whisper-small-coraa-l.1

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

• Loss: 0.9314
• Wer: 0.3360
• Cer: 0.4440

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: 8
• eval_batch_size: 2
• seed: 42
• 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
• num_epochs: 12

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.9302	1.0	500	0.7463	0.3403	0.4312
0.4418	2.0	1000	0.7186	0.3548	0.4452
0.2068	3.0	1500	0.7844	0.3875	0.4514
0.0737	4.0	2000	0.8452	0.3796	0.4520
0.03	5.0	2500	0.8570	0.3869	0.4585
0.0118	6.0	3000	0.9240	0.3396	0.4447
0.0083	7.0	3500	0.9640	0.3495	0.4458
0.0024	8.0	4000	0.9781	0.3538	0.4473
0.0027	9.0	4500	1.0058	0.3697	0.4569
0.0006	10.0	5000	1.0003	0.3849	0.4550
0.0005	11.0	5500	1.0013	0.3555	0.4461
0.0004	12.0	6000	0.9906	0.3627	0.4481

Framework versions

• Transformers 4.55.2
• Pytorch 2.7.0+cu126
• Datasets 4.0.0
• Tokenizers 0.21.4