Fix initialization of optimizers in DDP Strategy (#11952)

Co-authored-by: Rohit Gupta <[email protected]>
Co-authored-by: Carlos Mocholí <[email protected]>

Author: 3 people

CHANGELOG.md

@@ -30,6 +30,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - Fixed torchelastic detection with non-distributed installations ([#13142](https://github.com/PyTorchLightning/pytorch-lightning/pull/13142))
 - Fixed logging's step values when multiple dataloaders are used during evaluation ([#12184](https://github.com/PyTorchLightning/pytorch-lightning/pull/12184))
 - Fixed epoch logging on train epoch end ([#13025](https://github.com/PyTorchLightning/pytorch-lightning/pull/13025))
+- Fixed `DDPStrategy` and `DDPSpawnStrategy` to initialize optimizers only after moving the module to the device ([#11952](https://github.com/PyTorchLightning/pytorch-lightning/pull/11952))
 
 
 ## [1.6.3] - 2022-05-03

pytorch_lightning/strategies/bagua.py

@@ -16,9 +16,11 @@
 from pytorch_lightning.plugins.precision import PrecisionPlugin
 from pytorch_lightning.strategies.ddp import DDPStrategy
 from pytorch_lightning.strategies.strategy import TBroadcast
+from pytorch_lightning.trainer.states import TrainerFn
 from pytorch_lightning.utilities.distributed import ReduceOp
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from pytorch_lightning.utilities.imports import _BAGUA_AVAILABLE
+from pytorch_lightning.utilities.optimizer import optimizers_to_device
 from pytorch_lightning.utilities.seed import reset_seed
 
 if _BAGUA_AVAILABLE:
@@ -152,6 +154,33 @@ def _set_node_environment_variables(self) -> None:
         os.environ["WORLD_SIZE"] = str(self.world_size)
         os.environ["LOCAL_RANK"] = str(self.local_rank)
 
+    def setup(self, trainer: "pl.Trainer") -> None:
+        self._rank_0_will_call_children_scripts = self.broadcast(self._rank_0_will_call_children_scripts)
+        if self._should_run_deadlock_detection():
+            self._share_information_to_prevent_deadlock()
+
+        self.accelerator.setup(trainer)
+
+        # move the model to the correct device
+        self.model_to_device()
+
+        trainer_fn = trainer.state.fn
+
+        if trainer_fn == TrainerFn.FITTING:
+            if self._layer_sync and self.model:
+                self.model = self._layer_sync.apply(self.model)
+
+        self.setup_precision_plugin()
+
+        if trainer_fn == TrainerFn.FITTING:
+            # set up optimizers after the module has been moved to the device
+            # but before the module has been wrapped
+            self.setup_optimizers(trainer)
+            optimizers_to_device(self.optimizers, self.root_device)
+
+            # skip wrapping the model if we are not fitting as no gradients need to be exchanged
+            self._configure_bagua_model(trainer)
+
     def _check_qadam_optimizer(self) -> None:
         has_qadam_optimizer = any([isinstance(opt, QAdamOptimizer) for opt in self.optimizers])
 
@@ -160,13 +189,12 @@ def _check_qadam_optimizer(self) -> None:
 
         self._bagua_kwargs["q_adam_optimizer"] = self.optimizers[0]
 
-    def configure_ddp(self) -> None:
+    def _configure_bagua_model(self, trainer: "pl.Trainer") -> None:
         model = LightningBaguaModule(self.model)  # type: ignore[arg-type]
         self._model = self._setup_model(model)
 
         # start the background communication for async algorithm
-        assert self.lightning_module.trainer is not None
-        if self.lightning_module.trainer.training and self._bagua_algorithm == "async":
+        if trainer.training and self._bagua_algorithm == "async":
             self.model.bagua_algorithm.resume(self.model)  # type: ignore
 
     def _setup_model(self, model: Module) -> BaguaDistributedDataParallel:

pytorch_lightning/strategies/ddp.py

@@ -56,6 +56,7 @@
     _TORCH_GREATER_EQUAL_1_10,
     _TORCH_GREATER_EQUAL_1_11,
 )
+from pytorch_lightning.utilities.optimizer import optimizers_to_device
 from pytorch_lightning.utilities.rank_zero import rank_zero_info, rank_zero_only, rank_zero_warn
 from pytorch_lightning.utilities.seed import reset_seed
 from pytorch_lightning.utilities.types import STEP_OUTPUT
@@ -152,24 +153,37 @@ def setup_environment(self) -> None:
         super().setup_environment()
 
     def setup(self, trainer: "pl.Trainer") -> None:
-        super().setup(trainer)
         # share ddp pids to all processes
         self._rank_0_will_call_children_scripts = self.broadcast(self._rank_0_will_call_children_scripts)
         if self._should_run_deadlock_detection():
             self._share_information_to_prevent_deadlock()
 
+        self.accelerator.setup(trainer)
+
         # move the model to the correct device
         self.model_to_device()
 
         # skip wrapping the model if we are not fitting as no gradients need to be exchanged
         trainer_fn = trainer.state.fn
-        if trainer_fn != TrainerFn.FITTING:
-            return
 
-        if self._layer_sync:
-            self.model = self._layer_sync.apply(self.model)
+        if trainer_fn == TrainerFn.FITTING:
+            if self._layer_sync:
+                self.model = self._layer_sync.apply(self.model)
+
+        self.setup_precision_plugin()
+
+        if trainer_fn == TrainerFn.FITTING:
+            self.configure_ddp()
 
-        self.configure_ddp()
+            # set up optimizers after the wrapped module has been moved to the device
+            self.setup_optimizers(trainer)
+            optimizers_to_device(self.optimizers, self.root_device)
+
+        if _TORCH_GREATER_EQUAL_1_10 and trainer_fn == TrainerFn.FITTING:
+            import torch.distributed.algorithms.ddp_comm_hooks.post_localSGD_hook as post_localSGD
+
+            if isinstance(self._ddp_comm_state, post_localSGD.PostLocalSGDState):
+                self._enable_model_averaging()
 
     def _setup_model(self, model: Module) -> DistributedDataParallel:
         """Wraps the model into a :class:`~torch.nn.parallel.distributed.DistributedDataParallel` module."""
@@ -223,12 +237,6 @@ def _register_ddp_hooks(self) -> None:
                 ddp_comm_wrapper=self._ddp_comm_wrapper,
             )
 
-            if _TORCH_GREATER_EQUAL_1_10 and self.lightning_module.trainer.state.fn == TrainerFn.FITTING:
-                import torch.distributed.algorithms.ddp_comm_hooks.post_localSGD_hook as post_localSGD
-
-                if isinstance(self._ddp_comm_state, post_localSGD.PostLocalSGDState):
-                    self._enable_model_averaging()
-
     def _enable_model_averaging(self) -> None:
         # Only called when PyTorch version >= 1.10
         log.detail(f"{self.__class__.__name__}: reinitializing optimizers with post localSGD")

pytorch_lightning/strategies/ddp_spawn.py

@@ -42,6 +42,7 @@
     sync_ddp_if_available,
 )
 from pytorch_lightning.utilities.imports import _TORCH_GREATER_EQUAL_1_11
+from pytorch_lightning.utilities.optimizer import optimizers_to_device
 from pytorch_lightning.utilities.rank_zero import rank_zero_info, rank_zero_only
 from pytorch_lightning.utilities.seed import reset_seed
 from pytorch_lightning.utilities.types import STEP_OUTPUT
@@ -122,20 +123,22 @@ def _configure_launcher(self):
 
     def setup(self, trainer: "pl.Trainer") -> None:
         os.environ["MASTER_PORT"] = str(self.cluster_environment.main_port)
-        super().setup(trainer)
+
+        self.accelerator.setup(trainer)
 
         # move the model to the correct device
         self.model_to_device()
 
-        trainer_fn = self.lightning_module.trainer.state.fn
-        if trainer_fn != TrainerFn.FITTING:
-            return
+        # skip wrapping the model if we are not fitting as no gradients need to be exchanged
+        trainer_fn = trainer.state.fn
+        if trainer_fn == TrainerFn.FITTING:
+            if self._layer_sync:
+                self.model = self._layer_sync.apply(self.model)
 
-        if self._layer_sync:
-            self.model = self._layer_sync.apply(self.model)
+        self.setup_precision_plugin()
 
-        # skip wrapping the model if we are not fitting as no gradients need to be exchanged
-        self.configure_ddp()
+        if trainer_fn == TrainerFn.FITTING:
+            self.configure_ddp()
 
     def _setup_model(self, model: Module) -> DistributedDataParallel:
         """Wraps the model into a :class:`~torch.nn.parallel.distributed.DistributedDataParallel` module."""
@@ -186,6 +189,10 @@ def configure_ddp(self) -> None:
         self.model = self._setup_model(LightningDistributedModule(self.model))
         self._register_ddp_hooks()
 
+        # set up optimizers after the wrapped module has been moved to the device
+        self.setup_optimizers(self.lightning_module.trainer)
+        optimizers_to_device(self.optimizers, self.root_device)
+
     def determine_ddp_device_ids(self):
         if self.root_device.type == "cpu":
             return None

pytorch_lightning/strategies/fully_sharded.py

@@ -139,14 +139,16 @@ def setup_distributed(self) -> None:
     def setup(self, trainer: "pl.Trainer") -> None:
         self.accelerator.setup(trainer)
 
