search

stevenygd / PointFlow

PointFlow : 3D Point Cloud Generation with Continuous Normalizing Flows
https://www.guandaoyang.com/PointFlow/
MIT License
720 stars 101 forks source link

How does the prior loss behave with and without batchnorm ? #6

Closed dawood95 closed 4 years ago

dawood95 commented 4 years ago

I'm trying to train a similar model. My prior loss doesn't decrease and stays around 60k (for a vector of dimension 256).

stevenygd commented 4 years ago

Are you referring to the batch norm in point CNF or latent CNF?

dawood95 commented 4 years ago

Both. If you could share loss curves, that would be very helpful.

stevenygd commented 4 years ago

The loss doesn't decrease without the batch norm from my end.