Create a 1D, 2D, or 3D convolution module.

Read and unpack the next bytes from a binary file. :param fid: :param num_bytes: Sum of combination of {2, 4, 8}, e.g. 2, 6, 16, 30, etc.

@param x: B,N,4,H,W @param t_embed: B,t_dim @param v_embed: B,N,v_dim @param target_poses: N,3,4

@param x_input: B,4,H,W @param spatial_volume: B,C,V,V,V @param clip_embed: B,1,768 @param t_embed:

Make a standard normalization layer. :param channels: number of input channels. :return: an nn.Module for normalization.

Loads the schedulers state. Args: state_dict (dict): scheduler state. Should be an object returned from a call to

Evaluate a function without caching intermediate activations, allowing for reduced memory at the expense of extra compute in the backward pas

Create a linear module.

Create sinusoidal timestep embeddings. :param timesteps: a 1-D Tensor of N indices, one per batch element. These may be

F.binary_cross_entropy is not numerically stable in mixed-precision training.

@param x: B,N,4,H,W @param t_embed: B,t_dim @param v_embed: B,N,v_dim @param target_poses: N,3,4

@param x_input: B,4,H,W @param spatial_volume: B,C,V,V,V @param clip_embed: B,1,768 @param t_embed:

Zero out the parameters of a module and return it.

Zero out the parameters of a module and return it.

@param x_ratio: float @param y_ratio: float @param Ks: b,3,3 @param poses: b,3,4 @return: