Add openpose-multi Singularity build definition file

openpose/CMakeLists.patch

@@ -0,0 +1,29 @@
+diff --git a/CMakeLists.txt b/CMakeLists.txt
+index 0e269073..d0d4265d 100644
+--- a/CMakeLists.txt
++++ b/CMakeLists.txt
+@@ -220,6 +220,8 @@ if (${GPU_MODE} MATCHES "OPENCL")
+   find_package(OpenCL)
+ endif (${GPU_MODE} MATCHES "OPENCL")
+
++find_package(Boost COMPONENTS system filesystem REQUIRED)
++
+ # Code to avoid crash at compiling time if OpenCL is not found
+ if (NOT OpenCL_FOUND)
+   set(OpenCL_LIBRARIES "")
+@@ -786,6 +788,7 @@ if (UNIX OR APPLE)
+           PREFIX ${CAFFE_PREFIX}
+           CMAKE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=<INSTALL_DIR>
+           -DCMAKE_TOOLCHAIN_FILE=${CMAKE_TOOLCHAIN_FILE}
++          -DCMAKE_CXX_FLAGS="-w"
+           -DMKLDNN_INSTALL_DIR:PATH=<INSTALL_DIR>
+           -DUSE_MKL2017_AS_DEFAULT_ENGINE=${CAFFE_CPU_ONLY}
+           -DUSE_CUDNN=${USE_CUDNN}
+@@ -807,6 +810,7 @@ if (UNIX OR APPLE)
+           PREFIX ${CAFFE_PREFIX}
+           CMAKE_ARGS -DCMAKE_INSTALL_PREFIX:PATH=<INSTALL_DIR>
+           -DCMAKE_TOOLCHAIN_FILE=${CMAKE_TOOLCHAIN_FILE}
++          -DCMAKE_CXX_FLAGS="-w"
+           -DUSE_CUDNN=${USE_CUDNN}
+           -DCUDA_ARCH_NAME=${CUDA_ARCH}
+           -DCUDA_ARCH_BIN=${CUDA_ARCH_BIN}

