norm_adv_by_std_in_grpo should be from rllm.algorithm.norm_adv_by_std_in_grpo?

[gang-architect]

Description

https://github.com/rllm-org/rllm/blob/main/rllm/experimental/unified_trainer.py#L216

Here we do norm_adv_by_std_in_grpo=self.rllm_config.stepwise_advantage.get("norm_adv_by_std_in_grpo", True), but if I'm not wrong, the norm_adv_by_std_in_grpo flag is defined at rllm.algorithm.norm_adv_by_std_in_grpo in the config files.

Steps to Reproduce

n/a

Error Output / Traceback

rLLM Version

latest main

Training Backend

tinker

Python Version

3.11.14

GPU / CUDA Version

No response

vLLM Version (if applicable)

No response

Training Script / Config

Additional Context

No response