search

Nullspace-Colombia / unray-bridge

Rllib framework for using Unreal Engine 5 (UE5) as external environment for Reinforced Learning training process
MIT License
8 stars 2 forks source link

Documentation clarification #12

Closed f00d4tehg0dz closed 10 months ago

f00d4tehg0dz commented 10 months ago

Hey Team,

I'm running into issues trying out both examples you have provided.

Issue

When running in command prompt python parallel_multiagentArena.py I receive this error: OverflowError: int too big to convert And when running in command prompt python main_cartpole.py I receive this error [ CONNECTION ] connecting to localhost port 10010 Trying to connect... Trying to connect... Trying to connect... Trying to connect... Connection Timeout!

What's interesting is when running in command prompt netstat -a | findstr :10011 I do get a response from UE5.3 TCP 0.0.0.0:10011 www:0 LISTENING

Steps to reproduce

Create a new VirtualEnv with Anaconda 3.

  1. Use Python 3.10.x (have tried 3.8.x as well)
  2. git clone https://github.com/Nullspace-Colombia/unray-bridge.git cd unray-bridge pip install -r requirements.txt pip install tensorflow
  3. In Anaconda 3 select your new Virtual Environment and click the green play button and select Open Terminal
  4. Enable the UE 5.3 plugin Remote Control API
  5. Restart UE 5.3
  6. Inside Project Settings, select Plugins - Remote Control
  7. Under Remote Control WebSocket Server Port, enter in 10011
  8. Select map MultiAgentArena
  9. Press the Green Play button in the UE 5.3 Editor
  10. In your Terminal (command prompt)
  11. Run in command prompt python main_multiagentArena.py
  12. Wait for error

Hardware

f00d4tehg0dz commented 10 months ago

Disregard. My initial problem was I was unable to connect via localhost from your Python script, so I assumed improperly that I needed to use the remote control plugin.

Closing this as I'm good now.

GDiaz16 commented 10 months ago

@f00d4tehg0dz Great that you solved the issue!