Price: $58.99 - $56.04
(as of Dec 17,2024 07:08:38 UTC – Details)

From the Publisher

Book Overview

The book is for anyone who wants to build LLM products that can serve real use cases today. It explores various methods to adapt “foundational” LLMs to specific tasks with enhanced accuracy, reliability, and scalability. It tackles the lack of reliability of “out of the box” LLMs by teaching the AI developer tech stack of the future; Prompting, Fine-Tuning, RAG, and Tools Use.

LLMs are a fast-evolving and competitive field and new models and techniques will appear. These will unlock new capabilities, but today’s LLM developer stack is transferable and will also be essential for adapting next-generation models to specific data and industries. Those using the models of today are best placed to take advantage of the models of the future! We focus on teaching the core principles of building production products with LLMs which will keep this book relevant as models change.

LLMs are very different from other software technologies and are already widely deployed online to hundreds of millions of users. As they continue to advance, it’s crucial for workers across all sectors to adapt and develop skills that complement AI capabilities. There will never be a better time to learn how LLMs work and how to develop with them!

This book comes with access to our webpage where we also share lots of additional up-to-date content, code, notebooks, and resources.

This book breaks down techniques that are scalable for enterprise-level workflows, helping both independent developers and small companies with limited resources create AI products that deliver value to paying customers.

Who is it for?

AI Practitioners
AI/ML Engineers
Students/Researchers
Computer Science Professionals
Programmers
Tinkerers
Job Seekers

LLM Fundamentals, Architecture, & LLMs in Practice

Foundations

Building blocks of LLMs: language modeling, tokenization, embeddings, emergent abilities, scaling laws, context size…
Transformer Architecture: attention mechanism, design choices, encoder-only transformers, decoder-only transformers, encoder-decoder transformers, GPT Architecture, Masked Self-Attention, MinGPT

LLMs in Practice

Hallucinations & Biases: Mitigation strategies, controlling LLM outputs
Decoding methods: greedy search, sampling, beam search, top-k sampling, top-p sampling
Objective functions and evaluation metrics: perplexity metric and GLUE, SuperGLUE, BIG-Bench, HELM, FLASK Benchmarks…

Prompting & Frameworks

Prompting

Prompting techniques: zero-shot, in context, few-shot, role, chains, and chain-of-thought…
Prompt Injection and Prompt Hacking

Frameworks

LangChain: prompt templates, output parsers, summarization chain, QA chains
LlamaIndex: vector stores, embeddings, data connectors, nodes, indexes

RAG & Fine-Tuning

Retrieval-Augmented Generation Components

Data Ingestion(PDFs, web pages, Google Drive), text splitters, embeddings, LangChain Chains
Querying in LlamaIndex: query construction, expansion, transformation, splitting, customizing a retriever engine…
Reranking Documents: recursive, small-to-big
RAG Metrics: Mean Reciprocal Rank (MRR), Hit Rate, Mean Average Precision (MAP), and Normalized Discounted Cumulative Gain (NDCG)…
Evaluation Tools: evaluating with ragas, custom evaluation of RAG pipelines

Fine-Tuning Optimization Techniques

LoRA, QLoRA, supervised fine-tuning, SFT RLHF

Agents, Optimization & Deployment

Agents

Using AutoGPT & BabyAGI with LangChain
Agent Simulation Project: CAMEL, Generative Agents
Building Agents, LangGPT, OpenAI Assistants

Optimization & Deployment

Challenges, quantization, pruning, distillation, cloud deployment, CPU and GPU optimization & deployment, creating APIs from open-source LLMs

ASIN ‏ : ‎ B0D4FFPFW8
Publisher ‏ : ‎ Independently published (May 21, 2024)
Language ‏ : ‎ English
Paperback ‏ : ‎ 463 pages
ISBN-13 ‏ : ‎ 979-8324731472
Item Weight ‏ : ‎ 1.85 pounds
Dimensions ‏ : ‎ 7.44 x 1.05 x 9.69 inches

Building large language models (LLMs) for production is a complex and challenging task that requires careful consideration of various factors to ensure their optimal performance. In order to enhance the abilities and reliability of LLMs, developers often turn to techniques such as prompting, fine-tuning, and using the RAG (Retrieval-Augmented Generation) framework.

Prompting is a technique where specific instructions or cues are provided to the LLM to guide its responses in a desired direction. By providing relevant prompts, developers can steer the model towards generating more accurate and contextually appropriate outputs, ultimately improving its overall performance.

Fine-tuning is another crucial step in the development of LLMs for production. This process involves training the model on a specific dataset or task to adapt it to a particular domain or set of requirements. By fine-tuning the LLM, developers can tailor its capabilities to better suit the needs of a specific application, ultimately improving its reliability and effectiveness.

The RAG framework, which combines retrieval-based and generative approaches, is another powerful tool for enhancing LLM abilities and reliability. By incorporating a retrieval mechanism that retrieves relevant information from a large knowledge base before generating responses, the RAG framework can improve the model’s accuracy and coherence, making it more reliable in real-world applications.

In conclusion, building LLMs for production requires a combination of techniques such as prompting, fine-tuning, and leveraging frameworks like RAG to enhance their abilities and reliability. By carefully considering these factors and implementing best practices in model development, developers can create LLMs that deliver high-quality outputs and meet the needs of a wide range of applications.
#Building #LLMs #Production #Enhancing #LLM #Abilities #Reliability #Prompting #FineTuning #RAG