ML-Tech-Test/inception_module.py at master · mariovp/ML-Tech-Test · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
import tensorflow as tf

from tensorflow.keras import layers
from tensorflow import keras

import time


class InceptionV1Module(layers.Layer):

    def __init__(self,
                 conv1x1_filters=64, conv3x3_reduce_filters=96, conv3x3_filters=128, conv5x5_reduce_filters=16, conv5x5_filters=32, pooling_conv_filters=32,
                 **kwargs):
        super(InceptionV1Module, self).__init__(**kwargs)

        self.conv1x1_filters = conv1x1_filters
        self.conv3x3_reduce_filters = conv3x3_reduce_filters
        self.conv3x3_filters = conv3x3_filters
        self.conv5x5_reduce_filters = conv5x5_reduce_filters
        self.conv5x5_filters = conv5x5_filters
        self.pooling_conv_filters = pooling_conv_filters

        self.conv_1x1 = layers.Conv2D(
            conv1x1_filters, (1, 1), padding='same', activation='relu')
        self.bn_1x1 = layers.BatchNormalization()

        self.conv_3x3_1 = layers.Conv2D(
            conv3x3_reduce_filters, (1, 1), padding='same', activation='relu')
        self.conv_3x3_2 = layers.Conv2D(
            conv3x3_filters, (3, 3), padding='same', activation='relu')
        self.bn_3x3_1 = layers.BatchNormalization()
        self.bn_3x3_2 = layers.BatchNormalization()

        self.conv_5x5_1 = layers.Conv2D(
            conv5x5_reduce_filters, (1, 1), padding='same', activation='relu')
        self.conv_5x5_2 = layers.Conv2D(
            conv5x5_filters, (5, 5), padding='same', activation='relu')

        self.bn_5x5_1 = layers.BatchNormalization()
        self.bn_5x5_2 = layers.BatchNormalization()

        self.pooling_1 = layers.MaxPooling2D(
            (3, 3), strides=(1, 1), padding='same')
        self.pooling_2 = layers.Conv2D(
            pooling_conv_filters, (1, 1), padding='same', activation='relu')


    def call(self, inputs):

        res_conv_1x1 = self.conv_1x1(inputs)
        res_conv_1x1 = self.bn_1x1(res_conv_1x1)

        res_conv_3x3 = self.conv_3x3_1(inputs)
        res_conv_3x3 = self.bn_3x3_1(res_conv_3x3)
        res_conv_3x3 = self.conv_3x3_2(res_conv_3x3)
        res_conv_3x3 = self.bn_3x3_2(res_conv_3x3)

        res_conv_5x5 = self.conv_5x5_1(inputs)
        res_conv_5x5 = self.bn_5x5_1(res_conv_5x5)
        res_conv_5x5 = self.conv_5x5_2(res_conv_5x5)
        res_conv_5x5 = self.bn_5x5_2(res_conv_5x5)

        res_pooling = self.pooling_1(inputs)
        res_pooling = self.pooling_2(res_pooling)

        result = layers.concatenate(
            [res_conv_1x1, res_conv_3x3, res_conv_5x5, res_pooling])

        return result

    def get_config(self):
        config = super().get_config().copy()
        config.update({
            'conv1x1_filters': self.conv1x1_filters,
            'conv3x3_reduce_filters': self.conv3x3_reduce_filters,
            'conv3x3_filters': self.conv3x3_filters,
            'conv5x5_reduce_filters': self.conv5x5_reduce_filters,
            'conv5x5_filters': self.conv5x5_filters,
            'pooling_conv_filters': self.pooling_conv_filters
        })
        return config

    def from_config(self, cls, config):
        # raise ValueError("From config")
        return cls(**config)


print("Num GPUs Available: ", len(
    tf.config.experimental.list_physical_devices('GPU')))

fashion_mnist = keras.datasets.cifar10

(train_images, train_labels), (test_images,
                               test_labels) = fashion_mnist.load_data()
train_images = train_images / 255.0
test_images = test_images / 255.0


class_names = ['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat',
               'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']

# strategy = tf.distribute.OneDeviceStrategy(device="/gpu:0")
strategy = tf.distribute.MirroredStrategy()
with strategy.scope():
    model = keras.Sequential([
        # keras.layers.Reshape(input_shape=(28, 28), target_shape=(28, 28, 1)),
        keras.layers.InputLayer(input_shape=(32, 32, 3)),
        InceptionV1Module(),
        keras.layers.MaxPool2D(3, 2, padding="same"),
        keras.layers.Dropout(0.3),
        InceptionV1Module(),
        keras.layers.MaxPool2D(3, 2, padding="same"),
        keras.layers.Dropout(0.4),
        InceptionV1Module(),
        keras.layers.Dropout(0.5),
        keras.layers.Flatten(),
        keras.layers.Dense(10, activation='softmax')
    ])
    # Hasta 85% en Cifar-10 en 85 epochs

    model.summary()

    optimizer = tf.keras.optimizers.RMSprop(learning_rate=0.01, decay=1e-6)

    model.compile(optimizer="adam",
                  loss='sparse_categorical_crossentropy', metrics=['accuracy'])

    """tf.keras.utils.plot_model(
        model,
        to_file='model.png',
        show_shapes=True,
        show_layer_names=True,
        rankdir='LR',
        expand_nested=True,
        dpi=96
    )"""

    start_time = time.time()
    model.fit(train_images, train_labels, epochs=250, batch_size=256, validation_data=(test_images, test_labels))
    elapsed_seconds = time.time() - start_time
    elapsed_time = time.strftime('%H:%M:%S', time.gmtime(elapsed_seconds))

    test_loss, test_acc = model.evaluate(test_images, test_labels, verbose=0)
    print("Test accuracy: ", test_acc)
    print("Training took", elapsed_time, "(hh:mm:ss)", elapsed_seconds, "(seconds)")