CodeLlama's actual combat plan for local deployment

  Hello everyone, I am herosunly. Graduated from the 985 college with a master's degree, he is currently working as an algorithm researcher, and is keen on the research and application of machine learning algorithms. He won the first place in the Aliyun Tianchi competition, the second place in the CCF competition, and the third place in the HKUST Xunfei competition. Has a number of invention patents. Have your own unique insights into machine learning and deep learning. I have coached several non-computer majors to enter the algorithm industry for employment. Hope to grow and progress together with you.

  This article mainly introduces the actual combat plan of CodeLlama's local deployment, hoping to help students who study large language models.

1 Introduction

  Code LLAMA is generated by a set of pre-trained and fine-tuned codes 模型家族, including three types of models of 7B, , 13Band . 34BIn addition to (Code LLAMA), each category 基础模型also includes versions of 擅长Python编程(Code LLAMA - Python) and (Code LLAMA - INSTRUCT)经过指令微调