Matches in SemOpenAlex for { <https://semopenalex.org/work/W4287169069> ?p ?o ?g. }
- W4287169069 abstract "Tremendous success of machine learning (ML) and the unabated growth in ML model complexity motivated many ML-specific designs in both CPU and accelerator architectures to speed up the model inference. While these architectures are diverse, highly optimized low-precision arithmetic is a component shared by most. Impressive compute throughputs are indeed often exhibited by these architectures on benchmark ML models. Nevertheless, production models such as recommendation systems important to Facebook's personalization services are demanding and complex: These systems must serve billions of users per month responsively with low latency while maintaining high prediction accuracy, notwithstanding computations with many tens of billions parameters per inference. Do these low-precision architectures work well with our production recommendation systems? They do. But not without significant effort. We share in this paper our search strategies to adapt reference recommendation models to low-precision hardware, our optimization of low-precision compute kernels, and the design and development of tool chain so as to maintain our models' accuracy throughout their lifespan during which topic trends and users' interests inevitably evolve. Practicing these low-precision technologies helped us save datacenter capacities while deploying models with up to 5X complexity that would otherwise not be deployed on traditional general-purpose CPUs. We believe these lessons from the trenches promote better co-design between hardware architecture and software engineering and advance the state of the art of ML in industry." @default.
- W4287169069 created "2022-07-25" @default.
- W4287169069 creator A5001698250 @default.
- W4287169069 creator A5002591266 @default.
- W4287169069 creator A5007535563 @default.
- W4287169069 creator A5007892541 @default.
- W4287169069 creator A5008108990 @default.
- W4287169069 creator A5028220093 @default.
- W4287169069 creator A5038603364 @default.
- W4287169069 creator A5039508118 @default.
- W4287169069 creator A5046749324 @default.
- W4287169069 creator A5047403657 @default.
- W4287169069 creator A5047513659 @default.
- W4287169069 creator A5048895100 @default.
- W4287169069 creator A5051151162 @default.
- W4287169069 creator A5059637342 @default.
- W4287169069 creator A5060639309 @default.
- W4287169069 creator A5060661843 @default.
- W4287169069 creator A5061948021 @default.
- W4287169069 creator A5069883905 @default.
- W4287169069 creator A5090570610 @default.
- W4287169069 creator A5039488725 @default.
- W4287169069 date "2021-05-26" @default.
- W4287169069 modified "2023-09-28" @default.
- W4287169069 title "Low-Precision Hardware Architectures Meet Recommendation Model Inference at Scale" @default.
- W4287169069 hasPublicationYear "2021" @default.
- W4287169069 type Work @default.
- W4287169069 citedByCount "0" @default.
- W4287169069 crossrefType "posted-content" @default.
- W4287169069 hasAuthorship W4287169069A5001698250 @default.
- W4287169069 hasAuthorship W4287169069A5002591266 @default.
- W4287169069 hasAuthorship W4287169069A5007535563 @default.
- W4287169069 hasAuthorship W4287169069A5007892541 @default.
- W4287169069 hasAuthorship W4287169069A5008108990 @default.
- W4287169069 hasAuthorship W4287169069A5028220093 @default.
- W4287169069 hasAuthorship W4287169069A5038603364 @default.
- W4287169069 hasAuthorship W4287169069A5039488725 @default.
- W4287169069 hasAuthorship W4287169069A5039508118 @default.
- W4287169069 hasAuthorship W4287169069A5046749324 @default.
- W4287169069 hasAuthorship W4287169069A5047403657 @default.
- W4287169069 hasAuthorship W4287169069A5047513659 @default.
- W4287169069 hasAuthorship W4287169069A5048895100 @default.
- W4287169069 hasAuthorship W4287169069A5051151162 @default.
- W4287169069 hasAuthorship W4287169069A5059637342 @default.
- W4287169069 hasAuthorship W4287169069A5060639309 @default.
- W4287169069 hasAuthorship W4287169069A5060661843 @default.
- W4287169069 hasAuthorship W4287169069A5061948021 @default.
- W4287169069 hasAuthorship W4287169069A5069883905 @default.
- W4287169069 hasAuthorship W4287169069A5090570610 @default.
- W4287169069 hasBestOaLocation W42871690691 @default.
- W4287169069 hasConcept C111919701 @default.
- W4287169069 hasConcept C113775141 @default.
- W4287169069 hasConcept C118524514 @default.
- W4287169069 hasConcept C119857082 @default.
- W4287169069 hasConcept C120314980 @default.
- W4287169069 hasConcept C121332964 @default.
- W4287169069 hasConcept C123657996 @default.
- W4287169069 hasConcept C13280743 @default.
- W4287169069 hasConcept C136764020 @default.
- W4287169069 hasConcept C142362112 @default.
- W4287169069 hasConcept C153349607 @default.
- W4287169069 hasConcept C154945302 @default.
- W4287169069 hasConcept C168167062 @default.
- W4287169069 hasConcept C183003079 @default.
- W4287169069 hasConcept C185798385 @default.
- W4287169069 hasConcept C205649164 @default.
- W4287169069 hasConcept C2776214188 @default.
- W4287169069 hasConcept C2777904410 @default.
- W4287169069 hasConcept C31258907 @default.
- W4287169069 hasConcept C41008148 @default.
- W4287169069 hasConcept C46637626 @default.
- W4287169069 hasConcept C557471498 @default.
- W4287169069 hasConcept C97355855 @default.
- W4287169069 hasConceptScore W4287169069C111919701 @default.
- W4287169069 hasConceptScore W4287169069C113775141 @default.
- W4287169069 hasConceptScore W4287169069C118524514 @default.
- W4287169069 hasConceptScore W4287169069C119857082 @default.
- W4287169069 hasConceptScore W4287169069C120314980 @default.
- W4287169069 hasConceptScore W4287169069C121332964 @default.
- W4287169069 hasConceptScore W4287169069C123657996 @default.
- W4287169069 hasConceptScore W4287169069C13280743 @default.
- W4287169069 hasConceptScore W4287169069C136764020 @default.
- W4287169069 hasConceptScore W4287169069C142362112 @default.
- W4287169069 hasConceptScore W4287169069C153349607 @default.
- W4287169069 hasConceptScore W4287169069C154945302 @default.
- W4287169069 hasConceptScore W4287169069C168167062 @default.
- W4287169069 hasConceptScore W4287169069C183003079 @default.
- W4287169069 hasConceptScore W4287169069C185798385 @default.
- W4287169069 hasConceptScore W4287169069C205649164 @default.
- W4287169069 hasConceptScore W4287169069C2776214188 @default.
- W4287169069 hasConceptScore W4287169069C2777904410 @default.
- W4287169069 hasConceptScore W4287169069C31258907 @default.
- W4287169069 hasConceptScore W4287169069C41008148 @default.
- W4287169069 hasConceptScore W4287169069C46637626 @default.
- W4287169069 hasConceptScore W4287169069C557471498 @default.
- W4287169069 hasConceptScore W4287169069C97355855 @default.
- W4287169069 hasLocation W42871690691 @default.
- W4287169069 hasOpenAccess W4287169069 @default.
- W4287169069 hasPrimaryLocation W42871690691 @default.
- W4287169069 hasRelatedWork W10010518 @default.