Instructions to use KBlueLeaf/MOSS-Image-heal with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use KBlueLeaf/MOSS-Image-heal with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("KBlueLeaf/MOSS-Image-heal", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
Purpose?
#2
by sorryhyun - opened
Hi, I'm quite interested in your work. So what would be the final purpose of this work? I expect enabling fp16 training for both accuracy and stability if being ambitious. Thanks for the nice work anw
It try to provide a MOSS-image-base that is easier to finetune, and some community member already report that with healed version the finetune is way more stable and effective.
It also allow FP16 to finetune/inference the model without special handling as well.
Basically it is trained to fix the bad numerical behavior in MOSS-image models, while try to align with the original output.
I will also use similar infra to train KohakuMOSS in future I guess