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 17, 05:30 UTC
[Minor] Fix false warning when TP=1 (#2674)
Woosuk Kwon committed
Jan 30, 2024 at 14:39 UTC
105a40f53a8001ce7f4282462a4636b31383f393
1 file changed
+4
vllm/model_executor/parallel_utils/custom_all_reduce.py
+4
@@ -29,6 +29,10 @@ def init_custom_ar() -> None:
29
return
30
rank = get_tensor_model_parallel_rank()
31
world_size = get_tensor_model_parallel_world_size()
32
+ if world_size == 1:
33
+ # No need to initialize custom allreduce for single GPU case.
34
+ return
35
+
36
if world_size not in _SUPPORTED_WORLD_SIZES:
37
logger.warn(
38
"Custom allreduce is disabled due to an unsupported world size: "