Each UCR/UEA subfolder corresponds to a different value of K (1, 2, 5 and 10 for UCR, and 5, 10 and 20 for UEA) and contains an encoder model, an SVM model and a hyperparameter file per UCR/UEA dataset. They also containe pretrained classifiers for all UCR/UEA datasets over concatenated representations from different encoders (see Section 5.1 of the paper).
The 'IHEPC' folder includes pretrained models and precomputed features for the IHEPC dataset.