gpt4all/gpt4all-bindings/python
Max Cembalest 0989c8d962 formatted note callouts
Signed-off-by: Max Cembalest <mbcembalest@gmail.com>
2024-07-10 09:32:50 -04:00
..
docs formatted note callouts 2024-07-10 09:32:50 -04:00
gpt4all python: fix docstring grammar (#2529) 2024-07-05 12:44:28 -04:00
.gitignore transfer python bindings code 2023-05-10 13:38:32 -04:00
.isort.cfg Python Bindings: Improved unit tests, documentation and unification of API (#1090) 2023-06-30 16:02:02 -04:00
LICENSE.txt transfer python bindings code 2023-05-10 13:38:32 -04:00
makefile python bindings: typing fixes, misc fixes (#1131) 2023-07-03 21:30:24 -04:00
MANIFEST.in transfer python bindings code 2023-05-10 13:38:32 -04:00
mkdocs.yml Create googledrive.md (#2621) 2024-07-09 22:47:23 -04:00
README.md Fix path in Readme (#2339) 2024-05-31 17:20:41 -04:00
setup.py V3 docs max (#2488) 2024-07-01 13:00:14 -04:00

Python GPT4All

This package contains a set of Python bindings around the llmodel C-API.

Package on PyPI: https://pypi.org/project/gpt4all/

Documentation

https://docs.gpt4all.io/gpt4all_python.html

Installation

The easiest way to install the Python bindings for GPT4All is to use pip:

pip install gpt4all

This will download the latest version of the gpt4all package from PyPI.

Local Build

As an alternative to downloading via pip, you may build the Python bindings from source.

Prerequisites

You will need a compiler. On Windows, you should install Visual Studio with the C++ Development components. On macOS, you will need the full version of Xcode—Xcode Command Line Tools lacks certain required tools. On Linux, you will need a GCC or Clang toolchain with C++ support.

On Windows and Linux, building GPT4All with full GPU support requires the Vulkan SDK and the latest CUDA Toolkit.

Building the python bindings

  1. Clone GPT4All and change directory:
git clone --recurse-submodules https://github.com/nomic-ai/gpt4all.git
cd gpt4all/gpt4all-backend
  1. Build the backend.

If you are using Windows and have Visual Studio installed:

cmake -B build
cmake --build build --parallel --config RelWithDebInfo

For all other platforms:

cmake -B build -DCMAKE_BUILD_TYPE=RelWithDebInfo
cmake --build build --parallel

RelWithDebInfo is a good default, but you can also use Release or Debug depending on the situation.

  1. Install the Python package:
cd ../gpt4all-bindings/python
pip install -e .

Usage

Test it out! In a Python script or console:

from gpt4all import GPT4All
model = GPT4All("orca-mini-3b-gguf2-q4_0.gguf")
output = model.generate("The capital of France is ", max_tokens=3)
print(output)

GPU Usage

from gpt4all import GPT4All
model = GPT4All("orca-mini-3b-gguf2-q4_0.gguf", device='gpu') # device='amd', device='intel'
output = model.generate("The capital of France is ", max_tokens=3)
print(output)

Troubleshooting a Local Build

  • If you're on Windows and have compiled with a MinGW toolchain, you might run into an error like:

    FileNotFoundError: Could not find module '<...>\gpt4all-bindings\python\gpt4all\llmodel_DO_NOT_MODIFY\build\libllmodel.dll'
    (or one of its dependencies). Try using the full path with constructor syntax.
    

    The key phrase in this case is "or one of its dependencies". The Python interpreter you're using probably doesn't see the MinGW runtime dependencies. At the moment, the following three are required: libgcc_s_seh-1.dll, libstdc++-6.dll and libwinpthread-1.dll. You should copy them from MinGW into a folder where Python will see them, preferably next to libllmodel.dll.

  • Note regarding the Microsoft toolchain: Compiling with MSVC is possible, but not the official way to go about it at the moment. MSVC doesn't produce DLLs with a lib prefix, which the bindings expect. You'd have to amend that yourself.