A HiDiffusion node for ComfyUI
HiDiffusion From: link
--增加adapter_style支持,SDXL需求的显存较大,虽然能跑CPU,但是不推荐,会爆显存,SD1.5测试没问题。
--修复controlnet菜单的加载列表的问题;
--Adding adapter_style support, SDXL requires a large amount of graphics memory. Although it can run on CPU, it is not recommended as it may cause graphics memory "explosion". SD1.5 testing is not a problem.
--Fix the issue with loading lists in the Controlnet menu;
1、ParlerTTS node:ComfyUI_ParlerTTS
2、Llama3_8B node:ComfyUI_Llama3_8B
3、HiDiffusion node:ComfyUI_HiDiffusion_Pro
4、ID_Animator node: ComfyUI_ID_Animator
5、StoryDiffusion node:ComfyUI_StoryDiffusion
6、Pops node:ComfyUI_Pops
7、stable-audio-open-1.0 node :ComfyUI_StableAudio_Open
8、GLM4 node:ComfyUI_ChatGLM_API
9、CustomNet node:ComfyUI_CustomNet
10、Pipeline_Tool node :ComfyUI_Pipeline_Tool
11、Pic2Story node :ComfyUI_Pic2Story
12、PBR_Maker node:ComfyUI_PBR_Maker
13、ComfyUI_Streamv2v_Plus node:ComfyUI_Streamv2v_Plus
14、ComfyUI_MS_Diffusion node:ComfyUI_MS_Diffusion
-- 增加 manne加速Lora
-- 加入controlnet-tile-sdxl的支持,内置图片预处理,默认512尺寸,新增apply_window_attn 条件控制。
--修复节点连接逻辑,现在文生图模式,无需接入image,无controlnet也无需接入control_image
--支持SDXL-lighting\Hyper\LCM\DMD2\的加速Unet,建议适当提高步数;
--基于官方的更新,加入lora支持,需要填关键词;
--加入skip,去掉意义不大的其他参数;
--支持所有的SDXL controlnet模型(仅支持配置config文件的diffuser加载模式);
--你可以修改model.yaml文件,添加其他的可能支持SDXL“扩散模型”或者“controlnet”或者SDXL“unet”模型;
-- add manne lighting lora
--Added support for control net file sdxl, built-in image preprocessing, default size of 512, and added condition control for apply_window_attn.
--Fix node connection logic, now in text-based graphics mode, there is no need to connect to image, no controllnet, and no need to connect to controll_image
--Support acceleration Unet for SDXL lighting, Hyper, LCM, and DMD2. It is recommended to increase the number of steps appropriately;
--Based on official updates, adding support for Lora requires filling in keywords;
--Add skip and remove other parameters that are not significant;
--Supports all SDXL controllnet models (only supports diffuser loading mode for configuring config files);
--You can modify the model.yaml file and add other models that may support SDXL "diffusion models" or "controllnet" or SDXL "unet" models;
1.1 In the .\ComfyUI \ custom_node directory, run the following:
git clone https://github.com/smthemex/ComfyUI_HiDiffusion_Pro.git
1.2 using it
diffusers >=0.28.0
yaml
3种调用模型节点的方法
一种是:在repo_id填写诸如"stabilityai/stable-diffusion-xl-base-1.0" 这样的标准抱脸的repo_id,系统会自己下载或者查找.cache路径,这个要开全局外网或者你开启了hf镜像
一种是:把你下载好的模型放在comfyUI的"models/diffusers"路径下,记得不要改最尾的模型路径名“比如stable-diffusion-xl-base-1.0”,就可以用菜单直接调用。(注意:要用菜单的方式,必须删掉repo_id默认的"stabilityai/stable-diffusion-xl-base-1.0",让repo_id留空,controlnet_id一样,要调用菜单,必须留空controlnet_id)
一种是:你在repo_id或controlnet_id直接填写诸如“x:/xx/xx/stabilityai/stable-diffusion-xl-base-1.0” 这样的已经下载好的扩散模型的绝对路径也能用
Three methods for calling model nodes
Filling in the corresponding repo_id will download the model
or
in the diffuse directory, store and use the downloaded model (Notice:in this method,repo_id or controlnet_repo_id must be empty)
or
using repo_id,and Fill in the absolute path of your diffusers model,using"/"
controlnet模型存放示例:
├── ComfyUI/models/diffusers/
| ├──xinsir/controlnet-openpose-sdxl-1.0
| ├── config.json
| ├── diffusion_pytorch_model.fp16.safetensors
| ├──xinsir/controlnet-scribble-sdxl-1.0
| ├── config.json
| ├── diffusion_pytorch_model.fp16.safetensors
| ├──xinsir/controlnet-tile-sdxl-1.0
| ├── config.json
| ├── diffusion_pytorch_model.fp16.safetensors
| ├──diffusers/controlnet-canny-sdxl-1.0
| ├── config.json
| ├── diffusion_pytorch_model.fp16.safetensors
| ├──diffusers/controlnet-depth-sdxl-1.0
| ├── config.json
| ├── diffusion_pytorch_model.fp16.safetensors
| ├──TheMistoAI/MistoLine
| ├── config.json
| ├── diffusion_pytorch_model.fp16.safetensors
ip_adapter model 模型存放示例:
如果不存放对应,运行时会自动下载。 If the corresponding file is not stored, it will be automatically downloaded at runtime.
├── ComfyUI/custom_nodes/ComfyUI_HiDiffusion_Pro/weights/
| ├──models
| ├── ip-adapter_sd15.bin
| ├── image_encoder/
| ├── config.json
| ├── model.safetensors
| ├──sdxl_models
| ├── ip-adapter_sdxl.bin
| ├── image_encoder/
| ├── config.json
| ├── model.safetensors
部分支持的模型请查看model.yaml 文件
For partially supported models, please refer to the model.yaml file
txt2img using ip_adapter_style
img2img use controlnet and lora 图生图加controlnet和lora
txt2img use XL/sd1.5/XL turbo/background... 文生图
txt2img + controlnet 文生图加controlnet .
img2img use Hyper unet 图生图加加速unet
@article{zhang2023hidiffusion,
title={HiDiffusion: Unlocking Higher-Resolution Creativity and Efficiency in Pretrained Diffusion Models},
author={Zhang, Shen and Chen, Zhaowei and Zhao, Zhenyu and Chen, Yuhao and Tang, Yao and Liang, Jiajun},
journal={arXiv preprint arXiv:2311.17528},
year={2023}
}