biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[VLM] Add Qwen3-VL generation test (#25185)

Browse Source 
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: Roger Wang <hey@rogerw.io>
Co-authored-by: Roger Wang <hey@rogerw.io>
This commit is contained in:
Isotr0py
2025-10-29 20:19:37 +08:00
committed by GitHub
parent 3481e40743
commit ad3ec89532
7 changed files with 108 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
tests/models/multimodal/generation/test_common.py
View File

@@ -159,6 +159,28 @@ VLM_TEST_SETTINGS = {
image_size_factors=[(), (0.25,), (0.25, 0.25, 0.25), (0.25, 0.2, 0.15)],
marks=[pytest.mark.core_model, pytest.mark.cpu_model],
),
"qwen3_vl": VLMTestInfo(
models=["Qwen/Qwen3-VL-4B-Instruct"],
test_type=(
VLMTestType.IMAGE,
VLMTestType.MULTI_IMAGE,
VLMTestType.VIDEO,
),
needs_video_metadata=True,
prompt_formatter=lambda img_prompt: f"<|im_start|>User\n{img_prompt}<|im_end|>\n<|im_start|>assistant\n", # noqa: E501
img_idx_to_prompt=lambda idx: "<|vision_start|><|image_pad|><|vision_end|>", # noqa: E501
video_idx_to_prompt=lambda idx: "<|vision_start|><|video_pad|><|vision_end|>", # noqa: E501
max_model_len=4096,
max_num_seqs=2,
num_logprobs=20,
auto_cls=AutoModelForImageTextToText,
vllm_output_post_proc=model_utils.qwen2_vllm_to_hf_output,
patch_hf_runner=model_utils.qwen3_vl_patch_hf_runner,
image_size_factors=[(), (0.25,), (0.25, 0.25, 0.25), (0.25, 0.2, 0.15)],
marks=[
pytest.mark.core_model,
],
),
"ultravox": VLMTestInfo(
models=["fixie-ai/ultravox-v0_5-llama-3_2-1b"],
test_type=VLMTestType.AUDIO,