add context manager that safely patches :func:`torch.cuda.is_availabl…

…e` with its NVML-based version if possible

src/lightning_lite/accelerators/cuda.py

@@ -13,8 +13,9 @@
 # limitations under the License.
 import os
 import warnings
+from contextlib import contextmanager
 from functools import lru_cache
-from typing import Dict, List, Optional, Set, Union
+from typing import Dict, Generator, List, Optional, Set, Union
 
 import torch
 
@@ -77,6 +78,25 @@ def _get_all_available_cuda_gpus() -> List[int]:
     return list(range(num_cuda_devices()))
 
 
+@contextmanager
+def patch_cuda_is_available() -> Generator:
+    """Context manager that safely patches :func:`torch.cuda.is_available` with its NVML-based version if
+    possible."""
+    orig_check = None
+    new_check = torch.cuda.device_count if _TORCH_GREATER_EQUAL_1_13 else _device_count_nvml
+
+    if hasattr(torch._C, "_cuda_getDeviceCount") and _device_count_nvml() >= 0:
+        # we can safely patch is_available if both torch has CUDA compiled and the NVML count is succeeding
+        # otherwise, patching is_available could lead to attribute errors or infinite recursion
+        orig_check = torch.cuda.is_available
+        torch.cuda.is_available = new_check  # type: ignore[assignment]
+    try:
+        yield
+    finally:
+        if orig_check:
+            torch.cuda.is_available = orig_check
+
+
 @lru_cache(1)
 def num_cuda_devices() -> int:
     """Returns the number of available CUDA devices.

src/lightning_lite/plugins/precision/native_amp.py

@@ -20,7 +20,7 @@
 from torch.optim import LBFGS
 from typing_extensions import Literal
 
-from lightning_lite.accelerators.cuda import is_cuda_available
+from lightning_lite.accelerators.cuda import patch_cuda_is_available
 from lightning_lite.plugins.precision.precision import Precision
 from lightning_lite.plugins.precision.utils import _convert_fp_tensor
 from lightning_lite.utilities.imports import _TORCH_GREATER_EQUAL_1_10
@@ -48,9 +48,9 @@ def __init__(
         if precision == "bf16" and not _TORCH_GREATER_EQUAL_1_10:
             raise ImportError("To use bfloat16 with native amp you must install torch greater or equal to 1.10.")
         if scaler is None and precision == 16:
-            # if possible, we defer CUDA initialization to support strategies that will attempt forks
-            torch.cuda.is_available = is_cuda_available
-            scaler = torch.cuda.amp.GradScaler()
+            with patch_cuda_is_available():
+                # if possible, we defer CUDA initialization to support strategies that will attempt forks
+                scaler = torch.cuda.amp.GradScaler()
         if scaler is not None and precision == "bf16":
             raise ValueError(f"`precision='bf16'` does not use a scaler, found {scaler}.")
         self.precision = precision

src/pytorch_lightning/plugins/precision/native_amp.py

@@ -19,7 +19,7 @@
 from torch.optim import LBFGS
 
 import pytorch_lightning as pl
-from lightning_lite.accelerators.cuda import is_cuda_available
+from lightning_lite.accelerators.cuda import patch_cuda_is_available
 from lightning_lite.utilities.types import Steppable
 from pytorch_lightning.plugins.precision.precision_plugin import PrecisionPlugin
 from pytorch_lightning.utilities import _TORCH_GREATER_EQUAL_1_10, AMPType
@@ -51,9 +51,9 @@ def __init__(
                 "To use bfloat16 with native amp you must install torch greater or equal to 1.10."
             )
         if scaler is None and precision == 16:
-            # if possible, we defer CUDA initialization to support strategies that will attempt forks
-            torch.cuda.is_available = is_cuda_available
-            scaler = torch.cuda.amp.GradScaler()
+            with patch_cuda_is_available():
+                # if possible, we defer CUDA initialization to support strategies that will attempt forks
+                scaler = torch.cuda.amp.GradScaler()
         if scaler is not None and precision == "bf16":
             raise MisconfigurationException(f"`precision='bf16'` does not use a scaler, found {scaler}.")
         self.precision = precision