simplestories-trigger-organisms / simple-names /multi /function_additivity-joint-e5kl3small-single-e3kl3small.json
| { | |
| "config": "configs/simplestories_multi.json", | |
| "joint_tag": "e5kl3small", | |
| "single_tag": "e3kl3small", | |
| "positions": 2192, | |
| "trigger_positions": 822, | |
| "fits": { | |
| "trained/contrast/r1": { | |
| "singles_same_seed": 0.15722104204237175, | |
| "singles_other_seed": 0.15118580011499372, | |
| "plain_sum_same_seed": -0.31199193577187656, | |
| "joint_other_seed": 0.3143820837102935, | |
| "single_other_seed": 0.47735655007467603, | |
| "mean_alpha": [ | |
| 0.20449777768688307, | |
| 0.256917270014922, | |
| 0.26778826706619824 | |
| ] | |
| }, | |
| "trained/contrast/r3": { | |
| "singles_same_seed": 0.18311470034211252, | |
| "singles_other_seed": 0.18252321686433703, | |
| "plain_sum_same_seed": -0.3024654124778825, | |
| "joint_other_seed": 0.5866823451817111, | |
| "single_other_seed": 0.47735655007467603, | |
| "mean_alpha": [ | |
| 0.22563983991812353, | |
| 0.30545997716918855, | |
| 0.31497479540744927 | |
| ] | |
| }, | |
| "trained/contrast/r16": { | |
| "singles_same_seed": 0.18956465949918366, | |
| "singles_other_seed": 0.19712998680099625, | |
| "plain_sum_same_seed": -0.4061425253169264, | |
| "joint_other_seed": 0.7293115931667272, | |
| "single_other_seed": 0.47735655007467603, | |
| "mean_alpha": [ | |
| 0.24208609782074328, | |
| 0.31242847245090477, | |
| 0.3321112621010148 | |
| ] | |
| }, | |
| "trained/trigger/r1": { | |
| "singles_same_seed": 0.14451527922172314, | |
| "singles_other_seed": 0.13954749424936996, | |
| "plain_sum_same_seed": -0.31698027272316587, | |
| "joint_other_seed": 0.2791917246230651, | |
| "single_other_seed": 0.3922550950079986, | |
| "mean_alpha": [ | |
| 0.1803879713724893, | |
| 0.21562589899355902, | |
| 0.23239086005589113 | |
| ] | |
| }, | |
| "trained/trigger/r3": { | |
| "singles_same_seed": 0.14950411283136056, | |
| "singles_other_seed": 0.1437155031488253, | |
| "plain_sum_same_seed": -0.23467530937399816, | |
| "joint_other_seed": 0.44263358131634, | |
| "single_other_seed": 0.3922550950079986, | |
| "mean_alpha": [ | |
| 0.2226391193412599, | |
| 0.2596216393012871, | |
| 0.27410332874731264 | |
| ] | |
| }, | |
| "trained/trigger/r16": { | |
| "singles_same_seed": 0.13125135319974723, | |
| "singles_other_seed": 0.13180656972957439, | |
| "plain_sum_same_seed": -0.24751704295132715, | |
| "joint_other_seed": 0.5576826654447682, | |
| "single_other_seed": 0.3922550950079986, | |
| "mean_alpha": [ | |
| 0.24828708204089678, | |
| 0.271789741981405, | |
| 0.2863557302378121 | |
| ] | |
| }, | |
| "trained/all/r1": { | |
| "singles_same_seed": 0.12763152862126248, | |
| "singles_other_seed": 0.12395210508511545, | |
| "plain_sum_same_seed": -0.35361201471287473, | |
| "joint_other_seed": 0.26660045325494397, | |
| "single_other_seed": 0.3587347579194623, | |
| "mean_alpha": [ | |
| 0.17753660452201567, | |
| 0.20810464654965066, | |
| 0.23040555122465847 | |
| ] | |
| }, | |
| "trained/all/r3": { | |
| "singles_same_seed": 0.12426102847455722, | |
| "singles_other_seed": 0.11983588865585959, | |
| "plain_sum_same_seed": -0.2566101194431716, | |
| "joint_other_seed": 0.3841497199036973, | |
| "single_other_seed": 0.3587347579194623, | |
| "mean_alpha": [ | |
| 0.22095685130371417, | |
| 0.2526101448452137, | |
| 0.27184673247185337 | |
| ] | |
| }, | |
| "trained/all/r16": { | |
| "singles_same_seed": 0.10499413690111237, | |
| "singles_other_seed": 0.10437008868330633, | |
| "plain_sum_same_seed": -0.23840699007986244, | |
| "joint_other_seed": 0.5010189390246579, | |
| "single_other_seed": 0.3587347579194623, | |
| "mean_alpha": [ | |
| 0.24614201612181089, | |
| 0.2668605466710038, | |
| 0.2855909937079848 | |
| ] | |
| }, | |
| "learned/contrast/r1": { | |
| "singles_same_seed": 0.15725313698567311, | |
| "singles_other_seed": 0.15125384481176485, | |
| "plain_sum_same_seed": -0.3117731277554263, | |
| "joint_other_seed": 0.3145630138667902, | |
| "single_other_seed": 0.47768719633479495, | |
| "mean_alpha": [ | |
| 0.20456697257078116, | |
| 0.25683154445291695, | |
| 0.2679635143447538 | |
| ] | |
| }, | |
| "learned/contrast/r3": { | |
| "singles_same_seed": 0.18294239385112843, | |
| "singles_other_seed": 0.18253792956046877, | |
| "plain_sum_same_seed": -0.3028288178035212, | |
| "joint_other_seed": 0.5867736225720839, | |
| "single_other_seed": 0.47768719633479495, | |
| "mean_alpha": [ | |
| 0.22606165359158897, | |
| 0.3051313643022334, | |
| 0.31479379224326715 | |
| ] | |
| }, | |
| "learned/contrast/r16": { | |
| "singles_same_seed": 0.189426106161116, | |
| "singles_other_seed": 0.19713641808807436, | |
| "plain_sum_same_seed": -0.4063577105886733, | |
| "joint_other_seed": 0.7293671093306947, | |
| "single_other_seed": 0.47768719633479495, | |
| "mean_alpha": [ | |
| 0.24238962241649878, | |
| 0.31235422367518323, | |
| 0.3319392851101131 | |
| ] | |
| }, | |
| "learned/trigger/r1": { | |
| "singles_same_seed": 0.1444868901097356, | |
| "singles_other_seed": 0.1395954580607608, | |
| "plain_sum_same_seed": -0.31677900147846816, | |
| "joint_other_seed": 0.279262468079983, | |
| "single_other_seed": 0.3924589814436419, | |
| "mean_alpha": [ | |
| 0.18046467274735545, | |
| 0.21556713014091483, | |
| 0.2325496653697522 | |
| ] | |
| }, | |
| "learned/trigger/r3": { | |
| "singles_same_seed": 0.14941298855894403, | |
| "singles_other_seed": 0.1437138295151933, | |
| "plain_sum_same_seed": -0.23483022706009793, | |
| "joint_other_seed": 0.4426919751500006, | |
| "single_other_seed": 0.3924589814436419, | |
| "mean_alpha": [ | |
| 0.22262756866320926, | |
| 0.25940925424289335, | |
| 0.27390210404816556 | |
| ] | |
| }, | |
| "learned/trigger/r16": { | |
| "singles_same_seed": 0.13121800426617913, | |
| "singles_other_seed": 0.1317981821522234, | |
| "plain_sum_same_seed": -0.24753107333185587, | |
| "joint_other_seed": 0.5577253448124586, | |
| "single_other_seed": 0.3924589814436419, | |
| "mean_alpha": [ | |
| 0.24825966796867313, | |
| 0.27165401629209224, | |
| 0.2861211441763555 | |
| ] | |
| }, | |
| "learned/all/r1": { | |
| "singles_same_seed": 0.12760054607152593, | |
| "singles_other_seed": 0.12397603322869855, | |
| "plain_sum_same_seed": -0.3533804801371443, | |
| "joint_other_seed": 0.2666200372884032, | |
| "single_other_seed": 0.3588921907515229, | |
| "mean_alpha": [ | |
| 0.1776558673862108, | |
| 0.2080522060309632, | |
| 0.23055076121012721 | |
| ] | |
| }, | |
| "learned/all/r3": { | |
| "singles_same_seed": 0.12420293627613421, | |
| "singles_other_seed": 0.11983278988434098, | |
| "plain_sum_same_seed": -0.2566885002058761, | |
| "joint_other_seed": 0.38421088001617765, | |
| "single_other_seed": 0.3588921907515229, | |
| "mean_alpha": [ | |
| 0.22093265161143674, | |
| 0.2523648124317535, | |
| 0.2716933809738654 | |
| ] | |
| }, | |
| "learned/all/r16": { | |
| "singles_same_seed": 0.10498196012268879, | |
| "singles_other_seed": 0.10437337243687561, | |
| "plain_sum_same_seed": -0.23837462749188854, | |
| "joint_other_seed": 0.5010556729497015, | |
| "single_other_seed": 0.3588921907515229, | |
| "mean_alpha": [ | |
| 0.2461076730202573, | |
| 0.2666819781445647, | |
| 0.28540823677192545 | |
| ] | |
| } | |
| }, | |
| "per_module": { | |
| "trained/contrast/r1": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.12274977777751228, | |
| "joint_other_seed": 0.2648990959306884, | |
| "single_other_seed": 0.48587948332480346, | |
| "energy": 22890.353515625 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.3186467253937689, | |
| "joint_other_seed": 0.5925843620309077, | |
| "single_other_seed": 0.7243091349223783, | |
| "energy": 3784.9230143229165 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.21654944925673025, | |
| "joint_other_seed": 0.5730861357918005, | |
| "single_other_seed": 0.6368519846345397, | |
| "energy": 309.9208984375 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.30641798265881626, | |
| "joint_other_seed": 0.6399689840724809, | |
| "single_other_seed": 0.760534927951505, | |
| "energy": 0.3500038683414459 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.1090865185427583, | |
| "joint_other_seed": 0.18564067870909085, | |
| "single_other_seed": 0.5441413325132178, | |
| "energy": 2226.26025390625 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.1248003152305488, | |
| "joint_other_seed": 0.413910078590367, | |
| "single_other_seed": 0.1932504611807233, | |
| "energy": 2026.356689453125 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.11004480997959853, | |
| "joint_other_seed": 0.15915483301282507, | |
| "single_other_seed": 0.4552290351023257, | |
| "energy": 12.238382577896118 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.3199219791350615, | |
| "joint_other_seed": 0.6598530503816253, | |
| "single_other_seed": 0.2647264410806427, | |
| "energy": 475.1502685546875 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.23265024056774075, | |
| "joint_other_seed": 0.26815278931381525, | |
| "single_other_seed": 0.5878925509667523, | |
| "energy": 705.0514322916666 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.46904195846484253, | |
| "joint_other_seed": 0.6359591493206762, | |
| "single_other_seed": 0.8011841023733169, | |
| "energy": 24.170740127563477 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.3931492404038426, | |
| "joint_other_seed": 0.6115013494828851, | |
| "single_other_seed": 0.6177728718745246, | |
| "energy": 0.10852570335070293 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.13920598490167435, | |
| "joint_other_seed": 0.26105119910815144, | |
| "single_other_seed": 0.15763238084996323, | |
| "energy": 76.66505304972331 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.136168188343613, | |
| "joint_other_seed": 0.4279265032273333, | |
| "single_other_seed": 0.27112513696978946, | |
| "energy": 61.2096004486084 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.2610536989709984, | |
| "joint_other_seed": 0.128278177294887, | |
| "single_other_seed": 0.37701066731825883, | |
| "energy": 0.175991578027606 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.2474406453315695, | |
| "joint_other_seed": 0.33485525063006427, | |
| "single_other_seed": 0.3343101885625702, | |
| "energy": 239.59706115722656 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.3421404435110442, | |
| "joint_other_seed": 0.32068260829065676, | |
| "single_other_seed": 0.296530556530737, | |
| "energy": 71.28492482503255 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.1696864609742843, | |
| "joint_other_seed": 0.1482662636256591, | |
| "single_other_seed": 0.42978460011696584, | |
| "energy": 0.9450852672259012 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.1409251359013891, | |
| "joint_other_seed": 0.1002191075517066, | |
| "single_other_seed": 0.3366246367264273, | |
| "energy": 0.027544548735022545 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.13696755623185775, | |
| "joint_other_seed": 0.35726860077526157, | |
| "single_other_seed": 0.4651934549202552, | |
| "energy": 230.4182268778483 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.20393665081753365, | |
| "joint_other_seed": 0.5113728337013659, | |
| "single_other_seed": 0.486970793053885, | |
| "energy": 253.22417704264322 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.0730525525897383, | |
| "joint_other_seed": 0.20864407224736847, | |
| "single_other_seed": 0.12299144114097349, | |
| "energy": 0.2294423778851827 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.07797853339848992, | |
| "joint_other_seed": 0.25615123523098315, | |
| "single_other_seed": 0.4152346533083575, | |
| "energy": 986.9556884765625 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.100991308119318, | |
| "joint_other_seed": 0.005989840041099148, | |
| "single_other_seed": 0.11030511850948532, | |
| "energy": 58.93562157948812 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.07979539202480013, | |
| "joint_other_seed": 0.2306801109224047, | |
| "single_other_seed": 0.05167793859276493, | |
| "energy": 0.4098975360393524 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.027799329201548817, | |
| "joint_other_seed": 0.009714762092688511, | |
| "single_other_seed": 0.35150394918660266, | |
| "energy": 0.12453945974508922 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.12076618539209705, | |
| "joint_other_seed": 0.08693269667848025, | |
| "single_other_seed": 0.26805661626541666, | |
| "energy": 240.28058878580728 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.07839345967585154, | |
| "joint_other_seed": 0.18766955734294366, | |
| "single_other_seed": 0.2717621192367111, | |
| "energy": 355.1099497477214 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.049259799493530654, | |
| "joint_other_seed": 0.0567922144927665, | |
| "single_other_seed": 0.06782404834330469, | |
| "energy": 0.5091984868049622 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.25598103936285743, | |
| "joint_other_seed": 0.6704044919107881, | |
| "single_other_seed": 0.38853736803957095, | |
| "energy": 65.78683090209961 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.04194278918763472, | |
| "joint_other_seed": 0.024916997377660288, | |
| "single_other_seed": 0.09271719625673758, | |
| "energy": 3.910970449447632 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.019571694814601637, | |
| "joint_other_seed": 0.02249341602533217, | |
| "single_other_seed": 0.008274084377512092, | |
| "energy": 0.36615753670533496 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.01757133229877883, | |
| "joint_other_seed": 0.016361502824112195, | |
| "single_other_seed": 0.14473034896500814, | |
| "energy": 0.013498639377454916 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.0644252249950944, | |
| "joint_other_seed": 0.08404644497872238, | |
| "single_other_seed": 0.435110237073146, | |
| "energy": 205.56495412190756 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.11944457569773093, | |
| "joint_other_seed": 0.10250252812251337, | |
| "single_other_seed": 0.3246355327060052, | |
| "energy": 276.10450744628906 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.010071919057357714, | |
| "joint_other_seed": 0.012641579501144973, | |
| "single_other_seed": 0.04923025870806208, | |
| "energy": 0.12704000373681387 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.08811399845066048, | |
| "joint_other_seed": 0.17538287258109087, | |
| "single_other_seed": 0.5181498942594329, | |
| "energy": 91.31846809387207 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.24987131263796758, | |
| "joint_other_seed": 0.33641574318747225, | |
| "single_other_seed": 0.15697565887361753, | |
| "energy": 158.9761505126953 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.15734447309778554, | |
| "joint_other_seed": 0.2964696177935535, | |
| "single_other_seed": 0.1973466321030315, | |
| "energy": 3.5031625827153525 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.11609155650603076, | |
| "joint_other_seed": 0.3165940442537887, | |
| "single_other_seed": 0.5224868217359202, | |
| "energy": 0.8077931106090546 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.16019138511405973, | |
| "joint_other_seed": 0.3234354203796891, | |
| "single_other_seed": 0.15117191476887493, | |
| "energy": 45.62259737650553 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.2080828239005672, | |
| "joint_other_seed": 0.5249625390196974, | |
| "single_other_seed": 0.5634331571340561, | |
| "energy": 181.7551472981771 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.0029291541523294162, | |
| "joint_other_seed": 0.01898450294671228, | |
| "single_other_seed": 0.09654409277635945, | |
| "energy": 0.2462572604417801 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.21375293987390953, | |
| "joint_other_seed": 0.34964838478193705, | |
| "single_other_seed": 0.6922795806250662, | |
| "energy": 273.95740000406903 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.34682920461929534, | |
| "joint_other_seed": 0.6634094663490182, | |
| "single_other_seed": 0.5542114947727996, | |
| "energy": 292.32372029622394 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.0660694191650778, | |
| "joint_other_seed": 0.006562160813818441, | |
| "single_other_seed": 0.34817693558750884, | |
| "energy": 5.190321326255798 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.08043267534546072, | |
| "joint_other_seed": 0.037947020189179614, | |
| "single_other_seed": 0.7035667897393909, | |
| "energy": 4.532786289850871 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.20084686037005683, | |
| "joint_other_seed": 0.327548473043617, | |
| "single_other_seed": 0.4565658516217236, | |
| "energy": 63.22435506184896 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.10526187869659775, | |
| "joint_other_seed": 0.33869959801687555, | |
| "single_other_seed": 0.2602501633700697, | |
| "energy": 35.235870361328125 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.009668515650887333, | |
| "joint_other_seed": 0.0016489020101244645, | |
| "single_other_seed": 0.10943172991174543, | |
| "energy": 0.4844776789347331 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.031216267881020943, | |
| "joint_other_seed": 0.24237898454839646, | |
| "single_other_seed": 0.12258662755143679, | |
| "energy": 73.96189626057942 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.012812565726644576, | |
| "joint_other_seed": 0.0026271060174521965, | |
| "single_other_seed": 0.01896949130803771, | |
| "energy": 4.602675358454387 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.02447048528881479, | |
| "joint_other_seed": 0.02821779492385923, | |
| "single_other_seed": 0.09990828160962073, | |
| "energy": 0.2361130565404892 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.03690455806165216, | |
| "joint_other_seed": 0.12005388129326222, | |
| "single_other_seed": 0.15212360030252592, | |
| "energy": 0.04467291136582693 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.08233981353276598, | |
| "joint_other_seed": 0.05551739958286217, | |
| "single_other_seed": 0.37480263007423154, | |
| "energy": 12.342458407084147 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.17862053367733188, | |
| "joint_other_seed": 0.2136184444578032, | |
| "single_other_seed": 0.533085126560215, | |
| "energy": 32.257493336995445 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.006644243510924408, | |
| "joint_other_seed": 0.06857123050442337, | |
| "single_other_seed": 0.0737956205272633, | |
| "energy": 0.2529975970586141 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.11888354476409302, | |
| "joint_other_seed": 0.4427935721603006, | |
| "single_other_seed": 0.45819984064497443, | |
| "energy": 36.800679524739586 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.013577718953705298, | |
| "joint_other_seed": 0.0908133883981892, | |
| "single_other_seed": 0.15490209856885284, | |
| "energy": 4.818599859873454 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.009859675829198111, | |
| "joint_other_seed": 0.03961219093285584, | |
| "single_other_seed": 0.02591546913443521, | |
| "energy": 0.585266629854838 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.02386575739905661, | |
| "joint_other_seed": 0.024757051464210256, | |
| "single_other_seed": 0.10571904046179226, | |
| "energy": 0.03798249860604604 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.16912695047095005, | |
| "joint_other_seed": 0.41188416692372143, | |
| "single_other_seed": 0.3332116614851647, | |
| "energy": 12.836616516113281 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.08338884922687673, | |
| "joint_other_seed": 0.15520865495411895, | |
| "single_other_seed": 0.2610238804870265, | |
| "energy": 16.0059757232666 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.011886860935393595, | |
| "joint_other_seed": 0.0623970493795003, | |
| "single_other_seed": 0.044884367788452506, | |
| "energy": 0.2197528084119161 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.12040908554211416, | |
| "joint_other_seed": 0.2244765934479207, | |
| "single_other_seed": 0.5169659376043727, | |
| "energy": 21.059315363566082 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.13328141712462882, | |
| "joint_other_seed": 0.3432152163325981, | |
| "single_other_seed": 0.27519547969684754, | |
| "energy": 2.7877233028411865 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.06063984409680656, | |
| "joint_other_seed": 0.18804262127091043, | |
| "single_other_seed": 0.10343897429672366, | |
| "energy": 1.217750072479248 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.07963849694900804, | |
| "joint_other_seed": 0.1117491346934909, | |
| "single_other_seed": 0.036070010636146455, | |
| "energy": 0.3018124004205068 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.17368232588644084, | |
| "joint_other_seed": 0.47371559050678935, | |
| "single_other_seed": 0.41685351024797757, | |
| "energy": 12.722970008850098 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.17804087719369976, | |
| "joint_other_seed": 0.5499924986569786, | |
| "single_other_seed": 0.4136367904158723, | |
| "energy": 15.783662478129068 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.01674669385817218, | |
| "joint_other_seed": 0.06647551427195242, | |
| "single_other_seed": 0.07608046237333232, | |
| "energy": 0.4720759689807892 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.05032503440568037, | |
| "joint_other_seed": 0.15518685408688215, | |
| "single_other_seed": 0.2986795979905751, | |
| "energy": 32.04583549499512 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.020990434677321756, | |
| "joint_other_seed": 0.1312590507325342, | |
| "single_other_seed": 0.020644501107136448, | |
| "energy": 5.2509307861328125 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.10619219849422645, | |
| "joint_other_seed": 0.32921464636615205, | |
| "single_other_seed": 0.29108682944240805, | |
| "energy": 2.2301189502080283 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.14839042745626663, | |
| "joint_other_seed": 0.18015084384817537, | |
| "single_other_seed": 0.389346712695, | |
| "energy": 0.42935073375701904 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.18390349356601843, | |
| "joint_other_seed": 0.6652992484906987, | |
| "single_other_seed": 0.38803882271023393, | |
| "energy": 29.28092892964681 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.2129625068769233, | |
| "joint_other_seed": 0.27690136391172226, | |
| "single_other_seed": 0.40672036489866437, | |
| "energy": 35.2887929280599 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.041382917739086755, | |
| "joint_other_seed": 0.13040831960672306, | |
| "single_other_seed": 0.11039686489547995, | |
| "energy": 1.1214876572291057 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.04607750293626789, | |
| "joint_other_seed": 0.31669782283338166, | |
| "single_other_seed": 0.2374804749622339, | |
| "energy": 87.47171401977539 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.002036688817975807, | |
| "joint_other_seed": 0.040256971536408614, | |
| "single_other_seed": 0.05083346159488045, | |
| "energy": 8.736092408498129 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.24394925013613822, | |
| "joint_other_seed": 0.5769253407091057, | |
| "single_other_seed": 0.37963640165343016, | |
| "energy": 4.416626373926799 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.580255185838451, | |
| "joint_other_seed": 0.7959975599108051, | |
| "single_other_seed": 0.699225835530336, | |
| "energy": 3.9649970531463623 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.29325362936175225, | |
| "joint_other_seed": 0.5632772735353162, | |
| "single_other_seed": 0.5687990590898997, | |
| "energy": 40.220689137776695 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.24710212366182982, | |
| "joint_other_seed": 0.07182455849938008, | |
| "single_other_seed": 0.6401642965556913, | |
| "energy": 32.089619636535645 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.2644630988092015, | |
| "joint_other_seed": 0.4545564315925388, | |
| "single_other_seed": 0.3482208047103039, | |
| "energy": 3.213616132736206 | |
| } | |
| }, | |
| "trained/contrast/r3": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.17010646471733104, | |
| "joint_other_seed": 0.6139125464174384, | |
| "single_other_seed": 0.48587948332480346, | |
| "energy": 19234.0166015625 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.39931908280201633, | |
| "joint_other_seed": 0.8202753921984236, | |
| "single_other_seed": 0.7243091349223783, | |
| "energy": 2317.2754720052085 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.23986821364563138, | |
| "joint_other_seed": 0.8005799066166266, | |
| "single_other_seed": 0.6368519846345397, | |
| "energy": 270.02516682942706 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.35140311190654705, | |
| "joint_other_seed": 0.7543561469729885, | |
| "single_other_seed": 0.760534927951505, | |
| "energy": 0.40501729647318524 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.16198830323494506, | |
| "joint_other_seed": 0.46996482577022547, | |
| "single_other_seed": 0.5441413325132178, | |
| "energy": 1662.8070068359375 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.12350800408745766, | |
| "joint_other_seed": 0.3446890402169489, | |
| "single_other_seed": 0.1932504611807233, | |
| "energy": 1003.241709391276 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.18036117862282175, | |
| "joint_other_seed": 0.6662307910475372, | |
| "single_other_seed": 0.4552290351023257, | |
| "energy": 9.164018313090006 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.273826687012923, | |
| "joint_other_seed": 0.5175818590514629, | |
| "single_other_seed": 0.2647264410806427, | |
| "energy": 404.1960754394531 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.3377876623182037, | |
| "joint_other_seed": 0.7526240466031844, | |
| "single_other_seed": 0.5878925509667523, | |
| "energy": 584.2193298339844 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.49528071620755, | |
| "joint_other_seed": 0.7120327942920848, | |
| "single_other_seed": 0.8011841023733169, | |
| "energy": 20.387056986490887 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.2977124936288189, | |
| "joint_other_seed": 0.5230387285387146, | |
| "single_other_seed": 0.6177728718745246, | |
| "energy": 0.15216578294833502 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.07002903258879745, | |
| "joint_other_seed": 0.04400453702605367, | |
| "single_other_seed": 0.15763238084996323, | |
| "energy": 180.77486165364584 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.05072261062466177, | |
| "joint_other_seed": 0.43821464806617005, | |
| "single_other_seed": 0.27112513696978946, | |
| "energy": 203.35321044921875 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.06711414282983874, | |
| "joint_other_seed": 0.10550510450528228, | |
| "single_other_seed": 0.37701066731825883, | |
| "energy": 0.2867518166700999 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.20627557375563899, | |
| "joint_other_seed": 0.49387991205121173, | |
| "single_other_seed": 0.3343101885625702, | |
| "energy": 344.57752482096356 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.36992652155067185, | |
| "joint_other_seed": 0.6355475982575012, | |
| "single_other_seed": 0.296530556530737, | |
| "energy": 33.63620249430338 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.04835089932189551, | |
| "joint_other_seed": 0.2630878968932385, | |
| "single_other_seed": 0.42978460011696584, | |
| "energy": 3.210548480351766 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.07240134739162085, | |
| "joint_other_seed": 0.15346638104317634, | |
| "single_other_seed": 0.3366246367264273, | |
| "energy": 0.04390159621834755 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.03207733617213049, | |
| "joint_other_seed": 0.5507778956646228, | |
| "single_other_seed": 0.4651934549202552, | |
| "energy": 382.1139322916667 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.07373349473784407, | |
| "joint_other_seed": 0.44906737924040807, | |
| "single_other_seed": 0.486970793053885, | |
| "energy": 373.6040445963542 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.03262270372275522, | |
| "joint_other_seed": 0.06590262017357751, | |
| "single_other_seed": 0.12299144114097349, | |
| "energy": 0.3059568752845128 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.14015037760426985, | |
| "joint_other_seed": 0.6566351623755946, | |
| "single_other_seed": 0.4152346533083575, | |
| "energy": 680.1048787434896 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.09586793916795266, | |
| "joint_other_seed": 0.18295765436576994, | |
| "single_other_seed": 0.11030511850948532, | |
| "energy": 28.581904729207356 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.05033709736550296, | |
| "joint_other_seed": 0.20359081259848455, | |
| "single_other_seed": 0.05167793859276493, | |
| "energy": 0.6510884364446005 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.019591357238056162, | |
| "joint_other_seed": 0.09403029516079109, | |
| "single_other_seed": 0.35150394918660266, | |
| "energy": 0.26111594835917157 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.06901325238342326, | |
| "joint_other_seed": 0.22967774118706585, | |
| "single_other_seed": 0.26805661626541666, | |
| "energy": 312.4721984863281 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.10824674489073072, | |
| "joint_other_seed": 0.36042471176304525, | |
| "single_other_seed": 0.2717621192367111, | |
| "energy": 402.05442301432294 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.05414666169367044, | |
| "joint_other_seed": 0.16931133893358452, | |
| "single_other_seed": 0.06782404834330469, | |
| "energy": 1.0880316297213237 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.16697375573756368, | |
| "joint_other_seed": 0.4445715528637888, | |
| "single_other_seed": 0.38853736803957095, | |
| "energy": 58.59997049967448 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.10072531135355034, | |
| "joint_other_seed": 0.22344405378066756, | |
| "single_other_seed": 0.09271719625673758, | |
| "energy": 6.0073879559834795 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.028431890065034104, | |
| "joint_other_seed": 0.04650076768630507, | |
| "single_other_seed": 0.008274084377512092, | |
| "energy": 0.7193136413892111 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.010594817831487746, | |
| "joint_other_seed": 0.03830888505382635, | |
| "single_other_seed": 0.14473034896500814, | |
| "energy": 0.02969462548693021 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.033362334504530744, | |
| "joint_other_seed": 0.20235829672978078, | |
| "single_other_seed": 0.435110237073146, | |
| "energy": 192.49266560872397 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.05678759619451412, | |
| "joint_other_seed": 0.36413811516761574, | |
| "single_other_seed": 0.3246355327060052, | |
| "energy": 337.13086954752606 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.02569671576281396, | |
| "joint_other_seed": 0.02795171778437988, | |
| "single_other_seed": 0.04923025870806208, | |
| "energy": 0.24716521799564362 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.07830969002775057, | |
| "joint_other_seed": 0.2949446603135516, | |
| "single_other_seed": 0.5181498942594329, | |
| "energy": 87.84032440185547 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.24124483186834833, | |
| "joint_other_seed": 0.6275505012716596, | |
| "single_other_seed": 0.15697565887361753, | |
| "energy": 84.80374654134114 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.1236041794117052, | |
| "joint_other_seed": 0.20778886871238356, | |
| "single_other_seed": 0.1973466321030315, | |
| "energy": 4.096524079640706 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.19709919505241547, | |
| "joint_other_seed": 0.6201092682235163, | |
| "single_other_seed": 0.5224868217359202, | |
| "energy": 0.828405479590098 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.05306469417376636, | |
| "joint_other_seed": 0.28685552282020677, | |
| "single_other_seed": 0.15117191476887493, | |
| "energy": 216.2823486328125 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.09472261302472729, | |
| "joint_other_seed": 0.40071012141954987, | |
| "single_other_seed": 0.5634331571340561, | |
| "energy": 214.1713409423828 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.004162861433483767, | |
| "joint_other_seed": 0.08161122015457911, | |
| "single_other_seed": 0.09654409277635945, | |
| "energy": 0.419976403315862 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.33599743935621507, | |
| "joint_other_seed": 0.7738538906165858, | |
| "single_other_seed": 0.6922795806250662, | |
| "energy": 231.91500854492188 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.31859726466927707, | |
| "joint_other_seed": 0.8496534432157805, | |
| "single_other_seed": 0.5542114947727996, | |
| "energy": 315.8881022135417 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.11371856778832456, | |
| "joint_other_seed": 0.28154375851636293, | |
| "single_other_seed": 0.34817693558750884, | |
| "energy": 6.452053785324097 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.2806103215032736, | |
| "joint_other_seed": 0.5836512343889243, | |
| "single_other_seed": 0.7035667897393909, | |
| "energy": 10.61309560139974 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.10515256558690589, | |
| "joint_other_seed": 0.2723187728957762, | |
| "single_other_seed": 0.4565658516217236, | |
| "energy": 91.83759053548177 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.10312125311075981, | |
| "joint_other_seed": 0.2742967105050157, | |
| "single_other_seed": 0.2602501633700697, | |
| "energy": 74.77923583984375 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.01630543173994914, | |
| "joint_other_seed": 0.10255171280757513, | |
| "single_other_seed": 0.10943172991174543, | |
| "energy": 0.7992507417996725 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.039199866201983014, | |
| "joint_other_seed": 0.13391920429671444, | |
| "single_other_seed": 0.12258662755143679, | |
| "energy": 55.89650217692057 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.02116274670346463, | |
| "joint_other_seed": 0.03902716644816373, | |
| "single_other_seed": 0.01896949130803771, | |
| "energy": 4.380404710769653 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.017683349299963513, | |
| "joint_other_seed": 0.015793376597674436, | |
| "single_other_seed": 0.09990828160962073, | |
| "energy": 0.4485498269399007 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.03459282023016186, | |
| "joint_other_seed": 0.09095274086822824, | |
| "single_other_seed": 0.15212360030252592, | |
| "energy": 0.07613001018762589 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.058345793297831294, | |
| "joint_other_seed": 0.147186128879496, | |
| "single_other_seed": 0.37480263007423154, | |
| "energy": 28.415136337280273 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.13218856087577002, | |
| "joint_other_seed": 0.27742916548231, | |
| "single_other_seed": 0.533085126560215, | |
| "energy": 41.585428873697914 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.005310514741734607, | |
| "joint_other_seed": 0.050885187520966046, | |
| "single_other_seed": 0.0737956205272633, | |
| "energy": 0.42966462175051373 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.1001009406927903, | |
| "joint_other_seed": 0.3688822300755886, | |
| "single_other_seed": 0.45819984064497443, | |
| "energy": 37.600624084472656 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.022476585059574616, | |
| "joint_other_seed": 0.09096512232275067, | |
| "single_other_seed": 0.15490209856885284, | |
| "energy": 4.557456811269124 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.014481512185820278, | |
| "joint_other_seed": 0.08641492292378239, | |
| "single_other_seed": 0.02591546913443521, | |
| "energy": 0.7493685086568197 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.020749543498994605, | |
| "joint_other_seed": 0.18188579109113845, | |
| "single_other_seed": 0.10571904046179226, | |
| "energy": 0.07350461681683858 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.11565839749408191, | |
| "joint_other_seed": 0.26092236803292934, | |
| "single_other_seed": 0.3332116614851647, | |
| "energy": 20.78656768798828 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.03901216826030052, | |
| "joint_other_seed": 0.2083521059642066, | |
| "single_other_seed": 0.2610238804870265, | |
| "energy": 22.987594604492188 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.007854452042568849, | |
| "joint_other_seed": 0.06950063410848273, | |
| "single_other_seed": 0.044884367788452506, | |
| "energy": 0.46611035863558453 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.07221559599647076, | |
| "joint_other_seed": 0.19578997422382782, | |
| "single_other_seed": 0.5169659376043727, | |
| "energy": 23.161670049031574 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.06692507366832656, | |
| "joint_other_seed": 0.14451320229641673, | |
| "single_other_seed": 0.27519547969684754, | |
| "energy": 3.663823048273722 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.03449965271544009, | |
| "joint_other_seed": 0.2349300888882998, | |
| "single_other_seed": 0.10343897429672366, | |
| "energy": 1.3879763682683308 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.08178952922327298, | |
| "joint_other_seed": 0.2611321546255678, | |
| "single_other_seed": 0.036070010636146455, | |
| "energy": 0.6819352805614471 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.10865489282105101, | |
| "joint_other_seed": 0.22981370595882233, | |
| "single_other_seed": 0.41685351024797757, | |
| "energy": 21.52891794840495 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.10816062099572654, | |
| "joint_other_seed": 0.27913500806068503, | |
| "single_other_seed": 0.4136367904158723, | |
| "energy": 19.8917662302653 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.02145218945760874, | |
| "joint_other_seed": 0.13872450576284867, | |
| "single_other_seed": 0.07608046237333232, | |
| "energy": 0.7780904769897461 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.07030609351822037, | |
| "joint_other_seed": 0.22695084935010879, | |
| "single_other_seed": 0.2986795979905751, | |
| "energy": 31.024306615193684 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.02690176318302373, | |
| "joint_other_seed": 0.1924226964648329, | |
| "single_other_seed": 0.020644501107136448, | |
| "energy": 4.280163129170735 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.1224008557143893, | |
| "joint_other_seed": 0.5389797370369038, | |
| "single_other_seed": 0.29108682944240805, | |
| "energy": 2.329963525136312 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.2085677213543662, | |
| "joint_other_seed": 0.4459070460190897, | |
| "single_other_seed": 0.389346712695, | |
| "energy": 0.6365281144777933 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.16165843457427442, | |
| "joint_other_seed": 0.37762383746347117, | |
| "single_other_seed": 0.38803882271023393, | |
| "energy": 32.18395868937174 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.13872992000091278, | |
| "joint_other_seed": 0.39768011034928297, | |
| "single_other_seed": 0.40672036489866437, | |
| "energy": 37.444541931152344 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.07334681022372425, | |
| "joint_other_seed": 0.2924288066139894, | |
| "single_other_seed": 0.11039686489547995, | |
| "energy": 1.7441280682881672 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.06379038821861538, | |
| "joint_other_seed": 0.3207119583695762, | |
| "single_other_seed": 0.2374804749622339, | |
| "energy": 43.84257888793945 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.01597525878293926, | |
| "joint_other_seed": 0.05437653462311187, | |
| "single_other_seed": 0.05083346159488045, | |
| "energy": 5.665256023406982 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.23212328005251837, | |
| "joint_other_seed": 0.7413908991270375, | |
| "single_other_seed": 0.37963640165343016, | |
| "energy": 4.366843541463216 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.5458349141230188, | |
| "joint_other_seed": 0.8022970662325721, | |
| "single_other_seed": 0.699225835530336, | |
| "energy": 3.3591512044270835 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.25232811188271265, | |
| "joint_other_seed": 0.5281766994529856, | |
| "single_other_seed": 0.5687990590898997, | |
| "energy": 49.47330093383789 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.2563433036077226, | |
| "joint_other_seed": 0.5931803216129194, | |
| "single_other_seed": 0.6401642965556913, | |
| "energy": 52.74034118652344 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.22000002672162522, | |
| "joint_other_seed": 0.547862350418886, | |
| "single_other_seed": 0.3482208047103039, | |
| "energy": 5.829241434733073 | |
| } | |
| }, | |
| "trained/contrast/r16": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.18985009381070006, | |
| "joint_other_seed": 0.7540195402893833, | |
| "single_other_seed": 0.48587948332480346, | |
| "energy": 15344.231770833334 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.45101715873772064, | |
| "joint_other_seed": 0.8594516160746031, | |
| "single_other_seed": 0.7243091349223783, | |
| "energy": 1683.475341796875 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.283275282175223, | |
| "joint_other_seed": 0.8292548299959472, | |
| "single_other_seed": 0.6368519846345397, | |
| "energy": 248.01134236653647 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.38063448667623423, | |
| "joint_other_seed": 0.828929373791224, | |
| "single_other_seed": 0.760534927951505, | |
| "energy": 0.4551767408847809 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.20414724490755098, | |
| "joint_other_seed": 0.6623772057301531, | |
| "single_other_seed": 0.5441413325132178, | |
| "energy": 1099.0692952473958 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.13897022201621312, | |
| "joint_other_seed": 0.6852409083906779, | |
| "single_other_seed": 0.1932504611807233, | |
| "energy": 1189.9245198567708 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.23152568826637365, | |
| "joint_other_seed": 0.7924636721276561, | |
| "single_other_seed": 0.4552290351023257, | |
| "energy": 6.657144228617351 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.3212446617316256, | |
| "joint_other_seed": 0.5708896243282546, | |
| "single_other_seed": 0.2647264410806427, | |
| "energy": 194.62755584716797 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.3654073695505575, | |
| "joint_other_seed": 0.8310044092823425, | |
| "single_other_seed": 0.5878925509667523, | |
| "energy": 479.6976725260417 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.5040240980802042, | |
| "joint_other_seed": 0.7802856022175768, | |
| "single_other_seed": 0.8011841023733169, | |
| "energy": 13.89929453531901 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.28904725731434, | |
| "joint_other_seed": 0.7176526653317586, | |
| "single_other_seed": 0.6177728718745246, | |
| "energy": 0.2126318315664927 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.056885939734942714, | |
| "joint_other_seed": 0.637434179140163, | |
| "single_other_seed": 0.15763238084996323, | |
| "energy": 264.76697794596356 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.05807131446091607, | |
| "joint_other_seed": 0.6174308103042802, | |
| "single_other_seed": 0.27112513696978946, | |
| "energy": 153.7741241455078 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.0784768944733708, | |
| "joint_other_seed": 0.34422300336928624, | |
| "single_other_seed": 0.37701066731825883, | |
| "energy": 0.24573844174544016 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.19387933329150042, | |
| "joint_other_seed": 0.7664164472489479, | |
| "single_other_seed": 0.3343101885625702, | |
| "energy": 318.2289326985677 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.4016325292812538, | |
| "joint_other_seed": 0.7760581860027339, | |
| "single_other_seed": 0.296530556530737, | |
| "energy": 30.47664515177409 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.03417249645399467, | |
| "joint_other_seed": 0.5058557624399165, | |
| "single_other_seed": 0.42978460011696584, | |
| "energy": 3.553188959757487 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.052878427250320104, | |
| "joint_other_seed": 0.38462999138585907, | |
| "single_other_seed": 0.3366246367264273, | |
| "energy": 0.0603558657070001 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.041006916978658214, | |
| "joint_other_seed": 0.6349544130753763, | |
| "single_other_seed": 0.4651934549202552, | |
| "energy": 365.3296712239583 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.05764545795746667, | |
| "joint_other_seed": 0.6625374045403936, | |
| "single_other_seed": 0.486970793053885, | |
| "energy": 388.6643473307292 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.036680075749190966, | |
| "joint_other_seed": 0.35821378424260986, | |
| "single_other_seed": 0.12299144114097349, | |
| "energy": 0.43217435479164124 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.1699524798247436, | |
| "joint_other_seed": 0.7683815893865202, | |
| "single_other_seed": 0.4152346533083575, | |
| "energy": 555.8263142903646 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.08815452889714781, | |
| "joint_other_seed": 0.49898013191754764, | |
| "single_other_seed": 0.11030511850948532, | |
| "energy": 19.417185465494793 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.03613891784424972, | |
| "joint_other_seed": 0.2132431192464017, | |
| "single_other_seed": 0.05167793859276493, | |
| "energy": 0.8393538594245911 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.0376706645668472, | |
| "joint_other_seed": 0.3381581211335255, | |
| "single_other_seed": 0.35150394918660266, | |
| "energy": 0.3541514078776042 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.07688059173203088, | |
| "joint_other_seed": 0.644783011545036, | |
| "single_other_seed": 0.26805661626541666, | |
| "energy": 375.2253824869792 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.1477479742380179, | |
| "joint_other_seed": 0.5375040074529149, | |
| "single_other_seed": 0.2717621192367111, | |
| "energy": 418.7093912760417 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.07582311374125628, | |
| "joint_other_seed": 0.4553860488090261, | |
| "single_other_seed": 0.06782404834330469, | |
| "energy": 1.402848521868388 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.17344994736467345, | |
| "joint_other_seed": 0.5671999849124485, | |
| "single_other_seed": 0.38853736803957095, | |
| "energy": 50.16293970743815 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.10738854246057183, | |
| "joint_other_seed": 0.41714187543811576, | |
| "single_other_seed": 0.09271719625673758, | |
| "energy": 5.3013261159261065 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.033173968507707485, | |
| "joint_other_seed": 0.1996202980180255, | |
| "single_other_seed": 0.008274084377512092, | |
| "energy": 0.8904021978378296 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.010002582338407118, | |
| "joint_other_seed": 0.15612439877644124, | |
| "single_other_seed": 0.14473034896500814, | |
| "energy": 0.07287070900201797 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.0381113566997795, | |
| "joint_other_seed": 0.5338590212550816, | |
| "single_other_seed": 0.435110237073146, | |
| "energy": 207.63192240397134 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.05484967280727948, | |
| "joint_other_seed": 0.6068045122730348, | |
| "single_other_seed": 0.3246355327060052, | |
| "energy": 304.06798299153644 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.022039297535048852, | |
| "joint_other_seed": 0.2601171732590105, | |
| "single_other_seed": 0.04923025870806208, | |
| "energy": 0.49099764227867126 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.0992815399315431, | |
| "joint_other_seed": 0.4871709283054805, | |
| "single_other_seed": 0.5181498942594329, | |
| "energy": 77.91193389892578 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.2469707032920726, | |
| "joint_other_seed": 0.7403906896275427, | |
| "single_other_seed": 0.15697565887361753, | |
| "energy": 63.39901606241862 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.11449803169632633, | |
| "joint_other_seed": 0.3780711270559418, | |
| "single_other_seed": 0.1973466321030315, | |
| "energy": 2.5935325622558594 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.20969773484691068, | |
| "joint_other_seed": 0.6434087840250995, | |
| "single_other_seed": 0.5224868217359202, | |
| "energy": 0.7576847871144613 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.0561805217368152, | |
| "joint_other_seed": 0.558746470518558, | |
| "single_other_seed": 0.15117191476887493, | |
| "energy": 197.9088389078776 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.09732407603924509, | |
| "joint_other_seed": 0.5815517126361314, | |
| "single_other_seed": 0.5634331571340561, | |
| "energy": 210.0519053141276 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.006178914007003766, | |
| "joint_other_seed": 0.201936089554839, | |
| "single_other_seed": 0.09654409277635945, | |
| "energy": 0.6815288066864014 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.3414555203967319, | |
| "joint_other_seed": 0.8429381152370687, | |
| "single_other_seed": 0.6922795806250662, | |
| "energy": 217.7068634033203 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.2947861436953314, | |
| "joint_other_seed": 0.911387783585256, | |
| "single_other_seed": 0.5542114947727996, | |
| "energy": 303.73572794596356 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.11586555473825655, | |
| "joint_other_seed": 0.36834491464493824, | |
| "single_other_seed": 0.34817693558750884, | |
| "energy": 5.3632386525472 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.29705314902096425, | |
| "joint_other_seed": 0.7590221465528184, | |
| "single_other_seed": 0.7035667897393909, | |
| "energy": 14.356690406799316 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.1036655661105742, | |
| "joint_other_seed": 0.4268025382378291, | |
| "single_other_seed": 0.4565658516217236, | |
| "energy": 109.40492502848308 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.09706373069055085, | |
| "joint_other_seed": 0.3666119860498459, | |
| "single_other_seed": 0.2602501633700697, | |
| "energy": 78.06507110595703 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.01471235077380495, | |
| "joint_other_seed": 0.25518347617664794, | |
| "single_other_seed": 0.10943172991174543, | |
| "energy": 0.8949112892150879 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.05480415614338252, | |
| "joint_other_seed": 0.2674799084943881, | |
| "single_other_seed": 0.12258662755143679, | |
| "energy": 51.31196594238281 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.02050703440522006, | |
| "joint_other_seed": 0.11508459931721615, | |
| "single_other_seed": 0.01896949130803771, | |
| "energy": 2.9043431282043457 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.02260929139504886, | |
| "joint_other_seed": 0.14310310491309874, | |
| "single_other_seed": 0.09990828160962073, | |
| "energy": 0.6199389298756918 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.02793260972167, | |
| "joint_other_seed": 0.2149228908954981, | |
| "single_other_seed": 0.15212360030252592, | |
| "energy": 0.11352902899185817 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.04638156543060701, | |
| "joint_other_seed": 0.2827485733527385, | |
| "single_other_seed": 0.37480263007423154, | |
| "energy": 42.27107493082682 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.12245382042567043, | |
| "joint_other_seed": 0.3860382995340359, | |
| "single_other_seed": 0.533085126560215, | |
| "energy": 40.3581288655599 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.006426476997143826, | |
| "joint_other_seed": 0.15695932689895875, | |
| "single_other_seed": 0.0737956205272633, | |
| "energy": 0.748854915301005 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.09179933886898306, | |
| "joint_other_seed": 0.4642385740112299, | |
| "single_other_seed": 0.45819984064497443, | |
| "energy": 37.367777506510414 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.025925332611485774, | |
| "joint_other_seed": 0.1847525616354304, | |
| "single_other_seed": 0.15490209856885284, | |
| "energy": 4.394463221232097 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.014417064796492185, | |
| "joint_other_seed": 0.19685430695972927, | |
| "single_other_seed": 0.02591546913443521, | |
| "energy": 0.6992279291152954 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.015458842833984021, | |
| "joint_other_seed": 0.2146271383555356, | |
| "single_other_seed": 0.10571904046179226, | |
| "energy": 0.11557281017303467 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.07782698181225167, | |
| "joint_other_seed": 0.3263550114615953, | |
| "single_other_seed": 0.3332116614851647, | |
| "energy": 26.72911834716797 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.045237035977576445, | |
| "joint_other_seed": 0.3433471427805347, | |
| "single_other_seed": 0.2610238804870265, | |
| "energy": 33.62341054280599 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.007861832479782893, | |
| "joint_other_seed": 0.17520023720584577, | |
| "single_other_seed": 0.044884367788452506, | |
| "energy": 0.7809054652849833 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.09004299631599877, | |
| "joint_other_seed": 0.38183741084716194, | |
| "single_other_seed": 0.5169659376043727, | |
| "energy": 20.474997838338215 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.07646014772421306, | |
| "joint_other_seed": 0.2630814326301312, | |
| "single_other_seed": 0.27519547969684754, | |
| "energy": 2.8846232891082764 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.03337616828139678, | |
| "joint_other_seed": 0.3892454649481979, | |
| "single_other_seed": 0.10343897429672366, | |
| "energy": 1.300816535949707 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.059128910959483694, | |
| "joint_other_seed": 0.35168972879034843, | |
| "single_other_seed": 0.036070010636146455, | |
| "energy": 1.4811168511708577 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.07748043367422885, | |
| "joint_other_seed": 0.3162782414081521, | |
| "single_other_seed": 0.41685351024797757, | |
| "energy": 29.08724848429362 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.0720680353869817, | |
| "joint_other_seed": 0.3474552427486016, | |
| "single_other_seed": 0.4136367904158723, | |
| "energy": 28.287079493204754 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.029480187459467677, | |
| "joint_other_seed": 0.296103668978273, | |
| "single_other_seed": 0.07608046237333232, | |
| "energy": 1.1812771558761597 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.07966883748496023, | |
| "joint_other_seed": 0.5023108230875075, | |
| "single_other_seed": 0.2986795979905751, | |
| "energy": 28.200546264648438 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.018608023580117705, | |
| "joint_other_seed": 0.19160980379977655, | |
| "single_other_seed": 0.020644501107136448, | |
| "energy": 3.4827138582865396 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.12411742490327027, | |
| "joint_other_seed": 0.6225407113318102, | |
| "single_other_seed": 0.29108682944240805, | |
| "energy": 2.1847098668416343 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.230178152659275, | |
| "joint_other_seed": 0.6723589985156556, | |
| "single_other_seed": 0.389346712695, | |
| "energy": 0.6721501350402832 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.12941112070643035, | |
| "joint_other_seed": 0.4855999956261845, | |
| "single_other_seed": 0.38803882271023393, | |
| "energy": 33.08419926961263 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.11798009372502216, | |
| "joint_other_seed": 0.49233619965226405, | |
| "single_other_seed": 0.40672036489866437, | |
| "energy": 41.387980143229164 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.06846057508376202, | |
| "joint_other_seed": 0.468334187586116, | |
| "single_other_seed": 0.11039686489547995, | |
| "energy": 1.712796966234843 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.08939512128132465, | |
| "joint_other_seed": 0.4713296022365111, | |
| "single_other_seed": 0.2374804749622339, | |
| "energy": 34.23164304097494 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.020934242018224358, | |
| "joint_other_seed": 0.19795934505683999, | |
| "single_other_seed": 0.05083346159488045, | |
| "energy": 3.6320525805155435 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.2543662744019305, | |
| "joint_other_seed": 0.8280128916053853, | |
| "single_other_seed": 0.37963640165343016, | |
| "energy": 3.930346647898356 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.5631248943055185, | |
| "joint_other_seed": 0.8478407778436997, | |
| "single_other_seed": 0.699225835530336, | |
| "energy": 3.2172398567199707 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.23538688275469644, | |
| "joint_other_seed": 0.6553207365123596, | |
| "single_other_seed": 0.5687990590898997, | |
| "energy": 57.91364288330078 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.2553434994027529, | |
| "joint_other_seed": 0.6396393133285614, | |
| "single_other_seed": 0.6401642965556913, | |
| "energy": 49.91226323445638 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.21590142649507582, | |
| "joint_other_seed": 0.6752822337641643, | |
| "single_other_seed": 0.3482208047103039, | |
| "energy": 4.492636044820149 | |
| } | |
| }, | |
| "trained/trigger/r1": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.15451051906034963, | |
| "joint_other_seed": 0.25066257595661257, | |
| "single_other_seed": 0.4640814665196763, | |
| "energy": 35014.656901041664 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.31352552625839525, | |
| "joint_other_seed": 0.5783778181832208, | |
| "single_other_seed": 0.7058980561512407, | |
| "energy": 4641.806640625 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.171673751704552, | |
| "joint_other_seed": 0.5431096901657956, | |
| "single_other_seed": 0.6257225234283108, | |
| "energy": 153.92414093017578 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.3204090380682034, | |
| "joint_other_seed": 0.6434533534785254, | |
| "single_other_seed": 0.7067756317147811, | |
| "energy": 2.1194138526916504 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.13620408577661358, | |
| "joint_other_seed": 0.2099288799712361, | |
| "single_other_seed": 0.5279037902428164, | |
| "energy": 3455.5336100260415 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.10269234862698606, | |
| "joint_other_seed": 0.3689170297341574, | |
| "single_other_seed": 0.10972752188308754, | |
| "energy": 2980.1641845703125 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.07574471098207314, | |
| "joint_other_seed": 0.16660416376310716, | |
| "single_other_seed": 0.2146551506678699, | |
| "energy": 10.779833674430847 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.3344307623617483, | |
| "joint_other_seed": 0.6543391877658598, | |
| "single_other_seed": 0.5243573719231198, | |
| "energy": 3145.2178548177085 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.09581331015630912, | |
| "joint_other_seed": 0.1403275252713574, | |
| "single_other_seed": 0.42690857302758467, | |
| "energy": 704.4025472005209 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.35534861672886653, | |
| "joint_other_seed": 0.5209144262021221, | |
| "single_other_seed": 0.6310774005552311, | |
| "energy": 32.8062629699707 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.26120407435821613, | |
| "joint_other_seed": 0.4550251971651835, | |
| "single_other_seed": 0.5317387276711343, | |
| "energy": 1.001305103302002 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.08405194807098737, | |
| "joint_other_seed": 0.15020904989410067, | |
| "single_other_seed": 0.10110999403717336, | |
| "energy": 333.06341552734375 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.15981920235391878, | |
| "joint_other_seed": 0.44237214004394165, | |
| "single_other_seed": 0.24561641327817219, | |
| "energy": 449.7085876464844 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.08602302842930998, | |
| "joint_other_seed": 0.24913834058601825, | |
| "single_other_seed": 0.3126659865167023, | |
| "energy": 0.9550197124481201 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.23170519569089099, | |
| "joint_other_seed": 0.33559180427097723, | |
| "single_other_seed": 0.31195102073494996, | |
| "energy": 2671.9983723958335 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.1168924835579418, | |
| "joint_other_seed": 0.15076345001299707, | |
| "single_other_seed": 0.17731184302562808, | |
| "energy": 312.36696370442706 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.055312610913512916, | |
| "joint_other_seed": 0.1824052828084186, | |
| "single_other_seed": 0.3641591963136199, | |
| "energy": 7.446090221405029 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.09874231846584536, | |
| "joint_other_seed": 0.13384105976802407, | |
| "single_other_seed": 0.26231468890071385, | |
| "energy": 0.6842431227366129 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.1296925886630289, | |
| "joint_other_seed": 0.24163760001059176, | |
| "single_other_seed": 0.4054313432429959, | |
| "energy": 1074.9021708170574 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.17378388206260498, | |
| "joint_other_seed": 0.43969849327729493, | |
| "single_other_seed": 0.34236778218793673, | |
| "energy": 1003.5683390299479 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.021581440081735237, | |
| "joint_other_seed": 0.18045882911775835, | |
| "single_other_seed": 0.1449921097768857, | |
| "energy": 1.4082310199737549 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.05314729217083569, | |
| "joint_other_seed": 0.15754033131415213, | |
| "single_other_seed": 0.33217953036158376, | |
| "energy": 3102.4209798177085 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.04547947705087341, | |
| "joint_other_seed": 0.053758899533292714, | |
| "single_other_seed": 0.04703559916233768, | |
| "energy": 199.97027333577475 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.05410097329293029, | |
| "joint_other_seed": 0.1761015965187417, | |
| "single_other_seed": 0.11428907680284839, | |
| "energy": 6.9742452303568525 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.0272776082621059, | |
| "joint_other_seed": 0.01792797594075579, | |
| "single_other_seed": 0.14680320279747328, | |
| "energy": 1.1320514678955078 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.10119611128876072, | |
| "joint_other_seed": 0.08268314389904197, | |
| "single_other_seed": 0.21309014883994815, | |
| "energy": 1006.6614990234375 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.0741734226803552, | |
| "joint_other_seed": 0.15646290037733124, | |
| "single_other_seed": 0.23485787347265896, | |
| "energy": 917.2846476236979 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.009786919383027448, | |
| "joint_other_seed": 0.042687232660148844, | |
| "single_other_seed": 0.06404832232692637, | |
| "energy": 2.2056153615315757 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.13652585513278848, | |
| "joint_other_seed": 0.4645968613572671, | |
| "single_other_seed": 0.32597502466099537, | |
| "energy": 622.0616251627604 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.039584319677306246, | |
| "joint_other_seed": 0.1254189142519866, | |
| "single_other_seed": 0.1107848176380748, | |
| "energy": 128.2563705444336 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.006548068984222444, | |
| "joint_other_seed": 0.012847325678347232, | |
| "single_other_seed": 0.01053277394584129, | |
| "energy": 6.244124571482341 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.026709151244181273, | |
| "joint_other_seed": 0.011050399324222915, | |
| "single_other_seed": 0.12247141934528094, | |
| "energy": 1.0542799631754558 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.0845095175575209, | |
| "joint_other_seed": 0.10289642254962095, | |
| "single_other_seed": 0.3083956216804902, | |
| "energy": 1165.7970377604167 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.08337776217178183, | |
| "joint_other_seed": 0.09413464233812263, | |
| "single_other_seed": 0.17717077114997415, | |
| "energy": 1307.1453857421875 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.0013202706211939974, | |
| "joint_other_seed": 0.006927317817230216, | |
| "single_other_seed": 0.011090593093745826, | |
| "energy": 2.694811701774597 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.05561257900418781, | |
| "joint_other_seed": 0.3179029830763076, | |
| "single_other_seed": 0.32635573441951116, | |
| "energy": 2546.486083984375 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.12886231222988742, | |
| "joint_other_seed": 0.2655248413992862, | |
| "single_other_seed": 0.0822492566724442, | |
| "energy": 708.91796875 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.032620563497636645, | |
| "joint_other_seed": 0.1842249932203146, | |
| "single_other_seed": 0.19101030765684135, | |
| "energy": 18.467161178588867 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.04466604525529927, | |
| "joint_other_seed": 0.1845628448687482, | |
| "single_other_seed": 0.20053346330625577, | |
| "energy": 2.783588091532389 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.12282884366491831, | |
| "joint_other_seed": 0.21842188200642143, | |
| "single_other_seed": 0.17376045014722985, | |
| "energy": 507.5249430338542 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.151027032598249, | |
| "joint_other_seed": 0.457861350435116, | |
| "single_other_seed": 0.31521473347233675, | |
| "energy": 1115.677734375 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.002054114333100432, | |
| "joint_other_seed": 0.009783797754429835, | |
| "single_other_seed": 0.03958595341675741, | |
| "energy": 4.574832280476888 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.05336890407659112, | |
| "joint_other_seed": 0.2057686105496136, | |
| "single_other_seed": 0.5512641641259381, | |
| "energy": 6751.358072916667 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.17925242859309776, | |
| "joint_other_seed": 0.48761845677170235, | |
| "single_other_seed": 0.4493563751423172, | |
| "energy": 987.1809895833334 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.020944039525827574, | |
| "joint_other_seed": 0.004274741480557509, | |
| "single_other_seed": 0.1530650712676246, | |
| "energy": 27.06903330485026 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.029017402351267707, | |
| "joint_other_seed": 0.014508880895149123, | |
| "single_other_seed": 0.5926320611047866, | |
| "energy": 5.748664379119873 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.10630676437901432, | |
| "joint_other_seed": 0.13996799256899783, | |
| "single_other_seed": 0.28004040888245474, | |
| "energy": 352.15465291341144 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.06474751638019616, | |
| "joint_other_seed": 0.25692423879867954, | |
| "single_other_seed": 0.1531286417764227, | |
| "energy": 339.28663126627606 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.0019275143473038714, | |
| "joint_other_seed": 0.0008934211536729686, | |
| "single_other_seed": 0.05571162213978631, | |
| "energy": 5.429182529449463 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.010725817185994413, | |
| "joint_other_seed": 0.15634657175791822, | |
| "single_other_seed": 0.046613092208021234, | |
| "energy": 1985.802734375 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.0014324899920949985, | |
| "joint_other_seed": 0.000534641490533394, | |
| "single_other_seed": 0.008714388740887702, | |
| "energy": 118.8533935546875 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.006288193411478238, | |
| "joint_other_seed": 0.04741371745855527, | |
| "single_other_seed": 0.10173713811795682, | |
| "energy": 8.984479268391928 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.01736934092911291, | |
| "joint_other_seed": 0.12773849201346588, | |
| "single_other_seed": 0.07718146054211683, | |
| "energy": 4.2615540822347 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.04976779858742675, | |
| "joint_other_seed": 0.030085014493132722, | |
| "single_other_seed": 0.13987515849153442, | |
| "energy": 364.0401611328125 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.07392213174624052, | |
| "joint_other_seed": 0.09366414808029078, | |
| "single_other_seed": 0.2947735213465493, | |
| "energy": 296.0897216796875 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.007490025557472801, | |
| "joint_other_seed": 0.04478942368376915, | |
| "single_other_seed": 0.07003675436717427, | |
| "energy": 7.879288037618001 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.20968695153437278, | |
| "joint_other_seed": 0.4817874383580202, | |
| "single_other_seed": 0.4036489963166986, | |
| "energy": 2000.923095703125 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.011063352692287864, | |
| "joint_other_seed": 0.04172490288468689, | |
| "single_other_seed": 0.08589367374924217, | |
| "energy": 180.4958953857422 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.0025555375444354933, | |
| "joint_other_seed": 0.01376186508893995, | |
| "single_other_seed": 0.03131645248795023, | |
| "energy": 13.400068600972494 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.010353990344803998, | |
| "joint_other_seed": 0.012900645100649921, | |
| "single_other_seed": 0.0720903165224952, | |
| "energy": 3.747355302174886 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.12614973595820556, | |
| "joint_other_seed": 0.38054834254942194, | |
| "single_other_seed": 0.28726696871990964, | |
| "energy": 380.05322265625 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.04814537276129971, | |
| "joint_other_seed": 0.14387237815114862, | |
| "single_other_seed": 0.1490903507489325, | |
| "energy": 558.9799397786459 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.008159758034267717, | |
| "joint_other_seed": 0.0561823470717695, | |
| "single_other_seed": 0.06471306796361498, | |
| "energy": 12.512087186177572 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.19991890609154006, | |
| "joint_other_seed": 0.30932590512505753, | |
| "single_other_seed": 0.4346874608327894, | |
| "energy": 1733.0971272786458 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.03442595951471206, | |
| "joint_other_seed": 0.2628155645128045, | |
| "single_other_seed": 0.18605141054193586, | |
| "energy": 178.40953572591147 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.018796279737676502, | |
| "joint_other_seed": 0.11610484425116296, | |
| "single_other_seed": 0.04564375872725601, | |
| "energy": 24.76534652709961 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.058348265699042945, | |
| "joint_other_seed": 0.07362502688932526, | |
| "single_other_seed": 0.07379616173301262, | |
| "energy": 11.489717165629068 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.14415171068909471, | |
| "joint_other_seed": 0.3656237205657926, | |
| "single_other_seed": 0.4077865584515338, | |
| "energy": 385.66351318359375 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.15180439526253897, | |
| "joint_other_seed": 0.43408123224837647, | |
| "single_other_seed": 0.39646240931103316, | |
| "energy": 402.3367513020833 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.011884816190246505, | |
| "joint_other_seed": 0.07386887131263331, | |
| "single_other_seed": 0.04666607555200741, | |
| "energy": 19.841841379801433 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.024861101837834965, | |
| "joint_other_seed": 0.08090283868874548, | |
| "single_other_seed": 0.2019864700250521, | |
| "energy": 1602.7982584635417 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.043751273135832824, | |
| "joint_other_seed": 0.14421238584488838, | |
| "single_other_seed": 0.03622367276169591, | |
| "energy": 251.86400858561197 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.050811510684154015, | |
| "joint_other_seed": 0.21684660903506692, | |
| "single_other_seed": 0.15444637955424534, | |
| "energy": 46.246229807535805 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.13840528784585515, | |
| "joint_other_seed": 0.21146714488710094, | |
| "single_other_seed": 0.3204860112740461, | |
| "energy": 19.20721435546875 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.12168291459553127, | |
| "joint_other_seed": 0.4183822766723061, | |
| "single_other_seed": 0.2967651150147088, | |
| "energy": 489.37652587890625 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.1160631571240257, | |
| "joint_other_seed": 0.1635855773488215, | |
| "single_other_seed": 0.274685485429328, | |
| "energy": 498.50095621744794 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.027177902052434977, | |
| "joint_other_seed": 0.08506485260130914, | |
| "single_other_seed": 0.10415474710253214, | |
| "energy": 47.57082748413086 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.0340467925926109, | |
| "joint_other_seed": 0.20409377564294817, | |
| "single_other_seed": 0.20648462611953777, | |
| "energy": 2573.97998046875 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.013134292631765354, | |
| "joint_other_seed": 0.047093001416166125, | |
| "single_other_seed": 0.07115237101009253, | |
| "energy": 255.36797078450522 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.20065129828445816, | |
| "joint_other_seed": 0.48601694305786125, | |
| "single_other_seed": 0.3931325246733607, | |
| "energy": 86.99032084147136 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.333485986653296, | |
| "joint_other_seed": 0.6331033625592899, | |
| "single_other_seed": 0.6503589322572455, | |
| "energy": 76.38252512613933 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.14469757290646257, | |
| "joint_other_seed": 0.3195897032637943, | |
| "single_other_seed": 0.43973067637614727, | |
| "energy": 488.00844319661456 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.08222530804046024, | |
| "joint_other_seed": 0.07736517726422276, | |
| "single_other_seed": 0.46652912189590057, | |
| "energy": 617.4787394205729 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.30571873150704487, | |
| "joint_other_seed": 0.4379921150770907, | |
| "single_other_seed": 0.4484410589151423, | |
| "energy": 155.59408060709634 | |
| } | |
| }, | |
| "trained/trigger/r3": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.21009524450126782, | |
| "joint_other_seed": 0.6098550582424869, | |
| "single_other_seed": 0.4640814665196763, | |
| "energy": 29869.397135416668 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.38905274833454484, | |
| "joint_other_seed": 0.8026972091345913, | |
| "single_other_seed": 0.7058980561512407, | |
| "energy": 2555.3976236979165 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.1841183969291208, | |
| "joint_other_seed": 0.7734998199463077, | |
| "single_other_seed": 0.6257225234283108, | |
| "energy": 159.78815205891928 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.3447420915921503, | |
| "joint_other_seed": 0.688135907085818, | |
| "single_other_seed": 0.7067756317147811, | |
| "energy": 2.0818745295206704 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.1576366127015718, | |
| "joint_other_seed": 0.46320992996954335, | |
| "single_other_seed": 0.5279037902428164, | |
| "energy": 2883.5923665364585 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.09015084217685365, | |
| "joint_other_seed": 0.2803434740379427, | |
| "single_other_seed": 0.10972752188308754, | |
| "energy": 1530.539794921875 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.10543447384523752, | |
| "joint_other_seed": 0.6066372957744347, | |
| "single_other_seed": 0.2146551506678699, | |
| "energy": 7.896860281626384 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.3342801242583646, | |
| "joint_other_seed": 0.5679869465130141, | |
| "single_other_seed": 0.5243573719231198, | |
| "energy": 2509.7342122395835 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.11675629206312899, | |
| "joint_other_seed": 0.6456465417801542, | |
| "single_other_seed": 0.42690857302758467, | |
| "energy": 1071.7510375976562 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.30394754650538625, | |
| "joint_other_seed": 0.5324395397769988, | |
| "single_other_seed": 0.6310774005552311, | |
| "energy": 34.49395942687988 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.16525292404054825, | |
| "joint_other_seed": 0.39013612973351497, | |
| "single_other_seed": 0.5317387276711343, | |
| "energy": 1.2610977490743 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.059689847346412694, | |
| "joint_other_seed": 0.06939440587612888, | |
| "single_other_seed": 0.10110999403717336, | |
| "energy": 750.9500122070312 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.07314469592532047, | |
| "joint_other_seed": 0.4109727477707038, | |
| "single_other_seed": 0.24561641327817219, | |
| "energy": 1100.2390950520833 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.10722984147617431, | |
| "joint_other_seed": 0.23734425819903748, | |
| "single_other_seed": 0.3126659865167023, | |
| "energy": 1.5579127073287964 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.22979490003534875, | |
| "joint_other_seed": 0.49512508581649217, | |
| "single_other_seed": 0.31195102073494996, | |
| "energy": 3013.357666015625 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.11522335363465841, | |
| "joint_other_seed": 0.49748488967628657, | |
| "single_other_seed": 0.17731184302562808, | |
| "energy": 236.57011922200522 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.021653565953150267, | |
| "joint_other_seed": 0.2016622183227622, | |
| "single_other_seed": 0.3641591963136199, | |
| "energy": 19.950766881306965 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.06995690942731592, | |
| "joint_other_seed": 0.16873855150483552, | |
| "single_other_seed": 0.26231468890071385, | |
| "energy": 1.1827263434727986 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.04842286162716203, | |
| "joint_other_seed": 0.46981466540175776, | |
| "single_other_seed": 0.4054313432429959, | |
| "energy": 1664.3075764973958 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.08926592169540493, | |
| "joint_other_seed": 0.3753045019232124, | |
| "single_other_seed": 0.34236778218793673, | |
| "energy": 1465.8033040364583 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.0118081767538335, | |
| "joint_other_seed": 0.0990709710527804, | |
| "single_other_seed": 0.1449921097768857, | |
| "energy": 2.559579372406006 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.125469408584867, | |
| "joint_other_seed": 0.5080006471967955, | |
| "single_other_seed": 0.33217953036158376, | |
| "energy": 2776.1451416015625 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.05897107382637634, | |
| "joint_other_seed": 0.1535397236488143, | |
| "single_other_seed": 0.04703559916233768, | |
| "energy": 207.63753763834634 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.03950351834944421, | |
| "joint_other_seed": 0.11814032153667575, | |
| "single_other_seed": 0.11428907680284839, | |
| "energy": 13.560680707295736 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.02044875164050064, | |
| "joint_other_seed": 0.15930316000232791, | |
| "single_other_seed": 0.14680320279747328, | |
| "energy": 2.201943596204122 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.0605832781656408, | |
| "joint_other_seed": 0.22731091367924616, | |
| "single_other_seed": 0.21309014883994815, | |
| "energy": 1344.961181640625 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.10121009710852436, | |
| "joint_other_seed": 0.31509993732124353, | |
| "single_other_seed": 0.23485787347265896, | |
| "energy": 1336.1524251302083 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.010079080932155815, | |
| "joint_other_seed": 0.07495620891966348, | |
| "single_other_seed": 0.06404832232692637, | |
| "energy": 4.351117928822835 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.09495608730797112, | |
| "joint_other_seed": 0.33595051955633903, | |
| "single_other_seed": 0.32597502466099537, | |
| "energy": 1208.7977498372395 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.04953715034173345, | |
| "joint_other_seed": 0.20246973571810567, | |
| "single_other_seed": 0.1107848176380748, | |
| "energy": 175.77812703450522 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.010811675352572325, | |
| "joint_other_seed": 0.023198690950766543, | |
| "single_other_seed": 0.01053277394584129, | |
| "energy": 13.435932795206705 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.02013239792310606, | |
| "joint_other_seed": 0.05970349912253059, | |
| "single_other_seed": 0.12247141934528094, | |
| "energy": 1.8066553274790447 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.03936160868436145, | |
| "joint_other_seed": 0.17568190537846262, | |
| "single_other_seed": 0.3083956216804902, | |
| "energy": 1160.2025756835938 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.03737813050101901, | |
| "joint_other_seed": 0.2928963971383791, | |
| "single_other_seed": 0.17717077114997415, | |
| "energy": 1728.6854654947917 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.0041853144639468765, | |
| "joint_other_seed": 0.08248859340545411, | |
| "single_other_seed": 0.011090593093745826, | |
| "energy": 6.123222827911377 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.05398364173822956, | |
| "joint_other_seed": 0.4477021425863629, | |
| "single_other_seed": 0.32635573441951116, | |
| "energy": 2264.9981282552085 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.1493322266390511, | |
| "joint_other_seed": 0.4302003678129185, | |
| "single_other_seed": 0.0822492566724442, | |
| "energy": 660.5603942871094 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.021182693491755278, | |
| "joint_other_seed": 0.08553377406442864, | |
| "single_other_seed": 0.19101030765684135, | |
| "energy": 24.704872767130535 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.04302398482490144, | |
| "joint_other_seed": 0.29776510711068993, | |
| "single_other_seed": 0.20053346330625577, | |
| "energy": 4.459458986918132 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.05768126823548725, | |
| "joint_other_seed": 0.23520246686521273, | |
| "single_other_seed": 0.17376045014722985, | |
| "energy": 1280.2759602864583 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.055587780726568115, | |
| "joint_other_seed": 0.3085298995123684, | |
| "single_other_seed": 0.31521473347233675, | |
| "energy": 1442.7530924479167 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.004296891391255997, | |
| "joint_other_seed": 0.08804636238563723, | |
| "single_other_seed": 0.03958595341675741, | |
| "energy": 8.681829452514648 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.10893565242901426, | |
| "joint_other_seed": 0.6002126571719206, | |
| "single_other_seed": 0.5512641641259381, | |
| "energy": 5465.91357421875 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.1943467896488249, | |
| "joint_other_seed": 0.7339822948857915, | |
| "single_other_seed": 0.4493563751423172, | |
| "energy": 1079.5183919270833 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.020632041412626534, | |
| "joint_other_seed": 0.12601713159697436, | |
| "single_other_seed": 0.1530650712676246, | |
| "energy": 31.176286697387695 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.16784505153418572, | |
| "joint_other_seed": 0.41927380094703115, | |
| "single_other_seed": 0.5926320611047866, | |
| "energy": 12.323846817016602 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.04265516722919249, | |
| "joint_other_seed": 0.10575452834075882, | |
| "single_other_seed": 0.28004040888245474, | |
| "energy": 922.4414265950521 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.04164256322564064, | |
| "joint_other_seed": 0.1608886962987799, | |
| "single_other_seed": 0.1531286417764227, | |
| "energy": 837.8826497395834 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.005636959114580953, | |
| "joint_other_seed": 0.0801947121050971, | |
| "single_other_seed": 0.05571162213978631, | |
| "energy": 11.719293594360352 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.015472195325151095, | |
| "joint_other_seed": 0.0875619494721277, | |
| "single_other_seed": 0.046613092208021234, | |
| "energy": 1887.984375 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.005086069386286751, | |
| "joint_other_seed": 0.016357315586759236, | |
| "single_other_seed": 0.008714388740887702, | |
| "energy": 136.93621317545572 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.008294717369532911, | |
| "joint_other_seed": 0.02246524381375503, | |
| "single_other_seed": 0.10173713811795682, | |
| "energy": 14.856172879536947 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.01634675714402367, | |
| "joint_other_seed": 0.07300948338750846, | |
| "single_other_seed": 0.07718146054211683, | |
| "energy": 6.623775005340576 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.03866416694260366, | |
| "joint_other_seed": 0.08142601331503174, | |
| "single_other_seed": 0.13987515849153442, | |
| "energy": 820.6852010091146 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.04068667114039415, | |
| "joint_other_seed": 0.13564239582692927, | |
| "single_other_seed": 0.2947735213465493, | |
| "energy": 780.9871419270834 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.004948177481184701, | |
| "joint_other_seed": 0.048159097527244565, | |
| "single_other_seed": 0.07003675436717427, | |
| "energy": 14.079201062520346 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.16416151956467284, | |
| "joint_other_seed": 0.3249543181373149, | |
| "single_other_seed": 0.4036489963166986, | |
| "energy": 1860.8416748046875 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.009654591097026377, | |
| "joint_other_seed": 0.04063717308813242, | |
| "single_other_seed": 0.08589367374924217, | |
| "energy": 200.22602335611978 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.009638737180587276, | |
| "joint_other_seed": 0.038055526802757145, | |
| "single_other_seed": 0.03131645248795023, | |
| "energy": 29.126463572184246 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.012446391653692102, | |
| "joint_other_seed": 0.11167049749255886, | |
| "single_other_seed": 0.0720903165224952, | |
| "energy": 8.500863711039225 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.05718110834112333, | |
| "joint_other_seed": 0.15321963683377093, | |
| "single_other_seed": 0.28726696871990964, | |
| "energy": 1008.3535359700521 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.018423503755845116, | |
| "joint_other_seed": 0.1615016108505528, | |
| "single_other_seed": 0.1490903507489325, | |
| "energy": 1015.0167439778646 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.0037758684859517944, | |
| "joint_other_seed": 0.06626306845671082, | |
| "single_other_seed": 0.06471306796361498, | |
| "energy": 23.534467697143555 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.10566646762889886, | |
| "joint_other_seed": 0.18203903191188198, | |
| "single_other_seed": 0.4346874608327894, | |
| "energy": 1569.9856363932292 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.02599151623265583, | |
| "joint_other_seed": 0.15386133355432285, | |
| "single_other_seed": 0.18605141054193586, | |
| "energy": 240.83792114257812 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.021022870002894933, | |
| "joint_other_seed": 0.1631836611125833, | |
| "single_other_seed": 0.04564375872725601, | |
| "energy": 39.84698486328125 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.07668904422016243, | |
| "joint_other_seed": 0.2381073336655124, | |
| "single_other_seed": 0.07379616173301262, | |
| "energy": 15.168261528015137 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.08151684781951618, | |
| "joint_other_seed": 0.13365664784611475, | |
| "single_other_seed": 0.4077865584515338, | |
| "energy": 841.3798014322916 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.06993978590293537, | |
| "joint_other_seed": 0.11898220365581647, | |
| "single_other_seed": 0.39646240931103316, | |
| "energy": 850.9099731445312 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.02218522842232895, | |
| "joint_other_seed": 0.12463057957732249, | |
| "single_other_seed": 0.04666607555200741, | |
| "energy": 35.725528717041016 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.06276269350113106, | |
| "joint_other_seed": 0.24865455324935415, | |
| "single_other_seed": 0.2019864700250521, | |
| "energy": 2228.9109700520835 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.03044900688582386, | |
| "joint_other_seed": 0.16269469525766445, | |
| "single_other_seed": 0.03622367276169591, | |
| "energy": 252.93541463216147 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.04603579771165367, | |
| "joint_other_seed": 0.2965099158953445, | |
| "single_other_seed": 0.15444637955424534, | |
| "energy": 54.6469357808431 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.1546306708400622, | |
| "joint_other_seed": 0.34086771752850487, | |
| "single_other_seed": 0.3204860112740461, | |
| "energy": 27.418489456176758 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.08254062034548115, | |
| "joint_other_seed": 0.1906945887912772, | |
| "single_other_seed": 0.2967651150147088, | |
| "energy": 950.6585489908854 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.06266560736131999, | |
| "joint_other_seed": 0.1396495356925288, | |
| "single_other_seed": 0.274685485429328, | |
| "energy": 993.1823120117188 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.07856823187872874, | |
| "joint_other_seed": 0.2639419515209618, | |
| "single_other_seed": 0.10415474710253214, | |
| "energy": 87.14122772216797 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.03575690070548454, | |
| "joint_other_seed": 0.1781567250221634, | |
| "single_other_seed": 0.20648462611953777, | |
| "energy": 2114.8819173177085 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.014891303062880001, | |
| "joint_other_seed": 0.0352869199641106, | |
| "single_other_seed": 0.07115237101009253, | |
| "energy": 284.39866638183594 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.17300305243247516, | |
| "joint_other_seed": 0.45909842932186795, | |
| "single_other_seed": 0.3931325246733607, | |
| "energy": 84.2755355834961 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.37563893669591475, | |
| "joint_other_seed": 0.6476381457673199, | |
| "single_other_seed": 0.6503589322572455, | |
| "energy": 84.66402180989583 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.11251627318257684, | |
| "joint_other_seed": 0.24550085611155648, | |
| "single_other_seed": 0.43973067637614727, | |
| "energy": 1025.4613850911458 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.096672946744633, | |
| "joint_other_seed": 0.4156533308978557, | |
| "single_other_seed": 0.46652912189590057, | |
| "energy": 1082.1835123697917 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.2791877207437786, | |
| "joint_other_seed": 0.5065620137735427, | |
| "single_other_seed": 0.4484410589151423, | |
| "energy": 185.51702372233072 | |
| } | |
| }, | |
| "trained/trigger/r16": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.23956529817668315, | |
| "joint_other_seed": 0.7439295206267449, | |
| "single_other_seed": 0.4640814665196763, | |
| "energy": 23567.040364583332 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.44068084170873045, | |
| "joint_other_seed": 0.8361723373471214, | |
| "single_other_seed": 0.7058980561512407, | |
| "energy": 1729.0692952473958 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.2204873042264911, | |
| "joint_other_seed": 0.7995986534368174, | |
| "single_other_seed": 0.6257225234283108, | |
| "energy": 155.14470672607422 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.352787306630282, | |
| "joint_other_seed": 0.8006213725510376, | |
| "single_other_seed": 0.7067756317147811, | |
| "energy": 2.270211696624756 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.1588222933086626, | |
| "joint_other_seed": 0.6056090646732178, | |
| "single_other_seed": 0.5279037902428164, | |
| "energy": 1987.2820638020833 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.09766755224264878, | |
| "joint_other_seed": 0.5841400758676678, | |
| "single_other_seed": 0.10972752188308754, | |
| "energy": 1813.6748860677083 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.10263743548756032, | |
| "joint_other_seed": 0.6302705158387132, | |
| "single_other_seed": 0.2146551506678699, | |
| "energy": 4.887747128804524 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.4014702821707604, | |
| "joint_other_seed": 0.6935147792875231, | |
| "single_other_seed": 0.5243573719231198, | |
| "energy": 2188.1373697916665 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.12595740957179147, | |
| "joint_other_seed": 0.7935080968080241, | |
| "single_other_seed": 0.42690857302758467, | |
| "energy": 981.8397623697916 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.2771973939171458, | |
| "joint_other_seed": 0.6445420656462676, | |
| "single_other_seed": 0.6310774005552311, | |
| "energy": 27.736914952596027 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.1652008164345784, | |
| "joint_other_seed": 0.5746290923301559, | |
| "single_other_seed": 0.5317387276711343, | |
| "energy": 1.8973779678344727 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.04595584391048246, | |
| "joint_other_seed": 0.5606741722155495, | |
| "single_other_seed": 0.10110999403717336, | |
| "energy": 1224.8262329101562 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.08004910159344936, | |
| "joint_other_seed": 0.5199599939102049, | |
| "single_other_seed": 0.24561641327817219, | |
| "energy": 892.9406331380209 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.11791951991724103, | |
| "joint_other_seed": 0.42931031701789446, | |
| "single_other_seed": 0.3126659865167023, | |
| "energy": 1.559438983599345 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.2342763891034082, | |
| "joint_other_seed": 0.7388012860945825, | |
| "single_other_seed": 0.31195102073494996, | |
| "energy": 2981.14404296875 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.12633491524719864, | |
| "joint_other_seed": 0.6105737518170843, | |
| "single_other_seed": 0.17731184302562808, | |
| "energy": 168.52887980143228 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.015220825997927956, | |
| "joint_other_seed": 0.3731195658909795, | |
| "single_other_seed": 0.3641591963136199, | |
| "energy": 21.65782101949056 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.05037207747809114, | |
| "joint_other_seed": 0.3245276707330227, | |
| "single_other_seed": 0.26231468890071385, | |
| "energy": 2.052087386449178 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.043379833371746335, | |
| "joint_other_seed": 0.5372361841472452, | |
| "single_other_seed": 0.4054313432429959, | |
| "energy": 1773.7633056640625 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.05998825430688115, | |
| "joint_other_seed": 0.5731346921951921, | |
| "single_other_seed": 0.34236778218793673, | |
| "energy": 1682.3819580078125 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.014056973386137428, | |
| "joint_other_seed": 0.24600350851761188, | |
| "single_other_seed": 0.1449921097768857, | |
| "energy": 3.065732002258301 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.1364255236694495, | |
| "joint_other_seed": 0.6907265455212401, | |
| "single_other_seed": 0.33217953036158376, | |
| "energy": 2419.9708658854165 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.07736792353598232, | |
| "joint_other_seed": 0.47501173812374187, | |
| "single_other_seed": 0.04703559916233768, | |
| "energy": 175.48029073079428 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.036159346089055244, | |
| "joint_other_seed": 0.17227221306363272, | |
| "single_other_seed": 0.11428907680284839, | |
| "energy": 12.444373766581217 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.019275795657666017, | |
| "joint_other_seed": 0.3233499798991054, | |
| "single_other_seed": 0.14680320279747328, | |
| "energy": 3.4976386229197183 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.0651266747776772, | |
| "joint_other_seed": 0.5256686631117512, | |
| "single_other_seed": 0.21309014883994815, | |
| "energy": 1967.73095703125 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.10929325402671286, | |
| "joint_other_seed": 0.4484768023674789, | |
| "single_other_seed": 0.23485787347265896, | |
| "energy": 1587.3733317057292 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.014059702908502558, | |
| "joint_other_seed": 0.24628035171376395, | |
| "single_other_seed": 0.06404832232692637, | |
| "energy": 6.339756329854329 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.09061817941568817, | |
| "joint_other_seed": 0.4877610870704132, | |
| "single_other_seed": 0.32597502466099537, | |
| "energy": 1225.5948893229167 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.06052784257561472, | |
| "joint_other_seed": 0.3769681867345729, | |
| "single_other_seed": 0.1107848176380748, | |
| "energy": 180.45411173502603 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.009950519143413383, | |
| "joint_other_seed": 0.16474049986639694, | |
| "single_other_seed": 0.01053277394584129, | |
| "energy": 19.986724853515625 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.00924478545457541, | |
| "joint_other_seed": 0.18345849271572112, | |
| "single_other_seed": 0.12247141934528094, | |
| "energy": 4.560928185780843 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.019558621514232488, | |
| "joint_other_seed": 0.41865472340018706, | |
| "single_other_seed": 0.3083956216804902, | |
| "energy": 1749.2527262369792 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.02449986492008227, | |
| "joint_other_seed": 0.4732149887243011, | |
| "single_other_seed": 0.17717077114997415, | |
| "energy": 2023.2103271484375 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.00349182661709313, | |
| "joint_other_seed": 0.20932163612106736, | |
| "single_other_seed": 0.011090593093745826, | |
| "energy": 9.837234814961752 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.06344422065902944, | |
| "joint_other_seed": 0.5707860145523422, | |
| "single_other_seed": 0.32635573441951116, | |
| "energy": 2120.2982584635415 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.1675207990159523, | |
| "joint_other_seed": 0.6431957437355751, | |
| "single_other_seed": 0.0822492566724442, | |
| "energy": 537.4306844075521 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.014587405940677645, | |
| "joint_other_seed": 0.22079170190680897, | |
| "single_other_seed": 0.19101030765684135, | |
| "energy": 27.71540896097819 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.037111360477846356, | |
| "joint_other_seed": 0.38444043484602797, | |
| "single_other_seed": 0.20053346330625577, | |
| "energy": 7.087292989095052 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.033197307921317254, | |
| "joint_other_seed": 0.40748705193158274, | |
| "single_other_seed": 0.17376045014722985, | |
| "energy": 1667.5139567057292 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.03648306521650713, | |
| "joint_other_seed": 0.4130396573691508, | |
| "single_other_seed": 0.31521473347233675, | |
| "energy": 1812.8243815104167 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.005999124997113747, | |
| "joint_other_seed": 0.20372671074470405, | |
| "single_other_seed": 0.03958595341675741, | |
| "energy": 13.753936449686686 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.11842236000278994, | |
| "joint_other_seed": 0.7995976178880961, | |
| "single_other_seed": 0.5512641641259381, | |
| "energy": 4716.045084635417 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.20718158894096295, | |
| "joint_other_seed": 0.8090572103175722, | |
| "single_other_seed": 0.4493563751423172, | |
| "energy": 1051.5061848958333 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.018562734131589487, | |
| "joint_other_seed": 0.24491853885036508, | |
| "single_other_seed": 0.1530650712676246, | |
| "energy": 32.068963368733726 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.1806915391971533, | |
| "joint_other_seed": 0.6300511779060085, | |
| "single_other_seed": 0.5926320611047866, | |
| "energy": 17.441326141357422 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.02042106615182504, | |
| "joint_other_seed": 0.32122164274674525, | |
| "single_other_seed": 0.28004040888245474, | |
| "energy": 1720.2058919270833 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.019955826376333974, | |
| "joint_other_seed": 0.2587315911898759, | |
| "single_other_seed": 0.1531286417764227, | |
| "energy": 1552.8030192057292 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.005631972360255255, | |
| "joint_other_seed": 0.18807960800017653, | |
| "single_other_seed": 0.05571162213978631, | |
| "energy": 19.020296732584637 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.022810635739770824, | |
| "joint_other_seed": 0.2640173342508154, | |
| "single_other_seed": 0.046613092208021234, | |
| "energy": 1886.2245279947917 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.004547864199530793, | |
| "joint_other_seed": 0.06746355040340235, | |
| "single_other_seed": 0.008714388740887702, | |
| "energy": 106.59467061360677 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.013773086258906386, | |
| "joint_other_seed": 0.14685144961735863, | |
| "single_other_seed": 0.10173713811795682, | |
| "energy": 21.311541239420574 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.01200618477715043, | |
| "joint_other_seed": 0.17185652070532964, | |
| "single_other_seed": 0.07718146054211683, | |
| "energy": 12.723722775777182 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.01759516820790874, | |
| "joint_other_seed": 0.24738354799838524, | |
| "single_other_seed": 0.13987515849153442, | |
| "energy": 1554.455810546875 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.021127990836844135, | |
| "joint_other_seed": 0.2652995294836137, | |
| "single_other_seed": 0.2947735213465493, | |
| "energy": 1449.6674397786458 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.006220296643583256, | |
| "joint_other_seed": 0.13857532384111648, | |
| "single_other_seed": 0.07003675436717427, | |
| "energy": 25.713788986206055 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.1416923433231546, | |
| "joint_other_seed": 0.42080148960408714, | |
| "single_other_seed": 0.4036489963166986, | |
| "energy": 1771.1400960286458 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.012699936554469993, | |
| "joint_other_seed": 0.20940170221189225, | |
| "single_other_seed": 0.08589367374924217, | |
| "energy": 193.48401387532553 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.012628646398129328, | |
| "joint_other_seed": 0.17514657294913438, | |
| "single_other_seed": 0.03131645248795023, | |
| "energy": 33.39300282796224 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.011944060254155953, | |
| "joint_other_seed": 0.20715810636823517, | |
| "single_other_seed": 0.0720903165224952, | |
| "energy": 18.093185424804688 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.027507549274330523, | |
| "joint_other_seed": 0.31998372278701503, | |
| "single_other_seed": 0.28726696871990964, | |
| "energy": 1799.4704996744792 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.01452879451061384, | |
| "joint_other_seed": 0.27238862454919854, | |
| "single_other_seed": 0.1490903507489325, | |
| "energy": 1613.044189453125 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.00470982271015663, | |
| "joint_other_seed": 0.18779231338304458, | |
| "single_other_seed": 0.06471306796361498, | |
| "energy": 40.46242141723633 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.1324734769378162, | |
| "joint_other_seed": 0.4184789605927691, | |
| "single_other_seed": 0.4346874608327894, | |
| "energy": 1511.3125 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.029794257248565326, | |
| "joint_other_seed": 0.3142581187602105, | |
| "single_other_seed": 0.18605141054193586, | |
| "energy": 209.24992879231772 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.01880634618978627, | |
| "joint_other_seed": 0.24540136579025118, | |
| "single_other_seed": 0.04564375872725601, | |
| "energy": 38.44870630900065 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.07080815509040983, | |
| "joint_other_seed": 0.36956688317763686, | |
| "single_other_seed": 0.07379616173301262, | |
| "energy": 26.468076705932617 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.044151836019340474, | |
| "joint_other_seed": 0.3009688119378603, | |
| "single_other_seed": 0.4077865584515338, | |
| "energy": 1571.4049072265625 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.03382527744076247, | |
| "joint_other_seed": 0.27485670646779226, | |
| "single_other_seed": 0.39646240931103316, | |
| "energy": 1386.3585205078125 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.024191368783886724, | |
| "joint_other_seed": 0.26940686749097054, | |
| "single_other_seed": 0.04666607555200741, | |
| "energy": 56.6503054300944 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.061972548169708874, | |
| "joint_other_seed": 0.4853224173688978, | |
| "single_other_seed": 0.2019864700250521, | |
| "energy": 2325.8543294270835 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.03265892402728854, | |
| "joint_other_seed": 0.2435901343268685, | |
| "single_other_seed": 0.03622367276169591, | |
| "energy": 250.16190083821616 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.04225602012331664, | |
| "joint_other_seed": 0.40343807534531173, | |
| "single_other_seed": 0.15444637955424534, | |
| "energy": 56.99618657430013 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.14601801945554915, | |
| "joint_other_seed": 0.5190473407445564, | |
| "single_other_seed": 0.3204860112740461, | |
| "energy": 36.497580210367836 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.044591496300785405, | |
| "joint_other_seed": 0.3386625168862746, | |
| "single_other_seed": 0.2967651150147088, | |
| "energy": 1611.1257731119792 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.030403904536046467, | |
| "joint_other_seed": 0.3227014232705727, | |
| "single_other_seed": 0.274685485429328, | |
| "energy": 1578.5759684244792 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.07136406016186025, | |
| "joint_other_seed": 0.4515262698460585, | |
| "single_other_seed": 0.10415474710253214, | |
| "energy": 98.33413950602214 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.045305809007151865, | |
| "joint_other_seed": 0.3423868832982909, | |
| "single_other_seed": 0.20648462611953777, | |
| "energy": 2184.5830891927085 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.026958511292807796, | |
| "joint_other_seed": 0.21982296117590927, | |
| "single_other_seed": 0.07115237101009253, | |
| "energy": 262.53094482421875 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.21183526439534694, | |
| "joint_other_seed": 0.6468190148049491, | |
| "single_other_seed": 0.3931325246733607, | |
| "energy": 83.57801564534505 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.42807649267749104, | |
| "joint_other_seed": 0.7485589519716371, | |
| "single_other_seed": 0.6503589322572455, | |
| "energy": 92.94496154785156 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.07938948548305269, | |
| "joint_other_seed": 0.461746241686557, | |
| "single_other_seed": 0.43973067637614727, | |
| "energy": 1699.9314778645833 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.09306391953688453, | |
| "joint_other_seed": 0.4617417354295261, | |
| "single_other_seed": 0.46652912189590057, | |
| "energy": 1457.084228515625 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.23480331871857396, | |
| "joint_other_seed": 0.6436824936880499, | |
| "single_other_seed": 0.4484410589151423, | |
| "energy": 209.22176106770834 | |
| } | |
| }, | |
| "trained/all/r1": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.1532229316810666, | |
| "joint_other_seed": 0.24173209681836483, | |
| "single_other_seed": 0.44811679067370147, | |
| "energy": 39861.90625 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.29187843693716753, | |
| "joint_other_seed": 0.5501560906444152, | |
| "single_other_seed": 0.6774332899431854, | |
| "energy": 5260.834147135417 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.16147795964630085, | |
| "joint_other_seed": 0.5206910074398361, | |
| "single_other_seed": 0.5934244918414519, | |
| "energy": 231.35144551595053 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.3113501828026284, | |
| "joint_other_seed": 0.6193379697947259, | |
| "single_other_seed": 0.6616902130971496, | |
| "energy": 4.03853702545166 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.12387680568320181, | |
| "joint_other_seed": 0.1988931774768935, | |
| "single_other_seed": 0.5044815243219982, | |
| "energy": 4038.4915364583335 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.08419514051233123, | |
| "joint_other_seed": 0.3237029303955925, | |
| "single_other_seed": 0.07558693602861832, | |
| "energy": 3493.6555989583335 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.050327160631030154, | |
| "joint_other_seed": 0.1428950988705634, | |
| "single_other_seed": 0.25326322743638896, | |
| "energy": 12.800386746724447 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.3607413232113291, | |
| "joint_other_seed": 0.651907595112061, | |
| "single_other_seed": 0.5592731941078808, | |
| "energy": 7315.093098958333 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.07590275309963801, | |
| "joint_other_seed": 0.119960598186301, | |
| "single_other_seed": 0.3976547874967953, | |
| "energy": 1367.0419514973958 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.2835387932741069, | |
| "joint_other_seed": 0.43967985148361755, | |
| "single_other_seed": 0.5264533063317531, | |
| "energy": 51.64264933268229 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.24207885284443476, | |
| "joint_other_seed": 0.44174754186098175, | |
| "single_other_seed": 0.5249094087397959, | |
| "energy": 2.5362467765808105 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.08220788873573581, | |
| "joint_other_seed": 0.20974588694736074, | |
| "single_other_seed": 0.10463864309003279, | |
| "energy": 769.1956176757812 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.15528909088541965, | |
| "joint_other_seed": 0.42726925837593843, | |
| "single_other_seed": 0.2413112291362709, | |
| "energy": 993.876220703125 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.10120309594307197, | |
| "joint_other_seed": 0.31998133522551986, | |
| "single_other_seed": 0.32670545615153346, | |
| "energy": 2.442079544067383 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.21762443241608576, | |
| "joint_other_seed": 0.3390843660705197, | |
| "single_other_seed": 0.3055934700224835, | |
| "energy": 5806.13916015625 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.07875495192860626, | |
| "joint_other_seed": 0.11134294773390996, | |
| "single_other_seed": 0.15272817951550088, | |
| "energy": 640.0982767740885 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.05000209350577167, | |
| "joint_other_seed": 0.20609369802734778, | |
| "single_other_seed": 0.36210041857182396, | |
| "energy": 17.963436126708984 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.100716523203635, | |
| "joint_other_seed": 0.12555034966527423, | |
| "single_other_seed": 0.26899980793177847, | |
| "energy": 1.7883254488309224 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.11108977574923995, | |
| "joint_other_seed": 0.20181464396292526, | |
| "single_other_seed": 0.38398963494243255, | |
| "energy": 1801.3724772135417 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.1554991879394464, | |
| "joint_other_seed": 0.39260292981300726, | |
| "single_other_seed": 0.3167369804690122, | |
| "energy": 1761.218994140625 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.0223601739110304, | |
| "joint_other_seed": 0.17450977990387562, | |
| "single_other_seed": 0.14961398668920167, | |
| "energy": 3.419227441151937 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.03830031334644568, | |
| "joint_other_seed": 0.11771072148806949, | |
| "single_other_seed": 0.2929816260751952, | |
| "energy": 5136.373942057292 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.05270271307586105, | |
| "joint_other_seed": 0.06645910208933492, | |
| "single_other_seed": 0.0393132383309347, | |
