{"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/on-the-variance-of-the-adaptive-learning-rate","title":"On the Variance of the Adaptive Learning Rate and Beyond","arxiv_id":"1908.03265","date":"2019-08-08","proceeding":"ICLR 2020 1","authors":["Liyuan Liu","Haoming Jiang","Pengcheng He","Weizhu Chen","Xiaodong Liu","Jianfeng Gao","Jiawei Han"],"abstract":"The learning rate warmup heuristic achieves remarkable success in stabilizing training, accelerating convergence and improving generalization for adaptive stochastic optimization algorithms like RMSprop and Adam. Here, we study its mechanism in details. Pursuing the theory behind warmup, we identify a problem of the adaptive learning rate (i.e., it has problematically large variance in the early stage), suggest warmup works as a variance reduction technique, and provide both empirical and theoretical evidence to verify our hypothesis. We further propose RAdam, a new variant of Adam, by introducing a term to rectify the variance of the adaptive learning rate. Extensive experimental results on image classification, language modeling, and neural machine translation verify our intuition and demonstrate the effectiveness and robustness of our proposed method. All implementations are available at: https://github.com/LiyuanLucasLiu/RAdam.","url_abs":"https://arxiv.org/abs/1908.03265v4","url_pdf":"https://arxiv.org/pdf/1908.03265v4.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":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/LiyuanLucasLiu/RAdam","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/201419/Optimizer-PyTorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/AnirudhMaiya/RAdam","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/HamadYA/GhostFaceNets","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/Kitsunetic/Optimizer-Visualization","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/SdahlSean/RangerOptimizerTensorflow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/Tony-Y/pytorch_warmup","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/allen108108/Model-Optimizer_Implementation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/bond005/yandex-shifts-weather","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/chunhuililili/mt_dnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/float256/rectified-adam-keras","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"GPL-3.0"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/frgfm/Holocron","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/kpe/params-flow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/microsoft/MT-DNN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/miyamankyushu/DCGAN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/mnikitin/RAdam","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"mxnet","reach":{"status":"ok"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/nachiket273/lookahead_pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/namisan/mt-dnn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/shun601/4th-tellus-satellite-challenge","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/MaximeVandegar/Papers-in-100-Lines-of-Code/tree/main/On_the_Variance_of_the_Adaptive_Learning_Rate_and_Beyond","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"on-the-variance-of-the-adaptive-learning-rate","repo_url":"https://github.com/labmlai/annotated_deep_learning_paper_implementations","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"language-modeling","task_name":"Language Modeling"},{"task_slug":"language-modelling","task_name":"Language Modelling"},{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"stochastic-optimization","task_name":"Stochastic Optimization"},{"task_slug":"translation","task_name":"Translation"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[{"method_slug":"adam","method_name":"Adam"},{"method_slug":"radam","method_name":"RAdam"},{"method_slug":"rmsprop","method_name":"RMSProp"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1908.03265","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1908.03265"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/shun601/4th-tellus-satellite-challenge","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/kpe/params-flow","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/mnikitin/RAdam","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/frgfm/Holocron","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/namisan/mt-dnn","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/LiyuanLucasLiu/RAdam","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/float256/rectified-adam-keras","reach":{"status":"ok","spdx":"GPL-3.