minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[misc] Use --cuda-graph-max-bs-decode in tests, examples, and docs (#29591)

Browse Source
This commit is contained in:
Liangsheng Yin
2026-06-28 18:38:28 -07:00
committed by GitHub
parent 3217410cf6
commit 909123ddb8
140 changed files with 304 additions and 290 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/registered/ep/test_deepep_large.py
+1 -1
View File
@@ -176,7 +176,7 @@ class TestDeepseekV32TBO(CustomTestCase):
"--enable-two-batch-overlap",
"--moe-a2a-backend",
"deepep",
"--cuda-graph-max-bs",
"--cuda-graph-max-bs-decode",
"256",
"--model-loader-extra-config",
'{"enable_multithread_load": true, "num_threads": 64}',
test/registered/ep/test_deepep_small.py
+10 -10
View File
@@ -37,7 +37,7 @@ class TestPureDP(CustomTestCase):
"4",
"--moe-a2a-backend",
"deepep",
"--cuda-graph-max-bs",
"--cuda-graph-max-bs-decode",
"128",
"--max-running-requests",
"512",
@@ -84,7 +84,7 @@ class TestHybridDPTP(CustomTestCase):
"2",
"--moe-a2a-backend",
"deepep",
"--cuda-graph-max-bs",
"--cuda-graph-max-bs-decode",
"128",
"--max-running-requests",
"256",
@@ -126,7 +126,7 @@ class TestTP(CustomTestCase):
"4",
"--moe-a2a-backend",
"deepep",
"--cuda-graph-max-bs",
"--cuda-graph-max-bs-decode",
"128",
"--max-running-requests",
"128",
@@ -175,7 +175,7 @@ class TestNoGatherdBuffer(CustomTestCase):
"--enable-dp-lm-head",
"--moe-a2a-backend",
"deepep",
"--cuda-graph-max-bs",
"--cuda-graph-max-bs-decode",
"32",
"--max-running-requests",
"512",
@@ -223,7 +223,7 @@ class TestTBO(CustomTestCase):
"--moe-a2a-backend",
"deepep",
"--enable-two-batch-overlap",
"--cuda-graph-max-bs",
"--cuda-graph-max-bs-decode",
"128",
"--max-running-requests",
"512",
@@ -270,7 +270,7 @@ class TestTBOWithTPAttn(CustomTestCase):
"--moe-a2a-backend",
"deepep",
"--enable-two-batch-overlap",
"--cuda-graph-max-bs",
"--cuda-graph-max-bs-decode",
"128",
"--max-running-requests",
"512",
@@ -323,7 +323,7 @@ class TestTBOWithTPAttnAndDenseDP(CustomTestCase):
"--moe-a2a-backend",
"deepep",
"--enable-two-batch-overlap",
"--cuda-graph-max-bs",
"--cuda-graph-max-bs-decode",
"128",
"--max-running-requests",
"512",
@@ -386,7 +386,7 @@ class TestMTP(CustomTestCase):
"3",
"--speculative-num-draft-tokens",
"3",
"--cuda-graph-max-bs",
"--cuda-graph-max-bs-decode",
"32",
"--max-running-requests",
"64",
@@ -456,7 +456,7 @@ class TestMTPWithTBO(CustomTestCase):
DEFAULT_MODEL_NAME_FOR_TEST_MLA_NEXTN,
"--chunked-prefill-size",
"256",
"--cuda-graph-max-bs",
"--cuda-graph-max-bs-decode",
"32",
"--max-running-requests",
"128",
@@ -530,7 +530,7 @@ class TestMTPWithTPAttnAndTBO(CustomTestCase):
DEFAULT_MODEL_NAME_FOR_TEST_MLA_NEXTN,
"--chunked-prefill-size",
"256",
"--cuda-graph-max-bs",
"--cuda-graph-max-bs-decode",
"32",
"--max-running-requests",
"128",
test/registered/ep/test_mooncake_ep_small.py
+1 -1
View File
@@ -53,7 +53,7 @@ class TestTP(CustomTestCase):
"72",
"--chunked-prefill-size",
"512",
"--cuda-graph-max-bs",
"--cuda-graph-max-bs-decode",
"128",
"--max-running-requests",
"512",