From the course: Getting Hands-On with GPT Models: Tips and Tricks

Unlock this course with a free trial

Join today to access over 24,700 courses taught by industry experts.

Using completion sequences and few shot

Using completion sequences and few shot

From the course: Getting Hands-On with GPT Models: Tips and Tricks

Start my 1-month free trial Buy for my team

Using completion sequences and few shot

“

- [Instructor] We usually understand a concept better when someone gives us an example. LLMs work in a very similar way. Providing examples is a technique called few-shot learning. Let me show you how this can be helpful using ChatGPT. Let's go ahead and go into the Exercise Files. I'm here in 02_03, with prompts.txt open. Let's copy the first prompt, and we'll paste it in, and hit Enter. So, our task is to classify these reviews. Let's see what ChatGPT outputted. We have Positive, Negative, Positive, Negative, Neutral, Negative, Strongly Negative, Strongly Negative, and Unclassifiable. Now, this is good, but not exactly what I wanted. I want a more specific format, so let's go back and copy the second prompt. Let's paste it in. Now, in this prompt, I provide some categories that we want to classify the reviews into. I have Promoter, Neutral, Churn risk, Churned, Detractor, and Random. Let's hit Enter, and this is better. I get the responses that I want, but I still have this preamble…

Contents