Hanhpt23/whisper-small-Encod-frenchmed

openai/whisper-small

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

• Loss: 1.6120
• Wer: 44.5015

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
1.2076	1.0	215	1.2220	65.9091
0.7142	2.0	430	1.3108	57.4047
0.3943	3.0	645	1.3278	56.9648
0.227	4.0	860	1.4398	46.9208
0.1701	5.0	1075	1.4437	46.3343
0.1137	6.0	1290	1.5016	48.8270
0.0942	7.0	1505	1.6104	47.0674
0.0627	8.0	1720	1.5725	48.1672
0.0484	9.0	1935	1.6233	45.5279
0.0454	10.0	2150	1.6325	49.7801
0.0276	11.0	2365	1.6426	45.7478
0.0246	12.0	2580	1.6756	44.7214
0.0189	13.0	2795	1.6642	46.1144
0.0135	14.0	3010	1.6463	43.1818
0.0036	15.0	3225	1.6292	44.3548
0.0066	16.0	3440	1.6133	45.0147
0.0022	17.0	3655	1.6154	45.8211
0.0015	18.0	3870	1.6086	45.5279
0.0002	19.0	4085	1.6105	45.2346
0.0005	20.0	4300	1.6120	44.5015

Framework versions

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