/chainer

A flexible framework of neural networks for deep learning

Primary LanguagePythonMIT LicenseMIT

chainer: Neural network framework

Requirements

Minimum requirements:

  • Python 2.7+, 3.4+
  • NumPy
  • Six 1.9+

Requirements for some features:

  • CUDA support
    • CUDA 6.5+
    • PyCUDA
    • scikits.cuda (pip install scikits.cuda>=0.5.0b2,!=0.042)
    • Mako (depending through PyCUDA)
  • CuDNN support
    • CuDNN v2
  • Caffe model support
    • Python 2.7+ (Py3 is not supported)
    • Protocol Buffers (pip install protobuf)
  • Testing utilities
    • Nose

Installation

Install Chainer via PyPI:

pip install chainer

You can also install Chainer from the source code:

python setup.py install

If you want to enable CUDA, first you have to install CUDA and set the environment variable PATH and LD_LIBRARY_PATH for CUDA executables and libraries. For example, if you are using Ubuntu and CUDA is installed by the official distribution, then CUDA is installed at /usr/local/cuda. In this case, you have to add the following line to .bashrc or .zshrc (choose which you are using):

export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH

Do not forget to restart your terminal session (or source it) to enable this change. Then, install CUDA-related dependent packages via pip:

pip install chainer-cuda-deps

or, from the source:

python cuda_deps/setup.py install

License

MIT License (see LICENSE file).