{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,15]],"date-time":"2026-07-15T22:56:28Z","timestamp":1784156188182,"version":"3.55.0"},"reference-count":63,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"4","license":[{"start":{"date-parts":[[2023,4,1]],"date-time":"2023-04-01T00:00:00Z","timestamp":1680307200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2023,4,1]],"date-time":"2023-04-01T00:00:00Z","timestamp":1680307200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2023,4,1]],"date-time":"2023-04-01T00:00:00Z","timestamp":1680307200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE J. Solid-State Circuits"],"published-print":{"date-parts":[[2023,4]]},"DOI":"10.1109\/jssc.2023.3234893","type":"journal-article","created":{"date-parts":[[2023,1,18]],"date-time":"2023-01-18T18:44:59Z","timestamp":1674067499000},"page":"1129-1141","source":"Crossref","is-referenced-by-count":61,"title":["A 95.6-TOPS\/W Deep Learning Inference Accelerator With Per-Vector Scaled 4-bit Quantization in 5 nm"],"prefix":"10.1109","volume":"58","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-8117-1412","authenticated-orcid":false,"given":"Ben","family":"Keller","sequence":"first","affiliation":[{"name":"NVIDIA, Inc., Santa Clara, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Rangharajan","family":"Venkatesan","sequence":"additional","affiliation":[{"name":"NVIDIA, Inc., Santa Clara, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5045-1964","authenticated-orcid":false,"given":"Steve","family":"Dai","sequence":"additional","affiliation":[{"name":"NVIDIA, Inc., Santa Clara, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Stephen G.","family":"Tell","sequence":"additional","affiliation":[{"name":"NVIDIA, Inc., Durham, NC, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9997-3141","authenticated-orcid":false,"given":"Brian","family":"Zimmer","sequence":"additional","affiliation":[{"name":"NVIDIA, Inc., Santa Clara, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Charbel","family":"Sakr","sequence":"additional","affiliation":[{"name":"NVIDIA, Inc., Santa Clara, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4632-2876","authenticated-orcid":false,"given":"William J.","family":"Dally","sequence":"additional","affiliation":[{"name":"NVIDIA, Inc., Santa Clara, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5137-5617","authenticated-orcid":false,"given":"C. Thomas","family":"Gray","sequence":"additional","affiliation":[{"name":"NVIDIA, Inc., Durham, NC, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7584-3489","authenticated-orcid":false,"given":"Brucek","family":"Khailany","sequence":"additional","affiliation":[{"name":"NVIDIA, Inc., Austin, TX, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref2","article-title":"BERT: Pre-training of deep bidirectional transformers for language understanding","volume-title":"arXiv:1810.04805","author":"Devlin","year":"2018"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-58452-8_24"},{"key":"ref4","article-title":"Explaining how a deep neural network trained with end-to-end learning steers a car","volume-title":"arxiv:1704.07911","author":"Bojarski","year":"2017"},{"key":"ref5","doi-asserted-by":"publisher","DOI":"10.1145\/3065386"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"},{"key":"ref7","article-title":"Megatron-LM: Training multi-billion parameter language models using model parallelism","volume-title":"arXiv:1909.08053","author":"Shoeybi","year":"2019"},{"key":"ref8","first-page":"1","article-title":"An image is worth 16\u00d716 words: Transformers for image recognition at scale","volume-title":"Proc. Int. Conf. Learn. Represent. (ICLR)","author":"Dosovitskiy"},{"key":"ref9","article-title":"DoReFa-net: Training low bitwidth convolutional neural networks with low bitwidth gradients","volume-title":"arXiv:1606.06160","author":"Zhou","year":"2016"},{"key":"ref10","first-page":"3007","article-title":"Analytical guarantees on numerical precision of deep neural networks","volume-title":"Proc. 34th Int. Conf. Mach. Learn.","author":"Sakr"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/isscc42613.2021.9365791"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1145\/3079856.3080254"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/VLSITechnologyandCir46769.2022.9830458"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/ISSCC42614.2022.9731716"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/isscc42614.2022.9731686"},{"key":"ref17","article-title":"Integer quantization for deep learning inference: Principles and empirical evaluation","volume-title":"arXiv:2004.09602","author":"Wu","year":"2020"},{"key":"ref18","first-page":"6105","article-title":"EfficientNet: Rethinking model scaling for convolutional neural networks","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Tan"},{"key":"ref19","first-page":"10096","article-title":"EfficientNetv2: Smaller models and faster training","volume-title":"Proc. Int. Conf. Mach. Learn. (ICML)","author":"Tan"},{"key":"ref21","first-page":"873","article-title":"VS-Quant: Per-vector scaled quantization for accurate