[Inductor] Add triton.autotune support for user defined triton kernels with constant/simple grids #112228

oulgen · 2023-10-27T02:34:37Z

Stack from ghstack (oldest at bottom):

Support calling user defined triton kernels with kernel.run #112292
[Inductor] Add triton.autotune support for user defined triton kernels with complex grids #112290
-> [Inductor] Add triton.autotune support for user defined triton kernels with constant/simple grids #112228

cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @peterbell10 @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @aakhundov @ColinPeppler

…s with constant/simple grids [ghstack-poisoned]

pytorch-bot · 2023-10-27T02:34:40Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/112228

📄 Preview Python docs built from this PR
📄 Preview C++ docs built from this PR
❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit 356c5b1 with merge base 8d44999 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

…s with constant/simple grids ghstack-source-id: 37bf2369d049574f82b200610f1f1f35e7120f94 Pull Request resolved: #112228

…iton kernels with constant/simple grids" cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]

…s with constant/simple grids ghstack-source-id: 65a5279fb602e9cb58da1df0a55b9a1c6f853b3a Pull Request resolved: #112228

oulgen · 2023-10-27T02:51:10Z

This PR only handles simple grids, on the next PR I will deal with complex grids

oulgen · 2023-10-27T02:53:20Z

torch/_inductor/triton_heuristics.py

@@ -362,16 +363,17 @@ def kernel_call():
                    {**dict(zip(self.arg_names, args)), **launcher.config.kwargs}


Hmm, I'm not really sure what this pre_hook stuff is. I assume I need to put **kwargs here too, how to trigger/test this?

…iton kernels with constant/simple grids" cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]

…s with constant/simple grids ghstack-source-id: db33a033857213c6ce6756309174eddd2de8d3fb Pull Request resolved: #112228

oulgen · 2023-10-27T21:38:09Z

@pytorchbot merge

pytorchmergebot · 2023-10-27T21:40:03Z

Merge started

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

Advanced Debugging

Check the merge workflow status
here

huydhn · 2023-10-28T01:49:49Z

@pytorchbot revert -m 'Sorry for reverting your change but it is failing ROCm test in trunk https://hud.pytorch.org/pytorch/pytorch/commit/dbb31a2984fa616b4bb6fac7abb2a06ec0533eb1' -c nosignal

You can add ciflow/rocm to your PR to trigger the failure.

pytorchmergebot · 2023-10-28T01:51:28Z

@pytorchbot successfully started a revert job. Check the current status here.
Questions? Feedback? Please reach out to the PyTorch DevX Team

pytorchmergebot · 2023-10-28T01:51:37Z

@oulgen your PR has been successfully reverted.

…n kernels with constant/simple grids (#112228)" This reverts commit dbb31a2. Reverted #112228 on behalf of https://github.com/huydhn due to Sorry for reverting your change but it is failing ROCm test in trunk https://hud.pytorch.org/pytorch/pytorch/commit/dbb31a2984fa616b4bb6fac7abb2a06ec0533eb1 ([comment](#112228 (comment)))

…iton kernels with constant/simple grids" cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx peterbell10 ipiszy yf225 chenyang78 kadeng muchulee8 aakhundov ColinPeppler [ghstack-poisoned]

oulgen · 2023-10-28T02:37:36Z

@huydhn No worries, I just disabled this test under ROCM. If rocm mode is valuable, could we run these tests by default so that we don't have to revert diffs?

oulgen · 2023-10-28T17:27:39Z

@pytorchbot merge

pytorchmergebot · 2023-10-28T17:29:47Z

Merge started

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

Advanced Debugging

Check the merge workflow status
here

huydhn · 2023-10-30T07:03:13Z

@huydhn No worries, I just disabled this test under ROCM. If rocm mode is valuable, could we run these tests by default so that we don't have to revert diffs?

We have a smaller number of ROCM runners than Linux CUDA. So this is a trade-off between CI coverage and queuing time. There is a work in progress #110511 to run only a subset of tests on ROCm for every PR. Adding dynamo tests in the set could help avoid similar cases.

jansel · 2023-10-30T17:39:35Z

Let's just disable the test on rocm.

oulgen · 2023-10-30T17:40:27Z

@jansel already done and pushed with the disable

…s with constant/simple grids (pytorch#112228) Pull Request resolved: pytorch#112228 Approved by: https://github.com/jansel

…n kernels with constant/simple grids (pytorch#112228)" This reverts commit dbb31a2. Reverted pytorch#112228 on behalf of https://github.com/huydhn due to Sorry for reverting your change but it is failing ROCm test in trunk https://hud.pytorch.org/pytorch/pytorch/commit/dbb31a2984fa616b4bb6fac7abb2a06ec0533eb1 ([comment](pytorch#112228 (comment)))

…s with constant/simple grids (pytorch#112228) Pull Request resolved: pytorch#112228 Approved by: https://github.com/jansel

…n kernels with constant/simple grids (pytorch#112228)" This reverts commit dbb31a2. Reverted pytorch#112228 on behalf of https://github.com/huydhn due to Sorry for reverting your change but it is failing ROCm test in trunk https://hud.pytorch.org/pytorch/pytorch/commit/dbb31a2984fa616b4bb6fac7abb2a06ec0533eb1 ([comment](pytorch#112228 (comment)))

…s with constant/simple grids (pytorch#112228) Pull Request resolved: pytorch#112228 Approved by: https://github.com/jansel

