ArjunJ/QWEN3-ASR-ML

This model is a fine-tuned version of Qwen-ASR for Malayalam.

Training Data

This model was trained using the IndicTTS dataset.

Evaluation Results

Evaluated on the FLEURS Malayalam test set:

• Word Error Rate (WER): 0.5706
• Character Error Rate (CER): 0.1843

License

Creative Commons Attribution 4.0 International (CC BY 4.0)

Usage

To use this model, you can use the following code snippet:

import torch
from qwen_asr import Qwen3ASRModel

model = Qwen3ASRModel.from_pretrained(
    "ArjunJ/QWEN3-ASR-ML",
    dtype=torch.bfloat16,
    device_map="auto",
)

results = model.transcribe(audio="path_to_your_audio.wav")
print(results[0].text)