enable/disable allreduce_fusion

mindspore/ccsrc/parallel/allreduce_fusion/step_allreduce_fusion.cc

@@ -31,10 +31,11 @@ bool StepAllreduceFusion(const FuncGraphPtr &root, const opt::OptimizerPtr &opti
   MS_EXCEPTION_IF_NULL(optimizer);
   MS_EXCEPTION_IF_NULL(ParallelContext::GetInstance());
   std::string parallel_mode = ParallelContext::GetInstance()->parallel_mode();
+  bool enable_all_reduce_fusion = ParallelContext::GetInstance()->enable_all_reduce_fusion();
   // assume no change to graph
   bool changes = false;
   // control whether use model_parallel mode
-  if (((parallel_mode != AUTO_PARALLEL) && (parallel_mode != SEMI_AUTO_PARALLEL)) ||
+  if (((parallel_mode != AUTO_PARALLEL) && (parallel_mode != SEMI_AUTO_PARALLEL)) || (!enable_all_reduce_fusion) ||
       (root->has_flag(ALLREDUCE_FUSION_RUN_ONCE_ONLY))) {
     return changes;
   }

mindspore/ccsrc/parallel/context.cc

@@ -55,6 +55,7 @@ void ParallelContext::Reset() {
   parallel_mode_ = STAND_ALONE;
   parameter_broadcast_ = false;
   parameter_broadcast_is_set_ = false;
+  enable_all_reduce_fusion_ = false;
 }
 
 void ParallelContext::set_device_num(int32_t device_num) {

mindspore/ccsrc/parallel/context.h

@@ -80,6 +80,10 @@ class ParallelContext {
   const std::vector<uint32_t> all_reduce_fusion_split_indices() const;
   void set_all_reduce_fusion_split_sizes(const std::vector<uint32_t> sizes);
   const std::vector<uint32_t> all_reduce_fusion_split_sizes() const;
+  void set_enable_all_reduce_fusion(bool enable_all_reduce_fusion) {
+    enable_all_reduce_fusion_ = enable_all_reduce_fusion;
+  }
+  bool enable_all_reduce_fusion() const { return enable_all_reduce_fusion_; }
 
   void Reset();
 
@@ -98,6 +102,7 @@ class ParallelContext {
   bool device_num_is_set_;
   bool global_rank_is_set_;
   bool parameter_broadcast_is_set_;
+  bool enable_all_reduce_fusion_;
   std::vector<uint32_t> all_reduce_fusion_split_indices_;
   std::vector<uint32_t> all_reduce_fusion_split_sizes_;
 };

mindspore/ccsrc/pipeline/init.cc

@@ -183,6 +183,10 @@ PYBIND11_MODULE(_c_expression, m) {
          "Set all reduce fusion split sizes.")
     .def("get_all_reduce_fusion_split_sizes", &ParallelContext::all_reduce_fusion_split_sizes,
          "Get all reduce fusion split sizes.")
+    .def("set_enable_all_reduce_fusion", &ParallelContext::set_enable_all_reduce_fusion,
+         "Set enable/disable all reduce fusion.")
+    .def("get_enable_all_reduce_fusion", &ParallelContext::enable_all_reduce_fusion,
+         "Get enable/disable all reduce fusion.")
     .def("get_parameter_broadcast", &ParallelContext::parameter_broadcast, "Get parameter broadcast.")
     .def("get_parameter_broadcast_is_set", &ParallelContext::parameter_broadcast_is_set,
          "Get parameter broadcast is set.")

mindspore/parallel/_auto_parallel_context.py

@@ -259,6 +259,23 @@ class _AutoParallelContext:
         self.check_context_handle()
         return self._context_handle.get_all_reduce_fusion_split_sizes()
 
