--- title: SRPO-for-ComfyUI canonical_url: "https://www.modelscope.cn/models/wikeeyang/SRPO-for-ComfyUI" md_url: "https://www.modelscope.cn/models/wikeeyang/SRPO-for-ComfyUI.md" repository: wikeeyang/SRPO-for-ComfyUI last_updated: 2025-09-24 license: other pipeline_tag: text-to-image-synthesis tasks: - text-to-image-synthesis base_model: - tencent/SRPO base_model_relation: quantized parameters: 23.8B tensor_type: - BF16 - F8_E4M3 library_name: - safetensors - gguf - pytorch frameworks: - PyTorch language: - en downloads: 492 stars: 6 tags: - gguf --- # SRPO-for-ComfyUI > SRPO-for-ComfyUI - wikeeyang 在 ModelScope 开源的模型。SRPO-for-ComfyUI wikeeyang/SRPO-for-ComfyUI 是 ModelScope 魔搭社区上的 23.8B 参数text-to-image-synthesis模型，采用 other 许可，基于 tencent/SRPO 构建。 - **Repository**: wikeeyang/SRPO-for-ComfyUI - **License**: other - **Tasks**: text-to-image-synthesis - **Parameters**: 23.8B - **Base model**: tencent/SRPO - **Tags**: gguf - **Downloads**: 492 - **Stars**: 6 - **Last updated**: 2025-09-24 Source: https://www.modelscope.cn/models/wikeeyang/SRPO-for-ComfyUI --- =================================================================================== # SRPO-for-ComfyUI 本模型为 https://www.modelscope.cn/models/tencent-community/SRPO 模型的 转换 和 bf16 / fp8_e4m3fn / Q8_0 / Q4_1 量化版本，以适配 ComfyUI 用户环境正常加载和出图，保持原模型正常的出图效果。 This model is the converted and quantized version of the model: https://www.modelscope.cn/models/tencent-community/SRPO, To adapt the ComfyUI environment for normal loading and output of images, maintaining the original model's normal effects.

## License Agreement Please fall under SRPO license refer license.txt file and refer to the FLUX.1 [dev] Non-Commercial License. 以下部分引用自原模型说明内容： ===================================================================================

Directly Aligning the Full Diffusion Trajectory with Fine-Grained Human Preference

     
Xiangwei Shen1,2*, Zhimin Li1*, Zhantao Yang1, Shiyi Zhang3, Yingfang Zhang1, Donghao Li1,
Chunyu Wang1, Qinglin Lu1, Yansong Tang3,✝
1Hunyuan, Tencent 
2School of Science and Engineering, The Chinese University of Hong Kong, Shenzhen 
3Shenzhen International Graduate School, Tsinghua University 
*Equal contribution  ✝Corresponding author
## Abstract Recent studies have demonstrated the effectiveness of directly aligning diffusion models with human preferences using differentiable reward. However, they exhibit two primary challenges: (1) they rely on multistep denoising with gradient computation for reward scoring, which is computationally expensive, thus restricting optimization to only a few diffusion steps; (2) they often need continuous offline adaptation of reward models in order to achieve desired aesthetic quality, such as photorealism or precise lighting effects. To address the limitation of multistep denoising, we propose Direct-Align, a method that predefines a noise prior to effectively recover original images from any time steps via interpolation, leveraging the equation that diffusion states are interpolations between noise and target images, which effectively avoids over-optimization in late timesteps. Furthermore, we introduce Semantic Relative Preference Optimization (SRPO), in which rewards are formulated as text-conditioned signals. This approach enables online adjustment of rewards in response to positive and negative prompt augmentation, thereby reducing the reliance on offline reward fine-tuning. By fine-tuning the FLUX.1.dev model with optimized denoising and online reward adjustment, we improve its human-evaluated realism and aesthetic quality by over 3x. ### Checkpoints The `diffusion_pytorch_model.safetensors` is online version of SRPO based on [FLUX.1 Dev](https://huggingface.co/black-forest-labs/FLUX.1-dev), trained on HPD dataset with [HPSv2](https://github.com/tgxs002/HPSv2) ### License SRPO is licensed under the License Terms of SRPO. See `./License.txt` for more details. ## Citation If you use SRPO for your research, please cite our paper: ```bibtex @misc{shen2025directlyaligningdiffusiontrajectory, title={Directly Aligning the Full Diffusion Trajectory with Fine-Grained Human Preference}, author={Xiangwei Shen and Zhimin Li and Zhantao Yang and Shiyi Zhang and Yingfang Zhang and Donghao Li and Chunyu Wang and Qinglin Lu and Yansong Tang}, year={2025}, eprint={2509.06942}, archivePrefix={arXiv}, primaryClass={cs.AI}, url={https://arxiv.org/abs/2509.06942}, } ```