Created by: CUITCHE

使用官方的“快速上手”代码，得到几组词的向量

import numpy as np
import paddle.fluid.dygraph as D
from ernie.tokenizing_ernie import ErnieTokenizer
from ernie.modeling_ernie import ErnieModel

D.guard().__enter__() # activate paddle `dygrpah` mode

model = ErnieModel.from_pretrained('ernie-1.0')    # Try to get pretrained model from server, make sure you have network connection
tokenizer = ErnieTokenizer.from_pretrained('ernie-1.0')

ids, _ = tokenizer.encode('hello world')
ids = D.to_variable(np.expand_dims(ids, 0))  # insert extra `batch` dimension
pooled, encoded = model(ids)                 # eager execution
print(pooled.numpy())                        # convert  results to numpy

我根据词向量计算余弦相似度，缺没得到预期的值：

def cos_sim(vector_a, vector_b):
    """计算两组向量的余弦值"""
    vector_a = np.mat(vector_a)
    vector_b = np.mat(vector_b)
    num = float(vector_a * vector_b.T)
    denom = np.linalg.norm(vector_a) * np.linalg.norm(vector_b)
    cos = num / denom
    sim = 0.5 + 0.5 * cos
    return sim

def calc_cos_sim(ws):
    res = list()
    for w in ws:
        ids, _ = tokenizer.encode(w)
        ids = D.to_variable(np.expand_dims(ids, 0))
        pooled, encoded = model(ids)
        res.append(list(pooled.numpy()[0]))
    return cos_sim(res[0], res[1])


if __name__ == '__main__':
    # activate paddle `dygrpah` mode
    with D.guard():
        model = ErnieModel.from_pretrained('ernie-1.0')
        tokenizer = ErnieTokenizer.from_pretrained('ernie-1.0')
        pairs = [('法师', '律师'), ('医院', '法院'), ('展架', '展柜')]
        for p in pairs:
            print(f'{p} {calc_cos_sim(p)}')

A、('法师', '律师') 0.9383215354753538 B、('医院', '法院') 0.978857960588891 C、('展架', '展柜') 0.9633232497133388

实际上，C组应该更接近1，而A、B更远离1。

另：每次计算出来的值也不完全相同，有小的波动，如我再次运行： A、('法师', '律师') 0.956882570176395 B、('医院', '法院') 0.9377995660564593 C、('展架', '展柜') 0.9481575568652009