whisper-cv-1

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: 0.3732
• Wer: 0.2347
• Cer: 0.2844

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
0.2043	1.0	500	0.2237	0.2451	0.2885
0.1082	2.0	1000	0.2422	0.2451	0.2887
0.0315	3.0	1500	0.2712	0.2440	0.2838
0.0188	4.0	2000	0.2902	0.2429	0.2873
0.0214	5.0	2500	0.3111	0.2429	0.2863
0.0054	6.0	3000	0.3255	0.2395	0.2818
0.0004	7.0	3500	0.3459	0.2436	0.2879
0.0003	8.0	4000	0.3520	0.2365	0.2858
0.0002	9.0	4500	0.3713	0.2354	0.2830
0.0002	10.0	5000	0.3752	0.2351	0.2847
0.0002	11.0	5500	0.3838	0.2343	0.2853
0.0001	12.0	6000	0.3787	0.2354	0.2829

Framework versions

• Transformers 4.55.2
• Pytorch 2.7.0+cu126
• Datasets 2.19.1
• Tokenizers 0.21.4