Skip to content

Latest commit

 

History

History
21 lines (18 loc) · 528 Bytes

README.md

File metadata and controls

21 lines (18 loc) · 528 Bytes

llama-local

Convert LLama-2/3 Weights to HuggingFace compatible and run locally.

Set-up

  1. The bash script assumes that you've already downloaded your weights from https://llama.meta.com/llama-downloads.
  2. Run bash script.
chmod +x convert_to_hugging_face.sh
./convert_to_hugging_face.sh /home/llama /home/llama/llama-2-7b-chat 7B
  1. Install requirements.
pip install -r requirements.txt
  1. Run python script.
python local_llama.py

Note: Output path will be e.g. /home/llama/llama-2-7b-chat-hf.