Rename LightningLite (1/n) (#15932)

Author: awaelchli

docs/source-pytorch/starter/lightning_lite.rst

@@ -7,7 +7,7 @@ Lightning Lite
 on any kind of device while retaining full control over their own loops and optimization logic.
 
 .. image:: https://pl-public-data.s3.amazonaws.com/docs/static/images/lite/lightning_lite.gif
-    :alt: Animation showing how to convert your PyTorch code to LightningLite.
+    :alt: Animation showing how to convert your PyTorch code to Fabric.
     :width: 500
     :align: center
 
@@ -72,8 +72,8 @@ The ``train`` function contains a standard training loop used to train ``MyModel
 ----------
 
 
-Convert to LightningLite
-========================
+Convert to Fabric
+=================
 
 Here are five easy steps to let :class:`~pytorch_lightning.lite.LightningLite` scale your PyTorch models.
 
@@ -89,7 +89,7 @@ Here are five easy steps to let :class:`~pytorch_lightning.lite.LightningLite` s
     import torch
     from torch import nn
     from torch.utils.data import DataLoader, Dataset
-    from lightning.lite import LightningLite
+    from lightning.lite import Fabric
 
 
     class MyModel(nn.Module):
@@ -102,7 +102,7 @@ Here are five easy steps to let :class:`~pytorch_lightning.lite.LightningLite` s
 
     def train(args):
 
-        lite = LightningLite()
+        lite = Fabric()
 
         model = MyModel(...)
         optimizer = torch.optim.SGD(model.parameters(), ...)
@@ -124,7 +124,7 @@ Here are five easy steps to let :class:`~pytorch_lightning.lite.LightningLite` s
 
 
 That's all you need to do to your code. You can now train on any kind of device and scale your training.
-Check out `this <https://github.com/Lightning-AI/lightning/blob/master/examples/lite/image_classifier_2_lite.py>`_ full MNIST training example with LightningLite.
+Check out `this <https://github.com/Lightning-AI/lightning/blob/master/examples/lite/image_classifier_2_lite.py>`_ full MNIST training example with Fabric.
 
 Here is how to train on eight GPUs with `torch.bfloat16 <https://pytorch.org/docs/1.10.0/generated/torch.Tensor.bfloat16.html>`_ precision:
 
@@ -149,7 +149,7 @@ You can also easily use distributed collectives if required.
 
 .. code-block:: python
 
-    lite = LightningLite()
+    lite = Fabric()
 
     # Transfer and concatenate tensors across processes
     lite.all_gather(...)

examples/app_multi_node/train_lite.py

@@ -2,7 +2,7 @@
 
 import lightning as L
 from lightning.app.components import LiteMultiNode
-from lightning.lite import LightningLite
+from lightning.lite import Fabric
 
 
 class LitePyTorchDistributed(L.LightningWork):
@@ -14,8 +14,8 @@ def run(self):
             torch.nn.Linear(1, 1),
         )
 
-        # 2. Create LightningLite.
-        lite = LightningLite(strategy="ddp", precision=16)
+        # 2. Create Fabric.
+        lite = Fabric(strategy="ddp", precision=16)
         model, optimizer = lite.setup(model, torch.optim.SGD(model.parameters(), lr=0.01))
         criterion = torch.nn.MSELoss()
 

examples/lite/image_classifier_2_lite.py

@@ -12,11 +12,11 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-"""Here are 4 easy steps to use LightningLite in your PyTorch code.
+"""Here are 4 easy steps to use Fabric in your PyTorch code.
 
 1. Create the Lightning Lite object at the beginning of your script.
 
-2. Remove all ``.to`` and ``.cuda`` calls since LightningLite will take care of it.
+2. Remove all ``.to`` and ``.cuda`` calls since Fabric will take care of it.
 
 3. Apply ``setup`` over each model and optimizers pair, ``setup_dataloaders`` on all your dataloaders,
 and replace ``loss.backward()`` with ``self.backward(loss)``.
@@ -40,7 +40,7 @@
 from torchmetrics.classification import Accuracy
 from torchvision.datasets import MNIST
 
-from lightning.lite import LightningLite  # import LightningLite
+from lightning.lite import Fabric  # import Fabric
 from lightning.lite import seed_everything
 
