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 2019/09/06 20:52:41 UTC

[GitHub] [incubator-mxnet] reminisce commented on issue #16111: Incompatible data type for numpy ops?

reminisce commented on issue #16111: Incompatible data type for numpy ops?
URL: https://github.com/apache/incubator-mxnet/issues/16111#issuecomment-529010437
 
 
   Yeah, we are already aware of this and has been working on this. It requires us to re-write the kernel of those ops and that would become much less tedious and efforts after we use TVM op module developed by @yzhliu . One catch here is that we may not be able to provide backward compute function for int input types as the backward pass shares the same infer type function as the forward. So for those ops taking integer inputs, we give error message that the gradient calculation is not supported and ask users to use floats as inputs. What do you guys think?

----------------------------------------------------------------
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