Fix subgroup optimizer metadata inconsistency#7820
Draft
st-bang97 wants to merge 5 commits intodeepspeedai:masterfrom
Draft
Fix subgroup optimizer metadata inconsistency#7820st-bang97 wants to merge 5 commits intodeepspeedai:masterfrom
st-bang97 wants to merge 5 commits intodeepspeedai:masterfrom
Conversation
Collaborator
|
Hi @st-bang97, Overall this PR looks good to me. Here are some suggestions:
|
Contributor
Author
|
Thanks for the review and feedback. I’m currently finalizing a paper submission and preparing for reviews, so I won’t be able to address this immediately. I’ll take a closer look and update the PR as soon as I have time. |
delock
pushed a commit
that referenced
this pull request
Apr 23, 2026
Fix CPUAdam same-step subgroup drift in ZeRO-3 (#7819) This PR ports the fix from #7820 to the latest DeepSpeed version. It makes `Adam_Optimizer::IncrementStep` idempotent for repeated calls at the same logical step and avoids unnecessary recomputation when the step has not changed. ZeRO-3/SuperOffload can invoke multiple subgroup updates within a single logical step on a shared native optimizer object. The previous logic mixed multiply and recompute paths, producing non-bit-identical bias-correction metadata across subgroup calls. This change aligns the step-transition logic in both the CPU and XPU headers, clarifies first-step and non-sequential-step behavior, and prevents unnecessary work on repeated same-step updates. It also adds CPUAdam regression tests covering subgroup-style repeated same-step updates through both `step_subgroup()` and `step()` with parameter swapping. Signed-off-by: st_bang <st.bang@dgist.ac.kr>
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
3 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.
Description
This PR addresses Issue #7819.
When using ZeRO Stage-3 + CPU offloaded optimizer (CPUAdam) with 2+ subgroups,
ds_adam_stepcan be invoked multiple times within a single global optimizer step. Before this fix, the internal bias-correction-related state (e.g.,_betta2_t→_bias_correction2) could become inconsistent across subgroup invocations within the same step, leading to subgroup-wise optimizer state divergence.Solution
Make
IncrementStep()step-consistent under repeated calls in the same global step.Change
Update
IncrementStep()to only advance or recompute state whenstep != _step, preventing subgroup-to-subgroup drift inside a single step.