[DeviceMesh][ez] Extract the pg creation as a util function#163930
Closed
fduwjj wants to merge 2 commits intogh/fduwjj/210/basefrom
Closed
[DeviceMesh][ez] Extract the pg creation as a util function#163930fduwjj wants to merge 2 commits intogh/fduwjj/210/basefrom
fduwjj wants to merge 2 commits intogh/fduwjj/210/basefrom
Conversation
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/163930
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit bccf9e9 with merge base 5fcde74 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
the
oncall: distributed
fduwjj
added
ciflow/trunk
fegin
approved these changes
Sep 26, 2025
This is just to extract common logic into a util function because we will use it many times for the following stack of Device Mesh refactoring. cc H-Huang awgu wanchaol fegin wz337 wconstab d4l3k pragupta msaroufim dcci [ghstack-poisoned]
Contributor
Author
|
@pytorchbot merge |
Collaborator
Merge startedYour 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 |
pytorchmergebot
pushed a commit
that referenced
this pull request
Sep 27, 2025
) While refactoring the bookkeeping for DeviceMesh while leveraging CuTe layout, we found that we need to have two more util functions. One is to check whether one layout has overlap inside it or not. For example, (2,2):(2:1) has no overlap while (2,2):(2:2) has overlap. Pull Request resolved: #163367 Approved by: https://github.com/fegin ghstack dependencies: #163212, #163288, #163928, #163930
This was referenced Sep 26, 2025
jainapurva
pushed a commit
that referenced
this pull request
Sep 29, 2025
This is just to extract common logic into a util function because we will use it many times for the following stack of Device Mesh refactoring. Pull Request resolved: #163930 Approved by: https://github.com/fegin ghstack dependencies: #163212, #163288, #163928
jainapurva
pushed a commit
that referenced
this pull request
Sep 29, 2025
) While refactoring the bookkeeping for DeviceMesh while leveraging CuTe layout, we found that we need to have two more util functions. One is to check whether one layout has overlap inside it or not. For example, (2,2):(2:1) has no overlap while (2,2):(2:2) has overlap. Pull Request resolved: #163367 Approved by: https://github.com/fegin ghstack dependencies: #163212, #163288, #163928, #163930
lw
reviewed
Sep 29, 2025
Comment on lines
+896
to
+904
| mesh = DeviceMesh( | ||
| device_type, | ||
| mesh_nd, | ||
| mesh_dim_names=mesh_dim_names, | ||
| backend_override=backend_override, | ||
| _init_backend=_init_backend, | ||
| ) | ||
| if cur_rank in mesh_nd: | ||
| res_mesh = mesh |
Contributor
There was a problem hiding this comment.
Thanks for splitting this out as I had asked! I didn't get to review this before it was landed, but there's still something I don't understand. I get that we need to call the "PG creation API" multiple times on each rank, even for the ranks that don't participate, but I don't get why we need to call the DeviceMesh constructor multiple times!
Could we instead call the PG creation API directly and just invoke the DeviceMesh constructor once with the right PG?
maggiemoss
pushed a commit
to maggiemoss/pytorch
that referenced
this pull request
Sep 29, 2025
…163930) This is just to extract common logic into a util function because we will use it many times for the following stack of Device Mesh refactoring. Pull Request resolved: pytorch#163930 Approved by: https://github.com/fegin ghstack dependencies: pytorch#163212, pytorch#163288, pytorch#163928
maggiemoss
pushed a commit
to maggiemoss/pytorch
that referenced
this pull request
Sep 29, 2025
…rch#163367) While refactoring the bookkeeping for DeviceMesh while leveraging CuTe layout, we found that we need to have two more util functions. One is to check whether one layout has overlap inside it or not. For example, (2,2):(2:1) has no overlap while (2,2):(2:2) has overlap. Pull Request resolved: pytorch#163367 Approved by: https://github.com/fegin ghstack dependencies: pytorch#163212, pytorch#163288, pytorch#163928, pytorch#163930
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):
This is just to extract common logic into a util function because we will use it many times for the following stack of Device Mesh refactoring.
cc @H-Huang @awgu @wanchaol @fegin @wz337 @wconstab @d4l3k @pragupta @msaroufim @dcci