Hanhpt23/whisper-tiny-chinesemed-full

openai/whisper-tiny

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

• Loss: 1.4458
• Wer: 91.7485

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
0.9094	1.0	161	1.0980	107.4656
0.5868	2.0	322	1.1400	169.7446
0.3831	3.0	483	1.1572	103.7328
0.1839	4.0	644	1.2091	99.2141
0.1032	5.0	805	1.2858	99.6071
0.0669	6.0	966	1.3358	94.3026
0.0445	7.0	1127	1.3355	95.6778
0.0279	8.0	1288	1.3960	93.1238
0.0218	9.0	1449	1.3946	95.4813
0.0163	10.0	1610	1.4312	91.7485
0.007	11.0	1771	1.4203	94.6955
0.0049	12.0	1932	1.4495	92.1415
0.0045	13.0	2093	1.4258	92.7308
0.0029	14.0	2254	1.4203	92.7308
0.0036	15.0	2415	1.4382	91.5521
0.0023	16.0	2576	1.4340	91.3556
0.0009	17.0	2737	1.4496	91.3556
0.0009	18.0	2898	1.4467	92.7308
0.0007	19.0	3059	1.4446	91.9450
0.0006	20.0	3220	1.4458	91.7485

Framework versions

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