Whisper_large_v3_turbo_v2

This model is a fine-tuned version of openai/whisper-large-v3-turbo on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6363
• Wer: 31.7384

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: 0.0001
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8106	1.1834	500	1.0268	93.4686
0.5518	2.3669	1000	0.8523	56.8544
0.4203	3.5503	1500	0.7787	52.2696
0.2934	4.7337	2000	0.7357	48.8402
0.2243	5.9172	2500	0.7544	49.3678
0.1262	7.1006	3000	0.7770	49.9682
0.1038	8.2840	3500	0.7445	43.7824
0.0791	9.4675	4000	0.7615	44.6193
0.057	10.6509	4500	0.7432	41.0079
0.0441	11.8343	5000	0.7307	40.3166
0.0313	13.0178	5500	0.7222	38.7519
0.0147	14.2012	6000	0.7173	37.2965
0.0091	15.3846	6500	0.6866	34.8949
0.0022	16.5680	7000	0.6540	33.5031
0.0025	17.7515	7500	0.6488	32.5298
0.0004	18.9349	8000	0.6363	31.7384

Framework versions

• Transformers 4.45.2
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3