Parallelize epilogue/prologue benchmarking #143408
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/143408
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 04386ff with merge base c986eba ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang aakhundov [ghstack-poisoned]
cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang aakhundov [ghstack-poisoned]
eellison
changed the title
|
@pytorchbot merge |
Merge startedYour 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 |
pytorchmergebot
pushed a commit
that referenced
this pull request
Jan 28, 2025
Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): Previously, we would finalize the config of a triton template after its first fusion. this maintains multiple configs, in case we epilogue fuse, then prologue fuse, and prologue fusion has a new better config. Pull Request resolved: #145103 Approved by: https://github.com/jansel, https://github.com/shunting314 ghstack dependencies: #143408
jamesjwu
added a commit
that referenced
this pull request
Feb 11, 2025
Fix a bug introduced by D69123174: because triton kernels now are returned directly by the worker, each future created by the triton kernel should only be used once per compile. Otherwise, a long running process that does something like in :
```
compiled_1 = torch.compile("max-autotune", fullgraph=True)(fn)
# run compiled_1
out_compiled = compiled_1
compiled_2 = torch.compile("max-autotune", fullgraph=True)(fn2)
```
Where fn1 and fn2 are very similar (i.e. same source code) would result in us using the launcher for the first autotuning run, and setting the launcher to None after running, and then using the same future again.
This does not remove the caching support for @eellison's caching for prologue benchmarking, because that happens under the same compile: #143408
Differential Revision: [D69476856](https://our.internmc.facebook.com/intern/diff/D69476856/)
**NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D69476856/)!
[ghstack-poisoned]
jamesjwu
added a commit
that referenced
this pull request
Feb 11, 2025
Fix a bug introduced by D69123174: because triton kernels now are returned directly by the worker, each future created by the triton kernel should only be used once per compile. Otherwise, a long running process that does something like in :
```
compiled_1 = torch.compile("max-autotune", fullgraph=True)(fn)
# run compiled_1
out_compiled = compiled_1
compiled_2 = torch.compile("max-autotune", fullgraph=True)(fn2)
```
Where fn1 and fn2 are very similar (i.e. same source code) would result in us using the launcher for the first autotuning run, and setting the launcher to None after running, and then using the same future again.
This does not remove the caching support for eellison's caching for prologue benchmarking, because that happens under the same compile: #143408
Differential Revision: [D69476856](https://our.internmc.facebook.com/intern/diff/D69476856/)
**NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D69476856/)!
ghstack-source-id: 265875622
Pull Request resolved: #146925
jamesjwu
added a commit
that referenced
this pull request
Feb 11, 2025
Fix a bug introduced by D69123174: because triton kernels now are returned directly by the worker, each future created by the triton kernel should only be used once per compile. Otherwise, a long running process that does something like in :
```
compiled_1 = torch.compile("max-autotune", fullgraph=True)(fn)
# run compiled_1
out_compiled = compiled_1
compiled_2 = torch.compile("max-autotune", fullgraph=True)(fn2)
```
Where fn1 and fn2 are very similar (i.e. would generate the same triton kernel source code) would result in us using the launcher for the first autotuning run, and setting the launcher to None after running, and then using the same future/kernel again without regenerating the launcher.
Found this bug testing internal inference models.
This does not remove the caching support for eellison's caching for prologue benchmarking, because that happens under the same compile: #143408
Differential Revision: [D69476856](https://our.internmc.facebook.com/intern/diff/D69476856/)
**NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D69476856/)!
cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 amjames chauhang aakhundov
[ghstack-poisoned]
jamesjwu
added a commit
that referenced
this pull request
Feb 11, 2025
…h inductor compile"
Fix a bug introduced by D69123174: because triton kernels now are returned directly by the worker, each future created by the triton kernel should only be used once per compile. Otherwise, a long running process that does something like in :
```
compiled_1 = torch.compile("max-autotune", fullgraph=True)(fn)
# run compiled_1
out_compiled = compiled_1
compiled_2 = torch.compile("max-autotune", fullgraph=True)(fn2)
```
Where fn1 and fn2 are very similar (i.e. would generate the same triton kernel source code) would result in us using the launcher for the first autotuning run, and setting the launcher to None after running, and then using the same future/kernel again without regenerating the launcher.
Found this bug testing internal inference models.
