TensorBoard

For convenience, the project natively supports TensorBoard for monitoring the training progress. To use TensorBoard, run using the --tensorboard flag, i.e.

$ python -m run.adversarial.train --train --tensorboard

The output TensorBoard is published to http://localhost:6006 on the running server. If the code is run through ssh, it is still possible to access the results locally, by doing

$ ssh <user>@<host> -L 16006:127.0.0.1:6006

and navigating to http://localhost:16006 on the local machine. The file outputs from running with TensorBoard callbacks are stored in the logs/ directory of the project, and running TensorBoard manually is possible by doing

$ tensorboard --logdir logs/<timestamp>

Notice that TensorBoard requires using the TensorFlow backend. (This might not be strictly true, but it's asserted nonetheless.)