vectorization_parallel_env

Vectorize PettingZoo ParallelEnv

Installation

pip install git+https://github.com/Raffaelbdl/vectorization_parallel_env

Example Usage

from vec_parallel_env import SubProcVecParallelEnv
from pettingzoo.mpe import simple_spread_v3

env_fn = simple_spread_v3.parallel_env()
envs = SubProcVecParallelEnv([lambda: env_fn for _ in range(5)])

# training loop

envs.close()

API details

envs.step takes a dictionary of batched actions

action_spaces = envs.action_spaces
actions = {
    agent: np.stack([space.sample() for _ in range(envs.num_agents)])
    for agent, space in action_spaces.items()
}

envs.step returns a gymnasium-style tuple where each element is a dictionary of batched values

observation_spaces = envs.observation_spaces
observations = {
    agent: np.stack([space.sample() for _ in range(envs.num_agents)])
    for agent, space in observation_spaces.items()
}