A pipeline example showing how to use kwargs in a function, and how to use FunctionTransformer from sklearn to decode URLs

functiontransformer_example_with_text_data.py

def fit_model(X, y, **kwargs):
    print(kwargs)
    print(kwargs['max_iter'])
    pipeline = Pipeline([
        ('decode', FunctionTransformer(func=lambda x: x.apply(
            lambda url: parse.unquote(parse.unquote(url))), validate=False)),
        ('cvect', CountVectorizer(binary=True, max_features=1000, stop_words='english',
                                  token_pattern=r'\b\w[\w\.\-\,]+\b')),
        ('clf', MLPClassifier(verbose=1, solver='sgd', max_iter=kwargs['max_iter'] or 1000,
                              tol=0.00001, learning_rate='adaptive', learning_rate_init=0.05)),
        ])

    pipeline.fit(X, y)
    return pipeline