What's Changed

• Ensure LinaerKernel stores ard_num_dims property. by @gpleiss in #2635
• Export "constraints" submodule from package root by @chrisyeh96 in #2656
• [Docs] Fixes #2569: Incorrect Matern docs by @samjenner1 in #2657
• Loosen numeric tolerances in certain unit tests by @esantorella in #2661
• Replace deprecated np.Inf with np.inf by @saitcakmak in #2663
• Implement Hadamard Gaussian noise by @TobyBoyne in #2481
• Update precommit by @sdaulton in #2665
• Supporting passing full train inputs, rather than only supporting task indices to in HadamardGaussianLikelihood by @sdaulton in #2664
• RFF/linear prediction strategy fix by @gpleiss in #2671

New Contributors

• @samjenner1 made their first contribution in #2657
• @TobyBoyne made their first contribution in #2481

Full Changelog: v1.14...v1.14.1

What's Changed

• DOC: fix a typo, Hyperparamters -> Hyperparameters by @partev in #2588
• Allow for dimension-specific variance in linear kernel by @sdaulton in #2593
• Fix bug #2606: Setting mixing_weights=False in SoftmaxLikelihood still adds the learnable parameter W by @DiegoFreitasH in #2607
• Add typing annotations to gpytorch.Module by @chrisyeh96 in #2608
• Improve typing annotations for linear kernel by @chrisyeh96 in #2613
• Remove unused imports in GPyTorch Fully Bayesian example by @chrisyeh96 in #2612
• gpytorch.module: fix typing annotation by @chrisyeh96 in #2611
• Remove deprecated lazy tensor by @saitcakmak in #2615
• Support broadcasting batch shapes in MTMVN.from_independent_mvns by @saitcakmak in #2621
• Fix typo in Pyro_GPyTorch_High_Level.ipynb by @williamjsdavis in #2574
• Fix error in IndexKernel docstring by @parrangoiz in #2625
• Implement MVN.unsqueeze by @saitcakmak in #2624
• Update MVN.expand() to support non-lazy MVN & reuse scale_tril where possible by @saitcakmak in #2623
• Unpin jaxtyping by @saitcakmak in #2626
• Detach new_covar_cache to enable JIT tracing of models after fantasization by @SaiAakash in #2605
• DOC: fix typos in likelihood.py by @chrisyeh96 in #2486
• sign fix for gradient blocks by @m-julian in #2598
• Start adding typing annotations to ExactGP by @chrisyeh96 in #2436
• Hotfix: fix .diagonal() calls for keops kernel matrices. by @gpleiss in #2590
• Version bump (Python 3.10) by @gpleiss in #2627

New Contributors

• @DiegoFreitasH made their first contribution in #2607
• @williamjsdavis made their first contribution in #2574
• @parrangoiz made their first contribution in #2625
• @SaiAakash made their first contribution in #2605

Full Changelog: v1.13...v1.14

What's Changed

• main and develop branches by @gpleiss in #2542
• Include jaxtyping to allow for Tensor/LinearOperator typehints with sizes. by @gpleiss in #2543
• fix: replace deprecated scipy.integrate.cumtrapz with cumulative_trapezoid by @natsukium in #2545
• use common notation for normal distribution N(\mu, \sigma^2) by @partev in #2547
• fix broken link in Simple_GP_Regression.ipynb by @partev in #2546
• Deprecate last_dim_is_batch (bump PyTorch version to >= 2.0) by @gpleiss in #2549
• Added ability for priors of transformed distributions to have their p… by @hvarfner in #2551
• Avoid unnecessary memory allocation for covariance downdate in SGPR prediction strategy by @JonathanWenger in #2559
• Fix VNNGP with batches by @LuhuanWu in #2375
• fix a typo by @partev in #2570
• fix a typo by @partev in #2571
• DOC: improve the formatting in the documentation by @partev in #2578

New Contributors

• @natsukium made their first contribution in #2545
• @hvarfner made their first contribution in #2551

Full Changelog: v1.12...v1.13

What's Changed

• Minor patch to Matern covariances by @j-wilson in #2378
• Fix error messages for ApproximateGP.get_fantasy_model by @gpleiss in #2374
• Fix lazy kernel slicing when there are multiple outputs by @douglas-boubert in #2376
• Fix training status of noise model of HeteroskedasticNoise after exceptions by @fjzzq2002 in #2382
• Stop rbf_kernel_grad and rbf_kernel_gradgrad creating the full covariance matrix unnecessarily by @douglas-boubert in #2388
• Likelihood bugfix by @gpleiss in #2395
• Update RTD configuration, and linear_operator requirement. by @gpleiss in #2399
• Better support for missing labels by @Turakar in #2288
• Fix latex of gradients in docs by @jlperla in #2404
• Skip the warning in gpytorch.lazy.__getattr__ if name starts with _ by @saitcakmak in #2423
• Fix KeOps regressions from #2296. by @gpleiss in #2413
• Update index.rst by @mkomod in #2449
• python should also be a runtime dependency by @jaimergp in #2457
• fix a typo: cannonical -> canonical by @partev in #2461
• Update distributions.rst by @chrisyeh96 in #2487
• Fix flaky SVGP classification test by @gpleiss in #2495
• DOC: Fix typo in docstring. by @johanneskopton in #2493
• fix a typo by @partev in #2464
• DOC: fix formatting issue in RFFKernel documentation by @partev in #2463
• DOC: fix broken formatting in leave_one_out_pseudo_likelihood.py by @partev in #2462
• ConstantKernel by @SebastianAment in #2511
• DOC: fix broken URL in periodic_kernel.py by @partev in #2513
• Bug: Exploit Structure in get_fantasy_strategy by @naefjo in #2494
• Matern52 grad by @m-julian in #2512
• Added optional kwargs to ExactMarginalLogLikelihood call by @rafaol in #2522
• Corrected configuration of exclude statements in pre-commit configuration by @JonathanWenger in #2541

