Невозможно сохранить и восстановить обученную модель TensorFlow

Я только что прочитал Deep MNIST для экспертов и изменил код mnist_deep.py для сохранения обученной модели с использованием saver = tf.train.Saver() перед созданием сеанса и saver.save(sess, './mnist_deep_model', global_step=2000) после цикла for для обучения модели. Кажется, он правильно сохранен, поскольку в моей рабочей папке есть четыре следующих файла:

• контрольно-пропускной пункт
• mnist_deep_model-2000.data-00000-of-00001
• mnist_deep_model-2000.indexs
• mnist_deep_model-2000.meta

Я также изменил mnist_deep.py, добавив две следующие функции, чтобы иметь возможность тестировать модель на отдельных тестовых изображениях, одно за другим:

def indexMax(list):
  """indexMax returns the index of the max element of the list."""
  return list.index(max(list))


def identifyDigitInImage(sess, x, y_conv, keep_prob, image):
  """identifyDigitInImage apply the trained model to given image to identify the represented digit."""
  result = sess.run(y_conv, {x:[image], keep_prob: 1.0})[0].tolist()
  return indexMax(result)

Я также добавил в конце основной функции цикл, в котором я случайным образом выбираю одно тестовое изображение в тестовом наборе и пытаюсь применить обученную модель к каждому из них с помощью этой функции. Кажется, это работает, так как я получаю ту же точность в этом тестовом цикле: 99,2%

Затем я написал вторую программу: mnist_deep_restore_trained_model.py (также основанную на исходном коде mnist_deep.py), пытаясь восстановить ранее сохраненную обученную модель и применить к ней тестовые изображения, надеясь получить такую ​​же точность.

Конечно, я удалил из этой программы весь код, необходимый для создания, обучения и тестирования модели (функция deepnn() и все связанные с ней функции, создание тензоров: x = tf.placeholder(tf.float32, [None, 784]), y_conv, keep_prob = deepnn(x), loss, optimizer и прочее о точности ...) и Я просто восстановил сохраненную модель таким образом: (после открытия сеанса)

saver = tf.train.import_meta_graph('./mnist_deep_model-2000.meta')
saver.restore(sess, tf.train.latest_checkpoint('./'))

Я также удалил инициализацию глобальных переменных в начале сеанса, поскольку значения глобальных переменных должны были быть восстановлены из обученной модели:

Но, чтобы иметь возможность применить модель для определения цифры данного тестового изображения (см. function identifyDigitInImage(sess, x, y_conv, keep_prob, image)), мне все еще нужно получить тензорные переменные x, y_conv и keep_prob. Поэтому после восстановления модели с диска я добавил следующие строки:

graph = tf.get_default_graph()
x = graph.get_tensor_by_name("x:0")
keep_prob = graph.get_tensor_by_name("keep_prob:0")
y_conv = graph.get_tensor_by_name("y_conv:0")

Наконец, я также добавил в конце этой второй программы тот же тестовый цикл, что и в mnist_deep.py, ожидая получить те же результаты от этой восстановленной модели ...

К сожалению, я получаю исключение при первом вызове get_tensor_by_name ():

x = graph.get_tensor_by_name("x:0")
KeyError: "The name 'x:0' refers to a Tensor which does not exist. The operation 'x', does not exist in the graph."

Остальные get_tensor_by_name() вызовы также вызовут это же исключение.

Что я делаю неправильно ? Почему нельзя получить эти тензоры таким образом?

Вот мой полный исходный код mnist_deep.py:

# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================

"""A deep MNIST classifier using convolutional layers.
See extensive documentation at
https://www.tensorflow.org/get_started/mnist/pros
"""
# Disable linter warnings to maintain consistency with tutorial.
# pylint: disable=invalid-name
# pylint: disable=g-bad-import-order

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

import argparse
import sys
import tempfile
import random

from tensorflow.examples.tutorials.mnist import input_data

import tensorflow as tf

FLAGS = None


def deepnn(x):
  """deepnn builds the graph for a deep net for classifying digits.
  Args:
    x: an input tensor with the dimensions (N_examples, 784), where 784 is the
    number of pixels in a standard MNIST image.
  Returns:
    A tuple (y, keep_prob). y is a tensor of shape (N_examples, 10), with values
    equal to the logits of classifying the digit into one of 10 classes (the
    digits 0-9). keep_prob is a scalar placeholder for the probability of
    dropout.
  """
  # Reshape to use within a convolutional neural net.
  # Last dimension is for "features" - there is only one here, since images are
  # grayscale -- it would be 3 for an RGB image, 4 for RGBA, etc.
  with tf.name_scope('reshape'):
    x_image = tf.reshape(x, [-1, 28, 28, 1])

