Improve typing for plugins

pyproject.toml

@@ -88,9 +88,6 @@ module = [
     "pytorch_lightning.plugins.environments.lsf_environment",
     "pytorch_lightning.plugins.environments.slurm_environment",
     "pytorch_lightning.plugins.environments.torchelastic_environment",
-    "pytorch_lightning.plugins.precision.deepspeed",
-    "pytorch_lightning.plugins.precision.native_amp",
-    "pytorch_lightning.plugins.precision.precision_plugin",
     "pytorch_lightning.plugins.training_type.ddp",
     "pytorch_lightning.plugins.training_type.ddp2",
     "pytorch_lightning.plugins.training_type.ddp_spawn",

pytorch_lightning/plugins/precision/deepspeed.py

@@ -49,7 +49,9 @@ def backward(self, model: "pl.LightningModule", closure_loss: Tensor, *args: Any
         deepspeed_engine: DeepSpeedEngine = model.trainer.model
         deepspeed_engine.backward(closure_loss, *args, **kwargs)
 
-    def _run_backward(self, tensor: Tensor, model: Module, *args: Any, **kwargs: Any) -> None:
+    def _run_backward(self, tensor: Tensor, model: Optional["DeepSpeedEngine"], *args: Any, **kwargs: Any) -> None:
+        if model is None:
+            raise ValueError("Please provide the model as input to `backward`.")
         model.backward(tensor, *args, **kwargs)
 
     def optimizer_step(

pytorch_lightning/plugins/precision/native_amp.py

@@ -25,9 +25,9 @@
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 
 if _TORCH_GREATER_EQUAL_1_10:
-    from torch import autocast
+    from torch import autocast as new_autocast
 else:
-    from torch.cuda.amp import autocast
+    from torch.cuda.amp import autocast as old_autocast
 
 
 class NativeMixedPrecisionPlugin(MixedPrecisionPlugin):
@@ -62,7 +62,7 @@ def pre_backward(self, model: "pl.LightningModule", closure_loss: torch.Tensor)
             closure_loss = self.scaler.scale(closure_loss)
         return super().pre_backward(model, closure_loss)
 
-    def _run_backward(self, tensor: Tensor, model: Module, *args: Any, **kwargs: Any) -> None:
+    def _run_backward(self, tensor: Tensor, model: Optional[Module], *args: Any, **kwargs: Any) -> None:
         if self.scaler is not None:
             tensor = self.scaler.scale(tensor)
         super()._run_backward(tensor, model, *args, **kwargs)
@@ -93,12 +93,12 @@ def optimizer_step(
             self.scaler.step(optimizer, **kwargs)
             self.scaler.update()
 
-    def autocast_context_manager(self) -> autocast:
+    def autocast_context_manager(self) -> Union["old_autocast", "new_autocast"]:
         if _TORCH_GREATER_EQUAL_1_10:
             # the dtype could be automatically inferred but we need to manually set it due to a bug upstream
             # https://github.com/pytorch/pytorch/issues/67233
-            return autocast(self.device, dtype=torch.bfloat16 if self.precision == "bf16" else torch.half)
-        return autocast()
+            return new_autocast(self.device, dtype=torch.bfloat16 if self.precision == "bf16" else torch.half)
+        return old_autocast()
 
     @contextmanager
     def forward_context(self) -> Generator[None, None, None]:

pytorch_lightning/plugins/precision/precision_plugin.py

@@ -147,7 +147,7 @@ def optimizer_step(
         """Hook to run the optimizer step."""
         if isinstance(model, pl.LightningModule):
             closure = partial(self._wrap_closure, model, optimizer, optimizer_idx, closure)
-        optimizer.step(closure=closure, **kwargs)
+        optimizer.step(closure=closure, **kwargs)  # type: ignore[call-arg]
 
     def _track_grad_norm(self, trainer: "pl.Trainer") -> None:
         if trainer.track_grad_norm == -1: