Please share the information related to randomly selected samples from the AMASS dataset for training & evaluation as mentioned in the paper. Also, do you guys plan to release the training code for the draping network?