# coding: utf-8
"""
 Convolutional Neural Network (CNN) with MNIST
"""
import numpy as np
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data

# Training Data
mnist = input_data.read_data_sets('data/', one_hot=True)
trainimg   = mnist.train.images
trainlabel = mnist.train.labels
testimg    = mnist.test.images
testlabel  = mnist.test.labels

# Define convolutional neural network architecture 

# Parameters
learning_rate   = 0.001
training_epochs = 5
batch_size      = 100
display_step    = 1

# Network
n_input  = 784
n_output = 10
weights  = {
    # 3x3 filters of 1 depth(black and white) and 64 of them.
    'wc1': tf.Variable(tf.random_normal([3, 3, 1, 64], stddev=0.1)),
    'wc2': tf.Variable(tf.random_normal([3, 3, 64, 128], stddev=0.1)),
    'wd1': tf.Variable(tf.random_normal([7*7*128, 1024], stddev=0.1)),
    'wd2': tf.Variable(tf.random_normal([1024, n_output], stddev=0.1))
}
biases   = {
    'bc1': tf.Variable(tf.random_normal([64], stddev=0.1)),
    'bc2': tf.Variable(tf.random_normal([128], stddev=0.1)),
    'bd1': tf.Variable(tf.random_normal([1024], stddev=0.1)),
    'bd2': tf.Variable(tf.random_normal([n_output], stddev=0.1))
}

def conv_basic(_input, _w, _b, _keepratio):
    """Defines the conv+pool + fully connected layers 
    arg:                                                                                    
      _input: the input 
      _w: the filter size or weight
      _b: bias
      _keepratio: dropout probability
    return: 
      out: a dictionary containing the output of each of the layers.
      
    """   
    # Input [batch,h,w,channel]
    _input_r = tf.reshape(_input, shape=[-1, 28, 28, 1])
    # this is a useful node that allows you to see how variables are being changed
    # _input_r = tf.Print(_input_r, [tf.shape(_input_r)], message = "This is _input_r: ", summarize=4)
    # input_shape = tf.shape(_input_r)
    # Conv1
    # SAME padding rounds up, meaning the partial windows are included.
    _conv1 = tf.nn.relu(tf.nn.bias_add(
            tf.nn.conv2d(_input_r, _w['wc1'], strides=[1, 1, 1, 1], padding='SAME')
            , _b['bc1']))
    # Pool1 using a kernel of 2x2x1 where the F = 2 for width and height, and depth = input_depth 1,
    # in this case. S = 2 is chosen. A -1x14x14x1 will be outputted.
    _pool1 = tf.nn.max_pool(_conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
    _pool_dr1 = tf.nn.dropout(_pool1, _keepratio)
    # _pool_dr1 shape is [100,14,14,64]
    input_shape = tf.shape(_pool_dr1)
    # Conv2
    _conv2 = tf.nn.relu(tf.nn.bias_add(
            tf.nn.conv2d(_pool_dr1, _w['wc2'], strides=[1, 1, 1, 1], padding='SAME')
            , _b['bc2']))
    _pool2 = tf.nn.max_pool(_conv2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
    _pool_dr2 = tf.nn.dropout(_pool2, _keepratio)
    
    # Vectorize to match the FC weight's dim
    _dense1 = tf.reshape(_pool_dr2, [-1, _w['wd1'].get_shape().as_list()[0]])
    
    # Fc1
    _fc1 = tf.nn.relu(tf.add(tf.matmul(_dense1, _w['wd1']), _b['bd1']))
    _fc_dr1 = tf.nn.dropout(_fc1, _keepratio)
    
    # Fc2
    _out = tf.add(tf.matmul(_fc_dr1, _w['wd2']), _b['bd2'])
    
    # Return everything
    out = {
        'input_r': _input_r,
        'input_shape': input_shape,
        'conv1': _conv1,
        'pool1': _pool1,
        'pool1_dr1': _pool_dr1,
        'conv2': _conv2,
        'pool2': _pool2,
        'pool_dr2': _pool_dr2,
        'dense1': _dense1,
        'fc1': _fc1,
        'fc_dr1': _fc_dr1,
        'out': _out
    }
    return out


# tf Graph input
x = tf.placeholder(tf.float32, [None, n_input])
y = tf.placeholder(tf.float32, [None, n_output])
keepratio = tf.placeholder(tf.float32)

# Functions! 
conv_dict = conv_basic(x, weights, biases, keepratio)
_pred = conv_dict['out']
re_input = conv_dict['input_shape']
# _pred contains the 'unscaled' log probabilities. or jsut output of last layer.
# compute the mean of the prediction cost on the whole batch.
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(_pred, y))
optm = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)
_corr = tf.equal(tf.argmax(_pred,1), tf.argmax(y,1)) # Count corrects
accr = tf.reduce_mean(tf.cast(_corr, tf.float32)) # Accuracy
init = tf.initialize_all_variables()

# Saver 
save_step = 1;
saver = tf.train.Saver(max_to_keep=training_epochs) 

print ("Network Ready to Go!")

do_train = 1

sess = tf.Session()
sess.run(init)

if do_train == 1:
    for epoch in range(training_epochs):
        avg_cost = 0.
        total_batch = int(mnist.train.num_examples/batch_size)
        # Loop over all batches
        for i in range(total_batch):
            batch_xs, batch_ys = mnist.train.next_batch(batch_size)
            # Feed training using batch data
            sess.run(optm, feed_dict={x: batch_xs, y: batch_ys, keepratio:0.7})
            # Compute average loss
            avg_cost += sess.run(cost, feed_dict={x: batch_xs, y: batch_ys, keepratio:1.})/total_batch
            print(sess.run(re_input, feed_dict={x: batch_xs, keepratio:1.}))
            print("loop: ", i)

        # Display logs per epoch step
        if epoch % display_step == 0: 
            print ("Epoch: %03d/%03d cost: %.9f" % (epoch, training_epochs, avg_cost))
            train_acc = sess.run(accr, feed_dict={x: batch_xs, y: batch_ys, keepratio:1.})
            print (" Training accuracy: %.3f" % (train_acc))
            test_acc = sess.run(accr, feed_dict={x: testimg, y: testlabel, keepratio:1.})
            print (" Test accuracy: %.3f" % (test_acc))

        # Save Net
        if epoch % save_step == 0:
            saver.save(sess, "nets/cnn_basic.ckpt-" + str(epoch))

    print ("Optimization Finished.")
        


# In[6]:

# Restore net
if do_train == 0:
    epoch = 4
    saver.restore(sess, "nets/cnn_basic.ckpt-" + str(epoch))


# In[7]:

# Compute test accuracy
test_acc = sess.run(accr, feed_dict={x: testimg, y: testlabel, keepratio:1.})
print (" Test accuracy: %.3f" % (test_acc))