Multistage docker (flashlight#396)

Summary:
**Original Issue**: This dicussion started [here](flashlight#225 (comment))

The main point of this PR is to update dockerfiles to use a multistage build, reducing build time and final image size. But it also adds to small but important features:

- `.dockerignore` to avoid passing unnecessary files to the build context
- `ccache` support to CMake

> I'm sure that most of the packages installed with apt on some builds and especially on the final image can be removed. But you know better which ones to take out. This would further reduce the final image size

| image                          | build-time | size   |
|--------------------------------|------------|--------|
| cpu-base-consolidation-latest  | 454.4s     | 4.76GB |
| cpu-latest                     | 433.6s     | 3.38GB |
| cuda-base-consolidation-latest | 970.3s     | 13.7GB |
| cuda-latest                    | 440.3s     | 9.31GB |

Pull Request resolved: flashlight#396

Reviewed By: jacobkahn

Differential Revision: D25820782

Pulled By: tlikhomanenko

fbshipit-source-id: eef3d0550dee862d9bbfb24aa06f716c92ef75e8

.docker/Dockerfile-CPU

@@ -4,18 +4,75 @@
 # flashlight       master     (git, CPU backend)
 # ==================================================================
 
-FROM flml/flashlight:cpu-base-consolidation-latest
-
-ENV MKLROOT="/opt/intel/mkl"
+FROM flml/flashlight:cpu-base-consolidation-latest as build
 
 # ==================================================================
 # flashlight with CPU backend
 # ------------------------------------------------------------------
 # Setup and build flashlight
-RUN mkdir /root/flashlight
+RUN mkdir /tmp/flashlight
+
+COPY . /tmp/flashlight
+
+RUN mkdir -p /tmp/flashlight/build && \
+    cd /tmp/flashlight/build && \ 
+    cmake -DCMAKE_BUILD_TYPE="Release" -DCMAKE_INSTALL_PREFIX="/opt/flashlight" -DFL_BACKEND="CPU" -DFL_LIBRARIES_USE_CUDA="OFF" -DFL_BUILD_TESTS="ON" -DDNNL_DIR="/opt/onednn/lib/cmake/dnnl" -DGloo_INCLUDE_DIR="/opt/gloo/include" -DGloo_NATIVE_LIBRARY="/opt/gloo/lib/libgloo.a" .. && \
+    make -j$(nproc) && \
+    make install -j$(nproc)
+
+
 
-COPY . /root/flashlight
+#############################################################################
+#                            SECOND STAGE                                   #
+#############################################################################
 
-RUN cd /root/flashlight && mkdir -p build && \
-    cd build && cmake .. -DFL_BACKEND=CPU -DFL_LIBRARIES_USE_CUDA=OFF && \
-    make -j$(nproc) && make install
+FROM ubuntu:18.04
+
+ENV DEBIAN_FRONTEND=noninteractive
+
+# install dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+        g++ \
+        # for cmake
+        zlib1g-dev libcurl4-openssl-dev \
+        # for MKL
+        apt-transport-https \
+        gpg-agent gnupg2 \
+        # for arrayfire CPU backend
+        # OpenBLAS
+        libopenblas-dev libfftw3-dev liblapacke-dev \
+        # ATLAS
+        libatlas3-base libatlas-base-dev libfftw3-dev liblapacke-dev \
+        # ssh for OpenMPI
+        openssh-server openssh-client \
+        # OpenMPI
+        libopenmpi-dev libomp-dev openmpi-bin \
+        # libsndfile
+        libsndfile1-dev \
+        # for libsndfile for ubuntu 18.04
+        libopus-dev \
+        # FFTW
+        libfftw3-dev \
+        # for kenlm
+        zlib1g-dev libbz2-dev liblzma-dev \
+        # gflags
+        libgflags-dev libgflags2.2 \
+        # for glog
+        libgoogle-glog-dev libgoogle-glog0v5 \
+        # for receipts data processing
+        sox \
+        # for python
+        python3-dev python3-pip python3-distutils && \
+    apt-get clean && \
+    apt-get -y autoremove && \
+    rm -rf /var/lib/apt/lists/*
+
+COPY --from=build /opt/arrayfire  /opt/arrayfire
+COPY --from=build /opt/kenlm      /opt/kenlm
+COPY --from=build /opt/intel      /opt/intel
+COPY --from=build /opt/boost      /opt/boost
+COPY --from=build /opt/flashlight /opt/flashlight
+
+ENV KENLM_ROOT_DIR=/opt/kenlm
+ENV MKLROOT="/opt/intel/mkl"
+ENV LD_LIBRARY_PATH="/opt/boost/lib:$LD_LIBRARY_PATH"

