lintrunner -a

Author: chilo-ms

plugin_execution_providers/tensorrt/cuda/cu_inc/unary_elementwise_impl.cuh

@@ -35,7 +35,7 @@ __global__ void _UnaryElementWise(
   InT value[NumElementsPerThread];
 
   CUDA_LONG id = start;
-  #pragma unroll
+#pragma unroll
   for (int i = 0; i < NumElementsPerThread; i++) {
     if (id < N) {
       value[i] = input_data[id];
@@ -44,7 +44,7 @@ __global__ void _UnaryElementWise(
   }
 
   id = start;
-  #pragma unroll
+#pragma unroll
   for (int i = 0; i < NumElementsPerThread; i++) {
     if (id < N) {
       output_data[id] = functor(value[i]);

plugin_execution_providers/tensorrt/cuda/unary_elementwise_ops_impl.cu

@@ -9,7 +9,6 @@
 #endif
 #include <cuda_fp16.h>
 
-
 namespace cuda {
 
 // the postfix of means the types supported by the op:
@@ -72,7 +71,7 @@ struct OP_Cast {
   IMPL_CAST_IMPL(T, uint32_t)  \
   IMPL_CAST_IMPL(T, uint64_t)  \
   IMPL_CAST_IMPL(T, bool)      \
-  //IMPL_CAST_IMPL(T, BFloat16)
+  // IMPL_CAST_IMPL(T, BFloat16)
 
 IMPL_CAST_IMPL_FROM(half)
 IMPL_CAST_IMPL_FROM(float)
@@ -86,6 +85,6 @@ IMPL_CAST_IMPL_FROM(uint16_t)
 IMPL_CAST_IMPL_FROM(uint32_t)
 IMPL_CAST_IMPL_FROM(uint64_t)
 IMPL_CAST_IMPL_FROM(bool)
-//IMPL_CAST_IMPL_FROM(BFloat16)
+// IMPL_CAST_IMPL_FROM(BFloat16)
 
 }  // namespace cuda

plugin_execution_providers/tensorrt/cuda/unary_elementwise_ops_impl.h

@@ -27,7 +27,7 @@ namespace cuda {
   DECL_IMPL_CAST(T, uint32_t)  \
   DECL_IMPL_CAST(T, uint64_t)  \
   DECL_IMPL_CAST(T, bool)      \
-  //DECL_IMPL_CAST(T, BFloat16)
+  // DECL_IMPL_CAST(T, BFloat16)
 
 DECL_IMPL_CAST_FROM(half)
 DECL_IMPL_CAST_FROM(float)
@@ -41,7 +41,7 @@ DECL_IMPL_CAST_FROM(uint16_t)
 DECL_IMPL_CAST_FROM(uint32_t)
 DECL_IMPL_CAST_FROM(uint64_t)
 DECL_IMPL_CAST_FROM(bool)
-//DECL_IMPL_CAST_FROM(BFloat16)
+// DECL_IMPL_CAST_FROM(BFloat16)
 
 template <typename InT, typename OutT>
 void Impl_Cast(cudaStream_t stream, const InT* input_data, OutT* output_data, size_t count) {

plugin_execution_providers/tensorrt/cuda_allocator.h

@@ -15,7 +15,7 @@ struct CUDAAllocator : OrtAllocator {
     OrtAllocator::Info = [](const OrtAllocator* this_) { return static_cast<const CUDAAllocator*>(this_)->Info(); };
     OrtAllocator::Reserve = nullptr;
     OrtAllocator::GetStats = nullptr;
-    OrtAllocator::AllocOnStream = nullptr; // Allocate memory, handling usage across different Streams. Not used for TRT EP.
+    OrtAllocator::AllocOnStream = nullptr;  // Allocate memory, handling usage across different Streams. Not used for TRT EP.
   }
   // TODO: Handle destructor
   //~CUDAAllocator();

plugin_execution_providers/tensorrt/onnx_ctx_model_helper.cc

@@ -46,7 +46,6 @@ OrtStatus* EPContextNodeHelper::CreateEPContextNode(const std::string& engine_ca
                                                     const std::string& compute_capability,
                                                     const std::string& onnx_model_path,
                                                     OrtNode** ep_context_node) {
-
   // Helper to collect input or output names from an array of OrtValueInfo instances.
   auto collect_input_output_names = [&](gsl::span<const OrtValueInfo* const> value_infos,
                                         std::vector<const char*>& result) -> OrtStatus* {
@@ -99,16 +98,14 @@ OrtStatus* EPContextNodeHelper::CreateEPContextNode(const std::string& engine_ca
     RETURN_IF_ERROR(ort_api.CreateOpAttr("ep_cache_context", engine_cache_path.c_str(), engine_cache_path.size(), ORT_OP_ATTR_STRING, &attributes[1]));
   }
 
