tensorflow API：tf.set_random_seed同一随机种子不改变的设置

会话级种子

设置随机函数的seed参数，对应的变量可以跨session生成相同的随机数：

• 例子

tf.reset_default_graph()
a = tf.random_uniform([1], seed=1)
b = tf.random_normal([1])

# Repeatedly running this block with the same graph will generate the same
# sequence of values for 'a', but different sequences of values for 'b'.
print("Session 1")
with tf.Session() as sess1:
  print(sess1.run(a))  # generates 'A1'
  print(sess1.run(a))  # generates 'A2'
  print(sess1.run(b))  # generates 'B1'
  print(sess1.run(b))  # generates 'B2'

print("Session 2")
with tf.Session() as sess2:
  print(sess2.run(a))  # generates 'A1'
  print(sess2.run(a))  # generates 'A2'
  print(sess2.run(b))  # generates 'B3'
  print(sess2.run(b))  # generates 'B4'

结果：变量a跨会话生成的随机数相同，b则不同。

Session 1
[0.2390374]
[0.22267115]
[0.9374042]
[0.57995176]
Session 2
[0.2390374]
[0.22267115]
[-1.6857139]
[0.6809292]

图级种子

通过tf.set_random_seed设定种子数，后面定义的全部变量都可以跨会话生成相同的随机数。
* 例子：

tf.reset_default_graph()
tf.set_random_seed(1234)
a = tf.random_uniform([1])
b = tf.random_normal([1])

# Repeatedly running this block with the same graph will generate the same
# sequences of 'a' and 'b'.
print("Session 1")
with tf.Session() as sess1:
  print(sess1.run(a))  # generates 'A1'
  print(sess1.run(a))  # generates 'A2'
  print(sess1.run(b))  # generates 'B1'
  print(sess1.run(b))  # generates 'B2'

print("Session 2")
with tf.Session() as sess2:
  print(sess2.run(a))  # generates 'A1'
  print(sess2.run(a))  # generates 'A2'
  print(sess2.run(b))  # generates 'B1'
  print(sess2.run(b))  # generates 'B2'

结果：

Session 1
[0.96046877]
[0.8362156]
[0.4987599]
[0.54880583]
Session 2
[0.96046877]
[0.8362156]
[0.4987599]
[0.54880583]

同一个sess生成两次结果不一样

情况1：设置了seed参数但是同样的两次结果不一样：

import tensorflow as tf
tf.reset_default_graph()
embedding1 = tf.random_uniform(seed=1234,minval=0,maxval=10,shape=(5,5))
with tf.Session() as sess:
    print("embedding1:",sess.run(embedding1))
    print("embedding1:",sess.run(embedding1))

结果：

embedding1: [[8.48307    3.2357132  3.067001   0.69699764 9.138565  ]
 [1.7047906  2.833712   3.5627055  5.4155626  0.75256824]
 [0.7449007  8.65954    5.5905952  0.09341955 0.67013025]
 [4.9774456  2.989055   8.094423   6.584983   4.8154235 ]
 [1.5170729  5.910175   3.3441043  1.5515935  2.4549198 ]]
embedding1: [[9.550312   0.6379819  1.1600447  9.080754   1.3742483 ]
 [6.237757   5.177211   2.1648896  5.644026   5.9398937 ]
 [1.3058889  5.9784565  1.05986    5.3199635  4.053446  ]
 [4.7531548  2.615583   5.401784   5.989144   3.236668  ]
 [8.574158   1.7371535  0.89173794 7.255272   8.292801  ]]

情况2：定义的两个变量的随机生成函数一样，种子一样，结果一样：

import tensorflow as tf
tf.reset_default_graph()
embedding1 = tf.random_uniform(seed=1234,minval=0,maxval=10,shape=(5,5))
embedding2 = tf.random_uniform(seed=1234,minval=0,maxval=10,shape=(5,5))
sess = tf.Session()
print("embedding1:",sess.run(embedding1))
print("embedding2:",sess.run(embedding2))

结果：

embedding1: [[8.48307    3.2357132  3.067001   0.69699764 9.138565  ]
 [1.7047906  2.833712   3.5627055  5.4155626  0.75256824]
 [0.7449007  8.65954    5.5905952  0.09341955 0.67013025]
 [4.9774456  2.989055   8.094423   6.584983   4.8154235 ]
 [1.5170729  5.910175   3.3441043  1.5515935  2.4549198 ]]
embedding2: [[8.48307    3.2357132  3.067001   0.69699764 9.138565  ]
 [1.7047906  2.833712   3.5627055  5.4155626  0.75256824]
 [0.7449007  8.65954    5.5905952  0.09341955 0.67013025]
 [4.9774456  2.989055   8.094423   6.584983   4.8154235 ]
 [1.5170729  5.910175   3.3441043  1.5515935  2.4549198 ]]

为了一样，在一个sess.run中运行：

import tensorflow as tf
tf.reset_default_graph()


tf.set_random_seed(1234)
embedding1 = tf.random_uniform(minval=0,maxval=10,shape=(5,5))
embedding2 = tf.random_uniform(minval=0,maxval=10,shape=(5,5))
ids = [2,1,0]
some_embedding = tf.nn.embedding_lookup(embedding1,ids=ids)
with tf.Session() as sess:
    print("embedding table:{}n result:{}".format(*sess.run([embedding1,some_embedding])))
    #print("result：",sess.run(some_embedding))
    #print(*sess.run([embedding1,some_embedding]))

结果：

embedding table:[[9.604688   5.811516   6.4159     9.621765   0.5434954 ]
 [4.1893444  5.8865128  7.9785547  8.296125   8.388672  ]
 [0.41017294 5.350975   4.223858   9.372683   9.035423  ]
 [1.5520871  1.4448678  3.6297297  8.929963   5.167904  ]
 [1.5287185  6.8655777  8.099522   1.5997577  6.136037  ]]
 result:[[0.41017294 5.350975   4.223858   9.372683   9.035423  ]
 [4.1893444  5.8865128  7.9785547  8.296125   8.388672  ]
 [9.604688   5.811516   6.4159     9.621765   0.5434954 ]]

设为变量variable，得到同一个session可复用的结果：

f_tf = tf.Variable(tf.random_normal([1, 3, 1, 1]))
# ...
init_op = tf.global_variables_initializer()
# ...
with tf.Session() as sess:
    sess.run(init_op)
    print(sess.run(f_tf))
    print(sess.run(f_tf))

结果：

[[[[1.0966153 ]]

  [[0.2081248 ]]

  [[0.15507936]]]]
[[[[1.0966153 ]]

  [[0.2081248 ]]

  [[0.15507936]]]]

原因：TensorFlow has several ops that create random tensors with different distributions. The random ops are stateful, and create new random values each time they are evaluated.

参考：https://stackoverflow.com/questions/41637974/running-session-multiple-times-with-tf-random-returns-different-values-for-conv2

最后

以上就是纯真芝麻为你收集整理的tensorflow API：tf.set_random_seed同一随机种子不改变的设置的全部内容，希望文章能够帮你解决tensorflow API：tf.set_random_seed同一随机种子不改变的设置所遇到的程序开发问题。

如果觉得靠谱客网站的内容还不错，欢迎将靠谱客网站推荐给程序员好友。