.docker/Dockerfile-CPU-Base

@@ -3,38 +3,144 @@
 # ------------------------------------------------------------------
 # Ubuntu           18.04
 # OpenMPI          latest       (apt)
-# cmake            3.10         (git)
-# MKL              2018.4-057   (apt)
-# arrayfire        3.7.1        (git, CPU backend)
-# oneDNN           2.0          (git)
-# Gloo             b7e0906      (git)
-# libsndfile       4bdd741      (git)
+# cmake            3.10         (oficial binaries)
+# MKL              2020.4-912   (apt)
+# arrayfire        3.7.3        (git, CPU backend)
+# libsndfile       latest       (apt, v1.0.28-4)
+# oneDNN           master       (git)
+# Gloo             1da2117      (git)
 # FFTW             latest       (apt)
 # KenLM            4a27753      (git)
 # GLOG             latest       (apt)
 # gflags           latest       (apt)
-# python           3.6          (apt)
+# python3          latest       (apt)
+# boost            1.75.0       (source)
 # ==================================================================
 
-FROM ubuntu:18.04
+FROM ubuntu:18.04 as cpu_base_builder
 
-ENV APT_INSTALL="apt-get install -y --no-install-recommends"
-ENV MKLROOT="/opt/intel/mkl"
+ENV DEBIAN_FRONTEND=noninteractive
 
