Dhananjay Ram

Training large language models more efficiently

Large language models (LLMs) go through several stages of training on mixed datasets with different distributions, stages that include pretraining, instruction tuning, and reinforcement learning from...