Save pandas dataframe with numpy arrays column

you can pickle the data instead.

df.to_pickle('test.csv')
df = pd.read_pickle('test.csv')

This will ensure that the format remains the same. However, it is not human readable

If human readability is an issue, I would recommend converting it to a json file

df.to_json('abc.json')
df = pd.read_json('abc.json')