-RUN apt-get update && \
-    DEBIAN_FRONTEND=noninteractive $APT_INSTALL \
+RUN apt-get update && apt-get install -y --no-install-recommends \
         build-essential \
         ca-certificates \
-        wget \
+        curl \
+        git \
+        g++
+        # for cmake
+        #zlib1g-dev libcurl4-openssl-dev \
+        # for MKL
+        #apt-transport-https \
+        #gpg-agent gnupg2 \
+        # ssh for OpenMPI
+        #openssh-server openssh-client \
+        ## gflags
+        #libgflags-dev libgflags2.2 \
+        ## for glog
+        #libgoogle-glog-dev libgoogle-glog0v5 \
+
+# Install boost 1.75
+RUN cd /tmp && \
+    curl -sSLo - https://dl.bintray.com/boostorg/release/1.75.0/source/boost_1_75_0.tar.bz2 | tar --bzip2 -x && \
+    cd boost_1_75_0 && \
+    ./bootstrap.sh --prefix=/opt/boost && \
+    ./b2 && \
+    ./b2 install
+ENV LD_LIBRARY_PATH="/opt/boost/lib:$LD_LIBRARY_PATH"
+
+# cmake 3.10
+RUN curl -sSLo - https://cmake.org/files/v3.10/cmake-3.10.3-Linux-x86_64.tar.gz | tar -xz -C /opt/cmake --strip-components 1
+ENV PATH="/opt/cmake/bin:$PATH"
+# ==================================================================
+# arrayfire with CPU backend https://github.com/arrayfire/arrayfire/wiki/Build-Instructions-for-Linux
+# ------------------------------------------------------------------
+FROM cpu_base_builder as cpu_arrayfire
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+        # OpenBLAS
+        libopenblas-dev liblapacke-dev \
+        # ATLAS
+        libatlas3-base libatlas-base-dev liblapacke-dev \
+        # FFTW
+        libfftw3-dev
+
+# Install ArrayFire
+RUN cd /tmp && \
+    git clone --branch v3.7.3 --depth 1 --recursive --shallow-submodules https://github.com/arrayfire/arrayfire.git && \
+    mkdir -p arrayfire/build && \
+    cd arrayfire/build && \
+    cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/opt/arrayfire -DAF_BUILD_CUDA=OFF -DAF_BUILD_CPU=ON -DAF_BUILD_OPENCL=OFF -DAF_BUILD_EXAMPLES=OFF -DBUILD_TESTING=OFF -DAF_WITH_IMAGEIO=OFF .. && \
+    make -j$(nproc) && \
+    make install -j$(nproc)
+
+# ==================================================================
+# oneDNN https://github.com/oneapi-src/oneDNN#requirements-for-building-from-source
+# ------------------------------------------------------------------
+FROM cpu_base_builder as cpu_onednn
+
+#Install OpenMP and TBB
+RUN apt-get update && apt-get install -y --no-install-recommends \
+        ocl-icd-opencl-dev libtbb-dev
+
+RUN cd /tmp && \
+    git clone --branch v2.0 --depth 1 https://github.com/oneapi-src/onednn.git && \
+    mkdir -p onednn/build && \
+    cd onednn/build && \
+    cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/opt/onednn -DWITH_EXAMPLE=OFF -DWITH_TEST=OFF .. && \
+    make -j$(nproc) && \
+    make install -j$(nproc)
+# ==================================================================
+# Gloo https://github.com/facebookincubator/gloo.git
+# ------------------------------------------------------------------
+FROM cpu_base_builder as cpu_gloo
+
+# Install OpenMPI
+RUN apt-get update && apt-get install -y --no-install-recommends \
+        libopenmpi-dev
+
+RUN cd /tmp && \
+    git clone --depth 1 https://github.com/facebookincubator/gloo.git && \
+    mkdir -p gloo/build && \
+    cd gloo/build && \
+    cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/opt/gloo -DUSE_MPI=ON .. && \
+    make -j$(nproc) && \
+    make install -j$(nproc)
+# ==================================================================
+# KenLM https://github.com/kpu/kenlm/blob/master/BUILDING
+# ------------------------------------------------------------------
+FROM cpu_base_builder as cpu_kenlm
+
+# Instal gzip, bz2 and xz
+RUN apt-get update && apt-get install -y --no-install-recommends \
+        zlib1g-dev libbz2-dev liblzma-dev
+
+RUN cd /tmp && \
+    git clone --depth 1 https://github.com/kpu/kenlm.git && \
+    mkdir -p kenlm/build && \
+    cd kenlm/build && \
+    cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/opt/kenlm -DCMAKE_POSITION_INDEPENDENT_CODE=ON .. && \
+    make -j$(nproc) && \
+    make install -j$(nproc)
+
+
+
+#############################################################################
+#                            SECOND STAGE                                   #
+#############################################################################
+
+FROM cpu_base_builder as final
+
+# install dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+        build-essential \
+        ca-certificates \
+        curl \
         git \
-        vim \
-        emacs \
-        nano \
-        htop \
         g++ \
+        # for cmake
+        zlib1g-dev libcurl4-openssl-dev \
         # for MKL
-        apt-transport-https \
-        gpg-agent gnupg2 \
+        apt-transport-https gpg-agent gnupg2 \
         # for arrayfire CPU backend
         # OpenBLAS
         libopenblas-dev libfftw3-dev liblapacke-dev \
@@ -44,108 +150,43 @@ RUN apt-get update && \
         openssh-server openssh-client \
         # OpenMPI
         libopenmpi-dev libomp-dev openmpi-bin \
-        # for libsndfile
-        autoconf automake autogen build-essential libasound2-dev \
-        libflac-dev libogg-dev libtool libvorbis-dev pkg-config python \
+        # libsndfile
+        libsndfile1-dev \
         # for libsndfile for ubuntu 18.04
         libopus-dev \
         # FFTW
         libfftw3-dev \
         # for kenlm
-        zlib1g-dev libbz2-dev liblzma-dev libboost-all-dev \
+        zlib1g-dev libbz2-dev liblzma-dev \
         # gflags
         libgflags-dev libgflags2.2 \
         # for glog
         libgoogle-glog-dev libgoogle-glog0v5 \
-        # for python sox
-        sox
-# ==================================================================
-# cmake 3.10
-# ------------------------------------------------------------------
