Install Guide

Before installing Chainer

We recommend these platforms.

Chainer is supported on Python 2.7.6+, 3.4.3+, 3.5.1+, 3.6.0+. Chainer uses C++ compiler such as g++. You need to install it before installing Chainer. This is typical installation method for each platform:

# Ubuntu 14.04
$ apt-get install g++

# CentOS 7
$ yum install gcc-c++

If you use old setuptools, upgrade it:

$ pip install -U setuptools

Install Chainer

Chainer depends on these Python packages:

CUDA support

cuDNN support

  • cuDNN v2, v3, v4, v5, v5.1, v6

Caffe model support

All these libraries are automatically installed with pip or

Image dataset is optional

HDF5 serialization is optional

Install Chainer via pip

We recommend to install Chainer via pip:

$ pip install chainer

Install Chainer from source

You can use to install Chainer from source:

$ tar zxf chainer-x.x.x.tar.gz
$ cd chainer-x.x.x
$ python install

When an error occurs...

Use -vvvv option with pip command. That shows all logs of installation. It may helps you:

$ pip install chainer -vvvv

Install Chainer with CUDA

You need to install CUDA Toolkit before installing Chainer. If you have CUDA in a default directory or set CUDA_PATH correctly, Chainer installer finds CUDA automatically:

$ pip install chainer


Chainer installer looks up CUDA_PATH environment variable first. If it is empty, the installer looks for nvcc command from PATH environment variable and use its parent directory as the root directory of CUDA installation. If nvcc command is also not found, the installer tries to use the default directory for Ubuntu /usr/local/cuda.

If you installed CUDA into a non-default directory, you need to specify the directory with CUDA_PATH environment variable:

$ CUDA_PATH=/opt/nvidia/cuda pip install chainer


If you want to use sudo to install Chainer, note that sudo command initializes all environment variables. Please specify CUDA_PATH environment variable inside sudo like this:

$ sudo CUDA_PATH=/opt/nvidia/cuda pip install chainer

Install Chainer with CUDA and cuDNN

cuDNN is a library for Deep Neural Networks that NVIDIA provides. Chainer can use cuDNN. If you want to enable cuDNN, install cuDNN and CUDA before installing Chainer. We recommend you to install developer library of deb package of cuDNN.

If you want to install tar-gz version, we recommend you to install it to CUDA directory. For example if you uses Ubuntu Linux, copy .h files to include directory and .so files to lib64 directory:

$ cp /path/to/cudnn.h $CUDA_PATH/include
$ cp /path/to/* $CUDA_PATH/lib64

The destination directories depend on your environment.

If you want to use cuDNN installed in other directory, please use CFLAGS, LDFLAGS and LD_LIBRARY_PATH environment variables before installing Chainer:

export CFLAGS=-I/path/to/cudnn/include
export LDFLAGS=-L/path/to/cudnn/lib
export LD_LIBRARY_PATH=/path/to/cudnn/lib:$LD_LIBRARY_PATH

Install Chainer for developers

Chainer uses Cython (>=0.24). Developers need to use Cython to regenerate C++ sources from pyx files. We recommend to use pip with -e option for editable mode:

$ pip install -U cython
$ cd /path/to/chainer/source
$ pip install -e .

Users need not to install Cython as a distribution package of Chainer only contains generated sources.

Support image dataset

Install Pillow manually to activate image dataset. This feature is optional:

$ pip install pillow

Support HDF5 serialization

Install h5py manually to activate HDF5 serialization. This feature is optional:

$ pip install h5py

Before installing h5py, you need to install libhdf5. It depends on your environment:

# Ubuntu 14.04
$ apt-get install libhdf5-dev

# CentOS 7
$ yum -y install epel-release
$ yum install hdf5-devel

Uninstall Chainer

Use pip to uninstall Chainer:

$ pip uninstall chainer


When you upgrade Chainer, pip sometimes installed various version of Chainer in site-packages. Please uninstall it repeatedly until pip returns an error.

Upgrade Chainer

Just use pip with -U option:

$ pip install -U chainer

Reinstall Chainer

If you want to reinstall Chainer, please uninstall Chainer and then install it. We recommend to use --no-cache-dir option as pip sometimes uses cache:

$ pip uninstall chainer
$ pip install chainer --no-cache-dir

When you install Chainer without CUDA, and after that you want to use CUDA, please reinstall Chainer. You need to reinstall Chainer when you want to upgrade CUDA.

Run Chainer with Docker

We provide the official Docker image. Use nvidia-docker command to run Chainer image with GPU. You can login to the environment with bash, and run the Python interpreter:

$ nvidia-docker run -it chainer/chainer /bin/bash

Or, run the interpreter directly:

$ nvidia-docker run -it chainer/chainer /usr/bin/python

What “recommend” means?

We tests Chainer automatically with Jenkins. All supported environments are tested in this environment. We cannot guarantee that Chainer works on other environments.


The installer says “hdf5.h is not found”

You don’t have libhdf5. Please install hdf5. See Before installing Chainer.

MemoryError happens

You maybe failed to install Cython. Please install it manually. See When an error occurs....

Examples says “cuDNN is not enabled”

You failed to build Chainer with cuDNN. If you don’t need cuDNN, ignore this message. Otherwise, retry to install Chainer with cuDNN. -vvvv option helps you. See Install Chainer with CUDA and cuDNN.