lucidrains / imagen-pytorch

Implementation of Imagen, Google's Text-to-Image Neural Network, in Pytorch
MIT License
8.09k stars 768 forks source link

Improvements for beginners #307

Open TheFusion21 opened 1 year ago

TheFusion21 commented 1 year ago
lucidrains commented 1 year ago

yea, all good points

you want to take on any of this?

lucidrains commented 1 year ago

ok, let's make this a community effort. I'll get around to this after I wrap up all the remaining video todos

TheFusion21 commented 1 year ago

Will do preencoding as I was planing on doing it just haven't had time. Docs for training can do. maybe add a wiki? Requirements should be easy software and hardware

Slurm don't know enough Webdataset maybe have to take a look into it

lucidrains commented 1 year ago

@TheFusion21 yea, i'm stretched for time too

i really want to get started on some exciting new protein diffusion work, which will take up a month at least

lucidrains commented 1 year ago

@TheFusion21 no hurries, we can just keep this open and knock out items on the list when time allows

jacobwjs commented 1 year ago

Will do preencoding as I was planing on doing it just haven't had time. Docs for training can do. maybe add a wiki? Requirements should be easy software and hardware

Slurm don't know enough Webdataset maybe have to take a look into it

@TheFusion21 Did you ever get around to work on precomputing text embeddings?