madclam
commented
6 years ago The ventilation of files should be as follows:
Input_data (or public_data):
DataName_train.data DataName_train.solution DataName_valid.data DataName_test.data DataName_feat.name DataName_label.name DataName_feat.type DataName_public.info [anything else that is NOT xxx_valid.solution, xxxtest.solution not xxxprivate.info]
Reference data for phase 1
DataName_valid.solution DataName_public.info DataName_private.info
Reference data for phase 2
DataName_test.solution DataName_public.info DataName_private.info
In this version of ChaLab, we use the same datasets for input_data and public_data. According to the instructions on how to prepare data, users can supply private informations in files DataName_private.info. Such files should be bundled with all other files ONLY in the input_data zip file (this will correspond to the data that can be re-downloaded from ChaLab by the organizers but not by participants), but DataName_private.info should NOT be part of the zip file of public_data. Alternatively, to avoid duplicating the dataset, we can include DataName_private.info in the competition bundle at the top level, not in the data zip file.