aipanjab/whisper-small-pa

Whisper-small

This model is a fine-tuned version of openai/whisper-small on various datasets. It achieves the following results on the evaluation set:

• Loss: 0.2283
• Wer: 27.8113

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: 16
• 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: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1168	1.2804	1000	0.1364	35.1815
0.073	2.5608	2000	0.1149	30.2039
0.0443	3.8412	3000	0.1193	28.8568
0.0128	5.1216	4000	0.1494	28.9790
0.0073	6.4020	5000	0.1684	28.7293
0.0025	7.6825	6000	0.1877	28.5655
0.0018	8.9629	7000	0.2025	28.3184
0.0005	10.2433	8000	0.2150	27.9179
0.0002	11.5237	9000	0.2231	27.8347
0.0002	12.8041	10000	0.2283	27.8113

Framework versions

• Transformers 4.46.3
• Pytorch 2.1.0+cu118
• Datasets 3.1.0
• Tokenizers 0.20.3