roywei opened a new issue #15211: MXNet built from source does not work with horovod URL: https://github.com/apache/incubator-mxnet/issues/15211 From https://lists.apache.org/thread.html/6b498b466123b11df023f746fb233117bc58a68818383b0952a9c9e2@%3Cdev.mxnet.apache.org%3E I was able to verify and reproduce the error. using MXNet built from source is not able to install Horovod Steps to reproduce: Using AWS Base DLAMI on P3.2xLarge. checkout tag 1.5.0.rc0 ``` sudo make -j$(nproc) USE_CUDA=1 USE_CUDA_PATH=/usr/local/cuda USE_CUDNN=1 USE_NCCL=1 USE_BLAS=openblas ``` ``` cd python pip install -e . --user ``` Able to build successfully and run a simple mnist example. ``` HOROVOD_GPU_ALLREDUCE=NCCL pip install horovod --user --no-cache-dir -U ``` error ``` home/ubuntu/src/incubator-mxnet/python/mxnet/../../include/mkldnn/mkldnn.h:55:28: fatal error: mkldnn_version.h: No such file or directory compilation terminated. INFO: Unable to build MXNet plugin, will skip it. ```
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to 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
