{"id":"W3100985894","doi":"10.1109/micro50266.2020.00071","title":"GOBO: Quantizing Attention-Based NLP Models for Low Latency and Energy Efficient Inference","year":2020,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Topic Modeling","field":"Computer Science","cited_by":163,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Computer science; Quantization (signal processing); Inference; Computation; Latency (audio); Language model; Parallel computing; Computer engineering; Computer hardware; Artificial intelligence; Algorithm","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0001718774,0.0003024619,0.0003379644,0.0001983316,0.0001821642,0.0001668848,0.001161973,0.0002209903,0.000003134556],"category_scores_gemma":[0.00004176932,0.0003592453,0.0001899054,0.0003035652,0.00006223171,0.0002176302,0.001272743,0.0002845346,0.000005464555],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001101075,"about_ca_system_score_gemma":0.0002687956,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002443685,"about_ca_topic_score_gemma":0.0000284257,"domain_scores_codex":[0.9978526,0.00007364412,0.000245408,0.001369498,0.0001085855,0.0003502339],"domain_scores_gemma":[0.9983659,0.0001986761,0.0002243487,0.0008311899,0.0001856374,0.0001942983],"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":[0.00000961796,0.00002959441,0.0003061196,0.0001012718,0.00001884322,0.00002960357,0.00006939939,0.6855415,0.00005032133,0.3132216,0.000008157251,0.0006139381],"study_design_scores_gemma":[0.0004338663,0.00003575314,0.0001103035,0.0001543528,0.0000364294,6.011832e-7,0.00001769329,0.9381905,0.00006076521,0.06056452,0.00003817777,0.0003570604],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1300904,0.00006754708,0.8685606,0.0002403987,0.0003655637,0.0001772058,0.00001335576,0.0002070917,0.0002778624],"genre_scores_gemma":[0.9832288,0.00004892419,0.01628911,0.0001693501,0.00004992414,0.000003210818,0.0000150252,0.0000174286,0.0001782209],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8531384,"threshold_uncertainty_score":0.999886,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.11445244292019,"score_gpt":0.2010061101867137,"score_spread":0.08655366726652367,"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."}}