def simple_predict(model, text, length_to_predict):
    targets = []

    bpe_string = bpe.encode(text, output_type=yttm.OutputType.SUBWORD, bos=True, eos=False)

    bpe_string_id = []
    for s in bpe_string:
        bpe_string_id.append(bpe.subword_to_id(s))

    for i in range(length_to_predict - len(bpe_string)):
        x = torch.LongTensor([bpe_string_id])
        predict = model(x)[0]
        pred1 = torch.argmax(torch.sigmoid(predict.cpu().contiguous().view(-1, num_tokens)), dim=1)[-1]
        bpe_string_id.append(pred1)
        bpe_string.append(bpe.id_to_subword(pred1))
        if bpe_string[-1] == '<EOS>':
            break

    return bpe_string_id