Baselines: Running into issues on example execution

[ampnirvana]

Get this error when I run the first example python3 -m baselines.deepq.experiments.train_cartpole:

/usr/bin/python3: Error while finding spec for 'baselines.deepq.experiments.train_cartpole' (<class 'ImportError'>: cannot import name 'deepq')

I have both Python 2 and 3 installed. Thus I installed baselines with pip3.
Any suggestions?

[chenmin1107]

I had the same issue as above. Any help will be appreciated !

[siemanko]

I cannot reproduce it neither on linux nor on max. Can you provide more details?

[chenmin1107]

i did my setup within virtualenv

1. i installed tensorflow with pip3
2. i installed baselines with pip3
3. i installed gym with pip3 (minimal install)

i tried running one of the example command line: python3 -m baselines.deepq.experiments.train_cartpole

and i got the following error:
/usr/bin/python3: Error while finding spec for 'baselines.deepq.experiments.train_cartpole' (: cannot import name 'deepq')

When i tested it in python3:

1) import baselines
This is ok
2) from baselines import deepq
i got the following error:
Python 3.4.3 (default, Nov 17 2016, 01:08:31)
[GCC 4.8.4] on linux
Type "help", "copyright", "credits" or "license" for more information.

import baselines
from baselines import deepq

i got the following error:

Traceback (most recent call last):
File "", line 1, in
File "/usr/local/lib/python3.4/dist-packages/baselines/deepq/__init__.py", line 4, in
from baselines.deepq.simple import learn, load # noqa
File "/usr/local/lib/python3.4/dist-packages/baselines/deepq/simple.py", line 12, in
from baselines import deepq
ImportError: cannot import name 'deepq'

[ampnirvana]

I managed to solve the issue by installing Anaconda 3 and creating a Conda environment conda create -n tensorflow and then running the installation in this environment. More instructions are provided in the following:
https://www.tensorflow.org/install/install_linux#installing_with_anaconda

[siemanko]

It seems like there's a workaround so I will close this issue.

If the issue persists, since it is difficult to reproduce, please specify a Dockerfile needed to reproduce it before reopening.

[Zhang-Yong]

Do not install testflow seperately, install gym and baselines with pip3 and run with python3, everything should be ok.

[chenmin1107]

But the program needs tensorflow as the backend, right? You must have installed it somewhere.

[ghost]

the issue seemed to be stemming from a recursive import error (i don't have know what i'm talking about) of deepq from baselines.deepq.simple.py
I changed some some code in simple.py to more explicitly point at the desired modules and it worked
modifications include:
edited line 12 to "from baselines.deepq.build_graph import build_act, build_train"
edited line 172 to " act, train, update_target, debug = baselines.deepq.build_graph.build_train("
added "import baselines" to file

[Zhang-Yong]

@chenmin1107 , "pip3 baselines" will install testflow, if you look at lastest test flow installation, it can installed by pip3, do not have to install in virtualenv, I guess the virtualenv cause the issue.

[yuke93]

I think it is a recursive import error and @Dangerpuss offers a practical solution. But it's still weird because in my case one machine with python 3.5 does not report this error while the other machine with python3.4 reports this error.

[chenmin1107]

@Zhang-Yong thanks for the suggestion, but I want to set it up in a virtualenv so it won't mess my other things.

@yuke93 Same thing here, @Dangerpuss 's way works on my side as well. But I have not tried python3.5. Maybe it is just a python version issue.

[sbittner12]

Thanks @Dangerpuss I'm also confident that it's a python 3 versioning issue as @yuke93 indicates.

from baselines import deepq

Python 3.4: recursive import error, dangerpuss's edits fixed problem
Python 3.6: no issue

I'd say Python 3.5 and greater should work based on the responses.

[IQ17]

in simple.py
Also edit line 25 to: act = baselines.deepq.build_graph.build_act(**act_params)

This is required we you run "python -m baselines.deepq.experiments.enjoy_cartpole"

[Bhargav5]

You can replace,
from baselines import deepq by
from baselines.deepq.build_graph import build_act, build_train

it will work. This problem I faced while working with python 2.7