LogoLogo
Continuum WebsiteContinuum ApplicationsContinuum KnowledgeAxolotl Platform
Continuum Knowledge
Continuum Knowledge
  • Continuum
  • Data
    • Datasets
      • Pre Training Data
      • Types of Fine Tuning
      • Self Instruct Paper
      • Self-Alignment with Instruction Backtranslation
      • Systematic Evaluation of Instruction-Tuned Large Language Models on Open Datasets
      • Instruction Tuning
      • Instruction Fine Tuning - Alpagasus
      • Less is More For Alignment
      • Enhanced Supervised Fine Tuning
      • Visualising Data using t-SNE
      • UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
      • Training and Evaluation Datasets
      • What is perplexity?
  • MODELS
    • Foundation Models
      • The leaderboard
      • Foundation Models
      • LLama 2 - Analysis
      • Analysis of Llama 3
      • Llama 3.1 series
      • Google Gemini 1.5
      • Platypus: Quick, Cheap, and Powerful Refinement of LLMs
      • Mixtral of Experts
      • Mixture-of-Agents (MoA)
      • Phi 1.5
        • Refining the Art of AI Training: A Deep Dive into Phi 1.5's Innovative Approach
      • Phi 2.0
      • Phi-3 Technical Report
  • Training
    • The Fine Tuning Process
      • Why fine tune?
        • Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?
        • Explanations in Fine Tuning
      • Tokenization
        • Tokenization Is More Than Compression
        • Tokenization - SentencePiece
        • Tokenization explore
        • Tokenizer Choice For LLM Training: Negligible or Crucial?
        • Getting the most out of your tokenizer for pre-training and domain adaptation
        • TokenMonster
      • Parameter Efficient Fine Tuning
        • P-Tuning
          • The Power of Scale for Parameter-Efficient Prompt Tuning
        • Prefix-Tuning: Optimizing Continuous Prompts for Generation
        • Harnessing the Power of PEFT: A Smarter Approach to Fine-tuning Pre-trained Models
        • What is Low-Rank Adaptation (LoRA) - explained by the inventor
        • Low Rank Adaptation (Lora)
        • Practical Tips for Fine-tuning LMs Using LoRA (Low-Rank Adaptation)
        • QLORA: Efficient Finetuning of Quantized LLMs
        • Bits and Bytes
        • The Magic behind Qlora
        • Practical Guide to LoRA: Tips and Tricks for Effective Model Adaptation
        • The quantization constant
        • QLORA: Efficient Finetuning of Quantized Language Models
        • QLORA and Fine-Tuning of Quantized Language Models (LMs)
        • ReLoRA: High-Rank Training Through Low-Rank Updates
        • SLoRA: Federated Parameter Efficient Fine-Tuning of Language Models
        • GaLora: Memory-Efficient LLM Training by Gradient Low-Rank Projection
      • Hyperparameters
        • Batch Size
        • Padding Tokens
        • Mixed precision training
        • FP8 Formats for Deep Learning
        • Floating Point Numbers
        • Batch Size and Model loss
        • Batch Normalisation
        • Rethinking Learning Rate Tuning in the Era of Language Models
        • Sample Packing
        • Gradient accumulation
        • A process for choosing the learning rate
        • Learning Rate Scheduler
        • Checkpoints
        • A Survey on Efficient Training of Transformers
        • Sequence Length Warmup
        • Understanding Training vs. Evaluation Data Splits
        • Cross-entropy loss
        • Weight Decay
        • Optimiser
        • Caching
      • Training Processes
        • Extending the context window
        • PyTorch Fully Sharded Data Parallel (FSDP)
        • Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation
        • YaRN: Efficient Context Window Extension of Large Language Models
        • Sliding Window Attention
        • LongRoPE
        • Reinforcement Learning
        • An introduction to reinforcement learning
        • Reinforcement Learning from Human Feedback (RLHF)
        • Direct Preference Optimization: Your Language Model is Secretly a Reward Model
  • INFERENCE
    • Why is inference important?
      • Grouped Query Attention
      • Key Value Cache
      • Flash Attention
      • Flash Attention 2
      • StreamingLLM
      • Paged Attention and vLLM
      • TensorRT-LLM
      • Torchscript
      • NVIDIA L40S GPU
      • Triton Inference Server - Introduction
      • Triton Inference Server
      • FiDO: Fusion-in-Decoder optimised for stronger performance and faster inference
      • Is PUE a useful measure of data centre performance?
      • SLORA
  • KNOWLEDGE
    • Vector Databases
      • A Comprehensive Survey on Vector Databases
      • Vector database management systems: Fundamental concepts, use-cases, and current challenges
      • Using the Output Embedding to Improve Language Models
      • Decoding Sentence-BERT
      • ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT
      • SimCSE: Simple Contrastive Learning of Sentence Embeddings
      • Questions Are All You Need to Train a Dense Passage Retriever
      • Improving Text Embeddings with Large Language Models
      • Massive Text Embedding Benchmark
      • RocketQAv2: A Joint Training Method for Dense Passage Retrieval and Passage Re-ranking
      • LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders
      • Embedding and Fine-Tuning in Neural Language Models
      • Embedding Model Construction
      • Demystifying Embedding Spaces using Large Language Models
      • Fine-Tuning Llama for Multi-Stage Text Retrieval
      • Large Language Model Based Text Augmentation Enhanced Personality Detection Model
      • One Embedder, Any Task: Instruction-Finetuned Text Embeddings
      • Vector Databases are not the only solution
      • Knowledge Graphs
        • Harnessing Knowledge Graphs to Elevate AI: A Technical Exploration
        • Unifying Large Language Models and Knowledge Graphs: A Roadmap
      • Approximate Nearest Neighbor (ANN)
      • High Dimensional Data
      • Principal Component Analysis (PCA)
      • Vector Similarity Search - HNSW
      • FAISS (Facebook AI Similarity Search)
      • Unsupervised Dense Retrievers
    • Retrieval Augmented Generation
      • Retrieval-Augmented Generation for Large Language Models: A Survey
      • Fine-Tuning or Retrieval?
      • Revolutionising Information Retrieval: The Power of RAG in Language Models
      • A Survey on Retrieval-Augmented Text Generation
      • REALM: Retrieval-Augmented Language Model Pre-Training
      • Retrieve Anything To Augment Large Language Models
      • Generate Rather Than Retrieve: Large Language Models Are Strong Context Generators
      • Active Retrieval Augmented Generation
      • DSPy: LM Assertions: Enhancing Language Model Pipelines with Computational Constraints
      • DSPy: Compiling Declarative Language Model Calls
      • DSPy: In-Context Learning for Extreme Multi-Label Classification
      • Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs
      • HYDE: Revolutionising Search with Hypothetical Document Embeddings
      • Enhancing Recommender Systems with Large Language Model Reasoning Graphs
      • Retrieval Augmented Generation (RAG) versus fine tuning
      • RAFT: Adapting Language Model to Domain Specific RAG
      • Summarisation Methods and RAG
      • Lessons Learned on LLM RAG Solutions
      • Stanford: Retrieval Augmented Language Models
      • Overview of RAG Approaches with Vector Databases
      • Mastering Chunking in Retrieval-Augmented Generation (RAG) Systems
    • Semantic Routing
    • Resource Description Framework (RDF)
  • AGENTS
    • What is agency?
      • Rephrase and Respond: Let Large Language Models Ask Better Questions for Themselves
      • Types of Agents
      • The risk of AI agency
      • Understanding Personality in Large Language Models: A New Frontier in AI Psychology
      • AI Agents - Reasoning, Planning, and Tool Calling
      • Personality and Brand
      • Agent Interaction via APIs
      • Bridging Minds and Machines: The Legacy of Newell, Shaw, and Simon
      • A Survey on Language Model based Autonomous Agents
      • Large Language Models as Agents
      • AI Reasoning: A Deep Dive into Chain-of-Thought Prompting
      • Enhancing AI Reasoning with Self-Taught Reasoner (STaR)
      • Exploring the Frontier of AI: The "Tree of Thoughts" Framework
      • Toolformer: Revolutionising Language Models with API Integration - An Analysis
      • TaskMatrix.AI: Bridging Foundational AI Models with Specialised Systems for Enhanced Task Completion
      • Unleashing the Power of LLMs in API Integration: The Rise of Gorilla
      • Andrew Ng's presentation on AI agents
      • Making AI accessible with Andrej Karpathy and Stephanie Zhan
  • Regulation and Ethics
