06-lstm-tensorflow.py 文件源码

def get_book(name, language):
    book = lzma.open('../data/{}-common.vpl.xz'.format(name), 'rt').read().splitlines()
    book = [
        [ language.get(w, veczero) for w in words(l)]
        + ([veczero] * (n_steps - len(words(l))))
        for l in book ]
    lens = np.array([len(l) for l in book], dtype=np.int32)

    for verse in book:
        assert len(verse) <= n_steps, "n_steps should be at least {}".format(len(verse))
    return (book, lens)