Open Ianyliu opened 4 months ago
Hi!
Currently there is no CUDA support for the calculations, but there is a work in progress for heuristic algorithm calculation which will allow to derive a subset of irredundant systems for larger number of conditions. I will update you once it's available.
When attempting to run functions such as
get_irredundant_systems()
orget_irredundant_sums()
on datasets with >= 10 conditions, I almost always run into kernel crashes or out of memory errors. There is an option to limit the max_depth forget_irredundant_sums()
, but no such option forget_irredundant_systems()
to the best of my knowledge.Is there any possibility/plan of using GPU calculations to allow for causal discovery of larger number of conditions?