whisper-small-taiwanese-full

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

• Loss: 0.4080
• Cer: 29.8836

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: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 300
• training_steps: 4500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.7008	0.5282	750	0.6961	43.8228
0.3366	1.0563	1500	0.5263	36.2712
0.3311	1.5845	2250	0.4640	32.0063
0.1384	2.1127	3000	0.4337	31.4976
0.1616	2.6408	3750	0.4143	30.0890
0.0742	3.1690	4500	0.4080	29.8836

Framework versions

• Transformers 4.52.4
• Pytorch 2.1.0+cu118
• Datasets 3.6.0
• Tokenizers 0.21.1