-        if trainer.state.fn == TrainerFn.FITTING and self._layer_sync:
-            self.model = self._layer_sync.apply(self.model)
+        if trainer.state.fn == TrainerFn.FITTING:
+            self.setup_optimizers(trainer)
+            optimizers_to_device(self.optimizers, self.root_device)
 
+            if self._layer_sync:
+                self.model = self._layer_sync.apply(self.model)
+
+        self.setup_precision_plugin()
         self.configure_ddp()
         self.barrier()
-        self.setup_optimizers(trainer)
-        optimizers_to_device(self.optimizers, self.root_device)
-        self.setup_precision_plugin()
 
     @contextlib.contextmanager
     def model_sharded_context(self) -> Generator:
@@ -183,6 +185,9 @@ def configure_ddp(self) -> None:
             # (TODO: need to figure out solution)
             self.model_to_device()
 
+        # setup optimizers after fully sharded has wrapped the lightning module
+        self.setup_optimizers(self.lightning_module.trainer)
+
     def model_to_device(self) -> None:
         log.detail(f"{self.__class__.__name__}: moving model to device [{self.root_device}]...")
         # ensure we update the device type in the lightning module

pytorch_lightning/strategies/sharded.py

@@ -25,6 +25,7 @@
 from pytorch_lightning.utilities.enums import PrecisionType
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from pytorch_lightning.utilities.imports import _FAIRSCALE_AVAILABLE, _FAIRSCALE_OSS_FP16_BROADCAST_AVAILABLE
+from pytorch_lightning.utilities.optimizer import optimizers_to_device
 from pytorch_lightning.utilities.rank_zero import rank_zero_only
 
 if _FAIRSCALE_AVAILABLE:
@@ -40,16 +41,41 @@ class DDPShardedStrategy(DDPStrategy):
     strategy_name = "ddp_sharded"
     _REDUCE_BUFFER_SIZE_DEFAULT: int = 2**23  # 8M
 
-    def configure_ddp(self) -> None:
-        trainer = self.lightning_module.trainer
-        if "reduce_buffer_size" not in self._ddp_kwargs:
-            # For multi-node training, enabling bucketing will improve performance.
-            self._ddp_kwargs["reduce_buffer_size"] = self._REDUCE_BUFFER_SIZE_DEFAULT if self.num_nodes > 1 else 0
+    def setup(self, trainer: "pl.Trainer") -> None:
+        # share ddp pids to all processes
+        self._rank_0_will_call_children_scripts = self.broadcast(self._rank_0_will_call_children_scripts)
+        if self._should_run_deadlock_detection():
+            self._share_information_to_prevent_deadlock()
+
+        self.accelerator.setup(trainer)
+
+        # move the model to the correct device
+        self.model_to_device()
+
+        # skip wrapping the model if we are not fitting as no gradients need to be exchanged
+        trainer_fn = trainer.state.fn
+        if trainer_fn == TrainerFn.FITTING:
+            if self._layer_sync:
+                self.model = self._layer_sync.apply(self.model)
 
+        self.setup_precision_plugin()
+
+        if trainer_fn == TrainerFn.FITTING:
+            self.configure_ddp()
+
+    def configure_ddp(self) -> None:
+        self._set_ddp_kwargs()
+        self.setup_optimizers(self.model.trainer)
         self.model, self.optimizers = self._setup_model_and_optimizers(
             model=LightningShardedDataParallel(self.model),
-            optimizers=trainer.optimizers,
+            optimizers=self.optimizers,
         )
+        optimizers_to_device(self.optimizers, self.root_device)
+
+    def _set_ddp_kwargs(self) -> None:
+        if "reduce_buffer_size" not in self._ddp_kwargs:
+            # For multi-node training, enabling bucketing will improve performance.
+            self._ddp_kwargs["reduce_buffer_size"] = self._REDUCE_BUFFER_SIZE_DEFAULT if self.num_nodes > 1 else 0
 
     def _setup_model_and_optimizers(self, model: Module, optimizers: List[Optimizer]) -> Tuple[Module, List[Optimizer]]:
         """Wraps the model and optimizers with fairscale components.
@@ -62,6 +88,12 @@ def _setup_model_and_optimizers(self, model: Module, optimizers: List[Optimizer]
         model = ShardedDataParallel(model, sharded_optimizer=optimizers, **self._ddp_kwargs)
         return model, optimizers
 
+    def _wrap_optimizers(self, optimizers: List[Optimizer]) -> List["OSS"]:
+        if self.model is not None and self.model.trainer.state.fn != TrainerFn.FITTING:
+            return optimizers
+
+        return self._reinit_optimizers_with_oss(optimizers)
+
     def _reinit_optimizers_with_oss(self, optimizers: List[Union[Optimizer, LightningOptimizer]]) -> List["OSS"]:
         for x, optimizer in enumerate(optimizers):
             if isinstance(optimizer, LightningOptimizer):
@@ -79,12 +111,6 @@ def _reinit_optimizers_with_oss(self, optimizers: List[Union[Optimizer, Lightnin
                 del optimizer
         return optimizers
 
