We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix pytorch nightly compilation for cuda linking
Fixes #ISSUE_NUMBER
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @peterbell10 @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @aakhundov @ColinPeppler @amjames @desertfire @chauhang
Sorry, something went wrong.
Note: Links to docs will display an error until the docs builds have been completed.
As of commit 51bc5fb with merge base c4a1570 ():
👉 Rebase onto the `viable/strict` branch to avoid these failures
DebertaForQuestionAnswering
llava
>>> Lint for test/onnx/test_fx_op_consistency.py:
This comment was automatically generated by Dr. CI and updates every 15 minutes.
Please seek CI approval before scheduling CIFlow labels
More details at: triton-lang/triton#3053
Why have we not detected this with the tests on our side last time we have updated the pin?
Can we wait for this before we cut the next nightly?
Do I need to do something specific for: https://github.com/pytorch/pytorch/actions/runs/7832512037/job/21371840351 or it was already failing?
It seems it was just an http failure right?
There seems to be some real failure caused by interface change of AttrsDescriptor.
Bert also wants to try out the OSS triton pin update process. So cc @bertmaher
In even newer triton, divisible_by_8 is gone: https://github.com/openai/triton/blob/0f74acc4d92ea506105d25b2724439df85133b19/python/triton/compiler/compiler.py#L21-L22 . Not sure if this is what @aakhundov mean
divisible_by_8
@shunting314 is it still used here?
pytorch/test/inductor/test_metrics.py
Line 20 in 3381f28
@bertmaher @bhack re. broken test_triton_kernel_equal_to_1_arg, yeah, so ids_of_folded_args seems to be gone in the Triton version this PR updates to (can't find it if grep jit.py). An easy fix would be removing the lines 1002 and 1005 from the unit test below:
test_triton_kernel_equal_to_1_arg
ids_of_folded_args
pytorch/test/inductor/test_triton_kernels.py
Lines 998 to 1005 in 3381f28
I see we're already skipping ids_of_folded_args in the Inductor code if unavailable, so this part should be good:
pytorch/torch/_inductor/utils.py
Lines 710 to 712 in 3381f28
I'm wondering why we didn't catch this error, though. Was the PR not rebased before landing? I've added the test mid-last week.
In even newer triton, divisible_by_8 is gone:
In the Triton version that this PR updates to, divisible_by_8 is still in: https://github.com/openai/triton/blob/a9bc1a36470eefafe0e2ab2503b8698f1e89e7e3/python/triton/runtime/jit.py#L227-L231
When I've adapted the interface divisible_by_8 was still available #119636. It was only removed ids_of_folded_args. Are we talking about different triton shas?
Remove ids_of_folded_args from test_triton_kernel_equal_to_1_arg
97f60b1
Summary: Due to the Triton pin update in #119457, `test_triton_kernel_equal_to_1_arg` started to break, as `ids_of_folded_args` has vanished from the upstream Triton codebase. Test Plan: ``` $ python test/inductor/test_triton_kernels.py -k test_triton_kernel_equal_to_1_arg ... ---------------------------------------------------------------------- Ran 6 tests in 6.790s OK ``` [ghstack-poisoned]
5f780e3
Summary: Due to the Triton pin update in #119457, `test_triton_kernel_equal_to_1_arg` started to break, as `ids_of_folded_args` has vanished from the upstream Triton codebase. Test Plan: ``` $ python test/inductor/test_triton_kernels.py -k test_triton_kernel_equal_to_1_arg ... ---------------------------------------------------------------------- Ran 6 tests in 6.790s OK ``` ghstack-source-id: 1a7ef6f Pull Request resolved: #121192
@bhack yeah, it's just ids_of_folded_args. Here's a forward fix for this test: #121192
Remove ids_of_folded_args from test_triton_kernel_equal_to_1_arg (#12…
9ccff0a
…1192) Summary: Due to the Triton pin update in #119457, `test_triton_kernel_equal_to_1_arg` started to break, as `ids_of_folded_args` has vanished from the upstream Triton codebase. Test Plan: ``` $ python test/inductor/test_triton_kernels.py -k test_triton_kernel_equal_to_1_arg ... ---------------------------------------------------------------------- Ran 6 tests in 6.790s OK ``` Pull Request resolved: #121192 Approved by: https://github.com/oulgen, https://github.com/bertmaher
@pytorchbot merge -r main
This PR needs to be approved by an authorized maintainer before merge.
@pytorchbot started a rebase job onto refs/remotes/origin/main. Check the current status here
Rebase failed due to Command git -C /home/runner/work/pytorch/pytorch rebase refs/remotes/origin/main pull/119457/head returned non-zero exit code 1
git -C /home/runner/work/pytorch/pytorch rebase refs/remotes/origin/main pull/119457/head
Rebasing (1/1) Auto-merging .ci/docker/ci_commit_pins/triton.txt CONFLICT (content): Merge conflict in .ci/docker/ci_commit_pins/ 628C triton.txt error: could not apply 4554a5a53cb... Update triton hint: Resolve all conflicts manually, mark them as resolved with hint: "git add/rm <conflicted_files>", then run "git rebase --continue". hint: You can instead skip this commit: run "git rebase --skip". hint: To abort and get back to the state before "git rebase", run "git rebase --abort". Could not apply 4554a5a53cb... Update triton
Raised by https://github.com/pytorch/pytorch/actions/runs/8153738367
Given the revert, it needs a manual rebase. Oh well.
Update triton
51bc5fb
b6112de
@pytorchbot merge
Your change will be merged once all checks pass (ETA 0-4 Hours).
Learn more about merging in the wiki.
Questions? Feedback? Please reach out to the PyTorch DevX Team
fbf36d0
lezcano lezcano approved these changes
bertmaher bertmaher left review comments
desertfire Awaiting requested review from desertfire desertfire is a code owner
Chillee Awaiting requested review from Chillee Chillee is a code owner
eellison Awaiting requested review from eellison eellison is a code owner
shunting314 Awaiting requested review from shunting314 shunting314 is a code owner
jeffdaily Awaiting requested review from jeffdaily 9499 jeffdaily is a code owner
jataylo Awaiting requested review from jataylo jataylo is a code owner
jithunnair-amd Awaiting requested review from jithunnair-amd jithunnair-amd is a code owner
pruthvistony Awaiting requested review from pruthvistony pruthvistony is a code owner
Skylion007 Awaiting requested review from Skylion007
Successfully merging this pull request may close these issues.