list(map()) bug

[MMeecatfish]

注意到在PR#202中为了加快速度，将list()替换成[]，但是将所有的list(map())都替换成了[map()]，例如AgentBase.py第282行：

traj_list1 = [map(list, zip(*traj_list))]  # state, reward, done, action, noise

这不会引起迭代展开，而是生成了单个元素的list对象，如下：

>>> list(map(list, zip(*((1,2,3),(4,5,6)))))
[[1, 4], [2, 5], [3, 6]]
>>> [map(list, zip(*((1,2,3),(4,5,6))))]
[<map object at 0x7f96184ec910>]

[YangletLiu]

非常感谢您的反馈！！ 这两天我们陆续收到了 多个反馈，正努力修复。 这个 PR 接收得 有点快； 另一方面，也是因为 缺少 优良的Unit Tests....补课补课。

注意到在PR#202中为了加快速度，将list()替换成[]，但是将所有的list(map())都替换成了[map()]，例如AgentBase.py第282行：

traj_list1 = [map(list, zip(*traj_list))]  # state, reward, done, action, noise

这不会引起迭代展开，而是生成了单个元素的list对象，如下：

>>> list(map(list, zip(*((1,2,3),(4,5,6)))))
[[1, 4], [2, 5], [3, 6]]
>>> [map(list, zip(*((1,2,3),(4,5,6))))]
[<map object at 0x7f96184ec910>]

非常感谢您的反馈！！ 这两天我们陆续收到了 多个反馈，正努力修复。 这个 PR 接收得 有点快； 另一方面，也是因为 缺少 优良的Unit Tests....补课补课。

[sboal94]

Same problem, when will it be solved?

I have tried changing [] to list() in the line on AgentBase.py:

traj_list1 = list(map(list, zip(*traj_list)))

It doesn't work either. It gives error in this notebook

[jimmydengpeng]

把所有[map]换成list(map)就可解决。map()生成的是一个可迭代对象(iterable)，直接放进方括号[]得到的list也只能迭代使用，而不能通过index索引，后面的代码用的index索引，所以报错。

[sboal94]

@jimmydengpeng Thank you for your contribution. I have replaced this

traj_list1 = [map(list, zip(*traj_list))] # state, reward, done, action, noise

with this

traj_list1 = list(map(list, zip(*traj_list))) # state, reward, done, action, noise

Same error

 289         traj_state = torch.stack(traj_list1[0])
    290         traj_action = torch.stack(traj_list1[3])
    291 

TypeError: stack(): argument 'tensors' (position 1) must be tuple of Tensors, not map

[LotuSrc]

@sboal94 Do you install elegantrl with -e flag? traj_list1[0] is a map style. It seems that the modification doesn't work.

[quantumiracle]

just replace states, rewards, masks, actions = [torch.cat(item, dim=0) for item in traj_items] with [states, rewards, masks, actions] = traj_list since the samples are already shaped.

[Yonv1943]

The follow Pull request fix this bug ↓ Fix bug for vec env and agentbase init https://github.com/AI4Finance-Foundation/ElegantRL/pull/248

[YangletLiu]

just replace states, rewards, masks, actions = [torch.cat(item, dim=0) for item in traj_items] with [states, rewards, masks, actions] = traj_list since the samples are already shaped.

Thanks very much!