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

[CI] Fix stale per-token group quant callers (#32047)

Browse Source
This commit is contained in:
Mohammad Miadh Angkad
2026-07-22 19:45:29 +08:00
committed by GitHub
parent 977ea336cd
commit 4f88206393
2 changed files with 12 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files
sgl-kernel/tests/test_per_token_group_quant_8bit.py
+2 -5
View File
@@ -1,8 +1,5 @@
import itertools
import os
import sys
import time
from pathlib import Path
import pytest
import torch
@@ -17,7 +14,7 @@ from sglang.kernels.ops.quantization.fp8_kernel import (
from sglang.kernels.ops.quantization.fp8_kernel import (
sglang_per_token_group_quant_8bit,
)
from sglang.srt.utils import get_bool_env_var, is_hip
from sglang.srt.utils import is_hip
_is_hip = is_hip()
fp8_type_ = torch.float8_e4m3fnuz if _is_hip else torch.float8_e4m3fn
@@ -156,7 +153,7 @@ def test_per_token_group_quant_with_column_major(
*triton_per_token_group_quant_8bit(**execute_kwargs)
)
x_q_sglang, x_s_sglang = _postprocess(
*sglang_per_token_group_quant_8bit(**execute_kwargs, enable_v2=True)
*sglang_per_token_group_quant_8bit(**execute_kwargs)
)
try: