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

[Fix] memory leak by overlap + retract (#11981)

Browse Source 
Co-authored-by: Liangsheng Yin <lsyincs@gmail.com>
This commit is contained in:
cctry
2025-10-23 07:59:23 -07:00
committed by GitHub
parent 6c18addb6f
commit b0b4f71679
9 changed files with 132 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
test/srt/run_suite.py
View File

@@ -112,7 +112,7 @@ suites = {
TestFile("test_reasoning_parser.py", 5),
TestFile("test_regex_constrained.py", 64),
TestFile("test_request_queue_validation.py", 30),
TestFile("test_retract_decode.py", 54),
TestFile("test_retract_decode.py", 90),
TestFile("test_score_api.py", 310),
TestFile("test_server_args.py", 1),
TestFile("test_skip_tokenizer_init.py", 117),

47
test/srt/test_retract_decode.py
View File

@@ -1,7 +1,8 @@
import os
import time
import unittest
from types import SimpleNamespace
from sglang.srt.environ import envs
from sglang.srt.utils import kill_process_tree
from sglang.test.run_eval import run_eval
from sglang.test.test_utils import (
@@ -16,13 +17,12 @@ from sglang.test.test_utils import (
class TestRetractDecode(CustomTestCase):
@classmethod
def setUpClass(cls):
os.environ["SGLANG_TEST_RETRACT"] = "1"
cls.model = DEFAULT_MODEL_NAME_FOR_TEST
cls.base_url = DEFAULT_URL_FOR_TEST
cls.process = popen_launch_server(
cls.model, cls.base_url, timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH
)
with envs.SGLANG_TEST_RETRACT.override(True):
cls.process = popen_launch_server(
cls.model, cls.base_url, timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH
)
@classmethod
def tearDownClass(cls):
@@ -39,22 +39,43 @@ class TestRetractDecode(CustomTestCase):
metrics = run_eval(args)
self.assertGreaterEqual(metrics["score"], 0.65)
time.sleep(1) # wait for mem check
assert self.process.poll() is None, "Server crashed during test"
class TestRetractDecodeChunkCache(CustomTestCase):
@classmethod
def setUpClass(cls):
os.environ["SGLANG_TEST_RETRACT"] = "1"
cls.model = DEFAULT_MODEL_NAME_FOR_TEST
cls.base_url = DEFAULT_URL_FOR_TEST
cls.process = popen_launch_server(
cls.model,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
other_args=["--disable-radix-cache", "--chunked-prefill-size", 128],
with envs.SGLANG_TEST_RETRACT.override(True):
cls.process = popen_launch_server(
cls.model,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
other_args=["--disable-radix-cache", "--chunked-prefill-size", 128],
)
def test_mmlu(self):
args = SimpleNamespace(
base_url=self.base_url,
model=self.model,
eval_name="mmlu",
num_examples=64,
num_threads=32,
)
metrics = run_eval(args)
self.assertGreaterEqual(metrics["score"], 0.65)
time.sleep(1) # wait for mem check
assert self.process.poll() is None, "Server crashed during test"
@classmethod
def tearDownClass(cls):
kill_process_tree(cls.process.pid)
if __name__ == "__main__":
unittest.main()