[dynamo] Support namedtuple subclass #153982

StrongerXi · 2025-05-20T21:08:46Z

Stack from ghstack (oldest at bottom):

Fixes #133762. This involves

support tuple subclass constructed inside compile region.
handle the "fake" global scope associated with NamedTuple-generated
__new__.
handle namedtuple._tuplegetter more faithfully.

cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @chenyang78 @kadeng @chauhang @amjames

Differential Revision: D75488091

Fixes #133762. We just need to 1. handle the "fake" global scope associated with NamedTuple generated `__new__`. 2. handle `namedtuple._tuplegetter` more faithfully. [ghstack-poisoned]

pytorch-bot · 2025-05-20T21:08:50Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/153982

📄 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.

✅ You can merge normally! (3 Unrelated Failures)

As of commit 4c25a19 with merge base 7ae204c ():

FLAKY - The following jobs failed but were likely due to flakiness present on trunk:

inductor / linux-jammy-cpu-py3.9-gcc11-inductor / test (cpu_inductor_torchbench, 2, 2, linux.8xlarge.amx) (gh) (similar failure)
opacus_cifar10
inductor / linux-jammy-cpu-py3.9-gcc11-inductor / test (dynamic_cpu_inductor_torchbench, 2, 2, linux.8xlarge.amx) (gh) (similar failure)
opacus_cifar10

BROKEN TRUNK - The following job failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

inductor / cuda12.8-py3.10-gcc9-sm86 / test (inductor_torchbench, 2, 2, linux.g5.4xlarge.nvidia.gpu) (gh) (trunk failure)
opacus_cifar10

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

Fixes #133762. We just need to 1. handle the "fake" global scope associated with NamedTuple generated `__new__`. 2. handle `namedtuple._tuplegetter` more faithfully. ghstack-source-id: 118f9d0 Pull Request resolved: #153982

Fixes #133762. We just need to 1. handle the "fake" global scope associated with NamedTuple generated `__new__`. 2. handle `namedtuple._tuplegetter` more faithfully. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames [ghstack-poisoned]

Fixes #133762. This involves 1. support tuple subclass constructed inside compile region. 2. handle the "fake" global scope associated with NamedTuple-generated `__new__`. 3. handle `namedtuple._tuplegetter` more faithfully. 4. use `object.__getattribute__(obj, "__dict__")[key] = value` to replay side effects onto pre-existing user-defined objects, because `object.__setattr__` can still trigger user code on via descriptors setter. ghstack-source-id: c61116a Pull Request resolved: #153982

Fixes #133762. This involves 1. support tuple subclass constructed inside compile region. 2. handle the "fake" global scope associated with NamedTuple-generated `__new__`. 3. handle `namedtuple._tuplegetter` more faithfully. 4. use `object.__getattribute__(obj, "__dict__")[key] = value` to replay side effects onto pre-existing user-defined objects, because `object.__setattr__` can still trigger user code on via descriptors setter. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames [ghstack-poisoned]

Fixes #133762. This involves 1. support tuple subclass constructed inside compile region. 2. handle the "fake" global scope associated with NamedTuple-generated `__new__`. 3. handle `namedtuple._tuplegetter` more faithfully. 4. use `object.__getattribute__(obj, "__dict__")[key] = value` to replay side effects onto pre-existing user-defined objects, because `object.__setattr__` can still trigger user code on via descriptors setter. 5. fixing `is` and `is not` handling when `GetAttrVariable` is involved. This was exposed by a numpy test case under dynamo: test_broadcast_single_arg, see comments in the patch for more. ghstack-source-id: 8bbe65b Pull Request resolved: #153982

Fixes #133762. This involves 1. support tuple subclass constructed inside compile region. 2. handle the "fake" global scope associated with NamedTuple-generated `__new__`. 3. handle `namedtuple._tuplegetter` more faithfully. 4. use `object.__getattribute__(obj, "__dict__")[key] = value` to replay side effects onto pre-existing user-defined objects, because `object.__setattr__` can still trigger user code on via descriptors setter. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames [ghstack-poisoned]

Fixes #133762. This involves 1. support tuple subclass constructed inside compile region. 2. handle the "fake" global scope associated with NamedTuple-generated `__new__`. 3. handle `namedtuple._tuplegetter` more faithfully. 4. use `object.__getattribute__(obj, "__dict__")[key] = value` to replay side effects onto pre-existing user-defined objects, because `object.__setattr__` can still trigger user code on via descriptors setter. 5. fixing `is` and `is not` handling when `GetAttrVariable` is involved. This was exposed by a numpy test case under dynamo: test_broadcast_single_arg, see comments in the patch for more. ghstack-source-id: 594b14c Pull Request resolved: #153982

Fixes #133762. This involves 1. support tuple subclass constructed inside compile region. 2. handle the "fake" global scope associated with NamedTuple-generated `__new__`. 3. handle `namedtuple._tuplegetter` more faithfully. 4. use `object.__getattribute__(obj, "__dict__")[key] = value` to replay side effects onto pre-existing user-defined objects, because `object.__setattr__` can still trigger user code on via descriptors setter. 5. fixing `is` and `is not` handling when `GetAttrVariable` is involved. This was exposed by a numpy test case under dynamo: test_broadcast_single_arg, see comments in the patch for more. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames [ghstack-poisoned]

