【paddle.fleet】paddle.fleet -> paddle.distributed.fleet. (#26186)

* move paddle.fleet to paddle.distributed.fleet

paddle/fluid/framework/CMakeLists.txt

@@ -164,23 +164,23 @@ if(WITH_PYTHON)
   if (NOT WIN32)
     add_custom_command(TARGET framework_py_proto POST_BUILD
       COMMAND ${CMAKE_COMMAND} -E make_directory ${PADDLE_BINARY_DIR}/python/paddle/fluid/proto
-      COMMAND ${CMAKE_COMMAND} -E make_directory ${PADDLE_BINARY_DIR}/python/paddle/fleet/proto
-      COMMAND ${CMAKE_COMMAND} -E touch ${PADDLE_BINARY_DIR}/python/paddle/fleet/proto/__init__.py
+      COMMAND ${CMAKE_COMMAND} -E make_directory ${PADDLE_BINARY_DIR}/python/paddle/distributed/fleet/proto
+      COMMAND ${CMAKE_COMMAND} -E touch ${PADDLE_BINARY_DIR}/python/paddle/distributed/fleet/proto/__init__.py
       COMMAND cp *.py ${PADDLE_BINARY_DIR}/python/paddle/fluid/proto/
-      COMMAND cp distributed_strategy_*.py ${PADDLE_BINARY_DIR}/python/paddle/fleet/proto
+      COMMAND cp distributed_strategy_*.py ${PADDLE_BINARY_DIR}/python/paddle/distributed/fleet/proto
       COMMENT "Copy generated python proto into directory paddle/fluid/proto."
       WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR})
   else(NOT WIN32)
     string(REPLACE "/" "\\" proto_dstpath "${PADDLE_BINARY_DIR}/python/paddle/fluid/proto/")
-    string(REPLACE "/" "\\" fleet_proto_dstpath "${PADDLE_BINARY_DIR}/python/paddle/fleet/proto/")
+    string(REPLACE "/" "\\" fleet_proto_dstpath "${PADDLE_BINARY_DIR}/python/paddle/distributed/fleet/proto/")
     add_custom_command(TARGET framework_py_proto POST_BUILD
           COMMAND ${CMAKE_COMMAND} -E make_directory ${PADDLE_BINARY_DIR}/python/paddle/fluid/proto
-	  COMMAND ${CMAKE_COMMAND} -E make_directory ${PADDLE_BINARY_DIR}/python/paddle/fleet/proto
-	  COMMAND ${CMAKE_COMMAND} -E touch ${PADDLE_BINARY_DIR}/python/paddle/fleet/proto/__init__.py
+	  COMMAND ${CMAKE_COMMAND} -E make_directory ${PADDLE_BINARY_DIR}/python/paddle/distributed/fleet/proto
+	  COMMAND ${CMAKE_COMMAND} -E touch ${PADDLE_BINARY_DIR}/python/paddle/distributed/fleet/proto/__init__.py
           COMMAND copy /Y *.py ${proto_dstpath}
 	  COMMAND copy /Y distributed_strategy_*.py ${fleet_proto_dstpath}
           COMMENT "Copy generated python proto into directory paddle/fluid/proto."
-	  COMMENT "Copy generated python proto into directory paddle/fleet/proto."
+	  COMMENT "Copy generated python proto into directory paddle/distributed/fleet/proto."
           WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR})
   endif(NOT WIN32)
 endif()

python/paddle/__init__.py

@@ -36,7 +36,7 @@ import paddle.distributed
 import paddle.sysconfig
 import paddle.tensor
 import paddle.nn
-import paddle.fleet
+import paddle.distributed.fleet
 import paddle.framework
 import paddle.optimizer
 import paddle.metric

python/paddle/fleet/base/distributed_strategy.py → python/paddle/distributed/fleet/base/distributed_strategy.py

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 import paddle
-from paddle.fleet.proto import distributed_strategy_pb2
+from paddle.distributed.fleet.proto import distributed_strategy_pb2
 from paddle.fluid.framework import Variable
 import google.protobuf.text_format
 
@@ -103,7 +103,7 @@ class DistributedStrategy(object):
         Examples:
           .. code-block:: python
         
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             strategy = fleet.DistributedStrategy()
             strategy.dgc = True
             strategy.recompute = True
@@ -120,7 +120,7 @@ class DistributedStrategy(object):
         Examples:
           .. code-block:: python
 
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             strategy = fleet.DistributedStrategy()
             strategy.load_from_prototxt("dist_strategy.protoxt")
         """
@@ -141,7 +141,7 @@ class DistributedStrategy(object):
             exe_strategy.num_iteration_per_drop_scope = 10
             exe_strategy.num_iteration_per_run = 10
 
-            strategy = paddle.fleet.DistributedStrategy()
+            strategy = paddle.distributed.fleet.DistributedStrategy()
             strategy.execution_strategy = exe_strategy
         """
         execution_strategy = paddle.fluid.ExecutionStrategy()
@@ -178,7 +178,7 @@ class DistributedStrategy(object):
             build_strategy.fuse_all_optimizer_ops = True
             build_strategy.enable_inplace = True
             
-            strategy = paddle.fleet.DistributedStrategy()
+            strategy = paddle.distributed.fleet.DistributedStrategy()
             strategy.build_strategy = build_strategy
         """
 
@@ -211,7 +211,7 @@ class DistributedStrategy(object):
         Examples:
           .. code-block:: python
 
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             role_maker = fleet.PaddleCloudRoleMaker()
             fleet.init(role_maker)
 
@@ -253,7 +253,7 @@ class DistributedStrategy(object):
         Examples:
           .. code-block:: python
 
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             role_maker = fleet.PaddleCloudRoleMaker()
             fleet.init(role_maker)
 
@@ -282,7 +282,7 @@ class DistributedStrategy(object):
         Examples:
           .. code-block:: python
 
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             strategy = fleet.DistributedStrategy()
             strategy.amp = True # by default this is false
 
@@ -314,7 +314,7 @@ class DistributedStrategy(object):
         Examples:
           .. code-block:: python
 
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             strategy = fleet.DistributedStrategy()
             strategy.recompute = True
             # suppose x and y are names of checkpoint tensors for recomputation
@@ -432,7 +432,7 @@ class DistributedStrategy(object):
         Examples:
           .. code-block:: python
         
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             strategy = fleet.DistributedStrategy()
             strategy.recompute = True
             strategy.recompute_configs = {"checkpionts": ["x", "y"]}
@@ -457,7 +457,7 @@ class DistributedStrategy(object):
         Examples:
           .. code-block:: python
         
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             strategy = fleet.DistributedStrategy()
             strategy.pipeline = True
 
@@ -490,7 +490,7 @@ class DistributedStrategy(object):
         Examples:
           .. code-block:: python
         
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             strategy = fleet.DistributedStrategy()
             strategy.pipeline = True
             strategy.pipeline_configs = {"micro_batch": 12}
@@ -560,7 +560,7 @@ class DistributedStrategy(object):
 
         Examples:
         .. code-block:: python
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             strategy = fleet.DistributedStrategy()
             strategy.gradient_merge = True
             strategy.gradient_merge_configs = {"k_steps": 4, "avg": True}
@@ -583,7 +583,7 @@ class DistributedStrategy(object):
             avg (bool): whether to average the gradients of each mini-batch,
                 the default value is `True`
         Example:
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             strategy = fleet.DistributedStrategy()
             strategy.gradient_merge = True
             strategy.gradient_merge_configs = {"k_steps": 4, "avg": True}

python/paddle/fleet/base/fleet_base.py → python/paddle/distributed/fleet/base/fleet_base.py

