whisper-medium-en-cv-6.2

This model is a fine-tuned version of openai/whisper-medium.en on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:

• Loss: 1.1366
• Wer: 31.6595

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-05
• train_batch_size: 48
• eval_batch_size: 4
• 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: 750
• training_steps: 7500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0	0	2.4185	46.5401
0.6822	0.1	750	0.9972	36.9871
0.2058	1.1	1500	1.0039	48.4997
0.0635	2.1	2250	1.0966	42.9884
0.0275	3.1	3000	1.1136	35.3950
0.0149	4.1	3750	1.1359	33.1598
0.0075	5.1	4500	1.1148	37.3546
0.0043	6.1	5250	1.1232	33.9865
0.0008	7.1	6000	1.1331	35.3644
0.0005	8.1	6750	1.1354	31.4452
0.0004	9.1	7500	1.1366	31.6595

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.1
• Tokenizers 0.21.1