[Bugfix][Kernel] Skip the fused silu-mul block-quant fast path when a swiglu clamp is set (#57984)

Signed-off-by: Garrett Goon <garrett@primeintellect.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Garrett Goon
2026-09-21 22:38:38 +00:00
committed by GitHub
co-authored by Claude Opus 5
parent 8644d2af2f
commit 8bf051cd51
2 changed files with 58 additions and 0 deletions
@@ -479,8 +479,11 @@ class TritonExperts(LoRAExpertsMixin, mk.FusedMoEExpertsModular):
# Fuse SiLU+Mul + FP8 block quantize into a single kernel
# when conditions permit (gated SiLU, fp8 block quant with
# group_size=128, no LoRA requiring the BF16 intermediate).
# The fused kernel has no clamp parameter, so a configured
# SwiGLU clamp limit falls through to the unfused path.
if (
activation == MoEActivation.SILU
and self.activation_config.clamp_limit is None
and self.quant_config.use_fp8_w8a8
and self.block_shape == [128, 128]
and lora_context is None