2 three modeling methods

Three modeling methods

Sequential Method Sequence model Function model Subclassing model
Functional expression API Method
Model Subclass custom models

One 、Sequential model

# The first one is Sequential model、
from tensorflow.keras import layers
import tensorflow as tf

model = tf.keras.Sequential()
model.add(layers.Dense(64, activation='relu')) # first floor 
model.add(layers.Dense(64, activation='relu')) # The second floor 
model.add(layers.Dense(10)) # The third level 
# I can go ahead and add 

# The second kind 
model = tf.keras.Sequential([
    layers.Dense(64, activation='relu', input_shape=(32,)), # first floor 
    layers.Dense(64, activation='relu'), # The second floor 
    layers.Dense(10) # The third level 
    # I can go ahead and add 
])

#
model.compile(optimizer=tf.keras.optimizers.Adam(0.01),
             loss=tf.keras.losses.CategoricalCrossentropy(from_logits=True),
             metrics=['accuracy'])

import numpy as np

data = np.random.random((1000,32))
labels = np.random.random((1000,10))

# model training 
model.fit(data, labels, epochs=10, batch_size=32)

Two 、Functional model

A functional model is a way to create a model , This model is better than tf.keras.Sequential More flexible . Functional models can deal with models with nonlinear topology , Models with shared layers, models with multiple inputs or outputs, and so on

Deep learning models are usually directed acyclic graphs of layers （DAG） The main idea of . therefore , Functional models are a way to build layer diagrams .

for instance ： “”" (input: 32-dimensional vectors)
↧ [Dense (64 units, relu activation)]
↧ [Dense (64 units, relu activation)]
↧ [Dense (10 units, softmax activation)]
↧ (output: logits of a probability distribution over 10 classes) “”"

inputs = tf.keras.Input(shape=(32,))
x = layers.Dense(64, activation='relu')(inputs) # first floor 
x = layers.Dense(64, activation='relu')(x) # The second floor 
predictions = layers.Dense(10)(x) # The third level 

model = tf.keras.Model(inputs=inputs, outputs=predictions)

model.compile(optimizer=tf.keras.optimizers.RMSprop(0.001),
           loss=tf.keras.losses.CategoricalCrossentropy(from_logits=True),
             metrics=['accuracy'])

import tensorflow as tf
data = np.random.random((1000,32))
labels = np.random.random((1000,10))
model.fit(data, labels, batch_size=1, epochs=5)

Model subclassing
By subclassing tf.keras.Model And define your own forward propagation model to build a fully customizable model . and eager execution Models complement each other .