Instructions to use waifu-diffusion/wd-1-5-beta with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use waifu-diffusion/wd-1-5-beta with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("waifu-diffusion/wd-1-5-beta", 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
General training discoveries
#2
by CodeExplode - opened
Hi!
I'm just wondering if there are any general discoveries regarding training that the creators are willing to share? e.g. Does caption dropout seem to help? Did you settle on randomizing all tags or sometimes leave some at the front? (e.g. 1man, 1woman, etc)
Most of our info is in the release notes. Partial caption dropout seems to work well, but we don't really have the resources to do a proper test. Tags are completely randomized. The training code is here if you want to see https://github.com/waifu-diffusion/network-trainer
Thanks!
CodeExplode changed discussion status to closed