Cannot run the run.py because mpi4py has no module "MPI"

[Niels-Sch]

(tf-nightly) C:\Users\Niels\Documents\GitHub\large-scale-curiosity>python run.py
no OpenGL.GLU
Traceback (most recent call last):
  File "run.py", line 15, in <module>
    from mpi4py import MPI
ImportError: DLL load failed: The specified module could not be found.

I installed mpi4py through pip install. Any got any idea how to fix this?

[kevin5k]

@Niels-ns + @jachiam,
I have recently been having encountering this issue on Win10 via the anaconda environment and I could not find anyone else encountering this issue. Hence, I would like to share the solution here -->

The issue I encountered was that when I executed the ppo installtest command, I faced the same error message which you had reported. It appears that is more of a Windows 10 environment issue than the anaconda issue.

Solution

1. Download and install the Windows MPI SDK and exe from microsoft.
2. Upon completion, close all instances of Anaconda
3. Launch a new instance of Anaconda and activate your environment (eg. spinningup)
4. Execute "python -m spinup.run ppo --hid "[32,32]" --env LunarLander-v2 --exp_name installtest --gamma 0.999"

[Hsgngr]

The issue I encountered was that when I executed the ppo installtest command, I faced the same error message which you had reported. It appears that is more of a Windows 10 environment issue than the anaconda issue.

Solution
This doesn't work.

[Hsgngr]

Currently the support of OpenAI for Windows is weak, everything works smoothly in WLS or directly ubuntu.