Bug fix: Make MBM acquisition pass the correct args to each BoTorch optimizer; cleanup #2571
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: D59354709 |
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #2571 +/- ##
=======================================
Coverage 95.22% 95.23%
=======================================
Files 489 489
Lines 47612 47661 +49
=======================================
+ Hits 45338 45389 +51
+ Misses 2274 2272 -2 ☔ View full report in Codecov by Sentry. |
|
This pull request was exported from Phabricator. Differential Revision: D59354709 |
esantorella
force-pushed
the
export-D59354709
branch
from
b461ad3
to
9a1b70f
Compare
esantorella
added a commit
to esantorella/Ax
that referenced
this pull request
Jul 9, 2024
…ptimizer; cleanup (facebook#2571) Summary: Pull Request resolved: facebook#2571 Motivation: MBM can dispatch to four different BoTorch optimizers depending on the search space. Currently, `optimize_acqf_discrete_local_search` is failing because it is passed the inappropriate argument `sequential`. This diff * Makes the logic more clear: First, Acquisition.optimize determines which of the four optimizers is appropriate. then it constructs arguments based on that optimizer. Then it constructs arguments for the optimizers using `optimizer_argparse`. Then it does any optimizer_specific logic and calls the optimizer. * Fixes optimizer_argparse so that inappropriate arguments such as `sequential` are not passed to optimizers they don't apply to. * Removes special-casing for qNEHVI and qMES in optimizer_argparse that wasn't actually doing anything * Extends unit tests for `optimizer_argparse` to check all optimizers * Reduces the usage and scope of mocks in test_acquisition so that `optimize_acqf` and its variants are actually run as much as possible. Differential Revision: D59354709
esantorella
added a commit
to esantorella/Ax
that referenced
this pull request
Jul 9, 2024
…ptimizer; cleanup (facebook#2571) Summary: Pull Request resolved: facebook#2571 Motivation: MBM can dispatch to four different BoTorch optimizers depending on the search space. Currently, `optimize_acqf_discrete_local_search` is failing because it is passed the inappropriate argument `sequential`. This diff * Makes the logic more clear: First, Acquisition.optimize determines which of the four optimizers is appropriate. then it constructs arguments based on that optimizer. Then it constructs arguments for the optimizers using `optimizer_argparse`. Then it does any optimizer_specific logic and calls the optimizer. * Fixes optimizer_argparse so that inappropriate arguments such as `sequential` are not passed to optimizers they don't apply to. * Removes special-casing for qNEHVI and qMES in optimizer_argparse that wasn't actually doing anything * Extends unit tests for `optimizer_argparse` to check all optimizers * Reduces the usage and scope of mocks in test_acquisition so that `optimize_acqf` and its variants are actually run as much as possible. Differential Revision: D59354709
|
This pull request was exported from Phabricator. Differential Revision: D59354709 |
esantorella
added a commit
to esantorella/Ax
that referenced
this pull request
Jul 9, 2024
…ptimizer; cleanup (facebook#2571) Summary: Pull Request resolved: facebook#2571 Motivation: MBM can dispatch to four different BoTorch optimizers depending on the search space. Currently, `optimize_acqf_discrete_local_search` is failing because it is passed the inappropriate argument `sequential`. This diff * Makes the logic more clear: First, Acquisition.optimize determines which of the four optimizers is appropriate. then it constructs arguments based on that optimizer. Then it constructs arguments for the optimizers using `optimizer_argparse`. Then it does any optimizer_specific logic and calls the optimizer. * Fixes optimizer_argparse so that inappropriate arguments such as `sequential` are not passed to optimizers they don't apply to. * Removes special-casing for qNEHVI and qMES in optimizer_argparse that wasn't actually doing anything * Extends unit tests for `optimizer_argparse` to check all optimizers * Reduces the usage and scope of mocks in test_acquisition so that `optimize_acqf` and its variants are actually run as much as possible. Differential Revision: D59354709
esantorella
force-pushed
the
export-D59354709
branch
from
9a1b70f
to
6b0ad55
Compare
…ptimizer; cleanup (facebook#2571) Summary: Pull Request resolved: facebook#2571 # Context The current flow is 1. MBM `Acquisition.optimize` constructs arguments passed to BoTorch optimizers in `optimizer_argparse` without knowing which of the four optimizers it will be passing the arguments to. It has only a boolean flag indicating whether the optimizer is discrete. 2. `Acquisition.optimize` decides which of the four optimizers to use and does optimizer specific-logic (these two parts are not really sequential). 3. `Acquisition.optimize` calls a BoTorch optimizer, passing some arguments that may not apply to the optimizer that actually got used. 4. Prior to pytorch/botorch#2390, the inappropriate arguments would be silently ignored, but now they raise an exception. MBM can dispatch to four different BoTorch optimizers depending on the search space. Currently, `optimize_acqf_discrete_local_search` is failing because it is passed the inappropriate argument `sequential`. # This diff * Makes the flow within `Acquisition.optimize` more clear and changes `optimizer_argparse` so that inappropriate arguments such as `sequential` are not passed to optimizers they don't apply to: 1. `Acquisition.optimize` determines which of the four optimizers is appropriate. 2. `Acquisition.optimize` constructs arguments based on that optimizer, only constructing needed arguments. 3. Then it does any optimizer-specific logic. 4. Then it calls a BoTorch optimizer; there is no longer an error because only appropriate arguments were passed. * Extends unit tests for `optimizer_argparse` to check all optimizers * Reduces the usage and scope of mocks in test_acquisition so that `optimize_acqf` and its variants are actually run as much as possible. Differential Revision: D59354709
|
This pull request was exported from Phabricator. Differential Revision: D59354709 |
esantorella
force-pushed
the
export-D59354709
branch
from
6b0ad55
to
8a00600
Compare
|
This pull request has been merged in 1a6faa6. |
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:
Motivation: MBM can dispatch to four different BoTorch optimizers depending on the search space. Currently,
optimize_acqf_discrete_local_searchis failing because it is passed the inappropriate argumentsequential.This diff
optimizer_argparse. Then it does any optimizer_specific logic and calls the optimizer.sequentialare not passed to optimizers they don't apply to.optimizer_argparseto check all optimizersoptimize_acqfand its variants are actually run as much as possible.Differential Revision: D59354709