{"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/zero-offload-democratizing-billion-scale","title":"ZeRO-Offload: Democratizing Billion-Scale Model Training","arxiv_id":"2101.06840","date":"2021-01-18","proceeding":null,"authors":["Jie Ren","Samyam Rajbhandari","Reza Yazdani Aminabadi","Olatunji Ruwase","Shuangyan Yang","Minjia Zhang","Dong Li","Yuxiong He"],"abstract":"Large-scale model training has been a playing ground for a limited few requiring complex model refactoring and access to prohibitively expensive GPU clusters. ZeRO-Offload changes the large model training landscape by making large model training accessible to nearly everyone. It can train models with over 13 billion parameters on a single GPU, a 10x increase in size compared to popular framework such as PyTorch, and it does so without requiring any model change from the data scientists or sacrificing computational efficiency. ZeRO-Offload enables large model training by offloading data and compute to CPU. To preserve compute efficiency, it is designed to minimize the data movement to/from GPU, and reduce CPU compute time while maximizing memory savings on GPU. As a result, ZeRO-Offload can achieve 40 TFlops/GPU on a single NVIDIA V100 GPU for 10B parameter model compared to 30TF using PyTorch alone for a 1.4B parameter model, the largest that can be trained without running out of memory. ZeRO-Offload is also designed to scale on multiple-GPUs when available, offering near linear speedup on up to 128 GPUs. Additionally, it can work together with model parallelism to train models with over 70 billion parameters on a single DGX-2 box, a 4.5x increase in model size compared to using model parallelism alone. By combining compute and memory efficiency with ease-of-use, ZeRO-Offload democratizes large-scale model training making it accessible to even data scientists with access to just a single GPU.","url_abs":"https://arxiv.org/abs/2101.06840v1","url_pdf":"https://arxiv.org/pdf/2101.06840v1.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":"zero-offload-democratizing-billion-scale","repo_url":"https://github.com/microsoft/DeepSpeed","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"zero-offload-democratizing-billion-scale","repo_url":"https://github.com/ShadenSmith/ghpages-test","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"zero-offload-democratizing-billion-scale","repo_url":"https://github.com/facebookresearch/fairscale","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":null,"task_name":"CPU"},{"task_slug":"computational-efficiency","task_name":"Computational Efficiency"},{"task_slug":null,"task_name":"GPU"},{"task_slug":"model","task_name":"model"}],"methods":[{"method_slug":"zero-offload","method_name":"ZeRO-Offload"}],"datasets_introduced":[],"methods_introduced":[{"slug":"zero-offload","name":"ZeRO-Offload","full_name":"ZeRO-Offload"}],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2101.06840","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}