Improvements to Jitting: broadcasted measurements and counts on all wires #6108
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
albi3ro
commented
Aug 16, 2024
albi3ro
commented
Aug 16, 2024
albi3ro
commented
Aug 16, 2024
albi3ro
commented
Aug 16, 2024
albi3ro
commented
Aug 16, 2024
albi3ro
commented
Aug 16, 2024
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## master #6108 +/- ##
==========================================
- Coverage 99.67% 99.66% -0.01%
==========================================
Files 432 443 +11
Lines 41839 42240 +401
==========================================
+ Hits 41702 42098 +396
- Misses 137 142 +5 ☔ View full report in Codecov by Sentry. |
albi3ro
added a commit
that referenced
this pull request
Aug 28, 2024
mudit2812
pushed a commit
that referenced
this pull request
Sep 10, 2024
…ires (#6108) **Context:** When using jax jit and non-backprop, we need to know the exact shape of the result value in order to use a `pure_callback` for the device execution. The framework we use to determine this shape currently does not work for measurements on all available wires when the device does not specify a wire order. It also explicitly errors out with `qml.counts`, even when `all_outcomes=True`. But when `all_outcomes=True`, we know the exact resulting shape, so we can integrate it with jax-jit anyway. **Description of the Change:** This PR makes a breaking change to `MeasurementProcess.shape`. Now it's call signature is `(self, shots: Optional[int]=None, num_device_wires:int =0)`. `num_device_wires` can take on the value of `len(tape.wires)` when the measurement is broadcasted on all available wires, but the device does not specify a number of wires. **Benefits:** Improved jit support. **Possible Drawbacks:** Breaking changes always cause some draw backs. We don't know who may be relying on this method somewhere in the wild. But given we only use the method for the jax-jit interface, I think we are safe to make a breaking change here. **Related GitHub Issues:** [sc-65313] [sc-59327] Fixes #5813
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.
Context:
When using jax jit and non-backprop, we need to know the exact shape of the result value in order to use a
pure_callbackfor the device execution. The framework we use to determine this shape currently does not work for measurements on all available wires when the device does not specify a wire order.It also explicitly errors out with
qml.counts, even whenall_outcomes=True. But whenall_outcomes=True, we know the exact resulting shape, so we can integrate it with jax-jit anyway.Description of the Change:
This PR makes a breaking change to
MeasurementProcess.shape. Now it's call signature is(self, shots: Optional[int]=None, num_device_wires:int =0).num_device_wirescan take on the value oflen(tape.wires)when the measurement is broadcasted on all available wires, but the device does not specify a number of wires.Benefits:
Improved jit support.
Possible Drawbacks:
Breaking changes always cause some draw backs. We don't know who may be relying on this method somewhere in the wild. But given we only use the method for the jax-jit interface, I think we are safe to make a breaking change here.
Related GitHub Issues:
Fixes #5813
Related Shortcut Stories:
[sc-65313] [sc-59327]