DLR-RM/stable-baselines3

Assert that the VecNormalize wrapper handles the new truncations correctly

Open

#1,093 创建于 2022年10月4日

在 GitHub 查看
 (2 评论) (0 反应) (0 负责人)Python (6,550 star) (1,407 fork)batch import
help wanted

描述

I need to think more about that one, but my guess is that it should not matter much, the scaling is meant to avoid large values and to evolve with the agent performance. Still, we should double check that. Could you open an issue so we don't forget?

Originally posted by @araffin in https://github.com/DLR-RM/stable-baselines3/pull/780#discussion_r985247261

贡献者指南

技术栈
pythonpytorch
领域
machine learningai
议题类型
research
难度面向新贡献者的预计实现难度，1 表示很小改动，5 表示专家级工作。
3
预计时间有经验贡献者完成调查、实现、测试并准备 pull request 的粗略时间范围。
1-3 hours
活动状态议题当前的可参与程度：新鲜、活跃、陈旧、阻塞或等待维护者输入。
stale
清晰度议题是否清楚说明期望改动、验收标准和下一步。
mostly clear
前置要求
familiarity with VecNormalize wrapperunderstanding of RL truncation vs termination
新手友好度1-100 的估计分数，表示该议题对首次贡献者的友好程度。
30
研究方向
Investigate the VecNormalize implementation in stable baselines3 to ensure that truncation steps (as opposed to terminal steps) are correctly handled. Review the discussion in PR #780, particularly the comment by @araffin, and examine the relevant source code in vec normalize.py. Look for any potential issues with normalization updates during non terminal transitions and verify that the scaling logic accommodates truncations without bias.