Wesley Maddox

Sorry for the imprecision, botorch main

This looks great! Yeah, I really struggled with input transforms with variational GPs (don't think the version in Botorch really handles them super well now) and had to place them...

You're correct that it probably shouldn't be passed / used, as the loss functions use the current pars after we return the callable.

Ugh, yeah, that looks like a bug. Let me put up a PR to fix (idk who has write access to this still however). For the second question, you shouldn't...

Hi Kirk, You're reading the plot incorrectly - beneath the blue line shows that both SGD and SWAG are overconfident in that situation (confidence > accuracy). With that being said,...

Just following up, I checked and we never seem to have run calibration on CIFAR 5+5, but it's not terribly surprising that both SGD and SWAG (somewhat less so) are...

I believe we sampled those randomly at one point, so it's a holdover from that. No, I haven't noticed that.

Hmm.. I'll have to look into it. Maybe @izmailovpavel can be of help?

Hi, Just a word of caution here -- we never could really get the segmentation code to reproduce the results in the original Tiramisu paper so I don't know what...

What was the command that you used?