How to install LLama 2.0 on virtual Envirnoment

Question

I am trying to run llama on my local server. After running Quickstart.ipynb Iam getting error llama_recipes not found. I tried to install llama_recipes from directory but didn't work. Also while loading, I am getting error about accelerator

Answer 1

1. Create a virtual environment
2. Clone repo: https://github.com/facebookresearch/llama-recipes
3. Install requirements from dev_requirement.txt an requirements.txt using command pip install -r [filename].txt
4. Move llama_recipes folder from src directory and quickstart.ipynb from examples directory to same directory.
5. Install/upgrade jupyter and ipywidgets if you are running jupyter notebook on VSCode.
6. if you still face dependecy error about accelerator/bitsandbytes edit load_in_8bit=True to load_in_8bit=False.