@@ -34,9 +34,8 @@ class Fleet(object):
     Examples:
         .. code-block:: python
 
-            import paddle.fleet as fleet
-            import paddle.fluid.incubate.fleet.base.role_maker as role_maker
-            role = role_maker.PaddleCloudRoleMaker(is_collective=True)
+            import paddle.distributed.fleet as fleet
+            role = fleet.role_maker.PaddleCloudRoleMaker(is_collective=True)
             fleet.init(role)
             strategy = fleet.DistributedStrategy()
             optimizer = paddle.optimizer.SGD(learning_rate=0.001)
@@ -218,9 +217,8 @@ class Fleet(object):
 
         Examples:
             .. code-block:: python
-            import paddle.fleet as fleet
-            import paddle.fluid.incubate.fleet.base.role_maker as role_maker
-            role = role_maker.PaddleCloudRoleMaker(is_collective=True)
+            import paddle.distributed.fleet as fleet
+            role = fleet.role_maker.PaddleCloudRoleMaker(is_collective=True)
             fleet.init(role)
             strategy = fleet.DistributedStrategy()
             optimizer = paddle.optimizer.SGD(learning_rate=0.001)
@@ -260,8 +258,7 @@ class Fleet(object):
 
         Examples:
             import paddle
-            import paddle.fleet as fleet
-            import paddle.fluid.incubate.fleet.base.role_maker as role_maker
+            import paddle.distributed.fleet as fleet
 
             fc_1 = paddle.layers.fc(input=input_x, size=hid_dim, act='tanh')
             fc_2 = paddlen.layers.fc(input=fc_1, size=hid_dim, act='tanh')
@@ -269,7 +266,7 @@ class Fleet(object):
             cost = paddle.layers.cross_entropy(input=prediction, label=input_y)
             avg_cost = paddle.layers.mean(x=cost)
 
-            role = role_maker.PaddleCloudRoleMaker(is_collective=True)
+            role = fleet.role_maker.PaddleCloudRoleMaker(is_collective=True)
             fleet.init(role)
             strategy = fleet.DistributedStrategy()
             optimizer = paddle.optimizer.SGD(learning_rate=0.001)

python/paddle/fleet/base/role_maker.py → python/paddle/distributed/fleet/base/role_maker.py

@@ -481,7 +481,7 @@ class PaddleCloudRoleMaker(RoleMakerBase):
         return "lo"
 
     def __start_kv_server(self, http_server_d, size_d):
-        from paddle.fleet.utils import KVServer
+        from paddle.distributed.fleet.utils import KVServer
         http_server = KVServer(int(self._http_ip_port[1]), size_d)
         http_server.start()
         wait_seconds = 5

python/paddle/fleet/base/util_factory.py → python/paddle/distributed/fleet/base/util_factory.py

@@ -55,8 +55,8 @@ class UtilBase(object):
 
     def set_file_system(self, fs_client):
         assert isinstance(
-            fs_client,
-            FS), "fs_client must be the instance of paddle.fleet.utils.FS"
+            fs_client, FS
+        ), "fs_client must be the instance of paddle.distributed.fleet.utils.FS"
         self.fs_client = fs_client
 
     def __check_comm_world(self, comm_world="worker"):

python/paddle/fleet/cloud_utils.py → python/paddle/distributed/fleet/cloud_utils.py

@@ -14,7 +14,7 @@
 
 import os
 import paddle
-from paddle.fleet.launch_utils import get_cluster, logger
+from paddle.distributed.fleet.launch_utils import get_cluster, logger
 
 
 def get_cloud_cluster(args_node_ips, selected_gpus, args_port=6170):

python/paddle/fleet/launch.py → python/paddle/distributed/fleet/launch.py

@@ -66,8 +66,8 @@ from argparse import ArgumentParser, REMAINDER
 import paddle
 import paddle.fluid as fluid
 
-from paddle.fleet.launch_utils import *
-import paddle.fleet.cloud_utils as cloud_utils
+from paddle.distributed.fleet.launch_utils import *
+import paddle.distributed.fleet.cloud_utils as cloud_utils
 
 
 def _print_arguments(args):

python/paddle/fleet/metrics/metric.py → python/paddle/distributed/fleet/metrics/metric.py

@@ -43,7 +43,7 @@ def sum(input, scope=None):
           
           # in train.py, after train or infer
           res = np.array(scope.find_var(global_cnt.name).get_tensor())
-          print("sum array: ", paddle.fleet.sum(res))
+          print("sum array: ", paddle.distributed.fleet.sum(res))
     """
     fleet._role_maker._barrier_worker()
     if scope is None:
@@ -82,7 +82,7 @@ def max(input, scope=None):
 
           # in train.py, after train or infer
           res = np.array(scope.find_var(global_cnt.name).get_tensor())
-          print("max array: ", paddle.fleet.max(res))
+          print("max array: ", paddle.distributed.fleet.max(res))
     """
     fleet._role_maker._barrier_worker()
     if scope is None:
@@ -121,7 +121,7 @@ def min(input, scope=None):
 
           # in train.py, after train or infer
           res = np.array(scope.find_var(global_cnt.name).get_tensor())
-          print("min array: ", paddle.fleet.min(res))
+          print("min array: ", paddle.distributed.fleet.min(res))
     """
     fleet._role_maker._barrier_worker()
     if scope is None:
@@ -162,7 +162,7 @@ def auc(stat_pos, stat_neg, scope=None):
           # in train.py, after train or infer
           pos = np.array(scope.find_var(stat_pos.name).get_tensor())
           neg = np.array(scope.find_var(stat_neg.name).get_tensor())
-          print("auc: ", paddle.fleet.auc(pos, neg))
+          print("auc: ", paddle.distributed.fleet.auc(pos, neg))
     """
     fleet._role_maker._barrier_worker()
     if scope is None:
@@ -240,7 +240,7 @@ def mae(abserr, total_ins_num, scope=None):
 
           # in train.py, after train or infer
           res = np.array(scope.find_var(abserr.name).get_tensor())
-          print("mae: ", paddle.fleet.mae(res, total_ins_num))
+          print("mae: ", paddle.distributed.fleet.mae(res, total_ins_num))
     """
     fleet._role_maker._barrier_worker()
     if scope is None:
@@ -278,7 +278,7 @@ def rmse(sqrerr, total_ins_num, scope=None):
 
           # in train.py, after train or infer
           res = np.array(scope.find_var(sqrerr.name).get_tensor())
-          print("rmse: ", paddle.fleet.rmse(res, total_ins_num))
+          print("rmse: ", paddle.distributed.fleet.rmse(res, total_ins_num))
     """
     fleet._role_maker._barrier_worker()
     if scope is None:
@@ -316,7 +316,7 @@ def mse(sqrerr, total_ins_num, scope=None):
 
