Hanhpt23/whisper-small-engmed-v1

openai/whisper-small

This model is a fine-tuned version of openai/whisper-small on the pphuc25/EngMed dataset. It achieves the following results on the evaluation set:

• Loss: 1.1935
• Wer: 27.2679

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5346	1.0	323	0.6308	30.4852
0.2978	2.0	646	0.7524	25.0527
0.1478	3.0	969	0.7934	29.2194
0.1208	4.0	1292	0.8305	34.0190
0.0667	5.0	1615	0.8781	22.2046
0.0562	6.0	1938	1.0234	27.0042
0.0519	7.0	2261	0.9954	17.1941
0.0311	8.0	2584	1.0153	35.2321
0.0183	9.0	2907	1.1135	48.5759
0.015	10.0	3230	1.1428	35.6540
0.0099	11.0	3553	1.1355	39.3987
0.0054	12.0	3876	1.1408	27.3207
0.0037	13.0	4199	1.1651	28.6920
0.0024	14.0	4522	1.1648	40.2954
0.0004	15.0	4845	1.1480	44.0401
0.0004	16.0	5168	1.1564	23.1540
0.0003	17.0	5491	1.1754	28.2173
0.0007	18.0	5814	1.1884	26.0021
0.0001	19.0	6137	1.1925	27.9536
0.0001	20.0	6460	1.1935	27.2679

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0
• Datasets 2.19.1
• Tokenizers 0.19.1