The question is short: what does the 'relative' mean in tensorboard?
I understand that step correspond to the 'step' parameter in summary_writer.add_summary()
call, but the 'relative' change x axis to something I don't understand.
the 'wall' is probably for wall time.
Also, when using 'step', multiple runs will have zigzag plot, but that is another issue.
Thanks for you input.
With the relative-option the x-axis shows the wall time relative to the first data point. That typically means that it shows the number of hours since the training run was started.
The Scalars dashboard shows how the loss and metrics change with every epoch. You can use it to also track training speed, learning rate, and other scalar values.
Head over to localhost:6006 to see TensorBoard on your local machine. We can see some of the scalar metrics that are provided by default With the linear Classifier. We can also Expand and Zoom into any of these graphs. Double-clicking allows us to zoom out.
With the relative-option the x-axis shows the wall time relative to the first data point. That typically means that it shows the number of hours since the training run was started.
This option is for example very useful if you want to compare the performance of two or more different training runs that weren't started simultaneously.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With