A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 13, 09:30 UTC
[KVConnector] Enable get_block_ids_with_load_errors() in LMCache connector (#27978)
Signed-off-by: Zirui Liu <ziliu@ddn.com> Signed-off-by: ziruiliu <ziliu@ddn.com> Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> Co-authored-by: Nicolò Lucchesi <nlucches@redhat.com>
ziruiliu committed
Nov 12, 2025 at 18:44 UTC
d143152308ede994406196aa484566a25a68fe3a
1 file changed
+15
vllm/distributed/kv_transfer/kv_connector/v1/lmcache_connector.py
+15
@@ -136,6 +136,21 @@ class LMCacheConnectorV1(KVConnectorBase_V1):
136
"""
137
return self._lmcache_engine.get_finished(finished_req_ids)
138
139
+ def get_block_ids_with_load_errors(self) -> set[int]:
140
+ """
141
+ Get the set of block IDs that failed to load.
142
+
143
+ Returns:
144
+ Set of block IDs that encountered load errors.
145
+ Empty set if no load errors occurred.
146
+ """
147
+ method = getattr(self._lmcache_engine, "get_block_ids_with_load_errors", None)
148
+ if callable(method):
149
+ return method()
150
+
151
+ # Fallback for older versions that don't support this method
152
+ return set()
153
+
154
# ==============================
155
# Scheduler-side methods
156
# ==============================