llm-recipes

Sarvam AI

Using local indic LLM

sarvam-1
- https://huggingface.co/sarvamai/sarvam-1
- Collab Demo
- Spaces - Huggingface
How to build a Instruction following model using Base model
TODO
- Part A
  - Download model from HuggingFace
  - Fine tune dataset with saamvad ?
    - English - Hindi
  - Download finetuned model
  - Convert model to ollama
  - Deploy with gradio for sarvam-1
- Part B
  - Download weights
  - make gguf file for ollama with llama.cpp
  - upload gguf to ollama
With Ollama
- sarvam-2b
  - Ollama - https://ollama.com/gaganyatri/sarvam-2b-v0.5
- python -m venv venv
- source venv/bin/activate
- pip install -r requirements-ollama.txt
- Download Model
  - python -m sarvam_ollama load_sarvam
- Execute Code
  - python sarvam_ollama.py
Prequisities
- ollama should be running on Docker
  - For CPU
    - docker compose -f cpu-ollama.yml up -d
  - For GPU
    - docker compose -f gpu-ollama.yml up -d
Use With Huggingface
- sarvam-2b
  - HF Model - https://huggingface.co/sarvamai/sarvam-2b-v0.5
  - python -m venv venv
  - source venv/bin/activate
  - pip install -r requirements-hf.txt
  - huggingface-cli download sarvamai/sarvam-2b-v0.5
  - python text_sarvam.py
- Voice ASR - Shuka
  - HF Model -https://huggingface.co/sarvamai/shuka_v1
  - python -m venv venv
  - source venv/bin/activate
  - pip install -r requirements-shuka.txt
  - huggingface-cli download sarvamai/shuka_v1
  - python voice_shuka.py
Shuka with ollama ?
Is built on https://huggingface.co/fixie-ai/ultravox-v0_3

This site is open source. Improve this page.