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

[AMD] Fix int8 per-token quant Triton portability + register test for AMD nightly CI (#29694)

Browse Source
This commit is contained in:
Michael
2026-07-01 12:56:33 -07:00
committed by GitHub
parent c312cdd3a7
commit 8361561c24
2 changed files with 13 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/registered/quant/test_int8_kernel.py
+2 -1
View File
@@ -8,10 +8,11 @@ from sglang.srt.layers.moe.moe_runner.triton_utils.fused_moe import fused_moe
from sglang.srt.layers.moe.topk import TopKConfig, select_experts
from sglang.srt.layers.quantization.int8_kernel import per_token_quant_int8
from sglang.srt.server_args import ServerArgs, set_global_server_args_for_scheduler
from sglang.test.ci.ci_register import register_cuda_ci
from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
from sglang.test.test_utils import CustomTestCase
register_cuda_ci(est_time=15, stage="base-b", runner_config="1-gpu-small")
register_amd_ci(est_time=15, suite="nightly-amd-kernel-1-gpu", nightly=True)
def native_w8a8_per_token_matmul(A, B, As, Bs, output_dtype=torch.float16):