bagasshw/whisper-large-v2

Whisper Large-v2 Java - HQ TTS

This model is a fine-tuned version of openai/whisper-large-v2 on the jv_id_tts dataset. It achieves the following results on the evaluation set:

• Loss: 0.1792
• Wer: 9.2409

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• 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: 100
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2752	0.8587	1000	0.2629	18.0899
0.1117	1.7170	2000	0.2091	14.7071
0.0656	2.5754	3000	0.1855	12.2112
0.0294	3.4337	4000	0.1709	10.8911
0.0182	4.2920	5000	0.1662	10.4992
0.01	5.1503	6000	0.1709	10.0660
0.0084	6.0086	7000	0.1681	9.6328
0.0057	6.8673	8000	0.1689	9.0965
0.0019	7.7256	9000	0.1780	9.2409
0.0005	8.5839	10000	0.1792	9.2409

Framework versions

• Transformers 4.51.3
• Pytorch 2.7.0+cu126
• Datasets 2.18.0
• Tokenizers 0.21.1