How to improve prediction speed?

[Leafaeolian]

Report

Thanks a lot for this impressive tool.

When I follow the ComboSciPlex tutorial, I notice that predicting only 7 conditions already takes about 9 minutes. This makes benchmarking on large-scale datasets (e.g., genome-wide perturbation datasets) quite time-consuming.

I was wondering if there are any recommended strategies to speed up the prediction process?

Additionally, is there any plan to include early stopping in the framework to make benchmarking more fair with other work?

Thanks for your help!

Version information

No response