Use pre-downloaded model of Llama-3

#6
by Gupta-Rushil - opened

Hey,

While running the script mentioned in the model card, it seems that the command tries to download the original Llama-3 from the Meta's repository on huggingface. However, I have the Llama-3 model already downloaded on my system. Can I pass some path somewhere so that it looks for the model checkpoints in that directory instead of downloading it fresh?

Many thanks!

Sign up or log in to comment