!18950 adapt for multi-frontend cj

Merge pull request !18950 from qingshanxiaozi/adapt_for_cj

mindspore/ccsrc/backend/optimizer/ascend/mindir/sparse_softmax_cross_entropy_with_logits_unify_mindir.cc

@@ -419,8 +419,9 @@ const BaseRef GradSparseSoftmaxCrossEntropyWithLogitsUnifyMindIR::DefinePattern(
   VarPtr x1 = std::make_shared<Var>();
   VarPtr x2 = std::make_shared<Var>();
   VarPtr x3 = std::make_shared<Var>();
+  VarPtr x4 = std::make_shared<Var>();
   VectorRef sparse_softmax_cross_entropy_with_logits_grad({prim::kPrimSparseSoftmaxCrossEntropyWithLogits, x1, x2});
-  VectorRef sparse_softmax_cross_entropy_with_logits({prim::kPrimSparseSoftmaxCrossEntropyWithLogits, x1, x2});
+  VectorRef sparse_softmax_cross_entropy_with_logits({prim::kPrimSparseSoftmaxCrossEntropyWithLogits, x1, x4});
   VectorRef depend(
     {prim::kPrimDepend, sparse_softmax_cross_entropy_with_logits_grad, sparse_softmax_cross_entropy_with_logits});
   return VectorRef({prim::kPrimMul, depend, x3});
@@ -518,19 +519,21 @@ const AnfNodePtr PynativeSparseSoftmaxCrossEntropyWithLogitsUnifyMindIR::Process
   MS_EXCEPTION_IF_NULL(sparse_softmax_node);
   CheckCNodeInputSize(sparse_softmax_node, kSparseSoftmaxCrossEntropyWithLogitsInputTensorNum);
 
-  if (AnfAlgo::HasNodeAttr(kAttrIsGrad, sparse_softmax_node) &&
-      AnfAlgo::GetNodeAttr<bool>(sparse_softmax_node, kAttrIsGrad)) {
-    return nullptr;
-  }
-
   CNodePtr softmax_node;
   auto one_hot_node = CreateOneHot(graph, sparse_softmax_node, true);
   softmax_node = CreateSoftmaxCrossEntropyWithLogits(graph, sparse_softmax_node, one_hot_node);
 
   std::vector<AnfNodePtr> softmax_node_outputs;
   CreateMultipleOutputsOfAnfNode(graph, softmax_node, kSoftmaxCrossEntropyWithLogitsOutputNum, &softmax_node_outputs);
-  auto reduce_node = CreateReduceMean(graph, sparse_softmax_node, softmax_node_outputs[0], true);
+  // Both of the forward loss function and the backward loss function from cangjie will match this pattern,
-  return reduce_node;
+  // the true branch is for the backward loss function, and the false branch is for the other one.
+  if (AnfAlgo::HasNodeAttr(kAttrIsGrad, sparse_softmax_node) &&
+      AnfAlgo::GetNodeAttr<bool>(sparse_softmax_node, kAttrIsGrad)) {
+    return softmax_node_outputs[1];
+  } else {
+    auto reduce_node = CreateReduceMean(graph, sparse_softmax_node, softmax_node_outputs[0], true);
+    return reduce_node;
+  }
 }
 
 const BaseRef PynativeGradSparseSoftmaxCrossEntropyWithLogitsUnifyMindIR::DefinePattern() const {

