[DSV4.1] Big fused wo_a quant (#39957)

Co-authored-by: BBuf <1182563586@qq.com>
This commit is contained in:
DarkSharpness
2026-09-19 19:46:53 +08:00
committed by GitHub
co-authored by BBuf
parent cb22f2451e
commit d1acbe0746
12 changed files with 1070 additions and 27 deletions
@@ -303,7 +303,7 @@ class TestBlockFp8AsMxfp8Linear(_LinearBackendCheck):
def test_mxfp8_view_and_swizzled_input(self):
if "flashinfer_cutedsl" not in _block32_backends():
self.skipTest(f"cutedsl not in SM{get_device_sm()} backend set")
from sglang.kernels.ops.attention.dsv4.wo_a_bf16 import (
from sglang.kernels.ops.attention.dsv4.wo_a import (
_quantize_partial,
_wo_a_reduce,
)