metadata

license: apache-2.0
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: openai/whisper-small-en
    results:
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: myst-test
          type: asr
          config: en
          split: test
        metrics:
          - type: wer
            value: 12.11
            name: WER
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: cslu_scripted
          type: asr
          config: en
          split: test
        metrics:
          - type: wer
            value: 2.74
            name: WER
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: cslu_spontaneous
          type: asr
          config: en
          split: test
        metrics:
          - type: wer
            value: 32.72
            name: WER
      - task:
          type: automatic-speech-recognition
          name: Automatic Speech Recognition
        dataset:
          name: librispeech
          type: asr
          config: en
          split: testclean
        metrics:
          - type: wer
            value: 7.97
            name: WER

openai/whisper-small-en

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

Loss: 0.21313807368278503
Wer: 8.312998407517968

Training and evaluation data

Training data: Myst Train (125 hours) + CSLU Scripted train (35 hours) Evaluation data: Myst Dev (20.9 hours) + CSLU Scripted Dev(4.8)

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e-05
train_batch_size: 32
eval_batch_size: 32
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
converged_after: 1500