This matches the implementation in Denoising Diffusion Probabilistic Models: From Fairseq. Build sinusoidal embeddings. This matches

r""" Sample images in a grid Args: fname: the file name nrow: the number of rows of the grid ncol: the number of columns o

r""" Sample from the reverse model p(x0|x1)...p(xN-1|xN)p(xN) proposed in DDPM, DDIM and Analytic-DPM

r""" for n>=1, betas[n] is the variance of q(xn|xn-1) for n=0, betas[0]=0

Create a 1D, 2D, or 3D average pooling module.

Create a beta schedule that discretizes the given alpha_t_bar function, which defines the cumulative product of (1-beta) over time from t = [

r""" -inputs * log (sigmoid(logits)) - (1 - inputs) * log (1 - sigmoid(logits)) element wise with automatically expand dimensions

Calculation of the statistics used by the FID. Params: -- files : List of image files paths -- model : Instance of inception m

Numpy implementation of the Frechet Distance. The Frechet distance between two multivariate Gaussians X_1 ~ N(mu_1, C_1) and X_2 ~ N(mu_2, C_2

tensordot(x, y, 1).

Fused `tf.nn.conv2d()` followed by `downsample_2d()`. Padding is performed only once at the beginning, not between the operations. The fused

r""" Create an instance of the optimizer described in the profile Args: config: a config describing the optimizer optimizer: the o

r""" Create an instance of the optimizer described in the profile Args: config: a config describing the optimizer Example: { "

r""" Args: it: the iteration when the evaluated models is saved

r""" Evaluate during training Args: it: the iteration of training

Build pretrained Inception model for FID computation The Inception model for FID computation uses a different set of weights and has a slight

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.

Calculates the activations of the pool_3 layer for all images. Params: -- files : List of image files paths -- model : Instan

Apply the model and return all of the intermediate tensors. :param x: an [N x C x ...] Tensor of inputs. :param timesteps: a