[MM][Perf][CG] Support ViT full CUDA graph for glm4_1v image and video inference (#40576)

Signed-off-by: grYe99 <guorongye99@gmail.com>
Co-authored-by: grYe99 <guorongye99@gmail.com>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>
This commit is contained in:
GuorongYe
2026-06-09 11:13:56 +08:00
committed by GitHub
co-authored by grYe99 Isotr0py
parent d3de61502f
commit 9f153aa781
4 changed files with 480 additions and 25 deletions
@@ -137,6 +137,28 @@ MODEL_CONFIGS: dict[str, VitCudagraphTestConfig] = {
),
},
),
"glm4_1v": VitCudagraphTestConfig(
model="zai-org/GLM-4.1V-9B-Thinking",
image_prompt=(
"[gMASK]<sop><|system|>\nYou are a helpful assistant.<|user|>\n"
"<|begin_of_image|><|image|><|end_of_image|>"
"What is in this image?<|assistant|>assistant\n"
),
video_prompt=(
"[gMASK]<sop><|system|>\nYou are a helpful assistant.<|user|>\n"
"<|begin_of_video|><|video|><|end_of_video|>"
"Describe this video in one sentence<|assistant|>assistant\n"
),
needs_video_metadata=True,
marks=[pytest.mark.core_model],
vllm_runner_kwargs={
"load_format": "dummy",
"hf_overrides": partial(
dummy_hf_overrides,
model_arch="Glm4vForConditionalGeneration",
),
},
),
}