Search

Your search keyword '"Dubois, Lionel"' showing total 234 results
Start Over Author "Dubois, Lionel" Publication Year Range Last 50 years Language english
234 results on '"Dubois, Lionel"'

5. Towards EXtreme scale technologies and accelerators for euROhpc hw/Sw supercomputing applications for exascale: The TEXTAROSSA approach

Author

Agosta, Giovanni, Aldinucci, Marco, Alvarez, Carlos, Ammendola, Roberto, Arfat, Yasir, Beaumont, Olivier, Bernaschi, Massimo, Biagioni, Andrea, Boccali, Tommaso, Bramas, Berenger, Brandolese, Carlo, Cantalupo, Barbara, Carrozzo, Mauro, Cattaneo, Daniele, Celestini, Alessandro, Celino, Massimo, Colonnelli, Iacopo, Cretaro, Paolo, D’Ambra, Pasqua, Danelutto, Marco, Esposito, Roberto, Eyraud-Dubois, Lionel, Filgueras, Antonio, Fornaciari, William, Frezza, Ottorino, Galimberti, Andrea, Giacomini, Francesco, Goglin, Brice, Gregori, Daniele, Guermouche, Abdou, Iannone, Francesco, Kulczewski, Michal, Lo Cicero, Francesca, Lonardo, Alessandro, Martinelli, Alberto R., Martinelli, Michele, Martorell, Xavier, Massari, Giuseppe, Montangero, Simone, Mittone, Gianluca, Namyst, Raymond, Oleksiak, Ariel, Palazzari, Paolo, Paolucci, Pier Stanislao, Reghenzani, Federico, Rossi, Cristian, Saponara, Sergio, Simula, Francesco, Terraneo, Federico, Thibault, Samuel, Torquati, Massimo, Turisini, Matteo, Vicini, Piero, Vidal, Miquel, Zoni, Davide, and Zummo, Giuseppe

Published
2022
Full Text
View/download PDF

7. Optimal Re-Materialization Strategies for Heterogeneous Chains: How to Train Deep Neural Networks with Limited Memory.

Author

BEAUMONT, OLIVIER, EYRAUD-DUBOIS, LIONEL, HERRMANN, JULIEN, JOLY, ALEXIS, and SHILOVA, ALENA

Subjects
*ARTIFICIAL neural networks, *AUTOMATIC differentiation, *MEMORY, *DATA warehousing
Abstract

Training in Feed Forward Deep Neural Networks is a memory-intensive operation which is usually performed on GPUs with limited memory capacities. This may force data scientists to limit the depth of the models or the resolution of the input data if data does not fit in the GPU memory. The re-materialization technique, whose idea comes from the checkpointing strategies developed in the Automatic Differentiation literature, allows data scientists to limit the memory requirements related to the storage of intermediate data (activations), at the cost of an increase in the computational cost. This paper introduces a new strategy of re-materialization of activations that significantly reduces memory usage. It consists in selecting which activations are saved and which activations are deleted during the forward phase and then recomputing the deleted activations when they are needed during the backward phase. We propose an original computation model that combines two types of activation savings: either only storing the layer inputs or recording the complete history of operations that produced the outputs. This paper focuses on the fully heterogeneous case, where the computation time and the memory requirement of each layer is different. We prove that finding the optimal solution is NP-hard and that classical techniques from Automatic Differentiation literature do not apply. Moreover, the classical assumption of memory persistence of materialized activations, used to simplify the search of optimal solutions, does not hold anymore. Thus, we propose a weak memory persistence property and provide a dynamic program to compute the optimal sequence of computations. This algorithm is made available through the Rotor software, a PyTorch plug-in dealing with any network consisting of a sequence of layers, each of them having an arbitrarily complex structure. Through extensive experiments, we show that our implementation consistently outperforms existing re-materialization approaches for a large class of networks, image sizes, and batch sizes. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

10. Study of pathways to reduce the energy consumption of the CO2 capture process by absorption-regeneration

Author

Dubois Lionel, Costa Alexis, De Weireld Guy, and Thomas Diane

Subjects
capture du co2, absorption-régénération, solvants aminés, simulations, procédés innovants, Engineering (General). Civil engineering (General), TA1-2040
Abstract

Several industrial sectors, such as for example cement manufacturers and lime producers, produce so-called “unavoidable” CO2 emissions because these ones are intrinsically linked to the industrial process itself (decarbonation of calcium carbonate). In order to reduce these emissions, it is necessary to implement a Carbon Capture, Utilization and/or Storage (CCUS) process chain, whose step of capture, although already technologically mature (especially the absorption-regeneration process using amine(s)-based solvents), leads to very high energy consumption. Three pathways to reduce this consumption have been investigated (experimentally and/or through the development of Aspen PlusTM simulations), namely: (i) upstream of the process thanks to the increase of the flue gas CO2 content (by partial oxy-combustion and/or flue gas recirculation), (ii) within the process (using more efficient and innovative mixtures of solvents such as demixing solutions), and (iii) at the configurational level by using advanced configurations in the capture process. It emerged that the use of a demixing process such as the mixture composed of diethylethanolamine (DEEA) and methylamino-propylamine (MAPA), or the implementation of an advanced process configuration (InterCooling Absorber + Rich Vapor Compression + Rich Solvent Splitting and Preheating, with methyldiethanolamine (MDEA) + piperazine (PZ) as a solvent) are the most energy reducing pathways for the absorption-regeneration process, i.e. more than 40% in comparison with a conventional process using monoethanolamine (MEA). Moreover, from an economical point of view, and compared to a basic configuration with MEA, the demixing technology has the advantage of being able to achieve such energy performance with a more limited investment (CAPEX) (+1.6%) than with advanced process configurations (+8.8%).

