gpytorch
gpytorch copied to clipboard
A highly efficient implementation of Gaussian Processes in PyTorch
Hi, I try to reproduce the GPflow example ([https://gpflow.github.io/GPflow/2.5.2/notebooks/advanced/heteroskedastic.html](https://gpflow.github.io/GPflow/2.5.2/notebooks/advanced/heteroskedastic.html)) of heteroskedastic GP with an integration of gpytorch and pyro. Although the model fits the data well,  the model parameters...
# 🐛 Bug Using Deep Kernel Learning with the `InducingPointKernel` produces a non-positive-definite covariance matrix on test data (not the training data). The issue appears after the DNN weights +...
Greetings Devs and Community! I am trying to setup a basic multi-input multi-output variational GP (essentially modifying the Mulit-output Deep GP example) with 2 inputs and 2 outputs. In this...
Hi, I figured that the discussions are not getting much attention here, and since there are already similar issues (#1388 , #1353), I decided to open another one. ### Discussed...
I noticed that the index kernel's prior is [registered without a setting closure](https://github.com/cornellius-gp/gpytorch/blob/271f53dd06f22d01f3f94df19d913b10f7ee8290/gpytorch/kernels/index_kernel.py#L71). This means when sampling from the prior, the value of the parameter won't be changed. Was this...
# 🐛 Bug I am not sure this is a bug or something I just made a mistake. I do notice there was a post (issue #1129) very similar to...
# 📚 Documentation/Examples ** Is documentation wrong? ** The link to the PeriodicKernel reference (http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.81.1927&rep=rep1&type=pdf) is broken.  https://docs.gpytorch.ai/en/v1.5.1/kernels.html#gpytorch.kernels.PeriodicKernel
I'm encountering issues with CG Tolerance when using KISS-GP on a SpectralMixtureKernel to interpret (quasi-)periodic time variability in astronomical data. This warning only appears for some values of the mixture...
# 🐛 Bug I am finding that including ``with gpytorch.settings.fast_computations(covar_root_decomposition=False, log_prob=False, solves=False):`` unexpectedly **improves runtime** by 5x (and produces different MLL value). I will provide the full reproducible code at...
The LazyTensor functionality is moving to a separate repository (GitHub.com/cornellius-gp/linear_operator). This PR deprecates LazyTensor and instead uses the LinearOperator implementation. (Note that this PR does not change any GPyTorch code;...