…n kernels with constant/simple grids (pytorch#112228)" This reverts commit dbb31a2. Reverted pytorch#112228 on behalf of https://github.com/huydhn due to Sorry for reverting your change but it is failing ROCm test in trunk https://hud.pytorch.org/pytorch/pytorch/commit/dbb31a2984fa616b4bb6fac7abb2a06ec0533eb1 ([comment](pytorch#112228 (comment)))

…s with constant/simple grids (pytorch#112228) Pull Request resolved: pytorch#112228 Approved by: https://github.com/jansel

…s with constant/simple grids (pytorch#112228)

[Inductor] Add triton.autotune support for user defined triton kernel…

c30fb6c

…s with constant/simple grids [ghstack-poisoned]

oulgen added a commit that referenced this pull request Oct 27, 2023

[Inductor] Add triton.autotune support for user defined triton kernel…

7f12b9f

…s with constant/simple grids ghstack-source-id: 37bf2369d049574f82b200610f1f1f35e7120f94 Pull Request resolved: #112228

github-actions bot added module: inductor module: dynamo ciflow/inductor labels Oct 27, 2023

oulgen added a commit that referenced this pull request Oct 27, 2023

[Inductor] Add triton.autotune support for user defined triton kernel…

2186812

…s with constant/simple grids ghstack-source-id: 65a5279fb602e9cb58da1df0a55b9a1c6f853b3a Pull Request resolved: #112228

oulgen requested review from jansel, zou3519 and Chillee October 27, 2023 02:51

oulgen commented Oct 27, 2023

View reviewed changes

oulgen added a commit that referenced this pull request Oct 27, 2023

[Inductor] Add triton.autotune support for user defined triton kernel…

bc7193a

…s with constant/simple grids ghstack-source-id: db33a033857213c6ce6756309174eddd2de8d3fb Pull Request resolved: #112228

oulgen added the ciflow/trunk Trigger trunk jobs on your pull request label Oct 27, 2023

jansel approved these changes Oct 27, 2023

View reviewed changes

oulgen mentioned this pull request Oct 27, 2023

[Inductor] Add triton.autotune support for user defined triton kernels with complex grids #112290

Closed

oulgen added the release notes: inductor label Oct 27, 2023

oulgen mentioned this pull request Oct 27, 2023

Support calling user defined triton kernels with kernel.run #112292

Closed

pytorchmergebot added the merging label Oct 27, 2023

pytorchmergebot added Merged and removed merging labels Oct 27, 2023

pytorchmergebot closed this in dbb31a2 Oct 27, 2023

huydhn added the ciflow/rocm label Oct 28, 2023

pytorchmergebot added the Reverted label Oct 28, 2023

oulgen reopened this Oct 28, 2023

pytorchmergebot added the merging label Oct 28, 2023

pytorchmergebot removed the merging label Oct 28, 2023

pytorchmergebot closed this in c14c4ef Oct 28, 2023

facebook-github-bot deleted the gh/oulgen/17/head branch November 1, 2023 14:25

kit1980 removed the Reverted label Nov 1, 2023

Stonepia added a commit to Stonepia/pytorch that referenced this pull request Jan 31, 2024

[Inductor] Add triton.autotune support for user defined triton kernel…

d6b1e2a

…s with constant/simple grids (pytorch#112228)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Inductor] Add triton.autotune support for user defined triton kernels with constant/simple grids #112228

[Inductor] Add triton.autotune support for user defined triton kernels with constant/simple grids #112228

oulgen commented Oct 27, 2023 •

edited

pytorch-bot bot commented Oct 27, 2023 •

edited

oulgen commented Oct 27, 2023

oulgen Oct 27, 2023

oulgen commented Oct 27, 2023

pytorchmergebot commented Oct 27, 2023

huydhn commented Oct 28, 2023

pytorchmergebot commented Oct 28, 2023

pytorchmergebot commented Oct 28, 2023

oulgen commented Oct 28, 2023

oulgen commented Oct 28, 2023

pytorchmergebot commented Oct 28, 2023

huydhn commented Oct 30, 2023

jansel commented Oct 30, 2023

oulgen commented Oct 30, 2023

		@@ -362,16 +363,17 @@ def kernel_call():
		{dict(zip(self.arg_names, args)), launcher.config.kwargs}

[Inductor] Add triton.autotune support for user defined triton kernels with constant/simple grids #112228

[Inductor] Add triton.autotune support for user defined triton kernels with constant/simple grids #112228

Conversation

oulgen commented Oct 27, 2023 • edited

pytorch-bot bot commented Oct 27, 2023 • edited

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/112228

✅ No Failures

oulgen commented Oct 27, 2023

oulgen Oct 27, 2023

Choose a reason for hiding this comment

oulgen commented Oct 27, 2023

pytorchmergebot commented Oct 27, 2023

Merge started

huydhn commented Oct 28, 2023

pytorchmergebot commented Oct 28, 2023

pytorchmergebot commented Oct 28, 2023

oulgen commented Oct 28, 2023

oulgen commented Oct 28, 2023

pytorchmergebot commented Oct 28, 2023

Merge started

huydhn commented Oct 30, 2023

jansel commented Oct 30, 2023

oulgen commented Oct 30, 2023

oulgen commented Oct 27, 2023 •

edited

pytorch-bot bot commented Oct 27, 2023 •

edited