Dockerfile

# Use images provided by Nvida with CUDA and CUDNN preinstalled
FROM nvidia/cuda:8.0-cudnn5-runtime-ubuntu16.04

# These are apparently tensorflow deps
RUN apt-get update && apt-get install -y --no-install-recommends \
        netbase \
        build-essential \
        curl \
        libfreetype6-dev \
        libpng12-dev \
        libzmq3-dev \
        pkg-config \
	      python3 \
        python3-dev \
        python3-pip \
        libcupti-dev \ 
        rsync \
        software-properties-common \
        unzip \
        libgtk2.0-0 \
        git \
	      tcl-dev \
	      tk-dev \
        && \
    apt-get clean && \
    rm -rf /var/lib/apt/lists/*

# Upgrade pip and install setuptools
RUN pip3 install --upgrade pip setuptools

# Install scipy packages and some utils
RUN pip3 install --no-cache-dir numpy \
        sklearn \
        Pillow \
        eventlet \
        Flask \  
        python-socketio \
        moviepy \
        scipy \
        opencv-python \
        matplotlib \
        ipython \
        jupyter \
        pandas \
        sympy \
        nose \
        h5py

# Install tensorflow with GPU support and keras
RUN pip3 install --no-cache-dir tensorflow-gpu keras

# Setup our workdir
RUN mkdir /src
WORKDIR "/src"

# For CUDA profiling, TensorFlow requires CUPTI.
ENV LD_LIBRARY_PATH /usr/local/cuda/extras/CUPTI/lib64:$LD_LIBRARY_PATH

COPY jupyter_notebook_config.py /root/.jupyter/
# TensorBoard
EXPOSE 6006

# Flask Server
EXPOSE 4567

# Jupyter notebook
EXPOSE 8888