bfshi / TOAST

Official code for "TOAST: Transfer Learning via Attention Steering"

bfshi/TOAST Issues

Motivation for top-down input added to value matrix
Closed 7 months ago2
for visualization problems
Updated 8 months ago
License
Updated a year ago2
LLamaTopDownForCausalLM.from_pretrained Error in with Pre-tuning on OpenWebText
Updated a year ago2
How to pre-tune in medical images
Closed a year ago1
Parameter initialization problem.
Updated a year ago2
Question about parameters of TOAST-Lite?
Updated a year ago2
Confusion about pretrained-model
Closed a year ago3
code
Updated a year ago4
When using 8bit for loading, an error is reported
Closed a year ago3