syncing from master #125
Merged
syncing from master #125
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
* first attempt * axis = None failing * axis = None passes * added ND grad tests * simplification * remove special case code * added future warning for cumsum(scalar) behavior * remove graph_implementaiton * clarify 0D warning * fix dqcp test raising future warning * create lower triangular sparse matrix directly, rathern than through dense to sparse
🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 4.5 <[email protected]>
* minimal step towards axis arg for sum_squares * more proper canonicalization * more full featured implementation * vectorized SOC constraints * inherit from AxisAtom * address claude's comments * Remove redundant sum_squares_axis test Removed redundant test for sum of squares as it is covered by another test. * Clean up comments in quad_over_lin.py Removed unnecessary comments about inherited methods.
* attempting a more systematic validation of _grad
Add expression-level gradcheck utility and fix gradient bugs
Implement a PyTorch-style gradcheck for systematically validating
expression-level gradients (expr.grad[var]) against numerical finite
differences.
New test infrastructure (test_expression_gradcheck.py):
- expression_gradcheck() using central finite differences
- expression_gradcheck_multi() for multi-variable atoms
- AtomInputGenerator for domain-aware input generation
- Parametrized tests covering ~50 atoms with multiple random seeds
- 185 passing tests, 64 skipped (atoms without gradient support)
Bug fixes discovered by gradcheck:
1. cumsum._grad (cumsum.py):
- Was returning standard Jacobian (lower triangular) instead of
CVXPY convention grad[i,j] = d(out[j])/d(in[i]) (upper triangular)
- Fixed axis=None case with correct F-order permutation handling
2. quad_over_lin._grad (quad_over_lin.py):
- Used np.reshape() without order='F', returning C-order instead
of CVXPY's F-order vectorization convention
- Fixed scalar handling when y is a 1-element array
Updated test_grad.py to use corrected gradient conventions.
Known issues (skipped in tests):
- pnorm_inf/norm_inf: Gradient not implemented
- matmul/multiply: Bugs in binary_operators.py for bilinear gradients
* fixed matmul and mul elemwise grads
* validated more atoms
* pass at consolidation
* fixed domains
* added more ND checks
* consolidating test_grad
* consolidate grad further
* address claude's comments
* remove tril
* Inline test ID helper functions per review feedback
Remove get_atom_config_ids(), get_multi_var_config_ids(), and
get_domain_violation_ids() functions and inline their list
comprehensions directly at the call sites.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <[email protected]>
---------
Co-authored-by: Claude Opus 4.5 <[email protected]>
…ling (#3044) * Add ignore_nan option to solver and validation tests for NaN/Inf handling * Moved to ParamConeProg and made some performance enhancements. * Update doc/source/tutorial/solvers/index.rst Co-authored-by: Copilot <[email protected]> * Update cvxpy/problems/problem.py Co-authored-by: Copilot <[email protected]> * stripped down nan checking * clarify where inf is allowed * moved nan check * clarifying comment * more efficient nan check * better test --------- Co-authored-by: Matthew Glover <[email protected]> Co-authored-by: Copilot <[email protected]> Co-authored-by: Steven Diamond <[email protected]>
* add CSR option * Changes what order argument does --------- Co-authored-by: Parth Nobel <[email protected]>
* document current sum of list behaviour * support generator input * use GeneratorType to make generator support cleaner --------- Co-authored-by: Muhammad Yasirroni <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 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.
Description
Please include a short summary of the change.
Issue link (if applicable):
Type of change
Contribution checklist