": 0.9577, + "step": 227870 + }, + { + "epoch": 19.72, + "learning_rate": 3.0301481417309192e-05, + "loss": 0.943, + "step": 227880 + }, + { + "epoch": 19.73, + "learning_rate": 3.0300615091397383e-05, + "loss": 0.9908, + "step": 227890 + }, + { + "epoch": 19.73, + "learning_rate": 3.0299748765485574e-05, + "loss": 0.9877, + "step": 227900 + }, + { + "epoch": 19.73, + "learning_rate": 3.029888243957377e-05, + "loss": 0.967, + "step": 227910 + }, + { + "epoch": 19.73, + "learning_rate": 3.029801611366196e-05, + "loss": 0.9765, + "step": 227920 + }, + { + "epoch": 19.73, + "learning_rate": 3.029714978775015e-05, + "loss": 1.0304, + "step": 227930 + }, + { + "epoch": 19.73, + "learning_rate": 3.029628346183834e-05, + "loss": 0.9884, + "step": 227940 + }, + { + "epoch": 19.73, + "learning_rate": 3.0295417135926536e-05, + "loss": 0.8977, + "step": 227950 + }, + { + "epoch": 19.73, + "learning_rate": 3.0294550810014727e-05, + "loss": 0.9373, + "step": 227960 + }, + { + "epoch": 19.73, + "learning_rate": 3.0293684484102918e-05, + "loss": 0.9609, + "step": 227970 + }, + { + "epoch": 19.73, + "learning_rate": 3.029281815819111e-05, + "loss": 0.9437, + "step": 227980 + }, + { + "epoch": 19.73, + "learning_rate": 3.0291951832279304e-05, + "loss": 0.9625, + "step": 227990 + }, + { + "epoch": 19.73, + "learning_rate": 3.0291085506367495e-05, + "loss": 0.9673, + "step": 228000 + }, + { + "epoch": 19.74, + "learning_rate": 3.0290219180455686e-05, + "loss": 0.991, + "step": 228010 + }, + { + "epoch": 19.74, + "learning_rate": 3.028935285454388e-05, + "loss": 0.9424, + "step": 228020 + }, + { + "epoch": 19.74, + "learning_rate": 3.028848652863207e-05, + "loss": 0.9866, + "step": 228030 + }, + { + "epoch": 19.74, + "learning_rate": 3.0287620202720262e-05, + "loss": 0.9631, + "step": 228040 + }, + { + "epoch": 19.74, + "learning_rate": 3.0286753876808453e-05, + "loss": 0.9944, + "step": 228050 + }, + { + "epoch": 19.74, + "learning_rate": 3.0285887550896648e-05, + "loss": 0.9654, + "step": 228060 + }, + { + "epoch": 19.74, + "learning_rate": 3.028502122498484e-05, + "loss": 0.9605, + "step": 228070 + }, + { + "epoch": 19.74, + "learning_rate": 3.028415489907303e-05, + "loss": 0.9514, + "step": 228080 + }, + { + "epoch": 19.74, + "learning_rate": 3.028328857316122e-05, + "loss": 0.9565, + "step": 228090 + }, + { + "epoch": 19.74, + "learning_rate": 3.0282422247249415e-05, + "loss": 0.9663, + "step": 228100 + }, + { + "epoch": 19.74, + "learning_rate": 3.0281555921337607e-05, + "loss": 0.9495, + "step": 228110 + }, + { + "epoch": 19.74, + "learning_rate": 3.0280689595425798e-05, + "loss": 0.9653, + "step": 228120 + }, + { + "epoch": 19.75, + "learning_rate": 3.0279823269513992e-05, + "loss": 0.9905, + "step": 228130 + }, + { + "epoch": 19.75, + "learning_rate": 3.0278956943602183e-05, + "loss": 0.9269, + "step": 228140 + }, + { + "epoch": 19.75, + "learning_rate": 3.0278090617690374e-05, + "loss": 1.0036, + "step": 228150 + }, + { + "epoch": 19.75, + "learning_rate": 3.0277224291778565e-05, + "loss": 1.0023, + "step": 228160 + }, + { + "epoch": 19.75, + "learning_rate": 3.027635796586676e-05, + "loss": 0.9316, + "step": 228170 + }, + { + "epoch": 19.75, + "learning_rate": 3.027549163995495e-05, + "loss": 0.9442, + "step": 228180 + }, + { + "epoch": 19.75, + "learning_rate": 3.0274625314043142e-05, + "loss": 0.9893, + "step": 228190 + }, + { + "epoch": 19.75, + "learning_rate": 3.0273758988131333e-05, + "loss": 0.9769, + "step": 228200 + }, + { + "epoch": 19.75, + "learning_rate": 3.0272892662219527e-05, + "loss": 0.9867, + "step": 228210 + }, + { + "epoch": 19.75, + "learning_rate": 3.027202633630772e-05, + "loss": 0.893, + "step": 228220 + }, + { + "epoch": 19.75, + "learning_rate": 3.027116001039591e-05, + "loss": 0.9608, + "step": 228230 + }, + { + "epoch": 19.76, + "learning_rate": 3.0270293684484104e-05, + "loss": 1.0149, + "step": 228240 + }, + { + "epoch": 19.76, + "learning_rate": 3.0269427358572295e-05, + "loss": 1.0054, + "step": 228250 + }, + { + "epoch": 19.76, + "learning_rate": 3.0268561032660486e-05, + "loss": 0.9742, + "step": 228260 + }, + { + "epoch": 19.76, + "learning_rate": 3.0267694706748677e-05, + "loss": 0.9855, + "step": 228270 + }, + { + "epoch": 19.76, + "learning_rate": 3.026682838083687e-05, + "loss": 0.9818, + "step": 228280 + }, + { + "epoch": 19.76, + "learning_rate": 3.0265962054925063e-05, + "loss": 0.9523, + "step": 228290 + }, + { + "epoch": 19.76, + "learning_rate": 3.0265095729013254e-05, + "loss": 1.0147, + "step": 228300 + }, + { + "epoch": 19.76, + "learning_rate": 3.0264229403101445e-05, + "loss": 0.9148, + "step": 228310 + }, + { + "epoch": 19.76, + "learning_rate": 3.026336307718964e-05, + "loss": 0.9637, + "step": 228320 + }, + { + "epoch": 19.76, + "learning_rate": 3.026249675127783e-05, + "loss": 1.0238, + "step": 228330 + }, + { + "epoch": 19.76, + "learning_rate": 3.026163042536602e-05, + "loss": 0.9905, + "step": 228340 + }, + { + "epoch": 19.76, + "learning_rate": 3.0260764099454212e-05, + "loss": 0.9541, + "step": 228350 + }, + { + "epoch": 19.77, + "learning_rate": 3.0259897773542407e-05, + "loss": 0.9905, + "step": 228360 + }, + { + "epoch": 19.77, + "learning_rate": 3.0259031447630598e-05, + "loss": 0.938, + "step": 228370 + }, + { + "epoch": 19.77, + "learning_rate": 3.025816512171879e-05, + "loss": 1.01, + "step": 228380 + }, + { + "epoch": 19.77, + "learning_rate": 3.0257298795806983e-05, + "loss": 0.9902, + "step": 228390 + }, + { + "epoch": 19.77, + "learning_rate": 3.0256432469895175e-05, + "loss": 0.9931, + "step": 228400 + }, + { + "epoch": 19.77, + "learning_rate": 3.0255566143983366e-05, + "loss": 0.9953, + "step": 228410 + }, + { + "epoch": 19.77, + "learning_rate": 3.0254699818071557e-05, + "loss": 0.9541, + "step": 228420 + }, + { + "epoch": 19.77, + "learning_rate": 3.025383349215975e-05, + "loss": 0.9958, + "step": 228430 + }, + { + "epoch": 19.77, + "learning_rate": 3.0252967166247942e-05, + "loss": 0.9414, + "step": 228440 + }, + { + "epoch": 19.77, + "learning_rate": 3.0252100840336133e-05, + "loss": 0.9732, + "step": 228450 + }, + { + "epoch": 19.77, + "learning_rate": 3.0251234514424324e-05, + "loss": 1.0038, + "step": 228460 + }, + { + "epoch": 19.78, + "learning_rate": 3.025036818851252e-05, + "loss": 0.959, + "step": 228470 + }, + { + "epoch": 19.78, + "learning_rate": 3.024950186260071e-05, + "loss": 1.0275, + "step": 228480 + }, + { + "epoch": 19.78, + "learning_rate": 3.02486355366889e-05, + "loss": 1.0161, + "step": 228490 + }, + { + "epoch": 19.78, + "learning_rate": 3.0247769210777095e-05, + "loss": 0.9195, + "step": 228500 + }, + { + "epoch": 19.78, + "learning_rate": 3.0246902884865286e-05, + "loss": 0.9084, + "step": 228510 + }, + { + "epoch": 19.78, + "learning_rate": 3.0246036558953477e-05, + "loss": 0.9887, + "step": 228520 + }, + { + "epoch": 19.78, + "learning_rate": 3.024517023304167e-05, + "loss": 0.9382, + "step": 228530 + }, + { + "epoch": 19.78, + "learning_rate": 3.0244303907129863e-05, + "loss": 0.9959, + "step": 228540 + }, + { + "epoch": 19.78, + "learning_rate": 3.0243437581218054e-05, + "loss": 0.9751, + "step": 228550 + }, + { + "epoch": 19.78, + "learning_rate": 3.0242571255306245e-05, + "loss": 0.988, + "step": 228560 + }, + { + "epoch": 19.78, + "learning_rate": 3.0241704929394436e-05, + "loss": 1.0417, + "step": 228570 + }, + { + "epoch": 19.78, + "learning_rate": 3.024083860348263e-05, + "loss": 0.9287, + "step": 228580 + }, + { + "epoch": 19.79, + "learning_rate": 3.0239972277570822e-05, + "loss": 1.0185, + "step": 228590 + }, + { + "epoch": 19.79, + "learning_rate": 3.0239105951659013e-05, + "loss": 0.9862, + "step": 228600 + }, + { + "epoch": 19.79, + "learning_rate": 3.0238239625747207e-05, + "loss": 1.0085, + "step": 228610 + }, + { + "epoch": 19.79, + "learning_rate": 3.0237373299835398e-05, + "loss": 0.9275, + "step": 228620 + }, + { + "epoch": 19.79, + "learning_rate": 3.023650697392359e-05, + "loss": 0.9656, + "step": 228630 + }, + { + "epoch": 19.79, + "learning_rate": 3.023564064801178e-05, + "loss": 1.0344, + "step": 228640 + }, + { + "epoch": 19.79, + "learning_rate": 3.0234774322099975e-05, + "loss": 0.9248, + "step": 228650 + }, + { + "epoch": 19.79, + "learning_rate": 3.0233907996188166e-05, + "loss": 0.9491, + "step": 228660 + }, + { + "epoch": 19.79, + "learning_rate": 3.0233041670276357e-05, + "loss": 0.9802, + "step": 228670 + }, + { + "epoch": 19.79, + "learning_rate": 3.0232175344364548e-05, + "loss": 0.9888, + "step": 228680 + }, + { + "epoch": 19.79, + "learning_rate": 3.0231309018452743e-05, + "loss": 1.059, + "step": 228690 + }, + { + "epoch": 19.8, + "learning_rate": 3.0230442692540934e-05, + "loss": 0.964, + "step": 228700 + }, + { + "epoch": 19.8, + "learning_rate": 3.0229576366629125e-05, + "loss": 0.9707, + "step": 228710 + }, + { + "epoch": 19.8, + "learning_rate": 3.0228710040717316e-05, + "loss": 1.0177, + "step": 228720 + }, + { + "epoch": 19.8, + "learning_rate": 3.022784371480551e-05, + "loss": 0.9088, + "step": 228730 + }, + { + "epoch": 19.8, + "learning_rate": 3.02269773888937e-05, + "loss": 0.9832, + "step": 228740 + }, + { + "epoch": 19.8, + "learning_rate": 3.0226111062981892e-05, + "loss": 0.9693, + "step": 228750 + }, + { + "epoch": 19.8, + "learning_rate": 3.0225244737070087e-05, + "loss": 0.9339, + "step": 228760 + }, + { + "epoch": 19.8, + "learning_rate": 3.0224378411158278e-05, + "loss": 0.9753, + "step": 228770 + }, + { + "epoch": 19.8, + "learning_rate": 3.022351208524647e-05, + "loss": 0.9706, + "step": 228780 + }, + { + "epoch": 19.8, + "learning_rate": 3.022264575933466e-05, + "loss": 0.9533, + "step": 228790 + }, + { + "epoch": 19.8, + "learning_rate": 3.0221779433422854e-05, + "loss": 0.9885, + "step": 228800 + }, + { + "epoch": 19.8, + "learning_rate": 3.0220913107511045e-05, + "loss": 0.9606, + "step": 228810 + }, + { + "epoch": 19.81, + "learning_rate": 3.0220046781599237e-05, + "loss": 0.9338, + "step": 228820 + }, + { + "epoch": 19.81, + "learning_rate": 3.0219180455687428e-05, + "loss": 0.9889, + "step": 228830 + }, + { + "epoch": 19.81, + "learning_rate": 3.0218314129775622e-05, + "loss": 0.9517, + "step": 228840 + }, + { + "epoch": 19.81, + "learning_rate": 3.0217447803863813e-05, + "loss": 0.9814, + "step": 228850 + }, + { + "epoch": 19.81, + "learning_rate": 3.0216581477952004e-05, + "loss": 0.9929, + "step": 228860 + }, + { + "epoch": 19.81, + "learning_rate": 3.02157151520402e-05, + "loss": 0.997, + "step": 228870 + }, + { + "epoch": 19.81, + "learning_rate": 3.021484882612839e-05, + "loss": 0.957, + "step": 228880 + }, + { + "epoch": 19.81, + "learning_rate": 3.021398250021658e-05, + "loss": 0.9799, + "step": 228890 + }, + { + "epoch": 19.81, + "learning_rate": 3.0213116174304772e-05, + "loss": 0.9221, + "step": 228900 + }, + { + "epoch": 19.81, + "learning_rate": 3.0212249848392966e-05, + "loss": 0.9773, + "step": 228910 + }, + { + "epoch": 19.81, + "learning_rate": 3.0211383522481157e-05, + "loss": 0.9118, + "step": 228920 + }, + { + "epoch": 19.82, + "learning_rate": 3.021051719656935e-05, + "loss": 0.9778, + "step": 228930 + }, + { + "epoch": 19.82, + "learning_rate": 3.020965087065754e-05, + "loss": 0.9924, + "step": 228940 + }, + { + "epoch": 19.82, + "learning_rate": 3.0208784544745734e-05, + "loss": 0.9864, + "step": 228950 + }, + { + "epoch": 19.82, + "learning_rate": 3.0207918218833925e-05, + "loss": 0.9589, + "step": 228960 + }, + { + "epoch": 19.82, + "learning_rate": 3.0207051892922116e-05, + "loss": 0.9501, + "step": 228970 + }, + { + "epoch": 19.82, + "learning_rate": 3.020618556701031e-05, + "loss": 0.9779, + "step": 228980 + }, + { + "epoch": 19.82, + "learning_rate": 3.02053192410985e-05, + "loss": 0.9442, + "step": 228990 + }, + { + "epoch": 19.82, + "learning_rate": 3.0204452915186693e-05, + "loss": 0.9427, + "step": 229000 + }, + { + "epoch": 19.82, + "learning_rate": 3.0203586589274884e-05, + "loss": 0.9689, + "step": 229010 + }, + { + "epoch": 19.82, + "learning_rate": 3.0202720263363078e-05, + "loss": 0.9508, + "step": 229020 + }, + { + "epoch": 19.82, + "learning_rate": 3.020185393745127e-05, + "loss": 0.9404, + "step": 229030 + }, + { + "epoch": 19.82, + "learning_rate": 3.020098761153946e-05, + "loss": 0.9128, + "step": 229040 + }, + { + "epoch": 19.83, + "learning_rate": 3.020012128562765e-05, + "loss": 1.0217, + "step": 229050 + }, + { + "epoch": 19.83, + "learning_rate": 3.0199254959715846e-05, + "loss": 0.9884, + "step": 229060 + }, + { + "epoch": 19.83, + "learning_rate": 3.0198388633804037e-05, + "loss": 0.9523, + "step": 229070 + }, + { + "epoch": 19.83, + "learning_rate": 3.0197522307892228e-05, + "loss": 1.0078, + "step": 229080 + }, + { + "epoch": 19.83, + "learning_rate": 3.019665598198042e-05, + "loss": 0.9447, + "step": 229090 + }, + { + "epoch": 19.83, + "learning_rate": 3.0195789656068613e-05, + "loss": 0.9274, + "step": 229100 + }, + { + "epoch": 19.83, + "learning_rate": 3.0194923330156805e-05, + "loss": 0.9588, + "step": 229110 + }, + { + "epoch": 19.83, + "learning_rate": 3.0194057004244996e-05, + "loss": 1.0029, + "step": 229120 + }, + { + "epoch": 19.83, + "learning_rate": 3.019319067833319e-05, + "loss": 0.9646, + "step": 229130 + }, + { + "epoch": 19.83, + "learning_rate": 3.019232435242138e-05, + "loss": 0.9893, + "step": 229140 + }, + { + "epoch": 19.83, + "learning_rate": 3.0191458026509572e-05, + "loss": 0.9548, + "step": 229150 + }, + { + "epoch": 19.83, + "learning_rate": 3.0190591700597763e-05, + "loss": 0.9647, + "step": 229160 + }, + { + "epoch": 19.84, + "learning_rate": 3.0189725374685958e-05, + "loss": 0.9439, + "step": 229170 + }, + { + "epoch": 19.84, + "learning_rate": 3.018885904877415e-05, + "loss": 0.9497, + "step": 229180 + }, + { + "epoch": 19.84, + "learning_rate": 3.018799272286234e-05, + "loss": 0.9914, + "step": 229190 + }, + { + "epoch": 19.84, + "learning_rate": 3.018712639695053e-05, + "loss": 0.9043, + "step": 229200 + }, + { + "epoch": 19.84, + "learning_rate": 3.0186260071038725e-05, + "loss": 1.0222, + "step": 229210 + }, + { + "epoch": 19.84, + "learning_rate": 3.0185393745126916e-05, + "loss": 0.9292, + "step": 229220 + }, + { + "epoch": 19.84, + "learning_rate": 3.0184527419215107e-05, + "loss": 0.9499, + "step": 229230 + }, + { + "epoch": 19.84, + "learning_rate": 3.0183661093303302e-05, + "loss": 0.9352, + "step": 229240 + }, + { + "epoch": 19.84, + "learning_rate": 3.0182794767391493e-05, + "loss": 0.9599, + "step": 229250 + }, + { + "epoch": 19.84, + "learning_rate": 3.0181928441479684e-05, + "loss": 0.9839, + "step": 229260 + }, + { + "epoch": 19.84, + "learning_rate": 3.0181062115567875e-05, + "loss": 0.9002, + "step": 229270 + }, + { + "epoch": 19.85, + "learning_rate": 3.018019578965607e-05, + "loss": 0.9982, + "step": 229280 + }, + { + "epoch": 19.85, + "learning_rate": 3.017932946374426e-05, + "loss": 0.9675, + "step": 229290 + }, + { + "epoch": 19.85, + "learning_rate": 3.017846313783245e-05, + "loss": 0.9248, + "step": 229300 + }, + { + "epoch": 19.85, + "learning_rate": 3.0177596811920643e-05, + "loss": 0.9221, + "step": 229310 + }, + { + "epoch": 19.85, + "learning_rate": 3.0176730486008837e-05, + "loss": 1.0309, + "step": 229320 + }, + { + "epoch": 19.85, + "learning_rate": 3.0175864160097028e-05, + "loss": 0.9838, + "step": 229330 + }, + { + "epoch": 19.85, + "learning_rate": 3.017499783418522e-05, + "loss": 0.9422, + "step": 229340 + }, + { + "epoch": 19.85, + "learning_rate": 3.0174131508273414e-05, + "loss": 0.9575, + "step": 229350 + }, + { + "epoch": 19.85, + "learning_rate": 3.0173265182361605e-05, + "loss": 0.9832, + "step": 229360 + }, + { + "epoch": 19.85, + "learning_rate": 3.0172398856449796e-05, + "loss": 0.9716, + "step": 229370 + }, + { + "epoch": 19.85, + "learning_rate": 3.0171532530537987e-05, + "loss": 0.951, + "step": 229380 + }, + { + "epoch": 19.85, + "learning_rate": 3.017066620462618e-05, + "loss": 0.9163, + "step": 229390 + }, + { + "epoch": 19.86, + "learning_rate": 3.0169799878714372e-05, + "loss": 0.942, + "step": 229400 + }, + { + "epoch": 19.86, + "learning_rate": 3.0168933552802564e-05, + "loss": 0.9764, + "step": 229410 + }, + { + "epoch": 19.86, + "learning_rate": 3.0168067226890755e-05, + "loss": 0.9443, + "step": 229420 + }, + { + "epoch": 19.86, + "learning_rate": 3.016720090097895e-05, + "loss": 0.9921, + "step": 229430 + }, + { + "epoch": 19.86, + "learning_rate": 3.016633457506714e-05, + "loss": 0.9601, + "step": 229440 + }, + { + "epoch": 19.86, + "learning_rate": 3.016546824915533e-05, + "loss": 0.9517, + "step": 229450 + }, + { + "epoch": 19.86, + "learning_rate": 3.0164601923243522e-05, + "loss": 0.929, + "step": 229460 + }, + { + "epoch": 19.86, + "learning_rate": 3.0163735597331717e-05, + "loss": 0.9975, + "step": 229470 + }, + { + "epoch": 19.86, + "learning_rate": 3.0162869271419908e-05, + "loss": 0.9684, + "step": 229480 + }, + { + "epoch": 19.86, + "learning_rate": 3.01620029455081e-05, + "loss": 1.0079, + "step": 229490 + }, + { + "epoch": 19.86, + "learning_rate": 3.0161136619596293e-05, + "loss": 0.9585, + "step": 229500 + }, + { + "epoch": 19.87, + "learning_rate": 3.0160270293684484e-05, + "loss": 0.9396, + "step": 229510 + }, + { + "epoch": 19.87, + "learning_rate": 3.0159403967772675e-05, + "loss": 0.9132, + "step": 229520 + }, + { + "epoch": 19.87, + "learning_rate": 3.0158537641860867e-05, + "loss": 0.9993, + "step": 229530 + }, + { + "epoch": 19.87, + "learning_rate": 3.015767131594906e-05, + "loss": 0.9851, + "step": 229540 + }, + { + "epoch": 19.87, + "learning_rate": 3.0156804990037252e-05, + "loss": 0.9807, + "step": 229550 + }, + { + "epoch": 19.87, + "learning_rate": 3.0155938664125443e-05, + "loss": 0.9894, + "step": 229560 + }, + { + "epoch": 19.87, + "learning_rate": 3.0155072338213634e-05, + "loss": 0.9468, + "step": 229570 + }, + { + "epoch": 19.87, + "learning_rate": 3.015420601230183e-05, + "loss": 0.9708, + "step": 229580 + }, + { + "epoch": 19.87, + "learning_rate": 3.015333968639002e-05, + "loss": 0.9943, + "step": 229590 + }, + { + "epoch": 19.87, + "learning_rate": 3.015247336047821e-05, + "loss": 1.0123, + "step": 229600 + }, + { + "epoch": 19.87, + "learning_rate": 3.0151607034566405e-05, + "loss": 0.9638, + "step": 229610 + }, + { + "epoch": 19.87, + "learning_rate": 3.0150740708654596e-05, + "loss": 0.9183, + "step": 229620 + }, + { + "epoch": 19.88, + "learning_rate": 3.0149874382742787e-05, + "loss": 0.957, + "step": 229630 + }, + { + "epoch": 19.88, + "learning_rate": 3.014900805683098e-05, + "loss": 0.9562, + "step": 229640 + }, + { + "epoch": 19.88, + "learning_rate": 3.0148141730919173e-05, + "loss": 0.9999, + "step": 229650 + }, + { + "epoch": 19.88, + "learning_rate": 3.0147275405007364e-05, + "loss": 0.9885, + "step": 229660 + }, + { + "epoch": 19.88, + "learning_rate": 3.0146409079095555e-05, + "loss": 1.0139, + "step": 229670 + }, + { + "epoch": 19.88, + "learning_rate": 3.0145542753183746e-05, + "loss": 0.9046, + "step": 229680 + }, + { + "epoch": 19.88, + "learning_rate": 3.014467642727194e-05, + "loss": 0.929, + "step": 229690 + }, + { + "epoch": 19.88, + "learning_rate": 3.014381010136013e-05, + "loss": 0.958, + "step": 229700 + }, + { + "epoch": 19.88, + "learning_rate": 3.0142943775448323e-05, + "loss": 1.023, + "step": 229710 + }, + { + "epoch": 19.88, + "learning_rate": 3.0142077449536517e-05, + "loss": 0.9723, + "step": 229720 + }, + { + "epoch": 19.88, + "learning_rate": 3.0141211123624708e-05, + "loss": 0.9851, + "step": 229730 + }, + { + "epoch": 19.89, + "learning_rate": 3.01403447977129e-05, + "loss": 0.9748, + "step": 229740 + }, + { + "epoch": 19.89, + "learning_rate": 3.013947847180109e-05, + "loss": 0.8836, + "step": 229750 + }, + { + "epoch": 19.89, + "learning_rate": 3.0138612145889285e-05, + "loss": 1.0044, + "step": 229760 + }, + { + "epoch": 19.89, + "learning_rate": 3.0137745819977476e-05, + "loss": 0.9431, + "step": 229770 + }, + { + "epoch": 19.89, + "learning_rate": 3.0136879494065667e-05, + "loss": 0.965, + "step": 229780 + }, + { + "epoch": 19.89, + "learning_rate": 3.0136013168153858e-05, + "loss": 0.9339, + "step": 229790 + }, + { + "epoch": 19.89, + "learning_rate": 3.0135146842242052e-05, + "loss": 0.9551, + "step": 229800 + }, + { + "epoch": 19.89, + "learning_rate": 3.0134280516330243e-05, + "loss": 0.9848, + "step": 229810 + }, + { + "epoch": 19.89, + "learning_rate": 3.0133414190418434e-05, + "loss": 0.9585, + "step": 229820 + }, + { + "epoch": 19.89, + "learning_rate": 3.013254786450663e-05, + "loss": 0.9808, + "step": 229830 + }, + { + "epoch": 19.89, + "learning_rate": 3.013168153859482e-05, + "loss": 1.0054, + "step": 229840 + }, + { + "epoch": 19.89, + "learning_rate": 3.013081521268301e-05, + "loss": 0.9634, + "step": 229850 + }, + { + "epoch": 19.9, + "learning_rate": 3.0129948886771202e-05, + "loss": 0.9523, + "step": 229860 + }, + { + "epoch": 19.9, + "learning_rate": 3.0129082560859397e-05, + "loss": 0.9898, + "step": 229870 + }, + { + "epoch": 19.9, + "learning_rate": 3.0128216234947588e-05, + "loss": 0.9461, + "step": 229880 + }, + { + "epoch": 19.9, + "learning_rate": 3.012734990903578e-05, + "loss": 0.9059, + "step": 229890 + }, + { + "epoch": 19.9, + "learning_rate": 3.012648358312397e-05, + "loss": 0.9804, + "step": 229900 + }, + { + "epoch": 19.9, + "learning_rate": 3.0125617257212164e-05, + "loss": 0.9751, + "step": 229910 + }, + { + "epoch": 19.9, + "learning_rate": 3.0124750931300355e-05, + "loss": 0.9886, + "step": 229920 + }, + { + "epoch": 19.9, + "learning_rate": 3.0123884605388546e-05, + "loss": 0.9154, + "step": 229930 + }, + { + "epoch": 19.9, + "learning_rate": 3.0123018279476737e-05, + "loss": 0.9899, + "step": 229940 + }, + { + "epoch": 19.9, + "learning_rate": 3.0122151953564932e-05, + "loss": 0.9386, + "step": 229950 + }, + { + "epoch": 19.9, + "learning_rate": 3.0121285627653123e-05, + "loss": 1.0122, + "step": 229960 + }, + { + "epoch": 19.91, + "learning_rate": 3.0120419301741314e-05, + "loss": 0.9745, + "step": 229970 + }, + { + "epoch": 19.91, + "learning_rate": 3.011955297582951e-05, + "loss": 0.9423, + "step": 229980 + }, + { + "epoch": 19.91, + "learning_rate": 3.01186866499177e-05, + "loss": 0.9689, + "step": 229990 + }, + { + "epoch": 19.91, + "learning_rate": 3.011782032400589e-05, + "loss": 0.9718, + "step": 230000 + }, + { + "epoch": 19.91, + "learning_rate": 3.011695399809408e-05, + "loss": 1.0153, + "step": 230010 + }, + { + "epoch": 19.91, + "learning_rate": 3.0116087672182276e-05, + "loss": 0.9646, + "step": 230020 + }, + { + "epoch": 19.91, + "learning_rate": 3.0115221346270467e-05, + "loss": 0.9989, + "step": 230030 + }, + { + "epoch": 19.91, + "learning_rate": 3.0114355020358658e-05, + "loss": 0.992, + "step": 230040 + }, + { + "epoch": 19.91, + "learning_rate": 3.011348869444685e-05, + "loss": 0.9925, + "step": 230050 + }, + { + "epoch": 19.91, + "learning_rate": 3.0112622368535044e-05, + "loss": 0.958, + "step": 230060 + }, + { + "epoch": 19.91, + "learning_rate": 3.0111756042623235e-05, + "loss": 0.9868, + "step": 230070 + }, + { + "epoch": 19.91, + "learning_rate": 3.0110889716711426e-05, + "loss": 0.957, + "step": 230080 + }, + { + "epoch": 19.92, + "learning_rate": 3.011002339079962e-05, + "loss": 0.9965, + "step": 230090 + }, + { + "epoch": 19.92, + "learning_rate": 3.010915706488781e-05, + "loss": 0.9857, + "step": 230100 + }, + { + "epoch": 19.92, + "learning_rate": 3.0108290738976002e-05, + "loss": 0.9509, + "step": 230110 + }, + { + "epoch": 19.92, + "learning_rate": 3.0107424413064194e-05, + "loss": 0.9948, + "step": 230120 + }, + { + "epoch": 19.92, + "learning_rate": 3.0106558087152388e-05, + "loss": 0.9944, + "step": 230130 + }, + { + "epoch": 19.92, + "learning_rate": 3.010569176124058e-05, + "loss": 0.9096, + "step": 230140 + }, + { + "epoch": 19.92, + "learning_rate": 3.010482543532877e-05, + "loss": 0.9541, + "step": 230150 + }, + { + "epoch": 19.92, + "learning_rate": 3.010395910941696e-05, + "loss": 0.9161, + "step": 230160 + }, + { + "epoch": 19.92, + "learning_rate": 3.0103092783505156e-05, + "loss": 0.9779, + "step": 230170 + }, + { + "epoch": 19.92, + "learning_rate": 3.0102226457593347e-05, + "loss": 0.9313, + "step": 230180 + }, + { + "epoch": 19.92, + "learning_rate": 3.0101360131681538e-05, + "loss": 0.9571, + "step": 230190 + }, + { + "epoch": 19.92, + "learning_rate": 3.0100493805769732e-05, + "loss": 0.9632, + "step": 230200 + }, + { + "epoch": 19.93, + "learning_rate": 3.0099627479857923e-05, + "loss": 0.9903, + "step": 230210 + }, + { + "epoch": 19.93, + "learning_rate": 3.0098761153946114e-05, + "loss": 0.9258, + "step": 230220 + }, + { + "epoch": 19.93, + "learning_rate": 3.0097894828034305e-05, + "loss": 0.9016, + "step": 230230 + }, + { + "epoch": 19.93, + "learning_rate": 3.00970285021225e-05, + "loss": 0.9875, + "step": 230240 + }, + { + "epoch": 19.93, + "learning_rate": 3.009616217621069e-05, + "loss": 1.0239, + "step": 230250 + }, + { + "epoch": 19.93, + "learning_rate": 3.0095295850298882e-05, + "loss": 0.9821, + "step": 230260 + }, + { + "epoch": 19.93, + "learning_rate": 3.0094429524387073e-05, + "loss": 0.9444, + "step": 230270 + }, + { + "epoch": 19.93, + "learning_rate": 3.0093563198475268e-05, + "loss": 0.9108, + "step": 230280 + }, + { + "epoch": 19.93, + "learning_rate": 3.009269687256346e-05, + "loss": 1.0147, + "step": 230290 + }, + { + "epoch": 19.93, + "learning_rate": 3.009183054665165e-05, + "loss": 0.9428, + "step": 230300 + }, + { + "epoch": 19.93, + "learning_rate": 3.009096422073984e-05, + "loss": 0.9546, + "step": 230310 + }, + { + "epoch": 19.94, + "learning_rate": 3.0090097894828035e-05, + "loss": 0.9257, + "step": 230320 + }, + { + "epoch": 19.94, + "learning_rate": 3.0089231568916226e-05, + "loss": 0.9768, + "step": 230330 + }, + { + "epoch": 19.94, + "learning_rate": 3.0088365243004417e-05, + "loss": 0.9541, + "step": 230340 + }, + { + "epoch": 19.94, + "learning_rate": 3.0087498917092612e-05, + "loss": 0.8773, + "step": 230350 + }, + { + "epoch": 19.94, + "learning_rate": 3.0086632591180803e-05, + "loss": 0.9853, + "step": 230360 + }, + { + "epoch": 19.94, + "learning_rate": 3.0085766265268994e-05, + "loss": 0.915, + "step": 230370 + }, + { + "epoch": 19.94, + "learning_rate": 3.0084899939357185e-05, + "loss": 0.8982, + "step": 230380 + }, + { + "epoch": 19.94, + "learning_rate": 3.008403361344538e-05, + "loss": 0.9698, + "step": 230390 + }, + { + "epoch": 19.94, + "learning_rate": 3.008316728753357e-05, + "loss": 0.9315, + "step": 230400 + }, + { + "epoch": 19.94, + "learning_rate": 3.008230096162176e-05, + "loss": 0.9631, + "step": 230410 + }, + { + "epoch": 19.94, + "learning_rate": 3.0081434635709953e-05, + "loss": 1.0232, + "step": 230420 + }, + { + "epoch": 19.94, + "learning_rate": 3.0080568309798147e-05, + "loss": 0.9314, + "step": 230430 + }, + { + "epoch": 19.95, + "learning_rate": 3.0079701983886338e-05, + "loss": 0.9878, + "step": 230440 + }, + { + "epoch": 19.95, + "learning_rate": 3.007883565797453e-05, + "loss": 0.9773, + "step": 230450 + }, + { + "epoch": 19.95, + "learning_rate": 3.0077969332062724e-05, + "loss": 0.9474, + "step": 230460 + }, + { + "epoch": 19.95, + "learning_rate": 3.0077103006150915e-05, + "loss": 0.9943, + "step": 230470 + }, + { + "epoch": 19.95, + "learning_rate": 3.0076236680239106e-05, + "loss": 0.9341, + "step": 230480 + }, + { + "epoch": 19.95, + "learning_rate": 3.0075370354327297e-05, + "loss": 0.9846, + "step": 230490 + }, + { + "epoch": 19.95, + "learning_rate": 3.007450402841549e-05, + "loss": 1.0417, + "step": 230500 + }, + { + "epoch": 19.95, + "learning_rate": 3.0073637702503682e-05, + "loss": 1.0226, + "step": 230510 + }, + { + "epoch": 19.95, + "learning_rate": 3.0072771376591873e-05, + "loss": 0.9054, + "step": 230520 + }, + { + "epoch": 19.95, + "learning_rate": 3.0071905050680064e-05, + "loss": 0.9367, + "step": 230530 + }, + { + "epoch": 19.95, + "learning_rate": 3.007103872476826e-05, + "loss": 1.0004, + "step": 230540 + }, + { + "epoch": 19.96, + "learning_rate": 3.007017239885645e-05, + "loss": 0.9482, + "step": 230550 + }, + { + "epoch": 19.96, + "learning_rate": 3.006930607294464e-05, + "loss": 0.9719, + "step": 230560 + }, + { + "epoch": 19.96, + "learning_rate": 3.0068439747032835e-05, + "loss": 0.9307, + "step": 230570 + }, + { + "epoch": 19.96, + "learning_rate": 3.0067573421121027e-05, + "loss": 0.9908, + "step": 230580 + }, + { + "epoch": 19.96, + "learning_rate": 3.0066707095209218e-05, + "loss": 0.9127, + "step": 230590 + }, + { + "epoch": 19.96, + "learning_rate": 3.006584076929741e-05, + "loss": 0.941, + "step": 230600 + }, + { + "epoch": 19.96, + "learning_rate": 3.0064974443385603e-05, + "loss": 0.9279, + "step": 230610 + }, + { + "epoch": 19.96, + "learning_rate": 3.0064108117473794e-05, + "loss": 0.914, + "step": 230620 + }, + { + "epoch": 19.96, + "learning_rate": 3.0063241791561985e-05, + "loss": 0.9336, + "step": 230630 + }, + { + "epoch": 19.96, + "learning_rate": 3.0062375465650176e-05, + "loss": 0.9653, + "step": 230640 + }, + { + "epoch": 19.96, + "learning_rate": 3.006150913973837e-05, + "loss": 0.9164, + "step": 230650 + }, + { + "epoch": 19.96, + "learning_rate": 3.0060642813826562e-05, + "loss": 0.9711, + "step": 230660 + }, + { + "epoch": 19.97, + "learning_rate": 3.0059776487914753e-05, + "loss": 1.0203, + "step": 230670 + }, + { + "epoch": 19.97, + "learning_rate": 3.0058910162002944e-05, + "loss": 0.9219, + "step": 230680 + }, + { + "epoch": 19.97, + "learning_rate": 3.005804383609114e-05, + "loss": 0.9158, + "step": 230690 + }, + { + "epoch": 19.97, + "learning_rate": 3.005717751017933e-05, + "loss": 0.9832, + "step": 230700 + }, + { + "epoch": 19.97, + "learning_rate": 3.005631118426752e-05, + "loss": 0.9349, + "step": 230710 + }, + { + "epoch": 19.97, + "learning_rate": 3.0055444858355715e-05, + "loss": 0.9705, + "step": 230720 + }, + { + "epoch": 19.97, + "learning_rate": 3.0054578532443906e-05, + "loss": 0.9371, + "step": 230730 + }, + { + "epoch": 19.97, + "learning_rate": 3.0053712206532097e-05, + "loss": 0.9158, + "step": 230740 + }, + { + "epoch": 19.97, + "learning_rate": 3.0052845880620288e-05, + "loss": 0.9585, + "step": 230750 + }, + { + "epoch": 19.97, + "learning_rate": 3.0051979554708483e-05, + "loss": 0.9594, + "step": 230760 + }, + { + "epoch": 19.97, + "learning_rate": 3.0051113228796674e-05, + "loss": 0.908, + "step": 230770 + }, + { + "epoch": 19.98, + "learning_rate": 3.0050246902884865e-05, + "loss": 1.0002, + "step": 230780 + }, + { + "epoch": 19.98, + "learning_rate": 3.0049380576973056e-05, + "loss": 0.9821, + "step": 230790 + }, + { + "epoch": 19.98, + "learning_rate": 3.004851425106125e-05, + "loss": 0.9836, + "step": 230800 + }, + { + "epoch": 19.98, + "learning_rate": 3.004764792514944e-05, + "loss": 0.9362, + "step": 230810 + }, + { + "epoch": 19.98, + "learning_rate": 3.0046781599237632e-05, + "loss": 1.0556, + "step": 230820 + }, + { + "epoch": 19.98, + "learning_rate": 3.0045915273325827e-05, + "loss": 0.9074, + "step": 230830 + }, + { + "epoch": 19.98, + "learning_rate": 3.0045048947414018e-05, + "loss": 0.937, + "step": 230840 + }, + { + "epoch": 19.98, + "learning_rate": 3.004418262150221e-05, + "loss": 0.9714, + "step": 230850 + }, + { + "epoch": 19.98, + "learning_rate": 3.00433162955904e-05, + "loss": 0.9594, + "step": 230860 + }, + { + "epoch": 19.98, + "learning_rate": 3.0042449969678595e-05, + "loss": 0.9447, + "step": 230870 + }, + { + "epoch": 19.98, + "learning_rate": 3.0041583643766786e-05, + "loss": 1.014, + "step": 230880 + }, + { + "epoch": 19.98, + "learning_rate": 3.0040717317854977e-05, + "loss": 0.9657, + "step": 230890 + }, + { + "epoch": 19.99, + "learning_rate": 3.0039850991943168e-05, + "loss": 0.9427, + "step": 230900 + }, + { + "epoch": 19.99, + "learning_rate": 3.0038984666031362e-05, + "loss": 0.9856, + "step": 230910 + }, + { + "epoch": 19.99, + "learning_rate": 3.0038118340119553e-05, + "loss": 0.9754, + "step": 230920 + }, + { + "epoch": 19.99, + "learning_rate": 3.0037252014207744e-05, + "loss": 0.9661, + "step": 230930 + }, + { + "epoch": 19.99, + "learning_rate": 3.003638568829594e-05, + "loss": 1.0202, + "step": 230940 + }, + { + "epoch": 19.99, + "learning_rate": 3.003551936238413e-05, + "loss": 0.9489, + "step": 230950 + }, + { + "epoch": 19.99, + "learning_rate": 3.003465303647232e-05, + "loss": 0.9848, + "step": 230960 + }, + { + "epoch": 19.99, + "learning_rate": 3.0033786710560512e-05, + "loss": 0.9017, + "step": 230970 + }, + { + "epoch": 19.99, + "learning_rate": 3.0032920384648706e-05, + "loss": 1.002, + "step": 230980 + }, + { + "epoch": 19.99, + "learning_rate": 3.0032054058736897e-05, + "loss": 0.9099, + "step": 230990 + }, + { + "epoch": 19.99, + "learning_rate": 3.003118773282509e-05, + "loss": 0.964, + "step": 231000 + }, + { + "epoch": 20.0, + "learning_rate": 3.003032140691328e-05, + "loss": 0.9307, + "step": 231010 + }, + { + "epoch": 20.0, + "learning_rate": 3.0029455081001474e-05, + "loss": 1.0065, + "step": 231020 + }, + { + "epoch": 20.0, + "learning_rate": 3.0028588755089665e-05, + "loss": 0.96, + "step": 231030 + }, + { + "epoch": 20.0, + "learning_rate": 3.0027722429177856e-05, + "loss": 0.9774, + "step": 231040 + }, + { + "epoch": 20.0, + "learning_rate": 3.0026856103266047e-05, + "loss": 0.952, + "step": 231050 + }, + { + "epoch": 20.0, + "learning_rate": 3.0025989777354242e-05, + "loss": 0.9675, + "step": 231060 + }, + { + "epoch": 20.0, + "eval_Bleu_1": 0.042785894125759054, + "eval_Bleu_2": 2.73439467776094e-11, + "eval_Bleu_3": 2.42928382867649e-14, + "eval_Bleu_4": 7.428000264458903e-16, + "eval_ROUGE_L": 0.08699374467430514, + "eval_cer": 0.9931186863222807, + "eval_em": 0, + "eval_f1": 0.11565704094856626, + "eval_loss": 0.9652414917945862, + "eval_runtime": 3459.5901, + "eval_samples_per_second": 1.484, + "eval_steps_per_second": 1.484, + "eval_wer": 0.9679246179261133, + "step": 231067 + }, + { + "epoch": 20.0, + "learning_rate": 1.6689300470975756e-05, + "loss": 0.9375, + "step": 231070 + }, + { + "epoch": 20.0, + "learning_rate": 1.6687855760062413e-05, + "loss": 0.9504, + "step": 231080 + }, + { + "epoch": 20.0, + "learning_rate": 1.6686411049149067e-05, + "loss": 0.9677, + "step": 231090 + }, + { + "epoch": 20.0, + "learning_rate": 1.668496633823572e-05, + "loss": 0.8723, + "step": 231100 + }, + { + "epoch": 20.0, + "learning_rate": 1.6683521627322373e-05, + "loss": 0.8952, + "step": 231110 + }, + { + "epoch": 20.01, + "learning_rate": 1.668207691640903e-05, + "loss": 0.8954, + "step": 231120 + }, + { + "epoch": 20.01, + "learning_rate": 1.668063220549568e-05, + "loss": 0.9386, + "step": 231130 + }, + { + "epoch": 20.01, + "learning_rate": 1.6679187494582334e-05, + "loss": 0.9299, + "step": 231140 + }, + { + "epoch": 20.01, + "learning_rate": 1.667774278366899e-05, + "loss": 0.9258, + "step": 231150 + }, + { + "epoch": 20.01, + "learning_rate": 1.667629807275564e-05, + "loss": 0.9844, + "step": 231160 + }, + { + "epoch": 20.01, + "learning_rate": 1.6674853361842297e-05, + "loss": 0.9618, + "step": 231170 + }, + { + "epoch": 20.01, + "learning_rate": 1.667340865092895e-05, + "loss": 0.9693, + "step": 231180 + }, + { + "epoch": 20.01, + "learning_rate": 1.6671963940015604e-05, + "loss": 0.9057, + "step": 231190 + }, + { + "epoch": 20.01, + "learning_rate": 1.6670519229102257e-05, + "loss": 0.9532, + "step": 231200 + }, + { + "epoch": 20.01, + "learning_rate": 1.6669074518188914e-05, + "loss": 0.9378, + "step": 231210 + }, + { + "epoch": 20.01, + "learning_rate": 1.6667629807275564e-05, + "loss": 0.9521, + "step": 231220 + }, + { + "epoch": 20.01, + "learning_rate": 1.6666185096362217e-05, + "loss": 0.9635, + "step": 231230 + }, + { + "epoch": 20.02, + "learning_rate": 1.6664740385448874e-05, + "loss": 0.8977, + "step": 231240 + }, + { + "epoch": 20.02, + "learning_rate": 1.6663295674535524e-05, + "loss": 0.9167, + "step": 231250 + }, + { + "epoch": 20.02, + "learning_rate": 1.666185096362218e-05, + "loss": 0.9591, + "step": 231260 + }, + { + "epoch": 20.02, + "learning_rate": 1.6660406252708834e-05, + "loss": 0.934, + "step": 231270 + }, + { + "epoch": 20.02, + "learning_rate": 1.6658961541795488e-05, + "loss": 0.9223, + "step": 231280 + }, + { + "epoch": 20.02, + "learning_rate": 1.665751683088214e-05, + "loss": 0.9081, + "step": 231290 + }, + { + "epoch": 20.02, + "learning_rate": 1.6656072119968798e-05, + "loss": 0.8832, + "step": 231300 + }, + { + "epoch": 20.02, + "learning_rate": 1.6654627409055448e-05, + "loss": 0.9132, + "step": 231310 + }, + { + "epoch": 20.02, + "learning_rate": 1.66531826981421e-05, + "loss": 0.9358, + "step": 231320 + }, + { + "epoch": 20.02, + "learning_rate": 1.6651737987228758e-05, + "loss": 0.9005, + "step": 231330 + }, + { + "epoch": 20.02, + "learning_rate": 1.6650293276315408e-05, + "loss": 0.9414, + "step": 231340 + }, + { + "epoch": 20.03, + "learning_rate": 1.6648848565402065e-05, + "loss": 0.9173, + "step": 231350 + }, + { + "epoch": 20.03, + "learning_rate": 1.664740385448872e-05, + "loss": 0.8917, + "step": 231360 + }, + { + "epoch": 20.03, + "learning_rate": 1.6645959143575372e-05, + "loss": 0.9679, + "step": 231370 + }, + { + "epoch": 20.03, + "learning_rate": 1.6644514432662025e-05, + "loss": 0.9529, + "step": 231380 + }, + { + "epoch": 20.03, + "learning_rate": 1.664306972174868e-05, + "loss": 0.931, + "step": 231390 + }, + { + "epoch": 20.03, + "learning_rate": 1.6641625010835332e-05, + "loss": 0.897, + "step": 231400 + }, + { + "epoch": 20.03, + "learning_rate": 1.6640180299921985e-05, + "loss": 0.9674, + "step": 231410 + }, + { + "epoch": 20.03, + "learning_rate": 1.663873558900864e-05, + "loss": 0.9959, + "step": 231420 + }, + { + "epoch": 20.03, + "learning_rate": 1.6637290878095292e-05, + "loss": 0.9688, + "step": 231430 + }, + { + "epoch": 20.03, + "learning_rate": 1.663584616718195e-05, + "loss": 0.9332, + "step": 231440 + }, + { + "epoch": 20.03, + "learning_rate": 1.66344014562686e-05, + "loss": 1.0018, + "step": 231450 + }, + { + "epoch": 20.03, + "learning_rate": 1.6632956745355256e-05, + "loss": 0.9561, + "step": 231460 + }, + { + "epoch": 20.04, + "learning_rate": 1.663151203444191e-05, + "loss": 0.9199, + "step": 231470 + }, + { + "epoch": 20.04, + "learning_rate": 1.6630067323528563e-05, + "loss": 0.981, + "step": 231480 + }, + { + "epoch": 20.04, + "learning_rate": 1.6628622612615216e-05, + "loss": 0.9379, + "step": 231490 + }, + { + "epoch": 20.04, + "learning_rate": 1.662717790170187e-05, + "loss": 0.9179, + "step": 231500 + }, + { + "epoch": 20.04, + "learning_rate": 1.6625733190788523e-05, + "loss": 0.9847, + "step": 231510 + }, + { + "epoch": 20.04, + "learning_rate": 1.6624288479875176e-05, + "loss": 0.8622, + "step": 231520 + }, + { + "epoch": 20.04, + "learning_rate": 1.6622843768961833e-05, + "loss": 0.9495, + "step": 231530 + }, + { + "epoch": 20.04, + "learning_rate": 1.6621399058048483e-05, + "loss": 0.884, + "step": 231540 + }, + { + "epoch": 20.04, + "learning_rate": 1.661995434713514e-05, + "loss": 0.9616, + "step": 231550 + }, + { + "epoch": 20.04, + "learning_rate": 1.6618509636221793e-05, + "loss": 0.9715, + "step": 231560 + }, + { + "epoch": 20.04, + "learning_rate": 1.6617064925308446e-05, + "loss": 0.9267, + "step": 231570 + }, + { + "epoch": 20.05, + "learning_rate": 1.66156202143951e-05, + "loss": 0.9482, + "step": 231580 + }, + { + "epoch": 20.05, + "learning_rate": 1.6614175503481753e-05, + "loss": 0.9864, + "step": 231590 + }, + { + "epoch": 20.05, + "learning_rate": 1.6612730792568407e-05, + "loss": 0.9349, + "step": 231600 + }, + { + "epoch": 20.05, + "learning_rate": 1.661128608165506e-05, + "loss": 0.9847, + "step": 231610 + }, + { + "epoch": 20.05, + "learning_rate": 1.6609841370741717e-05, + "loss": 0.9198, + "step": 231620 + }, + { + "epoch": 20.05, + "learning_rate": 1.6608396659828367e-05, + "loss": 0.9673, + "step": 231630 + }, + { + "epoch": 20.05, + "learning_rate": 1.6606951948915024e-05, + "loss": 0.9406, + "step": 231640 + }, + { + "epoch": 20.05, + "learning_rate": 1.6605507238001677e-05, + "loss": 0.9571, + "step": 231650 + }, + { + "epoch": 20.05, + "learning_rate": 1.660406252708833e-05, + "loss": 0.971, + "step": 231660 + }, + { + "epoch": 20.05, + "learning_rate": 1.6602617816174984e-05, + "loss": 0.9561, + "step": 231670 + }, + { + "epoch": 20.05, + "learning_rate": 1.6601173105261637e-05, + "loss": 0.9715, + "step": 231680 + }, + { + "epoch": 20.05, + "learning_rate": 1.659972839434829e-05, + "loss": 0.9272, + "step": 231690 + }, + { + "epoch": 20.06, + "learning_rate": 1.6598283683434944e-05, + "loss": 0.9311, + "step": 231700 + }, + { + "epoch": 20.06, + "learning_rate": 1.65968389725216e-05, + "loss": 0.9664, + "step": 231710 + }, + { + "epoch": 20.06, + "learning_rate": 1.659539426160825e-05, + "loss": 1.0201, + "step": 231720 + }, + { + "epoch": 20.06, + "learning_rate": 1.6593949550694908e-05, + "loss": 0.9809, + "step": 231730 + }, + { + "epoch": 20.06, + "learning_rate": 1.659250483978156e-05, + "loss": 0.9966, + "step": 231740 + }, + { + "epoch": 20.06, + "learning_rate": 1.6591060128868214e-05, + "loss": 0.9668, + "step": 231750 + }, + { + "epoch": 20.06, + "learning_rate": 1.6589615417954868e-05, + "loss": 0.9988, + "step": 231760 + }, + { + "epoch": 20.06, + "learning_rate": 1.658817070704152e-05, + "loss": 0.8984, + "step": 231770 + }, + { + "epoch": 20.06, + "learning_rate": 1.6586725996128175e-05, + "loss": 0.9656, + "step": 231780 + }, + { + "epoch": 20.06, + "learning_rate": 1.6585281285214828e-05, + "loss": 0.9615, + "step": 231790 + }, + { + "epoch": 20.06, + "learning_rate": 1.6583836574301485e-05, + "loss": 0.93, + "step": 231800 + }, + { + "epoch": 20.06, + "learning_rate": 1.6582391863388135e-05, + "loss": 0.9151, + "step": 231810 + }, + { + "epoch": 20.07, + "learning_rate": 1.658094715247479e-05, + "loss": 0.988, + "step": 231820 + }, + { + "epoch": 20.07, + "learning_rate": 1.6579502441561445e-05, + "loss": 0.9558, + "step": 231830 + }, + { + "epoch": 20.07, + "learning_rate": 1.6578057730648098e-05, + "loss": 0.9082, + "step": 231840 + }, + { + "epoch": 20.07, + "learning_rate": 1.657661301973475e-05, + "loss": 0.8847, + "step": 231850 + }, + { + "epoch": 20.07, + "learning_rate": 1.6575168308821405e-05, + "loss": 0.9104, + "step": 231860 + }, + { + "epoch": 20.07, + "learning_rate": 1.657372359790806e-05, + "loss": 0.9181, + "step": 231870 + }, + { + "epoch": 20.07, + "learning_rate": 1.6572278886994712e-05, + "loss": 0.9441, + "step": 231880 + }, + { + "epoch": 20.07, + "learning_rate": 1.657083417608137e-05, + "loss": 0.8974, + "step": 231890 + }, + { + "epoch": 20.07, + "learning_rate": 1.656938946516802e-05, + "loss": 0.9607, + "step": 231900 + }, + { + "epoch": 20.07, + "learning_rate": 1.6567944754254675e-05, + "loss": 0.9101, + "step": 231910 + }, + { + "epoch": 20.07, + "learning_rate": 1.656650004334133e-05, + "loss": 0.9807, + "step": 231920 + }, + { + "epoch": 20.08, + "learning_rate": 1.6565055332427982e-05, + "loss": 0.9192, + "step": 231930 + }, + { + "epoch": 20.08, + "learning_rate": 1.6563610621514636e-05, + "loss": 0.9545, + "step": 231940 + }, + { + "epoch": 20.08, + "learning_rate": 1.656216591060129e-05, + "loss": 0.934, + "step": 231950 + }, + { + "epoch": 20.08, + "learning_rate": 1.6560721199687942e-05, + "loss": 0.9666, + "step": 231960 + }, + { + "epoch": 20.08, + "learning_rate": 1.6559276488774596e-05, + "loss": 0.9852, + "step": 231970 + }, + { + "epoch": 20.08, + "learning_rate": 1.6557831777861253e-05, + "loss": 0.9172, + "step": 231980 + }, + { + "epoch": 20.08, + "learning_rate": 1.6556387066947903e-05, + "loss": 0.943, + "step": 231990 + }, + { + "epoch": 20.08, + "learning_rate": 1.655494235603456e-05, + "loss": 0.941, + "step": 232000 + }, + { + "epoch": 20.08, + "learning_rate": 1.6553497645121213e-05, + "loss": 0.9512, + "step": 232010 + }, + { + "epoch": 20.08, + "learning_rate": 1.6552052934207866e-05, + "loss": 0.9637, + "step": 232020 + }, + { + "epoch": 20.08, + "learning_rate": 1.655060822329452e-05, + "loss": 0.9882, + "step": 232030 + }, + { + "epoch": 20.08, + "learning_rate": 1.6549163512381173e-05, + "loss": 0.9475, + "step": 232040 + }, + { + "epoch": 20.09, + "learning_rate": 1.6547718801467826e-05, + "loss": 0.9727, + "step": 232050 + }, + { + "epoch": 20.09, + "learning_rate": 1.654627409055448e-05, + "loss": 0.9573, + "step": 232060 + }, + { + "epoch": 20.09, + "learning_rate": 1.6544829379641137e-05, + "loss": 0.9455, + "step": 232070 + }, + { + "epoch": 20.09, + "learning_rate": 1.6543384668727787e-05, + "loss": 0.9221, + "step": 232080 + }, + { + "epoch": 20.09, + "learning_rate": 1.6541939957814443e-05, + "loss": 0.9572, + "step": 232090 + }, + { + "epoch": 20.09, + "learning_rate": 1.6540495246901097e-05, + "loss": 0.9743, + "step": 232100 + }, + { + "epoch": 20.09, + "learning_rate": 1.653905053598775e-05, + "loss": 0.8984, + "step": 232110 + }, + { + "epoch": 20.09, + "learning_rate": 1.6537605825074404e-05, + "loss": 0.9298, + "step": 232120 + }, + { + "epoch": 20.09, + "learning_rate": 1.6536161114161057e-05, + "loss": 0.9457, + "step": 232130 + }, + { + "epoch": 20.09, + "learning_rate": 1.653471640324771e-05, + "loss": 0.9638, + "step": 232140 + }, + { + "epoch": 20.09, + "learning_rate": 1.6533271692334364e-05, + "loss": 0.9171, + "step": 232150 + }, + { + "epoch": 20.1, + "learning_rate": 1.653182698142102e-05, + "loss": 0.9601, + "step": 232160 + }, + { + "epoch": 20.1, + "learning_rate": 1.653038227050767e-05, + "loss": 0.8807, + "step": 232170 + }, + { + "epoch": 20.1, + "learning_rate": 1.6528937559594327e-05, + "loss": 0.9526, + "step": 232180 + }, + { + "epoch": 20.1, + "learning_rate": 1.652749284868098e-05, + "loss": 0.975, + "step": 232190 + }, + { + "epoch": 20.1, + "learning_rate": 1.6526048137767634e-05, + "loss": 0.9302, + "step": 232200 + }, + { + "epoch": 20.1, + "learning_rate": 1.6524603426854287e-05, + "loss": 0.9016, + "step": 232210 + }, + { + "epoch": 20.1, + "learning_rate": 1.652315871594094e-05, + "loss": 0.9787, + "step": 232220 + }, + { + "epoch": 20.1, + "learning_rate": 1.6521714005027594e-05, + "loss": 1.0251, + "step": 232230 + }, + { + "epoch": 20.1, + "learning_rate": 1.6520269294114248e-05, + "loss": 0.9082, + "step": 232240 + }, + { + "epoch": 20.1, + "learning_rate": 1.6518824583200904e-05, + "loss": 0.949, + "step": 232250 + }, + { + "epoch": 20.1, + "learning_rate": 1.6517379872287554e-05, + "loss": 0.9471, + "step": 232260 + }, + { + "epoch": 20.1, + "learning_rate": 1.651593516137421e-05, + "loss": 0.9702, + "step": 232270 + }, + { + "epoch": 20.11, + "learning_rate": 1.6514490450460865e-05, + "loss": 0.9549, + "step": 232280 + }, + { + "epoch": 20.11, + "learning_rate": 1.6513045739547518e-05, + "loss": 0.9084, + "step": 232290 + }, + { + "epoch": 20.11, + "learning_rate": 1.651160102863417e-05, + "loss": 0.9381, + "step": 232300 + }, + { + "epoch": 20.11, + "learning_rate": 1.6510156317720825e-05, + "loss": 0.8558, + "step": 232310 + }, + { + "epoch": 20.11, + "learning_rate": 1.6508711606807478e-05, + "loss": 0.9151, + "step": 232320 + }, + { + "epoch": 20.11, + "learning_rate": 1.650726689589413e-05, + "loss": 0.9284, + "step": 232330 + }, + { + "epoch": 20.11, + "learning_rate": 1.6505822184980785e-05, + "loss": 0.9687, + "step": 232340 + }, + { + "epoch": 20.11, + "learning_rate": 1.650437747406744e-05, + "loss": 0.9237, + "step": 232350 + }, + { + "epoch": 20.11, + "learning_rate": 1.6502932763154095e-05, + "loss": 1.0241, + "step": 232360 + }, + { + "epoch": 20.11, + "learning_rate": 1.6501488052240745e-05, + "loss": 0.9468, + "step": 232370 + }, + { + "epoch": 20.11, + "learning_rate": 1.6500043341327402e-05, + "loss": 0.9363, + "step": 232380 + }, + { + "epoch": 20.12, + "learning_rate": 1.6498598630414055e-05, + "loss": 0.8863, + "step": 232390 + }, + { + "epoch": 20.12, + "learning_rate": 1.649715391950071e-05, + "loss": 0.9523, + "step": 232400 + }, + { + "epoch": 20.12, + "learning_rate": 1.6495709208587362e-05, + "loss": 0.9889, + "step": 232410 + }, + { + "epoch": 20.12, + "learning_rate": 1.6494264497674016e-05, + "loss": 0.9523, + "step": 232420 + }, + { + "epoch": 20.12, + "learning_rate": 1.649281978676067e-05, + "loss": 0.8933, + "step": 232430 + }, + { + "epoch": 20.12, + "learning_rate": 1.6491375075847322e-05, + "loss": 0.9564, + "step": 232440 + }, + { + "epoch": 20.12, + "learning_rate": 1.648993036493398e-05, + "loss": 0.9511, + "step": 232450 + }, + { + "epoch": 20.12, + "learning_rate": 1.648848565402063e-05, + "loss": 1.0246, + "step": 232460 + }, + { + "epoch": 20.12, + "learning_rate": 1.6487040943107286e-05, + "loss": 0.9403, + "step": 232470 + }, + { + "epoch": 20.12, + "learning_rate": 1.648559623219394e-05, + "loss": 1.0138, + "step": 232480 + }, + { + "epoch": 20.12, + "learning_rate": 1.6484151521280593e-05, + "loss": 0.9649, + "step": 232490 + }, + { + "epoch": 20.12, + "learning_rate": 1.6482706810367246e-05, + "loss": 0.9336, + "step": 232500 + }, + { + "epoch": 20.13, + "learning_rate": 1.64812620994539e-05, + "loss": 0.9387, + "step": 232510 + }, + { + "epoch": 20.13, + "learning_rate": 1.6479817388540553e-05, + "loss": 0.9613, + "step": 232520 + }, + { + "epoch": 20.13, + "learning_rate": 1.6478372677627206e-05, + "loss": 0.95, + "step": 232530 + }, + { + "epoch": 20.13, + "learning_rate": 1.6476927966713863e-05, + "loss": 0.958, + "step": 232540 + }, + { + "epoch": 20.13, + "learning_rate": 1.6475483255800513e-05, + "loss": 0.9614, + "step": 232550 + }, + { + "epoch": 20.13, + "learning_rate": 1.647403854488717e-05, + "loss": 0.899, + "step": 232560 + }, + { + "epoch": 20.13, + "learning_rate": 1.6472593833973823e-05, + "loss": 0.9511, + "step": 232570 + }, + { + "epoch": 20.13, + "learning_rate": 1.6471149123060477e-05, + "loss": 0.9202, + "step": 232580 + }, + { + "epoch": 20.13, + "learning_rate": 1.646970441214713e-05, + "loss": 1.0037, + "step": 232590 + }, + { + "epoch": 20.13, + "learning_rate": 1.6468259701233783e-05, + "loss": 1.0083, + "step": 232600 + }, + { + "epoch": 20.13, + "learning_rate": 1.6466814990320437e-05, + "loss": 0.8911, + "step": 232610 + }, + { + "epoch": 20.14, + "learning_rate": 1.646537027940709e-05, + "loss": 0.9323, + "step": 232620 + }, + { + "epoch": 20.14, + "learning_rate": 1.6463925568493747e-05, + "loss": 0.9615, + "step": 232630 + }, + { + "epoch": 20.14, + "learning_rate": 1.6462480857580397e-05, + "loss": 0.942, + "step": 232640 + }, + { + "epoch": 20.14, + "learning_rate": 1.6461036146667054e-05, + "loss": 0.9605, + "step": 232650 + }, + { + "epoch": 20.14, + "learning_rate": 1.6459591435753707e-05, + "loss": 0.9966, + "step": 232660 + }, + { + "epoch": 20.14, + "learning_rate": 1.645814672484036e-05, + "loss": 0.9467, + "step": 232670 + }, + { + "epoch": 20.14, + "learning_rate": 1.6456702013927014e-05, + "loss": 0.904, + "step": 232680 + }, + { + "epoch": 20.14, + "learning_rate": 1.6455257303013667e-05, + "loss": 0.9475, + "step": 232690 + }, + { + "epoch": 20.14, + "learning_rate": 1.645381259210032e-05, + "loss": 0.9006, + "step": 232700 + }, + { + "epoch": 20.14, + "learning_rate": 1.6452367881186974e-05, + "loss": 0.9408, + "step": 232710 + }, + { + "epoch": 20.14, + "learning_rate": 1.645092317027363e-05, + "loss": 0.9421, + "step": 232720 + }, + { + "epoch": 20.14, + "learning_rate": 1.644947845936028e-05, + "loss": 0.935, + "step": 232730 + }, + { + "epoch": 20.15, + "learning_rate": 1.6448033748446938e-05, + "loss": 0.9017, + "step": 232740 + }, + { + "epoch": 20.15, + "learning_rate": 1.644658903753359e-05, + "loss": 0.9398, + "step": 232750 + }, + { + "epoch": 20.15, + "learning_rate": 1.6445144326620244e-05, + "loss": 0.9757, + "step": 232760 + }, + { + "epoch": 20.15, + "learning_rate": 1.6443699615706898e-05, + "loss": 0.9478, + "step": 232770 + }, + { + "epoch": 20.15, + "learning_rate": 1.644225490479355e-05, + "loss": 0.9086, + "step": 232780 + }, + { + "epoch": 20.15, + "learning_rate": 1.6440810193880205e-05, + "loss": 0.9449, + "step": 232790 + }, + { + "epoch": 20.15, + "learning_rate": 1.6439365482966858e-05, + "loss": 0.9138, + "step": 232800 + }, + { + "epoch": 20.15, + "learning_rate": 1.6437920772053515e-05, + "loss": 0.9278, + "step": 232810 + }, + { + "epoch": 20.15, + "learning_rate": 1.6436476061140165e-05, + "loss": 0.9518, + "step": 232820 + }, + { + "epoch": 20.15, + "learning_rate": 1.643503135022682e-05, + "loss": 0.9545, + "step": 232830 + }, + { + "epoch": 20.15, + "learning_rate": 1.6433586639313475e-05, + "loss": 0.9108, + "step": 232840 + }, + { + "epoch": 20.15, + "learning_rate": 1.643214192840013e-05, + "loss": 0.9299, + "step": 232850 + }, + { + "epoch": 20.16, + "learning_rate": 1.6430697217486782e-05, + "loss": 1.0, + "step": 232860 + }, + { + "epoch": 20.16, + "learning_rate": 1.6429252506573435e-05, + "loss": 0.9566, + "step": 232870 + }, + { + "epoch": 20.16, + "learning_rate": 1.642780779566009e-05, + "loss": 0.9919, + "step": 232880 + }, + { + "epoch": 20.16, + "learning_rate": 1.6426363084746742e-05, + "loss": 0.9376, + "step": 232890 + }, + { + "epoch": 20.16, + "learning_rate": 1.64249183738334e-05, + "loss": 0.9595, + "step": 232900 + }, + { + "epoch": 20.16, + "learning_rate": 1.642347366292005e-05, + "loss": 0.9482, + "step": 232910 + }, + { + "epoch": 20.16, + "learning_rate": 1.6422028952006706e-05, + "loss": 0.9571, + "step": 232920 + }, + { + "epoch": 20.16, + "learning_rate": 1.642058424109336e-05, + "loss": 0.9115, + "step": 232930 + }, + { + "epoch": 20.16, + "learning_rate": 1.6419139530180012e-05, + "loss": 0.9402, + "step": 232940 + }, + { + "epoch": 20.16, + "learning_rate": 1.6417694819266666e-05, + "loss": 1.0169, + "step": 232950 + }, + { + "epoch": 20.16, + "learning_rate": 1.641625010835332e-05, + "loss": 0.9298, + "step": 232960 + }, + { + "epoch": 20.17, + "learning_rate": 1.6414805397439973e-05, + "loss": 0.9316, + "step": 232970 + }, + { + "epoch": 20.17, + "learning_rate": 1.6413360686526626e-05, + "loss": 0.8983, + "step": 232980 + }, + { + "epoch": 20.17, + "learning_rate": 1.6411915975613283e-05, + "loss": 0.9126, + "step": 232990 + }, + { + "epoch": 20.17, + "learning_rate": 1.6410471264699933e-05, + "loss": 0.9282, + "step": 233000 + }, + { + "epoch": 20.17, + "learning_rate": 1.640902655378659e-05, + "loss": 0.9799, + "step": 233010 + }, + { + "epoch": 20.17, + "learning_rate": 1.6407581842873243e-05, + "loss": 0.955, + "step": 233020 + }, + { + "epoch": 20.17, + "learning_rate": 1.6406137131959896e-05, + "loss": 0.9599, + "step": 233030 + }, + { + "epoch": 20.17, + "learning_rate": 1.640469242104655e-05, + "loss": 0.9685, + "step": 233040 + }, + { + "epoch": 20.17, + "learning_rate": 1.6403247710133203e-05, + "loss": 0.9335, + "step": 233050 + }, + { + "epoch": 20.17, + "learning_rate": 1.6401802999219857e-05, + "loss": 0.9885, + "step": 233060 + }, + { + "epoch": 20.17, + "learning_rate": 1.640035828830651e-05, + "loss": 0.9459, + "step": 233070 + }, + { + "epoch": 20.17, + "learning_rate": 1.6398913577393167e-05, + "loss": 0.9529, + "step": 233080 + }, + { + "epoch": 20.18, + "learning_rate": 1.6397468866479817e-05, + "loss": 0.9749, + "step": 233090 + }, + { + "epoch": 20.18, + "learning_rate": 1.6396024155566473e-05, + "loss": 0.9292, + "step": 233100 + }, + { + "epoch": 20.18, + "learning_rate": 1.6394579444653127e-05, + "loss": 0.9373, + "step": 233110 + }, + { + "epoch": 20.18, + "learning_rate": 1.639313473373978e-05, + "loss": 0.9061, + "step": 233120 + }, + { + "epoch": 20.18, + "learning_rate": 1.6391690022826434e-05, + "loss": 0.9025, + "step": 233130 + }, + { + "epoch": 20.18, + "learning_rate": 1.6390245311913087e-05, + "loss": 0.9598, + "step": 233140 + }, + { + "epoch": 20.18, + "learning_rate": 1.638880060099974e-05, + "loss": 0.921, + "step": 233150 + }, + { + "epoch": 20.18, + "learning_rate": 1.6387355890086394e-05, + "loss": 1.0265, + "step": 233160 + }, + { + "epoch": 20.18, + "learning_rate": 1.638591117917305e-05, + "loss": 0.9133, + "step": 233170 + }, + { + "epoch": 20.18, + "learning_rate": 1.63844664682597e-05, + "loss": 0.9286, + "step": 233180 + }, + { + "epoch": 20.18, + "learning_rate": 1.6383021757346357e-05, + "loss": 0.8967, + "step": 233190 + }, + { + "epoch": 20.19, + "learning_rate": 1.6381577046433007e-05, + "loss": 0.924, + "step": 233200 + }, + { + "epoch": 20.19, + "learning_rate": 1.6380132335519664e-05, + "loss": 0.9789, + "step": 233210 + }, + { + "epoch": 20.19, + "learning_rate": 1.6378687624606318e-05, + "loss": 0.9403, + "step": 233220 + }, + { + "epoch": 20.19, + "learning_rate": 1.637724291369297e-05, + "loss": 0.9801, + "step": 233230 + }, + { + "epoch": 20.19, + "learning_rate": 1.6375798202779624e-05, + "loss": 0.9494, + "step": 233240 + }, + { + "epoch": 20.19, + "learning_rate": 1.6374353491866278e-05, + "loss": 0.8893, + "step": 233250 + }, + { + "epoch": 20.19, + "learning_rate": 1.637290878095293e-05, + "loss": 0.9662, + "step": 233260 + }, + { + "epoch": 20.19, + "learning_rate": 1.6371464070039585e-05, + "loss": 0.9257, + "step": 233270 + }, + { + "epoch": 20.19, + "learning_rate": 1.637001935912624e-05, + "loss": 0.9139, + "step": 233280 + }, + { + "epoch": 20.19, + "learning_rate": 1.636857464821289e-05, + "loss": 0.983, + "step": 233290 + }, + { + "epoch": 20.19, + "learning_rate": 1.6367129937299548e-05, + "loss": 0.9162, + "step": 233300 + }, + { + "epoch": 20.19, + "learning_rate": 1.63656852263862e-05, + "loss": 0.924, + "step": 233310 + }, + { + "epoch": 20.2, + "learning_rate": 1.6364240515472855e-05, + "loss": 0.9617, + "step": 233320 + }, + { + "epoch": 20.2, + "learning_rate": 1.636279580455951e-05, + "loss": 0.9435, + "step": 233330 + }, + { + "epoch": 20.2, + "learning_rate": 1.6361351093646162e-05, + "loss": 0.8991, + "step": 233340 + }, + { + "epoch": 20.2, + "learning_rate": 1.6359906382732815e-05, + "loss": 0.9689, + "step": 233350 + }, + { + "epoch": 20.2, + "learning_rate": 1.635846167181947e-05, + "loss": 0.9575, + "step": 233360 + }, + { + "epoch": 20.2, + "learning_rate": 1.6357016960906125e-05, + "loss": 0.9858, + "step": 233370 + }, + { + "epoch": 20.2, + "learning_rate": 1.6355572249992775e-05, + "loss": 0.9361, + "step": 233380 + }, + { + "epoch": 20.2, + "learning_rate": 1.6354127539079432e-05, + "loss": 0.9556, + "step": 233390 + }, + { + "epoch": 20.2, + "learning_rate": 1.6352682828166085e-05, + "loss": 0.9333, + "step": 233400 + }, + { + "epoch": 20.2, + "learning_rate": 1.635123811725274e-05, + "loss": 0.9217, + "step": 233410 + }, + { + "epoch": 20.2, + "learning_rate": 1.6349793406339392e-05, + "loss": 0.915, + "step": 233420 + }, + { + "epoch": 20.21, + "learning_rate": 1.6348348695426046e-05, + "loss": 0.9715, + "step": 233430 + }, + { + "epoch": 20.21, + "learning_rate": 1.63469039845127e-05, + "loss": 0.9462, + "step": 233440 + }, + { + "epoch": 20.21, + "learning_rate": 1.6345459273599352e-05, + "loss": 0.9739, + "step": 233450 + }, + { + "epoch": 20.21, + "learning_rate": 1.634401456268601e-05, + "loss": 0.9593, + "step": 233460 + }, + { + "epoch": 20.21, + "learning_rate": 1.634256985177266e-05, + "loss": 0.8712, + "step": 233470 + }, + { + "epoch": 20.21, + "learning_rate": 1.6341125140859316e-05, + "loss": 0.9588, + "step": 233480 + }, + { + "epoch": 20.21, + "learning_rate": 1.633968042994597e-05, + "loss": 0.9475, + "step": 233490 + }, + { + "epoch": 20.21, + "learning_rate": 1.6338235719032623e-05, + "loss": 0.9693, + "step": 233500 + }, + { + "epoch": 20.21, + "learning_rate": 1.6336791008119276e-05, + "loss": 0.9317, + "step": 233510 + }, + { + "epoch": 20.21, + "learning_rate": 1.633534629720593e-05, + "loss": 0.9846, + "step": 233520 + }, + { + "epoch": 20.21, + "learning_rate": 1.6333901586292583e-05, + "loss": 0.9559, + "step": 233530 + }, + { + "epoch": 20.21, + "learning_rate": 1.6332456875379236e-05, + "loss": 0.975, + "step": 233540 + }, + { + "epoch": 20.22, + "learning_rate": 1.6331012164465893e-05, + "loss": 0.9747, + "step": 233550 + }, + { + "epoch": 20.22, + "learning_rate": 1.6329567453552543e-05, + "loss": 0.9203, + "step": 233560 + }, + { + "epoch": 20.22, + "learning_rate": 1.63281227426392e-05, + "loss": 0.9147, + "step": 233570 + }, + { + "epoch": 20.22, + "learning_rate": 1.6326678031725853e-05, + "loss": 0.9536, + "step": 233580 + }, + { + "epoch": 20.22, + "learning_rate": 1.6325233320812507e-05, + "loss": 0.9205, + "step": 233590 + }, + { + "epoch": 20.22, + "learning_rate": 1.632378860989916e-05, + "loss": 0.928, + "step": 233600 + }, + { + "epoch": 20.22, + "learning_rate": 1.6322343898985814e-05, + "loss": 0.9493, + "step": 233610 + }, + { + "epoch": 20.22, + "learning_rate": 1.6320899188072467e-05, + "loss": 0.9232, + "step": 233620 + }, + { + "epoch": 20.22, + "learning_rate": 1.631945447715912e-05, + "loss": 0.9622, + "step": 233630 + }, + { + "epoch": 20.22, + "learning_rate": 1.6318009766245777e-05, + "loss": 0.8863, + "step": 233640 + }, + { + "epoch": 20.22, + "learning_rate": 1.6316565055332427e-05, + "loss": 0.9162, + "step": 233650 + }, + { + "epoch": 20.23, + "learning_rate": 1.6315120344419084e-05, + "loss": 0.9548, + "step": 233660 + }, + { + "epoch": 20.23, + "learning_rate": 1.6313675633505737e-05, + "loss": 0.9229, + "step": 233670 + }, + { + "epoch": 20.23, + "learning_rate": 1.631223092259239e-05, + "loss": 0.9691, + "step": 233680 + }, + { + "epoch": 20.23, + "learning_rate": 1.6310786211679044e-05, + "loss": 0.9339, + "step": 233690 + }, + { + "epoch": 20.23, + "learning_rate": 1.6309341500765698e-05, + "loss": 0.9041, + "step": 233700 + }, + { + "epoch": 20.23, + "learning_rate": 1.630789678985235e-05, + "loss": 0.932, + "step": 233710 + }, + { + "epoch": 20.23, + "learning_rate": 1.6306452078939004e-05, + "loss": 0.9737, + "step": 233720 + }, + { + "epoch": 20.23, + "learning_rate": 1.630500736802566e-05, + "loss": 0.9594, + "step": 233730 + }, + { + "epoch": 20.23, + "learning_rate": 1.630356265711231e-05, + "loss": 0.9318, + "step": 233740 + }, + { + "epoch": 20.23, + "learning_rate": 1.6302117946198968e-05, + "loss": 0.9404, + "step": 233750 + }, + { + "epoch": 20.23, + "learning_rate": 1.630067323528562e-05, + "loss": 0.9306, + "step": 233760 + }, + { + "epoch": 20.23, + "learning_rate": 1.6299228524372275e-05, + "loss": 0.9318, + "step": 233770 + }, + { + "epoch": 20.24, + "learning_rate": 1.6297783813458928e-05, + "loss": 0.946, + "step": 233780 + }, + { + "epoch": 20.24, + "learning_rate": 1.629633910254558e-05, + "loss": 0.9293, + "step": 233790 + }, + { + "epoch": 20.24, + "learning_rate": 1.6294894391632235e-05, + "loss": 0.9716, + "step": 233800 + }, + { + "epoch": 20.24, + "learning_rate": 1.6293449680718888e-05, + "loss": 0.9633, + "step": 233810 + }, + { + "epoch": 20.24, + "learning_rate": 1.6292004969805545e-05, + "loss": 0.9066, + "step": 233820 + }, + { + "epoch": 20.24, + "learning_rate": 1.6290560258892195e-05, + "loss": 0.9504, + "step": 233830 + }, + { + "epoch": 20.24, + "learning_rate": 1.6289115547978852e-05, + "loss": 0.9573, + "step": 233840 + }, + { + "epoch": 20.24, + "learning_rate": 1.6287670837065505e-05, + "loss": 0.9441, + "step": 233850 + }, + { + "epoch": 20.24, + "learning_rate": 1.628622612615216e-05, + "loss": 0.8772, + "step": 233860 + }, + { + "epoch": 20.24, + "learning_rate": 1.6284781415238812e-05, + "loss": 0.9361, + "step": 233870 + }, + { + "epoch": 20.24, + "learning_rate": 1.6283336704325465e-05, + "loss": 0.9342, + "step": 233880 + }, + { + "epoch": 20.24, + "learning_rate": 1.628189199341212e-05, + "loss": 1.0127, + "step": 233890 + }, + { + "epoch": 20.25, + "learning_rate": 1.6280447282498772e-05, + "loss": 0.9809, + "step": 233900 + }, + { + "epoch": 20.25, + "learning_rate": 1.627900257158543e-05, + "loss": 1.0362, + "step": 233910 + }, + { + "epoch": 20.25, + "learning_rate": 1.627755786067208e-05, + "loss": 0.9381, + "step": 233920 + }, + { + "epoch": 20.25, + "learning_rate": 1.6276113149758736e-05, + "loss": 1.0134, + "step": 233930 + }, + { + "epoch": 20.25, + "learning_rate": 1.627466843884539e-05, + "loss": 0.9356, + "step": 233940 + }, + { + "epoch": 20.25, + "learning_rate": 1.6273223727932043e-05, + "loss": 0.9572, + "step": 233950 + }, + { + "epoch": 20.25, + "learning_rate": 1.6271779017018696e-05, + "loss": 0.8769, + "step": 233960 + }, + { + "epoch": 20.25, + "learning_rate": 1.627033430610535e-05, + "loss": 0.9396, + "step": 233970 + }, + { + "epoch": 20.25, + "learning_rate": 1.6268889595192003e-05, + "loss": 0.9935, + "step": 233980 + }, + { + "epoch": 20.25, + "learning_rate": 1.6267444884278656e-05, + "loss": 0.9782, + "step": 233990 + }, + { + "epoch": 20.25, + "learning_rate": 1.6266000173365313e-05, + "loss": 0.9702, + "step": 234000 + }, + { + "epoch": 20.26, + "learning_rate": 1.6264555462451963e-05, + "loss": 0.8849, + "step": 234010 + }, + { + "epoch": 20.26, + "learning_rate": 1.626311075153862e-05, + "loss": 0.8841, + "step": 234020 + }, + { + "epoch": 20.26, + "learning_rate": 1.6261666040625273e-05, + "loss": 0.9723, + "step": 234030 + }, + { + "epoch": 20.26, + "learning_rate": 1.6260221329711926e-05, + "loss": 0.947, + "step": 234040 + }, + { + "epoch": 20.26, + "learning_rate": 1.625877661879858e-05, + "loss": 0.9288, + "step": 234050 + }, + { + "epoch": 20.26, + "learning_rate": 1.6257331907885233e-05, + "loss": 0.9505, + "step": 234060 + }, + { + "epoch": 20.26, + "learning_rate": 1.6255887196971887e-05, + "loss": 0.9008, + "step": 234070 + }, + { + "epoch": 20.26, + "learning_rate": 1.625444248605854e-05, + "loss": 0.9923, + "step": 234080 + }, + { + "epoch": 20.26, + "learning_rate": 1.6252997775145193e-05, + "loss": 0.9981, + "step": 234090 + }, + { + "epoch": 20.26, + "learning_rate": 1.6251553064231847e-05, + "loss": 0.8907, + "step": 234100 + }, + { + "epoch": 20.26, + "learning_rate": 1.6250108353318504e-05, + "loss": 0.9448, + "step": 234110 + }, + { + "epoch": 20.26, + "learning_rate": 1.6248663642405154e-05, + "loss": 0.9359, + "step": 234120 + }, + { + "epoch": 20.27, + "learning_rate": 1.624721893149181e-05, + "loss": 0.9642, + "step": 234130 + }, + { + "epoch": 20.27, + "learning_rate": 1.6245774220578464e-05, + "loss": 0.909, + "step": 234140 + }, + { + "epoch": 20.27, + "learning_rate": 1.6244329509665114e-05, + "loss": 1.0263, + "step": 234150 + }, + { + "epoch": 20.27, + "learning_rate": 1.624288479875177e-05, + "loss": 0.9036, + "step": 234160 + }, + { + "epoch": 20.27, + "learning_rate": 1.6241440087838424e-05, + "loss": 0.961, + "step": 234170 + }, + { + "epoch": 20.27, + "learning_rate": 1.6239995376925077e-05, + "loss": 0.953, + "step": 234180 + }, + { + "epoch": 20.27, + "learning_rate": 1.623855066601173e-05, + "loss": 0.9832, + "step": 234190 + }, + { + "epoch": 20.27, + "learning_rate": 1.6237105955098388e-05, + "loss": 0.9026, + "step": 234200 + }, + { + "epoch": 20.27, + "learning_rate": 1.6235661244185038e-05, + "loss": 1.0194, + "step": 234210 + }, + { + "epoch": 20.27, + "learning_rate": 1.6234216533271694e-05, + "loss": 0.917, + "step": 234220 + }, + { + "epoch": 20.27, + "learning_rate": 1.6232771822358348e-05, + "loss": 0.9733, + "step": 234230 + }, + { + "epoch": 20.28, + "learning_rate": 1.6231327111444998e-05, + "loss": 0.9862, + "step": 234240 + }, + { + "epoch": 20.28, + "learning_rate": 1.6229882400531655e-05, + "loss": 0.9003, + "step": 234250 + }, + { + "epoch": 20.28, + "learning_rate": 1.6228437689618308e-05, + "loss": 0.9534, + "step": 234260 + }, + { + "epoch": 20.28, + "learning_rate": 1.622699297870496e-05, + "loss": 1.0465, + "step": 234270 + }, + { + "epoch": 20.28, + "learning_rate": 1.6225548267791615e-05, + "loss": 0.8553, + "step": 234280 + }, + { + "epoch": 20.28, + "learning_rate": 1.622410355687827e-05, + "loss": 0.9764, + "step": 234290 + }, + { + "epoch": 20.28, + "learning_rate": 1.622265884596492e-05, + "loss": 0.9363, + "step": 234300 + }, + { + "epoch": 20.28, + "learning_rate": 1.622121413505158e-05, + "loss": 0.9621, + "step": 234310 + }, + { + "epoch": 20.28, + "learning_rate": 1.6219769424138232e-05, + "loss": 0.9914, + "step": 234320 + }, + { + "epoch": 20.28, + "learning_rate": 1.6218324713224882e-05, + "loss": 0.917, + "step": 234330 + }, + { + "epoch": 20.28, + "learning_rate": 1.621688000231154e-05, + "loss": 0.849, + "step": 234340 + }, + { + "epoch": 20.28, + "learning_rate": 1.6215435291398192e-05, + "loss": 0.901, + "step": 234350 + }, + { + "epoch": 20.29, + "learning_rate": 1.6213990580484845e-05, + "loss": 0.8663, + "step": 234360 + }, + { + "epoch": 20.29, + "learning_rate": 1.62125458695715e-05, + "loss": 0.9327, + "step": 234370 + }, + { + "epoch": 20.29, + "learning_rate": 1.6211101158658155e-05, + "loss": 0.9012, + "step": 234380 + }, + { + "epoch": 20.29, + "learning_rate": 1.6209656447744805e-05, + "loss": 0.9357, + "step": 234390 + }, + { + "epoch": 20.29, + "learning_rate": 1.6208211736831462e-05, + "loss": 0.9665, + "step": 234400 + }, + { + "epoch": 20.29, + "learning_rate": 1.6206767025918116e-05, + "loss": 0.8971, + "step": 234410 + }, + { + "epoch": 20.29, + "learning_rate": 1.6205322315004766e-05, + "loss": 0.9709, + "step": 234420 + }, + { + "epoch": 20.29, + "learning_rate": 1.6203877604091422e-05, + "loss": 0.9126, + "step": 234430 + }, + { + "epoch": 20.29, + "learning_rate": 1.6202432893178076e-05, + "loss": 0.9171, + "step": 234440 + }, + { + "epoch": 20.29, + "learning_rate": 1.620098818226473e-05, + "loss": 1.0011, + "step": 234450 + }, + { + "epoch": 20.29, + "learning_rate": 1.6199543471351383e-05, + "loss": 0.9431, + "step": 234460 + }, + { + "epoch": 20.3, + "learning_rate": 1.619809876043804e-05, + "loss": 0.949, + "step": 234470 + }, + { + "epoch": 20.3, + "learning_rate": 1.619665404952469e-05, + "loss": 0.9569, + "step": 234480 + }, + { + "epoch": 20.3, + "learning_rate": 1.6195209338611346e-05, + "loss": 0.9404, + "step": 234490 + }, + { + "epoch": 20.3, + "learning_rate": 1.6193764627698e-05, + "loss": 0.9891, + "step": 234500 + }, + { + "epoch": 20.3, + "learning_rate": 1.619231991678465e-05, + "loss": 0.9301, + "step": 234510 + }, + { + "epoch": 20.3, + "learning_rate": 1.6190875205871306e-05, + "loss": 0.9399, + "step": 234520 + }, + { + "epoch": 20.3, + "learning_rate": 1.618943049495796e-05, + "loss": 0.8529, + "step": 234530 + }, + { + "epoch": 20.3, + "learning_rate": 1.6187985784044613e-05, + "loss": 0.8778, + "step": 234540 + }, + { + "epoch": 20.3, + "learning_rate": 1.6186541073131267e-05, + "loss": 0.9361, + "step": 234550 + }, + { + "epoch": 20.3, + "learning_rate": 1.6185096362217923e-05, + "loss": 1.0012, + "step": 234560 + }, + { + "epoch": 20.3, + "learning_rate": 1.6183651651304573e-05, + "loss": 0.9283, + "step": 234570 + }, + { + "epoch": 20.3, + "learning_rate": 1.618220694039123e-05, + "loss": 0.9288, + "step": 234580 + }, + { + "epoch": 20.31, + "learning_rate": 1.6180762229477884e-05, + "loss": 0.9661, + "step": 234590 + }, + { + "epoch": 20.31, + "learning_rate": 1.6179317518564534e-05, + "loss": 0.9177, + "step": 234600 + }, + { + "epoch": 20.31, + "learning_rate": 1.617787280765119e-05, + "loss": 0.9014, + "step": 234610 + }, + { + "epoch": 20.31, + "learning_rate": 1.6176428096737844e-05, + "loss": 0.9683, + "step": 234620 + }, + { + "epoch": 20.31, + "learning_rate": 1.6174983385824497e-05, + "loss": 0.8921, + "step": 234630 + }, + { + "epoch": 20.31, + "learning_rate": 1.617353867491115e-05, + "loss": 0.867, + "step": 234640 + }, + { + "epoch": 20.31, + "learning_rate": 1.6172093963997807e-05, + "loss": 0.9396, + "step": 234650 + }, + { + "epoch": 20.31, + "learning_rate": 1.6170649253084457e-05, + "loss": 0.966, + "step": 234660 + }, + { + "epoch": 20.31, + "learning_rate": 1.6169204542171114e-05, + "loss": 0.9688, + "step": 234670 + }, + { + "epoch": 20.31, + "learning_rate": 1.6167759831257767e-05, + "loss": 0.9136, + "step": 234680 + }, + { + "epoch": 20.31, + "learning_rate": 1.6166315120344417e-05, + "loss": 0.919, + "step": 234690 + }, + { + "epoch": 20.32, + "learning_rate": 1.6164870409431074e-05, + "loss": 0.9569, + "step": 234700 + }, + { + "epoch": 20.32, + "learning_rate": 1.6163425698517728e-05, + "loss": 0.945, + "step": 234710 + }, + { + "epoch": 20.32, + "learning_rate": 1.616198098760438e-05, + "loss": 0.9172, + "step": 234720 + }, + { + "epoch": 20.32, + "learning_rate": 1.6160536276691034e-05, + "loss": 0.9197, + "step": 234730 + }, + { + "epoch": 20.32, + "learning_rate": 1.615909156577769e-05, + "loss": 0.9594, + "step": 234740 + }, + { + "epoch": 20.32, + "learning_rate": 1.615764685486434e-05, + "loss": 0.9681, + "step": 234750 + }, + { + "epoch": 20.32, + "learning_rate": 1.6156202143950998e-05, + "loss": 0.9587, + "step": 234760 + }, + { + "epoch": 20.32, + "learning_rate": 1.615475743303765e-05, + "loss": 0.9324, + "step": 234770 + }, + { + "epoch": 20.32, + "learning_rate": 1.61533127221243e-05, + "loss": 0.9148, + "step": 234780 + }, + { + "epoch": 20.32, + "learning_rate": 1.6151868011210958e-05, + "loss": 0.978, + "step": 234790 + }, + { + "epoch": 20.32, + "learning_rate": 1.615042330029761e-05, + "loss": 0.9467, + "step": 234800 + }, + { + "epoch": 20.32, + "learning_rate": 1.6148978589384265e-05, + "loss": 0.9842, + "step": 234810 + }, + { + "epoch": 20.33, + "learning_rate": 1.614753387847092e-05, + "loss": 1.0002, + "step": 234820 + }, + { + "epoch": 20.33, + "learning_rate": 1.6146089167557575e-05, + "loss": 0.966, + "step": 234830 + }, + { + "epoch": 20.33, + "learning_rate": 1.6144644456644225e-05, + "loss": 0.9431, + "step": 234840 + }, + { + "epoch": 20.33, + "learning_rate": 1.6143199745730882e-05, + "loss": 0.9303, + "step": 234850 + }, + { + "epoch": 20.33, + "learning_rate": 1.6141755034817535e-05, + "loss": 0.9079, + "step": 234860 + }, + { + "epoch": 20.33, + "learning_rate": 1.6140310323904185e-05, + "loss": 0.9204, + "step": 234870 + }, + { + "epoch": 20.33, + "learning_rate": 1.6138865612990842e-05, + "loss": 0.9593, + "step": 234880 + }, + { + "epoch": 20.33, + "learning_rate": 1.6137420902077496e-05, + "loss": 0.9815, + "step": 234890 + }, + { + "epoch": 20.33, + "learning_rate": 1.613597619116415e-05, + "loss": 0.988, + "step": 234900 + }, + { + "epoch": 20.33, + "learning_rate": 1.6134531480250802e-05, + "loss": 0.9334, + "step": 234910 + }, + { + "epoch": 20.33, + "learning_rate": 1.613308676933746e-05, + "loss": 0.9838, + "step": 234920 + }, + { + "epoch": 20.33, + "learning_rate": 1.613164205842411e-05, + "loss": 0.9358, + "step": 234930 + }, + { + "epoch": 20.34, + "learning_rate": 1.6130197347510766e-05, + "loss": 0.9423, + "step": 234940 + }, + { + "epoch": 20.34, + "learning_rate": 1.612875263659742e-05, + "loss": 0.9409, + "step": 234950 + }, + { + "epoch": 20.34, + "learning_rate": 1.612730792568407e-05, + "loss": 0.9103, + "step": 234960 + }, + { + "epoch": 20.34, + "learning_rate": 1.6125863214770726e-05, + "loss": 0.9297, + "step": 234970 + }, + { + "epoch": 20.34, + "learning_rate": 1.612441850385738e-05, + "loss": 1.0065, + "step": 234980 + }, + { + "epoch": 20.34, + "learning_rate": 1.6122973792944033e-05, + "loss": 0.916, + "step": 234990 + }, + { + "epoch": 20.34, + "learning_rate": 1.6121529082030686e-05, + "loss": 0.9344, + "step": 235000 + }, + { + "epoch": 20.34, + "learning_rate": 1.612008437111734e-05, + "loss": 0.9329, + "step": 235010 + }, + { + "epoch": 20.34, + "learning_rate": 1.6118639660203993e-05, + "loss": 0.993, + "step": 235020 + }, + { + "epoch": 20.34, + "learning_rate": 1.611719494929065e-05, + "loss": 0.9146, + "step": 235030 + }, + { + "epoch": 20.34, + "learning_rate": 1.61157502383773e-05, + "loss": 0.8582, + "step": 235040 + }, + { + "epoch": 20.35, + "learning_rate": 1.6114305527463953e-05, + "loss": 0.9994, + "step": 235050 + }, + { + "epoch": 20.35, + "learning_rate": 1.611286081655061e-05, + "loss": 0.9498, + "step": 235060 + }, + { + "epoch": 20.35, + "learning_rate": 1.611141610563726e-05, + "loss": 0.9005, + "step": 235070 + }, + { + "epoch": 20.35, + "learning_rate": 1.6109971394723917e-05, + "loss": 0.9119, + "step": 235080 + }, + { + "epoch": 20.35, + "learning_rate": 1.610852668381057e-05, + "loss": 0.9914, + "step": 235090 + }, + { + "epoch": 20.35, + "learning_rate": 1.6107081972897224e-05, + "loss": 0.9249, + "step": 235100 + }, + { + "epoch": 20.35, + "learning_rate": 1.6105637261983877e-05, + "loss": 0.924, + "step": 235110 + }, + { + "epoch": 20.35, + "learning_rate": 1.6104192551070534e-05, + "loss": 0.9385, + "step": 235120 + }, + { + "epoch": 20.35, + "learning_rate": 1.6102747840157184e-05, + "loss": 0.9436, + "step": 235130 + }, + { + "epoch": 20.35, + "learning_rate": 1.6101303129243837e-05, + "loss": 0.9162, + "step": 235140 + }, + { + "epoch": 20.35, + "learning_rate": 1.6099858418330494e-05, + "loss": 0.929, + "step": 235150 + }, + { + "epoch": 20.35, + "learning_rate": 1.6098413707417144e-05, + "loss": 0.9459, + "step": 235160 + }, + { + "epoch": 20.36, + "learning_rate": 1.60969689965038e-05, + "loss": 0.9698, + "step": 235170 + }, + { + "epoch": 20.36, + "learning_rate": 1.6095524285590454e-05, + "loss": 0.907, + "step": 235180 + }, + { + "epoch": 20.36, + "learning_rate": 1.6094079574677108e-05, + "loss": 0.9611, + "step": 235190 + }, + { + "epoch": 20.36, + "learning_rate": 1.609263486376376e-05, + "loss": 0.8761, + "step": 235200 + }, + { + "epoch": 20.36, + "learning_rate": 1.6091190152850418e-05, + "loss": 0.8893, + "step": 235210 + }, + { + "epoch": 20.36, + "learning_rate": 1.6089745441937068e-05, + "loss": 0.9606, + "step": 235220 + }, + { + "epoch": 20.36, + "learning_rate": 1.608830073102372e-05, + "loss": 1.0, + "step": 235230 + }, + { + "epoch": 20.36, + "learning_rate": 1.6086856020110378e-05, + "loss": 0.9352, + "step": 235240 + }, + { + "epoch": 20.36, + "learning_rate": 1.6085411309197028e-05, + "loss": 0.9292, + "step": 235250 + }, + { + "epoch": 20.36, + "learning_rate": 1.6083966598283685e-05, + "loss": 0.9758, + "step": 235260 + }, + { + "epoch": 20.36, + "learning_rate": 1.6082521887370338e-05, + "loss": 0.9557, + "step": 235270 + }, + { + "epoch": 20.37, + "learning_rate": 1.608107717645699e-05, + "loss": 0.9522, + "step": 235280 + }, + { + "epoch": 20.37, + "learning_rate": 1.6079632465543645e-05, + "loss": 0.9807, + "step": 235290 + }, + { + "epoch": 20.37, + "learning_rate": 1.60781877546303e-05, + "loss": 0.9141, + "step": 235300 + }, + { + "epoch": 20.37, + "learning_rate": 1.607674304371695e-05, + "loss": 1.0002, + "step": 235310 + }, + { + "epoch": 20.37, + "learning_rate": 1.6075298332803605e-05, + "loss": 0.9014, + "step": 235320 + }, + { + "epoch": 20.37, + "learning_rate": 1.6073853621890262e-05, + "loss": 0.9589, + "step": 235330 + }, + { + "epoch": 20.37, + "learning_rate": 1.6072408910976912e-05, + "loss": 0.9309, + "step": 235340 + }, + { + "epoch": 20.37, + "learning_rate": 1.607096420006357e-05, + "loss": 0.8788, + "step": 235350 + }, + { + "epoch": 20.37, + "learning_rate": 1.6069519489150222e-05, + "loss": 0.9304, + "step": 235360 + }, + { + "epoch": 20.37, + "learning_rate": 1.6068074778236875e-05, + "loss": 0.8875, + "step": 235370 + }, + { + "epoch": 20.37, + "learning_rate": 1.606663006732353e-05, + "loss": 0.9555, + "step": 235380 + }, + { + "epoch": 20.37, + "learning_rate": 1.6065185356410186e-05, + "loss": 0.9608, + "step": 235390 + }, + { + "epoch": 20.38, + "learning_rate": 1.6063740645496836e-05, + "loss": 0.9434, + "step": 235400 + }, + { + "epoch": 20.38, + "learning_rate": 1.606229593458349e-05, + "loss": 0.9997, + "step": 235410 + }, + { + "epoch": 20.38, + "learning_rate": 1.6060851223670146e-05, + "loss": 0.9346, + "step": 235420 + }, + { + "epoch": 20.38, + "learning_rate": 1.6059406512756796e-05, + "loss": 1.0046, + "step": 235430 + }, + { + "epoch": 20.38, + "learning_rate": 1.6057961801843453e-05, + "loss": 1.0267, + "step": 235440 + }, + { + "epoch": 20.38, + "learning_rate": 1.6056517090930106e-05, + "loss": 0.9137, + "step": 235450 + }, + { + "epoch": 20.38, + "learning_rate": 1.605507238001676e-05, + "loss": 0.9639, + "step": 235460 + }, + { + "epoch": 20.38, + "learning_rate": 1.6053627669103413e-05, + "loss": 0.8825, + "step": 235470 + }, + { + "epoch": 20.38, + "learning_rate": 1.605218295819007e-05, + "loss": 0.9316, + "step": 235480 + }, + { + "epoch": 20.38, + "learning_rate": 1.605073824727672e-05, + "loss": 0.9338, + "step": 235490 + }, + { + "epoch": 20.38, + "learning_rate": 1.6049293536363373e-05, + "loss": 0.9566, + "step": 235500 + }, + { + "epoch": 20.39, + "learning_rate": 1.604784882545003e-05, + "loss": 0.9555, + "step": 235510 + }, + { + "epoch": 20.39, + "learning_rate": 1.604640411453668e-05, + "loss": 1.0083, + "step": 235520 + }, + { + "epoch": 20.39, + "learning_rate": 1.6044959403623337e-05, + "loss": 0.9673, + "step": 235530 + }, + { + "epoch": 20.39, + "learning_rate": 1.604351469270999e-05, + "loss": 0.9876, + "step": 235540 + }, + { + "epoch": 20.39, + "learning_rate": 1.6042069981796643e-05, + "loss": 0.9466, + "step": 235550 + }, + { + "epoch": 20.39, + "learning_rate": 1.6040625270883297e-05, + "loss": 0.9817, + "step": 235560 + }, + { + "epoch": 20.39, + "learning_rate": 1.6039180559969954e-05, + "loss": 0.9703, + "step": 235570 + }, + { + "epoch": 20.39, + "learning_rate": 1.6037735849056604e-05, + "loss": 0.9476, + "step": 235580 + }, + { + "epoch": 20.39, + "learning_rate": 1.6036291138143257e-05, + "loss": 0.9188, + "step": 235590 + }, + { + "epoch": 20.39, + "learning_rate": 1.6034846427229914e-05, + "loss": 0.9591, + "step": 235600 + }, + { + "epoch": 20.39, + "learning_rate": 1.6033401716316564e-05, + "loss": 0.9391, + "step": 235610 + }, + { + "epoch": 20.39, + "learning_rate": 1.603195700540322e-05, + "loss": 0.9675, + "step": 235620 + }, + { + "epoch": 20.4, + "learning_rate": 1.6030512294489874e-05, + "loss": 0.9015, + "step": 235630 + }, + { + "epoch": 20.4, + "learning_rate": 1.6029067583576527e-05, + "loss": 0.9246, + "step": 235640 + }, + { + "epoch": 20.4, + "learning_rate": 1.602762287266318e-05, + "loss": 0.9115, + "step": 235650 + }, + { + "epoch": 20.4, + "learning_rate": 1.6026178161749837e-05, + "loss": 0.9124, + "step": 235660 + }, + { + "epoch": 20.4, + "learning_rate": 1.6024733450836487e-05, + "loss": 0.964, + "step": 235670 + }, + { + "epoch": 20.4, + "learning_rate": 1.602328873992314e-05, + "loss": 0.9061, + "step": 235680 + }, + { + "epoch": 20.4, + "learning_rate": 1.6021844029009798e-05, + "loss": 0.9314, + "step": 235690 + }, + { + "epoch": 20.4, + "learning_rate": 1.6020399318096448e-05, + "loss": 0.9819, + "step": 235700 + }, + { + "epoch": 20.4, + "learning_rate": 1.6018954607183104e-05, + "loss": 0.9032, + "step": 235710 + }, + { + "epoch": 20.4, + "learning_rate": 1.6017509896269758e-05, + "loss": 0.9753, + "step": 235720 + }, + { + "epoch": 20.4, + "learning_rate": 1.601606518535641e-05, + "loss": 0.9119, + "step": 235730 + }, + { + "epoch": 20.41, + "learning_rate": 1.6014620474443065e-05, + "loss": 0.9696, + "step": 235740 + }, + { + "epoch": 20.41, + "learning_rate": 1.601317576352972e-05, + "loss": 0.9362, + "step": 235750 + }, + { + "epoch": 20.41, + "learning_rate": 1.601173105261637e-05, + "loss": 0.9614, + "step": 235760 + }, + { + "epoch": 20.41, + "learning_rate": 1.6010286341703025e-05, + "loss": 0.9669, + "step": 235770 + }, + { + "epoch": 20.41, + "learning_rate": 1.600884163078968e-05, + "loss": 0.965, + "step": 235780 + }, + { + "epoch": 20.41, + "learning_rate": 1.600739691987633e-05, + "loss": 0.9713, + "step": 235790 + }, + { + "epoch": 20.41, + "learning_rate": 1.600595220896299e-05, + "loss": 0.9639, + "step": 235800 + }, + { + "epoch": 20.41, + "learning_rate": 1.6004507498049642e-05, + "loss": 0.9101, + "step": 235810 + }, + { + "epoch": 20.41, + "learning_rate": 1.6003062787136295e-05, + "loss": 0.9212, + "step": 235820 + }, + { + "epoch": 20.41, + "learning_rate": 1.600161807622295e-05, + "loss": 0.947, + "step": 235830 + }, + { + "epoch": 20.41, + "learning_rate": 1.6000173365309602e-05, + "loss": 0.8845, + "step": 235840 + }, + { + "epoch": 20.41, + "learning_rate": 1.5998728654396255e-05, + "loss": 0.9246, + "step": 235850 + }, + { + "epoch": 20.42, + "learning_rate": 1.599728394348291e-05, + "loss": 0.9051, + "step": 235860 + }, + { + "epoch": 20.42, + "learning_rate": 1.5995839232569566e-05, + "loss": 0.9862, + "step": 235870 + }, + { + "epoch": 20.42, + "learning_rate": 1.5994394521656216e-05, + "loss": 1.0033, + "step": 235880 + }, + { + "epoch": 20.42, + "learning_rate": 1.5992949810742872e-05, + "loss": 0.9453, + "step": 235890 + }, + { + "epoch": 20.42, + "learning_rate": 1.5991505099829522e-05, + "loss": 0.9312, + "step": 235900 + }, + { + "epoch": 20.42, + "learning_rate": 1.599006038891618e-05, + "loss": 0.9713, + "step": 235910 + }, + { + "epoch": 20.42, + "learning_rate": 1.5988615678002833e-05, + "loss": 0.878, + "step": 235920 + }, + { + "epoch": 20.42, + "learning_rate": 1.5987170967089486e-05, + "loss": 0.9076, + "step": 235930 + }, + { + "epoch": 20.42, + "learning_rate": 1.598572625617614e-05, + "loss": 0.9158, + "step": 235940 + }, + { + "epoch": 20.42, + "learning_rate": 1.5984281545262793e-05, + "loss": 0.989, + "step": 235950 + }, + { + "epoch": 20.42, + "learning_rate": 1.5982836834349446e-05, + "loss": 0.9462, + "step": 235960 + }, + { + "epoch": 20.42, + "learning_rate": 1.59813921234361e-05, + "loss": 0.9834, + "step": 235970 + }, + { + "epoch": 20.43, + "learning_rate": 1.5979947412522756e-05, + "loss": 0.9468, + "step": 235980 + }, + { + "epoch": 20.43, + "learning_rate": 1.5978502701609406e-05, + "loss": 0.9647, + "step": 235990 + }, + { + "epoch": 20.43, + "learning_rate": 1.5977057990696063e-05, + "loss": 0.9334, + "step": 236000 + }, + { + "epoch": 20.43, + "learning_rate": 1.5975613279782716e-05, + "loss": 0.9671, + "step": 236010 + }, + { + "epoch": 20.43, + "learning_rate": 1.597416856886937e-05, + "loss": 0.8954, + "step": 236020 + }, + { + "epoch": 20.43, + "learning_rate": 1.5972723857956023e-05, + "loss": 0.9399, + "step": 236030 + }, + { + "epoch": 20.43, + "learning_rate": 1.5971279147042677e-05, + "loss": 0.968, + "step": 236040 + }, + { + "epoch": 20.43, + "learning_rate": 1.596983443612933e-05, + "loss": 0.9622, + "step": 236050 + }, + { + "epoch": 20.43, + "learning_rate": 1.5968389725215983e-05, + "loss": 0.9296, + "step": 236060 + }, + { + "epoch": 20.43, + "learning_rate": 1.596694501430264e-05, + "loss": 0.9273, + "step": 236070 + }, + { + "epoch": 20.43, + "learning_rate": 1.596550030338929e-05, + "loss": 0.9245, + "step": 236080 + }, + { + "epoch": 20.44, + "learning_rate": 1.5964055592475947e-05, + "loss": 0.9614, + "step": 236090 + }, + { + "epoch": 20.44, + "learning_rate": 1.59626108815626e-05, + "loss": 0.9477, + "step": 236100 + }, + { + "epoch": 20.44, + "learning_rate": 1.5961166170649254e-05, + "loss": 0.9321, + "step": 236110 + }, + { + "epoch": 20.44, + "learning_rate": 1.5959721459735907e-05, + "loss": 0.9631, + "step": 236120 + }, + { + "epoch": 20.44, + "learning_rate": 1.595827674882256e-05, + "loss": 0.971, + "step": 236130 + }, + { + "epoch": 20.44, + "learning_rate": 1.5956832037909214e-05, + "loss": 0.9492, + "step": 236140 + }, + { + "epoch": 20.44, + "learning_rate": 1.5955387326995867e-05, + "loss": 0.9513, + "step": 236150 + }, + { + "epoch": 20.44, + "learning_rate": 1.5953942616082524e-05, + "loss": 0.9343, + "step": 236160 + }, + { + "epoch": 20.44, + "learning_rate": 1.5952497905169174e-05, + "loss": 0.9364, + "step": 236170 + }, + { + "epoch": 20.44, + "learning_rate": 1.595105319425583e-05, + "loss": 0.9499, + "step": 236180 + }, + { + "epoch": 20.44, + "learning_rate": 1.5949608483342484e-05, + "loss": 0.8922, + "step": 236190 + }, + { + "epoch": 20.44, + "learning_rate": 1.5948163772429138e-05, + "loss": 0.9695, + "step": 236200 + }, + { + "epoch": 20.45, + "learning_rate": 1.594671906151579e-05, + "loss": 0.9657, + "step": 236210 + }, + { + "epoch": 20.45, + "learning_rate": 1.5945274350602445e-05, + "loss": 0.9054, + "step": 236220 + }, + { + "epoch": 20.45, + "learning_rate": 1.5943829639689098e-05, + "loss": 0.9406, + "step": 236230 + }, + { + "epoch": 20.45, + "learning_rate": 1.594238492877575e-05, + "loss": 0.9248, + "step": 236240 + }, + { + "epoch": 20.45, + "learning_rate": 1.5940940217862408e-05, + "loss": 0.9347, + "step": 236250 + }, + { + "epoch": 20.45, + "learning_rate": 1.5939495506949058e-05, + "loss": 0.9594, + "step": 236260 + }, + { + "epoch": 20.45, + "learning_rate": 1.5938050796035715e-05, + "loss": 0.9263, + "step": 236270 + }, + { + "epoch": 20.45, + "learning_rate": 1.5936606085122368e-05, + "loss": 0.9728, + "step": 236280 + }, + { + "epoch": 20.45, + "learning_rate": 1.593516137420902e-05, + "loss": 0.9047, + "step": 236290 + }, + { + "epoch": 20.45, + "learning_rate": 1.5933716663295675e-05, + "loss": 0.9741, + "step": 236300 + }, + { + "epoch": 20.45, + "learning_rate": 1.593227195238233e-05, + "loss": 0.9521, + "step": 236310 + }, + { + "epoch": 20.46, + "learning_rate": 1.5930827241468982e-05, + "loss": 0.9396, + "step": 236320 + }, + { + "epoch": 20.46, + "learning_rate": 1.5929382530555635e-05, + "loss": 0.9408, + "step": 236330 + }, + { + "epoch": 20.46, + "learning_rate": 1.5927937819642292e-05, + "loss": 0.9712, + "step": 236340 + }, + { + "epoch": 20.46, + "learning_rate": 1.5926493108728942e-05, + "loss": 0.9426, + "step": 236350 + }, + { + "epoch": 20.46, + "learning_rate": 1.59250483978156e-05, + "loss": 0.9609, + "step": 236360 + }, + { + "epoch": 20.46, + "learning_rate": 1.5923603686902252e-05, + "loss": 0.9882, + "step": 236370 + }, + { + "epoch": 20.46, + "learning_rate": 1.5922158975988906e-05, + "loss": 0.898, + "step": 236380 + }, + { + "epoch": 20.46, + "learning_rate": 1.592071426507556e-05, + "loss": 0.9475, + "step": 236390 + }, + { + "epoch": 20.46, + "learning_rate": 1.5919269554162212e-05, + "loss": 0.9664, + "step": 236400 + }, + { + "epoch": 20.46, + "learning_rate": 1.5917824843248866e-05, + "loss": 0.9228, + "step": 236410 + }, + { + "epoch": 20.46, + "learning_rate": 1.591638013233552e-05, + "loss": 0.9541, + "step": 236420 + }, + { + "epoch": 20.46, + "learning_rate": 1.5914935421422176e-05, + "loss": 0.8929, + "step": 236430 + }, + { + "epoch": 20.47, + "learning_rate": 1.5913490710508826e-05, + "loss": 0.9015, + "step": 236440 + }, + { + "epoch": 20.47, + "learning_rate": 1.5912045999595483e-05, + "loss": 0.9923, + "step": 236450 + }, + { + "epoch": 20.47, + "learning_rate": 1.5910601288682136e-05, + "loss": 0.9267, + "step": 236460 + }, + { + "epoch": 20.47, + "learning_rate": 1.590915657776879e-05, + "loss": 0.9296, + "step": 236470 + }, + { + "epoch": 20.47, + "learning_rate": 1.5907711866855443e-05, + "loss": 0.9279, + "step": 236480 + }, + { + "epoch": 20.47, + "learning_rate": 1.5906267155942096e-05, + "loss": 0.9956, + "step": 236490 + }, + { + "epoch": 20.47, + "learning_rate": 1.590482244502875e-05, + "loss": 0.9513, + "step": 236500 + }, + { + "epoch": 20.47, + "learning_rate": 1.5903377734115403e-05, + "loss": 0.9606, + "step": 236510 + }, + { + "epoch": 20.47, + "learning_rate": 1.590193302320206e-05, + "loss": 0.899, + "step": 236520 + }, + { + "epoch": 20.47, + "learning_rate": 1.590048831228871e-05, + "loss": 0.9473, + "step": 236530 + }, + { + "epoch": 20.47, + "learning_rate": 1.5899043601375367e-05, + "loss": 0.9705, + "step": 236540 + }, + { + "epoch": 20.48, + "learning_rate": 1.589759889046202e-05, + "loss": 0.9501, + "step": 236550 + }, + { + "epoch": 20.48, + "learning_rate": 1.5896154179548673e-05, + "loss": 0.9355, + "step": 236560 + }, + { + "epoch": 20.48, + "learning_rate": 1.5894709468635327e-05, + "loss": 0.9548, + "step": 236570 + }, + { + "epoch": 20.48, + "learning_rate": 1.589326475772198e-05, + "loss": 0.8952, + "step": 236580 + }, + { + "epoch": 20.48, + "learning_rate": 1.5891820046808634e-05, + "loss": 0.9137, + "step": 236590 + }, + { + "epoch": 20.48, + "learning_rate": 1.5890375335895287e-05, + "loss": 0.9419, + "step": 236600 + }, + { + "epoch": 20.48, + "learning_rate": 1.5888930624981944e-05, + "loss": 0.9055, + "step": 236610 + }, + { + "epoch": 20.48, + "learning_rate": 1.5887485914068594e-05, + "loss": 0.9939, + "step": 236620 + }, + { + "epoch": 20.48, + "learning_rate": 1.588604120315525e-05, + "loss": 0.9597, + "step": 236630 + }, + { + "epoch": 20.48, + "learning_rate": 1.5884596492241904e-05, + "loss": 1.0456, + "step": 236640 + }, + { + "epoch": 20.48, + "learning_rate": 1.5883151781328557e-05, + "loss": 0.9463, + "step": 236650 + }, + { + "epoch": 20.48, + "learning_rate": 1.588170707041521e-05, + "loss": 0.9374, + "step": 236660 + }, + { + "epoch": 20.49, + "learning_rate": 1.5880262359501864e-05, + "loss": 0.9536, + "step": 236670 + }, + { + "epoch": 20.49, + "learning_rate": 1.5878817648588518e-05, + "loss": 0.9336, + "step": 236680 + }, + { + "epoch": 20.49, + "learning_rate": 1.587737293767517e-05, + "loss": 0.9522, + "step": 236690 + }, + { + "epoch": 20.49, + "learning_rate": 1.5875928226761828e-05, + "loss": 0.9742, + "step": 236700 + }, + { + "epoch": 20.49, + "learning_rate": 1.5874483515848478e-05, + "loss": 0.8746, + "step": 236710 + }, + { + "epoch": 20.49, + "learning_rate": 1.5873038804935135e-05, + "loss": 0.8699, + "step": 236720 + }, + { + "epoch": 20.49, + "learning_rate": 1.5871594094021788e-05, + "loss": 0.9773, + "step": 236730 + }, + { + "epoch": 20.49, + "learning_rate": 1.587014938310844e-05, + "loss": 0.9376, + "step": 236740 + }, + { + "epoch": 20.49, + "learning_rate": 1.5868704672195095e-05, + "loss": 0.9445, + "step": 236750 + }, + { + "epoch": 20.49, + "learning_rate": 1.5867259961281748e-05, + "loss": 0.9933, + "step": 236760 + }, + { + "epoch": 20.49, + "learning_rate": 1.58658152503684e-05, + "loss": 0.9545, + "step": 236770 + }, + { + "epoch": 20.5, + "learning_rate": 1.5864370539455055e-05, + "loss": 0.9619, + "step": 236780 + }, + { + "epoch": 20.5, + "learning_rate": 1.5862925828541712e-05, + "loss": 0.958, + "step": 236790 + }, + { + "epoch": 20.5, + "learning_rate": 1.5861481117628362e-05, + "loss": 0.9629, + "step": 236800 + }, + { + "epoch": 20.5, + "learning_rate": 1.586003640671502e-05, + "loss": 0.9534, + "step": 236810 + }, + { + "epoch": 20.5, + "learning_rate": 1.585859169580167e-05, + "loss": 0.9461, + "step": 236820 + }, + { + "epoch": 20.5, + "learning_rate": 1.5857146984888325e-05, + "loss": 0.9962, + "step": 236830 + }, + { + "epoch": 20.5, + "learning_rate": 1.585570227397498e-05, + "loss": 0.9444, + "step": 236840 + }, + { + "epoch": 20.5, + "learning_rate": 1.5854257563061632e-05, + "loss": 0.968, + "step": 236850 + }, + { + "epoch": 20.5, + "learning_rate": 1.5852812852148286e-05, + "loss": 0.9406, + "step": 236860 + }, + { + "epoch": 20.5, + "learning_rate": 1.585136814123494e-05, + "loss": 0.964, + "step": 236870 + }, + { + "epoch": 20.5, + "learning_rate": 1.5849923430321592e-05, + "loss": 0.9624, + "step": 236880 + }, + { + "epoch": 20.5, + "learning_rate": 1.5848478719408246e-05, + "loss": 0.9576, + "step": 236890 + }, + { + "epoch": 20.51, + "learning_rate": 1.5847034008494902e-05, + "loss": 0.9194, + "step": 236900 + }, + { + "epoch": 20.51, + "learning_rate": 1.5845589297581552e-05, + "loss": 0.9533, + "step": 236910 + }, + { + "epoch": 20.51, + "learning_rate": 1.584414458666821e-05, + "loss": 0.9927, + "step": 236920 + }, + { + "epoch": 20.51, + "learning_rate": 1.5842699875754863e-05, + "loss": 0.9093, + "step": 236930 + }, + { + "epoch": 20.51, + "learning_rate": 1.5841255164841516e-05, + "loss": 0.9569, + "step": 236940 + }, + { + "epoch": 20.51, + "learning_rate": 1.583981045392817e-05, + "loss": 0.9608, + "step": 236950 + }, + { + "epoch": 20.51, + "learning_rate": 1.5838365743014823e-05, + "loss": 0.8912, + "step": 236960 + }, + { + "epoch": 20.51, + "learning_rate": 1.5836921032101476e-05, + "loss": 0.9711, + "step": 236970 + }, + { + "epoch": 20.51, + "learning_rate": 1.583547632118813e-05, + "loss": 0.956, + "step": 236980 + }, + { + "epoch": 20.51, + "learning_rate": 1.5834031610274786e-05, + "loss": 0.9893, + "step": 236990 + }, + { + "epoch": 20.51, + "learning_rate": 1.5832586899361436e-05, + "loss": 0.951, + "step": 237000 + }, + { + "epoch": 20.51, + "learning_rate": 1.5831142188448093e-05, + "loss": 0.9573, + "step": 237010 + }, + { + "epoch": 20.52, + "learning_rate": 1.5829697477534747e-05, + "loss": 0.9609, + "step": 237020 + }, + { + "epoch": 20.52, + "learning_rate": 1.58282527666214e-05, + "loss": 0.9835, + "step": 237030 + }, + { + "epoch": 20.52, + "learning_rate": 1.5826808055708053e-05, + "loss": 0.9285, + "step": 237040 + }, + { + "epoch": 20.52, + "learning_rate": 1.5825363344794707e-05, + "loss": 0.9231, + "step": 237050 + }, + { + "epoch": 20.52, + "learning_rate": 1.582391863388136e-05, + "loss": 0.9914, + "step": 237060 + }, + { + "epoch": 20.52, + "learning_rate": 1.5822473922968014e-05, + "loss": 0.9558, + "step": 237070 + }, + { + "epoch": 20.52, + "learning_rate": 1.582102921205467e-05, + "loss": 0.9336, + "step": 237080 + }, + { + "epoch": 20.52, + "learning_rate": 1.581958450114132e-05, + "loss": 0.9265, + "step": 237090 + }, + { + "epoch": 20.52, + "learning_rate": 1.5818139790227977e-05, + "loss": 0.9301, + "step": 237100 + }, + { + "epoch": 20.52, + "learning_rate": 1.581669507931463e-05, + "loss": 1.0083, + "step": 237110 + }, + { + "epoch": 20.52, + "learning_rate": 1.5815250368401284e-05, + "loss": 0.9819, + "step": 237120 + }, + { + "epoch": 20.53, + "learning_rate": 1.5813805657487937e-05, + "loss": 0.9625, + "step": 237130 + }, + { + "epoch": 20.53, + "learning_rate": 1.581236094657459e-05, + "loss": 0.9517, + "step": 237140 + }, + { + "epoch": 20.53, + "learning_rate": 1.5810916235661244e-05, + "loss": 0.9303, + "step": 237150 + }, + { + "epoch": 20.53, + "learning_rate": 1.5809471524747898e-05, + "loss": 0.9809, + "step": 237160 + }, + { + "epoch": 20.53, + "learning_rate": 1.5808026813834554e-05, + "loss": 0.9672, + "step": 237170 + }, + { + "epoch": 20.53, + "learning_rate": 1.5806582102921204e-05, + "loss": 0.9461, + "step": 237180 + }, + { + "epoch": 20.53, + "learning_rate": 1.580513739200786e-05, + "loss": 0.9274, + "step": 237190 + }, + { + "epoch": 20.53, + "learning_rate": 1.5803692681094514e-05, + "loss": 0.9214, + "step": 237200 + }, + { + "epoch": 20.53, + "learning_rate": 1.5802247970181168e-05, + "loss": 0.9767, + "step": 237210 + }, + { + "epoch": 20.53, + "learning_rate": 1.580080325926782e-05, + "loss": 0.9807, + "step": 237220 + }, + { + "epoch": 20.53, + "learning_rate": 1.5799358548354475e-05, + "loss": 0.9683, + "step": 237230 + }, + { + "epoch": 20.53, + "learning_rate": 1.5797913837441128e-05, + "loss": 0.9599, + "step": 237240 + }, + { + "epoch": 20.54, + "learning_rate": 1.579646912652778e-05, + "loss": 0.9942, + "step": 237250 + }, + { + "epoch": 20.54, + "learning_rate": 1.5795024415614438e-05, + "loss": 0.937, + "step": 237260 + }, + { + "epoch": 20.54, + "learning_rate": 1.5793579704701088e-05, + "loss": 0.9507, + "step": 237270 + }, + { + "epoch": 20.54, + "learning_rate": 1.5792134993787745e-05, + "loss": 0.9234, + "step": 237280 + }, + { + "epoch": 20.54, + "learning_rate": 1.57906902828744e-05, + "loss": 0.9912, + "step": 237290 + }, + { + "epoch": 20.54, + "learning_rate": 1.5789245571961052e-05, + "loss": 0.9707, + "step": 237300 + }, + { + "epoch": 20.54, + "learning_rate": 1.5787800861047705e-05, + "loss": 0.9333, + "step": 237310 + }, + { + "epoch": 20.54, + "learning_rate": 1.578635615013436e-05, + "loss": 0.9534, + "step": 237320 + }, + { + "epoch": 20.54, + "learning_rate": 1.5784911439221012e-05, + "loss": 0.9681, + "step": 237330 + }, + { + "epoch": 20.54, + "learning_rate": 1.5783466728307665e-05, + "loss": 0.9217, + "step": 237340 + }, + { + "epoch": 20.54, + "learning_rate": 1.5782022017394322e-05, + "loss": 0.9213, + "step": 237350 + }, + { + "epoch": 20.55, + "learning_rate": 1.5780577306480972e-05, + "loss": 0.966, + "step": 237360 + }, + { + "epoch": 20.55, + "learning_rate": 1.577913259556763e-05, + "loss": 0.9622, + "step": 237370 + }, + { + "epoch": 20.55, + "learning_rate": 1.5777687884654282e-05, + "loss": 0.9343, + "step": 237380 + }, + { + "epoch": 20.55, + "learning_rate": 1.5776243173740936e-05, + "loss": 0.9746, + "step": 237390 + }, + { + "epoch": 20.55, + "learning_rate": 1.577479846282759e-05, + "loss": 0.9499, + "step": 237400 + }, + { + "epoch": 20.55, + "learning_rate": 1.5773353751914243e-05, + "loss": 0.9562, + "step": 237410 + }, + { + "epoch": 20.55, + "learning_rate": 1.5771909041000896e-05, + "loss": 0.9476, + "step": 237420 + }, + { + "epoch": 20.55, + "learning_rate": 1.577046433008755e-05, + "loss": 1.0125, + "step": 237430 + }, + { + "epoch": 20.55, + "learning_rate": 1.5769019619174206e-05, + "loss": 0.9341, + "step": 237440 + }, + { + "epoch": 20.55, + "learning_rate": 1.5767574908260856e-05, + "loss": 0.9029, + "step": 237450 + }, + { + "epoch": 20.55, + "learning_rate": 1.5766130197347513e-05, + "loss": 0.9205, + "step": 237460 + }, + { + "epoch": 20.55, + "learning_rate": 1.5764685486434166e-05, + "loss": 0.925, + "step": 237470 + }, + { + "epoch": 20.56, + "learning_rate": 1.576324077552082e-05, + "loss": 0.9116, + "step": 237480 + }, + { + "epoch": 20.56, + "learning_rate": 1.5761796064607473e-05, + "loss": 0.9723, + "step": 237490 + }, + { + "epoch": 20.56, + "learning_rate": 1.5760351353694127e-05, + "loss": 0.9462, + "step": 237500 + }, + { + "epoch": 20.56, + "learning_rate": 1.575890664278078e-05, + "loss": 0.932, + "step": 237510 + }, + { + "epoch": 20.56, + "learning_rate": 1.5757461931867433e-05, + "loss": 0.9124, + "step": 237520 + }, + { + "epoch": 20.56, + "learning_rate": 1.575601722095409e-05, + "loss": 1.0083, + "step": 237530 + }, + { + "epoch": 20.56, + "learning_rate": 1.575457251004074e-05, + "loss": 0.926, + "step": 237540 + }, + { + "epoch": 20.56, + "learning_rate": 1.5753127799127397e-05, + "loss": 0.9752, + "step": 237550 + }, + { + "epoch": 20.56, + "learning_rate": 1.575168308821405e-05, + "loss": 0.975, + "step": 237560 + }, + { + "epoch": 20.56, + "learning_rate": 1.5750238377300704e-05, + "loss": 0.9221, + "step": 237570 + }, + { + "epoch": 20.56, + "learning_rate": 1.5748793666387357e-05, + "loss": 0.912, + "step": 237580 + }, + { + "epoch": 20.57, + "learning_rate": 1.574734895547401e-05, + "loss": 1.0387, + "step": 237590 + }, + { + "epoch": 20.57, + "learning_rate": 1.5745904244560664e-05, + "loss": 0.9557, + "step": 237600 + }, + { + "epoch": 20.57, + "learning_rate": 1.5744459533647317e-05, + "loss": 0.8899, + "step": 237610 + }, + { + "epoch": 20.57, + "learning_rate": 1.5743014822733974e-05, + "loss": 0.9271, + "step": 237620 + }, + { + "epoch": 20.57, + "learning_rate": 1.5741570111820624e-05, + "loss": 0.9428, + "step": 237630 + }, + { + "epoch": 20.57, + "learning_rate": 1.574012540090728e-05, + "loss": 0.8968, + "step": 237640 + }, + { + "epoch": 20.57, + "learning_rate": 1.5738680689993934e-05, + "loss": 0.8832, + "step": 237650 + }, + { + "epoch": 20.57, + "learning_rate": 1.5737235979080588e-05, + "loss": 0.999, + "step": 237660 + }, + { + "epoch": 20.57, + "learning_rate": 1.573579126816724e-05, + "loss": 0.9691, + "step": 237670 + }, + { + "epoch": 20.57, + "learning_rate": 1.5734346557253894e-05, + "loss": 0.9106, + "step": 237680 + }, + { + "epoch": 20.57, + "learning_rate": 1.5732901846340548e-05, + "loss": 0.9225, + "step": 237690 + }, + { + "epoch": 20.57, + "learning_rate": 1.57314571354272e-05, + "loss": 0.9086, + "step": 237700 + }, + { + "epoch": 20.58, + "learning_rate": 1.5730012424513855e-05, + "loss": 1.0081, + "step": 237710 + }, + { + "epoch": 20.58, + "learning_rate": 1.5728567713600508e-05, + "loss": 0.9599, + "step": 237720 + }, + { + "epoch": 20.58, + "learning_rate": 1.5727123002687165e-05, + "loss": 0.9269, + "step": 237730 + }, + { + "epoch": 20.58, + "learning_rate": 1.5725678291773815e-05, + "loss": 0.9681, + "step": 237740 + }, + { + "epoch": 20.58, + "learning_rate": 1.572423358086047e-05, + "loss": 0.9148, + "step": 237750 + }, + { + "epoch": 20.58, + "learning_rate": 1.5722788869947125e-05, + "loss": 0.9315, + "step": 237760 + }, + { + "epoch": 20.58, + "learning_rate": 1.572134415903378e-05, + "loss": 0.9602, + "step": 237770 + }, + { + "epoch": 20.58, + "learning_rate": 1.5719899448120432e-05, + "loss": 0.899, + "step": 237780 + }, + { + "epoch": 20.58, + "learning_rate": 1.5718454737207085e-05, + "loss": 0.9425, + "step": 237790 + }, + { + "epoch": 20.58, + "learning_rate": 1.571701002629374e-05, + "loss": 1.0101, + "step": 237800 + }, + { + "epoch": 20.58, + "learning_rate": 1.5715565315380392e-05, + "loss": 0.9718, + "step": 237810 + }, + { + "epoch": 20.59, + "learning_rate": 1.571412060446705e-05, + "loss": 0.9555, + "step": 237820 + }, + { + "epoch": 20.59, + "learning_rate": 1.57126758935537e-05, + "loss": 0.9481, + "step": 237830 + }, + { + "epoch": 20.59, + "learning_rate": 1.5711231182640355e-05, + "loss": 0.9675, + "step": 237840 + }, + { + "epoch": 20.59, + "learning_rate": 1.570978647172701e-05, + "loss": 0.9333, + "step": 237850 + }, + { + "epoch": 20.59, + "learning_rate": 1.5708341760813662e-05, + "loss": 0.9292, + "step": 237860 + }, + { + "epoch": 20.59, + "learning_rate": 1.5706897049900316e-05, + "loss": 0.8781, + "step": 237870 + }, + { + "epoch": 20.59, + "learning_rate": 1.570545233898697e-05, + "loss": 0.9516, + "step": 237880 + }, + { + "epoch": 20.59, + "learning_rate": 1.5704007628073622e-05, + "loss": 0.9704, + "step": 237890 + }, + { + "epoch": 20.59, + "learning_rate": 1.5702562917160276e-05, + "loss": 0.9494, + "step": 237900 + }, + { + "epoch": 20.59, + "learning_rate": 1.5701118206246933e-05, + "loss": 0.9804, + "step": 237910 + }, + { + "epoch": 20.59, + "learning_rate": 1.5699673495333583e-05, + "loss": 0.975, + "step": 237920 + }, + { + "epoch": 20.59, + "learning_rate": 1.569822878442024e-05, + "loss": 1.0315, + "step": 237930 + }, + { + "epoch": 20.6, + "learning_rate": 1.5696784073506893e-05, + "loss": 0.9246, + "step": 237940 + }, + { + "epoch": 20.6, + "learning_rate": 1.5695339362593546e-05, + "loss": 0.9242, + "step": 237950 + }, + { + "epoch": 20.6, + "learning_rate": 1.56938946516802e-05, + "loss": 0.9076, + "step": 237960 + }, + { + "epoch": 20.6, + "learning_rate": 1.5692449940766853e-05, + "loss": 0.9876, + "step": 237970 + }, + { + "epoch": 20.6, + "learning_rate": 1.5691005229853506e-05, + "loss": 0.9859, + "step": 237980 + }, + { + "epoch": 20.6, + "learning_rate": 1.568956051894016e-05, + "loss": 0.9428, + "step": 237990 + }, + { + "epoch": 20.6, + "learning_rate": 1.5688115808026817e-05, + "loss": 0.9298, + "step": 238000 + }, + { + "epoch": 20.6, + "learning_rate": 1.5686671097113467e-05, + "loss": 0.9718, + "step": 238010 + }, + { + "epoch": 20.6, + "learning_rate": 1.5685226386200123e-05, + "loss": 0.9562, + "step": 238020 + }, + { + "epoch": 20.6, + "learning_rate": 1.5683781675286777e-05, + "loss": 0.9234, + "step": 238030 + }, + { + "epoch": 20.6, + "learning_rate": 1.568233696437343e-05, + "loss": 0.991, + "step": 238040 + }, + { + "epoch": 20.61, + "learning_rate": 1.5680892253460084e-05, + "loss": 0.9259, + "step": 238050 + }, + { + "epoch": 20.61, + "learning_rate": 1.5679447542546737e-05, + "loss": 0.9495, + "step": 238060 + }, + { + "epoch": 20.61, + "learning_rate": 1.567800283163339e-05, + "loss": 0.904, + "step": 238070 + }, + { + "epoch": 20.61, + "learning_rate": 1.5676558120720044e-05, + "loss": 0.9716, + "step": 238080 + }, + { + "epoch": 20.61, + "learning_rate": 1.56751134098067e-05, + "loss": 0.9301, + "step": 238090 + }, + { + "epoch": 20.61, + "learning_rate": 1.567366869889335e-05, + "loss": 0.929, + "step": 238100 + }, + { + "epoch": 20.61, + "learning_rate": 1.5672223987980007e-05, + "loss": 0.9072, + "step": 238110 + }, + { + "epoch": 20.61, + "learning_rate": 1.567077927706666e-05, + "loss": 0.9554, + "step": 238120 + }, + { + "epoch": 20.61, + "learning_rate": 1.5669334566153314e-05, + "loss": 0.8513, + "step": 238130 + }, + { + "epoch": 20.61, + "learning_rate": 1.5667889855239967e-05, + "loss": 0.9694, + "step": 238140 + }, + { + "epoch": 20.61, + "learning_rate": 1.566644514432662e-05, + "loss": 0.9759, + "step": 238150 + }, + { + "epoch": 20.61, + "learning_rate": 1.5665000433413274e-05, + "loss": 0.9533, + "step": 238160 + }, + { + "epoch": 20.62, + "learning_rate": 1.5663555722499928e-05, + "loss": 0.9541, + "step": 238170 + }, + { + "epoch": 20.62, + "learning_rate": 1.5662111011586584e-05, + "loss": 0.9505, + "step": 238180 + }, + { + "epoch": 20.62, + "learning_rate": 1.5660666300673234e-05, + "loss": 0.9311, + "step": 238190 + }, + { + "epoch": 20.62, + "learning_rate": 1.565922158975989e-05, + "loss": 0.9114, + "step": 238200 + }, + { + "epoch": 20.62, + "learning_rate": 1.5657776878846545e-05, + "loss": 0.9307, + "step": 238210 + }, + { + "epoch": 20.62, + "learning_rate": 1.5656332167933198e-05, + "loss": 0.9281, + "step": 238220 + }, + { + "epoch": 20.62, + "learning_rate": 1.565488745701985e-05, + "loss": 0.934, + "step": 238230 + }, + { + "epoch": 20.62, + "learning_rate": 1.5653442746106505e-05, + "loss": 0.964, + "step": 238240 + }, + { + "epoch": 20.62, + "learning_rate": 1.5651998035193158e-05, + "loss": 0.9475, + "step": 238250 + }, + { + "epoch": 20.62, + "learning_rate": 1.565055332427981e-05, + "loss": 0.9378, + "step": 238260 + }, + { + "epoch": 20.62, + "learning_rate": 1.564910861336647e-05, + "loss": 0.9821, + "step": 238270 + }, + { + "epoch": 20.62, + "learning_rate": 1.564766390245312e-05, + "loss": 0.983, + "step": 238280 + }, + { + "epoch": 20.63, + "learning_rate": 1.5646219191539775e-05, + "loss": 0.9172, + "step": 238290 + }, + { + "epoch": 20.63, + "learning_rate": 1.564477448062643e-05, + "loss": 0.9878, + "step": 238300 + }, + { + "epoch": 20.63, + "learning_rate": 1.5643329769713082e-05, + "loss": 0.9765, + "step": 238310 + }, + { + "epoch": 20.63, + "learning_rate": 1.5641885058799735e-05, + "loss": 0.8836, + "step": 238320 + }, + { + "epoch": 20.63, + "learning_rate": 1.564044034788639e-05, + "loss": 1.0143, + "step": 238330 + }, + { + "epoch": 20.63, + "learning_rate": 1.5638995636973042e-05, + "loss": 0.9953, + "step": 238340 + }, + { + "epoch": 20.63, + "learning_rate": 1.5637550926059696e-05, + "loss": 0.967, + "step": 238350 + }, + { + "epoch": 20.63, + "learning_rate": 1.5636106215146352e-05, + "loss": 0.9521, + "step": 238360 + }, + { + "epoch": 20.63, + "learning_rate": 1.5634661504233002e-05, + "loss": 0.9084, + "step": 238370 + }, + { + "epoch": 20.63, + "learning_rate": 1.563321679331966e-05, + "loss": 0.9342, + "step": 238380 + }, + { + "epoch": 20.63, + "learning_rate": 1.5631772082406313e-05, + "loss": 0.9451, + "step": 238390 + }, + { + "epoch": 20.64, + "learning_rate": 1.5630327371492966e-05, + "loss": 0.9952, + "step": 238400 + }, + { + "epoch": 20.64, + "learning_rate": 1.562888266057962e-05, + "loss": 0.9731, + "step": 238410 + }, + { + "epoch": 20.64, + "learning_rate": 1.5627437949666273e-05, + "loss": 0.9714, + "step": 238420 + }, + { + "epoch": 20.64, + "learning_rate": 1.5625993238752926e-05, + "loss": 0.9573, + "step": 238430 + }, + { + "epoch": 20.64, + "learning_rate": 1.562454852783958e-05, + "loss": 0.9845, + "step": 238440 + }, + { + "epoch": 20.64, + "learning_rate": 1.5623103816926236e-05, + "loss": 0.8772, + "step": 238450 + }, + { + "epoch": 20.64, + "learning_rate": 1.5621659106012886e-05, + "loss": 0.9313, + "step": 238460 + }, + { + "epoch": 20.64, + "learning_rate": 1.5620214395099543e-05, + "loss": 0.8696, + "step": 238470 + }, + { + "epoch": 20.64, + "learning_rate": 1.5618769684186196e-05, + "loss": 0.9507, + "step": 238480 + }, + { + "epoch": 20.64, + "learning_rate": 1.561732497327285e-05, + "loss": 0.923, + "step": 238490 + }, + { + "epoch": 20.64, + "learning_rate": 1.5615880262359503e-05, + "loss": 0.8825, + "step": 238500 + }, + { + "epoch": 20.64, + "learning_rate": 1.5614435551446157e-05, + "loss": 0.9621, + "step": 238510 + }, + { + "epoch": 20.65, + "learning_rate": 1.561299084053281e-05, + "loss": 0.9217, + "step": 238520 + }, + { + "epoch": 20.65, + "learning_rate": 1.5611546129619463e-05, + "loss": 0.9321, + "step": 238530 + }, + { + "epoch": 20.65, + "learning_rate": 1.561010141870612e-05, + "loss": 0.8862, + "step": 238540 + }, + { + "epoch": 20.65, + "learning_rate": 1.560865670779277e-05, + "loss": 0.9515, + "step": 238550 + }, + { + "epoch": 20.65, + "learning_rate": 1.5607211996879427e-05, + "loss": 0.9165, + "step": 238560 + }, + { + "epoch": 20.65, + "learning_rate": 1.560576728596608e-05, + "loss": 0.9537, + "step": 238570 + }, + { + "epoch": 20.65, + "learning_rate": 1.5604322575052734e-05, + "loss": 0.913, + "step": 238580 + }, + { + "epoch": 20.65, + "learning_rate": 1.5602877864139387e-05, + "loss": 0.9712, + "step": 238590 + }, + { + "epoch": 20.65, + "learning_rate": 1.560143315322604e-05, + "loss": 0.988, + "step": 238600 + }, + { + "epoch": 20.65, + "learning_rate": 1.5599988442312694e-05, + "loss": 0.9796, + "step": 238610 + }, + { + "epoch": 20.65, + "learning_rate": 1.5598543731399347e-05, + "loss": 0.9717, + "step": 238620 + }, + { + "epoch": 20.66, + "learning_rate": 1.5597099020486e-05, + "loss": 0.9564, + "step": 238630 + }, + { + "epoch": 20.66, + "learning_rate": 1.5595654309572654e-05, + "loss": 0.9633, + "step": 238640 + }, + { + "epoch": 20.66, + "learning_rate": 1.559420959865931e-05, + "loss": 0.9484, + "step": 238650 + }, + { + "epoch": 20.66, + "learning_rate": 1.559276488774596e-05, + "loss": 0.9286, + "step": 238660 + }, + { + "epoch": 20.66, + "learning_rate": 1.5591320176832618e-05, + "loss": 0.9359, + "step": 238670 + }, + { + "epoch": 20.66, + "learning_rate": 1.558987546591927e-05, + "loss": 0.9451, + "step": 238680 + }, + { + "epoch": 20.66, + "learning_rate": 1.558843075500592e-05, + "loss": 0.9499, + "step": 238690 + }, + { + "epoch": 20.66, + "learning_rate": 1.5586986044092578e-05, + "loss": 0.9072, + "step": 238700 + }, + { + "epoch": 20.66, + "learning_rate": 1.558554133317923e-05, + "loss": 0.9059, + "step": 238710 + }, + { + "epoch": 20.66, + "learning_rate": 1.5584096622265885e-05, + "loss": 0.943, + "step": 238720 + }, + { + "epoch": 20.66, + "learning_rate": 1.5582651911352538e-05, + "loss": 0.9316, + "step": 238730 + }, + { + "epoch": 20.66, + "learning_rate": 1.5581207200439195e-05, + "loss": 0.9454, + "step": 238740 + }, + { + "epoch": 20.67, + "learning_rate": 1.5579762489525845e-05, + "loss": 0.9432, + "step": 238750 + }, + { + "epoch": 20.67, + "learning_rate": 1.5578317778612502e-05, + "loss": 0.9659, + "step": 238760 + }, + { + "epoch": 20.67, + "learning_rate": 1.5576873067699155e-05, + "loss": 0.9338, + "step": 238770 + }, + { + "epoch": 20.67, + "learning_rate": 1.5575428356785805e-05, + "loss": 0.9342, + "step": 238780 + }, + { + "epoch": 20.67, + "learning_rate": 1.5573983645872462e-05, + "loss": 0.9554, + "step": 238790 + }, + { + "epoch": 20.67, + "learning_rate": 1.5572538934959115e-05, + "loss": 0.9275, + "step": 238800 + }, + { + "epoch": 20.67, + "learning_rate": 1.557109422404577e-05, + "loss": 0.9389, + "step": 238810 + }, + { + "epoch": 20.67, + "learning_rate": 1.5569649513132422e-05, + "loss": 0.9571, + "step": 238820 + }, + { + "epoch": 20.67, + "learning_rate": 1.556820480221908e-05, + "loss": 0.9567, + "step": 238830 + }, + { + "epoch": 20.67, + "learning_rate": 1.556676009130573e-05, + "loss": 0.9651, + "step": 238840 + }, + { + "epoch": 20.67, + "learning_rate": 1.5565315380392386e-05, + "loss": 1.0345, + "step": 238850 + }, + { + "epoch": 20.68, + "learning_rate": 1.556387066947904e-05, + "loss": 0.9242, + "step": 238860 + }, + { + "epoch": 20.68, + "learning_rate": 1.556242595856569e-05, + "loss": 0.9306, + "step": 238870 + }, + { + "epoch": 20.68, + "learning_rate": 1.5560981247652346e-05, + "loss": 0.9242, + "step": 238880 + }, + { + "epoch": 20.68, + "learning_rate": 1.5559536536739e-05, + "loss": 0.9435, + "step": 238890 + }, + { + "epoch": 20.68, + "learning_rate": 1.5558091825825653e-05, + "loss": 0.9781, + "step": 238900 + }, + { + "epoch": 20.68, + "learning_rate": 1.5556647114912306e-05, + "loss": 0.9377, + "step": 238910 + }, + { + "epoch": 20.68, + "learning_rate": 1.5555202403998963e-05, + "loss": 0.9745, + "step": 238920 + }, + { + "epoch": 20.68, + "learning_rate": 1.5553757693085613e-05, + "loss": 0.9093, + "step": 238930 + }, + { + "epoch": 20.68, + "learning_rate": 1.555231298217227e-05, + "loss": 0.9616, + "step": 238940 + }, + { + "epoch": 20.68, + "learning_rate": 1.5550868271258923e-05, + "loss": 0.9528, + "step": 238950 + }, + { + "epoch": 20.68, + "learning_rate": 1.5549423560345573e-05, + "loss": 0.9086, + "step": 238960 + }, + { + "epoch": 20.68, + "learning_rate": 1.554797884943223e-05, + "loss": 0.9506, + "step": 238970 + }, + { + "epoch": 20.69, + "learning_rate": 1.5546534138518883e-05, + "loss": 0.9877, + "step": 238980 + }, + { + "epoch": 20.69, + "learning_rate": 1.5545089427605537e-05, + "loss": 0.9305, + "step": 238990 + }, + { + "epoch": 20.69, + "learning_rate": 1.554364471669219e-05, + "loss": 0.9025, + "step": 239000 + }, + { + "epoch": 20.69, + "learning_rate": 1.5542200005778847e-05, + "loss": 0.9585, + "step": 239010 + }, + { + "epoch": 20.69, + "learning_rate": 1.5540755294865497e-05, + "loss": 0.9608, + "step": 239020 + }, + { + "epoch": 20.69, + "learning_rate": 1.5539310583952154e-05, + "loss": 0.9443, + "step": 239030 + }, + { + "epoch": 20.69, + "learning_rate": 1.5537865873038807e-05, + "loss": 0.9116, + "step": 239040 + }, + { + "epoch": 20.69, + "learning_rate": 1.5536421162125457e-05, + "loss": 0.9479, + "step": 239050 + }, + { + "epoch": 20.69, + "learning_rate": 1.5534976451212114e-05, + "loss": 0.9581, + "step": 239060 + }, + { + "epoch": 20.69, + "learning_rate": 1.5533531740298767e-05, + "loss": 0.9213, + "step": 239070 + }, + { + "epoch": 20.69, + "learning_rate": 1.553208702938542e-05, + "loss": 0.9084, + "step": 239080 + }, + { + "epoch": 20.7, + "learning_rate": 1.5530642318472074e-05, + "loss": 0.9952, + "step": 239090 + }, + { + "epoch": 20.7, + "learning_rate": 1.552919760755873e-05, + "loss": 0.9145, + "step": 239100 + }, + { + "epoch": 20.7, + "learning_rate": 1.552775289664538e-05, + "loss": 0.925, + "step": 239110 + }, + { + "epoch": 20.7, + "learning_rate": 1.5526308185732037e-05, + "loss": 0.954, + "step": 239120 + }, + { + "epoch": 20.7, + "learning_rate": 1.552486347481869e-05, + "loss": 0.9689, + "step": 239130 + }, + { + "epoch": 20.7, + "learning_rate": 1.552341876390534e-05, + "loss": 0.9087, + "step": 239140 + }, + { + "epoch": 20.7, + "learning_rate": 1.5521974052991998e-05, + "loss": 0.958, + "step": 239150 + }, + { + "epoch": 20.7, + "learning_rate": 1.552052934207865e-05, + "loss": 0.9555, + "step": 239160 + }, + { + "epoch": 20.7, + "learning_rate": 1.5519084631165304e-05, + "loss": 0.9188, + "step": 239170 + }, + { + "epoch": 20.7, + "learning_rate": 1.5517639920251958e-05, + "loss": 0.9305, + "step": 239180 + }, + { + "epoch": 20.7, + "learning_rate": 1.5516195209338615e-05, + "loss": 0.9784, + "step": 239190 + }, + { + "epoch": 20.7, + "learning_rate": 1.5514750498425265e-05, + "loss": 0.9364, + "step": 239200 + }, + { + "epoch": 20.71, + "learning_rate": 1.551330578751192e-05, + "loss": 0.9929, + "step": 239210 + }, + { + "epoch": 20.71, + "learning_rate": 1.5511861076598575e-05, + "loss": 1.028, + "step": 239220 + }, + { + "epoch": 20.71, + "learning_rate": 1.5510416365685225e-05, + "loss": 0.9495, + "step": 239230 + }, + { + "epoch": 20.71, + "learning_rate": 1.550897165477188e-05, + "loss": 0.9981, + "step": 239240 + }, + { + "epoch": 20.71, + "learning_rate": 1.5507526943858535e-05, + "loss": 0.9448, + "step": 239250 + }, + { + "epoch": 20.71, + "learning_rate": 1.550608223294519e-05, + "loss": 1.0004, + "step": 239260 + }, + { + "epoch": 20.71, + "learning_rate": 1.5504637522031842e-05, + "loss": 0.9776, + "step": 239270 + }, + { + "epoch": 20.71, + "learning_rate": 1.55031928111185e-05, + "loss": 0.9869, + "step": 239280 + }, + { + "epoch": 20.71, + "learning_rate": 1.550174810020515e-05, + "loss": 0.9265, + "step": 239290 + }, + { + "epoch": 20.71, + "learning_rate": 1.5500303389291805e-05, + "loss": 0.9006, + "step": 239300 + }, + { + "epoch": 20.71, + "learning_rate": 1.549885867837846e-05, + "loss": 0.9495, + "step": 239310 + }, + { + "epoch": 20.71, + "learning_rate": 1.549741396746511e-05, + "loss": 0.9784, + "step": 239320 + }, + { + "epoch": 20.72, + "learning_rate": 1.5495969256551766e-05, + "loss": 0.9103, + "step": 239330 + }, + { + "epoch": 20.72, + "learning_rate": 1.549452454563842e-05, + "loss": 0.8854, + "step": 239340 + }, + { + "epoch": 20.72, + "learning_rate": 1.5493079834725072e-05, + "loss": 0.9124, + "step": 239350 + }, + { + "epoch": 20.72, + "learning_rate": 1.5491635123811726e-05, + "loss": 0.9244, + "step": 239360 + }, + { + "epoch": 20.72, + "learning_rate": 1.5490190412898383e-05, + "loss": 1.0098, + "step": 239370 + }, + { + "epoch": 20.72, + "learning_rate": 1.5488745701985033e-05, + "loss": 1.0214, + "step": 239380 + }, + { + "epoch": 20.72, + "learning_rate": 1.548730099107169e-05, + "loss": 0.911, + "step": 239390 + }, + { + "epoch": 20.72, + "learning_rate": 1.5485856280158343e-05, + "loss": 0.9342, + "step": 239400 + }, + { + "epoch": 20.72, + "learning_rate": 1.5484411569244993e-05, + "loss": 0.9609, + "step": 239410 + }, + { + "epoch": 20.72, + "learning_rate": 1.548296685833165e-05, + "loss": 0.9572, + "step": 239420 + }, + { + "epoch": 20.72, + "learning_rate": 1.5481522147418303e-05, + "loss": 0.9106, + "step": 239430 + }, + { + "epoch": 20.73, + "learning_rate": 1.5480077436504956e-05, + "loss": 0.9862, + "step": 239440 + }, + { + "epoch": 20.73, + "learning_rate": 1.547863272559161e-05, + "loss": 0.9642, + "step": 239450 + }, + { + "epoch": 20.73, + "learning_rate": 1.5477188014678266e-05, + "loss": 0.9506, + "step": 239460 + }, + { + "epoch": 20.73, + "learning_rate": 1.5475743303764916e-05, + "loss": 0.9139, + "step": 239470 + }, + { + "epoch": 20.73, + "learning_rate": 1.5474298592851573e-05, + "loss": 0.9351, + "step": 239480 + }, + { + "epoch": 20.73, + "learning_rate": 1.5472853881938227e-05, + "loss": 0.901, + "step": 239490 + }, + { + "epoch": 20.73, + "learning_rate": 1.5471409171024877e-05, + "loss": 0.9947, + "step": 239500 + }, + { + "epoch": 20.73, + "learning_rate": 1.5469964460111533e-05, + "loss": 0.9316, + "step": 239510 + }, + { + "epoch": 20.73, + "learning_rate": 1.5468519749198183e-05, + "loss": 0.9862, + "step": 239520 + }, + { + "epoch": 20.73, + "learning_rate": 1.546707503828484e-05, + "loss": 0.9317, + "step": 239530 + }, + { + "epoch": 20.73, + "learning_rate": 1.5465630327371494e-05, + "loss": 0.9623, + "step": 239540 + }, + { + "epoch": 20.73, + "learning_rate": 1.5464185616458147e-05, + "loss": 0.9842, + "step": 239550 + }, + { + "epoch": 20.74, + "learning_rate": 1.54627409055448e-05, + "loss": 0.907, + "step": 239560 + }, + { + "epoch": 20.74, + "learning_rate": 1.5461296194631457e-05, + "loss": 0.9561, + "step": 239570 + }, + { + "epoch": 20.74, + "learning_rate": 1.5459851483718107e-05, + "loss": 0.9205, + "step": 239580 + }, + { + "epoch": 20.74, + "learning_rate": 1.545840677280476e-05, + "loss": 0.9499, + "step": 239590 + }, + { + "epoch": 20.74, + "learning_rate": 1.5456962061891417e-05, + "loss": 0.9119, + "step": 239600 + }, + { + "epoch": 20.74, + "learning_rate": 1.5455517350978067e-05, + "loss": 1.0085, + "step": 239610 + }, + { + "epoch": 20.74, + "learning_rate": 1.5454072640064724e-05, + "loss": 0.9608, + "step": 239620 + }, + { + "epoch": 20.74, + "learning_rate": 1.5452627929151378e-05, + "loss": 0.9616, + "step": 239630 + }, + { + "epoch": 20.74, + "learning_rate": 1.545118321823803e-05, + "loss": 0.8842, + "step": 239640 + }, + { + "epoch": 20.74, + "learning_rate": 1.5449738507324684e-05, + "loss": 0.9571, + "step": 239650 + }, + { + "epoch": 20.74, + "learning_rate": 1.544829379641134e-05, + "loss": 0.9498, + "step": 239660 + }, + { + "epoch": 20.75, + "learning_rate": 1.544684908549799e-05, + "loss": 0.9339, + "step": 239670 + }, + { + "epoch": 20.75, + "learning_rate": 1.5445404374584645e-05, + "loss": 0.9053, + "step": 239680 + }, + { + "epoch": 20.75, + "learning_rate": 1.54439596636713e-05, + "loss": 0.9859, + "step": 239690 + }, + { + "epoch": 20.75, + "learning_rate": 1.544251495275795e-05, + "loss": 0.975, + "step": 239700 + }, + { + "epoch": 20.75, + "learning_rate": 1.5441070241844608e-05, + "loss": 0.9305, + "step": 239710 + }, + { + "epoch": 20.75, + "learning_rate": 1.543962553093126e-05, + "loss": 0.9075, + "step": 239720 + }, + { + "epoch": 20.75, + "learning_rate": 1.5438180820017915e-05, + "loss": 1.0027, + "step": 239730 + }, + { + "epoch": 20.75, + "learning_rate": 1.5436736109104568e-05, + "loss": 1.008, + "step": 239740 + }, + { + "epoch": 20.75, + "learning_rate": 1.5435291398191225e-05, + "loss": 0.9969, + "step": 239750 + }, + { + "epoch": 20.75, + "learning_rate": 1.5433846687277875e-05, + "loss": 0.9414, + "step": 239760 + }, + { + "epoch": 20.75, + "learning_rate": 1.543240197636453e-05, + "loss": 0.9653, + "step": 239770 + }, + { + "epoch": 20.75, + "learning_rate": 1.5430957265451185e-05, + "loss": 0.936, + "step": 239780 + }, + { + "epoch": 20.76, + "learning_rate": 1.5429512554537835e-05, + "loss": 0.9567, + "step": 239790 + }, + { + "epoch": 20.76, + "learning_rate": 1.5428067843624492e-05, + "loss": 0.9643, + "step": 239800 + }, + { + "epoch": 20.76, + "learning_rate": 1.5426623132711145e-05, + "loss": 0.9404, + "step": 239810 + }, + { + "epoch": 20.76, + "learning_rate": 1.54251784217978e-05, + "loss": 0.903, + "step": 239820 + }, + { + "epoch": 20.76, + "learning_rate": 1.5423733710884452e-05, + "loss": 0.9518, + "step": 239830 + }, + { + "epoch": 20.76, + "learning_rate": 1.5422288999971106e-05, + "loss": 0.9784, + "step": 239840 + }, + { + "epoch": 20.76, + "learning_rate": 1.542084428905776e-05, + "loss": 0.9494, + "step": 239850 + }, + { + "epoch": 20.76, + "learning_rate": 1.5419399578144412e-05, + "loss": 0.9236, + "step": 239860 + }, + { + "epoch": 20.76, + "learning_rate": 1.541795486723107e-05, + "loss": 0.9015, + "step": 239870 + }, + { + "epoch": 20.76, + "learning_rate": 1.541651015631772e-05, + "loss": 0.9856, + "step": 239880 + }, + { + "epoch": 20.76, + "learning_rate": 1.5415065445404376e-05, + "loss": 0.9242, + "step": 239890 + }, + { + "epoch": 20.77, + "learning_rate": 1.541362073449103e-05, + "loss": 0.9759, + "step": 239900 + }, + { + "epoch": 20.77, + "learning_rate": 1.5412176023577683e-05, + "loss": 0.9358, + "step": 239910 + }, + { + "epoch": 20.77, + "learning_rate": 1.5410731312664336e-05, + "loss": 0.9679, + "step": 239920 + }, + { + "epoch": 20.77, + "learning_rate": 1.540928660175099e-05, + "loss": 0.9583, + "step": 239930 + }, + { + "epoch": 20.77, + "learning_rate": 1.5407841890837643e-05, + "loss": 0.9765, + "step": 239940 + }, + { + "epoch": 20.77, + "learning_rate": 1.5406397179924296e-05, + "loss": 0.9339, + "step": 239950 + }, + { + "epoch": 20.77, + "learning_rate": 1.5404952469010953e-05, + "loss": 0.95, + "step": 239960 + }, + { + "epoch": 20.77, + "learning_rate": 1.5403507758097603e-05, + "loss": 0.9602, + "step": 239970 + }, + { + "epoch": 20.77, + "learning_rate": 1.540206304718426e-05, + "loss": 0.9276, + "step": 239980 + }, + { + "epoch": 20.77, + "learning_rate": 1.5400618336270913e-05, + "loss": 1.0016, + "step": 239990 + }, + { + "epoch": 20.77, + "learning_rate": 1.5399173625357567e-05, + "loss": 0.9177, + "step": 240000 + }, + { + "epoch": 20.77, + "learning_rate": 1.539772891444422e-05, + "loss": 0.9192, + "step": 240010 + }, + { + "epoch": 20.78, + "learning_rate": 1.5396284203530874e-05, + "loss": 0.9594, + "step": 240020 + }, + { + "epoch": 20.78, + "learning_rate": 1.5394839492617527e-05, + "loss": 0.9544, + "step": 240030 + }, + { + "epoch": 20.78, + "learning_rate": 1.539339478170418e-05, + "loss": 0.9455, + "step": 240040 + }, + { + "epoch": 20.78, + "learning_rate": 1.5391950070790837e-05, + "loss": 0.9466, + "step": 240050 + }, + { + "epoch": 20.78, + "learning_rate": 1.5390505359877487e-05, + "loss": 0.8778, + "step": 240060 + }, + { + "epoch": 20.78, + "learning_rate": 1.5389060648964144e-05, + "loss": 0.9583, + "step": 240070 + }, + { + "epoch": 20.78, + "learning_rate": 1.5387615938050797e-05, + "loss": 0.9489, + "step": 240080 + }, + { + "epoch": 20.78, + "learning_rate": 1.538617122713745e-05, + "loss": 0.9844, + "step": 240090 + }, + { + "epoch": 20.78, + "learning_rate": 1.5384726516224104e-05, + "loss": 0.9385, + "step": 240100 + }, + { + "epoch": 20.78, + "learning_rate": 1.5383281805310757e-05, + "loss": 0.9325, + "step": 240110 + }, + { + "epoch": 20.78, + "learning_rate": 1.538183709439741e-05, + "loss": 0.9834, + "step": 240120 + }, + { + "epoch": 20.79, + "learning_rate": 1.5380392383484064e-05, + "loss": 0.9404, + "step": 240130 + }, + { + "epoch": 20.79, + "learning_rate": 1.537894767257072e-05, + "loss": 0.9566, + "step": 240140 + }, + { + "epoch": 20.79, + "learning_rate": 1.537750296165737e-05, + "loss": 0.9366, + "step": 240150 + }, + { + "epoch": 20.79, + "learning_rate": 1.5376058250744028e-05, + "loss": 0.8916, + "step": 240160 + }, + { + "epoch": 20.79, + "learning_rate": 1.537461353983068e-05, + "loss": 0.9479, + "step": 240170 + }, + { + "epoch": 20.79, + "learning_rate": 1.5373168828917335e-05, + "loss": 0.9869, + "step": 240180 + }, + { + "epoch": 20.79, + "learning_rate": 1.5371724118003988e-05, + "loss": 0.9662, + "step": 240190 + }, + { + "epoch": 20.79, + "learning_rate": 1.537027940709064e-05, + "loss": 0.9047, + "step": 240200 + }, + { + "epoch": 20.79, + "learning_rate": 1.5368834696177295e-05, + "loss": 0.9245, + "step": 240210 + }, + { + "epoch": 20.79, + "learning_rate": 1.5367389985263948e-05, + "loss": 0.9215, + "step": 240220 + }, + { + "epoch": 20.79, + "learning_rate": 1.5365945274350605e-05, + "loss": 0.8829, + "step": 240230 + }, + { + "epoch": 20.79, + "learning_rate": 1.5364500563437255e-05, + "loss": 0.9398, + "step": 240240 + }, + { + "epoch": 20.8, + "learning_rate": 1.5363055852523912e-05, + "loss": 0.9134, + "step": 240250 + }, + { + "epoch": 20.8, + "learning_rate": 1.5361611141610565e-05, + "loss": 1.0144, + "step": 240260 + }, + { + "epoch": 20.8, + "learning_rate": 1.536016643069722e-05, + "loss": 0.9968, + "step": 240270 + }, + { + "epoch": 20.8, + "learning_rate": 1.5358721719783872e-05, + "loss": 0.9469, + "step": 240280 + }, + { + "epoch": 20.8, + "learning_rate": 1.5357277008870525e-05, + "loss": 0.9138, + "step": 240290 + }, + { + "epoch": 20.8, + "learning_rate": 1.535583229795718e-05, + "loss": 0.9597, + "step": 240300 + }, + { + "epoch": 20.8, + "learning_rate": 1.5354387587043832e-05, + "loss": 0.8924, + "step": 240310 + }, + { + "epoch": 20.8, + "learning_rate": 1.535294287613049e-05, + "loss": 0.9819, + "step": 240320 + }, + { + "epoch": 20.8, + "learning_rate": 1.535149816521714e-05, + "loss": 0.9003, + "step": 240330 + }, + { + "epoch": 20.8, + "learning_rate": 1.5350053454303796e-05, + "loss": 0.8978, + "step": 240340 + }, + { + "epoch": 20.8, + "learning_rate": 1.534860874339045e-05, + "loss": 0.9647, + "step": 240350 + }, + { + "epoch": 20.8, + "learning_rate": 1.5347164032477102e-05, + "loss": 0.9239, + "step": 240360 + }, + { + "epoch": 20.81, + "learning_rate": 1.5345719321563756e-05, + "loss": 0.9278, + "step": 240370 + }, + { + "epoch": 20.81, + "learning_rate": 1.534427461065041e-05, + "loss": 0.9704, + "step": 240380 + }, + { + "epoch": 20.81, + "learning_rate": 1.5342829899737063e-05, + "loss": 0.9344, + "step": 240390 + }, + { + "epoch": 20.81, + "learning_rate": 1.5341385188823716e-05, + "loss": 0.9358, + "step": 240400 + }, + { + "epoch": 20.81, + "learning_rate": 1.5339940477910373e-05, + "loss": 0.9861, + "step": 240410 + }, + { + "epoch": 20.81, + "learning_rate": 1.5338495766997023e-05, + "loss": 0.9749, + "step": 240420 + }, + { + "epoch": 20.81, + "learning_rate": 1.533705105608368e-05, + "loss": 0.9185, + "step": 240430 + }, + { + "epoch": 20.81, + "learning_rate": 1.533560634517033e-05, + "loss": 0.9396, + "step": 240440 + }, + { + "epoch": 20.81, + "learning_rate": 1.5334161634256986e-05, + "loss": 0.9294, + "step": 240450 + }, + { + "epoch": 20.81, + "learning_rate": 1.533271692334364e-05, + "loss": 0.968, + "step": 240460 + }, + { + "epoch": 20.81, + "learning_rate": 1.5331272212430293e-05, + "loss": 0.928, + "step": 240470 + }, + { + "epoch": 20.82, + "learning_rate": 1.5329827501516947e-05, + "loss": 0.928, + "step": 240480 + }, + { + "epoch": 20.82, + "learning_rate": 1.53283827906036e-05, + "loss": 0.9159, + "step": 240490 + }, + { + "epoch": 20.82, + "learning_rate": 1.5326938079690253e-05, + "loss": 0.9783, + "step": 240500 + }, + { + "epoch": 20.82, + "learning_rate": 1.5325493368776907e-05, + "loss": 0.9481, + "step": 240510 + }, + { + "epoch": 20.82, + "learning_rate": 1.5324048657863564e-05, + "loss": 0.8988, + "step": 240520 + }, + { + "epoch": 20.82, + "learning_rate": 1.5322603946950214e-05, + "loss": 0.9338, + "step": 240530 + }, + { + "epoch": 20.82, + "learning_rate": 1.532115923603687e-05, + "loss": 0.903, + "step": 240540 + }, + { + "epoch": 20.82, + "learning_rate": 1.5319714525123524e-05, + "loss": 0.9348, + "step": 240550 + }, + { + "epoch": 20.82, + "learning_rate": 1.5318269814210177e-05, + "loss": 0.9502, + "step": 240560 + }, + { + "epoch": 20.82, + "learning_rate": 1.531682510329683e-05, + "loss": 0.9765, + "step": 240570 + }, + { + "epoch": 20.82, + "learning_rate": 1.5315380392383484e-05, + "loss": 0.9981, + "step": 240580 + }, + { + "epoch": 20.82, + "learning_rate": 1.5313935681470137e-05, + "loss": 0.9192, + "step": 240590 + }, + { + "epoch": 20.83, + "learning_rate": 1.531249097055679e-05, + "loss": 0.9169, + "step": 240600 + }, + { + "epoch": 20.83, + "learning_rate": 1.5311046259643448e-05, + "loss": 0.977, + "step": 240610 + }, + { + "epoch": 20.83, + "learning_rate": 1.5309601548730098e-05, + "loss": 0.9309, + "step": 240620 + }, + { + "epoch": 20.83, + "learning_rate": 1.5308156837816754e-05, + "loss": 0.897, + "step": 240630 + }, + { + "epoch": 20.83, + "learning_rate": 1.5306712126903408e-05, + "loss": 0.9254, + "step": 240640 + }, + { + "epoch": 20.83, + "learning_rate": 1.530526741599006e-05, + "loss": 0.9031, + "step": 240650 + }, + { + "epoch": 20.83, + "learning_rate": 1.5303822705076715e-05, + "loss": 0.9196, + "step": 240660 + }, + { + "epoch": 20.83, + "learning_rate": 1.5302377994163368e-05, + "loss": 0.8742, + "step": 240670 + }, + { + "epoch": 20.83, + "learning_rate": 1.530093328325002e-05, + "loss": 0.9169, + "step": 240680 + }, + { + "epoch": 20.83, + "learning_rate": 1.5299488572336675e-05, + "loss": 0.9853, + "step": 240690 + }, + { + "epoch": 20.83, + "learning_rate": 1.529804386142333e-05, + "loss": 0.9453, + "step": 240700 + }, + { + "epoch": 20.84, + "learning_rate": 1.529659915050998e-05, + "loss": 0.9263, + "step": 240710 + }, + { + "epoch": 20.84, + "learning_rate": 1.5295154439596638e-05, + "loss": 0.932, + "step": 240720 + }, + { + "epoch": 20.84, + "learning_rate": 1.529370972868329e-05, + "loss": 0.9125, + "step": 240730 + }, + { + "epoch": 20.84, + "learning_rate": 1.5292265017769945e-05, + "loss": 0.9352, + "step": 240740 + }, + { + "epoch": 20.84, + "learning_rate": 1.52908203068566e-05, + "loss": 0.9242, + "step": 240750 + }, + { + "epoch": 20.84, + "learning_rate": 1.5289375595943252e-05, + "loss": 0.9465, + "step": 240760 + }, + { + "epoch": 20.84, + "learning_rate": 1.5287930885029905e-05, + "loss": 0.9298, + "step": 240770 + }, + { + "epoch": 20.84, + "learning_rate": 1.528648617411656e-05, + "loss": 0.8998, + "step": 240780 + }, + { + "epoch": 20.84, + "learning_rate": 1.5285041463203215e-05, + "loss": 1.0356, + "step": 240790 + }, + { + "epoch": 20.84, + "learning_rate": 1.5283596752289865e-05, + "loss": 0.9725, + "step": 240800 + }, + { + "epoch": 20.84, + "learning_rate": 1.5282152041376522e-05, + "loss": 0.9504, + "step": 240810 + }, + { + "epoch": 20.84, + "learning_rate": 1.5280707330463176e-05, + "loss": 0.9416, + "step": 240820 + }, + { + "epoch": 20.85, + "learning_rate": 1.527926261954983e-05, + "loss": 0.8958, + "step": 240830 + }, + { + "epoch": 20.85, + "learning_rate": 1.5277817908636482e-05, + "loss": 0.9274, + "step": 240840 + }, + { + "epoch": 20.85, + "learning_rate": 1.5276373197723136e-05, + "loss": 0.908, + "step": 240850 + }, + { + "epoch": 20.85, + "learning_rate": 1.527492848680979e-05, + "loss": 0.9406, + "step": 240860 + }, + { + "epoch": 20.85, + "learning_rate": 1.5273483775896443e-05, + "loss": 0.9362, + "step": 240870 + }, + { + "epoch": 20.85, + "learning_rate": 1.52720390649831e-05, + "loss": 0.9434, + "step": 240880 + }, + { + "epoch": 20.85, + "learning_rate": 1.527059435406975e-05, + "loss": 0.9428, + "step": 240890 + }, + { + "epoch": 20.85, + "learning_rate": 1.5269149643156406e-05, + "loss": 0.8533, + "step": 240900 + }, + { + "epoch": 20.85, + "learning_rate": 1.526770493224306e-05, + "loss": 0.9768, + "step": 240910 + }, + { + "epoch": 20.85, + "learning_rate": 1.5266260221329713e-05, + "loss": 0.9017, + "step": 240920 + }, + { + "epoch": 20.85, + "learning_rate": 1.5264815510416366e-05, + "loss": 0.9548, + "step": 240930 + }, + { + "epoch": 20.86, + "learning_rate": 1.526337079950302e-05, + "loss": 0.933, + "step": 240940 + }, + { + "epoch": 20.86, + "learning_rate": 1.5261926088589673e-05, + "loss": 0.9674, + "step": 240950 + }, + { + "epoch": 20.86, + "learning_rate": 1.5260481377676327e-05, + "loss": 0.9723, + "step": 240960 + }, + { + "epoch": 20.86, + "learning_rate": 1.5259036666762983e-05, + "loss": 1.0009, + "step": 240970 + }, + { + "epoch": 20.86, + "learning_rate": 1.5257591955849635e-05, + "loss": 0.9377, + "step": 240980 + }, + { + "epoch": 20.86, + "learning_rate": 1.5256147244936288e-05, + "loss": 0.9332, + "step": 240990 + }, + { + "epoch": 20.86, + "learning_rate": 1.5254702534022943e-05, + "loss": 0.9245, + "step": 241000 + }, + { + "epoch": 20.86, + "learning_rate": 1.5253257823109595e-05, + "loss": 0.975, + "step": 241010 + }, + { + "epoch": 20.86, + "learning_rate": 1.525181311219625e-05, + "loss": 0.9465, + "step": 241020 + }, + { + "epoch": 20.86, + "learning_rate": 1.5250368401282905e-05, + "loss": 0.9579, + "step": 241030 + }, + { + "epoch": 20.86, + "learning_rate": 1.5248923690369557e-05, + "loss": 0.955, + "step": 241040 + }, + { + "epoch": 20.86, + "learning_rate": 1.5247478979456212e-05, + "loss": 0.9653, + "step": 241050 + }, + { + "epoch": 20.87, + "learning_rate": 1.5246034268542867e-05, + "loss": 0.9825, + "step": 241060 + }, + { + "epoch": 20.87, + "learning_rate": 1.5244589557629519e-05, + "loss": 0.9464, + "step": 241070 + }, + { + "epoch": 20.87, + "learning_rate": 1.5243144846716172e-05, + "loss": 0.937, + "step": 241080 + }, + { + "epoch": 20.87, + "learning_rate": 1.5241700135802827e-05, + "loss": 0.9142, + "step": 241090 + }, + { + "epoch": 20.87, + "learning_rate": 1.5240255424889479e-05, + "loss": 0.9525, + "step": 241100 + }, + { + "epoch": 20.87, + "learning_rate": 1.5238810713976134e-05, + "loss": 0.9846, + "step": 241110 + }, + { + "epoch": 20.87, + "learning_rate": 1.523736600306279e-05, + "loss": 0.921, + "step": 241120 + }, + { + "epoch": 20.87, + "learning_rate": 1.5235921292149441e-05, + "loss": 0.8902, + "step": 241130 + }, + { + "epoch": 20.87, + "learning_rate": 1.5234476581236096e-05, + "loss": 0.9757, + "step": 241140 + }, + { + "epoch": 20.87, + "learning_rate": 1.5233031870322751e-05, + "loss": 0.9769, + "step": 241150 + }, + { + "epoch": 20.87, + "learning_rate": 1.5231587159409403e-05, + "loss": 0.9347, + "step": 241160 + }, + { + "epoch": 20.88, + "learning_rate": 1.5230142448496056e-05, + "loss": 0.9395, + "step": 241170 + }, + { + "epoch": 20.88, + "learning_rate": 1.5228697737582711e-05, + "loss": 0.9062, + "step": 241180 + }, + { + "epoch": 20.88, + "learning_rate": 1.5227253026669363e-05, + "loss": 0.9403, + "step": 241190 + }, + { + "epoch": 20.88, + "learning_rate": 1.5225808315756018e-05, + "loss": 0.9695, + "step": 241200 + }, + { + "epoch": 20.88, + "learning_rate": 1.5224363604842673e-05, + "loss": 0.9421, + "step": 241210 + }, + { + "epoch": 20.88, + "learning_rate": 1.5222918893929325e-05, + "loss": 0.9365, + "step": 241220 + }, + { + "epoch": 20.88, + "learning_rate": 1.522147418301598e-05, + "loss": 0.9561, + "step": 241230 + }, + { + "epoch": 20.88, + "learning_rate": 1.5220029472102635e-05, + "loss": 1.0119, + "step": 241240 + }, + { + "epoch": 20.88, + "learning_rate": 1.5218584761189287e-05, + "loss": 0.9439, + "step": 241250 + }, + { + "epoch": 20.88, + "learning_rate": 1.521714005027594e-05, + "loss": 0.8612, + "step": 241260 + }, + { + "epoch": 20.88, + "learning_rate": 1.5215695339362595e-05, + "loss": 0.9556, + "step": 241270 + }, + { + "epoch": 20.88, + "learning_rate": 1.5214250628449247e-05, + "loss": 0.9929, + "step": 241280 + }, + { + "epoch": 20.89, + "learning_rate": 1.5212805917535902e-05, + "loss": 0.9706, + "step": 241290 + }, + { + "epoch": 20.89, + "learning_rate": 1.5211361206622557e-05, + "loss": 0.8977, + "step": 241300 + }, + { + "epoch": 20.89, + "learning_rate": 1.5209916495709209e-05, + "loss": 0.9313, + "step": 241310 + }, + { + "epoch": 20.89, + "learning_rate": 1.5208471784795864e-05, + "loss": 0.9499, + "step": 241320 + }, + { + "epoch": 20.89, + "learning_rate": 1.5207027073882516e-05, + "loss": 0.9034, + "step": 241330 + }, + { + "epoch": 20.89, + "learning_rate": 1.520558236296917e-05, + "loss": 0.99, + "step": 241340 + }, + { + "epoch": 20.89, + "learning_rate": 1.5204137652055824e-05, + "loss": 0.9161, + "step": 241350 + }, + { + "epoch": 20.89, + "learning_rate": 1.5202692941142478e-05, + "loss": 0.9213, + "step": 241360 + }, + { + "epoch": 20.89, + "learning_rate": 1.5201248230229131e-05, + "loss": 0.9507, + "step": 241370 + }, + { + "epoch": 20.89, + "learning_rate": 1.5199803519315786e-05, + "loss": 0.9654, + "step": 241380 + }, + { + "epoch": 20.89, + "learning_rate": 1.5198358808402438e-05, + "loss": 1.0133, + "step": 241390 + }, + { + "epoch": 20.89, + "learning_rate": 1.5196914097489093e-05, + "loss": 0.96, + "step": 241400 + }, + { + "epoch": 20.9, + "learning_rate": 1.5195469386575748e-05, + "loss": 0.9769, + "step": 241410 + }, + { + "epoch": 20.9, + "learning_rate": 1.51940246756624e-05, + "loss": 1.0042, + "step": 241420 + }, + { + "epoch": 20.9, + "learning_rate": 1.5192579964749055e-05, + "loss": 0.9384, + "step": 241430 + }, + { + "epoch": 20.9, + "learning_rate": 1.5191135253835708e-05, + "loss": 0.9097, + "step": 241440 + }, + { + "epoch": 20.9, + "learning_rate": 1.5189690542922362e-05, + "loss": 0.9664, + "step": 241450 + }, + { + "epoch": 20.9, + "learning_rate": 1.5188245832009015e-05, + "loss": 0.9253, + "step": 241460 + }, + { + "epoch": 20.9, + "learning_rate": 1.518680112109567e-05, + "loss": 0.9549, + "step": 241470 + }, + { + "epoch": 20.9, + "learning_rate": 1.5185356410182322e-05, + "loss": 0.9074, + "step": 241480 + }, + { + "epoch": 20.9, + "learning_rate": 1.5183911699268977e-05, + "loss": 0.9162, + "step": 241490 + }, + { + "epoch": 20.9, + "learning_rate": 1.5182466988355632e-05, + "loss": 0.9241, + "step": 241500 + }, + { + "epoch": 20.9, + "learning_rate": 1.5181022277442284e-05, + "loss": 0.9408, + "step": 241510 + }, + { + "epoch": 20.91, + "learning_rate": 1.5179577566528939e-05, + "loss": 0.9512, + "step": 241520 + }, + { + "epoch": 20.91, + "learning_rate": 1.5178132855615592e-05, + "loss": 0.948, + "step": 241530 + }, + { + "epoch": 20.91, + "learning_rate": 1.5176688144702245e-05, + "loss": 0.9249, + "step": 241540 + }, + { + "epoch": 20.91, + "learning_rate": 1.5175243433788899e-05, + "loss": 0.9624, + "step": 241550 + }, + { + "epoch": 20.91, + "learning_rate": 1.5173798722875554e-05, + "loss": 0.9747, + "step": 241560 + }, + { + "epoch": 20.91, + "learning_rate": 1.5172354011962206e-05, + "loss": 0.9129, + "step": 241570 + }, + { + "epoch": 20.91, + "learning_rate": 1.517090930104886e-05, + "loss": 0.9404, + "step": 241580 + }, + { + "epoch": 20.91, + "learning_rate": 1.5169464590135516e-05, + "loss": 1.0005, + "step": 241590 + }, + { + "epoch": 20.91, + "learning_rate": 1.5168019879222168e-05, + "loss": 0.9741, + "step": 241600 + }, + { + "epoch": 20.91, + "learning_rate": 1.5166575168308823e-05, + "loss": 0.8931, + "step": 241610 + }, + { + "epoch": 20.91, + "learning_rate": 1.5165130457395476e-05, + "loss": 0.9743, + "step": 241620 + }, + { + "epoch": 20.91, + "learning_rate": 1.516368574648213e-05, + "loss": 0.9597, + "step": 241630 + }, + { + "epoch": 20.92, + "learning_rate": 1.5162241035568783e-05, + "loss": 0.9449, + "step": 241640 + }, + { + "epoch": 20.92, + "learning_rate": 1.5160796324655438e-05, + "loss": 0.9895, + "step": 241650 + }, + { + "epoch": 20.92, + "learning_rate": 1.515935161374209e-05, + "loss": 0.92, + "step": 241660 + }, + { + "epoch": 20.92, + "learning_rate": 1.5157906902828745e-05, + "loss": 0.9612, + "step": 241670 + }, + { + "epoch": 20.92, + "learning_rate": 1.51564621919154e-05, + "loss": 0.9473, + "step": 241680 + }, + { + "epoch": 20.92, + "learning_rate": 1.5155017481002051e-05, + "loss": 0.98, + "step": 241690 + }, + { + "epoch": 20.92, + "learning_rate": 1.5153572770088707e-05, + "loss": 0.9089, + "step": 241700 + }, + { + "epoch": 20.92, + "learning_rate": 1.515212805917536e-05, + "loss": 0.9162, + "step": 241710 + }, + { + "epoch": 20.92, + "learning_rate": 1.5150683348262013e-05, + "loss": 0.9673, + "step": 241720 + }, + { + "epoch": 20.92, + "learning_rate": 1.5149238637348667e-05, + "loss": 0.9715, + "step": 241730 + }, + { + "epoch": 20.92, + "learning_rate": 1.5147793926435322e-05, + "loss": 1.0055, + "step": 241740 + }, + { + "epoch": 20.93, + "learning_rate": 1.5146349215521974e-05, + "loss": 0.9357, + "step": 241750 + }, + { + "epoch": 20.93, + "learning_rate": 1.5144904504608629e-05, + "loss": 0.9864, + "step": 241760 + }, + { + "epoch": 20.93, + "learning_rate": 1.5143459793695284e-05, + "loss": 0.9168, + "step": 241770 + }, + { + "epoch": 20.93, + "learning_rate": 1.5142015082781935e-05, + "loss": 0.8792, + "step": 241780 + }, + { + "epoch": 20.93, + "learning_rate": 1.514057037186859e-05, + "loss": 0.9038, + "step": 241790 + }, + { + "epoch": 20.93, + "learning_rate": 1.5139125660955244e-05, + "loss": 0.916, + "step": 241800 + }, + { + "epoch": 20.93, + "learning_rate": 1.5137680950041897e-05, + "loss": 0.961, + "step": 241810 + }, + { + "epoch": 20.93, + "learning_rate": 1.513623623912855e-05, + "loss": 0.9844, + "step": 241820 + }, + { + "epoch": 20.93, + "learning_rate": 1.5134791528215206e-05, + "loss": 0.9781, + "step": 241830 + }, + { + "epoch": 20.93, + "learning_rate": 1.5133346817301857e-05, + "loss": 0.9838, + "step": 241840 + }, + { + "epoch": 20.93, + "learning_rate": 1.5131902106388513e-05, + "loss": 0.8919, + "step": 241850 + }, + { + "epoch": 20.93, + "learning_rate": 1.5130457395475168e-05, + "loss": 0.9192, + "step": 241860 + }, + { + "epoch": 20.94, + "learning_rate": 1.512901268456182e-05, + "loss": 0.9549, + "step": 241870 + }, + { + "epoch": 20.94, + "learning_rate": 1.5127567973648474e-05, + "loss": 0.9355, + "step": 241880 + }, + { + "epoch": 20.94, + "learning_rate": 1.5126123262735128e-05, + "loss": 0.943, + "step": 241890 + }, + { + "epoch": 20.94, + "learning_rate": 1.5124678551821781e-05, + "loss": 0.9709, + "step": 241900 + }, + { + "epoch": 20.94, + "learning_rate": 1.5123233840908435e-05, + "loss": 0.9219, + "step": 241910 + }, + { + "epoch": 20.94, + "learning_rate": 1.512178912999509e-05, + "loss": 0.9369, + "step": 241920 + }, + { + "epoch": 20.94, + "learning_rate": 1.5120344419081741e-05, + "loss": 0.9481, + "step": 241930 + }, + { + "epoch": 20.94, + "learning_rate": 1.5118899708168396e-05, + "loss": 0.9877, + "step": 241940 + }, + { + "epoch": 20.94, + "learning_rate": 1.5117454997255052e-05, + "loss": 0.9598, + "step": 241950 + }, + { + "epoch": 20.94, + "learning_rate": 1.5116010286341703e-05, + "loss": 0.9356, + "step": 241960 + }, + { + "epoch": 20.94, + "learning_rate": 1.5114565575428358e-05, + "loss": 0.8889, + "step": 241970 + }, + { + "epoch": 20.95, + "learning_rate": 1.5113120864515012e-05, + "loss": 0.9557, + "step": 241980 + }, + { + "epoch": 20.95, + "learning_rate": 1.5111676153601665e-05, + "loss": 0.8615, + "step": 241990 + }, + { + "epoch": 20.95, + "learning_rate": 1.5110231442688319e-05, + "loss": 0.9283, + "step": 242000 + }, + { + "epoch": 20.95, + "learning_rate": 1.5108786731774974e-05, + "loss": 0.9612, + "step": 242010 + }, + { + "epoch": 20.95, + "learning_rate": 1.5107342020861625e-05, + "loss": 0.9975, + "step": 242020 + }, + { + "epoch": 20.95, + "learning_rate": 1.510589730994828e-05, + "loss": 0.9193, + "step": 242030 + }, + { + "epoch": 20.95, + "learning_rate": 1.5104452599034936e-05, + "loss": 0.9878, + "step": 242040 + }, + { + "epoch": 20.95, + "learning_rate": 1.5103007888121587e-05, + "loss": 0.9614, + "step": 242050 + }, + { + "epoch": 20.95, + "learning_rate": 1.5101563177208242e-05, + "loss": 0.9469, + "step": 242060 + }, + { + "epoch": 20.95, + "learning_rate": 1.5100118466294896e-05, + "loss": 0.9832, + "step": 242070 + }, + { + "epoch": 20.95, + "learning_rate": 1.5098673755381549e-05, + "loss": 0.9427, + "step": 242080 + }, + { + "epoch": 20.95, + "learning_rate": 1.5097229044468203e-05, + "loss": 0.9663, + "step": 242090 + }, + { + "epoch": 20.96, + "learning_rate": 1.5095784333554858e-05, + "loss": 0.9457, + "step": 242100 + }, + { + "epoch": 20.96, + "learning_rate": 1.509433962264151e-05, + "loss": 0.9372, + "step": 242110 + }, + { + "epoch": 20.96, + "learning_rate": 1.5092894911728164e-05, + "loss": 0.9693, + "step": 242120 + }, + { + "epoch": 20.96, + "learning_rate": 1.509145020081482e-05, + "loss": 0.9224, + "step": 242130 + }, + { + "epoch": 20.96, + "learning_rate": 1.5090005489901471e-05, + "loss": 0.9746, + "step": 242140 + }, + { + "epoch": 20.96, + "learning_rate": 1.5088560778988126e-05, + "loss": 0.9459, + "step": 242150 + }, + { + "epoch": 20.96, + "learning_rate": 1.508711606807478e-05, + "loss": 0.8643, + "step": 242160 + }, + { + "epoch": 20.96, + "learning_rate": 1.5085671357161433e-05, + "loss": 0.9143, + "step": 242170 + }, + { + "epoch": 20.96, + "learning_rate": 1.5084226646248086e-05, + "loss": 0.8883, + "step": 242180 + }, + { + "epoch": 20.96, + "learning_rate": 1.5082781935334742e-05, + "loss": 0.9319, + "step": 242190 + }, + { + "epoch": 20.96, + "learning_rate": 1.5081337224421393e-05, + "loss": 0.9547, + "step": 242200 + }, + { + "epoch": 20.97, + "learning_rate": 1.5079892513508048e-05, + "loss": 0.8937, + "step": 242210 + }, + { + "epoch": 20.97, + "learning_rate": 1.5078447802594703e-05, + "loss": 0.9824, + "step": 242220 + }, + { + "epoch": 20.97, + "learning_rate": 1.5077003091681355e-05, + "loss": 0.8999, + "step": 242230 + }, + { + "epoch": 20.97, + "learning_rate": 1.507555838076801e-05, + "loss": 0.9818, + "step": 242240 + }, + { + "epoch": 20.97, + "learning_rate": 1.5074113669854662e-05, + "loss": 0.916, + "step": 242250 + }, + { + "epoch": 20.97, + "learning_rate": 1.5072668958941315e-05, + "loss": 0.894, + "step": 242260 + }, + { + "epoch": 20.97, + "learning_rate": 1.507122424802797e-05, + "loss": 0.919, + "step": 242270 + }, + { + "epoch": 20.97, + "learning_rate": 1.5069779537114622e-05, + "loss": 0.9292, + "step": 242280 + }, + { + "epoch": 20.97, + "learning_rate": 1.5068334826201277e-05, + "loss": 0.9133, + "step": 242290 + }, + { + "epoch": 20.97, + "learning_rate": 1.5066890115287932e-05, + "loss": 0.9381, + "step": 242300 + }, + { + "epoch": 20.97, + "learning_rate": 1.5065445404374584e-05, + "loss": 0.9643, + "step": 242310 + }, + { + "epoch": 20.97, + "learning_rate": 1.5064000693461239e-05, + "loss": 1.0015, + "step": 242320 + }, + { + "epoch": 20.98, + "learning_rate": 1.5062555982547894e-05, + "loss": 0.9645, + "step": 242330 + }, + { + "epoch": 20.98, + "learning_rate": 1.5061111271634546e-05, + "loss": 0.8853, + "step": 242340 + }, + { + "epoch": 20.98, + "learning_rate": 1.50596665607212e-05, + "loss": 0.8991, + "step": 242350 + }, + { + "epoch": 20.98, + "learning_rate": 1.5058221849807854e-05, + "loss": 0.9397, + "step": 242360 + }, + { + "epoch": 20.98, + "learning_rate": 1.5056777138894506e-05, + "loss": 0.9705, + "step": 242370 + }, + { + "epoch": 20.98, + "learning_rate": 1.5055332427981161e-05, + "loss": 0.9665, + "step": 242380 + }, + { + "epoch": 20.98, + "learning_rate": 1.5053887717067816e-05, + "loss": 0.8962, + "step": 242390 + }, + { + "epoch": 20.98, + "learning_rate": 1.5052443006154468e-05, + "loss": 0.8581, + "step": 242400 + }, + { + "epoch": 20.98, + "learning_rate": 1.5050998295241123e-05, + "loss": 0.9378, + "step": 242410 + }, + { + "epoch": 20.98, + "learning_rate": 1.5049553584327778e-05, + "loss": 0.9331, + "step": 242420 + }, + { + "epoch": 20.98, + "learning_rate": 1.504810887341443e-05, + "loss": 0.9157, + "step": 242430 + }, + { + "epoch": 20.98, + "learning_rate": 1.5046664162501083e-05, + "loss": 1.0101, + "step": 242440 + }, + { + "epoch": 20.99, + "learning_rate": 1.5045219451587738e-05, + "loss": 0.9288, + "step": 242450 + }, + { + "epoch": 20.99, + "learning_rate": 1.504377474067439e-05, + "loss": 0.9559, + "step": 242460 + }, + { + "epoch": 20.99, + "learning_rate": 1.5042330029761045e-05, + "loss": 0.9331, + "step": 242470 + }, + { + "epoch": 20.99, + "learning_rate": 1.50408853188477e-05, + "loss": 0.944, + "step": 242480 + }, + { + "epoch": 20.99, + "learning_rate": 1.5039440607934352e-05, + "loss": 0.9171, + "step": 242490 + }, + { + "epoch": 20.99, + "learning_rate": 1.5037995897021007e-05, + "loss": 0.9035, + "step": 242500 + }, + { + "epoch": 20.99, + "learning_rate": 1.5036551186107662e-05, + "loss": 0.9636, + "step": 242510 + }, + { + "epoch": 20.99, + "learning_rate": 1.5035106475194314e-05, + "loss": 0.9796, + "step": 242520 + }, + { + "epoch": 20.99, + "learning_rate": 1.5033661764280967e-05, + "loss": 0.9331, + "step": 242530 + }, + { + "epoch": 20.99, + "learning_rate": 1.5032217053367622e-05, + "loss": 0.9449, + "step": 242540 + }, + { + "epoch": 20.99, + "learning_rate": 1.5030772342454274e-05, + "loss": 0.9632, + "step": 242550 + }, + { + "epoch": 21.0, + "learning_rate": 1.5029327631540929e-05, + "loss": 0.9506, + "step": 242560 + }, + { + "epoch": 21.0, + "learning_rate": 1.5027882920627584e-05, + "loss": 0.9246, + "step": 242570 + }, + { + "epoch": 21.0, + "learning_rate": 1.5026438209714236e-05, + "loss": 0.9451, + "step": 242580 + }, + { + "epoch": 21.0, + "learning_rate": 1.5024993498800891e-05, + "loss": 0.9279, + "step": 242590 + }, + { + "epoch": 21.0, + "learning_rate": 1.5023548787887546e-05, + "loss": 0.9835, + "step": 242600 + }, + { + "epoch": 21.0, + "learning_rate": 1.5022104076974198e-05, + "loss": 0.9399, + "step": 242610 + }, + { + "epoch": 21.0, + "eval_cer": 0.993128401201183, + "eval_em": 0, + "eval_f1": 0.11730630725049648, + "eval_loss": 0.9501730799674988, + "eval_runtime": 1329.865, + "eval_samples_per_second": 3.861, + "eval_steps_per_second": 0.483, + "eval_wer": 0.9673334963609078, + "step": 242613 + }, + { + "epoch": 21.0, + "learning_rate": 1.5020659366060851e-05, + "loss": 0.9515, + "step": 242620 + }, + { + "epoch": 21.0, + "learning_rate": 1.5019214655147506e-05, + "loss": 0.9312, + "step": 242630 + }, + { + "epoch": 21.0, + "learning_rate": 1.5017769944234158e-05, + "loss": 0.9056, + "step": 242640 + }, + { + "epoch": 21.0, + "learning_rate": 1.5016325233320813e-05, + "loss": 0.9176, + "step": 242650 + }, + { + "epoch": 21.0, + "learning_rate": 1.5014880522407468e-05, + "loss": 1.0053, + "step": 242660 + }, + { + "epoch": 21.0, + "learning_rate": 1.501343581149412e-05, + "loss": 0.942, + "step": 242670 + }, + { + "epoch": 21.01, + "learning_rate": 1.5011991100580775e-05, + "loss": 0.905, + "step": 242680 + }, + { + "epoch": 21.01, + "learning_rate": 1.501054638966743e-05, + "loss": 0.884, + "step": 242690 + }, + { + "epoch": 21.01, + "learning_rate": 1.5009101678754082e-05, + "loss": 0.9021, + "step": 242700 + }, + { + "epoch": 21.01, + "learning_rate": 1.5007656967840735e-05, + "loss": 0.967, + "step": 242710 + }, + { + "epoch": 21.01, + "learning_rate": 1.500621225692739e-05, + "loss": 0.9469, + "step": 242720 + }, + { + "epoch": 21.01, + "learning_rate": 1.5004767546014042e-05, + "loss": 0.9372, + "step": 242730 + }, + { + "epoch": 21.01, + "learning_rate": 1.5003322835100697e-05, + "loss": 0.9272, + "step": 242740 + }, + { + "epoch": 21.01, + "learning_rate": 1.5001878124187352e-05, + "loss": 0.9271, + "step": 242750 + }, + { + "epoch": 21.01, + "learning_rate": 1.5000433413274004e-05, + "loss": 0.9155, + "step": 242760 + }, + { + "epoch": 21.01, + "learning_rate": 1.4998988702360659e-05, + "loss": 0.9317, + "step": 242770 + }, + { + "epoch": 21.01, + "learning_rate": 1.4997543991447314e-05, + "loss": 0.9745, + "step": 242780 + }, + { + "epoch": 21.02, + "learning_rate": 1.4996099280533966e-05, + "loss": 0.9177, + "step": 242790 + }, + { + "epoch": 21.02, + "learning_rate": 1.4994654569620619e-05, + "loss": 0.9544, + "step": 242800 + }, + { + "epoch": 21.02, + "learning_rate": 1.4993209858707274e-05, + "loss": 0.9301, + "step": 242810 + }, + { + "epoch": 21.02, + "learning_rate": 1.4991765147793926e-05, + "loss": 0.9384, + "step": 242820 + }, + { + "epoch": 21.02, + "learning_rate": 1.499032043688058e-05, + "loss": 0.896, + "step": 242830 + }, + { + "epoch": 21.02, + "learning_rate": 1.4988875725967236e-05, + "loss": 0.9214, + "step": 242840 + }, + { + "epoch": 21.02, + "learning_rate": 1.4987431015053888e-05, + "loss": 0.9264, + "step": 242850 + }, + { + "epoch": 21.02, + "learning_rate": 1.4985986304140543e-05, + "loss": 0.96, + "step": 242860 + }, + { + "epoch": 21.02, + "learning_rate": 1.4984541593227198e-05, + "loss": 0.9558, + "step": 242870 + }, + { + "epoch": 21.02, + "learning_rate": 1.498309688231385e-05, + "loss": 0.9015, + "step": 242880 + }, + { + "epoch": 21.02, + "learning_rate": 1.4981652171400503e-05, + "loss": 0.9749, + "step": 242890 + }, + { + "epoch": 21.02, + "learning_rate": 1.4980207460487158e-05, + "loss": 1.0066, + "step": 242900 + }, + { + "epoch": 21.03, + "learning_rate": 1.497876274957381e-05, + "loss": 0.967, + "step": 242910 + }, + { + "epoch": 21.03, + "learning_rate": 1.4977318038660465e-05, + "loss": 0.8708, + "step": 242920 + }, + { + "epoch": 21.03, + "learning_rate": 1.497587332774712e-05, + "loss": 0.8791, + "step": 242930 + }, + { + "epoch": 21.03, + "learning_rate": 1.4974428616833772e-05, + "loss": 0.9018, + "step": 242940 + }, + { + "epoch": 21.03, + "learning_rate": 1.4972983905920427e-05, + "loss": 0.9761, + "step": 242950 + }, + { + "epoch": 21.03, + "learning_rate": 1.4971539195007082e-05, + "loss": 0.9297, + "step": 242960 + }, + { + "epoch": 21.03, + "learning_rate": 1.4970094484093733e-05, + "loss": 0.9912, + "step": 242970 + }, + { + "epoch": 21.03, + "learning_rate": 1.4968649773180387e-05, + "loss": 1.0018, + "step": 242980 + }, + { + "epoch": 21.03, + "learning_rate": 1.4967205062267042e-05, + "loss": 0.9779, + "step": 242990 + }, + { + "epoch": 21.03, + "learning_rate": 1.4965760351353694e-05, + "loss": 0.9187, + "step": 243000 + }, + { + "epoch": 21.03, + "learning_rate": 1.4964315640440349e-05, + "loss": 0.9143, + "step": 243010 + }, + { + "epoch": 21.04, + "learning_rate": 1.4962870929527004e-05, + "loss": 0.8723, + "step": 243020 + }, + { + "epoch": 21.04, + "learning_rate": 1.4961426218613656e-05, + "loss": 0.9155, + "step": 243030 + }, + { + "epoch": 21.04, + "learning_rate": 1.495998150770031e-05, + "loss": 0.9495, + "step": 243040 + }, + { + "epoch": 21.04, + "learning_rate": 1.4958536796786966e-05, + "loss": 0.8945, + "step": 243050 + }, + { + "epoch": 21.04, + "learning_rate": 1.4957092085873617e-05, + "loss": 0.8975, + "step": 243060 + }, + { + "epoch": 21.04, + "learning_rate": 1.495564737496027e-05, + "loss": 0.9245, + "step": 243070 + }, + { + "epoch": 21.04, + "learning_rate": 1.4954202664046926e-05, + "loss": 0.8946, + "step": 243080 + }, + { + "epoch": 21.04, + "learning_rate": 1.4952757953133578e-05, + "loss": 0.949, + "step": 243090 + }, + { + "epoch": 21.04, + "learning_rate": 1.4951313242220233e-05, + "loss": 0.9414, + "step": 243100 + }, + { + "epoch": 21.04, + "learning_rate": 1.4949868531306888e-05, + "loss": 0.9858, + "step": 243110 + }, + { + "epoch": 21.04, + "learning_rate": 1.494842382039354e-05, + "loss": 0.9242, + "step": 243120 + }, + { + "epoch": 21.04, + "learning_rate": 1.4946979109480195e-05, + "loss": 0.9794, + "step": 243130 + }, + { + "epoch": 21.05, + "learning_rate": 1.4945534398566846e-05, + "loss": 0.9774, + "step": 243140 + }, + { + "epoch": 21.05, + "learning_rate": 1.4944089687653501e-05, + "loss": 0.963, + "step": 243150 + }, + { + "epoch": 21.05, + "learning_rate": 1.4942644976740155e-05, + "loss": 0.9544, + "step": 243160 + }, + { + "epoch": 21.05, + "learning_rate": 1.4941200265826808e-05, + "loss": 0.9439, + "step": 243170 + }, + { + "epoch": 21.05, + "learning_rate": 1.4939755554913462e-05, + "loss": 0.9097, + "step": 243180 + }, + { + "epoch": 21.05, + "learning_rate": 1.4938310844000117e-05, + "loss": 0.9196, + "step": 243190 + }, + { + "epoch": 21.05, + "learning_rate": 1.4936866133086768e-05, + "loss": 0.9089, + "step": 243200 + }, + { + "epoch": 21.05, + "learning_rate": 1.4935421422173423e-05, + "loss": 0.9497, + "step": 243210 + }, + { + "epoch": 21.05, + "learning_rate": 1.4933976711260078e-05, + "loss": 0.9149, + "step": 243220 + }, + { + "epoch": 21.05, + "learning_rate": 1.493253200034673e-05, + "loss": 0.9488, + "step": 243230 + }, + { + "epoch": 21.05, + "learning_rate": 1.4931087289433385e-05, + "loss": 0.9422, + "step": 243240 + }, + { + "epoch": 21.06, + "learning_rate": 1.4929642578520039e-05, + "loss": 0.9188, + "step": 243250 + }, + { + "epoch": 21.06, + "learning_rate": 1.4928197867606692e-05, + "loss": 0.9145, + "step": 243260 + }, + { + "epoch": 21.06, + "learning_rate": 1.4926753156693345e-05, + "loss": 0.9494, + "step": 243270 + }, + { + "epoch": 21.06, + "learning_rate": 1.492530844578e-05, + "loss": 0.9065, + "step": 243280 + }, + { + "epoch": 21.06, + "learning_rate": 1.4923863734866652e-05, + "loss": 0.9143, + "step": 243290 + }, + { + "epoch": 21.06, + "learning_rate": 1.4922419023953307e-05, + "loss": 0.9145, + "step": 243300 + }, + { + "epoch": 21.06, + "learning_rate": 1.4920974313039962e-05, + "loss": 0.9321, + "step": 243310 + }, + { + "epoch": 21.06, + "learning_rate": 1.4919529602126614e-05, + "loss": 0.9005, + "step": 243320 + }, + { + "epoch": 21.06, + "learning_rate": 1.491808489121327e-05, + "loss": 0.9557, + "step": 243330 + }, + { + "epoch": 21.06, + "learning_rate": 1.4916640180299923e-05, + "loss": 0.9226, + "step": 243340 + }, + { + "epoch": 21.06, + "learning_rate": 1.4915195469386576e-05, + "loss": 0.8768, + "step": 243350 + }, + { + "epoch": 21.06, + "learning_rate": 1.491375075847323e-05, + "loss": 0.9094, + "step": 243360 + }, + { + "epoch": 21.07, + "learning_rate": 1.4912306047559884e-05, + "loss": 0.9486, + "step": 243370 + }, + { + "epoch": 21.07, + "learning_rate": 1.4910861336646536e-05, + "loss": 0.8943, + "step": 243380 + }, + { + "epoch": 21.07, + "learning_rate": 1.4909416625733191e-05, + "loss": 0.9575, + "step": 243390 + }, + { + "epoch": 21.07, + "learning_rate": 1.4907971914819846e-05, + "loss": 0.8834, + "step": 243400 + }, + { + "epoch": 21.07, + "learning_rate": 1.4906527203906498e-05, + "loss": 0.9497, + "step": 243410 + }, + { + "epoch": 21.07, + "learning_rate": 1.4905082492993153e-05, + "loss": 0.9632, + "step": 243420 + }, + { + "epoch": 21.07, + "learning_rate": 1.4903637782079807e-05, + "loss": 0.9623, + "step": 243430 + }, + { + "epoch": 21.07, + "learning_rate": 1.490219307116646e-05, + "loss": 0.9299, + "step": 243440 + }, + { + "epoch": 21.07, + "learning_rate": 1.4900748360253113e-05, + "loss": 0.9381, + "step": 243450 + }, + { + "epoch": 21.07, + "learning_rate": 1.4899303649339768e-05, + "loss": 0.9513, + "step": 243460 + }, + { + "epoch": 21.07, + "learning_rate": 1.489785893842642e-05, + "loss": 0.9527, + "step": 243470 + }, + { + "epoch": 21.07, + "learning_rate": 1.4896414227513075e-05, + "loss": 0.8908, + "step": 243480 + }, + { + "epoch": 21.08, + "learning_rate": 1.489496951659973e-05, + "loss": 0.9211, + "step": 243490 + }, + { + "epoch": 21.08, + "learning_rate": 1.4893524805686382e-05, + "loss": 0.9659, + "step": 243500 + }, + { + "epoch": 21.08, + "learning_rate": 1.4892080094773037e-05, + "loss": 0.9383, + "step": 243510 + }, + { + "epoch": 21.08, + "learning_rate": 1.489063538385969e-05, + "loss": 0.9036, + "step": 243520 + }, + { + "epoch": 21.08, + "learning_rate": 1.4889190672946344e-05, + "loss": 0.9561, + "step": 243530 + }, + { + "epoch": 21.08, + "learning_rate": 1.4887745962032997e-05, + "loss": 0.9286, + "step": 243540 + }, + { + "epoch": 21.08, + "learning_rate": 1.4886301251119652e-05, + "loss": 0.9327, + "step": 243550 + }, + { + "epoch": 21.08, + "learning_rate": 1.4884856540206304e-05, + "loss": 0.9547, + "step": 243560 + }, + { + "epoch": 21.08, + "learning_rate": 1.488341182929296e-05, + "loss": 0.959, + "step": 243570 + }, + { + "epoch": 21.08, + "learning_rate": 1.4881967118379614e-05, + "loss": 0.9222, + "step": 243580 + }, + { + "epoch": 21.08, + "learning_rate": 1.4880522407466266e-05, + "loss": 0.8879, + "step": 243590 + }, + { + "epoch": 21.09, + "learning_rate": 1.4879077696552921e-05, + "loss": 0.9453, + "step": 243600 + }, + { + "epoch": 21.09, + "learning_rate": 1.4877632985639574e-05, + "loss": 0.9978, + "step": 243610 + }, + { + "epoch": 21.09, + "learning_rate": 1.4876188274726228e-05, + "loss": 0.9724, + "step": 243620 + }, + { + "epoch": 21.09, + "learning_rate": 1.4874743563812881e-05, + "loss": 0.9057, + "step": 243630 + }, + { + "epoch": 21.09, + "learning_rate": 1.4873298852899536e-05, + "loss": 0.9423, + "step": 243640 + }, + { + "epoch": 21.09, + "learning_rate": 1.4871854141986188e-05, + "loss": 0.9677, + "step": 243650 + }, + { + "epoch": 21.09, + "learning_rate": 1.4870409431072843e-05, + "loss": 0.9582, + "step": 243660 + }, + { + "epoch": 21.09, + "learning_rate": 1.4868964720159498e-05, + "loss": 0.8659, + "step": 243670 + }, + { + "epoch": 21.09, + "learning_rate": 1.486752000924615e-05, + "loss": 0.954, + "step": 243680 + }, + { + "epoch": 21.09, + "learning_rate": 1.4866075298332805e-05, + "loss": 0.8907, + "step": 243690 + }, + { + "epoch": 21.09, + "learning_rate": 1.4864630587419458e-05, + "loss": 0.9174, + "step": 243700 + }, + { + "epoch": 21.09, + "learning_rate": 1.4863185876506112e-05, + "loss": 0.9356, + "step": 243710 + }, + { + "epoch": 21.1, + "learning_rate": 1.4861741165592765e-05, + "loss": 0.933, + "step": 243720 + }, + { + "epoch": 21.1, + "learning_rate": 1.486029645467942e-05, + "loss": 0.9872, + "step": 243730 + }, + { + "epoch": 21.1, + "learning_rate": 1.4858851743766072e-05, + "loss": 0.8816, + "step": 243740 + }, + { + "epoch": 21.1, + "learning_rate": 1.4857407032852727e-05, + "loss": 0.9757, + "step": 243750 + }, + { + "epoch": 21.1, + "learning_rate": 1.4855962321939382e-05, + "loss": 0.925, + "step": 243760 + }, + { + "epoch": 21.1, + "learning_rate": 1.4854517611026034e-05, + "loss": 0.9443, + "step": 243770 + }, + { + "epoch": 21.1, + "learning_rate": 1.4853072900112689e-05, + "loss": 0.8975, + "step": 243780 + }, + { + "epoch": 21.1, + "learning_rate": 1.4851628189199342e-05, + "loss": 0.9475, + "step": 243790 + }, + { + "epoch": 21.1, + "learning_rate": 1.4850183478285996e-05, + "loss": 0.9881, + "step": 243800 + }, + { + "epoch": 21.1, + "learning_rate": 1.4848738767372649e-05, + "loss": 0.9683, + "step": 243810 + }, + { + "epoch": 21.1, + "learning_rate": 1.4847294056459304e-05, + "loss": 0.9678, + "step": 243820 + }, + { + "epoch": 21.11, + "learning_rate": 1.4845849345545956e-05, + "loss": 0.9368, + "step": 243830 + }, + { + "epoch": 21.11, + "learning_rate": 1.4844404634632611e-05, + "loss": 0.9562, + "step": 243840 + }, + { + "epoch": 21.11, + "learning_rate": 1.4842959923719266e-05, + "loss": 0.884, + "step": 243850 + }, + { + "epoch": 21.11, + "learning_rate": 1.4841515212805918e-05, + "loss": 0.9366, + "step": 243860 + }, + { + "epoch": 21.11, + "learning_rate": 1.4840070501892573e-05, + "loss": 0.9352, + "step": 243870 + }, + { + "epoch": 21.11, + "learning_rate": 1.4838625790979226e-05, + "loss": 0.8974, + "step": 243880 + }, + { + "epoch": 21.11, + "learning_rate": 1.483718108006588e-05, + "loss": 0.9138, + "step": 243890 + }, + { + "epoch": 21.11, + "learning_rate": 1.4835736369152533e-05, + "loss": 0.9088, + "step": 243900 + }, + { + "epoch": 21.11, + "learning_rate": 1.4834291658239188e-05, + "loss": 0.9817, + "step": 243910 + }, + { + "epoch": 21.11, + "learning_rate": 1.483284694732584e-05, + "loss": 0.9244, + "step": 243920 + }, + { + "epoch": 21.11, + "learning_rate": 1.4831402236412495e-05, + "loss": 0.9117, + "step": 243930 + }, + { + "epoch": 21.11, + "learning_rate": 1.482995752549915e-05, + "loss": 0.9631, + "step": 243940 + }, + { + "epoch": 21.12, + "learning_rate": 1.4828512814585802e-05, + "loss": 0.9373, + "step": 243950 + }, + { + "epoch": 21.12, + "learning_rate": 1.4827068103672457e-05, + "loss": 0.9453, + "step": 243960 + }, + { + "epoch": 21.12, + "learning_rate": 1.482562339275911e-05, + "loss": 0.9784, + "step": 243970 + }, + { + "epoch": 21.12, + "learning_rate": 1.4824178681845764e-05, + "loss": 0.9635, + "step": 243980 + }, + { + "epoch": 21.12, + "learning_rate": 1.4822733970932417e-05, + "loss": 0.927, + "step": 243990 + }, + { + "epoch": 21.12, + "learning_rate": 1.4821289260019072e-05, + "loss": 0.9288, + "step": 244000 + }, + { + "epoch": 21.12, + "learning_rate": 1.4819844549105724e-05, + "loss": 0.943, + "step": 244010 + }, + { + "epoch": 21.12, + "learning_rate": 1.4818399838192379e-05, + "loss": 0.9343, + "step": 244020 + }, + { + "epoch": 21.12, + "learning_rate": 1.481695512727903e-05, + "loss": 0.9667, + "step": 244030 + }, + { + "epoch": 21.12, + "learning_rate": 1.4815510416365686e-05, + "loss": 0.9274, + "step": 244040 + }, + { + "epoch": 21.12, + "learning_rate": 1.481406570545234e-05, + "loss": 0.9137, + "step": 244050 + }, + { + "epoch": 21.13, + "learning_rate": 1.4812620994538992e-05, + "loss": 0.9698, + "step": 244060 + }, + { + "epoch": 21.13, + "learning_rate": 1.4811176283625648e-05, + "loss": 0.9268, + "step": 244070 + }, + { + "epoch": 21.13, + "learning_rate": 1.4809731572712301e-05, + "loss": 0.9409, + "step": 244080 + }, + { + "epoch": 21.13, + "learning_rate": 1.4808286861798953e-05, + "loss": 0.9374, + "step": 244090 + }, + { + "epoch": 21.13, + "learning_rate": 1.4806842150885608e-05, + "loss": 0.8942, + "step": 244100 + }, + { + "epoch": 21.13, + "learning_rate": 1.4805397439972263e-05, + "loss": 0.9291, + "step": 244110 + }, + { + "epoch": 21.13, + "learning_rate": 1.4803952729058915e-05, + "loss": 0.9463, + "step": 244120 + }, + { + "epoch": 21.13, + "learning_rate": 1.480250801814557e-05, + "loss": 0.929, + "step": 244130 + }, + { + "epoch": 21.13, + "learning_rate": 1.4801063307232225e-05, + "loss": 0.9299, + "step": 244140 + }, + { + "epoch": 21.13, + "learning_rate": 1.4799618596318876e-05, + "loss": 0.9129, + "step": 244150 + }, + { + "epoch": 21.13, + "learning_rate": 1.4798173885405531e-05, + "loss": 0.9241, + "step": 244160 + }, + { + "epoch": 21.13, + "learning_rate": 1.4796729174492185e-05, + "loss": 0.9181, + "step": 244170 + }, + { + "epoch": 21.14, + "learning_rate": 1.4795284463578837e-05, + "loss": 0.9552, + "step": 244180 + }, + { + "epoch": 21.14, + "learning_rate": 1.4793839752665492e-05, + "loss": 0.9431, + "step": 244190 + }, + { + "epoch": 21.14, + "learning_rate": 1.4792395041752147e-05, + "loss": 0.8592, + "step": 244200 + }, + { + "epoch": 21.14, + "learning_rate": 1.4790950330838798e-05, + "loss": 0.9763, + "step": 244210 + }, + { + "epoch": 21.14, + "learning_rate": 1.4789505619925454e-05, + "loss": 0.8794, + "step": 244220 + }, + { + "epoch": 21.14, + "learning_rate": 1.4788060909012109e-05, + "loss": 0.9912, + "step": 244230 + }, + { + "epoch": 21.14, + "learning_rate": 1.478661619809876e-05, + "loss": 0.9375, + "step": 244240 + }, + { + "epoch": 21.14, + "learning_rate": 1.4785171487185415e-05, + "loss": 0.9329, + "step": 244250 + }, + { + "epoch": 21.14, + "learning_rate": 1.4783726776272069e-05, + "loss": 0.986, + "step": 244260 + }, + { + "epoch": 21.14, + "learning_rate": 1.478228206535872e-05, + "loss": 0.9218, + "step": 244270 + }, + { + "epoch": 21.14, + "learning_rate": 1.4780837354445376e-05, + "loss": 0.9362, + "step": 244280 + }, + { + "epoch": 21.15, + "learning_rate": 1.477939264353203e-05, + "loss": 0.9125, + "step": 244290 + }, + { + "epoch": 21.15, + "learning_rate": 1.4777947932618682e-05, + "loss": 0.9588, + "step": 244300 + }, + { + "epoch": 21.15, + "learning_rate": 1.4776503221705338e-05, + "loss": 0.9319, + "step": 244310 + }, + { + "epoch": 21.15, + "learning_rate": 1.4775058510791993e-05, + "loss": 0.9605, + "step": 244320 + }, + { + "epoch": 21.15, + "learning_rate": 1.4773613799878644e-05, + "loss": 0.9416, + "step": 244330 + }, + { + "epoch": 21.15, + "learning_rate": 1.47721690889653e-05, + "loss": 0.9397, + "step": 244340 + }, + { + "epoch": 21.15, + "learning_rate": 1.4770724378051953e-05, + "loss": 0.9176, + "step": 244350 + }, + { + "epoch": 21.15, + "learning_rate": 1.4769279667138604e-05, + "loss": 0.9004, + "step": 244360 + }, + { + "epoch": 21.15, + "learning_rate": 1.476783495622526e-05, + "loss": 0.9307, + "step": 244370 + }, + { + "epoch": 21.15, + "learning_rate": 1.4766390245311915e-05, + "loss": 0.9221, + "step": 244380 + }, + { + "epoch": 21.15, + "learning_rate": 1.4764945534398566e-05, + "loss": 0.9421, + "step": 244390 + }, + { + "epoch": 21.15, + "learning_rate": 1.4763500823485221e-05, + "loss": 0.8807, + "step": 244400 + }, + { + "epoch": 21.16, + "learning_rate": 1.4762056112571877e-05, + "loss": 0.8999, + "step": 244410 + }, + { + "epoch": 21.16, + "learning_rate": 1.4760611401658528e-05, + "loss": 1.0058, + "step": 244420 + }, + { + "epoch": 21.16, + "learning_rate": 1.4759166690745183e-05, + "loss": 0.9255, + "step": 244430 + }, + { + "epoch": 21.16, + "learning_rate": 1.4757721979831837e-05, + "loss": 0.9583, + "step": 244440 + }, + { + "epoch": 21.16, + "learning_rate": 1.4756277268918488e-05, + "loss": 0.9249, + "step": 244450 + }, + { + "epoch": 21.16, + "learning_rate": 1.4754832558005144e-05, + "loss": 0.9135, + "step": 244460 + }, + { + "epoch": 21.16, + "learning_rate": 1.4753387847091799e-05, + "loss": 0.9412, + "step": 244470 + }, + { + "epoch": 21.16, + "learning_rate": 1.475194313617845e-05, + "loss": 0.9299, + "step": 244480 + }, + { + "epoch": 21.16, + "learning_rate": 1.4750498425265105e-05, + "loss": 0.9492, + "step": 244490 + }, + { + "epoch": 21.16, + "learning_rate": 1.474905371435176e-05, + "loss": 0.9859, + "step": 244500 + }, + { + "epoch": 21.16, + "learning_rate": 1.4747609003438412e-05, + "loss": 0.988, + "step": 244510 + }, + { + "epoch": 21.17, + "learning_rate": 1.4746164292525067e-05, + "loss": 0.9579, + "step": 244520 + }, + { + "epoch": 21.17, + "learning_rate": 1.474471958161172e-05, + "loss": 0.8742, + "step": 244530 + }, + { + "epoch": 21.17, + "learning_rate": 1.4743274870698372e-05, + "loss": 0.9219, + "step": 244540 + }, + { + "epoch": 21.17, + "learning_rate": 1.4741830159785027e-05, + "loss": 0.9955, + "step": 244550 + }, + { + "epoch": 21.17, + "learning_rate": 1.4740385448871683e-05, + "loss": 0.9623, + "step": 244560 + }, + { + "epoch": 21.17, + "learning_rate": 1.4738940737958334e-05, + "loss": 0.879, + "step": 244570 + }, + { + "epoch": 21.17, + "learning_rate": 1.473749602704499e-05, + "loss": 0.9741, + "step": 244580 + }, + { + "epoch": 21.17, + "learning_rate": 1.4736051316131644e-05, + "loss": 0.9129, + "step": 244590 + }, + { + "epoch": 21.17, + "learning_rate": 1.4734606605218296e-05, + "loss": 0.9075, + "step": 244600 + }, + { + "epoch": 21.17, + "learning_rate": 1.4733161894304951e-05, + "loss": 0.9327, + "step": 244610 + }, + { + "epoch": 21.17, + "learning_rate": 1.4731717183391605e-05, + "loss": 0.9046, + "step": 244620 + }, + { + "epoch": 21.17, + "learning_rate": 1.4730272472478256e-05, + "loss": 0.9629, + "step": 244630 + }, + { + "epoch": 21.18, + "learning_rate": 1.4728827761564911e-05, + "loss": 0.8935, + "step": 244640 + }, + { + "epoch": 21.18, + "learning_rate": 1.4727383050651566e-05, + "loss": 0.9297, + "step": 244650 + }, + { + "epoch": 21.18, + "learning_rate": 1.4725938339738218e-05, + "loss": 0.8901, + "step": 244660 + }, + { + "epoch": 21.18, + "learning_rate": 1.4724493628824873e-05, + "loss": 0.9459, + "step": 244670 + }, + { + "epoch": 21.18, + "learning_rate": 1.4723048917911528e-05, + "loss": 0.959, + "step": 244680 + }, + { + "epoch": 21.18, + "learning_rate": 1.472160420699818e-05, + "loss": 0.9779, + "step": 244690 + }, + { + "epoch": 21.18, + "learning_rate": 1.4720159496084835e-05, + "loss": 0.9415, + "step": 244700 + }, + { + "epoch": 21.18, + "learning_rate": 1.4718714785171489e-05, + "loss": 0.9753, + "step": 244710 + }, + { + "epoch": 21.18, + "learning_rate": 1.471727007425814e-05, + "loss": 0.961, + "step": 244720 + }, + { + "epoch": 21.18, + "learning_rate": 1.4715825363344795e-05, + "loss": 0.9264, + "step": 244730 + }, + { + "epoch": 21.18, + "learning_rate": 1.471438065243145e-05, + "loss": 0.9062, + "step": 244740 + }, + { + "epoch": 21.18, + "learning_rate": 1.4712935941518102e-05, + "loss": 0.9183, + "step": 244750 + }, + { + "epoch": 21.19, + "learning_rate": 1.4711491230604757e-05, + "loss": 0.9018, + "step": 244760 + }, + { + "epoch": 21.19, + "learning_rate": 1.4710046519691412e-05, + "loss": 0.9691, + "step": 244770 + }, + { + "epoch": 21.19, + "learning_rate": 1.4708601808778064e-05, + "loss": 0.9108, + "step": 244780 + }, + { + "epoch": 21.19, + "learning_rate": 1.4707157097864719e-05, + "loss": 0.9767, + "step": 244790 + }, + { + "epoch": 21.19, + "learning_rate": 1.4705712386951372e-05, + "loss": 0.9061, + "step": 244800 + }, + { + "epoch": 21.19, + "learning_rate": 1.4704267676038024e-05, + "loss": 0.9263, + "step": 244810 + }, + { + "epoch": 21.19, + "learning_rate": 1.470282296512468e-05, + "loss": 0.9339, + "step": 244820 + }, + { + "epoch": 21.19, + "learning_rate": 1.4701378254211334e-05, + "loss": 0.9757, + "step": 244830 + }, + { + "epoch": 21.19, + "learning_rate": 1.4699933543297986e-05, + "loss": 0.9001, + "step": 244840 + }, + { + "epoch": 21.19, + "learning_rate": 1.4698488832384641e-05, + "loss": 0.9761, + "step": 244850 + }, + { + "epoch": 21.19, + "learning_rate": 1.4697044121471296e-05, + "loss": 0.9352, + "step": 244860 + }, + { + "epoch": 21.2, + "learning_rate": 1.4695599410557948e-05, + "loss": 0.9689, + "step": 244870 + }, + { + "epoch": 21.2, + "learning_rate": 1.4694154699644603e-05, + "loss": 0.9524, + "step": 244880 + }, + { + "epoch": 21.2, + "learning_rate": 1.4692709988731256e-05, + "loss": 0.9515, + "step": 244890 + }, + { + "epoch": 21.2, + "learning_rate": 1.4691265277817908e-05, + "loss": 0.9629, + "step": 244900 + }, + { + "epoch": 21.2, + "learning_rate": 1.4689820566904563e-05, + "loss": 0.848, + "step": 244910 + }, + { + "epoch": 21.2, + "learning_rate": 1.4688375855991218e-05, + "loss": 0.9368, + "step": 244920 + }, + { + "epoch": 21.2, + "learning_rate": 1.468693114507787e-05, + "loss": 0.9537, + "step": 244930 + }, + { + "epoch": 21.2, + "learning_rate": 1.4685486434164525e-05, + "loss": 0.9418, + "step": 244940 + }, + { + "epoch": 21.2, + "learning_rate": 1.4684041723251177e-05, + "loss": 0.9308, + "step": 244950 + }, + { + "epoch": 21.2, + "learning_rate": 1.4682597012337832e-05, + "loss": 0.909, + "step": 244960 + }, + { + "epoch": 21.2, + "learning_rate": 1.4681152301424487e-05, + "loss": 0.9506, + "step": 244970 + }, + { + "epoch": 21.2, + "learning_rate": 1.4679707590511139e-05, + "loss": 0.9403, + "step": 244980 + }, + { + "epoch": 21.21, + "learning_rate": 1.4678262879597792e-05, + "loss": 0.9224, + "step": 244990 + }, + { + "epoch": 21.21, + "learning_rate": 1.4676818168684447e-05, + "loss": 0.9544, + "step": 245000 + }, + { + "epoch": 21.21, + "learning_rate": 1.4675373457771099e-05, + "loss": 0.8904, + "step": 245010 + }, + { + "epoch": 21.21, + "learning_rate": 1.4673928746857754e-05, + "loss": 0.9766, + "step": 245020 + }, + { + "epoch": 21.21, + "learning_rate": 1.4672484035944409e-05, + "loss": 0.9253, + "step": 245030 + }, + { + "epoch": 21.21, + "learning_rate": 1.467103932503106e-05, + "loss": 0.9802, + "step": 245040 + }, + { + "epoch": 21.21, + "learning_rate": 1.4669594614117716e-05, + "loss": 0.9391, + "step": 245050 + }, + { + "epoch": 21.21, + "learning_rate": 1.4668149903204371e-05, + "loss": 1.0154, + "step": 245060 + }, + { + "epoch": 21.21, + "learning_rate": 1.4666705192291023e-05, + "loss": 0.8899, + "step": 245070 + }, + { + "epoch": 21.21, + "learning_rate": 1.4665260481377676e-05, + "loss": 0.9599, + "step": 245080 + }, + { + "epoch": 21.21, + "learning_rate": 1.4663815770464331e-05, + "loss": 0.9133, + "step": 245090 + }, + { + "epoch": 21.22, + "learning_rate": 1.4662371059550983e-05, + "loss": 0.9129, + "step": 245100 + }, + { + "epoch": 21.22, + "learning_rate": 1.4660926348637638e-05, + "loss": 0.9035, + "step": 245110 + }, + { + "epoch": 21.22, + "learning_rate": 1.4659481637724293e-05, + "loss": 0.9849, + "step": 245120 + }, + { + "epoch": 21.22, + "learning_rate": 1.4658036926810945e-05, + "loss": 0.9113, + "step": 245130 + }, + { + "epoch": 21.22, + "learning_rate": 1.46565922158976e-05, + "loss": 0.9812, + "step": 245140 + }, + { + "epoch": 21.22, + "learning_rate": 1.4655147504984255e-05, + "loss": 0.9248, + "step": 245150 + }, + { + "epoch": 21.22, + "learning_rate": 1.4653702794070907e-05, + "loss": 0.9098, + "step": 245160 + }, + { + "epoch": 21.22, + "learning_rate": 1.465225808315756e-05, + "loss": 0.9501, + "step": 245170 + }, + { + "epoch": 21.22, + "learning_rate": 1.4650813372244215e-05, + "loss": 0.958, + "step": 245180 + }, + { + "epoch": 21.22, + "learning_rate": 1.4649368661330867e-05, + "loss": 0.9479, + "step": 245190 + }, + { + "epoch": 21.22, + "learning_rate": 1.4647923950417522e-05, + "loss": 0.9349, + "step": 245200 + }, + { + "epoch": 21.22, + "learning_rate": 1.4646479239504177e-05, + "loss": 0.8908, + "step": 245210 + }, + { + "epoch": 21.23, + "learning_rate": 1.4645034528590829e-05, + "loss": 0.9364, + "step": 245220 + }, + { + "epoch": 21.23, + "learning_rate": 1.4643589817677484e-05, + "loss": 0.9127, + "step": 245230 + }, + { + "epoch": 21.23, + "learning_rate": 1.4642145106764139e-05, + "loss": 0.9871, + "step": 245240 + }, + { + "epoch": 21.23, + "learning_rate": 1.464070039585079e-05, + "loss": 0.9278, + "step": 245250 + }, + { + "epoch": 21.23, + "learning_rate": 1.4639255684937444e-05, + "loss": 0.9629, + "step": 245260 + }, + { + "epoch": 21.23, + "learning_rate": 1.4637810974024099e-05, + "loss": 0.9982, + "step": 245270 + }, + { + "epoch": 21.23, + "learning_rate": 1.463636626311075e-05, + "loss": 0.9204, + "step": 245280 + }, + { + "epoch": 21.23, + "learning_rate": 1.4634921552197406e-05, + "loss": 0.9537, + "step": 245290 + }, + { + "epoch": 21.23, + "learning_rate": 1.4633476841284061e-05, + "loss": 0.9308, + "step": 245300 + }, + { + "epoch": 21.23, + "learning_rate": 1.4632032130370713e-05, + "loss": 0.8869, + "step": 245310 + }, + { + "epoch": 21.23, + "learning_rate": 1.4630587419457368e-05, + "loss": 0.9956, + "step": 245320 + }, + { + "epoch": 21.24, + "learning_rate": 1.4629142708544023e-05, + "loss": 0.871, + "step": 245330 + }, + { + "epoch": 21.24, + "learning_rate": 1.4627697997630674e-05, + "loss": 0.9188, + "step": 245340 + }, + { + "epoch": 21.24, + "learning_rate": 1.4626253286717328e-05, + "loss": 0.9478, + "step": 245350 + }, + { + "epoch": 21.24, + "learning_rate": 1.4624808575803983e-05, + "loss": 0.9092, + "step": 245360 + }, + { + "epoch": 21.24, + "learning_rate": 1.4623363864890635e-05, + "loss": 0.9443, + "step": 245370 + }, + { + "epoch": 21.24, + "learning_rate": 1.462191915397729e-05, + "loss": 0.9438, + "step": 245380 + }, + { + "epoch": 21.24, + "learning_rate": 1.4620474443063945e-05, + "loss": 0.9463, + "step": 245390 + }, + { + "epoch": 21.24, + "learning_rate": 1.4619029732150597e-05, + "loss": 0.9904, + "step": 245400 + }, + { + "epoch": 21.24, + "learning_rate": 1.4617585021237252e-05, + "loss": 0.9221, + "step": 245410 + }, + { + "epoch": 21.24, + "learning_rate": 1.4616140310323907e-05, + "loss": 0.9493, + "step": 245420 + }, + { + "epoch": 21.24, + "learning_rate": 1.4614695599410558e-05, + "loss": 0.9451, + "step": 245430 + }, + { + "epoch": 21.24, + "learning_rate": 1.4613250888497212e-05, + "loss": 0.9344, + "step": 245440 + }, + { + "epoch": 21.25, + "learning_rate": 1.4611806177583867e-05, + "loss": 0.9396, + "step": 245450 + }, + { + "epoch": 21.25, + "learning_rate": 1.4610361466670519e-05, + "loss": 0.9219, + "step": 245460 + }, + { + "epoch": 21.25, + "learning_rate": 1.4608916755757174e-05, + "loss": 0.9968, + "step": 245470 + }, + { + "epoch": 21.25, + "learning_rate": 1.4607472044843829e-05, + "loss": 0.9531, + "step": 245480 + }, + { + "epoch": 21.25, + "learning_rate": 1.460602733393048e-05, + "loss": 0.8923, + "step": 245490 + }, + { + "epoch": 21.25, + "learning_rate": 1.4604582623017136e-05, + "loss": 0.975, + "step": 245500 + }, + { + "epoch": 21.25, + "learning_rate": 1.4603137912103789e-05, + "loss": 0.9574, + "step": 245510 + }, + { + "epoch": 21.25, + "learning_rate": 1.4601693201190442e-05, + "loss": 0.8907, + "step": 245520 + }, + { + "epoch": 21.25, + "learning_rate": 1.4600248490277096e-05, + "loss": 0.8924, + "step": 245530 + }, + { + "epoch": 21.25, + "learning_rate": 1.459880377936375e-05, + "loss": 0.9125, + "step": 245540 + }, + { + "epoch": 21.25, + "learning_rate": 1.4597359068450403e-05, + "loss": 0.8832, + "step": 245550 + }, + { + "epoch": 21.26, + "learning_rate": 1.4595914357537058e-05, + "loss": 0.9426, + "step": 245560 + }, + { + "epoch": 21.26, + "learning_rate": 1.4594469646623713e-05, + "loss": 0.9483, + "step": 245570 + }, + { + "epoch": 21.26, + "learning_rate": 1.4593024935710364e-05, + "loss": 0.9009, + "step": 245580 + }, + { + "epoch": 21.26, + "learning_rate": 1.459158022479702e-05, + "loss": 0.9756, + "step": 245590 + }, + { + "epoch": 21.26, + "learning_rate": 1.4590135513883673e-05, + "loss": 0.9895, + "step": 245600 + }, + { + "epoch": 21.26, + "learning_rate": 1.4588690802970326e-05, + "loss": 0.9031, + "step": 245610 + }, + { + "epoch": 21.26, + "learning_rate": 1.458724609205698e-05, + "loss": 0.9077, + "step": 245620 + }, + { + "epoch": 21.26, + "learning_rate": 1.4585801381143635e-05, + "loss": 0.9468, + "step": 245630 + }, + { + "epoch": 21.26, + "learning_rate": 1.4584356670230286e-05, + "loss": 0.8995, + "step": 245640 + }, + { + "epoch": 21.26, + "learning_rate": 1.4582911959316942e-05, + "loss": 0.9108, + "step": 245650 + }, + { + "epoch": 21.26, + "learning_rate": 1.4581467248403597e-05, + "loss": 0.9521, + "step": 245660 + }, + { + "epoch": 21.26, + "learning_rate": 1.4580022537490248e-05, + "loss": 0.8977, + "step": 245670 + }, + { + "epoch": 21.27, + "learning_rate": 1.4578577826576903e-05, + "loss": 0.9272, + "step": 245680 + }, + { + "epoch": 21.27, + "learning_rate": 1.4577133115663557e-05, + "loss": 0.9259, + "step": 245690 + }, + { + "epoch": 21.27, + "learning_rate": 1.457568840475021e-05, + "loss": 0.8921, + "step": 245700 + }, + { + "epoch": 21.27, + "learning_rate": 1.4574243693836864e-05, + "loss": 0.9276, + "step": 245710 + }, + { + "epoch": 21.27, + "learning_rate": 1.4572798982923519e-05, + "loss": 0.961, + "step": 245720 + }, + { + "epoch": 21.27, + "learning_rate": 1.457135427201017e-05, + "loss": 0.9779, + "step": 245730 + }, + { + "epoch": 21.27, + "learning_rate": 1.4569909561096825e-05, + "loss": 0.9332, + "step": 245740 + }, + { + "epoch": 21.27, + "learning_rate": 1.456846485018348e-05, + "loss": 0.9004, + "step": 245750 + }, + { + "epoch": 21.27, + "learning_rate": 1.4567020139270132e-05, + "loss": 0.9295, + "step": 245760 + }, + { + "epoch": 21.27, + "learning_rate": 1.4565575428356787e-05, + "loss": 0.9227, + "step": 245770 + }, + { + "epoch": 21.27, + "learning_rate": 1.456413071744344e-05, + "loss": 0.943, + "step": 245780 + }, + { + "epoch": 21.27, + "learning_rate": 1.4562686006530094e-05, + "loss": 0.9243, + "step": 245790 + }, + { + "epoch": 21.28, + "learning_rate": 1.4561241295616748e-05, + "loss": 0.9581, + "step": 245800 + }, + { + "epoch": 21.28, + "learning_rate": 1.4559796584703403e-05, + "loss": 0.9036, + "step": 245810 + }, + { + "epoch": 21.28, + "learning_rate": 1.4558351873790054e-05, + "loss": 0.9078, + "step": 245820 + }, + { + "epoch": 21.28, + "learning_rate": 1.455690716287671e-05, + "loss": 0.9739, + "step": 245830 + }, + { + "epoch": 21.28, + "learning_rate": 1.4555462451963361e-05, + "loss": 0.9541, + "step": 245840 + }, + { + "epoch": 21.28, + "learning_rate": 1.4554017741050016e-05, + "loss": 0.9105, + "step": 245850 + }, + { + "epoch": 21.28, + "learning_rate": 1.4552573030136671e-05, + "loss": 0.8992, + "step": 245860 + }, + { + "epoch": 21.28, + "learning_rate": 1.4551128319223323e-05, + "loss": 0.9633, + "step": 245870 + }, + { + "epoch": 21.28, + "learning_rate": 1.4549683608309978e-05, + "loss": 0.895, + "step": 245880 + }, + { + "epoch": 21.28, + "learning_rate": 1.4548238897396632e-05, + "loss": 0.9668, + "step": 245890 + }, + { + "epoch": 21.28, + "learning_rate": 1.4546794186483285e-05, + "loss": 0.9647, + "step": 245900 + }, + { + "epoch": 21.29, + "learning_rate": 1.4545349475569938e-05, + "loss": 0.8836, + "step": 245910 + }, + { + "epoch": 21.29, + "learning_rate": 1.4543904764656593e-05, + "loss": 0.9023, + "step": 245920 + }, + { + "epoch": 21.29, + "learning_rate": 1.4542460053743245e-05, + "loss": 0.9391, + "step": 245930 + }, + { + "epoch": 21.29, + "learning_rate": 1.45410153428299e-05, + "loss": 0.9435, + "step": 245940 + }, + { + "epoch": 21.29, + "learning_rate": 1.4539570631916555e-05, + "loss": 0.9498, + "step": 245950 + }, + { + "epoch": 21.29, + "learning_rate": 1.4538125921003207e-05, + "loss": 0.9633, + "step": 245960 + }, + { + "epoch": 21.29, + "learning_rate": 1.4536681210089862e-05, + "loss": 0.9264, + "step": 245970 + }, + { + "epoch": 21.29, + "learning_rate": 1.4535236499176515e-05, + "loss": 0.9472, + "step": 245980 + }, + { + "epoch": 21.29, + "learning_rate": 1.4533791788263169e-05, + "loss": 0.9083, + "step": 245990 + }, + { + "epoch": 21.29, + "learning_rate": 1.4532347077349822e-05, + "loss": 0.883, + "step": 246000 + }, + { + "epoch": 21.29, + "learning_rate": 1.4530902366436477e-05, + "loss": 1.0026, + "step": 246010 + }, + { + "epoch": 21.29, + "learning_rate": 1.4529457655523129e-05, + "loss": 0.9017, + "step": 246020 + }, + { + "epoch": 21.3, + "learning_rate": 1.4528012944609784e-05, + "loss": 0.9604, + "step": 246030 + }, + { + "epoch": 21.3, + "learning_rate": 1.452656823369644e-05, + "loss": 0.95, + "step": 246040 + }, + { + "epoch": 21.3, + "learning_rate": 1.4525123522783091e-05, + "loss": 0.9657, + "step": 246050 + }, + { + "epoch": 21.3, + "learning_rate": 1.4523678811869746e-05, + "loss": 0.936, + "step": 246060 + }, + { + "epoch": 21.3, + "learning_rate": 1.45222341009564e-05, + "loss": 0.9227, + "step": 246070 + }, + { + "epoch": 21.3, + "learning_rate": 1.4520789390043053e-05, + "loss": 0.9318, + "step": 246080 + }, + { + "epoch": 21.3, + "learning_rate": 1.4519344679129706e-05, + "loss": 0.9543, + "step": 246090 + }, + { + "epoch": 21.3, + "learning_rate": 1.4517899968216361e-05, + "loss": 0.9413, + "step": 246100 + }, + { + "epoch": 21.3, + "learning_rate": 1.4516455257303013e-05, + "loss": 0.9046, + "step": 246110 + }, + { + "epoch": 21.3, + "learning_rate": 1.4515010546389668e-05, + "loss": 0.9336, + "step": 246120 + }, + { + "epoch": 21.3, + "learning_rate": 1.4513565835476323e-05, + "loss": 0.8767, + "step": 246130 + }, + { + "epoch": 21.31, + "learning_rate": 1.4512121124562975e-05, + "loss": 0.9426, + "step": 246140 + }, + { + "epoch": 21.31, + "learning_rate": 1.451067641364963e-05, + "loss": 0.9494, + "step": 246150 + }, + { + "epoch": 21.31, + "learning_rate": 1.4509231702736283e-05, + "loss": 0.9717, + "step": 246160 + }, + { + "epoch": 21.31, + "learning_rate": 1.4507786991822935e-05, + "loss": 0.9492, + "step": 246170 + }, + { + "epoch": 21.31, + "learning_rate": 1.450634228090959e-05, + "loss": 0.9029, + "step": 246180 + }, + { + "epoch": 21.31, + "learning_rate": 1.4504897569996245e-05, + "loss": 0.948, + "step": 246190 + }, + { + "epoch": 21.31, + "learning_rate": 1.4503452859082897e-05, + "loss": 0.9698, + "step": 246200 + }, + { + "epoch": 21.31, + "learning_rate": 1.4502008148169552e-05, + "loss": 0.9474, + "step": 246210 + }, + { + "epoch": 21.31, + "learning_rate": 1.4500563437256207e-05, + "loss": 0.8923, + "step": 246220 + }, + { + "epoch": 21.31, + "learning_rate": 1.4499118726342859e-05, + "loss": 0.9296, + "step": 246230 + }, + { + "epoch": 21.31, + "learning_rate": 1.4497674015429514e-05, + "loss": 0.973, + "step": 246240 + }, + { + "epoch": 21.31, + "learning_rate": 1.4496229304516167e-05, + "loss": 0.8844, + "step": 246250 + }, + { + "epoch": 21.32, + "learning_rate": 1.4494784593602819e-05, + "loss": 0.9366, + "step": 246260 + }, + { + "epoch": 21.32, + "learning_rate": 1.4493339882689474e-05, + "loss": 0.9205, + "step": 246270 + }, + { + "epoch": 21.32, + "learning_rate": 1.449189517177613e-05, + "loss": 0.9448, + "step": 246280 + }, + { + "epoch": 21.32, + "learning_rate": 1.4490450460862781e-05, + "loss": 0.9321, + "step": 246290 + }, + { + "epoch": 21.32, + "learning_rate": 1.4489005749949436e-05, + "loss": 0.9055, + "step": 246300 + }, + { + "epoch": 21.32, + "learning_rate": 1.4487561039036091e-05, + "loss": 0.9587, + "step": 246310 + }, + { + "epoch": 21.32, + "learning_rate": 1.4486116328122743e-05, + "loss": 0.9526, + "step": 246320 + }, + { + "epoch": 21.32, + "learning_rate": 1.4484671617209398e-05, + "loss": 0.9196, + "step": 246330 + }, + { + "epoch": 21.32, + "learning_rate": 1.4483226906296051e-05, + "loss": 0.926, + "step": 246340 + }, + { + "epoch": 21.32, + "learning_rate": 1.4481782195382703e-05, + "loss": 0.9505, + "step": 246350 + }, + { + "epoch": 21.32, + "learning_rate": 1.4480337484469358e-05, + "loss": 0.9317, + "step": 246360 + }, + { + "epoch": 21.33, + "learning_rate": 1.4478892773556013e-05, + "loss": 0.8378, + "step": 246370 + }, + { + "epoch": 21.33, + "learning_rate": 1.4477448062642665e-05, + "loss": 0.8853, + "step": 246380 + }, + { + "epoch": 21.33, + "learning_rate": 1.447600335172932e-05, + "loss": 0.9325, + "step": 246390 + }, + { + "epoch": 21.33, + "learning_rate": 1.4474558640815975e-05, + "loss": 0.9257, + "step": 246400 + }, + { + "epoch": 21.33, + "learning_rate": 1.4473113929902627e-05, + "loss": 0.9121, + "step": 246410 + }, + { + "epoch": 21.33, + "learning_rate": 1.4471669218989282e-05, + "loss": 0.9801, + "step": 246420 + }, + { + "epoch": 21.33, + "learning_rate": 1.4470224508075935e-05, + "loss": 1.0149, + "step": 246430 + }, + { + "epoch": 21.33, + "learning_rate": 1.4468779797162587e-05, + "loss": 0.9147, + "step": 246440 + }, + { + "epoch": 21.33, + "learning_rate": 1.4467335086249242e-05, + "loss": 0.9697, + "step": 246450 + }, + { + "epoch": 21.33, + "learning_rate": 1.4465890375335897e-05, + "loss": 1.0013, + "step": 246460 + }, + { + "epoch": 21.33, + "learning_rate": 1.4464445664422549e-05, + "loss": 0.9649, + "step": 246470 + }, + { + "epoch": 21.33, + "learning_rate": 1.4463000953509204e-05, + "loss": 0.9465, + "step": 246480 + }, + { + "epoch": 21.34, + "learning_rate": 1.4461556242595859e-05, + "loss": 0.8845, + "step": 246490 + }, + { + "epoch": 21.34, + "learning_rate": 1.446011153168251e-05, + "loss": 0.967, + "step": 246500 + }, + { + "epoch": 21.34, + "learning_rate": 1.4458666820769166e-05, + "loss": 0.9464, + "step": 246510 + }, + { + "epoch": 21.34, + "learning_rate": 1.4457222109855819e-05, + "loss": 0.8581, + "step": 246520 + }, + { + "epoch": 21.34, + "learning_rate": 1.445577739894247e-05, + "loss": 0.883, + "step": 246530 + }, + { + "epoch": 21.34, + "learning_rate": 1.4454332688029126e-05, + "loss": 0.9774, + "step": 246540 + }, + { + "epoch": 21.34, + "learning_rate": 1.4452887977115781e-05, + "loss": 0.9592, + "step": 246550 + }, + { + "epoch": 21.34, + "learning_rate": 1.4451443266202433e-05, + "loss": 0.9264, + "step": 246560 + }, + { + "epoch": 21.34, + "learning_rate": 1.4449998555289088e-05, + "loss": 0.9488, + "step": 246570 + }, + { + "epoch": 21.34, + "learning_rate": 1.4448553844375743e-05, + "loss": 1.0144, + "step": 246580 + }, + { + "epoch": 21.34, + "learning_rate": 1.4447109133462395e-05, + "loss": 0.9678, + "step": 246590 + }, + { + "epoch": 21.35, + "learning_rate": 1.444566442254905e-05, + "loss": 0.9093, + "step": 246600 + }, + { + "epoch": 21.35, + "learning_rate": 1.4444219711635703e-05, + "loss": 0.8904, + "step": 246610 + }, + { + "epoch": 21.35, + "learning_rate": 1.4442775000722355e-05, + "loss": 0.8909, + "step": 246620 + }, + { + "epoch": 21.35, + "learning_rate": 1.444133028980901e-05, + "loss": 0.9459, + "step": 246630 + }, + { + "epoch": 21.35, + "learning_rate": 1.4439885578895665e-05, + "loss": 0.8913, + "step": 246640 + }, + { + "epoch": 21.35, + "learning_rate": 1.4438440867982317e-05, + "loss": 0.9091, + "step": 246650 + }, + { + "epoch": 21.35, + "learning_rate": 1.4436996157068972e-05, + "loss": 0.917, + "step": 246660 + }, + { + "epoch": 21.35, + "learning_rate": 1.4435551446155627e-05, + "loss": 0.9319, + "step": 246670 + }, + { + "epoch": 21.35, + "learning_rate": 1.4434106735242279e-05, + "loss": 0.9378, + "step": 246680 + }, + { + "epoch": 21.35, + "learning_rate": 1.4432662024328934e-05, + "loss": 0.8935, + "step": 246690 + }, + { + "epoch": 21.35, + "learning_rate": 1.4431217313415587e-05, + "loss": 0.9364, + "step": 246700 + }, + { + "epoch": 21.35, + "learning_rate": 1.4429772602502239e-05, + "loss": 0.8972, + "step": 246710 + }, + { + "epoch": 21.36, + "learning_rate": 1.4428327891588894e-05, + "loss": 0.9311, + "step": 246720 + }, + { + "epoch": 21.36, + "learning_rate": 1.4426883180675549e-05, + "loss": 0.9219, + "step": 246730 + }, + { + "epoch": 21.36, + "learning_rate": 1.44254384697622e-05, + "loss": 0.9088, + "step": 246740 + }, + { + "epoch": 21.36, + "learning_rate": 1.4423993758848856e-05, + "loss": 0.9685, + "step": 246750 + }, + { + "epoch": 21.36, + "learning_rate": 1.4422549047935507e-05, + "loss": 0.9204, + "step": 246760 + }, + { + "epoch": 21.36, + "learning_rate": 1.4421104337022162e-05, + "loss": 0.9516, + "step": 246770 + }, + { + "epoch": 21.36, + "learning_rate": 1.4419659626108818e-05, + "loss": 1.0409, + "step": 246780 + }, + { + "epoch": 21.36, + "learning_rate": 1.441821491519547e-05, + "loss": 0.9867, + "step": 246790 + }, + { + "epoch": 21.36, + "learning_rate": 1.4416770204282123e-05, + "loss": 0.951, + "step": 246800 + }, + { + "epoch": 21.36, + "learning_rate": 1.4415325493368778e-05, + "loss": 0.8945, + "step": 246810 + }, + { + "epoch": 21.36, + "learning_rate": 1.441388078245543e-05, + "loss": 0.9948, + "step": 246820 + }, + { + "epoch": 21.36, + "learning_rate": 1.4412436071542085e-05, + "loss": 0.9548, + "step": 246830 + }, + { + "epoch": 21.37, + "learning_rate": 1.441099136062874e-05, + "loss": 1.0011, + "step": 246840 + }, + { + "epoch": 21.37, + "learning_rate": 1.4409546649715391e-05, + "loss": 0.9491, + "step": 246850 + }, + { + "epoch": 21.37, + "learning_rate": 1.4408101938802046e-05, + "loss": 0.9963, + "step": 246860 + }, + { + "epoch": 21.37, + "learning_rate": 1.4406657227888701e-05, + "loss": 0.9322, + "step": 246870 + }, + { + "epoch": 21.37, + "learning_rate": 1.4405212516975353e-05, + "loss": 0.9002, + "step": 246880 + }, + { + "epoch": 21.37, + "learning_rate": 1.4403767806062007e-05, + "loss": 0.9674, + "step": 246890 + }, + { + "epoch": 21.37, + "learning_rate": 1.4402323095148662e-05, + "loss": 0.8903, + "step": 246900 + }, + { + "epoch": 21.37, + "learning_rate": 1.4400878384235313e-05, + "loss": 0.943, + "step": 246910 + }, + { + "epoch": 21.37, + "learning_rate": 1.4399433673321968e-05, + "loss": 0.8741, + "step": 246920 + }, + { + "epoch": 21.37, + "learning_rate": 1.4397988962408624e-05, + "loss": 0.9825, + "step": 246930 + }, + { + "epoch": 21.37, + "learning_rate": 1.4396544251495275e-05, + "loss": 0.9259, + "step": 246940 + }, + { + "epoch": 21.38, + "learning_rate": 1.439509954058193e-05, + "loss": 0.9775, + "step": 246950 + }, + { + "epoch": 21.38, + "learning_rate": 1.4393654829668585e-05, + "loss": 0.9389, + "step": 246960 + }, + { + "epoch": 21.38, + "learning_rate": 1.4392210118755237e-05, + "loss": 0.9808, + "step": 246970 + }, + { + "epoch": 21.38, + "learning_rate": 1.439076540784189e-05, + "loss": 0.9553, + "step": 246980 + }, + { + "epoch": 21.38, + "learning_rate": 1.4389320696928546e-05, + "loss": 0.8744, + "step": 246990 + }, + { + "epoch": 21.38, + "learning_rate": 1.4387875986015197e-05, + "loss": 0.9117, + "step": 247000 + }, + { + "epoch": 21.38, + "learning_rate": 1.4386431275101852e-05, + "loss": 0.9167, + "step": 247010 + }, + { + "epoch": 21.38, + "learning_rate": 1.4384986564188507e-05, + "loss": 0.9027, + "step": 247020 + }, + { + "epoch": 21.38, + "learning_rate": 1.438354185327516e-05, + "loss": 0.9568, + "step": 247030 + }, + { + "epoch": 21.38, + "learning_rate": 1.4382097142361814e-05, + "loss": 0.9363, + "step": 247040 + }, + { + "epoch": 21.38, + "learning_rate": 1.438065243144847e-05, + "loss": 0.9551, + "step": 247050 + }, + { + "epoch": 21.38, + "learning_rate": 1.4379207720535121e-05, + "loss": 0.9194, + "step": 247060 + }, + { + "epoch": 21.39, + "learning_rate": 1.4377763009621774e-05, + "loss": 0.9397, + "step": 247070 + }, + { + "epoch": 21.39, + "learning_rate": 1.437631829870843e-05, + "loss": 0.9195, + "step": 247080 + }, + { + "epoch": 21.39, + "learning_rate": 1.4374873587795081e-05, + "loss": 0.9384, + "step": 247090 + }, + { + "epoch": 21.39, + "learning_rate": 1.4373428876881736e-05, + "loss": 0.9906, + "step": 247100 + }, + { + "epoch": 21.39, + "learning_rate": 1.4371984165968391e-05, + "loss": 0.9374, + "step": 247110 + }, + { + "epoch": 21.39, + "learning_rate": 1.4370539455055043e-05, + "loss": 0.8897, + "step": 247120 + }, + { + "epoch": 21.39, + "learning_rate": 1.4369094744141698e-05, + "loss": 0.9506, + "step": 247130 + }, + { + "epoch": 21.39, + "learning_rate": 1.4367650033228353e-05, + "loss": 0.9392, + "step": 247140 + }, + { + "epoch": 21.39, + "learning_rate": 1.4366205322315005e-05, + "loss": 0.9075, + "step": 247150 + }, + { + "epoch": 21.39, + "learning_rate": 1.4364760611401658e-05, + "loss": 0.9547, + "step": 247160 + }, + { + "epoch": 21.39, + "learning_rate": 1.4363315900488313e-05, + "loss": 0.8958, + "step": 247170 + }, + { + "epoch": 21.4, + "learning_rate": 1.4361871189574965e-05, + "loss": 0.9503, + "step": 247180 + }, + { + "epoch": 21.4, + "learning_rate": 1.436042647866162e-05, + "loss": 0.879, + "step": 247190 + }, + { + "epoch": 21.4, + "learning_rate": 1.4358981767748275e-05, + "loss": 0.9314, + "step": 247200 + }, + { + "epoch": 21.4, + "learning_rate": 1.4357537056834927e-05, + "loss": 0.9859, + "step": 247210 + }, + { + "epoch": 21.4, + "learning_rate": 1.4356092345921582e-05, + "loss": 0.9266, + "step": 247220 + }, + { + "epoch": 21.4, + "learning_rate": 1.4354647635008237e-05, + "loss": 0.9471, + "step": 247230 + }, + { + "epoch": 21.4, + "learning_rate": 1.4353202924094889e-05, + "loss": 0.8946, + "step": 247240 + }, + { + "epoch": 21.4, + "learning_rate": 1.4351758213181542e-05, + "loss": 0.9686, + "step": 247250 + }, + { + "epoch": 21.4, + "learning_rate": 1.4350313502268197e-05, + "loss": 0.9267, + "step": 247260 + }, + { + "epoch": 21.4, + "learning_rate": 1.4348868791354849e-05, + "loss": 0.9194, + "step": 247270 + }, + { + "epoch": 21.4, + "learning_rate": 1.4347424080441504e-05, + "loss": 0.9408, + "step": 247280 + }, + { + "epoch": 21.4, + "learning_rate": 1.434597936952816e-05, + "loss": 0.912, + "step": 247290 + }, + { + "epoch": 21.41, + "learning_rate": 1.4344534658614811e-05, + "loss": 0.9705, + "step": 247300 + }, + { + "epoch": 21.41, + "learning_rate": 1.4343089947701466e-05, + "loss": 0.9622, + "step": 247310 + }, + { + "epoch": 21.41, + "learning_rate": 1.4341645236788121e-05, + "loss": 0.9757, + "step": 247320 + }, + { + "epoch": 21.41, + "learning_rate": 1.4340200525874773e-05, + "loss": 0.8684, + "step": 247330 + }, + { + "epoch": 21.41, + "learning_rate": 1.4338755814961426e-05, + "loss": 0.8897, + "step": 247340 + }, + { + "epoch": 21.41, + "learning_rate": 1.4337311104048081e-05, + "loss": 0.9189, + "step": 247350 + }, + { + "epoch": 21.41, + "learning_rate": 1.4335866393134733e-05, + "loss": 0.9795, + "step": 247360 + }, + { + "epoch": 21.41, + "learning_rate": 1.4334421682221388e-05, + "loss": 0.9495, + "step": 247370 + }, + { + "epoch": 21.41, + "learning_rate": 1.4332976971308043e-05, + "loss": 0.8705, + "step": 247380 + }, + { + "epoch": 21.41, + "learning_rate": 1.4331532260394695e-05, + "loss": 0.9367, + "step": 247390 + }, + { + "epoch": 21.41, + "learning_rate": 1.433008754948135e-05, + "loss": 0.8933, + "step": 247400 + }, + { + "epoch": 21.42, + "learning_rate": 1.4328642838568005e-05, + "loss": 0.9675, + "step": 247410 + }, + { + "epoch": 21.42, + "learning_rate": 1.4327198127654657e-05, + "loss": 0.9339, + "step": 247420 + }, + { + "epoch": 21.42, + "learning_rate": 1.432575341674131e-05, + "loss": 0.9939, + "step": 247430 + }, + { + "epoch": 21.42, + "learning_rate": 1.4324308705827965e-05, + "loss": 0.9355, + "step": 247440 + }, + { + "epoch": 21.42, + "learning_rate": 1.4322863994914617e-05, + "loss": 0.9074, + "step": 247450 + }, + { + "epoch": 21.42, + "learning_rate": 1.4321419284001272e-05, + "loss": 0.942, + "step": 247460 + }, + { + "epoch": 21.42, + "learning_rate": 1.4319974573087927e-05, + "loss": 0.8873, + "step": 247470 + }, + { + "epoch": 21.42, + "learning_rate": 1.4318529862174579e-05, + "loss": 0.897, + "step": 247480 + }, + { + "epoch": 21.42, + "learning_rate": 1.4317085151261234e-05, + "loss": 0.9558, + "step": 247490 + }, + { + "epoch": 21.42, + "learning_rate": 1.4315640440347889e-05, + "loss": 0.9605, + "step": 247500 + }, + { + "epoch": 21.42, + "learning_rate": 1.431419572943454e-05, + "loss": 0.873, + "step": 247510 + }, + { + "epoch": 21.42, + "learning_rate": 1.4312751018521194e-05, + "loss": 0.8841, + "step": 247520 + }, + { + "epoch": 21.43, + "learning_rate": 1.431130630760785e-05, + "loss": 0.9728, + "step": 247530 + }, + { + "epoch": 21.43, + "learning_rate": 1.4309861596694501e-05, + "loss": 0.8684, + "step": 247540 + }, + { + "epoch": 21.43, + "learning_rate": 1.4308416885781156e-05, + "loss": 0.9164, + "step": 247550 + }, + { + "epoch": 21.43, + "learning_rate": 1.4306972174867811e-05, + "loss": 0.9137, + "step": 247560 + }, + { + "epoch": 21.43, + "learning_rate": 1.4305527463954463e-05, + "loss": 0.9408, + "step": 247570 + }, + { + "epoch": 21.43, + "learning_rate": 1.4304082753041118e-05, + "loss": 0.9691, + "step": 247580 + }, + { + "epoch": 21.43, + "learning_rate": 1.4302638042127773e-05, + "loss": 0.9103, + "step": 247590 + }, + { + "epoch": 21.43, + "learning_rate": 1.4301193331214425e-05, + "loss": 0.9146, + "step": 247600 + }, + { + "epoch": 21.43, + "learning_rate": 1.4299748620301078e-05, + "loss": 0.9275, + "step": 247610 + }, + { + "epoch": 21.43, + "learning_rate": 1.4298303909387733e-05, + "loss": 0.9458, + "step": 247620 + }, + { + "epoch": 21.43, + "learning_rate": 1.4296859198474385e-05, + "loss": 0.8459, + "step": 247630 + }, + { + "epoch": 21.44, + "learning_rate": 1.429541448756104e-05, + "loss": 0.9462, + "step": 247640 + }, + { + "epoch": 21.44, + "learning_rate": 1.4293969776647692e-05, + "loss": 0.9559, + "step": 247650 + }, + { + "epoch": 21.44, + "learning_rate": 1.4292525065734347e-05, + "loss": 0.9192, + "step": 247660 + }, + { + "epoch": 21.44, + "learning_rate": 1.4291080354821002e-05, + "loss": 0.8915, + "step": 247670 + }, + { + "epoch": 21.44, + "learning_rate": 1.4289635643907654e-05, + "loss": 0.8844, + "step": 247680 + }, + { + "epoch": 21.44, + "learning_rate": 1.4288190932994309e-05, + "loss": 0.926, + "step": 247690 + }, + { + "epoch": 21.44, + "learning_rate": 1.4286746222080962e-05, + "loss": 0.9909, + "step": 247700 + }, + { + "epoch": 21.44, + "learning_rate": 1.4285301511167615e-05, + "loss": 0.8967, + "step": 247710 + }, + { + "epoch": 21.44, + "learning_rate": 1.4283856800254269e-05, + "loss": 0.9495, + "step": 247720 + }, + { + "epoch": 21.44, + "learning_rate": 1.4282412089340924e-05, + "loss": 0.9338, + "step": 247730 + }, + { + "epoch": 21.44, + "learning_rate": 1.4280967378427576e-05, + "loss": 0.9456, + "step": 247740 + }, + { + "epoch": 21.44, + "learning_rate": 1.427952266751423e-05, + "loss": 0.9653, + "step": 247750 + }, + { + "epoch": 21.45, + "learning_rate": 1.4278077956600886e-05, + "loss": 0.9152, + "step": 247760 + }, + { + "epoch": 21.45, + "learning_rate": 1.4276633245687538e-05, + "loss": 0.9698, + "step": 247770 + }, + { + "epoch": 21.45, + "learning_rate": 1.4275188534774193e-05, + "loss": 0.9037, + "step": 247780 + }, + { + "epoch": 21.45, + "learning_rate": 1.4273743823860846e-05, + "loss": 0.9343, + "step": 247790 + }, + { + "epoch": 21.45, + "learning_rate": 1.42722991129475e-05, + "loss": 0.9704, + "step": 247800 + }, + { + "epoch": 21.45, + "learning_rate": 1.4270854402034153e-05, + "loss": 0.9797, + "step": 247810 + }, + { + "epoch": 21.45, + "learning_rate": 1.4269409691120808e-05, + "loss": 0.9711, + "step": 247820 + }, + { + "epoch": 21.45, + "learning_rate": 1.426796498020746e-05, + "loss": 0.944, + "step": 247830 + }, + { + "epoch": 21.45, + "learning_rate": 1.4266520269294115e-05, + "loss": 0.8849, + "step": 247840 + }, + { + "epoch": 21.45, + "learning_rate": 1.426507555838077e-05, + "loss": 0.9347, + "step": 247850 + }, + { + "epoch": 21.45, + "learning_rate": 1.4263630847467421e-05, + "loss": 0.9856, + "step": 247860 + }, + { + "epoch": 21.45, + "learning_rate": 1.4262186136554077e-05, + "loss": 0.9593, + "step": 247870 + }, + { + "epoch": 21.46, + "learning_rate": 1.426074142564073e-05, + "loss": 0.9153, + "step": 247880 + }, + { + "epoch": 21.46, + "learning_rate": 1.4259296714727383e-05, + "loss": 0.9186, + "step": 247890 + }, + { + "epoch": 21.46, + "learning_rate": 1.4257852003814037e-05, + "loss": 0.9432, + "step": 247900 + }, + { + "epoch": 21.46, + "learning_rate": 1.4256407292900692e-05, + "loss": 0.9342, + "step": 247910 + }, + { + "epoch": 21.46, + "learning_rate": 1.4254962581987344e-05, + "loss": 0.9742, + "step": 247920 + }, + { + "epoch": 21.46, + "learning_rate": 1.4253517871073999e-05, + "loss": 0.931, + "step": 247930 + }, + { + "epoch": 21.46, + "learning_rate": 1.4252073160160654e-05, + "loss": 0.9626, + "step": 247940 + }, + { + "epoch": 21.46, + "learning_rate": 1.4250628449247305e-05, + "loss": 0.9063, + "step": 247950 + }, + { + "epoch": 21.46, + "learning_rate": 1.424918373833396e-05, + "loss": 0.999, + "step": 247960 + }, + { + "epoch": 21.46, + "learning_rate": 1.4247739027420614e-05, + "loss": 0.8791, + "step": 247970 + }, + { + "epoch": 21.46, + "learning_rate": 1.4246294316507267e-05, + "loss": 0.9991, + "step": 247980 + }, + { + "epoch": 21.47, + "learning_rate": 1.424484960559392e-05, + "loss": 0.9219, + "step": 247990 + }, + { + "epoch": 21.47, + "learning_rate": 1.4243404894680576e-05, + "loss": 0.9035, + "step": 248000 + }, + { + "epoch": 21.47, + "learning_rate": 1.4241960183767227e-05, + "loss": 0.9168, + "step": 248010 + }, + { + "epoch": 21.47, + "learning_rate": 1.4240515472853883e-05, + "loss": 0.8879, + "step": 248020 + }, + { + "epoch": 21.47, + "learning_rate": 1.4239070761940538e-05, + "loss": 0.913, + "step": 248030 + }, + { + "epoch": 21.47, + "learning_rate": 1.423762605102719e-05, + "loss": 0.9464, + "step": 248040 + }, + { + "epoch": 21.47, + "learning_rate": 1.4236181340113844e-05, + "loss": 0.9111, + "step": 248050 + }, + { + "epoch": 21.47, + "learning_rate": 1.4234736629200498e-05, + "loss": 0.9392, + "step": 248060 + }, + { + "epoch": 21.47, + "learning_rate": 1.4233291918287151e-05, + "loss": 0.9556, + "step": 248070 + }, + { + "epoch": 21.47, + "learning_rate": 1.4231847207373805e-05, + "loss": 0.9467, + "step": 248080 + }, + { + "epoch": 21.47, + "learning_rate": 1.423040249646046e-05, + "loss": 0.9635, + "step": 248090 + }, + { + "epoch": 21.47, + "learning_rate": 1.4228957785547111e-05, + "loss": 0.9372, + "step": 248100 + }, + { + "epoch": 21.48, + "learning_rate": 1.4227513074633767e-05, + "loss": 0.9773, + "step": 248110 + }, + { + "epoch": 21.48, + "learning_rate": 1.4226068363720422e-05, + "loss": 0.9237, + "step": 248120 + }, + { + "epoch": 21.48, + "learning_rate": 1.4224623652807073e-05, + "loss": 0.9815, + "step": 248130 + }, + { + "epoch": 21.48, + "learning_rate": 1.4223178941893728e-05, + "loss": 0.9153, + "step": 248140 + }, + { + "epoch": 21.48, + "learning_rate": 1.4221734230980382e-05, + "loss": 0.8833, + "step": 248150 + }, + { + "epoch": 21.48, + "learning_rate": 1.4220289520067035e-05, + "loss": 0.9496, + "step": 248160 + }, + { + "epoch": 21.48, + "learning_rate": 1.4218844809153689e-05, + "loss": 0.8916, + "step": 248170 + }, + { + "epoch": 21.48, + "learning_rate": 1.4217400098240344e-05, + "loss": 0.9574, + "step": 248180 + }, + { + "epoch": 21.48, + "learning_rate": 1.4215955387326995e-05, + "loss": 0.9579, + "step": 248190 + }, + { + "epoch": 21.48, + "learning_rate": 1.421451067641365e-05, + "loss": 0.94, + "step": 248200 + }, + { + "epoch": 21.48, + "learning_rate": 1.4213065965500306e-05, + "loss": 0.9723, + "step": 248210 + }, + { + "epoch": 21.49, + "learning_rate": 1.4211621254586957e-05, + "loss": 0.9823, + "step": 248220 + }, + { + "epoch": 21.49, + "learning_rate": 1.4210176543673612e-05, + "loss": 0.9327, + "step": 248230 + }, + { + "epoch": 21.49, + "learning_rate": 1.4208731832760266e-05, + "loss": 0.9233, + "step": 248240 + }, + { + "epoch": 21.49, + "learning_rate": 1.4207287121846919e-05, + "loss": 0.947, + "step": 248250 + }, + { + "epoch": 21.49, + "learning_rate": 1.4205842410933573e-05, + "loss": 0.9602, + "step": 248260 + }, + { + "epoch": 21.49, + "learning_rate": 1.4204397700020228e-05, + "loss": 0.9546, + "step": 248270 + }, + { + "epoch": 21.49, + "learning_rate": 1.420295298910688e-05, + "loss": 0.9816, + "step": 248280 + }, + { + "epoch": 21.49, + "learning_rate": 1.4201508278193534e-05, + "loss": 0.9562, + "step": 248290 + }, + { + "epoch": 21.49, + "learning_rate": 1.420006356728019e-05, + "loss": 0.971, + "step": 248300 + }, + { + "epoch": 21.49, + "learning_rate": 1.4198618856366841e-05, + "loss": 0.9786, + "step": 248310 + }, + { + "epoch": 21.49, + "learning_rate": 1.4197174145453496e-05, + "loss": 0.9599, + "step": 248320 + }, + { + "epoch": 21.49, + "learning_rate": 1.419572943454015e-05, + "loss": 0.9485, + "step": 248330 + }, + { + "epoch": 21.5, + "learning_rate": 1.4194284723626803e-05, + "loss": 0.9239, + "step": 248340 + }, + { + "epoch": 21.5, + "learning_rate": 1.4192840012713456e-05, + "loss": 0.9408, + "step": 248350 + }, + { + "epoch": 21.5, + "learning_rate": 1.4191395301800112e-05, + "loss": 0.968, + "step": 248360 + }, + { + "epoch": 21.5, + "learning_rate": 1.4189950590886763e-05, + "loss": 0.8717, + "step": 248370 + }, + { + "epoch": 21.5, + "learning_rate": 1.4188505879973418e-05, + "loss": 0.9353, + "step": 248380 + }, + { + "epoch": 21.5, + "learning_rate": 1.4187061169060073e-05, + "loss": 0.8905, + "step": 248390 + }, + { + "epoch": 21.5, + "learning_rate": 1.4185616458146725e-05, + "loss": 0.8993, + "step": 248400 + }, + { + "epoch": 21.5, + "learning_rate": 1.418417174723338e-05, + "loss": 0.9867, + "step": 248410 + }, + { + "epoch": 21.5, + "learning_rate": 1.4182727036320034e-05, + "loss": 0.9238, + "step": 248420 + }, + { + "epoch": 21.5, + "learning_rate": 1.4181282325406687e-05, + "loss": 0.9234, + "step": 248430 + }, + { + "epoch": 21.5, + "learning_rate": 1.417983761449334e-05, + "loss": 0.9616, + "step": 248440 + }, + { + "epoch": 21.51, + "learning_rate": 1.4178392903579995e-05, + "loss": 0.9682, + "step": 248450 + }, + { + "epoch": 21.51, + "learning_rate": 1.4176948192666647e-05, + "loss": 0.8846, + "step": 248460 + }, + { + "epoch": 21.51, + "learning_rate": 1.4175503481753302e-05, + "loss": 0.939, + "step": 248470 + }, + { + "epoch": 21.51, + "learning_rate": 1.4174058770839957e-05, + "loss": 0.9764, + "step": 248480 + }, + { + "epoch": 21.51, + "learning_rate": 1.4172614059926609e-05, + "loss": 0.8982, + "step": 248490 + }, + { + "epoch": 21.51, + "learning_rate": 1.4171169349013264e-05, + "loss": 0.9462, + "step": 248500 + }, + { + "epoch": 21.51, + "learning_rate": 1.4169724638099918e-05, + "loss": 0.9578, + "step": 248510 + }, + { + "epoch": 21.51, + "learning_rate": 1.4168279927186571e-05, + "loss": 0.9223, + "step": 248520 + }, + { + "epoch": 21.51, + "learning_rate": 1.4166835216273224e-05, + "loss": 0.9673, + "step": 248530 + }, + { + "epoch": 21.51, + "learning_rate": 1.416539050535988e-05, + "loss": 0.9133, + "step": 248540 + }, + { + "epoch": 21.51, + "learning_rate": 1.4163945794446531e-05, + "loss": 0.9284, + "step": 248550 + }, + { + "epoch": 21.51, + "learning_rate": 1.4162501083533186e-05, + "loss": 0.9913, + "step": 248560 + }, + { + "epoch": 21.52, + "learning_rate": 1.4161056372619838e-05, + "loss": 0.8798, + "step": 248570 + }, + { + "epoch": 21.52, + "learning_rate": 1.4159611661706493e-05, + "loss": 0.9006, + "step": 248580 + }, + { + "epoch": 21.52, + "learning_rate": 1.4158166950793148e-05, + "loss": 0.8843, + "step": 248590 + }, + { + "epoch": 21.52, + "learning_rate": 1.41567222398798e-05, + "loss": 0.9226, + "step": 248600 + }, + { + "epoch": 21.52, + "learning_rate": 1.4155277528966455e-05, + "loss": 0.9257, + "step": 248610 + }, + { + "epoch": 21.52, + "learning_rate": 1.4153832818053108e-05, + "loss": 0.9266, + "step": 248620 + }, + { + "epoch": 21.52, + "learning_rate": 1.415238810713976e-05, + "loss": 0.899, + "step": 248630 + }, + { + "epoch": 21.52, + "learning_rate": 1.4150943396226415e-05, + "loss": 0.9182, + "step": 248640 + }, + { + "epoch": 21.52, + "learning_rate": 1.414949868531307e-05, + "loss": 0.9751, + "step": 248650 + }, + { + "epoch": 21.52, + "learning_rate": 1.4148053974399722e-05, + "loss": 0.9403, + "step": 248660 + }, + { + "epoch": 21.52, + "learning_rate": 1.4146609263486377e-05, + "loss": 0.9575, + "step": 248670 + }, + { + "epoch": 21.53, + "learning_rate": 1.4145164552573032e-05, + "loss": 0.953, + "step": 248680 + }, + { + "epoch": 21.53, + "learning_rate": 1.4143719841659684e-05, + "loss": 0.9101, + "step": 248690 + }, + { + "epoch": 21.53, + "learning_rate": 1.4142275130746339e-05, + "loss": 0.9495, + "step": 248700 + }, + { + "epoch": 21.53, + "learning_rate": 1.4140830419832992e-05, + "loss": 0.8972, + "step": 248710 + }, + { + "epoch": 21.53, + "learning_rate": 1.4139385708919644e-05, + "loss": 0.9813, + "step": 248720 + }, + { + "epoch": 21.53, + "learning_rate": 1.4137940998006299e-05, + "loss": 0.9557, + "step": 248730 + }, + { + "epoch": 21.53, + "learning_rate": 1.4136496287092954e-05, + "loss": 0.9506, + "step": 248740 + }, + { + "epoch": 21.53, + "learning_rate": 1.4135051576179606e-05, + "loss": 0.985, + "step": 248750 + }, + { + "epoch": 21.53, + "learning_rate": 1.4133606865266261e-05, + "loss": 0.9301, + "step": 248760 + }, + { + "epoch": 21.53, + "learning_rate": 1.4132162154352916e-05, + "loss": 0.9194, + "step": 248770 + }, + { + "epoch": 21.53, + "learning_rate": 1.4130717443439568e-05, + "loss": 0.9719, + "step": 248780 + }, + { + "epoch": 21.53, + "learning_rate": 1.4129272732526223e-05, + "loss": 0.9483, + "step": 248790 + }, + { + "epoch": 21.54, + "learning_rate": 1.4127828021612876e-05, + "loss": 0.9556, + "step": 248800 + }, + { + "epoch": 21.54, + "learning_rate": 1.4126383310699528e-05, + "loss": 0.9349, + "step": 248810 + }, + { + "epoch": 21.54, + "learning_rate": 1.4124938599786183e-05, + "loss": 0.9135, + "step": 248820 + }, + { + "epoch": 21.54, + "learning_rate": 1.4123493888872838e-05, + "loss": 0.9211, + "step": 248830 + }, + { + "epoch": 21.54, + "learning_rate": 1.412204917795949e-05, + "loss": 0.9202, + "step": 248840 + }, + { + "epoch": 21.54, + "learning_rate": 1.4120604467046145e-05, + "loss": 0.9373, + "step": 248850 + }, + { + "epoch": 21.54, + "learning_rate": 1.41191597561328e-05, + "loss": 0.9165, + "step": 248860 + }, + { + "epoch": 21.54, + "learning_rate": 1.4117715045219452e-05, + "loss": 0.9759, + "step": 248870 + }, + { + "epoch": 21.54, + "learning_rate": 1.4116270334306107e-05, + "loss": 0.9502, + "step": 248880 + }, + { + "epoch": 21.54, + "learning_rate": 1.411482562339276e-05, + "loss": 0.9199, + "step": 248890 + }, + { + "epoch": 21.54, + "learning_rate": 1.4113380912479412e-05, + "loss": 0.9109, + "step": 248900 + }, + { + "epoch": 21.54, + "learning_rate": 1.4111936201566067e-05, + "loss": 0.874, + "step": 248910 + }, + { + "epoch": 21.55, + "learning_rate": 1.4110491490652722e-05, + "loss": 0.9617, + "step": 248920 + }, + { + "epoch": 21.55, + "learning_rate": 1.4109046779739374e-05, + "loss": 0.9175, + "step": 248930 + }, + { + "epoch": 21.55, + "learning_rate": 1.4107602068826029e-05, + "loss": 0.8857, + "step": 248940 + }, + { + "epoch": 21.55, + "learning_rate": 1.4106157357912684e-05, + "loss": 0.8972, + "step": 248950 + }, + { + "epoch": 21.55, + "learning_rate": 1.4104712646999336e-05, + "loss": 0.9293, + "step": 248960 + }, + { + "epoch": 21.55, + "learning_rate": 1.410326793608599e-05, + "loss": 0.9452, + "step": 248970 + }, + { + "epoch": 21.55, + "learning_rate": 1.4101823225172644e-05, + "loss": 0.9603, + "step": 248980 + }, + { + "epoch": 21.55, + "learning_rate": 1.4100378514259296e-05, + "loss": 0.9314, + "step": 248990 + }, + { + "epoch": 21.55, + "learning_rate": 1.409893380334595e-05, + "loss": 0.9172, + "step": 249000 + }, + { + "epoch": 21.55, + "learning_rate": 1.4097489092432606e-05, + "loss": 0.8547, + "step": 249010 + }, + { + "epoch": 21.55, + "learning_rate": 1.4096044381519258e-05, + "loss": 0.8792, + "step": 249020 + }, + { + "epoch": 21.56, + "learning_rate": 1.4094599670605913e-05, + "loss": 0.8723, + "step": 249030 + }, + { + "epoch": 21.56, + "learning_rate": 1.4093154959692568e-05, + "loss": 0.9439, + "step": 249040 + }, + { + "epoch": 21.56, + "learning_rate": 1.409171024877922e-05, + "loss": 0.9059, + "step": 249050 + }, + { + "epoch": 21.56, + "learning_rate": 1.4090265537865875e-05, + "loss": 0.9616, + "step": 249060 + }, + { + "epoch": 21.56, + "learning_rate": 1.4088820826952528e-05, + "loss": 0.9736, + "step": 249070 + }, + { + "epoch": 21.56, + "learning_rate": 1.408737611603918e-05, + "loss": 0.9448, + "step": 249080 + }, + { + "epoch": 21.56, + "learning_rate": 1.4085931405125835e-05, + "loss": 0.97, + "step": 249090 + }, + { + "epoch": 21.56, + "learning_rate": 1.408448669421249e-05, + "loss": 0.9275, + "step": 249100 + }, + { + "epoch": 21.56, + "learning_rate": 1.4083041983299142e-05, + "loss": 0.9647, + "step": 249110 + }, + { + "epoch": 21.56, + "learning_rate": 1.4081597272385797e-05, + "loss": 0.9647, + "step": 249120 + }, + { + "epoch": 21.56, + "learning_rate": 1.4080152561472452e-05, + "loss": 0.9595, + "step": 249130 + }, + { + "epoch": 21.56, + "learning_rate": 1.4078707850559103e-05, + "loss": 0.9757, + "step": 249140 + }, + { + "epoch": 21.57, + "learning_rate": 1.4077263139645759e-05, + "loss": 0.9704, + "step": 249150 + }, + { + "epoch": 21.57, + "learning_rate": 1.4075818428732412e-05, + "loss": 0.9699, + "step": 249160 + }, + { + "epoch": 21.57, + "learning_rate": 1.4074373717819064e-05, + "loss": 0.9768, + "step": 249170 + }, + { + "epoch": 21.57, + "learning_rate": 1.4072929006905719e-05, + "loss": 0.9424, + "step": 249180 + }, + { + "epoch": 21.57, + "learning_rate": 1.4071484295992374e-05, + "loss": 0.9827, + "step": 249190 + }, + { + "epoch": 21.57, + "learning_rate": 1.4070039585079026e-05, + "loss": 0.9306, + "step": 249200 + }, + { + "epoch": 21.57, + "learning_rate": 1.406859487416568e-05, + "loss": 0.9559, + "step": 249210 + }, + { + "epoch": 21.57, + "learning_rate": 1.4067150163252336e-05, + "loss": 0.9727, + "step": 249220 + }, + { + "epoch": 21.57, + "learning_rate": 1.4065705452338987e-05, + "loss": 0.9359, + "step": 249230 + }, + { + "epoch": 21.57, + "learning_rate": 1.4064260741425642e-05, + "loss": 0.9281, + "step": 249240 + }, + { + "epoch": 21.57, + "learning_rate": 1.4062816030512296e-05, + "loss": 0.9866, + "step": 249250 + }, + { + "epoch": 21.58, + "learning_rate": 1.4061371319598948e-05, + "loss": 0.8841, + "step": 249260 + }, + { + "epoch": 21.58, + "learning_rate": 1.4059926608685603e-05, + "loss": 0.9584, + "step": 249270 + }, + { + "epoch": 21.58, + "learning_rate": 1.4058481897772258e-05, + "loss": 0.9191, + "step": 249280 + }, + { + "epoch": 21.58, + "learning_rate": 1.405703718685891e-05, + "loss": 0.9003, + "step": 249290 + }, + { + "epoch": 21.58, + "learning_rate": 1.4055592475945565e-05, + "loss": 0.957, + "step": 249300 + }, + { + "epoch": 21.58, + "learning_rate": 1.405414776503222e-05, + "loss": 0.9007, + "step": 249310 + }, + { + "epoch": 21.58, + "learning_rate": 1.4052703054118871e-05, + "loss": 0.9378, + "step": 249320 + }, + { + "epoch": 21.58, + "learning_rate": 1.4051258343205526e-05, + "loss": 0.8874, + "step": 249330 + }, + { + "epoch": 21.58, + "learning_rate": 1.404981363229218e-05, + "loss": 0.9468, + "step": 249340 + }, + { + "epoch": 21.58, + "learning_rate": 1.4048368921378832e-05, + "loss": 0.9686, + "step": 249350 + }, + { + "epoch": 21.58, + "learning_rate": 1.4046924210465487e-05, + "loss": 0.8772, + "step": 249360 + }, + { + "epoch": 21.58, + "learning_rate": 1.4045479499552142e-05, + "loss": 0.8767, + "step": 249370 + }, + { + "epoch": 21.59, + "learning_rate": 1.4044034788638793e-05, + "loss": 0.9377, + "step": 249380 + }, + { + "epoch": 21.59, + "learning_rate": 1.4042590077725448e-05, + "loss": 0.9037, + "step": 249390 + }, + { + "epoch": 21.59, + "learning_rate": 1.4041145366812104e-05, + "loss": 0.8974, + "step": 249400 + }, + { + "epoch": 21.59, + "learning_rate": 1.4039700655898755e-05, + "loss": 0.9543, + "step": 249410 + }, + { + "epoch": 21.59, + "learning_rate": 1.403825594498541e-05, + "loss": 0.9831, + "step": 249420 + }, + { + "epoch": 21.59, + "learning_rate": 1.4036811234072064e-05, + "loss": 0.9621, + "step": 249430 + }, + { + "epoch": 21.59, + "learning_rate": 1.4035366523158715e-05, + "loss": 0.9464, + "step": 249440 + }, + { + "epoch": 21.59, + "learning_rate": 1.403392181224537e-05, + "loss": 0.8937, + "step": 249450 + }, + { + "epoch": 21.59, + "learning_rate": 1.4032477101332022e-05, + "loss": 0.9095, + "step": 249460 + }, + { + "epoch": 21.59, + "learning_rate": 1.4031032390418677e-05, + "loss": 0.967, + "step": 249470 + }, + { + "epoch": 21.59, + "learning_rate": 1.4029587679505332e-05, + "loss": 1.0022, + "step": 249480 + }, + { + "epoch": 21.6, + "learning_rate": 1.4028142968591984e-05, + "loss": 0.9512, + "step": 249490 + }, + { + "epoch": 21.6, + "learning_rate": 1.402669825767864e-05, + "loss": 0.9204, + "step": 249500 + }, + { + "epoch": 21.6, + "learning_rate": 1.4025253546765293e-05, + "loss": 0.9744, + "step": 249510 + }, + { + "epoch": 21.6, + "learning_rate": 1.4023808835851946e-05, + "loss": 0.9718, + "step": 249520 + }, + { + "epoch": 21.6, + "learning_rate": 1.40223641249386e-05, + "loss": 0.9975, + "step": 249530 + }, + { + "epoch": 21.6, + "learning_rate": 1.4020919414025254e-05, + "loss": 0.9435, + "step": 249540 + }, + { + "epoch": 21.6, + "learning_rate": 1.4019474703111906e-05, + "loss": 0.9086, + "step": 249550 + }, + { + "epoch": 21.6, + "learning_rate": 1.4018029992198561e-05, + "loss": 0.9755, + "step": 249560 + }, + { + "epoch": 21.6, + "learning_rate": 1.4016585281285216e-05, + "loss": 0.9401, + "step": 249570 + }, + { + "epoch": 21.6, + "learning_rate": 1.4015140570371868e-05, + "loss": 0.9342, + "step": 249580 + }, + { + "epoch": 21.6, + "learning_rate": 1.4013695859458523e-05, + "loss": 0.8938, + "step": 249590 + }, + { + "epoch": 21.6, + "learning_rate": 1.4012251148545177e-05, + "loss": 0.9182, + "step": 249600 + }, + { + "epoch": 21.61, + "learning_rate": 1.401080643763183e-05, + "loss": 0.9682, + "step": 249610 + }, + { + "epoch": 21.61, + "learning_rate": 1.4009361726718483e-05, + "loss": 0.9391, + "step": 249620 + }, + { + "epoch": 21.61, + "learning_rate": 1.4007917015805138e-05, + "loss": 0.9393, + "step": 249630 + }, + { + "epoch": 21.61, + "learning_rate": 1.400647230489179e-05, + "loss": 0.8973, + "step": 249640 + }, + { + "epoch": 21.61, + "learning_rate": 1.4005027593978445e-05, + "loss": 0.9553, + "step": 249650 + }, + { + "epoch": 21.61, + "learning_rate": 1.40035828830651e-05, + "loss": 0.9062, + "step": 249660 + }, + { + "epoch": 21.61, + "learning_rate": 1.4002138172151752e-05, + "loss": 0.9696, + "step": 249670 + }, + { + "epoch": 21.61, + "learning_rate": 1.4000693461238407e-05, + "loss": 0.9073, + "step": 249680 + }, + { + "epoch": 21.61, + "learning_rate": 1.399924875032506e-05, + "loss": 0.9127, + "step": 249690 + }, + { + "epoch": 21.61, + "learning_rate": 1.3997804039411714e-05, + "loss": 0.9603, + "step": 249700 + }, + { + "epoch": 21.61, + "learning_rate": 1.3996359328498367e-05, + "loss": 0.9434, + "step": 249710 + }, + { + "epoch": 21.62, + "learning_rate": 1.3994914617585022e-05, + "loss": 0.979, + "step": 249720 + }, + { + "epoch": 21.62, + "learning_rate": 1.3993469906671674e-05, + "loss": 0.9493, + "step": 249730 + }, + { + "epoch": 21.62, + "learning_rate": 1.399202519575833e-05, + "loss": 0.8911, + "step": 249740 + }, + { + "epoch": 21.62, + "learning_rate": 1.3990580484844984e-05, + "loss": 0.9807, + "step": 249750 + }, + { + "epoch": 21.62, + "learning_rate": 1.3989135773931636e-05, + "loss": 0.9614, + "step": 249760 + }, + { + "epoch": 21.62, + "learning_rate": 1.3987691063018291e-05, + "loss": 0.9467, + "step": 249770 + }, + { + "epoch": 21.62, + "learning_rate": 1.3986246352104944e-05, + "loss": 0.902, + "step": 249780 + }, + { + "epoch": 21.62, + "learning_rate": 1.3984801641191598e-05, + "loss": 0.8945, + "step": 249790 + }, + { + "epoch": 21.62, + "learning_rate": 1.3983356930278251e-05, + "loss": 0.9273, + "step": 249800 + }, + { + "epoch": 21.62, + "learning_rate": 1.3981912219364906e-05, + "loss": 0.9623, + "step": 249810 + }, + { + "epoch": 21.62, + "learning_rate": 1.3980467508451558e-05, + "loss": 0.9532, + "step": 249820 + }, + { + "epoch": 21.62, + "learning_rate": 1.3979022797538213e-05, + "loss": 0.9465, + "step": 249830 + }, + { + "epoch": 21.63, + "learning_rate": 1.3977578086624868e-05, + "loss": 0.8926, + "step": 249840 + }, + { + "epoch": 21.63, + "learning_rate": 1.397613337571152e-05, + "loss": 0.8928, + "step": 249850 + }, + { + "epoch": 21.63, + "learning_rate": 1.3974688664798175e-05, + "loss": 0.9417, + "step": 249860 + }, + { + "epoch": 21.63, + "learning_rate": 1.3973243953884828e-05, + "loss": 0.9246, + "step": 249870 + }, + { + "epoch": 21.63, + "learning_rate": 1.3971799242971482e-05, + "loss": 0.9416, + "step": 249880 + }, + { + "epoch": 21.63, + "learning_rate": 1.3970354532058135e-05, + "loss": 0.9566, + "step": 249890 + }, + { + "epoch": 21.63, + "learning_rate": 1.396890982114479e-05, + "loss": 0.929, + "step": 249900 + }, + { + "epoch": 21.63, + "learning_rate": 1.3967465110231442e-05, + "loss": 0.889, + "step": 249910 + }, + { + "epoch": 21.63, + "learning_rate": 1.3966020399318097e-05, + "loss": 0.8596, + "step": 249920 + }, + { + "epoch": 21.63, + "learning_rate": 1.3964575688404752e-05, + "loss": 0.9261, + "step": 249930 + }, + { + "epoch": 21.63, + "learning_rate": 1.3963130977491404e-05, + "loss": 0.9557, + "step": 249940 + }, + { + "epoch": 21.64, + "learning_rate": 1.3961686266578059e-05, + "loss": 0.9313, + "step": 249950 + }, + { + "epoch": 21.64, + "learning_rate": 1.3960241555664712e-05, + "loss": 0.9437, + "step": 249960 + }, + { + "epoch": 21.64, + "learning_rate": 1.3958796844751366e-05, + "loss": 0.9777, + "step": 249970 + }, + { + "epoch": 21.64, + "learning_rate": 1.3957352133838019e-05, + "loss": 0.9205, + "step": 249980 + }, + { + "epoch": 21.64, + "learning_rate": 1.3955907422924674e-05, + "loss": 0.924, + "step": 249990 + }, + { + "epoch": 21.64, + "learning_rate": 1.3954462712011326e-05, + "loss": 0.909, + "step": 250000 + }, + { + "epoch": 21.64, + "learning_rate": 1.3953018001097981e-05, + "loss": 0.9843, + "step": 250010 + }, + { + "epoch": 21.64, + "learning_rate": 1.3951573290184636e-05, + "loss": 0.8898, + "step": 250020 + }, + { + "epoch": 21.64, + "learning_rate": 1.3950128579271288e-05, + "loss": 0.9027, + "step": 250030 + }, + { + "epoch": 21.64, + "learning_rate": 1.3948683868357943e-05, + "loss": 0.891, + "step": 250040 + }, + { + "epoch": 21.64, + "learning_rate": 1.3947239157444596e-05, + "loss": 0.8996, + "step": 250050 + }, + { + "epoch": 21.64, + "learning_rate": 1.394579444653125e-05, + "loss": 0.974, + "step": 250060 + }, + { + "epoch": 21.65, + "learning_rate": 1.3944349735617903e-05, + "loss": 0.9326, + "step": 250070 + }, + { + "epoch": 21.65, + "learning_rate": 1.3942905024704558e-05, + "loss": 0.9126, + "step": 250080 + }, + { + "epoch": 21.65, + "learning_rate": 1.394146031379121e-05, + "loss": 0.9788, + "step": 250090 + }, + { + "epoch": 21.65, + "learning_rate": 1.3940015602877865e-05, + "loss": 0.9174, + "step": 250100 + }, + { + "epoch": 21.65, + "learning_rate": 1.393857089196452e-05, + "loss": 0.9218, + "step": 250110 + }, + { + "epoch": 21.65, + "learning_rate": 1.3937126181051172e-05, + "loss": 0.9092, + "step": 250120 + }, + { + "epoch": 21.65, + "learning_rate": 1.3935681470137827e-05, + "loss": 0.9399, + "step": 250130 + }, + { + "epoch": 21.65, + "learning_rate": 1.393423675922448e-05, + "loss": 0.9782, + "step": 250140 + }, + { + "epoch": 21.65, + "learning_rate": 1.3932792048311134e-05, + "loss": 0.9008, + "step": 250150 + }, + { + "epoch": 21.65, + "learning_rate": 1.3931347337397787e-05, + "loss": 0.9502, + "step": 250160 + }, + { + "epoch": 21.65, + "learning_rate": 1.3929902626484442e-05, + "loss": 0.9268, + "step": 250170 + }, + { + "epoch": 21.65, + "learning_rate": 1.3928457915571094e-05, + "loss": 0.9354, + "step": 250180 + }, + { + "epoch": 21.66, + "learning_rate": 1.3927013204657749e-05, + "loss": 0.9185, + "step": 250190 + }, + { + "epoch": 21.66, + "learning_rate": 1.3925568493744404e-05, + "loss": 0.8562, + "step": 250200 + }, + { + "epoch": 21.66, + "learning_rate": 1.3924123782831056e-05, + "loss": 0.9045, + "step": 250210 + }, + { + "epoch": 21.66, + "learning_rate": 1.392267907191771e-05, + "loss": 0.9533, + "step": 250220 + }, + { + "epoch": 21.66, + "learning_rate": 1.3921234361004364e-05, + "loss": 0.9088, + "step": 250230 + }, + { + "epoch": 21.66, + "learning_rate": 1.3919789650091018e-05, + "loss": 0.8958, + "step": 250240 + }, + { + "epoch": 21.66, + "learning_rate": 1.3918344939177671e-05, + "loss": 0.9418, + "step": 250250 + }, + { + "epoch": 21.66, + "learning_rate": 1.3916900228264326e-05, + "loss": 0.958, + "step": 250260 + }, + { + "epoch": 21.66, + "learning_rate": 1.3915455517350978e-05, + "loss": 0.9481, + "step": 250270 + }, + { + "epoch": 21.66, + "learning_rate": 1.3914010806437633e-05, + "loss": 0.9861, + "step": 250280 + }, + { + "epoch": 21.66, + "learning_rate": 1.3912566095524288e-05, + "loss": 0.9336, + "step": 250290 + }, + { + "epoch": 21.67, + "learning_rate": 1.391112138461094e-05, + "loss": 0.9019, + "step": 250300 + }, + { + "epoch": 21.67, + "learning_rate": 1.3909676673697595e-05, + "loss": 0.9666, + "step": 250310 + }, + { + "epoch": 21.67, + "learning_rate": 1.3908231962784248e-05, + "loss": 0.9061, + "step": 250320 + }, + { + "epoch": 21.67, + "learning_rate": 1.3906787251870901e-05, + "loss": 0.9266, + "step": 250330 + }, + { + "epoch": 21.67, + "learning_rate": 1.3905342540957555e-05, + "loss": 0.9564, + "step": 250340 + }, + { + "epoch": 21.67, + "learning_rate": 1.390389783004421e-05, + "loss": 0.9925, + "step": 250350 + }, + { + "epoch": 21.67, + "learning_rate": 1.3902453119130862e-05, + "loss": 0.8617, + "step": 250360 + }, + { + "epoch": 21.67, + "learning_rate": 1.3901008408217517e-05, + "loss": 0.9863, + "step": 250370 + }, + { + "epoch": 21.67, + "learning_rate": 1.3899563697304168e-05, + "loss": 0.918, + "step": 250380 + }, + { + "epoch": 21.67, + "learning_rate": 1.3898118986390824e-05, + "loss": 0.9951, + "step": 250390 + }, + { + "epoch": 21.67, + "learning_rate": 1.3896674275477479e-05, + "loss": 0.9431, + "step": 250400 + }, + { + "epoch": 21.67, + "learning_rate": 1.389522956456413e-05, + "loss": 0.9257, + "step": 250410 + }, + { + "epoch": 21.68, + "learning_rate": 1.3893784853650785e-05, + "loss": 0.9286, + "step": 250420 + }, + { + "epoch": 21.68, + "learning_rate": 1.3892340142737439e-05, + "loss": 0.9508, + "step": 250430 + }, + { + "epoch": 21.68, + "learning_rate": 1.389089543182409e-05, + "loss": 0.9663, + "step": 250440 + }, + { + "epoch": 21.68, + "learning_rate": 1.3889450720910746e-05, + "loss": 0.9408, + "step": 250450 + }, + { + "epoch": 21.68, + "learning_rate": 1.38880060099974e-05, + "loss": 0.9213, + "step": 250460 + }, + { + "epoch": 21.68, + "learning_rate": 1.3886561299084052e-05, + "loss": 0.8727, + "step": 250470 + }, + { + "epoch": 21.68, + "learning_rate": 1.3885116588170708e-05, + "loss": 0.9149, + "step": 250480 + }, + { + "epoch": 21.68, + "learning_rate": 1.3883671877257363e-05, + "loss": 0.9804, + "step": 250490 + }, + { + "epoch": 21.68, + "learning_rate": 1.3882227166344014e-05, + "loss": 0.9428, + "step": 250500 + }, + { + "epoch": 21.68, + "learning_rate": 1.388078245543067e-05, + "loss": 0.9695, + "step": 250510 + }, + { + "epoch": 21.68, + "learning_rate": 1.3879337744517323e-05, + "loss": 0.9524, + "step": 250520 + }, + { + "epoch": 21.69, + "learning_rate": 1.3877893033603974e-05, + "loss": 0.8604, + "step": 250530 + }, + { + "epoch": 21.69, + "learning_rate": 1.387644832269063e-05, + "loss": 0.8948, + "step": 250540 + }, + { + "epoch": 21.69, + "learning_rate": 1.3875003611777285e-05, + "loss": 0.8845, + "step": 250550 + }, + { + "epoch": 21.69, + "learning_rate": 1.3873558900863936e-05, + "loss": 1.0101, + "step": 250560 + }, + { + "epoch": 21.69, + "learning_rate": 1.3872114189950591e-05, + "loss": 0.9238, + "step": 250570 + }, + { + "epoch": 21.69, + "learning_rate": 1.3870669479037247e-05, + "loss": 0.9667, + "step": 250580 + }, + { + "epoch": 21.69, + "learning_rate": 1.3869224768123898e-05, + "loss": 0.9573, + "step": 250590 + }, + { + "epoch": 21.69, + "learning_rate": 1.3867780057210553e-05, + "loss": 0.8919, + "step": 250600 + }, + { + "epoch": 21.69, + "learning_rate": 1.3866335346297207e-05, + "loss": 0.9458, + "step": 250610 + }, + { + "epoch": 21.69, + "learning_rate": 1.3864890635383858e-05, + "loss": 0.9349, + "step": 250620 + }, + { + "epoch": 21.69, + "learning_rate": 1.3863445924470514e-05, + "loss": 0.947, + "step": 250630 + }, + { + "epoch": 21.69, + "learning_rate": 1.3862001213557169e-05, + "loss": 0.8749, + "step": 250640 + }, + { + "epoch": 21.7, + "learning_rate": 1.386055650264382e-05, + "loss": 0.873, + "step": 250650 + }, + { + "epoch": 21.7, + "learning_rate": 1.3859111791730475e-05, + "loss": 0.9088, + "step": 250660 + }, + { + "epoch": 21.7, + "learning_rate": 1.385766708081713e-05, + "loss": 0.9435, + "step": 250670 + }, + { + "epoch": 21.7, + "learning_rate": 1.3856222369903782e-05, + "loss": 0.9505, + "step": 250680 + }, + { + "epoch": 21.7, + "learning_rate": 1.3854777658990437e-05, + "loss": 0.8827, + "step": 250690 + }, + { + "epoch": 21.7, + "learning_rate": 1.385333294807709e-05, + "loss": 0.9087, + "step": 250700 + }, + { + "epoch": 21.7, + "learning_rate": 1.3851888237163742e-05, + "loss": 0.8643, + "step": 250710 + }, + { + "epoch": 21.7, + "learning_rate": 1.3850443526250397e-05, + "loss": 0.8856, + "step": 250720 + }, + { + "epoch": 21.7, + "learning_rate": 1.3848998815337053e-05, + "loss": 0.8979, + "step": 250730 + }, + { + "epoch": 21.7, + "learning_rate": 1.3847554104423704e-05, + "loss": 0.9544, + "step": 250740 + }, + { + "epoch": 21.7, + "learning_rate": 1.384610939351036e-05, + "loss": 0.8989, + "step": 250750 + }, + { + "epoch": 21.71, + "learning_rate": 1.3844664682597014e-05, + "loss": 0.8573, + "step": 250760 + }, + { + "epoch": 21.71, + "learning_rate": 1.3843219971683666e-05, + "loss": 0.9719, + "step": 250770 + }, + { + "epoch": 21.71, + "learning_rate": 1.3841775260770321e-05, + "loss": 0.9384, + "step": 250780 + }, + { + "epoch": 21.71, + "learning_rate": 1.3840330549856975e-05, + "loss": 0.9192, + "step": 250790 + }, + { + "epoch": 21.71, + "learning_rate": 1.3838885838943626e-05, + "loss": 0.8857, + "step": 250800 + }, + { + "epoch": 21.71, + "learning_rate": 1.3837441128030281e-05, + "loss": 0.9538, + "step": 250810 + }, + { + "epoch": 21.71, + "learning_rate": 1.3835996417116936e-05, + "loss": 1.0405, + "step": 250820 + }, + { + "epoch": 21.71, + "learning_rate": 1.3834551706203588e-05, + "loss": 0.9952, + "step": 250830 + }, + { + "epoch": 21.71, + "learning_rate": 1.3833106995290243e-05, + "loss": 0.9663, + "step": 250840 + }, + { + "epoch": 21.71, + "learning_rate": 1.3831662284376898e-05, + "loss": 0.8646, + "step": 250850 + }, + { + "epoch": 21.71, + "learning_rate": 1.383021757346355e-05, + "loss": 0.9701, + "step": 250860 + }, + { + "epoch": 21.71, + "learning_rate": 1.3828772862550205e-05, + "loss": 0.9456, + "step": 250870 + }, + { + "epoch": 21.72, + "learning_rate": 1.3827328151636859e-05, + "loss": 0.8884, + "step": 250880 + }, + { + "epoch": 21.72, + "learning_rate": 1.382588344072351e-05, + "loss": 0.9737, + "step": 250890 + }, + { + "epoch": 21.72, + "learning_rate": 1.3824438729810165e-05, + "loss": 0.9152, + "step": 250900 + }, + { + "epoch": 21.72, + "learning_rate": 1.382299401889682e-05, + "loss": 0.9275, + "step": 250910 + }, + { + "epoch": 21.72, + "learning_rate": 1.3821549307983472e-05, + "loss": 0.9546, + "step": 250920 + }, + { + "epoch": 21.72, + "learning_rate": 1.3820104597070127e-05, + "loss": 0.9643, + "step": 250930 + }, + { + "epoch": 21.72, + "learning_rate": 1.3818659886156782e-05, + "loss": 0.9435, + "step": 250940 + }, + { + "epoch": 21.72, + "learning_rate": 1.3817215175243434e-05, + "loss": 0.9163, + "step": 250950 + }, + { + "epoch": 21.72, + "learning_rate": 1.3815770464330089e-05, + "loss": 0.9465, + "step": 250960 + }, + { + "epoch": 21.72, + "learning_rate": 1.3814325753416742e-05, + "loss": 0.9729, + "step": 250970 + }, + { + "epoch": 21.72, + "learning_rate": 1.3812881042503394e-05, + "loss": 0.9386, + "step": 250980 + }, + { + "epoch": 21.73, + "learning_rate": 1.381143633159005e-05, + "loss": 0.8959, + "step": 250990 + }, + { + "epoch": 21.73, + "learning_rate": 1.3809991620676704e-05, + "loss": 0.9214, + "step": 251000 + }, + { + "epoch": 21.73, + "learning_rate": 1.3808546909763356e-05, + "loss": 0.989, + "step": 251010 + }, + { + "epoch": 21.73, + "learning_rate": 1.3807102198850011e-05, + "loss": 0.9881, + "step": 251020 + }, + { + "epoch": 21.73, + "learning_rate": 1.3805657487936666e-05, + "loss": 0.9354, + "step": 251030 + }, + { + "epoch": 21.73, + "learning_rate": 1.3804212777023318e-05, + "loss": 0.9687, + "step": 251040 + }, + { + "epoch": 21.73, + "learning_rate": 1.3802768066109973e-05, + "loss": 0.9896, + "step": 251050 + }, + { + "epoch": 21.73, + "learning_rate": 1.3801323355196626e-05, + "loss": 0.908, + "step": 251060 + }, + { + "epoch": 21.73, + "learning_rate": 1.3799878644283278e-05, + "loss": 0.9564, + "step": 251070 + }, + { + "epoch": 21.73, + "learning_rate": 1.3798433933369933e-05, + "loss": 0.9068, + "step": 251080 + }, + { + "epoch": 21.73, + "learning_rate": 1.3796989222456588e-05, + "loss": 0.952, + "step": 251090 + }, + { + "epoch": 21.73, + "learning_rate": 1.379554451154324e-05, + "loss": 0.9201, + "step": 251100 + }, + { + "epoch": 21.74, + "learning_rate": 1.3794099800629895e-05, + "loss": 0.9169, + "step": 251110 + }, + { + "epoch": 21.74, + "learning_rate": 1.379265508971655e-05, + "loss": 0.874, + "step": 251120 + }, + { + "epoch": 21.74, + "learning_rate": 1.3791210378803202e-05, + "loss": 0.9332, + "step": 251130 + }, + { + "epoch": 21.74, + "learning_rate": 1.3789765667889857e-05, + "loss": 0.9524, + "step": 251140 + }, + { + "epoch": 21.74, + "learning_rate": 1.378832095697651e-05, + "loss": 0.9541, + "step": 251150 + }, + { + "epoch": 21.74, + "learning_rate": 1.3786876246063162e-05, + "loss": 0.9619, + "step": 251160 + }, + { + "epoch": 21.74, + "learning_rate": 1.3785431535149817e-05, + "loss": 0.9732, + "step": 251170 + }, + { + "epoch": 21.74, + "learning_rate": 1.3783986824236472e-05, + "loss": 0.9901, + "step": 251180 + }, + { + "epoch": 21.74, + "learning_rate": 1.3782542113323124e-05, + "loss": 0.9364, + "step": 251190 + }, + { + "epoch": 21.74, + "learning_rate": 1.3781097402409779e-05, + "loss": 1.0318, + "step": 251200 + }, + { + "epoch": 21.74, + "learning_rate": 1.3779652691496434e-05, + "loss": 0.9114, + "step": 251210 + }, + { + "epoch": 21.74, + "learning_rate": 1.3778207980583086e-05, + "loss": 0.9466, + "step": 251220 + }, + { + "epoch": 21.75, + "learning_rate": 1.3776763269669741e-05, + "loss": 0.8674, + "step": 251230 + }, + { + "epoch": 21.75, + "learning_rate": 1.3775318558756394e-05, + "loss": 0.9109, + "step": 251240 + }, + { + "epoch": 21.75, + "learning_rate": 1.3773873847843046e-05, + "loss": 0.9272, + "step": 251250 + }, + { + "epoch": 21.75, + "learning_rate": 1.3772429136929701e-05, + "loss": 0.9513, + "step": 251260 + }, + { + "epoch": 21.75, + "learning_rate": 1.3770984426016353e-05, + "loss": 0.9162, + "step": 251270 + }, + { + "epoch": 21.75, + "learning_rate": 1.3769539715103008e-05, + "loss": 0.941, + "step": 251280 + }, + { + "epoch": 21.75, + "learning_rate": 1.3768095004189663e-05, + "loss": 0.8982, + "step": 251290 + }, + { + "epoch": 21.75, + "learning_rate": 1.3766650293276315e-05, + "loss": 0.9166, + "step": 251300 + }, + { + "epoch": 21.75, + "learning_rate": 1.376520558236297e-05, + "loss": 0.9286, + "step": 251310 + }, + { + "epoch": 21.75, + "learning_rate": 1.3763760871449625e-05, + "loss": 0.9765, + "step": 251320 + }, + { + "epoch": 21.75, + "learning_rate": 1.3762316160536277e-05, + "loss": 0.9531, + "step": 251330 + }, + { + "epoch": 21.76, + "learning_rate": 1.376087144962293e-05, + "loss": 0.9715, + "step": 251340 + }, + { + "epoch": 21.76, + "learning_rate": 1.3759426738709585e-05, + "loss": 0.9681, + "step": 251350 + }, + { + "epoch": 21.76, + "learning_rate": 1.3757982027796237e-05, + "loss": 0.9582, + "step": 251360 + }, + { + "epoch": 21.76, + "learning_rate": 1.3756537316882892e-05, + "loss": 0.9088, + "step": 251370 + }, + { + "epoch": 21.76, + "learning_rate": 1.3755092605969547e-05, + "loss": 0.9862, + "step": 251380 + }, + { + "epoch": 21.76, + "learning_rate": 1.3753647895056199e-05, + "loss": 0.9629, + "step": 251390 + }, + { + "epoch": 21.76, + "learning_rate": 1.3752203184142854e-05, + "loss": 0.9003, + "step": 251400 + }, + { + "epoch": 21.76, + "learning_rate": 1.3750758473229509e-05, + "loss": 0.8703, + "step": 251410 + }, + { + "epoch": 21.76, + "learning_rate": 1.374931376231616e-05, + "loss": 0.8818, + "step": 251420 + }, + { + "epoch": 21.76, + "learning_rate": 1.3747869051402814e-05, + "loss": 0.9464, + "step": 251430 + }, + { + "epoch": 21.76, + "learning_rate": 1.3746424340489469e-05, + "loss": 0.9342, + "step": 251440 + }, + { + "epoch": 21.76, + "learning_rate": 1.374497962957612e-05, + "loss": 0.9648, + "step": 251450 + }, + { + "epoch": 21.77, + "learning_rate": 1.3743534918662776e-05, + "loss": 0.8829, + "step": 251460 + }, + { + "epoch": 21.77, + "learning_rate": 1.3742090207749431e-05, + "loss": 0.8891, + "step": 251470 + }, + { + "epoch": 21.77, + "learning_rate": 1.3740645496836083e-05, + "loss": 0.9231, + "step": 251480 + }, + { + "epoch": 21.77, + "learning_rate": 1.3739200785922738e-05, + "loss": 0.9941, + "step": 251490 + }, + { + "epoch": 21.77, + "learning_rate": 1.3737756075009393e-05, + "loss": 0.9205, + "step": 251500 + }, + { + "epoch": 21.77, + "learning_rate": 1.3736311364096044e-05, + "loss": 0.912, + "step": 251510 + }, + { + "epoch": 21.77, + "learning_rate": 1.3734866653182698e-05, + "loss": 0.8989, + "step": 251520 + }, + { + "epoch": 21.77, + "learning_rate": 1.3733421942269353e-05, + "loss": 0.8973, + "step": 251530 + }, + { + "epoch": 21.77, + "learning_rate": 1.3731977231356005e-05, + "loss": 0.9027, + "step": 251540 + }, + { + "epoch": 21.77, + "learning_rate": 1.373053252044266e-05, + "loss": 0.9778, + "step": 251550 + }, + { + "epoch": 21.77, + "learning_rate": 1.3729087809529315e-05, + "loss": 0.995, + "step": 251560 + }, + { + "epoch": 21.78, + "learning_rate": 1.3727643098615967e-05, + "loss": 0.983, + "step": 251570 + }, + { + "epoch": 21.78, + "learning_rate": 1.3726198387702622e-05, + "loss": 1.0094, + "step": 251580 + }, + { + "epoch": 21.78, + "learning_rate": 1.3724753676789277e-05, + "loss": 0.9892, + "step": 251590 + }, + { + "epoch": 21.78, + "learning_rate": 1.3723308965875928e-05, + "loss": 0.9932, + "step": 251600 + }, + { + "epoch": 21.78, + "learning_rate": 1.3721864254962582e-05, + "loss": 0.9302, + "step": 251610 + }, + { + "epoch": 21.78, + "learning_rate": 1.3720419544049237e-05, + "loss": 0.9375, + "step": 251620 + }, + { + "epoch": 21.78, + "learning_rate": 1.3718974833135889e-05, + "loss": 0.8868, + "step": 251630 + }, + { + "epoch": 21.78, + "learning_rate": 1.3717530122222544e-05, + "loss": 0.9169, + "step": 251640 + }, + { + "epoch": 21.78, + "learning_rate": 1.3716085411309199e-05, + "loss": 0.9449, + "step": 251650 + }, + { + "epoch": 21.78, + "learning_rate": 1.371464070039585e-05, + "loss": 0.9318, + "step": 251660 + }, + { + "epoch": 21.78, + "learning_rate": 1.3713195989482506e-05, + "loss": 0.9879, + "step": 251670 + }, + { + "epoch": 21.78, + "learning_rate": 1.371175127856916e-05, + "loss": 0.9139, + "step": 251680 + }, + { + "epoch": 21.79, + "learning_rate": 1.3710306567655812e-05, + "loss": 0.9945, + "step": 251690 + }, + { + "epoch": 21.79, + "learning_rate": 1.3708861856742466e-05, + "loss": 0.9193, + "step": 251700 + }, + { + "epoch": 21.79, + "learning_rate": 1.370741714582912e-05, + "loss": 0.9707, + "step": 251710 + }, + { + "epoch": 21.79, + "learning_rate": 1.3705972434915773e-05, + "loss": 0.9384, + "step": 251720 + }, + { + "epoch": 21.79, + "learning_rate": 1.3704527724002428e-05, + "loss": 0.9072, + "step": 251730 + }, + { + "epoch": 21.79, + "learning_rate": 1.3703083013089083e-05, + "loss": 0.8845, + "step": 251740 + }, + { + "epoch": 21.79, + "learning_rate": 1.3701638302175734e-05, + "loss": 0.9326, + "step": 251750 + }, + { + "epoch": 21.79, + "learning_rate": 1.370019359126239e-05, + "loss": 0.9293, + "step": 251760 + }, + { + "epoch": 21.79, + "learning_rate": 1.3698748880349045e-05, + "loss": 0.9359, + "step": 251770 + }, + { + "epoch": 21.79, + "learning_rate": 1.3697304169435696e-05, + "loss": 0.9786, + "step": 251780 + }, + { + "epoch": 21.79, + "learning_rate": 1.369585945852235e-05, + "loss": 0.8616, + "step": 251790 + }, + { + "epoch": 21.8, + "learning_rate": 1.3694414747609005e-05, + "loss": 0.9295, + "step": 251800 + }, + { + "epoch": 21.8, + "learning_rate": 1.3692970036695656e-05, + "loss": 0.9434, + "step": 251810 + }, + { + "epoch": 21.8, + "learning_rate": 1.3691525325782312e-05, + "loss": 0.9627, + "step": 251820 + }, + { + "epoch": 21.8, + "learning_rate": 1.3690080614868967e-05, + "loss": 0.9026, + "step": 251830 + }, + { + "epoch": 21.8, + "learning_rate": 1.3688635903955618e-05, + "loss": 0.9589, + "step": 251840 + }, + { + "epoch": 21.8, + "learning_rate": 1.3687191193042273e-05, + "loss": 0.9292, + "step": 251850 + }, + { + "epoch": 21.8, + "learning_rate": 1.3685746482128929e-05, + "loss": 0.9274, + "step": 251860 + }, + { + "epoch": 21.8, + "learning_rate": 1.368430177121558e-05, + "loss": 0.9295, + "step": 251870 + }, + { + "epoch": 21.8, + "learning_rate": 1.3682857060302234e-05, + "loss": 0.8994, + "step": 251880 + }, + { + "epoch": 21.8, + "learning_rate": 1.3681412349388889e-05, + "loss": 0.8868, + "step": 251890 + }, + { + "epoch": 21.8, + "learning_rate": 1.367996763847554e-05, + "loss": 0.9376, + "step": 251900 + }, + { + "epoch": 21.8, + "learning_rate": 1.3678522927562196e-05, + "loss": 0.9135, + "step": 251910 + }, + { + "epoch": 21.81, + "learning_rate": 1.367707821664885e-05, + "loss": 0.9179, + "step": 251920 + }, + { + "epoch": 21.81, + "learning_rate": 1.3675633505735502e-05, + "loss": 0.9465, + "step": 251930 + }, + { + "epoch": 21.81, + "learning_rate": 1.3674188794822157e-05, + "loss": 0.9427, + "step": 251940 + }, + { + "epoch": 21.81, + "learning_rate": 1.3672744083908812e-05, + "loss": 0.9018, + "step": 251950 + }, + { + "epoch": 21.81, + "learning_rate": 1.3671299372995464e-05, + "loss": 0.9521, + "step": 251960 + }, + { + "epoch": 21.81, + "learning_rate": 1.3669854662082118e-05, + "loss": 0.8899, + "step": 251970 + }, + { + "epoch": 21.81, + "learning_rate": 1.3668409951168773e-05, + "loss": 0.9529, + "step": 251980 + }, + { + "epoch": 21.81, + "learning_rate": 1.3666965240255424e-05, + "loss": 0.921, + "step": 251990 + }, + { + "epoch": 21.81, + "learning_rate": 1.366552052934208e-05, + "loss": 0.9312, + "step": 252000 + }, + { + "epoch": 21.81, + "learning_rate": 1.3664075818428735e-05, + "loss": 0.9047, + "step": 252010 + }, + { + "epoch": 21.81, + "learning_rate": 1.3662631107515386e-05, + "loss": 0.9206, + "step": 252020 + }, + { + "epoch": 21.82, + "learning_rate": 1.3661186396602041e-05, + "loss": 1.0024, + "step": 252030 + }, + { + "epoch": 21.82, + "learning_rate": 1.3659741685688696e-05, + "loss": 0.9946, + "step": 252040 + }, + { + "epoch": 21.82, + "learning_rate": 1.3658296974775348e-05, + "loss": 0.9346, + "step": 252050 + }, + { + "epoch": 21.82, + "learning_rate": 1.3656852263862002e-05, + "loss": 0.9489, + "step": 252060 + }, + { + "epoch": 21.82, + "learning_rate": 1.3655407552948657e-05, + "loss": 0.918, + "step": 252070 + }, + { + "epoch": 21.82, + "learning_rate": 1.3653962842035308e-05, + "loss": 0.898, + "step": 252080 + }, + { + "epoch": 21.82, + "learning_rate": 1.3652518131121963e-05, + "loss": 0.916, + "step": 252090 + }, + { + "epoch": 21.82, + "learning_rate": 1.3651073420208618e-05, + "loss": 0.9006, + "step": 252100 + }, + { + "epoch": 21.82, + "learning_rate": 1.364962870929527e-05, + "loss": 0.9168, + "step": 252110 + }, + { + "epoch": 21.82, + "learning_rate": 1.3648183998381925e-05, + "loss": 0.9305, + "step": 252120 + }, + { + "epoch": 21.82, + "learning_rate": 1.364673928746858e-05, + "loss": 0.907, + "step": 252130 + }, + { + "epoch": 21.82, + "learning_rate": 1.3645294576555232e-05, + "loss": 0.9562, + "step": 252140 + }, + { + "epoch": 21.83, + "learning_rate": 1.3643849865641885e-05, + "loss": 0.9123, + "step": 252150 + }, + { + "epoch": 21.83, + "learning_rate": 1.364240515472854e-05, + "loss": 0.9335, + "step": 252160 + }, + { + "epoch": 21.83, + "learning_rate": 1.3640960443815192e-05, + "loss": 0.9296, + "step": 252170 + }, + { + "epoch": 21.83, + "learning_rate": 1.3639515732901847e-05, + "loss": 0.872, + "step": 252180 + }, + { + "epoch": 21.83, + "learning_rate": 1.3638071021988499e-05, + "loss": 0.983, + "step": 252190 + }, + { + "epoch": 21.83, + "learning_rate": 1.3636626311075154e-05, + "loss": 0.9167, + "step": 252200 + }, + { + "epoch": 21.83, + "learning_rate": 1.363518160016181e-05, + "loss": 0.9663, + "step": 252210 + }, + { + "epoch": 21.83, + "learning_rate": 1.3633736889248461e-05, + "loss": 0.9367, + "step": 252220 + }, + { + "epoch": 21.83, + "learning_rate": 1.3632292178335116e-05, + "loss": 0.9068, + "step": 252230 + }, + { + "epoch": 21.83, + "learning_rate": 1.363084746742177e-05, + "loss": 0.9175, + "step": 252240 + }, + { + "epoch": 21.83, + "learning_rate": 1.3629402756508423e-05, + "loss": 0.8889, + "step": 252250 + }, + { + "epoch": 21.83, + "learning_rate": 1.3627958045595076e-05, + "loss": 0.9138, + "step": 252260 + }, + { + "epoch": 21.84, + "learning_rate": 1.3626513334681731e-05, + "loss": 0.945, + "step": 252270 + }, + { + "epoch": 21.84, + "learning_rate": 1.3625068623768383e-05, + "loss": 0.9306, + "step": 252280 + }, + { + "epoch": 21.84, + "learning_rate": 1.3623623912855038e-05, + "loss": 0.9551, + "step": 252290 + }, + { + "epoch": 21.84, + "learning_rate": 1.3622179201941693e-05, + "loss": 0.9639, + "step": 252300 + }, + { + "epoch": 21.84, + "learning_rate": 1.3620734491028345e-05, + "loss": 0.9836, + "step": 252310 + }, + { + "epoch": 21.84, + "learning_rate": 1.3619289780115e-05, + "loss": 0.9968, + "step": 252320 + }, + { + "epoch": 21.84, + "learning_rate": 1.3617845069201653e-05, + "loss": 0.9733, + "step": 252330 + }, + { + "epoch": 21.84, + "learning_rate": 1.3616400358288307e-05, + "loss": 0.9041, + "step": 252340 + }, + { + "epoch": 21.84, + "learning_rate": 1.361495564737496e-05, + "loss": 0.9627, + "step": 252350 + }, + { + "epoch": 21.84, + "learning_rate": 1.3613510936461615e-05, + "loss": 0.9312, + "step": 252360 + }, + { + "epoch": 21.84, + "learning_rate": 1.3612066225548267e-05, + "loss": 0.857, + "step": 252370 + }, + { + "epoch": 21.85, + "learning_rate": 1.3610621514634922e-05, + "loss": 0.9362, + "step": 252380 + }, + { + "epoch": 21.85, + "learning_rate": 1.3609176803721577e-05, + "loss": 0.9061, + "step": 252390 + }, + { + "epoch": 21.85, + "learning_rate": 1.3607732092808229e-05, + "loss": 0.9378, + "step": 252400 + }, + { + "epoch": 21.85, + "learning_rate": 1.3606287381894884e-05, + "loss": 0.9374, + "step": 252410 + }, + { + "epoch": 21.85, + "learning_rate": 1.3604842670981537e-05, + "loss": 0.9821, + "step": 252420 + }, + { + "epoch": 21.85, + "learning_rate": 1.360339796006819e-05, + "loss": 0.9637, + "step": 252430 + }, + { + "epoch": 21.85, + "learning_rate": 1.3601953249154844e-05, + "loss": 0.9134, + "step": 252440 + }, + { + "epoch": 21.85, + "learning_rate": 1.36005085382415e-05, + "loss": 0.9724, + "step": 252450 + }, + { + "epoch": 21.85, + "learning_rate": 1.3599063827328151e-05, + "loss": 0.9178, + "step": 252460 + }, + { + "epoch": 21.85, + "learning_rate": 1.3597619116414806e-05, + "loss": 0.9423, + "step": 252470 + }, + { + "epoch": 21.85, + "learning_rate": 1.3596174405501461e-05, + "loss": 0.9761, + "step": 252480 + }, + { + "epoch": 21.85, + "learning_rate": 1.3594729694588113e-05, + "loss": 0.9093, + "step": 252490 + }, + { + "epoch": 21.86, + "learning_rate": 1.3593284983674768e-05, + "loss": 0.9439, + "step": 252500 + }, + { + "epoch": 21.86, + "learning_rate": 1.3591840272761421e-05, + "loss": 0.9096, + "step": 252510 + }, + { + "epoch": 21.86, + "learning_rate": 1.3590395561848075e-05, + "loss": 0.9597, + "step": 252520 + }, + { + "epoch": 21.86, + "learning_rate": 1.3588950850934728e-05, + "loss": 0.9529, + "step": 252530 + }, + { + "epoch": 21.86, + "learning_rate": 1.3587506140021383e-05, + "loss": 0.9654, + "step": 252540 + }, + { + "epoch": 21.86, + "learning_rate": 1.3586061429108035e-05, + "loss": 0.9599, + "step": 252550 + }, + { + "epoch": 21.86, + "learning_rate": 1.358461671819469e-05, + "loss": 0.9554, + "step": 252560 + }, + { + "epoch": 21.86, + "learning_rate": 1.3583172007281345e-05, + "loss": 0.8979, + "step": 252570 + }, + { + "epoch": 21.86, + "learning_rate": 1.3581727296367997e-05, + "loss": 0.9225, + "step": 252580 + }, + { + "epoch": 21.86, + "learning_rate": 1.3580282585454652e-05, + "loss": 0.9292, + "step": 252590 + }, + { + "epoch": 21.86, + "learning_rate": 1.3578837874541305e-05, + "loss": 0.9746, + "step": 252600 + }, + { + "epoch": 21.87, + "learning_rate": 1.3577393163627959e-05, + "loss": 0.9735, + "step": 252610 + }, + { + "epoch": 21.87, + "learning_rate": 1.3575948452714612e-05, + "loss": 0.9626, + "step": 252620 + }, + { + "epoch": 21.87, + "learning_rate": 1.3574503741801267e-05, + "loss": 0.8438, + "step": 252630 + }, + { + "epoch": 21.87, + "learning_rate": 1.3573059030887919e-05, + "loss": 0.9775, + "step": 252640 + }, + { + "epoch": 21.87, + "learning_rate": 1.3571614319974574e-05, + "loss": 0.9099, + "step": 252650 + }, + { + "epoch": 21.87, + "learning_rate": 1.3570169609061229e-05, + "loss": 0.9529, + "step": 252660 + }, + { + "epoch": 21.87, + "learning_rate": 1.356872489814788e-05, + "loss": 0.976, + "step": 252670 + }, + { + "epoch": 21.87, + "learning_rate": 1.3567280187234536e-05, + "loss": 0.9563, + "step": 252680 + }, + { + "epoch": 21.87, + "learning_rate": 1.3565835476321189e-05, + "loss": 0.9284, + "step": 252690 + }, + { + "epoch": 21.87, + "learning_rate": 1.3564390765407843e-05, + "loss": 1.0009, + "step": 252700 + }, + { + "epoch": 21.87, + "learning_rate": 1.3562946054494496e-05, + "loss": 0.9066, + "step": 252710 + }, + { + "epoch": 21.87, + "learning_rate": 1.3561501343581151e-05, + "loss": 0.8749, + "step": 252720 + }, + { + "epoch": 21.88, + "learning_rate": 1.3560056632667803e-05, + "loss": 0.9412, + "step": 252730 + }, + { + "epoch": 21.88, + "learning_rate": 1.3558611921754458e-05, + "loss": 0.9632, + "step": 252740 + }, + { + "epoch": 21.88, + "learning_rate": 1.3557167210841113e-05, + "loss": 0.9778, + "step": 252750 + }, + { + "epoch": 21.88, + "learning_rate": 1.3555722499927765e-05, + "loss": 0.9675, + "step": 252760 + }, + { + "epoch": 21.88, + "learning_rate": 1.355427778901442e-05, + "loss": 0.9439, + "step": 252770 + }, + { + "epoch": 21.88, + "learning_rate": 1.3552833078101073e-05, + "loss": 0.9403, + "step": 252780 + }, + { + "epoch": 21.88, + "learning_rate": 1.3551388367187726e-05, + "loss": 0.9528, + "step": 252790 + }, + { + "epoch": 21.88, + "learning_rate": 1.354994365627438e-05, + "loss": 0.9079, + "step": 252800 + }, + { + "epoch": 21.88, + "learning_rate": 1.3548498945361035e-05, + "loss": 0.9455, + "step": 252810 + }, + { + "epoch": 21.88, + "learning_rate": 1.3547054234447687e-05, + "loss": 0.9639, + "step": 252820 + }, + { + "epoch": 21.88, + "learning_rate": 1.3545609523534342e-05, + "loss": 0.9379, + "step": 252830 + }, + { + "epoch": 21.89, + "learning_rate": 1.3544164812620997e-05, + "loss": 0.9179, + "step": 252840 + }, + { + "epoch": 21.89, + "learning_rate": 1.3542720101707649e-05, + "loss": 0.9986, + "step": 252850 + }, + { + "epoch": 21.89, + "learning_rate": 1.3541275390794304e-05, + "loss": 0.9037, + "step": 252860 + }, + { + "epoch": 21.89, + "learning_rate": 1.3539830679880957e-05, + "loss": 0.9999, + "step": 252870 + }, + { + "epoch": 21.89, + "learning_rate": 1.353838596896761e-05, + "loss": 0.9621, + "step": 252880 + }, + { + "epoch": 21.89, + "learning_rate": 1.3536941258054264e-05, + "loss": 0.9151, + "step": 252890 + }, + { + "epoch": 21.89, + "learning_rate": 1.3535496547140919e-05, + "loss": 0.9458, + "step": 252900 + }, + { + "epoch": 21.89, + "learning_rate": 1.353405183622757e-05, + "loss": 0.8931, + "step": 252910 + }, + { + "epoch": 21.89, + "learning_rate": 1.3532607125314226e-05, + "loss": 0.9669, + "step": 252920 + }, + { + "epoch": 21.89, + "learning_rate": 1.353116241440088e-05, + "loss": 0.939, + "step": 252930 + }, + { + "epoch": 21.89, + "learning_rate": 1.3529717703487532e-05, + "loss": 0.903, + "step": 252940 + }, + { + "epoch": 21.89, + "learning_rate": 1.3528272992574188e-05, + "loss": 0.9304, + "step": 252950 + }, + { + "epoch": 21.9, + "learning_rate": 1.3526828281660841e-05, + "loss": 0.9332, + "step": 252960 + }, + { + "epoch": 21.9, + "learning_rate": 1.3525383570747494e-05, + "loss": 1.0124, + "step": 252970 + }, + { + "epoch": 21.9, + "learning_rate": 1.3523938859834148e-05, + "loss": 0.9246, + "step": 252980 + }, + { + "epoch": 21.9, + "learning_rate": 1.3522494148920803e-05, + "loss": 0.925, + "step": 252990 + }, + { + "epoch": 21.9, + "learning_rate": 1.3521049438007455e-05, + "loss": 0.87, + "step": 253000 + }, + { + "epoch": 21.9, + "learning_rate": 1.351960472709411e-05, + "loss": 0.9529, + "step": 253010 + }, + { + "epoch": 21.9, + "learning_rate": 1.3518160016180765e-05, + "loss": 0.9502, + "step": 253020 + }, + { + "epoch": 21.9, + "learning_rate": 1.3516715305267416e-05, + "loss": 0.9851, + "step": 253030 + }, + { + "epoch": 21.9, + "learning_rate": 1.3515270594354071e-05, + "loss": 0.8976, + "step": 253040 + }, + { + "epoch": 21.9, + "learning_rate": 1.3513825883440725e-05, + "loss": 0.9562, + "step": 253050 + }, + { + "epoch": 21.9, + "learning_rate": 1.3512381172527378e-05, + "loss": 0.9845, + "step": 253060 + }, + { + "epoch": 21.91, + "learning_rate": 1.3510936461614032e-05, + "loss": 0.949, + "step": 253070 + }, + { + "epoch": 21.91, + "learning_rate": 1.3509491750700683e-05, + "loss": 0.9292, + "step": 253080 + }, + { + "epoch": 21.91, + "learning_rate": 1.3508047039787338e-05, + "loss": 0.8619, + "step": 253090 + }, + { + "epoch": 21.91, + "learning_rate": 1.3506602328873994e-05, + "loss": 0.9992, + "step": 253100 + }, + { + "epoch": 21.91, + "learning_rate": 1.3505157617960645e-05, + "loss": 0.9346, + "step": 253110 + }, + { + "epoch": 21.91, + "learning_rate": 1.35037129070473e-05, + "loss": 0.9212, + "step": 253120 + }, + { + "epoch": 21.91, + "learning_rate": 1.3502268196133955e-05, + "loss": 0.9027, + "step": 253130 + }, + { + "epoch": 21.91, + "learning_rate": 1.3500823485220607e-05, + "loss": 0.9985, + "step": 253140 + }, + { + "epoch": 21.91, + "learning_rate": 1.3499378774307262e-05, + "loss": 0.9561, + "step": 253150 + }, + { + "epoch": 21.91, + "learning_rate": 1.3497934063393916e-05, + "loss": 0.934, + "step": 253160 + }, + { + "epoch": 21.91, + "learning_rate": 1.3496489352480567e-05, + "loss": 0.9097, + "step": 253170 + }, + { + "epoch": 21.91, + "learning_rate": 1.3495044641567222e-05, + "loss": 0.9044, + "step": 253180 + }, + { + "epoch": 21.92, + "learning_rate": 1.3493599930653877e-05, + "loss": 0.9446, + "step": 253190 + }, + { + "epoch": 21.92, + "learning_rate": 1.349215521974053e-05, + "loss": 0.9473, + "step": 253200 + }, + { + "epoch": 21.92, + "learning_rate": 1.3490710508827184e-05, + "loss": 0.9549, + "step": 253210 + }, + { + "epoch": 21.92, + "learning_rate": 1.348926579791384e-05, + "loss": 0.9417, + "step": 253220 + }, + { + "epoch": 21.92, + "learning_rate": 1.3487821087000491e-05, + "loss": 0.9357, + "step": 253230 + }, + { + "epoch": 21.92, + "learning_rate": 1.3486376376087146e-05, + "loss": 0.9146, + "step": 253240 + }, + { + "epoch": 21.92, + "learning_rate": 1.34849316651738e-05, + "loss": 0.9355, + "step": 253250 + }, + { + "epoch": 21.92, + "learning_rate": 1.3483486954260451e-05, + "loss": 0.9249, + "step": 253260 + }, + { + "epoch": 21.92, + "learning_rate": 1.3482042243347106e-05, + "loss": 0.9438, + "step": 253270 + }, + { + "epoch": 21.92, + "learning_rate": 1.3480597532433761e-05, + "loss": 0.9347, + "step": 253280 + }, + { + "epoch": 21.92, + "learning_rate": 1.3479152821520413e-05, + "loss": 0.9115, + "step": 253290 + }, + { + "epoch": 21.92, + "learning_rate": 1.3477708110607068e-05, + "loss": 0.9659, + "step": 253300 + }, + { + "epoch": 21.93, + "learning_rate": 1.3476263399693723e-05, + "loss": 0.9229, + "step": 253310 + }, + { + "epoch": 21.93, + "learning_rate": 1.3474818688780375e-05, + "loss": 0.9256, + "step": 253320 + }, + { + "epoch": 21.93, + "learning_rate": 1.347337397786703e-05, + "loss": 0.957, + "step": 253330 + }, + { + "epoch": 21.93, + "learning_rate": 1.3471929266953683e-05, + "loss": 0.8998, + "step": 253340 + }, + { + "epoch": 21.93, + "learning_rate": 1.3470484556040335e-05, + "loss": 0.9458, + "step": 253350 + }, + { + "epoch": 21.93, + "learning_rate": 1.346903984512699e-05, + "loss": 0.9292, + "step": 253360 + }, + { + "epoch": 21.93, + "learning_rate": 1.3467595134213645e-05, + "loss": 0.889, + "step": 253370 + }, + { + "epoch": 21.93, + "learning_rate": 1.3466150423300297e-05, + "loss": 0.9102, + "step": 253380 + }, + { + "epoch": 21.93, + "learning_rate": 1.3464705712386952e-05, + "loss": 0.9281, + "step": 253390 + }, + { + "epoch": 21.93, + "learning_rate": 1.3463261001473607e-05, + "loss": 0.9205, + "step": 253400 + }, + { + "epoch": 21.93, + "learning_rate": 1.3461816290560259e-05, + "loss": 0.9671, + "step": 253410 + }, + { + "epoch": 21.94, + "learning_rate": 1.3460371579646914e-05, + "loss": 0.902, + "step": 253420 + }, + { + "epoch": 21.94, + "learning_rate": 1.3458926868733567e-05, + "loss": 0.934, + "step": 253430 + }, + { + "epoch": 21.94, + "learning_rate": 1.3457482157820219e-05, + "loss": 0.9326, + "step": 253440 + }, + { + "epoch": 21.94, + "learning_rate": 1.3456037446906874e-05, + "loss": 0.8856, + "step": 253450 + }, + { + "epoch": 21.94, + "learning_rate": 1.345459273599353e-05, + "loss": 0.9287, + "step": 253460 + }, + { + "epoch": 21.94, + "learning_rate": 1.3453148025080181e-05, + "loss": 0.9517, + "step": 253470 + }, + { + "epoch": 21.94, + "learning_rate": 1.3451703314166836e-05, + "loss": 0.9974, + "step": 253480 + }, + { + "epoch": 21.94, + "learning_rate": 1.3450258603253491e-05, + "loss": 0.9032, + "step": 253490 + }, + { + "epoch": 21.94, + "learning_rate": 1.3448813892340143e-05, + "loss": 0.9555, + "step": 253500 + }, + { + "epoch": 21.94, + "learning_rate": 1.3447369181426796e-05, + "loss": 0.9599, + "step": 253510 + }, + { + "epoch": 21.94, + "learning_rate": 1.3445924470513451e-05, + "loss": 0.9047, + "step": 253520 + }, + { + "epoch": 21.94, + "learning_rate": 1.3444479759600103e-05, + "loss": 0.9192, + "step": 253530 + }, + { + "epoch": 21.95, + "learning_rate": 1.3443035048686758e-05, + "loss": 0.8795, + "step": 253540 + }, + { + "epoch": 21.95, + "learning_rate": 1.3441590337773413e-05, + "loss": 0.9018, + "step": 253550 + }, + { + "epoch": 21.95, + "learning_rate": 1.3440145626860065e-05, + "loss": 1.0008, + "step": 253560 + }, + { + "epoch": 21.95, + "learning_rate": 1.343870091594672e-05, + "loss": 0.9254, + "step": 253570 + }, + { + "epoch": 21.95, + "learning_rate": 1.3437256205033375e-05, + "loss": 0.9178, + "step": 253580 + }, + { + "epoch": 21.95, + "learning_rate": 1.3435811494120027e-05, + "loss": 0.9022, + "step": 253590 + }, + { + "epoch": 21.95, + "learning_rate": 1.343436678320668e-05, + "loss": 0.9586, + "step": 253600 + }, + { + "epoch": 21.95, + "learning_rate": 1.3432922072293335e-05, + "loss": 0.9556, + "step": 253610 + }, + { + "epoch": 21.95, + "learning_rate": 1.3431477361379987e-05, + "loss": 0.8746, + "step": 253620 + }, + { + "epoch": 21.95, + "learning_rate": 1.3430032650466642e-05, + "loss": 0.9735, + "step": 253630 + }, + { + "epoch": 21.95, + "learning_rate": 1.3428587939553297e-05, + "loss": 0.9227, + "step": 253640 + }, + { + "epoch": 21.96, + "learning_rate": 1.3427143228639949e-05, + "loss": 0.87, + "step": 253650 + }, + { + "epoch": 21.96, + "learning_rate": 1.3425698517726604e-05, + "loss": 0.9077, + "step": 253660 + }, + { + "epoch": 21.96, + "learning_rate": 1.3424253806813259e-05, + "loss": 0.8978, + "step": 253670 + }, + { + "epoch": 21.96, + "learning_rate": 1.342280909589991e-05, + "loss": 0.9416, + "step": 253680 + }, + { + "epoch": 21.96, + "learning_rate": 1.3421364384986564e-05, + "loss": 0.9468, + "step": 253690 + }, + { + "epoch": 21.96, + "learning_rate": 1.341991967407322e-05, + "loss": 0.9448, + "step": 253700 + }, + { + "epoch": 21.96, + "learning_rate": 1.3418474963159871e-05, + "loss": 0.9051, + "step": 253710 + }, + { + "epoch": 21.96, + "learning_rate": 1.3417030252246526e-05, + "loss": 0.9185, + "step": 253720 + }, + { + "epoch": 21.96, + "learning_rate": 1.3415585541333181e-05, + "loss": 0.9302, + "step": 253730 + }, + { + "epoch": 21.96, + "learning_rate": 1.3414140830419833e-05, + "loss": 0.8988, + "step": 253740 + }, + { + "epoch": 21.96, + "learning_rate": 1.3412696119506488e-05, + "loss": 0.9787, + "step": 253750 + }, + { + "epoch": 21.96, + "learning_rate": 1.3411251408593143e-05, + "loss": 0.9163, + "step": 253760 + }, + { + "epoch": 21.97, + "learning_rate": 1.3409806697679795e-05, + "loss": 1.0078, + "step": 253770 + }, + { + "epoch": 21.97, + "learning_rate": 1.3408361986766448e-05, + "loss": 0.9265, + "step": 253780 + }, + { + "epoch": 21.97, + "learning_rate": 1.3406917275853103e-05, + "loss": 0.913, + "step": 253790 + }, + { + "epoch": 21.97, + "learning_rate": 1.3405472564939755e-05, + "loss": 0.962, + "step": 253800 + }, + { + "epoch": 21.97, + "learning_rate": 1.340402785402641e-05, + "loss": 0.9177, + "step": 253810 + }, + { + "epoch": 21.97, + "learning_rate": 1.3402583143113065e-05, + "loss": 0.9588, + "step": 253820 + }, + { + "epoch": 21.97, + "learning_rate": 1.3401138432199717e-05, + "loss": 0.9582, + "step": 253830 + }, + { + "epoch": 21.97, + "learning_rate": 1.3399693721286372e-05, + "loss": 0.9448, + "step": 253840 + }, + { + "epoch": 21.97, + "learning_rate": 1.3398249010373027e-05, + "loss": 0.9286, + "step": 253850 + }, + { + "epoch": 21.97, + "learning_rate": 1.3396804299459679e-05, + "loss": 0.9523, + "step": 253860 + }, + { + "epoch": 21.97, + "learning_rate": 1.3395359588546332e-05, + "loss": 0.8934, + "step": 253870 + }, + { + "epoch": 21.98, + "learning_rate": 1.3393914877632987e-05, + "loss": 0.9122, + "step": 253880 + }, + { + "epoch": 21.98, + "learning_rate": 1.3392470166719639e-05, + "loss": 0.977, + "step": 253890 + }, + { + "epoch": 21.98, + "learning_rate": 1.3391025455806294e-05, + "loss": 0.9377, + "step": 253900 + }, + { + "epoch": 21.98, + "learning_rate": 1.3389580744892949e-05, + "loss": 0.9979, + "step": 253910 + }, + { + "epoch": 21.98, + "learning_rate": 1.33881360339796e-05, + "loss": 0.94, + "step": 253920 + }, + { + "epoch": 21.98, + "learning_rate": 1.3386691323066256e-05, + "loss": 0.9607, + "step": 253930 + }, + { + "epoch": 21.98, + "learning_rate": 1.3385246612152911e-05, + "loss": 0.9843, + "step": 253940 + }, + { + "epoch": 21.98, + "learning_rate": 1.3383801901239563e-05, + "loss": 0.9381, + "step": 253950 + }, + { + "epoch": 21.98, + "learning_rate": 1.3382357190326216e-05, + "loss": 0.993, + "step": 253960 + }, + { + "epoch": 21.98, + "learning_rate": 1.338091247941287e-05, + "loss": 0.9955, + "step": 253970 + }, + { + "epoch": 21.98, + "learning_rate": 1.3379467768499523e-05, + "loss": 0.9659, + "step": 253980 + }, + { + "epoch": 21.98, + "learning_rate": 1.3378023057586178e-05, + "loss": 0.9185, + "step": 253990 + }, + { + "epoch": 21.99, + "learning_rate": 1.337657834667283e-05, + "loss": 0.9127, + "step": 254000 + }, + { + "epoch": 21.99, + "learning_rate": 1.3375133635759485e-05, + "loss": 0.9211, + "step": 254010 + }, + { + "epoch": 21.99, + "learning_rate": 1.337368892484614e-05, + "loss": 0.8645, + "step": 254020 + }, + { + "epoch": 21.99, + "learning_rate": 1.3372244213932791e-05, + "loss": 0.9688, + "step": 254030 + }, + { + "epoch": 21.99, + "learning_rate": 1.3370799503019447e-05, + "loss": 0.9016, + "step": 254040 + }, + { + "epoch": 21.99, + "learning_rate": 1.33693547921061e-05, + "loss": 0.9014, + "step": 254050 + }, + { + "epoch": 21.99, + "learning_rate": 1.3367910081192753e-05, + "loss": 0.9475, + "step": 254060 + }, + { + "epoch": 21.99, + "learning_rate": 1.3366465370279407e-05, + "loss": 0.9143, + "step": 254070 + }, + { + "epoch": 21.99, + "learning_rate": 1.3365020659366062e-05, + "loss": 0.8851, + "step": 254080 + }, + { + "epoch": 21.99, + "learning_rate": 1.3363575948452714e-05, + "loss": 0.9371, + "step": 254090 + }, + { + "epoch": 21.99, + "learning_rate": 1.3362131237539369e-05, + "loss": 0.9151, + "step": 254100 + }, + { + "epoch": 22.0, + "learning_rate": 1.3360686526626024e-05, + "loss": 0.8891, + "step": 254110 + }, + { + "epoch": 22.0, + "learning_rate": 1.3359241815712675e-05, + "loss": 0.8715, + "step": 254120 + }, + { + "epoch": 22.0, + "learning_rate": 1.335779710479933e-05, + "loss": 0.9455, + "step": 254130 + }, + { + "epoch": 22.0, + "learning_rate": 1.3356352393885984e-05, + "loss": 0.9093, + "step": 254140 + }, + { + "epoch": 22.0, + "learning_rate": 1.3354907682972637e-05, + "loss": 0.9346, + "step": 254150 + }, + { + "epoch": 22.0, + "learning_rate": 1.335346297205929e-05, + "loss": 0.939, + "step": 254160 + }, + { + "epoch": 22.0, + "eval_cer": 0.9931778335330519, + "eval_em": 0, + "eval_f1": 0.11718386329668186, + "eval_loss": 0.9457400441169739, + "eval_runtime": 1320.1209, + "eval_samples_per_second": 3.89, + "eval_steps_per_second": 0.486, + "eval_wer": 0.9676871137258075, + "step": 254167 + }, + { + "epoch": 22.0, + "learning_rate": 1.3352018261145946e-05, + "loss": 0.9105, + "step": 254170 + }, + { + "epoch": 22.0, + "learning_rate": 1.3350573550232597e-05, + "loss": 0.9177, + "step": 254180 + }, + { + "epoch": 22.0, + "learning_rate": 1.3349128839319253e-05, + "loss": 0.8954, + "step": 254190 + }, + { + "epoch": 22.0, + "learning_rate": 1.3347684128405908e-05, + "loss": 0.9096, + "step": 254200 + }, + { + "epoch": 22.0, + "learning_rate": 1.334623941749256e-05, + "loss": 0.9265, + "step": 254210 + }, + { + "epoch": 22.0, + "learning_rate": 1.3344794706579214e-05, + "loss": 0.9467, + "step": 254220 + }, + { + "epoch": 22.01, + "learning_rate": 1.3343349995665868e-05, + "loss": 1.022, + "step": 254230 + }, + { + "epoch": 22.01, + "learning_rate": 1.3341905284752521e-05, + "loss": 0.8978, + "step": 254240 + }, + { + "epoch": 22.01, + "learning_rate": 1.3340460573839175e-05, + "loss": 0.9351, + "step": 254250 + }, + { + "epoch": 22.01, + "learning_rate": 1.333901586292583e-05, + "loss": 0.9497, + "step": 254260 + }, + { + "epoch": 22.01, + "learning_rate": 1.3337571152012481e-05, + "loss": 0.9537, + "step": 254270 + }, + { + "epoch": 22.01, + "learning_rate": 1.3336126441099137e-05, + "loss": 0.8913, + "step": 254280 + }, + { + "epoch": 22.01, + "learning_rate": 1.3334681730185792e-05, + "loss": 0.9897, + "step": 254290 + }, + { + "epoch": 22.01, + "learning_rate": 1.3333237019272443e-05, + "loss": 0.9483, + "step": 254300 + }, + { + "epoch": 22.01, + "learning_rate": 1.3331792308359098e-05, + "loss": 0.8896, + "step": 254310 + }, + { + "epoch": 22.01, + "learning_rate": 1.3330347597445752e-05, + "loss": 0.9533, + "step": 254320 + }, + { + "epoch": 22.01, + "learning_rate": 1.3328902886532405e-05, + "loss": 0.9533, + "step": 254330 + }, + { + "epoch": 22.01, + "learning_rate": 1.3327458175619059e-05, + "loss": 0.9183, + "step": 254340 + }, + { + "epoch": 22.02, + "learning_rate": 1.3326013464705714e-05, + "loss": 0.936, + "step": 254350 + }, + { + "epoch": 22.02, + "learning_rate": 1.3324568753792365e-05, + "loss": 0.9632, + "step": 254360 + }, + { + "epoch": 22.02, + "learning_rate": 1.332312404287902e-05, + "loss": 0.9895, + "step": 254370 + }, + { + "epoch": 22.02, + "learning_rate": 1.3321679331965676e-05, + "loss": 0.9161, + "step": 254380 + }, + { + "epoch": 22.02, + "learning_rate": 1.3320234621052327e-05, + "loss": 0.8771, + "step": 254390 + }, + { + "epoch": 22.02, + "learning_rate": 1.3318789910138982e-05, + "loss": 0.9621, + "step": 254400 + }, + { + "epoch": 22.02, + "learning_rate": 1.3317345199225636e-05, + "loss": 0.9168, + "step": 254410 + }, + { + "epoch": 22.02, + "learning_rate": 1.3315900488312289e-05, + "loss": 0.9465, + "step": 254420 + }, + { + "epoch": 22.02, + "learning_rate": 1.3314455777398943e-05, + "loss": 0.9734, + "step": 254430 + }, + { + "epoch": 22.02, + "learning_rate": 1.3313011066485598e-05, + "loss": 0.9432, + "step": 254440 + }, + { + "epoch": 22.02, + "learning_rate": 1.331156635557225e-05, + "loss": 0.892, + "step": 254450 + }, + { + "epoch": 22.03, + "learning_rate": 1.3310121644658904e-05, + "loss": 0.923, + "step": 254460 + }, + { + "epoch": 22.03, + "learning_rate": 1.330867693374556e-05, + "loss": 0.8908, + "step": 254470 + }, + { + "epoch": 22.03, + "learning_rate": 1.3307232222832211e-05, + "loss": 0.9417, + "step": 254480 + }, + { + "epoch": 22.03, + "learning_rate": 1.3305787511918866e-05, + "loss": 0.8899, + "step": 254490 + }, + { + "epoch": 22.03, + "learning_rate": 1.330434280100552e-05, + "loss": 0.9164, + "step": 254500 + }, + { + "epoch": 22.03, + "learning_rate": 1.3302898090092173e-05, + "loss": 0.9718, + "step": 254510 + }, + { + "epoch": 22.03, + "learning_rate": 1.3301453379178826e-05, + "loss": 0.9155, + "step": 254520 + }, + { + "epoch": 22.03, + "learning_rate": 1.3300008668265482e-05, + "loss": 0.9163, + "step": 254530 + }, + { + "epoch": 22.03, + "learning_rate": 1.3298563957352133e-05, + "loss": 0.9059, + "step": 254540 + }, + { + "epoch": 22.03, + "learning_rate": 1.3297119246438788e-05, + "loss": 0.923, + "step": 254550 + }, + { + "epoch": 22.03, + "learning_rate": 1.3295674535525443e-05, + "loss": 0.9466, + "step": 254560 + }, + { + "epoch": 22.03, + "learning_rate": 1.3294229824612095e-05, + "loss": 0.9003, + "step": 254570 + }, + { + "epoch": 22.04, + "learning_rate": 1.329278511369875e-05, + "loss": 0.9207, + "step": 254580 + }, + { + "epoch": 22.04, + "learning_rate": 1.3291340402785404e-05, + "loss": 0.927, + "step": 254590 + }, + { + "epoch": 22.04, + "learning_rate": 1.3289895691872057e-05, + "loss": 0.9651, + "step": 254600 + }, + { + "epoch": 22.04, + "learning_rate": 1.328845098095871e-05, + "loss": 0.9264, + "step": 254610 + }, + { + "epoch": 22.04, + "learning_rate": 1.3287006270045365e-05, + "loss": 0.8889, + "step": 254620 + }, + { + "epoch": 22.04, + "learning_rate": 1.3285561559132017e-05, + "loss": 0.954, + "step": 254630 + }, + { + "epoch": 22.04, + "learning_rate": 1.3284116848218672e-05, + "loss": 0.9636, + "step": 254640 + }, + { + "epoch": 22.04, + "learning_rate": 1.3282672137305327e-05, + "loss": 0.8373, + "step": 254650 + }, + { + "epoch": 22.04, + "learning_rate": 1.3281227426391979e-05, + "loss": 0.9641, + "step": 254660 + }, + { + "epoch": 22.04, + "learning_rate": 1.3279782715478634e-05, + "loss": 0.9319, + "step": 254670 + }, + { + "epoch": 22.04, + "learning_rate": 1.3278338004565288e-05, + "loss": 0.9303, + "step": 254680 + }, + { + "epoch": 22.05, + "learning_rate": 1.3276893293651941e-05, + "loss": 0.9039, + "step": 254690 + }, + { + "epoch": 22.05, + "learning_rate": 1.3275448582738594e-05, + "loss": 0.9102, + "step": 254700 + }, + { + "epoch": 22.05, + "learning_rate": 1.327400387182525e-05, + "loss": 0.9013, + "step": 254710 + }, + { + "epoch": 22.05, + "learning_rate": 1.3272559160911901e-05, + "loss": 0.9166, + "step": 254720 + }, + { + "epoch": 22.05, + "learning_rate": 1.3271114449998556e-05, + "loss": 0.8978, + "step": 254730 + }, + { + "epoch": 22.05, + "learning_rate": 1.3269669739085211e-05, + "loss": 0.9425, + "step": 254740 + }, + { + "epoch": 22.05, + "learning_rate": 1.3268225028171863e-05, + "loss": 0.9371, + "step": 254750 + }, + { + "epoch": 22.05, + "learning_rate": 1.3266780317258518e-05, + "loss": 0.9285, + "step": 254760 + }, + { + "epoch": 22.05, + "learning_rate": 1.3265335606345171e-05, + "loss": 0.9179, + "step": 254770 + }, + { + "epoch": 22.05, + "learning_rate": 1.3263890895431825e-05, + "loss": 0.873, + "step": 254780 + }, + { + "epoch": 22.05, + "learning_rate": 1.3262446184518478e-05, + "loss": 0.9325, + "step": 254790 + }, + { + "epoch": 22.05, + "learning_rate": 1.3261001473605133e-05, + "loss": 0.982, + "step": 254800 + }, + { + "epoch": 22.06, + "learning_rate": 1.3259556762691785e-05, + "loss": 0.9255, + "step": 254810 + }, + { + "epoch": 22.06, + "learning_rate": 1.325811205177844e-05, + "loss": 0.9343, + "step": 254820 + }, + { + "epoch": 22.06, + "learning_rate": 1.3256667340865095e-05, + "loss": 0.9771, + "step": 254830 + }, + { + "epoch": 22.06, + "learning_rate": 1.3255222629951747e-05, + "loss": 0.8711, + "step": 254840 + }, + { + "epoch": 22.06, + "learning_rate": 1.3253777919038402e-05, + "loss": 0.9399, + "step": 254850 + }, + { + "epoch": 22.06, + "learning_rate": 1.3252333208125055e-05, + "loss": 0.8857, + "step": 254860 + }, + { + "epoch": 22.06, + "learning_rate": 1.3250888497211709e-05, + "loss": 0.9552, + "step": 254870 + }, + { + "epoch": 22.06, + "learning_rate": 1.3249443786298362e-05, + "loss": 0.9307, + "step": 254880 + }, + { + "epoch": 22.06, + "learning_rate": 1.3247999075385014e-05, + "loss": 0.9363, + "step": 254890 + }, + { + "epoch": 22.06, + "learning_rate": 1.3246554364471669e-05, + "loss": 0.9468, + "step": 254900 + }, + { + "epoch": 22.06, + "learning_rate": 1.3245109653558324e-05, + "loss": 0.9676, + "step": 254910 + }, + { + "epoch": 22.07, + "learning_rate": 1.3243664942644976e-05, + "loss": 0.8822, + "step": 254920 + }, + { + "epoch": 22.07, + "learning_rate": 1.3242220231731631e-05, + "loss": 0.8984, + "step": 254930 + }, + { + "epoch": 22.07, + "learning_rate": 1.3240775520818286e-05, + "loss": 0.8915, + "step": 254940 + }, + { + "epoch": 22.07, + "learning_rate": 1.3239330809904938e-05, + "loss": 0.9301, + "step": 254950 + }, + { + "epoch": 22.07, + "learning_rate": 1.3237886098991593e-05, + "loss": 0.9975, + "step": 254960 + }, + { + "epoch": 22.07, + "learning_rate": 1.3236441388078246e-05, + "loss": 0.9169, + "step": 254970 + }, + { + "epoch": 22.07, + "learning_rate": 1.3234996677164898e-05, + "loss": 0.9947, + "step": 254980 + }, + { + "epoch": 22.07, + "learning_rate": 1.3233551966251553e-05, + "loss": 0.9002, + "step": 254990 + }, + { + "epoch": 22.07, + "learning_rate": 1.3232107255338208e-05, + "loss": 0.9632, + "step": 255000 + }, + { + "epoch": 22.07, + "learning_rate": 1.323066254442486e-05, + "loss": 0.9254, + "step": 255010 + }, + { + "epoch": 22.07, + "learning_rate": 1.3229217833511515e-05, + "loss": 0.9067, + "step": 255020 + }, + { + "epoch": 22.07, + "learning_rate": 1.322777312259817e-05, + "loss": 0.9138, + "step": 255030 + }, + { + "epoch": 22.08, + "learning_rate": 1.3226328411684822e-05, + "loss": 0.9427, + "step": 255040 + }, + { + "epoch": 22.08, + "learning_rate": 1.3224883700771477e-05, + "loss": 0.9492, + "step": 255050 + }, + { + "epoch": 22.08, + "learning_rate": 1.322343898985813e-05, + "loss": 0.9338, + "step": 255060 + }, + { + "epoch": 22.08, + "learning_rate": 1.3221994278944782e-05, + "loss": 0.9085, + "step": 255070 + }, + { + "epoch": 22.08, + "learning_rate": 1.3220549568031437e-05, + "loss": 0.9283, + "step": 255080 + }, + { + "epoch": 22.08, + "learning_rate": 1.3219104857118092e-05, + "loss": 0.9414, + "step": 255090 + }, + { + "epoch": 22.08, + "learning_rate": 1.3217660146204744e-05, + "loss": 0.9405, + "step": 255100 + }, + { + "epoch": 22.08, + "learning_rate": 1.3216215435291399e-05, + "loss": 0.9251, + "step": 255110 + }, + { + "epoch": 22.08, + "learning_rate": 1.3214770724378054e-05, + "loss": 0.9041, + "step": 255120 + }, + { + "epoch": 22.08, + "learning_rate": 1.3213326013464706e-05, + "loss": 0.9906, + "step": 255130 + }, + { + "epoch": 22.08, + "learning_rate": 1.321188130255136e-05, + "loss": 0.9273, + "step": 255140 + }, + { + "epoch": 22.09, + "learning_rate": 1.3210436591638014e-05, + "loss": 0.9504, + "step": 255150 + }, + { + "epoch": 22.09, + "learning_rate": 1.3208991880724666e-05, + "loss": 0.8913, + "step": 255160 + }, + { + "epoch": 22.09, + "learning_rate": 1.320754716981132e-05, + "loss": 0.9053, + "step": 255170 + }, + { + "epoch": 22.09, + "learning_rate": 1.3206102458897976e-05, + "loss": 0.9244, + "step": 255180 + }, + { + "epoch": 22.09, + "learning_rate": 1.3204657747984628e-05, + "loss": 0.9539, + "step": 255190 + }, + { + "epoch": 22.09, + "learning_rate": 1.3203213037071283e-05, + "loss": 0.9653, + "step": 255200 + }, + { + "epoch": 22.09, + "learning_rate": 1.3201768326157938e-05, + "loss": 0.899, + "step": 255210 + }, + { + "epoch": 22.09, + "learning_rate": 1.320032361524459e-05, + "loss": 0.9365, + "step": 255220 + }, + { + "epoch": 22.09, + "learning_rate": 1.3198878904331245e-05, + "loss": 0.9561, + "step": 255230 + }, + { + "epoch": 22.09, + "learning_rate": 1.3197434193417898e-05, + "loss": 0.9744, + "step": 255240 + }, + { + "epoch": 22.09, + "learning_rate": 1.319598948250455e-05, + "loss": 0.9419, + "step": 255250 + }, + { + "epoch": 22.09, + "learning_rate": 1.3194544771591205e-05, + "loss": 0.9556, + "step": 255260 + }, + { + "epoch": 22.1, + "learning_rate": 1.319310006067786e-05, + "loss": 0.9241, + "step": 255270 + }, + { + "epoch": 22.1, + "learning_rate": 1.3191655349764512e-05, + "loss": 0.9524, + "step": 255280 + }, + { + "epoch": 22.1, + "learning_rate": 1.3190210638851167e-05, + "loss": 0.9486, + "step": 255290 + }, + { + "epoch": 22.1, + "learning_rate": 1.3188765927937822e-05, + "loss": 0.9187, + "step": 255300 + }, + { + "epoch": 22.1, + "learning_rate": 1.3187321217024473e-05, + "loss": 0.985, + "step": 255310 + }, + { + "epoch": 22.1, + "learning_rate": 1.3185876506111129e-05, + "loss": 0.9429, + "step": 255320 + }, + { + "epoch": 22.1, + "learning_rate": 1.3184431795197782e-05, + "loss": 0.8585, + "step": 255330 + }, + { + "epoch": 22.1, + "learning_rate": 1.3182987084284434e-05, + "loss": 0.9526, + "step": 255340 + }, + { + "epoch": 22.1, + "learning_rate": 1.3181542373371089e-05, + "loss": 0.9285, + "step": 255350 + }, + { + "epoch": 22.1, + "learning_rate": 1.3180097662457744e-05, + "loss": 0.9313, + "step": 255360 + }, + { + "epoch": 22.1, + "learning_rate": 1.3178652951544396e-05, + "loss": 0.9911, + "step": 255370 + }, + { + "epoch": 22.1, + "learning_rate": 1.317720824063105e-05, + "loss": 0.9415, + "step": 255380 + }, + { + "epoch": 22.11, + "learning_rate": 1.3175763529717706e-05, + "loss": 0.9268, + "step": 255390 + }, + { + "epoch": 22.11, + "learning_rate": 1.3174318818804357e-05, + "loss": 0.9593, + "step": 255400 + }, + { + "epoch": 22.11, + "learning_rate": 1.3172874107891012e-05, + "loss": 0.888, + "step": 255410 + }, + { + "epoch": 22.11, + "learning_rate": 1.3171429396977666e-05, + "loss": 0.9135, + "step": 255420 + }, + { + "epoch": 22.11, + "learning_rate": 1.3169984686064318e-05, + "loss": 0.9799, + "step": 255430 + }, + { + "epoch": 22.11, + "learning_rate": 1.3168539975150973e-05, + "loss": 0.9767, + "step": 255440 + }, + { + "epoch": 22.11, + "learning_rate": 1.3167095264237628e-05, + "loss": 0.8786, + "step": 255450 + }, + { + "epoch": 22.11, + "learning_rate": 1.316565055332428e-05, + "loss": 0.8933, + "step": 255460 + }, + { + "epoch": 22.11, + "learning_rate": 1.3164205842410935e-05, + "loss": 0.9529, + "step": 255470 + }, + { + "epoch": 22.11, + "learning_rate": 1.316276113149759e-05, + "loss": 0.9074, + "step": 255480 + }, + { + "epoch": 22.11, + "learning_rate": 1.3161316420584241e-05, + "loss": 0.9632, + "step": 255490 + }, + { + "epoch": 22.12, + "learning_rate": 1.3159871709670896e-05, + "loss": 0.9047, + "step": 255500 + }, + { + "epoch": 22.12, + "learning_rate": 1.315842699875755e-05, + "loss": 0.9328, + "step": 255510 + }, + { + "epoch": 22.12, + "learning_rate": 1.3156982287844202e-05, + "loss": 0.859, + "step": 255520 + }, + { + "epoch": 22.12, + "learning_rate": 1.3155537576930857e-05, + "loss": 0.9144, + "step": 255530 + }, + { + "epoch": 22.12, + "learning_rate": 1.3154092866017512e-05, + "loss": 0.8872, + "step": 255540 + }, + { + "epoch": 22.12, + "learning_rate": 1.3152648155104163e-05, + "loss": 0.9158, + "step": 255550 + }, + { + "epoch": 22.12, + "learning_rate": 1.3151203444190818e-05, + "loss": 0.9604, + "step": 255560 + }, + { + "epoch": 22.12, + "learning_rate": 1.3149758733277474e-05, + "loss": 0.9037, + "step": 255570 + }, + { + "epoch": 22.12, + "learning_rate": 1.3148314022364125e-05, + "loss": 0.8972, + "step": 255580 + }, + { + "epoch": 22.12, + "learning_rate": 1.314686931145078e-05, + "loss": 0.9829, + "step": 255590 + }, + { + "epoch": 22.12, + "learning_rate": 1.3145424600537434e-05, + "loss": 0.9506, + "step": 255600 + }, + { + "epoch": 22.12, + "learning_rate": 1.3143979889624085e-05, + "loss": 0.8818, + "step": 255610 + }, + { + "epoch": 22.13, + "learning_rate": 1.314253517871074e-05, + "loss": 0.9051, + "step": 255620 + }, + { + "epoch": 22.13, + "learning_rate": 1.3141090467797396e-05, + "loss": 1.0057, + "step": 255630 + }, + { + "epoch": 22.13, + "learning_rate": 1.3139645756884047e-05, + "loss": 0.9106, + "step": 255640 + }, + { + "epoch": 22.13, + "learning_rate": 1.3138201045970702e-05, + "loss": 0.9028, + "step": 255650 + }, + { + "epoch": 22.13, + "learning_rate": 1.3136756335057358e-05, + "loss": 0.9538, + "step": 255660 + }, + { + "epoch": 22.13, + "learning_rate": 1.313531162414401e-05, + "loss": 0.9171, + "step": 255670 + }, + { + "epoch": 22.13, + "learning_rate": 1.3133866913230664e-05, + "loss": 0.9552, + "step": 255680 + }, + { + "epoch": 22.13, + "learning_rate": 1.3132422202317318e-05, + "loss": 0.9876, + "step": 255690 + }, + { + "epoch": 22.13, + "learning_rate": 1.313097749140397e-05, + "loss": 0.934, + "step": 255700 + }, + { + "epoch": 22.13, + "learning_rate": 1.3129532780490625e-05, + "loss": 0.9392, + "step": 255710 + }, + { + "epoch": 22.13, + "learning_rate": 1.312808806957728e-05, + "loss": 0.9329, + "step": 255720 + }, + { + "epoch": 22.14, + "learning_rate": 1.3126643358663931e-05, + "loss": 0.8952, + "step": 255730 + }, + { + "epoch": 22.14, + "learning_rate": 1.3125198647750586e-05, + "loss": 0.9356, + "step": 255740 + }, + { + "epoch": 22.14, + "learning_rate": 1.3123753936837241e-05, + "loss": 0.8729, + "step": 255750 + }, + { + "epoch": 22.14, + "learning_rate": 1.3122309225923893e-05, + "loss": 0.8758, + "step": 255760 + }, + { + "epoch": 22.14, + "learning_rate": 1.3120864515010548e-05, + "loss": 0.9546, + "step": 255770 + }, + { + "epoch": 22.14, + "learning_rate": 1.31194198040972e-05, + "loss": 0.8866, + "step": 255780 + }, + { + "epoch": 22.14, + "learning_rate": 1.3117975093183853e-05, + "loss": 0.9606, + "step": 255790 + }, + { + "epoch": 22.14, + "learning_rate": 1.3116530382270508e-05, + "loss": 0.8812, + "step": 255800 + }, + { + "epoch": 22.14, + "learning_rate": 1.311508567135716e-05, + "loss": 0.9867, + "step": 255810 + }, + { + "epoch": 22.14, + "learning_rate": 1.3113640960443815e-05, + "loss": 0.9259, + "step": 255820 + }, + { + "epoch": 22.14, + "learning_rate": 1.311219624953047e-05, + "loss": 0.9704, + "step": 255830 + }, + { + "epoch": 22.14, + "learning_rate": 1.3110751538617122e-05, + "loss": 0.9461, + "step": 255840 + }, + { + "epoch": 22.15, + "learning_rate": 1.3109306827703777e-05, + "loss": 0.9149, + "step": 255850 + }, + { + "epoch": 22.15, + "learning_rate": 1.3107862116790432e-05, + "loss": 0.9123, + "step": 255860 + }, + { + "epoch": 22.15, + "learning_rate": 1.3106417405877084e-05, + "loss": 0.8828, + "step": 255870 + }, + { + "epoch": 22.15, + "learning_rate": 1.3104972694963737e-05, + "loss": 0.9346, + "step": 255880 + }, + { + "epoch": 22.15, + "learning_rate": 1.3103527984050392e-05, + "loss": 0.9466, + "step": 255890 + }, + { + "epoch": 22.15, + "learning_rate": 1.3102083273137044e-05, + "loss": 0.9094, + "step": 255900 + }, + { + "epoch": 22.15, + "learning_rate": 1.31006385622237e-05, + "loss": 0.9615, + "step": 255910 + }, + { + "epoch": 22.15, + "learning_rate": 1.3099193851310354e-05, + "loss": 0.9613, + "step": 255920 + }, + { + "epoch": 22.15, + "learning_rate": 1.3097749140397006e-05, + "loss": 0.8722, + "step": 255930 + }, + { + "epoch": 22.15, + "learning_rate": 1.3096304429483661e-05, + "loss": 0.9322, + "step": 255940 + }, + { + "epoch": 22.15, + "learning_rate": 1.3094859718570316e-05, + "loss": 0.9606, + "step": 255950 + }, + { + "epoch": 22.16, + "learning_rate": 1.3093415007656968e-05, + "loss": 0.8873, + "step": 255960 + }, + { + "epoch": 22.16, + "learning_rate": 1.3091970296743621e-05, + "loss": 0.9191, + "step": 255970 + }, + { + "epoch": 22.16, + "learning_rate": 1.3090525585830276e-05, + "loss": 0.897, + "step": 255980 + }, + { + "epoch": 22.16, + "learning_rate": 1.3089080874916928e-05, + "loss": 0.8887, + "step": 255990 + }, + { + "epoch": 22.16, + "learning_rate": 1.3087636164003583e-05, + "loss": 0.932, + "step": 256000 + }, + { + "epoch": 22.16, + "learning_rate": 1.3086191453090238e-05, + "loss": 0.9097, + "step": 256010 + }, + { + "epoch": 22.16, + "learning_rate": 1.308474674217689e-05, + "loss": 0.9244, + "step": 256020 + }, + { + "epoch": 22.16, + "learning_rate": 1.3083302031263545e-05, + "loss": 0.87, + "step": 256030 + }, + { + "epoch": 22.16, + "learning_rate": 1.30818573203502e-05, + "loss": 0.8924, + "step": 256040 + }, + { + "epoch": 22.16, + "learning_rate": 1.3080412609436852e-05, + "loss": 0.8502, + "step": 256050 + }, + { + "epoch": 22.16, + "learning_rate": 1.3078967898523505e-05, + "loss": 0.9606, + "step": 256060 + }, + { + "epoch": 22.16, + "learning_rate": 1.307752318761016e-05, + "loss": 0.9236, + "step": 256070 + }, + { + "epoch": 22.17, + "learning_rate": 1.3076078476696812e-05, + "loss": 0.888, + "step": 256080 + }, + { + "epoch": 22.17, + "learning_rate": 1.3074633765783467e-05, + "loss": 0.9019, + "step": 256090 + }, + { + "epoch": 22.17, + "learning_rate": 1.3073189054870122e-05, + "loss": 0.9767, + "step": 256100 + }, + { + "epoch": 22.17, + "learning_rate": 1.3071744343956774e-05, + "loss": 0.9537, + "step": 256110 + }, + { + "epoch": 22.17, + "learning_rate": 1.3070299633043429e-05, + "loss": 0.8835, + "step": 256120 + }, + { + "epoch": 22.17, + "learning_rate": 1.3068854922130084e-05, + "loss": 0.9091, + "step": 256130 + }, + { + "epoch": 22.17, + "learning_rate": 1.3067410211216736e-05, + "loss": 0.9492, + "step": 256140 + }, + { + "epoch": 22.17, + "learning_rate": 1.3065965500303389e-05, + "loss": 0.8724, + "step": 256150 + }, + { + "epoch": 22.17, + "learning_rate": 1.3064520789390044e-05, + "loss": 0.8797, + "step": 256160 + }, + { + "epoch": 22.17, + "learning_rate": 1.3063076078476696e-05, + "loss": 0.9333, + "step": 256170 + }, + { + "epoch": 22.17, + "learning_rate": 1.3061631367563351e-05, + "loss": 0.9261, + "step": 256180 + }, + { + "epoch": 22.18, + "learning_rate": 1.3060186656650006e-05, + "loss": 0.9213, + "step": 256190 + }, + { + "epoch": 22.18, + "learning_rate": 1.3058741945736658e-05, + "loss": 0.8858, + "step": 256200 + }, + { + "epoch": 22.18, + "learning_rate": 1.3057297234823313e-05, + "loss": 0.906, + "step": 256210 + }, + { + "epoch": 22.18, + "learning_rate": 1.3055852523909968e-05, + "loss": 0.9583, + "step": 256220 + }, + { + "epoch": 22.18, + "learning_rate": 1.305440781299662e-05, + "loss": 0.9457, + "step": 256230 + }, + { + "epoch": 22.18, + "learning_rate": 1.3052963102083273e-05, + "loss": 0.9108, + "step": 256240 + }, + { + "epoch": 22.18, + "learning_rate": 1.3051518391169928e-05, + "loss": 0.9036, + "step": 256250 + }, + { + "epoch": 22.18, + "learning_rate": 1.305007368025658e-05, + "loss": 0.9473, + "step": 256260 + }, + { + "epoch": 22.18, + "learning_rate": 1.3048628969343235e-05, + "loss": 0.9585, + "step": 256270 + }, + { + "epoch": 22.18, + "learning_rate": 1.304718425842989e-05, + "loss": 0.9428, + "step": 256280 + }, + { + "epoch": 22.18, + "learning_rate": 1.3045739547516542e-05, + "loss": 0.9719, + "step": 256290 + }, + { + "epoch": 22.18, + "learning_rate": 1.3044294836603197e-05, + "loss": 0.8633, + "step": 256300 + }, + { + "epoch": 22.19, + "learning_rate": 1.3042850125689852e-05, + "loss": 0.9015, + "step": 256310 + }, + { + "epoch": 22.19, + "learning_rate": 1.3041405414776504e-05, + "loss": 0.9698, + "step": 256320 + }, + { + "epoch": 22.19, + "learning_rate": 1.3039960703863157e-05, + "loss": 0.9447, + "step": 256330 + }, + { + "epoch": 22.19, + "learning_rate": 1.3038515992949812e-05, + "loss": 0.9034, + "step": 256340 + }, + { + "epoch": 22.19, + "learning_rate": 1.3037071282036464e-05, + "loss": 0.8938, + "step": 256350 + }, + { + "epoch": 22.19, + "learning_rate": 1.3035626571123119e-05, + "loss": 0.9344, + "step": 256360 + }, + { + "epoch": 22.19, + "learning_rate": 1.3034181860209774e-05, + "loss": 0.9277, + "step": 256370 + }, + { + "epoch": 22.19, + "learning_rate": 1.3032737149296426e-05, + "loss": 0.9155, + "step": 256380 + }, + { + "epoch": 22.19, + "learning_rate": 1.303129243838308e-05, + "loss": 0.9173, + "step": 256390 + }, + { + "epoch": 22.19, + "learning_rate": 1.3029847727469736e-05, + "loss": 0.9133, + "step": 256400 + }, + { + "epoch": 22.19, + "learning_rate": 1.3028403016556388e-05, + "loss": 0.938, + "step": 256410 + }, + { + "epoch": 22.2, + "learning_rate": 1.3026958305643041e-05, + "loss": 0.9005, + "step": 256420 + }, + { + "epoch": 22.2, + "learning_rate": 1.3025513594729696e-05, + "loss": 0.8714, + "step": 256430 + }, + { + "epoch": 22.2, + "learning_rate": 1.3024068883816348e-05, + "loss": 0.8771, + "step": 256440 + }, + { + "epoch": 22.2, + "learning_rate": 1.3022624172903003e-05, + "loss": 0.9261, + "step": 256450 + }, + { + "epoch": 22.2, + "learning_rate": 1.3021179461989658e-05, + "loss": 0.9549, + "step": 256460 + }, + { + "epoch": 22.2, + "learning_rate": 1.301973475107631e-05, + "loss": 0.8968, + "step": 256470 + }, + { + "epoch": 22.2, + "learning_rate": 1.3018290040162965e-05, + "loss": 0.9254, + "step": 256480 + }, + { + "epoch": 22.2, + "learning_rate": 1.301684532924962e-05, + "loss": 0.9659, + "step": 256490 + }, + { + "epoch": 22.2, + "learning_rate": 1.3015400618336272e-05, + "loss": 0.936, + "step": 256500 + }, + { + "epoch": 22.2, + "learning_rate": 1.3013955907422925e-05, + "loss": 0.941, + "step": 256510 + }, + { + "epoch": 22.2, + "learning_rate": 1.301251119650958e-05, + "loss": 0.9964, + "step": 256520 + }, + { + "epoch": 22.2, + "learning_rate": 1.3011066485596232e-05, + "loss": 0.9579, + "step": 256530 + }, + { + "epoch": 22.21, + "learning_rate": 1.3009621774682887e-05, + "loss": 0.8923, + "step": 256540 + }, + { + "epoch": 22.21, + "learning_rate": 1.3008177063769542e-05, + "loss": 0.9305, + "step": 256550 + }, + { + "epoch": 22.21, + "learning_rate": 1.3006732352856194e-05, + "loss": 0.9804, + "step": 256560 + }, + { + "epoch": 22.21, + "learning_rate": 1.3005287641942849e-05, + "loss": 0.8849, + "step": 256570 + }, + { + "epoch": 22.21, + "learning_rate": 1.3003842931029504e-05, + "loss": 0.9101, + "step": 256580 + }, + { + "epoch": 22.21, + "learning_rate": 1.3002398220116155e-05, + "loss": 0.8787, + "step": 256590 + }, + { + "epoch": 22.21, + "learning_rate": 1.3000953509202809e-05, + "loss": 0.9036, + "step": 256600 + }, + { + "epoch": 22.21, + "learning_rate": 1.2999508798289464e-05, + "loss": 0.9523, + "step": 256610 + }, + { + "epoch": 22.21, + "learning_rate": 1.2998064087376116e-05, + "loss": 0.9369, + "step": 256620 + }, + { + "epoch": 22.21, + "learning_rate": 1.299661937646277e-05, + "loss": 0.937, + "step": 256630 + }, + { + "epoch": 22.21, + "learning_rate": 1.2995174665549426e-05, + "loss": 0.8836, + "step": 256640 + }, + { + "epoch": 22.21, + "learning_rate": 1.2993729954636078e-05, + "loss": 0.9303, + "step": 256650 + }, + { + "epoch": 22.22, + "learning_rate": 1.2992285243722733e-05, + "loss": 0.9375, + "step": 256660 + }, + { + "epoch": 22.22, + "learning_rate": 1.2990840532809388e-05, + "loss": 0.9621, + "step": 256670 + }, + { + "epoch": 22.22, + "learning_rate": 1.298939582189604e-05, + "loss": 0.9089, + "step": 256680 + }, + { + "epoch": 22.22, + "learning_rate": 1.2987951110982693e-05, + "loss": 0.9198, + "step": 256690 + }, + { + "epoch": 22.22, + "learning_rate": 1.2986506400069346e-05, + "loss": 0.9395, + "step": 256700 + }, + { + "epoch": 22.22, + "learning_rate": 1.2985061689156e-05, + "loss": 0.8241, + "step": 256710 + }, + { + "epoch": 22.22, + "learning_rate": 1.2983616978242655e-05, + "loss": 0.9542, + "step": 256720 + }, + { + "epoch": 22.22, + "learning_rate": 1.2982172267329306e-05, + "loss": 0.9373, + "step": 256730 + }, + { + "epoch": 22.22, + "learning_rate": 1.2980727556415961e-05, + "loss": 0.9347, + "step": 256740 + }, + { + "epoch": 22.22, + "learning_rate": 1.2979282845502617e-05, + "loss": 0.9381, + "step": 256750 + }, + { + "epoch": 22.22, + "learning_rate": 1.2977838134589268e-05, + "loss": 0.9917, + "step": 256760 + }, + { + "epoch": 22.23, + "learning_rate": 1.2976393423675923e-05, + "loss": 0.9371, + "step": 256770 + }, + { + "epoch": 22.23, + "learning_rate": 1.2974948712762577e-05, + "loss": 0.9054, + "step": 256780 + }, + { + "epoch": 22.23, + "learning_rate": 1.297350400184923e-05, + "loss": 0.9485, + "step": 256790 + }, + { + "epoch": 22.23, + "learning_rate": 1.2972059290935884e-05, + "loss": 0.9376, + "step": 256800 + }, + { + "epoch": 22.23, + "learning_rate": 1.2970614580022539e-05, + "loss": 0.8594, + "step": 256810 + }, + { + "epoch": 22.23, + "learning_rate": 1.296916986910919e-05, + "loss": 0.9331, + "step": 256820 + }, + { + "epoch": 22.23, + "learning_rate": 1.2967725158195845e-05, + "loss": 0.9663, + "step": 256830 + }, + { + "epoch": 22.23, + "learning_rate": 1.29662804472825e-05, + "loss": 0.9614, + "step": 256840 + }, + { + "epoch": 22.23, + "learning_rate": 1.2964835736369152e-05, + "loss": 0.9253, + "step": 256850 + }, + { + "epoch": 22.23, + "learning_rate": 1.2963391025455807e-05, + "loss": 0.8992, + "step": 256860 + }, + { + "epoch": 22.23, + "learning_rate": 1.296194631454246e-05, + "loss": 0.8929, + "step": 256870 + }, + { + "epoch": 22.23, + "learning_rate": 1.2960501603629114e-05, + "loss": 0.9304, + "step": 256880 + }, + { + "epoch": 22.24, + "learning_rate": 1.2959056892715767e-05, + "loss": 0.8997, + "step": 256890 + }, + { + "epoch": 22.24, + "learning_rate": 1.2957612181802423e-05, + "loss": 0.899, + "step": 256900 + }, + { + "epoch": 22.24, + "learning_rate": 1.2956167470889074e-05, + "loss": 0.9845, + "step": 256910 + }, + { + "epoch": 22.24, + "learning_rate": 1.295472275997573e-05, + "loss": 0.9143, + "step": 256920 + }, + { + "epoch": 22.24, + "learning_rate": 1.2953278049062384e-05, + "loss": 0.9384, + "step": 256930 + }, + { + "epoch": 22.24, + "learning_rate": 1.2951833338149036e-05, + "loss": 0.9846, + "step": 256940 + }, + { + "epoch": 22.24, + "learning_rate": 1.2950388627235691e-05, + "loss": 0.875, + "step": 256950 + }, + { + "epoch": 22.24, + "learning_rate": 1.2948943916322345e-05, + "loss": 0.9212, + "step": 256960 + }, + { + "epoch": 22.24, + "learning_rate": 1.2947499205408998e-05, + "loss": 0.9259, + "step": 256970 + }, + { + "epoch": 22.24, + "learning_rate": 1.2946054494495651e-05, + "loss": 0.9619, + "step": 256980 + }, + { + "epoch": 22.24, + "learning_rate": 1.2944609783582306e-05, + "loss": 0.9665, + "step": 256990 + }, + { + "epoch": 22.25, + "learning_rate": 1.2943165072668958e-05, + "loss": 0.9078, + "step": 257000 + }, + { + "epoch": 22.25, + "learning_rate": 1.2941720361755613e-05, + "loss": 0.8943, + "step": 257010 + }, + { + "epoch": 22.25, + "learning_rate": 1.2940275650842268e-05, + "loss": 0.9279, + "step": 257020 + }, + { + "epoch": 22.25, + "learning_rate": 1.293883093992892e-05, + "loss": 0.8731, + "step": 257030 + }, + { + "epoch": 22.25, + "learning_rate": 1.2937386229015575e-05, + "loss": 0.9185, + "step": 257040 + }, + { + "epoch": 22.25, + "learning_rate": 1.2935941518102229e-05, + "loss": 0.9541, + "step": 257050 + }, + { + "epoch": 22.25, + "learning_rate": 1.2934496807188882e-05, + "loss": 0.9261, + "step": 257060 + }, + { + "epoch": 22.25, + "learning_rate": 1.2933052096275535e-05, + "loss": 0.9023, + "step": 257070 + }, + { + "epoch": 22.25, + "learning_rate": 1.293160738536219e-05, + "loss": 0.8853, + "step": 257080 + }, + { + "epoch": 22.25, + "learning_rate": 1.2930162674448842e-05, + "loss": 0.9237, + "step": 257090 + }, + { + "epoch": 22.25, + "learning_rate": 1.2928717963535497e-05, + "loss": 0.9258, + "step": 257100 + }, + { + "epoch": 22.25, + "learning_rate": 1.2927273252622152e-05, + "loss": 0.926, + "step": 257110 + }, + { + "epoch": 22.26, + "learning_rate": 1.2925828541708804e-05, + "loss": 0.9166, + "step": 257120 + }, + { + "epoch": 22.26, + "learning_rate": 1.2924383830795459e-05, + "loss": 0.9823, + "step": 257130 + }, + { + "epoch": 22.26, + "learning_rate": 1.2922939119882112e-05, + "loss": 0.942, + "step": 257140 + }, + { + "epoch": 22.26, + "learning_rate": 1.2921494408968766e-05, + "loss": 0.9265, + "step": 257150 + }, + { + "epoch": 22.26, + "learning_rate": 1.292004969805542e-05, + "loss": 0.8771, + "step": 257160 + }, + { + "epoch": 22.26, + "learning_rate": 1.2918604987142074e-05, + "loss": 0.9231, + "step": 257170 + }, + { + "epoch": 22.26, + "learning_rate": 1.2917160276228726e-05, + "loss": 0.9437, + "step": 257180 + }, + { + "epoch": 22.26, + "learning_rate": 1.2915715565315381e-05, + "loss": 0.9469, + "step": 257190 + }, + { + "epoch": 22.26, + "learning_rate": 1.2914270854402036e-05, + "loss": 0.9514, + "step": 257200 + }, + { + "epoch": 22.26, + "learning_rate": 1.2912826143488688e-05, + "loss": 0.9451, + "step": 257210 + }, + { + "epoch": 22.26, + "learning_rate": 1.2911381432575343e-05, + "loss": 0.9221, + "step": 257220 + }, + { + "epoch": 22.27, + "learning_rate": 1.2909936721661996e-05, + "loss": 0.9512, + "step": 257230 + }, + { + "epoch": 22.27, + "learning_rate": 1.290849201074865e-05, + "loss": 0.9593, + "step": 257240 + }, + { + "epoch": 22.27, + "learning_rate": 1.2907047299835303e-05, + "loss": 0.9016, + "step": 257250 + }, + { + "epoch": 22.27, + "learning_rate": 1.2905602588921958e-05, + "loss": 0.9525, + "step": 257260 + }, + { + "epoch": 22.27, + "learning_rate": 1.290415787800861e-05, + "loss": 0.9388, + "step": 257270 + }, + { + "epoch": 22.27, + "learning_rate": 1.2902713167095265e-05, + "loss": 0.8878, + "step": 257280 + }, + { + "epoch": 22.27, + "learning_rate": 1.290126845618192e-05, + "loss": 0.9009, + "step": 257290 + }, + { + "epoch": 22.27, + "learning_rate": 1.2899823745268572e-05, + "loss": 0.9075, + "step": 257300 + }, + { + "epoch": 22.27, + "learning_rate": 1.2898379034355227e-05, + "loss": 0.9371, + "step": 257310 + }, + { + "epoch": 22.27, + "learning_rate": 1.289693432344188e-05, + "loss": 0.925, + "step": 257320 + }, + { + "epoch": 22.27, + "learning_rate": 1.2895489612528534e-05, + "loss": 0.9567, + "step": 257330 + }, + { + "epoch": 22.27, + "learning_rate": 1.2894044901615187e-05, + "loss": 0.9335, + "step": 257340 + }, + { + "epoch": 22.28, + "learning_rate": 1.2892600190701842e-05, + "loss": 0.9122, + "step": 257350 + }, + { + "epoch": 22.28, + "learning_rate": 1.2891155479788494e-05, + "loss": 0.9148, + "step": 257360 + }, + { + "epoch": 22.28, + "learning_rate": 1.2889710768875149e-05, + "loss": 0.92, + "step": 257370 + }, + { + "epoch": 22.28, + "learning_rate": 1.2888266057961804e-05, + "loss": 0.9807, + "step": 257380 + }, + { + "epoch": 22.28, + "learning_rate": 1.2886821347048456e-05, + "loss": 0.8965, + "step": 257390 + }, + { + "epoch": 22.28, + "learning_rate": 1.2885376636135111e-05, + "loss": 0.9649, + "step": 257400 + }, + { + "epoch": 22.28, + "learning_rate": 1.2883931925221764e-05, + "loss": 0.8954, + "step": 257410 + }, + { + "epoch": 22.28, + "learning_rate": 1.2882487214308418e-05, + "loss": 0.9021, + "step": 257420 + }, + { + "epoch": 22.28, + "learning_rate": 1.2881042503395071e-05, + "loss": 0.9295, + "step": 257430 + }, + { + "epoch": 22.28, + "learning_rate": 1.2879597792481726e-05, + "loss": 0.8921, + "step": 257440 + }, + { + "epoch": 22.28, + "learning_rate": 1.2878153081568378e-05, + "loss": 0.8945, + "step": 257450 + }, + { + "epoch": 22.29, + "learning_rate": 1.2876708370655033e-05, + "loss": 0.9214, + "step": 257460 + }, + { + "epoch": 22.29, + "learning_rate": 1.2875263659741688e-05, + "loss": 0.9352, + "step": 257470 + }, + { + "epoch": 22.29, + "learning_rate": 1.287381894882834e-05, + "loss": 0.9332, + "step": 257480 + }, + { + "epoch": 22.29, + "learning_rate": 1.2872374237914995e-05, + "loss": 0.8661, + "step": 257490 + }, + { + "epoch": 22.29, + "learning_rate": 1.2870929527001648e-05, + "loss": 0.9846, + "step": 257500 + }, + { + "epoch": 22.29, + "learning_rate": 1.28694848160883e-05, + "loss": 0.8699, + "step": 257510 + }, + { + "epoch": 22.29, + "learning_rate": 1.2868040105174955e-05, + "loss": 0.9059, + "step": 257520 + }, + { + "epoch": 22.29, + "learning_rate": 1.286659539426161e-05, + "loss": 0.9526, + "step": 257530 + }, + { + "epoch": 22.29, + "learning_rate": 1.2865150683348262e-05, + "loss": 0.9544, + "step": 257540 + }, + { + "epoch": 22.29, + "learning_rate": 1.2863705972434917e-05, + "loss": 0.947, + "step": 257550 + }, + { + "epoch": 22.29, + "learning_rate": 1.2862261261521572e-05, + "loss": 0.9605, + "step": 257560 + }, + { + "epoch": 22.29, + "learning_rate": 1.2860816550608224e-05, + "loss": 0.8638, + "step": 257570 + }, + { + "epoch": 22.3, + "learning_rate": 1.2859371839694879e-05, + "loss": 0.9594, + "step": 257580 + }, + { + "epoch": 22.3, + "learning_rate": 1.285792712878153e-05, + "loss": 0.8738, + "step": 257590 + }, + { + "epoch": 22.3, + "learning_rate": 1.2856482417868184e-05, + "loss": 0.9378, + "step": 257600 + }, + { + "epoch": 22.3, + "learning_rate": 1.2855037706954839e-05, + "loss": 0.9379, + "step": 257610 + }, + { + "epoch": 22.3, + "learning_rate": 1.285359299604149e-05, + "loss": 0.9491, + "step": 257620 + }, + { + "epoch": 22.3, + "learning_rate": 1.2852148285128146e-05, + "loss": 0.9313, + "step": 257630 + }, + { + "epoch": 22.3, + "learning_rate": 1.2850703574214801e-05, + "loss": 0.905, + "step": 257640 + }, + { + "epoch": 22.3, + "learning_rate": 1.2849258863301453e-05, + "loss": 0.8886, + "step": 257650 + }, + { + "epoch": 22.3, + "learning_rate": 1.2847814152388108e-05, + "loss": 0.9253, + "step": 257660 + }, + { + "epoch": 22.3, + "learning_rate": 1.2846369441474763e-05, + "loss": 0.9473, + "step": 257670 + }, + { + "epoch": 22.3, + "learning_rate": 1.2844924730561414e-05, + "loss": 0.9179, + "step": 257680 + }, + { + "epoch": 22.3, + "learning_rate": 1.2843480019648068e-05, + "loss": 0.9558, + "step": 257690 + }, + { + "epoch": 22.31, + "learning_rate": 1.2842035308734723e-05, + "loss": 0.9484, + "step": 257700 + }, + { + "epoch": 22.31, + "learning_rate": 1.2840590597821375e-05, + "loss": 0.929, + "step": 257710 + }, + { + "epoch": 22.31, + "learning_rate": 1.283914588690803e-05, + "loss": 0.9198, + "step": 257720 + }, + { + "epoch": 22.31, + "learning_rate": 1.2837701175994685e-05, + "loss": 0.9784, + "step": 257730 + }, + { + "epoch": 22.31, + "learning_rate": 1.2836256465081337e-05, + "loss": 0.8902, + "step": 257740 + }, + { + "epoch": 22.31, + "learning_rate": 1.2834811754167992e-05, + "loss": 0.9506, + "step": 257750 + }, + { + "epoch": 22.31, + "learning_rate": 1.2833367043254647e-05, + "loss": 0.9263, + "step": 257760 + }, + { + "epoch": 22.31, + "learning_rate": 1.2831922332341298e-05, + "loss": 0.9452, + "step": 257770 + }, + { + "epoch": 22.31, + "learning_rate": 1.2830477621427952e-05, + "loss": 0.9553, + "step": 257780 + }, + { + "epoch": 22.31, + "learning_rate": 1.2829032910514607e-05, + "loss": 0.9645, + "step": 257790 + }, + { + "epoch": 22.31, + "learning_rate": 1.2827588199601259e-05, + "loss": 0.8408, + "step": 257800 + }, + { + "epoch": 22.32, + "learning_rate": 1.2826143488687914e-05, + "loss": 0.8948, + "step": 257810 + }, + { + "epoch": 22.32, + "learning_rate": 1.2824698777774569e-05, + "loss": 0.8835, + "step": 257820 + }, + { + "epoch": 22.32, + "learning_rate": 1.282325406686122e-05, + "loss": 0.942, + "step": 257830 + }, + { + "epoch": 22.32, + "learning_rate": 1.2821809355947876e-05, + "loss": 0.9101, + "step": 257840 + }, + { + "epoch": 22.32, + "learning_rate": 1.282036464503453e-05, + "loss": 0.9332, + "step": 257850 + }, + { + "epoch": 22.32, + "learning_rate": 1.2818919934121182e-05, + "loss": 0.9383, + "step": 257860 + }, + { + "epoch": 22.32, + "learning_rate": 1.2817475223207836e-05, + "loss": 0.9573, + "step": 257870 + }, + { + "epoch": 22.32, + "learning_rate": 1.281603051229449e-05, + "loss": 0.9752, + "step": 257880 + }, + { + "epoch": 22.32, + "learning_rate": 1.2814585801381143e-05, + "loss": 0.9503, + "step": 257890 + }, + { + "epoch": 22.32, + "learning_rate": 1.2813141090467798e-05, + "loss": 0.9462, + "step": 257900 + }, + { + "epoch": 22.32, + "learning_rate": 1.2811696379554453e-05, + "loss": 0.9928, + "step": 257910 + }, + { + "epoch": 22.32, + "learning_rate": 1.2810251668641104e-05, + "loss": 0.901, + "step": 257920 + }, + { + "epoch": 22.33, + "learning_rate": 1.280880695772776e-05, + "loss": 0.9305, + "step": 257930 + }, + { + "epoch": 22.33, + "learning_rate": 1.2807362246814415e-05, + "loss": 0.9532, + "step": 257940 + }, + { + "epoch": 22.33, + "learning_rate": 1.2805917535901066e-05, + "loss": 0.9131, + "step": 257950 + }, + { + "epoch": 22.33, + "learning_rate": 1.280447282498772e-05, + "loss": 0.9085, + "step": 257960 + }, + { + "epoch": 22.33, + "learning_rate": 1.2803028114074375e-05, + "loss": 0.8786, + "step": 257970 + }, + { + "epoch": 22.33, + "learning_rate": 1.2801583403161026e-05, + "loss": 0.9135, + "step": 257980 + }, + { + "epoch": 22.33, + "learning_rate": 1.2800138692247682e-05, + "loss": 0.9333, + "step": 257990 + }, + { + "epoch": 22.33, + "learning_rate": 1.2798693981334337e-05, + "loss": 0.9276, + "step": 258000 + }, + { + "epoch": 22.33, + "learning_rate": 1.2797249270420988e-05, + "loss": 0.9363, + "step": 258010 + }, + { + "epoch": 22.33, + "learning_rate": 1.2795804559507643e-05, + "loss": 0.9141, + "step": 258020 + }, + { + "epoch": 22.33, + "learning_rate": 1.2794359848594299e-05, + "loss": 0.9813, + "step": 258030 + }, + { + "epoch": 22.34, + "learning_rate": 1.279291513768095e-05, + "loss": 0.8627, + "step": 258040 + }, + { + "epoch": 22.34, + "learning_rate": 1.2791470426767604e-05, + "loss": 0.9162, + "step": 258050 + }, + { + "epoch": 22.34, + "learning_rate": 1.2790025715854259e-05, + "loss": 0.9244, + "step": 258060 + }, + { + "epoch": 22.34, + "learning_rate": 1.278858100494091e-05, + "loss": 0.9208, + "step": 258070 + }, + { + "epoch": 22.34, + "learning_rate": 1.2787136294027566e-05, + "loss": 0.934, + "step": 258080 + }, + { + "epoch": 22.34, + "learning_rate": 1.278569158311422e-05, + "loss": 0.9213, + "step": 258090 + }, + { + "epoch": 22.34, + "learning_rate": 1.2784246872200872e-05, + "loss": 0.9192, + "step": 258100 + }, + { + "epoch": 22.34, + "learning_rate": 1.2782802161287527e-05, + "loss": 0.9402, + "step": 258110 + }, + { + "epoch": 22.34, + "learning_rate": 1.2781357450374182e-05, + "loss": 0.9312, + "step": 258120 + }, + { + "epoch": 22.34, + "learning_rate": 1.2779912739460834e-05, + "loss": 0.951, + "step": 258130 + }, + { + "epoch": 22.34, + "learning_rate": 1.2778468028547488e-05, + "loss": 0.9355, + "step": 258140 + }, + { + "epoch": 22.34, + "learning_rate": 1.2777023317634143e-05, + "loss": 0.9053, + "step": 258150 + }, + { + "epoch": 22.35, + "learning_rate": 1.2775578606720794e-05, + "loss": 0.9202, + "step": 258160 + }, + { + "epoch": 22.35, + "learning_rate": 1.277413389580745e-05, + "loss": 0.9098, + "step": 258170 + }, + { + "epoch": 22.35, + "learning_rate": 1.2772689184894105e-05, + "loss": 0.9196, + "step": 258180 + }, + { + "epoch": 22.35, + "learning_rate": 1.2771244473980756e-05, + "loss": 0.9188, + "step": 258190 + }, + { + "epoch": 22.35, + "learning_rate": 1.2769799763067411e-05, + "loss": 0.9726, + "step": 258200 + }, + { + "epoch": 22.35, + "learning_rate": 1.2768355052154066e-05, + "loss": 0.9566, + "step": 258210 + }, + { + "epoch": 22.35, + "learning_rate": 1.2766910341240718e-05, + "loss": 0.9073, + "step": 258220 + }, + { + "epoch": 22.35, + "learning_rate": 1.2765465630327372e-05, + "loss": 0.9208, + "step": 258230 + }, + { + "epoch": 22.35, + "learning_rate": 1.2764020919414027e-05, + "loss": 0.9189, + "step": 258240 + }, + { + "epoch": 22.35, + "learning_rate": 1.2762576208500678e-05, + "loss": 0.9594, + "step": 258250 + }, + { + "epoch": 22.35, + "learning_rate": 1.2761131497587333e-05, + "loss": 0.9262, + "step": 258260 + }, + { + "epoch": 22.36, + "learning_rate": 1.2759686786673988e-05, + "loss": 0.8767, + "step": 258270 + }, + { + "epoch": 22.36, + "learning_rate": 1.275824207576064e-05, + "loss": 0.9471, + "step": 258280 + }, + { + "epoch": 22.36, + "learning_rate": 1.2756797364847295e-05, + "loss": 0.9275, + "step": 258290 + }, + { + "epoch": 22.36, + "learning_rate": 1.275535265393395e-05, + "loss": 0.8842, + "step": 258300 + }, + { + "epoch": 22.36, + "learning_rate": 1.2753907943020602e-05, + "loss": 0.9325, + "step": 258310 + }, + { + "epoch": 22.36, + "learning_rate": 1.2752463232107255e-05, + "loss": 0.9165, + "step": 258320 + }, + { + "epoch": 22.36, + "learning_rate": 1.275101852119391e-05, + "loss": 0.9402, + "step": 258330 + }, + { + "epoch": 22.36, + "learning_rate": 1.2749573810280562e-05, + "loss": 0.8987, + "step": 258340 + }, + { + "epoch": 22.36, + "learning_rate": 1.2748129099367217e-05, + "loss": 0.8786, + "step": 258350 + }, + { + "epoch": 22.36, + "learning_rate": 1.2746684388453872e-05, + "loss": 0.9055, + "step": 258360 + }, + { + "epoch": 22.36, + "learning_rate": 1.2745239677540524e-05, + "loss": 0.9764, + "step": 258370 + }, + { + "epoch": 22.36, + "learning_rate": 1.274379496662718e-05, + "loss": 0.9866, + "step": 258380 + }, + { + "epoch": 22.37, + "learning_rate": 1.2742350255713834e-05, + "loss": 0.9009, + "step": 258390 + }, + { + "epoch": 22.37, + "learning_rate": 1.2740905544800486e-05, + "loss": 0.9482, + "step": 258400 + }, + { + "epoch": 22.37, + "learning_rate": 1.273946083388714e-05, + "loss": 0.9283, + "step": 258410 + }, + { + "epoch": 22.37, + "learning_rate": 1.2738016122973794e-05, + "loss": 0.8927, + "step": 258420 + }, + { + "epoch": 22.37, + "learning_rate": 1.2736571412060446e-05, + "loss": 0.8926, + "step": 258430 + }, + { + "epoch": 22.37, + "learning_rate": 1.2735126701147101e-05, + "loss": 0.907, + "step": 258440 + }, + { + "epoch": 22.37, + "learning_rate": 1.2733681990233756e-05, + "loss": 0.9534, + "step": 258450 + }, + { + "epoch": 22.37, + "learning_rate": 1.2732237279320408e-05, + "loss": 0.9267, + "step": 258460 + }, + { + "epoch": 22.37, + "learning_rate": 1.2730792568407063e-05, + "loss": 0.9123, + "step": 258470 + }, + { + "epoch": 22.37, + "learning_rate": 1.2729347857493718e-05, + "loss": 0.9641, + "step": 258480 + }, + { + "epoch": 22.37, + "learning_rate": 1.272790314658037e-05, + "loss": 0.9138, + "step": 258490 + }, + { + "epoch": 22.38, + "learning_rate": 1.2726458435667023e-05, + "loss": 0.9127, + "step": 258500 + }, + { + "epoch": 22.38, + "learning_rate": 1.2725013724753677e-05, + "loss": 0.9707, + "step": 258510 + }, + { + "epoch": 22.38, + "learning_rate": 1.272356901384033e-05, + "loss": 0.9236, + "step": 258520 + }, + { + "epoch": 22.38, + "learning_rate": 1.2722124302926985e-05, + "loss": 0.9401, + "step": 258530 + }, + { + "epoch": 22.38, + "learning_rate": 1.2720679592013637e-05, + "loss": 0.9126, + "step": 258540 + }, + { + "epoch": 22.38, + "learning_rate": 1.2719234881100292e-05, + "loss": 0.8713, + "step": 258550 + }, + { + "epoch": 22.38, + "learning_rate": 1.2717790170186947e-05, + "loss": 0.8854, + "step": 258560 + }, + { + "epoch": 22.38, + "learning_rate": 1.2716345459273599e-05, + "loss": 0.9903, + "step": 258570 + }, + { + "epoch": 22.38, + "learning_rate": 1.2714900748360254e-05, + "loss": 0.8942, + "step": 258580 + }, + { + "epoch": 22.38, + "learning_rate": 1.2713456037446907e-05, + "loss": 0.923, + "step": 258590 + }, + { + "epoch": 22.38, + "learning_rate": 1.271201132653356e-05, + "loss": 0.8918, + "step": 258600 + }, + { + "epoch": 22.38, + "learning_rate": 1.2710566615620214e-05, + "loss": 0.8722, + "step": 258610 + }, + { + "epoch": 22.39, + "learning_rate": 1.270912190470687e-05, + "loss": 0.9252, + "step": 258620 + }, + { + "epoch": 22.39, + "learning_rate": 1.2707677193793521e-05, + "loss": 0.9269, + "step": 258630 + }, + { + "epoch": 22.39, + "learning_rate": 1.2706232482880176e-05, + "loss": 0.9635, + "step": 258640 + }, + { + "epoch": 22.39, + "learning_rate": 1.2704787771966831e-05, + "loss": 0.9109, + "step": 258650 + }, + { + "epoch": 22.39, + "learning_rate": 1.2703343061053483e-05, + "loss": 0.9329, + "step": 258660 + }, + { + "epoch": 22.39, + "learning_rate": 1.2701898350140138e-05, + "loss": 0.8853, + "step": 258670 + }, + { + "epoch": 22.39, + "learning_rate": 1.2700453639226791e-05, + "loss": 0.9902, + "step": 258680 + }, + { + "epoch": 22.39, + "learning_rate": 1.2699008928313445e-05, + "loss": 0.9396, + "step": 258690 + }, + { + "epoch": 22.39, + "learning_rate": 1.2697564217400098e-05, + "loss": 0.9432, + "step": 258700 + }, + { + "epoch": 22.39, + "learning_rate": 1.2696119506486753e-05, + "loss": 0.9474, + "step": 258710 + }, + { + "epoch": 22.39, + "learning_rate": 1.2694674795573405e-05, + "loss": 0.9129, + "step": 258720 + }, + { + "epoch": 22.39, + "learning_rate": 1.269323008466006e-05, + "loss": 0.9429, + "step": 258730 + }, + { + "epoch": 22.4, + "learning_rate": 1.2691785373746715e-05, + "loss": 0.9156, + "step": 258740 + }, + { + "epoch": 22.4, + "learning_rate": 1.2690340662833367e-05, + "loss": 0.8646, + "step": 258750 + }, + { + "epoch": 22.4, + "learning_rate": 1.2688895951920022e-05, + "loss": 0.9887, + "step": 258760 + }, + { + "epoch": 22.4, + "learning_rate": 1.2687451241006675e-05, + "loss": 0.8804, + "step": 258770 + }, + { + "epoch": 22.4, + "learning_rate": 1.2686006530093329e-05, + "loss": 0.9415, + "step": 258780 + }, + { + "epoch": 22.4, + "learning_rate": 1.2684561819179982e-05, + "loss": 0.9357, + "step": 258790 + }, + { + "epoch": 22.4, + "learning_rate": 1.2683117108266637e-05, + "loss": 0.8944, + "step": 258800 + }, + { + "epoch": 22.4, + "learning_rate": 1.2681672397353289e-05, + "loss": 0.9339, + "step": 258810 + }, + { + "epoch": 22.4, + "learning_rate": 1.2680227686439944e-05, + "loss": 0.898, + "step": 258820 + }, + { + "epoch": 22.4, + "learning_rate": 1.2678782975526599e-05, + "loss": 0.9206, + "step": 258830 + }, + { + "epoch": 22.4, + "learning_rate": 1.267733826461325e-05, + "loss": 0.9084, + "step": 258840 + }, + { + "epoch": 22.41, + "learning_rate": 1.2675893553699906e-05, + "loss": 0.9622, + "step": 258850 + }, + { + "epoch": 22.41, + "learning_rate": 1.2674448842786559e-05, + "loss": 0.9602, + "step": 258860 + }, + { + "epoch": 22.41, + "learning_rate": 1.2673004131873213e-05, + "loss": 0.9705, + "step": 258870 + }, + { + "epoch": 22.41, + "learning_rate": 1.2671559420959866e-05, + "loss": 0.9383, + "step": 258880 + }, + { + "epoch": 22.41, + "learning_rate": 1.2670114710046521e-05, + "loss": 0.8999, + "step": 258890 + }, + { + "epoch": 22.41, + "learning_rate": 1.2668669999133173e-05, + "loss": 0.9045, + "step": 258900 + }, + { + "epoch": 22.41, + "learning_rate": 1.2667225288219828e-05, + "loss": 0.9907, + "step": 258910 + }, + { + "epoch": 22.41, + "learning_rate": 1.2665780577306483e-05, + "loss": 0.8913, + "step": 258920 + }, + { + "epoch": 22.41, + "learning_rate": 1.2664335866393135e-05, + "loss": 0.9698, + "step": 258930 + }, + { + "epoch": 22.41, + "learning_rate": 1.266289115547979e-05, + "loss": 0.9218, + "step": 258940 + }, + { + "epoch": 22.41, + "learning_rate": 1.2661446444566443e-05, + "loss": 0.9594, + "step": 258950 + }, + { + "epoch": 22.41, + "learning_rate": 1.2660001733653096e-05, + "loss": 0.9569, + "step": 258960 + }, + { + "epoch": 22.42, + "learning_rate": 1.265855702273975e-05, + "loss": 0.9098, + "step": 258970 + }, + { + "epoch": 22.42, + "learning_rate": 1.2657112311826405e-05, + "loss": 0.9223, + "step": 258980 + }, + { + "epoch": 22.42, + "learning_rate": 1.2655667600913057e-05, + "loss": 0.933, + "step": 258990 + }, + { + "epoch": 22.42, + "learning_rate": 1.2654222889999712e-05, + "loss": 0.881, + "step": 259000 + }, + { + "epoch": 22.42, + "learning_rate": 1.2652778179086367e-05, + "loss": 0.9433, + "step": 259010 + }, + { + "epoch": 22.42, + "learning_rate": 1.2651333468173019e-05, + "loss": 0.9275, + "step": 259020 + }, + { + "epoch": 22.42, + "learning_rate": 1.2649888757259674e-05, + "loss": 0.9258, + "step": 259030 + }, + { + "epoch": 22.42, + "learning_rate": 1.2648444046346327e-05, + "loss": 0.9581, + "step": 259040 + }, + { + "epoch": 22.42, + "learning_rate": 1.264699933543298e-05, + "loss": 0.933, + "step": 259050 + }, + { + "epoch": 22.42, + "learning_rate": 1.2645554624519634e-05, + "loss": 0.901, + "step": 259060 + }, + { + "epoch": 22.42, + "learning_rate": 1.2644109913606289e-05, + "loss": 0.9292, + "step": 259070 + }, + { + "epoch": 22.43, + "learning_rate": 1.264266520269294e-05, + "loss": 0.898, + "step": 259080 + }, + { + "epoch": 22.43, + "learning_rate": 1.2641220491779596e-05, + "loss": 0.8635, + "step": 259090 + }, + { + "epoch": 22.43, + "learning_rate": 1.263977578086625e-05, + "loss": 0.898, + "step": 259100 + }, + { + "epoch": 22.43, + "learning_rate": 1.2638331069952902e-05, + "loss": 0.8995, + "step": 259110 + }, + { + "epoch": 22.43, + "learning_rate": 1.2636886359039558e-05, + "loss": 0.9654, + "step": 259120 + }, + { + "epoch": 22.43, + "learning_rate": 1.2635441648126211e-05, + "loss": 0.9607, + "step": 259130 + }, + { + "epoch": 22.43, + "learning_rate": 1.2633996937212864e-05, + "loss": 0.9286, + "step": 259140 + }, + { + "epoch": 22.43, + "learning_rate": 1.2632552226299518e-05, + "loss": 0.8999, + "step": 259150 + }, + { + "epoch": 22.43, + "learning_rate": 1.2631107515386173e-05, + "loss": 0.8542, + "step": 259160 + }, + { + "epoch": 22.43, + "learning_rate": 1.2629662804472825e-05, + "loss": 0.9462, + "step": 259170 + }, + { + "epoch": 22.43, + "learning_rate": 1.262821809355948e-05, + "loss": 0.9076, + "step": 259180 + }, + { + "epoch": 22.43, + "learning_rate": 1.2626773382646135e-05, + "loss": 0.903, + "step": 259190 + }, + { + "epoch": 22.44, + "learning_rate": 1.2625328671732786e-05, + "loss": 0.9419, + "step": 259200 + }, + { + "epoch": 22.44, + "learning_rate": 1.2623883960819441e-05, + "loss": 0.9044, + "step": 259210 + }, + { + "epoch": 22.44, + "learning_rate": 1.2622439249906095e-05, + "loss": 0.9255, + "step": 259220 + }, + { + "epoch": 22.44, + "learning_rate": 1.2620994538992748e-05, + "loss": 0.8912, + "step": 259230 + }, + { + "epoch": 22.44, + "learning_rate": 1.2619549828079402e-05, + "loss": 0.9642, + "step": 259240 + }, + { + "epoch": 22.44, + "learning_rate": 1.2618105117166057e-05, + "loss": 0.9425, + "step": 259250 + }, + { + "epoch": 22.44, + "learning_rate": 1.2616660406252708e-05, + "loss": 0.89, + "step": 259260 + }, + { + "epoch": 22.44, + "learning_rate": 1.2615215695339364e-05, + "loss": 0.9251, + "step": 259270 + }, + { + "epoch": 22.44, + "learning_rate": 1.2613770984426019e-05, + "loss": 0.9404, + "step": 259280 + }, + { + "epoch": 22.44, + "learning_rate": 1.261232627351267e-05, + "loss": 0.9363, + "step": 259290 + }, + { + "epoch": 22.44, + "learning_rate": 1.2610881562599325e-05, + "loss": 0.9513, + "step": 259300 + }, + { + "epoch": 22.45, + "learning_rate": 1.2609436851685979e-05, + "loss": 0.9634, + "step": 259310 + }, + { + "epoch": 22.45, + "learning_rate": 1.2607992140772632e-05, + "loss": 0.9114, + "step": 259320 + }, + { + "epoch": 22.45, + "learning_rate": 1.2606547429859286e-05, + "loss": 0.953, + "step": 259330 + }, + { + "epoch": 22.45, + "learning_rate": 1.260510271894594e-05, + "loss": 0.915, + "step": 259340 + }, + { + "epoch": 22.45, + "learning_rate": 1.2603658008032592e-05, + "loss": 0.9229, + "step": 259350 + }, + { + "epoch": 22.45, + "learning_rate": 1.2602213297119247e-05, + "loss": 0.9415, + "step": 259360 + }, + { + "epoch": 22.45, + "learning_rate": 1.2600768586205903e-05, + "loss": 0.9304, + "step": 259370 + }, + { + "epoch": 22.45, + "learning_rate": 1.2599323875292554e-05, + "loss": 0.9151, + "step": 259380 + }, + { + "epoch": 22.45, + "learning_rate": 1.259787916437921e-05, + "loss": 0.9293, + "step": 259390 + }, + { + "epoch": 22.45, + "learning_rate": 1.2596434453465861e-05, + "loss": 0.9562, + "step": 259400 + }, + { + "epoch": 22.45, + "learning_rate": 1.2594989742552516e-05, + "loss": 0.8923, + "step": 259410 + }, + { + "epoch": 22.45, + "learning_rate": 1.259354503163917e-05, + "loss": 0.8747, + "step": 259420 + }, + { + "epoch": 22.46, + "learning_rate": 1.2592100320725821e-05, + "loss": 0.8987, + "step": 259430 + }, + { + "epoch": 22.46, + "learning_rate": 1.2590655609812476e-05, + "loss": 0.9083, + "step": 259440 + }, + { + "epoch": 22.46, + "learning_rate": 1.2589210898899131e-05, + "loss": 0.9289, + "step": 259450 + }, + { + "epoch": 22.46, + "learning_rate": 1.2587766187985783e-05, + "loss": 0.9438, + "step": 259460 + }, + { + "epoch": 22.46, + "learning_rate": 1.2586321477072438e-05, + "loss": 0.8832, + "step": 259470 + }, + { + "epoch": 22.46, + "learning_rate": 1.2584876766159093e-05, + "loss": 0.8713, + "step": 259480 + }, + { + "epoch": 22.46, + "learning_rate": 1.2583432055245745e-05, + "loss": 0.9335, + "step": 259490 + }, + { + "epoch": 22.46, + "learning_rate": 1.25819873443324e-05, + "loss": 0.9198, + "step": 259500 + }, + { + "epoch": 22.46, + "learning_rate": 1.2580542633419054e-05, + "loss": 0.9332, + "step": 259510 + }, + { + "epoch": 22.46, + "learning_rate": 1.2579097922505705e-05, + "loss": 0.8646, + "step": 259520 + }, + { + "epoch": 22.46, + "learning_rate": 1.257765321159236e-05, + "loss": 0.9422, + "step": 259530 + }, + { + "epoch": 22.47, + "learning_rate": 1.2576208500679015e-05, + "loss": 0.8915, + "step": 259540 + }, + { + "epoch": 22.47, + "learning_rate": 1.2574763789765667e-05, + "loss": 0.9744, + "step": 259550 + }, + { + "epoch": 22.47, + "learning_rate": 1.2573319078852322e-05, + "loss": 0.9304, + "step": 259560 + }, + { + "epoch": 22.47, + "learning_rate": 1.2571874367938977e-05, + "loss": 0.8826, + "step": 259570 + }, + { + "epoch": 22.47, + "learning_rate": 1.2570429657025629e-05, + "loss": 0.8845, + "step": 259580 + }, + { + "epoch": 22.47, + "learning_rate": 1.2568984946112284e-05, + "loss": 0.884, + "step": 259590 + }, + { + "epoch": 22.47, + "learning_rate": 1.2567540235198937e-05, + "loss": 0.9187, + "step": 259600 + }, + { + "epoch": 22.47, + "learning_rate": 1.2566095524285589e-05, + "loss": 0.9179, + "step": 259610 + }, + { + "epoch": 22.47, + "learning_rate": 1.2564650813372244e-05, + "loss": 0.9501, + "step": 259620 + }, + { + "epoch": 22.47, + "learning_rate": 1.25632061024589e-05, + "loss": 0.8683, + "step": 259630 + }, + { + "epoch": 22.47, + "learning_rate": 1.2561761391545551e-05, + "loss": 0.9964, + "step": 259640 + }, + { + "epoch": 22.47, + "learning_rate": 1.2560316680632206e-05, + "loss": 0.9089, + "step": 259650 + }, + { + "epoch": 22.48, + "learning_rate": 1.2558871969718861e-05, + "loss": 0.8386, + "step": 259660 + }, + { + "epoch": 22.48, + "learning_rate": 1.2557427258805513e-05, + "loss": 0.9125, + "step": 259670 + }, + { + "epoch": 22.48, + "learning_rate": 1.2555982547892168e-05, + "loss": 0.8888, + "step": 259680 + }, + { + "epoch": 22.48, + "learning_rate": 1.2554537836978821e-05, + "loss": 0.875, + "step": 259690 + }, + { + "epoch": 22.48, + "learning_rate": 1.2553093126065473e-05, + "loss": 0.9201, + "step": 259700 + }, + { + "epoch": 22.48, + "learning_rate": 1.2551648415152128e-05, + "loss": 0.9211, + "step": 259710 + }, + { + "epoch": 22.48, + "learning_rate": 1.2550203704238783e-05, + "loss": 0.9886, + "step": 259720 + }, + { + "epoch": 22.48, + "learning_rate": 1.2548758993325435e-05, + "loss": 0.9202, + "step": 259730 + }, + { + "epoch": 22.48, + "learning_rate": 1.254731428241209e-05, + "loss": 0.9097, + "step": 259740 + }, + { + "epoch": 22.48, + "learning_rate": 1.2545869571498745e-05, + "loss": 0.9524, + "step": 259750 + }, + { + "epoch": 22.48, + "learning_rate": 1.2544424860585397e-05, + "loss": 0.9431, + "step": 259760 + }, + { + "epoch": 22.48, + "learning_rate": 1.2542980149672052e-05, + "loss": 0.9298, + "step": 259770 + }, + { + "epoch": 22.49, + "learning_rate": 1.2541535438758705e-05, + "loss": 0.9271, + "step": 259780 + }, + { + "epoch": 22.49, + "learning_rate": 1.2540090727845357e-05, + "loss": 0.9229, + "step": 259790 + }, + { + "epoch": 22.49, + "learning_rate": 1.2538646016932012e-05, + "loss": 0.8874, + "step": 259800 + }, + { + "epoch": 22.49, + "learning_rate": 1.2537201306018667e-05, + "loss": 0.8497, + "step": 259810 + }, + { + "epoch": 22.49, + "learning_rate": 1.2535756595105319e-05, + "loss": 0.921, + "step": 259820 + }, + { + "epoch": 22.49, + "learning_rate": 1.2534311884191974e-05, + "loss": 0.9003, + "step": 259830 + }, + { + "epoch": 22.49, + "learning_rate": 1.2532867173278629e-05, + "loss": 0.9086, + "step": 259840 + }, + { + "epoch": 22.49, + "learning_rate": 1.253142246236528e-05, + "loss": 0.9332, + "step": 259850 + }, + { + "epoch": 22.49, + "learning_rate": 1.2529977751451936e-05, + "loss": 0.9949, + "step": 259860 + }, + { + "epoch": 22.49, + "learning_rate": 1.252853304053859e-05, + "loss": 0.9412, + "step": 259870 + }, + { + "epoch": 22.49, + "learning_rate": 1.2527088329625241e-05, + "loss": 0.9758, + "step": 259880 + }, + { + "epoch": 22.5, + "learning_rate": 1.2525643618711896e-05, + "loss": 0.9466, + "step": 259890 + }, + { + "epoch": 22.5, + "learning_rate": 1.2524198907798551e-05, + "loss": 0.9564, + "step": 259900 + }, + { + "epoch": 22.5, + "learning_rate": 1.2522754196885203e-05, + "loss": 0.9979, + "step": 259910 + }, + { + "epoch": 22.5, + "learning_rate": 1.2521309485971858e-05, + "loss": 0.9114, + "step": 259920 + }, + { + "epoch": 22.5, + "learning_rate": 1.2519864775058513e-05, + "loss": 0.9219, + "step": 259930 + }, + { + "epoch": 22.5, + "learning_rate": 1.2518420064145165e-05, + "loss": 0.9143, + "step": 259940 + }, + { + "epoch": 22.5, + "learning_rate": 1.251697535323182e-05, + "loss": 0.9075, + "step": 259950 + }, + { + "epoch": 22.5, + "learning_rate": 1.2515530642318473e-05, + "loss": 0.9527, + "step": 259960 + }, + { + "epoch": 22.5, + "learning_rate": 1.2514085931405125e-05, + "loss": 0.9336, + "step": 259970 + }, + { + "epoch": 22.5, + "learning_rate": 1.251264122049178e-05, + "loss": 0.8953, + "step": 259980 + }, + { + "epoch": 22.5, + "learning_rate": 1.2511196509578435e-05, + "loss": 0.9623, + "step": 259990 + }, + { + "epoch": 22.5, + "learning_rate": 1.2509751798665087e-05, + "loss": 0.9073, + "step": 260000 + }, + { + "epoch": 22.51, + "learning_rate": 1.2508307087751742e-05, + "loss": 0.9564, + "step": 260010 + }, + { + "epoch": 22.51, + "learning_rate": 1.2506862376838397e-05, + "loss": 0.8668, + "step": 260020 + }, + { + "epoch": 22.51, + "learning_rate": 1.2505417665925049e-05, + "loss": 0.9365, + "step": 260030 + }, + { + "epoch": 22.51, + "learning_rate": 1.2503972955011704e-05, + "loss": 0.8909, + "step": 260040 + }, + { + "epoch": 22.51, + "learning_rate": 1.2502528244098357e-05, + "loss": 0.8851, + "step": 260050 + }, + { + "epoch": 22.51, + "learning_rate": 1.2501083533185009e-05, + "loss": 0.9595, + "step": 260060 + }, + { + "epoch": 22.51, + "learning_rate": 1.2499638822271664e-05, + "loss": 0.9595, + "step": 260070 + }, + { + "epoch": 22.51, + "learning_rate": 1.2498194111358317e-05, + "loss": 0.8969, + "step": 260080 + }, + { + "epoch": 22.51, + "learning_rate": 1.249674940044497e-05, + "loss": 0.9395, + "step": 260090 + }, + { + "epoch": 22.51, + "learning_rate": 1.2495304689531626e-05, + "loss": 0.8975, + "step": 260100 + }, + { + "epoch": 22.51, + "learning_rate": 1.249385997861828e-05, + "loss": 0.9141, + "step": 260110 + }, + { + "epoch": 22.52, + "learning_rate": 1.2492415267704933e-05, + "loss": 0.9578, + "step": 260120 + }, + { + "epoch": 22.52, + "learning_rate": 1.2490970556791588e-05, + "loss": 0.9146, + "step": 260130 + }, + { + "epoch": 22.52, + "learning_rate": 1.248952584587824e-05, + "loss": 0.9229, + "step": 260140 + }, + { + "epoch": 22.52, + "learning_rate": 1.2488081134964893e-05, + "loss": 0.9946, + "step": 260150 + }, + { + "epoch": 22.52, + "learning_rate": 1.2486636424051548e-05, + "loss": 1.0259, + "step": 260160 + }, + { + "epoch": 22.52, + "learning_rate": 1.2485191713138201e-05, + "loss": 0.9195, + "step": 260170 + }, + { + "epoch": 22.52, + "learning_rate": 1.2483747002224855e-05, + "loss": 0.8973, + "step": 260180 + }, + { + "epoch": 22.52, + "learning_rate": 1.248230229131151e-05, + "loss": 0.938, + "step": 260190 + }, + { + "epoch": 22.52, + "learning_rate": 1.2480857580398163e-05, + "loss": 0.8562, + "step": 260200 + }, + { + "epoch": 22.52, + "learning_rate": 1.2479412869484817e-05, + "loss": 0.9196, + "step": 260210 + }, + { + "epoch": 22.52, + "learning_rate": 1.2477968158571472e-05, + "loss": 0.9802, + "step": 260220 + }, + { + "epoch": 22.52, + "learning_rate": 1.2476523447658123e-05, + "loss": 0.9093, + "step": 260230 + }, + { + "epoch": 22.53, + "learning_rate": 1.2475078736744777e-05, + "loss": 0.8843, + "step": 260240 + }, + { + "epoch": 22.53, + "learning_rate": 1.2473634025831432e-05, + "loss": 0.9563, + "step": 260250 + }, + { + "epoch": 22.53, + "learning_rate": 1.2472189314918085e-05, + "loss": 0.944, + "step": 260260 + }, + { + "epoch": 22.53, + "learning_rate": 1.2470744604004739e-05, + "loss": 0.9307, + "step": 260270 + }, + { + "epoch": 22.53, + "learning_rate": 1.2469299893091394e-05, + "loss": 0.9223, + "step": 260280 + }, + { + "epoch": 22.53, + "learning_rate": 1.2467855182178047e-05, + "loss": 0.9538, + "step": 260290 + }, + { + "epoch": 22.53, + "learning_rate": 1.24664104712647e-05, + "loss": 0.9257, + "step": 260300 + }, + { + "epoch": 22.53, + "learning_rate": 1.2464965760351356e-05, + "loss": 0.9323, + "step": 260310 + }, + { + "epoch": 22.53, + "learning_rate": 1.2463521049438007e-05, + "loss": 0.8721, + "step": 260320 + }, + { + "epoch": 22.53, + "learning_rate": 1.246207633852466e-05, + "loss": 0.9266, + "step": 260330 + }, + { + "epoch": 22.53, + "learning_rate": 1.2460631627611316e-05, + "loss": 0.9165, + "step": 260340 + }, + { + "epoch": 22.54, + "learning_rate": 1.245918691669797e-05, + "loss": 0.9047, + "step": 260350 + }, + { + "epoch": 22.54, + "learning_rate": 1.2457742205784623e-05, + "loss": 0.9581, + "step": 260360 + }, + { + "epoch": 22.54, + "learning_rate": 1.2456297494871278e-05, + "loss": 0.9066, + "step": 260370 + }, + { + "epoch": 22.54, + "learning_rate": 1.2454852783957931e-05, + "loss": 0.9223, + "step": 260380 + }, + { + "epoch": 22.54, + "learning_rate": 1.2453408073044584e-05, + "loss": 0.947, + "step": 260390 + }, + { + "epoch": 22.54, + "learning_rate": 1.245196336213124e-05, + "loss": 0.9171, + "step": 260400 + }, + { + "epoch": 22.54, + "learning_rate": 1.2450518651217891e-05, + "loss": 0.9495, + "step": 260410 + }, + { + "epoch": 22.54, + "learning_rate": 1.2449073940304545e-05, + "loss": 0.9319, + "step": 260420 + }, + { + "epoch": 22.54, + "learning_rate": 1.24476292293912e-05, + "loss": 0.9218, + "step": 260430 + }, + { + "epoch": 22.54, + "learning_rate": 1.2446184518477853e-05, + "loss": 0.9855, + "step": 260440 + }, + { + "epoch": 22.54, + "learning_rate": 1.2444739807564507e-05, + "loss": 0.9415, + "step": 260450 + }, + { + "epoch": 22.54, + "learning_rate": 1.2443295096651162e-05, + "loss": 0.8643, + "step": 260460 + }, + { + "epoch": 22.55, + "learning_rate": 1.2441850385737815e-05, + "loss": 0.897, + "step": 260470 + }, + { + "epoch": 22.55, + "learning_rate": 1.2440405674824468e-05, + "loss": 0.9621, + "step": 260480 + }, + { + "epoch": 22.55, + "learning_rate": 1.2438960963911123e-05, + "loss": 0.9357, + "step": 260490 + }, + { + "epoch": 22.55, + "learning_rate": 1.2437516252997775e-05, + "loss": 0.8511, + "step": 260500 + }, + { + "epoch": 22.55, + "learning_rate": 1.2436071542084429e-05, + "loss": 0.9766, + "step": 260510 + }, + { + "epoch": 22.55, + "learning_rate": 1.2434626831171082e-05, + "loss": 0.931, + "step": 260520 + }, + { + "epoch": 22.55, + "learning_rate": 1.2433182120257737e-05, + "loss": 0.9029, + "step": 260530 + }, + { + "epoch": 22.55, + "learning_rate": 1.243173740934439e-05, + "loss": 0.9006, + "step": 260540 + }, + { + "epoch": 22.55, + "learning_rate": 1.2430292698431044e-05, + "loss": 0.9469, + "step": 260550 + }, + { + "epoch": 22.55, + "learning_rate": 1.2428847987517699e-05, + "loss": 0.9319, + "step": 260560 + }, + { + "epoch": 22.55, + "learning_rate": 1.2427403276604352e-05, + "loss": 0.9376, + "step": 260570 + }, + { + "epoch": 22.56, + "learning_rate": 1.2425958565691006e-05, + "loss": 0.9565, + "step": 260580 + }, + { + "epoch": 22.56, + "learning_rate": 1.2424513854777659e-05, + "loss": 0.9267, + "step": 260590 + }, + { + "epoch": 22.56, + "learning_rate": 1.2423069143864313e-05, + "loss": 0.8955, + "step": 260600 + }, + { + "epoch": 22.56, + "learning_rate": 1.2421624432950966e-05, + "loss": 0.9761, + "step": 260610 + }, + { + "epoch": 22.56, + "learning_rate": 1.2420179722037621e-05, + "loss": 0.9438, + "step": 260620 + }, + { + "epoch": 22.56, + "learning_rate": 1.2418735011124274e-05, + "loss": 0.9372, + "step": 260630 + }, + { + "epoch": 22.56, + "learning_rate": 1.2417290300210928e-05, + "loss": 0.9042, + "step": 260640 + }, + { + "epoch": 22.56, + "learning_rate": 1.2415845589297583e-05, + "loss": 0.9552, + "step": 260650 + }, + { + "epoch": 22.56, + "learning_rate": 1.2414400878384236e-05, + "loss": 0.8909, + "step": 260660 + }, + { + "epoch": 22.56, + "learning_rate": 1.241295616747089e-05, + "loss": 0.9922, + "step": 260670 + }, + { + "epoch": 22.56, + "learning_rate": 1.2411511456557543e-05, + "loss": 0.9087, + "step": 260680 + }, + { + "epoch": 22.56, + "learning_rate": 1.2410066745644196e-05, + "loss": 0.8838, + "step": 260690 + }, + { + "epoch": 22.57, + "learning_rate": 1.240862203473085e-05, + "loss": 0.9201, + "step": 260700 + }, + { + "epoch": 22.57, + "learning_rate": 1.2407177323817505e-05, + "loss": 0.9112, + "step": 260710 + }, + { + "epoch": 22.57, + "learning_rate": 1.2405732612904158e-05, + "loss": 0.8835, + "step": 260720 + }, + { + "epoch": 22.57, + "learning_rate": 1.2404287901990812e-05, + "loss": 0.921, + "step": 260730 + }, + { + "epoch": 22.57, + "learning_rate": 1.2402843191077467e-05, + "loss": 0.9708, + "step": 260740 + }, + { + "epoch": 22.57, + "learning_rate": 1.240139848016412e-05, + "loss": 0.925, + "step": 260750 + }, + { + "epoch": 22.57, + "learning_rate": 1.2399953769250774e-05, + "loss": 0.9133, + "step": 260760 + }, + { + "epoch": 22.57, + "learning_rate": 1.2398509058337427e-05, + "loss": 0.8935, + "step": 260770 + }, + { + "epoch": 22.57, + "learning_rate": 1.239706434742408e-05, + "loss": 0.8829, + "step": 260780 + }, + { + "epoch": 22.57, + "learning_rate": 1.2395619636510734e-05, + "loss": 0.941, + "step": 260790 + }, + { + "epoch": 22.57, + "learning_rate": 1.2394174925597389e-05, + "loss": 0.947, + "step": 260800 + }, + { + "epoch": 22.57, + "learning_rate": 1.2392730214684042e-05, + "loss": 0.9611, + "step": 260810 + }, + { + "epoch": 22.58, + "learning_rate": 1.2391285503770696e-05, + "loss": 0.884, + "step": 260820 + }, + { + "epoch": 22.58, + "learning_rate": 1.238984079285735e-05, + "loss": 0.9316, + "step": 260830 + }, + { + "epoch": 22.58, + "learning_rate": 1.2388396081944004e-05, + "loss": 0.9048, + "step": 260840 + }, + { + "epoch": 22.58, + "learning_rate": 1.2386951371030658e-05, + "loss": 0.9107, + "step": 260850 + }, + { + "epoch": 22.58, + "learning_rate": 1.2385506660117311e-05, + "loss": 0.9251, + "step": 260860 + }, + { + "epoch": 22.58, + "learning_rate": 1.2384061949203964e-05, + "loss": 0.9233, + "step": 260870 + }, + { + "epoch": 22.58, + "learning_rate": 1.2382617238290618e-05, + "loss": 0.9138, + "step": 260880 + }, + { + "epoch": 22.58, + "learning_rate": 1.2381172527377273e-05, + "loss": 0.9438, + "step": 260890 + }, + { + "epoch": 22.58, + "learning_rate": 1.2379727816463926e-05, + "loss": 0.9189, + "step": 260900 + }, + { + "epoch": 22.58, + "learning_rate": 1.237828310555058e-05, + "loss": 0.95, + "step": 260910 + }, + { + "epoch": 22.58, + "learning_rate": 1.2376838394637235e-05, + "loss": 0.9191, + "step": 260920 + }, + { + "epoch": 22.59, + "learning_rate": 1.2375393683723888e-05, + "loss": 0.8908, + "step": 260930 + }, + { + "epoch": 22.59, + "learning_rate": 1.2373948972810541e-05, + "loss": 0.9015, + "step": 260940 + }, + { + "epoch": 22.59, + "learning_rate": 1.2372504261897195e-05, + "loss": 0.8919, + "step": 260950 + }, + { + "epoch": 22.59, + "learning_rate": 1.2371059550983848e-05, + "loss": 0.9132, + "step": 260960 + }, + { + "epoch": 22.59, + "learning_rate": 1.2369614840070502e-05, + "loss": 0.9305, + "step": 260970 + }, + { + "epoch": 22.59, + "learning_rate": 1.2368170129157155e-05, + "loss": 0.9649, + "step": 260980 + }, + { + "epoch": 22.59, + "learning_rate": 1.236672541824381e-05, + "loss": 0.9939, + "step": 260990 + }, + { + "epoch": 22.59, + "learning_rate": 1.2365280707330464e-05, + "loss": 0.9205, + "step": 261000 + }, + { + "epoch": 22.59, + "learning_rate": 1.2363835996417117e-05, + "loss": 0.9669, + "step": 261010 + }, + { + "epoch": 22.59, + "learning_rate": 1.2362391285503772e-05, + "loss": 0.8995, + "step": 261020 + }, + { + "epoch": 22.59, + "learning_rate": 1.2360946574590425e-05, + "loss": 0.9858, + "step": 261030 + }, + { + "epoch": 22.59, + "learning_rate": 1.2359501863677079e-05, + "loss": 0.9724, + "step": 261040 + }, + { + "epoch": 22.6, + "learning_rate": 1.2358057152763732e-05, + "loss": 0.919, + "step": 261050 + }, + { + "epoch": 22.6, + "learning_rate": 1.2356612441850386e-05, + "loss": 0.8521, + "step": 261060 + }, + { + "epoch": 22.6, + "learning_rate": 1.2355167730937039e-05, + "loss": 0.9441, + "step": 261070 + }, + { + "epoch": 22.6, + "learning_rate": 1.2353723020023694e-05, + "loss": 0.8991, + "step": 261080 + }, + { + "epoch": 22.6, + "learning_rate": 1.2352278309110348e-05, + "loss": 0.9509, + "step": 261090 + }, + { + "epoch": 22.6, + "learning_rate": 1.2350833598197001e-05, + "loss": 0.9565, + "step": 261100 + }, + { + "epoch": 22.6, + "learning_rate": 1.2349388887283656e-05, + "loss": 0.9103, + "step": 261110 + }, + { + "epoch": 22.6, + "learning_rate": 1.234794417637031e-05, + "loss": 0.9295, + "step": 261120 + }, + { + "epoch": 22.6, + "learning_rate": 1.2346499465456963e-05, + "loss": 0.9165, + "step": 261130 + }, + { + "epoch": 22.6, + "learning_rate": 1.2345054754543616e-05, + "loss": 0.9288, + "step": 261140 + }, + { + "epoch": 22.6, + "learning_rate": 1.234361004363027e-05, + "loss": 0.8908, + "step": 261150 + }, + { + "epoch": 22.61, + "learning_rate": 1.2342165332716923e-05, + "loss": 0.8765, + "step": 261160 + }, + { + "epoch": 22.61, + "learning_rate": 1.2340720621803578e-05, + "loss": 0.9439, + "step": 261170 + }, + { + "epoch": 22.61, + "learning_rate": 1.2339275910890231e-05, + "loss": 0.9429, + "step": 261180 + }, + { + "epoch": 22.61, + "learning_rate": 1.2337831199976885e-05, + "loss": 0.8858, + "step": 261190 + }, + { + "epoch": 22.61, + "learning_rate": 1.233638648906354e-05, + "loss": 0.915, + "step": 261200 + }, + { + "epoch": 22.61, + "learning_rate": 1.2334941778150193e-05, + "loss": 0.9295, + "step": 261210 + }, + { + "epoch": 22.61, + "learning_rate": 1.2333497067236847e-05, + "loss": 0.9852, + "step": 261220 + }, + { + "epoch": 22.61, + "learning_rate": 1.23320523563235e-05, + "loss": 0.8989, + "step": 261230 + }, + { + "epoch": 22.61, + "learning_rate": 1.2330607645410154e-05, + "loss": 0.935, + "step": 261240 + }, + { + "epoch": 22.61, + "learning_rate": 1.2329162934496807e-05, + "loss": 0.991, + "step": 261250 + }, + { + "epoch": 22.61, + "learning_rate": 1.2327718223583462e-05, + "loss": 0.9187, + "step": 261260 + }, + { + "epoch": 22.61, + "learning_rate": 1.2326273512670115e-05, + "loss": 0.9777, + "step": 261270 + }, + { + "epoch": 22.62, + "learning_rate": 1.2324828801756769e-05, + "loss": 0.9555, + "step": 261280 + }, + { + "epoch": 22.62, + "learning_rate": 1.2323384090843424e-05, + "loss": 0.8759, + "step": 261290 + }, + { + "epoch": 22.62, + "learning_rate": 1.2321939379930077e-05, + "loss": 0.9178, + "step": 261300 + }, + { + "epoch": 22.62, + "learning_rate": 1.232049466901673e-05, + "loss": 0.9159, + "step": 261310 + }, + { + "epoch": 22.62, + "learning_rate": 1.2319049958103384e-05, + "loss": 0.9068, + "step": 261320 + }, + { + "epoch": 22.62, + "learning_rate": 1.2317605247190037e-05, + "loss": 0.9505, + "step": 261330 + }, + { + "epoch": 22.62, + "learning_rate": 1.2316160536276691e-05, + "loss": 0.8895, + "step": 261340 + }, + { + "epoch": 22.62, + "learning_rate": 1.2314715825363346e-05, + "loss": 0.9305, + "step": 261350 + }, + { + "epoch": 22.62, + "learning_rate": 1.231327111445e-05, + "loss": 0.9424, + "step": 261360 + }, + { + "epoch": 22.62, + "learning_rate": 1.2311826403536653e-05, + "loss": 0.9385, + "step": 261370 + }, + { + "epoch": 22.62, + "learning_rate": 1.2310381692623308e-05, + "loss": 0.9241, + "step": 261380 + }, + { + "epoch": 22.63, + "learning_rate": 1.2308936981709961e-05, + "loss": 0.881, + "step": 261390 + }, + { + "epoch": 22.63, + "learning_rate": 1.2307492270796615e-05, + "loss": 0.9235, + "step": 261400 + }, + { + "epoch": 22.63, + "learning_rate": 1.2306047559883268e-05, + "loss": 0.9425, + "step": 261410 + }, + { + "epoch": 22.63, + "learning_rate": 1.2304602848969921e-05, + "loss": 0.9478, + "step": 261420 + }, + { + "epoch": 22.63, + "learning_rate": 1.2303158138056575e-05, + "loss": 0.9303, + "step": 261430 + }, + { + "epoch": 22.63, + "learning_rate": 1.2301713427143228e-05, + "loss": 0.9158, + "step": 261440 + }, + { + "epoch": 22.63, + "learning_rate": 1.2300268716229883e-05, + "loss": 0.9451, + "step": 261450 + }, + { + "epoch": 22.63, + "learning_rate": 1.2298824005316537e-05, + "loss": 0.918, + "step": 261460 + }, + { + "epoch": 22.63, + "learning_rate": 1.229737929440319e-05, + "loss": 0.9107, + "step": 261470 + }, + { + "epoch": 22.63, + "learning_rate": 1.2295934583489845e-05, + "loss": 0.9587, + "step": 261480 + }, + { + "epoch": 22.63, + "learning_rate": 1.2294489872576499e-05, + "loss": 0.9111, + "step": 261490 + }, + { + "epoch": 22.63, + "learning_rate": 1.2293045161663152e-05, + "loss": 0.959, + "step": 261500 + }, + { + "epoch": 22.64, + "learning_rate": 1.2291600450749805e-05, + "loss": 0.9409, + "step": 261510 + }, + { + "epoch": 22.64, + "learning_rate": 1.2290155739836459e-05, + "loss": 0.9467, + "step": 261520 + }, + { + "epoch": 22.64, + "learning_rate": 1.2288711028923112e-05, + "loss": 0.9002, + "step": 261530 + }, + { + "epoch": 22.64, + "learning_rate": 1.2287266318009767e-05, + "loss": 0.9852, + "step": 261540 + }, + { + "epoch": 22.64, + "learning_rate": 1.228582160709642e-05, + "loss": 0.9113, + "step": 261550 + }, + { + "epoch": 22.64, + "learning_rate": 1.2284376896183074e-05, + "loss": 0.9251, + "step": 261560 + }, + { + "epoch": 22.64, + "learning_rate": 1.2282932185269729e-05, + "loss": 0.9276, + "step": 261570 + }, + { + "epoch": 22.64, + "learning_rate": 1.2281487474356382e-05, + "loss": 1.0012, + "step": 261580 + }, + { + "epoch": 22.64, + "learning_rate": 1.2280042763443036e-05, + "loss": 0.8536, + "step": 261590 + }, + { + "epoch": 22.64, + "learning_rate": 1.227859805252969e-05, + "loss": 0.9201, + "step": 261600 + }, + { + "epoch": 22.64, + "learning_rate": 1.2277153341616343e-05, + "loss": 0.8889, + "step": 261610 + }, + { + "epoch": 22.65, + "learning_rate": 1.2275708630702996e-05, + "loss": 0.944, + "step": 261620 + }, + { + "epoch": 22.65, + "learning_rate": 1.2274263919789651e-05, + "loss": 0.9399, + "step": 261630 + }, + { + "epoch": 22.65, + "learning_rate": 1.2272819208876305e-05, + "loss": 0.9079, + "step": 261640 + }, + { + "epoch": 22.65, + "learning_rate": 1.2271374497962958e-05, + "loss": 0.9235, + "step": 261650 + }, + { + "epoch": 22.65, + "learning_rate": 1.2269929787049613e-05, + "loss": 0.932, + "step": 261660 + }, + { + "epoch": 22.65, + "learning_rate": 1.2268485076136266e-05, + "loss": 0.9299, + "step": 261670 + }, + { + "epoch": 22.65, + "learning_rate": 1.226704036522292e-05, + "loss": 0.9598, + "step": 261680 + }, + { + "epoch": 22.65, + "learning_rate": 1.2265595654309573e-05, + "loss": 0.8678, + "step": 261690 + }, + { + "epoch": 22.65, + "learning_rate": 1.2264150943396227e-05, + "loss": 0.8683, + "step": 261700 + }, + { + "epoch": 22.65, + "learning_rate": 1.226270623248288e-05, + "loss": 0.9223, + "step": 261710 + }, + { + "epoch": 22.65, + "learning_rate": 1.2261261521569535e-05, + "loss": 0.9266, + "step": 261720 + }, + { + "epoch": 22.65, + "learning_rate": 1.2259816810656188e-05, + "loss": 0.9422, + "step": 261730 + }, + { + "epoch": 22.66, + "learning_rate": 1.2258372099742842e-05, + "loss": 0.9497, + "step": 261740 + }, + { + "epoch": 22.66, + "learning_rate": 1.2256927388829497e-05, + "loss": 0.9134, + "step": 261750 + }, + { + "epoch": 22.66, + "learning_rate": 1.225548267791615e-05, + "loss": 0.9593, + "step": 261760 + }, + { + "epoch": 22.66, + "learning_rate": 1.2254037967002804e-05, + "loss": 0.8971, + "step": 261770 + }, + { + "epoch": 22.66, + "learning_rate": 1.2252593256089457e-05, + "loss": 0.9245, + "step": 261780 + }, + { + "epoch": 22.66, + "learning_rate": 1.225114854517611e-05, + "loss": 0.9538, + "step": 261790 + }, + { + "epoch": 22.66, + "learning_rate": 1.2249703834262764e-05, + "loss": 0.9686, + "step": 261800 + }, + { + "epoch": 22.66, + "learning_rate": 1.2248259123349419e-05, + "loss": 0.903, + "step": 261810 + }, + { + "epoch": 22.66, + "learning_rate": 1.2246814412436072e-05, + "loss": 0.9781, + "step": 261820 + }, + { + "epoch": 22.66, + "learning_rate": 1.2245369701522726e-05, + "loss": 0.9466, + "step": 261830 + }, + { + "epoch": 22.66, + "learning_rate": 1.2243924990609381e-05, + "loss": 0.9597, + "step": 261840 + }, + { + "epoch": 22.66, + "learning_rate": 1.2242480279696034e-05, + "loss": 0.9292, + "step": 261850 + }, + { + "epoch": 22.67, + "learning_rate": 1.2241035568782688e-05, + "loss": 0.8922, + "step": 261860 + }, + { + "epoch": 22.67, + "learning_rate": 1.2239590857869341e-05, + "loss": 0.9218, + "step": 261870 + }, + { + "epoch": 22.67, + "learning_rate": 1.2238146146955995e-05, + "loss": 0.9545, + "step": 261880 + }, + { + "epoch": 22.67, + "learning_rate": 1.2236701436042648e-05, + "loss": 0.8403, + "step": 261890 + }, + { + "epoch": 22.67, + "learning_rate": 1.2235256725129301e-05, + "loss": 0.8881, + "step": 261900 + }, + { + "epoch": 22.67, + "learning_rate": 1.2233812014215956e-05, + "loss": 0.9582, + "step": 261910 + }, + { + "epoch": 22.67, + "learning_rate": 1.223236730330261e-05, + "loss": 0.9351, + "step": 261920 + }, + { + "epoch": 22.67, + "learning_rate": 1.2230922592389263e-05, + "loss": 0.9025, + "step": 261930 + }, + { + "epoch": 22.67, + "learning_rate": 1.2229477881475918e-05, + "loss": 0.8911, + "step": 261940 + }, + { + "epoch": 22.67, + "learning_rate": 1.2228033170562572e-05, + "loss": 0.9213, + "step": 261950 + }, + { + "epoch": 22.67, + "learning_rate": 1.2226588459649223e-05, + "loss": 0.9176, + "step": 261960 + }, + { + "epoch": 22.68, + "learning_rate": 1.2225143748735878e-05, + "loss": 0.9612, + "step": 261970 + }, + { + "epoch": 22.68, + "learning_rate": 1.2223699037822532e-05, + "loss": 0.933, + "step": 261980 + }, + { + "epoch": 22.68, + "learning_rate": 1.2222254326909185e-05, + "loss": 0.9194, + "step": 261990 + }, + { + "epoch": 22.68, + "learning_rate": 1.222080961599584e-05, + "loss": 0.9102, + "step": 262000 + }, + { + "epoch": 22.68, + "learning_rate": 1.2219364905082494e-05, + "loss": 0.9396, + "step": 262010 + }, + { + "epoch": 22.68, + "learning_rate": 1.2217920194169147e-05, + "loss": 0.9529, + "step": 262020 + }, + { + "epoch": 22.68, + "learning_rate": 1.2216475483255802e-05, + "loss": 0.9118, + "step": 262030 + }, + { + "epoch": 22.68, + "learning_rate": 1.2215030772342456e-05, + "loss": 0.939, + "step": 262040 + }, + { + "epoch": 22.68, + "learning_rate": 1.2213586061429107e-05, + "loss": 0.9329, + "step": 262050 + }, + { + "epoch": 22.68, + "learning_rate": 1.2212141350515762e-05, + "loss": 0.9024, + "step": 262060 + }, + { + "epoch": 22.68, + "learning_rate": 1.2210696639602416e-05, + "loss": 0.9595, + "step": 262070 + }, + { + "epoch": 22.68, + "learning_rate": 1.220925192868907e-05, + "loss": 0.8835, + "step": 262080 + }, + { + "epoch": 22.69, + "learning_rate": 1.2207807217775724e-05, + "loss": 0.9479, + "step": 262090 + }, + { + "epoch": 22.69, + "learning_rate": 1.2206362506862378e-05, + "loss": 0.8501, + "step": 262100 + }, + { + "epoch": 22.69, + "learning_rate": 1.2204917795949031e-05, + "loss": 0.9101, + "step": 262110 + }, + { + "epoch": 22.69, + "learning_rate": 1.2203473085035686e-05, + "loss": 0.974, + "step": 262120 + }, + { + "epoch": 22.69, + "learning_rate": 1.220202837412234e-05, + "loss": 0.8852, + "step": 262130 + }, + { + "epoch": 22.69, + "learning_rate": 1.2200583663208991e-05, + "loss": 0.9231, + "step": 262140 + }, + { + "epoch": 22.69, + "learning_rate": 1.2199138952295646e-05, + "loss": 0.9102, + "step": 262150 + }, + { + "epoch": 22.69, + "learning_rate": 1.21976942413823e-05, + "loss": 0.9339, + "step": 262160 + }, + { + "epoch": 22.69, + "learning_rate": 1.2196249530468953e-05, + "loss": 0.9534, + "step": 262170 + }, + { + "epoch": 22.69, + "learning_rate": 1.2194804819555608e-05, + "loss": 0.9197, + "step": 262180 + }, + { + "epoch": 22.69, + "learning_rate": 1.2193360108642262e-05, + "loss": 0.9668, + "step": 262190 + }, + { + "epoch": 22.7, + "learning_rate": 1.2191915397728915e-05, + "loss": 0.9577, + "step": 262200 + }, + { + "epoch": 22.7, + "learning_rate": 1.219047068681557e-05, + "loss": 0.9106, + "step": 262210 + }, + { + "epoch": 22.7, + "learning_rate": 1.2189025975902223e-05, + "loss": 0.946, + "step": 262220 + }, + { + "epoch": 22.7, + "learning_rate": 1.2187581264988875e-05, + "loss": 0.9287, + "step": 262230 + }, + { + "epoch": 22.7, + "learning_rate": 1.218613655407553e-05, + "loss": 0.9412, + "step": 262240 + }, + { + "epoch": 22.7, + "learning_rate": 1.2184691843162184e-05, + "loss": 0.8997, + "step": 262250 + }, + { + "epoch": 22.7, + "learning_rate": 1.2183247132248837e-05, + "loss": 0.8626, + "step": 262260 + }, + { + "epoch": 22.7, + "learning_rate": 1.2181802421335492e-05, + "loss": 0.8367, + "step": 262270 + }, + { + "epoch": 22.7, + "learning_rate": 1.2180357710422146e-05, + "loss": 0.9189, + "step": 262280 + }, + { + "epoch": 22.7, + "learning_rate": 1.2178912999508799e-05, + "loss": 0.9048, + "step": 262290 + }, + { + "epoch": 22.7, + "learning_rate": 1.2177468288595454e-05, + "loss": 0.9793, + "step": 262300 + }, + { + "epoch": 22.7, + "learning_rate": 1.2176023577682107e-05, + "loss": 0.8771, + "step": 262310 + }, + { + "epoch": 22.71, + "learning_rate": 1.2174578866768759e-05, + "loss": 0.909, + "step": 262320 + }, + { + "epoch": 22.71, + "learning_rate": 1.2173134155855413e-05, + "loss": 0.9471, + "step": 262330 + }, + { + "epoch": 22.71, + "learning_rate": 1.2171689444942068e-05, + "loss": 0.9632, + "step": 262340 + }, + { + "epoch": 22.71, + "learning_rate": 1.2170244734028721e-05, + "loss": 0.9402, + "step": 262350 + }, + { + "epoch": 22.71, + "learning_rate": 1.2168800023115374e-05, + "loss": 0.9037, + "step": 262360 + }, + { + "epoch": 22.71, + "learning_rate": 1.216735531220203e-05, + "loss": 0.8864, + "step": 262370 + }, + { + "epoch": 22.71, + "learning_rate": 1.2165910601288683e-05, + "loss": 0.9832, + "step": 262380 + }, + { + "epoch": 22.71, + "learning_rate": 1.2164465890375336e-05, + "loss": 0.9588, + "step": 262390 + }, + { + "epoch": 22.71, + "learning_rate": 1.2163021179461991e-05, + "loss": 0.9511, + "step": 262400 + }, + { + "epoch": 22.71, + "learning_rate": 1.2161576468548643e-05, + "loss": 0.8678, + "step": 262410 + }, + { + "epoch": 22.71, + "learning_rate": 1.2160131757635296e-05, + "loss": 0.979, + "step": 262420 + }, + { + "epoch": 22.72, + "learning_rate": 1.2158687046721952e-05, + "loss": 0.9418, + "step": 262430 + }, + { + "epoch": 22.72, + "learning_rate": 1.2157242335808605e-05, + "loss": 0.9115, + "step": 262440 + }, + { + "epoch": 22.72, + "learning_rate": 1.2155797624895258e-05, + "loss": 0.9082, + "step": 262450 + }, + { + "epoch": 22.72, + "learning_rate": 1.2154352913981913e-05, + "loss": 0.9313, + "step": 262460 + }, + { + "epoch": 22.72, + "learning_rate": 1.2152908203068567e-05, + "loss": 0.8911, + "step": 262470 + }, + { + "epoch": 22.72, + "learning_rate": 1.215146349215522e-05, + "loss": 0.9531, + "step": 262480 + }, + { + "epoch": 22.72, + "learning_rate": 1.2150018781241875e-05, + "loss": 0.8829, + "step": 262490 + }, + { + "epoch": 22.72, + "learning_rate": 1.2148574070328527e-05, + "loss": 0.9086, + "step": 262500 + }, + { + "epoch": 22.72, + "learning_rate": 1.214712935941518e-05, + "loss": 0.9285, + "step": 262510 + }, + { + "epoch": 22.72, + "learning_rate": 1.2145684648501835e-05, + "loss": 0.9443, + "step": 262520 + }, + { + "epoch": 22.72, + "learning_rate": 1.2144239937588489e-05, + "loss": 0.8955, + "step": 262530 + }, + { + "epoch": 22.72, + "learning_rate": 1.2142795226675142e-05, + "loss": 0.955, + "step": 262540 + }, + { + "epoch": 22.73, + "learning_rate": 1.2141350515761797e-05, + "loss": 0.9281, + "step": 262550 + }, + { + "epoch": 22.73, + "learning_rate": 1.213990580484845e-05, + "loss": 0.9882, + "step": 262560 + }, + { + "epoch": 22.73, + "learning_rate": 1.2138461093935104e-05, + "loss": 0.922, + "step": 262570 + }, + { + "epoch": 22.73, + "learning_rate": 1.213701638302176e-05, + "loss": 0.8898, + "step": 262580 + }, + { + "epoch": 22.73, + "learning_rate": 1.2135571672108411e-05, + "loss": 0.9281, + "step": 262590 + }, + { + "epoch": 22.73, + "learning_rate": 1.2134126961195064e-05, + "loss": 0.9876, + "step": 262600 + }, + { + "epoch": 22.73, + "learning_rate": 1.213268225028172e-05, + "loss": 0.8807, + "step": 262610 + }, + { + "epoch": 22.73, + "learning_rate": 1.2131237539368373e-05, + "loss": 0.9293, + "step": 262620 + }, + { + "epoch": 22.73, + "learning_rate": 1.2129792828455026e-05, + "loss": 0.9293, + "step": 262630 + }, + { + "epoch": 22.73, + "learning_rate": 1.2128348117541681e-05, + "loss": 0.9297, + "step": 262640 + }, + { + "epoch": 22.73, + "learning_rate": 1.2126903406628335e-05, + "loss": 0.9395, + "step": 262650 + }, + { + "epoch": 22.74, + "learning_rate": 1.2125458695714988e-05, + "loss": 0.9633, + "step": 262660 + }, + { + "epoch": 22.74, + "learning_rate": 1.2124013984801643e-05, + "loss": 0.9235, + "step": 262670 + }, + { + "epoch": 22.74, + "learning_rate": 1.2122569273888295e-05, + "loss": 0.9568, + "step": 262680 + }, + { + "epoch": 22.74, + "learning_rate": 1.2121124562974948e-05, + "loss": 0.9525, + "step": 262690 + }, + { + "epoch": 22.74, + "learning_rate": 1.2119679852061603e-05, + "loss": 0.9493, + "step": 262700 + }, + { + "epoch": 22.74, + "learning_rate": 1.2118235141148257e-05, + "loss": 0.9578, + "step": 262710 + }, + { + "epoch": 22.74, + "learning_rate": 1.211679043023491e-05, + "loss": 0.9433, + "step": 262720 + }, + { + "epoch": 22.74, + "learning_rate": 1.2115345719321565e-05, + "loss": 0.9648, + "step": 262730 + }, + { + "epoch": 22.74, + "learning_rate": 1.2113901008408219e-05, + "loss": 0.9679, + "step": 262740 + }, + { + "epoch": 22.74, + "learning_rate": 1.2112456297494872e-05, + "loss": 0.8881, + "step": 262750 + }, + { + "epoch": 22.74, + "learning_rate": 1.2111011586581525e-05, + "loss": 0.9493, + "step": 262760 + }, + { + "epoch": 22.74, + "learning_rate": 1.2109566875668179e-05, + "loss": 0.9406, + "step": 262770 + }, + { + "epoch": 22.75, + "learning_rate": 1.2108122164754832e-05, + "loss": 0.9262, + "step": 262780 + }, + { + "epoch": 22.75, + "learning_rate": 1.2106677453841486e-05, + "loss": 0.9237, + "step": 262790 + }, + { + "epoch": 22.75, + "learning_rate": 1.210523274292814e-05, + "loss": 0.904, + "step": 262800 + }, + { + "epoch": 22.75, + "learning_rate": 1.2103788032014794e-05, + "loss": 0.9194, + "step": 262810 + }, + { + "epoch": 22.75, + "learning_rate": 1.2102343321101448e-05, + "loss": 0.9408, + "step": 262820 + }, + { + "epoch": 22.75, + "learning_rate": 1.2100898610188103e-05, + "loss": 0.9331, + "step": 262830 + }, + { + "epoch": 22.75, + "learning_rate": 1.2099453899274756e-05, + "loss": 0.9751, + "step": 262840 + }, + { + "epoch": 22.75, + "learning_rate": 1.209800918836141e-05, + "loss": 0.9716, + "step": 262850 + }, + { + "epoch": 22.75, + "learning_rate": 1.2096564477448063e-05, + "loss": 0.9761, + "step": 262860 + }, + { + "epoch": 22.75, + "learning_rate": 1.2095119766534716e-05, + "loss": 0.8824, + "step": 262870 + }, + { + "epoch": 22.75, + "learning_rate": 1.209367505562137e-05, + "loss": 0.9938, + "step": 262880 + }, + { + "epoch": 22.76, + "learning_rate": 1.2092230344708025e-05, + "loss": 0.8978, + "step": 262890 + }, + { + "epoch": 22.76, + "learning_rate": 1.2090785633794678e-05, + "loss": 0.9052, + "step": 262900 + }, + { + "epoch": 22.76, + "learning_rate": 1.2089340922881331e-05, + "loss": 0.9303, + "step": 262910 + }, + { + "epoch": 22.76, + "learning_rate": 1.2087896211967987e-05, + "loss": 0.9186, + "step": 262920 + }, + { + "epoch": 22.76, + "learning_rate": 1.208645150105464e-05, + "loss": 0.8999, + "step": 262930 + }, + { + "epoch": 22.76, + "learning_rate": 1.2085006790141293e-05, + "loss": 0.9796, + "step": 262940 + }, + { + "epoch": 22.76, + "learning_rate": 1.2083562079227947e-05, + "loss": 0.9238, + "step": 262950 + }, + { + "epoch": 22.76, + "learning_rate": 1.20821173683146e-05, + "loss": 0.9293, + "step": 262960 + }, + { + "epoch": 22.76, + "learning_rate": 1.2080672657401254e-05, + "loss": 0.8854, + "step": 262970 + }, + { + "epoch": 22.76, + "learning_rate": 1.2079227946487909e-05, + "loss": 0.9058, + "step": 262980 + }, + { + "epoch": 22.76, + "learning_rate": 1.2077783235574562e-05, + "loss": 0.9651, + "step": 262990 + }, + { + "epoch": 22.76, + "learning_rate": 1.2076338524661215e-05, + "loss": 0.9701, + "step": 263000 + }, + { + "epoch": 22.77, + "learning_rate": 1.207489381374787e-05, + "loss": 0.9519, + "step": 263010 + }, + { + "epoch": 22.77, + "learning_rate": 1.2073449102834524e-05, + "loss": 0.8538, + "step": 263020 + }, + { + "epoch": 22.77, + "learning_rate": 1.2072004391921177e-05, + "loss": 0.9522, + "step": 263030 + }, + { + "epoch": 22.77, + "learning_rate": 1.207055968100783e-05, + "loss": 0.9263, + "step": 263040 + }, + { + "epoch": 22.77, + "learning_rate": 1.2069114970094484e-05, + "loss": 0.892, + "step": 263050 + }, + { + "epoch": 22.77, + "learning_rate": 1.2067670259181137e-05, + "loss": 0.9467, + "step": 263060 + }, + { + "epoch": 22.77, + "learning_rate": 1.2066225548267793e-05, + "loss": 0.9254, + "step": 263070 + }, + { + "epoch": 22.77, + "learning_rate": 1.2064780837354446e-05, + "loss": 0.9564, + "step": 263080 + }, + { + "epoch": 22.77, + "learning_rate": 1.20633361264411e-05, + "loss": 0.966, + "step": 263090 + }, + { + "epoch": 22.77, + "learning_rate": 1.2061891415527754e-05, + "loss": 0.9899, + "step": 263100 + }, + { + "epoch": 22.77, + "learning_rate": 1.2060446704614408e-05, + "loss": 0.9116, + "step": 263110 + }, + { + "epoch": 22.77, + "learning_rate": 1.2059001993701061e-05, + "loss": 0.9609, + "step": 263120 + }, + { + "epoch": 22.78, + "learning_rate": 1.2057557282787715e-05, + "loss": 0.9691, + "step": 263130 + }, + { + "epoch": 22.78, + "learning_rate": 1.2056112571874368e-05, + "loss": 0.9333, + "step": 263140 + }, + { + "epoch": 22.78, + "learning_rate": 1.2054667860961021e-05, + "loss": 0.9497, + "step": 263150 + }, + { + "epoch": 22.78, + "learning_rate": 1.2053223150047676e-05, + "loss": 0.965, + "step": 263160 + }, + { + "epoch": 22.78, + "learning_rate": 1.205177843913433e-05, + "loss": 0.942, + "step": 263170 + }, + { + "epoch": 22.78, + "learning_rate": 1.2050333728220983e-05, + "loss": 0.9553, + "step": 263180 + }, + { + "epoch": 22.78, + "learning_rate": 1.2048889017307638e-05, + "loss": 0.8804, + "step": 263190 + }, + { + "epoch": 22.78, + "learning_rate": 1.2047444306394292e-05, + "loss": 0.8936, + "step": 263200 + }, + { + "epoch": 22.78, + "learning_rate": 1.2045999595480945e-05, + "loss": 0.8875, + "step": 263210 + }, + { + "epoch": 22.78, + "learning_rate": 1.2044554884567599e-05, + "loss": 0.937, + "step": 263220 + }, + { + "epoch": 22.78, + "learning_rate": 1.2043110173654252e-05, + "loss": 0.9381, + "step": 263230 + }, + { + "epoch": 22.79, + "learning_rate": 1.2041665462740905e-05, + "loss": 0.9178, + "step": 263240 + }, + { + "epoch": 22.79, + "learning_rate": 1.2040220751827559e-05, + "loss": 0.9678, + "step": 263250 + }, + { + "epoch": 22.79, + "learning_rate": 1.2038776040914214e-05, + "loss": 0.9192, + "step": 263260 + }, + { + "epoch": 22.79, + "learning_rate": 1.2037331330000867e-05, + "loss": 0.9517, + "step": 263270 + }, + { + "epoch": 22.79, + "learning_rate": 1.203588661908752e-05, + "loss": 0.9011, + "step": 263280 + }, + { + "epoch": 22.79, + "learning_rate": 1.2034441908174176e-05, + "loss": 0.9496, + "step": 263290 + }, + { + "epoch": 22.79, + "learning_rate": 1.2032997197260829e-05, + "loss": 0.972, + "step": 263300 + }, + { + "epoch": 22.79, + "learning_rate": 1.2031552486347483e-05, + "loss": 0.9333, + "step": 263310 + }, + { + "epoch": 22.79, + "learning_rate": 1.2030107775434136e-05, + "loss": 0.8761, + "step": 263320 + }, + { + "epoch": 22.79, + "learning_rate": 1.202866306452079e-05, + "loss": 0.933, + "step": 263330 + }, + { + "epoch": 22.79, + "learning_rate": 1.2027218353607443e-05, + "loss": 0.8728, + "step": 263340 + }, + { + "epoch": 22.79, + "learning_rate": 1.2025773642694098e-05, + "loss": 0.9255, + "step": 263350 + }, + { + "epoch": 22.8, + "learning_rate": 1.2024328931780751e-05, + "loss": 0.9407, + "step": 263360 + }, + { + "epoch": 22.8, + "learning_rate": 1.2022884220867405e-05, + "loss": 0.8614, + "step": 263370 + }, + { + "epoch": 22.8, + "learning_rate": 1.202143950995406e-05, + "loss": 1.0002, + "step": 263380 + }, + { + "epoch": 22.8, + "learning_rate": 1.2019994799040713e-05, + "loss": 0.9356, + "step": 263390 + }, + { + "epoch": 22.8, + "learning_rate": 1.2018550088127366e-05, + "loss": 0.9447, + "step": 263400 + }, + { + "epoch": 22.8, + "learning_rate": 1.201710537721402e-05, + "loss": 0.9405, + "step": 263410 + }, + { + "epoch": 22.8, + "learning_rate": 1.2015660666300673e-05, + "loss": 0.9212, + "step": 263420 + }, + { + "epoch": 22.8, + "learning_rate": 1.2014215955387327e-05, + "loss": 0.9685, + "step": 263430 + }, + { + "epoch": 22.8, + "learning_rate": 1.2012771244473982e-05, + "loss": 0.9392, + "step": 263440 + }, + { + "epoch": 22.8, + "learning_rate": 1.2011326533560635e-05, + "loss": 0.9256, + "step": 263450 + }, + { + "epoch": 22.8, + "learning_rate": 1.2009881822647289e-05, + "loss": 0.9521, + "step": 263460 + }, + { + "epoch": 22.81, + "learning_rate": 1.2008437111733944e-05, + "loss": 0.9085, + "step": 263470 + }, + { + "epoch": 22.81, + "learning_rate": 1.2006992400820597e-05, + "loss": 0.8912, + "step": 263480 + }, + { + "epoch": 22.81, + "learning_rate": 1.200554768990725e-05, + "loss": 0.9088, + "step": 263490 + }, + { + "epoch": 22.81, + "learning_rate": 1.2004102978993904e-05, + "loss": 0.9281, + "step": 263500 + }, + { + "epoch": 22.81, + "learning_rate": 1.2002658268080557e-05, + "loss": 0.9342, + "step": 263510 + }, + { + "epoch": 22.81, + "learning_rate": 1.200121355716721e-05, + "loss": 0.9177, + "step": 263520 + }, + { + "epoch": 22.81, + "learning_rate": 1.1999768846253866e-05, + "loss": 0.9819, + "step": 263530 + }, + { + "epoch": 22.81, + "learning_rate": 1.1998324135340519e-05, + "loss": 0.9378, + "step": 263540 + }, + { + "epoch": 22.81, + "learning_rate": 1.1996879424427172e-05, + "loss": 0.9158, + "step": 263550 + }, + { + "epoch": 22.81, + "learning_rate": 1.1995434713513828e-05, + "loss": 0.9328, + "step": 263560 + }, + { + "epoch": 22.81, + "learning_rate": 1.1993990002600481e-05, + "loss": 0.9195, + "step": 263570 + }, + { + "epoch": 22.81, + "learning_rate": 1.1992545291687134e-05, + "loss": 0.9255, + "step": 263580 + }, + { + "epoch": 22.82, + "learning_rate": 1.1991100580773788e-05, + "loss": 0.9449, + "step": 263590 + }, + { + "epoch": 22.82, + "learning_rate": 1.1989655869860441e-05, + "loss": 0.9055, + "step": 263600 + }, + { + "epoch": 22.82, + "learning_rate": 1.1988211158947095e-05, + "loss": 0.943, + "step": 263610 + }, + { + "epoch": 22.82, + "learning_rate": 1.198676644803375e-05, + "loss": 0.8578, + "step": 263620 + }, + { + "epoch": 22.82, + "learning_rate": 1.1985321737120403e-05, + "loss": 0.9575, + "step": 263630 + }, + { + "epoch": 22.82, + "learning_rate": 1.1983877026207056e-05, + "loss": 0.9586, + "step": 263640 + }, + { + "epoch": 22.82, + "learning_rate": 1.1982432315293711e-05, + "loss": 0.9394, + "step": 263650 + }, + { + "epoch": 22.82, + "learning_rate": 1.1980987604380365e-05, + "loss": 0.9487, + "step": 263660 + }, + { + "epoch": 22.82, + "learning_rate": 1.1979542893467018e-05, + "loss": 0.8903, + "step": 263670 + }, + { + "epoch": 22.82, + "learning_rate": 1.1978098182553672e-05, + "loss": 0.9122, + "step": 263680 + }, + { + "epoch": 22.82, + "learning_rate": 1.1976653471640325e-05, + "loss": 0.9305, + "step": 263690 + }, + { + "epoch": 22.83, + "learning_rate": 1.1975208760726978e-05, + "loss": 0.9184, + "step": 263700 + }, + { + "epoch": 22.83, + "learning_rate": 1.1973764049813632e-05, + "loss": 0.9124, + "step": 263710 + }, + { + "epoch": 22.83, + "learning_rate": 1.1972319338900287e-05, + "loss": 0.9738, + "step": 263720 + }, + { + "epoch": 22.83, + "learning_rate": 1.197087462798694e-05, + "loss": 0.9251, + "step": 263730 + }, + { + "epoch": 22.83, + "learning_rate": 1.1969429917073594e-05, + "loss": 0.9635, + "step": 263740 + }, + { + "epoch": 22.83, + "learning_rate": 1.1967985206160249e-05, + "loss": 0.9289, + "step": 263750 + }, + { + "epoch": 22.83, + "learning_rate": 1.1966540495246902e-05, + "loss": 0.8919, + "step": 263760 + }, + { + "epoch": 22.83, + "learning_rate": 1.1965095784333556e-05, + "loss": 0.8145, + "step": 263770 + }, + { + "epoch": 22.83, + "learning_rate": 1.1963651073420209e-05, + "loss": 0.9354, + "step": 263780 + }, + { + "epoch": 22.83, + "learning_rate": 1.1962206362506862e-05, + "loss": 0.9503, + "step": 263790 + }, + { + "epoch": 22.83, + "learning_rate": 1.1960761651593516e-05, + "loss": 0.9047, + "step": 263800 + }, + { + "epoch": 22.83, + "learning_rate": 1.1959316940680171e-05, + "loss": 0.9476, + "step": 263810 + }, + { + "epoch": 22.84, + "learning_rate": 1.1957872229766824e-05, + "loss": 0.9378, + "step": 263820 + }, + { + "epoch": 22.84, + "learning_rate": 1.1956427518853478e-05, + "loss": 0.9243, + "step": 263830 + }, + { + "epoch": 22.84, + "learning_rate": 1.1954982807940133e-05, + "loss": 0.9247, + "step": 263840 + }, + { + "epoch": 22.84, + "learning_rate": 1.1953538097026786e-05, + "loss": 0.8671, + "step": 263850 + }, + { + "epoch": 22.84, + "learning_rate": 1.195209338611344e-05, + "loss": 0.8719, + "step": 263860 + }, + { + "epoch": 22.84, + "learning_rate": 1.1950648675200093e-05, + "loss": 0.9225, + "step": 263870 + }, + { + "epoch": 22.84, + "learning_rate": 1.1949203964286746e-05, + "loss": 1.0182, + "step": 263880 + }, + { + "epoch": 22.84, + "learning_rate": 1.19477592533734e-05, + "loss": 0.8768, + "step": 263890 + }, + { + "epoch": 22.84, + "learning_rate": 1.1946314542460055e-05, + "loss": 0.904, + "step": 263900 + }, + { + "epoch": 22.84, + "learning_rate": 1.1944869831546708e-05, + "loss": 0.915, + "step": 263910 + }, + { + "epoch": 22.84, + "learning_rate": 1.1943425120633362e-05, + "loss": 0.94, + "step": 263920 + }, + { + "epoch": 22.85, + "learning_rate": 1.1941980409720017e-05, + "loss": 0.9254, + "step": 263930 + }, + { + "epoch": 22.85, + "learning_rate": 1.194053569880667e-05, + "loss": 0.9016, + "step": 263940 + }, + { + "epoch": 22.85, + "learning_rate": 1.1939090987893323e-05, + "loss": 0.9202, + "step": 263950 + }, + { + "epoch": 22.85, + "learning_rate": 1.1937646276979977e-05, + "loss": 0.945, + "step": 263960 + }, + { + "epoch": 22.85, + "learning_rate": 1.193620156606663e-05, + "loss": 0.978, + "step": 263970 + }, + { + "epoch": 22.85, + "learning_rate": 1.1934756855153284e-05, + "loss": 0.9784, + "step": 263980 + }, + { + "epoch": 22.85, + "learning_rate": 1.1933312144239939e-05, + "loss": 0.9547, + "step": 263990 + }, + { + "epoch": 22.85, + "learning_rate": 1.1931867433326592e-05, + "loss": 0.9533, + "step": 264000 + }, + { + "epoch": 22.85, + "learning_rate": 1.1930422722413246e-05, + "loss": 0.9379, + "step": 264010 + }, + { + "epoch": 22.85, + "learning_rate": 1.19289780114999e-05, + "loss": 0.8766, + "step": 264020 + }, + { + "epoch": 22.85, + "learning_rate": 1.1927533300586554e-05, + "loss": 0.9173, + "step": 264030 + }, + { + "epoch": 22.85, + "learning_rate": 1.1926088589673207e-05, + "loss": 0.9198, + "step": 264040 + }, + { + "epoch": 22.86, + "learning_rate": 1.192464387875986e-05, + "loss": 0.9436, + "step": 264050 + }, + { + "epoch": 22.86, + "learning_rate": 1.1923199167846514e-05, + "loss": 0.9502, + "step": 264060 + }, + { + "epoch": 22.86, + "learning_rate": 1.1921754456933168e-05, + "loss": 0.8935, + "step": 264070 + }, + { + "epoch": 22.86, + "learning_rate": 1.1920309746019823e-05, + "loss": 0.9468, + "step": 264080 + }, + { + "epoch": 22.86, + "learning_rate": 1.1918865035106476e-05, + "loss": 0.9039, + "step": 264090 + }, + { + "epoch": 22.86, + "learning_rate": 1.191742032419313e-05, + "loss": 0.9341, + "step": 264100 + }, + { + "epoch": 22.86, + "learning_rate": 1.1915975613279785e-05, + "loss": 0.9292, + "step": 264110 + }, + { + "epoch": 22.86, + "learning_rate": 1.1914530902366438e-05, + "loss": 0.9177, + "step": 264120 + }, + { + "epoch": 22.86, + "learning_rate": 1.1913086191453091e-05, + "loss": 0.9252, + "step": 264130 + }, + { + "epoch": 22.86, + "learning_rate": 1.1911641480539743e-05, + "loss": 0.9277, + "step": 264140 + }, + { + "epoch": 22.86, + "learning_rate": 1.1910196769626398e-05, + "loss": 0.987, + "step": 264150 + }, + { + "epoch": 22.86, + "learning_rate": 1.1908752058713052e-05, + "loss": 0.9604, + "step": 264160 + }, + { + "epoch": 22.87, + "learning_rate": 1.1907307347799705e-05, + "loss": 0.9744, + "step": 264170 + }, + { + "epoch": 22.87, + "learning_rate": 1.190586263688636e-05, + "loss": 0.8996, + "step": 264180 + }, + { + "epoch": 22.87, + "learning_rate": 1.1904417925973013e-05, + "loss": 0.9929, + "step": 264190 + }, + { + "epoch": 22.87, + "learning_rate": 1.1902973215059667e-05, + "loss": 0.9277, + "step": 264200 + }, + { + "epoch": 22.87, + "learning_rate": 1.1901528504146322e-05, + "loss": 0.9443, + "step": 264210 + }, + { + "epoch": 22.87, + "learning_rate": 1.1900083793232975e-05, + "loss": 0.9606, + "step": 264220 + }, + { + "epoch": 22.87, + "learning_rate": 1.1898639082319627e-05, + "loss": 0.9014, + "step": 264230 + }, + { + "epoch": 22.87, + "learning_rate": 1.1897194371406282e-05, + "loss": 0.8834, + "step": 264240 + }, + { + "epoch": 22.87, + "learning_rate": 1.1895749660492936e-05, + "loss": 0.9499, + "step": 264250 + }, + { + "epoch": 22.87, + "learning_rate": 1.1894304949579589e-05, + "loss": 0.9395, + "step": 264260 + }, + { + "epoch": 22.87, + "learning_rate": 1.1892860238666244e-05, + "loss": 0.9028, + "step": 264270 + }, + { + "epoch": 22.88, + "learning_rate": 1.1891415527752897e-05, + "loss": 0.8816, + "step": 264280 + }, + { + "epoch": 22.88, + "learning_rate": 1.188997081683955e-05, + "loss": 0.8851, + "step": 264290 + }, + { + "epoch": 22.88, + "learning_rate": 1.1888526105926206e-05, + "loss": 0.915, + "step": 264300 + }, + { + "epoch": 22.88, + "learning_rate": 1.188708139501286e-05, + "loss": 0.9614, + "step": 264310 + }, + { + "epoch": 22.88, + "learning_rate": 1.1885636684099511e-05, + "loss": 0.9209, + "step": 264320 + }, + { + "epoch": 22.88, + "learning_rate": 1.1884191973186166e-05, + "loss": 0.9307, + "step": 264330 + }, + { + "epoch": 22.88, + "learning_rate": 1.188274726227282e-05, + "loss": 0.9234, + "step": 264340 + }, + { + "epoch": 22.88, + "learning_rate": 1.1881302551359473e-05, + "loss": 0.9457, + "step": 264350 + }, + { + "epoch": 22.88, + "learning_rate": 1.1879857840446128e-05, + "loss": 0.8775, + "step": 264360 + }, + { + "epoch": 22.88, + "learning_rate": 1.1878413129532781e-05, + "loss": 0.9364, + "step": 264370 + }, + { + "epoch": 22.88, + "learning_rate": 1.1876968418619435e-05, + "loss": 0.91, + "step": 264380 + }, + { + "epoch": 22.88, + "learning_rate": 1.187552370770609e-05, + "loss": 0.9431, + "step": 264390 + }, + { + "epoch": 22.89, + "learning_rate": 1.1874078996792743e-05, + "loss": 0.9065, + "step": 264400 + }, + { + "epoch": 22.89, + "learning_rate": 1.1872634285879395e-05, + "loss": 0.8539, + "step": 264410 + }, + { + "epoch": 22.89, + "learning_rate": 1.187118957496605e-05, + "loss": 0.9825, + "step": 264420 + }, + { + "epoch": 22.89, + "learning_rate": 1.1869744864052703e-05, + "loss": 0.9657, + "step": 264430 + }, + { + "epoch": 22.89, + "learning_rate": 1.1868300153139357e-05, + "loss": 0.8929, + "step": 264440 + }, + { + "epoch": 22.89, + "learning_rate": 1.1866855442226012e-05, + "loss": 0.9573, + "step": 264450 + }, + { + "epoch": 22.89, + "learning_rate": 1.1865410731312665e-05, + "loss": 0.9712, + "step": 264460 + }, + { + "epoch": 22.89, + "learning_rate": 1.1863966020399319e-05, + "loss": 0.9652, + "step": 264470 + }, + { + "epoch": 22.89, + "learning_rate": 1.1862521309485974e-05, + "loss": 0.9767, + "step": 264480 + }, + { + "epoch": 22.89, + "learning_rate": 1.1861076598572627e-05, + "loss": 0.8354, + "step": 264490 + }, + { + "epoch": 22.89, + "learning_rate": 1.1859631887659279e-05, + "loss": 0.9093, + "step": 264500 + }, + { + "epoch": 22.9, + "learning_rate": 1.1858187176745934e-05, + "loss": 0.9426, + "step": 264510 + }, + { + "epoch": 22.9, + "learning_rate": 1.1856742465832587e-05, + "loss": 0.9364, + "step": 264520 + }, + { + "epoch": 22.9, + "learning_rate": 1.185529775491924e-05, + "loss": 0.9659, + "step": 264530 + }, + { + "epoch": 22.9, + "learning_rate": 1.1853853044005896e-05, + "loss": 0.9107, + "step": 264540 + }, + { + "epoch": 22.9, + "learning_rate": 1.185240833309255e-05, + "loss": 0.9488, + "step": 264550 + }, + { + "epoch": 22.9, + "learning_rate": 1.1850963622179203e-05, + "loss": 0.9411, + "step": 264560 + }, + { + "epoch": 22.9, + "learning_rate": 1.1849518911265858e-05, + "loss": 0.9143, + "step": 264570 + }, + { + "epoch": 22.9, + "learning_rate": 1.1848074200352511e-05, + "loss": 0.9283, + "step": 264580 + }, + { + "epoch": 22.9, + "learning_rate": 1.1846629489439163e-05, + "loss": 0.9232, + "step": 264590 + }, + { + "epoch": 22.9, + "learning_rate": 1.1845184778525816e-05, + "loss": 0.9108, + "step": 264600 + }, + { + "epoch": 22.9, + "learning_rate": 1.1843740067612471e-05, + "loss": 0.9877, + "step": 264610 + }, + { + "epoch": 22.9, + "learning_rate": 1.1842295356699125e-05, + "loss": 0.897, + "step": 264620 + }, + { + "epoch": 22.91, + "learning_rate": 1.1840850645785778e-05, + "loss": 0.9196, + "step": 264630 + }, + { + "epoch": 22.91, + "learning_rate": 1.1839405934872433e-05, + "loss": 0.8946, + "step": 264640 + }, + { + "epoch": 22.91, + "learning_rate": 1.1837961223959087e-05, + "loss": 0.9261, + "step": 264650 + }, + { + "epoch": 22.91, + "learning_rate": 1.183651651304574e-05, + "loss": 0.9487, + "step": 264660 + }, + { + "epoch": 22.91, + "learning_rate": 1.1835071802132395e-05, + "loss": 1.0073, + "step": 264670 + }, + { + "epoch": 22.91, + "learning_rate": 1.1833627091219047e-05, + "loss": 1.002, + "step": 264680 + }, + { + "epoch": 22.91, + "learning_rate": 1.18321823803057e-05, + "loss": 0.8749, + "step": 264690 + }, + { + "epoch": 22.91, + "learning_rate": 1.1830737669392355e-05, + "loss": 0.9446, + "step": 264700 + }, + { + "epoch": 22.91, + "learning_rate": 1.1829292958479009e-05, + "loss": 0.9583, + "step": 264710 + }, + { + "epoch": 22.91, + "learning_rate": 1.1827848247565662e-05, + "loss": 0.9437, + "step": 264720 + }, + { + "epoch": 22.91, + "learning_rate": 1.1826403536652317e-05, + "loss": 0.9798, + "step": 264730 + }, + { + "epoch": 22.92, + "learning_rate": 1.182495882573897e-05, + "loss": 0.8586, + "step": 264740 + }, + { + "epoch": 22.92, + "learning_rate": 1.1823514114825624e-05, + "loss": 0.9681, + "step": 264750 + }, + { + "epoch": 22.92, + "learning_rate": 1.1822069403912277e-05, + "loss": 0.9834, + "step": 264760 + }, + { + "epoch": 22.92, + "learning_rate": 1.182062469299893e-05, + "loss": 0.8965, + "step": 264770 + }, + { + "epoch": 22.92, + "learning_rate": 1.1819179982085584e-05, + "loss": 0.8906, + "step": 264780 + }, + { + "epoch": 22.92, + "learning_rate": 1.181773527117224e-05, + "loss": 0.9297, + "step": 264790 + }, + { + "epoch": 22.92, + "learning_rate": 1.1816290560258893e-05, + "loss": 0.9384, + "step": 264800 + }, + { + "epoch": 22.92, + "learning_rate": 1.1814845849345546e-05, + "loss": 0.9033, + "step": 264810 + }, + { + "epoch": 22.92, + "learning_rate": 1.1813401138432201e-05, + "loss": 0.981, + "step": 264820 + }, + { + "epoch": 22.92, + "learning_rate": 1.1811956427518854e-05, + "loss": 0.8601, + "step": 264830 + }, + { + "epoch": 22.92, + "learning_rate": 1.1810511716605508e-05, + "loss": 0.9312, + "step": 264840 + }, + { + "epoch": 22.92, + "learning_rate": 1.1809067005692161e-05, + "loss": 0.8937, + "step": 264850 + }, + { + "epoch": 22.93, + "learning_rate": 1.1807622294778815e-05, + "loss": 0.9722, + "step": 264860 + }, + { + "epoch": 22.93, + "learning_rate": 1.1806177583865468e-05, + "loss": 0.9247, + "step": 264870 + }, + { + "epoch": 22.93, + "learning_rate": 1.1804732872952123e-05, + "loss": 0.9598, + "step": 264880 + }, + { + "epoch": 22.93, + "learning_rate": 1.1803288162038777e-05, + "loss": 0.8783, + "step": 264890 + }, + { + "epoch": 22.93, + "learning_rate": 1.180184345112543e-05, + "loss": 0.9035, + "step": 264900 + }, + { + "epoch": 22.93, + "learning_rate": 1.1800398740212085e-05, + "loss": 0.8854, + "step": 264910 + }, + { + "epoch": 22.93, + "learning_rate": 1.1798954029298738e-05, + "loss": 0.9579, + "step": 264920 + }, + { + "epoch": 22.93, + "learning_rate": 1.1797509318385392e-05, + "loss": 0.9253, + "step": 264930 + }, + { + "epoch": 22.93, + "learning_rate": 1.1796064607472045e-05, + "loss": 0.9621, + "step": 264940 + }, + { + "epoch": 22.93, + "learning_rate": 1.1794619896558699e-05, + "loss": 0.9732, + "step": 264950 + }, + { + "epoch": 22.93, + "learning_rate": 1.1793175185645352e-05, + "loss": 0.9746, + "step": 264960 + }, + { + "epoch": 22.94, + "learning_rate": 1.1791730474732007e-05, + "loss": 0.9668, + "step": 264970 + }, + { + "epoch": 22.94, + "learning_rate": 1.179028576381866e-05, + "loss": 0.8722, + "step": 264980 + }, + { + "epoch": 22.94, + "learning_rate": 1.1788841052905314e-05, + "loss": 0.9015, + "step": 264990 + }, + { + "epoch": 22.94, + "learning_rate": 1.1787396341991969e-05, + "loss": 0.9435, + "step": 265000 + }, + { + "epoch": 22.94, + "learning_rate": 1.1785951631078622e-05, + "loss": 0.9055, + "step": 265010 + }, + { + "epoch": 22.94, + "learning_rate": 1.1784506920165276e-05, + "loss": 0.9451, + "step": 265020 + }, + { + "epoch": 22.94, + "learning_rate": 1.1783062209251929e-05, + "loss": 0.9478, + "step": 265030 + }, + { + "epoch": 22.94, + "learning_rate": 1.1781617498338583e-05, + "loss": 0.914, + "step": 265040 + }, + { + "epoch": 22.94, + "learning_rate": 1.1780172787425236e-05, + "loss": 0.8897, + "step": 265050 + }, + { + "epoch": 22.94, + "learning_rate": 1.177872807651189e-05, + "loss": 0.9332, + "step": 265060 + }, + { + "epoch": 22.94, + "learning_rate": 1.1777283365598544e-05, + "loss": 0.9553, + "step": 265070 + }, + { + "epoch": 22.94, + "learning_rate": 1.1775838654685198e-05, + "loss": 0.9317, + "step": 265080 + }, + { + "epoch": 22.95, + "learning_rate": 1.1774393943771851e-05, + "loss": 0.982, + "step": 265090 + }, + { + "epoch": 22.95, + "learning_rate": 1.1772949232858506e-05, + "loss": 0.9335, + "step": 265100 + }, + { + "epoch": 22.95, + "learning_rate": 1.177150452194516e-05, + "loss": 0.9474, + "step": 265110 + }, + { + "epoch": 22.95, + "learning_rate": 1.1770059811031813e-05, + "loss": 0.927, + "step": 265120 + }, + { + "epoch": 22.95, + "learning_rate": 1.1768615100118466e-05, + "loss": 0.9414, + "step": 265130 + }, + { + "epoch": 22.95, + "learning_rate": 1.176717038920512e-05, + "loss": 0.9297, + "step": 265140 + }, + { + "epoch": 22.95, + "learning_rate": 1.1765725678291773e-05, + "loss": 0.9487, + "step": 265150 + }, + { + "epoch": 22.95, + "learning_rate": 1.1764280967378428e-05, + "loss": 0.9526, + "step": 265160 + }, + { + "epoch": 22.95, + "learning_rate": 1.1762836256465082e-05, + "loss": 0.9523, + "step": 265170 + }, + { + "epoch": 22.95, + "learning_rate": 1.1761391545551735e-05, + "loss": 0.9476, + "step": 265180 + }, + { + "epoch": 22.95, + "learning_rate": 1.175994683463839e-05, + "loss": 0.9099, + "step": 265190 + }, + { + "epoch": 22.95, + "learning_rate": 1.1758502123725044e-05, + "loss": 0.9041, + "step": 265200 + }, + { + "epoch": 22.96, + "learning_rate": 1.1757057412811697e-05, + "loss": 0.8942, + "step": 265210 + }, + { + "epoch": 22.96, + "learning_rate": 1.175561270189835e-05, + "loss": 0.8957, + "step": 265220 + }, + { + "epoch": 22.96, + "learning_rate": 1.1754167990985004e-05, + "loss": 0.9993, + "step": 265230 + }, + { + "epoch": 22.96, + "learning_rate": 1.1752723280071657e-05, + "loss": 0.886, + "step": 265240 + }, + { + "epoch": 22.96, + "learning_rate": 1.1751278569158312e-05, + "loss": 0.9428, + "step": 265250 + }, + { + "epoch": 22.96, + "learning_rate": 1.1749833858244966e-05, + "loss": 0.9336, + "step": 265260 + }, + { + "epoch": 22.96, + "learning_rate": 1.1748389147331619e-05, + "loss": 0.9583, + "step": 265270 + }, + { + "epoch": 22.96, + "learning_rate": 1.1746944436418274e-05, + "loss": 0.9597, + "step": 265280 + }, + { + "epoch": 22.96, + "learning_rate": 1.1745499725504928e-05, + "loss": 0.9499, + "step": 265290 + }, + { + "epoch": 22.96, + "learning_rate": 1.1744055014591581e-05, + "loss": 0.9168, + "step": 265300 + }, + { + "epoch": 22.96, + "learning_rate": 1.1742610303678234e-05, + "loss": 0.9589, + "step": 265310 + }, + { + "epoch": 22.97, + "learning_rate": 1.1741165592764888e-05, + "loss": 0.9566, + "step": 265320 + }, + { + "epoch": 22.97, + "learning_rate": 1.1739720881851541e-05, + "loss": 0.9079, + "step": 265330 + }, + { + "epoch": 22.97, + "learning_rate": 1.1738276170938196e-05, + "loss": 0.9448, + "step": 265340 + }, + { + "epoch": 22.97, + "learning_rate": 1.173683146002485e-05, + "loss": 0.9048, + "step": 265350 + }, + { + "epoch": 22.97, + "learning_rate": 1.1735386749111503e-05, + "loss": 0.8994, + "step": 265360 + }, + { + "epoch": 22.97, + "learning_rate": 1.1733942038198158e-05, + "loss": 0.9719, + "step": 265370 + }, + { + "epoch": 22.97, + "learning_rate": 1.1732497327284811e-05, + "loss": 0.8478, + "step": 265380 + }, + { + "epoch": 22.97, + "learning_rate": 1.1731052616371465e-05, + "loss": 0.9333, + "step": 265390 + }, + { + "epoch": 22.97, + "learning_rate": 1.1729607905458118e-05, + "loss": 0.9317, + "step": 265400 + }, + { + "epoch": 22.97, + "learning_rate": 1.1728163194544772e-05, + "loss": 0.9736, + "step": 265410 + }, + { + "epoch": 22.97, + "learning_rate": 1.1726718483631425e-05, + "loss": 0.896, + "step": 265420 + }, + { + "epoch": 22.97, + "learning_rate": 1.172527377271808e-05, + "loss": 0.888, + "step": 265430 + }, + { + "epoch": 22.98, + "learning_rate": 1.1723829061804734e-05, + "loss": 0.9764, + "step": 265440 + }, + { + "epoch": 22.98, + "learning_rate": 1.1722384350891387e-05, + "loss": 0.959, + "step": 265450 + }, + { + "epoch": 22.98, + "learning_rate": 1.1720939639978042e-05, + "loss": 0.8869, + "step": 265460 + }, + { + "epoch": 22.98, + "learning_rate": 1.1719494929064695e-05, + "loss": 0.8935, + "step": 265470 + }, + { + "epoch": 22.98, + "learning_rate": 1.1718050218151349e-05, + "loss": 0.8888, + "step": 265480 + }, + { + "epoch": 22.98, + "learning_rate": 1.1716605507238002e-05, + "loss": 0.9473, + "step": 265490 + }, + { + "epoch": 22.98, + "learning_rate": 1.1715160796324656e-05, + "loss": 0.9033, + "step": 265500 + }, + { + "epoch": 22.98, + "learning_rate": 1.1713716085411309e-05, + "loss": 0.9771, + "step": 265510 + }, + { + "epoch": 22.98, + "learning_rate": 1.1712271374497962e-05, + "loss": 0.9327, + "step": 265520 + }, + { + "epoch": 22.98, + "learning_rate": 1.1710826663584617e-05, + "loss": 0.9422, + "step": 265530 + }, + { + "epoch": 22.98, + "learning_rate": 1.1709381952671271e-05, + "loss": 0.9121, + "step": 265540 + }, + { + "epoch": 22.99, + "learning_rate": 1.1707937241757924e-05, + "loss": 0.9322, + "step": 265550 + }, + { + "epoch": 22.99, + "learning_rate": 1.170649253084458e-05, + "loss": 0.9295, + "step": 265560 + }, + { + "epoch": 22.99, + "learning_rate": 1.1705047819931233e-05, + "loss": 0.946, + "step": 265570 + }, + { + "epoch": 22.99, + "learning_rate": 1.1703603109017886e-05, + "loss": 0.9403, + "step": 265580 + }, + { + "epoch": 22.99, + "learning_rate": 1.170215839810454e-05, + "loss": 0.9404, + "step": 265590 + }, + { + "epoch": 22.99, + "learning_rate": 1.1700713687191193e-05, + "loss": 0.9889, + "step": 265600 + }, + { + "epoch": 22.99, + "learning_rate": 1.1699268976277846e-05, + "loss": 0.909, + "step": 265610 + }, + { + "epoch": 22.99, + "learning_rate": 1.1697824265364501e-05, + "loss": 0.9387, + "step": 265620 + }, + { + "epoch": 22.99, + "learning_rate": 1.1696379554451155e-05, + "loss": 0.9432, + "step": 265630 + }, + { + "epoch": 22.99, + "learning_rate": 1.1694934843537808e-05, + "loss": 0.9374, + "step": 265640 + }, + { + "epoch": 22.99, + "learning_rate": 1.1693490132624463e-05, + "loss": 0.9416, + "step": 265650 + }, + { + "epoch": 22.99, + "learning_rate": 1.1692045421711117e-05, + "loss": 0.8589, + "step": 265660 + }, + { + "epoch": 23.0, + "learning_rate": 1.169060071079777e-05, + "loss": 0.9434, + "step": 265670 + }, + { + "epoch": 23.0, + "learning_rate": 1.1689155999884424e-05, + "loss": 0.926, + "step": 265680 + }, + { + "epoch": 23.0, + "learning_rate": 1.1687711288971077e-05, + "loss": 0.9882, + "step": 265690 + }, + { + "epoch": 23.0, + "learning_rate": 1.168626657805773e-05, + "loss": 0.9292, + "step": 265700 + }, + { + "epoch": 23.0, + "learning_rate": 1.1684821867144385e-05, + "loss": 0.9085, + "step": 265710 + }, + { + "epoch": 23.0, + "learning_rate": 1.1683377156231039e-05, + "loss": 0.9411, + "step": 265720 + }, + { + "epoch": 23.0, + "eval_cer": 0.9931626177616802, + "eval_em": 0, + "eval_f1": 0.1187918125427299, + "eval_loss": 0.9417679905891418, + "eval_runtime": 1272.4449, + "eval_samples_per_second": 4.036, + "eval_steps_per_second": 0.505, + "eval_wer": 0.9666931146652686, + "step": 265720 + }, + { + "epoch": 23.0, + "learning_rate": 1.1681932445317692e-05, + "loss": 0.9044, + "step": 265730 + }, + { + "epoch": 23.0, + "learning_rate": 1.1680487734404347e-05, + "loss": 0.9374, + "step": 265740 + }, + { + "epoch": 23.0, + "learning_rate": 1.1679043023491e-05, + "loss": 0.9031, + "step": 265750 + }, + { + "epoch": 23.0, + "learning_rate": 1.1677598312577654e-05, + "loss": 0.906, + "step": 265760 + }, + { + "epoch": 23.0, + "learning_rate": 1.1676153601664307e-05, + "loss": 0.9102, + "step": 265770 + }, + { + "epoch": 23.01, + "learning_rate": 1.167470889075096e-05, + "loss": 0.9192, + "step": 265780 + }, + { + "epoch": 23.01, + "learning_rate": 1.1673264179837614e-05, + "loss": 0.8772, + "step": 265790 + }, + { + "epoch": 23.01, + "learning_rate": 1.167181946892427e-05, + "loss": 0.9181, + "step": 265800 + }, + { + "epoch": 23.01, + "learning_rate": 1.1670374758010923e-05, + "loss": 0.949, + "step": 265810 + }, + { + "epoch": 23.01, + "learning_rate": 1.1668930047097576e-05, + "loss": 0.8814, + "step": 265820 + }, + { + "epoch": 23.01, + "learning_rate": 1.1667485336184231e-05, + "loss": 0.978, + "step": 265830 + }, + { + "epoch": 23.01, + "learning_rate": 1.1666040625270885e-05, + "loss": 0.9025, + "step": 265840 + }, + { + "epoch": 23.01, + "learning_rate": 1.1664595914357538e-05, + "loss": 0.96, + "step": 265850 + }, + { + "epoch": 23.01, + "learning_rate": 1.1663151203444191e-05, + "loss": 0.9116, + "step": 265860 + }, + { + "epoch": 23.01, + "learning_rate": 1.1661706492530845e-05, + "loss": 0.9423, + "step": 265870 + }, + { + "epoch": 23.01, + "learning_rate": 1.1660261781617498e-05, + "loss": 0.8861, + "step": 265880 + }, + { + "epoch": 23.01, + "learning_rate": 1.1658817070704153e-05, + "loss": 0.9231, + "step": 265890 + }, + { + "epoch": 23.02, + "learning_rate": 1.1657372359790807e-05, + "loss": 0.8253, + "step": 265900 + }, + { + "epoch": 23.02, + "learning_rate": 1.165592764887746e-05, + "loss": 0.9338, + "step": 265910 + }, + { + "epoch": 23.02, + "learning_rate": 1.1654482937964115e-05, + "loss": 0.9661, + "step": 265920 + }, + { + "epoch": 23.02, + "learning_rate": 1.1653038227050769e-05, + "loss": 0.8782, + "step": 265930 + }, + { + "epoch": 23.02, + "learning_rate": 1.1651593516137422e-05, + "loss": 0.8851, + "step": 265940 + }, + { + "epoch": 23.02, + "learning_rate": 1.1650148805224075e-05, + "loss": 0.9184, + "step": 265950 + }, + { + "epoch": 23.02, + "learning_rate": 1.1648704094310729e-05, + "loss": 0.9416, + "step": 265960 + }, + { + "epoch": 23.02, + "learning_rate": 1.1647259383397382e-05, + "loss": 0.9451, + "step": 265970 + }, + { + "epoch": 23.02, + "learning_rate": 1.1645814672484036e-05, + "loss": 0.9016, + "step": 265980 + }, + { + "epoch": 23.02, + "learning_rate": 1.164436996157069e-05, + "loss": 0.9407, + "step": 265990 + }, + { + "epoch": 23.02, + "learning_rate": 1.1642925250657344e-05, + "loss": 0.9465, + "step": 266000 + }, + { + "epoch": 23.03, + "learning_rate": 1.1641480539743997e-05, + "loss": 0.9346, + "step": 266010 + }, + { + "epoch": 23.03, + "learning_rate": 1.1640035828830652e-05, + "loss": 0.9141, + "step": 266020 + }, + { + "epoch": 23.03, + "learning_rate": 1.1638591117917306e-05, + "loss": 0.9338, + "step": 266030 + }, + { + "epoch": 23.03, + "learning_rate": 1.163714640700396e-05, + "loss": 0.9732, + "step": 266040 + }, + { + "epoch": 23.03, + "learning_rate": 1.1635701696090613e-05, + "loss": 0.8967, + "step": 266050 + }, + { + "epoch": 23.03, + "learning_rate": 1.1634256985177266e-05, + "loss": 0.9396, + "step": 266060 + }, + { + "epoch": 23.03, + "learning_rate": 1.163281227426392e-05, + "loss": 0.9561, + "step": 266070 + }, + { + "epoch": 23.03, + "learning_rate": 1.1631367563350575e-05, + "loss": 0.9307, + "step": 266080 + }, + { + "epoch": 23.03, + "learning_rate": 1.1629922852437228e-05, + "loss": 0.8557, + "step": 266090 + }, + { + "epoch": 23.03, + "learning_rate": 1.1628478141523881e-05, + "loss": 0.9371, + "step": 266100 + }, + { + "epoch": 23.03, + "learning_rate": 1.1627033430610536e-05, + "loss": 0.9056, + "step": 266110 + }, + { + "epoch": 23.03, + "learning_rate": 1.162558871969719e-05, + "loss": 0.8544, + "step": 266120 + }, + { + "epoch": 23.04, + "learning_rate": 1.1624144008783843e-05, + "loss": 0.9142, + "step": 266130 + }, + { + "epoch": 23.04, + "learning_rate": 1.1622699297870497e-05, + "loss": 0.9371, + "step": 266140 + }, + { + "epoch": 23.04, + "learning_rate": 1.162125458695715e-05, + "loss": 0.9722, + "step": 266150 + }, + { + "epoch": 23.04, + "learning_rate": 1.1619809876043803e-05, + "loss": 0.9502, + "step": 266160 + }, + { + "epoch": 23.04, + "learning_rate": 1.1618365165130458e-05, + "loss": 0.8865, + "step": 266170 + }, + { + "epoch": 23.04, + "learning_rate": 1.1616920454217112e-05, + "loss": 0.8759, + "step": 266180 + }, + { + "epoch": 23.04, + "learning_rate": 1.1615475743303765e-05, + "loss": 0.9446, + "step": 266190 + }, + { + "epoch": 23.04, + "learning_rate": 1.161403103239042e-05, + "loss": 0.889, + "step": 266200 + }, + { + "epoch": 23.04, + "learning_rate": 1.1612586321477074e-05, + "loss": 0.9408, + "step": 266210 + }, + { + "epoch": 23.04, + "learning_rate": 1.1611141610563727e-05, + "loss": 0.9199, + "step": 266220 + }, + { + "epoch": 23.04, + "learning_rate": 1.160969689965038e-05, + "loss": 0.9828, + "step": 266230 + }, + { + "epoch": 23.05, + "learning_rate": 1.1608252188737034e-05, + "loss": 0.9453, + "step": 266240 + }, + { + "epoch": 23.05, + "learning_rate": 1.1606807477823687e-05, + "loss": 0.9027, + "step": 266250 + }, + { + "epoch": 23.05, + "learning_rate": 1.1605362766910342e-05, + "loss": 0.9347, + "step": 266260 + }, + { + "epoch": 23.05, + "learning_rate": 1.1603918055996996e-05, + "loss": 0.9019, + "step": 266270 + }, + { + "epoch": 23.05, + "learning_rate": 1.160247334508365e-05, + "loss": 0.9161, + "step": 266280 + }, + { + "epoch": 23.05, + "learning_rate": 1.1601028634170304e-05, + "loss": 0.9418, + "step": 266290 + }, + { + "epoch": 23.05, + "learning_rate": 1.1599583923256958e-05, + "loss": 0.9295, + "step": 266300 + }, + { + "epoch": 23.05, + "learning_rate": 1.1598139212343611e-05, + "loss": 0.9122, + "step": 266310 + }, + { + "epoch": 23.05, + "learning_rate": 1.1596694501430264e-05, + "loss": 0.9415, + "step": 266320 + }, + { + "epoch": 23.05, + "learning_rate": 1.1595249790516918e-05, + "loss": 0.9503, + "step": 266330 + }, + { + "epoch": 23.05, + "learning_rate": 1.1593805079603571e-05, + "loss": 0.9583, + "step": 266340 + }, + { + "epoch": 23.05, + "learning_rate": 1.1592360368690226e-05, + "loss": 0.9269, + "step": 266350 + }, + { + "epoch": 23.06, + "learning_rate": 1.159091565777688e-05, + "loss": 0.8963, + "step": 266360 + }, + { + "epoch": 23.06, + "learning_rate": 1.1589470946863533e-05, + "loss": 0.995, + "step": 266370 + }, + { + "epoch": 23.06, + "learning_rate": 1.1588026235950188e-05, + "loss": 0.9409, + "step": 266380 + }, + { + "epoch": 23.06, + "learning_rate": 1.1586581525036842e-05, + "loss": 0.9952, + "step": 266390 + }, + { + "epoch": 23.06, + "learning_rate": 1.1585136814123495e-05, + "loss": 0.9269, + "step": 266400 + }, + { + "epoch": 23.06, + "learning_rate": 1.1583692103210147e-05, + "loss": 0.9605, + "step": 266410 + }, + { + "epoch": 23.06, + "learning_rate": 1.1582247392296802e-05, + "loss": 0.9282, + "step": 266420 + }, + { + "epoch": 23.06, + "learning_rate": 1.1580802681383455e-05, + "loss": 0.9497, + "step": 266430 + }, + { + "epoch": 23.06, + "learning_rate": 1.1579357970470109e-05, + "loss": 0.9616, + "step": 266440 + }, + { + "epoch": 23.06, + "learning_rate": 1.1577913259556764e-05, + "loss": 0.9359, + "step": 266450 + }, + { + "epoch": 23.06, + "learning_rate": 1.1576468548643417e-05, + "loss": 0.9329, + "step": 266460 + }, + { + "epoch": 23.07, + "learning_rate": 1.157502383773007e-05, + "loss": 0.86, + "step": 266470 + }, + { + "epoch": 23.07, + "learning_rate": 1.1573579126816726e-05, + "loss": 0.9188, + "step": 266480 + }, + { + "epoch": 23.07, + "learning_rate": 1.1572134415903379e-05, + "loss": 0.907, + "step": 266490 + }, + { + "epoch": 23.07, + "learning_rate": 1.157068970499003e-05, + "loss": 0.9242, + "step": 266500 + }, + { + "epoch": 23.07, + "learning_rate": 1.1569244994076686e-05, + "loss": 0.8748, + "step": 266510 + }, + { + "epoch": 23.07, + "learning_rate": 1.156780028316334e-05, + "loss": 0.8476, + "step": 266520 + }, + { + "epoch": 23.07, + "learning_rate": 1.1566355572249993e-05, + "loss": 0.9055, + "step": 266530 + }, + { + "epoch": 23.07, + "learning_rate": 1.1564910861336648e-05, + "loss": 0.8761, + "step": 266540 + }, + { + "epoch": 23.07, + "learning_rate": 1.1563466150423301e-05, + "loss": 0.9435, + "step": 266550 + }, + { + "epoch": 23.07, + "learning_rate": 1.1562021439509954e-05, + "loss": 0.9405, + "step": 266560 + }, + { + "epoch": 23.07, + "learning_rate": 1.156057672859661e-05, + "loss": 0.9375, + "step": 266570 + }, + { + "epoch": 23.07, + "learning_rate": 1.1559132017683263e-05, + "loss": 0.8965, + "step": 266580 + }, + { + "epoch": 23.08, + "learning_rate": 1.1557687306769915e-05, + "loss": 0.9178, + "step": 266590 + }, + { + "epoch": 23.08, + "learning_rate": 1.155624259585657e-05, + "loss": 0.9227, + "step": 266600 + }, + { + "epoch": 23.08, + "learning_rate": 1.1554797884943223e-05, + "loss": 0.9649, + "step": 266610 + }, + { + "epoch": 23.08, + "learning_rate": 1.1553353174029877e-05, + "loss": 0.9419, + "step": 266620 + }, + { + "epoch": 23.08, + "learning_rate": 1.1551908463116532e-05, + "loss": 0.9367, + "step": 266630 + }, + { + "epoch": 23.08, + "learning_rate": 1.1550463752203185e-05, + "loss": 0.9311, + "step": 266640 + }, + { + "epoch": 23.08, + "learning_rate": 1.1549019041289838e-05, + "loss": 0.9164, + "step": 266650 + }, + { + "epoch": 23.08, + "learning_rate": 1.1547574330376493e-05, + "loss": 0.9404, + "step": 266660 + }, + { + "epoch": 23.08, + "learning_rate": 1.1546129619463147e-05, + "loss": 0.9146, + "step": 266670 + }, + { + "epoch": 23.08, + "learning_rate": 1.1544684908549799e-05, + "loss": 0.9413, + "step": 266680 + }, + { + "epoch": 23.08, + "learning_rate": 1.1543240197636454e-05, + "loss": 0.9614, + "step": 266690 + }, + { + "epoch": 23.08, + "learning_rate": 1.1541795486723107e-05, + "loss": 0.9214, + "step": 266700 + }, + { + "epoch": 23.09, + "learning_rate": 1.154035077580976e-05, + "loss": 0.8938, + "step": 266710 + }, + { + "epoch": 23.09, + "learning_rate": 1.1538906064896416e-05, + "loss": 0.9779, + "step": 266720 + }, + { + "epoch": 23.09, + "learning_rate": 1.1537461353983069e-05, + "loss": 0.9534, + "step": 266730 + }, + { + "epoch": 23.09, + "learning_rate": 1.1536016643069722e-05, + "loss": 0.8869, + "step": 266740 + }, + { + "epoch": 23.09, + "learning_rate": 1.1534571932156377e-05, + "loss": 0.927, + "step": 266750 + }, + { + "epoch": 23.09, + "learning_rate": 1.1533127221243029e-05, + "loss": 0.9211, + "step": 266760 + }, + { + "epoch": 23.09, + "learning_rate": 1.1531682510329683e-05, + "loss": 0.9217, + "step": 266770 + }, + { + "epoch": 23.09, + "learning_rate": 1.1530237799416338e-05, + "loss": 0.9079, + "step": 266780 + }, + { + "epoch": 23.09, + "learning_rate": 1.1528793088502991e-05, + "loss": 0.9215, + "step": 266790 + }, + { + "epoch": 23.09, + "learning_rate": 1.1527348377589644e-05, + "loss": 0.9141, + "step": 266800 + }, + { + "epoch": 23.09, + "learning_rate": 1.15259036666763e-05, + "loss": 0.9611, + "step": 266810 + }, + { + "epoch": 23.1, + "learning_rate": 1.1524458955762953e-05, + "loss": 0.9225, + "step": 266820 + }, + { + "epoch": 23.1, + "learning_rate": 1.1523014244849606e-05, + "loss": 0.9222, + "step": 266830 + }, + { + "epoch": 23.1, + "learning_rate": 1.1521569533936261e-05, + "loss": 0.9552, + "step": 266840 + }, + { + "epoch": 23.1, + "learning_rate": 1.1520124823022913e-05, + "loss": 0.9036, + "step": 266850 + }, + { + "epoch": 23.1, + "learning_rate": 1.1518680112109566e-05, + "loss": 0.9064, + "step": 266860 + }, + { + "epoch": 23.1, + "learning_rate": 1.151723540119622e-05, + "loss": 0.9916, + "step": 266870 + }, + { + "epoch": 23.1, + "learning_rate": 1.1515790690282875e-05, + "loss": 0.919, + "step": 266880 + }, + { + "epoch": 23.1, + "learning_rate": 1.1514345979369528e-05, + "loss": 0.9657, + "step": 266890 + }, + { + "epoch": 23.1, + "learning_rate": 1.1512901268456182e-05, + "loss": 0.9165, + "step": 266900 + }, + { + "epoch": 23.1, + "learning_rate": 1.1511456557542837e-05, + "loss": 0.9707, + "step": 266910 + }, + { + "epoch": 23.1, + "learning_rate": 1.151001184662949e-05, + "loss": 0.9405, + "step": 266920 + }, + { + "epoch": 23.1, + "learning_rate": 1.1508567135716144e-05, + "loss": 0.9388, + "step": 266930 + }, + { + "epoch": 23.11, + "learning_rate": 1.1507122424802797e-05, + "loss": 0.9548, + "step": 266940 + }, + { + "epoch": 23.11, + "learning_rate": 1.150567771388945e-05, + "loss": 0.8551, + "step": 266950 + }, + { + "epoch": 23.11, + "learning_rate": 1.1504233002976104e-05, + "loss": 0.8595, + "step": 266960 + }, + { + "epoch": 23.11, + "learning_rate": 1.1502788292062759e-05, + "loss": 0.9244, + "step": 266970 + }, + { + "epoch": 23.11, + "learning_rate": 1.1501343581149412e-05, + "loss": 0.9732, + "step": 266980 + }, + { + "epoch": 23.11, + "learning_rate": 1.1499898870236066e-05, + "loss": 0.9244, + "step": 266990 + }, + { + "epoch": 23.11, + "learning_rate": 1.149845415932272e-05, + "loss": 0.9441, + "step": 267000 + }, + { + "epoch": 23.11, + "learning_rate": 1.1497009448409374e-05, + "loss": 0.9291, + "step": 267010 + }, + { + "epoch": 23.11, + "learning_rate": 1.1495564737496028e-05, + "loss": 0.9487, + "step": 267020 + }, + { + "epoch": 23.11, + "learning_rate": 1.1494120026582681e-05, + "loss": 0.9116, + "step": 267030 + }, + { + "epoch": 23.11, + "learning_rate": 1.1492675315669334e-05, + "loss": 0.9303, + "step": 267040 + }, + { + "epoch": 23.12, + "learning_rate": 1.1491230604755988e-05, + "loss": 0.9431, + "step": 267050 + }, + { + "epoch": 23.12, + "learning_rate": 1.1489785893842643e-05, + "loss": 1.0059, + "step": 267060 + }, + { + "epoch": 23.12, + "learning_rate": 1.1488341182929296e-05, + "loss": 0.9424, + "step": 267070 + }, + { + "epoch": 23.12, + "learning_rate": 1.148689647201595e-05, + "loss": 0.8912, + "step": 267080 + }, + { + "epoch": 23.12, + "learning_rate": 1.1485451761102605e-05, + "loss": 0.9052, + "step": 267090 + }, + { + "epoch": 23.12, + "learning_rate": 1.1484007050189258e-05, + "loss": 0.9068, + "step": 267100 + }, + { + "epoch": 23.12, + "learning_rate": 1.1482562339275912e-05, + "loss": 0.8779, + "step": 267110 + }, + { + "epoch": 23.12, + "learning_rate": 1.1481117628362565e-05, + "loss": 0.9284, + "step": 267120 + }, + { + "epoch": 23.12, + "learning_rate": 1.1479672917449218e-05, + "loss": 0.8721, + "step": 267130 + }, + { + "epoch": 23.12, + "learning_rate": 1.1478228206535872e-05, + "loss": 0.9104, + "step": 267140 + }, + { + "epoch": 23.12, + "learning_rate": 1.1476783495622527e-05, + "loss": 0.9164, + "step": 267150 + }, + { + "epoch": 23.12, + "learning_rate": 1.147533878470918e-05, + "loss": 0.905, + "step": 267160 + }, + { + "epoch": 23.13, + "learning_rate": 1.1473894073795834e-05, + "loss": 0.8926, + "step": 267170 + }, + { + "epoch": 23.13, + "learning_rate": 1.1472449362882489e-05, + "loss": 0.9372, + "step": 267180 + }, + { + "epoch": 23.13, + "learning_rate": 1.1471004651969142e-05, + "loss": 0.9301, + "step": 267190 + }, + { + "epoch": 23.13, + "learning_rate": 1.1469559941055795e-05, + "loss": 0.8582, + "step": 267200 + }, + { + "epoch": 23.13, + "learning_rate": 1.1468115230142449e-05, + "loss": 0.9223, + "step": 267210 + }, + { + "epoch": 23.13, + "learning_rate": 1.1466670519229102e-05, + "loss": 0.9525, + "step": 267220 + }, + { + "epoch": 23.13, + "learning_rate": 1.1465225808315756e-05, + "loss": 0.9473, + "step": 267230 + }, + { + "epoch": 23.13, + "learning_rate": 1.146378109740241e-05, + "loss": 0.9631, + "step": 267240 + }, + { + "epoch": 23.13, + "learning_rate": 1.1462336386489064e-05, + "loss": 0.9705, + "step": 267250 + }, + { + "epoch": 23.13, + "learning_rate": 1.1460891675575718e-05, + "loss": 0.9718, + "step": 267260 + }, + { + "epoch": 23.13, + "learning_rate": 1.1459446964662373e-05, + "loss": 0.9112, + "step": 267270 + }, + { + "epoch": 23.14, + "learning_rate": 1.1458002253749026e-05, + "loss": 0.9348, + "step": 267280 + }, + { + "epoch": 23.14, + "learning_rate": 1.145655754283568e-05, + "loss": 0.9747, + "step": 267290 + }, + { + "epoch": 23.14, + "learning_rate": 1.1455112831922333e-05, + "loss": 0.9306, + "step": 267300 + }, + { + "epoch": 23.14, + "learning_rate": 1.1453668121008986e-05, + "loss": 0.8761, + "step": 267310 + }, + { + "epoch": 23.14, + "learning_rate": 1.145222341009564e-05, + "loss": 0.884, + "step": 267320 + }, + { + "epoch": 23.14, + "learning_rate": 1.1450778699182293e-05, + "loss": 0.9467, + "step": 267330 + }, + { + "epoch": 23.14, + "learning_rate": 1.1449333988268948e-05, + "loss": 0.9215, + "step": 267340 + }, + { + "epoch": 23.14, + "learning_rate": 1.1447889277355601e-05, + "loss": 0.9157, + "step": 267350 + }, + { + "epoch": 23.14, + "learning_rate": 1.1446444566442255e-05, + "loss": 0.947, + "step": 267360 + }, + { + "epoch": 23.14, + "learning_rate": 1.144499985552891e-05, + "loss": 0.872, + "step": 267370 + }, + { + "epoch": 23.14, + "learning_rate": 1.1443555144615563e-05, + "loss": 0.9505, + "step": 267380 + }, + { + "epoch": 23.14, + "learning_rate": 1.1442110433702217e-05, + "loss": 0.9446, + "step": 267390 + }, + { + "epoch": 23.15, + "learning_rate": 1.144066572278887e-05, + "loss": 0.8806, + "step": 267400 + }, + { + "epoch": 23.15, + "learning_rate": 1.1439221011875524e-05, + "loss": 0.9131, + "step": 267410 + }, + { + "epoch": 23.15, + "learning_rate": 1.1437776300962177e-05, + "loss": 0.9139, + "step": 267420 + }, + { + "epoch": 23.15, + "learning_rate": 1.1436331590048832e-05, + "loss": 0.9136, + "step": 267430 + }, + { + "epoch": 23.15, + "learning_rate": 1.1434886879135485e-05, + "loss": 0.9251, + "step": 267440 + }, + { + "epoch": 23.15, + "learning_rate": 1.1433442168222139e-05, + "loss": 0.8975, + "step": 267450 + }, + { + "epoch": 23.15, + "learning_rate": 1.1431997457308794e-05, + "loss": 0.9094, + "step": 267460 + }, + { + "epoch": 23.15, + "learning_rate": 1.1430552746395447e-05, + "loss": 0.8844, + "step": 267470 + }, + { + "epoch": 23.15, + "learning_rate": 1.14291080354821e-05, + "loss": 0.9459, + "step": 267480 + }, + { + "epoch": 23.15, + "learning_rate": 1.1427663324568754e-05, + "loss": 0.9227, + "step": 267490 + }, + { + "epoch": 23.15, + "learning_rate": 1.1426218613655407e-05, + "loss": 0.8826, + "step": 267500 + }, + { + "epoch": 23.16, + "learning_rate": 1.1424773902742061e-05, + "loss": 0.9127, + "step": 267510 + }, + { + "epoch": 23.16, + "learning_rate": 1.1423329191828716e-05, + "loss": 0.8698, + "step": 267520 + }, + { + "epoch": 23.16, + "learning_rate": 1.142188448091537e-05, + "loss": 0.9139, + "step": 267530 + }, + { + "epoch": 23.16, + "learning_rate": 1.1420439770002023e-05, + "loss": 0.9404, + "step": 267540 + }, + { + "epoch": 23.16, + "learning_rate": 1.1418995059088678e-05, + "loss": 0.923, + "step": 267550 + }, + { + "epoch": 23.16, + "learning_rate": 1.1417550348175331e-05, + "loss": 0.9517, + "step": 267560 + }, + { + "epoch": 23.16, + "learning_rate": 1.1416105637261985e-05, + "loss": 0.9162, + "step": 267570 + }, + { + "epoch": 23.16, + "learning_rate": 1.1414660926348638e-05, + "loss": 0.8857, + "step": 267580 + }, + { + "epoch": 23.16, + "learning_rate": 1.1413216215435291e-05, + "loss": 0.8991, + "step": 267590 + }, + { + "epoch": 23.16, + "learning_rate": 1.1411771504521945e-05, + "loss": 0.9161, + "step": 267600 + }, + { + "epoch": 23.16, + "learning_rate": 1.14103267936086e-05, + "loss": 0.953, + "step": 267610 + }, + { + "epoch": 23.16, + "learning_rate": 1.1408882082695253e-05, + "loss": 0.8902, + "step": 267620 + }, + { + "epoch": 23.17, + "learning_rate": 1.1407437371781907e-05, + "loss": 0.9298, + "step": 267630 + }, + { + "epoch": 23.17, + "learning_rate": 1.1405992660868562e-05, + "loss": 0.8461, + "step": 267640 + }, + { + "epoch": 23.17, + "learning_rate": 1.1404547949955215e-05, + "loss": 0.9411, + "step": 267650 + }, + { + "epoch": 23.17, + "learning_rate": 1.1403103239041869e-05, + "loss": 0.8719, + "step": 267660 + }, + { + "epoch": 23.17, + "learning_rate": 1.1401658528128522e-05, + "loss": 0.899, + "step": 267670 + }, + { + "epoch": 23.17, + "learning_rate": 1.1400213817215175e-05, + "loss": 0.9637, + "step": 267680 + }, + { + "epoch": 23.17, + "learning_rate": 1.1398769106301829e-05, + "loss": 0.9138, + "step": 267690 + }, + { + "epoch": 23.17, + "learning_rate": 1.1397324395388484e-05, + "loss": 0.9183, + "step": 267700 + }, + { + "epoch": 23.17, + "learning_rate": 1.1395879684475137e-05, + "loss": 0.9001, + "step": 267710 + }, + { + "epoch": 23.17, + "learning_rate": 1.139443497356179e-05, + "loss": 0.9335, + "step": 267720 + }, + { + "epoch": 23.17, + "learning_rate": 1.1392990262648446e-05, + "loss": 0.9537, + "step": 267730 + }, + { + "epoch": 23.17, + "learning_rate": 1.1391545551735099e-05, + "loss": 0.9258, + "step": 267740 + }, + { + "epoch": 23.18, + "learning_rate": 1.1390100840821752e-05, + "loss": 0.9421, + "step": 267750 + }, + { + "epoch": 23.18, + "learning_rate": 1.1388656129908406e-05, + "loss": 0.9201, + "step": 267760 + }, + { + "epoch": 23.18, + "learning_rate": 1.138721141899506e-05, + "loss": 0.9289, + "step": 267770 + }, + { + "epoch": 23.18, + "learning_rate": 1.1385766708081713e-05, + "loss": 0.9233, + "step": 267780 + }, + { + "epoch": 23.18, + "learning_rate": 1.1384321997168366e-05, + "loss": 0.9069, + "step": 267790 + }, + { + "epoch": 23.18, + "learning_rate": 1.1382877286255021e-05, + "loss": 0.9489, + "step": 267800 + }, + { + "epoch": 23.18, + "learning_rate": 1.1381432575341675e-05, + "loss": 0.9328, + "step": 267810 + }, + { + "epoch": 23.18, + "learning_rate": 1.1379987864428328e-05, + "loss": 0.928, + "step": 267820 + }, + { + "epoch": 23.18, + "learning_rate": 1.1378543153514983e-05, + "loss": 0.8354, + "step": 267830 + }, + { + "epoch": 23.18, + "learning_rate": 1.1377098442601636e-05, + "loss": 0.8834, + "step": 267840 + }, + { + "epoch": 23.18, + "learning_rate": 1.137565373168829e-05, + "loss": 0.9222, + "step": 267850 + }, + { + "epoch": 23.19, + "learning_rate": 1.1374209020774943e-05, + "loss": 0.9722, + "step": 267860 + }, + { + "epoch": 23.19, + "learning_rate": 1.1372764309861597e-05, + "loss": 0.9568, + "step": 267870 + }, + { + "epoch": 23.19, + "learning_rate": 1.137131959894825e-05, + "loss": 0.9306, + "step": 267880 + }, + { + "epoch": 23.19, + "learning_rate": 1.1369874888034905e-05, + "loss": 0.9034, + "step": 267890 + }, + { + "epoch": 23.19, + "learning_rate": 1.1368430177121559e-05, + "loss": 0.9202, + "step": 267900 + }, + { + "epoch": 23.19, + "learning_rate": 1.1366985466208212e-05, + "loss": 0.9086, + "step": 267910 + }, + { + "epoch": 23.19, + "learning_rate": 1.1365540755294867e-05, + "loss": 0.8602, + "step": 267920 + }, + { + "epoch": 23.19, + "learning_rate": 1.136409604438152e-05, + "loss": 0.8909, + "step": 267930 + }, + { + "epoch": 23.19, + "learning_rate": 1.1362651333468174e-05, + "loss": 0.9417, + "step": 267940 + }, + { + "epoch": 23.19, + "learning_rate": 1.1361206622554827e-05, + "loss": 0.9377, + "step": 267950 + }, + { + "epoch": 23.19, + "learning_rate": 1.135976191164148e-05, + "loss": 0.9815, + "step": 267960 + }, + { + "epoch": 23.19, + "learning_rate": 1.1358317200728134e-05, + "loss": 0.8982, + "step": 267970 + }, + { + "epoch": 23.2, + "learning_rate": 1.1356872489814789e-05, + "loss": 0.9277, + "step": 267980 + }, + { + "epoch": 23.2, + "learning_rate": 1.1355427778901442e-05, + "loss": 0.8967, + "step": 267990 + }, + { + "epoch": 23.2, + "learning_rate": 1.1353983067988096e-05, + "loss": 0.9187, + "step": 268000 + }, + { + "epoch": 23.2, + "learning_rate": 1.1352538357074751e-05, + "loss": 0.9221, + "step": 268010 + }, + { + "epoch": 23.2, + "learning_rate": 1.1351093646161404e-05, + "loss": 0.9185, + "step": 268020 + }, + { + "epoch": 23.2, + "learning_rate": 1.1349648935248058e-05, + "loss": 1.0045, + "step": 268030 + }, + { + "epoch": 23.2, + "learning_rate": 1.1348204224334711e-05, + "loss": 0.907, + "step": 268040 + }, + { + "epoch": 23.2, + "learning_rate": 1.1346759513421365e-05, + "loss": 0.8416, + "step": 268050 + }, + { + "epoch": 23.2, + "learning_rate": 1.1345314802508018e-05, + "loss": 0.9277, + "step": 268060 + }, + { + "epoch": 23.2, + "learning_rate": 1.1343870091594673e-05, + "loss": 0.9404, + "step": 268070 + }, + { + "epoch": 23.2, + "learning_rate": 1.1342425380681326e-05, + "loss": 0.8345, + "step": 268080 + }, + { + "epoch": 23.21, + "learning_rate": 1.134098066976798e-05, + "loss": 0.9179, + "step": 268090 + }, + { + "epoch": 23.21, + "learning_rate": 1.1339535958854635e-05, + "loss": 0.9207, + "step": 268100 + }, + { + "epoch": 23.21, + "learning_rate": 1.1338091247941288e-05, + "loss": 0.9202, + "step": 268110 + }, + { + "epoch": 23.21, + "learning_rate": 1.1336646537027942e-05, + "loss": 0.9279, + "step": 268120 + }, + { + "epoch": 23.21, + "learning_rate": 1.1335201826114595e-05, + "loss": 0.8578, + "step": 268130 + }, + { + "epoch": 23.21, + "learning_rate": 1.1333757115201248e-05, + "loss": 0.8954, + "step": 268140 + }, + { + "epoch": 23.21, + "learning_rate": 1.1332312404287902e-05, + "loss": 1.016, + "step": 268150 + }, + { + "epoch": 23.21, + "learning_rate": 1.1330867693374557e-05, + "loss": 0.9391, + "step": 268160 + }, + { + "epoch": 23.21, + "learning_rate": 1.132942298246121e-05, + "loss": 0.9367, + "step": 268170 + }, + { + "epoch": 23.21, + "learning_rate": 1.1327978271547864e-05, + "loss": 0.8909, + "step": 268180 + }, + { + "epoch": 23.21, + "learning_rate": 1.1326533560634519e-05, + "loss": 0.837, + "step": 268190 + }, + { + "epoch": 23.21, + "learning_rate": 1.1325088849721172e-05, + "loss": 0.9658, + "step": 268200 + }, + { + "epoch": 23.22, + "learning_rate": 1.1323644138807826e-05, + "loss": 0.8817, + "step": 268210 + }, + { + "epoch": 23.22, + "learning_rate": 1.1322199427894479e-05, + "loss": 0.9207, + "step": 268220 + }, + { + "epoch": 23.22, + "learning_rate": 1.1320754716981132e-05, + "loss": 0.9099, + "step": 268230 + }, + { + "epoch": 23.22, + "learning_rate": 1.1319310006067786e-05, + "loss": 0.9173, + "step": 268240 + }, + { + "epoch": 23.22, + "learning_rate": 1.131786529515444e-05, + "loss": 0.9718, + "step": 268250 + }, + { + "epoch": 23.22, + "learning_rate": 1.1316420584241094e-05, + "loss": 0.864, + "step": 268260 + }, + { + "epoch": 23.22, + "learning_rate": 1.1314975873327748e-05, + "loss": 0.9248, + "step": 268270 + }, + { + "epoch": 23.22, + "learning_rate": 1.1313531162414401e-05, + "loss": 0.9761, + "step": 268280 + }, + { + "epoch": 23.22, + "learning_rate": 1.1312086451501056e-05, + "loss": 0.8941, + "step": 268290 + }, + { + "epoch": 23.22, + "learning_rate": 1.131064174058771e-05, + "loss": 0.8929, + "step": 268300 + }, + { + "epoch": 23.22, + "learning_rate": 1.1309197029674363e-05, + "loss": 0.8947, + "step": 268310 + }, + { + "epoch": 23.23, + "learning_rate": 1.1307752318761016e-05, + "loss": 0.8717, + "step": 268320 + }, + { + "epoch": 23.23, + "learning_rate": 1.130630760784767e-05, + "loss": 0.9248, + "step": 268330 + }, + { + "epoch": 23.23, + "learning_rate": 1.1304862896934323e-05, + "loss": 0.9064, + "step": 268340 + }, + { + "epoch": 23.23, + "learning_rate": 1.1303418186020978e-05, + "loss": 0.8821, + "step": 268350 + }, + { + "epoch": 23.23, + "learning_rate": 1.1301973475107632e-05, + "loss": 0.9922, + "step": 268360 + }, + { + "epoch": 23.23, + "learning_rate": 1.1300528764194285e-05, + "loss": 0.8896, + "step": 268370 + }, + { + "epoch": 23.23, + "learning_rate": 1.129908405328094e-05, + "loss": 0.8922, + "step": 268380 + }, + { + "epoch": 23.23, + "learning_rate": 1.1297639342367593e-05, + "loss": 0.9063, + "step": 268390 + }, + { + "epoch": 23.23, + "learning_rate": 1.1296194631454247e-05, + "loss": 0.8543, + "step": 268400 + }, + { + "epoch": 23.23, + "learning_rate": 1.12947499205409e-05, + "loss": 0.9513, + "step": 268410 + }, + { + "epoch": 23.23, + "learning_rate": 1.1293305209627554e-05, + "loss": 0.9588, + "step": 268420 + }, + { + "epoch": 23.23, + "learning_rate": 1.1291860498714207e-05, + "loss": 0.9434, + "step": 268430 + }, + { + "epoch": 23.24, + "learning_rate": 1.1290415787800862e-05, + "loss": 0.9125, + "step": 268440 + }, + { + "epoch": 23.24, + "learning_rate": 1.1288971076887516e-05, + "loss": 0.878, + "step": 268450 + }, + { + "epoch": 23.24, + "learning_rate": 1.1287526365974169e-05, + "loss": 0.9415, + "step": 268460 + }, + { + "epoch": 23.24, + "learning_rate": 1.1286081655060824e-05, + "loss": 0.912, + "step": 268470 + }, + { + "epoch": 23.24, + "learning_rate": 1.1284636944147477e-05, + "loss": 0.9526, + "step": 268480 + }, + { + "epoch": 23.24, + "learning_rate": 1.128319223323413e-05, + "loss": 0.8934, + "step": 268490 + }, + { + "epoch": 23.24, + "learning_rate": 1.1281747522320784e-05, + "loss": 0.9318, + "step": 268500 + }, + { + "epoch": 23.24, + "learning_rate": 1.1280302811407438e-05, + "loss": 0.8914, + "step": 268510 + }, + { + "epoch": 23.24, + "learning_rate": 1.1278858100494091e-05, + "loss": 0.8953, + "step": 268520 + }, + { + "epoch": 23.24, + "learning_rate": 1.1277413389580746e-05, + "loss": 0.8963, + "step": 268530 + }, + { + "epoch": 23.24, + "learning_rate": 1.12759686786674e-05, + "loss": 0.9353, + "step": 268540 + }, + { + "epoch": 23.25, + "learning_rate": 1.1274523967754053e-05, + "loss": 0.8878, + "step": 268550 + }, + { + "epoch": 23.25, + "learning_rate": 1.1273079256840708e-05, + "loss": 1.0016, + "step": 268560 + }, + { + "epoch": 23.25, + "learning_rate": 1.1271634545927361e-05, + "loss": 0.9635, + "step": 268570 + }, + { + "epoch": 23.25, + "learning_rate": 1.1270189835014015e-05, + "loss": 0.87, + "step": 268580 + }, + { + "epoch": 23.25, + "learning_rate": 1.1268745124100668e-05, + "loss": 0.8975, + "step": 268590 + }, + { + "epoch": 23.25, + "learning_rate": 1.1267300413187322e-05, + "loss": 0.9109, + "step": 268600 + }, + { + "epoch": 23.25, + "learning_rate": 1.1265855702273975e-05, + "loss": 0.8938, + "step": 268610 + }, + { + "epoch": 23.25, + "learning_rate": 1.126441099136063e-05, + "loss": 0.9077, + "step": 268620 + }, + { + "epoch": 23.25, + "learning_rate": 1.1262966280447283e-05, + "loss": 0.9477, + "step": 268630 + }, + { + "epoch": 23.25, + "learning_rate": 1.1261521569533937e-05, + "loss": 0.9509, + "step": 268640 + }, + { + "epoch": 23.25, + "learning_rate": 1.1260076858620592e-05, + "loss": 0.8576, + "step": 268650 + }, + { + "epoch": 23.25, + "learning_rate": 1.1258632147707245e-05, + "loss": 0.9356, + "step": 268660 + }, + { + "epoch": 23.26, + "learning_rate": 1.1257187436793899e-05, + "loss": 0.8713, + "step": 268670 + }, + { + "epoch": 23.26, + "learning_rate": 1.125574272588055e-05, + "loss": 0.9994, + "step": 268680 + }, + { + "epoch": 23.26, + "learning_rate": 1.1254298014967206e-05, + "loss": 0.8041, + "step": 268690 + }, + { + "epoch": 23.26, + "learning_rate": 1.1252853304053859e-05, + "loss": 0.9354, + "step": 268700 + }, + { + "epoch": 23.26, + "learning_rate": 1.1251408593140512e-05, + "loss": 0.9528, + "step": 268710 + }, + { + "epoch": 23.26, + "learning_rate": 1.1249963882227167e-05, + "loss": 0.9356, + "step": 268720 + }, + { + "epoch": 23.26, + "learning_rate": 1.124851917131382e-05, + "loss": 0.928, + "step": 268730 + }, + { + "epoch": 23.26, + "learning_rate": 1.1247074460400474e-05, + "loss": 0.9358, + "step": 268740 + }, + { + "epoch": 23.26, + "learning_rate": 1.124562974948713e-05, + "loss": 0.9202, + "step": 268750 + }, + { + "epoch": 23.26, + "learning_rate": 1.1244185038573781e-05, + "loss": 0.8883, + "step": 268760 + }, + { + "epoch": 23.26, + "learning_rate": 1.1242740327660434e-05, + "loss": 0.8969, + "step": 268770 + }, + { + "epoch": 23.26, + "learning_rate": 1.124129561674709e-05, + "loss": 0.9452, + "step": 268780 + }, + { + "epoch": 23.27, + "learning_rate": 1.1239850905833743e-05, + "loss": 0.9595, + "step": 268790 + }, + { + "epoch": 23.27, + "learning_rate": 1.1238406194920396e-05, + "loss": 0.9281, + "step": 268800 + }, + { + "epoch": 23.27, + "learning_rate": 1.1236961484007051e-05, + "loss": 0.8633, + "step": 268810 + }, + { + "epoch": 23.27, + "learning_rate": 1.1235516773093705e-05, + "loss": 0.8948, + "step": 268820 + }, + { + "epoch": 23.27, + "learning_rate": 1.1234072062180358e-05, + "loss": 0.9468, + "step": 268830 + }, + { + "epoch": 23.27, + "learning_rate": 1.1232627351267013e-05, + "loss": 0.9485, + "step": 268840 + }, + { + "epoch": 23.27, + "learning_rate": 1.1231182640353665e-05, + "loss": 0.9091, + "step": 268850 + }, + { + "epoch": 23.27, + "learning_rate": 1.1229737929440318e-05, + "loss": 0.9423, + "step": 268860 + }, + { + "epoch": 23.27, + "learning_rate": 1.1228293218526973e-05, + "loss": 0.9422, + "step": 268870 + }, + { + "epoch": 23.27, + "learning_rate": 1.1226848507613627e-05, + "loss": 0.9181, + "step": 268880 + }, + { + "epoch": 23.27, + "learning_rate": 1.122540379670028e-05, + "loss": 0.8945, + "step": 268890 + }, + { + "epoch": 23.28, + "learning_rate": 1.1223959085786935e-05, + "loss": 0.9017, + "step": 268900 + }, + { + "epoch": 23.28, + "learning_rate": 1.1222514374873589e-05, + "loss": 0.9957, + "step": 268910 + }, + { + "epoch": 23.28, + "learning_rate": 1.1221069663960242e-05, + "loss": 0.9236, + "step": 268920 + }, + { + "epoch": 23.28, + "learning_rate": 1.1219624953046897e-05, + "loss": 0.9315, + "step": 268930 + }, + { + "epoch": 23.28, + "learning_rate": 1.1218180242133549e-05, + "loss": 0.912, + "step": 268940 + }, + { + "epoch": 23.28, + "learning_rate": 1.1216735531220202e-05, + "loss": 0.8644, + "step": 268950 + }, + { + "epoch": 23.28, + "learning_rate": 1.1215290820306857e-05, + "loss": 0.9216, + "step": 268960 + }, + { + "epoch": 23.28, + "learning_rate": 1.121384610939351e-05, + "loss": 0.9539, + "step": 268970 + }, + { + "epoch": 23.28, + "learning_rate": 1.1212401398480164e-05, + "loss": 0.9478, + "step": 268980 + }, + { + "epoch": 23.28, + "learning_rate": 1.121095668756682e-05, + "loss": 0.9309, + "step": 268990 + }, + { + "epoch": 23.28, + "learning_rate": 1.1209511976653473e-05, + "loss": 0.8986, + "step": 269000 + }, + { + "epoch": 23.28, + "learning_rate": 1.1208067265740126e-05, + "loss": 0.9272, + "step": 269010 + }, + { + "epoch": 23.29, + "learning_rate": 1.1206622554826781e-05, + "loss": 0.8845, + "step": 269020 + }, + { + "epoch": 23.29, + "learning_rate": 1.1205177843913433e-05, + "loss": 0.8989, + "step": 269030 + }, + { + "epoch": 23.29, + "learning_rate": 1.1203733133000086e-05, + "loss": 0.8243, + "step": 269040 + }, + { + "epoch": 23.29, + "learning_rate": 1.1202288422086741e-05, + "loss": 0.9329, + "step": 269050 + }, + { + "epoch": 23.29, + "learning_rate": 1.1200843711173395e-05, + "loss": 0.9469, + "step": 269060 + }, + { + "epoch": 23.29, + "learning_rate": 1.1199399000260048e-05, + "loss": 0.9017, + "step": 269070 + }, + { + "epoch": 23.29, + "learning_rate": 1.1197954289346703e-05, + "loss": 0.847, + "step": 269080 + }, + { + "epoch": 23.29, + "learning_rate": 1.1196509578433357e-05, + "loss": 0.9371, + "step": 269090 + }, + { + "epoch": 23.29, + "learning_rate": 1.119506486752001e-05, + "loss": 0.9782, + "step": 269100 + }, + { + "epoch": 23.29, + "learning_rate": 1.1193620156606663e-05, + "loss": 0.9149, + "step": 269110 + }, + { + "epoch": 23.29, + "learning_rate": 1.1192175445693317e-05, + "loss": 0.9417, + "step": 269120 + }, + { + "epoch": 23.3, + "learning_rate": 1.119073073477997e-05, + "loss": 0.9588, + "step": 269130 + }, + { + "epoch": 23.3, + "learning_rate": 1.1189286023866624e-05, + "loss": 0.9429, + "step": 269140 + }, + { + "epoch": 23.3, + "learning_rate": 1.1187841312953279e-05, + "loss": 0.8795, + "step": 269150 + }, + { + "epoch": 23.3, + "learning_rate": 1.1186396602039932e-05, + "loss": 0.9262, + "step": 269160 + }, + { + "epoch": 23.3, + "learning_rate": 1.1184951891126585e-05, + "loss": 0.8612, + "step": 269170 + }, + { + "epoch": 23.3, + "learning_rate": 1.118350718021324e-05, + "loss": 0.9657, + "step": 269180 + }, + { + "epoch": 23.3, + "learning_rate": 1.1182062469299894e-05, + "loss": 0.9234, + "step": 269190 + }, + { + "epoch": 23.3, + "learning_rate": 1.1180617758386547e-05, + "loss": 0.8948, + "step": 269200 + }, + { + "epoch": 23.3, + "learning_rate": 1.11791730474732e-05, + "loss": 0.9902, + "step": 269210 + }, + { + "epoch": 23.3, + "learning_rate": 1.1177728336559854e-05, + "loss": 0.9113, + "step": 269220 + }, + { + "epoch": 23.3, + "learning_rate": 1.1176283625646507e-05, + "loss": 0.9341, + "step": 269230 + }, + { + "epoch": 23.3, + "learning_rate": 1.1174838914733163e-05, + "loss": 0.8852, + "step": 269240 + }, + { + "epoch": 23.31, + "learning_rate": 1.1173394203819816e-05, + "loss": 0.8653, + "step": 269250 + }, + { + "epoch": 23.31, + "learning_rate": 1.117194949290647e-05, + "loss": 0.9289, + "step": 269260 + }, + { + "epoch": 23.31, + "learning_rate": 1.1170504781993124e-05, + "loss": 0.9471, + "step": 269270 + }, + { + "epoch": 23.31, + "learning_rate": 1.1169060071079778e-05, + "loss": 0.8943, + "step": 269280 + }, + { + "epoch": 23.31, + "learning_rate": 1.1167615360166431e-05, + "loss": 0.9505, + "step": 269290 + }, + { + "epoch": 23.31, + "learning_rate": 1.1166170649253085e-05, + "loss": 0.9205, + "step": 269300 + }, + { + "epoch": 23.31, + "learning_rate": 1.1164725938339738e-05, + "loss": 0.9285, + "step": 269310 + }, + { + "epoch": 23.31, + "learning_rate": 1.1163281227426391e-05, + "loss": 0.8929, + "step": 269320 + }, + { + "epoch": 23.31, + "learning_rate": 1.1161836516513046e-05, + "loss": 0.9145, + "step": 269330 + }, + { + "epoch": 23.31, + "learning_rate": 1.11603918055997e-05, + "loss": 0.9333, + "step": 269340 + }, + { + "epoch": 23.31, + "learning_rate": 1.1158947094686353e-05, + "loss": 0.9147, + "step": 269350 + }, + { + "epoch": 23.32, + "learning_rate": 1.1157502383773008e-05, + "loss": 0.897, + "step": 269360 + }, + { + "epoch": 23.32, + "learning_rate": 1.1156057672859662e-05, + "loss": 0.9621, + "step": 269370 + }, + { + "epoch": 23.32, + "learning_rate": 1.1154612961946315e-05, + "loss": 0.9314, + "step": 269380 + }, + { + "epoch": 23.32, + "learning_rate": 1.1153168251032969e-05, + "loss": 0.9167, + "step": 269390 + }, + { + "epoch": 23.32, + "learning_rate": 1.1151723540119622e-05, + "loss": 0.9141, + "step": 269400 + }, + { + "epoch": 23.32, + "learning_rate": 1.1150278829206275e-05, + "loss": 0.9464, + "step": 269410 + }, + { + "epoch": 23.32, + "learning_rate": 1.114883411829293e-05, + "loss": 0.9178, + "step": 269420 + }, + { + "epoch": 23.32, + "learning_rate": 1.1147389407379584e-05, + "loss": 0.9052, + "step": 269430 + }, + { + "epoch": 23.32, + "learning_rate": 1.1145944696466237e-05, + "loss": 0.9083, + "step": 269440 + }, + { + "epoch": 23.32, + "learning_rate": 1.1144499985552892e-05, + "loss": 0.9129, + "step": 269450 + }, + { + "epoch": 23.32, + "learning_rate": 1.1143055274639546e-05, + "loss": 0.8815, + "step": 269460 + }, + { + "epoch": 23.32, + "learning_rate": 1.1141610563726199e-05, + "loss": 0.89, + "step": 269470 + }, + { + "epoch": 23.33, + "learning_rate": 1.1140165852812853e-05, + "loss": 0.936, + "step": 269480 + }, + { + "epoch": 23.33, + "learning_rate": 1.1138721141899506e-05, + "loss": 0.9577, + "step": 269490 + }, + { + "epoch": 23.33, + "learning_rate": 1.113727643098616e-05, + "loss": 0.9582, + "step": 269500 + }, + { + "epoch": 23.33, + "learning_rate": 1.1135831720072814e-05, + "loss": 0.8992, + "step": 269510 + }, + { + "epoch": 23.33, + "learning_rate": 1.1134387009159468e-05, + "loss": 0.9148, + "step": 269520 + }, + { + "epoch": 23.33, + "learning_rate": 1.1132942298246121e-05, + "loss": 0.962, + "step": 269530 + }, + { + "epoch": 23.33, + "learning_rate": 1.1131497587332776e-05, + "loss": 0.939, + "step": 269540 + }, + { + "epoch": 23.33, + "learning_rate": 1.113005287641943e-05, + "loss": 0.9381, + "step": 269550 + }, + { + "epoch": 23.33, + "learning_rate": 1.1128608165506083e-05, + "loss": 0.8976, + "step": 269560 + }, + { + "epoch": 23.33, + "learning_rate": 1.1127163454592736e-05, + "loss": 0.8698, + "step": 269570 + }, + { + "epoch": 23.33, + "learning_rate": 1.112571874367939e-05, + "loss": 0.9094, + "step": 269580 + }, + { + "epoch": 23.34, + "learning_rate": 1.1124274032766043e-05, + "loss": 0.9088, + "step": 269590 + }, + { + "epoch": 23.34, + "learning_rate": 1.1122829321852697e-05, + "loss": 0.9344, + "step": 269600 + }, + { + "epoch": 23.34, + "learning_rate": 1.1121384610939352e-05, + "loss": 0.9098, + "step": 269610 + }, + { + "epoch": 23.34, + "learning_rate": 1.1119939900026005e-05, + "loss": 0.9094, + "step": 269620 + }, + { + "epoch": 23.34, + "learning_rate": 1.1118495189112659e-05, + "loss": 0.8621, + "step": 269630 + }, + { + "epoch": 23.34, + "learning_rate": 1.1117050478199314e-05, + "loss": 0.8837, + "step": 269640 + }, + { + "epoch": 23.34, + "learning_rate": 1.1115605767285967e-05, + "loss": 0.9354, + "step": 269650 + }, + { + "epoch": 23.34, + "learning_rate": 1.111416105637262e-05, + "loss": 0.9276, + "step": 269660 + }, + { + "epoch": 23.34, + "learning_rate": 1.1112716345459274e-05, + "loss": 0.9238, + "step": 269670 + }, + { + "epoch": 23.34, + "learning_rate": 1.1111271634545927e-05, + "loss": 0.9282, + "step": 269680 + }, + { + "epoch": 23.34, + "learning_rate": 1.110982692363258e-05, + "loss": 0.9419, + "step": 269690 + }, + { + "epoch": 23.34, + "learning_rate": 1.1108382212719236e-05, + "loss": 0.8757, + "step": 269700 + }, + { + "epoch": 23.35, + "learning_rate": 1.1106937501805889e-05, + "loss": 0.9188, + "step": 269710 + }, + { + "epoch": 23.35, + "learning_rate": 1.1105492790892542e-05, + "loss": 0.8944, + "step": 269720 + }, + { + "epoch": 23.35, + "learning_rate": 1.1104048079979198e-05, + "loss": 0.9239, + "step": 269730 + }, + { + "epoch": 23.35, + "learning_rate": 1.1102603369065851e-05, + "loss": 0.8928, + "step": 269740 + }, + { + "epoch": 23.35, + "learning_rate": 1.1101158658152504e-05, + "loss": 0.939, + "step": 269750 + }, + { + "epoch": 23.35, + "learning_rate": 1.1099713947239158e-05, + "loss": 0.9227, + "step": 269760 + }, + { + "epoch": 23.35, + "learning_rate": 1.1098269236325811e-05, + "loss": 0.913, + "step": 269770 + }, + { + "epoch": 23.35, + "learning_rate": 1.1096824525412465e-05, + "loss": 0.9313, + "step": 269780 + }, + { + "epoch": 23.35, + "learning_rate": 1.109537981449912e-05, + "loss": 0.9039, + "step": 269790 + }, + { + "epoch": 23.35, + "learning_rate": 1.1093935103585773e-05, + "loss": 0.8788, + "step": 269800 + }, + { + "epoch": 23.35, + "learning_rate": 1.1092490392672426e-05, + "loss": 0.9013, + "step": 269810 + }, + { + "epoch": 23.35, + "learning_rate": 1.1091045681759081e-05, + "loss": 0.9083, + "step": 269820 + }, + { + "epoch": 23.36, + "learning_rate": 1.1089600970845735e-05, + "loss": 0.9199, + "step": 269830 + }, + { + "epoch": 23.36, + "learning_rate": 1.1088156259932388e-05, + "loss": 0.9106, + "step": 269840 + }, + { + "epoch": 23.36, + "learning_rate": 1.1086711549019042e-05, + "loss": 0.9374, + "step": 269850 + }, + { + "epoch": 23.36, + "learning_rate": 1.1085266838105695e-05, + "loss": 0.9224, + "step": 269860 + }, + { + "epoch": 23.36, + "learning_rate": 1.1083822127192348e-05, + "loss": 0.9435, + "step": 269870 + }, + { + "epoch": 23.36, + "learning_rate": 1.1082377416279004e-05, + "loss": 0.9236, + "step": 269880 + }, + { + "epoch": 23.36, + "learning_rate": 1.1080932705365657e-05, + "loss": 0.915, + "step": 269890 + }, + { + "epoch": 23.36, + "learning_rate": 1.107948799445231e-05, + "loss": 0.8694, + "step": 269900 + }, + { + "epoch": 23.36, + "learning_rate": 1.1078043283538965e-05, + "loss": 0.9534, + "step": 269910 + }, + { + "epoch": 23.36, + "learning_rate": 1.1076598572625619e-05, + "loss": 0.8876, + "step": 269920 + }, + { + "epoch": 23.36, + "learning_rate": 1.1075153861712272e-05, + "loss": 0.9262, + "step": 269930 + }, + { + "epoch": 23.37, + "learning_rate": 1.1073709150798926e-05, + "loss": 0.8842, + "step": 269940 + }, + { + "epoch": 23.37, + "learning_rate": 1.1072264439885579e-05, + "loss": 0.9522, + "step": 269950 + }, + { + "epoch": 23.37, + "learning_rate": 1.1070819728972232e-05, + "loss": 0.8819, + "step": 269960 + }, + { + "epoch": 23.37, + "learning_rate": 1.1069375018058887e-05, + "loss": 0.9415, + "step": 269970 + }, + { + "epoch": 23.37, + "learning_rate": 1.1067930307145541e-05, + "loss": 0.9103, + "step": 269980 + }, + { + "epoch": 23.37, + "learning_rate": 1.1066485596232194e-05, + "loss": 0.9069, + "step": 269990 + }, + { + "epoch": 23.37, + "learning_rate": 1.106504088531885e-05, + "loss": 0.9163, + "step": 270000 + }, + { + "epoch": 23.37, + "learning_rate": 1.1063596174405503e-05, + "loss": 0.9153, + "step": 270010 + }, + { + "epoch": 23.37, + "learning_rate": 1.1062151463492156e-05, + "loss": 0.9257, + "step": 270020 + }, + { + "epoch": 23.37, + "learning_rate": 1.106070675257881e-05, + "loss": 0.9164, + "step": 270030 + }, + { + "epoch": 23.37, + "learning_rate": 1.1059262041665463e-05, + "loss": 0.9472, + "step": 270040 + }, + { + "epoch": 23.37, + "learning_rate": 1.1057817330752116e-05, + "loss": 0.9455, + "step": 270050 + }, + { + "epoch": 23.38, + "learning_rate": 1.105637261983877e-05, + "loss": 0.9665, + "step": 270060 + }, + { + "epoch": 23.38, + "learning_rate": 1.1054927908925425e-05, + "loss": 0.9086, + "step": 270070 + }, + { + "epoch": 23.38, + "learning_rate": 1.1053483198012078e-05, + "loss": 0.8896, + "step": 270080 + }, + { + "epoch": 23.38, + "learning_rate": 1.1052038487098732e-05, + "loss": 0.9594, + "step": 270090 + }, + { + "epoch": 23.38, + "learning_rate": 1.1050593776185387e-05, + "loss": 0.9116, + "step": 270100 + }, + { + "epoch": 23.38, + "learning_rate": 1.104914906527204e-05, + "loss": 0.9764, + "step": 270110 + }, + { + "epoch": 23.38, + "learning_rate": 1.1047704354358693e-05, + "loss": 0.8964, + "step": 270120 + }, + { + "epoch": 23.38, + "learning_rate": 1.1046259643445347e-05, + "loss": 0.9138, + "step": 270130 + }, + { + "epoch": 23.38, + "learning_rate": 1.1044814932532e-05, + "loss": 0.9016, + "step": 270140 + }, + { + "epoch": 23.38, + "learning_rate": 1.1043370221618654e-05, + "loss": 0.9746, + "step": 270150 + }, + { + "epoch": 23.38, + "learning_rate": 1.1041925510705309e-05, + "loss": 0.9403, + "step": 270160 + }, + { + "epoch": 23.39, + "learning_rate": 1.1040480799791962e-05, + "loss": 0.8811, + "step": 270170 + }, + { + "epoch": 23.39, + "learning_rate": 1.1039036088878616e-05, + "loss": 0.8903, + "step": 270180 + }, + { + "epoch": 23.39, + "learning_rate": 1.103759137796527e-05, + "loss": 0.8716, + "step": 270190 + }, + { + "epoch": 23.39, + "learning_rate": 1.1036146667051924e-05, + "loss": 0.8898, + "step": 270200 + }, + { + "epoch": 23.39, + "learning_rate": 1.1034701956138577e-05, + "loss": 0.9023, + "step": 270210 + }, + { + "epoch": 23.39, + "learning_rate": 1.103325724522523e-05, + "loss": 0.952, + "step": 270220 + }, + { + "epoch": 23.39, + "learning_rate": 1.1031812534311884e-05, + "loss": 0.8761, + "step": 270230 + }, + { + "epoch": 23.39, + "learning_rate": 1.1030367823398538e-05, + "loss": 0.9355, + "step": 270240 + }, + { + "epoch": 23.39, + "learning_rate": 1.1028923112485193e-05, + "loss": 0.8955, + "step": 270250 + }, + { + "epoch": 23.39, + "learning_rate": 1.1027478401571846e-05, + "loss": 0.9781, + "step": 270260 + }, + { + "epoch": 23.39, + "learning_rate": 1.10260336906585e-05, + "loss": 0.8671, + "step": 270270 + }, + { + "epoch": 23.39, + "learning_rate": 1.1024588979745155e-05, + "loss": 0.9015, + "step": 270280 + }, + { + "epoch": 23.4, + "learning_rate": 1.1023144268831808e-05, + "loss": 0.9252, + "step": 270290 + }, + { + "epoch": 23.4, + "learning_rate": 1.1021699557918461e-05, + "loss": 0.9, + "step": 270300 + }, + { + "epoch": 23.4, + "learning_rate": 1.1020254847005115e-05, + "loss": 0.8596, + "step": 270310 + }, + { + "epoch": 23.4, + "learning_rate": 1.1018810136091768e-05, + "loss": 0.9584, + "step": 270320 + }, + { + "epoch": 23.4, + "learning_rate": 1.1017365425178422e-05, + "loss": 0.9378, + "step": 270330 + }, + { + "epoch": 23.4, + "learning_rate": 1.1015920714265077e-05, + "loss": 0.8622, + "step": 270340 + }, + { + "epoch": 23.4, + "learning_rate": 1.101447600335173e-05, + "loss": 0.9533, + "step": 270350 + }, + { + "epoch": 23.4, + "learning_rate": 1.1013031292438383e-05, + "loss": 0.935, + "step": 270360 + }, + { + "epoch": 23.4, + "learning_rate": 1.1011586581525039e-05, + "loss": 0.8761, + "step": 270370 + }, + { + "epoch": 23.4, + "learning_rate": 1.1010141870611692e-05, + "loss": 0.9322, + "step": 270380 + }, + { + "epoch": 23.4, + "learning_rate": 1.1008697159698345e-05, + "loss": 0.91, + "step": 270390 + }, + { + "epoch": 23.41, + "learning_rate": 1.1007252448784999e-05, + "loss": 0.9352, + "step": 270400 + }, + { + "epoch": 23.41, + "learning_rate": 1.1005807737871652e-05, + "loss": 0.9481, + "step": 270410 + }, + { + "epoch": 23.41, + "learning_rate": 1.1004363026958306e-05, + "loss": 0.8947, + "step": 270420 + }, + { + "epoch": 23.41, + "learning_rate": 1.100291831604496e-05, + "loss": 0.9795, + "step": 270430 + }, + { + "epoch": 23.41, + "learning_rate": 1.1001473605131614e-05, + "loss": 0.9304, + "step": 270440 + }, + { + "epoch": 23.41, + "learning_rate": 1.1000028894218267e-05, + "loss": 0.9072, + "step": 270450 + }, + { + "epoch": 23.41, + "learning_rate": 1.099858418330492e-05, + "loss": 0.9362, + "step": 270460 + }, + { + "epoch": 23.41, + "learning_rate": 1.0997139472391576e-05, + "loss": 0.9163, + "step": 270470 + }, + { + "epoch": 23.41, + "learning_rate": 1.099569476147823e-05, + "loss": 0.9525, + "step": 270480 + }, + { + "epoch": 23.41, + "learning_rate": 1.0994250050564883e-05, + "loss": 0.929, + "step": 270490 + }, + { + "epoch": 23.41, + "learning_rate": 1.0992805339651536e-05, + "loss": 0.9549, + "step": 270500 + }, + { + "epoch": 23.41, + "learning_rate": 1.099136062873819e-05, + "loss": 0.9697, + "step": 270510 + }, + { + "epoch": 23.42, + "learning_rate": 1.0989915917824843e-05, + "loss": 0.8978, + "step": 270520 + }, + { + "epoch": 23.42, + "learning_rate": 1.0988471206911498e-05, + "loss": 0.9473, + "step": 270530 + }, + { + "epoch": 23.42, + "learning_rate": 1.0987026495998151e-05, + "loss": 0.9806, + "step": 270540 + }, + { + "epoch": 23.42, + "learning_rate": 1.0985581785084805e-05, + "loss": 0.9299, + "step": 270550 + }, + { + "epoch": 23.42, + "learning_rate": 1.098413707417146e-05, + "loss": 0.886, + "step": 270560 + }, + { + "epoch": 23.42, + "learning_rate": 1.0982692363258113e-05, + "loss": 0.9256, + "step": 270570 + }, + { + "epoch": 23.42, + "learning_rate": 1.0981247652344767e-05, + "loss": 0.9437, + "step": 270580 + }, + { + "epoch": 23.42, + "learning_rate": 1.097980294143142e-05, + "loss": 0.9146, + "step": 270590 + }, + { + "epoch": 23.42, + "learning_rate": 1.0978358230518073e-05, + "loss": 0.9338, + "step": 270600 + }, + { + "epoch": 23.42, + "learning_rate": 1.0976913519604727e-05, + "loss": 0.9334, + "step": 270610 + }, + { + "epoch": 23.42, + "learning_rate": 1.0975468808691382e-05, + "loss": 0.947, + "step": 270620 + }, + { + "epoch": 23.43, + "learning_rate": 1.0974024097778035e-05, + "loss": 0.8674, + "step": 270630 + }, + { + "epoch": 23.43, + "learning_rate": 1.0972579386864689e-05, + "loss": 0.9637, + "step": 270640 + }, + { + "epoch": 23.43, + "learning_rate": 1.0971134675951344e-05, + "loss": 0.9233, + "step": 270650 + }, + { + "epoch": 23.43, + "learning_rate": 1.0969689965037997e-05, + "loss": 0.9067, + "step": 270660 + }, + { + "epoch": 23.43, + "learning_rate": 1.096824525412465e-05, + "loss": 0.941, + "step": 270670 + }, + { + "epoch": 23.43, + "learning_rate": 1.0966800543211304e-05, + "loss": 0.9986, + "step": 270680 + }, + { + "epoch": 23.43, + "learning_rate": 1.0965355832297957e-05, + "loss": 0.9174, + "step": 270690 + }, + { + "epoch": 23.43, + "learning_rate": 1.096391112138461e-05, + "loss": 0.8964, + "step": 270700 + }, + { + "epoch": 23.43, + "learning_rate": 1.0962466410471266e-05, + "loss": 0.9444, + "step": 270710 + }, + { + "epoch": 23.43, + "learning_rate": 1.096102169955792e-05, + "loss": 0.943, + "step": 270720 + }, + { + "epoch": 23.43, + "learning_rate": 1.0959576988644573e-05, + "loss": 0.9045, + "step": 270730 + }, + { + "epoch": 23.43, + "learning_rate": 1.0958132277731228e-05, + "loss": 0.9153, + "step": 270740 + }, + { + "epoch": 23.44, + "learning_rate": 1.0956687566817881e-05, + "loss": 0.9914, + "step": 270750 + }, + { + "epoch": 23.44, + "learning_rate": 1.0955242855904533e-05, + "loss": 0.8992, + "step": 270760 + }, + { + "epoch": 23.44, + "learning_rate": 1.0953798144991188e-05, + "loss": 0.9075, + "step": 270770 + }, + { + "epoch": 23.44, + "learning_rate": 1.0952353434077841e-05, + "loss": 0.9202, + "step": 270780 + }, + { + "epoch": 23.44, + "learning_rate": 1.0950908723164495e-05, + "loss": 0.9288, + "step": 270790 + }, + { + "epoch": 23.44, + "learning_rate": 1.094946401225115e-05, + "loss": 0.9402, + "step": 270800 + }, + { + "epoch": 23.44, + "learning_rate": 1.0948019301337803e-05, + "loss": 0.8953, + "step": 270810 + }, + { + "epoch": 23.44, + "learning_rate": 1.0946574590424457e-05, + "loss": 0.9057, + "step": 270820 + }, + { + "epoch": 23.44, + "learning_rate": 1.0945129879511112e-05, + "loss": 0.9408, + "step": 270830 + }, + { + "epoch": 23.44, + "learning_rate": 1.0943685168597765e-05, + "loss": 0.9212, + "step": 270840 + }, + { + "epoch": 23.44, + "learning_rate": 1.0942240457684417e-05, + "loss": 0.8794, + "step": 270850 + }, + { + "epoch": 23.44, + "learning_rate": 1.0940795746771072e-05, + "loss": 0.9596, + "step": 270860 + }, + { + "epoch": 23.45, + "learning_rate": 1.0939351035857725e-05, + "loss": 0.8741, + "step": 270870 + }, + { + "epoch": 23.45, + "learning_rate": 1.0937906324944379e-05, + "loss": 0.9479, + "step": 270880 + }, + { + "epoch": 23.45, + "learning_rate": 1.0936461614031034e-05, + "loss": 0.9396, + "step": 270890 + }, + { + "epoch": 23.45, + "learning_rate": 1.0935016903117687e-05, + "loss": 0.8911, + "step": 270900 + }, + { + "epoch": 23.45, + "learning_rate": 1.093357219220434e-05, + "loss": 0.9458, + "step": 270910 + }, + { + "epoch": 23.45, + "learning_rate": 1.0932127481290994e-05, + "loss": 0.8456, + "step": 270920 + }, + { + "epoch": 23.45, + "learning_rate": 1.0930682770377649e-05, + "loss": 0.9696, + "step": 270930 + }, + { + "epoch": 23.45, + "learning_rate": 1.09292380594643e-05, + "loss": 0.9446, + "step": 270940 + }, + { + "epoch": 23.45, + "learning_rate": 1.0927793348550954e-05, + "loss": 0.8862, + "step": 270950 + }, + { + "epoch": 23.45, + "learning_rate": 1.092634863763761e-05, + "loss": 0.9518, + "step": 270960 + }, + { + "epoch": 23.45, + "learning_rate": 1.0924903926724263e-05, + "loss": 0.8712, + "step": 270970 + }, + { + "epoch": 23.46, + "learning_rate": 1.0923459215810916e-05, + "loss": 0.9767, + "step": 270980 + }, + { + "epoch": 23.46, + "learning_rate": 1.0922014504897571e-05, + "loss": 0.9062, + "step": 270990 + }, + { + "epoch": 23.46, + "learning_rate": 1.0920569793984224e-05, + "loss": 0.8863, + "step": 271000 + }, + { + "epoch": 23.46, + "learning_rate": 1.0919125083070878e-05, + "loss": 0.9229, + "step": 271010 + }, + { + "epoch": 23.46, + "learning_rate": 1.0917680372157533e-05, + "loss": 0.9894, + "step": 271020 + }, + { + "epoch": 23.46, + "learning_rate": 1.0916235661244185e-05, + "loss": 0.8472, + "step": 271030 + }, + { + "epoch": 23.46, + "learning_rate": 1.0914790950330838e-05, + "loss": 0.9309, + "step": 271040 + }, + { + "epoch": 23.46, + "learning_rate": 1.0913346239417493e-05, + "loss": 0.9063, + "step": 271050 + }, + { + "epoch": 23.46, + "learning_rate": 1.0911901528504147e-05, + "loss": 0.9081, + "step": 271060 + }, + { + "epoch": 23.46, + "learning_rate": 1.09104568175908e-05, + "loss": 0.8651, + "step": 271070 + }, + { + "epoch": 23.46, + "learning_rate": 1.0909012106677455e-05, + "loss": 0.9461, + "step": 271080 + }, + { + "epoch": 23.46, + "learning_rate": 1.0907567395764108e-05, + "loss": 0.8807, + "step": 271090 + }, + { + "epoch": 23.47, + "learning_rate": 1.0906122684850762e-05, + "loss": 0.9032, + "step": 271100 + }, + { + "epoch": 23.47, + "learning_rate": 1.0904677973937417e-05, + "loss": 0.9239, + "step": 271110 + }, + { + "epoch": 23.47, + "learning_rate": 1.0903233263024069e-05, + "loss": 0.9028, + "step": 271120 + }, + { + "epoch": 23.47, + "learning_rate": 1.0901788552110722e-05, + "loss": 0.9485, + "step": 271130 + }, + { + "epoch": 23.47, + "learning_rate": 1.0900343841197377e-05, + "loss": 0.9487, + "step": 271140 + }, + { + "epoch": 23.47, + "learning_rate": 1.089889913028403e-05, + "loss": 0.9145, + "step": 271150 + }, + { + "epoch": 23.47, + "learning_rate": 1.0897454419370684e-05, + "loss": 0.9604, + "step": 271160 + }, + { + "epoch": 23.47, + "learning_rate": 1.0896009708457339e-05, + "loss": 0.9264, + "step": 271170 + }, + { + "epoch": 23.47, + "learning_rate": 1.0894564997543992e-05, + "loss": 0.972, + "step": 271180 + }, + { + "epoch": 23.47, + "learning_rate": 1.0893120286630646e-05, + "loss": 0.8883, + "step": 271190 + }, + { + "epoch": 23.47, + "learning_rate": 1.08916755757173e-05, + "loss": 0.9794, + "step": 271200 + }, + { + "epoch": 23.48, + "learning_rate": 1.0890230864803953e-05, + "loss": 0.9134, + "step": 271210 + }, + { + "epoch": 23.48, + "learning_rate": 1.0888786153890606e-05, + "loss": 0.9117, + "step": 271220 + }, + { + "epoch": 23.48, + "learning_rate": 1.0887341442977261e-05, + "loss": 0.9465, + "step": 271230 + }, + { + "epoch": 23.48, + "learning_rate": 1.0885896732063914e-05, + "loss": 0.9119, + "step": 271240 + }, + { + "epoch": 23.48, + "learning_rate": 1.0884452021150568e-05, + "loss": 0.9635, + "step": 271250 + }, + { + "epoch": 23.48, + "learning_rate": 1.0883007310237223e-05, + "loss": 0.9457, + "step": 271260 + }, + { + "epoch": 23.48, + "learning_rate": 1.0881562599323876e-05, + "loss": 0.9208, + "step": 271270 + }, + { + "epoch": 23.48, + "learning_rate": 1.088011788841053e-05, + "loss": 0.9598, + "step": 271280 + }, + { + "epoch": 23.48, + "learning_rate": 1.0878673177497185e-05, + "loss": 0.9554, + "step": 271290 + }, + { + "epoch": 23.48, + "learning_rate": 1.0877228466583836e-05, + "loss": 0.9366, + "step": 271300 + }, + { + "epoch": 23.48, + "learning_rate": 1.087578375567049e-05, + "loss": 0.8875, + "step": 271310 + }, + { + "epoch": 23.48, + "learning_rate": 1.0874339044757145e-05, + "loss": 0.897, + "step": 271320 + }, + { + "epoch": 23.49, + "learning_rate": 1.0872894333843798e-05, + "loss": 0.9106, + "step": 271330 + }, + { + "epoch": 23.49, + "learning_rate": 1.0871449622930452e-05, + "loss": 0.9278, + "step": 271340 + }, + { + "epoch": 23.49, + "learning_rate": 1.0870004912017107e-05, + "loss": 0.9457, + "step": 271350 + }, + { + "epoch": 23.49, + "learning_rate": 1.086856020110376e-05, + "loss": 0.9055, + "step": 271360 + }, + { + "epoch": 23.49, + "learning_rate": 1.0867115490190414e-05, + "loss": 0.9047, + "step": 271370 + }, + { + "epoch": 23.49, + "learning_rate": 1.0865670779277067e-05, + "loss": 0.8765, + "step": 271380 + }, + { + "epoch": 23.49, + "learning_rate": 1.086422606836372e-05, + "loss": 0.9581, + "step": 271390 + }, + { + "epoch": 23.49, + "learning_rate": 1.0862781357450374e-05, + "loss": 0.8721, + "step": 271400 + }, + { + "epoch": 23.49, + "learning_rate": 1.0861336646537027e-05, + "loss": 0.9205, + "step": 271410 + }, + { + "epoch": 23.49, + "learning_rate": 1.0859891935623682e-05, + "loss": 0.879, + "step": 271420 + }, + { + "epoch": 23.49, + "learning_rate": 1.0858447224710336e-05, + "loss": 0.9353, + "step": 271430 + }, + { + "epoch": 23.5, + "learning_rate": 1.0857002513796989e-05, + "loss": 0.909, + "step": 271440 + }, + { + "epoch": 23.5, + "learning_rate": 1.0855557802883644e-05, + "loss": 0.9152, + "step": 271450 + }, + { + "epoch": 23.5, + "learning_rate": 1.0854113091970298e-05, + "loss": 0.8846, + "step": 271460 + }, + { + "epoch": 23.5, + "learning_rate": 1.0852668381056951e-05, + "loss": 0.8846, + "step": 271470 + }, + { + "epoch": 23.5, + "learning_rate": 1.0851223670143604e-05, + "loss": 0.9217, + "step": 271480 + }, + { + "epoch": 23.5, + "learning_rate": 1.0849778959230258e-05, + "loss": 0.8904, + "step": 271490 + }, + { + "epoch": 23.5, + "learning_rate": 1.0848334248316911e-05, + "loss": 0.9381, + "step": 271500 + }, + { + "epoch": 23.5, + "learning_rate": 1.0846889537403566e-05, + "loss": 0.928, + "step": 271510 + }, + { + "epoch": 23.5, + "learning_rate": 1.084544482649022e-05, + "loss": 0.9053, + "step": 271520 + }, + { + "epoch": 23.5, + "learning_rate": 1.0844000115576873e-05, + "loss": 0.9247, + "step": 271530 + }, + { + "epoch": 23.5, + "learning_rate": 1.0842555404663528e-05, + "loss": 0.9387, + "step": 271540 + }, + { + "epoch": 23.5, + "learning_rate": 1.0841110693750181e-05, + "loss": 0.9452, + "step": 271550 + }, + { + "epoch": 23.51, + "learning_rate": 1.0839665982836835e-05, + "loss": 0.95, + "step": 271560 + }, + { + "epoch": 23.51, + "learning_rate": 1.0838221271923488e-05, + "loss": 0.9106, + "step": 271570 + }, + { + "epoch": 23.51, + "learning_rate": 1.0836776561010142e-05, + "loss": 0.9686, + "step": 271580 + }, + { + "epoch": 23.51, + "learning_rate": 1.0835331850096795e-05, + "loss": 0.9498, + "step": 271590 + }, + { + "epoch": 23.51, + "learning_rate": 1.083388713918345e-05, + "loss": 0.9545, + "step": 271600 + }, + { + "epoch": 23.51, + "learning_rate": 1.0832442428270104e-05, + "loss": 0.9334, + "step": 271610 + }, + { + "epoch": 23.51, + "learning_rate": 1.0830997717356757e-05, + "loss": 0.9045, + "step": 271620 + }, + { + "epoch": 23.51, + "learning_rate": 1.0829553006443412e-05, + "loss": 0.922, + "step": 271630 + }, + { + "epoch": 23.51, + "learning_rate": 1.0828108295530065e-05, + "loss": 0.9538, + "step": 271640 + }, + { + "epoch": 23.51, + "learning_rate": 1.0826663584616719e-05, + "loss": 0.8896, + "step": 271650 + }, + { + "epoch": 23.51, + "learning_rate": 1.0825218873703372e-05, + "loss": 0.9298, + "step": 271660 + }, + { + "epoch": 23.52, + "learning_rate": 1.0823774162790026e-05, + "loss": 0.9066, + "step": 271670 + }, + { + "epoch": 23.52, + "learning_rate": 1.0822329451876679e-05, + "loss": 0.89, + "step": 271680 + }, + { + "epoch": 23.52, + "learning_rate": 1.0820884740963334e-05, + "loss": 0.9235, + "step": 271690 + }, + { + "epoch": 23.52, + "learning_rate": 1.0819440030049988e-05, + "loss": 0.9364, + "step": 271700 + }, + { + "epoch": 23.52, + "learning_rate": 1.0817995319136641e-05, + "loss": 0.877, + "step": 271710 + }, + { + "epoch": 23.52, + "learning_rate": 1.0816550608223296e-05, + "loss": 0.9486, + "step": 271720 + }, + { + "epoch": 23.52, + "learning_rate": 1.081510589730995e-05, + "loss": 0.8723, + "step": 271730 + }, + { + "epoch": 23.52, + "learning_rate": 1.0813661186396603e-05, + "loss": 0.9406, + "step": 271740 + }, + { + "epoch": 23.52, + "learning_rate": 1.0812216475483256e-05, + "loss": 0.9226, + "step": 271750 + }, + { + "epoch": 23.52, + "learning_rate": 1.081077176456991e-05, + "loss": 0.9492, + "step": 271760 + }, + { + "epoch": 23.52, + "learning_rate": 1.0809327053656563e-05, + "loss": 0.8387, + "step": 271770 + }, + { + "epoch": 23.52, + "learning_rate": 1.0807882342743218e-05, + "loss": 0.9698, + "step": 271780 + }, + { + "epoch": 23.53, + "learning_rate": 1.0806437631829871e-05, + "loss": 0.9027, + "step": 271790 + }, + { + "epoch": 23.53, + "learning_rate": 1.0804992920916525e-05, + "loss": 0.9331, + "step": 271800 + }, + { + "epoch": 23.53, + "learning_rate": 1.080354821000318e-05, + "loss": 0.889, + "step": 271810 + }, + { + "epoch": 23.53, + "learning_rate": 1.0802103499089833e-05, + "loss": 0.892, + "step": 271820 + }, + { + "epoch": 23.53, + "learning_rate": 1.0800658788176487e-05, + "loss": 0.9619, + "step": 271830 + }, + { + "epoch": 23.53, + "learning_rate": 1.079921407726314e-05, + "loss": 0.9345, + "step": 271840 + }, + { + "epoch": 23.53, + "learning_rate": 1.0797769366349794e-05, + "loss": 0.9551, + "step": 271850 + }, + { + "epoch": 23.53, + "learning_rate": 1.0796324655436447e-05, + "loss": 0.8787, + "step": 271860 + }, + { + "epoch": 23.53, + "learning_rate": 1.07948799445231e-05, + "loss": 0.8885, + "step": 271870 + }, + { + "epoch": 23.53, + "learning_rate": 1.0793435233609755e-05, + "loss": 0.8922, + "step": 271880 + }, + { + "epoch": 23.53, + "learning_rate": 1.0791990522696409e-05, + "loss": 0.905, + "step": 271890 + }, + { + "epoch": 23.53, + "learning_rate": 1.0790545811783062e-05, + "loss": 0.9265, + "step": 271900 + }, + { + "epoch": 23.54, + "learning_rate": 1.0789101100869717e-05, + "loss": 0.9016, + "step": 271910 + }, + { + "epoch": 23.54, + "learning_rate": 1.078765638995637e-05, + "loss": 0.9231, + "step": 271920 + }, + { + "epoch": 23.54, + "learning_rate": 1.0786211679043024e-05, + "loss": 0.8502, + "step": 271930 + }, + { + "epoch": 23.54, + "learning_rate": 1.0784766968129677e-05, + "loss": 0.9606, + "step": 271940 + }, + { + "epoch": 23.54, + "learning_rate": 1.078332225721633e-05, + "loss": 0.8723, + "step": 271950 + }, + { + "epoch": 23.54, + "learning_rate": 1.0781877546302984e-05, + "loss": 0.9534, + "step": 271960 + }, + { + "epoch": 23.54, + "learning_rate": 1.078043283538964e-05, + "loss": 0.8495, + "step": 271970 + }, + { + "epoch": 23.54, + "learning_rate": 1.0778988124476293e-05, + "loss": 0.8973, + "step": 271980 + }, + { + "epoch": 23.54, + "learning_rate": 1.0777543413562946e-05, + "loss": 0.9712, + "step": 271990 + }, + { + "epoch": 23.54, + "learning_rate": 1.0776098702649601e-05, + "loss": 0.9176, + "step": 272000 + }, + { + "epoch": 23.54, + "learning_rate": 1.0774653991736255e-05, + "loss": 0.9114, + "step": 272010 + }, + { + "epoch": 23.55, + "learning_rate": 1.0773209280822908e-05, + "loss": 0.9298, + "step": 272020 + }, + { + "epoch": 23.55, + "learning_rate": 1.0771764569909561e-05, + "loss": 0.9463, + "step": 272030 + }, + { + "epoch": 23.55, + "learning_rate": 1.0770319858996215e-05, + "loss": 0.8913, + "step": 272040 + }, + { + "epoch": 23.55, + "learning_rate": 1.0768875148082868e-05, + "loss": 0.9574, + "step": 272050 + }, + { + "epoch": 23.55, + "learning_rate": 1.0767430437169523e-05, + "loss": 0.9398, + "step": 272060 + }, + { + "epoch": 23.55, + "learning_rate": 1.0765985726256177e-05, + "loss": 0.9478, + "step": 272070 + }, + { + "epoch": 23.55, + "learning_rate": 1.076454101534283e-05, + "loss": 0.9523, + "step": 272080 + }, + { + "epoch": 23.55, + "learning_rate": 1.0763096304429485e-05, + "loss": 0.9569, + "step": 272090 + }, + { + "epoch": 23.55, + "learning_rate": 1.0761651593516139e-05, + "loss": 0.9246, + "step": 272100 + }, + { + "epoch": 23.55, + "learning_rate": 1.0760206882602792e-05, + "loss": 0.8805, + "step": 272110 + }, + { + "epoch": 23.55, + "learning_rate": 1.0758762171689445e-05, + "loss": 0.875, + "step": 272120 + }, + { + "epoch": 23.55, + "learning_rate": 1.0757317460776099e-05, + "loss": 0.9549, + "step": 272130 + }, + { + "epoch": 23.56, + "learning_rate": 1.0755872749862752e-05, + "loss": 0.9371, + "step": 272140 + }, + { + "epoch": 23.56, + "learning_rate": 1.0754428038949407e-05, + "loss": 0.9438, + "step": 272150 + }, + { + "epoch": 23.56, + "learning_rate": 1.075298332803606e-05, + "loss": 0.91, + "step": 272160 + }, + { + "epoch": 23.56, + "learning_rate": 1.0751538617122714e-05, + "loss": 0.8434, + "step": 272170 + }, + { + "epoch": 23.56, + "learning_rate": 1.0750093906209369e-05, + "loss": 0.869, + "step": 272180 + }, + { + "epoch": 23.56, + "learning_rate": 1.0748649195296022e-05, + "loss": 0.9159, + "step": 272190 + }, + { + "epoch": 23.56, + "learning_rate": 1.0747204484382676e-05, + "loss": 0.897, + "step": 272200 + }, + { + "epoch": 23.56, + "learning_rate": 1.074575977346933e-05, + "loss": 0.913, + "step": 272210 + }, + { + "epoch": 23.56, + "learning_rate": 1.0744315062555983e-05, + "loss": 0.9463, + "step": 272220 + }, + { + "epoch": 23.56, + "learning_rate": 1.0742870351642636e-05, + "loss": 0.9187, + "step": 272230 + }, + { + "epoch": 23.56, + "learning_rate": 1.0741425640729291e-05, + "loss": 