Hanhpt23/whisper-base-chinesemed-v1

openai/whisper-base

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

• Loss: 5.0851
• Wer: 123.8122

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
2.9911	1.0	2222	2.9811	126.1019
2.4471	2.0	4444	2.9707	127.8764
1.8547	3.0	6666	3.2495	108.1282
1.2595	4.0	8888	3.5609	127.3039
0.9103	5.0	11110	3.9172	114.0813
0.593	6.0	13332	4.2574	108.8151
0.4738	7.0	15554	4.4006	108.0137
0.3788	8.0	17776	4.5577	136.2335
0.3916	9.0	19998	4.6187	128.9067
0.3148	10.0	22220	4.7217	121.5799
0.3413	11.0	24442	4.8141	122.3812
0.2903	12.0	26664	4.8305	117.5157
0.3044	13.0	28886	4.8859	129.0212
0.2648	14.0	31108	4.9314	111.5054
0.3343	15.0	33330	4.9714	111.4482
0.2693	16.0	35552	5.0438	109.9599
0.2677	17.0	37774	5.0470	108.0710
0.2834	18.0	39996	5.0293	120.2633
0.2198	19.0	42218	5.0545	123.6978
0.2242	20.0	44440	5.0851	123.8122

Framework versions

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