SENetConfig

class lucid.models.SENetConfig(block: Literal['se_basic', 'bottleneck'], layers: tuple[int, int, int, int] | list[int], reduction: int = 16, cardinality: int = 1, base_width: int = 64, num_classes: int = 1000, in_channels: int = 3, stem_width: int = 64, stem_type: Literal['deep'] | None = None, avg_down: bool = False, channels: tuple[int, int, int, int] | list[int] = (64, 128, 256, 512), block_args: dict[str, typing.Any] = <factory>)

SENetConfig stores the squeeze-and-excitation backbone choices used by lucid.models.SENet.

Class Signature

@dataclass
class SENetConfig:
    block: Literal["se_basic", "bottleneck"]
    layers: tuple[int, int, int, int] | list[int]
    reduction: int = 16
    cardinality: int = 1
    base_width: int = 64
    num_classes: int = 1000
    in_channels: int = 3
    stem_width: int = 64
    stem_type: Literal["deep"] | None = None
    avg_down: bool = False
    channels: tuple[int, int, int, int] | list[int] = (64, 128, 256, 512)
    block_args: dict[str, Any] = field(default_factory=dict)