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, 09:30 UTC
[Doc] update gpu-memory-utilization flag docs (#9507)
Signed-off-by: Joe Runde <Joseph.Runde@ibm.com>
Joe Runde committed
Oct 18, 2024 at 22:26 UTC
82c25151ec54f723de8589ccc3ad24d4a1817e90
1 file changed
+5
-1
vllm/engine/arg_utils.py
+5
-1
@@ -428,7 +428,11 @@ class EngineArgs:
428
help='The fraction of GPU memory to be used for the model '
429
'executor, which can range from 0 to 1. For example, a value of '
430
'0.5 would imply 50%% GPU memory utilization. If unspecified, '
431
- 'will use the default value of 0.9.')
431
+ 'will use the default value of 0.9. This is a global gpu memory '
432
+ 'utilization limit, for example if 50%% of the gpu memory is '
433
+ 'already used before vLLM starts and --gpu-memory-utilization is '
434
+ 'set to 0.9, then only 40%% of the gpu memory will be allocated '
435
+ 'to the model executor.')
436
parser.add_argument(
437
'--num-gpu-blocks-override',
438
type=int,