Hugging Face's logo

qualcomm
/
Segment-Anything-Model

Image Segmentation
PyTorch
LiteRT
ONNX
foundation
android
Model card Files Community

Segment-Anything-Model: Optimized for Mobile Deployment

High-quality segmentation mask generation around any object in an image with simple input prompt

Transformer based encoder-decoder where prompts specify what to segment in an image thereby allowing segmentation without the need for additional training. The image encoder generates embeddings and the lightweight decoder operates on the embeddings for point and mask based image segmentation.

This model is an implementation of Segment-Anything-Model found here.

This repository provides scripts to run Segment-Anything-Model on Qualcomm® devices. More details on model performance across various devices, can be found here.

Model Details

  • Model Type: Semantic segmentation
  • Model Stats:
    • Model checkpoint: vit_l
    • Input resolution: 720p (720x1280)
    • Number of parameters (SAMDecoder): 5.11M
    • Model size (SAMDecoder): 19.6 MB
Model Device Chipset Target Runtime Inference Time (ms) Peak Memory Range (MB) Precision Primary Compute Unit Target Model
SAMDecoder Samsung Galaxy S23 Snapdragon® 8 Gen 2 TFLITE 7.368 ms 0 - 29 MB FP16 NPU Segment-Anything-Model.tflite
SAMDecoder Samsung Galaxy S23 Snapdragon® 8 Gen 2 QNN 6.139 ms 4 - 7 MB FP16 NPU Segment-Anything-Model.so
SAMDecoder Samsung Galaxy S23 Snapdragon® 8 Gen 2 ONNX 8.724 ms 3 - 59 MB FP16 NPU Segment-Anything-Model.onnx
SAMDecoder Samsung Galaxy S24 Snapdragon® 8 Gen 3 TFLITE 5.183 ms 0 - 47 MB FP16 NPU Segment-Anything-Model.tflite
SAMDecoder Samsung Galaxy S24 Snapdragon® 8 Gen 3 QNN 4.206 ms 4 - 23 MB FP16 NPU Segment-Anything-Model.so
SAMDecoder Samsung Galaxy S24 Snapdragon® 8 Gen 3 ONNX 5.79 ms 4 - 65 MB FP16 NPU Segment-Anything-Model.onnx
SAMDecoder Snapdragon 8 Elite QRD Snapdragon® 8 Elite TFLITE 5.19 ms 0 - 45 MB FP16 NPU Segment-Anything-Model.tflite
SAMDecoder Snapdragon 8 Elite QRD Snapdragon® 8 Elite QNN 4.385 ms 4 - 42 MB FP16 NPU Use Export Script
SAMDecoder Snapdragon 8 Elite QRD Snapdragon® 8 Elite ONNX 5.586 ms 6 - 59 MB FP16 NPU Segment-Anything-Model.onnx
SAMDecoder SA7255P ADP SA7255P TFLITE 53.085 ms 0 - 40 MB FP16 NPU Segment-Anything-Model.tflite
SAMDecoder SA7255P ADP SA7255P QNN 48.936 ms 1 - 8 MB FP16 NPU Use Export Script
SAMDecoder SA8255 (Proxy) SA8255P Proxy TFLITE 7.379 ms 0 - 32 MB FP16 NPU Segment-Anything-Model.tflite
SAMDecoder SA8255 (Proxy) SA8255P Proxy QNN 6.142 ms 4 - 7 MB FP16 NPU Use Export Script
SAMDecoder SA8295P ADP SA8295P TFLITE 9.633 ms 0 - 38 MB FP16 NPU Segment-Anything-Model.tflite
SAMDecoder SA8295P ADP SA8295P QNN 7.489 ms 0 - 14 MB FP16 NPU Use Export Script
SAMDecoder SA8650 (Proxy) SA8650P Proxy TFLITE 7.374 ms 0 - 28 MB FP16 NPU Segment-Anything-Model.tflite
SAMDecoder SA8650 (Proxy) SA8650P Proxy QNN 6.18 ms 4 - 6 MB FP16 NPU Use Export Script
SAMDecoder SA8775P ADP SA8775P TFLITE 10.424 ms 0 - 40 MB FP16 NPU Segment-Anything-Model.tflite
SAMDecoder SA8775P ADP SA8775P QNN 8.823 ms 1 - 11 MB FP16 NPU Use Export Script
SAMDecoder QCS8275 (Proxy) QCS8275 Proxy TFLITE 53.085 ms 0 - 40 MB FP16 NPU Segment-Anything-Model.tflite
SAMDecoder QCS8275 (Proxy) QCS8275 Proxy QNN 48.936 ms 1 - 8 MB FP16 NPU Use Export Script
SAMDecoder QCS8550 (Proxy) QCS8550 Proxy TFLITE 7.377 ms 0 - 29 MB FP16 NPU Segment-Anything-Model.tflite
SAMDecoder QCS8550 (Proxy) QCS8550 Proxy QNN 6.15 ms 4 - 7 MB FP16 NPU Use Export Script
SAMDecoder QCS9075 (Proxy) QCS9075 Proxy TFLITE 10.424 ms 0 - 40 MB FP16 NPU Segment-Anything-Model.tflite
SAMDecoder QCS9075 (Proxy) QCS9075 Proxy QNN 8.823 ms 1 - 11 MB FP16 NPU Use Export Script
SAMDecoder QCS8450 (Proxy) QCS8450 Proxy TFLITE 8.425 ms 0 - 43 MB FP16 NPU Segment-Anything-Model.tflite
SAMDecoder QCS8450 (Proxy) QCS8450 Proxy QNN 7.554 ms 4 - 40 MB FP16 NPU Use Export Script
SAMDecoder Snapdragon X Elite CRD Snapdragon® X Elite QNN 6.804 ms 4 - 4 MB FP16 NPU Use Export Script
SAMDecoder Snapdragon X Elite CRD Snapdragon® X Elite ONNX 8.733 ms 11 - 11 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart1 Samsung Galaxy S23 Snapdragon® 8 Gen 2 TFLITE 248.723 ms 12 - 110 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart1 Samsung Galaxy S23 Snapdragon® 8 Gen 2 QNN 217.855 ms 12 - 15 MB FP16 NPU Segment-Anything-Model.so
SAMEncoderPart1 Samsung Galaxy S23 Snapdragon® 8 Gen 2 ONNX 244.864 ms 10 - 214 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart1 Samsung Galaxy S24 Snapdragon® 8 Gen 3 TFLITE 171.86 ms 9 - 940 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart1 Samsung Galaxy S24 Snapdragon® 8 Gen 3 QNN 150.079 ms 12 - 31 MB FP16 NPU Segment-Anything-Model.so
