data check

Hello, I have two problems which need to check with the authors.
1. authors mentioned 'alignet_train.npz' in readme, but only found train_cluster_border_500.npz in ./data/release_1.1/uncertainty_distillation/. So, are they the same file? When I actually used the later one to train my CNN model, the loss was never reduced. And I checked the image triplets and their similarities and doubt they are not matched. I wonder whether image tirplets were ranked by [sim,sim,odd].
2. the first problem insights me to verify the matched problem using the 'representations.npz'. But my python cannot read this file, maybe because of the file was borken, so would the authors please provide this representation file in .h5 format, which would be more robust for big-size data.