semaj87 / summarise-dialogue-flan-t5

Performing the task of dialogue summarisation using Generative AI, whilst comparing the effects of zero shot, one shot and few shot prompt engineering. These steps are used to enhance the completion of Large Language Models (LLMs))

Summarising dialogue using FLAN-T5

This is a simple project that was undertaken to perform the task of dialogue summary using Generative AI. Through the use of different techniques to inference process, the exploration of how different prompts can affect the completion of the model were performed. Prompt engineering was carried out, by comparing zero shot, one shot and few shot inferences, with the intention to see how to best enhance the generative output of the LLM. This is part of Adapt & Align model, which is only one of ghe steps involved during the generative AI project lifecycle. Additional changes were also made to the configuration parameters, to best understand the influence they make, in terms of the model's ability to make a final decision about next word generation. Configuration parameters that were changed: do_sample, temperature, top_k & top_p

Generative AI project lifecycle

Model and dataset

Model: FLAN-T5
Dataset: DialogueSum

Prerequisites

• Python3
• Jupyter
• Hugging Face
• PyTorch

License

Distributed under the MIT License. See LICENSE for more information.