Hanhpt23/whisper-small-chinesemed-free_ED0-8

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: 0.9626
• Wer: 102.9470
• Cer: 27.5178

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
0.5624	1.0	161	0.7102	103.7328	36.8093
0.2753	2.0	322	0.7326	117.6817	44.2291
0.127	3.0	483	0.7775	86.0511	31.1720
0.0519	4.0	644	0.8256	103.5363	31.2834
0.0423	5.0	805	0.8966	119.4499	36.9207
0.0251	6.0	966	0.8908	105.8939	29.3449
0.0137	7.0	1127	0.9214	84.6758	26.2478
0.0169	8.0	1288	0.9114	87.0334	24.3761
0.0107	9.0	1449	0.9319	104.9116	28.4759
0.0025	10.0	1610	0.9353	103.5363	27.3173
0.0007	11.0	1771	0.9370	105.6974	28.9216
0.0017	12.0	1932	0.9412	103.3399	27.8075
0.0006	13.0	2093	0.9438	102.3576	27.7852
0.0003	14.0	2254	0.9575	103.5363	29.3449
0.0003	15.0	2415	0.9568	102.3576	27.8743
0.0002	16.0	2576	0.9591	103.1434	27.6738
0.0002	17.0	2737	0.9601	102.9470	27.5401
0.0002	18.0	2898	0.9613	102.7505	27.5178
0.0002	19.0	3059	0.9622	102.9470	27.5401
0.0002	20.0	3220	0.9626	102.9470	27.5178

Framework versions

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