Skip to content

Fix GB300 torch_dist checkpoint save crashes from forked local writers#57

Open
zyzshishui wants to merge 1 commit into
radixark:miles-mainfrom
zyzshishui:fix-save
Open

Fix GB300 torch_dist checkpoint save crashes from forked local writers#57
zyzshishui wants to merge 1 commit into
radixark:miles-mainfrom
zyzshishui:fix-save

Conversation

@zyzshishui

Copy link
Copy Markdown

What does this PR do ?

⚠️ For major changes (either in lines of code or in its impact), please make sure to first share a design doc with the team. If you're unsure what's the best way to do so, contact the @mcore-oncall.

Contribution process

Summary

Fix GB300 torch_dist checkpoint save crashes from forked local writers.

Symptom & Reproduction

  • Symptom: GB300 checkpoint save produced partial iter_0000000 directories and failed before .metadata was written.
  • Reproduction: Run the Qwen3.5-35B-A3B direct-save repro on a 2-node GB300 job with torch_dist checkpointing enabled.

Root Cause

  1. write_preloaded_data_multiproc forks one writer per local bucket.
  2. write_preloaded_data serializes parent-staged CPU tensors.
  3. PyTorchStreamWriter::writeRecord crashes in crc32_16bytes.

Fix

Run local checkpoint buckets sequentially in the current writer process instead of forking local writer subprocesses. This preserves the existing public async-writer entry point and global result queue, while removing the unsafe local process boundary around staged checkpoint tensors.

Note: This does not make --async-save supported on GB300. --async-save still moves the checkpoint write into an outer async worker process, and that path reproduced the same crc32_16bytes crash in validation.

Verification

  • sync no-MSC save succeeded with 16 .distcp files, .metadata, and metadata.json.
  • sync MSC save succeeded with 16 .distcp files, .metadata, and metadata.json.
  • load/resume from the saved checkpoint succeeded and produced a complete follow-up checkpoint.

Pre-checks

  • I want this PR in a versioned release and have added the appropriate Milestone (e.g., Core 0.8)
  • I have added relevant unit tests
  • I have added relevant functional tests
  • I have added proper typing to my code Typing guidelines
  • I have added relevant documentation
  • I have run the autoformatter.sh on my PR

Code review

The following process is enforced via the CODEOWNERS file for changes into megatron/core. For changes outside of megatron/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

  1. Attach the Expert Review label when your PR is ready for review.
  2. GitHub auto-assigns expert reviewers based on your changes. They will get notified and pick up your PR soon.

⚠️ Only proceed to the next step once all reviewers have approved, merge-conflict are resolved and the CI is passing.
Final Review might get declined if these requirements are not fulfilled.

(Step 3): Final Review

  1. Add Final Review label
  2. GitHub auto-assigns final reviewers based on your changes. They will get notified and pick up your PR soon.

(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, select Cherry-pick to 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.com or zijiey@nvidia.com.

Merging your PR

Any member of core-adlr and core-nemo will be able to merge your PR.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant