I have downloaded the llama2 model to local path, but program always try downloaded the llama2 from huggingface

I have downloaded the llama2 model to local path, but program always try downloaded the llama2 from huggingface, how do I let the program know the model in my local path? or is there any way to customize the path like `AutoModel.from_pretrained`？Thank you.

The report names no repository file, test, or entry point. Start by tracing how Llama2 model loading handles a local path versus Hugging Face, and compare the desired behavior with AutoModel.from_pretrained; done means a configured local path is used without attempting a Hugging Face download.

Written by the indexing model from the issue text.