{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,30]],"date-time":"2026-06-30T15:37:42Z","timestamp":1782833862389,"version":"3.54.5"},"reference-count":57,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"1","license":[{"start":{"date-parts":[[2024,9,1]],"date-time":"2024-09-01T00:00:00Z","timestamp":1725148800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2024,9,1]],"date-time":"2024-09-01T00:00:00Z","timestamp":1725148800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2024,9,1]],"date-time":"2024-09-01T00:00:00Z","timestamp":1725148800000},"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 Trans. Circuits Syst. Artif. Intel."],"published-print":{"date-parts":[[2024,9]]},"DOI":"10.1109\/tcasai.2024.3459412","type":"journal-article","created":{"date-parts":[[2024,9,11]],"date-time":"2024-09-11T17:42:53Z","timestamp":1726076573000},"page":"37-52","source":"Crossref","is-referenced-by-count":7,"title":["Optimizing Foundation Model Inference on a Many-Tiny-Core Open-Source RISC-V Platform"],"prefix":"10.1109","volume":"1","author":[{"ORCID":"https:\/\/orcid.org\/0009-0004-9412-6081","authenticated-orcid":false,"given":"Viviane","family":"Potocnik","sequence":"first","affiliation":[{"name":"Integrated Systems Laboratory (IIS) of ETH Zürich, ETZ, Züurich, Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7986-1975","authenticated-orcid":false,"given":"Luca","family":"Colagrande","sequence":"additional","affiliation":[{"name":"Integrated Systems Laboratory (IIS) of ETH Zürich, ETZ, Züurich, Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0007-9700-1286","authenticated-orcid":false,"given":"Tim","family":"Fischer","sequence":"additional","affiliation":[{"name":"Integrated Systems Laboratory (IIS) of ETH Zürich, ETZ, Züurich, Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-3011-6368","authenticated-orcid":false,"given":"Luca","family":"Bertaccini","sequence":"additional","affiliation":[{"name":"Integrated Systems Laboratory (IIS) of ETH Zürich, ETZ, Züurich, Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2872-7071","authenticated-orcid":false,"given":"Daniele Jahier","family":"Pagliari","sequence":"additional","affiliation":[{"name":"Department of Control and Computer Engineering, Politecnico di Torino, Turin, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-6215-8220","authenticated-orcid":false,"given":"Alessio","family":"Burrello","sequence":"additional","affiliation":[{"name":"Department of Electrical, Electronic and Information Engineering, University of Bologna, Bologna, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8068-3806","authenticated-orcid":false,"given":"Luca","family":"Benini","sequence":"additional","affiliation":[{"name":"Integrated Systems Laboratory (IIS) of ETH Zürich, ETZ, Züurich, Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"A survey of large language models","author":"Zhao","year":"2023"},{"key":"ref2","article-title":"An image is worth 16x16 words: Transformers for image recognition at scale","author":"Dosovitskiy","year":"2020"},{"key":"ref3","article-title":"A comprehensive survey on pretrained foundation models: A history from BERT to ChatGPT","author":"Zhou","year":"2023"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"},{"key":"ref5","article-title":"On the opportunities and risks of foundation models","author":"Bommasani","year":"2022"},{"key":"ref6","first-page":"1877","article-title":"Language models are few-shot learners","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"33,","author":"Brown","year":"2020"},{"key":"ref7","article-title":"Whisper","year":"2024"},{"key":"ref8","article-title":"SparQ attention: Bandwidth-efficient LLM inference","author":"Ribar","year":"2024"},{"key":"ref9","first-page":"5156","article-title":"Transformers are RNNs: Fast autoregressive transformers with linear attention","volume-title":"Proc. 37th Int. Conf. Mach. Learn., (ICML)","volume":"119","author":"Katharopoulos","year":"2020"},{"key":"ref10","article-title":"FlashAttention-2: Faster attention with better parallelism and work partitioning","author":"Dao","year":"2023"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1109\/MM.2021.3061394"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/MM.2023.3256796"},{"key":"ref13","article-title":"Cerebras-GPT: Open compute-optimal language models trained on the cerebras wafer-scale cluster","author":"Dey","year":"2023"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1145\/3470496.3527405"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/TC.2023.3329930"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1109\/TC.2020.3027900"},{"key":"ref17","article-title":"A study of BFLOAT16 for deep learning training","author":"Kalamkar","year":"2019"},{"key":"ref18","article-title":"FP8 formats for deep learning","author":"Micikevicius","year":"2022"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1810.04805"},{"key":"ref20","article-title":"Llama 2: Open foundation and fine-tuned chat models","author":"Touvron","year":"2023"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1109\/WACV57701.2024.00127"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.1109\/JAS.2023.123618"},{"key":"ref23","first-page":"30318","article-title":"GPT3.int8(): 8-bit matrix multiplication for transformers at