Making Numpy depedency in Local Tensor optional to fix broken Torchao CI (#165938)

In recent change LocalTensor introduced dependency on Numpy and has broken Torchao CI.
This dependency cna be made optional and required only when Local Tensor is used.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/165938
Approved by: https://github.com/atalman

torch/distributed/_local_tensor/__init__.py

@@ -51,8 +51,15 @@ from collections.abc import Sequence
 from types import TracebackType
 from typing import Any, Callable, Generator, Optional, Union
 
+
+try:
     import numpy as np
 
+    HAS_NUMPY = True
+except ModuleNotFoundError:
+    HAS_NUMPY = False
+    np = None  # type: ignore[assignment]
+
 import torch
 from torch import Size, SymBool, SymInt, Tensor
 from torch._C import DispatchKey, DispatchKeySet, ScriptObject
@@ -524,8 +531,13 @@ class LocalTensor(torch.Tensor):
         with LocalTensorMode(local_tensor._ranks):
             return func(*args, **kwargs)
 
-    def numpy(self, *, force: bool = False) -> np.ndarray:
+    def numpy(
+        self, *, force: bool = False
+    ) -> np.ndarray:  # pyrefly: ignore  # missing-attribute
+        if HAS_NUMPY:
             return self.reconcile().numpy(force=force)
+        else:
+            raise RuntimeError("Numpy is not available")
 
     def __lt__(
         self, other: torch.Tensor | bool | complex | float | int