It seems that the filenames S0766.CSV to S0770.CSV are used for different files in the train_features folder and the val_features folder. Same problem with filenames S1059.CSV to S1068.CSV in the val_features and test_features folder.
Or am I missing something?
Hi @akrueger67. Thank you for your question. We had a mistake on our end when we created the zip archives that led to some extra misnamed files being included. These have been fixed now, with those extra files removed. (We’ve also removed some of the extraneous layers of directories.) Please delete the previous copies of the features and redownload train_features.zip, val_features.zip, test_features.zip, and supplemental_features.zip to ensure that you have the correct features.
The metadata files metadata.csv and supplemental_metadata.csv are unchanged and are still the authority on which samples are part of which data subsets, as well as the expected contents of each of the features files via their MD5 hash.