Behavior Suite

DeepMind Behavior Suite

Behavior Suite is a collection of carefully-designed experiments that investigate various aspects of agent behavior through shared benchmarks.

Shimmy provides compatibility wrappers to convert Behavior Suite environments to Gymnasium.

Installation

To install shimmy and required dependencies:

pip install shimmy[bsuite]

We also provide a Dockerfile for reproducibility and cross-platform compatibility:

curl https://raw.githubusercontent.com/Farama-Foundation/Shimmy/main/bin/bsuite.Dockerfile | docker build -t bsuite -f - . && docker run -it bsuite

Usage

Load a bsuite environment:

import gymnasium as gym

env = gym.make("bsuite/catch-v0", render_mode="human")

Run the environment:

observation, info = env.reset(seed=42)
for _ in range(1000):
   action = env.action_space.sample()  # this is where you would insert your policy
   observation, reward, terminated, truncated, info = env.step(action)

   if terminated or truncated:
      observation, info = env.reset()
env.close()

To get a list of all available bsuite environments (23 total):

from gymnasium.envs.registration import registry
BSUITE_ENV_IDS = [
    env_id
    for env_id in registry
    if env_id.startswith("bsuite") and env_id != "bsuite/compatibility-env-v0"
]
print(BSUITE_ENV_IDS)

Class Description

class shimmy.bsuite_compatibility.BSuiteCompatibilityV0(env: Environment, render_mode: str | None = None)

A compatibility wrapper that converts a BSuite environment into a gymnasium environment.

Note

Bsuite uses np.random.RandomState, a legacy random number generator while gymnasium uses np.random.Generator, therefore the return type of np_random is different from expected.

Initialises the environment with a render mode along with render information.

metadata: dict[str, Any] = {'render_modes': []}

observation_space: spaces.Space[ObsType]

action_space: spaces.Space[ActType]

reset(*, seed: int | None = None, options: dict[str, Any] | None = None) → tuple[ObsType, dict[str, Any]]

Resets the bsuite environment.

step(action: int) → tuple[ObsType, float, bool, bool, dict[str, Any]]

Steps through the bsuite environment.

render() → np.ndarray | None

Renders the bsuite env.

close()

Closes the environment.

property np_random: RandomState

This should be np.random.Generator but bsuite uses np.random.RandomState.