Skip to content

tf2.0 Implementation of Masked Autoregressive Flows for Density Estimation

Notifications You must be signed in to change notification settings

johnpjust/MAF_GQ_images_tf20

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

MAF_GQ_images_tf20

tf2.0 Implementation of Masked Autoregressive Flows for Density Estimation

There's probably some room for optimization here, such as using tf.function. Runs a bit slower than graph mode and the LL is less. In graph mode the LL obtained was > BNAF slightly, now it is slightly less than it. From previous work though higher LL doesn't ensure there is a tangible/noticeable difference for most applications.

Note when saving the model, one needs to save the trainable parameters, non-trainable parameters (running mean/stdev for batch norm when used), and random permutations for each layer.

About

tf2.0 Implementation of Masked Autoregressive Flows for Density Estimation

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages