八、Keras 高层接口

八、Keras 高层接口

内容参考来自https://github.com/dragen1860/Deep-Learning-with-TensorFlow-book开源书籍《TensorFlow2深度学习》，这只是我做的简单的学习笔记，方便以后复习

Keras 是一个主要由 Python 语言开发的开源神经网络计算库，最初由 François Chollet编写，它被设计为高度模块化和易扩展的高层神经网络接口，使得用户可以不需要过多的专业知识就可以简洁、快速地完成模型的搭建与训练。

1.常见功能模块

Keras 提供了一系列高层的神经网络相关类和函数，如经典数据集加载函数、网络层类、模型容器、损失函数类、优化器类、经典模型类等。

对于常见的网络层，我们一般直接使用层方式来完成模型的搭建，在 tf.keras.layers 命名空间(下文使用 layers 指代 tf.keras.layers)中提供了大量常见网络层的类，如全连接层、激活函数层、池化层、卷积层、循环神经网络层等。

以softmax为例

import tensorflow as tf
# 导入 keras 模型，不能使用 import keras，它导入的是标准的 Keras 库
from tensorflow import keras
from tensorflow.keras import Sequential,layers, optimizers,losses # 导入常见网络层类
x = tf.constant([2.,1.,0.1]) # 创建输入张量
layer = layers.Softmax(axis=-1) # 创建 Softmax 层
out = layer(x) # 调用 softmax 前向计算，输出为 out

可以通过 Keras 提供的网络容器 Sequential 将多个网络层封装成一个大网络模型，只需要调用网络模型的实例一次即可完成数据从第一层到最末层的顺序传播运算。

# 导入 Sequential 容器
from tensorflow.keras import layers, Sequential
network = Sequential([ # 封装为一个网络
layers.Dense(3, activation=None), # 全连接层，此处不使用激活函数
layers.ReLU(),#激活函数层
layers.Dense(2, activation=None), # 全连接层，此处不使用激活函数
layers.ReLU() #激活函数层
])
x = tf.random.normal([4,3])
out = network(x) # 输入从第一层开始，逐层传播至输出层，并返回输出层的输出

2.模型装配、训练与测试

compile,fit,predict

import tensorflow as tf
from tensorflow.keras import datasets, layers, optimizers, Sequential, metrics


def preprocess(x, y):
    """
    x is a simple image, not a batch
    """
    x = tf.cast(x, dtype=tf.float32) / 255.
    x = tf.reshape(x, [28 * 28])
    y = tf.cast(y, dtype=tf.int32)
    y = tf.one_hot(y, depth=10)
    return x, y


batchsz = 128
(x, y), (x_val, y_val) = datasets.mnist.load_data()
print('datasets:', x.shape, y.shape, x.min(), x.max())

db = tf.data.Dataset.from_tensor_slices((x, y))
db = db.map(preprocess).shuffle(60000).batch(batchsz)
ds_val = tf.data.Dataset.from_tensor_slices((x_val, y_val))
ds_val = ds_val.map(preprocess).batch(batchsz)

sample = next(iter(db))
print(sample[0].shape, sample[1].shape)

network = Sequential([layers.Dense(256, activation='relu'),
                      layers.Dense(128, activation='relu'),
                      layers.Dense(64, activation='relu'),
                      layers.Dense(32, activation='relu'),
                      layers.Dense(10)])
network.build(input_shape=(None, 28 * 28))
network.summary()

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

network.fit(db, epochs=5, validation_data=ds_val, validation_freq=2)

network.evaluate(ds_val)

sample = next(iter(ds_val))
x = sample[0]
y = sample[1]  # one-hot
pred = network.predict(x)  # [b, 10]
# convert back to number
y = tf.argmax(y, axis=1)
pred = tf.argmax(pred, axis=1)

print(pred)
print(y)

3.模型保存与加载

3.1张量方式

network.save_weights('weights.ckpt')
network.load_weights('weights.ckpt')

