Hanhpt23/whisper-medium-chinesemed-v1

openai/whisper-medium

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

• Loss: 5.2602
• Wer: 128.5060

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
3.1376	1.0	2222	3.0423	114.8827
2.4455	2.0	4444	3.0159	122.3240
1.5794	3.0	6666	3.4702	109.4448
0.885	4.0	8888	4.0015	122.7247
0.6963	5.0	11110	4.3781	109.8454
0.4926	6.0	13332	4.5503	105.9531
0.4615	7.0	15554	4.6635	108.9296
0.3744	8.0	17776	4.8384	113.3371
0.395	9.0	19998	4.9563	113.7378
0.3312	10.0	22220	4.9508	109.5592
0.3324	11.0	24442	5.0560	120.2633
0.3027	12.0	26664	4.9751	117.0006
0.3127	13.0	28886	5.1136	124.7853
0.2792	14.0	31108	5.1324	111.6772
0.3383	15.0	33330	5.1389	110.5896
0.2695	16.0	35552	5.1925	111.6199
0.2773	17.0	37774	5.1852	111.9061
0.2829	18.0	39996	5.2192	121.5226
0.2195	19.0	42218	5.2098	126.5598
0.2254	20.0	44440	5.2602	128.5060

Framework versions

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