chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

[CI][NPU] Bugfix import sgl-kernel error (#18173)

Browse Source
This commit is contained in:
Even Zhou
2026-02-03 11:39:38 -08:00
committed by GitHub
parent 495290aefd
commit d48bbe3bed
1 changed files with 5 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/layers/moe/moe_runner/flashinfer_trtllm.py
+5 -1
View File
@@ -22,6 +22,7 @@ from sglang.srt.layers.quantization.fp8_kernel import (
scaled_fp8_quant, scaled_fp8_quant,
) )
from sglang.srt.utils.common import ( from sglang.srt.utils.common import (
is_cuda_alike,
is_flashinfer_available, is_flashinfer_available,
is_sm120_supported, is_sm120_supported,
next_power_of_2, next_power_of_2,
@@ -32,10 +33,13 @@ if TYPE_CHECKING:
StandardCombineInput, StandardCombineInput,
StandardDispatchOutput, StandardDispatchOutput,
) )
if is_flashinfer_available() and is_sm120_supported(): if is_flashinfer_available() and is_sm120_supported():
from flashinfer import fp4_quantize from flashinfer import fp4_quantize
else: elif is_cuda_alike():
from sgl_kernel import scaled_fp4_quant as fp4_quantize from sgl_kernel import scaled_fp4_quant as fp4_quantize
else:
fp4_quantize = None
def align_fp8_moe_weights_for_flashinfer_trtllm( def align_fp8_moe_weights_for_flashinfer_trtllm(