Consider replace sklearn.metrics.roc_auc_scorewith torchmetrics.AUROC

#2
by 2catycm - opened

It is very slow to run the evaluation on CPU, if we have pytorch evalutaion, it should be prefered.
Here is an experiment of the comparison of CPU and GPU roc-auc speed on my dataset with 1 2080Ti GPU.

image.png

Sign up or log in to comment