This post hasn't been updated for 2 years
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
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:
WORKDIR/workdir is the directory for working with
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)
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.
Execute container, access to
$ docker-compose up -d
Access via Browser
Chainer version and execution environment
Try to launch the
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.
All Rights Reserved