| "energy": 463.82497151692706 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.052535951697118555, | |
| "joint_other_seed": 0.17663348857338812, | |
| "single_other_seed": 0.12592987452065718, | |
| "energy": 18.66283925374349 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.0317949201963722, | |
| "joint_other_seed": 0.01944380618340391, | |
| "single_other_seed": 0.14692226186879032, | |
| "energy": 2.968983014424642 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.0815330746365488, | |
| "joint_other_seed": 0.07170404050172641, | |
| "single_other_seed": 0.1889701344004625, | |
| "energy": 1651.6423950195312 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.05500375825127524, | |
| "joint_other_seed": 0.1274242803834129, | |
| "single_other_seed": 0.21109406152264532, | |
| "energy": 1386.8681233723958 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.011313975350634609, | |
| "joint_other_seed": 0.0593350288034008, | |
| "single_other_seed": 0.0732304776958198, | |
| "energy": 5.468466758728027 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.12149351367330152, | |
| "joint_other_seed": 0.45939952159922925, | |
| "single_other_seed": 0.31437607038787596, | |
| "energy": 1521.7087809244792 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.039547889213428, | |
| "joint_other_seed": 0.1276870106859357, | |
| "single_other_seed": 0.11663940224099856, | |
| "energy": 339.7416636149089 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.0073744321096170795, | |
| "joint_other_seed": 0.014811597925167175, | |
| "single_other_seed": 0.010774844085124191, | |
| "energy": 17.02111053466797 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.027131266020042122, | |
| "joint_other_seed": 0.011308472762817656, | |
| "single_other_seed": 0.1205982594903281, | |
| "energy": 2.8582187493642173 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.08004968823361325, | |
| "joint_other_seed": 0.1007608995383259, | |
| "single_other_seed": 0.27611364685869666, | |
| "energy": 2020.39404296875 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.07222624768110786, | |
| "joint_other_seed": 0.092180236230333, | |
| "single_other_seed": 0.1700221941926361, | |
| "energy": 2232.225341796875 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.0014220265177348201, | |
| "joint_other_seed": 0.0066749289888459185, | |
| "single_other_seed": 0.00984234044827248, | |
| "energy": 7.182994206746419 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.05253536809118242, | |
| "joint_other_seed": 0.3119700814494194, | |
| "single_other_seed": 0.31249733395227464, | |
| "energy": 6641.9912109375 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.10200475622331573, | |
| "joint_other_seed": 0.2216031300316864, | |
| "single_other_seed": 0.10002790597170608, | |
| "energy": 1341.9746500651042 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.03260599387452401, | |
| "joint_other_seed": 0.1640302845732784, | |
| "single_other_seed": 0.20930823713398128, | |
| "energy": 46.52178192138672 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.04362450205359325, | |
| "joint_other_seed": 0.1743919905115302, | |
| "single_other_seed": 0.17358361082812723, | |
| "energy": 6.241325855255127 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.0994644108931968, | |
| "joint_other_seed": 0.21211778193823438, | |
| "single_other_seed": 0.16639605862354157, | |
| "energy": 1135.3727213541667 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.11387102263864612, | |
| "joint_other_seed": 0.4145166533573672, | |
| "single_other_seed": 0.26266388980276245, | |
| "energy": 1994.8168538411458 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.002358597535624951, | |
| "joint_other_seed": 0.01301688871428952, | |
| "single_other_seed": 0.041263623481190614, | |
| "energy": 12.034358660380045 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.053948601133005036, | |
| "joint_other_seed": 0.20358720659374252, | |
| "single_other_seed": 0.5409156305566543, | |
| "energy": 17523.400716145832 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.12462141518382719, | |
| "joint_other_seed": 0.43550322692067045, | |
| "single_other_seed": 0.4127825079777623, | |
| "energy": 2060.1400553385415 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.01385037701035785, | |
| "joint_other_seed": 0.004817349319983677, | |
| "single_other_seed": 0.13229671573213586, | |
| "energy": 60.51694107055664 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.030900762147843097, | |
| "joint_other_seed": 0.0180963324234813, | |
| "single_other_seed": 0.5243198653941963, | |
| "energy": 10.989652633666992 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.08950248326659238, | |
| "joint_other_seed": 0.11724101346719362, | |
| "single_other_seed": 0.2561361224819235, | |
| "energy": 782.7784830729166 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.062263370121286354, | |
| "joint_other_seed": 0.23388370935970793, | |
| "single_other_seed": 0.1416972722214367, | |
| "energy": 849.0794270833334 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.0018017840096553606, | |
| "joint_other_seed": 0.000841272888470912, | |
| "single_other_seed": 0.05344201643102686, | |
| "energy": 14.341851234436035 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.008236775029841104, | |
| "joint_other_seed": 0.14768772276759345, | |
| "single_other_seed": 0.03879788879264867, | |
| "energy": 4885.623372395833 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.0012610186513136723, | |
| "joint_other_seed": 0.0005281861125400189, | |
| "single_other_seed": 0.007674959365953304, | |
| "energy": 310.5619812011719 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.00577397115623679, | |
| "joint_other_seed": 0.04936789551777243, | |
| "single_other_seed": 0.10182615658890926, | |
| "energy": 23.367394765218098 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.016537966811607907, | |
| "joint_other_seed": 0.12469832469715525, | |
| "single_other_seed": 0.07743756724573647, | |
| "energy": 11.063886006673178 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.0457688744121032, | |
| "joint_other_seed": 0.03328871676332701, | |
| "single_other_seed": 0.12826154619762153, | |
| "energy": 951.8951416015625 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.06010015809154978, | |
| "joint_other_seed": 0.07931015899452054, | |
| "single_other_seed": 0.2710579694821311, | |
| "energy": 709.7340901692709 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.00823912176698438, | |
| "joint_other_seed": 0.046731286210325296, | |
| "single_other_seed": 0.06917540793465327, | |
| "energy": 20.793126424153645 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.20426877343064076, | |
| "joint_other_seed": 0.47380172828209965, | |
| "single_other_seed": 0.40000490722120674, | |
| "energy": 5094.351155598958 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.011759645098414317, | |
| "joint_other_seed": 0.04520789222125987, | |
| "single_other_seed": 0.08777484586826956, | |
| "energy": 490.17311604817706 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.0023657335083341335, | |
| "joint_other_seed": 0.0135118413097155, | |
| "single_other_seed": 0.030195208681685792, | |
| "energy": 35.128004709879555 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.010474788221236375, | |
| "joint_other_seed": 0.012766874986773868, | |
| "single_other_seed": 0.0725305279306323, | |
| "energy": 10.004602432250977 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.12191608610507687, | |
| "joint_other_seed": 0.37853892283363355, | |
| "single_other_seed": 0.28307664810102695, | |
| "energy": 968.9420166015625 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.041244590538082826, | |
| "joint_other_seed": 0.13788839190713964, | |
| "single_other_seed": 0.1517876894030512, | |
| "energy": 1396.6142578125 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.008361783733775163, | |
| "joint_other_seed": 0.05681230166591, | |
| "single_other_seed": 0.0642206656561638, | |
| "energy": 33.3339958190918 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.19362889783697146, | |
| "joint_other_seed": 0.3020562780242645, | |
| "single_other_seed": 0.4231721111403465, | |
| "energy": 4476.537109375 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.0346836764012437, | |
| "joint_other_seed": 0.26826154560332105, | |
| "single_other_seed": 0.17922029899926128, | |
| "energy": 485.92859903971356 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.016793669173569614, | |
| "joint_other_seed": 0.11239858716348727, | |
| "single_other_seed": 0.04469842287671, | |
| "energy": 64.57457478841145 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.057413276595798646, | |
| "joint_other_seed": 0.07142772506655592, | |
| "single_other_seed": 0.0762138161992767, | |
| "energy": 30.101455052693684 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.137906085904001, | |
| "joint_other_seed": 0.35607099913907225, | |
| "single_other_seed": 0.40287580669562406, | |
| "energy": 980.5436808268229 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.14676167992746525, | |
| "joint_other_seed": 0.4231113708495231, | |
| "single_other_seed": 0.39129970592240937, | |
| "energy": 1024.4737955729167 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.011672650426000075, | |
| "joint_other_seed": 0.07227114559104553, | |
| "single_other_seed": 0.04747030511104855, | |
| "energy": 52.47013473510742 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.023210088788507832, | |
| "joint_other_seed": 0.08194282961499233, | |
| "single_other_seed": 0.1978132273597456, | |
| "energy": 4277.32177734375 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.04394387643448361, | |
| "joint_other_seed": 0.1454025435989068, | |
| "single_other_seed": 0.0354728814912888, | |
| "energy": 676.4292805989584 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.04999282195609811, | |
| "joint_other_seed": 0.21124646732355734, | |
| "single_other_seed": 0.1514474531991653, | |
| "energy": 119.04248555501302 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.1343576089276627, | |
| "joint_other_seed": 0.21090794619169373, | |
| "single_other_seed": 0.3181661841830635, | |
| "energy": 50.07338269551595 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.1128255134554298, | |
| "joint_other_seed": 0.39902003802018027, | |
| "single_other_seed": 0.29169242240146276, | |
| "energy": 1234.6209513346355 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.1090290649290443, | |
| "joint_other_seed": 0.1585697357098068, | |
| "single_other_seed": 0.27255315041742323, | |
| "energy": 1263.3814697265625 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.026699118762772867, | |
| "joint_other_seed": 0.08589073033368155, | |
| "single_other_seed": 0.10362526313668152, | |
| "energy": 125.33080037434895 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.032628907577550104, | |
| "joint_other_seed": 0.20146211458529195, | |
| "single_other_seed": 0.2042873631030948, | |
| "energy": 6968.204752604167 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.013307019815613996, | |
| "joint_other_seed": 0.044308992734428565, | |
| "single_other_seed": 0.07144907017433422, | |
| "energy": 683.8508097330729 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.1985794295817328, | |
| "joint_other_seed": 0.4811443217673078, | |
| "single_other_seed": 0.3912756424245397, | |
| "energy": 225.73617553710938 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.3525198638900877, | |
| "joint_other_seed": 0.6398612987527702, | |
| "single_other_seed": 0.65317524746437, | |
| "energy": 208.31158447265625 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.13206184175311364, | |
| "joint_other_seed": 0.29983453629370993, | |
| "single_other_seed": 0.4343533199384628, | |
| "energy": 1208.9947916666667 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.07877971354093638, | |
| "joint_other_seed": 0.08410525254240864, | |
| "single_other_seed": 0.4655687911571512, | |
| "energy": 1635.3168538411458 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.30259603706477023, | |
| "joint_other_seed": 0.4360328466531883, | |
| "single_other_seed": 0.44770980648376535, | |
| "energy": 417.6214904785156 | |
| } | |
| }, | |
| "trained/all/r3": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.20493758876546453, | |
| "joint_other_seed": 0.5922433360435454, | |
| "single_other_seed": 0.44811679067370147, | |
| "energy": 34247.076822916664 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.3568185253207169, | |
| "joint_other_seed": 0.7513176479666662, | |
| "single_other_seed": 0.6774332899431854, | |
| "energy": 2931.798095703125 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.17868203665640758, | |
| "joint_other_seed": 0.7347692585603147, | |
| "single_other_seed": 0.5934244918414519, | |
| "energy": 218.74850463867188 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.31132536679199196, | |
| "joint_other_seed": 0.6443164706447893, | |
| "single_other_seed": 0.6616902130971496, | |
| "energy": 3.822299003601074 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.12887105462810844, | |
| "joint_other_seed": 0.398353445259498, | |
| "single_other_seed": 0.5044815243219982, | |
| "energy": 3746.671630859375 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.0698099857706379, | |
| "joint_other_seed": 0.2461752499482096, | |
| "single_other_seed": 0.07558693602861832, | |
| "energy": 2238.37060546875 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.13871963914051477, | |
| "joint_other_seed": 0.5171645056679663, | |
| "single_other_seed": 0.25326322743638896, | |
| "energy": 12.871987660725912 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.36518199991724587, | |
| "joint_other_seed": 0.6339060357154043, | |
| "single_other_seed": 0.5592731941078808, | |
| "energy": 5825.649576822917 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.07372235013895767, | |
| "joint_other_seed": 0.5898421752753964, | |
| "single_other_seed": 0.3976547874967953, | |
| "energy": 1888.3025716145833 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.2030457279715827, | |
| "joint_other_seed": 0.43371361843650136, | |
| "single_other_seed": 0.5264533063317531, | |
| "energy": 58.73718134562174 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.1572896237358251, | |
| "joint_other_seed": 0.3828382351261681, | |
| "single_other_seed": 0.5249094087397959, | |
| "energy": 3.1801485220591226 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.061282416582584265, | |
| "joint_other_seed": 0.09992223818690182, | |
| "single_other_seed": 0.10463864309003279, | |
| "energy": 1587.536376953125 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.08506991833901267, | |
| "joint_other_seed": 0.3787538912815547, | |
| "single_other_seed": 0.2413112291362709, | |
| "energy": 2132.7166748046875 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.11669171664498024, | |
| "joint_other_seed": 0.25146581511174293, | |
| "single_other_seed": 0.32670545615153346, | |
| "energy": 4.056080420811971 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.21358361142516225, | |
| "joint_other_seed": 0.48000159412906807, | |
| "single_other_seed": 0.3055934700224835, | |
| "energy": 6313.857747395833 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.08923211442019208, | |
| "joint_other_seed": 0.4830050767377442, | |
| "single_other_seed": 0.15272817951550088, | |
| "energy": 556.9349975585938 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.023728565164232718, | |
| "joint_other_seed": 0.17894946463227102, | |
| "single_other_seed": 0.36210041857182396, | |
| "energy": 42.913893381754555 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.06738211602790589, | |
| "joint_other_seed": 0.15864931441587338, | |
| "single_other_seed": 0.26899980793177847, | |
| "energy": 3.10943603515625 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.046233442858023555, | |
| "joint_other_seed": 0.410227881188164, | |
| "single_other_seed": 0.38398963494243255, | |
| "energy": 2851.2565104166665 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.09020031100193182, | |
| "joint_other_seed": 0.338935701086454, | |
| "single_other_seed": 0.3167369804690122, | |
| "energy": 2576.83203125 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.01249026185733837, | |
| "joint_other_seed": 0.09569057891073351, | |
| "single_other_seed": 0.14961398668920167, | |
| "energy": 6.588569164276123 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.1065386647796922, | |
| "joint_other_seed": 0.43609347233220586, | |
| "single_other_seed": 0.2929816260751952, | |
| "energy": 5094.317464192708 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.06520152956153424, | |
| "joint_other_seed": 0.15385421286100154, | |
| "single_other_seed": 0.0393132383309347, | |
| "energy": 489.5119222005208 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.04208609996583634, | |
| "joint_other_seed": 0.11811135868756018, | |
| "single_other_seed": 0.12592987452065718, | |
| "energy": 35.20250701904297 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.025755359062081706, | |
| "joint_other_seed": 0.1655239197853241, | |
| "single_other_seed": 0.14692226186879032, | |
| "energy": 5.755976517995198 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.05040963742242255, | |
| "joint_other_seed": 0.19830998245628806, | |
| "single_other_seed": 0.1889701344004625, | |
| "energy": 2377.15380859375 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.08035399364419504, | |
| "joint_other_seed": 0.2690612515480758, | |
| "single_other_seed": 0.21109406152264532, | |
| "energy": 2234.8234049479165 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.01747161291204602, | |
| "joint_other_seed": 0.09490253877533328, | |
| "single_other_seed": 0.0732304776958198, | |
| "energy": 12.951597531636557 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.0877830571469248, | |
| "joint_other_seed": 0.3218064846773896, | |
| "single_other_seed": 0.31437607038787596, | |
| "energy": 2963.8301595052085 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.045301789932716274, | |
| "joint_other_seed": 0.20183038081134141, | |
| "single_other_seed": 0.11663940224099856, | |
| "energy": 464.7893880208333 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.010761082002549882, | |
| "joint_other_seed": 0.025273542479149147, | |
| "single_other_seed": 0.010774844085124191, | |
| "energy": 35.594825744628906 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.020726792797737592, | |
| "joint_other_seed": 0.06072340259019593, | |
| "single_other_seed": 0.1205982594903281, | |
| "energy": 4.845906734466553 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.03816978278032401, | |
| "joint_other_seed": 0.14955345615275994, | |
| "single_other_seed": 0.27611364685869666, | |
| "energy": 2281.7662760416665 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.03250744383813009, | |
| "joint_other_seed": 0.263693117721021, | |
| "single_other_seed": 0.1700221941926361, | |
| "energy": 3143.2203776041665 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.003889064932446229, | |
| "joint_other_seed": 0.08276532023178906, | |
| "single_other_seed": 0.00984234044827248, | |
| "energy": 16.17855676015218 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.0535502384912907, | |
| "joint_other_seed": 0.44906145257438856, | |
| "single_other_seed": 0.31249733395227464, | |
| "energy": 5838.910481770833 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.12590634484704744, | |
| "joint_other_seed": 0.3789476942070156, | |
| "single_other_seed": 0.10002790597170608, | |
| "energy": 1377.4964803059895 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.01603903371906179, | |
| "joint_other_seed": 0.08157581419150861, | |
| "single_other_seed": 0.20930823713398128, | |
| "energy": 59.644466400146484 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.0315973053460753, | |
| "joint_other_seed": 0.2684046062983094, | |
| "single_other_seed": 0.17358361082812723, | |
| "energy": 10.80898920694987 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.04987715772322143, | |
| "joint_other_seed": 0.19636313902672498, | |
| "single_other_seed": 0.16639605862354157, | |
| "energy": 2479.6190592447915 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.04302161819175254, | |
| "joint_other_seed": 0.27544482510144863, | |
| "single_other_seed": 0.26266388980276245, | |
| "energy": 2969.455322265625 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.004587764178095033, | |
| "joint_other_seed": 0.09020841364533205, | |
| "single_other_seed": 0.041263623481190614, | |
| "energy": 22.997177124023438 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.10946824108752262, | |
| "joint_other_seed": 0.6018268867107877, | |
| "single_other_seed": 0.5409156305566543, | |
| "energy": 14387.345052083334 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.14025174882232494, | |
| "joint_other_seed": 0.6878820501427926, | |
| "single_other_seed": 0.4127825079777623, | |
| "energy": 2122.4639485677085 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.012929119336335337, | |
| "joint_other_seed": 0.10539420279435241, | |
| "single_other_seed": 0.13229671573213586, | |
| "energy": 73.1492207845052 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.13329897825973458, | |
| "joint_other_seed": 0.37422351979669854, | |
| "single_other_seed": 0.5243198653941963, | |
| "energy": 20.738697687784832 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.03613967690454678, | |
| "joint_other_seed": 0.09832705285925702, | |
| "single_other_seed": 0.2561361224819235, | |
| "energy": 2108.285847981771 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.038264194799290885, | |
| "joint_other_seed": 0.1539541311315048, | |
| "single_other_seed": 0.1416972722214367, | |
| "energy": 2042.7308756510417 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.004868399917688504, | |
| "joint_other_seed": 0.07922934939669757, | |
| "single_other_seed": 0.05344201643102686, | |
| "energy": 30.442265192667644 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.0134232719222791, | |
| "joint_other_seed": 0.08216534965981126, | |
| "single_other_seed": 0.03879788879264867, | |
| "energy": 4845.46484375 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.005165527247514179, | |
| "joint_other_seed": 0.01624354588074195, | |
| "single_other_seed": 0.007674959365953304, | |
| "energy": 369.19054158528644 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.008528806843184467, | |
| "joint_other_seed": 0.02286604017621663, | |
| "single_other_seed": 0.10182615658890926, | |
| "energy": 39.27280044555664 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.015437231563047549, | |
| "joint_other_seed": 0.0687119883282108, | |
| "single_other_seed": 0.07743756724573647, | |
| "energy": 17.305023829142254 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.0359656940333859, | |
| "joint_other_seed": 0.07757832161781875, | |
| "single_other_seed": 0.12826154619762153, | |
| "energy": 2106.0096842447915 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.034587657906638536, | |
| "joint_other_seed": 0.13405610782007538, | |
| "single_other_seed": 0.2710579694821311, | |
| "energy": 1940.6547037760417 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.005374388356733983, | |
| "joint_other_seed": 0.04962735710159227, | |
| "single_other_seed": 0.06917540793465327, | |
| "energy": 37.5213737487793 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.15791589222815716, | |
| "joint_other_seed": 0.31573499867854826, | |
| "single_other_seed": 0.40000490722120674, | |
| "energy": 4720.409342447917 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.009454287604582445, | |
| "joint_other_seed": 0.03965217306221869, | |
| "single_other_seed": 0.08777484586826956, | |
| "energy": 530.2259826660156 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.00851979163414951, | |
| "joint_other_seed": 0.03466836049298527, | |
| "single_other_seed": 0.030195208681685792, | |
| "energy": 76.25372314453125 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.012732466210383264, | |
| "joint_other_seed": 0.11255944323974099, | |
| "single_other_seed": 0.0725305279306323, | |
| "energy": 22.838110605875652 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.054497679879154326, | |
| "joint_other_seed": 0.15116616176024888, | |
| "single_other_seed": 0.28307664810102695, | |
| "energy": 2598.927490234375 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.017180691618906896, | |
| "joint_other_seed": 0.1564319616439932, | |
| "single_other_seed": 0.1517876894030512, | |
| "energy": 2579.8938802083335 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.00383356593462425, | |
| "joint_other_seed": 0.06537614132194847, | |
| "single_other_seed": 0.0642206656561638, | |
| "energy": 62.55110295613607 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.10093376225029549, | |
| "joint_other_seed": 0.1795571091355791, | |
| "single_other_seed": 0.4231721111403465, | |
| "energy": 4122.190104166667 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.025499762262757963, | |
| "joint_other_seed": 0.15635148847335767, | |
| "single_other_seed": 0.17922029899926128, | |
| "energy": 639.673350016276 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.020697568930068244, | |
| "joint_other_seed": 0.16216048261097882, | |
| "single_other_seed": 0.04469842287671, | |
| "energy": 104.88719177246094 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.07665580124374505, | |
| "joint_other_seed": 0.23631443093507387, | |
| "single_other_seed": 0.0762138161992767, | |
| "energy": 40.19543965657552 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.079116640572234, | |
| "joint_other_seed": 0.1310640297970297, | |
| "single_other_seed": 0.40287580669562406, | |
| "energy": 2190.405517578125 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.06684283933803492, | |
| "joint_other_seed": 0.11394081032029213, | |
| "single_other_seed": 0.39129970592240937, | |
| "energy": 2236.2395833333335 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.021940486944835857, | |
| "joint_other_seed": 0.1252539298096309, | |
| "single_other_seed": 0.04747030511104855, | |
| "energy": 94.62638600667317 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.05932250361413993, | |
| "joint_other_seed": 0.2436909568847807, | |
| "single_other_seed": 0.1978132273597456, | |
| "energy": 5843.65625 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.03043526154291898, | |
| "joint_other_seed": 0.1639837924342915, | |
| "single_other_seed": 0.0354728814912888, | |
| "energy": 676.35302734375 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.04485744422945237, | |
| "joint_other_seed": 0.29172619934060073, | |
| "single_other_seed": 0.1514474531991653, | |
| "energy": 144.1634063720703 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.14958475494196122, | |
| "joint_other_seed": 0.3339670074205132, | |
| "single_other_seed": 0.3181661841830635, | |
| "energy": 70.99815622965495 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.07813248064935441, | |
| "joint_other_seed": 0.1819036621678456, | |
| "single_other_seed": 0.29169242240146276, | |
| "energy": 2448.826456705729 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.058825691734890495, | |
| "joint_other_seed": 0.12932386701516274, | |
| "single_other_seed": 0.27255315041742323, | |
| "energy": 2572.5001627604165 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.07799667151288381, | |
| "joint_other_seed": 0.2642130841674992, | |
| "single_other_seed": 0.10362526313668152, | |
| "energy": 230.16250610351562 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.0357791145377061, | |
| "joint_other_seed": 0.1797613020509935, | |
| "single_other_seed": 0.2042873631030948, | |
| "energy": 5672.9208984375 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.014867296353960557, | |
| "joint_other_seed": 0.03577550283284736, | |
| "single_other_seed": 0.07144907017433422, | |
| "energy": 758.6786499023438 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.16998670338042077, | |
| "joint_other_seed": 0.4596181717835577, | |
| "single_other_seed": 0.3912756424245397, | |
| "energy": 224.1661376953125 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.3924236629943536, | |
| "joint_other_seed": 0.6554453334163672, | |
| "single_other_seed": 0.65317524746437, | |
| "energy": 231.95325215657553 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.10398103634114853, | |
| "joint_other_seed": 0.22875650968650152, | |
| "single_other_seed": 0.4343533199384628, | |
| "energy": 2590.44677734375 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.09758399837035969, | |
| "joint_other_seed": 0.41354622078658676, | |
| "single_other_seed": 0.4655687911571512, | |
| "energy": 2852.431640625 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.27729916261911713, | |
| "joint_other_seed": 0.5037805869355674, | |
| "single_other_seed": 0.44770980648376535, | |
| "energy": 494.2400716145833 | |
| } | |
| }, | |
| "trained/all/r16": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.23406707293590134, | |
| "joint_other_seed": 0.7324826461294813, | |
| "single_other_seed": 0.44811679067370147, | |
| "energy": 26929.387369791668 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.3867894560975088, | |
| "joint_other_seed": 0.7770775003254051, | |
| "single_other_seed": 0.6774332899431854, | |
| "energy": 2094.49462890625 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.20723516488958982, | |
| "joint_other_seed": 0.7564579480786008, | |
| "single_other_seed": 0.5934244918414519, | |
| "energy": 210.3242390950521 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.30677165769501713, | |
| "joint_other_seed": 0.7740168341302718, | |
| "single_other_seed": 0.6616902130971496, | |
| "energy": 4.16606593132019 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.10997016121567016, | |
| "joint_other_seed": 0.5262380453445952, | |
| "single_other_seed": 0.5044815243219982, | |
| "energy": 2929.2613932291665 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.07431695058630916, | |
| "joint_other_seed": 0.49644246615817067, | |
| "single_other_seed": 0.07558693602861832, | |
| "energy": 2762.851806640625 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.15051901944152712, | |
| "joint_other_seed": 0.6216857206845152, | |
| "single_other_seed": 0.25326322743638896, | |
| "energy": 11.328424135843912 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.42276112206816185, | |
| "joint_other_seed": 0.7402413756943412, | |
| "single_other_seed": 0.5592731941078808, | |
| "energy": 5412.20068359375 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.07672622034182308, | |
| "joint_other_seed": 0.7543613203937463, | |
| "single_other_seed": 0.3976547874967953, | |
| "energy": 1721.7012532552083 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.17051795244646117, | |
| "joint_other_seed": 0.5654550645369527, | |
| "single_other_seed": 0.5264533063317531, | |
| "energy": 50.66081110636393 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.159105889436158, | |
| "joint_other_seed": 0.5700942916762719, | |
| "single_other_seed": 0.5249094087397959, | |
| "energy": 4.53758446375529 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.054507750190625925, | |
| "joint_other_seed": 0.5168102400250439, | |
| "single_other_seed": 0.10463864309003279, | |
| "energy": 2386.4290364583335 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.08788947701228011, | |
| "joint_other_seed": 0.4813288341429701, | |
| "single_other_seed": 0.2413112291362709, | |
| "energy": 1840.771728515625 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.12848873518501558, | |
| "joint_other_seed": 0.42907501698788775, | |
| "single_other_seed": 0.32670545615153346, | |
| "energy": 4.358217875162761 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.21873714701861147, | |
| "joint_other_seed": 0.7217082558223875, | |
| "single_other_seed": 0.3055934700224835, | |
| "energy": 6211.3203125 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.10273487824880767, | |
| "joint_other_seed": 0.6127515968347023, | |
| "single_other_seed": 0.15272817951550088, | |
| "energy": 426.0592346191406 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.01646698926664225, | |
| "joint_other_seed": 0.33471492243475964, | |
| "single_other_seed": 0.36210041857182396, | |
| "energy": 46.86688486735026 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.049075492110751684, | |
| "joint_other_seed": 0.322637931491461, | |
| "single_other_seed": 0.26899980793177847, | |
| "energy": 5.403188546498616 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.037244104080477625, | |
| "joint_other_seed": 0.46985554725619816, | |
| "single_other_seed": 0.38398963494243255, | |
| "energy": 3304.0525716145835 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.055221596854505894, | |
| "joint_other_seed": 0.5079373668859292, | |
| "single_other_seed": 0.3167369804690122, | |
| "energy": 3097.9747721354165 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.01366512744521231, | |
| "joint_other_seed": 0.24604173367886162, | |
| "single_other_seed": 0.14961398668920167, | |
| "energy": 8.124719937642416 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.10931616926574395, | |
| "joint_other_seed": 0.6470982163867718, | |
| "single_other_seed": 0.2929816260751952, | |
| "energy": 4530.374674479167 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.08270163846548273, | |
| "joint_other_seed": 0.46402745536754764, | |
| "single_other_seed": 0.0393132383309347, | |
| "energy": 436.46240234375 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.03878286061324354, | |
| "joint_other_seed": 0.17067653429919494, | |
| "single_other_seed": 0.12592987452065718, | |
| "energy": 32.478108723958336 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.027790464687792005, | |
| "joint_other_seed": 0.33152925360525376, | |
| "single_other_seed": 0.14692226186879032, | |
| "energy": 9.426867802937826 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.051769447491392784, | |
| "joint_other_seed": 0.4600540433474335, | |
| "single_other_seed": 0.1889701344004625, | |
| "energy": 3796.3533528645835 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.07483536747528619, | |
| "joint_other_seed": 0.38615060368870013, | |
| "single_other_seed": 0.21109406152264532, | |
| "energy": 3004.9365234375 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.022069110154462335, | |
| "joint_other_seed": 0.2714606414027003, | |
| "single_other_seed": 0.0732304776958198, | |
| "energy": 18.06484095255534 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.0835299757118646, | |
| "joint_other_seed": 0.47089534634891883, | |
| "single_other_seed": 0.31437607038787596, | |
| "energy": 3010.025634765625 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.05717956040120886, | |
| "joint_other_seed": 0.37482755052327543, | |
| "single_other_seed": 0.11663940224099856, | |
| "energy": 476.27268473307294 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.010706462176292564, | |
| "joint_other_seed": 0.17137315486971638, | |
| "single_other_seed": 0.010774844085124191, | |
| "energy": 53.871988932291664 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.009314562071010335, | |
| "joint_other_seed": 0.1829568683296634, | |
| "single_other_seed": 0.1205982594903281, | |
| "energy": 12.167282422383627 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.01638605976235119, | |
| "joint_other_seed": 0.37290951674036654, | |
| "single_other_seed": 0.27611364685869666, | |
| "energy": 3772.97900390625 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.018328923044348522, | |
| "joint_other_seed": 0.41434112285197366, | |
| "single_other_seed": 0.1700221941926361, | |
| "energy": 4128.683186848958 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.0035141183142919363, | |
| "joint_other_seed": 0.21209301133437963, | |
| "single_other_seed": 0.00984234044827248, | |
| "energy": 26.293458938598633 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.0600908863037769, | |
| "joint_other_seed": 0.5738581099971661, | |
| "single_other_seed": 0.31249733395227464, | |
| "energy": 5469.968098958333 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.14317295861243906, | |
| "joint_other_seed": 0.6029377943156407, | |
| "single_other_seed": 0.10002790597170608, | |
| "energy": 1145.67041015625 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.012008449901506735, | |
| "joint_other_seed": 0.20814544923744482, | |
| "single_other_seed": 0.20930823713398128, | |
| "energy": 69.27089436848958 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.029864552009264522, | |
| "joint_other_seed": 0.3705614123023202, | |
| "single_other_seed": 0.17358361082812723, | |
| "energy": 17.978864034016926 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.026998371818568567, | |
| "joint_other_seed": 0.36491718353545255, | |
| "single_other_seed": 0.16639605862354157, | |
| "energy": 3659.3521321614585 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.02883516008703224, | |
| "joint_other_seed": 0.37557780451920675, | |
| "single_other_seed": 0.26266388980276245, | |
| "energy": 4046.4083658854165 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.006104952126649521, | |
| "joint_other_seed": 0.20202927731541542, | |
| "single_other_seed": 0.041263623481190614, | |
| "energy": 36.24147033691406 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.11679382871908683, | |
| "joint_other_seed": 0.802728832078026, | |
| "single_other_seed": 0.5409156305566543, | |
