using a subset of AMASS

[mateuszwyszynski]

By default the script data.sample_poses and then all the next data preparation and training steps process the entire AMASS dataset. The best option would be for the user to be able to choose, if needed, which data is used.

This can currently be done by changing the default amass_splits inside data/data_splits.py, but I believe we need a better design for this process.

On a related note, the function responsible for data preparation is prepare_vposer_datasets. It is possible to use it just for a chosen data split (train, test or vald) by changing the mode parameter:

prepare_vposer_datasets(posendf_data_dir, amass_splits, amass_dir, mode='train')