jeff7522553/whisper-medium-gy

Whisper medium TW - from gy

This model is a fine-tuned version of openai/whisper-medium on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0280
• Cer: 6.6526

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: 1
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 2
• optimizer: Use OptimizerNames.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: 500
• training_steps: 6000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.0463	0.3597	1000	0.0990	7.1173
0.0299	0.7194	2000	0.0628	6.0608
0.0262	1.0791	3000	0.0465	7.0165
0.0332	1.4388	4000	0.0360	7.3738
0.0131	1.7986	5000	0.0316	5.1994
0.0055	2.1583	6000	0.0280	6.6526

Framework versions

• Transformers 4.50.3
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.21.4