whisper-tiny-en-US

This model is a fine-tuned version of openai/whisper-tiny on the PolyAI/minds14 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6166
• Wer Ortho: 0.3504
• Wer: 0.3536

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: 50
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.7658	1.7857	50	0.5871	0.3948	0.3932
0.2602	3.5714	100	0.4866	0.3504	0.3501
0.0796	5.3571	150	0.5121	0.3424	0.3453
0.0316	7.1429	200	0.5443	0.3374	0.3418
0.0116	8.9286	250	0.5672	0.3202	0.3253
0.0034	10.7143	300	0.5966	0.3529	0.3566
0.0026	12.5	350	0.6046	0.3541	0.3583
0.002	14.2857	400	0.6098	0.3498	0.3536
0.002	16.0714	450	0.6146	0.3510	0.3542
0.002	17.8571	500	0.6166	0.3504	0.3536

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.0
• Tokenizers 0.19.1