FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2912976659> ?p ?o ?g. }

• W2912976659 abstract "Typically, Ultra-deep neural network(UDNN) tends to yield high-quality model, but its training process is usually resource intensive and time-consuming. Modern GPU's scarce DRAM capacity is the primary bottleneck that hinders the trainability and the training efficiency of UDNN. In this paper, we present AccUDNN, an accelerator that aims to make the utmost use of finite GPU memory resources to speed up the training process of UDNN. AccUDNN mainly includes two modules: memory optimizer and hyperparameter tuner. Memory optimizer develops a performance-model guided dynamic swap out/in strategy, by offloading appropriate data to host memory, GPU memory footprint can be significantly slashed to overcome the restriction of trainability of UDNN. After applying the memory optimization strategy, hyperparameter tuner is designed to explore the efficiency-optimal minibatch size and the matched learning rate. Evaluations demonstrate that AccUDNN cuts down the GPU memory requirement of ResNet-152 from more than 24GB to 8GB. In turn, given 12GB GPU memory budget, the efficiency-optimal minibatch size can reach 4.2x larger than original Caffe. Benefiting from better utilization of single GPU's computing resources and fewer parameter synchronization of large minibatch size, 7.7x speed-up is achieved by 8 GPUs' cluster without any communication optimization and no accuracy losses." @default.
• W2912976659 created "2019-02-21" @default.
• W2912976659 creator A5011069823 @default.
• W2912976659 creator A5018288181 @default.
• W2912976659 creator A5019603661 @default.
• W2912976659 creator A5038603194 @default.
• W2912976659 creator A5039670436 @default.
• W2912976659 creator A5058924557 @default.
• W2912976659 creator A5091348611 @default.
• W2912976659 date "2019-01-20" @default.
• W2912976659 modified "2023-10-16" @default.
• W2912976659 title "AccUDNN: A GPU Memory Efficient Accelerator for Training Ultra-deep Neural Networks" @default.
• W2912976659 cites W1667652561 @default.
• W2912976659 cites W1686810756 @default.
• W2912976659 cites W2094233035 @default.
• W2912976659 cites W2155893237 @default.
• W2912976659 cites W2162390675 @default.
• W2912976659 cites W2168231600 @default.
• W2912976659 cites W2194775991 @default.
• W2912976659 cites W2338908902 @default.
• W2912976659 cites W2339765813 @default.
• W2912976659 cites W2402144811 @default.
• W2912976659 cites W2580688187 @default.
• W2912976659 cites W2622263826 @default.
• W2912976659 cites W2759465730 @default.
• W2912976659 cites W2766164908 @default.
• W2912976659 cites W2769856846 @default.
• W2912976659 cites W2809901035 @default.
• W2912976659 cites W2883283076 @default.
• W2912976659 cites W2963433607 @default.
• W2912976659 cites W2963513744 @default.
• W2912976659 cites W2964174152 @default.
• W2912976659 cites W3103894541 @default.
• W2912976659 doi "https://doi.org/10.48550/arxiv.1901.06773" @default.
• W2912976659 hasPublicationYear "2019" @default.
• W2912976659 type Work @default.
• W2912976659 sameAs 2912976659 @default.
• W2912976659 citedByCount "0" @default.
• W2912976659 crossrefType "posted-content" @default.
• W2912976659 hasAuthorship W2912976659A5011069823 @default.
• W2912976659 hasAuthorship W2912976659A5018288181 @default.
• W2912976659 hasAuthorship W2912976659A5019603661 @default.
• W2912976659 hasAuthorship W2912976659A5038603194 @default.
• W2912976659 hasAuthorship W2912976659A5039670436 @default.
• W2912976659 hasAuthorship W2912976659A5058924557 @default.
• W2912976659 hasAuthorship W2912976659A5091348611 @default.
• W2912976659 hasBestOaLocation W29129766591 @default.
• W2912976659 hasConcept C108583219 @default.
• W2912976659 hasConcept C111919701 @default.
• W2912976659 hasConcept C113775141 @default.
• W2912976659 hasConcept C149635348 @default.
• W2912976659 hasConcept C154945302 @default.
• W2912976659 hasConcept C173608175 @default.
• W2912976659 hasConcept C2780513914 @default.
• W2912976659 hasConcept C2781357197 @default.
• W2912976659 hasConcept C41008148 @default.
• W2912976659 hasConcept C50644808 @default.
• W2912976659 hasConcept C68339613 @default.
• W2912976659 hasConcept C7366592 @default.
• W2912976659 hasConcept C74064498 @default.
• W2912976659 hasConcept C74912251 @default.
• W2912976659 hasConcept C76155785 @default.
• W2912976659 hasConcept C8642999 @default.
• W2912976659 hasConcept C9390403 @default.
• W2912976659 hasConcept C98045186 @default.
• W2912976659 hasConcept C9819579 @default.
• W2912976659 hasConceptScore W2912976659C108583219 @default.
• W2912976659 hasConceptScore W2912976659C111919701 @default.
• W2912976659 hasConceptScore W2912976659C113775141 @default.
• W2912976659 hasConceptScore W2912976659C149635348 @default.
• W2912976659 hasConceptScore W2912976659C154945302 @default.
• W2912976659 hasConceptScore W2912976659C173608175 @default.
• W2912976659 hasConceptScore W2912976659C2780513914 @default.
• W2912976659 hasConceptScore W2912976659C2781357197 @default.
• W2912976659 hasConceptScore W2912976659C41008148 @default.
• W2912976659 hasConceptScore W2912976659C50644808 @default.
• W2912976659 hasConceptScore W2912976659C68339613 @default.
• W2912976659 hasConceptScore W2912976659C7366592 @default.
• W2912976659 hasConceptScore W2912976659C74064498 @default.
• W2912976659 hasConceptScore W2912976659C74912251 @default.
• W2912976659 hasConceptScore W2912976659C76155785 @default.
• W2912976659 hasConceptScore W2912976659C8642999 @default.
• W2912976659 hasConceptScore W2912976659C9390403 @default.
• W2912976659 hasConceptScore W2912976659C98045186 @default.
• W2912976659 hasConceptScore W2912976659C9819579 @default.
• W2912976659 hasLocation W29129766591 @default.
• W2912976659 hasOpenAccess W2912976659 @default.
• W2912976659 hasPrimaryLocation W29129766591 @default.
• W2912976659 hasRelatedWork W1188123746 @default.
• W2912976659 hasRelatedWork W2164287667 @default.
• W2912976659 hasRelatedWork W2724545582 @default.
• W2912976659 hasRelatedWork W2912976659 @default.
• W2912976659 hasRelatedWork W3005578915 @default.
• W2912976659 hasRelatedWork W3183058688 @default.
• W2912976659 hasRelatedWork W3207233976 @default.
• W2912976659 hasRelatedWork W4286901967 @default.
• W2912976659 hasRelatedWork W4306353333 @default.
• W2912976659 hasRelatedWork W4308613523 @default.
• W2912976659 isParatext "false" @default.
• W2912976659 isRetracted "false" @default.

• next