Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ ZENODOarrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
ZENODO
Conference object . 2025
License: CC BY
Data sources: ZENODO
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
ZENODO
Article . 2025
License: CC BY
Data sources: Datacite
ZENODO
Article . 2025
License: CC BY
Data sources: Datacite
versions View all 4 versions
addClaim

Rendimiento, eficiencia y consumo energético de frameworks de entrenamiento para Deep Learning

Authors: Cortés, Gustavo; de Castro, Manuel; Bregon, Anibal; Martínez Prieto, Miguel Ángel; Llanos, Diego R.;

Rendimiento, eficiencia y consumo energético de frameworks de entrenamiento para Deep Learning

Abstract

El entrenamiento de modelos de Deep Learning es una tarea computacionalmente exigente, con costes significativos en términos de tiempo, recursos y consumo energético. A medida que estos modelos crecen en complejidad y tamaño, la elección del framework adecuado se vuelve un factor clave para optimizar su eficiencia. Existen múltiples frameworks, como TensorFlow, PyTorch y JAX, además de librerías como Keras, ampliamente utilizadas en la comunidad de Deep Learning. En este análisis comparamos su rendimiento y eficiencia en tres tipos de redes: perceptrones multicapa, convolucionales y de memoria a corto y largo plazo (LSTM). Se evalúan métricas como el tiempo de entrenamiento y prueba, el uso de memoria en GPU y el consumo energético total y por hora. Los resultados muestran que PyTorch sin Keras logra los tiempos de entrenamiento más bajos, especialmente en la red LSTM. JAX mantiene un equilibrio entre rendimiento y consumo energético, destacando con Keras. En general, el uso de Keras tiende a disminuir el consumo por hora, pero aumenta el tiempo de entrenamiento, especialmente en PyTorch. TensorFlow, en general, presenta los tiempos más altos, aunque con un consumo energético por hora menor en MLP y CNN. Estos hallazgos destacan la influencia del framework en la eficiencia del entrenamiento, evidenciando la necesidad de elegir el más adecuado según el tipo de red y los recursos disponibles.

El presente trabajo ha sido financiado en parte por el proyecto NATASHA (PID2022-142292NB-I00), del Ministerio de Ciencia, Innovación y Universidades de España.

Producción Científica

Country
Spain
Related Organizations
Keywords

Informática, 3304 Tecnología de Los Ordenadores, Deep Learning, framework, Tensor Flow, PyTorch, JAX, Keras, eficiencia, GPU., 1203 Ciencia de Los Ordenadores

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Green