Self-learning hands-on for Chainer by Jupyter notebook
Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is 2 commits ahead of hido:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.

README.md

Run the Jupyter notebook

You can download and run the notebooks on your Jupyter-enabled machine.

Tested environment

  • Ubuntu 14.04 LTS
  • NVIDIA GPU (recommended GTX 970 or higher)
  • CUDA 7.5 (installed under /usr/local/cuda)
    • cuDNN v4 (optional)
  • Python 2.7 / 3.4
  • Chainer v1.20.0.1
    • All version after v1.5.0 should work
  • Additional python packages
    • pyparsing == '2.1.10'
    • pydot == '1.2.3'
    • jupyter
    • matplotlib

Set up

Copy .ipynb files, data.py, and image directory into a jupyter-managed directory.

Start

$ jupyter notebook

Build a Chainer & Jupyter environment from scratch on

This repository also provide helper scripts to install CUDA, Chainer, and Jupyter on Ubuntu 14.04.

Install CUDA 7.5 and CUDA toolkit

CUDA libraries will be installed into /usr/local/cuda/.

$ ./install-cuda.sh

Reboot is needed before proceeding. You can also install cuDNN by yourself at this point.

Install Chainer

Python packages will be installed under $HOME/.local/ with pip install --user

$ ./install-chainer.sh

Install Jupyter and visualization packages

$ ./install-jupyter.sh

Run the Jupyter notebook with preset config

Make sure that the environmental variables are loaded.

$ source ~/.bash_profile
$ jupyter notebook

The access URL is https://<ip address>:8888/. The password is "chainer".