Hanhpt23/whisper-small-chinesemed-free_ED3-11

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: 0.9508
• Wer: 170.3340
• Cer: 95.3654

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
0.5247	1.0	161	0.7107	187.2299	105.6595
0.3389	2.0	322	0.6834	225.9332	70.5882
0.2239	3.0	483	0.6913	213.7525	69.2736
0.1151	4.0	644	0.7414	246.9548	117.2460
0.0777	5.0	805	0.8040	221.2181	110.5169
0.037	6.0	966	0.8081	236.9352	87.0098
0.0228	7.0	1127	0.8673	132.8094	53.7433
0.0155	8.0	1288	0.9148	135.7564	42.8253
0.0086	9.0	1449	0.8912	183.1041	71.5463
0.0053	10.0	1610	0.9049	169.1552	102.2950
0.0034	11.0	1771	0.8989	203.7328	99.2424
0.0027	12.0	1932	0.9066	166.4047	92.2460
0.0016	13.0	2093	0.9231	250.6876	123.8191
0.0027	14.0	2254	0.9342	207.4656	107.6203
0.0011	15.0	2415	0.9385	208.2515	103.7433
0.0011	16.0	2576	0.9434	212.5737	104.7014
0.001	17.0	2737	0.9462	211.7878	104.5455
0.0009	18.0	2898	0.9487	169.9411	95.2094
0.0008	19.0	3059	0.9502	169.9411	95.2763
0.0008	20.0	3220	0.9508	170.3340	95.3654

Framework versions

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