Hanhpt23/whisper-small-engmed-free_E0-8

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: 0.0002
• Wer: 5.7320
• Cer: 4.8831

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	Cer
0.9946	1.0	386	0.4352	45.8326	35.2282
0.5738	2.0	772	0.2223	23.9456	21.5899
0.3113	3.0	1158	0.1615	22.5744	18.8772
0.1727	4.0	1544	0.1109	23.1553	19.4900
0.1512	5.0	1930	0.0792	16.6306	13.3710
0.1032	6.0	2316	0.0608	24.0367	22.4830
0.0948	7.0	2702	0.0428	24.7124	20.5852
0.0888	8.0	3088	0.0302	17.0837	13.1765
0.0364	9.0	3474	0.0295	13.1180	9.8240
0.039	10.0	3860	0.0168	12.1479	9.6012
0.0174	11.0	4246	0.0137	8.3940	6.9216
0.0183	12.0	4632	0.0099	9.6575	8.4382
0.0107	13.0	5018	0.0080	8.4862	6.9185
0.0074	14.0	5404	0.0038	8.0	6.0914
0.0026	15.0	5790	0.0020	6.6856	5.3293
0.0049	16.0	6176	0.0010	6.4868	5.2683
0.0034	17.0	6562	0.0005	6.1118	4.9500
0.0006	18.0	6948	0.0004	6.1402	5.1912
0.0006	19.0	7334	0.0003	5.6847	4.8742
0.0001	20.0	7720	0.0002	5.7320	4.8831

Framework versions

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