Fix QNN simulator failures after HadamardTransform support - #21843
Conversation
Summary: With certain simulator setups, the custom hadamard transfrom could be imported multiple times. Further, the QNN SDK version can crash due to repeated SDK version checks repeatedly loaded/unloaded libQnnHtp.so, corrupting native state. So, we now: 1) check to guard against multiple import 2) SDK build IDs are now cached per root Differential Revision: D116043878
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/21843
Note: Links to docs will display an error until the docs builds have been completed. ❌ 2 New FailuresAs of commit a8e784b with merge base ea3641d ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
@JakeStevens has exported this pull request. If you are a Meta employee, you can view the originating Diff in D116043878. |
This PR needs a
|
|
Thanks @JakeStevens for putting this PR |
|
@chenweng-quic et al: I am landing due to many internal tests breaking after the hadamard PR. I believe this is preferable to a revert. If you have any concerns with this approach, we can resolve in follow up! Thanks |
qti-horodnic
left a comment
There was a problem hiding this comment.
Approving, thanks for fixing the issue rather than reverting.
|
Thanks @qti-horodnic for confirming. and thank you @JakeStevens for your prompt fix . Lets merge |
Summary:
With certain simulator setups, the custom hadamard transfrom could be imported multiple times.
Further, the QNN SDK version can crash due to repeated SDK version checks repeatedly loaded/unloaded libQnnHtp.so, corrupting native state. So, we now:
Differential Revision: D116043878
cc @cbilgin @psiddh