SAMEncoderPart1 Samsung Galaxy S24 Snapdragon® 8 Gen 3 ONNX 172.144 ms 35 - 947 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart1 Snapdragon 8 Elite QRD Snapdragon® 8 Elite TFLITE 147.045 ms 11 - 926 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart1 Snapdragon 8 Elite QRD Snapdragon® 8 Elite QNN 176.233 ms 12 - 890 MB FP16 NPU Use Export Script
SAMEncoderPart1 Snapdragon 8 Elite QRD Snapdragon® 8 Elite ONNX 181.035 ms 36 - 908 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart1 SA7255P ADP SA7255P TFLITE 1332.42 ms 0 - 914 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart1 SA7255P ADP SA7255P QNN 1261.003 ms 12 - 19 MB FP16 NPU Use Export Script
SAMEncoderPart1 SA8255 (Proxy) SA8255P Proxy TFLITE 248.72 ms 12 - 120 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart1 SA8255 (Proxy) SA8255P Proxy QNN 218.365 ms 12 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart1 SA8295P ADP SA8295P TFLITE 293.443 ms 10 - 886 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart1 SA8295P ADP SA8295P QNN 251.686 ms 0 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart1 SA8650 (Proxy) SA8650P Proxy TFLITE 247.555 ms 12 - 114 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart1 SA8650 (Proxy) SA8650P Proxy QNN 215.12 ms 12 - 14 MB FP16 NPU Use Export Script
SAMEncoderPart1 SA8775P ADP SA8775P TFLITE 305.909 ms 12 - 926 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart1 SA8775P ADP SA8775P QNN 270.478 ms 1 - 11 MB FP16 NPU Use Export Script
SAMEncoderPart1 QCS8275 (Proxy) QCS8275 Proxy TFLITE 1332.42 ms 0 - 914 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart1 QCS8275 (Proxy) QCS8275 Proxy QNN 1261.003 ms 12 - 19 MB FP16 NPU Use Export Script
SAMEncoderPart1 QCS8550 (Proxy) QCS8550 Proxy TFLITE 249.603 ms 12 - 104 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart1 QCS8550 (Proxy) QCS8550 Proxy QNN 213.14 ms 12 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart1 QCS9075 (Proxy) QCS9075 Proxy TFLITE 305.909 ms 12 - 926 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart1 QCS9075 (Proxy) QCS9075 Proxy QNN 270.478 ms 1 - 11 MB FP16 NPU Use Export Script
SAMEncoderPart1 QCS8450 (Proxy) QCS8450 Proxy TFLITE 279.425 ms 12 - 1376 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart1 QCS8450 (Proxy) QCS8450 Proxy QNN 283.833 ms 12 - 974 MB FP16 NPU Use Export Script
SAMEncoderPart1 Snapdragon X Elite CRD Snapdragon® X Elite QNN 217.349 ms 12 - 12 MB FP16 NPU Use Export Script
SAMEncoderPart1 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 247.096 ms 46 - 46 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart2 Samsung Galaxy S23 Snapdragon® 8 Gen 2 TFLITE 680.259 ms 0 - 89 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart2 Samsung Galaxy S23 Snapdragon® 8 Gen 2 QNN 615.417 ms 12 - 15 MB FP16 NPU Segment-Anything-Model.so
SAMEncoderPart2 Samsung Galaxy S23 Snapdragon® 8 Gen 2 ONNX 782.889 ms 23 - 168 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart2 Samsung Galaxy S24 Snapdragon® 8 Gen 3 TFLITE 516.603 ms 11 - 793 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart2 Samsung Galaxy S24 Snapdragon® 8 Gen 3 QNN 461.482 ms 12 - 31 MB FP16 NPU Segment-Anything-Model.so
SAMEncoderPart2 Samsung Galaxy S24 Snapdragon® 8 Gen 3 ONNX 572.181 ms 22 - 773 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart2 Snapdragon 8 Elite QRD Snapdragon® 8 Elite TFLITE 440.054 ms 11 - 796 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart2 Snapdragon 8 Elite QRD Snapdragon® 8 Elite QNN 498.801 ms 12 - 761 MB FP16 NPU Use Export Script
SAMEncoderPart2 Snapdragon 8 Elite QRD Snapdragon® 8 Elite ONNX 486.671 ms 24 - 754 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart2 SA7255P ADP SA7255P TFLITE 2080.579 ms 0 - 785 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart2 SA7255P ADP SA7255P QNN 1976.601 ms 3 - 10 MB FP16 NPU Use Export Script
SAMEncoderPart2 SA8255 (Proxy) SA8255P Proxy TFLITE 666.319 ms 0 - 82 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart2 SA8255 (Proxy) SA8255P Proxy QNN 618.277 ms 12 - 14 MB FP16 NPU Use Export Script
SAMEncoderPart2 SA8295P ADP SA8295P TFLITE 763.543 ms 12 - 758 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart2 SA8295P ADP SA8295P QNN 681.034 ms 0 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart2 SA8650 (Proxy) SA8650P Proxy TFLITE 670.805 ms 12 - 95 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart2 SA8650 (Proxy) SA8650P Proxy QNN 618.227 ms 12 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart2 SA8775P ADP SA8775P TFLITE 767.518 ms 11 - 796 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart2 SA8775P ADP SA8775P QNN 693.98 ms 1 - 11 MB FP16 NPU Use Export Script
