Hanhpt23/whisper-small-germanmed-free_ED3-11

openai/whisper-small

This model is a fine-tuned version of openai/whisper-small on the Hanhpt23/GermanMed-full dataset. It achieves the following results on the evaluation set:

• Loss: 0.5327
• Wer: 22.8736
• Cer: 15.0096

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
0.4351	1.0	194	0.4422	66.5535	48.8609
0.2391	2.0	388	0.4250	54.4071	37.4409
0.1426	3.0	582	0.4290	56.2172	35.4555
0.0877	4.0	776	0.4469	55.1682	35.9389
0.0456	5.0	970	0.4723	29.5691	19.8569
0.0374	6.0	1164	0.4664	34.9275	22.5023
0.015	7.0	1358	0.4901	24.7249	16.3800
0.0179	8.0	1552	0.4805	29.3325	19.8240
0.0112	9.0	1746	0.4858	25.3317	16.7420
0.0064	10.0	1940	0.4905	25.7534	16.3488
0.0037	11.0	2134	0.5068	26.1648	16.7871
0.0027	12.0	2328	0.5067	23.2747	15.2504
0.0013	13.0	2522	0.5167	25.7842	16.4181
0.0009	14.0	2716	0.5195	23.0690	15.1586
0.0008	15.0	2910	0.5206	23.1719	15.2106
0.0007	16.0	3104	0.5260	23.0382	14.9230
0.0007	17.0	3298	0.5288	23.1719	15.1118
0.0006	18.0	3492	0.5310	23.1821	15.1465
0.0006	19.0	3686	0.5323	22.8633	15.0079
0.0006	20.0	3880	0.5327	22.8736	15.0096

Framework versions

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