Update strategy import statements (#11231)

docs/source/advanced/advanced_gpu.rst

@@ -291,7 +291,7 @@ Below we show an example of running `ZeRO-Offload <https://www.deepspeed.ai/tuto
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
 
     model = MyModel()
     trainer = Trainer(gpus=4, strategy="deepspeed_stage_2_offload", precision=16)
@@ -310,7 +310,7 @@ You can also modify the ZeRO-Offload parameters via the plugin as below.
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
 
     model = MyModel()
     trainer = Trainer(
@@ -335,7 +335,7 @@ For even more speed benefit, DeepSpeed offers an optimized CPU version of ADAM c
 
     import pytorch_lightning
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
     from deepspeed.ops.adam import DeepSpeedCPUAdam
 
 
@@ -379,7 +379,7 @@ Also please have a look at our :ref:`deepspeed-zero-stage-3-tips` which contains
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
     from deepspeed.ops.adam import FusedAdam
 
 
@@ -403,7 +403,7 @@ You can also use the Lightning Trainer to run predict or evaluate with DeepSpeed
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
 
 
     class MyModel(pl.LightningModule):
@@ -429,7 +429,7 @@ This reduces the time taken to initialize very large models, as well as ensure w
 
     import torch.nn as nn
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
     from deepspeed.ops.adam import FusedAdam
 
 
@@ -467,7 +467,7 @@ DeepSpeed ZeRO Stage 3 Offloads optimizer state, gradients to the host CPU to re
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
 
     # Enable CPU Offloading
     model = MyModel()
@@ -496,7 +496,7 @@ Additionally, DeepSpeed supports offloading to NVMe drives for even larger model
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
 
     # Enable CPU Offloading
     model = MyModel()
@@ -541,7 +541,7 @@ This saves memory when training larger models, however requires using a checkpoi
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
     import deepspeed
 
 
@@ -564,7 +564,7 @@ This saves memory when training larger models, however requires using a checkpoi
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
     import deepspeed
 
 
@@ -644,7 +644,7 @@ In some cases you may want to define your own DeepSpeed Config, to access all pa
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
 
     deepspeed_config = {
         "zero_allow_untested_optimizer": True,
@@ -687,7 +687,7 @@ We support taking the config as a json formatted file:
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
 
     model = MyModel()
     trainer = Trainer(gpus=4, strategy=DeepSpeedStrategy("/path/to/deepspeed_config.json"), precision=16)
@@ -722,7 +722,7 @@ This can reduce peak memory usage and throughput as saved memory will be equal t
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DDPStrategy
+    from pytorch_lightning.strategies import DDPStrategy
 
     model = MyModel()
     trainer = Trainer(gpus=4, strategy=DDPStrategy(gradient_as_bucket_view=True))
@@ -741,7 +741,7 @@ Enable `FP16 Compress Hook for multi-node throughput improvement <https://pytorc
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DDPStrategy
+    from pytorch_lightning.strategies import DDPStrategy
     from torch.distributed.algorithms.ddp_comm_hooks import (
         default_hooks as default,
         powerSGD_hook as powerSGD,
@@ -760,7 +760,7 @@ Enable `PowerSGD for multi-node throughput improvement <https://pytorch.org/docs
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DDPStrategy
+    from pytorch_lightning.strategies import DDPStrategy
     from torch.distributed.algorithms.ddp_comm_hooks import powerSGD_hook as powerSGD
 
     model = MyModel()
@@ -786,7 +786,7 @@ Combine hooks for accumulated benefit:
 .. code-block:: python
 
     from pytorch_lightning import Trainer
-    from pytorch_lightning.plugins import DDPStrategy
+    from pytorch_lightning.strategies import DDPStrategy
     from torch.distributed.algorithms.ddp_comm_hooks import (
         default_hooks as default,
         powerSGD_hook as powerSGD,

docs/source/advanced/ipu.rst

@@ -61,7 +61,7 @@ You can also use pure 16-bit training, where the weights are also in 16-bit prec
 .. code-block:: python
 
     import pytorch_lightning as pl
-    from pytorch_lightning.plugins import IPUStrategy
+    from pytorch_lightning.strategies import IPUStrategy
 
     model = MyLightningModule()
     model = model.half()
@@ -80,7 +80,7 @@ IPUs provide further optimizations to speed up training. By using the ``IPUStrat
 .. code-block:: python
 
     import pytorch_lightning as pl
-    from pytorch_lightning.plugins import IPUStrategy
+    from pytorch_lightning.strategies import IPUStrategy
 
