imitation.scripts.ingredients.environment#
This ingredient provides a vectorized gym environment.
Functions
|
Builds the vector environment for rollouts. |
|
Builds the vector environment. |
- imitation.scripts.ingredients.environment.make_rollout_venv(gym_id, num_vec, parallel, max_episode_steps, env_make_kwargs, _rnd)[source]#
Builds the vector environment for rollouts.
This environment does no logging, and it is wrapped in a RolloutInfoWrapper.
- Parameters
gym_id (
str) – The id of the environment to create.num_vec (
int) – Number of gym.Env instances to combine into a vector environment.parallel (
bool) – Whether to use “true” parallelism. If True, then use SubProcVecEnv. Otherwise, use DummyVecEnv which steps through environments serially.max_episode_steps (
int) – If not None, then a TimeLimit wrapper is applied to each environment to artificially limit the maximum number of timesteps in an episode.env_make_kwargs (
Mapping[str,Any]) – The kwargs passed to spec.make of a gym environment._rnd (
Generator) – Random number generator provided by Sacred.
- Yields
The constructed vector environment.
- Return type
Generator[VecEnv,None,None]
- imitation.scripts.ingredients.environment.make_venv(gym_id, num_vec, parallel, max_episode_steps, env_make_kwargs, _run, _rnd, **kwargs)[source]#
Builds the vector environment.
- Parameters
gym_id (
str) – The id of the environment to create.num_vec (
int) – Number of gym.Env instances to combine into a vector environment.parallel (
bool) – Whether to use “true” parallelism. If True, then use SubProcVecEnv. Otherwise, use DummyVecEnv which steps through environments serially.max_episode_steps (
int) – If not None, then a TimeLimit wrapper is applied to each environment to artificially limit the maximum number of timesteps in an episode.env_make_kwargs (
Mapping[str,Any]) – The kwargs passed to spec.make of a gym environment.kwargs – Passed through to util.make_vec_env.
- Yields
The constructed vector environment.
- Return type
Generator[VecEnv,None,None]