Straight to Zero: Why Linearly Decaying the Learning Rate to Zero Works Best for LLMs

Add code
Feb 21, 2025

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: