handle ObservationFeatures without trial_index in Relativize #2441
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
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D57126153 |
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #2441 +/- ##
========================================
Coverage 95.31% 95.32%
========================================
Files 495 497 +2
Lines 48069 48194 +125
========================================
+ Hits 45818 45940 +122
- Misses 2251 2254 +3 ☔ View full report in Codecov by Sentry. |
sdaulton
added a commit
to sdaulton/Ax-1
that referenced
this pull request
May 8, 2024
…k#2441) Summary: This adds support for untransforming observation features in Relativize that do not have a trial index. This can occur when the Modelbridge does not use TrialAsTask and there are multiple SQ observations (e.g. for different trials) and we try to untransform the predictions for arms in a GR. Differential Revision: D57126153
|
This pull request was exported from Phabricator. Differential Revision: D57126153 |
sdaulton
added a commit
to sdaulton/Ax-1
that referenced
this pull request
May 9, 2024
…k#2441) Summary: This adds support for untransforming observation features in Relativize that do not have a trial index. This can occur when the Modelbridge does not use TrialAsTask and there are multiple SQ observations (e.g. for different trials) and we try to untransform the predictions for arms in a GR. Reviewed By: bernardbeckerman Differential Revision: D57126153
|
This pull request was exported from Phabricator. Differential Revision: D57126153 |
Summary: This implements a transform for adding `start_time` and `duration` as features for modeling. Currently, this adds them as `RangeParameter`s (to unblock time-sensitive applications), but in the future it would be good to revise this with a better treatment of non-tunable contextual information. `duration` appears to lead to better model fits on the synthetic example (notebook) than using `end_time`. This also works better than using the midpoint between start time and end time. Reviewed By: bernardbeckerman, Balandat Differential Revision: D57082939
Summary: see title Reviewed By: bernardbeckerman Differential Revision: D57119551
Summary: This adds support for handling pending observations that were generated by a Modelbridge that did not use the TimeAsFeature transform and therefore does not have start_time and end_time set in the observation features. Reviewed By: bernardbeckerman Differential Revision: D57121601
…k#2441) Summary: This adds support for untransforming observation features in Relativize that do not have a trial index. This can occur when the Modelbridge does not use TrialAsTask and there are multiple SQ observations (e.g. for different trials) and we try to untransform the predictions for arms in a GR. Reviewed By: bernardbeckerman Differential Revision: D57126153
|
This pull request was exported from Phabricator. Differential Revision: D57126153 |
|
This pull request has been merged in 9a624a2. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in 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.
Summary: This adds support for untransforming observation features in Relativize that do not have a trial index. This can occur when the Modelbridge does not use TrialAsTask and there are multiple SQ observations (e.g. for different trials) and we try to untransform the predictions for arms in a GR.
Differential Revision: D57126153