Max Balandat

Sorry for the delay, busy times... > We want to use this probability directly as an outcome constraint in an optimization. To do so, one has to somehow invert the...

I think so? I guess my first idea here would be to just extract the latent (un-sigmoidified) model prediction and pass that in as the outcome together with the transformed...

@jackliu333 what makes you think that `fit_gpytorch_model` (now deprecated and superseded by `fit_gpytorch_mll`) does not work when the mll lives on the GPU? Are you getting errors when trying to...

Moving this to a new issue: #1566

Yeah I think we still want to do this, just correctly :) I think we should be able to use a slightly different reformulation of the problem to avoid the...

Not sure this is "fixed" rather than being avoided - I think the underlying issue is still the long-standing gpytorch issue https://github.com/cornellius-gp/gpytorch/issues/1853. But should be ok to close this BoTorch...

Thanks for the thorough investigation! Given the result, should we close out the issue as it doesn't seem to be a botorch or gpytorch issue and also doesn't seem to...

> Using in-place operations instead of creating and destroying small tensors. We'll have to be careful about doing that when on a path that should be differentiable - autograd doesn't...

Hmm yeah not obvious what is going on here; Note that this is not the "same" problem from the perspective of the code; under the hood we are fitting either...

@saitcakmak You have looked into the determinism of candidate generation in the past - do you happen to know what might cause this? No need to go deep if nothing...