如何重写tensorflow的检查点文件？

import tensorflow as tf
import os

# 定义模型结构
input_size = 100
output_size = 10
hidden_size = 50

# 定义输入和输出占位符
inputs = tf.placeholder(tf.float32, [None, input_size])
targets = tf.placeholder(tf.float32, [None, output_size])

# 定义模型参数
weights = tf.Variable(tf.random_normal([input_size, hidden_size]))
biases = tf.Variable(tf.zeros([hidden_size]))
output_weights = tf.Variable(tf.random_normal([hidden_size, output_size]))
output_biases = tf.Variable(tf.zeros([output_size]))

# 定义前向传播过程
hidden_layer = tf.matmul(inputs, weights) + biases
hidden_layer = tf.nn.relu(hidden_layer)
output_layer = tf.matmul(hidden_layer, output_weights) + output_biases

# 定义损失函数和优化器
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=output_layer, labels=targets))
optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.1).minimize(loss)

# 创建会话并初始化变量
sess = tf.Session()
sess.run(tf.global_variables_initializer())

# 保存检查点文件
saver = tf.train.Saver()
save_path = os.path.join(os.getcwd(), "checkpoint.ckpt")
saver.save(sess, save_path)

# 重写检查点文件
new_weights = tf.Variable(tf.random_normal([input_size, hidden_size]))
new_biases = tf.Variable(tf.zeros([hidden_size]))
new_output_weights = tf.Variable(tf.random_normal([hidden_size, output_size]))
new_output_biases = tf.Variable(tf.zeros([output_size]))

# 重新定义前向传播过程
new_hidden_layer = tf.matmul(inputs, new_weights) + new_biases
new_hidden_layer = tf.nn.relu(new_hidden_layer)
new_output_layer = tf.matmul(new_hidden_layer, new_output_weights) + new_output_biases

# 加载原有的检查点文件
saver.restore(sess, save_path)

# 保存新的检查点文件
saver.save(sess, save_path)