Downloads · 30 days
0
tripura-st-lab
vivekharry/tripura-st-lab
tripura-st-lab is a automatic speech recognition model from vivekharry. Use it when you need speech turned into text. It is set up for transformers. The card lists the license as mit.
Training code and dataset registry for Kokborok-first speech translation, then Bengali and Marathi on BhasaAnuvaad.
Access
Public
Updated Sep 8, 2026
Repo size
—
Likes
0
Public
Repo makeup
Click a slice to open those files.
.py27.9 KB · 83%
At a glance
- Task
- Automatic Speech Recognition
- Library
- transformers
- License
- mit
- Access
- Public
- Created
- Sep 8, 2026
- Updated
- Sep 8, 2026
- SHA
- 682fe9eb
Model card
From the Hugging Face model README
Tripura ST Lab
Training code and dataset registry for Kokborok-first speech translation, then Bengali and Marathi on BhasaAnuvaad.
Landing page: https://huggingface.co/spaces/vivekharry/tripura-st-lab-site
Kokborok AUDIO (16 kHz mono)
→ ASR (MMS-1B + NE-ASR adapter, or Whisper fine-tune)
Kokborok transcript
→ MT (NLLB / sdmy/kokborok)
English
Clone this repo, then:
pip install -e .
python -m indic_st.train_asr --dataset kokborok_asr --init-from openai/whisper-small
python -m indic_st.train_mt --dataset kokborok_mt
See the full README in the files tab for dataset IDs, tonal-augmentation rules, and citations.
Author: Vivek Das (vivekharry).
- Task
- Automatic Speech Recognition
- Library
- transformers
- License
- mit
- Languages
- trp, bn, mr, en
- Created
- Sep 8, 2026
- Updated
- Sep 8, 2026