Large page size aligned hierarchical caching (#4581)

This commit is contained in:
Zhiqiang Xie
2025-04-01 22:38:15 -07:00
committed by GitHub
parent 9eb49e878b
commit e119f04215
8 changed files with 242 additions and 71 deletions

View File

@@ -12,7 +12,7 @@ from sglang.test.test_utils import (
)
class TestPageSize(CustomTestCase):
class TestHiCache(CustomTestCase):
@classmethod
def setUpClass(cls):
cls.model = DEFAULT_MODEL_NAME_FOR_TEST
@@ -21,7 +21,9 @@ class TestPageSize(CustomTestCase):
cls.model,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
other_args=["--enable-hierarchical-cache"],
other_args=[
"--enable-hierarchical-cache",
],
)
@classmethod

View File

@@ -21,7 +21,10 @@ class TestHierarchicalMLA(CustomTestCase):
cls.model,
cls.base_url,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
other_args=["--trust-remote-code", "--enable-hierarchical-cache"],
other_args=[
"--trust-remote-code",
"--enable-hierarchical-cache",
],
)
@classmethod

View File

@@ -0,0 +1,49 @@
import unittest
from types import SimpleNamespace
from sglang.srt.utils import kill_process_tree
from sglang.test.run_eval import run_eval
from sglang.test.test_utils import (
DEFAULT_MODEL_NAME_FOR_TEST,
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
DEFAULT_URL_FOR_TEST,
CustomTestCase,
popen_launch_server,
)
class TestHiCachePage(CustomTestCase):
@classmethod
def setUpClass(cls):
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=[
"--enable-hierarchical-cache",
"--page-size",
"32",
],
)
@classmethod
def tearDownClass(cls):
kill_process_tree(cls.process.pid)
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)
if __name__ == "__main__":
unittest.main()