update, test=develop (#33588)

Author: lilong12

python/paddle/distributed/fleet/meta_optimizers/sharding_optimizer.py

@@ -428,59 +428,33 @@ def _init_comm(self):
 
         # pp ring
         if self.pp_degree > 1:
-            if self.schedule_mode == 'F-then-B':  # GPipe
-                self._collective_helper._init_communicator(
-                    self._startup_program,
-                    self.current_endpoint,
-                    self.pp_group_endpoints,
-                    self.pp_rank,
-                    self.pp_ring_id,
-                    False,
-                    global_ring_id=self.global_ring_id,
-                    sync=False)
-                # append_naive_sync(startup_block, self.startup_prog_sync_var,
-                #                   self.global_ring_id)
+            for pair in self.pipeline_pair:
+                pair_key = pair[0] * 1000 + pair[1]
+                ring_id = self.pp_ring_map[pair_key]
+                print("pp pair:{}, ring_id: {}".format(pair, ring_id))
+                if self.pp_rank not in pair: continue
+                pp_group_endpoints = [
+                    self.pp_group_endpoints[pair[0]],
+                    self.pp_group_endpoints[pair[1]],
+                ]
+                if pair[0] < pair[1]:
+                    start_ring_id = self.pp_ring_id + pair[1] - pair[0] - 1
+                else:
+                    start_ring_id = self.pp_ring_id + 2 + pair[0] - pair[1] - 1
+                pp_rank = 0 if self.pp_rank == pair[0] else 1
                 self._collective_helper._init_communicator(
                     self._startup_program,
                     self.current_endpoint,
-                    self.pp_group_endpoints,
-                    self.pp_rank,
-                    self.pp_ring_id + 2,
+                    pp_group_endpoints,
+                    pp_rank,
+                    ring_id,
                     False,
                     global_ring_id=self.global_ring_id,
                     sync=False)
                 # append_naive_sync(startup_block, self.startup_prog_sync_var,
                 #                   self.global_ring_id)
-            else:
-                assert self.schedule_mode == '1F1B'
-                for pair in self.pipeline_pair:
-                    pair_key = pair[0] * 1000 + pair[1]
-                    ring_id = self.pp_ring_map[pair_key]
-                    print("pp pair:{}, ring_id: {}".format(pair, ring_id))
-                    if self.pp_rank not in pair: continue
-                    pp_group_endpoints = [
-                        self.pp_group_endpoints[pair[0]],
-                        self.pp_group_endpoints[pair[1]],
-                    ]
-                    if pair[0] < pair[1]:
-                        start_ring_id = self.pp_ring_id + pair[1] - pair[0] - 1
-                    else:
-                        start_ring_id = self.pp_ring_id + 2 + pair[0] - pair[
-                            1] - 1
-                    pp_rank = 0 if self.pp_rank == pair[0] else 1
-                    self._collective_helper._init_communicator(
-                        self._startup_program,
-                        self.current_endpoint,
-                        pp_group_endpoints,
-                        pp_rank,
-                        ring_id,
-                        False,
-                        global_ring_id=self.global_ring_id,
-                        sync=False)
-                    # append_naive_sync(startup_block, self.startup_prog_sync_var,
-                    #                   self.global_ring_id)
-
-                # TODO (JZ-LIANG) to unify this shit 
+
+            # TODO (JZ-LIANG) to unify this shit 
             assert self.pp_rank_ == self.pp_rank, "pp rank for pp opt [{}], pp rank for sharding opt [{}]".format(
                 self.pp_rank_, self.pp_rank)