Fixes #133762. This involves 1. support tuple subclass constructed inside compile region. 2. handle the "fake" global scope associated with NamedTuple-generated `__new__`. 3. handle `namedtuple._tuplegetter` more faithfully. 4. use `object.__getattribute__(obj, "__dict__")[key] = value` to replay side effects onto pre-existing user-defined objects, because `object.__setattr__` can still trigger user code on via descriptors setter. ghstack-source-id: 4a18e55 Pull Request resolved: #153982

Fixes #133762. This involves 1. support tuple subclass constructed inside compile region. 2. handle the "fake" global scope associated with NamedTuple-generated `__new__`. 3. handle `namedtuple._tuplegetter` more faithfully. 4. use `object.__getattribute__(obj, "__dict__")[key] = value` to replay side effects onto pre-existing user-defined objects, because `object.__setattr__` can still trigger user code on via descriptors setter. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames [ghstack-poisoned]

Fixes #133762. This involves 1. support tuple subclass constructed inside compile region. 2. handle the "fake" global scope associated with NamedTuple-generated `__new__`. 3. handle `namedtuple._tuplegetter` more faithfully. 4. use `object.__getattribute__(obj, "__dict__")[key] = value` to replay side effects onto pre-existing user-defined objects, because `object.__setattr__` can still trigger user code on via descriptors setter. ghstack-source-id: e886a52 Pull Request resolved: #153982

Fixes #133762. This involves 1. support tuple subclass constructed inside compile region. 2. handle the "fake" global scope associated with NamedTuple-generated `__new__`. 3. handle `namedtuple._tuplegetter` more faithfully. 4. use `object.__getattribute__(obj, "__dict__")[key] = value` to replay side effects onto pre-existing user-defined objects, because `object.__setattr__` can still trigger user code on via descriptors setter. cc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang amjames [ghstack-poisoned]

Fixes #133762. This involves 1. support tuple subclass constructed inside compile region. 2. handle the "fake" global scope associated with NamedTuple-generated `__new__`. 3. handle `namedtuple._tuplegetter` more faithfully. 4. use `object.__getattribute__(obj, "__dict__")[key] = value` to replay side effects onto pre-existing user-defined objects, because `object.__setattr__` can still trigger user code on via descriptors setter. ghstack-source-id: 5bd1192 Pull Request resolved: #153982

StrongerXi · 2025-05-27T23:15:08Z

@StrongerXi has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

StrongerXi · 2025-05-30T16:06:20Z

@pytorchbot merge

pytorchmergebot · 2025-05-30T16:08:29Z

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

Fixes pytorch#133762. This involves 1. support tuple subclass constructed inside compile region. 2. handle the "fake" global scope associated with NamedTuple-generated `__new__`. 3. handle `namedtuple._tuplegetter` more faithfully. Differential Revision: [D75488091](https://our.internmc.facebook.com/intern/diff/D75488091) Pull Request resolved: pytorch#153982 Approved by: https://github.com/jansel ghstack dependencies: pytorch#154176

[dynamo] Support namedtuple subclass

af8f482

Fixes #133762. We just need to 1. handle the "fake" global scope associated with NamedTuple generated `__new__`. 2. handle `namedtuple._tuplegetter` more faithfully. [ghstack-poisoned]

StrongerXi mentioned this pull request May 20, 2025

[dynamo] Support tuple subclass constructed within compile region #153981

Closed

pytorch-bot bot added ciflow/inductor module: dynamo labels May 20, 2025

StrongerXi linked an issue May 20, 2025 that may be closed by this pull request

Cannot override __add__ in NamedTuple with __new__ + torch.compile #133762

Closed

StrongerXi added the release notes: dynamo label May 20, 2025

StrongerXi mentioned this pull request May 23, 2025

[dynamo] Trace into descriptor with __set__ #154176

Closed

pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label May 27, 2025

StrongerXi requested review from jansel and anijain2305 and removed request for jansel May 28, 2025 04:53

jansel approved these changes May 28, 2025

View reviewed changes

pytorchmergebot added the merging label May 30, 2025

pytorchmergebot added the Merged label May 30, 2025

pytorchmergebot closed this in 7183f52 May 30, 2025

pytorchmergebot removed the merging label May 30, 2025

github-actions bot deleted the gh/StrongerXi/116/head branch July 2, 2025 02:19

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[dynamo] Support namedtuple subclass #153982

[dynamo] Support namedtuple subclass #153982

Uh oh!

StrongerXi commented May 20, 2025 •

edited

Loading

Uh oh!

pytorch-bot bot commented May 20, 2025 •

edited

Loading

Uh oh!

StrongerXi commented May 27, 2025

Uh oh!

StrongerXi commented May 30, 2025

Uh oh!

pytorchmergebot commented May 30, 2025

Uh oh!

Uh oh!

[dynamo] Support namedtuple subclass #153982

[dynamo] Support namedtuple subclass #153982

Uh oh!

Conversation

StrongerXi commented May 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented May 20, 2025 • 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/153982

✅ You can merge normally! (3 Unrelated Failures)

Uh oh!

StrongerXi commented May 27, 2025

Uh oh!

StrongerXi commented May 30, 2025

Uh oh!

pytorchmergebot commented May 30, 2025

Merge started

Uh oh!

Uh oh!

StrongerXi commented May 20, 2025 •

edited

Loading

pytorch-bot bot commented May 20, 2025 •

edited

Loading