It looks like the training manifest (_/data/imputation_challenge/training_data/metadata_training_data.tsv_) is missing one file that's actually in the folder: C25M22. This is also listed in _Imputation Challenge Metadata_ (https://docs.google.com/spreadsheets/d/111mzOWe1h3NSYv001xK3xBMrj9zO5dZw940qSTDhLos/edit#gid=1876022763) as training data . Another view: 267 training files in _Imputation Challenge Metadata_, only 266 files in_ metadata_training_data.tsv_: ``` >>> impute_meta.groupby('data_split')[['filename']].count() data_split filename B 56 T 267 V 45 >> training_data_manifest.shape (266, 7) >>> validation_data_manifest.shape (45, 7) ```