eric-haibin-lin commented on a change in pull request #17270: [WIP] Dynamic
custom operator GPU support
URL: https://github.com/apache/incubator-mxnet/pull/17270#discussion_r367199659
##########
File path: src/c_api/c_api.cc
##########
@@ -720,8 +751,11 @@ int MXLoadLib(const char *path) {
gradOp.set_attr<bool>("TIsLayerOpBackward", true, plevel);
gradOp.set_attr<FStatefulComputeEx>("FStatefulComputeEx<cpu>",
fstateful_backward, plevel);
+ gradOp.set_attr<FStatefulComputeEx>("FStatefulComputeEx<gpu>",
Review comment:
Are you in favor of approach 1 or approach 2? They have different indication
for the recommended nnvm_registration API for custom ops
----------------------------------------------------------------
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
