linshoufan/linshoufanfork-whisper-small-nan-tw-pinyin

Whisper Small Taiwanese

This model is a fine-tuned version of openai/whisper-small on the Common Voice 15.0 and 16.1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3916
• Wer: 68.5703

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: 16
• 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: 500
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4172	0.32	1000	0.4641	82.9494
0.2962	0.64	2000	0.3834	73.8040
0.229	0.97	3000	0.3537	70.0423
0.1994	1.29	4000	0.3685	71.1599
0.1693	1.61	5000	0.3551	67.8206
0.1398	1.93	6000	0.3526	67.6707
0.1032	2.25	7000	0.3836	69.4834
0.0745	2.58	8000	0.3839	68.5566
0.0558	2.9	9000	0.3916	68.5703

Framework versions

• Transformers 4.38.2
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.2