test_hicache_mla.py 1.67 KB
Newer Older
1
2
3
import unittest
from types import SimpleNamespace

4
from sglang.srt.utils import is_hip, kill_process_tree
5
6
7
8
9
from sglang.test.run_eval import run_eval
from sglang.test.test_utils import (
    DEFAULT_MLA_MODEL_NAME_FOR_TEST,
    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
    DEFAULT_URL_FOR_TEST,
10
    CustomTestCase,
11
12
13
    popen_launch_server,
)

14
15
16
17
18
19
_is_hip = is_hip()
if _is_hip:
    hicache_args = ["--hicache-size", 200]
else:
    hicache_args = ["--hicache-ratio", 2]

20

21
class TestHierarchicalMLA(CustomTestCase):
22
23
24
25
26
27
28
29
    @classmethod
    def setUpClass(cls):
        cls.model = DEFAULT_MLA_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,
30
31
32
            other_args=[
                "--trust-remote-code",
                "--enable-hierarchical-cache",
33
34
            ]
            + hicache_args,
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
        )

    @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.assertGreater(metrics["score"], 0.5)

    def test_mgsm_en(self):
        args = SimpleNamespace(
            base_url=self.base_url,
            model=self.model,
            eval_name="mgsm_en",
            num_examples=None,
            num_threads=1024,
        )

        metrics = run_eval(args)
        self.assertGreater(metrics["score"], 0.8)


if __name__ == "__main__":
    unittest.main()