whisper-tiny-zh-tw

This model is a fine-tuned version of Wellyowo/whisper-tiny-zh-tw on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5683
• Wer Ortho: 60.0
• Wer: 59.2233

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.1844	0.69	500	0.5523	64.0	65.0485
0.1105	1.38	1000	0.5425	65.0	64.0777
0.0475	2.06	1500	0.5324	63.0	64.0777
0.0687	2.75	2000	0.5128	62.0	61.1650
0.0371	3.44	2500	0.5496	64.0	64.0777
0.0181	4.13	3000	0.5339	62.0	63.1068
0.0202	4.82	3500	0.5474	65.0	64.0777
0.011	5.51	4000	0.5683	60.0	59.2233

Framework versions

• Transformers 4.37.2
• Pytorch 2.0.1+cu117
• Datasets 2.16.1
• Tokenizers 0.15.1