-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathneural_structured_learning_model.py
234 lines (202 loc) · 9.84 KB
/
neural_structured_learning_model.py
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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
"""
This module uses adversarial regularization from NSL (neural structured learning)
to predict the breed (flat) of the animal (a multi-class classification problem from 37 different classes)
"""
import tensorflow as tf
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from tensorflow.keras.applications.inception_resnet_v2 import InceptionResNetV2
from tensorflow.keras.applications.inception_resnet_v2 import preprocess_input as preprocess_input_inception_resnet_v2
from keras.callbacks import EarlyStopping, ModelCheckpoint, ReduceLROnPlateau
import neural_structured_learning as nsl
import pandas as pd
import numpy as np
import sys
IMAGE_INPUT_NAME = 'input_1' # if experiencing problems with this change to value of 'model.layers[0].name'
LABEL_INPUT_NAME = 'label'
INPUT_SHAPE = [299, 299, 3]
model_name = sys.argv[1] if len(sys.argv) > 1 else ""
TRAIN_BATCH_SIZE = 16
multiplier, adv_step_size, adv_grad_norm = float(sys.argv[2]), float(sys.argv[3]), sys.argv[4]
print('\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n')
print('NEW RUN FOR NSL MODEL')
print(f'MODEL = {model_name}, BATCH_SIZE = {TRAIN_BATCH_SIZE}')
print(f'multiplier = {multiplier}, adv_step_size = {adv_step_size}, adv_grad_norm = {adv_grad_norm}')
def convert_to_dictionaries(image, label):
return {IMAGE_INPUT_NAME: image, LABEL_INPUT_NAME: label}
"""LOAD DATAFRAMES"""
df = pd.read_csv("data_advanced_model_linux.csv")
df['cat/dog'] = df['cat/dog'].astype(str)
df['breed'] = df['breed'].astype(str)
train_df = df[df['train/test'] == 'train'][['path', 'cat/dog', 'breed']]
val_df = df[df['train/test'] == 'validation'][['path', 'cat/dog', 'breed']]
test_df = df[df['train/test'] == 'test'][['path', 'cat/dog', 'breed']]
num_of_classes = len(set(train_df['breed']))
"""CREATE IMAGE GENERATORS"""
# TODO only good for inception_resnet_v2
pre_process = preprocess_input_inception_resnet_v2
train_data_gen = ImageDataGenerator(shear_range=0.2, zoom_range=0.2, horizontal_flip=True,
preprocessing_function=pre_process)
train_generator = train_data_gen.flow_from_dataframe(dataframe=train_df, x_col="path", y_col="breed",
class_mode="categorical", target_size=INPUT_SHAPE[:2],
batch_size=TRAIN_BATCH_SIZE)
val_data_gen = ImageDataGenerator(preprocessing_function=pre_process)
val_generator = val_data_gen.flow_from_dataframe(dataframe=val_df, x_col="path", y_col="breed",
class_mode="categorical", target_size=INPUT_SHAPE[:2],
batch_size=1, shuffle=False)
test_data_gen = ImageDataGenerator(preprocessing_function=pre_process)
test_generator = test_data_gen.flow_from_dataframe(dataframe=test_df, x_col="path", y_col="breed",
class_mode="categorical", target_size=INPUT_SHAPE[:2],
batch_size=1, shuffle=False)
"""PREPARE TENSORFLOW DATASETS FOR TRAIN, TEST"""
train_dataset = tf.data.Dataset.from_generator(
lambda: train_generator,
output_types=(tf.float32, tf.float32))
# convert the dataset to the desired format of NSL (dictionaries)
train_dataset = train_dataset.map(convert_to_dictionaries)
val_dataset = tf.data.Dataset.from_generator(
lambda: val_generator,
output_types=(tf.float32, tf.float32))
val_dataset = val_dataset.map(convert_to_dictionaries)
val_dataset = val_dataset.take(len(val_df))
# same for test data
test_dataset = tf.data.Dataset.from_generator(
lambda: test_generator,
output_types=(tf.float32, tf.float32))
test_dataset = test_dataset.map(convert_to_dictionaries)
# for test data we dont want to generate infinite data, we just want the amount of data in the test (that's why take())
test_dataset = test_dataset.take(len(test_df)) # Note: test_generator must have shuffle=False
"""DEFINE BASE MODEL"""
model = InceptionResNetV2(weights=None, classes=num_of_classes)
"""NSL"""
# TODO play with parm of adv_config
adversarial_config = nsl.configs.make_adv_reg_config(multiplier=multiplier,
adv_step_size=adv_step_size,
adv_grad_norm=adv_grad_norm)
adversarial_model = nsl.keras.AdversarialRegularization(model, label_keys=[LABEL_INPUT_NAME],
adv_config=adversarial_config)
checkpoint = ModelCheckpoint(filepath=f'nsl_weights_{model_name}_{multiplier}_{adv_step_size}_{adv_grad_norm}.hdf5',
save_best_only=True, verbose=1)
early_stopping = EarlyStopping(monitor='val_loss', patience=10, verbose=1)
reduce_lr = ReduceLROnPlateau(patience=5, verbose=1)
