Huy Do
Huy Do
One use case requires this is to force merge old PR, but this check has been removed because the field from GitHub has been deprecated (and become None)
May be we could fallback to regular merge instead of a force merge. However, we also to raise the awareness of when to use force merge. Eli: we could do...
@pytorchbot drci (please ignore this, I'm testing Dr.CI)
AFAICT, nothing depends on `types-dataclasses` to build or run PyTorch which comes from this https://github.com/pytorch/pytorch/pull/61879 (I don't even have that package in my PyTorch local env). It makes more sense...
Just FYI, the authentication part with AWS doesn't work with forked PR, i.e. `kirklandsign:android-prebuilt` from your fork, so it's failing expectedly https://github.com/pytorch/executorch/actions/runs/8792470295/job/24128672297?pr=3215. I guess we can ignore it in this...
@pytorchbot drci (please ignore this, I'm testing Dr.CI)
@pytorchbot revert -m 'This breaks XLA test https://hud.pytorch.org/pytorch/pytorch/commit/34e8eb229db76f7f5eb8f18c062dbd1ee47f8b12' -c landrace
> ROCm CI is stable now after skipping multiprocessing tests. Thank @jeffdaily. This is great. We don't need this after all. I'll monitor HUD a bit longer with more commits...
It looks like a similar flaky test as https://github.com/pytorch/pytorch/issues/92178 and https://github.com/pytorch/pytorch/issues/92042