Published
2023
Full Text
View/download PDF

23. Lithium isotope tracing in silicon-based electrodes using solid-state MAS NMR: a powerful comprehensive tool for the characterization of lithium batteries.

Author

Berthault, Manon, Buzlukov, Anton, Dubois, Lionel, Bayle, Pierre-Alain, Porcher, Willy, Gutel, Thibaut, De Vito, Eric, and Bardet, Michel

Abstract

The introduction of lithiated components with different 7Li/6Li isotopic ratios, also called isotopic tracing, can give access to better understanding of lithium transport and lithiation processes in lithium-ion batteries. In this work, we propose a simple methodology based on high-resolution solid-state NMR for the determination of the 7Li/6Li ratio in silicon electrodes following different strategies of isotopic tracing. The 6Li and 7Li MAS NMR experiments allow obtaining resolved spectra whose spectral components can be assigned to different moieties of the materials. In order to measure the ratio of the 6Li/7Li NMR integrals, a silicon electrode with a natural 7Li/6Li isotope abundance was used as a reference. This calibration can then be used to determine the 7Li/6Li ratio of any similar samples. This method was applied to study the phenomena occurring at the interface between a silicon electrode and a labeled electrolyte, which is an essential step for isotopic tracing experiments in systems after the formation of the solid electrolyte interphase (SEI). Beyond the isotopic exchanges between the SEI and the electrolyte already observed in the literature, our results show that isotopic exchanges also involve Li–Si alloys in the electrode bulk. Within a 52-hour contact, the electrolyte labeling disappeared: isotopic concentrations of the electrolyte and electrode become practically homogenized. However, at the electrode level, different silicides are characterized by rather different isotopic enrichment. In the present work, ToF SIMS and liquid state NMR were also used to cross-check and discuss the solid-state NMR method we have proposed. [ABSTRACT FROM AUTHOR]

Published
2023
Full Text
View/download PDF

24. On the Arithmetic Intensity of Distributed-Memory Dense Matrix Multiplication Involving a Symmetric Input Matrix (SYMM)

Author

Agullo, Emmanuel, Buttari, Alfredo, Coulaud, Olivier, Eyraud-Dubois, Lionel, Faverge, Mathieu, Franc, Alain, Guermouche, Abdou, Jego, Antoine, Peressoni, Romain, Pruvost, Florent, COmposabilité Numerique and parallèle pour le CAlcul haute performanCE (CONCACE), Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique (CERFACS)-Airbus [France]-Inria Bordeaux - Sud-Ouest, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria), Algorithmes Parallèles et Optimisation (IRIT-APO), Institut de recherche en informatique de Toulouse (IRIT), Université Toulouse Capitole (UT Capitole), Université de Toulouse (UT)-Université de Toulouse (UT)-Université Toulouse - Jean Jaurès (UT2J), Université de Toulouse (UT)-Université Toulouse III - Paul Sabatier (UT3), Université de Toulouse (UT)-Centre National de la Recherche Scientifique (CNRS)-Institut National Polytechnique (Toulouse) (Toulouse INP), Université de Toulouse (UT)-Toulouse Mind & Brain Institut (TMBI), Université Toulouse - Jean Jaurès (UT2J), Université de Toulouse (UT)-Université de Toulouse (UT)-Université Toulouse III - Paul Sabatier (UT3), Université de Toulouse (UT)-Université Toulouse Capitole (UT Capitole), Université de Toulouse (UT), Centre National de la Recherche Scientifique (CNRS), Laboratoire Bordelais de Recherche en Informatique (LaBRI), Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS), Outils et Optimisations pour le Calcul Haute Performance et l'Apprentissage (TOPAL), Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Inria Bordeaux - Sud-Ouest, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Institut Polytechnique de Bordeaux (Bordeaux INP), Biodiversité, Gènes & Communautés (BioGeCo), Université de Bordeaux (UB)-Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement (INRAE), from patterns to models in computational biodiversity and biotechnology (PLEIADE), Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Biodiversité, Gènes & Communautés (BioGeCo), Université de Bordeaux (UB)-Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement (INRAE)-Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement (INRAE), Service Expérimentation et Développement [Bordeaux] (SED), Inria Bordeaux - Sud-Ouest, Région Nouvelle-Aquitaine (2018-1R50119 HPC scalable ecosystem), IEEE, and ANR-19-CE46-0009,SOLHARIS,Solveurs pour architectures hétérogènes utilisant des supports d'exécution, objectif scalabilité(2019)

Subjects
Matrix multiplication, Symmetric, GEMM, 2DBC, SYMM, task-based programming, TBC, [INFO.INFO-DC]Computer Science [cs]/Distributed, Parallel, and Cluster Computing [cs.DC], [INFO.INFO-BI]Computer Science [cs]/Bioinformatics [q-bio.QM], 2.5D, SBC, 3D
Abstract

International audience; Dense matrix multiplication involving a symmetric input matrix (SYMM) is implemented in reference distributed-memory codes with the same data distribution as its general analogue (GEMM). We show that, when the symmetric matrix is dominant, such a 2D block-cyclic (2D BC) scheme leads to a lower arithmetic intensity (AI) of SYMM than that of GEMM by a factor of 2. We propose alternative data distributions preserving the memory benefit of SYMM of storing only half of the matrix while achieving up to the same AI as GEMM. We also show that, in the case we can afford the same memory footprint as GEMM, SYMM can achieve a higher AI. We propose a task-based design of SYMM independent of the data distribution. This design allows for scalable A-stationary SYMM with which all discussed data distributions, may they be very irregular, can be easily assessed. We have integrated the resulting code in a reduction dimension algorithm involving a randomized singular value decomposition dominated by SYMM. An experimental study shows a compelling impact on performance.

Published
2023

25. A 3D-printed device for in situ monitoring of an organic redox-flow battery via NMR/MRI

Author

Caja-Munoz, Borja, Chighine, Kévin, Dognon, Jean-Pierre, Dubois, Lionel, Berthault, Patrick, Laboratoire Structure et Dynamique par Résonance Magnétique (LCF) (LSDRM), Nanosciences et Innovation pour les Matériaux, la Biomédecine et l'Energie (ex SIS2M) (NIMBE UMR 3685), Institut Rayonnement Matière de Saclay (IRAMIS), Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Université Paris-Saclay-Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Université Paris-Saclay-Institut de Chimie du CNRS (INC)-Centre National de la Recherche Scientifique (CNRS)-Institut Rayonnement Matière de Saclay (IRAMIS), Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Université Paris-Saclay-Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Université Paris-Saclay-Institut de Chimie du CNRS (INC)-Centre National de la Recherche Scientifique (CNRS), Conception d’Architectures Moléculaires et Processus Electroniques (CAMPE ), SYstèmes Moléculaires et nanoMatériaux pour l’Energie et la Santé (SYMMES), Institut de Chimie du CNRS (INC)-Centre National de la Recherche Scientifique (CNRS)-Institut de Recherche Interdisciplinaire de Grenoble (IRIG), Direction de Recherche Fondamentale (CEA) (DRF (CEA)), Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Direction de Recherche Fondamentale (CEA) (DRF (CEA)), Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Université Grenoble Alpes (UGA)-Institut de Chimie du CNRS (INC)-Centre National de la Recherche Scientifique (CNRS)-Institut de Recherche Interdisciplinaire de Grenoble (IRIG), Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Université Grenoble Alpes (UGA), Programmes Transverses de Compétence CEA (Projet ROBO3D), and ANR-17-LCV2-0002,DESIR,Détection Efficace et Sensible d'Intermédiaires Réactionnels par RMN(2017)

Subjects
Redox-flow, [SPI]Engineering Sciences [physics], Battery Redox-flow NMR MRI Operando 3D printing Anthraquinone, Battery, [CHIM]Chemical Sciences, Anthraquinone, Operando, 3D printing, NMR, MRI
Abstract

International audience; A mini organic redox-flow battery pluggable on the basis of a high-resolution NMR probehead has been conceived and built mainly by 3D printing. This device allows the realization of all modern spectroscopy experiments as well as imaging experiments. It has been tested for the real-time monitoring of redox cycling of 9,10-Anthraquinone-2,7-disulfonic acid disodium salt (2,7-AQDS) in acidic conditions, which has revealed the preponderant role of dimerization in the processes of oxidation and reduction. Determination of the thermodynamic properties of homo-and heterodimer formation through quantum chemical, multilevel modeling workflows confirm our hypotheses about the molecular processes occurring during charge and discharge.

Published
2023

26. Symmetric Block-Cyclic Distribution: Fewer Communications Leads to Faster Dense Cholesky Factorization

Author

Beaumont, Olivier, Duchon, Philippe, Eyraud-Dubois, Lionel, Langou, Julien, Vérité, Mathieu, High-End Parallel Algorithms for Challenging Numerical Simulations (HiePACS), Laboratoire Bordelais de Recherche en Informatique (LaBRI), Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Inria Bordeaux - Sud-Ouest, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria), Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS), University of Colorado [Denver], Projet Région Nouvelle-Aquitaine 2018-1R50119 'HPC scalable ecosystem'., ANR-19-CE46-0009,SOLHARIS,Solveurs pour architectures hétérogènes utilisant des supports d'exécution, objectif scalabilité(2019), European Project: 956831,TEXTAROSSA(2021), Eyraud-Dubois, Lionel, Solveurs pour architectures hétérogènes utilisant des supports d'exécution, objectif scalabilité - - SOLHARIS2019 - ANR-19-CE46-0009 - AAPG2019 - VALID, and Towards EXtreme scale Technologies and Accelerators for euROhpc hw/Sw Supercomputing Applications for exascale - TEXTAROSSA - 2021-01-01 - 2024-01-01 - 956831 - VALID