adversarial_model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
# TODO different loss?
print('============ fit adversarial model ============')
adversarial_model.fit(train_dataset, epochs=100, steps_per_epoch=np.ceil(len(train_df) / TRAIN_BATCH_SIZE),
validation_data=val_dataset, callbacks=[checkpoint, early_stopping, reduce_lr])
adversarial_model.load_weights(filepath=f'nsl_weights_{model_name}_{multiplier}_{adv_step_size}_{adv_grad_norm}.hdf5')
print('================== inference ==================')
# predictions = adv_model.predict(test_dataset) # TODO
result = adversarial_model.evaluate(test_dataset)
print(f'#RESULTS# NSL model: \n{dict(zip(adversarial_model.metrics_names, result))}\n'
f'model_name: {model_name}\n'
f'multiplier: {multiplier}\n'
f'adv_step_size: {adv_step_size}\n'
f'adv_grad_norm: {adv_grad_norm}')
"""########################"""
#
# def simple_model():
# class HParams(object):
# def __init__(self):
# self.input_shape = INPUT_SHAPE
# self.num_classes = 37
# self.conv_filters = [32, 64, 64]
# self.kernel_size = (3, 3)
# self.pool_size = (2, 2)
# self.num_fc_units = [64]
# self.batch_size = 32
# self.epochs = 30
# self.adv_multiplier = 0.2
# self.adv_step_size = 0.2
# self.adv_grad_norm = 'infinity'
#
# HPARAMS = HParams()
#
# def build_base_model(hparams):
# """Builds a model according to the architecture defined in `hparams`."""
# inputs = tf.keras.Input(
# shape=hparams.input_shape, dtype=tf.float32, name=IMAGE_INPUT_NAME)
#
# x = inputs
# for i, num_filters in enumerate(hparams.conv_filters):
# x = tf.keras.layers.Conv2D(
# num_filters, hparams.kernel_size, activation='relu')(
# x)
# if i < len(hparams.conv_filters) - 1:
# # max pooling between convolutional layers
# x = tf.keras.layers.MaxPooling2D(hparams.pool_size)(x)
# x = tf.keras.layers.Flatten()(x)
# for num_units in hparams.num_fc_units:
# x = tf.keras.layers.Dense(num_units, activation='relu')(x)
# pred = tf.keras.layers.Dense(hparams.num_classes, activation='softmax')(x)
# model = tf.keras.Model(inputs=inputs, outputs=pred)
# return model
#
# base_model = build_base_model(HPARAMS)
#
# base_model.compile(optimizer='adam', loss='categorical_crossentropy',
# metrics=['acc'])
# """try simple model"""
# print('================== fit simple model ==================')
# print(base_model.summary())
# base_model.fit(train_generator, epochs=30, steps_per_epoch=np.ceil(len(train_df) / TRAIN_BATCH_SIZE))
# print('================== inference ==================')
# result = base_model.evaluate(test_dataset)
# print(dict(zip(base_model.metrics_names, result)))
#
"""
############################################
############################################
############################################
DEAD CODE ISLAND
############################################
############################################
############################################
"""
"""GRAPH SIMILARITY REPRESENTATION - GET THE LAST LAYER'S OUTPUT BEFORE PREDICTIONS"""
# from keras import backend as K
# layer_function = K.function([model.layers[0].input], [model.layers[-2].output])
# # that's the representation of the layer one before the last
# layer_output = layer_function(train_dataset.__iter__().next())[0]
# def normalize(features):
# features[IMAGE_INPUT_NAME] = tf.cast(
# features[IMAGE_INPUT_NAME], dtype=tf.float32) / 255.0
# return features
#
#
# def convert_to_tuples(features):
# return features[IMAGE_INPUT_NAME], features[LABEL_INPUT_NAME]
# datasets = tfds.load('mnist')
# mnist_train_dataset = datasets['train']
# mnist_test_dataset = datasets['test']
#
# train_dataset = mnist_train_dataset.map(normalize).shuffle(10000).batch(HPARAMS.batch_size).map(convert_to_tuples)
# test_dataset = mnist_test_dataset.map(normalize).batch(BATCH_SIZE).map(convert_to_tuples)
#
#
#
# base_model = build_base_model(HPARAMS)
# base_model.summary()
#
# adv_config = nsl.configs.make_adv_reg_config(
# multiplier=HPARAMS.adv_multiplier,
# adv_step_size=HPARAMS.adv_step_size,
# adv_grad_norm=HPARAMS.adv_grad_norm
# )
#
# base_adv_model = build_base_model(HPARAMS)
# adv_model = nsl.keras.AdversarialRegularization(
# base_adv_model,
# label_keys=[LABEL_INPUT_NAME],
# adv_config=adv_config
# )
#
# train_set_for_adv_model = train_dataset.map(convert_to_dictionaries)
# test_set_for_adv_model = test_dataset.map(convert_to_dictionaries)
# flow_from_dataframe_args = [train_df, None, 'path', 'breed', None, (128, 128),
# 'rgb', None, 'categorical', 1, True, None, None,
# '', 'png', None, 'nearest', True]
# code for renaming layer's name - doesn't help
# for layer in model.layers:
# # change only the first layer name
# layer._name = 'image'
# break