EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Fix CI tests (#3132)

Browse Source
This commit is contained in:
Lianmin Zheng
2025-01-25 17:43:39 -08:00
committed by GitHub
parent 9286740eff
commit da6f8081f6
2 changed files with 8 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
.github/workflows/pr-test.yml vendored
View File

@@ -43,6 +43,8 @@ jobs:
- name: Run test
timeout-minutes: 10
env:
OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
run: |
cd test/lang
python3 run_suite.py --suite per-commit

12
test/srt/test_bench_serving.py
View File

@@ -49,7 +49,7 @@ class TestBenchServing(unittest.TestCase):
)
# There is a regression with torch 2.5
# This number was 950 for torch 2.4
self.assertGreater(res["output_throughput"], 850)
self.assertGreater(res["output_throughput"], 1000)
def test_offline_throughput_without_radix_cache(self):
res = run_bench_serving(
@@ -114,7 +114,7 @@ class TestBenchServing(unittest.TestCase):
f"### test_offline_throughput_default_fp8\n"
f'Output throughput: {res["output_throughput"]:.2f} token/s\n'
)
self.assertGreater(res["output_throughput"], 3850)
self.assertGreater(res["output_throughput"], 3900)
def test_online_latency_default(self):
res = run_bench_serving(
@@ -129,7 +129,7 @@ class TestBenchServing(unittest.TestCase):
f"### test_online_latency_default\n"
f'median_e2e_latency_ms : {res["median_e2e_latency_ms"]:.2f} ms\n'
)
self.assertLess(res["median_e2e_latency_ms"], 12000)
self.assertLess(res["median_e2e_latency_ms"], 11000)
self.assertLess(res["median_ttft_ms"], 86)
self.assertLess(res["median_itl_ms"], 10)
@@ -161,7 +161,7 @@ class TestBenchServing(unittest.TestCase):
f"### test_online_latency_eagle\n"
f'median_e2e_latency_ms : {res["median_e2e_latency_ms"]:.2f} ms\n'
)
self.assertLess(res["median_e2e_latency_ms"], 10000)
self.assertLess(res["median_e2e_latency_ms"], 450)
def test_moe_offline_throughput_default(self):
res = run_bench_serving(
@@ -176,7 +176,7 @@ class TestBenchServing(unittest.TestCase):
f"### test_moe_offline_throughput_default\n"
f'Output throughput: {res["output_throughput"]:.2f} token/s\n'
)
self.assertGreater(res["output_throughput"], 2150)
self.assertGreater(res["output_throughput"], 2200)
def test_moe_offline_throughput_without_radix_cache(self):
res = run_bench_serving(
@@ -191,7 +191,7 @@ class TestBenchServing(unittest.TestCase):
f"### test_moe_offline_throughput_without_radix_cache\n"
f'Output throughput: {res["output_throughput"]:.2f} token/s\n'
)
self.assertGreater(res["output_throughput"], 2150)
self.assertGreater(res["output_throughput"], 2200)
if __name__ == "__main__":