asr_java_result / README.md
bagasshw's picture
End of training
861b11a verified
metadata
library_name: transformers
language:
  - jv
license: apache-2.0
base_model: openai/whisper-tiny
tags:
  - whisper
  - javanese
  - asr
  - generated_from_trainer
datasets:
  - jv_id_asr_split
metrics:
  - wer
model-index:
  - name: Whisper Tiny Java
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: jv_id_asr_split
          type: jv_id_asr_split
          config: jv_id_asr_source
          split: validation
          args: jv_id_asr_source
        metrics:
          - name: Wer
            type: wer
            value: 0.6624243173112566

Whisper Tiny Java

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

  • Loss: 1.0129
  • Wer: 0.6624

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: 16
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 128
  • optimizer: Use OptimizerNames.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: 30
  • training_steps: 150
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
3.1804 0.4110 30 1.8947 0.8897
1.5393 0.8219 60 1.2656 0.7560
1.1714 1.2329 90 1.1013 0.7068
1.0264 1.6438 120 1.0346 0.6828
0.9896 2.0548 150 1.0129 0.6624

Framework versions

  • Transformers 4.50.0.dev0
  • Pytorch 2.5.1+cu124
  • Datasets 3.3.2
  • Tokenizers 0.21.0