This does not remove the caching support for eellison's caching for prologue benchmarking, because that happens under the same compile: #143408
Differential Revision: [D69476856](https://our.internmc.facebook.com/intern/diff/D69476856/)
**NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D69476856/)!
cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 amjames chauhang aakhundov
[ghstack-poisoned]
jamesjwu
added a commit
that referenced
this pull request
Feb 11, 2025
Fix a bug introduced by D69123174: because triton kernels now are returned directly by the worker, each future created by the triton kernel should only be used once per compile. Otherwise, a long running process that does something like in :
```
compiled_1 = torch.compile("max-autotune", fullgraph=True)(fn)
# run compiled_1
out_compiled = compiled_1
compiled_2 = torch.compile("max-autotune", fullgraph=True)(fn2)
```
Where fn1 and fn2 are very similar (i.e. would generate the same triton kernel source code) would result in us using the launcher for the first autotuning run, and setting the launcher to None after running, and then using the same future/kernel again without regenerating the launcher.
Found this bug testing internal inference models.
This does not remove the caching support for eellison's caching for prologue benchmarking, because that happens under the same compile: #143408
Differential Revision: [D69476856](https://our.internmc.facebook.com/intern/diff/D69476856/)
**NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D69476856/)!
cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 amjames chauhang aakhundov
[ghstack-poisoned]
jamesjwu
added a commit
that referenced
this pull request
Feb 11, 2025
Pull Request resolved: #146925 Fix a bug introduced by D69123174: because triton kernels now are returned directly by the worker, each future created by the triton kernel should only be used once per compile. Otherwise, a long running process that does something like in : ``` compiled_1 = torch.compile("max-autotune", fullgraph=True)(fn) # run compiled_1 out_compiled = compiled_1 compiled_2 = torch.compile("max-autotune", fullgraph=True)(fn2) ``` Where fn1 and fn2 are very similar (i.e. same source code) would result in us using the launcher for the first autotuning run, and setting the launcher to None after running, and then using the same future again. This does not remove the caching support for @eellison's caching for prologue benchmarking, because that happens under the same compile: #143408 ghstack-source-id: 265898129 Differential Revision: [D69476856](https://our.internmc.facebook.com/intern/diff/D69476856/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D69476856/)!
pytorchmergebot
pushed a commit
that referenced
this pull request
Feb 12, 2025
Fix a bug introduced by D69123174: because triton kernels now are returned directly by the worker, each future created by the triton kernel should only be used once per compile. Otherwise, a long running process that does something like in :
```
compiled_1 = torch.compile("max-autotune", fullgraph=True)(fn)
# run compiled_1
out_compiled = compiled_1
compiled_2 = torch.compile("max-autotune", fullgraph=True)(fn2)
```
Where fn1 and fn2 are very similar (i.e. would generate the same triton kernel source code) would result in us using the launcher for the first autotuning run, and setting the launcher to None after running, and then using the same future/kernel again without regenerating the launcher.
Found this bug testing internal inference models.
This does not remove the caching support for @eellison's caching for prologue benchmarking, because that happens under the same compile: #143408
Differential Revision: [D69476856](https://our.internmc.facebook.com/intern/diff/D69476856/)
**NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D69476856/)!
Pull Request resolved: #146925
Approved by: https://github.com/laithsakka, https://github.com/jansel
ghstack dependencies: #146417
desai0007
pushed a commit
to desai0007/test-repo-pytorch
that referenced
this pull request
Feb 26, 2025
ghstack-source-id: 956b795 Pull Request resolved: pytorch/pytorch#143408
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Stack from ghstack (oldest at bottom):
When we attempt prologue or epilogue fusion with a TritonTemplate, we benchmark it at compile time in order to determine profitability. This avoids slowdowns/register spilling, and allows us to pick fusion when a base triton template is slower than cublas but faster when considering an epilogue. However, that fused benchmarking does not do the same async compilation as we do for the base TritonTemplate. The Base TritonTemplate is async compiled during lowering, then later waited on and benchmarked.
This PR extends a similar process to benchmarking fused TritonTemplates in the scheduler. We keep a list of pending fusions which have async compilations. And we resolve any pending fusions a node is in prior to attempting to fuse it with any other node.
Initially, I saw some slowdowns with this because we kick off async compilations of identical fusions in parallel. To address this I added source code caching at the
async_compilelevel (we also already cache benchmark runs, but that would not happen in parallel).Compilation speedups:
This also should let us be a bit more aggressive with either configs, or benchmarking other fusions which are hard to determine profitability of.
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @desertfire @chauhang @aakhundov