Closed kevinyamauchi closed 1 year ago
not 100% how this improves logging but happy for you to merge if you're happy!
@alisterburt, sorry for the vague description. Basically, I just added a step that logs the summary statistics of the validation loss and metric (i.e., mean over all batches). When we attach a tensorboard logger, these will be viewable as realtime plots there.
this PR adds some improved logging to the UNet by writing to the LightningModule logger the summary statistics of the validation step (both metric and loss).