| "energy": 12271.940755208334 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.16005882175459607, | |
| "joint_other_seed": 0.7634163687838359, | |
| "single_other_seed": 0.4127825079777623, | |
| "energy": 1975.7147216796875 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.01150498578023007, | |
| "joint_other_seed": 0.22104473450500595, | |
| "single_other_seed": 0.13229671573213586, | |
| "energy": 77.99509938557942 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.1409453481804842, | |
| "joint_other_seed": 0.5650721785069709, | |
| "single_other_seed": 0.5243198653941963, | |
| "energy": 30.205067952473957 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.01776252336575368, | |
| "joint_other_seed": 0.3093899017124394, | |
| "single_other_seed": 0.2561361224819235, | |
| "energy": 4211.59033203125 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.018606390776848336, | |
| "joint_other_seed": 0.25624484571856776, | |
| "single_other_seed": 0.1416972722214367, | |
| "energy": 3968.627685546875 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.005273454435498537, | |
| "joint_other_seed": 0.18604771485504212, | |
| "single_other_seed": 0.05344201643102686, | |
| "energy": 50.05035527547201 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.020091887407558517, | |
| "joint_other_seed": 0.26470880019096077, | |
| "single_other_seed": 0.03879788879264867, | |
| "energy": 4839.836751302083 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.004603718290691394, | |
| "joint_other_seed": 0.06778613717496262, | |
| "single_other_seed": 0.007674959365953304, | |
| "energy": 286.3142496744792 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.014196133126836652, | |
| "joint_other_seed": 0.14508185983412547, | |
| "single_other_seed": 0.10182615658890926, | |
| "energy": 56.53026835123698 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.011502875229082776, | |
| "joint_other_seed": 0.16978894606652717, | |
| "single_other_seed": 0.07743756724573647, | |
| "energy": 33.75439580281576 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.017565959784498153, | |
| "joint_other_seed": 0.24407879429805202, | |
| "single_other_seed": 0.12826154619762153, | |
| "energy": 4009.77734375 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.018775775374452852, | |
| "joint_other_seed": 0.2596254514990527, | |
| "single_other_seed": 0.2710579694821311, | |
| "energy": 3740.771240234375 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.006419375889149204, | |
| "joint_other_seed": 0.140109005777317, | |
| "single_other_seed": 0.06917540793465327, | |
| "energy": 68.14927927652995 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.13419722666988476, | |
| "joint_other_seed": 0.4106605668603365, | |
| "single_other_seed": 0.40000490722120674, | |
| "energy": 4504.693196614583 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.012379883828700144, | |
| "joint_other_seed": 0.20908806679215153, | |
| "single_other_seed": 0.08777484586826956, | |
| "energy": 510.5575459798177 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.011497443914622485, | |
| "joint_other_seed": 0.1702193983026471, | |
| "single_other_seed": 0.030195208681685792, | |
| "energy": 86.99686686197917 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.012304556094985153, | |
| "joint_other_seed": 0.20801161035522808, | |
| "single_other_seed": 0.0725305279306323, | |
| "energy": 48.32108688354492 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.026427394926708092, | |
| "joint_other_seed": 0.3167644973368231, | |
| "single_other_seed": 0.28307664810102695, | |
| "energy": 4703.8486328125 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.013745990987642065, | |
| "joint_other_seed": 0.2669794249622459, | |
| "single_other_seed": 0.1517876894030512, | |
| "energy": 4176.990234375 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.004687042804614627, | |
| "joint_other_seed": 0.1858797113910482, | |
| "single_other_seed": 0.0642206656561638, | |
| "energy": 107.42502848307292 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.1263247970734157, | |
| "joint_other_seed": 0.4175920676110656, | |
| "single_other_seed": 0.4231721111403465, | |
| "energy": 3974.5655110677085 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.02941033313717112, | |
| "joint_other_seed": 0.3150203227634607, | |
| "single_other_seed": 0.17922029899926128, | |
| "energy": 556.3469950358073 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.018579223879341522, | |
| "joint_other_seed": 0.2449401815520874, | |
| "single_other_seed": 0.04469842287671, | |
| "energy": 101.09486389160156 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.07032984059396441, | |
| "joint_other_seed": 0.3701988520149212, | |
| "single_other_seed": 0.0762138161992767, | |
| "energy": 70.29944864908855 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.043627448141379865, | |
| "joint_other_seed": 0.2983272208472789, | |
| "single_other_seed": 0.40287580669562406, | |
| "energy": 4131.12841796875 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.032863438922197785, | |
| "joint_other_seed": 0.271737094539541, | |
| "single_other_seed": 0.39129970592240937, | |
| "energy": 3650.54931640625 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.023810379798211323, | |
| "joint_other_seed": 0.2687827233845965, | |
| "single_other_seed": 0.04747030511104855, | |
| "energy": 150.10559590657553 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.059104861306159995, | |
| "joint_other_seed": 0.48282208378121505, | |
| "single_other_seed": 0.1978132273597456, | |
| "energy": 6109.519368489583 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.03243743627999618, | |
| "joint_other_seed": 0.24353893122002568, | |
| "single_other_seed": 0.0354728814912888, | |
| "energy": 668.5298461914062 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.04144724202328853, | |
| "joint_other_seed": 0.40086929200881904, | |
| "single_other_seed": 0.1514474531991653, | |
| "energy": 152.27714029947916 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.14081688398438585, | |
| "joint_other_seed": 0.5119835274614595, | |
| "single_other_seed": 0.3181661841830635, | |
| "energy": 95.81234232584636 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.04326796302230773, | |
| "joint_other_seed": 0.3342780042194443, | |
| "single_other_seed": 0.29169242240146276, | |
| "energy": 4254.623860677083 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.028323078821068215, | |
| "joint_other_seed": 0.31882513437699844, | |
| "single_other_seed": 0.27255315041742323, | |
| "energy": 4150.495442708333 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.07078854953438785, | |
| "joint_other_seed": 0.44983650487345017, | |
| "single_other_seed": 0.10362526313668152, | |
| "energy": 260.761967976888 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.04410127402986408, | |
| "joint_other_seed": 0.3403254736021497, | |
| "single_other_seed": 0.2042873631030948, | |
| "energy": 5809.93798828125 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.02689813114706352, | |
| "joint_other_seed": 0.2197395550354042, | |
| "single_other_seed": 0.07144907017433422, | |
| "energy": 697.1281534830729 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.20769234090135913, | |
| "joint_other_seed": 0.6521861911509768, | |
| "single_other_seed": 0.3912756424245397, | |
| "energy": 225.02582295735678 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.4436876709117512, | |
| "joint_other_seed": 0.7561767022636587, | |
| "single_other_seed": 0.65317524746437, | |
| "energy": 256.0254414876302 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.07543803836029095, | |
| "joint_other_seed": 0.4559043382507185, | |
| "single_other_seed": 0.4343533199384628, | |
| "energy": 4424.90283203125 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.095965736364518, | |
| "joint_other_seed": 0.4648327168817463, | |
| "single_other_seed": 0.4655687911571512, | |
| "energy": 3866.8645833333335 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.23250709751284449, | |
| "joint_other_seed": 0.6429156333316934, | |
| "single_other_seed": 0.44770980648376535, | |
| "energy": 557.8719278971354 | |
| } | |
| }, | |
| "learned/contrast/r1": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.12286428103793867, | |
| "joint_other_seed": 0.2651533863148765, | |
| "single_other_seed": 0.48618083276082924, | |
| "energy": 22901.712239583332 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.3182673343674174, | |
| "joint_other_seed": 0.5926605219051344, | |
| "single_other_seed": 0.7262603629953079, | |
| "energy": 3791.4009602864585 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.21550027087200263, | |
| "joint_other_seed": 0.57404590266586, | |
| "single_other_seed": 0.636953706173659, | |
| "energy": 309.2820129394531 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.30643930194857605, | |
| "joint_other_seed": 0.6401317592814624, | |
| "single_other_seed": 0.7606565417492044, | |
| "energy": 0.35029323399066925 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.10898760916126338, | |
| "joint_other_seed": 0.18562716279267066, | |
| "single_other_seed": 0.5442148827862126, | |
| "energy": 2225.4239908854165 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.12492599021934603, | |
| "joint_other_seed": 0.413861724755327, | |
| "single_other_seed": 0.19317809779895123, | |
| "energy": 2026.8141276041667 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.10983054974448136, | |
| "joint_other_seed": 0.1589655466510156, | |
| "single_other_seed": 0.4555644699754827, | |
| "energy": 12.237568457921347 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.31966263222112234, | |
| "joint_other_seed": 0.6590486465473548, | |
| "single_other_seed": 0.264540392850826, | |
| "energy": 475.47231038411456 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.23399183984581018, | |
| "joint_other_seed": 0.2687803346277643, | |
| "single_other_seed": 0.5888040822325997, | |
| "energy": 702.3123881022135 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.468012001448574, | |
| "joint_other_seed": 0.6354566354610669, | |
| "single_other_seed": 0.794095946816052, | |
| "energy": 23.995681762695312 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.3935327218193718, | |
| "joint_other_seed": 0.6121401419607863, | |
| "single_other_seed": 0.6178030537252293, | |
| "energy": 0.10837081323067348 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.13903930939302744, | |
| "joint_other_seed": 0.2608517762600354, | |
| "single_other_seed": 0.15757835979430157, | |
| "energy": 76.67850748697917 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.13617083823304177, | |
| "joint_other_seed": 0.4277403106420967, | |
| "single_other_seed": 0.2710334954656072, | |
| "energy": 61.23396746317545 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.26070729341911403, | |
| "joint_other_seed": 0.12807442059255092, | |
| "single_other_seed": 0.3774171822217186, | |
| "energy": 0.1759335845708847 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.2482601114934344, | |
| "joint_other_seed": 0.33495092070406574, | |
| "single_other_seed": 0.3351431001647752, | |
| "energy": 239.677001953125 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.3408814872474569, | |
| "joint_other_seed": 0.320671077012209, | |
| "single_other_seed": 0.29451519427575235, | |
| "energy": 71.13290786743164 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.1674406061200696, | |
| "joint_other_seed": 0.14651934938019961, | |
| "single_other_seed": 0.42886259080040146, | |
| "energy": 0.9446612000465393 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.14019014164947363, | |
| "joint_other_seed": 0.10003355719455087, | |
| "single_other_seed": 0.33648888161789847, | |
| "energy": 0.02753126434981823 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.13667766734870168, | |
| "joint_other_seed": 0.3573322034607883, | |
| "single_other_seed": 0.4649530321186872, | |
| "energy": 230.46971893310547 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.20395212331564927, | |
| "joint_other_seed": 0.5115546086266373, | |
| "single_other_seed": 0.4869640513674657, | |
| "energy": 253.3546600341797 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.07281640458457006, | |
| "joint_other_seed": 0.2096760778148581, | |
| "single_other_seed": 0.12230449318126191, | |
| "energy": 0.22955509026845297 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.07801184988937437, | |
| "joint_other_seed": 0.2556117576262079, | |
| "single_other_seed": 0.41468748682001444, | |
| "energy": 986.7871602376302 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.10048149208195867, | |
| "joint_other_seed": 0.006122397598963714, | |
| "single_other_seed": 0.10946890989919589, | |
| "energy": 58.940550804138184 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.076937191457212, | |
| "joint_other_seed": 0.2296395406943852, | |
| "single_other_seed": 0.05247028029026857, | |
| "energy": 0.41371147831281024 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.02748679214272498, | |
| "joint_other_seed": 0.009614721531405857, | |
| "single_other_seed": 0.35057642036418457, | |
| "energy": 0.12439110378424327 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.12081495201183834, | |
| "joint_other_seed": 0.08693779684858871, | |
| "single_other_seed": 0.2682200882309549, | |
| "energy": 240.42602030436197 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.07832333540268044, | |
| "joint_other_seed": 0.18739677248287234, | |
| "single_other_seed": 0.27171680926975217, | |
| "energy": 355.21673583984375 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.04966019115930265, | |
| "joint_other_seed": 0.05653983891804004, | |
| "single_other_seed": 0.06791322857231885, | |
| "energy": 0.5109515289465586 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.25513148528210894, | |
| "joint_other_seed": 0.6691708544831959, | |
| "single_other_seed": 0.3882469086969097, | |
| "energy": 65.75610860188802 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.04264049013135418, | |
| "joint_other_seed": 0.025153052371095108, | |
| "single_other_seed": 0.09334792370916878, | |
| "energy": 3.880064010620117 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.01852613787324786, | |
| "joint_other_seed": 0.022833890597847795, | |
| "single_other_seed": 0.008420181910096636, | |
| "energy": 0.3671884437402089 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.017466796669591833, | |
| "joint_other_seed": 0.016316431874527138, | |
| "single_other_seed": 0.14491317985256214, | |
| "energy": 0.013528230600059032 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.06444103441073668, | |
| "joint_other_seed": 0.08412903867320598, | |
| "single_other_seed": 0.43486394626663594, | |
| "energy": 205.6046396891276 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.11968936126760743, | |
| "joint_other_seed": 0.10256932573844302, | |
| "single_other_seed": 0.3244772761426219, | |
| "energy": 275.95021565755206 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.009963650499806409, | |
| "joint_other_seed": 0.012500031949879808, | |
| "single_other_seed": 0.049132571432760286, | |
| "energy": 0.12728608896334967 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.08807330097196342, | |
| "joint_other_seed": 0.17523739086319512, | |
| "single_other_seed": 0.5184225239559133, | |
| "energy": 91.20072555541992 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.24996647685393303, | |
| "joint_other_seed": 0.33705754559131895, | |
| "single_other_seed": 0.15775015215516175, | |
| "energy": 159.19631958007812 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.15629717163673562, | |
| "joint_other_seed": 0.30040399490110603, | |
| "single_other_seed": 0.19602363780776, | |
| "energy": 3.4860304991404214 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.11564497659575967, | |
| "joint_other_seed": 0.31609687478295156, | |
| "single_other_seed": 0.5217404599068344, | |
| "energy": 0.8083964586257935 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.1602447820814784, | |
| "joint_other_seed": 0.3235134632123271, | |
| "single_other_seed": 0.15111415309571063, | |
| "energy": 45.63208516438802 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.20836587526499029, | |
| "joint_other_seed": 0.525128426854, | |
| "single_other_seed": 0.563161741919724, | |
| "energy": 181.70380147298178 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.0028726791461577417, | |
| "joint_other_seed": 0.01906995963869329, | |
| "single_other_seed": 0.09641240155362496, | |
| "energy": 0.24612701932589212 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.21368187256518534, | |
| "joint_other_seed": 0.349381515852101, | |
| "single_other_seed": 0.6923714340090474, | |
| "energy": 274.1408754984538 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.3478286642365238, | |
| "joint_other_seed": 0.6620070851865225, | |
| "single_other_seed": 0.5572470405855015, | |
| "energy": 292.0780537923177 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.06486670344963097, | |
| "joint_other_seed": 0.006908616112365373, | |
| "single_other_seed": 0.3485456490432278, | |
| "energy": 5.235250631968181 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.08067308860793165, | |
| "joint_other_seed": 0.03792022769728076, | |
| "single_other_seed": 0.7037768138336717, | |
| "energy": 4.529441674550374 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.20098193701293765, | |
| "joint_other_seed": 0.3276326402693275, | |
| "single_other_seed": 0.456616220616337, | |
| "energy": 63.24528884887695 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.10533349333715336, | |
| "joint_other_seed": 0.3385079886681076, | |
| "single_other_seed": 0.2601667715820517, | |
| "energy": 35.248020807902016 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.009681008599218277, | |
| "joint_other_seed": 0.0016463190142137485, | |
| "single_other_seed": 0.10964224386530597, | |
| "energy": 0.484230101108551 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.03112642378884689, | |
| "joint_other_seed": 0.2423496726830865, | |
| "single_other_seed": 0.12316772375131324, | |
| "energy": 73.93813578287761 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.012704320345809705, | |
| "joint_other_seed": 0.0025010350053236152, | |
| "single_other_seed": 0.01921614350994416, | |
| "energy": 4.577728350957234 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.024638687049082786, | |
| "joint_other_seed": 0.027157670762029573, | |
| "single_other_seed": 0.09903304311479666, | |
| "energy": 0.2396919627984365 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.0372968619063061, | |
| "joint_other_seed": 0.12028214932231296, | |
| "single_other_seed": 0.15256205428566627, | |
| "energy": 0.0447481299440066 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.08243755101414108, | |
| "joint_other_seed": 0.05553314094235784, | |
| "single_other_seed": 0.37502767011563776, | |
| "energy": 12.339414914449057 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.17852010736566348, | |
| "joint_other_seed": 0.21373484283260225, | |
| "single_other_seed": 0.5330037798171711, | |
| "energy": 32.25030390421549 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.0066978646952425835, | |
| "joint_other_seed": 0.06875367850484415, | |
| "single_other_seed": 0.07378363938351762, | |
| "energy": 0.25368601083755493 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.11892443341909198, | |
| "joint_other_seed": 0.4433493545846485, | |
| "single_other_seed": 0.45822458416630096, | |
| "energy": 36.84699694315592 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.01359364154557341, | |
| "joint_other_seed": 0.08971905482689513, | |
| "single_other_seed": 0.15623901833412324, | |
| "energy": 4.812166452407837 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.009577152921625975, | |
| "joint_other_seed": 0.03875026182654806, | |
| "single_other_seed": 0.025393060622785568, | |
| "energy": 0.5905129412810007 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.02376780732374628, | |
| "joint_other_seed": 0.024734860266378278, | |
| "single_other_seed": 0.10560073073632037, | |
| "energy": 0.03795967002709707 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.16897557698161858, | |
| "joint_other_seed": 0.4116188812868248, | |
| "single_other_seed": 0.33321677942553374, | |
| "energy": 12.832961718241373 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.08332194598382636, | |
| "joint_other_seed": 0.15524934017720315, | |
| "single_other_seed": 0.2609375432656098, | |
| "energy": 16.014588991800945 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.01193962452250985, | |
| "joint_other_seed": 0.06214743450029481, | |
| "single_other_seed": 0.044835795137279866, | |
| "energy": 0.21946248412132263 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.12071905682226425, | |
| "joint_other_seed": 0.22456940601277553, | |
| "single_other_seed": 0.517111134169827, | |
| "energy": 21.075810114542644 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.1328178341710734, | |
| "joint_other_seed": 0.3416602610812326, | |
| "single_other_seed": 0.2755029587459753, | |
| "energy": 2.801560322443644 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.06053918663280822, | |
| "joint_other_seed": 0.18460066425858357, | |
| "single_other_seed": 0.10472344564303182, | |
| "energy": 1.2233876585960388 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.07985805484709729, | |
| "joint_other_seed": 0.11164946460882441, | |
| "single_other_seed": 0.03618061943068562, | |
| "energy": 0.30101268490155536 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.17375581676763252, | |
| "joint_other_seed": 0.4735254315923747, | |
| "single_other_seed": 0.4168090504983608, | |
| "energy": 12.726617495218912 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.1779545740429177, | |
| "joint_other_seed": 0.5498557706555418, | |
| "single_other_seed": 0.4136492835316682, | |
| "energy": 15.789840698242188 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.016768058674552777, | |
| "joint_other_seed": 0.06628519112646603, | |
| "single_other_seed": 0.07640359053796553, | |
| "energy": 0.47260130445162457 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.05029410198698703, | |
| "joint_other_seed": 0.1551411903546841, | |
| "single_other_seed": 0.2986384887643364, | |
| "energy": 32.0254275004069 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.020717825083280017, | |
| "joint_other_seed": 0.1318484391194641, | |
| "single_other_seed": 0.02067663848853829, | |
| "energy": 5.2392152150472 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.10623606591085155, | |
| "joint_other_seed": 0.33001526375372714, | |
| "single_other_seed": 0.2931924783052734, | |
| "energy": 2.2324507236480713 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.14847221783302522, | |
| "joint_other_seed": 0.18012135922667016, | |
| "single_other_seed": 0.38891425463576923, | |
| "energy": 0.4293999473253886 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.18401272802102386, | |
| "joint_other_seed": 0.6654712636751177, | |
| "single_other_seed": 0.38801351515503035, | |
| "energy": 29.286324818929035 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.21264583318612906, | |
| "joint_other_seed": 0.27664808299639776, | |
| "single_other_seed": 0.4068551848733063, | |
| "energy": 35.27225240071615 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.041326048756914036, | |
| "joint_other_seed": 0.13054153484088427, | |
| "single_other_seed": 0.11020948587380763, | |
| "energy": 1.1235290169715881 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.04612774635917816, | |
| "joint_other_seed": 0.3169104200196442, | |
| "single_other_seed": 0.2375995472995207, | |
| "energy": 87.49025853474934 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.0020428553897725577, | |
| "joint_other_seed": 0.04023568750363549, | |
| "single_other_seed": 0.05095066735116949, | |
| "energy": 8.71691862742106 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.2427744754915363, | |
| "joint_other_seed": 0.5773133243839965, | |
| "single_other_seed": 0.38034398209421716, | |
| "energy": 4.420498530069987 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.5798741426242487, | |
| "joint_other_seed": 0.7954655322002272, | |
| "single_other_seed": 0.6987350110949648, | |
| "energy": 3.9688194592793784 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.29317876485411154, | |
| "joint_other_seed": 0.5630919714195417, | |
| "single_other_seed": 0.5686077190807861, | |
| "energy": 40.210375467936196 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.24715463899838586, | |
| "joint_other_seed": 0.0717793974294938, | |
| "single_other_seed": 0.6401279363532715, | |
| "energy": 32.091957092285156 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.26430108613657044, | |
| "joint_other_seed": 0.4543908051960366, | |
| "single_other_seed": 0.34838439716448355, | |
| "energy": 3.2152388095855713 | |
| } | |
| }, | |
| "learned/contrast/r3": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.17004236581000765, | |
| "joint_other_seed": 0.6138245451061912, | |
| "single_other_seed": 0.48618083276082924, | |
| "energy": 19232.75390625 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.399851876329688, | |
| "joint_other_seed": 0.8210375338598853, | |
| "single_other_seed": 0.7262603629953079, | |
| "energy": 2318.388427734375 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.24094148116230352, | |
| "joint_other_seed": 0.8064207265805252, | |
| "single_other_seed": 0.636953706173659, | |
| "energy": 269.0149434407552 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.3516235641064388, | |
| "joint_other_seed": 0.7541482343041264, | |
| "single_other_seed": 0.7606565417492044, | |
| "energy": 0.40534820159276325 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.1620186859485353, | |
| "joint_other_seed": 0.4701196257524103, | |
| "single_other_seed": 0.5442148827862126, | |
| "energy": 1662.6458740234375 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.12367681192907132, | |
| "joint_other_seed": 0.3449885657843659, | |
| "single_other_seed": 0.19317809779895123, | |
| "energy": 1003.1679484049479 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.18014428670697524, | |
| "joint_other_seed": 0.6655762074314966, | |
| "single_other_seed": 0.4555644699754827, | |
| "energy": 9.160479227701822 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.27368414275841935, | |
| "joint_other_seed": 0.5174727862671895, | |
| "single_other_seed": 0.264540392850826, | |
| "energy": 404.20619201660156 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.3368972837180113, | |
| "joint_other_seed": 0.7530975580828396, | |
| "single_other_seed": 0.5888040822325997, | |
| "energy": 582.6297403971354 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.4926666688321466, | |
| "joint_other_seed": 0.7115051396725154, | |
| "single_other_seed": 0.794095946816052, | |
| "energy": 20.25833511352539 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.29744973357242865, | |
| "joint_other_seed": 0.5234791497257434, | |
| "single_other_seed": 0.6178030537252293, | |
| "energy": 0.15204481532176337 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.07015391436068763, | |
| "joint_other_seed": 0.04403412477353219, | |
| "single_other_seed": 0.15757835979430157, | |
| "energy": 180.70311991373697 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.050756396776025614, | |
| "joint_other_seed": 0.4383066580301128, | |
| "single_other_seed": 0.2710334954656072, | |
| "energy": 203.33539326985678 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.06714885387681302, | |
| "joint_other_seed": 0.1056631805617948, | |
| "single_other_seed": 0.3774171822217186, | |
| "energy": 0.2866048167149226 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.20700158376720523, | |
| "joint_other_seed": 0.4942058597763126, | |
| "single_other_seed": 0.3351431001647752, | |
| "energy": 344.6257019042969 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.3694694393103043, | |
| "joint_other_seed": 0.6336833213195399, | |
| "single_other_seed": 0.29451519427575235, | |
| "energy": 33.631343841552734 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.0481846663532729, | |
| "joint_other_seed": 0.2599180710568668, | |
| "single_other_seed": 0.42886259080040146, | |
| "energy": 3.182020664215088 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.07261911009265254, | |
| "joint_other_seed": 0.15388577512894008, | |
| "single_other_seed": 0.33648888161789847, | |
| "energy": 0.04387051612138748 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.03202845578049662, | |
| "joint_other_seed": 0.5507221003087741, | |
| "single_other_seed": 0.4649530321186872, | |
| "energy": 382.21337890625 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.07383782126398897, | |
| "joint_other_seed": 0.44907340893429304, | |
| "single_other_seed": 0.4869640513674657, | |
| "energy": 373.6364440917969 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.03259110080818961, | |
| "joint_other_seed": 0.06614128364323024, | |
| "single_other_seed": 0.12230449318126191, | |
| "energy": 0.30566519498825073 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.1402870583236141, | |
| "joint_other_seed": 0.6567666344728835, | |
| "single_other_seed": 0.41468748682001444, | |
| "energy": 679.8792521158854 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.09414917573363046, | |
| "joint_other_seed": 0.1835824537600781, | |
| "single_other_seed": 0.10946890989919589, | |
| "energy": 28.632724126180012 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.05020280156511381, | |
| "joint_other_seed": 0.20482756030504767, | |
| "single_other_seed": 0.05247028029026857, | |
| "energy": 0.6514701147874197 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.019620826266740332, | |
| "joint_other_seed": 0.09429272243438402, | |
| "single_other_seed": 0.35057642036418457, | |
| "energy": 0.260831614335378 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.06900889361270246, | |
| "joint_other_seed": 0.22958660201455516, | |
| "single_other_seed": 0.2682200882309549, | |
| "energy": 312.4665069580078 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.10815075846733924, | |
| "joint_other_seed": 0.3603840333236995, | |
| "single_other_seed": 0.27171680926975217, | |
| "energy": 401.9179178873698 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.054307073018716734, | |
| "joint_other_seed": 0.16988901084349028, | |
| "single_other_seed": 0.06791322857231885, | |
| "energy": 1.0880539417266846 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.1667063389411081, | |
| "joint_other_seed": 0.44440990401502745, | |
| "single_other_seed": 0.3882469086969097, | |
| "energy": 58.55854415893555 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.10205461377247249, | |
| "joint_other_seed": 0.22491154185563056, | |
| "single_other_seed": 0.09334792370916878, | |
| "energy": 6.0050614674886065 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.02870826917279168, | |
| "joint_other_seed": 0.04712190954657758, | |
| "single_other_seed": 0.008420181910096636, | |
| "energy": 0.7154272397359213 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.010496809509546012, | |
| "joint_other_seed": 0.03834477240733447, | |
| "single_other_seed": 0.14491317985256214, | |
| "energy": 0.029711920768022537 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.0333871039093081, | |
| "joint_other_seed": 0.20260813279389045, | |
| "single_other_seed": 0.43486394626663594, | |
| "energy": 192.46937052408853 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.057010177189465217, | |
| "joint_other_seed": 0.36408182526961186, | |
| "single_other_seed": 0.3244772761426219, | |
| "energy": 337.0934651692708 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.025490466117478294, | |
| "joint_other_seed": 0.02789989010595173, | |
| "single_other_seed": 0.049132571432760286, | |
| "energy": 0.24669378995895386 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.07824170021470955, | |
| "joint_other_seed": 0.29507999829527953, | |
| "single_other_seed": 0.5184225239559133, | |
| "energy": 87.77853139241536 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.2417393689314724, | |
| "joint_other_seed": 0.6293675092698894, | |
| "single_other_seed": 0.15775015215516175, | |
| "energy": 84.77686309814453 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.12118029481103276, | |
| "joint_other_seed": 0.20952807061424594, | |
| "single_other_seed": 0.19602363780776, | |
| "energy": 4.08453106880188 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.19699399037031262, | |
| "joint_other_seed": 0.6198419852314851, | |
| "single_other_seed": 0.5217404599068344, | |
| "energy": 0.8281291325887045 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.053028980251891844, | |
| "joint_other_seed": 0.2867130507397715, | |
| "single_other_seed": 0.15111415309571063, | |
| "energy": 216.25977579752603 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.09467255104895222, | |
| "joint_other_seed": 0.4007661886640542, | |
| "single_other_seed": 0.563161741919724, | |
| "energy": 214.08226521809897 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.004144534148157297, | |
| "joint_other_seed": 0.08154757619150155, | |
| "single_other_seed": 0.09641240155362496, | |
| "energy": 0.4191741744677226 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.3360939720059134, | |
| "joint_other_seed": 0.7737842945083195, | |
| "single_other_seed": 0.6923714340090474, | |
| "energy": 231.93643697102866 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.31937685245639413, | |
| "joint_other_seed": 0.8502398578263967, | |
| "single_other_seed": 0.5572470405855015, | |
| "energy": 315.93918863932294 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.11098147141362757, | |
| "joint_other_seed": 0.28092367896873516, | |
| "single_other_seed": 0.3485456490432278, | |
| "energy": 6.476069609324138 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.2799477441797862, | |
| "joint_other_seed": 0.5827885275533549, | |
| "single_other_seed": 0.7037768138336717, | |
| "energy": 10.61158561706543 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.10516892554323376, | |
| "joint_other_seed": 0.27253372987901847, | |
| "single_other_seed": 0.456616220616337, | |
| "energy": 91.83975982666016 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.10319123601343239, | |
| "joint_other_seed": 0.27427434886874724, | |
| "single_other_seed": 0.2601667715820517, | |
| "energy": 74.7890739440918 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.016122546054429532, | |
| "joint_other_seed": 0.10235176605259433, | |
| "single_other_seed": 0.10964224386530597, | |
| "energy": 0.7985434532165527 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.03915894331392489, | |
| "joint_other_seed": 0.13396612812978037, | |
| "single_other_seed": 0.12316772375131324, | |
| "energy": 55.886189778645836 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.021423426568727533, | |
| "joint_other_seed": 0.039217845378733185, | |
| "single_other_seed": 0.01921614350994416, | |
| "energy": 4.370078166325887 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.018341922060633342, | |
| "joint_other_seed": 0.01610898515346315, | |
| "single_other_seed": 0.09903304311479666, | |
| "energy": 0.44588852922121686 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.03500225577378746, | |
| "joint_other_seed": 0.0910622715904662, | |
| "single_other_seed": 0.15256205428566627, | |
| "energy": 0.07612619052330653 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.058415040767751725, | |
| "joint_other_seed": 0.1472925742396276, | |
| "single_other_seed": 0.37502767011563776, | |
| "energy": 28.41834831237793 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.13208226770337517, | |
| "joint_other_seed": 0.2774598884189462, | |
| "single_other_seed": 0.5330037798171711, | |
| "energy": 41.56993865966797 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.005336710434096781, | |
| "joint_other_seed": 0.050906709346249746, | |
| "single_other_seed": 0.07378363938351762, | |
| "energy": 0.42929046352704364 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.09996062966965813, | |
| "joint_other_seed": 0.3690614530500653, | |
| "single_other_seed": 0.45822458416630096, | |
| "energy": 37.60566838582357 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.02246098260220708, | |
| "joint_other_seed": 0.09119573062439179, | |
| "single_other_seed": 0.15623901833412324, | |
| "energy": 4.5531392097473145 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.014369384034718546, | |
| "joint_other_seed": 0.0852664936644274, | |
| "single_other_seed": 0.025393060622785568, | |
| "energy": 0.7440875569979349 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.02084598816025729, | |
| "joint_other_seed": 0.18187526556955536, | |
| "single_other_seed": 0.10560073073632037, | |
| "energy": 0.07342253625392914 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.11564064144820607, | |
| "joint_other_seed": 0.26093218079879577, | |
| "single_other_seed": 0.33321677942553374, | |
| "energy": 20.785385767618816 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.038986904662007826, | |
| "joint_other_seed": 0.20841960959367223, | |
| "single_other_seed": 0.2609375432656098, | |
| "energy": 22.995131810506184 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.007843243514039755, | |
| "joint_other_seed": 0.06947030799594978, | |
