data
GoogLeNetConfig
extends
ModelConfigGoogLeNetConfig(num_classes: int = 1000, in_channels: int = 3, aux_logits: bool = True, transform_input: bool = False, dropout: float = 0.4, aux_dropout: float = 0.7)Configuration for GoogLeNet (Inception v1).
aux_logits enables the two auxiliary classifiers that were used
during training in the original paper. They are ignored at inference
(eval mode); the auxiliary loss is typically weighted 0.3 in the total
training objective.
dropout — main classifier dropout rate (0.4 in the paper).
aux_dropout — auxiliary classifier dropout rate (0.7 in the paper).
Examples
>>> from lucid.models.vision.googlenet._config import GoogLeNetConfig
>>> cfg = GoogLeNetConfig()
>>> cfg.model_type
'googlenet'
>>> cfg.num_classes, cfg.in_channels
(1000, 3)