     model = MyLightningModule()
     trainer = pl.Trainer(ipus=8, strategy=IPUStrategy(device_iterations=32))
@@ -92,7 +92,7 @@ Note that by default we return the last device iteration loss. You can override
 
     import poptorch
     import pytorch_lightning as pl
-    from pytorch_lightning.plugins import IPUStrategy
+    from pytorch_lightning.strategies import IPUStrategy
 
     model = MyLightningModule()
     inference_opts = poptorch.Options()
@@ -121,7 +121,7 @@ Lightning supports dumping all reports to a directory to open using the tool.
 .. code-block:: python
 
     import pytorch_lightning as pl
-    from pytorch_lightning.plugins import IPUStrategy
+    from pytorch_lightning.strategies import IPUStrategy
 
     model = MyLightningModule()
     trainer = pl.Trainer(ipus=8, strategy=IPUStrategy(autoreport_dir="report_dir/"))

docs/source/advanced/plugins_registry.rst

@@ -25,7 +25,7 @@ Additionally, you can pass your custom registered training type plugins to the `
 
 .. code-block:: python
 
-    from pytorch_lightning.plugins import DDPStrategy, TrainingTypePluginsRegistry, CheckpointIO
+    from pytorch_lightning.strategies import DDPStrategy, TrainingTypePluginsRegistry, CheckpointIO
 
 
     class CustomCheckpointIO(CheckpointIO):

docs/source/advanced/training_tricks.rst

@@ -178,7 +178,7 @@ For example, when training Graph Neural Networks, a common strategy is to load t
 
 A simple way to prevent redundant dataset replicas is to rely on :obj:`torch.multiprocessing` to share the `data automatically between spawned processes via shared memory <https://pytorch.org/docs/stable/notes/multiprocessing.html>`_.
 For this, all data pre-loading should be done on the main process inside :meth:`DataModule.__init__`.
-As a result, all tensor-data will get automatically shared when using the :class:`~pytorch_lightning.plugins.DDPSpawnStrategy` training type strategy:
+As a result, all tensor-data will get automatically shared when using the :class:`~pytorch_lightning.strategies.DDPSpawnStrategy` training type strategy:
 
 .. warning::
 

docs/source/common/trainer.rst

@@ -1416,7 +1416,7 @@ Supports passing different training strategies with aliases (ddp, ddp_spawn, etc
 
 .. code-block:: python
 
-    from pytorch_lightning.plugins import DDPStrategy
+    from pytorch_lightning.strategies import DDPStrategy
 
 
     class CustomDDPStrategy(DDPStrategy):

docs/source/extensions/accelerators.rst

@@ -23,7 +23,8 @@ One to handle differences from the training routine and one to handle different
 
     from pytorch_lightning import Trainer
     from pytorch_lightning.accelerators import GPUAccelerator
-    from pytorch_lightning.plugins import NativeMixedPrecisionPlugin, DDPStrategy
+    from pytorch_lightning.plugins import NativeMixedPrecisionPlugin
+    from pytorch_lightning.strategies import DDPStrategy
 
     accelerator = GPUAccelerator()
     precision_plugin = NativeMixedPrecisionPlugin(precision=16, device="cuda")

docs/source/extensions/plugins.rst

@@ -60,7 +60,7 @@ Expert users may choose to extend an existing plugin by overriding its methods .
 
 .. code-block:: python
 
-    from pytorch_lightning.plugins import DDPStrategy
+    from pytorch_lightning.strategies import DDPStrategy
 
 
     class CustomDDPStrategy(DDPStrategy):

docs/source/guides/speed.rst

@@ -86,7 +86,7 @@ This by default comes with a performance hit, and can be disabled in most cases.
 
 .. code-block:: python
 
-    from pytorch_lightning.plugins import DDPStrategy
+    from pytorch_lightning.strategies import DDPStrategy
 
     trainer = pl.Trainer(
         gpus=2,
@@ -95,7 +95,7 @@ This by default comes with a performance hit, and can be disabled in most cases.
 