Subjects
[INFO.INFO-DC] Computer Science [cs]/Distributed, Parallel, and Cluster Computing [cs.DC], [INFO.INFO-DC]Computer Science [cs]/Distributed, Parallel, and Cluster Computing [cs.DC]
Abstract

International audience; We consider the distributed Cholesky factorization on homogeneous nodes. Inspired by recent progress on asymptotic lower bounds on the total communication volume required to perform Cholesky factorization, we present an original data distribution, Symmetric Block Cyclic (SBC), designed to take advantage of the symmetry of the matrix. We prove that SBC reduces the overall communication volume between nodes by a factor of square root of 2 compared to the standard 2D blockcyclic distribution. SBC can easily be implemented within the paradigm of task-based runtime systems. Experiments using the Chameleon library over the StarPU runtime system demonstrate that the SBC distribution reduces the communication volume as expected, and also achieves better performance and scalability than the classical 2D block-cyclic allocation scheme in all configurations. We also propose a 2.5D variant of SBC and prove that it further improves the communication and performance benefits.

Published
2022

28. Energy, exergy, economic and environmental (4E) analysis of integrated direct air capture and CO2 methanation under uncertainty

Author

Coppitters, Diederik, Costa, Alexis, Chauvy, Remi, Dubois, Lionel, De Paepe, Ward, Thomas, Diane, De Weireld, Guy, Contino, Francesco, and UCL - SST/IMMC/TFL - Thermodynamics and fluid mechanics

Abstract

Direct Air Capture (DAC) technologies are gaining interest in the concept of carbon utilization and Power-to-Gas (PtG), as the economic valorization of the CO into methane provides a viable pathway to allow DAC systems to mature. However, research on DAC mainly focuses on isolated systems, and the system performance depends on parameters that are highly uncertain. To study the integration of DAC in PtG, we developed a DAC-PtG model, performed an Energy, Exergy, Economic and Environmental (4E) analysis and implemented uncertainty quantification to consider the uncertain environment. The results illustrate that the DAC-PtG system is autothermal when introducing a two-stage mechanical vapor recompression unit at the DAC outlet. The exergy efficiency ranges between 51.3% and 52.6% within 3 standard deviations, for which the uncertainty is driven by the ambient conditions and the uncertain heat of desorption. The methane issued from DAC-PtG has a lower carbon footprint than fossil methane when the carbon footprint of the electricity supply is below or equal to 0.12 kg˙CO˙2-eq /kWh. The Levelized Cost of Synthetic Natural Gas (LCSNG) ranges between 130 €/ MWh and 744 €/ MWh, following an uncertain electricity price and uncertain expenses related to DAC and electrolysis. Therefore, bulk manufacturing, further maturing of these technologies and more demonstration projects are required to reduce the uncertainty of the LCSNG. Future works will consider intermittent renewable energy sources to supply power.

Published
2023

31. Weight Offloading Strategies for Training Large DNN Models

Author

Beaumont, Olivier, Eyraud-Dubois, Lionel, Shilova, Alena, Zhao, Xunyi, Eyraud-Dubois, Lionel, High-End Parallel Algorithms for Challenging Numerical Simulations (HiePACS), Laboratoire Bordelais de Recherche en Informatique (LaBRI), Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Inria Bordeaux - Sud-Ouest, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria), Scool (Scool), Inria Lille - Nord Europe, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 (CRIStAL), and Centrale Lille-Université de Lille-Centre National de la Recherche Scientifique (CNRS)-Centrale Lille-Université de Lille-Centre National de la Recherche Scientifique (CNRS)

Subjects
Scheduling, Offloading Strategies, [INFO.INFO-DC] Computer Science [cs]/Distributed, Parallel, and Cluster Computing [cs.DC], Training of DNNs, [INFO.INFO-DC]Computer Science [cs]/Distributed, Parallel, and Cluster Computing [cs.DC]
Abstract

The limited memory of GPUs induces serious problems in the training phase of deep neural networks (DNNs). Indeed, with the recent tremendous increase in the size of DNN models, which can now routinely include hundreds of billions or even trillions of parameters, it is impossible to store these models in the memory of a GPU and several strategies have been devised to solve this problem. In this paper, we analyze in detail the strategy that consists in offloading the weights of some model layers from the GPU to the CPU when they are not used. Since the PCI bus bandwidth between the GPU and the CPU is limited, it is crucial to know which layers should be transferred (offloaded and prefetched) and when. We prove that this problem is in general NP-Complete in the strong sense and we propose a lower bound formulation in the form of an Integer Linear Program (ILP). We propose heuristics to select the layers to offload and to build the schedule of data transfers. We show that this approach allows to build near-optimal weight offloading strategies on realistic size DNNs and architectures.

