20/08/2019, 14:23

Build Chainer + JupyterLab with Docker

Since I know that there is the framework called Chainer of Deep Learning and Machine Learning Here is what I have done: https://github.com/tubutubumustard/chainer_docker.git Dockerfile FROM ubuntu:16.04 WORKDIR /workdir RUN apt-get update -y && apt-get install -y ...

Since I know that there is the framework called Chainer of Deep Learning and Machine Learning

Here is what I have done: https://github.com/tubutubumustard/chainer_docker.git

  • Dockerfile
FROM ubuntu:16.04

WORKDIR /workdir

RUN apt-get update -y && 
    apt-get install -y --no-install-recommends 
    python3-dev 
    python3-pip 
    python3-wheel 
    python3-setuptools 
    git 
    g++ 
    make 
    cmake 
    libblas3 
    libblas-dev 
    && 
    rm -rf /var/lib/apt/lists/* /var/cache/apt/archives/*

RUN pip3 install --upgrade pip

RUN export CHAINER_BUILD_CHAINERX=1
RUN export CHAINERX_BUILD_CUDA=1

RUN pip3 install -U --no-cache-dir 
    'ideep4py<2.1' 
    cupy-cuda92==6.2.0 
    chainer==6.2.0 
    scikit-learn 
    pandas 
    matplotlib 
    jupyterlab

I made the Dockerfile in reference of the followings:

https://hub.docker.com/r/chainer/chainer/dockerfile

https://github.com/chainer/chainer/blob/master/docker/intel/python3/Dockerfile

WORKDIR/workdir is the directory for working with JupyterLab.

I installed iDeep and CuPy because my laptop does n’t have an NVIDIA GPU,

iDeep will be used to calculate.

If you want to use CuPy with Docker, you have to set various things.

(Let's try it if you feel to be interested in)

  • docker-compose.yml
version: '3'
services:
  chainer:
    build: ./
    volumes:
      - ./workdir:/workdir
    command: jupyter lab --port 8000 --ip=0.0.0.0 --allow-root --NotebookApp.token='
    ports:
      - 50020:8000

docker-compose.yml is for practising, so JupyterLab has root permission and token is disabled.

  • sh

$ docker-compose up -d

Access via Browser

Check Chainer version and execution environment

Try to launch the Chainer's tutorial

By the way I tried to used CuPy, but not succeeded

If you want to be enthusiast in machine learning and deep learning, it seems that there is some more useful site to read.

0