How were human poses generated with VPoser?

[kingsleykimm]

In the data/humanoids/humanoid_data, the README.md says that the left_hand and right_hand keys in a given motion data .pkl file show the grid of 48 poses generated from the VPoser model (https://github.com/nghorbani/human_body_prior). However, there doesn't seem to be any info on how this was done, and looking at the VPoser model all that is said is that you can either use the IKEngine and input key joints, but there's limited documentation.