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/Tony-Y/pytorch_warmup","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/Kitsunetic/Optimizer-Visualization","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/bond005/yandex-shifts-weather","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/MaximeVandegar/Papers-in-100-Lines-of-Code/tree/main/On_the_Variance_of_the_Adaptive_Learning_Rate_and_Beyond","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/SdahlSean/RangerOptimizerTensorflow","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/nachiket273/lookahead_pytorch","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/allen108108/Model-Optimizer_Implementation","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/201419/Optimizer-PyTorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/AnirudhMaiya/RAdam","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/labmlai/annotated_deep_learning_paper_implementations","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/microsoft/MT-DNN","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/miyamankyushu/DCGAN","reach":{"status":"ok"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/chunhuililili/mt_dnn","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/HamadYA/GhostFaceNets","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran":1,"ran_draft_wrong":1,"ran_honours":2,"unverified":9},"by_repo_kind":{"official":{"samples":5,"ran":2,"repositories":1},"listed":{"samples":8,"ran":2,"repositories":2}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"2c92cec199cf12cd","entry":"get_mean_and_std","repo":"LiyuanLucasLiu/RAdam","repo_kind":"official","path":"cifar_imagenet/utils/misc.py","file_url":"https://github.com/LiyuanLucasLiu/RAdam/blob/HEAD/cifar_imagenet/utils/misc.py","link_basis":"plan_row","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"2c92cec199cf12cd"}},{"code_sha256_prefix":"50afa2863ffd9fde","entry":"plot_overlap","repo":"LiyuanLucasLiu/RAdam","repo_kind":"official","path":"cifar_imagenet/utils/logger.py","file_url":"https://github.com/LiyuanLucasLiu/RAdam/blob/HEAD/cifar_imagenet/utils/logger.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"50afa2863ffd9fde"}},{"code_sha256_prefix":"d979441c66cfc515","entry":"scaled_sign","repo":"201419/Optimizer-PyTorch","repo_kind":"listed","path":"errorfeedbacksgd.py","file_url":"https://github.com/201419/Optimizer-PyTorch/blob/HEAD/errorfeedbacksgd.py","link_basis":"harvester_set","language":"python","status":"ran_honours","verification_level":1,"contract_check":"HONOURS","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"d979441c66cfc515"}},{"code_sha256_prefix":"f45a2479358fd34b","entry":"unscaled_sign","repo":"201419/Optimizer-PyTorch","repo_kind":"listed","path":"errorfeedbacksgd.py","file_url":"https://github.com/201419/Optimizer-PyTorch/blob/HEAD/errorfeedbacksgd.py","link_basis":"harvester_set","language":"python","status":"ran_honours","verification_level":1,"contract_check":"HONOURS","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"f45a2479358fd34b"}},{"code_sha256_prefix":"c87e7665fde6cdfc","entry":"colorize","repo":"LiyuanLucasLiu/RAdam","repo_kind":"official","path":"cifar_imagenet/utils/visualize.py","file_url":"https://github.com/LiyuanLucasLiu/RAdam/blob/HEAD/cifar_imagenet/utils/visualize.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"c87e7665fde6cdfc"}},{"code_sha256_prefix":"d97f2e791492ba33","entry":"create_one_cycle_lr_scheduler","repo":"kpe/params-flow","repo_kind":"listed","path":"params_flow/utils/learn_scheduler.py","file_url":"https://github.com/kpe/params-flow/blob/HEAD/params_flow/utils/learn_scheduler.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"d97f2e791492ba33"}},{"code_sha256_prefix":"1c47760cdb987d2d","entry":"fetch_url","repo":"kpe/params-flow","repo_kind":"listed","path":"params_flow/utils/fetch_unpack.py","file_url":"https://github.com/kpe/params-flow/blob/HEAD/params_flow/utils/fetch_unpack.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"1c47760cdb987d2d"}},{"code_sha256_prefix":"d66f0a3cbb86204c","entry":"gauss","repo":"LiyuanLucasLiu/RAdam","repo_kind":"official","path":"cifar_imagenet/utils/visualize.py","file_url":"https://github.com/LiyuanLucasLiu/RAdam/blob/HEAD/cifar_imagenet/utils/visualize.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"d66f0a3cbb86204c"}},{"code_sha256_prefix":"e6d5b83d23a4800c","entry":"gelu","repo":"kpe/params-flow","repo_kind":"listed","path":"params_flow/activations.py","file_url":"https://github.com/kpe/params-flow/blob/HEAD/params_flow/activations.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"e6d5b83d23a4800c"}},{"code_sha256_prefix":"3dbb04436d7c7aa9","entry":"gelu_exact","repo":"kpe/params-flow","repo_kind":"listed","path":"params_flow/activations.py","file_url":"https://github.com/kpe/params-flow/blob/HEAD/params_flow/activations.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"3dbb04436d7c7aa9"}},{"code_sha256_prefix":"528aa6f214e15748","entry":"get_activation","repo":"kpe/params-flow","repo_kind":"listed","path":"params_flow/activations.py","file_url":"https://github.com/kpe/params-flow/blob/HEAD/params_flow/activations.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"528aa6f214e15748"}},{"code_sha256_prefix":"1394422d51c4b126","entry":"make_image","repo":"LiyuanLucasLiu/RAdam","repo_kind":"official","path":"cifar_imagenet/utils/visualize.py","file_url":"https://github.com/LiyuanLucasLiu/RAdam/blob/HEAD/cifar_imagenet/utils/visualize.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"1394422d51c4b126"}},{"code_sha256_prefix":"335085134ae18f02","entry":"unpack_archive","repo":"kpe/params-flow","repo_kind":"listed","path":"params_flow/utils/fetch_unpack.py","file_url":"https://github.com/kpe/params-flow/blob/HEAD/params_flow/utils/fetch_unpack.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"335085134ae18f02"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}