Generalize GenNode.apply_input_constructors
#4704
Open
+283
−241
Conversation
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
|
@lena-kashtelyan has exported this pull request. If you are a Meta employee, you can view the originating Diff in D80175141. |
meta-cla
bot
added
the
CLA Signed
lena-kashtelyan
force-pushed
the
export-D80175141
branch
from
87ce03e to
e19aba2
Compare
lena-kashtelyan
pushed a commit
to lena-kashtelyan/Ax
that referenced
this pull request
Dec 25, 2025
Summary: Long-overdue diff to stop applying input constructors in a manual way that defeats their purpose (if each is applied through a helper, what's the point of keeping them in a mapping from the name of the arg they are going to construct, to a callable that constructs it?) This diff: 1. Moves behavior related to injection of "disabled parameters" in the search space, from input constructor (where it's not even guaranteed to apply since that input constructor isn't always used), to generation node (for now; proposed move to adapter is in D89779057 –– let's consider alternatives on that diff, too), with an underlying utility moved to `SearchSpace` to be co-located with the rest of the disabled parameter logic. 2. Reaps `GenerationStrategy.DEFAULT_N`, which was actually unused (or used in one place but always left at its default of 1, only complicating things). D89772029 will deal with this better, with a top-level concurrency setting added via an `ExperimentDesign` object. 3. Applies input constructors generically, in a loop, instead of via helpers per input constructor purpose. This ends up removing a bunch of duplicate logic that lived at different layers within. Differential Revision: D80175141 Privacy Context Container: L1307644
lena-kashtelyan
pushed a commit
to lena-kashtelyan/Ax
that referenced
this pull request
Dec 26, 2025
Summary: Long-overdue diff to stop applying input constructors in a manual way that defeats their purpose (if each is applied through a helper, what's the point of keeping them in a mapping from the name of the arg they are going to construct, to a callable that constructs it?) This diff: 1. Moves behavior related to injection of "disabled parameters" in the search space, from input constructor (where it's not even guaranteed to apply since that input constructor isn't always used), to generation node (for now; proposed move to adapter is in D89779057 –– let's consider alternatives on that diff, too), with an underlying utility moved to `SearchSpace` to be co-located with the rest of the disabled parameter logic. 2. Reaps `GenerationStrategy.DEFAULT_N`, which was actually unused (or used in one place but always left at its default of 1, only complicating things). D89772029 will deal with this better, with a top-level concurrency setting added via an `ExperimentDesign` object. 3. Applies input constructors generically, in a loop, instead of via helpers per input constructor purpose. This ends up removing a bunch of duplicate logic that lived at different layers within. Differential Revision: D80175141 Privacy Context Container: L1307644
lena-kashtelyan
pushed a commit
to lena-kashtelyan/Ax
that referenced
this pull request
Dec 26, 2025
Summary: Long-overdue diff to stop applying input constructors in a manual way that defeats their purpose (if each is applied through a helper, what's the point of keeping them in a mapping from the name of the arg they are going to construct, to a callable that constructs it?) This diff: 1. Moves behavior related to injection of "disabled parameters" in the search space, from input constructor (where it's not even guaranteed to apply since that input constructor isn't always used), to generation node (for now; proposed move to adapter is in D89779057 –– let's consider alternatives on that diff, too), with an underlying utility moved to `SearchSpace` to be co-located with the rest of the disabled parameter logic. 2. Reaps `GenerationStrategy.DEFAULT_N`, which was actually unused (or used in one place but always left at its default of 1, only complicating things). D89772029 will deal with this better, with a top-level concurrency setting added via an `ExperimentDesign` object. 3. Applies input constructors generically, in a loop, instead of via helpers per input constructor purpose. This ends up removing a bunch of duplicate logic that lived at different layers within. Differential Revision: D80175141 Privacy Context Container: L1307644
lena-kashtelyan
force-pushed
the
export-D80175141
branch
2 times, most recently
from
85b2933 to
c57ac71
Compare
lena-kashtelyan
pushed a commit
to lena-kashtelyan/Ax
that referenced
this pull request
Dec 26, 2025
Summary: Long-overdue diff to stop applying input constructors in a manual way that defeats their purpose (if each is applied through a helper, what's the point of keeping them in a mapping from the name of the arg they are going to construct, to a callable that constructs it?) This diff: 1. Moves behavior related to injection of "disabled parameters" in the search space, from input constructor (where it's not even guaranteed to apply since that input constructor isn't always used), to generation node (for now; proposed move to adapter is in D89779057 –– let's consider alternatives on that diff, too), with an underlying utility moved to `SearchSpace` to be co-located with the rest of the disabled parameter logic. 2. Reaps `GenerationStrategy.DEFAULT_N`, which was actually unused (or used in one place but always left at its default of 1, only complicating things). D89772029 will deal with this better, with a top-level concurrency setting added via an `ExperimentDesign` object. 3. Applies input constructors generically, in a loop, instead of via helpers per input constructor purpose. This ends up removing a bunch of duplicate logic that lived at different layers within. Differential Revision: D80175141 Privacy Context Container: L1307644
lena-kashtelyan
pushed a commit
to lena-kashtelyan/Ax
that referenced
this pull request
Dec 26, 2025
Summary: Long-overdue diff to stop applying input constructors in a manual way that defeats their purpose (if each is applied through a helper, what's the point of keeping them in a mapping from the name of the arg they are going to construct, to a callable that constructs it?) This diff: 1. Moves behavior related to injection of "disabled parameters" in the search space, from input constructor (where it's not even guaranteed to apply since that input constructor isn't always used), to generation node (for now; proposed move to adapter is in D89779057 –– let's consider alternatives on that diff, too), with an underlying utility moved to `SearchSpace` to be co-located with the rest of the disabled parameter logic. 2. Reaps `GenerationStrategy.DEFAULT_N`, which was actually unused (or used in one place but always left at its default of 1, only complicating things). D89772029 will deal with this better, with a top-level concurrency setting added via an `ExperimentDesign` object. 3. Applies input constructors generically, in a loop, instead of via helpers per input constructor purpose. This ends up removing a bunch of duplicate logic that lived at different layers within. Differential Revision: D80175141 Privacy Context Container: L1307644
lena-kashtelyan
force-pushed
the
export-D80175141
branch
from
c57ac71 to
04794e0
Compare
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #4704 +/- ##
==========================================
- Coverage 96.71% 96.71% -0.01%
==========================================
Files 582 582
Lines 60713 60735 +22
==========================================
+ Hits 58717 58738 +21
- Misses 1996 1997 +1 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
lena-kashtelyan
pushed a commit
to lena-kashtelyan/Ax
that referenced
this pull request
Dec 26, 2025
Summary: Long-overdue diff to stop applying input constructors in a manual way that defeats their purpose (if each is applied through a helper, what's the point of keeping them in a mapping from the name of the arg they are going to construct, to a callable that constructs it?) This diff: 1. Moves behavior related to injection of "disabled parameters" in the search space, from input constructor (where it's not even guaranteed to apply since that input constructor isn't always used), to generation node (for now; proposed move to adapter is in D89779057 –– let's consider alternatives on that diff, too), with an underlying utility moved to `SearchSpace` to be co-located with the rest of the disabled parameter logic. 2. Reaps `GenerationStrategy.DEFAULT_N`, which was actually unused (or used in one place but always left at its default of 1, only complicating things). D89772029 will deal with this better, with a top-level concurrency setting added via an `ExperimentDesign` object. 3. Applies input constructors generically, in a loop, instead of via helpers per input constructor purpose. This ends up removing a bunch of duplicate logic that lived at different layers within. Differential Revision: D80175141 Privacy Context Container: L1307644
lena-kashtelyan
force-pushed
the
export-D80175141
branch
from
04794e0 to
7c7355b
Compare
lena-kashtelyan
pushed a commit
to lena-kashtelyan/Ax
that referenced
this pull request
Dec 27, 2025
