{"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/sift-sparse-iso-flop-transformations-for","title":"Sparse-IFT: Sparse Iso-FLOP Transformations for Maximizing Training Efficiency","arxiv_id":"2303.11525","date":"2023-03-21","proceeding":null,"authors":["Vithursan Thangarasa","Shreyas Saxena","Abhay Gupta","Sean Lie"],"abstract":"Recent research has focused on weight sparsity in deep neural network training to reduce FLOPs, aiming for improved efficiency (test accuracy w.r.t training FLOPs). However, sparse weight training often compromises accuracy, requiring extended training schedules to attain the accuracy of dense models. In contrast, our approach, Sparse Iso-FLOP Transformations (Sparse-IFT), uses sparsity to improve accuracy while maintaining dense model FLOPs. Using a single hyperparameter (i.e., the sparsity level), Sparse-IFTs efficiently replace dense layers, expanding the search space for optimal sparse masks. In addition, dynamic sparse training (DST) with Sparse-IFT models effectively navigate this larger sparse mask-weight space, which is evidenced by a spectral analysis using Ramanujan graph properties. Our study reveals a robust correlation among mask topology, weights, and final performance. Notably, without adjusting any training hyperparameters, replacing dense layers with Sparse-IFT yields significant improvements, such as a +3.5% boost for ResNet-18 on ImageNet and +0.9% for GPT-3 Small on the Open LLM leaderboard. To the best of our knowledge, this is the first work to demonstrate the use of sparsity for improving the accuracy of dense models through a set of simple-to-use sparse transformations. Code is available at: https://github.com/CerebrasResearch/Sparse-IFT.","url_abs":"https://arxiv.org/abs/2303.11525v4","url_pdf":"https://arxiv.org/pdf/2303.11525v4.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":"sift-sparse-iso-flop-transformations-for","repo_url":"https://github.com/cerebrasresearch/sift","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"sift-sparse-iso-flop-transformations-for","repo_url":"https://github.com/cerebrasresearch/sparse-ift","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"navigate","task_name":"Navigate"}],"methods":[{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"attention-dropout","method_name":"Attention Dropout"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"cosine-annealing","method_name":"Cosine Annealing"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"gpt-3","method_name":"GPT-3"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"linear-warmup-with-cosine-annealing","method_name":"Linear Warmup With Cosine Annealing"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"weight-decay","method_name":"Weight Decay"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2303.11525","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2303.11525"}},"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":"deterministic:regex_extraction","url":"https://github.com/CerebrasResearch/Sparse-IFT","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/cerebrasresearch/sparse-ift","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/cerebrasresearch/sift","reach":{"status":"ok","spdx":"Apache-2.0"}}],"summary":{"ran_fixture":1,"unverified":23},"by_repo_kind":{"official":{"samples":24,"ran":1,"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":"0d95e28fac98427f","entry":"accuracy","repo":"cerebrasresearch/sift","repo_kind":"official","path":"ComputerVision/ImageNet/utils/imagenet_utils.py","file_url":"https://github.com/cerebrasresearch/sift/blob/HEAD/ComputerVision/ImageNet/utils/imagenet_utils.py","link_basis":"harvester_set","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":"invariant","behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"0d95e28fac98427f"}},{"code_sha256_prefix":"8e0b681246bf0f5c","entry":"add_drop_connect_specific_args","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/dropconnect/utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/dropconnect/utils.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":"8e0b681246bf0f5c"}},{"code_sha256_prefix":"3d8211706172e297","entry":"add_mask_optimizer_specific_args","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/utils/argparse_utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/utils/argparse_utils.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":"3d8211706172e297"}},{"code_sha256_prefix":"b77615413ecbb1c6","entry":"add_mask_optimizer_specific_args_cv","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/utils/argparse_utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/utils/argparse_utils.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":"b77615413ecbb1c6"}},{"code_sha256_prefix":"428bfab7ca67279f","entry":"calc_sparsity","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/optimizers/utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/optimizers/utils.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":"428bfab7ca67279f"}},{"code_sha256_prefix":"3aa3c043ad3d461f","entry":"count_to_string","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/utils/profiler_utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/utils/profiler_utils.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":"3aa3c043ad3d461f"}},{"code_sha256_prefix":"60ad7d8d0303d9a5","entry":"create_run_command","repo":"cerebrasresearch/sift","repo_kind":"official","path":"ComputerVision/launch_utils/prepare_job_commands.py","file_url":"https://github.com/cerebrasresearch/sift/blob/HEAD/ComputerVision/launch_utils/prepare_job_commands.