def get_one_vec(**kwargs):
text = kwargs.get('text', '默认')
is_lite = kwargs.get('is_lite', 0)
vecs = text2vec(text)
if is_lite:
vecs = vecs[:, range(0, 768, 4)] # 64*3 = 192维度
return normalize(vecs).tolist()
def get_batch_vec(**kwargs):
texts = kwargs.get('texts', ['默认'])
is_lite = kwargs.get('is_lite', 0)
vecs = texts2vec(texts)
if is_lite:
vecs = vecs[:, range(0, 768, 4)] # 64*3 = 192维度
return normalize(vecs).tolist()
def sent2vec(sent, w2v, tokenizer=str.split):
return normalize(np.row_stack([w2v[w] for w in tokenizer(sent)]).mean(0))