mindspore/ccsrc/backend/session/ascend_session.cc

@@ -294,8 +294,10 @@ void AscendSession::UnifyMindIR(const KernelGraphPtr &graph) {
     unify_mindir_pm->AddPass(std::make_shared<opt::GradSparseSoftmaxCrossEntropyWithLogitsUnifyMindIRV2>());
     unify_mindir_pm->AddPass(std::make_shared<opt::SparseSoftmaxCrossEntropyWithLogitsUnifyMindIR>());
   } else {
-    unify_mindir_pm->AddPass(std::make_shared<opt::PynativeSparseSoftmaxCrossEntropyWithLogitsUnifyMindIR>());
+    // Add PynativeGradSparseSoftmaxCrossEntropyWithLogitsUnifyMindIR pass first to avoid the backward loss function
+    // from the python frontend matching the pattern defined in PynativeSparseSoftmaxCrossEntropyWithLogitsUnifyMindIR.
     unify_mindir_pm->AddPass(std::make_shared<opt::PynativeGradSparseSoftmaxCrossEntropyWithLogitsUnifyMindIR>());
+    unify_mindir_pm->AddPass(std::make_shared<opt::PynativeSparseSoftmaxCrossEntropyWithLogitsUnifyMindIR>());
   }
   unify_mindir_pm->AddPass(std::make_shared<opt::DropoutUnifyMindIR1>());
   unify_mindir_pm->AddPass(std::make_shared<opt::DropoutGradUnifyMindIR>());

mindspore/ccsrc/backend/session/session_basic.h

@@ -300,9 +300,6 @@ class SessionBasic : public std::enable_shared_from_this<SessionBasic> {
 #if !defined(_WIN32) && !defined(_WIN64)
   std::shared_ptr<Debugger> debugger_;
 #endif
-#if (ENABLE_CPU && !_WIN32)
-  bool initialized_ps_cache_{false};
-#endif
 };
 
 using SessionPtr = std::shared_ptr<session::SessionBasic>;

mindspore/ccsrc/pipeline/jit/CMakeLists.txt

@@ -26,4 +26,9 @@ if(ENABLE_GE OR ENABLE_D)
     list(APPEND _PIPELINE_SRC_FILES ${_PIPELINE_GE_SRC_FILES})
 endif()
 
+if("${ENABLE_HIDDEN}" STREQUAL "OFF")
+    string(REPLACE " -Werror " " " CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
+    string(REPLACE " -fvisibility=hidden" " -fvisibility=default" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
+endif()
+
 add_library(_mindspore_pipeline_jit_obj OBJECT ${_PIPELINE_SRC_FILES})

mindspore/ccsrc/runtime/device/kernel_runtime_manager.cc

@@ -51,6 +51,11 @@ void KernelRuntimeManager::ClearGraphResource(uint32_t graph_id, const std::vect
   }
 }
 
+KernelRuntimeManager &KernelRuntimeManager::Instance() {
+  static KernelRuntimeManager instance{};
+  return instance;
+}
+
 void KernelRuntimeManager::Register(const std::string &device_name, KernelRuntimeCreator &&runtime_creator) {
   if (runtime_creators_.find(device_name) == runtime_creators_.end()) {
     (void)runtime_creators_.emplace(device_name, runtime_creator);

mindspore/ccsrc/runtime/device/kernel_runtime_manager.h

@@ -32,10 +32,7 @@ using KernelRuntimeCreator = std::function<std::shared_ptr<KernelRuntime>()>;
 
 class KernelRuntimeManager {
  public:
-  static KernelRuntimeManager &Instance() {
+  static KernelRuntimeManager &Instance();
-    static KernelRuntimeManager instance;
-    return instance;
-  }
   void Register(const std::string &device_name, KernelRuntimeCreator &&runtime_creator);
   KernelRuntime *GetKernelRuntime(const std::string &device_name, uint32_t device_id);
   KernelRuntime *GetCurrentKernelRuntime();

mindspore/core/utils/ms_utils.cc

@@ -17,9 +17,11 @@
 
 namespace mindspore {
 namespace common {
+namespace {
 const int CACHED_STR_NUM = 1 << 8;
 const int CACHED_STR_MASK = CACHED_STR_NUM - 1;
 std::vector<std::string> STR_HOLDER(CACHED_STR_NUM);
+}  // namespace
 const char *SafeCStr(const std::string &&str) {
   static std::atomic<uint32_t> index{0};
   uint32_t cur_index = index++;