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 11, 15:00 UTC
[LMCache] Relax lmcache version requirement (#30425)
Signed-off-by: Nick Hill <nhill@redhat.com>
Nick Hill committed
Dec 11, 2025 at 19:18 UTC
947dfda9c281c2b2d779a29e73bbc20170dcfab3
2 files changed
+9
-2
requirements/kv_connectors.txt
+1
-1
@@ -1,2 +1,2 @@
1
-lmcache >= 0.3.10.post1
1
+lmcache
2
nixl >= 0.7.1 # Required for disaggregated prefill
vllm/distributed/kv_transfer/kv_connector/v1/lmcache_integration/vllm_v1_adapter.py
+8
-1
@@ -27,7 +27,14 @@ from lmcache.v1.lookup_client.lmcache_async_lookup_client import (
27
LMCacheAsyncLookupServer,
28
)
29
from lmcache.v1.offload_server.zmq_server import ZMQOffloadServer
30
-from lmcache.v1.plugin.runtime_plugin_launcher import RuntimePluginLauncher
30
+
31
+try:
32
+ from lmcache.v1.plugin.runtime_plugin_launcher import RuntimePluginLauncher
33
+except ImportError:
34
+ # Backwards compatibility for lmcache <= 0.3.10-post1
35
+ from lmcache.v1.plugin.plugin_launcher import (
36
+ PluginLauncher as RuntimePluginLauncher,
37
+ )
38
39
from vllm.attention.backends.abstract import AttentionMetadata
40
from vllm.config import VllmConfig