Hi,
I noticed that you are using Botorch to compute the hypervolume:
|
def compute_hypervolume(pareto_front: torch.Tensor, ref_point: torch.Tensor) -> float: |
Botorch implementation is notoriously slow (10000x slower sometimes). You can see a comparison with other Python packages here: https://multi-objective.github.io/moocore/python/#exact-computation-of-hypervolume
Hi,
I noticed that you are using Botorch to compute the hypervolume:
boat/boat/bayesopt/loop/utils.py
Line 7 in 7ecad33
Botorch implementation is notoriously slow (10000x slower sometimes). You can see a comparison with other Python packages here: https://multi-objective.github.io/moocore/python/#exact-computation-of-hypervolume