Skip to content

MinariCollectionSettings

Full path: schola.scripts.minari.settings.MinariCollectionSettings

Dataclass for configuring Minari dataset collection parameters.

MinariCollectionSettings(dataset_id = None, num_steps = 1000, seed = None,
                         author = None, author_email = None,
                         code_permalink = None, algorithm_name = None,
                         description = None, record_infos = False,
                         data_path = None)

Parameters

  • dataset_id (str)

  • num_steps (Annotated)

  • seed (int)

  • author (str)

  • author_email (Any)

  • code_permalink (Any)

  • algorithm_name (str)

  • description (str)

  • record_infos (bool)

  • data_path (Annotated)

Methods

init

__init__(dataset_id = None, num_steps = 1000, seed = None, author = None,
         author_email = None, code_permalink = None, algorithm_name = None,
         description = None, record_infos = False, data_path = None)

Parameters

  • dataset_id (str)

  • num_steps (Annotated)

  • seed (int)

  • author (str)

  • author_email (Any)

  • code_permalink (Any)

  • algorithm_name (str)

  • description (str)

  • record_infos (bool)

  • data_path (Annotated)

Attributes

algorithm_name

algorithm_name

Name of the algorithm or policy used to collect the data.

author

author

Author name for the dataset metadata.

author_email

author_email

Author email for the dataset metadata.

code_permalink

URL to the code or repository used to generate the dataset.

data_path

data_path

Directory path where Minari datasets will be stored. If None, uses the default Minari datasets directory (MINARI_DATASETS_PATH environment variable or ~/.minari/datasets/).

dataset_id

dataset_id

Unique identifier for the Minari dataset. This will be used to name the dataset when it is created.

description

description

Description of the dataset.

num_steps

num_steps

Total number of steps to collect for the dataset. This is the total number of environment steps that will be recorded.

record_infos

record_infos

Whether to record the info dictionaries in the dataset. If False, only observations, actions, rewards, terminations, and truncations are recorded.

seed

seed

Random seed for reproducibility. If None, the environment will use a random seed.