 .. code-block:: python
 
-    from pytorch_lightning.plugins import DDPSpawnStrategy
+    from pytorch_lightning.strategies import DDPSpawnStrategy
 
     trainer = pl.Trainer(
         gpus=2,

docs/source/starter/lightning_lite.rst

@@ -389,7 +389,7 @@ Additionally, you can pass in your custom training type strategy by configuring
 
 .. code-block:: python
 
-    from pytorch_lightning.plugins import DeepSpeedStrategy
+    from pytorch_lightning.strategies import DeepSpeedStrategy
 
     lite = Lite(strategy=DeepSpeedStrategy(stage=2), accelerator="gpu", devices=2)
 

pytorch_lightning/lite/lite.py

@@ -26,7 +26,8 @@ from torch.utils.data import DataLoader, DistributedSampler, RandomSampler, Sequ
 
 from pytorch_lightning.accelerators.accelerator import Accelerator
 from pytorch_lightning.lite.wrappers import _LiteDataLoader, _LiteModule, _LiteOptimizer
-from pytorch_lightning.plugins import DDPSpawnStrategy, DeepSpeedStrategy, PLUGIN_INPUT, Strategy, TPUSpawnStrategy
+from pytorch_lightning.plugins import PLUGIN_INPUT
+from pytorch_lightning.strategies import DDPSpawnStrategy, DeepSpeedStrategy, Strategy, TPUSpawnStrategy
 from pytorch_lightning.strategies.training_type_plugin import TBroadcast
 from pytorch_lightning.trainer.connectors.accelerator_connector import AcceleratorConnector
 from pytorch_lightning.utilities import _AcceleratorType, _StrategyType, move_data_to_device

pytorch_lightning/lite/wrappers.py

@@ -20,7 +20,8 @@ from torch.optim import Optimizer
 from torch.utils.data import DataLoader
 
 from pytorch_lightning.core.mixins import DeviceDtypeModuleMixin
-from pytorch_lightning.plugins import PrecisionPlugin, Strategy
+from pytorch_lightning.plugins import PrecisionPlugin
+from pytorch_lightning.strategies import Strategy
 from pytorch_lightning.utilities.apply_func import apply_to_collection, move_data_to_device
 
 

pytorch_lightning/loops/dataloader/prediction_loop.py

@@ -5,7 +5,7 @@ from torch.utils.data import DataLoader
 
 from pytorch_lightning.loops.dataloader.dataloader_loop import DataLoaderLoop
 from pytorch_lightning.loops.epoch.prediction_epoch_loop import PredictionEpochLoop
-from pytorch_lightning.plugins import DDPSpawnStrategy
+from pytorch_lightning.strategies import DDPSpawnStrategy
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from pytorch_lightning.utilities.types import _PREDICT_OUTPUT
 

pytorch_lightning/loops/utilities.py

@@ -22,7 +22,7 @@ import torch
 from torch.optim import Optimizer
 
 import pytorch_lightning as pl
-from pytorch_lightning.plugins import ParallelStrategy
+from pytorch_lightning.strategies import ParallelStrategy
 from pytorch_lightning.utilities import rank_zero_warn
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from pytorch_lightning.utilities.fetching import AbstractDataFetcher, DataLoaderIterDataFetcher

