--- license: apache-2.0 library_name: transformers pipeline_tag: text-generation tags: - merge --- Merge of top 7B models with PASS method > mergekit is a toolkit for merging pre-trained language models. mergekit uses an out-of-core approach to perform unreasonably elaborate merges in resource-constrained situations. Merges can be run entirely on CPU or accelerated with as little as 8 GB of VRAM. Many merging algorithms are supported, with more coming as they catch my attention. ## Eval ```python { "all": { "acc": 0.6152059168567449, "acc_stderr": 0.031951119145286845, "acc_norm": 0.6274010157580394, "acc_norm_stderr": 0.032831804892806175, "mc1": 0.25091799265605874, "mc1_stderr": 0.015176985027707694, "mc2": 0.5006656333594469, "mc2_stderr": 0.01636490303268174 }, "harness|arc:challenge|25": { "acc": 0.3447098976109215, "acc_stderr": 0.013888816286782112, "acc_norm": 0.3856655290102389, "acc_norm_stderr": 0.01422425097325717 }, "harness|hellaswag|10": { "acc": 0.34116709818761204, "acc_stderr": 0.004731324409133264, "acc_norm": 0.515435172276439, "acc_norm_stderr": 0.004987403268345035 }, "harness|hendrycksTest-abstract_algebra|5": { "acc": 0.27, "acc_stderr": 0.04461960433384741, "acc_norm": 0.27, "acc_norm_stderr": 0.04461960433384741 }, "harness|hendrycksTest-anatomy|5": { "acc": 0.5703703703703704, "acc_stderr": 0.042763494943765995, "acc_norm": 0.5703703703703704, "acc_norm_stderr": 0.042763494943765995 }, "harness|hendrycksTest-astronomy|5": { "acc": 0.6842105263157895, "acc_stderr": 0.0378272898086547, "acc_norm": 0.6842105263157895, "acc_norm_stderr": 0.0378272898086547 }, "harness|hendrycksTest-business_ethics|5": { "acc": 0.62, "acc_stderr": 0.048783173121456316, "acc_norm": 0.62, "acc_norm_stderr": 0.048783173121456316 }, "harness|hendrycksTest-clinical_knowledge|5": { "acc": 0.7169811320754716, "acc_stderr": 0.027724236492700918, "acc_norm": 0.7169811320754716, "acc_norm_stderr": 0.027724236492700918 }, "harness|hendrycksTest-college_biology|5": { "acc": 0.7638888888888888, "acc_stderr": 0.03551446610810826, "acc_norm": 0.7638888888888888, "acc_norm_stderr": 0.03551446610810826 }, "harness|hendrycksTest-college_chemistry|5": { "acc": 0.45, "acc_stderr": 0.05, "acc_norm": 0.45, "acc_norm_stderr": 0.05 }, "harness|hendrycksTest-college_computer_science|5": { "acc": 0.46, "acc_stderr": 0.05009082659620333, "acc_norm": 0.46, "acc_norm_stderr": 0.05009082659620333 }, "harness|hendrycksTest-college_mathematics|5": { "acc": 0.32, "acc_stderr": 0.04688261722621504, "acc_norm": 0.32, "acc_norm_stderr": 0.04688261722621504 }, "harness|hendrycksTest-college_medicine|5": { "acc": 0.6358381502890174, "acc_stderr": 0.03669072477416907, "acc_norm": 0.6358381502890174, "acc_norm_stderr": 0.03669072477416907 }, "harness|hendrycksTest-college_physics|5": { "acc": 0.4019607843137255, "acc_stderr": 0.048786087144669955, "acc_norm": 0.4019607843137255, "acc_norm_stderr": 0.048786087144669955 }, "harness|hendrycksTest-computer_security|5": { "acc": 0.76, "acc_stderr": 0.04292346959909283, "acc_norm": 0.76, "acc_norm_stderr": 0.04292346959909283 }, "harness|hendrycksTest-conceptual_physics|5": { "acc": 0.5446808510638298, "acc_stderr": 0.03255525359340355, "acc_norm": 0.5446808510638298, "acc_norm_stderr": 0.03255525359340355 }, "harness|hendrycksTest-econometrics|5": { "acc": 0.4824561403508772, "acc_stderr": 0.04700708033551038, "acc_norm": 0.4824561403508772, "acc_norm_stderr": 0.04700708033551038 }, "harness|hendrycksTest-electrical_engineering|5": { "acc": 0.5172413793103449, "acc_stderr": 0.04164188720169375, "acc_norm": 0.5172413793103449, "acc_norm_stderr": 0.04164188720169375 }, "harness|hendrycksTest-elementary_mathematics|5": { "acc": 0.42857142857142855, "acc_stderr": 0.025487187147859372, "acc_norm": 0.42857142857142855, "acc_norm_stderr": 0.025487187147859372 }, "harness|hendrycksTest-formal_logic|5": { "acc": 