Padding with zeros and replacing missing vals with NaNs

[nadthoob]

In function load_from_tsfile (momentfm/utils/data.py) you replace missing vals (replace_missing_vals_with) with NaN by default. However, when padding you use zero, as in ClassificationDataset.__getitem__ (momentfm/data/classification_dataset.py):
timeseries = np.pad(timeseries, (self.seq_len - timeseries_len, 0))

Why not both pad and fill missing values with zeros, or with NaN; why use two different filler values?