Published
2022

32. A complete biomimetic iron-sulfur cubane redox series

Author

Grunwald, Liam, Clémancey, Martin, Klose, Daniel, Dubois, Lionel, Gambarelli, Serge, Jeschke, Gunnar, Wörle, Michael, Blondin, Geneviève, Mougel, Victor, Department of Chemistry and Applied Biosciences [ETH Zürich] (D-CHAB), Eidgenössische Technische Hochschule - Swiss Federal Institute of Technology [Zürich] (ETH Zürich), Physiochimie des Métaux (PMB), Laboratoire de Chimie et Biologie des Métaux (LCBM - UMR 5249), Institut de Chimie du CNRS (INC)-Centre National de la Recherche Scientifique (CNRS)-Institut de Recherche Interdisciplinaire de Grenoble (IRIG), Direction de Recherche Fondamentale (CEA) (DRF (CEA)), Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Direction de Recherche Fondamentale (CEA) (DRF (CEA)), Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Université Grenoble Alpes (UGA)-Institut de Chimie du CNRS (INC)-Centre National de la Recherche Scientifique (CNRS)-Institut de Recherche Interdisciplinaire de Grenoble (IRIG), Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Université Grenoble Alpes (UGA), Conception d’Architectures Moléculaires et Processus Electroniques (CAMPE ), and SYstèmes Moléculaires et nanoMatériaux pour l’Energie et la Santé (SYMMES)

Subjects
Multidisciplinary, Mössbauer spectroscopy, Iron, Coenzymes, [SDV.BBM.BM]Life Sciences [q-bio]/Biochemistry, Molecular Biology/Molecular biology, all-ferrous cubane, [CHIM.MATE]Chemical Sciences/Material chemistry, iron-sulfur clusters, Hydrocarbons, electrochemistry, Biomimetic Materials, nitrogenase, Nitrogenase, [CHIM]Chemical Sciences, [PHYS.PHYS.PHYS-CHEM-PH]Physics [physics]/Physics [physics]/Chemical Physics [physics.chem-ph], Oxidation-Reduction, Sulfur
Abstract

Synthetic iron-sulfur cubanes are models for biological cofactors, which are essential to delineate oxidation states in the more complex enzymatic systems. However, a complete series of [Fe4S4]n complexes spanning all redox states accessible by 1-electron transformations of the individual iron atoms (n = 0-4+) has never been prepared, deterring the methodical comparison of structure and spectroscopic signature. Here, we demonstrate that the use of a bulky arylthiolate ligand promoting the encapsulation of alkali-metal cations in the vicinity of the cubane enables the synthesis of such a series. Characterization by EPR, 57Fe Mössbauer spectroscopy, UV-visible electronic absorption, variable-temperature X-ray diffraction analysis, and cyclic voltammetry reveals key trends for the geometry of the Fe4S4 core as well as for the Mössbauer isomer shift, which both correlate systematically with oxidation state. Furthermore, we confirm the S = 4 electronic ground state of the most reduced member of the series, [Fe4S4]0, and provide electrochemical evidence that it is accessible within 0.82 V from the [Fe4S4]2+ state, highlighting its relevance as a mimic of the nitrogenase iron protein cluster., Proceedings of the National Academy of Sciences of the United States of America, 119 (31), ISSN:0027-8424, ISSN:1091-6490

Published
2022

33. Survey on Large Scale Neural Network Training

Author

Gusak, Julia, Cherniuk, Daria, Shilova, Alena, Katrutsa, Alexandr, Bershatsky, Daniel, Zhao, Xunyi, Eyraud-Dubois, Lionel, Shliazhko, Oleh, Dimitrov, Denis, Oseledets, Ivan, Beaumont, Olivier, Skolkovo Institute of Science and Technology [Moscow] (Skoltech), Scool (Scool), Inria Lille - Nord Europe, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 (CRIStAL), Centrale Lille-Université de Lille-Centre National de la Recherche Scientifique (CNRS)-Centrale Lille-Université de Lille-Centre National de la Recherche Scientifique (CNRS), High-End Parallel Algorithms for Challenging Numerical Simulations (HiePACS), Laboratoire Bordelais de Recherche en Informatique (LaBRI), Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Inria Bordeaux - Sud-Ouest, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria), Auteur indépendant, Artificial Intelligence Research Institute (AIRI), Skolkovo Institute of Science and Technology [Moscow] [Skoltech], Scool [Scool], High-End Parallel Algorithms for Challenging Numerical Simulations [HiePACS], and Artificial Intelligence Research Institute [AIRI]

Subjects
FOS: Computer and information sciences, Computer Science - Machine Learning, Artificial Intelligence (cs.AI), Computer Science - Artificial Intelligence, Machine Learning (cs.LG), [INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI]
Abstract

International audience; Modern Deep Neural Networks (DNNs) require significant memory to store weight, activations, and other intermediate tensors during training. Hence, many models don't fit one GPU device or can be trained using only a small per-GPU batch size. This survey provides a systematic overview of the approaches that enable more efficient DNNs training. We analyze techniques that save memory and make good use of computation and communication resources on architectures with a single or several GPUs. We summarize the main categories of strategies and compare strategies within and across categories. Along with approaches proposed in the literature, we discuss available implementations.

Published
2022

39. Carbon Capture Performance Assessment Applied to Combined Cycle Gas Turbine Under Part-Load Operation.

Author

Verhaeghe, Antoine, Dubois, Lionel, Bricteux, Laurent, Thomas, Diane, Blondeau, Julien, and De Paepe, Ward

Abstract

The growing share of renewable energies in our electricity production, together with the still lacking storage capacity, strongly reinforces the need for more flexible electricity production units. In this context, combined cycle gas turbines (CCGTs) have a role to play, both in the current and future electricity production system due to their high efficiency, high load flexibility, and low CO2 emissions compared to other conventional thermal power plants. Nevertheless, bearing in mind our current challenges concerning climate change, the CO2 emissions of these CCGTs need to be reduced drastically. The amine-based absorption carbon capture (CC) process is currently the most mature and applicable CC technology. This process is known to require a considerable amount of thermal energy, degrading plant performance. However, to back-up renewable production, CCGTs will operate most of the time under part-load conditions. The impact of these part-load operations on the CC is still relatively unknown. Within this framework, this study aims to assess the performance of the CC process applied to a typical CCGT under part-load operation using specific simulation models. The CC plant model has been successfully validated against experimental data from a pilot-scale capture facility. Then, the CC plant has been scaled-up to the CCGT scale and the process has been optimized for each operating condition. The simulation results show that the specific reboiler duty increases for part-load operation, while the specific cooling requirements decrease. Moreover, the analysis of the yearly CCGT operation highlights a relative increase in CC energy penalty of 21% for an annual CCGT load factor of 0.5, impacting significantly plant performance. The next step will involve reducing this energy penalty. [ABSTRACT FROM AUTHOR]

Published
2023
Full Text
View/download PDF

40. Toward the Improvement of Silicon-Based Composite Electrodes via an In-Situ Si@C-Graphene Composite Synthesis for Li-Ion Battery Applications.

Author

Mery, Adrien, Chenavier, Yves, Marcucci, Coralie, Benayad, Anass, Alper, John P., Dubois, Lionel, Haon, Cédric, Boime, Nathalie Herlin, Sadki, Saïd, and Duclairoir, Florence

Subjects
LITHIUM-ion batteries, HYBRID materials, ELECTRODES, GRAPHENE, HYDROGELS, GREENHOUSE gases
Abstract

Using Si as anode materials for Li-ion batteries remain challenging due to its morphological evolution and SEI modification upon cycling. The present work aims at developing a composite consisting of carbon-coated Si nanoparticles (Si@C NPs) intimately embedded in a three-dimensional (3D) graphene hydrogel (GHG) architecture to stabilize Si inside LiB electrodes. Instead of simply mixing both components, the novelty of the synthesis procedure lies in the in situ hydrothermal process, which was shown to successfully yield graphene oxide reduction, 3D graphene assembly production, and homogeneous distribution of Si@C NPs in the GHG matrix. Electrochemical characterizations in half-cells, on electrodes not containing additional conductive additive, revealed the importance of the protective C shell to achieve high specific capacity (up to 2200 mAh.g−1), along with good stability (200 cycles with an average Ceff > 99%). These performances are far superior to that of electrodes made with non-C-coated Si NPs or prepared by mixing both components. These observations highlight the synergetic effects of C shell on Si NPs, and of the single-step in situ preparation that enables the yield of a Si@C-GHG hybrid composite with physicochemical, structural, and morphological properties promoting sample conductivity and Li-ion diffusion pathways. [ABSTRACT FROM AUTHOR]

Published
2023
Full Text
View/download PDF

43. An Integer Linear Programming Approach for Pipelined Model Parallelism

Author

Beaumont, Olivier, Eyraud-Dubois, Lionel, Shilova, Alena, Beaumont, Olivier, High-End Parallel Algorithms for Challenging Numerical Simulations (HiePACS), Laboratoire Bordelais de Recherche en Informatique (LaBRI), Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Inria Bordeaux - Sud-Ouest, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria), Scool (Scool), Inria Lille - Nord Europe, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 (CRIStAL), Centrale Lille-Université de Lille-Centre National de la Recherche Scientifique (CNRS)-Centrale Lille-Université de Lille-Centre National de la Recherche Scientifique (CNRS), and Inria

Subjects
Integer Linear Programming, [INFO.INFO-AI] Computer Science [cs]/Artificial Intelligence [cs.AI], Model Parallelism, Memory, [INFO.INFO-DC] Computer Science [cs]/Distributed, Parallel, and Cluster Computing [cs.DC], Training, Parallélisme de modèle, Programmation linéaire en nombres entiers, Mémoire, [INFO.INFO-DC]Computer Science [cs]/Distributed, Parallel, and Cluster Computing [cs.DC], Apprentissage, [INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI]
Abstract

