{"id":"W4416746178","doi":"10.1016/j.parco.2025.103165","title":"Butterfly factorization for vision transformers on multi-IPU systems","year":2025,"lang":"en","type":"article","venue":"Parallel Computing","topic":"Advanced Memory and Neural Computing","field":"Engineering","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"","keywords":"Butterfly; Memory footprint; Transformer; Factorization; Computation","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0001299105,0.0001818024,0.000202341,0.0001080613,0.0002179698,0.00004948065,0.0001123159,0.000068849,9.263134e-7],"category_scores_gemma":[0.00002001305,0.000181982,0.00007329116,0.0001724072,0.00001098158,0.00009654484,0.0000156563,0.0001548021,0.000007539858],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00007123905,"about_ca_system_score_gemma":0.000009202425,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000002432566,"about_ca_topic_score_gemma":8.337827e-7,"domain_scores_codex":[0.9990768,0.00002273429,0.0002911288,0.000226673,0.00008344954,0.0002991876],"domain_scores_gemma":[0.9995359,0.000241289,0.00003743449,0.0001078632,0.00003587177,0.00004162121],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00002230683,0.00001211829,0.00006249137,0.0002468197,0.00002611559,9.906117e-7,0.0001654164,0.9733028,0.01042202,0.0006837048,0.0001389833,0.01491617],"study_design_scores_gemma":[0.000779792,0.00006057231,0.0004424994,0.000277533,0.00001305063,0.000001362164,0.0000848399,0.9890133,0.006896189,0.0000746138,0.002157745,0.0001984623],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1088395,0.0002034833,0.8877846,0.00003089748,0.001322174,0.0004380475,0.000003608596,0.0004389746,0.000938706],"genre_scores_gemma":[0.9952077,0.00001177275,0.004345286,0.0000686771,0.0001306005,0.00001064348,0.00001752892,0.00002663275,0.00018115],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8863682,"threshold_uncertainty_score":0.7421012,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02213822854525998,"score_gpt":0.2953664193748585,"score_spread":0.2732281908295985,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}