Does it provide the LR scheduler 'ReduceOnPlateau'?
Model/Dataset/Scheduler description
If the loss does not decrease for a certain period in training stage, i want to adjust the learning rate automatically.
However, there does not seem to be a 'ReuceOnPlateau" in the lr_scheduler list provided by mmcv, mmdetection, mmrotate.
Maybe I couldn't find it ? If it is available, can you provide an example how it is used ?
Thanks in advance for your answer.
Open source status
- [ ] The model implementation is available
- [ ] The model weights are available.
Provide useful links for the implementation
No response
Hi @Jongmin1991, we have support ReduceOnPlateau in MMEngine, which you can directly use by
param_scheduler = [
dict(
type='ReduceOnPlateauLR',
monitor='loss',
rule='less',
patience=50),
]
Please refer: https://github.com/open-mmlab/mmengine/blob/a7d4b7c7428061c8b85360a0e4e07d95eb1d5261/mmengine/optim/scheduler/param_scheduler.py#L1289
@Jongmin1991 by any chance did you manage to get ReduceOnPlateauLR to work on the 'loss'? I have the following error here https://github.com/open-mmlab/mmrotate/issues/1002 Many thanks!