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 15, 08:30 UTC
Minor fix about MooncakeStoreConnector (#18721)
Signed-off-by: baoloongmao <baoloongmao@tencent.com>
maobaolong committed
May 27, 2025 at 16:02 UTC
c24b1572ac23bc5c495c3680a70d9f6c36ff7143
1 file changed
+5
-4
vllm/distributed/kv_transfer/kv_connector/mooncake_store_connector.py
+5
-4
@@ -31,12 +31,12 @@ class MooncakeStoreConnector(KVConnectorBase):
31
local_rank: int,
32
config: VllmConfig,
33
):
34
- self.config = config.kv_transfer_config
34
+ self.kv_transfer_config = config.kv_transfer_config
35
self.kv_helper = kv_helper(config)
36
self.local_tp_rank = local_rank
37
38
# Init kv_store
39
- if self.config.kv_connector == "MooncakeStoreConnector":
39
+ if self.kv_transfer_config.kv_connector == "MooncakeStoreConnector":
40
# Check if MOONCAKE_CONFIG_PATH is set
41
import os
42
use_mooncake_store = os.getenv('MOONCAKE_CONFIG_PATH') is not None
@@ -50,10 +50,11 @@ class MooncakeStoreConnector(KVConnectorBase):
50
MooncakeStore)
51
logger.info(
52
"Initializing KVStoreConnector under kv_transfer_config %s",
53
- self.config)
53
+ self.kv_transfer_config)
54
self.kv_store = MooncakeStore(config)
55
else:
56
- logger.error("Can not find %s", self.config.kv_connector)
56
+ logger.error("Can not find %s",
57
+ self.kv_transfer_config.kv_connector)
58
59
assert self.kv_store is not None
60