def q_sample()

in agents/diffusion.py [0:0]

• 8 lines of code
• 2 McCabe index (conditional complexity)

    def q_sample(self, x_start, t, noise=None):
        if noise is None:
            noise = torch.randn_like(x_start)

        sample = (
                extract(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start +
                extract(self.sqrt_one_minus_alphas_cumprod, t, x_start.shape) * noise
        )

        return sample