Added some tests to prevent regressions in partitioning and flexattention #136826

Chillee · 2024-09-27T02:32:07Z

Stack from ghstack (oldest at bottom):

Add type annotations for higher order ops/flex_attention #137065
set FlexAttention devices properly during tracing #137049
Allow automatic dynamic shapes for closures and set current node properly in flexattention subgraph lowering #137043
-> Added some tests to prevent regressions in partitioning and flexattention #136826

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

…tion [ghstack-poisoned]

pytorch-bot · 2024-09-27T02:32:12Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/136826

📄 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.

❌ 2 New Failures

As of commit 86e03a9 with merge base b825848 ():

NEW FAILURES - The following jobs have failed:

trunk / linux-focal-rocm6.2-py3.10 / test (default, 1, 2, linux.rocm.gpu) (gh)
test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_enable_gqa_False_n_heads0_cuda_bfloat16
trunk / linux-focal-rocm6.2-py3.10 / test (distributed, 1, 1, linux.rocm.gpu) (gh)
test_transformers.py::TestSDPACudaOnlyCUDA::test_flash_attention_vs_math_ref_grads_batch_size_1_seq_len_q_143_seq_len_k_2048_head_dim_203_is_causal_False_dropout_p_0_0_bfloat16_scale_l1_enable_gqa_False_n_heads0_cuda_bfloat16

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

yanboliang · 2024-09-27T04:17:41Z

test/inductor/test_flex_attention.py

        out.sum().backward()

+    @supported_platform
+    def test_compiled_block_mask(self):


should we move this under class TestBlockMask? since it's a block mask specific test?

Oh, I didn't know we had that section lol

…d flexattention" cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]

drisspg · 2024-09-27T16:48:15Z

test/inductor/test_perf.py

        inp = (T(10, grad=True), T(10, grad=True))
        self.assertExpectedInline(count_numel_train(f, *inp), """70""")

+    def test_partitioning_relu(self):


is this related?

Chillee · 2024-09-27T17:32:29Z

@pytorchbot merge

pytorchmergebot · 2024-09-27T17:34:21Z

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

pytorchmergebot · 2024-09-27T17:39:47Z

Merge failed

Reason: Command git -C /home/runner/work/pytorch/pytorch cherry-pick -x ee3ef1fc0aff6d953867bc475acf37d2184d2725 returned non-zero exit code 1

Auto-merging test/inductor/test_flex_attention.py
CONFLICT (content): Merge conflict in test/inductor/test_flex_attention.py
error: could not apply ee3ef1fc0a... Added some tests to prevent regressions in partitioning and flexattention
hint: After resolving the conflicts, mark them with
hint: "git add/rm <pathspec>", then run
hint: "git cherry-pick --continue".
hint: You can instead skip this commit with "git cherry-pick --skip".
hint: To abort and get back to the state before "git cherry-pick",
hint: run "git cherry-pick --abort".
hint: Disable this message with "git config advice.mergeConflict false"

Details for Dev Infra team

Raised by workflow job

…d flexattention" cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]

…tion ghstack-source-id: 2295a37 Pull Request resolved: #136826

Chillee · 2024-09-30T19:02:38Z

@pytorchbot merge

pytorchmergebot · 2024-09-30T19:05:09Z

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

pytorchmergebot · 2024-09-30T19:10:28Z

Merge failed

Reason: Command git -C /home/runner/work/pytorch/pytorch cherry-pick -x 6fd6537c760979a35b51e8fc82701fb7afda0da1 returned non-zero exit code 1

Auto-merging test/inductor/test_flex_attention.py
CONFLICT (content): Merge conflict in test/inductor/test_flex_attention.py
error: could not apply 6fd6537c76... Added some tests to prevent regressions in partitioning and flexattention
hint: After resolving the conflicts, mark them with
hint: "git add/rm <pathspec>", then run
hint: "git cherry-pick --continue".
hint: You can instead skip this commit with "git cherry-pick --skip".
hint: To abort and get back to the state before "git cherry-pick",
hint: run "git cherry-pick --abort".
hint: Disable this message with "git config advice.mergeConflict false"

Details for Dev Infra team

Raised by workflow job

…d flexattention" cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang [ghstack-poisoned]

Chillee · 2024-09-30T22:48:22Z

@pytorchbot merge -i

pytorchmergebot · 2024-09-30T22:50:02Z

Merge started

Your change will be merged while ignoring the following 2 checks: trunk / linux-focal-rocm6.2-py3.10 / test (default, 1, 2, linux.rocm.gpu), trunk / linux-focal-rocm6.2-py3.10 / test (distributed, 1, 1, linux.rocm.gpu)

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

…erly in flexattention subgraph lowering (#137043) Pull Request resolved: #137043 Approved by: https://github.com/drisspg ghstack dependencies: #136826

Pull Request resolved: #137049 Approved by: https://github.com/zou3519, https://github.com/drisspg, https://github.com/yanboliang ghstack dependencies: #136826, #137043

