Carl Hvarfner
Carl Hvarfner
@mfeurer Great, thanks for the clarification! Do you know if this can be disabled? I'm perfectly happy hacking around in the source code if needed.
Okay, thanks! I'll look at it on my end and let you know if I find a simple way around it.
@sdaulton for sure! I currently observe similar things for JES, but I'm not sure whether the found points are actually higher in acquisition function value or not (for either LogEI...
I have not quite figured out why the test coverage is not there, since I thought I addressed it today. I will also figure out the conflicts ASAP!
Thanks! I'm trying to sort something out on my end too, but it may be a bit patchy if one is to avoid GPyTorch.
So, I have a suggested solution to this problem that seems to be relatively pain-free. However, it doesn't quite fit within the BOTorch ecosystem, as it involves duplicating the training...
That is the way that I have understood things. The FBGP doesn't have batch dimension in the inputs, since there's only one set of training data. The length- and outputscales...
Not quite - the same issue still persists! It seems like the forward pass through the GP https://github.com/cornellius-gp/gpytorch/blob/batch_shape/gpytorch/models/exact_gp.py#L229 outputs a `batch_shape x batch_shape x num_training_data` which causes the aforementioned `view(...)`...