Whisper Small Tel

🚀 Whisper Telugu - Fine-tuned

This model is a fine - tuned version of openai/whisper-large-v2 on the Telugu Audio Dataset. It offers significant value in the field of automatic speech recognition for the Telugu language, providing a more accurate and efficient solution for transcribing Telugu audio.

📚 Documentation

Model Information

Model Performance

This model achieves the following results on the evaluation set:

• Loss: 3.5889
• Wer: 92.3967

Model Index

• Name: Whisper Telugu - Fine-tuned
• Results:
  • Task:
    • Name: Automatic Speech Recognition
    • Type: automatic-speech-recognition
  • Dataset:
    • Name: Telugu Audio Dataset
    • Type: sagarchapara/telugu-audio
    • Config: te_in
    • Split: None
    • Args: 'split: train'
  • Metrics:
    • Name: Wer
    • Type: wer
    • Value: 92.39665881345041

Training and Evaluation

Training Hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0002
• train_batch_size: 4
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• 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: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training Results

Framework Versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0

📄 License

This model is released under the apache - 2.0 license.