Lecture 05 - Prompt Engineering.pptx
- 1. Lecture 05 Irfan Malik Dr. Sheraz Naseer
- 2. ● Prompt Engineering ● Chat GPT PlayGround Agenda
- 4. Tokenization Definition: Tokenization is the process of breaking down text into smaller units called tokens, which can be words, phrases, or even characters. 4
- 5. Tokens in Language Models Definition: Tokens are the fundamental units of text used by language models. They can be as short as a single character or as long as a word or phrase. Importance: Language models process text at the token level, allowing them to analyze and generate text based on the context and relationships between individual tokens. 5
- 6. Prompt Limits in Language Models Definition: Prompt limits refer to the constraints imposed on the input provided to a language model. Reasons for Prompt Limits: ● Memory limitations: Language models have a finite capacity to process and store information. ● Computational constraints: Longer prompts require more processing time and computational resources. ● Response coherence: Extremely long prompts may result in irrelevant outputs. 6
- 8. What is Prompt Engineering? ● Prompt engineering involves crafting tailored instructions or queries to guide the language model towards desired outputs. ● It helps shape the behavior of the model and improves response quality. 8
- 9. Anatomy of Prompt Definition: In the context of language models, a prompt refers to the input or instruction provided to the model to generate a specific response or output 9 Parts of a Good Prompt (Not all are necessary for each and every prompt) 1. Simulate Persona 2. Task 3. Steps to complete Task 4. Context/Constraints 5. Goal 6. Format Output
- 10. Prompting Example 01 Simulate Persona: You are a customer service representative for a tech company. Task: Assist a customer in resolving a software issue. Steps to complete Task: Greet the customer and ask for details about the issue. Provide step-by-step instructions to troubleshoot the problem. If the issue persists, escalate it to the technical support team. Context/Constraints: The customer is not tech-savvy, and you must communicate in a friendly and non-technical manner. Goal: Resolve the customer's issue satisfactorily. Format Output: Provide a summary of the troubleshooting steps and any additional information the customer might need. 10
- 11. Prompting Example 02 Simulate Persona: Imagine you are a mentor guiding a group of individuals new to prompt engineering. Task: Identify and address common issues in prompt creation to improve overall effectiveness. Steps to complete Task: Analyze the prompt for Simulate Persona, Task, Steps to complete Task, Context/constraints, goals and Format Output Identify and rectify instances of ambiguity or vagueness in prompt language. Ensure that prompts provide sufficient information for the desired task without unnecessary complexity. Context/Constraints: Consider the diverse backgrounds and skill levels of those engaging with the prompts. Goal: Enhance the quality of prompts by eliminating potential sources of confusion or misinterpretation. Format Output: Provide a revised version of a poorly constructed prompt, highlighting the changes made and explaining how each modification contributes to better prompt engineering. 11
- 12. Context in Prompting Definition: Context refers to the information and preceding text that informs the language model about the desired output or influences its response. Role of Context: ● Context helps shape the model's behavior and generate more accurate responses. ● Relevance: Context ensures continuity and relevance in the generated text. ● Interpretation: The language model uses context to understand the desired meaning and tone of the prompt. 12
- 13. Tips for Effective Prompt Engineering ● Be explicit: Clearly specify the desired format or output. ● Provide context: Offer relevant information or background to guide the model. ● Control output length: Instruct the model to generate responses of a specific length. ● Restrict responses: Utilize techniques like temperature adjustment to refine output quality. ● Experiment and iterate: Refine prompts through experimentation and feedback loops. 13
- 15. Summary 1. Prompt Engineering 2. Best Practices 3. ChatGPT PlayGround 4. Sample prompt practise 15