Hanhpt23/whisper-small-chinesemed-free_E3-11

openai/whisper-small

This model is a fine-tuned version of openai/whisper-small on the pphuc25/ChiMed dataset. It achieves the following results on the evaluation set:

• Loss: 1.1603
• Wer: 84.4794
• Cer: 25.0446

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
0.6517	1.0	161	0.8312	272.4951	128.1194
0.3421	2.0	322	0.7985	171.1198	63.7478
0.2206	3.0	483	0.9271	152.0629	59.8039
0.103	4.0	644	1.0397	85.2652	28.4091
0.0834	5.0	805	1.0335	92.7308	26.6266
0.0578	6.0	966	1.0893	86.2475	28.6319
0.0617	7.0	1127	1.1243	88.2122	28.0303
0.0407	8.0	1288	1.1577	84.6758	28.6988
0.032	9.0	1449	1.1847	83.3006	26.7157
0.0223	10.0	1610	1.1404	94.1061	32.5980
0.0144	11.0	1771	1.1927	88.0157	28.3645
0.0087	12.0	1932	1.1309	88.4086	26.4483
0.0126	13.0	2093	1.1613	84.0864	25.5570
0.0047	14.0	2254	1.1863	86.4440	25.7353
0.0061	15.0	2415	1.1674	83.6935	27.0276
0.0013	16.0	2576	1.1762	83.4971	25.2005
0.0007	17.0	2737	1.1697	84.8723	25.4679
0.0006	18.0	2898	1.1614	83.8900	25.2897
0.0002	19.0	3059	1.1597	84.4794	25.0446
0.0002	20.0	3220	1.1603	84.4794	25.0446

Framework versions

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