Reorder columns

[nikita-galileo]

Is there a way to reorder the columns in a dataset? I notice remove_columns and rename_columns and have even tried the following to no avail:

def reorder_cols(sample):
    sample = {col: sample[col] for col in new_cols_order}
    return sample

new_ds = new_ds.map(reorder_cols)

[mariosasko]

Hi! The columns can be reordered either by performing a Dataset.cast with the reordered features dictionary or by replacing the initial features with the reordered ones (ds.info.features = reordered_features). I suggest you use the former approach as the latter relies on the internal decoding logic.