|
What are the advantages of tying embedding and output matrices in language models? |
Answered by
Ultron09
Aug 28, 2026
Replies: 1 comment
|
Weight tying shares parameters between input lookup and output classification, halving parameter count. |
0 replies
Answer selected by
Ultron09
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Weight tying shares parameters between input lookup and output classification, halving parameter count.