0.9061, + "step": 272240 + }, + { + "epoch": 23.57, + "learning_rate": 1.0739980929815945e-05, + "loss": 0.9314, + "step": 272250 + }, + { + "epoch": 23.57, + "learning_rate": 1.0738536218902598e-05, + "loss": 0.9339, + "step": 272260 + }, + { + "epoch": 23.57, + "learning_rate": 1.0737091507989251e-05, + "loss": 0.8734, + "step": 272270 + }, + { + "epoch": 23.57, + "learning_rate": 1.0735646797075906e-05, + "loss": 0.9158, + "step": 272280 + }, + { + "epoch": 23.57, + "learning_rate": 1.073420208616256e-05, + "loss": 0.9031, + "step": 272290 + }, + { + "epoch": 23.57, + "learning_rate": 1.0732757375249213e-05, + "loss": 0.9745, + "step": 272300 + }, + { + "epoch": 23.57, + "learning_rate": 1.0731312664335867e-05, + "loss": 0.9305, + "step": 272310 + }, + { + "epoch": 23.57, + "learning_rate": 1.072986795342252e-05, + "loss": 0.9132, + "step": 272320 + }, + { + "epoch": 23.57, + "learning_rate": 1.0728423242509173e-05, + "loss": 0.8841, + "step": 272330 + }, + { + "epoch": 23.57, + "learning_rate": 1.0726978531595828e-05, + "loss": 0.8666, + "step": 272340 + }, + { + "epoch": 23.57, + "learning_rate": 1.0725533820682482e-05, + "loss": 0.8398, + "step": 272350 + }, + { + "epoch": 23.57, + "learning_rate": 1.0724089109769135e-05, + "loss": 0.9131, + "step": 272360 + }, + { + "epoch": 23.58, + "learning_rate": 1.072264439885579e-05, + "loss": 0.9289, + "step": 272370 + }, + { + "epoch": 23.58, + "learning_rate": 1.0721199687942444e-05, + "loss": 0.906, + "step": 272380 + }, + { + "epoch": 23.58, + "learning_rate": 1.0719754977029097e-05, + "loss": 0.9623, + "step": 272390 + }, + { + "epoch": 23.58, + "learning_rate": 1.071831026611575e-05, + "loss": 0.9125, + "step": 272400 + }, + { + "epoch": 23.58, + "learning_rate": 1.0716865555202404e-05, + "loss": 0.8744, + "step": 272410 + }, + { + "epoch": 23.58, + "learning_rate": 1.0715420844289057e-05, + "loss": 0.9299, + "step": 272420 + }, + { + "epoch": 23.58, + "learning_rate": 1.0713976133375712e-05, + "loss": 0.9388, + "step": 272430 + }, + { + "epoch": 23.58, + "learning_rate": 1.0712531422462366e-05, + "loss": 0.9371, + "step": 272440 + }, + { + "epoch": 23.58, + "learning_rate": 1.071108671154902e-05, + "loss": 0.9023, + "step": 272450 + }, + { + "epoch": 23.58, + "learning_rate": 1.0709642000635674e-05, + "loss": 0.9245, + "step": 272460 + }, + { + "epoch": 23.58, + "learning_rate": 1.0708197289722328e-05, + "loss": 0.9144, + "step": 272470 + }, + { + "epoch": 23.59, + "learning_rate": 1.0706752578808981e-05, + "loss": 0.949, + "step": 272480 + }, + { + "epoch": 23.59, + "learning_rate": 1.0705307867895635e-05, + "loss": 0.89, + "step": 272490 + }, + { + "epoch": 23.59, + "learning_rate": 1.0703863156982288e-05, + "loss": 0.9003, + "step": 272500 + }, + { + "epoch": 23.59, + "learning_rate": 1.0702418446068941e-05, + "loss": 0.9084, + "step": 272510 + }, + { + "epoch": 23.59, + "learning_rate": 1.0700973735155596e-05, + "loss": 0.9082, + "step": 272520 + }, + { + "epoch": 23.59, + "learning_rate": 1.069952902424225e-05, + "loss": 0.9462, + "step": 272530 + }, + { + "epoch": 23.59, + "learning_rate": 1.0698084313328903e-05, + "loss": 0.9005, + "step": 272540 + }, + { + "epoch": 23.59, + "learning_rate": 1.0696639602415558e-05, + "loss": 0.9246, + "step": 272550 + }, + { + "epoch": 23.59, + "learning_rate": 1.0695194891502212e-05, + "loss": 0.9115, + "step": 272560 + }, + { + "epoch": 23.59, + "learning_rate": 1.0693750180588865e-05, + "loss": 0.9649, + "step": 272570 + }, + { + "epoch": 23.59, + "learning_rate": 1.0692305469675518e-05, + "loss": 0.9017, + "step": 272580 + }, + { + "epoch": 23.59, + "learning_rate": 1.0690860758762172e-05, + "loss": 0.8775, + "step": 272590 + }, + { + "epoch": 23.6, + "learning_rate": 1.0689416047848825e-05, + "loss": 0.9032, + "step": 272600 + }, + { + "epoch": 23.6, + "learning_rate": 1.068797133693548e-05, + "loss": 0.9142, + "step": 272610 + }, + { + "epoch": 23.6, + "learning_rate": 1.0686526626022134e-05, + "loss": 0.9332, + "step": 272620 + }, + { + "epoch": 23.6, + "learning_rate": 1.0685081915108787e-05, + "loss": 0.9226, + "step": 272630 + }, + { + "epoch": 23.6, + "learning_rate": 1.0683637204195442e-05, + "loss": 0.8597, + "step": 272640 + }, + { + "epoch": 23.6, + "learning_rate": 1.0682192493282096e-05, + "loss": 0.9318, + "step": 272650 + }, + { + "epoch": 23.6, + "learning_rate": 1.0680747782368749e-05, + "loss": 0.9647, + "step": 272660 + }, + { + "epoch": 23.6, + "learning_rate": 1.0679303071455402e-05, + "loss": 0.9133, + "step": 272670 + }, + { + "epoch": 23.6, + "learning_rate": 1.0677858360542056e-05, + "loss": 0.8954, + "step": 272680 + }, + { + "epoch": 23.6, + "learning_rate": 1.067641364962871e-05, + "loss": 0.9361, + "step": 272690 + }, + { + "epoch": 23.6, + "learning_rate": 1.0674968938715364e-05, + "loss": 0.9204, + "step": 272700 + }, + { + "epoch": 23.61, + "learning_rate": 1.0673524227802018e-05, + "loss": 0.9413, + "step": 272710 + }, + { + "epoch": 23.61, + "learning_rate": 1.0672079516888671e-05, + "loss": 0.901, + "step": 272720 + }, + { + "epoch": 23.61, + "learning_rate": 1.0670634805975324e-05, + "loss": 0.9107, + "step": 272730 + }, + { + "epoch": 23.61, + "learning_rate": 1.066919009506198e-05, + "loss": 0.9593, + "step": 272740 + }, + { + "epoch": 23.61, + "learning_rate": 1.0667745384148633e-05, + "loss": 0.9671, + "step": 272750 + }, + { + "epoch": 23.61, + "learning_rate": 1.0666300673235285e-05, + "loss": 0.9025, + "step": 272760 + }, + { + "epoch": 23.61, + "learning_rate": 1.066485596232194e-05, + "loss": 0.9188, + "step": 272770 + }, + { + "epoch": 23.61, + "learning_rate": 1.0663411251408593e-05, + "loss": 0.9145, + "step": 272780 + }, + { + "epoch": 23.61, + "learning_rate": 1.0661966540495247e-05, + "loss": 0.9451, + "step": 272790 + }, + { + "epoch": 23.61, + "learning_rate": 1.0660521829581902e-05, + "loss": 0.9253, + "step": 272800 + }, + { + "epoch": 23.61, + "learning_rate": 1.0659077118668555e-05, + "loss": 0.9552, + "step": 272810 + }, + { + "epoch": 23.61, + "learning_rate": 1.0657632407755208e-05, + "loss": 0.9521, + "step": 272820 + }, + { + "epoch": 23.62, + "learning_rate": 1.0656187696841863e-05, + "loss": 0.9014, + "step": 272830 + }, + { + "epoch": 23.62, + "learning_rate": 1.0654742985928517e-05, + "loss": 0.9538, + "step": 272840 + }, + { + "epoch": 23.62, + "learning_rate": 1.0653298275015169e-05, + "loss": 0.9276, + "step": 272850 + }, + { + "epoch": 23.62, + "learning_rate": 1.0651853564101824e-05, + "loss": 0.883, + "step": 272860 + }, + { + "epoch": 23.62, + "learning_rate": 1.0650408853188477e-05, + "loss": 0.9207, + "step": 272870 + }, + { + "epoch": 23.62, + "learning_rate": 1.064896414227513e-05, + "loss": 0.8806, + "step": 272880 + }, + { + "epoch": 23.62, + "learning_rate": 1.0647519431361786e-05, + "loss": 0.8987, + "step": 272890 + }, + { + "epoch": 23.62, + "learning_rate": 1.0646074720448439e-05, + "loss": 0.9624, + "step": 272900 + }, + { + "epoch": 23.62, + "learning_rate": 1.0644630009535092e-05, + "loss": 0.9041, + "step": 272910 + }, + { + "epoch": 23.62, + "learning_rate": 1.0643185298621747e-05, + "loss": 0.9083, + "step": 272920 + }, + { + "epoch": 23.62, + "learning_rate": 1.06417405877084e-05, + "loss": 0.9013, + "step": 272930 + }, + { + "epoch": 23.63, + "learning_rate": 1.0640295876795053e-05, + "loss": 0.981, + "step": 272940 + }, + { + "epoch": 23.63, + "learning_rate": 1.0638851165881708e-05, + "loss": 0.9113, + "step": 272950 + }, + { + "epoch": 23.63, + "learning_rate": 1.0637406454968361e-05, + "loss": 0.8892, + "step": 272960 + }, + { + "epoch": 23.63, + "learning_rate": 1.0635961744055014e-05, + "loss": 0.9177, + "step": 272970 + }, + { + "epoch": 23.63, + "learning_rate": 1.063451703314167e-05, + "loss": 0.9515, + "step": 272980 + }, + { + "epoch": 23.63, + "learning_rate": 1.0633072322228323e-05, + "loss": 0.9071, + "step": 272990 + }, + { + "epoch": 23.63, + "learning_rate": 1.0631627611314976e-05, + "loss": 0.9296, + "step": 273000 + }, + { + "epoch": 23.63, + "learning_rate": 1.0630182900401631e-05, + "loss": 0.9489, + "step": 273010 + }, + { + "epoch": 23.63, + "learning_rate": 1.0628738189488285e-05, + "loss": 0.866, + "step": 273020 + }, + { + "epoch": 23.63, + "learning_rate": 1.0627293478574936e-05, + "loss": 0.9243, + "step": 273030 + }, + { + "epoch": 23.63, + "learning_rate": 1.0625848767661592e-05, + "loss": 0.912, + "step": 273040 + }, + { + "epoch": 23.63, + "learning_rate": 1.0624404056748245e-05, + "loss": 0.9623, + "step": 273050 + }, + { + "epoch": 23.64, + "learning_rate": 1.0622959345834898e-05, + "loss": 0.9399, + "step": 273060 + }, + { + "epoch": 23.64, + "learning_rate": 1.0621514634921553e-05, + "loss": 0.9352, + "step": 273070 + }, + { + "epoch": 23.64, + "learning_rate": 1.0620069924008207e-05, + "loss": 0.948, + "step": 273080 + }, + { + "epoch": 23.64, + "learning_rate": 1.061862521309486e-05, + "loss": 0.9151, + "step": 273090 + }, + { + "epoch": 23.64, + "learning_rate": 1.0617180502181515e-05, + "loss": 0.8501, + "step": 273100 + }, + { + "epoch": 23.64, + "learning_rate": 1.0615735791268169e-05, + "loss": 0.8933, + "step": 273110 + }, + { + "epoch": 23.64, + "learning_rate": 1.061429108035482e-05, + "loss": 0.9253, + "step": 273120 + }, + { + "epoch": 23.64, + "learning_rate": 1.0612846369441475e-05, + "loss": 0.9773, + "step": 273130 + }, + { + "epoch": 23.64, + "learning_rate": 1.0611401658528129e-05, + "loss": 0.9064, + "step": 273140 + }, + { + "epoch": 23.64, + "learning_rate": 1.0609956947614782e-05, + "loss": 0.9402, + "step": 273150 + }, + { + "epoch": 23.64, + "learning_rate": 1.0608512236701437e-05, + "loss": 0.9085, + "step": 273160 + }, + { + "epoch": 23.64, + "learning_rate": 1.060706752578809e-05, + "loss": 0.8899, + "step": 273170 + }, + { + "epoch": 23.65, + "learning_rate": 1.0605622814874744e-05, + "loss": 0.8787, + "step": 273180 + }, + { + "epoch": 23.65, + "learning_rate": 1.0604178103961398e-05, + "loss": 0.9829, + "step": 273190 + }, + { + "epoch": 23.65, + "learning_rate": 1.0602733393048053e-05, + "loss": 0.9222, + "step": 273200 + }, + { + "epoch": 23.65, + "learning_rate": 1.0601288682134704e-05, + "loss": 0.9616, + "step": 273210 + }, + { + "epoch": 23.65, + "learning_rate": 1.0599843971221358e-05, + "loss": 0.9461, + "step": 273220 + }, + { + "epoch": 23.65, + "learning_rate": 1.0598399260308013e-05, + "loss": 0.9457, + "step": 273230 + }, + { + "epoch": 23.65, + "learning_rate": 1.0596954549394666e-05, + "loss": 0.912, + "step": 273240 + }, + { + "epoch": 23.65, + "learning_rate": 1.059550983848132e-05, + "loss": 1.0135, + "step": 273250 + }, + { + "epoch": 23.65, + "learning_rate": 1.0594065127567975e-05, + "loss": 0.8917, + "step": 273260 + }, + { + "epoch": 23.65, + "learning_rate": 1.0592620416654628e-05, + "loss": 0.9374, + "step": 273270 + }, + { + "epoch": 23.65, + "learning_rate": 1.0591175705741282e-05, + "loss": 0.9143, + "step": 273280 + }, + { + "epoch": 23.66, + "learning_rate": 1.0589730994827937e-05, + "loss": 0.8753, + "step": 273290 + }, + { + "epoch": 23.66, + "learning_rate": 1.0588286283914588e-05, + "loss": 0.9532, + "step": 273300 + }, + { + "epoch": 23.66, + "learning_rate": 1.0586841573001242e-05, + "loss": 0.9404, + "step": 273310 + }, + { + "epoch": 23.66, + "learning_rate": 1.0585396862087897e-05, + "loss": 0.9109, + "step": 273320 + }, + { + "epoch": 23.66, + "learning_rate": 1.058395215117455e-05, + "loss": 0.9476, + "step": 273330 + }, + { + "epoch": 23.66, + "learning_rate": 1.0582507440261204e-05, + "loss": 0.9201, + "step": 273340 + }, + { + "epoch": 23.66, + "learning_rate": 1.0581062729347859e-05, + "loss": 0.9807, + "step": 273350 + }, + { + "epoch": 23.66, + "learning_rate": 1.0579618018434512e-05, + "loss": 0.9222, + "step": 273360 + }, + { + "epoch": 23.66, + "learning_rate": 1.0578173307521165e-05, + "loss": 0.8783, + "step": 273370 + }, + { + "epoch": 23.66, + "learning_rate": 1.057672859660782e-05, + "loss": 0.8884, + "step": 273380 + }, + { + "epoch": 23.66, + "learning_rate": 1.0575283885694472e-05, + "loss": 0.9199, + "step": 273390 + }, + { + "epoch": 23.66, + "learning_rate": 1.0573839174781126e-05, + "loss": 0.9045, + "step": 273400 + }, + { + "epoch": 23.67, + "learning_rate": 1.057239446386778e-05, + "loss": 0.9692, + "step": 273410 + }, + { + "epoch": 23.67, + "learning_rate": 1.0570949752954434e-05, + "loss": 0.9073, + "step": 273420 + }, + { + "epoch": 23.67, + "learning_rate": 1.0569505042041088e-05, + "loss": 0.9079, + "step": 273430 + }, + { + "epoch": 23.67, + "learning_rate": 1.0568060331127743e-05, + "loss": 0.8901, + "step": 273440 + }, + { + "epoch": 23.67, + "learning_rate": 1.0566615620214396e-05, + "loss": 0.8923, + "step": 273450 + }, + { + "epoch": 23.67, + "learning_rate": 1.056517090930105e-05, + "loss": 0.8801, + "step": 273460 + }, + { + "epoch": 23.67, + "learning_rate": 1.0563726198387704e-05, + "loss": 0.9756, + "step": 273470 + }, + { + "epoch": 23.67, + "learning_rate": 1.0562281487474356e-05, + "loss": 0.9341, + "step": 273480 + }, + { + "epoch": 23.67, + "learning_rate": 1.056083677656101e-05, + "loss": 0.9518, + "step": 273490 + }, + { + "epoch": 23.67, + "learning_rate": 1.0559392065647665e-05, + "loss": 0.897, + "step": 273500 + }, + { + "epoch": 23.67, + "learning_rate": 1.0557947354734318e-05, + "loss": 0.9357, + "step": 273510 + }, + { + "epoch": 23.68, + "learning_rate": 1.0556502643820971e-05, + "loss": 0.8842, + "step": 273520 + }, + { + "epoch": 23.68, + "learning_rate": 1.0555057932907627e-05, + "loss": 0.9315, + "step": 273530 + }, + { + "epoch": 23.68, + "learning_rate": 1.055361322199428e-05, + "loss": 0.9058, + "step": 273540 + }, + { + "epoch": 23.68, + "learning_rate": 1.0552168511080933e-05, + "loss": 0.9324, + "step": 273550 + }, + { + "epoch": 23.68, + "learning_rate": 1.0550723800167588e-05, + "loss": 0.9032, + "step": 273560 + }, + { + "epoch": 23.68, + "learning_rate": 1.054927908925424e-05, + "loss": 0.8559, + "step": 273570 + }, + { + "epoch": 23.68, + "learning_rate": 1.0547834378340894e-05, + "loss": 1.0206, + "step": 273580 + }, + { + "epoch": 23.68, + "learning_rate": 1.0546389667427549e-05, + "loss": 0.8718, + "step": 273590 + }, + { + "epoch": 23.68, + "learning_rate": 1.0544944956514202e-05, + "loss": 0.9008, + "step": 273600 + }, + { + "epoch": 23.68, + "learning_rate": 1.0543500245600855e-05, + "loss": 0.9069, + "step": 273610 + }, + { + "epoch": 23.68, + "learning_rate": 1.054205553468751e-05, + "loss": 0.9097, + "step": 273620 + }, + { + "epoch": 23.68, + "learning_rate": 1.0540610823774164e-05, + "loss": 0.8775, + "step": 273630 + }, + { + "epoch": 23.69, + "learning_rate": 1.0539166112860817e-05, + "loss": 0.9517, + "step": 273640 + }, + { + "epoch": 23.69, + "learning_rate": 1.053772140194747e-05, + "loss": 0.9638, + "step": 273650 + }, + { + "epoch": 23.69, + "learning_rate": 1.0536276691034124e-05, + "loss": 0.918, + "step": 273660 + }, + { + "epoch": 23.69, + "learning_rate": 1.0534831980120777e-05, + "loss": 0.9657, + "step": 273670 + }, + { + "epoch": 23.69, + "learning_rate": 1.0533387269207431e-05, + "loss": 0.9677, + "step": 273680 + }, + { + "epoch": 23.69, + "learning_rate": 1.0531942558294086e-05, + "loss": 0.9295, + "step": 273690 + }, + { + "epoch": 23.69, + "learning_rate": 1.053049784738074e-05, + "loss": 0.9373, + "step": 273700 + }, + { + "epoch": 23.69, + "learning_rate": 1.0529053136467393e-05, + "loss": 0.9196, + "step": 273710 + }, + { + "epoch": 23.69, + "learning_rate": 1.0527608425554048e-05, + "loss": 0.9143, + "step": 273720 + }, + { + "epoch": 23.69, + "learning_rate": 1.0526163714640701e-05, + "loss": 0.9491, + "step": 273730 + }, + { + "epoch": 23.69, + "learning_rate": 1.0524719003727355e-05, + "loss": 0.9305, + "step": 273740 + }, + { + "epoch": 23.7, + "learning_rate": 1.0523274292814008e-05, + "loss": 0.9331, + "step": 273750 + }, + { + "epoch": 23.7, + "learning_rate": 1.0521829581900661e-05, + "loss": 0.8841, + "step": 273760 + }, + { + "epoch": 23.7, + "learning_rate": 1.0520384870987315e-05, + "loss": 0.9019, + "step": 273770 + }, + { + "epoch": 23.7, + "learning_rate": 1.051894016007397e-05, + "loss": 0.9282, + "step": 273780 + }, + { + "epoch": 23.7, + "learning_rate": 1.0517495449160623e-05, + "loss": 0.9684, + "step": 273790 + }, + { + "epoch": 23.7, + "learning_rate": 1.0516050738247277e-05, + "loss": 0.8919, + "step": 273800 + }, + { + "epoch": 23.7, + "learning_rate": 1.0514606027333932e-05, + "loss": 0.9497, + "step": 273810 + }, + { + "epoch": 23.7, + "learning_rate": 1.0513161316420585e-05, + "loss": 0.9523, + "step": 273820 + }, + { + "epoch": 23.7, + "learning_rate": 1.0511716605507239e-05, + "loss": 0.9241, + "step": 273830 + }, + { + "epoch": 23.7, + "learning_rate": 1.0510271894593892e-05, + "loss": 0.9785, + "step": 273840 + }, + { + "epoch": 23.7, + "learning_rate": 1.0508827183680545e-05, + "loss": 0.8744, + "step": 273850 + }, + { + "epoch": 23.7, + "learning_rate": 1.0507382472767199e-05, + "loss": 0.9036, + "step": 273860 + }, + { + "epoch": 23.71, + "learning_rate": 1.0505937761853854e-05, + "loss": 0.9377, + "step": 273870 + }, + { + "epoch": 23.71, + "learning_rate": 1.0504493050940507e-05, + "loss": 0.9231, + "step": 273880 + }, + { + "epoch": 23.71, + "learning_rate": 1.050304834002716e-05, + "loss": 0.9159, + "step": 273890 + }, + { + "epoch": 23.71, + "learning_rate": 1.0501603629113816e-05, + "loss": 0.9332, + "step": 273900 + }, + { + "epoch": 23.71, + "learning_rate": 1.0500158918200469e-05, + "loss": 0.9128, + "step": 273910 + }, + { + "epoch": 23.71, + "learning_rate": 1.0498714207287122e-05, + "loss": 0.9324, + "step": 273920 + }, + { + "epoch": 23.71, + "learning_rate": 1.0497269496373776e-05, + "loss": 0.9224, + "step": 273930 + }, + { + "epoch": 23.71, + "learning_rate": 1.049582478546043e-05, + "loss": 0.9297, + "step": 273940 + }, + { + "epoch": 23.71, + "learning_rate": 1.0494380074547083e-05, + "loss": 0.9121, + "step": 273950 + }, + { + "epoch": 23.71, + "learning_rate": 1.0492935363633738e-05, + "loss": 0.881, + "step": 273960 + }, + { + "epoch": 23.71, + "learning_rate": 1.0491490652720391e-05, + "loss": 0.8618, + "step": 273970 + }, + { + "epoch": 23.72, + "learning_rate": 1.0490045941807045e-05, + "loss": 0.9457, + "step": 273980 + }, + { + "epoch": 23.72, + "learning_rate": 1.04886012308937e-05, + "loss": 0.9433, + "step": 273990 + }, + { + "epoch": 23.72, + "learning_rate": 1.0487156519980353e-05, + "loss": 0.943, + "step": 274000 + }, + { + "epoch": 23.72, + "learning_rate": 1.0485711809067006e-05, + "loss": 0.9293, + "step": 274010 + }, + { + "epoch": 23.72, + "learning_rate": 1.048426709815366e-05, + "loss": 0.8908, + "step": 274020 + }, + { + "epoch": 23.72, + "learning_rate": 1.0482822387240313e-05, + "loss": 0.9486, + "step": 274030 + }, + { + "epoch": 23.72, + "learning_rate": 1.0481377676326967e-05, + "loss": 0.8873, + "step": 274040 + }, + { + "epoch": 23.72, + "learning_rate": 1.0479932965413622e-05, + "loss": 0.9019, + "step": 274050 + }, + { + "epoch": 23.72, + "learning_rate": 1.0478488254500275e-05, + "loss": 0.8876, + "step": 274060 + }, + { + "epoch": 23.72, + "learning_rate": 1.0477043543586929e-05, + "loss": 0.8978, + "step": 274070 + }, + { + "epoch": 23.72, + "learning_rate": 1.0475598832673582e-05, + "loss": 0.9116, + "step": 274080 + }, + { + "epoch": 23.72, + "learning_rate": 1.0474154121760237e-05, + "loss": 0.9321, + "step": 274090 + }, + { + "epoch": 23.73, + "learning_rate": 1.047270941084689e-05, + "loss": 0.9262, + "step": 274100 + }, + { + "epoch": 23.73, + "learning_rate": 1.0471264699933544e-05, + "loss": 0.9327, + "step": 274110 + }, + { + "epoch": 23.73, + "learning_rate": 1.0469819989020197e-05, + "loss": 0.8879, + "step": 274120 + }, + { + "epoch": 23.73, + "learning_rate": 1.046837527810685e-05, + "loss": 0.8919, + "step": 274130 + }, + { + "epoch": 23.73, + "learning_rate": 1.0466930567193504e-05, + "loss": 0.913, + "step": 274140 + }, + { + "epoch": 23.73, + "learning_rate": 1.0465485856280159e-05, + "loss": 0.9309, + "step": 274150 + }, + { + "epoch": 23.73, + "learning_rate": 1.0464041145366812e-05, + "loss": 0.9, + "step": 274160 + }, + { + "epoch": 23.73, + "learning_rate": 1.0462596434453466e-05, + "loss": 0.9059, + "step": 274170 + }, + { + "epoch": 23.73, + "learning_rate": 1.0461151723540121e-05, + "loss": 0.9402, + "step": 274180 + }, + { + "epoch": 23.73, + "learning_rate": 1.0459707012626774e-05, + "loss": 0.9235, + "step": 274190 + }, + { + "epoch": 23.73, + "learning_rate": 1.0458262301713428e-05, + "loss": 0.9587, + "step": 274200 + }, + { + "epoch": 23.73, + "learning_rate": 1.0456817590800081e-05, + "loss": 0.9627, + "step": 274210 + }, + { + "epoch": 23.74, + "learning_rate": 1.0455372879886735e-05, + "loss": 0.9224, + "step": 274220 + }, + { + "epoch": 23.74, + "learning_rate": 1.0453928168973388e-05, + "loss": 0.9208, + "step": 274230 + }, + { + "epoch": 23.74, + "learning_rate": 1.0452483458060043e-05, + "loss": 0.9592, + "step": 274240 + }, + { + "epoch": 23.74, + "learning_rate": 1.0451038747146696e-05, + "loss": 0.9489, + "step": 274250 + }, + { + "epoch": 23.74, + "learning_rate": 1.044959403623335e-05, + "loss": 0.9014, + "step": 274260 + }, + { + "epoch": 23.74, + "learning_rate": 1.0448149325320005e-05, + "loss": 0.8561, + "step": 274270 + }, + { + "epoch": 23.74, + "learning_rate": 1.0446704614406658e-05, + "loss": 0.9758, + "step": 274280 + }, + { + "epoch": 23.74, + "learning_rate": 1.0445259903493312e-05, + "loss": 0.9315, + "step": 274290 + }, + { + "epoch": 23.74, + "learning_rate": 1.0443815192579965e-05, + "loss": 0.9054, + "step": 274300 + }, + { + "epoch": 23.74, + "learning_rate": 1.0442370481666618e-05, + "loss": 0.9137, + "step": 274310 + }, + { + "epoch": 23.74, + "learning_rate": 1.0440925770753272e-05, + "loss": 0.8944, + "step": 274320 + }, + { + "epoch": 23.75, + "learning_rate": 1.0439481059839927e-05, + "loss": 0.9816, + "step": 274330 + }, + { + "epoch": 23.75, + "learning_rate": 1.043803634892658e-05, + "loss": 0.9183, + "step": 274340 + }, + { + "epoch": 23.75, + "learning_rate": 1.0436591638013234e-05, + "loss": 0.9387, + "step": 274350 + }, + { + "epoch": 23.75, + "learning_rate": 1.0435146927099889e-05, + "loss": 0.8693, + "step": 274360 + }, + { + "epoch": 23.75, + "learning_rate": 1.0433702216186542e-05, + "loss": 0.899, + "step": 274370 + }, + { + "epoch": 23.75, + "learning_rate": 1.0432257505273196e-05, + "loss": 0.8667, + "step": 274380 + }, + { + "epoch": 23.75, + "learning_rate": 1.0430812794359849e-05, + "loss": 0.9768, + "step": 274390 + }, + { + "epoch": 23.75, + "learning_rate": 1.0429368083446502e-05, + "loss": 0.866, + "step": 274400 + }, + { + "epoch": 23.75, + "learning_rate": 1.0427923372533156e-05, + "loss": 0.947, + "step": 274410 + }, + { + "epoch": 23.75, + "learning_rate": 1.0426478661619811e-05, + "loss": 0.9246, + "step": 274420 + }, + { + "epoch": 23.75, + "learning_rate": 1.0425033950706464e-05, + "loss": 1.0114, + "step": 274430 + }, + { + "epoch": 23.75, + "learning_rate": 1.0423589239793118e-05, + "loss": 0.9151, + "step": 274440 + }, + { + "epoch": 23.76, + "learning_rate": 1.0422144528879773e-05, + "loss": 0.9413, + "step": 274450 + }, + { + "epoch": 23.76, + "learning_rate": 1.0420699817966426e-05, + "loss": 0.9078, + "step": 274460 + }, + { + "epoch": 23.76, + "learning_rate": 1.041925510705308e-05, + "loss": 0.8909, + "step": 274470 + }, + { + "epoch": 23.76, + "learning_rate": 1.0417810396139733e-05, + "loss": 0.8815, + "step": 274480 + }, + { + "epoch": 23.76, + "learning_rate": 1.0416365685226386e-05, + "loss": 0.9561, + "step": 274490 + }, + { + "epoch": 23.76, + "learning_rate": 1.041492097431304e-05, + "loss": 0.9436, + "step": 274500 + }, + { + "epoch": 23.76, + "learning_rate": 1.0413476263399695e-05, + "loss": 0.935, + "step": 274510 + }, + { + "epoch": 23.76, + "learning_rate": 1.0412031552486348e-05, + "loss": 0.9497, + "step": 274520 + }, + { + "epoch": 23.76, + "learning_rate": 1.0410586841573002e-05, + "loss": 0.9767, + "step": 274530 + }, + { + "epoch": 23.76, + "learning_rate": 1.0409142130659655e-05, + "loss": 0.9332, + "step": 274540 + }, + { + "epoch": 23.76, + "learning_rate": 1.040769741974631e-05, + "loss": 0.887, + "step": 274550 + }, + { + "epoch": 23.77, + "learning_rate": 1.0406252708832963e-05, + "loss": 0.9303, + "step": 274560 + }, + { + "epoch": 23.77, + "learning_rate": 1.0404807997919617e-05, + "loss": 0.8966, + "step": 274570 + }, + { + "epoch": 23.77, + "learning_rate": 1.040336328700627e-05, + "loss": 0.8777, + "step": 274580 + }, + { + "epoch": 23.77, + "learning_rate": 1.0401918576092924e-05, + "loss": 0.922, + "step": 274590 + }, + { + "epoch": 23.77, + "learning_rate": 1.0400473865179577e-05, + "loss": 0.9197, + "step": 274600 + }, + { + "epoch": 23.77, + "learning_rate": 1.0399029154266232e-05, + "loss": 0.901, + "step": 274610 + }, + { + "epoch": 23.77, + "learning_rate": 1.0397584443352886e-05, + "loss": 0.9202, + "step": 274620 + }, + { + "epoch": 23.77, + "learning_rate": 1.0396139732439539e-05, + "loss": 0.9347, + "step": 274630 + }, + { + "epoch": 23.77, + "learning_rate": 1.0394695021526194e-05, + "loss": 0.9935, + "step": 274640 + }, + { + "epoch": 23.77, + "learning_rate": 1.0393250310612847e-05, + "loss": 0.9146, + "step": 274650 + }, + { + "epoch": 23.77, + "learning_rate": 1.03918055996995e-05, + "loss": 0.9124, + "step": 274660 + }, + { + "epoch": 23.77, + "learning_rate": 1.0390360888786154e-05, + "loss": 0.9332, + "step": 274670 + }, + { + "epoch": 23.78, + "learning_rate": 1.0388916177872808e-05, + "loss": 0.9009, + "step": 274680 + }, + { + "epoch": 23.78, + "learning_rate": 1.0387471466959461e-05, + "loss": 0.9301, + "step": 274690 + }, + { + "epoch": 23.78, + "learning_rate": 1.0386026756046116e-05, + "loss": 0.9437, + "step": 274700 + }, + { + "epoch": 23.78, + "learning_rate": 1.038458204513277e-05, + "loss": 0.9171, + "step": 274710 + }, + { + "epoch": 23.78, + "learning_rate": 1.0383137334219423e-05, + "loss": 0.9069, + "step": 274720 + }, + { + "epoch": 23.78, + "learning_rate": 1.0381692623306078e-05, + "loss": 0.8997, + "step": 274730 + }, + { + "epoch": 23.78, + "learning_rate": 1.0380247912392731e-05, + "loss": 0.9644, + "step": 274740 + }, + { + "epoch": 23.78, + "learning_rate": 1.0378803201479385e-05, + "loss": 0.9693, + "step": 274750 + }, + { + "epoch": 23.78, + "learning_rate": 1.0377358490566038e-05, + "loss": 0.8872, + "step": 274760 + }, + { + "epoch": 23.78, + "learning_rate": 1.0375913779652692e-05, + "loss": 0.894, + "step": 274770 + }, + { + "epoch": 23.78, + "learning_rate": 1.0374469068739345e-05, + "loss": 0.9079, + "step": 274780 + }, + { + "epoch": 23.79, + "learning_rate": 1.0373024357826e-05, + "loss": 0.8948, + "step": 274790 + }, + { + "epoch": 23.79, + "learning_rate": 1.0371579646912653e-05, + "loss": 0.9677, + "step": 274800 + }, + { + "epoch": 23.79, + "learning_rate": 1.0370134935999307e-05, + "loss": 0.9422, + "step": 274810 + }, + { + "epoch": 23.79, + "learning_rate": 1.0368690225085962e-05, + "loss": 0.8849, + "step": 274820 + }, + { + "epoch": 23.79, + "learning_rate": 1.0367245514172615e-05, + "loss": 0.9715, + "step": 274830 + }, + { + "epoch": 23.79, + "learning_rate": 1.0365800803259269e-05, + "loss": 0.9484, + "step": 274840 + }, + { + "epoch": 23.79, + "learning_rate": 1.0364356092345922e-05, + "loss": 0.9535, + "step": 274850 + }, + { + "epoch": 23.79, + "learning_rate": 1.0362911381432576e-05, + "loss": 0.9072, + "step": 274860 + }, + { + "epoch": 23.79, + "learning_rate": 1.0361466670519229e-05, + "loss": 0.9242, + "step": 274870 + }, + { + "epoch": 23.79, + "learning_rate": 1.0360021959605884e-05, + "loss": 0.9241, + "step": 274880 + }, + { + "epoch": 23.79, + "learning_rate": 1.0358577248692537e-05, + "loss": 0.9245, + "step": 274890 + }, + { + "epoch": 23.79, + "learning_rate": 1.035713253777919e-05, + "loss": 0.8782, + "step": 274900 + }, + { + "epoch": 23.8, + "learning_rate": 1.0355687826865846e-05, + "loss": 0.9124, + "step": 274910 + }, + { + "epoch": 23.8, + "learning_rate": 1.03542431159525e-05, + "loss": 0.9217, + "step": 274920 + }, + { + "epoch": 23.8, + "learning_rate": 1.0352798405039153e-05, + "loss": 0.875, + "step": 274930 + }, + { + "epoch": 23.8, + "learning_rate": 1.0351353694125806e-05, + "loss": 0.9145, + "step": 274940 + }, + { + "epoch": 23.8, + "learning_rate": 1.034990898321246e-05, + "loss": 0.9135, + "step": 274950 + }, + { + "epoch": 23.8, + "learning_rate": 1.0348464272299113e-05, + "loss": 0.9283, + "step": 274960 + }, + { + "epoch": 23.8, + "learning_rate": 1.0347019561385768e-05, + "loss": 0.9156, + "step": 274970 + }, + { + "epoch": 23.8, + "learning_rate": 1.0345574850472421e-05, + "loss": 0.9225, + "step": 274980 + }, + { + "epoch": 23.8, + "learning_rate": 1.0344130139559075e-05, + "loss": 0.9241, + "step": 274990 + }, + { + "epoch": 23.8, + "learning_rate": 1.0342685428645728e-05, + "loss": 0.951, + "step": 275000 + }, + { + "epoch": 23.8, + "learning_rate": 1.0341240717732383e-05, + "loss": 0.902, + "step": 275010 + }, + { + "epoch": 23.81, + "learning_rate": 1.0339796006819037e-05, + "loss": 0.9342, + "step": 275020 + }, + { + "epoch": 23.81, + "learning_rate": 1.0338351295905688e-05, + "loss": 0.9591, + "step": 275030 + }, + { + "epoch": 23.81, + "learning_rate": 1.0336906584992343e-05, + "loss": 0.9298, + "step": 275040 + }, + { + "epoch": 23.81, + "learning_rate": 1.0335461874078997e-05, + "loss": 0.8456, + "step": 275050 + }, + { + "epoch": 23.81, + "learning_rate": 1.033401716316565e-05, + "loss": 0.9124, + "step": 275060 + }, + { + "epoch": 23.81, + "learning_rate": 1.0332572452252305e-05, + "loss": 0.8479, + "step": 275070 + }, + { + "epoch": 23.81, + "learning_rate": 1.0331127741338959e-05, + "loss": 0.9578, + "step": 275080 + }, + { + "epoch": 23.81, + "learning_rate": 1.0329683030425612e-05, + "loss": 0.8884, + "step": 275090 + }, + { + "epoch": 23.81, + "learning_rate": 1.0328238319512267e-05, + "loss": 0.8704, + "step": 275100 + }, + { + "epoch": 23.81, + "learning_rate": 1.032679360859892e-05, + "loss": 0.92, + "step": 275110 + }, + { + "epoch": 23.81, + "learning_rate": 1.0325348897685572e-05, + "loss": 0.9157, + "step": 275120 + }, + { + "epoch": 23.81, + "learning_rate": 1.0323904186772227e-05, + "loss": 0.9218, + "step": 275130 + }, + { + "epoch": 23.82, + "learning_rate": 1.032245947585888e-05, + "loss": 0.9247, + "step": 275140 + }, + { + "epoch": 23.82, + "learning_rate": 1.0321014764945534e-05, + "loss": 0.9394, + "step": 275150 + }, + { + "epoch": 23.82, + "learning_rate": 1.031957005403219e-05, + "loss": 0.8453, + "step": 275160 + }, + { + "epoch": 23.82, + "learning_rate": 1.0318125343118843e-05, + "loss": 0.9503, + "step": 275170 + }, + { + "epoch": 23.82, + "learning_rate": 1.0316680632205496e-05, + "loss": 0.9336, + "step": 275180 + }, + { + "epoch": 23.82, + "learning_rate": 1.0315235921292151e-05, + "loss": 0.8933, + "step": 275190 + }, + { + "epoch": 23.82, + "learning_rate": 1.0313791210378804e-05, + "loss": 0.9079, + "step": 275200 + }, + { + "epoch": 23.82, + "learning_rate": 1.0312346499465456e-05, + "loss": 0.9547, + "step": 275210 + }, + { + "epoch": 23.82, + "learning_rate": 1.0310901788552111e-05, + "loss": 0.8903, + "step": 275220 + }, + { + "epoch": 23.82, + "learning_rate": 1.0309457077638765e-05, + "loss": 0.8378, + "step": 275230 + }, + { + "epoch": 23.82, + "learning_rate": 1.0308012366725418e-05, + "loss": 0.9293, + "step": 275240 + }, + { + "epoch": 23.82, + "learning_rate": 1.0306567655812073e-05, + "loss": 0.9501, + "step": 275250 + }, + { + "epoch": 23.83, + "learning_rate": 1.0305122944898727e-05, + "loss": 0.9566, + "step": 275260 + }, + { + "epoch": 23.83, + "learning_rate": 1.030367823398538e-05, + "loss": 0.9166, + "step": 275270 + }, + { + "epoch": 23.83, + "learning_rate": 1.0302233523072035e-05, + "loss": 0.9079, + "step": 275280 + }, + { + "epoch": 23.83, + "learning_rate": 1.0300788812158688e-05, + "loss": 0.87, + "step": 275290 + }, + { + "epoch": 23.83, + "learning_rate": 1.029934410124534e-05, + "loss": 0.9219, + "step": 275300 + }, + { + "epoch": 23.83, + "learning_rate": 1.0297899390331995e-05, + "loss": 0.8569, + "step": 275310 + }, + { + "epoch": 23.83, + "learning_rate": 1.0296454679418649e-05, + "loss": 0.9432, + "step": 275320 + }, + { + "epoch": 23.83, + "learning_rate": 1.0295009968505302e-05, + "loss": 0.9407, + "step": 275330 + }, + { + "epoch": 23.83, + "learning_rate": 1.0293565257591957e-05, + "loss": 0.9499, + "step": 275340 + }, + { + "epoch": 23.83, + "learning_rate": 1.029212054667861e-05, + "loss": 0.9386, + "step": 275350 + }, + { + "epoch": 23.83, + "learning_rate": 1.0290675835765264e-05, + "loss": 0.9286, + "step": 275360 + }, + { + "epoch": 23.84, + "learning_rate": 1.0289231124851919e-05, + "loss": 0.9981, + "step": 275370 + }, + { + "epoch": 23.84, + "learning_rate": 1.0287786413938572e-05, + "loss": 0.8628, + "step": 275380 + }, + { + "epoch": 23.84, + "learning_rate": 1.0286341703025224e-05, + "loss": 1.0409, + "step": 275390 + }, + { + "epoch": 23.84, + "learning_rate": 1.028489699211188e-05, + "loss": 0.9219, + "step": 275400 + }, + { + "epoch": 23.84, + "learning_rate": 1.0283452281198533e-05, + "loss": 0.913, + "step": 275410 + }, + { + "epoch": 23.84, + "learning_rate": 1.0282007570285186e-05, + "loss": 0.91, + "step": 275420 + }, + { + "epoch": 23.84, + "learning_rate": 1.028056285937184e-05, + "loss": 0.9077, + "step": 275430 + }, + { + "epoch": 23.84, + "learning_rate": 1.0279118148458494e-05, + "loss": 1.0236, + "step": 275440 + }, + { + "epoch": 23.84, + "learning_rate": 1.0277673437545148e-05, + "loss": 0.9421, + "step": 275450 + }, + { + "epoch": 23.84, + "learning_rate": 1.0276228726631801e-05, + "loss": 0.9433, + "step": 275460 + }, + { + "epoch": 23.84, + "learning_rate": 1.0274784015718456e-05, + "loss": 0.9228, + "step": 275470 + }, + { + "epoch": 23.84, + "learning_rate": 1.0273339304805108e-05, + "loss": 0.9309, + "step": 275480 + }, + { + "epoch": 23.85, + "learning_rate": 1.0271894593891761e-05, + "loss": 0.8848, + "step": 275490 + }, + { + "epoch": 23.85, + "learning_rate": 1.0270449882978417e-05, + "loss": 0.9196, + "step": 275500 + }, + { + "epoch": 23.85, + "learning_rate": 1.026900517206507e-05, + "loss": 0.9383, + "step": 275510 + }, + { + "epoch": 23.85, + "learning_rate": 1.0267560461151723e-05, + "loss": 0.9624, + "step": 275520 + }, + { + "epoch": 23.85, + "learning_rate": 1.0266115750238378e-05, + "loss": 0.955, + "step": 275530 + }, + { + "epoch": 23.85, + "learning_rate": 1.0264671039325032e-05, + "loss": 0.9659, + "step": 275540 + }, + { + "epoch": 23.85, + "learning_rate": 1.0263226328411685e-05, + "loss": 0.8059, + "step": 275550 + }, + { + "epoch": 23.85, + "learning_rate": 1.026178161749834e-05, + "loss": 0.9388, + "step": 275560 + }, + { + "epoch": 23.85, + "learning_rate": 1.0260336906584992e-05, + "loss": 0.925, + "step": 275570 + }, + { + "epoch": 23.85, + "learning_rate": 1.0258892195671645e-05, + "loss": 0.9173, + "step": 275580 + }, + { + "epoch": 23.85, + "learning_rate": 1.02574474847583e-05, + "loss": 0.9619, + "step": 275590 + }, + { + "epoch": 23.86, + "learning_rate": 1.0256002773844954e-05, + "loss": 0.9089, + "step": 275600 + }, + { + "epoch": 23.86, + "learning_rate": 1.0254558062931607e-05, + "loss": 0.9514, + "step": 275610 + }, + { + "epoch": 23.86, + "learning_rate": 1.0253113352018262e-05, + "loss": 0.949, + "step": 275620 + }, + { + "epoch": 23.86, + "learning_rate": 1.0251668641104916e-05, + "loss": 0.9306, + "step": 275630 + }, + { + "epoch": 23.86, + "learning_rate": 1.0250223930191569e-05, + "loss": 0.922, + "step": 275640 + }, + { + "epoch": 23.86, + "learning_rate": 1.0248779219278224e-05, + "loss": 0.94, + "step": 275650 + }, + { + "epoch": 23.86, + "learning_rate": 1.0247334508364876e-05, + "loss": 0.8731, + "step": 275660 + }, + { + "epoch": 23.86, + "learning_rate": 1.024588979745153e-05, + "loss": 0.9399, + "step": 275670 + }, + { + "epoch": 23.86, + "learning_rate": 1.0244445086538184e-05, + "loss": 0.9398, + "step": 275680 + }, + { + "epoch": 23.86, + "learning_rate": 1.0243000375624838e-05, + "loss": 0.9541, + "step": 275690 + }, + { + "epoch": 23.86, + "learning_rate": 1.0241555664711491e-05, + "loss": 0.9514, + "step": 275700 + }, + { + "epoch": 23.86, + "learning_rate": 1.0240110953798146e-05, + "loss": 0.9473, + "step": 275710 + }, + { + "epoch": 23.87, + "learning_rate": 1.02386662428848e-05, + "loss": 0.9484, + "step": 275720 + }, + { + "epoch": 23.87, + "learning_rate": 1.0237221531971453e-05, + "loss": 0.9253, + "step": 275730 + }, + { + "epoch": 23.87, + "learning_rate": 1.0235776821058108e-05, + "loss": 0.9297, + "step": 275740 + }, + { + "epoch": 23.87, + "learning_rate": 1.023433211014476e-05, + "loss": 0.8911, + "step": 275750 + }, + { + "epoch": 23.87, + "learning_rate": 1.0232887399231413e-05, + "loss": 0.9056, + "step": 275760 + }, + { + "epoch": 23.87, + "learning_rate": 1.0231442688318068e-05, + "loss": 0.9351, + "step": 275770 + }, + { + "epoch": 23.87, + "learning_rate": 1.0229997977404722e-05, + "loss": 0.9595, + "step": 275780 + }, + { + "epoch": 23.87, + "learning_rate": 1.0228553266491375e-05, + "loss": 0.9095, + "step": 275790 + }, + { + "epoch": 23.87, + "learning_rate": 1.022710855557803e-05, + "loss": 0.8667, + "step": 275800 + }, + { + "epoch": 23.87, + "learning_rate": 1.0225663844664684e-05, + "loss": 0.9337, + "step": 275810 + }, + { + "epoch": 23.87, + "learning_rate": 1.0224219133751337e-05, + "loss": 0.951, + "step": 275820 + }, + { + "epoch": 23.88, + "learning_rate": 1.0222774422837992e-05, + "loss": 0.9455, + "step": 275830 + }, + { + "epoch": 23.88, + "learning_rate": 1.0221329711924644e-05, + "loss": 0.8938, + "step": 275840 + }, + { + "epoch": 23.88, + "learning_rate": 1.0219885001011297e-05, + "loss": 0.9111, + "step": 275850 + }, + { + "epoch": 23.88, + "learning_rate": 1.0218440290097952e-05, + "loss": 0.8839, + "step": 275860 + }, + { + "epoch": 23.88, + "learning_rate": 1.0216995579184606e-05, + "loss": 0.9307, + "step": 275870 + }, + { + "epoch": 23.88, + "learning_rate": 1.0215550868271259e-05, + "loss": 0.9058, + "step": 275880 + }, + { + "epoch": 23.88, + "learning_rate": 1.0214106157357912e-05, + "loss": 0.9149, + "step": 275890 + }, + { + "epoch": 23.88, + "learning_rate": 1.0212661446444568e-05, + "loss": 0.9282, + "step": 275900 + }, + { + "epoch": 23.88, + "learning_rate": 1.0211216735531221e-05, + "loss": 0.9196, + "step": 275910 + }, + { + "epoch": 23.88, + "learning_rate": 1.0209772024617874e-05, + "loss": 0.9722, + "step": 275920 + }, + { + "epoch": 23.88, + "learning_rate": 1.0208327313704528e-05, + "loss": 0.8729, + "step": 275930 + }, + { + "epoch": 23.88, + "learning_rate": 1.0206882602791181e-05, + "loss": 0.9104, + "step": 275940 + }, + { + "epoch": 23.89, + "learning_rate": 1.0205437891877835e-05, + "loss": 0.9088, + "step": 275950 + }, + { + "epoch": 23.89, + "learning_rate": 1.020399318096449e-05, + "loss": 0.9667, + "step": 275960 + }, + { + "epoch": 23.89, + "learning_rate": 1.0202548470051143e-05, + "loss": 0.859, + "step": 275970 + }, + { + "epoch": 23.89, + "learning_rate": 1.0201103759137796e-05, + "loss": 0.8799, + "step": 275980 + }, + { + "epoch": 23.89, + "learning_rate": 1.0199659048224451e-05, + "loss": 0.8958, + "step": 275990 + }, + { + "epoch": 23.89, + "learning_rate": 1.0198214337311105e-05, + "loss": 0.947, + "step": 276000 + }, + { + "epoch": 23.89, + "learning_rate": 1.0196769626397758e-05, + "loss": 0.8767, + "step": 276010 + }, + { + "epoch": 23.89, + "learning_rate": 1.0195324915484412e-05, + "loss": 0.8977, + "step": 276020 + }, + { + "epoch": 23.89, + "learning_rate": 1.0193880204571065e-05, + "loss": 0.9482, + "step": 276030 + }, + { + "epoch": 23.89, + "learning_rate": 1.0192435493657718e-05, + "loss": 0.9213, + "step": 276040 + }, + { + "epoch": 23.89, + "learning_rate": 1.0190990782744374e-05, + "loss": 0.9626, + "step": 276050 + }, + { + "epoch": 23.9, + "learning_rate": 1.0189546071831027e-05, + "loss": 0.9319, + "step": 276060 + }, + { + "epoch": 23.9, + "learning_rate": 1.018810136091768e-05, + "loss": 0.8911, + "step": 276070 + }, + { + "epoch": 23.9, + "learning_rate": 1.0186656650004335e-05, + "loss": 0.9264, + "step": 276080 + }, + { + "epoch": 23.9, + "learning_rate": 1.0185211939090989e-05, + "loss": 0.901, + "step": 276090 + }, + { + "epoch": 23.9, + "learning_rate": 1.0183767228177642e-05, + "loss": 0.901, + "step": 276100 + }, + { + "epoch": 23.9, + "learning_rate": 1.0182322517264296e-05, + "loss": 0.975, + "step": 276110 + }, + { + "epoch": 23.9, + "learning_rate": 1.0180877806350949e-05, + "loss": 0.9133, + "step": 276120 + }, + { + "epoch": 23.9, + "learning_rate": 1.0179433095437602e-05, + "loss": 0.9574, + "step": 276130 + }, + { + "epoch": 23.9, + "learning_rate": 1.0177988384524257e-05, + "loss": 0.9578, + "step": 276140 + }, + { + "epoch": 23.9, + "learning_rate": 1.0176543673610911e-05, + "loss": 0.9224, + "step": 276150 + }, + { + "epoch": 23.9, + "learning_rate": 1.0175098962697564e-05, + "loss": 0.926, + "step": 276160 + }, + { + "epoch": 23.9, + "learning_rate": 1.017365425178422e-05, + "loss": 0.8283, + "step": 276170 + }, + { + "epoch": 23.91, + "learning_rate": 1.0172209540870873e-05, + "loss": 0.9336, + "step": 276180 + }, + { + "epoch": 23.91, + "learning_rate": 1.0170764829957526e-05, + "loss": 0.9106, + "step": 276190 + }, + { + "epoch": 23.91, + "learning_rate": 1.016932011904418e-05, + "loss": 0.9161, + "step": 276200 + }, + { + "epoch": 23.91, + "learning_rate": 1.0167875408130833e-05, + "loss": 0.8957, + "step": 276210 + }, + { + "epoch": 23.91, + "learning_rate": 1.0166430697217486e-05, + "loss": 0.9337, + "step": 276220 + }, + { + "epoch": 23.91, + "learning_rate": 1.0164985986304141e-05, + "loss": 0.9488, + "step": 276230 + }, + { + "epoch": 23.91, + "learning_rate": 1.0163541275390795e-05, + "loss": 0.9382, + "step": 276240 + }, + { + "epoch": 23.91, + "learning_rate": 1.0162096564477448e-05, + "loss": 0.8796, + "step": 276250 + }, + { + "epoch": 23.91, + "learning_rate": 1.0160651853564103e-05, + "loss": 0.9157, + "step": 276260 + }, + { + "epoch": 23.91, + "learning_rate": 1.0159207142650757e-05, + "loss": 0.8836, + "step": 276270 + }, + { + "epoch": 23.91, + "learning_rate": 1.015776243173741e-05, + "loss": 0.9271, + "step": 276280 + }, + { + "epoch": 23.91, + "learning_rate": 1.0156317720824064e-05, + "loss": 0.9295, + "step": 276290 + }, + { + "epoch": 23.92, + "learning_rate": 1.0154873009910717e-05, + "loss": 0.9385, + "step": 276300 + }, + { + "epoch": 23.92, + "learning_rate": 1.015342829899737e-05, + "loss": 0.926, + "step": 276310 + }, + { + "epoch": 23.92, + "learning_rate": 1.0151983588084025e-05, + "loss": 0.946, + "step": 276320 + }, + { + "epoch": 23.92, + "learning_rate": 1.0150538877170679e-05, + "loss": 0.9272, + "step": 276330 + }, + { + "epoch": 23.92, + "learning_rate": 1.0149094166257332e-05, + "loss": 0.9124, + "step": 276340 + }, + { + "epoch": 23.92, + "learning_rate": 1.0147649455343986e-05, + "loss": 0.8941, + "step": 276350 + }, + { + "epoch": 23.92, + "learning_rate": 1.014620474443064e-05, + "loss": 0.9122, + "step": 276360 + }, + { + "epoch": 23.92, + "learning_rate": 1.0144760033517294e-05, + "loss": 0.9409, + "step": 276370 + }, + { + "epoch": 23.92, + "learning_rate": 1.0143315322603947e-05, + "loss": 0.9093, + "step": 276380 + }, + { + "epoch": 23.92, + "learning_rate": 1.01418706116906e-05, + "loss": 0.8666, + "step": 276390 + }, + { + "epoch": 23.92, + "learning_rate": 1.0140425900777254e-05, + "loss": 0.925, + "step": 276400 + }, + { + "epoch": 23.93, + "learning_rate": 1.0138981189863908e-05, + "loss": 0.9266, + "step": 276410 + }, + { + "epoch": 23.93, + "learning_rate": 1.0137536478950563e-05, + "loss": 0.9387, + "step": 276420 + }, + { + "epoch": 23.93, + "learning_rate": 1.0136091768037216e-05, + "loss": 0.9539, + "step": 276430 + }, + { + "epoch": 23.93, + "learning_rate": 1.013464705712387e-05, + "loss": 0.866, + "step": 276440 + }, + { + "epoch": 23.93, + "learning_rate": 1.0133202346210525e-05, + "loss": 0.9018, + "step": 276450 + }, + { + "epoch": 23.93, + "learning_rate": 1.0131757635297178e-05, + "loss": 0.9286, + "step": 276460 + }, + { + "epoch": 23.93, + "learning_rate": 1.0130312924383831e-05, + "loss": 0.9958, + "step": 276470 + }, + { + "epoch": 23.93, + "learning_rate": 1.0128868213470485e-05, + "loss": 0.8753, + "step": 276480 + }, + { + "epoch": 23.93, + "learning_rate": 1.0127423502557138e-05, + "loss": 0.9795, + "step": 276490 + }, + { + "epoch": 23.93, + "learning_rate": 1.0125978791643792e-05, + "loss": 0.931, + "step": 276500 + }, + { + "epoch": 23.93, + "learning_rate": 1.0124534080730447e-05, + "loss": 0.8948, + "step": 276510 + }, + { + "epoch": 23.93, + "learning_rate": 1.01230893698171e-05, + "loss": 0.9146, + "step": 276520 + }, + { + "epoch": 23.94, + "learning_rate": 1.0121644658903753e-05, + "loss": 0.9183, + "step": 276530 + }, + { + "epoch": 23.94, + "learning_rate": 1.0120199947990409e-05, + "loss": 0.9334, + "step": 276540 + }, + { + "epoch": 23.94, + "learning_rate": 1.0118755237077062e-05, + "loss": 0.8772, + "step": 276550 + }, + { + "epoch": 23.94, + "learning_rate": 1.0117310526163715e-05, + "loss": 0.9091, + "step": 276560 + }, + { + "epoch": 23.94, + "learning_rate": 1.0115865815250369e-05, + "loss": 0.9666, + "step": 276570 + }, + { + "epoch": 23.94, + "learning_rate": 1.0114421104337022e-05, + "loss": 0.8946, + "step": 276580 + }, + { + "epoch": 23.94, + "learning_rate": 1.0112976393423676e-05, + "loss": 0.9189, + "step": 276590 + }, + { + "epoch": 23.94, + "learning_rate": 1.011153168251033e-05, + "loss": 0.8754, + "step": 276600 + }, + { + "epoch": 23.94, + "learning_rate": 1.0110086971596984e-05, + "loss": 0.9643, + "step": 276610 + }, + { + "epoch": 23.94, + "learning_rate": 1.0108642260683637e-05, + "loss": 0.8973, + "step": 276620 + }, + { + "epoch": 23.94, + "learning_rate": 1.0107197549770292e-05, + "loss": 0.9466, + "step": 276630 + }, + { + "epoch": 23.95, + "learning_rate": 1.0105752838856946e-05, + "loss": 0.8516, + "step": 276640 + }, + { + "epoch": 23.95, + "learning_rate": 1.01043081279436e-05, + "loss": 0.9226, + "step": 276650 + }, + { + "epoch": 23.95, + "learning_rate": 1.0102863417030253e-05, + "loss": 0.9045, + "step": 276660 + }, + { + "epoch": 23.95, + "learning_rate": 1.0101418706116906e-05, + "loss": 0.936, + "step": 276670 + }, + { + "epoch": 23.95, + "learning_rate": 1.009997399520356e-05, + "loss": 0.9551, + "step": 276680 + }, + { + "epoch": 23.95, + "learning_rate": 1.0098529284290215e-05, + "loss": 0.9253, + "step": 276690 + }, + { + "epoch": 23.95, + "learning_rate": 1.0097084573376868e-05, + "loss": 0.925, + "step": 276700 + }, + { + "epoch": 23.95, + "learning_rate": 1.0095639862463521e-05, + "loss": 0.8845, + "step": 276710 + }, + { + "epoch": 23.95, + "learning_rate": 1.0094195151550176e-05, + "loss": 0.9214, + "step": 276720 + }, + { + "epoch": 23.95, + "learning_rate": 1.009275044063683e-05, + "loss": 0.8796, + "step": 276730 + }, + { + "epoch": 23.95, + "learning_rate": 1.0091305729723483e-05, + "loss": 0.9564, + "step": 276740 + }, + { + "epoch": 23.95, + "learning_rate": 1.0089861018810137e-05, + "loss": 0.9506, + "step": 276750 + }, + { + "epoch": 23.96, + "learning_rate": 1.008841630789679e-05, + "loss": 0.9241, + "step": 276760 + }, + { + "epoch": 23.96, + "learning_rate": 1.0086971596983443e-05, + "loss": 0.9323, + "step": 276770 + }, + { + "epoch": 23.96, + "learning_rate": 1.0085526886070098e-05, + "loss": 0.9302, + "step": 276780 + }, + { + "epoch": 23.96, + "learning_rate": 1.0084082175156752e-05, + "loss": 0.9246, + "step": 276790 + }, + { + "epoch": 23.96, + "learning_rate": 1.0082637464243405e-05, + "loss": 0.9316, + "step": 276800 + }, + { + "epoch": 23.96, + "learning_rate": 1.0081192753330059e-05, + "loss": 0.9859, + "step": 276810 + }, + { + "epoch": 23.96, + "learning_rate": 1.0079748042416714e-05, + "loss": 0.9218, + "step": 276820 + }, + { + "epoch": 23.96, + "learning_rate": 1.0078303331503367e-05, + "loss": 0.9197, + "step": 276830 + }, + { + "epoch": 23.96, + "learning_rate": 1.007685862059002e-05, + "loss": 0.9337, + "step": 276840 + }, + { + "epoch": 23.96, + "learning_rate": 1.0075413909676674e-05, + "loss": 0.9712, + "step": 276850 + }, + { + "epoch": 23.96, + "learning_rate": 1.0073969198763327e-05, + "loss": 0.9404, + "step": 276860 + }, + { + "epoch": 23.97, + "learning_rate": 1.007252448784998e-05, + "loss": 0.9253, + "step": 276870 + }, + { + "epoch": 23.97, + "learning_rate": 1.0071079776936636e-05, + "loss": 0.9651, + "step": 276880 + }, + { + "epoch": 23.97, + "learning_rate": 1.006963506602329e-05, + "loss": 0.9272, + "step": 276890 + }, + { + "epoch": 23.97, + "learning_rate": 1.0068190355109943e-05, + "loss": 0.9546, + "step": 276900 + }, + { + "epoch": 23.97, + "learning_rate": 1.0066745644196598e-05, + "loss": 0.9421, + "step": 276910 + }, + { + "epoch": 23.97, + "learning_rate": 1.0065300933283251e-05, + "loss": 0.9306, + "step": 276920 + }, + { + "epoch": 23.97, + "learning_rate": 1.0063856222369904e-05, + "loss": 0.9876, + "step": 276930 + }, + { + "epoch": 23.97, + "learning_rate": 1.0062411511456558e-05, + "loss": 0.9273, + "step": 276940 + }, + { + "epoch": 23.97, + "learning_rate": 1.0060966800543211e-05, + "loss": 0.9115, + "step": 276950 + }, + { + "epoch": 23.97, + "learning_rate": 1.0059522089629865e-05, + "loss": 0.947, + "step": 276960 + }, + { + "epoch": 23.97, + "learning_rate": 1.005807737871652e-05, + "loss": 0.9042, + "step": 276970 + }, + { + "epoch": 23.97, + "learning_rate": 1.0056632667803173e-05, + "loss": 0.8936, + "step": 276980 + }, + { + "epoch": 23.98, + "learning_rate": 1.0055187956889827e-05, + "loss": 0.8951, + "step": 276990 + }, + { + "epoch": 23.98, + "learning_rate": 1.0053743245976482e-05, + "loss": 0.8612, + "step": 277000 + }, + { + "epoch": 23.98, + "learning_rate": 1.0052298535063135e-05, + "loss": 0.8502, + "step": 277010 + }, + { + "epoch": 23.98, + "learning_rate": 1.0050853824149788e-05, + "loss": 0.9102, + "step": 277020 + }, + { + "epoch": 23.98, + "learning_rate": 1.0049409113236442e-05, + "loss": 0.8759, + "step": 277030 + }, + { + "epoch": 23.98, + "learning_rate": 1.0047964402323095e-05, + "loss": 0.9312, + "step": 277040 + }, + { + "epoch": 23.98, + "learning_rate": 1.0046519691409749e-05, + "loss": 0.919, + "step": 277050 + }, + { + "epoch": 23.98, + "learning_rate": 1.0045074980496404e-05, + "loss": 0.8713, + "step": 277060 + }, + { + "epoch": 23.98, + "learning_rate": 1.0043630269583057e-05, + "loss": 0.9056, + "step": 277070 + }, + { + "epoch": 23.98, + "learning_rate": 1.004218555866971e-05, + "loss": 0.9058, + "step": 277080 + }, + { + "epoch": 23.98, + "learning_rate": 1.0040740847756366e-05, + "loss": 0.9575, + "step": 277090 + }, + { + "epoch": 23.99, + "learning_rate": 1.0039296136843019e-05, + "loss": 0.9695, + "step": 277100 + }, + { + "epoch": 23.99, + "learning_rate": 1.0037851425929672e-05, + "loss": 0.9176, + "step": 277110 + }, + { + "epoch": 23.99, + "learning_rate": 1.0036406715016326e-05, + "loss": 0.9081, + "step": 277120 + }, + { + "epoch": 23.99, + "learning_rate": 1.003496200410298e-05, + "loss": 0.915, + "step": 277130 + }, + { + "epoch": 23.99, + "learning_rate": 1.0033517293189633e-05, + "loss": 0.9691, + "step": 277140 + }, + { + "epoch": 23.99, + "learning_rate": 1.0032072582276288e-05, + "loss": 0.8759, + "step": 277150 + }, + { + "epoch": 23.99, + "learning_rate": 1.0030627871362941e-05, + "loss": 0.9548, + "step": 277160 + }, + { + "epoch": 23.99, + "learning_rate": 1.0029183160449594e-05, + "loss": 0.9034, + "step": 277170 + }, + { + "epoch": 23.99, + "learning_rate": 1.002773844953625e-05, + "loss": 0.842, + "step": 277180 + }, + { + "epoch": 23.99, + "learning_rate": 1.0026293738622903e-05, + "loss": 0.8966, + "step": 277190 + }, + { + "epoch": 23.99, + "learning_rate": 1.0024849027709556e-05, + "loss": 0.9466, + "step": 277200 + }, + { + "epoch": 23.99, + "learning_rate": 1.002340431679621e-05, + "loss": 0.8854, + "step": 277210 + }, + { + "epoch": 24.0, + "learning_rate": 1.0021959605882863e-05, + "loss": 0.9064, + "step": 277220 + }, + { + "epoch": 24.0, + "learning_rate": 1.0020514894969517e-05, + "loss": 0.9312, + "step": 277230 + }, + { + "epoch": 24.0, + "learning_rate": 1.001907018405617e-05, + "loss": 0.8773, + "step": 277240 + }, + { + "epoch": 24.0, + "learning_rate": 1.0017625473142825e-05, + "loss": 0.9374, + "step": 277250 + }, + { + "epoch": 24.0, + "learning_rate": 1.0016180762229478e-05, + "loss": 0.9345, + "step": 277260 + }, + { + "epoch": 24.0, + "learning_rate": 1.0014736051316132e-05, + "loss": 0.9848, + "step": 277270 + }, + { + "epoch": 24.0, + "eval_cer": 0.9930738212451684, + "eval_em": 0, + "eval_f1": 0.11994211257845032, + "eval_loss": 0.9396665096282959, + "eval_runtime": 778.0499, + "eval_samples_per_second": 6.6, + "eval_steps_per_second": 0.825, + "eval_wer": 0.9666931146652686, + "step": 277272 + }, + { + "epoch": 24.0, + "learning_rate": 1.0013291340402787e-05, + "loss": 0.9085, + "step": 277280 + }, + { + "epoch": 24.0, + "learning_rate": 1.001184662948944e-05, + "loss": 0.9187, + "step": 277290 + }, + { + "epoch": 24.0, + "learning_rate": 1.0010401918576092e-05, + "loss": 0.9096, + "step": 277300 + }, + { + "epoch": 24.0, + "learning_rate": 1.0008957207662747e-05, + "loss": 0.9518, + "step": 277310 + }, + { + "epoch": 24.0, + "learning_rate": 1.00075124967494e-05, + "loss": 0.9412, + "step": 277320 + }, + { + "epoch": 24.01, + "learning_rate": 1.0006067785836054e-05, + "loss": 0.8944, + "step": 277330 + }, + { + "epoch": 24.01, + "learning_rate": 1.0004623074922709e-05, + "loss": 0.8972, + "step": 277340 + }, + { + "epoch": 24.01, + "learning_rate": 1.0003178364009362e-05, + "loss": 0.9663, + "step": 277350 + }, + { + "epoch": 24.01, + "learning_rate": 1.0001733653096016e-05, + "loss": 0.9209, + "step": 277360 + }, + { + "epoch": 24.01, + "learning_rate": 1.000028894218267e-05, + "loss": 0.8641, + "step": 277370 + }, + { + "epoch": 24.01, + "learning_rate": 9.998844231269324e-06, + "loss": 0.9072, + "step": 277380 + }, + { + "epoch": 24.01, + "learning_rate": 9.997399520355976e-06, + "loss": 0.8715, + "step": 277390 + }, + { + "epoch": 24.01, + "learning_rate": 9.995954809442631e-06, + "loss": 0.9235, + "step": 277400 + }, + { + "epoch": 24.01, + "learning_rate": 9.994510098529284e-06, + "loss": 0.8803, + "step": 277410 + }, + { + "epoch": 24.01, + "learning_rate": 9.993065387615938e-06, + "loss": 0.9238, + "step": 277420 + }, + { + "epoch": 24.01, + "learning_rate": 9.991620676702593e-06, + "loss": 0.9421, + "step": 277430 + }, + { + "epoch": 24.01, + "learning_rate": 9.990175965789246e-06, + "loss": 0.9302, + "step": 277440 + }, + { + "epoch": 24.02, + "learning_rate": 9.9887312548759e-06, + "loss": 0.9243, + "step": 277450 + }, + { + "epoch": 24.02, + "learning_rate": 9.987286543962555e-06, + "loss": 0.8953, + "step": 277460 + }, + { + "epoch": 24.02, + "learning_rate": 9.985841833049208e-06, + "loss": 0.891, + "step": 277470 + }, + { + "epoch": 24.02, + "learning_rate": 9.98439712213586e-06, + "loss": 1.0047, + "step": 277480 + }, + { + "epoch": 24.02, + "learning_rate": 9.982952411222515e-06, + "loss": 0.9774, + "step": 277490 + }, + { + "epoch": 24.02, + "learning_rate": 9.981507700309168e-06, + "loss": 0.9373, + "step": 277500 + }, + { + "epoch": 24.02, + "learning_rate": 9.980062989395822e-06, + "loss": 0.9609, + "step": 277510 + }, + { + "epoch": 24.02, + "learning_rate": 9.978618278482477e-06, + "loss": 0.9312, + "step": 277520 + }, + { + "epoch": 24.02, + "learning_rate": 9.97717356756913e-06, + "loss": 0.8803, + "step": 277530 + }, + { + "epoch": 24.02, + "learning_rate": 9.975728856655784e-06, + "loss": 0.9343, + "step": 277540 + }, + { + "epoch": 24.02, + "learning_rate": 9.974284145742439e-06, + "loss": 0.9118, + "step": 277550 + }, + { + "epoch": 24.02, + "learning_rate": 9.972839434829092e-06, + "loss": 0.9202, + "step": 277560 + }, + { + "epoch": 24.03, + "learning_rate": 9.971394723915744e-06, + "loss": 0.9784, + "step": 277570 + }, + { + "epoch": 24.03, + "learning_rate": 9.969950013002399e-06, + "loss": 0.8529, + "step": 277580 + }, + { + "epoch": 24.03, + "learning_rate": 9.968505302089052e-06, + "loss": 0.9114, + "step": 277590 + }, + { + "epoch": 24.03, + "learning_rate": 9.967060591175706e-06, + "loss": 0.9026, + "step": 277600 + }, + { + "epoch": 24.03, + "learning_rate": 9.96561588026236e-06, + "loss": 0.9238, + "step": 277610 + }, + { + "epoch": 24.03, + "learning_rate": 9.964171169349014e-06, + "loss": 0.8855, + "step": 277620 + }, + { + "epoch": 24.03, + "learning_rate": 9.962726458435668e-06, + "loss": 0.8805, + "step": 277630 + }, + { + "epoch": 24.03, + "learning_rate": 9.961281747522323e-06, + "loss": 0.943, + "step": 277640 + }, + { + "epoch": 24.03, + "learning_rate": 9.959837036608976e-06, + "loss": 0.9387, + "step": 277650 + }, + { + "epoch": 24.03, + "learning_rate": 9.958392325695628e-06, + "loss": 0.9104, + "step": 277660 + }, + { + "epoch": 24.03, + "learning_rate": 9.956947614782283e-06, + "loss": 0.9099, + "step": 277670 + }, + { + "epoch": 24.04, + "learning_rate": 9.955502903868936e-06, + "loss": 0.9471, + "step": 277680 + }, + { + "epoch": 24.04, + "learning_rate": 9.95405819295559e-06, + "loss": 0.916, + "step": 277690 + }, + { + "epoch": 24.04, + "learning_rate": 9.952613482042243e-06, + "loss": 0.9417, + "step": 277700 + }, + { + "epoch": 24.04, + "learning_rate": 9.951168771128898e-06, + "loss": 0.9382, + "step": 277710 + }, + { + "epoch": 24.04, + "learning_rate": 9.949724060215551e-06, + "loss": 0.8777, + "step": 277720 + }, + { + "epoch": 24.04, + "learning_rate": 9.948279349302205e-06, + "loss": 0.9079, + "step": 277730 + }, + { + "epoch": 24.04, + "learning_rate": 9.94683463838886e-06, + "loss": 0.8988, + "step": 277740 + }, + { + "epoch": 24.04, + "learning_rate": 9.945389927475512e-06, + "loss": 0.9353, + "step": 277750 + }, + { + "epoch": 24.04, + "learning_rate": 9.943945216562165e-06, + "loss": 0.9353, + "step": 277760 + }, + { + "epoch": 24.04, + "learning_rate": 9.94250050564882e-06, + "loss": 0.9925, + "step": 277770 + }, + { + "epoch": 24.04, + "learning_rate": 9.941055794735474e-06, + "loss": 0.8929, + "step": 277780 + }, + { + "epoch": 24.04, + "learning_rate": 9.939611083822127e-06, + "loss": 0.942, + "step": 277790 + }, + { + "epoch": 24.05, + "learning_rate": 9.938166372908782e-06, + "loss": 0.9254, + "step": 277800 + }, + { + "epoch": 24.05, + "learning_rate": 9.936721661995435e-06, + "loss": 0.9089, + "step": 277810 + }, + { + "epoch": 24.05, + "learning_rate": 9.935276951082089e-06, + "loss": 0.9642, + "step": 277820 + }, + { + "epoch": 24.05, + "learning_rate": 9.933832240168744e-06, + "loss": 0.9542, + "step": 277830 + }, + { + "epoch": 24.05, + "learning_rate": 9.932387529255396e-06, + "loss": 0.9043, + "step": 277840 + }, + { + "epoch": 24.05, + "learning_rate": 9.930942818342049e-06, + "loss": 0.922, + "step": 277850 + }, + { + "epoch": 24.05, + "learning_rate": 9.929498107428704e-06, + "loss": 0.9075, + "step": 277860 + }, + { + "epoch": 24.05, + "learning_rate": 9.928053396515358e-06, + "loss": 0.9846, + "step": 277870 + }, + { + "epoch": 24.05, + "learning_rate": 9.926608685602011e-06, + "loss": 0.913, + "step": 277880 + }, + { + "epoch": 24.05, + "learning_rate": 9.925163974688666e-06, + "loss": 0.8737, + "step": 277890 + }, + { + "epoch": 24.05, + "learning_rate": 9.92371926377532e-06, + "loss": 0.881, + "step": 277900 + }, + { + "epoch": 24.06, + "learning_rate": 9.922274552861973e-06, + "loss": 0.8682, + "step": 277910 + }, + { + "epoch": 24.06, + "learning_rate": 9.920829841948628e-06, + "loss": 0.881, + "step": 277920 + }, + { + "epoch": 24.06, + "learning_rate": 9.91938513103528e-06, + "loss": 0.8801, + "step": 277930 + }, + { + "epoch": 24.06, + "learning_rate": 9.917940420121933e-06, + "loss": 0.9177, + "step": 277940 + }, + { + "epoch": 24.06, + "learning_rate": 9.916495709208588e-06, + "loss": 0.9489, + "step": 277950 + }, + { + "epoch": 24.06, + "learning_rate": 9.915050998295241e-06, + "loss": 0.9285, + "step": 277960 + }, + { + "epoch": 24.06, + "learning_rate": 9.913606287381895e-06, + "loss": 0.8947, + "step": 277970 + }, + { + "epoch": 24.06, + "learning_rate": 9.91216157646855e-06, + "loss": 0.9714, + "step": 277980 + }, + { + "epoch": 24.06, + "learning_rate": 9.910716865555203e-06, + "loss": 0.9218, + "step": 277990 + }, + { + "epoch": 24.06, + "learning_rate": 9.909272154641857e-06, + "loss": 0.9162, + "step": 278000 + }, + { + "epoch": 24.06, + "learning_rate": 9.90782744372851e-06, + "loss": 0.8639, + "step": 278010 + }, + { + "epoch": 24.06, + "learning_rate": 9.906382732815164e-06, + "loss": 0.9916, + "step": 278020 + }, + { + "epoch": 24.07, + "learning_rate": 9.904938021901817e-06, + "loss": 0.9074, + "step": 278030 + }, + { + "epoch": 24.07, + "learning_rate": 9.903493310988472e-06, + "loss": 0.8769, + "step": 278040 + }, + { + "epoch": 24.07, + "learning_rate": 9.902048600075125e-06, + "loss": 0.964, + "step": 278050 + }, + { + "epoch": 24.07, + "learning_rate": 9.900603889161779e-06, + "loss": 0.8678, + "step": 278060 + }, + { + "epoch": 24.07, + "learning_rate": 9.899159178248434e-06, + "loss": 0.8932, + "step": 278070 + }, + { + "epoch": 24.07, + "learning_rate": 9.897714467335087e-06, + "loss": 0.9646, + "step": 278080 + }, + { + "epoch": 24.07, + "learning_rate": 9.89626975642174e-06, + "loss": 0.9244, + "step": 278090 + }, + { + "epoch": 24.07, + "learning_rate": 9.894825045508394e-06, + "loss": 0.9445, + "step": 278100 + }, + { + "epoch": 24.07, + "learning_rate": 9.893380334595047e-06, + "loss": 0.9083, + "step": 278110 + }, + { + "epoch": 24.07, + "learning_rate": 9.891935623681701e-06, + "loss": 0.9732, + "step": 278120 + }, + { + "epoch": 24.07, + "learning_rate": 9.890490912768356e-06, + "loss": 0.9597, + "step": 278130 + }, + { + "epoch": 24.08, + "learning_rate": 9.88904620185501e-06, + "loss": 0.9171, + "step": 278140 + }, + { + "epoch": 24.08, + "learning_rate": 9.887601490941663e-06, + "loss": 0.9165, + "step": 278150 + }, + { + "epoch": 24.08, + "learning_rate": 9.886156780028316e-06, + "loss": 0.9176, + "step": 278160 + }, + { + "epoch": 24.08, + "learning_rate": 9.884712069114971e-06, + "loss": 0.8906, + "step": 278170 + }, + { + "epoch": 24.08, + "learning_rate": 9.883267358201625e-06, + "loss": 0.8816, + "step": 278180 + }, + { + "epoch": 24.08, + "learning_rate": 9.881822647288278e-06, + "loss": 0.8576, + "step": 278190 + }, + { + "epoch": 24.08, + "learning_rate": 9.880377936374931e-06, + "loss": 0.9397, + "step": 278200 + }, + { + "epoch": 24.08, + "learning_rate": 9.878933225461585e-06, + "loss": 0.9169, + "step": 278210 + }, + { + "epoch": 24.08, + "learning_rate": 9.877488514548238e-06, + "loss": 0.9458, + "step": 278220 + }, + { + "epoch": 24.08, + "learning_rate": 9.876043803634893e-06, + "loss": 0.9207, + "step": 278230 + }, + { + "epoch": 24.08, + "learning_rate": 9.874599092721547e-06, + "loss": 0.9508, + "step": 278240 + }, + { + "epoch": 24.08, + "learning_rate": 9.8731543818082e-06, + "loss": 0.8689, + "step": 278250 + }, + { + "epoch": 24.09, + "learning_rate": 9.871709670894855e-06, + "loss": 0.8939, + "step": 278260 + }, + { + "epoch": 24.09, + "learning_rate": 9.870264959981509e-06, + "loss": 0.9111, + "step": 278270 + }, + { + "epoch": 24.09, + "learning_rate": 9.868820249068162e-06, + "loss": 0.8808, + "step": 278280 + }, + { + "epoch": 24.09, + "learning_rate": 9.867375538154815e-06, + "loss": 0.942, + "step": 278290 + }, + { + "epoch": 24.09, + "learning_rate": 9.865930827241469e-06, + "loss": 0.8559, + "step": 278300 + }, + { + "epoch": 24.09, + "learning_rate": 9.864486116328122e-06, + "loss": 0.9246, + "step": 278310 + }, + { + "epoch": 24.09, + "learning_rate": 9.863041405414777e-06, + "loss": 0.9582, + "step": 278320 + }, + { + "epoch": 24.09, + "learning_rate": 9.86159669450143e-06, + "loss": 0.9289, + "step": 278330 + }, + { + "epoch": 24.09, + "learning_rate": 9.860151983588084e-06, + "loss": 0.9387, + "step": 278340 + }, + { + "epoch": 24.09, + "learning_rate": 9.858707272674739e-06, + "loss": 0.848, + "step": 278350 + }, + { + "epoch": 24.09, + "learning_rate": 9.857262561761392e-06, + "loss": 0.8894, + "step": 278360 + }, + { + "epoch": 24.1, + "learning_rate": 9.855817850848046e-06, + "loss": 0.9085, + "step": 278370 + }, + { + "epoch": 24.1, + "learning_rate": 9.8543731399347e-06, + "loss": 0.8721, + "step": 278380 + }, + { + "epoch": 24.1, + "learning_rate": 9.852928429021353e-06, + "loss": 0.9795, + "step": 278390 + }, + { + "epoch": 24.1, + "learning_rate": 9.851483718108006e-06, + "loss": 0.8991, + "step": 278400 + }, + { + "epoch": 24.1, + "learning_rate": 9.850039007194661e-06, + "loss": 0.9105, + "step": 278410 + }, + { + "epoch": 24.1, + "learning_rate": 9.848594296281315e-06, + "loss": 0.8522, + "step": 278420 + }, + { + "epoch": 24.1, + "learning_rate": 9.847149585367968e-06, + "loss": 0.8958, + "step": 278430 + }, + { + "epoch": 24.1, + "learning_rate": 9.845704874454623e-06, + "loss": 0.9154, + "step": 278440 + }, + { + "epoch": 24.1, + "learning_rate": 9.844260163541276e-06, + "loss": 0.8912, + "step": 278450 + }, + { + "epoch": 24.1, + "learning_rate": 9.84281545262793e-06, + "loss": 0.8682, + "step": 278460 + }, + { + "epoch": 24.1, + "learning_rate": 9.841370741714583e-06, + "loss": 0.8869, + "step": 278470 + }, + { + "epoch": 24.1, + "learning_rate": 9.839926030801237e-06, + "loss": 0.982, + "step": 278480 + }, + { + "epoch": 24.11, + "learning_rate": 9.83848131988789e-06, + "loss": 0.9108, + "step": 278490 + }, + { + "epoch": 24.11, + "learning_rate": 9.837036608974545e-06, + "loss": 0.9216, + "step": 278500 + }, + { + "epoch": 24.11, + "learning_rate": 9.835591898061199e-06, + "loss": 0.8741, + "step": 278510 + }, + { + "epoch": 24.11, + "learning_rate": 9.834147187147852e-06, + "loss": 0.9573, + "step": 278520 + }, + { + "epoch": 24.11, + "learning_rate": 9.832702476234507e-06, + "loss": 0.8435, + "step": 278530 + }, + { + "epoch": 24.11, + "learning_rate": 9.83125776532116e-06, + "loss": 0.874, + "step": 278540 + }, + { + "epoch": 24.11, + "learning_rate": 9.829813054407814e-06, + "loss": 0.9098, + "step": 278550 + }, + { + "epoch": 24.11, + "learning_rate": 9.828368343494467e-06, + "loss": 0.933, + "step": 278560 + }, + { + "epoch": 24.11, + "learning_rate": 9.82692363258112e-06, + "loss": 0.8573, + "step": 278570 + }, + { + "epoch": 24.11, + "learning_rate": 9.825478921667774e-06, + "loss": 0.9514, + "step": 278580 + }, + { + "epoch": 24.11, + "learning_rate": 9.824034210754429e-06, + "loss": 0.8867, + "step": 278590 + }, + { + "epoch": 24.11, + "learning_rate": 9.822589499841082e-06, + "loss": 0.9554, + "step": 278600 + }, + { + "epoch": 24.12, + "learning_rate": 9.821144788927736e-06, + "loss": 0.8875, + "step": 278610 + }, + { + "epoch": 24.12, + "learning_rate": 9.81970007801439e-06, + "loss": 0.9364, + "step": 278620 + }, + { + "epoch": 24.12, + "learning_rate": 9.818255367101044e-06, + "loss": 0.9291, + "step": 278630 + }, + { + "epoch": 24.12, + "learning_rate": 9.816810656187698e-06, + "loss": 0.9501, + "step": 278640 + }, + { + "epoch": 24.12, + "learning_rate": 9.815365945274351e-06, + "loss": 0.9108, + "step": 278650 + }, + { + "epoch": 24.12, + "learning_rate": 9.813921234361005e-06, + "loss": 0.8958, + "step": 278660 + }, + { + "epoch": 24.12, + "learning_rate": 9.812476523447658e-06, + "loss": 0.9479, + "step": 278670 + }, + { + "epoch": 24.12, + "learning_rate": 9.811031812534311e-06, + "loss": 0.9646, + "step": 278680 + }, + { + "epoch": 24.12, + "learning_rate": 9.809587101620966e-06, + "loss": 0.9103, + "step": 278690 + }, + { + "epoch": 24.12, + "learning_rate": 9.80814239070762e-06, + "loss": 0.9294, + "step": 278700 + }, + { + "epoch": 24.12, + "learning_rate": 9.806697679794273e-06, + "loss": 0.8659, + "step": 278710 + }, + { + "epoch": 24.13, + "learning_rate": 9.805252968880928e-06, + "loss": 0.9662, + "step": 278720 + }, + { + "epoch": 24.13, + "learning_rate": 9.803808257967582e-06, + "loss": 0.8783, + "step": 278730 + }, + { + "epoch": 24.13, + "learning_rate": 9.802363547054235e-06, + "loss": 0.9628, + "step": 278740 + }, + { + "epoch": 24.13, + "learning_rate": 9.800918836140888e-06, + "loss": 0.9268, + "step": 278750 + }, + { + "epoch": 24.13, + "learning_rate": 9.799474125227542e-06, + "loss": 0.8965, + "step": 278760 + }, + { + "epoch": 24.13, + "learning_rate": 9.798029414314195e-06, + "loss": 0.9043, + "step": 278770 + }, + { + "epoch": 24.13, + "learning_rate": 9.79658470340085e-06, + "loss": 0.9068, + "step": 278780 + }, + { + "epoch": 24.13, + "learning_rate": 9.795139992487504e-06, + "loss": 0.9015, + "step": 278790 + }, + { + "epoch": 24.13, + "learning_rate": 9.793695281574157e-06, + "loss": 0.8358, + "step": 278800 + }, + { + "epoch": 24.13, + "learning_rate": 9.792250570660812e-06, + "loss": 0.9356, + "step": 278810 + }, + { + "epoch": 24.13, + "learning_rate": 9.790805859747466e-06, + "loss": 0.9012, + "step": 278820 + }, + { + "epoch": 24.13, + "learning_rate": 9.789361148834119e-06, + "loss": 0.8879, + "step": 278830 + }, + { + "epoch": 24.14, + "learning_rate": 9.787916437920772e-06, + "loss": 0.9132, + "step": 278840 + }, + { + "epoch": 24.14, + "learning_rate": 9.786471727007426e-06, + "loss": 0.8994, + "step": 278850 + }, + { + "epoch": 24.14, + "learning_rate": 9.78502701609408e-06, + "loss": 0.9107, + "step": 278860 + }, + { + "epoch": 24.14, + "learning_rate": 9.783582305180734e-06, + "loss": 0.9084, + "step": 278870 + }, + { + "epoch": 24.14, + "learning_rate": 9.782137594267388e-06, + "loss": 0.8877, + "step": 278880 + }, + { + "epoch": 24.14, + "learning_rate": 9.780692883354041e-06, + "loss": 0.9232, + "step": 278890 + }, + { + "epoch": 24.14, + "learning_rate": 9.779248172440696e-06, + "loss": 0.9268, + "step": 278900 + }, + { + "epoch": 24.14, + "learning_rate": 9.77780346152735e-06, + "loss": 0.8968, + "step": 278910 + }, + { + "epoch": 24.14, + "learning_rate": 9.776358750614003e-06, + "loss": 0.8837, + "step": 278920 + }, + { + "epoch": 24.14, + "learning_rate": 9.774914039700656e-06, + "loss": 0.9682, + "step": 278930 + }, + { + "epoch": 24.14, + "learning_rate": 9.77346932878731e-06, + "loss": 0.8904, + "step": 278940 + }, + { + "epoch": 24.15, + "learning_rate": 9.772024617873963e-06, + "loss": 0.8521, + "step": 278950 + }, + { + "epoch": 24.15, + "learning_rate": 9.770579906960618e-06, + "loss": 0.9353, + "step": 278960 + }, + { + "epoch": 24.15, + "learning_rate": 9.769135196047272e-06, + "loss": 0.9085, + "step": 278970 + }, + { + "epoch": 24.15, + "learning_rate": 9.767690485133925e-06, + "loss": 0.9342, + "step": 278980 + }, + { + "epoch": 24.15, + "learning_rate": 9.76624577422058e-06, + "loss": 0.9345, + "step": 278990 + }, + { + "epoch": 24.15, + "learning_rate": 9.764801063307233e-06, + "loss": 0.8992, + "step": 279000 + }, + { + "epoch": 24.15, + "learning_rate": 9.763356352393887e-06, + "loss": 0.8696, + "step": 279010 + }, + { + "epoch": 24.15, + "learning_rate": 9.76191164148054e-06, + "loss": 0.9278, + "step": 279020 + }, + { + "epoch": 24.15, + "learning_rate": 9.760466930567194e-06, + "loss": 0.8847, + "step": 279030 + }, + { + "epoch": 24.15, + "learning_rate": 9.759022219653847e-06, + "loss": 0.9807, + "step": 279040 + }, + { + "epoch": 24.15, + "learning_rate": 9.7575775087405e-06, + "loss": 0.8878, + "step": 279050 + }, + { + "epoch": 24.15, + "learning_rate": 9.756132797827156e-06, + "loss": 0.9124, + "step": 279060 + }, + { + "epoch": 24.16, + "learning_rate": 9.754688086913809e-06, + "loss": 0.9315, + "step": 279070 + }, + { + "epoch": 24.16, + "learning_rate": 9.753243376000462e-06, + "loss": 0.9374, + "step": 279080 + }, + { + "epoch": 24.16, + "learning_rate": 9.751798665087117e-06, + "loss": 0.9274, + "step": 279090 + }, + { + "epoch": 24.16, + "learning_rate": 9.75035395417377e-06, + "loss": 0.9753, + "step": 279100 + }, + { + "epoch": 24.16, + "learning_rate": 9.748909243260424e-06, + "loss": 0.9149, + "step": 279110 + }, + { + "epoch": 24.16, + "learning_rate": 9.747464532347078e-06, + "loss": 0.8843, + "step": 279120 + }, + { + "epoch": 24.16, + "learning_rate": 9.746019821433731e-06, + "loss": 0.9427, + "step": 279130 + }, + { + "epoch": 24.16, + "learning_rate": 9.744575110520384e-06, + "loss": 0.8683, + "step": 279140 + }, + { + "epoch": 24.16, + "learning_rate": 9.74313039960704e-06, + "loss": 0.8723, + "step": 279150 + }, + { + "epoch": 24.16, + "learning_rate": 9.741685688693693e-06, + "loss": 0.9374, + "step": 279160 + }, + { + "epoch": 24.16, + "learning_rate": 9.740240977780346e-06, + "loss": 0.9631, + "step": 279170 + }, + { + "epoch": 24.17, + "learning_rate": 9.738796266867001e-06, + "loss": 0.9399, + "step": 279180 + }, + { + "epoch": 24.17, + "learning_rate": 9.737351555953655e-06, + "loss": 0.9553, + "step": 279190 + }, + { + "epoch": 24.17, + "learning_rate": 9.735906845040308e-06, + "loss": 0.9163, + "step": 279200 + }, + { + "epoch": 24.17, + "learning_rate": 9.734462134126962e-06, + "loss": 0.8741, + "step": 279210 + }, + { + "epoch": 24.17, + "learning_rate": 9.733017423213615e-06, + "loss": 0.9117, + "step": 279220 + }, + { + "epoch": 24.17, + "learning_rate": 9.731572712300268e-06, + "loss": 0.9214, + "step": 279230 + }, + { + "epoch": 24.17, + "learning_rate": 9.730128001386923e-06, + "loss": 0.9401, + "step": 279240 + }, + { + "epoch": 24.17, + "learning_rate": 9.728683290473577e-06, + "loss": 0.8703, + "step": 279250 + }, + { + "epoch": 24.17, + "learning_rate": 9.72723857956023e-06, + "loss": 0.8706, + "step": 279260 + }, + { + "epoch": 24.17, + "learning_rate": 9.725793868646885e-06, + "loss": 0.9413, + "step": 279270 + }, + { + "epoch": 24.17, + "learning_rate": 9.724349157733539e-06, + "loss": 0.9285, + "step": 279280 + }, + { + "epoch": 24.17, + "learning_rate": 9.722904446820192e-06, + "loss": 0.9012, + "step": 279290 + }, + { + "epoch": 24.18, + "learning_rate": 9.721459735906846e-06, + "loss": 0.8756, + "step": 279300 + }, + { + "epoch": 24.18, + "learning_rate": 9.720015024993499e-06, + "loss": 0.8817, + "step": 279310 + }, + { + "epoch": 24.18, + "learning_rate": 9.718570314080152e-06, + "loss": 0.9693, + "step": 279320 + }, + { + "epoch": 24.18, + "learning_rate": 9.717125603166807e-06, + "loss": 0.8877, + "step": 279330 + }, + { + "epoch": 24.18, + "learning_rate": 9.71568089225346e-06, + "loss": 0.8919, + "step": 279340 + }, + { + "epoch": 24.18, + "learning_rate": 9.714236181340114e-06, + "loss": 0.8989, + "step": 279350 + }, + { + "epoch": 24.18, + "learning_rate": 9.71279147042677e-06, + "loss": 0.8618, + "step": 279360 + }, + { + "epoch": 24.18, + "learning_rate": 9.711346759513423e-06, + "loss": 0.8665, + "step": 279370 + }, + { + "epoch": 24.18, + "learning_rate": 9.709902048600076e-06, + "loss": 0.9549, + "step": 279380 + }, + { + "epoch": 24.18, + "learning_rate": 9.70845733768673e-06, + "loss": 0.9238, + "step": 279390 + }, + { + "epoch": 24.18, + "learning_rate": 9.707012626773383e-06, + "loss": 0.9123, + "step": 279400 + }, + { + "epoch": 24.19, + "learning_rate": 9.705567915860036e-06, + "loss": 0.8897, + "step": 279410 + }, + { + "epoch": 24.19, + "learning_rate": 9.704123204946691e-06, + "loss": 0.9465, + "step": 279420 + }, + { + "epoch": 24.19, + "learning_rate": 9.702678494033345e-06, + "loss": 0.9231, + "step": 279430 + }, + { + "epoch": 24.19, + "learning_rate": 9.701233783119998e-06, + "loss": 0.9169, + "step": 279440 + }, + { + "epoch": 24.19, + "learning_rate": 9.699789072206653e-06, + "loss": 0.9254, + "step": 279450 + }, + { + "epoch": 24.19, + "learning_rate": 9.698344361293307e-06, + "loss": 0.9501, + "step": 279460 + }, + { + "epoch": 24.19, + "learning_rate": 9.69689965037996e-06, + "loss": 0.9406, + "step": 279470 + }, + { + "epoch": 24.19, + "learning_rate": 9.695454939466613e-06, + "loss": 0.9294, + "step": 279480 + }, + { + "epoch": 24.19, + "learning_rate": 9.694010228553267e-06, + "loss": 0.9023, + "step": 279490 + }, + { + "epoch": 24.19, + "learning_rate": 9.69256551763992e-06, + "loss": 0.9116, + "step": 279500 + }, + { + "epoch": 24.19, + "learning_rate": 9.691120806726574e-06, + "loss": 0.8868, + "step": 279510 + }, + { + "epoch": 24.19, + "learning_rate": 9.689676095813229e-06, + "loss": 0.8909, + "step": 279520 + }, + { + "epoch": 24.2, + "learning_rate": 9.688231384899882e-06, + "loss": 0.9669, + "step": 279530 + }, + { + "epoch": 24.2, + "learning_rate": 9.686786673986535e-06, + "loss": 0.9211, + "step": 279540 + }, + { + "epoch": 24.2, + "learning_rate": 9.68534196307319e-06, + "loss": 1.0063, + "step": 279550 + }, + { + "epoch": 24.2, + "learning_rate": 9.683897252159844e-06, + "loss": 0.9455, + "step": 279560 + }, + { + "epoch": 24.2, + "learning_rate": 9.682452541246496e-06, + "loss": 0.9256, + "step": 279570 + }, + { + "epoch": 24.2, + "learning_rate": 9.68100783033315e-06, + "loss": 0.9406, + "step": 279580 + }, + { + "epoch": 24.2, + "learning_rate": 9.679563119419804e-06, + "loss": 0.9083, + "step": 279590 + }, + { + "epoch": 24.2, + "learning_rate": 9.678118408506458e-06, + "loss": 0.9246, + "step": 279600 + }, + { + "epoch": 24.2, + "learning_rate": 9.676673697593113e-06, + "loss": 0.922, + "step": 279610 + }, + { + "epoch": 24.2, + "learning_rate": 9.675228986679766e-06, + "loss": 0.9771, + "step": 279620 + }, + { + "epoch": 24.2, + "learning_rate": 9.67378427576642e-06, + "loss": 0.9324, + "step": 279630 + }, + { + "epoch": 24.2, + "learning_rate": 9.672339564853074e-06, + "loss": 0.9544, + "step": 279640 + }, + { + "epoch": 24.21, + "learning_rate": 9.670894853939728e-06, + "loss": 0.9662, + "step": 279650 + }, + { + "epoch": 24.21, + "learning_rate": 9.66945014302638e-06, + "loss": 0.9338, + "step": 279660 + }, + { + "epoch": 24.21, + "learning_rate": 9.668005432113035e-06, + "loss": 0.9222, + "step": 279670 + }, + { + "epoch": 24.21, + "learning_rate": 9.666560721199688e-06, + "loss": 0.9357, + "step": 279680 + }, + { + "epoch": 24.21, + "learning_rate": 9.665116010286341e-06, + "loss": 0.9536, + "step": 279690 + }, + { + "epoch": 24.21, + "learning_rate": 9.663671299372997e-06, + "loss": 0.8637, + "step": 279700 + }, + { + "epoch": 24.21, + "learning_rate": 9.66222658845965e-06, + "loss": 0.9149, + "step": 279710 + }, + { + "epoch": 24.21, + "learning_rate": 9.660781877546303e-06, + "loss": 0.9871, + "step": 279720 + }, + { + "epoch": 24.21, + "learning_rate": 9.659337166632958e-06, + "loss": 0.9582, + "step": 279730 + }, + { + "epoch": 24.21, + "learning_rate": 9.657892455719612e-06, + "loss": 0.9323, + "step": 279740 + }, + { + "epoch": 24.21, + "learning_rate": 9.656447744806264e-06, + "loss": 0.9843, + "step": 279750 + }, + { + "epoch": 24.22, + "learning_rate": 9.655003033892919e-06, + "loss": 0.919, + "step": 279760 + }, + { + "epoch": 24.22, + "learning_rate": 9.653558322979572e-06, + "loss": 0.8484, + "step": 279770 + }, + { + "epoch": 24.22, + "learning_rate": 9.652113612066225e-06, + "loss": 0.8942, + "step": 279780 + }, + { + "epoch": 24.22, + "learning_rate": 9.65066890115288e-06, + "loss": 0.9352, + "step": 279790 + }, + { + "epoch": 24.22, + "learning_rate": 9.649224190239534e-06, + "loss": 0.9205, + "step": 279800 + }, + { + "epoch": 24.22, + "learning_rate": 9.647779479326187e-06, + "loss": 0.9444, + "step": 279810 + }, + { + "epoch": 24.22, + "learning_rate": 9.646334768412842e-06, + "loss": 0.873, + "step": 279820 + }, + { + "epoch": 24.22, + "learning_rate": 9.644890057499496e-06, + "loss": 0.9112, + "step": 279830 + }, + { + "epoch": 24.22, + "learning_rate": 9.643445346586147e-06, + "loss": 0.8853, + "step": 279840 + }, + { + "epoch": 24.22, + "learning_rate": 9.642000635672803e-06, + "loss": 0.9342, + "step": 279850 + }, + { + "epoch": 24.22, + "learning_rate": 9.640555924759456e-06, + "loss": 0.8789, + "step": 279860 + }, + { + "epoch": 24.22, + "learning_rate": 9.63911121384611e-06, + "loss": 0.9353, + "step": 279870 + }, + { + "epoch": 24.23, + "learning_rate": 9.637666502932764e-06, + "loss": 0.934, + "step": 279880 + }, + { + "epoch": 24.23, + "learning_rate": 9.636221792019418e-06, + "loss": 0.925, + "step": 279890 + }, + { + "epoch": 24.23, + "learning_rate": 9.634777081106071e-06, + "loss": 0.941, + "step": 279900 + }, + { + "epoch": 24.23, + "learning_rate": 9.633332370192726e-06, + "loss": 0.9276, + "step": 279910 + }, + { + "epoch": 24.23, + "learning_rate": 9.63188765927938e-06, + "loss": 0.8887, + "step": 279920 + }, + { + "epoch": 24.23, + "learning_rate": 9.630442948366031e-06, + "loss": 0.8772, + "step": 279930 + }, + { + "epoch": 24.23, + "learning_rate": 9.628998237452686e-06, + "loss": 0.9342, + "step": 279940 + }, + { + "epoch": 24.23, + "learning_rate": 9.62755352653934e-06, + "loss": 0.9153, + "step": 279950 + }, + { + "epoch": 24.23, + "learning_rate": 9.626108815625993e-06, + "loss": 0.9181, + "step": 279960 + }, + { + "epoch": 24.23, + "learning_rate": 9.624664104712647e-06, + "loss": 0.8604, + "step": 279970 + }, + { + "epoch": 24.23, + "learning_rate": 9.623219393799302e-06, + "loss": 0.886, + "step": 279980 + }, + { + "epoch": 24.24, + "learning_rate": 9.621774682885955e-06, + "loss": 0.9127, + "step": 279990 + }, + { + "epoch": 24.24, + "learning_rate": 9.620329971972609e-06, + "loss": 0.9468, + "step": 280000 + }, + { + "epoch": 24.24, + "learning_rate": 9.618885261059262e-06, + "loss": 0.9483, + "step": 280010 + }, + { + "epoch": 24.24, + "learning_rate": 9.617440550145915e-06, + "loss": 0.92, + "step": 280020 + }, + { + "epoch": 24.24, + "learning_rate": 9.615995839232569e-06, + "loss": 0.9821, + "step": 280030 + }, + { + "epoch": 24.24, + "learning_rate": 9.614551128319224e-06, + "loss": 0.8964, + "step": 280040 + }, + { + "epoch": 24.24, + "learning_rate": 9.613106417405877e-06, + "loss": 0.9824, + "step": 280050 + }, + { + "epoch": 24.24, + "learning_rate": 9.61166170649253e-06, + "loss": 0.9246, + "step": 280060 + }, + { + "epoch": 24.24, + "learning_rate": 9.610216995579186e-06, + "loss": 0.8878, + "step": 280070 + }, + { + "epoch": 24.24, + "learning_rate": 9.608772284665839e-06, + "loss": 0.952, + "step": 280080 + }, + { + "epoch": 24.24, + "learning_rate": 9.607327573752493e-06, + "loss": 0.9185, + "step": 280090 + }, + { + "epoch": 24.24, + "learning_rate": 9.605882862839146e-06, + "loss": 0.9111, + "step": 280100 + }, + { + "epoch": 24.25, + "learning_rate": 9.6044381519258e-06, + "loss": 0.8939, + "step": 280110 + }, + { + "epoch": 24.25, + "learning_rate": 9.602993441012453e-06, + "loss": 0.912, + "step": 280120 + }, + { + "epoch": 24.25, + "learning_rate": 9.601548730099108e-06, + "loss": 0.897, + "step": 280130 + }, + { + "epoch": 24.25, + "learning_rate": 9.600104019185761e-06, + "loss": 0.8852, + "step": 280140 + }, + { + "epoch": 24.25, + "learning_rate": 9.598659308272415e-06, + "loss": 0.905, + "step": 280150 + }, + { + "epoch": 24.25, + "learning_rate": 9.59721459735907e-06, + "loss": 0.8904, + "step": 280160 + }, + { + "epoch": 24.25, + "learning_rate": 9.595769886445723e-06, + "loss": 0.8965, + "step": 280170 + }, + { + "epoch": 24.25, + "learning_rate": 9.594325175532376e-06, + "loss": 0.9014, + "step": 280180 + }, + { + "epoch": 24.25, + "learning_rate": 9.59288046461903e-06, + "loss": 0.9135, + "step": 280190 + }, + { + "epoch": 24.25, + "learning_rate": 9.591435753705683e-06, + "loss": 0.9409, + "step": 280200 + }, + { + "epoch": 24.25, + "learning_rate": 9.589991042792337e-06, + "loss": 0.8536, + "step": 280210 + }, + { + "epoch": 24.26, + "learning_rate": 9.588546331878992e-06, + "loss": 0.9015, + "step": 280220 + }, + { + "epoch": 24.26, + "learning_rate": 9.587101620965645e-06, + "loss": 0.9466, + "step": 280230 + }, + { + "epoch": 24.26, + "learning_rate": 9.585656910052299e-06, + "loss": 0.9506, + "step": 280240 + }, + { + "epoch": 24.26, + "learning_rate": 9.584212199138954e-06, + "loss": 0.849, + "step": 280250 + }, + { + "epoch": 24.26, + "learning_rate": 9.582767488225607e-06, + "loss": 0.8936, + "step": 280260 + }, + { + "epoch": 24.26, + "learning_rate": 9.58132277731226e-06, + "loss": 0.9135, + "step": 280270 + }, + { + "epoch": 24.26, + "learning_rate": 9.579878066398914e-06, + "loss": 0.926, + "step": 280280 + }, + { + "epoch": 24.26, + "learning_rate": 9.578433355485567e-06, + "loss": 0.8785, + "step": 280290 + }, + { + "epoch": 24.26, + "learning_rate": 9.57698864457222e-06, + "loss": 0.9036, + "step": 280300 + }, + { + "epoch": 24.26, + "learning_rate": 9.575543933658876e-06, + "loss": 0.9786, + "step": 280310 + }, + { + "epoch": 24.26, + "learning_rate": 9.574099222745529e-06, + "loss": 0.9115, + "step": 280320 + }, + { + "epoch": 24.26, + "learning_rate": 9.572654511832182e-06, + "loss": 0.8591, + "step": 280330 + }, + { + "epoch": 24.27, + "learning_rate": 9.571209800918838e-06, + "loss": 0.8845, + "step": 280340 + }, + { + "epoch": 24.27, + "learning_rate": 9.569765090005491e-06, + "loss": 0.9843, + "step": 280350 + }, + { + "epoch": 24.27, + "learning_rate": 9.568320379092144e-06, + "loss": 0.9152, + "step": 280360 + }, + { + "epoch": 24.27, + "learning_rate": 9.566875668178798e-06, + "loss": 0.9043, + "step": 280370 + }, + { + "epoch": 24.27, + "learning_rate": 9.565430957265451e-06, + "loss": 0.9078, + "step": 280380 + }, + { + "epoch": 24.27, + "learning_rate": 9.563986246352105e-06, + "loss": 0.8864, + "step": 280390 + }, + { + "epoch": 24.27, + "learning_rate": 9.562541535438758e-06, + "loss": 0.885, + "step": 280400 + }, + { + "epoch": 24.27, + "learning_rate": 9.561096824525413e-06, + "loss": 0.8932, + "step": 280410 + }, + { + "epoch": 24.27, + "learning_rate": 9.559652113612066e-06, + "loss": 0.9594, + "step": 280420 + }, + { + "epoch": 24.27, + "learning_rate": 9.55820740269872e-06, + "loss": 0.9233, + "step": 280430 + }, + { + "epoch": 24.27, + "learning_rate": 9.556762691785375e-06, + "loss": 0.9503, + "step": 280440 + }, + { + "epoch": 24.28, + "learning_rate": 9.555317980872028e-06, + "loss": 0.8654, + "step": 280450 + }, + { + "epoch": 24.28, + "learning_rate": 9.553873269958682e-06, + "loss": 0.94, + "step": 280460 + }, + { + "epoch": 24.28, + "learning_rate": 9.552428559045335e-06, + "loss": 0.9401, + "step": 280470 + }, + { + "epoch": 24.28, + "learning_rate": 9.550983848131988e-06, + "loss": 0.9593, + "step": 280480 + }, + { + "epoch": 24.28, + "learning_rate": 9.549539137218642e-06, + "loss": 0.8717, + "step": 280490 + }, + { + "epoch": 24.28, + "learning_rate": 9.548094426305297e-06, + "loss": 0.8271, + "step": 280500 + }, + { + "epoch": 24.28, + "learning_rate": 9.54664971539195e-06, + "loss": 0.9721, + "step": 280510 + }, + { + "epoch": 24.28, + "learning_rate": 9.545205004478604e-06, + "loss": 0.9634, + "step": 280520 + }, + { + "epoch": 24.28, + "learning_rate": 9.543760293565259e-06, + "loss": 0.9486, + "step": 280530 + }, + { + "epoch": 24.28, + "learning_rate": 9.542315582651912e-06, + "loss": 0.9502, + "step": 280540 + }, + { + "epoch": 24.28, + "learning_rate": 9.540870871738566e-06, + "loss": 0.9597, + "step": 280550 + }, + { + "epoch": 24.28, + "learning_rate": 9.539426160825219e-06, + "loss": 0.9193, + "step": 280560 + }, + { + "epoch": 24.29, + "learning_rate": 9.537981449911872e-06, + "loss": 0.8682, + "step": 280570 + }, + { + "epoch": 24.29, + "learning_rate": 9.536536738998526e-06, + "loss": 0.918, + "step": 280580 + }, + { + "epoch": 24.29, + "learning_rate": 9.535092028085181e-06, + "loss": 0.8737, + "step": 280590 + }, + { + "epoch": 24.29, + "learning_rate": 9.533647317171834e-06, + "loss": 0.8752, + "step": 280600 + }, + { + "epoch": 24.29, + "learning_rate": 9.532202606258488e-06, + "loss": 0.8459, + "step": 280610 + }, + { + "epoch": 24.29, + "learning_rate": 9.530757895345143e-06, + "loss": 0.8664, + "step": 280620 + }, + { + "epoch": 24.29, + "learning_rate": 9.529313184431796e-06, + "loss": 0.9375, + "step": 280630 + }, + { + "epoch": 24.29, + "learning_rate": 9.52786847351845e-06, + "loss": 0.8746, + "step": 280640 + }, + { + "epoch": 24.29, + "learning_rate": 9.526423762605103e-06, + "loss": 0.8178, + "step": 280650 + }, + { + "epoch": 24.29, + "learning_rate": 9.524979051691756e-06, + "loss": 0.8814, + "step": 280660 + }, + { + "epoch": 24.29, + "learning_rate": 9.52353434077841e-06, + "loss": 0.891, + "step": 280670 + }, + { + "epoch": 24.29, + "learning_rate": 9.522089629865065e-06, + "loss": 0.9362, + "step": 280680 + }, + { + "epoch": 24.3, + "learning_rate": 9.520644918951718e-06, + "loss": 0.9364, + "step": 280690 + }, + { + "epoch": 24.3, + "learning_rate": 9.519200208038372e-06, + "loss": 0.9134, + "step": 280700 + }, + { + "epoch": 24.3, + "learning_rate": 9.517755497125027e-06, + "loss": 0.9693, + "step": 280710 + }, + { + "epoch": 24.3, + "learning_rate": 9.51631078621168e-06, + "loss": 0.9383, + "step": 280720 + }, + { + "epoch": 24.3, + "learning_rate": 9.514866075298333e-06, + "loss": 0.8964, + "step": 280730 + }, + { + "epoch": 24.3, + "learning_rate": 9.513421364384987e-06, + "loss": 0.8902, + "step": 280740 + }, + { + "epoch": 24.3, + "learning_rate": 9.51197665347164e-06, + "loss": 0.9365, + "step": 280750 + }, + { + "epoch": 24.3, + "learning_rate": 9.510531942558294e-06, + "loss": 0.9266, + "step": 280760 + }, + { + "epoch": 24.3, + "learning_rate": 9.509087231644949e-06, + "loss": 0.9463, + "step": 280770 + }, + { + "epoch": 24.3, + "learning_rate": 9.507642520731602e-06, + "loss": 0.9404, + "step": 280780 + }, + { + "epoch": 24.3, + "learning_rate": 9.506197809818256e-06, + "loss": 0.947, + "step": 280790 + }, + { + "epoch": 24.31, + "learning_rate": 9.50475309890491e-06, + "loss": 0.9372, + "step": 280800 + }, + { + "epoch": 24.31, + "learning_rate": 9.503308387991564e-06, + "loss": 0.9036, + "step": 280810 + }, + { + "epoch": 24.31, + "learning_rate": 9.501863677078217e-06, + "loss": 0.924, + "step": 280820 + }, + { + "epoch": 24.31, + "learning_rate": 9.50041896616487e-06, + "loss": 0.954, + "step": 280830 + }, + { + "epoch": 24.31, + "learning_rate": 9.498974255251524e-06, + "loss": 0.8674, + "step": 280840 + }, + { + "epoch": 24.31, + "learning_rate": 9.497529544338178e-06, + "loss": 0.9351, + "step": 280850 + }, + { + "epoch": 24.31, + "learning_rate": 9.496084833424831e-06, + "loss": 0.9397, + "step": 280860 + }, + { + "epoch": 24.31, + "learning_rate": 9.494640122511486e-06, + "loss": 0.9191, + "step": 280870 + }, + { + "epoch": 24.31, + "learning_rate": 9.49319541159814e-06, + "loss": 0.9015, + "step": 280880 + }, + { + "epoch": 24.31, + "learning_rate": 9.491750700684793e-06, + "loss": 0.9029, + "step": 280890 + }, + { + "epoch": 24.31, + "learning_rate": 9.490305989771448e-06, + "loss": 0.8472, + "step": 280900 + }, + { + "epoch": 24.31, + "learning_rate": 9.488861278858101e-06, + "loss": 0.9099, + "step": 280910 + }, + { + "epoch": 24.32, + "learning_rate": 9.487416567944755e-06, + "loss": 0.9428, + "step": 280920 + }, + { + "epoch": 24.32, + "learning_rate": 9.485971857031408e-06, + "loss": 0.8876, + "step": 280930 + }, + { + "epoch": 24.32, + "learning_rate": 9.484527146118062e-06, + "loss": 0.9041, + "step": 280940 + }, + { + "epoch": 24.32, + "learning_rate": 9.483082435204715e-06, + "loss": 0.9409, + "step": 280950 + }, + { + "epoch": 24.32, + "learning_rate": 9.48163772429137e-06, + "loss": 0.9005, + "step": 280960 + }, + { + "epoch": 24.32, + "learning_rate": 9.480193013378023e-06, + "loss": 0.9446, + "step": 280970 + }, + { + "epoch": 24.32, + "learning_rate": 9.478748302464677e-06, + "loss": 0.9177, + "step": 280980 + }, + { + "epoch": 24.32, + "learning_rate": 9.477303591551332e-06, + "loss": 0.9121, + "step": 280990 + }, + { + "epoch": 24.32, + "learning_rate": 9.475858880637985e-06, + "loss": 0.939, + "step": 281000 + }, + { + "epoch": 24.32, + "learning_rate": 9.474414169724639e-06, + "loss": 0.8889, + "step": 281010 + }, + { + "epoch": 24.32, + "learning_rate": 9.472969458811292e-06, + "loss": 0.8984, + "step": 281020 + }, + { + "epoch": 24.33, + "learning_rate": 9.471524747897946e-06, + "loss": 0.8742, + "step": 281030 + }, + { + "epoch": 24.33, + "learning_rate": 9.470080036984599e-06, + "loss": 0.9065, + "step": 281040 + }, + { + "epoch": 24.33, + "learning_rate": 9.468635326071254e-06, + "loss": 0.9126, + "step": 281050 + }, + { + "epoch": 24.33, + "learning_rate": 9.467190615157907e-06, + "loss": 0.9836, + "step": 281060 + }, + { + "epoch": 24.33, + "learning_rate": 9.46574590424456e-06, + "loss": 0.9385, + "step": 281070 + }, + { + "epoch": 24.33, + "learning_rate": 9.464301193331216e-06, + "loss": 0.917, + "step": 281080 + }, + { + "epoch": 24.33, + "learning_rate": 9.46285648241787e-06, + "loss": 0.944, + "step": 281090 + }, + { + "epoch": 24.33, + "learning_rate": 9.461411771504523e-06, + "loss": 0.9196, + "step": 281100 + }, + { + "epoch": 24.33, + "learning_rate": 9.459967060591176e-06, + "loss": 0.964, + "step": 281110 + }, + { + "epoch": 24.33, + "learning_rate": 9.45852234967783e-06, + "loss": 0.8883, + "step": 281120 + }, + { + "epoch": 24.33, + "learning_rate": 9.457077638764483e-06, + "loss": 0.9354, + "step": 281130 + }, + { + "epoch": 24.33, + "learning_rate": 9.455632927851138e-06, + "loss": 0.8934, + "step": 281140 + }, + { + "epoch": 24.34, + "learning_rate": 9.454188216937791e-06, + "loss": 0.9878, + "step": 281150 + }, + { + "epoch": 24.34, + "learning_rate": 9.452743506024445e-06, + "loss": 0.8688, + "step": 281160 + }, + { + "epoch": 24.34, + "learning_rate": 9.4512987951111e-06, + "loss": 0.8909, + "step": 281170 + }, + { + "epoch": 24.34, + "learning_rate": 9.449854084197753e-06, + "loss": 0.9257, + "step": 281180 + }, + { + "epoch": 24.34, + "learning_rate": 9.448409373284407e-06, + "loss": 0.9601, + "step": 281190 + }, + { + "epoch": 24.34, + "learning_rate": 9.44696466237106e-06, + "loss": 0.8976, + "step": 281200 + }, + { + "epoch": 24.34, + "learning_rate": 9.445519951457713e-06, + "loss": 0.9824, + "step": 281210 + }, + { + "epoch": 24.34, + "learning_rate": 9.444075240544367e-06, + "loss": 0.9193, + "step": 281220 + }, + { + "epoch": 24.34, + "learning_rate": 9.442630529631022e-06, + "loss": 0.9131, + "step": 281230 + }, + { + "epoch": 24.34, + "learning_rate": 9.441185818717675e-06, + "loss": 0.8667, + "step": 281240 + }, + { + "epoch": 24.34, + "learning_rate": 9.439741107804329e-06, + "loss": 0.9205, + "step": 281250 + }, + { + "epoch": 24.35, + "learning_rate": 9.438296396890984e-06, + "loss": 0.9052, + "step": 281260 + }, + { + "epoch": 24.35, + "learning_rate": 9.436851685977637e-06, + "loss": 0.8933, + "step": 281270 + }, + { + "epoch": 24.35, + "learning_rate": 9.43540697506429e-06, + "loss": 0.9186, + "step": 281280 + }, + { + "epoch": 24.35, + "learning_rate": 9.433962264150944e-06, + "loss": 0.9539, + "step": 281290 + }, + { + "epoch": 24.35, + "learning_rate": 9.432517553237597e-06, + "loss": 0.8966, + "step": 281300 + }, + { + "epoch": 24.35, + "learning_rate": 9.43107284232425e-06, + "loss": 0.8465, + "step": 281310 + }, + { + "epoch": 24.35, + "learning_rate": 9.429628131410904e-06, + "loss": 0.8591, + "step": 281320 + }, + { + "epoch": 24.35, + "learning_rate": 9.42818342049756e-06, + "loss": 0.9115, + "step": 281330 + }, + { + "epoch": 24.35, + "learning_rate": 9.426738709584213e-06, + "loss": 0.886, + "step": 281340 + }, + { + "epoch": 24.35, + "learning_rate": 9.425293998670866e-06, + "loss": 0.9175, + "step": 281350 + }, + { + "epoch": 24.35, + "learning_rate": 9.423849287757521e-06, + "loss": 0.9238, + "step": 281360 + }, + { + "epoch": 24.35, + "learning_rate": 9.422404576844174e-06, + "loss": 0.9073, + "step": 281370 + }, + { + "epoch": 24.36, + "learning_rate": 9.420959865930828e-06, + "loss": 0.9925, + "step": 281380 + }, + { + "epoch": 24.36, + "learning_rate": 9.419515155017481e-06, + "loss": 0.9053, + "step": 281390 + }, + { + "epoch": 24.36, + "learning_rate": 9.418070444104135e-06, + "loss": 0.9389, + "step": 281400 + }, + { + "epoch": 24.36, + "learning_rate": 9.416625733190788e-06, + "loss": 0.9111, + "step": 281410 + }, + { + "epoch": 24.36, + "learning_rate": 9.415181022277443e-06, + "loss": 0.9315, + "step": 281420 + }, + { + "epoch": 24.36, + "learning_rate": 9.413736311364097e-06, + "loss": 0.9456, + "step": 281430 + }, + { + "epoch": 24.36, + "learning_rate": 9.41229160045075e-06, + "loss": 0.9113, + "step": 281440 + }, + { + "epoch": 24.36, + "learning_rate": 9.410846889537405e-06, + "loss": 0.9011, + "step": 281450 + }, + { + "epoch": 24.36, + "learning_rate": 9.409402178624058e-06, + "loss": 0.8836, + "step": 281460 + }, + { + "epoch": 24.36, + "learning_rate": 9.407957467710712e-06, + "loss": 0.9586, + "step": 281470 + }, + { + "epoch": 24.36, + "learning_rate": 9.406512756797365e-06, + "loss": 0.8825, + "step": 281480 + }, + { + "epoch": 24.37, + "learning_rate": 9.405068045884019e-06, + "loss": 0.9036, + "step": 281490 + }, + { + "epoch": 24.37, + "learning_rate": 9.403623334970672e-06, + "loss": 0.8806, + "step": 281500 + }, + { + "epoch": 24.37, + "learning_rate": 9.402178624057327e-06, + "loss": 0.9413, + "step": 281510 + }, + { + "epoch": 24.37, + "learning_rate": 9.40073391314398e-06, + "loss": 0.884, + "step": 281520 + }, + { + "epoch": 24.37, + "learning_rate": 9.399289202230634e-06, + "loss": 0.8824, + "step": 281530 + }, + { + "epoch": 24.37, + "learning_rate": 9.397844491317289e-06, + "loss": 0.9546, + "step": 281540 + }, + { + "epoch": 24.37, + "learning_rate": 9.396399780403942e-06, + "loss": 0.9066, + "step": 281550 + }, + { + "epoch": 24.37, + "learning_rate": 9.394955069490596e-06, + "loss": 0.9296, + "step": 281560 + }, + { + "epoch": 24.37, + "learning_rate": 9.39351035857725e-06, + "loss": 0.9479, + "step": 281570 + }, + { + "epoch": 24.37, + "learning_rate": 9.392065647663903e-06, + "loss": 0.9378, + "step": 281580 + }, + { + "epoch": 24.37, + "learning_rate": 9.390620936750556e-06, + "loss": 0.9074, + "step": 281590 + }, + { + "epoch": 24.37, + "learning_rate": 9.389176225837211e-06, + "loss": 0.9208, + "step": 281600 + }, + { + "epoch": 24.38, + "learning_rate": 9.387731514923864e-06, + "loss": 0.8977, + "step": 281610 + }, + { + "epoch": 24.38, + "learning_rate": 9.386286804010518e-06, + "loss": 0.9831, + "step": 281620 + }, + { + "epoch": 24.38, + "learning_rate": 9.384842093097173e-06, + "loss": 0.8858, + "step": 281630 + }, + { + "epoch": 24.38, + "learning_rate": 9.383397382183826e-06, + "loss": 0.9579, + "step": 281640 + }, + { + "epoch": 24.38, + "learning_rate": 9.38195267127048e-06, + "loss": 0.908, + "step": 281650 + }, + { + "epoch": 24.38, + "learning_rate": 9.380507960357133e-06, + "loss": 0.9511, + "step": 281660 + }, + { + "epoch": 24.38, + "learning_rate": 9.379063249443787e-06, + "loss": 0.8814, + "step": 281670 + }, + { + "epoch": 24.38, + "learning_rate": 9.37761853853044e-06, + "loss": 0.9491, + "step": 281680 + }, + { + "epoch": 24.38, + "learning_rate": 9.376173827617095e-06, + "loss": 0.9081, + "step": 281690 + }, + { + "epoch": 24.38, + "learning_rate": 9.374729116703748e-06, + "loss": 0.8899, + "step": 281700 + }, + { + "epoch": 24.38, + "learning_rate": 9.373284405790402e-06, + "loss": 0.9146, + "step": 281710 + }, + { + "epoch": 24.38, + "learning_rate": 9.371839694877057e-06, + "loss": 0.9053, + "step": 281720 + }, + { + "epoch": 24.39, + "learning_rate": 9.37039498396371e-06, + "loss": 0.9623, + "step": 281730 + }, + { + "epoch": 24.39, + "learning_rate": 9.368950273050364e-06, + "loss": 0.896, + "step": 281740 + }, + { + "epoch": 24.39, + "learning_rate": 9.367505562137017e-06, + "loss": 0.9143, + "step": 281750 + }, + { + "epoch": 24.39, + "learning_rate": 9.36606085122367e-06, + "loss": 0.8851, + "step": 281760 + }, + { + "epoch": 24.39, + "learning_rate": 9.364616140310324e-06, + "loss": 0.915, + "step": 281770 + }, + { + "epoch": 24.39, + "learning_rate": 9.363171429396977e-06, + "loss": 0.9016, + "step": 281780 + }, + { + "epoch": 24.39, + "learning_rate": 9.361726718483632e-06, + "loss": 0.9166, + "step": 281790 + }, + { + "epoch": 24.39, + "learning_rate": 9.360282007570286e-06, + "loss": 0.9444, + "step": 281800 + }, + { + "epoch": 24.39, + "learning_rate": 9.358837296656939e-06, + "loss": 0.8892, + "step": 281810 + }, + { + "epoch": 24.39, + "learning_rate": 9.357392585743594e-06, + "loss": 0.9005, + "step": 281820 + }, + { + "epoch": 24.39, + "learning_rate": 9.355947874830248e-06, + "loss": 0.9361, + "step": 281830 + }, + { + "epoch": 24.4, + "learning_rate": 9.3545031639169e-06, + "loss": 0.9346, + "step": 281840 + }, + { + "epoch": 24.4, + "learning_rate": 9.353058453003554e-06, + "loss": 0.8876, + "step": 281850 + }, + { + "epoch": 24.4, + "learning_rate": 9.351613742090208e-06, + "loss": 0.9139, + "step": 281860 + }, + { + "epoch": 24.4, + "learning_rate": 9.350169031176861e-06, + "loss": 0.9453, + "step": 281870 + }, + { + "epoch": 24.4, + "learning_rate": 9.348724320263516e-06, + "loss": 0.9155, + "step": 281880 + }, + { + "epoch": 24.4, + "learning_rate": 9.34727960935017e-06, + "loss": 0.971, + "step": 281890 + }, + { + "epoch": 24.4, + "learning_rate": 9.345834898436823e-06, + "loss": 0.9227, + "step": 281900 + }, + { + "epoch": 24.4, + "learning_rate": 9.344390187523478e-06, + "loss": 0.8599, + "step": 281910 + }, + { + "epoch": 24.4, + "learning_rate": 9.342945476610132e-06, + "loss": 0.9162, + "step": 281920 + }, + { + "epoch": 24.4, + "learning_rate": 9.341500765696783e-06, + "loss": 0.9252, + "step": 281930 + }, + { + "epoch": 24.4, + "learning_rate": 9.340056054783438e-06, + "loss": 0.8907, + "step": 281940 + }, + { + "epoch": 24.4, + "learning_rate": 9.338611343870092e-06, + "loss": 0.9525, + "step": 281950 + }, + { + "epoch": 24.41, + "learning_rate": 9.337166632956745e-06, + "loss": 0.8789, + "step": 281960 + }, + { + "epoch": 24.41, + "learning_rate": 9.3357219220434e-06, + "loss": 0.8646, + "step": 281970 + }, + { + "epoch": 24.41, + "learning_rate": 9.334277211130054e-06, + "loss": 0.8957, + "step": 281980 + }, + { + "epoch": 24.41, + "learning_rate": 9.332832500216707e-06, + "loss": 0.8669, + "step": 281990 + }, + { + "epoch": 24.41, + "learning_rate": 9.331387789303362e-06, + "loss": 0.9288, + "step": 282000 + }, + { + "epoch": 24.41, + "learning_rate": 9.329943078390014e-06, + "loss": 0.9224, + "step": 282010 + }, + { + "epoch": 24.41, + "learning_rate": 9.328498367476667e-06, + "loss": 0.8862, + "step": 282020 + }, + { + "epoch": 24.41, + "learning_rate": 9.327053656563322e-06, + "loss": 0.9138, + "step": 282030 + }, + { + "epoch": 24.41, + "learning_rate": 9.325608945649976e-06, + "loss": 0.9109, + "step": 282040 + }, + { + "epoch": 24.41, + "learning_rate": 9.324164234736629e-06, + "loss": 1.011, + "step": 282050 + }, + { + "epoch": 24.41, + "learning_rate": 9.322719523823284e-06, + "loss": 0.9707, + "step": 282060 + }, + { + "epoch": 24.42, + "learning_rate": 9.321274812909938e-06, + "loss": 0.945, + "step": 282070 + }, + { + "epoch": 24.42, + "learning_rate": 9.319830101996591e-06, + "loss": 0.9292, + "step": 282080 + }, + { + "epoch": 24.42, + "learning_rate": 9.318385391083246e-06, + "loss": 0.9213, + "step": 282090 + }, + { + "epoch": 24.42, + "learning_rate": 9.316940680169898e-06, + "loss": 0.8847, + "step": 282100 + }, + { + "epoch": 24.42, + "learning_rate": 9.315495969256551e-06, + "loss": 0.9328, + "step": 282110 + }, + { + "epoch": 24.42, + "learning_rate": 9.314051258343206e-06, + "loss": 0.8846, + "step": 282120 + }, + { + "epoch": 24.42, + "learning_rate": 9.31260654742986e-06, + "loss": 0.9115, + "step": 282130 + }, + { + "epoch": 24.42, + "learning_rate": 9.311161836516513e-06, + "loss": 0.9193, + "step": 282140 + }, + { + "epoch": 24.42, + "learning_rate": 9.309717125603168e-06, + "loss": 0.883, + "step": 282150 + }, + { + "epoch": 24.42, + "learning_rate": 9.308272414689821e-06, + "loss": 0.9266, + "step": 282160 + }, + { + "epoch": 24.42, + "learning_rate": 9.306827703776475e-06, + "loss": 0.895, + "step": 282170 + }, + { + "epoch": 24.42, + "learning_rate": 9.30538299286313e-06, + "loss": 0.9168, + "step": 282180 + }, + { + "epoch": 24.43, + "learning_rate": 9.303938281949782e-06, + "loss": 0.9371, + "step": 282190 + }, + { + "epoch": 24.43, + "learning_rate": 9.302493571036435e-06, + "loss": 0.9195, + "step": 282200 + }, + { + "epoch": 24.43, + "learning_rate": 9.301048860123088e-06, + "loss": 0.949, + "step": 282210 + }, + { + "epoch": 24.43, + "learning_rate": 9.299604149209744e-06, + "loss": 0.9452, + "step": 282220 + }, + { + "epoch": 24.43, + "learning_rate": 9.298159438296397e-06, + "loss": 0.8903, + "step": 282230 + }, + { + "epoch": 24.43, + "learning_rate": 9.29671472738305e-06, + "loss": 0.8917, + "step": 282240 + }, + { + "epoch": 24.43, + "learning_rate": 9.295270016469705e-06, + "loss": 0.9442, + "step": 282250 + }, + { + "epoch": 24.43, + "learning_rate": 9.293825305556359e-06, + "loss": 0.8794, + "step": 282260 + }, + { + "epoch": 24.43, + "learning_rate": 9.292380594643012e-06, + "loss": 0.9006, + "step": 282270 + }, + { + "epoch": 24.43, + "learning_rate": 9.290935883729666e-06, + "loss": 0.9144, + "step": 282280 + }, + { + "epoch": 24.43, + "learning_rate": 9.289491172816319e-06, + "loss": 0.9226, + "step": 282290 + }, + { + "epoch": 24.44, + "learning_rate": 9.288046461902972e-06, + "loss": 0.9334, + "step": 282300 + }, + { + "epoch": 24.44, + "learning_rate": 9.286601750989628e-06, + "loss": 0.9016, + "step": 282310 + }, + { + "epoch": 24.44, + "learning_rate": 9.285157040076281e-06, + "loss": 0.9255, + "step": 282320 + }, + { + "epoch": 24.44, + "learning_rate": 9.283712329162934e-06, + "loss": 0.9518, + "step": 282330 + }, + { + "epoch": 24.44, + "learning_rate": 9.28226761824959e-06, + "loss": 0.9455, + "step": 282340 + }, + { + "epoch": 24.44, + "learning_rate": 9.280822907336243e-06, + "loss": 0.9633, + "step": 282350 + }, + { + "epoch": 24.44, + "learning_rate": 9.279378196422896e-06, + "loss": 0.9035, + "step": 282360 + }, + { + "epoch": 24.44, + "learning_rate": 9.27793348550955e-06, + "loss": 0.9312, + "step": 282370 + }, + { + "epoch": 24.44, + "learning_rate": 9.276488774596203e-06, + "loss": 0.927, + "step": 282380 + }, + { + "epoch": 24.44, + "learning_rate": 9.275044063682856e-06, + "loss": 0.947, + "step": 282390 + }, + { + "epoch": 24.44, + "learning_rate": 9.273599352769511e-06, + "loss": 0.8464, + "step": 282400 + }, + { + "epoch": 24.44, + "learning_rate": 9.272154641856165e-06, + "loss": 0.9037, + "step": 282410 + }, + { + "epoch": 24.45, + "learning_rate": 9.270709930942818e-06, + "loss": 0.9176, + "step": 282420 + }, + { + "epoch": 24.45, + "learning_rate": 9.269265220029473e-06, + "loss": 0.9086, + "step": 282430 + }, + { + "epoch": 24.45, + "learning_rate": 9.267820509116127e-06, + "loss": 0.9193, + "step": 282440 + }, + { + "epoch": 24.45, + "learning_rate": 9.26637579820278e-06, + "loss": 0.9394, + "step": 282450 + }, + { + "epoch": 24.45, + "learning_rate": 9.264931087289434e-06, + "loss": 0.9243, + "step": 282460 + }, + { + "epoch": 24.45, + "learning_rate": 9.263486376376087e-06, + "loss": 0.917, + "step": 282470 + }, + { + "epoch": 24.45, + "learning_rate": 9.26204166546274e-06, + "loss": 0.8728, + "step": 282480 + }, + { + "epoch": 24.45, + "learning_rate": 9.260596954549395e-06, + "loss": 0.9525, + "step": 282490 + }, + { + "epoch": 24.45, + "learning_rate": 9.259152243636049e-06, + "loss": 0.9348, + "step": 282500 + }, + { + "epoch": 24.45, + "learning_rate": 9.257707532722702e-06, + "loss": 0.9121, + "step": 282510 + }, + { + "epoch": 24.45, + "learning_rate": 9.256262821809357e-06, + "loss": 0.9291, + "step": 282520 + }, + { + "epoch": 24.46, + "learning_rate": 9.25481811089601e-06, + "loss": 0.8837, + "step": 282530 + }, + { + "epoch": 24.46, + "learning_rate": 9.253373399982664e-06, + "loss": 0.9191, + "step": 282540 + }, + { + "epoch": 24.46, + "learning_rate": 9.251928689069317e-06, + "loss": 0.9642, + "step": 282550 + }, + { + "epoch": 24.46, + "learning_rate": 9.25048397815597e-06, + "loss": 0.8409, + "step": 282560 + }, + { + "epoch": 24.46, + "learning_rate": 9.249039267242624e-06, + "loss": 0.9519, + "step": 282570 + }, + { + "epoch": 24.46, + "learning_rate": 9.24759455632928e-06, + "loss": 0.8612, + "step": 282580 + }, + { + "epoch": 24.46, + "learning_rate": 9.246149845415933e-06, + "loss": 0.8677, + "step": 282590 + }, + { + "epoch": 24.46, + "learning_rate": 9.244705134502586e-06, + "loss": 0.9164, + "step": 282600 + }, + { + "epoch": 24.46, + "learning_rate": 9.243260423589241e-06, + "loss": 0.968, + "step": 282610 + }, + { + "epoch": 24.46, + "learning_rate": 9.241815712675895e-06, + "loss": 0.9221, + "step": 282620 + }, + { + "epoch": 24.46, + "learning_rate": 9.240371001762548e-06, + "loss": 0.9218, + "step": 282630 + }, + { + "epoch": 24.46, + "learning_rate": 9.238926290849201e-06, + "loss": 0.9043, + "step": 282640 + }, + { + "epoch": 24.47, + "learning_rate": 9.237481579935855e-06, + "loss": 0.9066, + "step": 282650 + }, + { + "epoch": 24.47, + "learning_rate": 9.236036869022508e-06, + "loss": 0.9743, + "step": 282660 + }, + { + "epoch": 24.47, + "learning_rate": 9.234592158109162e-06, + "loss": 0.9333, + "step": 282670 + }, + { + "epoch": 24.47, + "learning_rate": 9.233147447195817e-06, + "loss": 0.8858, + "step": 282680 + }, + { + "epoch": 24.47, + "learning_rate": 9.23170273628247e-06, + "loss": 0.9291, + "step": 282690 + }, + { + "epoch": 24.47, + "learning_rate": 9.230258025369123e-06, + "loss": 0.9513, + "step": 282700 + }, + { + "epoch": 24.47, + "learning_rate": 9.228813314455779e-06, + "loss": 0.8953, + "step": 282710 + }, + { + "epoch": 24.47, + "learning_rate": 9.227368603542432e-06, + "loss": 0.9705, + "step": 282720 + }, + { + "epoch": 24.47, + "learning_rate": 9.225923892629085e-06, + "loss": 0.8446, + "step": 282730 + }, + { + "epoch": 24.47, + "learning_rate": 9.224479181715739e-06, + "loss": 0.8996, + "step": 282740 + }, + { + "epoch": 24.47, + "learning_rate": 9.223034470802392e-06, + "loss": 0.8438, + "step": 282750 + }, + { + "epoch": 24.48, + "learning_rate": 9.221589759889046e-06, + "loss": 0.9433, + "step": 282760 + }, + { + "epoch": 24.48, + "learning_rate": 9.2201450489757e-06, + "loss": 0.8675, + "step": 282770 + }, + { + "epoch": 24.48, + "learning_rate": 9.218700338062354e-06, + "loss": 0.8558, + "step": 282780 + }, + { + "epoch": 24.48, + "learning_rate": 9.217255627149007e-06, + "loss": 0.8438, + "step": 282790 + }, + { + "epoch": 24.48, + "learning_rate": 9.215810916235662e-06, + "loss": 0.9438, + "step": 282800 + }, + { + "epoch": 24.48, + "learning_rate": 9.214366205322316e-06, + "loss": 0.9757, + "step": 282810 + }, + { + "epoch": 24.48, + "learning_rate": 9.21292149440897e-06, + "loss": 0.8543, + "step": 282820 + }, + { + "epoch": 24.48, + "learning_rate": 9.211476783495623e-06, + "loss": 0.8669, + "step": 282830 + }, + { + "epoch": 24.48, + "learning_rate": 9.210032072582276e-06, + "loss": 0.9317, + "step": 282840 + }, + { + "epoch": 24.48, + "learning_rate": 9.20858736166893e-06, + "loss": 0.9442, + "step": 282850 + }, + { + "epoch": 24.48, + "learning_rate": 9.207142650755585e-06, + "loss": 0.9145, + "step": 282860 + }, + { + "epoch": 24.48, + "learning_rate": 9.205697939842238e-06, + "loss": 0.9145, + "step": 282870 + }, + { + "epoch": 24.49, + "learning_rate": 9.204253228928891e-06, + "loss": 0.9249, + "step": 282880 + }, + { + "epoch": 24.49, + "learning_rate": 9.202808518015546e-06, + "loss": 0.9307, + "step": 282890 + }, + { + "epoch": 24.49, + "learning_rate": 9.2013638071022e-06, + "loss": 0.9445, + "step": 282900 + }, + { + "epoch": 24.49, + "learning_rate": 9.199919096188853e-06, + "loss": 0.9277, + "step": 282910 + }, + { + "epoch": 24.49, + "learning_rate": 9.198474385275507e-06, + "loss": 0.9233, + "step": 282920 + }, + { + "epoch": 24.49, + "learning_rate": 9.19702967436216e-06, + "loss": 0.7818, + "step": 282930 + }, + { + "epoch": 24.49, + "learning_rate": 9.195584963448813e-06, + "loss": 0.8882, + "step": 282940 + }, + { + "epoch": 24.49, + "learning_rate": 9.194140252535468e-06, + "loss": 0.9364, + "step": 282950 + }, + { + "epoch": 24.49, + "learning_rate": 9.192695541622122e-06, + "loss": 0.9382, + "step": 282960 + }, + { + "epoch": 24.49, + "learning_rate": 9.191250830708775e-06, + "loss": 0.932, + "step": 282970 + }, + { + "epoch": 24.49, + "learning_rate": 9.18980611979543e-06, + "loss": 0.9134, + "step": 282980 + }, + { + "epoch": 24.49, + "learning_rate": 9.188361408882084e-06, + "loss": 0.8636, + "step": 282990 + }, + { + "epoch": 24.5, + "learning_rate": 9.186916697968737e-06, + "loss": 0.9358, + "step": 283000 + }, + { + "epoch": 24.5, + "learning_rate": 9.18547198705539e-06, + "loss": 0.8871, + "step": 283010 + }, + { + "epoch": 24.5, + "learning_rate": 9.184027276142044e-06, + "loss": 0.9337, + "step": 283020 + }, + { + "epoch": 24.5, + "learning_rate": 9.182582565228697e-06, + "loss": 0.8916, + "step": 283030 + }, + { + "epoch": 24.5, + "learning_rate": 9.181137854315352e-06, + "loss": 0.9304, + "step": 283040 + }, + { + "epoch": 24.5, + "learning_rate": 9.179693143402006e-06, + "loss": 0.9314, + "step": 283050 + }, + { + "epoch": 24.5, + "learning_rate": 9.17824843248866e-06, + "loss": 0.9275, + "step": 283060 + }, + { + "epoch": 24.5, + "learning_rate": 9.176803721575314e-06, + "loss": 0.9603, + "step": 283070 + }, + { + "epoch": 24.5, + "learning_rate": 9.175359010661968e-06, + "loss": 0.861, + "step": 283080 + }, + { + "epoch": 24.5, + "learning_rate": 9.173914299748621e-06, + "loss": 0.937, + "step": 283090 + }, + { + "epoch": 24.5, + "learning_rate": 9.172469588835275e-06, + "loss": 0.9897, + "step": 283100 + }, + { + "epoch": 24.51, + "learning_rate": 9.171024877921928e-06, + "loss": 0.897, + "step": 283110 + }, + { + "epoch": 24.51, + "learning_rate": 9.169580167008581e-06, + "loss": 0.9436, + "step": 283120 + }, + { + "epoch": 24.51, + "learning_rate": 9.168135456095235e-06, + "loss": 0.8831, + "step": 283130 + }, + { + "epoch": 24.51, + "learning_rate": 9.16669074518189e-06, + "loss": 0.9216, + "step": 283140 + }, + { + "epoch": 24.51, + "learning_rate": 9.165246034268543e-06, + "loss": 0.9385, + "step": 283150 + }, + { + "epoch": 24.51, + "learning_rate": 9.163801323355197e-06, + "loss": 0.8688, + "step": 283160 + }, + { + "epoch": 24.51, + "learning_rate": 9.162356612441852e-06, + "loss": 0.9187, + "step": 283170 + }, + { + "epoch": 24.51, + "learning_rate": 9.160911901528505e-06, + "loss": 0.8718, + "step": 283180 + }, + { + "epoch": 24.51, + "learning_rate": 9.159467190615158e-06, + "loss": 0.8712, + "step": 283190 + }, + { + "epoch": 24.51, + "learning_rate": 9.158022479701812e-06, + "loss": 0.939, + "step": 283200 + }, + { + "epoch": 24.51, + "learning_rate": 9.156577768788465e-06, + "loss": 0.8999, + "step": 283210 + }, + { + "epoch": 24.51, + "learning_rate": 9.155133057875119e-06, + "loss": 0.8865, + "step": 283220 + }, + { + "epoch": 24.52, + "learning_rate": 9.153688346961774e-06, + "loss": 0.9676, + "step": 283230 + }, + { + "epoch": 24.52, + "learning_rate": 9.152243636048427e-06, + "loss": 0.8529, + "step": 283240 + }, + { + "epoch": 24.52, + "learning_rate": 9.15079892513508e-06, + "loss": 0.8808, + "step": 283250 + }, + { + "epoch": 24.52, + "learning_rate": 9.149354214221736e-06, + "loss": 0.9415, + "step": 283260 + }, + { + "epoch": 24.52, + "learning_rate": 9.147909503308389e-06, + "loss": 0.9201, + "step": 283270 + }, + { + "epoch": 24.52, + "learning_rate": 9.146464792395042e-06, + "loss": 0.9049, + "step": 283280 + }, + { + "epoch": 24.52, + "learning_rate": 9.145020081481696e-06, + "loss": 0.9254, + "step": 283290 + }, + { + "epoch": 24.52, + "learning_rate": 9.14357537056835e-06, + "loss": 0.9245, + "step": 283300 + }, + { + "epoch": 24.52, + "learning_rate": 9.142130659655003e-06, + "loss": 0.9417, + "step": 283310 + }, + { + "epoch": 24.52, + "learning_rate": 9.140685948741658e-06, + "loss": 0.9678, + "step": 283320 + }, + { + "epoch": 24.52, + "learning_rate": 9.139241237828311e-06, + "loss": 0.9253, + "step": 283330 + }, + { + "epoch": 24.53, + "learning_rate": 9.137796526914964e-06, + "loss": 0.9378, + "step": 283340 + }, + { + "epoch": 24.53, + "learning_rate": 9.13635181600162e-06, + "loss": 0.933, + "step": 283350 + }, + { + "epoch": 24.53, + "learning_rate": 9.134907105088273e-06, + "loss": 0.9459, + "step": 283360 + }, + { + "epoch": 24.53, + "learning_rate": 9.133462394174926e-06, + "loss": 0.9317, + "step": 283370 + }, + { + "epoch": 24.53, + "learning_rate": 9.13201768326158e-06, + "loss": 0.8648, + "step": 283380 + }, + { + "epoch": 24.53, + "learning_rate": 9.130572972348233e-06, + "loss": 0.9488, + "step": 283390 + }, + { + "epoch": 24.53, + "learning_rate": 9.129128261434887e-06, + "loss": 0.9149, + "step": 283400 + }, + { + "epoch": 24.53, + "learning_rate": 9.127683550521542e-06, + "loss": 0.9308, + "step": 283410 + }, + { + "epoch": 24.53, + "learning_rate": 9.126238839608195e-06, + "loss": 0.958, + "step": 283420 + }, + { + "epoch": 24.53, + "learning_rate": 9.124794128694848e-06, + "loss": 0.8201, + "step": 283430 + }, + { + "epoch": 24.53, + "learning_rate": 9.123349417781503e-06, + "loss": 0.8776, + "step": 283440 + }, + { + "epoch": 24.53, + "learning_rate": 9.121904706868157e-06, + "loss": 0.9048, + "step": 283450 + }, + { + "epoch": 24.54, + "learning_rate": 9.12045999595481e-06, + "loss": 0.9049, + "step": 283460 + }, + { + "epoch": 24.54, + "learning_rate": 9.119015285041464e-06, + "loss": 0.9158, + "step": 283470 + }, + { + "epoch": 24.54, + "learning_rate": 9.117570574128117e-06, + "loss": 0.9486, + "step": 283480 + }, + { + "epoch": 24.54, + "learning_rate": 9.11612586321477e-06, + "loss": 0.9177, + "step": 283490 + }, + { + "epoch": 24.54, + "learning_rate": 9.114681152301426e-06, + "loss": 0.9596, + "step": 283500 + }, + { + "epoch": 24.54, + "learning_rate": 9.113236441388079e-06, + "loss": 0.8658, + "step": 283510 + }, + { + "epoch": 24.54, + "learning_rate": 9.111791730474732e-06, + "loss": 0.8626, + "step": 283520 + }, + { + "epoch": 24.54, + "learning_rate": 9.110347019561387e-06, + "loss": 0.866, + "step": 283530 + }, + { + "epoch": 24.54, + "learning_rate": 9.10890230864804e-06, + "loss": 0.8416, + "step": 283540 + }, + { + "epoch": 24.54, + "learning_rate": 9.107457597734694e-06, + "loss": 0.8819, + "step": 283550 + }, + { + "epoch": 24.54, + "learning_rate": 9.106012886821348e-06, + "loss": 0.9449, + "step": 283560 + }, + { + "epoch": 24.55, + "learning_rate": 9.104568175908001e-06, + "loss": 0.9144, + "step": 283570 + }, + { + "epoch": 24.55, + "learning_rate": 9.103123464994654e-06, + "loss": 0.8215, + "step": 283580 + }, + { + "epoch": 24.55, + "learning_rate": 9.101678754081308e-06, + "loss": 0.9136, + "step": 283590 + }, + { + "epoch": 24.55, + "learning_rate": 9.100234043167963e-06, + "loss": 0.9329, + "step": 283600 + }, + { + "epoch": 24.55, + "learning_rate": 9.098789332254616e-06, + "loss": 0.9609, + "step": 283610 + }, + { + "epoch": 24.55, + "learning_rate": 9.09734462134127e-06, + "loss": 0.8938, + "step": 283620 + }, + { + "epoch": 24.55, + "learning_rate": 9.095899910427925e-06, + "loss": 0.857, + "step": 283630 + }, + { + "epoch": 24.55, + "learning_rate": 9.094455199514578e-06, + "loss": 0.8969, + "step": 283640 + }, + { + "epoch": 24.55, + "learning_rate": 9.093010488601232e-06, + "loss": 0.9409, + "step": 283650 + }, + { + "epoch": 24.55, + "learning_rate": 9.091565777687885e-06, + "loss": 0.8974, + "step": 283660 + }, + { + "epoch": 24.55, + "learning_rate": 9.090121066774538e-06, + "loss": 0.896, + "step": 283670 + }, + { + "epoch": 24.55, + "learning_rate": 9.088676355861192e-06, + "loss": 0.9298, + "step": 283680 + }, + { + "epoch": 24.56, + "learning_rate": 9.087231644947847e-06, + "loss": 0.887, + "step": 283690 + }, + { + "epoch": 24.56, + "learning_rate": 9.0857869340345e-06, + "loss": 0.9111, + "step": 283700 + }, + { + "epoch": 24.56, + "learning_rate": 9.084342223121154e-06, + "loss": 0.9323, + "step": 283710 + }, + { + "epoch": 24.56, + "learning_rate": 9.082897512207809e-06, + "loss": 0.9221, + "step": 283720 + }, + { + "epoch": 24.56, + "learning_rate": 9.081452801294462e-06, + "loss": 0.9444, + "step": 283730 + }, + { + "epoch": 24.56, + "learning_rate": 9.080008090381115e-06, + "loss": 0.9541, + "step": 283740 + }, + { + "epoch": 24.56, + "learning_rate": 9.078563379467769e-06, + "loss": 0.944, + "step": 283750 + }, + { + "epoch": 24.56, + "learning_rate": 9.077118668554422e-06, + "loss": 0.8825, + "step": 283760 + }, + { + "epoch": 24.56, + "learning_rate": 9.075673957641076e-06, + "loss": 0.8581, + "step": 283770 + }, + { + "epoch": 24.56, + "learning_rate": 9.07422924672773e-06, + "loss": 0.8658, + "step": 283780 + }, + { + "epoch": 24.56, + "learning_rate": 9.072784535814384e-06, + "loss": 0.8753, + "step": 283790 + }, + { + "epoch": 24.57, + "learning_rate": 9.071339824901038e-06, + "loss": 0.937, + "step": 283800 + }, + { + "epoch": 24.57, + "learning_rate": 9.069895113987693e-06, + "loss": 0.9761, + "step": 283810 + }, + { + "epoch": 24.57, + "learning_rate": 9.068450403074346e-06, + "loss": 0.9424, + "step": 283820 + }, + { + "epoch": 24.57, + "learning_rate": 9.067005692161e-06, + "loss": 0.8289, + "step": 283830 + }, + { + "epoch": 24.57, + "learning_rate": 9.065560981247653e-06, + "loss": 0.9411, + "step": 283840 + }, + { + "epoch": 24.57, + "learning_rate": 9.064116270334306e-06, + "loss": 0.9114, + "step": 283850 + }, + { + "epoch": 24.57, + "learning_rate": 9.06267155942096e-06, + "loss": 0.8977, + "step": 283860 + }, + { + "epoch": 24.57, + "learning_rate": 9.061226848507615e-06, + "loss": 0.9341, + "step": 283870 + }, + { + "epoch": 24.57, + "learning_rate": 9.059782137594268e-06, + "loss": 0.8474, + "step": 283880 + }, + { + "epoch": 24.57, + "learning_rate": 9.058337426680922e-06, + "loss": 0.9242, + "step": 283890 + }, + { + "epoch": 24.57, + "learning_rate": 9.056892715767577e-06, + "loss": 0.929, + "step": 283900 + }, + { + "epoch": 24.57, + "learning_rate": 9.05544800485423e-06, + "loss": 0.8998, + "step": 283910 + }, + { + "epoch": 24.58, + "learning_rate": 9.054003293940883e-06, + "loss": 0.9219, + "step": 283920 + }, + { + "epoch": 24.58, + "learning_rate": 9.052558583027537e-06, + "loss": 0.8798, + "step": 283930 + }, + { + "epoch": 24.58, + "learning_rate": 9.05111387211419e-06, + "loss": 0.8673, + "step": 283940 + }, + { + "epoch": 24.58, + "learning_rate": 9.049669161200844e-06, + "loss": 0.9769, + "step": 283950 + }, + { + "epoch": 24.58, + "learning_rate": 9.048224450287499e-06, + "loss": 0.9294, + "step": 283960 + }, + { + "epoch": 24.58, + "learning_rate": 9.046779739374152e-06, + "loss": 0.9201, + "step": 283970 + }, + { + "epoch": 24.58, + "learning_rate": 9.045335028460805e-06, + "loss": 0.8989, + "step": 283980 + }, + { + "epoch": 24.58, + "learning_rate": 9.04389031754746e-06, + "loss": 0.924, + "step": 283990 + }, + { + "epoch": 24.58, + "learning_rate": 9.042445606634114e-06, + "loss": 0.8895, + "step": 284000 + }, + { + "epoch": 24.58, + "learning_rate": 9.041000895720766e-06, + "loss": 0.9434, + "step": 284010 + }, + { + "epoch": 24.58, + "learning_rate": 9.039556184807419e-06, + "loss": 0.9033, + "step": 284020 + }, + { + "epoch": 24.58, + "learning_rate": 9.038111473894074e-06, + "loss": 0.8728, + "step": 284030 + }, + { + "epoch": 24.59, + "learning_rate": 9.036666762980728e-06, + "loss": 0.9199, + "step": 284040 + }, + { + "epoch": 24.59, + "learning_rate": 9.035222052067381e-06, + "loss": 0.9133, + "step": 284050 + }, + { + "epoch": 24.59, + "learning_rate": 9.033777341154036e-06, + "loss": 0.8668, + "step": 284060 + }, + { + "epoch": 24.59, + "learning_rate": 9.03233263024069e-06, + "loss": 0.9744, + "step": 284070 + }, + { + "epoch": 24.59, + "learning_rate": 9.030887919327343e-06, + "loss": 0.8823, + "step": 284080 + }, + { + "epoch": 24.59, + "learning_rate": 9.029443208413998e-06, + "loss": 0.8856, + "step": 284090 + }, + { + "epoch": 24.59, + "learning_rate": 9.02799849750065e-06, + "loss": 0.9408, + "step": 284100 + }, + { + "epoch": 24.59, + "learning_rate": 9.026553786587303e-06, + "loss": 0.9343, + "step": 284110 + }, + { + "epoch": 24.59, + "learning_rate": 9.025109075673958e-06, + "loss": 0.9484, + "step": 284120 + }, + { + "epoch": 24.59, + "learning_rate": 9.023664364760611e-06, + "loss": 0.8855, + "step": 284130 + }, + { + "epoch": 24.59, + "learning_rate": 9.022219653847265e-06, + "loss": 0.8633, + "step": 284140 + }, + { + "epoch": 24.6, + "learning_rate": 9.02077494293392e-06, + "loss": 0.906, + "step": 284150 + }, + { + "epoch": 24.6, + "learning_rate": 9.019330232020573e-06, + "loss": 0.8929, + "step": 284160 + }, + { + "epoch": 24.6, + "learning_rate": 9.017885521107227e-06, + "loss": 0.9235, + "step": 284170 + }, + { + "epoch": 24.6, + "learning_rate": 9.016440810193882e-06, + "loss": 0.9323, + "step": 284180 + }, + { + "epoch": 24.6, + "learning_rate": 9.014996099280534e-06, + "loss": 0.9903, + "step": 284190 + }, + { + "epoch": 24.6, + "learning_rate": 9.013551388367187e-06, + "loss": 0.9883, + "step": 284200 + }, + { + "epoch": 24.6, + "learning_rate": 9.012106677453842e-06, + "loss": 0.9544, + "step": 284210 + }, + { + "epoch": 24.6, + "learning_rate": 9.010661966540495e-06, + "loss": 0.9576, + "step": 284220 + }, + { + "epoch": 24.6, + "learning_rate": 9.009217255627149e-06, + "loss": 0.9034, + "step": 284230 + }, + { + "epoch": 24.6, + "learning_rate": 9.007772544713804e-06, + "loss": 0.8732, + "step": 284240 + }, + { + "epoch": 24.6, + "learning_rate": 9.006327833800457e-06, + "loss": 0.8723, + "step": 284250 + }, + { + "epoch": 24.6, + "learning_rate": 9.00488312288711e-06, + "loss": 0.9595, + "step": 284260 + }, + { + "epoch": 24.61, + "learning_rate": 9.003438411973766e-06, + "loss": 0.9333, + "step": 284270 + }, + { + "epoch": 24.61, + "learning_rate": 9.001993701060417e-06, + "loss": 0.9164, + "step": 284280 + }, + { + "epoch": 24.61, + "learning_rate": 9.000548990147071e-06, + "loss": 0.8974, + "step": 284290 + }, + { + "epoch": 24.61, + "learning_rate": 8.999104279233726e-06, + "loss": 0.8743, + "step": 284300 + }, + { + "epoch": 24.61, + "learning_rate": 8.99765956832038e-06, + "loss": 0.9195, + "step": 284310 + }, + { + "epoch": 24.61, + "learning_rate": 8.996214857407033e-06, + "loss": 0.9373, + "step": 284320 + }, + { + "epoch": 24.61, + "learning_rate": 8.994770146493688e-06, + "loss": 0.873, + "step": 284330 + }, + { + "epoch": 24.61, + "learning_rate": 8.993325435580341e-06, + "loss": 0.8742, + "step": 284340 + }, + { + "epoch": 24.61, + "learning_rate": 8.991880724666995e-06, + "loss": 0.891, + "step": 284350 + }, + { + "epoch": 24.61, + "learning_rate": 8.99043601375365e-06, + "loss": 0.8813, + "step": 284360 + }, + { + "epoch": 24.61, + "learning_rate": 8.988991302840301e-06, + "loss": 0.9261, + "step": 284370 + }, + { + "epoch": 24.62, + "learning_rate": 8.987546591926955e-06, + "loss": 0.852, + "step": 284380 + }, + { + "epoch": 24.62, + "learning_rate": 8.98610188101361e-06, + "loss": 0.933, + "step": 284390 + }, + { + "epoch": 24.62, + "learning_rate": 8.984657170100263e-06, + "loss": 0.9752, + "step": 284400 + }, + { + "epoch": 24.62, + "learning_rate": 8.983212459186917e-06, + "loss": 0.8833, + "step": 284410 + }, + { + "epoch": 24.62, + "learning_rate": 8.981767748273572e-06, + "loss": 0.916, + "step": 284420 + }, + { + "epoch": 24.62, + "learning_rate": 8.980323037360225e-06, + "loss": 0.9635, + "step": 284430 + }, + { + "epoch": 24.62, + "learning_rate": 8.978878326446879e-06, + "loss": 0.9504, + "step": 284440 + }, + { + "epoch": 24.62, + "learning_rate": 8.977433615533534e-06, + "loss": 0.8963, + "step": 284450 + }, + { + "epoch": 24.62, + "learning_rate": 8.975988904620185e-06, + "loss": 0.9365, + "step": 284460 + }, + { + "epoch": 24.62, + "learning_rate": 8.974544193706839e-06, + "loss": 0.9034, + "step": 284470 + }, + { + "epoch": 24.62, + "learning_rate": 8.973099482793492e-06, + "loss": 0.9482, + "step": 284480 + }, + { + "epoch": 24.62, + "learning_rate": 8.971654771880147e-06, + "loss": 0.8762, + "step": 284490 + }, + { + "epoch": 24.63, + "learning_rate": 8.9702100609668e-06, + "loss": 0.9082, + "step": 284500 + }, + { + "epoch": 24.63, + "learning_rate": 8.968765350053454e-06, + "loss": 0.9162, + "step": 284510 + }, + { + "epoch": 24.63, + "learning_rate": 8.967320639140109e-06, + "loss": 0.9359, + "step": 284520 + }, + { + "epoch": 24.63, + "learning_rate": 8.965875928226762e-06, + "loss": 0.8686, + "step": 284530 + }, + { + "epoch": 24.63, + "learning_rate": 8.964431217313416e-06, + "loss": 0.9489, + "step": 284540 + }, + { + "epoch": 24.63, + "learning_rate": 8.96298650640007e-06, + "loss": 0.9274, + "step": 284550 + }, + { + "epoch": 24.63, + "learning_rate": 8.961541795486723e-06, + "loss": 0.8986, + "step": 284560 + }, + { + "epoch": 24.63, + "learning_rate": 8.960097084573376e-06, + "loss": 0.9164, + "step": 284570 + }, + { + "epoch": 24.63, + "learning_rate": 8.958652373660031e-06, + "loss": 0.9082, + "step": 284580 + }, + { + "epoch": 24.63, + "learning_rate": 8.957207662746685e-06, + "loss": 0.8508, + "step": 284590 + }, + { + "epoch": 24.63, + "learning_rate": 8.955762951833338e-06, + "loss": 0.9269, + "step": 284600 + }, + { + "epoch": 24.64, + "learning_rate": 8.954318240919993e-06, + "loss": 0.8866, + "step": 284610 + }, + { + "epoch": 24.64, + "learning_rate": 8.952873530006646e-06, + "loss": 0.8977, + "step": 284620 + }, + { + "epoch": 24.64, + "learning_rate": 8.9514288190933e-06, + "loss": 0.9041, + "step": 284630 + }, + { + "epoch": 24.64, + "learning_rate": 8.949984108179953e-06, + "loss": 0.8985, + "step": 284640 + }, + { + "epoch": 24.64, + "learning_rate": 8.948539397266607e-06, + "loss": 0.9625, + "step": 284650 + }, + { + "epoch": 24.64, + "learning_rate": 8.94709468635326e-06, + "loss": 0.932, + "step": 284660 + }, + { + "epoch": 24.64, + "learning_rate": 8.945649975439915e-06, + "loss": 0.9007, + "step": 284670 + }, + { + "epoch": 24.64, + "learning_rate": 8.944205264526569e-06, + "loss": 0.8831, + "step": 284680 + }, + { + "epoch": 24.64, + "learning_rate": 8.942760553613222e-06, + "loss": 0.9016, + "step": 284690 + }, + { + "epoch": 24.64, + "learning_rate": 8.941315842699877e-06, + "loss": 0.9326, + "step": 284700 + }, + { + "epoch": 24.64, + "learning_rate": 8.93987113178653e-06, + "loss": 0.9437, + "step": 284710 + }, + { + "epoch": 24.64, + "learning_rate": 8.938426420873184e-06, + "loss": 0.8859, + "step": 284720 + }, + { + "epoch": 24.65, + "learning_rate": 8.936981709959837e-06, + "loss": 0.8902, + "step": 284730 + }, + { + "epoch": 24.65, + "learning_rate": 8.93553699904649e-06, + "loss": 0.8844, + "step": 284740 + }, + { + "epoch": 24.65, + "learning_rate": 8.934092288133144e-06, + "loss": 0.8922, + "step": 284750 + }, + { + "epoch": 24.65, + "learning_rate": 8.932647577219799e-06, + "loss": 0.8619, + "step": 284760 + }, + { + "epoch": 24.65, + "learning_rate": 8.931202866306452e-06, + "loss": 0.9346, + "step": 284770 + }, + { + "epoch": 24.65, + "learning_rate": 8.929758155393106e-06, + "loss": 0.8602, + "step": 284780 + }, + { + "epoch": 24.65, + "learning_rate": 8.928313444479761e-06, + "loss": 0.9204, + "step": 284790 + }, + { + "epoch": 24.65, + "learning_rate": 8.926868733566414e-06, + "loss": 0.9487, + "step": 284800 + }, + { + "epoch": 24.65, + "learning_rate": 8.925424022653068e-06, + "loss": 0.8497, + "step": 284810 + }, + { + "epoch": 24.65, + "learning_rate": 8.923979311739721e-06, + "loss": 0.9145, + "step": 284820 + }, + { + "epoch": 24.65, + "learning_rate": 8.922534600826375e-06, + "loss": 0.9123, + "step": 284830 + }, + { + "epoch": 24.66, + "learning_rate": 8.921089889913028e-06, + "loss": 0.915, + "step": 284840 + }, + { + "epoch": 24.66, + "learning_rate": 8.919645178999683e-06, + "loss": 0.955, + "step": 284850 + }, + { + "epoch": 24.66, + "learning_rate": 8.918200468086336e-06, + "loss": 0.9639, + "step": 284860 + }, + { + "epoch": 24.66, + "learning_rate": 8.91675575717299e-06, + "loss": 0.9014, + "step": 284870 + }, + { + "epoch": 24.66, + "learning_rate": 8.915311046259645e-06, + "loss": 0.8933, + "step": 284880 + }, + { + "epoch": 24.66, + "learning_rate": 8.913866335346298e-06, + "loss": 0.9047, + "step": 284890 + }, + { + "epoch": 24.66, + "learning_rate": 8.912421624432952e-06, + "loss": 0.969, + "step": 284900 + }, + { + "epoch": 24.66, + "learning_rate": 8.910976913519605e-06, + "loss": 0.9077, + "step": 284910 + }, + { + "epoch": 24.66, + "learning_rate": 8.909532202606258e-06, + "loss": 0.8555, + "step": 284920 + }, + { + "epoch": 24.66, + "learning_rate": 8.908087491692912e-06, + "loss": 0.9281, + "step": 284930 + }, + { + "epoch": 24.66, + "learning_rate": 8.906642780779565e-06, + "loss": 0.992, + "step": 284940 + }, + { + "epoch": 24.66, + "learning_rate": 8.90519806986622e-06, + "loss": 0.9666, + "step": 284950 + }, + { + "epoch": 24.67, + "learning_rate": 8.903753358952874e-06, + "loss": 0.9696, + "step": 284960 + }, + { + "epoch": 24.67, + "learning_rate": 8.902308648039527e-06, + "loss": 0.9765, + "step": 284970 + }, + { + "epoch": 24.67, + "learning_rate": 8.900863937126182e-06, + "loss": 0.8809, + "step": 284980 + }, + { + "epoch": 24.67, + "learning_rate": 8.899419226212836e-06, + "loss": 0.872, + "step": 284990 + }, + { + "epoch": 24.67, + "learning_rate": 8.897974515299489e-06, + "loss": 0.8766, + "step": 285000 + }, + { + "epoch": 24.67, + "learning_rate": 8.896529804386142e-06, + "loss": 0.9116, + "step": 285010 + }, + { + "epoch": 24.67, + "learning_rate": 8.895085093472796e-06, + "loss": 0.9334, + "step": 285020 + }, + { + "epoch": 24.67, + "learning_rate": 8.89364038255945e-06, + "loss": 0.9423, + "step": 285030 + }, + { + "epoch": 24.67, + "learning_rate": 8.892195671646104e-06, + "loss": 0.9414, + "step": 285040 + }, + { + "epoch": 24.67, + "learning_rate": 8.890750960732758e-06, + "loss": 0.9181, + "step": 285050 + }, + { + "epoch": 24.67, + "learning_rate": 8.889306249819411e-06, + "loss": 0.9165, + "step": 285060 + }, + { + "epoch": 24.67, + "learning_rate": 8.887861538906066e-06, + "loss": 0.8728, + "step": 285070 + }, + { + "epoch": 24.68, + "learning_rate": 8.88641682799272e-06, + "loss": 0.9113, + "step": 285080 + }, + { + "epoch": 24.68, + "learning_rate": 8.884972117079373e-06, + "loss": 0.8566, + "step": 285090 + }, + { + "epoch": 24.68, + "learning_rate": 8.883527406166026e-06, + "loss": 0.9436, + "step": 285100 + }, + { + "epoch": 24.68, + "learning_rate": 8.88208269525268e-06, + "loss": 0.9439, + "step": 285110 + }, + { + "epoch": 24.68, + "learning_rate": 8.880637984339333e-06, + "loss": 0.9012, + "step": 285120 + }, + { + "epoch": 24.68, + "learning_rate": 8.879193273425988e-06, + "loss": 0.8903, + "step": 285130 + }, + { + "epoch": 24.68, + "learning_rate": 8.877748562512642e-06, + "loss": 0.9213, + "step": 285140 + }, + { + "epoch": 24.68, + "learning_rate": 8.876303851599295e-06, + "loss": 0.8583, + "step": 285150 + }, + { + "epoch": 24.68, + "learning_rate": 8.87485914068595e-06, + "loss": 0.8791, + "step": 285160 + }, + { + "epoch": 24.68, + "learning_rate": 8.873414429772603e-06, + "loss": 0.9664, + "step": 285170 + }, + { + "epoch": 24.68, + "learning_rate": 8.871969718859257e-06, + "loss": 0.871, + "step": 285180 + }, + { + "epoch": 24.69, + "learning_rate": 8.87052500794591e-06, + "loss": 0.9132, + "step": 285190 + }, + { + "epoch": 24.69, + "learning_rate": 8.869080297032564e-06, + "loss": 0.9003, + "step": 285200 + }, + { + "epoch": 24.69, + "learning_rate": 8.867635586119217e-06, + "loss": 0.9472, + "step": 285210 + }, + { + "epoch": 24.69, + "learning_rate": 8.866190875205872e-06, + "loss": 0.9606, + "step": 285220 + }, + { + "epoch": 24.69, + "learning_rate": 8.864746164292526e-06, + "loss": 0.9728, + "step": 285230 + }, + { + "epoch": 24.69, + "learning_rate": 8.863301453379179e-06, + "loss": 0.9295, + "step": 285240 + }, + { + "epoch": 24.69, + "learning_rate": 8.861856742465834e-06, + "loss": 0.8737, + "step": 285250 + }, + { + "epoch": 24.69, + "learning_rate": 8.860412031552487e-06, + "loss": 0.841, + "step": 285260 + }, + { + "epoch": 24.69, + "learning_rate": 8.85896732063914e-06, + "loss": 0.9265, + "step": 285270 + }, + { + "epoch": 24.69, + "learning_rate": 8.857522609725794e-06, + "loss": 0.9489, + "step": 285280 + }, + { + "epoch": 24.69, + "learning_rate": 8.856077898812448e-06, + "loss": 0.8525, + "step": 285290 + }, + { + "epoch": 24.69, + "learning_rate": 8.854633187899101e-06, + "loss": 0.8998, + "step": 285300 + }, + { + "epoch": 24.7, + "learning_rate": 8.853188476985756e-06, + "loss": 0.9094, + "step": 285310 + }, + { + "epoch": 24.7, + "learning_rate": 8.85174376607241e-06, + "loss": 0.9261, + "step": 285320 + }, + { + "epoch": 24.7, + "learning_rate": 8.850299055159063e-06, + "loss": 0.9403, + "step": 285330 + }, + { + "epoch": 24.7, + "learning_rate": 8.848854344245718e-06, + "loss": 0.9104, + "step": 285340 + }, + { + "epoch": 24.7, + "learning_rate": 8.847409633332371e-06, + "loss": 0.871, + "step": 285350 + }, + { + "epoch": 24.7, + "learning_rate": 8.845964922419025e-06, + "loss": 0.8818, + "step": 285360 + }, + { + "epoch": 24.7, + "learning_rate": 8.844520211505678e-06, + "loss": 0.8689, + "step": 285370 + }, + { + "epoch": 24.7, + "learning_rate": 8.843075500592332e-06, + "loss": 0.8786, + "step": 285380 + }, + { + "epoch": 24.7, + "learning_rate": 8.841630789678985e-06, + "loss": 0.9201, + "step": 285390 + }, + { + "epoch": 24.7, + "learning_rate": 8.840186078765638e-06, + "loss": 0.8592, + "step": 285400 + }, + { + "epoch": 24.7, + "learning_rate": 8.838741367852293e-06, + "loss": 0.9133, + "step": 285410 + }, + { + "epoch": 24.71, + "learning_rate": 8.837296656938947e-06, + "loss": 0.958, + "step": 285420 + }, + { + "epoch": 24.71, + "learning_rate": 8.8358519460256e-06, + "loss": 0.8993, + "step": 285430 + }, + { + "epoch": 24.71, + "learning_rate": 8.834407235112255e-06, + "loss": 0.8965, + "step": 285440 + }, + { + "epoch": 24.71, + "learning_rate": 8.832962524198909e-06, + "loss": 0.8949, + "step": 285450 + }, + { + "epoch": 24.71, + "learning_rate": 8.831517813285562e-06, + "loss": 0.8948, + "step": 285460 + }, + { + "epoch": 24.71, + "learning_rate": 8.830073102372216e-06, + "loss": 0.8995, + "step": 285470 + }, + { + "epoch": 24.71, + "learning_rate": 8.828628391458869e-06, + "loss": 0.9272, + "step": 285480 + }, + { + "epoch": 24.71, + "learning_rate": 8.827183680545522e-06, + "loss": 0.8959, + "step": 285490 + }, + { + "epoch": 24.71, + "learning_rate": 8.825738969632177e-06, + "loss": 0.9081, + "step": 285500 + }, + { + "epoch": 24.71, + "learning_rate": 8.82429425871883e-06, + "loss": 0.8785, + "step": 285510 + }, + { + "epoch": 24.71, + "learning_rate": 8.822849547805484e-06, + "loss": 0.9538, + "step": 285520 + }, + { + "epoch": 24.71, + "learning_rate": 8.82140483689214e-06, + "loss": 0.9229, + "step": 285530 + }, + { + "epoch": 24.72, + "learning_rate": 8.819960125978793e-06, + "loss": 0.9779, + "step": 285540 + }, + { + "epoch": 24.72, + "learning_rate": 8.818515415065446e-06, + "loss": 0.9202, + "step": 285550 + }, + { + "epoch": 24.72, + "learning_rate": 8.8170707041521e-06, + "loss": 0.8636, + "step": 285560 + }, + { + "epoch": 24.72, + "learning_rate": 8.815625993238753e-06, + "loss": 0.9214, + "step": 285570 + }, + { + "epoch": 24.72, + "learning_rate": 8.814181282325406e-06, + "loss": 0.9239, + "step": 285580 + }, + { + "epoch": 24.72, + "learning_rate": 8.812736571412061e-06, + "loss": 0.9668, + "step": 285590 + }, + { + "epoch": 24.72, + "learning_rate": 8.811291860498715e-06, + "loss": 0.923, + "step": 285600 + }, + { + "epoch": 24.72, + "learning_rate": 8.809847149585368e-06, + "loss": 0.9156, + "step": 285610 + }, + { + "epoch": 24.72, + "learning_rate": 8.808402438672023e-06, + "loss": 0.8908, + "step": 285620 + }, + { + "epoch": 24.72, + "learning_rate": 8.806957727758677e-06, + "loss": 0.9509, + "step": 285630 + }, + { + "epoch": 24.72, + "learning_rate": 8.80551301684533e-06, + "loss": 0.9317, + "step": 285640 + }, + { + "epoch": 24.73, + "learning_rate": 8.804068305931983e-06, + "loss": 0.8989, + "step": 285650 + }, + { + "epoch": 24.73, + "learning_rate": 8.802623595018637e-06, + "loss": 0.8858, + "step": 285660 + }, + { + "epoch": 24.73, + "learning_rate": 8.80117888410529e-06, + "loss": 0.8482, + "step": 285670 + }, + { + "epoch": 24.73, + "learning_rate": 8.799734173191945e-06, + "loss": 0.8877, + "step": 285680 + }, + { + "epoch": 24.73, + "learning_rate": 8.798289462278599e-06, + "loss": 0.8602, + "step": 285690 + }, + { + "epoch": 24.73, + "learning_rate": 8.796844751365252e-06, + "loss": 0.9141, + "step": 285700 + }, + { + "epoch": 24.73, + "learning_rate": 8.795400040451907e-06, + "loss": 0.9261, + "step": 285710 + }, + { + "epoch": 24.73, + "learning_rate": 8.79395532953856e-06, + "loss": 0.8926, + "step": 285720 + }, + { + "epoch": 24.73, + "learning_rate": 8.792510618625214e-06, + "loss": 0.8826, + "step": 285730 + }, + { + "epoch": 24.73, + "learning_rate": 8.791065907711867e-06, + "loss": 0.9519, + "step": 285740 + }, + { + "epoch": 24.73, + "learning_rate": 8.78962119679852e-06, + "loss": 0.9287, + "step": 285750 + }, + { + "epoch": 24.73, + "learning_rate": 8.788176485885174e-06, + "loss": 0.9359, + "step": 285760 + }, + { + "epoch": 24.74, + "learning_rate": 8.78673177497183e-06, + "loss": 0.8997, + "step": 285770 + }, + { + "epoch": 24.74, + "learning_rate": 8.785287064058483e-06, + "loss": 0.9243, + "step": 285780 + }, + { + "epoch": 24.74, + "learning_rate": 8.783842353145136e-06, + "loss": 0.9509, + "step": 285790 + }, + { + "epoch": 24.74, + "learning_rate": 8.782397642231791e-06, + "loss": 0.8972, + "step": 285800 + }, + { + "epoch": 24.74, + "learning_rate": 8.780952931318444e-06, + "loss": 0.9183, + "step": 285810 + }, + { + "epoch": 24.74, + "learning_rate": 8.779508220405098e-06, + "loss": 0.954, + "step": 285820 + }, + { + "epoch": 24.74, + "learning_rate": 8.778063509491751e-06, + "loss": 0.909, + "step": 285830 + }, + { + "epoch": 24.74, + "learning_rate": 8.776618798578405e-06, + "loss": 0.9308, + "step": 285840 + }, + { + "epoch": 24.74, + "learning_rate": 8.775174087665058e-06, + "loss": 0.8832, + "step": 285850 + }, + { + "epoch": 24.74, + "learning_rate": 8.773729376751711e-06, + "loss": 0.9189, + "step": 285860 + }, + { + "epoch": 24.74, + "learning_rate": 8.772284665838367e-06, + "loss": 0.8953, + "step": 285870 + }, + { + "epoch": 24.75, + "learning_rate": 8.77083995492502e-06, + "loss": 0.94, + "step": 285880 + }, + { + "epoch": 24.75, + "learning_rate": 8.769395244011673e-06, + "loss": 0.9134, + "step": 285890 + }, + { + "epoch": 24.75, + "learning_rate": 8.767950533098328e-06, + "loss": 0.932, + "step": 285900 + }, + { + "epoch": 24.75, + "learning_rate": 8.766505822184982e-06, + "loss": 0.985, + "step": 285910 + }, + { + "epoch": 24.75, + "learning_rate": 8.765061111271635e-06, + "loss": 0.9259, + "step": 285920 + }, + { + "epoch": 24.75, + "learning_rate": 8.763616400358289e-06, + "loss": 0.9038, + "step": 285930 + }, + { + "epoch": 24.75, + "learning_rate": 8.762171689444942e-06, + "loss": 0.8933, + "step": 285940 + }, + { + "epoch": 24.75, + "learning_rate": 8.760726978531595e-06, + "loss": 0.9376, + "step": 285950 + }, + { + "epoch": 24.75, + "learning_rate": 8.75928226761825e-06, + "loss": 0.9203, + "step": 285960 + }, + { + "epoch": 24.75, + "learning_rate": 8.757837556704904e-06, + "loss": 0.8993, + "step": 285970 + }, + { + "epoch": 24.75, + "learning_rate": 8.756392845791557e-06, + "loss": 0.8951, + "step": 285980 + }, + { + "epoch": 24.75, + "learning_rate": 8.754948134878212e-06, + "loss": 0.9142, + "step": 285990 + }, + { + "epoch": 24.76, + "learning_rate": 8.753503423964866e-06, + "loss": 0.8891, + "step": 286000 + }, + { + "epoch": 24.76, + "learning_rate": 8.752058713051517e-06, + "loss": 0.9087, + "step": 286010 + }, + { + "epoch": 24.76, + "learning_rate": 8.750614002138173e-06, + "loss": 0.8976, + "step": 286020 + }, + { + "epoch": 24.76, + "learning_rate": 8.749169291224826e-06, + "loss": 0.9092, + "step": 286030 + }, + { + "epoch": 24.76, + "learning_rate": 8.74772458031148e-06, + "loss": 0.8719, + "step": 286040 + }, + { + "epoch": 24.76, + "learning_rate": 8.746279869398134e-06, + "loss": 0.9393, + "step": 286050 + }, + { + "epoch": 24.76, + "learning_rate": 8.744835158484788e-06, + "loss": 0.8536, + "step": 286060 + }, + { + "epoch": 24.76, + "learning_rate": 8.743390447571441e-06, + "loss": 0.9844, + "step": 286070 + }, + { + "epoch": 24.76, + "learning_rate": 8.741945736658096e-06, + "loss": 0.9368, + "step": 286080 + }, + { + "epoch": 24.76, + "learning_rate": 8.74050102574475e-06, + "loss": 0.9353, + "step": 286090 + }, + { + "epoch": 24.76, + "learning_rate": 8.739056314831401e-06, + "loss": 0.897, + "step": 286100 + }, + { + "epoch": 24.76, + "learning_rate": 8.737611603918056e-06, + "loss": 0.8746, + "step": 286110 + }, + { + "epoch": 24.77, + "learning_rate": 8.73616689300471e-06, + "loss": 0.9259, + "step": 286120 + }, + { + "epoch": 24.77, + "learning_rate": 8.734722182091363e-06, + "loss": 0.9019, + "step": 286130 + }, + { + "epoch": 24.77, + "learning_rate": 8.733277471178018e-06, + "loss": 0.9024, + "step": 286140 + }, + { + "epoch": 24.77, + "learning_rate": 8.731832760264672e-06, + "loss": 0.8457, + "step": 286150 + }, + { + "epoch": 24.77, + "learning_rate": 8.730388049351325e-06, + "loss": 0.9202, + "step": 286160 + }, + { + "epoch": 24.77, + "learning_rate": 8.72894333843798e-06, + "loss": 0.8783, + "step": 286170 + }, + { + "epoch": 24.77, + "learning_rate": 8.727498627524634e-06, + "loss": 0.9072, + "step": 286180 + }, + { + "epoch": 24.77, + "learning_rate": 8.726053916611285e-06, + "loss": 0.9142, + "step": 286190 + }, + { + "epoch": 24.77, + "learning_rate": 8.72460920569794e-06, + "loss": 0.8891, + "step": 286200 + }, + { + "epoch": 24.77, + "learning_rate": 8.723164494784594e-06, + "loss": 0.9467, + "step": 286210 + }, + { + "epoch": 24.77, + "learning_rate": 8.721719783871247e-06, + "loss": 0.9178, + "step": 286220 + }, + { + "epoch": 24.78, + "learning_rate": 8.720275072957902e-06, + "loss": 0.8646, + "step": 286230 + }, + { + "epoch": 24.78, + "learning_rate": 8.718830362044556e-06, + "loss": 0.8945, + "step": 286240 + }, + { + "epoch": 24.78, + "learning_rate": 8.717385651131209e-06, + "loss": 0.9454, + "step": 286250 + }, + { + "epoch": 24.78, + "learning_rate": 8.715940940217864e-06, + "loss": 0.9417, + "step": 286260 + }, + { + "epoch": 24.78, + "learning_rate": 8.714496229304518e-06, + "loss": 0.9295, + "step": 286270 + }, + { + "epoch": 24.78, + "learning_rate": 8.71305151839117e-06, + "loss": 0.9457, + "step": 286280 + }, + { + "epoch": 24.78, + "learning_rate": 8.711606807477823e-06, + "loss": 0.9425, + "step": 286290 + }, + { + "epoch": 24.78, + "learning_rate": 8.710162096564478e-06, + "loss": 0.9412, + "step": 286300 + }, + { + "epoch": 24.78, + "learning_rate": 8.708717385651131e-06, + "loss": 0.8543, + "step": 286310 + }, + { + "epoch": 24.78, + "learning_rate": 8.707272674737785e-06, + "loss": 0.9078, + "step": 286320 + }, + { + "epoch": 24.78, + "learning_rate": 8.70582796382444e-06, + "loss": 0.9628, + "step": 286330 + }, + { + "epoch": 24.78, + "learning_rate": 8.704383252911093e-06, + "loss": 0.8721, + "step": 286340 + }, + { + "epoch": 24.79, + "learning_rate": 8.702938541997746e-06, + "loss": 0.9863, + "step": 286350 + }, + { + "epoch": 24.79, + "learning_rate": 8.701493831084402e-06, + "loss": 0.9614, + "step": 286360 + }, + { + "epoch": 24.79, + "learning_rate": 8.700049120171053e-06, + "loss": 0.8894, + "step": 286370 + }, + { + "epoch": 24.79, + "learning_rate": 8.698604409257707e-06, + "loss": 0.86, + "step": 286380 + }, + { + "epoch": 24.79, + "learning_rate": 8.697159698344362e-06, + "loss": 0.8995, + "step": 286390 + }, + { + "epoch": 24.79, + "learning_rate": 8.695714987431015e-06, + "loss": 0.9179, + "step": 286400 + }, + { + "epoch": 24.79, + "learning_rate": 8.694270276517669e-06, + "loss": 0.9193, + "step": 286410 + }, + { + "epoch": 24.79, + "learning_rate": 8.692825565604324e-06, + "loss": 0.9347, + "step": 286420 + }, + { + "epoch": 24.79, + "learning_rate": 8.691380854690977e-06, + "loss": 0.9336, + "step": 286430 + }, + { + "epoch": 24.79, + "learning_rate": 8.68993614377763e-06, + "loss": 0.9325, + "step": 286440 + }, + { + "epoch": 24.79, + "learning_rate": 8.688491432864285e-06, + "loss": 0.9069, + "step": 286450 + }, + { + "epoch": 24.8, + "learning_rate": 8.687046721950937e-06, + "loss": 0.9055, + "step": 286460 + }, + { + "epoch": 24.8, + "learning_rate": 8.68560201103759e-06, + "loss": 0.9083, + "step": 286470 + }, + { + "epoch": 24.8, + "learning_rate": 8.684157300124246e-06, + "loss": 0.8995, + "step": 286480 + }, + { + "epoch": 24.8, + "learning_rate": 8.682712589210899e-06, + "loss": 0.9619, + "step": 286490 + }, + { + "epoch": 24.8, + "learning_rate": 8.681267878297552e-06, + "loss": 0.9129, + "step": 286500 + }, + { + "epoch": 24.8, + "learning_rate": 8.679823167384208e-06, + "loss": 0.9082, + "step": 286510 + }, + { + "epoch": 24.8, + "learning_rate": 8.678378456470861e-06, + "loss": 0.9903, + "step": 286520 + }, + { + "epoch": 24.8, + "learning_rate": 8.676933745557514e-06, + "loss": 0.9062, + "step": 286530 + }, + { + "epoch": 24.8, + "learning_rate": 8.67548903464417e-06, + "loss": 0.9169, + "step": 286540 + }, + { + "epoch": 24.8, + "learning_rate": 8.674044323730821e-06, + "loss": 0.953, + "step": 286550 + }, + { + "epoch": 24.8, + "learning_rate": 8.672599612817475e-06, + "loss": 0.9183, + "step": 286560 + }, + { + "epoch": 24.8, + "learning_rate": 8.67115490190413e-06, + "loss": 0.9021, + "step": 286570 + }, + { + "epoch": 24.81, + "learning_rate": 8.669710190990783e-06, + "loss": 0.9183, + "step": 286580 + }, + { + "epoch": 24.81, + "learning_rate": 8.668265480077436e-06, + "loss": 0.8817, + "step": 286590 + }, + { + "epoch": 24.81, + "learning_rate": 8.666820769164091e-06, + "loss": 0.8794, + "step": 286600 + }, + { + "epoch": 24.81, + "learning_rate": 8.665376058250745e-06, + "loss": 0.8751, + "step": 286610 + }, + { + "epoch": 24.81, + "learning_rate": 8.663931347337398e-06, + "loss": 0.9289, + "step": 286620 + }, + { + "epoch": 24.81, + "learning_rate": 8.662486636424053e-06, + "loss": 0.9474, + "step": 286630 + }, + { + "epoch": 24.81, + "learning_rate": 8.661041925510705e-06, + "loss": 0.9179, + "step": 286640 + }, + { + "epoch": 24.81, + "learning_rate": 8.659597214597358e-06, + "loss": 0.9315, + "step": 286650 + }, + { + "epoch": 24.81, + "learning_rate": 8.658152503684014e-06, + "loss": 0.9462, + "step": 286660 + }, + { + "epoch": 24.81, + "learning_rate": 8.656707792770667e-06, + "loss": 0.9318, + "step": 286670 + }, + { + "epoch": 24.81, + "learning_rate": 8.65526308185732e-06, + "loss": 0.91, + "step": 286680 + }, + { + "epoch": 24.82, + "learning_rate": 8.653818370943975e-06, + "loss": 0.914, + "step": 286690 + }, + { + "epoch": 24.82, + "learning_rate": 8.652373660030629e-06, + "loss": 0.9147, + "step": 286700 + }, + { + "epoch": 24.82, + "learning_rate": 8.650928949117282e-06, + "loss": 0.9611, + "step": 286710 + }, + { + "epoch": 24.82, + "learning_rate": 8.649484238203937e-06, + "loss": 0.9022, + "step": 286720 + }, + { + "epoch": 24.82, + "learning_rate": 8.648039527290589e-06, + "loss": 0.9247, + "step": 286730 + }, + { + "epoch": 24.82, + "learning_rate": 8.646594816377242e-06, + "loss": 0.8708, + "step": 286740 + }, + { + "epoch": 24.82, + "learning_rate": 8.645150105463896e-06, + "loss": 0.9255, + "step": 286750 + }, + { + "epoch": 24.82, + "learning_rate": 8.643705394550551e-06, + "loss": 0.908, + "step": 286760 + }, + { + "epoch": 24.82, + "learning_rate": 8.642260683637204e-06, + "loss": 0.9385, + "step": 286770 + }, + { + "epoch": 24.82, + "learning_rate": 8.640815972723858e-06, + "loss": 0.9181, + "step": 286780 + }, + { + "epoch": 24.82, + "learning_rate": 8.639371261810513e-06, + "loss": 0.9325, + "step": 286790 + }, + { + "epoch": 24.82, + "learning_rate": 8.637926550897166e-06, + "loss": 0.867, + "step": 286800 + }, + { + "epoch": 24.83, + "learning_rate": 8.63648183998382e-06, + "loss": 0.9271, + "step": 286810 + }, + { + "epoch": 24.83, + "learning_rate": 8.635037129070473e-06, + "loss": 0.9185, + "step": 286820 + }, + { + "epoch": 24.83, + "learning_rate": 8.633592418157126e-06, + "loss": 0.9258, + "step": 286830 + }, + { + "epoch": 24.83, + "learning_rate": 8.63214770724378e-06, + "loss": 0.9276, + "step": 286840 + }, + { + "epoch": 24.83, + "learning_rate": 8.630702996330435e-06, + "loss": 0.9034, + "step": 286850 + }, + { + "epoch": 24.83, + "learning_rate": 8.629258285417088e-06, + "loss": 0.8738, + "step": 286860 + }, + { + "epoch": 24.83, + "learning_rate": 8.627813574503742e-06, + "loss": 0.9385, + "step": 286870 + }, + { + "epoch": 24.83, + "learning_rate": 8.626368863590397e-06, + "loss": 0.9201, + "step": 286880 + }, + { + "epoch": 24.83, + "learning_rate": 8.62492415267705e-06, + "loss": 0.9146, + "step": 286890 + }, + { + "epoch": 24.83, + "learning_rate": 8.623479441763704e-06, + "loss": 0.9505, + "step": 286900 + }, + { + "epoch": 24.83, + "learning_rate": 8.622034730850357e-06, + "loss": 0.9098, + "step": 286910 + }, + { + "epoch": 24.84, + "learning_rate": 8.62059001993701e-06, + "loss": 0.9555, + "step": 286920 + }, + { + "epoch": 24.84, + "learning_rate": 8.619145309023664e-06, + "loss": 0.9425, + "step": 286930 + }, + { + "epoch": 24.84, + "learning_rate": 8.617700598110319e-06, + "loss": 0.9686, + "step": 286940 + }, + { + "epoch": 24.84, + "learning_rate": 8.616255887196972e-06, + "loss": 0.9736, + "step": 286950 + }, + { + "epoch": 24.84, + "learning_rate": 8.614811176283626e-06, + "loss": 0.9004, + "step": 286960 + }, + { + "epoch": 24.84, + "learning_rate": 8.61336646537028e-06, + "loss": 0.9514, + "step": 286970 + }, + { + "epoch": 24.84, + "learning_rate": 8.611921754456934e-06, + "loss": 0.8848, + "step": 286980 + }, + { + "epoch": 24.84, + "learning_rate": 8.610477043543587e-06, + "loss": 0.9294, + "step": 286990 + }, + { + "epoch": 24.84, + "learning_rate": 8.60903233263024e-06, + "loss": 0.9243, + "step": 287000 + }, + { + "epoch": 24.84, + "learning_rate": 8.607587621716894e-06, + "loss": 0.8994, + "step": 287010 + }, + { + "epoch": 24.84, + "learning_rate": 8.606142910803548e-06, + "loss": 0.9401, + "step": 287020 + }, + { + "epoch": 24.84, + "learning_rate": 8.604698199890203e-06, + "loss": 0.9361, + "step": 287030 + }, + { + "epoch": 24.85, + "learning_rate": 8.603253488976856e-06, + "loss": 0.9827, + "step": 287040 + }, + { + "epoch": 24.85, + "learning_rate": 8.60180877806351e-06, + "loss": 0.8578, + "step": 287050 + }, + { + "epoch": 24.85, + "learning_rate": 8.600364067150165e-06, + "loss": 0.9361, + "step": 287060 + }, + { + "epoch": 24.85, + "learning_rate": 8.598919356236818e-06, + "loss": 0.9349, + "step": 287070 + }, + { + "epoch": 24.85, + "learning_rate": 8.597474645323471e-06, + "loss": 0.9019, + "step": 287080 + }, + { + "epoch": 24.85, + "learning_rate": 8.596029934410125e-06, + "loss": 0.9613, + "step": 287090 + }, + { + "epoch": 24.85, + "learning_rate": 8.594585223496778e-06, + "loss": 0.9233, + "step": 287100 + }, + { + "epoch": 24.85, + "learning_rate": 8.593140512583432e-06, + "loss": 0.9166, + "step": 287110 + }, + { + "epoch": 24.85, + "learning_rate": 8.591695801670087e-06, + "loss": 0.9194, + "step": 287120 + }, + { + "epoch": 24.85, + "learning_rate": 8.59025109075674e-06, + "loss": 0.8891, + "step": 287130 + }, + { + "epoch": 24.85, + "learning_rate": 8.588806379843393e-06, + "loss": 0.8715, + "step": 287140 + }, + { + "epoch": 24.85, + "learning_rate": 8.587361668930049e-06, + "loss": 0.9233, + "step": 287150 + }, + { + "epoch": 24.86, + "learning_rate": 8.585916958016702e-06, + "loss": 0.9967, + "step": 287160 + }, + { + "epoch": 24.86, + "learning_rate": 8.584472247103355e-06, + "loss": 0.9317, + "step": 287170 + }, + { + "epoch": 24.86, + "learning_rate": 8.583027536190009e-06, + "loss": 0.9029, + "step": 287180 + }, + { + "epoch": 24.86, + "learning_rate": 8.581582825276662e-06, + "loss": 0.9449, + "step": 287190 + }, + { + "epoch": 24.86, + "learning_rate": 8.580138114363316e-06, + "loss": 0.931, + "step": 287200 + }, + { + "epoch": 24.86, + "learning_rate": 8.578693403449969e-06, + "loss": 0.8584, + "step": 287210 + }, + { + "epoch": 24.86, + "learning_rate": 8.577248692536624e-06, + "loss": 0.9427, + "step": 287220 + }, + { + "epoch": 24.86, + "learning_rate": 8.575803981623277e-06, + "loss": 0.9301, + "step": 287230 + }, + { + "epoch": 24.86, + "learning_rate": 8.57435927070993e-06, + "loss": 0.8307, + "step": 287240 + }, + { + "epoch": 24.86, + "learning_rate": 8.572914559796586e-06, + "loss": 0.8795, + "step": 287250 + }, + { + "epoch": 24.86, + "learning_rate": 8.57146984888324e-06, + "loss": 0.9031, + "step": 287260 + }, + { + "epoch": 24.87, + "learning_rate": 8.570025137969893e-06, + "loss": 0.9467, + "step": 287270 + }, + { + "epoch": 24.87, + "learning_rate": 8.568580427056546e-06, + "loss": 0.8726, + "step": 287280 + }, + { + "epoch": 24.87, + "learning_rate": 8.5671357161432e-06, + "loss": 0.9127, + "step": 287290 + }, + { + "epoch": 24.87, + "learning_rate": 8.565691005229853e-06, + "loss": 0.8669, + "step": 287300 + }, + { + "epoch": 24.87, + "learning_rate": 8.564246294316508e-06, + "loss": 0.9317, + "step": 287310 + }, + { + "epoch": 24.87, + "learning_rate": 8.562801583403161e-06, + "loss": 1.023, + "step": 287320 + }, + { + "epoch": 24.87, + "learning_rate": 8.561356872489815e-06, + "loss": 0.9408, + "step": 287330 + }, + { + "epoch": 24.87, + "learning_rate": 8.55991216157647e-06, + "loss": 0.9579, + "step": 287340 + }, + { + "epoch": 24.87, + "learning_rate": 8.558467450663123e-06, + "loss": 0.9477, + "step": 287350 + }, + { + "epoch": 24.87, + "learning_rate": 8.557022739749777e-06, + "loss": 0.932, + "step": 287360 + }, + { + "epoch": 24.87, + "learning_rate": 8.55557802883643e-06, + "loss": 0.9172, + "step": 287370 + }, + { + "epoch": 24.87, + "learning_rate": 8.554133317923083e-06, + "loss": 0.9505, + "step": 287380 + }, + { + "epoch": 24.88, + "learning_rate": 8.552688607009737e-06, + "loss": 0.9373, + "step": 287390 + }, + { + "epoch": 24.88, + "learning_rate": 8.551243896096392e-06, + "loss": 0.871, + "step": 287400 + }, + { + "epoch": 24.88, + "learning_rate": 8.549799185183045e-06, + "loss": 0.8693, + "step": 287410 + }, + { + "epoch": 24.88, + "learning_rate": 8.548354474269699e-06, + "loss": 0.9238, + "step": 287420 + }, + { + "epoch": 24.88, + "learning_rate": 8.546909763356354e-06, + "loss": 0.8942, + "step": 287430 + }, + { + "epoch": 24.88, + "learning_rate": 8.545465052443007e-06, + "loss": 0.9556, + "step": 287440 + }, + { + "epoch": 24.88, + "learning_rate": 8.54402034152966e-06, + "loss": 0.8844, + "step": 287450 + }, + { + "epoch": 24.88, + "learning_rate": 8.542575630616314e-06, + "loss": 0.8932, + "step": 287460 + }, + { + "epoch": 24.88, + "learning_rate": 8.541130919702967e-06, + "loss": 0.9525, + "step": 287470 + }, + { + "epoch": 24.88, + "learning_rate": 8.53968620878962e-06, + "loss": 0.8964, + "step": 287480 + }, + { + "epoch": 24.88, + "learning_rate": 8.538241497876276e-06, + "loss": 0.9307, + "step": 287490 + }, + { + "epoch": 24.89, + "learning_rate": 8.53679678696293e-06, + "loss": 0.9766, + "step": 287500 + }, + { + "epoch": 24.89, + "learning_rate": 8.535352076049583e-06, + "loss": 0.8975, + "step": 287510 + }, + { + "epoch": 24.89, + "learning_rate": 8.533907365136238e-06, + "loss": 0.9303, + "step": 287520 + }, + { + "epoch": 24.89, + "learning_rate": 8.532462654222891e-06, + "loss": 0.9912, + "step": 287530 + }, + { + "epoch": 24.89, + "learning_rate": 8.531017943309544e-06, + "loss": 0.8837, + "step": 287540 + }, + { + "epoch": 24.89, + "learning_rate": 8.529573232396198e-06, + "loss": 0.8808, + "step": 287550 + }, + { + "epoch": 24.89, + "learning_rate": 8.528128521482851e-06, + "loss": 0.91, + "step": 287560 + }, + { + "epoch": 24.89, + "learning_rate": 8.526683810569505e-06, + "loss": 0.9297, + "step": 287570 + }, + { + "epoch": 24.89, + "learning_rate": 8.52523909965616e-06, + "loss": 0.9349, + "step": 287580 + }, + { + "epoch": 24.89, + "learning_rate": 8.523794388742813e-06, + "loss": 0.9586, + "step": 287590 + }, + { + "epoch": 24.89, + "learning_rate": 8.522349677829467e-06, + "loss": 0.9217, + "step": 287600 + }, + { + "epoch": 24.89, + "learning_rate": 8.520904966916122e-06, + "loss": 0.9319, + "step": 287610 + }, + { + "epoch": 24.9, + "learning_rate": 8.519460256002775e-06, + "loss": 0.9297, + "step": 287620 + }, + { + "epoch": 24.9, + "learning_rate": 8.518015545089428e-06, + "loss": 0.9289, + "step": 287630 + }, + { + "epoch": 24.9, + "learning_rate": 8.516570834176082e-06, + "loss": 0.8916, + "step": 287640 + }, + { + "epoch": 24.9, + "learning_rate": 8.515126123262735e-06, + "loss": 0.9016, + "step": 287650 + }, + { + "epoch": 24.9, + "learning_rate": 8.513681412349389e-06, + "loss": 0.8845, + "step": 287660 + }, + { + "epoch": 24.9, + "learning_rate": 8.512236701436042e-06, + "loss": 0.9296, + "step": 287670 + }, + { + "epoch": 24.9, + "learning_rate": 8.510791990522697e-06, + "loss": 0.8862, + "step": 287680 + }, + { + "epoch": 24.9, + "learning_rate": 8.50934727960935e-06, + "loss": 0.8394, + "step": 287690 + }, + { + "epoch": 24.9, + "learning_rate": 8.507902568696004e-06, + "loss": 0.9486, + "step": 287700 + }, + { + "epoch": 24.9, + "learning_rate": 8.506457857782659e-06, + "loss": 0.9177, + "step": 287710 + }, + { + "epoch": 24.9, + "learning_rate": 8.505013146869312e-06, + "loss": 0.8535, + "step": 287720 + }, + { + "epoch": 24.91, + "learning_rate": 8.503568435955966e-06, + "loss": 0.9017, + "step": 287730 + }, + { + "epoch": 24.91, + "learning_rate": 8.50212372504262e-06, + "loss": 0.8892, + "step": 287740 + }, + { + "epoch": 24.91, + "learning_rate": 8.500679014129273e-06, + "loss": 0.8846, + "step": 287750 + }, + { + "epoch": 24.91, + "learning_rate": 8.499234303215926e-06, + "loss": 0.974, + "step": 287760 + }, + { + "epoch": 24.91, + "learning_rate": 8.497789592302581e-06, + "loss": 0.9282, + "step": 287770 + }, + { + "epoch": 24.91, + "learning_rate": 8.496344881389234e-06, + "loss": 0.8696, + "step": 287780 + }, + { + "epoch": 24.91, + "learning_rate": 8.494900170475888e-06, + "loss": 0.9082, + "step": 287790 + }, + { + "epoch": 24.91, + "learning_rate": 8.493455459562543e-06, + "loss": 0.9402, + "step": 287800 + }, + { + "epoch": 24.91, + "learning_rate": 8.492010748649196e-06, + "loss": 0.9372, + "step": 287810 + }, + { + "epoch": 24.91, + "learning_rate": 8.49056603773585e-06, + "loss": 0.9818, + "step": 287820 + }, + { + "epoch": 24.91, + "learning_rate": 8.489121326822503e-06, + "loss": 0.8832, + "step": 287830 + }, + { + "epoch": 24.91, + "learning_rate": 8.487676615909157e-06, + "loss": 0.9211, + "step": 287840 + }, + { + "epoch": 24.92, + "learning_rate": 8.48623190499581e-06, + "loss": 0.8819, + "step": 287850 + }, + { + "epoch": 24.92, + "learning_rate": 8.484787194082465e-06, + "loss": 0.9247, + "step": 287860 + }, + { + "epoch": 24.92, + "learning_rate": 8.483342483169118e-06, + "loss": 0.9184, + "step": 287870 + }, + { + "epoch": 24.92, + "learning_rate": 8.481897772255772e-06, + "loss": 0.9078, + "step": 287880 + }, + { + "epoch": 24.92, + "learning_rate": 8.480453061342427e-06, + "loss": 0.9207, + "step": 287890 + }, + { + "epoch": 24.92, + "learning_rate": 8.47900835042908e-06, + "loss": 0.9427, + "step": 287900 + }, + { + "epoch": 24.92, + "learning_rate": 8.477563639515734e-06, + "loss": 0.9013, + "step": 287910 + }, + { + "epoch": 24.92, + "learning_rate": 8.476118928602387e-06, + "loss": 0.9023, + "step": 287920 + }, + { + "epoch": 24.92, + "learning_rate": 8.47467421768904e-06, + "loss": 0.9018, + "step": 287930 + }, + { + "epoch": 24.92, + "learning_rate": 8.473229506775694e-06, + "loss": 0.9018, + "step": 287940 + }, + { + "epoch": 24.92, + "learning_rate": 8.471784795862349e-06, + "loss": 0.8443, + "step": 287950 + }, + { + "epoch": 24.93, + "learning_rate": 8.470340084949002e-06, + "loss": 0.9073, + "step": 287960 + }, + { + "epoch": 24.93, + "learning_rate": 8.468895374035656e-06, + "loss": 0.9309, + "step": 287970 + }, + { + "epoch": 24.93, + "learning_rate": 8.46745066312231e-06, + "loss": 0.9125, + "step": 287980 + }, + { + "epoch": 24.93, + "learning_rate": 8.466005952208964e-06, + "loss": 0.8901, + "step": 287990 + }, + { + "epoch": 24.93, + "learning_rate": 8.464561241295618e-06, + "loss": 0.9938, + "step": 288000 + }, + { + "epoch": 24.93, + "learning_rate": 8.463116530382271e-06, + "loss": 0.9665, + "step": 288010 + }, + { + "epoch": 24.93, + "learning_rate": 8.461671819468924e-06, + "loss": 0.9179, + "step": 288020 + }, + { + "epoch": 24.93, + "learning_rate": 8.460227108555578e-06, + "loss": 0.9471, + "step": 288030 + }, + { + "epoch": 24.93, + "learning_rate": 8.458782397642233e-06, + "loss": 0.932, + "step": 288040 + }, + { + "epoch": 24.93, + "learning_rate": 8.457337686728886e-06, + "loss": 0.9447, + "step": 288050 + }, + { + "epoch": 24.93, + "learning_rate": 8.45589297581554e-06, + "loss": 0.9554, + "step": 288060 + }, + { + "epoch": 24.93, + "learning_rate": 8.454448264902195e-06, + "loss": 0.9176, + "step": 288070 + }, + { + "epoch": 24.94, + "learning_rate": 8.453003553988848e-06, + "loss": 0.8571, + "step": 288080 + }, + { + "epoch": 24.94, + "learning_rate": 8.451558843075502e-06, + "loss": 0.9006, + "step": 288090 + }, + { + "epoch": 24.94, + "learning_rate": 8.450114132162153e-06, + "loss": 0.8821, + "step": 288100 + }, + { + "epoch": 24.94, + "learning_rate": 8.448669421248808e-06, + "loss": 0.9515, + "step": 288110 + }, + { + "epoch": 24.94, + "learning_rate": 8.447224710335462e-06, + "loss": 0.9291, + "step": 288120 + }, + { + "epoch": 24.94, + "learning_rate": 8.445779999422115e-06, + "loss": 0.9328, + "step": 288130 + }, + { + "epoch": 24.94, + "learning_rate": 8.44433528850877e-06, + "loss": 0.8883, + "step": 288140 + }, + { + "epoch": 24.94, + "learning_rate": 8.442890577595424e-06, + "loss": 0.8988, + "step": 288150 + }, + { + "epoch": 24.94, + "learning_rate": 8.441445866682077e-06, + "loss": 0.8978, + "step": 288160 + }, + { + "epoch": 24.94, + "learning_rate": 8.440001155768732e-06, + "loss": 0.8337, + "step": 288170 + }, + { + "epoch": 24.94, + "learning_rate": 8.438556444855385e-06, + "loss": 0.9198, + "step": 288180 + }, + { + "epoch": 24.94, + "learning_rate": 8.437111733942037e-06, + "loss": 0.8726, + "step": 288190 + }, + { + "epoch": 24.95, + "learning_rate": 8.435667023028692e-06, + "loss": 0.9149, + "step": 288200 + }, + { + "epoch": 24.95, + "learning_rate": 8.434222312115346e-06, + "loss": 0.8787, + "step": 288210 + }, + { + "epoch": 24.95, + "learning_rate": 8.432777601201999e-06, + "loss": 0.8601, + "step": 288220 + }, + { + "epoch": 24.95, + "learning_rate": 8.431332890288654e-06, + "loss": 0.9083, + "step": 288230 + }, + { + "epoch": 24.95, + "learning_rate": 8.429888179375308e-06, + "loss": 0.9014, + "step": 288240 + }, + { + "epoch": 24.95, + "learning_rate": 8.428443468461961e-06, + "loss": 0.8908, + "step": 288250 + }, + { + "epoch": 24.95, + "learning_rate": 8.426998757548616e-06, + "loss": 0.9585, + "step": 288260 + }, + { + "epoch": 24.95, + "learning_rate": 8.42555404663527e-06, + "loss": 0.9376, + "step": 288270 + }, + { + "epoch": 24.95, + "learning_rate": 8.424109335721921e-06, + "loss": 0.8772, + "step": 288280 + }, + { + "epoch": 24.95, + "learning_rate": 8.422664624808576e-06, + "loss": 0.8771, + "step": 288290 + }, + { + "epoch": 24.95, + "learning_rate": 8.42121991389523e-06, + "loss": 0.9034, + "step": 288300 + }, + { + "epoch": 24.96, + "learning_rate": 8.419775202981883e-06, + "loss": 0.9177, + "step": 288310 + }, + { + "epoch": 24.96, + "learning_rate": 8.418330492068538e-06, + "loss": 0.9307, + "step": 288320 + }, + { + "epoch": 24.96, + "learning_rate": 8.416885781155191e-06, + "loss": 0.9601, + "step": 288330 + }, + { + "epoch": 24.96, + "learning_rate": 8.415441070241845e-06, + "loss": 0.8482, + "step": 288340 + }, + { + "epoch": 24.96, + "learning_rate": 8.4139963593285e-06, + "loss": 0.919, + "step": 288350 + }, + { + "epoch": 24.96, + "learning_rate": 8.412551648415153e-06, + "loss": 0.8927, + "step": 288360 + }, + { + "epoch": 24.96, + "learning_rate": 8.411106937501805e-06, + "loss": 0.908, + "step": 288370 + }, + { + "epoch": 24.96, + "learning_rate": 8.40966222658846e-06, + "loss": 0.9293, + "step": 288380 + }, + { + "epoch": 24.96, + "learning_rate": 8.408217515675114e-06, + "loss": 0.8934, + "step": 288390 + }, + { + "epoch": 24.96, + "learning_rate": 8.406772804761767e-06, + "loss": 0.9536, + "step": 288400 + }, + { + "epoch": 24.96, + "learning_rate": 8.405328093848422e-06, + "loss": 0.9381, + "step": 288410 + }, + { + "epoch": 24.96, + "learning_rate": 8.403883382935075e-06, + "loss": 0.9191, + "step": 288420 + }, + { + "epoch": 24.97, + "learning_rate": 8.402438672021729e-06, + "loss": 0.8956, + "step": 288430 + }, + { + "epoch": 24.97, + "learning_rate": 8.400993961108384e-06, + "loss": 0.9164, + "step": 288440 + }, + { + "epoch": 24.97, + "learning_rate": 8.399549250195037e-06, + "loss": 0.8982, + "step": 288450 + }, + { + "epoch": 24.97, + "learning_rate": 8.398104539281689e-06, + "loss": 0.902, + "step": 288460 + }, + { + "epoch": 24.97, + "learning_rate": 8.396659828368344e-06, + "loss": 0.932, + "step": 288470 + }, + { + "epoch": 24.97, + "learning_rate": 8.395215117454998e-06, + "loss": 0.8818, + "step": 288480 + }, + { + "epoch": 24.97, + "learning_rate": 8.393770406541651e-06, + "loss": 0.929, + "step": 288490 + }, + { + "epoch": 24.97, + "learning_rate": 8.392325695628306e-06, + "loss": 0.9092, + "step": 288500 + }, + { + "epoch": 24.97, + "learning_rate": 8.39088098471496e-06, + "loss": 0.8872, + "step": 288510 + }, + { + "epoch": 24.97, + "learning_rate": 8.389436273801613e-06, + "loss": 0.8986, + "step": 288520 + }, + { + "epoch": 24.97, + "learning_rate": 8.387991562888268e-06, + "loss": 0.9065, + "step": 288530 + }, + { + "epoch": 24.98, + "learning_rate": 8.386546851974921e-06, + "loss": 0.9281, + "step": 288540 + }, + { + "epoch": 24.98, + "learning_rate": 8.385102141061573e-06, + "loss": 0.937, + "step": 288550 + }, + { + "epoch": 24.98, + "learning_rate": 8.383657430148226e-06, + "loss": 0.9651, + "step": 288560 + }, + { + "epoch": 24.98, + "learning_rate": 8.382212719234881e-06, + "loss": 0.9768, + "step": 288570 + }, + { + "epoch": 24.98, + "learning_rate": 8.380768008321535e-06, + "loss": 0.8976, + "step": 288580 + }, + { + "epoch": 24.98, + "learning_rate": 8.379323297408188e-06, + "loss": 0.9593, + "step": 288590 + }, + { + "epoch": 24.98, + "learning_rate": 8.377878586494843e-06, + "loss": 0.9203, + "step": 288600 + }, + { + "epoch": 24.98, + "learning_rate": 8.376433875581497e-06, + "loss": 0.9221, + "step": 288610 + }, + { + "epoch": 24.98, + "learning_rate": 8.37498916466815e-06, + "loss": 0.9474, + "step": 288620 + }, + { + "epoch": 24.98, + "learning_rate": 8.373544453754805e-06, + "loss": 0.9341, + "step": 288630 + }, + { + "epoch": 24.98, + "learning_rate": 8.372099742841457e-06, + "loss": 0.953, + "step": 288640 + }, + { + "epoch": 24.98, + "learning_rate": 8.37065503192811e-06, + "loss": 0.882, + "step": 288650 + }, + { + "epoch": 24.99, + "learning_rate": 8.369210321014765e-06, + "loss": 0.8548, + "step": 288660 + }, + { + "epoch": 24.99, + "learning_rate": 8.367765610101419e-06, + "loss": 0.9372, + "step": 288670 + }, + { + "epoch": 24.99, + "learning_rate": 8.366320899188072e-06, + "loss": 0.8787, + "step": 288680 + }, + { + "epoch": 24.99, + "learning_rate": 8.364876188274727e-06, + "loss": 0.906, + "step": 288690 + }, + { + "epoch": 24.99, + "learning_rate": 8.36343147736138e-06, + "loss": 0.9196, + "step": 288700 + }, + { + "epoch": 24.99, + "learning_rate": 8.361986766448034e-06, + "loss": 0.9182, + "step": 288710 + }, + { + "epoch": 24.99, + "learning_rate": 8.360542055534689e-06, + "loss": 0.8793, + "step": 288720 + }, + { + "epoch": 24.99, + "learning_rate": 8.359097344621341e-06, + "loss": 0.9787, + "step": 288730 + }, + { + "epoch": 24.99, + "learning_rate": 8.357652633707994e-06, + "loss": 0.8827, + "step": 288740 + }, + { + "epoch": 24.99, + "learning_rate": 8.35620792279465e-06, + "loss": 0.9334, + "step": 288750 + }, + { + "epoch": 24.99, + "learning_rate": 8.354763211881303e-06, + "loss": 0.9437, + "step": 288760 + }, + { + "epoch": 25.0, + "learning_rate": 8.353318500967956e-06, + "loss": 0.9143, + "step": 288770 + }, + { + "epoch": 25.0, + "learning_rate": 8.351873790054611e-06, + "loss": 0.9524, + "step": 288780 + }, + { + "epoch": 25.0, + "learning_rate": 8.350429079141265e-06, + "loss": 0.8755, + "step": 288790 + }, + { + "epoch": 25.0, + "learning_rate": 8.348984368227918e-06, + "loss": 0.9208, + "step": 288800 + }, + { + "epoch": 25.0, + "learning_rate": 8.347539657314573e-06, + "loss": 0.8792, + "step": 288810 + }, + { + "epoch": 25.0, + "learning_rate": 8.346094946401225e-06, + "loss": 0.9103, + "step": 288820 + }, + { + "epoch": 25.0, + "eval_cer": 0.9930788174686038, + "eval_em": 0, + "eval_f1": 0.12078545847419246, + "eval_loss": 0.9365286827087402, + "eval_runtime": 899.8398, + "eval_samples_per_second": 5.707, + "eval_steps_per_second": 0.713, + "eval_wer": 0.9660228250332946, + "step": 288825 + }, + { + "epoch": 25.0, + "learning_rate": 8.344650235487878e-06, + "loss": 0.8436, + "step": 288830 + }, + { + "epoch": 25.0, + "learning_rate": 8.343205524574533e-06, + "loss": 0.8648, + "step": 288840 + }, + { + "epoch": 25.0, + "learning_rate": 8.341760813661187e-06, + "loss": 0.9667, + "step": 288850 + }, + { + "epoch": 25.0, + "learning_rate": 8.34031610274784e-06, + "loss": 0.9356, + "step": 288860 + }, + { + "epoch": 25.0, + "learning_rate": 8.338871391834495e-06, + "loss": 0.9373, + "step": 288870 + }, + { + "epoch": 25.0, + "learning_rate": 8.337426680921149e-06, + "loss": 0.9024, + "step": 288880 + }, + { + "epoch": 25.01, + "learning_rate": 8.335981970007802e-06, + "loss": 0.9375, + "step": 288890 + }, + { + "epoch": 25.01, + "learning_rate": 8.334537259094457e-06, + "loss": 0.8825, + "step": 288900 + }, + { + "epoch": 25.01, + "learning_rate": 8.333092548181109e-06, + "loss": 0.8539, + "step": 288910 + }, + { + "epoch": 25.01, + "learning_rate": 8.331647837267762e-06, + "loss": 0.9128, + "step": 288920 + }, + { + "epoch": 25.01, + "learning_rate": 8.330203126354417e-06, + "loss": 0.9284, + "step": 288930 + }, + { + "epoch": 25.01, + "learning_rate": 8.32875841544107e-06, + "loss": 0.9529, + "step": 288940 + }, + { + "epoch": 25.01, + "learning_rate": 8.327313704527724e-06, + "loss": 0.882, + "step": 288950 + }, + { + "epoch": 25.01, + "learning_rate": 8.325868993614379e-06, + "loss": 0.8919, + "step": 288960 + }, + { + "epoch": 25.01, + "learning_rate": 8.324424282701032e-06, + "loss": 0.8916, + "step": 288970 + }, + { + "epoch": 25.01, + "learning_rate": 8.322979571787686e-06, + "loss": 0.8681, + "step": 288980 + }, + { + "epoch": 25.01, + "learning_rate": 8.32153486087434e-06, + "loss": 0.9259, + "step": 288990 + }, + { + "epoch": 25.02, + "learning_rate": 8.320090149960993e-06, + "loss": 0.8883, + "step": 289000 + }, + { + "epoch": 25.02, + "learning_rate": 8.318645439047646e-06, + "loss": 0.8692, + "step": 289010 + }, + { + "epoch": 25.02, + "learning_rate": 8.3172007281343e-06, + "loss": 0.9109, + "step": 289020 + }, + { + "epoch": 25.02, + "learning_rate": 8.315756017220955e-06, + "loss": 0.9321, + "step": 289030 + }, + { + "epoch": 25.02, + "learning_rate": 8.314311306307608e-06, + "loss": 0.9205, + "step": 289040 + }, + { + "epoch": 25.02, + "learning_rate": 8.312866595394261e-06, + "loss": 0.8991, + "step": 289050 + }, + { + "epoch": 25.02, + "learning_rate": 8.311421884480916e-06, + "loss": 0.881, + "step": 289060 + }, + { + "epoch": 25.02, + "learning_rate": 8.30997717356757e-06, + "loss": 0.9252, + "step": 289070 + }, + { + "epoch": 25.02, + "learning_rate": 8.308532462654223e-06, + "loss": 0.9037, + "step": 289080 + }, + { + "epoch": 25.02, + "learning_rate": 8.307087751740877e-06, + "loss": 0.9063, + "step": 289090 + }, + { + "epoch": 25.02, + "learning_rate": 8.30564304082753e-06, + "loss": 0.8358, + "step": 289100 + }, + { + "epoch": 25.02, + "learning_rate": 8.304198329914183e-06, + "loss": 0.9144, + "step": 289110 + }, + { + "epoch": 25.03, + "learning_rate": 8.302753619000838e-06, + "loss": 0.9109, + "step": 289120 + }, + { + "epoch": 25.03, + "learning_rate": 8.301308908087492e-06, + "loss": 0.9161, + "step": 289130 + }, + { + "epoch": 25.03, + "learning_rate": 8.299864197174145e-06, + "loss": 0.8741, + "step": 289140 + }, + { + "epoch": 25.03, + "learning_rate": 8.2984194862608e-06, + "loss": 0.9191, + "step": 289150 + }, + { + "epoch": 25.03, + "learning_rate": 8.296974775347454e-06, + "loss": 0.9348, + "step": 289160 + }, + { + "epoch": 25.03, + "learning_rate": 8.295530064434107e-06, + "loss": 0.9209, + "step": 289170 + }, + { + "epoch": 25.03, + "learning_rate": 8.29408535352076e-06, + "loss": 0.9446, + "step": 289180 + }, + { + "epoch": 25.03, + "learning_rate": 8.292640642607414e-06, + "loss": 0.9627, + "step": 289190 + }, + { + "epoch": 25.03, + "learning_rate": 8.291195931694067e-06, + "loss": 0.9597, + "step": 289200 + }, + { + "epoch": 25.03, + "learning_rate": 8.289751220780722e-06, + "loss": 0.9441, + "step": 289210 + }, + { + "epoch": 25.03, + "learning_rate": 8.288306509867376e-06, + "loss": 0.898, + "step": 289220 + }, + { + "epoch": 25.04, + "learning_rate": 8.28686179895403e-06, + "loss": 0.907, + "step": 289230 + }, + { + "epoch": 25.04, + "learning_rate": 8.285417088040684e-06, + "loss": 0.9337, + "step": 289240 + }, + { + "epoch": 25.04, + "learning_rate": 8.283972377127338e-06, + "loss": 0.8898, + "step": 289250 + }, + { + "epoch": 25.04, + "learning_rate": 8.282527666213991e-06, + "loss": 0.8768, + "step": 289260 + }, + { + "epoch": 25.04, + "learning_rate": 8.281082955300645e-06, + "loss": 0.928, + "step": 289270 + }, + { + "epoch": 25.04, + "learning_rate": 8.279638244387298e-06, + "loss": 0.9004, + "step": 289280 + }, + { + "epoch": 25.04, + "learning_rate": 8.278193533473951e-06, + "loss": 0.9338, + "step": 289290 + }, + { + "epoch": 25.04, + "learning_rate": 8.276748822560606e-06, + "loss": 0.9481, + "step": 289300 + }, + { + "epoch": 25.04, + "learning_rate": 8.27530411164726e-06, + "loss": 0.9331, + "step": 289310 + }, + { + "epoch": 25.04, + "learning_rate": 8.273859400733913e-06, + "loss": 0.9183, + "step": 289320 + }, + { + "epoch": 25.04, + "learning_rate": 8.272414689820568e-06, + "loss": 0.9304, + "step": 289330 + }, + { + "epoch": 25.04, + "learning_rate": 8.270969978907222e-06, + "loss": 0.9156, + "step": 289340 + }, + { + "epoch": 25.05, + "learning_rate": 8.269525267993875e-06, + "loss": 0.9425, + "step": 289350 + }, + { + "epoch": 25.05, + "learning_rate": 8.268080557080528e-06, + "loss": 0.8985, + "step": 289360 + }, + { + "epoch": 25.05, + "learning_rate": 8.266635846167182e-06, + "loss": 0.9775, + "step": 289370 + }, + { + "epoch": 25.05, + "learning_rate": 8.265191135253835e-06, + "loss": 0.976, + "step": 289380 + }, + { + "epoch": 25.05, + "learning_rate": 8.26374642434049e-06, + "loss": 0.9364, + "step": 289390 + }, + { + "epoch": 25.05, + "learning_rate": 8.262301713427144e-06, + "loss": 0.9494, + "step": 289400 + }, + { + "epoch": 25.05, + "learning_rate": 8.260857002513797e-06, + "loss": 0.9304, + "step": 289410 + }, + { + "epoch": 25.05, + "learning_rate": 8.259412291600452e-06, + "loss": 0.9068, + "step": 289420 + }, + { + "epoch": 25.05, + "learning_rate": 8.257967580687106e-06, + "loss": 0.9707, + "step": 289430 + }, + { + "epoch": 25.05, + "learning_rate": 8.256522869773759e-06, + "loss": 0.8589, + "step": 289440 + }, + { + "epoch": 25.05, + "learning_rate": 8.255078158860412e-06, + "loss": 0.8806, + "step": 289450 + }, + { + "epoch": 25.05, + "learning_rate": 8.253633447947066e-06, + "loss": 0.9205, + "step": 289460 + }, + { + "epoch": 25.06, + "learning_rate": 8.25218873703372e-06, + "loss": 0.9416, + "step": 289470 + }, + { + "epoch": 25.06, + "learning_rate": 8.250744026120373e-06, + "loss": 0.9554, + "step": 289480 + }, + { + "epoch": 25.06, + "learning_rate": 8.249299315207028e-06, + "loss": 0.9004, + "step": 289490 + }, + { + "epoch": 25.06, + "learning_rate": 8.247854604293681e-06, + "loss": 0.8898, + "step": 289500 + }, + { + "epoch": 25.06, + "learning_rate": 8.246409893380334e-06, + "loss": 0.9001, + "step": 289510 + }, + { + "epoch": 25.06, + "learning_rate": 8.24496518246699e-06, + "loss": 0.8999, + "step": 289520 + }, + { + "epoch": 25.06, + "learning_rate": 8.243520471553643e-06, + "loss": 0.9839, + "step": 289530 + }, + { + "epoch": 25.06, + "learning_rate": 8.242075760640296e-06, + "loss": 0.9385, + "step": 289540 + }, + { + "epoch": 25.06, + "learning_rate": 8.24063104972695e-06, + "loss": 0.9038, + "step": 289550 + }, + { + "epoch": 25.06, + "learning_rate": 8.239186338813603e-06, + "loss": 0.9142, + "step": 289560 + }, + { + "epoch": 25.06, + "learning_rate": 8.237741627900257e-06, + "loss": 0.8978, + "step": 289570 + }, + { + "epoch": 25.07, + "learning_rate": 8.236296916986912e-06, + "loss": 0.9086, + "step": 289580 + }, + { + "epoch": 25.07, + "learning_rate": 8.234852206073565e-06, + "loss": 0.8551, + "step": 289590 + }, + { + "epoch": 25.07, + "learning_rate": 8.233407495160218e-06, + "loss": 0.8616, + "step": 289600 + }, + { + "epoch": 25.07, + "learning_rate": 8.231962784246873e-06, + "loss": 0.875, + "step": 289610 + }, + { + "epoch": 25.07, + "learning_rate": 8.230518073333527e-06, + "loss": 0.9183, + "step": 289620 + }, + { + "epoch": 25.07, + "learning_rate": 8.22907336242018e-06, + "loss": 0.8902, + "step": 289630 + }, + { + "epoch": 25.07, + "learning_rate": 8.227628651506834e-06, + "loss": 0.9188, + "step": 289640 + }, + { + "epoch": 25.07, + "learning_rate": 8.226183940593487e-06, + "loss": 0.9269, + "step": 289650 + }, + { + "epoch": 25.07, + "learning_rate": 8.22473922968014e-06, + "loss": 0.9831, + "step": 289660 + }, + { + "epoch": 25.07, + "learning_rate": 8.223294518766796e-06, + "loss": 0.8887, + "step": 289670 + }, + { + "epoch": 25.07, + "learning_rate": 8.221849807853449e-06, + "loss": 0.9141, + "step": 289680 + }, + { + "epoch": 25.07, + "learning_rate": 8.220405096940102e-06, + "loss": 0.8987, + "step": 289690 + }, + { + "epoch": 25.08, + "learning_rate": 8.218960386026757e-06, + "loss": 0.9198, + "step": 289700 + }, + { + "epoch": 25.08, + "learning_rate": 8.21751567511341e-06, + "loss": 0.9124, + "step": 289710 + }, + { + "epoch": 25.08, + "learning_rate": 8.216070964200064e-06, + "loss": 0.9301, + "step": 289720 + }, + { + "epoch": 25.08, + "learning_rate": 8.214626253286718e-06, + "loss": 0.9199, + "step": 289730 + }, + { + "epoch": 25.08, + "learning_rate": 8.213181542373371e-06, + "loss": 0.8734, + "step": 289740 + }, + { + "epoch": 25.08, + "learning_rate": 8.211736831460024e-06, + "loss": 0.9706, + "step": 289750 + }, + { + "epoch": 25.08, + "learning_rate": 8.21029212054668e-06, + "loss": 0.8927, + "step": 289760 + }, + { + "epoch": 25.08, + "learning_rate": 8.208847409633333e-06, + "loss": 0.9472, + "step": 289770 + }, + { + "epoch": 25.08, + "learning_rate": 8.207402698719986e-06, + "loss": 0.906, + "step": 289780 + }, + { + "epoch": 25.08, + "learning_rate": 8.205957987806641e-06, + "loss": 0.9523, + "step": 289790 + }, + { + "epoch": 25.08, + "learning_rate": 8.204513276893295e-06, + "loss": 0.8649, + "step": 289800 + }, + { + "epoch": 25.09, + "learning_rate": 8.203068565979948e-06, + "loss": 0.9297, + "step": 289810 + }, + { + "epoch": 25.09, + "learning_rate": 8.201623855066602e-06, + "loss": 0.8467, + "step": 289820 + }, + { + "epoch": 25.09, + "learning_rate": 8.200179144153255e-06, + "loss": 0.8797, + "step": 289830 + }, + { + "epoch": 25.09, + "learning_rate": 8.198734433239908e-06, + "loss": 0.8999, + "step": 289840 + }, + { + "epoch": 25.09, + "learning_rate": 8.197289722326563e-06, + "loss": 0.9408, + "step": 289850 + }, + { + "epoch": 25.09, + "learning_rate": 8.195845011413217e-06, + "loss": 0.9145, + "step": 289860 + }, + { + "epoch": 25.09, + "learning_rate": 8.19440030049987e-06, + "loss": 0.8957, + "step": 289870 + }, + { + "epoch": 25.09, + "learning_rate": 8.192955589586525e-06, + "loss": 0.9163, + "step": 289880 + }, + { + "epoch": 25.09, + "learning_rate": 8.191510878673179e-06, + "loss": 0.889, + "step": 289890 + }, + { + "epoch": 25.09, + "learning_rate": 8.190066167759832e-06, + "loss": 0.9017, + "step": 289900 + }, + { + "epoch": 25.09, + "learning_rate": 8.188621456846485e-06, + "loss": 0.9, + "step": 289910 + }, + { + "epoch": 25.09, + "learning_rate": 8.187176745933139e-06, + "loss": 0.9142, + "step": 289920 + }, + { + "epoch": 25.1, + "learning_rate": 8.185732035019792e-06, + "loss": 0.8854, + "step": 289930 + }, + { + "epoch": 25.1, + "learning_rate": 8.184287324106446e-06, + "loss": 0.8856, + "step": 289940 + }, + { + "epoch": 25.1, + "learning_rate": 8.1828426131931e-06, + "loss": 0.8912, + "step": 289950 + }, + { + "epoch": 25.1, + "learning_rate": 8.181397902279754e-06, + "loss": 0.9821, + "step": 289960 + }, + { + "epoch": 25.1, + "learning_rate": 8.179953191366408e-06, + "loss": 0.9142, + "step": 289970 + }, + { + "epoch": 25.1, + "learning_rate": 8.178508480453063e-06, + "loss": 0.8929, + "step": 289980 + }, + { + "epoch": 25.1, + "learning_rate": 8.177063769539716e-06, + "loss": 0.8482, + "step": 289990 + }, + { + "epoch": 25.1, + "learning_rate": 8.17561905862637e-06, + "loss": 0.8469, + "step": 290000 + }, + { + "epoch": 25.1, + "learning_rate": 8.174174347713023e-06, + "loss": 0.9485, + "step": 290010 + }, + { + "epoch": 25.1, + "learning_rate": 8.172729636799676e-06, + "loss": 0.9443, + "step": 290020 + }, + { + "epoch": 25.1, + "learning_rate": 8.17128492588633e-06, + "loss": 0.8872, + "step": 290030 + }, + { + "epoch": 25.11, + "learning_rate": 8.169840214972985e-06, + "loss": 0.8839, + "step": 290040 + }, + { + "epoch": 25.11, + "learning_rate": 8.168395504059638e-06, + "loss": 0.9142, + "step": 290050 + }, + { + "epoch": 25.11, + "learning_rate": 8.166950793146292e-06, + "loss": 0.9774, + "step": 290060 + }, + { + "epoch": 25.11, + "learning_rate": 8.165506082232947e-06, + "loss": 0.9277, + "step": 290070 + }, + { + "epoch": 25.11, + "learning_rate": 8.1640613713196e-06, + "loss": 0.9268, + "step": 290080 + }, + { + "epoch": 25.11, + "learning_rate": 8.162616660406253e-06, + "loss": 0.9589, + "step": 290090 + }, + { + "epoch": 25.11, + "learning_rate": 8.161171949492907e-06, + "loss": 0.8969, + "step": 290100 + }, + { + "epoch": 25.11, + "learning_rate": 8.15972723857956e-06, + "loss": 0.872, + "step": 290110 + }, + { + "epoch": 25.11, + "learning_rate": 8.158282527666214e-06, + "loss": 0.8991, + "step": 290120 + }, + { + "epoch": 25.11, + "learning_rate": 8.156837816752869e-06, + "loss": 0.9356, + "step": 290130 + }, + { + "epoch": 25.11, + "learning_rate": 8.155393105839522e-06, + "loss": 0.9068, + "step": 290140 + }, + { + "epoch": 25.11, + "learning_rate": 8.153948394926175e-06, + "loss": 0.9155, + "step": 290150 + }, + { + "epoch": 25.12, + "learning_rate": 8.15250368401283e-06, + "loss": 0.9428, + "step": 290160 + }, + { + "epoch": 25.12, + "learning_rate": 8.151058973099484e-06, + "loss": 0.9119, + "step": 290170 + }, + { + "epoch": 25.12, + "learning_rate": 8.149614262186137e-06, + "loss": 0.8614, + "step": 290180 + }, + { + "epoch": 25.12, + "learning_rate": 8.14816955127279e-06, + "loss": 0.9566, + "step": 290190 + }, + { + "epoch": 25.12, + "learning_rate": 8.146724840359444e-06, + "loss": 0.9315, + "step": 290200 + }, + { + "epoch": 25.12, + "learning_rate": 8.145280129446098e-06, + "loss": 0.906, + "step": 290210 + }, + { + "epoch": 25.12, + "learning_rate": 8.143835418532753e-06, + "loss": 0.8792, + "step": 290220 + }, + { + "epoch": 25.12, + "learning_rate": 8.142390707619406e-06, + "loss": 0.8373, + "step": 290230 + }, + { + "epoch": 25.12, + "learning_rate": 8.14094599670606e-06, + "loss": 0.9043, + "step": 290240 + }, + { + "epoch": 25.12, + "learning_rate": 8.139501285792714e-06, + "loss": 0.8712, + "step": 290250 + }, + { + "epoch": 25.12, + "learning_rate": 8.138056574879368e-06, + "loss": 0.9432, + "step": 290260 + }, + { + "epoch": 25.13, + "learning_rate": 8.136611863966021e-06, + "loss": 0.9009, + "step": 290270 + }, + { + "epoch": 25.13, + "learning_rate": 8.135167153052675e-06, + "loss": 0.9317, + "step": 290280 + }, + { + "epoch": 25.13, + "learning_rate": 8.133722442139328e-06, + "loss": 0.8499, + "step": 290290 + }, + { + "epoch": 25.13, + "learning_rate": 8.132277731225981e-06, + "loss": 0.9179, + "step": 290300 + }, + { + "epoch": 25.13, + "learning_rate": 8.130833020312637e-06, + "loss": 0.9097, + "step": 290310 + }, + { + "epoch": 25.13, + "learning_rate": 8.12938830939929e-06, + "loss": 0.9285, + "step": 290320 + }, + { + "epoch": 25.13, + "learning_rate": 8.127943598485943e-06, + "loss": 0.9035, + "step": 290330 + }, + { + "epoch": 25.13, + "learning_rate": 8.126498887572597e-06, + "loss": 0.8975, + "step": 290340 + }, + { + "epoch": 25.13, + "learning_rate": 8.125054176659252e-06, + "loss": 0.9536, + "step": 290350 + }, + { + "epoch": 25.13, + "learning_rate": 8.123609465745905e-06, + "loss": 0.886, + "step": 290360 + }, + { + "epoch": 25.13, + "learning_rate": 8.122164754832557e-06, + "loss": 0.8557, + "step": 290370 + }, + { + "epoch": 25.13, + "learning_rate": 8.120720043919212e-06, + "loss": 0.8356, + "step": 290380 + }, + { + "epoch": 25.14, + "learning_rate": 8.119275333005865e-06, + "loss": 0.8936, + "step": 290390 + }, + { + "epoch": 25.14, + "learning_rate": 8.117830622092519e-06, + "loss": 0.8785, + "step": 290400 + }, + { + "epoch": 25.14, + "learning_rate": 8.116385911179174e-06, + "loss": 0.9151, + "step": 290410 + }, + { + "epoch": 25.14, + "learning_rate": 8.114941200265827e-06, + "loss": 0.9184, + "step": 290420 + }, + { + "epoch": 25.14, + "learning_rate": 8.11349648935248e-06, + "loss": 0.8432, + "step": 290430 + }, + { + "epoch": 25.14, + "learning_rate": 8.112051778439136e-06, + "loss": 0.95, + "step": 290440 + }, + { + "epoch": 25.14, + "learning_rate": 8.11060706752579e-06, + "loss": 0.9185, + "step": 290450 + }, + { + "epoch": 25.14, + "learning_rate": 8.109162356612441e-06, + "loss": 0.8654, + "step": 290460 + }, + { + "epoch": 25.14, + "learning_rate": 8.107717645699096e-06, + "loss": 0.9369, + "step": 290470 + }, + { + "epoch": 25.14, + "learning_rate": 8.10627293478575e-06, + "loss": 0.8964, + "step": 290480 + }, + { + "epoch": 25.14, + "learning_rate": 8.104828223872403e-06, + "loss": 0.9058, + "step": 290490 + }, + { + "epoch": 25.14, + "learning_rate": 8.103383512959058e-06, + "loss": 0.9132, + "step": 290500 + }, + { + "epoch": 25.15, + "learning_rate": 8.101938802045711e-06, + "loss": 0.9768, + "step": 290510 + }, + { + "epoch": 25.15, + "learning_rate": 8.100494091132365e-06, + "loss": 0.8732, + "step": 290520 + }, + { + "epoch": 25.15, + "learning_rate": 8.09904938021902e-06, + "loss": 0.9425, + "step": 290530 + }, + { + "epoch": 25.15, + "learning_rate": 8.097604669305673e-06, + "loss": 0.934, + "step": 290540 + }, + { + "epoch": 25.15, + "learning_rate": 8.096159958392325e-06, + "loss": 0.9679, + "step": 290550 + }, + { + "epoch": 25.15, + "learning_rate": 8.09471524747898e-06, + "loss": 0.913, + "step": 290560 + }, + { + "epoch": 25.15, + "learning_rate": 8.093270536565633e-06, + "loss": 0.9019, + "step": 290570 + }, + { + "epoch": 25.15, + "learning_rate": 8.091825825652287e-06, + "loss": 0.8633, + "step": 290580 + }, + { + "epoch": 25.15, + "learning_rate": 8.090381114738942e-06, + "loss": 0.9077, + "step": 290590 + }, + { + "epoch": 25.15, + "learning_rate": 8.088936403825595e-06, + "loss": 0.8858, + "step": 290600 + }, + { + "epoch": 25.15, + "learning_rate": 8.087491692912249e-06, + "loss": 0.9452, + "step": 290610 + }, + { + "epoch": 25.16, + "learning_rate": 8.086046981998904e-06, + "loss": 0.8998, + "step": 290620 + }, + { + "epoch": 25.16, + "learning_rate": 8.084602271085557e-06, + "loss": 0.9721, + "step": 290630 + }, + { + "epoch": 25.16, + "learning_rate": 8.083157560172209e-06, + "loss": 0.9343, + "step": 290640 + }, + { + "epoch": 25.16, + "learning_rate": 8.081712849258864e-06, + "loss": 0.9101, + "step": 290650 + }, + { + "epoch": 25.16, + "learning_rate": 8.080268138345517e-06, + "loss": 0.9113, + "step": 290660 + }, + { + "epoch": 25.16, + "learning_rate": 8.07882342743217e-06, + "loss": 0.9278, + "step": 290670 + }, + { + "epoch": 25.16, + "learning_rate": 8.077378716518826e-06, + "loss": 0.884, + "step": 290680 + }, + { + "epoch": 25.16, + "learning_rate": 8.075934005605479e-06, + "loss": 0.9567, + "step": 290690 + }, + { + "epoch": 25.16, + "learning_rate": 8.074489294692133e-06, + "loss": 0.8224, + "step": 290700 + }, + { + "epoch": 25.16, + "learning_rate": 8.073044583778788e-06, + "loss": 0.8775, + "step": 290710 + }, + { + "epoch": 25.16, + "learning_rate": 8.071599872865441e-06, + "loss": 0.8943, + "step": 290720 + }, + { + "epoch": 25.16, + "learning_rate": 8.070155161952093e-06, + "loss": 0.9265, + "step": 290730 + }, + { + "epoch": 25.17, + "learning_rate": 8.068710451038748e-06, + "loss": 0.9204, + "step": 290740 + }, + { + "epoch": 25.17, + "learning_rate": 8.067265740125401e-06, + "loss": 0.9152, + "step": 290750 + }, + { + "epoch": 25.17, + "learning_rate": 8.065821029212055e-06, + "loss": 0.8078, + "step": 290760 + }, + { + "epoch": 25.17, + "learning_rate": 8.06437631829871e-06, + "loss": 0.8933, + "step": 290770 + }, + { + "epoch": 25.17, + "learning_rate": 8.062931607385363e-06, + "loss": 0.9052, + "step": 290780 + }, + { + "epoch": 25.17, + "learning_rate": 8.061486896472016e-06, + "loss": 0.9215, + "step": 290790 + }, + { + "epoch": 25.17, + "learning_rate": 8.06004218555867e-06, + "loss": 0.9546, + "step": 290800 + }, + { + "epoch": 25.17, + "learning_rate": 8.058597474645325e-06, + "loss": 0.917, + "step": 290810 + }, + { + "epoch": 25.17, + "learning_rate": 8.057152763731977e-06, + "loss": 0.8688, + "step": 290820 + }, + { + "epoch": 25.17, + "learning_rate": 8.05570805281863e-06, + "loss": 0.8504, + "step": 290830 + }, + { + "epoch": 25.17, + "learning_rate": 8.054263341905285e-06, + "loss": 0.9558, + "step": 290840 + }, + { + "epoch": 25.18, + "learning_rate": 8.052818630991939e-06, + "loss": 0.8604, + "step": 290850 + }, + { + "epoch": 25.18, + "learning_rate": 8.051373920078592e-06, + "loss": 0.9368, + "step": 290860 + }, + { + "epoch": 25.18, + "learning_rate": 8.049929209165247e-06, + "loss": 0.9276, + "step": 290870 + }, + { + "epoch": 25.18, + "learning_rate": 8.0484844982519e-06, + "loss": 0.8807, + "step": 290880 + }, + { + "epoch": 25.18, + "learning_rate": 8.047039787338554e-06, + "loss": 0.8623, + "step": 290890 + }, + { + "epoch": 25.18, + "learning_rate": 8.045595076425209e-06, + "loss": 0.9386, + "step": 290900 + }, + { + "epoch": 25.18, + "learning_rate": 8.04415036551186e-06, + "loss": 0.9679, + "step": 290910 + }, + { + "epoch": 25.18, + "learning_rate": 8.042705654598514e-06, + "loss": 0.8923, + "step": 290920 + }, + { + "epoch": 25.18, + "learning_rate": 8.041260943685169e-06, + "loss": 0.9125, + "step": 290930 + }, + { + "epoch": 25.18, + "learning_rate": 8.039816232771822e-06, + "loss": 0.9317, + "step": 290940 + }, + { + "epoch": 25.18, + "learning_rate": 8.038371521858476e-06, + "loss": 0.9106, + "step": 290950 + }, + { + "epoch": 25.18, + "learning_rate": 8.036926810945131e-06, + "loss": 0.9418, + "step": 290960 + }, + { + "epoch": 25.19, + "learning_rate": 8.035482100031784e-06, + "loss": 0.9694, + "step": 290970 + }, + { + "epoch": 25.19, + "learning_rate": 8.034037389118438e-06, + "loss": 0.88, + "step": 290980 + }, + { + "epoch": 25.19, + "learning_rate": 8.032592678205093e-06, + "loss": 0.94, + "step": 290990 + }, + { + "epoch": 25.19, + "learning_rate": 8.031147967291745e-06, + "loss": 0.9375, + "step": 291000 + }, + { + "epoch": 25.19, + "learning_rate": 8.029703256378398e-06, + "loss": 0.8816, + "step": 291010 + }, + { + "epoch": 25.19, + "learning_rate": 8.028258545465053e-06, + "loss": 0.9048, + "step": 291020 + }, + { + "epoch": 25.19, + "learning_rate": 8.026813834551706e-06, + "loss": 0.9148, + "step": 291030 + }, + { + "epoch": 25.19, + "learning_rate": 8.02536912363836e-06, + "loss": 0.8441, + "step": 291040 + }, + { + "epoch": 25.19, + "learning_rate": 8.023924412725015e-06, + "loss": 0.9735, + "step": 291050 + }, + { + "epoch": 25.19, + "learning_rate": 8.022479701811668e-06, + "loss": 0.9167, + "step": 291060 + }, + { + "epoch": 25.19, + "learning_rate": 8.021034990898322e-06, + "loss": 0.8852, + "step": 291070 + }, + { + "epoch": 25.2, + "learning_rate": 8.019590279984977e-06, + "loss": 0.8652, + "step": 291080 + }, + { + "epoch": 25.2, + "learning_rate": 8.018145569071628e-06, + "loss": 0.8996, + "step": 291090 + }, + { + "epoch": 25.2, + "learning_rate": 8.016700858158282e-06, + "loss": 0.8877, + "step": 291100 + }, + { + "epoch": 25.2, + "learning_rate": 8.015256147244937e-06, + "loss": 0.8953, + "step": 291110 + }, + { + "epoch": 25.2, + "learning_rate": 8.01381143633159e-06, + "loss": 0.9213, + "step": 291120 + }, + { + "epoch": 25.2, + "learning_rate": 8.012366725418244e-06, + "loss": 0.8957, + "step": 291130 + }, + { + "epoch": 25.2, + "learning_rate": 8.010922014504899e-06, + "loss": 0.8422, + "step": 291140 + }, + { + "epoch": 25.2, + "learning_rate": 8.009477303591552e-06, + "loss": 0.8684, + "step": 291150 + }, + { + "epoch": 25.2, + "learning_rate": 8.008032592678206e-06, + "loss": 0.8503, + "step": 291160 + }, + { + "epoch": 25.2, + "learning_rate": 8.00658788176486e-06, + "loss": 0.8871, + "step": 291170 + }, + { + "epoch": 25.2, + "learning_rate": 8.005143170851512e-06, + "loss": 0.9116, + "step": 291180 + }, + { + "epoch": 25.2, + "learning_rate": 8.003698459938166e-06, + "loss": 0.9234, + "step": 291190 + }, + { + "epoch": 25.21, + "learning_rate": 8.002253749024821e-06, + "loss": 0.93, + "step": 291200 + }, + { + "epoch": 25.21, + "learning_rate": 8.000809038111474e-06, + "loss": 0.977, + "step": 291210 + }, + { + "epoch": 25.21, + "learning_rate": 7.999364327198128e-06, + "loss": 0.9442, + "step": 291220 + }, + { + "epoch": 25.21, + "learning_rate": 7.997919616284783e-06, + "loss": 0.9088, + "step": 291230 + }, + { + "epoch": 25.21, + "learning_rate": 7.996474905371436e-06, + "loss": 0.9692, + "step": 291240 + }, + { + "epoch": 25.21, + "learning_rate": 7.99503019445809e-06, + "loss": 0.8598, + "step": 291250 + }, + { + "epoch": 25.21, + "learning_rate": 7.993585483544743e-06, + "loss": 0.8656, + "step": 291260 + }, + { + "epoch": 25.21, + "learning_rate": 7.992140772631396e-06, + "loss": 0.9174, + "step": 291270 + }, + { + "epoch": 25.21, + "learning_rate": 7.99069606171805e-06, + "loss": 0.8669, + "step": 291280 + }, + { + "epoch": 25.21, + "learning_rate": 7.989251350804703e-06, + "loss": 0.8606, + "step": 291290 + }, + { + "epoch": 25.21, + "learning_rate": 7.987806639891358e-06, + "loss": 0.8889, + "step": 291300 + }, + { + "epoch": 25.22, + "learning_rate": 7.986361928978012e-06, + "loss": 0.9018, + "step": 291310 + }, + { + "epoch": 25.22, + "learning_rate": 7.984917218064665e-06, + "loss": 0.8423, + "step": 291320 + }, + { + "epoch": 25.22, + "learning_rate": 7.98347250715132e-06, + "loss": 0.9238, + "step": 291330 + }, + { + "epoch": 25.22, + "learning_rate": 7.982027796237973e-06, + "loss": 0.8993, + "step": 291340 + }, + { + "epoch": 25.22, + "learning_rate": 7.980583085324627e-06, + "loss": 0.8544, + "step": 291350 + }, + { + "epoch": 25.22, + "learning_rate": 7.97913837441128e-06, + "loss": 0.9008, + "step": 291360 + }, + { + "epoch": 25.22, + "learning_rate": 7.977693663497934e-06, + "loss": 0.9041, + "step": 291370 + }, + { + "epoch": 25.22, + "learning_rate": 7.976248952584587e-06, + "loss": 0.9357, + "step": 291380 + }, + { + "epoch": 25.22, + "learning_rate": 7.974804241671242e-06, + "loss": 0.888, + "step": 291390 + }, + { + "epoch": 25.22, + "learning_rate": 7.973359530757896e-06, + "loss": 0.89, + "step": 291400 + }, + { + "epoch": 25.22, + "learning_rate": 7.971914819844549e-06, + "loss": 0.9216, + "step": 291410 + }, + { + "epoch": 25.22, + "learning_rate": 7.970470108931204e-06, + "loss": 0.8897, + "step": 291420 + }, + { + "epoch": 25.23, + "learning_rate": 7.969025398017857e-06, + "loss": 0.8772, + "step": 291430 + }, + { + "epoch": 25.23, + "learning_rate": 7.96758068710451e-06, + "loss": 0.9203, + "step": 291440 + }, + { + "epoch": 25.23, + "learning_rate": 7.966135976191164e-06, + "loss": 0.8704, + "step": 291450 + }, + { + "epoch": 25.23, + "learning_rate": 7.964691265277818e-06, + "loss": 0.8968, + "step": 291460 + }, + { + "epoch": 25.23, + "learning_rate": 7.963246554364471e-06, + "loss": 0.9688, + "step": 291470 + }, + { + "epoch": 25.23, + "learning_rate": 7.961801843451126e-06, + "loss": 0.8784, + "step": 291480 + }, + { + "epoch": 25.23, + "learning_rate": 7.96035713253778e-06, + "loss": 0.9236, + "step": 291490 + }, + { + "epoch": 25.23, + "learning_rate": 7.958912421624433e-06, + "loss": 0.9712, + "step": 291500 + }, + { + "epoch": 25.23, + "learning_rate": 7.957467710711088e-06, + "loss": 0.9216, + "step": 291510 + }, + { + "epoch": 25.23, + "learning_rate": 7.956022999797741e-06, + "loss": 0.9283, + "step": 291520 + }, + { + "epoch": 25.23, + "learning_rate": 7.954578288884395e-06, + "loss": 0.8587, + "step": 291530 + }, + { + "epoch": 25.23, + "learning_rate": 7.953133577971048e-06, + "loss": 0.9058, + "step": 291540 + }, + { + "epoch": 25.24, + "learning_rate": 7.951688867057702e-06, + "loss": 0.9834, + "step": 291550 + }, + { + "epoch": 25.24, + "learning_rate": 7.950244156144355e-06, + "loss": 0.8667, + "step": 291560 + }, + { + "epoch": 25.24, + "learning_rate": 7.94879944523101e-06, + "loss": 0.9273, + "step": 291570 + }, + { + "epoch": 25.24, + "learning_rate": 7.947354734317663e-06, + "loss": 0.9807, + "step": 291580 + }, + { + "epoch": 25.24, + "learning_rate": 7.945910023404317e-06, + "loss": 0.9003, + "step": 291590 + }, + { + "epoch": 25.24, + "learning_rate": 7.944465312490972e-06, + "loss": 0.868, + "step": 291600 + }, + { + "epoch": 25.24, + "learning_rate": 7.943020601577625e-06, + "loss": 0.8905, + "step": 291610 + }, + { + "epoch": 25.24, + "learning_rate": 7.941575890664279e-06, + "loss": 0.8529, + "step": 291620 + }, + { + "epoch": 25.24, + "learning_rate": 7.940131179750932e-06, + "loss": 0.9041, + "step": 291630 + }, + { + "epoch": 25.24, + "learning_rate": 7.938686468837586e-06, + "loss": 0.8679, + "step": 291640 + }, + { + "epoch": 25.24, + "learning_rate": 7.937241757924239e-06, + "loss": 0.8753, + "step": 291650 + }, + { + "epoch": 25.25, + "learning_rate": 7.935797047010894e-06, + "loss": 0.8882, + "step": 291660 + }, + { + "epoch": 25.25, + "learning_rate": 7.934352336097547e-06, + "loss": 0.9439, + "step": 291670 + }, + { + "epoch": 25.25, + "learning_rate": 7.9329076251842e-06, + "loss": 0.9191, + "step": 291680 + }, + { + "epoch": 25.25, + "learning_rate": 7.931462914270856e-06, + "loss": 0.8767, + "step": 291690 + }, + { + "epoch": 25.25, + "learning_rate": 7.93001820335751e-06, + "loss": 0.8977, + "step": 291700 + }, + { + "epoch": 25.25, + "learning_rate": 7.928573492444163e-06, + "loss": 0.8325, + "step": 291710 + }, + { + "epoch": 25.25, + "learning_rate": 7.927128781530816e-06, + "loss": 0.9518, + "step": 291720 + }, + { + "epoch": 25.25, + "learning_rate": 7.92568407061747e-06, + "loss": 0.915, + "step": 291730 + }, + { + "epoch": 25.25, + "learning_rate": 7.924239359704123e-06, + "loss": 0.8771, + "step": 291740 + }, + { + "epoch": 25.25, + "learning_rate": 7.922794648790776e-06, + "loss": 0.931, + "step": 291750 + }, + { + "epoch": 25.25, + "learning_rate": 7.921349937877431e-06, + "loss": 0.9369, + "step": 291760 + }, + { + "epoch": 25.25, + "learning_rate": 7.919905226964085e-06, + "loss": 0.867, + "step": 291770 + }, + { + "epoch": 25.26, + "learning_rate": 7.918460516050738e-06, + "loss": 0.9504, + "step": 291780 + }, + { + "epoch": 25.26, + "learning_rate": 7.917015805137393e-06, + "loss": 0.9116, + "step": 291790 + }, + { + "epoch": 25.26, + "learning_rate": 7.915571094224047e-06, + "loss": 0.9408, + "step": 291800 + }, + { + "epoch": 25.26, + "learning_rate": 7.9141263833107e-06, + "loss": 0.8302, + "step": 291810 + }, + { + "epoch": 25.26, + "learning_rate": 7.912681672397353e-06, + "loss": 0.8979, + "step": 291820 + }, + { + "epoch": 25.26, + "learning_rate": 7.911236961484007e-06, + "loss": 0.896, + "step": 291830 + }, + { + "epoch": 25.26, + "learning_rate": 7.90979225057066e-06, + "loss": 0.963, + "step": 291840 + }, + { + "epoch": 25.26, + "learning_rate": 7.908347539657315e-06, + "loss": 0.9152, + "step": 291850 + }, + { + "epoch": 25.26, + "learning_rate": 7.906902828743969e-06, + "loss": 0.9846, + "step": 291860 + }, + { + "epoch": 25.26, + "learning_rate": 7.905458117830622e-06, + "loss": 0.8844, + "step": 291870 + }, + { + "epoch": 25.26, + "learning_rate": 7.904013406917277e-06, + "loss": 0.9588, + "step": 291880 + }, + { + "epoch": 25.27, + "learning_rate": 7.90256869600393e-06, + "loss": 0.9417, + "step": 291890 + }, + { + "epoch": 25.27, + "learning_rate": 7.901123985090584e-06, + "loss": 0.8893, + "step": 291900 + }, + { + "epoch": 25.27, + "learning_rate": 7.899679274177237e-06, + "loss": 0.9216, + "step": 291910 + }, + { + "epoch": 25.27, + "learning_rate": 7.89823456326389e-06, + "loss": 0.9048, + "step": 291920 + }, + { + "epoch": 25.27, + "learning_rate": 7.896789852350544e-06, + "loss": 0.8899, + "step": 291930 + }, + { + "epoch": 25.27, + "learning_rate": 7.8953451414372e-06, + "loss": 0.9584, + "step": 291940 + }, + { + "epoch": 25.27, + "learning_rate": 7.893900430523853e-06, + "loss": 0.8881, + "step": 291950 + }, + { + "epoch": 25.27, + "learning_rate": 7.892455719610506e-06, + "loss": 0.8819, + "step": 291960 + }, + { + "epoch": 25.27, + "learning_rate": 7.891011008697161e-06, + "loss": 0.8867, + "step": 291970 + }, + { + "epoch": 25.27, + "learning_rate": 7.889566297783814e-06, + "loss": 0.9186, + "step": 291980 + }, + { + "epoch": 25.27, + "learning_rate": 7.888121586870468e-06, + "loss": 0.8826, + "step": 291990 + }, + { + "epoch": 25.27, + "learning_rate": 7.886676875957121e-06, + "loss": 0.9192, + "step": 292000 + }, + { + "epoch": 25.28, + "learning_rate": 7.885232165043775e-06, + "loss": 0.9493, + "step": 292010 + }, + { + "epoch": 25.28, + "learning_rate": 7.883787454130428e-06, + "loss": 0.8863, + "step": 292020 + }, + { + "epoch": 25.28, + "learning_rate": 7.882342743217083e-06, + "loss": 0.8996, + "step": 292030 + }, + { + "epoch": 25.28, + "learning_rate": 7.880898032303737e-06, + "loss": 0.907, + "step": 292040 + }, + { + "epoch": 25.28, + "learning_rate": 7.87945332139039e-06, + "loss": 0.9546, + "step": 292050 + }, + { + "epoch": 25.28, + "learning_rate": 7.878008610477045e-06, + "loss": 0.8923, + "step": 292060 + }, + { + "epoch": 25.28, + "learning_rate": 7.876563899563698e-06, + "loss": 0.9271, + "step": 292070 + }, + { + "epoch": 25.28, + "learning_rate": 7.875119188650352e-06, + "loss": 0.9494, + "step": 292080 + }, + { + "epoch": 25.28, + "learning_rate": 7.873674477737005e-06, + "loss": 0.8503, + "step": 292090 + }, + { + "epoch": 25.28, + "learning_rate": 7.872229766823659e-06, + "loss": 0.9272, + "step": 292100 + }, + { + "epoch": 25.28, + "learning_rate": 7.870785055910312e-06, + "loss": 0.9265, + "step": 292110 + }, + { + "epoch": 25.29, + "learning_rate": 7.869340344996967e-06, + "loss": 0.8885, + "step": 292120 + }, + { + "epoch": 25.29, + "learning_rate": 7.86789563408362e-06, + "loss": 0.8754, + "step": 292130 + }, + { + "epoch": 25.29, + "learning_rate": 7.866450923170274e-06, + "loss": 0.9073, + "step": 292140 + }, + { + "epoch": 25.29, + "learning_rate": 7.865006212256927e-06, + "loss": 0.9071, + "step": 292150 + }, + { + "epoch": 25.29, + "learning_rate": 7.863561501343582e-06, + "loss": 0.9425, + "step": 292160 + }, + { + "epoch": 25.29, + "learning_rate": 7.862116790430236e-06, + "loss": 0.934, + "step": 292170 + }, + { + "epoch": 25.29, + "learning_rate": 7.86067207951689e-06, + "loss": 0.9346, + "step": 292180 + }, + { + "epoch": 25.29, + "learning_rate": 7.859227368603543e-06, + "loss": 0.8932, + "step": 292190 + }, + { + "epoch": 25.29, + "learning_rate": 7.857782657690196e-06, + "loss": 0.9226, + "step": 292200 + }, + { + "epoch": 25.29, + "learning_rate": 7.85633794677685e-06, + "loss": 0.8476, + "step": 292210 + }, + { + "epoch": 25.29, + "learning_rate": 7.854893235863504e-06, + "loss": 0.8885, + "step": 292220 + }, + { + "epoch": 25.29, + "learning_rate": 7.853448524950158e-06, + "loss": 0.8817, + "step": 292230 + }, + { + "epoch": 25.3, + "learning_rate": 7.852003814036811e-06, + "loss": 0.9014, + "step": 292240 + }, + { + "epoch": 25.3, + "learning_rate": 7.850559103123466e-06, + "loss": 0.8365, + "step": 292250 + }, + { + "epoch": 25.3, + "learning_rate": 7.84911439221012e-06, + "loss": 0.8498, + "step": 292260 + }, + { + "epoch": 25.3, + "learning_rate": 7.847669681296773e-06, + "loss": 0.851, + "step": 292270 + }, + { + "epoch": 25.3, + "learning_rate": 7.846224970383427e-06, + "loss": 0.9427, + "step": 292280 + }, + { + "epoch": 25.3, + "learning_rate": 7.84478025947008e-06, + "loss": 0.9515, + "step": 292290 + }, + { + "epoch": 25.3, + "learning_rate": 7.843335548556733e-06, + "loss": 0.9237, + "step": 292300 + }, + { + "epoch": 25.3, + "learning_rate": 7.841890837643388e-06, + "loss": 0.9072, + "step": 292310 + }, + { + "epoch": 25.3, + "learning_rate": 7.840446126730042e-06, + "loss": 0.9009, + "step": 292320 + }, + { + "epoch": 25.3, + "learning_rate": 7.839001415816695e-06, + "loss": 0.9, + "step": 292330 + }, + { + "epoch": 25.3, + "learning_rate": 7.83755670490335e-06, + "loss": 0.9295, + "step": 292340 + }, + { + "epoch": 25.31, + "learning_rate": 7.836111993990004e-06, + "loss": 0.9612, + "step": 292350 + }, + { + "epoch": 25.31, + "learning_rate": 7.834667283076657e-06, + "loss": 0.9098, + "step": 292360 + }, + { + "epoch": 25.31, + "learning_rate": 7.83322257216331e-06, + "loss": 0.9423, + "step": 292370 + }, + { + "epoch": 25.31, + "learning_rate": 7.831777861249964e-06, + "loss": 0.9359, + "step": 292380 + }, + { + "epoch": 25.31, + "learning_rate": 7.830333150336617e-06, + "loss": 0.9428, + "step": 292390 + }, + { + "epoch": 25.31, + "learning_rate": 7.828888439423272e-06, + "loss": 0.9199, + "step": 292400 + }, + { + "epoch": 25.31, + "learning_rate": 7.827443728509926e-06, + "loss": 0.8732, + "step": 292410 + }, + { + "epoch": 25.31, + "learning_rate": 7.825999017596579e-06, + "loss": 0.9223, + "step": 292420 + }, + { + "epoch": 25.31, + "learning_rate": 7.824554306683234e-06, + "loss": 0.9667, + "step": 292430 + }, + { + "epoch": 25.31, + "learning_rate": 7.823109595769888e-06, + "loss": 0.8919, + "step": 292440 + }, + { + "epoch": 25.31, + "learning_rate": 7.821664884856541e-06, + "loss": 0.9257, + "step": 292450 + }, + { + "epoch": 25.31, + "learning_rate": 7.820220173943194e-06, + "loss": 0.9546, + "step": 292460 + }, + { + "epoch": 25.32, + "learning_rate": 7.818775463029848e-06, + "loss": 0.9697, + "step": 292470 + }, + { + "epoch": 25.32, + "learning_rate": 7.817330752116501e-06, + "loss": 0.8981, + "step": 292480 + }, + { + "epoch": 25.32, + "learning_rate": 7.815886041203156e-06, + "loss": 0.8695, + "step": 292490 + }, + { + "epoch": 25.32, + "learning_rate": 7.81444133028981e-06, + "loss": 0.9146, + "step": 292500 + }, + { + "epoch": 25.32, + "learning_rate": 7.812996619376463e-06, + "loss": 0.9593, + "step": 292510 + }, + { + "epoch": 25.32, + "learning_rate": 7.811551908463118e-06, + "loss": 0.8829, + "step": 292520 + }, + { + "epoch": 25.32, + "learning_rate": 7.810107197549772e-06, + "loss": 0.9113, + "step": 292530 + }, + { + "epoch": 25.32, + "learning_rate": 7.808662486636425e-06, + "loss": 0.9145, + "step": 292540 + }, + { + "epoch": 25.32, + "learning_rate": 7.807217775723078e-06, + "loss": 0.9278, + "step": 292550 + }, + { + "epoch": 25.32, + "learning_rate": 7.805773064809732e-06, + "loss": 0.8987, + "step": 292560 + }, + { + "epoch": 25.32, + "learning_rate": 7.804328353896385e-06, + "loss": 0.9632, + "step": 292570 + }, + { + "epoch": 25.32, + "learning_rate": 7.80288364298304e-06, + "loss": 0.9681, + "step": 292580 + }, + { + "epoch": 25.33, + "learning_rate": 7.801438932069694e-06, + "loss": 0.8714, + "step": 292590 + }, + { + "epoch": 25.33, + "learning_rate": 7.799994221156347e-06, + "loss": 0.8378, + "step": 292600 + }, + { + "epoch": 25.33, + "learning_rate": 7.798549510243e-06, + "loss": 0.9481, + "step": 292610 + }, + { + "epoch": 25.33, + "learning_rate": 7.797104799329655e-06, + "loss": 0.879, + "step": 292620 + }, + { + "epoch": 25.33, + "learning_rate": 7.795660088416309e-06, + "loss": 0.8288, + "step": 292630 + }, + { + "epoch": 25.33, + "learning_rate": 7.79421537750296e-06, + "loss": 0.9199, + "step": 292640 + }, + { + "epoch": 25.33, + "learning_rate": 7.792770666589616e-06, + "loss": 0.9339, + "step": 292650 + }, + { + "epoch": 25.33, + "learning_rate": 7.791325955676269e-06, + "loss": 0.9252, + "step": 292660 + }, + { + "epoch": 25.33, + "learning_rate": 7.789881244762922e-06, + "loss": 0.812, + "step": 292670 + }, + { + "epoch": 25.33, + "learning_rate": 7.788436533849578e-06, + "loss": 0.8766, + "step": 292680 + }, + { + "epoch": 25.33, + "learning_rate": 7.786991822936231e-06, + "loss": 0.9567, + "step": 292690 + }, + { + "epoch": 25.34, + "learning_rate": 7.785547112022884e-06, + "loss": 0.857, + "step": 292700 + }, + { + "epoch": 25.34, + "learning_rate": 7.78410240110954e-06, + "loss": 0.921, + "step": 292710 + }, + { + "epoch": 25.34, + "learning_rate": 7.782657690196193e-06, + "loss": 0.9087, + "step": 292720 + }, + { + "epoch": 25.34, + "learning_rate": 7.781212979282845e-06, + "loss": 0.8411, + "step": 292730 + }, + { + "epoch": 25.34, + "learning_rate": 7.7797682683695e-06, + "loss": 0.9721, + "step": 292740 + }, + { + "epoch": 25.34, + "learning_rate": 7.778323557456153e-06, + "loss": 0.9156, + "step": 292750 + }, + { + "epoch": 25.34, + "learning_rate": 7.776878846542806e-06, + "loss": 0.944, + "step": 292760 + }, + { + "epoch": 25.34, + "learning_rate": 7.775434135629461e-06, + "loss": 0.9497, + "step": 292770 + }, + { + "epoch": 25.34, + "learning_rate": 7.773989424716115e-06, + "loss": 0.8717, + "step": 292780 + }, + { + "epoch": 25.34, + "learning_rate": 7.772544713802768e-06, + "loss": 0.9369, + "step": 292790 + }, + { + "epoch": 25.34, + "learning_rate": 7.771100002889423e-06, + "loss": 0.9096, + "step": 292800 + }, + { + "epoch": 25.34, + "learning_rate": 7.769655291976077e-06, + "loss": 0.9294, + "step": 292810 + }, + { + "epoch": 25.35, + "learning_rate": 7.768210581062728e-06, + "loss": 0.9246, + "step": 292820 + }, + { + "epoch": 25.35, + "learning_rate": 7.766765870149384e-06, + "loss": 0.955, + "step": 292830 + }, + { + "epoch": 25.35, + "learning_rate": 7.765321159236037e-06, + "loss": 0.9181, + "step": 292840 + }, + { + "epoch": 25.35, + "learning_rate": 7.76387644832269e-06, + "loss": 0.8682, + "step": 292850 + }, + { + "epoch": 25.35, + "learning_rate": 7.762431737409345e-06, + "loss": 0.9349, + "step": 292860 + }, + { + "epoch": 25.35, + "learning_rate": 7.760987026495999e-06, + "loss": 0.917, + "step": 292870 + }, + { + "epoch": 25.35, + "learning_rate": 7.759542315582652e-06, + "loss": 0.9665, + "step": 292880 + }, + { + "epoch": 25.35, + "learning_rate": 7.758097604669307e-06, + "loss": 0.924, + "step": 292890 + }, + { + "epoch": 25.35, + "learning_rate": 7.75665289375596e-06, + "loss": 0.886, + "step": 292900 + }, + { + "epoch": 25.35, + "learning_rate": 7.755208182842612e-06, + "loss": 0.9003, + "step": 292910 + }, + { + "epoch": 25.35, + "learning_rate": 7.753763471929267e-06, + "loss": 0.9528, + "step": 292920 + }, + { + "epoch": 25.36, + "learning_rate": 7.752318761015921e-06, + "loss": 0.9311, + "step": 292930 + }, + { + "epoch": 25.36, + "learning_rate": 7.750874050102574e-06, + "loss": 0.9304, + "step": 292940 + }, + { + "epoch": 25.36, + "learning_rate": 7.74942933918923e-06, + "loss": 0.9043, + "step": 292950 + }, + { + "epoch": 25.36, + "learning_rate": 7.747984628275883e-06, + "loss": 0.8794, + "step": 292960 + }, + { + "epoch": 25.36, + "learning_rate": 7.746539917362536e-06, + "loss": 0.9518, + "step": 292970 + }, + { + "epoch": 25.36, + "learning_rate": 7.745095206449191e-06, + "loss": 0.8709, + "step": 292980 + }, + { + "epoch": 25.36, + "learning_rate": 7.743650495535845e-06, + "loss": 0.8951, + "step": 292990 + }, + { + "epoch": 25.36, + "learning_rate": 7.742205784622496e-06, + "loss": 0.9056, + "step": 293000 + }, + { + "epoch": 25.36, + "learning_rate": 7.740761073709151e-06, + "loss": 0.9003, + "step": 293010 + }, + { + "epoch": 25.36, + "learning_rate": 7.739316362795805e-06, + "loss": 0.9442, + "step": 293020 + }, + { + "epoch": 25.36, + "learning_rate": 7.737871651882458e-06, + "loss": 0.8977, + "step": 293030 + }, + { + "epoch": 25.36, + "learning_rate": 7.736426940969113e-06, + "loss": 0.9248, + "step": 293040 + }, + { + "epoch": 25.37, + "learning_rate": 7.734982230055767e-06, + "loss": 0.881, + "step": 293050 + }, + { + "epoch": 25.37, + "learning_rate": 7.73353751914242e-06, + "loss": 0.8918, + "step": 293060 + }, + { + "epoch": 25.37, + "learning_rate": 7.732092808229074e-06, + "loss": 0.9539, + "step": 293070 + }, + { + "epoch": 25.37, + "learning_rate": 7.730648097315729e-06, + "loss": 0.9004, + "step": 293080 + }, + { + "epoch": 25.37, + "learning_rate": 7.72920338640238e-06, + "loss": 0.9185, + "step": 293090 + }, + { + "epoch": 25.37, + "learning_rate": 7.727758675489034e-06, + "loss": 0.9218, + "step": 293100 + }, + { + "epoch": 25.37, + "learning_rate": 7.726313964575689e-06, + "loss": 0.9254, + "step": 293110 + }, + { + "epoch": 25.37, + "learning_rate": 7.724869253662342e-06, + "loss": 0.9206, + "step": 293120 + }, + { + "epoch": 25.37, + "learning_rate": 7.723424542748996e-06, + "loss": 0.8771, + "step": 293130 + }, + { + "epoch": 25.37, + "learning_rate": 7.72197983183565e-06, + "loss": 0.9298, + "step": 293140 + }, + { + "epoch": 25.37, + "learning_rate": 7.720535120922304e-06, + "loss": 0.8843, + "step": 293150 + }, + { + "epoch": 25.38, + "learning_rate": 7.719090410008957e-06, + "loss": 0.9355, + "step": 293160 + }, + { + "epoch": 25.38, + "learning_rate": 7.717645699095613e-06, + "loss": 0.9237, + "step": 293170 + }, + { + "epoch": 25.38, + "learning_rate": 7.716200988182264e-06, + "loss": 0.942, + "step": 293180 + }, + { + "epoch": 25.38, + "learning_rate": 7.714756277268918e-06, + "loss": 0.878, + "step": 293190 + }, + { + "epoch": 25.38, + "learning_rate": 7.713311566355573e-06, + "loss": 0.8885, + "step": 293200 + }, + { + "epoch": 25.38, + "learning_rate": 7.711866855442226e-06, + "loss": 0.9124, + "step": 293210 + }, + { + "epoch": 25.38, + "learning_rate": 7.71042214452888e-06, + "loss": 0.8436, + "step": 293220 + }, + { + "epoch": 25.38, + "learning_rate": 7.708977433615535e-06, + "loss": 0.8841, + "step": 293230 + }, + { + "epoch": 25.38, + "learning_rate": 7.707532722702188e-06, + "loss": 0.9306, + "step": 293240 + }, + { + "epoch": 25.38, + "learning_rate": 7.706088011788841e-06, + "loss": 0.9393, + "step": 293250 + }, + { + "epoch": 25.38, + "learning_rate": 7.704643300875495e-06, + "loss": 0.9074, + "step": 293260 + }, + { + "epoch": 25.38, + "learning_rate": 7.703198589962148e-06, + "loss": 0.9505, + "step": 293270 + }, + { + "epoch": 25.39, + "learning_rate": 7.701753879048802e-06, + "loss": 0.8537, + "step": 293280 + }, + { + "epoch": 25.39, + "learning_rate": 7.700309168135457e-06, + "loss": 0.907, + "step": 293290 + }, + { + "epoch": 25.39, + "learning_rate": 7.69886445722211e-06, + "loss": 0.9475, + "step": 293300 + }, + { + "epoch": 25.39, + "learning_rate": 7.697419746308763e-06, + "loss": 0.8945, + "step": 293310 + }, + { + "epoch": 25.39, + "learning_rate": 7.695975035395419e-06, + "loss": 0.848, + "step": 293320 + }, + { + "epoch": 25.39, + "learning_rate": 7.694530324482072e-06, + "loss": 0.797, + "step": 293330 + }, + { + "epoch": 25.39, + "learning_rate": 7.693085613568725e-06, + "loss": 0.8835, + "step": 293340 + }, + { + "epoch": 25.39, + "learning_rate": 7.691640902655379e-06, + "loss": 0.8873, + "step": 293350 + }, + { + "epoch": 25.39, + "learning_rate": 7.690196191742032e-06, + "loss": 0.9427, + "step": 293360 + }, + { + "epoch": 25.39, + "learning_rate": 7.688751480828686e-06, + "loss": 0.9877, + "step": 293370 + }, + { + "epoch": 25.39, + "learning_rate": 7.68730676991534e-06, + "loss": 0.9407, + "step": 293380 + }, + { + "epoch": 25.4, + "learning_rate": 7.685862059001994e-06, + "loss": 0.8927, + "step": 293390 + }, + { + "epoch": 25.4, + "learning_rate": 7.684417348088647e-06, + "loss": 0.9225, + "step": 293400 + }, + { + "epoch": 25.4, + "learning_rate": 7.682972637175302e-06, + "loss": 0.9194, + "step": 293410 + }, + { + "epoch": 25.4, + "learning_rate": 7.681527926261956e-06, + "loss": 0.8563, + "step": 293420 + }, + { + "epoch": 25.4, + "learning_rate": 7.68008321534861e-06, + "loss": 0.9118, + "step": 293430 + }, + { + "epoch": 25.4, + "learning_rate": 7.678638504435263e-06, + "loss": 0.9099, + "step": 293440 + }, + { + "epoch": 25.4, + "learning_rate": 7.677193793521916e-06, + "loss": 0.9487, + "step": 293450 + }, + { + "epoch": 25.4, + "learning_rate": 7.67574908260857e-06, + "loss": 0.9422, + "step": 293460 + }, + { + "epoch": 25.4, + "learning_rate": 7.674304371695225e-06, + "loss": 0.8854, + "step": 293470 + }, + { + "epoch": 25.4, + "learning_rate": 7.672859660781878e-06, + "loss": 0.8943, + "step": 293480 + }, + { + "epoch": 25.4, + "learning_rate": 7.671414949868531e-06, + "loss": 0.8599, + "step": 293490 + }, + { + "epoch": 25.4, + "learning_rate": 7.669970238955186e-06, + "loss": 0.8869, + "step": 293500 + }, + { + "epoch": 25.41, + "learning_rate": 7.66852552804184e-06, + "loss": 0.87, + "step": 293510 + }, + { + "epoch": 25.41, + "learning_rate": 7.667080817128493e-06, + "loss": 0.8851, + "step": 293520 + }, + { + "epoch": 25.41, + "learning_rate": 7.665636106215147e-06, + "loss": 0.9178, + "step": 293530 + }, + { + "epoch": 25.41, + "learning_rate": 7.6641913953018e-06, + "loss": 0.9059, + "step": 293540 + }, + { + "epoch": 25.41, + "learning_rate": 7.662746684388453e-06, + "loss": 0.8769, + "step": 293550 + }, + { + "epoch": 25.41, + "learning_rate": 7.661301973475107e-06, + "loss": 0.9437, + "step": 293560 + }, + { + "epoch": 25.41, + "learning_rate": 7.659857262561762e-06, + "loss": 0.8836, + "step": 293570 + }, + { + "epoch": 25.41, + "learning_rate": 7.658412551648415e-06, + "loss": 0.9043, + "step": 293580 + }, + { + "epoch": 25.41, + "learning_rate": 7.656967840735069e-06, + "loss": 0.9524, + "step": 293590 + }, + { + "epoch": 25.41, + "learning_rate": 7.655523129821724e-06, + "loss": 0.9226, + "step": 293600 + }, + { + "epoch": 25.41, + "learning_rate": 7.654078418908377e-06, + "loss": 0.9545, + "step": 293610 + }, + { + "epoch": 25.41, + "learning_rate": 7.65263370799503e-06, + "loss": 0.9447, + "step": 293620 + }, + { + "epoch": 25.42, + "learning_rate": 7.651188997081684e-06, + "loss": 0.9923, + "step": 293630 + }, + { + "epoch": 25.42, + "learning_rate": 7.649744286168337e-06, + "loss": 0.8943, + "step": 293640 + }, + { + "epoch": 25.42, + "learning_rate": 7.64829957525499e-06, + "loss": 0.9183, + "step": 293650 + }, + { + "epoch": 25.42, + "learning_rate": 7.646854864341646e-06, + "loss": 0.8941, + "step": 293660 + }, + { + "epoch": 25.42, + "learning_rate": 7.6454101534283e-06, + "loss": 0.8628, + "step": 293670 + }, + { + "epoch": 25.42, + "learning_rate": 7.643965442514953e-06, + "loss": 0.9097, + "step": 293680 + }, + { + "epoch": 25.42, + "learning_rate": 7.642520731601608e-06, + "loss": 0.8913, + "step": 293690 + }, + { + "epoch": 25.42, + "learning_rate": 7.641076020688261e-06, + "loss": 0.8684, + "step": 293700 + }, + { + "epoch": 25.42, + "learning_rate": 7.639631309774914e-06, + "loss": 0.9088, + "step": 293710 + }, + { + "epoch": 25.42, + "learning_rate": 7.638186598861568e-06, + "loss": 0.8771, + "step": 293720 + }, + { + "epoch": 25.42, + "learning_rate": 7.636741887948221e-06, + "loss": 0.9378, + "step": 293730 + }, + { + "epoch": 25.43, + "learning_rate": 7.635297177034875e-06, + "loss": 0.9439, + "step": 293740 + }, + { + "epoch": 25.43, + "learning_rate": 7.63385246612153e-06, + "loss": 0.8372, + "step": 293750 + }, + { + "epoch": 25.43, + "learning_rate": 7.632407755208183e-06, + "loss": 0.9536, + "step": 293760 + }, + { + "epoch": 25.43, + "learning_rate": 7.630963044294837e-06, + "loss": 0.8989, + "step": 293770 + }, + { + "epoch": 25.43, + "learning_rate": 7.629518333381492e-06, + "loss": 0.9161, + "step": 293780 + }, + { + "epoch": 25.43, + "learning_rate": 7.628073622468144e-06, + "loss": 0.9476, + "step": 293790 + }, + { + "epoch": 25.43, + "learning_rate": 7.626628911554798e-06, + "loss": 0.8821, + "step": 293800 + }, + { + "epoch": 25.43, + "learning_rate": 7.625184200641453e-06, + "loss": 0.9523, + "step": 293810 + }, + { + "epoch": 25.43, + "learning_rate": 7.623739489728106e-06, + "loss": 0.9454, + "step": 293820 + }, + { + "epoch": 25.43, + "learning_rate": 7.6222947788147595e-06, + "loss": 0.9097, + "step": 293830 + }, + { + "epoch": 25.43, + "learning_rate": 7.620850067901414e-06, + "loss": 0.8961, + "step": 293840 + }, + { + "epoch": 25.43, + "learning_rate": 7.619405356988067e-06, + "loss": 0.8652, + "step": 293850 + }, + { + "epoch": 25.44, + "learning_rate": 7.6179606460747205e-06, + "loss": 0.9067, + "step": 293860 + }, + { + "epoch": 25.44, + "learning_rate": 7.616515935161376e-06, + "loss": 0.8947, + "step": 293870 + }, + { + "epoch": 25.44, + "learning_rate": 7.615071224248028e-06, + "loss": 0.9476, + "step": 293880 + }, + { + "epoch": 25.44, + "learning_rate": 7.6136265133346815e-06, + "loss": 0.872, + "step": 293890 + }, + { + "epoch": 25.44, + "learning_rate": 7.612181802421337e-06, + "loss": 0.9105, + "step": 293900 + }, + { + "epoch": 25.44, + "learning_rate": 7.61073709150799e-06, + "loss": 0.9456, + "step": 293910 + }, + { + "epoch": 25.44, + "learning_rate": 7.609292380594643e-06, + "loss": 0.9044, + "step": 293920 + }, + { + "epoch": 25.44, + "learning_rate": 7.607847669681298e-06, + "loss": 0.9091, + "step": 293930 + }, + { + "epoch": 25.44, + "learning_rate": 7.606402958767951e-06, + "loss": 0.9055, + "step": 293940 + }, + { + "epoch": 25.44, + "learning_rate": 7.6049582478546044e-06, + "loss": 0.9067, + "step": 293950 + }, + { + "epoch": 25.44, + "learning_rate": 7.603513536941258e-06, + "loss": 0.9415, + "step": 293960 + }, + { + "epoch": 25.45, + "learning_rate": 7.602068826027912e-06, + "loss": 0.9336, + "step": 293970 + }, + { + "epoch": 25.45, + "learning_rate": 7.6006241151145655e-06, + "loss": 0.9172, + "step": 293980 + }, + { + "epoch": 25.45, + "learning_rate": 7.599179404201219e-06, + "loss": 0.9443, + "step": 293990 + }, + { + "epoch": 25.45, + "learning_rate": 7.597734693287874e-06, + "loss": 0.8939, + "step": 294000 + }, + { + "epoch": 25.45, + "learning_rate": 7.596289982374527e-06, + "loss": 0.9444, + "step": 294010 + }, + { + "epoch": 25.45, + "learning_rate": 7.594845271461181e-06, + "loss": 0.9458, + "step": 294020 + }, + { + "epoch": 25.45, + "learning_rate": 7.593400560547835e-06, + "loss": 0.947, + "step": 294030 + }, + { + "epoch": 25.45, + "learning_rate": 7.591955849634488e-06, + "loss": 0.8875, + "step": 294040 + }, + { + "epoch": 25.45, + "learning_rate": 7.590511138721142e-06, + "loss": 0.9211, + "step": 294050 + }, + { + "epoch": 25.45, + "learning_rate": 7.589066427807796e-06, + "loss": 0.9531, + "step": 294060 + }, + { + "epoch": 25.45, + "learning_rate": 7.5876217168944494e-06, + "loss": 0.8979, + "step": 294070 + }, + { + "epoch": 25.45, + "learning_rate": 7.586177005981103e-06, + "loss": 0.8694, + "step": 294080 + }, + { + "epoch": 25.46, + "learning_rate": 7.584732295067758e-06, + "loss": 0.9078, + "step": 294090 + }, + { + "epoch": 25.46, + "learning_rate": 7.583287584154411e-06, + "loss": 0.9071, + "step": 294100 + }, + { + "epoch": 25.46, + "learning_rate": 7.581842873241065e-06, + "loss": 0.8931, + "step": 294110 + }, + { + "epoch": 25.46, + "learning_rate": 7.580398162327719e-06, + "loss": 0.9137, + "step": 294120 + }, + { + "epoch": 25.46, + "learning_rate": 7.578953451414372e-06, + "loss": 0.9169, + "step": 294130 + }, + { + "epoch": 25.46, + "learning_rate": 7.577508740501026e-06, + "loss": 0.9493, + "step": 294140 + }, + { + "epoch": 25.46, + "learning_rate": 7.57606402958768e-06, + "loss": 0.9339, + "step": 294150 + }, + { + "epoch": 25.46, + "learning_rate": 7.574619318674333e-06, + "loss": 0.8698, + "step": 294160 + }, + { + "epoch": 25.46, + "learning_rate": 7.573174607760987e-06, + "loss": 0.8685, + "step": 294170 + }, + { + "epoch": 25.46, + "learning_rate": 7.571729896847642e-06, + "loss": 0.9501, + "step": 294180 + }, + { + "epoch": 25.46, + "learning_rate": 7.570285185934295e-06, + "loss": 0.9314, + "step": 294190 + }, + { + "epoch": 25.47, + "learning_rate": 7.568840475020949e-06, + "loss": 0.9497, + "step": 294200 + }, + { + "epoch": 25.47, + "learning_rate": 7.567395764107603e-06, + "loss": 0.9671, + "step": 294210 + }, + { + "epoch": 25.47, + "learning_rate": 7.565951053194256e-06, + "loss": 0.9045, + "step": 294220 + }, + { + "epoch": 25.47, + "learning_rate": 7.56450634228091e-06, + "loss": 0.8751, + "step": 294230 + }, + { + "epoch": 25.47, + "learning_rate": 7.563061631367564e-06, + "loss": 0.8893, + "step": 294240 + }, + { + "epoch": 25.47, + "learning_rate": 7.561616920454217e-06, + "loss": 0.8329, + "step": 294250 + }, + { + "epoch": 25.47, + "learning_rate": 7.560172209540871e-06, + "loss": 0.8908, + "step": 294260 + }, + { + "epoch": 25.47, + "learning_rate": 7.558727498627526e-06, + "loss": 0.9474, + "step": 294270 + }, + { + "epoch": 25.47, + "learning_rate": 7.557282787714179e-06, + "loss": 0.9063, + "step": 294280 + }, + { + "epoch": 25.47, + "learning_rate": 7.555838076800833e-06, + "loss": 0.8614, + "step": 294290 + }, + { + "epoch": 25.47, + "learning_rate": 7.554393365887487e-06, + "loss": 0.8785, + "step": 294300 + }, + { + "epoch": 25.47, + "learning_rate": 7.55294865497414e-06, + "loss": 0.9495, + "step": 294310 + }, + { + "epoch": 25.48, + "learning_rate": 7.551503944060794e-06, + "loss": 0.9348, + "step": 294320 + }, + { + "epoch": 25.48, + "learning_rate": 7.550059233147448e-06, + "loss": 0.9398, + "step": 294330 + }, + { + "epoch": 25.48, + "learning_rate": 7.548614522234101e-06, + "loss": 0.8858, + "step": 294340 + }, + { + "epoch": 25.48, + "learning_rate": 7.547169811320755e-06, + "loss": 0.9203, + "step": 294350 + }, + { + "epoch": 25.48, + "learning_rate": 7.54572510040741e-06, + "loss": 0.936, + "step": 294360 + }, + { + "epoch": 25.48, + "learning_rate": 7.544280389494063e-06, + "loss": 0.9226, + "step": 294370 + }, + { + "epoch": 25.48, + "learning_rate": 7.5428356785807165e-06, + "loss": 0.9134, + "step": 294380 + }, + { + "epoch": 25.48, + "learning_rate": 7.541390967667371e-06, + "loss": 0.9436, + "step": 294390 + }, + { + "epoch": 25.48, + "learning_rate": 7.539946256754024e-06, + "loss": 0.8945, + "step": 294400 + }, + { + "epoch": 25.48, + "learning_rate": 7.5385015458406776e-06, + "loss": 0.9192, + "step": 294410 + }, + { + "epoch": 25.48, + "learning_rate": 7.537056834927331e-06, + "loss": 0.9181, + "step": 294420 + }, + { + "epoch": 25.49, + "learning_rate": 7.535612124013985e-06, + "loss": 0.9075, + "step": 294430 + }, + { + "epoch": 25.49, + "learning_rate": 7.534167413100639e-06, + "loss": 0.8975, + "step": 294440 + }, + { + "epoch": 25.49, + "learning_rate": 7.532722702187292e-06, + "loss": 0.9214, + "step": 294450 + }, + { + "epoch": 25.49, + "learning_rate": 7.531277991273947e-06, + "loss": 0.8524, + "step": 294460 + }, + { + "epoch": 25.49, + "learning_rate": 7.5298332803606e-06, + "loss": 0.908, + "step": 294470 + }, + { + "epoch": 25.49, + "learning_rate": 7.528388569447253e-06, + "loss": 0.921, + "step": 294480 + }, + { + "epoch": 25.49, + "learning_rate": 7.526943858533908e-06, + "loss": 0.9394, + "step": 294490 + }, + { + "epoch": 25.49, + "learning_rate": 7.5254991476205615e-06, + "loss": 0.9293, + "step": 294500 + }, + { + "epoch": 25.49, + "learning_rate": 7.524054436707215e-06, + "loss": 0.9125, + "step": 294510 + }, + { + "epoch": 25.49, + "learning_rate": 7.522609725793869e-06, + "loss": 0.8854, + "step": 294520 + }, + { + "epoch": 25.49, + "learning_rate": 7.5211650148805225e-06, + "loss": 0.908, + "step": 294530 + }, + { + "epoch": 25.49, + "learning_rate": 7.519720303967176e-06, + "loss": 0.8904, + "step": 294540 + }, + { + "epoch": 25.5, + "learning_rate": 7.518275593053831e-06, + "loss": 0.877, + "step": 294550 + }, + { + "epoch": 25.5, + "learning_rate": 7.5168308821404836e-06, + "loss": 0.8982, + "step": 294560 + }, + { + "epoch": 25.5, + "learning_rate": 7.515386171227137e-06, + "loss": 0.9417, + "step": 294570 + }, + { + "epoch": 25.5, + "learning_rate": 7.513941460313792e-06, + "loss": 0.9234, + "step": 294580 + }, + { + "epoch": 25.5, + "learning_rate": 7.5124967494004454e-06, + "loss": 0.9067, + "step": 294590 + }, + { + "epoch": 25.5, + "learning_rate": 7.511052038487099e-06, + "loss": 0.9128, + "step": 294600 + }, + { + "epoch": 25.5, + "learning_rate": 7.509607327573753e-06, + "loss": 0.9416, + "step": 294610 + }, + { + "epoch": 25.5, + "learning_rate": 7.5081626166604065e-06, + "loss": 0.8984, + "step": 294620 + }, + { + "epoch": 25.5, + "learning_rate": 7.50671790574706e-06, + "loss": 0.9324, + "step": 294630 + }, + { + "epoch": 25.5, + "learning_rate": 7.505273194833715e-06, + "loss": 0.865, + "step": 294640 + }, + { + "epoch": 25.5, + "learning_rate": 7.5038284839203675e-06, + "loss": 0.9259, + "step": 294650 + }, + { + "epoch": 25.5, + "learning_rate": 7.502383773007021e-06, + "loss": 0.857, + "step": 294660 + }, + { + "epoch": 25.51, + "learning_rate": 7.500939062093676e-06, + "loss": 0.9367, + "step": 294670 + }, + { + "epoch": 25.51, + "learning_rate": 7.499494351180329e-06, + "loss": 0.8607, + "step": 294680 + }, + { + "epoch": 25.51, + "learning_rate": 7.498049640266983e-06, + "loss": 0.9219, + "step": 294690 + }, + { + "epoch": 25.51, + "learning_rate": 7.496604929353637e-06, + "loss": 0.947, + "step": 294700 + }, + { + "epoch": 25.51, + "learning_rate": 7.49516021844029e-06, + "loss": 0.8514, + "step": 294710 + }, + { + "epoch": 25.51, + "learning_rate": 7.493715507526944e-06, + "loss": 0.9626, + "step": 294720 + }, + { + "epoch": 25.51, + "learning_rate": 7.492270796613599e-06, + "loss": 0.8599, + "step": 294730 + }, + { + "epoch": 25.51, + "learning_rate": 7.4908260857002515e-06, + "loss": 0.8839, + "step": 294740 + }, + { + "epoch": 25.51, + "learning_rate": 7.489381374786905e-06, + "loss": 0.8763, + "step": 294750 + }, + { + "epoch": 25.51, + "learning_rate": 7.48793666387356e-06, + "loss": 0.9411, + "step": 294760 + }, + { + "epoch": 25.51, + "learning_rate": 7.486491952960213e-06, + "loss": 0.8885, + "step": 294770 + }, + { + "epoch": 25.52, + "learning_rate": 7.485047242046867e-06, + "loss": 0.9054, + "step": 294780 + }, + { + "epoch": 25.52, + "learning_rate": 7.483602531133521e-06, + "loss": 0.9408, + "step": 294790 + }, + { + "epoch": 25.52, + "learning_rate": 7.482157820220174e-06, + "loss": 0.9351, + "step": 294800 + }, + { + "epoch": 25.52, + "learning_rate": 7.480713109306828e-06, + "loss": 0.8898, + "step": 294810 + }, + { + "epoch": 25.52, + "learning_rate": 7.479268398393483e-06, + "loss": 0.9153, + "step": 294820 + }, + { + "epoch": 25.52, + "learning_rate": 7.477823687480135e-06, + "loss": 0.9158, + "step": 294830 + }, + { + "epoch": 25.52, + "learning_rate": 7.476378976566789e-06, + "loss": 0.9943, + "step": 294840 + }, + { + "epoch": 25.52, + "learning_rate": 7.474934265653444e-06, + "loss": 0.9299, + "step": 294850 + }, + { + "epoch": 25.52, + "learning_rate": 7.473489554740097e-06, + "loss": 0.898, + "step": 294860 + }, + { + "epoch": 25.52, + "learning_rate": 7.472044843826751e-06, + "loss": 0.8836, + "step": 294870 + }, + { + "epoch": 25.52, + "learning_rate": 7.470600132913404e-06, + "loss": 0.905, + "step": 294880 + }, + { + "epoch": 25.52, + "learning_rate": 7.469155422000058e-06, + "loss": 0.9175, + "step": 294890 + }, + { + "epoch": 25.53, + "learning_rate": 7.467710711086712e-06, + "loss": 0.9882, + "step": 294900 + }, + { + "epoch": 25.53, + "learning_rate": 7.466266000173365e-06, + "loss": 0.8861, + "step": 294910 + }, + { + "epoch": 25.53, + "learning_rate": 7.464821289260019e-06, + "loss": 0.8699, + "step": 294920 + }, + { + "epoch": 25.53, + "learning_rate": 7.463376578346673e-06, + "loss": 0.9649, + "step": 294930 + }, + { + "epoch": 25.53, + "learning_rate": 7.461931867433326e-06, + "loss": 0.9352, + "step": 294940 + }, + { + "epoch": 25.53, + "learning_rate": 7.460487156519981e-06, + "loss": 0.8836, + "step": 294950 + }, + { + "epoch": 25.53, + "learning_rate": 7.459042445606635e-06, + "loss": 0.9325, + "step": 294960 + }, + { + "epoch": 25.53, + "learning_rate": 7.457597734693288e-06, + "loss": 0.8623, + "step": 294970 + }, + { + "epoch": 25.53, + "learning_rate": 7.456153023779942e-06, + "loss": 0.9373, + "step": 294980 + }, + { + "epoch": 25.53, + "learning_rate": 7.454708312866596e-06, + "loss": 0.9491, + "step": 294990 + }, + { + "epoch": 25.53, + "learning_rate": 7.453263601953249e-06, + "loss": 0.8955, + "step": 295000 + }, + { + "epoch": 25.54, + "learning_rate": 7.451818891039903e-06, + "loss": 0.8767, + "step": 295010 + }, + { + "epoch": 25.54, + "learning_rate": 7.450374180126557e-06, + "loss": 0.9217, + "step": 295020 + }, + { + "epoch": 25.54, + "learning_rate": 7.44892946921321e-06, + "loss": 0.9104, + "step": 295030 + }, + { + "epoch": 25.54, + "learning_rate": 7.447484758299865e-06, + "loss": 0.945, + "step": 295040 + }, + { + "epoch": 25.54, + "learning_rate": 7.4460400473865186e-06, + "loss": 0.8625, + "step": 295050 + }, + { + "epoch": 25.54, + "learning_rate": 7.444595336473172e-06, + "loss": 0.9029, + "step": 295060 + }, + { + "epoch": 25.54, + "learning_rate": 7.443150625559826e-06, + "loss": 0.9404, + "step": 295070 + }, + { + "epoch": 25.54, + "learning_rate": 7.44170591464648e-06, + "loss": 0.9501, + "step": 295080 + }, + { + "epoch": 25.54, + "learning_rate": 7.440261203733133e-06, + "loss": 0.9535, + "step": 295090 + }, + { + "epoch": 25.54, + "learning_rate": 7.438816492819787e-06, + "loss": 0.8958, + "step": 295100 + }, + { + "epoch": 25.54, + "learning_rate": 7.437371781906441e-06, + "loss": 0.937, + "step": 295110 + }, + { + "epoch": 25.54, + "learning_rate": 7.435927070993094e-06, + "loss": 0.9278, + "step": 295120 + }, + { + "epoch": 25.55, + "learning_rate": 7.434482360079749e-06, + "loss": 0.9366, + "step": 295130 + }, + { + "epoch": 25.55, + "learning_rate": 7.4330376491664025e-06, + "loss": 0.9146, + "step": 295140 + }, + { + "epoch": 25.55, + "learning_rate": 7.431592938253056e-06, + "loss": 0.9292, + "step": 295150 + }, + { + "epoch": 25.55, + "learning_rate": 7.43014822733971e-06, + "loss": 0.8795, + "step": 295160 + }, + { + "epoch": 25.55, + "learning_rate": 7.4287035164263635e-06, + "loss": 0.9055, + "step": 295170 + }, + { + "epoch": 25.55, + "learning_rate": 7.427258805513017e-06, + "loss": 0.9479, + "step": 295180 + }, + { + "epoch": 25.55, + "learning_rate": 7.425814094599671e-06, + "loss": 0.904, + "step": 295190 + }, + { + "epoch": 25.55, + "learning_rate": 7.4243693836863246e-06, + "loss": 0.9714, + "step": 295200 + }, + { + "epoch": 25.55, + "learning_rate": 7.422924672772978e-06, + "loss": 0.9196, + "step": 295210 + }, + { + "epoch": 25.55, + "learning_rate": 7.421479961859633e-06, + "loss": 0.8758, + "step": 295220 + }, + { + "epoch": 25.55, + "learning_rate": 7.4200352509462864e-06, + "loss": 0.8998, + "step": 295230 + }, + { + "epoch": 25.56, + "learning_rate": 7.41859054003294e-06, + "loss": 0.9325, + "step": 295240 + }, + { + "epoch": 25.56, + "learning_rate": 7.417145829119594e-06, + "loss": 0.8887, + "step": 295250 + }, + { + "epoch": 25.56, + "learning_rate": 7.4157011182062475e-06, + "loss": 0.8832, + "step": 295260 + }, + { + "epoch": 25.56, + "learning_rate": 7.414256407292901e-06, + "loss": 0.9121, + "step": 295270 + }, + { + "epoch": 25.56, + "learning_rate": 7.412811696379555e-06, + "loss": 0.9358, + "step": 295280 + }, + { + "epoch": 25.56, + "learning_rate": 7.4113669854662085e-06, + "loss": 0.8941, + "step": 295290 + }, + { + "epoch": 25.56, + "learning_rate": 7.409922274552862e-06, + "loss": 0.9073, + "step": 295300 + }, + { + "epoch": 25.56, + "learning_rate": 7.408477563639515e-06, + "loss": 0.8974, + "step": 295310 + }, + { + "epoch": 25.56, + "learning_rate": 7.40703285272617e-06, + "loss": 0.9161, + "step": 295320 + }, + { + "epoch": 25.56, + "learning_rate": 7.405588141812824e-06, + "loss": 0.9012, + "step": 295330 + }, + { + "epoch": 25.56, + "learning_rate": 7.404143430899476e-06, + "loss": 0.9243, + "step": 295340 + }, + { + "epoch": 25.56, + "learning_rate": 7.402698719986131e-06, + "loss": 0.8522, + "step": 295350 + }, + { + "epoch": 25.57, + "learning_rate": 7.401254009072785e-06, + "loss": 0.9236, + "step": 295360 + }, + { + "epoch": 25.57, + "learning_rate": 7.399809298159438e-06, + "loss": 0.9509, + "step": 295370 + }, + { + "epoch": 25.57, + "learning_rate": 7.3983645872460924e-06, + "loss": 0.9099, + "step": 295380 + }, + { + "epoch": 25.57, + "learning_rate": 7.396919876332746e-06, + "loss": 0.9566, + "step": 295390 + }, + { + "epoch": 25.57, + "learning_rate": 7.395475165419399e-06, + "loss": 0.8943, + "step": 295400 + }, + { + "epoch": 25.57, + "learning_rate": 7.394030454506054e-06, + "loss": 0.8689, + "step": 295410 + }, + { + "epoch": 25.57, + "learning_rate": 7.392585743592708e-06, + "loss": 0.9263, + "step": 295420 + }, + { + "epoch": 25.57, + "learning_rate": 7.39114103267936e-06, + "loss": 0.8861, + "step": 295430 + }, + { + "epoch": 25.57, + "learning_rate": 7.389696321766015e-06, + "loss": 0.9239, + "step": 295440 + }, + { + "epoch": 25.57, + "learning_rate": 7.388251610852669e-06, + "loss": 0.9441, + "step": 295450 + }, + { + "epoch": 25.57, + "learning_rate": 7.386806899939322e-06, + "loss": 0.9126, + "step": 295460 + }, + { + "epoch": 25.58, + "learning_rate": 7.385362189025976e-06, + "loss": 0.9894, + "step": 295470 + }, + { + "epoch": 25.58, + "learning_rate": 7.38391747811263e-06, + "loss": 0.8658, + "step": 295480 + }, + { + "epoch": 25.58, + "learning_rate": 7.382472767199283e-06, + "loss": 0.8475, + "step": 295490 + }, + { + "epoch": 25.58, + "learning_rate": 7.381028056285938e-06, + "loss": 0.9472, + "step": 295500 + }, + { + "epoch": 25.58, + "learning_rate": 7.379583345372592e-06, + "loss": 0.951, + "step": 295510 + }, + { + "epoch": 25.58, + "learning_rate": 7.378138634459244e-06, + "loss": 0.9026, + "step": 295520 + }, + { + "epoch": 25.58, + "learning_rate": 7.376693923545899e-06, + "loss": 0.9504, + "step": 295530 + }, + { + "epoch": 25.58, + "learning_rate": 7.375249212632553e-06, + "loss": 0.9352, + "step": 295540 + }, + { + "epoch": 25.58, + "learning_rate": 7.373804501719206e-06, + "loss": 0.9032, + "step": 295550 + }, + { + "epoch": 25.58, + "learning_rate": 7.37235979080586e-06, + "loss": 0.8669, + "step": 295560 + }, + { + "epoch": 25.58, + "learning_rate": 7.370915079892514e-06, + "loss": 0.8685, + "step": 295570 + }, + { + "epoch": 25.58, + "learning_rate": 7.369470368979167e-06, + "loss": 0.8707, + "step": 295580 + }, + { + "epoch": 25.59, + "learning_rate": 7.368025658065822e-06, + "loss": 0.9, + "step": 295590 + }, + { + "epoch": 25.59, + "learning_rate": 7.366580947152476e-06, + "loss": 0.8903, + "step": 295600 + }, + { + "epoch": 25.59, + "learning_rate": 7.365136236239128e-06, + "loss": 0.8681, + "step": 295610 + }, + { + "epoch": 25.59, + "learning_rate": 7.363691525325783e-06, + "loss": 0.8928, + "step": 295620 + }, + { + "epoch": 25.59, + "learning_rate": 7.362246814412437e-06, + "loss": 0.8852, + "step": 295630 + }, + { + "epoch": 25.59, + "learning_rate": 7.36080210349909e-06, + "loss": 0.8669, + "step": 295640 + }, + { + "epoch": 25.59, + "learning_rate": 7.359357392585744e-06, + "loss": 0.8553, + "step": 295650 + }, + { + "epoch": 25.59, + "learning_rate": 7.357912681672398e-06, + "loss": 0.9541, + "step": 295660 + }, + { + "epoch": 25.59, + "learning_rate": 7.356467970759051e-06, + "loss": 0.8906, + "step": 295670 + }, + { + "epoch": 25.59, + "learning_rate": 7.355023259845706e-06, + "loss": 0.8614, + "step": 295680 + }, + { + "epoch": 25.59, + "learning_rate": 7.3535785489323595e-06, + "loss": 0.9197, + "step": 295690 + }, + { + "epoch": 25.6, + "learning_rate": 7.352133838019012e-06, + "loss": 0.8759, + "step": 295700 + }, + { + "epoch": 25.6, + "learning_rate": 7.350689127105667e-06, + "loss": 0.9049, + "step": 295710 + }, + { + "epoch": 25.6, + "learning_rate": 7.349244416192321e-06, + "loss": 0.8978, + "step": 295720 + }, + { + "epoch": 25.6, + "learning_rate": 7.347799705278974e-06, + "loss": 0.9117, + "step": 295730 + }, + { + "epoch": 25.6, + "learning_rate": 7.346354994365628e-06, + "loss": 0.9127, + "step": 295740 + }, + { + "epoch": 25.6, + "learning_rate": 7.344910283452282e-06, + "loss": 0.8953, + "step": 295750 + }, + { + "epoch": 25.6, + "learning_rate": 7.343465572538935e-06, + "loss": 0.9379, + "step": 295760 + }, + { + "epoch": 25.6, + "learning_rate": 7.342020861625588e-06, + "loss": 0.9258, + "step": 295770 + }, + { + "epoch": 25.6, + "learning_rate": 7.3405761507122435e-06, + "loss": 0.9599, + "step": 295780 + }, + { + "epoch": 25.6, + "learning_rate": 7.339131439798896e-06, + "loss": 0.9046, + "step": 295790 + }, + { + "epoch": 25.6, + "learning_rate": 7.3376867288855494e-06, + "loss": 0.8843, + "step": 295800 + }, + { + "epoch": 25.6, + "learning_rate": 7.3362420179722045e-06, + "loss": 0.9396, + "step": 295810 + }, + { + "epoch": 25.61, + "learning_rate": 7.334797307058858e-06, + "loss": 0.8921, + "step": 295820 + }, + { + "epoch": 25.61, + "learning_rate": 7.333352596145511e-06, + "loss": 0.898, + "step": 295830 + }, + { + "epoch": 25.61, + "learning_rate": 7.3319078852321656e-06, + "loss": 0.9011, + "step": 295840 + }, + { + "epoch": 25.61, + "learning_rate": 7.330463174318819e-06, + "loss": 0.8569, + "step": 295850 + }, + { + "epoch": 25.61, + "learning_rate": 7.329018463405472e-06, + "loss": 0.8799, + "step": 295860 + }, + { + "epoch": 25.61, + "learning_rate": 7.3275737524921274e-06, + "loss": 0.9195, + "step": 295870 + }, + { + "epoch": 25.61, + "learning_rate": 7.32612904157878e-06, + "loss": 0.9183, + "step": 295880 + }, + { + "epoch": 25.61, + "learning_rate": 7.324684330665433e-06, + "loss": 0.8373, + "step": 295890 + }, + { + "epoch": 25.61, + "learning_rate": 7.3232396197520885e-06, + "loss": 0.9327, + "step": 295900 + }, + { + "epoch": 25.61, + "learning_rate": 7.321794908838742e-06, + "loss": 0.9251, + "step": 295910 + }, + { + "epoch": 25.61, + "learning_rate": 7.320350197925395e-06, + "loss": 0.982, + "step": 295920 + }, + { + "epoch": 25.61, + "learning_rate": 7.3189054870120495e-06, + "loss": 0.8958, + "step": 295930 + }, + { + "epoch": 25.62, + "learning_rate": 7.317460776098703e-06, + "loss": 0.9323, + "step": 295940 + }, + { + "epoch": 25.62, + "learning_rate": 7.316016065185356e-06, + "loss": 0.8795, + "step": 295950 + }, + { + "epoch": 25.62, + "learning_rate": 7.314571354272011e-06, + "loss": 0.877, + "step": 295960 + }, + { + "epoch": 25.62, + "learning_rate": 7.313126643358664e-06, + "loss": 0.8835, + "step": 295970 + }, + { + "epoch": 25.62, + "learning_rate": 7.311681932445317e-06, + "loss": 0.9426, + "step": 295980 + }, + { + "epoch": 25.62, + "learning_rate": 7.310237221531972e-06, + "loss": 0.9728, + "step": 295990 + }, + { + "epoch": 25.62, + "learning_rate": 7.308792510618626e-06, + "loss": 0.9153, + "step": 296000 + }, + { + "epoch": 25.62, + "learning_rate": 7.307347799705279e-06, + "loss": 0.8834, + "step": 296010 + }, + { + "epoch": 25.62, + "learning_rate": 7.3059030887919334e-06, + "loss": 0.9149, + "step": 296020 + }, + { + "epoch": 25.62, + "learning_rate": 7.304458377878587e-06, + "loss": 0.9578, + "step": 296030 + }, + { + "epoch": 25.62, + "learning_rate": 7.30301366696524e-06, + "loss": 0.8904, + "step": 296040 + }, + { + "epoch": 25.63, + "learning_rate": 7.3015689560518945e-06, + "loss": 0.9204, + "step": 296050 + }, + { + "epoch": 25.63, + "learning_rate": 7.300124245138548e-06, + "loss": 0.9513, + "step": 296060 + }, + { + "epoch": 25.63, + "learning_rate": 7.298679534225201e-06, + "loss": 0.9242, + "step": 296070 + }, + { + "epoch": 25.63, + "learning_rate": 7.297234823311856e-06, + "loss": 0.9619, + "step": 296080 + }, + { + "epoch": 25.63, + "learning_rate": 7.29579011239851e-06, + "loss": 0.9191, + "step": 296090 + }, + { + "epoch": 25.63, + "learning_rate": 7.294345401485163e-06, + "loss": 0.8962, + "step": 296100 + }, + { + "epoch": 25.63, + "learning_rate": 7.292900690571817e-06, + "loss": 0.9272, + "step": 296110 + }, + { + "epoch": 25.63, + "learning_rate": 7.291455979658471e-06, + "loss": 0.8969, + "step": 296120 + }, + { + "epoch": 25.63, + "learning_rate": 7.290011268745124e-06, + "loss": 0.9085, + "step": 296130 + }, + { + "epoch": 25.63, + "learning_rate": 7.288566557831778e-06, + "loss": 0.84, + "step": 296140 + }, + { + "epoch": 25.63, + "learning_rate": 7.287121846918432e-06, + "loss": 0.896, + "step": 296150 + }, + { + "epoch": 25.63, + "learning_rate": 7.285677136005085e-06, + "loss": 0.893, + "step": 296160 + }, + { + "epoch": 25.64, + "learning_rate": 7.28423242509174e-06, + "loss": 0.9283, + "step": 296170 + }, + { + "epoch": 25.64, + "learning_rate": 7.282787714178394e-06, + "loss": 0.8771, + "step": 296180 + }, + { + "epoch": 25.64, + "learning_rate": 7.281343003265047e-06, + "loss": 0.8887, + "step": 296190 + }, + { + "epoch": 25.64, + "learning_rate": 7.279898292351701e-06, + "loss": 0.8403, + "step": 296200 + }, + { + "epoch": 25.64, + "learning_rate": 7.278453581438355e-06, + "loss": 0.9024, + "step": 296210 + }, + { + "epoch": 25.64, + "learning_rate": 7.277008870525008e-06, + "loss": 0.8941, + "step": 296220 + }, + { + "epoch": 25.64, + "learning_rate": 7.2755641596116615e-06, + "loss": 0.8526, + "step": 296230 + }, + { + "epoch": 25.64, + "learning_rate": 7.274119448698316e-06, + "loss": 0.8688, + "step": 296240 + }, + { + "epoch": 25.64, + "learning_rate": 7.272674737784969e-06, + "loss": 0.9608, + "step": 296250 + }, + { + "epoch": 25.64, + "learning_rate": 7.2712300268716225e-06, + "loss": 0.9096, + "step": 296260 + }, + { + "epoch": 25.64, + "learning_rate": 7.269785315958278e-06, + "loss": 0.8685, + "step": 296270 + }, + { + "epoch": 25.65, + "learning_rate": 7.268340605044931e-06, + "loss": 0.9158, + "step": 296280 + }, + { + "epoch": 25.65, + "learning_rate": 7.266895894131584e-06, + "loss": 0.8518, + "step": 296290 + }, + { + "epoch": 25.65, + "learning_rate": 7.265451183218239e-06, + "loss": 0.8532, + "step": 296300 + }, + { + "epoch": 25.65, + "learning_rate": 7.264006472304892e-06, + "loss": 0.903, + "step": 296310 + }, + { + "epoch": 25.65, + "learning_rate": 7.2625617613915455e-06, + "loss": 0.8962, + "step": 296320 + }, + { + "epoch": 25.65, + "learning_rate": 7.2611170504782e-06, + "loss": 0.9799, + "step": 296330 + }, + { + "epoch": 25.65, + "learning_rate": 7.259672339564853e-06, + "loss": 0.9034, + "step": 296340 + }, + { + "epoch": 25.65, + "learning_rate": 7.2582276286515065e-06, + "loss": 0.9204, + "step": 296350 + }, + { + "epoch": 25.65, + "learning_rate": 7.2567829177381616e-06, + "loss": 0.9398, + "step": 296360 + }, + { + "epoch": 25.65, + "learning_rate": 7.255338206824815e-06, + "loss": 0.9221, + "step": 296370 + }, + { + "epoch": 25.65, + "learning_rate": 7.2538934959114675e-06, + "loss": 0.8659, + "step": 296380 + }, + { + "epoch": 25.65, + "learning_rate": 7.252448784998123e-06, + "loss": 0.8814, + "step": 296390 + }, + { + "epoch": 25.66, + "learning_rate": 7.251004074084776e-06, + "loss": 0.9219, + "step": 296400 + }, + { + "epoch": 25.66, + "learning_rate": 7.249559363171429e-06, + "loss": 0.9751, + "step": 296410 + }, + { + "epoch": 25.66, + "learning_rate": 7.248114652258084e-06, + "loss": 0.9407, + "step": 296420 + }, + { + "epoch": 25.66, + "learning_rate": 7.246669941344737e-06, + "loss": 0.8626, + "step": 296430 + }, + { + "epoch": 25.66, + "learning_rate": 7.2452252304313904e-06, + "loss": 0.9428, + "step": 296440 + }, + { + "epoch": 25.66, + "learning_rate": 7.2437805195180455e-06, + "loss": 0.9408, + "step": 296450 + }, + { + "epoch": 25.66, + "learning_rate": 7.242335808604699e-06, + "loss": 0.8986, + "step": 296460 + }, + { + "epoch": 25.66, + "learning_rate": 7.2408910976913515e-06, + "loss": 0.9616, + "step": 296470 + }, + { + "epoch": 25.66, + "learning_rate": 7.2394463867780065e-06, + "loss": 0.9413, + "step": 296480 + }, + { + "epoch": 25.66, + "learning_rate": 7.23800167586466e-06, + "loss": 0.8719, + "step": 296490 + }, + { + "epoch": 25.66, + "learning_rate": 7.236556964951313e-06, + "loss": 0.9057, + "step": 296500 + }, + { + "epoch": 25.67, + "learning_rate": 7.235112254037968e-06, + "loss": 0.9263, + "step": 296510 + }, + { + "epoch": 25.67, + "learning_rate": 7.233667543124621e-06, + "loss": 0.9189, + "step": 296520 + }, + { + "epoch": 25.67, + "learning_rate": 7.232222832211274e-06, + "loss": 0.9255, + "step": 296530 + }, + { + "epoch": 25.67, + "learning_rate": 7.2307781212979295e-06, + "loss": 0.922, + "step": 296540 + }, + { + "epoch": 25.67, + "learning_rate": 7.229333410384583e-06, + "loss": 0.8814, + "step": 296550 + }, + { + "epoch": 25.67, + "learning_rate": 7.227888699471235e-06, + "loss": 0.8604, + "step": 296560 + }, + { + "epoch": 25.67, + "learning_rate": 7.2264439885578905e-06, + "loss": 0.9323, + "step": 296570 + }, + { + "epoch": 25.67, + "learning_rate": 7.224999277644544e-06, + "loss": 0.9223, + "step": 296580 + }, + { + "epoch": 25.67, + "learning_rate": 7.223554566731197e-06, + "loss": 0.8629, + "step": 296590 + }, + { + "epoch": 25.67, + "learning_rate": 7.2221098558178515e-06, + "loss": 0.9063, + "step": 296600 + }, + { + "epoch": 25.67, + "learning_rate": 7.220665144904505e-06, + "loss": 0.9204, + "step": 296610 + }, + { + "epoch": 25.67, + "learning_rate": 7.219220433991158e-06, + "loss": 0.9193, + "step": 296620 + }, + { + "epoch": 25.68, + "learning_rate": 7.217775723077813e-06, + "loss": 0.9199, + "step": 296630 + }, + { + "epoch": 25.68, + "learning_rate": 7.216331012164467e-06, + "loss": 0.8726, + "step": 296640 + }, + { + "epoch": 25.68, + "learning_rate": 7.214886301251119e-06, + "loss": 0.8829, + "step": 296650 + }, + { + "epoch": 25.68, + "learning_rate": 7.2134415903377744e-06, + "loss": 0.9295, + "step": 296660 + }, + { + "epoch": 25.68, + "learning_rate": 7.211996879424428e-06, + "loss": 0.9173, + "step": 296670 + }, + { + "epoch": 25.68, + "learning_rate": 7.210552168511081e-06, + "loss": 0.9007, + "step": 296680 + }, + { + "epoch": 25.68, + "learning_rate": 7.209107457597735e-06, + "loss": 0.9321, + "step": 296690 + }, + { + "epoch": 25.68, + "learning_rate": 7.207662746684389e-06, + "loss": 0.939, + "step": 296700 + }, + { + "epoch": 25.68, + "learning_rate": 7.206218035771042e-06, + "loss": 0.9026, + "step": 296710 + }, + { + "epoch": 25.68, + "learning_rate": 7.204773324857696e-06, + "loss": 0.8587, + "step": 296720 + }, + { + "epoch": 25.68, + "learning_rate": 7.203328613944351e-06, + "loss": 0.934, + "step": 296730 + }, + { + "epoch": 25.69, + "learning_rate": 7.201883903031003e-06, + "loss": 0.9257, + "step": 296740 + }, + { + "epoch": 25.69, + "learning_rate": 7.200439192117657e-06, + "loss": 0.8944, + "step": 296750 + }, + { + "epoch": 25.69, + "learning_rate": 7.198994481204312e-06, + "loss": 0.9327, + "step": 296760 + }, + { + "epoch": 25.69, + "learning_rate": 7.197549770290965e-06, + "loss": 0.9251, + "step": 296770 + }, + { + "epoch": 25.69, + "learning_rate": 7.1961050593776186e-06, + "loss": 0.9221, + "step": 296780 + }, + { + "epoch": 25.69, + "learning_rate": 7.194660348464273e-06, + "loss": 0.8638, + "step": 296790 + }, + { + "epoch": 25.69, + "learning_rate": 7.193215637550926e-06, + "loss": 0.9313, + "step": 296800 + }, + { + "epoch": 25.69, + "learning_rate": 7.19177092663758e-06, + "loss": 0.8778, + "step": 296810 + }, + { + "epoch": 25.69, + "learning_rate": 7.190326215724235e-06, + "loss": 0.9684, + "step": 296820 + }, + { + "epoch": 25.69, + "learning_rate": 7.188881504810887e-06, + "loss": 0.996, + "step": 296830 + }, + { + "epoch": 25.69, + "learning_rate": 7.187436793897541e-06, + "loss": 0.9209, + "step": 296840 + }, + { + "epoch": 25.69, + "learning_rate": 7.185992082984196e-06, + "loss": 0.9327, + "step": 296850 + }, + { + "epoch": 25.7, + "learning_rate": 7.184547372070849e-06, + "loss": 0.8131, + "step": 296860 + }, + { + "epoch": 25.7, + "learning_rate": 7.1831026611575025e-06, + "loss": 0.9298, + "step": 296870 + }, + { + "epoch": 25.7, + "learning_rate": 7.181657950244157e-06, + "loss": 0.9282, + "step": 296880 + }, + { + "epoch": 25.7, + "learning_rate": 7.18021323933081e-06, + "loss": 0.9132, + "step": 296890 + }, + { + "epoch": 25.7, + "learning_rate": 7.1787685284174635e-06, + "loss": 0.9271, + "step": 296900 + }, + { + "epoch": 25.7, + "learning_rate": 7.177323817504119e-06, + "loss": 0.8698, + "step": 296910 + }, + { + "epoch": 25.7, + "learning_rate": 7.175879106590771e-06, + "loss": 0.905, + "step": 296920 + }, + { + "epoch": 25.7, + "learning_rate": 7.1744343956774246e-06, + "loss": 0.8423, + "step": 296930 + }, + { + "epoch": 25.7, + "learning_rate": 7.17298968476408e-06, + "loss": 0.9083, + "step": 296940 + }, + { + "epoch": 25.7, + "learning_rate": 7.171544973850733e-06, + "loss": 0.9206, + "step": 296950 + }, + { + "epoch": 25.7, + "learning_rate": 7.1701002629373864e-06, + "loss": 0.9197, + "step": 296960 + }, + { + "epoch": 25.7, + "learning_rate": 7.168655552024041e-06, + "loss": 0.8579, + "step": 296970 + }, + { + "epoch": 25.71, + "learning_rate": 7.167210841110694e-06, + "loss": 0.8713, + "step": 296980 + }, + { + "epoch": 25.71, + "learning_rate": 7.1657661301973475e-06, + "loss": 0.931, + "step": 296990 + }, + { + "epoch": 25.71, + "learning_rate": 7.1643214192840026e-06, + "loss": 0.9448, + "step": 297000 + }, + { + "epoch": 25.71, + "learning_rate": 7.162876708370655e-06, + "loss": 0.9635, + "step": 297010 + }, + { + "epoch": 25.71, + "learning_rate": 7.1614319974573085e-06, + "loss": 0.8405, + "step": 297020 + }, + { + "epoch": 25.71, + "learning_rate": 7.159987286543964e-06, + "loss": 0.8795, + "step": 297030 + }, + { + "epoch": 25.71, + "learning_rate": 7.158542575630617e-06, + "loss": 0.9671, + "step": 297040 + }, + { + "epoch": 25.71, + "learning_rate": 7.15709786471727e-06, + "loss": 0.8912, + "step": 297050 + }, + { + "epoch": 25.71, + "learning_rate": 7.155653153803925e-06, + "loss": 0.8783, + "step": 297060 + }, + { + "epoch": 25.71, + "learning_rate": 7.154208442890578e-06, + "loss": 0.924, + "step": 297070 + }, + { + "epoch": 25.71, + "learning_rate": 7.152763731977231e-06, + "loss": 0.8758, + "step": 297080 + }, + { + "epoch": 25.72, + "learning_rate": 7.1513190210638865e-06, + "loss": 0.9569, + "step": 297090 + }, + { + "epoch": 25.72, + "learning_rate": 7.149874310150539e-06, + "loss": 0.9268, + "step": 297100 + }, + { + "epoch": 25.72, + "learning_rate": 7.1484295992371925e-06, + "loss": 0.8919, + "step": 297110 + }, + { + "epoch": 25.72, + "learning_rate": 7.146984888323846e-06, + "loss": 0.933, + "step": 297120 + }, + { + "epoch": 25.72, + "learning_rate": 7.145540177410501e-06, + "loss": 0.9255, + "step": 297130 + }, + { + "epoch": 25.72, + "learning_rate": 7.144095466497154e-06, + "loss": 0.9026, + "step": 297140 + }, + { + "epoch": 25.72, + "learning_rate": 7.142650755583808e-06, + "loss": 0.9472, + "step": 297150 + }, + { + "epoch": 25.72, + "learning_rate": 7.141206044670462e-06, + "loss": 0.8557, + "step": 297160 + }, + { + "epoch": 25.72, + "learning_rate": 7.139761333757115e-06, + "loss": 0.8932, + "step": 297170 + }, + { + "epoch": 25.72, + "learning_rate": 7.138316622843769e-06, + "loss": 0.9524, + "step": 297180 + }, + { + "epoch": 25.72, + "learning_rate": 7.136871911930423e-06, + "loss": 0.9187, + "step": 297190 + }, + { + "epoch": 25.72, + "learning_rate": 7.135427201017076e-06, + "loss": 0.9408, + "step": 297200 + }, + { + "epoch": 25.73, + "learning_rate": 7.13398249010373e-06, + "loss": 0.9259, + "step": 297210 + }, + { + "epoch": 25.73, + "learning_rate": 7.132537779190385e-06, + "loss": 0.9146, + "step": 297220 + }, + { + "epoch": 25.73, + "learning_rate": 7.131093068277038e-06, + "loss": 0.9163, + "step": 297230 + }, + { + "epoch": 25.73, + "learning_rate": 7.129648357363692e-06, + "loss": 0.9543, + "step": 297240 + }, + { + "epoch": 25.73, + "learning_rate": 7.128203646450346e-06, + "loss": 0.8231, + "step": 297250 + }, + { + "epoch": 25.73, + "learning_rate": 7.126758935536999e-06, + "loss": 0.9169, + "step": 297260 + }, + { + "epoch": 25.73, + "learning_rate": 7.125314224623653e-06, + "loss": 0.9321, + "step": 297270 + }, + { + "epoch": 25.73, + "learning_rate": 7.123869513710307e-06, + "loss": 0.8727, + "step": 297280 + }, + { + "epoch": 25.73, + "learning_rate": 7.12242480279696e-06, + "loss": 0.8811, + "step": 297290 + }, + { + "epoch": 25.73, + "learning_rate": 7.120980091883614e-06, + "loss": 0.946, + "step": 297300 + }, + { + "epoch": 25.73, + "learning_rate": 7.119535380970269e-06, + "loss": 0.9233, + "step": 297310 + }, + { + "epoch": 25.74, + "learning_rate": 7.118090670056922e-06, + "loss": 0.9136, + "step": 297320 + }, + { + "epoch": 25.74, + "learning_rate": 7.116645959143576e-06, + "loss": 0.9281, + "step": 297330 + }, + { + "epoch": 25.74, + "learning_rate": 7.11520124823023e-06, + "loss": 0.9186, + "step": 297340 + }, + { + "epoch": 25.74, + "learning_rate": 7.113756537316883e-06, + "loss": 0.897, + "step": 297350 + }, + { + "epoch": 25.74, + "learning_rate": 7.112311826403537e-06, + "loss": 0.9656, + "step": 297360 + }, + { + "epoch": 25.74, + "learning_rate": 7.110867115490191e-06, + "loss": 0.9343, + "step": 297370 + }, + { + "epoch": 25.74, + "learning_rate": 7.109422404576844e-06, + "loss": 0.8812, + "step": 297380 + }, + { + "epoch": 25.74, + "learning_rate": 7.107977693663498e-06, + "loss": 0.8581, + "step": 297390 + }, + { + "epoch": 25.74, + "learning_rate": 7.106532982750153e-06, + "loss": 0.934, + "step": 297400 + }, + { + "epoch": 25.74, + "learning_rate": 7.105088271836806e-06, + "loss": 0.8621, + "step": 297410 + }, + { + "epoch": 25.74, + "learning_rate": 7.1036435609234596e-06, + "loss": 0.8783, + "step": 297420 + }, + { + "epoch": 25.74, + "learning_rate": 7.102198850010114e-06, + "loss": 0.956, + "step": 297430 + }, + { + "epoch": 25.75, + "learning_rate": 7.100754139096767e-06, + "loss": 0.9273, + "step": 297440 + }, + { + "epoch": 25.75, + "learning_rate": 7.099309428183421e-06, + "loss": 0.9608, + "step": 297450 + }, + { + "epoch": 25.75, + "learning_rate": 7.097864717270075e-06, + "loss": 0.9017, + "step": 297460 + }, + { + "epoch": 25.75, + "learning_rate": 7.096420006356728e-06, + "loss": 0.8297, + "step": 297470 + }, + { + "epoch": 25.75, + "learning_rate": 7.094975295443382e-06, + "loss": 0.889, + "step": 297480 + }, + { + "epoch": 25.75, + "learning_rate": 7.093530584530037e-06, + "loss": 0.8954, + "step": 297490 + }, + { + "epoch": 25.75, + "learning_rate": 7.09208587361669e-06, + "loss": 0.9041, + "step": 297500 + }, + { + "epoch": 25.75, + "learning_rate": 7.0906411627033435e-06, + "loss": 0.8997, + "step": 297510 + }, + { + "epoch": 25.75, + "learning_rate": 7.089196451789998e-06, + "loss": 0.8962, + "step": 297520 + }, + { + "epoch": 25.75, + "learning_rate": 7.087751740876651e-06, + "loss": 0.9389, + "step": 297530 + }, + { + "epoch": 25.75, + "learning_rate": 7.0863070299633045e-06, + "loss": 0.9013, + "step": 297540 + }, + { + "epoch": 25.76, + "learning_rate": 7.084862319049959e-06, + "loss": 0.8635, + "step": 297550 + }, + { + "epoch": 25.76, + "learning_rate": 7.083417608136612e-06, + "loss": 0.9191, + "step": 297560 + }, + { + "epoch": 25.76, + "learning_rate": 7.0819728972232656e-06, + "loss": 0.8897, + "step": 297570 + }, + { + "epoch": 25.76, + "learning_rate": 7.080528186309919e-06, + "loss": 0.8371, + "step": 297580 + }, + { + "epoch": 25.76, + "learning_rate": 7.079083475396574e-06, + "loss": 0.9698, + "step": 297590 + }, + { + "epoch": 25.76, + "learning_rate": 7.0776387644832274e-06, + "loss": 0.9207, + "step": 297600 + }, + { + "epoch": 25.76, + "learning_rate": 7.07619405356988e-06, + "loss": 0.927, + "step": 297610 + }, + { + "epoch": 25.76, + "learning_rate": 7.074749342656535e-06, + "loss": 0.9165, + "step": 297620 + }, + { + "epoch": 25.76, + "learning_rate": 7.0733046317431885e-06, + "loss": 0.8877, + "step": 297630 + }, + { + "epoch": 25.76, + "learning_rate": 7.071859920829842e-06, + "loss": 0.9435, + "step": 297640 + }, + { + "epoch": 25.76, + "learning_rate": 7.070415209916496e-06, + "loss": 0.9385, + "step": 297650 + }, + { + "epoch": 25.76, + "learning_rate": 7.0689704990031495e-06, + "loss": 0.962, + "step": 297660 + }, + { + "epoch": 25.77, + "learning_rate": 7.067525788089803e-06, + "loss": 0.9281, + "step": 297670 + }, + { + "epoch": 25.77, + "learning_rate": 7.066081077176458e-06, + "loss": 0.9299, + "step": 297680 + }, + { + "epoch": 25.77, + "learning_rate": 7.064636366263111e-06, + "loss": 0.9495, + "step": 297690 + }, + { + "epoch": 25.77, + "learning_rate": 7.063191655349764e-06, + "loss": 0.9262, + "step": 297700 + }, + { + "epoch": 25.77, + "learning_rate": 7.061746944436419e-06, + "loss": 0.9305, + "step": 297710 + }, + { + "epoch": 25.77, + "learning_rate": 7.060302233523072e-06, + "loss": 0.9478, + "step": 297720 + }, + { + "epoch": 25.77, + "learning_rate": 7.058857522609726e-06, + "loss": 0.9483, + "step": 297730 + }, + { + "epoch": 25.77, + "learning_rate": 7.05741281169638e-06, + "loss": 0.9142, + "step": 297740 + }, + { + "epoch": 25.77, + "learning_rate": 7.0559681007830334e-06, + "loss": 0.8454, + "step": 297750 + }, + { + "epoch": 25.77, + "learning_rate": 7.054523389869687e-06, + "loss": 0.9499, + "step": 297760 + }, + { + "epoch": 25.77, + "learning_rate": 7.053078678956342e-06, + "loss": 0.8725, + "step": 297770 + }, + { + "epoch": 25.78, + "learning_rate": 7.051633968042995e-06, + "loss": 0.9205, + "step": 297780 + }, + { + "epoch": 25.78, + "learning_rate": 7.050189257129648e-06, + "loss": 0.9307, + "step": 297790 + }, + { + "epoch": 25.78, + "learning_rate": 7.048744546216303e-06, + "loss": 0.9008, + "step": 297800 + }, + { + "epoch": 25.78, + "learning_rate": 7.047299835302956e-06, + "loss": 0.8843, + "step": 297810 + }, + { + "epoch": 25.78, + "learning_rate": 7.04585512438961e-06, + "loss": 0.9278, + "step": 297820 + }, + { + "epoch": 25.78, + "learning_rate": 7.044410413476264e-06, + "loss": 0.8948, + "step": 297830 + }, + { + "epoch": 25.78, + "learning_rate": 7.042965702562917e-06, + "loss": 0.8901, + "step": 297840 + }, + { + "epoch": 25.78, + "learning_rate": 7.041520991649571e-06, + "loss": 0.9051, + "step": 297850 + }, + { + "epoch": 25.78, + "learning_rate": 7.040076280736226e-06, + "loss": 0.8611, + "step": 297860 + }, + { + "epoch": 25.78, + "learning_rate": 7.038631569822879e-06, + "loss": 0.8851, + "step": 297870 + }, + { + "epoch": 25.78, + "learning_rate": 7.037186858909532e-06, + "loss": 0.927, + "step": 297880 + }, + { + "epoch": 25.78, + "learning_rate": 7.035742147996187e-06, + "loss": 0.8576, + "step": 297890 + }, + { + "epoch": 25.79, + "learning_rate": 7.03429743708284e-06, + "loss": 0.9521, + "step": 297900 + }, + { + "epoch": 25.79, + "learning_rate": 7.032852726169494e-06, + "loss": 0.9183, + "step": 297910 + }, + { + "epoch": 25.79, + "learning_rate": 7.031408015256148e-06, + "loss": 0.8999, + "step": 297920 + }, + { + "epoch": 25.79, + "learning_rate": 7.029963304342801e-06, + "loss": 0.9438, + "step": 297930 + }, + { + "epoch": 25.79, + "learning_rate": 7.028518593429455e-06, + "loss": 0.9528, + "step": 297940 + }, + { + "epoch": 25.79, + "learning_rate": 7.02707388251611e-06, + "loss": 0.9527, + "step": 297950 + }, + { + "epoch": 25.79, + "learning_rate": 7.025629171602763e-06, + "loss": 0.8508, + "step": 297960 + }, + { + "epoch": 25.79, + "learning_rate": 7.024184460689416e-06, + "loss": 0.8699, + "step": 297970 + }, + { + "epoch": 25.79, + "learning_rate": 7.022739749776071e-06, + "loss": 0.913, + "step": 297980 + }, + { + "epoch": 25.79, + "learning_rate": 7.021295038862724e-06, + "loss": 0.9686, + "step": 297990 + }, + { + "epoch": 25.79, + "learning_rate": 7.019850327949378e-06, + "loss": 0.8625, + "step": 298000 + }, + { + "epoch": 25.79, + "learning_rate": 7.018405617036032e-06, + "loss": 0.9577, + "step": 298010 + }, + { + "epoch": 25.8, + "learning_rate": 7.016960906122685e-06, + "loss": 0.9076, + "step": 298020 + }, + { + "epoch": 25.8, + "learning_rate": 7.015516195209339e-06, + "loss": 0.9138, + "step": 298030 + }, + { + "epoch": 25.8, + "learning_rate": 7.014071484295992e-06, + "loss": 0.9354, + "step": 298040 + }, + { + "epoch": 25.8, + "learning_rate": 7.012626773382646e-06, + "loss": 0.9015, + "step": 298050 + }, + { + "epoch": 25.8, + "learning_rate": 7.0111820624693e-06, + "loss": 0.9399, + "step": 298060 + }, + { + "epoch": 25.8, + "learning_rate": 7.009737351555953e-06, + "loss": 0.9238, + "step": 298070 + }, + { + "epoch": 25.8, + "learning_rate": 7.008292640642608e-06, + "loss": 0.938, + "step": 298080 + }, + { + "epoch": 25.8, + "learning_rate": 7.006847929729262e-06, + "loss": 0.9214, + "step": 298090 + }, + { + "epoch": 25.8, + "learning_rate": 7.005403218815915e-06, + "loss": 0.866, + "step": 298100 + }, + { + "epoch": 25.8, + "learning_rate": 7.003958507902569e-06, + "loss": 0.9264, + "step": 298110 + }, + { + "epoch": 25.8, + "learning_rate": 7.002513796989223e-06, + "loss": 0.8503, + "step": 298120 + }, + { + "epoch": 25.81, + "learning_rate": 7.001069086075876e-06, + "loss": 0.9417, + "step": 298130 + }, + { + "epoch": 25.81, + "learning_rate": 6.99962437516253e-06, + "loss": 0.9454, + "step": 298140 + }, + { + "epoch": 25.81, + "learning_rate": 6.998179664249184e-06, + "loss": 0.8731, + "step": 298150 + }, + { + "epoch": 25.81, + "learning_rate": 6.996734953335837e-06, + "loss": 0.8969, + "step": 298160 + }, + { + "epoch": 25.81, + "learning_rate": 6.995290242422492e-06, + "loss": 0.9205, + "step": 298170 + }, + { + "epoch": 25.81, + "learning_rate": 6.9938455315091455e-06, + "loss": 0.9041, + "step": 298180 + }, + { + "epoch": 25.81, + "learning_rate": 6.992400820595799e-06, + "loss": 0.9565, + "step": 298190 + }, + { + "epoch": 25.81, + "learning_rate": 6.990956109682453e-06, + "loss": 0.9507, + "step": 298200 + }, + { + "epoch": 25.81, + "learning_rate": 6.9895113987691066e-06, + "loss": 0.9029, + "step": 298210 + }, + { + "epoch": 25.81, + "learning_rate": 6.98806668785576e-06, + "loss": 0.9057, + "step": 298220 + }, + { + "epoch": 25.81, + "learning_rate": 6.986621976942414e-06, + "loss": 0.9408, + "step": 298230 + }, + { + "epoch": 25.81, + "learning_rate": 6.985177266029068e-06, + "loss": 0.9001, + "step": 298240 + }, + { + "epoch": 25.82, + "learning_rate": 6.983732555115721e-06, + "loss": 0.8949, + "step": 298250 + }, + { + "epoch": 25.82, + "learning_rate": 6.982287844202376e-06, + "loss": 0.8404, + "step": 298260 + }, + { + "epoch": 25.82, + "learning_rate": 6.9808431332890295e-06, + "loss": 0.9676, + "step": 298270 + }, + { + "epoch": 25.82, + "learning_rate": 6.979398422375683e-06, + "loss": 0.8663, + "step": 298280 + }, + { + "epoch": 25.82, + "learning_rate": 6.977953711462337e-06, + "loss": 0.8772, + "step": 298290 + }, + { + "epoch": 25.82, + "learning_rate": 6.9765090005489905e-06, + "loss": 0.9241, + "step": 298300 + }, + { + "epoch": 25.82, + "learning_rate": 6.975064289635644e-06, + "loss": 0.8856, + "step": 298310 + }, + { + "epoch": 25.82, + "learning_rate": 6.973619578722298e-06, + "loss": 0.9007, + "step": 298320 + }, + { + "epoch": 25.82, + "learning_rate": 6.9721748678089515e-06, + "loss": 0.9476, + "step": 298330 + }, + { + "epoch": 25.82, + "learning_rate": 6.970730156895605e-06, + "loss": 0.8933, + "step": 298340 + }, + { + "epoch": 25.82, + "learning_rate": 6.96928544598226e-06, + "loss": 0.909, + "step": 298350 + }, + { + "epoch": 25.83, + "learning_rate": 6.967840735068913e-06, + "loss": 0.9379, + "step": 298360 + }, + { + "epoch": 25.83, + "learning_rate": 6.966396024155567e-06, + "loss": 0.9063, + "step": 298370 + }, + { + "epoch": 25.83, + "learning_rate": 6.964951313242221e-06, + "loss": 0.8686, + "step": 298380 + }, + { + "epoch": 25.83, + "learning_rate": 6.9635066023288744e-06, + "loss": 0.9107, + "step": 298390 + }, + { + "epoch": 25.83, + "learning_rate": 6.962061891415528e-06, + "loss": 0.9444, + "step": 298400 + }, + { + "epoch": 25.83, + "learning_rate": 6.960617180502182e-06, + "loss": 0.8949, + "step": 298410 + }, + { + "epoch": 25.83, + "learning_rate": 6.9591724695888355e-06, + "loss": 0.8261, + "step": 298420 + }, + { + "epoch": 25.83, + "learning_rate": 6.957727758675489e-06, + "loss": 0.9168, + "step": 298430 + }, + { + "epoch": 25.83, + "learning_rate": 6.956283047762144e-06, + "loss": 0.9439, + "step": 298440 + }, + { + "epoch": 25.83, + "learning_rate": 6.954838336848797e-06, + "loss": 0.8822, + "step": 298450 + }, + { + "epoch": 25.83, + "learning_rate": 6.953393625935451e-06, + "loss": 0.9059, + "step": 298460 + }, + { + "epoch": 25.83, + "learning_rate": 6.951948915022105e-06, + "loss": 0.9265, + "step": 298470 + }, + { + "epoch": 25.84, + "learning_rate": 6.950504204108758e-06, + "loss": 0.9268, + "step": 298480 + }, + { + "epoch": 25.84, + "learning_rate": 6.949059493195412e-06, + "loss": 0.8935, + "step": 298490 + }, + { + "epoch": 25.84, + "learning_rate": 6.947614782282065e-06, + "loss": 0.8697, + "step": 298500 + }, + { + "epoch": 25.84, + "learning_rate": 6.946170071368719e-06, + "loss": 0.9346, + "step": 298510 + }, + { + "epoch": 25.84, + "learning_rate": 6.944725360455373e-06, + "loss": 0.9161, + "step": 298520 + }, + { + "epoch": 25.84, + "learning_rate": 6.943280649542026e-06, + "loss": 0.9219, + "step": 298530 + }, + { + "epoch": 25.84, + "learning_rate": 6.941835938628681e-06, + "loss": 0.9348, + "step": 298540 + }, + { + "epoch": 25.84, + "learning_rate": 6.940391227715335e-06, + "loss": 0.8957, + "step": 298550 + }, + { + "epoch": 25.84, + "learning_rate": 6.938946516801987e-06, + "loss": 0.8982, + "step": 298560 + }, + { + "epoch": 25.84, + "learning_rate": 6.937501805888642e-06, + "loss": 0.9342, + "step": 298570 + }, + { + "epoch": 25.84, + "learning_rate": 6.936057094975296e-06, + "loss": 0.9361, + "step": 298580 + }, + { + "epoch": 25.85, + "learning_rate": 6.934612384061949e-06, + "loss": 0.9354, + "step": 298590 + }, + { + "epoch": 25.85, + "learning_rate": 6.933167673148603e-06, + "loss": 0.9442, + "step": 298600 + }, + { + "epoch": 25.85, + "learning_rate": 6.931722962235257e-06, + "loss": 0.9124, + "step": 298610 + }, + { + "epoch": 25.85, + "learning_rate": 6.93027825132191e-06, + "loss": 0.9172, + "step": 298620 + }, + { + "epoch": 25.85, + "learning_rate": 6.928833540408565e-06, + "loss": 0.8631, + "step": 298630 + }, + { + "epoch": 25.85, + "learning_rate": 6.927388829495219e-06, + "loss": 0.9489, + "step": 298640 + }, + { + "epoch": 25.85, + "learning_rate": 6.925944118581871e-06, + "loss": 0.8761, + "step": 298650 + }, + { + "epoch": 25.85, + "learning_rate": 6.924499407668526e-06, + "loss": 0.8975, + "step": 298660 + }, + { + "epoch": 25.85, + "learning_rate": 6.92305469675518e-06, + "loss": 0.9389, + "step": 298670 + }, + { + "epoch": 25.85, + "learning_rate": 6.921609985841833e-06, + "loss": 0.8246, + "step": 298680 + }, + { + "epoch": 25.85, + "learning_rate": 6.920165274928487e-06, + "loss": 0.9224, + "step": 298690 + }, + { + "epoch": 25.85, + "learning_rate": 6.918720564015141e-06, + "loss": 0.9304, + "step": 298700 + }, + { + "epoch": 25.86, + "learning_rate": 6.917275853101794e-06, + "loss": 0.8673, + "step": 298710 + }, + { + "epoch": 25.86, + "learning_rate": 6.915831142188449e-06, + "loss": 0.9423, + "step": 298720 + }, + { + "epoch": 25.86, + "learning_rate": 6.914386431275103e-06, + "loss": 0.8911, + "step": 298730 + }, + { + "epoch": 25.86, + "learning_rate": 6.912941720361755e-06, + "loss": 0.8561, + "step": 298740 + }, + { + "epoch": 25.86, + "learning_rate": 6.91149700944841e-06, + "loss": 0.8305, + "step": 298750 + }, + { + "epoch": 25.86, + "learning_rate": 6.910052298535064e-06, + "loss": 0.9129, + "step": 298760 + }, + { + "epoch": 25.86, + "learning_rate": 6.908607587621717e-06, + "loss": 0.9027, + "step": 298770 + }, + { + "epoch": 25.86, + "learning_rate": 6.907162876708371e-06, + "loss": 0.8287, + "step": 298780 + }, + { + "epoch": 25.86, + "learning_rate": 6.905718165795025e-06, + "loss": 0.8749, + "step": 298790 + }, + { + "epoch": 25.86, + "learning_rate": 6.904273454881678e-06, + "loss": 0.9197, + "step": 298800 + }, + { + "epoch": 25.86, + "learning_rate": 6.902828743968333e-06, + "loss": 0.8996, + "step": 298810 + }, + { + "epoch": 25.87, + "learning_rate": 6.9013840330549865e-06, + "loss": 0.9144, + "step": 298820 + }, + { + "epoch": 25.87, + "learning_rate": 6.899939322141639e-06, + "loss": 0.8907, + "step": 298830 + }, + { + "epoch": 25.87, + "learning_rate": 6.898494611228294e-06, + "loss": 0.9453, + "step": 298840 + }, + { + "epoch": 25.87, + "learning_rate": 6.8970499003149476e-06, + "loss": 0.8833, + "step": 298850 + }, + { + "epoch": 25.87, + "learning_rate": 6.895605189401601e-06, + "loss": 0.9083, + "step": 298860 + }, + { + "epoch": 25.87, + "learning_rate": 6.894160478488255e-06, + "loss": 0.9098, + "step": 298870 + }, + { + "epoch": 25.87, + "learning_rate": 6.892715767574909e-06, + "loss": 0.8571, + "step": 298880 + }, + { + "epoch": 25.87, + "learning_rate": 6.891271056661562e-06, + "loss": 0.8896, + "step": 298890 + }, + { + "epoch": 25.87, + "learning_rate": 6.889826345748217e-06, + "loss": 0.9057, + "step": 298900 + }, + { + "epoch": 25.87, + "learning_rate": 6.8883816348348705e-06, + "loss": 0.8605, + "step": 298910 + }, + { + "epoch": 25.87, + "learning_rate": 6.886936923921523e-06, + "loss": 0.9161, + "step": 298920 + }, + { + "epoch": 25.87, + "learning_rate": 6.885492213008176e-06, + "loss": 0.8774, + "step": 298930 + }, + { + "epoch": 25.88, + "learning_rate": 6.8840475020948315e-06, + "loss": 0.9508, + "step": 298940 + }, + { + "epoch": 25.88, + "learning_rate": 6.882602791181485e-06, + "loss": 0.8692, + "step": 298950 + }, + { + "epoch": 25.88, + "learning_rate": 6.881158080268138e-06, + "loss": 0.9638, + "step": 298960 + }, + { + "epoch": 25.88, + "learning_rate": 6.8797133693547925e-06, + "loss": 0.8393, + "step": 298970 + }, + { + "epoch": 25.88, + "learning_rate": 6.878268658441446e-06, + "loss": 0.9282, + "step": 298980 + }, + { + "epoch": 25.88, + "learning_rate": 6.876823947528099e-06, + "loss": 0.9416, + "step": 298990 + }, + { + "epoch": 25.88, + "learning_rate": 6.875379236614754e-06, + "loss": 0.9265, + "step": 299000 + }, + { + "epoch": 25.88, + "learning_rate": 6.873934525701407e-06, + "loss": 0.8742, + "step": 299010 + }, + { + "epoch": 25.88, + "learning_rate": 6.87248981478806e-06, + "loss": 0.9001, + "step": 299020 + }, + { + "epoch": 25.88, + "learning_rate": 6.8710451038747154e-06, + "loss": 0.9461, + "step": 299030 + }, + { + "epoch": 25.88, + "learning_rate": 6.869600392961369e-06, + "loss": 0.8989, + "step": 299040 + }, + { + "epoch": 25.88, + "learning_rate": 6.868155682048022e-06, + "loss": 0.9496, + "step": 299050 + }, + { + "epoch": 25.89, + "learning_rate": 6.8667109711346765e-06, + "loss": 0.9093, + "step": 299060 + }, + { + "epoch": 25.89, + "learning_rate": 6.86526626022133e-06, + "loss": 0.9438, + "step": 299070 + }, + { + "epoch": 25.89, + "learning_rate": 6.863821549307983e-06, + "loss": 0.9073, + "step": 299080 + }, + { + "epoch": 25.89, + "learning_rate": 6.862376838394638e-06, + "loss": 0.9027, + "step": 299090 + }, + { + "epoch": 25.89, + "learning_rate": 6.860932127481291e-06, + "loss": 0.9494, + "step": 299100 + }, + { + "epoch": 25.89, + "learning_rate": 6.859487416567944e-06, + "loss": 0.8848, + "step": 299110 + }, + { + "epoch": 25.89, + "learning_rate": 6.858042705654599e-06, + "loss": 0.9203, + "step": 299120 + }, + { + "epoch": 25.89, + "learning_rate": 6.856597994741253e-06, + "loss": 0.9448, + "step": 299130 + }, + { + "epoch": 25.89, + "learning_rate": 6.855153283827906e-06, + "loss": 0.9218, + "step": 299140 + }, + { + "epoch": 25.89, + "learning_rate": 6.85370857291456e-06, + "loss": 0.8863, + "step": 299150 + }, + { + "epoch": 25.89, + "learning_rate": 6.852263862001214e-06, + "loss": 0.9604, + "step": 299160 + }, + { + "epoch": 25.9, + "learning_rate": 6.850819151087867e-06, + "loss": 0.9299, + "step": 299170 + }, + { + "epoch": 25.9, + "learning_rate": 6.849374440174522e-06, + "loss": 0.9102, + "step": 299180 + }, + { + "epoch": 25.9, + "learning_rate": 6.847929729261175e-06, + "loss": 0.9098, + "step": 299190 + }, + { + "epoch": 25.9, + "learning_rate": 6.846485018347828e-06, + "loss": 0.9105, + "step": 299200 + }, + { + "epoch": 25.9, + "learning_rate": 6.845040307434483e-06, + "loss": 0.9573, + "step": 299210 + }, + { + "epoch": 25.9, + "learning_rate": 6.843595596521137e-06, + "loss": 0.889, + "step": 299220 + }, + { + "epoch": 25.9, + "learning_rate": 6.84215088560779e-06, + "loss": 0.9247, + "step": 299230 + }, + { + "epoch": 25.9, + "learning_rate": 6.840706174694444e-06, + "loss": 0.9746, + "step": 299240 + }, + { + "epoch": 25.9, + "learning_rate": 6.839261463781098e-06, + "loss": 0.8946, + "step": 299250 + }, + { + "epoch": 25.9, + "learning_rate": 6.837816752867751e-06, + "loss": 0.9645, + "step": 299260 + }, + { + "epoch": 25.9, + "learning_rate": 6.836372041954406e-06, + "loss": 0.9236, + "step": 299270 + }, + { + "epoch": 25.9, + "learning_rate": 6.834927331041059e-06, + "loss": 0.9078, + "step": 299280 + }, + { + "epoch": 25.91, + "learning_rate": 6.833482620127712e-06, + "loss": 0.928, + "step": 299290 + }, + { + "epoch": 25.91, + "learning_rate": 6.832037909214367e-06, + "loss": 0.8737, + "step": 299300 + }, + { + "epoch": 25.91, + "learning_rate": 6.830593198301021e-06, + "loss": 0.9246, + "step": 299310 + }, + { + "epoch": 25.91, + "learning_rate": 6.829148487387674e-06, + "loss": 0.8789, + "step": 299320 + }, + { + "epoch": 25.91, + "learning_rate": 6.827703776474328e-06, + "loss": 0.8696, + "step": 299330 + }, + { + "epoch": 25.91, + "learning_rate": 6.826259065560982e-06, + "loss": 0.874, + "step": 299340 + }, + { + "epoch": 25.91, + "learning_rate": 6.824814354647635e-06, + "loss": 0.9089, + "step": 299350 + }, + { + "epoch": 25.91, + "learning_rate": 6.82336964373429e-06, + "loss": 0.9425, + "step": 299360 + }, + { + "epoch": 25.91, + "learning_rate": 6.821924932820943e-06, + "loss": 0.9337, + "step": 299370 + }, + { + "epoch": 25.91, + "learning_rate": 6.820480221907596e-06, + "loss": 0.8911, + "step": 299380 + }, + { + "epoch": 25.91, + "learning_rate": 6.8190355109942495e-06, + "loss": 0.9236, + "step": 299390 + }, + { + "epoch": 25.92, + "learning_rate": 6.817590800080905e-06, + "loss": 0.9409, + "step": 299400 + }, + { + "epoch": 25.92, + "learning_rate": 6.816146089167558e-06, + "loss": 0.8972, + "step": 299410 + }, + { + "epoch": 25.92, + "learning_rate": 6.814701378254211e-06, + "loss": 0.921, + "step": 299420 + }, + { + "epoch": 25.92, + "learning_rate": 6.813256667340866e-06, + "loss": 0.9196, + "step": 299430 + }, + { + "epoch": 25.92, + "learning_rate": 6.811811956427519e-06, + "loss": 0.963, + "step": 299440 + }, + { + "epoch": 25.92, + "learning_rate": 6.8103672455141724e-06, + "loss": 0.9016, + "step": 299450 + }, + { + "epoch": 25.92, + "learning_rate": 6.808922534600827e-06, + "loss": 0.8884, + "step": 299460 + }, + { + "epoch": 25.92, + "learning_rate": 6.80747782368748e-06, + "loss": 0.8593, + "step": 299470 + }, + { + "epoch": 25.92, + "learning_rate": 6.8060331127741335e-06, + "loss": 0.8863, + "step": 299480 + }, + { + "epoch": 25.92, + "learning_rate": 6.8045884018607885e-06, + "loss": 0.8892, + "step": 299490 + }, + { + "epoch": 25.92, + "learning_rate": 6.803143690947442e-06, + "loss": 0.9509, + "step": 299500 + }, + { + "epoch": 25.92, + "learning_rate": 6.801698980034095e-06, + "loss": 0.9172, + "step": 299510 + }, + { + "epoch": 25.93, + "learning_rate": 6.80025426912075e-06, + "loss": 0.9084, + "step": 299520 + }, + { + "epoch": 25.93, + "learning_rate": 6.798809558207403e-06, + "loss": 0.8857, + "step": 299530 + }, + { + "epoch": 25.93, + "learning_rate": 6.797364847294056e-06, + "loss": 0.9016, + "step": 299540 + }, + { + "epoch": 25.93, + "learning_rate": 6.795920136380711e-06, + "loss": 0.8898, + "step": 299550 + }, + { + "epoch": 25.93, + "learning_rate": 6.794475425467364e-06, + "loss": 0.9129, + "step": 299560 + }, + { + "epoch": 25.93, + "learning_rate": 6.793030714554017e-06, + "loss": 0.9024, + "step": 299570 + }, + { + "epoch": 25.93, + "learning_rate": 6.7915860036406725e-06, + "loss": 0.9011, + "step": 299580 + }, + { + "epoch": 25.93, + "learning_rate": 6.790141292727326e-06, + "loss": 0.9054, + "step": 299590 + }, + { + "epoch": 25.93, + "learning_rate": 6.788696581813979e-06, + "loss": 0.9099, + "step": 299600 + }, + { + "epoch": 25.93, + "learning_rate": 6.7872518709006335e-06, + "loss": 0.9002, + "step": 299610 + }, + { + "epoch": 25.93, + "learning_rate": 6.785807159987287e-06, + "loss": 0.8512, + "step": 299620 + }, + { + "epoch": 25.94, + "learning_rate": 6.78436244907394e-06, + "loss": 0.8923, + "step": 299630 + }, + { + "epoch": 25.94, + "learning_rate": 6.7829177381605946e-06, + "loss": 0.8807, + "step": 299640 + }, + { + "epoch": 25.94, + "learning_rate": 6.781473027247248e-06, + "loss": 0.9014, + "step": 299650 + }, + { + "epoch": 25.94, + "learning_rate": 6.780028316333901e-06, + "loss": 0.929, + "step": 299660 + }, + { + "epoch": 25.94, + "learning_rate": 6.7785836054205564e-06, + "loss": 0.9775, + "step": 299670 + }, + { + "epoch": 25.94, + "learning_rate": 6.77713889450721e-06, + "loss": 0.9331, + "step": 299680 + }, + { + "epoch": 25.94, + "learning_rate": 6.775694183593863e-06, + "loss": 0.9223, + "step": 299690 + }, + { + "epoch": 25.94, + "learning_rate": 6.7742494726805175e-06, + "loss": 0.9199, + "step": 299700 + }, + { + "epoch": 25.94, + "learning_rate": 6.772804761767171e-06, + "loss": 0.8806, + "step": 299710 + }, + { + "epoch": 25.94, + "learning_rate": 6.771360050853824e-06, + "loss": 0.9723, + "step": 299720 + }, + { + "epoch": 25.94, + "learning_rate": 6.7699153399404785e-06, + "loss": 0.8796, + "step": 299730 + }, + { + "epoch": 25.94, + "learning_rate": 6.768470629027132e-06, + "loss": 0.9113, + "step": 299740 + }, + { + "epoch": 25.95, + "learning_rate": 6.767025918113785e-06, + "loss": 0.8787, + "step": 299750 + }, + { + "epoch": 25.95, + "learning_rate": 6.76558120720044e-06, + "loss": 0.9232, + "step": 299760 + }, + { + "epoch": 25.95, + "learning_rate": 6.764136496287094e-06, + "loss": 0.952, + "step": 299770 + }, + { + "epoch": 25.95, + "learning_rate": 6.762691785373747e-06, + "loss": 0.9193, + "step": 299780 + }, + { + "epoch": 25.95, + "learning_rate": 6.761247074460401e-06, + "loss": 1.0129, + "step": 299790 + }, + { + "epoch": 25.95, + "learning_rate": 6.759802363547055e-06, + "loss": 0.8851, + "step": 299800 + }, + { + "epoch": 25.95, + "learning_rate": 6.758357652633708e-06, + "loss": 0.934, + "step": 299810 + }, + { + "epoch": 25.95, + "learning_rate": 6.7569129417203624e-06, + "loss": 0.9189, + "step": 299820 + }, + { + "epoch": 25.95, + "learning_rate": 6.755468230807016e-06, + "loss": 0.9496, + "step": 299830 + }, + { + "epoch": 25.95, + "learning_rate": 6.754023519893669e-06, + "loss": 0.9248, + "step": 299840 + }, + { + "epoch": 25.95, + "learning_rate": 6.752578808980323e-06, + "loss": 0.8683, + "step": 299850 + }, + { + "epoch": 25.96, + "learning_rate": 6.751134098066978e-06, + "loss": 0.8942, + "step": 299860 + }, + { + "epoch": 25.96, + "learning_rate": 6.749689387153631e-06, + "loss": 0.9115, + "step": 299870 + }, + { + "epoch": 25.96, + "learning_rate": 6.748244676240284e-06, + "loss": 0.8458, + "step": 299880 + }, + { + "epoch": 25.96, + "learning_rate": 6.746799965326939e-06, + "loss": 0.8747, + "step": 299890 + }, + { + "epoch": 25.96, + "learning_rate": 6.745355254413592e-06, + "loss": 0.9621, + "step": 299900 + }, + { + "epoch": 25.96, + "learning_rate": 6.7439105435002455e-06, + "loss": 0.8872, + "step": 299910 + }, + { + "epoch": 25.96, + "learning_rate": 6.7424658325869e-06, + "loss": 0.8623, + "step": 299920 + }, + { + "epoch": 25.96, + "learning_rate": 6.741021121673553e-06, + "loss": 0.912, + "step": 299930 + }, + { + "epoch": 25.96, + "learning_rate": 6.7395764107602066e-06, + "loss": 0.9241, + "step": 299940 + }, + { + "epoch": 25.96, + "learning_rate": 6.738131699846862e-06, + "loss": 0.8902, + "step": 299950 + }, + { + "epoch": 25.96, + "learning_rate": 6.736686988933515e-06, + "loss": 0.8574, + "step": 299960 + }, + { + "epoch": 25.96, + "learning_rate": 6.735242278020168e-06, + "loss": 0.9076, + "step": 299970 + }, + { + "epoch": 25.97, + "learning_rate": 6.733797567106823e-06, + "loss": 0.8889, + "step": 299980 + }, + { + "epoch": 25.97, + "learning_rate": 6.732352856193476e-06, + "loss": 0.9153, + "step": 299990 + }, + { + "epoch": 25.97, + "learning_rate": 6.7309081452801295e-06, + "loss": 0.8546, + "step": 300000 + }, + { + "epoch": 25.97, + "learning_rate": 6.729463434366784e-06, + "loss": 0.9063, + "step": 300010 + }, + { + "epoch": 25.97, + "learning_rate": 6.728018723453437e-06, + "loss": 0.8867, + "step": 300020 + }, + { + "epoch": 25.97, + "learning_rate": 6.7265740125400905e-06, + "loss": 0.8736, + "step": 300030 + }, + { + "epoch": 25.97, + "learning_rate": 6.725129301626746e-06, + "loss": 0.8709, + "step": 300040 + }, + { + "epoch": 25.97, + "learning_rate": 6.723684590713398e-06, + "loss": 0.9485, + "step": 300050 + }, + { + "epoch": 25.97, + "learning_rate": 6.7222398798000515e-06, + "loss": 0.9021, + "step": 300060 + }, + { + "epoch": 25.97, + "learning_rate": 6.720795168886707e-06, + "loss": 0.8571, + "step": 300070 + }, + { + "epoch": 25.97, + "learning_rate": 6.71935045797336e-06, + "loss": 0.936, + "step": 300080 + }, + { + "epoch": 25.97, + "learning_rate": 6.717905747060013e-06, + "loss": 0.9404, + "step": 300090 + }, + { + "epoch": 25.98, + "learning_rate": 6.716461036146668e-06, + "loss": 0.9347, + "step": 300100 + }, + { + "epoch": 25.98, + "learning_rate": 6.715016325233321e-06, + "loss": 0.9013, + "step": 300110 + }, + { + "epoch": 25.98, + "learning_rate": 6.7135716143199745e-06, + "loss": 0.896, + "step": 300120 + }, + { + "epoch": 25.98, + "learning_rate": 6.7121269034066295e-06, + "loss": 0.9427, + "step": 300130 + }, + { + "epoch": 25.98, + "learning_rate": 6.710682192493282e-06, + "loss": 0.844, + "step": 300140 + }, + { + "epoch": 25.98, + "learning_rate": 6.7092374815799355e-06, + "loss": 0.9198, + "step": 300150 + }, + { + "epoch": 25.98, + "learning_rate": 6.7077927706665906e-06, + "loss": 0.965, + "step": 300160 + }, + { + "epoch": 25.98, + "learning_rate": 6.706348059753244e-06, + "loss": 0.8816, + "step": 300170 + }, + { + "epoch": 25.98, + "learning_rate": 6.704903348839897e-06, + "loss": 0.9283, + "step": 300180 + }, + { + "epoch": 25.98, + "learning_rate": 6.703458637926552e-06, + "loss": 0.872, + "step": 300190 + }, + { + "epoch": 25.98, + "learning_rate": 6.702013927013205e-06, + "loss": 0.8764, + "step": 300200 + }, + { + "epoch": 25.99, + "learning_rate": 6.700569216099858e-06, + "loss": 0.9168, + "step": 300210 + }, + { + "epoch": 25.99, + "learning_rate": 6.6991245051865135e-06, + "loss": 0.927, + "step": 300220 + }, + { + "epoch": 25.99, + "learning_rate": 6.697679794273166e-06, + "loss": 0.9403, + "step": 300230 + }, + { + "epoch": 25.99, + "learning_rate": 6.6962350833598194e-06, + "loss": 0.9192, + "step": 300240 + }, + { + "epoch": 25.99, + "learning_rate": 6.6947903724464745e-06, + "loss": 0.8963, + "step": 300250 + }, + { + "epoch": 25.99, + "learning_rate": 6.693345661533128e-06, + "loss": 0.9214, + "step": 300260 + }, + { + "epoch": 25.99, + "learning_rate": 6.691900950619781e-06, + "loss": 0.9418, + "step": 300270 + }, + { + "epoch": 25.99, + "learning_rate": 6.690456239706435e-06, + "loss": 0.9909, + "step": 300280 + }, + { + "epoch": 25.99, + "learning_rate": 6.689011528793089e-06, + "loss": 0.8943, + "step": 300290 + }, + { + "epoch": 25.99, + "learning_rate": 6.687566817879742e-06, + "loss": 0.9078, + "step": 300300 + }, + { + "epoch": 25.99, + "learning_rate": 6.686122106966396e-06, + "loss": 0.8903, + "step": 300310 + }, + { + "epoch": 25.99, + "learning_rate": 6.68467739605305e-06, + "loss": 0.9196, + "step": 300320 + }, + { + "epoch": 26.0, + "learning_rate": 6.683232685139703e-06, + "loss": 0.8954, + "step": 300330 + }, + { + "epoch": 26.0, + "learning_rate": 6.681787974226357e-06, + "loss": 0.9136, + "step": 300340 + }, + { + "epoch": 26.0, + "learning_rate": 6.680343263313012e-06, + "loss": 0.9037, + "step": 300350 + }, + { + "epoch": 26.0, + "learning_rate": 6.678898552399665e-06, + "loss": 0.9403, + "step": 300360 + }, + { + "epoch": 26.0, + "learning_rate": 6.677453841486319e-06, + "loss": 0.9272, + "step": 300370 + }, + { + "epoch": 26.0, + "eval_cer": 0.9930666549448872, + "eval_em": 0, + "eval_f1": 0.12098808287566994, + "eval_loss": 0.9345543384552002, + "eval_runtime": 894.5831, + "eval_samples_per_second": 5.74, + "eval_steps_per_second": 0.718, + "eval_wer": 0.9665963536947737, + "step": 300379 + }, + { + "epoch": 26.0, + "learning_rate": 6.676009130572973e-06, + "loss": 0.9554, + "step": 300380 + }, + { + "epoch": 26.0, + "learning_rate": 6.674564419659626e-06, + "loss": 0.9645, + "step": 300390 + }, + { + "epoch": 26.0, + "learning_rate": 6.67311970874628e-06, + "loss": 0.8583, + "step": 300400 + }, + { + "epoch": 26.0, + "learning_rate": 6.671674997832934e-06, + "loss": 0.8853, + "step": 300410 + }, + { + "epoch": 26.0, + "learning_rate": 6.670230286919587e-06, + "loss": 0.8522, + "step": 300420 + }, + { + "epoch": 26.0, + "learning_rate": 6.668785576006241e-06, + "loss": 0.8891, + "step": 300430 + }, + { + "epoch": 26.01, + "learning_rate": 6.667340865092896e-06, + "loss": 0.9082, + "step": 300440 + }, + { + "epoch": 26.01, + "learning_rate": 6.665896154179549e-06, + "loss": 0.9322, + "step": 300450 + }, + { + "epoch": 26.01, + "learning_rate": 6.664451443266203e-06, + "loss": 0.8848, + "step": 300460 + }, + { + "epoch": 26.01, + "learning_rate": 6.663006732352857e-06, + "loss": 0.9052, + "step": 300470 + }, + { + "epoch": 26.01, + "learning_rate": 6.66156202143951e-06, + "loss": 0.9132, + "step": 300480 + }, + { + "epoch": 26.01, + "learning_rate": 6.660117310526164e-06, + "loss": 0.9463, + "step": 300490 + }, + { + "epoch": 26.01, + "learning_rate": 6.658672599612818e-06, + "loss": 0.8761, + "step": 300500 + }, + { + "epoch": 26.01, + "learning_rate": 6.657227888699471e-06, + "loss": 0.9348, + "step": 300510 + }, + { + "epoch": 26.01, + "learning_rate": 6.655783177786125e-06, + "loss": 0.8469, + "step": 300520 + }, + { + "epoch": 26.01, + "learning_rate": 6.65433846687278e-06, + "loss": 0.9191, + "step": 300530 + }, + { + "epoch": 26.01, + "learning_rate": 6.652893755959433e-06, + "loss": 0.9059, + "step": 300540 + }, + { + "epoch": 26.01, + "learning_rate": 6.6514490450460865e-06, + "loss": 0.9361, + "step": 300550 + }, + { + "epoch": 26.02, + "learning_rate": 6.650004334132741e-06, + "loss": 0.907, + "step": 300560 + }, + { + "epoch": 26.02, + "learning_rate": 6.648559623219394e-06, + "loss": 0.8884, + "step": 300570 + }, + { + "epoch": 26.02, + "learning_rate": 6.6471149123060476e-06, + "loss": 0.8967, + "step": 300580 + }, + { + "epoch": 26.02, + "learning_rate": 6.645670201392702e-06, + "loss": 0.9116, + "step": 300590 + }, + { + "epoch": 26.02, + "learning_rate": 6.644225490479355e-06, + "loss": 0.8505, + "step": 300600 + }, + { + "epoch": 26.02, + "learning_rate": 6.642780779566009e-06, + "loss": 0.9012, + "step": 300610 + }, + { + "epoch": 26.02, + "learning_rate": 6.641336068652664e-06, + "loss": 0.8653, + "step": 300620 + }, + { + "epoch": 26.02, + "learning_rate": 6.639891357739317e-06, + "loss": 0.926, + "step": 300630 + }, + { + "epoch": 26.02, + "learning_rate": 6.6384466468259705e-06, + "loss": 0.9007, + "step": 300640 + }, + { + "epoch": 26.02, + "learning_rate": 6.637001935912625e-06, + "loss": 0.8642, + "step": 300650 + }, + { + "epoch": 26.02, + "learning_rate": 6.635557224999278e-06, + "loss": 0.9109, + "step": 300660 + }, + { + "epoch": 26.03, + "learning_rate": 6.6341125140859315e-06, + "loss": 0.8796, + "step": 300670 + }, + { + "epoch": 26.03, + "learning_rate": 6.632667803172586e-06, + "loss": 0.9387, + "step": 300680 + }, + { + "epoch": 26.03, + "learning_rate": 6.631223092259239e-06, + "loss": 0.895, + "step": 300690 + }, + { + "epoch": 26.03, + "learning_rate": 6.6297783813458925e-06, + "loss": 0.9395, + "step": 300700 + }, + { + "epoch": 26.03, + "learning_rate": 6.628333670432548e-06, + "loss": 0.8424, + "step": 300710 + }, + { + "epoch": 26.03, + "learning_rate": 6.626888959519201e-06, + "loss": 0.9375, + "step": 300720 + }, + { + "epoch": 26.03, + "learning_rate": 6.625444248605854e-06, + "loss": 0.9333, + "step": 300730 + }, + { + "epoch": 26.03, + "learning_rate": 6.623999537692507e-06, + "loss": 0.9235, + "step": 300740 + }, + { + "epoch": 26.03, + "learning_rate": 6.622554826779162e-06, + "loss": 0.8817, + "step": 300750 + }, + { + "epoch": 26.03, + "learning_rate": 6.6211101158658154e-06, + "loss": 0.9119, + "step": 300760 + }, + { + "epoch": 26.03, + "learning_rate": 6.619665404952469e-06, + "loss": 0.8894, + "step": 300770 + }, + { + "epoch": 26.03, + "learning_rate": 6.618220694039123e-06, + "loss": 0.8812, + "step": 300780 + }, + { + "epoch": 26.04, + "learning_rate": 6.6167759831257765e-06, + "loss": 0.8986, + "step": 300790 + }, + { + "epoch": 26.04, + "learning_rate": 6.61533127221243e-06, + "loss": 0.8605, + "step": 300800 + }, + { + "epoch": 26.04, + "learning_rate": 6.613886561299085e-06, + "loss": 0.8679, + "step": 300810 + }, + { + "epoch": 26.04, + "learning_rate": 6.612441850385738e-06, + "loss": 0.9049, + "step": 300820 + }, + { + "epoch": 26.04, + "learning_rate": 6.610997139472391e-06, + "loss": 0.9151, + "step": 300830 + }, + { + "epoch": 26.04, + "learning_rate": 6.609552428559046e-06, + "loss": 0.8697, + "step": 300840 + }, + { + "epoch": 26.04, + "learning_rate": 6.608107717645699e-06, + "loss": 0.9157, + "step": 300850 + }, + { + "epoch": 26.04, + "learning_rate": 6.606663006732353e-06, + "loss": 0.9499, + "step": 300860 + }, + { + "epoch": 26.04, + "learning_rate": 6.605218295819007e-06, + "loss": 0.9599, + "step": 300870 + }, + { + "epoch": 26.04, + "learning_rate": 6.60377358490566e-06, + "loss": 0.9115, + "step": 300880 + }, + { + "epoch": 26.04, + "learning_rate": 6.602328873992314e-06, + "loss": 0.9087, + "step": 300890 + }, + { + "epoch": 26.05, + "learning_rate": 6.600884163078969e-06, + "loss": 0.8888, + "step": 300900 + }, + { + "epoch": 26.05, + "learning_rate": 6.599439452165622e-06, + "loss": 0.853, + "step": 300910 + }, + { + "epoch": 26.05, + "learning_rate": 6.597994741252275e-06, + "loss": 0.9456, + "step": 300920 + }, + { + "epoch": 26.05, + "learning_rate": 6.59655003033893e-06, + "loss": 0.8917, + "step": 300930 + }, + { + "epoch": 26.05, + "learning_rate": 6.595105319425583e-06, + "loss": 0.9197, + "step": 300940 + }, + { + "epoch": 26.05, + "learning_rate": 6.593660608512237e-06, + "loss": 0.903, + "step": 300950 + }, + { + "epoch": 26.05, + "learning_rate": 6.592215897598891e-06, + "loss": 0.9063, + "step": 300960 + }, + { + "epoch": 26.05, + "learning_rate": 6.590771186685544e-06, + "loss": 0.9641, + "step": 300970 + }, + { + "epoch": 26.05, + "learning_rate": 6.589326475772198e-06, + "loss": 0.9184, + "step": 300980 + }, + { + "epoch": 26.05, + "learning_rate": 6.587881764858853e-06, + "loss": 0.9088, + "step": 300990 + }, + { + "epoch": 26.05, + "learning_rate": 6.586437053945506e-06, + "loss": 0.8782, + "step": 301000 + }, + { + "epoch": 26.05, + "learning_rate": 6.584992343032159e-06, + "loss": 0.9381, + "step": 301010 + }, + { + "epoch": 26.06, + "learning_rate": 6.583547632118814e-06, + "loss": 0.856, + "step": 301020 + }, + { + "epoch": 26.06, + "learning_rate": 6.582102921205467e-06, + "loss": 0.8704, + "step": 301030 + }, + { + "epoch": 26.06, + "learning_rate": 6.580658210292121e-06, + "loss": 0.8566, + "step": 301040 + }, + { + "epoch": 26.06, + "learning_rate": 6.579213499378775e-06, + "loss": 0.8874, + "step": 301050 + }, + { + "epoch": 26.06, + "learning_rate": 6.577768788465428e-06, + "loss": 0.8983, + "step": 301060 + }, + { + "epoch": 26.06, + "learning_rate": 6.576324077552082e-06, + "loss": 0.881, + "step": 301070 + }, + { + "epoch": 26.06, + "learning_rate": 6.574879366638737e-06, + "loss": 0.9721, + "step": 301080 + }, + { + "epoch": 26.06, + "learning_rate": 6.57343465572539e-06, + "loss": 0.8981, + "step": 301090 + }, + { + "epoch": 26.06, + "learning_rate": 6.571989944812043e-06, + "loss": 0.8691, + "step": 301100 + }, + { + "epoch": 26.06, + "learning_rate": 6.570545233898698e-06, + "loss": 0.9863, + "step": 301110 + }, + { + "epoch": 26.06, + "learning_rate": 6.569100522985351e-06, + "loss": 0.9206, + "step": 301120 + }, + { + "epoch": 26.07, + "learning_rate": 6.567655812072005e-06, + "loss": 0.9086, + "step": 301130 + }, + { + "epoch": 26.07, + "learning_rate": 6.566211101158659e-06, + "loss": 0.9422, + "step": 301140 + }, + { + "epoch": 26.07, + "learning_rate": 6.564766390245312e-06, + "loss": 0.8947, + "step": 301150 + }, + { + "epoch": 26.07, + "learning_rate": 6.563321679331966e-06, + "loss": 0.9679, + "step": 301160 + }, + { + "epoch": 26.07, + "learning_rate": 6.561876968418621e-06, + "loss": 0.8195, + "step": 301170 + }, + { + "epoch": 26.07, + "learning_rate": 6.560432257505274e-06, + "loss": 0.9562, + "step": 301180 + }, + { + "epoch": 26.07, + "learning_rate": 6.558987546591927e-06, + "loss": 0.8863, + "step": 301190 + }, + { + "epoch": 26.07, + "learning_rate": 6.55754283567858e-06, + "loss": 0.9603, + "step": 301200 + }, + { + "epoch": 26.07, + "learning_rate": 6.556098124765235e-06, + "loss": 0.9102, + "step": 301210 + }, + { + "epoch": 26.07, + "learning_rate": 6.5546534138518886e-06, + "loss": 0.9135, + "step": 301220 + }, + { + "epoch": 26.07, + "learning_rate": 6.553208702938542e-06, + "loss": 0.851, + "step": 301230 + }, + { + "epoch": 26.07, + "learning_rate": 6.551763992025196e-06, + "loss": 0.8939, + "step": 301240 + }, + { + "epoch": 26.08, + "learning_rate": 6.55031928111185e-06, + "loss": 0.8357, + "step": 301250 + }, + { + "epoch": 26.08, + "learning_rate": 6.548874570198503e-06, + "loss": 0.8885, + "step": 301260 + }, + { + "epoch": 26.08, + "learning_rate": 6.547429859285158e-06, + "loss": 0.8871, + "step": 301270 + }, + { + "epoch": 26.08, + "learning_rate": 6.545985148371811e-06, + "loss": 0.9297, + "step": 301280 + }, + { + "epoch": 26.08, + "learning_rate": 6.544540437458464e-06, + "loss": 0.8734, + "step": 301290 + }, + { + "epoch": 26.08, + "learning_rate": 6.543095726545119e-06, + "loss": 0.9475, + "step": 301300 + }, + { + "epoch": 26.08, + "learning_rate": 6.5416510156317725e-06, + "loss": 0.8878, + "step": 301310 + }, + { + "epoch": 26.08, + "learning_rate": 6.540206304718426e-06, + "loss": 0.8805, + "step": 301320 + }, + { + "epoch": 26.08, + "learning_rate": 6.53876159380508e-06, + "loss": 0.8558, + "step": 301330 + }, + { + "epoch": 26.08, + "learning_rate": 6.5373168828917335e-06, + "loss": 0.8774, + "step": 301340 + }, + { + "epoch": 26.08, + "learning_rate": 6.535872171978387e-06, + "loss": 0.8806, + "step": 301350 + }, + { + "epoch": 26.08, + "learning_rate": 6.534427461065042e-06, + "loss": 0.9107, + "step": 301360 + }, + { + "epoch": 26.09, + "learning_rate": 6.5329827501516946e-06, + "loss": 0.9474, + "step": 301370 + }, + { + "epoch": 26.09, + "learning_rate": 6.531538039238348e-06, + "loss": 0.9053, + "step": 301380 + }, + { + "epoch": 26.09, + "learning_rate": 6.530093328325003e-06, + "loss": 0.8988, + "step": 301390 + }, + { + "epoch": 26.09, + "learning_rate": 6.5286486174116564e-06, + "loss": 0.8507, + "step": 301400 + }, + { + "epoch": 26.09, + "learning_rate": 6.52720390649831e-06, + "loss": 0.9402, + "step": 301410 + }, + { + "epoch": 26.09, + "learning_rate": 6.525759195584964e-06, + "loss": 0.8904, + "step": 301420 + }, + { + "epoch": 26.09, + "learning_rate": 6.5243144846716175e-06, + "loss": 0.8811, + "step": 301430 + }, + { + "epoch": 26.09, + "learning_rate": 6.522869773758271e-06, + "loss": 0.9735, + "step": 301440 + }, + { + "epoch": 26.09, + "learning_rate": 6.521425062844926e-06, + "loss": 0.8891, + "step": 301450 + }, + { + "epoch": 26.09, + "learning_rate": 6.5199803519315785e-06, + "loss": 0.9146, + "step": 301460 + }, + { + "epoch": 26.09, + "learning_rate": 6.518535641018232e-06, + "loss": 0.9271, + "step": 301470 + }, + { + "epoch": 26.1, + "learning_rate": 6.517090930104887e-06, + "loss": 0.8587, + "step": 301480 + }, + { + "epoch": 26.1, + "learning_rate": 6.51564621919154e-06, + "loss": 0.8491, + "step": 301490 + }, + { + "epoch": 26.1, + "learning_rate": 6.514201508278194e-06, + "loss": 0.9345, + "step": 301500 + }, + { + "epoch": 26.1, + "learning_rate": 6.512756797364848e-06, + "loss": 0.8939, + "step": 301510 + }, + { + "epoch": 26.1, + "learning_rate": 6.511312086451501e-06, + "loss": 0.8806, + "step": 301520 + }, + { + "epoch": 26.1, + "learning_rate": 6.509867375538155e-06, + "loss": 0.952, + "step": 301530 + }, + { + "epoch": 26.1, + "learning_rate": 6.50842266462481e-06, + "loss": 0.8754, + "step": 301540 + }, + { + "epoch": 26.1, + "learning_rate": 6.5069779537114624e-06, + "loss": 0.8962, + "step": 301550 + }, + { + "epoch": 26.1, + "learning_rate": 6.505533242798116e-06, + "loss": 0.9518, + "step": 301560 + }, + { + "epoch": 26.1, + "learning_rate": 6.504088531884771e-06, + "loss": 0.9523, + "step": 301570 + }, + { + "epoch": 26.1, + "learning_rate": 6.502643820971424e-06, + "loss": 0.8984, + "step": 301580 + }, + { + "epoch": 26.1, + "learning_rate": 6.501199110058078e-06, + "loss": 0.8833, + "step": 301590 + }, + { + "epoch": 26.11, + "learning_rate": 6.499754399144732e-06, + "loss": 0.8905, + "step": 301600 + }, + { + "epoch": 26.11, + "learning_rate": 6.498309688231385e-06, + "loss": 0.9477, + "step": 301610 + }, + { + "epoch": 26.11, + "learning_rate": 6.496864977318039e-06, + "loss": 0.9198, + "step": 301620 + }, + { + "epoch": 26.11, + "learning_rate": 6.495420266404694e-06, + "loss": 0.9514, + "step": 301630 + }, + { + "epoch": 26.11, + "learning_rate": 6.493975555491346e-06, + "loss": 0.8844, + "step": 301640 + }, + { + "epoch": 26.11, + "learning_rate": 6.492530844578e-06, + "loss": 0.9209, + "step": 301650 + }, + { + "epoch": 26.11, + "learning_rate": 6.491086133664653e-06, + "loss": 0.9028, + "step": 301660 + }, + { + "epoch": 26.11, + "learning_rate": 6.489641422751308e-06, + "loss": 0.9169, + "step": 301670 + }, + { + "epoch": 26.11, + "learning_rate": 6.488196711837962e-06, + "loss": 0.9333, + "step": 301680 + }, + { + "epoch": 26.11, + "learning_rate": 6.486752000924615e-06, + "loss": 0.9143, + "step": 301690 + }, + { + "epoch": 26.11, + "learning_rate": 6.485307290011269e-06, + "loss": 0.8822, + "step": 301700 + }, + { + "epoch": 26.12, + "learning_rate": 6.483862579097923e-06, + "loss": 0.896, + "step": 301710 + }, + { + "epoch": 26.12, + "learning_rate": 6.482417868184576e-06, + "loss": 0.8603, + "step": 301720 + }, + { + "epoch": 26.12, + "learning_rate": 6.48097315727123e-06, + "loss": 0.9221, + "step": 301730 + }, + { + "epoch": 26.12, + "learning_rate": 6.479528446357884e-06, + "loss": 0.962, + "step": 301740 + }, + { + "epoch": 26.12, + "learning_rate": 6.478083735444537e-06, + "loss": 0.9246, + "step": 301750 + }, + { + "epoch": 26.12, + "learning_rate": 6.476639024531192e-06, + "loss": 0.8917, + "step": 301760 + }, + { + "epoch": 26.12, + "learning_rate": 6.475194313617846e-06, + "loss": 0.8765, + "step": 301770 + }, + { + "epoch": 26.12, + "learning_rate": 6.473749602704499e-06, + "loss": 0.8564, + "step": 301780 + }, + { + "epoch": 26.12, + "learning_rate": 6.472304891791153e-06, + "loss": 0.9641, + "step": 301790 + }, + { + "epoch": 26.12, + "learning_rate": 6.470860180877807e-06, + "loss": 0.8741, + "step": 301800 + }, + { + "epoch": 26.12, + "learning_rate": 6.46941546996446e-06, + "loss": 0.9154, + "step": 301810 + }, + { + "epoch": 26.12, + "learning_rate": 6.467970759051114e-06, + "loss": 0.9172, + "step": 301820 + }, + { + "epoch": 26.13, + "learning_rate": 6.466526048137768e-06, + "loss": 0.9128, + "step": 301830 + }, + { + "epoch": 26.13, + "learning_rate": 6.465081337224421e-06, + "loss": 0.98, + "step": 301840 + }, + { + "epoch": 26.13, + "learning_rate": 6.463636626311076e-06, + "loss": 0.9435, + "step": 301850 + }, + { + "epoch": 26.13, + "learning_rate": 6.4621919153977296e-06, + "loss": 0.8744, + "step": 301860 + }, + { + "epoch": 26.13, + "learning_rate": 6.460747204484383e-06, + "loss": 0.8672, + "step": 301870 + }, + { + "epoch": 26.13, + "learning_rate": 6.459302493571037e-06, + "loss": 0.873, + "step": 301880 + }, + { + "epoch": 26.13, + "learning_rate": 6.457857782657691e-06, + "loss": 0.8813, + "step": 301890 + }, + { + "epoch": 26.13, + "learning_rate": 6.456413071744344e-06, + "loss": 0.9129, + "step": 301900 + }, + { + "epoch": 26.13, + "learning_rate": 6.454968360830998e-06, + "loss": 0.891, + "step": 301910 + }, + { + "epoch": 26.13, + "learning_rate": 6.453523649917652e-06, + "loss": 0.8559, + "step": 301920 + }, + { + "epoch": 26.13, + "learning_rate": 6.452078939004305e-06, + "loss": 0.8921, + "step": 301930 + }, + { + "epoch": 26.14, + "learning_rate": 6.45063422809096e-06, + "loss": 0.9201, + "step": 301940 + }, + { + "epoch": 26.14, + "learning_rate": 6.4491895171776135e-06, + "loss": 0.8551, + "step": 301950 + }, + { + "epoch": 26.14, + "learning_rate": 6.447744806264267e-06, + "loss": 0.9313, + "step": 301960 + }, + { + "epoch": 26.14, + "learning_rate": 6.446300095350921e-06, + "loss": 0.8925, + "step": 301970 + }, + { + "epoch": 26.14, + "learning_rate": 6.4448553844375745e-06, + "loss": 0.8892, + "step": 301980 + }, + { + "epoch": 26.14, + "learning_rate": 6.443410673524228e-06, + "loss": 0.9029, + "step": 301990 + }, + { + "epoch": 26.14, + "learning_rate": 6.441965962610882e-06, + "loss": 0.8598, + "step": 302000 + }, + { + "epoch": 26.14, + "learning_rate": 6.4405212516975356e-06, + "loss": 0.9178, + "step": 302010 + }, + { + "epoch": 26.14, + "learning_rate": 6.439076540784189e-06, + "loss": 0.8536, + "step": 302020 + }, + { + "epoch": 26.14, + "learning_rate": 