| "single_other_seed": 0.044835795137279866, | |
| "energy": 0.46579739451408386 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.07235846428476167, | |
| "joint_other_seed": 0.19577407987694118, | |
| "single_other_seed": 0.517111134169827, | |
| "energy": 23.169044494628906 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.06731874586481923, | |
| "joint_other_seed": 0.14532575511151857, | |
| "single_other_seed": 0.2755029587459753, | |
| "energy": 3.661405404408773 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.03430341922076774, | |
| "joint_other_seed": 0.233869600987897, | |
| "single_other_seed": 0.10472344564303182, | |
| "energy": 1.385449767112732 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.08168924425891355, | |
| "joint_other_seed": 0.26063059336631145, | |
| "single_other_seed": 0.03618061943068562, | |
| "energy": 0.6819771925608317 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.10874486252667581, | |
| "joint_other_seed": 0.22997708800078243, | |
| "single_other_seed": 0.4168090504983608, | |
| "energy": 21.528710047403973 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.10815805027241998, | |
| "joint_other_seed": 0.2792229567033554, | |
| "single_other_seed": 0.4136492835316682, | |
| "energy": 19.88881492614746 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.02139373089466634, | |
| "joint_other_seed": 0.13872732918937158, | |
| "single_other_seed": 0.07640359053796553, | |
| "energy": 0.7773478825887045 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.07024788025756312, | |
| "joint_other_seed": 0.22693546721486968, | |
| "single_other_seed": 0.2986384887643364, | |
| "energy": 31.011003494262695 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.026901487287366854, | |
| "joint_other_seed": 0.19197671465454802, | |
| "single_other_seed": 0.02067663848853829, | |
| "energy": 4.272938172022502 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.1229259187661893, | |
| "joint_other_seed": 0.541684802906705, | |
| "single_other_seed": 0.2931924783052734, | |
| "energy": 2.3281158606211343 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.2086585655048684, | |
| "joint_other_seed": 0.4463195931122328, | |
| "single_other_seed": 0.38891425463576923, | |
| "energy": 0.6363293528556824 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.16171160127167972, | |
| "joint_other_seed": 0.37783853967044756, | |
| "single_other_seed": 0.38801351515503035, | |
| "energy": 32.18111673990885 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.1385016320260276, | |
| "joint_other_seed": 0.3974353983469673, | |
| "single_other_seed": 0.4068551848733063, | |
| "energy": 37.43267822265625 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.07332000925788751, | |
| "joint_other_seed": 0.29226016943447763, | |
| "single_other_seed": 0.11020948587380763, | |
| "energy": 1.743136207262675 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.0638201273640725, | |
| "joint_other_seed": 0.3206326925679248, | |
| "single_other_seed": 0.2375995472995207, | |
| "energy": 43.83546829223633 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.015855962014519, | |
| "joint_other_seed": 0.05487576068356437, | |
| "single_other_seed": 0.05095066735116949, | |
| "energy": 5.645834445953369 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.2313141699799084, | |
| "joint_other_seed": 0.7407377133214557, | |
| "single_other_seed": 0.38034398209421716, | |
| "energy": 4.3596146901448565 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.5457954473218873, | |
| "joint_other_seed": 0.8023157649044788, | |
| "single_other_seed": 0.6987350110949648, | |
| "energy": 3.361826181411743 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.25228278775220336, | |
| "joint_other_seed": 0.5281936011284748, | |
| "single_other_seed": 0.5686077190807861, | |
| "energy": 49.47104136149088 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.2562584311083085, | |
| "joint_other_seed": 0.5932172716263173, | |
| "single_other_seed": 0.6401279363532715, | |
| "energy": 52.73363240559896 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.21996843921138132, | |
| "joint_other_seed": 0.5476724975512358, | |
| "single_other_seed": 0.34838439716448355, | |
| "energy": 5.826353232065837 | |
| } | |
| }, | |
| "learned/contrast/r16": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.18969629628891094, | |
| "joint_other_seed": 0.754035718544253, | |
| "single_other_seed": 0.48618083276082924, | |
| "energy": 15345.241861979166 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.4524771557165933, | |
| "joint_other_seed": 0.859630529030676, | |
| "single_other_seed": 0.7262603629953079, | |
| "energy": 1683.2373860677083 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.2825894070003202, | |
| "joint_other_seed": 0.8315462067350513, | |
| "single_other_seed": 0.636953706173659, | |
| "energy": 247.49786376953125 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.3812651171196182, | |
| "joint_other_seed": 0.8288118522364716, | |
| "single_other_seed": 0.7606565417492044, | |
| "energy": 0.45515358448028564 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.20413924475268952, | |
| "joint_other_seed": 0.6624498493418213, | |
| "single_other_seed": 0.5442148827862126, | |
| "energy": 1098.8770141601562 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.13911270075803972, | |
| "joint_other_seed": 0.6853567016288029, | |
| "single_other_seed": 0.19317809779895123, | |
| "energy": 1189.8199259440105 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.23137813152505615, | |
| "joint_other_seed": 0.7921477757857599, | |
| "single_other_seed": 0.4555644699754827, | |
| "energy": 6.650694529215495 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.32140164033644214, | |
| "joint_other_seed": 0.5708037310048013, | |
| "single_other_seed": 0.264540392850826, | |
| "energy": 194.61229197184244 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.36497008112860374, | |
| "joint_other_seed": 0.8310479377033952, | |
| "single_other_seed": 0.5888040822325997, | |
| "energy": 479.17397054036456 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.501386251801518, | |
| "joint_other_seed": 0.7813842522042268, | |
| "single_other_seed": 0.794095946816052, | |
| "energy": 13.904192288716635 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.2884986431495927, | |
| "joint_other_seed": 0.7176593754037828, | |
| "single_other_seed": 0.6178030537252293, | |
| "energy": 0.21262029310067496 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.0569305373665564, | |
| "joint_other_seed": 0.6375347138711656, | |
| "single_other_seed": 0.15757835979430157, | |
| "energy": 264.7279764811198 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.058122019309734596, | |
| "joint_other_seed": 0.61736201401231, | |
| "single_other_seed": 0.2710334954656072, | |
| "energy": 153.7758331298828 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.07884960798903606, | |
| "joint_other_seed": 0.3443462457348896, | |
| "single_other_seed": 0.3774171822217186, | |
| "energy": 0.24554108083248138 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.19444703455446066, | |
| "joint_other_seed": 0.766307411661413, | |
| "single_other_seed": 0.3351431001647752, | |
| "energy": 318.3415934244792 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.40148646951274714, | |
| "joint_other_seed": 0.7756315600381014, | |
| "single_other_seed": 0.29451519427575235, | |
| "energy": 30.446348190307617 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.033493953306994784, | |
| "joint_other_seed": 0.5066288043850011, | |
| "single_other_seed": 0.42886259080040146, | |
| "energy": 3.5517707665761313 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.053026852594695084, | |
| "joint_other_seed": 0.3846339928042739, | |
| "single_other_seed": 0.33648888161789847, | |
| "energy": 0.06030271450678507 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.04095299470861402, | |
| "joint_other_seed": 0.6349196710230524, | |
| "single_other_seed": 0.4649530321186872, | |
| "energy": 365.31142171223956 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.05772208266031231, | |
| "joint_other_seed": 0.6625434929709926, | |
| "single_other_seed": 0.4869640513674657, | |
| "energy": 388.6562194824219 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.036673645669949795, | |
| "joint_other_seed": 0.35883286312919704, | |
| "single_other_seed": 0.12230449318126191, | |
| "energy": 0.4319249391555786 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.17006226289019433, | |
| "joint_other_seed": 0.7686185674783201, | |
| "single_other_seed": 0.41468748682001444, | |
| "energy": 555.9290466308594 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.08731383253746294, | |
| "joint_other_seed": 0.49911438756038373, | |
| "single_other_seed": 0.10946890989919589, | |
| "energy": 19.391607920328777 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.034952614274505356, | |
| "joint_other_seed": 0.21359892096136845, | |
| "single_other_seed": 0.05247028029026857, | |
| "energy": 0.8381545543670654 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.03765732777070274, | |
| "joint_other_seed": 0.33855501726154924, | |
| "single_other_seed": 0.35057642036418457, | |
| "energy": 0.35410964985688526 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.07691497134036762, | |
| "joint_other_seed": 0.6447171907446413, | |
| "single_other_seed": 0.2682200882309549, | |
| "energy": 375.1974589029948 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.14771774056011777, | |
| "joint_other_seed": 0.5375615251034569, | |
| "single_other_seed": 0.27171680926975217, | |
| "energy": 418.7221272786458 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.07592348664275284, | |
| "joint_other_seed": 0.4554667593531323, | |
| "single_other_seed": 0.06791322857231885, | |
| "energy": 1.4027678966522217 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.17334926613541668, | |
| "joint_other_seed": 0.5671879503554975, | |
| "single_other_seed": 0.3882469086969097, | |
| "energy": 50.15190633138021 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.10893131333887807, | |
| "joint_other_seed": 0.4175999154044201, | |
| "single_other_seed": 0.09334792370916878, | |
| "energy": 5.298467477162679 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.032926359045277076, | |
| "joint_other_seed": 0.19953946133938327, | |
| "single_other_seed": 0.008420181910096636, | |
| "energy": 0.8873077233632406 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.009933007624458257, | |
| "joint_other_seed": 0.15608003960378347, | |
| "single_other_seed": 0.14491317985256214, | |
| "energy": 0.0728130613764127 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.038105504714470606, | |
| "joint_other_seed": 0.5338310349734499, | |
| "single_other_seed": 0.43486394626663594, | |
| "energy": 207.63184102376303 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.05498577023015094, | |
| "joint_other_seed": 0.6067588913766628, | |
| "single_other_seed": 0.3244772761426219, | |
| "energy": 303.9904479980469 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.02191169887916422, | |
| "joint_other_seed": 0.26013129223784326, | |
| "single_other_seed": 0.049132571432760286, | |
| "energy": 0.49068917830785114 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.09923383590544343, | |
| "joint_other_seed": 0.4872299168473103, | |
| "single_other_seed": 0.5184225239559133, | |
| "energy": 77.88979085286458 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.24705868348262627, | |
| "joint_other_seed": 0.7401988675045513, | |
| "single_other_seed": 0.15775015215516175, | |
| "energy": 63.31670633951823 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.11263965626305217, | |
| "joint_other_seed": 0.378991815832289, | |
| "single_other_seed": 0.19602363780776, | |
| "energy": 2.587951898574829 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.20962020797060632, | |
| "joint_other_seed": 0.6436777196685923, | |
| "single_other_seed": 0.5217404599068344, | |
| "energy": 0.7573725779851278 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.05620384697845682, | |
| "joint_other_seed": 0.5587285523806552, | |
| "single_other_seed": 0.15111415309571063, | |
| "energy": 197.87670389811197 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.09725054527861765, | |
| "joint_other_seed": 0.58158688712169, | |
| "single_other_seed": 0.563161741919724, | |
| "energy": 210.04161071777344 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.006102128187526534, | |
| "joint_other_seed": 0.2020866236207638, | |
| "single_other_seed": 0.09641240155362496, | |
| "energy": 0.6809099912643433 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.3418752013414744, | |
| "joint_other_seed": 0.8429678708648637, | |
| "single_other_seed": 0.6923714340090474, | |
| "energy": 217.72577921549478 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.29516740391791596, | |
| "joint_other_seed": 0.9114159548234324, | |
| "single_other_seed": 0.5572470405855015, | |
| "energy": 303.6115010579427 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.11424981257508249, | |
| "joint_other_seed": 0.3680001161892885, | |
| "single_other_seed": 0.3485456490432278, | |
| "energy": 5.359782616297404 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.2970113528077698, | |
| "joint_other_seed": 0.7589367901173103, | |
| "single_other_seed": 0.7037768138336717, | |
| "energy": 14.353742917378744 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.1036193558580123, | |
| "joint_other_seed": 0.42674549662646843, | |
| "single_other_seed": 0.456616220616337, | |
| "energy": 109.3924560546875 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.09707011310435965, | |
| "joint_other_seed": 0.3666466758988556, | |
| "single_other_seed": 0.2601667715820517, | |
| "energy": 78.05718739827473 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.014657885873168963, | |
| "joint_other_seed": 0.25525279081859387, | |
| "single_other_seed": 0.10964224386530597, | |
| "energy": 0.8939587076505026 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.05471774238017246, | |
| "joint_other_seed": 0.26737829211132413, | |
| "single_other_seed": 0.12316772375131324, | |
| "energy": 51.30871454874674 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.02087144065877844, | |
| "joint_other_seed": 0.1151530476959608, | |
| "single_other_seed": 0.01921614350994416, | |
| "energy": 2.9003668626149497 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.02306317411084058, | |
| "joint_other_seed": 0.1429818683149393, | |
| "single_other_seed": 0.09903304311479666, | |
| "energy": 0.6159220735232035 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.02827043291031739, | |
| "joint_other_seed": 0.2150555083262653, | |
| "single_other_seed": 0.15256205428566627, | |
| "energy": 0.11352109163999557 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.04637694545373583, | |
| "joint_other_seed": 0.2827995354167274, | |
| "single_other_seed": 0.37502767011563776, | |
| "energy": 42.26479848225912 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.12237803996975483, | |
| "joint_other_seed": 0.3860866352900876, | |
| "single_other_seed": 0.5330037798171711, | |
| "energy": 40.356824239095054 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.006427156008231183, | |
| "joint_other_seed": 0.15703005792059754, | |
| "single_other_seed": 0.07378363938351762, | |
| "energy": 0.7481204469998678 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.09177955821548545, | |
| "joint_other_seed": 0.46421994884388307, | |
| "single_other_seed": 0.45822458416630096, | |
| "energy": 37.367383321126304 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.02583348831146437, | |
| "joint_other_seed": 0.1843468034570781, | |
| "single_other_seed": 0.15623901833412324, | |
| "energy": 4.392136096954346 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.014287542125987152, | |
| "joint_other_seed": 0.1965192106643816, | |
| "single_other_seed": 0.025393060622785568, | |
| "energy": 0.6970632274945577 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.015506278056095057, | |
| "joint_other_seed": 0.2146337327349237, | |
| "single_other_seed": 0.10560073073632037, | |
| "energy": 0.11549265931049983 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.07780073261952025, | |
| "joint_other_seed": 0.3263035049039509, | |
| "single_other_seed": 0.33321677942553374, | |
| "energy": 26.725253423055012 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.04516860750742491, | |
| "joint_other_seed": 0.34335019206702855, | |
| "single_other_seed": 0.2609375432656098, | |
| "energy": 33.621838887532554 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.007882570768818598, | |
| "joint_other_seed": 0.17519073025496235, | |
| "single_other_seed": 0.044835795137279866, | |
| "energy": 0.7802271842956543 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.09018611091742945, | |
| "joint_other_seed": 0.381907345345876, | |
| "single_other_seed": 0.517111134169827, | |
| "energy": 20.470768610636394 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.07671767420317499, | |
| "joint_other_seed": 0.2628910553454288, | |
| "single_other_seed": 0.2755029587459753, | |
| "energy": 2.8836283683776855 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.03297732485768812, | |
| "joint_other_seed": 0.38784590177084144, | |
| "single_other_seed": 0.10472344564303182, | |
| "energy": 1.2969170808792114 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.05895615696323453, | |
| "joint_other_seed": 0.3515825367871429, | |
| "single_other_seed": 0.03618061943068562, | |
| "energy": 1.4809504747390747 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.07750818654390135, | |
| "joint_other_seed": 0.31632449317751576, | |
| "single_other_seed": 0.4168090504983608, | |
| "energy": 29.084396362304688 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.07205176071784447, | |
| "joint_other_seed": 0.34752323877972335, | |
| "single_other_seed": 0.4136492835316682, | |
| "energy": 28.281922658284504 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.02944769410377368, | |
| "joint_other_seed": 0.29605292666328364, | |
| "single_other_seed": 0.07640359053796553, | |
| "energy": 1.1800579229990642 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.0796855557616919, | |
| "joint_other_seed": 0.5024027133720625, | |
| "single_other_seed": 0.2986384887643364, | |
| "energy": 28.193771998087566 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.01859138952252093, | |
| "joint_other_seed": 0.19154578396895938, | |
| "single_other_seed": 0.02067663848853829, | |
| "energy": 3.4775284131368003 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.1250145780021883, | |
| "joint_other_seed": 0.6233645076013712, | |
| "single_other_seed": 0.2931924783052734, | |
| "energy": 2.18062416712443 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.2301385894128396, | |
| "joint_other_seed": 0.672441836568128, | |
| "single_other_seed": 0.38891425463576923, | |
| "energy": 0.6719926397005717 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.12949418285239978, | |
| "joint_other_seed": 0.48565391244503153, | |
| "single_other_seed": 0.38801351515503035, | |
| "energy": 33.08369700113932 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.1179273218996526, | |
| "joint_other_seed": 0.49237504388206443, | |
| "single_other_seed": 0.4068551848733063, | |
| "energy": 41.386243184407554 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.06844568598760747, | |
| "joint_other_seed": 0.46831528974538206, | |
| "single_other_seed": 0.11020948587380763, | |
| "energy": 1.7113364537556965 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.08944341760439023, | |
| "joint_other_seed": 0.4712793977145794, | |
| "single_other_seed": 0.2375995472995207, | |
| "energy": 34.227755864461265 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.02098885522927063, | |
| "joint_other_seed": 0.19810657137816323, | |
| "single_other_seed": 0.05095066735116949, | |
| "energy": 3.624799966812134 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.25428770260796574, | |
| "joint_other_seed": 0.8278430561320871, | |
| "single_other_seed": 0.38034398209421716, | |
| "energy": 3.926519791285197 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.5630815386594165, | |
| "joint_other_seed": 0.8478365656265289, | |
| "single_other_seed": 0.6987350110949648, | |
| "energy": 3.2167115211486816 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.2353515682253824, | |
| "joint_other_seed": 0.655384798396274, | |
| "single_other_seed": 0.5686077190807861, | |
| "energy": 57.908941904703774 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.25531229583980686, | |
| "joint_other_seed": 0.6396774593489912, | |
| "single_other_seed": 0.6401279363532715, | |
| "energy": 49.90599695841471 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.21591582052729907, | |
| "joint_other_seed": 0.6751092983485162, | |
| "single_other_seed": 0.34838439716448355, | |
| "energy": 4.489993572235107 | |
| } | |
| }, | |
| "learned/trigger/r1": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.1546703251519188, | |
| "joint_other_seed": 0.25089384417948957, | |
| "single_other_seed": 0.4643610002664895, | |
| "energy": 35032.001953125 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.3129561507314581, | |
| "joint_other_seed": 0.5784378145097739, | |
| "single_other_seed": 0.7078430244434457, | |
| "energy": 4650.430338541667 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.17172785364806697, | |
| "joint_other_seed": 0.5444824692930682, | |
| "single_other_seed": 0.6254937739311894, | |
| "energy": 153.5638427734375 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.3205745350428997, | |
| "joint_other_seed": 0.6435544404143523, | |
| "single_other_seed": 0.7068658729525884, | |
| "energy": 2.121134042739868 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.1360881380582878, | |
| "joint_other_seed": 0.20990079471745715, | |
| "single_other_seed": 0.5280085196131685, | |
| "energy": 3454.4305013020835 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.102848456067694, | |
| "joint_other_seed": 0.3688656747654868, | |
| "single_other_seed": 0.10965674975114792, | |
| "energy": 2980.877156575521 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.07576024588362396, | |
| "joint_other_seed": 0.1663724772856946, | |
| "single_other_seed": 0.21501616339504448, | |
| "energy": 10.779874920845032 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.3342225754010222, | |
| "joint_other_seed": 0.6535651090538729, | |
| "single_other_seed": 0.5239160353421755, | |
| "energy": 3147.5149739583335 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.09596919854205077, | |
| "joint_other_seed": 0.14057566842600444, | |
| "single_other_seed": 0.42768268943738197, | |
| "energy": 701.9873250325521 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.35498508338358514, | |
| "joint_other_seed": 0.5204172090109904, | |
| "single_other_seed": 0.6246131809184639, | |
| "energy": 32.56611633300781 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.26141193802477636, | |
| "joint_other_seed": 0.45549979483218717, | |
| "single_other_seed": 0.5317400852723304, | |
| "energy": 0.9999579588572184 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.08398633846397774, | |
| "joint_other_seed": 0.15010744630303519, | |
| "single_other_seed": 0.10112784213831483, | |
| "energy": 333.10606892903644 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.15983114391253384, | |
| "joint_other_seed": 0.4421944171211054, | |
| "single_other_seed": 0.24557368809346902, | |
| "energy": 449.8955993652344 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.08609976114887385, | |
| "joint_other_seed": 0.2491684837182091, | |
| "single_other_seed": 0.31293591565942924, | |
| "energy": 0.9549757838249207 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.2320677616978113, | |
| "joint_other_seed": 0.33555596279214006, | |
| "single_other_seed": 0.312575398497833, | |
| "energy": 2672.5738932291665 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.11629608891637255, | |
| "joint_other_seed": 0.1507468764602784, | |
| "single_other_seed": 0.17627260980388315, | |
| "energy": 312.1925964355469 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.054244400827739625, | |
| "joint_other_seed": 0.18033367524135724, | |
| "single_other_seed": 0.36344939213391136, | |
| "energy": 7.467344760894775 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.09832815484214896, | |
| "joint_other_seed": 0.13362580578931138, | |
| "single_other_seed": 0.26265183763983824, | |
| "energy": 0.6839713851610819 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.12935425500180264, | |
| "joint_other_seed": 0.24156716809649204, | |
| "single_other_seed": 0.405295006479894, | |
| "energy": 1075.1372884114583 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.17388493682266037, | |
| "joint_other_seed": 0.4398594346995265, | |
| "single_other_seed": 0.34238913790364267, | |
| "energy": 1004.1082763671875 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.02159172667934417, | |
| "joint_other_seed": 0.1807550629291288, | |
| "single_other_seed": 0.14465841866469964, | |
| "energy": 1.408408522605896 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.053221566105547846, | |
| "joint_other_seed": 0.1572379040404438, | |
| "single_other_seed": 0.3317354676177363, | |
| "energy": 3101.6552734375 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.044917141136429195, | |
| "joint_other_seed": 0.05435175799051839, | |
| "single_other_seed": 0.046958031505577424, | |
| "energy": 200.66782887776694 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.05330585094823259, | |
| "joint_other_seed": 0.17372439755040614, | |
| "single_other_seed": 0.11405181310116108, | |
| "energy": 7.039542992909749 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.026884431657743235, | |
| "joint_other_seed": 0.017661146777662062, | |
| "single_other_seed": 0.1465730484204707, | |
| "energy": 1.1304617722829182 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.10125931071139094, | |
| "joint_other_seed": 0.0826898852475696, | |
| "single_other_seed": 0.2132464968480683, | |
| "energy": 1007.2298380533854 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.07412375888474376, | |
| "joint_other_seed": 0.1562357869238652, | |
| "single_other_seed": 0.2348274117132253, | |
| "energy": 917.5702819824219 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.009825642360095754, | |
| "joint_other_seed": 0.04253759411250946, | |
| "single_other_seed": 0.06416346206832779, | |
| "energy": 2.213597059249878 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.13623733487339384, | |
| "joint_other_seed": 0.46383834147466296, | |
| "single_other_seed": 0.3257371207656048, | |
| "energy": 621.84716796875 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.04009485170748017, | |
| "joint_other_seed": 0.12545716808123175, | |
| "single_other_seed": 0.10972062519563952, | |
| "energy": 127.4470698038737 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.006111496857449161, | |
| "joint_other_seed": 0.01310148123123056, | |
| "single_other_seed": 0.01076623488390005, | |
| "energy": 6.258768876393636 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.026601628568984292, | |
| "joint_other_seed": 0.01094414399040633, | |
| "single_other_seed": 0.12252833888459036, | |
| "energy": 1.0562293529510498 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.08447124446137688, | |
| "joint_other_seed": 0.10297950578568771, | |
| "single_other_seed": 0.30826957926228105, | |
| "energy": 1165.8702799479167 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.08349567161583538, | |
| "joint_other_seed": 0.09420094342231611, | |
| "single_other_seed": 0.1771233688062973, | |
| "energy": 1306.5553385416667 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.001303436177478523, | |
| "joint_other_seed": 0.006850052637976223, | |
| "single_other_seed": 0.011064333726355069, | |
| "energy": 2.7000145514806113 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.05544308280678367, | |
| "joint_other_seed": 0.3176865128962727, | |
| "single_other_seed": 0.3262937511652213, | |
| "energy": 2543.838134765625 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.12865245939768, | |
| "joint_other_seed": 0.26609997905016214, | |
| "single_other_seed": 0.0827586578851297, | |
| "energy": 709.4117228190104 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.032531677303624607, | |
| "joint_other_seed": 0.1866218800217584, | |
| "single_other_seed": 0.1912205301181017, | |
| "energy": 18.373641967773438 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.04457565706273958, | |
| "joint_other_seed": 0.1843424909854754, | |
| "single_other_seed": 0.20058234535111552, | |
| "energy": 2.7851791381835938 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.12286546494719464, | |
| "joint_other_seed": 0.21847462781686, | |
| "single_other_seed": 0.1736529479210811, | |
| "energy": 507.6237080891927 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.1511941873524133, | |
| "joint_other_seed": 0.45801667077722896, | |
| "single_other_seed": 0.31508037284667034, | |
| "energy": 1115.4426676432292 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.0020604431443869142, | |
| "joint_other_seed": 0.009821738745344755, | |
| "single_other_seed": 0.03959086654286714, | |
| "energy": 4.568957328796387 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.05327508811550408, | |
| "joint_other_seed": 0.20551618083855028, | |
| "single_other_seed": 0.551340363870114, | |
| "energy": 6758.125162760417 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.17958122125059084, | |
| "joint_other_seed": 0.4865962910785736, | |
| "single_other_seed": 0.4514000502020741, | |
| "energy": 986.3439737955729 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.020613204620685055, | |
| "joint_other_seed": 0.004503412036544918, | |
| "single_other_seed": 0.1530421958698822, | |
| "energy": 27.289644241333008 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.02908878885469457, | |
| "joint_other_seed": 0.01449958850774713, | |
| "single_other_seed": 0.5928075335238887, | |
| "energy": 5.7441708246866865 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.10637711077252665, | |
| "joint_other_seed": 0.14002795506421095, | |
| "single_other_seed": 0.2800391159234179, | |
| "energy": 352.26588948567706 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.06478642507222199, | |
| "joint_other_seed": 0.25679329254562633, | |
| "single_other_seed": 0.1530719336310502, | |
| "energy": 339.365966796875 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.001939438106235425, | |
| "joint_other_seed": 0.000915126984955883, | |
| "single_other_seed": 0.05580152779456893, | |
| "energy": 5.429978688557942 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.010693457982618892, | |
| "joint_other_seed": 0.15633095844634393, | |
| "single_other_seed": 0.04681202784501459, | |
| "energy": 1985.257568359375 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.0015054176260415053, | |
| "joint_other_seed": 0.0005002270118624752, | |
| "single_other_seed": 0.009026070276891618, | |
| "energy": 118.12060546875 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.0061641962801997074, | |
| "joint_other_seed": 0.04562607895353654, | |
| "single_other_seed": 0.10113378852231658, | |
| "energy": 9.118872165679932 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.017484815615677052, | |
| "joint_other_seed": 0.1279709110235906, | |
| "single_other_seed": 0.07758205212436686, | |
| "energy": 4.268757979075114 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.04982470212301637, | |
| "joint_other_seed": 0.030094572364743444, | |
| "single_other_seed": 0.13996706298494344, | |
| "energy": 363.9580078125 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.07389550046862892, | |
| "joint_other_seed": 0.0937157665008258, | |
| "single_other_seed": 0.2947041378316383, | |
| "energy": 296.01833089192706 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.007576939948085726, | |
| "joint_other_seed": 0.04492397282536997, | |
| "single_other_seed": 0.06999139934415854, | |
| "energy": 7.901806672414144 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.20924256505972783, | |
| "joint_other_seed": 0.4822984165754606, | |
| "single_other_seed": 0.4036124417826097, | |
| "energy": 2002.469970703125 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.01089061716473061, | |
| "joint_other_seed": 0.041165413926182226, | |
| "single_other_seed": 0.08649947121915735, | |
| "energy": 180.32162475585938 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.0025760661871601087, | |
| "joint_other_seed": 0.013437399471888933, | |
| "single_other_seed": 0.030532612694742278, | |
| "energy": 13.481903394063314 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.010258952254022624, | |
| "joint_other_seed": 0.012873484256370582, | |
| "single_other_seed": 0.07188525528755918, | |
| "energy": 3.744297663370768 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.1260403423805695, | |
| "joint_other_seed": 0.380336323579484, | |
| "single_other_seed": 0.28723323308988, | |
| "energy": 379.96303304036456 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.04805821214715145, | |
| "joint_other_seed": 0.14388179830553305, | |
| "single_other_seed": 0.1490021643779822, | |
| "energy": 559.2771199544271 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.008198476167609825, | |
| "joint_other_seed": 0.055902886810462404, | |
| "single_other_seed": 0.06463263680347098, | |
| "energy": 12.493381182352701 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.20051679978688064, | |
| "joint_other_seed": 0.30948216867287326, | |
| "single_other_seed": 0.43487974602703017, | |
| "energy": 1734.8336995442708 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.03433853723785951, | |
| "joint_other_seed": 0.26169925512368375, | |
| "single_other_seed": 0.18651738518291217, | |
| "energy": 179.3256632486979 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.018676151042781308, | |
| "joint_other_seed": 0.1126941743099068, | |
| "single_other_seed": 0.04615714051364461, | |
| "energy": 24.835484822591145 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.05834056533559162, | |
| "joint_other_seed": 0.07343957104112964, | |
| "single_other_seed": 0.07374310973503384, | |
| "energy": 11.468725681304932 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.14420631881165513, | |
| "joint_other_seed": 0.3654549784517943, | |
| "single_other_seed": 0.4077134046114984, | |
| "energy": 385.7596740722656 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.15167255077821584, | |
| "joint_other_seed": 0.4339874284755616, | |
| "single_other_seed": 0.3965024777743623, | |
| "energy": 402.4934997558594 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.011870275563466204, | |
| "joint_other_seed": 0.07361605921074554, | |
| "single_other_seed": 0.04688538810397269, | |
| "energy": 19.86478106180827 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.02486378211825424, | |
| "joint_other_seed": 0.08089264201528867, | |
| "single_other_seed": 0.20198374013067566, | |
| "energy": 1601.6787516276042 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.04334226473858949, | |
| "joint_other_seed": 0.14487090105669187, | |
| "single_other_seed": 0.03613248878709541, | |
| "energy": 251.33082071940103 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.050518070958856964, | |
| "joint_other_seed": 0.2173891607755848, | |
| "single_other_seed": 0.15508192641958551, | |
| "energy": 46.29534403483073 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.13835511885329943, | |
| "joint_other_seed": 0.21139404321705957, | |
| "single_other_seed": 0.3202603229766112, | |
| "energy": 19.215836842854817 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.12171616309659983, | |
| "joint_other_seed": 0.41848448610193834, | |
| "single_other_seed": 0.2967436681277361, | |
| "energy": 489.45128377278644 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.11586620529538627, | |
| "joint_other_seed": 0.16339621930521753, | |
| "single_other_seed": 0.27471015323338144, | |
| "energy": 498.3363444010417 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.02711688822691216, | |
| "joint_other_seed": 0.08511837209853912, | |
| "single_other_seed": 0.1039725488415284, | |
| "energy": 47.66041946411133 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.03411171624321996, | |
| "joint_other_seed": 0.20427425336519367, | |
| "single_other_seed": 0.20660687446625292, | |
| "energy": 2575.0577799479165 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.012981853860886905, | |
| "joint_other_seed": 0.047179648417432274, | |
| "single_other_seed": 0.07161676143065276, | |
| "energy": 254.90038045247397 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.19961960711145466, | |
| "joint_other_seed": 0.48629435400299137, | |
| "single_other_seed": 0.39401552851159805, | |
| "energy": 87.07045491536458 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.3333274137003984, | |
| "joint_other_seed": 0.6327077071159841, | |
| "single_other_seed": 0.6500717756499006, | |
| "energy": 76.47052001953125 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.14467206919709086, | |
| "joint_other_seed": 0.3194616609445129, | |
| "single_other_seed": 0.4395782730099329, | |
| "energy": 487.90191650390625 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.08223739444159106, | |
| "joint_other_seed": 0.07731544305042037, | |
