1. import numpy as np
2. import tensorflow as tf
3. import tensorflow.keras as keras
4. from tensorflow.keras import Sequential
5. from tensorflow.keras.layers import Conv2D, AvgPool2D, Flatten, Dense, Dropout
6.  
7.  
8. def load_train(path):
9.     x_train = np.load(path + 'train_features.npy')
10.     y_train = np.load(path + 'train_target.npy')
11.     x_train = x_train.reshape(x_train.shape[0],28,28,1)/255.0
12.     return x_train, y_train
13.    
14. def create_model(input_shape):
15.     model = Sequential()
16.     model.add(Conv2D(filters = 32, kernel_size=(3,3),padding = 'same', activation = 'relu',input_shape=(28,28,1)))
17.     model.add(AvgPool2D(pool_size=(2, 2)))
18.     model.add(Conv2D(filters = 64, kernel_size=(3,3), activation = 'relu'))
19.     model.add(AvgPool2D(pool_size=(2, 2)))
20.     model.add(Conv2D(filters = 128, kernel_size=(3,3), activation = 'relu'))
21.     model.add(Flatten())
22.     model.add(Dense(units=100,activation = 'relu'))
23.     model.add(Dense(units=50,activation = 'relu'))
24.     model.add(Dense(units=30,activation = 'relu'))
25.     model.add(Dense(units=10,activation = 'softmax'))
26.     model.compile(loss='sparse_categorical_crossentropy', optimizer='adam', metrics=['acc'])
27.     return model
28.    
29. def train_model(model, train_data, test_data, batch_size=32, epochs=5,
30.                 steps_per_epoch=None, validation_steps=None):
31.     x_train, y_train = train_data
32.     x_test, y_test = test_data
33.     model.fit(train_data, validation_data=test_data,batch_size=batch_size,
34.               epochs=epochs, steps_per_epoch=steps_per_epoch,validation_steps=validation_steps,
35.               verbose=2, shuffle=True)
36.     return model