 DATASETS_PATH = path.join(path.dirname(__file__), "..", "..", "Datasets")
@@ -49,7 +49,7 @@
 def run(hparams):
     # Create the Lightning Lite object. The parameters like accelerator, strategy, devices etc. will be proided
     # by the command line. See all options: `lightning run model --help`
-    lite = LightningLite()
+    lite = Fabric()
 
     lite.hparams = hparams
     seed_everything(hparams.seed)  # instead of torch.manual_seed(...)
@@ -148,7 +148,7 @@ def run(hparams):
     # Arguments can be passed in through the CLI as normal and will be parsed here
     # Example:
     # lightning run model image_classifier.py accelerator=cuda --epochs=3
-    parser = argparse.ArgumentParser(description="LightningLite MNIST Example")
+    parser = argparse.ArgumentParser(description="Fabric MNIST Example")
     parser.add_argument(
         "--batch-size", type=int, default=64, metavar="N", help="input batch size for training (default: 64)"
     )

src/lightning/__init__.py

@@ -36,7 +36,7 @@ def _detail(self: Any, message: str, *args: Any, **kwargs: Any) -> None:
 from lightning.app.perf import pdb  # noqa: E402
 from lightning.app.utilities.packaging.build_config import BuildConfig  # noqa: E402
 from lightning.app.utilities.packaging.cloud_compute import CloudCompute  # noqa: E402
-from lightning.lite.lite import LightningLite  # noqa: E402
+from lightning.lite.lite import Fabric  # noqa: E402
 from lightning.pytorch.callbacks import Callback  # noqa: E402
 from lightning.pytorch.core import LightningDataModule, LightningModule  # noqa: E402
 from lightning.pytorch.trainer import Trainer  # noqa: E402
@@ -60,7 +60,7 @@ def _detail(self: Any, message: str, *args: Any, **kwargs: Any) -> None:
     "LightningModule",
     "Callback",
     "seed_everything",
-    "LightningLite",
+    "Fabric",
     "storage",
     "pdb",
 ]

src/lightning_app/utilities/introspection.py

@@ -261,7 +261,7 @@ class TorchMetricVisitor(LightningVisitor):
 
 
 class LightningLiteVisitor(LightningVisitor):
-    class_name = "LightningLite"
+    class_name = "Fabric"
 
 
 class LightningBaseProfilerVisitor(LightningVisitor):

src/lightning_lite/CHANGELOG.md

@@ -28,7 +28,7 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 - The `LightningLite.run()` method is no longer abstract ([#14992](https://github.com/Lightning-AI/lightning/issues/14992))
 
 
--
+- Renamed the class `LightningLite` to `Fabric` ([#15932](https://github.com/Lightning-AI/lightning/issues/15932))
 
 
 ### Deprecated

src/lightning_lite/__init__.py

@@ -20,10 +20,10 @@
 os.environ["PYTORCH_NVML_BASED_CUDA_CHECK"] = "1"
 
 
-from lightning_lite.lite import LightningLite  # noqa: E402
+from lightning_lite.lite import Fabric  # noqa: E402
 from lightning_lite.utilities.seed import seed_everything  # noqa: E402
 
-__all__ = ["LightningLite", "seed_everything"]
+__all__ = ["Fabric", "seed_everything"]
 
 # for compatibility with namespace packages
 __import__("pkg_resources").declare_namespace(__name__)

src/lightning_lite/cli.py

@@ -106,9 +106,9 @@
     )
     @click.argument("script_args", nargs=-1, type=click.UNPROCESSED)
     def _run_model(**kwargs: Any) -> None:
-        """Run a Lightning Lite script.
+        """Run a Lightning Fabric script.
 
-        SCRIPT is the path to the Python script with the code to run. The script must contain a LightningLite object.
+        SCRIPT is the path to the Python script with the code to run. The script must contain a Fabric object.
 
         SCRIPT_ARGS are the remaining arguments that you can pass to the script itself and are expected to be parsed
         there.
@@ -120,7 +120,7 @@ def _run_model(**kwargs: Any) -> None:
 def _set_env_variables(args: Namespace) -> None:
     """Set the environment variables for the new processes.
 
-    The Lite connector will parse the arguments set here.
+    The Fabric connector will parse the arguments set here.
     """
     os.environ["LT_CLI_USED"] = "1"
     os.environ["LT_ACCELERATOR"] = str(args.accelerator)
@@ -187,7 +187,7 @@ def main(args: Namespace, script_args: Optional[List[str]] = None) -> None:
 if __name__ == "__main__":
     if not _CLICK_AVAILABLE:  # pragma: no cover
         _log.error(
-            "To use the Lightning Lite CLI, you must have `click` installed."
+            "To use the Lightning Fabric CLI, you must have `click` installed."
             " Install it by running `pip install -U click`."
         )
         raise SystemExit(1)

src/lightning_lite/connector.py

@@ -65,7 +65,7 @@
 
 
 class _Connector:
-    """The Connector parses several Lite arguments and instantiates the Strategy including its owned components.
+    """The Connector parses several Fabric arguments and instantiates the Strategy including its owned components.
 
