Apply the block to a Tensor, conditioned on a timestep embedding. :param x: an [N x C x ...] Tensor of features. :param emb:

Apply QKV attention. :param qkv: an [N x (H * 3 * C) x T] tensor of Qs, Ks, and Vs. :return: an [N x (H * C) x T] tensor aft

Apply QKV attention. :param qkv: an [N x (3 * H * C) x T] tensor of Qs, Ks, and Vs. :return: an [N x (H * C) x T] tensor aft

Apply the model to an input batch. :param x: an [N x C x ...] Tensor of inputs. :param timesteps: a 1-D batch of timesteps.

Apply the model to an input batch. :param x: an [N x C x ...] Tensor of inputs. :param timesteps: a 1-D batch of timesteps.

Follow skimage.util.random_noise.

Convert flow into middlebury color code image.

IFFT with shifting DC to the corner of the image prior to transform

IFFT for multi-coil

Apply centered 2-dimensional Inverse Fast Fourier Transform. Args: data: Complex valued input data containing at least 3 dimensions:

Move gray maps to GPU, no normalization is done

Take the mean over all non-batch dimensions.

The function used for sampling from noise.

Transfer flow map to image. Part of code forked from flownet.

Convert flow into middlebury color code image.

Get the distribution q(x_t | x_0). :param x_start: the [N x C x ...] tensor of noiseless inputs. :param t: the number of dif

Compute the mean and variance of the diffusion posterior: q(x_{t-1} | x_t, x_0)

Scale the parameters of a module and return it.

Generate spatial discounting mask constant. Spatial discounting mask is first introduced in publication: Generative Image Inpainting with

Turn a tensor image with shape CxHxW to a numpy array image with shape HxWxC :param tensor_img: :return: a numpy array image with shape H

Update target parameters to be closer to those of source parameters using an exponential moving average. :param target_params: the targe