[Bugfix] fix swiglu limit issue for humming backend + deepseek v4 (#42541)

Signed-off-by: Jinzhen Lin <jinzhen.ljz@antgroup.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
This commit is contained in:
Jinzhen Lin
2026-05-18 17:32:26 +00:00
committed by GitHub
co-authored by Michael Goin
parent 67f58ce23f
commit 6859ca7615
3 changed files with 34 additions and 6 deletions
@@ -164,7 +164,12 @@ def prepare_humming_moe_layer(layer: RoutedExperts, quant_config: dict):
layer.register_buffer("locks", locks)
def get_humming_moe_quant_config(layer: RoutedExperts):
def get_humming_moe_quant_config(
layer: RoutedExperts,
gemm1_alpha: float | None = None,
gemm1_beta: float | None = None,
gemm1_clamp_limit: float | None = None,
):
input_schema = layer.input_schemas["w13"]
weight_schema = layer.weight_schemas["w13"]
@@ -211,4 +216,7 @@ def get_humming_moe_quant_config(layer: RoutedExperts):
_a2=a_quant_desc,
_w1=w1_quant_desc,
_w2=w2_quant_desc,
gemm1_alpha=gemm1_alpha,
gemm1_beta=gemm1_beta,
gemm1_clamp_limit=gemm1_clamp_limit,
)