[feat] distributed optimizer low GPU memory resume from checkpoint - #20
Merged
Conversation
…n checkpoint When a model has both fp32 and bf16 parameters (e.g., certain params kept in fp32 for numerical stability), _build_optimizer_group_ranges produces a parameter ordering that doesn't account for the fp32/fp16 split that happens during shard allocation. This causes checkpoint save/load to map optimizer states to wrong parameters. Fix: rebuild model_param_group_index_map inside _build_model_and_main_param_groups where the actual fp32 vs float16 grouping is known, ensuring correct parameter-to-optimizer-state mapping during checkpoint operations. Co-Authored-By: fzyzcjy <5236035+fzyzcjy@users.noreply.github.com>
When --low-memory-resume is set, optimizer states (exp_avg, exp_avg_sq) are initially allocated on CPU during checkpoint loading, then moved to GPU after the full state is loaded. This prevents GPU OOM for large models where optimizer state exceeds GPU memory during the loading phase. Changes: - arguments.py: add --low-memory-resume CLI argument - optimizer_config.py: add low_memory_resume field to OptimizerConfig - distrib_optimizer: allocate dummy tensors on CPU, move to GPU after load - fused_adam_patch: monkey-patch TE FusedAdam to allocate on CPU - mlp: merge sharded state dict tensors on CPU Co-Authored-By: fzyzcjy <5236035+fzyzcjy@users.noreply.github.com>
yueming-yuan
force-pushed
the
feat/low-mem-resume
branch
from
March 12, 2026 05:22
18cec25 to
3c0e083
Compare
fzyzcjy
approved these changes
Mar 12, 2026
fzyzcjy
left a comment
fzyzcjy
left a comment
Collaborator
There was a problem hiding this comment.
approve given it is trivially copied from a verified experiment
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.
What does this PR do ?
Contribution process
flowchart LR A[Pre-checks] --> B[PR Tests] subgraph Code Review/Approval C1[Expert Review] --> C2[Final Review] end B --> C1 C2 --> D[Merge]Pre-checks
Core 0.8)Code review
The following process is enforced via the CODEOWNERS file for changes into
megatron/core. For changes outside ofmegatron/core, it is up to the PR author whether or not to tag the Final Reviewer team.For MRs into `main` branch
Feel free to message or comment the @mcore-oncall to help accelerate your merge into main. The less complex your PR is, the faster it will be approved and merged!
(Step 1): Add PR label
Expert Review(Step 2): Collect the expert reviewers reviews
Expert Reviewlabel when your PR is ready for review.Final Review might get declined if these requirements are not fulfilled.
(Step 3): Final Review
Final Reviewlabel(Optional Step 4): Cherry-pick into release branch
If this PR also needs to be merged into
core_r*release branches, after this PR has been merged, selectCherry-pickto open a new PR into the release branch.For MRs into `dev` branch
The proposed review process for `dev` branch is under active discussion.MRs are mergable after one approval by either
eharper@nvidia.comorzijiey@nvidia.com.Merging your PR
Any member of core-adlr and
core-nemowill be able to merge your PR.