fix async tp - #4045
Open
acisseJZhong wants to merge 2 commits into
Open
Conversation
acisseJZhong
requested review from
IvanKobzarev,
SherlockNoMad,
aditvenk,
fegin,
sanketpurandare,
tianyu-l,
wconstab,
wwwjn and
xmfan
as code owners
August 1, 2026 00:02
meta-cla
Bot
added
the
CLA Signed
tianyu-l
reviewed
Aug 1, 2026
| ) | ||
|
|
||
| with warnings.catch_warnings(): | ||
| warnings.simplefilter("ignore", FutureWarning) |
| context_parallel_degree=1, | ||
| pipeline_parallel_degree=1, | ||
| enable_sequence_parallel=False, | ||
| enable_sequence_parallel=self.enable_sequence_parallel, |
Contributor
There was a problem hiding this comment.
we are not convinced by evidence that this is needed at all, let's wait
Comment on lines
+82
to
+84
| maybe_enable_async_tp( | ||
| parallelism, compile_config, parallel_dims.get_dense_tp_mesh() | ||
| ) |
Contributor
There was a problem hiding this comment.
I feel we can put this under compile enablement -- if model compile is enabled and inductor backend is used, we should always enable async TP (any reason not to?). And we should delete the enable_async_tp flag. WDYT?
My eventual goal is to delete the special parallelize_deepseekv3, having a unified one for decoder, and make it simple.
| spmd_backend: Literal["default", "spmd_types"] = "default" | ||
| """SPMD backend used by TorchTitan model parallelization in the generator.""" | ||
|
|
||
| enable_async_tensor_parallel: bool = False |
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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):
Summary
inheriting from #3869
Fix async TP setup across SPMD backends by:
ParallelDims.get_dense_tp_mesh()for backend-aware meshselection
inference config