Preface

• Who this book is for
• What this book covers
• To get the most out of this book
• Download the color images
• Conventions used
• Get in touch
• Share Your Thoughts

1. Part 1: Introduction and Data Preparation

2. Chapter 1: Introduction to LLM Design Patterns FREE CHAPTER

• Understanding LLMs
• Understanding design patterns
• Design patterns for LLM development
• Summary

3. Chapter 2: Data Cleaning for LLM Training

• Understanding the importance of clean data
• Common data quality issues in language datasets
• Text preprocessing techniques for LLMs
• Handling multilingual and code-mixed data
• Deduplication strategies for large text corpora
• Automated data cleaning pipelines
• Data validation and quality assurance
• Summary

4. Chapter 3: Data Augmentation

• Text data augmentation techniques
• Leveraging existing LLMs for data generation
• Multilingual data augmentation strategies
• Semantic preservation in text augmentation
• Balancing augmentation and data quality
• Evaluating the impact of data augmentation
• Summary

5. Chapter 4: Handling Large Datasets for LLM Training

• Challenges of large datasets
• Data sampling techniques
• Distributed data processing
• Data sharding and parallelization strategies
• Efficient data storage formats
• Streaming data processing for continuous LLM training
• Memory-efficient data loading techniques
• Summary

6. Chapter 5: Data Versioning

• Understanding the need for data versioning
• Data versioning strategies for large language datasets
• Tools for data versioning
• Integrating data versioning in training workflows
• Version control for text corpora
• Managing dataset variants and experiments
• Best practices for data versioning
• Summary

7. Chapter 6: Dataset Annotation and Labeling

• The importance of quality annotations
• Annotation strategies for different tasks
• Tools and platforms for large-scale text annotation
• Managing annotation quality
• Crowdsourcing annotations – benefits and challenges
• Semi-automated annotation techniques
• Scaling annotation processes for massive language datasets
• Annotation biases and mitigation strategies
• Summary

8. Part 2: Training and Optimization of Large Language Models

9. Chapter 7: Training Pipeline

• Components of a training pipeline
• Data input and preprocessing
• LLM architecture design considerations
• Loss functions and optimization strategies
• Logging
• Pipeline modularity and reusability
• Scaling your training pipeline for larger models
• Summary

10. Chapter 8: Hyperparameter Tuning

• Understanding hyperparameters
• Manual versus automated tuning
• Grid and random search
• Bayesian optimization
• Population-based methods
• Multi-objective hyperparameter optimization
• Hyperparameter tuning at scale – challenges and solutions
• Summary

11. Chapter 9: Regularization

• L2 regularization (Ridge regression)
• Dropout
• Layer-wise adaptive regularization
• Gradient clipping and noise injection
• Regularization in transfer learning and fine-tuning scenarios
• Emerging regularization techniques
• Summary

12. Chapter 10: Checkpointing and Recovery

• Why is checkpointing important?
• Checkpoint frequency and storage strategies
• Efficient checkpoint formats
• Recovering from failures
• Checkpointing in distributed LLM training
• Version control for LLM checkpoints
• Automated checkpointing and recovery systems
• Summary

13. Chapter 11: Fine-Tuning

• Implementing transfer learning and fine-tuning
• Strategies for freezing and unfreezing layers
• Learning rate scheduling
• Domain-specific fine-tuning techniques
• Few-shot and zero-shot fine-tuning
• Continual fine-tuning and catastrophic forgetting
• Summary

14. Chapter 12: Model Pruning

• Magnitude-based pruning
• Structured versus unstructured pruning
• Iterative pruning techniques
• Pruning during training versus post-training pruning
• Balancing pruning and model performance
• Combining pruning with other compression techniques
• Summary

15. Chapter 13: Quantization

• Understanding the basics
• Mixed-precision quantization
• Hardware-specific considerations
• Comparing quantization strategies
• Combining quantization with other optimization techniques
• Summary

16. Part 3: Evaluation and Interpretation of Large Language Models

17. Chapter 14: Evaluation Metrics

• NLU benchmarks
• Reasoning and problem-solving metrics
• Coding and programming evaluation
• Conversational ability assessment
• Commonsense and general knowledge benchmarks
• Other commonly used benchmarks
• Developing custom metrics and benchmarks
• Interpreting and comparing LLM evaluation results
• Summary

18. Chapter 15: Cross-Validation

• Pre-training and fine-tuning data splits
• Few-shot and zero-shot evaluation strategies
• Domain and task generalization
• Continual learning evaluation
• Cross-validation challenges and best practices
• Summary

