[bugfix]GLM-4V model (#17122)

This commit is contained in:
KnightLTC
2026-04-01 10:37:40 +08:00
committed by GitHub
parent f9debd6514
commit 2488233ad5
3 changed files with 38 additions and 3 deletions
@@ -0,0 +1,33 @@
import unittest
from sglang.test.ascend.vlm_utils import TestVLMModels
from sglang.test.ci.ci_register import register_npu_ci
register_npu_ci(est_time=400, suite="nightly-8-npu-a3", nightly=True)
class TestGLM4Models(TestVLMModels):
model = "/root/.cache/modelscope/hub/models/ZhipuAI/GLM-4.5V"
mmmu_accuracy = 0.2
other_args = [
"--trust-remote-code",
"--cuda-graph-max-bs",
"32",
"--enable-multimodal",
"--mem-fraction-static",
0.7,
"--log-level",
"info",
"--attention-backend",
"ascend",
"--disable-cuda-graph",
"--tp-size",
8,
]
def test_vlm_mmmu_benchmark(self):
self._run_vlm_mmmu_test()
if __name__ == "__main__":
unittest.main()