Informal chat where we read through a few papers and look at a recent project.
No central notebook for this lesson, but some resources we'll be talking about:
A brief shoutout to https://multimodal.art/ as a great way to keep up with things
CogView2: Faster and Better Text-to-Image Generation via Hierarchical Transformers (https://arxiv.org/abs/2204.14217)
Denoising Diffusion GAN (https://github.com/NVlabs/denoising-d...)
My project, CLOOB Conditioned Latent Denoising Diffusion GANs (https://johnowhitaker.github.io/cclddg/) and specifically the demo notebook (https://colab.research.google.com/dri...)
I forgot to mention a few things:
1) If you're curious how I organised the code into that library with nice docs and such: check out NBDev.
2) The demo grids shown run from no conditioning (left) to fairly extreme conditioning (right) using classifier free guidance.
-- Watch live at / johnowhitaker
コメント