built autoencoders → VAEs → GANs → WGANs → diffusion models from scratch over the past few months. no libraries, just math and code. moving on to latent diffusion
CHECK IT OUTTTTTTTTTT !!!!!!!!!!!!!!!!!!!! -> (https://t.co/70tsVdA3AG)
What I gifted to myself with my first ever salary from an internship
After months or years of feeling depressed and un-accomplishmentness, I guess things are becoming better and I wish that for everyone around myself... best of luck yall... may the power be with all of you ❤️
@rutu_3 Go see the documentary of Tan Youhui, the best example of real life recursion
A person hires a hitman to get his rival ded (1st recursion call)
The hitman hires another hitman to get his work done (2nd recursion call)
similarly this cycle goes on 4 more times -> 6 recursion calls
Obsidian's graph view has to be one of the most satisfying and useless stuff to ever exist. By useless I don't mean that it's bad, but I don't see any practical use to it but god it looks amazing after you've used it for a long time.
Idono why does my diffusion model keeps hitting this phase where it stops learning anything at all...
is it the model's size?
image's size?
not enough training time?
architectural problem?
the dataset?
well if anybody's willing to diagnose, i'll be really grateful for it man