Why density of Gaussian distribution (input samples) is ignored

[dzy1997]

Hi Vishakh,
I am trying to learn normalizing flow and find your tutorial really helpful! I just have one question on your implementation of the model:

normalizing-flows/src/normalizing_flows.py

Line 65 in ec2c666

sum_log_det = 0

Here you initialize sum_log_det to 0, ignoring the prob density of input Gaussian distribution. Would it make more sense if you compute and add that, since the model is transforming prob density? At the current state, how would it differ from a model trained from uniform distribution samples as input? Thank you!