make block more private to reduce conflicts with module

python/oneflow/nn/graph/block.py

@@ -203,7 +203,7 @@ class ModuleBlock(Block):
         # that hooks of nn.Modules are ignored. It is not recommended
         # to use hooks of nn.Module in nn.Graph for the moment.
         # result = self._origin.__class__.__call__(self, *args)
-        result = self._forward(*args)
+        result = self.__block_forward(*args)
 
         outputs = ()
         if not (type(result) is tuple or type(result) is list):
@@ -231,17 +231,17 @@ class ModuleBlock(Block):
 
         return result
 
-    def _forward(self, *args):
+    def __block_forward(self, *args):
         self._is_executing_forward = True
-        args = self._pre_forward_mapping_out_scope(*args)
+        args = self.__pre_forward_mapping_out_scope(*args)
         with self.scope_context():
             result = self._origin.__class__.forward(self, *args)
-        result = self._post_forward_mapping_out_scope(result)
+        result = self.__post_forward_mapping_out_scope(result)
         result = seq_to_func_return(result)
         self._is_executing_forward = False
         return result
 
-    def _pre_forward_mapping_out_scope(self, *args):
+    def __pre_forward_mapping_out_scope(self, *args):
         # Insert identity op when doing activation checkpointing or pipeline execution.
         # Identity op outside activation checkpointing scope will be the endpoint of an activation checkpointing segment.
         # Identity op as the first op of a pipeline stage will make backward op depends on the identity op within the stage,
@@ -254,11 +254,13 @@ class ModuleBlock(Block):
                 assert isinstance(t, Tensor)
                 return oneflow._C.identity(t)
 
-            args = self._mapping_io("input", insert_identity, "insert_identity", *args,)
+            args = self.__mapping_io(
+                "input", insert_identity, "insert_identity", *args,
+            )
 
         return args
 
-    def _post_forward_mapping_out_scope(self, *args):
+    def __post_forward_mapping_out_scope(self, *args):
         # Insert identity op when doing activation checkpointing or pipeline execution.
         if self.config.activation_checkpointing or (
             self.config.stage_id is not None and self.config.stage_id >= 0
@@ -268,7 +270,7 @@ class ModuleBlock(Block):
                 assert isinstance(t, Tensor)
                 return oneflow._C.identity(t)
 
-            args = self._mapping_io(
+            args = self.__mapping_io(
                 "output", insert_identity, "insert_identity", *args,
             )
         return args
@@ -298,7 +300,7 @@ class ModuleBlock(Block):
                 for m in module.modules(memo):
                     yield m
 
-    def _mapping_io(self, io_type, func, func_desc, *args):
+    def __mapping_io(self, io_type, func, func_desc, *args):
         assert isinstance(func_desc, str)
         assert io_type in ("input", "output")
         mapped_args = []
@@ -311,7 +313,7 @@ class ModuleBlock(Block):
             if isinstance(arg, list):
                 seq_args = list()
                 for i in range(len(arg)):
-                    is_tensor, name, repr_str = self._io_tensor_check_and_gen(
+                    is_tensor, name, repr_str = self.__io_tensor_check_and_gen(
                         arg[i], io_type, idx, i
                     )
                     if is_tensor:
@@ -330,7 +332,7 @@ class ModuleBlock(Block):
                         seq_args.append(arg[i])
                 mapped_args.append(seq_args)
             elif isinstance(arg, Tensor):
-                is_tensor, name, repr_str = self._io_tensor_check_and_gen(
+                is_tensor, name, repr_str = self.__io_tensor_check_and_gen(
                     arg, io_type, idx
                 )
                 assert is_tensor
@@ -341,7 +343,7 @@ class ModuleBlock(Block):
                     f"{repr_str} is a Tensor, {func_desc} transformation has been done.",
                 )
             else:
-                is_tensor, name, repr_str = self._io_tensor_check_and_gen(
+                is_tensor, name, repr_str = self.__io_tensor_check_and_gen(
                     arg, io_type, idx
                 )
                 assert not is_tensor
@@ -354,7 +356,7 @@ class ModuleBlock(Block):
 
         return tuple(mapped_args)
 
-    def _io_tensor_check_and_gen(self, item, io_type, idx, second_idx=None):
+    def __io_tensor_check_and_gen(self, item, io_type, idx, second_idx=None):
         assert io_type in ("input", "output")
         name = (
             "_"
@@ -383,7 +385,7 @@ class ModuleBlock(Block):
             )
             return False, name, repr_str
 
-    def _members(self, get_members_fn, recurse=True) -> Iterator["Block"]:
+    def __members(self, get_members_fn, recurse=True) -> Iterator["Block"]:
         assert self._type == BlockType.MODULE
         memo = set()
         modules = self.modules() if recurse else [self]
@@ -397,13 +399,13 @@ class ModuleBlock(Block):
 
     def parameters(self, recurse: bool = True) -> Iterator["Block"]:
         assert self._type == BlockType.MODULE
-        gen = self._members(lambda module: module._parameters.items(), recurse=recurse)
+        gen = self.__members(lambda module: module._parameters.items(), recurse=recurse)
         for elem in gen:
             yield elem
 
     def buffers(self, recurse: bool = True) -> Iterator["Block"]:
         assert self._type == BlockType.MODULE
-        gen = self._members(lambda module: module._buffers.items(), recurse=recurse)
+        gen = self.__members(lambda module: module._buffers.items(), recurse=recurse)
         for elem in gen:
             yield elem