Restore-trained-weights-after-network-modification

Any experimental setup code to demonstrate how you can restore saved checkpoint after altering network graph and resume training.
[TensorFlow v1.x]

Usage:
1. Run python original_model.py to train the base model for few epochs and create a checkpoint.
2. Run modified network in the given python files to load the checkpoint created above and resume training.

For more details: https://www.divakar-verma.com/post/restoring-trained-weights-after-network-modification

Base Network model

Case 1 & 2
Removing nodes from network. See Case_1-2.py

Case 3
Adding non-trainable variables to the network. See Case_3.py

Case 4
Adding trainable variables to the network. See Case_4.py

Feel free to play around with the code and to raise a pull request.