[Typing][B-42][BUAA] Add type annotations for `python/paddle/autograd…

…/py_layer.py` (PaddlePaddle#66328)

python/paddle/autograd/py_layer.py

@@ -12,18 +12,23 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, Any, Callable, Sequence, TypeVar
+
+from typing_extensions import Concatenate
+
 import paddle
 from paddle.base import core
 
-__all__ = []
+if TYPE_CHECKING:
+    from paddle import Tensor
 
 
-def with_metaclass(meta, *bases):
-    class impl(meta):
-        def __new__(cls, name, temp_bases, attrs):
-            return meta(name, bases, attrs)
+__all__ = []
+
 
-    return type.__new__(impl, "impl", (), {})
+_RetT = TypeVar('_RetT')
 
 
 class PyLayerContext:
@@ -52,7 +57,12 @@ class PyLayerContext:
             ...         return grad
     """
 
-    def save_for_backward(self, *tensors):
+    container: tuple[Tensor, ...]
+    not_inplace_tensors: tuple[Tensor, ...]
+    non_differentiable: tuple[Tensor, ...]
+    materialize_grads: bool
+
+    def save_for_backward(self, *tensors: Tensor) -> None:
         """
         Saves given tensors that backward need. Use ``saved_tensor`` in the `backward` to get the saved tensors.
 
@@ -90,7 +100,7 @@ def save_for_backward(self, *tensors):
         """
         self.container = tensors
 
-    def saved_tensor(self):
+    def saved_tensor(self) -> tuple[Tensor, ...]:
         """
         Get the tensors stored by ``save_for_backward``.
 
@@ -122,7 +132,7 @@ def saved_tensor(self):
         """
         return self.container
 
-    def mark_not_inplace(self, *args):
+    def mark_not_inplace(self, *args: Tensor) -> None:
         """
         Marks inputs as not inplace.
         This should be called at most once, only from inside the `forward` method,
@@ -163,7 +173,7 @@ def mark_not_inplace(self, *args):
         """
         self.not_inplace_tensors = args
 
-    def mark_non_differentiable(self, *args):
+    def mark_non_differentiable(self, *args: Tensor) -> None:
         """
         Marks outputs as non-differentiable.
         This should be called at most once, only from inside the `forward` method,
@@ -203,7 +213,7 @@ def mark_non_differentiable(self, *args):
         """
         self.non_differentiable = args
 
-    def set_materialize_grads(self, value: bool):
+    def set_materialize_grads(self, value: bool) -> None:
         """
         Sets whether to materialize output grad tensors. Default is True.
 
@@ -267,7 +277,7 @@ def __init__(cls, name, bases, attrs):
         return super().__init__(name, bases, attrs)
 
 
-class PyLayer(with_metaclass(PyLayerMeta, core.eager.PyLayer, PyLayerContext)):
+class PyLayer(core.eager.PyLayer, PyLayerContext, metaclass=PyLayerMeta):
     """
     Paddle implements Python custom operators on the PaddlePaddle framework by creating a subclass of
     ``PyLayer``, which must comply with the following rules:
@@ -322,7 +332,9 @@ class PyLayer(with_metaclass(PyLayerMeta, core.eager.PyLayer, PyLayerContext)):
     """
 
     @staticmethod
-    def forward(ctx, *args, **kwargs):
+    def forward(
+        ctx: PyLayerContext, *args: Any, **kwargs: Any
+    ) -> Tensor | Sequence[Tensor]:
         """
         It is to be overloaded by subclasses. It must accept a object of :ref:`api_paddle_autograd_PyLayerContext` as
         the first argument, followed by any number of arguments (tensors or other types).
@@ -361,7 +373,7 @@ def forward(ctx, *args, **kwargs):
         )
 
     @staticmethod
-    def backward(ctx, *args):
+    def backward(ctx: PyLayerContext, *args: Any) -> Tensor | Sequence[Tensor]:
         """
         This is a function to calculate the gradient. It is to be overloaded by subclasses.
         It must accept a object of :ref:`api_paddle_autograd_PyLayerContext` as the first
@@ -402,8 +414,10 @@ def backward(ctx, *args):
         )
 
 
-def once_differentiable(backward):
-    def wrapper(ctx, *args):
+def once_differentiable(
+    backward: Callable[Concatenate[PyLayerContext, ...], _RetT]
+) -> Callable[Concatenate[PyLayerContext, ...], _RetT]:
+    def wrapper(ctx: PyLayerContext, *args: Any) -> _RetT:
         with paddle.base.dygraph.no_grad():
             outputs = backward(ctx, *args)
         return outputs