1 paper · 1 filter
Zhenting Qi, Fan Nie, Alexandre Alahi +6
Modern language model (LM) training has been divided into multiple stages, making it difficult for downstream developers to evaluate the impact of design choices made at each stage…