primitiv: A Neural Network Toolkit. (Python frontend)
Project description
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)
scikit-build (0.6.1 or later, only for building)
(optional) CUDA (7.5 or later)
(optional) OpenCL (1.2 or later) and OpenCL C++ binding v2
Install dependencies:
pip3 install numpy cython 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
According to the manylinux1 policy described in PEP 513, binary packages are required to depend only on an extremely limited set of external shared libraries. Most users may install primitiv with CUDA and/or OpenCL backends that are not supported in the manylinux1 policy. For now, we provide only a source pacakge.
Resources
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.