openpose/Singularity.openpose-multi

@@ -0,0 +1,145 @@
+Bootstrap: docker
+From: nvcr.io/nvidia/caffe:20.03-py3
+
+%files
+    CMakeLists.patch
+    cudacap.cxx
+
+%post
+    # Packages
+    export LC_ALL=C
+    apt-get -y --no-install-recommends update && \
+        apt-get -y --no-install-recommends upgrade
+
+    apt-get install -y --no-install-recommends \
+        build-essential \
+        git \
+        wget \
+        nano \
+        dialog \
+        software-properties-common \
+        libatlas-base-dev \
+        libleveldb-dev \
+        libsnappy-dev \
+        libhdf5-serial-dev \
+        libboost-all-dev \
+        libgflags-dev \
+        libgoogle-glog-dev \
+        liblmdb-dev \
+        pciutils \
+        python3-setuptools \
+        python3-dev \
+        python3-pip \
+        opencl-headers \
+        ocl-icd-opencl-dev \
+        libviennacl-dev \
+        libavcodec-dev \
+        libavformat-dev \
+        libswscale-dev \
+        libv4l-dev \
+        libxvidcore-dev \
+        libx264-dev \
+        libgtk-3-dev \
+        gfortran \
+        pkg-config \
+        libssl-dev \
+        libcanberra-gtk-module && \
+        python3 -m pip install \
+        numpy \
+        opencv-python
+
+    # Up to date ffmpeg
+    add-apt-repository -y ppa:jonathonf/ffmpeg-4 && \
+        apt-get -y --no-install-recommends update && \
+        apt-get -y --no-install-recommends install ffmpeg \
+
+    # Up to date CMake (Ubuntu 18.04 includes 3.10 which is too old)
+    wget -O - \
+        https://apt.kitware.com/keys/kitware-archive-latest.asc | \
+        gpg --dearmor - | \
+        tee /etc/apt/trusted.gpg.d/kitware.gpg && \
+        apt-add-repository -y 'deb https://apt.kitware.com/ubuntu/ bionic main' && \
+        apt-get update && \
+        apt-get -y --no-install-recommends install cmake
+
+    # Get OpenPose
+    cd /opt && \
+        git clone --depth 1 https://github.com/CMU-Perceptual-Computing-Lab/openpose.git --branch v1.6.0 --single-branch && \
+        cp -R openpose openpose_cpu && \
+        cp -R openpose openpose_legacy_gpu
+
+    # Fix up stubs
+    ln -s /usr/local/cuda/lib64/stubs/libnvidia-ml.so /usr/local/cuda/lib64/stubs/libnvidia-ml.so.1
+
+    # Build GPU + NVCaffe OpenPose
+    mkdir -p /opt/openpose/build && \
+    cd /opt/openpose/ && git apply /CMakeLists.patch
+    cd build && \
+        cmake \
+        -DGPU_MODE=CUDA \
+        -DBUILD_PYTHON=ON \
+        -DCMAKE_EXE_LINKER_FLAGS="-Wl,-rpath-link=/usr/local/cuda/lib64/stubs" \
+        -DDL_FRAMEWORK=NV_CAFFE \
+        -DCaffe_INCLUDE_DIRS=/usr/local/include/caffe \
+        -DCaffe_LIBS_RELEASE=/usr/local/lib/libcaffe-nv.so \
+        -DCaffe_LIBS=/usr/local/lib/libcaffe-nv.so \
+        -DBUILD_CAFFE=OFF \
+        -DCUDA_ARCH=All \
+        .. && \
+        make -j`nproc`
+
+    # Build CPU OpenPose
+    mkdir -p /opt/openpose_cpu/build && \
+        cd /opt/openpose_cpu/ && git apply /CMakeLists.patch &&
+        cd build &&
+        cmake \
+        -DGPU_MODE=CPU_ONLY \
+        -DBUILD_PYTHON=ON \
+        .. && \
+        make -j`nproc`
+
+    # Build legacy GPU OpenPose
+    mkdir -p /opt/openpose_legacy_gpu/build && \
+        cd /opt/openpose_legacy_gpu/ && git apply /CMakeLists.patch &&
+        cd build &&
+        cmake \
+        -DGPU_MODE=CUDA \
+        -DBUILD_PYTHON=ON \
+        -DCMAKE_EXE_LINKER_FLAGS="-Wl,-rpath-link=/usr/local/cuda/lib64/stubs" \
+        -DCUDA_ARCH=All \
+        .. && \
+        make -j`nproc`
+
+    # Build CUDA capability checker
+    g++ /cudacap.cxx -I/usr/local/cuda/include -o /cudacap /usr/local/cuda/lib64/libcudart.so
+
+
+%environment
+    if [ -z "$LD_LIBRARY_PATH" ]; then
+        LD_LIBRARY_PATH="/.singularity.d/libs"
+    else
+        LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/.singularity.d/libs"
+    fi
+
+    if nvidia-smi; then
+        if [ "$(/cudacap)" -ge "50" ]; then
+            echo "#### USING GPU with NVCaffe ####"
+            export OPENPOSE_SRC=/opt/openpose
+            export OPENPOSE_VAR=gpu
+        else
+            echo "#### USING Legacy GPU with Custom Caffe ####"
+            export OPENPOSE_SRC=/opt/openpose_legacy_gpu
+            export OPENPOSE_VAR=legacy_gpu
+        fi
+    else
+        echo "#### USING CPU ####"
+        export OPENPOSE_SRC=/opt/openpose_cpu
+        export OPENPOSE_VAR=cpu
+    fi
+    export OPENPOSE=$OPENPOSE_SRC/build
+    export OPENPOSE_BIN=$OPENPOSE/examples/openpose/openpose.bin
+    export PYTHONPATH="$OPENPOSE/python:$PYTHONPATH"
+
+
+%runscript
+    cd $OPENPOSE_SRC && exec $OPENPOSE_BIN "$@"

openpose/cudacap.cxx

@@ -0,0 +1,31 @@
+#include <cstdio>
+#include <cstdlib>
+#include <cuda_runtime_api.h>
+
+int main(int argc, char *argv[])
+{
+    cudaDeviceProp prop;
+    cudaError_t status;
+    int device_count;
+    int device_index = 0;
+    if (argc > 1) {
+        device_index = atoi(argv[1]);
+    }
+
+    status = cudaGetDeviceCount(&device_count);
+    if (status != cudaSuccess) {
+        fprintf(stderr,"cudaGetDeviceCount() failed: %s\n", cudaGetErrorString(status));
+        return -1;
+    }
+    if (device_index >= device_count) {
+        fprintf(stderr, "Specified device index %d exceeds the maximum (the device count on this system is %d)\n", device_index, device_count);
+        return -1;
+    }
+    status = cudaGetDeviceProperties(&prop, device_index);
+    if (status != cudaSuccess) {
+        fprintf(stderr,"cudaGetDeviceProperties() for device device_index failed: %s\n", cudaGetErrorString(status));
+        return -1;
+    }
+    int v = prop.major * 10 + prop.minor;
+    printf("%d\n", v);
+}