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 2018/05/25 14:27:25 UTC

[GitHub] asitstands commented on issue #11031: Use dtype=int for the indices returned by TopK

asitstands commented on issue #11031: Use dtype=int for the indices returned by TopK
URL: https://github.com/apache/incubator-mxnet/issues/11031#issuecomment-392075252
 
 
   This could break existing code. Sometimes the resulting array of indices is subject to another operation. Then the result can become different if the dtype of the index array changes. Or it can make an error as some opertions can be applied to only floating point arrays.
   ```python
   nd.topk(nd.arange(10), k=2).mean()
   ```
   Currently the result is `[8.5]`, but the result is `[8]` if the dtype is integer. 
   ```python
   nd.gemm2(nd.topk(nd.arange(10), k=2), x)
   ```
   Currently this is allowed, but it'll cause an error if the dtype changes to integer.
   
   `multinomial` has a similar but opposite issue. It returns an index array of integer type, but it would be convenient to have floating point array sometimes (#10970). I think that it would be better if we can make `multinomial` and `topk` consistent.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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