tensorflow 只恢复部分模型参数的实例

技术  /  管理员 发布于 7年前   429

我就废话不多说了，直接上代码吧！

import tensorflow as tfdef model_1():  with tf.variable_scope("var_a"):    a = tf.Variable(initial_value=[1, 2, 3], name="a")  vars = [var for var in tf.trainable_variables() if var.name.startswith("var_a")]  print(len(vars))  return varsdef model_2():  vars1 = model_1()  with tf.variable_scope("var_b"):    a = tf.Variable(initial_value=[1, 2, 3], name="a")  vars2 = [var for var in tf.trainable_variables() if var.name.startswith("var")]  print(len(vars2))  return vars1def pretrain_model1():  print("-------- model 1 ------")  vars = model_1()  with tf.Session() as sess:    sess.run(tf.global_variables_initializer())    saver = tf.train.Saver()    saver.save(sess, "./model.ckpt")def train_model2():  print("-------- model 2 ------")  model1_vars = model_2()  with tf.Session() as sess:    sess.run(tf.global_variables_initializer())    saver = tf.train.Saver(var_list=model1_vars)    saver.restore(sess, "./model.ckpt")    vars = sess.run([model1_vars])    for var in vars:      print(var)step = 2if step == 1:  pretrain_model1()else:  train_model2()

以上这篇tensorflow 只恢复部分模型参数的实例就是小编分享给大家的全部内容了，希望能给大家一个参考，也希望大家多多支持。