[dtensor] add src_data_rank to distribute_tensor API #143883
Closed
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
As titled, this PR add a kwarg src_data_rank to the distribute_tensor API, to allow user specify a specific rank as the full tensor source data. Previously we by default specify group_rank=0 as the source of truth for single device semantic, this new option: * gives advanced user flexiblity to choose the source data rank * allow user to specify None explicity, which means we will skip the communications needed (scatter/broadcast) for the cases that does not care about single device semantic (i.e. loading from a checkpoint) [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/143883
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit cd83f97 with merge base d88a8c4 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
As titled, this PR add a kwarg src_data_rank to the distribute_tensor API, to allow user specify a specific rank as the full tensor source data. Previously we by default specify group_rank=0 as the source of truth for single device semantic, this new option: * gives advanced user flexiblity to choose the source data rank * allow user to specify None explicity, which means we will skip the communicat 10000 ions needed (scatter/broadcast) for the cases that does not care about single device semantic (i.e. loading from a checkpoint) cc H-Huang awgu kwen2501 fegin fduwjj wz337 wconstab d4l3k c-p-i-o [ghstack-poisoned]
wanchaol
added a commit
that referenced
this pull request
Dec 27, 2024
As titled, this PR add a kwarg src_data_rank to the distribute_tensor API, to allow user specify a specific rank as the full tensor source data. Previously we by default specify group_rank=0 as the source of truth for single device semantic, this new option: * gives advanced user flexiblity to choose the source data rank * allow user to specify None explicity, which means we will skip the communications needed (scatter/broadcast) for the cases that does not care about single device semantic (i.e. loading from a checkpoint) ghstack-source-id: 55510d9 Pull Request resolved: #143883
tianyu-l
reviewed
Dec 30, 2024
As titled, this PR add a kwarg src_data_rank to the distribute_tensor API, to allow user specify a specific rank as the full tensor source data. Previously we by default specify group_rank=0 as the source of truth for single device semantic, this new option: * gives advanced user flexiblity to choose the source data rank * allow user to specify None explicity, which means we will skip the communications needed (scatter/broadcast) for the cases that does not care about single device semantic (i.e. loading from a checkpoint) cc H-Huang awgu kwen2501 fegin fduwjj wz337 wconstab d4l3k c-p-i-o [ghstack-poisoned]
As titled, this PR add a kwarg src_data_rank to the distribute_tensor API, to allow user specify a specific rank as the full tensor source data. Previously we by default specify group_rank=0 as the source of truth for single device semantic, this new option: * gives advanced user flexiblity to choose the source data rank * allow user to specify None explicity, which means we will skip the communications needed (scatter/broadcast) for the cases that does not care about single device semantic (i.e. loading from a checkpoint) cc H-Huang awgu kwen2501 fegin fduwjj wz337 wconstab d4l3k c-p-i-o [ghstack-poisoned]
XilunWu
approved these changes
Dec 30, 2024
| @@ -41,15 +45,21 @@ def world_size(self) -> int: | |||
| return 4 | |||
|
|
|||
| @with_comms | |||
| def test_distribute_tensor(self): | |||
| def test_distribute_tensor_rank(self): | |||
There was a problem hiding this comment.
it would be good if we also test uneven sharding.
There was a problem hiding this comment.
I think this feature is somewhat orthogonal to even or uneven sharding, I'll try to update as a follow up later
pytorchmergebot
pushed a commit
that referenced
this pull request
Jan 2, 2025
as titled, this PR propagates the src_data_rank in the TP API, so that module level APIs could leverage the flexibility to choose src_data_rank, and avoid the communication if it does not need to Pull Request resolved: #144005 Approved by: https://github.com/tianyu-l ghstack dependencies: #143883
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.
Stack from ghstack (oldest at bottom):
As titled, this PR add a kwarg src_data_rank to the distribute_tensor
API, to allow user specify a specific rank as the full tensor source
data. Previously we by default specify group_rank=0 as the source of
truth for single device semantic, this new option:
communications needed (scatter/broadcast) for the cases that does not
care about single device semantic (i.e. loading from a checkpoint)
cc @H-Huang @awgu @kwen2501 @fegin @fduwjj @wz337 @wconstab @d4l3k @c-p-i-o