book

Designing Large Language Model Applications

by Suhas Pai

March 2025

Intermediate to advanced

366 pages

9h 31m

English

O'Reilly Media, Inc.

Read now

Unlock full access

Includes

Includes Quizzes

Who This Book Is ForHow This Book Is StructuredWhat This Book Is Not AboutHow to Read the BookConventions Used in This BookUsing Code ExamplesO’Reilly Online LearningHow to Contact UsAcknowledgments
Defining LLMsA Brief History of LLMsEarly YearsThe Modern LLM EraThe Impact of LLMsLLM Usage in the EnterprisePromptingZero-Shot PromptingFew-Shot PromptingChain-of-Thought PromptingPrompt ChainingAdversarial PromptingAccessing LLMs Through an APIStrengths and Limitations of LLMsBuilding Your First Chatbot PrototypeFrom Prototype to ProductionSummary
Ingredients of an LLMPre-Training Data RequirementsPopular Pre-Training DatasetsSynthetic Pre-Training DataTraining Data PreprocessingData Filtering and CleaningSelecting Quality DocumentsDeduplicationRemoving Personally Identifiable InformationTraining Set DecontaminationData MixturesEffect of Pre-Training Data on Downstream TasksBias and Fairness Issues in Pre-Training DatasetsSummary
VocabularyTokenizersTokenization PipelineNormalizationPre-TokenizationTokenizationByte Pair EncodingWordPieceSpecial TokensSummary
PreliminariesRepresenting MeaningThe Transformer ArchitectureSelf-AttentionPositional EncodingFeedforward NetworksLayer NormalizationLoss FunctionsIntrinsic Model EvaluationTransformer BackbonesEncoder-Only ArchitecturesEncoder-Decoder ArchitecturesDecoder-Only ArchitecturesMixture of ExpertsLearning ObjectivesFull Language ModelingPrefix Language ModelingMasked Language ModelingWhich Learning Objectives Are Better?Pre-Training ModelsSummary
Navigating the LLM LandscapeWho Are the LLM providers?Model FlavorsOpen Source LLMsHow to Choose an LLM for Your TaskOpen Source Versus Proprietary LLMsLLM EvaluationLoading LLMsHugging Face AccelerateOllamaLLM Inference APIsDecoding StrategiesGreedy DecodingBeam SearchTop-k SamplingTop-p SamplingRunning Inference on LLMsStructured OutputsModel Debugging and InterpretabilitySummary
The Need for Fine-TuningFine-Tuning: A Full ExampleLearning Algorithms ParametersMemory Optimization ParametersRegularization ParametersBatch SizeParameter-Efficient Fine-TuningWorking with Reduced PrecisionPutting It All TogetherFine-Tuning DatasetsUtilizing Publicly Available Instruction-Tuning DatasetsLLM-Generated Instruction-Tuning DatasetsSummary
Continual Pre-TrainingReplay (Memory)Parameter ExpansionParameter-Efficient Fine-TuningAdding New ParametersSubset MethodsCombining Multiple ModelsModel EnsemblingModel FusionAdapter MergingSummary

Defining Alignment TrainingReinforcement LearningTypes of Human FeedbackRLHF ExampleHallucinationsMitigating HallucinationsSelf-ConsistencyChain-of-ActionsRecitationSampling Methods for Addressing HallucinationDecoding by Contrasting LayersIn-Context HallucinationsHallucinations Due to Irrelevant InformationReasoningDeductive ReasoningInductive ReasoningAbductive ReasoningCommon Sense ReasoningInducing Reasoning in LLMsVerifiers for Improving ReasoningInference-Time ComputationFine-Tuning for ReasoningSummary
LLM Inference ChallengesInference Optimization TechniquesTechniques for Reducing ComputeK-V CachingEarly ExitKnowledge DistillationTechniques for Accelerating DecodingSpeculative DecodingParallel DecodingTechniques for Reducing Storage NeedsSymmetric QuantizationAsymmetric QuantizationSummary
LLM Interaction ParadigmsPassive ApproachThe Explicit ApproachThe Autonomous ApproachDefining AgentsAgentic WorkflowComponents of an Agentic SystemModelsToolsData StoresAgent Loop PromptGuardrails and VerifiersAgent Orchestration SoftwareSummary
Introduction to EmbeddingsSemantic SearchSimilarity MeasuresFine-Tuning Embedding ModelsBase ModelsTraining DatasetLoss FunctionsInstruction EmbeddingsOptimizing Embedding SizeMatryoshka EmbeddingsBinary and Integer EmbeddingsProduct QuantizationChunkingSliding Window ChunkingMetadata-Aware ChunkingLayout-Aware ChunkingSemantic ChunkingLate ChunkingVector DatabasesInterpreting EmbeddingsSummary
The Need for RAGTypical RAG ScenariosDeciding When to RetrieveThe RAG PipelineRewriteRetrieveRerankRefineInsertGenerateRAG for Memory ManagementRAG for Selecting In-Context Training ExamplesRAG for Model TrainingLimitations of RAGRAG Versus Long ContextRAG Versus Fine-TuningSummary
Multi-LLM ArchitecturesLLM CascadesRoutersTask-Specialized LLMsProgramming ParadigmsDSPyLMQLSummary

Content preview from Designing Large Language Model Applications

Chapter 7. Advanced Fine-Tuning Techniques

In the previous chapter, we presented the canonical way to fine-tune a typical LLM. In the real world, there are a wide variety of motivations for updating an LLM, and similarly there are multiple ways to update it. In this chapter, we will describe several advanced fine-tuning techniques and highlight the scenarios in which each technique would be suitable.

Why would you want to update the parameters of an LLM? We touched upon this in previous chapters but let’s go through it in more detail now:

Domain adaptation: The data that we work with belongs to a specialized domain that the LLM might not have been familiarized with during pre-training. In this case, we would like to update the model by training it on domain-specific data.
Task adaptation: We care about LLM performance on specific downstream tasks. To improve the LLM’s performance on these tasks, we can train it on task-specific data. This can be supervised or unsupervised.
Knowledge updating: We would like to keep the LLM’s knowledge up-to-date by continually training it on new data.
Controllability/steerability: We would like to control the behavior of the LLM, including making it more likely to follow user requests written in natural language, reject certain types of requests, and so on. Techniques to achieve this are collectively called alignment training. We will defer discussion of alignment training to Chapter 8.