tfds.as_dataframe throws an exception

/!\ IF YOU WANT PEOPLE TO HELP YOU, PLEASE GIVE AS MUCH DETAIL AS POSSIBLE, INCLUDING THE FULL STACKTRACE AND CODE SNIPPET

Short description
Calling tfds.as_dataframe on the coco_captions dataset returns an error having to do with the constructor

Environment information

• Operating System: Windows 10

• Python version: 3.7.10

• tensorflow-datasets/tfds-nightly version: 4.3.0

• tensorflow/tf-nightly version: 2.5.0

• Does the issue still exists with the last tfds-nightly package (pip install --upgrade tfds-nightly) ?
  Yes
  Reproduction instructions

import tensorflow_datasets as tfds
(coco_train, coco_val, coco_test), coco_info = tfds.load('coco_captions', split=['train','val','test'], with_info=True)
df = tfds.as_dataframe(coco_train.take(10), coco_info)

Link to logs
Leads to:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-320-2a3721f476e4> in <module>
----> 1 df = tfds.as_dataframe(coco_val.take(10))
      2 df.head()

~\Anaconda3\envs\RoastBot\lib\site-packages\tensorflow_datasets\core\as_dataframe.py in as_dataframe(ds, ds_info)
    220   columns = _make_columns(ds.element_spec, ds_info=ds_info)
    221   rows = [_make_row_dict(ex, columns) for ex in dataset_utils.as_numpy(ds)]
--> 222   df = StyledDataFrame(rows)
    223   df.current_style.format({c.name: c.format_fn for c in columns if c.format_fn})
    224   return df

~\Anaconda3\envs\RoastBot\lib\site-packages\tensorflow_datasets\core\as_dataframe.py in __init__(self, *args, **kwargs)
    146 
    147   def __init__(self, *args, **kwargs):
--> 148     super().__init__(*args, **kwargs)
    149     # Use name-mangling for forward-compatibility in case pandas
    150     # adds a `_styler` attribute in the future.

TypeError: object.__init__() takes exactly one argument (the instance to initialize)

Expected behavior
df should be a pandas dataframe.

Additional context
Add any other context about the problem here.

Start in tensorflow_datasets/core/as_dataframe.py at StyledDataFrame.init and the as_dataframe call shown in the traceback. Reproduce the coco_captions example, inspect the constructor failure, and verify that tfds.as_dataframe returns a pandas dataframe without the TypeError.