charlescao2023's picture
Upload folder using huggingface_hub
58cb504 verified

base: include Llama base model and medusa head head: only medusa head

convert

  1. prepare medusa
git clone https://github.com/FasterDecoding/Medusa.git
cd Medusa
pip install -e .
  1. python3 medusa_model_demo.py -m base -o head --use-full-key --verbose