Skip to content

refactor(mimo): migrate MegatronMIMO to model builders#4557

Draft
yaoyu-33 wants to merge 1 commit into
yuya/refactor-bridge-model-buildersfrom
yuya/refactor-megatron-mimo-model-builder
Draft

refactor(mimo): migrate MegatronMIMO to model builders#4557
yaoyu-33 wants to merge 1 commit into
yuya/refactor-bridge-model-buildersfrom
yuya/refactor-megatron-mimo-model-builder

Conversation

@yaoyu-33

@yaoyu-33 yaoyu-33 commented Jun 28, 2026

Copy link
Copy Markdown
Contributor

Summary

  • migrate MegatronMIMO construction, conversion, checkpoint IO, and data loading to serializable ModelConfig plus standalone ModelBuilder contracts
  • retain deprecated provider compatibility paths and released keyword aliases
  • use one provider-neutral MegatronMIMOInfra type across legacy and builder paths
  • pass runtime process grids explicitly from setup infrastructure instead of storing them in serializable configs
  • update Qwen3.5-VL MIMO spec builders and repository examples

Stack

Validation

  • uvx pre-commit run --all-files
  • two independent subagent review passes completed; all confirmed findings fixed
  • EOS: 628 targeted unit tests passed
  • EOS: 2-GPU MegatronMIMO checkpoint save/load round-trip passed

Signed-off-by: yaoyu-33 <yaoyu.094@gmail.com>
@copy-pr-bot

copy-pr-bot Bot commented Jun 28, 2026

Copy link
Copy Markdown

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@yaoyu-33

Copy link
Copy Markdown
Contributor Author

/ok to test 89f2971

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