Fix typing inconsistency in return of _check_input (#9564)

Fix #9560

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

CHANGELOG.md

@@ -81,6 +81,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed metaclass conflict in `Dataset` ([#8999](https://github.com/pyg-team/pytorch_geometric/pull/8999))
 - Fixed import errors on `MessagePassing` modules with nested inheritance ([#8973](https://github.com/pyg-team/pytorch_geometric/pull/8973))
 - Fixed bug in multi XPU training ([#9456](https://github.com/pyg-team/pytorch_geometric/pull/9456))
+- Fixed TorchScript compilation error for `MessagePassing._check_input` on older torch versions ([#9564](https://github.com/pyg-team/pytorch_geometric/pull/9564))
 
 ### Removed
 

torch_geometric/nn/conv/message_passing.py

@@ -204,7 +204,7 @@ def __repr__(self) -> str:
     def _check_input(
         self,
         edge_index: Union[Tensor, SparseTensor],
-        size: Optional[Tuple[int, int]],
+        size: Optional[Tuple[Optional[int], Optional[int]]],
     ) -> List[Optional[int]]:
 
         if not torch.jit.is_scripting() and isinstance(edge_index, EdgeIndex):