Do not crash when compiling quantized LORA models (pytorch#148435)

Fixes pytorch#148072

Pull Request resolved: pytorch#148435
Approved by: https://github.com/Valentine233, https://github.com/leslie-fang-intel

Author: Whadup

torch/_inductor/fx_passes/quantization.py

@@ -1071,6 +1071,10 @@ def _register_quantization_reshape():
 def _is_valid_woq_optimization_pattern():
     def fn(match):
         assert all(k in match.kwargs for k in ("x", "weight", "scales"))
+        if not all(
+            hasattr(match.kwargs[key], "meta") for key in ["x", "weight", "scales"]
+        ):
+            return False
         x = match.kwargs["x"].meta["val"]
         weight = match.kwargs["weight"].meta["val"]
         scales = match.kwargs["scales"].meta["val"]