[Question] Possible redundant Euler step in sampling loop ddpm/neuralsolver.py

First of all, I really enjoyed and found the paper very interesting — thanks for sharing this work!
While reading the code, I noticed what seems to be an extra Euler-like update in the sampling loop:

pred_trajectory.append(x0)
delta_lamda = lamda_next - lamda
x = (sigma_next/sigma) * x + sigma_next * (delta_lamda) * x0

This runs after the main coefficient-weighted update, so it might be redundant and could overwrite the intended solver step.

Is this behavior correct, or should the extra update be removed?
Thanks in advance! :)

Start in ddpm/neuralsolver.py and inspect the sampling loop around pred_trajectory.append(x0), delta_lamda, and the subsequent x update. Compare that update with the preceding coefficient-weighted solver step and the intended sampling behavior described in the paper. Done means determining whether the extra Euler-like update is intentional or redundant and documenting the conclusion or removing it if confirmed.