Max Balandat
Max Balandat
Hi, thanks for your question. This has most likely to do with the fact that gpytorch is switching from using Cholesky-based solves to using Linear Conjugate Gradient solves by default...
The only issue is that things may get slow, but other than that this shouldn’t be an issue.
What GPyTorch version do you have installed?
I meant the gpytorch version: ``` import gpytorch print(gpytorch.__version__) ```
Would you mind providing a full (non-)working example so we can debug from there? From a high level your approach seems to make sense but I'm worried how you're just...
That seems reasonable to me in general. One thing to note is that `gen_candidates_torch` in its current form is quite basic and may not work very well compared to `gen_candidates_scipy`....
> a problem in 100+ dimensions I'm curious, are you using any dimensionality reduction techniques here (DKL, random projections...)? Seems like a vanilla GP won't work very well in this...
> You are right since #207 it is hard to contrive an example where gen_candidates_scipy performs worse/slower/whatever! Great. Thanks also for the example. One thing to note is that the...
wrong reference in the PR above, sorry for the confusion
In light of this, it would be really nice to have (i) forward differentiation in pytorch (https://github.com/pytorch/pytorch/issues/10223) so we can efficiently implement Hessian computations (ii) the ability to directly return...