diffusion.py

import math

import numpy as np
import torch


def get_num_embedding(timesteps, embedding_dim):
    """
    This matches the implementation in Denoising Diffusion Probabilistic Models:
    From Fairseq.
    Build sinusoidal embeddings.
    This matches the implementation in tensor2tensor, but differs slightly
    from the description in Section 3.5 of "Attention Is All You Need".

    for time step embedding or num node embedding
    """
    assert len(timesteps.shape) == 1

    half_dim = embedding_dim // 2
    emb = math.log(10000) / (half_dim - 1)
    emb = torch.exp(torch.arange(half_dim, dtype=torch.float32) * -emb)
    emb = emb.to(device=timesteps.device)
    emb = timesteps.float()[:, None] * emb[None, :]
    emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
    if embedding_dim % 2 == 1:  # zero pad
        emb = torch.nn.functional.pad(emb, (0, 1, 0, 0))
    return emb


def nonlinearity(x):
    # swish
    return x * torch.sigmoid(x)


def get_beta_schedule(beta_schedule, beta_start, beta_end, num_diffusion_timesteps):
    def sigmoid(x):
        return 1 / (np.exp(-x) + 1)

    if beta_schedule == "quad":
        betas = (
                np.linspace(
                    beta_start ** 0.5,
                    beta_end ** 0.5,
                    num_diffusion_timesteps,
                    dtype=np.float64,
                )
                ** 2
        )
    elif beta_schedule == "linear":
        betas = np.linspace(
            beta_start, beta_end, num_diffusion_timesteps, dtype=np.float64
        )
    elif beta_schedule == "const":
        betas = beta_end * np.ones(num_diffusion_timesteps, dtype=np.float64)
    elif beta_schedule == "jsd":  # 1/T, 1/(T-1), 1/(T-2), ..., 1
        betas = 1.0 / np.linspace(
            num_diffusion_timesteps, 1, num_diffusion_timesteps, dtype=np.float64
        )
    elif beta_schedule == "sigmoid":
        betas = np.linspace(-6, 6, num_diffusion_timesteps)
        betas = sigmoid(betas) * (beta_end - beta_start) + beta_start
    elif beta_schedule == "cosine":
        return betas_for_alpha_bar(
            num_diffusion_timesteps,
            lambda t: math.cos((t + 0.008) / 1.008 * math.pi / 2) ** 2,
        )
    else:
        raise NotImplementedError(beta_schedule)
    assert betas.shape == (num_diffusion_timesteps,)
    return betas


def betas_for_alpha_bar(num_diffusion_timesteps, alpha_bar, max_beta=0.999):
    """
    Create a beta schedule that discretizes the given alpha_t_bar function,
    which defines the cumulative product of (1-beta) over time from t = [0,1].
    :param num_diffusion_timesteps: the number of betas to produce.
    :param alpha_bar: a lambda that takes an argument t from 0 to 1 and
                      produces the cumulative product of (1-beta) up to that
                      part of the diffusion process.
    :param max_beta: the maximum beta to use; use values lower than 1 to
                     prevent singularities.
    """
    betas = []
    for i in range(num_diffusion_timesteps):
        t1 = i / num_diffusion_timesteps
        t2 = (i + 1) / num_diffusion_timesteps
        betas.append(min(1 - alpha_bar(t2) / alpha_bar(t1), max_beta))
    return np.array(betas)