+    def set_enable_all_reduce_fusion(self, enable_all_reduce_fusion):
+        """
+        Set enable/disable all reduce fusion.
+
+        Args:
+            enable_all_reduce_fusion (bool): Enable/disable all reduce fusion.
+        """
+        self.check_context_handle()
+        if not isinstance(enable_all_reduce_fusion, bool):
+            raise TypeError('enable_all_reduce_fusion is invalid type')
+        self._context_handle.set_enable_all_reduce_fusion(enable_all_reduce_fusion)
+
+    def get_enable_all_reduce_fusion(self):
+        """Get all reduce fusion flag."""
+        self.check_context_handle()
+        return self._context_handle.get_enable_all_reduce_fusion()
+
     def get_device_num_is_set(self):
         """Get device number is set or not."""
         self.check_context_handle()

mindspore/parallel/_utils.py

@@ -117,6 +117,7 @@ _cast_before_mirror = None
 _loss_repeated_mean = None
 _communication_backend = None
 _has_checkpointed = False
+_enable_all_reduce_fusion = None
 
 
 def _checkpoint_auto_parallel_context():
@@ -133,6 +134,7 @@ def _checkpoint_auto_parallel_context():
     global _cast_before_mirror
     global _loss_repeated_mean
     global _communication_backend
+    global _enable_all_reduce_fusion
     _parallel_mode = auto_parallel_context().get_parallel_mode()
     _device_num = _get_device_num()
     _global_rank = _get_global_rank()
@@ -141,6 +143,7 @@ def _checkpoint_auto_parallel_context():
     _cast_before_mirror = auto_parallel_context().get_cast_before_mirror()
     _loss_repeated_mean = auto_parallel_context().get_loss_repeated_mean()
     _communication_backend = auto_parallel_context().get_communication_backend()
+    _enable_all_reduce_fusion = auto_parallel_context().get_enable_all_reduce_fusion()
     _has_checkpointed = True
 
 
@@ -154,10 +157,12 @@ def _restore_auto_parallel_context():
     global _cast_before_mirror
     global _loss_repeated_mean
     global _communication_backend
+    global _enable_all_reduce_fusion
     _set_auto_parallel_context(parallel_mode=_parallel_mode, device_num=_device_num, global_rank=_global_rank,
                                parameter_broadcast=_parameter_broadcast, mirror_mean=_mirror_mean,
                                cast_before_mirror=_cast_before_mirror, loss_repeated_mean=_loss_repeated_mean)
     auto_parallel_context().set_communication_backend(_communication_backend)
+    auto_parallel_context().set_enable_all_reduce_fusion(_enable_all_reduce_fusion)
 
 
 def _reset_checkpoint_auto_parallel_context():

tests/ut/python/parallel/__init__.py

@@ -13,10 +13,12 @@
 # limitations under the License.
 
 import mindspore.context as context
+from mindspore.parallel._auto_parallel_context import auto_parallel_context
 from mindspore.parallel._utils import _reset_op_id
 
 
 def setup_module(module):
+    auto_parallel_context().set_enable_all_reduce_fusion(enable_all_reduce_fusion=True)
     context.set_context(mode=context.GRAPH_MODE, device_target="Ascend", save_graphs=False)
     _reset_op_id()
 

tests/ut/python/parallel/test_allreduce_fusion.py

@@ -23,7 +23,7 @@ from tests.dataset_mock import MindData
 from mindspore import context
 from mindspore.common.api import _executor
 from mindspore.parallel import _cost_model_context as cost_model_context
-
+from mindspore.parallel._auto_parallel_context import auto_parallel_context
 
 
 class Dataset(MindData):
@@ -105,6 +105,7 @@ def train_common(net):
     epoch_size = 2
     device_num=4
     context.reset_auto_parallel_context()
+    auto_parallel_context().set_enable_all_reduce_fusion(enable_all_reduce_fusion=True)
     context.set_auto_parallel_context(parallel_mode=ParallelMode.SEMI_AUTO_PARALLEL, device_num=device_num, parameter_broadcast=False)
     context.set_context(mode=context.GRAPH_MODE)