ResNeStConfig¶

class lucid.models.ResNeStConfig(layers: tuple[int, int, int, int] | list[int], base_width: int = 64, stem_width: int = 32, cardinality: int = 1, radix: int = 2, avd: bool = True, num_classes: int = 1000, in_channels: int = 3, avg_down: bool = False, channels: tuple[int, int, int, int] | list[int] = (64, 128, 256, 512), block_args: dict[str, typing.Any] = <factory>)¶

ResNeStConfig stores the split-attention backbone choices used by lucid.models.ResNeSt.

Class Signature¶

@dataclass
class ResNeStConfig:
    layers: tuple[int, int, int, int] | list[int]
    base_width: int = 64
    stem_width: int = 32
    cardinality: int = 1
    radix: int = 2
    avd: bool = True
    num_classes: int = 1000
    in_channels: int = 3
    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)

Parameters¶

layers (tuple[int, int, int, int] | list[int]): Number of bottleneck blocks in each of the four stages.
base_width (int): Width parameter used to derive grouped bottleneck channels.
stem_width (int): Width of the deep stem.
cardinality (int): Number of groups used by split-attention bottlenecks.
radix (int): Number of attention splits per bottleneck. 0 falls back to the grouped-convolution path.
avd (bool): Whether average downsampling is enabled inside the bottleneck.
num_classes (int): Number of output classes.
in_channels (int): Number of channels in the input image tensor.
avg_down (bool): Whether projection shortcuts should downsample using average pooling before the 1x1 projection.
channels (tuple[int, int, int, int] | list[int]): Output width of each residual stage.
block_args (dict[str, Any]): Extra keyword arguments forwarded to the ResNeSt bottleneck, in addition to the preset split-attention settings.

Validation¶

layers and channels must each contain exactly four positive integers.
base_width, stem_width, cardinality, num_classes, and in_channels must be greater than 0.
radix must be greater than or equal to 0.
block_args must be a dictionary.