import tensorflow as tf
import numpy as np

W = tf.Variable(np.arange(6).reshape((2, 3)), dtype=tf.float32, name="weights")
b = tf.Variable(np.arange(3).reshape((1, 3)), dtype=tf.float32, name="biases")
 
saver = tf.train.Saver()
with tf.Session() as sess:  
    saver.restore(sess, "niu/save_net.ckpt") 
    print("weights:", sess.run(W))
    print("biases:", sess.run(b))

TF：利用TF的train.Saver載入曾經訓練好的variables(W、b)以供預測新的數據