bigger models?

#3
by jacek2024 - opened

Any plans for something similar but 34B?

currently its good with small models because of the overtraining "14 trillion tokens" which help it. so the bigger the size the more the tokens And the harder it is to train

Technology Innovation Institute org

@jacek2024 - thanks for the interest in bigger sized models ; Stay tuned for more updates !

puneeshkhanna changed discussion status to closed

Sign up or log in to comment