First of all, thanks for this interesting alternative to Tensorboard.
One feature I am still missing is the ability to set a display_interval per plot line instead of per graph. In my training runs, I usually run the validation steps less frequently than training steps. It seems the display_interval however, is over actual samples given to Losswise, not over the iteration number passed in the x parameter.
So, for example, if I pass validation samples at x values (100, 200, 300, ...), and set display_interval to 100, then the validation loss plot will only update in x value intervals of 10000 instead of 100, which is longer than I would like. Lowering display_interval makes the training samples too noisy, since I pass training samples at x values (1, 2, 3, ...). So the training and validation values are effectively averaged over largely different interval sizes in x (although over the same raw number of samples).
My current work-around of putting training and validation in different plots with different display_intervals is not ideal either, because x-axis of the plots is never quite synchronized.
One way this could be solved would be to allow an option to average over same intervals in x instead of same number of raw samples, or an alternative could be to allow setting different display_intervals for different plot lines in a Graph.
First of all, thanks for this interesting alternative to Tensorboard.
One feature I am still missing is the ability to set a
display_intervalper plot line instead of per graph. In my training runs, I usually run the validation steps less frequently than training steps. It seems thedisplay_intervalhowever, is over actual samples given to Losswise, not over the iteration number passed in thexparameter.So, for example, if I pass validation samples at
xvalues(100, 200, 300, ...), and setdisplay_intervalto100, then the validation loss plot will only update inxvalue intervals of10000instead of100, which is longer than I would like. Loweringdisplay_intervalmakes the training samples too noisy, since I pass training samples atxvalues(1, 2, 3, ...). So the training and validation values are effectively averaged over largely different interval sizes inx(although over the same raw number of samples).My current work-around of putting training and validation in different plots with different
display_intervals is not ideal either, because x-axis of the plots is never quite synchronized.One way this could be solved would be to allow an option to average over same intervals in
xinstead of same number of raw samples, or an alternative could be to allow setting differentdisplay_intervals for different plot lines in aGraph.