Load an existing model under no Internet condition

[Taimin]

I found that when no Internet is available, FastLanguageModel.from_pretrained will take a long time waiting. In unsloth/load.py, the following code appears:

from huggingface_hub.utils import disable_progress_bars, enable_progress_bars, are_progress_bars_disabled
was_disabled = are_progress_bars_disabled()
disable_progress_bars()

I suspect executing these codes will need access to Internet. If not, they keep waiting and waiting until access is available.
The behavior should be if a local model is available, load it immediately without checking.