scale","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"35","author":"Dettmers","year":"2022"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1109\/iccv48922.2021.00064"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00009"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D18-2012"},{"key":"ref27","article-title":"Efficiently scaling transformer inference","volume-title":"Proc. Mach. Learn. Syst.","volume":"5","author":"Pope","year":"2023"},{"key":"ref28","article-title":"MobileLLM: Optimizing sub-billion parameter language models for on-device use cases","author":"Liu","year":"2024"},{"key":"ref29","first-page":"2441","article-title":"Luna: Linear unified nested attention","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"34","author":"Ma","year":"2021"},{"key":"ref30","article-title":"Linformer: Self-attention with linear complexity","author":"Wang","year":"2020"},{"key":"ref31","first-page":"52342","article-title":"Scissorhands: Exploiting the persistence of importance hypothesis for LLM KV cache compression at test time","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"36","author":"Liu","year":"2023"},{"key":"ref32","first-page":"10088","article-title":"QLoRA: Efficient finetuning of quantized LLMs","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"36","author":"Dettmers","year":"2023"},{"key":"ref33","article-title":"NVIDIA data center deep learning product performance AI inference","year":"2024"},{"key":"ref34","article-title":"SambaNova DataScale\u00ae \u2014 The AI Platform for Innovation","author":"Systems","year":"2024"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1109\/MM.2023.3256384"},{"key":"ref36","article-title":"Intel Gaudi 2 neural network deep learning inference processor","year":"2024"},{"key":"ref37","article-title":"Intel Gaudi 3 AI accelerator","year":"2024"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1109\/HCS55958.2022.9895477"},{"issue":"11","key":"ref39","first-page":"4038","article-title":"AccelTran: A sparsity-aware accelerator for dynamic inference with transformers","volume-title":"IEEE Trans. Comput.-Aided Des. Integr. Circuits Syst.","volume":"42","author":"Tuli","year":"2023"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1109\/JSSC.2022.3213521"},{"key":"ref41","first-page":"368","article-title":"20.5 C-Transformer: A 2.6-18.1$\\mu$J\/Token homogeneous DNN-transformer\/spiking-transformer processor with big-little network and implicit weight generation for large language models","volume-title":"Proc. IEEE Int. Solid-State Circuits Conf. (ISSCC)","author":"Kim","year":"2024"},{"issue":"3","key":"ref42","first-page":"855","article-title":"FreFlex: A high-performance processor for convolution and attention computations via sparsity-adaptive dynamic frequency boosting","volume-title":"IEEE J. Solid-State Circuits","volume":"59","author":"Sundaram","year":"2024"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/jssc.2024.3397189"},{"key":"ref44","first-page":"342","article-title":"22.9 A 12nm 18.1TFLOPs\/W sparse transformer processor with entropy-based early exit, mixed-precision predication and fine-grained power management","volume-title":"Proc. IEEE Int. Solid-State Circuits Conf. (ISSCC)","author":"Tambe","year":"2023"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.1109\/HCS52781.2021.9567075"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1109\/VLSITechnologyandCir46783.2024.10631529"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1109\/tetc.2024.3365354"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1109\/ARITH.2019.00019"},{"key":"ref49","first-page":"p. 7686\u20137695","article-title":"Training deep neural networks with 8-bit floating point numbers","volume-title":"Proc. 32nd Int. Conf. Neural Inf. Process. Syst., (NIPS)","author":"Wang","year":"2018"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1109\/MM.2020.3045564"},{"key":"ref51","article-title":"HBM2E","year":"2024"},{"key":"ref52","article-title":"GTC: What\u2019s new? Grace, grace hopper, infiniband, arm HPC software ecosystem","year":"2024"},{"key":"ref53","article-title":"I-BERT: Integer-only BERT quantization","author":"Kim","year":"2021"},{"key":"ref54","first-page":"12934","article-title":"Efficientformer: Vision transformers at mobilenet speed","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"35","author":"Li","year":"2022"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.1109\/TPDS.2023.3322029"},{"key":"ref56","article-title":"A comprehensive performance study of large language models on novel AI accelerators","author":"Emani","year":"2023"},{"key":"ref57","article-title":"Benchmarking large language models on NVIDIA H100 GPUs with CoreWeave (Part 1)","year":"2023"}],"container-title":["IEEE Transactions on Circuits and Systems for Artificial Intelligence"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/10495160\/10584333\/10678895.pdf?arnumber=10678895","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,11,27]],"date-time":"2024-11-27T19:31:54Z","timestamp":1732735914000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/10678895\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,9]]},"references-count":57,"journal-issue":{"issue":"1"},"URL":"https:\/\/doi.org\/10.1109\/tcasai.2024.3459412","relation":{},"ISSN":["2996-6647"],"issn-type":[{"value":"2996-6647","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,9]]}}}