``nvidia-docker`` compatible ``Dockerfile`` for ``ray``

README.md

nvidia-docker compatible Dockerfile for ray

Work in progress Dockerfile to create a GPU enabled Ray runtime with Nvidia Docker.

Uses the environment variable RAY_CLUSTER to connect to an existing Ray cluster (given its Redis location).

Dockerfile

FROM nvidia/cuda:10.0-cudnn7-runtime-ubuntu18.04

# At build time we might require a proxy, depending on our network configuration
ARG http_proxy
ARG https_proxy
ENV http_proxy=$http_proxy
ENV https_proxy=$https_proxy
ENV HTTP_PROXY=$http_proxy
ENV HTTPS_PROXY=$https_proxy

ENV DEBIAN_FRONTEND noninteractive
ENV LD_LIBRARY_PATH /usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda-10.0/compat/

RUN apt-get update && apt-get install -y --fix-missing \
    nano \
    wget \
    python3.7 \
    python3-distutils \
    python3-dev

RUN wget -O get-pip.py 'https://bootstrap.pypa.io/get-pip.py';
RUN python3.7 get-pip.py \
      --disable-pip-version-check \
      --no-cache-dir

RUN python3.7 -m pip install ray

CMD python3.7 -c "import os; import ray; ray.init(redis_address=os.environ.get('RAY_CLUSTER') or None)"