Add batch_shape property to GP model class #2307
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Balandat
added a commit
to Balandat/botorch
that referenced
this pull request
Apr 19, 2023
This enables the use of `SingleTaskVariationalGP` with certain botorch features (e.g. with entropy-based acquistion functions as requested in pytorch#1795). This is a bit of a band-aid, the proper thing to do here is to fix up the PR upstreaming this to gpytorch (cornellius-gp/gpytorch#2307) to enable support for `batch_shape` on all approximate gpytorch models, and then just call that on the `model` in `ApproximateGPyTorchModel`.
facebook-github-bot
pushed a commit
to pytorch/botorch
that referenced
this pull request
Apr 19, 2023
Summary: This enables the use of `SingleTaskVariationalGP` with certain botorch features (e.g. with entropy-based acquistion functions as requested in #1795). This is a bit of a band-aid, the proper thing to do here is to fix up the PR upstreaming this to gpytorch (cornellius-gp/gpytorch#2307) to enable support for `batch_shape` on all approximate gpytorch models, and then just call that on the `model` in `ApproximateGPyTorchModel`. Pull Request resolved: #1799 Reviewed By: dme65 Differential Revision: D45107761 Pulled By: Balandat fbshipit-source-id: 44f1893d4ab915f744c17dfd5da4efd4923c66ae
Implements #2301. TODO: Verify compatibility with the botorch setup of other models
gpleiss
reviewed
Jul 6, 2023
| " \n", | ||
| " # Locations Z_{d} corresponding to u_{d}, they can be randomly initialized or \n", | ||
| " # regularly placed with shape (D x n_inducing x latent_dim).\n", | ||
| " self.inducing_inputs = torch.randn(data_dim, n_inducing, latent_dim)\n", | ||
| " \n", | ||
| " # Sparse Variational Formulation (inducing variables initialised as randn)\n", | ||
| " q_u = CholeskyVariationalDistribution(n_inducing, batch_shape=self.batch_shape) \n", | ||
| " q_u = CholeskyVariationalDistribution(n_inducing, batch_shape=torch.Size([data_dim])) \n", |
There was a problem hiding this comment.
Why aren't we using the self.batch_shape property here?
|
Let's hold off. I have been running into some issues with this approach (caused by the fact that in botorch the batch shape is not necessarily the same as the gpytorch-internal batch shape. So right now this would cause a bunch of headaches and bugs on our end rather than only resolving some. We'll have to rethink if this is the right approach. |
# for free
to join this conversation on GitHub.
Already have an account?
# to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Implements #2301.