Tools and methodology for performance analysis of large language models training on GPUs

This master thesis presents a tool, a methodology, a set of metrics, and practical examples for evaluating the performance of training large language models in HPC environments. NSYS2PRV is a tool that converts NVIDIA Nsight Systems reports into traces compatible with Paraver, enabling significantly...

Descripción completa

Detalles Bibliográficos
Autor: Clascà Ramírez, Marc
Tipo de recurso: tesis de maestría
Fecha de publicación:2026
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:dnet:upcommonspor::bdf4723b76e74b654b7915a6216ed512
Acceso en línea:https://hdl.handle.net/2117/460885
Access Level:acceso abierto
Palabra clave:Graphics processing units
High performance computing
Artificial intelligence
Computaciço d'alt rendiment
Eines de rendiment
Intel·ligència artificial
Anàlisi de rendiment
Visualització de dades
Unitats de processament gràfic
Models de llenguatge
High-performance computing
Performance tools
Performance analysis
Data visualization
Large language models
Scalability
Processadors gràfics
Càlcul intensiu (Informàtica)
Àrees temàtiques de la UPC::Informàtica::Intel·ligència artificial
Descripción
Sumario:This master thesis presents a tool, a methodology, a set of metrics, and practical examples for evaluating the performance of training large language models in HPC environments. NSYS2PRV is a tool that converts NVIDIA Nsight Systems reports into traces compatible with Paraver, enabling significantly enhanced insight compared to current performance analysis practices. By leveraging the capabilities of a well-established HPC performance analysis tool, we enable the comparison of execution traces and the quantification of microscopic-level differences to explain behaviors across hundreds or more computing devices. We argue that large-scale GPU applications and AI workloads can greatly benefit from the type of large-scale performance analysis introduced here, an approach that is not yet widely adopted in this domain. Translating nsys-generated traces to Paraver allows analysts to combine the fine-grained, highly accurate execution data obtainable from proprietary tools with the flexibility and scalability of an open-source, parallel performance analysis environment. Paraver also enables easy, customizable computation of efficiency metrics. This work demonstrates a more effective and insightful analysis experience than that offered by the native visualization tools in Nsight Systems. Additionally, we introduce a set of Paraver compatible metrics that guide the analysis process, and we showcase examples where these metrics were successfully applied to real-world AI and HPC workloads.