It includes two datasets that are used in the downstream tasks for evaluating UIBert: App Similar Element Retrieval data and Visual Item Selection (VIS) data. Both datasets are written TFRecords. - View it on GitHub