[DSV4] Add BF16 and MXFP8 A2A support for flashinfer a2a one sided (#40960)

Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Signed-off-by: Zijing Liu <liuzijing2014@gmail.com>
Co-authored-by: Zijing Liu <liuzijing2014@users.noreply.github.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Yongye Zhu
2026-04-30 15:33:12 -07:00
committed by GitHub
co-authored by Zijing Liu Claude Opus 4.7
parent 526927be94
commit b4806c8ee1
12 changed files with 107 additions and 68 deletions
@@ -254,6 +254,8 @@ class FusedMoEQuantConfig:
gemm1_beta: float | None = None
gemm1_clamp_limit: float | None = None
mx_alignment: int = 0
def __post_init__(self):
assert not self.per_act_token_quant or self.block_shape is None, (
"illegal quantization"
@@ -712,6 +714,7 @@ def mxfp4_mxfp8_moe_quant_config(
gemm1_alpha: float | None = None,
gemm1_beta: float | None = None,
gemm1_clamp_limit: float | None = None,
mx_alignment: int = 0,
) -> FusedMoEQuantConfig:
"""
Construct a quant config for mxfp4 activations and mxfp4 weights.
@@ -724,6 +727,7 @@ def mxfp4_mxfp8_moe_quant_config(
gemm1_alpha=gemm1_alpha,
gemm1_beta=gemm1_beta,
gemm1_clamp_limit=gemm1_clamp_limit,
mx_alignment=mx_alignment,
)