[Kernel] Add support for Flashinfer Mamba SSU algorithm selection (#50157)

Signed-off-by: amitz-nv <203509407+amitz-nv@users.noreply.github.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
This commit is contained in:
amitz-nv
2026-08-04 09:40:23 -07:00
committed by GitHub
co-authored by OpenAI Codex
parent 7ac2ec7582
commit 52c0e3cb08
4 changed files with 88 additions and 5 deletions
+8 -1
View File
@@ -77,7 +77,7 @@ from vllm.config.device import Device
from vllm.config.kernel import IrOpPriorityConfig, LinearBackend, MoEBackend
from vllm.config.load import SafetensorsLoadStrategy
from vllm.config.lora import MaxLoRARanks
from vllm.config.mamba import MambaBackendEnum
from vllm.config.mamba import MambaBackendEnum, MambaSSUAlgorithm
from vllm.config.model import (
ConvertOption,
HfOverrides,
@@ -707,6 +707,7 @@ class EngineArgs:
use_replayssm: bool = CacheConfig.use_replayssm
mamba_backend: MambaBackendEnum = MambaBackendEnum.TRITON
mamba_ssu_algorithm: MambaSSUAlgorithm | None = None
enable_mamba_cache_stochastic_rounding: bool = (
MambaConfig.enable_stochastic_rounding
)
@@ -961,6 +962,9 @@ class EngineArgs:
description=MambaConfig.__doc__,
)
mamba_group.add_argument("--mamba-backend", **mamba_kwargs["backend"])
mamba_group.add_argument(
"--mamba-ssu-algorithm", **mamba_kwargs["ssu_algorithm"]
)
mamba_group.add_argument(
"--enable-mamba-cache-stochastic-rounding",
**mamba_kwargs["enable_stochastic_rounding"],
@@ -2356,6 +2360,8 @@ class EngineArgs:
mamba_config.backend = MambaBackendEnum[self.mamba_backend.upper()]
else:
mamba_config.backend = self.mamba_backend
if self.mamba_ssu_algorithm is not None:
mamba_config.ssu_algorithm = self.mamba_ssu_algorithm
if self.enable_mamba_cache_stochastic_rounding:
mamba_config.enable_stochastic_rounding = (
self.enable_mamba_cache_stochastic_rounding
@@ -2364,6 +2370,7 @@ class EngineArgs:
mamba_config.stochastic_rounding_philox_rounds = (
self.mamba_cache_philox_rounds
)
mamba_config.validate_ssu_algorithm()
# Kernel config overrides
kernel_config = copy.deepcopy(self.kernel_config)