| Autore |
Luu Hien
|
| Edizione | [1st ed.] |
| Pubbl/distr/stampa |
Berkeley, CA : , : Apress L. P., , 2024
|
| Descrizione fisica |
1 online resource (342 pages)
|
| Altri autori (Persone) |
PumperlaMax
ZhangZhe
|
| ISBN |
9798868803765
9798868803758
|
| Formato |
Materiale a stampa  |
| Livello bibliografico |
Monografia |
| Lingua di pubblicazione |
eng
|
| Nota di contenuto |
Intro -- Table of Contents -- About the Authors -- About the Technical Reviewer -- Chapter 1: Introduction to MLOps -- MLOps Overview -- ML Projects -- ML Project Inputs and Artifacts -- MLOps: The Missing Element -- Operationalize ML Project Challenges -- Applying Machine Learning -- Garbage In, Garbage Out -- In the Beginning -- Team Sport -- Summary of Challenges -- Automation -- Reproducibility -- Monitoring -- MLOps: The Promise -- Paradigm -- Engineering Discipline -- Data Engineering -- Machine Learning -- DevOps -- Principles -- Automation -- Versioning -- Experiment Tracking -- Reproducibility -- Testing -- Data-Related Testing -- Model-Related Testing -- Continuous ML Training, Evaluation, and Deployment -- Continuous Monitoring -- MLOps Canonical Stack -- MLOps Blueprint -- MLOps Components -- Feature Engineering -- Feature Store -- Notebook Service -- Model Training -- Abstraction -- Compute Resource -- Continuous Training -- Consistency and Reproducibility -- Experimentation -- Model Store -- Managing ML Models and Lifecycle -- Model Traceability and Reproducibility -- Model Deployment -- Model Serving -- Prediction Store -- ML Observability -- Monitoring -- Observability -- Explainability -- MLOps Pillars -- Feature Engineering -- Model Training and Management -- Model Serving -- ML Observability -- Summary -- Chapter 2: MLOps Adoption Strategies and Case Studies -- Adoption Strategies -- Goals Alignment -- MLOps Need Assessment -- Use Cases -- Fraud Detection -- Churn Prediction -- Loan Approval and Credit Scoring -- Technology -- People -- Culture -- Risk Tolerance -- Execution Velocity -- Decision-Making Process -- Collaboration Style -- Maturity Level -- MLOps Infrastructure Approaches -- Build -- Buy -- Hybrid -- MLOps Landscape -- Platforms and Tools -- Case Studies -- Uber Michelangelo -- Key Takeaways and Lessons Learned.
Meta FBLearner -- Key Takeaways and Lessons Learned -- Summary -- Chapter 3: Feature Engineering Infrastructure -- Overview -- Benefits -- High-Level Architecture -- Feature Specification and Definition -- Feature Metadata -- Feature Metadata Format -- YAML-Based Feature Specification and Definition Examples -- Python-Based Feature Specification and Definition Examples -- Feature Registry -- Feature Orchestration -- Key Considerations -- Feature Store -- Offline Feature Store -- Key Considerations -- Online Feature Store -- Key Considerations -- Feature Upload -- Key Considerations -- Feature Serving -- Monitoring -- Build vs. Buy -- Important Factors -- Build -- Buy -- Organizational Challenges -- Data Availability -- Data Governance -- Case Studies -- Open Source -- Overview -- Concepts -- Architecture -- Assessments -- Strengths -- Opportunities -- In-House -- Overview -- Producer -- Consumer -- Concepts -- Architecture -- Assessment -- Strengths -- Opportunities -- Vendor Solutions -- Overview -- Concepts -- Architecture -- Assessment -- Strengths -- Opportunities -- Summary -- Chapter 4: Model Training Infrastructure -- Overview -- High-Level Architecture -- Model Development Environment -- Data Access -- Compute Resource Access -- Model Development Experience -- Reproducibility -- Experiment Tracking -- High-Level Architecture -- Model Training Pipelines -- Orchestration -- Orchestration Programming Style -- Continuous Model Training -- Model Training at Scale -- Distributed Model Training -- Model Registry -- High-Level Architecture -- Case Studies -- In-House -- Open Source -- MLflow Overview -- MLflow Feature Set -- MLflow Tracking -- MLflow Model Registry -- MLflow High-Level Architecture -- Summary -- Chapter 5: Model Serving Infrastructure -- Overview -- High-Level Architecture -- Feature Store -- Model Registry -- Metric Service.
Logging Service -- Inference Service -- Service Endpoint -- Inference Request Batching -- Model Loading and Unloading -- Feature Fetching -- Model Prediction -- Prediction Logging -- Prediction Service -- Machine Learning Framework -- Cost-Effectiveness -- Prediction Pre-processing and Post-processing -- Prediction Step Design Choice -- Embedded Prediction Step -- Remote Prediction Step -- Offline Inference -- Case Studies -- In-House -- LyftLearn Serving -- Reddit's Model Serving Architecture Evolution -- Open Source -- BentoML -- Seldon Core -- Ray Serve -- Python-Native Model Serving Application -- Flexible Scaling and Resource Allocation -- Multi-model Inference -- Summary -- Chapter 6: ML Observability Infrastructure -- Overview -- Model Performance -- Drift -- Data Quality -- Explainability -- High-Level Architecture -- Observability Store -- Feature Engineering -- Model Training -- Model Prediction -- Observability Store Implementation -- Case Studies -- Lyft: Model Monitoring -- Phase One -- Phase Two -- Cultural Shift for Adoption -- Open Source -- Great Expectations -- whylogs -- Evidently -- Summary -- Chapter 7: Ray Core -- Ray Core in a Nutshell -- Basic Concepts -- API Basics -- Architecture Basics -- Scheduling -- Fault Tolerance -- KubeRay -- Summary -- References -- Chapter 8: An Introduction to the Ray AI Libraries -- Overview -- What Are Ray's AI Libraries? -- Why and When to Use Ray for ML? -- AI Workloads to Run with Ray -- An Introduction to Ray's AI Libraries -- Datasets and Preprocessors -- Trainers -- Tuners and Checkpoints -- Running Batch Prediction -- Online Serving Deployments -- An Example of Training and Deploying Large Language Models with Ray -- Starting a Ray Cluster and Managing Dependencies -- Loading a Dataset and Preprocessing It -- Fine-Tuning a Language Model -- Training Runtime Considerations.
Generate Text from a Prompt -- Running Batch Inference for Our GPT-J Model -- Running Online Model Serving -- An Overview of Ray's Integrations -- How Ray Compares to Related Systems -- Distributed Python Frameworks -- Ray AI Libraries and the Broader ML Ecosystem -- How to Integrate Ray into Your ML Platform -- Summary -- Chapter 9: The Future of MLOps -- MLOps Landscape -- ML Development Lifecycle -- ML Infrastructure Architecture -- MLOps Maturity Model -- MLOps Solution Landscape -- AI/ML Landscape -- Generative AI -- Foundation Models -- Large Language Models (LLMs) -- AI Assistants -- Responsible AI -- Artificial General Intelligence (AGI) -- The Rise of LLMOps -- LLM Applications Archetypes -- Prompt Engineering Application Archetype -- Retrieval Augmented Generation (RAG) Application Archetype -- Fine-Tuning Application Archetype -- Model Training vs. LLM Fine-Tuning -- A Combined Application Archetype -- LLMOps Stack -- Summary -- Index.
|
| Record Nr. | UNINA-9910865243603321 |
Info
Mastering machine learning with spark 2.x : create scalable machine learning applications to power a modern data-driven business using spark / / Alex Tellez, Max Pumperla, Michal Malohlava
