English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The past few years have seen a surge of applying Deep Learning (DL) modelsfor a wide array of tasks such as image classification, object detection,machine translation, etc. While DL models provide an opportunity to solveotherwise intractable tasks, their adoption relies on them being optimized tomeet latency and resource requirements. Benchmarking is a key step in thisprocess but has been hampered in part due to the lack of representative andup-to-date benchmarking suites. This is exacerbated by the fast-evolving paceof DL models.  This paper proposes DLBricks, a composable benchmark generation design thatreduces the effort of developing, maintaining, and running DL benchmarks onCPUs. DLBricks decomposes DL models into a set of unique runnable networks andconstructs the original model's performance using the performance of thegenerated benchmarks. DLBricks leverages two key observations: DL layers arethe performance building blocks of DL models and layers are extensivelyrepeated within and across DL models. Since benchmarks are generatedautomatically and the benchmarking time is minimized, DLBricks can keepup-to-date with the latest proposed models, relieving the pressure of selectingrepresentative DL models. Moreover, DLBricks allows users to representproprietary models within benchmark suites. We evaluate DLBricks using $50$MXNet models spanning $5$ DL tasks on $4$ representative CPU systems. We showthat DLBricks provides an accurate performance estimate for the DL models andreduces the benchmarking time across systems (e.g. within $95\%$ accuracy andup to $4.4\times$ benchmarking time speedup on Amazon EC2 c5.xlarge).

DLBricks: Composable Benchmark Generation to Reduce Deep Learning Benchmarking Effort on CPUs (Extended)