6.437631829870844e-06, + "loss": 0.9178, + "step": 302030 + }, + { + "epoch": 26.14, + "learning_rate": 6.4361871189574974e-06, + "loss": 0.9768, + "step": 302040 + }, + { + "epoch": 26.14, + "learning_rate": 6.43474240804415e-06, + "loss": 0.9079, + "step": 302050 + }, + { + "epoch": 26.15, + "learning_rate": 6.433297697130805e-06, + "loss": 0.9135, + "step": 302060 + }, + { + "epoch": 26.15, + "learning_rate": 6.4318529862174585e-06, + "loss": 0.9449, + "step": 302070 + }, + { + "epoch": 26.15, + "learning_rate": 6.430408275304112e-06, + "loss": 0.9292, + "step": 302080 + }, + { + "epoch": 26.15, + "learning_rate": 6.428963564390765e-06, + "loss": 0.8952, + "step": 302090 + }, + { + "epoch": 26.15, + "learning_rate": 6.4275188534774195e-06, + "loss": 0.9622, + "step": 302100 + }, + { + "epoch": 26.15, + "learning_rate": 6.426074142564073e-06, + "loss": 0.8799, + "step": 302110 + }, + { + "epoch": 26.15, + "learning_rate": 6.424629431650726e-06, + "loss": 0.9008, + "step": 302120 + }, + { + "epoch": 26.15, + "learning_rate": 6.423184720737381e-06, + "loss": 0.9774, + "step": 302130 + }, + { + "epoch": 26.15, + "learning_rate": 6.421740009824034e-06, + "loss": 0.9425, + "step": 302140 + }, + { + "epoch": 26.15, + "learning_rate": 6.420295298910687e-06, + "loss": 0.8719, + "step": 302150 + }, + { + "epoch": 26.15, + "learning_rate": 6.418850587997342e-06, + "loss": 0.9095, + "step": 302160 + }, + { + "epoch": 26.16, + "learning_rate": 6.417405877083996e-06, + "loss": 0.9446, + "step": 302170 + }, + { + "epoch": 26.16, + "learning_rate": 6.415961166170649e-06, + "loss": 0.8641, + "step": 302180 + }, + { + "epoch": 26.16, + "learning_rate": 6.4145164552573034e-06, + "loss": 0.9113, + "step": 302190 + }, + { + "epoch": 26.16, + "learning_rate": 6.413071744343957e-06, + "loss": 0.9091, + "step": 302200 + }, + { + "epoch": 26.16, + "learning_rate": 6.41162703343061e-06, + "loss": 0.9485, + "step": 302210 + }, + { + "epoch": 26.16, + "learning_rate": 6.410182322517265e-06, + "loss": 0.9732, + "step": 302220 + }, + { + "epoch": 26.16, + "learning_rate": 6.408737611603918e-06, + "loss": 0.9159, + "step": 302230 + }, + { + "epoch": 26.16, + "learning_rate": 6.407292900690571e-06, + "loss": 0.901, + "step": 302240 + }, + { + "epoch": 26.16, + "learning_rate": 6.405848189777226e-06, + "loss": 0.8409, + "step": 302250 + }, + { + "epoch": 26.16, + "learning_rate": 6.40440347886388e-06, + "loss": 0.9483, + "step": 302260 + }, + { + "epoch": 26.16, + "learning_rate": 6.402958767950533e-06, + "loss": 0.8994, + "step": 302270 + }, + { + "epoch": 26.16, + "learning_rate": 6.401514057037187e-06, + "loss": 0.8829, + "step": 302280 + }, + { + "epoch": 26.17, + "learning_rate": 6.400069346123841e-06, + "loss": 0.9587, + "step": 302290 + }, + { + "epoch": 26.17, + "learning_rate": 6.398624635210494e-06, + "loss": 0.8512, + "step": 302300 + }, + { + "epoch": 26.17, + "learning_rate": 6.397179924297149e-06, + "loss": 0.9572, + "step": 302310 + }, + { + "epoch": 26.17, + "learning_rate": 6.395735213383802e-06, + "loss": 0.8885, + "step": 302320 + }, + { + "epoch": 26.17, + "learning_rate": 6.394290502470455e-06, + "loss": 0.9037, + "step": 302330 + }, + { + "epoch": 26.17, + "learning_rate": 6.39284579155711e-06, + "loss": 0.9302, + "step": 302340 + }, + { + "epoch": 26.17, + "learning_rate": 6.391401080643764e-06, + "loss": 0.9298, + "step": 302350 + }, + { + "epoch": 26.17, + "learning_rate": 6.389956369730417e-06, + "loss": 0.8515, + "step": 302360 + }, + { + "epoch": 26.17, + "learning_rate": 6.388511658817071e-06, + "loss": 0.8921, + "step": 302370 + }, + { + "epoch": 26.17, + "learning_rate": 6.387066947903725e-06, + "loss": 0.9161, + "step": 302380 + }, + { + "epoch": 26.17, + "learning_rate": 6.385622236990378e-06, + "loss": 0.9232, + "step": 302390 + }, + { + "epoch": 26.17, + "learning_rate": 6.384177526077033e-06, + "loss": 0.8798, + "step": 302400 + }, + { + "epoch": 26.18, + "learning_rate": 6.382732815163686e-06, + "loss": 0.9216, + "step": 302410 + }, + { + "epoch": 26.18, + "learning_rate": 6.381288104250339e-06, + "loss": 0.9249, + "step": 302420 + }, + { + "epoch": 26.18, + "learning_rate": 6.379843393336994e-06, + "loss": 0.86, + "step": 302430 + }, + { + "epoch": 26.18, + "learning_rate": 6.378398682423648e-06, + "loss": 0.9331, + "step": 302440 + }, + { + "epoch": 26.18, + "learning_rate": 6.376953971510301e-06, + "loss": 0.8981, + "step": 302450 + }, + { + "epoch": 26.18, + "learning_rate": 6.375509260596955e-06, + "loss": 0.9124, + "step": 302460 + }, + { + "epoch": 26.18, + "learning_rate": 6.374064549683609e-06, + "loss": 0.9126, + "step": 302470 + }, + { + "epoch": 26.18, + "learning_rate": 6.372619838770262e-06, + "loss": 0.9673, + "step": 302480 + }, + { + "epoch": 26.18, + "learning_rate": 6.371175127856917e-06, + "loss": 0.9061, + "step": 302490 + }, + { + "epoch": 26.18, + "learning_rate": 6.36973041694357e-06, + "loss": 0.8662, + "step": 302500 + }, + { + "epoch": 26.18, + "learning_rate": 6.368285706030223e-06, + "loss": 0.87, + "step": 302510 + }, + { + "epoch": 26.19, + "learning_rate": 6.366840995116878e-06, + "loss": 0.8809, + "step": 302520 + }, + { + "epoch": 26.19, + "learning_rate": 6.365396284203532e-06, + "loss": 0.9343, + "step": 302530 + }, + { + "epoch": 26.19, + "learning_rate": 6.363951573290185e-06, + "loss": 0.8856, + "step": 302540 + }, + { + "epoch": 26.19, + "learning_rate": 6.362506862376838e-06, + "loss": 0.9245, + "step": 302550 + }, + { + "epoch": 26.19, + "learning_rate": 6.361062151463493e-06, + "loss": 0.9255, + "step": 302560 + }, + { + "epoch": 26.19, + "learning_rate": 6.359617440550146e-06, + "loss": 0.9221, + "step": 302570 + }, + { + "epoch": 26.19, + "learning_rate": 6.358172729636799e-06, + "loss": 0.8893, + "step": 302580 + }, + { + "epoch": 26.19, + "learning_rate": 6.356728018723454e-06, + "loss": 0.855, + "step": 302590 + }, + { + "epoch": 26.19, + "learning_rate": 6.355283307810107e-06, + "loss": 0.9091, + "step": 302600 + }, + { + "epoch": 26.19, + "learning_rate": 6.3538385968967604e-06, + "loss": 0.8795, + "step": 302610 + }, + { + "epoch": 26.19, + "learning_rate": 6.3523938859834155e-06, + "loss": 0.9113, + "step": 302620 + }, + { + "epoch": 26.19, + "learning_rate": 6.350949175070069e-06, + "loss": 0.9115, + "step": 302630 + }, + { + "epoch": 26.2, + "learning_rate": 6.349504464156722e-06, + "loss": 0.9503, + "step": 302640 + }, + { + "epoch": 26.2, + "learning_rate": 6.3480597532433766e-06, + "loss": 0.9132, + "step": 302650 + }, + { + "epoch": 26.2, + "learning_rate": 6.34661504233003e-06, + "loss": 0.8899, + "step": 302660 + }, + { + "epoch": 26.2, + "learning_rate": 6.345170331416683e-06, + "loss": 0.8739, + "step": 302670 + }, + { + "epoch": 26.2, + "learning_rate": 6.343725620503338e-06, + "loss": 0.8984, + "step": 302680 + }, + { + "epoch": 26.2, + "learning_rate": 6.342280909589991e-06, + "loss": 0.903, + "step": 302690 + }, + { + "epoch": 26.2, + "learning_rate": 6.340836198676644e-06, + "loss": 0.9073, + "step": 302700 + }, + { + "epoch": 26.2, + "learning_rate": 6.3393914877632995e-06, + "loss": 0.9154, + "step": 302710 + }, + { + "epoch": 26.2, + "learning_rate": 6.337946776849953e-06, + "loss": 0.9003, + "step": 302720 + }, + { + "epoch": 26.2, + "learning_rate": 6.336502065936606e-06, + "loss": 0.8779, + "step": 302730 + }, + { + "epoch": 26.2, + "learning_rate": 6.3350573550232605e-06, + "loss": 0.8782, + "step": 302740 + }, + { + "epoch": 26.21, + "learning_rate": 6.333612644109914e-06, + "loss": 0.9335, + "step": 302750 + }, + { + "epoch": 26.21, + "learning_rate": 6.332167933196567e-06, + "loss": 0.9856, + "step": 302760 + }, + { + "epoch": 26.21, + "learning_rate": 6.3307232222832215e-06, + "loss": 0.9532, + "step": 302770 + }, + { + "epoch": 26.21, + "learning_rate": 6.329278511369875e-06, + "loss": 0.9779, + "step": 302780 + }, + { + "epoch": 26.21, + "learning_rate": 6.327833800456528e-06, + "loss": 0.8489, + "step": 302790 + }, + { + "epoch": 26.21, + "learning_rate": 6.326389089543183e-06, + "loss": 0.9097, + "step": 302800 + }, + { + "epoch": 26.21, + "learning_rate": 6.324944378629837e-06, + "loss": 0.9007, + "step": 302810 + }, + { + "epoch": 26.21, + "learning_rate": 6.32349966771649e-06, + "loss": 0.8774, + "step": 302820 + }, + { + "epoch": 26.21, + "learning_rate": 6.3220549568031444e-06, + "loss": 0.9326, + "step": 302830 + }, + { + "epoch": 26.21, + "learning_rate": 6.320610245889798e-06, + "loss": 0.919, + "step": 302840 + }, + { + "epoch": 26.21, + "learning_rate": 6.319165534976451e-06, + "loss": 0.9311, + "step": 302850 + }, + { + "epoch": 26.21, + "learning_rate": 6.3177208240631055e-06, + "loss": 0.9429, + "step": 302860 + }, + { + "epoch": 26.22, + "learning_rate": 6.316276113149759e-06, + "loss": 0.9641, + "step": 302870 + }, + { + "epoch": 26.22, + "learning_rate": 6.314831402236412e-06, + "loss": 0.8526, + "step": 302880 + }, + { + "epoch": 26.22, + "learning_rate": 6.313386691323067e-06, + "loss": 0.9053, + "step": 302890 + }, + { + "epoch": 26.22, + "learning_rate": 6.311941980409721e-06, + "loss": 0.8738, + "step": 302900 + }, + { + "epoch": 26.22, + "learning_rate": 6.310497269496374e-06, + "loss": 0.9666, + "step": 302910 + }, + { + "epoch": 26.22, + "learning_rate": 6.309052558583028e-06, + "loss": 0.9771, + "step": 302920 + }, + { + "epoch": 26.22, + "learning_rate": 6.307607847669682e-06, + "loss": 0.8882, + "step": 302930 + }, + { + "epoch": 26.22, + "learning_rate": 6.306163136756335e-06, + "loss": 0.9212, + "step": 302940 + }, + { + "epoch": 26.22, + "learning_rate": 6.304718425842989e-06, + "loss": 0.855, + "step": 302950 + }, + { + "epoch": 26.22, + "learning_rate": 6.303273714929643e-06, + "loss": 0.8857, + "step": 302960 + }, + { + "epoch": 26.22, + "learning_rate": 6.301829004016296e-06, + "loss": 0.9432, + "step": 302970 + }, + { + "epoch": 26.23, + "learning_rate": 6.300384293102951e-06, + "loss": 0.9164, + "step": 302980 + }, + { + "epoch": 26.23, + "learning_rate": 6.298939582189605e-06, + "loss": 0.9169, + "step": 302990 + }, + { + "epoch": 26.23, + "learning_rate": 6.297494871276258e-06, + "loss": 0.884, + "step": 303000 + }, + { + "epoch": 26.23, + "learning_rate": 6.296050160362911e-06, + "loss": 0.892, + "step": 303010 + }, + { + "epoch": 26.23, + "learning_rate": 6.294605449449566e-06, + "loss": 0.8456, + "step": 303020 + }, + { + "epoch": 26.23, + "learning_rate": 6.293160738536219e-06, + "loss": 0.9099, + "step": 303030 + }, + { + "epoch": 26.23, + "learning_rate": 6.2917160276228725e-06, + "loss": 0.9022, + "step": 303040 + }, + { + "epoch": 26.23, + "learning_rate": 6.290271316709527e-06, + "loss": 0.8778, + "step": 303050 + }, + { + "epoch": 26.23, + "learning_rate": 6.28882660579618e-06, + "loss": 0.9165, + "step": 303060 + }, + { + "epoch": 26.23, + "learning_rate": 6.2873818948828335e-06, + "loss": 0.888, + "step": 303070 + }, + { + "epoch": 26.23, + "learning_rate": 6.285937183969489e-06, + "loss": 0.9137, + "step": 303080 + }, + { + "epoch": 26.23, + "learning_rate": 6.284492473056142e-06, + "loss": 0.9609, + "step": 303090 + }, + { + "epoch": 26.24, + "learning_rate": 6.2830477621427946e-06, + "loss": 0.9159, + "step": 303100 + }, + { + "epoch": 26.24, + "learning_rate": 6.28160305122945e-06, + "loss": 0.8852, + "step": 303110 + }, + { + "epoch": 26.24, + "learning_rate": 6.280158340316103e-06, + "loss": 0.8972, + "step": 303120 + }, + { + "epoch": 26.24, + "learning_rate": 6.2787136294027565e-06, + "loss": 0.7904, + "step": 303130 + }, + { + "epoch": 26.24, + "learning_rate": 6.277268918489411e-06, + "loss": 0.8636, + "step": 303140 + }, + { + "epoch": 26.24, + "learning_rate": 6.275824207576064e-06, + "loss": 0.9787, + "step": 303150 + }, + { + "epoch": 26.24, + "learning_rate": 6.2743794966627175e-06, + "loss": 0.9049, + "step": 303160 + }, + { + "epoch": 26.24, + "learning_rate": 6.2729347857493726e-06, + "loss": 0.9454, + "step": 303170 + }, + { + "epoch": 26.24, + "learning_rate": 6.271490074836026e-06, + "loss": 0.9432, + "step": 303180 + }, + { + "epoch": 26.24, + "learning_rate": 6.2700453639226785e-06, + "loss": 0.8994, + "step": 303190 + }, + { + "epoch": 26.24, + "learning_rate": 6.268600653009334e-06, + "loss": 0.9639, + "step": 303200 + }, + { + "epoch": 26.25, + "learning_rate": 6.267155942095987e-06, + "loss": 0.8756, + "step": 303210 + }, + { + "epoch": 26.25, + "learning_rate": 6.26571123118264e-06, + "loss": 0.867, + "step": 303220 + }, + { + "epoch": 26.25, + "learning_rate": 6.264266520269295e-06, + "loss": 0.9307, + "step": 303230 + }, + { + "epoch": 26.25, + "learning_rate": 6.262821809355948e-06, + "loss": 0.9122, + "step": 303240 + }, + { + "epoch": 26.25, + "learning_rate": 6.2613770984426014e-06, + "loss": 0.8323, + "step": 303250 + }, + { + "epoch": 26.25, + "learning_rate": 6.2599323875292565e-06, + "loss": 0.8769, + "step": 303260 + }, + { + "epoch": 26.25, + "learning_rate": 6.25848767661591e-06, + "loss": 0.9079, + "step": 303270 + }, + { + "epoch": 26.25, + "learning_rate": 6.2570429657025625e-06, + "loss": 0.9157, + "step": 303280 + }, + { + "epoch": 26.25, + "learning_rate": 6.2555982547892175e-06, + "loss": 0.9128, + "step": 303290 + }, + { + "epoch": 26.25, + "learning_rate": 6.254153543875871e-06, + "loss": 0.9321, + "step": 303300 + }, + { + "epoch": 26.25, + "learning_rate": 6.252708832962524e-06, + "loss": 0.8914, + "step": 303310 + }, + { + "epoch": 26.25, + "learning_rate": 6.251264122049179e-06, + "loss": 0.939, + "step": 303320 + }, + { + "epoch": 26.26, + "learning_rate": 6.249819411135832e-06, + "loss": 0.9174, + "step": 303330 + }, + { + "epoch": 26.26, + "learning_rate": 6.248374700222485e-06, + "loss": 0.9014, + "step": 303340 + }, + { + "epoch": 26.26, + "learning_rate": 6.24692998930914e-06, + "loss": 0.8721, + "step": 303350 + }, + { + "epoch": 26.26, + "learning_rate": 6.245485278395794e-06, + "loss": 0.9079, + "step": 303360 + }, + { + "epoch": 26.26, + "learning_rate": 6.244040567482446e-06, + "loss": 0.8493, + "step": 303370 + }, + { + "epoch": 26.26, + "learning_rate": 6.242595856569101e-06, + "loss": 0.8885, + "step": 303380 + }, + { + "epoch": 26.26, + "learning_rate": 6.241151145655755e-06, + "loss": 0.8471, + "step": 303390 + }, + { + "epoch": 26.26, + "learning_rate": 6.239706434742408e-06, + "loss": 0.8924, + "step": 303400 + }, + { + "epoch": 26.26, + "learning_rate": 6.238261723829062e-06, + "loss": 0.9267, + "step": 303410 + }, + { + "epoch": 26.26, + "learning_rate": 6.236817012915716e-06, + "loss": 0.8475, + "step": 303420 + }, + { + "epoch": 26.26, + "learning_rate": 6.235372302002369e-06, + "loss": 0.9512, + "step": 303430 + }, + { + "epoch": 26.26, + "learning_rate": 6.2339275910890236e-06, + "loss": 0.86, + "step": 303440 + }, + { + "epoch": 26.27, + "learning_rate": 6.232482880175678e-06, + "loss": 0.9082, + "step": 303450 + }, + { + "epoch": 26.27, + "learning_rate": 6.23103816926233e-06, + "loss": 0.9155, + "step": 303460 + }, + { + "epoch": 26.27, + "learning_rate": 6.229593458348985e-06, + "loss": 0.9229, + "step": 303470 + }, + { + "epoch": 26.27, + "learning_rate": 6.228148747435639e-06, + "loss": 0.8718, + "step": 303480 + }, + { + "epoch": 26.27, + "learning_rate": 6.226704036522292e-06, + "loss": 0.865, + "step": 303490 + }, + { + "epoch": 26.27, + "learning_rate": 6.225259325608946e-06, + "loss": 0.8693, + "step": 303500 + }, + { + "epoch": 26.27, + "learning_rate": 6.2238146146956e-06, + "loss": 0.95, + "step": 303510 + }, + { + "epoch": 26.27, + "learning_rate": 6.222369903782253e-06, + "loss": 0.8919, + "step": 303520 + }, + { + "epoch": 26.27, + "learning_rate": 6.2209251928689075e-06, + "loss": 0.8691, + "step": 303530 + }, + { + "epoch": 26.27, + "learning_rate": 6.219480481955562e-06, + "loss": 0.8552, + "step": 303540 + }, + { + "epoch": 26.27, + "learning_rate": 6.218035771042214e-06, + "loss": 0.8647, + "step": 303550 + }, + { + "epoch": 26.28, + "learning_rate": 6.2165910601288685e-06, + "loss": 0.8728, + "step": 303560 + }, + { + "epoch": 26.28, + "learning_rate": 6.215146349215522e-06, + "loss": 0.9891, + "step": 303570 + }, + { + "epoch": 26.28, + "learning_rate": 6.213701638302176e-06, + "loss": 0.9435, + "step": 303580 + }, + { + "epoch": 26.28, + "learning_rate": 6.2122569273888296e-06, + "loss": 0.8791, + "step": 303590 + }, + { + "epoch": 26.28, + "learning_rate": 6.210812216475483e-06, + "loss": 0.8816, + "step": 303600 + }, + { + "epoch": 26.28, + "learning_rate": 6.209367505562137e-06, + "loss": 0.8554, + "step": 303610 + }, + { + "epoch": 26.28, + "learning_rate": 6.2079227946487914e-06, + "loss": 0.8657, + "step": 303620 + }, + { + "epoch": 26.28, + "learning_rate": 6.206478083735445e-06, + "loss": 0.8874, + "step": 303630 + }, + { + "epoch": 26.28, + "learning_rate": 6.205033372822098e-06, + "loss": 0.9589, + "step": 303640 + }, + { + "epoch": 26.28, + "learning_rate": 6.2035886619087525e-06, + "loss": 0.9231, + "step": 303650 + }, + { + "epoch": 26.28, + "learning_rate": 6.202143950995406e-06, + "loss": 0.9257, + "step": 303660 + }, + { + "epoch": 26.28, + "learning_rate": 6.20069924008206e-06, + "loss": 0.9081, + "step": 303670 + }, + { + "epoch": 26.29, + "learning_rate": 6.1992545291687135e-06, + "loss": 0.9175, + "step": 303680 + }, + { + "epoch": 26.29, + "learning_rate": 6.197809818255367e-06, + "loss": 0.8172, + "step": 303690 + }, + { + "epoch": 26.29, + "learning_rate": 6.196365107342021e-06, + "loss": 0.9307, + "step": 303700 + }, + { + "epoch": 26.29, + "learning_rate": 6.194920396428675e-06, + "loss": 0.9471, + "step": 303710 + }, + { + "epoch": 26.29, + "learning_rate": 6.193475685515329e-06, + "loss": 0.9035, + "step": 303720 + }, + { + "epoch": 26.29, + "learning_rate": 6.192030974601982e-06, + "loss": 0.8489, + "step": 303730 + }, + { + "epoch": 26.29, + "learning_rate": 6.190586263688636e-06, + "loss": 0.9261, + "step": 303740 + }, + { + "epoch": 26.29, + "learning_rate": 6.18914155277529e-06, + "loss": 0.8876, + "step": 303750 + }, + { + "epoch": 26.29, + "learning_rate": 6.187696841861944e-06, + "loss": 0.9042, + "step": 303760 + }, + { + "epoch": 26.29, + "learning_rate": 6.1862521309485974e-06, + "loss": 0.9078, + "step": 303770 + }, + { + "epoch": 26.29, + "learning_rate": 6.184807420035251e-06, + "loss": 0.8463, + "step": 303780 + }, + { + "epoch": 26.3, + "learning_rate": 6.183362709121905e-06, + "loss": 0.8805, + "step": 303790 + }, + { + "epoch": 26.3, + "learning_rate": 6.1819179982085585e-06, + "loss": 0.8943, + "step": 303800 + }, + { + "epoch": 26.3, + "learning_rate": 6.180473287295213e-06, + "loss": 0.8497, + "step": 303810 + }, + { + "epoch": 26.3, + "learning_rate": 6.179028576381866e-06, + "loss": 0.8701, + "step": 303820 + }, + { + "epoch": 26.3, + "learning_rate": 6.1775838654685195e-06, + "loss": 0.9254, + "step": 303830 + }, + { + "epoch": 26.3, + "learning_rate": 6.176139154555174e-06, + "loss": 0.8911, + "step": 303840 + }, + { + "epoch": 26.3, + "learning_rate": 6.174694443641828e-06, + "loss": 0.9551, + "step": 303850 + }, + { + "epoch": 26.3, + "learning_rate": 6.173249732728481e-06, + "loss": 0.9057, + "step": 303860 + }, + { + "epoch": 26.3, + "learning_rate": 6.171805021815135e-06, + "loss": 0.8621, + "step": 303870 + }, + { + "epoch": 26.3, + "learning_rate": 6.170360310901789e-06, + "loss": 0.9054, + "step": 303880 + }, + { + "epoch": 26.3, + "learning_rate": 6.168915599988442e-06, + "loss": 0.8958, + "step": 303890 + }, + { + "epoch": 26.3, + "learning_rate": 6.167470889075097e-06, + "loss": 0.9224, + "step": 303900 + }, + { + "epoch": 26.31, + "learning_rate": 6.16602617816175e-06, + "loss": 0.8829, + "step": 303910 + }, + { + "epoch": 26.31, + "learning_rate": 6.1645814672484035e-06, + "loss": 0.889, + "step": 303920 + }, + { + "epoch": 26.31, + "learning_rate": 6.163136756335058e-06, + "loss": 0.8853, + "step": 303930 + }, + { + "epoch": 26.31, + "learning_rate": 6.161692045421712e-06, + "loss": 0.9465, + "step": 303940 + }, + { + "epoch": 26.31, + "learning_rate": 6.160247334508365e-06, + "loss": 0.8967, + "step": 303950 + }, + { + "epoch": 26.31, + "learning_rate": 6.158802623595019e-06, + "loss": 0.9614, + "step": 303960 + }, + { + "epoch": 26.31, + "learning_rate": 6.157357912681673e-06, + "loss": 0.9042, + "step": 303970 + }, + { + "epoch": 26.31, + "learning_rate": 6.155913201768326e-06, + "loss": 0.9286, + "step": 303980 + }, + { + "epoch": 26.31, + "learning_rate": 6.154468490854981e-06, + "loss": 0.9061, + "step": 303990 + }, + { + "epoch": 26.31, + "learning_rate": 6.153023779941634e-06, + "loss": 0.8668, + "step": 304000 + }, + { + "epoch": 26.31, + "learning_rate": 6.151579069028287e-06, + "loss": 0.9742, + "step": 304010 + }, + { + "epoch": 26.32, + "learning_rate": 6.150134358114942e-06, + "loss": 0.8704, + "step": 304020 + }, + { + "epoch": 26.32, + "learning_rate": 6.148689647201595e-06, + "loss": 0.888, + "step": 304030 + }, + { + "epoch": 26.32, + "learning_rate": 6.147244936288249e-06, + "loss": 0.8809, + "step": 304040 + }, + { + "epoch": 26.32, + "learning_rate": 6.145800225374903e-06, + "loss": 0.861, + "step": 304050 + }, + { + "epoch": 26.32, + "learning_rate": 6.144355514461556e-06, + "loss": 0.9057, + "step": 304060 + }, + { + "epoch": 26.32, + "learning_rate": 6.14291080354821e-06, + "loss": 0.9041, + "step": 304070 + }, + { + "epoch": 26.32, + "learning_rate": 6.1414660926348645e-06, + "loss": 0.9259, + "step": 304080 + }, + { + "epoch": 26.32, + "learning_rate": 6.140021381721518e-06, + "loss": 0.9109, + "step": 304090 + }, + { + "epoch": 26.32, + "learning_rate": 6.138576670808171e-06, + "loss": 0.8498, + "step": 304100 + }, + { + "epoch": 26.32, + "learning_rate": 6.137131959894826e-06, + "loss": 0.9155, + "step": 304110 + }, + { + "epoch": 26.32, + "learning_rate": 6.135687248981479e-06, + "loss": 0.9235, + "step": 304120 + }, + { + "epoch": 26.32, + "learning_rate": 6.134242538068133e-06, + "loss": 0.9281, + "step": 304130 + }, + { + "epoch": 26.33, + "learning_rate": 6.132797827154787e-06, + "loss": 0.9237, + "step": 304140 + }, + { + "epoch": 26.33, + "learning_rate": 6.13135311624144e-06, + "loss": 0.9409, + "step": 304150 + }, + { + "epoch": 26.33, + "learning_rate": 6.129908405328094e-06, + "loss": 0.9046, + "step": 304160 + }, + { + "epoch": 26.33, + "learning_rate": 6.1284636944147485e-06, + "loss": 0.9773, + "step": 304170 + }, + { + "epoch": 26.33, + "learning_rate": 6.127018983501402e-06, + "loss": 0.8798, + "step": 304180 + }, + { + "epoch": 26.33, + "learning_rate": 6.125574272588055e-06, + "loss": 0.9172, + "step": 304190 + }, + { + "epoch": 26.33, + "learning_rate": 6.1241295616747095e-06, + "loss": 0.9311, + "step": 304200 + }, + { + "epoch": 26.33, + "learning_rate": 6.122684850761363e-06, + "loss": 0.89, + "step": 304210 + }, + { + "epoch": 26.33, + "learning_rate": 6.121240139848017e-06, + "loss": 0.9071, + "step": 304220 + }, + { + "epoch": 26.33, + "learning_rate": 6.1197954289346706e-06, + "loss": 0.8904, + "step": 304230 + }, + { + "epoch": 26.33, + "learning_rate": 6.118350718021324e-06, + "loss": 0.9351, + "step": 304240 + }, + { + "epoch": 26.34, + "learning_rate": 6.116906007107978e-06, + "loss": 0.9156, + "step": 304250 + }, + { + "epoch": 26.34, + "learning_rate": 6.115461296194632e-06, + "loss": 0.9148, + "step": 304260 + }, + { + "epoch": 26.34, + "learning_rate": 6.114016585281286e-06, + "loss": 0.8736, + "step": 304270 + }, + { + "epoch": 26.34, + "learning_rate": 6.112571874367939e-06, + "loss": 0.9458, + "step": 304280 + }, + { + "epoch": 26.34, + "learning_rate": 6.111127163454593e-06, + "loss": 0.8462, + "step": 304290 + }, + { + "epoch": 26.34, + "learning_rate": 6.109682452541247e-06, + "loss": 0.943, + "step": 304300 + }, + { + "epoch": 26.34, + "learning_rate": 6.108237741627901e-06, + "loss": 0.9634, + "step": 304310 + }, + { + "epoch": 26.34, + "learning_rate": 6.106793030714554e-06, + "loss": 0.9113, + "step": 304320 + }, + { + "epoch": 26.34, + "learning_rate": 6.105348319801208e-06, + "loss": 0.9152, + "step": 304330 + }, + { + "epoch": 26.34, + "learning_rate": 6.103903608887862e-06, + "loss": 0.8993, + "step": 304340 + }, + { + "epoch": 26.34, + "learning_rate": 6.1024588979745155e-06, + "loss": 0.9691, + "step": 304350 + }, + { + "epoch": 26.34, + "learning_rate": 6.10101418706117e-06, + "loss": 0.8958, + "step": 304360 + }, + { + "epoch": 26.35, + "learning_rate": 6.099569476147823e-06, + "loss": 0.8879, + "step": 304370 + }, + { + "epoch": 26.35, + "learning_rate": 6.0981247652344766e-06, + "loss": 0.9626, + "step": 304380 + }, + { + "epoch": 26.35, + "learning_rate": 6.096680054321131e-06, + "loss": 0.9128, + "step": 304390 + }, + { + "epoch": 26.35, + "learning_rate": 6.095235343407785e-06, + "loss": 0.89, + "step": 304400 + }, + { + "epoch": 26.35, + "learning_rate": 6.093790632494438e-06, + "loss": 0.8722, + "step": 304410 + }, + { + "epoch": 26.35, + "learning_rate": 6.092345921581092e-06, + "loss": 0.8629, + "step": 304420 + }, + { + "epoch": 26.35, + "learning_rate": 6.090901210667746e-06, + "loss": 0.8994, + "step": 304430 + }, + { + "epoch": 26.35, + "learning_rate": 6.0894564997543995e-06, + "loss": 0.9151, + "step": 304440 + }, + { + "epoch": 26.35, + "learning_rate": 6.088011788841054e-06, + "loss": 0.8448, + "step": 304450 + }, + { + "epoch": 26.35, + "learning_rate": 6.086567077927706e-06, + "loss": 0.8528, + "step": 304460 + }, + { + "epoch": 26.35, + "learning_rate": 6.0851223670143605e-06, + "loss": 0.8805, + "step": 304470 + }, + { + "epoch": 26.35, + "learning_rate": 6.083677656101015e-06, + "loss": 0.9513, + "step": 304480 + }, + { + "epoch": 26.36, + "learning_rate": 6.082232945187668e-06, + "loss": 0.9183, + "step": 304490 + }, + { + "epoch": 26.36, + "learning_rate": 6.0807882342743215e-06, + "loss": 0.8987, + "step": 304500 + }, + { + "epoch": 26.36, + "learning_rate": 6.079343523360976e-06, + "loss": 0.9325, + "step": 304510 + }, + { + "epoch": 26.36, + "learning_rate": 6.077898812447629e-06, + "loss": 0.8987, + "step": 304520 + }, + { + "epoch": 26.36, + "learning_rate": 6.076454101534283e-06, + "loss": 0.8945, + "step": 304530 + }, + { + "epoch": 26.36, + "learning_rate": 6.075009390620938e-06, + "loss": 0.8837, + "step": 304540 + }, + { + "epoch": 26.36, + "learning_rate": 6.07356467970759e-06, + "loss": 0.8824, + "step": 304550 + }, + { + "epoch": 26.36, + "learning_rate": 6.0721199687942444e-06, + "loss": 0.857, + "step": 304560 + }, + { + "epoch": 26.36, + "learning_rate": 6.070675257880899e-06, + "loss": 0.8881, + "step": 304570 + }, + { + "epoch": 26.36, + "learning_rate": 6.069230546967552e-06, + "loss": 0.8868, + "step": 304580 + }, + { + "epoch": 26.36, + "learning_rate": 6.0677858360542055e-06, + "loss": 0.8449, + "step": 304590 + }, + { + "epoch": 26.37, + "learning_rate": 6.06634112514086e-06, + "loss": 0.9271, + "step": 304600 + }, + { + "epoch": 26.37, + "learning_rate": 6.064896414227513e-06, + "loss": 0.8726, + "step": 304610 + }, + { + "epoch": 26.37, + "learning_rate": 6.063451703314167e-06, + "loss": 0.8694, + "step": 304620 + }, + { + "epoch": 26.37, + "learning_rate": 6.062006992400822e-06, + "loss": 0.8719, + "step": 304630 + }, + { + "epoch": 26.37, + "learning_rate": 6.060562281487474e-06, + "loss": 0.9062, + "step": 304640 + }, + { + "epoch": 26.37, + "learning_rate": 6.059117570574128e-06, + "loss": 0.907, + "step": 304650 + }, + { + "epoch": 26.37, + "learning_rate": 6.057672859660783e-06, + "loss": 0.8439, + "step": 304660 + }, + { + "epoch": 26.37, + "learning_rate": 6.056228148747436e-06, + "loss": 0.9568, + "step": 304670 + }, + { + "epoch": 26.37, + "learning_rate": 6.054783437834089e-06, + "loss": 0.8901, + "step": 304680 + }, + { + "epoch": 26.37, + "learning_rate": 6.053338726920743e-06, + "loss": 0.9352, + "step": 304690 + }, + { + "epoch": 26.37, + "learning_rate": 6.051894016007397e-06, + "loss": 0.8524, + "step": 304700 + }, + { + "epoch": 26.37, + "learning_rate": 6.050449305094051e-06, + "loss": 0.9041, + "step": 304710 + }, + { + "epoch": 26.38, + "learning_rate": 6.049004594180705e-06, + "loss": 0.8925, + "step": 304720 + }, + { + "epoch": 26.38, + "learning_rate": 6.047559883267358e-06, + "loss": 0.9231, + "step": 304730 + }, + { + "epoch": 26.38, + "learning_rate": 6.046115172354012e-06, + "loss": 0.8588, + "step": 304740 + }, + { + "epoch": 26.38, + "learning_rate": 6.044670461440666e-06, + "loss": 0.9055, + "step": 304750 + }, + { + "epoch": 26.38, + "learning_rate": 6.04322575052732e-06, + "loss": 0.9364, + "step": 304760 + }, + { + "epoch": 26.38, + "learning_rate": 6.041781039613973e-06, + "loss": 0.9357, + "step": 304770 + }, + { + "epoch": 26.38, + "learning_rate": 6.040336328700627e-06, + "loss": 0.9271, + "step": 304780 + }, + { + "epoch": 26.38, + "learning_rate": 6.038891617787281e-06, + "loss": 0.8646, + "step": 304790 + }, + { + "epoch": 26.38, + "learning_rate": 6.037446906873935e-06, + "loss": 0.9482, + "step": 304800 + }, + { + "epoch": 26.38, + "learning_rate": 6.036002195960589e-06, + "loss": 0.9067, + "step": 304810 + }, + { + "epoch": 26.38, + "learning_rate": 6.034557485047242e-06, + "loss": 0.8944, + "step": 304820 + }, + { + "epoch": 26.39, + "learning_rate": 6.033112774133896e-06, + "loss": 0.8897, + "step": 304830 + }, + { + "epoch": 26.39, + "learning_rate": 6.03166806322055e-06, + "loss": 0.9076, + "step": 304840 + }, + { + "epoch": 26.39, + "learning_rate": 6.030223352307204e-06, + "loss": 0.8981, + "step": 304850 + }, + { + "epoch": 26.39, + "learning_rate": 6.028778641393857e-06, + "loss": 0.9264, + "step": 304860 + }, + { + "epoch": 26.39, + "learning_rate": 6.027333930480511e-06, + "loss": 0.8988, + "step": 304870 + }, + { + "epoch": 26.39, + "learning_rate": 6.025889219567165e-06, + "loss": 0.9256, + "step": 304880 + }, + { + "epoch": 26.39, + "learning_rate": 6.024444508653819e-06, + "loss": 0.8788, + "step": 304890 + }, + { + "epoch": 26.39, + "learning_rate": 6.022999797740473e-06, + "loss": 0.9045, + "step": 304900 + }, + { + "epoch": 26.39, + "learning_rate": 6.021555086827126e-06, + "loss": 0.8325, + "step": 304910 + }, + { + "epoch": 26.39, + "learning_rate": 6.020110375913779e-06, + "loss": 0.8474, + "step": 304920 + }, + { + "epoch": 26.39, + "learning_rate": 6.018665665000434e-06, + "loss": 0.8995, + "step": 304930 + }, + { + "epoch": 26.39, + "learning_rate": 6.017220954087088e-06, + "loss": 0.8876, + "step": 304940 + }, + { + "epoch": 26.4, + "learning_rate": 6.015776243173741e-06, + "loss": 0.9515, + "step": 304950 + }, + { + "epoch": 26.4, + "learning_rate": 6.014331532260395e-06, + "loss": 0.9181, + "step": 304960 + }, + { + "epoch": 26.4, + "learning_rate": 6.012886821347049e-06, + "loss": 0.8939, + "step": 304970 + }, + { + "epoch": 26.4, + "learning_rate": 6.011442110433702e-06, + "loss": 0.9505, + "step": 304980 + }, + { + "epoch": 26.4, + "learning_rate": 6.0099973995203565e-06, + "loss": 0.8807, + "step": 304990 + }, + { + "epoch": 26.4, + "learning_rate": 6.00855268860701e-06, + "loss": 0.8603, + "step": 305000 + }, + { + "epoch": 26.4, + "learning_rate": 6.007107977693663e-06, + "loss": 0.9225, + "step": 305010 + }, + { + "epoch": 26.4, + "learning_rate": 6.0056632667803176e-06, + "loss": 0.8843, + "step": 305020 + }, + { + "epoch": 26.4, + "learning_rate": 6.004218555866972e-06, + "loss": 0.9238, + "step": 305030 + }, + { + "epoch": 26.4, + "learning_rate": 6.002773844953625e-06, + "loss": 0.8812, + "step": 305040 + }, + { + "epoch": 26.4, + "learning_rate": 6.001329134040279e-06, + "loss": 0.8047, + "step": 305050 + }, + { + "epoch": 26.41, + "learning_rate": 5.999884423126933e-06, + "loss": 0.924, + "step": 305060 + }, + { + "epoch": 26.41, + "learning_rate": 5.998439712213586e-06, + "loss": 0.962, + "step": 305070 + }, + { + "epoch": 26.41, + "learning_rate": 5.9969950013002405e-06, + "loss": 0.9037, + "step": 305080 + }, + { + "epoch": 26.41, + "learning_rate": 5.995550290386894e-06, + "loss": 0.9244, + "step": 305090 + }, + { + "epoch": 26.41, + "learning_rate": 5.994105579473547e-06, + "loss": 0.9001, + "step": 305100 + }, + { + "epoch": 26.41, + "learning_rate": 5.9926608685602015e-06, + "loss": 0.9276, + "step": 305110 + }, + { + "epoch": 26.41, + "learning_rate": 5.991216157646856e-06, + "loss": 0.914, + "step": 305120 + }, + { + "epoch": 26.41, + "learning_rate": 5.989771446733509e-06, + "loss": 0.934, + "step": 305130 + }, + { + "epoch": 26.41, + "learning_rate": 5.9883267358201625e-06, + "loss": 0.9573, + "step": 305140 + }, + { + "epoch": 26.41, + "learning_rate": 5.986882024906816e-06, + "loss": 0.9335, + "step": 305150 + }, + { + "epoch": 26.41, + "learning_rate": 5.98543731399347e-06, + "loss": 0.883, + "step": 305160 + }, + { + "epoch": 26.41, + "learning_rate": 5.983992603080124e-06, + "loss": 0.8897, + "step": 305170 + }, + { + "epoch": 26.42, + "learning_rate": 5.982547892166778e-06, + "loss": 0.9436, + "step": 305180 + }, + { + "epoch": 26.42, + "learning_rate": 5.981103181253431e-06, + "loss": 0.8816, + "step": 305190 + }, + { + "epoch": 26.42, + "learning_rate": 5.9796584703400854e-06, + "loss": 0.9266, + "step": 305200 + }, + { + "epoch": 26.42, + "learning_rate": 5.978213759426739e-06, + "loss": 0.9321, + "step": 305210 + }, + { + "epoch": 26.42, + "learning_rate": 5.976769048513393e-06, + "loss": 0.9038, + "step": 305220 + }, + { + "epoch": 26.42, + "learning_rate": 5.9753243376000465e-06, + "loss": 0.918, + "step": 305230 + }, + { + "epoch": 26.42, + "learning_rate": 5.9738796266867e-06, + "loss": 0.8969, + "step": 305240 + }, + { + "epoch": 26.42, + "learning_rate": 5.972434915773354e-06, + "loss": 0.8306, + "step": 305250 + }, + { + "epoch": 26.42, + "learning_rate": 5.970990204860008e-06, + "loss": 0.9424, + "step": 305260 + }, + { + "epoch": 26.42, + "learning_rate": 5.969545493946662e-06, + "loss": 0.9117, + "step": 305270 + }, + { + "epoch": 26.42, + "learning_rate": 5.968100783033315e-06, + "loss": 0.9312, + "step": 305280 + }, + { + "epoch": 26.43, + "learning_rate": 5.966656072119969e-06, + "loss": 0.8629, + "step": 305290 + }, + { + "epoch": 26.43, + "learning_rate": 5.965211361206623e-06, + "loss": 0.9482, + "step": 305300 + }, + { + "epoch": 26.43, + "learning_rate": 5.963766650293277e-06, + "loss": 0.9415, + "step": 305310 + }, + { + "epoch": 26.43, + "learning_rate": 5.96232193937993e-06, + "loss": 0.9376, + "step": 305320 + }, + { + "epoch": 26.43, + "learning_rate": 5.960877228466584e-06, + "loss": 0.9186, + "step": 305330 + }, + { + "epoch": 26.43, + "learning_rate": 5.959432517553238e-06, + "loss": 0.8971, + "step": 305340 + }, + { + "epoch": 26.43, + "learning_rate": 5.957987806639892e-06, + "loss": 0.9369, + "step": 305350 + }, + { + "epoch": 26.43, + "learning_rate": 5.956543095726546e-06, + "loss": 0.8821, + "step": 305360 + }, + { + "epoch": 26.43, + "learning_rate": 5.955098384813199e-06, + "loss": 0.9425, + "step": 305370 + }, + { + "epoch": 26.43, + "learning_rate": 5.9536536738998525e-06, + "loss": 0.9525, + "step": 305380 + }, + { + "epoch": 26.43, + "learning_rate": 5.952208962986507e-06, + "loss": 0.9075, + "step": 305390 + }, + { + "epoch": 26.43, + "learning_rate": 5.950764252073161e-06, + "loss": 0.9228, + "step": 305400 + }, + { + "epoch": 26.44, + "learning_rate": 5.9493195411598135e-06, + "loss": 0.9068, + "step": 305410 + }, + { + "epoch": 26.44, + "learning_rate": 5.947874830246468e-06, + "loss": 0.9652, + "step": 305420 + }, + { + "epoch": 26.44, + "learning_rate": 5.946430119333122e-06, + "loss": 0.8564, + "step": 305430 + }, + { + "epoch": 26.44, + "learning_rate": 5.944985408419775e-06, + "loss": 0.9025, + "step": 305440 + }, + { + "epoch": 26.44, + "learning_rate": 5.94354069750643e-06, + "loss": 0.9332, + "step": 305450 + }, + { + "epoch": 26.44, + "learning_rate": 5.942095986593083e-06, + "loss": 0.9048, + "step": 305460 + }, + { + "epoch": 26.44, + "learning_rate": 5.940651275679736e-06, + "loss": 0.9033, + "step": 305470 + }, + { + "epoch": 26.44, + "learning_rate": 5.939206564766391e-06, + "loss": 0.8988, + "step": 305480 + }, + { + "epoch": 26.44, + "learning_rate": 5.937761853853045e-06, + "loss": 0.9024, + "step": 305490 + }, + { + "epoch": 26.44, + "learning_rate": 5.9363171429396975e-06, + "loss": 0.8954, + "step": 305500 + }, + { + "epoch": 26.44, + "learning_rate": 5.934872432026352e-06, + "loss": 0.8731, + "step": 305510 + }, + { + "epoch": 26.44, + "learning_rate": 5.933427721113006e-06, + "loss": 0.9343, + "step": 305520 + }, + { + "epoch": 26.45, + "learning_rate": 5.931983010199659e-06, + "loss": 0.8857, + "step": 305530 + }, + { + "epoch": 26.45, + "learning_rate": 5.9305382992863136e-06, + "loss": 0.9081, + "step": 305540 + }, + { + "epoch": 26.45, + "learning_rate": 5.929093588372967e-06, + "loss": 0.8921, + "step": 305550 + }, + { + "epoch": 26.45, + "learning_rate": 5.92764887745962e-06, + "loss": 0.8964, + "step": 305560 + }, + { + "epoch": 26.45, + "learning_rate": 5.926204166546275e-06, + "loss": 0.8954, + "step": 305570 + }, + { + "epoch": 26.45, + "learning_rate": 5.924759455632929e-06, + "loss": 0.9859, + "step": 305580 + }, + { + "epoch": 26.45, + "learning_rate": 5.923314744719581e-06, + "loss": 0.8847, + "step": 305590 + }, + { + "epoch": 26.45, + "learning_rate": 5.921870033806236e-06, + "loss": 0.9206, + "step": 305600 + }, + { + "epoch": 26.45, + "learning_rate": 5.920425322892889e-06, + "loss": 0.8567, + "step": 305610 + }, + { + "epoch": 26.45, + "learning_rate": 5.918980611979543e-06, + "loss": 0.8877, + "step": 305620 + }, + { + "epoch": 26.45, + "learning_rate": 5.9175359010661975e-06, + "loss": 0.9181, + "step": 305630 + }, + { + "epoch": 26.46, + "learning_rate": 5.91609119015285e-06, + "loss": 0.9154, + "step": 305640 + }, + { + "epoch": 26.46, + "learning_rate": 5.914646479239504e-06, + "loss": 0.9404, + "step": 305650 + }, + { + "epoch": 26.46, + "learning_rate": 5.9132017683261586e-06, + "loss": 0.8826, + "step": 305660 + }, + { + "epoch": 26.46, + "learning_rate": 5.911757057412812e-06, + "loss": 0.9182, + "step": 305670 + }, + { + "epoch": 26.46, + "learning_rate": 5.910312346499465e-06, + "loss": 0.9113, + "step": 305680 + }, + { + "epoch": 26.46, + "learning_rate": 5.90886763558612e-06, + "loss": 0.9078, + "step": 305690 + }, + { + "epoch": 26.46, + "learning_rate": 5.907422924672773e-06, + "loss": 0.8936, + "step": 305700 + }, + { + "epoch": 26.46, + "learning_rate": 5.905978213759427e-06, + "loss": 0.9241, + "step": 305710 + }, + { + "epoch": 26.46, + "learning_rate": 5.904533502846081e-06, + "loss": 0.9179, + "step": 305720 + }, + { + "epoch": 26.46, + "learning_rate": 5.903088791932734e-06, + "loss": 0.8862, + "step": 305730 + }, + { + "epoch": 26.46, + "learning_rate": 5.901644081019388e-06, + "loss": 0.866, + "step": 305740 + }, + { + "epoch": 26.46, + "learning_rate": 5.9001993701060425e-06, + "loss": 0.888, + "step": 305750 + }, + { + "epoch": 26.47, + "learning_rate": 5.898754659192696e-06, + "loss": 0.8541, + "step": 305760 + }, + { + "epoch": 26.47, + "learning_rate": 5.897309948279349e-06, + "loss": 0.9081, + "step": 305770 + }, + { + "epoch": 26.47, + "learning_rate": 5.8958652373660035e-06, + "loss": 0.9009, + "step": 305780 + }, + { + "epoch": 26.47, + "learning_rate": 5.894420526452657e-06, + "loss": 0.8742, + "step": 305790 + }, + { + "epoch": 26.47, + "learning_rate": 5.892975815539311e-06, + "loss": 0.9139, + "step": 305800 + }, + { + "epoch": 26.47, + "learning_rate": 5.8915311046259646e-06, + "loss": 0.9023, + "step": 305810 + }, + { + "epoch": 26.47, + "learning_rate": 5.890086393712618e-06, + "loss": 0.8649, + "step": 305820 + }, + { + "epoch": 26.47, + "learning_rate": 5.888641682799272e-06, + "loss": 0.8261, + "step": 305830 + }, + { + "epoch": 26.47, + "learning_rate": 5.887196971885926e-06, + "loss": 0.934, + "step": 305840 + }, + { + "epoch": 26.47, + "learning_rate": 5.88575226097258e-06, + "loss": 0.8817, + "step": 305850 + }, + { + "epoch": 26.47, + "learning_rate": 5.884307550059233e-06, + "loss": 0.9379, + "step": 305860 + }, + { + "epoch": 26.48, + "learning_rate": 5.882862839145887e-06, + "loss": 0.9023, + "step": 305870 + }, + { + "epoch": 26.48, + "learning_rate": 5.881418128232541e-06, + "loss": 0.9663, + "step": 305880 + }, + { + "epoch": 26.48, + "learning_rate": 5.879973417319195e-06, + "loss": 0.8942, + "step": 305890 + }, + { + "epoch": 26.48, + "learning_rate": 5.8785287064058485e-06, + "loss": 0.9384, + "step": 305900 + }, + { + "epoch": 26.48, + "learning_rate": 5.877083995492502e-06, + "loss": 0.9034, + "step": 305910 + }, + { + "epoch": 26.48, + "learning_rate": 5.875639284579156e-06, + "loss": 0.9042, + "step": 305920 + }, + { + "epoch": 26.48, + "learning_rate": 5.8741945736658095e-06, + "loss": 0.96, + "step": 305930 + }, + { + "epoch": 26.48, + "learning_rate": 5.872749862752464e-06, + "loss": 0.872, + "step": 305940 + }, + { + "epoch": 26.48, + "learning_rate": 5.871305151839117e-06, + "loss": 0.9555, + "step": 305950 + }, + { + "epoch": 26.48, + "learning_rate": 5.8698604409257706e-06, + "loss": 0.8869, + "step": 305960 + }, + { + "epoch": 26.48, + "learning_rate": 5.868415730012425e-06, + "loss": 0.9136, + "step": 305970 + }, + { + "epoch": 26.48, + "learning_rate": 5.866971019099079e-06, + "loss": 0.8894, + "step": 305980 + }, + { + "epoch": 26.49, + "learning_rate": 5.8655263081857324e-06, + "loss": 0.8966, + "step": 305990 + }, + { + "epoch": 26.49, + "learning_rate": 5.864081597272386e-06, + "loss": 0.9152, + "step": 306000 + }, + { + "epoch": 26.49, + "learning_rate": 5.86263688635904e-06, + "loss": 0.9342, + "step": 306010 + }, + { + "epoch": 26.49, + "learning_rate": 5.8611921754456935e-06, + "loss": 0.9554, + "step": 306020 + }, + { + "epoch": 26.49, + "learning_rate": 5.859747464532348e-06, + "loss": 0.9326, + "step": 306030 + }, + { + "epoch": 26.49, + "learning_rate": 5.858302753619001e-06, + "loss": 0.9336, + "step": 306040 + }, + { + "epoch": 26.49, + "learning_rate": 5.8568580427056545e-06, + "loss": 0.9327, + "step": 306050 + }, + { + "epoch": 26.49, + "learning_rate": 5.855413331792309e-06, + "loss": 0.9434, + "step": 306060 + }, + { + "epoch": 26.49, + "learning_rate": 5.853968620878962e-06, + "loss": 0.9077, + "step": 306070 + }, + { + "epoch": 26.49, + "learning_rate": 5.852523909965616e-06, + "loss": 0.9322, + "step": 306080 + }, + { + "epoch": 26.49, + "learning_rate": 5.85107919905227e-06, + "loss": 0.915, + "step": 306090 + }, + { + "epoch": 26.5, + "learning_rate": 5.849634488138923e-06, + "loss": 0.9381, + "step": 306100 + }, + { + "epoch": 26.5, + "learning_rate": 5.848189777225577e-06, + "loss": 0.8459, + "step": 306110 + }, + { + "epoch": 26.5, + "learning_rate": 5.846745066312232e-06, + "loss": 0.8695, + "step": 306120 + }, + { + "epoch": 26.5, + "learning_rate": 5.845300355398885e-06, + "loss": 0.9088, + "step": 306130 + }, + { + "epoch": 26.5, + "learning_rate": 5.8438556444855384e-06, + "loss": 0.8036, + "step": 306140 + }, + { + "epoch": 26.5, + "learning_rate": 5.842410933572193e-06, + "loss": 0.8611, + "step": 306150 + }, + { + "epoch": 26.5, + "learning_rate": 5.840966222658846e-06, + "loss": 0.8321, + "step": 306160 + }, + { + "epoch": 26.5, + "learning_rate": 5.8395215117455e-06, + "loss": 0.9517, + "step": 306170 + }, + { + "epoch": 26.5, + "learning_rate": 5.838076800832154e-06, + "loss": 0.88, + "step": 306180 + }, + { + "epoch": 26.5, + "learning_rate": 5.836632089918807e-06, + "loss": 0.859, + "step": 306190 + }, + { + "epoch": 26.5, + "learning_rate": 5.835187379005461e-06, + "loss": 0.8714, + "step": 306200 + }, + { + "epoch": 26.5, + "learning_rate": 5.833742668092116e-06, + "loss": 0.8955, + "step": 306210 + }, + { + "epoch": 26.51, + "learning_rate": 5.832297957178769e-06, + "loss": 0.92, + "step": 306220 + }, + { + "epoch": 26.51, + "learning_rate": 5.830853246265422e-06, + "loss": 0.965, + "step": 306230 + }, + { + "epoch": 26.51, + "learning_rate": 5.829408535352077e-06, + "loss": 0.902, + "step": 306240 + }, + { + "epoch": 26.51, + "learning_rate": 5.82796382443873e-06, + "loss": 0.9103, + "step": 306250 + }, + { + "epoch": 26.51, + "learning_rate": 5.826519113525384e-06, + "loss": 0.9278, + "step": 306260 + }, + { + "epoch": 26.51, + "learning_rate": 5.825074402612038e-06, + "loss": 0.9274, + "step": 306270 + }, + { + "epoch": 26.51, + "learning_rate": 5.823629691698691e-06, + "loss": 0.9158, + "step": 306280 + }, + { + "epoch": 26.51, + "learning_rate": 5.822184980785345e-06, + "loss": 0.9229, + "step": 306290 + }, + { + "epoch": 26.51, + "learning_rate": 5.820740269871999e-06, + "loss": 0.9284, + "step": 306300 + }, + { + "epoch": 26.51, + "learning_rate": 5.819295558958653e-06, + "loss": 0.8661, + "step": 306310 + }, + { + "epoch": 26.51, + "learning_rate": 5.817850848045306e-06, + "loss": 0.8898, + "step": 306320 + }, + { + "epoch": 26.52, + "learning_rate": 5.81640613713196e-06, + "loss": 0.9163, + "step": 306330 + }, + { + "epoch": 26.52, + "learning_rate": 5.814961426218614e-06, + "loss": 0.9159, + "step": 306340 + }, + { + "epoch": 26.52, + "learning_rate": 5.813516715305268e-06, + "loss": 0.928, + "step": 306350 + }, + { + "epoch": 26.52, + "learning_rate": 5.812072004391922e-06, + "loss": 0.8874, + "step": 306360 + }, + { + "epoch": 26.52, + "learning_rate": 5.810627293478575e-06, + "loss": 0.9157, + "step": 306370 + }, + { + "epoch": 26.52, + "learning_rate": 5.809182582565229e-06, + "loss": 0.8766, + "step": 306380 + }, + { + "epoch": 26.52, + "learning_rate": 5.807737871651883e-06, + "loss": 0.883, + "step": 306390 + }, + { + "epoch": 26.52, + "learning_rate": 5.806293160738537e-06, + "loss": 0.8771, + "step": 306400 + }, + { + "epoch": 26.52, + "learning_rate": 5.80484844982519e-06, + "loss": 0.9044, + "step": 306410 + }, + { + "epoch": 26.52, + "learning_rate": 5.803403738911844e-06, + "loss": 0.9028, + "step": 306420 + }, + { + "epoch": 26.52, + "learning_rate": 5.801959027998498e-06, + "loss": 0.9041, + "step": 306430 + }, + { + "epoch": 26.52, + "learning_rate": 5.800514317085152e-06, + "loss": 0.8972, + "step": 306440 + }, + { + "epoch": 26.53, + "learning_rate": 5.7990696061718056e-06, + "loss": 0.8589, + "step": 306450 + }, + { + "epoch": 26.53, + "learning_rate": 5.797624895258459e-06, + "loss": 0.917, + "step": 306460 + }, + { + "epoch": 26.53, + "learning_rate": 5.796180184345113e-06, + "loss": 0.9543, + "step": 306470 + }, + { + "epoch": 26.53, + "learning_rate": 5.794735473431767e-06, + "loss": 0.9533, + "step": 306480 + }, + { + "epoch": 26.53, + "learning_rate": 5.793290762518421e-06, + "loss": 0.907, + "step": 306490 + }, + { + "epoch": 26.53, + "learning_rate": 5.791846051605073e-06, + "loss": 0.8748, + "step": 306500 + }, + { + "epoch": 26.53, + "learning_rate": 5.790401340691728e-06, + "loss": 0.9079, + "step": 306510 + }, + { + "epoch": 26.53, + "learning_rate": 5.788956629778382e-06, + "loss": 0.8252, + "step": 306520 + }, + { + "epoch": 26.53, + "learning_rate": 5.787511918865035e-06, + "loss": 0.9572, + "step": 306530 + }, + { + "epoch": 26.53, + "learning_rate": 5.7860672079516895e-06, + "loss": 0.8895, + "step": 306540 + }, + { + "epoch": 26.53, + "learning_rate": 5.784622497038343e-06, + "loss": 0.9357, + "step": 306550 + }, + { + "epoch": 26.53, + "learning_rate": 5.783177786124996e-06, + "loss": 0.9079, + "step": 306560 + }, + { + "epoch": 26.54, + "learning_rate": 5.7817330752116505e-06, + "loss": 0.8831, + "step": 306570 + }, + { + "epoch": 26.54, + "learning_rate": 5.780288364298305e-06, + "loss": 0.9229, + "step": 306580 + }, + { + "epoch": 26.54, + "learning_rate": 5.778843653384957e-06, + "loss": 0.8921, + "step": 306590 + }, + { + "epoch": 26.54, + "learning_rate": 5.7773989424716116e-06, + "loss": 0.872, + "step": 306600 + }, + { + "epoch": 26.54, + "learning_rate": 5.775954231558266e-06, + "loss": 0.9288, + "step": 306610 + }, + { + "epoch": 26.54, + "learning_rate": 5.774509520644919e-06, + "loss": 0.8662, + "step": 306620 + }, + { + "epoch": 26.54, + "learning_rate": 5.7730648097315734e-06, + "loss": 0.9166, + "step": 306630 + }, + { + "epoch": 26.54, + "learning_rate": 5.771620098818227e-06, + "loss": 0.9544, + "step": 306640 + }, + { + "epoch": 26.54, + "learning_rate": 5.77017538790488e-06, + "loss": 0.8571, + "step": 306650 + }, + { + "epoch": 26.54, + "learning_rate": 5.7687306769915345e-06, + "loss": 0.9002, + "step": 306660 + }, + { + "epoch": 26.54, + "learning_rate": 5.767285966078189e-06, + "loss": 0.9256, + "step": 306670 + }, + { + "epoch": 26.55, + "learning_rate": 5.765841255164841e-06, + "loss": 0.9886, + "step": 306680 + }, + { + "epoch": 26.55, + "learning_rate": 5.7643965442514955e-06, + "loss": 0.8916, + "step": 306690 + }, + { + "epoch": 26.55, + "learning_rate": 5.76295183333815e-06, + "loss": 0.9072, + "step": 306700 + }, + { + "epoch": 26.55, + "learning_rate": 5.761507122424803e-06, + "loss": 0.7965, + "step": 306710 + }, + { + "epoch": 26.55, + "learning_rate": 5.7600624115114565e-06, + "loss": 0.9023, + "step": 306720 + }, + { + "epoch": 26.55, + "learning_rate": 5.75861770059811e-06, + "loss": 0.8923, + "step": 306730 + }, + { + "epoch": 26.55, + "learning_rate": 5.757172989684764e-06, + "loss": 0.9651, + "step": 306740 + }, + { + "epoch": 26.55, + "learning_rate": 5.755728278771418e-06, + "loss": 0.9089, + "step": 306750 + }, + { + "epoch": 26.55, + "learning_rate": 5.754283567858072e-06, + "loss": 0.9193, + "step": 306760 + }, + { + "epoch": 26.55, + "learning_rate": 5.752838856944725e-06, + "loss": 0.884, + "step": 306770 + }, + { + "epoch": 26.55, + "learning_rate": 5.7513941460313794e-06, + "loss": 0.9263, + "step": 306780 + }, + { + "epoch": 26.55, + "learning_rate": 5.749949435118033e-06, + "loss": 0.8975, + "step": 306790 + }, + { + "epoch": 26.56, + "learning_rate": 5.748504724204687e-06, + "loss": 0.8656, + "step": 306800 + }, + { + "epoch": 26.56, + "learning_rate": 5.7470600132913405e-06, + "loss": 0.9087, + "step": 306810 + }, + { + "epoch": 26.56, + "learning_rate": 5.745615302377994e-06, + "loss": 0.9556, + "step": 306820 + }, + { + "epoch": 26.56, + "learning_rate": 5.744170591464648e-06, + "loss": 0.8985, + "step": 306830 + }, + { + "epoch": 26.56, + "learning_rate": 5.742725880551302e-06, + "loss": 0.9426, + "step": 306840 + }, + { + "epoch": 26.56, + "learning_rate": 5.741281169637956e-06, + "loss": 0.8804, + "step": 306850 + }, + { + "epoch": 26.56, + "learning_rate": 5.739836458724609e-06, + "loss": 0.9075, + "step": 306860 + }, + { + "epoch": 26.56, + "learning_rate": 5.738391747811263e-06, + "loss": 0.8621, + "step": 306870 + }, + { + "epoch": 26.56, + "learning_rate": 5.736947036897917e-06, + "loss": 0.8971, + "step": 306880 + }, + { + "epoch": 26.56, + "learning_rate": 5.735502325984571e-06, + "loss": 0.9433, + "step": 306890 + }, + { + "epoch": 26.56, + "learning_rate": 5.734057615071224e-06, + "loss": 0.8964, + "step": 306900 + }, + { + "epoch": 26.57, + "learning_rate": 5.732612904157878e-06, + "loss": 0.8677, + "step": 306910 + }, + { + "epoch": 26.57, + "learning_rate": 5.731168193244532e-06, + "loss": 0.8864, + "step": 306920 + }, + { + "epoch": 26.57, + "learning_rate": 5.729723482331186e-06, + "loss": 0.9744, + "step": 306930 + }, + { + "epoch": 26.57, + "learning_rate": 5.72827877141784e-06, + "loss": 0.8912, + "step": 306940 + }, + { + "epoch": 26.57, + "learning_rate": 5.726834060504493e-06, + "loss": 0.8883, + "step": 306950 + }, + { + "epoch": 26.57, + "learning_rate": 5.7253893495911465e-06, + "loss": 0.8846, + "step": 306960 + }, + { + "epoch": 26.57, + "learning_rate": 5.723944638677801e-06, + "loss": 0.9794, + "step": 306970 + }, + { + "epoch": 26.57, + "learning_rate": 5.722499927764455e-06, + "loss": 0.8479, + "step": 306980 + }, + { + "epoch": 26.57, + "learning_rate": 5.721055216851108e-06, + "loss": 0.8869, + "step": 306990 + }, + { + "epoch": 26.57, + "learning_rate": 5.719610505937762e-06, + "loss": 0.9181, + "step": 307000 + }, + { + "epoch": 26.57, + "learning_rate": 5.718165795024416e-06, + "loss": 0.9003, + "step": 307010 + }, + { + "epoch": 26.57, + "learning_rate": 5.716721084111069e-06, + "loss": 0.9205, + "step": 307020 + }, + { + "epoch": 26.58, + "learning_rate": 5.715276373197724e-06, + "loss": 0.9111, + "step": 307030 + }, + { + "epoch": 26.58, + "learning_rate": 5.713831662284377e-06, + "loss": 0.9403, + "step": 307040 + }, + { + "epoch": 26.58, + "learning_rate": 5.7123869513710304e-06, + "loss": 0.8821, + "step": 307050 + }, + { + "epoch": 26.58, + "learning_rate": 5.710942240457685e-06, + "loss": 0.9652, + "step": 307060 + }, + { + "epoch": 26.58, + "learning_rate": 5.709497529544339e-06, + "loss": 0.9197, + "step": 307070 + }, + { + "epoch": 26.58, + "learning_rate": 5.708052818630992e-06, + "loss": 0.8925, + "step": 307080 + }, + { + "epoch": 26.58, + "learning_rate": 5.706608107717646e-06, + "loss": 0.9255, + "step": 307090 + }, + { + "epoch": 26.58, + "learning_rate": 5.7051633968043e-06, + "loss": 0.9334, + "step": 307100 + }, + { + "epoch": 26.58, + "learning_rate": 5.703718685890953e-06, + "loss": 0.8982, + "step": 307110 + }, + { + "epoch": 26.58, + "learning_rate": 5.702273974977608e-06, + "loss": 0.9106, + "step": 307120 + }, + { + "epoch": 26.58, + "learning_rate": 5.700829264064261e-06, + "loss": 0.9308, + "step": 307130 + }, + { + "epoch": 26.59, + "learning_rate": 5.699384553150914e-06, + "loss": 0.913, + "step": 307140 + }, + { + "epoch": 26.59, + "learning_rate": 5.697939842237569e-06, + "loss": 0.9329, + "step": 307150 + }, + { + "epoch": 26.59, + "learning_rate": 5.696495131324223e-06, + "loss": 0.9459, + "step": 307160 + }, + { + "epoch": 26.59, + "learning_rate": 5.695050420410876e-06, + "loss": 0.9208, + "step": 307170 + }, + { + "epoch": 26.59, + "learning_rate": 5.69360570949753e-06, + "loss": 0.898, + "step": 307180 + }, + { + "epoch": 26.59, + "learning_rate": 5.692160998584183e-06, + "loss": 0.8905, + "step": 307190 + }, + { + "epoch": 26.59, + "learning_rate": 5.690716287670837e-06, + "loss": 0.8488, + "step": 307200 + }, + { + "epoch": 26.59, + "learning_rate": 5.6892715767574915e-06, + "loss": 0.9276, + "step": 307210 + }, + { + "epoch": 26.59, + "learning_rate": 5.687826865844145e-06, + "loss": 0.9381, + "step": 307220 + }, + { + "epoch": 26.59, + "learning_rate": 5.686382154930798e-06, + "loss": 0.9232, + "step": 307230 + }, + { + "epoch": 26.59, + "learning_rate": 5.6849374440174526e-06, + "loss": 0.8778, + "step": 307240 + }, + { + "epoch": 26.59, + "learning_rate": 5.683492733104106e-06, + "loss": 0.8946, + "step": 307250 + }, + { + "epoch": 26.6, + "learning_rate": 5.68204802219076e-06, + "loss": 0.854, + "step": 307260 + }, + { + "epoch": 26.6, + "learning_rate": 5.680603311277414e-06, + "loss": 0.9083, + "step": 307270 + }, + { + "epoch": 26.6, + "learning_rate": 5.679158600364067e-06, + "loss": 0.9722, + "step": 307280 + }, + { + "epoch": 26.6, + "learning_rate": 5.677713889450721e-06, + "loss": 0.8366, + "step": 307290 + }, + { + "epoch": 26.6, + "learning_rate": 5.6762691785373755e-06, + "loss": 0.921, + "step": 307300 + }, + { + "epoch": 26.6, + "learning_rate": 5.674824467624029e-06, + "loss": 0.8394, + "step": 307310 + }, + { + "epoch": 26.6, + "learning_rate": 5.673379756710682e-06, + "loss": 0.9635, + "step": 307320 + }, + { + "epoch": 26.6, + "learning_rate": 5.6719350457973365e-06, + "loss": 0.8502, + "step": 307330 + }, + { + "epoch": 26.6, + "learning_rate": 5.67049033488399e-06, + "loss": 0.9663, + "step": 307340 + }, + { + "epoch": 26.6, + "learning_rate": 5.669045623970644e-06, + "loss": 0.9311, + "step": 307350 + }, + { + "epoch": 26.6, + "learning_rate": 5.6676009130572975e-06, + "loss": 0.9375, + "step": 307360 + }, + { + "epoch": 26.61, + "learning_rate": 5.666156202143951e-06, + "loss": 0.9275, + "step": 307370 + }, + { + "epoch": 26.61, + "learning_rate": 5.664711491230605e-06, + "loss": 0.8781, + "step": 307380 + }, + { + "epoch": 26.61, + "learning_rate": 5.663266780317259e-06, + "loss": 0.911, + "step": 307390 + }, + { + "epoch": 26.61, + "learning_rate": 5.661822069403913e-06, + "loss": 0.9098, + "step": 307400 + }, + { + "epoch": 26.61, + "learning_rate": 5.660377358490566e-06, + "loss": 0.8746, + "step": 307410 + }, + { + "epoch": 26.61, + "learning_rate": 5.65893264757722e-06, + "loss": 0.9283, + "step": 307420 + }, + { + "epoch": 26.61, + "learning_rate": 5.657487936663874e-06, + "loss": 0.9027, + "step": 307430 + }, + { + "epoch": 26.61, + "learning_rate": 5.656043225750528e-06, + "loss": 0.915, + "step": 307440 + }, + { + "epoch": 26.61, + "learning_rate": 5.6545985148371815e-06, + "loss": 0.8495, + "step": 307450 + }, + { + "epoch": 26.61, + "learning_rate": 5.653153803923835e-06, + "loss": 0.9348, + "step": 307460 + }, + { + "epoch": 26.61, + "learning_rate": 5.651709093010489e-06, + "loss": 0.9114, + "step": 307470 + }, + { + "epoch": 26.61, + "learning_rate": 5.6502643820971425e-06, + "loss": 0.8957, + "step": 307480 + }, + { + "epoch": 26.62, + "learning_rate": 5.648819671183797e-06, + "loss": 0.8765, + "step": 307490 + }, + { + "epoch": 26.62, + "learning_rate": 5.64737496027045e-06, + "loss": 0.9095, + "step": 307500 + }, + { + "epoch": 26.62, + "learning_rate": 5.6459302493571035e-06, + "loss": 0.8882, + "step": 307510 + }, + { + "epoch": 26.62, + "learning_rate": 5.644485538443758e-06, + "loss": 0.8631, + "step": 307520 + }, + { + "epoch": 26.62, + "learning_rate": 5.643040827530412e-06, + "loss": 0.9246, + "step": 307530 + }, + { + "epoch": 26.62, + "learning_rate": 5.641596116617065e-06, + "loss": 0.8955, + "step": 307540 + }, + { + "epoch": 26.62, + "learning_rate": 5.640151405703719e-06, + "loss": 0.8705, + "step": 307550 + }, + { + "epoch": 26.62, + "learning_rate": 5.638706694790373e-06, + "loss": 0.8852, + "step": 307560 + }, + { + "epoch": 26.62, + "learning_rate": 5.6372619838770264e-06, + "loss": 0.8961, + "step": 307570 + }, + { + "epoch": 26.62, + "learning_rate": 5.635817272963681e-06, + "loss": 0.8583, + "step": 307580 + }, + { + "epoch": 26.62, + "learning_rate": 5.634372562050334e-06, + "loss": 0.8627, + "step": 307590 + }, + { + "epoch": 26.63, + "learning_rate": 5.6329278511369875e-06, + "loss": 0.9055, + "step": 307600 + }, + { + "epoch": 26.63, + "learning_rate": 5.631483140223642e-06, + "loss": 0.9952, + "step": 307610 + }, + { + "epoch": 26.63, + "learning_rate": 5.630038429310296e-06, + "loss": 0.9083, + "step": 307620 + }, + { + "epoch": 26.63, + "learning_rate": 5.628593718396949e-06, + "loss": 0.9252, + "step": 307630 + }, + { + "epoch": 26.63, + "learning_rate": 5.627149007483603e-06, + "loss": 0.867, + "step": 307640 + }, + { + "epoch": 26.63, + "learning_rate": 5.625704296570256e-06, + "loss": 0.8805, + "step": 307650 + }, + { + "epoch": 26.63, + "learning_rate": 5.62425958565691e-06, + "loss": 0.9164, + "step": 307660 + }, + { + "epoch": 26.63, + "learning_rate": 5.622814874743565e-06, + "loss": 0.8711, + "step": 307670 + }, + { + "epoch": 26.63, + "learning_rate": 5.621370163830217e-06, + "loss": 0.8464, + "step": 307680 + }, + { + "epoch": 26.63, + "learning_rate": 5.619925452916871e-06, + "loss": 0.926, + "step": 307690 + }, + { + "epoch": 26.63, + "learning_rate": 5.618480742003526e-06, + "loss": 0.8759, + "step": 307700 + }, + { + "epoch": 26.63, + "learning_rate": 5.617036031090179e-06, + "loss": 0.8674, + "step": 307710 + }, + { + "epoch": 26.64, + "learning_rate": 5.6155913201768325e-06, + "loss": 0.8706, + "step": 307720 + }, + { + "epoch": 26.64, + "learning_rate": 5.614146609263487e-06, + "loss": 0.9516, + "step": 307730 + }, + { + "epoch": 26.64, + "learning_rate": 5.61270189835014e-06, + "loss": 0.8466, + "step": 307740 + }, + { + "epoch": 26.64, + "learning_rate": 5.611257187436794e-06, + "loss": 0.9749, + "step": 307750 + }, + { + "epoch": 26.64, + "learning_rate": 5.6098124765234486e-06, + "loss": 0.8775, + "step": 307760 + }, + { + "epoch": 26.64, + "learning_rate": 5.608367765610101e-06, + "loss": 0.8993, + "step": 307770 + }, + { + "epoch": 26.64, + "learning_rate": 5.606923054696755e-06, + "loss": 0.8739, + "step": 307780 + }, + { + "epoch": 26.64, + "learning_rate": 5.60547834378341e-06, + "loss": 0.979, + "step": 307790 + }, + { + "epoch": 26.64, + "learning_rate": 5.604033632870063e-06, + "loss": 0.8976, + "step": 307800 + }, + { + "epoch": 26.64, + "learning_rate": 5.602588921956716e-06, + "loss": 0.9211, + "step": 307810 + }, + { + "epoch": 26.64, + "learning_rate": 5.601144211043371e-06, + "loss": 0.9073, + "step": 307820 + }, + { + "epoch": 26.64, + "learning_rate": 5.599699500130024e-06, + "loss": 0.8877, + "step": 307830 + }, + { + "epoch": 26.65, + "learning_rate": 5.598254789216678e-06, + "loss": 0.9613, + "step": 307840 + }, + { + "epoch": 26.65, + "learning_rate": 5.596810078303332e-06, + "loss": 0.9196, + "step": 307850 + }, + { + "epoch": 26.65, + "learning_rate": 5.595365367389985e-06, + "loss": 0.8842, + "step": 307860 + }, + { + "epoch": 26.65, + "learning_rate": 5.593920656476639e-06, + "loss": 0.8866, + "step": 307870 + }, + { + "epoch": 26.65, + "learning_rate": 5.592475945563293e-06, + "loss": 0.8536, + "step": 307880 + }, + { + "epoch": 26.65, + "learning_rate": 5.591031234649947e-06, + "loss": 0.9252, + "step": 307890 + }, + { + "epoch": 26.65, + "learning_rate": 5.5895865237366e-06, + "loss": 0.8949, + "step": 307900 + }, + { + "epoch": 26.65, + "learning_rate": 5.588141812823254e-06, + "loss": 0.9223, + "step": 307910 + }, + { + "epoch": 26.65, + "learning_rate": 5.586697101909908e-06, + "loss": 0.9212, + "step": 307920 + }, + { + "epoch": 26.65, + "learning_rate": 5.585252390996562e-06, + "loss": 0.9215, + "step": 307930 + }, + { + "epoch": 26.65, + "learning_rate": 5.583807680083216e-06, + "loss": 0.8157, + "step": 307940 + }, + { + "epoch": 26.66, + "learning_rate": 5.582362969169869e-06, + "loss": 0.9012, + "step": 307950 + }, + { + "epoch": 26.66, + "learning_rate": 5.580918258256523e-06, + "loss": 0.882, + "step": 307960 + }, + { + "epoch": 26.66, + "learning_rate": 5.579473547343177e-06, + "loss": 0.885, + "step": 307970 + }, + { + "epoch": 26.66, + "learning_rate": 5.578028836429831e-06, + "loss": 0.9491, + "step": 307980 + }, + { + "epoch": 26.66, + "learning_rate": 5.576584125516484e-06, + "loss": 0.8933, + "step": 307990 + }, + { + "epoch": 26.66, + "learning_rate": 5.575139414603138e-06, + "loss": 0.9737, + "step": 308000 + }, + { + "epoch": 26.66, + "learning_rate": 5.573694703689792e-06, + "loss": 0.884, + "step": 308010 + }, + { + "epoch": 26.66, + "learning_rate": 5.572249992776446e-06, + "loss": 0.9177, + "step": 308020 + }, + { + "epoch": 26.66, + "learning_rate": 5.5708052818630996e-06, + "loss": 0.866, + "step": 308030 + }, + { + "epoch": 26.66, + "learning_rate": 5.569360570949753e-06, + "loss": 0.9067, + "step": 308040 + }, + { + "epoch": 26.66, + "learning_rate": 5.567915860036407e-06, + "loss": 0.9091, + "step": 308050 + }, + { + "epoch": 26.66, + "learning_rate": 5.566471149123061e-06, + "loss": 0.9513, + "step": 308060 + }, + { + "epoch": 26.67, + "learning_rate": 5.565026438209715e-06, + "loss": 0.9456, + "step": 308070 + }, + { + "epoch": 26.67, + "learning_rate": 5.563581727296368e-06, + "loss": 0.9336, + "step": 308080 + }, + { + "epoch": 26.67, + "learning_rate": 5.562137016383022e-06, + "loss": 0.8404, + "step": 308090 + }, + { + "epoch": 26.67, + "learning_rate": 5.560692305469676e-06, + "loss": 0.926, + "step": 308100 + }, + { + "epoch": 26.67, + "learning_rate": 5.559247594556329e-06, + "loss": 0.9088, + "step": 308110 + }, + { + "epoch": 26.67, + "learning_rate": 5.5578028836429835e-06, + "loss": 0.867, + "step": 308120 + }, + { + "epoch": 26.67, + "learning_rate": 5.556358172729637e-06, + "loss": 0.9074, + "step": 308130 + }, + { + "epoch": 26.67, + "learning_rate": 5.55491346181629e-06, + "loss": 0.9197, + "step": 308140 + }, + { + "epoch": 26.67, + "learning_rate": 5.5534687509029445e-06, + "loss": 0.9079, + "step": 308150 + }, + { + "epoch": 26.67, + "learning_rate": 5.552024039989599e-06, + "loss": 0.8575, + "step": 308160 + }, + { + "epoch": 26.67, + "learning_rate": 5.550579329076252e-06, + "loss": 0.8663, + "step": 308170 + }, + { + "epoch": 26.68, + "learning_rate": 5.5491346181629056e-06, + "loss": 0.9044, + "step": 308180 + }, + { + "epoch": 26.68, + "learning_rate": 5.54768990724956e-06, + "loss": 0.9121, + "step": 308190 + }, + { + "epoch": 26.68, + "learning_rate": 5.546245196336213e-06, + "loss": 0.8953, + "step": 308200 + }, + { + "epoch": 26.68, + "learning_rate": 5.5448004854228674e-06, + "loss": 0.9491, + "step": 308210 + }, + { + "epoch": 26.68, + "learning_rate": 5.543355774509521e-06, + "loss": 0.9144, + "step": 308220 + }, + { + "epoch": 26.68, + "learning_rate": 5.541911063596174e-06, + "loss": 0.8803, + "step": 308230 + }, + { + "epoch": 26.68, + "learning_rate": 5.5404663526828285e-06, + "loss": 0.9831, + "step": 308240 + }, + { + "epoch": 26.68, + "learning_rate": 5.539021641769483e-06, + "loss": 0.8891, + "step": 308250 + }, + { + "epoch": 26.68, + "learning_rate": 5.537576930856136e-06, + "loss": 0.9254, + "step": 308260 + }, + { + "epoch": 26.68, + "learning_rate": 5.5361322199427895e-06, + "loss": 0.9008, + "step": 308270 + }, + { + "epoch": 26.68, + "learning_rate": 5.534687509029444e-06, + "loss": 0.9105, + "step": 308280 + }, + { + "epoch": 26.68, + "learning_rate": 5.533242798116097e-06, + "loss": 0.9428, + "step": 308290 + }, + { + "epoch": 26.69, + "learning_rate": 5.531798087202751e-06, + "loss": 0.9435, + "step": 308300 + }, + { + "epoch": 26.69, + "learning_rate": 5.530353376289405e-06, + "loss": 0.8622, + "step": 308310 + }, + { + "epoch": 26.69, + "learning_rate": 5.528908665376058e-06, + "loss": 0.9075, + "step": 308320 + }, + { + "epoch": 26.69, + "learning_rate": 5.527463954462712e-06, + "loss": 0.8904, + "step": 308330 + }, + { + "epoch": 26.69, + "learning_rate": 5.526019243549366e-06, + "loss": 0.8905, + "step": 308340 + }, + { + "epoch": 26.69, + "learning_rate": 5.52457453263602e-06, + "loss": 0.9014, + "step": 308350 + }, + { + "epoch": 26.69, + "learning_rate": 5.5231298217226734e-06, + "loss": 0.8796, + "step": 308360 + }, + { + "epoch": 26.69, + "learning_rate": 5.521685110809327e-06, + "loss": 0.9373, + "step": 308370 + }, + { + "epoch": 26.69, + "learning_rate": 5.520240399895981e-06, + "loss": 0.8395, + "step": 308380 + }, + { + "epoch": 26.69, + "learning_rate": 5.518795688982635e-06, + "loss": 0.8941, + "step": 308390 + }, + { + "epoch": 26.69, + "learning_rate": 5.517350978069289e-06, + "loss": 0.8564, + "step": 308400 + }, + { + "epoch": 26.7, + "learning_rate": 5.515906267155942e-06, + "loss": 0.9222, + "step": 308410 + }, + { + "epoch": 26.7, + "learning_rate": 5.514461556242596e-06, + "loss": 0.9652, + "step": 308420 + }, + { + "epoch": 26.7, + "learning_rate": 5.51301684532925e-06, + "loss": 0.8919, + "step": 308430 + }, + { + "epoch": 26.7, + "learning_rate": 5.511572134415904e-06, + "loss": 0.9486, + "step": 308440 + }, + { + "epoch": 26.7, + "learning_rate": 5.510127423502557e-06, + "loss": 0.9626, + "step": 308450 + }, + { + "epoch": 26.7, + "learning_rate": 5.508682712589211e-06, + "loss": 0.9174, + "step": 308460 + }, + { + "epoch": 26.7, + "learning_rate": 5.507238001675865e-06, + "loss": 0.9372, + "step": 308470 + }, + { + "epoch": 26.7, + "learning_rate": 5.505793290762519e-06, + "loss": 0.8896, + "step": 308480 + }, + { + "epoch": 26.7, + "learning_rate": 5.504348579849173e-06, + "loss": 0.8968, + "step": 308490 + }, + { + "epoch": 26.7, + "learning_rate": 5.502903868935826e-06, + "loss": 0.836, + "step": 308500 + }, + { + "epoch": 26.7, + "learning_rate": 5.50145915802248e-06, + "loss": 0.9436, + "step": 308510 + }, + { + "epoch": 26.7, + "learning_rate": 5.500014447109134e-06, + "loss": 0.8933, + "step": 308520 + }, + { + "epoch": 26.71, + "learning_rate": 5.498569736195788e-06, + "loss": 0.9327, + "step": 308530 + }, + { + "epoch": 26.71, + "learning_rate": 5.497125025282441e-06, + "loss": 0.9048, + "step": 308540 + }, + { + "epoch": 26.71, + "learning_rate": 5.495680314369095e-06, + "loss": 0.9085, + "step": 308550 + }, + { + "epoch": 26.71, + "learning_rate": 5.494235603455749e-06, + "loss": 0.8981, + "step": 308560 + }, + { + "epoch": 26.71, + "learning_rate": 5.492790892542402e-06, + "loss": 0.8753, + "step": 308570 + }, + { + "epoch": 26.71, + "learning_rate": 5.491346181629057e-06, + "loss": 0.9507, + "step": 308580 + }, + { + "epoch": 26.71, + "learning_rate": 5.48990147071571e-06, + "loss": 0.8958, + "step": 308590 + }, + { + "epoch": 26.71, + "learning_rate": 5.488456759802363e-06, + "loss": 0.8955, + "step": 308600 + }, + { + "epoch": 26.71, + "learning_rate": 5.487012048889018e-06, + "loss": 0.8812, + "step": 308610 + }, + { + "epoch": 26.71, + "learning_rate": 5.485567337975672e-06, + "loss": 0.9405, + "step": 308620 + }, + { + "epoch": 26.71, + "learning_rate": 5.484122627062325e-06, + "loss": 0.8743, + "step": 308630 + }, + { + "epoch": 26.72, + "learning_rate": 5.482677916148979e-06, + "loss": 0.8649, + "step": 308640 + }, + { + "epoch": 26.72, + "learning_rate": 5.481233205235633e-06, + "loss": 0.9166, + "step": 308650 + }, + { + "epoch": 26.72, + "learning_rate": 5.479788494322286e-06, + "loss": 0.8793, + "step": 308660 + }, + { + "epoch": 26.72, + "learning_rate": 5.4783437834089405e-06, + "loss": 0.8876, + "step": 308670 + }, + { + "epoch": 26.72, + "learning_rate": 5.476899072495594e-06, + "loss": 0.8453, + "step": 308680 + }, + { + "epoch": 26.72, + "learning_rate": 5.475454361582247e-06, + "loss": 0.8659, + "step": 308690 + }, + { + "epoch": 26.72, + "learning_rate": 5.474009650668902e-06, + "loss": 0.8994, + "step": 308700 + }, + { + "epoch": 26.72, + "learning_rate": 5.472564939755556e-06, + "loss": 0.9114, + "step": 308710 + }, + { + "epoch": 26.72, + "learning_rate": 5.471120228842208e-06, + "loss": 0.8906, + "step": 308720 + }, + { + "epoch": 26.72, + "learning_rate": 5.469675517928863e-06, + "loss": 0.9018, + "step": 308730 + }, + { + "epoch": 26.72, + "learning_rate": 5.468230807015517e-06, + "loss": 0.9113, + "step": 308740 + }, + { + "epoch": 26.72, + "learning_rate": 5.46678609610217e-06, + "loss": 0.8373, + "step": 308750 + }, + { + "epoch": 26.73, + "learning_rate": 5.4653413851888245e-06, + "loss": 0.9424, + "step": 308760 + }, + { + "epoch": 26.73, + "learning_rate": 5.463896674275477e-06, + "loss": 0.8946, + "step": 308770 + }, + { + "epoch": 26.73, + "learning_rate": 5.462451963362131e-06, + "loss": 0.9197, + "step": 308780 + }, + { + "epoch": 26.73, + "learning_rate": 5.4610072524487855e-06, + "loss": 0.8473, + "step": 308790 + }, + { + "epoch": 26.73, + "learning_rate": 5.459562541535439e-06, + "loss": 0.888, + "step": 308800 + }, + { + "epoch": 26.73, + "learning_rate": 5.458117830622092e-06, + "loss": 0.928, + "step": 308810 + }, + { + "epoch": 26.73, + "learning_rate": 5.4566731197087466e-06, + "loss": 0.9196, + "step": 308820 + }, + { + "epoch": 26.73, + "learning_rate": 5.4552284087954e-06, + "loss": 0.8577, + "step": 308830 + }, + { + "epoch": 26.73, + "learning_rate": 5.453783697882054e-06, + "loss": 0.9572, + "step": 308840 + }, + { + "epoch": 26.73, + "learning_rate": 5.4523389869687084e-06, + "loss": 0.9382, + "step": 308850 + }, + { + "epoch": 26.73, + "learning_rate": 5.450894276055361e-06, + "loss": 0.9305, + "step": 308860 + }, + { + "epoch": 26.73, + "learning_rate": 5.449449565142015e-06, + "loss": 0.9641, + "step": 308870 + }, + { + "epoch": 26.74, + "learning_rate": 5.4480048542286695e-06, + "loss": 0.8851, + "step": 308880 + }, + { + "epoch": 26.74, + "learning_rate": 5.446560143315323e-06, + "loss": 0.8828, + "step": 308890 + }, + { + "epoch": 26.74, + "learning_rate": 5.445115432401976e-06, + "loss": 0.837, + "step": 308900 + }, + { + "epoch": 26.74, + "learning_rate": 5.4436707214886305e-06, + "loss": 0.88, + "step": 308910 + }, + { + "epoch": 26.74, + "learning_rate": 5.442226010575284e-06, + "loss": 0.937, + "step": 308920 + }, + { + "epoch": 26.74, + "learning_rate": 5.440781299661938e-06, + "loss": 0.8922, + "step": 308930 + }, + { + "epoch": 26.74, + "learning_rate": 5.439336588748592e-06, + "loss": 0.9212, + "step": 308940 + }, + { + "epoch": 26.74, + "learning_rate": 5.437891877835245e-06, + "loss": 0.9004, + "step": 308950 + }, + { + "epoch": 26.74, + "learning_rate": 5.436447166921899e-06, + "loss": 0.9324, + "step": 308960 + }, + { + "epoch": 26.74, + "learning_rate": 5.435002456008553e-06, + "loss": 0.944, + "step": 308970 + }, + { + "epoch": 26.74, + "learning_rate": 5.433557745095207e-06, + "loss": 0.9212, + "step": 308980 + }, + { + "epoch": 26.75, + "learning_rate": 5.43211303418186e-06, + "loss": 0.9059, + "step": 308990 + }, + { + "epoch": 26.75, + "learning_rate": 5.430668323268514e-06, + "loss": 0.8873, + "step": 309000 + }, + { + "epoch": 26.75, + "learning_rate": 5.429223612355168e-06, + "loss": 0.9224, + "step": 309010 + }, + { + "epoch": 26.75, + "learning_rate": 5.427778901441822e-06, + "loss": 0.923, + "step": 309020 + }, + { + "epoch": 26.75, + "learning_rate": 5.4263341905284755e-06, + "loss": 0.9825, + "step": 309030 + }, + { + "epoch": 26.75, + "learning_rate": 5.424889479615129e-06, + "loss": 0.9107, + "step": 309040 + }, + { + "epoch": 26.75, + "learning_rate": 5.423444768701783e-06, + "loss": 0.9148, + "step": 309050 + }, + { + "epoch": 26.75, + "learning_rate": 5.4220000577884365e-06, + "loss": 0.9161, + "step": 309060 + }, + { + "epoch": 26.75, + "learning_rate": 5.420555346875091e-06, + "loss": 0.8866, + "step": 309070 + }, + { + "epoch": 26.75, + "learning_rate": 5.419110635961744e-06, + "loss": 0.9303, + "step": 309080 + }, + { + "epoch": 26.75, + "learning_rate": 5.4176659250483975e-06, + "loss": 0.8791, + "step": 309090 + }, + { + "epoch": 26.75, + "learning_rate": 5.416221214135052e-06, + "loss": 0.9533, + "step": 309100 + }, + { + "epoch": 26.76, + "learning_rate": 5.414776503221706e-06, + "loss": 0.9202, + "step": 309110 + }, + { + "epoch": 26.76, + "learning_rate": 5.413331792308359e-06, + "loss": 0.9543, + "step": 309120 + }, + { + "epoch": 26.76, + "learning_rate": 5.411887081395013e-06, + "loss": 0.8939, + "step": 309130 + }, + { + "epoch": 26.76, + "learning_rate": 5.410442370481667e-06, + "loss": 0.8943, + "step": 309140 + }, + { + "epoch": 26.76, + "learning_rate": 5.4089976595683204e-06, + "loss": 0.8944, + "step": 309150 + }, + { + "epoch": 26.76, + "learning_rate": 5.407552948654975e-06, + "loss": 0.8721, + "step": 309160 + }, + { + "epoch": 26.76, + "learning_rate": 5.406108237741628e-06, + "loss": 0.8713, + "step": 309170 + }, + { + "epoch": 26.76, + "learning_rate": 5.4046635268282815e-06, + "loss": 0.8922, + "step": 309180 + }, + { + "epoch": 26.76, + "learning_rate": 5.403218815914936e-06, + "loss": 0.9208, + "step": 309190 + }, + { + "epoch": 26.76, + "learning_rate": 5.40177410500159e-06, + "loss": 0.8811, + "step": 309200 + }, + { + "epoch": 26.76, + "learning_rate": 5.400329394088243e-06, + "loss": 0.9034, + "step": 309210 + }, + { + "epoch": 26.77, + "learning_rate": 5.398884683174897e-06, + "loss": 0.8821, + "step": 309220 + }, + { + "epoch": 26.77, + "learning_rate": 5.39743997226155e-06, + "loss": 0.9391, + "step": 309230 + }, + { + "epoch": 26.77, + "learning_rate": 5.395995261348204e-06, + "loss": 0.8934, + "step": 309240 + }, + { + "epoch": 26.77, + "learning_rate": 5.394550550434859e-06, + "loss": 0.8958, + "step": 309250 + }, + { + "epoch": 26.77, + "learning_rate": 5.393105839521512e-06, + "loss": 0.8936, + "step": 309260 + }, + { + "epoch": 26.77, + "learning_rate": 5.391661128608165e-06, + "loss": 0.9126, + "step": 309270 + }, + { + "epoch": 26.77, + "learning_rate": 5.39021641769482e-06, + "loss": 0.8844, + "step": 309280 + }, + { + "epoch": 26.77, + "learning_rate": 5.388771706781473e-06, + "loss": 0.9191, + "step": 309290 + }, + { + "epoch": 26.77, + "learning_rate": 5.387326995868127e-06, + "loss": 0.919, + "step": 309300 + }, + { + "epoch": 26.77, + "learning_rate": 5.385882284954781e-06, + "loss": 0.9027, + "step": 309310 + }, + { + "epoch": 26.77, + "learning_rate": 5.384437574041434e-06, + "loss": 0.9147, + "step": 309320 + }, + { + "epoch": 26.77, + "learning_rate": 5.382992863128088e-06, + "loss": 0.8999, + "step": 309330 + }, + { + "epoch": 26.78, + "learning_rate": 5.3815481522147426e-06, + "loss": 0.9118, + "step": 309340 + }, + { + "epoch": 26.78, + "learning_rate": 5.380103441301396e-06, + "loss": 0.9273, + "step": 309350 + }, + { + "epoch": 26.78, + "learning_rate": 5.378658730388049e-06, + "loss": 0.8988, + "step": 309360 + }, + { + "epoch": 26.78, + "learning_rate": 5.377214019474704e-06, + "loss": 0.8604, + "step": 309370 + }, + { + "epoch": 26.78, + "learning_rate": 5.375769308561357e-06, + "loss": 0.8968, + "step": 309380 + }, + { + "epoch": 26.78, + "learning_rate": 5.374324597648011e-06, + "loss": 1.0225, + "step": 309390 + }, + { + "epoch": 26.78, + "learning_rate": 5.372879886734665e-06, + "loss": 0.9207, + "step": 309400 + }, + { + "epoch": 26.78, + "learning_rate": 5.371435175821318e-06, + "loss": 0.9257, + "step": 309410 + }, + { + "epoch": 26.78, + "learning_rate": 5.369990464907972e-06, + "loss": 0.9092, + "step": 309420 + }, + { + "epoch": 26.78, + "learning_rate": 5.368545753994626e-06, + "loss": 0.9859, + "step": 309430 + }, + { + "epoch": 26.78, + "learning_rate": 5.36710104308128e-06, + "loss": 0.9505, + "step": 309440 + }, + { + "epoch": 26.79, + "learning_rate": 5.365656332167933e-06, + "loss": 0.9332, + "step": 309450 + }, + { + "epoch": 26.79, + "learning_rate": 5.364211621254587e-06, + "loss": 0.8878, + "step": 309460 + }, + { + "epoch": 26.79, + "learning_rate": 5.362766910341241e-06, + "loss": 0.9016, + "step": 309470 + }, + { + "epoch": 26.79, + "learning_rate": 5.361322199427895e-06, + "loss": 0.9157, + "step": 309480 + }, + { + "epoch": 26.79, + "learning_rate": 5.359877488514549e-06, + "loss": 0.8783, + "step": 309490 + }, + { + "epoch": 26.79, + "learning_rate": 5.358432777601202e-06, + "loss": 0.8739, + "step": 309500 + }, + { + "epoch": 26.79, + "learning_rate": 5.356988066687856e-06, + "loss": 0.9148, + "step": 309510 + }, + { + "epoch": 26.79, + "learning_rate": 5.35554335577451e-06, + "loss": 0.904, + "step": 309520 + }, + { + "epoch": 26.79, + "learning_rate": 5.354098644861164e-06, + "loss": 0.8984, + "step": 309530 + }, + { + "epoch": 26.79, + "learning_rate": 5.352653933947817e-06, + "loss": 0.8832, + "step": 309540 + }, + { + "epoch": 26.79, + "learning_rate": 5.351209223034471e-06, + "loss": 0.8748, + "step": 309550 + }, + { + "epoch": 26.79, + "learning_rate": 5.349764512121125e-06, + "loss": 0.8889, + "step": 309560 + }, + { + "epoch": 26.8, + "learning_rate": 5.348319801207779e-06, + "loss": 0.9053, + "step": 309570 + }, + { + "epoch": 26.8, + "learning_rate": 5.3468750902944325e-06, + "loss": 0.9112, + "step": 309580 + }, + { + "epoch": 26.8, + "learning_rate": 5.345430379381086e-06, + "loss": 0.8535, + "step": 309590 + }, + { + "epoch": 26.8, + "learning_rate": 5.34398566846774e-06, + "loss": 0.9474, + "step": 309600 + }, + { + "epoch": 26.8, + "learning_rate": 5.3425409575543936e-06, + "loss": 0.911, + "step": 309610 + }, + { + "epoch": 26.8, + "learning_rate": 5.341096246641048e-06, + "loss": 0.9498, + "step": 309620 + }, + { + "epoch": 26.8, + "learning_rate": 5.339651535727701e-06, + "loss": 0.8804, + "step": 309630 + }, + { + "epoch": 26.8, + "learning_rate": 5.338206824814355e-06, + "loss": 0.9212, + "step": 309640 + }, + { + "epoch": 26.8, + "learning_rate": 5.336762113901009e-06, + "loss": 0.95, + "step": 309650 + }, + { + "epoch": 26.8, + "learning_rate": 5.335317402987662e-06, + "loss": 0.8853, + "step": 309660 + }, + { + "epoch": 26.8, + "learning_rate": 5.3338726920743165e-06, + "loss": 0.9206, + "step": 309670 + }, + { + "epoch": 26.81, + "learning_rate": 5.33242798116097e-06, + "loss": 0.9528, + "step": 309680 + }, + { + "epoch": 26.81, + "learning_rate": 5.330983270247623e-06, + "loss": 0.9108, + "step": 309690 + }, + { + "epoch": 26.81, + "learning_rate": 5.3295385593342775e-06, + "loss": 0.9054, + "step": 309700 + }, + { + "epoch": 26.81, + "learning_rate": 5.328093848420932e-06, + "loss": 0.9169, + "step": 309710 + }, + { + "epoch": 26.81, + "learning_rate": 5.326649137507584e-06, + "loss": 0.8884, + "step": 309720 + }, + { + "epoch": 26.81, + "learning_rate": 5.3252044265942385e-06, + "loss": 0.8896, + "step": 309730 + }, + { + "epoch": 26.81, + "learning_rate": 5.323759715680893e-06, + "loss": 0.9217, + "step": 309740 + }, + { + "epoch": 26.81, + "learning_rate": 5.322315004767546e-06, + "loss": 0.8484, + "step": 309750 + }, + { + "epoch": 26.81, + "learning_rate": 5.3208702938542e-06, + "loss": 0.9332, + "step": 309760 + }, + { + "epoch": 26.81, + "learning_rate": 5.319425582940854e-06, + "loss": 0.9116, + "step": 309770 + }, + { + "epoch": 26.81, + "learning_rate": 5.317980872027507e-06, + "loss": 0.9138, + "step": 309780 + }, + { + "epoch": 26.81, + "learning_rate": 5.3165361611141614e-06, + "loss": 0.8954, + "step": 309790 + }, + { + "epoch": 26.82, + "learning_rate": 5.315091450200816e-06, + "loss": 0.9521, + "step": 309800 + }, + { + "epoch": 26.82, + "learning_rate": 5.313646739287468e-06, + "loss": 0.8651, + "step": 309810 + }, + { + "epoch": 26.82, + "learning_rate": 5.3122020283741225e-06, + "loss": 0.9084, + "step": 309820 + }, + { + "epoch": 26.82, + "learning_rate": 5.310757317460777e-06, + "loss": 0.8584, + "step": 309830 + }, + { + "epoch": 26.82, + "learning_rate": 5.30931260654743e-06, + "loss": 0.9199, + "step": 309840 + }, + { + "epoch": 26.82, + "learning_rate": 5.307867895634084e-06, + "loss": 0.9386, + "step": 309850 + }, + { + "epoch": 26.82, + "learning_rate": 5.306423184720738e-06, + "loss": 0.8292, + "step": 309860 + }, + { + "epoch": 26.82, + "learning_rate": 5.304978473807391e-06, + "loss": 0.9155, + "step": 309870 + }, + { + "epoch": 26.82, + "learning_rate": 5.303533762894045e-06, + "loss": 0.9294, + "step": 309880 + }, + { + "epoch": 26.82, + "learning_rate": 5.302089051980699e-06, + "loss": 0.8878, + "step": 309890 + }, + { + "epoch": 26.82, + "learning_rate": 5.300644341067352e-06, + "loss": 0.8635, + "step": 309900 + }, + { + "epoch": 26.82, + "learning_rate": 5.299199630154006e-06, + "loss": 0.9082, + "step": 309910 + }, + { + "epoch": 26.83, + "learning_rate": 5.29775491924066e-06, + "loss": 0.945, + "step": 309920 + }, + { + "epoch": 26.83, + "learning_rate": 5.296310208327314e-06, + "loss": 0.9549, + "step": 309930 + }, + { + "epoch": 26.83, + "learning_rate": 5.294865497413968e-06, + "loss": 0.9089, + "step": 309940 + }, + { + "epoch": 26.83, + "learning_rate": 5.293420786500621e-06, + "loss": 0.875, + "step": 309950 + }, + { + "epoch": 26.83, + "learning_rate": 5.291976075587275e-06, + "loss": 0.8664, + "step": 309960 + }, + { + "epoch": 26.83, + "learning_rate": 5.290531364673929e-06, + "loss": 0.9627, + "step": 309970 + }, + { + "epoch": 26.83, + "learning_rate": 5.289086653760583e-06, + "loss": 0.8939, + "step": 309980 + }, + { + "epoch": 26.83, + "learning_rate": 5.287641942847236e-06, + "loss": 0.8957, + "step": 309990 + }, + { + "epoch": 26.83, + "learning_rate": 5.28619723193389e-06, + "loss": 0.9096, + "step": 310000 + }, + { + "epoch": 26.83, + "learning_rate": 5.284752521020544e-06, + "loss": 0.9256, + "step": 310010 + }, + { + "epoch": 26.83, + "learning_rate": 5.283307810107198e-06, + "loss": 0.8982, + "step": 310020 + }, + { + "epoch": 26.84, + "learning_rate": 5.281863099193852e-06, + "loss": 0.9343, + "step": 310030 + }, + { + "epoch": 26.84, + "learning_rate": 5.280418388280505e-06, + "loss": 0.8735, + "step": 310040 + }, + { + "epoch": 26.84, + "learning_rate": 5.278973677367159e-06, + "loss": 0.9224, + "step": 310050 + }, + { + "epoch": 26.84, + "learning_rate": 5.277528966453813e-06, + "loss": 0.8873, + "step": 310060 + }, + { + "epoch": 26.84, + "learning_rate": 5.276084255540467e-06, + "loss": 0.9425, + "step": 310070 + }, + { + "epoch": 26.84, + "learning_rate": 5.27463954462712e-06, + "loss": 0.8806, + "step": 310080 + }, + { + "epoch": 26.84, + "learning_rate": 5.273194833713774e-06, + "loss": 0.909, + "step": 310090 + }, + { + "epoch": 26.84, + "learning_rate": 5.271750122800428e-06, + "loss": 0.9109, + "step": 310100 + }, + { + "epoch": 26.84, + "learning_rate": 5.270305411887082e-06, + "loss": 0.9299, + "step": 310110 + }, + { + "epoch": 26.84, + "learning_rate": 5.268860700973735e-06, + "loss": 0.9297, + "step": 310120 + }, + { + "epoch": 26.84, + "learning_rate": 5.267415990060389e-06, + "loss": 0.9305, + "step": 310130 + }, + { + "epoch": 26.84, + "learning_rate": 5.265971279147043e-06, + "loss": 0.9222, + "step": 310140 + }, + { + "epoch": 26.85, + "learning_rate": 5.264526568233696e-06, + "loss": 0.8768, + "step": 310150 + }, + { + "epoch": 26.85, + "learning_rate": 5.263081857320351e-06, + "loss": 0.8566, + "step": 310160 + }, + { + "epoch": 26.85, + "learning_rate": 5.261637146407004e-06, + "loss": 0.9177, + "step": 310170 + }, + { + "epoch": 26.85, + "learning_rate": 5.260192435493657e-06, + "loss": 0.9193, + "step": 310180 + }, + { + "epoch": 26.85, + "learning_rate": 5.258747724580312e-06, + "loss": 0.922, + "step": 310190 + }, + { + "epoch": 26.85, + "learning_rate": 5.257303013666966e-06, + "loss": 0.89, + "step": 310200 + }, + { + "epoch": 26.85, + "learning_rate": 5.255858302753619e-06, + "loss": 0.9505, + "step": 310210 + }, + { + "epoch": 26.85, + "learning_rate": 5.254413591840273e-06, + "loss": 0.8664, + "step": 310220 + }, + { + "epoch": 26.85, + "learning_rate": 5.252968880926927e-06, + "loss": 0.9648, + "step": 310230 + }, + { + "epoch": 26.85, + "learning_rate": 5.25152417001358e-06, + "loss": 0.9089, + "step": 310240 + }, + { + "epoch": 26.85, + "learning_rate": 5.2500794591002346e-06, + "loss": 0.8635, + "step": 310250 + }, + { + "epoch": 26.86, + "learning_rate": 5.248634748186888e-06, + "loss": 0.9333, + "step": 310260 + }, + { + "epoch": 26.86, + "learning_rate": 5.247190037273541e-06, + "loss": 0.9217, + "step": 310270 + }, + { + "epoch": 26.86, + "learning_rate": 5.245745326360196e-06, + "loss": 0.8872, + "step": 310280 + }, + { + "epoch": 26.86, + "learning_rate": 5.24430061544685e-06, + "loss": 0.9206, + "step": 310290 + }, + { + "epoch": 26.86, + "learning_rate": 5.242855904533503e-06, + "loss": 0.8879, + "step": 310300 + }, + { + "epoch": 26.86, + "learning_rate": 5.241411193620157e-06, + "loss": 0.8617, + "step": 310310 + }, + { + "epoch": 26.86, + "learning_rate": 5.239966482706811e-06, + "loss": 0.9246, + "step": 310320 + }, + { + "epoch": 26.86, + "learning_rate": 5.238521771793464e-06, + "loss": 0.9484, + "step": 310330 + }, + { + "epoch": 26.86, + "learning_rate": 5.2370770608801185e-06, + "loss": 0.9303, + "step": 310340 + }, + { + "epoch": 26.86, + "learning_rate": 5.235632349966772e-06, + "loss": 0.9092, + "step": 310350 + }, + { + "epoch": 26.86, + "learning_rate": 5.234187639053425e-06, + "loss": 0.9271, + "step": 310360 + }, + { + "epoch": 26.86, + "learning_rate": 5.2327429281400795e-06, + "loss": 0.9086, + "step": 310370 + }, + { + "epoch": 26.87, + "learning_rate": 5.231298217226733e-06, + "loss": 0.9182, + "step": 310380 + }, + { + "epoch": 26.87, + "learning_rate": 5.229853506313387e-06, + "loss": 0.9155, + "step": 310390 + }, + { + "epoch": 26.87, + "learning_rate": 5.2284087954000406e-06, + "loss": 0.925, + "step": 310400 + }, + { + "epoch": 26.87, + "learning_rate": 5.226964084486694e-06, + "loss": 0.9226, + "step": 310410 + }, + { + "epoch": 26.87, + "learning_rate": 5.225519373573348e-06, + "loss": 0.9201, + "step": 310420 + }, + { + "epoch": 26.87, + "learning_rate": 5.2240746626600024e-06, + "loss": 0.9229, + "step": 310430 + }, + { + "epoch": 26.87, + "learning_rate": 5.222629951746656e-06, + "loss": 0.8676, + "step": 310440 + }, + { + "epoch": 26.87, + "learning_rate": 5.221185240833309e-06, + "loss": 0.8081, + "step": 310450 + }, + { + "epoch": 26.87, + "learning_rate": 5.2197405299199635e-06, + "loss": 0.9472, + "step": 310460 + }, + { + "epoch": 26.87, + "learning_rate": 5.218295819006617e-06, + "loss": 0.8789, + "step": 310470 + }, + { + "epoch": 26.87, + "learning_rate": 5.216851108093271e-06, + "loss": 0.9168, + "step": 310480 + }, + { + "epoch": 26.88, + "learning_rate": 5.2154063971799245e-06, + "loss": 0.9223, + "step": 310490 + }, + { + "epoch": 26.88, + "learning_rate": 5.213961686266578e-06, + "loss": 0.9983, + "step": 310500 + }, + { + "epoch": 26.88, + "learning_rate": 5.212516975353232e-06, + "loss": 0.8621, + "step": 310510 + }, + { + "epoch": 26.88, + "learning_rate": 5.211072264439886e-06, + "loss": 0.9252, + "step": 310520 + }, + { + "epoch": 26.88, + "learning_rate": 5.20962755352654e-06, + "loss": 0.9705, + "step": 310530 + }, + { + "epoch": 26.88, + "learning_rate": 5.208182842613193e-06, + "loss": 0.9525, + "step": 310540 + }, + { + "epoch": 26.88, + "learning_rate": 5.206738131699847e-06, + "loss": 0.904, + "step": 310550 + }, + { + "epoch": 26.88, + "learning_rate": 5.205293420786501e-06, + "loss": 0.9031, + "step": 310560 + }, + { + "epoch": 26.88, + "learning_rate": 5.203848709873155e-06, + "loss": 0.8747, + "step": 310570 + }, + { + "epoch": 26.88, + "learning_rate": 5.2024039989598084e-06, + "loss": 0.8838, + "step": 310580 + }, + { + "epoch": 26.88, + "learning_rate": 5.200959288046462e-06, + "loss": 0.9265, + "step": 310590 + }, + { + "epoch": 26.88, + "learning_rate": 5.199514577133116e-06, + "loss": 0.879, + "step": 310600 + }, + { + "epoch": 26.89, + "learning_rate": 5.1980698662197695e-06, + "loss": 0.9042, + "step": 310610 + }, + { + "epoch": 26.89, + "learning_rate": 5.196625155306424e-06, + "loss": 0.9052, + "step": 310620 + }, + { + "epoch": 26.89, + "learning_rate": 5.195180444393077e-06, + "loss": 0.9258, + "step": 310630 + }, + { + "epoch": 26.89, + "learning_rate": 5.1937357334797305e-06, + "loss": 0.8999, + "step": 310640 + }, + { + "epoch": 26.89, + "learning_rate": 5.192291022566385e-06, + "loss": 0.9548, + "step": 310650 + }, + { + "epoch": 26.89, + "learning_rate": 5.190846311653039e-06, + "loss": 0.9155, + "step": 310660 + }, + { + "epoch": 26.89, + "learning_rate": 5.189401600739692e-06, + "loss": 0.8965, + "step": 310670 + }, + { + "epoch": 26.89, + "learning_rate": 5.187956889826346e-06, + "loss": 0.9006, + "step": 310680 + }, + { + "epoch": 26.89, + "learning_rate": 5.186512178913e-06, + "loss": 0.9501, + "step": 310690 + }, + { + "epoch": 26.89, + "learning_rate": 5.185067467999653e-06, + "loss": 0.8717, + "step": 310700 + }, + { + "epoch": 26.89, + "learning_rate": 5.183622757086308e-06, + "loss": 0.9046, + "step": 310710 + }, + { + "epoch": 26.9, + "learning_rate": 5.182178046172961e-06, + "loss": 0.9272, + "step": 310720 + }, + { + "epoch": 26.9, + "learning_rate": 5.1807333352596145e-06, + "loss": 0.9442, + "step": 310730 + }, + { + "epoch": 26.9, + "learning_rate": 5.179288624346269e-06, + "loss": 0.8676, + "step": 310740 + }, + { + "epoch": 26.9, + "learning_rate": 5.177843913432923e-06, + "loss": 0.9222, + "step": 310750 + }, + { + "epoch": 26.9, + "learning_rate": 5.176399202519576e-06, + "loss": 0.8875, + "step": 310760 + }, + { + "epoch": 26.9, + "learning_rate": 5.17495449160623e-06, + "loss": 0.9423, + "step": 310770 + }, + { + "epoch": 26.9, + "learning_rate": 5.173509780692884e-06, + "loss": 0.9466, + "step": 310780 + }, + { + "epoch": 26.9, + "learning_rate": 5.172065069779537e-06, + "loss": 0.9673, + "step": 310790 + }, + { + "epoch": 26.9, + "learning_rate": 5.170620358866192e-06, + "loss": 0.902, + "step": 310800 + }, + { + "epoch": 26.9, + "learning_rate": 5.169175647952844e-06, + "loss": 0.902, + "step": 310810 + }, + { + "epoch": 26.9, + "learning_rate": 5.167730937039498e-06, + "loss": 0.8499, + "step": 310820 + }, + { + "epoch": 26.9, + "learning_rate": 5.166286226126153e-06, + "loss": 0.9301, + "step": 310830 + }, + { + "epoch": 26.91, + "learning_rate": 5.164841515212806e-06, + "loss": 0.8803, + "step": 310840 + }, + { + "epoch": 26.91, + "learning_rate": 5.16339680429946e-06, + "loss": 0.8567, + "step": 310850 + }, + { + "epoch": 26.91, + "learning_rate": 5.161952093386114e-06, + "loss": 0.8759, + "step": 310860 + }, + { + "epoch": 26.91, + "learning_rate": 5.160507382472767e-06, + "loss": 0.9106, + "step": 310870 + }, + { + "epoch": 26.91, + "learning_rate": 5.159062671559421e-06, + "loss": 0.88, + "step": 310880 + }, + { + "epoch": 26.91, + "learning_rate": 5.1576179606460755e-06, + "loss": 0.8922, + "step": 310890 + }, + { + "epoch": 26.91, + "learning_rate": 5.156173249732728e-06, + "loss": 0.8882, + "step": 310900 + }, + { + "epoch": 26.91, + "learning_rate": 5.154728538819382e-06, + "loss": 0.8968, + "step": 310910 + }, + { + "epoch": 26.91, + "learning_rate": 5.153283827906037e-06, + "loss": 0.8683, + "step": 310920 + }, + { + "epoch": 26.91, + "learning_rate": 5.15183911699269e-06, + "loss": 0.9076, + "step": 310930 + }, + { + "epoch": 26.91, + "learning_rate": 5.150394406079344e-06, + "loss": 0.9486, + "step": 310940 + }, + { + "epoch": 26.91, + "learning_rate": 5.148949695165998e-06, + "loss": 0.9475, + "step": 310950 + }, + { + "epoch": 26.92, + "learning_rate": 5.147504984252651e-06, + "loss": 0.9507, + "step": 310960 + }, + { + "epoch": 26.92, + "learning_rate": 5.146060273339305e-06, + "loss": 0.9671, + "step": 310970 + }, + { + "epoch": 26.92, + "learning_rate": 5.1446155624259595e-06, + "loss": 0.8819, + "step": 310980 + }, + { + "epoch": 26.92, + "learning_rate": 5.143170851512612e-06, + "loss": 0.9188, + "step": 310990 + }, + { + "epoch": 26.92, + "learning_rate": 5.141726140599266e-06, + "loss": 0.9095, + "step": 311000 + }, + { + "epoch": 26.92, + "learning_rate": 5.14028142968592e-06, + "loss": 0.8733, + "step": 311010 + }, + { + "epoch": 26.92, + "learning_rate": 5.138836718772574e-06, + "loss": 0.9304, + "step": 311020 + }, + { + "epoch": 26.92, + "learning_rate": 5.137392007859228e-06, + "loss": 0.8913, + "step": 311030 + }, + { + "epoch": 26.92, + "learning_rate": 5.135947296945881e-06, + "loss": 0.8839, + "step": 311040 + }, + { + "epoch": 26.92, + "learning_rate": 5.134502586032535e-06, + "loss": 0.9094, + "step": 311050 + }, + { + "epoch": 26.92, + "learning_rate": 5.133057875119189e-06, + "loss": 0.896, + "step": 311060 + }, + { + "epoch": 26.93, + "learning_rate": 5.131613164205843e-06, + "loss": 0.9341, + "step": 311070 + }, + { + "epoch": 26.93, + "learning_rate": 5.130168453292496e-06, + "loss": 0.9009, + "step": 311080 + }, + { + "epoch": 26.93, + "learning_rate": 5.12872374237915e-06, + "loss": 1.0104, + "step": 311090 + }, + { + "epoch": 26.93, + "learning_rate": 5.127279031465804e-06, + "loss": 0.8896, + "step": 311100 + }, + { + "epoch": 26.93, + "learning_rate": 5.125834320552458e-06, + "loss": 0.9393, + "step": 311110 + }, + { + "epoch": 26.93, + "learning_rate": 5.124389609639112e-06, + "loss": 0.8959, + "step": 311120 + }, + { + "epoch": 26.93, + "learning_rate": 5.122944898725765e-06, + "loss": 0.8798, + "step": 311130 + }, + { + "epoch": 26.93, + "learning_rate": 5.121500187812419e-06, + "loss": 0.8904, + "step": 311140 + }, + { + "epoch": 26.93, + "learning_rate": 5.120055476899073e-06, + "loss": 0.8893, + "step": 311150 + }, + { + "epoch": 26.93, + "learning_rate": 5.1186107659857265e-06, + "loss": 0.8747, + "step": 311160 + }, + { + "epoch": 26.93, + "learning_rate": 5.11716605507238e-06, + "loss": 0.9515, + "step": 311170 + }, + { + "epoch": 26.93, + "learning_rate": 5.115721344159034e-06, + "loss": 0.8902, + "step": 311180 + }, + { + "epoch": 26.94, + "learning_rate": 5.1142766332456876e-06, + "loss": 0.8564, + "step": 311190 + }, + { + "epoch": 26.94, + "learning_rate": 5.112831922332342e-06, + "loss": 0.8944, + "step": 311200 + }, + { + "epoch": 26.94, + "learning_rate": 5.111387211418996e-06, + "loss": 0.8728, + "step": 311210 + }, + { + "epoch": 26.94, + "learning_rate": 5.109942500505649e-06, + "loss": 0.9155, + "step": 311220 + }, + { + "epoch": 26.94, + "learning_rate": 5.108497789592303e-06, + "loss": 0.9046, + "step": 311230 + }, + { + "epoch": 26.94, + "learning_rate": 5.107053078678956e-06, + "loss": 0.8975, + "step": 311240 + }, + { + "epoch": 26.94, + "learning_rate": 5.1056083677656105e-06, + "loss": 0.8902, + "step": 311250 + }, + { + "epoch": 26.94, + "learning_rate": 5.104163656852264e-06, + "loss": 0.9081, + "step": 311260 + }, + { + "epoch": 26.94, + "learning_rate": 5.102718945938917e-06, + "loss": 0.934, + "step": 311270 + }, + { + "epoch": 26.94, + "learning_rate": 5.1012742350255715e-06, + "loss": 0.9868, + "step": 311280 + }, + { + "epoch": 26.94, + "learning_rate": 5.099829524112226e-06, + "loss": 0.8931, + "step": 311290 + }, + { + "epoch": 26.95, + "learning_rate": 5.098384813198879e-06, + "loss": 0.8774, + "step": 311300 + }, + { + "epoch": 26.95, + "learning_rate": 5.0969401022855325e-06, + "loss": 0.8804, + "step": 311310 + }, + { + "epoch": 26.95, + "learning_rate": 5.095495391372187e-06, + "loss": 0.8732, + "step": 311320 + }, + { + "epoch": 26.95, + "learning_rate": 5.09405068045884e-06, + "loss": 0.9047, + "step": 311330 + }, + { + "epoch": 26.95, + "learning_rate": 5.092605969545494e-06, + "loss": 0.9577, + "step": 311340 + }, + { + "epoch": 26.95, + "learning_rate": 5.091161258632148e-06, + "loss": 0.8951, + "step": 311350 + }, + { + "epoch": 26.95, + "learning_rate": 5.089716547718801e-06, + "loss": 0.9838, + "step": 311360 + }, + { + "epoch": 26.95, + "learning_rate": 5.0882718368054554e-06, + "loss": 0.9074, + "step": 311370 + }, + { + "epoch": 26.95, + "learning_rate": 5.08682712589211e-06, + "loss": 0.8652, + "step": 311380 + }, + { + "epoch": 26.95, + "learning_rate": 5.085382414978763e-06, + "loss": 0.9366, + "step": 311390 + }, + { + "epoch": 26.95, + "learning_rate": 5.0839377040654165e-06, + "loss": 0.9285, + "step": 311400 + }, + { + "epoch": 26.95, + "learning_rate": 5.082492993152071e-06, + "loss": 0.8968, + "step": 311410 + }, + { + "epoch": 26.96, + "learning_rate": 5.081048282238724e-06, + "loss": 0.9002, + "step": 311420 + }, + { + "epoch": 26.96, + "learning_rate": 5.079603571325378e-06, + "loss": 0.9594, + "step": 311430 + }, + { + "epoch": 26.96, + "learning_rate": 5.078158860412032e-06, + "loss": 0.8866, + "step": 311440 + }, + { + "epoch": 26.96, + "learning_rate": 5.076714149498685e-06, + "loss": 0.9117, + "step": 311450 + }, + { + "epoch": 26.96, + "learning_rate": 5.075269438585339e-06, + "loss": 0.9455, + "step": 311460 + }, + { + "epoch": 26.96, + "learning_rate": 5.073824727671993e-06, + "loss": 0.9255, + "step": 311470 + }, + { + "epoch": 26.96, + "learning_rate": 5.072380016758647e-06, + "loss": 0.9039, + "step": 311480 + }, + { + "epoch": 26.96, + "learning_rate": 5.0709353058453e-06, + "loss": 0.8777, + "step": 311490 + }, + { + "epoch": 26.96, + "learning_rate": 5.069490594931954e-06, + "loss": 0.8567, + "step": 311500 + }, + { + "epoch": 26.96, + "learning_rate": 5.068045884018608e-06, + "loss": 0.8663, + "step": 311510 + }, + { + "epoch": 26.96, + "learning_rate": 5.066601173105262e-06, + "loss": 0.9323, + "step": 311520 + }, + { + "epoch": 26.97, + "learning_rate": 5.065156462191916e-06, + "loss": 0.8798, + "step": 311530 + }, + { + "epoch": 26.97, + "learning_rate": 5.063711751278569e-06, + "loss": 0.9118, + "step": 311540 + }, + { + "epoch": 26.97, + "learning_rate": 5.062267040365223e-06, + "loss": 0.9329, + "step": 311550 + }, + { + "epoch": 26.97, + "learning_rate": 5.060822329451877e-06, + "loss": 0.8827, + "step": 311560 + }, + { + "epoch": 26.97, + "learning_rate": 5.059377618538531e-06, + "loss": 0.8642, + "step": 311570 + }, + { + "epoch": 26.97, + "learning_rate": 5.057932907625184e-06, + "loss": 0.8908, + "step": 311580 + }, + { + "epoch": 26.97, + "learning_rate": 5.056488196711838e-06, + "loss": 0.9409, + "step": 311590 + }, + { + "epoch": 26.97, + "learning_rate": 5.055043485798492e-06, + "loss": 0.8687, + "step": 311600 + }, + { + "epoch": 26.97, + "learning_rate": 5.053598774885146e-06, + "loss": 0.9265, + "step": 311610 + }, + { + "epoch": 26.97, + "learning_rate": 5.0521540639718e-06, + "loss": 0.8518, + "step": 311620 + }, + { + "epoch": 26.97, + "learning_rate": 5.050709353058453e-06, + "loss": 0.9393, + "step": 311630 + }, + { + "epoch": 26.97, + "learning_rate": 5.049264642145107e-06, + "loss": 0.9397, + "step": 311640 + }, + { + "epoch": 26.98, + "learning_rate": 5.047819931231761e-06, + "loss": 0.9299, + "step": 311650 + }, + { + "epoch": 26.98, + "learning_rate": 5.046375220318415e-06, + "loss": 0.8983, + "step": 311660 + }, + { + "epoch": 26.98, + "learning_rate": 5.044930509405068e-06, + "loss": 0.8901, + "step": 311670 + }, + { + "epoch": 26.98, + "learning_rate": 5.043485798491722e-06, + "loss": 0.893, + "step": 311680 + }, + { + "epoch": 26.98, + "learning_rate": 5.042041087578376e-06, + "loss": 0.8788, + "step": 311690 + }, + { + "epoch": 26.98, + "learning_rate": 5.040596376665029e-06, + "loss": 0.9088, + "step": 311700 + }, + { + "epoch": 26.98, + "learning_rate": 5.039151665751684e-06, + "loss": 0.8644, + "step": 311710 + }, + { + "epoch": 26.98, + "learning_rate": 5.037706954838337e-06, + "loss": 0.9134, + "step": 311720 + }, + { + "epoch": 26.98, + "learning_rate": 5.03626224392499e-06, + "loss": 0.9472, + "step": 311730 + }, + { + "epoch": 26.98, + "learning_rate": 5.034817533011645e-06, + "loss": 0.9787, + "step": 311740 + }, + { + "epoch": 26.98, + "learning_rate": 5.033372822098299e-06, + "loss": 0.9148, + "step": 311750 + }, + { + "epoch": 26.99, + "learning_rate": 5.031928111184952e-06, + "loss": 0.9163, + "step": 311760 + }, + { + "epoch": 26.99, + "learning_rate": 5.030483400271606e-06, + "loss": 0.8816, + "step": 311770 + }, + { + "epoch": 26.99, + "learning_rate": 5.02903868935826e-06, + "loss": 0.8798, + "step": 311780 + }, + { + "epoch": 26.99, + "learning_rate": 5.027593978444913e-06, + "loss": 0.9129, + "step": 311790 + }, + { + "epoch": 26.99, + "learning_rate": 5.0261492675315675e-06, + "loss": 0.9245, + "step": 311800 + }, + { + "epoch": 26.99, + "learning_rate": 5.024704556618221e-06, + "loss": 0.9018, + "step": 311810 + }, + { + "epoch": 26.99, + "learning_rate": 5.023259845704874e-06, + "loss": 0.9393, + "step": 311820 + }, + { + "epoch": 26.99, + "learning_rate": 5.0218151347915286e-06, + "loss": 0.9033, + "step": 311830 + }, + { + "epoch": 26.99, + "learning_rate": 5.020370423878183e-06, + "loss": 0.852, + "step": 311840 + }, + { + "epoch": 26.99, + "learning_rate": 5.018925712964836e-06, + "loss": 0.9086, + "step": 311850 + }, + { + "epoch": 26.99, + "learning_rate": 5.01748100205149e-06, + "loss": 0.931, + "step": 311860 + }, + { + "epoch": 26.99, + "learning_rate": 5.016036291138144e-06, + "loss": 0.8866, + "step": 311870 + }, + { + "epoch": 27.0, + "learning_rate": 5.014591580224797e-06, + "loss": 0.9278, + "step": 311880 + }, + { + "epoch": 27.0, + "learning_rate": 5.0131468693114515e-06, + "loss": 0.9678, + "step": 311890 + }, + { + "epoch": 27.0, + "learning_rate": 5.011702158398105e-06, + "loss": 0.9435, + "step": 311900 + }, + { + "epoch": 27.0, + "learning_rate": 5.010257447484758e-06, + "loss": 0.9203, + "step": 311910 + }, + { + "epoch": 27.0, + "learning_rate": 5.0088127365714125e-06, + "loss": 0.9126, + "step": 311920 + }, + { + "epoch": 27.0, + "learning_rate": 5.007368025658066e-06, + "loss": 0.9219, + "step": 311930 + }, + { + "epoch": 27.0, + "eval_cer": 0.9930125039575515, + "eval_em": 0, + "eval_f1": 0.12251770767890295, + "eval_loss": 0.9328575730323792, + "eval_runtime": 900.6553, + "eval_samples_per_second": 5.701, + "eval_steps_per_second": 0.713, + "eval_wer": 0.965943656966526, + "step": 311932 + }, + { + "epoch": 27.0, + "learning_rate": 5.00592331474472e-06, + "loss": 0.8964, + "step": 311940 + }, + { + "epoch": 27.0, + "learning_rate": 5.0044786038313735e-06, + "loss": 0.9435, + "step": 311950 + }, + { + "epoch": 27.0, + "learning_rate": 5.003033892918027e-06, + "loss": 0.9353, + "step": 311960 + }, + { + "epoch": 27.0, + "learning_rate": 5.001589182004681e-06, + "loss": 0.9148, + "step": 311970 + }, + { + "epoch": 27.0, + "learning_rate": 5.000144471091335e-06, + "loss": 0.921, + "step": 311980 + }, + { + "epoch": 27.0, + "learning_rate": 4.998699760177988e-06, + "loss": 0.9466, + "step": 311990 + }, + { + "epoch": 27.01, + "learning_rate": 4.997255049264642e-06, + "loss": 0.9074, + "step": 312000 + }, + { + "epoch": 27.01, + "learning_rate": 4.9958103383512964e-06, + "loss": 0.867, + "step": 312010 + }, + { + "epoch": 27.01, + "learning_rate": 4.99436562743795e-06, + "loss": 0.8855, + "step": 312020 + }, + { + "epoch": 27.01, + "learning_rate": 4.992920916524604e-06, + "loss": 0.8629, + "step": 312030 + }, + { + "epoch": 27.01, + "learning_rate": 4.9914762056112575e-06, + "loss": 0.8403, + "step": 312040 + }, + { + "epoch": 27.01, + "learning_rate": 4.990031494697911e-06, + "loss": 0.9972, + "step": 312050 + }, + { + "epoch": 27.01, + "learning_rate": 4.988586783784565e-06, + "loss": 0.9431, + "step": 312060 + }, + { + "epoch": 27.01, + "learning_rate": 4.987142072871219e-06, + "loss": 0.9049, + "step": 312070 + }, + { + "epoch": 27.01, + "learning_rate": 4.985697361957872e-06, + "loss": 0.8824, + "step": 312080 + }, + { + "epoch": 27.01, + "learning_rate": 4.984252651044526e-06, + "loss": 0.8349, + "step": 312090 + }, + { + "epoch": 27.01, + "learning_rate": 4.98280794013118e-06, + "loss": 0.903, + "step": 312100 + }, + { + "epoch": 27.02, + "learning_rate": 4.981363229217834e-06, + "loss": 0.9053, + "step": 312110 + }, + { + "epoch": 27.02, + "learning_rate": 4.979918518304488e-06, + "loss": 0.9024, + "step": 312120 + }, + { + "epoch": 27.02, + "learning_rate": 4.978473807391141e-06, + "loss": 0.8684, + "step": 312130 + }, + { + "epoch": 27.02, + "learning_rate": 4.977029096477795e-06, + "loss": 0.8365, + "step": 312140 + }, + { + "epoch": 27.02, + "learning_rate": 4.975584385564449e-06, + "loss": 0.8525, + "step": 312150 + }, + { + "epoch": 27.02, + "learning_rate": 4.9741396746511024e-06, + "loss": 0.8887, + "step": 312160 + }, + { + "epoch": 27.02, + "learning_rate": 4.972694963737756e-06, + "loss": 0.9607, + "step": 312170 + }, + { + "epoch": 27.02, + "learning_rate": 4.97125025282441e-06, + "loss": 0.9115, + "step": 312180 + }, + { + "epoch": 27.02, + "learning_rate": 4.9698055419110635e-06, + "loss": 0.8964, + "step": 312190 + }, + { + "epoch": 27.02, + "learning_rate": 4.968360830997718e-06, + "loss": 0.9058, + "step": 312200 + }, + { + "epoch": 27.02, + "learning_rate": 4.966916120084372e-06, + "loss": 0.9001, + "step": 312210 + }, + { + "epoch": 27.02, + "learning_rate": 4.9654714091710245e-06, + "loss": 0.8845, + "step": 312220 + }, + { + "epoch": 27.03, + "learning_rate": 4.964026698257679e-06, + "loss": 0.8222, + "step": 312230 + }, + { + "epoch": 27.03, + "learning_rate": 4.962581987344333e-06, + "loss": 0.8654, + "step": 312240 + }, + { + "epoch": 27.03, + "learning_rate": 4.961137276430986e-06, + "loss": 0.8984, + "step": 312250 + }, + { + "epoch": 27.03, + "learning_rate": 4.95969256551764e-06, + "loss": 0.879, + "step": 312260 + }, + { + "epoch": 27.03, + "learning_rate": 4.958247854604294e-06, + "loss": 0.894, + "step": 312270 + }, + { + "epoch": 27.03, + "learning_rate": 4.956803143690947e-06, + "loss": 0.9159, + "step": 312280 + }, + { + "epoch": 27.03, + "learning_rate": 4.955358432777602e-06, + "loss": 0.8665, + "step": 312290 + }, + { + "epoch": 27.03, + "learning_rate": 4.953913721864255e-06, + "loss": 0.9149, + "step": 312300 + }, + { + "epoch": 27.03, + "learning_rate": 4.9524690109509085e-06, + "loss": 0.8828, + "step": 312310 + }, + { + "epoch": 27.03, + "learning_rate": 4.951024300037563e-06, + "loss": 0.9201, + "step": 312320 + }, + { + "epoch": 27.03, + "learning_rate": 4.949579589124217e-06, + "loss": 0.9229, + "step": 312330 + }, + { + "epoch": 27.04, + "learning_rate": 4.94813487821087e-06, + "loss": 0.8964, + "step": 312340 + }, + { + "epoch": 27.04, + "learning_rate": 4.946690167297524e-06, + "loss": 0.857, + "step": 312350 + }, + { + "epoch": 27.04, + "learning_rate": 4.945245456384178e-06, + "loss": 0.9304, + "step": 312360 + }, + { + "epoch": 27.04, + "learning_rate": 4.943800745470831e-06, + "loss": 0.8854, + "step": 312370 + }, + { + "epoch": 27.04, + "learning_rate": 4.942356034557486e-06, + "loss": 0.8677, + "step": 312380 + }, + { + "epoch": 27.04, + "learning_rate": 4.940911323644139e-06, + "loss": 0.9024, + "step": 312390 + }, + { + "epoch": 27.04, + "learning_rate": 4.939466612730792e-06, + "loss": 0.8762, + "step": 312400 + }, + { + "epoch": 27.04, + "learning_rate": 4.938021901817447e-06, + "loss": 0.9228, + "step": 312410 + }, + { + "epoch": 27.04, + "learning_rate": 4.9365771909041e-06, + "loss": 0.8937, + "step": 312420 + }, + { + "epoch": 27.04, + "learning_rate": 4.935132479990754e-06, + "loss": 0.9098, + "step": 312430 + }, + { + "epoch": 27.04, + "learning_rate": 4.933687769077408e-06, + "loss": 0.8967, + "step": 312440 + }, + { + "epoch": 27.04, + "learning_rate": 4.932243058164061e-06, + "loss": 0.8901, + "step": 312450 + }, + { + "epoch": 27.05, + "learning_rate": 4.930798347250715e-06, + "loss": 0.95, + "step": 312460 + }, + { + "epoch": 27.05, + "learning_rate": 4.9293536363373695e-06, + "loss": 0.9194, + "step": 312470 + }, + { + "epoch": 27.05, + "learning_rate": 4.927908925424023e-06, + "loss": 0.9472, + "step": 312480 + }, + { + "epoch": 27.05, + "learning_rate": 4.926464214510676e-06, + "loss": 0.8881, + "step": 312490 + }, + { + "epoch": 27.05, + "learning_rate": 4.925019503597331e-06, + "loss": 0.9115, + "step": 312500 + }, + { + "epoch": 27.05, + "learning_rate": 4.923574792683984e-06, + "loss": 0.8695, + "step": 312510 + }, + { + "epoch": 27.05, + "learning_rate": 4.922130081770638e-06, + "loss": 0.8667, + "step": 312520 + }, + { + "epoch": 27.05, + "learning_rate": 4.920685370857292e-06, + "loss": 0.8977, + "step": 312530 + }, + { + "epoch": 27.05, + "learning_rate": 4.919240659943945e-06, + "loss": 0.8749, + "step": 312540 + }, + { + "epoch": 27.05, + "learning_rate": 4.917795949030599e-06, + "loss": 0.8644, + "step": 312550 + }, + { + "epoch": 27.05, + "learning_rate": 4.9163512381172535e-06, + "loss": 0.9218, + "step": 312560 + }, + { + "epoch": 27.06, + "learning_rate": 4.914906527203907e-06, + "loss": 0.9747, + "step": 312570 + }, + { + "epoch": 27.06, + "learning_rate": 4.91346181629056e-06, + "loss": 0.9487, + "step": 312580 + }, + { + "epoch": 27.06, + "learning_rate": 4.9120171053772145e-06, + "loss": 0.9193, + "step": 312590 + }, + { + "epoch": 27.06, + "learning_rate": 4.910572394463868e-06, + "loss": 0.9454, + "step": 312600 + }, + { + "epoch": 27.06, + "learning_rate": 4.909127683550522e-06, + "loss": 0.9259, + "step": 312610 + }, + { + "epoch": 27.06, + "learning_rate": 4.9076829726371756e-06, + "loss": 0.894, + "step": 312620 + }, + { + "epoch": 27.06, + "learning_rate": 4.906238261723829e-06, + "loss": 0.8935, + "step": 312630 + }, + { + "epoch": 27.06, + "learning_rate": 4.904793550810483e-06, + "loss": 0.9358, + "step": 312640 + }, + { + "epoch": 27.06, + "learning_rate": 4.903348839897137e-06, + "loss": 0.8848, + "step": 312650 + }, + { + "epoch": 27.06, + "learning_rate": 4.901904128983791e-06, + "loss": 0.9102, + "step": 312660 + }, + { + "epoch": 27.06, + "learning_rate": 4.900459418070444e-06, + "loss": 0.8984, + "step": 312670 + }, + { + "epoch": 27.06, + "learning_rate": 4.899014707157098e-06, + "loss": 0.8667, + "step": 312680 + }, + { + "epoch": 27.07, + "learning_rate": 4.897569996243752e-06, + "loss": 0.8791, + "step": 312690 + }, + { + "epoch": 27.07, + "learning_rate": 4.896125285330406e-06, + "loss": 0.8559, + "step": 312700 + }, + { + "epoch": 27.07, + "learning_rate": 4.8946805744170595e-06, + "loss": 0.896, + "step": 312710 + }, + { + "epoch": 27.07, + "learning_rate": 4.893235863503713e-06, + "loss": 0.9095, + "step": 312720 + }, + { + "epoch": 27.07, + "learning_rate": 4.891791152590367e-06, + "loss": 0.9301, + "step": 312730 + }, + { + "epoch": 27.07, + "learning_rate": 4.8903464416770205e-06, + "loss": 0.8969, + "step": 312740 + }, + { + "epoch": 27.07, + "learning_rate": 4.888901730763675e-06, + "loss": 0.8975, + "step": 312750 + }, + { + "epoch": 27.07, + "learning_rate": 4.887457019850328e-06, + "loss": 0.8658, + "step": 312760 + }, + { + "epoch": 27.07, + "learning_rate": 4.8860123089369816e-06, + "loss": 0.8696, + "step": 312770 + }, + { + "epoch": 27.07, + "learning_rate": 4.884567598023636e-06, + "loss": 0.9864, + "step": 312780 + }, + { + "epoch": 27.07, + "learning_rate": 4.88312288711029e-06, + "loss": 0.8866, + "step": 312790 + }, + { + "epoch": 27.08, + "learning_rate": 4.8816781761969434e-06, + "loss": 0.8755, + "step": 312800 + }, + { + "epoch": 27.08, + "learning_rate": 4.880233465283597e-06, + "loss": 0.9019, + "step": 312810 + }, + { + "epoch": 27.08, + "learning_rate": 4.87878875437025e-06, + "loss": 0.9105, + "step": 312820 + }, + { + "epoch": 27.08, + "learning_rate": 4.8773440434569045e-06, + "loss": 0.9611, + "step": 312830 + }, + { + "epoch": 27.08, + "learning_rate": 4.875899332543559e-06, + "loss": 0.8989, + "step": 312840 + }, + { + "epoch": 27.08, + "learning_rate": 4.874454621630212e-06, + "loss": 0.8624, + "step": 312850 + }, + { + "epoch": 27.08, + "learning_rate": 4.8730099107168655e-06, + "loss": 0.9168, + "step": 312860 + }, + { + "epoch": 27.08, + "learning_rate": 4.87156519980352e-06, + "loss": 0.9164, + "step": 312870 + }, + { + "epoch": 27.08, + "learning_rate": 4.870120488890173e-06, + "loss": 0.85, + "step": 312880 + }, + { + "epoch": 27.08, + "learning_rate": 4.868675777976827e-06, + "loss": 0.8606, + "step": 312890 + }, + { + "epoch": 27.08, + "learning_rate": 4.867231067063481e-06, + "loss": 0.9082, + "step": 312900 + }, + { + "epoch": 27.08, + "learning_rate": 4.865786356150134e-06, + "loss": 0.8445, + "step": 312910 + }, + { + "epoch": 27.09, + "learning_rate": 4.864341645236788e-06, + "loss": 0.9099, + "step": 312920 + }, + { + "epoch": 27.09, + "learning_rate": 4.862896934323443e-06, + "loss": 0.8694, + "step": 312930 + }, + { + "epoch": 27.09, + "learning_rate": 4.861452223410096e-06, + "loss": 0.8804, + "step": 312940 + }, + { + "epoch": 27.09, + "learning_rate": 4.8600075124967494e-06, + "loss": 0.9016, + "step": 312950 + }, + { + "epoch": 27.09, + "learning_rate": 4.858562801583404e-06, + "loss": 0.9559, + "step": 312960 + }, + { + "epoch": 27.09, + "learning_rate": 4.857118090670057e-06, + "loss": 0.9007, + "step": 312970 + }, + { + "epoch": 27.09, + "learning_rate": 4.855673379756711e-06, + "loss": 0.8984, + "step": 312980 + }, + { + "epoch": 27.09, + "learning_rate": 4.854228668843365e-06, + "loss": 0.9286, + "step": 312990 + }, + { + "epoch": 27.09, + "learning_rate": 4.852783957930018e-06, + "loss": 0.8864, + "step": 313000 + }, + { + "epoch": 27.09, + "learning_rate": 4.851339247016672e-06, + "loss": 0.9378, + "step": 313010 + }, + { + "epoch": 27.09, + "learning_rate": 4.849894536103327e-06, + "loss": 0.8855, + "step": 313020 + }, + { + "epoch": 27.09, + "learning_rate": 4.84844982518998e-06, + "loss": 0.9205, + "step": 313030 + }, + { + "epoch": 27.1, + "learning_rate": 4.847005114276633e-06, + "loss": 0.9484, + "step": 313040 + }, + { + "epoch": 27.1, + "learning_rate": 4.845560403363287e-06, + "loss": 0.8858, + "step": 313050 + }, + { + "epoch": 27.1, + "learning_rate": 4.844115692449941e-06, + "loss": 0.9214, + "step": 313060 + }, + { + "epoch": 27.1, + "learning_rate": 4.842670981536595e-06, + "loss": 0.9137, + "step": 313070 + }, + { + "epoch": 27.1, + "learning_rate": 4.841226270623248e-06, + "loss": 0.8785, + "step": 313080 + }, + { + "epoch": 27.1, + "learning_rate": 4.839781559709902e-06, + "loss": 0.9112, + "step": 313090 + }, + { + "epoch": 27.1, + "learning_rate": 4.838336848796556e-06, + "loss": 0.9094, + "step": 313100 + }, + { + "epoch": 27.1, + "learning_rate": 4.83689213788321e-06, + "loss": 0.9114, + "step": 313110 + }, + { + "epoch": 27.1, + "learning_rate": 4.835447426969864e-06, + "loss": 0.9075, + "step": 313120 + }, + { + "epoch": 27.1, + "learning_rate": 4.834002716056517e-06, + "loss": 0.8871, + "step": 313130 + }, + { + "epoch": 27.1, + "learning_rate": 4.832558005143171e-06, + "loss": 0.8979, + "step": 313140 + }, + { + "epoch": 27.11, + "learning_rate": 4.831113294229825e-06, + "loss": 0.8931, + "step": 313150 + }, + { + "epoch": 27.11, + "learning_rate": 4.829668583316479e-06, + "loss": 0.9085, + "step": 313160 + }, + { + "epoch": 27.11, + "learning_rate": 4.828223872403132e-06, + "loss": 0.8914, + "step": 313170 + }, + { + "epoch": 27.11, + "learning_rate": 4.826779161489786e-06, + "loss": 0.8257, + "step": 313180 + }, + { + "epoch": 27.11, + "learning_rate": 4.82533445057644e-06, + "loss": 0.8885, + "step": 313190 + }, + { + "epoch": 27.11, + "learning_rate": 4.823889739663094e-06, + "loss": 0.9215, + "step": 313200 + }, + { + "epoch": 27.11, + "learning_rate": 4.822445028749748e-06, + "loss": 0.901, + "step": 313210 + }, + { + "epoch": 27.11, + "learning_rate": 4.821000317836401e-06, + "loss": 0.9055, + "step": 313220 + }, + { + "epoch": 27.11, + "learning_rate": 4.819555606923055e-06, + "loss": 0.8958, + "step": 313230 + }, + { + "epoch": 27.11, + "learning_rate": 4.818110896009709e-06, + "loss": 0.8708, + "step": 313240 + }, + { + "epoch": 27.11, + "learning_rate": 4.816666185096363e-06, + "loss": 0.8992, + "step": 313250 + }, + { + "epoch": 27.11, + "learning_rate": 4.815221474183016e-06, + "loss": 0.8784, + "step": 313260 + }, + { + "epoch": 27.12, + "learning_rate": 4.81377676326967e-06, + "loss": 0.9799, + "step": 313270 + }, + { + "epoch": 27.12, + "learning_rate": 4.812332052356323e-06, + "loss": 0.9865, + "step": 313280 + }, + { + "epoch": 27.12, + "learning_rate": 4.810887341442978e-06, + "loss": 0.9222, + "step": 313290 + }, + { + "epoch": 27.12, + "learning_rate": 4.809442630529631e-06, + "loss": 0.8635, + "step": 313300 + }, + { + "epoch": 27.12, + "learning_rate": 4.807997919616284e-06, + "loss": 0.922, + "step": 313310 + }, + { + "epoch": 27.12, + "learning_rate": 4.806553208702939e-06, + "loss": 0.9503, + "step": 313320 + }, + { + "epoch": 27.12, + "learning_rate": 4.805108497789593e-06, + "loss": 0.9165, + "step": 313330 + }, + { + "epoch": 27.12, + "learning_rate": 4.803663786876246e-06, + "loss": 0.8871, + "step": 313340 + }, + { + "epoch": 27.12, + "learning_rate": 4.8022190759629e-06, + "loss": 0.9854, + "step": 313350 + }, + { + "epoch": 27.12, + "learning_rate": 4.800774365049554e-06, + "loss": 0.906, + "step": 313360 + }, + { + "epoch": 27.12, + "learning_rate": 4.799329654136207e-06, + "loss": 0.8595, + "step": 313370 + }, + { + "epoch": 27.13, + "learning_rate": 4.7978849432228615e-06, + "loss": 0.8934, + "step": 313380 + }, + { + "epoch": 27.13, + "learning_rate": 4.796440232309515e-06, + "loss": 0.8774, + "step": 313390 + }, + { + "epoch": 27.13, + "learning_rate": 4.794995521396168e-06, + "loss": 0.8755, + "step": 313400 + }, + { + "epoch": 27.13, + "learning_rate": 4.7935508104828226e-06, + "loss": 0.9375, + "step": 313410 + }, + { + "epoch": 27.13, + "learning_rate": 4.792106099569477e-06, + "loss": 0.9474, + "step": 313420 + }, + { + "epoch": 27.13, + "learning_rate": 4.79066138865613e-06, + "loss": 0.8902, + "step": 313430 + }, + { + "epoch": 27.13, + "learning_rate": 4.789216677742784e-06, + "loss": 0.9066, + "step": 313440 + }, + { + "epoch": 27.13, + "learning_rate": 4.787771966829438e-06, + "loss": 0.8824, + "step": 313450 + }, + { + "epoch": 27.13, + "learning_rate": 4.786327255916091e-06, + "loss": 0.9155, + "step": 313460 + }, + { + "epoch": 27.13, + "learning_rate": 4.7848825450027455e-06, + "loss": 0.9019, + "step": 313470 + }, + { + "epoch": 27.13, + "learning_rate": 4.783437834089399e-06, + "loss": 0.913, + "step": 313480 + }, + { + "epoch": 27.13, + "learning_rate": 4.781993123176052e-06, + "loss": 0.8819, + "step": 313490 + }, + { + "epoch": 27.14, + "learning_rate": 4.7805484122627065e-06, + "loss": 0.8915, + "step": 313500 + }, + { + "epoch": 27.14, + "learning_rate": 4.77910370134936e-06, + "loss": 0.8643, + "step": 313510 + }, + { + "epoch": 27.14, + "learning_rate": 4.777658990436014e-06, + "loss": 0.9081, + "step": 313520 + }, + { + "epoch": 27.14, + "learning_rate": 4.7762142795226675e-06, + "loss": 0.9176, + "step": 313530 + }, + { + "epoch": 27.14, + "learning_rate": 4.774769568609321e-06, + "loss": 0.8711, + "step": 313540 + }, + { + "epoch": 27.14, + "learning_rate": 4.773324857695975e-06, + "loss": 0.8728, + "step": 313550 + }, + { + "epoch": 27.14, + "learning_rate": 4.771880146782629e-06, + "loss": 0.8877, + "step": 313560 + }, + { + "epoch": 27.14, + "learning_rate": 4.770435435869283e-06, + "loss": 0.8882, + "step": 313570 + }, + { + "epoch": 27.14, + "learning_rate": 4.768990724955936e-06, + "loss": 0.8984, + "step": 313580 + }, + { + "epoch": 27.14, + "learning_rate": 4.7675460140425904e-06, + "loss": 0.9111, + "step": 313590 + }, + { + "epoch": 27.14, + "learning_rate": 4.766101303129244e-06, + "loss": 0.8714, + "step": 313600 + }, + { + "epoch": 27.15, + "learning_rate": 4.764656592215898e-06, + "loss": 0.9154, + "step": 313610 + }, + { + "epoch": 27.15, + "learning_rate": 4.7632118813025515e-06, + "loss": 0.9143, + "step": 313620 + }, + { + "epoch": 27.15, + "learning_rate": 4.761767170389205e-06, + "loss": 0.9333, + "step": 313630 + }, + { + "epoch": 27.15, + "learning_rate": 4.760322459475859e-06, + "loss": 0.9175, + "step": 313640 + }, + { + "epoch": 27.15, + "learning_rate": 4.758877748562513e-06, + "loss": 0.8513, + "step": 313650 + }, + { + "epoch": 27.15, + "learning_rate": 4.757433037649167e-06, + "loss": 0.9102, + "step": 313660 + }, + { + "epoch": 27.15, + "learning_rate": 4.75598832673582e-06, + "loss": 0.9304, + "step": 313670 + }, + { + "epoch": 27.15, + "learning_rate": 4.754543615822474e-06, + "loss": 0.8899, + "step": 313680 + }, + { + "epoch": 27.15, + "learning_rate": 4.753098904909128e-06, + "loss": 0.8803, + "step": 313690 + }, + { + "epoch": 27.15, + "learning_rate": 4.751654193995782e-06, + "loss": 0.8712, + "step": 313700 + }, + { + "epoch": 27.15, + "learning_rate": 4.750209483082435e-06, + "loss": 0.9317, + "step": 313710 + }, + { + "epoch": 27.15, + "learning_rate": 4.748764772169089e-06, + "loss": 0.848, + "step": 313720 + }, + { + "epoch": 27.16, + "learning_rate": 4.747320061255743e-06, + "loss": 0.8705, + "step": 313730 + }, + { + "epoch": 27.16, + "learning_rate": 4.7458753503423964e-06, + "loss": 0.8561, + "step": 313740 + }, + { + "epoch": 27.16, + "learning_rate": 4.744430639429051e-06, + "loss": 0.8411, + "step": 313750 + }, + { + "epoch": 27.16, + "learning_rate": 4.742985928515704e-06, + "loss": 0.9088, + "step": 313760 + }, + { + "epoch": 27.16, + "learning_rate": 4.7415412176023575e-06, + "loss": 0.8668, + "step": 313770 + }, + { + "epoch": 27.16, + "learning_rate": 4.740096506689012e-06, + "loss": 0.8337, + "step": 313780 + }, + { + "epoch": 27.16, + "learning_rate": 4.738651795775666e-06, + "loss": 0.8829, + "step": 313790 + }, + { + "epoch": 27.16, + "learning_rate": 4.737207084862319e-06, + "loss": 0.9069, + "step": 313800 + }, + { + "epoch": 27.16, + "learning_rate": 4.735762373948973e-06, + "loss": 0.9843, + "step": 313810 + }, + { + "epoch": 27.16, + "learning_rate": 4.734317663035627e-06, + "loss": 0.9202, + "step": 313820 + }, + { + "epoch": 27.16, + "learning_rate": 4.73287295212228e-06, + "loss": 0.9168, + "step": 313830 + }, + { + "epoch": 27.17, + "learning_rate": 4.731428241208935e-06, + "loss": 0.8933, + "step": 313840 + }, + { + "epoch": 27.17, + "learning_rate": 4.729983530295588e-06, + "loss": 0.9286, + "step": 313850 + }, + { + "epoch": 27.17, + "learning_rate": 4.7285388193822414e-06, + "loss": 0.9267, + "step": 313860 + }, + { + "epoch": 27.17, + "learning_rate": 4.727094108468896e-06, + "loss": 0.901, + "step": 313870 + }, + { + "epoch": 27.17, + "learning_rate": 4.72564939755555e-06, + "loss": 0.9271, + "step": 313880 + }, + { + "epoch": 27.17, + "learning_rate": 4.724204686642203e-06, + "loss": 0.9606, + "step": 313890 + }, + { + "epoch": 27.17, + "learning_rate": 4.722759975728857e-06, + "loss": 0.8987, + "step": 313900 + }, + { + "epoch": 27.17, + "learning_rate": 4.721315264815511e-06, + "loss": 0.8878, + "step": 313910 + }, + { + "epoch": 27.17, + "learning_rate": 4.719870553902164e-06, + "loss": 0.8772, + "step": 313920 + }, + { + "epoch": 27.17, + "learning_rate": 4.718425842988819e-06, + "loss": 0.8447, + "step": 313930 + }, + { + "epoch": 27.17, + "learning_rate": 4.716981132075472e-06, + "loss": 0.874, + "step": 313940 + }, + { + "epoch": 27.17, + "learning_rate": 4.715536421162125e-06, + "loss": 0.9174, + "step": 313950 + }, + { + "epoch": 27.18, + "learning_rate": 4.71409171024878e-06, + "loss": 0.8257, + "step": 313960 + }, + { + "epoch": 27.18, + "learning_rate": 4.712646999335433e-06, + "loss": 0.9083, + "step": 313970 + }, + { + "epoch": 27.18, + "learning_rate": 4.711202288422087e-06, + "loss": 0.8532, + "step": 313980 + }, + { + "epoch": 27.18, + "learning_rate": 4.709757577508741e-06, + "loss": 0.9349, + "step": 313990 + }, + { + "epoch": 27.18, + "learning_rate": 4.708312866595394e-06, + "loss": 0.9165, + "step": 314000 + }, + { + "epoch": 27.18, + "learning_rate": 4.706868155682048e-06, + "loss": 0.8706, + "step": 314010 + }, + { + "epoch": 27.18, + "learning_rate": 4.7054234447687025e-06, + "loss": 0.9265, + "step": 314020 + }, + { + "epoch": 27.18, + "learning_rate": 4.703978733855356e-06, + "loss": 0.8986, + "step": 314030 + }, + { + "epoch": 27.18, + "learning_rate": 4.702534022942009e-06, + "loss": 0.9231, + "step": 314040 + }, + { + "epoch": 27.18, + "learning_rate": 4.7010893120286636e-06, + "loss": 0.8623, + "step": 314050 + }, + { + "epoch": 27.18, + "learning_rate": 4.699644601115317e-06, + "loss": 0.8906, + "step": 314060 + }, + { + "epoch": 27.19, + "learning_rate": 4.698199890201971e-06, + "loss": 0.9313, + "step": 314070 + }, + { + "epoch": 27.19, + "learning_rate": 4.696755179288625e-06, + "loss": 0.9115, + "step": 314080 + }, + { + "epoch": 27.19, + "learning_rate": 4.695310468375278e-06, + "loss": 0.9323, + "step": 314090 + }, + { + "epoch": 27.19, + "learning_rate": 4.693865757461932e-06, + "loss": 0.8813, + "step": 314100 + }, + { + "epoch": 27.19, + "learning_rate": 4.6924210465485865e-06, + "loss": 0.9455, + "step": 314110 + }, + { + "epoch": 27.19, + "learning_rate": 4.69097633563524e-06, + "loss": 0.9371, + "step": 314120 + }, + { + "epoch": 27.19, + "learning_rate": 4.689531624721893e-06, + "loss": 0.9176, + "step": 314130 + }, + { + "epoch": 27.19, + "learning_rate": 4.6880869138085475e-06, + "loss": 0.9059, + "step": 314140 + }, + { + "epoch": 27.19, + "learning_rate": 4.686642202895201e-06, + "loss": 0.9381, + "step": 314150 + }, + { + "epoch": 27.19, + "learning_rate": 4.685197491981855e-06, + "loss": 0.8747, + "step": 314160 + }, + { + "epoch": 27.19, + "learning_rate": 4.6837527810685085e-06, + "loss": 0.9139, + "step": 314170 + }, + { + "epoch": 27.19, + "learning_rate": 4.682308070155162e-06, + "loss": 0.8939, + "step": 314180 + }, + { + "epoch": 27.2, + "learning_rate": 4.680863359241816e-06, + "loss": 0.8875, + "step": 314190 + }, + { + "epoch": 27.2, + "learning_rate": 4.6794186483284696e-06, + "loss": 0.9129, + "step": 314200 + }, + { + "epoch": 27.2, + "learning_rate": 4.677973937415124e-06, + "loss": 0.8671, + "step": 314210 + }, + { + "epoch": 27.2, + "learning_rate": 4.676529226501777e-06, + "loss": 0.9016, + "step": 314220 + }, + { + "epoch": 27.2, + "learning_rate": 4.675084515588431e-06, + "loss": 0.9216, + "step": 314230 + }, + { + "epoch": 27.2, + "learning_rate": 4.673639804675085e-06, + "loss": 0.9277, + "step": 314240 + }, + { + "epoch": 27.2, + "learning_rate": 4.672195093761739e-06, + "loss": 0.9116, + "step": 314250 + }, + { + "epoch": 27.2, + "learning_rate": 4.670750382848392e-06, + "loss": 0.8904, + "step": 314260 + }, + { + "epoch": 27.2, + "learning_rate": 4.669305671935046e-06, + "loss": 0.8476, + "step": 314270 + }, + { + "epoch": 27.2, + "learning_rate": 4.6678609610217e-06, + "loss": 0.9393, + "step": 314280 + }, + { + "epoch": 27.2, + "learning_rate": 4.6664162501083535e-06, + "loss": 0.8664, + "step": 314290 + }, + { + "epoch": 27.2, + "learning_rate": 4.664971539195007e-06, + "loss": 0.9108, + "step": 314300 + }, + { + "epoch": 27.21, + "learning_rate": 4.663526828281661e-06, + "loss": 0.866, + "step": 314310 + }, + { + "epoch": 27.21, + "learning_rate": 4.6620821173683145e-06, + "loss": 0.8671, + "step": 314320 + }, + { + "epoch": 27.21, + "learning_rate": 4.660637406454969e-06, + "loss": 0.8959, + "step": 314330 + }, + { + "epoch": 27.21, + "learning_rate": 4.659192695541623e-06, + "loss": 0.9161, + "step": 314340 + }, + { + "epoch": 27.21, + "learning_rate": 4.6577479846282756e-06, + "loss": 0.8776, + "step": 314350 + }, + { + "epoch": 27.21, + "learning_rate": 4.65630327371493e-06, + "loss": 0.8673, + "step": 314360 + }, + { + "epoch": 27.21, + "learning_rate": 4.654858562801584e-06, + "loss": 0.9219, + "step": 314370 + }, + { + "epoch": 27.21, + "learning_rate": 4.6534138518882374e-06, + "loss": 0.9007, + "step": 314380 + }, + { + "epoch": 27.21, + "learning_rate": 4.651969140974891e-06, + "loss": 0.9372, + "step": 314390 + }, + { + "epoch": 27.21, + "learning_rate": 4.650524430061544e-06, + "loss": 0.9111, + "step": 314400 + }, + { + "epoch": 27.21, + "learning_rate": 4.6490797191481985e-06, + "loss": 0.8572, + "step": 314410 + }, + { + "epoch": 27.22, + "learning_rate": 4.647635008234853e-06, + "loss": 0.927, + "step": 314420 + }, + { + "epoch": 27.22, + "learning_rate": 4.646190297321506e-06, + "loss": 0.9891, + "step": 314430 + }, + { + "epoch": 27.22, + "learning_rate": 4.6447455864081595e-06, + "loss": 0.8996, + "step": 314440 + }, + { + "epoch": 27.22, + "learning_rate": 4.643300875494814e-06, + "loss": 0.8831, + "step": 314450 + }, + { + "epoch": 27.22, + "learning_rate": 4.641856164581467e-06, + "loss": 0.8863, + "step": 314460 + }, + { + "epoch": 27.22, + "learning_rate": 4.640411453668121e-06, + "loss": 0.9276, + "step": 314470 + }, + { + "epoch": 27.22, + "learning_rate": 4.638966742754775e-06, + "loss": 0.8633, + "step": 314480 + }, + { + "epoch": 27.22, + "learning_rate": 4.637522031841428e-06, + "loss": 0.9748, + "step": 314490 + }, + { + "epoch": 27.22, + "learning_rate": 4.636077320928082e-06, + "loss": 0.9787, + "step": 314500 + }, + { + "epoch": 27.22, + "learning_rate": 4.634632610014737e-06, + "loss": 0.8633, + "step": 314510 + }, + { + "epoch": 27.22, + "learning_rate": 4.63318789910139e-06, + "loss": 0.8984, + "step": 314520 + }, + { + "epoch": 27.22, + "learning_rate": 4.6317431881880435e-06, + "loss": 0.932, + "step": 314530 + }, + { + "epoch": 27.23, + "learning_rate": 4.630298477274698e-06, + "loss": 0.934, + "step": 314540 + }, + { + "epoch": 27.23, + "learning_rate": 4.628853766361351e-06, + "loss": 0.8444, + "step": 314550 + }, + { + "epoch": 27.23, + "learning_rate": 4.627409055448005e-06, + "loss": 0.9328, + "step": 314560 + }, + { + "epoch": 27.23, + "learning_rate": 4.625964344534659e-06, + "loss": 0.8933, + "step": 314570 + }, + { + "epoch": 27.23, + "learning_rate": 4.624519633621312e-06, + "loss": 0.9279, + "step": 314580 + }, + { + "epoch": 27.23, + "learning_rate": 4.623074922707966e-06, + "loss": 0.8683, + "step": 314590 + }, + { + "epoch": 27.23, + "learning_rate": 4.621630211794621e-06, + "loss": 0.8692, + "step": 314600 + }, + { + "epoch": 27.23, + "learning_rate": 4.620185500881274e-06, + "loss": 0.8871, + "step": 314610 + }, + { + "epoch": 27.23, + "learning_rate": 4.618740789967927e-06, + "loss": 0.941, + "step": 314620 + }, + { + "epoch": 27.23, + "learning_rate": 4.617296079054581e-06, + "loss": 0.8627, + "step": 314630 + }, + { + "epoch": 27.23, + "learning_rate": 4.615851368141235e-06, + "loss": 0.9253, + "step": 314640 + }, + { + "epoch": 27.24, + "learning_rate": 4.614406657227889e-06, + "loss": 0.8912, + "step": 314650 + }, + { + "epoch": 27.24, + "learning_rate": 4.612961946314543e-06, + "loss": 0.8779, + "step": 314660 + }, + { + "epoch": 27.24, + "learning_rate": 4.611517235401196e-06, + "loss": 0.8997, + "step": 314670 + }, + { + "epoch": 27.24, + "learning_rate": 4.61007252448785e-06, + "loss": 0.9146, + "step": 314680 + }, + { + "epoch": 27.24, + "learning_rate": 4.608627813574504e-06, + "loss": 0.8354, + "step": 314690 + }, + { + "epoch": 27.24, + "learning_rate": 4.607183102661158e-06, + "loss": 0.875, + "step": 314700 + }, + { + "epoch": 27.24, + "learning_rate": 4.605738391747811e-06, + "loss": 0.9107, + "step": 314710 + }, + { + "epoch": 27.24, + "learning_rate": 4.604293680834465e-06, + "loss": 0.8768, + "step": 314720 + }, + { + "epoch": 27.24, + "learning_rate": 4.602848969921119e-06, + "loss": 0.8825, + "step": 314730 + }, + { + "epoch": 27.24, + "learning_rate": 4.601404259007773e-06, + "loss": 0.9726, + "step": 314740 + }, + { + "epoch": 27.24, + "learning_rate": 4.599959548094427e-06, + "loss": 0.8933, + "step": 314750 + }, + { + "epoch": 27.24, + "learning_rate": 4.59851483718108e-06, + "loss": 0.9317, + "step": 314760 + }, + { + "epoch": 27.25, + "learning_rate": 4.597070126267734e-06, + "loss": 0.9497, + "step": 314770 + }, + { + "epoch": 27.25, + "learning_rate": 4.595625415354388e-06, + "loss": 0.9321, + "step": 314780 + }, + { + "epoch": 27.25, + "learning_rate": 4.594180704441042e-06, + "loss": 0.929, + "step": 314790 + }, + { + "epoch": 27.25, + "learning_rate": 4.592735993527695e-06, + "loss": 0.8418, + "step": 314800 + }, + { + "epoch": 27.25, + "learning_rate": 4.591291282614349e-06, + "loss": 0.9228, + "step": 314810 + }, + { + "epoch": 27.25, + "learning_rate": 4.589846571701003e-06, + "loss": 0.9344, + "step": 314820 + }, + { + "epoch": 27.25, + "learning_rate": 4.588401860787657e-06, + "loss": 0.9075, + "step": 314830 + }, + { + "epoch": 27.25, + "learning_rate": 4.5869571498743106e-06, + "loss": 0.9221, + "step": 314840 + }, + { + "epoch": 27.25, + "learning_rate": 4.585512438960964e-06, + "loss": 0.9061, + "step": 314850 + }, + { + "epoch": 27.25, + "learning_rate": 4.584067728047617e-06, + "loss": 0.8728, + "step": 314860 + }, + { + "epoch": 27.25, + "learning_rate": 4.582623017134272e-06, + "loss": 0.9047, + "step": 314870 + }, + { + "epoch": 27.26, + "learning_rate": 4.581178306220926e-06, + "loss": 0.9355, + "step": 314880 + }, + { + "epoch": 27.26, + "learning_rate": 4.579733595307579e-06, + "loss": 0.899, + "step": 314890 + }, + { + "epoch": 27.26, + "learning_rate": 4.578288884394233e-06, + "loss": 0.8697, + "step": 314900 + }, + { + "epoch": 27.26, + "learning_rate": 4.576844173480887e-06, + "loss": 0.8485, + "step": 314910 + }, + { + "epoch": 27.26, + "learning_rate": 4.57539946256754e-06, + "loss": 0.8953, + "step": 314920 + }, + { + "epoch": 27.26, + "learning_rate": 4.5739547516541945e-06, + "loss": 0.8775, + "step": 314930 + }, + { + "epoch": 27.26, + "learning_rate": 4.572510040740848e-06, + "loss": 0.9248, + "step": 314940 + }, + { + "epoch": 27.26, + "learning_rate": 4.571065329827501e-06, + "loss": 0.9405, + "step": 314950 + }, + { + "epoch": 27.26, + "learning_rate": 4.5696206189141555e-06, + "loss": 0.8733, + "step": 314960 + }, + { + "epoch": 27.26, + "learning_rate": 4.56817590800081e-06, + "loss": 0.9457, + "step": 314970 + }, + { + "epoch": 27.26, + "learning_rate": 4.566731197087463e-06, + "loss": 0.8826, + "step": 314980 + }, + { + "epoch": 27.26, + "learning_rate": 4.5652864861741166e-06, + "loss": 0.912, + "step": 314990 + }, + { + "epoch": 27.27, + "learning_rate": 4.563841775260771e-06, + "loss": 0.9696, + "step": 315000 + }, + { + "epoch": 27.27, + "learning_rate": 4.562397064347424e-06, + "loss": 0.9342, + "step": 315010 + }, + { + "epoch": 27.27, + "learning_rate": 4.5609523534340784e-06, + "loss": 0.9026, + "step": 315020 + }, + { + "epoch": 27.27, + "learning_rate": 4.559507642520732e-06, + "loss": 0.9001, + "step": 315030 + }, + { + "epoch": 27.27, + "learning_rate": 4.558062931607385e-06, + "loss": 0.8837, + "step": 315040 + }, + { + "epoch": 27.27, + "learning_rate": 4.5566182206940395e-06, + "loss": 0.8952, + "step": 315050 + }, + { + "epoch": 27.27, + "learning_rate": 4.555173509780694e-06, + "loss": 0.8847, + "step": 315060 + }, + { + "epoch": 27.27, + "learning_rate": 4.553728798867347e-06, + "loss": 0.9196, + "step": 315070 + }, + { + "epoch": 27.27, + "learning_rate": 4.5522840879540005e-06, + "loss": 0.9317, + "step": 315080 + }, + { + "epoch": 27.27, + "learning_rate": 4.550839377040654e-06, + "loss": 0.9094, + "step": 315090 + }, + { + "epoch": 27.27, + "learning_rate": 4.549394666127308e-06, + "loss": 0.9162, + "step": 315100 + }, + { + "epoch": 27.28, + "learning_rate": 4.547949955213962e-06, + "loss": 0.8488, + "step": 315110 + }, + { + "epoch": 27.28, + "learning_rate": 4.546505244300616e-06, + "loss": 0.9388, + "step": 315120 + }, + { + "epoch": 27.28, + "learning_rate": 4.545060533387269e-06, + "loss": 0.8842, + "step": 315130 + }, + { + "epoch": 27.28, + "learning_rate": 4.543615822473923e-06, + "loss": 0.9043, + "step": 315140 + }, + { + "epoch": 27.28, + "learning_rate": 4.542171111560577e-06, + "loss": 0.9297, + "step": 315150 + }, + { + "epoch": 27.28, + "learning_rate": 4.540726400647231e-06, + "loss": 0.9201, + "step": 315160 + }, + { + "epoch": 27.28, + "learning_rate": 4.5392816897338844e-06, + "loss": 0.9219, + "step": 315170 + }, + { + "epoch": 27.28, + "learning_rate": 4.537836978820538e-06, + "loss": 0.8213, + "step": 315180 + }, + { + "epoch": 27.28, + "learning_rate": 4.536392267907192e-06, + "loss": 0.9083, + "step": 315190 + }, + { + "epoch": 27.28, + "learning_rate": 4.534947556993846e-06, + "loss": 0.9311, + "step": 315200 + }, + { + "epoch": 27.28, + "learning_rate": 4.5335028460805e-06, + "loss": 0.9162, + "step": 315210 + }, + { + "epoch": 27.28, + "learning_rate": 4.532058135167153e-06, + "loss": 0.873, + "step": 315220 + }, + { + "epoch": 27.29, + "learning_rate": 4.530613424253807e-06, + "loss": 0.9353, + "step": 315230 + }, + { + "epoch": 27.29, + "learning_rate": 4.529168713340461e-06, + "loss": 0.9194, + "step": 315240 + }, + { + "epoch": 27.29, + "learning_rate": 4.527724002427115e-06, + "loss": 0.8257, + "step": 315250 + }, + { + "epoch": 27.29, + "learning_rate": 4.526279291513768e-06, + "loss": 0.8671, + "step": 315260 + }, + { + "epoch": 27.29, + "learning_rate": 4.524834580600422e-06, + "loss": 0.9277, + "step": 315270 + }, + { + "epoch": 27.29, + "learning_rate": 4.523389869687076e-06, + "loss": 0.8891, + "step": 315280 + }, + { + "epoch": 27.29, + "learning_rate": 4.52194515877373e-06, + "loss": 0.8546, + "step": 315290 + }, + { + "epoch": 27.29, + "learning_rate": 4.520500447860383e-06, + "loss": 0.9016, + "step": 315300 + }, + { + "epoch": 27.29, + "learning_rate": 4.519055736947037e-06, + "loss": 0.9329, + "step": 315310 + }, + { + "epoch": 27.29, + "learning_rate": 4.5176110260336905e-06, + "loss": 0.8946, + "step": 315320 + }, + { + "epoch": 27.29, + "learning_rate": 4.516166315120345e-06, + "loss": 0.8876, + "step": 315330 + }, + { + "epoch": 27.29, + "learning_rate": 4.514721604206999e-06, + "loss": 0.8984, + "step": 315340 + }, + { + "epoch": 27.3, + "learning_rate": 4.5132768932936515e-06, + "loss": 0.9785, + "step": 315350 + }, + { + "epoch": 27.3, + "learning_rate": 4.511832182380306e-06, + "loss": 0.8956, + "step": 315360 + }, + { + "epoch": 27.3, + "learning_rate": 4.51038747146696e-06, + "loss": 0.8658, + "step": 315370 + }, + { + "epoch": 27.3, + "learning_rate": 4.508942760553613e-06, + "loss": 0.9182, + "step": 315380 + }, + { + "epoch": 27.3, + "learning_rate": 4.507498049640267e-06, + "loss": 0.9727, + "step": 315390 + }, + { + "epoch": 27.3, + "learning_rate": 4.506053338726921e-06, + "loss": 0.9026, + "step": 315400 + }, + { + "epoch": 27.3, + "learning_rate": 4.504608627813574e-06, + "loss": 0.9226, + "step": 315410 + }, + { + "epoch": 27.3, + "learning_rate": 4.503163916900229e-06, + "loss": 0.8865, + "step": 315420 + }, + { + "epoch": 27.3, + "learning_rate": 4.501719205986883e-06, + "loss": 0.9472, + "step": 315430 + }, + { + "epoch": 27.3, + "learning_rate": 4.5002744950735354e-06, + "loss": 0.9593, + "step": 315440 + }, + { + "epoch": 27.3, + "learning_rate": 4.49882978416019e-06, + "loss": 0.8573, + "step": 315450 + }, + { + "epoch": 27.31, + "learning_rate": 4.497385073246844e-06, + "loss": 0.8208, + "step": 315460 + }, + { + "epoch": 27.31, + "learning_rate": 4.495940362333497e-06, + "loss": 0.8825, + "step": 315470 + }, + { + "epoch": 27.31, + "learning_rate": 4.494495651420151e-06, + "loss": 0.928, + "step": 315480 + }, + { + "epoch": 27.31, + "learning_rate": 4.493050940506805e-06, + "loss": 0.8898, + "step": 315490 + }, + { + "epoch": 27.31, + "learning_rate": 4.491606229593458e-06, + "loss": 0.8813, + "step": 315500 + }, + { + "epoch": 27.31, + "learning_rate": 4.490161518680113e-06, + "loss": 0.9331, + "step": 315510 + }, + { + "epoch": 27.31, + "learning_rate": 4.488716807766767e-06, + "loss": 0.8956, + "step": 315520 + }, + { + "epoch": 27.31, + "learning_rate": 4.487272096853419e-06, + "loss": 0.904, + "step": 315530 + }, + { + "epoch": 27.31, + "learning_rate": 4.485827385940074e-06, + "loss": 0.9586, + "step": 315540 + }, + { + "epoch": 27.31, + "learning_rate": 4.484382675026727e-06, + "loss": 0.8785, + "step": 315550 + }, + { + "epoch": 27.31, + "learning_rate": 4.482937964113381e-06, + "loss": 0.9273, + "step": 315560 + }, + { + "epoch": 27.31, + "learning_rate": 4.481493253200035e-06, + "loss": 0.9072, + "step": 315570 + }, + { + "epoch": 27.32, + "learning_rate": 4.480048542286688e-06, + "loss": 0.8722, + "step": 315580 + }, + { + "epoch": 27.32, + "learning_rate": 4.478603831373342e-06, + "loss": 0.9185, + "step": 315590 + }, + { + "epoch": 27.32, + "learning_rate": 4.4771591204599965e-06, + "loss": 0.9202, + "step": 315600 + }, + { + "epoch": 27.32, + "learning_rate": 4.47571440954665e-06, + "loss": 0.9377, + "step": 315610 + }, + { + "epoch": 27.32, + "learning_rate": 4.474269698633303e-06, + "loss": 0.901, + "step": 315620 + }, + { + "epoch": 27.32, + "learning_rate": 4.4728249877199576e-06, + "loss": 0.8888, + "step": 315630 + }, + { + "epoch": 27.32, + "learning_rate": 4.471380276806611e-06, + "loss": 0.9145, + "step": 315640 + }, + { + "epoch": 27.32, + "learning_rate": 4.469935565893265e-06, + "loss": 0.8435, + "step": 315650 + }, + { + "epoch": 27.32, + "learning_rate": 4.468490854979919e-06, + "loss": 0.9115, + "step": 315660 + }, + { + "epoch": 27.32, + "learning_rate": 4.467046144066572e-06, + "loss": 0.8688, + "step": 315670 + }, + { + "epoch": 27.32, + "learning_rate": 4.465601433153226e-06, + "loss": 0.9142, + "step": 315680 + }, + { + "epoch": 27.33, + "learning_rate": 4.4641567222398805e-06, + "loss": 0.9081, + "step": 315690 + }, + { + "epoch": 27.33, + "learning_rate": 4.462712011326534e-06, + "loss": 0.8497, + "step": 315700 + }, + { + "epoch": 27.33, + "learning_rate": 4.461267300413187e-06, + "loss": 0.8961, + "step": 315710 + }, + { + "epoch": 27.33, + "learning_rate": 4.4598225894998415e-06, + "loss": 0.8959, + "step": 315720 + }, + { + "epoch": 27.33, + "learning_rate": 4.458377878586495e-06, + "loss": 0.8939, + "step": 315730 + }, + { + "epoch": 27.33, + "learning_rate": 4.456933167673149e-06, + "loss": 0.9071, + "step": 315740 + }, + { + "epoch": 27.33, + "learning_rate": 4.4554884567598025e-06, + "loss": 0.9196, + "step": 315750 + }, + { + "epoch": 27.33, + "learning_rate": 4.454043745846456e-06, + "loss": 0.8886, + "step": 315760 + }, + { + "epoch": 27.33, + "learning_rate": 4.45259903493311e-06, + "loss": 0.9322, + "step": 315770 + }, + { + "epoch": 27.33, + "learning_rate": 4.4511543240197636e-06, + "loss": 0.8902, + "step": 315780 + }, + { + "epoch": 27.33, + "learning_rate": 4.449709613106418e-06, + "loss": 0.9397, + "step": 315790 + }, + { + "epoch": 27.33, + "learning_rate": 4.448264902193071e-06, + "loss": 0.8547, + "step": 315800 + }, + { + "epoch": 27.34, + "learning_rate": 4.446820191279725e-06, + "loss": 0.878, + "step": 315810 + }, + { + "epoch": 27.34, + "learning_rate": 4.445375480366379e-06, + "loss": 0.8597, + "step": 315820 + }, + { + "epoch": 27.34, + "learning_rate": 4.443930769453033e-06, + "loss": 0.9252, + "step": 315830 + }, + { + "epoch": 27.34, + "learning_rate": 4.4424860585396865e-06, + "loss": 0.9022, + "step": 315840 + }, + { + "epoch": 27.34, + "learning_rate": 4.44104134762634e-06, + "loss": 0.8979, + "step": 315850 + }, + { + "epoch": 27.34, + "learning_rate": 4.439596636712994e-06, + "loss": 0.8763, + "step": 315860 + }, + { + "epoch": 27.34, + "learning_rate": 4.4381519257996475e-06, + "loss": 0.9125, + "step": 315870 + }, + { + "epoch": 27.34, + "learning_rate": 4.436707214886302e-06, + "loss": 0.8922, + "step": 315880 + }, + { + "epoch": 27.34, + "learning_rate": 4.435262503972955e-06, + "loss": 0.8763, + "step": 315890 + }, + { + "epoch": 27.34, + "learning_rate": 4.4338177930596085e-06, + "loss": 0.8794, + "step": 315900 + }, + { + "epoch": 27.34, + "learning_rate": 4.432373082146263e-06, + "loss": 0.9404, + "step": 315910 + }, + { + "epoch": 27.35, + "learning_rate": 4.430928371232917e-06, + "loss": 0.8842, + "step": 315920 + }, + { + "epoch": 27.35, + "learning_rate": 4.42948366031957e-06, + "loss": 0.911, + "step": 315930 + }, + { + "epoch": 27.35, + "learning_rate": 4.428038949406224e-06, + "loss": 0.8991, + "step": 315940 + }, + { + "epoch": 27.35, + "learning_rate": 4.426594238492878e-06, + "loss": 0.9018, + "step": 315950 + }, + { + "epoch": 27.35, + "learning_rate": 4.4251495275795314e-06, + "loss": 0.8634, + "step": 315960 + }, + { + "epoch": 27.35, + "learning_rate": 4.423704816666186e-06, + "loss": 0.8524, + "step": 315970 + }, + { + "epoch": 27.35, + "learning_rate": 4.422260105752839e-06, + "loss": 0.8832, + "step": 315980 + }, + { + "epoch": 27.35, + "learning_rate": 4.4208153948394925e-06, + "loss": 0.9713, + "step": 315990 + }, + { + "epoch": 27.35, + "learning_rate": 4.419370683926147e-06, + "loss": 0.9223, + "step": 316000 + }, + { + "epoch": 27.35, + "learning_rate": 4.4179259730128e-06, + "loss": 0.8447, + "step": 316010 + }, + { + "epoch": 27.35, + "learning_rate": 4.416481262099454e-06, + "loss": 0.8121, + "step": 316020 + }, + { + "epoch": 27.35, + "learning_rate": 4.415036551186108e-06, + "loss": 0.8661, + "step": 316030 + }, + { + "epoch": 27.36, + "learning_rate": 4.413591840272761e-06, + "loss": 0.9143, + "step": 316040 + }, + { + "epoch": 27.36, + "learning_rate": 4.412147129359415e-06, + "loss": 0.9382, + "step": 316050 + }, + { + "epoch": 27.36, + "learning_rate": 4.41070241844607e-06, + "loss": 0.9178, + "step": 316060 + }, + { + "epoch": 27.36, + "learning_rate": 4.409257707532723e-06, + "loss": 0.9416, + "step": 316070 + }, + { + "epoch": 27.36, + "learning_rate": 4.407812996619376e-06, + "loss": 0.9315, + "step": 316080 + }, + { + "epoch": 27.36, + "learning_rate": 4.406368285706031e-06, + "loss": 0.8851, + "step": 316090 + }, + { + "epoch": 27.36, + "learning_rate": 4.404923574792684e-06, + "loss": 0.8653, + "step": 316100 + }, + { + "epoch": 27.36, + "learning_rate": 4.403478863879338e-06, + "loss": 0.952, + "step": 316110 + }, + { + "epoch": 27.36, + "learning_rate": 4.402034152965992e-06, + "loss": 0.9073, + "step": 316120 + }, + { + "epoch": 27.36, + "learning_rate": 4.400589442052645e-06, + "loss": 0.8996, + "step": 316130 + }, + { + "epoch": 27.36, + "learning_rate": 4.399144731139299e-06, + "loss": 0.9078, + "step": 316140 + }, + { + "epoch": 27.37, + "learning_rate": 4.3977000202259536e-06, + "loss": 0.9008, + "step": 316150 + }, + { + "epoch": 27.37, + "learning_rate": 4.396255309312607e-06, + "loss": 0.8924, + "step": 316160 + }, + { + "epoch": 27.37, + "learning_rate": 4.39481059839926e-06, + "loss": 0.8589, + "step": 316170 + }, + { + "epoch": 27.37, + "learning_rate": 4.393365887485915e-06, + "loss": 0.903, + "step": 316180 + }, + { + "epoch": 27.37, + "learning_rate": 4.391921176572568e-06, + "loss": 0.9445, + "step": 316190 + }, + { + "epoch": 27.37, + "learning_rate": 4.390476465659222e-06, + "loss": 0.9267, + "step": 316200 + }, + { + "epoch": 27.37, + "learning_rate": 4.389031754745876e-06, + "loss": 0.8859, + "step": 316210 + }, + { + "epoch": 27.37, + "learning_rate": 4.387587043832529e-06, + "loss": 0.883, + "step": 316220 + }, + { + "epoch": 27.37, + "learning_rate": 4.386142332919183e-06, + "loss": 0.9131, + "step": 316230 + }, + { + "epoch": 27.37, + "learning_rate": 4.384697622005837e-06, + "loss": 0.8954, + "step": 316240 + }, + { + "epoch": 27.37, + "learning_rate": 4.383252911092491e-06, + "loss": 0.8552, + "step": 316250 + }, + { + "epoch": 27.37, + "learning_rate": 4.381808200179144e-06, + "loss": 0.9023, + "step": 316260 + }, + { + "epoch": 27.38, + "learning_rate": 4.380363489265798e-06, + "loss": 0.8778, + "step": 316270 + }, + { + "epoch": 27.38, + "learning_rate": 4.378918778352452e-06, + "loss": 0.9067, + "step": 316280 + }, + { + "epoch": 27.38, + "learning_rate": 4.377474067439106e-06, + "loss": 0.9018, + "step": 316290 + }, + { + "epoch": 27.38, + "learning_rate": 4.376029356525759e-06, + "loss": 0.8724, + "step": 316300 + }, + { + "epoch": 27.38, + "learning_rate": 4.374584645612413e-06, + "loss": 0.9074, + "step": 316310 + }, + { + "epoch": 27.38, + "learning_rate": 4.373139934699067e-06, + "loss": 0.8909, + "step": 316320 + }, + { + "epoch": 27.38, + "learning_rate": 4.371695223785721e-06, + "loss": 0.8977, + "step": 316330 + }, + { + "epoch": 27.38, + "learning_rate": 4.370250512872375e-06, + "loss": 0.8709, + "step": 316340 + }, + { + "epoch": 27.38, + "learning_rate": 4.368805801959028e-06, + "loss": 0.8714, + "step": 316350 + }, + { + "epoch": 27.38, + "learning_rate": 4.367361091045682e-06, + "loss": 0.906, + "step": 316360 + }, + { + "epoch": 27.38, + "learning_rate": 4.365916380132336e-06, + "loss": 0.9251, + "step": 316370 + }, + { + "epoch": 27.38, + "learning_rate": 4.36447166921899e-06, + "loss": 0.9541, + "step": 316380 + }, + { + "epoch": 27.39, + "learning_rate": 4.363026958305643e-06, + "loss": 0.8968, + "step": 316390 + }, + { + "epoch": 27.39, + "learning_rate": 4.361582247392297e-06, + "loss": 0.885, + "step": 316400 + }, + { + "epoch": 27.39, + "learning_rate": 4.360137536478951e-06, + "loss": 0.8732, + "step": 316410 + }, + { + "epoch": 27.39, + "learning_rate": 4.3586928255656046e-06, + "loss": 0.8696, + "step": 316420 + }, + { + "epoch": 27.39, + "learning_rate": 4.357248114652259e-06, + "loss": 0.8538, + "step": 316430 + }, + { + "epoch": 27.39, + "learning_rate": 4.355803403738911e-06, + "loss": 0.8971, + "step": 316440 + }, + { + "epoch": 27.39, + "learning_rate": 4.354358692825566e-06, + "loss": 0.893, + "step": 316450 + }, + { + "epoch": 27.39, + "learning_rate": 4.35291398191222e-06, + "loss": 0.8332, + "step": 316460 + }, + { + "epoch": 27.39, + "learning_rate": 4.351469270998873e-06, + "loss": 0.9739, + "step": 316470 + }, + { + "epoch": 27.39, + "learning_rate": 4.350024560085527e-06, + "loss": 0.8056, + "step": 316480 + }, + { + "epoch": 27.39, + "learning_rate": 4.348579849172181e-06, + "loss": 0.8946, + "step": 316490 + }, + { + "epoch": 27.4, + "learning_rate": 4.347135138258834e-06, + "loss": 0.8961, + "step": 316500 + }, + { + "epoch": 27.4, + "learning_rate": 4.3456904273454885e-06, + "loss": 0.8767, + "step": 316510 + }, + { + "epoch": 27.4, + "learning_rate": 4.344245716432143e-06, + "loss": 0.9286, + "step": 316520 + }, + { + "epoch": 27.4, + "learning_rate": 4.342801005518795e-06, + "loss": 0.8392, + "step": 316530 + }, + { + "epoch": 27.4, + "learning_rate": 4.3413562946054495e-06, + "loss": 0.9322, + "step": 316540 + }, + { + "epoch": 27.4, + "learning_rate": 4.339911583692104e-06, + "loss": 0.9335, + "step": 316550 + }, + { + "epoch": 27.4, + "learning_rate": 4.338466872778757e-06, + "loss": 0.8936, + "step": 316560 + }, + { + "epoch": 27.4, + "learning_rate": 4.3370221618654106e-06, + "loss": 0.9322, + "step": 316570 + }, + { + "epoch": 27.4, + "learning_rate": 4.335577450952065e-06, + "loss": 0.8617, + "step": 316580 + }, + { + "epoch": 27.4, + "learning_rate": 4.334132740038718e-06, + "loss": 0.8581, + "step": 316590 + }, + { + "epoch": 27.4, + "learning_rate": 4.3326880291253724e-06, + "loss": 0.9072, + "step": 316600 + }, + { + "epoch": 27.4, + "learning_rate": 4.331243318212027e-06, + "loss": 0.8579, + "step": 316610 + }, + { + "epoch": 27.41, + "learning_rate": 4.329798607298679e-06, + "loss": 0.8809, + "step": 316620 + }, + { + "epoch": 27.41, + "learning_rate": 4.3283538963853335e-06, + "loss": 0.9121, + "step": 316630 + }, + { + "epoch": 27.41, + "learning_rate": 4.326909185471988e-06, + "loss": 0.8993, + "step": 316640 + }, + { + "epoch": 27.41, + "learning_rate": 4.325464474558641e-06, + "loss": 0.9359, + "step": 316650 + }, + { + "epoch": 27.41, + "learning_rate": 4.3240197636452945e-06, + "loss": 0.8848, + "step": 316660 + }, + { + "epoch": 27.41, + "learning_rate": 4.322575052731948e-06, + "loss": 0.907, + "step": 316670 + }, + { + "epoch": 27.41, + "learning_rate": 4.321130341818602e-06, + "loss": 0.8815, + "step": 316680 + }, + { + "epoch": 27.41, + "learning_rate": 4.319685630905256e-06, + "loss": 0.9012, + "step": 316690 + }, + { + "epoch": 27.41, + "learning_rate": 4.31824091999191e-06, + "loss": 0.9437, + "step": 316700 + }, + { + "epoch": 27.41, + "learning_rate": 4.316796209078563e-06, + "loss": 0.9679, + "step": 316710 + }, + { + "epoch": 27.41, + "learning_rate": 4.315351498165217e-06, + "loss": 0.9533, + "step": 316720 + }, + { + "epoch": 27.42, + "learning_rate": 4.313906787251871e-06, + "loss": 0.8912, + "step": 316730 + }, + { + "epoch": 27.42, + "learning_rate": 4.312462076338525e-06, + "loss": 0.8716, + "step": 316740 + }, + { + "epoch": 27.42, + "learning_rate": 4.3110173654251784e-06, + "loss": 0.8783, + "step": 316750 + }, + { + "epoch": 27.42, + "learning_rate": 4.309572654511832e-06, + "loss": 0.8993, + "step": 316760 + }, + { + "epoch": 27.42, + "learning_rate": 4.308127943598486e-06, + "loss": 0.8662, + "step": 316770 + }, + { + "epoch": 27.42, + "learning_rate": 4.30668323268514e-06, + "loss": 0.9377, + "step": 316780 + }, + { + "epoch": 27.42, + "learning_rate": 4.305238521771794e-06, + "loss": 0.8778, + "step": 316790 + }, + { + "epoch": 27.42, + "learning_rate": 4.303793810858447e-06, + "loss": 0.8728, + "step": 316800 + }, + { + "epoch": 27.42, + "learning_rate": 4.302349099945101e-06, + "loss": 0.927, + "step": 316810 + }, + { + "epoch": 27.42, + "learning_rate": 4.300904389031755e-06, + "loss": 0.8873, + "step": 316820 + }, + { + "epoch": 27.42, + "learning_rate": 4.299459678118409e-06, + "loss": 0.8837, + "step": 316830 + }, + { + "epoch": 27.42, + "learning_rate": 4.298014967205062e-06, + "loss": 0.932, + "step": 316840 + }, + { + "epoch": 27.43, + "learning_rate": 4.296570256291716e-06, + "loss": 0.9148, + "step": 316850 + }, + { + "epoch": 27.43, + "learning_rate": 4.29512554537837e-06, + "loss": 0.8707, + "step": 316860 + }, + { + "epoch": 27.43, + "learning_rate": 4.293680834465024e-06, + "loss": 0.9446, + "step": 316870 + }, + { + "epoch": 27.43, + "learning_rate": 4.292236123551678e-06, + "loss": 0.9028, + "step": 316880 + }, + { + "epoch": 27.43, + "learning_rate": 4.290791412638331e-06, + "loss": 0.8751, + "step": 316890 + }, + { + "epoch": 27.43, + "learning_rate": 4.2893467017249845e-06, + "loss": 0.8628, + "step": 316900 + }, + { + "epoch": 27.43, + "learning_rate": 4.287901990811639e-06, + "loss": 0.9675, + "step": 316910 + }, + { + "epoch": 27.43, + "learning_rate": 4.286457279898293e-06, + "loss": 0.9433, + "step": 316920 + }, + { + "epoch": 27.43, + "learning_rate": 4.285012568984946e-06, + "loss": 0.957, + "step": 316930 + }, + { + "epoch": 27.43, + "learning_rate": 4.2835678580716e-06, + "loss": 0.9261, + "step": 316940 + }, + { + "epoch": 27.43, + "learning_rate": 4.282123147158254e-06, + "loss": 0.9369, + "step": 316950 + }, + { + "epoch": 27.44, + "learning_rate": 4.280678436244907e-06, + "loss": 0.9351, + "step": 316960 + }, + { + "epoch": 27.44, + "learning_rate": 4.279233725331562e-06, + "loss": 0.9348, + "step": 316970 + }, + { + "epoch": 27.44, + "learning_rate": 4.277789014418215e-06, + "loss": 0.9391, + "step": 316980 + }, + { + "epoch": 27.44, + "learning_rate": 4.276344303504868e-06, + "loss": 0.8623, + "step": 316990 + }, + { + "epoch": 27.44, + "learning_rate": 4.274899592591523e-06, + "loss": 0.9345, + "step": 317000 + }, + { + "epoch": 27.44, + "learning_rate": 4.273454881678177e-06, + "loss": 0.9126, + "step": 317010 + }, + { + "epoch": 27.44, + "learning_rate": 4.27201017076483e-06, + "loss": 0.8905, + "step": 317020 + }, + { + "epoch": 27.44, + "learning_rate": 4.270565459851484e-06, + "loss": 0.8776, + "step": 317030 + }, + { + "epoch": 27.44, + "learning_rate": 4.269120748938138e-06, + "loss": 0.8969, + "step": 317040 + }, + { + "epoch": 27.44, + "learning_rate": 4.267676038024791e-06, + "loss": 0.8835, + "step": 317050 + }, + { + "epoch": 27.44, + "learning_rate": 4.2662313271114456e-06, + "loss": 0.9346, + "step": 317060 + }, + { + "epoch": 27.44, + "learning_rate": 4.264786616198099e-06, + "loss": 0.8667, + "step": 317070 + }, + { + "epoch": 27.45, + "learning_rate": 4.263341905284752e-06, + "loss": 0.909, + "step": 317080 + }, + { + "epoch": 27.45, + "learning_rate": 4.261897194371407e-06, + "loss": 0.8773, + "step": 317090 + }, + { + "epoch": 27.45, + "learning_rate": 4.260452483458061e-06, + "loss": 0.9667, + "step": 317100 + }, + { + "epoch": 27.45, + "learning_rate": 4.259007772544714e-06, + "loss": 0.9361, + "step": 317110 + }, + { + "epoch": 27.45, + "learning_rate": 4.257563061631368e-06, + "loss": 0.9112, + "step": 317120 + }, + { + "epoch": 27.45, + "learning_rate": 4.256118350718021e-06, + "loss": 0.8904, + "step": 317130 + }, + { + "epoch": 27.45, + "learning_rate": 4.254673639804675e-06, + "loss": 0.9524, + "step": 317140 + }, + { + "epoch": 27.45, + "learning_rate": 4.2532289288913295e-06, + "loss": 0.8901, + "step": 317150 + }, + { + "epoch": 27.45, + "learning_rate": 4.251784217977983e-06, + "loss": 0.87, + "step": 317160 + }, + { + "epoch": 27.45, + "learning_rate": 4.250339507064636e-06, + "loss": 0.9397, + "step": 317170 + }, + { + "epoch": 27.45, + "learning_rate": 4.2488947961512905e-06, + "loss": 0.9032, + "step": 317180 + }, + { + "epoch": 27.46, + "learning_rate": 4.247450085237944e-06, + "loss": 0.843, + "step": 317190 + }, + { + "epoch": 27.46, + "learning_rate": 4.246005374324598e-06, + "loss": 0.8594, + "step": 317200 + }, + { + "epoch": 27.46, + "learning_rate": 4.2445606634112516e-06, + "loss": 0.877, + "step": 317210 + }, + { + "epoch": 27.46, + "learning_rate": 4.243115952497905e-06, + "loss": 0.944, + "step": 317220 + }, + { + "epoch": 27.46, + "learning_rate": 4.241671241584559e-06, + "loss": 0.9467, + "step": 317230 + }, + { + "epoch": 27.46, + "learning_rate": 4.2402265306712134e-06, + "loss": 0.9308, + "step": 317240 + }, + { + "epoch": 27.46, + "learning_rate": 4.238781819757867e-06, + "loss": 0.8717, + "step": 317250 + }, + { + "epoch": 27.46, + "learning_rate": 4.23733710884452e-06, + "loss": 0.8821, + "step": 317260 + }, + { + "epoch": 27.46, + "learning_rate": 4.2358923979311745e-06, + "loss": 0.8519, + "step": 317270 + }, + { + "epoch": 27.46, + "learning_rate": 4.234447687017828e-06, + "loss": 0.8868, + "step": 317280 + }, + { + "epoch": 27.46, + "learning_rate": 4.233002976104482e-06, + "loss": 0.8756, + "step": 317290 + }, + { + "epoch": 27.46, + "learning_rate": 4.2315582651911355e-06, + "loss": 0.9379, + "step": 317300 + }, + { + "epoch": 27.47, + "learning_rate": 4.230113554277789e-06, + "loss": 0.9298, + "step": 317310 + }, + { + "epoch": 27.47, + "learning_rate": 4.228668843364443e-06, + "loss": 0.8939, + "step": 317320 + }, + { + "epoch": 27.47, + "learning_rate": 4.227224132451097e-06, + "loss": 0.9355, + "step": 317330 + }, + { + "epoch": 27.47, + "learning_rate": 4.225779421537751e-06, + "loss": 0.9645, + "step": 317340 + }, + { + "epoch": 27.47, + "learning_rate": 4.224334710624404e-06, + "loss": 0.8693, + "step": 317350 + }, + { + "epoch": 27.47, + "learning_rate": 4.2228899997110576e-06, + "loss": 0.8996, + "step": 317360 + }, + { + "epoch": 27.47, + "learning_rate": 4.221445288797712e-06, + "loss": 0.9127, + "step": 317370 + }, + { + "epoch": 27.47, + "learning_rate": 4.220000577884366e-06, + "loss": 0.9482, + "step": 317380 + }, + { + "epoch": 27.47, + "learning_rate": 4.218555866971019e-06, + "loss": 0.9317, + "step": 317390 + }, + { + "epoch": 27.47, + "learning_rate": 4.217111156057673e-06, + "loss": 0.9681, + "step": 317400 + }, + { + "epoch": 27.47, + "learning_rate": 4.215666445144327e-06, + "loss": 0.8873, + "step": 317410 + }, + { + "epoch": 27.47, + "learning_rate": 4.2142217342309805e-06, + "loss": 0.9387, + "step": 317420 + }, + { + "epoch": 27.48, + "learning_rate": 4.212777023317635e-06, + "loss": 0.9082, + "step": 317430 + }, + { + "epoch": 27.48, + "learning_rate": 4.211332312404288e-06, + "loss": 0.8814, + "step": 317440 + }, + { + "epoch": 27.48, + "learning_rate": 4.2098876014909415e-06, + "loss": 0.8797, + "step": 317450 + }, + { + "epoch": 27.48, + "learning_rate": 4.208442890577596e-06, + "loss": 0.947, + "step": 317460 + }, + { + "epoch": 27.48, + "learning_rate": 4.20699817966425e-06, + "loss": 0.9061, + "step": 317470 + }, + { + "epoch": 27.48, + "learning_rate": 4.2055534687509025e-06, + "loss": 0.9316, + "step": 317480 + }, + { + "epoch": 27.48, + "learning_rate": 4.204108757837557e-06, + "loss": 0.9122, + "step": 317490 + }, + { + "epoch": 27.48, + "learning_rate": 4.202664046924211e-06, + "loss": 0.8964, + "step": 317500 + }, + { + "epoch": 27.48, + "learning_rate": 4.201219336010864e-06, + "loss": 0.9398, + "step": 317510 + }, + { + "epoch": 27.48, + "learning_rate": 4.199774625097519e-06, + "loss": 0.9031, + "step": 317520 + }, + { + "epoch": 27.48, + "learning_rate": 4.198329914184172e-06, + "loss": 0.8922, + "step": 317530 + }, + { + "epoch": 27.49, + "learning_rate": 4.1968852032708254e-06, + "loss": 0.8294, + "step": 317540 + }, + { + "epoch": 27.49, + "learning_rate": 4.19544049235748e-06, + "loss": 0.8889, + "step": 317550 + }, + { + "epoch": 27.49, + "learning_rate": 4.193995781444134e-06, + "loss": 0.8907, + "step": 317560 + }, + { + "epoch": 27.49, + "learning_rate": 4.1925510705307865e-06, + "loss": 0.9102, + "step": 317570 + }, + { + "epoch": 27.49, + "learning_rate": 4.191106359617441e-06, + "loss": 0.9436, + "step": 317580 + }, + { + "epoch": 27.49, + "learning_rate": 4.189661648704094e-06, + "loss": 0.9422, + "step": 317590 + }, + { + "epoch": 27.49, + "learning_rate": 4.188216937790748e-06, + "loss": 0.891, + "step": 317600 + }, + { + "epoch": 27.49, + "learning_rate": 4.186772226877403e-06, + "loss": 0.8482, + "step": 317610 + }, + { + "epoch": 27.49, + "learning_rate": 4.185327515964055e-06, + "loss": 0.9085, + "step": 317620 + }, + { + "epoch": 27.49, + "learning_rate": 4.183882805050709e-06, + "loss": 0.8832, + "step": 317630 + }, + { + "epoch": 27.49, + "learning_rate": 4.182438094137364e-06, + "loss": 0.9492, + "step": 317640 + }, + { + "epoch": 27.49, + "learning_rate": 4.180993383224017e-06, + "loss": 0.9247, + "step": 317650 + }, + { + "epoch": 27.5, + "learning_rate": 4.1795486723106704e-06, + "loss": 0.8586, + "step": 317660 + }, + { + "epoch": 27.5, + "learning_rate": 4.178103961397325e-06, + "loss": 0.9551, + "step": 317670 + }, + { + "epoch": 27.5, + "learning_rate": 4.176659250483978e-06, + "loss": 0.9123, + "step": 317680 + }, + { + "epoch": 27.5, + "learning_rate": 4.175214539570632e-06, + "loss": 0.8919, + "step": 317690 + }, + { + "epoch": 27.5, + "learning_rate": 4.1737698286572865e-06, + "loss": 0.8967, + "step": 317700 + }, + { + "epoch": 27.5, + "learning_rate": 4.172325117743939e-06, + "loss": 0.9179, + "step": 317710 + }, + { + "epoch": 27.5, + "learning_rate": 4.170880406830593e-06, + "loss": 0.9088, + "step": 317720 + }, + { + "epoch": 27.5, + "learning_rate": 4.169435695917248e-06, + "loss": 0.9141, + "step": 317730 + }, + { + "epoch": 27.5, + "learning_rate": 4.167990985003901e-06, + "loss": 0.8948, + "step": 317740 + }, + { + "epoch": 27.5, + "learning_rate": 4.166546274090554e-06, + "loss": 0.8831, + "step": 317750 + }, + { + "epoch": 27.5, + "learning_rate": 4.165101563177209e-06, + "loss": 0.8904, + "step": 317760 + }, + { + "epoch": 27.51, + "learning_rate": 4.163656852263862e-06, + "loss": 0.8868, + "step": 317770 + }, + { + "epoch": 27.51, + "learning_rate": 4.162212141350516e-06, + "loss": 0.9212, + "step": 317780 + }, + { + "epoch": 27.51, + "learning_rate": 4.16076743043717e-06, + "loss": 0.8974, + "step": 317790 + }, + { + "epoch": 27.51, + "learning_rate": 4.159322719523823e-06, + "loss": 0.9213, + "step": 317800 + }, + { + "epoch": 27.51, + "learning_rate": 4.157878008610477e-06, + "loss": 0.8218, + "step": 317810 + }, + { + "epoch": 27.51, + "learning_rate": 4.156433297697131e-06, + "loss": 0.8989, + "step": 317820 + }, + { + "epoch": 27.51, + "learning_rate": 4.154988586783785e-06, + "loss": 0.8811, + "step": 317830 + }, + { + "epoch": 27.51, + "learning_rate": 4.153543875870438e-06, + "loss": 0.8667, + "step": 317840 + }, + { + "epoch": 27.51, + "learning_rate": 4.152099164957092e-06, + "loss": 0.8222, + "step": 317850 + }, + { + "epoch": 27.51, + "learning_rate": 4.150654454043746e-06, + "loss": 0.8799, + "step": 317860 + }, + { + "epoch": 27.51, + "learning_rate": 4.1492097431304e-06, + "loss": 0.9314, + "step": 317870 + }, + { + "epoch": 27.51, + "learning_rate": 4.147765032217054e-06, + "loss": 0.9074, + "step": 317880 + }, + { + "epoch": 27.52, + "learning_rate": 4.146320321303707e-06, + "loss": 0.8688, + "step": 317890 + }, + { + "epoch": 27.52, + "learning_rate": 4.144875610390361e-06, + "loss": 0.8953, + "step": 317900 + }, + { + "epoch": 27.52, + "learning_rate": 4.143430899477015e-06, + "loss": 0.8963, + "step": 317910 + }, + { + "epoch": 27.52, + "learning_rate": 4.141986188563669e-06, + "loss": 0.8986, + "step": 317920 + }, + { + "epoch": 27.52, + "learning_rate": 4.140541477650322e-06, + "loss": 0.8675, + "step": 317930 + }, + { + "epoch": 27.52, + "learning_rate": 4.139096766736976e-06, + "loss": 0.9217, + "step": 317940 + }, + { + "epoch": 27.52, + "learning_rate": 4.13765205582363e-06, + "loss": 0.8685, + "step": 317950 + }, + { + "epoch": 27.52, + "learning_rate": 4.136207344910284e-06, + "loss": 0.8509, + "step": 317960 + }, + { + "epoch": 27.52, + "learning_rate": 4.1347626339969375e-06, + "loss": 0.8929, + "step": 317970 + }, + { + "epoch": 27.52, + "learning_rate": 4.133317923083591e-06, + "loss": 0.86, + "step": 317980 + }, + { + "epoch": 27.52, + "learning_rate": 4.131873212170245e-06, + "loss": 0.9046, + "step": 317990 + }, + { + "epoch": 27.53, + "learning_rate": 4.1304285012568986e-06, + "loss": 0.9099, + "step": 318000 + }, + { + "epoch": 27.53, + "learning_rate": 4.128983790343553e-06, + "loss": 0.9325, + "step": 318010 + }, + { + "epoch": 27.53, + "learning_rate": 4.127539079430206e-06, + "loss": 0.9222, + "step": 318020 + }, + { + "epoch": 27.53, + "learning_rate": 4.12609436851686e-06, + "loss": 0.8806, + "step": 318030 + }, + { + "epoch": 27.53, + "learning_rate": 4.124649657603514e-06, + "loss": 0.9379, + "step": 318040 + }, + { + "epoch": 27.53, + "learning_rate": 4.123204946690167e-06, + "loss": 0.8533, + "step": 318050 + }, + { + "epoch": 27.53, + "learning_rate": 4.1217602357768215e-06, + "loss": 0.9072, + "step": 318060 + }, + { + "epoch": 27.53, + "learning_rate": 4.120315524863475e-06, + "loss": 0.9141, + "step": 318070 + }, + { + "epoch": 27.53, + "learning_rate": 4.118870813950128e-06, + "loss": 0.9346, + "step": 318080 + }, + { + "epoch": 27.53, + "learning_rate": 4.1174261030367825e-06, + "loss": 0.9004, + "step": 318090 + }, + { + "epoch": 27.53, + "learning_rate": 4.115981392123437e-06, + "loss": 0.9224, + "step": 318100 + }, + { + "epoch": 27.53, + "learning_rate": 4.11453668121009e-06, + "loss": 0.8943, + "step": 318110 + }, + { + "epoch": 27.54, + "learning_rate": 4.1130919702967435e-06, + "loss": 0.9692, + "step": 318120 + }, + { + "epoch": 27.54, + "learning_rate": 4.111647259383398e-06, + "loss": 0.8734, + "step": 318130 + }, + { + "epoch": 27.54, + "learning_rate": 4.110202548470051e-06, + "loss": 0.9387, + "step": 318140 + }, + { + "epoch": 27.54, + "learning_rate": 4.108757837556705e-06, + "loss": 0.9256, + "step": 318150 + }, + { + "epoch": 27.54, + "learning_rate": 4.107313126643359e-06, + "loss": 0.9167, + "step": 318160 + }, + { + "epoch": 27.54, + "learning_rate": 4.105868415730012e-06, + "loss": 0.8948, + "step": 318170 + }, + { + "epoch": 27.54, + "learning_rate": 4.1044237048166664e-06, + "loss": 0.9209, + "step": 318180 + }, + { + "epoch": 27.54, + "learning_rate": 4.102978993903321e-06, + "loss": 0.9032, + "step": 318190 + }, + { + "epoch": 27.54, + "learning_rate": 4.101534282989974e-06, + "loss": 0.8838, + "step": 318200 + }, + { + "epoch": 27.54, + "learning_rate": 4.1000895720766275e-06, + "loss": 0.8872, + "step": 318210 + }, + { + "epoch": 27.54, + "learning_rate": 4.098644861163282e-06, + "loss": 0.912, + "step": 318220 + }, + { + "epoch": 27.55, + "learning_rate": 4.097200150249935e-06, + "loss": 0.9361, + "step": 318230 + }, + { + "epoch": 27.55, + "learning_rate": 4.095755439336589e-06, + "loss": 0.8696, + "step": 318240 + }, + { + "epoch": 27.55, + "learning_rate": 4.094310728423243e-06, + "loss": 0.8804, + "step": 318250 + }, + { + "epoch": 27.55, + "learning_rate": 4.092866017509896e-06, + "loss": 0.9438, + "step": 318260 + }, + { + "epoch": 27.55, + "learning_rate": 4.09142130659655e-06, + "loss": 0.9117, + "step": 318270 + }, + { + "epoch": 27.55, + "learning_rate": 4.089976595683204e-06, + "loss": 0.9338, + "step": 318280 + }, + { + "epoch": 27.55, + "learning_rate": 4.088531884769858e-06, + "loss": 0.8422, + "step": 318290 + }, + { + "epoch": 27.55, + "learning_rate": 4.087087173856511e-06, + "loss": 0.9276, + "step": 318300 + }, + { + "epoch": 27.55, + "learning_rate": 4.085642462943165e-06, + "loss": 0.9006, + "step": 318310 + }, + { + "epoch": 27.55, + "learning_rate": 4.084197752029819e-06, + "loss": 0.9066, + "step": 318320 + }, + { + "epoch": 27.55, + "learning_rate": 4.082753041116473e-06, + "loss": 0.9153, + "step": 318330 + }, + { + "epoch": 27.55, + "learning_rate": 4.081308330203127e-06, + "loss": 0.8604, + "step": 318340 + }, + { + "epoch": 27.56, + "learning_rate": 4.07986361928978e-06, + "loss": 0.951, + "step": 318350 + }, + { + "epoch": 27.56, + "learning_rate": 4.078418908376434e-06, + "loss": 0.9441, + "step": 318360 + }, + { + "epoch": 27.56, + "learning_rate": 4.076974197463088e-06, + "loss": 0.9325, + "step": 318370 + }, + { + "epoch": 27.56, + "learning_rate": 4.075529486549742e-06, + "loss": 0.9261, + "step": 318380 + }, + { + "epoch": 27.56, + "learning_rate": 4.074084775636395e-06, + "loss": 1.0094, + "step": 318390 + }, + { + "epoch": 27.56, + "learning_rate": 4.072640064723049e-06, + "loss": 0.911, + "step": 318400 + }, + { + "epoch": 27.56, + "learning_rate": 4.071195353809703e-06, + "loss": 0.8889, + "step": 318410 + }, + { + "epoch": 27.56, + "learning_rate": 4.069750642896357e-06, + "loss": 0.9287, + "step": 318420 + }, + { + "epoch": 27.56, + "learning_rate": 4.068305931983011e-06, + "loss": 0.8397, + "step": 318430 + }, + { + "epoch": 27.56, + "learning_rate": 4.066861221069664e-06, + "loss": 0.9066, + "step": 318440 + }, + { + "epoch": 27.56, + "learning_rate": 4.065416510156318e-06, + "loss": 0.9112, + "step": 318450 + }, + { + "epoch": 27.56, + "learning_rate": 4.063971799242972e-06, + "loss": 0.9168, + "step": 318460 + }, + { + "epoch": 27.57, + "learning_rate": 4.062527088329626e-06, + "loss": 0.9488, + "step": 318470 + }, + { + "epoch": 27.57, + "learning_rate": 4.0610823774162785e-06, + "loss": 0.9106, + "step": 318480 + }, + { + "epoch": 27.57, + "learning_rate": 4.059637666502933e-06, + "loss": 0.8479, + "step": 318490 + }, + { + "epoch": 27.57, + "learning_rate": 4.058192955589587e-06, + "loss": 0.9418, + "step": 318500 + }, + { + "epoch": 27.57, + "learning_rate": 4.05674824467624e-06, + "loss": 0.8592, + "step": 318510 + }, + { + "epoch": 27.57, + "learning_rate": 4.055303533762895e-06, + "loss": 0.8931, + "step": 318520 + }, + { + "epoch": 27.57, + "learning_rate": 4.053858822849548e-06, + "loss": 0.9858, + "step": 318530 + }, + { + "epoch": 27.57, + "learning_rate": 4.052414111936201e-06, + "loss": 0.9001, + "step": 318540 + }, + { + "epoch": 27.57, + "learning_rate": 4.050969401022856e-06, + "loss": 0.8802, + "step": 318550 + }, + { + "epoch": 27.57, + "learning_rate": 4.04952469010951e-06, + "loss": 0.9561, + "step": 318560 + }, + { + "epoch": 27.57, + "learning_rate": 4.048079979196162e-06, + "loss": 0.8965, + "step": 318570 + }, + { + "epoch": 27.58, + "learning_rate": 4.046635268282817e-06, + "loss": 0.9357, + "step": 318580 + }, + { + "epoch": 27.58, + "learning_rate": 4.045190557369471e-06, + "loss": 0.8906, + "step": 318590 + }, + { + "epoch": 27.58, + "learning_rate": 4.043745846456124e-06, + "loss": 0.8824, + "step": 318600 + }, + { + "epoch": 27.58, + "learning_rate": 4.0423011355427785e-06, + "loss": 0.8365, + "step": 318610 + }, + { + "epoch": 27.58, + "learning_rate": 4.040856424629432e-06, + "loss": 0.864, + "step": 318620 + }, + { + "epoch": 27.58, + "learning_rate": 4.039411713716085e-06, + "loss": 1.0024, + "step": 318630 + }, + { + "epoch": 27.58, + "learning_rate": 4.0379670028027396e-06, + "loss": 0.9562, + "step": 318640 + }, + { + "epoch": 27.58, + "learning_rate": 4.036522291889394e-06, + "loss": 0.8526, + "step": 318650 + }, + { + "epoch": 27.58, + "learning_rate": 4.035077580976046e-06, + "loss": 0.9318, + "step": 318660 + }, + { + "epoch": 27.58, + "learning_rate": 4.033632870062701e-06, + "loss": 0.9449, + "step": 318670 + }, + { + "epoch": 27.58, + "learning_rate": 4.032188159149355e-06, + "loss": 0.8861, + "step": 318680 + }, + { + "epoch": 27.58, + "learning_rate": 4.030743448236008e-06, + "loss": 0.9398, + "step": 318690 + }, + { + "epoch": 27.59, + "learning_rate": 4.0292987373226625e-06, + "loss": 0.9424, + "step": 318700 + }, + { + "epoch": 27.59, + "learning_rate": 4.027854026409315e-06, + "loss": 0.8855, + "step": 318710 + }, + { + "epoch": 27.59, + "learning_rate": 4.026409315495969e-06, + "loss": 0.8823, + "step": 318720 + }, + { + "epoch": 27.59, + "learning_rate": 4.0249646045826235e-06, + "loss": 0.9357, + "step": 318730 + }, + { + "epoch": 27.59, + "learning_rate": 4.023519893669277e-06, + "loss": 0.9738, + "step": 318740 + }, + { + "epoch": 27.59, + "learning_rate": 4.02207518275593e-06, + "loss": 0.9354, + "step": 318750 + }, + { + "epoch": 27.59, + "learning_rate": 4.0206304718425845e-06, + "loss": 0.912, + "step": 318760 + }, + { + "epoch": 27.59, + "learning_rate": 4.019185760929238e-06, + "loss": 0.9144, + "step": 318770 + }, + { + "epoch": 27.59, + "learning_rate": 4.017741050015892e-06, + "loss": 0.9162, + "step": 318780 + }, + { + "epoch": 27.59, + "learning_rate": 4.016296339102546e-06, + "loss": 0.8574, + "step": 318790 + }, + { + "epoch": 27.59, + "learning_rate": 4.014851628189199e-06, + "loss": 0.8992, + "step": 318800 + }, + { + "epoch": 27.6, + "learning_rate": 4.013406917275853e-06, + "loss": 1.0013, + "step": 318810 + }, + { + "epoch": 27.6, + "learning_rate": 4.0119622063625074e-06, + "loss": 0.9359, + "step": 318820 + }, + { + "epoch": 27.6, + "learning_rate": 4.010517495449161e-06, + "loss": 0.8509, + "step": 318830 + }, + { + "epoch": 27.6, + "learning_rate": 4.009072784535814e-06, + "loss": 0.9022, + "step": 318840 + }, + { + "epoch": 27.6, + "learning_rate": 4.0076280736224685e-06, + "loss": 0.93, + "step": 318850 + }, + { + "epoch": 27.6, + "learning_rate": 4.006183362709122e-06, + "loss": 0.913, + "step": 318860 + }, + { + "epoch": 27.6, + "learning_rate": 4.004738651795776e-06, + "loss": 0.9016, + "step": 318870 + }, + { + "epoch": 27.6, + "learning_rate": 4.00329394088243e-06, + "loss": 0.902, + "step": 318880 + }, + { + "epoch": 27.6, + "learning_rate": 4.001849229969083e-06, + "loss": 0.9274, + "step": 318890 + }, + { + "epoch": 27.6, + "learning_rate": 4.000404519055737e-06, + "loss": 0.9267, + "step": 318900 + }, + { + "epoch": 27.6, + "learning_rate": 3.998959808142391e-06, + "loss": 0.8655, + "step": 318910 + }, + { + "epoch": 27.6, + "learning_rate": 3.997515097229045e-06, + "loss": 0.8567, + "step": 318920 + }, + { + "epoch": 27.61, + "learning_rate": 3.996070386315698e-06, + "loss": 0.8753, + "step": 318930 + }, + { + "epoch": 27.61, + "learning_rate": 3.9946256754023516e-06, + "loss": 0.918, + "step": 318940 + }, + { + "epoch": 27.61, + "learning_rate": 3.993180964489006e-06, + "loss": 0.9214, + "step": 318950 + }, + { + "epoch": 27.61, + "learning_rate": 3.99173625357566e-06, + "loss": 0.9094, + "step": 318960 + }, + { + "epoch": 27.61, + "learning_rate": 3.9902915426623134e-06, + "loss": 0.9569, + "step": 318970 + }, + { + "epoch": 27.61, + "learning_rate": 3.988846831748967e-06, + "loss": 0.9333, + "step": 318980 + }, + { + "epoch": 27.61, + "learning_rate": 3.987402120835621e-06, + "loss": 0.8645, + "step": 318990 + }, + { + "epoch": 27.61, + "learning_rate": 3.9859574099222745e-06, + "loss": 0.9769, + "step": 319000 + }, + { + "epoch": 27.61, + "learning_rate": 3.984512699008929e-06, + "loss": 0.866, + "step": 319010 + }, + { + "epoch": 27.61, + "learning_rate": 3.983067988095582e-06, + "loss": 0.9106, + "step": 319020 + }, + { + "epoch": 27.61, + "learning_rate": 3.9816232771822355e-06, + "loss": 0.8496, + "step": 319030 + }, + { + "epoch": 27.62, + "learning_rate": 3.98017856626889e-06, + "loss": 0.879, + "step": 319040 + }, + { + "epoch": 27.62, + "learning_rate": 3.978733855355544e-06, + "loss": 0.8397, + "step": 319050 + }, + { + "epoch": 27.62, + "learning_rate": 3.977289144442197e-06, + "loss": 0.8781, + "step": 319060 + }, + { + "epoch": 27.62, + "learning_rate": 3.975844433528851e-06, + "loss": 0.8717, + "step": 319070 + }, + { + "epoch": 27.62, + "learning_rate": 3.974399722615505e-06, + "loss": 0.9365, + "step": 319080 + }, + { + "epoch": 27.62, + "learning_rate": 3.972955011702158e-06, + "loss": 0.8878, + "step": 319090 + }, + { + "epoch": 27.62, + "learning_rate": 3.971510300788813e-06, + "loss": 0.8981, + "step": 319100 + }, + { + "epoch": 27.62, + "learning_rate": 3.970065589875466e-06, + "loss": 0.9068, + "step": 319110 + }, + { + "epoch": 27.62, + "learning_rate": 3.9686208789621195e-06, + "loss": 0.9109, + "step": 319120 + }, + { + "epoch": 27.62, + "learning_rate": 3.967176168048774e-06, + "loss": 0.9057, + "step": 319130 + }, + { + "epoch": 27.62, + "learning_rate": 3.965731457135428e-06, + "loss": 0.8963, + "step": 319140 + }, + { + "epoch": 27.62, + "learning_rate": 3.964286746222081e-06, + "loss": 0.9353, + "step": 319150 + }, + { + "epoch": 27.63, + "learning_rate": 3.962842035308735e-06, + "loss": 0.8964, + "step": 319160 + }, + { + "epoch": 27.63, + "learning_rate": 3.961397324395388e-06, + "loss": 0.9401, + "step": 319170 + }, + { + "epoch": 27.63, + "learning_rate": 3.959952613482042e-06, + "loss": 0.8416, + "step": 319180 + }, + { + "epoch": 27.63, + "learning_rate": 3.958507902568697e-06, + "loss": 0.887, + "step": 319190 + }, + { + "epoch": 27.63, + "learning_rate": 3.95706319165535e-06, + "loss": 0.9462, + "step": 319200 + }, + { + "epoch": 27.63, + "learning_rate": 3.955618480742003e-06, + "loss": 0.8822, + "step": 319210 + }, + { + "epoch": 27.63, + "learning_rate": 3.954173769828658e-06, + "loss": 0.9057, + "step": 319220 + }, + { + "epoch": 27.63, + "learning_rate": 3.952729058915311e-06, + "loss": 0.8634, + "step": 319230 + }, + { + "epoch": 27.63, + "learning_rate": 3.951284348001965e-06, + "loss": 0.8763, + "step": 319240 + }, + { + "epoch": 27.63, + "learning_rate": 3.949839637088619e-06, + "loss": 0.9192, + "step": 319250 + }, + { + "epoch": 27.63, + "learning_rate": 3.948394926175272e-06, + "loss": 0.8937, + "step": 319260 + }, + { + "epoch": 27.64, + "learning_rate": 3.946950215261926e-06, + "loss": 0.8881, + "step": 319270 + }, + { + "epoch": 27.64, + "learning_rate": 3.9455055043485805e-06, + "loss": 0.9028, + "step": 319280 + }, + { + "epoch": 27.64, + "learning_rate": 3.944060793435234e-06, + "loss": 0.9811, + "step": 319290 + }, + { + "epoch": 27.64, + "learning_rate": 3.942616082521887e-06, + "loss": 0.9052, + "step": 319300 + }, + { + "epoch": 27.64, + "learning_rate": 3.941171371608542e-06, + "loss": 0.8953, + "step": 319310 + }, + { + "epoch": 27.64, + "learning_rate": 3.939726660695195e-06, + "loss": 0.9209, + "step": 319320 + }, + { + "epoch": 27.64, + "learning_rate": 3.938281949781849e-06, + "loss": 0.8521, + "step": 319330 + }, + { + "epoch": 27.64, + "learning_rate": 3.936837238868503e-06, + "loss": 0.8821, + "step": 319340 + }, + { + "epoch": 27.64, + "learning_rate": 3.935392527955156e-06, + "loss": 0.8979, + "step": 319350 + }, + { + "epoch": 27.64, + "learning_rate": 3.93394781704181e-06, + "loss": 0.9164, + "step": 319360 + }, + { + "epoch": 27.64, + "learning_rate": 3.932503106128464e-06, + "loss": 0.9122, + "step": 319370 + }, + { + "epoch": 27.64, + "learning_rate": 3.931058395215118e-06, + "loss": 0.9484, + "step": 319380 + }, + { + "epoch": 27.65, + "learning_rate": 3.929613684301771e-06, + "loss": 0.8956, + "step": 319390 + }, + { + "epoch": 27.65, + "learning_rate": 3.928168973388425e-06, + "loss": 0.907, + "step": 319400 + }, + { + "epoch": 27.65, + "learning_rate": 3.926724262475079e-06, + "loss": 0.8268, + "step": 319410 + }, + { + "epoch": 27.65, + "learning_rate": 3.925279551561733e-06, + "loss": 0.8941, + "step": 319420 + }, + { + "epoch": 27.65, + "learning_rate": 3.9238348406483866e-06, + "loss": 0.8728, + "step": 319430 + }, + { + "epoch": 27.65, + "learning_rate": 3.92239012973504e-06, + "loss": 0.9465, + "step": 319440 + }, + { + "epoch": 27.65, + "learning_rate": 3.920945418821694e-06, + "loss": 0.9293, + "step": 319450 + }, + { + "epoch": 27.65, + "learning_rate": 3.919500707908348e-06, + "loss": 0.8709, + "step": 319460 + }, + { + "epoch": 27.65, + "learning_rate": 3.918055996995002e-06, + "loss": 0.855, + "step": 319470 + }, + { + "epoch": 27.65, + "learning_rate": 3.916611286081655e-06, + "loss": 0.9274, + "step": 319480 + }, + { + "epoch": 27.65, + "learning_rate": 3.915166575168309e-06, + "loss": 0.9091, + "step": 319490 + }, + { + "epoch": 27.65, + "learning_rate": 3.913721864254963e-06, + "loss": 0.9259, + "step": 319500 + }, + { + "epoch": 27.66, + "learning_rate": 3.912277153341617e-06, + "loss": 0.8996, + "step": 319510 + }, + { + "epoch": 27.66, + "learning_rate": 3.9108324424282705e-06, + "loss": 0.9185, + "step": 319520 + }, + { + "epoch": 27.66, + "learning_rate": 3.909387731514924e-06, + "loss": 0.8892, + "step": 319530 + }, + { + "epoch": 27.66, + "learning_rate": 3.907943020601578e-06, + "loss": 0.8563, + "step": 319540 + }, + { + "epoch": 27.66, + "learning_rate": 3.9064983096882315e-06, + "loss": 0.8701, + "step": 319550 + }, + { + "epoch": 27.66, + "learning_rate": 3.905053598774886e-06, + "loss": 0.8548, + "step": 319560 + }, + { + "epoch": 27.66, + "learning_rate": 3.903608887861539e-06, + "loss": 0.8864, + "step": 319570 + }, + { + "epoch": 27.66, + "learning_rate": 3.9021641769481926e-06, + "loss": 0.8935, + "step": 319580 + }, + { + "epoch": 27.66, + "learning_rate": 3.900719466034847e-06, + "loss": 0.878, + "step": 319590 + }, + { + "epoch": 27.66, + "learning_rate": 3.8992747551215e-06, + "loss": 0.8511, + "step": 319600 + }, + { + "epoch": 27.66, + "learning_rate": 3.8978300442081544e-06, + "loss": 0.928, + "step": 319610 + }, + { + "epoch": 27.67, + "learning_rate": 3.896385333294808e-06, + "loss": 0.9277, + "step": 319620 + }, + { + "epoch": 27.67, + "learning_rate": 3.894940622381461e-06, + "loss": 0.9441, + "step": 319630 + }, + { + "epoch": 27.67, + "learning_rate": 3.8934959114681155e-06, + "loss": 0.9278, + "step": 319640 + }, + { + "epoch": 27.67, + "learning_rate": 3.89205120055477e-06, + "loss": 0.8716, + "step": 319650 + }, + { + "epoch": 27.67, + "learning_rate": 3.890606489641422e-06, + "loss": 0.9005, + "step": 319660 + }, + { + "epoch": 27.67, + "learning_rate": 3.8891617787280765e-06, + "loss": 0.9456, + "step": 319670 + }, + { + "epoch": 27.67, + "learning_rate": 3.887717067814731e-06, + "loss": 0.9018, + "step": 319680 + }, + { + "epoch": 27.67, + "learning_rate": 3.886272356901384e-06, + "loss": 0.8618, + "step": 319690 + }, + { + "epoch": 27.67, + "learning_rate": 3.884827645988038e-06, + "loss": 0.8958, + "step": 319700 + }, + { + "epoch": 27.67, + "learning_rate": 3.883382935074692e-06, + "loss": 0.9423, + "step": 319710 + }, + { + "epoch": 27.67, + "learning_rate": 3.881938224161345e-06, + "loss": 0.8936, + "step": 319720 + }, + { + "epoch": 27.67, + "learning_rate": 3.880493513247999e-06, + "loss": 0.934, + "step": 319730 + }, + { + "epoch": 27.68, + "learning_rate": 3.879048802334654e-06, + "loss": 0.8367, + "step": 319740 + }, + { + "epoch": 27.68, + "learning_rate": 3.877604091421306e-06, + "loss": 0.9522, + "step": 319750 + }, + { + "epoch": 27.68, + "learning_rate": 3.8761593805079604e-06, + "loss": 0.8581, + "step": 319760 + }, + { + "epoch": 27.68, + "learning_rate": 3.874714669594615e-06, + "loss": 0.9431, + "step": 319770 + }, + { + "epoch": 27.68, + "learning_rate": 3.873269958681268e-06, + "loss": 0.9603, + "step": 319780 + }, + { + "epoch": 27.68, + "learning_rate": 3.871825247767922e-06, + "loss": 0.8968, + "step": 319790 + }, + { + "epoch": 27.68, + "learning_rate": 3.870380536854576e-06, + "loss": 0.9558, + "step": 319800 + }, + { + "epoch": 27.68, + "learning_rate": 3.868935825941229e-06, + "loss": 0.9653, + "step": 319810 + }, + { + "epoch": 27.68, + "learning_rate": 3.867491115027883e-06, + "loss": 0.9233, + "step": 319820 + }, + { + "epoch": 27.68, + "learning_rate": 3.866046404114537e-06, + "loss": 0.9116, + "step": 319830 + }, + { + "epoch": 27.68, + "learning_rate": 3.86460169320119e-06, + "loss": 0.875, + "step": 319840 + }, + { + "epoch": 27.69, + "learning_rate": 3.863156982287844e-06, + "loss": 0.8984, + "step": 319850 + }, + { + "epoch": 27.69, + "learning_rate": 3.861712271374498e-06, + "loss": 0.8556, + "step": 319860 + }, + { + "epoch": 27.69, + "learning_rate": 3.860267560461152e-06, + "loss": 0.8978, + "step": 319870 + }, + { + "epoch": 27.69, + "learning_rate": 3.858822849547806e-06, + "loss": 0.9167, + "step": 319880 + }, + { + "epoch": 27.69, + "learning_rate": 3.857378138634459e-06, + "loss": 0.8966, + "step": 319890 + }, + { + "epoch": 27.69, + "learning_rate": 3.855933427721113e-06, + "loss": 0.8897, + "step": 319900 + }, + { + "epoch": 27.69, + "learning_rate": 3.854488716807767e-06, + "loss": 0.9077, + "step": 319910 + }, + { + "epoch": 27.69, + "learning_rate": 3.853044005894421e-06, + "loss": 0.8695, + "step": 319920 + }, + { + "epoch": 27.69, + "learning_rate": 3.851599294981074e-06, + "loss": 0.8937, + "step": 319930 + }, + { + "epoch": 27.69, + "learning_rate": 3.850154584067728e-06, + "loss": 0.924, + "step": 319940 + }, + { + "epoch": 27.69, + "learning_rate": 3.848709873154382e-06, + "loss": 0.9143, + "step": 319950 + }, + { + "epoch": 27.69, + "learning_rate": 3.847265162241036e-06, + "loss": 0.9451, + "step": 319960 + }, + { + "epoch": 27.7, + "learning_rate": 3.845820451327689e-06, + "loss": 0.9244, + "step": 319970 + }, + { + "epoch": 27.7, + "learning_rate": 3.844375740414343e-06, + "loss": 0.9522, + "step": 319980 + }, + { + "epoch": 27.7, + "learning_rate": 3.842931029500997e-06, + "loss": 0.8641, + "step": 319990 + }, + { + "epoch": 27.7, + "learning_rate": 3.841486318587651e-06, + "loss": 0.9324, + "step": 320000 + }, + { + "epoch": 27.7, + "learning_rate": 3.840041607674305e-06, + "loss": 0.8663, + "step": 320010 + }, + { + "epoch": 27.7, + "learning_rate": 3.838596896760958e-06, + "loss": 0.8636, + "step": 320020 + }, + { + "epoch": 27.7, + "learning_rate": 3.837152185847612e-06, + "loss": 0.9156, + "step": 320030 + }, + { + "epoch": 27.7, + "learning_rate": 3.835707474934266e-06, + "loss": 0.9711, + "step": 320040 + }, + { + "epoch": 27.7, + "learning_rate": 3.83426276402092e-06, + "loss": 0.9127, + "step": 320050 + }, + { + "epoch": 27.7, + "learning_rate": 3.832818053107573e-06, + "loss": 0.8945, + "step": 320060 + }, + { + "epoch": 27.7, + "learning_rate": 3.831373342194227e-06, + "loss": 0.9207, + "step": 320070 + }, + { + "epoch": 27.71, + "learning_rate": 3.829928631280881e-06, + "loss": 0.9489, + "step": 320080 + }, + { + "epoch": 27.71, + "learning_rate": 3.828483920367534e-06, + "loss": 0.8817, + "step": 320090 + }, + { + "epoch": 27.71, + "learning_rate": 3.827039209454189e-06, + "loss": 0.886, + "step": 320100 + }, + { + "epoch": 27.71, + "learning_rate": 3.825594498540842e-06, + "loss": 0.8845, + "step": 320110 + }, + { + "epoch": 27.71, + "learning_rate": 3.824149787627495e-06, + "loss": 0.9512, + "step": 320120 + }, + { + "epoch": 27.71, + "learning_rate": 3.82270507671415e-06, + "loss": 0.8389, + "step": 320130 + }, + { + "epoch": 27.71, + "learning_rate": 3.821260365800804e-06, + "loss": 0.9272, + "step": 320140 + }, + { + "epoch": 27.71, + "learning_rate": 3.819815654887457e-06, + "loss": 0.8506, + "step": 320150 + }, + { + "epoch": 27.71, + "learning_rate": 3.818370943974111e-06, + "loss": 0.9065, + "step": 320160 + }, + { + "epoch": 27.71, + "learning_rate": 3.816926233060765e-06, + "loss": 0.8895, + "step": 320170 + }, + { + "epoch": 27.71, + "learning_rate": 3.815481522147418e-06, + "loss": 0.9071, + "step": 320180 + }, + { + "epoch": 27.71, + "learning_rate": 3.814036811234072e-06, + "loss": 0.8969, + "step": 320190 + }, + { + "epoch": 27.72, + "learning_rate": 3.8125921003207263e-06, + "loss": 0.8986, + "step": 320200 + }, + { + "epoch": 27.72, + "learning_rate": 3.8111473894073797e-06, + "loss": 0.8462, + "step": 320210 + }, + { + "epoch": 27.72, + "learning_rate": 3.8097026784940336e-06, + "loss": 0.8208, + "step": 320220 + }, + { + "epoch": 27.72, + "learning_rate": 3.808257967580688e-06, + "loss": 0.9482, + "step": 320230 + }, + { + "epoch": 27.72, + "learning_rate": 3.8068132566673408e-06, + "loss": 0.937, + "step": 320240 + }, + { + "epoch": 27.72, + "learning_rate": 3.805368545753995e-06, + "loss": 0.9322, + "step": 320250 + }, + { + "epoch": 27.72, + "learning_rate": 3.803923834840649e-06, + "loss": 0.893, + "step": 320260 + }, + { + "epoch": 27.72, + "learning_rate": 3.8024791239273022e-06, + "loss": 0.9233, + "step": 320270 + }, + { + "epoch": 27.72, + "learning_rate": 3.801034413013956e-06, + "loss": 0.9109, + "step": 320280 + }, + { + "epoch": 27.72, + "learning_rate": 3.7995897021006094e-06, + "loss": 0.9052, + "step": 320290 + }, + { + "epoch": 27.72, + "learning_rate": 3.7981449911872637e-06, + "loss": 0.9246, + "step": 320300 + }, + { + "epoch": 27.73, + "learning_rate": 3.7967002802739175e-06, + "loss": 0.9162, + "step": 320310 + }, + { + "epoch": 27.73, + "learning_rate": 3.795255569360571e-06, + "loss": 0.8756, + "step": 320320 + }, + { + "epoch": 27.73, + "learning_rate": 3.7938108584472247e-06, + "loss": 0.849, + "step": 320330 + }, + { + "epoch": 27.73, + "learning_rate": 3.792366147533879e-06, + "loss": 0.9481, + "step": 320340 + }, + { + "epoch": 27.73, + "learning_rate": 3.7909214366205323e-06, + "loss": 0.8757, + "step": 320350 + }, + { + "epoch": 27.73, + "learning_rate": 3.789476725707186e-06, + "loss": 0.9062, + "step": 320360 + }, + { + "epoch": 27.73, + "learning_rate": 3.78803201479384e-06, + "loss": 0.8588, + "step": 320370 + }, + { + "epoch": 27.73, + "learning_rate": 3.7865873038804934e-06, + "loss": 0.9428, + "step": 320380 + }, + { + "epoch": 27.73, + "learning_rate": 3.7851425929671476e-06, + "loss": 0.8883, + "step": 320390 + }, + { + "epoch": 27.73, + "learning_rate": 3.7836978820538014e-06, + "loss": 0.8837, + "step": 320400 + }, + { + "epoch": 27.73, + "learning_rate": 3.782253171140455e-06, + "loss": 0.8802, + "step": 320410 + }, + { + "epoch": 27.73, + "learning_rate": 3.7808084602271087e-06, + "loss": 0.9054, + "step": 320420 + }, + { + "epoch": 27.74, + "learning_rate": 3.779363749313763e-06, + "loss": 0.8565, + "step": 320430 + }, + { + "epoch": 27.74, + "learning_rate": 3.7779190384004163e-06, + "loss": 0.8957, + "step": 320440 + }, + { + "epoch": 27.74, + "learning_rate": 3.77647432748707e-06, + "loss": 0.915, + "step": 320450 + }, + { + "epoch": 27.74, + "learning_rate": 3.775029616573724e-06, + "loss": 0.9029, + "step": 320460 + }, + { + "epoch": 27.74, + "learning_rate": 3.7735849056603773e-06, + "loss": 0.92, + "step": 320470 + }, + { + "epoch": 27.74, + "learning_rate": 3.7721401947470316e-06, + "loss": 0.8976, + "step": 320480 + }, + { + "epoch": 27.74, + "learning_rate": 3.7706954838336854e-06, + "loss": 0.9094, + "step": 320490 + }, + { + "epoch": 27.74, + "learning_rate": 3.7692507729203388e-06, + "loss": 0.8853, + "step": 320500 + }, + { + "epoch": 27.74, + "learning_rate": 3.7678060620069926e-06, + "loss": 0.8464, + "step": 320510 + }, + { + "epoch": 27.74, + "learning_rate": 3.766361351093646e-06, + "loss": 0.9355, + "step": 320520 + }, + { + "epoch": 27.74, + "learning_rate": 3.7649166401803e-06, + "loss": 0.892, + "step": 320530 + }, + { + "epoch": 27.75, + "learning_rate": 3.763471929266954e-06, + "loss": 0.8869, + "step": 320540 + }, + { + "epoch": 27.75, + "learning_rate": 3.7620272183536074e-06, + "loss": 0.9432, + "step": 320550 + }, + { + "epoch": 27.75, + "learning_rate": 3.7605825074402613e-06, + "loss": 0.9441, + "step": 320560 + }, + { + "epoch": 27.75, + "learning_rate": 3.7591377965269155e-06, + "loss": 0.8945, + "step": 320570 + }, + { + "epoch": 27.75, + "learning_rate": 3.7576930856135685e-06, + "loss": 0.9247, + "step": 320580 + }, + { + "epoch": 27.75, + "learning_rate": 3.7562483747002227e-06, + "loss": 0.9066, + "step": 320590 + }, + { + "epoch": 27.75, + "learning_rate": 3.7548036637868765e-06, + "loss": 0.9365, + "step": 320600 + }, + { + "epoch": 27.75, + "learning_rate": 3.75335895287353e-06, + "loss": 0.92, + "step": 320610 + }, + { + "epoch": 27.75, + "learning_rate": 3.7519142419601838e-06, + "loss": 0.9027, + "step": 320620 + }, + { + "epoch": 27.75, + "learning_rate": 3.750469531046838e-06, + "loss": 0.9516, + "step": 320630 + }, + { + "epoch": 27.75, + "learning_rate": 3.7490248201334914e-06, + "loss": 0.9035, + "step": 320640 + }, + { + "epoch": 27.75, + "learning_rate": 3.747580109220145e-06, + "loss": 0.8744, + "step": 320650 + }, + { + "epoch": 27.76, + "learning_rate": 3.7461353983067995e-06, + "loss": 0.9498, + "step": 320660 + }, + { + "epoch": 27.76, + "learning_rate": 3.7446906873934524e-06, + "loss": 0.9737, + "step": 320670 + }, + { + "epoch": 27.76, + "learning_rate": 3.7432459764801067e-06, + "loss": 0.9135, + "step": 320680 + }, + { + "epoch": 27.76, + "learning_rate": 3.7418012655667605e-06, + "loss": 0.8873, + "step": 320690 + }, + { + "epoch": 27.76, + "learning_rate": 3.740356554653414e-06, + "loss": 0.9601, + "step": 320700 + }, + { + "epoch": 27.76, + "learning_rate": 3.7389118437400677e-06, + "loss": 0.8346, + "step": 320710 + }, + { + "epoch": 27.76, + "learning_rate": 3.737467132826722e-06, + "loss": 0.8086, + "step": 320720 + }, + { + "epoch": 27.76, + "learning_rate": 3.7360224219133753e-06, + "loss": 0.942, + "step": 320730 + }, + { + "epoch": 27.76, + "learning_rate": 3.734577711000029e-06, + "loss": 0.8993, + "step": 320740 + }, + { + "epoch": 27.76, + "learning_rate": 3.7331330000866825e-06, + "loss": 0.9384, + "step": 320750 + }, + { + "epoch": 27.76, + "learning_rate": 3.7316882891733364e-06, + "loss": 0.9511, + "step": 320760 + }, + { + "epoch": 27.76, + "learning_rate": 3.7302435782599906e-06, + "loss": 0.8774, + "step": 320770 + }, + { + "epoch": 27.77, + "learning_rate": 3.728798867346644e-06, + "loss": 0.9051, + "step": 320780 + }, + { + "epoch": 27.77, + "learning_rate": 3.727354156433298e-06, + "loss": 0.9151, + "step": 320790 + }, + { + "epoch": 27.77, + "learning_rate": 3.7259094455199516e-06, + "loss": 0.9086, + "step": 320800 + }, + { + "epoch": 27.77, + "learning_rate": 3.724464734606605e-06, + "loss": 0.9486, + "step": 320810 + }, + { + "epoch": 27.77, + "learning_rate": 3.7230200236932593e-06, + "loss": 0.8431, + "step": 320820 + }, + { + "epoch": 27.77, + "learning_rate": 3.721575312779913e-06, + "loss": 0.8597, + "step": 320830 + }, + { + "epoch": 27.77, + "learning_rate": 3.7201306018665665e-06, + "loss": 0.8944, + "step": 320840 + }, + { + "epoch": 27.77, + "learning_rate": 3.7186858909532203e-06, + "loss": 0.8661, + "step": 320850 + }, + { + "epoch": 27.77, + "learning_rate": 3.7172411800398746e-06, + "loss": 0.8641, + "step": 320860 + }, + { + "epoch": 27.77, + "learning_rate": 3.715796469126528e-06, + "loss": 0.9263, + "step": 320870 + }, + { + "epoch": 27.77, + "learning_rate": 3.7143517582131818e-06, + "loss": 0.937, + "step": 320880 + }, + { + "epoch": 27.78, + "learning_rate": 3.7129070472998356e-06, + "loss": 0.9489, + "step": 320890 + }, + { + "epoch": 27.78, + "learning_rate": 3.711462336386489e-06, + "loss": 0.8779, + "step": 320900 + }, + { + "epoch": 27.78, + "learning_rate": 3.7100176254731432e-06, + "loss": 0.9344, + "step": 320910 + }, + { + "epoch": 27.78, + "learning_rate": 3.708572914559797e-06, + "loss": 0.8774, + "step": 320920 + }, + { + "epoch": 27.78, + "learning_rate": 3.7071282036464504e-06, + "loss": 0.9457, + "step": 320930 + }, + { + "epoch": 27.78, + "learning_rate": 3.7056834927331043e-06, + "loss": 0.9437, + "step": 320940 + }, + { + "epoch": 27.78, + "learning_rate": 3.7042387818197576e-06, + "loss": 0.9336, + "step": 320950 + }, + { + "epoch": 27.78, + "learning_rate": 3.702794070906412e-06, + "loss": 0.9143, + "step": 320960 + }, + { + "epoch": 27.78, + "learning_rate": 3.7013493599930657e-06, + "loss": 0.8841, + "step": 320970 + }, + { + "epoch": 27.78, + "learning_rate": 3.699904649079719e-06, + "loss": 0.8479, + "step": 320980 + }, + { + "epoch": 27.78, + "learning_rate": 3.698459938166373e-06, + "loss": 0.9117, + "step": 320990 + }, + { + "epoch": 27.78, + "learning_rate": 3.697015227253027e-06, + "loss": 0.8833, + "step": 321000 + }, + { + "epoch": 27.79, + "learning_rate": 3.69557051633968e-06, + "loss": 0.951, + "step": 321010 + }, + { + "epoch": 27.79, + "learning_rate": 3.6941258054263344e-06, + "loss": 0.8753, + "step": 321020 + }, + { + "epoch": 27.79, + "learning_rate": 3.692681094512988e-06, + "loss": 0.8849, + "step": 321030 + }, + { + "epoch": 27.79, + "learning_rate": 3.6912363835996416e-06, + "loss": 0.9404, + "step": 321040 + }, + { + "epoch": 27.79, + "learning_rate": 3.689791672686296e-06, + "loss": 0.8826, + "step": 321050 + }, + { + "epoch": 27.79, + "learning_rate": 3.6883469617729496e-06, + "loss": 0.9165, + "step": 321060 + }, + { + "epoch": 27.79, + "learning_rate": 3.686902250859603e-06, + "loss": 0.8861, + "step": 321070 + }, + { + "epoch": 27.79, + "learning_rate": 3.685457539946257e-06, + "loss": 0.937, + "step": 321080 + }, + { + "epoch": 27.79, + "learning_rate": 3.684012829032911e-06, + "loss": 0.9525, + "step": 321090 + }, + { + "epoch": 27.79, + "learning_rate": 3.682568118119564e-06, + "loss": 0.8835, + "step": 321100 + }, + { + "epoch": 27.79, + "learning_rate": 3.6811234072062183e-06, + "loss": 0.9267, + "step": 321110 + }, + { + "epoch": 27.8, + "learning_rate": 3.679678696292872e-06, + "loss": 0.8306, + "step": 321120 + }, + { + "epoch": 27.8, + "learning_rate": 3.6782339853795255e-06, + "loss": 0.8613, + "step": 321130 + }, + { + "epoch": 27.8, + "learning_rate": 3.6767892744661798e-06, + "loss": 0.8765, + "step": 321140 + }, + { + "epoch": 27.8, + "learning_rate": 3.6753445635528336e-06, + "loss": 0.8763, + "step": 321150 + }, + { + "epoch": 27.8, + "learning_rate": 3.673899852639487e-06, + "loss": 0.9063, + "step": 321160 + }, + { + "epoch": 27.8, + "learning_rate": 3.672455141726141e-06, + "loss": 0.8842, + "step": 321170 + }, + { + "epoch": 27.8, + "learning_rate": 3.671010430812794e-06, + "loss": 0.8606, + "step": 321180 + }, + { + "epoch": 27.8, + "learning_rate": 3.669565719899448e-06, + "loss": 0.9179, + "step": 321190 + }, + { + "epoch": 27.8, + "learning_rate": 3.6681210089861023e-06, + "loss": 0.9584, + "step": 321200 + }, + { + "epoch": 27.8, + "learning_rate": 3.6666762980727557e-06, + "loss": 0.9334, + "step": 321210 + }, + { + "epoch": 27.8, + "learning_rate": 3.6652315871594095e-06, + "loss": 0.8988, + "step": 321220 + }, + { + "epoch": 27.8, + "learning_rate": 3.6637868762460637e-06, + "loss": 0.9436, + "step": 321230 + }, + { + "epoch": 27.81, + "learning_rate": 3.6623421653327167e-06, + "loss": 0.8404, + "step": 321240 + }, + { + "epoch": 27.81, + "learning_rate": 3.660897454419371e-06, + "loss": 0.9229, + "step": 321250 + }, + { + "epoch": 27.81, + "learning_rate": 3.6594527435060247e-06, + "loss": 0.8591, + "step": 321260 + }, + { + "epoch": 27.81, + "learning_rate": 3.658008032592678e-06, + "loss": 0.8918, + "step": 321270 + }, + { + "epoch": 27.81, + "learning_rate": 3.656563321679332e-06, + "loss": 0.8809, + "step": 321280 + }, + { + "epoch": 27.81, + "learning_rate": 3.655118610765986e-06, + "loss": 0.8985, + "step": 321290 + }, + { + "epoch": 27.81, + "learning_rate": 3.6536738998526396e-06, + "loss": 0.8983, + "step": 321300 + }, + { + "epoch": 27.81, + "learning_rate": 3.6522291889392934e-06, + "loss": 0.9396, + "step": 321310 + }, + { + "epoch": 27.81, + "learning_rate": 3.6507844780259472e-06, + "loss": 0.8966, + "step": 321320 + }, + { + "epoch": 27.81, + "learning_rate": 3.6493397671126006e-06, + "loss": 0.9066, + "step": 321330 + }, + { + "epoch": 27.81, + "learning_rate": 3.647895056199255e-06, + "loss": 0.8975, + "step": 321340 + }, + { + "epoch": 27.82, + "learning_rate": 3.6464503452859087e-06, + "loss": 0.9348, + "step": 321350 + }, + { + "epoch": 27.82, + "learning_rate": 3.645005634372562e-06, + "loss": 0.9562, + "step": 321360 + }, + { + "epoch": 27.82, + "learning_rate": 3.643560923459216e-06, + "loss": 0.9236, + "step": 321370 + }, + { + "epoch": 27.82, + "learning_rate": 3.64211621254587e-06, + "loss": 0.9102, + "step": 321380 + }, + { + "epoch": 27.82, + "learning_rate": 3.6406715016325235e-06, + "loss": 0.9089, + "step": 321390 + }, + { + "epoch": 27.82, + "learning_rate": 3.6392267907191774e-06, + "loss": 0.8795, + "step": 321400 + }, + { + "epoch": 27.82, + "learning_rate": 3.6377820798058308e-06, + "loss": 0.9244, + "step": 321410 + }, + { + "epoch": 27.82, + "learning_rate": 3.6363373688924846e-06, + "loss": 0.9595, + "step": 321420 + }, + { + "epoch": 27.82, + "learning_rate": 3.634892657979139e-06, + "loss": 0.9241, + "step": 321430 + }, + { + "epoch": 27.82, + "learning_rate": 3.633447947065792e-06, + "loss": 0.8951, + "step": 321440 + }, + { + "epoch": 27.82, + "learning_rate": 3.632003236152446e-06, + "loss": 0.9137, + "step": 321450 + }, + { + "epoch": 27.82, + "learning_rate": 3.6305585252391e-06, + "loss": 0.8785, + "step": 321460 + }, + { + "epoch": 27.83, + "learning_rate": 3.6291138143257532e-06, + "loss": 0.8613, + "step": 321470 + }, + { + "epoch": 27.83, + "learning_rate": 3.6276691034124075e-06, + "loss": 0.8676, + "step": 321480 + }, + { + "epoch": 27.83, + "learning_rate": 3.6262243924990613e-06, + "loss": 0.9445, + "step": 321490 + }, + { + "epoch": 27.83, + "learning_rate": 3.6247796815857147e-06, + "loss": 0.9272, + "step": 321500 + }, + { + "epoch": 27.83, + "learning_rate": 3.6233349706723685e-06, + "loss": 0.8843, + "step": 321510 + }, + { + "epoch": 27.83, + "learning_rate": 3.6218902597590228e-06, + "loss": 0.9062, + "step": 321520 + }, + { + "epoch": 27.83, + "learning_rate": 3.6204455488456757e-06, + "loss": 0.9702, + "step": 321530 + }, + { + "epoch": 27.83, + "learning_rate": 3.61900083793233e-06, + "loss": 0.9222, + "step": 321540 + }, + { + "epoch": 27.83, + "learning_rate": 3.617556127018984e-06, + "loss": 0.9197, + "step": 321550 + }, + { + "epoch": 27.83, + "learning_rate": 3.616111416105637e-06, + "loss": 0.9605, + "step": 321560 + }, + { + "epoch": 27.83, + "learning_rate": 3.6146667051922914e-06, + "loss": 0.8689, + "step": 321570 + }, + { + "epoch": 27.84, + "learning_rate": 3.6132219942789452e-06, + "loss": 0.9166, + "step": 321580 + }, + { + "epoch": 27.84, + "learning_rate": 3.6117772833655986e-06, + "loss": 0.8608, + "step": 321590 + }, + { + "epoch": 27.84, + "learning_rate": 3.6103325724522525e-06, + "loss": 0.874, + "step": 321600 + }, + { + "epoch": 27.84, + "learning_rate": 3.6088878615389067e-06, + "loss": 0.9253, + "step": 321610 + }, + { + "epoch": 27.84, + "learning_rate": 3.6074431506255597e-06, + "loss": 0.9018, + "step": 321620 + }, + { + "epoch": 27.84, + "learning_rate": 3.605998439712214e-06, + "loss": 0.9145, + "step": 321630 + }, + { + "epoch": 27.84, + "learning_rate": 3.6045537287988673e-06, + "loss": 0.9224, + "step": 321640 + }, + { + "epoch": 27.84, + "learning_rate": 3.603109017885521e-06, + "loss": 0.9188, + "step": 321650 + }, + { + "epoch": 27.84, + "learning_rate": 3.6016643069721754e-06, + "loss": 0.8555, + "step": 321660 + }, + { + "epoch": 27.84, + "learning_rate": 3.6002195960588283e-06, + "loss": 0.9538, + "step": 321670 + }, + { + "epoch": 27.84, + "learning_rate": 3.5987748851454826e-06, + "loss": 0.8727, + "step": 321680 + }, + { + "epoch": 27.84, + "learning_rate": 3.5973301742321364e-06, + "loss": 0.8164, + "step": 321690 + }, + { + "epoch": 27.85, + "learning_rate": 3.59588546331879e-06, + "loss": 0.8988, + "step": 321700 + }, + { + "epoch": 27.85, + "learning_rate": 3.5944407524054436e-06, + "loss": 0.8934, + "step": 321710 + }, + { + "epoch": 27.85, + "learning_rate": 3.592996041492098e-06, + "loss": 0.9244, + "step": 321720 + }, + { + "epoch": 27.85, + "learning_rate": 3.5915513305787513e-06, + "loss": 0.9238, + "step": 321730 + }, + { + "epoch": 27.85, + "learning_rate": 3.590106619665405e-06, + "loss": 0.8472, + "step": 321740 + }, + { + "epoch": 27.85, + "learning_rate": 3.5886619087520593e-06, + "loss": 0.8955, + "step": 321750 + }, + { + "epoch": 27.85, + "learning_rate": 3.5872171978387123e-06, + "loss": 0.9403, + "step": 321760 + }, + { + "epoch": 27.85, + "learning_rate": 3.5857724869253665e-06, + "loss": 0.9573, + "step": 321770 + }, + { + "epoch": 27.85, + "learning_rate": 3.5843277760120203e-06, + "loss": 0.912, + "step": 321780 + }, + { + "epoch": 27.85, + "learning_rate": 3.5828830650986737e-06, + "loss": 0.8876, + "step": 321790 + }, + { + "epoch": 27.85, + "learning_rate": 3.5814383541853276e-06, + "loss": 0.911, + "step": 321800 + }, + { + "epoch": 27.85, + "learning_rate": 3.579993643271982e-06, + "loss": 0.8652, + "step": 321810 + }, + { + "epoch": 27.86, + "learning_rate": 3.578548932358635e-06, + "loss": 0.9566, + "step": 321820 + }, + { + "epoch": 27.86, + "learning_rate": 3.577104221445289e-06, + "loss": 0.9647, + "step": 321830 + }, + { + "epoch": 27.86, + "learning_rate": 3.5756595105319433e-06, + "loss": 0.9637, + "step": 321840 + }, + { + "epoch": 27.86, + "learning_rate": 3.5742147996185962e-06, + "loss": 0.9267, + "step": 321850 + }, + { + "epoch": 27.86, + "learning_rate": 3.5727700887052505e-06, + "loss": 0.9113, + "step": 321860 + }, + { + "epoch": 27.86, + "learning_rate": 3.571325377791904e-06, + "loss": 0.8764, + "step": 321870 + }, + { + "epoch": 27.86, + "learning_rate": 3.5698806668785577e-06, + "loss": 0.9782, + "step": 321880 + }, + { + "epoch": 27.86, + "learning_rate": 3.5684359559652115e-06, + "loss": 0.8779, + "step": 321890 + }, + { + "epoch": 27.86, + "learning_rate": 3.566991245051865e-06, + "loss": 0.8314, + "step": 321900 + }, + { + "epoch": 27.86, + "learning_rate": 3.565546534138519e-06, + "loss": 0.8478, + "step": 321910 + }, + { + "epoch": 27.86, + "learning_rate": 3.564101823225173e-06, + "loss": 0.9179, + "step": 321920 + }, + { + "epoch": 27.87, + "learning_rate": 3.5626571123118264e-06, + "loss": 0.9216, + "step": 321930 + }, + { + "epoch": 27.87, + "learning_rate": 3.56121240139848e-06, + "loss": 0.9252, + "step": 321940 + }, + { + "epoch": 27.87, + "learning_rate": 3.5597676904851344e-06, + "loss": 0.8892, + "step": 321950 + }, + { + "epoch": 27.87, + "learning_rate": 3.558322979571788e-06, + "loss": 0.9026, + "step": 321960 + }, + { + "epoch": 27.87, + "learning_rate": 3.5568782686584416e-06, + "loss": 0.8692, + "step": 321970 + }, + { + "epoch": 27.87, + "learning_rate": 3.5554335577450954e-06, + "loss": 0.9324, + "step": 321980 + }, + { + "epoch": 27.87, + "learning_rate": 3.553988846831749e-06, + "loss": 0.9567, + "step": 321990 + }, + { + "epoch": 27.87, + "learning_rate": 3.552544135918403e-06, + "loss": 0.8932, + "step": 322000 + }, + { + "epoch": 27.87, + "learning_rate": 3.551099425005057e-06, + "loss": 0.8879, + "step": 322010 + }, + { + "epoch": 27.87, + "learning_rate": 3.5496547140917103e-06, + "loss": 0.8984, + "step": 322020 + }, + { + "epoch": 27.87, + "learning_rate": 3.548210003178364e-06, + "loss": 0.9191, + "step": 322030 + }, + { + "epoch": 27.87, + "learning_rate": 3.5467652922650184e-06, + "loss": 0.9071, + "step": 322040 + }, + { + "epoch": 27.88, + "learning_rate": 3.5453205813516717e-06, + "loss": 0.9052, + "step": 322050 + }, + { + "epoch": 27.88, + "learning_rate": 3.5438758704383256e-06, + "loss": 0.8844, + "step": 322060 + }, + { + "epoch": 27.88, + "learning_rate": 3.5424311595249794e-06, + "loss": 0.9784, + "step": 322070 + }, + { + "epoch": 27.88, + "learning_rate": 3.5409864486116328e-06, + "loss": 0.8853, + "step": 322080 + }, + { + "epoch": 27.88, + "learning_rate": 3.539541737698287e-06, + "loss": 0.8781, + "step": 322090 + }, + { + "epoch": 27.88, + "learning_rate": 3.53809702678494e-06, + "loss": 0.9227, + "step": 322100 + }, + { + "epoch": 27.88, + "learning_rate": 3.5366523158715942e-06, + "loss": 0.9089, + "step": 322110 + }, + { + "epoch": 27.88, + "learning_rate": 3.535207604958248e-06, + "loss": 0.9562, + "step": 322120 + }, + { + "epoch": 27.88, + "learning_rate": 3.5337628940449015e-06, + "loss": 0.9449, + "step": 322130 + }, + { + "epoch": 27.88, + "learning_rate": 3.5323181831315557e-06, + "loss": 0.888, + "step": 322140 + }, + { + "epoch": 27.88, + "learning_rate": 3.5308734722182095e-06, + "loss": 0.7977, + "step": 322150 + }, + { + "epoch": 27.89, + "learning_rate": 3.529428761304863e-06, + "loss": 0.9015, + "step": 322160 + }, + { + "epoch": 27.89, + "learning_rate": 3.5279840503915167e-06, + "loss": 0.941, + "step": 322170 + }, + { + "epoch": 27.89, + "learning_rate": 3.526539339478171e-06, + "loss": 0.9276, + "step": 322180 + }, + { + "epoch": 27.89, + "learning_rate": 3.525094628564824e-06, + "loss": 0.8303, + "step": 322190 + }, + { + "epoch": 27.89, + "learning_rate": 3.523649917651478e-06, + "loss": 0.9056, + "step": 322200 + }, + { + "epoch": 27.89, + "learning_rate": 3.522205206738132e-06, + "loss": 0.8852, + "step": 322210 + }, + { + "epoch": 27.89, + "learning_rate": 3.5207604958247854e-06, + "loss": 0.9112, + "step": 322220 + }, + { + "epoch": 27.89, + "learning_rate": 3.5193157849114396e-06, + "loss": 0.9318, + "step": 322230 + }, + { + "epoch": 27.89, + "learning_rate": 3.5178710739980935e-06, + "loss": 0.8567, + "step": 322240 + }, + { + "epoch": 27.89, + "learning_rate": 3.516426363084747e-06, + "loss": 0.8929, + "step": 322250 + }, + { + "epoch": 27.89, + "learning_rate": 3.5149816521714007e-06, + "loss": 0.8873, + "step": 322260 + }, + { + "epoch": 27.89, + "learning_rate": 3.513536941258055e-06, + "loss": 0.9024, + "step": 322270 + }, + { + "epoch": 27.9, + "learning_rate": 3.512092230344708e-06, + "loss": 0.9133, + "step": 322280 + }, + { + "epoch": 27.9, + "learning_rate": 3.510647519431362e-06, + "loss": 0.8896, + "step": 322290 + }, + { + "epoch": 27.9, + "learning_rate": 3.509202808518016e-06, + "loss": 0.8882, + "step": 322300 + }, + { + "epoch": 27.9, + "learning_rate": 3.5077580976046693e-06, + "loss": 0.8908, + "step": 322310 + }, + { + "epoch": 27.9, + "learning_rate": 3.506313386691323e-06, + "loss": 0.8977, + "step": 322320 + }, + { + "epoch": 27.9, + "learning_rate": 3.5048686757779765e-06, + "loss": 0.9217, + "step": 322330 + }, + { + "epoch": 27.9, + "learning_rate": 3.503423964864631e-06, + "loss": 0.9043, + "step": 322340 + }, + { + "epoch": 27.9, + "learning_rate": 3.5019792539512846e-06, + "loss": 0.9434, + "step": 322350 + }, + { + "epoch": 27.9, + "learning_rate": 3.500534543037938e-06, + "loss": 0.9308, + "step": 322360 + }, + { + "epoch": 27.9, + "learning_rate": 3.499089832124592e-06, + "loss": 0.8953, + "step": 322370 + }, + { + "epoch": 27.9, + "learning_rate": 3.497645121211246e-06, + "loss": 0.9432, + "step": 322380 + }, + { + "epoch": 27.91, + "learning_rate": 3.4962004102978995e-06, + "loss": 0.9017, + "step": 322390 + }, + { + "epoch": 27.91, + "learning_rate": 3.4947556993845533e-06, + "loss": 0.8974, + "step": 322400 + }, + { + "epoch": 27.91, + "learning_rate": 3.493310988471207e-06, + "loss": 0.8728, + "step": 322410 + }, + { + "epoch": 27.91, + "learning_rate": 3.4918662775578605e-06, + "loss": 0.9165, + "step": 322420 + }, + { + "epoch": 27.91, + "learning_rate": 3.4904215666445147e-06, + "loss": 0.9, + "step": 322430 + }, + { + "epoch": 27.91, + "learning_rate": 3.4889768557311686e-06, + "loss": 0.8971, + "step": 322440 + }, + { + "epoch": 27.91, + "learning_rate": 3.487532144817822e-06, + "loss": 0.8653, + "step": 322450 + }, + { + "epoch": 27.91, + "learning_rate": 3.4860874339044758e-06, + "loss": 0.883, + "step": 322460 + }, + { + "epoch": 27.91, + "learning_rate": 3.48464272299113e-06, + "loss": 0.927, + "step": 322470 + }, + { + "epoch": 27.91, + "learning_rate": 3.4831980120777834e-06, + "loss": 0.9068, + "step": 322480 + }, + { + "epoch": 27.91, + "learning_rate": 3.4817533011644372e-06, + "loss": 0.8705, + "step": 322490 + }, + { + "epoch": 27.91, + "learning_rate": 3.480308590251091e-06, + "loss": 0.9055, + "step": 322500 + }, + { + "epoch": 27.92, + "learning_rate": 3.4788638793377444e-06, + "loss": 0.8898, + "step": 322510 + }, + { + "epoch": 27.92, + "learning_rate": 3.4774191684243987e-06, + "loss": 0.9411, + "step": 322520 + }, + { + "epoch": 27.92, + "learning_rate": 3.4759744575110525e-06, + "loss": 0.9031, + "step": 322530 + }, + { + "epoch": 27.92, + "learning_rate": 3.474529746597706e-06, + "loss": 0.9462, + "step": 322540 + }, + { + "epoch": 27.92, + "learning_rate": 3.4730850356843597e-06, + "loss": 0.8851, + "step": 322550 + }, + { + "epoch": 27.92, + "learning_rate": 3.471640324771013e-06, + "loss": 0.9145, + "step": 322560 + }, + { + "epoch": 27.92, + "learning_rate": 3.4701956138576673e-06, + "loss": 0.9102, + "step": 322570 + }, + { + "epoch": 27.92, + "learning_rate": 3.468750902944321e-06, + "loss": 0.8823, + "step": 322580 + }, + { + "epoch": 27.92, + "learning_rate": 3.4673061920309746e-06, + "loss": 0.8557, + "step": 322590 + }, + { + "epoch": 27.92, + "learning_rate": 3.4658614811176284e-06, + "loss": 0.9322, + "step": 322600 + }, + { + "epoch": 27.92, + "learning_rate": 3.4644167702042826e-06, + "loss": 0.8666, + "step": 322610 + }, + { + "epoch": 27.93, + "learning_rate": 3.4629720592909356e-06, + "loss": 0.9021, + "step": 322620 + }, + { + "epoch": 27.93, + "learning_rate": 3.46152734837759e-06, + "loss": 0.8566, + "step": 322630 + }, + { + "epoch": 27.93, + "learning_rate": 3.4600826374642437e-06, + "loss": 0.9275, + "step": 322640 + }, + { + "epoch": 27.93, + "learning_rate": 3.458637926550897e-06, + "loss": 0.9476, + "step": 322650 + }, + { + "epoch": 27.93, + "learning_rate": 3.4571932156375513e-06, + "loss": 0.885, + "step": 322660 + }, + { + "epoch": 27.93, + "learning_rate": 3.455748504724205e-06, + "loss": 0.8814, + "step": 322670 + }, + { + "epoch": 27.93, + "learning_rate": 3.4543037938108585e-06, + "loss": 0.9574, + "step": 322680 + }, + { + "epoch": 27.93, + "learning_rate": 3.4528590828975123e-06, + "loss": 0.9051, + "step": 322690 + }, + { + "epoch": 27.93, + "learning_rate": 3.4514143719841666e-06, + "loss": 0.8417, + "step": 322700 + }, + { + "epoch": 27.93, + "learning_rate": 3.4499696610708195e-06, + "loss": 0.8767, + "step": 322710 + }, + { + "epoch": 27.93, + "learning_rate": 3.4485249501574738e-06, + "loss": 0.8609, + "step": 322720 + }, + { + "epoch": 27.93, + "learning_rate": 3.4470802392441276e-06, + "loss": 0.8908, + "step": 322730 + }, + { + "epoch": 27.94, + "learning_rate": 3.445635528330781e-06, + "loss": 0.8784, + "step": 322740 + }, + { + "epoch": 27.94, + "learning_rate": 3.4441908174174352e-06, + "loss": 0.8823, + "step": 322750 + }, + { + "epoch": 27.94, + "learning_rate": 3.442746106504088e-06, + "loss": 0.8946, + "step": 322760 + }, + { + "epoch": 27.94, + "learning_rate": 3.4413013955907424e-06, + "loss": 0.9018, + "step": 322770 + }, + { + "epoch": 27.94, + "learning_rate": 3.4398566846773963e-06, + "loss": 0.866, + "step": 322780 + }, + { + "epoch": 27.94, + "learning_rate": 3.4384119737640497e-06, + "loss": 0.9418, + "step": 322790 + }, + { + "epoch": 27.94, + "learning_rate": 3.4369672628507035e-06, + "loss": 0.9083, + "step": 322800 + }, + { + "epoch": 27.94, + "learning_rate": 3.4355225519373577e-06, + "loss": 0.897, + "step": 322810 + }, + { + "epoch": 27.94, + "learning_rate": 3.434077841024011e-06, + "loss": 0.9146, + "step": 322820 + }, + { + "epoch": 27.94, + "learning_rate": 3.432633130110665e-06, + "loss": 0.8989, + "step": 322830 + }, + { + "epoch": 27.94, + "learning_rate": 3.431188419197319e-06, + "loss": 0.9404, + "step": 322840 + }, + { + "epoch": 27.94, + "learning_rate": 3.429743708283972e-06, + "loss": 0.8672, + "step": 322850 + }, + { + "epoch": 27.95, + "learning_rate": 3.4282989973706264e-06, + "loss": 0.9059, + "step": 322860 + }, + { + "epoch": 27.95, + "learning_rate": 3.42685428645728e-06, + "loss": 0.9379, + "step": 322870 + }, + { + "epoch": 27.95, + "learning_rate": 3.4254095755439336e-06, + "loss": 0.8757, + "step": 322880 + }, + { + "epoch": 27.95, + "learning_rate": 3.4239648646305874e-06, + "loss": 0.8494, + "step": 322890 + }, + { + "epoch": 27.95, + "learning_rate": 3.4225201537172417e-06, + "loss": 0.817, + "step": 322900 + }, + { + "epoch": 27.95, + "learning_rate": 3.421075442803895e-06, + "loss": 0.8654, + "step": 322910 + }, + { + "epoch": 27.95, + "learning_rate": 3.419630731890549e-06, + "loss": 0.8969, + "step": 322920 + }, + { + "epoch": 27.95, + "learning_rate": 3.418186020977203e-06, + "loss": 0.8605, + "step": 322930 + }, + { + "epoch": 27.95, + "learning_rate": 3.416741310063856e-06, + "loss": 0.8765, + "step": 322940 + }, + { + "epoch": 27.95, + "learning_rate": 3.4152965991505103e-06, + "loss": 0.9396, + "step": 322950 + }, + { + "epoch": 27.95, + "learning_rate": 3.413851888237164e-06, + "loss": 0.8436, + "step": 322960 + }, + { + "epoch": 27.96, + "learning_rate": 3.4124071773238175e-06, + "loss": 0.8649, + "step": 322970 + }, + { + "epoch": 27.96, + "learning_rate": 3.4109624664104714e-06, + "loss": 0.8676, + "step": 322980 + }, + { + "epoch": 27.96, + "learning_rate": 3.4095177554971248e-06, + "loss": 0.9244, + "step": 322990 + }, + { + "epoch": 27.96, + "learning_rate": 3.408073044583779e-06, + "loss": 0.9285, + "step": 323000 + }, + { + "epoch": 27.96, + "learning_rate": 3.406628333670433e-06, + "loss": 0.8684, + "step": 323010 + }, + { + "epoch": 27.96, + "learning_rate": 3.4051836227570862e-06, + "loss": 0.8553, + "step": 323020 + }, + { + "epoch": 27.96, + "learning_rate": 3.40373891184374e-06, + "loss": 0.9081, + "step": 323030 + }, + { + "epoch": 27.96, + "learning_rate": 3.4022942009303943e-06, + "loss": 0.9622, + "step": 323040 + }, + { + "epoch": 27.96, + "learning_rate": 3.4008494900170477e-06, + "loss": 0.8495, + "step": 323050 + }, + { + "epoch": 27.96, + "learning_rate": 3.3994047791037015e-06, + "loss": 0.9234, + "step": 323060 + }, + { + "epoch": 27.96, + "learning_rate": 3.3979600681903553e-06, + "loss": 0.9031, + "step": 323070 + }, + { + "epoch": 27.96, + "learning_rate": 3.3965153572770087e-06, + "loss": 0.9208, + "step": 323080 + }, + { + "epoch": 27.97, + "learning_rate": 3.395070646363663e-06, + "loss": 0.9251, + "step": 323090 + }, + { + "epoch": 27.97, + "learning_rate": 3.3936259354503168e-06, + "loss": 0.8612, + "step": 323100 + }, + { + "epoch": 27.97, + "learning_rate": 3.39218122453697e-06, + "loss": 0.8762, + "step": 323110 + }, + { + "epoch": 27.97, + "learning_rate": 3.390736513623624e-06, + "loss": 0.8865, + "step": 323120 + }, + { + "epoch": 27.97, + "learning_rate": 3.3892918027102782e-06, + "loss": 0.9013, + "step": 323130 + }, + { + "epoch": 27.97, + "learning_rate": 3.3878470917969316e-06, + "loss": 0.8551, + "step": 323140 + }, + { + "epoch": 27.97, + "learning_rate": 3.3864023808835854e-06, + "loss": 0.8511, + "step": 323150 + }, + { + "epoch": 27.97, + "learning_rate": 3.3849576699702392e-06, + "loss": 0.9298, + "step": 323160 + }, + { + "epoch": 27.97, + "learning_rate": 3.3835129590568926e-06, + "loss": 0.8867, + "step": 323170 + }, + { + "epoch": 27.97, + "learning_rate": 3.382068248143547e-06, + "loss": 0.9186, + "step": 323180 + }, + { + "epoch": 27.97, + "learning_rate": 3.3806235372302007e-06, + "loss": 0.9252, + "step": 323190 + }, + { + "epoch": 27.98, + "learning_rate": 3.379178826316854e-06, + "loss": 0.9134, + "step": 323200 + }, + { + "epoch": 27.98, + "learning_rate": 3.377734115403508e-06, + "loss": 0.9376, + "step": 323210 + }, + { + "epoch": 27.98, + "learning_rate": 3.3762894044901613e-06, + "loss": 0.9089, + "step": 323220 + }, + { + "epoch": 27.98, + "learning_rate": 3.3748446935768156e-06, + "loss": 0.9525, + "step": 323230 + }, + { + "epoch": 27.98, + "learning_rate": 3.3733999826634694e-06, + "loss": 0.8724, + "step": 323240 + }, + { + "epoch": 27.98, + "learning_rate": 3.3719552717501228e-06, + "loss": 0.8813, + "step": 323250 + }, + { + "epoch": 27.98, + "learning_rate": 3.3705105608367766e-06, + "loss": 0.8819, + "step": 323260 + }, + { + "epoch": 27.98, + "learning_rate": 3.369065849923431e-06, + "loss": 0.9162, + "step": 323270 + }, + { + "epoch": 27.98, + "learning_rate": 3.367621139010084e-06, + "loss": 0.883, + "step": 323280 + }, + { + "epoch": 27.98, + "learning_rate": 3.366176428096738e-06, + "loss": 0.8653, + "step": 323290 + }, + { + "epoch": 27.98, + "learning_rate": 3.364731717183392e-06, + "loss": 0.9106, + "step": 323300 + }, + { + "epoch": 27.98, + "learning_rate": 3.3632870062700453e-06, + "loss": 0.9246, + "step": 323310 + }, + { + "epoch": 27.99, + "learning_rate": 3.361842295356699e-06, + "loss": 0.8998, + "step": 323320 + }, + { + "epoch": 27.99, + "learning_rate": 3.3603975844433533e-06, + "loss": 0.9319, + "step": 323330 + }, + { + "epoch": 27.99, + "learning_rate": 3.3589528735300067e-06, + "loss": 0.9121, + "step": 323340 + }, + { + "epoch": 27.99, + "learning_rate": 3.3575081626166605e-06, + "loss": 0.9392, + "step": 323350 + }, + { + "epoch": 27.99, + "learning_rate": 3.3560634517033148e-06, + "loss": 0.8809, + "step": 323360 + }, + { + "epoch": 27.99, + "learning_rate": 3.3546187407899677e-06, + "loss": 0.8842, + "step": 323370 + }, + { + "epoch": 27.99, + "learning_rate": 3.353174029876622e-06, + "loss": 0.9213, + "step": 323380 + }, + { + "epoch": 27.99, + "learning_rate": 3.351729318963276e-06, + "loss": 0.905, + "step": 323390 + }, + { + "epoch": 27.99, + "learning_rate": 3.350284608049929e-06, + "loss": 0.879, + "step": 323400 + }, + { + "epoch": 27.99, + "learning_rate": 3.348839897136583e-06, + "loss": 0.9033, + "step": 323410 + }, + { + "epoch": 27.99, + "learning_rate": 3.3473951862232373e-06, + "loss": 0.8725, + "step": 323420 + }, + { + "epoch": 28.0, + "learning_rate": 3.3459504753098907e-06, + "loss": 0.9162, + "step": 323430 + }, + { + "epoch": 28.0, + "learning_rate": 3.3445057643965445e-06, + "loss": 0.8211, + "step": 323440 + }, + { + "epoch": 28.0, + "learning_rate": 3.343061053483198e-06, + "loss": 0.882, + "step": 323450 + }, + { + "epoch": 28.0, + "learning_rate": 3.3416163425698517e-06, + "loss": 0.9546, + "step": 323460 + }, + { + "epoch": 28.0, + "learning_rate": 3.340171631656506e-06, + "loss": 0.9083, + "step": 323470 + }, + { + "epoch": 28.0, + "learning_rate": 3.3387269207431593e-06, + "loss": 0.8682, + "step": 323480 + }, + { + "epoch": 28.0, + "eval_cer": 0.9930416233608066, + "eval_em": 0, + "eval_f1": 0.1225159414300325, + "eval_loss": 0.9318199753761292, + "eval_runtime": 891.491, + "eval_samples_per_second": 5.76, + "eval_steps_per_second": 0.72, + "eval_wer": 0.9659366198050354, + "step": 323486 + }, + { + "epoch": 28.0, + "learning_rate": 3.337282209829813e-06, + "loss": 0.9138, + "step": 323490 + }, + { + "epoch": 28.0, + "learning_rate": 3.335837498916467e-06, + "loss": 0.8655, + "step": 323500 + }, + { + "epoch": 28.0, + "learning_rate": 3.3343927880031204e-06, + "loss": 0.8689, + "step": 323510 + }, + { + "epoch": 28.0, + "learning_rate": 3.3329480770897746e-06, + "loss": 0.9326, + "step": 323520 + }, + { + "epoch": 28.0, + "learning_rate": 3.3315033661764284e-06, + "loss": 0.9356, + "step": 323530 + }, + { + "epoch": 28.0, + "learning_rate": 3.330058655263082e-06, + "loss": 0.8853, + "step": 323540 + }, + { + "epoch": 28.01, + "learning_rate": 3.3286139443497356e-06, + "loss": 0.9009, + "step": 323550 + }, + { + "epoch": 28.01, + "learning_rate": 3.32716923343639e-06, + "loss": 0.8761, + "step": 323560 + }, + { + "epoch": 28.01, + "learning_rate": 3.3257245225230433e-06, + "loss": 0.914, + "step": 323570 + }, + { + "epoch": 28.01, + "learning_rate": 3.324279811609697e-06, + "loss": 0.9225, + "step": 323580 + }, + { + "epoch": 28.01, + "learning_rate": 3.322835100696351e-06, + "loss": 0.9512, + "step": 323590 + }, + { + "epoch": 28.01, + "learning_rate": 3.3213903897830043e-06, + "loss": 0.9363, + "step": 323600 + }, + { + "epoch": 28.01, + "learning_rate": 3.3199456788696585e-06, + "loss": 0.9237, + "step": 323610 + }, + { + "epoch": 28.01, + "learning_rate": 3.3185009679563124e-06, + "loss": 0.8804, + "step": 323620 + }, + { + "epoch": 28.01, + "learning_rate": 3.3170562570429658e-06, + "loss": 0.9063, + "step": 323630 + }, + { + "epoch": 28.01, + "learning_rate": 3.3156115461296196e-06, + "loss": 0.9192, + "step": 323640 + }, + { + "epoch": 28.01, + "learning_rate": 3.314166835216274e-06, + "loss": 0.9736, + "step": 323650 + }, + { + "epoch": 28.02, + "learning_rate": 3.312722124302927e-06, + "loss": 0.933, + "step": 323660 + }, + { + "epoch": 28.02, + "learning_rate": 3.311277413389581e-06, + "loss": 0.8482, + "step": 323670 + }, + { + "epoch": 28.02, + "learning_rate": 3.3098327024762344e-06, + "loss": 0.9482, + "step": 323680 + }, + { + "epoch": 28.02, + "learning_rate": 3.3083879915628882e-06, + "loss": 0.9335, + "step": 323690 + }, + { + "epoch": 28.02, + "learning_rate": 3.3069432806495425e-06, + "loss": 0.9067, + "step": 323700 + }, + { + "epoch": 28.02, + "learning_rate": 3.3054985697361955e-06, + "loss": 0.902, + "step": 323710 + }, + { + "epoch": 28.02, + "learning_rate": 3.3040538588228497e-06, + "loss": 0.9071, + "step": 323720 + }, + { + "epoch": 28.02, + "learning_rate": 3.3026091479095035e-06, + "loss": 0.8615, + "step": 323730 + }, + { + "epoch": 28.02, + "learning_rate": 3.301164436996157e-06, + "loss": 0.9528, + "step": 323740 + }, + { + "epoch": 28.02, + "learning_rate": 3.299719726082811e-06, + "loss": 0.8871, + "step": 323750 + }, + { + "epoch": 28.02, + "learning_rate": 3.298275015169465e-06, + "loss": 0.9039, + "step": 323760 + }, + { + "epoch": 28.02, + "learning_rate": 3.2968303042561184e-06, + "loss": 0.9412, + "step": 323770 + }, + { + "epoch": 28.03, + "learning_rate": 3.295385593342772e-06, + "loss": 0.9104, + "step": 323780 + }, + { + "epoch": 28.03, + "learning_rate": 3.2939408824294264e-06, + "loss": 0.8811, + "step": 323790 + }, + { + "epoch": 28.03, + "learning_rate": 3.2924961715160794e-06, + "loss": 0.8821, + "step": 323800 + }, + { + "epoch": 28.03, + "learning_rate": 3.2910514606027336e-06, + "loss": 0.8929, + "step": 323810 + }, + { + "epoch": 28.03, + "learning_rate": 3.2896067496893875e-06, + "loss": 0.9187, + "step": 323820 + }, + { + "epoch": 28.03, + "learning_rate": 3.288162038776041e-06, + "loss": 0.9399, + "step": 323830 + }, + { + "epoch": 28.03, + "learning_rate": 3.286717327862695e-06, + "loss": 0.9543, + "step": 323840 + }, + { + "epoch": 28.03, + "learning_rate": 3.285272616949349e-06, + "loss": 0.9054, + "step": 323850 + }, + { + "epoch": 28.03, + "learning_rate": 3.2838279060360023e-06, + "loss": 0.8936, + "step": 323860 + }, + { + "epoch": 28.03, + "learning_rate": 3.282383195122656e-06, + "loss": 0.8985, + "step": 323870 + }, + { + "epoch": 28.03, + "learning_rate": 3.2809384842093104e-06, + "loss": 0.981, + "step": 323880 + }, + { + "epoch": 28.03, + "learning_rate": 3.2794937732959633e-06, + "loss": 0.8957, + "step": 323890 + }, + { + "epoch": 28.04, + "learning_rate": 3.2780490623826176e-06, + "loss": 0.9024, + "step": 323900 + }, + { + "epoch": 28.04, + "learning_rate": 3.276604351469271e-06, + "loss": 0.844, + "step": 323910 + }, + { + "epoch": 28.04, + "learning_rate": 3.275159640555925e-06, + "loss": 0.9173, + "step": 323920 + }, + { + "epoch": 28.04, + "learning_rate": 3.273714929642579e-06, + "loss": 0.9024, + "step": 323930 + }, + { + "epoch": 28.04, + "learning_rate": 3.272270218729232e-06, + "loss": 0.8893, + "step": 323940 + }, + { + "epoch": 28.04, + "learning_rate": 3.2708255078158862e-06, + "loss": 0.8446, + "step": 323950 + }, + { + "epoch": 28.04, + "learning_rate": 3.26938079690254e-06, + "loss": 0.9245, + "step": 323960 + }, + { + "epoch": 28.04, + "learning_rate": 3.2679360859891935e-06, + "loss": 0.8756, + "step": 323970 + }, + { + "epoch": 28.04, + "learning_rate": 3.2664913750758473e-06, + "loss": 0.9361, + "step": 323980 + }, + { + "epoch": 28.04, + "learning_rate": 3.2650466641625015e-06, + "loss": 0.9134, + "step": 323990 + }, + { + "epoch": 28.04, + "learning_rate": 3.263601953249155e-06, + "loss": 0.9171, + "step": 324000 + }, + { + "epoch": 28.05, + "learning_rate": 3.2621572423358087e-06, + "loss": 0.8833, + "step": 324010 + }, + { + "epoch": 28.05, + "learning_rate": 3.260712531422463e-06, + "loss": 0.862, + "step": 324020 + }, + { + "epoch": 28.05, + "learning_rate": 3.259267820509116e-06, + "loss": 0.9031, + "step": 324030 + }, + { + "epoch": 28.05, + "learning_rate": 3.25782310959577e-06, + "loss": 0.8826, + "step": 324040 + }, + { + "epoch": 28.05, + "learning_rate": 3.256378398682424e-06, + "loss": 0.8998, + "step": 324050 + }, + { + "epoch": 28.05, + "learning_rate": 3.2549336877690774e-06, + "loss": 0.8757, + "step": 324060 + }, + { + "epoch": 28.05, + "learning_rate": 3.2534889768557312e-06, + "loss": 0.9518, + "step": 324070 + }, + { + "epoch": 28.05, + "learning_rate": 3.2520442659423855e-06, + "loss": 0.8939, + "step": 324080 + }, + { + "epoch": 28.05, + "learning_rate": 3.250599555029039e-06, + "loss": 0.9148, + "step": 324090 + }, + { + "epoch": 28.05, + "learning_rate": 3.2491548441156927e-06, + "loss": 0.8737, + "step": 324100 + }, + { + "epoch": 28.05, + "learning_rate": 3.247710133202347e-06, + "loss": 0.8249, + "step": 324110 + }, + { + "epoch": 28.05, + "learning_rate": 3.246265422289e-06, + "loss": 0.9016, + "step": 324120 + }, + { + "epoch": 28.06, + "learning_rate": 3.244820711375654e-06, + "loss": 0.8941, + "step": 324130 + }, + { + "epoch": 28.06, + "learning_rate": 3.2433760004623075e-06, + "loss": 0.9207, + "step": 324140 + }, + { + "epoch": 28.06, + "learning_rate": 3.2419312895489613e-06, + "loss": 0.9013, + "step": 324150 + }, + { + "epoch": 28.06, + "learning_rate": 3.240486578635615e-06, + "loss": 0.9255, + "step": 324160 + }, + { + "epoch": 28.06, + "learning_rate": 3.2390418677222686e-06, + "loss": 0.8965, + "step": 324170 + }, + { + "epoch": 28.06, + "learning_rate": 3.237597156808923e-06, + "loss": 0.8673, + "step": 324180 + }, + { + "epoch": 28.06, + "learning_rate": 3.2361524458955766e-06, + "loss": 0.9086, + "step": 324190 + }, + { + "epoch": 28.06, + "learning_rate": 3.23470773498223e-06, + "loss": 0.8604, + "step": 324200 + }, + { + "epoch": 28.06, + "learning_rate": 3.233263024068884e-06, + "loss": 0.8787, + "step": 324210 + }, + { + "epoch": 28.06, + "learning_rate": 3.231818313155538e-06, + "loss": 0.91, + "step": 324220 + }, + { + "epoch": 28.06, + "learning_rate": 3.2303736022421915e-06, + "loss": 0.9368, + "step": 324230 + }, + { + "epoch": 28.07, + "learning_rate": 3.2289288913288453e-06, + "loss": 0.9143, + "step": 324240 + }, + { + "epoch": 28.07, + "learning_rate": 3.227484180415499e-06, + "loss": 0.8812, + "step": 324250 + }, + { + "epoch": 28.07, + "learning_rate": 3.2260394695021525e-06, + "loss": 0.8667, + "step": 324260 + }, + { + "epoch": 28.07, + "learning_rate": 3.2245947585888067e-06, + "loss": 0.8519, + "step": 324270 + }, + { + "epoch": 28.07, + "learning_rate": 3.2231500476754606e-06, + "loss": 0.8998, + "step": 324280 + }, + { + "epoch": 28.07, + "learning_rate": 3.221705336762114e-06, + "loss": 0.9188, + "step": 324290 + }, + { + "epoch": 28.07, + "learning_rate": 3.2202606258487678e-06, + "loss": 0.9493, + "step": 324300 + }, + { + "epoch": 28.07, + "learning_rate": 3.218815914935422e-06, + "loss": 0.9473, + "step": 324310 + }, + { + "epoch": 28.07, + "learning_rate": 3.217371204022075e-06, + "loss": 0.8748, + "step": 324320 + }, + { + "epoch": 28.07, + "learning_rate": 3.2159264931087292e-06, + "loss": 0.8816, + "step": 324330 + }, + { + "epoch": 28.07, + "learning_rate": 3.2144817821953826e-06, + "loss": 0.8681, + "step": 324340 + }, + { + "epoch": 28.07, + "learning_rate": 3.2130370712820364e-06, + "loss": 0.8754, + "step": 324350 + }, + { + "epoch": 28.08, + "learning_rate": 3.2115923603686907e-06, + "loss": 0.8992, + "step": 324360 + }, + { + "epoch": 28.08, + "learning_rate": 3.2101476494553437e-06, + "loss": 0.9359, + "step": 324370 + }, + { + "epoch": 28.08, + "learning_rate": 3.208702938541998e-06, + "loss": 0.8954, + "step": 324380 + }, + { + "epoch": 28.08, + "learning_rate": 3.2072582276286517e-06, + "loss": 0.8921, + "step": 324390 + }, + { + "epoch": 28.08, + "learning_rate": 3.205813516715305e-06, + "loss": 0.9182, + "step": 324400 + }, + { + "epoch": 28.08, + "learning_rate": 3.204368805801959e-06, + "loss": 0.8645, + "step": 324410 + }, + { + "epoch": 28.08, + "learning_rate": 3.202924094888613e-06, + "loss": 0.9295, + "step": 324420 + }, + { + "epoch": 28.08, + "learning_rate": 3.2014793839752666e-06, + "loss": 0.9362, + "step": 324430 + }, + { + "epoch": 28.08, + "learning_rate": 3.2000346730619204e-06, + "loss": 0.9531, + "step": 324440 + }, + { + "epoch": 28.08, + "learning_rate": 3.1985899621485746e-06, + "loss": 0.91, + "step": 324450 + }, + { + "epoch": 28.08, + "learning_rate": 3.1971452512352276e-06, + "loss": 0.881, + "step": 324460 + }, + { + "epoch": 28.09, + "learning_rate": 3.195700540321882e-06, + "loss": 0.9342, + "step": 324470 + }, + { + "epoch": 28.09, + "learning_rate": 3.1942558294085357e-06, + "loss": 0.8549, + "step": 324480 + }, + { + "epoch": 28.09, + "learning_rate": 3.192811118495189e-06, + "loss": 0.9567, + "step": 324490 + }, + { + "epoch": 28.09, + "learning_rate": 3.191366407581843e-06, + "loss": 0.9213, + "step": 324500 + }, + { + "epoch": 28.09, + "learning_rate": 3.189921696668497e-06, + "loss": 0.8941, + "step": 324510 + }, + { + "epoch": 28.09, + "learning_rate": 3.1884769857551505e-06, + "loss": 0.9062, + "step": 324520 + }, + { + "epoch": 28.09, + "learning_rate": 3.1870322748418043e-06, + "loss": 0.9031, + "step": 324530 + }, + { + "epoch": 28.09, + "learning_rate": 3.1855875639284586e-06, + "loss": 0.9188, + "step": 324540 + }, + { + "epoch": 28.09, + "learning_rate": 3.1841428530151115e-06, + "loss": 0.9265, + "step": 324550 + }, + { + "epoch": 28.09, + "learning_rate": 3.182698142101766e-06, + "loss": 0.8372, + "step": 324560 + }, + { + "epoch": 28.09, + "learning_rate": 3.181253431188419e-06, + "loss": 0.9299, + "step": 324570 + }, + { + "epoch": 28.09, + "learning_rate": 3.179808720275073e-06, + "loss": 0.8888, + "step": 324580 + }, + { + "epoch": 28.1, + "learning_rate": 3.178364009361727e-06, + "loss": 0.8934, + "step": 324590 + }, + { + "epoch": 28.1, + "learning_rate": 3.1769192984483802e-06, + "loss": 0.8572, + "step": 324600 + }, + { + "epoch": 28.1, + "learning_rate": 3.1754745875350345e-06, + "loss": 0.8588, + "step": 324610 + }, + { + "epoch": 28.1, + "learning_rate": 3.1740298766216883e-06, + "loss": 0.9421, + "step": 324620 + }, + { + "epoch": 28.1, + "learning_rate": 3.1725851657083417e-06, + "loss": 0.8475, + "step": 324630 + }, + { + "epoch": 28.1, + "learning_rate": 3.1711404547949955e-06, + "loss": 0.9294, + "step": 324640 + }, + { + "epoch": 28.1, + "learning_rate": 3.1696957438816497e-06, + "loss": 0.8408, + "step": 324650 + }, + { + "epoch": 28.1, + "learning_rate": 3.168251032968303e-06, + "loss": 0.874, + "step": 324660 + }, + { + "epoch": 28.1, + "learning_rate": 3.166806322054957e-06, + "loss": 0.8657, + "step": 324670 + }, + { + "epoch": 28.1, + "learning_rate": 3.1653616111416108e-06, + "loss": 0.9178, + "step": 324680 + }, + { + "epoch": 28.1, + "learning_rate": 3.163916900228264e-06, + "loss": 0.9324, + "step": 324690 + }, + { + "epoch": 28.11, + "learning_rate": 3.1624721893149184e-06, + "loss": 0.9125, + "step": 324700 + }, + { + "epoch": 28.11, + "learning_rate": 3.1610274784015722e-06, + "loss": 0.9562, + "step": 324710 + }, + { + "epoch": 28.11, + "learning_rate": 3.1595827674882256e-06, + "loss": 0.897, + "step": 324720 + }, + { + "epoch": 28.11, + "learning_rate": 3.1581380565748794e-06, + "loss": 0.8743, + "step": 324730 + }, + { + "epoch": 28.11, + "learning_rate": 3.1566933456615337e-06, + "loss": 0.8959, + "step": 324740 + }, + { + "epoch": 28.11, + "learning_rate": 3.155248634748187e-06, + "loss": 0.8793, + "step": 324750 + }, + { + "epoch": 28.11, + "learning_rate": 3.153803923834841e-06, + "loss": 0.855, + "step": 324760 + }, + { + "epoch": 28.11, + "learning_rate": 3.1523592129214947e-06, + "loss": 0.9026, + "step": 324770 + }, + { + "epoch": 28.11, + "learning_rate": 3.150914502008148e-06, + "loss": 0.8952, + "step": 324780 + }, + { + "epoch": 28.11, + "learning_rate": 3.1494697910948023e-06, + "loss": 0.9634, + "step": 324790 + }, + { + "epoch": 28.11, + "learning_rate": 3.1480250801814553e-06, + "loss": 0.9622, + "step": 324800 + }, + { + "epoch": 28.11, + "learning_rate": 3.1465803692681096e-06, + "loss": 0.8861, + "step": 324810 + }, + { + "epoch": 28.12, + "learning_rate": 3.1451356583547634e-06, + "loss": 0.9082, + "step": 324820 + }, + { + "epoch": 28.12, + "learning_rate": 3.1436909474414168e-06, + "loss": 0.9139, + "step": 324830 + }, + { + "epoch": 28.12, + "learning_rate": 3.142246236528071e-06, + "loss": 0.9436, + "step": 324840 + }, + { + "epoch": 28.12, + "learning_rate": 3.140801525614725e-06, + "loss": 0.8699, + "step": 324850 + }, + { + "epoch": 28.12, + "learning_rate": 3.1393568147013782e-06, + "loss": 0.9143, + "step": 324860 + }, + { + "epoch": 28.12, + "learning_rate": 3.137912103788032e-06, + "loss": 0.8805, + "step": 324870 + }, + { + "epoch": 28.12, + "learning_rate": 3.1364673928746863e-06, + "loss": 0.8648, + "step": 324880 + }, + { + "epoch": 28.12, + "learning_rate": 3.1350226819613393e-06, + "loss": 0.9305, + "step": 324890 + }, + { + "epoch": 28.12, + "learning_rate": 3.1335779710479935e-06, + "loss": 0.9397, + "step": 324900 + }, + { + "epoch": 28.12, + "learning_rate": 3.1321332601346473e-06, + "loss": 0.8835, + "step": 324910 + }, + { + "epoch": 28.12, + "learning_rate": 3.1306885492213007e-06, + "loss": 0.8599, + "step": 324920 + }, + { + "epoch": 28.12, + "learning_rate": 3.129243838307955e-06, + "loss": 0.913, + "step": 324930 + }, + { + "epoch": 28.13, + "learning_rate": 3.1277991273946088e-06, + "loss": 0.8877, + "step": 324940 + }, + { + "epoch": 28.13, + "learning_rate": 3.126354416481262e-06, + "loss": 0.883, + "step": 324950 + }, + { + "epoch": 28.13, + "learning_rate": 3.124909705567916e-06, + "loss": 0.9831, + "step": 324960 + }, + { + "epoch": 28.13, + "learning_rate": 3.12346499465457e-06, + "loss": 0.9188, + "step": 324970 + }, + { + "epoch": 28.13, + "learning_rate": 3.122020283741223e-06, + "loss": 0.8692, + "step": 324980 + }, + { + "epoch": 28.13, + "learning_rate": 3.1205755728278774e-06, + "loss": 0.9219, + "step": 324990 + }, + { + "epoch": 28.13, + "learning_rate": 3.119130861914531e-06, + "loss": 0.8446, + "step": 325000 + }, + { + "epoch": 28.13, + "learning_rate": 3.1176861510011847e-06, + "loss": 0.9423, + "step": 325010 + }, + { + "epoch": 28.13, + "learning_rate": 3.116241440087839e-06, + "loss": 0.8738, + "step": 325020 + }, + { + "epoch": 28.13, + "learning_rate": 3.1147967291744923e-06, + "loss": 0.8749, + "step": 325030 + }, + { + "epoch": 28.13, + "learning_rate": 3.113352018261146e-06, + "loss": 0.9478, + "step": 325040 + }, + { + "epoch": 28.14, + "learning_rate": 3.1119073073478e-06, + "loss": 0.8433, + "step": 325050 + }, + { + "epoch": 28.14, + "learning_rate": 3.1104625964344537e-06, + "loss": 0.9112, + "step": 325060 + }, + { + "epoch": 28.14, + "learning_rate": 3.109017885521107e-06, + "loss": 0.946, + "step": 325070 + }, + { + "epoch": 28.14, + "learning_rate": 3.107573174607761e-06, + "loss": 0.9242, + "step": 325080 + }, + { + "epoch": 28.14, + "learning_rate": 3.1061284636944148e-06, + "loss": 0.886, + "step": 325090 + }, + { + "epoch": 28.14, + "learning_rate": 3.1046837527810686e-06, + "loss": 0.8344, + "step": 325100 + }, + { + "epoch": 28.14, + "learning_rate": 3.1032390418677224e-06, + "loss": 0.9249, + "step": 325110 + }, + { + "epoch": 28.14, + "learning_rate": 3.1017943309543762e-06, + "loss": 0.875, + "step": 325120 + }, + { + "epoch": 28.14, + "learning_rate": 3.10034962004103e-06, + "loss": 0.8505, + "step": 325130 + }, + { + "epoch": 28.14, + "learning_rate": 3.0989049091276834e-06, + "loss": 0.9203, + "step": 325140 + }, + { + "epoch": 28.14, + "learning_rate": 3.0974601982143377e-06, + "loss": 0.9152, + "step": 325150 + }, + { + "epoch": 28.14, + "learning_rate": 3.096015487300991e-06, + "loss": 0.9409, + "step": 325160 + }, + { + "epoch": 28.15, + "learning_rate": 3.094570776387645e-06, + "loss": 0.9059, + "step": 325170 + }, + { + "epoch": 28.15, + "learning_rate": 3.0931260654742987e-06, + "loss": 0.865, + "step": 325180 + }, + { + "epoch": 28.15, + "learning_rate": 3.0916813545609525e-06, + "loss": 0.9363, + "step": 325190 + }, + { + "epoch": 28.15, + "learning_rate": 3.0902366436476064e-06, + "loss": 0.8435, + "step": 325200 + }, + { + "epoch": 28.15, + "learning_rate": 3.0887919327342598e-06, + "loss": 0.9302, + "step": 325210 + }, + { + "epoch": 28.15, + "learning_rate": 3.087347221820914e-06, + "loss": 0.9051, + "step": 325220 + }, + { + "epoch": 28.15, + "learning_rate": 3.0859025109075674e-06, + "loss": 0.8957, + "step": 325230 + }, + { + "epoch": 28.15, + "learning_rate": 3.084457799994221e-06, + "loss": 0.8937, + "step": 325240 + }, + { + "epoch": 28.15, + "learning_rate": 3.083013089080875e-06, + "loss": 0.8589, + "step": 325250 + }, + { + "epoch": 28.15, + "learning_rate": 3.081568378167529e-06, + "loss": 0.8324, + "step": 325260 + }, + { + "epoch": 28.15, + "learning_rate": 3.0801236672541827e-06, + "loss": 0.884, + "step": 325270 + }, + { + "epoch": 28.16, + "learning_rate": 3.0786789563408365e-06, + "loss": 0.8777, + "step": 325280 + }, + { + "epoch": 28.16, + "learning_rate": 3.0772342454274903e-06, + "loss": 0.8953, + "step": 325290 + }, + { + "epoch": 28.16, + "learning_rate": 3.0757895345141437e-06, + "loss": 0.9027, + "step": 325300 + }, + { + "epoch": 28.16, + "learning_rate": 3.0743448236007975e-06, + "loss": 0.8541, + "step": 325310 + }, + { + "epoch": 28.16, + "learning_rate": 3.0729001126874513e-06, + "loss": 0.919, + "step": 325320 + }, + { + "epoch": 28.16, + "learning_rate": 3.071455401774105e-06, + "loss": 0.8967, + "step": 325330 + }, + { + "epoch": 28.16, + "learning_rate": 3.070010690860759e-06, + "loss": 0.9138, + "step": 325340 + }, + { + "epoch": 28.16, + "learning_rate": 3.068565979947413e-06, + "loss": 0.9109, + "step": 325350 + }, + { + "epoch": 28.16, + "learning_rate": 3.0671212690340666e-06, + "loss": 0.8816, + "step": 325360 + }, + { + "epoch": 28.16, + "learning_rate": 3.06567655812072e-06, + "loss": 0.9451, + "step": 325370 + }, + { + "epoch": 28.16, + "learning_rate": 3.0642318472073742e-06, + "loss": 0.8763, + "step": 325380 + }, + { + "epoch": 28.16, + "learning_rate": 3.0627871362940276e-06, + "loss": 0.8931, + "step": 325390 + }, + { + "epoch": 28.17, + "learning_rate": 3.0613424253806815e-06, + "loss": 0.8879, + "step": 325400 + }, + { + "epoch": 28.17, + "learning_rate": 3.0598977144673353e-06, + "loss": 0.8931, + "step": 325410 + }, + { + "epoch": 28.17, + "learning_rate": 3.058453003553989e-06, + "loss": 0.9245, + "step": 325420 + }, + { + "epoch": 28.17, + "learning_rate": 3.057008292640643e-06, + "loss": 0.905, + "step": 325430 + }, + { + "epoch": 28.17, + "learning_rate": 3.0555635817272963e-06, + "loss": 0.9094, + "step": 325440 + }, + { + "epoch": 28.17, + "learning_rate": 3.0541188708139506e-06, + "loss": 0.8396, + "step": 325450 + }, + { + "epoch": 28.17, + "learning_rate": 3.052674159900604e-06, + "loss": 0.8835, + "step": 325460 + }, + { + "epoch": 28.17, + "learning_rate": 3.0512294489872578e-06, + "loss": 0.9059, + "step": 325470 + }, + { + "epoch": 28.17, + "learning_rate": 3.0497847380739116e-06, + "loss": 0.8281, + "step": 325480 + }, + { + "epoch": 28.17, + "learning_rate": 3.0483400271605654e-06, + "loss": 0.8259, + "step": 325490 + }, + { + "epoch": 28.17, + "learning_rate": 3.046895316247219e-06, + "loss": 0.9127, + "step": 325500 + }, + { + "epoch": 28.18, + "learning_rate": 3.045450605333873e-06, + "loss": 0.9145, + "step": 325510 + }, + { + "epoch": 28.18, + "learning_rate": 3.044005894420527e-06, + "loss": 0.9018, + "step": 325520 + }, + { + "epoch": 28.18, + "learning_rate": 3.0425611835071803e-06, + "loss": 0.9073, + "step": 325530 + }, + { + "epoch": 28.18, + "learning_rate": 3.041116472593834e-06, + "loss": 0.9843, + "step": 325540 + }, + { + "epoch": 28.18, + "learning_rate": 3.039671761680488e-06, + "loss": 0.8972, + "step": 325550 + }, + { + "epoch": 28.18, + "learning_rate": 3.0382270507671417e-06, + "loss": 0.8805, + "step": 325560 + }, + { + "epoch": 28.18, + "learning_rate": 3.036782339853795e-06, + "loss": 0.8229, + "step": 325570 + }, + { + "epoch": 28.18, + "learning_rate": 3.0353376289404493e-06, + "loss": 0.8595, + "step": 325580 + }, + { + "epoch": 28.18, + "learning_rate": 3.0338929180271027e-06, + "loss": 0.9544, + "step": 325590 + }, + { + "epoch": 28.18, + "learning_rate": 3.0324482071137566e-06, + "loss": 0.897, + "step": 325600 + }, + { + "epoch": 28.18, + "learning_rate": 3.031003496200411e-06, + "loss": 0.8905, + "step": 325610 + }, + { + "epoch": 28.18, + "learning_rate": 3.029558785287064e-06, + "loss": 0.9156, + "step": 325620 + }, + { + "epoch": 28.19, + "learning_rate": 3.028114074373718e-06, + "loss": 0.9207, + "step": 325630 + }, + { + "epoch": 28.19, + "learning_rate": 3.0266693634603714e-06, + "loss": 0.9008, + "step": 325640 + }, + { + "epoch": 28.19, + "learning_rate": 3.0252246525470257e-06, + "loss": 0.8778, + "step": 325650 + }, + { + "epoch": 28.19, + "learning_rate": 3.023779941633679e-06, + "loss": 0.9134, + "step": 325660 + }, + { + "epoch": 28.19, + "learning_rate": 3.022335230720333e-06, + "loss": 0.8543, + "step": 325670 + }, + { + "epoch": 28.19, + "learning_rate": 3.0208905198069867e-06, + "loss": 0.9118, + "step": 325680 + }, + { + "epoch": 28.19, + "learning_rate": 3.0194458088936405e-06, + "loss": 0.9246, + "step": 325690 + }, + { + "epoch": 28.19, + "learning_rate": 3.0180010979802943e-06, + "loss": 1.0266, + "step": 325700 + }, + { + "epoch": 28.19, + "learning_rate": 3.016556387066948e-06, + "loss": 0.9387, + "step": 325710 + }, + { + "epoch": 28.19, + "learning_rate": 3.015111676153602e-06, + "loss": 0.912, + "step": 325720 + }, + { + "epoch": 28.19, + "learning_rate": 3.0136669652402554e-06, + "loss": 0.9347, + "step": 325730 + }, + { + "epoch": 28.2, + "learning_rate": 3.0122222543269096e-06, + "loss": 0.9222, + "step": 325740 + }, + { + "epoch": 28.2, + "learning_rate": 3.010777543413563e-06, + "loss": 0.8723, + "step": 325750 + }, + { + "epoch": 28.2, + "learning_rate": 3.009332832500217e-06, + "loss": 0.907, + "step": 325760 + }, + { + "epoch": 28.2, + "learning_rate": 3.0078881215868706e-06, + "loss": 0.8711, + "step": 325770 + }, + { + "epoch": 28.2, + "learning_rate": 3.0064434106735244e-06, + "loss": 0.9207, + "step": 325780 + }, + { + "epoch": 28.2, + "learning_rate": 3.0049986997601783e-06, + "loss": 0.8989, + "step": 325790 + }, + { + "epoch": 28.2, + "learning_rate": 3.0035539888468317e-06, + "loss": 0.8962, + "step": 325800 + }, + { + "epoch": 28.2, + "learning_rate": 3.002109277933486e-06, + "loss": 0.8423, + "step": 325810 + }, + { + "epoch": 28.2, + "learning_rate": 3.0006645670201393e-06, + "loss": 0.9204, + "step": 325820 + }, + { + "epoch": 28.2, + "learning_rate": 2.999219856106793e-06, + "loss": 0.9347, + "step": 325830 + }, + { + "epoch": 28.2, + "learning_rate": 2.997775145193447e-06, + "loss": 0.8231, + "step": 325840 + }, + { + "epoch": 28.2, + "learning_rate": 2.9963304342801007e-06, + "loss": 0.9037, + "step": 325850 + }, + { + "epoch": 28.21, + "learning_rate": 2.9948857233667546e-06, + "loss": 0.9318, + "step": 325860 + }, + { + "epoch": 28.21, + "learning_rate": 2.993441012453408e-06, + "loss": 0.9186, + "step": 325870 + }, + { + "epoch": 28.21, + "learning_rate": 2.991996301540062e-06, + "loss": 0.9326, + "step": 325880 + }, + { + "epoch": 28.21, + "learning_rate": 2.9905515906267156e-06, + "loss": 0.9187, + "step": 325890 + }, + { + "epoch": 28.21, + "learning_rate": 2.9891068797133694e-06, + "loss": 0.8198, + "step": 325900 + }, + { + "epoch": 28.21, + "learning_rate": 2.9876621688000232e-06, + "loss": 0.9115, + "step": 325910 + }, + { + "epoch": 28.21, + "learning_rate": 2.986217457886677e-06, + "loss": 0.8796, + "step": 325920 + }, + { + "epoch": 28.21, + "learning_rate": 2.984772746973331e-06, + "loss": 0.8446, + "step": 325930 + }, + { + "epoch": 28.21, + "learning_rate": 2.9833280360599847e-06, + "loss": 0.9356, + "step": 325940 + }, + { + "epoch": 28.21, + "learning_rate": 2.9818833251466385e-06, + "loss": 0.9115, + "step": 325950 + }, + { + "epoch": 28.21, + "learning_rate": 2.980438614233292e-06, + "loss": 0.9061, + "step": 325960 + }, + { + "epoch": 28.21, + "learning_rate": 2.978993903319946e-06, + "loss": 0.9477, + "step": 325970 + }, + { + "epoch": 28.22, + "learning_rate": 2.9775491924065995e-06, + "loss": 0.861, + "step": 325980 + }, + { + "epoch": 28.22, + "learning_rate": 2.9761044814932534e-06, + "loss": 0.8515, + "step": 325990 + }, + { + "epoch": 28.22, + "learning_rate": 2.9746597705799068e-06, + "loss": 0.9151, + "step": 326000 + }, + { + "epoch": 28.22, + "learning_rate": 2.973215059666561e-06, + "loss": 0.9099, + "step": 326010 + }, + { + "epoch": 28.22, + "learning_rate": 2.971770348753215e-06, + "loss": 0.8229, + "step": 326020 + }, + { + "epoch": 28.22, + "learning_rate": 2.970325637839868e-06, + "loss": 0.9643, + "step": 326030 + }, + { + "epoch": 28.22, + "learning_rate": 2.9688809269265225e-06, + "loss": 0.9277, + "step": 326040 + }, + { + "epoch": 28.22, + "learning_rate": 2.967436216013176e-06, + "loss": 0.9145, + "step": 326050 + }, + { + "epoch": 28.22, + "learning_rate": 2.9659915050998297e-06, + "loss": 0.9292, + "step": 326060 + }, + { + "epoch": 28.22, + "learning_rate": 2.9645467941864835e-06, + "loss": 0.8163, + "step": 326070 + }, + { + "epoch": 28.22, + "learning_rate": 2.9631020832731373e-06, + "loss": 0.9309, + "step": 326080 + }, + { + "epoch": 28.23, + "learning_rate": 2.9616573723597907e-06, + "loss": 0.9163, + "step": 326090 + }, + { + "epoch": 28.23, + "learning_rate": 2.9602126614464445e-06, + "loss": 0.902, + "step": 326100 + }, + { + "epoch": 28.23, + "learning_rate": 2.9587679505330988e-06, + "loss": 0.9179, + "step": 326110 + }, + { + "epoch": 28.23, + "learning_rate": 2.957323239619752e-06, + "loss": 0.9105, + "step": 326120 + }, + { + "epoch": 28.23, + "learning_rate": 2.955878528706406e-06, + "loss": 0.8472, + "step": 326130 + }, + { + "epoch": 28.23, + "learning_rate": 2.95443381779306e-06, + "loss": 0.982, + "step": 326140 + }, + { + "epoch": 28.23, + "learning_rate": 2.9529891068797136e-06, + "loss": 0.9093, + "step": 326150 + }, + { + "epoch": 28.23, + "learning_rate": 2.951544395966367e-06, + "loss": 0.9261, + "step": 326160 + }, + { + "epoch": 28.23, + "learning_rate": 2.9500996850530212e-06, + "loss": 0.9516, + "step": 326170 + }, + { + "epoch": 28.23, + "learning_rate": 2.9486549741396746e-06, + "loss": 0.9125, + "step": 326180 + }, + { + "epoch": 28.23, + "learning_rate": 2.9472102632263285e-06, + "loss": 0.9192, + "step": 326190 + }, + { + "epoch": 28.23, + "learning_rate": 2.9457655523129823e-06, + "loss": 0.9039, + "step": 326200 + }, + { + "epoch": 28.24, + "learning_rate": 2.944320841399636e-06, + "loss": 0.9406, + "step": 326210 + }, + { + "epoch": 28.24, + "learning_rate": 2.94287613048629e-06, + "loss": 0.9192, + "step": 326220 + }, + { + "epoch": 28.24, + "learning_rate": 2.9414314195729433e-06, + "loss": 0.9015, + "step": 326230 + }, + { + "epoch": 28.24, + "learning_rate": 2.9399867086595976e-06, + "loss": 0.9221, + "step": 326240 + }, + { + "epoch": 28.24, + "learning_rate": 2.938541997746251e-06, + "loss": 0.8936, + "step": 326250 + }, + { + "epoch": 28.24, + "learning_rate": 2.9370972868329048e-06, + "loss": 0.8885, + "step": 326260 + }, + { + "epoch": 28.24, + "learning_rate": 2.9356525759195586e-06, + "loss": 0.8694, + "step": 326270 + }, + { + "epoch": 28.24, + "learning_rate": 2.9342078650062124e-06, + "loss": 0.8706, + "step": 326280 + }, + { + "epoch": 28.24, + "learning_rate": 2.9327631540928662e-06, + "loss": 0.8448, + "step": 326290 + }, + { + "epoch": 28.24, + "learning_rate": 2.93131844317952e-06, + "loss": 0.915, + "step": 326300 + }, + { + "epoch": 28.24, + "learning_rate": 2.929873732266174e-06, + "loss": 0.9409, + "step": 326310 + }, + { + "epoch": 28.25, + "learning_rate": 2.9284290213528273e-06, + "loss": 0.9351, + "step": 326320 + }, + { + "epoch": 28.25, + "learning_rate": 2.926984310439481e-06, + "loss": 0.8636, + "step": 326330 + }, + { + "epoch": 28.25, + "learning_rate": 2.925539599526135e-06, + "loss": 0.899, + "step": 326340 + }, + { + "epoch": 28.25, + "learning_rate": 2.9240948886127887e-06, + "loss": 0.8895, + "step": 326350 + }, + { + "epoch": 28.25, + "learning_rate": 2.9226501776994425e-06, + "loss": 0.9466, + "step": 326360 + }, + { + "epoch": 28.25, + "learning_rate": 2.9212054667860963e-06, + "loss": 0.9066, + "step": 326370 + }, + { + "epoch": 28.25, + "learning_rate": 2.91976075587275e-06, + "loss": 0.8622, + "step": 326380 + }, + { + "epoch": 28.25, + "learning_rate": 2.9183160449594036e-06, + "loss": 0.891, + "step": 326390 + }, + { + "epoch": 28.25, + "learning_rate": 2.916871334046058e-06, + "loss": 0.9395, + "step": 326400 + }, + { + "epoch": 28.25, + "learning_rate": 2.915426623132711e-06, + "loss": 0.9446, + "step": 326410 + }, + { + "epoch": 28.25, + "learning_rate": 2.913981912219365e-06, + "loss": 0.8489, + "step": 326420 + }, + { + "epoch": 28.25, + "learning_rate": 2.912537201306019e-06, + "loss": 0.8618, + "step": 326430 + }, + { + "epoch": 28.26, + "learning_rate": 2.9110924903926727e-06, + "loss": 0.8483, + "step": 326440 + }, + { + "epoch": 28.26, + "learning_rate": 2.9096477794793265e-06, + "loss": 0.9085, + "step": 326450 + }, + { + "epoch": 28.26, + "learning_rate": 2.90820306856598e-06, + "loss": 0.8085, + "step": 326460 + }, + { + "epoch": 28.26, + "learning_rate": 2.906758357652634e-06, + "loss": 0.8466, + "step": 326470 + }, + { + "epoch": 28.26, + "learning_rate": 2.9053136467392875e-06, + "loss": 0.9175, + "step": 326480 + }, + { + "epoch": 28.26, + "learning_rate": 2.9038689358259413e-06, + "loss": 0.8711, + "step": 326490 + }, + { + "epoch": 28.26, + "learning_rate": 2.902424224912595e-06, + "loss": 0.916, + "step": 326500 + }, + { + "epoch": 28.26, + "learning_rate": 2.900979513999249e-06, + "loss": 0.9322, + "step": 326510 + }, + { + "epoch": 28.26, + "learning_rate": 2.8995348030859028e-06, + "loss": 0.926, + "step": 326520 + }, + { + "epoch": 28.26, + "learning_rate": 2.8980900921725566e-06, + "loss": 0.8548, + "step": 326530 + }, + { + "epoch": 28.26, + "learning_rate": 2.8966453812592104e-06, + "loss": 0.8752, + "step": 326540 + }, + { + "epoch": 28.27, + "learning_rate": 2.895200670345864e-06, + "loss": 0.9163, + "step": 326550 + }, + { + "epoch": 28.27, + "learning_rate": 2.8937559594325176e-06, + "loss": 0.9233, + "step": 326560 + }, + { + "epoch": 28.27, + "learning_rate": 2.8923112485191714e-06, + "loss": 0.8905, + "step": 326570 + }, + { + "epoch": 28.27, + "learning_rate": 2.8908665376058253e-06, + "loss": 0.8686, + "step": 326580 + }, + { + "epoch": 28.27, + "learning_rate": 2.8894218266924787e-06, + "loss": 0.9543, + "step": 326590 + }, + { + "epoch": 28.27, + "learning_rate": 2.887977115779133e-06, + "loss": 0.8812, + "step": 326600 + }, + { + "epoch": 28.27, + "learning_rate": 2.8865324048657867e-06, + "loss": 0.8581, + "step": 326610 + }, + { + "epoch": 28.27, + "learning_rate": 2.88508769395244e-06, + "loss": 0.8144, + "step": 326620 + }, + { + "epoch": 28.27, + "learning_rate": 2.8836429830390944e-06, + "loss": 0.9293, + "step": 326630 + }, + { + "epoch": 28.27, + "learning_rate": 2.8821982721257478e-06, + "loss": 0.9318, + "step": 326640 + }, + { + "epoch": 28.27, + "learning_rate": 2.8807535612124016e-06, + "loss": 0.9449, + "step": 326650 + }, + { + "epoch": 28.27, + "learning_rate": 2.879308850299055e-06, + "loss": 0.8837, + "step": 326660 + }, + { + "epoch": 28.28, + "learning_rate": 2.877864139385709e-06, + "loss": 0.9359, + "step": 326670 + }, + { + "epoch": 28.28, + "learning_rate": 2.8764194284723626e-06, + "loss": 0.8645, + "step": 326680 + }, + { + "epoch": 28.28, + "learning_rate": 2.8749747175590164e-06, + "loss": 0.9382, + "step": 326690 + }, + { + "epoch": 28.28, + "learning_rate": 2.8735300066456702e-06, + "loss": 0.8472, + "step": 326700 + }, + { + "epoch": 28.28, + "learning_rate": 2.872085295732324e-06, + "loss": 0.8941, + "step": 326710 + }, + { + "epoch": 28.28, + "learning_rate": 2.870640584818978e-06, + "loss": 0.9283, + "step": 326720 + }, + { + "epoch": 28.28, + "learning_rate": 2.8691958739056317e-06, + "loss": 0.8536, + "step": 326730 + }, + { + "epoch": 28.28, + "learning_rate": 2.8677511629922855e-06, + "loss": 0.9176, + "step": 326740 + }, + { + "epoch": 28.28, + "learning_rate": 2.866306452078939e-06, + "loss": 0.9087, + "step": 326750 + }, + { + "epoch": 28.28, + "learning_rate": 2.864861741165593e-06, + "loss": 0.8682, + "step": 326760 + }, + { + "epoch": 28.28, + "learning_rate": 2.8634170302522465e-06, + "loss": 0.9502, + "step": 326770 + }, + { + "epoch": 28.29, + "learning_rate": 2.8619723193389004e-06, + "loss": 0.8625, + "step": 326780 + }, + { + "epoch": 28.29, + "learning_rate": 2.860527608425554e-06, + "loss": 0.932, + "step": 326790 + }, + { + "epoch": 28.29, + "learning_rate": 2.859082897512208e-06, + "loss": 0.9208, + "step": 326800 + }, + { + "epoch": 28.29, + "learning_rate": 2.857638186598862e-06, + "loss": 0.9035, + "step": 326810 + }, + { + "epoch": 28.29, + "learning_rate": 2.8561934756855152e-06, + "loss": 0.873, + "step": 326820 + }, + { + "epoch": 28.29, + "learning_rate": 2.8547487647721695e-06, + "loss": 0.863, + "step": 326830 + }, + { + "epoch": 28.29, + "learning_rate": 2.853304053858823e-06, + "loss": 0.8724, + "step": 326840 + }, + { + "epoch": 28.29, + "learning_rate": 2.8518593429454767e-06, + "loss": 0.9444, + "step": 326850 + }, + { + "epoch": 28.29, + "learning_rate": 2.8504146320321305e-06, + "loss": 0.9283, + "step": 326860 + }, + { + "epoch": 28.29, + "learning_rate": 2.8489699211187843e-06, + "loss": 0.9115, + "step": 326870 + }, + { + "epoch": 28.29, + "learning_rate": 2.847525210205438e-06, + "loss": 0.9315, + "step": 326880 + }, + { + "epoch": 28.29, + "learning_rate": 2.8460804992920915e-06, + "loss": 0.8908, + "step": 326890 + }, + { + "epoch": 28.3, + "learning_rate": 2.8446357883787458e-06, + "loss": 0.9562, + "step": 326900 + }, + { + "epoch": 28.3, + "learning_rate": 2.843191077465399e-06, + "loss": 0.9111, + "step": 326910 + }, + { + "epoch": 28.3, + "learning_rate": 2.841746366552053e-06, + "loss": 0.8667, + "step": 326920 + }, + { + "epoch": 28.3, + "learning_rate": 2.840301655638707e-06, + "loss": 0.8608, + "step": 326930 + }, + { + "epoch": 28.3, + "learning_rate": 2.8388569447253606e-06, + "loss": 0.8575, + "step": 326940 + }, + { + "epoch": 28.3, + "learning_rate": 2.8374122338120144e-06, + "loss": 0.9116, + "step": 326950 + }, + { + "epoch": 28.3, + "learning_rate": 2.8359675228986682e-06, + "loss": 0.8594, + "step": 326960 + }, + { + "epoch": 28.3, + "learning_rate": 2.834522811985322e-06, + "loss": 0.9166, + "step": 326970 + }, + { + "epoch": 28.3, + "learning_rate": 2.8330781010719755e-06, + "loss": 0.8774, + "step": 326980 + }, + { + "epoch": 28.3, + "learning_rate": 2.8316333901586297e-06, + "loss": 0.9065, + "step": 326990 + }, + { + "epoch": 28.3, + "learning_rate": 2.830188679245283e-06, + "loss": 0.8731, + "step": 327000 + }, + { + "epoch": 28.31, + "learning_rate": 2.828743968331937e-06, + "loss": 0.9705, + "step": 327010 + }, + { + "epoch": 28.31, + "learning_rate": 2.8272992574185907e-06, + "loss": 0.9092, + "step": 327020 + }, + { + "epoch": 28.31, + "learning_rate": 2.8258545465052446e-06, + "loss": 0.9025, + "step": 327030 + }, + { + "epoch": 28.31, + "learning_rate": 2.8244098355918984e-06, + "loss": 0.8865, + "step": 327040 + }, + { + "epoch": 28.31, + "learning_rate": 2.8229651246785518e-06, + "loss": 0.8869, + "step": 327050 + }, + { + "epoch": 28.31, + "learning_rate": 2.821520413765206e-06, + "loss": 0.8472, + "step": 327060 + }, + { + "epoch": 28.31, + "learning_rate": 2.8200757028518594e-06, + "loss": 0.8895, + "step": 327070 + }, + { + "epoch": 28.31, + "learning_rate": 2.8186309919385132e-06, + "loss": 0.8608, + "step": 327080 + }, + { + "epoch": 28.31, + "learning_rate": 2.817186281025167e-06, + "loss": 0.9249, + "step": 327090 + }, + { + "epoch": 28.31, + "learning_rate": 2.815741570111821e-06, + "loss": 0.8765, + "step": 327100 + }, + { + "epoch": 28.31, + "learning_rate": 2.8142968591984747e-06, + "loss": 0.9597, + "step": 327110 + }, + { + "epoch": 28.31, + "learning_rate": 2.812852148285128e-06, + "loss": 0.8944, + "step": 327120 + }, + { + "epoch": 28.32, + "learning_rate": 2.8114074373717823e-06, + "loss": 0.8461, + "step": 327130 + }, + { + "epoch": 28.32, + "learning_rate": 2.8099627264584357e-06, + "loss": 0.9362, + "step": 327140 + }, + { + "epoch": 28.32, + "learning_rate": 2.8085180155450895e-06, + "loss": 0.887, + "step": 327150 + }, + { + "epoch": 28.32, + "learning_rate": 2.8070733046317433e-06, + "loss": 0.9458, + "step": 327160 + }, + { + "epoch": 28.32, + "learning_rate": 2.805628593718397e-06, + "loss": 0.9166, + "step": 327170 + }, + { + "epoch": 28.32, + "learning_rate": 2.8041838828050506e-06, + "loss": 0.8493, + "step": 327180 + }, + { + "epoch": 28.32, + "learning_rate": 2.802739171891705e-06, + "loss": 0.9177, + "step": 327190 + }, + { + "epoch": 28.32, + "learning_rate": 2.801294460978358e-06, + "loss": 0.9724, + "step": 327200 + }, + { + "epoch": 28.32, + "learning_rate": 2.799849750065012e-06, + "loss": 0.9303, + "step": 327210 + }, + { + "epoch": 28.32, + "learning_rate": 2.798405039151666e-06, + "loss": 0.8867, + "step": 327220 + }, + { + "epoch": 28.32, + "learning_rate": 2.7969603282383197e-06, + "loss": 0.9443, + "step": 327230 + }, + { + "epoch": 28.32, + "learning_rate": 2.7955156173249735e-06, + "loss": 0.868, + "step": 327240 + }, + { + "epoch": 28.33, + "learning_rate": 2.794070906411627e-06, + "loss": 0.8653, + "step": 327250 + }, + { + "epoch": 28.33, + "learning_rate": 2.792626195498281e-06, + "loss": 0.8368, + "step": 327260 + }, + { + "epoch": 28.33, + "learning_rate": 2.7911814845849345e-06, + "loss": 0.9055, + "step": 327270 + }, + { + "epoch": 28.33, + "learning_rate": 2.7897367736715883e-06, + "loss": 0.9085, + "step": 327280 + }, + { + "epoch": 28.33, + "learning_rate": 2.788292062758242e-06, + "loss": 0.9052, + "step": 327290 + }, + { + "epoch": 28.33, + "learning_rate": 2.786847351844896e-06, + "loss": 0.9022, + "step": 327300 + }, + { + "epoch": 28.33, + "learning_rate": 2.7854026409315498e-06, + "loss": 0.8994, + "step": 327310 + }, + { + "epoch": 28.33, + "learning_rate": 2.7839579300182036e-06, + "loss": 0.9021, + "step": 327320 + }, + { + "epoch": 28.33, + "learning_rate": 2.7825132191048574e-06, + "loss": 0.9282, + "step": 327330 + }, + { + "epoch": 28.33, + "learning_rate": 2.781068508191511e-06, + "loss": 0.8453, + "step": 327340 + }, + { + "epoch": 28.33, + "learning_rate": 2.7796237972781646e-06, + "loss": 0.9405, + "step": 327350 + }, + { + "epoch": 28.34, + "learning_rate": 2.7781790863648184e-06, + "loss": 0.8877, + "step": 327360 + }, + { + "epoch": 28.34, + "learning_rate": 2.7767343754514723e-06, + "loss": 0.8784, + "step": 327370 + }, + { + "epoch": 28.34, + "learning_rate": 2.775289664538126e-06, + "loss": 0.9409, + "step": 327380 + }, + { + "epoch": 28.34, + "learning_rate": 2.77384495362478e-06, + "loss": 0.9406, + "step": 327390 + }, + { + "epoch": 28.34, + "learning_rate": 2.7724002427114337e-06, + "loss": 0.8803, + "step": 327400 + }, + { + "epoch": 28.34, + "learning_rate": 2.770955531798087e-06, + "loss": 0.8726, + "step": 327410 + }, + { + "epoch": 28.34, + "learning_rate": 2.7695108208847414e-06, + "loss": 0.9334, + "step": 327420 + }, + { + "epoch": 28.34, + "learning_rate": 2.7680661099713948e-06, + "loss": 0.9094, + "step": 327430 + }, + { + "epoch": 28.34, + "learning_rate": 2.7666213990580486e-06, + "loss": 0.8594, + "step": 327440 + }, + { + "epoch": 28.34, + "learning_rate": 2.7651766881447024e-06, + "loss": 0.9115, + "step": 327450 + }, + { + "epoch": 28.34, + "learning_rate": 2.763731977231356e-06, + "loss": 0.8954, + "step": 327460 + }, + { + "epoch": 28.34, + "learning_rate": 2.76228726631801e-06, + "loss": 0.8524, + "step": 327470 + }, + { + "epoch": 28.35, + "learning_rate": 2.7608425554046634e-06, + "loss": 0.9264, + "step": 327480 + }, + { + "epoch": 28.35, + "learning_rate": 2.7593978444913177e-06, + "loss": 0.8545, + "step": 327490 + }, + { + "epoch": 28.35, + "learning_rate": 2.757953133577971e-06, + "loss": 0.8967, + "step": 327500 + }, + { + "epoch": 28.35, + "learning_rate": 2.756508422664625e-06, + "loss": 0.8547, + "step": 327510 + }, + { + "epoch": 28.35, + "learning_rate": 2.7550637117512787e-06, + "loss": 0.9208, + "step": 327520 + }, + { + "epoch": 28.35, + "learning_rate": 2.7536190008379325e-06, + "loss": 0.8753, + "step": 327530 + }, + { + "epoch": 28.35, + "learning_rate": 2.7521742899245863e-06, + "loss": 0.9155, + "step": 327540 + }, + { + "epoch": 28.35, + "learning_rate": 2.75072957901124e-06, + "loss": 0.9209, + "step": 327550 + }, + { + "epoch": 28.35, + "learning_rate": 2.749284868097894e-06, + "loss": 0.8834, + "step": 327560 + }, + { + "epoch": 28.35, + "learning_rate": 2.7478401571845474e-06, + "loss": 0.9179, + "step": 327570 + }, + { + "epoch": 28.35, + "learning_rate": 2.746395446271201e-06, + "loss": 0.9117, + "step": 327580 + }, + { + "epoch": 28.36, + "learning_rate": 2.744950735357855e-06, + "loss": 0.9138, + "step": 327590 + }, + { + "epoch": 28.36, + "learning_rate": 2.743506024444509e-06, + "loss": 0.8802, + "step": 327600 + }, + { + "epoch": 28.36, + "learning_rate": 2.7420613135311626e-06, + "loss": 0.8935, + "step": 327610 + }, + { + "epoch": 28.36, + "learning_rate": 2.7406166026178165e-06, + "loss": 0.9153, + "step": 327620 + }, + { + "epoch": 28.36, + "learning_rate": 2.7391718917044703e-06, + "loss": 0.977, + "step": 327630 + }, + { + "epoch": 28.36, + "learning_rate": 2.7377271807911237e-06, + "loss": 0.8636, + "step": 327640 + }, + { + "epoch": 28.36, + "learning_rate": 2.736282469877778e-06, + "loss": 0.9146, + "step": 327650 + }, + { + "epoch": 28.36, + "learning_rate": 2.7348377589644313e-06, + "loss": 0.8236, + "step": 327660 + }, + { + "epoch": 28.36, + "learning_rate": 2.733393048051085e-06, + "loss": 0.912, + "step": 327670 + }, + { + "epoch": 28.36, + "learning_rate": 2.7319483371377385e-06, + "loss": 0.85, + "step": 327680 + }, + { + "epoch": 28.36, + "learning_rate": 2.7305036262243928e-06, + "loss": 0.9273, + "step": 327690 + }, + { + "epoch": 28.36, + "learning_rate": 2.729058915311046e-06, + "loss": 0.9197, + "step": 327700 + }, + { + "epoch": 28.37, + "learning_rate": 2.7276142043977e-06, + "loss": 0.8732, + "step": 327710 + }, + { + "epoch": 28.37, + "learning_rate": 2.7261694934843542e-06, + "loss": 0.8325, + "step": 327720 + }, + { + "epoch": 28.37, + "learning_rate": 2.7247247825710076e-06, + "loss": 0.9302, + "step": 327730 + }, + { + "epoch": 28.37, + "learning_rate": 2.7232800716576614e-06, + "loss": 0.8663, + "step": 327740 + }, + { + "epoch": 28.37, + "learning_rate": 2.7218353607443152e-06, + "loss": 0.8851, + "step": 327750 + }, + { + "epoch": 28.37, + "learning_rate": 2.720390649830969e-06, + "loss": 0.9313, + "step": 327760 + }, + { + "epoch": 28.37, + "learning_rate": 2.7189459389176225e-06, + "loss": 0.8499, + "step": 327770 + }, + { + "epoch": 28.37, + "learning_rate": 2.7175012280042767e-06, + "loss": 0.9231, + "step": 327780 + }, + { + "epoch": 28.37, + "learning_rate": 2.71605651709093e-06, + "loss": 0.8769, + "step": 327790 + }, + { + "epoch": 28.37, + "learning_rate": 2.714611806177584e-06, + "loss": 0.8874, + "step": 327800 + }, + { + "epoch": 28.37, + "learning_rate": 2.7131670952642377e-06, + "loss": 0.9103, + "step": 327810 + }, + { + "epoch": 28.38, + "learning_rate": 2.7117223843508916e-06, + "loss": 0.9035, + "step": 327820 + }, + { + "epoch": 28.38, + "learning_rate": 2.7102776734375454e-06, + "loss": 0.9135, + "step": 327830 + }, + { + "epoch": 28.38, + "learning_rate": 2.7088329625241988e-06, + "loss": 0.8784, + "step": 327840 + }, + { + "epoch": 28.38, + "learning_rate": 2.707388251610853e-06, + "loss": 0.9241, + "step": 327850 + }, + { + "epoch": 28.38, + "learning_rate": 2.7059435406975064e-06, + "loss": 0.9034, + "step": 327860 + }, + { + "epoch": 28.38, + "learning_rate": 2.7044988297841602e-06, + "loss": 0.9627, + "step": 327870 + }, + { + "epoch": 28.38, + "learning_rate": 2.703054118870814e-06, + "loss": 0.9342, + "step": 327880 + }, + { + "epoch": 28.38, + "learning_rate": 2.701609407957468e-06, + "loss": 0.9447, + "step": 327890 + }, + { + "epoch": 28.38, + "learning_rate": 2.7001646970441217e-06, + "loss": 0.9108, + "step": 327900 + }, + { + "epoch": 28.38, + "learning_rate": 2.698719986130775e-06, + "loss": 0.9063, + "step": 327910 + }, + { + "epoch": 28.38, + "learning_rate": 2.6972752752174293e-06, + "loss": 0.8949, + "step": 327920 + }, + { + "epoch": 28.38, + "learning_rate": 2.6958305643040827e-06, + "loss": 0.9036, + "step": 327930 + }, + { + "epoch": 28.39, + "learning_rate": 2.6943858533907365e-06, + "loss": 0.8712, + "step": 327940 + }, + { + "epoch": 28.39, + "learning_rate": 2.6929411424773903e-06, + "loss": 0.8363, + "step": 327950 + }, + { + "epoch": 28.39, + "learning_rate": 2.691496431564044e-06, + "loss": 0.8806, + "step": 327960 + }, + { + "epoch": 28.39, + "learning_rate": 2.690051720650698e-06, + "loss": 0.8324, + "step": 327970 + }, + { + "epoch": 28.39, + "learning_rate": 2.688607009737352e-06, + "loss": 0.8887, + "step": 327980 + }, + { + "epoch": 28.39, + "learning_rate": 2.6871622988240056e-06, + "loss": 0.9394, + "step": 327990 + }, + { + "epoch": 28.39, + "learning_rate": 2.685717587910659e-06, + "loss": 0.9239, + "step": 328000 + }, + { + "epoch": 28.39, + "learning_rate": 2.684272876997313e-06, + "loss": 0.8925, + "step": 328010 + }, + { + "epoch": 28.39, + "learning_rate": 2.6828281660839667e-06, + "loss": 0.8837, + "step": 328020 + }, + { + "epoch": 28.39, + "learning_rate": 2.6813834551706205e-06, + "loss": 0.8864, + "step": 328030 + }, + { + "epoch": 28.39, + "learning_rate": 2.6799387442572743e-06, + "loss": 0.938, + "step": 328040 + }, + { + "epoch": 28.4, + "learning_rate": 2.678494033343928e-06, + "loss": 0.8761, + "step": 328050 + }, + { + "epoch": 28.4, + "learning_rate": 2.677049322430582e-06, + "loss": 0.9283, + "step": 328060 + }, + { + "epoch": 28.4, + "learning_rate": 2.6756046115172353e-06, + "loss": 0.8764, + "step": 328070 + }, + { + "epoch": 28.4, + "learning_rate": 2.6741599006038896e-06, + "loss": 0.8969, + "step": 328080 + }, + { + "epoch": 28.4, + "learning_rate": 2.672715189690543e-06, + "loss": 0.9671, + "step": 328090 + }, + { + "epoch": 28.4, + "learning_rate": 2.6712704787771968e-06, + "loss": 0.9486, + "step": 328100 + }, + { + "epoch": 28.4, + "learning_rate": 2.6698257678638506e-06, + "loss": 0.8971, + "step": 328110 + }, + { + "epoch": 28.4, + "learning_rate": 2.6683810569505044e-06, + "loss": 0.8859, + "step": 328120 + }, + { + "epoch": 28.4, + "learning_rate": 2.6669363460371582e-06, + "loss": 0.8934, + "step": 328130 + }, + { + "epoch": 28.4, + "learning_rate": 2.6654916351238116e-06, + "loss": 0.9665, + "step": 328140 + }, + { + "epoch": 28.4, + "learning_rate": 2.664046924210466e-06, + "loss": 0.8773, + "step": 328150 + }, + { + "epoch": 28.4, + "learning_rate": 2.6626022132971193e-06, + "loss": 0.8824, + "step": 328160 + }, + { + "epoch": 28.41, + "learning_rate": 2.661157502383773e-06, + "loss": 0.8893, + "step": 328170 + }, + { + "epoch": 28.41, + "learning_rate": 2.659712791470427e-06, + "loss": 0.8813, + "step": 328180 + }, + { + "epoch": 28.41, + "learning_rate": 2.6582680805570807e-06, + "loss": 0.8882, + "step": 328190 + }, + { + "epoch": 28.41, + "learning_rate": 2.656823369643734e-06, + "loss": 0.8613, + "step": 328200 + }, + { + "epoch": 28.41, + "learning_rate": 2.6553786587303884e-06, + "loss": 0.9673, + "step": 328210 + }, + { + "epoch": 28.41, + "learning_rate": 2.653933947817042e-06, + "loss": 0.9413, + "step": 328220 + }, + { + "epoch": 28.41, + "learning_rate": 2.6524892369036956e-06, + "loss": 0.8287, + "step": 328230 + }, + { + "epoch": 28.41, + "learning_rate": 2.6510445259903494e-06, + "loss": 0.8606, + "step": 328240 + }, + { + "epoch": 28.41, + "learning_rate": 2.649599815077003e-06, + "loss": 0.8841, + "step": 328250 + }, + { + "epoch": 28.41, + "learning_rate": 2.648155104163657e-06, + "loss": 0.908, + "step": 328260 + }, + { + "epoch": 28.41, + "learning_rate": 2.6467103932503104e-06, + "loss": 0.914, + "step": 328270 + }, + { + "epoch": 28.41, + "learning_rate": 2.6452656823369647e-06, + "loss": 0.9205, + "step": 328280 + }, + { + "epoch": 28.42, + "learning_rate": 2.643820971423618e-06, + "loss": 0.8483, + "step": 328290 + }, + { + "epoch": 28.42, + "learning_rate": 2.642376260510272e-06, + "loss": 0.9053, + "step": 328300 + }, + { + "epoch": 28.42, + "learning_rate": 2.640931549596926e-06, + "loss": 0.9818, + "step": 328310 + }, + { + "epoch": 28.42, + "learning_rate": 2.6394868386835795e-06, + "loss": 0.9653, + "step": 328320 + }, + { + "epoch": 28.42, + "learning_rate": 2.6380421277702333e-06, + "loss": 0.8974, + "step": 328330 + }, + { + "epoch": 28.42, + "learning_rate": 2.636597416856887e-06, + "loss": 0.8319, + "step": 328340 + }, + { + "epoch": 28.42, + "learning_rate": 2.635152705943541e-06, + "loss": 0.9242, + "step": 328350 + }, + { + "epoch": 28.42, + "learning_rate": 2.6337079950301944e-06, + "loss": 0.8916, + "step": 328360 + }, + { + "epoch": 28.42, + "learning_rate": 2.632263284116848e-06, + "loss": 0.9222, + "step": 328370 + }, + { + "epoch": 28.42, + "learning_rate": 2.630818573203502e-06, + "loss": 0.9112, + "step": 328380 + }, + { + "epoch": 28.42, + "learning_rate": 2.629373862290156e-06, + "loss": 0.8899, + "step": 328390 + }, + { + "epoch": 28.43, + "learning_rate": 2.6279291513768096e-06, + "loss": 0.9057, + "step": 328400 + }, + { + "epoch": 28.43, + "learning_rate": 2.6264844404634635e-06, + "loss": 0.8622, + "step": 328410 + }, + { + "epoch": 28.43, + "learning_rate": 2.6250397295501173e-06, + "loss": 0.8802, + "step": 328420 + }, + { + "epoch": 28.43, + "learning_rate": 2.6235950186367707e-06, + "loss": 0.9444, + "step": 328430 + }, + { + "epoch": 28.43, + "learning_rate": 2.622150307723425e-06, + "loss": 0.8948, + "step": 328440 + }, + { + "epoch": 28.43, + "learning_rate": 2.6207055968100783e-06, + "loss": 0.9251, + "step": 328450 + }, + { + "epoch": 28.43, + "learning_rate": 2.619260885896732e-06, + "loss": 0.8894, + "step": 328460 + }, + { + "epoch": 28.43, + "learning_rate": 2.617816174983386e-06, + "loss": 0.8908, + "step": 328470 + }, + { + "epoch": 28.43, + "learning_rate": 2.6163714640700398e-06, + "loss": 0.8838, + "step": 328480 + }, + { + "epoch": 28.43, + "learning_rate": 2.6149267531566936e-06, + "loss": 0.9003, + "step": 328490 + }, + { + "epoch": 28.43, + "learning_rate": 2.613482042243347e-06, + "loss": 0.9381, + "step": 328500 + }, + { + "epoch": 28.43, + "learning_rate": 2.6120373313300012e-06, + "loss": 0.8898, + "step": 328510 + }, + { + "epoch": 28.44, + "learning_rate": 2.6105926204166546e-06, + "loss": 0.8989, + "step": 328520 + }, + { + "epoch": 28.44, + "learning_rate": 2.6091479095033084e-06, + "loss": 0.851, + "step": 328530 + }, + { + "epoch": 28.44, + "learning_rate": 2.6077031985899623e-06, + "loss": 0.8593, + "step": 328540 + }, + { + "epoch": 28.44, + "learning_rate": 2.606258487676616e-06, + "loss": 0.8778, + "step": 328550 + }, + { + "epoch": 28.44, + "learning_rate": 2.60481377676327e-06, + "loss": 0.9062, + "step": 328560 + }, + { + "epoch": 28.44, + "learning_rate": 2.6033690658499237e-06, + "loss": 0.8618, + "step": 328570 + }, + { + "epoch": 28.44, + "learning_rate": 2.6019243549365775e-06, + "loss": 0.9039, + "step": 328580 + }, + { + "epoch": 28.44, + "learning_rate": 2.600479644023231e-06, + "loss": 0.8483, + "step": 328590 + }, + { + "epoch": 28.44, + "learning_rate": 2.5990349331098847e-06, + "loss": 0.8978, + "step": 328600 + }, + { + "epoch": 28.44, + "learning_rate": 2.5975902221965386e-06, + "loss": 0.9191, + "step": 328610 + }, + { + "epoch": 28.44, + "learning_rate": 2.5961455112831924e-06, + "loss": 0.8651, + "step": 328620 + }, + { + "epoch": 28.45, + "learning_rate": 2.594700800369846e-06, + "loss": 0.8886, + "step": 328630 + }, + { + "epoch": 28.45, + "learning_rate": 2.5932560894565e-06, + "loss": 0.8538, + "step": 328640 + }, + { + "epoch": 28.45, + "learning_rate": 2.591811378543154e-06, + "loss": 0.9341, + "step": 328650 + }, + { + "epoch": 28.45, + "learning_rate": 2.5903666676298072e-06, + "loss": 0.9344, + "step": 328660 + }, + { + "epoch": 28.45, + "learning_rate": 2.5889219567164615e-06, + "loss": 0.9785, + "step": 328670 + }, + { + "epoch": 28.45, + "learning_rate": 2.587477245803115e-06, + "loss": 0.894, + "step": 328680 + }, + { + "epoch": 28.45, + "learning_rate": 2.5860325348897687e-06, + "loss": 0.9047, + "step": 328690 + }, + { + "epoch": 28.45, + "learning_rate": 2.584587823976422e-06, + "loss": 0.8748, + "step": 328700 + }, + { + "epoch": 28.45, + "learning_rate": 2.5831431130630763e-06, + "loss": 0.8707, + "step": 328710 + }, + { + "epoch": 28.45, + "learning_rate": 2.58169840214973e-06, + "loss": 0.9005, + "step": 328720 + }, + { + "epoch": 28.45, + "learning_rate": 2.5802536912363835e-06, + "loss": 0.9071, + "step": 328730 + }, + { + "epoch": 28.45, + "learning_rate": 2.5788089803230378e-06, + "loss": 0.9058, + "step": 328740 + }, + { + "epoch": 28.46, + "learning_rate": 2.577364269409691e-06, + "loss": 0.9145, + "step": 328750 + }, + { + "epoch": 28.46, + "learning_rate": 2.575919558496345e-06, + "loss": 0.857, + "step": 328760 + }, + { + "epoch": 28.46, + "learning_rate": 2.574474847582999e-06, + "loss": 0.8982, + "step": 328770 + }, + { + "epoch": 28.46, + "learning_rate": 2.5730301366696526e-06, + "loss": 0.8769, + "step": 328780 + }, + { + "epoch": 28.46, + "learning_rate": 2.571585425756306e-06, + "loss": 0.9161, + "step": 328790 + }, + { + "epoch": 28.46, + "learning_rate": 2.57014071484296e-06, + "loss": 0.8302, + "step": 328800 + }, + { + "epoch": 28.46, + "learning_rate": 2.568696003929614e-06, + "loss": 0.9076, + "step": 328810 + }, + { + "epoch": 28.46, + "learning_rate": 2.5672512930162675e-06, + "loss": 0.8882, + "step": 328820 + }, + { + "epoch": 28.46, + "learning_rate": 2.5658065821029213e-06, + "loss": 0.8935, + "step": 328830 + }, + { + "epoch": 28.46, + "learning_rate": 2.564361871189575e-06, + "loss": 0.871, + "step": 328840 + }, + { + "epoch": 28.46, + "learning_rate": 2.562917160276229e-06, + "loss": 0.8954, + "step": 328850 + }, + { + "epoch": 28.47, + "learning_rate": 2.5614724493628823e-06, + "loss": 0.8971, + "step": 328860 + }, + { + "epoch": 28.47, + "learning_rate": 2.5600277384495366e-06, + "loss": 0.9381, + "step": 328870 + }, + { + "epoch": 28.47, + "learning_rate": 2.55858302753619e-06, + "loss": 0.9236, + "step": 328880 + }, + { + "epoch": 28.47, + "learning_rate": 2.5571383166228438e-06, + "loss": 0.9029, + "step": 328890 + }, + { + "epoch": 28.47, + "learning_rate": 2.555693605709498e-06, + "loss": 0.9046, + "step": 328900 + }, + { + "epoch": 28.47, + "learning_rate": 2.5542488947961514e-06, + "loss": 0.8853, + "step": 328910 + }, + { + "epoch": 28.47, + "learning_rate": 2.5528041838828052e-06, + "loss": 0.9181, + "step": 328920 + }, + { + "epoch": 28.47, + "learning_rate": 2.5513594729694586e-06, + "loss": 0.9011, + "step": 328930 + }, + { + "epoch": 28.47, + "learning_rate": 2.549914762056113e-06, + "loss": 0.8974, + "step": 328940 + }, + { + "epoch": 28.47, + "learning_rate": 2.5484700511427663e-06, + "loss": 0.9596, + "step": 328950 + }, + { + "epoch": 28.47, + "learning_rate": 2.54702534022942e-06, + "loss": 0.9305, + "step": 328960 + }, + { + "epoch": 28.47, + "learning_rate": 2.545580629316074e-06, + "loss": 0.8855, + "step": 328970 + }, + { + "epoch": 28.48, + "learning_rate": 2.5441359184027277e-06, + "loss": 0.923, + "step": 328980 + }, + { + "epoch": 28.48, + "learning_rate": 2.5426912074893815e-06, + "loss": 0.8878, + "step": 328990 + }, + { + "epoch": 28.48, + "learning_rate": 2.5412464965760354e-06, + "loss": 0.8886, + "step": 329000 + }, + { + "epoch": 28.48, + "learning_rate": 2.539801785662689e-06, + "loss": 0.9398, + "step": 329010 + }, + { + "epoch": 28.48, + "learning_rate": 2.5383570747493426e-06, + "loss": 0.9032, + "step": 329020 + }, + { + "epoch": 28.48, + "learning_rate": 2.5369123638359964e-06, + "loss": 0.925, + "step": 329030 + }, + { + "epoch": 28.48, + "learning_rate": 2.53546765292265e-06, + "loss": 0.9544, + "step": 329040 + }, + { + "epoch": 28.48, + "learning_rate": 2.534022942009304e-06, + "loss": 0.9061, + "step": 329050 + }, + { + "epoch": 28.48, + "learning_rate": 2.532578231095958e-06, + "loss": 0.9007, + "step": 329060 + }, + { + "epoch": 28.48, + "learning_rate": 2.5311335201826117e-06, + "loss": 0.9211, + "step": 329070 + }, + { + "epoch": 28.48, + "learning_rate": 2.5296888092692655e-06, + "loss": 0.8976, + "step": 329080 + }, + { + "epoch": 28.49, + "learning_rate": 2.528244098355919e-06, + "loss": 0.9413, + "step": 329090 + }, + { + "epoch": 28.49, + "learning_rate": 2.526799387442573e-06, + "loss": 0.9322, + "step": 329100 + }, + { + "epoch": 28.49, + "learning_rate": 2.5253546765292265e-06, + "loss": 0.9028, + "step": 329110 + }, + { + "epoch": 28.49, + "learning_rate": 2.5239099656158803e-06, + "loss": 0.8835, + "step": 329120 + }, + { + "epoch": 28.49, + "learning_rate": 2.522465254702534e-06, + "loss": 0.9261, + "step": 329130 + }, + { + "epoch": 28.49, + "learning_rate": 2.521020543789188e-06, + "loss": 0.8758, + "step": 329140 + }, + { + "epoch": 28.49, + "learning_rate": 2.519575832875842e-06, + "loss": 0.921, + "step": 329150 + }, + { + "epoch": 28.49, + "learning_rate": 2.518131121962495e-06, + "loss": 0.8902, + "step": 329160 + }, + { + "epoch": 28.49, + "learning_rate": 2.5166864110491494e-06, + "loss": 0.8532, + "step": 329170 + }, + { + "epoch": 28.49, + "learning_rate": 2.515241700135803e-06, + "loss": 0.9119, + "step": 329180 + }, + { + "epoch": 28.49, + "learning_rate": 2.5137969892224566e-06, + "loss": 0.9489, + "step": 329190 + }, + { + "epoch": 28.49, + "learning_rate": 2.5123522783091105e-06, + "loss": 0.9418, + "step": 329200 + }, + { + "epoch": 28.5, + "learning_rate": 2.5109075673957643e-06, + "loss": 0.8604, + "step": 329210 + }, + { + "epoch": 28.5, + "learning_rate": 2.509462856482418e-06, + "loss": 0.8836, + "step": 329220 + }, + { + "epoch": 28.5, + "learning_rate": 2.508018145569072e-06, + "loss": 0.9153, + "step": 329230 + }, + { + "epoch": 28.5, + "learning_rate": 2.5065734346557257e-06, + "loss": 0.9062, + "step": 329240 + }, + { + "epoch": 28.5, + "learning_rate": 2.505128723742379e-06, + "loss": 0.9246, + "step": 329250 + }, + { + "epoch": 28.5, + "learning_rate": 2.503684012829033e-06, + "loss": 0.9001, + "step": 329260 + }, + { + "epoch": 28.5, + "learning_rate": 2.5022393019156868e-06, + "loss": 0.8943, + "step": 329270 + }, + { + "epoch": 28.5, + "learning_rate": 2.5007945910023406e-06, + "loss": 0.9169, + "step": 329280 + }, + { + "epoch": 28.5, + "learning_rate": 2.499349880088994e-06, + "loss": 0.941, + "step": 329290 + }, + { + "epoch": 28.5, + "learning_rate": 2.4979051691756482e-06, + "loss": 0.8384, + "step": 329300 + }, + { + "epoch": 28.5, + "learning_rate": 2.496460458262302e-06, + "loss": 0.8954, + "step": 329310 + }, + { + "epoch": 28.5, + "learning_rate": 2.4950157473489554e-06, + "loss": 0.8549, + "step": 329320 + }, + { + "epoch": 28.51, + "learning_rate": 2.4935710364356097e-06, + "loss": 0.8898, + "step": 329330 + }, + { + "epoch": 28.51, + "learning_rate": 2.492126325522263e-06, + "loss": 0.8967, + "step": 329340 + }, + { + "epoch": 28.51, + "learning_rate": 2.490681614608917e-06, + "loss": 0.9514, + "step": 329350 + }, + { + "epoch": 28.51, + "learning_rate": 2.4892369036955707e-06, + "loss": 0.9612, + "step": 329360 + }, + { + "epoch": 28.51, + "learning_rate": 2.4877921927822245e-06, + "loss": 0.9099, + "step": 329370 + }, + { + "epoch": 28.51, + "learning_rate": 2.486347481868878e-06, + "loss": 0.8838, + "step": 329380 + }, + { + "epoch": 28.51, + "learning_rate": 2.4849027709555317e-06, + "loss": 0.8692, + "step": 329390 + }, + { + "epoch": 28.51, + "learning_rate": 2.483458060042186e-06, + "loss": 0.9156, + "step": 329400 + }, + { + "epoch": 28.51, + "learning_rate": 2.4820133491288394e-06, + "loss": 0.874, + "step": 329410 + }, + { + "epoch": 28.51, + "learning_rate": 2.480568638215493e-06, + "loss": 0.866, + "step": 329420 + }, + { + "epoch": 28.51, + "learning_rate": 2.479123927302147e-06, + "loss": 0.913, + "step": 329430 + }, + { + "epoch": 28.52, + "learning_rate": 2.477679216388801e-06, + "loss": 0.8872, + "step": 329440 + }, + { + "epoch": 28.52, + "learning_rate": 2.4762345054754542e-06, + "loss": 0.9015, + "step": 329450 + }, + { + "epoch": 28.52, + "learning_rate": 2.4747897945621085e-06, + "loss": 0.8952, + "step": 329460 + }, + { + "epoch": 28.52, + "learning_rate": 2.473345083648762e-06, + "loss": 0.9642, + "step": 329470 + }, + { + "epoch": 28.52, + "learning_rate": 2.4719003727354157e-06, + "loss": 0.8492, + "step": 329480 + }, + { + "epoch": 28.52, + "learning_rate": 2.4704556618220695e-06, + "loss": 0.8759, + "step": 329490 + }, + { + "epoch": 28.52, + "learning_rate": 2.4690109509087233e-06, + "loss": 0.8725, + "step": 329500 + }, + { + "epoch": 28.52, + "learning_rate": 2.467566239995377e-06, + "loss": 0.8589, + "step": 329510 + }, + { + "epoch": 28.52, + "learning_rate": 2.4661215290820305e-06, + "loss": 0.8967, + "step": 329520 + }, + { + "epoch": 28.52, + "learning_rate": 2.4646768181686848e-06, + "loss": 0.8597, + "step": 329530 + }, + { + "epoch": 28.52, + "learning_rate": 2.463232107255338e-06, + "loss": 0.8589, + "step": 329540 + }, + { + "epoch": 28.52, + "learning_rate": 2.461787396341992e-06, + "loss": 0.9056, + "step": 329550 + }, + { + "epoch": 28.53, + "learning_rate": 2.460342685428646e-06, + "loss": 0.8811, + "step": 329560 + }, + { + "epoch": 28.53, + "learning_rate": 2.4588979745152996e-06, + "loss": 0.9128, + "step": 329570 + }, + { + "epoch": 28.53, + "learning_rate": 2.4574532636019534e-06, + "loss": 0.8962, + "step": 329580 + }, + { + "epoch": 28.53, + "learning_rate": 2.4560085526886073e-06, + "loss": 0.9389, + "step": 329590 + }, + { + "epoch": 28.53, + "learning_rate": 2.454563841775261e-06, + "loss": 0.8925, + "step": 329600 + }, + { + "epoch": 28.53, + "learning_rate": 2.4531191308619145e-06, + "loss": 0.8806, + "step": 329610 + }, + { + "epoch": 28.53, + "learning_rate": 2.4516744199485683e-06, + "loss": 0.9241, + "step": 329620 + }, + { + "epoch": 28.53, + "learning_rate": 2.450229709035222e-06, + "loss": 0.8952, + "step": 329630 + }, + { + "epoch": 28.53, + "learning_rate": 2.448784998121876e-06, + "loss": 0.9035, + "step": 329640 + }, + { + "epoch": 28.53, + "learning_rate": 2.4473402872085297e-06, + "loss": 0.899, + "step": 329650 + }, + { + "epoch": 28.53, + "learning_rate": 2.4458955762951836e-06, + "loss": 0.8554, + "step": 329660 + }, + { + "epoch": 28.54, + "learning_rate": 2.4444508653818374e-06, + "loss": 0.8937, + "step": 329670 + }, + { + "epoch": 28.54, + "learning_rate": 2.4430061544684908e-06, + "loss": 0.8552, + "step": 329680 + }, + { + "epoch": 28.54, + "learning_rate": 2.441561443555145e-06, + "loss": 0.9141, + "step": 329690 + }, + { + "epoch": 28.54, + "learning_rate": 2.4401167326417984e-06, + "loss": 0.8978, + "step": 329700 + }, + { + "epoch": 28.54, + "learning_rate": 2.4386720217284522e-06, + "loss": 0.8732, + "step": 329710 + }, + { + "epoch": 28.54, + "learning_rate": 2.437227310815106e-06, + "loss": 0.9085, + "step": 329720 + }, + { + "epoch": 28.54, + "learning_rate": 2.43578259990176e-06, + "loss": 0.917, + "step": 329730 + }, + { + "epoch": 28.54, + "learning_rate": 2.4343378889884137e-06, + "loss": 0.8665, + "step": 329740 + }, + { + "epoch": 28.54, + "learning_rate": 2.432893178075067e-06, + "loss": 0.8884, + "step": 329750 + }, + { + "epoch": 28.54, + "learning_rate": 2.4314484671617213e-06, + "loss": 0.9524, + "step": 329760 + }, + { + "epoch": 28.54, + "learning_rate": 2.4300037562483747e-06, + "loss": 0.9417, + "step": 329770 + }, + { + "epoch": 28.54, + "learning_rate": 2.4285590453350285e-06, + "loss": 0.9125, + "step": 329780 + }, + { + "epoch": 28.55, + "learning_rate": 2.4271143344216824e-06, + "loss": 0.823, + "step": 329790 + }, + { + "epoch": 28.55, + "learning_rate": 2.425669623508336e-06, + "loss": 0.9398, + "step": 329800 + }, + { + "epoch": 28.55, + "learning_rate": 2.42422491259499e-06, + "loss": 0.8892, + "step": 329810 + }, + { + "epoch": 28.55, + "learning_rate": 2.4227802016816434e-06, + "loss": 0.9188, + "step": 329820 + }, + { + "epoch": 28.55, + "learning_rate": 2.4213354907682976e-06, + "loss": 0.9299, + "step": 329830 + }, + { + "epoch": 28.55, + "learning_rate": 2.419890779854951e-06, + "loss": 0.9322, + "step": 329840 + }, + { + "epoch": 28.55, + "learning_rate": 2.418446068941605e-06, + "loss": 0.8497, + "step": 329850 + }, + { + "epoch": 28.55, + "learning_rate": 2.4170013580282587e-06, + "loss": 0.9057, + "step": 329860 + }, + { + "epoch": 28.55, + "learning_rate": 2.4155566471149125e-06, + "loss": 0.8457, + "step": 329870 + }, + { + "epoch": 28.55, + "learning_rate": 2.414111936201566e-06, + "loss": 0.9253, + "step": 329880 + }, + { + "epoch": 28.55, + "learning_rate": 2.41266722528822e-06, + "loss": 0.9008, + "step": 329890 + }, + { + "epoch": 28.56, + "learning_rate": 2.411222514374874e-06, + "loss": 0.8801, + "step": 329900 + }, + { + "epoch": 28.56, + "learning_rate": 2.4097778034615273e-06, + "loss": 0.8907, + "step": 329910 + }, + { + "epoch": 28.56, + "learning_rate": 2.4083330925481816e-06, + "loss": 0.9196, + "step": 329920 + }, + { + "epoch": 28.56, + "learning_rate": 2.406888381634835e-06, + "loss": 0.8274, + "step": 329930 + }, + { + "epoch": 28.56, + "learning_rate": 2.405443670721489e-06, + "loss": 0.8856, + "step": 329940 + }, + { + "epoch": 28.56, + "learning_rate": 2.403998959808142e-06, + "loss": 0.8881, + "step": 329950 + }, + { + "epoch": 28.56, + "learning_rate": 2.4025542488947964e-06, + "loss": 0.929, + "step": 329960 + }, + { + "epoch": 28.56, + "learning_rate": 2.40110953798145e-06, + "loss": 0.8523, + "step": 329970 + }, + { + "epoch": 28.56, + "learning_rate": 2.3996648270681036e-06, + "loss": 0.814, + "step": 329980 + }, + { + "epoch": 28.56, + "learning_rate": 2.3982201161547575e-06, + "loss": 0.9165, + "step": 329990 + }, + { + "epoch": 28.56, + "learning_rate": 2.3967754052414113e-06, + "loss": 0.8783, + "step": 330000 + }, + { + "epoch": 28.56, + "learning_rate": 2.395330694328065e-06, + "loss": 0.8941, + "step": 330010 + }, + { + "epoch": 28.57, + "learning_rate": 2.393885983414719e-06, + "loss": 0.8212, + "step": 330020 + }, + { + "epoch": 28.57, + "learning_rate": 2.3924412725013727e-06, + "loss": 0.9382, + "step": 330030 + }, + { + "epoch": 28.57, + "learning_rate": 2.390996561588026e-06, + "loss": 0.932, + "step": 330040 + }, + { + "epoch": 28.57, + "learning_rate": 2.38955185067468e-06, + "loss": 0.9085, + "step": 330050 + }, + { + "epoch": 28.57, + "learning_rate": 2.3881071397613338e-06, + "loss": 0.9102, + "step": 330060 + }, + { + "epoch": 28.57, + "learning_rate": 2.3866624288479876e-06, + "loss": 0.8751, + "step": 330070 + }, + { + "epoch": 28.57, + "learning_rate": 2.3852177179346414e-06, + "loss": 0.9408, + "step": 330080 + }, + { + "epoch": 28.57, + "learning_rate": 2.3837730070212952e-06, + "loss": 0.8416, + "step": 330090 + }, + { + "epoch": 28.57, + "learning_rate": 2.382328296107949e-06, + "loss": 0.9089, + "step": 330100 + }, + { + "epoch": 28.57, + "learning_rate": 2.3808835851946024e-06, + "loss": 0.859, + "step": 330110 + }, + { + "epoch": 28.57, + "learning_rate": 2.3794388742812567e-06, + "loss": 0.8498, + "step": 330120 + }, + { + "epoch": 28.58, + "learning_rate": 2.37799416336791e-06, + "loss": 0.8659, + "step": 330130 + }, + { + "epoch": 28.58, + "learning_rate": 2.376549452454564e-06, + "loss": 0.8889, + "step": 330140 + }, + { + "epoch": 28.58, + "learning_rate": 2.3751047415412177e-06, + "loss": 0.8808, + "step": 330150 + }, + { + "epoch": 28.58, + "learning_rate": 2.3736600306278715e-06, + "loss": 0.9399, + "step": 330160 + }, + { + "epoch": 28.58, + "learning_rate": 2.3722153197145253e-06, + "loss": 0.9374, + "step": 330170 + }, + { + "epoch": 28.58, + "learning_rate": 2.3707706088011787e-06, + "loss": 0.9195, + "step": 330180 + }, + { + "epoch": 28.58, + "learning_rate": 2.369325897887833e-06, + "loss": 0.8941, + "step": 330190 + }, + { + "epoch": 28.58, + "learning_rate": 2.3678811869744864e-06, + "loss": 0.9101, + "step": 330200 + }, + { + "epoch": 28.58, + "learning_rate": 2.36643647606114e-06, + "loss": 0.9229, + "step": 330210 + }, + { + "epoch": 28.58, + "learning_rate": 2.364991765147794e-06, + "loss": 0.8591, + "step": 330220 + }, + { + "epoch": 28.58, + "learning_rate": 2.363547054234448e-06, + "loss": 0.8109, + "step": 330230 + }, + { + "epoch": 28.58, + "learning_rate": 2.3621023433211017e-06, + "loss": 0.8747, + "step": 330240 + }, + { + "epoch": 28.59, + "learning_rate": 2.3606576324077555e-06, + "loss": 0.8921, + "step": 330250 + }, + { + "epoch": 28.59, + "learning_rate": 2.3592129214944093e-06, + "loss": 0.8519, + "step": 330260 + }, + { + "epoch": 28.59, + "learning_rate": 2.3577682105810627e-06, + "loss": 0.8891, + "step": 330270 + }, + { + "epoch": 28.59, + "learning_rate": 2.3563234996677165e-06, + "loss": 0.9498, + "step": 330280 + }, + { + "epoch": 28.59, + "learning_rate": 2.3548787887543703e-06, + "loss": 0.9518, + "step": 330290 + }, + { + "epoch": 28.59, + "learning_rate": 2.353434077841024e-06, + "loss": 0.9852, + "step": 330300 + }, + { + "epoch": 28.59, + "learning_rate": 2.351989366927678e-06, + "loss": 0.9542, + "step": 330310 + }, + { + "epoch": 28.59, + "learning_rate": 2.3505446560143318e-06, + "loss": 0.8777, + "step": 330320 + }, + { + "epoch": 28.59, + "learning_rate": 2.3490999451009856e-06, + "loss": 0.9242, + "step": 330330 + }, + { + "epoch": 28.59, + "learning_rate": 2.347655234187639e-06, + "loss": 0.9408, + "step": 330340 + }, + { + "epoch": 28.59, + "learning_rate": 2.3462105232742932e-06, + "loss": 0.8831, + "step": 330350 + }, + { + "epoch": 28.59, + "learning_rate": 2.3447658123609466e-06, + "loss": 0.8347, + "step": 330360 + }, + { + "epoch": 28.6, + "learning_rate": 2.3433211014476004e-06, + "loss": 0.8983, + "step": 330370 + }, + { + "epoch": 28.6, + "learning_rate": 2.3418763905342543e-06, + "loss": 0.927, + "step": 330380 + }, + { + "epoch": 28.6, + "learning_rate": 2.340431679620908e-06, + "loss": 0.8814, + "step": 330390 + }, + { + "epoch": 28.6, + "learning_rate": 2.338986968707562e-06, + "loss": 0.8855, + "step": 330400 + }, + { + "epoch": 28.6, + "learning_rate": 2.3375422577942153e-06, + "loss": 0.8773, + "step": 330410 + }, + { + "epoch": 28.6, + "learning_rate": 2.3360975468808695e-06, + "loss": 0.8986, + "step": 330420 + }, + { + "epoch": 28.6, + "learning_rate": 2.334652835967523e-06, + "loss": 0.9409, + "step": 330430 + }, + { + "epoch": 28.6, + "learning_rate": 2.3332081250541768e-06, + "loss": 0.9194, + "step": 330440 + }, + { + "epoch": 28.6, + "learning_rate": 2.3317634141408306e-06, + "loss": 0.8513, + "step": 330450 + }, + { + "epoch": 28.6, + "learning_rate": 2.3303187032274844e-06, + "loss": 0.9089, + "step": 330460 + }, + { + "epoch": 28.6, + "learning_rate": 2.3288739923141378e-06, + "loss": 0.9164, + "step": 330470 + }, + { + "epoch": 28.61, + "learning_rate": 2.327429281400792e-06, + "loss": 0.9138, + "step": 330480 + }, + { + "epoch": 28.61, + "learning_rate": 2.3259845704874454e-06, + "loss": 0.8545, + "step": 330490 + }, + { + "epoch": 28.61, + "learning_rate": 2.3245398595740992e-06, + "loss": 0.8278, + "step": 330500 + }, + { + "epoch": 28.61, + "learning_rate": 2.323095148660753e-06, + "loss": 0.9506, + "step": 330510 + }, + { + "epoch": 28.61, + "learning_rate": 2.321650437747407e-06, + "loss": 0.8631, + "step": 330520 + }, + { + "epoch": 28.61, + "learning_rate": 2.3202057268340607e-06, + "loss": 0.8417, + "step": 330530 + }, + { + "epoch": 28.61, + "learning_rate": 2.318761015920714e-06, + "loss": 0.9027, + "step": 330540 + }, + { + "epoch": 28.61, + "learning_rate": 2.3173163050073683e-06, + "loss": 0.9624, + "step": 330550 + }, + { + "epoch": 28.61, + "learning_rate": 2.3158715940940217e-06, + "loss": 0.8925, + "step": 330560 + }, + { + "epoch": 28.61, + "learning_rate": 2.3144268831806755e-06, + "loss": 0.9608, + "step": 330570 + }, + { + "epoch": 28.61, + "learning_rate": 2.3129821722673294e-06, + "loss": 0.8901, + "step": 330580 + }, + { + "epoch": 28.61, + "learning_rate": 2.311537461353983e-06, + "loss": 0.8665, + "step": 330590 + }, + { + "epoch": 28.62, + "learning_rate": 2.310092750440637e-06, + "loss": 0.8683, + "step": 330600 + }, + { + "epoch": 28.62, + "learning_rate": 2.3086480395272904e-06, + "loss": 0.8306, + "step": 330610 + }, + { + "epoch": 28.62, + "learning_rate": 2.3072033286139446e-06, + "loss": 0.9244, + "step": 330620 + }, + { + "epoch": 28.62, + "learning_rate": 2.305758617700598e-06, + "loss": 0.8351, + "step": 330630 + }, + { + "epoch": 28.62, + "learning_rate": 2.304313906787252e-06, + "loss": 0.857, + "step": 330640 + }, + { + "epoch": 28.62, + "learning_rate": 2.3028691958739057e-06, + "loss": 0.9011, + "step": 330650 + }, + { + "epoch": 28.62, + "learning_rate": 2.3014244849605595e-06, + "loss": 0.9215, + "step": 330660 + }, + { + "epoch": 28.62, + "learning_rate": 2.2999797740472133e-06, + "loss": 0.9539, + "step": 330670 + }, + { + "epoch": 28.62, + "learning_rate": 2.298535063133867e-06, + "loss": 0.8903, + "step": 330680 + }, + { + "epoch": 28.62, + "learning_rate": 2.297090352220521e-06, + "loss": 0.9096, + "step": 330690 + }, + { + "epoch": 28.62, + "learning_rate": 2.2956456413071743e-06, + "loss": 0.8741, + "step": 330700 + }, + { + "epoch": 28.63, + "learning_rate": 2.2942009303938286e-06, + "loss": 0.8785, + "step": 330710 + }, + { + "epoch": 28.63, + "learning_rate": 2.292756219480482e-06, + "loss": 0.913, + "step": 330720 + }, + { + "epoch": 28.63, + "learning_rate": 2.291311508567136e-06, + "loss": 0.8808, + "step": 330730 + }, + { + "epoch": 28.63, + "learning_rate": 2.2898667976537896e-06, + "loss": 0.9399, + "step": 330740 + }, + { + "epoch": 28.63, + "learning_rate": 2.2884220867404434e-06, + "loss": 0.867, + "step": 330750 + }, + { + "epoch": 28.63, + "learning_rate": 2.2869773758270972e-06, + "loss": 0.9216, + "step": 330760 + }, + { + "epoch": 28.63, + "learning_rate": 2.2855326649137506e-06, + "loss": 0.8807, + "step": 330770 + }, + { + "epoch": 28.63, + "learning_rate": 2.284087954000405e-06, + "loss": 0.8763, + "step": 330780 + }, + { + "epoch": 28.63, + "learning_rate": 2.2826432430870583e-06, + "loss": 0.902, + "step": 330790 + }, + { + "epoch": 28.63, + "learning_rate": 2.281198532173712e-06, + "loss": 0.8896, + "step": 330800 + }, + { + "epoch": 28.63, + "learning_rate": 2.279753821260366e-06, + "loss": 0.9359, + "step": 330810 + }, + { + "epoch": 28.63, + "learning_rate": 2.2783091103470197e-06, + "loss": 0.9035, + "step": 330820 + }, + { + "epoch": 28.64, + "learning_rate": 2.2768643994336736e-06, + "loss": 0.8866, + "step": 330830 + }, + { + "epoch": 28.64, + "learning_rate": 2.275419688520327e-06, + "loss": 0.88, + "step": 330840 + }, + { + "epoch": 28.64, + "learning_rate": 2.273974977606981e-06, + "loss": 0.8754, + "step": 330850 + }, + { + "epoch": 28.64, + "learning_rate": 2.2725302666936346e-06, + "loss": 0.9537, + "step": 330860 + }, + { + "epoch": 28.64, + "learning_rate": 2.2710855557802884e-06, + "loss": 0.8847, + "step": 330870 + }, + { + "epoch": 28.64, + "learning_rate": 2.2696408448669422e-06, + "loss": 0.8971, + "step": 330880 + }, + { + "epoch": 28.64, + "learning_rate": 2.268196133953596e-06, + "loss": 0.8731, + "step": 330890 + }, + { + "epoch": 28.64, + "learning_rate": 2.26675142304025e-06, + "loss": 0.9074, + "step": 330900 + }, + { + "epoch": 28.64, + "learning_rate": 2.2653067121269037e-06, + "loss": 0.9161, + "step": 330910 + }, + { + "epoch": 28.64, + "learning_rate": 2.2638620012135575e-06, + "loss": 0.8718, + "step": 330920 + }, + { + "epoch": 28.64, + "learning_rate": 2.262417290300211e-06, + "loss": 0.9022, + "step": 330930 + }, + { + "epoch": 28.65, + "learning_rate": 2.260972579386865e-06, + "loss": 0.857, + "step": 330940 + }, + { + "epoch": 28.65, + "learning_rate": 2.2595278684735185e-06, + "loss": 0.8313, + "step": 330950 + }, + { + "epoch": 28.65, + "learning_rate": 2.2580831575601723e-06, + "loss": 0.9147, + "step": 330960 + }, + { + "epoch": 28.65, + "learning_rate": 2.2566384466468257e-06, + "loss": 0.8902, + "step": 330970 + }, + { + "epoch": 28.65, + "learning_rate": 2.25519373573348e-06, + "loss": 0.9463, + "step": 330980 + }, + { + "epoch": 28.65, + "learning_rate": 2.2537490248201334e-06, + "loss": 0.8721, + "step": 330990 + }, + { + "epoch": 28.65, + "learning_rate": 2.252304313906787e-06, + "loss": 0.8833, + "step": 331000 + }, + { + "epoch": 28.65, + "learning_rate": 2.2508596029934414e-06, + "loss": 0.901, + "step": 331010 + }, + { + "epoch": 28.65, + "learning_rate": 2.249414892080095e-06, + "loss": 0.8737, + "step": 331020 + }, + { + "epoch": 28.65, + "learning_rate": 2.2479701811667487e-06, + "loss": 0.8645, + "step": 331030 + }, + { + "epoch": 28.65, + "learning_rate": 2.2465254702534025e-06, + "loss": 0.8831, + "step": 331040 + }, + { + "epoch": 28.65, + "learning_rate": 2.2450807593400563e-06, + "loss": 0.8975, + "step": 331050 + }, + { + "epoch": 28.66, + "learning_rate": 2.2436360484267097e-06, + "loss": 0.9128, + "step": 331060 + }, + { + "epoch": 28.66, + "learning_rate": 2.2421913375133635e-06, + "loss": 0.8511, + "step": 331070 + }, + { + "epoch": 28.66, + "learning_rate": 2.2407466266000173e-06, + "loss": 0.9066, + "step": 331080 + }, + { + "epoch": 28.66, + "learning_rate": 2.239301915686671e-06, + "loss": 0.9077, + "step": 331090 + }, + { + "epoch": 28.66, + "learning_rate": 2.237857204773325e-06, + "loss": 0.8989, + "step": 331100 + }, + { + "epoch": 28.66, + "learning_rate": 2.2364124938599788e-06, + "loss": 0.854, + "step": 331110 + }, + { + "epoch": 28.66, + "learning_rate": 2.2349677829466326e-06, + "loss": 0.8806, + "step": 331120 + }, + { + "epoch": 28.66, + "learning_rate": 2.233523072033286e-06, + "loss": 0.8286, + "step": 331130 + }, + { + "epoch": 28.66, + "learning_rate": 2.2320783611199402e-06, + "loss": 0.9141, + "step": 331140 + }, + { + "epoch": 28.66, + "learning_rate": 2.2306336502065936e-06, + "loss": 0.8553, + "step": 331150 + }, + { + "epoch": 28.66, + "learning_rate": 2.2291889392932474e-06, + "loss": 0.8975, + "step": 331160 + }, + { + "epoch": 28.67, + "learning_rate": 2.2277442283799013e-06, + "loss": 0.8996, + "step": 331170 + }, + { + "epoch": 28.67, + "learning_rate": 2.226299517466555e-06, + "loss": 0.8304, + "step": 331180 + }, + { + "epoch": 28.67, + "learning_rate": 2.224854806553209e-06, + "loss": 0.9135, + "step": 331190 + }, + { + "epoch": 28.67, + "learning_rate": 2.2234100956398623e-06, + "loss": 0.8718, + "step": 331200 + }, + { + "epoch": 28.67, + "learning_rate": 2.2219653847265165e-06, + "loss": 0.8766, + "step": 331210 + }, + { + "epoch": 28.67, + "learning_rate": 2.22052067381317e-06, + "loss": 0.9234, + "step": 331220 + }, + { + "epoch": 28.67, + "learning_rate": 2.2190759628998238e-06, + "loss": 0.9666, + "step": 331230 + }, + { + "epoch": 28.67, + "learning_rate": 2.2176312519864776e-06, + "loss": 0.8821, + "step": 331240 + }, + { + "epoch": 28.67, + "learning_rate": 2.2161865410731314e-06, + "loss": 0.9297, + "step": 331250 + }, + { + "epoch": 28.67, + "learning_rate": 2.214741830159785e-06, + "loss": 0.8786, + "step": 331260 + }, + { + "epoch": 28.67, + "learning_rate": 2.213297119246439e-06, + "loss": 0.8958, + "step": 331270 + }, + { + "epoch": 28.67, + "learning_rate": 2.211852408333093e-06, + "loss": 0.8787, + "step": 331280 + }, + { + "epoch": 28.68, + "learning_rate": 2.2104076974197462e-06, + "loss": 0.8641, + "step": 331290 + }, + { + "epoch": 28.68, + "learning_rate": 2.2089629865064e-06, + "loss": 0.886, + "step": 331300 + }, + { + "epoch": 28.68, + "learning_rate": 2.207518275593054e-06, + "loss": 0.8627, + "step": 331310 + }, + { + "epoch": 28.68, + "learning_rate": 2.2060735646797077e-06, + "loss": 0.8802, + "step": 331320 + }, + { + "epoch": 28.68, + "learning_rate": 2.2046288537663615e-06, + "loss": 0.9281, + "step": 331330 + }, + { + "epoch": 28.68, + "learning_rate": 2.2031841428530153e-06, + "loss": 0.8928, + "step": 331340 + }, + { + "epoch": 28.68, + "learning_rate": 2.201739431939669e-06, + "loss": 0.8653, + "step": 331350 + }, + { + "epoch": 28.68, + "learning_rate": 2.2002947210263225e-06, + "loss": 0.9281, + "step": 331360 + }, + { + "epoch": 28.68, + "learning_rate": 2.1988500101129768e-06, + "loss": 0.8765, + "step": 331370 + }, + { + "epoch": 28.68, + "learning_rate": 2.19740529919963e-06, + "loss": 0.9255, + "step": 331380 + }, + { + "epoch": 28.68, + "learning_rate": 2.195960588286284e-06, + "loss": 0.9049, + "step": 331390 + }, + { + "epoch": 28.68, + "learning_rate": 2.194515877372938e-06, + "loss": 0.933, + "step": 331400 + }, + { + "epoch": 28.69, + "learning_rate": 2.1930711664595916e-06, + "loss": 0.8742, + "step": 331410 + }, + { + "epoch": 28.69, + "learning_rate": 2.1916264555462455e-06, + "loss": 0.8487, + "step": 331420 + }, + { + "epoch": 28.69, + "learning_rate": 2.190181744632899e-06, + "loss": 0.8898, + "step": 331430 + }, + { + "epoch": 28.69, + "learning_rate": 2.188737033719553e-06, + "loss": 0.878, + "step": 331440 + }, + { + "epoch": 28.69, + "learning_rate": 2.1872923228062065e-06, + "loss": 0.9359, + "step": 331450 + }, + { + "epoch": 28.69, + "learning_rate": 2.1858476118928603e-06, + "loss": 0.9327, + "step": 331460 + }, + { + "epoch": 28.69, + "learning_rate": 2.184402900979514e-06, + "loss": 0.8738, + "step": 331470 + }, + { + "epoch": 28.69, + "learning_rate": 2.182958190066168e-06, + "loss": 0.9113, + "step": 331480 + }, + { + "epoch": 28.69, + "learning_rate": 2.1815134791528213e-06, + "loss": 0.8948, + "step": 331490 + }, + { + "epoch": 28.69, + "learning_rate": 2.1800687682394756e-06, + "loss": 0.9568, + "step": 331500 + }, + { + "epoch": 28.69, + "learning_rate": 2.1786240573261294e-06, + "loss": 0.8826, + "step": 331510 + }, + { + "epoch": 28.7, + "learning_rate": 2.177179346412783e-06, + "loss": 0.8682, + "step": 331520 + }, + { + "epoch": 28.7, + "learning_rate": 2.1757346354994366e-06, + "loss": 0.9038, + "step": 331530 + }, + { + "epoch": 28.7, + "learning_rate": 2.1742899245860904e-06, + "loss": 0.8941, + "step": 331540 + }, + { + "epoch": 28.7, + "learning_rate": 2.1728452136727442e-06, + "loss": 0.8598, + "step": 331550 + }, + { + "epoch": 28.7, + "learning_rate": 2.1714005027593976e-06, + "loss": 0.9172, + "step": 331560 + }, + { + "epoch": 28.7, + "learning_rate": 2.169955791846052e-06, + "loss": 0.8411, + "step": 331570 + }, + { + "epoch": 28.7, + "learning_rate": 2.1685110809327053e-06, + "loss": 0.8917, + "step": 331580 + }, + { + "epoch": 28.7, + "learning_rate": 2.167066370019359e-06, + "loss": 0.9013, + "step": 331590 + }, + { + "epoch": 28.7, + "learning_rate": 2.1656216591060133e-06, + "loss": 0.8859, + "step": 331600 + }, + { + "epoch": 28.7, + "learning_rate": 2.1641769481926667e-06, + "loss": 0.9429, + "step": 331610 + }, + { + "epoch": 28.7, + "learning_rate": 2.1627322372793206e-06, + "loss": 0.9115, + "step": 331620 + }, + { + "epoch": 28.7, + "learning_rate": 2.161287526365974e-06, + "loss": 0.913, + "step": 331630 + }, + { + "epoch": 28.71, + "learning_rate": 2.159842815452628e-06, + "loss": 0.9316, + "step": 331640 + }, + { + "epoch": 28.71, + "learning_rate": 2.1583981045392816e-06, + "loss": 0.8513, + "step": 331650 + }, + { + "epoch": 28.71, + "learning_rate": 2.1569533936259354e-06, + "loss": 0.9058, + "step": 331660 + }, + { + "epoch": 28.71, + "learning_rate": 2.1555086827125892e-06, + "loss": 0.9103, + "step": 331670 + }, + { + "epoch": 28.71, + "learning_rate": 2.154063971799243e-06, + "loss": 0.9274, + "step": 331680 + }, + { + "epoch": 28.71, + "learning_rate": 2.152619260885897e-06, + "loss": 0.8414, + "step": 331690 + }, + { + "epoch": 28.71, + "learning_rate": 2.1511745499725507e-06, + "loss": 0.9336, + "step": 331700 + }, + { + "epoch": 28.71, + "learning_rate": 2.1497298390592045e-06, + "loss": 0.8552, + "step": 331710 + }, + { + "epoch": 28.71, + "learning_rate": 2.148285128145858e-06, + "loss": 0.9297, + "step": 331720 + }, + { + "epoch": 28.71, + "learning_rate": 2.146840417232512e-06, + "loss": 0.8996, + "step": 331730 + }, + { + "epoch": 28.71, + "learning_rate": 2.1453957063191655e-06, + "loss": 0.9203, + "step": 331740 + }, + { + "epoch": 28.72, + "learning_rate": 2.1439509954058193e-06, + "loss": 0.8744, + "step": 331750 + }, + { + "epoch": 28.72, + "learning_rate": 2.142506284492473e-06, + "loss": 0.8816, + "step": 331760 + }, + { + "epoch": 28.72, + "learning_rate": 2.141061573579127e-06, + "loss": 0.8724, + "step": 331770 + }, + { + "epoch": 28.72, + "learning_rate": 2.139616862665781e-06, + "loss": 0.9002, + "step": 331780 + }, + { + "epoch": 28.72, + "learning_rate": 