def preprocess_bert_input()

in bert_layer.py [0:0]

• 25 lines of code
• 2 McCabe index (conditional complexity)

def preprocess_bert_input(X, y, max_seq_length, tokenizer, categories):
    # print('Converting data to InputExample format')
    # Convert data to InputExample format
    examples = convert_text_to_examples(X, y)
    # print('Converting data to features')
    # Convert to features
    (
        input_ids,
        input_masks,
        segment_ids,
        labels,
    ) = convert_examples_to_features(
        tokenizer, examples, max_seq_length=max_seq_length
    )
    # For inference label is nont
    if labels[0][0] is None:
        return (input_ids,
                input_masks,
                segment_ids,
                labels)
    enc = preprocessing.OneHotEncoder(categories=[range(categories)], sparse=False)
    col = np.array(labels)
    col = labels.reshape(len(col), 1)
    enc.fit(labels)
    onehotlabels = enc.transform(col)
    labels = onehotlabels

    return (input_ids,
            input_masks,
            segment_ids,
            labels)