finetune_v4

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: 0.2085
• Wer: 14.5161

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: 4
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• 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	6.6667	10	0.2449	19.4700
No log	13.3333	20	0.1970	14.5161
No log	20.0	30	0.1805	11.6359
No log	26.6667	40	0.1826	14.4009
0.0538	33.3333	50	0.1930	22.1198
0.0538	40.0	60	0.1967	36.5207
0.0538	46.6667	70	0.2035	35.3687
0.0538	53.3333	80	0.2085	14.5161

Framework versions

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