[GitHub] mbrookhart commented on a change in pull request #12502: [NGRAPH] MXNet - nGraph initial integration

[GitBox <gi...@apache.org>]

mbrookhart commented on a change in pull request #12502: [NGRAPH] MXNet - nGraph initial integration
URL: https://github.com/apache/incubator-mxnet/pull/12502#discussion_r249905980
 
 

 ##########
 File path: include/mxnet/ndarray.h
 ##########
 @@ -779,6 +783,12 @@ class NDArray {
   void UpdateMKLDNNMemDesc();
 #endif
 
+#if MXNET_USE_NGRAPH == 1
+  // create and return tensor_view with this ndarray mem
+  std::shared_ptr<ngraph::runtime::Tensor> &create_tensor(bool is_boolean,
+                                                          bool is_scalar);
 
 Review comment:
   Thanks, @zheng-da, we'll see what we can do. The reason we have this here is to prevent memory copies during the update step for non-CPU, non-Nvidia GPU hardware targets, since the update step (sgd/adam/etc), won't run through the subgraph operator. We'll think of other ways to do it. In the short term, for this CPU-focused PR, we can probably remove this feature for this PR. One possibility is the nGraph Context we talked about a few months ago.

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