SKU/Artículo: AMZ-B0GF1QPQYJ
Building LLM Inference Engines with C++23: Optimization Techniques for Deploying Generative AI on Edge Devices and Consumer Hardware.
Format:
Hardcover
Detalles del producto
Disponibilidad:
En stock
En stock
Peso con empaque:
0.59 kg
0.59 kg
Devolución:
Sí
Sí
Condición
Nuevo
Nuevo
Producto de:
Amazon
Amazon
Viaja desde
USA
USA
Sobre este producto
- The "Black Box" of AI Ends Today. You have fine-tuned models in PyTorch. You have built chatbots with LangChain. You have called the OpenAI API. But do you know what happens in the microseconds between a user’s prompt and the first token? Most developers treat Large Language Model inference as magic, a heavy Python script wrapped in a container. But for production systems, "magic" is a liability. Magic consumes too much RAM. Magic suffers from unpredictable latency spikes. Magic cannot run on the edge. To build the next generation of AI infrastructure, you must stop calling libraries and start engineering systems. It is time to go Bare Metal. This book is not a theoretical overview of neural networks. It is a rigorous, code-first engineering manual. It takes you from a blank C++ source file to a production-grade inference engine capable of running modern models like Llama 3 and Mistral with throughput that rivals the giants. Written for systems programmers, AI engineers, and performance-obsessed developers, this guide dismantles the abstractions of modern deep learning. We strip away the Python interpreter, the garbage collector, and the framework overhead to reveal the raw mathematics and silicon underneath. What You Will Build: In this comprehensive guide, you will architect a specialized runtime engine from the ground up using modern C++23. You will not just learn about these concepts; you will implement them line-by-line:Memory Mastery: Replace malloc with custom Arena Allocators and mmap loaders. Learn to manage the KV Cache using Paged Attention to eliminate fragmentation and waste.The Compute Core: Hand-write AVX-512 and NEON intrinsics to saturate your CPU’s floating-point units. Implement block-wise Quantization (INT8/INT4) to run massive models on consumer hardware.Parallel Architecture: Design a lock-free, work-stealing thread pool that balances workloads across Performance and Efficiency cores, avoiding the pitfalls of OS scheduling and False Sharing.Hardware Acceleration: Break free from the CPU. Implement a Heterogeneous Compute backend using Vulkan and Metal Compute Shaders with Zero-Copy memory sharing for integrated GPUs.Distributed Inference: Scale beyond a single machine. Build a custom, ultra-low-latency TCP/IP networking stack to split large models across a cluster using Tensor Parallelism and Pipeline Parallelism.Production Readiness: Bridge the gap between C++ speed and Python usability. Create seamless Python bindings with pybind11 and serve your engine via an OpenAI-compatible FastAPI server.Who This Book Is For:Software Engineers who want to transition into AI infrastructure and High-Performance Computing (HPC).Machine Learning Engineers tired of the "Python Ceiling" who need to optimize latency-critical applications.C++ Developers looking for a practical, deep-dive project into the world of Generative AI.The Tech Stack: C++23 AVX-512 Vulkan Metal Pybind11 TCP/IP Sockets CMake Stop waiting on the GPU. Stop paying the latency tax. Take control of your inference pipeline. Scroll up and grab your copy to start building the engine of the future today.
AR$117.202
44% OFF
AR$65.114
IMPORT EASILY
By purchasing this product you can deduct VAT with your RUT number
AR$117.202
44% OFF
AR$65.114
Pagá fácil y rápido con Mercado Pago o MODO
Llega en 8 a 12 días hábiles
con envío
Tienes garantía de entrega
Este producto viaja
de USA ![]()
a tus manos en
![]()
a tus manos en
Compra protegida
Disfruta de una experiencia de compra segura y confiable
Tarjetas de Crédito y Débito
Visa
Mastercard
American Express
Pagos a través de PayPal
Compra procesada en dólares con dinero en cuenta o tarjetas internacionales.
Paypal
Pagos a través de MODO
Compra procesada en pesos desde la App de tu banco.
PayPal