2.138172151752434e-06, + "loss": 0.9299, + "step": 331790 + }, + { + "epoch": 28.72, + "learning_rate": 2.1367274408390884e-06, + "loss": 0.8564, + "step": 331800 + }, + { + "epoch": 28.72, + "learning_rate": 2.135282729925742e-06, + "loss": 0.859, + "step": 331810 + }, + { + "epoch": 28.72, + "learning_rate": 2.1338380190123957e-06, + "loss": 0.9137, + "step": 331820 + }, + { + "epoch": 28.72, + "learning_rate": 2.1323933080990495e-06, + "loss": 0.9526, + "step": 331830 + }, + { + "epoch": 28.72, + "learning_rate": 2.1309485971857033e-06, + "loss": 0.895, + "step": 331840 + }, + { + "epoch": 28.72, + "learning_rate": 2.129503886272357e-06, + "loss": 0.9082, + "step": 331850 + }, + { + "epoch": 28.72, + "learning_rate": 2.1280591753590105e-06, + "loss": 0.9283, + "step": 331860 + }, + { + "epoch": 28.73, + "learning_rate": 2.1266144644456647e-06, + "loss": 0.9124, + "step": 331870 + }, + { + "epoch": 28.73, + "learning_rate": 2.125169753532318e-06, + "loss": 0.926, + "step": 331880 + }, + { + "epoch": 28.73, + "learning_rate": 2.123725042618972e-06, + "loss": 0.9102, + "step": 331890 + }, + { + "epoch": 28.73, + "learning_rate": 2.1222803317056258e-06, + "loss": 0.927, + "step": 331900 + }, + { + "epoch": 28.73, + "learning_rate": 2.1208356207922796e-06, + "loss": 0.9328, + "step": 331910 + }, + { + "epoch": 28.73, + "learning_rate": 2.1193909098789334e-06, + "loss": 0.8806, + "step": 331920 + }, + { + "epoch": 28.73, + "learning_rate": 2.1179461989655872e-06, + "loss": 0.9573, + "step": 331930 + }, + { + "epoch": 28.73, + "learning_rate": 2.116501488052241e-06, + "loss": 0.8635, + "step": 331940 + }, + { + "epoch": 28.73, + "learning_rate": 2.1150567771388944e-06, + "loss": 0.879, + "step": 331950 + }, + { + "epoch": 28.73, + "learning_rate": 2.1136120662255487e-06, + "loss": 0.8771, + "step": 331960 + }, + { + "epoch": 28.73, + "learning_rate": 2.112167355312202e-06, + "loss": 0.9055, + "step": 331970 + }, + { + "epoch": 28.74, + "learning_rate": 2.110722644398856e-06, + "loss": 0.8764, + "step": 331980 + }, + { + "epoch": 28.74, + "learning_rate": 2.1092779334855093e-06, + "loss": 0.9369, + "step": 331990 + }, + { + "epoch": 28.74, + "learning_rate": 2.1078332225721635e-06, + "loss": 0.835, + "step": 332000 + }, + { + "epoch": 28.74, + "learning_rate": 2.1063885116588174e-06, + "loss": 0.9208, + "step": 332010 + }, + { + "epoch": 28.74, + "learning_rate": 2.1049438007454708e-06, + "loss": 0.8351, + "step": 332020 + }, + { + "epoch": 28.74, + "learning_rate": 2.103499089832125e-06, + "loss": 0.8669, + "step": 332030 + }, + { + "epoch": 28.74, + "learning_rate": 2.1020543789187784e-06, + "loss": 0.8533, + "step": 332040 + }, + { + "epoch": 28.74, + "learning_rate": 2.100609668005432e-06, + "loss": 0.9327, + "step": 332050 + }, + { + "epoch": 28.74, + "learning_rate": 2.099164957092086e-06, + "loss": 0.8341, + "step": 332060 + }, + { + "epoch": 28.74, + "learning_rate": 2.09772024617874e-06, + "loss": 0.9384, + "step": 332070 + }, + { + "epoch": 28.74, + "learning_rate": 2.0962755352653932e-06, + "loss": 0.9285, + "step": 332080 + }, + { + "epoch": 28.74, + "learning_rate": 2.094830824352047e-06, + "loss": 0.9086, + "step": 332090 + }, + { + "epoch": 28.75, + "learning_rate": 2.0933861134387013e-06, + "loss": 0.8521, + "step": 332100 + }, + { + "epoch": 28.75, + "learning_rate": 2.0919414025253547e-06, + "loss": 0.8796, + "step": 332110 + }, + { + "epoch": 28.75, + "learning_rate": 2.0904966916120085e-06, + "loss": 0.9564, + "step": 332120 + }, + { + "epoch": 28.75, + "learning_rate": 2.0890519806986623e-06, + "loss": 0.9057, + "step": 332130 + }, + { + "epoch": 28.75, + "learning_rate": 2.087607269785316e-06, + "loss": 0.9097, + "step": 332140 + }, + { + "epoch": 28.75, + "learning_rate": 2.0861625588719695e-06, + "loss": 0.912, + "step": 332150 + }, + { + "epoch": 28.75, + "learning_rate": 2.084717847958624e-06, + "loss": 0.8649, + "step": 332160 + }, + { + "epoch": 28.75, + "learning_rate": 2.083273137045277e-06, + "loss": 0.9594, + "step": 332170 + }, + { + "epoch": 28.75, + "learning_rate": 2.081828426131931e-06, + "loss": 0.9636, + "step": 332180 + }, + { + "epoch": 28.75, + "learning_rate": 2.080383715218585e-06, + "loss": 0.8741, + "step": 332190 + }, + { + "epoch": 28.75, + "learning_rate": 2.0789390043052386e-06, + "loss": 0.965, + "step": 332200 + }, + { + "epoch": 28.76, + "learning_rate": 2.0774942933918925e-06, + "loss": 0.9089, + "step": 332210 + }, + { + "epoch": 28.76, + "learning_rate": 2.076049582478546e-06, + "loss": 0.9304, + "step": 332220 + }, + { + "epoch": 28.76, + "learning_rate": 2.0746048715652e-06, + "loss": 0.8459, + "step": 332230 + }, + { + "epoch": 28.76, + "learning_rate": 2.0731601606518535e-06, + "loss": 0.9277, + "step": 332240 + }, + { + "epoch": 28.76, + "learning_rate": 2.0717154497385073e-06, + "loss": 0.8443, + "step": 332250 + }, + { + "epoch": 28.76, + "learning_rate": 2.070270738825161e-06, + "loss": 0.9384, + "step": 332260 + }, + { + "epoch": 28.76, + "learning_rate": 2.068826027911815e-06, + "loss": 0.9192, + "step": 332270 + }, + { + "epoch": 28.76, + "learning_rate": 2.0673813169984688e-06, + "loss": 0.8656, + "step": 332280 + }, + { + "epoch": 28.76, + "learning_rate": 2.0659366060851226e-06, + "loss": 0.9492, + "step": 332290 + }, + { + "epoch": 28.76, + "learning_rate": 2.0644918951717764e-06, + "loss": 0.9266, + "step": 332300 + }, + { + "epoch": 28.76, + "learning_rate": 2.06304718425843e-06, + "loss": 0.9444, + "step": 332310 + }, + { + "epoch": 28.76, + "learning_rate": 2.0616024733450836e-06, + "loss": 0.9168, + "step": 332320 + }, + { + "epoch": 28.77, + "learning_rate": 2.0601577624317374e-06, + "loss": 0.9225, + "step": 332330 + }, + { + "epoch": 28.77, + "learning_rate": 2.0587130515183913e-06, + "loss": 0.8406, + "step": 332340 + }, + { + "epoch": 28.77, + "learning_rate": 2.057268340605045e-06, + "loss": 0.8766, + "step": 332350 + }, + { + "epoch": 28.77, + "learning_rate": 2.055823629691699e-06, + "loss": 0.8966, + "step": 332360 + }, + { + "epoch": 28.77, + "learning_rate": 2.0543789187783527e-06, + "loss": 0.8019, + "step": 332370 + }, + { + "epoch": 28.77, + "learning_rate": 2.052934207865006e-06, + "loss": 0.95, + "step": 332380 + }, + { + "epoch": 28.77, + "learning_rate": 2.0514894969516603e-06, + "loss": 0.8326, + "step": 332390 + }, + { + "epoch": 28.77, + "learning_rate": 2.0500447860383137e-06, + "loss": 0.893, + "step": 332400 + }, + { + "epoch": 28.77, + "learning_rate": 2.0486000751249676e-06, + "loss": 0.9067, + "step": 332410 + }, + { + "epoch": 28.77, + "learning_rate": 2.0471553642116214e-06, + "loss": 0.9083, + "step": 332420 + }, + { + "epoch": 28.77, + "learning_rate": 2.045710653298275e-06, + "loss": 0.9203, + "step": 332430 + }, + { + "epoch": 28.78, + "learning_rate": 2.044265942384929e-06, + "loss": 0.8666, + "step": 332440 + }, + { + "epoch": 28.78, + "learning_rate": 2.0428212314715824e-06, + "loss": 0.917, + "step": 332450 + }, + { + "epoch": 28.78, + "learning_rate": 2.0413765205582366e-06, + "loss": 0.9367, + "step": 332460 + }, + { + "epoch": 28.78, + "learning_rate": 2.03993180964489e-06, + "loss": 0.9683, + "step": 332470 + }, + { + "epoch": 28.78, + "learning_rate": 2.038487098731544e-06, + "loss": 0.9446, + "step": 332480 + }, + { + "epoch": 28.78, + "learning_rate": 2.0370423878181977e-06, + "loss": 0.8886, + "step": 332490 + }, + { + "epoch": 28.78, + "learning_rate": 2.0355976769048515e-06, + "loss": 0.97, + "step": 332500 + }, + { + "epoch": 28.78, + "learning_rate": 2.0341529659915053e-06, + "loss": 0.9022, + "step": 332510 + }, + { + "epoch": 28.78, + "learning_rate": 2.032708255078159e-06, + "loss": 0.9248, + "step": 332520 + }, + { + "epoch": 28.78, + "learning_rate": 2.031263544164813e-06, + "loss": 0.9353, + "step": 332530 + }, + { + "epoch": 28.78, + "learning_rate": 2.0298188332514663e-06, + "loss": 0.9377, + "step": 332540 + }, + { + "epoch": 28.78, + "learning_rate": 2.02837412233812e-06, + "loss": 0.9241, + "step": 332550 + }, + { + "epoch": 28.79, + "learning_rate": 2.026929411424774e-06, + "loss": 0.905, + "step": 332560 + }, + { + "epoch": 28.79, + "learning_rate": 2.025484700511428e-06, + "loss": 0.901, + "step": 332570 + }, + { + "epoch": 28.79, + "learning_rate": 2.024039989598081e-06, + "loss": 0.9488, + "step": 332580 + }, + { + "epoch": 28.79, + "learning_rate": 2.0225952786847354e-06, + "loss": 0.875, + "step": 332590 + }, + { + "epoch": 28.79, + "learning_rate": 2.0211505677713893e-06, + "loss": 0.8432, + "step": 332600 + }, + { + "epoch": 28.79, + "learning_rate": 2.0197058568580427e-06, + "loss": 0.9162, + "step": 332610 + }, + { + "epoch": 28.79, + "learning_rate": 2.018261145944697e-06, + "loss": 0.876, + "step": 332620 + }, + { + "epoch": 28.79, + "learning_rate": 2.0168164350313503e-06, + "loss": 0.9079, + "step": 332630 + }, + { + "epoch": 28.79, + "learning_rate": 2.015371724118004e-06, + "loss": 0.9106, + "step": 332640 + }, + { + "epoch": 28.79, + "learning_rate": 2.0139270132046575e-06, + "loss": 0.91, + "step": 332650 + }, + { + "epoch": 28.79, + "learning_rate": 2.0124823022913117e-06, + "loss": 0.9645, + "step": 332660 + }, + { + "epoch": 28.79, + "learning_rate": 2.011037591377965e-06, + "loss": 0.9395, + "step": 332670 + }, + { + "epoch": 28.8, + "learning_rate": 2.009592880464619e-06, + "loss": 0.8735, + "step": 332680 + }, + { + "epoch": 28.8, + "learning_rate": 2.008148169551273e-06, + "loss": 0.8919, + "step": 332690 + }, + { + "epoch": 28.8, + "learning_rate": 2.0067034586379266e-06, + "loss": 0.8686, + "step": 332700 + }, + { + "epoch": 28.8, + "learning_rate": 2.0052587477245804e-06, + "loss": 0.8842, + "step": 332710 + }, + { + "epoch": 28.8, + "learning_rate": 2.0038140368112342e-06, + "loss": 0.9377, + "step": 332720 + }, + { + "epoch": 28.8, + "learning_rate": 2.002369325897888e-06, + "loss": 0.9603, + "step": 332730 + }, + { + "epoch": 28.8, + "learning_rate": 2.0009246149845414e-06, + "loss": 0.8748, + "step": 332740 + }, + { + "epoch": 28.8, + "learning_rate": 1.9994799040711957e-06, + "loss": 0.8851, + "step": 332750 + }, + { + "epoch": 28.8, + "learning_rate": 1.998035193157849e-06, + "loss": 0.8951, + "step": 332760 + }, + { + "epoch": 28.8, + "learning_rate": 1.996590482244503e-06, + "loss": 0.9332, + "step": 332770 + }, + { + "epoch": 28.8, + "learning_rate": 1.9951457713311567e-06, + "loss": 0.8951, + "step": 332780 + }, + { + "epoch": 28.81, + "learning_rate": 1.9937010604178105e-06, + "loss": 0.8905, + "step": 332790 + }, + { + "epoch": 28.81, + "learning_rate": 1.9922563495044644e-06, + "loss": 0.8674, + "step": 332800 + }, + { + "epoch": 28.81, + "learning_rate": 1.9908116385911178e-06, + "loss": 0.8602, + "step": 332810 + }, + { + "epoch": 28.81, + "learning_rate": 1.989366927677772e-06, + "loss": 0.8417, + "step": 332820 + }, + { + "epoch": 28.81, + "learning_rate": 1.9879222167644254e-06, + "loss": 0.8867, + "step": 332830 + }, + { + "epoch": 28.81, + "learning_rate": 1.986477505851079e-06, + "loss": 0.8625, + "step": 332840 + }, + { + "epoch": 28.81, + "learning_rate": 1.985032794937733e-06, + "loss": 0.8545, + "step": 332850 + }, + { + "epoch": 28.81, + "learning_rate": 1.983588084024387e-06, + "loss": 0.9201, + "step": 332860 + }, + { + "epoch": 28.81, + "learning_rate": 1.9821433731110407e-06, + "loss": 0.8765, + "step": 332870 + }, + { + "epoch": 28.81, + "learning_rate": 1.980698662197694e-06, + "loss": 0.8813, + "step": 332880 + }, + { + "epoch": 28.81, + "learning_rate": 1.9792539512843483e-06, + "loss": 0.9412, + "step": 332890 + }, + { + "epoch": 28.81, + "learning_rate": 1.9778092403710017e-06, + "loss": 0.8984, + "step": 332900 + }, + { + "epoch": 28.82, + "learning_rate": 1.9763645294576555e-06, + "loss": 0.8378, + "step": 332910 + }, + { + "epoch": 28.82, + "learning_rate": 1.9749198185443093e-06, + "loss": 0.9228, + "step": 332920 + }, + { + "epoch": 28.82, + "learning_rate": 1.973475107630963e-06, + "loss": 0.9031, + "step": 332930 + }, + { + "epoch": 28.82, + "learning_rate": 1.972030396717617e-06, + "loss": 0.8863, + "step": 332940 + }, + { + "epoch": 28.82, + "learning_rate": 1.970585685804271e-06, + "loss": 0.8484, + "step": 332950 + }, + { + "epoch": 28.82, + "learning_rate": 1.9691409748909246e-06, + "loss": 0.8565, + "step": 332960 + }, + { + "epoch": 28.82, + "learning_rate": 1.967696263977578e-06, + "loss": 0.9214, + "step": 332970 + }, + { + "epoch": 28.82, + "learning_rate": 1.966251553064232e-06, + "loss": 0.9194, + "step": 332980 + }, + { + "epoch": 28.82, + "learning_rate": 1.9648068421508856e-06, + "loss": 0.9244, + "step": 332990 + }, + { + "epoch": 28.82, + "learning_rate": 1.9633621312375395e-06, + "loss": 0.9073, + "step": 333000 + }, + { + "epoch": 28.82, + "learning_rate": 1.9619174203241933e-06, + "loss": 0.9405, + "step": 333010 + }, + { + "epoch": 28.83, + "learning_rate": 1.960472709410847e-06, + "loss": 0.8523, + "step": 333020 + }, + { + "epoch": 28.83, + "learning_rate": 1.959027998497501e-06, + "loss": 0.9142, + "step": 333030 + }, + { + "epoch": 28.83, + "learning_rate": 1.9575832875841543e-06, + "loss": 0.9144, + "step": 333040 + }, + { + "epoch": 28.83, + "learning_rate": 1.9561385766708086e-06, + "loss": 0.9334, + "step": 333050 + }, + { + "epoch": 28.83, + "learning_rate": 1.954693865757462e-06, + "loss": 0.8845, + "step": 333060 + }, + { + "epoch": 28.83, + "learning_rate": 1.9532491548441158e-06, + "loss": 0.9034, + "step": 333070 + }, + { + "epoch": 28.83, + "learning_rate": 1.9518044439307696e-06, + "loss": 0.9111, + "step": 333080 + }, + { + "epoch": 28.83, + "learning_rate": 1.9503597330174234e-06, + "loss": 0.8686, + "step": 333090 + }, + { + "epoch": 28.83, + "learning_rate": 1.9489150221040772e-06, + "loss": 0.9004, + "step": 333100 + }, + { + "epoch": 28.83, + "learning_rate": 1.9474703111907306e-06, + "loss": 0.9143, + "step": 333110 + }, + { + "epoch": 28.83, + "learning_rate": 1.946025600277385e-06, + "loss": 0.8883, + "step": 333120 + }, + { + "epoch": 28.83, + "learning_rate": 1.9445808893640383e-06, + "loss": 0.893, + "step": 333130 + }, + { + "epoch": 28.84, + "learning_rate": 1.943136178450692e-06, + "loss": 0.8745, + "step": 333140 + }, + { + "epoch": 28.84, + "learning_rate": 1.941691467537346e-06, + "loss": 0.8468, + "step": 333150 + }, + { + "epoch": 28.84, + "learning_rate": 1.9402467566239997e-06, + "loss": 0.8921, + "step": 333160 + }, + { + "epoch": 28.84, + "learning_rate": 1.938802045710653e-06, + "loss": 0.8996, + "step": 333170 + }, + { + "epoch": 28.84, + "learning_rate": 1.9373573347973073e-06, + "loss": 0.8531, + "step": 333180 + }, + { + "epoch": 28.84, + "learning_rate": 1.935912623883961e-06, + "loss": 0.9081, + "step": 333190 + }, + { + "epoch": 28.84, + "learning_rate": 1.9344679129706146e-06, + "loss": 0.8489, + "step": 333200 + }, + { + "epoch": 28.84, + "learning_rate": 1.9330232020572684e-06, + "loss": 0.8962, + "step": 333210 + }, + { + "epoch": 28.84, + "learning_rate": 1.931578491143922e-06, + "loss": 0.9596, + "step": 333220 + }, + { + "epoch": 28.84, + "learning_rate": 1.930133780230576e-06, + "loss": 0.9144, + "step": 333230 + }, + { + "epoch": 28.84, + "learning_rate": 1.9286890693172294e-06, + "loss": 0.8916, + "step": 333240 + }, + { + "epoch": 28.85, + "learning_rate": 1.9272443584038837e-06, + "loss": 0.9414, + "step": 333250 + }, + { + "epoch": 28.85, + "learning_rate": 1.925799647490537e-06, + "loss": 0.8763, + "step": 333260 + }, + { + "epoch": 28.85, + "learning_rate": 1.924354936577191e-06, + "loss": 0.9202, + "step": 333270 + }, + { + "epoch": 28.85, + "learning_rate": 1.9229102256638447e-06, + "loss": 0.8968, + "step": 333280 + }, + { + "epoch": 28.85, + "learning_rate": 1.9214655147504985e-06, + "loss": 0.9002, + "step": 333290 + }, + { + "epoch": 28.85, + "learning_rate": 1.9200208038371523e-06, + "loss": 0.9018, + "step": 333300 + }, + { + "epoch": 28.85, + "learning_rate": 1.918576092923806e-06, + "loss": 0.8779, + "step": 333310 + }, + { + "epoch": 28.85, + "learning_rate": 1.91713138201046e-06, + "loss": 0.8886, + "step": 333320 + }, + { + "epoch": 28.85, + "learning_rate": 1.9156866710971134e-06, + "loss": 0.8552, + "step": 333330 + }, + { + "epoch": 28.85, + "learning_rate": 1.914241960183767e-06, + "loss": 0.8352, + "step": 333340 + }, + { + "epoch": 28.85, + "learning_rate": 1.912797249270421e-06, + "loss": 0.9015, + "step": 333350 + }, + { + "epoch": 28.85, + "learning_rate": 1.911352538357075e-06, + "loss": 0.7764, + "step": 333360 + }, + { + "epoch": 28.86, + "learning_rate": 1.9099078274437286e-06, + "loss": 0.8305, + "step": 333370 + }, + { + "epoch": 28.86, + "learning_rate": 1.9084631165303824e-06, + "loss": 0.9128, + "step": 333380 + }, + { + "epoch": 28.86, + "learning_rate": 1.907018405617036e-06, + "loss": 0.9193, + "step": 333390 + }, + { + "epoch": 28.86, + "learning_rate": 1.9055736947036899e-06, + "loss": 0.9109, + "step": 333400 + }, + { + "epoch": 28.86, + "learning_rate": 1.904128983790344e-06, + "loss": 0.8819, + "step": 333410 + }, + { + "epoch": 28.86, + "learning_rate": 1.9026842728769975e-06, + "loss": 0.8785, + "step": 333420 + }, + { + "epoch": 28.86, + "learning_rate": 1.9012395619636511e-06, + "loss": 0.8726, + "step": 333430 + }, + { + "epoch": 28.86, + "learning_rate": 1.8997948510503047e-06, + "loss": 0.9282, + "step": 333440 + }, + { + "epoch": 28.86, + "learning_rate": 1.8983501401369587e-06, + "loss": 0.8772, + "step": 333450 + }, + { + "epoch": 28.86, + "learning_rate": 1.8969054292236124e-06, + "loss": 0.8918, + "step": 333460 + }, + { + "epoch": 28.86, + "learning_rate": 1.8954607183102662e-06, + "loss": 0.9505, + "step": 333470 + }, + { + "epoch": 28.87, + "learning_rate": 1.89401600739692e-06, + "loss": 0.8921, + "step": 333480 + }, + { + "epoch": 28.87, + "learning_rate": 1.8925712964835738e-06, + "loss": 0.8834, + "step": 333490 + }, + { + "epoch": 28.87, + "learning_rate": 1.8911265855702274e-06, + "loss": 0.9166, + "step": 333500 + }, + { + "epoch": 28.87, + "learning_rate": 1.8896818746568814e-06, + "loss": 0.8691, + "step": 333510 + }, + { + "epoch": 28.87, + "learning_rate": 1.888237163743535e-06, + "loss": 0.823, + "step": 333520 + }, + { + "epoch": 28.87, + "learning_rate": 1.8867924528301887e-06, + "loss": 0.9453, + "step": 333530 + }, + { + "epoch": 28.87, + "learning_rate": 1.8853477419168427e-06, + "loss": 0.8501, + "step": 333540 + }, + { + "epoch": 28.87, + "learning_rate": 1.8839030310034963e-06, + "loss": 0.9101, + "step": 333550 + }, + { + "epoch": 28.87, + "learning_rate": 1.88245832009015e-06, + "loss": 0.9551, + "step": 333560 + }, + { + "epoch": 28.87, + "learning_rate": 1.8810136091768037e-06, + "loss": 0.9099, + "step": 333570 + }, + { + "epoch": 28.87, + "learning_rate": 1.8795688982634578e-06, + "loss": 0.906, + "step": 333580 + }, + { + "epoch": 28.87, + "learning_rate": 1.8781241873501114e-06, + "loss": 0.8749, + "step": 333590 + }, + { + "epoch": 28.88, + "learning_rate": 1.876679476436765e-06, + "loss": 0.8826, + "step": 333600 + }, + { + "epoch": 28.88, + "learning_rate": 1.875234765523419e-06, + "loss": 0.9179, + "step": 333610 + }, + { + "epoch": 28.88, + "learning_rate": 1.8737900546100726e-06, + "loss": 0.9504, + "step": 333620 + }, + { + "epoch": 28.88, + "learning_rate": 1.8723453436967262e-06, + "loss": 0.9429, + "step": 333630 + }, + { + "epoch": 28.88, + "learning_rate": 1.8709006327833802e-06, + "loss": 0.8746, + "step": 333640 + }, + { + "epoch": 28.88, + "learning_rate": 1.8694559218700338e-06, + "loss": 0.887, + "step": 333650 + }, + { + "epoch": 28.88, + "learning_rate": 1.8680112109566877e-06, + "loss": 0.923, + "step": 333660 + }, + { + "epoch": 28.88, + "learning_rate": 1.8665665000433413e-06, + "loss": 0.9519, + "step": 333670 + }, + { + "epoch": 28.88, + "learning_rate": 1.8651217891299953e-06, + "loss": 0.8575, + "step": 333680 + }, + { + "epoch": 28.88, + "learning_rate": 1.863677078216649e-06, + "loss": 0.8796, + "step": 333690 + }, + { + "epoch": 28.88, + "learning_rate": 1.8622323673033025e-06, + "loss": 0.9477, + "step": 333700 + }, + { + "epoch": 28.88, + "learning_rate": 1.8607876563899565e-06, + "loss": 0.886, + "step": 333710 + }, + { + "epoch": 28.89, + "learning_rate": 1.8593429454766102e-06, + "loss": 0.9608, + "step": 333720 + }, + { + "epoch": 28.89, + "learning_rate": 1.857898234563264e-06, + "loss": 0.9242, + "step": 333730 + }, + { + "epoch": 28.89, + "learning_rate": 1.8564535236499178e-06, + "loss": 1.0057, + "step": 333740 + }, + { + "epoch": 28.89, + "learning_rate": 1.8550088127365716e-06, + "loss": 0.8714, + "step": 333750 + }, + { + "epoch": 28.89, + "learning_rate": 1.8535641018232252e-06, + "loss": 0.9058, + "step": 333760 + }, + { + "epoch": 28.89, + "learning_rate": 1.8521193909098788e-06, + "loss": 0.8383, + "step": 333770 + }, + { + "epoch": 28.89, + "learning_rate": 1.8506746799965329e-06, + "loss": 0.8971, + "step": 333780 + }, + { + "epoch": 28.89, + "learning_rate": 1.8492299690831865e-06, + "loss": 0.9619, + "step": 333790 + }, + { + "epoch": 28.89, + "learning_rate": 1.84778525816984e-06, + "loss": 0.8305, + "step": 333800 + }, + { + "epoch": 28.89, + "learning_rate": 1.846340547256494e-06, + "loss": 0.8905, + "step": 333810 + }, + { + "epoch": 28.89, + "learning_rate": 1.844895836343148e-06, + "loss": 0.9299, + "step": 333820 + }, + { + "epoch": 28.9, + "learning_rate": 1.8434511254298015e-06, + "loss": 0.932, + "step": 333830 + }, + { + "epoch": 28.9, + "learning_rate": 1.8420064145164556e-06, + "loss": 0.8911, + "step": 333840 + }, + { + "epoch": 28.9, + "learning_rate": 1.8405617036031092e-06, + "loss": 0.9457, + "step": 333850 + }, + { + "epoch": 28.9, + "learning_rate": 1.8391169926897628e-06, + "loss": 0.8902, + "step": 333860 + }, + { + "epoch": 28.9, + "learning_rate": 1.8376722817764168e-06, + "loss": 0.8939, + "step": 333870 + }, + { + "epoch": 28.9, + "learning_rate": 1.8362275708630704e-06, + "loss": 0.9529, + "step": 333880 + }, + { + "epoch": 28.9, + "learning_rate": 1.834782859949724e-06, + "loss": 0.9381, + "step": 333890 + }, + { + "epoch": 28.9, + "learning_rate": 1.8333381490363778e-06, + "loss": 0.9144, + "step": 333900 + }, + { + "epoch": 28.9, + "learning_rate": 1.8318934381230319e-06, + "loss": 0.888, + "step": 333910 + }, + { + "epoch": 28.9, + "learning_rate": 1.8304487272096855e-06, + "loss": 0.8984, + "step": 333920 + }, + { + "epoch": 28.9, + "learning_rate": 1.829004016296339e-06, + "loss": 0.857, + "step": 333930 + }, + { + "epoch": 28.9, + "learning_rate": 1.827559305382993e-06, + "loss": 0.8668, + "step": 333940 + }, + { + "epoch": 28.91, + "learning_rate": 1.8261145944696467e-06, + "loss": 0.9258, + "step": 333950 + }, + { + "epoch": 28.91, + "learning_rate": 1.8246698835563003e-06, + "loss": 0.9437, + "step": 333960 + }, + { + "epoch": 28.91, + "learning_rate": 1.8232251726429543e-06, + "loss": 0.8959, + "step": 333970 + }, + { + "epoch": 28.91, + "learning_rate": 1.821780461729608e-06, + "loss": 0.8673, + "step": 333980 + }, + { + "epoch": 28.91, + "learning_rate": 1.8203357508162618e-06, + "loss": 0.9313, + "step": 333990 + }, + { + "epoch": 28.91, + "learning_rate": 1.8188910399029154e-06, + "loss": 0.8915, + "step": 334000 + }, + { + "epoch": 28.91, + "learning_rate": 1.8174463289895694e-06, + "loss": 0.8999, + "step": 334010 + }, + { + "epoch": 28.91, + "learning_rate": 1.816001618076223e-06, + "loss": 0.8822, + "step": 334020 + }, + { + "epoch": 28.91, + "learning_rate": 1.8145569071628766e-06, + "loss": 0.9408, + "step": 334030 + }, + { + "epoch": 28.91, + "learning_rate": 1.8131121962495307e-06, + "loss": 0.9002, + "step": 334040 + }, + { + "epoch": 28.91, + "learning_rate": 1.8116674853361843e-06, + "loss": 0.8798, + "step": 334050 + }, + { + "epoch": 28.92, + "learning_rate": 1.8102227744228379e-06, + "loss": 0.9208, + "step": 334060 + }, + { + "epoch": 28.92, + "learning_rate": 1.808778063509492e-06, + "loss": 0.8663, + "step": 334070 + }, + { + "epoch": 28.92, + "learning_rate": 1.8073333525961457e-06, + "loss": 0.934, + "step": 334080 + }, + { + "epoch": 28.92, + "learning_rate": 1.8058886416827993e-06, + "loss": 0.9107, + "step": 334090 + }, + { + "epoch": 28.92, + "learning_rate": 1.8044439307694534e-06, + "loss": 0.9142, + "step": 334100 + }, + { + "epoch": 28.92, + "learning_rate": 1.802999219856107e-06, + "loss": 0.8846, + "step": 334110 + }, + { + "epoch": 28.92, + "learning_rate": 1.8015545089427606e-06, + "loss": 0.9088, + "step": 334120 + }, + { + "epoch": 28.92, + "learning_rate": 1.8001097980294142e-06, + "loss": 0.9278, + "step": 334130 + }, + { + "epoch": 28.92, + "learning_rate": 1.7986650871160682e-06, + "loss": 0.848, + "step": 334140 + }, + { + "epoch": 28.92, + "learning_rate": 1.7972203762027218e-06, + "loss": 0.8639, + "step": 334150 + }, + { + "epoch": 28.92, + "learning_rate": 1.7957756652893756e-06, + "loss": 0.8842, + "step": 334160 + }, + { + "epoch": 28.92, + "learning_rate": 1.7943309543760297e-06, + "loss": 0.8649, + "step": 334170 + }, + { + "epoch": 28.93, + "learning_rate": 1.7928862434626833e-06, + "loss": 0.9089, + "step": 334180 + }, + { + "epoch": 28.93, + "learning_rate": 1.7914415325493369e-06, + "loss": 0.9247, + "step": 334190 + }, + { + "epoch": 28.93, + "learning_rate": 1.789996821635991e-06, + "loss": 0.8916, + "step": 334200 + }, + { + "epoch": 28.93, + "learning_rate": 1.7885521107226445e-06, + "loss": 0.9301, + "step": 334210 + }, + { + "epoch": 28.93, + "learning_rate": 1.7871073998092981e-06, + "loss": 0.8746, + "step": 334220 + }, + { + "epoch": 28.93, + "learning_rate": 1.785662688895952e-06, + "loss": 0.9285, + "step": 334230 + }, + { + "epoch": 28.93, + "learning_rate": 1.7842179779826058e-06, + "loss": 0.9549, + "step": 334240 + }, + { + "epoch": 28.93, + "learning_rate": 1.7827732670692596e-06, + "loss": 0.8919, + "step": 334250 + }, + { + "epoch": 28.93, + "learning_rate": 1.7813285561559132e-06, + "loss": 0.9134, + "step": 334260 + }, + { + "epoch": 28.93, + "learning_rate": 1.7798838452425672e-06, + "loss": 0.8769, + "step": 334270 + }, + { + "epoch": 28.93, + "learning_rate": 1.7784391343292208e-06, + "loss": 0.8693, + "step": 334280 + }, + { + "epoch": 28.94, + "learning_rate": 1.7769944234158744e-06, + "loss": 0.8919, + "step": 334290 + }, + { + "epoch": 28.94, + "learning_rate": 1.7755497125025284e-06, + "loss": 0.898, + "step": 334300 + }, + { + "epoch": 28.94, + "learning_rate": 1.774105001589182e-06, + "loss": 0.9101, + "step": 334310 + }, + { + "epoch": 28.94, + "learning_rate": 1.7726602906758359e-06, + "loss": 0.9262, + "step": 334320 + }, + { + "epoch": 28.94, + "learning_rate": 1.7712155797624897e-06, + "loss": 0.8903, + "step": 334330 + }, + { + "epoch": 28.94, + "learning_rate": 1.7697708688491435e-06, + "loss": 0.9554, + "step": 334340 + }, + { + "epoch": 28.94, + "learning_rate": 1.7683261579357971e-06, + "loss": 0.8895, + "step": 334350 + }, + { + "epoch": 28.94, + "learning_rate": 1.7668814470224507e-06, + "loss": 0.8592, + "step": 334360 + }, + { + "epoch": 28.94, + "learning_rate": 1.7654367361091048e-06, + "loss": 0.9158, + "step": 334370 + }, + { + "epoch": 28.94, + "learning_rate": 1.7639920251957584e-06, + "loss": 0.863, + "step": 334380 + }, + { + "epoch": 28.94, + "learning_rate": 1.762547314282412e-06, + "loss": 0.836, + "step": 334390 + }, + { + "epoch": 28.94, + "learning_rate": 1.761102603369066e-06, + "loss": 0.9362, + "step": 334400 + }, + { + "epoch": 28.95, + "learning_rate": 1.7596578924557198e-06, + "loss": 0.9129, + "step": 334410 + }, + { + "epoch": 28.95, + "learning_rate": 1.7582131815423734e-06, + "loss": 0.8514, + "step": 334420 + }, + { + "epoch": 28.95, + "learning_rate": 1.7567684706290275e-06, + "loss": 0.94, + "step": 334430 + }, + { + "epoch": 28.95, + "learning_rate": 1.755323759715681e-06, + "loss": 0.9399, + "step": 334440 + }, + { + "epoch": 28.95, + "learning_rate": 1.7538790488023347e-06, + "loss": 0.9358, + "step": 334450 + }, + { + "epoch": 28.95, + "learning_rate": 1.7524343378889883e-06, + "loss": 0.9156, + "step": 334460 + }, + { + "epoch": 28.95, + "learning_rate": 1.7509896269756423e-06, + "loss": 0.9595, + "step": 334470 + }, + { + "epoch": 28.95, + "learning_rate": 1.749544916062296e-06, + "loss": 0.9168, + "step": 334480 + }, + { + "epoch": 28.95, + "learning_rate": 1.7481002051489497e-06, + "loss": 0.9046, + "step": 334490 + }, + { + "epoch": 28.95, + "learning_rate": 1.7466554942356035e-06, + "loss": 0.9332, + "step": 334500 + }, + { + "epoch": 28.95, + "learning_rate": 1.7452107833222574e-06, + "loss": 0.9345, + "step": 334510 + }, + { + "epoch": 28.96, + "learning_rate": 1.743766072408911e-06, + "loss": 0.8612, + "step": 334520 + }, + { + "epoch": 28.96, + "learning_rate": 1.742321361495565e-06, + "loss": 0.874, + "step": 334530 + }, + { + "epoch": 28.96, + "learning_rate": 1.7408766505822186e-06, + "loss": 0.895, + "step": 334540 + }, + { + "epoch": 28.96, + "learning_rate": 1.7394319396688722e-06, + "loss": 0.865, + "step": 334550 + }, + { + "epoch": 28.96, + "learning_rate": 1.7379872287555262e-06, + "loss": 0.8875, + "step": 334560 + }, + { + "epoch": 28.96, + "learning_rate": 1.7365425178421799e-06, + "loss": 0.8559, + "step": 334570 + }, + { + "epoch": 28.96, + "learning_rate": 1.7350978069288337e-06, + "loss": 0.9187, + "step": 334580 + }, + { + "epoch": 28.96, + "learning_rate": 1.7336530960154873e-06, + "loss": 0.8635, + "step": 334590 + }, + { + "epoch": 28.96, + "learning_rate": 1.7322083851021413e-06, + "loss": 0.9381, + "step": 334600 + }, + { + "epoch": 28.96, + "learning_rate": 1.730763674188795e-06, + "loss": 0.8693, + "step": 334610 + }, + { + "epoch": 28.96, + "learning_rate": 1.7293189632754485e-06, + "loss": 0.9039, + "step": 334620 + }, + { + "epoch": 28.96, + "learning_rate": 1.7278742523621026e-06, + "loss": 0.8829, + "step": 334630 + }, + { + "epoch": 28.97, + "learning_rate": 1.7264295414487562e-06, + "loss": 0.9425, + "step": 334640 + }, + { + "epoch": 28.97, + "learning_rate": 1.7249848305354098e-06, + "loss": 0.8482, + "step": 334650 + }, + { + "epoch": 28.97, + "learning_rate": 1.7235401196220638e-06, + "loss": 0.8713, + "step": 334660 + }, + { + "epoch": 28.97, + "learning_rate": 1.7220954087087176e-06, + "loss": 0.8606, + "step": 334670 + }, + { + "epoch": 28.97, + "learning_rate": 1.7206506977953712e-06, + "loss": 0.8432, + "step": 334680 + }, + { + "epoch": 28.97, + "learning_rate": 1.7192059868820248e-06, + "loss": 0.8601, + "step": 334690 + }, + { + "epoch": 28.97, + "learning_rate": 1.7177612759686789e-06, + "loss": 0.9304, + "step": 334700 + }, + { + "epoch": 28.97, + "learning_rate": 1.7163165650553325e-06, + "loss": 0.8666, + "step": 334710 + }, + { + "epoch": 28.97, + "learning_rate": 1.714871854141986e-06, + "loss": 0.8879, + "step": 334720 + }, + { + "epoch": 28.97, + "learning_rate": 1.71342714322864e-06, + "loss": 0.8775, + "step": 334730 + }, + { + "epoch": 28.97, + "learning_rate": 1.7119824323152937e-06, + "loss": 0.9426, + "step": 334740 + }, + { + "epoch": 28.97, + "learning_rate": 1.7105377214019475e-06, + "loss": 0.9923, + "step": 334750 + }, + { + "epoch": 28.98, + "learning_rate": 1.7090930104886016e-06, + "loss": 0.959, + "step": 334760 + }, + { + "epoch": 28.98, + "learning_rate": 1.7076482995752552e-06, + "loss": 0.8914, + "step": 334770 + }, + { + "epoch": 28.98, + "learning_rate": 1.7062035886619088e-06, + "loss": 0.9003, + "step": 334780 + }, + { + "epoch": 28.98, + "learning_rate": 1.7047588777485624e-06, + "loss": 0.9306, + "step": 334790 + }, + { + "epoch": 28.98, + "learning_rate": 1.7033141668352164e-06, + "loss": 0.9357, + "step": 334800 + }, + { + "epoch": 28.98, + "learning_rate": 1.70186945592187e-06, + "loss": 0.9297, + "step": 334810 + }, + { + "epoch": 28.98, + "learning_rate": 1.7004247450085238e-06, + "loss": 0.9004, + "step": 334820 + }, + { + "epoch": 28.98, + "learning_rate": 1.6989800340951777e-06, + "loss": 0.9207, + "step": 334830 + }, + { + "epoch": 28.98, + "learning_rate": 1.6975353231818315e-06, + "loss": 0.9032, + "step": 334840 + }, + { + "epoch": 28.98, + "learning_rate": 1.696090612268485e-06, + "loss": 0.9227, + "step": 334850 + }, + { + "epoch": 28.98, + "learning_rate": 1.6946459013551391e-06, + "loss": 0.9062, + "step": 334860 + }, + { + "epoch": 28.99, + "learning_rate": 1.6932011904417927e-06, + "loss": 0.8415, + "step": 334870 + }, + { + "epoch": 28.99, + "learning_rate": 1.6917564795284463e-06, + "loss": 0.8473, + "step": 334880 + }, + { + "epoch": 28.99, + "learning_rate": 1.6903117686151004e-06, + "loss": 0.9244, + "step": 334890 + }, + { + "epoch": 28.99, + "learning_rate": 1.688867057701754e-06, + "loss": 0.9282, + "step": 334900 + }, + { + "epoch": 28.99, + "learning_rate": 1.6874223467884078e-06, + "loss": 0.9293, + "step": 334910 + }, + { + "epoch": 28.99, + "learning_rate": 1.6859776358750614e-06, + "loss": 0.9374, + "step": 334920 + }, + { + "epoch": 28.99, + "learning_rate": 1.6845329249617154e-06, + "loss": 0.917, + "step": 334930 + }, + { + "epoch": 28.99, + "learning_rate": 1.683088214048369e-06, + "loss": 0.9141, + "step": 334940 + }, + { + "epoch": 28.99, + "learning_rate": 1.6816435031350226e-06, + "loss": 0.9291, + "step": 334950 + }, + { + "epoch": 28.99, + "learning_rate": 1.6801987922216767e-06, + "loss": 0.9446, + "step": 334960 + }, + { + "epoch": 28.99, + "learning_rate": 1.6787540813083303e-06, + "loss": 0.8886, + "step": 334970 + }, + { + "epoch": 28.99, + "learning_rate": 1.6773093703949839e-06, + "loss": 0.9395, + "step": 334980 + }, + { + "epoch": 29.0, + "learning_rate": 1.675864659481638e-06, + "loss": 0.9107, + "step": 334990 + }, + { + "epoch": 29.0, + "learning_rate": 1.6744199485682915e-06, + "loss": 0.8865, + "step": 335000 + }, + { + "epoch": 29.0, + "learning_rate": 1.6729752376549453e-06, + "loss": 0.8888, + "step": 335010 + }, + { + "epoch": 29.0, + "learning_rate": 1.671530526741599e-06, + "loss": 0.9552, + "step": 335020 + }, + { + "epoch": 29.0, + "learning_rate": 1.670085815828253e-06, + "loss": 0.884, + "step": 335030 + }, + { + "epoch": 29.0, + "eval_cer": 0.9929984741584095, + "eval_em": 0, + "eval_f1": 0.12293878992466108, + "eval_loss": 0.9309141635894775, + "eval_runtime": 899.2971, + "eval_samples_per_second": 5.71, + "eval_steps_per_second": 0.714, + "eval_wer": 0.965904952578328, + "step": 335039 + }, + { + "epoch": 29.0, + "learning_rate": 1.6686411049149066e-06, + "loss": 0.9328, + "step": 335040 + }, + { + "epoch": 29.0, + "learning_rate": 1.6671963940015602e-06, + "loss": 0.8858, + "step": 335050 + }, + { + "epoch": 29.0, + "learning_rate": 1.6657516830882142e-06, + "loss": 0.9093, + "step": 335060 + }, + { + "epoch": 29.0, + "learning_rate": 1.6643069721748678e-06, + "loss": 0.9001, + "step": 335070 + }, + { + "epoch": 29.0, + "learning_rate": 1.6628622612615216e-06, + "loss": 0.9284, + "step": 335080 + }, + { + "epoch": 29.0, + "learning_rate": 1.6614175503481755e-06, + "loss": 0.9137, + "step": 335090 + }, + { + "epoch": 29.01, + "learning_rate": 1.6599728394348293e-06, + "loss": 0.8876, + "step": 335100 + }, + { + "epoch": 29.01, + "learning_rate": 1.6585281285214829e-06, + "loss": 0.8998, + "step": 335110 + }, + { + "epoch": 29.01, + "learning_rate": 1.657083417608137e-06, + "loss": 0.8853, + "step": 335120 + }, + { + "epoch": 29.01, + "learning_rate": 1.6556387066947905e-06, + "loss": 0.9275, + "step": 335130 + }, + { + "epoch": 29.01, + "learning_rate": 1.6541939957814441e-06, + "loss": 0.891, + "step": 335140 + }, + { + "epoch": 29.01, + "learning_rate": 1.6527492848680977e-06, + "loss": 0.9077, + "step": 335150 + }, + { + "epoch": 29.01, + "learning_rate": 1.6513045739547518e-06, + "loss": 0.8612, + "step": 335160 + }, + { + "epoch": 29.01, + "learning_rate": 1.6498598630414056e-06, + "loss": 0.9446, + "step": 335170 + }, + { + "epoch": 29.01, + "learning_rate": 1.6484151521280592e-06, + "loss": 0.9381, + "step": 335180 + }, + { + "epoch": 29.01, + "learning_rate": 1.6469704412147132e-06, + "loss": 0.9514, + "step": 335190 + }, + { + "epoch": 29.01, + "learning_rate": 1.6455257303013668e-06, + "loss": 0.8579, + "step": 335200 + }, + { + "epoch": 29.01, + "learning_rate": 1.6440810193880204e-06, + "loss": 0.9068, + "step": 335210 + }, + { + "epoch": 29.02, + "learning_rate": 1.6426363084746745e-06, + "loss": 0.8506, + "step": 335220 + }, + { + "epoch": 29.02, + "learning_rate": 1.641191597561328e-06, + "loss": 0.8682, + "step": 335230 + }, + { + "epoch": 29.02, + "learning_rate": 1.6397468866479817e-06, + "loss": 0.9406, + "step": 335240 + }, + { + "epoch": 29.02, + "learning_rate": 1.6383021757346355e-06, + "loss": 0.8637, + "step": 335250 + }, + { + "epoch": 29.02, + "learning_rate": 1.6368574648212895e-06, + "loss": 0.8908, + "step": 335260 + }, + { + "epoch": 29.02, + "learning_rate": 1.6354127539079431e-06, + "loss": 0.933, + "step": 335270 + }, + { + "epoch": 29.02, + "learning_rate": 1.6339680429945967e-06, + "loss": 0.8511, + "step": 335280 + }, + { + "epoch": 29.02, + "learning_rate": 1.6325233320812508e-06, + "loss": 0.9411, + "step": 335290 + }, + { + "epoch": 29.02, + "learning_rate": 1.6310786211679044e-06, + "loss": 0.8953, + "step": 335300 + }, + { + "epoch": 29.02, + "learning_rate": 1.629633910254558e-06, + "loss": 0.8629, + "step": 335310 + }, + { + "epoch": 29.02, + "learning_rate": 1.628189199341212e-06, + "loss": 0.8791, + "step": 335320 + }, + { + "epoch": 29.03, + "learning_rate": 1.6267444884278656e-06, + "loss": 0.8938, + "step": 335330 + }, + { + "epoch": 29.03, + "learning_rate": 1.6252997775145194e-06, + "loss": 0.8749, + "step": 335340 + }, + { + "epoch": 29.03, + "learning_rate": 1.6238550666011735e-06, + "loss": 0.9276, + "step": 335350 + }, + { + "epoch": 29.03, + "learning_rate": 1.622410355687827e-06, + "loss": 0.9595, + "step": 335360 + }, + { + "epoch": 29.03, + "learning_rate": 1.6209656447744807e-06, + "loss": 0.9151, + "step": 335370 + }, + { + "epoch": 29.03, + "learning_rate": 1.6195209338611343e-06, + "loss": 0.9085, + "step": 335380 + }, + { + "epoch": 29.03, + "learning_rate": 1.6180762229477883e-06, + "loss": 0.9335, + "step": 335390 + }, + { + "epoch": 29.03, + "learning_rate": 1.616631512034442e-06, + "loss": 0.8932, + "step": 335400 + }, + { + "epoch": 29.03, + "learning_rate": 1.6151868011210957e-06, + "loss": 0.9501, + "step": 335410 + }, + { + "epoch": 29.03, + "learning_rate": 1.6137420902077496e-06, + "loss": 0.9227, + "step": 335420 + }, + { + "epoch": 29.03, + "learning_rate": 1.6122973792944034e-06, + "loss": 0.9162, + "step": 335430 + }, + { + "epoch": 29.03, + "learning_rate": 1.610852668381057e-06, + "loss": 0.8948, + "step": 335440 + }, + { + "epoch": 29.04, + "learning_rate": 1.609407957467711e-06, + "loss": 0.9408, + "step": 335450 + }, + { + "epoch": 29.04, + "learning_rate": 1.6079632465543646e-06, + "loss": 0.868, + "step": 335460 + }, + { + "epoch": 29.04, + "learning_rate": 1.6065185356410182e-06, + "loss": 0.8673, + "step": 335470 + }, + { + "epoch": 29.04, + "learning_rate": 1.6050738247276718e-06, + "loss": 0.8742, + "step": 335480 + }, + { + "epoch": 29.04, + "learning_rate": 1.6036291138143259e-06, + "loss": 0.8798, + "step": 335490 + }, + { + "epoch": 29.04, + "learning_rate": 1.6021844029009795e-06, + "loss": 0.9253, + "step": 335500 + }, + { + "epoch": 29.04, + "learning_rate": 1.6007396919876333e-06, + "loss": 0.8794, + "step": 335510 + }, + { + "epoch": 29.04, + "learning_rate": 1.5992949810742873e-06, + "loss": 0.9089, + "step": 335520 + }, + { + "epoch": 29.04, + "learning_rate": 1.597850270160941e-06, + "loss": 0.8951, + "step": 335530 + }, + { + "epoch": 29.04, + "learning_rate": 1.5964055592475945e-06, + "loss": 0.8476, + "step": 335540 + }, + { + "epoch": 29.04, + "learning_rate": 1.5949608483342486e-06, + "loss": 0.8382, + "step": 335550 + }, + { + "epoch": 29.05, + "learning_rate": 1.5935161374209022e-06, + "loss": 0.9008, + "step": 335560 + }, + { + "epoch": 29.05, + "learning_rate": 1.5920714265075558e-06, + "loss": 0.9479, + "step": 335570 + }, + { + "epoch": 29.05, + "learning_rate": 1.5906267155942096e-06, + "loss": 0.9063, + "step": 335580 + }, + { + "epoch": 29.05, + "learning_rate": 1.5891820046808634e-06, + "loss": 0.8977, + "step": 335590 + }, + { + "epoch": 29.05, + "learning_rate": 1.5877372937675172e-06, + "loss": 0.856, + "step": 335600 + }, + { + "epoch": 29.05, + "learning_rate": 1.5862925828541708e-06, + "loss": 0.8367, + "step": 335610 + }, + { + "epoch": 29.05, + "learning_rate": 1.5848478719408249e-06, + "loss": 0.9467, + "step": 335620 + }, + { + "epoch": 29.05, + "learning_rate": 1.5834031610274785e-06, + "loss": 0.8869, + "step": 335630 + }, + { + "epoch": 29.05, + "learning_rate": 1.581958450114132e-06, + "loss": 0.9019, + "step": 335640 + }, + { + "epoch": 29.05, + "learning_rate": 1.5805137392007861e-06, + "loss": 0.8829, + "step": 335650 + }, + { + "epoch": 29.05, + "learning_rate": 1.5790690282874397e-06, + "loss": 0.8815, + "step": 335660 + }, + { + "epoch": 29.05, + "learning_rate": 1.5776243173740935e-06, + "loss": 0.8867, + "step": 335670 + }, + { + "epoch": 29.06, + "learning_rate": 1.5761796064607474e-06, + "loss": 0.8321, + "step": 335680 + }, + { + "epoch": 29.06, + "learning_rate": 1.5747348955474012e-06, + "loss": 0.9241, + "step": 335690 + }, + { + "epoch": 29.06, + "learning_rate": 1.5732901846340548e-06, + "loss": 0.9487, + "step": 335700 + }, + { + "epoch": 29.06, + "learning_rate": 1.5718454737207084e-06, + "loss": 0.8976, + "step": 335710 + }, + { + "epoch": 29.06, + "learning_rate": 1.5704007628073624e-06, + "loss": 0.915, + "step": 335720 + }, + { + "epoch": 29.06, + "learning_rate": 1.568956051894016e-06, + "loss": 0.9267, + "step": 335730 + }, + { + "epoch": 29.06, + "learning_rate": 1.5675113409806696e-06, + "loss": 0.9276, + "step": 335740 + }, + { + "epoch": 29.06, + "learning_rate": 1.5660666300673237e-06, + "loss": 0.848, + "step": 335750 + }, + { + "epoch": 29.06, + "learning_rate": 1.5646219191539775e-06, + "loss": 0.9466, + "step": 335760 + }, + { + "epoch": 29.06, + "learning_rate": 1.563177208240631e-06, + "loss": 0.9528, + "step": 335770 + }, + { + "epoch": 29.06, + "learning_rate": 1.561732497327285e-06, + "loss": 0.8699, + "step": 335780 + }, + { + "epoch": 29.06, + "learning_rate": 1.5602877864139387e-06, + "loss": 0.9252, + "step": 335790 + }, + { + "epoch": 29.07, + "learning_rate": 1.5588430755005923e-06, + "loss": 0.897, + "step": 335800 + }, + { + "epoch": 29.07, + "learning_rate": 1.5573983645872461e-06, + "loss": 0.824, + "step": 335810 + }, + { + "epoch": 29.07, + "learning_rate": 1.5559536536739e-06, + "loss": 0.9283, + "step": 335820 + }, + { + "epoch": 29.07, + "learning_rate": 1.5545089427605536e-06, + "loss": 0.909, + "step": 335830 + }, + { + "epoch": 29.07, + "learning_rate": 1.5530642318472074e-06, + "loss": 0.9088, + "step": 335840 + }, + { + "epoch": 29.07, + "learning_rate": 1.5516195209338612e-06, + "loss": 0.9267, + "step": 335850 + }, + { + "epoch": 29.07, + "learning_rate": 1.550174810020515e-06, + "loss": 0.8843, + "step": 335860 + }, + { + "epoch": 29.07, + "learning_rate": 1.5487300991071688e-06, + "loss": 0.8945, + "step": 335870 + }, + { + "epoch": 29.07, + "learning_rate": 1.5472853881938225e-06, + "loss": 0.864, + "step": 335880 + }, + { + "epoch": 29.07, + "learning_rate": 1.5458406772804763e-06, + "loss": 0.8932, + "step": 335890 + }, + { + "epoch": 29.07, + "learning_rate": 1.5443959663671299e-06, + "loss": 0.8779, + "step": 335900 + }, + { + "epoch": 29.08, + "learning_rate": 1.5429512554537837e-06, + "loss": 0.9053, + "step": 335910 + }, + { + "epoch": 29.08, + "learning_rate": 1.5415065445404375e-06, + "loss": 0.8942, + "step": 335920 + }, + { + "epoch": 29.08, + "learning_rate": 1.5400618336270913e-06, + "loss": 0.8958, + "step": 335930 + }, + { + "epoch": 29.08, + "learning_rate": 1.5386171227137452e-06, + "loss": 0.9153, + "step": 335940 + }, + { + "epoch": 29.08, + "learning_rate": 1.5371724118003988e-06, + "loss": 0.9073, + "step": 335950 + }, + { + "epoch": 29.08, + "learning_rate": 1.5357277008870526e-06, + "loss": 0.8369, + "step": 335960 + }, + { + "epoch": 29.08, + "learning_rate": 1.5342829899737064e-06, + "loss": 0.8447, + "step": 335970 + }, + { + "epoch": 29.08, + "learning_rate": 1.53283827906036e-06, + "loss": 0.9333, + "step": 335980 + }, + { + "epoch": 29.08, + "learning_rate": 1.5313935681470138e-06, + "loss": 0.9067, + "step": 335990 + }, + { + "epoch": 29.08, + "learning_rate": 1.5299488572336676e-06, + "loss": 0.969, + "step": 336000 + }, + { + "epoch": 29.08, + "learning_rate": 1.5285041463203215e-06, + "loss": 0.9131, + "step": 336010 + }, + { + "epoch": 29.08, + "learning_rate": 1.5270594354069753e-06, + "loss": 0.8869, + "step": 336020 + }, + { + "epoch": 29.09, + "learning_rate": 1.5256147244936289e-06, + "loss": 0.8906, + "step": 336030 + }, + { + "epoch": 29.09, + "learning_rate": 1.5241700135802827e-06, + "loss": 0.8822, + "step": 336040 + }, + { + "epoch": 29.09, + "learning_rate": 1.5227253026669365e-06, + "loss": 0.9322, + "step": 336050 + }, + { + "epoch": 29.09, + "learning_rate": 1.5212805917535901e-06, + "loss": 0.9144, + "step": 336060 + }, + { + "epoch": 29.09, + "learning_rate": 1.519835880840244e-06, + "loss": 0.8813, + "step": 336070 + }, + { + "epoch": 29.09, + "learning_rate": 1.5183911699268976e-06, + "loss": 0.9061, + "step": 336080 + }, + { + "epoch": 29.09, + "learning_rate": 1.5169464590135514e-06, + "loss": 0.8599, + "step": 336090 + }, + { + "epoch": 29.09, + "learning_rate": 1.5155017481002054e-06, + "loss": 0.878, + "step": 336100 + }, + { + "epoch": 29.09, + "learning_rate": 1.514057037186859e-06, + "loss": 0.8844, + "step": 336110 + }, + { + "epoch": 29.09, + "learning_rate": 1.5126123262735128e-06, + "loss": 0.9342, + "step": 336120 + }, + { + "epoch": 29.09, + "learning_rate": 1.5111676153601664e-06, + "loss": 0.9623, + "step": 336130 + }, + { + "epoch": 29.1, + "learning_rate": 1.5097229044468203e-06, + "loss": 0.8883, + "step": 336140 + }, + { + "epoch": 29.1, + "learning_rate": 1.508278193533474e-06, + "loss": 0.8787, + "step": 336150 + }, + { + "epoch": 29.1, + "learning_rate": 1.5068334826201277e-06, + "loss": 0.9959, + "step": 336160 + }, + { + "epoch": 29.1, + "learning_rate": 1.5053887717067815e-06, + "loss": 0.8415, + "step": 336170 + }, + { + "epoch": 29.1, + "learning_rate": 1.5039440607934353e-06, + "loss": 0.928, + "step": 336180 + }, + { + "epoch": 29.1, + "learning_rate": 1.5024993498800891e-06, + "loss": 0.8975, + "step": 336190 + }, + { + "epoch": 29.1, + "learning_rate": 1.501054638966743e-06, + "loss": 0.8024, + "step": 336200 + }, + { + "epoch": 29.1, + "learning_rate": 1.4996099280533966e-06, + "loss": 0.8097, + "step": 336210 + }, + { + "epoch": 29.1, + "learning_rate": 1.4981652171400504e-06, + "loss": 0.8696, + "step": 336220 + }, + { + "epoch": 29.1, + "learning_rate": 1.496720506226704e-06, + "loss": 0.8553, + "step": 336230 + }, + { + "epoch": 29.1, + "learning_rate": 1.4952757953133578e-06, + "loss": 0.8519, + "step": 336240 + }, + { + "epoch": 29.1, + "learning_rate": 1.4938310844000116e-06, + "loss": 0.9408, + "step": 336250 + }, + { + "epoch": 29.11, + "learning_rate": 1.4923863734866654e-06, + "loss": 0.9058, + "step": 336260 + }, + { + "epoch": 29.11, + "learning_rate": 1.4909416625733193e-06, + "loss": 0.8845, + "step": 336270 + }, + { + "epoch": 29.11, + "learning_rate": 1.489496951659973e-06, + "loss": 0.9329, + "step": 336280 + }, + { + "epoch": 29.11, + "learning_rate": 1.4880522407466267e-06, + "loss": 0.9515, + "step": 336290 + }, + { + "epoch": 29.11, + "learning_rate": 1.4866075298332805e-06, + "loss": 0.9395, + "step": 336300 + }, + { + "epoch": 29.11, + "learning_rate": 1.485162818919934e-06, + "loss": 0.8903, + "step": 336310 + }, + { + "epoch": 29.11, + "learning_rate": 1.483718108006588e-06, + "loss": 0.8945, + "step": 336320 + }, + { + "epoch": 29.11, + "learning_rate": 1.4822733970932417e-06, + "loss": 0.9115, + "step": 336330 + }, + { + "epoch": 29.11, + "learning_rate": 1.4808286861798953e-06, + "loss": 0.9013, + "step": 336340 + }, + { + "epoch": 29.11, + "learning_rate": 1.4793839752665494e-06, + "loss": 0.9152, + "step": 336350 + }, + { + "epoch": 29.11, + "learning_rate": 1.477939264353203e-06, + "loss": 0.8775, + "step": 336360 + }, + { + "epoch": 29.12, + "learning_rate": 1.4764945534398568e-06, + "loss": 0.8752, + "step": 336370 + }, + { + "epoch": 29.12, + "learning_rate": 1.4750498425265106e-06, + "loss": 0.8636, + "step": 336380 + }, + { + "epoch": 29.12, + "learning_rate": 1.4736051316131642e-06, + "loss": 0.8418, + "step": 336390 + }, + { + "epoch": 29.12, + "learning_rate": 1.472160420699818e-06, + "loss": 0.8429, + "step": 336400 + }, + { + "epoch": 29.12, + "learning_rate": 1.4707157097864717e-06, + "loss": 0.9423, + "step": 336410 + }, + { + "epoch": 29.12, + "learning_rate": 1.4692709988731255e-06, + "loss": 0.8751, + "step": 336420 + }, + { + "epoch": 29.12, + "learning_rate": 1.4678262879597793e-06, + "loss": 0.9259, + "step": 336430 + }, + { + "epoch": 29.12, + "learning_rate": 1.4663815770464331e-06, + "loss": 0.8764, + "step": 336440 + }, + { + "epoch": 29.12, + "learning_rate": 1.464936866133087e-06, + "loss": 0.9498, + "step": 336450 + }, + { + "epoch": 29.12, + "learning_rate": 1.4634921552197405e-06, + "loss": 0.9135, + "step": 336460 + }, + { + "epoch": 29.12, + "learning_rate": 1.4620474443063944e-06, + "loss": 0.9353, + "step": 336470 + }, + { + "epoch": 29.12, + "learning_rate": 1.4606027333930482e-06, + "loss": 0.9652, + "step": 336480 + }, + { + "epoch": 29.13, + "learning_rate": 1.4591580224797018e-06, + "loss": 0.9202, + "step": 336490 + }, + { + "epoch": 29.13, + "learning_rate": 1.4577133115663556e-06, + "loss": 0.8864, + "step": 336500 + }, + { + "epoch": 29.13, + "learning_rate": 1.4562686006530094e-06, + "loss": 0.913, + "step": 336510 + }, + { + "epoch": 29.13, + "learning_rate": 1.4548238897396632e-06, + "loss": 0.897, + "step": 336520 + }, + { + "epoch": 29.13, + "learning_rate": 1.453379178826317e-06, + "loss": 0.9095, + "step": 336530 + }, + { + "epoch": 29.13, + "learning_rate": 1.4519344679129707e-06, + "loss": 0.8687, + "step": 336540 + }, + { + "epoch": 29.13, + "learning_rate": 1.4504897569996245e-06, + "loss": 0.8871, + "step": 336550 + }, + { + "epoch": 29.13, + "learning_rate": 1.4490450460862783e-06, + "loss": 0.8774, + "step": 336560 + }, + { + "epoch": 29.13, + "learning_rate": 1.447600335172932e-06, + "loss": 0.9317, + "step": 336570 + }, + { + "epoch": 29.13, + "learning_rate": 1.4461556242595857e-06, + "loss": 0.9072, + "step": 336580 + }, + { + "epoch": 29.13, + "learning_rate": 1.4447109133462393e-06, + "loss": 0.8796, + "step": 336590 + }, + { + "epoch": 29.14, + "learning_rate": 1.4432662024328934e-06, + "loss": 0.922, + "step": 336600 + }, + { + "epoch": 29.14, + "learning_rate": 1.4418214915195472e-06, + "loss": 0.9343, + "step": 336610 + }, + { + "epoch": 29.14, + "learning_rate": 1.4403767806062008e-06, + "loss": 0.9212, + "step": 336620 + }, + { + "epoch": 29.14, + "learning_rate": 1.4389320696928546e-06, + "loss": 0.9722, + "step": 336630 + }, + { + "epoch": 29.14, + "learning_rate": 1.4374873587795082e-06, + "loss": 0.8915, + "step": 336640 + }, + { + "epoch": 29.14, + "learning_rate": 1.436042647866162e-06, + "loss": 0.8956, + "step": 336650 + }, + { + "epoch": 29.14, + "learning_rate": 1.4345979369528158e-06, + "loss": 0.7663, + "step": 336660 + }, + { + "epoch": 29.14, + "learning_rate": 1.4331532260394695e-06, + "loss": 0.9326, + "step": 336670 + }, + { + "epoch": 29.14, + "learning_rate": 1.4317085151261233e-06, + "loss": 0.8808, + "step": 336680 + }, + { + "epoch": 29.14, + "learning_rate": 1.430263804212777e-06, + "loss": 0.8549, + "step": 336690 + }, + { + "epoch": 29.14, + "learning_rate": 1.428819093299431e-06, + "loss": 0.8872, + "step": 336700 + }, + { + "epoch": 29.14, + "learning_rate": 1.4273743823860847e-06, + "loss": 0.9046, + "step": 336710 + }, + { + "epoch": 29.15, + "learning_rate": 1.4259296714727383e-06, + "loss": 0.8613, + "step": 336720 + }, + { + "epoch": 29.15, + "learning_rate": 1.4244849605593922e-06, + "loss": 0.8974, + "step": 336730 + }, + { + "epoch": 29.15, + "learning_rate": 1.4230402496460458e-06, + "loss": 0.8948, + "step": 336740 + }, + { + "epoch": 29.15, + "learning_rate": 1.4215955387326996e-06, + "loss": 0.8674, + "step": 336750 + }, + { + "epoch": 29.15, + "learning_rate": 1.4201508278193534e-06, + "loss": 0.8818, + "step": 336760 + }, + { + "epoch": 29.15, + "learning_rate": 1.4187061169060072e-06, + "loss": 0.9395, + "step": 336770 + }, + { + "epoch": 29.15, + "learning_rate": 1.417261405992661e-06, + "loss": 0.9146, + "step": 336780 + }, + { + "epoch": 29.15, + "learning_rate": 1.4158166950793149e-06, + "loss": 0.88, + "step": 336790 + }, + { + "epoch": 29.15, + "learning_rate": 1.4143719841659685e-06, + "loss": 0.9409, + "step": 336800 + }, + { + "epoch": 29.15, + "learning_rate": 1.4129272732526223e-06, + "loss": 0.8697, + "step": 336810 + }, + { + "epoch": 29.15, + "learning_rate": 1.4114825623392759e-06, + "loss": 0.9155, + "step": 336820 + }, + { + "epoch": 29.15, + "learning_rate": 1.4100378514259297e-06, + "loss": 0.9165, + "step": 336830 + }, + { + "epoch": 29.16, + "learning_rate": 1.4085931405125835e-06, + "loss": 0.92, + "step": 336840 + }, + { + "epoch": 29.16, + "learning_rate": 1.4071484295992373e-06, + "loss": 0.8258, + "step": 336850 + }, + { + "epoch": 29.16, + "learning_rate": 1.4057037186858912e-06, + "loss": 0.8958, + "step": 336860 + }, + { + "epoch": 29.16, + "learning_rate": 1.4042590077725448e-06, + "loss": 0.9263, + "step": 336870 + }, + { + "epoch": 29.16, + "learning_rate": 1.4028142968591986e-06, + "loss": 0.9034, + "step": 336880 + }, + { + "epoch": 29.16, + "learning_rate": 1.4013695859458524e-06, + "loss": 0.9294, + "step": 336890 + }, + { + "epoch": 29.16, + "learning_rate": 1.399924875032506e-06, + "loss": 0.9102, + "step": 336900 + }, + { + "epoch": 29.16, + "learning_rate": 1.3984801641191598e-06, + "loss": 0.8607, + "step": 336910 + }, + { + "epoch": 29.16, + "learning_rate": 1.3970354532058134e-06, + "loss": 0.8523, + "step": 336920 + }, + { + "epoch": 29.16, + "learning_rate": 1.3955907422924673e-06, + "loss": 0.9222, + "step": 336930 + }, + { + "epoch": 29.16, + "learning_rate": 1.394146031379121e-06, + "loss": 0.9104, + "step": 336940 + }, + { + "epoch": 29.17, + "learning_rate": 1.3927013204657749e-06, + "loss": 0.9005, + "step": 336950 + }, + { + "epoch": 29.17, + "learning_rate": 1.3912566095524287e-06, + "loss": 0.8782, + "step": 336960 + }, + { + "epoch": 29.17, + "learning_rate": 1.3898118986390823e-06, + "loss": 0.9077, + "step": 336970 + }, + { + "epoch": 29.17, + "learning_rate": 1.3883671877257361e-06, + "loss": 0.8655, + "step": 336980 + }, + { + "epoch": 29.17, + "learning_rate": 1.38692247681239e-06, + "loss": 0.8938, + "step": 336990 + }, + { + "epoch": 29.17, + "learning_rate": 1.3854777658990436e-06, + "loss": 0.9217, + "step": 337000 + }, + { + "epoch": 29.17, + "learning_rate": 1.3840330549856974e-06, + "loss": 0.9282, + "step": 337010 + }, + { + "epoch": 29.17, + "learning_rate": 1.3825883440723512e-06, + "loss": 0.9222, + "step": 337020 + }, + { + "epoch": 29.17, + "learning_rate": 1.381143633159005e-06, + "loss": 0.9057, + "step": 337030 + }, + { + "epoch": 29.17, + "learning_rate": 1.3796989222456588e-06, + "loss": 0.9075, + "step": 337040 + }, + { + "epoch": 29.17, + "learning_rate": 1.3782542113323124e-06, + "loss": 0.8689, + "step": 337050 + }, + { + "epoch": 29.17, + "learning_rate": 1.3768095004189663e-06, + "loss": 0.878, + "step": 337060 + }, + { + "epoch": 29.18, + "learning_rate": 1.37536478950562e-06, + "loss": 0.8346, + "step": 337070 + }, + { + "epoch": 29.18, + "learning_rate": 1.3739200785922737e-06, + "loss": 0.9264, + "step": 337080 + }, + { + "epoch": 29.18, + "learning_rate": 1.3724753676789275e-06, + "loss": 0.8689, + "step": 337090 + }, + { + "epoch": 29.18, + "learning_rate": 1.3710306567655813e-06, + "loss": 0.848, + "step": 337100 + }, + { + "epoch": 29.18, + "learning_rate": 1.3695859458522351e-06, + "loss": 0.8715, + "step": 337110 + }, + { + "epoch": 29.18, + "learning_rate": 1.368141234938889e-06, + "loss": 0.8574, + "step": 337120 + }, + { + "epoch": 29.18, + "learning_rate": 1.3666965240255426e-06, + "loss": 0.8784, + "step": 337130 + }, + { + "epoch": 29.18, + "learning_rate": 1.3652518131121964e-06, + "loss": 0.9291, + "step": 337140 + }, + { + "epoch": 29.18, + "learning_rate": 1.36380710219885e-06, + "loss": 0.9456, + "step": 337150 + }, + { + "epoch": 29.18, + "learning_rate": 1.3623623912855038e-06, + "loss": 0.9101, + "step": 337160 + }, + { + "epoch": 29.18, + "learning_rate": 1.3609176803721576e-06, + "loss": 0.8596, + "step": 337170 + }, + { + "epoch": 29.19, + "learning_rate": 1.3594729694588112e-06, + "loss": 0.8926, + "step": 337180 + }, + { + "epoch": 29.19, + "learning_rate": 1.358028258545465e-06, + "loss": 0.8611, + "step": 337190 + }, + { + "epoch": 29.19, + "learning_rate": 1.3565835476321189e-06, + "loss": 0.8969, + "step": 337200 + }, + { + "epoch": 29.19, + "learning_rate": 1.3551388367187727e-06, + "loss": 0.8551, + "step": 337210 + }, + { + "epoch": 29.19, + "learning_rate": 1.3536941258054265e-06, + "loss": 0.9547, + "step": 337220 + }, + { + "epoch": 29.19, + "learning_rate": 1.3522494148920801e-06, + "loss": 0.9222, + "step": 337230 + }, + { + "epoch": 29.19, + "learning_rate": 1.350804703978734e-06, + "loss": 0.9054, + "step": 337240 + }, + { + "epoch": 29.19, + "learning_rate": 1.3493599930653875e-06, + "loss": 0.8946, + "step": 337250 + }, + { + "epoch": 29.19, + "learning_rate": 1.3479152821520414e-06, + "loss": 0.9246, + "step": 337260 + }, + { + "epoch": 29.19, + "learning_rate": 1.3464705712386952e-06, + "loss": 0.9136, + "step": 337270 + }, + { + "epoch": 29.19, + "learning_rate": 1.345025860325349e-06, + "loss": 0.8955, + "step": 337280 + }, + { + "epoch": 29.19, + "learning_rate": 1.3435811494120028e-06, + "loss": 0.9146, + "step": 337290 + }, + { + "epoch": 29.2, + "learning_rate": 1.3421364384986564e-06, + "loss": 0.896, + "step": 337300 + }, + { + "epoch": 29.2, + "learning_rate": 1.3406917275853102e-06, + "loss": 0.9078, + "step": 337310 + }, + { + "epoch": 29.2, + "learning_rate": 1.339247016671964e-06, + "loss": 0.8957, + "step": 337320 + }, + { + "epoch": 29.2, + "learning_rate": 1.3378023057586177e-06, + "loss": 0.9201, + "step": 337330 + }, + { + "epoch": 29.2, + "learning_rate": 1.3363575948452715e-06, + "loss": 0.8976, + "step": 337340 + }, + { + "epoch": 29.2, + "learning_rate": 1.3349128839319253e-06, + "loss": 0.8979, + "step": 337350 + }, + { + "epoch": 29.2, + "learning_rate": 1.3334681730185791e-06, + "loss": 0.8505, + "step": 337360 + }, + { + "epoch": 29.2, + "learning_rate": 1.332023462105233e-06, + "loss": 0.8487, + "step": 337370 + }, + { + "epoch": 29.2, + "learning_rate": 1.3305787511918865e-06, + "loss": 0.9139, + "step": 337380 + }, + { + "epoch": 29.2, + "learning_rate": 1.3291340402785404e-06, + "loss": 0.8901, + "step": 337390 + }, + { + "epoch": 29.2, + "learning_rate": 1.3276893293651942e-06, + "loss": 0.9055, + "step": 337400 + }, + { + "epoch": 29.21, + "learning_rate": 1.3262446184518478e-06, + "loss": 0.907, + "step": 337410 + }, + { + "epoch": 29.21, + "learning_rate": 1.3247999075385016e-06, + "loss": 0.9042, + "step": 337420 + }, + { + "epoch": 29.21, + "learning_rate": 1.3233551966251552e-06, + "loss": 0.9176, + "step": 337430 + }, + { + "epoch": 29.21, + "learning_rate": 1.321910485711809e-06, + "loss": 0.9015, + "step": 337440 + }, + { + "epoch": 29.21, + "learning_rate": 1.320465774798463e-06, + "loss": 0.9017, + "step": 337450 + }, + { + "epoch": 29.21, + "learning_rate": 1.3190210638851167e-06, + "loss": 0.8997, + "step": 337460 + }, + { + "epoch": 29.21, + "learning_rate": 1.3175763529717705e-06, + "loss": 0.8821, + "step": 337470 + }, + { + "epoch": 29.21, + "learning_rate": 1.316131642058424e-06, + "loss": 0.9084, + "step": 337480 + }, + { + "epoch": 29.21, + "learning_rate": 1.314686931145078e-06, + "loss": 0.8876, + "step": 337490 + }, + { + "epoch": 29.21, + "learning_rate": 1.3132422202317317e-06, + "loss": 0.8917, + "step": 337500 + }, + { + "epoch": 29.21, + "learning_rate": 1.3117975093183853e-06, + "loss": 0.9002, + "step": 337510 + }, + { + "epoch": 29.21, + "learning_rate": 1.3103527984050392e-06, + "loss": 0.8815, + "step": 337520 + }, + { + "epoch": 29.22, + "learning_rate": 1.308908087491693e-06, + "loss": 0.8828, + "step": 337530 + }, + { + "epoch": 29.22, + "learning_rate": 1.3074633765783468e-06, + "loss": 0.8948, + "step": 337540 + }, + { + "epoch": 29.22, + "learning_rate": 1.3060186656650006e-06, + "loss": 0.9106, + "step": 337550 + }, + { + "epoch": 29.22, + "learning_rate": 1.3045739547516542e-06, + "loss": 0.9348, + "step": 337560 + }, + { + "epoch": 29.22, + "learning_rate": 1.303129243838308e-06, + "loss": 0.9575, + "step": 337570 + }, + { + "epoch": 29.22, + "learning_rate": 1.3016845329249619e-06, + "loss": 0.9576, + "step": 337580 + }, + { + "epoch": 29.22, + "learning_rate": 1.3002398220116155e-06, + "loss": 0.866, + "step": 337590 + }, + { + "epoch": 29.22, + "learning_rate": 1.2987951110982693e-06, + "loss": 0.8857, + "step": 337600 + }, + { + "epoch": 29.22, + "learning_rate": 1.297350400184923e-06, + "loss": 0.8767, + "step": 337610 + }, + { + "epoch": 29.22, + "learning_rate": 1.295905689271577e-06, + "loss": 0.8929, + "step": 337620 + }, + { + "epoch": 29.22, + "learning_rate": 1.2944609783582307e-06, + "loss": 0.8641, + "step": 337630 + }, + { + "epoch": 29.23, + "learning_rate": 1.2930162674448843e-06, + "loss": 0.8337, + "step": 337640 + }, + { + "epoch": 29.23, + "learning_rate": 1.2915715565315382e-06, + "loss": 0.8666, + "step": 337650 + }, + { + "epoch": 29.23, + "learning_rate": 1.2901268456181918e-06, + "loss": 0.8977, + "step": 337660 + }, + { + "epoch": 29.23, + "learning_rate": 1.2886821347048456e-06, + "loss": 0.8673, + "step": 337670 + }, + { + "epoch": 29.23, + "learning_rate": 1.2872374237914994e-06, + "loss": 0.911, + "step": 337680 + }, + { + "epoch": 29.23, + "learning_rate": 1.285792712878153e-06, + "loss": 0.8735, + "step": 337690 + }, + { + "epoch": 29.23, + "learning_rate": 1.284348001964807e-06, + "loss": 0.9144, + "step": 337700 + }, + { + "epoch": 29.23, + "learning_rate": 1.2829032910514606e-06, + "loss": 0.8754, + "step": 337710 + }, + { + "epoch": 29.23, + "learning_rate": 1.2814585801381145e-06, + "loss": 0.9037, + "step": 337720 + }, + { + "epoch": 29.23, + "learning_rate": 1.2800138692247683e-06, + "loss": 0.918, + "step": 337730 + }, + { + "epoch": 29.23, + "learning_rate": 1.2785691583114219e-06, + "loss": 0.8773, + "step": 337740 + }, + { + "epoch": 29.23, + "learning_rate": 1.2771244473980757e-06, + "loss": 0.8959, + "step": 337750 + }, + { + "epoch": 29.24, + "learning_rate": 1.2756797364847293e-06, + "loss": 0.9114, + "step": 337760 + }, + { + "epoch": 29.24, + "learning_rate": 1.2742350255713831e-06, + "loss": 0.9263, + "step": 337770 + }, + { + "epoch": 29.24, + "learning_rate": 1.272790314658037e-06, + "loss": 0.8619, + "step": 337780 + }, + { + "epoch": 29.24, + "learning_rate": 1.2713456037446908e-06, + "loss": 0.9328, + "step": 337790 + }, + { + "epoch": 29.24, + "learning_rate": 1.2699008928313446e-06, + "loss": 0.8793, + "step": 337800 + }, + { + "epoch": 29.24, + "learning_rate": 1.2684561819179982e-06, + "loss": 0.8783, + "step": 337810 + }, + { + "epoch": 29.24, + "learning_rate": 1.267011471004652e-06, + "loss": 0.9201, + "step": 337820 + }, + { + "epoch": 29.24, + "learning_rate": 1.2655667600913058e-06, + "loss": 0.9009, + "step": 337830 + }, + { + "epoch": 29.24, + "learning_rate": 1.2641220491779594e-06, + "loss": 0.9206, + "step": 337840 + }, + { + "epoch": 29.24, + "learning_rate": 1.2626773382646133e-06, + "loss": 0.8515, + "step": 337850 + }, + { + "epoch": 29.24, + "learning_rate": 1.261232627351267e-06, + "loss": 0.9311, + "step": 337860 + }, + { + "epoch": 29.24, + "learning_rate": 1.259787916437921e-06, + "loss": 0.9155, + "step": 337870 + }, + { + "epoch": 29.25, + "learning_rate": 1.2583432055245747e-06, + "loss": 0.899, + "step": 337880 + }, + { + "epoch": 29.25, + "learning_rate": 1.2568984946112283e-06, + "loss": 0.8909, + "step": 337890 + }, + { + "epoch": 29.25, + "learning_rate": 1.2554537836978821e-06, + "loss": 0.8596, + "step": 337900 + }, + { + "epoch": 29.25, + "learning_rate": 1.254009072784536e-06, + "loss": 0.919, + "step": 337910 + }, + { + "epoch": 29.25, + "learning_rate": 1.2525643618711896e-06, + "loss": 0.9218, + "step": 337920 + }, + { + "epoch": 29.25, + "learning_rate": 1.2511196509578434e-06, + "loss": 0.9235, + "step": 337930 + }, + { + "epoch": 29.25, + "learning_rate": 1.249674940044497e-06, + "loss": 0.9646, + "step": 337940 + }, + { + "epoch": 29.25, + "learning_rate": 1.248230229131151e-06, + "loss": 0.9264, + "step": 337950 + }, + { + "epoch": 29.25, + "learning_rate": 1.2467855182178048e-06, + "loss": 0.879, + "step": 337960 + }, + { + "epoch": 29.25, + "learning_rate": 1.2453408073044584e-06, + "loss": 0.8876, + "step": 337970 + }, + { + "epoch": 29.25, + "learning_rate": 1.2438960963911123e-06, + "loss": 0.9049, + "step": 337980 + }, + { + "epoch": 29.26, + "learning_rate": 1.2424513854777659e-06, + "loss": 0.8948, + "step": 337990 + }, + { + "epoch": 29.26, + "learning_rate": 1.2410066745644197e-06, + "loss": 0.8539, + "step": 338000 + }, + { + "epoch": 29.26, + "learning_rate": 1.2395619636510735e-06, + "loss": 0.9198, + "step": 338010 + }, + { + "epoch": 29.26, + "learning_rate": 1.2381172527377271e-06, + "loss": 0.8679, + "step": 338020 + }, + { + "epoch": 29.26, + "learning_rate": 1.236672541824381e-06, + "loss": 0.9181, + "step": 338030 + }, + { + "epoch": 29.26, + "learning_rate": 1.2352278309110348e-06, + "loss": 0.8906, + "step": 338040 + }, + { + "epoch": 29.26, + "learning_rate": 1.2337831199976886e-06, + "loss": 0.9435, + "step": 338050 + }, + { + "epoch": 29.26, + "learning_rate": 1.2323384090843424e-06, + "loss": 0.9146, + "step": 338060 + }, + { + "epoch": 29.26, + "learning_rate": 1.230893698170996e-06, + "loss": 0.9134, + "step": 338070 + }, + { + "epoch": 29.26, + "learning_rate": 1.2294489872576498e-06, + "loss": 0.8661, + "step": 338080 + }, + { + "epoch": 29.26, + "learning_rate": 1.2280042763443036e-06, + "loss": 0.8919, + "step": 338090 + }, + { + "epoch": 29.26, + "learning_rate": 1.2265595654309572e-06, + "loss": 0.9094, + "step": 338100 + }, + { + "epoch": 29.27, + "learning_rate": 1.225114854517611e-06, + "loss": 0.9397, + "step": 338110 + }, + { + "epoch": 29.27, + "learning_rate": 1.2236701436042649e-06, + "loss": 0.9645, + "step": 338120 + }, + { + "epoch": 29.27, + "learning_rate": 1.2222254326909187e-06, + "loss": 0.8882, + "step": 338130 + }, + { + "epoch": 29.27, + "learning_rate": 1.2207807217775725e-06, + "loss": 0.8591, + "step": 338140 + }, + { + "epoch": 29.27, + "learning_rate": 1.2193360108642261e-06, + "loss": 0.9263, + "step": 338150 + }, + { + "epoch": 29.27, + "learning_rate": 1.21789129995088e-06, + "loss": 0.8983, + "step": 338160 + }, + { + "epoch": 29.27, + "learning_rate": 1.2164465890375335e-06, + "loss": 0.8699, + "step": 338170 + }, + { + "epoch": 29.27, + "learning_rate": 1.2150018781241874e-06, + "loss": 0.927, + "step": 338180 + }, + { + "epoch": 29.27, + "learning_rate": 1.2135571672108412e-06, + "loss": 0.8967, + "step": 338190 + }, + { + "epoch": 29.27, + "learning_rate": 1.212112456297495e-06, + "loss": 0.8519, + "step": 338200 + }, + { + "epoch": 29.27, + "learning_rate": 1.2106677453841488e-06, + "loss": 0.8486, + "step": 338210 + }, + { + "epoch": 29.28, + "learning_rate": 1.2092230344708024e-06, + "loss": 0.8838, + "step": 338220 + }, + { + "epoch": 29.28, + "learning_rate": 1.2077783235574562e-06, + "loss": 0.8874, + "step": 338230 + }, + { + "epoch": 29.28, + "learning_rate": 1.20633361264411e-06, + "loss": 0.8998, + "step": 338240 + }, + { + "epoch": 29.28, + "learning_rate": 1.2048889017307637e-06, + "loss": 0.874, + "step": 338250 + }, + { + "epoch": 29.28, + "learning_rate": 1.2034441908174175e-06, + "loss": 0.8963, + "step": 338260 + }, + { + "epoch": 29.28, + "learning_rate": 1.201999479904071e-06, + "loss": 0.89, + "step": 338270 + }, + { + "epoch": 29.28, + "learning_rate": 1.200554768990725e-06, + "loss": 0.9076, + "step": 338280 + }, + { + "epoch": 29.28, + "learning_rate": 1.1991100580773787e-06, + "loss": 0.9115, + "step": 338290 + }, + { + "epoch": 29.28, + "learning_rate": 1.1976653471640325e-06, + "loss": 0.8715, + "step": 338300 + }, + { + "epoch": 29.28, + "learning_rate": 1.1962206362506864e-06, + "loss": 0.8945, + "step": 338310 + }, + { + "epoch": 29.28, + "learning_rate": 1.19477592533734e-06, + "loss": 0.8725, + "step": 338320 + }, + { + "epoch": 29.28, + "learning_rate": 1.1933312144239938e-06, + "loss": 0.8953, + "step": 338330 + }, + { + "epoch": 29.29, + "learning_rate": 1.1918865035106476e-06, + "loss": 0.9147, + "step": 338340 + }, + { + "epoch": 29.29, + "learning_rate": 1.1904417925973012e-06, + "loss": 0.8563, + "step": 338350 + }, + { + "epoch": 29.29, + "learning_rate": 1.188997081683955e-06, + "loss": 0.8667, + "step": 338360 + }, + { + "epoch": 29.29, + "learning_rate": 1.1875523707706089e-06, + "loss": 0.9253, + "step": 338370 + }, + { + "epoch": 29.29, + "learning_rate": 1.1861076598572627e-06, + "loss": 0.8935, + "step": 338380 + }, + { + "epoch": 29.29, + "learning_rate": 1.1846629489439165e-06, + "loss": 0.8879, + "step": 338390 + }, + { + "epoch": 29.29, + "learning_rate": 1.18321823803057e-06, + "loss": 0.9013, + "step": 338400 + }, + { + "epoch": 29.29, + "learning_rate": 1.181773527117224e-06, + "loss": 0.8973, + "step": 338410 + }, + { + "epoch": 29.29, + "learning_rate": 1.1803288162038777e-06, + "loss": 0.9097, + "step": 338420 + }, + { + "epoch": 29.29, + "learning_rate": 1.1788841052905313e-06, + "loss": 0.9488, + "step": 338430 + }, + { + "epoch": 29.29, + "learning_rate": 1.1774393943771852e-06, + "loss": 0.8848, + "step": 338440 + }, + { + "epoch": 29.3, + "learning_rate": 1.175994683463839e-06, + "loss": 0.9129, + "step": 338450 + }, + { + "epoch": 29.3, + "learning_rate": 1.1745499725504928e-06, + "loss": 0.8622, + "step": 338460 + }, + { + "epoch": 29.3, + "learning_rate": 1.1731052616371466e-06, + "loss": 0.8903, + "step": 338470 + }, + { + "epoch": 29.3, + "learning_rate": 1.1716605507238002e-06, + "loss": 0.9139, + "step": 338480 + }, + { + "epoch": 29.3, + "learning_rate": 1.170215839810454e-06, + "loss": 0.9137, + "step": 338490 + }, + { + "epoch": 29.3, + "learning_rate": 1.1687711288971076e-06, + "loss": 0.8957, + "step": 338500 + }, + { + "epoch": 29.3, + "learning_rate": 1.1673264179837615e-06, + "loss": 0.919, + "step": 338510 + }, + { + "epoch": 29.3, + "learning_rate": 1.1658817070704153e-06, + "loss": 0.886, + "step": 338520 + }, + { + "epoch": 29.3, + "learning_rate": 1.1644369961570689e-06, + "loss": 0.8687, + "step": 338530 + }, + { + "epoch": 29.3, + "learning_rate": 1.1629922852437227e-06, + "loss": 0.9361, + "step": 338540 + }, + { + "epoch": 29.3, + "learning_rate": 1.1615475743303765e-06, + "loss": 0.9574, + "step": 338550 + }, + { + "epoch": 29.3, + "learning_rate": 1.1601028634170303e-06, + "loss": 0.8792, + "step": 338560 + }, + { + "epoch": 29.31, + "learning_rate": 1.1586581525036842e-06, + "loss": 0.9554, + "step": 338570 + }, + { + "epoch": 29.31, + "learning_rate": 1.1572134415903378e-06, + "loss": 0.8731, + "step": 338580 + }, + { + "epoch": 29.31, + "learning_rate": 1.1557687306769916e-06, + "loss": 0.9569, + "step": 338590 + }, + { + "epoch": 29.31, + "learning_rate": 1.1543240197636452e-06, + "loss": 0.9172, + "step": 338600 + }, + { + "epoch": 29.31, + "learning_rate": 1.152879308850299e-06, + "loss": 0.8785, + "step": 338610 + }, + { + "epoch": 29.31, + "learning_rate": 1.1514345979369528e-06, + "loss": 0.8526, + "step": 338620 + }, + { + "epoch": 29.31, + "learning_rate": 1.1499898870236067e-06, + "loss": 0.8468, + "step": 338630 + }, + { + "epoch": 29.31, + "learning_rate": 1.1485451761102605e-06, + "loss": 0.8925, + "step": 338640 + }, + { + "epoch": 29.31, + "learning_rate": 1.1471004651969143e-06, + "loss": 0.9085, + "step": 338650 + }, + { + "epoch": 29.31, + "learning_rate": 1.145655754283568e-06, + "loss": 0.8942, + "step": 338660 + }, + { + "epoch": 29.31, + "learning_rate": 1.1442110433702217e-06, + "loss": 0.9365, + "step": 338670 + }, + { + "epoch": 29.32, + "learning_rate": 1.1427663324568753e-06, + "loss": 0.9272, + "step": 338680 + }, + { + "epoch": 29.32, + "learning_rate": 1.1413216215435291e-06, + "loss": 0.9227, + "step": 338690 + }, + { + "epoch": 29.32, + "learning_rate": 1.139876910630183e-06, + "loss": 0.9066, + "step": 338700 + }, + { + "epoch": 29.32, + "learning_rate": 1.1384321997168368e-06, + "loss": 0.8869, + "step": 338710 + }, + { + "epoch": 29.32, + "learning_rate": 1.1369874888034906e-06, + "loss": 0.9007, + "step": 338720 + }, + { + "epoch": 29.32, + "learning_rate": 1.1355427778901442e-06, + "loss": 0.9417, + "step": 338730 + }, + { + "epoch": 29.32, + "learning_rate": 1.134098066976798e-06, + "loss": 0.9149, + "step": 338740 + }, + { + "epoch": 29.32, + "learning_rate": 1.1326533560634518e-06, + "loss": 0.8371, + "step": 338750 + }, + { + "epoch": 29.32, + "learning_rate": 1.1312086451501054e-06, + "loss": 0.9174, + "step": 338760 + }, + { + "epoch": 29.32, + "learning_rate": 1.1297639342367593e-06, + "loss": 0.8885, + "step": 338770 + }, + { + "epoch": 29.32, + "learning_rate": 1.1283192233234129e-06, + "loss": 0.9013, + "step": 338780 + }, + { + "epoch": 29.32, + "learning_rate": 1.1268745124100667e-06, + "loss": 0.8628, + "step": 338790 + }, + { + "epoch": 29.33, + "learning_rate": 1.1254298014967207e-06, + "loss": 0.8597, + "step": 338800 + }, + { + "epoch": 29.33, + "learning_rate": 1.1239850905833743e-06, + "loss": 0.8569, + "step": 338810 + }, + { + "epoch": 29.33, + "learning_rate": 1.1225403796700281e-06, + "loss": 0.8523, + "step": 338820 + }, + { + "epoch": 29.33, + "learning_rate": 1.1210956687566818e-06, + "loss": 0.8663, + "step": 338830 + }, + { + "epoch": 29.33, + "learning_rate": 1.1196509578433356e-06, + "loss": 0.9547, + "step": 338840 + }, + { + "epoch": 29.33, + "learning_rate": 1.1182062469299894e-06, + "loss": 0.9131, + "step": 338850 + }, + { + "epoch": 29.33, + "learning_rate": 1.116761536016643e-06, + "loss": 0.84, + "step": 338860 + }, + { + "epoch": 29.33, + "learning_rate": 1.1153168251032968e-06, + "loss": 0.9103, + "step": 338870 + }, + { + "epoch": 29.33, + "learning_rate": 1.1138721141899506e-06, + "loss": 0.8732, + "step": 338880 + }, + { + "epoch": 29.33, + "learning_rate": 1.1124274032766045e-06, + "loss": 0.9427, + "step": 338890 + }, + { + "epoch": 29.33, + "learning_rate": 1.1109826923632583e-06, + "loss": 0.8186, + "step": 338900 + }, + { + "epoch": 29.34, + "learning_rate": 1.1095379814499119e-06, + "loss": 0.8587, + "step": 338910 + }, + { + "epoch": 29.34, + "learning_rate": 1.1080932705365657e-06, + "loss": 0.8747, + "step": 338920 + }, + { + "epoch": 29.34, + "learning_rate": 1.1066485596232195e-06, + "loss": 0.8606, + "step": 338930 + }, + { + "epoch": 29.34, + "learning_rate": 1.1052038487098731e-06, + "loss": 0.8602, + "step": 338940 + }, + { + "epoch": 29.34, + "learning_rate": 1.103759137796527e-06, + "loss": 0.8861, + "step": 338950 + }, + { + "epoch": 29.34, + "learning_rate": 1.1023144268831808e-06, + "loss": 0.904, + "step": 338960 + }, + { + "epoch": 29.34, + "learning_rate": 1.1008697159698346e-06, + "loss": 0.8392, + "step": 338970 + }, + { + "epoch": 29.34, + "learning_rate": 1.0994250050564884e-06, + "loss": 0.8711, + "step": 338980 + }, + { + "epoch": 29.34, + "learning_rate": 1.097980294143142e-06, + "loss": 0.8928, + "step": 338990 + }, + { + "epoch": 29.34, + "learning_rate": 1.0965355832297958e-06, + "loss": 0.9426, + "step": 339000 + }, + { + "epoch": 29.34, + "learning_rate": 1.0950908723164494e-06, + "loss": 0.8626, + "step": 339010 + }, + { + "epoch": 29.34, + "learning_rate": 1.0936461614031032e-06, + "loss": 0.8837, + "step": 339020 + }, + { + "epoch": 29.35, + "learning_rate": 1.092201450489757e-06, + "loss": 0.8472, + "step": 339030 + }, + { + "epoch": 29.35, + "learning_rate": 1.0907567395764107e-06, + "loss": 0.8987, + "step": 339040 + }, + { + "epoch": 29.35, + "learning_rate": 1.0893120286630647e-06, + "loss": 0.8963, + "step": 339050 + }, + { + "epoch": 29.35, + "learning_rate": 1.0878673177497183e-06, + "loss": 0.9718, + "step": 339060 + }, + { + "epoch": 29.35, + "learning_rate": 1.0864226068363721e-06, + "loss": 0.8569, + "step": 339070 + }, + { + "epoch": 29.35, + "learning_rate": 1.084977895923026e-06, + "loss": 0.8949, + "step": 339080 + }, + { + "epoch": 29.35, + "learning_rate": 1.0835331850096795e-06, + "loss": 0.9288, + "step": 339090 + }, + { + "epoch": 29.35, + "learning_rate": 1.0820884740963334e-06, + "loss": 0.8431, + "step": 339100 + }, + { + "epoch": 29.35, + "learning_rate": 1.080643763182987e-06, + "loss": 0.9725, + "step": 339110 + }, + { + "epoch": 29.35, + "learning_rate": 1.0791990522696408e-06, + "loss": 0.8959, + "step": 339120 + }, + { + "epoch": 29.35, + "learning_rate": 1.0777543413562946e-06, + "loss": 0.9001, + "step": 339130 + }, + { + "epoch": 29.35, + "learning_rate": 1.0763096304429484e-06, + "loss": 0.9197, + "step": 339140 + }, + { + "epoch": 29.36, + "learning_rate": 1.0748649195296022e-06, + "loss": 0.9094, + "step": 339150 + }, + { + "epoch": 29.36, + "learning_rate": 1.073420208616256e-06, + "loss": 0.8717, + "step": 339160 + }, + { + "epoch": 29.36, + "learning_rate": 1.0719754977029097e-06, + "loss": 0.8817, + "step": 339170 + }, + { + "epoch": 29.36, + "learning_rate": 1.0705307867895635e-06, + "loss": 0.8865, + "step": 339180 + }, + { + "epoch": 29.36, + "learning_rate": 1.069086075876217e-06, + "loss": 0.8943, + "step": 339190 + }, + { + "epoch": 29.36, + "learning_rate": 1.067641364962871e-06, + "loss": 0.8965, + "step": 339200 + }, + { + "epoch": 29.36, + "learning_rate": 1.0661966540495247e-06, + "loss": 0.9486, + "step": 339210 + }, + { + "epoch": 29.36, + "learning_rate": 1.0647519431361786e-06, + "loss": 0.8819, + "step": 339220 + }, + { + "epoch": 29.36, + "learning_rate": 1.0633072322228324e-06, + "loss": 0.8886, + "step": 339230 + }, + { + "epoch": 29.36, + "learning_rate": 1.061862521309486e-06, + "loss": 0.913, + "step": 339240 + }, + { + "epoch": 29.36, + "learning_rate": 1.0604178103961398e-06, + "loss": 0.8827, + "step": 339250 + }, + { + "epoch": 29.37, + "learning_rate": 1.0589730994827936e-06, + "loss": 0.9166, + "step": 339260 + }, + { + "epoch": 29.37, + "learning_rate": 1.0575283885694472e-06, + "loss": 0.8748, + "step": 339270 + }, + { + "epoch": 29.37, + "learning_rate": 1.056083677656101e-06, + "loss": 0.8469, + "step": 339280 + }, + { + "epoch": 29.37, + "learning_rate": 1.0546389667427546e-06, + "loss": 0.9061, + "step": 339290 + }, + { + "epoch": 29.37, + "learning_rate": 1.0531942558294087e-06, + "loss": 0.8928, + "step": 339300 + }, + { + "epoch": 29.37, + "learning_rate": 1.0517495449160625e-06, + "loss": 0.948, + "step": 339310 + }, + { + "epoch": 29.37, + "learning_rate": 1.050304834002716e-06, + "loss": 0.8686, + "step": 339320 + }, + { + "epoch": 29.37, + "learning_rate": 1.04886012308937e-06, + "loss": 0.8751, + "step": 339330 + }, + { + "epoch": 29.37, + "learning_rate": 1.0474154121760235e-06, + "loss": 0.8988, + "step": 339340 + }, + { + "epoch": 29.37, + "learning_rate": 1.0459707012626773e-06, + "loss": 0.9151, + "step": 339350 + }, + { + "epoch": 29.37, + "learning_rate": 1.0445259903493312e-06, + "loss": 0.887, + "step": 339360 + }, + { + "epoch": 29.37, + "learning_rate": 1.0430812794359848e-06, + "loss": 0.8945, + "step": 339370 + }, + { + "epoch": 29.38, + "learning_rate": 1.0416365685226386e-06, + "loss": 0.8602, + "step": 339380 + }, + { + "epoch": 29.38, + "learning_rate": 1.0401918576092924e-06, + "loss": 0.9366, + "step": 339390 + }, + { + "epoch": 29.38, + "learning_rate": 1.0387471466959462e-06, + "loss": 0.8285, + "step": 339400 + }, + { + "epoch": 29.38, + "learning_rate": 1.0373024357826e-06, + "loss": 0.9205, + "step": 339410 + }, + { + "epoch": 29.38, + "learning_rate": 1.0358577248692537e-06, + "loss": 0.8851, + "step": 339420 + }, + { + "epoch": 29.38, + "learning_rate": 1.0344130139559075e-06, + "loss": 0.9354, + "step": 339430 + }, + { + "epoch": 29.38, + "learning_rate": 1.0329683030425613e-06, + "loss": 0.8612, + "step": 339440 + }, + { + "epoch": 29.38, + "learning_rate": 1.031523592129215e-06, + "loss": 0.9608, + "step": 339450 + }, + { + "epoch": 29.38, + "learning_rate": 1.0300788812158687e-06, + "loss": 0.9019, + "step": 339460 + }, + { + "epoch": 29.38, + "learning_rate": 1.0286341703025225e-06, + "loss": 0.951, + "step": 339470 + }, + { + "epoch": 29.38, + "learning_rate": 1.0271894593891764e-06, + "loss": 0.8944, + "step": 339480 + }, + { + "epoch": 29.39, + "learning_rate": 1.0257447484758302e-06, + "loss": 0.8753, + "step": 339490 + }, + { + "epoch": 29.39, + "learning_rate": 1.0243000375624838e-06, + "loss": 0.9205, + "step": 339500 + }, + { + "epoch": 29.39, + "learning_rate": 1.0228553266491376e-06, + "loss": 0.9279, + "step": 339510 + }, + { + "epoch": 29.39, + "learning_rate": 1.0214106157357912e-06, + "loss": 0.9277, + "step": 339520 + }, + { + "epoch": 29.39, + "learning_rate": 1.019965904822445e-06, + "loss": 0.8748, + "step": 339530 + }, + { + "epoch": 29.39, + "learning_rate": 1.0185211939090988e-06, + "loss": 0.9251, + "step": 339540 + }, + { + "epoch": 29.39, + "learning_rate": 1.0170764829957527e-06, + "loss": 0.8883, + "step": 339550 + }, + { + "epoch": 29.39, + "learning_rate": 1.0156317720824065e-06, + "loss": 0.9059, + "step": 339560 + }, + { + "epoch": 29.39, + "learning_rate": 1.01418706116906e-06, + "loss": 0.9021, + "step": 339570 + }, + { + "epoch": 29.39, + "learning_rate": 1.012742350255714e-06, + "loss": 0.889, + "step": 339580 + }, + { + "epoch": 29.39, + "learning_rate": 1.0112976393423677e-06, + "loss": 0.9111, + "step": 339590 + }, + { + "epoch": 29.39, + "learning_rate": 1.0098529284290213e-06, + "loss": 0.9432, + "step": 339600 + }, + { + "epoch": 29.4, + "learning_rate": 1.0084082175156751e-06, + "loss": 0.9365, + "step": 339610 + }, + { + "epoch": 29.4, + "learning_rate": 1.0069635066023288e-06, + "loss": 0.9445, + "step": 339620 + }, + { + "epoch": 29.4, + "learning_rate": 1.0055187956889826e-06, + "loss": 0.9517, + "step": 339630 + }, + { + "epoch": 29.4, + "learning_rate": 1.0040740847756366e-06, + "loss": 0.9039, + "step": 339640 + }, + { + "epoch": 29.4, + "learning_rate": 1.0026293738622902e-06, + "loss": 0.8604, + "step": 339650 + }, + { + "epoch": 29.4, + "learning_rate": 1.001184662948944e-06, + "loss": 0.9132, + "step": 339660 + }, + { + "epoch": 29.4, + "learning_rate": 9.997399520355978e-07, + "loss": 0.9136, + "step": 339670 + }, + { + "epoch": 29.4, + "learning_rate": 9.982952411222515e-07, + "loss": 0.9036, + "step": 339680 + }, + { + "epoch": 29.4, + "learning_rate": 9.968505302089053e-07, + "loss": 0.9448, + "step": 339690 + }, + { + "epoch": 29.4, + "learning_rate": 9.954058192955589e-07, + "loss": 0.9721, + "step": 339700 + }, + { + "epoch": 29.4, + "learning_rate": 9.939611083822127e-07, + "loss": 0.9032, + "step": 339710 + }, + { + "epoch": 29.41, + "learning_rate": 9.925163974688665e-07, + "loss": 0.9347, + "step": 339720 + }, + { + "epoch": 29.41, + "learning_rate": 9.910716865555203e-07, + "loss": 0.9222, + "step": 339730 + }, + { + "epoch": 29.41, + "learning_rate": 9.896269756421742e-07, + "loss": 0.8537, + "step": 339740 + }, + { + "epoch": 29.41, + "learning_rate": 9.881822647288278e-07, + "loss": 0.9175, + "step": 339750 + }, + { + "epoch": 29.41, + "learning_rate": 9.867375538154816e-07, + "loss": 0.885, + "step": 339760 + }, + { + "epoch": 29.41, + "learning_rate": 9.852928429021354e-07, + "loss": 0.9474, + "step": 339770 + }, + { + "epoch": 29.41, + "learning_rate": 9.83848131988789e-07, + "loss": 0.8518, + "step": 339780 + }, + { + "epoch": 29.41, + "learning_rate": 9.824034210754428e-07, + "loss": 0.8964, + "step": 339790 + }, + { + "epoch": 29.41, + "learning_rate": 9.809587101620966e-07, + "loss": 0.8653, + "step": 339800 + }, + { + "epoch": 29.41, + "learning_rate": 9.795139992487505e-07, + "loss": 0.9651, + "step": 339810 + }, + { + "epoch": 29.41, + "learning_rate": 9.780692883354043e-07, + "loss": 0.8265, + "step": 339820 + }, + { + "epoch": 29.41, + "learning_rate": 9.766245774220579e-07, + "loss": 0.8825, + "step": 339830 + }, + { + "epoch": 29.42, + "learning_rate": 9.751798665087117e-07, + "loss": 0.9429, + "step": 339840 + }, + { + "epoch": 29.42, + "learning_rate": 9.737351555953653e-07, + "loss": 0.9068, + "step": 339850 + }, + { + "epoch": 29.42, + "learning_rate": 9.722904446820191e-07, + "loss": 0.9402, + "step": 339860 + }, + { + "epoch": 29.42, + "learning_rate": 9.70845733768673e-07, + "loss": 0.9121, + "step": 339870 + }, + { + "epoch": 29.42, + "learning_rate": 9.694010228553266e-07, + "loss": 0.9019, + "step": 339880 + }, + { + "epoch": 29.42, + "learning_rate": 9.679563119419806e-07, + "loss": 0.9001, + "step": 339890 + }, + { + "epoch": 29.42, + "learning_rate": 9.665116010286342e-07, + "loss": 0.8776, + "step": 339900 + }, + { + "epoch": 29.42, + "learning_rate": 9.65066890115288e-07, + "loss": 0.9578, + "step": 339910 + }, + { + "epoch": 29.42, + "learning_rate": 9.636221792019418e-07, + "loss": 0.8692, + "step": 339920 + }, + { + "epoch": 29.42, + "learning_rate": 9.621774682885954e-07, + "loss": 0.876, + "step": 339930 + }, + { + "epoch": 29.42, + "learning_rate": 9.607327573752493e-07, + "loss": 0.925, + "step": 339940 + }, + { + "epoch": 29.43, + "learning_rate": 9.59288046461903e-07, + "loss": 0.937, + "step": 339950 + }, + { + "epoch": 29.43, + "learning_rate": 9.578433355485567e-07, + "loss": 0.8843, + "step": 339960 + }, + { + "epoch": 29.43, + "learning_rate": 9.563986246352105e-07, + "loss": 0.8387, + "step": 339970 + }, + { + "epoch": 29.43, + "learning_rate": 9.549539137218643e-07, + "loss": 0.9039, + "step": 339980 + }, + { + "epoch": 29.43, + "learning_rate": 9.53509202808518e-07, + "loss": 0.8862, + "step": 339990 + }, + { + "epoch": 29.43, + "learning_rate": 9.52064491895172e-07, + "loss": 0.9158, + "step": 340000 + }, + { + "epoch": 29.43, + "learning_rate": 9.506197809818256e-07, + "loss": 0.8905, + "step": 340010 + }, + { + "epoch": 29.43, + "learning_rate": 9.491750700684794e-07, + "loss": 0.8919, + "step": 340020 + }, + { + "epoch": 29.43, + "learning_rate": 9.477303591551331e-07, + "loss": 0.9111, + "step": 340030 + }, + { + "epoch": 29.43, + "learning_rate": 9.462856482417869e-07, + "loss": 0.8964, + "step": 340040 + }, + { + "epoch": 29.43, + "learning_rate": 9.448409373284407e-07, + "loss": 0.9282, + "step": 340050 + }, + { + "epoch": 29.43, + "learning_rate": 9.433962264150943e-07, + "loss": 0.8922, + "step": 340060 + }, + { + "epoch": 29.44, + "learning_rate": 9.419515155017481e-07, + "loss": 0.9201, + "step": 340070 + }, + { + "epoch": 29.44, + "learning_rate": 9.405068045884019e-07, + "loss": 0.8777, + "step": 340080 + }, + { + "epoch": 29.44, + "learning_rate": 9.390620936750557e-07, + "loss": 0.8782, + "step": 340090 + }, + { + "epoch": 29.44, + "learning_rate": 9.376173827617095e-07, + "loss": 0.8652, + "step": 340100 + }, + { + "epoch": 29.44, + "learning_rate": 9.361726718483631e-07, + "loss": 0.9072, + "step": 340110 + }, + { + "epoch": 29.44, + "learning_rate": 9.347279609350169e-07, + "loss": 0.8772, + "step": 340120 + }, + { + "epoch": 29.44, + "learning_rate": 9.332832500216706e-07, + "loss": 0.8925, + "step": 340130 + }, + { + "epoch": 29.44, + "learning_rate": 9.318385391083245e-07, + "loss": 0.8988, + "step": 340140 + }, + { + "epoch": 29.44, + "learning_rate": 9.303938281949783e-07, + "loss": 0.8905, + "step": 340150 + }, + { + "epoch": 29.44, + "learning_rate": 9.28949117281632e-07, + "loss": 0.8574, + "step": 340160 + }, + { + "epoch": 29.44, + "learning_rate": 9.275044063682858e-07, + "loss": 0.901, + "step": 340170 + }, + { + "epoch": 29.44, + "learning_rate": 9.260596954549394e-07, + "loss": 0.8581, + "step": 340180 + }, + { + "epoch": 29.45, + "learning_rate": 9.246149845415932e-07, + "loss": 0.9418, + "step": 340190 + }, + { + "epoch": 29.45, + "learning_rate": 9.23170273628247e-07, + "loss": 0.8723, + "step": 340200 + }, + { + "epoch": 29.45, + "learning_rate": 9.217255627149008e-07, + "loss": 0.9405, + "step": 340210 + }, + { + "epoch": 29.45, + "learning_rate": 9.202808518015546e-07, + "loss": 0.8809, + "step": 340220 + }, + { + "epoch": 29.45, + "learning_rate": 9.188361408882084e-07, + "loss": 0.8426, + "step": 340230 + }, + { + "epoch": 29.45, + "learning_rate": 9.17391429974862e-07, + "loss": 0.9688, + "step": 340240 + }, + { + "epoch": 29.45, + "learning_rate": 9.159467190615159e-07, + "loss": 0.9294, + "step": 340250 + }, + { + "epoch": 29.45, + "learning_rate": 9.145020081481695e-07, + "loss": 0.951, + "step": 340260 + }, + { + "epoch": 29.45, + "learning_rate": 9.130572972348234e-07, + "loss": 0.8843, + "step": 340270 + }, + { + "epoch": 29.45, + "learning_rate": 9.116125863214772e-07, + "loss": 0.8732, + "step": 340280 + }, + { + "epoch": 29.45, + "learning_rate": 9.101678754081309e-07, + "loss": 0.8535, + "step": 340290 + }, + { + "epoch": 29.46, + "learning_rate": 9.087231644947847e-07, + "loss": 0.8802, + "step": 340300 + }, + { + "epoch": 29.46, + "learning_rate": 9.072784535814383e-07, + "loss": 0.89, + "step": 340310 + }, + { + "epoch": 29.46, + "learning_rate": 9.058337426680921e-07, + "loss": 0.9651, + "step": 340320 + }, + { + "epoch": 29.46, + "learning_rate": 9.04389031754746e-07, + "loss": 0.9183, + "step": 340330 + }, + { + "epoch": 29.46, + "learning_rate": 9.029443208413997e-07, + "loss": 0.9107, + "step": 340340 + }, + { + "epoch": 29.46, + "learning_rate": 9.014996099280535e-07, + "loss": 0.903, + "step": 340350 + }, + { + "epoch": 29.46, + "learning_rate": 9.000548990147071e-07, + "loss": 0.8721, + "step": 340360 + }, + { + "epoch": 29.46, + "learning_rate": 8.986101881013609e-07, + "loss": 0.8625, + "step": 340370 + }, + { + "epoch": 29.46, + "learning_rate": 8.971654771880148e-07, + "loss": 0.9009, + "step": 340380 + }, + { + "epoch": 29.46, + "learning_rate": 8.957207662746684e-07, + "loss": 0.8645, + "step": 340390 + }, + { + "epoch": 29.46, + "learning_rate": 8.942760553613223e-07, + "loss": 0.9463, + "step": 340400 + }, + { + "epoch": 29.46, + "learning_rate": 8.92831344447976e-07, + "loss": 0.8378, + "step": 340410 + }, + { + "epoch": 29.47, + "learning_rate": 8.913866335346298e-07, + "loss": 0.8437, + "step": 340420 + }, + { + "epoch": 29.47, + "learning_rate": 8.899419226212836e-07, + "loss": 0.9408, + "step": 340430 + }, + { + "epoch": 29.47, + "learning_rate": 8.884972117079372e-07, + "loss": 0.8538, + "step": 340440 + }, + { + "epoch": 29.47, + "learning_rate": 8.87052500794591e-07, + "loss": 0.8602, + "step": 340450 + }, + { + "epoch": 29.47, + "learning_rate": 8.856077898812448e-07, + "loss": 0.8991, + "step": 340460 + }, + { + "epoch": 29.47, + "learning_rate": 8.841630789678986e-07, + "loss": 0.8163, + "step": 340470 + }, + { + "epoch": 29.47, + "learning_rate": 8.827183680545524e-07, + "loss": 0.9448, + "step": 340480 + }, + { + "epoch": 29.47, + "learning_rate": 8.81273657141206e-07, + "loss": 0.8974, + "step": 340490 + }, + { + "epoch": 29.47, + "learning_rate": 8.798289462278599e-07, + "loss": 0.9146, + "step": 340500 + }, + { + "epoch": 29.47, + "learning_rate": 8.783842353145137e-07, + "loss": 0.8286, + "step": 340510 + }, + { + "epoch": 29.47, + "learning_rate": 8.769395244011673e-07, + "loss": 0.9376, + "step": 340520 + }, + { + "epoch": 29.48, + "learning_rate": 8.754948134878212e-07, + "loss": 0.8969, + "step": 340530 + }, + { + "epoch": 29.48, + "learning_rate": 8.740501025744749e-07, + "loss": 0.9129, + "step": 340540 + }, + { + "epoch": 29.48, + "learning_rate": 8.726053916611287e-07, + "loss": 0.9372, + "step": 340550 + }, + { + "epoch": 29.48, + "learning_rate": 8.711606807477825e-07, + "loss": 0.8882, + "step": 340560 + }, + { + "epoch": 29.48, + "learning_rate": 8.697159698344361e-07, + "loss": 0.8731, + "step": 340570 + }, + { + "epoch": 29.48, + "learning_rate": 8.682712589210899e-07, + "loss": 0.8743, + "step": 340580 + }, + { + "epoch": 29.48, + "learning_rate": 8.668265480077436e-07, + "loss": 0.9298, + "step": 340590 + }, + { + "epoch": 29.48, + "learning_rate": 8.653818370943975e-07, + "loss": 0.9343, + "step": 340600 + }, + { + "epoch": 29.48, + "learning_rate": 8.639371261810513e-07, + "loss": 0.8894, + "step": 340610 + }, + { + "epoch": 29.48, + "learning_rate": 8.624924152677049e-07, + "loss": 0.9146, + "step": 340620 + }, + { + "epoch": 29.48, + "learning_rate": 8.610477043543588e-07, + "loss": 0.9126, + "step": 340630 + }, + { + "epoch": 29.48, + "learning_rate": 8.596029934410124e-07, + "loss": 0.8724, + "step": 340640 + }, + { + "epoch": 29.49, + "learning_rate": 8.581582825276662e-07, + "loss": 0.9029, + "step": 340650 + }, + { + "epoch": 29.49, + "learning_rate": 8.5671357161432e-07, + "loss": 0.9251, + "step": 340660 + }, + { + "epoch": 29.49, + "learning_rate": 8.552688607009738e-07, + "loss": 0.931, + "step": 340670 + }, + { + "epoch": 29.49, + "learning_rate": 8.538241497876276e-07, + "loss": 0.9202, + "step": 340680 + }, + { + "epoch": 29.49, + "learning_rate": 8.523794388742812e-07, + "loss": 0.8839, + "step": 340690 + }, + { + "epoch": 29.49, + "learning_rate": 8.50934727960935e-07, + "loss": 0.8553, + "step": 340700 + }, + { + "epoch": 29.49, + "learning_rate": 8.494900170475888e-07, + "loss": 0.8786, + "step": 340710 + }, + { + "epoch": 29.49, + "learning_rate": 8.480453061342425e-07, + "loss": 0.8899, + "step": 340720 + }, + { + "epoch": 29.49, + "learning_rate": 8.466005952208964e-07, + "loss": 0.8712, + "step": 340730 + }, + { + "epoch": 29.49, + "learning_rate": 8.451558843075502e-07, + "loss": 0.8891, + "step": 340740 + }, + { + "epoch": 29.49, + "learning_rate": 8.437111733942039e-07, + "loss": 0.8495, + "step": 340750 + }, + { + "epoch": 29.5, + "learning_rate": 8.422664624808577e-07, + "loss": 0.9176, + "step": 340760 + }, + { + "epoch": 29.5, + "learning_rate": 8.408217515675113e-07, + "loss": 0.9226, + "step": 340770 + }, + { + "epoch": 29.5, + "learning_rate": 8.393770406541651e-07, + "loss": 0.8764, + "step": 340780 + }, + { + "epoch": 29.5, + "learning_rate": 8.37932329740819e-07, + "loss": 0.8815, + "step": 340790 + }, + { + "epoch": 29.5, + "learning_rate": 8.364876188274727e-07, + "loss": 0.8707, + "step": 340800 + }, + { + "epoch": 29.5, + "learning_rate": 8.350429079141265e-07, + "loss": 0.928, + "step": 340810 + }, + { + "epoch": 29.5, + "learning_rate": 8.335981970007801e-07, + "loss": 0.8852, + "step": 340820 + }, + { + "epoch": 29.5, + "learning_rate": 8.321534860874339e-07, + "loss": 0.9723, + "step": 340830 + }, + { + "epoch": 29.5, + "learning_rate": 8.307087751740877e-07, + "loss": 0.8536, + "step": 340840 + }, + { + "epoch": 29.5, + "learning_rate": 8.292640642607414e-07, + "loss": 0.9004, + "step": 340850 + }, + { + "epoch": 29.5, + "learning_rate": 8.278193533473953e-07, + "loss": 0.8526, + "step": 340860 + }, + { + "epoch": 29.5, + "learning_rate": 8.263746424340489e-07, + "loss": 0.9042, + "step": 340870 + }, + { + "epoch": 29.51, + "learning_rate": 8.249299315207028e-07, + "loss": 0.9092, + "step": 340880 + }, + { + "epoch": 29.51, + "learning_rate": 8.234852206073566e-07, + "loss": 0.8686, + "step": 340890 + }, + { + "epoch": 29.51, + "learning_rate": 8.220405096940102e-07, + "loss": 0.8866, + "step": 340900 + }, + { + "epoch": 29.51, + "learning_rate": 8.20595798780664e-07, + "loss": 0.8667, + "step": 340910 + }, + { + "epoch": 29.51, + "learning_rate": 8.191510878673177e-07, + "loss": 0.9072, + "step": 340920 + }, + { + "epoch": 29.51, + "learning_rate": 8.177063769539716e-07, + "loss": 0.8892, + "step": 340930 + }, + { + "epoch": 29.51, + "learning_rate": 8.162616660406254e-07, + "loss": 0.9522, + "step": 340940 + }, + { + "epoch": 29.51, + "learning_rate": 8.14816955127279e-07, + "loss": 0.9078, + "step": 340950 + }, + { + "epoch": 29.51, + "learning_rate": 8.133722442139328e-07, + "loss": 0.9216, + "step": 340960 + }, + { + "epoch": 29.51, + "learning_rate": 8.119275333005867e-07, + "loss": 0.9515, + "step": 340970 + }, + { + "epoch": 29.51, + "learning_rate": 8.104828223872403e-07, + "loss": 0.8553, + "step": 340980 + }, + { + "epoch": 29.52, + "learning_rate": 8.090381114738942e-07, + "loss": 0.9063, + "step": 340990 + }, + { + "epoch": 29.52, + "learning_rate": 8.075934005605479e-07, + "loss": 0.9626, + "step": 341000 + }, + { + "epoch": 29.52, + "learning_rate": 8.061486896472017e-07, + "loss": 0.8741, + "step": 341010 + }, + { + "epoch": 29.52, + "learning_rate": 8.047039787338555e-07, + "loss": 0.8757, + "step": 341020 + }, + { + "epoch": 29.52, + "learning_rate": 8.032592678205091e-07, + "loss": 0.8533, + "step": 341030 + }, + { + "epoch": 29.52, + "learning_rate": 8.018145569071629e-07, + "loss": 0.8939, + "step": 341040 + }, + { + "epoch": 29.52, + "learning_rate": 8.003698459938166e-07, + "loss": 0.8996, + "step": 341050 + }, + { + "epoch": 29.52, + "learning_rate": 7.989251350804705e-07, + "loss": 0.9016, + "step": 341060 + }, + { + "epoch": 29.52, + "learning_rate": 7.974804241671243e-07, + "loss": 0.9037, + "step": 341070 + }, + { + "epoch": 29.52, + "learning_rate": 7.960357132537779e-07, + "loss": 0.9128, + "step": 341080 + }, + { + "epoch": 29.52, + "learning_rate": 7.945910023404317e-07, + "loss": 0.8533, + "step": 341090 + }, + { + "epoch": 29.52, + "learning_rate": 7.931462914270854e-07, + "loss": 0.8642, + "step": 341100 + }, + { + "epoch": 29.53, + "learning_rate": 7.917015805137392e-07, + "loss": 0.9412, + "step": 341110 + }, + { + "epoch": 29.53, + "learning_rate": 7.902568696003931e-07, + "loss": 0.9601, + "step": 341120 + }, + { + "epoch": 29.53, + "learning_rate": 7.888121586870468e-07, + "loss": 0.8861, + "step": 341130 + }, + { + "epoch": 29.53, + "learning_rate": 7.873674477737006e-07, + "loss": 0.9134, + "step": 341140 + }, + { + "epoch": 29.53, + "learning_rate": 7.859227368603542e-07, + "loss": 0.9536, + "step": 341150 + }, + { + "epoch": 29.53, + "learning_rate": 7.84478025947008e-07, + "loss": 0.8748, + "step": 341160 + }, + { + "epoch": 29.53, + "learning_rate": 7.830333150336618e-07, + "loss": 0.9167, + "step": 341170 + }, + { + "epoch": 29.53, + "learning_rate": 7.815886041203155e-07, + "loss": 0.8376, + "step": 341180 + }, + { + "epoch": 29.53, + "learning_rate": 7.801438932069694e-07, + "loss": 0.8393, + "step": 341190 + }, + { + "epoch": 29.53, + "learning_rate": 7.786991822936231e-07, + "loss": 0.9239, + "step": 341200 + }, + { + "epoch": 29.53, + "learning_rate": 7.772544713802768e-07, + "loss": 0.8455, + "step": 341210 + }, + { + "epoch": 29.53, + "learning_rate": 7.758097604669306e-07, + "loss": 0.8878, + "step": 341220 + }, + { + "epoch": 29.54, + "learning_rate": 7.743650495535844e-07, + "loss": 0.9129, + "step": 341230 + }, + { + "epoch": 29.54, + "learning_rate": 7.729203386402381e-07, + "loss": 0.9024, + "step": 341240 + }, + { + "epoch": 29.54, + "learning_rate": 7.714756277268918e-07, + "loss": 0.9144, + "step": 341250 + }, + { + "epoch": 29.54, + "learning_rate": 7.700309168135457e-07, + "loss": 0.8692, + "step": 341260 + }, + { + "epoch": 29.54, + "learning_rate": 7.685862059001994e-07, + "loss": 0.885, + "step": 341270 + }, + { + "epoch": 29.54, + "learning_rate": 7.671414949868532e-07, + "loss": 0.8865, + "step": 341280 + }, + { + "epoch": 29.54, + "learning_rate": 7.656967840735069e-07, + "loss": 0.9006, + "step": 341290 + }, + { + "epoch": 29.54, + "learning_rate": 7.642520731601607e-07, + "loss": 0.892, + "step": 341300 + }, + { + "epoch": 29.54, + "learning_rate": 7.628073622468144e-07, + "loss": 0.8988, + "step": 341310 + }, + { + "epoch": 29.54, + "learning_rate": 7.613626513334683e-07, + "loss": 0.9056, + "step": 341320 + }, + { + "epoch": 29.54, + "learning_rate": 7.59917940420122e-07, + "loss": 0.8748, + "step": 341330 + }, + { + "epoch": 29.55, + "learning_rate": 7.584732295067757e-07, + "loss": 0.938, + "step": 341340 + }, + { + "epoch": 29.55, + "learning_rate": 7.570285185934295e-07, + "loss": 0.9259, + "step": 341350 + }, + { + "epoch": 29.55, + "learning_rate": 7.555838076800832e-07, + "loss": 0.8424, + "step": 341360 + }, + { + "epoch": 29.55, + "learning_rate": 7.54139096766737e-07, + "loss": 0.8874, + "step": 341370 + }, + { + "epoch": 29.55, + "learning_rate": 7.526943858533907e-07, + "loss": 0.9384, + "step": 341380 + }, + { + "epoch": 29.55, + "learning_rate": 7.512496749400446e-07, + "loss": 0.911, + "step": 341390 + }, + { + "epoch": 29.55, + "learning_rate": 7.498049640266983e-07, + "loss": 0.9604, + "step": 341400 + }, + { + "epoch": 29.55, + "learning_rate": 7.48360253113352e-07, + "loss": 0.9606, + "step": 341410 + }, + { + "epoch": 29.55, + "learning_rate": 7.469155422000058e-07, + "loss": 0.9123, + "step": 341420 + }, + { + "epoch": 29.55, + "learning_rate": 7.454708312866596e-07, + "loss": 0.9299, + "step": 341430 + }, + { + "epoch": 29.55, + "learning_rate": 7.440261203733133e-07, + "loss": 0.9024, + "step": 341440 + }, + { + "epoch": 29.55, + "learning_rate": 7.42581409459967e-07, + "loss": 0.8732, + "step": 341450 + }, + { + "epoch": 29.56, + "learning_rate": 7.411366985466209e-07, + "loss": 0.8762, + "step": 341460 + }, + { + "epoch": 29.56, + "learning_rate": 7.396919876332747e-07, + "loss": 0.8837, + "step": 341470 + }, + { + "epoch": 29.56, + "learning_rate": 7.382472767199284e-07, + "loss": 0.8987, + "step": 341480 + }, + { + "epoch": 29.56, + "learning_rate": 7.368025658065821e-07, + "loss": 0.9477, + "step": 341490 + }, + { + "epoch": 29.56, + "learning_rate": 7.353578548932358e-07, + "loss": 0.8876, + "step": 341500 + }, + { + "epoch": 29.56, + "learning_rate": 7.339131439798896e-07, + "loss": 0.8841, + "step": 341510 + }, + { + "epoch": 29.56, + "learning_rate": 7.324684330665435e-07, + "loss": 0.9349, + "step": 341520 + }, + { + "epoch": 29.56, + "learning_rate": 7.310237221531972e-07, + "loss": 0.8443, + "step": 341530 + }, + { + "epoch": 29.56, + "learning_rate": 7.295790112398509e-07, + "loss": 0.8472, + "step": 341540 + }, + { + "epoch": 29.56, + "learning_rate": 7.281343003265047e-07, + "loss": 0.8718, + "step": 341550 + }, + { + "epoch": 29.56, + "learning_rate": 7.266895894131585e-07, + "loss": 0.9097, + "step": 341560 + }, + { + "epoch": 29.57, + "learning_rate": 7.252448784998122e-07, + "loss": 0.871, + "step": 341570 + }, + { + "epoch": 29.57, + "learning_rate": 7.23800167586466e-07, + "loss": 0.9039, + "step": 341580 + }, + { + "epoch": 29.57, + "learning_rate": 7.223554566731197e-07, + "loss": 0.9131, + "step": 341590 + }, + { + "epoch": 29.57, + "learning_rate": 7.209107457597736e-07, + "loss": 0.9402, + "step": 341600 + }, + { + "epoch": 29.57, + "learning_rate": 7.194660348464273e-07, + "loss": 0.9026, + "step": 341610 + }, + { + "epoch": 29.57, + "learning_rate": 7.18021323933081e-07, + "loss": 0.9013, + "step": 341620 + }, + { + "epoch": 29.57, + "learning_rate": 7.165766130197347e-07, + "loss": 0.8515, + "step": 341630 + }, + { + "epoch": 29.57, + "learning_rate": 7.151319021063885e-07, + "loss": 0.9084, + "step": 341640 + }, + { + "epoch": 29.57, + "learning_rate": 7.136871911930424e-07, + "loss": 0.905, + "step": 341650 + }, + { + "epoch": 29.57, + "learning_rate": 7.122424802796961e-07, + "loss": 0.9586, + "step": 341660 + }, + { + "epoch": 29.57, + "learning_rate": 7.107977693663498e-07, + "loss": 0.9217, + "step": 341670 + }, + { + "epoch": 29.57, + "learning_rate": 7.093530584530036e-07, + "loss": 0.902, + "step": 341680 + }, + { + "epoch": 29.58, + "learning_rate": 7.079083475396574e-07, + "loss": 0.9176, + "step": 341690 + }, + { + "epoch": 29.58, + "learning_rate": 7.064636366263111e-07, + "loss": 0.9222, + "step": 341700 + }, + { + "epoch": 29.58, + "learning_rate": 7.050189257129649e-07, + "loss": 0.9439, + "step": 341710 + }, + { + "epoch": 29.58, + "learning_rate": 7.035742147996187e-07, + "loss": 0.9195, + "step": 341720 + }, + { + "epoch": 29.58, + "learning_rate": 7.021295038862724e-07, + "loss": 0.9427, + "step": 341730 + }, + { + "epoch": 29.58, + "learning_rate": 7.006847929729262e-07, + "loss": 0.9714, + "step": 341740 + }, + { + "epoch": 29.58, + "learning_rate": 6.992400820595799e-07, + "loss": 0.8696, + "step": 341750 + }, + { + "epoch": 29.58, + "learning_rate": 6.977953711462336e-07, + "loss": 0.9198, + "step": 341760 + }, + { + "epoch": 29.58, + "learning_rate": 6.963506602328874e-07, + "loss": 0.8997, + "step": 341770 + }, + { + "epoch": 29.58, + "learning_rate": 6.949059493195412e-07, + "loss": 0.9046, + "step": 341780 + }, + { + "epoch": 29.58, + "learning_rate": 6.93461238406195e-07, + "loss": 0.9051, + "step": 341790 + }, + { + "epoch": 29.59, + "learning_rate": 6.920165274928487e-07, + "loss": 0.9159, + "step": 341800 + }, + { + "epoch": 29.59, + "learning_rate": 6.905718165795025e-07, + "loss": 0.9232, + "step": 341810 + }, + { + "epoch": 29.59, + "learning_rate": 6.891271056661562e-07, + "loss": 0.8784, + "step": 341820 + }, + { + "epoch": 29.59, + "learning_rate": 6.8768239475281e-07, + "loss": 0.9387, + "step": 341830 + }, + { + "epoch": 29.59, + "learning_rate": 6.862376838394638e-07, + "loss": 0.9055, + "step": 341840 + }, + { + "epoch": 29.59, + "learning_rate": 6.847929729261176e-07, + "loss": 0.8853, + "step": 341850 + }, + { + "epoch": 29.59, + "learning_rate": 6.833482620127713e-07, + "loss": 0.8892, + "step": 341860 + }, + { + "epoch": 29.59, + "learning_rate": 6.81903551099425e-07, + "loss": 0.9492, + "step": 341870 + }, + { + "epoch": 29.59, + "learning_rate": 6.804588401860788e-07, + "loss": 0.9022, + "step": 341880 + }, + { + "epoch": 29.59, + "learning_rate": 6.790141292727325e-07, + "loss": 0.9199, + "step": 341890 + }, + { + "epoch": 29.59, + "learning_rate": 6.775694183593863e-07, + "loss": 0.8599, + "step": 341900 + }, + { + "epoch": 29.59, + "learning_rate": 6.761247074460401e-07, + "loss": 0.8782, + "step": 341910 + }, + { + "epoch": 29.6, + "learning_rate": 6.746799965326938e-07, + "loss": 0.9029, + "step": 341920 + }, + { + "epoch": 29.6, + "learning_rate": 6.732352856193476e-07, + "loss": 0.8705, + "step": 341930 + }, + { + "epoch": 29.6, + "learning_rate": 6.717905747060014e-07, + "loss": 0.8884, + "step": 341940 + }, + { + "epoch": 29.6, + "learning_rate": 6.703458637926551e-07, + "loss": 0.9148, + "step": 341950 + }, + { + "epoch": 29.6, + "learning_rate": 6.689011528793088e-07, + "loss": 0.8499, + "step": 341960 + }, + { + "epoch": 29.6, + "learning_rate": 6.674564419659626e-07, + "loss": 0.9049, + "step": 341970 + }, + { + "epoch": 29.6, + "learning_rate": 6.660117310526165e-07, + "loss": 0.9237, + "step": 341980 + }, + { + "epoch": 29.6, + "learning_rate": 6.645670201392702e-07, + "loss": 0.8725, + "step": 341990 + }, + { + "epoch": 29.6, + "learning_rate": 6.631223092259239e-07, + "loss": 0.9395, + "step": 342000 + }, + { + "epoch": 29.6, + "learning_rate": 6.616775983125776e-07, + "loss": 0.8897, + "step": 342010 + }, + { + "epoch": 29.6, + "learning_rate": 6.602328873992315e-07, + "loss": 0.8786, + "step": 342020 + }, + { + "epoch": 29.61, + "learning_rate": 6.587881764858852e-07, + "loss": 0.866, + "step": 342030 + }, + { + "epoch": 29.61, + "learning_rate": 6.57343465572539e-07, + "loss": 0.8903, + "step": 342040 + }, + { + "epoch": 29.61, + "learning_rate": 6.558987546591927e-07, + "loss": 0.9155, + "step": 342050 + }, + { + "epoch": 29.61, + "learning_rate": 6.544540437458465e-07, + "loss": 0.89, + "step": 342060 + }, + { + "epoch": 29.61, + "learning_rate": 6.530093328325003e-07, + "loss": 0.8483, + "step": 342070 + }, + { + "epoch": 29.61, + "learning_rate": 6.51564621919154e-07, + "loss": 0.8691, + "step": 342080 + }, + { + "epoch": 29.61, + "learning_rate": 6.501199110058077e-07, + "loss": 0.9317, + "step": 342090 + }, + { + "epoch": 29.61, + "learning_rate": 6.486752000924615e-07, + "loss": 0.8748, + "step": 342100 + }, + { + "epoch": 29.61, + "learning_rate": 6.472304891791154e-07, + "loss": 0.8047, + "step": 342110 + }, + { + "epoch": 29.61, + "learning_rate": 6.457857782657691e-07, + "loss": 0.86, + "step": 342120 + }, + { + "epoch": 29.61, + "learning_rate": 6.443410673524228e-07, + "loss": 0.886, + "step": 342130 + }, + { + "epoch": 29.61, + "learning_rate": 6.428963564390765e-07, + "loss": 0.9722, + "step": 342140 + }, + { + "epoch": 29.62, + "learning_rate": 6.414516455257303e-07, + "loss": 0.9144, + "step": 342150 + }, + { + "epoch": 29.62, + "learning_rate": 6.400069346123841e-07, + "loss": 0.9164, + "step": 342160 + }, + { + "epoch": 29.62, + "learning_rate": 6.385622236990379e-07, + "loss": 0.9211, + "step": 342170 + }, + { + "epoch": 29.62, + "learning_rate": 6.371175127856916e-07, + "loss": 0.8847, + "step": 342180 + }, + { + "epoch": 29.62, + "learning_rate": 6.356728018723454e-07, + "loss": 0.8613, + "step": 342190 + }, + { + "epoch": 29.62, + "learning_rate": 6.342280909589991e-07, + "loss": 0.8989, + "step": 342200 + }, + { + "epoch": 29.62, + "learning_rate": 6.327833800456529e-07, + "loss": 0.9303, + "step": 342210 + }, + { + "epoch": 29.62, + "learning_rate": 6.313386691323066e-07, + "loss": 0.9163, + "step": 342220 + }, + { + "epoch": 29.62, + "learning_rate": 6.298939582189604e-07, + "loss": 0.9218, + "step": 342230 + }, + { + "epoch": 29.62, + "learning_rate": 6.284492473056142e-07, + "loss": 0.893, + "step": 342240 + }, + { + "epoch": 29.62, + "learning_rate": 6.27004536392268e-07, + "loss": 0.819, + "step": 342250 + }, + { + "epoch": 29.62, + "learning_rate": 6.255598254789217e-07, + "loss": 0.8451, + "step": 342260 + }, + { + "epoch": 29.63, + "learning_rate": 6.241151145655755e-07, + "loss": 0.8909, + "step": 342270 + }, + { + "epoch": 29.63, + "learning_rate": 6.226704036522292e-07, + "loss": 0.8876, + "step": 342280 + }, + { + "epoch": 29.63, + "learning_rate": 6.212256927388829e-07, + "loss": 0.9399, + "step": 342290 + }, + { + "epoch": 29.63, + "learning_rate": 6.197809818255368e-07, + "loss": 0.9179, + "step": 342300 + }, + { + "epoch": 29.63, + "learning_rate": 6.183362709121905e-07, + "loss": 0.9094, + "step": 342310 + }, + { + "epoch": 29.63, + "learning_rate": 6.168915599988443e-07, + "loss": 0.956, + "step": 342320 + }, + { + "epoch": 29.63, + "learning_rate": 6.15446849085498e-07, + "loss": 0.9052, + "step": 342330 + }, + { + "epoch": 29.63, + "learning_rate": 6.140021381721518e-07, + "loss": 0.928, + "step": 342340 + }, + { + "epoch": 29.63, + "learning_rate": 6.125574272588055e-07, + "loss": 0.9015, + "step": 342350 + }, + { + "epoch": 29.63, + "learning_rate": 6.111127163454593e-07, + "loss": 0.8988, + "step": 342360 + }, + { + "epoch": 29.63, + "learning_rate": 6.096680054321131e-07, + "loss": 0.9026, + "step": 342370 + }, + { + "epoch": 29.64, + "learning_rate": 6.082232945187668e-07, + "loss": 0.9313, + "step": 342380 + }, + { + "epoch": 29.64, + "learning_rate": 6.067785836054206e-07, + "loss": 0.8909, + "step": 342390 + }, + { + "epoch": 29.64, + "learning_rate": 6.053338726920744e-07, + "loss": 0.9492, + "step": 342400 + }, + { + "epoch": 29.64, + "learning_rate": 6.038891617787281e-07, + "loss": 0.905, + "step": 342410 + }, + { + "epoch": 29.64, + "learning_rate": 6.024444508653818e-07, + "loss": 0.927, + "step": 342420 + }, + { + "epoch": 29.64, + "learning_rate": 6.009997399520355e-07, + "loss": 0.902, + "step": 342430 + }, + { + "epoch": 29.64, + "learning_rate": 5.995550290386894e-07, + "loss": 0.884, + "step": 342440 + }, + { + "epoch": 29.64, + "learning_rate": 5.981103181253432e-07, + "loss": 0.8956, + "step": 342450 + }, + { + "epoch": 29.64, + "learning_rate": 5.966656072119969e-07, + "loss": 0.9271, + "step": 342460 + }, + { + "epoch": 29.64, + "learning_rate": 5.952208962986506e-07, + "loss": 0.8932, + "step": 342470 + }, + { + "epoch": 29.64, + "learning_rate": 5.937761853853044e-07, + "loss": 0.9127, + "step": 342480 + }, + { + "epoch": 29.64, + "learning_rate": 5.923314744719582e-07, + "loss": 0.9193, + "step": 342490 + }, + { + "epoch": 29.65, + "learning_rate": 5.90886763558612e-07, + "loss": 0.8992, + "step": 342500 + }, + { + "epoch": 29.65, + "learning_rate": 5.894420526452657e-07, + "loss": 0.9099, + "step": 342510 + }, + { + "epoch": 29.65, + "learning_rate": 5.879973417319195e-07, + "loss": 0.8785, + "step": 342520 + }, + { + "epoch": 29.65, + "learning_rate": 5.865526308185733e-07, + "loss": 0.8758, + "step": 342530 + }, + { + "epoch": 29.65, + "learning_rate": 5.85107919905227e-07, + "loss": 0.8569, + "step": 342540 + }, + { + "epoch": 29.65, + "learning_rate": 5.836632089918807e-07, + "loss": 0.9084, + "step": 342550 + }, + { + "epoch": 29.65, + "learning_rate": 5.822184980785344e-07, + "loss": 0.831, + "step": 342560 + }, + { + "epoch": 29.65, + "learning_rate": 5.807737871651883e-07, + "loss": 0.8919, + "step": 342570 + }, + { + "epoch": 29.65, + "learning_rate": 5.793290762518421e-07, + "loss": 0.8276, + "step": 342580 + }, + { + "epoch": 29.65, + "learning_rate": 5.778843653384958e-07, + "loss": 0.8846, + "step": 342590 + }, + { + "epoch": 29.65, + "learning_rate": 5.764396544251495e-07, + "loss": 0.8307, + "step": 342600 + }, + { + "epoch": 29.66, + "learning_rate": 5.749949435118033e-07, + "loss": 0.8812, + "step": 342610 + }, + { + "epoch": 29.66, + "learning_rate": 5.735502325984571e-07, + "loss": 0.9119, + "step": 342620 + }, + { + "epoch": 29.66, + "learning_rate": 5.721055216851109e-07, + "loss": 0.9269, + "step": 342630 + }, + { + "epoch": 29.66, + "learning_rate": 5.706608107717646e-07, + "loss": 0.8805, + "step": 342640 + }, + { + "epoch": 29.66, + "learning_rate": 5.692160998584184e-07, + "loss": 0.9058, + "step": 342650 + }, + { + "epoch": 29.66, + "learning_rate": 5.677713889450721e-07, + "loss": 0.8689, + "step": 342660 + }, + { + "epoch": 29.66, + "learning_rate": 5.663266780317259e-07, + "loss": 0.932, + "step": 342670 + }, + { + "epoch": 29.66, + "learning_rate": 5.648819671183796e-07, + "loss": 0.8594, + "step": 342680 + }, + { + "epoch": 29.66, + "learning_rate": 5.634372562050333e-07, + "loss": 0.9217, + "step": 342690 + }, + { + "epoch": 29.66, + "learning_rate": 5.619925452916872e-07, + "loss": 0.8691, + "step": 342700 + }, + { + "epoch": 29.66, + "learning_rate": 5.605478343783409e-07, + "loss": 0.9376, + "step": 342710 + }, + { + "epoch": 29.66, + "learning_rate": 5.591031234649947e-07, + "loss": 0.9039, + "step": 342720 + }, + { + "epoch": 29.67, + "learning_rate": 5.576584125516484e-07, + "loss": 0.8969, + "step": 342730 + }, + { + "epoch": 29.67, + "learning_rate": 5.562137016383022e-07, + "loss": 0.9006, + "step": 342740 + }, + { + "epoch": 29.67, + "learning_rate": 5.547689907249559e-07, + "loss": 0.9058, + "step": 342750 + }, + { + "epoch": 29.67, + "learning_rate": 5.533242798116098e-07, + "loss": 0.8755, + "step": 342760 + }, + { + "epoch": 29.67, + "learning_rate": 5.518795688982635e-07, + "loss": 0.8454, + "step": 342770 + }, + { + "epoch": 29.67, + "learning_rate": 5.504348579849173e-07, + "loss": 0.8792, + "step": 342780 + }, + { + "epoch": 29.67, + "learning_rate": 5.48990147071571e-07, + "loss": 0.8915, + "step": 342790 + }, + { + "epoch": 29.67, + "learning_rate": 5.475454361582247e-07, + "loss": 0.9067, + "step": 342800 + }, + { + "epoch": 29.67, + "learning_rate": 5.461007252448785e-07, + "loss": 0.8765, + "step": 342810 + }, + { + "epoch": 29.67, + "learning_rate": 5.446560143315323e-07, + "loss": 0.983, + "step": 342820 + }, + { + "epoch": 29.67, + "learning_rate": 5.432113034181861e-07, + "loss": 0.8988, + "step": 342830 + }, + { + "epoch": 29.68, + "learning_rate": 5.417665925048398e-07, + "loss": 0.8847, + "step": 342840 + }, + { + "epoch": 29.68, + "learning_rate": 5.403218815914935e-07, + "loss": 0.8979, + "step": 342850 + }, + { + "epoch": 29.68, + "learning_rate": 5.388771706781473e-07, + "loss": 0.8421, + "step": 342860 + }, + { + "epoch": 29.68, + "learning_rate": 5.374324597648011e-07, + "loss": 0.9301, + "step": 342870 + }, + { + "epoch": 29.68, + "learning_rate": 5.359877488514548e-07, + "loss": 0.9025, + "step": 342880 + }, + { + "epoch": 29.68, + "learning_rate": 5.345430379381085e-07, + "loss": 0.8952, + "step": 342890 + }, + { + "epoch": 29.68, + "learning_rate": 5.330983270247624e-07, + "loss": 0.887, + "step": 342900 + }, + { + "epoch": 29.68, + "learning_rate": 5.316536161114162e-07, + "loss": 0.9172, + "step": 342910 + }, + { + "epoch": 29.68, + "learning_rate": 5.302089051980699e-07, + "loss": 0.8665, + "step": 342920 + }, + { + "epoch": 29.68, + "learning_rate": 5.287641942847236e-07, + "loss": 0.8752, + "step": 342930 + }, + { + "epoch": 29.68, + "learning_rate": 5.273194833713773e-07, + "loss": 0.9198, + "step": 342940 + }, + { + "epoch": 29.68, + "learning_rate": 5.258747724580312e-07, + "loss": 0.8741, + "step": 342950 + }, + { + "epoch": 29.69, + "learning_rate": 5.24430061544685e-07, + "loss": 0.9226, + "step": 342960 + }, + { + "epoch": 29.69, + "learning_rate": 5.229853506313387e-07, + "loss": 0.9359, + "step": 342970 + }, + { + "epoch": 29.69, + "learning_rate": 5.215406397179924e-07, + "loss": 0.9517, + "step": 342980 + }, + { + "epoch": 29.69, + "learning_rate": 5.200959288046462e-07, + "loss": 0.9357, + "step": 342990 + }, + { + "epoch": 29.69, + "learning_rate": 5.186512178913e-07, + "loss": 0.8658, + "step": 343000 + }, + { + "epoch": 29.69, + "learning_rate": 5.172065069779537e-07, + "loss": 0.893, + "step": 343010 + }, + { + "epoch": 29.69, + "learning_rate": 5.157617960646074e-07, + "loss": 0.8729, + "step": 343020 + }, + { + "epoch": 29.69, + "learning_rate": 5.143170851512613e-07, + "loss": 0.8692, + "step": 343030 + }, + { + "epoch": 29.69, + "learning_rate": 5.128723742379151e-07, + "loss": 0.9007, + "step": 343040 + }, + { + "epoch": 29.69, + "learning_rate": 5.114276633245688e-07, + "loss": 0.9665, + "step": 343050 + }, + { + "epoch": 29.69, + "learning_rate": 5.099829524112225e-07, + "loss": 0.9083, + "step": 343060 + }, + { + "epoch": 29.7, + "learning_rate": 5.085382414978763e-07, + "loss": 0.8769, + "step": 343070 + }, + { + "epoch": 29.7, + "learning_rate": 5.0709353058453e-07, + "loss": 0.8994, + "step": 343080 + }, + { + "epoch": 29.7, + "learning_rate": 5.056488196711839e-07, + "loss": 0.9246, + "step": 343090 + }, + { + "epoch": 29.7, + "learning_rate": 5.042041087578376e-07, + "loss": 0.8974, + "step": 343100 + }, + { + "epoch": 29.7, + "learning_rate": 5.027593978444913e-07, + "loss": 0.8418, + "step": 343110 + }, + { + "epoch": 29.7, + "learning_rate": 5.013146869311451e-07, + "loss": 0.8893, + "step": 343120 + }, + { + "epoch": 29.7, + "learning_rate": 4.998699760177989e-07, + "loss": 0.8852, + "step": 343130 + }, + { + "epoch": 29.7, + "learning_rate": 4.984252651044526e-07, + "loss": 0.9416, + "step": 343140 + }, + { + "epoch": 29.7, + "learning_rate": 4.969805541911063e-07, + "loss": 0.8837, + "step": 343150 + }, + { + "epoch": 29.7, + "learning_rate": 4.955358432777602e-07, + "loss": 0.8954, + "step": 343160 + }, + { + "epoch": 29.7, + "learning_rate": 4.940911323644139e-07, + "loss": 0.8967, + "step": 343170 + }, + { + "epoch": 29.7, + "learning_rate": 4.926464214510677e-07, + "loss": 0.843, + "step": 343180 + }, + { + "epoch": 29.71, + "learning_rate": 4.912017105377214e-07, + "loss": 0.9191, + "step": 343190 + }, + { + "epoch": 29.71, + "learning_rate": 4.897569996243752e-07, + "loss": 0.9138, + "step": 343200 + }, + { + "epoch": 29.71, + "learning_rate": 4.883122887110289e-07, + "loss": 0.886, + "step": 343210 + }, + { + "epoch": 29.71, + "learning_rate": 4.868675777976827e-07, + "loss": 0.893, + "step": 343220 + }, + { + "epoch": 29.71, + "learning_rate": 4.854228668843365e-07, + "loss": 0.8784, + "step": 343230 + }, + { + "epoch": 29.71, + "learning_rate": 4.839781559709903e-07, + "loss": 0.9609, + "step": 343240 + }, + { + "epoch": 29.71, + "learning_rate": 4.82533445057644e-07, + "loss": 0.8812, + "step": 343250 + }, + { + "epoch": 29.71, + "learning_rate": 4.810887341442977e-07, + "loss": 0.9507, + "step": 343260 + }, + { + "epoch": 29.71, + "learning_rate": 4.796440232309515e-07, + "loss": 0.9205, + "step": 343270 + }, + { + "epoch": 29.71, + "learning_rate": 4.781993123176052e-07, + "loss": 0.8835, + "step": 343280 + }, + { + "epoch": 29.71, + "learning_rate": 4.76754601404259e-07, + "loss": 0.929, + "step": 343290 + }, + { + "epoch": 29.71, + "learning_rate": 4.753098904909128e-07, + "loss": 0.8737, + "step": 343300 + }, + { + "epoch": 29.72, + "learning_rate": 4.7386517957756654e-07, + "loss": 0.8963, + "step": 343310 + }, + { + "epoch": 29.72, + "learning_rate": 4.7242046866422036e-07, + "loss": 0.8447, + "step": 343320 + }, + { + "epoch": 29.72, + "learning_rate": 4.709757577508741e-07, + "loss": 0.8826, + "step": 343330 + }, + { + "epoch": 29.72, + "learning_rate": 4.6953104683752784e-07, + "loss": 0.9306, + "step": 343340 + }, + { + "epoch": 29.72, + "learning_rate": 4.6808633592418155e-07, + "loss": 0.9578, + "step": 343350 + }, + { + "epoch": 29.72, + "learning_rate": 4.666416250108353e-07, + "loss": 0.8713, + "step": 343360 + }, + { + "epoch": 29.72, + "learning_rate": 4.6519691409748914e-07, + "loss": 0.9069, + "step": 343370 + }, + { + "epoch": 29.72, + "learning_rate": 4.637522031841429e-07, + "loss": 0.9006, + "step": 343380 + }, + { + "epoch": 29.72, + "learning_rate": 4.623074922707966e-07, + "loss": 0.9001, + "step": 343390 + }, + { + "epoch": 29.72, + "learning_rate": 4.608627813574504e-07, + "loss": 0.899, + "step": 343400 + }, + { + "epoch": 29.72, + "learning_rate": 4.594180704441042e-07, + "loss": 0.9993, + "step": 343410 + }, + { + "epoch": 29.73, + "learning_rate": 4.5797335953075796e-07, + "loss": 0.9751, + "step": 343420 + }, + { + "epoch": 29.73, + "learning_rate": 4.565286486174117e-07, + "loss": 0.8921, + "step": 343430 + }, + { + "epoch": 29.73, + "learning_rate": 4.5508393770406544e-07, + "loss": 0.8974, + "step": 343440 + }, + { + "epoch": 29.73, + "learning_rate": 4.5363922679071916e-07, + "loss": 0.897, + "step": 343450 + }, + { + "epoch": 29.73, + "learning_rate": 4.52194515877373e-07, + "loss": 0.9327, + "step": 343460 + }, + { + "epoch": 29.73, + "learning_rate": 4.5074980496402674e-07, + "loss": 0.9094, + "step": 343470 + }, + { + "epoch": 29.73, + "learning_rate": 4.4930509405068045e-07, + "loss": 0.8912, + "step": 343480 + }, + { + "epoch": 29.73, + "learning_rate": 4.478603831373342e-07, + "loss": 0.8283, + "step": 343490 + }, + { + "epoch": 29.73, + "learning_rate": 4.46415672223988e-07, + "loss": 0.8878, + "step": 343500 + }, + { + "epoch": 29.73, + "learning_rate": 4.449709613106418e-07, + "loss": 0.8452, + "step": 343510 + }, + { + "epoch": 29.73, + "learning_rate": 4.435262503972955e-07, + "loss": 0.8479, + "step": 343520 + }, + { + "epoch": 29.73, + "learning_rate": 4.420815394839493e-07, + "loss": 0.8082, + "step": 343530 + }, + { + "epoch": 29.74, + "learning_rate": 4.40636828570603e-07, + "loss": 0.9028, + "step": 343540 + }, + { + "epoch": 29.74, + "learning_rate": 4.3919211765725686e-07, + "loss": 0.9313, + "step": 343550 + }, + { + "epoch": 29.74, + "learning_rate": 4.377474067439106e-07, + "loss": 0.9135, + "step": 343560 + }, + { + "epoch": 29.74, + "learning_rate": 4.3630269583056434e-07, + "loss": 0.9355, + "step": 343570 + }, + { + "epoch": 29.74, + "learning_rate": 4.3485798491721805e-07, + "loss": 0.8569, + "step": 343580 + }, + { + "epoch": 29.74, + "learning_rate": 4.334132740038718e-07, + "loss": 0.8775, + "step": 343590 + }, + { + "epoch": 29.74, + "learning_rate": 4.3196856309052564e-07, + "loss": 0.8845, + "step": 343600 + }, + { + "epoch": 29.74, + "learning_rate": 4.305238521771794e-07, + "loss": 0.877, + "step": 343610 + }, + { + "epoch": 29.74, + "learning_rate": 4.290791412638331e-07, + "loss": 0.8684, + "step": 343620 + }, + { + "epoch": 29.74, + "learning_rate": 4.276344303504869e-07, + "loss": 0.9224, + "step": 343630 + }, + { + "epoch": 29.74, + "learning_rate": 4.261897194371406e-07, + "loss": 0.951, + "step": 343640 + }, + { + "epoch": 29.75, + "learning_rate": 4.247450085237944e-07, + "loss": 0.886, + "step": 343650 + }, + { + "epoch": 29.75, + "learning_rate": 4.233002976104482e-07, + "loss": 0.8953, + "step": 343660 + }, + { + "epoch": 29.75, + "learning_rate": 4.2185558669710194e-07, + "loss": 0.9223, + "step": 343670 + }, + { + "epoch": 29.75, + "learning_rate": 4.2041087578375566e-07, + "loss": 0.8973, + "step": 343680 + }, + { + "epoch": 29.75, + "learning_rate": 4.189661648704095e-07, + "loss": 0.9002, + "step": 343690 + }, + { + "epoch": 29.75, + "learning_rate": 4.1752145395706324e-07, + "loss": 0.9393, + "step": 343700 + }, + { + "epoch": 29.75, + "learning_rate": 4.1607674304371695e-07, + "loss": 0.8757, + "step": 343710 + }, + { + "epoch": 29.75, + "learning_rate": 4.146320321303707e-07, + "loss": 0.9152, + "step": 343720 + }, + { + "epoch": 29.75, + "learning_rate": 4.1318732121702443e-07, + "loss": 0.9349, + "step": 343730 + }, + { + "epoch": 29.75, + "learning_rate": 4.117426103036783e-07, + "loss": 0.8702, + "step": 343740 + }, + { + "epoch": 29.75, + "learning_rate": 4.10297899390332e-07, + "loss": 0.9179, + "step": 343750 + }, + { + "epoch": 29.75, + "learning_rate": 4.088531884769858e-07, + "loss": 0.9093, + "step": 343760 + }, + { + "epoch": 29.76, + "learning_rate": 4.074084775636395e-07, + "loss": 0.8816, + "step": 343770 + }, + { + "epoch": 29.76, + "learning_rate": 4.0596376665029337e-07, + "loss": 0.8549, + "step": 343780 + }, + { + "epoch": 29.76, + "learning_rate": 4.045190557369471e-07, + "loss": 0.9069, + "step": 343790 + }, + { + "epoch": 29.76, + "learning_rate": 4.0307434482360084e-07, + "loss": 0.8534, + "step": 343800 + }, + { + "epoch": 29.76, + "learning_rate": 4.0162963391025456e-07, + "loss": 0.9358, + "step": 343810 + }, + { + "epoch": 29.76, + "learning_rate": 4.001849229969083e-07, + "loss": 0.8825, + "step": 343820 + }, + { + "epoch": 29.76, + "learning_rate": 3.9874021208356214e-07, + "loss": 0.9279, + "step": 343830 + }, + { + "epoch": 29.76, + "learning_rate": 3.9729550117021585e-07, + "loss": 0.9021, + "step": 343840 + }, + { + "epoch": 29.76, + "learning_rate": 3.958507902568696e-07, + "loss": 0.8551, + "step": 343850 + }, + { + "epoch": 29.76, + "learning_rate": 3.944060793435234e-07, + "loss": 0.9289, + "step": 343860 + }, + { + "epoch": 29.76, + "learning_rate": 3.929613684301771e-07, + "loss": 0.8653, + "step": 343870 + }, + { + "epoch": 29.77, + "learning_rate": 3.915166575168309e-07, + "loss": 0.8811, + "step": 343880 + }, + { + "epoch": 29.77, + "learning_rate": 3.900719466034847e-07, + "loss": 0.9051, + "step": 343890 + }, + { + "epoch": 29.77, + "learning_rate": 3.886272356901384e-07, + "loss": 0.936, + "step": 343900 + }, + { + "epoch": 29.77, + "learning_rate": 3.871825247767922e-07, + "loss": 0.8943, + "step": 343910 + }, + { + "epoch": 29.77, + "learning_rate": 3.857378138634459e-07, + "loss": 0.8484, + "step": 343920 + }, + { + "epoch": 29.77, + "learning_rate": 3.842931029500997e-07, + "loss": 0.8967, + "step": 343930 + }, + { + "epoch": 29.77, + "learning_rate": 3.8284839203675346e-07, + "loss": 0.8976, + "step": 343940 + }, + { + "epoch": 29.77, + "learning_rate": 3.814036811234072e-07, + "loss": 0.8881, + "step": 343950 + }, + { + "epoch": 29.77, + "learning_rate": 3.79958970210061e-07, + "loss": 0.9307, + "step": 343960 + }, + { + "epoch": 29.77, + "learning_rate": 3.7851425929671475e-07, + "loss": 0.8961, + "step": 343970 + }, + { + "epoch": 29.77, + "learning_rate": 3.770695483833685e-07, + "loss": 0.9263, + "step": 343980 + }, + { + "epoch": 29.77, + "learning_rate": 3.756248374700223e-07, + "loss": 0.9523, + "step": 343990 + }, + { + "epoch": 29.78, + "learning_rate": 3.74180126556676e-07, + "loss": 0.9267, + "step": 344000 + }, + { + "epoch": 29.78, + "learning_rate": 3.727354156433298e-07, + "loss": 0.9534, + "step": 344010 + }, + { + "epoch": 29.78, + "learning_rate": 3.712907047299835e-07, + "loss": 0.8874, + "step": 344020 + }, + { + "epoch": 29.78, + "learning_rate": 3.6984599381663734e-07, + "loss": 0.964, + "step": 344030 + }, + { + "epoch": 29.78, + "learning_rate": 3.6840128290329106e-07, + "loss": 0.8945, + "step": 344040 + }, + { + "epoch": 29.78, + "learning_rate": 3.669565719899448e-07, + "loss": 0.9105, + "step": 344050 + }, + { + "epoch": 29.78, + "learning_rate": 3.655118610765986e-07, + "loss": 0.8615, + "step": 344060 + }, + { + "epoch": 29.78, + "learning_rate": 3.6406715016325235e-07, + "loss": 0.8825, + "step": 344070 + }, + { + "epoch": 29.78, + "learning_rate": 3.626224392499061e-07, + "loss": 0.9102, + "step": 344080 + }, + { + "epoch": 29.78, + "learning_rate": 3.6117772833655983e-07, + "loss": 0.8473, + "step": 344090 + }, + { + "epoch": 29.78, + "learning_rate": 3.5973301742321365e-07, + "loss": 0.911, + "step": 344100 + }, + { + "epoch": 29.79, + "learning_rate": 3.5828830650986736e-07, + "loss": 0.9076, + "step": 344110 + }, + { + "epoch": 29.79, + "learning_rate": 3.568435955965212e-07, + "loss": 0.85, + "step": 344120 + }, + { + "epoch": 29.79, + "learning_rate": 3.553988846831749e-07, + "loss": 0.9113, + "step": 344130 + }, + { + "epoch": 29.79, + "learning_rate": 3.539541737698287e-07, + "loss": 0.848, + "step": 344140 + }, + { + "epoch": 29.79, + "learning_rate": 3.525094628564824e-07, + "loss": 0.8902, + "step": 344150 + }, + { + "epoch": 29.79, + "learning_rate": 3.510647519431362e-07, + "loss": 0.9374, + "step": 344160 + }, + { + "epoch": 29.79, + "learning_rate": 3.4962004102978996e-07, + "loss": 0.9237, + "step": 344170 + }, + { + "epoch": 29.79, + "learning_rate": 3.481753301164437e-07, + "loss": 0.8608, + "step": 344180 + }, + { + "epoch": 29.79, + "learning_rate": 3.467306192030975e-07, + "loss": 0.8774, + "step": 344190 + }, + { + "epoch": 29.79, + "learning_rate": 3.4528590828975125e-07, + "loss": 0.8976, + "step": 344200 + }, + { + "epoch": 29.79, + "learning_rate": 3.43841197376405e-07, + "loss": 0.8967, + "step": 344210 + }, + { + "epoch": 29.79, + "learning_rate": 3.423964864630588e-07, + "loss": 0.926, + "step": 344220 + }, + { + "epoch": 29.8, + "learning_rate": 3.409517755497125e-07, + "loss": 0.8836, + "step": 344230 + }, + { + "epoch": 29.8, + "learning_rate": 3.3950706463636626e-07, + "loss": 0.9187, + "step": 344240 + }, + { + "epoch": 29.8, + "learning_rate": 3.3806235372302003e-07, + "loss": 0.8553, + "step": 344250 + }, + { + "epoch": 29.8, + "learning_rate": 3.366176428096738e-07, + "loss": 0.8834, + "step": 344260 + }, + { + "epoch": 29.8, + "learning_rate": 3.3517293189632756e-07, + "loss": 0.8553, + "step": 344270 + }, + { + "epoch": 29.8, + "learning_rate": 3.337282209829813e-07, + "loss": 0.9687, + "step": 344280 + }, + { + "epoch": 29.8, + "learning_rate": 3.322835100696351e-07, + "loss": 0.8478, + "step": 344290 + }, + { + "epoch": 29.8, + "learning_rate": 3.308387991562888e-07, + "loss": 0.8938, + "step": 344300 + }, + { + "epoch": 29.8, + "learning_rate": 3.293940882429426e-07, + "loss": 0.9304, + "step": 344310 + }, + { + "epoch": 29.8, + "learning_rate": 3.2794937732959633e-07, + "loss": 0.7879, + "step": 344320 + }, + { + "epoch": 29.8, + "learning_rate": 3.2650466641625015e-07, + "loss": 0.8533, + "step": 344330 + }, + { + "epoch": 29.8, + "learning_rate": 3.2505995550290386e-07, + "loss": 0.8783, + "step": 344340 + }, + { + "epoch": 29.81, + "learning_rate": 3.236152445895577e-07, + "loss": 0.8355, + "step": 344350 + }, + { + "epoch": 29.81, + "learning_rate": 3.221705336762114e-07, + "loss": 0.8499, + "step": 344360 + }, + { + "epoch": 29.81, + "learning_rate": 3.2072582276286516e-07, + "loss": 0.8744, + "step": 344370 + }, + { + "epoch": 29.81, + "learning_rate": 3.1928111184951893e-07, + "loss": 0.8956, + "step": 344380 + }, + { + "epoch": 29.81, + "learning_rate": 3.178364009361727e-07, + "loss": 0.8705, + "step": 344390 + }, + { + "epoch": 29.81, + "learning_rate": 3.1639169002282646e-07, + "loss": 0.9734, + "step": 344400 + }, + { + "epoch": 29.81, + "learning_rate": 3.149469791094802e-07, + "loss": 0.9342, + "step": 344410 + }, + { + "epoch": 29.81, + "learning_rate": 3.13502268196134e-07, + "loss": 0.8938, + "step": 344420 + }, + { + "epoch": 29.81, + "learning_rate": 3.1205755728278775e-07, + "loss": 0.8372, + "step": 344430 + }, + { + "epoch": 29.81, + "learning_rate": 3.1061284636944147e-07, + "loss": 0.8772, + "step": 344440 + }, + { + "epoch": 29.81, + "learning_rate": 3.0916813545609523e-07, + "loss": 0.8986, + "step": 344450 + }, + { + "epoch": 29.82, + "learning_rate": 3.07723424542749e-07, + "loss": 0.8535, + "step": 344460 + }, + { + "epoch": 29.82, + "learning_rate": 3.0627871362940276e-07, + "loss": 0.8919, + "step": 344470 + }, + { + "epoch": 29.82, + "learning_rate": 3.0483400271605653e-07, + "loss": 0.8635, + "step": 344480 + }, + { + "epoch": 29.82, + "learning_rate": 3.033892918027103e-07, + "loss": 0.9262, + "step": 344490 + }, + { + "epoch": 29.82, + "learning_rate": 3.0194458088936406e-07, + "loss": 0.8923, + "step": 344500 + }, + { + "epoch": 29.82, + "learning_rate": 3.0049986997601777e-07, + "loss": 0.8793, + "step": 344510 + }, + { + "epoch": 29.82, + "learning_rate": 2.990551590626716e-07, + "loss": 0.8791, + "step": 344520 + }, + { + "epoch": 29.82, + "learning_rate": 2.976104481493253e-07, + "loss": 0.8919, + "step": 344530 + }, + { + "epoch": 29.82, + "learning_rate": 2.961657372359791e-07, + "loss": 0.9182, + "step": 344540 + }, + { + "epoch": 29.82, + "learning_rate": 2.9472102632263284e-07, + "loss": 0.8872, + "step": 344550 + }, + { + "epoch": 29.82, + "learning_rate": 2.9327631540928665e-07, + "loss": 0.865, + "step": 344560 + }, + { + "epoch": 29.82, + "learning_rate": 2.9183160449594037e-07, + "loss": 0.906, + "step": 344570 + }, + { + "epoch": 29.83, + "learning_rate": 2.9038689358259413e-07, + "loss": 0.8624, + "step": 344580 + }, + { + "epoch": 29.83, + "learning_rate": 2.889421826692479e-07, + "loss": 0.8868, + "step": 344590 + }, + { + "epoch": 29.83, + "learning_rate": 2.8749747175590166e-07, + "loss": 0.89, + "step": 344600 + }, + { + "epoch": 29.83, + "learning_rate": 2.8605276084255543e-07, + "loss": 0.8991, + "step": 344610 + }, + { + "epoch": 29.83, + "learning_rate": 2.846080499292092e-07, + "loss": 0.8905, + "step": 344620 + }, + { + "epoch": 29.83, + "learning_rate": 2.8316333901586296e-07, + "loss": 0.9586, + "step": 344630 + }, + { + "epoch": 29.83, + "learning_rate": 2.8171862810251667e-07, + "loss": 0.9184, + "step": 344640 + }, + { + "epoch": 29.83, + "learning_rate": 2.8027391718917044e-07, + "loss": 0.8747, + "step": 344650 + }, + { + "epoch": 29.83, + "learning_rate": 2.788292062758242e-07, + "loss": 0.8308, + "step": 344660 + }, + { + "epoch": 29.83, + "learning_rate": 2.7738449536247797e-07, + "loss": 0.883, + "step": 344670 + }, + { + "epoch": 29.83, + "learning_rate": 2.7593978444913173e-07, + "loss": 0.8561, + "step": 344680 + }, + { + "epoch": 29.84, + "learning_rate": 2.744950735357855e-07, + "loss": 0.8604, + "step": 344690 + }, + { + "epoch": 29.84, + "learning_rate": 2.7305036262243927e-07, + "loss": 0.9166, + "step": 344700 + }, + { + "epoch": 29.84, + "learning_rate": 2.7160565170909303e-07, + "loss": 0.9116, + "step": 344710 + }, + { + "epoch": 29.84, + "learning_rate": 2.7016094079574674e-07, + "loss": 0.9219, + "step": 344720 + }, + { + "epoch": 29.84, + "learning_rate": 2.6871622988240056e-07, + "loss": 0.8506, + "step": 344730 + }, + { + "epoch": 29.84, + "learning_rate": 2.672715189690543e-07, + "loss": 0.8697, + "step": 344740 + }, + { + "epoch": 29.84, + "learning_rate": 2.658268080557081e-07, + "loss": 0.926, + "step": 344750 + }, + { + "epoch": 29.84, + "learning_rate": 2.643820971423618e-07, + "loss": 0.8742, + "step": 344760 + }, + { + "epoch": 29.84, + "learning_rate": 2.629373862290156e-07, + "loss": 0.9029, + "step": 344770 + }, + { + "epoch": 29.84, + "learning_rate": 2.6149267531566934e-07, + "loss": 0.8502, + "step": 344780 + }, + { + "epoch": 29.84, + "learning_rate": 2.600479644023231e-07, + "loss": 0.8799, + "step": 344790 + }, + { + "epoch": 29.84, + "learning_rate": 2.5860325348897687e-07, + "loss": 0.9393, + "step": 344800 + }, + { + "epoch": 29.85, + "learning_rate": 2.5715854257563063e-07, + "loss": 0.8539, + "step": 344810 + }, + { + "epoch": 29.85, + "learning_rate": 2.557138316622844e-07, + "loss": 0.8915, + "step": 344820 + }, + { + "epoch": 29.85, + "learning_rate": 2.5426912074893816e-07, + "loss": 0.9562, + "step": 344830 + }, + { + "epoch": 29.85, + "learning_rate": 2.5282440983559193e-07, + "loss": 0.9125, + "step": 344840 + }, + { + "epoch": 29.85, + "learning_rate": 2.5137969892224564e-07, + "loss": 0.8649, + "step": 344850 + }, + { + "epoch": 29.85, + "learning_rate": 2.4993498800889946e-07, + "loss": 0.9975, + "step": 344860 + }, + { + "epoch": 29.85, + "learning_rate": 2.484902770955532e-07, + "loss": 0.9072, + "step": 344870 + }, + { + "epoch": 29.85, + "learning_rate": 2.4704556618220694e-07, + "loss": 0.8598, + "step": 344880 + }, + { + "epoch": 29.85, + "learning_rate": 2.456008552688607e-07, + "loss": 0.8904, + "step": 344890 + }, + { + "epoch": 29.85, + "learning_rate": 2.4415614435551447e-07, + "loss": 0.8773, + "step": 344900 + }, + { + "epoch": 29.85, + "learning_rate": 2.4271143344216824e-07, + "loss": 0.9156, + "step": 344910 + }, + { + "epoch": 29.86, + "learning_rate": 2.41266722528822e-07, + "loss": 0.9028, + "step": 344920 + }, + { + "epoch": 29.86, + "learning_rate": 2.3982201161547577e-07, + "loss": 0.9478, + "step": 344930 + }, + { + "epoch": 29.86, + "learning_rate": 2.383773007021295e-07, + "loss": 0.9268, + "step": 344940 + }, + { + "epoch": 29.86, + "learning_rate": 2.3693258978878327e-07, + "loss": 0.8994, + "step": 344950 + }, + { + "epoch": 29.86, + "learning_rate": 2.3548787887543704e-07, + "loss": 0.8944, + "step": 344960 + }, + { + "epoch": 29.86, + "learning_rate": 2.3404316796209078e-07, + "loss": 0.9573, + "step": 344970 + }, + { + "epoch": 29.86, + "learning_rate": 2.3259845704874457e-07, + "loss": 0.8997, + "step": 344980 + }, + { + "epoch": 29.86, + "learning_rate": 2.311537461353983e-07, + "loss": 0.9279, + "step": 344990 + }, + { + "epoch": 29.86, + "learning_rate": 2.297090352220521e-07, + "loss": 0.8708, + "step": 345000 + }, + { + "epoch": 29.86, + "learning_rate": 2.2826432430870584e-07, + "loss": 0.857, + "step": 345010 + }, + { + "epoch": 29.86, + "learning_rate": 2.2681961339535958e-07, + "loss": 0.8954, + "step": 345020 + }, + { + "epoch": 29.86, + "learning_rate": 2.2537490248201337e-07, + "loss": 0.8659, + "step": 345030 + }, + { + "epoch": 29.87, + "learning_rate": 2.239301915686671e-07, + "loss": 0.902, + "step": 345040 + }, + { + "epoch": 29.87, + "learning_rate": 2.224854806553209e-07, + "loss": 0.9113, + "step": 345050 + }, + { + "epoch": 29.87, + "learning_rate": 2.2104076974197464e-07, + "loss": 0.8758, + "step": 345060 + }, + { + "epoch": 29.87, + "learning_rate": 2.1959605882862843e-07, + "loss": 0.9352, + "step": 345070 + }, + { + "epoch": 29.87, + "learning_rate": 2.1815134791528217e-07, + "loss": 0.8554, + "step": 345080 + }, + { + "epoch": 29.87, + "learning_rate": 2.167066370019359e-07, + "loss": 0.8703, + "step": 345090 + }, + { + "epoch": 29.87, + "learning_rate": 2.152619260885897e-07, + "loss": 0.8276, + "step": 345100 + }, + { + "epoch": 29.87, + "learning_rate": 2.1381721517524344e-07, + "loss": 0.907, + "step": 345110 + }, + { + "epoch": 29.87, + "learning_rate": 2.123725042618972e-07, + "loss": 0.9355, + "step": 345120 + }, + { + "epoch": 29.87, + "learning_rate": 2.1092779334855097e-07, + "loss": 0.8757, + "step": 345130 + }, + { + "epoch": 29.87, + "learning_rate": 2.0948308243520474e-07, + "loss": 0.8916, + "step": 345140 + }, + { + "epoch": 29.88, + "learning_rate": 2.0803837152185848e-07, + "loss": 0.8682, + "step": 345150 + }, + { + "epoch": 29.88, + "learning_rate": 2.0659366060851222e-07, + "loss": 0.8731, + "step": 345160 + }, + { + "epoch": 29.88, + "learning_rate": 2.05148949695166e-07, + "loss": 0.9673, + "step": 345170 + }, + { + "epoch": 29.88, + "learning_rate": 2.0370423878181975e-07, + "loss": 0.9668, + "step": 345180 + }, + { + "epoch": 29.88, + "learning_rate": 2.0225952786847354e-07, + "loss": 0.9474, + "step": 345190 + }, + { + "epoch": 29.88, + "learning_rate": 2.0081481695512728e-07, + "loss": 0.8465, + "step": 345200 + }, + { + "epoch": 29.88, + "learning_rate": 1.9937010604178107e-07, + "loss": 0.9161, + "step": 345210 + }, + { + "epoch": 29.88, + "learning_rate": 1.979253951284348e-07, + "loss": 0.918, + "step": 345220 + }, + { + "epoch": 29.88, + "learning_rate": 1.9648068421508855e-07, + "loss": 0.8858, + "step": 345230 + }, + { + "epoch": 29.88, + "learning_rate": 1.9503597330174234e-07, + "loss": 0.8756, + "step": 345240 + }, + { + "epoch": 29.88, + "learning_rate": 1.935912623883961e-07, + "loss": 0.8947, + "step": 345250 + }, + { + "epoch": 29.88, + "learning_rate": 1.9214655147504984e-07, + "loss": 0.9398, + "step": 345260 + }, + { + "epoch": 29.89, + "learning_rate": 1.907018405617036e-07, + "loss": 0.8695, + "step": 345270 + }, + { + "epoch": 29.89, + "learning_rate": 1.8925712964835738e-07, + "loss": 0.975, + "step": 345280 + }, + { + "epoch": 29.89, + "learning_rate": 1.8781241873501114e-07, + "loss": 0.9116, + "step": 345290 + }, + { + "epoch": 29.89, + "learning_rate": 1.863677078216649e-07, + "loss": 0.9556, + "step": 345300 + }, + { + "epoch": 29.89, + "learning_rate": 1.8492299690831867e-07, + "loss": 0.8813, + "step": 345310 + }, + { + "epoch": 29.89, + "learning_rate": 1.834782859949724e-07, + "loss": 0.8688, + "step": 345320 + }, + { + "epoch": 29.89, + "learning_rate": 1.8203357508162618e-07, + "loss": 0.9068, + "step": 345330 + }, + { + "epoch": 29.89, + "learning_rate": 1.8058886416827992e-07, + "loss": 0.97, + "step": 345340 + }, + { + "epoch": 29.89, + "learning_rate": 1.7914415325493368e-07, + "loss": 0.9398, + "step": 345350 + }, + { + "epoch": 29.89, + "learning_rate": 1.7769944234158745e-07, + "loss": 0.9171, + "step": 345360 + }, + { + "epoch": 29.89, + "learning_rate": 1.762547314282412e-07, + "loss": 0.9514, + "step": 345370 + }, + { + "epoch": 29.9, + "learning_rate": 1.7481002051489498e-07, + "loss": 0.934, + "step": 345380 + }, + { + "epoch": 29.9, + "learning_rate": 1.7336530960154874e-07, + "loss": 0.9096, + "step": 345390 + }, + { + "epoch": 29.9, + "learning_rate": 1.719205986882025e-07, + "loss": 0.8885, + "step": 345400 + }, + { + "epoch": 29.9, + "learning_rate": 1.7047588777485625e-07, + "loss": 0.9252, + "step": 345410 + }, + { + "epoch": 29.9, + "learning_rate": 1.6903117686151001e-07, + "loss": 0.8997, + "step": 345420 + }, + { + "epoch": 29.9, + "learning_rate": 1.6758646594816378e-07, + "loss": 0.9425, + "step": 345430 + }, + { + "epoch": 29.9, + "learning_rate": 1.6614175503481755e-07, + "loss": 0.9067, + "step": 345440 + }, + { + "epoch": 29.9, + "learning_rate": 1.646970441214713e-07, + "loss": 0.89, + "step": 345450 + }, + { + "epoch": 29.9, + "learning_rate": 1.6325233320812508e-07, + "loss": 0.9471, + "step": 345460 + }, + { + "epoch": 29.9, + "learning_rate": 1.6180762229477884e-07, + "loss": 0.894, + "step": 345470 + }, + { + "epoch": 29.9, + "learning_rate": 1.6036291138143258e-07, + "loss": 0.919, + "step": 345480 + }, + { + "epoch": 29.9, + "learning_rate": 1.5891820046808635e-07, + "loss": 0.8152, + "step": 345490 + }, + { + "epoch": 29.91, + "learning_rate": 1.574734895547401e-07, + "loss": 0.9623, + "step": 345500 + }, + { + "epoch": 29.91, + "learning_rate": 1.5602877864139388e-07, + "loss": 0.8943, + "step": 345510 + }, + { + "epoch": 29.91, + "learning_rate": 1.5458406772804762e-07, + "loss": 0.8912, + "step": 345520 + }, + { + "epoch": 29.91, + "learning_rate": 1.5313935681470138e-07, + "loss": 0.8654, + "step": 345530 + }, + { + "epoch": 29.91, + "learning_rate": 1.5169464590135515e-07, + "loss": 0.8658, + "step": 345540 + }, + { + "epoch": 29.91, + "learning_rate": 1.5024993498800889e-07, + "loss": 0.8253, + "step": 345550 + }, + { + "epoch": 29.91, + "learning_rate": 1.4880522407466265e-07, + "loss": 0.935, + "step": 345560 + }, + { + "epoch": 29.91, + "learning_rate": 1.4736051316131642e-07, + "loss": 0.8716, + "step": 345570 + }, + { + "epoch": 29.91, + "learning_rate": 1.4591580224797018e-07, + "loss": 0.9515, + "step": 345580 + }, + { + "epoch": 29.91, + "learning_rate": 1.4447109133462395e-07, + "loss": 0.9037, + "step": 345590 + }, + { + "epoch": 29.91, + "learning_rate": 1.4302638042127771e-07, + "loss": 0.8596, + "step": 345600 + }, + { + "epoch": 29.91, + "learning_rate": 1.4158166950793148e-07, + "loss": 0.8835, + "step": 345610 + }, + { + "epoch": 29.92, + "learning_rate": 1.4013695859458522e-07, + "loss": 0.9144, + "step": 345620 + }, + { + "epoch": 29.92, + "learning_rate": 1.3869224768123898e-07, + "loss": 0.9181, + "step": 345630 + }, + { + "epoch": 29.92, + "learning_rate": 1.3724753676789275e-07, + "loss": 0.8698, + "step": 345640 + }, + { + "epoch": 29.92, + "learning_rate": 1.3580282585454652e-07, + "loss": 0.9325, + "step": 345650 + }, + { + "epoch": 29.92, + "learning_rate": 1.3435811494120028e-07, + "loss": 0.8425, + "step": 345660 + }, + { + "epoch": 29.92, + "learning_rate": 1.3291340402785405e-07, + "loss": 0.9776, + "step": 345670 + }, + { + "epoch": 29.92, + "learning_rate": 1.314686931145078e-07, + "loss": 0.9052, + "step": 345680 + }, + { + "epoch": 29.92, + "learning_rate": 1.3002398220116155e-07, + "loss": 0.9348, + "step": 345690 + }, + { + "epoch": 29.92, + "learning_rate": 1.2857927128781532e-07, + "loss": 0.9319, + "step": 345700 + }, + { + "epoch": 29.92, + "learning_rate": 1.2713456037446908e-07, + "loss": 0.8799, + "step": 345710 + }, + { + "epoch": 29.92, + "learning_rate": 1.2568984946112282e-07, + "loss": 0.8894, + "step": 345720 + }, + { + "epoch": 29.93, + "learning_rate": 1.242451385477766e-07, + "loss": 0.9433, + "step": 345730 + }, + { + "epoch": 29.93, + "learning_rate": 1.2280042763443035e-07, + "loss": 0.9141, + "step": 345740 + }, + { + "epoch": 29.93, + "learning_rate": 1.2135571672108412e-07, + "loss": 0.872, + "step": 345750 + }, + { + "epoch": 29.93, + "learning_rate": 1.1991100580773788e-07, + "loss": 0.9525, + "step": 345760 + }, + { + "epoch": 29.93, + "learning_rate": 1.1846629489439164e-07, + "loss": 0.917, + "step": 345770 + }, + { + "epoch": 29.93, + "learning_rate": 1.1702158398104539e-07, + "loss": 0.8957, + "step": 345780 + }, + { + "epoch": 29.93, + "learning_rate": 1.1557687306769915e-07, + "loss": 0.9441, + "step": 345790 + }, + { + "epoch": 29.93, + "learning_rate": 1.1413216215435292e-07, + "loss": 0.9102, + "step": 345800 + }, + { + "epoch": 29.93, + "learning_rate": 1.1268745124100668e-07, + "loss": 0.8652, + "step": 345810 + }, + { + "epoch": 29.93, + "learning_rate": 1.1124274032766045e-07, + "loss": 0.8823, + "step": 345820 + }, + { + "epoch": 29.93, + "learning_rate": 1.0979802941431422e-07, + "loss": 0.9037, + "step": 345830 + }, + { + "epoch": 29.93, + "learning_rate": 1.0835331850096795e-07, + "loss": 0.915, + "step": 345840 + }, + { + "epoch": 29.94, + "learning_rate": 1.0690860758762172e-07, + "loss": 0.865, + "step": 345850 + }, + { + "epoch": 29.94, + "learning_rate": 1.0546389667427549e-07, + "loss": 0.9335, + "step": 345860 + }, + { + "epoch": 29.94, + "learning_rate": 1.0401918576092924e-07, + "loss": 0.9229, + "step": 345870 + }, + { + "epoch": 29.94, + "learning_rate": 1.02574474847583e-07, + "loss": 0.8812, + "step": 345880 + }, + { + "epoch": 29.94, + "learning_rate": 1.0112976393423677e-07, + "loss": 0.8807, + "step": 345890 + }, + { + "epoch": 29.94, + "learning_rate": 9.968505302089053e-08, + "loss": 0.8654, + "step": 345900 + }, + { + "epoch": 29.94, + "learning_rate": 9.824034210754427e-08, + "loss": 0.847, + "step": 345910 + }, + { + "epoch": 29.94, + "learning_rate": 9.679563119419805e-08, + "loss": 0.8587, + "step": 345920 + }, + { + "epoch": 29.94, + "learning_rate": 9.53509202808518e-08, + "loss": 0.8654, + "step": 345930 + }, + { + "epoch": 29.94, + "learning_rate": 9.390620936750557e-08, + "loss": 0.8665, + "step": 345940 + }, + { + "epoch": 29.94, + "learning_rate": 9.246149845415934e-08, + "loss": 0.8997, + "step": 345950 + }, + { + "epoch": 29.95, + "learning_rate": 9.101678754081309e-08, + "loss": 0.9121, + "step": 345960 + }, + { + "epoch": 29.95, + "learning_rate": 8.957207662746684e-08, + "loss": 0.8704, + "step": 345970 + }, + { + "epoch": 29.95, + "learning_rate": 8.81273657141206e-08, + "loss": 0.941, + "step": 345980 + }, + { + "epoch": 29.95, + "learning_rate": 8.668265480077437e-08, + "loss": 0.8502, + "step": 345990 + }, + { + "epoch": 29.95, + "learning_rate": 8.523794388742812e-08, + "loss": 0.9118, + "step": 346000 + }, + { + "epoch": 29.95, + "learning_rate": 8.379323297408189e-08, + "loss": 0.8431, + "step": 346010 + }, + { + "epoch": 29.95, + "learning_rate": 8.234852206073566e-08, + "loss": 0.9525, + "step": 346020 + }, + { + "epoch": 29.95, + "learning_rate": 8.090381114738942e-08, + "loss": 0.8594, + "step": 346030 + }, + { + "epoch": 29.95, + "learning_rate": 7.945910023404317e-08, + "loss": 0.9004, + "step": 346040 + }, + { + "epoch": 29.95, + "learning_rate": 7.801438932069694e-08, + "loss": 0.8729, + "step": 346050 + }, + { + "epoch": 29.95, + "learning_rate": 7.656967840735069e-08, + "loss": 0.8466, + "step": 346060 + }, + { + "epoch": 29.95, + "learning_rate": 7.512496749400444e-08, + "loss": 0.8623, + "step": 346070 + }, + { + "epoch": 29.96, + "learning_rate": 7.368025658065821e-08, + "loss": 0.9673, + "step": 346080 + }, + { + "epoch": 29.96, + "learning_rate": 7.223554566731197e-08, + "loss": 0.8957, + "step": 346090 + }, + { + "epoch": 29.96, + "learning_rate": 7.079083475396574e-08, + "loss": 0.8833, + "step": 346100 + }, + { + "epoch": 29.96, + "learning_rate": 6.934612384061949e-08, + "loss": 0.8689, + "step": 346110 + }, + { + "epoch": 29.96, + "learning_rate": 6.790141292727326e-08, + "loss": 0.7889, + "step": 346120 + }, + { + "epoch": 29.96, + "learning_rate": 6.645670201392702e-08, + "loss": 0.9303, + "step": 346130 + }, + { + "epoch": 29.96, + "learning_rate": 6.501199110058078e-08, + "loss": 0.8635, + "step": 346140 + }, + { + "epoch": 29.96, + "learning_rate": 6.356728018723454e-08, + "loss": 0.9019, + "step": 346150 + }, + { + "epoch": 29.96, + "learning_rate": 6.21225692738883e-08, + "loss": 0.9368, + "step": 346160 + }, + { + "epoch": 29.96, + "learning_rate": 6.067785836054206e-08, + "loss": 0.9053, + "step": 346170 + }, + { + "epoch": 29.96, + "learning_rate": 5.923314744719582e-08, + "loss": 0.9018, + "step": 346180 + }, + { + "epoch": 29.97, + "learning_rate": 5.778843653384958e-08, + "loss": 0.9057, + "step": 346190 + }, + { + "epoch": 29.97, + "learning_rate": 5.634372562050334e-08, + "loss": 0.9002, + "step": 346200 + }, + { + "epoch": 29.97, + "learning_rate": 5.489901470715711e-08, + "loss": 0.8953, + "step": 346210 + }, + { + "epoch": 29.97, + "learning_rate": 5.345430379381086e-08, + "loss": 0.91, + "step": 346220 + }, + { + "epoch": 29.97, + "learning_rate": 5.200959288046462e-08, + "loss": 0.88, + "step": 346230 + }, + { + "epoch": 29.97, + "learning_rate": 5.0564881967118385e-08, + "loss": 0.8986, + "step": 346240 + }, + { + "epoch": 29.97, + "learning_rate": 4.912017105377214e-08, + "loss": 0.9105, + "step": 346250 + }, + { + "epoch": 29.97, + "learning_rate": 4.76754601404259e-08, + "loss": 0.8518, + "step": 346260 + }, + { + "epoch": 29.97, + "learning_rate": 4.623074922707967e-08, + "loss": 0.8949, + "step": 346270 + }, + { + "epoch": 29.97, + "learning_rate": 4.478603831373342e-08, + "loss": 0.9033, + "step": 346280 + }, + { + "epoch": 29.97, + "learning_rate": 4.3341327400387186e-08, + "loss": 0.8734, + "step": 346290 + }, + { + "epoch": 29.97, + "learning_rate": 4.1896616487040945e-08, + "loss": 0.8905, + "step": 346300 + }, + { + "epoch": 29.98, + "learning_rate": 4.045190557369471e-08, + "loss": 0.9354, + "step": 346310 + }, + { + "epoch": 29.98, + "learning_rate": 3.900719466034847e-08, + "loss": 0.8351, + "step": 346320 + }, + { + "epoch": 29.98, + "learning_rate": 3.756248374700222e-08, + "loss": 0.8978, + "step": 346330 + }, + { + "epoch": 29.98, + "learning_rate": 3.611777283365599e-08, + "loss": 0.9599, + "step": 346340 + }, + { + "epoch": 29.98, + "learning_rate": 3.4673061920309746e-08, + "loss": 0.9148, + "step": 346350 + }, + { + "epoch": 29.98, + "learning_rate": 3.322835100696351e-08, + "loss": 0.9302, + "step": 346360 + }, + { + "epoch": 29.98, + "learning_rate": 3.178364009361727e-08, + "loss": 0.9322, + "step": 346370 + }, + { + "epoch": 29.98, + "learning_rate": 3.033892918027103e-08, + "loss": 0.9037, + "step": 346380 + }, + { + "epoch": 29.98, + "learning_rate": 2.889421826692479e-08, + "loss": 0.8831, + "step": 346390 + }, + { + "epoch": 29.98, + "learning_rate": 2.7449507353578554e-08, + "loss": 0.9448, + "step": 346400 + }, + { + "epoch": 29.98, + "learning_rate": 2.600479644023231e-08, + "loss": 0.9178, + "step": 346410 + }, + { + "epoch": 29.99, + "learning_rate": 2.456008552688607e-08, + "loss": 0.897, + "step": 346420 + }, + { + "epoch": 29.99, + "learning_rate": 2.3115374613539834e-08, + "loss": 0.9322, + "step": 346430 + }, + { + "epoch": 29.99, + "learning_rate": 2.1670663700193593e-08, + "loss": 0.8448, + "step": 346440 + }, + { + "epoch": 29.99, + "learning_rate": 2.0225952786847355e-08, + "loss": 0.9267, + "step": 346450 + }, + { + "epoch": 29.99, + "learning_rate": 1.878124187350111e-08, + "loss": 0.8116, + "step": 346460 + }, + { + "epoch": 29.99, + "learning_rate": 1.7336530960154873e-08, + "loss": 0.9208, + "step": 346470 + }, + { + "epoch": 29.99, + "learning_rate": 1.5891820046808635e-08, + "loss": 0.8489, + "step": 346480 + }, + { + "epoch": 29.99, + "learning_rate": 1.4447109133462394e-08, + "loss": 0.8504, + "step": 346490 + }, + { + "epoch": 29.99, + "learning_rate": 1.3002398220116155e-08, + "loss": 0.8359, + "step": 346500 + }, + { + "epoch": 29.99, + "learning_rate": 1.1557687306769917e-08, + "loss": 0.9308, + "step": 346510 + }, + { + "epoch": 29.99, + "learning_rate": 1.0112976393423678e-08, + "loss": 0.8765, + "step": 346520 + }, + { + "epoch": 29.99, + "learning_rate": 8.668265480077437e-09, + "loss": 0.8803, + "step": 346530 + }, + { + "epoch": 30.0, + "learning_rate": 7.223554566731197e-09, + "loss": 0.8781, + "step": 346540 + }, + { + "epoch": 30.0, + "learning_rate": 5.7788436533849585e-09, + "loss": 0.9003, + "step": 346550 + }, + { + "epoch": 30.0, + "learning_rate": 4.334132740038718e-09, + "loss": 0.8903, + "step": 346560 + }, + { + "epoch": 30.0, + "learning_rate": 2.8894218266924793e-09, + "loss": 0.8169, + "step": 346570 + }, + { + "epoch": 30.0, + "learning_rate": 1.4447109133462396e-09, + "loss": 0.8439, + "step": 346580 + }, + { + "epoch": 30.0, + "learning_rate": 0.0, + "loss": 0.8775, + "step": 346590 + }, + { + "epoch": 30.0, + "eval_cer": 0.9930033946814899, + "eval_em": 0, + "eval_f1": 0.1229099519237232, + "eval_loss": 0.9308269619941711, + "eval_runtime": 950.7232, + "eval_samples_per_second": 5.401, + "eval_steps_per_second": 0.675, + "eval_wer": 0.9658644888997574, + "step": 346590 + } + ], + "logging_steps": 10, + "max_steps": 346590, + "num_train_epochs": 30, + "save_steps": 500, + "total_flos": 2.6532919568382566e+17, + "trial_name": null, + "trial_params": null +}