A2Perf is a benchmarking suite for evaluating agents on sequential decision-making problems that are relevant to the real world.

This library contains a collection of environments from domains spanning computer chip-floorplanning, web navigation, and quadruped locomotion.

The Gymnasium interface allows users to initialize and interact with the A2Perf environments as follows:

import gymnasium as gym
from a2perf.domains import circuit_training
# from a2perf.domains import web_navigation
# from a2perf.domains import quadruped_locomotion

# Choose one of the A2Perf environments
env = gym.make("CircuitTraining-Ariane-v0")
# or env = gym.make("WebNavigation-Difficulty-01-v0")
# or env = gym.make("QuadrupedLocomotion-DogPace-v0")

observation, info = env.reset(seed=42)
for _ in range(1000):
    action = env.action_space.sample()  # Replace with your policy
    observation, reward, terminated, truncated, info = env.step(action)
    if terminated or truncated:
        observation, info = env.reset()
env.close()