python保存模型_python保存和加载模型

一、机器学习模型机器学习模型使用pickle来保存和加载

# 引用

import pickle

# 读取模型

with open('./model/xgb2.pkl', 'rb') as f:

xgb1 = pickle.load(f)

# 使用模型

xgb1.fit(...）

# 保存模型

with open('./model/xgb3.pkl', 'wb') as f:

pickle.dump(xgb1, f)

二、深度学习模型使用tensorflow的save()方法保存和加载模型

import tensorflow as tf

# 训练

...

# 创建Saver()节点

saver = tf.train.Saver()

with tf.Session() as sess:

# 恢复sess

ckpt = tf.train.get_checkpoint_state('./ckpt/')

if ckpt and ckpt.model_checkpoint_path:

saver.restore(sess, ckpt.model_checkpoint_path)

else:

sess.run(tf.global_variables_initializer())

...

save_path = saver.save(sess, "./ckpt/my_model_final.ckpt")