Hanhpt23/whisper-small-frenchmed-free_E0-8

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: 1.4691
• Wer: 40.9824
• Cer: 28.8290

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
1.054	1.0	215	1.0167	52.2727	33.9067
0.6007	2.0	430	1.0289	57.0381	35.2828
0.3026	3.0	645	1.1406	40.3226	27.7418
0.1617	4.0	860	1.2025	40.0293	25.8291
0.1133	5.0	1075	1.2893	42.0088	28.0583
0.0903	6.0	1290	1.2941	42.0088	29.6821
0.0821	7.0	1505	1.3255	41.5689	28.0033
0.0511	8.0	1720	1.4011	43.5484	28.2923
0.0391	9.0	1935	1.4531	39.3695	27.4804
0.0283	10.0	2150	1.4935	43.0352	29.9436
0.0181	11.0	2365	1.4941	43.0352	29.1042
0.0152	12.0	2580	1.4768	43.3284	31.3609
0.0191	13.0	2795	1.4495	43.9150	30.8380
0.0055	14.0	3010	1.4699	42.1554	29.2418
0.0032	15.0	3225	1.4493	39.8094	28.0859
0.0041	16.0	3440	1.4503	40.8358	28.7739
0.0018	17.0	3655	1.4598	41.6422	29.3931
0.0015	18.0	3870	1.4654	41.2023	28.7051
0.0002	19.0	4085	1.4676	40.8358	28.7051
0.0005	20.0	4300	1.4691	40.9824	28.8290

Framework versions

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