SAMEncoderPart2 QCS8275 (Proxy) QCS8275 Proxy TFLITE 2080.579 ms 0 - 785 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart2 QCS8275 (Proxy) QCS8275 Proxy QNN 1976.601 ms 3 - 10 MB FP16 NPU Use Export Script
SAMEncoderPart2 QCS8550 (Proxy) QCS8550 Proxy TFLITE 676.281 ms 0 - 92 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart2 QCS8550 (Proxy) QCS8550 Proxy QNN 611.067 ms 12 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart2 QCS9075 (Proxy) QCS9075 Proxy TFLITE 767.518 ms 11 - 796 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart2 QCS9075 (Proxy) QCS9075 Proxy QNN 693.98 ms 1 - 11 MB FP16 NPU Use Export Script
SAMEncoderPart2 QCS8450 (Proxy) QCS8450 Proxy TFLITE 743.25 ms 12 - 771 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart2 QCS8450 (Proxy) QCS8450 Proxy QNN 740.992 ms 12 - 729 MB FP16 NPU Use Export Script
SAMEncoderPart2 Snapdragon X Elite CRD Snapdragon® X Elite QNN 587.357 ms 12 - 12 MB FP16 NPU Use Export Script
SAMEncoderPart2 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 730.12 ms 36 - 36 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart3 Samsung Galaxy S23 Snapdragon® 8 Gen 2 TFLITE 672.821 ms 12 - 93 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart3 Samsung Galaxy S23 Snapdragon® 8 Gen 2 QNN 617.682 ms 12 - 15 MB FP16 NPU Segment-Anything-Model.so
SAMEncoderPart3 Samsung Galaxy S23 Snapdragon® 8 Gen 2 ONNX 779.263 ms 24 - 163 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart3 Samsung Galaxy S24 Snapdragon® 8 Gen 3 TFLITE 510.644 ms 11 - 794 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart3 Samsung Galaxy S24 Snapdragon® 8 Gen 3 QNN 463.147 ms 12 - 31 MB FP16 NPU Segment-Anything-Model.so
SAMEncoderPart3 Samsung Galaxy S24 Snapdragon® 8 Gen 3 ONNX 573.142 ms 35 - 789 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart3 Snapdragon 8 Elite QRD Snapdragon® 8 Elite TFLITE 497.103 ms 11 - 797 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart3 Snapdragon 8 Elite QRD Snapdragon® 8 Elite QNN 499.164 ms 12 - 762 MB FP16 NPU Use Export Script
SAMEncoderPart3 Snapdragon 8 Elite QRD Snapdragon® 8 Elite ONNX 535.876 ms 12 - 743 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart3 SA7255P ADP SA7255P TFLITE 2080.525 ms 0 - 784 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart3 SA7255P ADP SA7255P QNN 1967.875 ms 4 - 12 MB FP16 NPU Use Export Script
SAMEncoderPart3 SA8255 (Proxy) SA8255P Proxy TFLITE 680.28 ms 0 - 85 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart3 SA8255 (Proxy) SA8255P Proxy QNN 614.492 ms 12 - 14 MB FP16 NPU Use Export Script
SAMEncoderPart3 SA8295P ADP SA8295P TFLITE 764.476 ms 11 - 756 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart3 SA8295P ADP SA8295P QNN 680.306 ms 0 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart3 SA8650 (Proxy) SA8650P Proxy TFLITE 670.295 ms 12 - 103 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart3 SA8650 (Proxy) SA8650P Proxy QNN 616.884 ms 12 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart3 SA8775P ADP SA8775P TFLITE 767.838 ms 12 - 797 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart3 SA8775P ADP SA8775P QNN 694.435 ms 12 - 22 MB FP16 NPU Use Export Script
SAMEncoderPart3 QCS8275 (Proxy) QCS8275 Proxy TFLITE 2080.525 ms 0 - 784 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart3 QCS8275 (Proxy) QCS8275 Proxy QNN 1967.875 ms 4 - 12 MB FP16 NPU Use Export Script
SAMEncoderPart3 QCS8550 (Proxy) QCS8550 Proxy TFLITE 674.671 ms 0 - 84 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart3 QCS8550 (Proxy) QCS8550 Proxy QNN 616.901 ms 12 - 17 MB FP16 NPU Use Export Script
SAMEncoderPart3 QCS9075 (Proxy) QCS9075 Proxy TFLITE 767.838 ms 12 - 797 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart3 QCS9075 (Proxy) QCS9075 Proxy QNN 694.435 ms 12 - 22 MB FP16 NPU Use Export Script
SAMEncoderPart3 QCS8450 (Proxy) QCS8450 Proxy TFLITE 760.033 ms 12 - 768 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart3 QCS8450 (Proxy) QCS8450 Proxy QNN 758.456 ms 12 - 732 MB FP16 NPU Use Export Script
SAMEncoderPart3 Snapdragon X Elite CRD Snapdragon® X Elite QNN 586.318 ms 12 - 12 MB FP16 NPU Use Export Script
SAMEncoderPart3 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 735.286 ms 35 - 35 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart4 Samsung Galaxy S23 Snapdragon® 8 Gen 2 TFLITE 669.945 ms 0 - 87 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart4 Samsung Galaxy S23 Snapdragon® 8 Gen 2 QNN 617.1 ms 12 - 15 MB FP16 NPU Segment-Anything-Model.so
