How does the process of using HuggingFace models work? I am using an LLM with gated access right now. I requested access to it on the HuggingFace website, and can then download it on my computer to “C:\Users<Username>.cache\huggingface\hub” and use it there. How does HuggingFace know that my computer is connected to my account? HF_TOKEN does not seem to exist in my computer’s environmental variables (though I do have an access token listed in the Settings page for my account). However, when I turn off my Wi-Fi, the code refuses to load the locally-stored model, presumably because it cannot verify that I have the right to use it, thus indicating that something is being done to check that I can access the model.
Also, where is the place where models are stored specified? I see that they are at “C:\Users<Username>.cache\huggingface\hub”, but I cannot find HF_HOME in my list of environmental variables.
Based on these things, how would I Dockerize a script that relies on a HuggingFace model? How would I allow it to verify that it has permission to use a model, and tell it where in the container the model is located?