`fragile.core.walkers`#

Module Contents#

Classes#

`SimpleWalkers`	The WalkersAPI class defines the base functionality for managing walkers in a swarm simulation.
`ScoreMetric`	A base class for score metrics in a swarm of walkers.
`RewardScore`	A class representing the reward score metric for a swarm of walkers.
`SonicScore`	A base class for score metrics in a swarm of walkers.
`MarioScore`	A base class for score metrics in a swarm of walkers.
`DiversityMetric`	A base class for diversity metrics in a swarm of walkers.
`RandomDistance`	A class representing the random distance diversity metric for a swarm of walkers.
`Walkers`	Walkers class handles the walkers' dynamics, including scoring, diversity, cloning, and resetting.
`ExplorationWalkers`	Walkers class handles the walkers' dynamics, including scoring, diversity, cloning, and resetting.
`NoBalance`	A class representing walkers with no balancing behavior in a swarm.

Functions#

l2_norm(x, y)

Compute the L2 norm between two tensors.

class fragile.core.walkers.SimpleWalkers(accumulate_reward=True, score_scale=1.0, diversity_scale=1.0, minimize=False, **kwargs)[source]#

Bases: fragile.core.api_classes.WalkersAPI

The WalkersAPI class defines the base functionality for managing walkers in a swarm simulation. This class inherits from the SwarmComponent class.

Parameters

accumulate_reward (bool) –
score_scale (float) –
diversity_scale (float) –
minimize (bool) –

default_inputs#

default_param_dict#

default_outputs#

run_epoch(observs, rewards, scores, oobs=None, inplace=True, **kwargs)[source]#

Implement the functionality for running an epoch in the derived class. This method is called during the balance operation.

Parameters

inplace (bool, optional) – If True, updates the swarm state with the data generated during the epoch. If False, returns the data. Defaults to True.
**kwargs – Additional keyword arguments required for running the epoch.

Returns

A dictionary containing the data generated during the epoch if inplace is: False. Otherwise, returns None.

Return type

StateData

fragile.core.walkers.l2_norm(x, y)[source]#

Compute the L2 norm between two tensors.

Parameters

x (Tensor) – The first tensor.
y (Tensor) – The second tensor.

Returns

The L2 norm between the two input tensors.

Return type

Tensor

class fragile.core.walkers.ScoreMetric(swarm=None, param_dict=None, inputs=None, outputs=None)[source]#

Bases: fragile.core.api_classes.WalkersMetric

A base class for score metrics in a swarm of walkers.

Parameters

swarm (Optional[SwarmAPI]) –
param_dict (Optional[fragile.core.typing.StateDict]) –
inputs (Optional[fragile.core.typing.InputDict]) –
outputs (Optional[Tuple[str]]) –

default_param_dict#

Default parameters dictionary for the score metric.

Type: dict

default_outputs#

Tuple containing the default output keys.

Type: tuple

default_param_dict#

default_outputs#

class fragile.core.walkers.RewardScore(accumulate_reward=True, keep_max_reward=False, **kwargs)[source]#

Bases: ScoreMetric

A class representing the reward score metric for a swarm of walkers.

Parameters

accumulate_reward (bool) –
keep_max_reward (bool) –

default_inputs#

Default inputs dictionary for the reward score metric.

Type: dict

default_inputs#

property inputs#

Return a dictionary containing the data that this component needs to function.

Return type: fragile.core.typing.InputDict

calculate(rewards=None, scores=None, **kwargs)[source]#

Calculate the scores for the walkers based on rewards.

Parameters

rewards (array) – Array of walkers’ rewards.
scores (array, optional) – Array of walkers’ scores.
**kwargs – Additional keyword arguments.

Returns

A dictionary containing the scores.

Return type

dict

reset(inplace=True, root_walker=None, states=None, **kwargs)[source]#

Reset the reward score metric.

Parameters

inplace (bool, optional) – Whether to reset in-place. Defaults to True.
root_walker (StateData, optional) – The root walker to reset. Defaults to None.
states (Optional[StateData], optional) – The state data to reset. Defaults to None.
**kwargs – Additional keyword arguments.

