chenchenghao/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

dp-attention: add follow_bootstrap_room + auto load-balance; drop decode_round_robin (#16110)

Browse Source
This commit is contained in:
Mufeez Amjad
2025-12-30 09:33:06 -05:00
committed by GitHub
parent 4de59d83a1
commit cbff7ad985
4 changed files with 80 additions and 35 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
test/srt/test_server_args.py
View File

@@ -25,6 +25,20 @@ class TestPrepareServerArgs(CustomTestCase):
)
class TestLoadBalanceMethod(unittest.TestCase):
def test_non_pd_defaults_to_round_robin(self):
server_args = ServerArgs(model_path="dummy", disaggregation_mode="null")
self.assertEqual(server_args.load_balance_method, "round_robin")
def test_pd_prefill_defaults_to_follow_bootstrap_room(self):
server_args = ServerArgs(model_path="dummy", disaggregation_mode="prefill")
self.assertEqual(server_args.load_balance_method, "follow_bootstrap_room")
def test_pd_decode_defaults_to_round_robin(self):
server_args = ServerArgs(model_path="dummy", disaggregation_mode="decode")
self.assertEqual(server_args.load_balance_method, "round_robin")
class TestPortArgs(unittest.TestCase):
@patch("sglang.srt.server_args.is_port_available")
@patch("sglang.srt.server_args.tempfile.NamedTemporaryFile")