在Keras中，有時候需要對模型進行序列化與反序列化。進行模型序列化時，會將模型結果與模型權重儲存在不同的檔案中，模型權重通常儲存在HDF5檔案中，模型的結構可以儲存在JSON或者YAML檔案中。後二者方法大同小異，這裡以JSON為例說明一下Keras模型的儲存與載入。

from sklearn import datasets
import numpy as np
from keras.models import Sequential
from keras.layers import Dense
from keras.utils import to_categorical
from keras.models import model_from_json

#匯入資料
dataset = datasets.load_iris()

x = dataset.data
Y = dataset.target

#將標籤資料轉換為分類編碼
Y_labels = to_categorical(Y, num_classes=3)

#設定隨機種子
seed = 7
np.random.seed(seed)

#構建模型函式
def create_model(optimizer = 'rmsprop', init = 'glorot_uniform'):
    #構建模型
    model = Sequential()
    model.add(Dense(units=4, activation='relu', input_dim=4, kernel_initializer=init))
    model.add(Dense(units=6, activation='relu', kernel_initializer=init))
    model.add(Dense(units=3, activation='softmax', kernel_initializer=init))

    #編譯模型
    model.compile(loss='categorical_crossentropy', optimizer=optimizer, metrics=['accuracy'])

    return model

#構建模型
model = create_model()
model.fit(x,Y_labels, epochs=200, batch_size=5, verbose=0)
scores = model.evaluate(x,Y_labels, verbose=0)
print('%s: %.2f%%' % (model.metrics_names[1], scores[1]*100))

#模型儲存JSON檔案
model_json = model.to_json()
with open(r'F:\Python\pycharm\keras_deeplearning\model\modle.json', 'w') as file:
    file.write(model_json)

#儲存模型權重值
model.save_weights('model.json.h5')

#從JSON檔案中載入模型
with open(r'F:\Python\pycharm\keras_deeplearning\model\modle.json', 'r') as file:
    model_json1 = file.read()

#載入模型
new_model = model_from_json(model_json1)
new_model.load_weights('model.json.h5')

#編譯模型
new_model.compile(loss='categorical_crossentropy', optimizer='rmsprop',metrics=['accuracy'])

#評估載入之後的模型
scores1 = new_model.evaluate(x,Y_labels,verbose=0)
print('%s: %.2f%%' % (model.metrics_names[1], scores1[1]*100))
 

通過載入模型的方式建立新的模型後，必須先編譯模型，然後使用載入後的模型對新資料進行預測。

這裡的JSON檔案內容如下：

{"class_name": "Sequential", "config": [{"class_name": "Dense", "config": {"name": "dense_1", "trainable": true, "batch_input_shape": [null, 4], "dtype": "float32", "units": 4, "activation": "relu", "use_bias": true, "kernel_initializer": {"class_name": "VarianceScaling", "config": {"scale": 1.0, "mode": "fan_avg", "distribution": "uniform", "seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}}, {"class_name": "Dense", "config": {"name": "dense_2", "trainable": true, "units": 6, "activation": "relu", "use_bias": true, "kernel_initializer": {"class_name": "VarianceScaling", "config": {"scale": 1.0, "mode": "fan_avg", "distribution": "uniform", "seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}}, {"class_name": "Dense", "config": {"name": "dense_3", "trainable": true, "units": 3, "activation": "softmax", "use_bias": true, "kernel_initializer": {"class_name": "VarianceScaling", "config": {"scale": 1.0, "mode": "fan_avg", "distribution": "uniform", "seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}}], "keras_version": "2.2.2", "backend": "tensorflow"}