Restoring checkpoints and training parameters

[k22jung]

• What are your command line arguments?:

  python3 train.py --num_epochs 10 --checkpoint_step 1 --validation_step 1 --dataset /path/to/dataset --crop_width 512 --crop_height 512 --batch_size 2 --num_val_images 100 --h_flip 1 --v_flip 0 --brightness 0.5 --rotation 10 --model Encoder-Decoder-Skip --frontend MobileNetV2 --continue_training 1 --epoch_start_i 9

• Have you written any custom code?: No.
• What have you done to try and solve this issue?: Added a global step variable, but unsure what its impact is.
• TensorFlow version?: 1.10.1

Describe the problem

Describe the problem clearly here. Be sure to convey here why it's a bug or a feature request.

I wasn't entirely sure if this was addressed in the code already or not, so I just wanted to clarify a few things. When restoring a checkpoint and training from an epoch index that is non-zero, would the training parameters for the optimizer, including decayed learning rate and tensorflow's global_step also be restored correctly to the appropriate epoch that was left at from the checkpoint?