plismbench.models.meta module#

Models from Meta company.

class plismbench.models.meta.Dinov2ViTGiant(device: int | list[int] | None = -1, mixed_precision: bool = False)[source]#

Bases: Extractor

ViT-giant model trained with DINOv2 with 4 registers on ImageNet (1).

Note

  1. facebookresearch/dinov2

Parameters:

  • device (int | list[int] | None = DEFAULT_DEVICE,) – Compute resources to use. If None, will use all available GPUs. If -1, extraction will run on CPU.

  • mixed_precision (bool = True) – Whether to use mixed_precision.

property transform: Compose#

Transform method to apply element wise.