Implements a causal 3D convolution layer where each position only depends on previous timesteps and current spatial locations. This maintai

r""" Output of decoding method. Args: sample (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`):

r""" A Resnet block.

Base class for model's outputs that also contains a pooling of the last hidden states. Args: hidden_state (`torch.FloatTensor` of sh

MLP as used in Vision Transformer, MLP-Mixer and related networks

Modulation layer for DiT.

Embeds scalar timesteps into vector representations.

A 3D downsampling layer with an optional convolution.

Euler scheduler. This model inherits from [`SchedulerMixin`] and [`ConfigMixin`]. Check the superclass documentation for the generic met

Projects text embeddings. Also handles dropout for classifier-free guidance. Adapted from https://github.com/PixArt-alpha/PixArt-alpha/blob/

A 3D UNet mid-block [`UNetMidBlockCausal3D`] with multiple residual blocks and optional attention blocks.

A 3D upsampling layer with an optional convolution.

ODE solver class

r""" The `DecoderCausal3D` layer of a variational autoencoder that decodes its latent representation into an output sample.

r""" The `EncoderCausal3D` layer of a variational autoencoder that encodes its input into a latent representation.

The final layer of DiT.

Output class for the scheduler's `step` function output. Args: prev_sample (`torch.FloatTensor` of shape `(batch_size, num_channels,

HunyuanVideo Transformer backbone Inherited from ModelMixin and ConfigMixin for compatibility with diffusers' sampler StableDiffusionPipelin

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

copied from https://github.com/black-forest-labs/flux/blob/main/src/flux/modules/layers.py

A multimodal dit block with seperate modulation for text and image/video, see more details (SD3): https://arxiv.org/abs/2403.03206

A DiT block with parallel linear layers as described in https://arxiv.org/abs/2302.05442 and adapted modulation interface. Also refer to

2D Image to Patch Embedding Image to Patch Embedding using Conv2d A convolution based approach to patchifying a 2D image w/ embedding projec

A single token refiner block for llm text embedding refine.

SDE solver class

r""" A VAE model with KL loss for encoding images/videos into latents and decoding latent representations into images/videos. This model i

Linear Coupling Plan

Which type of output the model predicts.

Which type of path to use.

Sampler class for the transport model

class for VP path flow matching

Which type of weighting to use.