whisper-small-chinese-tw-minnan-hanzi

This model is a fine-tuned version of openai/whisper-small on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:

• Loss: 1.2235
• Wer: 86.7140
• Cer: 62.5226

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: 1e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0802	3.6364	1000	1.1582	91.0751	80.0120
0.0014	7.2727	2000	1.1876	85.8012	61.7399
0.0002	10.9091	3000	1.1944	85.4970	61.7098
0.0002	14.5455	4000	1.2139	85.8012	61.8603
0.0001	18.1818	5000	1.2235	86.7140	62.5226

Framework versions

• Transformers 4.46.3
• Pytorch 2.4.0+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3