AutoModelForCausalLM.from_pretrained gets stuck when loading model from local folder

Hi, I’m hosting my app on modal com.

When I define it like this, implying that is supposed to be pulled from the repo it works fine, with exception of the time I have to wait for the model to be pulled.

    tokenizer = AutoTokenizer.from_pretrained("NousResearch/Llama-2-7b-chat-hf")
    model = AutoModelForCausalLM.from_pretrained("NousResearch/Llama-2-7b-chat-hf",)

but when I download contents from here, and put them into folder in the container and define the model like this so it could be accessed locally it gets stuck:

    tokenizer = AutoTokenizer.from_pretrained(MODEL_FILES_PATH)
    model = AutoModelForCausalLM.from_pretrained(MODEL_FILES_PATH)

It’s the same environment, the files are accessible. What could be the cause?

P.S. I’ve tried enabling safetensors, but it didn’t change much.

Have a look at this page :