Hanhpt23/whisper-medium-frenchmed-v1

openai/whisper-medium

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

• Loss: 1.5876
• Wer: 35.7771

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
1.0291	1.0	215	1.0037	44.2815
0.6067	2.0	430	1.1008	35.7038
0.3327	3.0	645	1.1870	37.0235
0.1727	4.0	860	1.2912	37.5367
0.1213	5.0	1075	1.4015	36.7302
0.1029	6.0	1290	1.4085	42.7419
0.054	7.0	1505	1.5070	37.2434
0.0438	8.0	1720	1.5635	37.2434
0.0465	9.0	1935	1.5192	51.5396
0.0329	10.0	2150	1.5806	39.0029
0.0211	11.0	2365	1.5379	34.8974
0.0165	12.0	2580	1.5379	40.1026
0.0157	13.0	2795	1.5596	37.1701
0.0115	14.0	3010	1.5608	36.5836
0.0048	15.0	3225	1.5498	35.9238
0.0045	16.0	3440	1.5764	36.5836
0.0019	17.0	3655	1.5668	35.7771
0.0004	18.0	3870	1.5774	35.8504
0.0004	19.0	4085	1.5845	35.8504
0.0003	20.0	4300	1.5876	35.7771

Framework versions

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