完整例子

import  os
os.environ['TF_CPP_MIN_LOG_LEVEL']='2'

import  tensorflow as tf
from    tensorflow.keras import datasets, layers, optimizers, Sequential, metrics


def preprocess(x, y):
    """
    x is a simple image, not a batch
    """
    x = tf.cast(x, dtype=tf.float32) / 255.
    x = tf.reshape(x, [28*28])
    y = tf.cast(y, dtype=tf.int32)
    y = tf.one_hot(y, depth=10)
    return x,y


batchsz = 128
(x, y), (x_val, y_val) = datasets.mnist.load_data()
print('datasets:', x.shape, y.shape, x.min(), x.max())



db = tf.data.Dataset.from_tensor_slices((x,y))
db = db.map(preprocess).shuffle(60000).batch(batchsz)
ds_val = tf.data.Dataset.from_tensor_slices((x_val, y_val))
ds_val = ds_val.map(preprocess).batch(batchsz) 

sample = next(iter(db))
print(sample[0].shape, sample[1].shape)


network = Sequential([layers.Dense(256, activation='relu'),
                     layers.Dense(128, activation='relu'),
                     layers.Dense(64, activation='relu'),
                     layers.Dense(32, activation='relu'),
                     layers.Dense(10)])
network.build(input_shape=(None, 28*28))
network.summary()




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

network.fit(db, epochs=3, validation_data=ds_val, validation_freq=2)
 
network.evaluate(ds_val)

network.save_weights('weights.ckpt')
print('saved weights.')
del network

network = Sequential([layers.Dense(256, activation='relu'),
                     layers.Dense(128, activation='relu'),
                     layers.Dense(64, activation='relu'),
                     layers.Dense(32, activation='relu'),
                     layers.Dense(10)])
network.compile(optimizer=optimizers.Adam(lr=0.01),
		loss=tf.losses.CategoricalCrossentropy(from_logits=True),
		metrics=['accuracy']
	)
network.load_weights('weights.ckpt')
print('loaded weights!')
network.evaluate(ds_val)

3.2网络方式

# 保存模型结构与模型参数到文件
network.save('model.h5')
# 从文件恢复网络结构与网络参数
network = keras.models.load_model('model.h5')

完整的例子

import  os
os.environ['TF_CPP_MIN_LOG_LEVEL']='2'

import  tensorflow as tf
from    tensorflow.keras import datasets, layers, optimizers, Sequential, metrics


def preprocess(x, y):
    """
    x is a simple image, not a batch
    """
    x = tf.cast(x, dtype=tf.float32) / 255.
    x = tf.reshape(x, [28*28])
    y = tf.cast(y, dtype=tf.int32)
    y = tf.one_hot(y, depth=10)
    return x,y


batchsz = 128
(x, y), (x_val, y_val) = datasets.mnist.load_data()
print('datasets:', x.shape, y.shape, x.min(), x.max())



db = tf.data.Dataset.from_tensor_slices((x,y))
db = db.map(preprocess).shuffle(60000).batch(batchsz)
ds_val = tf.data.Dataset.from_tensor_slices((x_val, y_val))
ds_val = ds_val.map(preprocess).batch(batchsz) 

sample = next(iter(db))
print(sample[0].shape, sample[1].shape)


network = Sequential([layers.Dense(256, activation='relu'),
                     layers.Dense(128, activation='relu'),
                     layers.Dense(64, activation='relu'),
                     layers.Dense(32, activation='relu'),
                     layers.Dense(10)])
network.build(input_shape=(None, 28*28))
network.summary()




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

network.fit(db, epochs=3, validation_data=ds_val, validation_freq=2)
 
network.evaluate(ds_val)

network.save('model.h5')
print('saved total model.')
del network

print('loaded model from file.')
network = tf.keras.models.load_model('model.h5', compile=False)
network.compile(optimizer=optimizers.Adam(lr=0.01),
        loss=tf.losses.CategoricalCrossentropy(from_logits=True),
        metrics=['accuracy']
    )
x_val = tf.cast(x_val, dtype=tf.float32) / 255.
x_val = tf.reshape(x_val, [-1, 28*28])
y_val = tf.cast(y_val, dtype=tf.int32)
y_val = tf.one_hot(y_val, depth=10)
ds_val = tf.data.Dataset.from_tensor_slices((x_val, y_val)).batch(128)
network.evaluate(ds_val)

