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

[Model][VLM] Add Qwen2.5-Omni model support (thinker only) (#15130)

Browse Source 
Signed-off-by: fyabc <suyang.fy@alibaba-inc.com>
Signed-off-by: Roger Wang <ywang@roblox.com>
Co-authored-by: Roger Wang <136131678+ywang96@users.noreply.github.com>
Co-authored-by: Roger Wang <ywang@roblox.com>
Co-authored-by: Xiong Wang <wangxiongts@163.com>
This commit is contained in:
Yang Fan
2025-04-19 14:14:36 +08:00
committed by GitHub
parent 5c9121203c
commit 2c1bd848a6
23 changed files with 1855 additions and 85 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
tests/models/decoder_only/vision_language/test_models.py
View File

@@ -139,6 +139,23 @@ 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],
),
"qwen2_5_omni": VLMTestInfo(
models=["Qwen/Qwen2.5-Omni-7B"],
test_type=(
VLMTestType.IMAGE,
VLMTestType.MULTI_IMAGE,
VLMTestType.VIDEO
),
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_bos|><|IMAGE|><|vision_eos|>", # noqa: E501
video_idx_to_prompt=lambda idx: "<|vision_bos|><|VIDEO|><|vision_eos|>", # noqa: E501
max_model_len=4096,
max_num_seqs=2,
auto_cls=AutoModelForVision2Seq,
vllm_output_post_proc=model_utils.qwen2_vllm_to_hf_output,
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],
),
#### Extended model tests
"aria": VLMTestInfo(
models=["rhymes-ai/Aria"],