Edit model card

openai/whisper-medium

This model is a fine-tuned version of openai/whisper-medium on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2790
  • Wer: 8.3986
  • Cer: 5.2582

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: 32
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • training_steps: 10000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Cer
0.1691 1.01 1000 0.1871 10.1740 6.3509
0.0916 2.02 2000 0.1691 8.9797 5.5499
0.0452 3.03 3000 0.1902 8.9814 5.5867
0.0213 4.04 4000 0.2062 8.9375 5.6531
0.0096 5.05 5000 0.2284 8.7331 5.6202
0.0041 6.05 6000 0.2395 8.5051 5.3009
0.0022 7.06 7000 0.2535 8.5507 5.3640
0.001 8.07 8000 0.2656 8.5557 5.3791
0.0006 9.08 9000 0.2721 8.4037 5.2739
0.0004 10.09 10000 0.2790 8.3986 5.2582

Framework versions

  • Transformers 4.26.0.dev0
  • Pytorch 1.13.0+cu117
  • Datasets 2.7.1.dev0
  • Tokenizers 0.13.2
Downloads last month
4
Hosted inference API
or or
This model can be loaded on the Inference API on-demand.