I am trying to implement the disentangled VAE model according to this link. I want to understand the architecture of this model in order to customize it later. As infrastructure, I have a Linux kernel with 4 cores, 8 GB as the memory with only CPU support. but still, the model is taking hours to run.

Can anyone try to run this model and give me feedback? Is there any other simpler implementation of the disentangled VAE in python because I couldn't find any.


First of all , this is a deepmind's paper , so the computational level they operate on is much higher. Can you share your code files , so that I can see how you interpreted miyosuda's code. The disentangled representation has to be scaled down so that the model trains , and gets useful results in atleast a few hours.

4GB is too less RAM for training any kind of model. The best temporary solution I can suggest is to perform the python script of distangled VAE on Colab Colab Notebooks

The more permanent solution that can making your model training much easier given your system constraints , is to use GCP or Azure.

How do you plan to customize the architecture once you finish implementing distangled VAE.

  • $\begingroup$ Hi abunickabhi, sorry for the late reply I gave up on that model but still tried something simpler. Here is the link of my code github.com/emnajaoua/beta_variational_autoencoders , My dataset consists of many varieties of grid layouts. Those grid layouts are represented by 6 by 6 matrices which have different values. My objective to reconstruct those grids or vectors but the results that i am getting are not that accurate. $\endgroup$ – Emna Jaoua Aug 8 '18 at 12:14
  • $\begingroup$ github.com/emnajaoua/beta_variational_autoencoders/blob/master/… the github code link $\endgroup$ – Emna Jaoua Aug 10 '18 at 7:18
  • $\begingroup$ Got it! How far do you want to implement the distangled VAE. $\endgroup$ – abunickabhi Aug 10 '18 at 7:40
  • $\begingroup$ I am sorry but what do you mean by how far I want to implement ? $\endgroup$ – Emna Jaoua Aug 10 '18 at 9:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.