[SimpleFSDP] Add support for hsdp+tp #1343
Open
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
dc82ddf to
cbd4fd1
Compare
tianyu-l
reviewed
Jun 28, 2025
| target_spec=target_spec, | ||
| ) | ||
|
|
||
| for placement in placements: |
There was a problem hiding this comment.
Instead of doing this for loop over placements (which tries to be generic but actually hard to make every possible case correct), let's be more explicit about the 3 types DP / FSDP / HSDP, where there is at most 1 replicate dim, and at most 1 shard dim.
I think we can just have three if-else conditions (instead of DP / FSDP before), and one redistribute_local_tensor outside, just like before.
The FSDP2 code for dealing with HSDP is also here, for your reference
https://github.com/pytorch/pytorch/blob/dfef1e44085bb156abc4aff0f34a0b82a4a337b8/torch/distributed/fsdp/_fully_shard/_fsdp_param.py#L324
cbd4fd1 to
fe081b9
Compare
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Outdated suggestions cannot be applied.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.
As titled, this pr adds support for SimpleFSDP's HSDP + TP
The profile trace below shows the three streams for FSDP's All-gather/Reduce-scatter; DDP's All-reduce; and TP's communications.
The loss below shows SimpleFSDP & FSDP2's losses matches under HSDP + TP mode (seed=42).