New Contributors

• @douglas-boubert made their first contribution in #2376
• @fjzzq2002 made their first contribution in #2382
• @jlperla made their first contribution in #2404
• @mkomod made their first contribution in #2449
• @jaimergp made their first contribution in #2457
• @partev made their first contribution in #2461
• @chrisyeh96 made their first contribution in #2487
• @johanneskopton made their first contribution in #2493
• @naefjo made their first contribution in #2494
• @rafaol made their first contribution in #2522

Full Changelog: v1.11...v1.12

What's Changed

• Fix solve_triangular(Tensor, LinearOperator) not supported in VNNGP by @Turakar in #2323
• Metrics fixes and cleanup by @JonathanWenger in #2325
• Lock down doc requirements to prevent RTD failures. by @gpleiss in #2339
• Fix typos in multivariate_normal.py by @manuelhaussmann in #2331
• add Hamming IMQ kernel by @samuelstanton in #2327
• Use torch.cdist for dist by @esantorella in #2336
• Enable fantasy models for multitask GPs Reborn by @yyexela in #2317
• Clean up deprecation warnings by @saitcakmak in #2348
• More informative string representation of MultitaskMultivariateNormal distributions. by @gpleiss in #2333
• Mean and kernel functions for first and second derivatives by @ankushaggarwal in #2235
• Bugfix: double added log noise prior by @LuisAugenstein in #2355
• Remove Module.getattr by @saitcakmak in #2359
• Remove num_outputs from IndependentModelList by @saitcakmak in #2360
• keops periodic and keops kernels unit tests by @m-julian in #2296
• Deprecate checkpointing by @gpleiss in #2361

New Contributors

• @Turakar made their first contribution in #2323
• @manuelhaussmann made their first contribution in #2331
• @esantorella made their first contribution in #2336
• @yyexela made their first contribution in #2317
• @ankushaggarwal made their first contribution in #2235
• @LuisAugenstein made their first contribution in #2355

Full Changelog: v1.10...v1.11

What's Changed

• Re-add pyro + torch_master check by @gpleiss in #2241
• Fix silently ignored arguments in IndependentModelList by @saitcakmak in #2249
• fix bug in nearest_neighbor_variational_strategy by @LuhuanWu in #2243
• Move infinite interval bounds check into Interval constructor by @Balandat in #2259
• Use ufmt for code formatting and import sorting by @Balandat in #2262
• Update nearest_neighbors.py by @yw5aj in #2267
• Use raw strings to avoid "DeprecationWarning: invalid escape sequence" by @saitcakmak in #2282
• Fix handling of re-used priors by @Balandat in #2269
• Fix BernoulliLikelihood documentation by @gpleiss in #2285
• gpytorch.settings.variational_cholesky_jitter can be set dynamically. by @gpleiss in #2255
• Likelihood docs update by @gpleiss in #2292
• Improve development/contributing documentation by @gpleiss in #2293
• Use raw strings to avoid "DeprecationWarning: invalid escape sequence" by @saitcakmak in #2295
• Update SGPR notebook by @gpleiss in #2303
• Update linear operator dependency to 0.4.0 by @gpleiss in #2321

