Fix ChainedAssignmentError in warm_start_from_old_experiment #4799
+23
−18
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
meta-cla
bot
added
the
CLA Signed
Summary: I'm not sure why these are showing up as failures today, we migrated to Python scientific stack in july 2024, but these are deprecation warnings we should go ahead and take care of anyway. Figured would knock them out while they are causing me issues Prompted after some failures in exports not related to my changes: https://github.com/facebook/Ax/actions/runs/21225324302/job/61070638075?fbclid=IwY2xjawPeXMFleHRuA2FlbQIxMQBicmlkETFRTkR6WlE4NHVrd3IyQXNlc3J0YwZhcHBfaWQBMAABHjTAiZi71n24w95hvzEewrKNPKOGzJisgR7t4qJ3APRMYlusgFC-gu7RLiSb_aem_Zk3pmTDonCFsJvZCTkpeMA Add `.copy()` after `.numpy()` calls to ensure arrays are writeable. PyTorch tensors converted via `.detach().cpu().numpy()` return read-only arrays in some cases. The subsequent squeeze operations create read-only views, and the in-place assignment `loo_covs[:, diag_idx, diag_idx] = loo_vars` fails with "assignment destination is read-only" error. Differential Revision: D91185467
…ok#4795) Summary: Prompted after some failures in exports not related to my changes: https://github.com/facebook/Ax/actions/runs/21225324302/job/61070638075?fbclid=IwY2xjawPeXMFleHRuA2FlbQIxMQBicmlkETFRTkR6WlE4NHVrd3IyQXNlc3J0YwZhcHBfaWQBMAABHjTAiZi71n24w95hvzEewrKNPKOGzJisgR7t4qJ3APRMYlusgFC-gu7RLiSb_aem_Zk3pmTDonCFsJvZCTkpeMA Pandas 2.0+ changed default string column dtype from `object` to `StringDtype(na_value=nan)`. The `_safecast_df()` method doesn't properly handle the comparison between `StringDtype` and `np.dtype("O")` because they are different types that don't compare equal. Add explicit check for `pd.StringDtype` to force casting when needed. Differential Revision: D91185469
Summary: Prompted after some failures in exports not related to my changes: https://github.com/facebook/Ax/actions/runs/21225324302/job/61070638075?fbclid=IwY2xjawPeXMFleHRuA2FlbQIxMQBicmlkETFRTkR6WlE4NHVrd3IyQXNlc3J0YwZhcHBfaWQBMAABHjTAiZi71n24w95hvzEewrKNPKOGzJisgR7t4qJ3APRMYlusgFC-gu7RLiSb_aem_Zk3pmTDonCFsJvZCTkpeMA The `trial_indices` property returns `set[np.int64]` from `.unique()` on numpy-backed dataframes, but `experiment.trials.keys()` returns Python `int`. Set operations fail because `np.int64` and `int` are treated as different types in set comparisons. Convert numpy types to Python int using set comprehension. Differential Revision: D91185468
Summary: The `copy=False` parameter in `DataFrame.astype()` and `DataFrame.reindex()` was deprecated in pandas 2.0 and removed in pandas 3.0. These calls now raise warnings or errors. Remove the deprecated `copy=False` argument from: - `arm_data.astype(dtype=column_to_type, copy=False)` in `cast.py` - `df.reindex(columns=desired_order, copy=False)` in `data.py` This is part of the ongoing pandas 2.0+ compatibility fixes. Differential Revision: D91186870
Summary: Converting a list of numpy arrays directly to a PyTorch tensor is extremely slow because PyTorch iterates through each array individually. This fix converts the list to a single contiguous numpy array first using `np.array()`, which PyTorch can then convert efficiently in one operation. Performance benchmark (10 constraints × 1000 arms, 100 iterations): - Before: 1.546 ms per call - After: 0.428 ms per call - **Speedup: 3.6x faster** This eliminates the UserWarning that was cluttering test output and logs. Differential Revision: D91187902
…k#4799) Summary: Pandas 3.0 introduced Copy-on-Write (CoW) which makes chained assignment with `inplace=True` ineffective. The pattern `df["col"].replace(..., inplace=True)` no longer modifies the original DataFrame because the column selection creates a copy. Replace the chained `inplace=True` pattern with explicit assignment: - `df["col"].replace(..., inplace=True)` → `df["col"] = df["col"].replace(...)` This eliminates the `ChainedAssignmentError` warning and ensures the DataFrame is actually modified as intended. Differential Revision: D91187974
mgarrard
force-pushed
the
export-D91187974
branch
from
bfb32c5 to
f4b6157
Compare
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #4799 +/- ##
=======================================
Coverage 96.70% 96.70%
=======================================
Files 587 587
Lines 61294 61296 +2
=======================================
+ Hits 59273 59276 +3
+ Misses 2021 2020 -1 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
mgarrard
added a commit
to mgarrard/Ax
that referenced
this pull request
Jan 22, 2026
…k#4799) Summary: Pandas 3.0 introduced Copy-on-Write (CoW) which makes chained assignment with `inplace=True` ineffective. The pattern `df["col"].replace(..., inplace=True)` no longer modifies the original DataFrame because the column selection creates a copy. Replace the chained `inplace=True` pattern with explicit assignment: - `df["col"].replace(..., inplace=True)` → `df["col"] = df["col"].replace(...)` This eliminates the `ChainedAssignmentError` warning and ensures the DataFrame is actually modified as intended. Differential Revision: D91187974
mgarrard
added a commit
to mgarrard/Ax
that referenced
this pull request
Jan 22, 2026
…k#4799) Summary: Pandas 3.0 introduced Copy-on-Write (CoW) which makes chained assignment with `inplace=True` ineffective. The pattern `df["col"].replace(..., inplace=True)` no longer modifies the original DataFrame because the column selection creates a copy. Replace the chained `inplace=True` pattern with explicit assignment: - `df["col"].replace(..., inplace=True)` → `df["col"] = df["col"].replace(...)` This eliminates the `ChainedAssignmentError` warning and ensures the DataFrame is actually modified as intended. Differential Revision: D91187974
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:
Pandas 3.0 introduced Copy-on-Write (CoW) which makes chained assignment with
inplace=Trueineffective. The patterndf["col"].replace(..., inplace=True)no longer modifies the original DataFrame because the column selection creates a copy.Replace the chained
inplace=Truepattern with explicit assignment:df["col"].replace(..., inplace=True)→df["col"] = df["col"].replace(...)This eliminates the
ChainedAssignmentErrorwarning and ensures the DataFrame is actually modified as intended.Differential Revision: D91187974