Code for paper Generative Uncertainty in Diffusion Models.

• python = 3.9
• torch = 2.2.0
• laplace-torch

1. Clone or download this repo. cd yourself to it's root directory.
2. Create and activate python conda enviromnent: conda create --name diff-uq python=3.9
3. Activate conda environment: conda activate diff-uq
4. Install dependencies, using pip install -r requirements.txt

• For ADM experiments on ImageNet 128x128, see README.md in ./ADM
• For UViT experiments on ImageNet 256x256, see README.md in ./UViT

In case of issues with the code, open a github issue (preferred) or reach out to m.jazbec@uva.nl .

The Robert Bosch GmbH is acknowledged for financial support.