[Roadmap][Megatron, VLM] Support VLM
Open
Nobody has claimed this yet.
roadmap
- Dominant language
- Python
- Stars
- 8.5k
- Forks
- 1.3k
- Avg merge
- 5h 36m
- Merged PRs (30d)
- 22
Description
Megatron + VLM is ready https://github.com/THUDM/slime/issues/1097
TODO
-
Dataset
- Geo3k
- [Future] Support context parallelism for long-context datasets
-
Rollout - training backend agnostic
-
Model Support - Qwen for now
- Add support through the Megatron Bridge
- Support Qwen3-VL
- Support GLM
- [Future] Long term: adopt the AutoModel
-
Megatron Trainer
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the TODO checklist and the linked Megatron Bridge Qwen-VL implementation. The remaining scope includes context parallelism for long-context datasets, GLM support, and future AutoModel adoption, so a contributor must first narrow this roadmap into one defined task. Done means the selected checklist item is implemented and its relevant validation is complete.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- ai, machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100