import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("opendiffusionai/sd-bestbase", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]opendiffusionai/sd-bestbase
A convenience starting point who want to train something simple like SD1.5 base model... but also want to minimize stupidity.
This involves:
- replacing original CLIP with LongCLIP ("zer0int/LongCLIP-L-Diffusers")
- replacing original VAE with the later finetuned vae, which is waaay better on faces ("stabilityai/sd-vae-ft-mse")
For your convenience, I also include the script to regenerate it yourself: create-sdbest.py
- Downloads last month
- 15