I’m looking for an elegant way to log multiline plots (e.g. showing AUROC and AUPRC on one plot) whilst training, and using the wandb.log() function. As far as I can tell, the only way is to keep track of the values from the start of training up until the current iteration, then log a wandb.plot.line (according to Weights & Biases), however this is a different interface from the standard use of wandb.log(), wherein you only give the latest value.
Is there some way to do what I’m looking for (multiline plots that update each iteration of training, and only need to be given the latest value for each variable)?