fork download
copy 
  1. from __future__ import absolute_import
  2. from __future__ import division
  3. from __future__ import print_function
  4. # inference
  5. import argparse
  6. import sys
  7. import tempfile
  8.  
  9. #from tensorflow.examples.tutorials.mnist import input_data
  10.  
  11. import tensorflow as tf
  12.  
  13. FLAGS = None
  14.  
  15.  
  16. def deepnn(x):
  17.   """deepnn builds the graph for a deep net for classifying digits.
  18.   Args:
  19.     x: an input tensor with the dimensions (N_examples, 784), where 784 is the
  20.     number of pixels in a standard MNIST image.
  21.   Returns:
  22.     A tuple (y, keep_prob). y is a tensor of shape (N_examples, 10), with values
  23.     equal to the logits of classifying the digit into one of 10 classes (the
  24.     digits 0-9). keep_prob is a scalar placeholder for the probability of
  25.     dropout.
  26.   """
  27.   # Reshape to use within a convolutional neural net.
  28.   # Last dimension is for "features" - there is only one here, since images are
  29.   # grayscale -- it would be 3 for an RGB image, 4 for RGBA, etc.
  30.   with tf.name_scope('reshape'):
  31.     x_image = tf.reshape(x, [-1, 28, 28, 1])
  32.  
  33.   # First convolutional layer - maps one grayscale image to 32 feature maps.
  34.   with tf.name_scope('conv1'):
  35.     W_conv1 = weight_variable([5, 5, 1, 32])
  36.     b_conv1 = bias_variable([32])
  37.     h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
  38.  
  39.   # Pooling layer - downsamples by 2X.
  40.   with tf.name_scope('pool1'):
  41.     h_pool1 = max_pool_2x2(h_conv1)
  42.  
  43.   # Second convolutional layer -- maps 32 feature maps to 64.
  44.   with tf.name_scope('conv2'):
  45.     W_conv2 = weight_variable([5, 5, 32, 64])
  46.     b_conv2 = bias_variable([64])
  47.     h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)
  48.  
  49.   # Second pooling layer.
  50.   with tf.name_scope('pool2'):
  51.     h_pool2 = max_pool_2x2(h_conv2)
  52.  
  53.   # Fully connected layer 1 -- after 2 round of downsampling, our 28x28 image
  54.   # is down to 7x7x64 feature maps -- maps this to 1024 features.
  55.   with tf.name_scope('fc1'):
  56.     W_fc1 = weight_variable([7 * 7 * 64, 1024])
  57.     b_fc1 = bias_variable([1024])
  58.  
  59.     h_pool2_flat = tf.reshape(h_pool2, [-1, 7*7*64])
  60.     h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1) + b_fc1)
  61.  
  62.   #-
  63.   # Dropout - controls the complexity of the model, prevents co-adaptation of
  64.   # features.
  65.   #with tf.name_scope('dropout'):
  66.   #  keep_prob = tf.placeholder(tf.float32)
  67.   #  h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)
  68.  
  69.   # Map the 1024 features to 10 classes, one for each digit
  70.   with tf.name_scope('fc2'):
  71.     W_fc2 = weight_variable([1024, 10])
  72.     b_fc2 = bias_variable([10])
  73.  
  74.   #  y_conv = tf.matmul(h_fc1_drop, W_fc2) + b_fc2
  75.   #return y_conv, keep_prob
  76.     y_conv = tf.matmul(h_fc1, W_fc2) + b_fc2
  77.   return y_conv
  78.  
  79.  
  80. def conv2d(x, W):
  81.   """conv2d returns a 2d convolution layer with full stride."""
  82.   return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')
  83.  
  84.  
  85. def max_pool_2x2(x):
  86.   """max_pool_2x2 downsamples a feature map by 2X."""
  87.   return tf.nn.max_pool(x, ksize=[1, 2, 2, 1],
  88.                         strides=[1, 2, 2, 1], padding='SAME')
  89.  
  90.  
  91. def weight_variable(shape):
  92.   """weight_variable generates a weight variable of a given shape."""
  93.   initial = tf.truncated_normal(shape, stddev=0.1)
  94.   return tf.Variable(initial)
  95.  
  96.  
  97. def bias_variable(shape):
  98.   """bias_variable generates a bias variable of a given shape."""
  99.   initial = tf.constant(0.1, shape=shape)
  100.   return tf.Variable(initial)
  101.  
  102.  
  103. def main(_):
  104.   #-
  105.   # Import data
  106.   #mnist = input_data.read_data_sets(FLAGS.data_dir, one_hot=True)
  107.  
  108.   # Create the model
  109.   x = tf.placeholder(tf.float32, [None, 784],name="input")
  110.  
  111.   #-
  112.   # Define loss and optimizer
  113.   #y_ = tf.placeholder(tf.float32, [None, 10])
  114.  
  115.   #-
  116.   # Build the graph for the deep net
  117.   #y_conv, keep_prob = deepnn(x)
  118.   # No longer need keep_prob since removing dropout layers.
  119.   y_conv = deepnn(x)
  120.   output = tf.nn.softmax(y_conv, name='output')
  121. #  with tf.name_scope('loss'):
  122.  
  123. #  train_writer.add_graph(tf.get_default_graph())
  124.   saver = tf.train.Saver(tf.global_variables())
  125.  #saver = tf.train.Saver()
  126.  
  127.   with tf.Session() as sess:
  128.     sess.run(tf.global_variables_initializer())
  129.     sess.run(tf.local_variables_initializer())
  130.        # read the previously saved network.
  131.     saver.restore(sess, '.' + '/mnist_model')
  132.        # save the version of the network ready that can be compiled for NCS
  133.     saver.save(sess, '.' + '/mnist_inference')
  134.     #for i in range(3000):
  135.  
  136.     #save_path = saver.save(sess, graph_location + "/mnist_model")
  137.  
  138. if __name__ == '__main__':
  139.   parser = argparse.ArgumentParser()
  140.   parser.add_argument('--data_dir', type=str,
  141.                       default='/tmp/tensorflow/mnist/input_data',
  142.                       help='Directory for storing input data')
  143.   FLAGS, unparsed = parser.parse_known_args()
  144. tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)
Runtime error #stdin #stdout #stderr 0.03s 119808KB
comments (?)
stdin
copy 
Standard input is empty
stdout
copy 
Standard output is empty
stderr
copy 
Traceback (most recent call last):
  File "prog.py", line 11, in <module>
    import tensorflow as tf
ImportError: No module named tensorflow
https://ideone.com/RSAE2H
language:
Python (PyPy 2.7.13)
created:
5 years ago
visibility:
public

Share or Embed source code

Discover > Sphere Engine API

The brand new service which powers Ideone!

Discover > IDE Widget

Widget for compiling and running the source code in a web browser!