tests/accelerators/test_accelerator_connector.py

@@ -24,7 +24,14 @@ from pytorch_lightning import Trainer
 from pytorch_lightning.accelerators.accelerator import Accelerator
 from pytorch_lightning.accelerators.cpu import CPUAccelerator
 from pytorch_lightning.accelerators.gpu import GPUAccelerator
-from pytorch_lightning.plugins import (
+from pytorch_lightning.plugins import PrecisionPlugin
+from pytorch_lightning.plugins.environments import (
+    KubeflowEnvironment,
+    LightningEnvironment,
+    SLURMEnvironment,
+    TorchElasticEnvironment,
+)
+from pytorch_lightning.strategies import (
     DataParallelStrategy,
     DDP2Strategy,
     DDPShardedStrategy,
@@ -33,15 +40,8 @@ from pytorch_lightning.plugins import (
     DDPStrategy,
     DeepSpeedStrategy,
     ParallelStrategy,
-    PrecisionPlugin,
     SingleDeviceStrategy,
 )
-from pytorch_lightning.plugins.environments import (
-    KubeflowEnvironment,
-    LightningEnvironment,
-    SLURMEnvironment,
-    TorchElasticEnvironment,
-)
 from pytorch_lightning.utilities import _AcceleratorType, _StrategyType
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from tests.helpers.runif import RunIf
@@ -97,7 +97,7 @@ def test_accelerator_choice_ddp_spawn(cuda_available_mock, device_count_mock):
 )
 @mock.patch("torch.cuda.set_device")
 @mock.patch("torch.cuda.device_count", return_value=2)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_accelerator_choice_ddp_slurm(*_):
     with pytest.deprecated_call(match=r"accelerator='ddp'\)` has been deprecated in v1.5"):
         trainer = Trainer(fast_dev_run=True, accelerator="ddp", gpus=2)
@@ -122,7 +122,7 @@ def test_accelerator_choice_ddp_slurm(*_):
 )
 @mock.patch("torch.cuda.set_device")
 @mock.patch("torch.cuda.device_count", return_value=2)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_accelerator_choice_ddp2_slurm(*_):
     with pytest.deprecated_call(match=r"accelerator='ddp2'\)` has been deprecated in v1.5"):
         trainer = Trainer(fast_dev_run=True, accelerator="ddp2", gpus=2)
@@ -147,7 +147,7 @@ def test_accelerator_choice_ddp2_slurm(*_):
 )
 @mock.patch("torch.cuda.set_device")
 @mock.patch("torch.cuda.device_count", return_value=1)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_accelerator_choice_ddp_te(*_):
     with pytest.deprecated_call(match=r"accelerator='ddp'\)` has been deprecated in v1.5"):
         trainer = Trainer(fast_dev_run=True, accelerator="ddp", gpus=2)
@@ -171,7 +171,7 @@ def test_accelerator_choice_ddp_te(*_):
 )
 @mock.patch("torch.cuda.set_device")
 @mock.patch("torch.cuda.device_count", return_value=1)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_accelerator_choice_ddp2_te(*_):
     with pytest.deprecated_call(match=r"accelerator='ddp2'\)` has been deprecated in v1.5"):
         trainer = Trainer(fast_dev_run=True, accelerator="ddp2", gpus=2)
@@ -186,7 +186,7 @@ def test_accelerator_choice_ddp2_te(*_):
     os.environ, {"WORLD_SIZE": "2", "LOCAL_WORLD_SIZE": "2", "RANK": "1", "LOCAL_RANK": "1", "GROUP_RANK": "0"}
 )
 @mock.patch("torch.cuda.device_count", return_value=0)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_accelerator_choice_ddp_cpu_te(*_):
     trainer = Trainer(fast_dev_run=True, accelerator="ddp_cpu", num_processes=2)
     assert isinstance(trainer.accelerator, CPUAccelerator)
@@ -209,7 +209,7 @@ def test_accelerator_choice_ddp_cpu_te(*_):
 )
 @mock.patch("torch.cuda.set_device")
 @mock.patch("torch.cuda.device_count", return_value=1)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_accelerator_choice_ddp_kubeflow(*_):
     with pytest.deprecated_call(match=r"accelerator='ddp'\)` has been deprecated in v1.5"):
         trainer = Trainer(fast_dev_run=True, accelerator="ddp", gpus=1)
@@ -231,7 +231,7 @@ def test_accelerator_choice_ddp_kubeflow(*_):
     },
 )
 @mock.patch("torch.cuda.device_count", return_value=0)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_accelerator_choice_ddp_cpu_kubeflow(*_):
     trainer = Trainer(fast_dev_run=True, accelerator="ddp_cpu", num_processes=1)
     assert isinstance(trainer.accelerator, CPUAccelerator)
@@ -253,7 +253,7 @@ def test_accelerator_choice_ddp_cpu_kubeflow(*_):
     },
 )
 @mock.patch("torch.cuda.device_count", return_value=0)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_accelerator_choice_ddp_cpu_slurm(*_):
     trainer = Trainer(fast_dev_run=True, accelerator="ddp_cpu", num_processes=2)
     assert trainer._accelerator_connector._is_slurm_managing_tasks()
@@ -333,7 +333,7 @@ def test_accelerator_choice_ddp_cpu_custom_cluster(_, tmpdir):
     },
 )
 @mock.patch("torch.cuda.device_count", return_value=0)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_custom_accelerator(device_count_mock, setup_distributed_mock):
     class Accel(Accelerator):
         pass
@@ -374,7 +374,7 @@ def test_custom_accelerator(device_count_mock, setup_distributed_mock):
     },
 )
 @mock.patch("torch.cuda.device_count", return_value=0)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_dist_backend_accelerator_mapping(*_):
     trainer = Trainer(fast_dev_run=True, strategy="ddp_spawn", num_processes=2)
     assert isinstance(trainer.accelerator, CPUAccelerator)
@@ -707,7 +707,7 @@ def test_strategy_choice_ddp_spawn(cuda_available_mock, device_count_mock):
         "SLURM_LOCALID": "1",
     },
 )
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 @pytest.mark.parametrize("strategy", ["ddp", DDPStrategy()])
 def test_strategy_choice_ddp_slurm(setup_distributed_mock, strategy):
     trainer = Trainer(fast_dev_run=True, strategy=strategy, gpus=2)
@@ -732,7 +732,7 @@ def test_strategy_choice_ddp_slurm(setup_distributed_mock, strategy):
 )
 @mock.patch("torch.cuda.set_device")
 @mock.patch("torch.cuda.device_count", return_value=2)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 @pytest.mark.parametrize("strategy", ["ddp2", DDP2Strategy()])
 def test_strategy_choice_ddp2_slurm(set_device_mock, device_count_mock, setup_distributed_mock, strategy):
     trainer = Trainer(fast_dev_run=True, strategy=strategy, gpus=2)
@@ -757,7 +757,7 @@ def test_strategy_choice_ddp2_slurm(set_device_mock, device_count_mock, setup_di
 )
 @mock.patch("torch.cuda.set_device")
 @mock.patch("torch.cuda.device_count", return_value=2)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_strategy_choice_ddp_te(*_):
     trainer = Trainer(fast_dev_run=True, strategy="ddp", gpus=2)
     assert isinstance(trainer.accelerator, GPUAccelerator)
@@ -780,7 +780,7 @@ def test_strategy_choice_ddp_te(*_):
 )
 @mock.patch("torch.cuda.set_device")
 @mock.patch("torch.cuda.device_count", return_value=2)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_strategy_choice_ddp2_te(*_):
     trainer = Trainer(fast_dev_run=True, strategy="ddp2", gpus=2)
     assert isinstance(trainer.accelerator, GPUAccelerator)
@@ -794,7 +794,7 @@ def test_strategy_choice_ddp2_te(*_):
     os.environ, {"WORLD_SIZE": "2", "LOCAL_WORLD_SIZE": "2", "RANK": "1", "LOCAL_RANK": "1", "GROUP_RANK": "0"}
 )
 @mock.patch("torch.cuda.device_count", return_value=0)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_strategy_choice_ddp_cpu_te(*_):
     trainer = Trainer(fast_dev_run=True, strategy="ddp_spawn", num_processes=2)
     assert isinstance(trainer.accelerator, CPUAccelerator)
@@ -817,7 +817,7 @@ def test_strategy_choice_ddp_cpu_te(*_):
 )
 @mock.patch("torch.cuda.set_device")
 @mock.patch("torch.cuda.device_count", return_value=1)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_strategy_choice_ddp_kubeflow(*_):
     trainer = Trainer(fast_dev_run=True, strategy="ddp", gpus=1)
     assert isinstance(trainer.accelerator, GPUAccelerator)
@@ -838,7 +838,7 @@ def test_strategy_choice_ddp_kubeflow(*_):
     },
 )
 @mock.patch("torch.cuda.device_count", return_value=0)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 def test_strategy_choice_ddp_cpu_kubeflow(*_):
     trainer = Trainer(fast_dev_run=True, strategy="ddp_spawn", num_processes=2)
     assert isinstance(trainer.accelerator, CPUAccelerator)
@@ -860,7 +860,7 @@ def test_strategy_choice_ddp_cpu_kubeflow(*_):
     },
 )
 @mock.patch("torch.cuda.device_count", return_value=0)
-@mock.patch("pytorch_lightning.plugins.DDPStrategy.setup_distributed", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPStrategy.setup_distributed", autospec=True)
 @pytest.mark.parametrize("strategy", ["ddp", DDPStrategy()])
 def test_strategy_choice_ddp_cpu_slurm(device_count_mock, setup_distributed_mock, strategy):
     trainer = Trainer(fast_dev_run=True, strategy=strategy, num_processes=2)

