Created by: PengheLiu

paddle: 1.3.1 on mac cpu version 直接运行的这个例子。 http://paddlepaddle.org/documentation/docs/zh/1.3/api_cn/layers_cn.html#permalink-110-nce

import paddle as paddle
import paddle.fluid as fluid
from paddle.fluid import layers

import numpy as np

window_size = 5
words = []
for i in xrange(window_size):
        words.append(layers.data(
                name='word_{0}'.format(i), shape=[1], dtype='int64'))

dict_size = 10000
label_word = int(window_size / 2) + 1

embs = []
for i in xrange(window_size):
        if i == label_word:
                continue

        emb = layers.embedding(input=words[i], size=[dict_size, 32],
                                                   param_attr='emb.w', is_sparse=True)
        embs.append(emb)

embs = layers.concat(input=embs, axis=1)

#使用custom distribution
dist = fluid.layers.assign(input=np.array([0.05,0.5,0.1,0.3,0.05]).astype("float32"))
loss = layers.nce(input=embs, label=words[label_word],
                          num_total_classes=5, param_attr='nce.w',
                          bias_attr='nce.b',
                          num_neg_samples=3,
                          sampler="custom_dist",
                          custom_dist=dist)