-    def _wrap_optimizers(self, optimizers: List[Optimizer]) -> List["OSS"]:
-        if self.model is not None and self.model.trainer.state.fn != TrainerFn.FITTING:
-            return optimizers
-
-        return self._reinit_optimizers_with_oss(optimizers)
-
     def optimizer_state(self, optimizer: "OSS") -> Optional[dict]:
         if isinstance(optimizer, LightningOptimizer):
             optimizer = optimizer._optimizer

pytorch_lightning/strategies/sharded_spawn.py

@@ -23,6 +23,7 @@
 from pytorch_lightning.trainer.states import TrainerFn
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from pytorch_lightning.utilities.imports import _FAIRSCALE_AVAILABLE
+from pytorch_lightning.utilities.optimizer import optimizers_to_device
 from pytorch_lightning.utilities.rank_zero import rank_zero_only
 
 if _FAIRSCALE_AVAILABLE:
@@ -38,9 +39,12 @@ class DDPSpawnShardedStrategy(DDPSpawnStrategy):
     strategy_name = "ddp_sharded_spawn"
 
     def configure_ddp(self) -> None:
+        # set up optimizers after the wrapped module has been moved to the device
+        self.setup_optimizers(self.lightning_module.trainer)
         self.model, self.optimizers = self._setup_model_and_optimizers(
             model=LightningShardedDataParallel(self.model), optimizers=self.optimizers
         )
+        optimizers_to_device(self.optimizers, self.root_device)
 
     def _setup_model_and_optimizers(self, model: Module, optimizers: List[Optimizer]) -> Tuple[Module, List[Optimizer]]:
         """Wraps the model and optimizers with fairscale components.

pytorch_lightning/strategies/tpu_spawn.py

@@ -26,6 +26,7 @@
 from pytorch_lightning.strategies.ddp_spawn import DDPSpawnStrategy
 from pytorch_lightning.strategies.launchers.xla_spawn import _XLASpawnLauncher
 from pytorch_lightning.trainer.connectors.data_connector import DataConnector
+from pytorch_lightning.trainer.states import TrainerFn
 from pytorch_lightning.utilities import _TPU_AVAILABLE, find_shared_parameters, set_shared_parameters
 from pytorch_lightning.utilities.data import has_len
 from pytorch_lightning.utilities.distributed import ReduceOp
@@ -126,9 +127,6 @@ def _configure_launcher(self):
     def setup(self, trainer: "pl.Trainer") -> None:
         self.start_method = "fork"
         self.accelerator.setup(trainer)
-        self.setup_optimizers(trainer)
-        self.setup_precision_plugin()
-        optimizers_to_device(self.optimizers, self.root_device)
 
         if self.debug:
             os.environ["PT_XLA_DEBUG"] = str(1)
@@ -140,8 +138,11 @@ def setup(self, trainer: "pl.Trainer") -> None:
         else:
             set_shared_parameters(self.model.module, shared_params)
 
-        self.setup_optimizers(trainer)
-        self.precision_plugin.connect(self.model, None, None)
+        self.setup_precision_plugin()
+
+        if trainer.state.fn == TrainerFn.FITTING:
+            self.setup_optimizers(trainer)
+            optimizers_to_device(self.optimizers, self.root_device)
 
     def _setup_model(self, model: Module) -> Module:
         return model

tests/strategies/test_bagua_strategy.py

@@ -85,9 +85,9 @@ def test_configuration(algorithm, tmpdir):
     ), mock.patch("bagua.torch_api.communication.is_initialized", return_value=True):
         if algorithm == "qadam":
             with pytest.raises(MisconfigurationException, match="Bagua QAdam can only accept one QAdamOptimizer"):
-                trainer.strategy.configure_ddp()
+                trainer.strategy._configure_bagua_model(trainer)
         else:
-            trainer.strategy.configure_ddp()
+            trainer.strategy._configure_bagua_model(trainer)
 
 
 @RunIf(min_gpus=1, bagua=True)
@@ -109,7 +109,7 @@ def test_qadam_configuration(tmpdir):
     with mock.patch(
         "bagua.torch_api.data_parallel.bagua_distributed.BaguaDistributedDataParallel.__init__", return_value=None
     ), mock.patch("bagua.torch_api.communication.is_initialized", return_value=True):
-        trainer.strategy.configure_ddp()
+        trainer.strategy._configure_bagua_model(trainer)
 
 
 def test_bagua_not_available(monkeypatch):