`Distributed Reinforcement Learning (DistRL) is a Python project designed to implement distributed deep reinforcement learning in PyTorch with minimal dependencies. The primary goal is to create a flexible framework for experimenting with multi-objective and meta-learning in a distributed environment.
Requirements: Python >= 3.11.6
Use the package manager pip to install DistRL:
git clone https://github.com/JoseLuisC99/distributed-reinforcement-learning
cd distributed-reinforcement-learning
pip install -r requirements.txt
pip install -e .At the moment, only the next models are available:
- Policy-Parameter Server
- GORILA
- A3C
- IMPALA
- Ape-X
- R2D2
- SEED RL
This demo model only supports Gymnasium environments. modify the policy network in the file launcher.py and then execute the next command:
This demo model currently supports Gymnasium environments. To use it, modify the policy network in the file launcher.py and then execute the following command:
usage: launcher.py --workers WORKERS --master_port MASTER_PORT --env_name ENV_NAME --max_iters MAX_ITERS --max_episodes MAX_EPISODES --output_dir OUTPUT_DIREnsure that MASTER_PORT is a free port on your computer, and ENV_NAME is a valid environment ID.