A sequential module that passes timestep embeddings to the children that support it as an extra input.

A downsampling layer with an optional convolution. :param channels: Channels in the inputs and outputs. :param use_conv: A bool determin

A residual block that can optionally change the number of channels. :param channels: The number of input channels. :param emb_channels:

An upsampling layer with an optional convolution. :param channels: Channels in the inputs and outputs. :param use_conv: A bool determini

Inception-v1 I3D architecture. The model is introduced in: Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset

Latent action VAE.

Latent action VAE.

Uses xformers efficient implementation. NOTE: This is a single-head self-attention operation.

This is the base class for all autoencoders, including image autoencoders, image autoencoders with discriminators, unCLIP models, etc. Hence,

Base class for all image autoencoders that we train, like VQGAN or AutoencoderKL (we also restore them explicitly as special cases for legacy

Embeds each dimension independently and concatenates them.

Uses the OpenCLIP vision transformer encoder for images.

NOTE: Use with a base_lr of 1.0.

Supports repeated iterations, configurable via lists. NOTE: Use with a base_lr of 1.0.

Abstract LightningDataModule that represents a dataset we can train a Lightning module on.