Tensorflow2.0 save model

import tensorflow as tf
import json
import tensorflow.keras.models
# This is a separate package
import tensorflow.keras
mnist = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0

model = tf.keras.models.Sequential([
tf.keras.layers.Flatten(),
tf.keras.layers.Dense(128, activation=‘relu’),
tf.keras.layers.Dense(10, activation=‘softmax’)
])

model.compile(optimizer=‘adam’,
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
metrics=[‘sparse_categorical_accuracy’])

model.fit(x_train, y_train, batch_size=32, epochs=5, validation_data=(x_test, y_test), validation_freq=1)
model.summary()
‘’’
# Save as h5 Model ： Use model.save To save the model , The structure of the model 、 Weight parameters and optimizer settings , Setting of cost function ,metrics All settings of are saved
# Extraction method ：tensorflow.keras.models.load_model(“mymodel.h5”)
#model.save(“mymodel.h5”)
‘’’
‘’’
#SavedModel yes Tensorflow A model format in , Its advantage is that it has nothing to do with language , No addition h5 suffix ,pb structure ,variable A weight , Can be used in addition to Sequential Models other than
#model.save(“mymodel”)
# Just want to save the structure of the model , Without saving the weight of the model 、 Optimizer and cost function , The main use of json
‘’’
‘’’
json_config=model.to_json()
with open(“model.json”,“w”)as m:
json.dump(json_config,m)
# Read the model ：
with open(“model.json”) as m:
json_config=json.load(m)
model2=tensorflow.keras.models.model_from_json(json_config)
model2.summary()
‘’’
‘’’
Save only the parameters of the model ：
model.save_weights（‘path_to_my_model.h5’）
# Onload , You need to use json Import model structure , Or define the model structure
model.load_weights（‘path_to_my_model.h5’）
# You also need to define compile The content in , Including optimizer and cost function . because model.save() Will save compile The content in ,
and model.save_weights Only the parameters of the model will be saved . therefore ,load_weights It needs to be redefined in the future compile The content of , In this way, the model can be further trained .
‘’’