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

Fix GPU OOM (#6564)

Browse Source 
Co-authored-by: michael <michael.zhang@amd.com>
This commit is contained in:
kk
2025-05-25 07:38:39 +08:00
committed by GitHub
parent 24c035f2e3
commit 7a5e6ce1cb
8 changed files with 15 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
test/srt/test_bench_serving.py
View File

@@ -32,7 +32,7 @@ class TestBenchServing(CustomTestCase):
f'Output throughput: {res["output_throughput"]:.2f} token/s\n'
)
if os.getenv("SGLANG_AMD_CI") == "1":
self.assertGreater(res["output_throughput"], 3500)
self.assertGreater(res["output_throughput"], 3150)
else:
self.assertGreater(res["output_throughput"], 3800)
@@ -70,7 +70,7 @@ class TestBenchServing(CustomTestCase):
f'Output throughput: {res["output_throughput"]:.2f} token/s\n'
)
if os.getenv("SGLANG_AMD_CI") == "1":
self.assertGreater(res["output_throughput"], 3500)
self.assertGreater(res["output_throughput"], 3050)
else:
self.assertGreater(res["output_throughput"], 3800)
@@ -126,7 +126,7 @@ class TestBenchServing(CustomTestCase):
f'Output throughput: {res["output_throughput"]:.2f} token/s\n'
)
if os.getenv("SGLANG_AMD_CI") == "1":
self.assertGreater(res["output_throughput"], 4000)
self.assertGreater(res["output_throughput"], 3500)
else:
self.assertGreater(res["output_throughput"], 4300)