Conversation
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## master #861 +/- ##
==========================================
- Coverage 96.86% 95.37% -1.49%
==========================================
Files 128 128
Lines 26181 26212 +31
==========================================
- Hits 25359 25001 -358
- Misses 822 1211 +389 🚀 New features to boost your workflow:
|
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Types of changes
Description
ot.solve_sample(X_a, X_b, a, b, reg=..., lazy=True)returns anOTResultwhosevalueisNone: the lazy regularized branch setsvalue_linear, the potentials and the lazy plan, but never the regularized value. Because of this, the debiased lazy Sinkhorn divergence fails:The value is now computed as in the non-lazy
ot.solve(value_linear + reg * R(plan), with R the KL toa b^Tor the negative entropy). The plan is read fromlazy_planin batches ofbatch_sizerows, so memory use stays O(batch_size× n_b).The lazy solver is always the entropic Sinkhorn, so it now raises
NotImplementedErrorfor anotherreg_type(for example"L2") instead of silently solving a different problem, as the non-lazy path already does for unsupported values.How has this been tested
Added
test_solve_sample_lazy_value, which checksvalueagainst the non-lazy solver forreg_type"KL" and "entropy" (with batches smaller than the problem), and fordebias=Trueand"split". It fails on master.test/test_solvers.pypasses (337 passed, 5 skipped).Checklist