jethrowang/vanilla-whisper-tiny_tat

Whisper Tiny Taiwanese Condenser

This model is a fine-tuned version of openai/whisper-tiny on the TAT ASR Aligned dataset. It achieves the following results on the evaluation set:

• Loss: 0.6252
• Cer: 11.4109

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: 0.0001
• train_batch_size: 64
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 681
• training_steps: 6810
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.3117	0.9985	681	0.4649	17.1248
0.1805	1.9971	1362	0.4360	13.6667
0.108	2.9956	2043	0.4497	13.4248
0.0648	3.9941	2724	0.4710	12.8500
0.0349	4.9927	3405	0.5276	12.7592
0.0192	5.9912	4086	0.5607	12.4186
0.0089	6.9897	4767	0.5911	12.1183
0.0035	7.9883	5448	0.6032	11.6608
0.0009	8.9868	6129	0.6198	11.5311
0.0005	9.9853	6810	0.6252	11.4109

Framework versions

• Transformers 4.42.3
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1