Hanhpt23/whisper-small-germanmed-free_E0-8D0-8

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: 0.6821
• Wer: 26.4630
• Cer: 15.4774

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
0.5026	1.0	194	0.5290	29.2811	19.4792
0.2354	2.0	388	0.5282	34.3515	22.4451
0.1144	3.0	582	0.5396	32.5825	21.4992
0.073	4.0	776	0.5676	32.2226	22.5456
0.0465	5.0	970	0.6049	26.3499	16.4094
0.0375	6.0	1164	0.6197	33.1791	21.1216
0.0213	7.0	1358	0.6250	30.1759	20.1462
0.0229	8.0	1552	0.6453	31.4718	19.4914
0.0118	9.0	1746	0.6510	23.1924	14.5627
0.0138	10.0	1940	0.6604	27.9235	17.4974
0.0081	11.0	2134	0.6546	26.3705	16.3176
0.0029	12.0	2328	0.6527	25.3625	15.1725
0.0028	13.0	2522	0.6712	22.6473	14.5384
0.0003	14.0	2716	0.6743	30.7004	18.0015
0.0002	15.0	2910	0.6752	27.2035	16.0248
0.0001	16.0	3104	0.6787	24.8277	15.1292
0.0001	17.0	3298	0.6803	26.6893	15.6541
0.0001	18.0	3492	0.6813	26.5864	15.6021
0.0001	19.0	3686	0.6819	26.4836	15.4964
0.0001	20.0	3880	0.6821	26.4630	15.4774

Framework versions

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