From 70e7eaf1e863245caf06f9916fec15a81a85e44c Mon Sep 17 00:00:00 2001
From: Pranav Sharma <prs@microsoft.com>
Date: Thu, 17 Oct 2019 11:46:14 -0700
Subject: [PATCH] Update DML transformers with the new Graph API and re-enable
 DML in the GPU CI build. (#2147)

---
 .../core/providers/dml/GraphTransformers/bn_add_fusion.cc   | 4 ++--
 .../core/providers/dml/GraphTransformers/bn_mul_fusion.cc   | 6 +++---
 .../ci_build/github/azure-pipelines/win-gpu-ci-pipeline.yml | 3 ++-
 3 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/onnxruntime/core/providers/dml/GraphTransformers/bn_add_fusion.cc b/onnxruntime/core/providers/dml/GraphTransformers/bn_add_fusion.cc
index c879c4c186..5d458691a6 100644
--- a/onnxruntime/core/providers/dml/GraphTransformers/bn_add_fusion.cc
+++ b/onnxruntime/core/providers/dml/GraphTransformers/bn_add_fusion.cc
@@ -5,7 +5,7 @@
 
 #undef OPTIONAL
 #include "core/graph/graph_utils.h"
-#include "core/optimizer/initializer.h" 
+#include "core/optimizer/initializer.h"
 #include "core/optimizer/utils.h"
 #include "bn_add_fusion.h"
 
@@ -97,7 +97,7 @@ bool BatchNormalizationAddFusion::SatisfyCondition(const Graph& graph, const Nod
   const auto& next_node = *node.OutputNodesBegin();
   return !(!graph_utils::IsSupportedOptypeVersionAndDomain(next_node, "Add", {7}) ||
            next_node.GetExecutionProviderType() != node.GetExecutionProviderType() ||
-           next_node.GetInputEdgesCount() != 1 || graph.IsNodeOutputsInGraphOutputs(next_node));
+           next_node.GetInputEdgesCount() != 1 || !graph.GetNodeOutputsInGraphOutputs(next_node).empty());
 }
 
 }  // namespace onnxruntime
diff --git a/onnxruntime/core/providers/dml/GraphTransformers/bn_mul_fusion.cc b/onnxruntime/core/providers/dml/GraphTransformers/bn_mul_fusion.cc
index 808f66553c..4560cb9aad 100644
--- a/onnxruntime/core/providers/dml/GraphTransformers/bn_mul_fusion.cc
+++ b/onnxruntime/core/providers/dml/GraphTransformers/bn_mul_fusion.cc
@@ -5,7 +5,7 @@
 
 #undef OPTIONAL
 #include "core/graph/graph_utils.h"
-#include "core/optimizer/initializer.h" 
+#include "core/optimizer/initializer.h"
 #include "core/optimizer/utils.h"
 #include "bn_mul_fusion.h"
 
@@ -20,7 +20,7 @@ Status BatchNormalizationMulFusion::Apply(Graph& graph, Node& node, RewriteRuleE
   const auto& mul_inputs = mul_node.InputDefs();
 
   const ONNX_NAMESPACE::TensorProto* BatchNormalization_Scale_tensor_proto = nullptr;
-  if (!graph.GetInitializedTensor(BatchNormalization_inputs[1]->Name(), BatchNormalization_Scale_tensor_proto)){
+  if (!graph.GetInitializedTensor(BatchNormalization_inputs[1]->Name(), BatchNormalization_Scale_tensor_proto)) {
     return Status::OK();
   }
 
@@ -109,7 +109,7 @@ bool BatchNormalizationMulFusion::SatisfyCondition(const Graph& graph, const Nod
 
   const auto& next_node = *node.OutputNodesBegin();
   return !(!graph_utils::IsSupportedOptypeVersionAndDomain(next_node, "Mul", {7}) ||
-           next_node.GetInputEdgesCount() != 1 || graph.IsNodeOutputsInGraphOutputs(next_node) ||
+           next_node.GetInputEdgesCount() != 1 || !graph.GetNodeOutputsInGraphOutputs(next_node).empty() ||
            next_node.GetExecutionProviderType() != node.GetExecutionProviderType());
 }
 
diff --git a/tools/ci_build/github/azure-pipelines/win-gpu-ci-pipeline.yml b/tools/ci_build/github/azure-pipelines/win-gpu-ci-pipeline.yml
index dfe50a482a..b1dd980892 100644
--- a/tools/ci_build/github/azure-pipelines/win-gpu-ci-pipeline.yml
+++ b/tools/ci_build/github/azure-pipelines/win-gpu-ci-pipeline.yml
@@ -2,9 +2,10 @@ jobs:
 - template: templates/win-ci.yml
   parameters:
     AgentPool : 'Win-GPU-CUDA10'
+    AgentDemands: 'Has19H1WinSDK'
     DoDebugBuild: 'true'
     DoCompliance: 'false'
-    BuildCommand: '$(Build.SourcesDirectory)\tools\ci_build\build.py --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --cmake_path $(Build.BinariesDirectory)\cmake\bin\cmake.exe --ctest_path $(Build.BinariesDirectory)\cmake\bin\ctest.exe  --enable_pybind --use_openmp --use_mkldnn --build_shared_lib  --build_csharp --enable_onnx_tests --use_cuda --cuda_version=10.0 --cuda_home="C:\local\cuda_10.0.130_win10_trt6015dll" --cudnn_home="C:\local\cudnn-10.0-windows10-x64-v7.3.1.20\cuda" --gen_doc'
+    BuildCommand: '$(Build.SourcesDirectory)\tools\ci_build\build.py --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --cmake_path $(Build.BinariesDirectory)\cmake\bin\cmake.exe --ctest_path $(Build.BinariesDirectory)\cmake\bin\ctest.exe  --enable_pybind --use_openmp --use_mkldnn --use_dml --build_shared_lib  --build_csharp --enable_onnx_tests --use_cuda --cuda_version=10.0 --cuda_home="C:\local\cuda_10.0.130_win10_trt6015dll" --cudnn_home="C:\local\cudnn-10.0-windows10-x64-v7.3.1.20\cuda" --gen_doc'
     JobName: 'Windows_CI_GPU_Dev'
     DoNugetPack:  'false'
     NuPackScript : ''