# Stable Diffusion: Decouple GPU Ops from Code

[Check out Colab notebook](https://colab.research.google.com/drive/1ZA1q3h9sHd23O3ltYlvUmCGSvHekfIaQ?usp=sharing) with an example of Stable Diffusion running on Everinfer in a decoupled fashion - U-net runs on remote serverless hardware, while everything else runs locally on a CPU with **a 9.5s cold start time!**