Experimenting with Emerging RISC-V Systems for Decentralised Machine Learning
Decentralised Machine Learning (DML) enables collaborative ma- chine learning without centralised input data. Federated Learning (FL) and Edge Inference are examples of DML. While tools for DML (especially FL) are starting to flourish, many are not flexible and portable enough to experiment with nov...
Published in: | Proceedings of the 20th ACM International Conference on Computing Frontiers |
---|---|
Main Authors: | , , , , , , , , , , , , |
Other Authors: | |
Format: | Conference Object |
Language: | English |
Published: |
ACM
2023
|
Subjects: | |
Online Access: | https://hdl.handle.net/2318/1898473 https://doi.org/10.1145/3587135.3592211 https://arxiv.org/abs/2302.07946 |
id |
ftunivtorino:oai:iris.unito.it:2318/1898473 |
---|---|
record_format |
openpolar |
spelling |
ftunivtorino:oai:iris.unito.it:2318/1898473 2023-11-05T03:41:35+01:00 Experimenting with Emerging RISC-V Systems for Decentralised Machine Learning Gianluca Mittone Nicolò Tonci Robert Birke Iacopo Colonnelli Doriana Medic Andrea Bartolini Roberto Esposito Emanuele Parisi Francesco Beneventi Mirko Polato Massimo Torquati Luca Benini Marco Aldinucci Gianluca Mittone , Nicolò Tonci , Robert Birke , Iacopo Colonnelli , Doriana Medic , Andrea Bartolini , Roberto Esposito , Emanuele Parisi , Francesco Beneventi , Mirko Polato , Massimo Torquati , Luca Benini , Marco Aldinucci 2023 https://hdl.handle.net/2318/1898473 https://doi.org/10.1145/3587135.3592211 https://arxiv.org/abs/2302.07946 eng eng ACM country:USA place:New York City ispartofbook:Proceedings of the 20th ACM International Conference on Computing Frontiers The 20th ACM International Conference on Computing Frontiers firstpage:- lastpage:- numberofpages:- https://hdl.handle.net/2318/1898473 doi:10.1145/3587135.3592211 https://arxiv.org/abs/2302.07946 info:eu-repo/semantics/openAccess info:eu-repo/semantics/conferenceObject 2023 ftunivtorino https://doi.org/10.1145/3587135.3592211 2023-10-10T22:15:26Z Decentralised Machine Learning (DML) enables collaborative ma- chine learning without centralised input data. Federated Learning (FL) and Edge Inference are examples of DML. While tools for DML (especially FL) are starting to flourish, many are not flexible and portable enough to experiment with novel processors (e.g., RISC-V), non-fully connected network topologies, and asynchro- nous collaboration schemes. We overcome these limitations via a domain-specific language allowing us to map DML schemes to an underlying middleware, i.e. the FastFlow parallel programming library. We experiment with it by generating different working DML schemes on x86-64 and ARM platforms and an emerging RISC-V one. We characterise the performance and energy efficiency of the presented schemes and systems. As a byproduct, we intro- duce a RISC-V porting of the PyTorch framework, the first publicly available to our knowledge. Conference Object DML Università degli studi di Torino: AperTo (Archivio Istituzionale ad Accesso Aperto) Proceedings of the 20th ACM International Conference on Computing Frontiers 73 83 |
institution |
Open Polar |
collection |
Università degli studi di Torino: AperTo (Archivio Istituzionale ad Accesso Aperto) |
op_collection_id |
ftunivtorino |
language |
English |
description |
Decentralised Machine Learning (DML) enables collaborative ma- chine learning without centralised input data. Federated Learning (FL) and Edge Inference are examples of DML. While tools for DML (especially FL) are starting to flourish, many are not flexible and portable enough to experiment with novel processors (e.g., RISC-V), non-fully connected network topologies, and asynchro- nous collaboration schemes. We overcome these limitations via a domain-specific language allowing us to map DML schemes to an underlying middleware, i.e. the FastFlow parallel programming library. We experiment with it by generating different working DML schemes on x86-64 and ARM platforms and an emerging RISC-V one. We characterise the performance and energy efficiency of the presented schemes and systems. As a byproduct, we intro- duce a RISC-V porting of the PyTorch framework, the first publicly available to our knowledge. |
author2 |
Gianluca Mittone , Nicolò Tonci , Robert Birke , Iacopo Colonnelli , Doriana Medic , Andrea Bartolini , Roberto Esposito , Emanuele Parisi , Francesco Beneventi , Mirko Polato , Massimo Torquati , Luca Benini , Marco Aldinucci |
format |
Conference Object |
author |
Gianluca Mittone Nicolò Tonci Robert Birke Iacopo Colonnelli Doriana Medic Andrea Bartolini Roberto Esposito Emanuele Parisi Francesco Beneventi Mirko Polato Massimo Torquati Luca Benini Marco Aldinucci |
spellingShingle |
Gianluca Mittone Nicolò Tonci Robert Birke Iacopo Colonnelli Doriana Medic Andrea Bartolini Roberto Esposito Emanuele Parisi Francesco Beneventi Mirko Polato Massimo Torquati Luca Benini Marco Aldinucci Experimenting with Emerging RISC-V Systems for Decentralised Machine Learning |
author_facet |
Gianluca Mittone Nicolò Tonci Robert Birke Iacopo Colonnelli Doriana Medic Andrea Bartolini Roberto Esposito Emanuele Parisi Francesco Beneventi Mirko Polato Massimo Torquati Luca Benini Marco Aldinucci |
author_sort |
Gianluca Mittone |
title |
Experimenting with Emerging RISC-V Systems for Decentralised Machine Learning |
title_short |
Experimenting with Emerging RISC-V Systems for Decentralised Machine Learning |
title_full |
Experimenting with Emerging RISC-V Systems for Decentralised Machine Learning |
title_fullStr |
Experimenting with Emerging RISC-V Systems for Decentralised Machine Learning |
title_full_unstemmed |
Experimenting with Emerging RISC-V Systems for Decentralised Machine Learning |
title_sort |
experimenting with emerging risc-v systems for decentralised machine learning |
publisher |
ACM |
publishDate |
2023 |
url |
https://hdl.handle.net/2318/1898473 https://doi.org/10.1145/3587135.3592211 https://arxiv.org/abs/2302.07946 |
genre |
DML |
genre_facet |
DML |
op_relation |
ispartofbook:Proceedings of the 20th ACM International Conference on Computing Frontiers The 20th ACM International Conference on Computing Frontiers firstpage:- lastpage:- numberofpages:- https://hdl.handle.net/2318/1898473 doi:10.1145/3587135.3592211 https://arxiv.org/abs/2302.07946 |
op_rights |
info:eu-repo/semantics/openAccess |
op_doi |
https://doi.org/10.1145/3587135.3592211 |
container_title |
Proceedings of the 20th ACM International Conference on Computing Frontiers |
container_start_page |
73 |
op_container_end_page |
83 |
_version_ |
1781698025342631936 |