Merge branch 'ep_context_with_external_ini' of https://github.com/mic…

…rosoft/onnxruntime into ep_context_with_external_ini
microsoft · Jan 27, 2025 · 8d1d650 · 8d1d650
2 parents 403ded3 + fc75131
commit 8d1d650
Show file tree

Hide file tree

Showing 3 changed files with 5 additions and 3 deletions.
diff --git a/include/onnxruntime/core/session/onnxruntime_session_options_config_keys.h b/include/onnxruntime/core/session/onnxruntime_session_options_config_keys.h
@@ -287,7 +287,7 @@ static const char* const kOrtSessionOptionShareEpContexts = "ep.share_ep_context
 static const char* const kOrtSessionOptionsEpContextModelExternalInitializersFileName =
     "ep.context_model_external_initializers_file_name";
 
-// For nodes fallback to CPU use this config to control the minimum size of the initializer 
+// For nodes fallback to CPU use this config to control the minimum size of the initializer
 // when externalizing it during serialization for EP context model
 static const char* const kOrtSessionOptionsEpContextModelExternalInitializersMinSizeInBytes =
     "ep.context_model_external_initializers_min_size_in_bytes";

diff --git a/onnxruntime/python/tools/symbolic_shape_infer.py b/onnxruntime/python/tools/symbolic_shape_infer.py
@@ -205,6 +205,7 @@ def __init__(self, int_max, auto_merge, guess_output_rank, verbose, prefix=""):
             "GemmFastGelu": self._infer_GemmFastGelu,
             "GemmFloat8": self._infer_GemmFloat8,
             "GroupNorm": self._infer_GroupNorm,
+            "GroupNormalization": self._infer_GroupNorm,
             "GroupQueryAttention": self._infer_GroupQueryAttention,
             "LayerNormalization": self._infer_LayerNormalization,
             "LongformerAttention": self._infer_LongformerAttention,
@@ -474,6 +475,7 @@ def _onnx_infer_single_node(self, node):
             "PythonOp",
             "MultiHeadAttention",
             "GroupNorm",
+            "GroupNormalization",
             "GroupQueryAttention",
             "SparseAttention",
             "SkipGroupNorm",

diff --git a/onnxruntime/test/providers/qnn/qnn_ep_context_test.cc b/onnxruntime/test/providers/qnn/qnn_ep_context_test.cc
@@ -50,7 +50,7 @@ static const std::string& GetNodeAttr(const Node& node, const std::string& attr_
 static GetTestModelFn BuildGraphWithQAndNonQ(bool single_ep_node = true) {
   return [single_ep_node](ModelTestBuilder& builder) {
     // Creat non-quantized FusedMatMul node1
-    std::vector<float> data(200*200, 1.0f);
+    std::vector<float> data(200 * 200, 1.0f);
     NodeArg* input1 = MakeTestInput(builder, TestInputDef<float>({200, 200}, false, data));
     NodeArg* add1_ini_input2 = MakeTestInput(builder, TestInputDef<float>({200, 200}, true, data));
 
@@ -220,7 +220,7 @@ void EpCtxCpuNodeWithExternalIniFileTestBody(bool expect_external_ini_file) {
   if (expect_external_ini_file) {
     // Set the threshold to a small size so FusedMatMul node with weights float[200, 200] will dump to external data file
     so.AddConfigEntry(kOrtSessionOptionsEpContextModelExternalInitializersMinSizeInBytes, "1024");
-  } // otherwise it will use default value 1024,000, so the initializer is in Onnx file, no external data file generated
+  }  // otherwise it will use default value 1024,000, so the initializer is in Onnx file, no external data file generated
 
   Ort::Session session(*ort_env, ToPathString(model_with_ext).c_str(), so);