Hanhpt23/whisper-small-silvarmed

openai/whisper-small

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

• Loss: 0.1782
• Wer: 6.0842

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
0.0747	1.0	2438	0.2071	8.9008
0.0385	2.0	4876	0.2154	9.5935
0.0434	3.0	7314	0.1894	4.9993
0.0175	4.0	9752	0.2119	5.9665
0.0226	5.0	12190	0.1965	4.6334
0.0089	6.0	14628	0.2144	5.1954
0.0195	7.0	17066	0.2112	4.9536
0.0053	8.0	19504	0.1983	5.7313
0.0053	9.0	21942	0.2062	6.8226
0.0034	10.0	24380	0.1960	6.1822
0.0107	11.0	26818	0.2028	5.8947
0.0001	12.0	29256	0.2012	5.9012
0.0015	13.0	31694	0.1876	4.6138
0.0	14.0	34132	0.1878	6.8096
0.0032	15.0	36570	0.1962	6.0515
0.0	16.0	39008	0.1846	5.9796
0.0001	17.0	41446	0.1835	4.5680
0.002	18.0	43884	0.1813	5.9273
0.0	19.0	46322	0.1778	6.1495
0.0	20.0	48760	0.1782	6.0842

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0
• Datasets 3.2.0
• Tokenizers 0.19.1