embedding_lookup()的用法
这个函数真的很常用,尤其word2vec
tf.nn.embedding_lookup()就是根据input_ids中的id,寻找embeddings中的第id行。比如input_ids=[1,3,5],则找出embeddings中第1,3,5行,组成一个tensor返回。
实例 1
import tensorflow as tf
import numpy as np
input_ids = tf.placeholder(tf.int32, shape=[None], name="input_ids")
embedding = tf.Variable(np.identity(5, dtype=np.int32))
input_embedding = tf.nn.embedding_lookup(embedding, input_ids)
sess = tf.InteractiveSession()
sess.run(tf.global_variables_initializer())
print("embedding=\n", embedding.eval())
print("input_embedding=\n", sess.run(input_embedding, feed_dict={input_ids: [1, 2, 3, 0, 3, 2, 1]}))
结果
embedding=
[[1 0 0 0 0]
[0 1 0 0 0]
[0 0 1 0 0]
[0 0 0 1 0]
[0 0 0 0 1]]
input_embedding=
[[0 1 0 0 0]
[0 0 1 0 0]
[0 0 0 1 0]
[1 0 0 0 0]
[0 0 0 1 0]
[0 0 1 0 0]
[0 1 0 0 0]]
[Finished in 3.8s]
实例2
import tensorflow as tf
import numpy as np
input_ids = tf.placeholder(dtype=tf.int32, shape=[3, 2])
embedding = tf.Variable(np.identity(5, dtype=np.int32))
input_embedding = tf.nn.embedding_lookup(embedding, input_ids)
sess = tf.InteractiveSession()
sess.run(tf.global_variables_initializer())
print("embedding=\n", embedding.eval())
print("input_embedding=\n", sess.run(input_embedding, feed_dict={input_ids: [[1, 2], [2, 1], [3, 3]]}))
结果
embedding=
[[1 0 0 0 0]
[0 1 0 0 0]
[0 0 1 0 0]
[0 0 0 1 0]
[0 0 0 0 1]]
input_embedding=
[[[0 1 0 0 0]
[0 0 1 0 0]]
[[0 0 1 0 0]
[0 1 0 0 0]]
[[0 0 0 1 0]
[0 0 0 1 0]]]
[Finished in 4.0s]
来自:https://blog.youkuaiyun.com/u013041398/article/details/60955847
https://blog.youkuaiyun.com/laolu1573/article/details/77170407
本文详细介绍TensorFlow中embedding_lookup函数的使用方法,通过两个实例演示如何根据输入ID查找对应的embedding向量,并展示不同输入形状下的输出结果。
933

被折叠的 条评论
为什么被折叠?