low-precision neural network inference","volume-title":"Proc. Conf. Mach. Learn. Syst. (MLSys)","author":"Dai"},{"key":"ref22","first-page":"10347","article-title":"Training data-efficient image transformers & distillation through attention","volume-title":"Proc. 38th Int. Conf. Mach. Learn. (ICML)","author":"Touvron"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1109\/isscc42614.2022.9731639"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1109\/ISSCC42614.2022.9731668"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/VLSITechnologyandCir46769.2022.9830153"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1109\/isscc42613.2021.9365943"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/isscc19947.2020.9063111"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1145\/3352460.3358302"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/isscc42614.2022.9731659"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1109\/ISSCC42614.2022.9731654"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.23919\/VLSICircuits52068.2021.9492476"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.23919\/VLSICircuits52068.2021.9492403"},{"key":"ref33","article-title":"DistilBERT, a distilled version of BERT: Smaller, faster, cheaper and lighter","volume-title":"arXiv:1910.01108","author":"Sanh","year":"2019"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/N19-1133"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1109\/IJCNN55064.2022.9892797"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA47549.2020.00035"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA52012.2021.00060"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1145\/3489517.3530505"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1109\/EMC2-NIPS53020.2019.00016"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.24963\/ijcai.2020\/520"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA51647.2021.00018"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1145\/3534678.3539260"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1145\/3470496.3527423"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1109\/isscc42614.2022.9731645"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.1145\/3466752.3480095"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1109\/SOCC49529.2020.9524802"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1109\/ICCAD45719.2019.8942127"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1145\/3297858.3304025"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1109\/DAC18074.2021.9586134"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1109\/MM.2016.11"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.1109\/DAC.2018.8465897"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.1109\/TVLSI.2017.2776954"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1145\/3352460.3358260"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.1109\/MSSC.2019.2922889"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.1109\/JPROC.2017.2761740"},{"key":"ref56","article-title":"Estimating or propagating gradients through stochastic neurons for conditional computation","volume-title":"arXiv:1308.3432","author":"Bengio","year":"2013"},{"key":"ref57","first-page":"8026","article-title":"PyTorch: An imperative style, high-performance deep learning library","volume-title":"Proc. Conf. Neural Inf. Process. Syst. (NIPS)","author":"Paszke"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1109\/VLSITechnologyandCir46769.2022.9830277"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1109\/JSSC.2019.2960488"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D16-1264"},{"key":"ref61","first-page":"1","article-title":"Adam: A method for stochastic optimization","volume-title":"Proc. Int. Conf. Learn. Represent. (ICLR)","author":"Kingma"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2009.5206848"},{"key":"ref63","first-page":"19123","article-title":"Optimal clipping and magnitude-aware differentiation for improved quantization-aware training","volume-title":"Proc. Int. Conf. Mach. Learn. (ICML)","author":"Sakr"},{"key":"ref64","article-title":"PACT: Parameterized clipping activation for quantized neural networks","volume-title":"arXiv:1805.06085","author":"Choi","year":"2018"},{"key":"ref65","first-page":"1135","article-title":"Learning both weights and connections for efficient neural networks","volume-title":"Proc. Int. Conf. Neural Inf. Process. Syst.","author":"Han"}],"container-title":["IEEE Journal of Solid-State Circuits"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/4\/10082904\/10019275.pdf?arnumber=10019275","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,2,13]],"date-time":"2024-02-13T06:41:19Z","timestamp":1707806479000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/10019275\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,4]]},"references-count":63,"journal-issue":{"issue":"4"},"URL":"https:\/\/doi.org\/10.1109\/jssc.2023.3234893","relation":{},"ISSN":["0018-9200","1558-173X"],"issn-type":[{"value":"0018-9200","type":"print"},{"value":"1558-173X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2023,4]]}}}