Hanhpt23/whisper-base-chinesemed-full

openai/whisper-base

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

• Loss: 1.2560
• Wer: 85.6582

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
0.7187	1.0	161	0.9482	110.4126
0.4241	2.0	322	0.9742	94.1061
0.2284	3.0	483	1.0348	100.9823
0.1019	4.0	644	1.0740	89.9804
0.0663	5.0	805	1.1215	87.4263
0.0473	6.0	966	1.1625	88.4086
0.0354	7.0	1127	1.1645	89.5874
0.0288	8.0	1288	1.1893	105.6974
0.0194	9.0	1449	1.1955	88.0157
0.0144	10.0	1610	1.1969	123.5756
0.011	11.0	1771	1.2305	90.3733
0.005	12.0	1932	1.2550	88.8016
0.0058	13.0	2093	1.2345	87.8193
0.0018	14.0	2254	1.2281	86.8369
0.0006	15.0	2415	1.2489	85.8546
0.0012	16.0	2576	1.2419	86.8369
0.0005	17.0	2737	1.2564	85.6582
0.0004	18.0	2898	1.2542	85.6582
0.0004	19.0	3059	1.2548	85.8546
0.0004	20.0	3220	1.2560	85.6582

Framework versions

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