data.py 文件源码

def _get_word_vectors(self, desc, word_embedding):
        output = []
        len_desc = []
        for i in range(desc.shape[1]):
            words = self._nums2chars(desc[:, i])
            words = split_sentence_into_words(words)
            word_vecs = torch.Tensor([word_embedding[w] for w in words])
            # zero padding
            if len(words) < self.max_word_length:
                word_vecs = torch.cat((
                    word_vecs,
                    torch.zeros(self.max_word_length - len(words), word_vecs.size(1))
                ))
            output.append(word_vecs)
            len_desc.append(len(words))
        return torch.stack(output), len_desc