whisper-large-v3-lora-bn-en-banking

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

• Loss: 4.6914
• Wer: 98.5075
• Cer: 94.7368

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: 1
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 8
• optimizer: Use OptimizerNames.ADAFACTOR and the args are: No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
2.3917	1.0	1	4.8753	98.5075	94.7368
2.3917	2.0	2	4.8739	98.5075	94.7368
2.3911	3.0	3	4.8704	98.5075	94.7368
2.3893	4.0	4	4.8605	98.5075	94.7368
2.3856	5.0	5	4.8488	98.5075	94.7368
2.3801	6.0	6	4.8313	98.5075	94.7368
2.3728	7.0	7	4.8101	98.5075	94.7368
2.3642	8.0	8	4.7857	98.5075	94.7368
2.354	9.0	9	4.7571	98.5075	94.7368
2.3421	10.0	10	4.7262	98.5075	94.7368
2.3302	11.0	11	4.6914	98.5075	94.7368

Framework versions

• PEFT 0.15.0
• Transformers 4.49.0
• Pytorch 2.7.1+cu118
• Datasets 3.4.1
• Tokenizers 0.21.1