---
title: Align-Dont-Blend-VLM
canonical_url: "https://www.modelscope.cn/models/GaffeyGao/Align-Dont-Blend-VLM"
md_url: "https://www.modelscope.cn/models/GaffeyGao/Align-Dont-Blend-VLM.md"
repository: GaffeyGao/Align-Dont-Blend-VLM
last_updated: 2026-11-09
pipeline_tag: image-text-to-text
tasks:
  - image-text-to-text
base_model:
  - Qwen/Qwen2.5-VL-7B-Instruct
base_model_relation: adapter
parameters: 40.4M
tensor_type:
  - BF16
library_name:
  - lora
  - safetensors
downloads: 2
stars: 0
tags:
  - lora
  - ai-generated-image-detection
---

# Align-Dont-Blend-VLM

> Align-Dont-Blend-VLM - GaffeyGao 在 ModelScope 开源的模型。VLM LoRA adapter for Align, Don't Blend

GaffeyGao/Align-Dont-Blend-VLM 是 ModelScope 魔搭社区上的 40.4M 参数image-text-to-text模型，基于 Qwen/Qwen2.5-VL-7B-Instruct 构建。

- **Repository**: GaffeyGao/Align-Dont-Blend-VLM
- **Tasks**: image-text-to-text
- **Parameters**: 40.4M
- **Base model**: Qwen/Qwen2.5-VL-7B-Instruct
- **Tags**: lora, ai-generated-image-detection
- **Downloads**: 2
- **Stars**: 0
- **Last updated**: 2026-11-09

Source: https://www.modelscope.cn/models/GaffeyGao/Align-Dont-Blend-VLM

---

# Align, Don't Blend — VLM Adapter

VLM LoRA adapter for **Align, Don't Blend**. Load it with **Qwen2.5-VL-7B-Instruct**; base-model weights are downloaded separately.

Download the files into `weights/vlm_adapter/` and follow the [inference instructions](https://github.com/GzZning/Align-Dont-Blend#inference). The pixel-artifact expert is optional and is not included.

[Hugging Face](https://huggingface.co/Gaffeyzz/Align-Dont-Blend-VLM) | [ModelScope](https://modelscope.cn/models/GaffeyGao/Align-Dont-Blend-VLM)