Merged
Conversation
FurtherAI
approved these changes
Apr 8, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Since Megatron runs in a subprocess, the serverless backend cannot reliably snapshot live training state by copying files from the parent process during long SFT runs. This PR changes internal checkpointing so the Megatron worker saves its own periodic checkpoints inside ART, and the serverless backend only uploads those intermediate checkpoints as artifacts for resume. This keeps the training logic in one place, makes long-running Megatron SFT jobs resumable after pod restarts or deploys, and leaves the final checkpoint to the normal end-of-training save path.