Compute-efficient pretraining and scaling to trillion-parameter models

118 points | by ronfriedhaber 6 days ago

67 comments