Return type

None

class fragile.core.walkers.SonicScore(swarm=None, param_dict=None, inputs=None, outputs=None)[source]#

Bases: ScoreMetric

A base class for score metrics in a swarm of walkers.

Parameters

swarm (Optional[SwarmAPI]) –
param_dict (Optional[fragile.core.typing.StateDict]) –
inputs (Optional[fragile.core.typing.InputDict]) –
outputs (Optional[Tuple[str]]) –

default_param_dict#

Default parameters dictionary for the score metric.

Type: dict

default_outputs#

Tuple containing the default output keys.

Type: tuple

accumulate_reward = False#

default_inputs#

static score_from_info(info)[source]#

calculate(rewards, scores=None, **kwargs)[source]#

class fragile.core.walkers.MarioScore(swarm=None, param_dict=None, inputs=None, outputs=None)[source]#

Bases: ScoreMetric

A base class for score metrics in a swarm of walkers.

Parameters

swarm (Optional[SwarmAPI]) –
param_dict (Optional[fragile.core.typing.StateDict]) –
inputs (Optional[fragile.core.typing.InputDict]) –
outputs (Optional[Tuple[str]]) –

default_param_dict#

Default parameters dictionary for the score metric.

Type: dict

default_outputs#

Tuple containing the default output keys.

Type: tuple

name = MarioScore#

accumulate_reward = False#

default_inputs#

static score_from_info(info)[source]#

calculate(rewards=None, scores=None, **kwargs)[source]#

class fragile.core.walkers.DiversityMetric(swarm=None, param_dict=None, inputs=None, outputs=None)[source]#

Bases: fragile.core.api_classes.WalkersMetric

A base class for diversity metrics in a swarm of walkers.

Parameters

swarm (Optional[SwarmAPI]) –
param_dict (Optional[fragile.core.typing.StateDict]) –
inputs (Optional[fragile.core.typing.InputDict]) –
outputs (Optional[Tuple[str]]) –

default_param_dict#

Default parameters dictionary for the diversity metric.

Type: dict

default_outputs#

Tuple containing the default output keys.

Type: tuple

default_param_dict#

default_outputs#

class fragile.core.walkers.RandomDistance(swarm=None, param_dict=None, inputs=None, outputs=None)[source]#

Bases: DiversityMetric

A class representing the random distance diversity metric for a swarm of walkers.

Parameters

swarm (Optional[SwarmAPI]) –
param_dict (Optional[fragile.core.typing.StateDict]) –
inputs (Optional[fragile.core.typing.InputDict]) –
outputs (Optional[Tuple[str]]) –

default_inputs#

Default inputs dictionary for the random distance diversity metric.

Type: dict

use_pbc = False#

default_inputs#

calculate(observs, oobs, **kwargs)[source]#

Calculate the diversities for the walkers based on the L2 distance to another walker chosen at random.

Parameters

observs (array) – Array of walkers’ observations.
oobs (array) – Array of out-of-bounds flags for walkers.
**kwargs – Additional keyword arguments.

Returns

A dictionary containing the diversities.

Return type

dict

class fragile.core.walkers.Walkers(score=None, diversity=None, minimize=False, score_scale=1.0, diversity_scale=1.0, track_data=None, accumulate_reward=True, keep_max_reward=False, clone_period=1, freeze_walkers=True, **kwargs)[source]#

Bases: fragile.core.api_classes.WalkersAPI

Walkers class handles the walkers’ dynamics, including scoring, diversity, cloning, and resetting.

Inherits from WalkersAPI.

Parameters

score (ScoreMetric) –
diversity (DiversityMetric) –
minimize (bool) –
score_scale (float) –
diversity_scale (float) –
accumulate_reward (bool) –
keep_max_reward (bool) –
clone_period (int) –

default_param_dict#

default_outputs = ['compas_clone', 'virtual_rewards', 'clone_probs', 'will_clone']#

default_inputs#

property param_dict#

Return the dictionary defining all the data attributes that the component requires.

Return type: fragile.core.typing.StateDict

property inputs#

Return a dictionary containing the data that this component needs to function.

Return type: fragile.core.typing.InputDict

property outputs#

Return a tuple containing the names of the data attribute that the component outputs.

Return type: Tuple[str, Ellipsis]

setup(swarm)[source]#

Set up the Walkers object by initializing its swarm.

Parameters: swarm – Swarm object to be initialized.

balance(inplace=True, **kwargs)[source]#

Balance the walkers.

Parameters

inplace (bool, optional) – Whether to perform the operation inplace. Defaults to True.
**kwargs – Additional keyword arguments.

Returns

None if inplace=True, otherwise a StateData object.

Return type

Union[None, StateData]

run_epoch(inplace=True, oobs=None, **kwargs)[source]#

Execute an epoch in the walkers’ dynamics.

Parameters

inplace (bool, optional) – Whether to perform the operation inplace. Defaults to True.
oobs – Out of bounds flags.
**kwargs – Additional keyword arguments.

Returns

State data after the epoch execution.

Return type

StateData

calculate_virtual_reward(scores, diversities, **kwargs)[source]#

Calculate the virtual rewards for walkers based on scores and diversities.

Parameters

scores (array) – Array of walkers’ scores.
diversities (array) – Array of walkers’ diversities.
**kwargs – Additional keyword arguments.

Returns

A dictionary containing the virtual rewards.

Return type

dict

calculate_clones(virtual_rewards, oobs=None)[source]#

Calculate the walkers that will clone and their target companions.

Parameters

virtual_rewards (array) – Array of walkers’ virtual rewards.
oobs (array, optional) – Array of out-of-bounds flags for walkers. Defaults to None.

Returns

A dictionary containing clone probabilities, will_clone flags, companion clones, and active flags.

Return type

dict

reset(inplace=True, **kwargs)[source]#

Reset the Walkers object and its score and diversity components.

Parameters

inplace (bool, optional) – Whether to perform the operation inplace. Defaults to True.
**kwargs – Additional keyword arguments.

Return type

None

class fragile.core.walkers.ExplorationWalkers(exploration_scale=1.0, **kwargs)[source]#

Bases: Walkers

Walkers class handles the walkers’ dynamics, including scoring, diversity, cloning, and resetting.

Inherits from WalkersAPI.

Parameters: exploration_scale (float) –

property explore_counts#

Return type: Dict[Any, int]

calculate_virtual_reward(scores, diversities, **kwargs)[source]#: Apply the virtual reward formula to account for all the different goal scores.

get_explore_rewards()[source]#

get_coords_keys()[source]#

class fragile.core.walkers.NoBalance(score=None, diversity=None, minimize=False, score_scale=1.0, diversity_scale=1.0, track_data=None, accumulate_reward=True, keep_max_reward=False, clone_period=1, freeze_walkers=True, **kwargs)[source]#

Bases: Walkers

A class representing walkers with no balancing behavior in a swarm.

Inherits from the Walkers class and modifies the properties and methods to disable balancing.

Parameters

score (ScoreMetric) –
diversity (DiversityMetric) –
minimize (bool) –
score_scale (float) –
diversity_scale (float) –
accumulate_reward (bool) –
keep_max_reward (bool) –
clone_period (int) –

property param_dict#

Return the dictionary defining all the data attributes that the component requires.

Return type: fragile.core.typing.StateDict

property inputs#

Return a dictionary containing the data that this component needs to function.

Return type: fragile.core.typing.InputDict

property outputs#

Return a tuple containing the names of the data attribute that the component outputs.

Return type: Tuple[str, Ellipsis]

run_epoch(inplace=True, oobs=None, **kwargs)[source]#

Run an epoch for the walkers without balancing.

Parameters

inplace (bool, optional) – Whether to run the epoch in-place. Defaults to True.
oobs (array, optional) – Array of out-of-bounds walkers. Defaults to None.
**kwargs – Additional keyword arguments.

Returns

A dictionary containing the scores.

Return type

dict

fragile.core.walkers

Contents

fragile.core.walkers#

Module Contents#

Classes#

Functions#

`fragile.core.walkers`#