base_model:

• google/siglip-so400m-patch14-384 license: apache-2.0 pipeline_tag: image-feature-extraction library_name: slimm

CoMP-MM-1B 模型卡片

这是一个支持原生图像分辨率输入的视觉基础模型（VFM），基于SigLIP持续预训练而成。

模型来源

• 代码仓库: https://github.com/SliMM-X/CoMP-MM
• 论文: https://arxiv.org/abs/2503.18931
• 项目主页: https://slimm-x.github.io/comp

快速开始使用模型

安装GitHub仓库后，可通过以下代码快速使用模型：

import torch
from slimm.model.processor import SliMMQwen2VLProcessor
from slimm.model.utils_vl import process_vision_info
from slimm.model.vision_encoder import CoMPSiglipVisionModel
from PIL import Image

model_path = "SliMM-X/CoMP-SigLIP-So400M"

model = CoMPSiglipVisionModel.from_pretrained(
    model_path, torch_dtype="auto", device_map="cuda", w_merger=False
).to(torch.bfloat16)


processor = SliMMQwen2VLProcessor.from_pretrained(model_path)

image_input = Image.open("https://slimm-x.github.io/comp/figs/teaser.png")
inputs = processor(
    images=image_input,
    return_tensors="pt",
)

inputs = inputs.to("cuda")
output_feat = model(inputs.pixel_values.to(torch.bfloat16), inputs.image_grid_thw)
print(output_feat)

引用

BibTeX格式:

@article{comp2025,
      title={CoMP: Continual Multimodal Pre-training for Vision Foundation Models}, 
      author={Chen, Yitong and Meng, Lingchen and Peng, Wujian and Wu, Zuxuan and Jiang, Yu-Gang},
      year={2025},
      journal={arXiv preprint arXiv:2503.18931}, 
}