We need to figure out how to train small transformers / or layers of them independently on small / modest hardware and stack them incrementally, to get to the same performance as training 1 big one from the beginning - THIS WOULD COMPLETELY CHANGE THE PLAYINGFIELD :)
Videh da ovih dana trenduje rasprava o napuštanju fakulteta kao obrazovnih platformi u korist kurseva i usko specijalizovanih obuka, pa bih da ukažem na psihološki termin "Survivorship Bias, ili prevedeno - sklonost ka preživljavanju kao temi koja se već dugo izučava.
Today we release LLaMA, 4 foundation models ranging from 7B to 65B parameters.
LLaMA-13B outperforms OPT and GPT-3 175B on most benchmarks. LLaMA-65B is competitive with Chinchilla 70B and PaLM 540B.
The weights for all models are open and available at https://t.co/q51f2oPZlE
1/n
Victor Hugo is most famous for writing beloved novels like Les Misérables and The Hunchback of Notre Dame.
But he was also one of the greatest painters of the 19th century.
And Hugo's art, kept secret until after his death, is unlike anything you've seen before...
Danas, u jednoj osnovnoj školi u Srbiji, učenici su mačku, sa imenom Dobrila, bacili u čopor 6 pasa uličara i navijali, dok su je psi komadali.
To nije do SNSa, nije ni do države, to je dragi roditelji, do nas. Mi smo mačku ubili.