Instructions to use ByteDance/Hyper-SD with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use ByteDance/Hyper-SD with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("ByteDance/Hyper-SD") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
Support for fp8?
#23
by Pomfy - opened
I've noticed that Hyper-SD loras do not work when FP8 mode is enabled (in any UI).
Would the new version of these loras support it in the future?
Hi, @Pomfy
We conduct training on BF16 precision and have no plan in support of FP8 currently.
And we are currently working on CFG-Preserved Hyper-SD15/SDXL that facilitate negative prompts and larger guidance scales.
But we are welcoming for public contribution. Thanks for your attention to our work.
Pomfy changed discussion status to closed
Thanks, the new CFG-Preserved Hyper-SD15/SDXL loras appear to work when fp8 mode is enabled.