New Contributors

• @yw5aj made their first contribution in #2267

Full Changelog: v1.9.1...v1.10

What's Changed

• Fix LMCVariationalStrategy example in docs by @adamjstewart in #2112
• Accept closure argument in NGD optimizer step by @dannyfriar in #2118
• Fix bug with Multitask DeepGP predictive variances. by @gpleiss in #2123
• Autogenerate parameter types in documentation from python typehints by @gpleiss in #2125
• Retiring deprecated versions ofpsd_safe_cholesky, NotPSDError, and assert_allclose by @SebastianAment in #2130
• fix custom dtype_value_context setting by @sdaulton in #2132
• Include linear operator in installation instructions by @saitcakmak in #2131
• Fixes HalfCauchyPrior by @feynmanliang in #2137
• Fix return type of Kernel.covar_dist by @Balandat in #2138
• Change variable name for better understanding by @findoctorlin in #2135
• Expose jitter by @hughsalimbeni in #2136
• Add HalfNormal prior distribution for non-negative variables. by @ZitongZhou in #2147
• Fix multitask/added_loss_term bugs in SGPR regression by @gpleiss in #2121
• fix bugs in test half Cauchy prior. by @ZitongZhou in #2156
• Generalize RandomModule by @feynmanliang in #2164
• MMVN.to_data_independent_dist returns correct variance for non-interleaved MMVN distributions. by @gpleiss in #2172
• Update MSLL in metrics.py by @jongwonKim-1997 in #2177
• Update multitask example notebook by @gpleiss in #2190
• Fix exception message for missing kernel lazy kernel attribute by @dannyfriar in #2195
• Improving _sq_dist when x1_eq_x2 by @SebastianAment in #2204
• Fix docs/requirements.txt by @gpleiss in #2206
• As per issue '#2175 [Docs] GP Regression With Uncertain Inputs'. by @corwinjoy in #2200
• Avoid evaluating kernel when adding jitter by @gpleiss in #2189
• Avoid evaluating kernel in expand_batch by @dannyfriar in #2185
• Deprecating postprocess by @SebastianAment in #2205
• Make PiecewisePolynomialKernel GPU compatible by @gpleiss in #2217
• Let LazyEvaluatedKernelTensor recall the grad state at instantiation by @SebastianAment in #2229
• Doc Update for Posterior Model Distribution and Posterior Predictive Distribution by @varunagrawal in #2230
• Fix 08_Advanced_Usage links by @st-- in #2240
• Add device property to Kernels, add unit tests by @Balandat in #2234
• pass **kwargs to ApproximateGP.call in DeepGPLayer by @IdanAchituve in #2224

New Contributors

• @dannyfriar made their first contribution in #2118
• @SebastianAment made their first contribution in #2130
• @feynmanliang made their first contribution in #2137
• @findoctorlin made their first contribution in #2135
• @hughsalimbeni made their first contribution in #2136
• @ZitongZhou made their first contribution in #2147
• @jongwonKim-1997 made their first contribution in #2177
• @corwinjoy made their first contribution in #2200
• @varunagrawal made their first contribution in #2230
• @st-- made their first contribution in #2240
• @IdanAchituve made their first contribution in #2224

Full Changelog: v1.9.0...v1.9.1

Starting with this release, the LazyTensor functionality of GPyTorch has been pulled out into its own separate Python package, called linear_operator. Most users won't notice the difference (at the moment), but power users will notice a few changes.

If you have your own custom LazyTensor code, don't worry: this release is backwards compatible! However, you'll see a lot of annoying deprecation warnings 😄

LazyTensor -> LinearOperator

• All gpytorch.lazy.*LazyTensor classes now live in the linear_operator repo, and are now called linear_operator.operator.*LinearOperator.
  • For example, gpytorch.lazy.DiagLazyTensor is now linear_operator.operators.DiagLinearOperator
  • The only major naming change: NonLazyTensor is now DenseLinearOperator
• gpytorch.lazify and gpytorch.delazify are now linear_operator.to_linear_operator and linear_operator.to_dense, respectively.
• The _quad_form_derivative method has been renamed to _bilinear_derivative (a more accurate name!)
• LinearOperator method names now reflect their corresponding PyTorch names. This includes:
  • add_diag -> add_diagonal
  • diag -> diagonal
  • inv_matmul -> solve
  • symeig -> eigh and eigvalsh
• LinearOperator now has the mT property

torch_function functionality

LinearOperators are now compatible with the torch api! For example, the following code works:

diag_linear_op = linear_operator.operators.DiagLinearOperator(torch.randn(10))
torch.matmul(diag_linear_op, torch.randn(10, 2))  # returns a torch.Tensor!

Other files that have moved:

• gpytorch.functions - all of the core functions used by LazyTensors now live in the LinearOperator repo. This includes: diagonalization, dsmm, inv_quad, inv_quad_logdet, matmul, pivoted_cholesky, root_decomposition, solve (formally inv_matmul), and sqrt_inv_matmul
• gpytorch.utils - a few have moved to the LinearOperator repo. This includes: broadcasting, cholesky, contour_intergral_quad, getitem, interpolation, lanczos, linear_cg, minres, permutation, stable_pinverse, qr, sparse, SothcasticLQ, and toeplitz.

Full Changelog: v1.8.1...v1.9.0

Bug fixes

• MultitaskMultivariateNormal: fix tensor reshape issue by @adamjstewart in #2081
• Fix handling of prior terms in ExactMarginalLogLikelihood by @saitcakmak in #2039
• Fix bug in preconditioned KISS-GP / Hadamard Multitask GPs by @gpleiss in #2090
• Add constant_constraint to ConstantMean by @gpleiss in #2082

New Contributors

• @mone27 made their first contribution in #2076

Full Changelog: v1.8.0...v1.8.1

Major Features

• add variational nearest neighbor GP by @LuhuanWu in #2026

New Contributors

• @adamjstewart made their first contribution in #2061
• @m-julian made their first contribution in #2054
• @ngam made their first contribution in #2059
• @LuhuanWu made their first contribution in #2026

Full Changelog: v1.7.0...v1.8.0