whisper-small-spa / README.md
saul1917's picture
End of training
4e33981 verified
metadata
language:
  - spa
license: apache-2.0
base_model: openai/whisper-small
tags:
  - generated_from_trainer
datasets:
  - CIEMPIESS
metrics:
  - wer
model-index:
  - name: Whisper Small Spanish
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: CIEMPIESS
          type: CIEMPIESS
          args: 'config: spa, split: test'
        metrics:
          - name: Wer
            type: wer
            value: 0.38445504655148455

Whisper Small Spanish

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

  • Loss: 0.0152
  • Wer: 0.3845

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: 8
  • 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: 1000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.2277 0.4484 100 0.2420 9.5187
0.2788 0.8969 200 0.1965 7.9300
0.1799 1.3453 300 0.1541 6.9804
0.158 1.7937 400 0.1121 5.4762
0.0926 2.2422 500 0.0810 4.1792
0.0912 2.6906 600 0.0584 3.7438
0.0451 3.1390 700 0.0364 2.4318
0.0416 3.5874 800 0.0261 1.3271
0.0312 4.0359 900 0.0178 0.4736
0.015 4.4843 1000 0.0152 0.3845

Framework versions

  • Transformers 4.40.2
  • Pytorch 2.2.1+cu121
  • Datasets 2.19.1
  • Tokenizers 0.19.1