           # in train.py, after train or infer
           metric = np.array(scope.find_var(sqrerr.name).get_tensor())
-          print("mse: ", paddle.fleet.mse(metric, total_ins_num))
+          print("mse: ", paddle.distributed.fleet.mse(metric, total_ins_num))
     """
     fleet._role_maker._barrier_worker()
     if scope is None:
@@ -365,7 +365,7 @@ def acc(correct, total, scope=None):
           # in train.py, after train or infer
           correct_num = np.array(scope.find_var(correct.name).get_tensor())
           total_num = np.array(scope.find_var(total.name).get_tensor())
-          print("accuracy: ", paddle.fleet.acc(correct_num, total_num))
+          print("accuracy: ", paddle.distributed.fleet.acc(correct_num, total_num))
     """
     fleet._role_maker._barrier_worker()
     if scope is None:

python/paddle/fluid/incubate/checkpoint/auto_checkpoint.py

@@ -305,7 +305,7 @@ class TrainEpochRange(SerializableBase):
         if self._checker.ce_test:
             config = None
 
-        from paddle.fleet.utils.fs import HDFSClient
+        from paddle.distributed.fleet.utils.fs import HDFSClient
         self._hdfs = HDFSClient(self._checker.hdfs_home, config)
 
         self._cper = CheckpointSaver(self._hdfs)

python/paddle/fluid/incubate/checkpoint/checkpoint_saver.py

@@ -79,7 +79,7 @@ class CheckpointSaver(object):
         tmp_path = "{}.tmp".format(real_path)
         saved_path = tmp_path
 
-        from paddle.fleet.utils.fs import LocalFS
+        from paddle.distributed.fleet.utils.fs import LocalFS
         local_fs = LocalFS()
 
         cache_path = None
@@ -134,7 +134,7 @@ class CheckpointSaver(object):
             assert isinstance(checkpoint_no, int)
             assert checkpoint_no >= 0
 
-        from paddle.fleet.utils.fs import LocalFS
+        from paddle.distributed.fleet.utils.fs import LocalFS
         local_fs = LocalFS()
         if self._fs.need_upload_download():
             cache_path = "{}/{}.{}.load_cache".format(

python/paddle/fluid/incubate/fleet/base/fleet_base.py

@@ -21,7 +21,7 @@ from paddle.fluid.executor import Executor
 from paddle.fluid.optimizer import SGD
 
 from paddle.fluid.incubate.fleet.base.mode import Mode
-from paddle.fleet.base.role_maker import RoleMakerBase
+from paddle.distributed.fleet.base.role_maker import RoleMakerBase
 from paddle.fluid.contrib.mixed_precision.decorator import OptimizerWithMixedPrecision
 from . import mode
 

python/paddle/fluid/reader.py

@@ -1709,7 +1709,7 @@ class PyReader(DataLoaderBase):
 
 class DatasetLoader(DataLoaderBase):
     def __init__(self, dataset, places, drop_last):
-        assert isinstance(dataset, paddle.fleet.dataset.
+        assert isinstance(dataset, paddle.distributed.fleet.dataset.
                           DatasetBase), "dataset must be type of DatasetBase"
         assert not in_dygraph_mode(
         ), "DatasetLoader is not supported in dygraph mode yet"
@@ -1725,7 +1725,7 @@ class DatasetLoader(DataLoaderBase):
 
         dataset.set_thread(thread_num)
 
-        if isinstance(dataset, paddle.fleet.dataset.
+        if isinstance(dataset, paddle.distributed.fleet.dataset.
                       InMemoryDataset) and dataset.queue_num > thread_num:
             logging.warn("queue_num {} which is set in Dataset is ignored".
                          format(dataset.queue_num))

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

@@ -18,7 +18,7 @@ import unittest
 import os
 import paddle.fluid.core as core
 import paddle.fluid as fluid
-from paddle.fleet.base.private_helper_function import wait_server_ready
+from paddle.distributed.fleet.base.private_helper_function import wait_server_ready
 
 
 class TestCCommInitOp(unittest.TestCase):

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

@@ -28,7 +28,7 @@ import numpy as np
 
 import ctr_dataset_reader
 from test_dist_fleet_base import runtime_main, FleetDistRunnerBase
-from paddle.fleet.base.util_factory import fleet_util
+from paddle.distributed.fleet.base.util_factory import fleet_util
 
 # Fix seed for test
 fluid.default_startup_program().random_seed = 1
@@ -217,7 +217,7 @@ class TestDistCTR2x2(FleetDistRunnerBase):
             filelist.append(train_file_path)
 
         # config dataset
-        dataset = paddle.fleet.DatasetFactory().create_dataset()
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset()
         dataset.set_batch_size(batch_size)
         dataset.set_use_var(self.feeds)
         pipe_command = 'python ctr_dataset_reader.py'

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

@@ -20,7 +20,7 @@ from paddle.fluid.incubate.fleet.collective import CollectiveOptimizer, fleet
 import os
 import sys
 
-from paddle.fleet.utils.fs import LocalFS, HDFSClient
+from paddle.distributed.fleet.utils.fs import LocalFS, HDFSClient
 import paddle.fluid.incubate.checkpoint.auto_checkpoint as acp
 from paddle.fluid.incubate.checkpoint.checkpoint_saver import PaddleModel
 from paddle.fluid.framework import program_guard

python/paddle/fluid/tests/unittests/test_auto_checkpoint2.py

@@ -20,7 +20,7 @@ from paddle.fluid.incubate.fleet.collective import CollectiveOptimizer, fleet
 import os
 import sys
 
-from paddle.fleet.utils.fs import LocalFS, HDFSClient
+from paddle.distributed.fleet.utils.fs import LocalFS, HDFSClient
 import paddle.fluid.incubate.checkpoint.auto_checkpoint as acp
 from paddle.fluid.incubate.checkpoint.checkpoint_saver import PaddleModel
 from paddle.fluid.framework import program_guard

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

@@ -26,7 +26,7 @@ import paddle
 import paddle.fluid as fluid
 
 import paddle.fluid.incubate.fleet.base.role_maker as role_maker
-import paddle.fleet as fleet
+import paddle.distributed.fleet as fleet
 
 
 class TestCommunicatorGeoEnd2End(unittest.TestCase):
@@ -108,7 +108,7 @@ class TestCommunicatorGeoEnd2End(unittest.TestCase):
 
         role = role_maker.PaddleCloudRoleMaker()
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.a_sync = True
         strategy.a_sync_configs = {"k_steps": 100}
 
@@ -136,7 +136,7 @@ import paddle.fluid as fluid
 from paddle.fluid.communicator import Communicator
 import paddle.fluid.incubate.fleet.base.role_maker as role_maker
 from paddle.fluid.incubate.fleet.parameter_server.mode import DistributedMode
-import paddle.fleet as fleet
+import paddle.distributed.fleet as fleet
 
 from test_communicator_geo import TestCommunicatorGeoEnd2End
 

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

@@ -22,7 +22,7 @@ import paddle
 import paddle.fluid as fluid
 
 import paddle.fluid.incubate.fleet.base.role_maker as role_maker
-import paddle.fleet as fleet
+import paddle.distributed.fleet as fleet
 
 
 class TestCommunicator(unittest.TestCase):
@@ -51,7 +51,7 @@ class TestCommunicator(unittest.TestCase):
 
         optimizer = fluid.optimizer.SGD(0.01)
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.a_sync = False
 
         optimizer = fleet.distributed_optimizer(optimizer, strategy)

