[pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

Author: pre-commit-ci[bot]

Diff for: src/lightning/fabric/accelerators/cpu.py

@@ -49,7 +49,7 @@ def get_parallel_devices(devices: Union[int, str, List[int]]) -> List[torch.devi
         """Gets parallel devices for the Accelerator."""
         devices = _parse_cpu_cores(devices)
         return [torch.device("cpu")] * devices
-    
+
     @staticmethod
     @override
     def get_device() -> str:

Diff for: src/lightning/fabric/accelerators/cuda.py

@@ -59,7 +59,7 @@ def get_parallel_devices(devices: List[int]) -> List[torch.device]:
     @override
     def get_device() -> str:
         return "cuda"
-    
+
     @staticmethod
     @override
     def auto_device_count() -> int:

Diff for: src/lightning/fabric/accelerators/mps.py

@@ -64,7 +64,7 @@ def get_parallel_devices(devices: Union[int, str, List[int]]) -> List[torch.devi
     @override
     def get_device() -> str:
         return "mps"
-    
+
     @staticmethod
     @override
     def auto_device_count() -> int:

Diff for: src/lightning/fabric/accelerators/xla.py

@@ -68,7 +68,7 @@ def get_parallel_devices(devices: Union[int, List[int]]) -> List[torch.device]:
     @override
     def get_device() -> str:
         return "xla"
-    
+
     @staticmethod
     @override
     # XLA's multiprocessing will pop the TPU_NUM_DEVICES key, so we need to cache it

Diff for: src/lightning/fabric/connector.py

@@ -464,9 +464,9 @@ def _check_and_init_precision(self) -> Precision:
             return DeepSpeedPrecision(self._precision_input)  # type: ignore
         if isinstance(self.strategy, FSDPStrategy):
             return FSDPPrecision(
-                precision=self._precision_input, # type: ignore[arg-type]
+                precision=self._precision_input,  # type: ignore[arg-type]
                 device=self._accelerator_flag.get_device() if isinstance(self._accelerator_flag, Accelerator) else None,
-            )  
+            )
         mp_precision_supported = ("32-true", "bf16-mixed", "bf16-true", "16-true")
         if isinstance(self.strategy, ModelParallelStrategy) and self._precision_input not in mp_precision_supported:
             raise ValueError(

Diff for: src/lightning/fabric/plugins/precision/amp.py

@@ -55,8 +55,7 @@ def __init__(
                 if _TORCH_GREATER_EQUAL_2_4
                 else getattr(
                     torch,
-                    "cuda" if  not isinstance(device, str) or device.split(":")[0] == "cpu"
-                    else device.split(":")[0]
+                    "cuda" if not isinstance(device, str) or device.split(":")[0] == "cpu" else device.split(":")[0],
                 ).amp.GradScaler()
             )
         if scaler is not None and self.precision == "bf16-mixed":

Diff for: src/lightning/fabric/strategies/deepspeed.py

@@ -619,8 +619,10 @@ def _initialize_engine(
     @override
     def setup_environment(self) -> None:
         from deepspeed.runtime.utils import get_accelerator
-        if (not isinstance(self.accelerator, CUDAAccelerator)) and \
-            self.accelerator.get_device() != get_accelerator().device_name(): # type: ignore[union-attr]
+
+        if (
+            not isinstance(self.accelerator, CUDAAccelerator)
+        ) and self.accelerator.get_device() != get_accelerator().device_name():  # type: ignore[union-attr]
             raise RuntimeError(
                 f"The DeepSpeed strategy is only supported on {get_accelerator().device_name()} GPUs,"
                 f"but `{self.accelerator.__class__.__name__}` is used."

Diff for: src/lightning/pytorch/plugins/precision/amp.py

@@ -55,8 +55,7 @@ def __init__(
                 if _TORCH_GREATER_EQUAL_2_4
                 else getattr(
                     torch,
-                    "cuda" if not isinstance(device, str) or device.split(":")[0] == "cpu"
-                    else device.split(":")[0]
+                    "cuda" if not isinstance(device, str) or device.split(":")[0] == "cpu" else device.split(":")[0],
                 ).amp.GradScaler()
             )
         if scaler is not None and self.precision == "bf16-mixed":

Diff for: src/lightning/pytorch/strategies/deepspeed.py

@@ -317,8 +317,10 @@ def __init__(
     @override
     def setup_environment(self) -> None:
         from deepspeed.runtime.utils import get_accelerator
-        if (not isinstance(self.accelerator, CUDAAccelerator)) and \
-            self.accelerator.get_device() != get_accelerator().device_name(): # type: ignore[union-attr]
+
+        if (
+            not isinstance(self.accelerator, CUDAAccelerator)
+        ) and self.accelerator.get_device() != get_accelerator().device_name():  # type: ignore[union-attr]
             raise RuntimeError(
                 f"The DeepSpeed strategy is only supported on {get_accelerator().device_name()} GPUs,"
                 f"but `{self.accelerator.__class__.__name__}` is used."

Diff for: src/lightning/pytorch/trainer/connectors/accelerator_connector.py

@@ -508,7 +508,7 @@ def _check_and_init_precision(self) -> Precision:
             return DeepSpeedPrecision(self._precision_flag)  # type: ignore[arg-type]
         if isinstance(self.strategy, FSDPStrategy):
             return FSDPPrecision(
-                precision=self._precision_flag, # type: ignore[arg-type]
+                precision=self._precision_flag,  # type: ignore[arg-type]
                 device=self._accelerator_flag.get_device() if isinstance(self._accelerator_flag, Accelerator) else None,
             )
         if self._precision_flag in ("16-true", "bf16-true"):

Diff for: tests/tests_fabric/test_connector.py

@@ -178,7 +178,7 @@ def parse_devices(devices):
         @staticmethod
         def get_parallel_devices(devices):
             return [torch.device("cpu")] * devices
-        
+
         @staticmethod
         def get_device() -> str:
             return "cpu"

Diff for: tests/tests_pytorch/trainer/connectors/test_accelerator_connector.py

@@ -191,7 +191,7 @@ def parse_devices(devices):
         @staticmethod
         def get_parallel_devices(devices):
             return [torch.device("cpu")] * devices
-        
+
         @staticmethod
         def get_device() -> str:
             return "cpu"