Fix issues:

1.Reuse realtuplegetitem op infer.
2.Get ms_role by env.
3.Fix acc error for optimizer_split case.

mindspore/ccsrc/backend/common/pass/insert_type_transform_op.cc

@@ -18,6 +18,7 @@
 
 #include <memory>
 #include <vector>
+#include <algorithm>
 #include "backend/common/session/anf_runtime_algorithm.h"
 #include "include/common/utils/utils.h"
 #include "include/common/utils/anfalgo.h"
@@ -385,34 +386,6 @@ void GenerateKernelObjectTypeForNewCNode(const CNodePtr &cnode, std::vector<Kern
                << ". Output object types: " << *output_obj_type;
 }
 
-void UpdateAbsForTupleGetItem(const CNodePtr &tuple_get_item_node) {
-  MS_EXCEPTION_IF_NULL(tuple_get_item_node);
-  if (!IsPrimitiveCNode(tuple_get_item_node, prim::kPrimTupleGetItem)) {
-    MS_LOG(EXCEPTION) << "Node should be TupleGetItem, but got " << tuple_get_item_node->fullname_with_scope() << ", "
-                      << tuple_get_item_node->DebugString();
-  }
-  auto tuple_input = common::AnfAlgo::GetInputNode(tuple_get_item_node, kIndex0);
-  MS_EXCEPTION_IF_NULL(tuple_input);
-  auto input_abs = tuple_input->abstract();
-  MS_EXCEPTION_IF_NULL(input_abs);
-  if (!input_abs->isa<abstract::AbstractSequence>()) {
-    MS_LOG(EXCEPTION) << "TupleGetItem's first input abstract should be Sequence, but got " << input_abs->ToString();
-  }
-
-  auto seq_abs = input_abs->cast<abstract::AbstractSequencePtr>();
-  MS_EXCEPTION_IF_NULL(seq_abs);
-  AbstractBasePtrList seq_element = seq_abs->elements();
-  // This method is used for TupleGetItem to RealTupleGetItem converting, the tuple elements must be scalar for now.
-  for (const auto &ele : seq_element) {
-    if (!ele->isa<abstract::AbstractScalar>() && !ele->isa<abstract::AbstractTensor>()) {
-      MS_LOG(EXCEPTION) << "Element of the tuple should be scalar or tensor, but got " << ele->ToString();
-    }
-  }
-
-  int64_t item_index = GetGetitemIndex(tuple_get_item_node);
-  tuple_get_item_node->set_abstract(seq_element[item_index]);
-}
-
 // A map of kernel object type pairs to processing functions.
 static std::map<ObjectTypePair, ProcessTypeTransformFunc> kTypePairToProcessFunc;
 
@@ -680,8 +653,10 @@ AnfNodePtrList InsertTypeTransformOp::ProcessTupleToTupleUnfold(const FuncGraphP
     kg->AddValueNodeToGraph(index_input->cast<ValueNodePtr>());
   }
 
-  // Need to update TupleGetItem abstract.
-  UpdateAbsForTupleGetItem(node);
+  auto abs = GenerateAbsByOpInfer(prim::kPrimRealTupleGetItem, {input, index_input});
+  MS_EXCEPTION_IF_NULL(abs);
+  MS_LOG(DEBUG) << "Abstract for RealTupleGetItem op is " << abs->ToString();
+  node->set_abstract(abs);
 
   // The primitive of user is changed.
   *new_prim = true;

mindspore/ccsrc/backend/common/pass/insert_type_transform_op.h

@@ -109,8 +109,6 @@ std::string GenerateOutputFormatForNewCNode(const CNodePtr &cnode);
 void GenerateKernelObjectTypeForNewCNode(const CNodePtr &cnode, std::vector<KernelObjectType> *input_obj_type,
                                          std::vector<KernelObjectType> *output_obj_type);
 
-void UpdateAbsForTupleGetItem(const CNodePtr &tuple_get_item_node);
-
 // After kernel selection phase, one kernel's acquired input type may not be the same as the actual input type(the input
 // node's output type). We need this pass to transform these types to valid types.
 class BACKEND_EXPORT InsertTypeTransformOp : public PatternProcessPass {

mindspore/python/mindspore/parallel/_ps_context.py

@@ -40,7 +40,7 @@ def _need_reset_device_target_for_ps(target):
     For Ascend backend, the card can't be occupied by multiple processes in distributed traning,
     so we need to reset the device target for some roles.
     '''
-    is_server = (_get_ps_context("ms_role") in ["MS_PSERVER", "MS_SERVER", "MS_SCHED"])
+    is_server = (os.getenv('MS_ROLE') in ["MS_PSERVER", "MS_SERVER", "MS_SCHED"])
     return is_server and target == "Ascend"
 
 

tests/st/place/optimizer_split/test_entry_optimizer_split.py

@@ -30,9 +30,10 @@ def test_split_ref_without_optim():
     )
     if return_code != 0:
         os.system(f"echo '\n**************** Worker Log ****************'")
-        os.system(f"grep -E 'ERROR|Error|error' -C 15 ./worker*/worker*.log")
+        os.system(f"grep -E 'ERROR|Error' -C 15 ./worker*/worker*.log")
+        os.system(f"cat ./worker*/worker*.log | grep acc")
         os.system(f"echo '\n**************** Scheduler Log ****************'")
-        os.system(f"grep -E 'ERROR|Error|error' -C 15 ./sched/sched.log")
+        os.system(f"grep -E 'ERROR|Error' -C 15 ./sched/sched.log")
     assert return_code == 0
 
 
@@ -50,7 +51,8 @@ def test_split_optim():
     )
     if return_code != 0:
         os.system(f"echo '\n**************** Worker Log ****************'")
-        os.system(f"grep -E 'ERROR|Error|error' -C 15 ./worker*/worker*.log")
+        os.system(f"grep -E 'ERROR|Error' -C 15 ./worker*/worker*.log")
+        os.system(f"cat ./worker*/worker*.log | grep acc")
         os.system(f"echo '\n**************** Scheduler Log ****************'")
-        os.system(f"grep -E 'ERROR|Error|error' -C 15 ./sched/sched.log")
+        os.system(f"grep -E 'ERROR|Error' -C 15 ./sched/sched.log")
     assert return_code == 0

tests/st/place/optimizer_split/test_optimizer_split.py

@@ -54,7 +54,7 @@ net_need_split_opt_map = {
 def run_dist():
     init()
     net = net_name_map.get(net_name)(dist=True)
-    opt = get_optimizer(net, net_need_split_opt_map.get(net_name))
+    opt = get_optimizer(net, dist=net_need_split_opt_map.get(net_name))
     criterion = get_loss()
     model = Model(net, criterion, opt, metrics={"Accuracy": Accuracy()})