Can you provide a quantized version? So that it can run on machines with lower video memory.
As the title.
They've been up on HF for a few hours here: https://huggingface.co/city96/HiDream-I1-Full-gguf/tree/main
use the GGUF loader node linked on the model card there.
They've been up on HF for a few hours here: https://huggingface.co/city96/HiDream-I1-Full-gguf/tree/main
use the GGUF loader node linked on the model card there.
hey there, is it working for you? i've found the example workflow but it includes a custom node "Quadruple clip loader" which i cant find anywhere
edit: i was referring to older instruction page which includes that node "https://comfyanonymous.github.io/ComfyUI_examples/hidream/" but i found Calcius' wokrflow has GGUF version of this node and i could download it via comfy manager. workflow: "https://huggingface.co/calcuis/hidream-gguf/tree/main"
I tried using the GGUF model file provided for llama.cpp. It runs, but the generated images are very blurry.