/
tf_keras_sequential_verbose.py
308 lines (241 loc) · 10.6 KB
/
tf_keras_sequential_verbose.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
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
import tensorflow as tf
import numpy as np
from PIL import Image
print(tf.__version__)
# 2.1.0
tf.random.set_seed(0)
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()
print(type(x_train))
# <class 'numpy.ndarray'>
print(x_train.shape)
# (60000, 28, 28)
print(x_train.dtype)
# uint8
print(x_train.min(), '-', x_train.max())
# 0 - 255
Image.fromarray(x_train[0]).resize((256, 256)).save('../data/img/dst/mnist_sample_resize.png')
print(x_test.shape)
# (10000, 28, 28)
print(type(y_train))
# <class 'numpy.ndarray'>
print(y_train.shape)
# (60000,)
print(y_train.dtype)
# uint8
print(y_train.min(), '-', y_train.max())
# 0 - 9
print(np.unique(y_train))
# [0 1 2 3 4 5 6 7 8 9]
print(y_train[0])
# 5
print(y_test.shape)
# (10000,)
x_train = x_train / 255
x_test = x_test / 255
print(x_train.dtype)
# float64
print(x_train.min(), '-', x_train.max())
# 0.0 - 1.0
print(tf.keras.Sequential is tf.keras.models.Sequential)
# True
model = tf.keras.Sequential([
tf.keras.layers.Flatten(input_shape=(28, 28), name='flatten_layer'),
tf.keras.layers.Dense(128, activation='relu'),
tf.keras.layers.Dropout(0.2),
tf.keras.layers.Dense(10, activation='softmax')
], name='my_model')
model.summary()
# Model: "my_model"
# _________________________________________________________________
# Layer (type) Output Shape Param #
# =================================================================
# flatten_layer (Flatten) (None, 784) 0
# _________________________________________________________________
# dense (Dense) (None, 128) 100480
# _________________________________________________________________
# dropout (Dropout) (None, 128) 0
# _________________________________________________________________
# dense_1 (Dense) (None, 10) 1290
# =================================================================
# Total params: 101,770
# Trainable params: 101,770
# Non-trainable params: 0
# _________________________________________________________________
model_1 = tf.keras.Sequential(name='my_model_1')
model_1.add(tf.keras.layers.Flatten(input_shape=(28, 28), name='flatten_layer_1'))
model_1.add(tf.keras.layers.Dense(128, activation='relu'))
model_1.add(tf.keras.layers.Dropout(0.2))
model_1.add(tf.keras.layers.Dense(10, activation='softmax'))
model_1.summary()
# Model: "my_model_1"
# _________________________________________________________________
# Layer (type) Output Shape Param #
# =================================================================
# flatten_layer_1 (Flatten) (None, 784) 0
# _________________________________________________________________
# dense_2 (Dense) (None, 128) 100480
# _________________________________________________________________
# dropout_1 (Dropout) (None, 128) 0
# _________________________________________________________________
# dense_3 (Dense) (None, 10) 1290
# =================================================================
# Total params: 101,770
# Trainable params: 101,770
# Non-trainable params: 0
# _________________________________________________________________
from tensorflow.keras import Sequential
from tensorflow.keras.layers import Dense, Dropout, Flatten
model_2 = Sequential([
Flatten(input_shape=(28, 28), name='flatten_layer_2'),
Dense(128, activation='relu'),
Dropout(0.2),
Dense(10, activation='softmax')
], name='my_model_2')
model_2.summary()
# Model: "my_model_2"
# _________________________________________________________________
# Layer (type) Output Shape Param #
# =================================================================
# flatten_layer_2 (Flatten) (None, 784) 0
# _________________________________________________________________
# dense_4 (Dense) (None, 128) 100480
# _________________________________________________________________
# dropout_2 (Dropout) (None, 128) 0
# _________________________________________________________________
# dense_5 (Dense) (None, 10) 1290
# =================================================================
# Total params: 101,770
# Trainable params: 101,770
# Non-trainable params: 0
# _________________________________________________________________
print(tf.keras.Sequential is Sequential)
# True
import tensorflow.keras.layers as L
print(tf.keras.layers.Dense is L.Dense)
# True
model.compile(optimizer='adam',
loss='sparse_categorical_crossentropy',
metrics=['accuracy'])
model_1.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=0.01),
loss=tf.keras.losses.SparseCategoricalCrossentropy(),
metrics=[tf.keras.metrics.SparseCategoricalAccuracy()])
model_1.compile(optimizer=tf.keras.optimizers.Adam(),
loss=tf.keras.losses.sparse_categorical_crossentropy,
metrics=[tf.keras.metrics.sparse_categorical_accuracy])
callbacks = [tf.keras.callbacks.EarlyStopping(patience=2, restore_best_weights=True),
tf.keras.callbacks.ModelCheckpoint(
'../data/temp/mnist_sequential_{epoch:03d}_{val_loss:.4f}.h5',
save_best_only=True
)]
