larroy commented on a change in pull request #14613: [MXNET-978] Higher order
gradient support for some unary operators
URL: https://github.com/apache/incubator-mxnet/pull/14613#discussion_r287533229
##########
File path: src/operator/tensor/elemwise_binary_op_basic.cc
##########
@@ -233,7 +233,17 @@ The storage type of ``elemwise_mul`` output depends on
storage types of inputs
return
std::vector<ResourceRequest>{ResourceRequest::kTempSpace};
})
.add_alias("_mul").add_alias("_Mul")
-.set_attr<nnvm::FGradient>("FGradient", ElemwiseGradUseIn{"_backward_mul"});
+.set_attr<nnvm::FGradient>("FGradient",
+ [](const nnvm::NodePtr& n, const std::vector<nnvm::NodeEntry>& ograds) {
+ auto lhs_grad = MakeNode("elemwise_mul", n->attrs.name + "_backward_lhs",
+ {ograds[0], n->inputs[1]}, nullptr, &n);
+ auto rhs_grad = MakeNode("elemwise_mul", n->attrs.name + "_backward_rhs",
+ {ograds[0], n->inputs[0]}, nullptr, &n);
+ std::vector<nnvm::NodeEntry> ret;
+ ret.emplace_back(nnvm::NodeEntry{lhs_grad, 0, 0});
Review comment:
Can we simplify as per https://github.com/apache/incubator-mxnet/pull/14095 ?
ret.emplace_back(MakeNode(...));
----------------------------------------------------------------
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:
us...@infra.apache.org
With regards,
Apache Git Services