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":"60ad7d8d0303d9a5"}},{"code_sha256_prefix":"9ac3b31a2048c5ee","entry":"extract_number","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/optimizers/utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/optimizers/utils.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":"9ac3b31a2048c5ee"}},{"code_sha256_prefix":"411a869a5eb7061b","entry":"generic_cv_should_dropconnect","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/dropconnect/utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/dropconnect/utils.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":"411a869a5eb7061b"}},{"code_sha256_prefix":"7ece5276bcdc838d","entry":"generic_cv_should_sparsify","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/utils/should_sparsify_utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/utils/should_sparsify_utils.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":"7ece5276bcdc838d"}},{"code_sha256_prefix":"f131b1b13ef74337","entry":"get_model_complexity_info","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/utils/profiler_utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/utils/profiler_utils.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":"f131b1b13ef74337"}},{"code_sha256_prefix":"777b63f2b109a22a","entry":"get_named_parameters","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/optimizers/utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/optimizers/utils.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":"777b63f2b109a22a"}},{"code_sha256_prefix":"9cda89ca5488aa19","entry":"get_optimizer_and_lr_scheduler","repo":"cerebrasresearch/sift","repo_kind":"official","path":"ComputerVision/ImageNet/utils/optim_utils.py","file_url":"https://github.com/cerebrasresearch/sift/blob/HEAD/ComputerVision/ImageNet/utils/optim_utils.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":"9cda89ca5488aa19"}},{"code_sha256_prefix":"acaa35365153c4be","entry":"get_required_args","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/utils/maskopt_utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/utils/maskopt_utils.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":"acaa35365153c4be"}},{"code_sha256_prefix":"0415b1caa3cae4d4","entry":"get_run_name","repo":"cerebrasresearch/sift","repo_kind":"official","path":"ComputerVision/launch_utils/prepare_job_commands.py","file_url":"https://github.com/cerebrasresearch/sift/blob/HEAD/ComputerVision/launch_utils/prepare_job_commands.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":"0415b1caa3cae4d4"}},{"code_sha256_prefix":"6ea58b7d199dffa7","entry":"get_should_sparsify","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/utils/should_sparsify_utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/utils/should_sparsify_utils.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":"6ea58b7d199dffa7"}},{"code_sha256_prefix":"a13e346300ba99cf","entry":"initializations","repo":"cerebrasresearch/sift","repo_kind":"official","path":"ComputerVision/CIFAR/models/initializers.py","file_url":"https://github.com/cerebrasresearch/sift/blob/HEAD/ComputerVision/CIFAR/models/initializers.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":"a13e346300ba99cf"}},{"code_sha256_prefix":"a18d31a26dcd11fd","entry":"load_data","repo":"cerebrasresearch/sift","repo_kind":"official","path":"ComputerVision/ImageNet/utils/data_utils.py","file_url":"https://github.com/cerebrasresearch/sift/blob/HEAD/ComputerVision/ImageNet/utils/data_utils.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":"a18d31a26dcd11fd"}},{"code_sha256_prefix":"eb8c569a7f3c60ce","entry":"num_model_parameters","repo":"cerebrasresearch/sift","repo_kind":"official","path":"ComputerVision/ImageNet/utils/common_utils.py","file_url":"https://github.com/cerebrasresearch/sift/blob/HEAD/ComputerVision/ImageNet/utils/common_utils.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":"eb8c569a7f3c60ce"}},{"code_sha256_prefix":"adcedda264ea96b0","entry":"print_model_with_flops","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/utils/profiler_utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/utils/profiler_utils.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":"adcedda264ea96b0"}},{"code_sha256_prefix":"4283f68f4165ad0a","entry":"reduce_across_processes","repo":"cerebrasresearch/sift","repo_kind":"official","path":"ComputerVision/ImageNet/utils/imagenet_utils.py","file_url":"https://github.com/cerebrasresearch/sift/blob/HEAD/ComputerVision/ImageNet/utils/imagenet_utils.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":"4283f68f4165ad0a"}},{"code_sha256_prefix":"4100493b55e06307","entry":"resnet_should_sparsify","repo":"CerebrasResearch/Sparse-IFT","repo_kind":"official","path":"cbsparse/sparse/utils/should_sparsify_utils.py","file_url":"https://github.com/CerebrasResearch/Sparse-IFT/blob/HEAD/cbsparse/sparse/utils/should_sparsify_utils.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":"4100493b55e06307"}},{"code_sha256_prefix":"87e81f47b5aefd7c","entry":"set_weight_decay","repo":"cerebrasresearch/sift","repo_kind":"official","path":"ComputerVision/ImageNet/utils/optim_utils.py","file_url":"https://github.com/cerebrasresearch/sift/blob/HEAD/ComputerVision/ImageNet/utils/optim_utils.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":"87e81f47b5aefd7c"}},{"code_sha256_prefix":"24f161565d6a4013","entry":"str_to_class","repo":"cerebrasresearch/sift","repo_kind":"official","path":"ComputerVision/CIFAR/models/cifar_resnet18.py","file_url":"https://github.com/cerebrasresearch/sift/blob/HEAD/ComputerVision/CIFAR/models/cifar_resnet18.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":"24f161565d6a4013"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}