history = model.fit(x_train, y_train, batch_size=128, epochs=20,
validation_split=0.2, callbacks=callbacks)
# Train on 48000 samples, validate on 12000 samples
# Epoch 1/20
# 48000/48000 [==============================] - 1s 29us/sample - loss: 0.4443 - accuracy: 0.8751 - val_loss: 0.2079 - val_accuracy: 0.9427
# Epoch 2/20
# 48000/48000 [==============================] - 1s 22us/sample - loss: 0.2112 - accuracy: 0.9393 - val_loss: 0.1507 - val_accuracy: 0.9584
# Epoch 3/20
# 48000/48000 [==============================] - 1s 23us/sample - loss: 0.1590 - accuracy: 0.9540 - val_loss: 0.1238 - val_accuracy: 0.9647
# Epoch 4/20
# 48000/48000 [==============================] - 1s 23us/sample - loss: 0.1302 - accuracy: 0.9616 - val_loss: 0.1083 - val_accuracy: 0.9679
# Epoch 5/20
# 48000/48000 [==============================] - 1s 22us/sample - loss: 0.1111 - accuracy: 0.9671 - val_loss: 0.0992 - val_accuracy: 0.9709
# Epoch 6/20
# 48000/48000 [==============================] - 1s 23us/sample - loss: 0.0960 - accuracy: 0.9710 - val_loss: 0.0927 - val_accuracy: 0.9719
# Epoch 7/20
# 48000/48000 [==============================] - 1s 24us/sample - loss: 0.0855 - accuracy: 0.9742 - val_loss: 0.0880 - val_accuracy: 0.9732
# Epoch 8/20
# 48000/48000 [==============================] - 1s 23us/sample - loss: 0.0748 - accuracy: 0.9772 - val_loss: 0.0809 - val_accuracy: 0.9760
# Epoch 9/20
# 48000/48000 [==============================] - 1s 24us/sample - loss: 0.0691 - accuracy: 0.9787 - val_loss: 0.0819 - val_accuracy: 0.9743
# Epoch 10/20
# 48000/48000 [==============================] - 1s 23us/sample - loss: 0.0630 - accuracy: 0.9808 - val_loss: 0.0771 - val_accuracy: 0.9758
# Epoch 11/20
# 48000/48000 [==============================] - 1s 22us/sample - loss: 0.0569 - accuracy: 0.9831 - val_loss: 0.0801 - val_accuracy: 0.9753
# Epoch 12/20
# 48000/48000 [==============================] - 1s 22us/sample - loss: 0.0518 - accuracy: 0.9844 - val_loss: 0.0778 - val_accuracy: 0.9754
print(type(history))
# <class 'tensorflow.python.keras.callbacks.History'>
print(type(history.history))
# <class 'dict'>
print(history.history.keys())
# dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
print(history.history['accuracy'])
# [0.87514585, 0.93927085, 0.9539792, 0.96164584, 0.967125, 0.9710417, 0.9741875, 0.9771875, 0.97866666, 0.98075, 0.98310417, 0.98441666]
test_loss, test_acc = model.evaluate(x_test, y_test, verbose=0)
print(test_loss)
# 0.07156232252293267
print(test_acc)
# 0.9786
print(x_test.shape)
# (10000, 28, 28)
predictions = model.predict(x_test)
print(type(predictions))
# <class 'numpy.ndarray'>
print(predictions.shape)
# (10000, 10)
print(predictions[0])
# [3.22501933e-06 1.48081245e-08 2.89624350e-05 3.07988637e-04
# 2.93199742e-10 3.28161093e-07 2.59668814e-11 9.99620199e-01
# 5.64465245e-06 3.35659788e-05]
print(predictions[0].sum())
# 0.99999994
print(predictions[0].argmax())
# 7
results = predictions.argmax(axis=1)
print(results)
# [7 2 1 ... 4 5 6]
print(type(results))
# <class 'numpy.ndarray'>
print(results.shape)
# (10000,)
img = np.array(Image.open('../data/img/dst/mnist_sample_resize.png').resize((28, 28))) / 255
print(img.shape)
# (28, 28)
# predictions_single = model.predict(img)
# ValueError: Error when checking input: expected flatten_input to have 3 dimensions, but got array with shape (28, 28)
img_expand = img[np.newaxis, ...]
print(img_expand.shape)
# (1, 28, 28)
print(img[None, ...].shape)
# (1, 28, 28)
print(np.expand_dims(img, 0).shape)
# (1, 28, 28)
print(tf.newaxis)
# None
print(img[tf.newaxis, ...].shape)
# (1, 28, 28)
print(img[np.newaxis].shape)
# (1, 28, 28)
predictions_single = model.predict(img_expand)
print(predictions_single)
# [[6.8237398e-09 1.0004978e-08 4.0168429e-06 4.5704491e-02 2.8772252e-14
# 9.5429057e-01 5.5912805e-12 3.1738683e-08 6.9545142e-10 9.2607473e-07]]
print(predictions_single.shape)
# (1, 10)
print(predictions_single[0].argmax())
# 5
model.save('../data/temp/my_model.h5')
new_model = tf.keras.models.load_model('../data/temp/my_model.h5')
new_model.summary()
# Model: "my_model"
# _________________________________________________________________
# Layer (type) Output Shape Param #
# =================================================================
# flatten_layer (Flatten) (None, 784) 0
# _________________________________________________________________
# dense (Dense) (None, 128) 100480
# _________________________________________________________________
# dropout (Dropout) (None, 128) 0
# _________________________________________________________________
# dense_1 (Dense) (None, 10) 1290
# =================================================================
# Total params: 101,770
# Trainable params: 101,770
# Non-trainable params: 0
# _________________________________________________________________
print(new_model.evaluate(x_test, y_test, verbose=0))
# [0.07156232252293267, 0.9786]
print(new_model.predict(img_expand))
# [[6.8237398e-09 1.0004978e-08 4.0168429e-06 4.5704491e-02 2.8772252e-14
# 9.5429057e-01 5.5912805e-12 3.1738683e-08 6.9545142e-10 9.2607473e-07]]