Commit bccdbe63 authored by Abraham, Subil's avatar Abraham, Subil
Browse files

trying to add mpi build of horovod

parent e1992ae2

pytorchimage-centos-cuda/Dockerfile

+5 −1
Original line number Diff line number Diff line
@@ -165,8 +165,12 @@ RUN cd ${INSTALL_ROOT} \ 
	&& rm -rf opencv



# Install horovod

RUN pip install horovod

#RUN pip install horovod



ENV PATH=$PATH:/usr/mpi/gcc/openmpi-4.0.3rc4/bin

ENV CPATH=$CPATH:/usr/mpi/gcc/openmpi-4.0.3rc4/include

ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/mpi/gcc/openmpi-4.0.3rc4/lib64

RUN HOROVOD_WITH_MPI=1 CXX=/usr/bin/g++ CC=/usr/bin/gcc pip install --no-cache-dir --no-binary horovod horovod



## Install DALI (not working)

#COPY ./cmake_patch /code

tensorflowimage-centos-cuda/Dockerfile

+5 −1
Original line number Diff line number Diff line
@@ -109,4 +109,8 @@ RUN cd /tmp/tensorflow && /usr/bin/bazel build --config=v2 //tensorflow/tools/pi 
RUN cd /tmp/tensorflow && ./bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg  

RUN pip install /tmp/tensorflow_pkg/tensorflow-2.5.0-cp39-cp39-linux_ppc64le.whl



RUN pip install horovod
 
#RUN pip install horovod

ENV PATH=$PATH:/usr/mpi/gcc/openmpi-4.0.3rc4/bin

ENV CPATH=$CPATH:/usr/mpi/gcc/openmpi-4.0.3rc4/include

ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/mpi/gcc/openmpi-4.0.3rc4/lib64

RUN HOROVOD_WITH_MPI=1 CXX=/usr/bin/g++ CC=/usr/bin/gcc pip install --no-cache-dir --no-binary horovod horovod