由 K Clark 著作 · 2019 · 被引用 1503 次 — to downstream NLP tasks, they generally require large amounts of compute to be ... We call our approach ELECTRA1 for “Efficiently Learning an Encoder that ...
確定! 回上一頁