Runner should only need to know about outcomes, not objectives vs. constraints #2963
+148
−202
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: D64919207 |
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #2963 +/- ##
=======================================
Coverage 95.65% 95.66%
=======================================
Files 486 486
Lines 48796 48772 -24
=======================================
- Hits 46676 46656 -20
+ Misses 2120 2116 -4 ☔ View full report in Codecov by Sentry. |
esantorella
added a commit
to esantorella/Ax
that referenced
this pull request
Oct 25, 2024
…nstraints (facebook#2963) Summary: Context: In theory, a `BenchmarkRunner` should not have to know what metrics are objectives or constraints, and a test function should not have to be aware of that, either. They are just generating data. A `BenchmarkProblem` should only store knowledge of objectives and constraints on the `OptimizationConfig`, so that various `OptimizationConfigs` can be used without changing the runner and test function. For historical reasons, runners track objectives and constraints separately and add noise to them separately, because this mimics how BoTorch test functions handle this. However, we now can and should isolate the quirks of BoTorch test functions to `BoTorchTestProblem`. This diff: * Updates `ParamBasedTestFunction.evaluate_true` to return all outcomes, not just objectives, and gets rid of `ParamBasedTestFunction.evaluate_true`, which was for constraints * Removes `num_objectives` from `ParamBasedTestProblem`, leaving `ParamBasedTestProblem` with nothing but an `evaluate_true` method * Removes the argument `constraint_noise_std` from `create_problem_from_botorch` and from `ParamBasedTestProblemRunner`, in favor of just using `noise_std`. * Updates argument validation Tangentially related changes: * For simplicity, makes `get_noise_stds` always return a dict * Stops allowing `noise_std` to be `None` and defaults it to zero (it was eventually set to zero when it was None in the past) Differential Revision: D64919207
esantorella
force-pushed
the
export-D64919207
branch
from
017fe54
to
58e3b5b
Compare
|
This pull request was exported from Phabricator. Differential Revision: D64919207 |
esantorella
added a commit
to esantorella/Ax
that referenced
this pull request
Oct 25, 2024
…nstraints (facebook#2963) Summary: Context: In theory, a `BenchmarkRunner` should not have to know what metrics are objectives or constraints, and a test function should not have to be aware of that, either. They are just generating data. A `BenchmarkProblem` should only store knowledge of objectives and constraints on the `OptimizationConfig`, so that various `OptimizationConfigs` can be used without changing the runner and test function. For historical reasons, runners track objectives and constraints separately and add noise to them separately, because this mimics how BoTorch test functions handle this. However, we now can and should isolate the quirks of BoTorch test functions to `BoTorchTestProblem`. This diff: * Updates `ParamBasedTestFunction.evaluate_true` to return all outcomes, not just objectives, and gets rid of `ParamBasedTestFunction.evaluate_true`, which was for constraints * Removes `num_objectives` from `ParamBasedTestProblem`, leaving `ParamBasedTestProblem` with nothing but an `evaluate_true` method * Removes the argument `constraint_noise_std` from `create_problem_from_botorch` and from `ParamBasedTestProblemRunner`, in favor of just using `noise_std`. * Updates argument validation Tangentially related changes: * For simplicity, makes `get_noise_stds` always return a dict * Stops allowing `noise_std` to be `None` and defaults it to zero (it was eventually set to zero when it was None in the past) Differential Revision: D64919207
esantorella
added a commit
to esantorella/Ax
that referenced
this pull request
Oct 25, 2024
…nstraints (facebook#2963) Summary: Context: In theory, a `BenchmarkRunner` should not have to know what metrics are objectives or constraints, and a test function should not have to be aware of that, either. They are just generating data. A `BenchmarkProblem` should only store knowledge of objectives and constraints on the `OptimizationConfig`, so that various `OptimizationConfigs` can be used without changing the runner and test function. For historical reasons, runners track objectives and constraints separately and add noise to them separately, because this mimics how BoTorch test functions handle this. However, we now can and should isolate the quirks of BoTorch test functions to `BoTorchTestProblem`. This diff: * Updates `ParamBasedTestFunction.evaluate_true` to return all outcomes, not just objectives, and gets rid of `ParamBasedTestFunction.evaluate_true`, which was for constraints * Removes `num_objectives` from `ParamBasedTestProblem`, leaving `ParamBasedTestProblem` with nothing but an `evaluate_true` method * Removes the argument `constraint_noise_std` from `create_problem_from_botorch` and from `ParamBasedTestProblemRunner`, in favor of just using `noise_std`. * Updates argument validation Tangentially related changes: * For simplicity, makes `get_noise_stds` always return a dict * Stops allowing `noise_std` to be `None` and defaults it to zero (it was eventually set to zero when it was None in the past) Differential Revision: D64919207
esantorella
added a commit