The training phase in Deep Neural Networks has become an important source of computing resource usage and because of the resulting volume of computation, it is crucial to perform it efficiently on parallel architectures. Even today, data parallelism is the most widely used method, but the associated requirement to replicate all the weights on the totality of computation resources poses problems of memory at the level of each node and of collective communications at the level of the platform. In this context, the model parallelism, which consists in distributing the different layers of the network over the computing nodes, is an attractive alternative. Indeed, it is expected to better distribute weights (to cope with memory problems) and it does not imply large collective communications since only forward activations are communicated. However, to be efficient, it must be combined with a pipelined/streaming approach, which leads in turn to new memory costs. The goal of this paper is to model these memory costs in detail and to show that it is possible to formalize this optimization problem as an Integer Linear Program (ILP)., La phase d’apprentissage dans les réseaux neuronaux profonds est devenue une source importante d’utilisation des ressources de calcul et, en raison du volume de calcul qui en résulte, il est crucial de l’exécuter efficacement sur des architectures parallèles. Aujourd’hui encore, le parallélisme de données est la méthode la plus utilisée, mais l’exigence associée de répliquer tous les poids sur la totalité des ressources de calcul pose des problèmes de mémoire au niveau de chaque nœud et de communications collectives au niveau de la plateforme. Dans ce contexte, le parallélisme de modèle, qui consiste à répartir les différentes couches du réseau sur les nœuds de calcul, est une alternative intéressante. En effet, il est censé mieux répartir les poids (pour faire face aux problèmes de mémoire) et il n’implique pas de grosses communications collectives puisque seules les activations "forward" sont communiquées. Cependant, pour être efficace, elle doit être combinée avec une approche pipelinée/streaming, ce qui entraîne à son tour de nouveaux coûts mémoire. L’objectif de cet article est de modéliser ces coûts de mémoire en détail et de montrer qu’il est possible de formaliser ce problème d’optimisation comme un programme linéaire en nombre entier (ILP).

Published
2022

44. Efficient Combination of Rematerialization and Offloading for Training DNNs

Author

Beaumont, Olivier, Eyraud-Dubois, Lionel, Shilova, Alena, High-End Parallel Algorithms for Challenging Numerical Simulations (HiePACS), Laboratoire Bordelais de Recherche en Informatique (LaBRI), Université de Bordeaux (UB)-Centre National de la Recherche Scientifique (CNRS)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Université de Bordeaux (UB)-Centre National de la Recherche Scientifique (CNRS)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Inria Bordeaux - Sud-Ouest, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria), Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Inria Bordeaux - Sud-Ouest, and Beaumont, Olivier

Subjects
[INFO.INFO-DC] Computer Science [cs]/Distributed, Parallel, and Cluster Computing [cs.DC], [INFO.INFO-DC]Computer Science [cs]/Distributed, Parallel, and Cluster Computing [cs.DC]
Abstract

International audience; Rematerialization and offloading are two well known strategies to save memory during the training phase of deep neural networks, allowing data scientists to consider larger models, batch sizes or higher resolution data. Rematerialization trades memory for computation time, whereas Offloading trades memory for data movements. As these two resources are independent, it is appealing to consider the simultaneous combination of both strategies to save even more memory. We precisely model the costs and constraints corresponding to Deep Learning frameworks such as PyTorch or Tensorflow, we propose optimal algorithms to find a valid sequence of memory-constrained operations and finally, we evaluate the performance of proposed algorithms on realistic networks and computation platforms. Our experiments show that the possibility to offload can remove one third of the overhead of rematerialization, and that together they can reduce the memory used for activations by a factor 4 to 6, with an overhead below 20%.

Published
2021

47. Independent tasks on 2 resources with co-scheduling effects

Author

Eyraud-Dubois, Lionel, Bentes, Cristiana, Laboratoire Bordelais de Recherche en Informatique (LaBRI), Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS), High-End Parallel Algorithms for Challenging Numerical Simulations (HiePACS), Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Université de Bordeaux (UB)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Centre National de la Recherche Scientifique (CNRS)-Inria Bordeaux - Sud-Ouest, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria), Universidade do Estado do Rio de Janeiro [Rio de Janeiro] (UERJ), Université de Bordeaux (UB)-Centre National de la Recherche Scientifique (CNRS)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB), Université de Bordeaux (UB)-Centre National de la Recherche Scientifique (CNRS)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Université de Bordeaux (UB)-Centre National de la Recherche Scientifique (CNRS)-École Nationale Supérieure d'Électronique, Informatique et Radiocommunications de Bordeaux (ENSEIRB)-Inria Bordeaux - Sud-Ouest, and Eyraud-Dubois, Lionel

Subjects
Software_OPERATINGSYSTEMS, GPU Concurrency, Scheduling, TheoryofComputation_ANALYSISOFALGORITHMSANDPROBLEMCOMPLEXITY, [INFO.INFO-DC] Computer Science [cs]/Distributed, Parallel, and Cluster Computing [cs.DC], Preemption, [INFO.INFO-DC]Computer Science [cs]/Distributed, Parallel, and Cluster Computing [cs.DC]
Abstract