Summary: Long-overdue diff to stop applying input constructors in a manual way that defeats their purpose (if each is applied through a helper, what's the point of keeping them in a mapping from the name of the arg they are going to construct, to a callable that constructs it?) This diff: 1. Moves behavior related to injection of "disabled parameters" in the search space, from input constructor (where it's not even guaranteed to apply since that input constructor isn't always used), to generation node (for now; proposed move to adapter is in D89779057 –– let's consider alternatives on that diff, too), with an underlying utility moved to `SearchSpace` to be co-located with the rest of the disabled parameter logic. 2. Reaps `GenerationStrategy.DEFAULT_N`, which was actually unused (or used in one place but always left at its default of 1, only complicating things). D89772029 will deal with this better, with a top-level concurrency setting added via an `ExperimentDesign` object. 3. Applies input constructors generically, in a loop, instead of via helpers per input constructor purpose. This ends up removing a bunch of duplicate logic that lived at different layers within. Reviewed By: saitcakmak Differential Revision: D80175141 Privacy Context Container: L1307644
added 2 commits
December 27, 2025 10:04
Summary: When `gen_single_trial` does not call `gen`, we have a few problems, in order of importance: 1. Forgetting to specify `pending_observations` to `gen_single_trial` will result in no `pending_observations` use. This is different from the behavior in the main `gen`, so it's a huge gotcha and could already be causing silent bugs. 2. Unnecessary bifurcation, meaning that more behaviors like 1) could encroach in the future. 3. Repeated operations. Reviewed By: saitcakmak Differential Revision: D80174582 Privacy Context Container: L1307644
Summary: Long-overdue diff to stop applying input constructors in a manual way that defeats their purpose (if each is applied through a helper, what's the point of keeping them in a mapping from the name of the arg they are going to construct, to a callable that constructs it?) This diff: 1. Moves behavior related to injection of "disabled parameters" in the search space, from input constructor (where it's not even guaranteed to apply since that input constructor isn't always used), to generation node (for now; proposed move to adapter is in D89779057 –– let's consider alternatives on that diff, too), with an underlying utility moved to `SearchSpace` to be co-located with the rest of the disabled parameter logic. 2. Reaps `GenerationStrategy.DEFAULT_N`, which was actually unused (or used in one place but always left at its default of 1, only complicating things). D89772029 will deal with this better, with a top-level concurrency setting added via an `ExperimentDesign` object. 3. Applies input constructors generically, in a loop, instead of via helpers per input constructor purpose. This ends up removing a bunch of duplicate logic that lived at different layers within. Reviewed By: saitcakmak Differential Revision: D80175141 Privacy Context Container: L1307644
lena-kashtelyan
pushed a commit
to lena-kashtelyan/Ax
that referenced
this pull request
Dec 27, 2025
Summary: Long-overdue diff to stop applying input constructors in a manual way that defeats their purpose (if each is applied through a helper, what's the point of keeping them in a mapping from the name of the arg they are going to construct, to a callable that constructs it?) This diff: 1. Moves behavior related to injection of "disabled parameters" in the search space, from input constructor (where it's not even guaranteed to apply since that input constructor isn't always used), to generation node (for now; proposed move to adapter is in D89779057 –– let's consider alternatives on that diff, too), with an underlying utility moved to `SearchSpace` to be co-located with the rest of the disabled parameter logic. 2. Reaps `GenerationStrategy.DEFAULT_N`, which was actually unused (or used in one place but always left at its default of 1, only complicating things). D89772029 will deal with this better, with a top-level concurrency setting added via an `ExperimentDesign` object. 3. Applies input constructors generically, in a loop, instead of via helpers per input constructor purpose. This ends up removing a bunch of duplicate logic that lived at different layers within. Reviewed By: saitcakmak Differential Revision: D80175141 Privacy Context Container: L1307644
lena-kashtelyan
force-pushed
the
export-D80175141
branch
from
7c7355b to
5c1ebdf
Compare
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:
Long-overdue diff to stop applying input constructors in a manual way that defeats their purpose (if each is applied through a helper, what's the point of keeping them in a mapping from the name of the arg they are going to construct, to a callable that constructs it?)
This diff:
SearchSpaceto be co-located with the rest of the disabled parameter logic.GenerationStrategy.DEFAULT_N, which was actually unused (or used in one place but always left at its default of 1, only complicating things). D89772029 will deal with this better, with a top-level concurrency setting added via anExperimentDesignobject.Differential Revision:
D80175141
Privacy Context Container: L1307644