Hanhpt23/whisper-small-germanmed-free_E3-11

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: 0.6983
• Wer: 23.1307
• Cer: 14.7116

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
0.543	1.0	194	0.5713	37.3547	23.8778
0.2814	2.0	388	0.5356	27.0287	17.3085
0.1367	3.0	582	0.5955	35.5446	23.1433
0.0903	4.0	776	0.6163	31.9037	20.3333
0.0768	5.0	970	0.6262	27.4607	16.8373
0.0585	6.0	1164	0.6513	29.8262	18.5455
0.0507	7.0	1358	0.6889	28.8388	18.3237
0.0388	8.0	1552	0.6816	28.1497	17.8282
0.0277	9.0	1746	0.6981	23.3673	14.2699
0.0167	10.0	1940	0.6921	24.5192	15.6108
0.0136	11.0	2134	0.7047	24.4678	15.5224
0.0092	12.0	2328	0.7167	24.5089	15.4271
0.004	13.0	2522	0.7071	24.9511	15.7251
0.0009	14.0	2716	0.7150	23.0999	14.4743
0.0009	15.0	2910	0.6984	23.5318	14.8035
0.0002	16.0	3104	0.7018	23.6758	15.0980
0.0004	17.0	3298	0.6949	22.8222	14.2993
0.0001	18.0	3492	0.6972	23.2130	14.7255
0.0001	19.0	3686	0.6980	23.2027	14.7601
0.0003	20.0	3880	0.6983	23.1307	14.7116

Framework versions

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