SAMEncoderPart4 Samsung Galaxy S23 Snapdragon® 8 Gen 2 ONNX 769.838 ms 12 - 162 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart4 Samsung Galaxy S24 Snapdragon® 8 Gen 3 TFLITE 509.105 ms 11 - 793 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart4 Samsung Galaxy S24 Snapdragon® 8 Gen 3 QNN 458.614 ms 12 - 32 MB FP16 NPU Segment-Anything-Model.so
SAMEncoderPart4 Samsung Galaxy S24 Snapdragon® 8 Gen 3 ONNX 570.567 ms 24 - 775 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart4 Snapdragon 8 Elite QRD Snapdragon® 8 Elite TFLITE 437.498 ms 11 - 797 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart4 Snapdragon 8 Elite QRD Snapdragon® 8 Elite QNN 496.587 ms 12 - 761 MB FP16 NPU Use Export Script
SAMEncoderPart4 Snapdragon 8 Elite QRD Snapdragon® 8 Elite ONNX 536.53 ms 27 - 757 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart4 SA7255P ADP SA7255P TFLITE 2081.571 ms 0 - 784 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart4 SA7255P ADP SA7255P QNN 1976.132 ms 3 - 11 MB FP16 NPU Use Export Script
SAMEncoderPart4 SA8255 (Proxy) SA8255P Proxy TFLITE 669.017 ms 0 - 82 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart4 SA8255 (Proxy) SA8255P Proxy QNN 616.248 ms 12 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart4 SA8295P ADP SA8295P TFLITE 764.855 ms 12 - 757 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart4 SA8295P ADP SA8295P QNN 682.303 ms 0 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart4 SA8650 (Proxy) SA8650P Proxy TFLITE 669.419 ms 12 - 72 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart4 SA8650 (Proxy) SA8650P Proxy QNN 618.263 ms 12 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart4 SA8775P ADP SA8775P TFLITE 767.294 ms 12 - 796 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart4 SA8775P ADP SA8775P QNN 694.355 ms 1 - 11 MB FP16 NPU Use Export Script
SAMEncoderPart4 QCS8275 (Proxy) QCS8275 Proxy TFLITE 2081.571 ms 0 - 784 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart4 QCS8275 (Proxy) QCS8275 Proxy QNN 1976.132 ms 3 - 11 MB FP16 NPU Use Export Script
SAMEncoderPart4 QCS8550 (Proxy) QCS8550 Proxy TFLITE 672.575 ms 0 - 80 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart4 QCS8550 (Proxy) QCS8550 Proxy QNN 615.774 ms 12 - 14 MB FP16 NPU Use Export Script
SAMEncoderPart4 QCS9075 (Proxy) QCS9075 Proxy TFLITE 767.294 ms 12 - 796 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart4 QCS9075 (Proxy) QCS9075 Proxy QNN 694.355 ms 1 - 11 MB FP16 NPU Use Export Script
SAMEncoderPart4 QCS8450 (Proxy) QCS8450 Proxy TFLITE 754.379 ms 12 - 767 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart4 QCS8450 (Proxy) QCS8450 Proxy QNN 739.622 ms 12 - 731 MB FP16 NPU Use Export Script
SAMEncoderPart4 Snapdragon X Elite CRD Snapdragon® X Elite QNN 586.175 ms 12 - 12 MB FP16 NPU Use Export Script
SAMEncoderPart4 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 731.787 ms 35 - 35 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart5 Samsung Galaxy S23 Snapdragon® 8 Gen 2 TFLITE 673.83 ms 0 - 88 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart5 Samsung Galaxy S23 Snapdragon® 8 Gen 2 QNN 620.228 ms 14 - 16 MB FP16 NPU Segment-Anything-Model.so
SAMEncoderPart5 Samsung Galaxy S23 Snapdragon® 8 Gen 2 ONNX 774.021 ms 12 - 147 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart5 Samsung Galaxy S24 Snapdragon® 8 Gen 3 TFLITE 506.085 ms 10 - 796 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart5 Samsung Galaxy S24 Snapdragon® 8 Gen 3 QNN 458.994 ms 12 - 32 MB FP16 NPU Segment-Anything-Model.so
SAMEncoderPart5 Samsung Galaxy S24 Snapdragon® 8 Gen 3 ONNX 570.102 ms 24 - 776 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart5 Snapdragon 8 Elite QRD Snapdragon® 8 Elite TFLITE 498.671 ms 12 - 796 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart5 Snapdragon 8 Elite QRD Snapdragon® 8 Elite QNN 440.165 ms 12 - 760 MB FP16 NPU Use Export Script
SAMEncoderPart5 Snapdragon 8 Elite QRD Snapdragon® 8 Elite ONNX 538.191 ms 24 - 753 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart5 SA7255P ADP SA7255P TFLITE 2075.136 ms 0 - 783 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart5 SA7255P ADP SA7255P QNN 1975.907 ms 12 - 19 MB FP16 NPU Use Export Script
SAMEncoderPart5 SA8255 (Proxy) SA8255P Proxy TFLITE 671.333 ms 0 - 81 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart5 SA8255 (Proxy) SA8255P Proxy QNN 618.951 ms 12 - 14 MB FP16 NPU Use Export Script
SAMEncoderPart5 SA8295P ADP SA8295P TFLITE 762.394 ms 12 - 756 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart5 SA8295P ADP SA8295P QNN 680.206 ms 0 - 14 MB FP16 NPU Use Export Script
