reduce_lr_on_plateau_scheduler doesn't work

[yinruiqing]

Environment info

• Python version: 3.8
• PyTorch version (GPU?): 1.7.1
• pytorch-lightning version: 1.4

Information

Thank you for this wonderful project. Please take a look of optimizer and optimizer_connector in pytorch lightning. The scheduler["reduce_on_plateau"] is always false and the monitor (val_loss) cannot be passed to step method. You can run any script with reduce_lr_on_plateau_scheduler or warm_up_reduce_lr_on_plateau_scheduler to reproduce this bug.

[sooftware]

Thanks for letting us know. We'll check it. @hasangchun