multiplier works weird about pytorch-gradual-warmup-lr HOT 2 OPEN

I can confirm this behavior. I think line 31 in warmup_scheduler/scheduler.py is troublesome, and that

return self.after_scheduler.get_last_lr()

should rather be:

return self.after_scheduler.get_lr()

I do however think the whole scheduler would be easier / less error-prone to implement using the built-in PyTorch LR scheduler LinearLR for the warmup part, optionally chained with one or more other schedulers (the equivalent of "after_scheduler") using SequentialLR.

from pytorch-gradual-warmup-lr.

Just to nuance my comment: For some reason it actually seems users are not supposed to call the .get_lr() function. It generates a warning message if called from elsewhere than .step(), in which case this is indicated by a with _enable_get_lr_call(self): statement.

from pytorch-gradual-warmup-lr.