Pseudo labels for semi-supverised training not given

In train_pl.py it seems like we do not have the pseudo_labels necessary for the semi-supervised pl training.

Line 115 ff.:

pseudo_labels_path = f'../pseudo_labels/pseudo_labels_fold{opts.fold}.p'  # FIXME not given?
with open(pseudo_labels_path, 'rb') as f:
    long_preds, long_stds, short_preds, short_stds = pickle.load(f)

Update: Seems like we have to generate these labels first with predict.py after model training (step 2) is completed