Add DTensor dispatch key (but don't do anything special with it yet)#166369
Closed
swolchok wants to merge 12 commits intogh/swolchok/861/basefrom
Closed
Add DTensor dispatch key (but don't do anything special with it yet)#166369swolchok wants to merge 12 commits intogh/swolchok/861/basefrom
swolchok wants to merge 12 commits intogh/swolchok/861/basefrom
Conversation
Just a baby step to make sure we and CI are happy with this before moving on to a combined fast path. [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/166369
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: ✅ No FailuresAs of commit dd7f900 with merge base 84776e1 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This was referenced Oct 7, 2025
Skylion007
approved these changes
Oct 28, 2025
…thing special with it yet)" Just a baby step to make sure we and CI are happy with this before moving on to a combined fast path. [ghstack-poisoned]
Contributor
Author
|
@Skylion007 just want to make sure you saw and are OK with the deletion of the DeferredInit dispatch key. We're out of space. CC @ezyang as well. marking bc-breaking accordingly |
Contributor
Author
|
I think this one might break debug builds as-is because of the checks in python_variable.cpp for DTensor having |
…th it yet)" Just a baby step to make sure we and CI are happy with this before moving on to a combined fast path. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci ezyang gchanan [ghstack-poisoned]
Contributor
Author
Whoops, that shouldn't break until the next PR; this one doesn't remove torch_dispatch from DTensor. (If anyone claimed that these two should be combined, I wouldn't argue.) |
ezyang
approved these changes
Oct 29, 2025
…spatch key (but don't do anything special with it yet)" Just a baby step to make sure we and CI are happy with this before moving on to a combined fast path. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci ezyang gchanan [ghstack-poisoned]
Contributor
Author
|
noting for posterity that CI is currently green |
… dispatch key (but don't do anything special with it yet)" Just a baby step to make sure we and CI are happy with this before moving on to a combined fast path. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci ezyang gchanan [ghstack-poisoned]
…do anything special with it yet)" Just a baby step to make sure we and CI are happy with this before moving on to a combined fast path. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci ezyang gchanan [ghstack-poisoned]
…th it yet)" Just a baby step to make sure we and CI are happy with this before moving on to a combined fast path. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci ezyang gchanan [ghstack-poisoned]
…. fix our "upcall" from Python to DTensor dispatch key to be an explicit redispatch so that the preceding fix doesn't break it. on "Add DTensor dispatch key (but don't do anything special with it yet)" Just a baby step to make sure we and CI are happy with this before moving on to a combined fast path. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci ezyang gchanan [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 5, 2025
This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. ghstack-source-id: 0937067 Pull Request resolved: #167051
swolchok
added a commit
that referenced
this pull request
Nov 6, 2025
…dd C++ fast path for `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 6, 2025
This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. ghstack-source-id: 2f0ea67 Pull Request resolved: #167051
swolchok
added a commit
that referenced
this pull request
Nov 6, 2025
This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. ghstack-source-id: b3451db Pull Request resolved: #167051
swolchok
added a commit
that referenced
this pull request
Nov 6, 2025
…r `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 6, 2025
…ch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 6, 2025
…n "Add C++ fast path for `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 6, 2025
…path for `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 6, 2025
This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. ghstack-source-id: 672fa5d Pull Request resolved: #167051
swolchok
added a commit
that referenced
this pull request
Nov 6, 2025
…or.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 6, 2025
…tch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
Contributor
Author
|
superseded by #167051 |
swolchok
added a commit
that referenced
this pull request
Nov 7, 2025
… path for `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 7, 2025
…or.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 10, 2025
…patch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 10, 2025
This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 12, 2025
…th for `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 12, 2025
…__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 12, 2025
…nd add a couple comments on "Add C++ fast path for `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 12, 2025
…nd add a couple comments on "Add C++ fast path for `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 12, 2025
…comments on "Add C++ fast path for `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 12, 2025
… gating in is_dtensor on "Add C++ fast path for `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 12, 2025
…ensor on "Add C++ fast path for `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 13, 2025
…C++ fast path for `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
swolchok
added a commit
that referenced
this pull request
Nov 13, 2025
…r `DTensor.__torch_dispatch__`" This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
pytorchmergebot
pushed a commit
that referenced
this pull request
Nov 13, 2025
This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike #166370 and #166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of #166370. Pull Request resolved: #167051 Approved by: https://github.com/ezyang
Khanaksahu
pushed a commit
to Khanaksahu/pytorch
that referenced
this pull request
Nov 17, 2025
Just a baby step to make sure we and CI are happy with this before moving on to a combined fast path. ghstack-source-id: 3dcd716 Pull Request resolved: pytorch/pytorch#166369
Silv3S
pushed a commit
to Silv3S/pytorch
that referenced
this pull request
Nov 18, 2025
This patches the `__torch_dispatch__` machinery to detect DTensor and hand over control to a C++ fast path. Unlike pytorch#166370 and pytorch#166369 (which added a DTensor dispatch key and are intended to be replaced by this PR), this approach fundamentally *is* `__torch_dispatch__`, hopefully sidestepping all manner of thorny "does it work just like `__torch_dispatch__`?" that came up during development and review of pytorch#166370. Pull Request resolved: pytorch#167051 Approved by: https://github.com/ezyang
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 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):
Just a baby step to make sure we and CI are happy with this before moving on to a combined fast path.
cc @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @pragupta @msaroufim @dcci @ezyang @gchanan