A question for the physical residual

[NightGlow0826]

Hi,
Maybe the physical residual (computed by the governing vorticity equation) should be calculated by the denoised one rather than the noised one.
x = x0 * a.sqrt() + e * (1.0 - a).sqrt() flag = np.random.uniform(0, 1) if flag < p: output = model(x, t.float()) else: dx = voriticity_residual((x*x_scale + x_offset)) / x_scale
Adding noise would bring extra residual but that's irrelevant to the model.

Thank you.