ctcyang edited a comment on issue #10696: [MXNET-366]Extend MXNet Distributed 
Training by AllReduce
URL: https://github.com/apache/incubator-mxnet/pull/10696#issuecomment-401432358
 
 
   Thanks for your help @threeleafzerg I was able to build it with 
`USE_ALLREDUCE_DIST_KVSTORE = 1`. On AWS EC2 instances using the Deep Learning 
AMI, you need to do these additional steps:
   
   ```
   wget 
https://github.com/google/protobuf/releases/download/v3.5.1/protobuf-cpp-3.5.1.tar.gz
 && tar --no-same-owner -zxf protobuf-cpp-3.5.1.tar.gz
   cd protobuf-3.5.1 && export CFLAGS=-fPIC && export CXXFLAGS=-fPIC && 
./configure -prefix=/usr && sudo make -j16 && sudo make -j16 install
   conda remove protobuf
   conda remove libprotobuf
   rm -rf ~/anaconda3/bin/proto* && rm -rf ~/anaconda3/lib/libproto*
   
   sudo apt-get --assume-yes remove libprotobuf-dev
   sudo apt-get --assume-yes remove libprotobuf-lite9v5
   sudo apt-get --assume-yes remove libprotobuf9v5
   sudo apt-get --assume-yes remove libprotoc9v5
   
   sudo apt-get --assume-yes install mpich
   sudo apt-get --assume-yes remove libopenmpi-dev
   sudo apt-get --assume-yes remove libopenmpi1.10
   sudo apt-get --assume-yes remove mpi-default-bin
   sudo apt-get --assume-yes remove mpi-default-dev
   sudo apt-get --assume-yes purge openmpi-bin
   sudo apt-get --assume-yes remove openmpi-common
   
   sudo rm -rf /usr/local/mpi
   sudo ln -s /usr/lib/mpich/lib/libmpich.so.12 /usr/lib/mpich/lib/libmpi.so
   sudo ln -s /usr/lib/mpich/lib/libmpich.so.12 /usr/lib/mpich/lib/libmpi.so.12
   sudo ldconfig
   ```
   
   When finished, do `ldconfig -p` and verify no more occurrences of 
`libprotoc.so.9` and `libprotobuf.so.9` occur in the output. Due to ABI 
incompatibility between different protobuf versions--preinstalled version that 
comes with Deep Learning AMI (Ubuntu apt), preinstalled version that comes with 
Anaconda and version that gets auto-installed by the Makefile--you need to 
uninstall the 2 versions that come with apt and Anaconda.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
[email protected]


With regards,
Apache Git Services

Reply via email to