tensorflow运作方式

定义变量，初始化,一般初始化随机值，或者常值

weights = tf.Variable(tf.random_normal([784, 200],stddev=0.35),
                      name='weights')
from tensorflow.python.framework import ops
ops.reset_default_graph()

biases = tf.Variable(tf.zeros([200]), name='biases')
init_op = tf.global_variables_initializer()

with tf.Session() as sess:
    sess.run(init_op)
    #print sess.run(weights)

保存变量

from tensorflow.python.framework import ops
#ops.reset_default_graph()
g1 = tf.Graph()
print g1
with g1.as_default():
    # 由另一个变量初始化
    weights = tf.Variable(tf.random_normal([784, 200], stddev=0.35), 
                      name='weights')
    w2 =tf.Variable(weights.initialized_value(), name='w2')
    w_twice = tf.Variable(weights.initialized_value()*0.2,name='w_twice')

# 保存变量
    init_op = tf.global_variables_initializer()

    saver = tf.train.Saver()
with tf.Session(graph=g1) as sess:
    sess.run(init_op)
    print sess.run(weights)
    save_path = saver.save(sess, '/tmp/model.ckpt')
    print 'Model saved in file: ',save_path

恢复变量

#ops.reset_default_graph()
# 恢复变量
g2 = tf.Graph()
with g2.as_default():
    weightss = tf.Variable(tf.zeros([784,200]),name='weights')
    w_2 = tf.Variable(weightss, name='w2')
    w_t = tf.Variable(weightss, name='w_twice')
    print weightss.graph
    saver = tf.train.Saver()
with tf.Session(graph=g2) as sess:
    saver.restore(sess, '/tmp/model.ckpt')
    #print sess.run(weightss)
   # print sess.run(w_2)
    print sess.run(w_t)

保存部分变量

from tensorflow.python.framework import ops
ops.reset_default_graph()
g1 = tf.Graph()
print g1
with g1.as_default():
    # 由另一个变量初始化
    weights = tf.Variable(tf.random_normal([784, 200], stddev=0.35), 
                      name='weights')
    w2 =tf.Variable(weights.initialized_value(), name='w2')
    w_twice = tf.Variable(weights.initialized_value()*0.2,name='w_twice')

# 保存变量
    init_op = tf.global_variables_initializer()

    saver = tf.train.Saver({'my_w2':w2,"my_wt":w_twice})
with tf.Session(graph=g1) as sess:
    sess.run(init_op)
    print sess.run(weights)
    save_path = saver.save(sess, '/tmp/model.ckpt')
    print 'Model saved in file: ',save_path

恢复变量

g2 = tf.Graph()
with g2.as_default():
    w_2 = tf.Variable(tf.zeros([784,200]), name='my_w2')
    w_t = tf.Variable(tf.zeros([784,200]), name='my_wt')
    #weightss = tf.Variable(tf.zeros([784,200]),name='my_weight')
    init_op = tf.global_variables_initializer()
    saver = tf.train.Saver()

with tf.Session(graph=g2) as sess:
    sess.run(init_op)
    saver.restore(sess, '/tmp/model.ckpt')

    print sess.run(w_2)