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

Improve profiler and integrate profiler in bench_one_batch_server (#6787)

Browse Source
This commit is contained in:
Lianmin Zheng
2025-05-31 15:53:55 -07:00
committed by GitHub
parent b520d02888
commit 2d72fc47cf
25 changed files with 481 additions and 223 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
test/srt/test_bench_one_batch.py
View File

@@ -6,6 +6,7 @@ from sglang.test.test_utils import (
DEFAULT_MOE_MODEL_NAME_FOR_TEST,
DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
CustomTestCase,
is_in_amd_ci,
is_in_ci,
run_bench_offline_throughput,
run_bench_one_batch,
@@ -46,7 +47,7 @@ class TestBenchOneBatch(CustomTestCase):
f"### test_moe_tp2_bs1 (Mixtral-8x7B)\n"
f"output_throughput: {output_throughput:.2f} token/s\n"
)
if os.getenv("SGLANG_AMD_CI") == "1":
if is_in_amd_ci():
self.assertGreater(output_throughput, 85)
else:
self.assertGreater(output_throughput, 125)
@@ -62,7 +63,7 @@ class TestBenchOneBatch(CustomTestCase):
f"### test_torch_compile_tp2_bs1 (Mixtral-8x7B)\n"
f"output_throughput: {output_throughput:.2f} token/s\n"
)
if os.getenv("SGLANG_AMD_CI") == "1":
if is_in_amd_ci():
self.assertGreater(output_throughput, 200)
else:
self.assertGreater(output_throughput, 220)