8G显存可运行,GGUF版, 文生图。
comfyui更新到最新版
模型选择Q2_K量化版, 文件名 hidream-i1-full-Q2_K.gguf ,放models/unet/GGUF ,
也可以放models/unet,运行时改下地址即可,
国内镜像加速地址:https://hf-mirror.com/city96/HiDream-I1-Full-gguf/tree/main
四个clip模型 ,放models/text_encoders ,
国内镜像加速地址: https://hf-mirror.com/Comfy-Org/HiDream-I1_ComfyUI/tree/main/split_files/text_encoders
VAE用flux的vae即可,也即ae.safetensors ,
国内镜像加速地址: https://hf-mirror.com/Comfy-Org/HiDream-I1_ComfyUI/tree/main/split_files/vae
配置32G内存,8G显存,已验证可运行环境:
Total VRAM 8110 MB, total RAM 31666 MB
pytorch version: 2.6.0+cu126
xformers version: 0.0.29.post3
Set vram state to: NORMAL_VRAM
Disabling smart memory management
Device: cuda:0 NVIDIA P104-100 : native
Using xformers attention
Python version: 3.12.8
ComfyUI version: 0.3.28
ComfyUI frontend version: 1.15.13