Link Search Menu Expand Document
Pudding

Pudding 🍮

Welcome to the official site for the project Pudding!

Pudding enables you to run various machine learning algorithms on Nvidia’s GPU. It is written in C/C++ and CUDA. To make it easier for one to use, it also comes with a Python binding.

Installation

Currently Pudding has only been tested on Linux.

Dependencies

Compiling the Pudding shared libarary requires:

  • CMake >= 3.18
  • gcc >= 5.4.0
  • CUDA 10.1

Installation Steps

First you need to compile the shared libaray:

mkdir build
cd build
cmake --DCMAKE_BUILD_TYPE=Release ..
cmake --build .

Or you can simply use the provided scripts:

sh scripts/release.sh

You can run tests if you want to check the build:

cmake --build . -t test

If you need the Python binding, go back to the root directory and run:

cd python
python setup.py install

You can run tests if you want (again, assume you are in the root directory and be sure to install the pytest package in Python):

pytest python/tests