Concurrent kernel execution is a relatively new feature in modern GPUs, which was designed to improve hardware utilization and the overall system throughput. However, the decision on the simultaneous execution of tasks is performed by the hardware with a leftover policy, that assigns as many resources as possible for one task and then assigns the remaining resources to the next task. This can lead to unreasonable use of resources. In this work, we tackle the problem of co-scheduling for GPUs with and without preemption, with the focus on determining the kernels submission order to reduce the number of preemptions and the kernels makespan, respectively. We propose a graph-based theoretical model to build preemptive and non-preemptive schedules. We show that the optimal preemptive makespan can be computed by solving a Linear Program in polynomial time, and we propose an algorithm based on this solution which minimizes the number of preemptions. We also propose an algorithm that transforms a preemptive solution of optimal makespan into a non-preemptive solution with the smallest possible preemption overhead. We show, however, that finding the minimal amount of preemptions among all preemptive solutions of optimal makespan is a NP-hard problem, and computing the optimal non-preemptive schedule is also NP-hard. In addition, we study the non-preemptive problem, without searching first for a good preemptive solution, and present a Mixed Integer Linear Program solution to this problem. We performed experiments on real-world GPU applications and our approach can achieve optimal makespan by preempting 6 to 9% of the tasks. Our non-preemptive approach, on the other side, obtains makespan within 2.5% of the optimal preemptive schedules, while previous approaches exceed the preemptive makespan by 5 to 12%.

Published
2020

48. A complete biomimetic iron-sulfur cubane redox series

Author

Grunwald, Liam, Clémancey, Martin, Klose, Daniel, Dubois, Lionel, Gambarelli, Serge, Jeschke, Gunnar, Wörle, Michael, Blondin, Geneviève, and Mougel, Victor

Abstract

Synthetic iron-sulfur cubanes are essential models for biological cofactors in the more complex enzymatic environments. However, a complete series of [Fe4S4]n complexes spanning all biorelevant oxidation states (n = 0-3+) has never been prepared. Here, we demonstrate that the use of a bulky arylthiolate ligand promoting the encapsulation of alkali-metal cations in the vicinity of the cubane enables the synthesis of such a series. Characterization by EPR, 57Fe Mössbauer spectroscopy, UV-Vis electronic absorption and variable-temperature X-ray diffraction analysis reveals key trends for the Fe4S4 core’s geometry as well as for the Mössbauer isomer shift, which both correlate systematically with oxidation state. Furthermore, we confirm the S=4 electronic ground state of the most reduced member, [Fe4S4]0, in agreement with that proposed for the all-ferrous cubanes in Nature., ChemRxiv

Published
2021
Full Text
View/download PDF

49. Life cycle and techno‐economic assessments of direct air capture processes: An integrated review.

Author

Chauvy, Remi and Dubois, Lionel

Subjects
*PRODUCT life cycle assessment, *WASTE heat, *CARBON emissions, *CARBON offsetting, *SYNTHETIC fuels, *ECONOMIC indicators
Abstract

Summary: To limit the increase of the global average temperature in the range of 1.5°C to 2°C above pre‐industrial levels, it is mandatory to reduce anthropogenic CO2 emissions. Aggressive mitigation measures are thus needed to tackle these emissions leading to "net negative CO2 emissions." The present study focuses on direct air capture (DAC) processes among the diverse negative emissions approaches. DAC refers to man‐made technologies that selectively extract CO2 from ambient air and deliver it in a concentrated form for further use or storage. DAC technologies are currently developed at different levels of maturity and performance. They can be classified into three main approaches, the liquid sorbent approach, the solid sorbent approach, and a panel of more innovative technologies combining different approaches. They involve various unit operations and different materials and energy types (electrical and thermal). To better evaluate the status and both the environmental and economic performances, the present paper provides a literature review of the life cycle (LCA) and techno‐economic (TEA) assessments in relation to DAC process chains. It was emphasized that DAC could lead to negative emissions if paired with subsequent storage, while the production of synthetic fuels can at best be carbon neutral when using CO2 from the air. Building large DAC plants has an impact on the amount of energy required to operate them, as well as other environmental impacts with regard to land, water, and material use. Even if the carbon‐negative characteristic of DAC was confirmed, these technologies are still expensive. It was highlighted that large DAC costs ranges are currently provided in the literature, from €80/tCO2 to €1133/tCO2 for the current DAC processes, while estimations from €34 to €260/tCO2 are expected in the future. Different levers were identified to improve the environmental and economic performances of DAC processes, such as the availability of waste heat, the heat integration possibilities, and, among others, the improvement of contactors and sorbents properties. Highlights: The paper entitled "Life cycle and techno‐economic assessments of direct air capture processes: An integrated review," submitted by Dr Remi Chauvy and Dr Lionel Dubois, fills the current gap in terms of comprehensive systematic and impacts assessment research on Direct Air Capture (DAC) technologies.Indeed, to better evaluate the status and both the environmental and economic performances of DAC technologies, the present paper provides an up‐to‐date literature review of the life cycle (LCA) and techno‐economic (TEA) assessments in relation with DAC process chains.The paper goes beyond the literature review as key levers, related to technology developments, to the energy use and to the implementation characteristics, are also identified to improve the environmental and economic performances of DAC processes. [ABSTRACT FROM AUTHOR]

Published
2022
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results