Accurate yet efficient Deep Neural Networks (DNNs) are in high demand, especially for applications that require their execution on constrained edge devices. Finding such DNNs in a reasonable time for new applications requires automated optimization pipelines since the huge space of hyper-parameter combinations is impossible to explore extensively by hand. In this work, we propose PLiNIO, an open-source library implementing a comprehensive set of state-of-the-art DNN design automation techniques, all based on lightweight gradient-based optimization, under a unified and user-friendly interface. With experiments on several edge-relevant tasks, we show that combining the various optimizations available in PLiNIO leads to rich sets of solutions that Pareto-dominate the considered baselines in terms of accuracy vs model size. Noteworthy, PLiNIO achieves up to 94.34% memory reduction for a <1% accuracy drop compared to a baseline architecture.

PLiNIO: A User-Friendly Library of Gradient-based Methods for Complexity-aware DNN Optimization / Jahier Pagliari, Daniele; Risso, Matteo; Motetti, Beatrice; Burrello, Alessio. - ELETTRONICO. - (2023), pp. 1-8. (Intervento presentato al convegno Forum for Specification and Design Languages (FDL) tenutosi a Turin (Italy) nel September 13-15, 2023) [10.1109/FDL59689.2023.10272045].

PLiNIO: A User-Friendly Library of Gradient-based Methods for Complexity-aware DNN Optimization

Jahier Pagliari, Daniele;Risso, Matteo;Motetti, Beatrice;Burrello, Alessio
2023

Abstract

Accurate yet efficient Deep Neural Networks (DNNs) are in high demand, especially for applications that require their execution on constrained edge devices. Finding such DNNs in a reasonable time for new applications requires automated optimization pipelines since the huge space of hyper-parameter combinations is impossible to explore extensively by hand. In this work, we propose PLiNIO, an open-source library implementing a comprehensive set of state-of-the-art DNN design automation techniques, all based on lightweight gradient-based optimization, under a unified and user-friendly interface. With experiments on several edge-relevant tasks, we show that combining the various optimizations available in PLiNIO leads to rich sets of solutions that Pareto-dominate the considered baselines in terms of accuracy vs model size. Noteworthy, PLiNIO achieves up to 94.34% memory reduction for a <1% accuracy drop compared to a baseline architecture.
File in questo prodotto:
File Dimensione Formato  
2307.09488.pdf

accesso aperto

Tipologia: 2. Post-print / Author's Accepted Manuscript
Licenza: PUBBLICO - Tutti i diritti riservati
Dimensione 1.94 MB
Formato Adobe PDF
1.94 MB Adobe PDF Visualizza/Apri
PLiNIO_A_User-Friendly_Library_of_Gradient-Based_Methods_for_Complexity-Aware_DNN_Optimization.pdf

non disponibili

Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: Non Pubblico - Accesso privato/ristretto
Dimensione 2.14 MB
Formato Adobe PDF
2.14 MB Adobe PDF   Visualizza/Apri   Richiedi una copia
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/2982474