Files
sglang/test/registered/models_e2e/test_dsa_glm52_dp_mtp.py
T

29 lines
675 B
Python

import unittest
from sglang.test.ci.ci_register import register_cuda_ci
from sglang.test.kits.eval_accuracy_kit import GSM8KMixin
from sglang.test.kits.spec_decoding_kit import SpecDecodingMixin
from sglang.test.server_fixtures.dsa_mtp_fixture import (
DsaMtpEvalConfigDefaults,
DsaMtpServerBase,
)
register_cuda_ci(
est_time=400,
stage="base-c",
runner_config="8-gpu-h200",
)
class TestGLM52DPMTP(
DsaMtpServerBase, DsaMtpEvalConfigDefaults, GSM8KMixin, SpecDecodingMixin
):
model = "zai-org/GLM-5.2-FP8"
mem_fraction_static = 0.88
enable_dp_attention = True
bs_1_speed_thres = 70
if __name__ == "__main__":
unittest.main()