train/validation split in the Domino model for the DrivAerML data set

[kk98kk]

Hi everyone,

I’m currently exploring the Domino model and trying to reproduce some of the results.

I found this section about the dataset split:
https://github.com/NVIDIA/physicsnemo-cfd/blob/main/workflows/bench_example/drivaer_ml_files/README.md#drivaerml-dataset-files

From what I can see, it lists which runs are used for training and testing, but I couldn’t find any information on how the training data itself was further divided into training and validation subsets.

Does anybody know how that split was done (e.g., which runs were used for training, validation and testing explicitly)
Having that detail would be very helpful for reproducibility.

Thanks a lot in advance!