finetune_v11

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: 1.1406
• Wer: 60.3080

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: 4
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 5
• training_steps: 80
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	5.7143	10	1.2480	46.1078
No log	11.4286	20	1.0137	49.1018
No log	17.1429	30	1.0430	43.1993
No log	22.8571	40	1.0820	45.0813
0.2458	28.5714	50	1.1270	54.3199
0.2458	34.2857	60	1.1328	45.8512
0.2458	40.0	70	1.1348	44.7391
0.2458	45.7143	80	1.1406	60.3080

Framework versions

• Transformers 4.42.3
• Pytorch 2.2.0
• Datasets 2.20.0
• Tokenizers 0.19.1