• Branch master:
• Branch develop:

primitiv

A Neural Network Toolkit.

Features

• Dynamic and incremental graph construction (a.k.a. "define-by-run" style)
• On-demand memory allocation
• Automatic minibatch broadcasting
• Mostly device-independent
• Simple usage

Languages

This repository contains only the core C++ library and its C APIs. Some bindings for other programming languages (e.g., Python) can be found in the official repository.

Prerequisites

• CMake 3.1.0 or later
• C++11 compiler (GCC, Clang)
• (optional) Google Test
  • Required only when -DPRIMITIV_BUILD_TESTS=ON.
• (optional) CUDA 8.0 or later/cuDNN 7.0 or later
  • Required only when -DPRIMITIV_USE_CUDA=ON
• (optional) OpenCL 1.2/OpenCL C++ binding v2
  • Required only when -DPRIMITIV_USE_OPENCL=ON

Documentation

• Official documentation site describes various information including installation, usage and library references.
• Example code shows some actual usages of primitiv.

Contact

• primitiv Developer Group
• @odashi_t (maintainer) on Twitter

This project is supported by ASTREC in NICT.