[megatron] fix: megatron args for fully async recipe

[yyDing1]

What does this PR do?

Some config args are modified in the initialization of ray_trainer to fit the megatron setting, specifically

verl/verl/trainer/ppo/ray_trainer.py

Lines 386 to 394 in f904162

	try:
	OmegaConf.set_struct(self.config, True)
	with open_dict(self.config):
	if OmegaConf.select(self.config, "actor_rollout_ref.actor.optim"):
	self.config.actor_rollout_ref.actor.optim.total_training_steps = total_training_steps
	if OmegaConf.select(self.config, "critic.optim"):
	self.config.critic.optim.total_training_steps = total_training_steps
	except Exception as e:
	print(f"Warning: Could not set total_training_steps in config. Structure missing? Error: {e}")

The above code is only invoked in fully-async-rollout when creating the rollout dataloader, while the initialization of fully-async-trainer does not contain the above logic. The megatron trainer should rely on this arg in init_worker, so it will cause an error in the initialization of the megatron optimizer.

My solution:

1. init rollouter and get total_training_steps
2. the optimizer step should be total_training_steps * trigger_parameter_sync_step
3. update the config and use the updated config to init trainer

Checklist Before Starting

• Search for similar PRs. Paste at least one query link here: ...
• Format the PR title as [{modules}] {type}: {description} (This will be checked by the CI)
  • {modules} include fsdp, megatron, veomni, sglang, vllm, rollout, trainer, ci, training_utils, recipe, hardware, deployment, ray, worker, single_controller, misc, perf, model, algo, env, tool, ckpt, doc, data, cfg, reward
  • 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

For changes that can not be tested by CI (e.g., algorithm implementation, new model support), validate by experiment(s) and show results like training curve plots, evaluation results, etc.

API and Usage Example

Demonstrate how the API changes if any, and provide usage example(s) if possible.

# Add code snippet or script demonstrating how to use this

Design & Code Changes

Demonstrate the high-level design if this PR is complex, and list the specific changes.

Checklist Before Submitting

Important

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

• Read the Contribute Guide.
• Apply pre-commit checks: pre-commit install && pre-commit run --all-files --show-diff-on-failure --color=always
• Add / Update the documentation.
• Add unit or end-to-end test(s) to the CI workflow to cover all the code. If not feasible, explain why: ...
• Once your PR is ready for CI, send a message in the ci-request channel in the verl Slack workspace. (If not accessible, please try the Feishu group (飞书群).)
• If your PR is related to the recipe submodule, please also update the reference to the submodule commit via git submodule update --remote or cd recipe && git pull origin main.