RegMix: Data Mixture as Regression for Language Model Pre-training

Open in new window