hankaj opened a new pull request, #21122:
URL: https://github.com/apache/incubator-mxnet/pull/21122

   ## Description ##
   @anko-intel discovered very bad accuracy on fused floating point model in 
Resnet50v2. I changed fuse_relu flag usage to post-op implementation. Another 
problem was that batch norm with relu node in fused graph did not have batch 
norm attributes set properly.
   Now the accuracy seems to be similar to the original model.
   
   I also performed some refactoring so that batch norm operator would have 
more accurate name and deleted unused operator


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to