0.3968253968253968, "acc_stderr": 0.043758884927270605, "acc_norm": 0.3968253968253968, "acc_norm_stderr": 0.043758884927270605 }, "harness|hendrycksTest-global_facts|5": { "acc": 0.34, "acc_stderr": 0.04760952285695236, "acc_norm": 0.34, "acc_norm_stderr": 0.04760952285695236 }, "harness|hendrycksTest-high_school_biology|5": { "acc": 0.7741935483870968, "acc_stderr": 0.023785577884181015, "acc_norm": 0.7741935483870968, "acc_norm_stderr": 0.023785577884181015 }, "harness|hendrycksTest-high_school_chemistry|5": { "acc": 0.5123152709359606, "acc_stderr": 0.035169204442208966, "acc_norm": 0.5123152709359606, "acc_norm_stderr": 0.035169204442208966 }, "harness|hendrycksTest-high_school_computer_science|5": { "acc": 0.66, "acc_stderr": 0.04760952285695237, "acc_norm": 0.66, "acc_norm_stderr": 0.04760952285695237 }, "harness|hendrycksTest-high_school_european_history|5": { "acc": 0.7636363636363637, "acc_stderr": 0.03317505930009181, "acc_norm": 0.7636363636363637, "acc_norm_stderr": 0.03317505930009181 }, "harness|hendrycksTest-high_school_geography|5": { "acc": 0.7373737373737373, "acc_stderr": 0.03135305009533085, "acc_norm": 0.7373737373737373, "acc_norm_stderr": 0.03135305009533085 }, "harness|hendrycksTest-high_school_government_and_politics|5": { "acc": 0.8808290155440415, "acc_stderr": 0.023381935348121437, "acc_norm": 0.8808290155440415, "acc_norm_stderr": 0.023381935348121437 }, "harness|hendrycksTest-high_school_macroeconomics|5": { "acc": 0.617948717948718, "acc_stderr": 0.024635549163908237, "acc_norm": 0.617948717948718, "acc_norm_stderr": 0.024635549163908237 }, "harness|hendrycksTest-high_school_mathematics|5": { "acc": 0.2777777777777778, "acc_stderr": 0.027309140588230203, "acc_norm": 0.2777777777777778, "acc_norm_stderr": 0.027309140588230203 }, "harness|hendrycksTest-high_school_microeconomics|5": { "acc": 0.6512605042016807, "acc_stderr": 0.030956636328566545, "acc_norm": 0.6512605042016807, "acc_norm_stderr": 0.030956636328566545 }, "harness|hendrycksTest-high_school_physics|5": { "acc": 0.32450331125827814, "acc_stderr": 0.038227469376587525, "acc_norm": 0.32450331125827814, "acc_norm_stderr": 0.038227469376587525 }, "harness|hendrycksTest-high_school_psychology|5": { "acc": 0.8440366972477065, "acc_stderr": 0.015555802713590158, "acc_norm": 0.8440366972477065, "acc_norm_stderr": 0.015555802713590158 }, "harness|hendrycksTest-high_school_statistics|5": { "acc": 0.4722222222222222, "acc_stderr": 0.0340470532865388, "acc_norm": 0.4722222222222222, "acc_norm_stderr": 0.0340470532865388 }, "harness|hendrycksTest-high_school_us_history|5": { "acc": 0.8431372549019608, "acc_stderr": 0.025524722324553346, "acc_norm": 0.8431372549019608, "acc_norm_stderr": 0.025524722324553346 }, "harness|hendrycksTest-high_school_world_history|5": { "acc": 0.810126582278481, "acc_stderr": 0.025530100460233497, "acc_norm": 0.810126582278481, "acc_norm_stderr": 0.025530100460233497 }, "harness|hendrycksTest-human_aging|5": { "acc": 0.7174887892376681, "acc_stderr": 0.03021683101150877, "acc_norm": 0.7174887892376681, "acc_norm_stderr": 0.03021683101150877 }, "harness|hendrycksTest-human_sexuality|5": { "acc": 0.7786259541984732, "acc_stderr": 0.0364129708131373, "acc_norm": 0.7786259541984732, "acc_norm_stderr": 0.0364129708131373 }, "harness|hendrycksTest-international_law|5": { "acc": 0.7768595041322314, "acc_stderr": 0.03800754475228733, "acc_norm": 0.7768595041322314, "acc_norm_stderr": 0.03800754475228733 }, "harness|hendrycksTest-jurisprudence|5": { "acc": 0.8148148148148148, "acc_stderr": 0.03755265865037181, "acc_norm": 0.8148148148148148, "acc_norm_stderr": 0.03755265865037181 }, "harness|hendrycksTest-logical_fallacies|5": { "acc": 0.7914110429447853, "acc_stderr": 0.031921934489347235, "acc_norm": 0.7914110429447853, "acc_norm_stderr": 0.031921934489347235 }, "harness|hendrycksTest-machine_learning|5": { "acc": 0.5446428571428571, "acc_stderr": 0.04726835553719097, "acc_norm": 0.5446428571428571, "acc_norm_stderr": 0.04726835553719097 }, "harness|hendrycksTest-management|5": { "acc": 0.8349514563106796, "acc_stderr": 0.036756688322331886, "acc_norm": 0.8349514563106796, "acc_norm_stderr": 0.036756688322331886 }, "harness|hendrycksTest-marketing|5": { "acc": 0.8290598290598291, "acc_stderr": 0.024662496845209804, "acc_norm": 0.8290598290598291, "acc_norm_stderr": 0.024662496845209804 }, "harness|hendrycksTest-medical_genetics|5": { "acc": 0.69, "acc_stderr": 0.04648231987117316, "acc_norm": 0.69, "acc_norm_stderr": 0.04648231987117316 }, "harness|hendrycksTest-miscellaneous|5": { "acc": 0.8250319284802043, "acc_stderr": 0.013586619219903324, "acc_norm": 0.8250319284802043, "acc_norm_stderr": 0.013586619219903324 }, "harness|hendrycksTest-moral_disputes|5": { "acc": 0.7283236994219653, "acc_stderr": 0.023948512905468348, "acc_norm": 0.7283236994219653, "acc_norm_stderr": 0.023948512905468348 }, "harness|hendrycksTest-moral_scenarios|5": { "acc": 0.36312849162011174, "acc_stderr": 0.016083749986853704, "acc_norm": 0.36312849162011174, "acc_norm_stderr": 0.016083749986853704 }, "harness|hendrycksTest-nutrition|5": { "acc": 0.7450980392156863, "acc_stderr": 0.02495418432487991, "acc_norm": 0.7450980392156863, "acc_norm_stderr": 0.02495418432487991 }, "harness|hendrycksTest-philosophy|5": { "acc": 0.7202572347266881, "acc_stderr": 0.02549425935069491, "acc_norm": 0.7202572347266881, "acc_norm_stderr": 0.02549425935069491 }, "harness|hendrycksTest-prehistory|5": { "acc": 0.7530864197530864, "acc_stderr": 0.023993501709042114, "acc_norm": 0.7530864197530864, "acc_norm_stderr": 0.023993501709042114 }, "harness|hendrycksTest-professional_accounting|5": { "acc": 0.4787234042553192, "acc_stderr": 0.029800481645628693, "acc_norm": 0.4787234042553192, "acc_norm_stderr": 0.029800481645628693 }, "harness|hendrycksTest-professional_law|5": { "acc": 0.4367666232073012, "acc_stderr": 0.01266770191960366, "acc_norm": 0.4367666232073012, "acc_norm_stderr": 0.01266770191960366 }, "harness|hendrycksTest-professional_medicine|5": { "acc": 0.6176470588235294, "acc_stderr": 0.029520095697687765, "acc_norm": 0.6176470588235294, "acc_norm_stderr": 0.029520095697687765 }, "harness|hendrycksTest-professional_psychology|5": { "acc": 0.6699346405228758, "acc_stderr": 0.019023726160724553, "acc_norm": 0.6699346405228758, "acc_norm_stderr": 0.019023726160724553 }, "harness|hendrycksTest-public_relations|5": { "acc": 0.6545454545454545, "acc_stderr": 0.04554619617541054, "acc_norm": 0.6545454545454545, "acc_norm_stderr": 0.04554619617541054 }, "harness|hendrycksTest-security_studies|5": { "acc": 0.726530612244898, "acc_stderr": 0.028535560337128445, "acc_norm": 0.726530612244898, "acc_norm_stderr": 0.028535560337128445 }, "harness|hendrycksTest-sociology|5": { "acc": 0.845771144278607, "acc_stderr": 0.025538433368578334, "acc_norm": 0.845771144278607, "acc_norm_stderr": 0.025538433368578334 }, "harness|hendrycksTest-us_foreign_policy|5": { "acc": 0.85, "acc_stderr": 0.03588702812826371, "acc_norm": 0.85, "acc_norm_stderr": 0.03588702812826371 }, "harness|hendrycksTest-virology|5": { "acc": 0.5542168674698795, "acc_stderr": 0.03869543323472101, "acc_norm": 0.5542168674698795, "acc_norm_stderr": 0.03869543323472101 }, "harness|hendrycksTest-world_religions|5": { "acc": 0.8362573099415205, "acc_stderr": 0.028380919596145866, "acc_norm": 0.8362573099415205, "acc_norm_stderr": 0.028380919596145866 }, "harness|truthfulqa:mc|0": { "mc1": 0.25091799265605874, "mc1_stderr": 0.015176985027707694, "mc2": 0.5006656333594469, "mc2_stderr": 0.01636490303268174 }, "harness|winogrande|5": { "acc": 0.7261247040252565, "acc_stderr": 0.012533292732620296 }, "harness|gsm8k|5": { "acc": 0.0, "acc_stderr": 0.0 } } ```