Bug fix in load_llm_gguf: undefined gguf_filename and obsolete resume_download kwargs

[lmgiorni]

Hello! I am not a programmer, so please take this with a grain of salt.

While setting up the node to run the GGUF model, I encountered two bugs preventing execution in the load_llm_gguf function. I managed to find a workaround and fixed them locally, so I’m sharing the details here in case you want to implement a formal patch.

Issue 1: NameError (gguf_filename is not defined)
In nodes.py, inside the load_kwargs dictionary, the code attempts to pass gguf_filename, which was not defined in the function parameters.

• Fix: I changed "gguf_file": gguf_filename, to "gguf_file": gguf_file, (matching the function's argument).

Issue 2: TypeError with transformers kwargs
After fixing the variable name, AutoModelForCausalLM.from_pretrained throws a TypeError because newer versions of the transformers library no longer accept force_download and resume_download in this context.

• Fix: I commented "force_download": False, and "resume_download": False from the load_kwargs dictionary.

The node is now working perfectly for me with these changes. Thanks for the great work on this project!

[Yuxin-Wen]

Thanks for sharing the fix.

Issue 1 (nodes.py / GGUF): It seems you are using a third-party wrapper (like a ComfyUI custom node). Our official repository does not contain these files, so we recommend submitting this patch to the respective maintainer.

Issue 2 (transformers kwargs): Good catch. This TypeError is indeed due to recent transformers updates deprecating force_download and resume_download. We will review this and patch our code for better compatibility.

I will leave this open for now so others can reference your workaround.