TypeError: expected CFunctionType instance instead of function in _pyllmodel.py

[omin23]

Bug Report

When running gpt4all's GPT4ALL() function to initialise a model for use, the program throws an error:

Traceback (most recent call last):
  File "/path/to/script/trygpt.py", line 4, in <module>
    model = GPT4All("Meta-Llama-3-8B-Instruct.Q4_0.gguf")
  File "/path/to/packages/gpt4all/gpt4all-bindings/python/gpt4all/gpt4all.py", line 263, in __init__
    self.model = LLModel(self.config["path"], n_ctx, ngl, backend)
  File "/path/to/packages/gpt4all/gpt4all-bindings/python/gpt4all/_pyllmodel.py", line 294, in __init__
    llmodel.llmodel_model_foreach_special_token(
ctypes.ArgumentError: argument 2: TypeError: expected CFunctionType instance instead of function

Steps to Reproduce

Run test script:

from gpt4all import GPT4All

model = GPT4All("Meta-Llama-3-8B-Instruct.Q4_0.gguf")

with model.chat_session():
    print(model.generate("How can I run LLMs efficiently on my laptop?", max_tokens=1024))

Expected Behavior

The correct model to be downloaded and instantiated.

Your Environment

• GPT4All version: 2.8.3.dev0 (0f27359c)
• Python: 3.10.15
• Operating System: Ubuntu 18.04.6 LTS
• Chat model used (if applicable): n/a

[mgoltzsche]

Right, I also just ran into this while trying to build the latest release (v3.5.3) for arm64/raspberry pi and running the CLI app.
Apparently the issue exists since gpt4all v3.5.0. Finally I was able to build and run it using gpt4all v3.4.2.

Fwiw this is how I've built a working alpine-based gpt4all v3.4.2 python CLI container. Building it with --build-arg GPT4ALL_VERSION=v3.5.3 reproduces the issue.

Btw it is a pity that the latest gpt4all python package that was released to pypi (2.8.2) does not support arm64. I had to build the previously listed Dockerfile on arm64 myself in order to be able to run a recent gpt4all version on the Raspberry Pi.