Repo of the code from the Medium article

• llama.cpp: loading model from ./models/ggml-model-q4_0.bin 'std::runtime_error' what(): unexpectedly reached end of file

Buğra Çakır reported an issue, running the code on Python 3.11.3 (main, May 24 2023, 00:00:00) [GCC 13.1.1 20230511 (Red Hat 13.1.1-2)] on linux
He solved the issue installing a different llama-cpp-python version with

pip install llama-cpp-python==0.1.48

• Cannot install llama-cpp-python

This happens usually only on Windows users. Running the installation of llama-cpp-python, required by
LangChain with the llamaEmbeddings, on windows CMake C complier is not installed by default, so you cannot build from source.
On Mac Users with Xtools and on Linux, usually the C complier is already available on the OS.
To avoid the issue you MUST use pre complied wheel.
Go here https://github.com/abetlen/llama-cpp-python/releases
and look for the complied wheel for your architecture and python version - you MUST take Weels Version 0.1.49
because higher versions are not compatible.

In my case I have Windows 10, 64 bit, python 3.10
so my file is llama_cpp_python-0.1.49-cp310-cp310-win_amd64.whl

Update and bug fixes - 2023.05.23

• Some readers faces an issue with langchain.callbacks
  

ImportError: cannot import name 'CallbackManager' from 'langchain.callbacks.base' 
on the line importing CallbackManager in my_langchain.py.

• Michal Founě solved it follwing the issue as described in langchain-ai/chat-langchain#70 with pip install langchain==0.0.142
  
  
  
• On the GitHub repo there is already an issue solved related to GPT4All' object has no attribute '_ctx'. Fixed specifying the versions during pip install like this:

pip install pygpt4all==1.0.1 
pip install pygptj==1.0.10 
pip install pyllamacpp==1.0.6

• Another quite common issue is related to readers using Mac with M1 chip. The arm64 architecture is a little reluctant to work. As suggested to Emile Pretorius and to Yosef Agung Wicaksono you can try to fix it with the guidelines in this document https://docs.google.com/document/d/1JDMBTOjbRtJo49z1SKEnvfG1a4JQMjWFT7hvfZZU4vQ/edit .
  it seems to have worked with them.
  
  
• Bruce Wen suggested to avoid all the problems mentioned above giving already all the versions in the code: all the pip instructions are now updated.
  
  
• BEpshtein suggested to update the article and the Github repo: keep reporting the issues and I will try to reply including the smart troubleshooting of the readers.
  
  
• geert van kempen got the following issue:

File "/Users/XXXX/GPT4ALL_gvk/.venv/lib/python3.10/site-packages/pyllamacpp/model.py", line 402, in __del__
if self._ctx:
AttributeError: 'GPT4All' object has no attribute '_ctx'

He was using pygpt4all version 1.1.0, and pygptj 2.0.3 (and he was using version 1.0.6 of pyllamaccp) and he solved it with the GitHub solved Issue (see above).

• Norman Procope and Kon16ov got problems loading the pdfs from my GitHub repo. It was solved downloading again the pdf or checking if UTF-8 format was used or not: fixed with decoding text = text.decode()
  
  
• Alain Uro and other users got an error on the pygtp4all callbacks. As greatly explained and solved by Rajneesh Aggarwal this happens because the pygpt4all PyPI package will no longer by actively maintained and the bindings may diverge from the GPT4All model backends. He solved it installing instead of pygtp4all pip install gpt4all
  The code must be changed as well as follows:

import gpt4all
model_path = '.\models'
model = gpt4all.GPT4All(model_name='gpt4all-converted.bin', model_path=model_path, model_type='llama', allow_download=True)
model.generate("Once upon a time, ", streaming=True)

• Shamik Dhar also had an issue with the new_text_callback.

def new_text_callback(text):
print(text, end="")
model = GPT4All('./models/gpt4all-converted.bin')
model.generate("Once upon a time, ", n_predict=55, new_text_callback=new_text_callback)
I am getting an error here that there is no argument in 
the generate module of name new_text_callback

It was solved by Oscar Jeong changing the code generate() to cpp_generate() as follows:

# Just change generate() to cpp_generate() then it works.
def new_text_callback(text):
print(text, end="")
model = GPT4All('./models/gpt4all-converted.bin')
model.cpp_generate("Once upon a time, ", n_predict=55, new_text_callback=new_text_callback)