whisper-tiny-openslrdev

This model is a fine-tuned version of openai/whisper-tiny.en on the tedlium dataset. It achieves the following results on the evaluation set:

• Loss: 2.0820
• Wer: 90.4154

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: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• training_steps: 300
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.06	20	3.7027	35.3291
3.9098	0.13	40	3.3264	35.0647
3.0852	0.19	60	2.9769	34.0871
2.2682	0.26	80	2.7802	31.6309
1.6662	0.32	100	2.5284	27.7728
1.6662	0.38	120	2.4481	24.3668
1.2505	0.45	140	2.4118	21.6532
1.0859	0.51	160	2.3687	20.9087
0.9491	0.58	180	2.1924	19.6493
0.8746	0.64	200	2.1752	22.1229
0.8746	0.7	220	2.2546	29.7245
0.8064	0.77	240	2.1611	39.6326
0.733	0.83	260	2.1281	55.7334
0.7135	0.89	280	2.0406	75.1705
0.6806	0.96	300	2.0820	90.4154

Framework versions

• Transformers 4.39.2
• Pytorch 2.2.2+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2