reminisce edited a comment on issue #14954:  broadcast_to() behavior change 
after new 0D and uninitialized shape support added?
URL: 
https://github.com/apache/incubator-mxnet/issues/14954#issuecomment-493793725
 
 
   @DickJC123 Sorry for the late reply and thanks for the analysis and 
suggestion. I agree with what you proposed up there. The ndarrays in memory in 
the backend are indeed as what you described, i.e. -1 means unknown, 0 means 
known zero size, etc. The conversion to numpy shape in the NDArray's Save 
function was somehow missed to be implemented. Would you mind submitting a fix 
or if you are busy, I can take care of it? Thanks.
   
   Another question on side, since zero-size tensors cannot be digested by any 
operators in MXNet without numpy compatibility mode, what's the purpose of 
saving zero-size tensors in sockeye's tests?

----------------------------------------------------------------
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:
[email protected]


With regards,
Apache Git Services

Reply via email to