| "single_other_seed": 0.4665777826857431, | |
| "energy": 617.463134765625 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.30555038868447937, | |
| "joint_other_seed": 0.43782543527943707, | |
| "single_other_seed": 0.4487417294578403, | |
| "energy": 155.6717732747396 | |
| } | |
| }, | |
| "learned/trigger/r3": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.21000159578007827, | |
| "joint_other_seed": 0.6097259095679429, | |
| "single_other_seed": 0.4643610002664895, | |
| "energy": 29868.000651041668 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.38955380555116736, | |
| "joint_other_seed": 0.8036573968922973, | |
| "single_other_seed": 0.7078430244434457, | |
| "energy": 2556.13818359375 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.18488334105468318, | |
| "joint_other_seed": 0.7780399040688851, | |
| "single_other_seed": 0.6254937739311894, | |
| "energy": 159.1924031575521 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.34500194331827516, | |
| "joint_other_seed": 0.6879489453345276, | |
| "single_other_seed": 0.7068658729525884, | |
| "energy": 2.083418289820353 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.1576642488225376, | |
| "joint_other_seed": 0.46330343734480384, | |
| "single_other_seed": 0.5280085196131685, | |
| "energy": 2883.2149251302085 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.0902911074284567, | |
| "joint_other_seed": 0.2806061357727412, | |
| "single_other_seed": 0.10965674975114792, | |
| "energy": 1530.3282063802083 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.10526875559318949, | |
| "joint_other_seed": 0.6064907582459371, | |
| "single_other_seed": 0.21501616339504448, | |
| "energy": 7.896752834320068 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.333864663797965, | |
| "joint_other_seed": 0.5678273995993592, | |
| "single_other_seed": 0.5239160353421755, | |
| "energy": 2508.7984212239585 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.11639757116990124, | |
| "joint_other_seed": 0.6460491180630923, | |
| "single_other_seed": 0.42768268943738197, | |
| "energy": 1070.7188110351562 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.3016438963193001, | |
| "joint_other_seed": 0.5316951434695529, | |
| "single_other_seed": 0.6246131809184639, | |
| "energy": 34.32500394185384 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.16536807517580787, | |
| "joint_other_seed": 0.39065282494027287, | |
| "single_other_seed": 0.5317400852723304, | |
| "energy": 1.2600523034731548 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.05975117431457211, | |
| "joint_other_seed": 0.06940310915569681, | |
| "single_other_seed": 0.10112784213831483, | |
| "energy": 750.6941731770834 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.0732270458897557, | |
| "joint_other_seed": 0.4111415114697159, | |
| "single_other_seed": 0.24557368809346902, | |
| "energy": 1100.1739705403645 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.1074569206345124, | |
| "joint_other_seed": 0.2375148917333222, | |
| "single_other_seed": 0.31293591565942924, | |
| "energy": 1.5572519302368164 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.2303857636006259, | |
| "joint_other_seed": 0.4954255558190246, | |
| "single_other_seed": 0.312575398497833, | |
| "energy": 3013.2926432291665 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.11495996297989486, | |
| "joint_other_seed": 0.49662402272814266, | |
| "single_other_seed": 0.17627260980388315, | |
| "energy": 236.5339152018229 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.021928585888026375, | |
| "joint_other_seed": 0.1993612075308947, | |
| "single_other_seed": 0.36344939213391136, | |
| "energy": 19.824208577473957 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.07011946126197166, | |
| "joint_other_seed": 0.16880474480794827, | |
| "single_other_seed": 0.26265183763983824, | |
| "energy": 1.1811367670694988 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.04835111895209687, | |
| "joint_other_seed": 0.46975507612190726, | |
| "single_other_seed": 0.405295006479894, | |
| "energy": 1664.6710611979167 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.08933807244212211, | |
| "joint_other_seed": 0.3753368012744872, | |
| "single_other_seed": 0.34238913790364267, | |
| "energy": 1465.9949951171875 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.01182527595459526, | |
| "joint_other_seed": 0.09932971701134263, | |
| "single_other_seed": 0.14465841866469964, | |
| "energy": 2.557423988978068 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.12550589536467727, | |
| "joint_other_seed": 0.5080341591036066, | |
| "single_other_seed": 0.3317354676177363, | |
| "energy": 2775.3213704427085 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.05892289638374948, | |
| "joint_other_seed": 0.1546236697076787, | |
| "single_other_seed": 0.046958031505577424, | |
| "energy": 207.76212056477866 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.03900158754480687, | |
| "joint_other_seed": 0.11822528331135485, | |
| "single_other_seed": 0.11405181310116108, | |
| "energy": 13.570127169291178 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.02039522191355836, | |
| "joint_other_seed": 0.15954022035715054, | |
| "single_other_seed": 0.1465730484204707, | |
| "energy": 2.1993558009465537 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.0605854111376225, | |
| "joint_other_seed": 0.22721813979557393, | |
| "single_other_seed": 0.2132464968480683, | |
| "energy": 1345.0263264973958 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.10115686921079903, | |
| "joint_other_seed": 0.3151229711380513, | |
| "single_other_seed": 0.2348274117132253, | |
| "energy": 1335.7676188151042 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.010131513725027408, | |
| "joint_other_seed": 0.07515093085897917, | |
| "single_other_seed": 0.06416346206832779, | |
| "energy": 4.352579434712728 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.0948569270035966, | |
| "joint_other_seed": 0.3359924605494906, | |
| "single_other_seed": 0.3257371207656048, | |
| "energy": 1208.2819417317708 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.049957441903392796, | |
| "joint_other_seed": 0.20282992570431047, | |
| "single_other_seed": 0.10972062519563952, | |
| "energy": 175.43970235188803 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.010733125419850609, | |
| "joint_other_seed": 0.023529654188478744, | |
| "single_other_seed": 0.01076623488390005, | |
| "energy": 13.390735308329264 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.020088085626187246, | |
| "joint_other_seed": 0.05962480407997384, | |
| "single_other_seed": 0.12252833888459036, | |
| "energy": 1.80722443262736 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.03933820283440648, | |
| "joint_other_seed": 0.17583844586982034, | |
| "single_other_seed": 0.30826957926228105, | |
| "energy": 1160.0012613932292 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.03748600377444666, | |
| "joint_other_seed": 0.2928724806176095, | |
| "single_other_seed": 0.1771233688062973, | |
| "energy": 1728.5965576171875 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.004167951310677376, | |
| "joint_other_seed": 0.08243113270684721, | |
| "single_other_seed": 0.011064333726355069, | |
| "energy": 6.115744431813558 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.05397284020089166, | |
| "joint_other_seed": 0.4479126857204654, | |
| "single_other_seed": 0.3262937511652213, | |
| "energy": 2263.5730794270835 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.14941494175787184, | |
| "joint_other_seed": 0.4314022584409334, | |
| "single_other_seed": 0.0827586578851297, | |
| "energy": 659.6617635091146 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.020702654874375948, | |
| "joint_other_seed": 0.08550539295047199, | |
| "single_other_seed": 0.1912205301181017, | |
| "energy": 24.65747324625651 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.04299532296431263, | |
| "joint_other_seed": 0.2979250525640794, | |
| "single_other_seed": 0.20058234535111552, | |
| "energy": 4.455713748931885 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.057668750439398786, | |
| "joint_other_seed": 0.23512673151737618, | |
| "single_other_seed": 0.1736529479210811, | |
| "energy": 1280.2508138020833 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.05557967853298995, | |
| "joint_other_seed": 0.3086260429081569, | |
| "single_other_seed": 0.31508037284667034, | |
| "energy": 1442.3824869791667 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.004300476415965204, | |
| "joint_other_seed": 0.0881139664653725, | |
| "single_other_seed": 0.03959086654286714, | |
| "energy": 8.670859018961588 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.10891357402923296, | |
| "joint_other_seed": 0.6002836928151312, | |
| "single_other_seed": 0.551340363870114, | |
| "energy": 5466.15185546875 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.19468454200003094, | |
| "joint_other_seed": 0.7348265390951169, | |
| "single_other_seed": 0.4514000502020741, | |
| "energy": 1079.6573486328125 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.0202198232536539, | |
| "joint_other_seed": 0.12599606125102003, | |
| "single_other_seed": 0.1530421958698822, | |
| "energy": 31.191019694010418 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.16747866817113233, | |
| "joint_other_seed": 0.41869250937224545, | |
| "single_other_seed": 0.5928075335238887, | |
| "energy": 12.318988164265951 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.04268748715705406, | |
| "joint_other_seed": 0.10580390306433929, | |
| "single_other_seed": 0.2800391159234179, | |
| "energy": 922.2677612304688 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.04169754481085575, | |
| "joint_other_seed": 0.16087524656682262, | |
| "single_other_seed": 0.1530719336310502, | |
| "energy": 837.8507690429688 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.0056537763490086546, | |
| "joint_other_seed": 0.08009317447743047, | |
| "single_other_seed": 0.05580152779456893, | |
| "energy": 11.711844762166342 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.015471057832608048, | |
| "joint_other_seed": 0.08756621556671995, | |
| "single_other_seed": 0.04681202784501459, | |
| "energy": 1887.92724609375 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.005092810579708784, | |
| "joint_other_seed": 0.016260053968536914, | |
| "single_other_seed": 0.009026070276891618, | |
| "energy": 136.46482849121094 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.008775612936625321, | |
| "joint_other_seed": 0.02294216032475882, | |
| "single_other_seed": 0.10113378852231658, | |
| "energy": 14.780303637186686 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.016486457381866382, | |
| "joint_other_seed": 0.07308489350221548, | |
| "single_other_seed": 0.07758205212436686, | |
| "energy": 6.622793992360433 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.03867792365431888, | |
| "joint_other_seed": 0.0814505998679883, | |
| "single_other_seed": 0.13996706298494344, | |
| "energy": 820.757568359375 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.040657755482748525, | |
| "joint_other_seed": 0.13564409980340442, | |
| "single_other_seed": 0.2947041378316383, | |
| "energy": 780.6941528320312 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.004991602311054444, | |
| "joint_other_seed": 0.048271483696735173, | |
| "single_other_seed": 0.06999139934415854, | |
| "energy": 14.065078417460123 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.1638324385295521, | |
| "joint_other_seed": 0.3250461522620929, | |
| "single_other_seed": 0.4036124417826097, | |
| "energy": 1860.5442708333333 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.009624468640162975, | |
| "joint_other_seed": 0.040576014694001294, | |
| "single_other_seed": 0.08649947121915735, | |
| "energy": 200.06658426920572 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.009460169103147398, | |
| "joint_other_seed": 0.03812198457301619, | |
| "single_other_seed": 0.030532612694742278, | |
| "energy": 29.00717608133952 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.01241950443374118, | |
| "joint_other_seed": 0.1116582468455336, | |
| "single_other_seed": 0.07188525528755918, | |
| "energy": 8.492513020833334 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.057194153961437455, | |
| "joint_other_seed": 0.1531976163237343, | |
| "single_other_seed": 0.28723323308988, | |
| "energy": 1008.2615559895834 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.01838684606630092, | |
| "joint_other_seed": 0.16156943186052464, | |
| "single_other_seed": 0.1490021643779822, | |
| "energy": 1015.2965291341146 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.0038078752768707647, | |
| "joint_other_seed": 0.06625270248690962, | |
| "single_other_seed": 0.06463263680347098, | |
| "energy": 23.522315979003906 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.1059780430002949, | |
| "joint_other_seed": 0.18206970617427443, | |
| "single_other_seed": 0.43487974602703017, | |
| "energy": 1570.2373860677083 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.026033854847100746, | |
| "joint_other_seed": 0.15428037783586313, | |
| "single_other_seed": 0.18651738518291217, | |
| "energy": 240.32990519205728 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.021020909910545837, | |
| "joint_other_seed": 0.16186419688422504, | |
| "single_other_seed": 0.04615714051364461, | |
| "energy": 39.81398264567057 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.07667130104795661, | |
| "joint_other_seed": 0.23785451701453386, | |
| "single_other_seed": 0.07374310973503384, | |
| "energy": 15.164565722147623 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.08155099169834823, | |
| "joint_other_seed": 0.13367060531476904, | |
| "single_other_seed": 0.4077134046114984, | |
| "energy": 841.4068603515625 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.0699106213366345, | |
| "joint_other_seed": 0.11900723698934679, | |
| "single_other_seed": 0.3965024777743623, | |
| "energy": 850.7741495768229 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.022160143561786416, | |
| "joint_other_seed": 0.12471398430880531, | |
| "single_other_seed": 0.04688538810397269, | |
| "energy": 35.68913650512695 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.06272199491222319, | |
| "joint_other_seed": 0.2486675009455761, | |
| "single_other_seed": 0.20198374013067566, | |
| "energy": 2228.2008463541665 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.03017312566566499, | |
| "joint_other_seed": 0.162952572994025, | |
| "single_other_seed": 0.03613248878709541, | |
| "energy": 252.8126678466797 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.046271183277342254, | |
| "joint_other_seed": 0.29775088652482273, | |
| "single_other_seed": 0.15508192641958551, | |
| "energy": 54.50439453125 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.15462844357868588, | |
| "joint_other_seed": 0.34132893951438714, | |
| "single_other_seed": 0.3202603229766112, | |
| "energy": 27.402589797973633 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.08256750957240011, | |
| "joint_other_seed": 0.19073761648418752, | |
| "single_other_seed": 0.2967436681277361, | |
| "energy": 950.5420125325521 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.06260286625384626, | |
| "joint_other_seed": 0.1395199482639562, | |
| "single_other_seed": 0.27471015323338144, | |
| "energy": 993.1850179036459 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.07848624851962083, | |
| "joint_other_seed": 0.26376946861467576, | |
| "single_other_seed": 0.1039725488415284, | |
| "energy": 87.08853149414062 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.03577331628079761, | |
| "joint_other_seed": 0.17812488768484047, | |
| "single_other_seed": 0.20660687446625292, | |
| "energy": 2114.8390299479165 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.014977182295955349, | |
| "joint_other_seed": 0.0353030644307204, | |
| "single_other_seed": 0.07161676143065276, | |
| "energy": 283.75599161783856 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.17325366631408723, | |
| "joint_other_seed": 0.4596041525775685, | |
| "single_other_seed": 0.39401552851159805, | |
| "energy": 84.00181579589844 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.37567671167824135, | |
| "joint_other_seed": 0.6477898476411447, | |
| "single_other_seed": 0.6500717756499006, | |
| "energy": 84.69896697998047 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.11248549666112462, | |
| "joint_other_seed": 0.24551337162162556, | |
| "single_other_seed": 0.4395782730099329, | |
| "energy": 1025.452657063802 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.09664074184830151, | |
| "joint_other_seed": 0.41568964953873433, | |
| "single_other_seed": 0.4665777826857431, | |
| "energy": 1082.1524658203125 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.27902333572338045, | |
| "joint_other_seed": 0.5064579616341889, | |
| "single_other_seed": 0.4487417294578403, | |
| "energy": 185.43567403157553 | |
| } | |
| }, | |
| "learned/trigger/r16": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.2393509358355982, | |
| "joint_other_seed": 0.7439590151431529, | |
| "single_other_seed": 0.4643610002664895, | |
| "energy": 23568.6484375 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.44194432535929773, | |
| "joint_other_seed": 0.8364452003702744, | |
| "single_other_seed": 0.7078430244434457, | |
| "energy": 1728.377685546875 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.21964708602703087, | |
| "joint_other_seed": 0.8010329324858041, | |
| "single_other_seed": 0.6254937739311894, | |
| "energy": 154.83389790852866 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.35325376769870853, | |
| "joint_other_seed": 0.8006646627692904, | |
| "single_other_seed": 0.7068658729525884, | |
| "energy": 2.269587198893229 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.15882735970489126, | |
| "joint_other_seed": 0.6056799095464345, | |
| "single_other_seed": 0.5280085196131685, | |
| "energy": 1986.732421875 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.0977528714062662, | |
| "joint_other_seed": 0.584220055187216, | |
| "single_other_seed": 0.10965674975114792, | |
| "energy": 1813.6064453125 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.10244629864164312, | |
| "joint_other_seed": 0.6301900595170239, | |
| "single_other_seed": 0.21501616339504448, | |
| "energy": 4.8827511469523115 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.4014623445039557, | |
| "joint_other_seed": 0.6934833271224246, | |
| "single_other_seed": 0.5239160353421755, | |
| "energy": 2187.9468587239585 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.12569551891851272, | |
| "joint_other_seed": 0.7929309112068177, | |
| "single_other_seed": 0.42768268943738197, | |
| "energy": 981.14306640625 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.27643458029296264, | |
| "joint_other_seed": 0.6455880279417715, | |
| "single_other_seed": 0.6246131809184639, | |
| "energy": 27.72201665242513 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.16507843945130807, | |
| "joint_other_seed": 0.5747551665906652, | |
| "single_other_seed": 0.5317400852723304, | |
| "energy": 1.896855354309082 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.04597715479588893, | |
| "joint_other_seed": 0.5606880605431216, | |
| "single_other_seed": 0.10112784213831483, | |
| "energy": 1224.7264607747395 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.08010082541240204, | |
| "joint_other_seed": 0.5199316432784706, | |
| "single_other_seed": 0.24557368809346902, | |
| "energy": 892.9751383463541 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.11809337197419034, | |
| "joint_other_seed": 0.42934893924411877, | |
| "single_other_seed": 0.31293591565942924, | |
| "energy": 1.5584479570388794 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.23460078262702388, | |
| "joint_other_seed": 0.7387858743739553, | |
| "single_other_seed": 0.312575398497833, | |
| "energy": 2981.848388671875 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.12627030424846986, | |
| "joint_other_seed": 0.6104354299009137, | |
| "single_other_seed": 0.17627260980388315, | |
| "energy": 168.41558329264322 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.01495345706585749, | |
| "joint_other_seed": 0.37438145073243123, | |
| "single_other_seed": 0.36344939213391136, | |
| "energy": 21.59408187866211 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.05042192669844925, | |
| "joint_other_seed": 0.3246996614158898, | |
| "single_other_seed": 0.26265183763983824, | |
| "energy": 2.0501720905303955 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.04332477652373701, | |
| "joint_other_seed": 0.5372375051418521, | |
| "single_other_seed": 0.405295006479894, | |
| "energy": 1773.5168050130208 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.06001649675749754, | |
| "joint_other_seed": 0.5730991464364641, | |
| "single_other_seed": 0.34238913790364267, | |
| "energy": 1682.3855387369792 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.014099233612641426, | |
| "joint_other_seed": 0.24622077177376045, | |
| "single_other_seed": 0.14465841866469964, | |
| "energy": 3.064183473587036 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.13643864362982916, | |
| "joint_other_seed": 0.6907902700639406, | |
| "single_other_seed": 0.3317354676177363, | |
| "energy": 2420.23486328125 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.0769107793632261, | |
| "joint_other_seed": 0.4747734200277881, | |
| "single_other_seed": 0.046958031505577424, | |
| "energy": 175.2070566813151 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.03606433233537609, | |
| "joint_other_seed": 0.17233616725732837, | |
| "single_other_seed": 0.11405181310116108, | |
| "energy": 12.407109896341959 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.01914206487959713, | |
| "joint_other_seed": 0.32344281362110705, | |
| "single_other_seed": 0.1465730484204707, | |
| "energy": 3.496675650278727 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.0651900001007103, | |
| "joint_other_seed": 0.5255851759163763, | |
| "single_other_seed": 0.2132464968480683, | |
| "energy": 1967.6321207682292 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.10931404801320943, | |
| "joint_other_seed": 0.4485936268656441, | |
| "single_other_seed": 0.2348274117132253, | |
| "energy": 1587.4130859375 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.014116237569308687, | |
| "joint_other_seed": 0.24636316595566898, | |
| "single_other_seed": 0.06416346206832779, | |
| "energy": 6.336649099985759 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.09076863796613466, | |
| "joint_other_seed": 0.48781604474417584, | |
| "single_other_seed": 0.3257371207656048, | |
| "energy": 1225.3326009114583 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.06089431612679663, | |
| "joint_other_seed": 0.37691958544626136, | |
| "single_other_seed": 0.10972062519563952, | |
| "energy": 180.00018819173178 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.00966862474113006, | |
| "joint_other_seed": 0.16483535382252468, | |
| "single_other_seed": 0.01076623488390005, | |
| "energy": 19.90766779581706 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.009250317314534517, | |
| "joint_other_seed": 0.18345446204630755, | |
| "single_other_seed": 0.12252833888459036, | |
| "energy": 4.556894779205322 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.01956568691978966, | |
| "joint_other_seed": 0.41870405531773014, | |
| "single_other_seed": 0.30826957926228105, | |
| "energy": 1749.0577799479167 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.024579970248470917, | |
| "joint_other_seed": 0.47322944690138924, | |
| "single_other_seed": 0.1771233688062973, | |
| "energy": 2022.9396158854167 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.0034747852630446774, | |
| "joint_other_seed": 0.2094636713969399, | |
| "single_other_seed": 0.011064333726355069, | |
| "energy": 9.828438440958658 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.06343589611089828, | |
| "joint_other_seed": 0.5709743944402343, | |
| "single_other_seed": 0.3262937511652213, | |
| "energy": 2119.7451985677085 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.1671260480897624, | |
| "joint_other_seed": 0.6430619963053414, | |
| "single_other_seed": 0.0827586578851297, | |
| "energy": 536.8010762532552 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.014310702421926336, | |
| "joint_other_seed": 0.2212164667957438, | |
| "single_other_seed": 0.1912205301181017, | |
| "energy": 27.64642333984375 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.03703252836766335, | |
| "joint_other_seed": 0.3846630754535366, | |
| "single_other_seed": 0.20058234535111552, | |
| "energy": 7.08266003926595 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.03320410384564698, | |
| "joint_other_seed": 0.4075158750689425, | |
| "single_other_seed": 0.1736529479210811, | |
| "energy": 1667.3258463541667 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.0364732827991584, | |
| "joint_other_seed": 0.41305622815486365, | |
| "single_other_seed": 0.31508037284667034, | |
| "energy": 1812.7193196614583 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.006001208451163809, | |
| "joint_other_seed": 0.20373723966409585, | |
| "single_other_seed": 0.03959086654286714, | |
| "energy": 13.742619514465332 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.11842379652603152, | |
| "joint_other_seed": 0.7996157276847955, | |
| "single_other_seed": 0.551340363870114, | |
| "energy": 4716.506022135417 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.20733088805105027, | |
| "joint_other_seed": 0.8092226883746148, | |
| "single_other_seed": 0.4514000502020741, | |
| "energy": 1051.1422932942708 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.01830783710416173, | |
| "joint_other_seed": 0.2448213579756191, | |
| "single_other_seed": 0.1530421958698822, | |
| "energy": 31.96150080362956 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.1807002917008318, | |
| "joint_other_seed": 0.6299811072625567, | |
| "single_other_seed": 0.5928075335238887, | |
| "energy": 17.432868321736652 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.020415730696190737, | |
| "joint_other_seed": 0.32120817603907537, | |
| "single_other_seed": 0.2800391159234179, | |
| "energy": 1719.9827880859375 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.019985937785709118, | |
| "joint_other_seed": 0.25879701499257535, | |
| "single_other_seed": 0.1530719336310502, | |
| "energy": 1552.5778401692708 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.005644074710979052, | |
| "joint_other_seed": 0.18819235115008548, | |
| "single_other_seed": 0.05580152779456893, | |
| "energy": 19.00423812866211 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.022775088092975126, | |
| "joint_other_seed": 0.26392929348824445, | |
| "single_other_seed": 0.04681202784501459, | |
| "energy": 1886.1090087890625 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.004573067299856182, | |
| "joint_other_seed": 0.0672987687511497, | |
| "single_other_seed": 0.009026070276891618, | |
| "energy": 106.40843963623047 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.01372183690089579, | |
| "joint_other_seed": 0.14622076572287968, | |
| "single_other_seed": 0.10113378852231658, | |
| "energy": 21.186272939046223 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.012151373567195423, | |
| "joint_other_seed": 0.17190910236237322, | |
| "single_other_seed": 0.07758205212436686, | |
| "energy": 12.7195831934611 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.017610026999029027, | |
| "joint_other_seed": 0.24742938509076595, | |
| "single_other_seed": 0.13996706298494344, | |
| "energy": 1554.2671712239583 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.021126236909840168, | |
| "joint_other_seed": 0.26532228533878244, | |
| "single_other_seed": 0.2947041378316383, | |
| "energy": 1449.5672607421875 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.006238984889379906, | |
| "joint_other_seed": 0.13867488295035424, | |
| "single_other_seed": 0.06999139934415854, | |
| "energy": 25.688992182413738 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.14151113445544683, | |
| "joint_other_seed": 0.4207134853862965, | |
| "single_other_seed": 0.4036124417826097, | |
| "energy": 1770.9283447265625 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.012486510306161658, | |
| "joint_other_seed": 0.20918881645472542, | |
| "single_other_seed": 0.08649947121915735, | |
| "energy": 193.29840596516928 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.012099114321260984, | |
| "joint_other_seed": 0.17468109009281008, | |
| "single_other_seed": 0.030532612694742278, | |
| "energy": 33.30140813191732 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.011886970873026925, | |
| "joint_other_seed": 0.20703976705343374, | |
| "single_other_seed": 0.07188525528755918, | |
| "energy": 18.07653299967448 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.027495916538165188, | |
| "joint_other_seed": 0.3199986118581778, | |
| "single_other_seed": 0.28723323308988, | |
| "energy": 1799.2099609375 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.014505832240918615, | |
| "joint_other_seed": 0.27239003062898315, | |
| "single_other_seed": 0.1490021643779822, | |
| "energy": 1613.018310546875 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.004734464905797209, | |
| "joint_other_seed": 0.18782993654718272, | |
| "single_other_seed": 0.06463263680347098, | |
| "energy": 40.42357508341471 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.1327598375703538, | |
| "joint_other_seed": 0.4186338360157127, | |
| "single_other_seed": 0.43487974602703017, | |
| "energy": 1510.8804931640625 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.02970644269640399, | |
| "joint_other_seed": 0.31383201908926883, | |
| "single_other_seed": 0.18651738518291217, | |
| "energy": 208.9108683268229 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.018688075507759083, | |
| "joint_other_seed": 0.244309275914261, | |
| "single_other_seed": 0.04615714051364461, | |
| "energy": 38.33260599772135 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.07060494199351075, | |
| "joint_other_seed": 0.36963398022876837, | |
| "single_other_seed": 0.07374310973503384, | |
| "energy": 26.46473503112793 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.04414805075634698, | |
| "joint_other_seed": 0.30100040670581396, | |
| "single_other_seed": 0.4077134046114984, | |
| "energy": 1571.3539225260417 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.033820812078059426, | |
| "joint_other_seed": 0.2748984333661382, | |
| "single_other_seed": 0.3965024777743623, | |
| "energy": 1386.1114501953125 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.024152169460693207, | |
| "joint_other_seed": 0.26946584714664734, | |
| "single_other_seed": 0.04688538810397269, | |
| "energy": 56.59733581542969 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.0620360995477276, | |
| "joint_other_seed": 0.4853895931753428, | |
| "single_other_seed": 0.20198374013067566, | |
| "energy": 2325.3705240885415 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.03251052073236815, | |
| "joint_other_seed": 0.24364820482292449, | |
| "single_other_seed": 0.03613248878709541, | |
| "energy": 249.91416931152344 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.04232155113265579, | |
| "joint_other_seed": 0.4043307052186995, | |
| "single_other_seed": 0.15508192641958551, | |
| "energy": 56.85654067993164 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.14594130749619816, | |
| "joint_other_seed": 0.519336182791877, | |
| "single_other_seed": 0.3202603229766112, | |
| "energy": 36.489715576171875 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.04462741286976357, | |
| "joint_other_seed": 0.338725130742208, | |
| "single_other_seed": 0.2967436681277361, | |
| "energy": 1610.9884440104167 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.030409858750321983, | |
| "joint_other_seed": 0.32274021111870244, | |
| "single_other_seed": 0.27471015323338144, | |
| "energy": 1578.4742838541667 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.07135756812628369, | |
| "joint_other_seed": 0.45148643717854053, | |
| "single_other_seed": 0.1039725488415284, | |
| "energy": 98.24627176920573 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.045355262735821134, | |
| "joint_other_seed": 0.3424016407187721, | |
| "single_other_seed": 0.20660687446625292, | |
| "energy": 2184.3394368489585 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.027308853262417165, | |
| "joint_other_seed": 0.21965292655062674, | |
| "single_other_seed": 0.07161676143065276, | |
| "energy": 262.0708821614583 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.21239931596228523, | |
| "joint_other_seed": 0.6471667691398317, | |
| "single_other_seed": 0.39401552851159805, | |
| "energy": 83.4324213663737 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.4281700936658449, | |
| "joint_other_seed": 0.7486840384203224, | |
| "single_other_seed": 0.6500717756499006, | |
| "energy": 92.89778900146484 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.07939247399901717, | |
| "joint_other_seed": 0.4617982926958716, | |
| "single_other_seed": 0.4395782730099329, | |
| "energy": 1699.7139892578125 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.09304284240828398, | |
| "joint_other_seed": 0.46174343683573305, | |
| "single_other_seed": 0.4665777826857431, | |
| "energy": 1456.8698323567708 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.234730844624238, | |
| "joint_other_seed": 0.6436722057439155, | |
| "single_other_seed": 0.4487417294578403, | |
| "energy": 209.0805409749349 | |
| } | |
| }, | |
| "learned/all/r1": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.15338855590339628, | |
| "joint_other_seed": 0.241956020252988, | |
| "single_other_seed": 0.4483871063989089, | |
| "energy": 39881.4609375 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.29132394547746066, | |
| "joint_other_seed": 0.5502201949220555, | |
| "single_other_seed": 0.6793299562745521, | |
| "energy": 5270.469401041667 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.16103725355240817, | |
| "joint_other_seed": 0.5214372029570814, | |
| "single_other_seed": 0.5934555645300559, | |
| "energy": 230.8532257080078 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.311549283140661, | |
| "joint_other_seed": 0.6194313666633136, | |
| "single_other_seed": 0.6617750267175087, | |
| "energy": 4.041804949442546 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.12375039804008559, | |
| "joint_other_seed": 0.19886659188004874, | |
| "single_other_seed": 0.5045979891394741, | |
| "energy": 4037.1660970052085 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.0843353435970362, | |
| "joint_other_seed": 0.3236487455857382, | |
| "single_other_seed": 0.07554163319658491, | |
| "energy": 3494.484090169271 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.050206025966444634, | |
| "joint_other_seed": 0.14273091352666567, | |
| "single_other_seed": 0.2536833687673502, | |
| "energy": 12.799226760864258 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.3605183259149306, | |
| "joint_other_seed": 0.6511356122618523, | |
| "single_other_seed": 0.5587965946862371, | |
| "energy": 7320.42041015625 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.07595870428904916, | |
| "joint_other_seed": 0.12013082596946167, | |
| "single_other_seed": 0.39898832270843276, | |
| "energy": 1362.6550699869792 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.283344759220845, | |
| "joint_other_seed": 0.43917237235530016, | |
| "single_other_seed": 0.5208452433709779, | |
| "energy": 51.2648811340332 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.24227766230549486, | |
| "joint_other_seed": 0.4422010681141717, | |
| "single_other_seed": 0.5249676105685213, | |
| "energy": 2.5328447818756104 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.08215505792401923, | |
| "joint_other_seed": 0.20963071763337526, | |
| "single_other_seed": 0.10464691448035823, | |
| "energy": 769.2807820638021 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.15531680658466662, | |
| "joint_other_seed": 0.4271020300726387, | |
| "single_other_seed": 0.2412835961441585, | |
| "energy": 994.2952677408854 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.10127501493535229, | |
| "joint_other_seed": 0.31990608484516736, | |
| "single_other_seed": 0.3269644458557466, | |
| "energy": 2.442063291867574 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.21790830170599162, | |
| "joint_other_seed": 0.33902633308141694, | |
| "single_other_seed": 0.3061740426306926, | |
| "energy": 5807.264811197917 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.07837961321868603, | |
| "joint_other_seed": 0.11135607143701443, | |
| "single_other_seed": 0.15188466412747137, | |
| "energy": 639.8643900553385 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.04906406046841927, | |
| "joint_other_seed": 0.2037487406119064, | |
| "single_other_seed": 0.36124135372946575, | |
| "energy": 18.011241912841797 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.10024657963745232, | |
| "joint_other_seed": 0.12534982792859084, | |
| "single_other_seed": 0.26936184836528954, | |
| "energy": 1.7877552111943562 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.11079443095238417, | |
