diff --git a/python/paddle/fluid/optimizer.py b/python/paddle/fluid/optimizer.py
index a674b13bde061c253ea0281f4af0c801cf593206..ffb54861904f4a7c6d54ba958343f9cb00cb559b 100644
--- a/python/paddle/fluid/optimizer.py
+++ b/python/paddle/fluid/optimizer.py
@@ -4046,7 +4046,7 @@ class PipelineOptimizer(object):
         """
         prev_op = []
         for op in ops:
-            if op.type == 'send_v2' or op.type == 'recv_v2':
+            if op.type == 'send_v2' or op.type == 'recv_v2' or op.type == 'c_broadcast':
                 continue
             if op == cur_op:
                 break
@@ -4434,6 +4434,39 @@ class PipelineOptimizer(object):
                         ring_id = self.ring_id + 2 + prev_device_index - cur_device_index - 1
                     if pair not in self._pipeline_pair:
                         self._pipeline_pair.append(pair)
+                    if self.schedule_mode == 0:  # GPipe
+                        block._insert_op(
+                            index=index + extra_index,
+                            type='send_v2',
+                            inputs={'X': var},
+                            attrs={
+                                self._op_device_key: prev_device,
+                                self._op_role_key: op_role,
+                                'use_calc_stream': True,
+                                'peer': cur_device_index,
+                                'ring_id': self.ring_id
+                                if cur_device_index > prev_device_index else
+                                self.ring_id + 2,
+                            })
+                        extra_index += 1
+                        block._insert_op(
+                            index=index + extra_index,
+                            type='recv_v2',
+                            outputs={'Out': [var]},
+                            attrs={
+                                'out_shape': var.shape,
+                                'dtype': var.dtype,
+                                self._op_device_key: cur_device,
+                                self._op_role_key: op_role,
+                                'use_calc_stream': True,
+                                'peer': prev_device_index,
+                                'ring_id': self.ring_id
+                                if cur_device_index > prev_device_index else
+                                self.ring_id + 2,
+                            })
+                        extra_index += 1
+                        continue
+                    assert self.schedule_mode == 1
                     block._insert_op(
                         index=index + extra_index,
                         #type='send_v2',
@@ -4452,19 +4485,19 @@ class PipelineOptimizer(object):
                             'root': 0,
                         })
                     extra_index += 1
-                    block._insert_op(
-                        index=index + extra_index,
-                        type='c_sync_comm_stream',
-                        inputs={'X': [var]},
-                        outputs={'Out': [var]},
-                        attrs={
-                            self._op_device_key: cur_device,
-                            self._op_role_key:
-                            core.op_proto_and_checker_maker.OpRole.Backward,
-                            'ring_id': self.ring_id,
-                            #'ring_id': self.ring_id if prev_device_index > cur_device_index else self.ring_id + 2,
-                        })
-                    extra_index += 1
+                    #block._insert_op(
+                    #    index=index + extra_index,
+                    #    type='c_sync_comm_stream',
+                    #    inputs={'X': [var]},
+                    #    outputs={'Out': [var]},
+                    #    attrs={
+                    #        self._op_device_key: cur_device,
+                    #        self._op_role_key:
+                    #        core.op_proto_and_checker_maker.OpRole.Backward,
+                    #        'ring_id': self.ring_id,
+                    #        #'ring_id': self.ring_id if prev_device_index > cur_device_index else self.ring_id + 2,
+                    #    })
+                    #extra_index += 1
                     fill_shape = list(var.shape)
                     fill_shape[0] = 1
                     block._insert_op(
@@ -4509,6 +4542,7 @@ class PipelineOptimizer(object):
                         outputs={'Out': [var]},
                         attrs={
                             self._op_device_key: cur_device,
+                            #self._op_role_key: core.op_proto_and_checker_maker.OpRole.Backward,
                             self._op_role_key: op_role,
                             'ring_id': self.ring_id,
                             #'ring_id': self.ring_id if prev_device_index > cur_device_index else self.ring_id + 2,
@@ -4987,6 +5021,10 @@ class PipelineOptimizer(object):
             and 'local_rank' in main_block.program._pipeline_opt, \
             'Please use pipeline with fleet.'
         local_rank = main_block.program._pipeline_opt['local_rank']
+        schedule_mode = 0
+        if 'schedule_mode' in main_block.program._pipeline_opt:
+            schedule_mode = main_block.program._pipeline_opt['schedule_mode']
+        self.schedule_mode = schedule_mode
 
         self.use_sharding = False
         if 'use_sharding' in main_block.program._pipeline_opt:
@@ -5074,6 +5112,7 @@ class PipelineOptimizer(object):
             "inner_parallelism": len(device_list),
             "num_pipeline_stages": len(device_list),
             "pipeline_stage": local_rank,
+            "schedule_mode": schedule_mode,
             "section_program": program_list[local_rank],
             "place": place_list[local_rank],
             "place_id": place_id,