Keras实践笔记9——模型的保存与恢复

1import numpy as np
2from keras import Sequential
3from keras.callbacks import TensorBoard
4from keras.layers import Dense
5from keras.models import load_model
6
7x = np.linspace(-10, 10, 300)
8y = 3 * x + np.random.random(x.shape) * 0.44
9
10model = Sequential()
11model.add(Dense(1, activation='linear', input_shape=(1,)))
12model.compile(optimizer='SGD', loss='mean_squared_error', metrics=['accuracy'])
13
14model.summary()
15
16model.fit(x, y, epochs=100, validation_split=0.3, verbose=2,
17          callbacks=[TensorBoard(log_dir='./logs', histogram_freq=1)])
18model.save('../models/linear_model.h5')
19
20predict_model = load_model('../models/linear_model.h5')
21
22predict_y = predict_model.predict([8])
23print(predict_y)

训练完后的模型可以保存下来，供下一次再训练或者拿来做predict，保存的时候只需要简单的save下就好

model.save('../models/linear_model.h5')

恢复的时候调用下load_model，就可以直接拿回来用了

predict_model = load_model('../models/linear_model.h5')