{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/online-normalizer-calculation-for-softmax","title":"Online normalizer calculation for softmax","arxiv_id":"1805.02867","date":"2018-05-08","proceeding":null,"authors":["Maxim Milakov","Natalia Gimelshein"],"abstract":"The Softmax function is ubiquitous in machine learning, multiple previous\nworks suggested faster alternatives for it. In this paper we propose a way to\ncompute classical Softmax with fewer memory accesses and hypothesize that this\nreduction in memory accesses should improve Softmax performance on actual\nhardware. The benchmarks confirm this hypothesis: Softmax accelerates by up to\n1.3x and Softmax+TopK combined and fused by up to 5x.","url_abs":"http://arxiv.org/abs/1805.02867v2","url_pdf":"http://arxiv.org/pdf/1805.02867v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"online-normalizer-calculation-for-softmax","repo_url":"https://github.com/NVIDIA/online-softmax","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"BSD-3-Clause"}}],"tasks":[{"task_slug":"machine-learning","task_name":"BIG-bench Machine Learning"}],"methods":[{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1805.02867","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}