Zhong Yuan Lai

I've been trying to implement this in the Service API: [Ax Issue 2514](https://github.com/facebook/Ax/issues/2514). Might be useful here (if I can verify its correctness!)

In the end I was able to make the following work (for multiobjectives): ``` from botorch.test_functions.multi_objective_multi_fidelity import MOMFBraninCurrin import torch tkwargs = { # Tkwargs is a dictionary contaning data...

@saitcakmak Dear Sait, thanks a lot for the information! I have some follow-ups regarding the points you raised: > Using some mocks, I see that the MFKG acquisition function is...

@sgbaird If you are interested in updates

@saitcakmak Thanks a million！I'm keeping this open for now as I plan to come back and put up a working repro later, at which point I'll close it; in the...

> > Hence I was under the impression that if I leave these out the acqf automatically reverts to a non-MF one. Is this not the case? > > The...

@saitcakmak @esantorella Hi guys, so I went ahead and tried to set up the input constructor as so: ``` @acqf_input_constructor(qHypervolumeKnowledgeGradient) def construct_inputs_qHVKG( model: Model, training_data: MaybeDict[SupervisedDataset], objective_thresholds: Tensor, bounds: List[Tuple[float,...

> Hi @Abrikosoff, > > I am guessing that @Balandat intended to the constraint model to be pretrained outside of Ax. If you wanted to use non-linear constraints with scipy,...

> Hi @Abrikosoff, > > I am guessing that @Balandat intended to the constraint model to be pretrained outside of Ax. If you wanted to use non-linear constraints with scipy,...

Let me test this out @saitcakmak , thanks a lot!