tensorflow2.0学习笔记：模型的保存与载入

tensorflow2.0 模型的保存

1. 通过ModelCheckpoint保存模型，该方式只保存网络达到某一种要求状态下的参数(比如训练模型达到最好时的参数)，而不管其他的状态。
2. 通过saved_model保存，是模型的一种保存格式，可以把这个保存的模型直接交给用户来部署，而不需要提供源代码或相关的信息，保存的模型就包含了这些信息。
3. 通过model.save保存模型，是把模型所有的状态都保存起来，简单粗暴。

import matplotlib as mpl
import matplotlib.pyplot as plt
%matplotlib inline
import numpy as np
import sklearn
import pandas as pd
import os
import sys
import time
import tensorflow as tf

from tensorflow import keras
print(tf.__version__)

2.0.0

fashion_mnist = keras.datasets.fashion_mnist
(x_train_all, y_train_all), (x_test, y_test) = fashion_mnist.load_data()
x_valid, x_train = x_train_all[:5000], x_train_all[5000:]
y_valid, y_train = y_train_all[:5000], y_train_all[5000:]

print(x_valid.shape, y_valid.shape)
print(x_train.shape, y_train.shape)
print(x_test.shape, y_test.shape)

model = keras.models.Sequential([
    keras.layers.Flatten(input_shape=[28, 28]),
    keras.layers.Dense(300, activation='relu'),
    keras.layers.Dense(100, activation='relu'),
    keras.layers.Dense(10, activation='softmax')
])

model.compile(loss="sparse_categorical_crossentropy",
              optimizer = "sgd",
              metrics = ["accuracy"])

1. ModelCheckpoint保存模型

logdir = './graph_def_and_weights'
if not os.path.exists(logdir):
    os.mkdir(logdir)
output_model_file = os.path.join(logdir,
                                 "fashion_mnist_weights.h5")

callbacks = [
    keras.callbacks.TensorBoard(logdir),
    keras.callbacks.ModelCheckpoint(output_model_file, #保存模型
                                    save_best_only = True,
                                    save_weights_only = False),
    keras.callbacks.EarlyStopping(patience=5, min_delta=1e-3),
]
history = model.fit(x_train_scaled, y_train, epochs=10,
                    validation_data=(x_valid_scaled, y_valid),
                    callbacks = callbacks)

model.evaluate(x_test_scaled, y_test)

loaded_model = keras.models.load_model(output_model_file)  #载入模型
loaded_model.evaluate(x_test_scaled, y_test)

2. saved_model保存模型

tf.saved_model.save(model, "./keras_saved_graph")#保存模型 
!saved_model_cli show --dir ./keras_saved_graph --all   #查看模型，签名

loaded_saved_model = tf.saved_model.load('./keras_saved_graph') #载入模型
print(list(loaded_saved_model.signatures.keys())) #打印签名

inference = loaded_saved_model.signatures['serving_default'] #获取函数句柄
print(inference)
print(inference.structured_outputs)

results = inference(tf.constant(x_test_scaled[0:1])) #测试
print(results['dense_2'])

3. model.save保存模型

model.save('save_model.h5')
print('saved total model.')

print('loaded model from file.')
loaded_model = tf.keras.models.load_model('save_model.h5', compile=False)  # 不需要重新创建网络

loaded_model.evaluate(x_test_scaled, y_test)

参数的保存和载入

model.save_weights(os.path.join(logdir, "fashion_mnist_weights_2.h5")) #保存参数
logdir = './graph_def_and_weights'
if not os.path.exists(logdir):
    os.mkdir(logdir)
output_model_file = os.path.join(logdir,
                                 "fashion_mnist_weights_2.h5")
'''
# 载入参数前要先建立模型结构
model = keras.models.Sequential([
    keras.layers.Flatten(input_shape=[28, 28]),
    keras.layers.Dense(300, activation='relu'),
    keras.layers.Dense(100, activation='relu'),
    keras.layers.Dense(10, activation='softmax')
])

model.compile(loss="sparse_categorical_crossentropy",
              optimizer = "sgd",
              metrics = ["accuracy"])
'''
model.load_weights(output_model_file) #载入参数
model.evaluate(x_test_scaled, y_test)