Hi, thanks for the shared code.
I assume gaussian_identity_variance distribution of the decoder distribution is a standard normal distribution.
However, in the function compute_log_p_log_q_log_d, the variance was taken as logvar instead, i.e., the variance is e.
May I confirm this though it's a minor issue?
Hi, thanks for the shared code.
I assume
gaussian_identity_variancedistribution of the decoder distribution is a standard normal distribution.However, in the function
compute_log_p_log_q_log_d, the variance was taken as logvar instead, i.e., the variance ise.May I confirm this though it's a minor issue?