The folder contains a collection of scripts whose goals are described below

• generate_metadata.py goes through the different datasets that compose the Well to generate YAML files containing the metadata of the corresponding dataset.

• compute_statistics.py goes through the training set of the different datasets that compose the Well to compute their statistics in terms of mean and std for each tensor field.

• check_thewell_data.py checks for all the HDF5 files that compose the dataset if:

• the name of boundary conditions is consistent;

• the different tensor fields contain NAN values;

• the different tensor fields contain constant frames;

• the different tensor fields contain outliers compared to the mean and std computed on the fly. A default value of $5\sigma$ serves as threshold for characterizing outliers.

• check_thewell_formatting.py checks that a HDF5 file follows the format expected by the Well.

• plot_velocity.py for each dataset, plots the velocity field, if it exists, at four times of the first validation file $0$, $T/3$, $2T/3$ and $T$. Where $T$ is the original length of the simulation.

• create_gif.py creates a gif from a time series input. The script must be edited to first load the time series.

• huggingface folder contains a script to upload data to the Hugging Face hub (upload.py). It also contains a standalone application for visualizing some data of the Well that are hosted on the Hugging Face hub.