Input for the training are images of petrophysical properties with well information, and on output are pressure and saturation distribution. The repo should contain everything to run the training. They achieved a correlation of 97% on Cumulative Oil Production.
Interesting repo; consider checking it out and starring it. If you are willing to and have time to dig deeper into this topic, let me know.
Thanks to the authors [Eduardo Maldonado] for making this code open.
Link to the repo: https://github.com/emaldonadocruz/Robust-Spatial-Uncertainty-Modeling-Workflow-Surrogate-Flow-Model
Link to the LinkedIn post