{"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/compression-of-neural-machine-translation","title":"Compression of Neural Machine Translation Models via Pruning","arxiv_id":"1606.09274","date":"2016-06-29","proceeding":"CONLL 2016 8","authors":["Abigail See","Minh-Thang Luong","Christopher D. Manning"],"abstract":"Neural Machine Translation (NMT), like many other deep learning domains,\ntypically suffers from over-parameterization, resulting in large storage sizes.\nThis paper examines three simple magnitude-based pruning schemes to compress\nNMT models, namely class-blind, class-uniform, and class-distribution, which\ndiffer in terms of how pruning thresholds are computed for the different\nclasses of weights in the NMT architecture. We demonstrate the efficacy of\nweight pruning as a compression technique for a state-of-the-art NMT system. We\nshow that an NMT model with over 200 million parameters can be pruned by 40%\nwith very little performance loss as measured on the WMT'14 English-German\ntranslation task. This sheds light on the distribution of redundancy in the NMT\narchitecture. Our main result is that with retraining, we can recover and even\nsurpass the original performance with an 80%-pruned model.","url_abs":"http://arxiv.org/abs/1606.09274v1","url_pdf":"http://arxiv.org/pdf/1606.09274v1.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":"compression-of-neural-machine-translation","repo_url":"https://github.com/zepx/pytorch-weight-prune","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"nmt","task_name":"NMT"},{"task_slug":"translation","task_name":"Translation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1606.09274","atlas_url":"https://app.syntology.ai/?focus=1606.09274","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}