    • Regulation and Ethics
      • Privacy
      • Detecting AI Generated content
      • Navigating the IP Maze in AI: The Convergence of Blockchain, Web 3.0, and LLMs
      • Adverse Reactions to generative AI
      • Navigating the Ethical Minefield: The Challenge of Security in Large Language Models
      • Navigating the Uncharted Waters: The Risks of Autonomous AI in Military Decision-Making
  • DISRUPTION
    • Data Architecture
      • What is a data pipeline?
      • What is Reverse ETL?
      • Unstructured Data and Generatve AI
      • Resource Description Framework (RDF)
      • Integrating generative AI with the Semantic Web
    • Search
      • BM25 - Search Engine Ranking Function
      • BERT as a reranking engine
      • BERT and Google
      • Generative Engine Optimisation (GEO)
      • Billion-scale similarity search with GPUs
      • FOLLOWIR: Evaluating and Teaching Information Retrieval Models to Follow Instructions
      • Neural Collaborative Filtering
      • Federated Neural Collaborative Filtering
      • Latent Space versus Embedding Space
      • Improving Text Embeddings with Large Language Models
    • Recommendation Engines
      • On Interpretation and Measurement of Soft Attributes for Recommendation
      • A Survey on Large Language Models for Recommendation
      • Model driven recommendation systems
      • Recommender AI Agent: Integrating Large Language Models for Interactive Recommendations
      • Foundation Models for Recommender Systems
      • Exploring the Impact of Large Language Models on Recommender Systems: An Extensive Review
      • AI driven recommendations - harming autonomy?
    • Logging
      • A Taxonomy of Anomalies in Log Data
      • Deeplog
      • LogBERT: Log Anomaly Detection via BERT
      • Experience Report: Deep Learning-based System Log Analysis for Anomaly Detection
      • Log-based Anomaly Detection with Deep Learning: How Far Are We?
      • Deep Learning for Anomaly Detection in Log Data: A Survey
      • LogGPT
      • Adaptive Semantic Gate Networks (ASGNet) for log-based anomaly diagnosis
  • Infrastructure
    • The modern data centre
      • Enhancing Data Centre Efficiency: Strategies to Improve PUE
      • TCO of NVIDIA GPUs and falling barriers to entry
      • Maximising GPU Utilisation with Kubernetes and NVIDIA GPU Operator
      • Data Centres
      • Liquid Cooling
    • Servers and Chips
      • The NVIDIA H100 GPU
      • NVIDIA H100 NVL
      • Lambda Hyperplane 8-H100
      • NVIDIA DGX Servers
      • NVIDIA DGX-2
      • NVIDIA DGX H-100 System
      • NVLink Switch
      • Tensor Cores
      • NVIDIA Grace Hopper Superchip
      • NVIDIA Grace CPU Superchip
      • NVIDIA GB200 NVL72
      • Hopper versus Blackwell
      • HGX: High-Performance GPU Platforms
      • ARM Chips
      • ARM versus x86
      • RISC versus CISC
      • Introduction to RISC-V
    • Networking and Connectivity
      • Infiniband versus Ethernet
      • NVIDIA Quantum InfiniBand
      • PCIe (Peripheral Component Interconnect Express)
      • NVIDIA ConnectX InfiniBand adapters
      • NVMe (Non-Volatile Memory Express)
      • NVMe over Fabrics (NVMe-oF)
      • NVIDIA Spectrum-X
      • NVIDIA GPUDirect
      • Evaluating Modern GPU Interconnect
      • Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)
      • Next-generation networking in AI environments
      • NVIDIA Collective Communications Library (NCCL)
    • Data and Memory
      • NVIDIA BlueField Data Processing Units (DPUs)
      • Remote Direct Memory Access (RDMA)
      • High Bandwidth Memory (HBM3)
      • Flash Memory
      • Model Requirements
      • Calculating GPU memory for serving LLMs
      • Transformer training costs
      • GPU Performance Optimisation
    • Libraries and Complements
      • NVIDIA Base Command
      • NVIDIA AI Enterprise
      • CUDA - NVIDIA GTC 2024 presentation
      • RAPIDs
      • RAFT
    • Vast Data Platform
      • Vast Datastore
      • Vast Database
      • Vast Data Engine
      • DASE (Disaggregated and Shared Everything)
      • Dremio and VAST Data
    • Storage
      • WEKA: A High-Performance Storage Solution for AI Workloads
      • Introduction to NVIDIA GPUDirect Storage (GDS)
        • GDS cuFile API
      • NVIDIA Magnum IO GPUDirect Storage (GDS)
      • Vectors in Memory
Powered by GitBook
LogoLogo

Continuum - Accelerated Artificial Intelligence

  • Continuum Website
  • Axolotl Platform

Copyright Continuum Labs - 2023

On this page

Was this helpful?

  1. KNOWLEDGE
  2. Retrieval Augmented Generation

Overview of RAG Approaches with Vector Databases

Some random tips on RAG and vector databases

Significance in Customisation for Specific Use Cases: RAG is particularly useful for customising language models to specific organisational needs and datasets, making it a common initial use case in many companies.

Complexity in Retrieval Augmented Generation: The retrieval process in RAG involves identifying the most relevant document segments to be fed into the model for generating responses. This process is complicated by the need to balance the context size – the amount of data the LLM can process at once – and the specificity of the information retrieved.

Experimentation and Adaptation: The process of chunking, embedding generation, and retrieval requires experimentation and adaptation to the specific document set and application requirements. This is a trial-and-error process, where the chunk size and the method of handling overlaps are fine-tuned for optimal retrieval and context representation.

Challenges in Document Parsing: A significant challenge in implementing RAG is the parsing of documents. The context and structure of the documents are crucial. For example, policies and regulations have hierarchical structures that need to be retained in the chunks for the context to remain intact. Misrepresenting or losing this structure in parsing can lead to incomplete or incorrect context, affecting the response accuracy.

File Parsing Techniques and Challenges: Parsing various document formats (like PDFs, Word documents, or Excel spreadsheets) into a data structure that preserves meaning while facilitating model operations is a major challenge.

Practical Challenges and Techniques: The real-world application of RAG involves dealing with messy, diverse data. Standard libraries and tools may not always be sufficient. Customised solutions might be required to handle specific data types, formats, and structures effectively.

Data Diversity and Complexity: In real-world applications, data sources are not limited to plain text documents. They include various formats like PDFs, Word documents, SharePoint sites, and even complex elements like tables. This diversity adds complexity to the process of converting these various formats into a uniform structure suitable for embedding and retrieval.

Optimising Embedding Generation: Generating embeddings for appropriately sized chunks of text is crucial. Too small chunks may miss context, while too large ones may contain excessive information. Finding the right balance is key to creating useful embeddings that represent meaningful text segments for effective retrieval.

Handling Overlapping Contexts: To ensure comprehensive context coverage without missing crucial information, chunks or windows of text for embedding generation might need to overlap. This approach allows for a broader context to be captured, improving the effectiveness of the retrieval process.

Importance of Document Hierarchy and Context: Maintaining the hierarchical structure of documents is crucial for understanding the context. For example, a rule stated in one section of a document might have exceptions listed in another section. Ignoring this hierarchy can lead to misunderstanding or misinterpreting the information.

Document summary method: Documents are summarised, and these summaries are embedded and stored in the vector database. For generation, the original full document is used instead of the summary, providing rich context to the language model.

Knowledge Graph-Aided RAG: Knowledge graphs, which store data in nodes and edges rather than traditional rows and columns, are particularly effective for data with complex relationships. In RAG, knowledge graphs can be used instead of chunking documents. Nodes in the knowledge graph are embedded and stored in a vector database.

