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 2019/07/23 19:14:43 UTC

[GitHub] [incubator-mxnet] roywei edited a comment on issue #15640: Performance regression for MXNet 1.5.0

roywei edited a comment on issue #15640: Performance regression for MXNet 1.5.0
URL: https://github.com/apache/incubator-mxnet/issues/15640#issuecomment-514344026
 
 
   Hi @QueensGambit, thanks for your detailed reply.
   
   There was a lengthy discussion about operator performance regression here: https://github.com/apache/incubator-mxnet/issues/15429#issuecomment-508831150
   
   Summary is:
   1. The profiler code changed between 1.4.1 and 1.5.0, so it's not an apple to apple comparison. 
   2. Using `time` module, we didn't see significant regression on `batchnorm` and `broadcast add`, 1-3% at most and givent the variance between runs, we decided to go with the release.  Relu activation slow is expected for correctness.
   3. We can look into `convlution` you mentioned.
   
   Also if the `cp` `nodes` `depths` values are different, does it mean the time will be different?

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


With regards,
Apache Git Services