Fine-Tuning problems

Thank you for your excellent work on the Aurora model — it's been incredibly promising for high-resolution global weather forecasting.

I’m currently attempting to fine-tune Aurora on a custom ERA5 dataset (e.g., adding new surface variables like 2m dewpoint and specific pressure levels such as 950 hPa). While I have gone through the available documentation and supplementary materials, I’m still unclear on the full recommended procedure for implementing fine-tuning in practice.

Could you please provide:

A complete, minimal working example (preferably Python script) for fine-tuning Aurora using a local dataset;

Any official script or reference implementation used for fine-tuning in the paper or experiments.

If such an example already exists, I’d appreciate a link or guidance on where to find it.Thanks again for your help and for open-sourcing such an impactful model!

Review the available documentation and supplementary materials first, then locate any existing fine-tuning scripts or reference implementations for Aurora. Done means providing a complete minimal Python example for fine-tuning with a local ERA5 dataset, or linking to an official example if one already exists.