Matches in SemOpenAlex for { <https://semopenalex.org/work/W3035332806> ?p ?o ?g. }
Showing items 1 to 87 of
87
with 100 items per page.
- W3035332806 abstract "We present APQ for efficient deep learning inference on resource-constrained hardware. Unlike previous methods that separately search the neural architecture, pruning policy, and quantization policy, we optimize them in a joint manner. To deal with the larger design space it brings, a promising approach is to train a quantization-aware accuracy predictor to quickly get the accuracy of the quantized model and feed it to the search engine to select the best fit. However, training this quantization-aware accuracy predictor requires collecting a large number of quantized <model, accuracy> pairs, which involves quantization-aware finetuning and thus is highly time-consuming. To tackle this challenge, we propose to transfer the knowledge from a full-precision (i.e., fp32) accuracy predictor to the quantization-aware (i.e., int8) accuracy predictor, which greatly improves the sample efficiency. Besides, collecting the dataset for the fp32 accuracy predictor only requires to evaluate neural networks without any training cost by sampling from a pretrained once-for-all network, which is highly efficient. Extensive experiments on ImageNet demonstrate the benefits of our joint optimization approach. With the same accuracy, APQ reduces the latency/energy by 2x/1.3x over MobileNetV2+HAQ. Compared to the separate optimization approach (ProxylessNAS+AMC+HAQ), APQ achieves 2.3% higher ImageNet accuracy while reducing orders of magnitude GPU hours and CO2 emission, pushing the frontier for green AI that is environmental-friendly. The code and video are publicly available." @default.
- W3035332806 created "2020-06-19" @default.
- W3035332806 creator A5021398655 @default.
- W3035332806 creator A5022883365 @default.
- W3035332806 creator A5024743366 @default.
- W3035332806 creator A5052596733 @default.
- W3035332806 creator A5065001783 @default.
- W3035332806 creator A5070926896 @default.
- W3035332806 creator A5073936760 @default.
- W3035332806 creator A5087929301 @default.
- W3035332806 date "2020-06-01" @default.
- W3035332806 modified "2023-10-12" @default.
- W3035332806 title "APQ: Joint Search for Network Architecture, Pruning and Quantization Policy" @default.
- W3035332806 cites W2123469553 @default.
- W3035332806 cites W2794952988 @default.
- W3035332806 cites W2799238091 @default.
- W3035332806 cites W2962851801 @default.
- W3035332806 cites W2963122961 @default.
- W3035332806 cites W2963163009 @default.
- W3035332806 cites W2963363373 @default.
- W3035332806 cites W2963367920 @default.
- W3035332806 cites W2963809228 @default.
- W3035332806 cites W2963918968 @default.
- W3035332806 cites W2964081807 @default.
- W3035332806 cites W2965658867 @default.
- W3035332806 cites W2967733054 @default.
- W3035332806 cites W2980137827 @default.
- W3035332806 cites W2984618279 @default.
- W3035332806 doi "https://doi.org/10.1109/cvpr42600.2020.00215" @default.
- W3035332806 hasPublicationYear "2020" @default.
- W3035332806 type Work @default.
- W3035332806 sameAs 3035332806 @default.
- W3035332806 citedByCount "91" @default.
- W3035332806 countsByYear W30353328062020 @default.
- W3035332806 countsByYear W30353328062021 @default.
- W3035332806 countsByYear W30353328062022 @default.
- W3035332806 countsByYear W30353328062023 @default.
- W3035332806 crossrefType "proceedings-article" @default.
- W3035332806 hasAuthorship W3035332806A5021398655 @default.
- W3035332806 hasAuthorship W3035332806A5022883365 @default.
- W3035332806 hasAuthorship W3035332806A5024743366 @default.
- W3035332806 hasAuthorship W3035332806A5052596733 @default.
- W3035332806 hasAuthorship W3035332806A5065001783 @default.
- W3035332806 hasAuthorship W3035332806A5070926896 @default.
- W3035332806 hasAuthorship W3035332806A5073936760 @default.
- W3035332806 hasAuthorship W3035332806A5087929301 @default.
- W3035332806 hasBestOaLocation W30353328062 @default.
- W3035332806 hasConcept C108010975 @default.
- W3035332806 hasConcept C113775141 @default.
- W3035332806 hasConcept C11413529 @default.
- W3035332806 hasConcept C119857082 @default.
- W3035332806 hasConcept C154945302 @default.
- W3035332806 hasConcept C2776214188 @default.
- W3035332806 hasConcept C28855332 @default.
- W3035332806 hasConcept C41008148 @default.
- W3035332806 hasConcept C50644808 @default.
- W3035332806 hasConcept C6557445 @default.
- W3035332806 hasConcept C86803240 @default.
- W3035332806 hasConceptScore W3035332806C108010975 @default.
- W3035332806 hasConceptScore W3035332806C113775141 @default.
- W3035332806 hasConceptScore W3035332806C11413529 @default.
- W3035332806 hasConceptScore W3035332806C119857082 @default.
- W3035332806 hasConceptScore W3035332806C154945302 @default.
- W3035332806 hasConceptScore W3035332806C2776214188 @default.
- W3035332806 hasConceptScore W3035332806C28855332 @default.
- W3035332806 hasConceptScore W3035332806C41008148 @default.
- W3035332806 hasConceptScore W3035332806C50644808 @default.
- W3035332806 hasConceptScore W3035332806C6557445 @default.
- W3035332806 hasConceptScore W3035332806C86803240 @default.
- W3035332806 hasLocation W30353328061 @default.
- W3035332806 hasLocation W30353328062 @default.
- W3035332806 hasOpenAccess W3035332806 @default.
- W3035332806 hasPrimaryLocation W30353328061 @default.
- W3035332806 hasRelatedWork W11097404 @default.
- W3035332806 hasRelatedWork W13977875 @default.
- W3035332806 hasRelatedWork W14600564 @default.
- W3035332806 hasRelatedWork W2805907 @default.
- W3035332806 hasRelatedWork W5144792 @default.
- W3035332806 hasRelatedWork W5477828 @default.
- W3035332806 hasRelatedWork W807719 @default.
- W3035332806 hasRelatedWork W8499301 @default.
- W3035332806 hasRelatedWork W8595301 @default.
- W3035332806 hasRelatedWork W13734992 @default.
- W3035332806 isParatext "false" @default.
- W3035332806 isRetracted "false" @default.
- W3035332806 magId "3035332806" @default.
- W3035332806 workType "article" @default.