Saver 保存读取

保存

import tensorflow as tf
import numpy as np

## Save to file
# remember to define the same dtype and shape when 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')


saver = tf.train.Saver()

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

"""    
Save to path:  my_net/save_net.ckpt
"""

提取

# 先建立 W, b 的容器

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")

# 这里不需要初始化步骤 init= tf.initialize_all_variables()

saver = tf.train.Saver()

with tf.Session() as sess:

\# 提取变量

saver.restore\(sess, "my\_net/save\_net.ckpt"\)

print\("weights:", sess.run\(W\)\)

print\("biases:", sess.run\(b\)\)

"""

weights: [[ 1. 2. 3.]

      \[ 3.  4.  5.\]\]

biases: [[ 1. 2. 3.]]

"""

import tensorflow as tf
import matplotlib.pyplot as plt
import numpy as np

tf.set_random_seed(1)
np.random.seed(1)

# fake data
x = np.linspace(-1, 1, 100)[:, np.newaxis]          # shape (100, 1)
noise = np.random.normal(0, 0.1, size=x.shape)
y = np.power(x, 2) + noise                          # shape (100, 1) + some noise


def save():
    print('This is save')
    # build neural network
    tf_x = tf.placeholder(tf.float32, x.shape)  # input x
    tf_y = tf.placeholder(tf.float32, y.shape)  # input y
    l = tf.layers.dense(tf_x, 10, tf.nn.relu)          # hidden layer
    o = tf.layers.dense(l, 1)                     # output layer
    loss = tf.losses.mean_squared_error(tf_y, o)   # compute cost
    train_op = tf.train.GradientDescentOptimizer(learning_rate=0.5).minimize(loss)

    sess = tf.Session()
    sess.run(tf.global_variables_initializer())  # initialize var in graph

    saver = tf.train.Saver()  # define a saver for saving and restoring

    for step in range(100):                             # train
        sess.run(train_op, {tf_x: x, tf_y: y})

    saver.save(sess,'my_net/save_net.ckpt', write_meta_graph=False)  # meta_graph is not recommended

    # plotting
    pred, l = sess.run([o, loss], {tf_x: x, tf_y: y})
    plt.figure(1, figsize=(10, 5))
    plt.subplot(121)
    plt.scatter(x, y)
    plt.plot(x, pred, 'r-', lw=5)
    plt.text(-1, 1.2, 'Save Loss=%.4f' % l, fontdict={'size': 15, 'color': 'red'})


def reload():
    print('This is reload')
    # build entire net again and restore
    tf_x = tf.placeholder(tf.float32, x.shape)  # input x
    tf_y = tf.placeholder(tf.float32, y.shape)  # input y
    l_ = tf.layers.dense(tf_x, 10, tf.nn.relu)          # hidden layer
    o_ = tf.layers.dense(l_, 1)                     # output layer
    loss_ = tf.losses.mean_squared_error(tf_y, o_)   # compute cost

    sess = tf.Session()
    # don't need to initialize variables, just restoring trained variables
    saver = tf.train.Saver()  # define a saver for saving and restoring
    saver.restore(sess,'my_net/save_net.ckpt')

    # plotting
    pred, l = sess.run([o_, loss_], {tf_x: x, tf_y: y})
    plt.subplot(122)
    plt.scatter(x, y)
    plt.plot(x, pred, 'r-', lw=5)
    plt.text(-1, 1.2, 'Reload Loss=%.4f' % l, fontdict={'size': 15, 'color': 'red'})
    plt.show()


save()

# destroy previous net
tf.reset_default_graph()

reload()