Text-to-Image
Diffusers
Safetensors
English
Chinese
sana, sana-sprint
SANA-Sprint
1024px_based_image_size
BF16
One-step diffusion
Instructions to use Efficient-Large-Model/Sana_Sprint_1.6B_1024px_diffusers with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Inference
Intermediate timesteps for SCM is not supported when num_inference_steps != 2.
#3
by Simon-arctoon - opened
Hi,
what is the correct call for using Sana-Sprint with 4 steps ?
pipeline( prompt=prompt, num_inference_steps=4)
yields a
Intermediate timesteps for SCM is not supported when num_inference_steps != 2.
Solved : add intermediate_timesteps=None.
There should be a catch somewhere to automatically disable the default intermediate_timesteps when inference steps != 2.