Help on Learning Rate Scheduler with CombinedDataLoader

[leighgable]

Hello All,
I have data from a number of sites, which I have to keep seperate during training, so I am using the lightning combined dataloader. I would like to use a learning rate scheduler, but since the validation loss is prefixed by the dataloader index,
I am not sure what to pass to optim.lr_scheduler.ReduceLROnPlateau for the monitor value. Can I hijack whatever callback is producing the validation loss with the path added to the name? I'm using the CSVLogger by the way. And the model is an LSTM/Autoencoder.

Any help would be greatly appreciated.