ML-Agents SoccerTwos Model for Deep RL Course Unit 7
This is a trained multi-agent reinforcement learning model using the POCA (MA-POCA) trainer algorithm to play soccer in the Unity SoccerTwos environment.
This model was trained as part of the Hugging Face Deep Reinforcement Learning Course.
Environment Details
- Name: SoccerTwos
- Number of Teams: 2 (Blue vs Purple)
- Agents per Team: 2
- Goal: Outscore the opposing team within the time limit.
Training Configuration
The training was conducted with shortened parameters to optimize computation overhead:
- Max Steps: 500,000
- Hidden Units: 128
- Batch Size: 1024
- Trainer Type: poca
How to Use
To inspect or utilize this policy file locally within your ML-Agents environment workspace, look at the uploaded .onnx binary file inside this repository.