Instructions to use lightx2v/Qwen-Image-Lightning with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use lightx2v/Qwen-Image-Lightning with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("Qwen/Qwen-Image", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("lightx2v/Qwen-Image-Lightning") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps
- Draw Things
- DiffusionBee
bf16?
#12
by alexloops - opened
What is the difference between BF16 and the other in terms of quality? Which one is better than the other?
As far as I know BF16 highest widest precision (Almost like FP32) so much better compared to other.. if they are FP16 or lower FP8. But your GPU needs to support BF16 otherwise it will be software emulated to FP32 during runtime and it will be slower. I'm Just curious why the size of BF16 lower in files section than the other if they are FP16 or FP8.
curious why the size of BF16 lower
Cause BF16 is twice as small as FP32. That's what the other precision is.