• Melting Pot Contest (AICrowd)

Tess-v1 is a reinforcement learning solution developed for the NeurIPS Melting Pot Contest. The repository contains various training and model files, with models/model_v4_cont.py being applied in all the final experiments.

• Introduction
• Trainings
• Prerequisites
• License
• Citation

Tess-v1 is a reinforcement learning solution tailored for the challenges posed in the NeurIPS Melting Pot Contest. The repository encompasses various training and model files, with particular emphasis on models/model_v4_cont.py, which is employed in all the final experiments.

• Harvest: Trained using train_multitask_harvest.py
• Clean Up: Trained using train_multitask.py
• Prisoners: Trained using train_multitask.py
• Territory Rooms: Trained using train-v6.py

Note: For reproducibility, it is essential to configure the parameters in configs.py. The necessary configuration details will be shared in the next commit.

Before running Tess-v1, ensure you have the following prerequisites installed:

• PyTorch (version >= 2.0.1)
• Stable-Baselines3
• Gym (version 0.21.0)
• OpenCV-Python
• NumPy
• WandB (Weights & Biases)
• MeltingPot Suite

You can install these dependencies using the following:

pip install torch>=2.0.1 stable-baselines3 gym==0.21.0 opencv-python numpy wandb dm-meltingpot

This project is licensed under the MIT License

Version		: 1
Released 	: 2023-12-26
Author		:: Utku Erodoğanaras
Contacts 	: [email protected]

Please site this repository in this format.