How to change model size for latent diffusion

[karllandheer]

I am trying to modify the tutorial shown here:
https://github.com/Project-MONAI/GenerativeModels/blob/main/tutorials/generative/2d_diffusion_autoencoder/2d_diffusion_autoencoder_tutorial.ipynb

to accomodate my data which is larger in size. For example, my images are 12 channels with an image size of 256x256. Even with a batch size of 1 this gives OOM for GPUs with 16GB RAM. Does anyone know how to address this? One way is to do random cropping transforms, however this then makes inference a bit more confusing. Alternatively I could reduce the model size of Diffusion_AE, however I'm not 100% sure how to do that. Any ideas?