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

@@ -38,25 +38,25 @@ class TestDataset(unittest.TestCase):
     def test_dataset_create(self):
         """ Testcase for dataset create. """
         try:
-            dataset = paddle.fleet.DatasetFactory().create_dataset(
+            dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
                 "InMemoryDataset")
         except:
             self.assertTrue(False)
 
         try:
-            dataset = paddle.fleet.DatasetFactory().create_dataset(
+            dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
                 "QueueDataset")
         except:
             self.assertTrue(False)
 
         try:
-            dataset = paddle.fleet.DatasetFactory().create_dataset(
+            dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
                 "FileInstantDataset")
         except:
             self.assertTrue(False)
 
         try:
-            dataset = paddle.fleet.DatasetFactory().create_dataset(
+            dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
                 "MyOwnDataset")
             self.assertTrue(False)
         except:
@@ -95,7 +95,7 @@ class TestDataset(unittest.TestCase):
                 name=slot, shape=[1], dtype="int64", lod_level=1)
             slots_vars.append(var)
 
-        dataset = paddle.fleet.DatasetFactory().create_dataset(
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
             "InMemoryDataset")
         dataset.set_batch_size(32)
         dataset.set_thread(3)
@@ -176,7 +176,7 @@ class TestDataset(unittest.TestCase):
                 name=slot, shape=[1], dtype="int64", lod_level=1)
             slots_vars.append(var)
 
-        dataset = paddle.fleet.DatasetFactory().create_dataset(
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
             "InMemoryDataset")
         dataset.set_batch_size(32)
         dataset.set_thread(3)
@@ -228,7 +228,7 @@ class TestDataset(unittest.TestCase):
                 name=slot, shape=[1], dtype="int64", lod_level=1)
             slots_vars.append(var)
 
-        dataset = paddle.fleet.DatasetFactory().create_dataset(
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
             "InMemoryDataset")
         dataset.set_batch_size(32)
         dataset.set_thread(3)
@@ -300,7 +300,7 @@ class TestDataset(unittest.TestCase):
                     name=slot, shape=[1], dtype="float32", lod_level=1)
                 slots_vars.append(var)
 
-        dataset = paddle.fleet.DatasetFactory().create_dataset(
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
             "InMemoryDataset")
         dataset.set_batch_size(32)
         dataset.set_thread(1)
@@ -367,7 +367,7 @@ class TestDataset(unittest.TestCase):
                 name="slot4", shape=[1], dtype="float32", lod_level=0)
             slots_vars = [var1, var2, var3, var4]
 
-        dataset = paddle.fleet.DatasetFactory().create_dataset(
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
             "InMemoryDataset")
         dataset.set_batch_size(32)
         dataset.set_thread(1)
@@ -423,7 +423,7 @@ class TestDataset(unittest.TestCase):
                 name=slot, shape=[1], dtype="float32", lod_level=1)
             slots_vars.append(var)
 
-        dataset = paddle.fleet.DatasetFactory().create_dataset(
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
             "InMemoryDataset")
         dataset.set_batch_size(32)
         dataset.set_thread(3)
@@ -517,7 +517,8 @@ class TestDataset(unittest.TestCase):
                 name=slot, shape=[1], dtype="int64", lod_level=1)
             slots_vars.append(var)
 
-        dataset = paddle.fleet.DatasetFactory().create_dataset("QueueDataset")
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
+            "QueueDataset")
         dataset.set_batch_size(32)
         dataset.set_thread(3)
         dataset.set_filelist(
@@ -542,7 +543,8 @@ class TestDataset(unittest.TestCase):
                 except Exception as e:
                     self.assertTrue(False)
 
-        dataset2 = paddle.fleet.DatasetFactory().create_dataset("QueueDataset")
+        dataset2 = paddle.distributed.fleet.DatasetFactory().create_dataset(
+            "QueueDataset")
         dataset2.set_use_var(slots_vars)
         dataset2.set_batch_size(32)
         dataset2.set_thread(3)
@@ -583,7 +585,8 @@ class TestDataset(unittest.TestCase):
                 name=slot, shape=[1], dtype="float32", lod_level=1)
             slots_vars.append(var)
 
-        dataset = paddle.fleet.DatasetFactory().create_dataset("QueueDataset")
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
+            "QueueDataset")
         dataset.set_batch_size(32)
         dataset.set_thread(3)
         dataset.set_filelist(
@@ -638,7 +641,7 @@ class TestDataset(unittest.TestCase):
                 name=slot, shape=[None, 1], dtype="int64", lod_level=1)
             slots_vars.append(var)
 
-        dataset = paddle.fleet.DatasetFactory().create_dataset(
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
             "InMemoryDataset")
         dataset.set_input_type(1)
         dataset.set_batch_size(1)
@@ -718,7 +721,8 @@ class TestDatasetWithFetchHandler(unittest.TestCase):
             inputs(list): inputs of get_dataset
             files(list): files of  get_dataset
         """
-        dataset = paddle.fleet.DatasetFactory().create_dataset("QueueDataset")
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
+            "QueueDataset")
         dataset.set_batch_size(32)
         dataset.set_thread(3)
         dataset.set_filelist(files)
@@ -875,7 +879,7 @@ class TestDataset2(unittest.TestCase):
             except ImportError as e:
                 print("warning: no mpi4py")
             exe.run(startup_program)
-            dataset = paddle.fleet.DatasetFactory().create_dataset(
+            dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
                 "InMemoryDataset")
             dataset.set_batch_size(32)
             dataset.set_thread(3)
@@ -945,7 +949,7 @@ class TestDataset2(unittest.TestCase):
             except ImportError as e:
                 print("warning: no mpi4py")
             exe.run(startup_program)
-            dataset = paddle.fleet.DatasetFactory().create_dataset(
+            dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
                 "InMemoryDataset")
             dataset.set_batch_size(32)
             dataset.set_thread(3)
@@ -962,12 +966,12 @@ class TestDataset2(unittest.TestCase):
                 print("warning: catch expected error")
             fleet._opt_info = None
             fleet._fleet_ptr = None
-            dataset = paddle.fleet.DatasetFactory().create_dataset(
+            dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
                 "InMemoryDataset")
             dataset.set_rank_offset("")
             dataset.set_pv_batch_size(1)
             dataset.set_hdfs_config("", "")
-            d = paddle.fleet.DatasetBase()
+            d = paddle.distributed.fleet.DatasetBase()
             try:
                 dataset.set_feed_type("MultiSlotInMemoryDataFeed")
             except:
@@ -1000,7 +1004,7 @@ class TestDataset2(unittest.TestCase):
             dataset.get_pv_data_size()
             dataset.get_memory_data_size()
             dataset.get_shuffle_data_size()
-            dataset = paddle.fleet.DatasetFactory().create_dataset(
+            dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
                 "QueueDataset")
             try:
                 dataset.local_shuffle()
@@ -1010,7 +1014,7 @@ class TestDataset2(unittest.TestCase):
                 dataset.global_shuffle()
             except:
                 print("warning: catch expected error")
-            dataset = paddle.fleet.FileInstantDataset()
+            dataset = paddle.distributed.fleet.FileInstantDataset()
             try:
                 dataset.local_shuffle()
             except:

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

@@ -97,7 +97,7 @@ class DatasetLoaderTestBase(unittest.TestCase):
 
     def check_batch_number(self, place, randomize_batch_num=False):
         main_prog, startup_prog, feeds = self.build_network()
-        dataset = paddle.fleet.DatasetFactory().create_dataset(
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
             self.dataset_name)
         dataset.set_batch_size(BATCH_SIZE)
 

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

@@ -33,7 +33,7 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
 
     def test_a_sync_optimizer_trainer(self):
         os.environ["TRAINING_ROLE"] = "TRAINER"
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
 
         main_program = paddle.fluid.Program()
         startup_program = paddle.fluid.Program()
@@ -53,7 +53,7 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
             input=prediction, label=input_y)
         avg_cost = paddle.fluid.layers.mean(x=cost)
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.a_sync = True
         optimizer = paddle.optimizer.SGD(learning_rate=0.01)
         optimizer = fleet.distributed_optimizer(optimizer, strategy=strategy)
@@ -78,7 +78,7 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
 
     def test_a_sync_optimizer_pserver(self):
         os.environ["TRAINING_ROLE"] = "PSERVER"
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
 
         main_program = paddle.fluid.Program()
         startup_program = paddle.fluid.Program()
@@ -98,7 +98,7 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
             input=prediction, label=input_y)
         avg_cost = paddle.fluid.layers.mean(x=cost)
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.a_sync = True
         optimizer = paddle.optimizer.SGD(learning_rate=0.01)
         optimizer = fleet.distributed_optimizer(optimizer, strategy=strategy)

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

@@ -32,7 +32,7 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
 
     def test_a_sync_optimizer_trainer(self):
         os.environ["TRAINING_ROLE"] = "TRAINER"
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
 
         main_program = paddle.fluid.Program()
         startup_program = paddle.fluid.Program()
@@ -52,7 +52,7 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
             input=prediction, label=input_y)
         avg_cost = paddle.fluid.layers.mean(x=cost)
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.a_sync = True
         strategy.a_sync_configs = {"k_steps": 100}
         optimizer = paddle.optimizer.SGD(learning_rate=0.01)
@@ -75,7 +75,7 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
 
     def test_a_sync_optimizer_pserver(self):
         os.environ["TRAINING_ROLE"] = "PSERVER"
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
 
         main_program = paddle.fluid.Program()
         startup_program = paddle.fluid.Program()
@@ -95,7 +95,7 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
             input=prediction, label=input_y)
         avg_cost = paddle.fluid.layers.mean(x=cost)
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.a_sync = True
         strategy.a_sync_configs = {"k_steps": 100}
         optimizer = paddle.optimizer.SGD(learning_rate=0.01)

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

@@ -15,7 +15,7 @@
 import unittest
 import paddle
 import os
-import paddle.fleet as fleet
+import paddle.distributed.fleet as fleet
 import paddle.fluid.incubate.fleet.base.role_maker as role_maker
 import time
 
@@ -45,7 +45,7 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
             input=prediction, label=input_y)
         avg_cost = paddle.fluid.layers.mean(x=cost)
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.a_sync = False
         optimizer = paddle.optimizer.SGD(learning_rate=0.01)
         optimizer = fleet.distributed_optimizer(optimizer, strategy=strategy)

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

@@ -32,8 +32,8 @@ import tempfile
 import unittest
 
 import paddle.fluid as fluid
-import paddle.fleet.base.role_maker as role_maker
-from paddle.fleet.base.util_factory import fleet_util
+import paddle.distributed.fleet.base.role_maker as role_maker
+from paddle.distributed.fleet.base.util_factory import fleet_util
 from paddle.fluid.incubate.fleet.parameter_server.distribute_transpiler import fleet
 from paddle.fluid.incubate.fleet.parameter_server.distribute_transpiler.distributed_strategy import StrategyFactory
 

python/paddle/fluid/tests/unittests/test_dist_fleet_ps2.py

@@ -22,7 +22,7 @@ import shutil
 import paddle
 import paddle.fluid as fluid
 import paddle.fluid.incubate.fleet.base.role_maker as role_maker
-import paddle.fleet as fleet
+import paddle.distributed.fleet as fleet
 
 # For Net
 base_lr = 0.2
@@ -163,7 +163,7 @@ class TestPSPassWithBow(unittest.TestCase):
         fleet.init(role)
         loss, acc, _ = self.net()
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.a_sync = True
         optimizer = paddle.optimizer.SGD(learning_rate=0.01)
         optimizer = fleet.distributed_optimizer(optimizer, strategy=strategy)

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

@@ -23,7 +23,7 @@ class TestFleetAMPOptimizer(unittest.TestCase):
         os.environ["PADDLE_TRAINER_ENDPOINTS"] = "127.0.0.1:36001"
 
     def test_amp_optimizer(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -38,7 +38,7 @@ class TestFleetAMPOptimizer(unittest.TestCase):
             input=prediction, label=input_y)
         avg_cost = paddle.fluid.layers.mean(x=cost)
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.amp = True
         strategy.amp_configs = {
             "init_loss_scaling": 32768,

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

@@ -26,13 +26,13 @@ class TestFleetBase(unittest.TestCase):
                        "127.0.0.1:36001,127.0.0.2:36001"
 
     def test_init(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
 
     def test_is_first_worker(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -40,21 +40,21 @@ class TestFleetBase(unittest.TestCase):
             print("test fleet first worker done.")
 
     def test_worker_index(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
         print(fleet.worker_index())
 
     def test_worker_num(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
         print(fleet.worker_num())
 
     def test_is_worker(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -62,14 +62,14 @@ class TestFleetBase(unittest.TestCase):
             print("test fleet is worker")
 
     def test_worker_endpoints(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
         print(fleet.worker_endpoints(to_string=True))
 
     def test_server_num(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -77,7 +77,7 @@ class TestFleetBase(unittest.TestCase):
             print("fleet server num: {}".format(fleet.server_num()))
 
     def test_server_index(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -85,7 +85,7 @@ class TestFleetBase(unittest.TestCase):
             print("fleet server index: {}".format(fleet.server_index()))
 
     def test_server_endpoints(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -94,7 +94,7 @@ class TestFleetBase(unittest.TestCase):
                 fleet.server_endpoints(to_string=True)))
 
     def test_is_server(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -102,14 +102,14 @@ class TestFleetBase(unittest.TestCase):
             print("test fleet is server")
 
     def test_util(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
         self.assertEqual(fleet.util, None)
 
     def test_barrier_worker(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -117,7 +117,7 @@ class TestFleetBase(unittest.TestCase):
             fleet.barrier_worker()
 
     def test_init_worker(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -125,7 +125,7 @@ class TestFleetBase(unittest.TestCase):
             fleet.init_worker()
 
     def test_run_server(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -133,7 +133,7 @@ class TestFleetBase(unittest.TestCase):
             fleet.run_worker()
 
     def test_stop_worker(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -141,7 +141,7 @@ class TestFleetBase(unittest.TestCase):
             fleet.stop_worker()
 
     def test_distributed_optimizer(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -151,7 +151,7 @@ class TestFleetBase(unittest.TestCase):
 
     def test_minimize(self):
         import paddle
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
 
         input_x = paddle.fluid.layers.data(

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

@@ -16,7 +16,7 @@ import unittest
 import paddle
 from paddle import fluid
 import os
-import paddle.fleet as fleet
+import paddle.distributed.fleet as fleet
 import paddle.fluid.incubate.fleet.base.role_maker as role_maker
 