         A. accelerator flag could be:
             1. accelerator class
@@ -297,7 +297,7 @@ def _check_device_config_and_set_final_flags(
                 else self._accelerator_flag
             )
             raise ValueError(
-                f"`Lite(devices={self._devices_flag!r})` value is not a valid input"
+                f"`Fabric(devices={self._devices_flag!r})` value is not a valid input"
                 f" using {accelerator_name} accelerator."
             )
 
@@ -345,7 +345,7 @@ def _set_parallel_devices_and_init_accelerator(self) -> None:
                 f"`{accelerator_cls.__qualname__}` can not run on your system"
                 " since the accelerator is not available. The following accelerator(s)"
                 " is available and can be passed into `accelerator` argument of"
-                f" `Lite`: {available_accelerator}."
+                f" `Fabric`: {available_accelerator}."
             )
 
         self._set_devices_flag_if_auto_passed()
@@ -416,14 +416,14 @@ def _check_strategy_and_fallback(self) -> None:
             strategy_flag = "ddp"
         if strategy_flag in _DDP_FORK_ALIASES and "fork" not in torch.multiprocessing.get_all_start_methods():
             raise ValueError(
-                f"You selected `Lite(strategy='{strategy_flag}')` but process forking is not supported on this"
-                f" platform. We recommed `Lite(strategy='ddp_spawn')` instead."
+                f"You selected `Fabric(strategy='{strategy_flag}')` but process forking is not supported on this"
+                f" platform. We recommed `Fabric(strategy='ddp_spawn')` instead."
             )
         if (
             strategy_flag in _FSDP_ALIASES or isinstance(self._strategy_flag, FSDPStrategy)
         ) and self._accelerator_flag not in ("cuda", "gpu"):
             raise ValueError(
-                "You selected the FSDP strategy but FSDP is only available on GPU. Set `Lite(accelerator='gpu', ...)`"
+                "You selected the FSDP strategy but FSDP is only available on GPU. Set `Fabric(accelerator='gpu', ...)`"
                 " to continue or select a different strategy."
             )
         if strategy_flag:
@@ -449,7 +449,7 @@ def _check_and_init_precision(self) -> Precision:
             elif self._precision_input in (16, "bf16"):
                 if self._precision_input == 16:
                     rank_zero_warn(
-                        "You passed `Lite(accelerator='tpu', precision=16)` but AMP"
+                        "You passed `Fabric(accelerator='tpu', precision=16)` but AMP"
                         " is not supported with TPUs. Using `precision='bf16'` instead."
                     )
                 return TPUBf16Precision()
@@ -463,7 +463,7 @@ def _check_and_init_precision(self) -> Precision:
 
         if self._precision_input == 16 and self._accelerator_flag == "cpu":
             rank_zero_warn(
-                "You passed `Lite(accelerator='cpu', precision=16)` but native AMP is not supported on CPU."
+                "You passed `Fabric(accelerator='cpu', precision=16)` but native AMP is not supported on CPU."
                 " Using `precision='bf16'` instead."
             )
             self._precision_input = "bf16"
@@ -487,7 +487,7 @@ def _validate_precision_choice(self) -> None:
         if isinstance(self.accelerator, TPUAccelerator):
             if self._precision_input == 64:
                 raise NotImplementedError(
-                    "`Lite(accelerator='tpu', precision=64)` is not implemented."
+                    "`Fabric(accelerator='tpu', precision=64)` is not implemented."
                     " Please, open an issue in `https://github.com/Lightning-AI/lightning/issues`"
                     " requesting this feature."
                 )
@@ -519,10 +519,10 @@ def _lazy_init_strategy(self) -> None:
 
         if _IS_INTERACTIVE and self.strategy.launcher and not self.strategy.launcher.is_interactive_compatible:
             raise RuntimeError(
-                f"`Lite(strategy={self._strategy_flag!r})` is not compatible with an interactive"
+                f"`Fabric(strategy={self._strategy_flag!r})` is not compatible with an interactive"
                 " environment. Run your code as a script, or choose one of the compatible strategies:"
-                f" Lite(strategy=None|{'|'.join(_StrategyType.interactive_compatible_types())})."
-                " In case you are spawning processes yourself, make sure to include the Lite"
+                f" Fabric(strategy=None|{'|'.join(_StrategyType.interactive_compatible_types())})."
+                " In case you are spawning processes yourself, make sure to include the Fabric"
                 " creation inside the worker function."
             )
 
