Re-introduce the unit test of test_mooncake_ep_small (#16019)
This commit is contained in:
@@ -0,0 +1,146 @@
|
||||
import unittest
|
||||
from types import SimpleNamespace
|
||||
|
||||
from sglang.srt.utils import kill_process_tree
|
||||
from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
|
||||
from sglang.test.server_fixtures.disaggregation_fixture import get_rdma_devices_args
|
||||
from sglang.test.test_utils import (
|
||||
DEFAULT_MODEL_NAME_FOR_TEST_MLA,
|
||||
DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
|
||||
DEFAULT_URL_FOR_TEST,
|
||||
CustomTestCase,
|
||||
popen_launch_server,
|
||||
)
|
||||
|
||||
ib_devices = get_rdma_devices_args()
|
||||
|
||||
|
||||
class TestTP(CustomTestCase):
|
||||
extra_args = []
|
||||
|
||||
@classmethod
|
||||
def setUpClass(cls):
|
||||
cls.model = DEFAULT_MODEL_NAME_FOR_TEST_MLA
|
||||
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=[
|
||||
"--trust-remote-code",
|
||||
"--tp",
|
||||
"4",
|
||||
"--elastic-ep-backend",
|
||||
"mooncake",
|
||||
"--mooncake-ib-device",
|
||||
ib_devices,
|
||||
"--moe-a2a-backend",
|
||||
"mooncake",
|
||||
"--deepep-mode",
|
||||
"low_latency",
|
||||
"--chunked-prefill-size",
|
||||
"512",
|
||||
"--cuda-graph-max-bs",
|
||||
"128",
|
||||
"--max-running-requests",
|
||||
"512",
|
||||
"--mem-fraction-static",
|
||||
"0.5",
|
||||
*cls.extra_args,
|
||||
],
|
||||
)
|
||||
|
||||
@classmethod
|
||||
def tearDownClass(cls):
|
||||
kill_process_tree(cls.process.pid)
|
||||
|
||||
def test_gsm8k(self):
|
||||
args = SimpleNamespace(
|
||||
num_shots=5,
|
||||
data_path=None,
|
||||
num_questions=200,
|
||||
max_new_tokens=512,
|
||||
parallel=128,
|
||||
host="http://127.0.0.1",
|
||||
port=int(self.base_url.split(":")[-1]),
|
||||
)
|
||||
metrics = run_eval_few_shot_gsm8k(args)
|
||||
print(metrics)
|
||||
|
||||
self.assertGreater(metrics["accuracy"], 0.60)
|
||||
|
||||
|
||||
@unittest.skip("covered in TestMooncakeWithEPLB")
|
||||
class TestPureDP(TestTP):
|
||||
extra_args = [
|
||||
"--tp",
|
||||
"4",
|
||||
"--enable-dp-attention",
|
||||
"--dp",
|
||||
"4",
|
||||
]
|
||||
|
||||
|
||||
class TestHybridDPTP(TestTP):
|
||||
extra_args = [
|
||||
"--tp",
|
||||
"4",
|
||||
"--enable-dp-attention",
|
||||
"--dp",
|
||||
"2",
|
||||
]
|
||||
|
||||
|
||||
@unittest.skip("covered in TestMooncakeWithEPLB")
|
||||
class TestNoGatherdBuffer(TestTP):
|
||||
extra_args = [
|
||||
"--tp",
|
||||
"4",
|
||||
"--enable-dp-attention",
|
||||
"--dp",
|
||||
"4",
|
||||
"--moe-dense-tp-size",
|
||||
"1",
|
||||
]
|
||||
|
||||
|
||||
@unittest.skip("covered in TestMooncakeWithEPLB")
|
||||
class TestTBO(TestTP):
|
||||
extra_args = [
|
||||
"--tp",
|
||||
"4",
|
||||
"--enable-dp-attention",
|
||||
"--dp",
|
||||
"4",
|
||||
"--moe-dense-tp-size",
|
||||
"1",
|
||||
"--enable-two-batch-overlap",
|
||||
]
|
||||
|
||||
|
||||
class TestMooncakeWithEPLB(TestTP):
|
||||
extra_args = [
|
||||
"--tp",
|
||||
"4",
|
||||
"--enable-dp-attention",
|
||||
"--dp",
|
||||
"4",
|
||||
"--moe-dense-tp-size",
|
||||
"1",
|
||||
"--enable-two-batch-overlap",
|
||||
"--enable-eplb",
|
||||
"--ep-num-redundant-experts",
|
||||
"4",
|
||||
"--eplb-rebalance-num-iterations",
|
||||
"50",
|
||||
"--expert-distribution-recorder-buffer-size",
|
||||
"50",
|
||||
"--expert-distribution-recorder-mode",
|
||||
"stat",
|
||||
"--ep-dispatch-algorithm",
|
||||
"static",
|
||||
]
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
unittest.main()
|
||||
@@ -45,8 +45,7 @@ suites = {
|
||||
],
|
||||
"per-commit-4-gpu-deepep": [
|
||||
TestFile("ep/test_deepep_small.py", 531),
|
||||
# TODO: Add it back after mooncake supports torch 2.9
|
||||
# TestFile("ep/test_mooncake_ep_small.py", 450),
|
||||
TestFile("ep/test_mooncake_ep_small.py", 660),
|
||||
],
|
||||
"per-commit-8-gpu-h200-deepep": [
|
||||
TestFile("ep/test_deepep_large.py", 563),
|
||||
|
||||
Reference in New Issue
Block a user