 
@@ -47,7 +47,7 @@ class TestFleetDGCOptimizer(unittest.TestCase):
                     input=prediction, label=input_y)
                 avg_cost = paddle.fluid.layers.mean(x=cost)
 
-                strategy = paddle.fleet.DistributedStrategy()
+                strategy = paddle.distributed.fleet.DistributedStrategy()
                 strategy.dgc = True
                 strategy.dgc_configs = {
                     "rampup_begin_step": 128,

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

@@ -19,7 +19,7 @@ import os
 
 class TestStrategyConfig(unittest.TestCase):
     def test_amp(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.amp = True
         self.assertEqual(strategy.amp, True)
         strategy.amp = False
@@ -28,7 +28,7 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.amp, False)
 
     def test_amp_configs(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         configs = {
             "init_loss_scaling": 32768,
             "decr_every_n_nan_or_inf": 2,
@@ -41,7 +41,7 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.amp_configs["init_loss_scaling"], 32768)
 
     def test_recompute(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.recompute = True
         self.assertEqual(strategy.recompute, True)
         strategy.recompute = False
@@ -50,13 +50,13 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.recompute, False)
 
     def test_recompute_configs(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         configs = {"checkpoints": ["x", "y"]}
         strategy.recompute_configs = configs
         self.assertEqual(len(strategy.recompute_configs["checkpoints"]), 2)
 
     def test_pipeline(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.pipeline = True
         self.assertEqual(strategy.pipeline, True)
         strategy.pipeline = False
@@ -65,13 +65,13 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.pipeline, False)
 
     def test_pipeline_configs(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         configs = {"micro_batch": 4}
         strategy.pipeline_configs = configs
         self.assertEqual(strategy.pipeline_configs["micro_batch"], 4)
 
     def test_localsgd(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.localsgd = True
         self.assertEqual(strategy.localsgd, True)
         strategy.localsgd = False
@@ -80,13 +80,13 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.localsgd, False)
 
     def test_localsgd_configs(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         configs = {"k_steps": 4}
         strategy.localsgd_configs = configs
         self.assertEqual(strategy.localsgd_configs["k_steps"], 4)
 
     def test_dgc(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.dgc = True
         self.assertEqual(strategy.dgc, True)
         strategy.dgc = False
@@ -95,7 +95,7 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.dgc, False)
 
     def test_sync_nccl_allreduce(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.sync_nccl_allreduce = True
         self.assertEqual(strategy.sync_nccl_allreduce, True)
         strategy.sync_nccl_allreduce = False
@@ -104,14 +104,14 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.sync_nccl_allreduce, False)
 
     def test_nccl_comm_num(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.nccl_comm_num = 1
         self.assertEqual(strategy.nccl_comm_num, 1)
         strategy.nccl_comm_num = "2"
         self.assertEqual(strategy.nccl_comm_num, 1)
 
     def test_use_hierarchical_allreduce(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.use_hierarchical_allreduce = True
         self.assertEqual(strategy.use_hierarchical_allreduce, True)
         strategy.use_hierarchical_allreduce = False
@@ -120,14 +120,14 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.use_hierarchical_allreduce, False)
 
     def test_hierarchical_allreduce_inter_nranks(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.hierarchical_allreduce_inter_nranks = 8
         self.assertEqual(strategy.hierarchical_allreduce_inter_nranks, 8)
         strategy.hierarchical_allreduce_inter_nranks = "4"
         self.assertEqual(strategy.hierarchical_allreduce_inter_nranks, 8)
 
     def test_sync_batch_norm(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.sync_batch_norm = True
         self.assertEqual(strategy.sync_batch_norm, True)
         strategy.sync_batch_norm = False
@@ -136,7 +136,7 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.sync_batch_norm, False)
 
     def test_fuse_all_reduce_ops(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.fuse_all_reduce_ops = True
         self.assertEqual(strategy.fuse_all_reduce_ops, True)
         strategy.fuse_all_reduce_ops = False
@@ -145,21 +145,21 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.fuse_all_reduce_ops, False)
 
     def test_fuse_grad_size_in_MB(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.fuse_grad_size_in_MB = 50
         self.assertEqual(strategy.fuse_grad_size_in_MB, 50)
         strategy.fuse_grad_size_in_MB = "40"
         self.assertEqual(strategy.fuse_grad_size_in_MB, 50)
 
     def test_fuse_grad_size_in_TFLOPS(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy._fuse_grad_size_in_TFLOPS = 0.1
         self.assertGreater(strategy._fuse_grad_size_in_TFLOPS, 0.09)
         strategy._fuse_grad_size_in_TFLOPS = "0.3"
         self.assertGreater(strategy._fuse_grad_size_in_TFLOPS, 0.09)
 
     def test_gradient_merge(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.gradient_merge = True
         self.assertEqual(strategy.gradient_merge, True)
         strategy.gradient_merge = False
@@ -168,13 +168,13 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.gradient_merge, False)
 
     def test_gradient_merge_configs(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         configs = {"k_steps": 4}
         strategy.gradient_merge_configs = configs
         self.assertEqual(strategy.gradient_merge_configs["k_steps"], 4)
 
     def test_lars(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.lars = True
         self.assertEqual(strategy.lars, True)
         strategy.lars = False
@@ -183,7 +183,7 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.lars, False)
 
     def test_lamb(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.lamb = True
         self.assertEqual(strategy.lamb, True)
         strategy.lamb = False
@@ -192,7 +192,7 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.lamb, False)
 
     def test_a_sync(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.a_sync = True
         self.assertEqual(strategy.a_sync, True)
         strategy.a_sync = False
@@ -202,13 +202,13 @@ class TestStrategyConfig(unittest.TestCase):
             strategy.a_sync = "True"
 
     def test_a_sync_configs(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         configs = {"k_steps": 1000}
         strategy.a_sync_configs = configs
         self.assertEqual(strategy.a_sync_configs["k_steps"], 1000)
 
     def test_elastic(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.elastic = True
         self.assertEqual(strategy.elastic, True)
         strategy.elastic = False
@@ -217,7 +217,7 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.elastic, False)
 
     def test_auto(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.auto = True
         self.assertEqual(strategy.auto, True)
         strategy.auto = False
@@ -226,7 +226,7 @@ class TestStrategyConfig(unittest.TestCase):
         self.assertEqual(strategy.auto, False)
 
     def test_strategy_prototxt(self):
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.a_sync = True
         strategy.localsgd = True
         strategy.dgc = True
@@ -255,7 +255,7 @@ class TestStrategyConfig(unittest.TestCase):
         exe_strategy.num_iteration_per_run = 10
         strategy.execution_strategy = exe_strategy
         strategy.save_to_prototxt("dist_strategy.prototxt")
-        strategy2 = paddle.fleet.DistributedStrategy()
+        strategy2 = paddle.distributed.fleet.DistributedStrategy()
         strategy2.load_from_prototxt("dist_strategy.prototxt")
         self.assertEqual(strategy.dgc, strategy2.dgc)
 
@@ -277,7 +277,7 @@ class TestStrategyConfig(unittest.TestCase):
         build_strategy.enable_backward_optimizer_op_deps = True
         build_strategy.trainers_endpoints = ["1", "2"]
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.build_strategy = build_strategy
 
     def test_execution_strategy(self):
@@ -286,7 +286,7 @@ class TestStrategyConfig(unittest.TestCase):
         exe_strategy.num_iteration_per_drop_scope = 10
         exe_strategy.num_iteration_per_run = 10
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.execution_strategy = exe_strategy
 
