[AMD] Fix Quark W4A4 MXFP4 MoE clamped-SwiGLU on the AITER path (#34540)

Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
Tuan Nguyen Gia
2026-10-07 03:20:40 +00:00
committed by GitHub
co-authored by Cursor
parent 012cd3858b
commit c0df52be2d
2 changed files with 144 additions and 1 deletions
@@ -980,6 +980,7 @@ class QuarkW4A4MXFp4MoE(QuarkMoEScheme):
from sglang.srt.layers.moe.moe_runner.aiter import (
AiterMoeQuantInfo,
AiterQuantType,
aiter_swiglu_oai_limit,
)
if hasattr(torch, "float4_e2m1fn_x2"):
@@ -1015,7 +1016,11 @@ class QuarkW4A4MXFp4MoE(QuarkMoEScheme):
expert_mask=layer.dispatcher.expert_mask_gpu,
hidden_pad=getattr(layer, "hidden_pad", 0),
intermediate_pad=getattr(layer, "intermediate_pad", 0),
swiglu_limit=self.moe_runner_config.swiglu_limit or 0.0,
swiglu_limit=(
aiter_swiglu_oai_limit(self.moe_runner_config)
or self.moe_runner_config.swiglu_limit
or 0.0
),
fused_moe_kwargs=_fused_moe_kwargs,
)
return self.runner.run(dispatch_output, quant_info)