README.md

metadata

language:
  - fy
license: apache-2.0
tags:
  - whisper-event
  - generated_from_trainer
datasets:
  - mozilla-foundation/common_voice_11_0
metrics:
  - wer
model-index:
  - name: Whisper Small Frisian
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: mozilla-foundation/common_voice_11_0 fy-NL
          type: mozilla-foundation/common_voice_11_0
          config: fy-NL
          split: test
          args: fy-NL
        metrics:
          - name: Wer
            type: wer
            value: 21.03

Whisper Small fy-NL - RuudVelo

This model is a fine-tuned version of openai/whisper-small on the Common Voice 11.0 dataset. It achieves the following results on the test set:

• Loss: 0.1443
• Wer: 21.03

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: 64
• eval_batch_size: 8
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Step	Validation Loss	Wer
0.0053	1000	0.4201	21.64
0.0008	2000	0.4607	21.03
0.0004	3000	0.4853	21.11
0.0003	4000	0.5015	21.14
0.0002	5000	0.5084	21.20

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.12.1+cu113
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2