运行Dynamic_train.py时报错Exception in thread Thread-2：parl.remote.exceptions.RemoteError: [PARL remote error when calling function __init__]:

[Senwei-Huang]

作者您好，我运行QuadrupedalRobots/ETGRL/train.py训练是没有问题的，但是运行Dynamic_train.py时出现了下面3个问题，查看Dynamic_train.py同级目录里是有./model/Dynamic_parallel_model.py文件的，请问是什么原因造成的呢？

Exception in thread Thread-2:
parl.remote.exceptions.RemoteError: [PARL remote error when calling function __init__]:

[Errno 2] No such file or directory: './model/Dynamic_parallel_model.py'
FileNotFoundError: [Errno 2] No such file or directory: './model/Dynamic_parallel_model.py'

parl.remote.exceptions.FutureFunctionError: There is an error raised when calling the future function __init__.

完整报错信息

[04-26 20:29:14 MainThread @Dynamic_train.py:71] args:Namespace(K=20, alg='ga', eval=0, gamma=1, load='', outdir='Dynamic', sigma=0.1, steps=10000, suffix='exp0', thread=2, xparl='192.168.30.145:8037')
Exception in thread Thread-5:
Traceback (most recent call last):
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/threading.py", line 932, in _bootstrap_inner
self.run()
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/threading.py", line 870, in run
self._target(*self._args, **self._kwargs)
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/site-packages/parl/remote/future_mode/proxy_wrapper_nowait.py", line 92, in _run_object_in_backend
raise e
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/site-packages/parl/remote/future_mode/proxy_wrapper_nowait.py", line 82, in _run_object_in_backend
self._xparl_remote_wrapper_obj = remote_wrapper(
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/site-packages/parl/remote/remote_wrapper.py", line 107, in init
raise RemoteError('init', traceback_str)
parl.remote.exceptions.RemoteError: [PARL remote error when calling function __init__]:
[Errno 2] No such file or directory: './model/Dynamic_parallel_model.py'
traceback:
Traceback (most recent call last):
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/site-packages/parl/remote/job.py", line 297, in wait_for_connection
cls = load_remote_class(message[1])
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/site-packages/parl/remote/remote_class_serialization.py", line 207, in load_remote_class
with open(file_name + '.py') as t_file:
FileNotFoundError: [Errno 2] No such file or directory: './model/Dynamic_parallel_model.py'

Exception in thread Thread-4:
Traceback (most recent call last):
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/threading.py", line 932, in _bootstrap_inner
self.run()
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/threading.py", line 870, in run
self._target(*self._args, **self._kwargs)
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/site-packages/parl/remote/future_mode/proxy_wrapper_nowait.py", line 92, in _run_object_in_backend
raise e
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/site-packages/parl/remote/future_mode/proxy_wrapper_nowait.py", line 82, in _run_object_in_backend
self._xparl_remote_wrapper_obj = remote_wrapper(
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/site-packages/parl/remote/remote_wrapper.py", line 107, in init
raise RemoteError('init', traceback_str)
parl.remote.exceptions.RemoteError: [PARL remote error when calling function __init__]:
[Errno 2] No such file or directory: './model/Dynamic_parallel_model.py'
traceback:
Traceback (most recent call last):
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/site-packages/parl/remote/job.py", line 297, in wait_for_connection
cls = load_remote_class(message[1])
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/site-packages/parl/remote/remote_class_serialization.py", line 207, in load_remote_class
with open(file_name + '.py') as t_file:
FileNotFoundError: [Errno 2] No such file or directory: './model/Dynamic_parallel_model.py'

Traceback (most recent call last):
File "/mnt/hgfs/虚拟机/PaddleRobotics-main/QuadrupedalRobots/ETGRL/Dynamic_train.py", line 74, in
main()
File "/mnt/hgfs/虚拟机/PaddleRobotics-main/QuadrupedalRobots/ETGRL/Dynamic_train.py", line 72, in main
model.train(args.steps)
File "/mnt/hgfs/虚拟机/PaddleRobotics-main/QuadrupedalRobots/ETGRL/model/Dynamic_parallel_model.py", line 159, in train
mean_re = self.update(epoch)
File "/mnt/hgfs/虚拟机/PaddleRobotics-main/QuadrupedalRobots/ETGRL/model/Dynamic_parallel_model.py", line 128, in update
future_objects.append(self.agent_list[i].batch_sample_episodes(param=solutions[i*self.K:(i+1)*self.K,:],K = self.K))
File "/home/senweihuang/anaconda3/envs/parl/lib/python3.8/site-packages/parl/remote/future_mode/proxy_wrapper_nowait.py", line 144, in getattr
raise self._xparl_remote_object_exception
parl.remote.exceptions.FutureFunctionError: There is an error raised when calling the future function __init__.
You can see the detailed error message above, which is printed by another thread.

Process finished with exit code 1

环境

Ubuntu 18.04
python 3.8
parl = 1.4.0
torch = 1.7.0
rlschool = 1.0.2

@xueeinstein

[Senwei-Huang]

我在PaddlePaddle / PARL下找到了两个类似的问题：
第一个（2020.7.13）
主要的问题是当前运行的代码依赖于文件夹内的代码，xparl默认是只是分发当前文件夹中的.py文件。导致这个问题在于PARL/xparl的设计是针对多机并行设计的，需要把当前工作目录的代码分发到不同机器上（在单机上也用同样的逻辑），并提供了一个单机内的并行的解决思路：

export PYTHONPATH=./:$PYTHONPATH
xparl start --port XXXX 
python main.py

但是这个方法并不有效。

第二个（2020.8.7）
XPARL会自动分发当前文件夹中的所有文件，但其子文件夹除外。XPARL不可能分发所有子文件夹，因为用户可能有复杂的目录结构，有时甚至有大文件。要显式分发所需的文件，请尝试以下API：
请注意，您必须传递文件，而不是直接传递文件夹。
法1：
修改
parl.connect(xparl_addr)
为
parl.connect(xparl_addr, distributed_files=['./model/Dynamic_parallel_model.py','./alg/es.py','./model/__init__.py'])

这个方法也无效。

法2：

git clone -b xparl_submod https://github.com/PaddlePaddle/PARL/
cd PARL
pip install .

这个方法执行不了
报错：

正克隆到 'PARL'...
fatal: 远程分支 xparl_submod 在上游 origin 未发现

您好，请问你们当时是怎么跑通的，咋到我这就会出现问题了呢？能不能给点解决的建议，谢谢。@xueeinstein @TomorrowIsAnOtherDay

[TomorrowIsAnOtherDay]

运行Dynamic_train.py时出现了下面3个问题
我看了下Dynamic_train.py是没有并行代码的，好奇你这个并行的错误怎么来的，是你自己改造并行版本的吗？

[Senwei-Huang]

没有改造，Dynamic_train.py的这行代码：

model = ES_ParallelModel(mean_dict=MEAN_DICT,gait=GAIT_LIST,K=args.K,thread = args.thread,sigma=args.sigma,
                             dynamic_param=dynamic_param,outdir=outdir,alg=args.alg,xparl_addr = args.xparl)

调用了./model/Dynamic_parallel_model.py的ES_ParallelModel类，Dynamic_parallel_model.py里面有并行代码，就下面这行：

@parl.remote_class(wait=False)
class RemoteESAgent(object):

@TomorrowIsAnOtherDay

[TomorrowIsAnOtherDay]

抱歉，今天会议比较多，我们尽快看下这个问题。

[Senwei-Huang]

不着急，您先忙。
@TomorrowIsAnOtherDay