 

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

@@ -15,7 +15,7 @@
 import unittest
 import paddle
 import os
-import paddle.fleet as fleet
+import paddle.distributed.fleet as fleet
 import paddle.fluid.incubate.fleet.base.role_maker as role_maker
 
 
@@ -41,7 +41,7 @@ class TestFleetGradientMergeMetaOptimizer(unittest.TestCase):
             input=prediction, label=input_y)
         avg_cost = paddle.fluid.layers.mean(x=cost)
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.gradient_merge = True
         strategy.gradient_merge_configs = {"k_steps": 2, "avg": True}
         optimizer = paddle.optimizer.SGD(learning_rate=0.01)

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

@@ -39,7 +39,7 @@ class TestFleetGraphExecutionMetaOptimizer(unittest.TestCase):
         }
 
         def node_func():
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             import paddle.fluid.incubate.fleet.base.role_maker as role_maker
             role = role_maker.PaddleCloudRoleMaker(is_collective=True)
             fleet.init(role)
@@ -57,7 +57,7 @@ class TestFleetGraphExecutionMetaOptimizer(unittest.TestCase):
                 input=prediction, label=input_y)
             avg_cost = paddle.fluid.layers.mean(x=cost)
 
-            strategy = paddle.fleet.DistributedStrategy()
+            strategy = paddle.distributed.fleet.DistributedStrategy()
             optimizer = paddle.optimizer.SGD(learning_rate=0.01)
             optimizer = fleet.distributed_optimizer(
                 optimizer, strategy=strategy)
@@ -90,7 +90,7 @@ class TestFleetGraphExecutionMetaOptimizer(unittest.TestCase):
         }
 
         def node_func():
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             import paddle.fluid.incubate.fleet.base.role_maker as role_maker
             role = role_maker.PaddleCloudRoleMaker(is_collective=True)
             fleet.init(role)
@@ -108,7 +108,7 @@ class TestFleetGraphExecutionMetaOptimizer(unittest.TestCase):
                 input=prediction, label=input_y)
             avg_cost = paddle.fluid.layers.mean(x=cost)
 
-            strategy = paddle.fleet.DistributedStrategy()
+            strategy = paddle.distributed.fleet.DistributedStrategy()
             strategy.nccl_comm_num = 2
             strategy.sync_nccl_allreduce = True
             optimizer = paddle.optimizer.SGD(learning_rate=0.01)

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

@@ -39,7 +39,7 @@ class TestFleetGraphExecutionMetaOptimizer(unittest.TestCase):
         }
 
         def node_func():
-            import paddle.fleet as fleet
+            import paddle.distributed.fleet as fleet
             import paddle.fluid.incubate.fleet.base.role_maker as role_maker
             role = role_maker.PaddleCloudRoleMaker(is_collective=True)
             fleet.init(role)
@@ -57,7 +57,7 @@ class TestFleetGraphExecutionMetaOptimizer(unittest.TestCase):
                 input=prediction, label=input_y)
             avg_cost = paddle.fluid.layers.mean(x=cost)
 
-            strategy = paddle.fleet.DistributedStrategy()
+            strategy = paddle.distributed.fleet.DistributedStrategy()
             strategy.nccl_comm_num = 2
             strategy.sync_nccl_allreduce = True
             optimizer = paddle.optimizer.SGD(learning_rate=0.01)

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

@@ -16,7 +16,7 @@ import unittest
 import paddle
 from paddle import fluid
 import os
-import paddle.fleet as fleet
+import paddle.distributed.fleet as fleet
 import paddle.fluid.incubate.fleet.base.role_maker as role_maker
 
 
@@ -47,7 +47,7 @@ class TestFleetLambMetaOptimizer(unittest.TestCase):
                     input=prediction, label=input_y)
                 avg_cost = paddle.fluid.layers.mean(x=cost)
 
-                strategy = paddle.fleet.DistributedStrategy()
+                strategy = paddle.distributed.fleet.DistributedStrategy()
                 strategy.lamb = True
                 strategy.lamb_configs = {
                     'lamb_weight_decay': 0.01,

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

@@ -16,7 +16,7 @@ import unittest
 import paddle
 from paddle import fluid
 import os
-import paddle.fleet as fleet
+import paddle.distributed.fleet as fleet
 import paddle.fluid.incubate.fleet.base.role_maker as role_maker
 
 
@@ -47,7 +47,7 @@ class TestFleetLarsMetaOptimizer(unittest.TestCase):
                     input=prediction, label=input_y)
                 avg_cost = paddle.fluid.layers.mean(x=cost)
 
-                strategy = paddle.fleet.DistributedStrategy()
+                strategy = paddle.distributed.fleet.DistributedStrategy()
                 strategy.lars = True
                 strategy.lars_configs = {
                     "lars_coeff": 0.001,

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

@@ -16,7 +16,7 @@ import unittest
 import paddle
 import os
 
-import paddle.fleet as fleet
+import paddle.distributed.fleet as fleet
 import paddle.fluid.incubate.fleet.base.role_maker as role_maker
 
 
@@ -39,7 +39,7 @@ class TestFleetLocalSGDMetaOptimizer(unittest.TestCase):
             input=prediction, label=input_y)
         avg_cost = paddle.fluid.layers.mean(x=cost)
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.localsgd = True
         strategy.auto = True
         config = strategy.localsgd_configs

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

@@ -19,7 +19,7 @@ import paddle
 import paddle.fluid as fluid
 import os
 import unittest
-import paddle.fleet.metrics.metric as metric
+import paddle.distributed.fleet.metrics.metric as metric
 from paddle.fluid.incubate.fleet.parameter_server.distribute_transpiler import fleet
 
 

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

@@ -24,7 +24,7 @@ class TestFleetMetaOptimizer(unittest.TestCase):
             "PADDLE_TRAINER_ENDPOINTS"] = "127.0.0.1:36001,127.0.0.1:36002"
 
     def test_pipeline_optimizer(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -49,7 +49,7 @@ class TestFleetMetaOptimizer(unittest.TestCase):
                 input=prediction, label=input_y)
             avg_cost = paddle.fluid.layers.mean(x=cost)
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.pipeline = True
         strategy.pipeline_configs = {'micro_batch': 2}
 

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

@@ -36,7 +36,7 @@ class TestFleetPrivateFunction(unittest.TestCase):
         thr = threading.Thread(target=init_server, args=(9292, ))
         thr.start()
 
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         ep = ["127.0.0.1:9292"]
         fleet.base.private_helper_function.wait_server_ready(ep)
 

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

@@ -26,7 +26,7 @@ class TestFleetRecomputeMetaOptimizer(unittest.TestCase):
                        "127.0.0.1:36001,127.0.0.2:36001"
 
     def test_recompute_optimizer(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -41,7 +41,7 @@ class TestFleetRecomputeMetaOptimizer(unittest.TestCase):
             input=prediction, label=input_y)
         avg_cost = paddle.fluid.layers.mean(x=cost)
 
