Rename LiteMultiNode to FabricMultiNode (#16505)

Author: awaelchli

.github/checkgroup.yml

@@ -205,7 +205,7 @@ subprojects:
       - "examples/fabric/**"
       - "examples/run_fabric_examples.sh"
       - "tests/tests_fabric/run_standalone_*.sh"
-      - "tests/tests_pytorch/run_standalone_tests.sh"  # used by Lite through a symlink
+      - "tests/tests_pytorch/run_standalone_tests.sh"  # used by Fabric through a symlink
       - "requirements/fabric/**"
       - "src/lightning_fabric/**"
       - "tests/tests_fabric/**"

docs/source-app/api_reference/components.rst

@@ -28,7 +28,7 @@ ___________________
     ~serve.python_server.PythonServer
     ~serve.streamlit.ServeStreamlit
     ~multi_node.base.MultiNode
-    ~multi_node.lite.LiteMultiNode
+    ~multi_node.fabric.FabricMultiNode
     ~multi_node.pytorch_spawn.PyTorchSpawnMultiNode
     ~multi_node.trainer.LightningTrainerMultiNode
     ~serve.auto_scaler.AutoScaler

docs/source-pytorch/accelerators/gpu_basic.rst

@@ -102,10 +102,10 @@ use the following utility function to pick GPU indices that are "accessible", wi
     # Find two GPUs on the system that are not already occupied
     trainer = Trainer(accelerator="cuda", devices=find_usable_cuda_devices(2))
 
-    from lightning.lite.accelerators import find_usable_cuda_devices
+    from lightning.fabric.accelerators import find_usable_cuda_devices
 
-    # Works with LightningLite too
-    lite = LightningLite(accelerator="cuda", devices=find_usable_cuda_devices(2))
+    # Works with Fabric too
+    fabric = Fabric(accelerator="cuda", devices=find_usable_cuda_devices(2))
 
 
 This is especially useful when GPUs are configured to be in "exclusive compute mode", such that only one process at a time is allowed access to the device.

docs/source-pytorch/fabric/guide/multi_node/cloud.rst

@@ -42,7 +42,7 @@ Launch multi-node training in the cloud
     :caption: app.py
 
     import lightning as L
-    from lightning.app.components import LiteMultiNode
+    from lightning.app.components import FabricMultiNode
 
     # 1. Put your code inside a LightningWork
     class MyTrainingComponent(L.LightningWork):
@@ -58,16 +58,16 @@ Launch multi-node training in the cloud
             model, optimizer = fabric.setup(model, optimizer)
             ...
 
-**Step 2:** Init a :class:`~lightning_app.core.app.LightningApp` with the ``LiteMultiNode`` component.
+**Step 2:** Init a :class:`~lightning_app.core.app.LightningApp` with the ``FabricMultiNode`` component.
 Configure the number of nodes, the number of GPUs per node, and the type of GPU:
 
 .. code-block:: python
     :emphasize-lines: 5,7
     :caption: app.py
 
-    # 2. Create the app with the LiteMultiNode component inside
+    # 2. Create the app with the FabricMultiNode component inside
     app = L.LightningApp(
-        LiteMultiNode(
+        FabricMultiNode(
             MyTrainingComponent,
             # Run with 2 nodes
             num_nodes=2,

examples/app_multi_node/train_fabric.py

@@ -1,11 +1,11 @@
 import torch
 
 import lightning as L
-from lightning.app.components import LiteMultiNode
+from lightning.app.components import FabricMultiNode
 from lightning.fabric import Fabric
 
 
-class LitePyTorchDistributed(L.LightningWork):
+class FabricPyTorchDistributed(L.LightningWork):
     def run(self):
         # 1. Prepare the model
         model = torch.nn.Sequential(
@@ -33,8 +33,8 @@ def run(self):
 
