Hanhpt23/whisper-tiny-engmed-v1

openai/whisper-tiny

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

• Loss: 1.0605
• Wer: 23.1540

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
0.6316	1.0	323	0.7416	27.0570
0.3434	2.0	646	0.8113	45.2004
0.1758	3.0	969	0.9400	37.6055
0.0946	4.0	1292	0.9426	45.8861
0.0614	5.0	1615	1.0252	40.0844
0.0435	6.0	1938	1.0605	23.1540
0.0235	7.0	2261	1.1076	29.8523
0.02	8.0	2584	1.1745	47.3101
0.0197	9.0	2907	1.1967	34.4937
0.0094	10.0	3230	1.2394	28.4283
0.0065	11.0	3553	1.2382	53.3755
0.0024	12.0	3876	1.2605	37.8165
0.0036	13.0	4199	1.2512	29.9051
0.002	14.0	4522	1.2735	32.7004
0.001	15.0	4845	1.2824	25.4747
0.0005	16.0	5168	1.2930	24.8945
0.0003	17.0	5491	1.2919	27.5844
0.0004	18.0	5814	1.3033	25.1055
0.0002	19.0	6137	1.2996	27.6371
0.0002	20.0	6460	1.3026	27.9008

Framework versions

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