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 14, 05:30 UTC
Remove redundant assignment in xfer_buffers, This is a little fix (#24732)
Signed-off-by: ChenTaoyu-SJTU <ctynb@qq.com>
TaoYu Chen committed
Sep 13, 2025 at 16:11 UTC
15b8fef453b373b84406207a947005a4d9d68acc
1 file changed
-2
vllm/distributed/kv_transfer/kv_connector/v1/nixl_connector.py
-2
@@ -708,8 +708,6 @@ class NixlConnectorWorker:
708
caches_data = []
709
# With hybrid allocator, layers can share a kv cache tensor
710
seen_base_addresses = []
711
- xfer_buffers = (self.host_xfer_buffers
712
- if self.use_host_buffer else kv_caches)
711
712
# Note(tms): I modified this from the original region setup code.
713
# K and V are now in different regions. Advantage is that we can