o oi @s\ddlmZmZmZmZddlmZddlm Z ddl m Z ddl m Z Gddde ZdS) )AnyDictOptionalTuple)random_generator)IntensityAugmentationBase2D)Tensor)StableDiffusionDissolvingc seZdZdZ    ddeeefdeded ed ed d f fd d Z dde de ee fde eefde e d e f ddZ ZS)RandomDissolvinga:Perform dissolving transformation using StableDiffusion models. Based on :cite:`shi2024dissolving`, the dissolving transformation is essentially applying one-step reverse diffusion. Our implementation currently supports HuggingFace implementations of SD 1.4, 1.5 and 2.1. SD 1.X tends to remove more details than SD2.1. .. list-table:: Title :widths: 32 32 32 :header-rows: 1 * - SD 1.4 - SD 1.5 - SD 2.1 * - figure:: https://raw.githubusercontent.com/kornia/data/main/dslv-sd-1.4.png - figure:: https://raw.githubusercontent.com/kornia/data/main/dslv-sd-1.5.png - figure:: https://raw.githubusercontent.com/kornia/data/main/dslv-sd-2.1.png Args: p: probability of applying the transformation. version: the version of the stable diffusion model. step_range: the step range of the diffusion model steps. Higher the step, stronger the dissolving effects. keepdim: whether to keep the output shape the same as input (True) or broadcast it to the batch form (False). **kwargs: additional arguments for `.from_pretrained` for HF StableDiffusionPipeline. Shape: - Input: :math:`(C, H, W)` or :math:`(B, C, H, W)`. - Output: :math:`(B, C, H, W)` di2.1?F step_rangeversionpkeepdimkwargsreturnNc sDtj|d|d||_t|fi||_t|jdddf|_dS)NT)r same_on_batchrstep_range_factor)super__init__rr _dslvrgPlainUniformGenerator_param_generator)selfrrrrr __class__`/home/ubuntu/.local/lib/python3.10/site-packages/kornia/augmentation/_2d/intensity/dissolving.pyr;szRandomDissolving.__init__inputparamsflags transformcCs|||ddS)Nrr)rlongitem)rr"r#r$r%r r r!apply_transformHsz RandomDissolving.apply_transform)r r rF)N)__name__ __module__ __qualname____doc__rfloatstrboolrrrrrr( __classcell__r r rr!r s>"   r N)typingrrrrkornia.augmentationrr&kornia.augmentation._2d.intensity.baser kornia.corerkornia.filtersr r r r r r!s