Is it possible to run Llama 3.1 without HF libraries ?

[whatdhack]

System Info

latest , linux

Information

• The official example scripts
• My own modified scripts

🐛 Describe the bug

Looking through the llama-recipes, saw references to HuggingFace libraries in many places. In Llama 3.0 , it was possible to run the 8B and 70B without using any HF libraries such as transformers etc.

For example this - https://github.com/meta-llama/llama3/blob/main/example_text_completion.py

Error logs

N/A

Expected behavior

Retain ability to run with a minimal just PyTorch stack.

[wukaixingxp]

Hi! Here is the PR that provide examples to run Llama 3.1 without HF libraries. Please take a look and let me know if you have more questions!

[HamidShojanazeri]

@whatdhack llama-stack is also the new way to run it with/out HF dependencies. Please take a look at that too.

[whatdhack]

The PR is gone from llama-models main.