Halcyon trains text-to-image models without language pretraining
The authors report Halcyon-1.6B beating a larger JiT baseline on gFD with 275 times less pretraining compute.
Stories per day · 30 days
1 stories, newest first · one card per story
1 stories, newest first · one card per story
The authors report Halcyon-1.6B beating a larger JiT baseline on gFD with 275 times less pretraining compute.
Today 1K readers · 1.5K views · 1.1K reads · all time 159K views