|
2 | 2 |
|
3 | 3 | The release log for BoTorch.
|
4 | 4 |
|
| 5 | +## [0.6.0] - Dec 8, 2021 |
| 6 | + |
| 7 | +#### Compatibility |
| 8 | +* Require PyTorch >=1.9 (#1011). |
| 9 | +* Require GPyTorch >=1.6 (#1011). |
| 10 | + |
| 11 | +#### New Features |
| 12 | +* New `ApproximateGPyTorchModel` wrapper for various (variational) approximate GP models (#1012). |
| 13 | +* New `SingleTaskVariationalGP` stochastic variational Gaussian Process model (#1012). |
| 14 | +* Support for Multi-Output Risk Measures (#906, #965). |
| 15 | +* Introduce `ModelList` and `PosteriorList` (#829). |
| 16 | +* New Constraint Active Search tutorial (#1010). |
| 17 | +* Add additional multi-objective optimization test problems (#958). |
| 18 | + |
| 19 | +#### Other Changes |
| 20 | +* Add `covar_module` as an optional input of `MultiTaskGP` models (#941). |
| 21 | +* Add `min_range` argument to `Normalize` transform to prevent division by zero (#931). |
| 22 | +* Add initialization heuristic for acquisition function optimization that samples around best points (#987). |
| 23 | +* Update initialization heuristic to perturb a subset of the dimensions of the best points if the dimension is > 20 (#988). |
| 24 | +* Modify `apply_constraints` utility to work with multi-output objectives (#994). |
| 25 | +* Short-cut `t_batch_mode_transform` decorator on non-tensor inputs (#991). |
| 26 | + |
| 27 | +#### Performance Improvements |
| 28 | +* Use lazy covariance matrix in `BatchedMultiOutputGPyTorchModel.posterior` (#976). |
| 29 | +* Fast low-rank Cholesky updates for `qNoisyExpectedHypervolumeImprovement` (#747, #995, #996). |
| 30 | + |
| 31 | +#### Bug Fixes |
| 32 | +* Update error handling to new PyTorch linear algebra messages (#940). |
| 33 | +* Avoid test failures on Ampere devices (#944). |
| 34 | +* Fixes to the `Griewank` test function (#972). |
| 35 | +* Handle empty base_sample_shape in `Posterior.rsample` (#986). |
| 36 | +* Handle `NotPSDError` and hitting `maxiter` in `fit_gpytorch_model` (#1007). |
| 37 | +* Use TransformedPosterior for subclasses of GPyTorchPosterior (#983). |
| 38 | +* Propagate `best_f` argument to `qProbabilityOfImprovement` in input constructors (f5a5f8b6dc20413e67c6234e31783ac340797a8d). |
| 39 | + |
| 40 | + |
5 | 41 | ## [0.5.1] - Sep 2, 2021
|
6 | 42 |
|
7 | 43 | #### Compatibility
|
@@ -35,6 +71,7 @@ The release log for BoTorch.
|
35 | 71 | * Support negative weights for minimization objectives in `get_chebyshev_scalarization` (#884).
|
36 | 72 | * Move `train_inputs` transforms to `model.train/eval` calls (#894).
|
37 | 73 |
|
| 74 | + |
38 | 75 | ## [0.5.0] - Jun 29, 2021
|
39 | 76 |
|
40 | 77 | #### Compatibility
|
|
0 commit comments