ConvNeXtV2Config¶

class lucid.models.ConvNeXtV2Config(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)¶

ConvNeXtV2Config stores the stage layout and classifier settings used by lucid.models.ConvNeXt_V2. It defines the four-stage depth profile, stage widths, classifier size, and drop-path rate for ConvNeXt-v2 variants.

Class Signature¶

@dataclass
class ConvNeXtV2Config:
    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

Parameters¶

num_classes (int): Number of output classes.
depths: Four-stage block counts for the ConvNeXt-v2 hierarchy.
dims: Four-stage channel widths for the ConvNeXt-v2 hierarchy.
drop_path (float): Global drop-path rate distributed across the stage 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].

Usage¶

import lucid.models as models

config = models.ConvNeXtV2Config(
    num_classes=10,
    depths=(2, 2, 6, 2),
    dims=(40, 80, 160, 320),
    drop_path=0.1,
)
model = models.ConvNeXt_V2(config)