19. Chapter 16: Interpretability

• Attention visualization techniques
• Probing methods
• Explaining LLM predictions with attribution methods
• Interpretability in transformer-based LLMs
• Mechanistic interpretability
• Trade-offs between interpretability and performance
• Summary

20. Chapter 17: Fairness and Bias Detection

• Types of bias
• Fairness metrics for LLM text generation and understanding
• Detecting bias
• Debiasing strategies
• Fairness-aware training
• Ethical considerations
• Summary

21. Chapter 18: Adversarial Robustness

• Types of textual adversarial attacks
• Adversarial training techniques
• Evaluating robustness
• Trade-offs in the adversarial training of LLMs
• Real-world implications
• Summary

22. Chapter 19: Reinforcement Learning from Human Feedback

• Components of RLHF systems
• Scaling RLHF
• Limitations of RLHF in language modeling
• Applications of RLHF
• Summary

23. Part 4: Advanced Prompt Engineering Techniques

24. Chapter 20: Chain-of-Thought Prompting

• Designing effective CoT prompts
• Using CoT prompting for problem solving
• Combining CoT prompting with other techniques
• Evaluating CoT prompting outputs
• Limitations of CoT prompting
• Future directions
• Summary

25. Chapter 21: Tree-of-Thoughts Prompting

• Designing ToT prompts
• Search strategies
• Pruning and evaluation
• Applying ToT to solve a multi-step problem
• Challenges in implementation
• Future directions
• Summary

26. Chapter 22: Reasoning and Acting

• Implementing ReAct in LangChain
• Building ReAct agents with LangChain’s Expression Language
• Completing tasks and solving problems
• Evaluating ReAct’s performance
• Safety, control, and ethical considerations
• Limitations and future directions
• Summary

27. Chapter 23: Reasoning WithOut Observation

• Implementing ReWOO with LangGraph
• Advantages of ReWOO
• Evaluating quality and ethical considerations
• Future directions
• Summary

28. Chapter 24: Reflection Techniques

• Designing prompts for self-reflection
• Implementing iterative refinement
• Correcting errors
• Evaluating the impact of reflection
• Challenges in implementing effective reflection
• Future directions
• Summary

29. Chapter 25: Automatic Multi-Step Reasoning and Tool Use

• Designing prompts for complex task decomposition
• Integrating external tools
• Implementing automatic tool selection and use
• Complex problem solving
• Evaluating multi-step reasoning and tool use
• Challenges and future directions
• Summary

30. Part 5: Retrieval and Knowledge Integration in Large Language Models

31. Chapter 26: Retrieval-Augmented Generation

• Building a simple RAG system for LLMs
• Embeddings and indexing for retrieval in LLM applications
• Query formulation strategies in LLM-based RAG
• Integrating retrieved information with LLM generation
• Challenges and opportunities in RAG for LLMs
• Summary

32. Chapter 27: Graph-Based RAG

• Introduction to graph-based knowledge representation for LLMs
• Designing graph RAG architectures for LLMs
• Graph embedding techniques for LLM retrieval
• Query expansion using graph structures in LLMs
• Applications and use cases of graph RAG in LLMs
• Challenges and future directions in graph-based RAG
• Summary

33. Chapter 28: Advanced RAG

• Multi-step and iterative retrieval techniques for LLMs
• Adaptive retrieval based on context and task in LLMs
• Meta-learning for improved retrieval in LLMs
• Combining RAG with other LLM prompting techniques
• Handling ambiguity and uncertainty in LLM-based RAG
• Scaling RAG to very large knowledge bases
• Future directions in RAG research for LLMs
• Summary

34. Chapter 29: Evaluating RAG Systems

• Challenges in evaluating RAG systems for LLMs
• Metrics for assessing retrieval quality in LLM-based RAG
• Considerations for retrieval metrics in RAG
• Evaluating the relevance of retrieved information for LLMs
• Measuring the impact of retrieval on LLM generation
• End-to-end evaluation of RAG systems in LLMs
• Human evaluation techniques for LLM-based RAG
• Benchmarks and datasets for RAG evaluation
• Summary

35. Chapter 30: Agentic Patterns

• Introduction to agentic AI systems based on LLMs
• Goal-setting and planning in LLM-based agents
• Implementing memory and state management for LLM agents
• Decision-making and action selection in LLM-based agents
• Learning and adaptation in agentic LLM systems
• Ethical considerations and safety in LLM-based agentic AI
• Future prospects of agentic AI using LLMs
• Summary
• Future directions in LLM patterns and their development

36. Index

• Why subscribe?

37. Other Books You May Enjoy

• Packt is searching for authors like you
• Share Your Thoughts
• Download a free PDF copy of this book