Edit model card

Whisper Large v3 Physician Dictation GPT 4 turbo

This model is a fine-tuned version of openai/whisper-medium on the Physician Dictation GPT 4 Turbo dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1660
  • Wer: 5.3497

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
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • training_steps: 8500
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.0143 3.9683 500 0.1068 4.9948
0.0023 7.9365 1000 0.1287 5.0757
0.0026 11.9048 1500 0.1254 4.9409
0.0003 15.8730 2000 0.1298 4.7298
0.001 19.8413 2500 0.1312 5.0173
0.0001 23.8095 3000 0.1405 5.2374
0.0001 27.7778 3500 0.1454 4.9903
0.0001 31.7460 4000 0.1497 5.2104
0.0 35.7143 4500 0.1531 5.1835
0.0 39.6825 5000 0.1558 5.1431
0.0 43.6508 5500 0.1581 5.1296
0.0 47.6190 6000 0.1601 5.1700
0.0 51.5873 6500 0.1619 5.1925
0.0 55.5556 7000 0.1635 5.2329
0.0 59.5238 7500 0.1648 5.2733
0.0 63.4921 8000 0.1656 5.3362
0.0 67.4603 8500 0.1660 5.3497

Framework versions

  • Transformers 4.40.2
  • Pytorch 2.0.1+cu117
  • Datasets 2.19.1
  • Tokenizers 0.19.1
Downloads last month
13
Safetensors
Model size
764M params
Tensor type
F32
·

Finetuned from

Evaluation results