Support of dtensor redistribute with device order #160266
Open
+317
−74
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
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/160266
Note: Links to docs will display an error until the docs builds have been completed. ❌ 4 New Failures, 1 Unrelated FailureAs of commit 12fe67c with merge base 24257f5 ( NEW FAILURES - The following jobs have failed:
FLAKY - The following job failed but was likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
fmassa
reviewed
Aug 10, 2025
| ): | ||
| mesh_dim_size = device_mesh.size(mesh_dim=mesh_dim) | ||
| current_placement = sorted_dst_placement[mesh_dim] | ||
| assert isinstance(current_placement, Shard) |
There was a problem hiding this comment.
I might be doing something wrong when trying this out, but I'm hitting this assertion, where current_placement is Replicate.
There was a problem hiding this comment.
Do you have instructions on how to reproduce the issue? I tried python examples/example_autoparallel.py and it complained RuntimeError: Function CompiledFunctionBackward returned an invalid gradient at index 0 - got [1, 6144] but expected shape compatible with [24, 6144].
[Prototype; for RFC, not ready for review] Now redistribute dtensor honors the device ordering. If no order information specified, it will use the default device order [0,1,2,...]. We can specify `device_order` as follow: ``` sharded_dt = distribute_tensor(input_data, mesh, placement, device_order) ``` and ``` out_dt = sharded_dt.redistribute(mesh, placement, device_order) ``` Note that device order information is added into the DTensorSpec. So `redistribute_local_tensor` doesn't need the `src_device_order` and `dst_device_order`. I leave them here as a reference for AutoParallel (cc fmassa ). I will remove those order related args from redistributed related API in this PR. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta [ghstack-poisoned]
[Prototype; for RFC, not ready for review] Now redistribute dtensor honors the device ordering. If no order information specified, it will use the default device order [0,1,2,...]. We can specify `device_order` as follow: ``` sharded_dt = distribute_tensor(input_data, mesh, placement, device_order) ``` and ``` out_dt = sharded_dt.redistribute(mesh, placement, device_order) ``` Note that device order information is added into the DTensorSpec. So `redistribute_local_tensor` doesn't need the `src_device_order` and `dst_device_order`. I leave them here as a reference for AutoParallel (cc fmassa ). I will remove those order related args from redistributed related API in this PR. cc H-Huang awgu wanchaol fegin fduwjj wz337 wconstab d4l3k pragupta [ghstack-poisoned]
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
[Prototype; for RFC, not ready for review]
Now redistribute dtensor honors the device ordering. If no order information specified, it will use the default device order [0,1,2,...]. We can specify
device_orderas follow:and
Note that device order information is added into the DTensorSpec. So
redistribute_local_tensordoesn't need thesrc_device_orderanddst_device_order. I leave them here as a reference for AutoParallel (cc @fmassa ). I will remove those order related args from redistributed related API in this PR.Stack from ghstack (oldest at bottom):
cc @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @pragupta