按照我修改的代码,方可运行。花了我好几个小时,希望大家不要重复犯错误,浪费时间
首先导入rnn:
from tensorflow.contrib import rnn
lstm_cell = rnn.BasicLSTMCell(n_hidden,state_is_tuple=False)#必须加上False不然报错
# 28长度的sequence,所以是需要分解位28次
_X = tf.split(_X, n_steps, 0) # n_steps * (batch_size, n_hidden)
#x = tf.split(x, n_steps, 0) # tf.split(value, num_or_size_splits, axis) versions > 0.12.0
# 开始跑RNN那部分
outputs, states = rnn.static_rnn(lstm_cell, _X, initial_state=_istate)
在LSTM单元中,有2个状态值,分别是c和h,分别对应于下图中的c和h。其中h在作为当前时间段的输出的同时,也是下一时间段的输入的一部分。
当state_is_tuple=True的时候,state是元组形式,state=(c,h)。
如果是False,那么state是一个由c和h拼接起来的张量,state=tf.concat(1,[c,h])。
在运行时,则返回2值,一个是h,还有一个state。
#coding:utf-8
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("MNIST_data/",one_hot=True)
import tensorflow as tf
import numpy as np
from tensorflow.contrib import rnn
'''
MNIST的数据是一个28*28的图像,这里RNN测试,把他看成一行行的序列(28维度(28长的sequence)*28行)
'''
# RNN学习时使用的参数
learning_rate = 0.001
training_iters = 100000
batch_size = 128
display_step = 10
# 神经网络的参数
n_input = 28 # 输入层的n
n_steps = 28 # 28长度
n_hidden = 128 # 隐含层的特征数
n_classes = 10 # 输出的数量,因为是分类问题,0~9个数字,这里一共有10个
# 构建tensorflow的输入X的placeholder
x = tf.placeholder("float", [None, n_steps, n_input])
# tensorflow里的LSTM需要两倍于n_hidden的长度的状态,一个state和一个cell
# Tensorflow LSTM cell requires 2x n_hidden length (state & cell)
istate = tf.placeholder("float", [None, 2 * n_hidden])
# 输出Y
y = tf.placeholder("float", [None, n_classes])
# 随机初始化每一层的权值和偏置
weights = {
'hidden': tf.Variable(tf.random_normal([n_input, n_hidden])), # Hidden layer weights
'out': tf.Variable(tf.random_normal([n_hidden, n_classes]))
}
biases = {
'hidden': tf.Variable(tf.random_normal([n_hidden])),
'out': tf.Variable(tf.random_normal([n_classes]))
}
#构建RNN
def RNN(_X, _istate, _weights, _biases):
# 规整输入的数据
_X = tf.transpose(_X, [1, 0, 2]) # permute n_steps and batch_size
_X = tf.reshape(_X, [-1, n_input]) # (n_steps*batch_size, n_input)
# 输入层到隐含层,第一次是直接运算
_X = tf.matmul(_X, _weights['hidden']) + _biases['hidden']
# 之后使用LSTM
#lstm_cell = rnn_cell.LayerNormBasicLSTMCell(n_hidden, forget_bias=1.0)
lstm_cell = rnn.BasicLSTMCell(n_hidden,state_is_tuple=False)
# 28长度的sequence,所以是需要分解位28次
_X = tf.split(_X, n_steps, 0) # n_steps * (batch_size, n_hidden)
#x = tf.split(x, n_steps, 0) # tf.split(value, num_or_size_splits, axis) versions > 0.12.0
# 开始跑RNN那部分
outputs, states = rnn.static_rnn(lstm_cell, _X, initial_state=_istate)
#rnn.rnn.dynamic_rnn()
return tf.matmul(outputs[-1],_weights['out']) + biases['out']
pred = RNN(x, istate, weights, biases)
# 定义损失和优化方法,其中算是为softmax交叉熵,优化方法为Adam
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred, labels=y)) # Softmax loss
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost) # Adam Optimizer
# 进行模型的评估,argmax是取出取值最大的那一个的标签作为输出
correct_pred = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
# 初始化
init = tf.initialize_all_variables()
# 开始运行
with tf.Session() as sess:
sess.run(init)
step = 1
# 持续迭代
while step * batch_size < training_iters:
# 随机抽出这一次迭代训练时用的数据
batch_xs, batch_ys = mnist.train.next_batch(batch_size)
# 对数据进行处理,使得其符合输入
batch_xs = batch_xs.reshape((batch_size, n_steps, n_input))
# 迭代
sess.run(optimizer, feed_dict={x: batch_xs, y: batch_ys,
istate: np.zeros((batch_size, 2 * n_hidden))})
# 在特定的迭代回合进行数据的输出
if step % display_step == 0:
# Calculate batch accuracy
acc = sess.run(accuracy, feed_dict={x: batch_xs, y: batch_ys,
istate: np.zeros((batch_size, 2 * n_hidden))})
# Calculate batch loss
loss = sess.run(cost, feed_dict={x: batch_xs, y: batch_ys,
istate: np.zeros((batch_size, 2 * n_hidden))})
print "Iter " + str(step * batch_size) + ", Minibatch Loss= " + "{:.6f}".format(loss) + \
", Training Accuracy= " + "{:.5f}".format(acc)
step += 1
print "Optimization Finished!"
# 载入测试集进行测试
test_len = 256
test_data = mnist.test.images[:test_len].reshape((-1, n_steps, n_input))
test_label = mnist.test.labels[:test_len]
print "Testing Accuracy:", sess.run(accuracy, feed_dict={x: test_data, y: test_label,
istate: np.zeros((test_len, 2 * n_hidden))})
如果对你有帮助,请赞助哟,随意打赏,一分也是爱,哈哈。