OpenAI GPT For Python Developers

Table of contents

1. Preface
   1. About the Author
2. The Story of OpenAI and ChatGPT
   1. About This Guide
   2. The Companion Toolkit
   3. Stay Connected
3. How Does GPT Work?
4. Setting Up the Development Environment
   1. Notes
   2. Installing Python, pip, and a Virtual Development Environment
   3. Obtain Your OpenAI API Keys
   4. Install the Official Python Bindings
   5. Test our API Keys
5. Understanding the Available Models and Which One to Use
   1. OpenAI Available Models and Important Considerations
   2. Which Model to Use?
   3. OpenAI Model Series
      1. GPT-4 Series
      2. GPT-3.5 Series
      3. InstructGPT-3 Series
      4. Base GPT-3 Series
      5. Codex Series
      6. Content Filter
      7. DALL-E Series
      8. TTS Series
      9. Whisper Model
      10. Embedding Model
   4. OpenAI Models and Pricing
   5. What’s Next?
6. Using GPT Chat Completions
   1. An Introductory Example
   2. System, User, and Assistant Roles
      1. The System Role
      2. The User Role
      3. The Assistant Role
   3. Few-shot Learning with Chat Completions
   4. Formatting the Output
   5. Controlling the Output’s Token Count
   6. Controlling When the Completion Output Stops
   7. Temperature and Hallucination
   8. Sampling with Top_p
   9. Temperature vs Top_p: What’s the Difference? Which One Should I Use?
   10. Streaming the API Response
   11. Controlling Repetitiveness: Frequency and Presence Penalties
   12. Frequency vs. Presence Penalty
   13. Controlling the Number of Results from the API
   14. Conclusion
7. Advanced Examples and Prompt Engineering
   1. What is Prompt Engineering?
   2. Few Shot Learning: A Key Prompt Engineering Technique
   3. Overgeneration and Selection
   4. General Knowledge Prompting (GKP): Generating a Rap Song
   5. Context Stuffing: Is Apple a Fruit or a Company?
   6. Dynamic Max Tokens
   7. Creating an Interactive CLI-Based Assistant
   8. What’s Next?
8. Embedding
   1. What is an Embedding?
   2. Use Cases: From Modern Search Engines to Self-Driving Cars
      1. Tesla: How Embeddings Are Used in Self-Driving Cars
      2. Kalendar AI: The Power of Embeddings in Sales Outreach
      3. Notion: Enhanced Search Capabilities
      4. DALL·E 2: Text-to-Image Conversion
   3. Understanding Text Embedding
   4. Embeddings for Multiple Inputs
   5. Use case: Semantic Search
   6. Cosine Similarity: A Deeper Look
   7. Semantic Search and OpenAI’s Text Embeddings
   8. Behind the Scenes: How Embeddings Work
9. Advanced Embedding Examples
   1. Predicting Your Preferred Coffee
   2. Creating a “Fuzzier” Search
   3. Predicting News Category: Zero-Shot Classification with Embeddings
   4. Evaluating the Accuracy of a Zero-Shot Classifier
   5. Precision in Zero-Shot Classifier Applications: Examples
10. Fine-Tuning and Best Practices
    1. Few-Shot Learning
    2. Enhancing Few-Shot Learning
    3. Practical Application of Fine-Tuning
    4. Fine-Tuning Best Practices
       1. Choosing the Model
       2. Validating the Dataset
       3. Token Limit
       4. Dataset Size
       5. Testing and Improving Training (Hyperparameters)
       6. Epochs
          1. Learning Rate Multiplier
       7. Batch Size
       8. Consider Estimated Costs
       9. Dataset Quality
       10. Combining Fine-Tuning with Other Techniques
       11. Experiment and Learn
       12. Use a Validation Set
       13. Test the Model
       14. Analyze the Results
11. Advanced Fine-Tuning: Mental Health Coach
    1. Dataset Used in the Example
    2. Preparing the Data
    3. Using the Model in Real-World Applications and Challenges
12. Context & Memory: Making AI More Real
    1. The Problem: No Memory
    2. No Context = Chaos of Randomness and Confusion
    3. History = Context
    4. The Problem with Carrying Over History
    5. Last In First Out (LIFO) Memory
    6. The Problem with Last In, First Out Memory
    7. Selective Context
13. Using a Vector Database with OpenAI
    1. Introduction
    2. What is a Vector Database?
    3. Example 1: Using Weaviate to Make Our Model More Context-Aware
    4. Example 2: Using Weaviate and OpenAI in Semantic Search
    5. Example 3: Using Weaviate and OpenAI for Generative Search
14. Speech Recognition and Translation Using Whisper
    1. What is Whisper?
    2. How to Get Started?
    3. Transcribe and Translate
    4. Using Whisper SDK in Python
    5. Using OpenAI Speech to Text API
       1. Transcription API
       2. Translation API
    6. Improving Whisper Transcription
       1. Cleaning the Audio
       2. Using the Prompt Parameter
       3. Post-Processing the Transcription
15. Text-to-Speech with OpenAI TTS Models
16. Autonomous AI-to-AI Discussion Using OpenAI, Weaviate, and AI Avatars
    1. Generating the Audio Files
    2. Using AI Avatar Models
    3. What’s Next?
17. Afterword