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 16, 03:15 UTC
fix init error for MessageQueue when n_local_reader is zero (#11768)
XiaobingZhang committed
Jan 7, 2025 at 14:12 UTC
e512f76a898d61b8857b36b138bb9cf93ea04d03
1 file changed
+2
-1
vllm/distributed/device_communicators/shm_broadcast.py
+2
-1
@@ -247,7 +247,8 @@ class MessageQueue:
247
self.handle = Handle(
248
connect_ip=connect_ip,
249
local_reader_ranks=local_reader_ranks,
250
- buffer_handle=self.buffer.handle(),
250
+ buffer_handle=self.buffer.handle()
251
+ if self.buffer is not None else None,
252
local_subscribe_port=local_subscribe_port,
253
remote_subscribe_port=remote_subscribe_port,
254
)