mnist_tfrecord.py

'''MNIST dataset with TFRecords, the standard TensorFlow data format.

TFRecord is a data format supported throughout TensorFlow.
This example demonstrates how to load TFRecord data using
Input Tensors. Input Tensors differ from the normal Keras
workflow because instead of fitting to data loaded into a
a numpy array, data is supplied via a special tensor that
reads data from nodes that are wired directly into model
graph with the `Input(tensor=input_tensor)` parameter.

There are several advantages to using Input Tensors.
First, if a dataset is already in TFRecord format you
can load and train on that data directly in Keras.
Second, extended backend API capabilities such as TensorFlow
data augmentation is easy to integrate directly into your
Keras training scripts via input tensors.
Third, TensorFlow implements several data APIs for
TFRecords, some of which provide significantly faster
training performance than numpy arrays can provide because
they run via the C++ backend. Please note that this
example is tailored for brevity and clarity and not
to demonstrate performance or augmentation capabilities.

Input Tensors also have important disadvantages. In
particular, Input Tensors are fixed at model construction
because rewiring networks is not yet supported.
For this reason, changing the data input source means
model weights must be saved and the model rebuilt
from scratch to connect the new input data.
validation cannot currently be performed as training
progresses, and must be performed after training completes.
This example demonstrates how to train with input
tensors, save the model weights, and then evaluate the
model using the numpy based Keras API.

Gets to ~99.1% test accuracy after 5 epochs
(high variance from run to run: 98.9-99.3).
'''
import numpy as np
import os
import tensorflow as tf
import keras
from keras import backend as K
from keras import layers
from keras.callbacks import Callback

from tensorflow.contrib.learn.python.learn.datasets import mnist

if K.backend() != 'tensorflow':
    raise RuntimeError('This example can only run with the '
                       'TensorFlow backend, '
                       'because it requires TFRecords, which '
                       'are not supported on other platforms.')


class EvaluateInputTensor(Callback):
    """ Validate a model which does not expect external numpy data during training.

    Keras does not expect external numpy data at training time, and thus cannot
    accept numpy arrays for validation when all of a Keras Model's
    `Input(input_tensor)` layers are provided an  `input_tensor` parameter,
    and the call to `Model.compile(target_tensors)` defines all `target_tensors`.
    Instead, create a second model for validation which is also configured
    with input tensors and add it to the `EvaluateInputTensor` callback
    to perform validation.

    It is recommended that this callback be the first in the list of callbacks
    because it defines the validation variables required by many other callbacks,
    and Callbacks are made in order.

    # Arguments
        model: Keras model on which to call model.evaluate().
        steps: Integer or `None`.
            Total number of steps (batches of samples)
            before declaring the evaluation round finished.
            Ignored with the default value of `None`.
    """

    def __init__(self, model, steps, metrics_prefix='val', verbose=1):
        # parameter of callbacks passed during initialization
        # pass evalation mode directly
        super(EvaluateInputTensor, self).__init__()
        self.val_model = model
        self.num_steps = steps
        self.verbose = verbose
        self.metrics_prefix = metrics_prefix

    def on_epoch_end(self, epoch, logs={}):
        self.val_model.set_weights(self.model.get_weights())
        results = self.val_model.evaluate(None, None, steps=int(self.num_steps),
                                          verbose=self.verbose)
        metrics_str = '\n'
        for result, name in zip(results, self.val_model.metrics_names):
            metric_name = self.metrics_prefix + '_' + name
            logs[metric_name] = result
            if self.verbose > 0:
                metrics_str = metrics_str + metric_name + ': ' + str(result) + ' '

        if self.verbose > 0:
            print(metrics_str)


def cnn_layers(x_train_input):
    x = layers.Conv2D(32, (3, 3),
                      activation='relu', padding='valid')(x_train_input)
    x = layers.MaxPooling2D(pool_size=(2, 2))(x)
    x = layers.Conv2D(64, (3, 3), activation='relu')(x)
    x = layers.MaxPooling2D(pool_size=(2, 2))(x)
    x = layers.Flatten()(x)
    x = layers.Dense(512, activation='relu')(x)
    x = layers.Dropout(0.5)(x)
    x_train_out = layers.Dense(num_classes,
                               activation='softmax',
                               name='x_train_out')(x)
    return x_train_out

sess = K.get_session()

batch_size = 100
batch_shape = (batch_size, 28, 28, 1)
epochs = 5
num_classes = 10

# The capacity variable controls the maximum queue size
# allowed when prefetching data for training.
capacity = 10000

# min_after_dequeue is the minimum number elements in the queue
# after a dequeue, which ensures sufficient mixing of elements.
min_after_dequeue = 3000

# If `enqueue_many` is `False`, `tensors` is assumed to represent a
# single example.  An input tensor with shape `[x, y, z]` will be output
# as a tensor with shape `[batch_size, x, y, z]`.
#
# If `enqueue_many` is `True`, `tensors` is assumed to represent a
# batch of examples, where the first dimension is indexed by example,
# and all members of `tensors` should have the same size in the
# first dimension.  If an input tensor has shape `[*, x, y, z]`, the
# output will have shape `[batch_size, x, y, z]`.
enqueue_many = True

cache_dir = os.path.expanduser(
    os.path.join('~', '.keras', 'datasets', 'MNIST-data'))
data = mnist.read_data_sets(cache_dir, validation_size=0)

x_train_batch, y_train_batch = tf.train.shuffle_batch(
    tensors=[data.train.images, data.train.labels.astype(np.int32)],
    batch_size=batch_size,
    capacity=capacity,
    min_after_dequeue=min_after_dequeue,
    enqueue_many=enqueue_many,
    num_threads=8)

x_train_batch = tf.cast(x_train_batch, tf.float32)
x_train_batch = tf.reshape(x_train_batch, shape=batch_shape)

y_train_batch = tf.cast(y_train_batch, tf.int32)
y_train_batch = tf.one_hot(y_train_batch, num_classes)

x_batch_shape = x_train_batch.get_shape().as_list()
y_batch_shape = y_train_batch.get_shape().as_list()

model_input = layers.Input(tensor=x_train_batch)
model_output = cnn_layers(model_input)
train_model = keras.models.Model(inputs=model_input, outputs=model_output)

# Pass the target tensor `y_train_batch` to `compile`
# via the `target_tensors` keyword argument:
train_model.compile(optimizer=keras.optimizers.RMSprop(lr=2e-3, decay=1e-5),
                    loss='categorical_crossentropy',
                    metrics=['accuracy'],
                    target_tensors=[y_train_batch])
train_model.summary()

x_test_batch, y_test_batch = tf.train.batch(
    tensors=[data.test.images, data.test.labels.astype(np.int32)],
    batch_size=batch_size,
    capacity=capacity,
    enqueue_many=enqueue_many,
    num_threads=8)

# Create a separate test model
# to perform validation during training
x_test_batch = tf.cast(x_test_batch, tf.float32)
x_test_batch = tf.reshape(x_test_batch, shape=batch_shape)

y_test_batch = tf.cast(y_test_batch, tf.int32)
y_test_batch = tf.one_hot(y_test_batch, num_classes)

x_test_batch_shape = x_test_batch.get_shape().as_list()
y_test_batch_shape = y_test_batch.get_shape().as_list()

test_model_input = layers.Input(tensor=x_test_batch)
test_model_output = cnn_layers(test_model_input)
test_model = keras.models.Model(inputs=test_model_input, outputs=test_model_output)

# Pass the target tensor `y_test_batch` to `compile`
# via the `target_tensors` keyword argument:
test_model.compile(optimizer=keras.optimizers.RMSprop(lr=2e-3, decay=1e-5),
                   loss='categorical_crossentropy',
                   metrics=['accuracy'],
                   target_tensors=[y_test_batch])

# Fit the model using data from the TFRecord data tensors.
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess, coord)

train_model.fit(epochs=epochs,
                steps_per_epoch=int(np.ceil(data.train.num_examples / float(batch_size))),
                callbacks=[EvaluateInputTensor(test_model, steps=100)])

# Save the model weights.
train_model.save_weights('saved_wt.h5')

# Clean up the TF session.
coord.request_stop()
coord.join(threads)
K.clear_session()

# Second Session to test loading trained model without tensors
x_test = np.reshape(data.test.images, (data.test.images.shape[0], 28, 28, 1))
y_test = data.test.labels
x_test_inp = layers.Input(shape=(x_test.shape[1:]))
test_out = cnn_layers(x_test_inp)
test_model = keras.models.Model(inputs=x_test_inp, outputs=test_out)

test_model.load_weights('saved_wt.h5')
test_model.compile(optimizer='rmsprop',
                   loss='categorical_crossentropy',
                   metrics=['accuracy'])
test_model.summary()

loss, acc = test_model.evaluate(x_test,
                                keras.utils.to_categorical(y_test),
                                batch_size=batch_size)
print('\nTest accuracy: {0}'.format(acc))