wenet-stt-base-mn-v1

wenet-stt-base-mn-v1

This model is a WeNet Conformer model fine-tuned for Mongolian speech recognition.

Model Details

• Architecture: Conformer
• Model Size: base
• Language: Mongolian (mn)
• Framework: WeNet
• License: Apache 2.0

Model Configuration

• Output Size: 256
• Attention Heads: 4
• Number of Blocks: 12
• Base model - Good balance of speed and accuracy

Training Details

• Batch Size: 16
• Learning Rate: 0.001
• Max Epochs: 100
• Gradient Clipping: 5

Usage

# Example usage with WeNet
# See WeNet documentation for detailed inference instructions

Citation

If you use this model, please cite:

@misc{wenet-stt-base-mn-v1,
  author = {Batuka0901},
  title = {wenet-stt-base-mn-v1},
  year = {2025},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/Batuka0901/wenet-stt-base-mn-v1}}
}

Contact

For questions or issues, please open an issue on the model repository.