ModelHub Indic Tester
Chat with Airavata (Hindi instruction LLM from IndicInstruct) and other catalog models.
LLM to test
Repo: ai4bharat/Airavata
Gated: yes — set HF_TOKEN secret
64 1024
0 1.2
Speech models (run locally in ModelHub)
These are heavy for a free Space — use your Mac ModelHub app instead:
| Model | Task |
|---|---|
facebook/seamless-m4t-v2-large |
Indic speech-to-speech |
ai4bharat/indic-conformer-600m-multilingual |
Indic ASR |
ai4bharat/indic-parler-tts |
Indic TTS |
nvidia/personaplex-7b-v1 |
English live duplex (MLX) |
Gated models (Airavata)
- Accept the license on ai4bharat/Airavata.
- Space Settings → Repository secrets →
HF_TOKEN= your HF token. - Use ZeroGPU or T4 hardware (7B LLM needs GPU).
HF_TOKEN set on this Space: yes
Airavata citation
Gala et al., Airavata: Introducing Hindi Instruction-tuned LLM, arXiv:2401.15006, 2024.