Getting Started

Prerequisites

It is encouraged to use environment wrapper and package manager, conda is chosen as the reference solution. Please follow the installation section in their official guide.

Some of the codes require CUDA dependency, please download the binary release from the NVIDIA website.

Quick start

Create an empty workspace named demo, if you have your preferred environment, feel free to skip the following two lines. It is highly recommended to use numpy and scipy from conda instead of pip, since their version has mkl support directly.

conda create -n demo python=3 numpy scipy
conda activate demo

Install uToolbox using pip

pip install utoolbox

or try out pre-release by

pip install --pre utoolbox

Congratz! You can now try out stuff in the guides section.

Frequently Asked Questions

What environment combinations are tested?

Platform

Version

CUDA

Windows

Windows 7 (64-bit) SP1

9.2.88.1

10.1.168

Linux

Debian 8.10, Linux 3.12.72

8.0.44

macOS *

High Sierra 10.13.6, Darwin 17.7.0

9.2.64.1

*

Since I no longer owned a macOS environment, macOS support is currently stalled.