Matches in SemOpenAlex for { <https://semopenalex.org/work/W2894576937> ?p ?o ?g. }
Showing items 1 to 76 of
76
with 100 items per page.
- W2894576937 abstract "State-of-the-art deep learning systems rely on iterative distributed training to tackle the increasing complexity of models and input data. The iteration time in these communication-heavy systems depends on the computation time, communication time and the extent of overlap of computation and communication. In this work, we identify a shortcoming in systems with graph representation for computation, such as TensorFlow and PyTorch, that result in high variance in iteration time --- random order of received parameters across workers. We develop a system, TicTac, to improve the iteration time by fixing this issue in distributed deep learning with Parameter Servers while guaranteeing near-optimal overlap of communication and computation. TicTac identifies and enforces an order of network transfers which improves the iteration time using prioritization. Our system is implemented over TensorFlow and requires no changes to the model or developer inputs. TicTac improves the throughput by up to $37.7%$ in inference and $19.2%$ in training, while also reducing straggler effect by up to $2.3times$. Our code is publicly available." @default.
- W2894576937 created "2018-10-12" @default.
- W2894576937 creator A5056821502 @default.
- W2894576937 creator A5058967466 @default.
- W2894576937 creator A5065622216 @default.
- W2894576937 date "2018-03-08" @default.
- W2894576937 modified "2023-09-24" @default.
- W2894576937 title "TicTac: Accelerating Distributed Deep Learning with Communication Scheduling" @default.
- W2894576937 hasPublicationYear "2018" @default.
- W2894576937 type Work @default.
- W2894576937 sameAs 2894576937 @default.
- W2894576937 citedByCount "5" @default.
- W2894576937 countsByYear W28945769372019 @default.
- W2894576937 countsByYear W28945769372020 @default.
- W2894576937 countsByYear W28945769372021 @default.
- W2894576937 crossrefType "posted-content" @default.
- W2894576937 hasAuthorship W2894576937A5056821502 @default.
- W2894576937 hasAuthorship W2894576937A5058967466 @default.
- W2894576937 hasAuthorship W2894576937A5065622216 @default.
- W2894576937 hasConcept C101765175 @default.
- W2894576937 hasConcept C108583219 @default.
- W2894576937 hasConcept C11413529 @default.
- W2894576937 hasConcept C119857082 @default.
- W2894576937 hasConcept C120314980 @default.
- W2894576937 hasConcept C126255220 @default.
- W2894576937 hasConcept C154945302 @default.
- W2894576937 hasConcept C206729178 @default.
- W2894576937 hasConcept C2776214188 @default.
- W2894576937 hasConcept C31258907 @default.
- W2894576937 hasConcept C33923547 @default.
- W2894576937 hasConcept C41008148 @default.
- W2894576937 hasConcept C45374587 @default.
- W2894576937 hasConcept C80444323 @default.
- W2894576937 hasConcept C93996380 @default.
- W2894576937 hasConceptScore W2894576937C101765175 @default.
- W2894576937 hasConceptScore W2894576937C108583219 @default.
- W2894576937 hasConceptScore W2894576937C11413529 @default.
- W2894576937 hasConceptScore W2894576937C119857082 @default.
- W2894576937 hasConceptScore W2894576937C120314980 @default.
- W2894576937 hasConceptScore W2894576937C126255220 @default.
- W2894576937 hasConceptScore W2894576937C154945302 @default.
- W2894576937 hasConceptScore W2894576937C206729178 @default.
- W2894576937 hasConceptScore W2894576937C2776214188 @default.
- W2894576937 hasConceptScore W2894576937C31258907 @default.
- W2894576937 hasConceptScore W2894576937C33923547 @default.
- W2894576937 hasConceptScore W2894576937C41008148 @default.
- W2894576937 hasConceptScore W2894576937C45374587 @default.
- W2894576937 hasConceptScore W2894576937C80444323 @default.
- W2894576937 hasConceptScore W2894576937C93996380 @default.
- W2894576937 hasLocation W28945769371 @default.
- W2894576937 hasOpenAccess W2894576937 @default.
- W2894576937 hasPrimaryLocation W28945769371 @default.
- W2894576937 hasRelatedWork W1574696033 @default.
- W2894576937 hasRelatedWork W189448761 @default.
- W2894576937 hasRelatedWork W2051518507 @default.
- W2894576937 hasRelatedWork W2108598243 @default.
- W2894576937 hasRelatedWork W2194775991 @default.
- W2894576937 hasRelatedWork W2325941297 @default.
- W2894576937 hasRelatedWork W2339900379 @default.
- W2894576937 hasRelatedWork W2407022425 @default.
- W2894576937 hasRelatedWork W2584979567 @default.
- W2894576937 hasRelatedWork W2622263826 @default.
- W2894576937 hasRelatedWork W2769644379 @default.
- W2894576937 hasRelatedWork W2791016214 @default.
- W2894576937 hasRelatedWork W2807897761 @default.
- W2894576937 hasRelatedWork W2955863409 @default.
- W2894576937 hasRelatedWork W2968430711 @default.
- W2894576937 hasRelatedWork W2975712713 @default.
- W2894576937 hasRelatedWork W3101036738 @default.
- W2894576937 hasRelatedWork W3154864063 @default.
- W2894576937 hasRelatedWork W3200800229 @default.
- W2894576937 hasRelatedWork W3205353516 @default.
- W2894576937 isParatext "false" @default.
- W2894576937 isRetracted "false" @default.
- W2894576937 magId "2894576937" @default.
- W2894576937 workType "article" @default.