CODEWITHSUNDEEP
randomtensorflowconvolution
unknown_jy
unknown_jy

Reputation: 767

Running session multiple times with tf.random returns different values for conv2d

import tensorflow as tf 
import numpy as np 

x_tf = tf.placeholder('float',[None, 2, 5, 1])
x_np = np.random.noraml(0,1,[1,2,5,1])

# ======== filter option1 and option2 ===========
f_np = np.random.normal(0,1,[1,3,1,1])
f_tf = tf.constant(f_np,'float') # option 1
f_tf = tf.random_normal([1,3,1,1]) # option 2
# ===============================================

x_conv = tf.nn.conv2d(x_tf,f_tf,[1,1,1,1],'SAME')

with tf.Session() as sess:
     tf.global_variables_initializer().run()
     x_conv_np  = sess.run(x_conv, feed_dict={x_tf: x_np})
     x_conv_np2 = sess.run(x_conv, feed_dict={x_tf: x_np})

If I run the code above with option1, I get the same values for x_conv_np and x_conv_np2 However, when I run the above with option2, I get different values for x_conv_np and x_conv_np2.

I am guessing the tf.random_normal gets initialized every time the session is ran. Is this meant to happen? This happens even if I do the tf.set_random_seed Can someone explain how TensorFlow initializes its random variables when the session is ran?  

Upvotes: 2

Views: 1229

Answers (1)

mrry
mrry

Reputation: 126154

All of the random number ops in TensorFlow (including tf.random_normal()) sample a new random tensor each time they runs:

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.

If you want to sample the distribution once and then re-use the result, you should use a tf.Variable and initialize it by running tf.random_normal() once. For example, the following code will print the same random value twice:

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))

Upvotes: 3

Related Questions