Warm-up technique no longer works as documented

[murnanedaniel]

Since the update to > v1.0, this technique (and the instructions in the docs) breaks. Specifically, I get TypeError: optimizer_step() got an unexpected keyword argument 'epoch', since the optimizer_step() call has changed to self.optimizer_step(optimizer, opt_idx, batch_idx, train_step_and_backward_closure). I can see how to re-organise optimizer, opt_idx and batch_idx, but what should one do with train_step_and_backward_closure? Default it to None?

Originally posted by @murnanedaniel in #2934 (comment)

[github-actions]

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

[murnanedaniel]

#4455
I see that this issue will be fixed with new documentation. Specifically, changing current_epoch to epoch in the optimizer_step override, as well as changing optimizer.step() to optimizer.step(optimizer_closure).