to esantorella/Ax
that referenced
this pull request
Oct 25, 2024
…nstraints (facebook#2963) Summary: Context: In theory, a `BenchmarkRunner` should not have to know what metrics are objectives or constraints, and a test function should not have to be aware of that, either. They are just generating data. A `BenchmarkProblem` should only store knowledge of objectives and constraints on the `OptimizationConfig`, so that various `OptimizationConfigs` can be used without changing the runner and test function. For historical reasons, runners track objectives and constraints separately and add noise to them separately, because this mimics how BoTorch test functions handle this. However, we now can and should isolate the quirks of BoTorch test functions to `BoTorchTestProblem`. This diff: * Updates `ParamBasedTestFunction.evaluate_true` to return all outcomes, not just objectives, and gets rid of `ParamBasedTestFunction.evaluate_true`, which was for constraints * Removes `num_objectives` from `ParamBasedTestProblem`, leaving `ParamBasedTestProblem` with nothing but an `evaluate_true` method * Removes the argument `constraint_noise_std` from `create_problem_from_botorch` and from `ParamBasedTestProblemRunner`, in favor of just using `noise_std`. * Updates argument validation Tangentially related changes: * For simplicity, makes `get_noise_stds` always return a dict * Stops allowing `noise_std` to be `None` and defaults it to zero (it was eventually set to zero when it was None in the past) Differential Revision: D64919207
esantorella
added a commit
to esantorella/Ax
that referenced
this pull request
Oct 25, 2024
…nstraints (facebook#2963) Summary: Pull Request resolved: facebook#2963 Context: In theory, a `BenchmarkRunner` should not have to know what metrics are objectives or constraints, and a test function should not have to be aware of that, either. They are just generating data. A `BenchmarkProblem` should only store knowledge of objectives and constraints on the `OptimizationConfig`, so that various `OptimizationConfigs` can be used without changing the runner and test function. For historical reasons, runners track objectives and constraints separately and add noise to them separately, because this mimics how BoTorch test functions handle this. However, we now can and should isolate the quirks of BoTorch test functions to `BoTorchTestProblem`. This diff: * Updates `ParamBasedTestFunction.evaluate_true` to return all outcomes, not just objectives, and gets rid of `ParamBasedTestFunction.evaluate_true`, which was for constraints * Removes `num_objectives` from `ParamBasedTestProblem`, leaving `ParamBasedTestProblem` with nothing but an `evaluate_true` method * Removes the argument `constraint_noise_std` from `create_problem_from_botorch` and from `ParamBasedTestProblemRunner`, in favor of just using `noise_std`. * Updates argument validation Tangentially related changes: * For simplicity, makes `get_noise_stds` always return a dict * Stops allowing `noise_std` to be `None` and defaults it to zero (it was eventually set to zero when it was None in the past) Differential Revision: D64919207
…nstraints (facebook#2963) Summary: Context: In theory, a `BenchmarkRunner` should not have to know what metrics are objectives or constraints, and a test function should not have to be aware of that, either. They are just generating data. A `BenchmarkProblem` should only store knowledge of objectives and constraints on the `OptimizationConfig`, so that various `OptimizationConfigs` can be used without changing the runner and test function. For historical reasons, runners track objectives and constraints separately and add noise to them separately, because this mimics how BoTorch test functions handle this. However, we now can and should isolate the quirks of BoTorch test functions to `BoTorchTestProblem`. This diff: * Updates `ParamBasedTestFunction.evaluate_true` to return all outcomes, not just objectives, and gets rid of `ParamBasedTestFunction.evaluate_true`, which was for constraints * Removes `num_objectives` from `ParamBasedTestProblem`, leaving `ParamBasedTestProblem` with nothing but an `evaluate_true` method * Removes the argument `constraint_noise_std` from `create_problem_from_botorch` and from `ParamBasedTestProblemRunner`, in favor of just using `noise_std`. * Updates argument validation Tangentially related changes: * For simplicity, makes `get_noise_stds` always return a dict * Stops allowing `noise_std` to be `None` and defaults it to zero (it was eventually set to zero when it was None in the past) Reviewed By: saitcakmak Differential Revision: D64919207
esantorella
added a commit
to esantorella/Ax
that referenced
this pull request
Oct 25, 2024
