keras: Build Model_model.build

---

keras.layers

inputs = keras.layers.Input(shape=(784,))		# 同 keras.Input()

x = keras.layers.Dense(64)(x)

x = keras.layers.Flatten()(x)

x = keras.layers.average([y1, y2, y3])

x = keras.layers.concatenate([title_features, body_features, tags_input])

x = keras.layers.Embedding(input_dim=10, output_dim=32)(title_input)
1
2
3
4
5
6
7
8
9
10
11

• Convolutional layers: Conv1D, Conv2D, Conv3D, Conv2DTranspose
• Pooling layers: MaxPooling1D, MaxPooling2D, MaxPooling3D, AveragePooling1D, GlobalAveragePooling2D
• RNN layers: GRU, LSTM, ConvLSTM2D
• BatchNormalization, Dropout, Embedding

Activation

activation="relu"
activation="softmax"
1
2

utils

keras.utils.to_categorical(y_train, 10)
1

Build Model

Sequential & Model

keras.Sequenctial只适合简单的情况，复杂的情况需要keras.Model。下面是复杂的情况：

• 您的模型有多个输入或多个输出
  ticket的例子
• 任何层都有多个输入或多个输出
  ResNet的例子
• 您需要进行层共享
• 您想要非线性拓扑(例如，一个残余连接，一个多分支模型)

Code

inputs = keras.layers.Input(shape=(784,))
x = keras.layers.Dense(64, activation="relu")(inputs)
outputs = keras.layers.Dense(10)(x)

model = keras.Model(inputs=inputs, outputs=outputs)
1
2
3
4
5

每层详解

• 指定shape

inputs = keras.layers.Input(shape=(784,))
1

The shape of the data is set as a 784-dimensional vector. The batch size is always omitted since only the shape of each sample is specified.

• 可变尺寸

body_input = keras.layers.Input(shape=(None,))

# 输入200x200 RGB图像将具有形状(200,200,3)，但输入任何大小的RGB图像将具有形状(None, None, 3)。
inputs = keras.Input(shape=(None, None, 3))
1
2
3
4

查看详情

print(inputs.shape)
# TensorShape([None, 784])，表示784 dimensional vector
# 这是inputs 的输出的形状，而不是输入

print(inputs.dtype)
# tf.float32
1
2
3
4
5
6

model.summary()

'''
Model: "mnist_model"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         [(None, 784)]             0         
_________________________________________________________________
dense (Dense)                (None, 64)                50240     
_________________________________________________________________
dense_1 (Dense)              (None, 64)                4160      
_________________________________________________________________
dense_2 (Dense)              (None, 10)                650       
=================================================================
Total params: 55,050
Trainable params: 55,050
Non-trainable params: 0
_________________________________________________________________
'''
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

第一个None表示的是batch_size(Here the batch size is None, which indicates our model can process batches of any size)，input_1就是名字，[(None, 784)]就是输出的形状。

可视化

keras.utils.plot_model(model, "my_first_model.png")
1

keras.utils.plot_model(model, "my_first_model_with_shape_info.png", show_shapes=True)
1

Save & Load Model

This saved file includes the:

• model architecture
• model weight values (that were learned during training)
• model training config, if any (as passed to compile)
• optimizer and its state, if any (to restart training where you left off)

model.save("path_to_my_model")
1

model = keras.models.load_model("path_to_my_model")
1