Skip to content

[DTensor] Fix partial redistribution order#172277

Closed
wconstab wants to merge 8 commits intogh/wconstab/493/basefrom
gh/wconstab/493/head
Closed

[DTensor] Fix partial redistribution order#172277
wconstab wants to merge 8 commits intogh/wconstab/493/basefrom
gh/wconstab/493/head

Conversation

@wconstab
Copy link
Copy Markdown
Contributor

@wconstab wconstab commented Jan 12, 2026

Stack from ghstack (oldest at bottom):

Make an explicit statement in dtensor docs that defines the ordering of
partials as "left to right", explaining that ordering is semantically
significant.

  1. Ordering of Partial -> Replicate is numerically significant
  2. we need to define one order and always use it

Adds a test case and fixes bugs in redistribution planning which would
have violated partial ordering semantics.

Make an explicit statement in dtensor docs that defines the ordering of
partials as "left to right", explaining that ordering is semantically
significant.

Adds a test case and fixes bugs in redistribution planning which would
have violated partial ordering semantics.

[ghstack-poisoned]
@pytorch-bot
Copy link
Copy Markdown

pytorch-bot Bot commented Jan 12, 2026

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/172277

Note: Links to docs will display an error until the docs builds have been completed.

❌ 2 New Failures

As of commit 6b1a933 with merge base d803917 (image):

NEW FAILURES - The following jobs have failed:

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

Make an explicit statement in dtensor docs that defines the ordering of
partials as "left to right", explaining that ordering is semantically
significant.

Adds a test case and fixes bugs in redistribution planning which would
have violated partial ordering semantics.

[ghstack-poisoned]
@wconstab wconstab requested a review from zpcore January 13, 2026 00:01
Make an explicit statement in dtensor docs that defines the ordering of
partials as "left to right", explaining that ordering is semantically
significant.

Adds a test case and fixes bugs in redistribution planning which would
have violated partial ordering semantics.

[ghstack-poisoned]
Make an explicit statement in dtensor docs that defines the ordering of
partials as "left to right", explaining that ordering is semantically
significant.

Adds a test case and fixes bugs in redistribution planning which would
have violated partial ordering semantics.

[ghstack-poisoned]
Make an explicit statement in dtensor docs that defines the ordering of
partials as "left to right", explaining that ordering is semantically
significant.

1. Ordering of Partial -> Replicate is numerically significant
2. we need to define one order and always use it

Adds a test case and fixes bugs in redistribution planning which would
have violated partial ordering semantics.

[ghstack-poisoned]
Make an explicit statement in dtensor docs that defines the ordering of
partials as "left to right", explaining that ordering is semantically
significant.

1. Ordering of Partial -> Replicate is numerically significant
2. we need to define one order and always use it

Adds a test case and fixes bugs in redistribution planning which would
have violated partial ordering semantics.

[ghstack-poisoned]
Make an explicit statement in dtensor docs that defines the ordering of
partials as "left to right", explaining that ordering is semantically
significant.

1. Ordering of Partial -> Replicate is numerically significant
2. we need to define one order and always use it

Adds a test case and fixes bugs in redistribution planning which would
have violated partial ordering semantics.

[ghstack-poisoned]
SergeyTyshkevich pushed a commit to SergeyTyshkevich/chart2 that referenced this pull request Jan 19, 2026
Make an explicit statement in dtensor docs that defines the ordering of
partials as "left to right", explaining that ordering is semantically
significant.

Adds a test case and fixes bugs in redistribution planning which would
have violated partial ordering semantics.

ghstack-source-id: 8b90e58
Pull Request resolved: pytorch/pytorch#172277
Make an explicit statement in dtensor docs that defines the ordering of
partials as "left to right", explaining that ordering is semantically
significant.

1. Ordering of Partial -> Replicate is numerically significant
2. we need to define one order and always use it

Adds a test case and fixes bugs in redistribution planning which would
have violated partial ordering semantics.

[ghstack-poisoned]
@wconstab
Copy link
Copy Markdown
Contributor Author

closing as replaced by #172609

@wconstab wconstab closed this Jan 20, 2026
suncapitalllc007-star pushed a commit to suncapitalllc007-star/pytorch that referenced this pull request Jan 25, 2026
Make an explicit statement in dtensor docs that defines the ordering of
partials as "left to right", explaining that ordering is semantically
significant.

Adds a test case and fixes bugs in redistribution planning which would
have violated partial ordering semantics.

ghstack-source-id: 9df4d64
Pull Request resolved: pytorch/pytorch#172277
@github-actions github-actions Bot deleted the gh/wconstab/493/head branch February 20, 2026 02:22
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant