Method p_mean_variance

schedule/diffusionSample.py:53–62

(self, x_t, cond_, t)

        return extract(self.coeff1, t, x_t.shape) * x_t - extract(self.coeff2, t, x_t.shape) * eps

    def p_mean_variance(self, x_t, cond_, t):
        # below: only log_variance is used in the KL computations
        var = torch.cat([self.posterior_var[1:2], self.betas[1:]])
        #var = self.betas
        var = extract(var, t, x_t.shape)
        eps = self.model(torch.cat((x_t, cond_), dim=1), t)
        # nonEps = self.model(x_t, t, torch.zeros_like(labels).to(labels.device))
        # eps = (1. + self.w) * eps - self.w * nonEps
        xt_prev_mean = self.predict_xt_prev_mean_from_eps(x_t, t, eps=eps)
        return xt_prev_mean, var

    def noisy_image(self, t, y):
        """ Compute y_noisy according to (6) p15 of [2]"""