tensorflow loss summary with respect to iteration time

Question

Is there a way to logging tensorflow loss (along training iterations) with respect to both iteration number and time elapsed during training. Then in tensorboard, we can see two curves for the training loss?

Answer 1

TensorBoard has three options for choosing the X-axis display, which can be found in the left panel on the scalar dashboard. They are "step", "wall time" and "relative time". You can change the display to relative or wall if you want to get a chronological view of the data.