Arm backend: Fix bug in ConvertExpandCopyToRepeatPass #7199
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7199
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ No FailuresAs of commit 912cfe6 with merge base de74961 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
Hi @Aleksei-grovety! Thank you for your pull request and welcome to our community. Action RequiredIn order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you. ProcessIn order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA. Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with If you have received this in error or have any questions, please contact us at [email protected]. Thanks! |
zingo
added
partner: arm
|
Please seek CI approval before scheduling CIFlow labels |
|
Hi @Aleksei-grovety ! This needs a rebase pls |
Aleksei-grovety
force-pushed
the
fix-repeat-args-in-arm-pass
branch
from
62a2493 to
a378621
Compare
Hi! Thanks for the advice. Done. |
Aleksei-grovety
force-pushed
the
fix-repeat-args-in-arm-pass
branch
from
a378621 to
4d8a7e4
Compare
|
Thx @Aleksei-grovety! There seem to be a general issue in the CI. See #7285 (comment) |
|
There's two issues. 1) that you saw in your last log had to do with docker images being push ahead of PR merge. That should be resolved now. 2) this: #7285 (comment) |
|
@Aleksei-grovety sorry, you need to rebase again for the tests to pass :/ |
In the ConvertExpandCopyToRepeatPass the arguments for the repeat operation are formed incorrectly. For the torch.Tensor.expand operation passing -1 as the size for a dimension means that the size of that dimension does not change. For the DeiT-tiny case, torch.ones(1, 1, 192).expand(1, -1, -1) the pass will prepare arguments to the repeat operation as [1, -1, 1] which will cause an error, in this case the arguments should be [1, 1, 1].
Aleksei-grovety
force-pushed
the
fix-repeat-args-in-arm-pass
branch
from
4d8a7e4 to
912cfe6
Compare
|
No problem. |
|
@Aleksei-grovety nice, tests pass. Still can't merge though. CLA check has to pass. Do you have a CCLA? @digantdesai fyi - need some guidance in this. |
|
Actually check: https://github.com/pytorch/executorch/blob/main/CONTRIBUTING.md#contributing-workflow, that should probably do it. Apologies for merging taking time. We appreciate your contribution! |
This process is in progress, our manager has already sent a request from our company. |
facebook-github-bot
added
the
CLA Signed
5 participants
In the ConvertExpandCopyToRepeatPass the arguments for the repeat operation are formed incorrectly. For the torch.Tensor.expand operation passing -1 as the size for a dimension means that the size of that dimension does not change. For the DeiT-tiny case, torch.ones(1, 1, 192).expand(1, -1, -1) the pass will prepare arguments to the repeat operation as [1, -1, 1] which will cause an error, in this case the arguments should be [1, 1, 1].