ReduceLROnPlateau

[nbgundavarapu]

ReduceLROnPlateau by default assumes a "min" metric (https://pytorch.org/docs/stable/generated/torch.optim.lr_scheduler.ReduceLROnPlateau.html)

mode ([str](https://docs.python.org/3/library/stdtypes.html#str)) – One of min, max. In min mode, lr will be reduced when the quantity monitored has stopped decreasing; in max mode it will be reduced when the quantity monitored has stopped increasing. Default: ‘min’.

In LVU tasks it used for accuracy metric, but the mode is not changed. Is this expected?

ViS4mer/run_lvu.py

Line 94 in 2a2442b

scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer, patience=patience, factor=0.2)

ViS4mer/run_lvu.py

Line 345 in 2a2442b

scheduler.step(val_acc)

[md-mohaiminul]

Hi,
That's a very good observation. I guess it was not intended. You can try changing the mode of ReduceLROnPlateau to 'max'.
Thanks