Platypus:Quick,Cheap,and Powerful Refinement of LLMs
Introduction
Now that large models have achieved very good results, how to distill the capabilities of large models into small models and maintain the same level of capabilities in small models is still a problem.
However, the current instruction tuning method costs a lot of time and money when the model is large enough. Qlora is already available to alleviate this problem, but the author provides another solution.