whisper-small-malay / README.md
M00dler's picture
End of training
25bae8f verified
metadata
language:
  - my
license: apache-2.0
base_model: openai/whisper-small
tags:
  - generated_from_trainer
datasets:
  - malaysia-ai/malay-conversational-speech-corpus
metrics:
  - wer
model-index:
  - name: Whisper small Malay (4 batch size) - Gab
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: malay-conversational-speech-corpus
          type: malaysia-ai/malay-conversational-speech-corpus
          args: 'config: malay, split: test'
        metrics:
          - name: Wer
            type: wer
            value: 27.394540942928042

Whisper small Malay (4 batch size) - Gab

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

  • Loss: 0.7126
  • Wer: 27.3945

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: 16
  • eval_batch_size: 4
  • 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: 4000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.0217 6.1728 1000 0.5993 28.8586
0.0013 12.3457 2000 0.6816 28.0397
0.0003 18.5185 3000 0.7018 27.8660
0.0002 24.6914 4000 0.7126 27.3945

Framework versions

  • Transformers 4.41.1
  • Pytorch 2.3.0+cu121
  • Datasets 2.19.1
  • Tokenizers 0.19.1