tensorflow入门笔记-CSDN博客

本文链接：https://blog.csdn.net/thormas1996/article/details/80733567

最近一直在实现一个seq2seq模型，正好把相关tensorflow和RNN基础模型的知识整理回顾一下

（一）tensorflow入门笔记

（二）RNN，LSTM和GRU原理

（三）attention机制

（四）seq2seq实例详解

基础背景
tensorflow的核心是计算图，即op。
一个基于tensorflow的代码一般包含两步：
1. 创建计算图（op）
2. 运行会话（session）

环境
python3.6 + tensorflow1.8.0

以下是基于Github上的例子，不定时更新

两种打开session方式等价

import tensorflow as tf
y=tf.constant(23)
with tf.Session() as sess: #会自动关闭session
    print(sess.run(y))

sess = tf.Session()
print(sess.run(y))
sess.close()

这两种方式需要在启动session之前构建完计算图
更灵活的方式是

sess = tf.InteractiveSession()

sess.close()

常数和变量

#常数
y = tf.constant(23,name='y')
#变量
y2 = tf.Variable(23,name='y2')
y3 = tf.placeholder(tf.int32) #占位符，类似形参

init = tf.global_variables_initializer() #初始化节点
sess = tf.Session()
sess.run(init)
print(sess.run(y),sess.run(y2))

z = tf.identity(y3) #占位符无法引用自身，放入计算图至少需要一个操作
value = 23
print(sess.run(z,feed_dict={y3:value}))

设置参数

#第一个是参数名称，第二个参数是默认值，第三个是参数描述
tf.app.flags.DEFINE_string('str_name', 'value',"descrip1")
FLAGS = tf.app.flags.FLAGS

加法，乘法

x = tf.placeholder(tf.int32)
y = tf.placeholder(tf.int32)
add = tf.add(x, y)
mul = tf.multiply(x, y)

with tf.Session() as sess:
    print(sess.run(add, feed_dict={x: 2, y: 3}))
    print(sess.run(mul, feed_dict={x: 2, y: 3}))

矩阵运算

matrix1 = tf.constant([[3., 3.]])
matrix2 = tf.constant([[2.],[2.]])
product = tf.matmul(matrix1, matrix2)
add = tf.add(matrix1, matrix2)
with tf.Session() as sess:
    result1 = sess.run(product)
    result2 = sess.run(add)
    print(result1,result2)

线性函数

def linear_function():
    np.random.seed(1)
    X = tf.constant(np.random.randn(3,1),name='X')
    W = tf.constant(np.random.randn(4,3),name='W')
    b = tf.constant(np.random.randn(4,1),name='b')
    Y = tf.add(tf.matmul(W,X),b)
    a = tf.constant(2)
    sess = tf.Session()
    result = sess.run(Y)
    sess.close()
    return result
print(str(linear_function()))

sigmoid函数

def sigmoid(z):
    x = tf.placeholder(tf.float32,name = "x")
    sigmoid = tf.sigmoid(x)
    with tf.Session() as sess:
        result = sess.run(sigmoid,feed_dict = {x:z})
    return result
print(sigmoid(2))

转化为一个one hot向量

def one_hot_matrix(labels,C):
    depth = tf.constant(value=C,name='C')
    one_hot_matrix = tf.one_hot(labels,depth,axis=0)
    sess = tf.Session()

    one_hot = sess.run(one_hot_matrix)
    sess.close()
    return one_hot

labels = np.array([1,2,3,0,2,1])
one_hot = one_hot_matrix(labels, C = 4)
#4x6矩阵，四行分别对应0，1，2，3；元素为1表示出现的位置
print (str(one_hot))

保存模型
```
# 每训练50次保存一次模型
if i% 50 == 0:
    saver.save(sess, self.ckpt_path + self.model_name + '.ckpt', global_step=i)
```
会自动存为三个文件：
model_name.ckpt-iteration_step.data-00000-of-00001, model_name.ckpt-iteration_step.index, model_name.ckpt-iteration_step.meta
以及一个chekpoint文件保存最新模型的名字，内容：
model_checkpoint_path: "model_name.ckpt-iteration_step"
all_model_checkpoint_paths: "model_name.ckpt-0"

读取之前保存的模型

def restore_last_session(ckpt_path):
    saver = tf.train.Saver()
    sess = tf.Session()
    ckpt = tf.train.get_checkpoint_state(ckpt_path)
    if ckpt and ckpt.model_checkpoint_path:
        saver.restore(sess, ckpt.model_checkpoint_path)
    return sess

sess=restore_last_session('yourpath')

ckpt.model_checkpoint_path会自动在checkpoint文件读取最新模型的名字

线性拟合

import tensorflow as tf
import numpy
import matplotlib.pyplot as plt
rng = numpy.random

# Parameters
learning_rate = 0.01
training_epochs = 2000
display_step = 50

# Training Data
train_X = numpy.asarray([3.3,4.4,5.5,6.71,6.93,4.168,9.779,6.182,7.59,2.167,7.042,10.791,5.313,7.997,5.654,9.27,3.1])
train_Y = numpy.asarray([1.7,2.76,2.09,3.19,1.694,1.573,3.366,2.596,2.53,1.221,2.827,3.465,1.65,2.904,2.42,2.94,1.3])
n_samples = train_X.shape[0]

# tf Graph Input
X = tf.placeholder("float")
Y = tf.placeholder("float")

# Create Model

# Set model weights
W = tf.Variable(rng.randn(), name="weight")
b = tf.Variable(rng.randn(), name="bias")

# Construct a linear model
activation = tf.add(tf.multiply(X, W), b)

# Minimize the squared errors
cost = tf.reduce_sum(tf.pow(activation-Y, 2))/(2*n_samples) #L2 loss
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost) #Gradient descent

# Initializing the variables
init = tf.initialize_all_variables()

# Launch the graph
with tf.Session() as sess:
    sess.run(init)

    # Fit all training data
    for epoch in range(training_epochs):
        for (x, y) in zip(train_X, train_Y):
            sess.run(optimizer, feed_dict={X: x, Y: y})

        #Display logs per epoch step
        if epoch % display_step == 0:
            print("Epoch:", '%04d' % (epoch+1), "cost=", \
                "{:.9f}".format(sess.run(cost, feed_dict={X: train_X, Y:train_Y})), \
                "W=", sess.run(W), "b=", sess.run(b) )

    print("Optimization Finished!")
    print("cost=", sess.run(cost, feed_dict={X: train_X, Y: train_Y}), \
          "W=", sess.run(W), "b=", sess.run(b))

    #Graphic display
    plt.plot(train_X, train_Y, 'ro', label='Original data')
    plt.plot(train_X, sess.run(W) * train_X + sess.run(b), label='Fitted line')
    plt.legend()
    plt.show()