AttributeError: 'str' object has no attribute 'decode'. Did you mean: 'encode'?

[makoeppel]

There is a problem in the file aif360/sklearn/preprocessing/learning_fair_representations.py:167 the function scipy.optimize.minimize returns a OptimizeResult which has the attribute message which is already a str object. So the program will file with:

AttributeError: 'str' object has no attribute 'decode'. Did you mean: 'encode'?

Quite a constructed example but this does fail with the error code above. I use aif360==0.5.0 and tried scipy==1.8.1, scipy==1.9.0 and scipy==1.9.1. Fix is trivial just change:

warnings.warn('lbfgs failed to converge: {}'.format(
              res.message.decode()), ConvergenceWarning)

to

warnings.warn('lbfgs failed to converge: {}'.format(
              res.message), ConvergenceWarning)

from aif360.sklearn.preprocessing import LearnedFairRepresentations
import pandas as pd

x = pd.DataFrame(
    [[1, 0.030671, -1.063611, 1.134739,  0.148453],
    [1, 0.837109, -1.008707, 1.134739,  -0.145920],
    [1, -0.042642, 0.245079, -0.420060, -0.145920],
    [1, 1.057047,  0.425801, -1.197459, -0.145920],
    [1, 0.775768,  1.408176, 1.134739,  -0.145920],
    [1, -0.115955, 0.898201, 1.523438,  -0.145920],
    [1, 0.763796, -0.280358, -1.974858, -0.145920],
    [1, 0.983734,  0.188195, -0.420060, -0.145920],
    [1, -0.555830,-1.364279, 1.523438,  1.761142],
    [1, 0.250608, -0.287350, 1.134739,  0.555214]],
    columns=["sex", "1", "2", "3", "4"]
)
x.set_index("sex", inplace=True)
y = pd.DataFrame([0, 0, 0, 0, 0, 0, 0, 1, 1, 1])

model = LearnedFairRepresentations(
    n_prototypes=5,
    reconstruct_weight=1,
    target_weight=1,
    fairness_weight=1,
    verbose=2,
    prot_attr='sex'
)

model.fit(x, y, priv_group=1)