Repository for the course "Project Machine Learning" during WiSe 24/25 at TU Berlin consisting of a replication of the paper "Neural Discrete Representation Learning" (van den Oord et al., 2018).
uniform model interface: which involves the a loss and backward definition
tested it shortly. The functions work but as I stated in the issue I cannot load the imagenet images, though. dunno. Autoencoder seems to work fine (dunno if it is really learning but it works), VAE only works in the first iteration. :) Then I only get nan values. I dont know why maybe you see it faster.
I did this pull request anyway because I think it's a good project structure and the problems will be fixed soon.
I did this pull request anyway because I think it's a good project structure and the problems will be fixed soon.