调用tensorflow2中的tf.keras.applications.VGG16并进行微调

调用tensorflow2中的tf.keras.applications.VGG16并进行微调
环境配置：tensorflow版本2.3.0；cuda 10.1；python3.8
数据集：cifar10
代码如下：

import tensorflow as tf
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.cifar10.load_data()    #图像大小是32×32×3
base_model = tf.keras.applications.VGG16(include_top=False, weights='imagenet', input_shape=(32, 32, 3))

#方法1
x = tf.keras.layers.Flatten()(x)
x = tf.keras.layers.Dense(256, activation='relu')(x)
x = tf.keras.layers.Dense(10, activation='softmax')(x)
model = tf.keras.Model(inputs=base_model.input, outputs=x)

#方法2
# model = tf.keras.Sequential()
# for layer in base_model.layers:
#     model.add(layer)
# model.add(tf.keras.layers.Flatten())
# model.add(tf.keras.layers.Dense(256, activation='relu'))
# model.add(tf.keras.layers.Dense(10, activation='softmax'))

for layer in base_model.layers:
    layer.trainable = False    #冻结
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=0.0001), loss='sparse_categorical_crossentropy', metrics=['accuracy'])
model.fit(x_train, y_train, epochs=5, batch_size=32)
model.evaluate(x_test, y_test, verbose=2)

1563/1563 [==============================] - 26s 16ms/step - loss: 1.4198 - accuracy: 0.4906
Epoch 2/5
1563/1563 [==============================] - 25s 16ms/step - loss: 0.5911 - accuracy: 0.8026
Epoch 3/5
1563/1563 [==============================] - 26s 16ms/step - loss: 0.4155 - accuracy: 0.8612
Epoch 4/5
1563/1563 [==============================] - 26s 16ms/step - loss: 0.3003 - accuracy: 0.8984
Epoch 5/5
1563/1563 [==============================] - 26s 16ms/step - loss: 0.2215 - accuracy: 0.9261
313/313 - 2s - loss: 0.4944 - accuracy: 0.8448