Optimización de modelos de IA generativa en hardware de última generación
Loading...
Official URL
Full text at PDC
Publication date
2025
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
El objetivo de este Trabajo de Fin de Grado es analizar y mejorar la eficiencia y el rendimiento de modelos de inteligencia artificial generativa en hardware de última generación. En esta investigación identificaremos los principales cuellos de botella computacionales relacionados con la ejecución de modelos generativos y sugeriremos métodos para optimizarlos y aumentar su eficiencia. Para esto, se han analizado distintas configuraciones de hardware y software, utilizando diversas estrategias de disminución de latencia, gestión de recursos eficaz y ajuste de parámetros. Los resultados obtenidos permiten formular recomendaciones prácticas para la implementación de modelos de IA generativa en escenarios reales, y así poder tener un mejor equilibrio entre rendimiento, consumo energético y sostenibilidad.
The objective of this thesis is to analyze and improve the efficiency and performance of generative artificial intelligence models on next-generation hardware. In this research, we will identify the main computational bottlenecks related to the execution of generative models and suggest methods to optimize them and increase their efficiency. To this end, we have analyzed different hardware and software configurations, using various strategies to reduce latency, manage resources effectively, and adjust parameters. The results obtained allow us to formulate practical recommendations for the implementation of generative AI models in real-world scenarios, thereby achieving a better balance between performance, energy consumption, and sustainability.
The objective of this thesis is to analyze and improve the efficiency and performance of generative artificial intelligence models on next-generation hardware. In this research, we will identify the main computational bottlenecks related to the execution of generative models and suggest methods to optimize them and increase their efficiency. To this end, we have analyzed different hardware and software configurations, using various strategies to reduce latency, manage resources effectively, and adjust parameters. The results obtained allow us to formulate practical recommendations for the implementation of generative AI models in real-world scenarios, thereby achieving a better balance between performance, energy consumption, and sustainability.
Description
Trabajo de Fin de Grado en Ingeniería de Computadores, Facultad de Informática UCM, Departamento de Arquitectura de Computadores y Automática, Curso 2024/2025.