@@ -549,9 +549,9 @@ def _argument_from_env(name: str, current: Any, default: Any) -> Any:
 
         if env_value is not None and env_value != current and current != default:
             raise ValueError(
-                f"Your code has `LightningLite({name}={current!r}, ...)` but it conflicts with the value "
+                f"Your code has `Fabric({name}={current!r}, ...)` but it conflicts with the value "
                 f"`--{name}={current}` set through the CLI. "
-                " Remove it either from the CLI or from the Lightning Lite object."
+                " Remove it either from the CLI or from the Lightning Fabric object."
             )
         if env_value is None:
             return current

src/lightning_lite/lite.py

@@ -54,8 +54,8 @@
 from lightning_lite.wrappers import _LiteDataLoader, _LiteModule, _LiteOptimizer
 
 
-class LightningLite:
-    """Lite accelerates your PyTorch training or inference code with minimal changes required.
+class Fabric:
+    """Fabric accelerates your PyTorch training or inference code with minimal changes required.
 
     - Automatic placement of models and data onto the device.
     - Automatic support for mixed and double precision (smaller memory footprint).
@@ -139,7 +139,7 @@ def is_global_zero(self) -> bool:
         return self._strategy.is_global_zero
 
     def run(self, *args: Any, **kwargs: Any) -> Any:
-        """All the code inside this run method gets accelerated by Lite.
+        """All the code inside this run method gets accelerated by Fabric.
 
         You can pass arbitrary arguments to this function when overriding it.
         """
@@ -502,16 +502,16 @@ def load(self, filepath: Union[str, Path]) -> Any:
         """
         return self._strategy.load_checkpoint(filepath)
 
-    def launch(self, function: Optional[Callable[["LightningLite"], Any]] = None, *args: Any, **kwargs: Any) -> Any:
+    def launch(self, function: Optional[Callable[["Fabric"], Any]] = None, *args: Any, **kwargs: Any) -> Any:
         if _is_using_cli():
             raise RuntimeError(
                 "This script was launched through the CLI, and processes have already been created. Calling "
                 " `.launch()` again is not allowed."
             )
         if function is not None and not inspect.signature(function).parameters:
             raise TypeError(
-                "The function passed to `Lite.launch()` needs to take at least one argument. The launcher will pass"
-                " in the `LightningLite` object so you can use it inside the function."
+                "The function passed to `Fabric.launch()` needs to take at least one argument. The launcher will pass"
+                " in the `Fabric` object so you can use it inside the function."
             )
         function = partial(self._run_with_setup, function or _do_nothing)
         args = [self, *args]
@@ -550,9 +550,9 @@ def _move_model_to_device(self, model: nn.Module, optimizers: List[Optimizer]) -
         initial_device = next(model.parameters()).device
         if any(param.device != initial_device for param in model.parameters()):
             rank_zero_warn(
-                "The model passed to `Lite.setup()` has parameters on different devices. Since `move_to_device=True`,"
+                "The model passed to `Fabric.setup()` has parameters on different devices. Since `move_to_device=True`,"
                 " all parameters will be moved to the new device. If this is not desired, set "
-                " `Lite.setup(..., move_to_device=False)`.",
+                " `Fabric.setup(..., move_to_device=False)`.",
                 category=PossibleUserWarning,
             )
 
@@ -586,10 +586,10 @@ def _get_distributed_sampler(dataloader: DataLoader, **kwargs: Any) -> Distribut
         return DistributedSamplerWrapper(dataloader.sampler, **kwargs)
 
     def _prepare_run_method(self) -> None:
-        if is_overridden("run", self, LightningLite) and _is_using_cli():
+        if is_overridden("run", self, Fabric) and _is_using_cli():
             raise TypeError(
-                "Overriding `LightningLite.run()` and launching from the CLI is not allowed. Run the script normally,"
-                " or change your code to directly call `lite = LightningLite(...); lite.setup(...)` etc."
+                "Overriding `Fabric.run()` and launching from the CLI is not allowed. Run the script normally,"
+                " or change your code to directly call `lite = Fabric(...); lite.setup(...)` etc."
             )
         # wrap the run method, so we can inject setup logic or spawn processes for the user
         setattr(self, "run", partial(self._run_impl, self.run))