3.3SavedModel 方式

# 保存模型结构与模型参数到文件
tf.saved_model.save(network, 'model-savedmodel')
# 从文件恢复网络结构与网络参数
network = tf.saved_model.load('model-savedmodel')

完整的例子

# 例子能跑通，但是我没明白
import os

os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

import tensorflow as tf
from tensorflow.keras import datasets, layers, optimizers, Sequential, metrics


def preprocess(x, y):
    """
    x is a simple image, not a batch
    """
    x = tf.cast(x, dtype=tf.float32) / 255.
    x = tf.reshape(x, [28 * 28])
    y = tf.cast(y, dtype=tf.int32)
    y = tf.one_hot(y, depth=10)
    return x, y


batchsz = 128
(x, y), (x_val, y_val) = datasets.mnist.load_data()
print('datasets:', x.shape, y.shape, x.min(), x.max())

db = tf.data.Dataset.from_tensor_slices((x, y))
db = db.map(preprocess).shuffle(60000).batch(batchsz)
ds_val = tf.data.Dataset.from_tensor_slices((x_val, y_val))
ds_val = ds_val.map(preprocess).batch(batchsz)

sample = next(iter(db))
print(sample[0].shape, sample[1].shape)

network = Sequential([layers.Dense(256, activation='relu'),
                      layers.Dense(128, activation='relu'),
                      layers.Dense(64, activation='relu'),
                      layers.Dense(32, activation='relu'),
                      layers.Dense(10)])
network.build(input_shape=(None, 28 * 28))
network.summary()

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

network.fit(db, epochs=3, validation_data=ds_val, validation_freq=2)

network.evaluate(ds_val)

# 保存模型结构与模型参数到文件
tf.saved_model.save(network, 'model-savedmodel')
print('saving savedmodel.')
del network  # 删除网络对象

print('load savedmodel from file.')
# 从文件恢复网络结构与网络参数
network = tf.saved_model.load('model-savedmodel')
print(network)
# 准确率计量器
acc_meter = metrics.CategoricalAccuracy()
for x, y in ds_val:  # 遍历测试集
    pred = network(x)  # 前向计算
acc_meter.update_state(y_true=y, y_pred=pred)  # 更新准确率统计
# 打印准确率
print("Test Accuracy:%f" % acc_meter.result())

4.自定义网络

尽管 Keras 提供了很多的常用网络层类，但深度学习可以使用的网络层远远不止这些。科研工作者一般是自行实现了较为新颖的网络层，经过大量实验验证有效后，深度学习框架才会跟进，内置对这些网络层的支持。因此掌握自定义网络层、网络的实现非常重要。

import tensorflow as tf
from tensorflow.keras import datasets, layers, optimizers, Sequential, metrics
from tensorflow import keras


def preprocess(x, y):
    """
    x is a simple image, not a batch
    """
    x = tf.cast(x, dtype=tf.float32) / 255.
    x = tf.reshape(x, [28 * 28])
    y = tf.cast(y, dtype=tf.int32)
    y = tf.one_hot(y, depth=10)
    return x, y


batchsz = 128
(x, y), (x_val, y_val) = datasets.mnist.load_data()
print('datasets:', x.shape, y.shape, x.min(), x.max())

db = tf.data.Dataset.from_tensor_slices((x, y))
db = db.map(preprocess).shuffle(60000).batch(batchsz)
ds_val = tf.data.Dataset.from_tensor_slices((x_val, y_val))
ds_val = ds_val.map(preprocess).batch(batchsz)

sample = next(iter(db))
print(sample[0].shape, sample[1].shape)