SAMEncoderPart5 SA8650 (Proxy) SA8650P Proxy TFLITE 672.05 ms 12 - 79 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart5 SA8650 (Proxy) SA8650P Proxy QNN 619.339 ms 12 - 14 MB FP16 NPU Use Export Script
SAMEncoderPart5 SA8775P ADP SA8775P TFLITE 767.752 ms 12 - 795 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart5 SA8775P ADP SA8775P QNN 694.595 ms 1 - 11 MB FP16 NPU Use Export Script
SAMEncoderPart5 QCS8275 (Proxy) QCS8275 Proxy TFLITE 2075.136 ms 0 - 783 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart5 QCS8275 (Proxy) QCS8275 Proxy QNN 1975.907 ms 12 - 19 MB FP16 NPU Use Export Script
SAMEncoderPart5 QCS8550 (Proxy) QCS8550 Proxy QNN 617.494 ms 12 - 14 MB FP16 NPU Use Export Script
SAMEncoderPart5 QCS9075 (Proxy) QCS9075 Proxy TFLITE 767.752 ms 12 - 795 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart5 QCS9075 (Proxy) QCS9075 Proxy QNN 694.595 ms 1 - 11 MB FP16 NPU Use Export Script
SAMEncoderPart5 QCS8450 (Proxy) QCS8450 Proxy TFLITE 742.93 ms 12 - 767 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart5 QCS8450 (Proxy) QCS8450 Proxy QNN 737.86 ms 12 - 736 MB FP16 NPU Use Export Script
SAMEncoderPart5 Snapdragon X Elite CRD Snapdragon® X Elite QNN 587.284 ms 12 - 12 MB FP16 NPU Use Export Script
SAMEncoderPart5 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 736.257 ms 36 - 36 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart6 Samsung Galaxy S23 Snapdragon® 8 Gen 2 TFLITE 672.843 ms 0 - 89 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart6 Samsung Galaxy S23 Snapdragon® 8 Gen 2 QNN 621.563 ms 16 - 18 MB FP16 NPU Segment-Anything-Model.so
SAMEncoderPart6 Samsung Galaxy S23 Snapdragon® 8 Gen 2 ONNX 771.737 ms 5 - 158 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart6 Samsung Galaxy S24 Snapdragon® 8 Gen 3 TFLITE 520.741 ms 11 - 796 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart6 Samsung Galaxy S24 Snapdragon® 8 Gen 3 QNN 459.106 ms 12 - 32 MB FP16 NPU Segment-Anything-Model.so
SAMEncoderPart6 Samsung Galaxy S24 Snapdragon® 8 Gen 3 ONNX 569.357 ms 27 - 781 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart6 Snapdragon 8 Elite QRD Snapdragon® 8 Elite TFLITE 493.833 ms 10 - 797 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart6 Snapdragon 8 Elite QRD Snapdragon® 8 Elite QNN 502.431 ms 12 - 760 MB FP16 NPU Use Export Script
SAMEncoderPart6 Snapdragon 8 Elite QRD Snapdragon® 8 Elite ONNX 488.507 ms 24 - 753 MB FP16 NPU Segment-Anything-Model.onnx
SAMEncoderPart6 SA7255P ADP SA7255P TFLITE 2080.562 ms 12 - 797 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart6 SA7255P ADP SA7255P QNN 1976.791 ms 12 - 19 MB FP16 NPU Use Export Script
SAMEncoderPart6 SA8255 (Proxy) SA8255P Proxy TFLITE 668.835 ms 12 - 102 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart6 SA8255 (Proxy) SA8255P Proxy QNN 615.347 ms 12 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart6 SA8295P ADP SA8295P TFLITE 764.252 ms 12 - 758 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart6 SA8295P ADP SA8295P QNN 680.522 ms 0 - 14 MB FP16 NPU Use Export Script
SAMEncoderPart6 SA8650 (Proxy) SA8650P Proxy TFLITE 672.927 ms 12 - 77 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart6 SA8650 (Proxy) SA8650P Proxy QNN 618.049 ms 12 - 15 MB FP16 NPU Use Export Script
SAMEncoderPart6 SA8775P ADP SA8775P TFLITE 769.477 ms 12 - 798 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart6 SA8775P ADP SA8775P QNN 694.634 ms 1 - 11 MB FP16 NPU Use Export Script
SAMEncoderPart6 QCS8275 (Proxy) QCS8275 Proxy TFLITE 2080.562 ms 12 - 797 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart6 QCS8275 (Proxy) QCS8275 Proxy QNN 1976.791 ms 12 - 19 MB FP16 NPU Use Export Script
SAMEncoderPart6 QCS8550 (Proxy) QCS8550 Proxy TFLITE 671.146 ms 12 - 98 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart6 QCS8550 (Proxy) QCS8550 Proxy QNN 616.722 ms 12 - 14 MB FP16 NPU Use Export Script
SAMEncoderPart6 QCS9075 (Proxy) QCS9075 Proxy TFLITE 769.477 ms 12 - 798 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart6 QCS9075 (Proxy) QCS9075 Proxy QNN 694.634 ms 1 - 11 MB FP16 NPU Use Export Script
SAMEncoderPart6 QCS8450 (Proxy) QCS8450 Proxy TFLITE 741.638 ms 12 - 767 MB FP16 NPU Segment-Anything-Model.tflite
SAMEncoderPart6 QCS8450 (Proxy) QCS8450 Proxy QNN 744.304 ms 12 - 731 MB FP16 NPU Use Export Script
SAMEncoderPart6 Snapdragon X Elite CRD Snapdragon® X Elite QNN 585.577 ms 12 - 12 MB FP16 NPU Use Export Script
SAMEncoderPart6 Snapdragon X Elite CRD Snapdragon® X Elite ONNX 740.654 ms 35 - 35 MB FP16 NPU Segment-Anything-Model.onnx

