BBDrive / Meta-SGD-RL

Meta-SGD for RL in Pytorch
3 stars 0 forks source link

您好 请问这套代码是对Meta Reinforcement Learning with Task Embedding and Shared Policy这篇文章的pytorch实现么? #1

Open c4cld opened 3 years ago

c4cld commented 3 years ago

您好 请问这套代码是对Meta Reinforcement Learning with Task Embedding and Shared Policy这篇文章的pytorch实现么?

BBDrive commented 3 years ago

不是,这个是Meta-SGD: Learning to Learn Quickly for Few-Shot Learning的复现,Meta Reinforcement Learning with Task Embedding and Shared Policy我也有复现过,但是没上传

c4cld commented 3 years ago

我最近在做加强reinforcement learning 算法泛化性的工作,对这篇文章比较感兴趣,想通过代码加强理解。请问对Meta Reinforcement Learning with Task Embedding and Shared Policy的pytorch复现能私发我一份么?我的邮箱 patriotzjp@tju.edu.cn 提前谢谢了 ^_^

BBDrive commented 3 years ago

最近我上传到github吧,但是那篇文章我没有复现实验结果,所以复现的未必好。

c4cld commented 3 years ago

非常感谢 ^_^

BBDrive commented 3 years ago

代码我上传了(https://github.com/BBDrive/TESP),但是这个代码是我几年前写的,逻辑应该还可以,但是具体细节可能你要再看看了,不保证可行性

c4cld commented 3 years ago

@BBDrive 非常感谢您的帮助

c4cld commented 3 years ago

@BBDrive 非常感谢您的代码 昨天我跑了一下 代码能跑通 运行python main.py --env-name HalfCheetahDir-v1 --output-folder TESP-halfcheetah-dir --device cuda 目前运行结果截图如下: image 看起来还是比较正常的