The model we are training can be found at /home/harrison/prod/AZ-Go. We are not developing in this directory, but pulling relevant commits from GitHub as major updates are made.
I have stored previous models at /home/harrison/logs, and these have been logged in Training Logs if you wish to look at previous training sessions.
I would like to have a collaborative directory to store all of this rather than in my own directory, but currently we (the students) do not have the ability to modify any upper level directories.
When working on the model, you can always pull the most recent version of the model from GitHub. Depending on your comfort level with git, it may be worthwhile to focus on creating/merging branches when working on big features.
https://github.com/leathalman/AZ-Go
You may run into a few different errors while training; I will try to document them as best I can.
When the previous training session ends in an error (for whatever reason), the training process may stop in the middle of an iteration (without completing it). If you try to load the previous model and continue training, you may get this error:

To combat this, navigate to AZ-Go/logs/go/CNN_MCTS_SimModified_checkpoint/your_board_size. This directory holds the checkpoint and examples files after each training session. Remove the most recent checkpoints (from the incomplete iteration), and rerun the model.
Please commit changes as you complete feature requests in the Development Progress page, so all team members can access your improvements.