ChatGLM-6B Fine-Tuning for Cultural and Creative Products Advertising Words
Xiyuan Zhang, Xinyue Zhang, Ying Yu
Abstract
ChatGLM-6B is a bilingual conversational language model based on the General Language Model (GLM) architecture. It efficiently blends natural language responses, drawn from Chinese Q&A feedback, with supervised learning and feedback reinforcement strategies guided by humans. These efforts produce responses aligned with human tastes and expectations. As we find ourselves in an era where traditional culture is being disseminated through the dynamic growth of the creative and cultural industries, this study zeroes in on ChatGLM-6B's feature allowing local deployment on standard consumer graphics cards. Utilising a dataset generated for this purpose, the ChatGLM-6B model undergoes fine-tuning via the P-Tuning v2 method at an int-4 quantisation level. Consequently, this endows the model with the capacity to generate lively and captivating advertising phrases for cultural and creative goods.