InceptionConfig¶
- class lucid.models.InceptionConfig(variant: Literal['v1', 'v3', 'v4'], num_classes: int = 1000, in_channels: int = 3, use_aux: bool | None = None, dropout_prob: float | None = None)¶
InceptionConfig stores the architectural choices used by lucid.models.Inception.
It defines which Inception variant to build and the common runtime options shared
by the v1, v3, and v4 families.
Class Signature¶
@dataclass
class InceptionConfig:
variant: Literal["v1", "v3", "v4"]
num_classes: int = 1000
in_channels: int = 3
use_aux: bool | None = None
dropout_prob: float | None = None
Parameters¶
variant (Literal[“v1”, “v3”, “v4”]): Inception family variant to build.
num_classes (int): Number of output classes.
in_channels (int): Number of channels in the input image tensor.
use_aux (bool | None): Auxiliary classifier flag for v1/v3. For v4 it is normalized to False.
dropout_prob (float | None): Optional dropout override. If omitted, each variant uses its existing default.
Validation¶
variant must be one of “v1”, “v3”, or “v4”.
num_classes and in_channels must be greater than 0.
dropout_prob, when provided, must be in the range [0.0, 1.0).
use_aux must be boolean for v1 and v3.
Inception v4 does not support auxiliary classifiers.
Usage¶
import lucid.models as models
config = models.InceptionConfig(
variant="v4",
num_classes=10,
in_channels=1,
dropout_prob=0.25,
)
model = models.Inception(config)