Fix qnn.TorchLayer attributes
#4611
Merged
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
Sep 18, 2023
|
[sc-45054] |
1 task
timmysilv
reviewed
Sep 18, 2023
Codecov ReportPatch coverage:
Additional details and impacted files@@ Coverage Diff @@
## master #4611 +/- ##
==========================================
- Coverage 99.64% 99.64% -0.01%
==========================================
Files 375 375
Lines 33398 33397 -1
==========================================
- Hits 33280 33279 -1
Misses 118 118
☔ View full report in Codecov by Sentry. |
Co-authored-by: Matthew Silverman <matthews@xanadu.ai>
timmysilv
reviewed
Sep 19, 2023
Co-authored-by: Matthew Silverman <matthews@xanadu.ai>
timmysilv
approved these changes
Sep 19, 2023
mudit2812
approved these changes
Sep 19, 2023
mudit2812
pushed a commit
that referenced
this pull request
Sep 20, 2023
Fixes bug #4584 `torch.nn.Module` has some special `__getattr__` logic that we are ignoring if the qnode has been initialized. This bugfix makes it so the `qnn.TorchLayer.__getattr__` uses `torch.nn.Module.__getattr__` if the requested property is not on the QNode. --------- Co-authored-by: Matthew Silverman <matthews@xanadu.ai>
timmysilv
added a commit
that referenced
this pull request
Jan 5, 2024
**Context:** Fixes #4999. We re-implemented the default `setattr` and `getattr` dunders for `KerasLayer` (needed to modify the QNode saved on the layer) but we should have called `super()`, because `keras.layer.Layer` has its own custom implementation that we needed to use. Our `TorchLayer` already calls `super` in these cases. Note - our existing tests were passing because the weights used had custom names ("w1", "w2", etc.), and this issue only comes up when your weights are named "weights" (which is a rather common case, I'm sure). **Description of the Change:** - Replace our default-handling of `get/set-attr` dunders with a call to `super()` - add a test case matching the one in the bug report using the "weights" identifier **Benefits:** Models using `KerasLayer` can be saved and loaded properly! **Possible Drawbacks:** N/A **Related GitHub Issues:** Reported initially in #4999, same as #4611 but for Keras
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.
Fixes bug #4584
torch.nn.Modulehas some special__getattr__logic that we are ignoring if the qnode has been initialized. This bugfix makes it so theqnn.TorchLayer.__getattr__usestorch.nn.Module.__getattr__if the requested property is not on the QNode.