[AMD] CI - migrate perf test and fix stage-b-test-1-gpu-amd (#17340)

Co-authored-by: Bingxu Chen <bingxche@amd.com>
Co-authored-by: bingxche <Bingxu.Chen@amd.com>
Co-authored-by: michaelzhang-ai <michaelzhang.ai@users.noreply.github.com>
This commit is contained in:
YC Tseng
2026-01-22 18:45:05 -08:00
committed by GitHub
co-authored by Bingxu Chen bingxche michaelzhang-ai
parent 0fec8820d1
commit 04a10c9bc2
14 changed files with 68 additions and 174 deletions
@@ -4,7 +4,8 @@ Performance tests for single GPU that need H200 (80GB) - FP8 and EAGLE tests.
import unittest
from sglang.test.ci.ci_register import register_cuda_ci
from sglang.srt.utils import is_hip
from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
from sglang.test.test_utils import (
DEFAULT_DRAFT_MODEL_EAGLE,
DEFAULT_MODEL_NAME_FOR_TEST_FP8,
@@ -17,6 +18,7 @@ from sglang.test.test_utils import (
)
register_cuda_ci(est_time=300, suite="stage-b-test-large-1-gpu-performance")
register_amd_ci(est_time=300, suite="stage-b-test-large-1-gpu-performance-amd")
class TestBenchServing1GPULarge(CustomTestCase):
@@ -38,6 +40,7 @@ class TestBenchServing1GPULarge(CustomTestCase):
else:
self.assertGreater(res["output_throughput"], 4300)
@unittest.skipIf(is_hip(), "Skip Eagle test for ROCm")
def test_online_latency_eagle(self):
res = run_bench_serving(
model=DEFAULT_TARGET_MODEL_EAGLE,