tests/accelerators/test_cpu.py

@@ -8,9 +8,9 @@ import torch
 import pytorch_lightning as pl
 from pytorch_lightning import Trainer
 from pytorch_lightning.accelerators import CPUAccelerator
-from pytorch_lightning.plugins import SingleDeviceStrategy
 from pytorch_lightning.plugins.io.torch_plugin import TorchCheckpointIO
 from pytorch_lightning.plugins.precision.precision_plugin import PrecisionPlugin
+from pytorch_lightning.strategies import SingleDeviceStrategy
 from tests.helpers.boring_model import BoringModel
 
 

tests/accelerators/test_ipu.py

@@ -21,7 +21,8 @@ import torch.nn.functional as F
 from pytorch_lightning import Callback, seed_everything, Trainer
 from pytorch_lightning.accelerators import CPUAccelerator, IPUAccelerator
 from pytorch_lightning.core.lightning import LightningModule
-from pytorch_lightning.plugins import IPUPrecisionPlugin, IPUStrategy
+from pytorch_lightning.plugins import IPUPrecisionPlugin
+from pytorch_lightning.strategies.ipu import IPUStrategy
 from pytorch_lightning.trainer.states import RunningStage, TrainerFn
 from pytorch_lightning.trainer.supporters import CombinedLoader
 from pytorch_lightning.utilities import _AcceleratorType, _IPU_AVAILABLE

