Repository navigation
Conversation
Extract cholesky_vjp as a NumPy/BLAS numerical primitive with parity tests against the pure and Cython paths, and add an in-repo harness for runtime and peak-allocation comparisons before flipping the default.
Member
Author
|
Closing for now — keep cip000E-numpy-blas-primitives as a separate branch until we are ready to merge. |
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## main #1303 +/- ##
==========================================
+ Coverage 62.22% 62.25% +0.03%
==========================================
Files 206 206
Lines 21962 21978 +16
Branches 2564 2566 +2
==========================================
+ Hits 13665 13683 +18
+ Misses 7675 7674 -1
+ Partials 622 621 -1
Flags with carried forward coverage won't be shown. Click here to find out more. 🚀 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
cholesky_vjpas a pure NumPy/BLAS reverse-mode Cholesky primitive (no model/paramz coupling).benchmarks/cip000E/harness for runtime / peak-allocation comparisons before flipping the default.backprop_gradientwhen built; usecholesky_vjpas fallback.Test plan
pytest GPy/testing/test_cython.py::TestCholeskiesBackproppython benchmarks/cip000E/run_cholesky_backprop.pyon Linux CI / Accept hosts and record resultsbackprop_gradientwhen Cython is built