- 
   ort_api.CreateOpAttr("hardware_architecture", compute_capability.c_str(), compute_capability.size(), ORT_OP_ATTR_STRING, &attributes[2]);
   ort_api.CreateOpAttr("onnx_model_filename", std::filesystem::path(onnx_model_path).filename().string().c_str(), 1,
                        ORT_OP_ATTR_STRING, &attributes[3]);
 
-
   RETURN_IF_ERROR(model_editor_api.CreateNode("EPContext", "com.microsoft", fused_node_name, input_names.data(),
                                               input_names.size(), output_names.data(), output_names.size(),
                                               attributes.data(), attributes.size(), ep_context_node));
-  
+
   return nullptr;
 }
 
@@ -140,7 +137,7 @@ OrtStatus* EPContextNodeReader::GetEpContextFromGraph(const OrtGraph& graph) {
   const int64_t embed_mode = reinterpret_cast<const ONNX_NAMESPACE::AttributeProto*>(node_attr)->i();
 
   // Only make path checks if model not provided as byte buffer
-  //bool make_secure_path_checks = !GetModelPath(graph_viewer).empty();
+  // bool make_secure_path_checks = !GetModelPath(graph_viewer).empty();
   bool make_secure_path_checks = false;
 
   if (embed_mode) {
@@ -151,7 +148,7 @@ OrtStatus* EPContextNodeReader::GetEpContextFromGraph(const OrtGraph& graph) {
 
     *(trt_engine_) = std::unique_ptr<nvinfer1::ICudaEngine>(trt_runtime_->deserializeCudaEngine(const_cast<char*>(context_binary.c_str()),
                                                                                                 static_cast<size_t>(context_binary.length())));
-    //LOGS_DEFAULT(VERBOSE) << "[TensorRT EP] Read engine as binary data from \"ep_cache_context\" attribute of ep context node and deserialized it";
+    // LOGS_DEFAULT(VERBOSE) << "[TensorRT EP] Read engine as binary data from \"ep_cache_context\" attribute of ep context node and deserialized it";
     if (!(*trt_engine_)) {
       return ort_api.CreateStatus(ORT_EP_FAIL, "TensorRT EP could not deserialize engine from binary data");
     }

plugin_execution_providers/tensorrt/onnx_ctx_model_helper.h

@@ -5,7 +5,7 @@
 
 #include "tensorrt_execution_provider.h"
 #include "ep_utils.h"
-#include "nv_includes.h"
+// #include "nv_includes.h"
 
 #include <string>
 #include <filesystem>
@@ -21,14 +21,13 @@ class EPContextNodeHelper : public ApiPtrs {
 
   static bool GraphHasCtxNode(const OrtGraph* graph, const OrtApi& ort_api);
 
-  OrtStatus* CreateEPContextNode(const std::string& engine_cache_path, 
+  OrtStatus* CreateEPContextNode(const std::string& engine_cache_path,
                                  char* engine_data,
                                  size_t size,
                                  const int64_t embed_mode,
                                  const std::string& compute_capability,
                                  const std::string& onnx_model_path,
-                                 OrtNode** ep_context_node
-                                 ); 
+                                 OrtNode** ep_context_node);
 
  private:
   const OrtGraph* graph_ = nullptr;
@@ -63,7 +62,7 @@ class EPContextNodeReader : public ApiPtrs {
         detailed_build_log_(detailed_build_log) {
   }
 
-  //bool ValidateEPCtxNode(const OrtGraph& graph);
+  // bool ValidateEPCtxNode(const OrtGraph& graph);
 
   OrtStatus* GetEpContextFromGraph(const OrtGraph& graph);