Instructions to use stabilityai/sd-turbo with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use stabilityai/sd-turbo with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("stabilityai/sd-turbo", 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
- Draw Things
- DiffusionBee
Why SD-Turbo uses epsilon prediction?
#5
by sienna223 - opened
As stated in the paper, SD-Turbo is distilled from SD 2.1 which is a v-prediction model and they enforce zero-terminal SNR. These points let me think SD-Turbo is a v-prediction model, however it is a epsilon-prediction model as indicated in config files of this repo.
So, why is it a epsilon-prediction model?
From the config file it seems they do not enforce zero-terminal SNR.