…nstraints (facebook#2963) Summary: Context: In theory, a `BenchmarkRunner` should not have to know what metrics are objectives or constraints, and a test function should not have to be aware of that, either. They are just generating data. A `BenchmarkProblem` should only store knowledge of objectives and constraints on the `OptimizationConfig`, so that various `OptimizationConfigs` can be used without changing the runner and test function. For historical reasons, runners track objectives and constraints separately and add noise to them separately, because this mimics how BoTorch test functions handle this. However, we now can and should isolate the quirks of BoTorch test functions to `BoTorchTestProblem`. This diff: * Updates `ParamBasedTestFunction.evaluate_true` to return all outcomes, not just objectives, and gets rid of `ParamBasedTestFunction.evaluate_true`, which was for constraints * Removes `num_objectives` from `ParamBasedTestProblem`, leaving `ParamBasedTestProblem` with nothing but an `evaluate_true` method * Removes the argument `constraint_noise_std` from `create_problem_from_botorch` and from `ParamBasedTestProblemRunner`, in favor of just using `noise_std`. * Updates argument validation Tangentially related changes: * For simplicity, makes `get_noise_stds` always return a dict * Stops allowing `noise_std` to be `None` and defaults it to zero (it was eventually set to zero when it was None in the past) Reviewed By: saitcakmak Differential Revision: D64919207
esantorella
force-pushed
the
export-D64919207
branch
from
58e3b5b
to
82ccf93
Compare
|
This pull request was exported from Phabricator. Differential Revision: D64919207 |
esantorella
added a commit
to esantorella/Ax
that referenced
this pull request
Oct 25, 2024
…nstraints (facebook#2963) Summary: Context: In theory, a `BenchmarkRunner` should not have to know what metrics are objectives or constraints, and a test function should not have to be aware of that, either. They are just generating data. A `BenchmarkProblem` should only store knowledge of objectives and constraints on the `OptimizationConfig`, so that various `OptimizationConfigs` can be used without changing the runner and test function. For historical reasons, runners track objectives and constraints separately and add noise to them separately, because this mimics how BoTorch test functions handle this. However, we now can and should isolate the quirks of BoTorch test functions to `BoTorchTestProblem`. This diff: * Updates `ParamBasedTestFunction.evaluate_true` to return all outcomes, not just objectives, and gets rid of `ParamBasedTestFunction.evaluate_true`, which was for constraints * Removes `num_objectives` from `ParamBasedTestProblem`, leaving `ParamBasedTestProblem` with nothing but an `evaluate_true` method * Removes the argument `constraint_noise_std` from `create_problem_from_botorch` and from `ParamBasedTestProblemRunner`, in favor of just using `noise_std`. * Updates argument validation Tangentially related changes: * For simplicity, makes `get_noise_stds` always return a dict * Stops allowing `noise_std` to be `None` and defaults it to zero (it was eventually set to zero when it was None in the past) Reviewed By: saitcakmak Differential Revision: D64919207
esantorella
added a commit
to esantorella/Ax
that referenced
this pull request
Oct 25, 2024
…nstraints (facebook#2963) Summary: Context: In theory, a `BenchmarkRunner` should not have to know what metrics are objectives or constraints, and a test function should not have to be aware of that, either. They are just generating data. A `BenchmarkProblem` should only store knowledge of objectives and constraints on the `OptimizationConfig`, so that various `OptimizationConfigs` can be used without changing the runner and test function. For historical reasons, runners track objectives and constraints separately and add noise to them separately, because this mimics how BoTorch test functions handle this. However, we now can and should isolate the quirks of BoTorch test functions to `BoTorchTestProblem`. This diff: * Updates `ParamBasedTestFunction.evaluate_true` to return all outcomes, not just objectives, and gets rid of `ParamBasedTestFunction.evaluate_true`, which was for constraints * Removes `num_objectives` from `ParamBasedTestProblem`, leaving `ParamBasedTestProblem` with nothing but an `evaluate_true` method * Removes the argument `constraint_noise_std` from `create_problem_from_botorch` and from `ParamBasedTestProblemRunner`, in favor of just using `noise_std`. * Updates argument validation Tangentially related changes: * For simplicity, makes `get_noise_stds` always return a dict * Stops allowing `noise_std` to be `None` and defaults it to zero (it was eventually set to zero when it was None in the past) Reviewed By: saitcakmak Differential Revision: D64919207
|
This pull request has been merged in bdc5df7. |
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:
Context: In theory, a
BenchmarkRunnershould not have to know what metrics are objectives or constraints, and a test function should not have to be aware of that either. They are just generating data. That should be handled by theOptimizationConfig, so that various setups can be used without changing the runner and test function.For historical reasons, runners track objectives and constraints separately and add noise to them separately, because this mimics how BoTorch test functions handle this. However, we now can and should isolate the quirks of BoTorch test functions within
BoTorchTestProblem.This diff:
constraint_noise_stdfromcreate_problem_from_botorchand fromParamBasedTestProblemRunner, in favor of just usingnoise_std.num_objectivestonum_outcomesinParamBasedTestProblemRunnerevaluate_slack_truefromParamBasedTestProblem, changingevaluate_trueto return both objectives and constraintsTangentially related changes:
get_noise_stdsalways return a dictnoise_stdto beNoneand defaults it to zero (it was eventually set to zero when it was None in the past)Differential Revision: D64919207