SoggyKiwi/DeIT-Dreamer

bro really had them the wrong way around

wording change

set device correctly

i fogor 💀

remove tv loss, using BCE loss now

fix various total variation bugs

add tv_loss to method signature

add total variation loss + tuning changes

convert n_targets to int, wording change

use n random target classes to maximise activation for

use CLS token, tuning default params

nvm dont be live

set outputs correctly

prevent running if image not uploaded

gradio number doesn't have a type apparently

oops deadnamed pillow

skibidi

lmao lets goooo

chatgpt winter eepy disease is real

fix chatgpt mistake lmao

init

initial commit