Pytorch-lightning: stepwise learning rate scheduler

[Selimonder]

Hello,

I am trying to manage LR scheduling with my training with pL. Main methods are relying on epoch-wise updates on LR, is there a way to reduce this to step-wise?

• OS: [Linux]
• Packaging [pip]
• Version [1.0.4]

[github-actions[bot]]

Hi! thanks for your contribution!, great first issue!

[ayasyrev]

https://pytorch-lightning.readthedocs.io/en/latest/optimizers.html?highlight=scheduller
for example:

def configure_optimizers(self):
   optimizer = Adam(...)
   scheduler = {'scheduler': OneCycleLR(optimizer, ...), 
                        'interval': 'step'}
    return [optimizer], [scheduller] 

[s-rog]

I'll close this for now, feel free to re-open if you have more questions.

Also check out our forums: https://forums.pytorchlightning.ai/