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, 07:00 UTC
[Doc] Fix typo in arg_utils.py (#10264)
Signed-off-by: Xin Yang <xyang19@gmail.com>
Xin Yang committed
Nov 12, 2024 at 21:54 UTC
032fcf16ae9d924cc98a083c3c8464173f87a49e
1 file changed
+2
-2
vllm/engine/arg_utils.py
+2
-2
@@ -626,8 +626,8 @@ class EngineArgs:
626
type=int,
627
default=EngineArgs.max_cpu_loras,
628
help=('Maximum number of LoRAs to store in CPU memory. '
629
- 'Must be >= than max_num_seqs. '
630
- 'Defaults to max_num_seqs.'))
629
+ 'Must be >= than max_loras. '
630
+ 'Defaults to max_loras.'))
631
parser.add_argument(
632
'--fully-sharded-loras',
633
action='store_true',