Pull Request resolved: #137065 Approved by: https://github.com/drisspg, https://github.com/Skylion007 ghstack dependencies: #136826, #137043, #137049

…tion (pytorch#136826) Pull Request resolved: pytorch#136826 Approved by: https://github.com/yanboliang, https://github.com/drisspg

…erly in flexattention subgraph lowering (pytorch#137043) Pull Request resolved: pytorch#137043 Approved by: https://github.com/drisspg ghstack dependencies: pytorch#136826

Pull Request resolved: pytorch#137049 Approved by: https://github.com/zou3519, https://github.com/drisspg, https://github.com/yanboliang ghstack dependencies: pytorch#136826, pytorch#137043

) Pull Request resolved: pytorch#137065 Approved by: https://github.com/drisspg, https://github.com/Skylion007 ghstack dependencies: pytorch#136826, pytorch#137043, pytorch#137049

I think this should basically make no difference numerically, but it does have some ramifications on things like CSE. Pull Request resolved: #137079 Approved by: https://github.com/Skylion007 ghstack dependencies: #136826, #137043, #137049, #137065

…tion ghstack-source-id: 3d8895f Pull Request resolved: pytorch/pytorch#136826

Added some tests to prevent regressions in partitioning and flexatten…

6ebac55

…tion [ghstack-poisoned]

pytorch-bot bot added module: inductor topic: not user facing topic category labels Sep 27, 2024

github-actions bot requested a review from ezyang September 27, 2024 02:32

Chillee requested review from drisspg and yanboliang September 27, 2024 02:33

yanboliang approved these changes Sep 27, 2024

View reviewed changes

Chillee added the ciflow/trunk Trigger trunk jobs on your pull request label Sep 27, 2024

drisspg reviewed Sep 27, 2024

View reviewed changes

drisspg approved these changes Sep 27, 2024

View reviewed changes

pytorchmergebot added the merging label Sep 27, 2024

pytorchmergebot removed the merging label Sep 27, 2024

Chillee added a commit that referenced this pull request Sep 28, 2024

Added some tests to prevent regressions in partitioning and flexatten…

6fd6537

…tion ghstack-source-id: 2295a37 Pull Request resolved: #136826

pytorchmergebot added the merging label Sep 30, 2024

pytorchmergebot removed the merging label Sep 30, 2024

This was referenced Sep 30, 2024

Allow automatic dynamic shapes for closures and set current node properly in flexattention subgraph lowering #137043

Closed

set FlexAttention devices properly during tracing #137048

Closed

Chillee mentioned this pull request Sep 30, 2024

set FlexAttention devices properly during tracing #137049

Closed

pytorchmergebot added the merging label Sep 30, 2024

Chillee mentioned this pull request Oct 1, 2024

Add type annotations for higher order ops/flex_attention #137065

Closed

pytorchmergebot added the Merged label Oct 1, 2024

pytorchmergebot closed this in d9c400b Oct 1, 2024

pytorchmergebot removed the merging label Oct 1, 2024

This was referenced Oct 1, 2024

try adding the norm_input_output_chooser pass #137074

Closed

Fix typo in _normalize ref #137079

Closed

github-actions bot deleted the gh/chillee/355/head branch November 29, 2024 02:11

Esquains pushed a commit to Esquains/study1 that referenced this pull request Dec 15, 2024

Added some tests to prevent regressions in partitioning and flexatten…

cf718f9

…tion ghstack-source-id: 3d8895f Pull Request resolved: pytorch/pytorch#136826

Added some tests to prevent regressions in partitioning and flexattention #136826

Added some tests to prevent regressions in partitioning and flexattention #136826

Uh oh!

Conversation

Chillee commented Sep 27, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Sep 27, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/136826

❌ 2 New Failures

Uh oh!

yanboliang Sep 27, 2024

Choose a reason for hiding this comment

Uh oh!

Chillee Sep 27, 2024

Choose a reason for hiding this comment

Uh oh!

drisspg Sep 27, 2024

Choose a reason for hiding this comment

Uh oh!

Chillee Sep 30, 2024

Choose a reason for hiding this comment

Uh oh!

Chillee commented Sep 27, 2024

Uh oh!

pytorchmergebot commented Sep 27, 2024

Merge started

Uh oh!

pytorchmergebot commented Sep 27, 2024

Merge failed

Uh oh!

Chillee commented Sep 30, 2024

Uh oh!

pytorchmergebot commented Sep 30, 2024

Merge started

Uh oh!

pytorchmergebot commented Sep 30, 2024

Merge failed

Uh oh!

Chillee commented Sep 30, 2024

Uh oh!

pytorchmergebot commented Sep 30, 2024

Merge started

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Chillee commented Sep 27, 2024 •

edited

Loading

pytorch-bot bot commented Sep 27, 2024 •

edited

Loading