Fix some typos (persistbale, Retruns, etc.)(#61744)

co63oc · web-flow · commit 362be32b8b15 · 2024-02-19T16:10:53.000+08:00
diff --git a/python/paddle/distributed/io.py b/python/paddle/distributed/io.py
@@ -409,7 +409,7 @@ def save_persistables(executor, dirname, main_program=None, filename=None):
 
         dirname(str, optional): The saving directory path.
                             When you need to save the parameter to the memory, set it to None.
-        main_program(Program, optional): The program whose persistbale variables will
+        main_program(Program, optional): The program whose persistable variables will
                                          be saved. You can refer to
                                          :ref:`api_guide_Program_en` for more details.
                                          If it is None, the default main program will
diff --git a/python/paddle/distributed/parallel.py b/python/paddle/distributed/parallel.py
@@ -291,7 +291,7 @@ class DataParallel(layers.Layer):
         ``PyLayer`` is not supported in DataParallel. To solve problems of this kind,
         it's recommended to skip gradient synchronization among multiple cards by 'no_sync',
         and manually implement 'all_reduce' before model optimization. There is an example
-        showing specific implemetation processing.
+        showing specific implementation processing.
 
     Examples:
 
@@ -567,7 +567,7 @@ def state_dict(
             destination(dict, optional) : If provide, all the parameters and persistable buffers will be set to this dict . Default: None
             include_sublayers(bool, optional) : If true, also include the parameters and persistable buffers from sublayers. Default: True
 
-        Retruns:
+        Returns:
             dict: a dict contains all the parameters and persistable buffers.
 
         Examples:
@@ -1011,7 +1011,7 @@ def init_parallel_env():
         )
         return
     # NOTE(xiongkun): support cpu gloo only, add this environment variable to
-    #                 enable cpu only gloo prarllel training)
+    #                 enable cpu only gloo parallel training)
     backend = os.environ.get('PADDLE_DISTRI_BACKEND', 'auto')
     is_cpu_only = _is_cpuonly(backend)
     # 1. gpu xpu check, must be gpu or xpu,
@@ -1127,12 +1127,12 @@ def init_parallel_env():
         return group
 
     node_num = {i.split(":")[0] for i in parallel_env.trainer_endpoints}
-    # 3: init gloo context (step 1: httpsever start)
+    # 3: init gloo context (step 1: httpserver start)
     init_gloo = int(os.getenv("PADDLE_WITH_GLOO", "0"))
     if is_cpu_only or init_gloo or backend == "heter":
         ep_rank_0 = parallel_env.trainer_endpoints[0].split(":")
         manager = Manager()
-        # glboal dict to store status
+        # global dict to store status
         http_server_d = manager.dict()
         http_server_d["running"] = False
         if parallel_env.rank == 0:
@@ -1185,7 +1185,7 @@ def init_parallel_env():
     parallel_helper._init_parallel_ctx()
 
     # 5: init gloo context (step 2: gloo init)
-    # dividing init_gloo into two part beacause nccl and gloo
+    # dividing init_gloo into two part because nccl and gloo
     # are separately looking for free ports which sometimes
     # leads to port-conflict.
     if (is_cpu_only or backend == "heter") and parallel_env.rank == 0:
diff --git a/python/paddle/distributed/parallel_with_gloo.py b/python/paddle/distributed/parallel_with_gloo.py
@@ -95,7 +95,7 @@ def gloo_init_parallel_env(rank_id, rank_num, server_endpoint):
 
     assert (
         rank_num < 2
-    ) is False, "rank_num should greater than or equal to 2 for parallel environment initialzation."
+    ) is False, "rank_num should greater than or equal to 2 for parallel environment initialization."
 
     # init gloo context
     manager = Manager()
@@ -189,7 +189,7 @@ def gloo_barrier():
             ...     test_gloo_barrier_with_multiprocess(2)
     """
 
-    assert _global_gloo_ctx is not None, "gloo context is not initialzed."
+    assert _global_gloo_ctx is not None, "gloo context is not initialized."
     _global_gloo_ctx.barrier()
 
 
diff --git a/python/paddle/distributed/spawn.py b/python/paddle/distributed/spawn.py
@@ -145,16 +145,16 @@ def _get_node_ip(ips):
 
 def _get_subprocess_env_list(nprocs, options):
     # NOTE (xiongkun03) Why put backend deduction  here ?
-    # Becase _get_subprocess_env_list is used by many testcases.
-    # So for campability, we put backend deduction here
+    # Because _get_subprocess_env_list is used by many testcases.
+    # So for compatibility, we put backend deduction here
 
     # logic for handle backend option
     if 'backend' not in options or options['backend'] == 'auto':
         options['backend'] = _get_default_backend()
     check_backend(options['backend'])
     block_windows_and_macos(options['backend'])
 
-    # contruct processes env list
+    # construct processes env list
     processes_env_list = []
 
     # get args from kwargs
diff --git a/python/paddle/distributed/transpiler/collective.py b/python/paddle/distributed/transpiler/collective.py
@@ -352,7 +352,7 @@ def _insert_allreduce_ops(self):
                         )
                         offset += 1
 
-                    # As we search ops reversedly, we should insert c_allreduce_sum
+                    # As we search ops reversely, we should insert c_allreduce_sum
                     # op in the same way to keep the ring_id alternate
                     ring_id = (ring_id + 1) % self.nrings
                     block._insert_op(
@@ -517,7 +517,7 @@ def _transpile_startup_program(self):
         nodes_num = 0
         if len(self.endpoints) > 1:
             nodes_num = len({x.split(':')[0] for x in self.endpoints})
-        # diffent ip num is multi node
+        # different ip num is multi node
         if nodes_num > 1:
             self.nranks = nodes_num
             print("begin to _transpile_startup_program for multi-node")
@@ -834,7 +834,7 @@ def _insert_allgather_ops(self):
                         )
                         offset += 1
 
-                    # As we search ops reversedly, we should insert c_allgather
+                    # As we search ops reversely, we should insert c_allgather
                     # op in the same way to keep the ring_id alternate
                     ring_id = (ring_id + 1) % self.nrings
                     block._insert_op(
diff --git a/python/paddle/distributed/utils/moe_utils.py b/python/paddle/distributed/utils/moe_utils.py
@@ -29,7 +29,7 @@ def global_scatter(
     The global_count of the rank 0 is [2, 0, , ], rank 1 is [2, 0, ,](Due to the limited space, only the data calculated on rank 0 is shown here).
     In the global_scatter operator, local_count[i] represents sending local_count[i] data to the (i % n_expert)th expert of the (i // n_expert)th card,
     global_count[i] represents receiving global_count[i] data from the (i // n_expert)th card to the (i % n_expert)th expert of this card. The rank in the
-    figure respresent the rank of the current card in all cards.
+    figure represent the rank of the current card in all cards.
 
     The process of global_scatter sending data is as follows:
 
@@ -161,7 +161,7 @@ def global_gather(
     The global_count of the rank 0 is [2, 0, , ], rank 1 is [2, 0, ,](Due to the limited space, only the data calculated on rank 0 is shown here).
     In the global_gather operator, the meaning of the global_count and local_count is opposed to global_scatter, global_count[i] represents sending global_count[i] data to the (i % n_expert)th expert of the (i // n_expert)th card,
     local_count[i] represents receiving local_count[i] data from the (i // n_expert)th card to the (i % n_expert)th expert of this card. The data sent will be arranged according to the experts of each card.
-    The rank in the figure respresent the rank of the current card in all cards.
+    The rank in the figure represent the rank of the current card in all cards.
 
     The process of global_gather sending data is as follows:
 
