Gym: AttributeError: module 'Box2D._Box2D' has no attribute 'RAND_LIMIT'

[Jackiexiao]

I can't run the LunarLander-v2 env, error occurs

[2017-07-08 12:53:22,066] Making new env: LunarLander-v2
Traceback (most recent call last):
  File "gym_test_CartPole.py", line 3, in <module>
    env = gym.make('LunarLander-v2')
  File "/home/jackie/gym/gym/envs/registration.py", line 161, in make
    return registry.make(id)
  File "/home/jackie/gym/gym/envs/registration.py", line 119, in make
    env = spec.make()
  File "/home/jackie/gym/gym/envs/registration.py", line 85, in make
    cls = load(self._entry_point)
  File "/home/jackie/gym/gym/envs/registration.py", line 17, in load
    result = entry_point.load(False)
  File "/home/jackie/anaconda3/lib/python3.6/site-packages/pkg_resources/__init__.py", line 2303, in load
    return self.resolve()
  File "/home/jackie/anaconda3/lib/python3.6/site-packages/pkg_resources/__init__.py", line 2309, in resolve
    module = __import__(self.module_name, fromlist=['__name__'], level=0)
  File "/home/jackie/gym/gym/envs/box2d/__init__.py", line 1, in <module>
    from gym.envs.box2d.lunar_lander import LunarLander
  File "/home/jackie/gym/gym/envs/box2d/lunar_lander.py", line 4, in <module>
    import Box2D
  File "/home/jackie/anaconda3/lib/python3.6/site-packages/Box2D/__init__.py", line 20, in <module>
    from .Box2D import *
  File "/home/jackie/anaconda3/lib/python3.6/site-packages/Box2D/Box2D.py", line 434, in <module>
    RAND_LIMIT = _Box2D.RAND_LIMIT
AttributeError: module 'Box2D._Box2D' has no attribute 'RAND_LIMIT'

[onaclov2000]

This is a known issue, and there is information here:
https://github.com/openai/gym/issues/100

I'm not sure regarding an anaconda environment if those dependencies should already be setup right, but since you're having the error, perhaps the link above can help

[Jackiexiao]

As @indescribable say, try pip3 install box2d box2d-kengz

@onaclov2000 Thx!
I solve this problem by building from source following instructions found here.

$ conda create -n py34 python=3.4
$ source activate py34
$ conda install -c https://conda.anaconda.org/kne pybox2d

it works well

import gym
env = gym.make('LunarLander-v2')
env.reset()
for _ in range(500):
  env.render()
  env.step(env.action_space.sample())

[nbgraham]

I am trying to use TensorFlow with my GPU on the CarRacing environment. This is really frustrating because TensorFlow with GPU only supports python 3.5.* and 3.6. on Windows.
It seems that Box2D doesn't work with those, but only with 3.4.

(py35_tf_gpu) C:\Users\nickb>conda install -c https://conda.anaconda.org/kne pybox2d
Fetching package metadata ...............
Solving package specifications: .

UnsatisfiableError: The following specifications were found to be in conflict:
  - pybox2d -> python 3.4*
  - python 3.5*
Use "conda info <package>" to see the dependencies for each package.

I tried with the newest swig (3.0.12) and 3.0.2, the one mentioned in https://github.com/openai/gym/issues/100 as working.

Neither one worked, giving the same error:

Traceback (most recent call last):
  File "C:/Users/nickb/PycharmProjects/RL-Race-Car-Simulator/run.py", line 157, in <module>
    main()
  File "C:/Users/nickb/PycharmProjects/RL-Race-Car-Simulator/run.py", line 21, in main
    run_simulator(continue_from, N, name)
  File "C:/Users/nickb/PycharmProjects/RL-Race-Car-Simulator/run.py", line 62, in run_simulator
    env = gym.make('CarRacing-v0')
  File "C:\Users\nickb\Anaconda3\envs\py35_tf_gpu\lib\site-packages\gym\envs\registration.py", line 164, in make
    return registry.make(id)
  File "C:\Users\nickb\Anaconda3\envs\py35_tf_gpu\lib\site-packages\gym\envs\registration.py", line 122, in make
    env = spec.make()
  File "C:\Users\nickb\Anaconda3\envs\py35_tf_gpu\lib\site-packages\gym\envs\registration.py", line 88, in make
    cls = load(self._entry_point)
  File "C:\Users\nickb\Anaconda3\envs\py35_tf_gpu\lib\site-packages\gym\envs\registration.py", line 17, in load
    result = entry_point.load(False)
  File "C:\Users\nickb\Anaconda3\envs\py35_tf_gpu\lib\site-packages\pkg_resources\__init__.py", line 2409, in load
    return self.resolve()
  File "C:\Users\nickb\Anaconda3\envs\py35_tf_gpu\lib\site-packages\pkg_resources\__init__.py", line 2415, in resolve
    module = __import__(self.module_name, fromlist=['__name__'], level=0)
  File "C:\Users\nickb\Anaconda3\envs\py35_tf_gpu\lib\site-packages\gym\envs\box2d\__init__.py", line 1, in <module>
    from gym.envs.box2d.lunar_lander import LunarLander
  File "C:\Users\nickb\Anaconda3\envs\py35_tf_gpu\lib\site-packages\gym\envs\box2d\lunar_lander.py", line 4, in <module>
    import Box2D
  File "C:\Users\nickb\Anaconda3\envs\py35_tf_gpu\lib\site-packages\Box2D\__init__.py", line 20, in <module>
    from .Box2D import *
  File "C:\Users\nickb\Anaconda3\envs\py35_tf_gpu\lib\site-packages\Box2D\Box2D.py", line 435, in <module>
    _Box2D.RAND_LIMIT_swigconstant(_Box2D)
AttributeError: module '_Box2D' has no attribute 'RAND_LIMIT_swigconstant'

