[Refector] Remove mammothmodal2-preview example (#5335)

Signed-off-by: princepride <wangzhipeng628@gmail.com>
Signed-off-by: 汪志鹏 <wangzhipeng628@gmail.com>
This commit is contained in:
汪志鹏 authored and GitHub committed 2026-07-24 07:29:55 +08:00
1 parent 0fe5d73a63
commit a61d5c0fd8
4 files changed
+46 -178

No files matched your search

+46 -1
View File
@@ -6,7 +6,8 @@
- Vendor: ByteDance Research
- Model: `bytedance-research/MammothModa2-Preview`
- Task: Text-to-image generation (AR → DiT two-stage pipeline)
- Task: Text-to-image generation (AR → DiT two-stage pipeline), plus
image-to-text understanding (AR stage)
- Mode: Offline inference
- Maintainer: Community
@@ -108,3 +109,47 @@ exists and is a valid image:
ls -lh mammoth_t2i.png
python -c "from PIL import Image; print(Image.open('mammoth_t2i.png').size)"
```
### Image-to-Text (I2T, simplest)
For the simplest offline image-understanding run, use a short inline Python
snippet from the repository root:
```python
from PIL import Image
from vllm import SamplingParams
from vllm.multimodal.image import convert_image_mode
from vllm_omni import Omni
def main():
prompt = (
"<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n"
"<|im_start|>user\n<|vision_start|><|image_pad|><|vision_end|>"
"Summarize this image.<|im_end|>\n"
"<|im_start|>assistant\n"
)
omni = Omni(
model="./MammothModa2-Preview",
deploy_config="vllm_omni/deploy/mammoth_moda2_ar.yaml",
)
try:
outputs = list(
omni.generate(
[{
"prompt": prompt,
"multi_modal_data": {"image": convert_image_mode(Image.open("./image.png"), "RGB")},
"additional_information": {"omni_task": ["chat"]},
}],
[SamplingParams(temperature=0.2, top_p=0.9, top_k=-1, max_tokens=512, seed=42)],
)
)
finally:
omni.close()
ro = getattr(outputs[-1], "request_output", outputs[-1])
print(ro.outputs[0].text.strip())
if __name__ == "__main__":
main()
```