[Train] Try to fix the timeout for test_result

python/ray/train/v2/_internal/execution/storage.py

@@ -313,6 +313,17 @@ class StorageContext:
     """Shared context that holds the source of truth for all paths and
     storage utilities, passed along from the driver to workers.
 
+    Args:
+        storage_path: Path where all results and checkpoints are persisted.
+            Can be a local directory or a remote URI (e.g., s3://bucket/path).
+        experiment_dir_name: Name of the experiment directory within the storage path.
+        storage_filesystem: Optional custom PyArrow filesystem to use for storage.
+            If not provided, will be auto-resolved from the storage_path URI.
+        skip_validation: If True, skips creating and checking the storage validation
+            marker file. This should be set to True for read-only operations (e.g.,
+            restoring from an existing experiment directory) to avoid unnecessary
+            file system writes. Defaults to False.
+
     This object defines a few types of paths:
     1. *_fs_path: A path on the `storage_filesystem`. This is a regular path
         which has been prefix-stripped by pyarrow.fs.FileSystem.from_uri and
@@ -383,6 +394,7 @@ def __init__(
         storage_path: Union[str, os.PathLike],
         experiment_dir_name: str,
         storage_filesystem: Optional[pyarrow.fs.FileSystem] = None,
+        skip_validation: bool = False,
     ):
         self.custom_fs_provided = storage_filesystem is not None
 
@@ -395,8 +407,9 @@ def __init__(
         )
         self.storage_fs_path = Path(self.storage_fs_path).as_posix()
 
-        self._create_validation_file()
-        self._check_validation_file()
+        if not skip_validation:
+            self._create_validation_file()
+            self._check_validation_file()
 
     def __str__(self):
         return (

python/ray/train/v2/api/result.py

@@ -68,6 +68,7 @@ def from_path(
             storage_path=storage_path,
             experiment_dir_name=experiment_dir_name,
             storage_filesystem=fs,
+            skip_validation=True,
         )
 
         # Validate that the checkpoint manager snapshot file exists

python/ray/train/v2/tests/test_result.py

@@ -160,6 +160,7 @@ def test_result_restore(
         storage_context = StorageContext(
             storage_path=storage_path,
             experiment_dir_name=exp_name,
+            skip_validation=True,
         )
 
         trial_dir = storage_context.experiment_fs_path
@@ -187,12 +188,11 @@ def test_result_restore(
     assert len(result.best_checkpoints) == num_checkpoints
 
     """
-    Top-3 checkpoints with metrics:
+    Top-2 checkpoints with metrics:
 
                         | iter   | metric_a    metric_b
-    checkpoint_000004        4            4          -4
-    checkpoint_000003        3            3          -3
     checkpoint_000002        2            2          -2
+    checkpoint_000001        1            1          -1
     """
     # Check if the checkpoints bounded with correct metrics
     best_ckpt_a = result.get_best_checkpoint(metric="metric_a", mode="max")