yuweiiizz/whisper-small-taiwanese

Whisper Small Taiwanese

This model is a fine-tuned version of openai/whisper-small on the Common Voice 16.1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.9011
• Cer: 50.3995

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3.6

Training results

Training Loss	Epoch	Step	Cer	Validation Loss
1.1421	0.4	1000	61.1639	1.1692
1.0556	0.8	2000	51.7749	1.0215
0.7837	1.2	3000	54.1978	0.9572
0.7332	1.6	4000	50.3966	0.9230
0.6957	2.0	5000	50.5772	0.9064
0.6211	2.4	6000	0.9177	49.8590
0.5584	2.8	7000	0.8962	47.5366
0.3952	3.2	8000	0.9025	48.2925
0.4248	3.6	9000	0.9011	50.3995

Framework versions

• Transformers 4.40.2
• Pytorch 2.1.2
• Datasets 2.19.1
• Tokenizers 0.19.1