Remove dp balance metadata and minimul token balance. (#11170)

This commit is contained in:
Liangsheng Yin
2025-10-03 01:48:15 +08:00
committed by GitHub
parent bfcd9b2433
commit 7ff740a6ce
11 changed files with 77 additions and 316 deletions

View File

@@ -124,47 +124,5 @@ class TestDPAttentionDP2TP2DeepseekV3MTP(CustomTestCase):
self.assertGreater(avg_spec_accept_length, 2.5)
class TestDPAttentionMinimumTokenLoadBalance(CustomTestCase):
@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,
other_args=[
"--trust-remote-code",
"--tp",
"2",
"--enable-dp-attention",
"--dp",
"2",
"--enable-torch-compile",
"--torch-compile-max-bs",
"2",
"--load-balance-method",
"minimum_tokens",
],
)
@classmethod
def tearDownClass(cls):
kill_process_tree(cls.process.pid)
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)
print(f"{metrics=}")
self.assertGreater(metrics["score"], 0.8)
if __name__ == "__main__":
unittest.main()