diff --git a/vllm/model_executor/layers/fused_moe/utils.py b/vllm/model_executor/layers/fused_moe/utils.py
index a682f848b0c4..e5957474630c 100644
--- a/vllm/model_executor/layers/fused_moe/utils.py
+++ b/vllm/model_executor/layers/fused_moe/utils.py
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import functools
 from math import prod
 
 import torch
@@ -325,5 +326,6 @@ def activation_without_mul(activation: str) -> str:
 # Torch custom ops can't deal with outputs aliasing inputs so we need to
 # disable inplace for torch >= 2.9.
 # See https://github.com/vllm-project/vllm/issues/26378
+@functools.cache
 def disable_inplace() -> bool:
     return is_torch_equal_or_newer("2.9")
diff --git a/vllm/model_executor/models/bert.py b/vllm/model_executor/models/bert.py
index df302aee0bf6..6e81eb8dc91b 100644
--- a/vllm/model_executor/models/bert.py
+++ b/vllm/model_executor/models/bert.py
@@ -609,8 +609,8 @@ class SPLADESparsePooler(Pooler):
     def __init__(
         self,
         mlm_head: nn.Module,
-        cls_token_id: Optional[int] = 101,
-        sep_token_id: Optional[int] = 102,
+        cls_token_id: int | None = 101,
+        sep_token_id: int | None = 102,
         pooling: str = "max",
         remove_cls_sep: bool = True,
     ):