Fix multi nccl comm & wait server ready (#28663)

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

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 
+import copy
 import paddle
 from paddle.fluid.framework import core
 from paddle.fluid import compiler
@@ -51,13 +52,21 @@ class GraphExecutionOptimizer(MetaOptimizerBase):
     # should fix the variable
     def _setup_nccl_op(self, startup_program, main_program, build_strategy):
         trainer_endpoints = self.role_maker._get_trainer_endpoints()
-        trainers = trainer_endpoints
+        other_trainers = copy.copy(trainer_endpoints)
+
         trainer_id = self.role_maker._worker_index()
         current_endpoint = self.role_maker._get_trainer_endpoints()[trainer_id]
+        other_trainers.remove(current_endpoint)
+
         trainer_endpoints_env = ",".join(trainer_endpoints)
         trainers_num = self.role_maker._worker_num()
+
+        if trainer_id == 0:
+            wait_server_ready(other_trainers)
+
         nccl_id_var = startup_program.global_block().create_var(
             name="NCCLID", persistable=True, type=core.VarDesc.VarType.RAW)
+
         for i in range(1, build_strategy.nccl_comm_num):
             startup_program.global_block().create_var(
                 name="NCCLID_{}".format(i),
@@ -90,7 +99,6 @@ class GraphExecutionOptimizer(MetaOptimizerBase):
             })
 
     def _try_to_compile(self, startup_program, main_program, loss):
-        import copy
         dist_strategy = self.user_defined_strategy
         local_build_strategy = paddle.fluid.BuildStrategy()
         local_build_strategy.enable_sequential_execution = \
@@ -148,13 +156,12 @@ class GraphExecutionOptimizer(MetaOptimizerBase):
 
         sync_allreduce = dist_strategy.sync_nccl_allreduce
         if sync_allreduce:
-            exe_strategy.num_threads = local_build_strategy.nccl_comm_num + 1
+            exe_strategy.num_threads = max(
-            if local_build_strategy.use_hierarchical_allreduce:
+                local_build_strategy.nccl_comm_num + 1,
-                exe_strategy.num_threads = 2 * local_build_strategy.nccl_comm_num + 1
+                exe_strategy.num_threads)
-            if exe_strategy.num_threads > 4:
+            if local_build_strategy.nccl_comm_num > 1:
                 logging.warn(
-                    "if you use hierachical_allreduce or "
+                    "nccl_comm_num > 1, you may need to set sync_nccl_allreduce=False to ensure that different nccl comms can overlap"
-                    "with multi nccl comm, please set distributed_strategy.sync_nccl_allreduce=False"
                 )
 
         sync_batch_norm = local_build_strategy.sync_batch_norm
@@ -167,6 +174,11 @@ class GraphExecutionOptimizer(MetaOptimizerBase):
                 "set num_threads=1, nccl_comm_num=1, hierachical_allreduce=False."
             )
 
+        # NOTE. compatible with compiler, otherwise these values will be overwritten by compiler
+        main_program._nccl_comm_num = local_build_strategy.nccl_comm_num
+        main_program._use_hierarchical_allreduce = local_build_strategy.use_hierarchical_allreduce
+        main_program._hierarchical_allreduce_inter_nranks = local_build_strategy.hierarchical_allreduce_inter_nranks
+
         # TODO(guru4elephant): should be an independent optimizer
         self._setup_nccl_op(startup_program, main_program, local_build_strategy)
 

python/paddle/fluid/tests/unittests/test_fleet_graph_execution_meta_optimizer.py

@@ -75,6 +75,9 @@ class TestFleetGraphExecutionMetaOptimizer(unittest.TestCase):
                 optimizer, strategy=strategy)
             optimizer.minimize(avg_cost)
 
+            exe = paddle.fluid.Executor(place=paddle.fluid.CPUPlace())
+            exe.run(paddle.fluid.default_startup_program())
+
         proc_a = launch_func(node_func, node_a)
         proc_a.start()
         proc_b = launch_func(node_func, node_b)
@@ -197,6 +200,9 @@ class TestFleetGraphExecutionMetaOptimizer(unittest.TestCase):
                 optimizer, strategy=strategy)
             optimizer.minimize(avg_cost)
 
+            exe = paddle.fluid.Executor(place=paddle.fluid.CPUPlace())
+            exe.run(paddle.fluid.default_startup_program())
+
         proc_a = launch_func(node_func, node_a)
         proc_a.start()
         proc_b = launch_func(node_func, node_b)