import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("georgefen/Hi-SD3", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]Enable 2048x2048 Generation for Stable Diffusion 3
SD3 generates noticeable artifacts at high resolutions, indicating that the positional embeddings at high positions have not been well-trained.
I performed 2000 steps of full fine-tuning on the Stable Diffusion 3 medium model using a 2048x2048 dataset, enabling SD3 to achieve good generation results at this resolution.
- Downloads last month
- 1