whisper-turbo-v2

This model is a fine-tuned version of Moaaz5/whisper-turbo-v1 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5788
• Wer: 26.8435

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: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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: 200
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7302	1.0	78	0.4956	30.0734
0.4093	2.0	156	0.4684	29.4539
0.2818	3.0	234	0.4822	28.3592
0.1857	4.0	312	0.4989	29.6584
0.1239	5.0	390	0.5145	27.4329
0.0684	6.0	468	0.5231	28.5757
0.0523	7.0	546	0.5459	27.4029
0.0237	8.0	624	0.5593	27.1262
0.0132	9.0	702	0.5667	27.2465
0.0077	9.8774	770	0.5788	26.8435

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1