> predict the next word in 40GB of Internet text This could do wonders for lip r... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		mycall on Jan 23, 2020 \| parent \| context \| favorite \| on: Talking to myself: how I trained GPT2-1.5b for rub... > predict the next word in 40GB of Internet text This could do wonders for lip reading correction.

menmob on Jan 23, 2020 [–]

OpenAI trained the initial 1.5B model on ~160G of text.. so I’m sure it’s already going to give amazing results.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact