{"id":"W4312868944","doi":"10.1109/iscas48785.2022.9937567","title":"Efficient Fine-Tuning of BERT Models on the Edge","year":2022,"lang":"en","type":"article","venue":"2022 IEEE International Symposium on Circuits and Systems (ISCAS)","topic":"Advanced Neural Network Applications","field":"Computer Science","cited_by":27,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Computer science; Adaptability; Software deployment; Edge device; Fine-tuning; Computation; Inference; Distributed computing; Resource (disambiguation); Metric (unit); Resource allocation; Artificial intelligence; Computer engineering; Operating system; Computer network","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":[0.0004718707,0.0001533839,0.000175936,0.000106003,0.0004397249,0.00009196431,0.001090937,0.00002674847,0.00002524522],"category_scores_gemma":[0.00001537422,0.0001200723,0.00007620933,0.0003126452,0.00004072831,0.00008322237,0.0002820421,0.0002452146,0.00001215497],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001377975,"about_ca_system_score_gemma":0.00002412033,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002894934,"about_ca_topic_score_gemma":0.000002266614,"domain_scores_codex":[0.9980552,0.000152637,0.0003732716,0.0004436127,0.0007842986,0.0001910401],"domain_scores_gemma":[0.9986339,0.0004598266,0.0002539721,0.0004884743,0.0001038943,0.00005995623],"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":[0.00000480786,0.00008517834,0.00004130053,0.000005253954,0.00002304644,0.000004939358,0.000453966,0.5960106,0.004821332,0.3958717,0.001669791,0.001008107],"study_design_scores_gemma":[0.0002147011,0.0001212346,0.00009279907,0.00003448828,0.000004725711,0.0000672565,0.00009971636,0.9901181,0.0004346334,0.0009251584,0.007740679,0.0001465244],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6035157,0.0004728622,0.3023622,0.01979936,0.009337189,0.002025656,0.0002653884,0.0002709892,0.06195064],"genre_scores_gemma":[0.998158,0.00001565154,0.00002765874,0.0004045856,0.0002078303,0.0003139264,0.000007191776,0.00001481983,0.0008503618],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3949465,"threshold_uncertainty_score":0.4896405,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03886087501171164,"score_gpt":0.2612023358593475,"score_spread":0.2223414608476358,"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."}}