Various test fixes (#15068)

src/lightning_lite/accelerators/cuda.py

@@ -82,7 +82,7 @@ def _get_all_available_cuda_gpus() -> List[int]:
 def _patch_cuda_is_available() -> Generator:
     """Context manager that safely patches :func:`torch.cuda.is_available` with its NVML-based version if
     possible."""
-    if hasattr(torch._C, "_cuda_getDeviceCount") and _device_count_nvml() >= 0 and not _TORCH_GREATER_EQUAL_1_13:
+    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

src/lightning_lite/utilities/distributed.py

@@ -136,10 +136,10 @@ def sync_ddp(result: Tensor, group: Optional[Any] = None, reduce_op: Optional[Un
 
     # WA for HPU. HPU doesn't support Long types, forcefully set it to float
     if module_available("habana_frameworks.torch.utils.library_loader"):
-        from habana_frameworks.torch.utils.library_loader import is_habana_available
+        from habana_frameworks.torch.utils.library_loader import is_habana_avaialble
 
         if (
-            is_habana_available()
+            is_habana_avaialble()
             and os.environ.get("HCCL_DISTRIBUTED_BACKEND") == "1"
             and result.type() in ("torch.LongTensor", "torch.hpu.LongTensor")
         ):

tests/tests_pytorch/serve/test_servable_module_validator.py

@@ -1,8 +1,9 @@
 from typing import Dict
 
+import pytest
 import torch
 
-from pytorch_lightning import seed_everything, Trainer
+from pytorch_lightning import Trainer
 from pytorch_lightning.demos.boring_classes import BoringModel
 from pytorch_lightning.serve.servable_module_validator import ServableModule, ServableModuleValidator
 
@@ -29,14 +30,13 @@ class ServableBoringModel(BoringModel, ServableModule):
 
 
 def test_servable_module_validator():
-    seed_everything(42)
     model = ServableBoringModel()
     callback = ServableModuleValidator()
     callback.on_train_start(Trainer(), model)
 
 
+@pytest.mark.flaky(reruns=3)
 def test_servable_module_validator_with_trainer(tmpdir):
-    seed_everything(42)
     callback = ServableModuleValidator()
     trainer = Trainer(
         default_root_dir=tmpdir,