A high-throughput and memory-efficient inference and serving engine for LLMs
fix: prevent MM cache hang from stale LRU order keys (#43595)
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
jeff.ye committed
Jun 9, 2026 at 13:48 UTC
7c2aa3108ab405d98101b931fc6a0298f98365cc
1 file changed
+1
-3
vllm/utils/cache.py
+1
-3
@@ -118,10 +118,8 @@ class LRUCache(cachetools.LRUCache[_K, _V]):
118
return info
119
120
def touch(self, key: _K) -> None:
121
- try:
121
+ if key in self:
122
self._LRUCache__order.move_to_end(key) # type: ignore
123
- except KeyError:
124
- self._LRUCache__order[key] = None # type: ignore
123
124
@overload
125
def get(self, key: _K, /) -> _V | None: ...