-
Notifications
You must be signed in to change notification settings - Fork 2
Home
Sacha Lewin edited this page Sep 30, 2025
·
9 revisions
Welcome to the Appa wiki!
This wiki will guide you to reproduce our results or generate your own, from downloading and preparing data, to plotting inference results.
If your goal is to quickly reproduce our results, you can directly go to the "Pre-trained Weights" section, which guides you through setting up your directory structure for producing all the results of the paper.
If you wish to produce your own results, we suggest going through the following order:
- Installation.
- Download ERA5 data.
- Compute ERA5 statistics.
- Train an autoencoder.
- Encode ERA5 into latent space.
- Compute latent data statistics.
- Train a denoiser.
- Check your models by generating unconditional samples.
- Generate forecast ensembles.
- Generate reanalysis ensembles.
- Generate baseline data.
- Evaluate the quantitative performance of your models.
- Analyze the reconstruction quality of your AE.
- Analyze the power spectra of all your models and data.
- Analyze the physical consistency of your models and data.
- Render states and plot figures.