FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4312038605> ?p ?o ?g. }

• W4312038605 abstract "The ability to scale out training workloads has been one of the key performance enablers of deep learning. The main scaling approach is data-parallel GPU-based training, which has been boosted by hardware and software support for highly efficient point-to-point communication, and in particular via hardware bandwidth overprovisioning. Overprovisioning comes at a cost: there is an order of magnitude price difference between cloud-grade servers with such support, relative to their popular consumer-grade counterparts, although single server-grade and consumer-grade GPUs can have similar computational envelopes. In this paper, we show that the costly hardware overprovisioning approach can be supplanted via algorithmic and system design, and propose a framework called CGX, which provides efficient software support for compressed communication in ML applications, for both multi-GPU single-node training, as well as larger-scale multi-node training. CGX is based on two technical advances: emph{At the system level}, it relies on a re-developed communication stack for ML frameworks, which provides flexible, highly-efficient support for compressed communication. emph{At the application level}, it provides emph{seamless, parameter-free} integration with popular frameworks, so that end-users do not have to modify training recipes, nor significant training code. This is complemented by a emph{layer-wise adaptive compression} technique which dynamically balances compression gains with accuracy preservation. CGX integrates with popular ML frameworks, providing up to 3X speedups for multi-GPU nodes based on commodity hardware, and order-of-magnitude improvements in the multi-node setting, with negligible impact on accuracy." @default.
• W4312038605 created "2023-01-04" @default.
• W4312038605 creator A5007080821 @default.
• W4312038605 creator A5051243807 @default.
• W4312038605 creator A5083822059 @default.
• W4312038605 date "2022-11-07" @default.
• W4312038605 modified "2023-10-10" @default.
• W4312038605 title "CGX" @default.
• W4312038605 cites W2060393849 @default.
• W4312038605 cites W2108598243 @default.
• W4312038605 cites W2405578611 @default.
• W4312038605 cites W2407022425 @default.
• W4312038605 cites W2563343794 @default.
• W4312038605 cites W2730150110 @default.
• W4312038605 cites W2775776326 @default.
• W4312038605 cites W2903901007 @default.
• W4312038605 cites W2926767350 @default.
• W4312038605 cites W2975712713 @default.
• W4312038605 cites W2985108934 @default.
• W4312038605 cites W2997709794 @default.
• W4312038605 cites W3008591352 @default.
• W4312038605 cites W3016081801 @default.
• W4312038605 cites W3043050223 @default.
• W4312038605 cites W3091097978 @default.
• W4312038605 cites W3109982791 @default.
• W4312038605 cites W3112483307 @default.
• W4312038605 cites W3204921648 @default.
• W4312038605 cites W3206636350 @default.
• W4312038605 cites W4226057994 @default.
• W4312038605 doi "https://doi.org/10.1145/3528535.3565248" @default.
• W4312038605 hasPublicationYear "2022" @default.
• W4312038605 type Work @default.
• W4312038605 citedByCount "0" @default.
• W4312038605 crossrefType "proceedings-article" @default.
• W4312038605 hasAuthorship W4312038605A5007080821 @default.
• W4312038605 hasAuthorship W4312038605A5051243807 @default.
• W4312038605 hasAuthorship W4312038605A5083822059 @default.
• W4312038605 hasBestOaLocation W43120386051 @default.
• W4312038605 hasConcept C111919701 @default.
• W4312038605 hasConcept C113775141 @default.
• W4312038605 hasConcept C120314980 @default.
• W4312038605 hasConcept C127413603 @default.
• W4312038605 hasConcept C173608175 @default.
• W4312038605 hasConcept C2777904410 @default.
• W4312038605 hasConcept C41008148 @default.
• W4312038605 hasConcept C62611344 @default.
• W4312038605 hasConcept C66938386 @default.
• W4312038605 hasConcept C79974875 @default.
• W4312038605 hasConcept C84211073 @default.
• W4312038605 hasConcept C93996380 @default.
• W4312038605 hasConceptScore W4312038605C111919701 @default.
• W4312038605 hasConceptScore W4312038605C113775141 @default.
• W4312038605 hasConceptScore W4312038605C120314980 @default.
• W4312038605 hasConceptScore W4312038605C127413603 @default.
• W4312038605 hasConceptScore W4312038605C173608175 @default.
• W4312038605 hasConceptScore W4312038605C2777904410 @default.
• W4312038605 hasConceptScore W4312038605C41008148 @default.
• W4312038605 hasConceptScore W4312038605C62611344 @default.
• W4312038605 hasConceptScore W4312038605C66938386 @default.
• W4312038605 hasConceptScore W4312038605C79974875 @default.
• W4312038605 hasConceptScore W4312038605C84211073 @default.
• W4312038605 hasConceptScore W4312038605C93996380 @default.
• W4312038605 hasFunder F4320334678 @default.
• W4312038605 hasLocation W43120386051 @default.
• W4312038605 hasLocation W43120386052 @default.
• W4312038605 hasOpenAccess W4312038605 @default.
• W4312038605 hasPrimaryLocation W43120386051 @default.
• W4312038605 hasRelatedWork W1485627940 @default.
• W4312038605 hasRelatedWork W1995076703 @default.
• W4312038605 hasRelatedWork W2275108558 @default.
• W4312038605 hasRelatedWork W2344051412 @default.
• W4312038605 hasRelatedWork W2512832353 @default.
• W4312038605 hasRelatedWork W2564093197 @default.
• W4312038605 hasRelatedWork W2763342290 @default.
• W4312038605 hasRelatedWork W2944998523 @default.
• W4312038605 hasRelatedWork W4226479101 @default.
• W4312038605 hasRelatedWork W4230646269 @default.
• W4312038605 isParatext "false" @default.
• W4312038605 isRetracted "false" @default.
• W4312038605 workType "article" @default.