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 2018/06/15 18:54:29 UTC

[GitHub] eric-haibin-lin commented on issue #11206: Embedding(sparse_grad=True) + hybridization fallback issues

eric-haibin-lin commented on issue #11206: Embedding(sparse_grad=True) + hybridization fallback issues
URL: https://github.com/apache/incubator-mxnet/issues/11206#issuecomment-397712309
 
 
   The problem is that currently is output grad is not the immediate output of `_backward_CachedOp`, the storage types of `_backward_CachedOp` outputs are inferred as dense storage. This is because _backward_CachedOp didn't register `FInferStorage`, producing dense outputs by default. 
   
   In this example, the outputs of _backward_CachedOp are passed to `add_n` to produce row_sparse grad. 
   
   To solve this issue, we need to register `FInferStorage` for `_backward_CachedOp`, which performs subgraph storage type inference, and return the stype inference result to the caller. 

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