network = Sequential([layers.Dense(256, activation='relu'),
                      layers.Dense(128, activation='relu'),
                      layers.Dense(64, activation='relu'),
                      layers.Dense(32, activation='relu'),
                      layers.Dense(10)])
network.build(input_shape=(None, 28 * 28))
network.summary()


class MyDense(layers.Layer):

    def __init__(self, inp_dim, outp_dim):
        super(MyDense, self).__init__()

        self.kernel = self.add_weight('w', [inp_dim, outp_dim])
        self.bias = self.add_weight('b', [outp_dim])

    def call(self, inputs, training=None):
        out = inputs @ self.kernel + self.bias

        return out


class MyModel(keras.Model):

    def __init__(self):
        super(MyModel, self).__init__()

        self.fc1 = MyDense(28 * 28, 256)
        self.fc2 = MyDense(256, 128)
        self.fc3 = MyDense(128, 64)
        self.fc4 = MyDense(64, 32)
        self.fc5 = MyDense(32, 10)

    def call(self, inputs, training=None):
        x = self.fc1(inputs)
        x = tf.nn.relu(x)
        x = self.fc2(x)
        x = tf.nn.relu(x)
        x = self.fc3(x)
        x = tf.nn.relu(x)
        x = self.fc4(x)
        x = tf.nn.relu(x)
        x = self.fc5(x)

        return x


network = MyModel()

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

network.fit(db, epochs=5, validation_data=ds_val,
            validation_freq=2)

network.evaluate(ds_val)

sample = next(iter(ds_val))
x = sample[0]
y = sample[1]  # one-hot
pred = network.predict(x)  # [b, 10]
# convert back to number 
y = tf.argmax(y, axis=1)
pred = tf.argmax(pred, axis=1)

print(pred)
print(y)

5.模型乐园

对于常用的网络模型，如 ResNet、VGG 等，不需要手动创建网络，可以直接从keras.applications 子模块中通过一行代码即可创建并使用这些经典模型，同时还可以通过设置 weights 参数加载预训练的网络参数，非常方便。

import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import datasets, layers, optimizers, Sequential, metrics

# 加载预训练网络模型，并去掉最后一层
# 通过设置 include_top 参数为 False，可以选择去掉 ResNet50 最后一层
resnet = keras.applications.ResNet50(weights='imagenet', include_top=False)
resnet.summary()
# 测试网络的输出
x = tf.random.normal([4, 224, 224, 3])
out = resnet(x)
print(out.shape)

# 新建池化层
global_average_layer = tf.keras.layers.GlobalAveragePooling2D()
# 利用上一层的输出作为本层的输入，测试其输出
x = tf.random.normal([4, 7, 7, 2048])
out = global_average_layer(x)
print(out.shape)

# 新建全连接层
fc = tf.keras.layers.Dense(100)
# 利用上一层的输出作为本层的输入，测试其输出
x = tf.random.normal([4, 2048])
out = fc(x)
print(out.shape)

# 重新包裹成我们的网络模型
mynet = Sequential([resnet, global_average_layer, fc])
mynet.summary()

resnet.trainable = False
mynet.summary()

6.测量工具

Keras 的测量工具的使用方法一般有 4 个主要步骤：新建测量器，写入数据，读取统计数据和清零测量器。

6.1新建测量器

# 新建平均测量器，适合 Loss 数据
loss_meter = metrics.Mean()

6.2 写入数据

# 记录采样的数据，通过 float()函数将张量转换为普通数值
loss_meter.update_state(float(loss))

6.3 读取统计信息

# 打印统计期间的平均 loss
print(step, 'loss:', loss_meter.result())

6.4 清除状态

if step % 100 == 0:
    # 打印统计的平均 loss
    print(step, 'loss:', loss_meter.result())
    loss_meter.reset_states() # 打印完后，清零测量器

参考：TensorFlow2.0: keras.metrics的使用

7.可视化

这部分没看懂，以后有机会再更新。

欢迎关注的我的微信公众号，同步更新，嘻嘻