r""" Output class for the custom Stable Video Diffusion pipeline. Args: frames (`[List[List[PIL.Image.Image]]`, `np.ndarray`, `torch.T

A reusable pipeline for single-step video diffusion inference. This class encapsulates the models and the core inference logic, separati

The output of [`UNetSpatioTemporalConditionModel`]. Args: sample (`torch.Tensor` of shape `(batch_size, num_frames, num_channels, he

Dataset for loading video data stored as image sequences on S3. Includes functionality to generate a binary mask from the alpha channel a

Simple hook to capture intermediate features.

A simple 3-layer MLP for feature projection.

A PyTorch Dataset that generates synthetic video data on-the-fly. It merges the logic of a generator script with a dataloader, avoiding t

r""" A custom pipeline based on Stable Video Diffusion that accepts an additional mask for conditioning. This pipeline is designed to work wit

r""" A custom pipeline based on Stable Video Diffusion that accepts an additional mask for conditioning. This pipeline is designed to work wit

r""" A conditional Spatio-Temporal UNet model that takes a noisy video frames, conditional state, and a timestep and returns a sample shaped o