65 lines
2.0 KiB
Python
65 lines
2.0 KiB
Python
import unittest
|
|
|
|
from sglang.test.accuracy_test_runner import AccuracyTestParams
|
|
from sglang.test.ci.ci_register import register_cuda_ci
|
|
from sglang.test.performance_test_runner import PerformanceTestParams
|
|
from sglang.test.run_combined_tests import run_combined_tests
|
|
from sglang.test.test_utils import ModelLaunchSettings
|
|
|
|
# Runs on both H200 and B200: registered once per runner_config below
|
|
register_cuda_ci(est_time=1860, stage="nightly", runner_config="8-gpu-h200")
|
|
register_cuda_ci(est_time=1860, stage="nightly", runner_config="8-gpu-b200")
|
|
|
|
MINIMAX_M25_MODEL_PATH = "MiniMaxAI/MiniMax-M2.5"
|
|
|
|
|
|
class TestMiniMaxM25(unittest.TestCase):
|
|
"""Unified test class for MiniMax-M2.5 performance and accuracy.
|
|
|
|
Single variant with TP=8 + EP=8 configuration.
|
|
Runs BOTH:
|
|
- Performance test (using NightlyBenchmarkRunner with extra_bench_args)
|
|
- Accuracy test (using run_eval with gsm8k)
|
|
"""
|
|
|
|
def test_minimax_m25(self):
|
|
"""Run performance and accuracy for MiniMax-M2.5."""
|
|
base_args = [
|
|
"--trust-remote-code",
|
|
"--ep=8",
|
|
"--mem-fraction-static=0.85",
|
|
"--reasoning-parser=minimax-append-think",
|
|
]
|
|
dp_attn_args = base_args + [
|
|
"--enable-dp-attention",
|
|
"--dp=8",
|
|
]
|
|
|
|
variants = [
|
|
ModelLaunchSettings(
|
|
MINIMAX_M25_MODEL_PATH,
|
|
tp_size=8,
|
|
extra_args=base_args,
|
|
variant="TP8+EP8",
|
|
),
|
|
ModelLaunchSettings(
|
|
MINIMAX_M25_MODEL_PATH,
|
|
tp_size=8,
|
|
extra_args=dp_attn_args,
|
|
variant="TP8+DP8+EP8+DPAttn",
|
|
),
|
|
]
|
|
|
|
run_combined_tests(
|
|
models=variants,
|
|
test_name="MiniMax-M2.5",
|
|
accuracy_params=AccuracyTestParams(dataset="gsm8k", baseline_accuracy=0.80),
|
|
performance_params=PerformanceTestParams(
|
|
result_dir="performance_results_minimax_m25",
|
|
),
|
|
)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
unittest.main()
|