Sm120 scatter fallback (#31669)
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
This commit is contained in:
co-authored by
github-actions[bot]
parent
c4c405a46b
commit
01f558d905
@@ -17,11 +17,17 @@ from triton_kernels.matmul_ogs import (
|
|||||||
ScatterIndx,
|
ScatterIndx,
|
||||||
matmul_ogs,
|
matmul_ogs,
|
||||||
)
|
)
|
||||||
|
from triton_kernels.matmul_ogs_details.opt_flags import update_opt_flags_constraints
|
||||||
from triton_kernels.numerics import InFlexData
|
from triton_kernels.numerics import InFlexData
|
||||||
from triton_kernels.swiglu import swiglu_fn
|
from triton_kernels.swiglu import swiglu_fn
|
||||||
from triton_kernels.tensor import FP4
|
from triton_kernels.tensor import FP4
|
||||||
|
|
||||||
from sglang.srt.utils import is_cuda
|
from sglang.srt.utils import is_cuda
|
||||||
|
from sglang.srt.utils.common import is_sm120_supported
|
||||||
|
|
||||||
|
if is_sm120_supported():
|
||||||
|
# use the regular gather/scatter implementation for unsupported devices.
|
||||||
|
update_opt_flags_constraints({"is_persistent": False})
|
||||||
|
|
||||||
if is_cuda():
|
if is_cuda():
|
||||||
from sglang.jit_kernel.activation import gelu_and_mul, silu_and_mul
|
from sglang.jit_kernel.activation import gelu_and_mul, silu_and_mul
|
||||||
|
|||||||
Reference in New Issue
Block a user