supports forward offloading

Causal 3d convolusion.

Causal 3d convolusion.

r"""A weak reference to a Storage. The cdata member is a Python number containing the integer representation of the Storage pointer.

Causal self-attention with a single head.

Simple wrapper that delays loading the model, since the model isn't needed if the training data is already cached.

A flexible attention module supporting both self-attention and cross-attention mechanisms. This module implements a multi-head attention lay

Causal self-attention with a single head.

A transformer block that combines self-attention, cross-attention and MLP layers with AdaLN modulation. Each component (self-attention, cross

The final layer of video DiT.