whisper-small-el / README.md
farsipal's picture
Update README.md
3d701de
metadata
language:
  - el
license: apache-2.0
tags:
  - whisper-event
datasets:
  - mozilla-foundation/common_voice_11_0
metrics:
  - wer
model-index:
  - name: whisper-small-el
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: mozilla-foundation/common_voice_11_0 el
          type: mozilla-foundation/common_voice_11_0
          config: el
          split: test
          args: el
        metrics:
          - name: Wer
            type: wer
            value: 25.696508172362552

Whisper Small - Greek (el)

This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_11_0 el dataset for transcription in Greek. It achieves the following results on the evaluation set:

  • train_loss: 0.0615
  • Wer: 20.2080

Training results

Upon completion of training the best model was reloaded and tested with the following results extracted from the stdout log:

Loading best model from ./whisper-small-el/checkpoint-5000 (score: 20.208023774145616).

{'train_runtime': 73232.697,
'train_samples_per_second': 4.37,
'train_steps_per_second': 0.068,
'train_loss': 0.06146362095708027,
'epoch': 94.34}

TrainOutput(global_step=5000,
    training_loss=0.06146362095708027,
    metrics={'train_runtime': 73232.697,
        'train_samples_per_second': 4.37,
        'train_steps_per_second': 0.068,
        'train_loss': 0.06146362095708027,
        'epoch': 94.34})

Framework versions

  • Transformers 4.26.0.dev0
  • Pytorch 1.13.0
  • Datasets 2.7.1.dev0
  • Tokenizers 0.12.1