primitiv 0.4.0.dev148

primitiv: A Neural Network Toolkit. (Python frontend)

Features

• Dynamic and incremental graph construction

• On-demand memory allocation

• Automatic minibatch broadcasting

• Mostly device-independent

• Simple usage

Install

Prerequisites:

• Python 3 (3.5 or later)

• NumPy (1.11.0 or later)

• Cython (0.27 or later)

• CMake (3.1.0 or later)

• scikit-build (0.6.1 or later)

• (optional) CUDA (7.5 or later)

• (optional) OpenCL (1.2 or later) and OpenCL C++ binding v2

Install dependencies:

pip3 install numpy cython cmake scikit-build

Build and install primitiv without CUDA and OpenCL:

pip3 install primitiv

Build and install primitiv with CUDA and/or OpenCL support:

# Enable only CUDA
pip3 install primitiv --global-option --enable-cuda

# Enable both CUDA and OpenCL
pip3 install primitiv --global-option --enable-cuda --global-option --enable-opencl

Notes

For now, we provide only a source pacakge, and pip command downloads a source package and builds it before installing. This is useful for users to install this library with CUDA/OpenCL backends while keeping compatibility with the manylinux1 standard described in PEP 513.

Resources

• Homepage

• Examples