coding=utf-8

    from __future__ import absolute_import
    from __future__ import division
    from __future__ import print_function

    import gzip
    import os
    import sys
    import time

    import numpy as np
    from six.moves import urllib
    from six.moves import xrange  
    import tensorflow as tf
    from tensorflow.python.ops import rnn, rnn_cell
    import matplotlib.pyplot as plt
    from tensorflow.examples.tutorials.mnist import input_data
    # 载入数据
    mnist = input_data.read_data_sets('MNIST_data', one_hot=True)

    # 把每张图片看作一个序列，28个元素，每个元素28个点

    # Parameters
    learning_rate = 0.001
    training_iters = 100000
    batch_size = 128
    display_step = 10

    # 每行看成上一行的序列
    n_input = 28 # MNIST data input (img shape: 28*28)
    n_steps = 28 # timesteps
    # n_hidden个lstm单元
    n_hidden = 128 
    # 最终分类到0～9数字上
    n_classes = 10 # MNIST total classes (0-9 digits)

    # tf Graph input
    x = tf.placeholder("float", [None, n_steps, n_input])
    y = tf.placeholder("float", [None, n_classes])

    # 从n_hidden个lstm到n_classes个最后一层神经元的权重
    weights = {
        'out': tf.Variable(tf.random_normal([n_hidden, n_classes]))
    }
    biases = {
        'out': tf.Variable(tf.random_normal([n_classes]))
    }

    def RNN(x, weights, biases):

        # 为rnn作数据转换
        # 输入数据是: (batch_size, n_steps, n_input)
        # 需要转换为: 'n_steps' 个矩阵 (batch_size, n_input)
        
        # 把n_steps和batch_size交换下，矩阵变成[n_steps,batch_size,n_input]
        x = tf.transpose(x, [1, 0, 2])
        # 再变成 (n_steps*batch_size, n_input)
        x = tf.reshape(x, [-1, n_input])
        # 拆成n_steps个 (batch_size, n_input)
        x = tf.split(0, n_steps, x)

        # n_hidden个lstm单元
        lstm_cell = rnn_cell.BasicLSTMCell(n_hidden, forget_bias=1.0)

        # 计算
        outputs, states = rnn.rnn(lstm_cell, x, dtype=tf.float32)

        # 取最后一个output，rnn后再走一层网络
        return tf.matmul(outputs[-1], weights['out']) + biases['out']

    pred = RNN(x, weights, biases)

    # loss函数，优化方法
    cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(pred, y))
    optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)

    # 准确率评估
    correct_pred = tf.equal(tf.argmax(pred,1), tf.argmax(y,1))
    accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

    init = tf.initialize_all_variables()

    with tf.Session() as sess:
        sess.run(init)
        step = 1
        while step * batch_size < training_iters:
            batch_x, batch_y = mnist.train.next_batch(batch_size)
            # Reshape data to get 28 seq of 28 elements
            batch_x = batch_x.reshape((batch_size, n_steps, n_input))
            # Run optimization op (backprop)
            sess.run(optimizer, feed_dict={x: batch_x, y: batch_y})
            if step % display_step == 0:
                # 计算准确率和loss
                acc = sess.run(accuracy, feed_dict={x: batch_x, y: batch_y})
                loss = sess.run(cost, feed_dict={x: batch_x, y: batch_y})
                print ("Iter" + str(step*batch_size) + ", Minibatch Loss= " + \
                      "{:.6f}".format(loss) + ", Training Accuracy= " + \
                      "{:.5f}".format(acc))
            step += 1
        print ("Optimization Finished!")

        # 用测试数据再评估下
        test_len = 128
        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}))