  # First convolutional layer - maps one grayscale image to 32 feature maps.
  with tf.name_scope('conv1'):
    W_conv1 = weight_variable([5, 5, 1, 32])
    b_conv1 = bias_variable([32])
    h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)

  # Pooling layer - downsamples by 2X.
  with tf.name_scope('pool1'):
    h_pool1 = max_pool_2x2(h_conv1)

  # Second convolutional layer -- maps 32 feature maps to 64.
  with tf.name_scope('conv2'):
    W_conv2 = weight_variable([5, 5, 32, 64])
    b_conv2 = bias_variable([64])
    h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)

  # Second pooling layer.
  with tf.name_scope('pool2'):
    h_pool2 = max_pool_2x2(h_conv2)

  # Fully connected layer 1 -- after 2 round of downsampling, our 28x28 image
  # is down to 7x7x64 feature maps -- maps this to 1024 features.
  with tf.name_scope('fc1'):
    W_fc1 = weight_variable([7 * 7 * 64, 1024])
    b_fc1 = bias_variable([1024])

    h_pool2_flat = tf.reshape(h_pool2, [-1, 7*7*64])
    h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1) + b_fc1)

  # Dropout - controls the complexity of the model, prevents co-adaptation of
  # features.
  with tf.name_scope('dropout'):
    keep_prob = tf.placeholder(tf.float32)
    h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)

  # Map the 1024 features to 10 classes, one for each digit
  with tf.name_scope('fc2'):
    W_fc2 = weight_variable([1024, 10])
    b_fc2 = bias_variable([10])

    y_conv = tf.matmul(h_fc1_drop, W_fc2) + b_fc2
  return y_conv, keep_prob


def conv2d(x, W):
  """conv2d returns a 2d convolution layer with full stride."""
  return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')


def max_pool_2x2(x):
  """max_pool_2x2 downsamples a feature map by 2X."""
  return tf.nn.max_pool(x, ksize=[1, 2, 2, 1],
                        strides=[1, 2, 2, 1], padding='SAME')


def weight_variable(shape):
  """weight_variable generates a weight variable of a given shape."""
  initial = tf.truncated_normal(shape, stddev=0.1)
  return tf.Variable(initial)


def bias_variable(shape):
  """bias_variable generates a bias variable of a given shape."""
  initial = tf.constant(0.1, shape=shape)
  return tf.Variable(initial)


def indexMax(list):
  """indexMax returns the index of the max element of the list."""
  return list.index(max(list))


def identifyDigitInImage(sess, x, y_conv, keep_prob, image):
  """identifyDigitInImage apply the trained model to given image to identify the represented digit."""
  result = sess.run(y_conv, {x:[image], keep_prob: 1.0})[0].tolist()
  return indexMax(result)


def main(_):
  # Import data
  mnist = input_data.read_data_sets(FLAGS.data_dir, one_hot=True)

  # Create the model
  x = tf.placeholder(tf.float32, [None, 784])

  # Define loss and optimizer
  y_ = tf.placeholder(tf.float32, [None, 10])

  # Build the graph for the deep net
  y_conv, keep_prob = deepnn(x)

  with tf.name_scope('loss'):
    cross_entropy = tf.nn.softmax_cross_entropy_with_logits(labels=y_,
                                                            logits=y_conv)
  cross_entropy = tf.reduce_mean(cross_entropy)

  with tf.name_scope('adam_optimizer'):
    train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)

  with tf.name_scope('accuracy'):
    correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
    correct_prediction = tf.cast(correct_prediction, tf.float32)
  accuracy = tf.reduce_mean(correct_prediction)

  #graph_location = tempfile.mkdtemp()
  #print('Saving graph to: %s' % graph_location)
  #train_writer = tf.summary.FileWriter(graph_location)
  #train_writer.add_graph(tf.get_default_graph())

  # Prepare a saver to save the trained model:
  saver = tf.train.Saver()

  with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())

    # Save the untrained model:
    saver.save(sess, './mnist_deep_model')

    # Train the model:
    for i in range(20000):
      batch = mnist.train.next_batch(50)
      if i % 100 == 0:
        train_accuracy = accuracy.eval(feed_dict={
            x: batch[0], y_: batch[1], keep_prob: 1.0})
        print('step %d, training accuracy %g' % (i, train_accuracy))
      train_step.run(feed_dict={x: batch[0], y_: batch[1], keep_prob: 0.5})

    # Save the trained model:
    saver.save(sess, './mnist_deep_model', global_step=2000)

    # Display the test accuracy:
    print('test accuracy %g' % accuracy.eval(feed_dict={
        x: mnist.test.images, y_: mnist.test.labels, keep_prob: 1.0}))

    # Now try to apply the model to randomly choosen test images, one by one:
    stop = False
    count = 0
    ok_count = 0
    while not stop:
        # Choosing a test image index:
        test_image_index = random.randint(0, len(mnist.test.images) - 1)
        test_image = mnist.test.images[test_image_index]

        # Applying the trained model to identify the digit from the test image:
        identified_digit = identifyDigitInImage(sess, x, y_conv, keep_prob, test_image)

        # Display the identified digit:
        print("The written digit on the given image has been identified as a {}".format(identified_digit))

        # Check the expected_digit from the test label of the choosen test image:
        expected_digit = indexMax(mnist.test.labels[test_image_index].tolist())

        # Display the expected digit:
        print("Actually, the digit is a {}".format(expected_digit))

        # Count the correctly identified digits:
        if identified_digit == expected_digit:
            ok_count += 1

        # Stop the loop after 10000 iterations
        count += 1
        stop = count == 10000

        # Display the measured accuracy during the test loop:
    print("2nd Test accuracy = {}%".format(100 * (ok_count / count)))


if __name__ == '__main__':
  parser = argparse.ArgumentParser()
  parser.add_argument('--data_dir', type=str,
                      default='/tmp/tensorflow/mnist/input_data',
                      help='Directory for storing input data')
  FLAGS, unparsed = parser.parse_known_args()
tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)  

И вот мой полный исходный код mnist_deep_restore_trained_model.py:

# Copyright 2015 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# =============================================================================

"""A deep MNIST classifier using convolutional layers.
See extensive documentation at
https://www.tensorflow.org/get_started/mnist/pros
"""
# Disable linter warnings to maintain consistency with tutorial.
# pylint: disable=invalid-name
# pylint: disable=g-bad-import-order

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

import argparse
import sys
import random

from tensorflow.examples.tutorials.mnist import input_data

import tensorflow as tf

FLAGS = None

def indexMax(list):
  """indexMax returns the index of the max element of the list."""
  return list.index(max(list))


def identifyDigitInImage(sess, x, y_conv, keep_prob, image):
  """identifyDigitInImage apply the trained model to given image to identify the represented digit."""
  result = sess.run(y_conv, {x:[image], keep_prob: 1.0})[0].tolist()
  return indexMax(result)


def main(_):
  # Import data
  mnist = input_data.read_data_sets(FLAGS.data_dir, one_hot=True)

  with tf.Session() as sess:

    # Restoring the trained model previously saved:
    saver = tf.train.import_meta_graph('./mnist_deep_model-2000.meta')
    saver.restore(sess, tf.train.latest_checkpoint('./'))

    # Trying to get back some required tensors variables from the restored graph:
    graph = tf.get_default_graph()
    x = graph.get_tensor_by_name("x:0")
    # This call fails with the following exception:
    # KeyError: "The name 'x:0' refers to a Tensor which does not exist. The operation 'x', does not exist in the graph."
    keep_prob = graph.get_tensor_by_name("keep_prob:0")
    y_conv = graph.get_tensor_by_name("y_conv:0")

    # Now try to apply the model to randomly choosen test images, one by one:
    stop = False
    count = 0
    ok_count = 0
    while not stop:
      # Choosing a test image index:
      test_image_index = random.randint(0, len(mnist.test.images) - 1)
      test_image = mnist.test.images[test_image_index]

      # Applying the trained model to identify the digit from the test image:
      identified_digit = identifyDigitInImage(sess, x, y_conv, keep_prob, test_image)

      # Display the identified digit:
      print("The written digit on the given image has been identified as a {}".format(identified_digit))

      # Check the expected_digit from the test label of the choosen test image:
      expected_digit = indexMax(mnist.test.labels[test_image_index].tolist())

      # Display the expected digit:
      print("Actually, the digit is a {}".format(expected_digit))

      # Count the correctly identified digits:
      if identified_digit == expected_digit:
        ok_count += 1

      # Stop the loop after 10000 iterations
      count += 1
      stop = count == 10000

    # Display the measured accuracy during the test loop:
    print("Test accuracy = {}%".format(100 * (ok_count / count)))


if __name__ == '__main__':
  parser = argparse.ArgumentParser()
  parser.add_argument('--data_dir', type=str,
                      default='/tmp/tensorflow/mnist/input_data',
                      help='Directory for storing input data')
  FLAGS, unparsed = parser.parse_known_args()
tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)