Repository navigation
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
ZeroLinearOperatorrecords a dtype/device but drops them when materializing its dense value, indexing, and transposing. A CUDA double zero operator therefore produces CPU float32 values. Itslogdet()similarly returns a CPU float32 scalar even for a batched operator.Preserve placement in those constructors and return one negative-infinity log determinant per batch. Use the tuple-shaped
torch.zerosform for scalar advanced-index results as well. The regression checks dense conversion, scalar/slice/tensor indexing, point/batch transposition and batched log determinants for CPU/CUDA and float32/float64. All four new device/dtype subcases fail against unchanged main; the corrected full zero-operator module passes (9 tests, 4 subtests).ufmt, flake8 and
git diff --checkpass. This patch and its tests were prepared with Codex.