Hanhpt23/whisper-medium-GermanMed-full

openai/whisper-medium

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

• Loss: 0.6710
• Wer: 19.0888

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
0.5095	1.0	194	0.5514	29.2194
0.2795	2.0	388	0.5479	27.2138
0.177	3.0	582	0.5744	24.6323
0.1321	4.0	776	0.6230	26.0311
0.0829	5.0	970	0.6206	22.8222
0.0865	6.0	1164	0.6116	22.9456
0.0681	7.0	1358	0.6316	23.9638
0.0536	8.0	1552	0.6182	22.3696
0.0289	9.0	1746	0.6302	21.3309
0.0228	10.0	1940	0.6738	19.5002
0.0178	11.0	2134	0.6600	19.7367
0.011	12.0	2328	0.6765	20.1172
0.0073	13.0	2522	0.6586	21.3823
0.0042	14.0	2716	0.6455	18.9036
0.004	15.0	2910	0.6595	19.0785
0.0003	16.0	3104	0.6635	18.9036
0.0019	17.0	3298	0.6694	19.1093
0.0001	18.0	3492	0.6747	19.0270
0.0001	19.0	3686	0.6708	19.0990
0.0001	20.0	3880	0.6710	19.0888

Framework versions

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