Extract values from a 1-D numpy array for a batch of indices. :param arr: the 1-D numpy array. :param timesteps: a tensor of indices into

Compute lambda_t = log(alpha_t) - log(sigma_t) of a given continuous-time label t in [0, T].

Compute sigma_t of a given continuous-time label t in [0, T].

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

Create a linear module.

Compute log(alpha_t) of a given continuous-time label t in [0, T].

Convert the model to the noise prediction model or the data prediction model.

PyTorch version of cv2.filter2D Args: img (Tensor): (b, c, h, w) kernel (Tensor): (b, k, k)

2D sinc filter, ref: https://dsp.stackexchange.com/questions/58301/2-d-circularly-symmetric-low-pass-filter Args: cutoff (float): cutoff

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

grid_size: int of the grid height and width return: pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/

Center cropping implementation from ADM. https://github.com/openai/guided-diffusion/blob/8fb3ad9197f16bbc40620447b2742e13458d2831/guided_diff

Expand the tensor `v` to the dim `dims`. Args: `v`: a PyTorch tensor with shape [N]. `dim`: a `int`. Returns: a

Randomly generate mixed kernels. Args: kernel_list (tuple): a list name of kernel types, support ['iso', 'aniso', 'skew', '

Args: clip (torch.tensor): Video clip to be cropped. Size is (T, C, H, W)

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

Zero out the parameters of a module and return it.

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

Get logging logger Args: log_file: Log filename, if specified, file handler will be added to logger log_level: Loggi

Take the mean over all non-batch dimensions.