Hanhpt23/whisper-tiny-GermanMed-v1

openai/whisper-tiny

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

• Loss: 5.8433
• Wer: 107.4053

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
3.2151	1.0	765	3.6892	145.0057
2.4608	2.0	1530	3.7770	102.6980
1.8089	3.0	2295	4.0128	125.0287
1.3185	4.0	3060	4.2434	104.7072
0.8036	5.0	3825	4.5708	111.1940
0.464	6.0	4590	4.7913	128.3582
0.2689	7.0	5355	4.9963	117.8530
0.1899	8.0	6120	5.1768	108.7256
0.192	9.0	6885	5.2917	103.8462
0.128	10.0	7650	5.4184	105.2813
0.1435	11.0	8415	5.5027	115.8439
0.1201	12.0	9180	5.6452	110.9070
0.1328	13.0	9945	5.6889	122.4455
0.0979	14.0	10710	5.6940	113.2032
0.1165	15.0	11475	5.7213	108.5534
0.0978	16.0	12240	5.7604	105.7405
0.0927	17.0	13005	5.7672	105.9127
0.0829	18.0	13770	5.7777	106.1998
0.0974	19.0	14535	5.7957	110.3904
0.0998	20.0	15300	5.8433	107.4053

Framework versions

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