runtime dockerfile and devel dockerfile

[pineking]

The size of docker image built from the current dockerfile is too large, more than 5GB.
that's better if there are two different dockerfiles: runtime and devel dockerfile.
That would be a smaller size for runtime docker image.

[chenzhekl]

Virtualenv with pip provides a good enough isolated environment for runtime. I don't see a necessity for a runtime Dockerfile.

[ngimel]

The following runtime Dockerfile produces an 3 GB image, compared to ~3.5 GB image produced from the devel Dockerfile currently in the tree. If this reduction is worth it, I can submit a PR.

FROM ubuntu:16.04 

LABEL com.nvidia.volumes.needed="nvidia_driver"
RUN apt-get update && apt-get install -y --no-install-recommends \
         build-essential \ 
         git \
         curl \
         ca-certificates \
         libjpeg-dev \
         libpng-dev && \
     rm -rf /var/lib/apt/lists/*

RUN curl -o ~/miniconda.sh -O  https://repo.continuum.io/miniconda/Miniconda3-4.2.12-Linux-x86_64.sh  && \
     chmod +x ~/miniconda.sh && \
     ~/miniconda.sh -b -p /opt/conda && \     
     rm ~/miniconda.sh && \
     /opt/conda/bin/conda install conda-build && \
     /opt/conda/bin/conda create -y --name pytorch-py35 python=3.5.2 numpy pyyaml scipy ipython mkl&& \
     /opt/conda/bin/conda clean -ya 
ENV PATH /opt/conda/envs/pytorch-py35/bin:$PATH
RUN conda install --name pytorch-py35 -c soumith magma-cuda80 && /opt/conda/bin/conda clean -ya
RUN conda install --name pytorch-py35 pytorch torchvision cuda80 -c soumith && /opt/conda/bin/conda clean -ya

ENV LD_LIBRARY_PATH /usr/local/nvidia/lib:/usr/local/nvidia/lib64

WORKDIR /workspace
RUN chmod -R a+w /workspace

[soumith]

sure whynot!

[soumith]

fixed in #1732