Allow training type plugin to delay optimizer creation (FSDP 2/n)

pytorch_lightning/accelerators/accelerator.py

@@ -85,7 +85,8 @@ def setup(self, trainer: 'Trainer', model: LightningModule) -> None:
             model: the LightningModule
         """
         self.setup_training_type_plugin(self.training_type_plugin, model)
-        self.setup_optimizers(trainer)
+        if not self.training_type_plugin.setup_optimizers_in_pre_dispatch:
+            self.setup_optimizers(trainer)
         self.setup_precision_plugin(self.precision_plugin)
 
     def start_training(self, trainer: 'Trainer') -> None:
@@ -97,12 +98,14 @@ def start_evaluating(self, trainer: 'Trainer') -> None:
     def start_predicting(self, trainer: 'Trainer') -> None:
         self.training_type_plugin.start_predicting(trainer)
 
-    def pre_dispatch(self) -> None:
+    def pre_dispatch(self, trainer: 'Trainer') -> None:
         """Hook to do something before the training/evaluation/prediction starts."""
         self.training_type_plugin.pre_dispatch()
+        if self.training_type_plugin.setup_optimizers_in_pre_dispatch:
+            self.setup_optimizers(trainer)
         self.precision_plugin.pre_dispatch()
 
-    def post_dispatch(self) -> None:
+    def post_dispatch(self, trainer: 'Trainer') -> None:
         """Hook to do something before the training/evaluation/prediction starts."""
         self.training_type_plugin.post_dispatch()
         self.precision_plugin.post_dispatch()

pytorch_lightning/plugins/training_type/training_type_plugin.py

@@ -182,3 +182,13 @@ def init_optimizers(self, trainer: "Trainer", model: LightningModule):
 
     def optimizer_step(self, optimizer: torch.optim.Optimizer, lambda_closure: Callable, **kwargs):
         optimizer.step(closure=lambda_closure, **kwargs)
+
+    @property
+    def setup_optimizers_in_pre_dispatch(self) -> bool:
+        """
+        Override to delay setting optimizers and schedulers till after dispatch.
+        This is useful when the `TrainingTypePlugin` requires operating on the wrapped accelerator model.
+        However this may break certain precision plugins such as APEX which require optimizers to be set.
+        Returns: If True, delay setup optimizers till pre_dispatch, else call within setup.
+        """
+        return False

pytorch_lightning/trainer/trainer.py

@@ -495,7 +495,7 @@ def fit(
         return self.accelerator.results or 1
 
     def pre_dispatch(self):
-        self.accelerator.pre_dispatch()
+        self.accelerator.pre_dispatch(self)
 
         # log hyper-parameters
         if self.logger is not None:
@@ -505,7 +505,7 @@ def pre_dispatch(self):
             self.logger.save()
 
     def post_dispatch(self):
-        self.accelerator.post_dispatch()
+        self.accelerator.post_dispatch(self)
         self.accelerator.teardown()
 
     def dispatch(self):

tests/accelerators/test_cpu.py

@@ -2,11 +2,12 @@
 
 import pytest
 import torch
-
+from pytorch_lightning import Trainer
 from pytorch_lightning.accelerators import CPUAccelerator
 from pytorch_lightning.plugins import SingleDevicePlugin
 from pytorch_lightning.plugins.precision import MixedPrecisionPlugin
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
+from tests.helpers.boring_model import BoringModel
 
 
 def test_unsupported_precision_plugins():
@@ -18,3 +19,35 @@ def test_unsupported_precision_plugins():
     )
     with pytest.raises(MisconfigurationException, match=r"amp \+ cpu is not supported."):
         accelerator.setup(trainer=trainer, model=model)
+
+
+@pytest.mark.parametrize("delay_dispatch", [True, False])
+def test_plugin_setup_optimizers_in_pre_dispatch(tmpdir, delay_dispatch):
+    """
+    Test when using a custom training type plugin that delays setup optimizers,
+    we do not call setup optimizers till ``pre_dispatch``.
+    """
+
+    class TestModel(BoringModel):
+        def on_fit_start(self):
+            if delay_dispatch:
+                # Ensure we haven't setup optimizers if we've delayed dispatch
+                assert len(self.trainer.optimizers) == 0
+            else:
+                assert len(self.trainer.optimizers) > 0
+
+        def on_fit_end(self):
+            assert len(self.trainer.optimizers) > 0
+
+    class CustomPlugin(SingleDevicePlugin):
+        @property
+        def setup_optimizers_in_pre_dispatch(self) -> bool:
+            return delay_dispatch
+
+    model = TestModel()
+    trainer = Trainer(
+        default_root_dir=tmpdir,
+        fast_dev_run=True,
+        plugins=CustomPlugin(device=torch.device("cpu"))
+    )
+    trainer.fit(model)