Reimplement qubit unitary #6986
Conversation
Co-authored-by: Christina Lee <[email protected]>
|
@JerryChen97, are these CI tests supposed to fail? |
Sorry, forgot to fix them on e2e branch. I'll fix soon |
|
External fail is exactly the same as this morning. Just ignore it until we merge this PR to e2e branch and update there with master |
**Context:** In our on-going sparse matrix epic, we used to postponed the full implementation of controlled ops family due to others of higher priorities. However, in a related PR https://github.com/PennyLaneAI/pennylane/actions/runs/13464002433/job/37625640042?pr=6986 it seems inevitable to have the `wire_order` branch of `Controlled.sparse_matrix` method implemented. For the sake of traceability, and to prevent some the mentioned PR from being furthermore spaghettized with high entanglement, we implement the `wire_order` branch here with unit tests, and watch and fix whatever new impact introduced as an independent PR instead of hotfix overthere. **Description of the Change:** Direct call `expand_matrix` just as what `matrix()` did above `sparse_matrix`, right before format-convert. Convert the error-raise test into a consistency-check with `matrix(wire_order=...)` **Benefits:** Fully fledged sparse `Controlled` Unblock #6986 **Possible Drawbacks:** Small chance to still have incompatibility in the future, which is also why we would like to have this separated such that easy to revert. **Related GitHub Issues:** [sc-84949] --------- Co-authored-by: Andrija Paurevic <[email protected]>
|
😅Messed up with the git branch control |
JerryChen97
force-pushed
the
reimplement-QubitUnitary
branch
from
cfaa2aa to
7a24f16
Compare
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## e2e-sparse-default-qubit #6986 +/- ##
===========================================================
Coverage ? 99.59%
===========================================================
Files ? 483
Lines ? 46156
Branches ? 0
===========================================================
Hits ? 45971
Misses ? 185
Partials ? 0 ☔ View full report in Codecov by Sentry. |
|
@AmintorDusko @albi3ro Now we have cleared the errors. Major changes since last check: 1. |
Co-authored-by: Christina Lee <[email protected]>
|
Hey @JerryChen97, we still have some work to do regarding this error. |
It should work as inherited from parents
|
I wonder if you know what is happening with PennyLane wheels here. |
PL CI affected by setuptools updates. Fix coming soon. |
|
CI back and passed |
Context:
Our current sparse repr for
QubitUntiaryis quite confusing: itsmatrix()return a sparse matrix by default, if this obj was instantiated by a CSR, which is inconsistent from what we expected for other operators, thatmatrix()should return a dense andsparse_matrix()gives out the sparse, well separated from each other.Therefore, a slight re-work is needed to make things clearer. Basically, we will overload the
has_matrixandhas_sparsefrom parents and have a more friendly and straightforward usage.Description of the Change:
Benefits:
Possible Drawbacks:
Related GitHub Issues: