Tensorflow1.x save and load model

This repository is built to learn how tensorflow 1.x load models.

Environment

• python 3.6
• tensorflow-gpu 1.13

Usage

run mnist_train.py to get trained models.

test_with_network.py tests by rebuilding the network.

test_with_meta.py tests by loading .meta.

And there is a wrong use in incorrect_usage.py，which both rebuilding the network and loading .meta.

use Tensorboard to view network structure， tensorboard --logdir=./log/test_with_meta.

Result

Select the top 5000 MNIST datasets for testing.

loading mode	accuracy
test_with_network	0.9784
test_with_meta	0.9784
incorrect_usage	0.0982