Large Language Models Writing Graph Queries: Large language models can be used to transcribe natural language into graph queries, effectively querying knowledge graphs and retrieving relevant information.

Impact of Larger Context Windows in RAG: With advancements like GPT-4's 128k token context window, more data can be provided to the large language model, offering richer context.

Structured Retrieval as an Optimisation Method: Structured retrieval involves using metadata to assist the retrieval process. This method can be particularly useful when dealing with databases with additional contextual information, like movies with metadata such as genre, director, or ratings.

Large language models can be trained to understand and use metadata schemas, allowing them to automatically tag incoming queries with relevant metadata for more precise filtering. This approach is particularly useful for time-based retrieval, where embeddings are tagged with date and time metadata, allowing searches within specific time frames or for the most recent data.

Advanced Structured Retrieval: The concept of structured retrieval is where metadata is used to pre-filter the data before conducting a similarity search. This method enhances efficiency by narrowing down the search space.

Application in Real-World Scenarios: using metadata for time-based retrieval and handling different types of documents separately shows the practical applicability of RAG in diverse real-world scenarios, such as financial analysis, news aggregation, and historical data analysis.

Handling Hierarchical Data in RAG: For very large documents that don't fit into a single context window, a hierarchical chunking approach can be used. Documents are chunked into larger segments, and these segments are further broken down into smaller chunks.

Document-Specific Agents and Recursive Retrieval: In scenarios where documents are similar but should be treated separately (e.g., financial documents from different companies), document-specific agents can be used. Each document has its own retriever and summary, and an agent decides which path to take based on the user query.

Reranking for Improved Retrieval Accuracy: Reranking is a method to improve the retrieval of chunks in RAG. After the initial embedding and similarity search, a reranking process further refines the selection of relevant chunks. This process addresses the potential information loss that might occur during the embedding phase, enhancing the precision of the retrieved data.

Reranking for Enhanced Retrieval Accuracy: Reranking involves using a transformer or deep learning model after an initial vector search to refine and prioritise the most relevant chunks. This process addresses the potential inaccuracies of similarity searches in vector spaces, ensuring that the best possible chunks are retrieved for generation. Although reranking is more accurate, it's also more resource-intensive, which is why it's not feasible to use it for the entire dataset.

Sentence text windows: Documents are chunked at the sentence level for retrieval, but a larger text window is provided for generation to ensure adequate context.

Parent Document Retriever: Similar to the document summary method, chunks are linked back to their parent document. Relevant chunks lead to the entire parent document being passed to the language model, offering comprehensive context.

Hierarchical Chunking: This approach involves chunking large documents into progressively smaller units, such as page-size chunks, then paragraph-size, and finally sentence-size chunks. This method balances the granularity of smaller chunks with the context provided by larger chunks, improving the relevance and accuracy of retrieval.

Prompt Engineering for Response Length and Relevance: The issue of generating responses of varying lengths depending on the chunk size and use case is more related to prompt engineering rather than chunk size. Prompt engineering involves crafting the input to the language model in a way that guides the length and style of the response.

Enhancing Language Model Contextual Understanding: By linking summaries or specific chunks to their parent documents and using these in the generation process, RAG techniques demonstrate a commitment to enhancing the contextual understanding of language models. This approach ensures that responses generated are not just relevant but also contextually rich and accurate.

Retaining Document Structure in ML Models: there's often a need to flatten hierarchical document structures into a format more amenable to machine learning models. This involves transforming documents into a flat representation like a list or dictionary, where each element is tagged with keys representing the hierarchical structure (e.g., file name, chapter, subsection, paragraph).

Application-Specific Retrieval and Design: The relevance of retrieved information is highly application-specific and depends on careful design and user feedback. This process is integral to the RAG system's effectiveness and underlines the ongoing need for skilled data scientists and programmers in AI development.

PreviousStanford: Retrieval Augmented Language ModelsNextMastering Chunking in Retrieval-Augmented Generation (RAG) Systems

Last updated 1 year ago

Was this helpful?

Page cover image