Difficulty running Client Tools and Visualizer/Gym


I’ve downloaded the Client Tools from the bottom of the downloads page:

When I run client.py with Python3 I get this output:

Traceback (most recent call last):
  File "./client.py", line 9, in <module>
    from . import output
ImportError: cannot import name 'output' from '__main__' (./client.py)

I also tried with Python2 just in case, but it failed with a different error (not posted because the client.py file indicates that Python3 is correct).

Am I doing something incorrectly here, or should this run? I’m on Arch Linux with Python 3.7.0.

Also I see that the Halite Visualizer and Gym has no source or Linux version - will this remain the case? Is that tool different from the web-based editor on this site?

Just trying to work out my testing options before I embark on rolling my own solution.


Hi @Akarabesu – thanks for your questions! We will add better documentation around the offline tools. :slight_smile:

To run the client, run:
python -m hlt_client

Instructions on building the Visualizer and Gym for Linux are in our repo, here: https://github.com/HaliteChallenge/Halite-III/tree/master/tools/halite-in-a-box.

With this tool, you can play and visualize games with your local bot against your submitted and select other bots.


It doesnt work for me on python 3 either. I’ve changed all files in the next manner:
from . import output => import output as output


I’ve made some progress with the visualizer + gym. At first I was stuck on step 3, until I realised that “from here” means tools/halite-in-a-box.

Also, for step 4 on linux I think the command should be:
./node_modules/.bin/electron-builder -l

When I run ./linux-unpacked/halite-in-a-box I get this window:

Is this working correctly, or should there be content in that blank window? I’m not quite sure what to do from here. :slight_smile:


Is the Visualizer/Gym only for Python bots?? Looks nice and has a great interface.

I’m looking for a fast way to test multiple games with dif board sizes with either c++ or javascript. Currently running offline, but one game at a time doesn’t help for machine learning optimization.