Fix leftover quantum operations in classical functions during the gradient passes #439
Conversation
This file contains 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
- `0_` to avoid it appearing after the HLO pass in intermediate files - more descriptive name than "pipeline"
MLIR generated from the JAX frontend typically contains a lot of duplicate instructions. For instance, the extraction of a scalar value from the same scalar tensor could be repeated several times. Besides making the program larger than it needs to be, it also makes it more difficult to perform insert/extract elimination when dynamic wires are used. This was causing the gradient pass to fail in some instances, specifically when duplicating a hybrid function while removing all quantum code from it.
dime10
added
bug
dime10
force-pushed
the
fix-leftover-quantum
branch
from
ebdfe94
to
504331d
Compare
|
@erick-xanadu I implemented the single register & qubit reset in between transform tapes (626df4d), if you want to give it a look. |
erick-xanadu
approved these changes
Jan 12, 2024
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #439 +/- ##
=======================================
Coverage 99.56% 99.56%
=======================================
Files 43 43
Lines 7643 7662 +19
Branches 512 514 +2
=======================================
+ Hits 7610 7629 +19
Misses 17 17
Partials 16 16 ☔ View full report in Codecov by Sentry. |
Merged
josh146
added a commit
that referenced
this pull request
Jan 29, 2024
**Context:** v0.4.1 bugfix PR **Description of the Change:** Incorporates the following PRs: #439 #457 #437 #465 #471 --------- Co-authored-by: David Ittah <dime10@users.noreply.github.com> Co-authored-by: Ali Asadi <10773383+maliasadi@users.noreply.github.com> Co-authored-by: Romain Moyard <rmoyard@gmail.com> Co-authored-by: Rashid N H M <95639609+rashidnhm@users.noreply.github.com> Co-authored-by: erick-xanadu <110487834+erick-xanadu@users.noreply.github.com>
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.
Resolves an issue in the gradient dialect brought to light by #397. The error raised is as follows:
The error is thrown when when a quantum function is cloned but with all of its quantum operations removed, in order to retain only the classical code. This is done in several places in the gradient dialect.
In this case, certain input programs (see linked issue) would still cause the gradient pass to be unable to remove all quantum operations from a function. A snippet of the offending IR would look like:
Even though the qubit extraction and insertion are performed at the same index, the SSA value used in both operations is not the same, and so the operations are not automatically folded way. This folding is usually relied upon to remove all quantum operations from a function.
Applying CSE before the gradient pass resolves the issue.
[sc-53218]
fixes #437