thomas0104/whisper-medium-nan-tw-only-char

Whisper medium nan-tw only char

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

• Loss: 0.9944
• Wer: 45.2824
• Cer: 45.3667

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: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
0.5832	1.04	1000	1.0634	56.3053	56.4745
0.1467	2.08	2000	1.0407	50.9618	51.0112
0.016	3.13	3000	1.0226	46.4427	46.5137
0.0001	5.01	4000	0.9974	45.4656	45.6082
0.0001	6.05	5000	0.9944	45.2824	45.3667

Framework versions

• Transformers 4.27.0.dev0
• Pytorch 1.13.1+cu117
• Datasets 2.8.0
• Tokenizers 0.13.2