Args: x (Tensor): The query tensor of shape [B, Mq, K] context (Optional[Tensor]): The key tensor of shape [B, Mk, K]

With CP, the function assume that the input tensor is already split. It delegates the embedding generation to generate_embeddings fun

Apply the Fourier feature transformation to the input tensor. Args: x (torch.Tensor): The input tensor. gain

Forward pass of the PatchEmbed module. Parameters: - x (torch.Tensor): The input tensor of shape (B, C, T, H, W) where

Args: x: (B, C, T, H, W) tensor of spatial-temp inputs timesteps: (B, ) tensor of timesteps crossattn_emb