17. Save and read TensorFlow

Save and read TensorFlow

1. Save

Go directly to the code:

import tensorflow as tf
import numpy as np

#Save to file
# remember to define the same dtype and shape wher restore

W = tf.Variable([[1,2,3],[3,4,5]],dtype = tf.float32,name = 'weights')
b = tf.Variable([[1,2,3]],dtype = tf.float32,name='biases')

init = tf.global_variables_initializer()

saver = tf.train.Saver()

with tf.Session() as sess:
    sess.run(init)
    save_path = saver.save(sess,"my_net/save_net.ckpt")
    print("Save to path:",save_path)

Note that the folder must be created in the current directory first.

operation result:

2. Recovery

################################################
# restore variables
# redefine the same shape and same type for your variables
W = tf.Variable(np.arange(6).reshape((2, 3)), dtype=tf.float32, name="weights")
b = tf.Variable(np.arange(3).reshape((1, 3)), dtype=tf.float32, name="biases")

# not need init step

#目前来说只能保存Variable,不能保存整个神经网络

saver = tf.train.Saver()
with tf.Session() as sess:
    saver.restore(sess, "my_net/save_net.ckpt")
    print("weights:", sess.run(W))
    print("biases:", sess.run(b))

operation result:

Successfully read the value saved above.