Installation

Install the package via pip:

pip install "qai-hub-models[sam]"

Configure Qualcomm® AI Hub to run this model on a cloud-hosted device

Sign-in to Qualcomm® AI Hub with your Qualcomm® ID. Once signed in navigate to Account -> Settings -> API Token.

With this API token, you can configure your client to run models on the cloud hosted devices.

qai-hub configure --api_token API_TOKEN

Navigate to docs for more information.

Demo off target

The package contains a simple end-to-end demo that downloads pre-trained weights and runs this model on a sample input.

python -m qai_hub_models.models.sam.demo

The above demo runs a reference implementation of pre-processing, model inference, and post processing.

NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).

%run -m qai_hub_models.models.sam.demo

Run model on a cloud-hosted device

In addition to the demo, you can also run the model on a cloud-hosted Qualcomm® device. This script does the following:

  • Performance check on-device on a cloud-hosted device
  • Downloads compiled assets that can be deployed on-device for Android.
  • Accuracy check between PyTorch and on-device outputs.
python -m qai_hub_models.models.sam.export

Profiling Results
------------------------------------------------------------
SAMDecoder
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 7.4                    
Estimated peak memory usage (MB): [0, 29]                
Total # Ops                     : 845                    
Compute Unit(s)                 : NPU (845 ops)          

------------------------------------------------------------
SAMEncoderPart1
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 248.7                  
Estimated peak memory usage (MB): [12, 110]              
Total # Ops                     : 584                    
Compute Unit(s)                 : NPU (584 ops)          

------------------------------------------------------------
SAMEncoderPart2
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 680.3                  
Estimated peak memory usage (MB): [0, 89]                
Total # Ops                     : 572                    
Compute Unit(s)                 : NPU (572 ops)          

------------------------------------------------------------
SAMEncoderPart3
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 672.8                  
Estimated peak memory usage (MB): [12, 93]               
Total # Ops                     : 572                    
Compute Unit(s)                 : NPU (572 ops)          

------------------------------------------------------------
SAMEncoderPart4
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 669.9                  
Estimated peak memory usage (MB): [0, 87]                
Total # Ops                     : 572                    
Compute Unit(s)                 : NPU (572 ops)          

------------------------------------------------------------
SAMEncoderPart5
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 673.8                  
Estimated peak memory usage (MB): [0, 88]                
Total # Ops                     : 572                    
Compute Unit(s)                 : NPU (572 ops)          

------------------------------------------------------------
SAMEncoderPart6
Device                          : Samsung Galaxy S23 (13)
Runtime                         : TFLITE                 
Estimated inference time (ms)   : 672.8                  
Estimated peak memory usage (MB): [0, 89]                
Total # Ops                     : 573                    
Compute Unit(s)                 : NPU (573 ops)

How does this work?

