Hanhpt23/whisper-base-Encod-frenchmed

openai/whisper-base

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

• Loss: 1.6240
• Wer: 47.8006

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
1.2466	1.0	215	1.2507	115.8358
0.7568	2.0	430	1.2524	86.4370
0.4229	3.0	645	1.3556	72.1408
0.2402	4.0	860	1.4000	52.4927
0.1544	5.0	1075	1.4770	47.1408
0.1127	6.0	1290	1.5115	56.7449
0.0674	7.0	1505	1.5317	56.8182
0.0526	8.0	1720	1.5979	51.1730
0.0442	9.0	1935	1.5844	50.5865
0.038	10.0	2150	1.5793	55.4252
0.0193	11.0	2365	1.6582	50.3666
0.0138	12.0	2580	1.6064	56.8182
0.0144	13.0	2795	1.5930	48.3138
0.0125	14.0	3010	1.6400	53.4457
0.0061	15.0	3225	1.6139	48.9003
0.0024	16.0	3440	1.6124	46.1877
0.0017	17.0	3655	1.6168	50.2933
0.0015	18.0	3870	1.6247	45.5279
0.0004	19.0	4085	1.6212	45.8211
0.0005	20.0	4300	1.6240	47.8006

Framework versions

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