Support logging expert balancedness metrics (#6482)

This commit is contained in:
fzyzcjy
2025-05-22 14:05:33 +08:00
committed by GitHub
parent e9feb48838
commit 9484eba4ad
2 changed files with 113 additions and 1 deletions

View File

@@ -177,6 +177,7 @@ class ServerArgs:
Literal["stat", "per_pass", "per_token"]
] = None
expert_distribution_recorder_buffer_size: Optional[int] = None
enable_expert_distribution_metrics: bool = False
deepep_config: Optional[str] = None
enable_torch_compile: bool = False
torch_compile_max_bs: int = 32
@@ -1304,6 +1305,11 @@ class ServerArgs:
default=ServerArgs.expert_distribution_recorder_buffer_size,
help="Circular buffer size of expert distribution recorder. Set to -1 to denote infinite buffer.",
)
parser.add_argument(
"--enable-expert-distribution-metrics",
action="store_true",
help="Enable logging metrics for expert balancedness",
)
parser.add_argument(
"--deepep-config",
type=str,