[inductor][triton] support JITCallable._hash_lock #161768
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
Fixes #161618 Triton # 7974 introduces a threading.RLock() in JITCallable, which is not pickle-able. This PR adds this field to the list of un-pickleable fields that need to be handled specially. [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/161768
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ❌ 2 New FailuresAs of commit 3a68946 with merge base a6456bf ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
xuzhao9
approved these changes
Aug 29, 2025
Fixes #161618 Triton # 7974 introduces a threading.RLock() in JITCallable, which is not pickle-able. This PR adds this field to the list of un-pickleable fields that need to be handled specially. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 amjames chauhang aakhundov coconutruben mlazos [ghstack-poisoned]
Fixes #161618 Triton # 7974 introduces a threading.RLock() in JITCallable, which is not pickle-able. This PR adds this field to the list of un-pickleable fields that need to be handled specially. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 amjames chauhang aakhundov coconutruben mlazos [ghstack-poisoned]
|
@pytorchbot merge -f "windows build errors are unrelated" |
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
davidberard98
added a commit
that referenced
this pull request
Sep 5, 2025
Follow-up to #161768. It turns out that we do need to restore the _hash_lock, even in the non-triton_bundler case, for the MultiKernel case, where the hash lock is used. To do this, we add `restore_after_unpickle()` which will restore fields (or if the old fields are not provided, initialize just the hash_lock) [ghstack-poisoned]
davidberard98
added a commit
that referenced
this pull request
Sep 5, 2025
Follow-up to #161768. It turns out that we do need to restore the _hash_lock, even in the non-triton_bundler case, for the MultiKernel case, where the hash lock is used. To do this, we add `restore_after_unpickle()` which will restore fields (or if the old fields are not provided, initialize just the hash_lock) ghstack-source-id: 8603f81 Pull Request resolved: #162244
davidberard98
added a commit
that referenced
this pull request
Sep 5, 2025
…lock support" Follow-up to #161768. It turns out that we do need to restore the _hash_lock, even in the non-triton_bundler case, for the MultiKernel case, where the hash lock is used. To do this, we add `restore_after_unpickle()` which will restore fields (or if the old fields are not provided, initialize just the hash_lock) cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 amjames chauhang aakhundov coconutruben [ghstack-poisoned]
davidberard98
added a commit
that referenced
this pull request
Sep 5, 2025
Follow-up to #161768. It turns out that we do need to restore the _hash_lock, even in the non-triton_bundler case, for the MultiKernel case, where the hash lock is used. To do this, we add `restore_after_unpickle()` which will restore fields (or if the old fields are not provided, initialize just the hash_lock) cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 amjames chauhang aakhundov coconutruben [ghstack-poisoned]
davidberard98
added a commit
that referenced
this pull request
Sep 5, 2025
Follow-up to #161768. It turns out that we do need to restore the _hash_lock, even in the non-triton_bundler case, for the MultiKernel case, where the hash lock is used. To do this, we add `restore_after_unpickle()` which will restore fields (or if the old fields are not provided, initialize just the hash_lock) ghstack-source-id: ec3e66c Pull Request resolved: #162244
davidberard98
added a commit
that referenced
this pull request
Sep 5, 2025
…lock support" Follow-up to #161768. Context: ProcessPool pickles the outputs before sending them back to the main process. Triton kernels have some un-pickleable fields, so `prepare_for_pickle()` is used to strip out those fields. Previously, in the standard case (without triton_bundler.py), `prepare_for_pickle()` would strip out the un-pickleable fields and they would never be added back after unpickling, because the un-pickleable fields were not actually needed after compilation finished. In #161768 updated `prepare_for_pickle` to also strip out the `fn._hash_lock` field, a newly added field in JITCallable instances which is a `threading.RLock()`, which is not pickleable. It turns out that we do need to restore the `fn._hash_lock` field, even in the non-triton_bundler case - the MultiKernel case uses the hash lock. To do this, we add `restore_after_unpickle()` which will restore fields (or if the old fields are not provided, initialize just the hash_lock) Compile time benchmarks look good, maybe a very minor regression (see the comment below on the PR) cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 amjames chauhang aakhundov coconutruben [ghstack-poisoned]
davidberard98
added a commit
that referenced
this pull request
Sep 5, 2025
Follow-up to #161768. Context: ProcessPool pickles the outputs before sending them back to the main process. Triton kernels have some un-pickleable fields, so `prepare_for_pickle()` is used to strip out those fields. Previously, in the standard case (without triton_bundler.py), `prepare_for_pickle()` would strip out the un-pickleable fields and they would never be added back after unpickling, because the un-pickleable fields were not actually needed after compilation finished. In #161768 updated `prepare_for_pickle` to also strip out the `fn._hash_lock` field, a newly added field in JITCallable instances which is a `threading.RLock()`, which is not pickleable. It turns out that we do need to restore the `fn._hash_lock` field, even in the non-triton_bundler case - the MultiKernel case uses the hash lock. To do this, we add `restore_after_unpickle()` which will restore fields (or if the old fields are not provided, initialize just the hash_lock) Compile time benchmarks look good, maybe a very minor regression (see the comment below on the PR) cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy chenyang78 kadeng muchulee8 amjames chauhang aakhundov coconutruben [ghstack-poisoned]
pytorchmergebot
pushed a commit
that referenced
this pull request
Sep 8, 2025
Follow-up to #161768. Context: ProcessPool pickles the outputs before sending them back to the main process. Triton kernels have some un-pickleable fields, so `prepare_for_pickle()` is used to strip out those fields. Previously, in the standard case (without triton_bundler.py), `prepare_for_pickle()` would strip out the un-pickleable fields and they would never be added back after unpickling, because the un-pickleable fields were not actually needed after compilation finished. In #161768 updated `prepare_for_pickle` to also strip out the `fn._hash_lock` field, a newly added field in JITCallable instances which is a `threading.RLock()`, which is not pickleable. It turns out that we do need to restore the `fn._hash_lock` field, even in the non-triton_bundler case - the MultiKernel case uses the hash lock. To do this, we add `restore_after_unpickle()` which will restore fields (or if the old fields are not provided, initialize just the hash_lock) Compile time benchmarks look good, maybe a very minor regression (see the comment below on the PR) Pull Request resolved: #162244 Approved by: https://github.com/atalman
markc-614
pushed a commit
to markc-614/pytorch
that referenced
this pull request
Sep 17, 2025
Fixes pytorch#161618 Triton # 7974 introduces a threading.RLock() in JITCallable, which is not pickle-able. This PR adds this field to the list of un-pickleable fields that need to be handled specially. Pull Request resolved: pytorch#161768 Approved by: https://github.com/xuzhao9
markc-614
pushed a commit
to markc-614/pytorch
that referenced
this pull request
Sep 17, 2025
Follow-up to pytorch#161768. Context: ProcessPool pickles the outputs before sending them back to the main process. Triton kernels have some un-pickleable fields, so `prepare_for_pickle()` is used to strip out those fields. Previously, in the standard case (without triton_bundler.py), `prepare_for_pickle()` would strip out the un-pickleable fields and they would never be added back after unpickling, because the un-pickleable fields were not actually needed after compilation finished. In pytorch#161768 updated `prepare_for_pickle` to also strip out the `fn._hash_lock` field, a newly added field in JITCallable instances which is a `threading.RLock()`, which is not pickleable. It turns out that we do need to restore the `fn._hash_lock` field, even in the non-triton_bundler case - the MultiKernel case uses the hash lock. To do this, we add `restore_after_unpickle()` which will restore fields (or if the old fields are not provided, initialize just the hash_lock) Compile time benchmarks look good, maybe a very minor regression (see the comment below on the PR) Pull Request resolved: pytorch#162244 Approved by: https://github.com/atalman
mansiag05
pushed a commit
