feat(mpo): add Nemotron Omni Nano and Super training#3345
Draft
jinglinglingling wants to merge 4 commits into
Draft
feat(mpo): add Nemotron Omni Nano and Super training#3345jinglinglingling wants to merge 4 commits into
jinglinglingling wants to merge 4 commits into
Conversation
Migrate MPO to model-owned multimodal packing with pair-safe batches, MMPR recipes, and resumable Nano/Super qualification workflows.
github-actions
Bot
added
the
Documentation
jinglinglingling
added
the
CI:Lfast
Allow smoke segments to keep one scheduler horizon while forcing an early checkpoint for resume validation.
Contributor
Author
|
/ok to test efe1b3b |
added 2 commits
July 24, 2026 22:06
Prefix chained-run IDs with letters so OmegaConf cannot parse hexadecimal IDs as scientific notation.
Keep the Python ID generator valid inside the shell command substitution.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Validation results
Nano CP1
Both packed and no-pack runs completed successfully. Their loss, reward, and accuracy trends are consistent with the legacy implementation.
Super CP2 (packed)
Loss and accuracy match the legacy trend through step 58. The increasing chosen BCO reward and decreasing rejected BCO reward show the expected preference separation.
The previous post-step-58 segment is excluded because it resumed without optimizer state. The corrected chained-resume smoke successfully restored model, distributed optimizer, LR scheduler, MPO state, and W&B run identity. A fresh chained 100-step run is now in progress.
Test plan