whisper_15epoch_specaug

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

• Loss: 0.3102
• Cer: 10.6155

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: 3e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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
• num_epochs: 12
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.5709	1.0	160	0.4689	12.4801
0.3323	2.0	320	0.3196	11.2728
0.299	3.0	480	0.3059	10.9308
0.2705	4.0	640	0.3029	10.7597
0.2507	5.0	800	0.3024	10.6625
0.2383	6.0	960	0.3028	10.5703
0.23	7.0	1120	0.3058	10.6524
0.2248	8.0	1280	0.3066	10.5988
0.2173	9.0	1440	0.3085	10.6457
0.2128	10.0	1600	0.3102	10.6155

Framework versions

• Transformers 4.56.0
• Pytorch 2.8.0+cu129
• Datasets 4.0.0
• Tokenizers 0.22.0