Hanhpt23/whisper-small-frenchmed-v1

openai/whisper-small

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

• Loss: 1.5011
• Wer: 37.2434

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
1.0211	1.0	215	0.9633	38.9296
0.5896	2.0	430	1.0086	52.6393
0.3027	3.0	645	1.1072	38.9296
0.1371	4.0	860	1.2011	38.7097
0.1004	5.0	1075	1.2837	41.9355
0.0638	6.0	1290	1.3137	41.4223
0.0513	7.0	1505	1.3113	38.6364
0.0374	8.0	1720	1.3284	39.0029
0.0279	9.0	1935	1.4303	39.3695
0.0176	10.0	2150	1.4479	37.9765
0.0144	11.0	2365	1.4417	39.0029
0.0173	12.0	2580	1.4390	38.9296
0.0036	13.0	2795	1.4696	40.0293
0.0053	14.0	3010	1.4461	38.8563
0.0039	15.0	3225	1.4651	37.6100
0.0026	16.0	3440	1.4707	35.9971
0.0018	17.0	3655	1.4972	37.6833
0.0003	18.0	3870	1.4943	37.5367
0.0002	19.0	4085	1.4995	37.2434
0.0002	20.0	4300	1.5011	37.2434

Framework versions

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