convnext_large(pretrained: bool = False, overrides: object = {})ConvNeXt-Large backbone (Liu et al., 2022).
Builds the canonical ConvNeXt-L configuration:
depths=(3, 3, 27, 3), dims=(192, 384, 768, 1536). Final
feature width is 1536. Approximately 198M parameters — Swin-L
equivalent.
Model Size
Parameters
pretrainedbool= FalseIf
True, loads ImageNet-22k pretrained weights when
available. Defaults to False.**overridesobject= {}Keyword overrides on top of the canonical ConvNeXt-L config.
Returns
ConvNeXtA ConvNeXt backbone returning a flat
feature.
Notes
ConvNeXt-L reaches 84.3% top-1 on ImageNet-1k (224x224) and 86.6% top-1 at 224x224 — and 87.5% at 384x384 — after ImageNet-22k pretraining (Liu et al., 2022, Table 1 lower block; per-resolution detail in Table 12).
Examples
>>> import lucid
>>> from lucid.models.vision.convnext import convnext_large
>>> model = convnext_large()
>>> x = lucid.randn(1, 3, 224, 224)
>>> model.forward_features(x).shape
(1, 1536)