What you will learn in Hugging Face in Action
Build intelligent applications with transformers, agents, and RAG
Chapter by chapter
Chapter 3: Using Hugging Face transformers and pipelines for NLP tasks
What you learn in chapter 3 →Chapter 5: Exploring, tokenizing, and visualizing Hugging Face datasets
What you learn in chapter 5 →Chapter 6: Fine-tuning pretrained models and working with multimodal models
What you learn in chapter 6 →Chapter 7: Creating LLM-based applications using LangChain and LlamaIndex
What you learn in chapter 7 →Chapter 11: Building locally running LLM-based applications using GPT4All
What you learn in chapter 11 →Chapter 13: Bridging LLMs to the real world with the Model Context Protocol
What you learn in chapter 13 →Concepts explained in this book
- templatePromptTemplate is a LangChain class that represents a parameterized string template and formats user-provided input variables into a prompt.
- language modelA large language model is a high-capacity neural model trained on large text corpora to process and generate language.
- GradioA Gradio application is a machine-learning web application created with the Gradio library.
- MCP serverAn MCP server is a software service that implements the Model Context Protocol to expose data or executable capabilities to AI assistants and clients.
- LangGraphLangGraph is a framework for building stateful, multi-step applications whose components communicate through graph-based state transitions.
- tokenizationA tokenization strategy is a rule or method for dividing text into model-processing units.
- fine-tuningFine-tuning is the adaptation of a pretrained model by continuing training on data for a particular task or domain.
- toolA tool is an executable capability that an agent or language-model application can invoke to perform a task or access external functionality.
- transformersTransformers is an open-source library for working with pretrained machine-learning models, especially models for natural-language processing and computer vision.
- GPT4AllGPT4All is an open-source software project and Python library for running and interacting with language models on local hardware.
- classifierClassification is the task of assigning an input to one or more predefined categories.
- foundational modelA pretrained model is a machine learning model whose parameters have already been learned from a training dataset before being adapted or used for a particular task.
- AI agentAn agent is a specialized AI system that autonomously performs tasks by understanding language, reasoning and planning, selecting tools, and delivering results.
- CUDACUDA is NVIDIA's parallel computing platform and programming model for executing general-purpose computations on GPUs.
- softmaxSoftmax is a mathematical function that converts a vector of numbers into a probability distribution by exponentiating and normalizing its values.
- transformer architectureThe Transformer architecture is a neural-network architecture based on attention mechanisms for processing sequences and other structured inputs.
- encoderA text encoder is a neural network component that converts textual input into an internal numerical representation.
- ReActReAct is an agent technique in which a language model alternates between reasoning, taking actions such as tool calls, observing results, and continuing reasoning.
- facebook/detr-resnet-50A pretrained DETR model repository containing a ResNet-50-based object detection model.
- multimodalA multimodal model is a machine learning model that jointly processes or relates inputs from multiple data modalities, such as images and text.
- token idA token ID is an integer identifier representing a token in a model's vocabulary.
- Hugging FaceThe Hugging Face Hub is a platform for hosting, discovering, sharing, and collaborating on machine learning models and datasets.
- MCP resourceAn MCP resource is a server-provided data source that an MCP client can inspect or access through the Model Context Protocol.
- tokenA token is a unit of text that a machine learning model processes as a single input element.
- LangFlowLangflow is a visual tool for prototyping applications built around large language models and related components.
- auto classAutoTokenizer is a Transformers factory class that selects and loads the tokenizer appropriate for a specified pretrained model.
- confidenceA confidence score is a numerical value expressing a model's estimated certainty in a prediction.
- GPT-4o-minigpt-4o-mini is a compact OpenAI multimodal language model designed for capable, cost-efficient generation and chat tasks.
- object detectionObject detection is the machine learning task of identifying and locating objects within images or video.
- single-modal modelA single-modal model is a machine learning model that processes data from one modality, such as only images or only text.
- logitA logit is an unnormalized model output representing evidence for a class before conversion into normalized probabilities.
- graphA graph is a structure consisting of connected nodes that represent stages and transitions in a workflow.
- bounding boxA bounding box is a rectangular region specifying the position of an object within an image or video.
- dataset splitA dataset split is a partition of a dataset designated for a particular stage or purpose in machine-learning workflows.
- zero-shot learningZero-shot classification is the assignment of labels to inputs without task-specific labeled examples during model adaptation or inference.
- local aiA local model is a machine learning model executed on the user's own machine rather than accessed through a remote service.
- GPUA graphics processing unit is a massively parallel processor designed to accelerate computation, including machine learning workloads.
- sentiment analysisSentiment analysis is the computational classification of text according to the expressed attitude or polarity, such as positive or negative sentiment.
- byte pair encodingByte-pair encoding is a subword tokenization algorithm that iteratively merges frequent adjacent units into a vocabulary of subword tokens.
- launchlaunch is a Gradio method that starts a configured application and makes its interface available for interaction.
- SerpAPIWrapperSerpAPIWrapper is a wrapper that exposes SerpAPI web-search functionality through a LangChain-compatible tool interface.
- query engineA query engine is a software component that accepts a query and returns a response from an underlying data or knowledge source.
- CLIPCLIP is a multimodal model trained to align images and text in a shared representation space.
- mcp clientAn MCP client is an AI assistant or application component that sends requests to an MCP server to access external tools or data.
- padding tokenA padding token is a special token inserted into shorter sequences so batched sequences have a uniform length.
- token type IDsToken type IDs are integer labels that identify which input segment each token belongs to.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.