Skip to content

refactor(trainer): remove deprecated light-megatron backend path#585

Merged
Xiaoming-AMD merged 3 commits intomainfrom
refactor/remove-light-megatron-path
Mar 23, 2026
Merged

refactor(trainer): remove deprecated light-megatron backend path#585
Xiaoming-AMD merged 3 commits intomainfrom
refactor/remove-light-megatron-path

Conversation

@WangLingxun
Copy link
Collaborator

Remove the legacy light-megatron trainer implementation and clean up all related framework aliases. This change deletes the lightmegatron trainer modules and removes light-megatron routing from parser and hook dispatchers (train pretrain and projection performance), so framework resolution now follows megatron directly without a compatibility alias.

- Update third_party/Megatron-LM to core_v0.16.0.
- Adapt muon optimizer patch signatures/call paths to Megatron 0.16 (config_overrides, optimizer arg forwarding).
- Relax turbo TE spec provider gate to <0.17 for DeepSeek compatibility on v0.16.
- Fix DDP/FSDP grad-sync compatibility by allowing _BaseDataParallel.{start,finish}_grad_sync to accept extra kwargs (e.g. force_all_reduce).
@WangLingxun WangLingxun force-pushed the refactor/remove-light-megatron-path branch from f8a4952 to ed4533e Compare March 9, 2026 07:09
@Xiaoming-AMD Xiaoming-AMD merged commit 5ec7649 into main Mar 23, 2026
2 of 5 checks passed
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.

2 participants