Cyrus Majebi's comment thread on TwoCents

http://twocents.space/comment_thread-cyrus-majebi-919/

Cyrus Majebi

Co-founder @ TwoCents, plugME • Lagos • 2 years, 7 months ago

Vectorization does indeed make matrix operations much faster, especially when training huge LLMs with billions of parameters. Infact, GPUs are designed to specifically take advantage of vectorization at scale. Interesting that I recognize that code snippet - looks like it's from Andrew Ng's Deep Learning Specialization on Coursera.

1 Reply

Albert David Bangura Graduate Teaching and Research Assistant @ Bahcesehir Cyp... • Nicosia, Cyprus • 2 years, 7 months ago

I learnt vectorization from his Machine Learning Specialisation course. Yeah, the code snippet is similar but did some changes to it. Thank you for adding your insight!

edited