tests/accelerators/test_tpu.py

@@ -23,7 +23,8 @@ from torch.utils.data import DataLoader
 from pytorch_lightning import Trainer
 from pytorch_lightning.accelerators.cpu import CPUAccelerator
 from pytorch_lightning.accelerators.tpu import TPUAccelerator
-from pytorch_lightning.plugins import DDPStrategy, TPUPrecisionPlugin, TPUSpawnStrategy, XLACheckpointIO
+from pytorch_lightning.plugins import TPUPrecisionPlugin, XLACheckpointIO
+from pytorch_lightning.strategies import DDPStrategy, TPUSpawnStrategy
 from pytorch_lightning.utilities import find_shared_parameters
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from tests.helpers.boring_model import BoringModel, RandomDataset

tests/benchmarks/test_sharded_parity.py

@@ -19,7 +19,7 @@ import pytest
 import torch
 
 from pytorch_lightning import seed_everything, Trainer
-from pytorch_lightning.plugins import DDPSpawnShardedStrategy
+from pytorch_lightning.strategies import DDPSpawnShardedStrategy
 from tests.helpers.boring_model import BoringModel, RandomDataset
 from tests.helpers.runif import RunIf
 

tests/callbacks/test_stochastic_weight_avg.py

@@ -22,8 +22,7 @@ from torch.utils.data import DataLoader
 
 from pytorch_lightning import LightningModule, Trainer
 from pytorch_lightning.callbacks import StochasticWeightAveraging
-from pytorch_lightning.plugins import DDPSpawnStrategy
-from pytorch_lightning.strategies import Strategy
+from pytorch_lightning.strategies import DDPSpawnStrategy, Strategy
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from tests.helpers.boring_model import BoringModel, RandomDataset, RandomIterableDataset
 from tests.helpers.runif import RunIf

tests/deprecated_api/test_remove_1-7.py

@@ -27,7 +27,6 @@ from pytorch_lightning.callbacks.progress import ProgressBar
 from pytorch_lightning.callbacks.xla_stats_monitor import XLAStatsMonitor
 from pytorch_lightning.loggers import LoggerCollection, TestTubeLogger
 from pytorch_lightning.overrides.distributed import IndexBatchSamplerWrapper
-from pytorch_lightning.plugins import SingleDeviceStrategy
 from pytorch_lightning.plugins.environments import (
     KubeflowEnvironment,
     LightningEnvironment,
@@ -35,6 +34,7 @@ from pytorch_lightning.plugins.environments import (
     SLURMEnvironment,
     TorchElasticEnvironment,
 )
+from pytorch_lightning.strategies import SingleDeviceStrategy
 from tests.callbacks.test_callbacks import OldStatefulCallback
 from tests.deprecated_api import _soft_unimport_module
 from tests.helpers import BoringModel

tests/lite/test_lite.py

@@ -25,7 +25,8 @@ from torch.utils.data import DataLoader, DistributedSampler, Sampler
 
 from pytorch_lightning.lite import LightningLite
 from pytorch_lightning.lite.wrappers import _LiteDataLoader, _LiteModule, _LiteOptimizer
-from pytorch_lightning.plugins import DeepSpeedStrategy, PrecisionPlugin, Strategy
+from pytorch_lightning.plugins import PrecisionPlugin
+from pytorch_lightning.strategies import DeepSpeedStrategy, Strategy
 from pytorch_lightning.utilities import _StrategyType
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from pytorch_lightning.utilities.seed import pl_worker_init_function

tests/models/test_sync_batchnorm.py

@@ -17,8 +17,8 @@ import torch.nn as nn
 import torch.nn.functional as F
 
 from pytorch_lightning import LightningModule, seed_everything, Trainer
-from pytorch_lightning.plugins import DDPSpawnStrategy
 from pytorch_lightning.plugins.environments import LightningEnvironment
+from pytorch_lightning.strategies import DDPSpawnStrategy
 from pytorch_lightning.utilities import FLOAT16_EPSILON
 from tests.helpers.datamodules import MNISTDataModule
 from tests.helpers.runif import RunIf

tests/models/test_tpu.py

@@ -24,7 +24,7 @@ import tests.helpers.utils as tutils
 from pytorch_lightning import Trainer
 from pytorch_lightning.accelerators import TPUAccelerator
 from pytorch_lightning.callbacks import EarlyStopping
-from pytorch_lightning.plugins import TPUSpawnStrategy
+from pytorch_lightning.strategies import TPUSpawnStrategy
 from pytorch_lightning.trainer.connectors.logger_connector.result import _Sync
 from pytorch_lightning.utilities import _AcceleratorType, _TPU_AVAILABLE
 from pytorch_lightning.utilities.distributed import ReduceOp

tests/plugins/test_checkpoint_io_plugin.py

@@ -19,7 +19,8 @@ import torch
 
 from pytorch_lightning import Trainer
 from pytorch_lightning.callbacks import ModelCheckpoint
-from pytorch_lightning.plugins import CheckpointIO, SingleDeviceStrategy
+from pytorch_lightning.plugins import CheckpointIO
+from pytorch_lightning.strategies import SingleDeviceStrategy
 from pytorch_lightning.utilities.types import _PATH
 from tests.helpers.boring_model import BoringModel
 

tests/plugins/test_cluster_integration.py

@@ -18,8 +18,8 @@ import pytest
 import torch
 
 from pytorch_lightning import Trainer
-from pytorch_lightning.plugins import DDP2Strategy, DDPShardedStrategy, DDPStrategy, DeepSpeedStrategy
 from pytorch_lightning.plugins.environments import LightningEnvironment, SLURMEnvironment, TorchElasticEnvironment
+from pytorch_lightning.strategies import DDP2Strategy, DDPShardedStrategy, DDPStrategy, DeepSpeedStrategy
 from pytorch_lightning.utilities import rank_zero_only
 from tests.helpers.runif import RunIf
 

tests/plugins/test_custom_plugin.py

@@ -18,7 +18,7 @@ import pytest
 import torch
 
 from pytorch_lightning import Trainer
-from pytorch_lightning.plugins import DDPStrategy, SingleDeviceStrategy
+from pytorch_lightning.strategies import DDPStrategy, SingleDeviceStrategy
 from tests.helpers import BoringModel
 from tests.helpers.runif import RunIf
 

tests/plugins/test_ddp_fully_sharded_with_full_state_dict.py

@@ -7,7 +7,8 @@ import torch
 
 from pytorch_lightning import Trainer
 from pytorch_lightning.callbacks import ModelCheckpoint
-from pytorch_lightning.plugins import DDPFullyShardedStrategy, FullyShardedNativeMixedPrecisionPlugin
+from pytorch_lightning.plugins import FullyShardedNativeMixedPrecisionPlugin
+from pytorch_lightning.strategies import DDPFullyShardedStrategy
 from pytorch_lightning.utilities import _FAIRSCALE_FULLY_SHARDED_AVAILABLE
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from tests.helpers.boring_model import BoringModel

tests/plugins/test_ddp_plugin.py

@@ -19,8 +19,8 @@ import torch
 from torch.nn.parallel import DistributedDataParallel
 
 from pytorch_lightning import LightningModule, Trainer
-from pytorch_lightning.plugins import DDPStrategy
 from pytorch_lightning.plugins.environments import LightningEnvironment
+from pytorch_lightning.strategies import DDPStrategy
 from pytorch_lightning.trainer.states import TrainerFn
 from tests.helpers.boring_model import BoringModel
 from tests.helpers.runif import RunIf

tests/plugins/test_ddp_plugin_with_comm_hook.py

@@ -14,7 +14,7 @@
 import torch
 
 from pytorch_lightning import Trainer
-from pytorch_lightning.plugins import DDPSpawnStrategy, DDPStrategy
+from pytorch_lightning.strategies import DDPSpawnStrategy, DDPStrategy
 from pytorch_lightning.utilities import _TORCH_GREATER_EQUAL_1_8, _TORCH_GREATER_EQUAL_1_10
 from tests.helpers import BoringModel
 from tests.helpers.runif import RunIf

tests/plugins/test_ddp_spawn_plugin.py

@@ -19,7 +19,7 @@ import torch
 from torch.nn.parallel.distributed import DistributedDataParallel
 
 from pytorch_lightning import LightningModule, Trainer
-from pytorch_lightning.plugins import DDPSpawnStrategy
+from pytorch_lightning.strategies import DDPSpawnStrategy
 from pytorch_lightning.trainer.states import TrainerFn
 from tests.helpers.boring_model import BoringDataModule, BoringModel
 from tests.helpers.runif import RunIf

tests/plugins/test_deepspeed_plugin.py

@@ -15,7 +15,8 @@ from torchmetrics import Accuracy
 
 from pytorch_lightning import LightningDataModule, LightningModule, seed_everything, Trainer
 from pytorch_lightning.callbacks import Callback, LearningRateMonitor, ModelCheckpoint
-from pytorch_lightning.plugins import DeepSpeedPrecisionPlugin, DeepSpeedStrategy
+from pytorch_lightning.plugins import DeepSpeedPrecisionPlugin
+from pytorch_lightning.strategies import DeepSpeedStrategy
 from pytorch_lightning.strategies.deepspeed import LightningDeepSpeedModule
 from pytorch_lightning.utilities.exceptions import MisconfigurationException
 from pytorch_lightning.utilities.imports import _DEEPSPEED_AVAILABLE

tests/plugins/test_plugins_registry.py

@@ -14,8 +14,8 @@
 import pytest
 
 from pytorch_lightning import Trainer
-from pytorch_lightning.plugins import (
-    CheckpointIO,
+from pytorch_lightning.plugins import CheckpointIO, TrainingTypePluginsRegistry
+from pytorch_lightning.strategies import (
     DDPFullyShardedStrategy,
     DDPShardedStrategy,
     DDPSpawnShardedStrategy,
@@ -23,7 +23,6 @@ from pytorch_lightning.plugins import (
     DDPStrategy,
     DeepSpeedStrategy,
     TPUSpawnStrategy,
-    TrainingTypePluginsRegistry,
 )
 from tests.helpers.runif import RunIf
 

tests/plugins/test_sharded_plugin.py

@@ -6,7 +6,7 @@ import pytest
 import torch
 
 from pytorch_lightning import LightningModule, Trainer
-from pytorch_lightning.plugins import DDPShardedStrategy, DDPSpawnShardedStrategy
+from pytorch_lightning.strategies import DDPShardedStrategy, DDPSpawnShardedStrategy
 from pytorch_lightning.trainer.states import TrainerFn
 from pytorch_lightning.utilities import _FAIRSCALE_AVAILABLE
 from tests.helpers.boring_model import BoringModel
@@ -231,7 +231,7 @@ def test_configure_ddp(tmpdir):
 
 
 @RunIf(skip_windows=True, fairscale=True)
-@mock.patch("pytorch_lightning.plugins.DDPShardedStrategy._wrap_optimizers", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPShardedStrategy._wrap_optimizers", autospec=True)
 @pytest.mark.parametrize("cls", [DDPShardedStrategy, DDPSpawnShardedStrategy])
 def test_custom_kwargs_sharded(tmpdir, cls):
     """Tests to ensure that if custom kwargs are passed, they are set correctly."""
@@ -248,7 +248,7 @@ def test_custom_kwargs_sharded(tmpdir, cls):
 
 
 @RunIf(skip_windows=True, fairscale=True)
-@mock.patch("pytorch_lightning.plugins.DDPShardedStrategy._wrap_optimizers", autospec=True)
+@mock.patch("pytorch_lightning.strategies.DDPShardedStrategy._wrap_optimizers", autospec=True)
 @pytest.mark.parametrize(["params", "expected_buffer_size"], [(dict(), 0), (dict(reduce_buffer_size=128), 128)])
 @pytest.mark.parametrize("num_nodes", [1, 2])
 def test_custom_kwargs_sharded_reduce_buffer_size(tmpdir, params, expected_buffer_size, num_nodes):

tests/plugins/test_single_device_plugin.py

@@ -14,7 +14,7 @@
 import torch
 
 from pytorch_lightning import Trainer
-from pytorch_lightning.plugins import SingleDeviceStrategy
+from pytorch_lightning.strategies import SingleDeviceStrategy
 from tests.helpers.boring_model import BoringModel
 from tests.helpers.runif import RunIf
 

tests/trainer/test_trainer.py

@@ -37,7 +37,7 @@ from pytorch_lightning.callbacks.prediction_writer import BasePredictionWriter
 from pytorch_lightning.core.saving import load_hparams_from_tags_csv, load_hparams_from_yaml, save_hparams_to_tags_csv
 from pytorch_lightning.loggers import TensorBoardLogger
 from pytorch_lightning.overrides.distributed import IndexBatchSamplerWrapper, UnrepeatedDistributedSampler
-from pytorch_lightning.plugins import (
+from pytorch_lightning.strategies import (
     DataParallelStrategy,
     DDP2Strategy,
     DDPFullyShardedStrategy,

tests/utilities/test_deepspeed_collate_checkpoint.py

@@ -16,7 +16,7 @@ import os
 import torch
 
 from pytorch_lightning import Trainer
-from pytorch_lightning.plugins import DeepSpeedStrategy
+from pytorch_lightning.strategies import DeepSpeedStrategy
 from pytorch_lightning.utilities.deepspeed import convert_zero_checkpoint_to_fp32_state_dict
 from tests.helpers.boring_model import BoringModel
 from tests.helpers.runif import RunIf