to mansiag05/pytorch
that referenced
this pull request
Sep 22, 2025
Fixes pytorch#161618 Triton # 7974 introduces a threading.RLock() in JITCallable, which is not pickle-able. This PR adds this field to the list of un-pickleable fields that need to be handled specially. Pull Request resolved: pytorch#161768 Approved by: https://github.com/xuzhao9
mansiag05
pushed a commit
to mansiag05/pytorch
that referenced
this pull request
Sep 22, 2025
Follow-up to pytorch#161768. Context: ProcessPool pickles the outputs before sending them back to the main process. Triton kernels have some un-pickleable fields, so `prepare_for_pickle()` is used to strip out those fields. Previously, in the standard case (without triton_bundler.py), `prepare_for_pickle()` would strip out the un-pickleable fields and they would never be added back after unpickling, because the un-pickleable fields were not actually needed after compilation finished. In pytorch#161768 updated `prepare_for_pickle` to also strip out the `fn._hash_lock` field, a newly added field in JITCallable instances which is a `threading.RLock()`, which is not pickleable. It turns out that we do need to restore the `fn._hash_lock` field, even in the non-triton_bundler case - the MultiKernel case uses the hash lock. To do this, we add `restore_after_unpickle()` which will restore fields (or if the old fields are not provided, initialize just the hash_lock) Compile time benchmarks look good, maybe a very minor regression (see the comment below on the PR) Pull Request resolved: pytorch#162244 Approved by: https://github.com/atalman
cleonard530
pushed a commit
to cleonard530/pytorch
that referenced
this pull request
Sep 22, 2025
Follow-up to pytorch#161768. Context: ProcessPool pickles the outputs before sending them back to the main process. Triton kernels have some un-pickleable fields, so `prepare_for_pickle()` is used to strip out those fields. Previously, in the standard case (without triton_bundler.py), `prepare_for_pickle()` would strip out the un-pickleable fields and they would never be added back after unpickling, because the un-pickleable fields were not actually needed after compilation finished. In pytorch#161768 updated `prepare_for_pickle` to also strip out the `fn._hash_lock` field, a newly added field in JITCallable instances which is a `threading.RLock()`, which is not pickleable. It turns out that we do need to restore the `fn._hash_lock` field, even in the non-triton_bundler case - the MultiKernel case uses the hash lock. To do this, we add `restore_after_unpickle()` which will restore fields (or if the old fields are not provided, initialize just the hash_lock) Compile time benchmarks look good, maybe a very minor regression (see the comment below on the PR) Pull Request resolved: pytorch#162244 Approved by: https://github.com/atalman
dsashidh
pushed a commit
to dsashidh/pytorch
that referenced
this pull request
Sep 26, 2025
Follow-up to pytorch#161768. Context: ProcessPool pickles the outputs before sending them back to the main process. Triton kernels have some un-pickleable fields, so `prepare_for_pickle()` is used to strip out those fields. Previously, in the standard case (without triton_bundler.py), `prepare_for_pickle()` would strip out the un-pickleable fields and they would never be added back after unpickling, because the un-pickleable fields were not actually needed after compilation finished. In pytorch#161768 updated `prepare_for_pickle` to also strip out the `fn._hash_lock` field, a newly added field in JITCallable instances which is a `threading.RLock()`, which is not pickleable. It turns out that we do need to restore the `fn._hash_lock` field, even in the non-triton_bundler case - the MultiKernel case uses the hash lock. To do this, we add `restore_after_unpickle()` which will restore fields (or if the old fields are not provided, initialize just the hash_lock) Compile time benchmarks look good, maybe a very minor regression (see the comment below on the PR) Pull Request resolved: pytorch#162244 Approved by: https://github.com/atalman
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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):
Fixes #161618
Triton # 7974 introduces a threading.RLock() in JITCallable, which is not pickle-able. This PR adds this field to the list of un-pickleable fields that need to be handled specially.
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @chenyang78 @kadeng @muchulee8 @amjames @chauhang @aakhundov @coconutruben @mlazos