-RUN apt-get purge -y cmake && \
-    # for cmake
-    DEBIAN_FRONTEND=noninteractive $APT_INSTALL zlib1g-dev libcurl4-openssl-dev && \
-    cd /tmp && wget https://cmake.org/files/v3.10/cmake-3.10.3.tar.gz  && \
-    tar -xzvf cmake-3.10.3.tar.gz  && cd cmake-3.10.3  && \
-    ./bootstrap --system-curl && \
-    make -j$(nproc) &&  make install && cmake --version
-# ==================================================================
-# MKL https://software.intel.com/en-us/mkl
-# ------------------------------------------------------------------
-RUN cd /tmp && wget https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS-2019.PUB && \
-    apt-key add GPG-PUB-KEY-INTEL-SW-PRODUCTS-2019.PUB && \
-# for build in March 2020 this row should be removed to prevent from the error
-# E: Failed to fetch https://apt.repos.intel.com/intelpython/binary/Packages  Writing more data than expected (15520 > 15023) E: Some index files failed to download. They have been ignored, or old ones used instead.
-#    wget https://apt.repos.intel.com/setup/intelproducts.list -O /etc/apt/sources.list.d/intelproducts.list && \
-    sh -c 'echo deb https://apt.repos.intel.com/mkl all main > /etc/apt/sources.list.d/intel-mkl.list' && \
-    apt-get update && DEBIAN_FRONTEND=noninteractive $APT_INSTALL intel-mkl-64bit-2018.4-057
-# ==================================================================
-# arrayfire with CPU backend https://github.com/arrayfire/arrayfire/wiki/
-# ------------------------------------------------------------------
-RUN cd /tmp && git clone --recursive https://github.com/arrayfire/arrayfire.git && \
-    wget https://dl.bintray.com/boostorg/release/1.70.0/source/boost_1_70_0.tar.gz && tar xf boost_1_70_0.tar.gz && \
-    cd arrayfire && git checkout v3.7.1  && git submodule update --init --recursive && \
-    mkdir build && cd build && \
-    CXXFLAGS=-DOS_LNX cmake .. -DCMAKE_BUILD_TYPE=Release -DAF_BUILD_CUDA=OFF -DAF_BUILD_OPENCL=OFF -DAF_BUILD_EXAMPLES=OFF -DBOOST_INCLUDEDIR=/tmp/boost_1_70_0 && \
-    make -j$(nproc) && make install
-# ==================================================================
-# oneDNN https://github.com/oneapi-src/oneDNN
-# ------------------------------------------------------------------
-RUN cd /tmp && git clone https://github.com/oneapi-src/oneDNN && \
-    cd oneDNN && git checkout v2.0 && mkdir -p build && cd build && \
-    cmake .. && \
-    make -j$(nproc) && make install
-# ==================================================================
-# Gloo https://github.com/facebookincubator/gloo.git
-# ------------------------------------------------------------------
-RUN cd /tmp && git clone https://github.com/facebookincubator/gloo.git && \
-    cd gloo && git checkout b7e0906 && mkdir build && cd build && \
-    cmake .. -DUSE_MPI=ON && \
-    make -j$(nproc) && make install
-# ==================================================================
-# python (for bindings)
-# ------------------------------------------------------------------
-RUN PIP_INSTALL="python3 -m pip --no-cache-dir install --upgrade" && \
-    DEBIAN_FRONTEND=noninteractive $APT_INSTALL \
-        software-properties-common \
-        && \
-    add-apt-repository ppa:deadsnakes/ppa && \
-    apt-get update && \
-    DEBIAN_FRONTEND=noninteractive $APT_INSTALL \
-        python3.6 \
-        python3.6-dev \
-        && \
-    wget -O ~/get-pip.py \
-        https://bootstrap.pypa.io/get-pip.py && \
-    python3.6 ~/get-pip.py && \
-    ln -s /usr/bin/python3.6 /usr/local/bin/python3 && \
-    ln -s /usr/bin/python3.6 /usr/local/bin/python && \
-    $PIP_INSTALL \
-        setuptools \
-        && \
-    $PIP_INSTALL \
-        numpy \
+        # gtest
+        libgtest-dev \
+        # for receipts data processing
         sox \
-        tqdm
-# ==================================================================
-# libsndfile https://github.com/erikd/libsndfile.git
-# ------------------------------------------------------------------
-RUN cd /tmp && git clone https://github.com/erikd/libsndfile.git && \
-    cd libsndfile && git checkout 4bdd7414602946a18799b514001b0570e8693a47 && \
-    ./autogen.sh && ./configure --enable-werror && \
-    make && make check && make install
+        # for python
+        python3-dev python3-pip python3-distutils && \
+    apt-get clean && \
+    apt-get -y autoremove && \
+    rm -rf /var/lib/apt/lists/*
+
 # ==================================================================
-# KenLM https://github.com/kpu/kenlm
+# python (for bindings)
 # ------------------------------------------------------------------
-RUN cd /tmp && git clone https://github.com/kpu/kenlm.git && \
-    cd kenlm && git checkout 4a277534fd33da323205e6ec256e8fd0ff6ee6fa && \
-    mkdir build && cd build && \
-    cmake .. -DCMAKE_POSITION_INDEPENDENT_CODE=ON && \
-    make -j$(nproc) && make install
+RUN ln -s /usr/bin/python3 /usr/local/bin/python && \
+    python3 -m pip --no-cache-dir install --upgrade setuptools numpy sox tqdm
+
 # ==================================================================
-# config & cleanup
+# Intel MKL https://software.intel.com/en-us/mkl
 # ------------------------------------------------------------------
-RUN ldconfig && \
-    apt-get clean && \
-    apt-get -y autoremove && \
-    rm -rf /var/lib/apt/lists/* /tmp/*
+RUN cd /tmp && curl -sSLo - https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS-2019.PUB | apt-key add && \
+    sh -c 'echo deb https://apt.repos.intel.com/mkl all main > /etc/apt/sources.list.d/intel-mkl.list' && \
+    apt-get update && apt-get install -y intel-mkl-64bit-2020.4-912
+ENV MKLROOT="/opt/intel/mkl"
+
+COPY --from=cpu_arrayfire /opt/arrayfire /opt/arrayfire
+COPY --from=cpu_kenlm     /opt/kenlm     /opt/kenlm
+COPY --from=cpu_gloo      /opt/gloo      /opt/gloo
+COPY --from=cpu_onednn    /opt/onednn    /opt/onednn