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 15, 19:45 UTC
[Misc] Ensure out-of-tree quantization method recognize by cli args (#14328)
Signed-off-by: liuyanyi <wolfsonliu@163.com>
Yanyi Liu committed
Mar 9, 2025 at 20:13 UTC
a21076ed3a4077e79afe0a3b422f89f9a920154d
1 file changed
+4
-4
vllm/engine/arg_utils.py
+4
-4
@@ -1478,15 +1478,15 @@ class AsyncEngineArgs(EngineArgs):
1478
@staticmethod
1479
def add_cli_args(parser: FlexibleArgumentParser,
1480
async_args_only: bool = False) -> FlexibleArgumentParser:
1481
+ # Initialize plugin to update the parser, for example, The plugin may
1482
+ # adding a new kind of quantization method to --quantization argument or
1483
+ # a new device to --device argument.
1484
+ load_general_plugins()
1485
if not async_args_only:
1486
parser = EngineArgs.add_cli_args(parser)
1487
parser.add_argument('--disable-log-requests',
1488
action='store_true',
1489
help='Disable logging requests.')
1486
- # Initialize plugin to update the parser, for example, The plugin may
1487
- # adding a new kind of quantization method to --quantization argument or
1488
- # a new device to --device argument.
1489
- load_general_plugins()
1490
from vllm.platforms import current_platform
1491
current_platform.pre_register_and_update(parser)
1492
return parser