First and foremost, install the dependencies in your conda environment:
conda env update -n <your_env> --file environment.yaml
The two notebooks situated at the root of this github that can be used to reproduce testing results on OSE NRT (Near Real Time) data.
Use the data pipeline notebook in order to download and pre-process your desired data. This notebook uses files situated in the data folder.
2023 NRT and reprocessed altimetry data is already baked into this repository in the data folder, which lets you skip this step altogether if you don't need to reproduce the results for different data
Use the results reproducing notebook in order to apply your trained model on the OSE data obtained during the 1st step. The reconstructed outputs will be compared with the OSE reference data and metrics will be computed for each present and future leadtime.
This notebook uses files situated in the reproduce folder, and the code used to run the model (python files at the root of src) are directly coming from the 4DVarNet (forecast) repository.
You're gonna carry that .pth