Elizabeth Santorella

Support initializing matrices with Patsy?

3

I think we've discussed this, but I don't remember the conclusion and can't find an issue now. We recommend `from_pandas` as the way "most users" should construct tabmat objects. `from_pandas`...

Batch benchmark replications

4

Summary: Currently, benchmark methods have an attribute `distribute_replications` that indicates whether replications should be run fully in parallel or serially. This diff allows for batching the replications by replacing the...

Store floats exactly in Pandas DataFrames; require exact equality for Pandas DataFrames to be considered equal

8

Pandas `to_json` by default stores 10 decimal places of floats. This leads to precision loss. The fix is not completely trivial, because storing many more decimal places would not guarantee...

Remove unused 'reference_point' from MultiObjectiveBenchmarkProblem

2

Summary: As title Differential Revision: D47383874

Don't allow unused keyword arguments in Model.construct_inputs

8

Summary: BoTorch changes to go with the subsequent Ax diff. This will eventually need to be landed with this diff going before the Ax diff. Differential Revision: D53086323

[Bug] `sample_all_priors` doesn't work with `KroneckerMultiTaskGP`

5

# 🐛 Bug Moving this from #1323 . ## To reproduce ** Code snippet to reproduce ** ```python import torch from botorch import fit_gpytorch_model from botorch.models.multitask import KroneckerMultiTaskGP from botorch.models.transforms.outcome...

Making a new issue from #1391 . The qNEI tutorial should mention * Even when there is noise, qNEI won't outperform EI in 100% of cases, but rather does better...

[Bug] [relatively easy fix] `gen_one_shot_kg_initial_conditions` should error when provided inter-point constraints

1

# 🐛 Bug See https://github.com/pytorch/botorch/issues/1220 for context. `optimize_acqf` does not support inter-point constraints -- those whose `indices` are two-dimensional -- when the acquisition function is `qKnowledgeGradient`. This is because `gen_one_shot_kg_initial_conditions`...

[Feature Request] InputDataWarnings are are confusing

4

## Motivation It's common to see warnings like this: ```python import torch from botorch.models import SingleTaskGP x = torch.linspace(-5, 10, 100).unsqueeze(-1) model = SingleTaskGP(train_X=x, train_Y=x) ``` ``` /botorch/models/utils/assorted.py:173: InputDataWarning: Input...

Pull out mixin for setting `X_pending` in acquisition functions

2

Summary: Formerly, all subclasses of `Acquisition` had a method `set_X_pending`, even if they did not support pending points. This confused me while writing tests for optimizing acquisition functions. - Pulled...