Open st-- opened 2 years ago
Ugh, yes, that looks like some sort of improper dimension based error. I'll try to take a closer look this weekend.
Does it work if you enforce the dimension?
Thanks for responding! Could you clarify what you mean by "enforcing the dimension" (what & where)?
(Just to clarify, I used GPyTorch==1.4.0 and botorch==0.4.0, as the code doesn't run with recent versions due to a bunch of breaking changes in those libraries.)
Ugh, yes, not surprising that it no longer works, especially as the core contribution (fantasization) here is merged into gpytorch for gaussian likelihoods, but not botorch (one day i'll publish that PR).
"enforcing the dimension"
off the top of my head, i mean this: "svgp model getting init_y.view(-1) whereas the other models get init_y.view(-1, 1)"
Well, changing the svgp constructor arguments from .view(-1)
to provide either .view(-1, 1)
or .view(1, -1)
instead still doesn't work... would you be able to fix the script ?
I pushed a fix and it ran locally for me. Sorry for the long delay there.
When running
as given in the README.md, I get the following exception:
Calling the experiment script with
--random
, or running--model=exact
(with or without--random
) runs fine.Might this have something to do with the
svgp
model gettinginit_y.view(-1)
whereas the other models getinit_y.view(-1, 1)
(similarly forinit_y_var
)?