Mostrar o rexistro simple do ítem
STuning-DL: Model-Driven Autotuning of Sparse GPU Kernels for Deep Learning
dc.contributor.author | López Castro, Roberto | |
dc.contributor.author | Andrade, Diego | |
dc.contributor.author | Fraguela, Basilio B. | |
dc.date.accessioned | 2024-06-05T14:30:12Z | |
dc.date.available | 2024-06-05T14:30:12Z | |
dc.date.issued | 2024-05 | |
dc.identifier.citation | R. L. Castro, D. Andrade and B. B. Fraguela, "STuning-DL: Model-Driven Autotuning of Sparse GPU Kernels for Deep Learning," in IEEE Access, vol. 12, pp. 70581-70599, 2024, doi: 10.1109/ACCESS.2024.3402326. | es_ES |
dc.identifier.issn | 2169-3536 | |
dc.identifier.uri | http://hdl.handle.net/2183/36810 | |
dc.description.abstract | [Abstract]: The relentless growth of modern Machine Learning models has spurred the adoption of sparsification techniques to simplify their architectures and reduce the computational demands. Network pruning has demonstrated success in maintaining original network accuracy while shedding significant portions of the original weights. However, leveraging this sparsity efficiently remains challenging due to computational irregularities, particularly in GPU kernels. A new trend of template-based GPU kernels for semi-structured sparsity shows promise in efficiency but lacks autotuning capabilities to adapt to input dynamics, often underperforming in scenarios where they have not been meticulously hand-tuned. We present STuning-DL, the first pruning-aware autotuner for third-party template-based implementations enabling efficient optimization of sparse kernels for Deep Learning, spanning from high-level aspects (CUDA C++ level) down to GPU-native instructions specifics (assembly-level). STuning-DL tunes and optimizes at run-time sparse kernels’ performance for each input problem, yielding speedups of up to 5.42× on NVIDIA T4-16GB and up to 3.6× on NVIDIA A100-40GB GPU in sparse matrices from real world models compared to existing heuristics from sparse libraries like cuSparse and cuSparseLt. | es_ES |
dc.description.sponsorship | This work was supported by grant PID2022-136435NB-I00, funded by MCIN/AEI/10.13039/501100011033 and by ‘‘ERDF A way of making Europe’’, EU; also by Xunta de Galicia under the Consolidation Programme of Competitive Reference Groups, ref. ED431C 2021/30. The work of Roberto L. Castro was supported by a predoctoral grant from the Ministry of Science, Innovation and Universities, ref. FPU19/03974. | es_ES |
dc.description.sponsorship | Xunta de Galicia; ED431C 2021/30 | es_ES |
dc.language.iso | eng | es_ES |
dc.publisher | Institute of Electrical and Electronics Engineers | es_ES |
dc.relation | info:eu-repo/grantAgreement/AEI/Plan Estatal de Investigación Científica y Técnica y de Innovación 2021-2023/PID2022-136435NB-I00/ES/ARQUITECTURAS, FRAMEWORKS Y APLICACIONES DE LA COMPUTACION DE ALTAS PRESTACIONES | es_ES |
dc.relation | info:eu-repo/grantAgreement//Plan Estatal de Investigación Científica y Técnica y de Innovación 2017-2020/FPU19%2F03974/ES/ | es_ES |
dc.relation.uri | https://doi.org/10.1109/ACCESS.2024.3402326 | es_ES |
dc.rights | Atribución 4.0 Internacional | es_ES |
dc.rights.uri | http://creativecommons.org/licenses/by/3.0/es/ | * |
dc.subject | CUDA | es_ES |
dc.subject | GPU | es_ES |
dc.subject | Learning-based predictive model | es_ES |
dc.subject | Network pruning | es_ES |
dc.subject | Sparse computation | es_ES |
dc.subject | SpMM | es_ES |
dc.subject | Tensor Core | es_ES |
dc.title | STuning-DL: Model-Driven Autotuning of Sparse GPU Kernels for Deep Learning | es_ES |
dc.type | info:eu-repo/semantics/article | es_ES |
dc.rights.access | info:eu-repo/semantics/openAccess | es_ES |
UDC.journalTitle | IEEE Access | es_ES |
UDC.volume | 12 | es_ES |
UDC.startPage | 70581 | es_ES |
UDC.endPage | 70599 | es_ES |
dc.identifier.doi | 10.1109/ACCESS.2024.3402326 |
Ficheiros no ítem
Este ítem aparece na(s) seguinte(s) colección(s)
-
GI-GAC - Artigos [187]