Enable PP and EP overlap for MoE #1721
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
meta-cla
bot
added
the
CLA Signed
H-Huang
force-pushed
the
deepseek-v3-new-methods
branch
from
3a61b86 to
0f7a7c9
Compare
|
Running with: CUDA_LAUNCH_BLOCKING |
H-Huang
force-pushed
the
deepseek-v3-new-methods
branch
from
0f7a7c9 to
6584aac
Compare
H-Huang
force-pushed
the
deepseek-v3-new-methods
branch
from
a6e46c7 to
5810c54
Compare
H-Huang
requested review from
fegin,
tianyu-l,
wconstab and
wwwjn
as code owners
|
Just landed pytorch/pytorch#162016, so once CI picks up the nightly the errors should be fixed |
tianyu-l
reviewed
Oct 9, 2025
|
|
||
| [activation_checkpoint] | ||
| mode = "selective" # ["none", "selective", "full"] | ||
| mode = "none" # ["none", "selective", "full"] |
There was a problem hiding this comment.
AC/SAC are not supported since we split the backward into two parts.
| mscale=0.70, | ||
| use_flex_attn=True, | ||
| attn_mask_type="block_causal", | ||
| use_flex_attn=False, |
There was a problem hiding this comment.
Is FlexAttention not supported? It sounds unrelated.
H-Huang
force-pushed
the
deepseek-v3-new-methods
branch
from
9e43a67 to
7cf98e4
Compare
pytorchmergebot
pushed a commit
to pytorch/pytorch
that referenced
this pull request
Oct 15, 2025
Fixed one issue with FSDP last reshard not being called. Rest is mostly refactoring, changing some variables to be class variables so they can be used in pytorch/torchtitan#1721 Pull Request resolved: #165513 Approved by: https://github.com/fegin
wwwjn
reviewed
Oct 16, 2025
Chao1Han
pushed a commit
to Chao1Han/pytorch
that referenced
this pull request
Oct 21, 2025
…#165513) Fixed one issue with FSDP last reshard not being called. Rest is mostly refactoring, changing some variables to be class variables so they can be used in pytorch/torchtitan#1721 Pull Request resolved: pytorch#165513 Approved by: https://github.com/fegin
H-Huang
force-pushed
the
deepseek-v3-new-methods
branch
from
7cf98e4 to
c29fa82
Compare
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.
Option 2 of #1682
These changes add a custom
overlap_callbackfunction to replace the OVERLAP_F_B action that is run during the schedule execution. In the custom function, we writerun_forward()andrun_backward().run_backward()is run as a separate thread so that we can have both forward and backward running together side by side. Looks like this:In order for these changes to work with Expert Parallel, we also need to add custom autograd functions to act as the boundary points at which we do communication. We added hooks before and after expert parallel dispatch and combine to signal boundary points, so our figure from before now turns into:
Now in each of these red blocks, we use a global coordinator. We need
threading.Barrier(2).wait()so that the comm and compute from our forward and backward steps are scheduled in lock-step before continuing.DSv3 16B run command:
Trace examples: