cellmil.interfaces.AttentionExplainerConfig

Classes

Aggregation(value)	Attention aggregation methods for GraphMIL.
AttentionExplainerConfig(*, output_path[, ...])	Configuration for Attention explainability method.
Normalization(value)	Attention normalization methods.
VisualizationMode(value)	Visualization output modes.

class cellmil.interfaces.AttentionExplainerConfig.Aggregation(value)

Bases: str, Enum

Attention aggregation methods for GraphMIL.

pooling_only = 'pooling_only'

gnn_only = 'gnn_only'

gnn_layer = 'gnn_layer'

random_walk = 'random_walk'

_generate_next_value_(start, count, last_values)

Generate the next value when not given.

name: the name of the member start: the initial start value or None count: the number of existing members last_value: the last value assigned or None

class cellmil.interfaces.AttentionExplainerConfig.Normalization(value)

Bases: str, Enum

Attention normalization methods.

min_max = 'min_max'

z_score = 'z_score'

robust = 'robust'

softmax = 'softmax'

sigmoid = 'sigmoid'

none = 'none'

_generate_next_value_(start, count, last_values)

Generate the next value when not given.

name: the name of the member start: the initial start value or None count: the number of existing members last_value: the last value assigned or None

class cellmil.interfaces.AttentionExplainerConfig.VisualizationMode(value)

Bases: str, Enum

Visualization output modes.

geojson = 'geojson'

graph = 'graph'

all = 'all'

_generate_next_value_(start, count, last_values)

Generate the next value when not given.

name: the name of the member start: the initial start value or None count: the number of existing members last_value: the last value assigned or None

class cellmil.interfaces.AttentionExplainerConfig.AttentionExplainerConfig(*, output_path: Path, attention_aggregation: Aggregation = Aggregation.pooling_only, gnn_layer_index: Optional[int] = None, class_index: Optional[int] = None, attention_head: Optional[int] = None, visualization_mode: VisualizationMode = VisualizationMode.geojson, color_scheme: tuple[list[int], list[int]] = ([35, 92, 236], [255, 0, 0]), normalize_attention: bool = True, normalization: Normalization = Normalization.min_max, visualize_cell_types: bool = False)

Bases: BaseModel

Configuration for Attention explainability method.

output_path: Path

attention_aggregation: Aggregation

gnn_layer_index: Optional[int]

class_index: Optional[int]

attention_head: Optional[int]

visualization_mode: VisualizationMode

color_scheme: tuple[list[int], list[int]]

normalize_attention: bool

normalization: Normalization

visualize_cell_types: bool

classmethod validate_gnn_layer_index(v: Optional[int]) → Optional[int]

model_config: ClassVar[ConfigDict] = {}

Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].