This export script leverages Qualcomm® AI Hub to optimize, validate, and deploy this model on-device. Lets go through each step below in detail:

Step 1: Compile model for on-device deployment

To compile a PyTorch model for on-device deployment, we first trace the model in memory using the jit.trace and then call the submit_compile_job API.

import torch

import qai_hub as hub
from qai_hub_models.models.sam import Model

# Load the model
model = Model.from_pretrained()
decoder_model = model.decoder
encoder_splits[0]_model = model.encoder_splits[0]
encoder_splits[1]_model = model.encoder_splits[1]
encoder_splits[2]_model = model.encoder_splits[2]
encoder_splits[3]_model = model.encoder_splits[3]
encoder_splits[4]_model = model.encoder_splits[4]
encoder_splits[5]_model = model.encoder_splits[5]

# Device
device = hub.Device("Samsung Galaxy S23")

# Trace model
decoder_input_shape = decoder_model.get_input_spec()
decoder_sample_inputs = decoder_model.sample_inputs()

traced_decoder_model = torch.jit.trace(decoder_model, [torch.tensor(data[0]) for _, data in decoder_sample_inputs.items()])

# Compile model on a specific device
decoder_compile_job = hub.submit_compile_job(
    model=traced_decoder_model ,
    device=device,
    input_specs=decoder_model.get_input_spec(),
)

# Get target model to run on-device
decoder_target_model = decoder_compile_job.get_target_model()
# Trace model
encoder_splits[0]_input_shape = encoder_splits[0]_model.get_input_spec()
encoder_splits[0]_sample_inputs = encoder_splits[0]_model.sample_inputs()

traced_encoder_splits[0]_model = torch.jit.trace(encoder_splits[0]_model, [torch.tensor(data[0]) for _, data in encoder_splits[0]_sample_inputs.items()])

# Compile model on a specific device
encoder_splits[0]_compile_job = hub.submit_compile_job(
    model=traced_encoder_splits[0]_model ,
    device=device,
    input_specs=encoder_splits[0]_model.get_input_spec(),
)

# Get target model to run on-device
encoder_splits[0]_target_model = encoder_splits[0]_compile_job.get_target_model()
# Trace model
encoder_splits[1]_input_shape = encoder_splits[1]_model.get_input_spec()
encoder_splits[1]_sample_inputs = encoder_splits[1]_model.sample_inputs()

traced_encoder_splits[1]_model = torch.jit.trace(encoder_splits[1]_model, [torch.tensor(data[0]) for _, data in encoder_splits[1]_sample_inputs.items()])

# Compile model on a specific device
encoder_splits[1]_compile_job = hub.submit_compile_job(
    model=traced_encoder_splits[1]_model ,
    device=device,
    input_specs=encoder_splits[1]_model.get_input_spec(),
)

# Get target model to run on-device
encoder_splits[1]_target_model = encoder_splits[1]_compile_job.get_target_model()
# Trace model
encoder_splits[2]_input_shape = encoder_splits[2]_model.get_input_spec()
encoder_splits[2]_sample_inputs = encoder_splits[2]_model.sample_inputs()

traced_encoder_splits[2]_model = torch.jit.trace(encoder_splits[2]_model, [torch.tensor(data[0]) for _, data in encoder_splits[2]_sample_inputs.items()])

# Compile model on a specific device
encoder_splits[2]_compile_job = hub.submit_compile_job(
    model=traced_encoder_splits[2]_model ,
    device=device,
    input_specs=encoder_splits[2]_model.get_input_spec(),
)

# Get target model to run on-device
encoder_splits[2]_target_model = encoder_splits[2]_compile_job.get_target_model()
# Trace model
encoder_splits[3]_input_shape = encoder_splits[3]_model.get_input_spec()
encoder_splits[3]_sample_inputs = encoder_splits[3]_model.sample_inputs()

traced_encoder_splits[3]_model = torch.jit.trace(encoder_splits[3]_model, [torch.tensor(data[0]) for _, data in encoder_splits[3]_sample_inputs.items()])

# Compile model on a specific device
encoder_splits[3]_compile_job = hub.submit_compile_job(
    model=traced_encoder_splits[3]_model ,
    device=device,
    input_specs=encoder_splits[3]_model.get_input_spec(),
)

# Get target model to run on-device
encoder_splits[3]_target_model = encoder_splits[3]_compile_job.get_target_model()
# Trace model
encoder_splits[4]_input_shape = encoder_splits[4]_model.get_input_spec()
encoder_splits[4]_sample_inputs = encoder_splits[4]_model.sample_inputs()

traced_encoder_splits[4]_model = torch.jit.trace(encoder_splits[4]_model, [torch.tensor(data[0]) for _, data in encoder_splits[4]_sample_inputs.items()])

# Compile model on a specific device
encoder_splits[4]_compile_job = hub.submit_compile_job(
    model=traced_encoder_splits[4]_model ,
    device=device,
    input_specs=encoder_splits[4]_model.get_input_spec(),
)

# Get target model to run on-device
encoder_splits[4]_target_model = encoder_splits[4]_compile_job.get_target_model()
# Trace model
encoder_splits[5]_input_shape = encoder_splits[5]_model.get_input_spec()
encoder_splits[5]_sample_inputs = encoder_splits[5]_model.sample_inputs()

traced_encoder_splits[5]_model = torch.jit.trace(encoder_splits[5]_model, [torch.tensor(data[0]) for _, data in encoder_splits[5]_sample_inputs.items()])

