Use StratifiedStandardize for per-task Y standardization in TL (#5194)#5194
Open
hvarfner wants to merge 4 commits into
Open
Use StratifiedStandardize for per-task Y standardization in TL (#5194)#5194hvarfner wants to merge 4 commits into
hvarfner wants to merge 4 commits into
Conversation
meta-cla
Bot
added
the
CLA Signed
|
@hvarfner has exported this pull request. If you are a Meta employee, you can view the originating Diff in D102197139. |
hvarfner
pushed a commit
to hvarfner/Ax
that referenced
this pull request
Apr 29, 2026
…ook#5194) Summary: Adds per-task outcome standardization to the transfer learning adapter, ensuring each task's observations are standardized independently rather than jointly. Updates the default transform pipeline to use TL-specific outcome transforms. This removes ambiguity on whether the right transforms have been applied (e.g. QuickBO/warm-starting), where standardization is not performed across, but within experiments. Differential Revision: D102197139
hvarfner
force-pushed
the
export-D102197139
branch
from
8046f90 to
407478c
Compare
meta-codesync
Bot
changed the title
hvarfner
pushed a commit
to hvarfner/Ax
that referenced
this pull request
Apr 29, 2026
…ook#5194) Summary: Adds per-task outcome standardization to the transfer learning adapter, ensuring each task's observations are standardized independently rather than jointly. Updates the default transform pipeline to use TL-specific outcome transforms. This removes ambiguity on whether the right transforms have been applied (e.g. QuickBO/warm-starting), where standardization is not performed across, but within experiments. Differential Revision: D102197139
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #5194 +/- ##
==========================================
+ Coverage 96.38% 96.61% +0.23%
==========================================
Files 617 617
Lines 69579 69638 +59
==========================================
+ Hits 67065 67284 +219
+ Misses 2514 2354 -160 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
added 4 commits
May 12, 2026 13:29
…TaskGP (facebook#5192) Summary: X-link: meta-pytorch/botorch#3296 Automatically configures learned feature imputation for models that pad heterogeneous per-task data to the full joint feature space. Models with native heterogeneity support are excluded from this automatic configuration. Reviewed By: saitcakmak Differential Revision: D101841497
facebook#5193) Summary: Switches the default heterogeneous transfer learning model from a specialized per-task kernel model to a standard multi-task GP with learned feature imputation. The previous default model class is marked as deprecated. Reviewed By: saitcakmak Differential Revision: D102197137
…acebook#5200) Summary: Adds a `data_parameters` argument to `TorchAdapter._get_fit_args` that decouples SSD construction (model params) from data column extraction (target params). This lets the TL adapter set `_model_space` to include source-only RangeParameters directly, so the SSD naturally covers the full joint feature space -- eliminating the need for the `_expand_ssd_to_joint_space` post-hoc expansion. Overrides `_set_search_space` to add source-only RangeParameters from the joint search space to `_model_space` while preserving target bounds for shared params (Normalize stays anchored to target bounds). At gen time, `self.parameters` is temporarily swapped to target-only so `extract_search_space_digest` sees only params present in the gen-time search space. Deletes `_expand_ssd_to_joint_space` (~90 lines). Differential Revision: D104702983
…ook#5194) Summary: Adds per-task outcome standardization to the transfer learning adapter, ensuring each task's observations are standardized independently rather than jointly. Updates the default transform pipeline to use TL-specific outcome transforms. This removes ambiguity on whether the right transforms have been applied (e.g. QuickBO/warm-starting), where standardization is not performed across, but within experiments. Differential Revision: D102197139
hvarfner
force-pushed
the
export-D102197139
branch
from
407478c to
e407627
Compare
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Summary:
Adds per-task outcome standardization to the transfer learning adapter, ensuring each task's observations are standardized independently rather than jointly. Updates the default transform pipeline to use TL-specific outcome transforms.
This removes ambiguity on whether the right transforms have been applied (e.g. QuickBO/warm-starting), where standardization is not performed across, but within experiments.
Differential Revision: D102197139