Training & Pre-training
Months of thousands of GPUs predicting the next token, trillions of times.
Stai visualizzando una scheda concettuale: cosa significa questo termine in linguaggio chiaro, e i record che mette in relazione.
Pre-training runs a model over trillions of tokens, nudging weights after every prediction error, in synchronized runs across tens of thousands of accelerators — the workload the biggest clusters on this site were built for. A frontier run costs hundreds of millions of dollars in compute, fails in exotic ways (silent data corruption, stragglers, loss spikes), and is why checkpointing, interconnects and reliability engineering are front-page infrastructure topics.
Come si collega al resto dello stack
I collegamenti di questa scheda a colpo d'occhio — clicca un gruppo per raggiungerlo.
Correlati
Fonti
Llama 3 herd paper (training at scale) ↗
Record revisionato al 2026-01-15 — vedi Metodologia.