Causal 3d convolusion.

Causal 3d convolusion.

Causal self-attention with a single head.

Causal self-attention with a single head.

DPT Head for dense prediction tasks. This implementation follows the architecture described in "Vision Transformers for Dense Prediction"

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).

r""" Pipeline for text-to-video generation using Wan. This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for

r""" Output class for CogVideo pipelines. Args: video (`torch.Tensor`, `np.ndarray`, or List[List[PIL.Image.Image]]): Lis

Residual convolution module.

The Aggregator applies alternating-attention over input frames, as described in VGGT: Visual Geometry Grounded Transformer. Args:

A dataset class that provides balanced sampling across multiple datasets. This class creates multiple SGDatasetWithResizeCrop instances

CameraHead predicts camera parameters from token representations using iterative refinement. It applies a series of transformer blocks (the

A dataset class for text-to-video generation that resizes inputs to fixed dimensions. This class preprocesses videos by resizing them to spe

Feature fusion block.

2D image to patch embedding: (B,C,H,W) -> (B,N,D) Args: img_size: Image size. patch_size: Patch token size. in_chans

Generates and caches 2D spatial positions for patches in a grid. This class efficiently manages the generation of spatial coordinates for patches

2D Rotary Position Embedding implementation. This module applies rotary position embeddings to input tokens based on their 2D spatial positio

Timestep Embedding Aware Cache, a training-free caching approach that estimates and leverages the fluctuating differences among model outputs

XLMRobertaModel with no pooler and no LM head.

Base dataset class for geometry-aware video training. The dataset reads three metadata files (camera paths, prompt-file paths, and extra

`FlowDPMSolverMultistepScheduler` is a fast dedicated high-order solver for diffusion ODEs. This model inherits from [`SchedulerMixin`] and [

`UniPCMultistepScheduler` is a training-free framework designed for the fast sampling of diffusion models. This model inherits from [`Schedu

r""" Wan diffusion backbone supporting both text-to-video and image-to-video.