One DDIM Step
~12 mincode completion
Implement ddim_step(x_t, eps, alpha_bar_t, alpha_bar_prev) returning xprev, same shape as xt.
Where this shows up
~12 min
••••••••••••••••
8 employers weight this skill
4 autonomy companies, 1 enterprise vendor, 1 defense company, 1 health and bio company, 1 AI product company. Top match scores 87.
Python
import numpy as np
def ddim_step(x_t, eps, alpha_bar_t, alpha_bar_prev):
"""
One deterministic DDIM update (eta = 0).
Args:
x_t: noisy sample at step t, any shape
eps: predicted noise, same shape
alpha_bar_t: cumulative alpha at step t, in (0, 1]
alpha_bar_prev: cumulative alpha at the target step, in (0, 1]
Returns:
array, same shape as x_t
"""
# YOUR CODE HERE
pass