Remove deprecated FixedNoiseGP #2536

Balandat · 2024-09-14T14:27:04Z

Summary: This model has been deprecated for a while and replaced by SingleTaskGP.

Differential Revision: D62680578

facebook-github-bot · 2024-09-14T14:27:27Z

This pull request was exported from Phabricator. Differential Revision: D62680578

codecov · 2024-09-14T14:36:13Z

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.98%. Comparing base (c895a8d) to head (750d7f8).
Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2536      +/-   ##
==========================================
- Coverage   99.98%   99.98%   -0.01%     
==========================================
  Files         193      193              
  Lines       17000    17001       +1     
==========================================
  Hits        16998    16998              
- Misses          2        3       +1

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

…2532) Summary: Pull Request resolved: pytorch#2532 Makes models which had their priors updated in pytorch#2507 use the `Standardize` outcome transform by default, mimicking pytorch#2458 Also removes some deprecated functionality in the process, namely the `data_fidelity` argument to `SingleTaskMultiFidelityGP`, as well as the `FixedNoiseMultiFidelityGP` and `FixedNoiseLCEMGP` models. Differential Revision: D62552307 Reviewed By: saitcakmak, esantorella

facebook-github-bot · 2024-09-14T16:06:47Z

This pull request was exported from Phabricator. Differential Revision: D62680578

Summary: Pull Request resolved: pytorch#2536 This model has been deprecated for a while and replaced by `SingleTaskGP`. Differential Revision: D62680578

facebook-github-bot · 2024-09-14T16:11:08Z

This pull request was exported from Phabricator. Differential Revision: D62680578

esantorella · 2024-09-15T18:49:38Z

botorch/models/approximate_gp.py

+            outcome_transform: An outcome transform that is applied to the training
+                data during instantiation and to the posterior during inference.
+                NOTE: If this model is trained in minibatches, an outcome transform
+                with learnable parameters (such as `Standardize`) would update its
+                parameters for each minibatch, which is undesirable. If you do intend
+                to train in minibatches, we recommend you not use an outcome transform
+                and instead pre-transform your whole data set before fitting the model.


Thanks for adding these docstrings!

Good to be explicit about defaults now that many models have Standardize by default.

Suggested change

outcome_transform: An outcome transform that is applied to the training

data during instantiation and to the posterior during inference.

NOTE: If this model is trained in minibatches, an outcome transform

with learnable parameters (such as `Standardize`) would update its

parameters for each minibatch, which is undesirable. If you do intend

to train in minibatches, we recommend you not use an outcome transform

and instead pre-transform your whole data set before fitting the model.

outcome_transform: An outcome transform that is applied to the training

data during instantiation and to the posterior during inference. By

default, no outcome transform is used. It is not recommended to use

an outcome transform if this model is trained in minibatches, because

an outcome transform

with learnable parameters (such as `Standardize`) would update its

parameters for each minibatch, which is undesirable. If you do intend

to train in minibatches, we recommend you not use an outcome transform

and instead pre-transform your whole data set before fitting the model.

esantorella · 2024-09-15T18:56:41Z

botorch/models/contextual_multioutput.py

@@ -51,8 +51,8 @@ def __init__(
        embs_dim_list: Optional[list[int]] = None,
        output_tasks: Optional[list[int]] = None,
        all_tasks: Optional[list[int]] = None,
+        outcome_transform: OutcomeTransform | _DefaultType | None = DEFAULT,


Suggested change

outcome_transform: OutcomeTransform | _DefaultType | None = DEFAULT,

*,

outcome_transform: OutcomeTransform | _DefaultType | None = DEFAULT,

esantorella · 2024-09-15T18:57:41Z

botorch/models/contextual_multioutput.py

@@ -247,62 +249,3 @@ def construct_inputs(
        if embs_dim_list is not None:
            base_inputs["embs_dim_list"] = embs_dim_list
        return base_inputs
-
-
-class FixedNoiseLCEMGP(LCEMGP):


Let's mention this in the summary or title

esantorella · 2024-09-15T18:58:03Z

botorch/models/fully_bayesian_multitask.py

@@ -254,6 +255,11 @@ def __init__(
            task_feature=task_feature,
            output_tasks=output_tasks,
            rank=rank,
+            # We already transformed the data above, this avoids applying the


esantorella · 2024-09-15T19:07:42Z

botorch/models/gp_regression_fidelity.py

-                    training data during instantiation and to the posterior during
-                    inference (that is, the `Posterior` obtained by calling
-                    `.posterior` on the model will be on the original scale).
+                training data during instantiation and to the posterior during


thanks for fixing up all this documentation!

botorch/models/gp_regression_fidelity.py

esantorella · 2024-09-15T19:08:32Z

botorch/models/gp_regression_mixed.py

@@ -65,7 +66,7 @@ def __init__(
            Callable[[torch.Size, int, list[int]], Kernel]
        ] = None,
        likelihood: Optional[Likelihood] = None,
-        outcome_transform: Optional[OutcomeTransform] = None,  # TODO
+        outcome_transform: Optional[Union[OutcomeTransform, _DefaultType]] = DEFAULT,


Suggested change

outcome_transform: Optional[Union[OutcomeTransform, _DefaultType]] = DEFAULT,

outcome_transform: OutcomeTransform | _DefaultType | None = DEFAULT,

botorch/models/multitask.py

esantorella · 2024-09-15T19:12:05Z

botorch/models/multitask.py

+                Pass down `None` to use no outcome transform. NOTE: Standardization
+                should be applied in a stratified fashion, separately for each task.


Can you clarify how one would apply standardization in a stratified fashion, perhaps with an example? Does the default get it right, or should the user manually standardize their data, or do they need to specify the outcome_transform in a certain way?

esantorella · 2024-09-15T19:14:57Z

botorch/utils/test_helpers.py

+    if m > 1:
+        if not isinstance(model, BatchedMultiOutputGPyTorchModel):
+            raise UnsupportedError(
+                "`get_pvar_expected` only supports `BatchedMultiOutputGPyTorchModel`s."


Suggested change

"`get_pvar_expected` only supports `BatchedMultiOutputGPyTorchModel`s."

"`get_pvar_expected` only supports `BatchedMultiOutputGPyTorchModel`s when m>1."

facebook-github-bot · 2024-09-16T02:11:09Z

This pull request has been merged in 18eb95a.

facebook-github-bot added the CLA Signed Do not delete this pull request or issue due to inactivity. label Sep 14, 2024

facebook-github-bot added the fb-exported label Sep 14, 2024

Balandat force-pushed the export-D62680578 branch from 0f862e4 to 4cafd4f Compare September 14, 2024 16:06

Remove deprecated FixedNoiseGP (pytorch#2536)

750d7f8

Summary: Pull Request resolved: pytorch#2536 This model has been deprecated for a while and replaced by `SingleTaskGP`. Differential Revision: D62680578

Balandat force-pushed the export-D62680578 branch from 4cafd4f to 750d7f8 Compare September 14, 2024 16:11

esantorella reviewed Sep 15, 2024

View reviewed changes

botorch/models/gp_regression_fidelity.py Show resolved Hide resolved

esantorella reviewed Sep 15, 2024

View reviewed changes

botorch/models/multitask.py Show resolved Hide resolved

esantorella reviewed Sep 15, 2024

View reviewed changes

facebook-github-bot closed this in 18eb95a Sep 16, 2024

facebook-github-bot added the Merged label Sep 16, 2024

ronald-jaepel mentioned this pull request Oct 30, 2024

ImportError: cannot import name 'FixedNoiseGP' from 'botorch.models.gp_regression' fau-advanced-separations/CADET-Process#174

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Remove deprecated FixedNoiseGP #2536

Remove deprecated FixedNoiseGP #2536

Balandat commented Sep 14, 2024

facebook-github-bot commented Sep 14, 2024

codecov bot commented Sep 14, 2024 •

edited

Loading

facebook-github-bot commented Sep 14, 2024

facebook-github-bot commented Sep 14, 2024

esantorella Sep 15, 2024

esantorella Sep 15, 2024

esantorella Sep 15, 2024

esantorella Sep 15, 2024

esantorella Sep 15, 2024

esantorella Sep 15, 2024

esantorella Sep 15, 2024

esantorella Sep 15, 2024

facebook-github-bot commented Sep 16, 2024

	outcome_transform: OutcomeTransform \| _DefaultType \| None = DEFAULT,
	*,
	outcome_transform: OutcomeTransform \| _DefaultType \| None = DEFAULT,

	outcome_transform: Optional[Union[OutcomeTransform, _DefaultType]] = DEFAULT,
	outcome_transform: OutcomeTransform \| _DefaultType \| None = DEFAULT,

		Pass down `None` to use no outcome transform. NOTE: Standardization
		should be applied in a stratified fashion, separately for each task.

	"`get_pvar_expected` only supports `BatchedMultiOutputGPyTorchModel`s."
	"`get_pvar_expected` only supports `BatchedMultiOutputGPyTorchModel`s when m>1."

Remove deprecated FixedNoiseGP #2536

Remove deprecated FixedNoiseGP #2536

Conversation

Balandat commented Sep 14, 2024

facebook-github-bot commented Sep 14, 2024

codecov bot commented Sep 14, 2024 • edited Loading

Codecov Report

facebook-github-bot commented Sep 14, 2024

facebook-github-bot commented Sep 14, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

facebook-github-bot commented Sep 16, 2024

codecov bot commented Sep 14, 2024 •

edited

Loading