-        strategy = paddle.fleet.DistributedStrategy()
+        strategy = paddle.distributed.fleet.DistributedStrategy()
         strategy.recompute = True
         strategy.recompute_configs = {"checkpoints": ["fc2"]}
 

python/paddle/fluid/tests/unittests/test_fleet_rolemaker_2.py

@@ -163,7 +163,7 @@ class TestCloudRoleMaker2(unittest.TestCase):
             data = "1 1 1 1\n"
             f.write(data)
 
-        dataset = paddle.fleet.DatasetFactory().create_dataset(
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
             "InMemoryDataset")
         dataset.set_filelist(["test_fleet_gloo_role_maker_1.txt"])
         dataset.set_use_var([show, label])

python/paddle/fluid/tests/unittests/test_fleet_rolemaker_4.py

@@ -40,9 +40,9 @@ class TestCloudRoleMaker(unittest.TestCase):
             from paddle.fluid.incubate.fleet.parameter_server.pslib import PSLib
             from paddle.fluid.incubate.fleet.base.role_maker import \
                 GeneralRoleMaker
-            from paddle.fleet.utils import KVHandler
-            from paddle.fleet.utils import KVServer
-            from paddle.fleet.utils import KVHTTPServer
+            from paddle.distributed.fleet.utils import KVHandler
+            from paddle.distributed.fleet.utils import KVServer
+            from paddle.distributed.fleet.utils import KVHTTPServer
         except:
             print("warning: no fleet, skip test_pslib_4")
             return

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

@@ -16,7 +16,7 @@
 from __future__ import print_function
 import os
 import unittest
-import paddle.fleet.base.role_maker as role_maker
+import paddle.distributed.fleet.base.role_maker as role_maker
 
 
 class TestRoleMakerBase(unittest.TestCase):

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

@@ -19,16 +19,17 @@ import os
 
 class TestFleetRuntime(unittest.TestCase):
     def test_fleet_runtime_base(self):
-        import paddle.fleet.runtime
-        base = paddle.fleet.runtime.runtime_base.RuntimeBase()
+        import paddle.distributed.fleet.runtime
+        base = paddle.distributed.fleet.runtime.runtime_base.RuntimeBase()
         base._run_worker()
         base._init_server()
         base._run_server()
         base._stop_worker()
 
     def test_fleet_collective_runtime(self):
-        import paddle.fleet.runtime
-        collective_runtime = paddle.fleet.runtime.CollectiveRuntime()
+        import paddle.distributed.fleet.runtime
+        collective_runtime = paddle.distributed.fleet.runtime.CollectiveRuntime(
+        )
         collective_runtime._init_worker()
         collective_runtime._run_worker()
         collective_runtime._init_worker()

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

@@ -22,8 +22,8 @@ import tempfile
 import os
 import sys
 from paddle.dataset.common import download, DATA_HOME
-from paddle.fleet.base.util_factory import fleet_util
-import paddle.fleet.base.role_maker as role_maker
+from paddle.distributed.fleet.base.util_factory import fleet_util
+import paddle.distributed.fleet.base.role_maker as role_maker
 
 
 class TestFleetUtil(unittest.TestCase):
@@ -34,7 +34,7 @@ class TestFleetUtil(unittest.TestCase):
     train_dir = os.path.join("fleet_util_data", "train_program")
 
     def test_util_base(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         util = fleet.UtilBase()
         strategy = fleet.DistributedStrategy()
         util._set_strategy(strategy)
@@ -42,7 +42,7 @@ class TestFleetUtil(unittest.TestCase):
         util._set_role_maker(role_maker)
 
     def test_util_factory(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         factory = fleet.base.util_factory.UtilFactory()
         strategy = fleet.DistributedStrategy()
         role_maker = None  # should be fleet.PaddleCloudRoleMaker()
@@ -55,7 +55,7 @@ class TestFleetUtil(unittest.TestCase):
         self.assertEqual(util.role_maker, None)
 
     def test_get_util(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
         import paddle.fluid.incubate.fleet.base.role_maker as role_maker
         role = role_maker.PaddleCloudRoleMaker(is_collective=True)
         fleet.init(role)
@@ -63,7 +63,7 @@ class TestFleetUtil(unittest.TestCase):
         self.assertEqual(default_util, None)
 
     def test_set_user_defined_util(self):
-        import paddle.fleet as fleet
+        import paddle.distributed.fleet as fleet
 
         class UserDefinedUtil(fleet.UtilBase):
             def __init__(self):
@@ -81,7 +81,7 @@ class TestFleetUtil(unittest.TestCase):
         self.assertEqual(user_id, 10)
 
     def test_fs(self):
-        from paddle.fleet.utils import LocalFS
+        from paddle.distributed.fleet.utils import LocalFS
         fs = LocalFS()
         dirs, files = fs.ls_dir("test_tmp")
         dirs, files = fs.ls_dir("./")

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

@@ -20,7 +20,7 @@ import os
 import sys
 import inspect
 
-from paddle.fleet.utils import LocalFS, FS, HDFSClient, FSTimeOut, FSFileExistsError, FSFileNotExistsError
+from paddle.distributed.fleet.utils import LocalFS, FS, HDFSClient, FSTimeOut, FSFileExistsError, FSFileNotExistsError
 
 
 class FSTest(unittest.TestCase):

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

@@ -19,7 +19,7 @@ from paddle.fluid.incubate.fleet.collective import CollectiveOptimizer, fleet
 import os
 import sys
 
-from paddle.fleet.utils import LocalFS, HDFSClient, FSTimeOut, FSFileExistsError, FSFileNotExistsError
+from paddle.distributed.fleet.utils import LocalFS, HDFSClient, FSTimeOut, FSFileExistsError, FSFileNotExistsError
 
 java_home = os.environ["JAVA_HOME"]
 

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

@@ -52,7 +52,7 @@ class TestDatasetWithStat(unittest.TestCase):
                 name=slot, shape=[1], dtype="int64", lod_level=1)
             slots_vars.append(var)
 
-        dataset = paddle.fleet.DatasetFactory().create_dataset(
+        dataset = paddle.distributed.fleet.DatasetFactory().create_dataset(
             "InMemoryDataset")
         dataset.set_batch_size(32)
         dataset.set_thread(3)

python/setup.py.in

@@ -145,14 +145,14 @@ packages=['paddle',
           'paddle.incubate',
           'paddle.incubate.complex',
           'paddle.incubate.complex.tensor',
-          'paddle.fleet',
-          'paddle.fleet.base',
-          'paddle.fleet.meta_optimizers',
-          'paddle.fleet.runtime',
-          'paddle.fleet.dataset',
-          'paddle.fleet.metrics',
-          'paddle.fleet.proto',
-          'paddle.fleet.utils',
+          'paddle.distributed.fleet',
+          'paddle.distributed.fleet.base',
+          'paddle.distributed.fleet.meta_optimizers',
+          'paddle.distributed.fleet.runtime',
+          'paddle.distributed.fleet.dataset',
+          'paddle.distributed.fleet.metrics',
+          'paddle.distributed.fleet.proto',
+          'paddle.distributed.fleet.utils',
           'paddle.framework',
           'paddle.jit',
           'paddle.fluid',
@@ -482,7 +482,7 @@ with redirect_stdout():
         },
         entry_points={
             'console_scripts': [
-                'fleetrun = paddle.fleet.launch:launch'
+                'fleetrun = paddle.distributed.fleet.launch:launch'
             ]
         }
     )