eric-haibin-lin commented on a change in pull request #13599: fallback to dense
version for grad(reshape), grad(expand_dims)
URL: https://github.com/apache/incubator-mxnet/pull/13599#discussion_r242793381
##########
File path: src/operator/tensor/elemwise_unary_op_basic.cc
##########
@@ -236,6 +236,21 @@ NNVM_REGISTER_OP(_backward_copy)
return std::vector<bool>{true};
});
+NNVM_REGISTER_OP(_backward_reshape)
+.set_num_inputs(1)
+.set_num_outputs(1)
+.set_attr<nnvm::TIsBackward>("TIsBackward", true)
+.set_attr<nnvm::FInplaceOption>("FInplaceOption",
+ [](const NodeAttrs& attrs){
+ return std::vector<std::pair<int, int> >{{0,
0}};
+ })
+.set_attr<FInferStorageType>("FInferStorageType", ElemwiseStorageType<1, 1,
false, false, false>)
Review comment:
If an op only supports dense tensors and FCompute, FInferStorageType is not
needed
----------------------------------------------------------------
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:
[email protected]
With regards,
Apache Git Services
