You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@mxnet.apache.org by GitBox <gi...@apache.org> on 2018/03/31 17:48:40 UTC

[GitHub] ArmageddonKnight opened a new issue #10356: MXNet from Source is slower than Pip Package

ArmageddonKnight opened a new issue #10356: MXNet from Source is slower than Pip Package
URL: https://github.com/apache/incubator-mxnet/issues/10356
 
 
   Hi there,
   
   I have two copies of `MXNet` - one local copy that I kept for revision and another global copy which has been installed using `sudo -H pip3 install mxnet-cu80==0.12.1`. I built my local copy using the following command:
   
   ```
   make -j $(nproc) \
   	USE_OPENCV=1 USE_BLAS=openblas USE_CUDA=1 \
   	USE_CUDA_PATH=/usr/local/cuda USE_CUDNN=1 USE_CPP_PACKAGE=1
   ```
   
   and created a virtual environment for the python binding. The problem is that - when I run the `cudnn_lstm_bucketing.py` example, what I noticed is that the local copy (throughput is around `3600 samples/sec`) is around `1000 samples/sec` slower than the global copy (throughput is around `4600 samples/sec`).
   
   Could someone please tell me whether I am missing some build options in my `make` command? Thank you.

----------------------------------------------------------------
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:
users@infra.apache.org


With regards,
Apache Git Services