 # 8 GPUs: (2 nodes of 4 x v100)
 app = L.LightningApp(
-    LiteMultiNode(
-        LitePyTorchDistributed,
+    FabricMultiNode(
+        FabricPyTorchDistributed,
         cloud_compute=L.CloudCompute("gpu-fast-multi"),  # 4 x V100
         num_nodes=2,
     )

pyproject.toml

@@ -99,7 +99,7 @@ warn_no_return = "False"
 # the list can be generated with:
 # mypy --no-error-summary 2>&1 | tr ':' ' ' | awk '{print $1}' | sort | uniq | sed 's/\.py//g; s|src/||g;  s|\/|\.|g' | xargs -I {} echo '"{}",'
 module = [
-    "lightning_app.components.multi_node.lite",
+    "lightning_app.components.multi_node.fabric",
     "lightning_app.components.multi_node.base",
     "lightning_app.components.multi_node.pytorch_spawn",
     "lightning_app.components.multi_node.trainer",

src/lightning_app/CHANGELOG.md

@@ -18,6 +18,8 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 - Add support for async predict method in PythonServer and remove torch context ([#16453](https://github.com/Lightning-AI/lightning/pull/16453))
 
+- Renamed `lightning.app.components.LiteMultiNode` to `lightning.app.components.FabricMultiNode` ([#16505](https://github.com/Lightning-AI/lightning/pull/16505))
+
 
 ### Deprecated
 

src/lightning_app/components/__init__.py

@@ -1,8 +1,8 @@
 from lightning_app.components.database.client import DatabaseClient
 from lightning_app.components.database.server import Database
 from lightning_app.components.multi_node import (
+    FabricMultiNode,
     LightningTrainerMultiNode,
-    LiteMultiNode,
     MultiNode,
     PyTorchSpawnMultiNode,
 )
@@ -33,7 +33,7 @@
     "Category",
     "Text",
     "MultiNode",
-    "LiteMultiNode",
+    "FabricMultiNode",
     "LightningTrainerScript",
     "PyTorchLightningScriptRunner",
     "PyTorchSpawnMultiNode",

src/lightning_app/components/multi_node/__init__.py

@@ -1,6 +1,6 @@
 from lightning_app.components.multi_node.base import MultiNode
-from lightning_app.components.multi_node.lite import LiteMultiNode
+from lightning_app.components.multi_node.fabric import FabricMultiNode
 from lightning_app.components.multi_node.pytorch_spawn import PyTorchSpawnMultiNode
 from lightning_app.components.multi_node.trainer import LightningTrainerMultiNode
 
-__all__ = ["LiteMultiNode", "MultiNode", "PyTorchSpawnMultiNode", "LightningTrainerMultiNode"]
+__all__ = ["FabricMultiNode", "MultiNode", "PyTorchSpawnMultiNode", "LightningTrainerMultiNode"]

src/lightning_app/components/multi_node/lite.py renamed to src/lightning_app/components/multi_node/fabric.py

@@ -28,7 +28,7 @@
 
 
 @runtime_checkable
-class _LiteWorkProtocol(Protocol):
+class _FabricWorkProtocol(Protocol):
     @staticmethod
     def run() -> None:
         ...
@@ -71,11 +71,11 @@ def run(
         os.environ["LOCAL_WORLD_SIZE"] = str(nprocs)
         os.environ["TORCHELASTIC_RUN_ID"] = "1"
 
-        # Used to force Lite to setup the distributed environnement.
+        # Used to force Fabric to setup the distributed environnement.
         os.environ["LT_CLI_USED"] = "1"
 
-        # Used to pass information to Lite directly.
-        def pre_fn(lite, *args, **kwargs):
+        # Used to pass information to Fabric directly.
+        def pre_fn(fabric, *args, **kwargs):
             kwargs["devices"] = nprocs
             kwargs["num_nodes"] = num_nodes
 
@@ -110,7 +110,7 @@ def pre_fn(lite, *args, **kwargs):
         return ret_val
 
 
-class LiteMultiNode(MultiNode):
+class FabricMultiNode(MultiNode):
     def __init__(
         self,
         work_cls: Type["LightningWork"],
@@ -119,7 +119,7 @@ def __init__(
         *work_args: Any,
         **work_kwargs: Any,
     ) -> None:
-        assert issubclass(work_cls, _LiteWorkProtocol)
+        assert issubclass(work_cls, _FabricWorkProtocol)
 
         # Note: Private way to modify the work run executor
         # Probably exposed to the users in the future if needed.