verl icon indicating copy to clipboard operation
verl copied to clipboard

[megatron] fix: Pass optimizer config betas to Megatron optimizer config

Open guillemgt opened this issue 1 month ago • 0 comments

What does this PR do?

This PR fixes a bug where the value of the beta parameters for the Adam optimizer actor_rollout_ref.(actor|critic).optim.betas are ignored when using the Megatron backend.

Checklist Before Starting

  • [X] Search for similar PRs. Paste at least one query link here: beta; optimizer
  • [X] Format the PR title as [{modules}] {type}: {description} (This will be checked by the CI)
    • {modules} include fsdp, megatron, sglang, vllm, rollout, trainer, ci, training_utils, recipe, hardware, deployment, ray, worker, single_controller, misc, perf, model, algo, env, tool, ckpt, doc, data
    • If this PR involves multiple modules, separate them with , like [megatron, fsdp, doc]
    • {type} is in feat, fix, refactor, chore, test
    • If this PR breaks any API (CLI arguments, config, function signature, etc.), add [BREAKING] to the beginning of the title.
    • Example: [BREAKING][fsdp, megatron] feat: dynamic batching

Test

No changes

API and Usage Example

No changes

Design & Code Changes

No design choices or changes

Checklist Before Submitting

[!IMPORTANT] Please check all the following items before requesting a review, otherwise the reviewer might deprioritize this PR for review.

guillemgt avatar Nov 20 '25 16:11 guillemgt