FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2904283553> ?p ?o ?g. }

• W2904283553 abstract "Dynamic neural networks enable higher representation flexibility compared to networks with a fixed architecture and are extensively deployed in problems dealing with varying input-induced network structure, such as those in Natural Language Processing. One of the standard optimizations used in static net training is persistency of recurrent weights on the chip. In dynamic nets, possibly-inhomogeneous computation graph for every input prevents caching recurrent weights in GPU registers. Therefore, existing solutions suffer from excessive recurring off-chip memory loads as well as compounded kernel launch overheads leading to underutilization of GPU SMs. In this paper, we present a software system that enables persistency of weight matrices during the training of dynamic neural networks on the GPU. Before the training begins, our approach named Virtual Persistent Processor Specialization (VPPS) specializes a forward-backward propagation kernel that contains in-register caching and operation routines. VPPS virtualizes persistent kernel CTAs as CISC-like vector processors that can be guided to execute supplied instructions. VPPS greatly reduces the overall amount of off-chip loads by caching weight matrices on the chip, while simultaneously, provides maximum portability as it does not make any assumptions about the shape of the given computation graphs hence fulfilling dynamic net requirements. We implemented our solution on DyNet and abstracted away its design complexities by providing simple function calls to the user. Our experiments on a Volta micro-architecture shows that, unlike the most competitive solutions, VPPS shows excellent performance even in small batch sizes and delivers up to 6x speedup on training dynamic nets." @default.
• W2904283553 created "2018-12-22" @default.
• W2904283553 creator A5036836220 @default.
• W2904283553 creator A5043513001 @default.
• W2904283553 creator A5046230478 @default.
• W2904283553 creator A5059614371 @default.
• W2904283553 date "2018-10-01" @default.
• W2904283553 modified "2023-10-17" @default.
• W2904283553 title "In-Register Parameter Caching for Dynamic Neural Nets with Virtual Persistent Processor Specialization" @default.
• W2904283553 cites W1519503583 @default.
• W2904283553 cites W1968391520 @default.
• W2904283553 cites W1972783048 @default.
• W2904283553 cites W2009832130 @default.
• W2904283553 cites W2048266589 @default.
• W2904283553 cites W2049875313 @default.
• W2904283553 cites W2067313328 @default.
• W2904283553 cites W2067523571 @default.
• W2904283553 cites W2076304675 @default.
• W2904283553 cites W2106329447 @default.
• W2904283553 cites W2117130368 @default.
• W2904283553 cites W2120844411 @default.
• W2904283553 cites W2134427337 @default.
• W2904283553 cites W2143612262 @default.
• W2904283553 cites W2150851481 @default.
• W2904283553 cites W2153636750 @default.
• W2904283553 cites W2155893237 @default.
• W2904283553 cites W2171399035 @default.
• W2904283553 cites W2236252626 @default.
• W2904283553 cites W2238700765 @default.
• W2904283553 cites W2238992335 @default.
• W2904283553 cites W2285660444 @default.
• W2904283553 cites W2323693848 @default.
• W2904283553 cites W2330958039 @default.
• W2904283553 cites W2415973476 @default.
• W2904283553 cites W2489529491 @default.
• W2904283553 cites W2510980549 @default.
• W2904283553 cites W2590246587 @default.
• W2904283553 cites W2605178034 @default.
• W2904283553 cites W2605347906 @default.
• W2904283553 cites W2606722458 @default.
• W2904283553 cites W2620106252 @default.
• W2904283553 cites W2625200202 @default.
• W2904283553 cites W2625457103 @default.
• W2904283553 cites W2657126969 @default.
• W2904283553 cites W2756276247 @default.
• W2904283553 cites W2765315405 @default.
• W2904283553 cites W2765329037 @default.
• W2904283553 cites W2765927961 @default.
• W2904283553 cites W2789554134 @default.
• W2904283553 cites W2790925711 @default.
• W2904283553 cites W2884590322 @default.
• W2904283553 cites W2906043559 @default.
• W2904283553 cites W2964194679 @default.
• W2904283553 cites W4236713805 @default.
• W2904283553 cites W4240168186 @default.
• W2904283553 cites W4245199738 @default.
• W2904283553 cites W4251575795 @default.
• W2904283553 doi "https://doi.org/10.1109/micro.2018.00038" @default.
• W2904283553 hasPublicationYear "2018" @default.
• W2904283553 type Work @default.
• W2904283553 sameAs 2904283553 @default.
• W2904283553 citedByCount "16" @default.
• W2904283553 countsByYear W29042835532018 @default.
• W2904283553 countsByYear W29042835532019 @default.
• W2904283553 countsByYear W29042835532020 @default.
• W2904283553 countsByYear W29042835532021 @default.
• W2904283553 countsByYear W29042835532023 @default.
• W2904283553 crossrefType "proceedings-article" @default.
• W2904283553 hasAuthorship W2904283553A5036836220 @default.
• W2904283553 hasAuthorship W2904283553A5043513001 @default.
• W2904283553 hasAuthorship W2904283553A5046230478 @default.
• W2904283553 hasAuthorship W2904283553A5059614371 @default.
• W2904283553 hasConcept C105795698 @default.
• W2904283553 hasConcept C111919701 @default.
• W2904283553 hasConcept C11413529 @default.
• W2904283553 hasConcept C114614502 @default.
• W2904283553 hasConcept C115537543 @default.
• W2904283553 hasConcept C173608175 @default.
• W2904283553 hasConcept C2780598303 @default.
• W2904283553 hasConcept C33923547 @default.
• W2904283553 hasConcept C41008148 @default.
• W2904283553 hasConcept C45374587 @default.
• W2904283553 hasConcept C63000827 @default.
• W2904283553 hasConcept C68339613 @default.
• W2904283553 hasConcept C74193536 @default.
• W2904283553 hasConceptScore W2904283553C105795698 @default.
• W2904283553 hasConceptScore W2904283553C111919701 @default.
• W2904283553 hasConceptScore W2904283553C11413529 @default.
• W2904283553 hasConceptScore W2904283553C114614502 @default.
• W2904283553 hasConceptScore W2904283553C115537543 @default.
• W2904283553 hasConceptScore W2904283553C173608175 @default.
• W2904283553 hasConceptScore W2904283553C2780598303 @default.
• W2904283553 hasConceptScore W2904283553C33923547 @default.
• W2904283553 hasConceptScore W2904283553C41008148 @default.
• W2904283553 hasConceptScore W2904283553C45374587 @default.
• W2904283553 hasConceptScore W2904283553C63000827 @default.
• W2904283553 hasConceptScore W2904283553C68339613 @default.
• W2904283553 hasConceptScore W2904283553C74193536 @default.
• W2904283553 hasLocation W29042835531 @default.
• W2904283553 hasOpenAccess W2904283553 @default.

• next