2 ms·
If your definition of "competitive" is loose enough, you can write your own Markov chain in an evening. Transformer models rely on a lot of prior art that has t
by bigyabai 8mo ago
If your definition of "competitive" is loose enough, you can write your own Markov chain in an evening. Transformer models rely on a lot of prior art that has to be learned incrementally.
- jasonjmcghee 8mo agoNot that loose lol. I’m thinking it’s still llama / dense decoder only transformer.