Run mypy with PyTorch 1.12

.github/workflows/code-checks.yml

@@ -32,7 +32,7 @@ jobs:
 
     - name: Install dependencies
       run: |
-        pip install torch==1.11 --find-links https://download.pytorch.org/whl/cpu/torch_stable.html
+        pip install torch==1.12 --find-links https://download.pytorch.org/whl/cpu/torch_stable.html
         python ./requirements/pytorch/adjust-versions.py requirements/pytorch/extra.txt
         pip install -r requirements/pytorch/devel.txt --find-links https://download.pytorch.org/whl/cpu/torch_stable.html
         pip list

pyproject.toml

@@ -52,7 +52,6 @@ module = [
     "pytorch_lightning.callbacks.progress.rich_progress",
     "pytorch_lightning.callbacks.quantization",
     "pytorch_lightning.core.datamodule",
-    "pytorch_lightning.core.decorators",
     "pytorch_lightning.core.module",
     "pytorch_lightning.core.saving",
     "pytorch_lightning.demos.boring_classes",

src/pytorch_lightning/plugins/precision/fully_sharded_native_amp.py

@@ -23,7 +23,7 @@
 if _TORCH_GREATER_EQUAL_1_12:
     from torch.distributed.fsdp.fully_sharded_data_parallel import MixedPrecision
 else:
-    MixedPrecision = None
+    MixedPrecision = None  # type: ignore[misc,assignment]
 
 
 class FullyShardedNativeMixedPrecisionPlugin(ShardedNativeMixedPrecisionPlugin):

src/pytorch_lightning/strategies/fully_sharded_native.py

@@ -51,7 +51,7 @@
     )
     from torch.distributed.fsdp.wrap import enable_wrap
 else:
-    MixedPrecision = None
+    MixedPrecision = None  # type: ignore[misc,assignment]
     BackwardPrefetch = None  # type: ignore[misc,assignment]
     CPUOffload = None  # type: ignore[misc,assignment]
 

src/pytorch_lightning/strategies/launchers/multiprocessing.py

@@ -144,7 +144,7 @@ def _recover_results_in_main_process(self, worker_output: "_WorkerOutput", train
         # load last weights
         if worker_output.weights_path is not None:
             ckpt = self._strategy.checkpoint_io.load_checkpoint(worker_output.weights_path)
-            trainer.lightning_module.load_state_dict(ckpt)  # type: ignore[arg-type]
+            trainer.lightning_module.load_state_dict(ckpt)
             self._strategy.checkpoint_io.remove_checkpoint(worker_output.weights_path)
 
         trainer.state = worker_output.trainer_state

src/pytorch_lightning/utilities/cloud_io.py

@@ -22,14 +22,12 @@
 from fsspec.core import url_to_fs
 from fsspec.implementations.local import AbstractFileSystem
 
-from pytorch_lightning.utilities.types import _PATH
+from pytorch_lightning.utilities.types import _DEVICE, _PATH
 
 
 def load(
     path_or_url: Union[IO, _PATH],
-    map_location: Optional[
-        Union[str, Callable, torch.device, Dict[Union[str, torch.device], Union[str, torch.device]]]
-    ] = None,
+    map_location: Optional[Union[_DEVICE, Callable[[_DEVICE], _DEVICE], Dict[_DEVICE, _DEVICE]]] = None,
 ) -> Any:
     """Loads a checkpoint.
 
@@ -41,7 +39,10 @@ def load(
         # any sort of BytesIO or similar
         return torch.load(path_or_url, map_location=map_location)
     if str(path_or_url).startswith("http"):
-        return torch.hub.load_state_dict_from_url(str(path_or_url), map_location=map_location)
+        return torch.hub.load_state_dict_from_url(
+            str(path_or_url),
+            map_location=map_location,  # type: ignore[arg-type] # upstream annotation is not correct
+        )
     fs = get_filesystem(path_or_url)
     with fs.open(path_or_url, "rb") as f:
         return torch.load(f, map_location=map_location)