Hanhpt23/whisper-small-engmed-free_E3-11

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: 5.9083
• Cer: 4.9996

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
1.0176	1.0	386	0.4133	39.0807	31.4939
0.5708	2.0	772	0.2067	31.8558	30.2466
0.3127	3.0	1158	0.1413	23.2384	20.9818
0.162	4.0	1544	0.0908	21.8493	16.6565
0.1345	5.0	1930	0.0681	16.1810	14.7028
0.0821	6.0	2316	0.0547	12.3978	10.4709
0.0879	7.0	2702	0.0408	12.6889	10.4477
0.0826	8.0	3088	0.0299	8.9417	7.5959
0.043	9.0	3474	0.0220	11.2629	10.2332
0.0258	10.0	3860	0.0200	13.2150	10.7967
0.0147	11.0	4246	0.0122	6.9222	5.7379
0.0118	12.0	4632	0.0080	7.7291	6.3420
0.0096	13.0	5018	0.0054	7.6853	6.5296
0.0046	14.0	5404	0.0038	5.7841	4.8865
0.0047	15.0	5790	0.0018	5.8716	5.2142
0.0046	16.0	6176	0.0008	5.6800	4.7276
0.0036	17.0	6562	0.0005	5.8018	4.8929
0.0005	18.0	6948	0.0003	6.1615	5.1130
0.0008	19.0	7334	0.0002	5.9757	5.0725
0.0001	20.0	7720	0.0002	5.9083	4.9996

Framework versions

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