[Kernel] Support Nvfp4 Cutedsl Moe Swiglu-oai and Relu2(non-gated) Activation (#47106)

Signed-off-by: Tiekai Bi <tiekaib@nvidia.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Co-authored-by: Yongye Zhu <zyy1102000@gmail.com>
This commit is contained in:
Bi Tiekai
2026-08-06 17:42:45 -07:00
committed by GitHub
co-authored by Claude Opus 5 Yongye Zhu
parent b1e12d142d
commit e08111211b
9 changed files with 216 additions and 14 deletions
@@ -1591,6 +1591,8 @@ class ModelOptNvFp4FusedMoE(FusedMoEMethodBase):
a13_scale=layer.w13_input_scale,
a2_scale=layer.w2_input_scale,
swiglu_limit=getattr(layer, "swiglu_limit", None),
swiglu_alpha=getattr(layer, "swiglu_alpha", None),
swiglu_beta=getattr(layer, "swiglu_beta", None),
layer=layer,
)