News for hackers. Stuff that matters.
Re:Direct initialization of transformers using larger pretra... by turtles3 on Friday December 22, 2023 @03:36PM (#38738381)
I believe what you're looking for is Hypernetworks

https://arxiv.org/abs/1609.09106

Re:Direct initialization of transformers using larger pretra... by jacobsimon on Friday December 22, 2023 @05:04PM (#38739210)
Very relevant thanks!