check tensor address in the custom kernel launcher

src/pyjuice/utils/kernel_launcher.py

@@ -34,6 +34,13 @@ def wrapper(*args, **kwargs):
                 if k in self.constexpr_names:
                     signature_list.append((self.constexpr_names[k], v))
 
+            if "batch_size" in kwargs:
+                signature_list.append(("batch_size", kwargs["batch_size"]))
+
+            for i, arg in enumerate(args):
+                if isinstance(arg, torch.Tensor):
+                    signature_list.append((i, id(arg)))
+
             grid0 = grid[0]
             grid1 = grid[1] if len(grid) > 1 else 1
             grid2 = grid[2] if len(grid) > 2 else 1