import tensorflow as tf 
from tensorflow.examples.tutorials.mnist import input_data
 
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

tf.set_random_seed(1)
#导入数据

mnist = input_data.read_data_sets('E:/Program Files/Machine Learning/node/MNIST_data',one_hot = True)
#hyperparameters
lr = 0.001  #学习率
training_iters = 100000
batch_size = 128

n_inputs = 28    
n_steps = 28
n_hidden_units = 128
n_classes = 10 

#定义x,y的placeholder和weights,biases的初始状况
x = tf.placeholder(tf.float32,[None,n_steps,n_inputs])
y = tf.placeholder(tf.float32,[None,n_classes])

#对weights biases初始化的定义
weight = {
    #shape (28,128)
    'in':tf.Variable(tf.random_normal([n_inputs,n_hidden_units])),
    #shape (128,10)
    'out':tf.Variable(tf.random_normal([n_hidden_units,n_classes]))

}
biases = {
    #shape (128,)
    'in':tf.Variable(tf.constant(0.1,shape=[n_hidden_units,])),
    #shape (128,10)
    'out':tf.Variable(tf.constant(0.1,shape=[n_classes,]))

}



#定义RNN的主体结构
def RNN(X,weight,biases):
    #原始的X是3维数据，我们需要把它变成2维数据才能使用weight的矩阵乘法
    #X ==> (128batchs*28steps,28 inputs)
    X = tf.reshape(X,[-1,n_inputs])
    #X_in = W * X +b
    X_in = tf.matmul(X,weight['in']) + biases['in']
    #X_in==> (128batchs,28steps,28 inputs)换回3维
    X_in = tf.reshape(X_in,[-1,n_steps,n_hidden_units])
    #cell
    #使用basic LSTM Cell
    lstm_cell = tf.contrib.rnn.BasicLSTMCell(n_hidden_units,forget_bias=1.0,state_is_tuple=True)
    init_state = lstm_cell.zero_state(batch_size,dtype=tf.float32)#初始化全0 state
    outputs,final_state = tf.nn.dynamic_rnn(lstm_cell,X_in,initial_state=init_state,time_major=False)

    #
    results = tf.matmul(final_state[1],weight['out'])+biases['out']
    return results

pred = RNN(x,weight,biases)
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels =y ,logits=pred))
train_op = tf.train.AdamOptimizer(lr).minimize(cost)

#训练RNN
correct_pred = tf.equal(tf.argmax(pred,1),tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correct_pred,tf.float32))

init = tf.global_variables_initializer()
with tf.Session() as sess:
    sess.run(init)
    step = 0
    while step*batch_size < training_iters:
        batch_xs,batch_ys = mnist.train.next_batch(batch_size)
        batch_xs= batch_xs.reshape([batch_size,n_steps,n_inputs])
        sess.run([train_op],feed_dict={x:batch_xs,y:batch_ys,})

    if step % 20 == 0:
        print(sess.run(accuracy,feed_dict={x:batch_xs,y:batch_ys,}))
        step+=1