v5_Robust_Model

This model is a fine-tuned version of openai/whisper-large on the 29 Lines dataset. It achieves the following results on the evaluation set:

• Loss: 0.0001
• Wer: 31.2217

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 795
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.8614	1.25	50	0.2854	17.6471
0.1863	2.5	100	0.0002	25.7919
0.0009	3.75	150	0.0001	17.1946
0.0001	5.0	200	0.0001	19.0045
0.0001	6.25	250	0.0001	23.9819
0.0001	7.5	300	0.0001	25.7919
0.0001	8.75	350	0.0001	22.6244
0.0001	10.0	400	0.0001	24.8869
0.0001	11.25	450	0.0001	23.5294
0.0001	12.5	500	0.0001	27.1493
0.0001	13.75	550	0.0001	26.6968
0.0001	15.0	600	0.0001	28.5068
0.0001	16.25	650	0.0001	29.4118
0.0001	17.5	700	0.0001	29.8643
0.0001	18.75	750	0.0001	31.2217

Framework versions

• Transformers 4.48.0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0