Early stopping on nested metrics

[rubencart]

Hey, as suggested in the wandb docs, I would like to log metrics like

self.log('val', {'acc': 0.9})
# or
self.log_dict({'val': {'acc': 0.9}})

And then be able to access them for e.g. early stopping like

callbacks = [
    EarlyStopping('val.Corpus_F1', patience=opt.early_stop, mode="max")
]
trainer = pl.Trainer(callbacks=callbacks)

Is something similar possible? I keep getting this error.
pytorch_lightning.utilities.exceptions.MisconfigurationException: ModelCheckpoint(monitor='val.Corpus_F1') not found in the returned metrics: ['train', 'val'].