You are viewing a plain text version of this content. The canonical link for it is here.

Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2019/12/23 22:36:54 UTC

[GitHub] [incubator-tvm] anijain2305 commented on issue #4508: Broadcasting is broken with LayoutTransform

anijain2305 commented on issue #4508: Broadcasting is broken with LayoutTransform
URL: https://github.com/apache/incubator-tvm/issues/4508#issuecomment-568597152
 
 
   https://github.com/apache/incubator-tvm/pull/4577
   
   This is not just expand dims problem. The input shape is (1,) - if we take it from C to NCHW8c. Then , we first need to `repeat` to make the input shape atleast (8,). 
   
   Instead, a simpler way is to prevent the insertion of expand dims and layout transforms altogether for scalars. Scalars can be easily broadcasted even if the second tensor is transformed in layouts. The above PR does just that.

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