whisper-large-qve / README.md
cportoca's picture
End of training
bb76c1b verified
|
raw
history blame
2.35 kB
metadata
library_name: transformers
language:
  - qve
license: apache-2.0
base_model: openai/whisper-large
tags:
  - generated_from_trainer
datasets:
  - cportoca/Quechua_dataset
metrics:
  - wer
model-index:
  - name: Whisper Large Ja-Qve - cportoca
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Quechua_dataset
          type: cportoca/Quechua_dataset
          args: 'config: Qve, split: train/test'
        metrics:
          - name: Wer
            type: wer
            value: 14.02572952620019

Whisper Large Ja-Qve - cportoca

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

  • Loss: 0.1918
  • Wer: 14.0257

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: 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: 8000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.2039 1.3550 1000 0.2614 25.1961
0.0757 2.7100 2000 0.1891 33.3856
0.0211 4.0650 3000 0.1846 16.0966
0.0112 5.4201 4000 0.1876 15.8770
0.0027 6.7751 5000 0.1875 14.1826
0.0006 8.1301 6000 0.1888 14.2140
0.0002 9.4851 7000 0.1897 13.9316
0.0001 10.8401 8000 0.1918 14.0257

Framework versions

  • Transformers 4.46.3
  • Pytorch 2.4.1+cu121
  • Datasets 3.1.0
  • Tokenizers 0.20.3