Log the ROC or PRC curve of the best model when using EarlyStopping

Hi everyone, what would be the best way to log the ROC curve (or any list of metrics really) of the best model when training using EarlyStopping? For now, I just use the test method at the end, passing the validation data loader, but I was wondering if there was a more elegant way to do that

trainer.test(test_dataloaders=dm.val_dataloader(), ckpt_path='best')

trainer.validate() will be available in hopefully next release: https://github.com/PyTorchLightning/pytorch-lightning/pull/4945

1 Like