{"id":"W7115041457","doi":"","title":"HarMoEny: Efficient multi-GPU inference of Mixture of Experts models by scheduling experts across accelerators","year":2025,"lang":"en","type":"dissertation","venue":"eScholarship@McGill (McGill)","topic":"Mobile Crowdsensing and Crowdsourcing","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Inference; Scheduling (production processes); Mixture model; Job shop scheduling; Expert system","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001775052,0.001327834,0.001255376,0.0008228328,0.0006517544,0.001509876,0.003231426,0.001505561,0.003628566],"category_scores_gemma":[0.005246526,0.001038868,0.001438425,0.0006967916,0.00069079,0.002548258,0.002032917,0.002592513,0.001294287],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001378673,"about_ca_system_score_gemma":0.002931928,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.02299866,"about_ca_topic_score_gemma":0.0399075,"domain_scores_codex":[0.9991251,0.0002294196,0.00003677472,0.0002698252,0.0002181738,0.0001206522],"domain_scores_gemma":[0.9986407,0.0006511488,0.00009312836,0.0002907744,0.0001935241,0.0001306669],"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":[0.0008749733,0.0002787428,0.00559602,0.0001919391,0.0004456695,0.0002112023,0.0001707236,0.7637665,0.008583101,0.008194646,0.02083783,0.1908485],"study_design_scores_gemma":[0.00001852515,0.00001470406,0.0001240143,0.000002911491,0.00000760935,0.00001215093,0.000007239106,0.9963694,0.0009572147,0.00166697,0.0008130608,0.000006180602],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.03661908,0.0007666028,0.9403259,0.0006294187,0.0001786342,0.0001321761,0.0005113506,0.01785513,0.002981614],"genre_scores_gemma":[0.4017494,0.0003924583,0.5849779,0.0008148173,0.0001651434,0.0002285894,0.002040994,0.001385907,0.00824474],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.02299866,"threshold_uncertainty_score":0.04572958,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02738314471006957,"score_gpt":0.2804819732843613,"score_spread":0.2530988285742917,"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."}}