kazuma.py 文件源码

def load_word2emb(self, show_progress=True, batch_size=1000):
        fin_name = self.ensure_file('kazuma.tar.gz', url=self.url)
        seen = set()

        with tarfile.open(fin_name, 'r:gz') as fzip:
            ftxt = fzip.extractfile('charNgram.txt')
            content = ftxt.read()
            ftxt.close()
            lines = content.splitlines()
            if show_progress:
                lines = tqdm(lines)
            batch = []
            for line in lines:
                elems = line.decode().rstrip().split()
                vec = [float(n) for n in elems[-self.d_emb:]]
                word = ' '.join(elems[:-self.d_emb])
                if word in seen:
                    continue
                seen.add(word)
                batch.append((word, vec))
                if len(batch) == batch_size:
                    self.insert_batch(batch)
                    batch.clear()
            if batch:
                self.insert_batch(batch)