# Compile model on a specific device
encoder_splits[5]_compile_job = hub.submit_compile_job(
    model=traced_encoder_splits[5]_model ,
    device=device,
    input_specs=encoder_splits[5]_model.get_input_spec(),
)

# Get target model to run on-device
encoder_splits[5]_target_model = encoder_splits[5]_compile_job.get_target_model()

Step 2: Performance profiling on cloud-hosted device

After compiling models from step 1. Models can be profiled model on-device using the target_model. Note that this scripts runs the model on a device automatically provisioned in the cloud. Once the job is submitted, you can navigate to a provided job URL to view a variety of on-device performance metrics.

decoder_profile_job = hub.submit_profile_job(
    model=decoder_target_model,
    device=device,
)
encoder_splits[0]_profile_job = hub.submit_profile_job(
    model=encoder_splits[0]_target_model,
    device=device,
)
encoder_splits[1]_profile_job = hub.submit_profile_job(
    model=encoder_splits[1]_target_model,
    device=device,
)
encoder_splits[2]_profile_job = hub.submit_profile_job(
    model=encoder_splits[2]_target_model,
    device=device,
)
encoder_splits[3]_profile_job = hub.submit_profile_job(
    model=encoder_splits[3]_target_model,
    device=device,
)
encoder_splits[4]_profile_job = hub.submit_profile_job(
    model=encoder_splits[4]_target_model,
    device=device,
)
encoder_splits[5]_profile_job = hub.submit_profile_job(
    model=encoder_splits[5]_target_model,
    device=device,
)

Step 3: Verify on-device accuracy

To verify the accuracy of the model on-device, you can run on-device inference on sample input data on the same cloud hosted device.

decoder_input_data = decoder_model.sample_inputs()
decoder_inference_job = hub.submit_inference_job(
    model=decoder_target_model,
    device=device,
    inputs=decoder_input_data,
)
decoder_inference_job.download_output_data()
encoder_splits[0]_input_data = encoder_splits[0]_model.sample_inputs()
encoder_splits[0]_inference_job = hub.submit_inference_job(
    model=encoder_splits[0]_target_model,
    device=device,
    inputs=encoder_splits[0]_input_data,
)
encoder_splits[0]_inference_job.download_output_data()
encoder_splits[1]_input_data = encoder_splits[1]_model.sample_inputs()
encoder_splits[1]_inference_job = hub.submit_inference_job(
    model=encoder_splits[1]_target_model,
    device=device,
    inputs=encoder_splits[1]_input_data,
)
encoder_splits[1]_inference_job.download_output_data()
encoder_splits[2]_input_data = encoder_splits[2]_model.sample_inputs()
encoder_splits[2]_inference_job = hub.submit_inference_job(
    model=encoder_splits[2]_target_model,
    device=device,
    inputs=encoder_splits[2]_input_data,
)
encoder_splits[2]_inference_job.download_output_data()
encoder_splits[3]_input_data = encoder_splits[3]_model.sample_inputs()
encoder_splits[3]_inference_job = hub.submit_inference_job(
    model=encoder_splits[3]_target_model,
    device=device,
    inputs=encoder_splits[3]_input_data,
)
encoder_splits[3]_inference_job.download_output_data()
encoder_splits[4]_input_data = encoder_splits[4]_model.sample_inputs()
encoder_splits[4]_inference_job = hub.submit_inference_job(
    model=encoder_splits[4]_target_model,
    device=device,
    inputs=encoder_splits[4]_input_data,
)
encoder_splits[4]_inference_job.download_output_data()
encoder_splits[5]_input_data = encoder_splits[5]_model.sample_inputs()
encoder_splits[5]_inference_job = hub.submit_inference_job(
    model=encoder_splits[5]_target_model,
    device=device,
    inputs=encoder_splits[5]_input_data,
)
encoder_splits[5]_inference_job.download_output_data()

With the output of the model, you can compute like PSNR, relative errors or spot check the output with expected output.

Note: This on-device profiling and inference requires access to Qualcomm® AI Hub. Sign up for access.

Run demo on a cloud-hosted device

You can also run the demo on-device.

python -m qai_hub_models.models.sam.demo --on-device

NOTE: If you want running in a Jupyter Notebook or Google Colab like environment, please add the following to your cell (instead of the above).

%run -m qai_hub_models.models.sam.demo -- --on-device

Deploying compiled model to Android

The models can be deployed using multiple runtimes:

  • TensorFlow Lite (.tflite export): This tutorial provides a guide to deploy the .tflite model in an Android application.

  • QNN (.so export ): This sample app provides instructions on how to use the .so shared library in an Android application.

View on Qualcomm® AI Hub

Get more details on Segment-Anything-Model's performance across various devices here. Explore all available models on Qualcomm® AI Hub

License

  • The license for the original implementation of Segment-Anything-Model can be found here.
  • The license for the compiled assets for on-device deployment can be found here

References

Community

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
Image Segmentation
This model is not currently available via any of the supported Inference Providers.
The model cannot be deployed to the HF Inference API: The HF Inference API does not support image-segmentation models for pytorch library.