using env variable for INFER and TRAIN mode

mindspore/ccsrc/pipeline/jit/pipeline_ge.cc

@@ -21,7 +21,6 @@
 #include <cstdlib>
 #include <algorithm>
 
-#include "utils/config_manager.h"
 #include "utils/hash_map.h"
 #include "debug/anf_ir_dump.h"
 #include "ir/tensor.h"
@@ -130,7 +129,11 @@ bool InitExecDatasetGe(const std::string &queue_name, int64_t size, int64_t batc
     return false;
   }
 
-  auto training = ConfigManager::GetInstance().training();
+  auto env_training = common::GetEnv("MS_GE_TRAIN");
+  bool training = false;
+  if (env_training == "1") {
+    training = true;
+  }
   if (training) {
     (void)setenv("GE_TRAIN", "1", 1);
   } else {
@@ -246,7 +249,6 @@ FuncGraphPtr BuildDFGraph(const std::map<std::string, ExecutorInfoPtr> &info, co
     MS_LOG(EXCEPTION) << "No phase in executor:" << GetPhasePrefix(phase);
   }
   FuncGraphPtr anf_graph = info.at(phase)->func_graph;
-  ConfigManager::GetInstance().set_training(anf_graph->has_flag("training"));
 #ifdef ENABLE_DUMP_IR
   if (MsContext::GetInstance()->get_param<bool>(MS_CTX_SAVE_GRAPHS_FLAG)) {
     draw::Draw("anf_graph.dot", anf_graph);  // for debug
@@ -259,7 +261,11 @@ FuncGraphPtr BuildDFGraph(const std::map<std::string, ExecutorInfoPtr> &info, co
     return nullptr;
   }
 
-  auto training = ConfigManager::GetInstance().training();
+  auto env_training = common::GetEnv("MS_GE_TRAIN");
+  bool training = false;
+  if (env_training == "1") {
+    training = true;
+  }
   if (training) {
     (void)setenv("GE_TRAIN", "1", 1);
   } else {

mindspore/ccsrc/transform/graph_ir/convert.h

@@ -19,6 +19,7 @@
 
 #define DRAW_GE_GRAPH
 
+#include <cstdlib>
 #include <memory>
 #include <map>
 #include <set>
@@ -56,7 +57,13 @@ class DfGraphConvertor {
   explicit DfGraphConvertor(const AnfGraphPtr &anf_graph) : anf_graph_(anf_graph) {
     MS_EXCEPTION_IF_NULL(anf_graph);
     df_graph_ = std::make_shared<DfGraph>(anf_graph_->ToString());
-    training_ = anf_graph->has_flag("training");
+    auto env_ge = mindspore::common::GetEnv("MS_ENABLE_GE");
+    auto env_training = mindspore::common::GetEnv("MS_GE_TRAIN");
+    if (env_ge == "1" && env_training == "1") {
+      training_ = true;
+    } else {
+      training_ = anf_graph->has_flag("training");
+    }
     distribute_ = anf_graph->has_flag("broadcast_flag");
     if (anf_graph->has_flag("broadcast_flag")) {
       ConfigManager::GetInstance().set_parallel_strategy(ParallelStrategy::DISTRIBUTION);

mindspore/ccsrc/utils/config_manager.h

@@ -117,10 +117,6 @@ class ConfigManager {
 
   void set_gpu_loopsink_size(const int64_t size) { gpu_loopsink_size_ = size; }
 
-  bool training() const { return training_; }
-
-  void set_training(const bool training) { training_ = training; }
-
  private:
   ConfigManager() = default;
   ~ConfigManager() = default;
@@ -134,7 +130,6 @@ class ConfigManager {
   std::map<std::string, int16_t> queue_info_map;
   std::string dataset_phase_{""};
   int64_t gpu_loopsink_size_{1};
-  bool training_{false};
 };
 
 }  // namespace mindspore

mindspore/ccsrc/utils/context/context_extends.cc

@@ -196,7 +196,8 @@ void GetGeOptions(const std::shared_ptr<MsContext> &ms_context_ptr, std::map<std
     (*ge_options)["ge.variableMemoryMaxSize"] = ms_context_ptr->get_param<std::string>(MS_CTX_VARIABLE_MEMORY_MAX_SIZE);
   }
 
-  if (ConfigManager::GetInstance().training()) {
+  auto training = common::GetEnv("MS_GE_TRAIN");
+  if (training == "1") {
     (*ge_options)["ge.graphRunMode"] = "1";
   }