whisper-small-es / README.md
dashelruiz's picture
End of training
f5009d7 verified
metadata
language:
  - es
license: apache-2.0
base_model: openai/whisper-small
tags:
  - generated_from_trainer
datasets:
  - facebook/multilingual_librispeech
metrics:
  - wer
model-index:
  - name: Whisper Medium es - Dash Guitar
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: facebook/multilingual_librispeech
          type: facebook/multilingual_librispeech
          config: spanish
          split: test
          args: spanish
        metrics:
          - name: Wer
            type: wer
            value: 7.085875706214689

Whisper Medium es - Dash Guitar

This model is a fine-tuned version of openai/whisper-small on the facebook/multilingual_librispeech dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1535
  • Wer Ortho: 7.0848
  • Wer: 7.0859

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: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: constant_with_warmup
  • lr_scheduler_warmup_steps: 50
  • training_steps: 4000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Ortho Wer
0.3349 0.02 500 0.1782 8.1526 8.1571
0.309 0.04 1000 0.1702 7.5899 7.5921
0.2814 0.05 1500 0.1680 8.0103 8.0124
0.3067 0.07 2000 0.1665 8.1007 8.1028
0.3223 0.09 2500 0.1751 9.2272 9.2294
0.2696 0.11 3000 0.1583 7.2374 7.2395
0.3203 0.13 3500 0.1542 6.9560 6.9559
0.2655 0.14 4000 0.1535 7.0848 7.0859

Framework versions

  • Transformers 4.36.2
  • Pytorch 2.1.2
  • Datasets 2.16.1
  • Tokenizers 0.15.0