whisper-small-vi / README.md
huybunn's picture
Update README.md
25dfcf5 verified
metadata
library_name: transformers
language:
  - vi
license: apache-2.0
base_model: openai/whisper-small
tags:
  - generated_from_trainer
datasets:
  - doof-ferb/infore1_25hours
metrics:
  - wer
model-index:
  - name: Whisper Small Vietnamese - Huybunn
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Infore1 25hours(50%)
          type: doof-ferb/infore1_25hours
        metrics:
          - name: Wer
            type: wer
            value: 4.219436970929851

Whisper Small Vietnames - Huybunn

This model is a fine-tuned version of openai/whisper-small on the Infore1 25hours(50%) dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1043
  • Wer: 4.2194

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: 4
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 16
  • optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • 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
0.034 2.6747 1000 0.1046 5.2202
0.0017 5.3481 2000 0.0985 4.2982
0.0007 8.0214 3000 0.1024 4.2129
0.0005 10.6961 4000 0.1043 4.2194

Framework versions

  • Transformers 4.50.3
  • Pytorch 2.5.1+cu121
  • Datasets 3.5.0
  • Tokenizers 0.21.0