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

[CI] Improve github summary & enable fa3 for more models (#5796)

Browse Source
This commit is contained in:
Lianmin Zheng
2025-04-27 15:29:46 -07:00
committed by GitHub
parent ded04b2e0a
commit daed453e84
6 changed files with 26 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
test/srt/test_bench_serving.py
View File

@@ -98,7 +98,7 @@ class TestBenchServing(CustomTestCase):
f"### test_offline_throughput_with_triton_attention_backend\n"
f'Output throughput: {res["output_throughput"]:.2f} token/s\n'
)
self.assertGreater(res["output_throughput"], 3600)
self.assertGreater(res["output_throughput"], 3700)
def test_offline_throughput_default_fp8(self):
res = run_bench_serving(
@@ -113,7 +113,7 @@ class TestBenchServing(CustomTestCase):
f"### test_offline_throughput_default_fp8\n"
f'Output throughput: {res["output_throughput"]:.2f} token/s\n'
)
self.assertGreater(res["output_throughput"], 4200)
self.assertGreater(res["output_throughput"], 4300)
def test_online_latency_default(self):
res = run_bench_serving(
@@ -126,7 +126,7 @@ class TestBenchServing(CustomTestCase):
if is_in_ci():
write_github_step_summary(
f"### test_online_latency_default\n"
f'median_e2e_latency_ms : {res["median_e2e_latency_ms"]:.2f} ms\n'
f'median_e2e_latency_ms: {res["median_e2e_latency_ms"]:.2f} ms\n'
)
self.assertLess(res["median_e2e_latency_ms"], 11000)
self.assertLess(res["median_ttft_ms"], 86)
@@ -161,8 +161,8 @@ class TestBenchServing(CustomTestCase):
if is_in_ci():
write_github_step_summary(
f"### test_online_latency_eagle\n"
f'median_e2e_latency_ms : {res["median_e2e_latency_ms"]:.2f} ms\n'
f'accept_length : {res["accept_length"]:.2f} \n'
f'median_e2e_latency_ms: {res["median_e2e_latency_ms"]:.2f} ms\n'
f'accept_length: {res["accept_length"]:.2f} \n'
)
self.assertLess(res["median_e2e_latency_ms"], 900)
self.assertGreater(res["accept_length"], 3.0)