ElegantRL

ElegantRL copied to clipboard

Encounter cuda error running demo_A2C_PPO.py

3


valleysprings

A little bug in elegantrl/train/run.py

1

run.py 中第82行 exp_r = buffer_items[2].mean().item 当agent为AgentPPO时，由于AgentPPO中重写了上一行中的agent.explore_env，也就是调用的是elegantrl/agents/AgentPPO中的explore_one_env和explore_vec_env两个函数，返回的元组中第三位不再是reward而是logprobs。如果需要获取探索的奖励，这里应该改为 exp_r = buffer_items[3].mean().item，请确认下是否是这样。 同理对run.py中的多进程版本即第218行求exp_r也应该将buffer_items_tensor[2]改为buffer_items_tensor[3]。 当然为了使用其他的Agent依然正确，其实应该修改的是AgentPPO中的xplore_one_env和explore_vec_env返回值顺序，这样的话就要修改elegantrl/agents/AgentPPO.py的81、120、124行将rewards调整到第三个返回值位置上。

legao-2

bug也太多了吧，跑一个实例，改了一天bug

2

官方能不能好好试试为数不多的示例？就4个呀，问题是并不是实例bug，而是内核代码有buG.

flhang

Discussion class : Env of option

4

Discussion 1）看到了 #276 2）閱讀了 github.com/AI4Finance-Foundation/FinRL/tree/master/examples/FinRL_PaperTrading_Demo.ipynb 点贊. 盼望 ElegantRL有一个恆指期權 或 道指期權 (例如:HSI Option or dji Option) 的环境 (例如:IndexOption_TradingEnv). 拟仿照 class StockTradingEnv 建一个 class IndexOptions_TradingEnv, 哪位學長,師兄可界紹一个网站,能够方便取得 options data both live and historical...

LunbiWa

traj_list is wrong in demo_FinRL_ElegantRL_China_A_shares

2

`traj_list = [map(list, zip(*traj_list))]` should be changed into `traj_list = list(zip(*traj_list))` and `cur_items = [map(list, zip(*traj_list))]` should be changed into `cur_items = list(zip(*traj_list))`

niceban

请教一个多GPU训练的问题

1

我看文档说支持多GPU训练，比如tutorial_BipedalWalker-v3.py里面的例子 args.learner_gpus = [0, 1, 2, 3] train_agent_multiprocessing(args) 然后我看了一下源码，貌似learner_gpus这个参数后来根本没有用过 train_agent_multiprocessing这里面好像也没有对gpu_id分配的管理 不知道我是不是漏了关键代码了，请问这个多GPU训练是怎么实现的呢？谢谢。

liaoshuren

🎨 use yaml to save config for training instead of Python File.

3

Both YAML and JSON are commonly used for storing configuration files, and both have their pros and cons. Advantages of using YAML: - YAML is often considered more readable and...

Yonv1943

⚗️ ✨ Update OpenAI gym to gymnasium

1

Update OpenAI gym to gymnasium > In the github of gym, there are: The team that has been maintaining Gym since 2021 has moved all future development to [Gymnasium](https://github.com/Farama-Foundation/Gymnasium), a...

Yonv1943

🐛 Fix bug PER moudle.

2

After updating the vectorized env and the corresponding multiprocessing training module, support for the PER algorithm has been affected. Corresponding Pull Request: https://github.com/AI4Finance-Foundation/ElegantRL/pull/269 --- The related issue is as follows,...

Yonv1943

https://github.com/AI4Finance-Foundation/ElegantRL/blob/3721f6e677ea5a2720be23bbb18855a2f06767d7/elegantrl/train/run.py#L99 In order to be compatible with the WinOS system, it needs to make this change: ``` method = 'spawn' if os.name == 'nt' else 'forkserver' # os.name == 'nt'...

Yonv1943