mbaret commented on a change in pull request #6355:
URL: https://github.com/apache/incubator-tvm/pull/6355#discussion_r485936429
##########
File path: src/relay/backend/contrib/ethosn/codegen.cc
##########
@@ -304,6 +475,42 @@ EthosnError ConstructNetworkVisitor::MakeSplitLayer(const
Call& call, sl::Tensor
return EthosnError();
}
+EthosnError ConstructNetworkVisitor::MakeDepthToSpaceLayer(const Call& call,
+
sl::TensorAndId<sl::Operand>* out) {
+ DepthToSpaceParams params;
+ params.input_info = GetTensorInfo(tensor_table_, call);
+ if (auto err = EthosnAPI::DepthToSpace(call, ¶ms)) {
+ return err;
+ }
+
+ auto input = operand_table_[call->args[0]][0];
+
+ try {
+ *out = AddDepthToSpace(network_, *input, params.depth_info);
+ } catch (const sl::NotSupportedException& e) {
+ return EthosnError(e.what());
+ }
+ return EthosnError();
+}
+
+EthosnError ConstructNetworkVisitor::MakeReluLayer(const Call& call,
Review comment:
I think we won't reimplement this for now, but there are subtle
differences between these functions, mostly around the number and type of the
inputs/outputs + how quantization params should be handled.
----------------------------------------------------------------
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]