[vglazer]

@nbgraham, as far as I can tell the underlying SWIG regression, which is discussed in some detail here, was addressed in this patch.

However, it seems that when you pip install Box2D you pick up a package that was built before the issue was resolved.

One solution is to use box2d-py, as @Jackiexiao suggested above. I'm running Python 3.6, so I had to build box2d-py from source by following these instructions instead of installing the pre-built 3.4 package.

As long python setup.py build picks up a version of SWIG that doesn't have the swigconstant problem, you should be fine. On a Mac, brew install swig does the trick, but I verified that building SWIG 3.0.12 from source also works.

Hope this helps!

[nbgraham]

Installing pybox2d from source seemed to work. I have no idea what minGW and MSYS are, but I have Visual Studio installed, so maybe that's why it worked.

I got this error when trying to use the monitor

raise error.DependencyNotInstalled("""Found neither the ffmpeg nor avconv executables. On OS X, you can install ffmpeg via `brew install ffmpeg`. On most Ubuntu variants, `sudo apt-get install ffmpeg` should do it. On Ubuntu 14.04, however, you'll need to install avconv with `sudo apt-get install libav-tools`.""")
gym.error.DependencyNotInstalled: Found neither the ffmpeg nor avconv executables. On OS X, you can install ffmpeg via `brew install ffmpeg`. On most Ubuntu variants, `sudo apt-get install ffmpeg` should do it. On Ubuntu 14.04, however, you'll need to install avconv with `sudo apt-get install libav-tools`.

It didn't give instructions on how to install ffmpeg on Windows. I found the ffmpeg download page, downloaded the zip, extracted it, and added the bin folder to my path and got it working.

Everything seems to be working, thanks!

[indescribable]

pip3 install box2d box2d-kengz

[qsheeeeen]

python3.5 ubuntu16
pip3 install gym[box2d]
pip3 install box2d
pip3 uninstall box2d-kengz

[xionghhcs]

come into the same problem.

My environment is conda,python=3.5.And I use

conda install -c https://conda.anaconda.org/kne pybox2d

finally it works.@Jackiexiao

[DuaneNielsen]

This is still broken. building gym from source using these instructions..

https://tzattackblog.wordpress.com/2018/01/08/attributeerror-module-_box2d-has-no-attribute-rand_limit_swigconstant/comment-page-1/#comment-4

• installing

pip3 install box2d box2d-kengz

solved it for me, on mac OS

[NBJack]

Thanks for all the suggestions, but this is still broken for myself under Ubuntu 16.04 with Python 3.5.2.

This was marked as closed; what was done to resolve this fundamentally?

[Khev]

I had the same problem. Doing the simple

1. pip install --upgrade pip
2. pip install Box2D

Worked for me. I'm running python 2.7, on ubuntu.

[DuaneNielsen]

@Khev what os were you running? I think OS matters here, as I understand it, mac/windows/linux all need different steps to solve ;)

[Khev]

My bad -- have updated my comment :-)

[DuaneNielsen]

@Khev I realized my comment didnt have OS either :) Fixed it too!

[GoingMyWay]

pip3 install box2d box2d-kengz

It helps on Ubuntu and Windows 10.

[awarebayes]

conda install swig
pip install Box2D

TDS article with more details:
https://towardsdatascience.com/how-to-install-openai-gym-in-a-windows-environment-338969e24d30

[hai-h-nguyen]

python3.5 ubuntu16
pip3 install gym[box2d]
pip3 install box2d
pip3 uninstall box2d-kengz

Important here is the command pip3 install gym[box2d]. Only installing box2d and box2d-kengz does not solve the problem for me.

[VictoriaBentell]

pip install box2d-py worked for me on ubuntu 18.04 and python 3.7