ConvNeXtConfig¶

class lucid.models.ConvNeXtConfig(num_classes: int = 1000, depths: tuple[int, int, int, int] | list[int] = (3, 3, 9, 3), dims: tuple[int, int, int, int] | list[int] = (96, 192, 384, 768), drop_path: float = 0.0, layer_scale_init: float = 1e-06)¶

ConvNeXtConfig stores the stage layout and classifier settings used by lucid.models.ConvNeXt. It defines the four-stage depth profile, stage widths, classifier size, drop-path rate, and layer-scale initialization.

Class Signature¶

@dataclass
class ConvNeXtConfig:
    num_classes: int = 1000
    depths: tuple[int, int, int, int] | list[int] = (3, 3, 9, 3)
    dims: tuple[int, int, int, int] | list[int] = (96, 192, 384, 768)
    drop_path: float = 0.0
    layer_scale_init: float = 1e-6

Parameters¶

num_classes (int): Number of output classes.
depths: Four-stage block counts for the ConvNeXt hierarchy.
dims: Four-stage channel widths for the ConvNeXt hierarchy.
drop_path (float): Global drop-path rate distributed across the stage blocks.
layer_scale_init (float): Initial layer-scale value used inside ConvNeXt blocks.

Validation¶

num_classes must be greater than 0.
depths and dims must contain exactly four positive integers.
drop_path must be in the range [0, 1].
layer_scale_init must be greater than or equal to 0.

Usage¶

import lucid.models as models

config = models.ConvNeXtConfig(
    num_classes=10,
    depths=(2, 2, 6, 2),
    dims=(64, 128, 256, 512),
    drop_path=0.1,
    layer_scale_init=0.0,
)
model = models.ConvNeXt(config)