whisper-large-v3-mn-1

This model is a fine-tuned version of openai/whisper-large-v3 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4498
• Wer: 25.5383
• Cer: 8.1517

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: 16
• eval_batch_size: 8
• 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: 500
• training_steps: 9000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.1072	2.1008	1000	0.3427	33.6407	10.9806
0.0318	4.2017	2000	0.3507	30.6795	10.1042
0.0114	6.3025	3000	0.3679	28.8627	9.3444
0.0051	8.4034	4000	0.3961	28.0655	9.1306
0.0017	10.5042	5000	0.4063	27.0514	8.7835
0.0008	12.6050	6000	0.4152	26.8263	8.6231
0.0005	14.7059	7000	0.4203	26.1565	8.4644
0.0002	16.8067	8000	0.4412	25.6115	8.2113
0.0001	18.9076	9000	0.4498	25.5383	8.1517

Framework versions

• Transformers 4.48.0
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0