Hanhpt23/whisper-tiny-GermanMed-full

openai/whisper-tiny

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

• Loss: 0.9014
• Wer: 29.3839

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
0.747	1.0	194	0.7466	40.1728
0.4004	2.0	388	0.7261	37.1285
0.2012	3.0	582	0.7231	34.3721
0.1134	4.0	776	0.7465	34.0327
0.0606	5.0	970	0.7888	37.7044
0.0525	6.0	1164	0.8147	33.9813
0.0323	7.0	1358	0.8244	31.6569
0.027	8.0	1552	0.8383	31.8215
0.0149	9.0	1746	0.8643	32.1094
0.0119	10.0	1940	0.8747	31.7495
0.009	11.0	2134	0.8765	30.9781
0.0037	12.0	2328	0.8875	29.8879
0.0021	13.0	2522	0.8832	30.0936
0.0011	14.0	2716	0.8943	29.9496
0.0013	15.0	2910	0.8906	29.5485
0.0006	16.0	3104	0.8944	29.5999
0.0006	17.0	3298	0.8968	29.3942
0.0005	18.0	3492	0.8997	29.3839
0.0006	19.0	3686	0.9010	29.4251
0.0005	20.0	3880	0.9014	29.3839

Framework versions

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