whisper-small-orm-victor

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: 0.4453
• Wer: 0.2909
• Cer: 0.0822

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: 16
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• 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: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.9876	0.4195	500	0.5378	0.3606	0.1034
0.8545	0.8389	1000	0.4632	0.3263	0.0885
0.6583	1.2584	1500	0.4402	0.3086	0.0857
0.6417	1.6779	2000	0.4289	0.3044	0.0860
0.4597	2.0973	2500	0.4235	0.2947	0.0841
0.4930	2.5168	3000	0.4177	0.2875	0.0793
0.4916	2.9362	3500	0.4239	0.2902	0.0823
0.3579	3.3557	4000	0.4534	0.2921	0.0833
0.3583	3.7752	4500	0.4453	0.2909	0.0822

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2