minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

Disable async assert in Nemotron nightly tests (#27838)

Browse Source
This commit is contained in:
Baizhou Zhang
2026-06-10 16:28:14 -07:00
committed by GitHub
parent 16124fc9b2
commit 125ef88892
2 changed files with 52 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/registered/4-gpu-models/test_nvidia_nemotron_3_super_nvfp4.py
+15 -12
View File
@@ -1,6 +1,7 @@
import unittest import unittest
from types import SimpleNamespace from types import SimpleNamespace
from sglang.srt.environ import envs
from sglang.srt.utils import kill_process_tree from sglang.srt.utils import kill_process_tree
from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.ci.ci_register import register_cuda_ci
from sglang.test.run_eval import run_eval from sglang.test.run_eval import run_eval
@@ -69,12 +70,13 @@ class TestNvidiaNemotron3SuperNVFP4(CustomTestCase):
def setUpClass(cls): def setUpClass(cls):
cls.model = NEMOTRON_3_SUPER_NVFP4_MODEL cls.model = NEMOTRON_3_SUPER_NVFP4_MODEL
cls.base_url = DEFAULT_URL_FOR_TEST cls.base_url = DEFAULT_URL_FOR_TEST
cls.process = popen_launch_server( with envs.SGLANG_ENABLE_ASYNC_ASSERT.override(0):
cls.model, cls.process = popen_launch_server(
cls.base_url, cls.model,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH, cls.base_url,
other_args=NEMOTRON_3_SUPER_NVFP4_ARGS, timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
) other_args=NEMOTRON_3_SUPER_NVFP4_ARGS,
)
@classmethod @classmethod
def tearDownClass(cls): def tearDownClass(cls):
@@ -89,12 +91,13 @@ class TestNvidiaNemotron3SuperNVFP4MTP(CustomTestCase):
def setUpClass(cls): def setUpClass(cls):
cls.model = NEMOTRON_3_SUPER_NVFP4_MODEL cls.model = NEMOTRON_3_SUPER_NVFP4_MODEL
cls.base_url = DEFAULT_URL_FOR_TEST cls.base_url = DEFAULT_URL_FOR_TEST
cls.process = popen_launch_server( with envs.SGLANG_ENABLE_ASYNC_ASSERT.override(0):
cls.model, cls.process = popen_launch_server(
cls.base_url, cls.model,
timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH, cls.base_url,
other_args=NEMOTRON_3_SUPER_NVFP4_ARGS + MTP_ARGS, timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
) other_args=NEMOTRON_3_SUPER_NVFP4_ARGS + MTP_ARGS,
)
@classmethod @classmethod
def tearDownClass(cls): def tearDownClass(cls):
test/registered/8-gpu-models/test_nvidia_nemotron_3_super_nightly.py
+37 -34
View File
@@ -1,5 +1,6 @@
import unittest import unittest
from sglang.srt.environ import envs
from sglang.test.accuracy_test_runner import AccuracyTestParams from sglang.test.accuracy_test_runner import AccuracyTestParams
from sglang.test.ci.ci_register import register_cuda_ci from sglang.test.ci.ci_register import register_cuda_ci
from sglang.test.performance_test_runner import PerformanceTestParams from sglang.test.performance_test_runner import PerformanceTestParams
@@ -76,23 +77,24 @@ class TestNvidiaNemotron3SuperNightly(unittest.TestCase):
), ),
] ]
run_combined_tests( with envs.SGLANG_ENABLE_ASYNC_ASSERT.override(0):
models=variants, run_combined_tests(
test_name="Nemotron-3-Super-120B-BF16", models=variants,
accuracy_params=AccuracyTestParams( test_name="Nemotron-3-Super-120B-BF16",
dataset="gsm8k", accuracy_params=AccuracyTestParams(
baseline_accuracy=GSM8K_BASELINE, dataset="gsm8k",
num_examples=1314, baseline_accuracy=GSM8K_BASELINE,
num_threads=512, num_examples=1314,
max_tokens=16000, num_threads=512,
temperature=1.0, max_tokens=16000,
top_p=0.95, temperature=1.0,
repeat=1, top_p=0.95,
), repeat=1,
performance_params=PerformanceTestParams( ),
profile_dir="performance_profiles_nemotron_3_super_bf16", performance_params=PerformanceTestParams(
), profile_dir="performance_profiles_nemotron_3_super_bf16",
) ),
)
@unittest.skipIf(not is_blackwell_system(), "NVFP4 requires Blackwell") @unittest.skipIf(not is_blackwell_system(), "NVFP4 requires Blackwell")
def test_nemotron_3_super_nvfp4(self): def test_nemotron_3_super_nvfp4(self):
@@ -112,23 +114,24 @@ class TestNvidiaNemotron3SuperNightly(unittest.TestCase):
), ),
] ]
run_combined_tests( with envs.SGLANG_ENABLE_ASYNC_ASSERT.override(0):
models=variants, run_combined_tests(
test_name="Nemotron-3-Super-120B-NVFP4", models=variants,
accuracy_params=AccuracyTestParams( test_name="Nemotron-3-Super-120B-NVFP4",
dataset="gsm8k", accuracy_params=AccuracyTestParams(
baseline_accuracy=GSM8K_BASELINE, dataset="gsm8k",
num_examples=1314, baseline_accuracy=GSM8K_BASELINE,
num_threads=512, num_examples=1314,
max_tokens=16000, num_threads=512,
temperature=1.0, max_tokens=16000,
top_p=0.95, temperature=1.0,
repeat=1, top_p=0.95,
), repeat=1,
performance_params=PerformanceTestParams( ),
profile_dir="performance_profiles_nemotron_3_super_nvfp4", performance_params=PerformanceTestParams(
), profile_dir="performance_profiles_nemotron_3_super_nvfp4",
) ),
)
if __name__ == "__main__": if __name__ == "__main__":