huajsj commented on a change in pull request #8497:
URL: https://github.com/apache/tvm/pull/8497#discussion_r677986161
##########
File path: src/runtime/graph_executor/graph_executor.cc
##########
@@ -389,6 +434,13 @@ void GraphExecutor::SetupOpExecs() {
input_dltensors_[eid].push_back(static_cast<DLTensor*>(op_args->arg_values[i].v_handle));
}
}
+ // check if op output is model output
+ if (output_node_id.count(nid) > 0) {
+ for (uint32_t index = inode.inputs.size();
+ index < inode.param.num_outputs + inode.param.num_inputs; ++index) {
+
output_dltensors_.push_back(static_cast<DLTensor*>(op_args->arg_values[index].v_handle));
Review comment:
"sort the output_node" seems like not existing logic, without this new
logic, is the said use case "arg_nodes": [3,1,2, 0]" would cause any UT fail?
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
