data
SENetConfig
extends
ModelConfigSENetConfig(num_classes: int = 1000, in_channels: int = 3, layers: tuple[int, ...] = (3, 4, 6, 3), reduction: int = 16, block_type: str = 'bottleneck', legacy_pool: bool = False)Unified config for all SE-ResNet variants (Hu et al., 2017).
Paper: "Squeeze-and-Excitation Networks"
block_type selects BasicBlock (SE-ResNet-18/34) or Bottleneck
(SE-ResNet-50/101/152). layers is the per-stage repetition count.
reduction is the channel reduction ratio in the SE block (default 16).
legacy_pool selects the original SENet stem max-pool geometry
(padding=0, ceil_mode=True) used by the canonical SE-ResNet
line; False (default) uses the modern ResNet stem pool
(padding=1, ceil_mode=False).