Closed krisrs1128 closed 4 years ago
Right now, the losses we see are weighted according to lambda_gan and lambda_L. This makes comparison across runs tricky.
Assignment:
Do not remove the current loss display. That's still useful for understanding the optimization.
We can actually already see the terms in the losses in wandb. Just need to use the interface there.
Right now, the losses we see are weighted according to lambda_gan and lambda_L. This makes comparison across runs tricky.
Assignment:
Do not remove the current loss display. That's still useful for understanding the optimization.