Mominnadeempak·Nov 13Teaching a 135M Model to Summarize: RLHF from ScratchBy Momin Nadeem and Ayesha Tariq
Mominnadeempak·Oct 5We Built a Transformer From Scratch to Turn English Questions Into SQLAn encoder-decoder trained from random initialisation on WikiSQL: 7.6 million parameters, 21 minutes on a free GPU, and about 59% execution…