Fit_one_cycle(,....,max-lr=slice(1e-5,1e-3)) how does this part work?

Hi there,

I am trying to get my lr correct.
By passing slice(1e-5,1e-3) to max_lr, my guess is:
Each epoch the learner will pick up a Learning Rate starting from 1e-5 until 1e-3,
and the LR spreads throughout every epoch.

Is my understanding correct?

If not, please help to correct my understanding.

If it’s correct, then why do we need
ReduceLROnPlateauCallback