| "joint_other_seed": 0.2017235065145221, | |
| "single_other_seed": 0.3838742773934516, | |
| "energy": 1801.780985514323 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.15560046482012468, | |
| "joint_other_seed": 0.3927487829644464, | |
| "single_other_seed": 0.31676711116843315, | |
| "energy": 1762.1693929036458 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.022380738545650813, | |
| "joint_other_seed": 0.17468542794584663, | |
| "single_other_seed": 0.1492909954215249, | |
| "energy": 3.4193987051645913 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.03834175951633301, | |
| "joint_other_seed": 0.11746084838076787, | |
| "single_other_seed": 0.29262605440507505, | |
| "energy": 5134.925537109375 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.05209167764573275, | |
| "joint_other_seed": 0.06716997810510716, | |
| "single_other_seed": 0.039280669717608796, | |
| "energy": 465.41566975911456 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.05174956204502246, | |
| "joint_other_seed": 0.17411378283953582, | |
| "single_other_seed": 0.12560850942846002, | |
| "energy": 18.838092803955078 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.03132349888867325, | |
| "joint_other_seed": 0.0191625418628949, | |
| "single_other_seed": 0.14666737282782372, | |
| "energy": 2.9647676944732666 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.08158547696089191, | |
| "joint_other_seed": 0.07171080139793218, | |
| "single_other_seed": 0.18909359299335204, | |
| "energy": 1652.5175374348958 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.054970140183892346, | |
| "joint_other_seed": 0.12724010949136977, | |
| "single_other_seed": 0.21106607212446393, | |
| "energy": 1387.2740071614583 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.01133789951045061, | |
| "joint_other_seed": 0.0591026647868611, | |
| "single_other_seed": 0.07336679042964156, | |
| "energy": 5.488253116607666 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.12126850585573234, | |
| "joint_other_seed": 0.45867227171460334, | |
| "single_other_seed": 0.31411286993545795, | |
| "energy": 1521.1991780598958 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.040087968421274045, | |
| "joint_other_seed": 0.12772398181564293, | |
| "single_other_seed": 0.11545420099793169, | |
| "energy": 337.58428955078125 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.006865183065585478, | |
| "joint_other_seed": 0.015101792754000143, | |
| "single_other_seed": 0.010952643508143511, | |
| "energy": 17.0596186319987 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.027025463269728656, | |
| "joint_other_seed": 0.011201461874506768, | |
| "single_other_seed": 0.12064643064879921, | |
| "energy": 2.863530079523722 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.08000488494589009, | |
| "joint_other_seed": 0.10083933222451347, | |
| "single_other_seed": 0.27602314223408964, | |
| "energy": 2020.4505615234375 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.07231592383569385, | |
| "joint_other_seed": 0.09224731017422505, | |
| "single_other_seed": 0.169989040541751, | |
| "energy": 2231.3841145833335 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.0014039268077293832, | |
| "joint_other_seed": 0.006600425985965441, | |
| "single_other_seed": 0.009821833337982477, | |
| "energy": 7.1969709396362305 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.052353211813936595, | |
| "joint_other_seed": 0.3117473844207753, | |
| "single_other_seed": 0.3123821327549259, | |
| "energy": 6635.171875 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.10180392644501968, | |
| "joint_other_seed": 0.2221036939297112, | |
| "single_other_seed": 0.10059310950435674, | |
| "energy": 1342.6150309244792 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.03249950566780224, | |
| "joint_other_seed": 0.166140299251873, | |
| "single_other_seed": 0.20986998801821588, | |
| "energy": 46.28835805257162 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.04359354534681281, | |
| "joint_other_seed": 0.17422832075197858, | |
| "single_other_seed": 0.17374580439746723, | |
| "energy": 6.244580268859863 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.09948619760609068, | |
| "joint_other_seed": 0.21216567344448423, | |
| "single_other_seed": 0.1662815712719009, | |
| "energy": 1135.5994466145833 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.11399870799794731, | |
| "joint_other_seed": 0.4146602429171141, | |
| "single_other_seed": 0.2625536565083416, | |
| "energy": 1994.4418131510417 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.0023679855025592333, | |
| "joint_other_seed": 0.013069044902639049, | |
| "single_other_seed": 0.04128212263723752, | |
| "energy": 12.018329938252768 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.05384730656574732, | |
| "joint_other_seed": 0.2033347239247637, | |
| "single_other_seed": 0.5410054461993619, | |
| "energy": 17540.997721354168 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.12484863378668032, | |
| "joint_other_seed": 0.43458894126387027, | |
| "single_other_seed": 0.4146627765952352, | |
| "energy": 2058.4473876953125 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.013680449485359736, | |
| "joint_other_seed": 0.004974155717131334, | |
| "single_other_seed": 0.13228109444583425, | |
| "energy": 61.0179074605306 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.030960363432624205, | |
| "joint_other_seed": 0.018084927202645762, | |
| "single_other_seed": 0.5243670516590349, | |
| "energy": 10.982696851094564 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.08956618056575927, | |
| "joint_other_seed": 0.11728804099788326, | |
| "single_other_seed": 0.25613399472845655, | |
| "energy": 783.0172729492188 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.0623040612672755, | |
| "joint_other_seed": 0.23376389747430504, | |
| "single_other_seed": 0.1416467630350906, | |
| "energy": 849.2821451822916 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.0018164787450101194, | |
| "joint_other_seed": 0.0008614713472293589, | |
| "single_other_seed": 0.05353039268358195, | |
| "energy": 14.343783060709635 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.008215698983136677, | |
| "joint_other_seed": 0.14767683474430127, | |
| "single_other_seed": 0.038957378178193114, | |
| "energy": 4884.269205729167 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.0013186627895805936, | |
| "joint_other_seed": 0.0004958910355418178, | |
| "single_other_seed": 0.0079726877072831, | |
| "energy": 308.6481018066406 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.005643411442860513, | |
| "joint_other_seed": 0.047510842687842, | |
| "single_other_seed": 0.10130508571115149, | |
| "energy": 23.71794064839681 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.01664940412165261, | |
| "joint_other_seed": 0.12492048420183832, | |
| "single_other_seed": 0.077842908076036, | |
| "energy": 11.082512219746908 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.04582338723059587, | |
| "joint_other_seed": 0.03329932535461666, | |
| "single_other_seed": 0.12835020666546726, | |
| "energy": 951.6658121744791 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.06008420575878071, | |
| "joint_other_seed": 0.07935545053212023, | |
| "single_other_seed": 0.2709930669000443, | |
| "energy": 709.5588582356771 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.008331272231543951, | |
| "joint_other_seed": 0.04687480562834456, | |
| "single_other_seed": 0.06911898293569085, | |
| "energy": 20.852401733398438 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.20380257441343586, | |
| "joint_other_seed": 0.47429821048786436, | |
| "single_other_seed": 0.3999837686855341, | |
| "energy": 5098.1728515625 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.01157470490684509, | |
| "joint_other_seed": 0.04463511756602856, | |
| "single_other_seed": 0.08840016119708094, | |
| "energy": 489.7016296386719 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.0023912280429022603, | |
| "joint_other_seed": 0.013194041706628523, | |
| "single_other_seed": 0.02941813421287487, | |
| "energy": 35.33549499511719 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.010377424311392547, | |
| "joint_other_seed": 0.012739383070981103, | |
| "single_other_seed": 0.07232331464741404, | |
| "energy": 9.996460596720377 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.12181137041951451, | |
| "joint_other_seed": 0.378330908256706, | |
| "single_other_seed": 0.2830446572341965, | |
| "energy": 968.7154744466146 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.041165975115938, | |
| "joint_other_seed": 0.13790295037022204, | |
| "single_other_seed": 0.15169360345131822, | |
| "energy": 1397.3559163411458 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.008398822163785846, | |
| "joint_other_seed": 0.05652922531702986, | |
| "single_other_seed": 0.06414332605032513, | |
| "energy": 33.2838134765625 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.1942187975987294, | |
| "joint_other_seed": 0.3022123203849716, | |
| "single_other_seed": 0.4233531035832384, | |
| "energy": 4481.05322265625 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.03459883073158043, | |
| "joint_other_seed": 0.2671212861269485, | |
| "single_other_seed": 0.17966342118076561, | |
| "energy": 488.4266764322917 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.01670985179287221, | |
| "joint_other_seed": 0.10903827784866327, | |
| "single_other_seed": 0.04517116300473545, | |
| "energy": 64.7612813313802 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.05739924735993651, | |
| "joint_other_seed": 0.07124877842629318, | |
| "single_other_seed": 0.07614318612348714, | |
| "energy": 30.046279907226562 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.1379598682499873, | |
| "joint_other_seed": 0.35590603129586573, | |
| "single_other_seed": 0.4028023229884472, | |
| "energy": 980.7935587565104 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.14663359417271493, | |
| "joint_other_seed": 0.4230170201230068, | |
| "single_other_seed": 0.391338549547512, | |
| "energy": 1024.8572591145833 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.011660032949801669, | |
| "joint_other_seed": 0.07202283417723399, | |
| "single_other_seed": 0.047692506238434795, | |
| "energy": 52.53083928426107 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.02321216644018198, | |
| "joint_other_seed": 0.08193123748858544, | |
| "single_other_seed": 0.1978118973016012, | |
| "energy": 4274.335774739583 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.04353046304591757, | |
| "joint_other_seed": 0.1460675964067636, | |
| "single_other_seed": 0.03538017866274212, | |
| "energy": 674.996826171875 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.04968345697468857, | |
| "joint_other_seed": 0.2117756527917617, | |
| "single_other_seed": 0.15209762286262196, | |
| "energy": 119.1666742960612 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.13430269337943712, | |
| "joint_other_seed": 0.2108436945237646, | |
| "single_other_seed": 0.31794007114593115, | |
| "energy": 50.09579594930013 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.11285773476926297, | |
| "joint_other_seed": 0.3991162816538175, | |
| "single_other_seed": 0.29167323031406245, | |
| "energy": 1234.8294677734375 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.10884359856309755, | |
| "joint_other_seed": 0.15838503248411318, | |
| "single_other_seed": 0.2725821146826102, | |
| "energy": 1262.9683837890625 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.026640761823316872, | |
| "joint_other_seed": 0.08594824307975035, | |
| "single_other_seed": 0.10344464143675647, | |
| "energy": 125.56663004557292 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.03269141624065175, | |
| "joint_other_seed": 0.20163755062411803, | |
| "single_other_seed": 0.2044090064329085, | |
| "energy": 6971.009440104167 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.01314265564930106, | |
| "joint_other_seed": 0.04439290528074069, | |
| "single_other_seed": 0.07191660332068683, | |
| "energy": 682.5655110677084 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.1975331215005176, | |
| "joint_other_seed": 0.48140482384722827, | |
| "single_other_seed": 0.3921588775154543, | |
| "energy": 225.95068359375 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.3523418824930442, | |
| "joint_other_seed": 0.6394664460510574, | |
| "single_other_seed": 0.652889282105622, | |
| "energy": 208.55430603027344 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.13203583022688115, | |
| "joint_other_seed": 0.2997115194771903, | |
| "single_other_seed": 0.43420584420781094, | |
| "energy": 1208.7257893880208 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.07879322815440004, | |
| "joint_other_seed": 0.08405144077342352, | |
| "single_other_seed": 0.46562197736008193, | |
| "energy": 1635.26513671875 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.3024299987749067, | |
| "joint_other_seed": 0.43586650174327457, | |
| "single_other_seed": 0.44801295216340764, | |
| "energy": 417.8306477864583 | |
| } | |
| }, | |
| "learned/all/r3": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.20485223689243315, | |
| "joint_other_seed": 0.5920945112051836, | |
| "single_other_seed": 0.4483871063989089, | |
| "energy": 34246.380208333336 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.3573116421255398, | |
| "joint_other_seed": 0.7520437172529415, | |
| "single_other_seed": 0.6793299562745521, | |
| "energy": 2932.7308756510415 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.17949452181672254, | |
| "joint_other_seed": 0.7392035421452168, | |
| "single_other_seed": 0.5934555645300559, | |
| "energy": 217.8640594482422 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.3115867938045753, | |
| "joint_other_seed": 0.644132766515271, | |
| "single_other_seed": 0.6617750267175087, | |
| "energy": 3.8249698479970298 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.1288952927258209, | |
| "joint_other_seed": 0.398433040105468, | |
| "single_other_seed": 0.5045979891394741, | |
| "energy": 3746.07568359375 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.06988920301934665, | |
| "joint_other_seed": 0.24639903813394437, | |
| "single_other_seed": 0.07554163319658491, | |
| "energy": 2238.14892578125 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.13832515699513315, | |
| "joint_other_seed": 0.5166500305506848, | |
| "single_other_seed": 0.2536833687673502, | |
| "energy": 12.864169438680014 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.3647596145608327, | |
| "joint_other_seed": 0.6338237904768023, | |
| "single_other_seed": 0.5587965946862371, | |
| "energy": 5822.81884765625 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.0733466129532554, | |
| "joint_other_seed": 0.5904381455465599, | |
| "single_other_seed": 0.39898832270843276, | |
| "energy": 1887.6908365885417 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.20135397551345802, | |
| "joint_other_seed": 0.433132927839487, | |
| "single_other_seed": 0.5208452433709779, | |
| "energy": 58.525107065836586 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.1573991070420544, | |
| "joint_other_seed": 0.3833411550648179, | |
| "single_other_seed": 0.5249676105685213, | |
| "energy": 3.1775333086649575 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.06132899593037111, | |
| "joint_other_seed": 0.09994423994633372, | |
| "single_other_seed": 0.10464691448035823, | |
| "energy": 1587.1387532552083 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.08514503970163878, | |
| "joint_other_seed": 0.378892061398994, | |
| "single_other_seed": 0.2412835961441585, | |
| "energy": 2132.580769856771 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.11695201270961764, | |
| "joint_other_seed": 0.25169054685926595, | |
| "single_other_seed": 0.3269644458557466, | |
| "energy": 4.05422576268514 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.21412655795711044, | |
| "joint_other_seed": 0.4802735781889176, | |
| "single_other_seed": 0.3061740426306926, | |
| "energy": 6313.5908203125 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.08915117422547902, | |
| "joint_other_seed": 0.48229774407342807, | |
| "single_other_seed": 0.15188466412747137, | |
| "energy": 557.0216471354166 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.02400925095399198, | |
| "joint_other_seed": 0.17744005633408932, | |
| "single_other_seed": 0.36124135372946575, | |
| "energy": 42.68611526489258 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.06752510055540917, | |
| "joint_other_seed": 0.15871944468141375, | |
| "single_other_seed": 0.26936184836528954, | |
| "energy": 3.104984998703003 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.04616451510751429, | |
| "joint_other_seed": 0.41017787371399494, | |
| "single_other_seed": 0.3838742773934516, | |
| "energy": 2851.784423828125 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.09026460636647349, | |
| "joint_other_seed": 0.3389842338818475, | |
| "single_other_seed": 0.31676711116843315, | |
| "energy": 2577.2701009114585 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.0125222162959987, | |
| "joint_other_seed": 0.09594698221754672, | |
| "single_other_seed": 0.1492909954215249, | |
| "energy": 6.58300240834554 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.10655439593373683, | |
| "joint_other_seed": 0.43607980202507823, | |
| "single_other_seed": 0.29262605440507505, | |
| "energy": 5092.974609375 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.06520438509165882, | |
| "joint_other_seed": 0.15491802795168585, | |
| "single_other_seed": 0.039280669717608796, | |
| "energy": 489.748291015625 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.041601477561953426, | |
| "joint_other_seed": 0.11825717794849666, | |
| "single_other_seed": 0.12560850942846002, | |
| "energy": 35.22747802734375 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.025641910654470967, | |
| "joint_other_seed": 0.16568970822812223, | |
| "single_other_seed": 0.14666737282782372, | |
| "energy": 5.749405066172282 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.05041185006988924, | |
| "joint_other_seed": 0.19824749290829413, | |
| "single_other_seed": 0.18909359299335204, | |
| "energy": 2377.216552734375 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.08031222111601999, | |
| "joint_other_seed": 0.2690877251874597, | |
| "single_other_seed": 0.21106607212446393, | |
| "energy": 2234.236083984375 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.017505612063583698, | |
| "joint_other_seed": 0.0951293603596981, | |
| "single_other_seed": 0.07336679042964156, | |
| "energy": 12.956251780192057 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.08769843830360191, | |
| "joint_other_seed": 0.3218538716598144, | |
| "single_other_seed": 0.31411286993545795, | |
| "energy": 2962.601806640625 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.04568455940961513, | |
| "joint_other_seed": 0.20213881850664905, | |
| "single_other_seed": 0.11545420099793169, | |
| "energy": 463.90940348307294 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.01064764071925639, | |
| "joint_other_seed": 0.025625962050934148, | |
| "single_other_seed": 0.010952643508143511, | |
| "energy": 35.48167928059896 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.02068488523350831, | |
| "joint_other_seed": 0.06064637157519015, | |
| "single_other_seed": 0.12064643064879921, | |
| "energy": 4.847487131754558 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.038153259557785146, | |
| "joint_other_seed": 0.14966782311169358, | |
| "single_other_seed": 0.27602314223408964, | |
| "energy": 2281.346150716146 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.03257399070582867, | |
| "joint_other_seed": 0.2636804502951985, | |
| "single_other_seed": 0.169989040541751, | |
| "energy": 3143.10595703125 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.0038795735425803946, | |
| "joint_other_seed": 0.08271356706813282, | |
| "single_other_seed": 0.009821833337982477, | |
| "energy": 16.159591674804688 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.05353354326431248, | |
| "joint_other_seed": 0.4492585873837631, | |
| "single_other_seed": 0.3123821327549259, | |
| "energy": 5835.302897135417 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.12594730037393453, | |
| "joint_other_seed": 0.37995356345700115, | |
| "single_other_seed": 0.10059310950435674, | |
| "energy": 1375.4086303710938 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.015685254276397442, | |
| "joint_other_seed": 0.08116389545547742, | |
| "single_other_seed": 0.20986998801821588, | |
| "energy": 59.52655283610026 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.03158557265409534, | |
| "joint_other_seed": 0.2685901030012928, | |
| "single_other_seed": 0.17374580439746723, | |
| "energy": 10.7991304397583 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.04987360066908875, | |
| "joint_other_seed": 0.1963092717982069, | |
| "single_other_seed": 0.1662815712719009, | |
| "energy": 2479.6629231770835 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.04302418294502919, | |
| "joint_other_seed": 0.2755424087014008, | |
| "single_other_seed": 0.2625536565083416, | |
| "energy": 2968.7808430989585 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.0045898419172095695, | |
| "joint_other_seed": 0.09028080078273248, | |
| "single_other_seed": 0.04128212263723752, | |
| "energy": 22.968135833740234 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.10945039023353376, | |
| "joint_other_seed": 0.6019000475454479, | |
| "single_other_seed": 0.5410054461993619, | |
| "energy": 14387.960611979166 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.14050083202191177, | |
| "joint_other_seed": 0.6887776774107757, | |
| "single_other_seed": 0.4146627765952352, | |
| "energy": 2122.3829752604165 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.012784173544427846, | |
| "joint_other_seed": 0.10530398493698201, | |
| "single_other_seed": 0.13228109444583425, | |
| "energy": 73.17888895670573 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.13303875223930528, | |
| "joint_other_seed": 0.3737802501176031, | |
| "single_other_seed": 0.5243670516590349, | |
| "energy": 20.72862180074056 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.03617529640059225, | |
| "joint_other_seed": 0.09835275348904338, | |
| "single_other_seed": 0.25613399472845655, | |
| "energy": 2107.8296712239585 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.038307160351620406, | |
| "joint_other_seed": 0.153946511033493, | |
| "single_other_seed": 0.1416467630350906, | |
| "energy": 2042.6400960286458 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.004885979156638465, | |
| "joint_other_seed": 0.07913119168803329, | |
| "single_other_seed": 0.05353039268358195, | |
| "energy": 30.423954010009766 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.013430116019369054, | |
| "joint_other_seed": 0.0821722741037878, | |
| "single_other_seed": 0.038957378178193114, | |
| "energy": 4845.298177083333 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.005170172935487161, | |
| "joint_other_seed": 0.016158846819156025, | |
| "single_other_seed": 0.0079726877072831, | |
| "energy": 367.9163513183594 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.00902996853509297, | |
| "joint_other_seed": 0.023314093986942064, | |
| "single_other_seed": 0.10130508571115149, | |
| "energy": 39.078182220458984 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.015567131811384916, | |
| "joint_other_seed": 0.06878096117464017, | |
| "single_other_seed": 0.077842908076036, | |
| "energy": 17.30240249633789 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.035978105475466426, | |
| "joint_other_seed": 0.07760108806107979, | |
| "single_other_seed": 0.12835020666546726, | |
| "energy": 2106.1930338541665 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.03456611117468822, | |
| "joint_other_seed": 0.13405042174523396, | |
| "single_other_seed": 0.2709930669000443, | |
| "energy": 1939.9485677083333 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.005423304537180207, | |
| "joint_other_seed": 0.049741192846117444, | |
| "single_other_seed": 0.06911898293569085, | |
| "energy": 37.48411305745443 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.1575826687099956, | |
| "joint_other_seed": 0.3158245148157097, | |
| "single_other_seed": 0.3999837686855341, | |
| "energy": 4719.584309895833 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.009417102804265887, | |
| "joint_other_seed": 0.03958895004848051, | |
| "single_other_seed": 0.08840016119708094, | |
| "energy": 529.8346150716146 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.008379704522810782, | |
| "joint_other_seed": 0.03471049282337468, | |
| "single_other_seed": 0.02941813421287487, | |
| "energy": 75.92956797281902 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.012709089737200974, | |
| "joint_other_seed": 0.11254560855297491, | |
| "single_other_seed": 0.07232331464741404, | |
| "energy": 22.815759658813477 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.05451026526709557, | |
| "joint_other_seed": 0.15114568549133867, | |
| "single_other_seed": 0.2830446572341965, | |
| "energy": 2598.7159830729165 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.01714885091174867, | |
| "joint_other_seed": 0.15650098284961167, | |
| "single_other_seed": 0.15169360345131822, | |
| "energy": 2580.5530598958335 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.0038673622772994465, | |
| "joint_other_seed": 0.0653656835152231, | |
| "single_other_seed": 0.06414332605032513, | |
| "energy": 62.51854705810547 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.10123183721250328, | |
| "joint_other_seed": 0.17959098236425486, | |
| "single_other_seed": 0.4233531035832384, | |
| "energy": 4122.842041015625 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.025529842265228697, | |
| "joint_other_seed": 0.1568017189963754, | |
| "single_other_seed": 0.17966342118076561, | |
| "energy": 638.3365275065104 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.020696937639061774, | |
| "joint_other_seed": 0.16079263904198826, | |
| "single_other_seed": 0.04517116300473545, | |
| "energy": 104.80843353271484 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.07663722941345064, | |
| "joint_other_seed": 0.236074082157806, | |
| "single_other_seed": 0.07614318612348714, | |
| "energy": 40.18707021077474 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.07914998056794653, | |
| "joint_other_seed": 0.1310741049469093, | |
| "single_other_seed": 0.4028023229884472, | |
| "energy": 2190.500244140625 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.0668115511909001, | |
| "joint_other_seed": 0.11396372669355181, | |
| "single_other_seed": 0.391338549547512, | |
| "energy": 2235.891845703125 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.02191946961005864, | |
| "joint_other_seed": 0.1253408512585733, | |
| "single_other_seed": 0.047692506238434795, | |
| "energy": 94.52956899007161 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.059282580906446225, | |
| "joint_other_seed": 0.24370158109710094, | |
| "single_other_seed": 0.1978118973016012, | |
| "energy": 5841.813639322917 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.030144937201884714, | |
| "joint_other_seed": 0.16424524059185464, | |
| "single_other_seed": 0.03538017866274212, | |
| "energy": 676.0024617513021 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.045051213489544994, | |
| "joint_other_seed": 0.2930174013966669, | |
| "single_other_seed": 0.15209762286262196, | |
| "energy": 143.7727305094401 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.14958017888832476, | |
| "joint_other_seed": 0.33441318138834586, | |
| "single_other_seed": 0.31794007114593115, | |
| "energy": 70.95785776774089 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.07815650598221058, | |
| "joint_other_seed": 0.1819460813346655, | |
| "single_other_seed": 0.29167323031406245, | |
| "energy": 2448.5302734375 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.058770642418921715, | |
| "joint_other_seed": 0.1292029509628252, | |
| "single_other_seed": 0.2725821146826102, | |
| "energy": 2572.5294596354165 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.07791957857448484, | |
| "joint_other_seed": 0.26403511799690205, | |
| "single_other_seed": 0.10344464143675647, | |
| "energy": 230.0251719156901 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.03579490069201641, | |
| "joint_other_seed": 0.17973144790353002, | |
| "single_other_seed": 0.2044090064329085, | |
| "energy": 5672.842610677083 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.01495471989275965, | |
| "joint_other_seed": 0.035794538057918546, | |
| "single_other_seed": 0.07191660332068683, | |
| "energy": 756.9460245768229 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.17020446602076567, | |
| "joint_other_seed": 0.4601424600231079, | |
| "single_other_seed": 0.3921588775154543, | |
| "energy": 223.43350728352866 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.39245451818358645, | |
| "joint_other_seed": 0.6555999996756019, | |
| "single_other_seed": 0.652889282105622, | |
| "energy": 232.05036417643228 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.10395158305850005, | |
| "joint_other_seed": 0.22877314301404084, | |
| "single_other_seed": 0.43420584420781094, | |
| "energy": 2590.4475911458335 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.09755375340126982, | |
| "joint_other_seed": 0.4135828534631538, | |
| "single_other_seed": 0.46562197736008193, | |
| "energy": 2852.3282877604165 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.2771367962185871, | |
| "joint_other_seed": 0.5036786804301653, | |
| "single_other_seed": 0.44801295216340764, | |
| "energy": 494.02356974283856 | |
| } | |
| }, | |
| "learned/all/r16": { | |
| "model.layers.0.self_attn.q_proj": { | |
| "singles_other_seed": 0.23387075738450402, | |
| "joint_other_seed": 0.7325071958520368, | |
| "single_other_seed": 0.4483871063989089, | |
| "energy": 26930.888671875 | |
| }, | |
| "model.layers.0.self_attn.k_proj": { | |
| "singles_other_seed": 0.3878293409999547, | |
| "joint_other_seed": 0.7770571077440166, | |
| "single_other_seed": 0.6793299562745521, | |
| "energy": 2093.705769856771 | |
| }, | |
| "model.layers.0.self_attn.v_proj": { | |
| "singles_other_seed": 0.2066770773407055, | |
| "joint_other_seed": 0.758219909878391, | |
| "single_other_seed": 0.5934555645300559, | |
| "energy": 209.82774353027344 | |
| }, | |
| "model.layers.0.self_attn.o_proj": { | |
| "singles_other_seed": 0.30718982447807064, | |
| "joint_other_seed": 0.7740853167282982, | |
| "single_other_seed": 0.6617750267175087, | |
| "energy": 4.1647359530131025 | |
| }, | |
| "model.layers.0.mlp.gate_proj": { | |
| "singles_other_seed": 0.10997249221417726, | |
| "joint_other_seed": 0.5262636423850005, | |
| "single_other_seed": 0.5045979891394741, | |
| "energy": 2928.3849283854165 | |
| }, | |
| "model.layers.0.mlp.up_proj": { | |
| "singles_other_seed": 0.07436649063710499, | |
| "joint_other_seed": 0.4965499945771953, | |
| "single_other_seed": 0.07554163319658491, | |
| "energy": 2762.7947591145835 | |
| }, | |
| "model.layers.0.mlp.down_proj": { | |
| "singles_other_seed": 0.1502510260898935, | |
| "joint_other_seed": 0.6215727341964965, | |
| "single_other_seed": 0.2536833687673502, | |
| "energy": 11.31715202331543 | |
| }, | |
| "model.layers.1.self_attn.q_proj": { | |
| "singles_other_seed": 0.4227435976880056, | |
| "joint_other_seed": 0.740224127310426, | |
| "single_other_seed": 0.5587965946862371, | |
| "energy": 5411.8251953125 | |
| }, | |
| "model.layers.1.self_attn.k_proj": { | |
| "singles_other_seed": 0.07652925466342797, | |
| "joint_other_seed": 0.7537031854186876, | |
| "single_other_seed": 0.39898832270843276, | |
| "energy": 1720.5129801432292 | |
| }, | |
| "model.layers.1.self_attn.v_proj": { | |
| "singles_other_seed": 0.1699857712269448, | |
| "joint_other_seed": 0.5659631959994764, | |
| "single_other_seed": 0.5208452433709779, | |
| "energy": 50.59355672200521 | |
| }, | |
| "model.layers.1.self_attn.o_proj": { | |
| "singles_other_seed": 0.15897844679427908, | |
| "joint_other_seed": 0.5702306529030938, | |
| "single_other_seed": 0.5249676105685213, | |
| "energy": 4.536321957906087 | |
| }, | |
| "model.layers.1.mlp.gate_proj": { | |
| "singles_other_seed": 0.054522354429360154, | |
| "joint_other_seed": 0.5168173818034913, | |
| "single_other_seed": 0.10464691448035823, | |
| "energy": 2386.2920735677085 | |
| }, | |
| "model.layers.1.mlp.up_proj": { | |
| "singles_other_seed": 0.08792912238696593, | |
| "joint_other_seed": 0.4813053810099869, | |
| "single_other_seed": 0.2412835961441585, | |
| "energy": 1840.7886149088542 | |
| }, | |
| "model.layers.1.mlp.down_proj": { | |
| "singles_other_seed": 0.12866918813857398, | |
| "joint_other_seed": 0.4291034968151567, | |
| "single_other_seed": 0.3269644458557466, | |
| "energy": 4.355302015940349 | |
| }, | |
| "model.layers.2.self_attn.q_proj": { | |
| "singles_other_seed": 0.21901906490465817, | |
| "joint_other_seed": 0.7217131361480764, | |
| "single_other_seed": 0.3061740426306926, | |
| "energy": 6212.553059895833 | |
| }, | |
| "model.layers.2.self_attn.k_proj": { | |
| "singles_other_seed": 0.10280290486942512, | |
| "joint_other_seed": 0.6129417333315114, | |
| "single_other_seed": 0.15188466412747137, | |
| "energy": 425.8183186848958 | |
| }, | |
| "model.layers.2.self_attn.v_proj": { | |
| "singles_other_seed": 0.016164996444436475, | |
| "joint_other_seed": 0.33587391781440923, | |
| "single_other_seed": 0.36124135372946575, | |
| "energy": 46.700730641682945 | |
| }, | |
| "model.layers.2.self_attn.o_proj": { | |
| "singles_other_seed": 0.0491230929385893, | |
| "joint_other_seed": 0.3228205259572646, | |
| "single_other_seed": 0.26936184836528954, | |
| "energy": 5.398156642913818 | |
| }, | |
| "model.layers.2.mlp.gate_proj": { | |
| "singles_other_seed": 0.03719593940536525, | |
| "joint_other_seed": 0.4698715016759103, | |
| "single_other_seed": 0.3838742773934516, | |
| "energy": 3303.528564453125 | |
| }, | |
| "model.layers.2.mlp.up_proj": { | |
| "singles_other_seed": 0.05524505924409895, | |
| "joint_other_seed": 0.5079085823298903, | |
| "single_other_seed": 0.31676711116843315, | |
| "energy": 3098.0032552083335 | |
| }, | |
| "model.layers.2.mlp.down_proj": { | |
| "singles_other_seed": 0.013706134309480311, | |
| "joint_other_seed": 0.24625066139066465, | |
| "single_other_seed": 0.1492909954215249, | |
| "energy": 8.120729287465414 | |
| }, | |
| "model.layers.3.self_attn.q_proj": { | |
| "singles_other_seed": 0.10932380042961798, | |
| "joint_other_seed": 0.6471084794256778, | |
| "single_other_seed": 0.29262605440507505, | |
| "energy": 4530.885091145833 | |
| }, | |
| "model.layers.3.self_attn.k_proj": { | |
| "singles_other_seed": 0.08220730097408835, | |
| "joint_other_seed": 0.4637163758659635, | |
| "single_other_seed": 0.039280669717608796, | |
| "energy": 435.7821960449219 | |
| }, | |
| "model.layers.3.self_attn.v_proj": { | |
| "singles_other_seed": 0.038674991951751836, | |
| "joint_other_seed": 0.1707472605762269, | |
| "single_other_seed": 0.12560850942846002, | |
| "energy": 32.374711990356445 | |
| }, | |
| "model.layers.3.self_attn.o_proj": { | |
| "singles_other_seed": 0.027584311759265567, | |
| "joint_other_seed": 0.3315824216398535, | |
| "single_other_seed": 0.14666737282782372, | |
| "energy": 9.424128850301107 | |
| }, | |
| "model.layers.3.mlp.gate_proj": { | |
| "singles_other_seed": 0.05182184720281646, | |
| "joint_other_seed": 0.4599920626371017, | |
| "single_other_seed": 0.18909359299335204, | |
| "energy": 3796.099853515625 | |
| }, | |
| "model.layers.3.mlp.up_proj": { | |
| "singles_other_seed": 0.07485299172858173, | |
| "joint_other_seed": 0.3862637270528124, | |
| "single_other_seed": 0.21106607212446393, | |
| "energy": 3004.9335123697915 | |
| }, | |
| "model.layers.3.mlp.down_proj": { | |
| "singles_other_seed": 0.022100509252305334, | |
| "joint_other_seed": 0.27153333343192054, | |
| "single_other_seed": 0.07336679042964156, | |
| "energy": 18.057025909423828 | |
| }, | |
| "model.layers.4.self_attn.q_proj": { | |
| "singles_other_seed": 0.08366189025303761, | |
| "joint_other_seed": 0.4709564079781554, | |
| "single_other_seed": 0.31411286993545795, | |
| "energy": 3009.3076171875 | |
| }, | |
| "model.layers.4.self_attn.k_proj": { | |
| "singles_other_seed": 0.05751920696602786, | |
| "joint_other_seed": 0.37474905912697154, | |
| "single_other_seed": 0.11545420099793169, | |
| "energy": 475.0594482421875 | |
| }, | |
| "model.layers.4.self_attn.v_proj": { | |
| "singles_other_seed": 0.010421826721844551, | |
| "joint_other_seed": 0.17147423489185598, | |
| "single_other_seed": 0.010952643508143511, | |
| "energy": 53.66927591959635 | |
| }, | |
| "model.layers.4.self_attn.o_proj": { | |
| "singles_other_seed": 0.009321936412427956, | |
| "joint_other_seed": 0.18295732896100314, | |
| "single_other_seed": 0.12064643064879921, | |
| "energy": 12.156514803568522 | |
| }, | |
| "model.layers.4.mlp.gate_proj": { | |
| "singles_other_seed": 0.01639384860658355, | |
| "joint_other_seed": 0.3729556388711379, | |
| "single_other_seed": 0.27602314223408964, | |
| "energy": 3772.4398600260415 | |
| }, | |
| "model.layers.4.mlp.up_proj": { | |
| "singles_other_seed": 0.018382449333385065, | |
| "joint_other_seed": 0.4143718191368376, | |
| "single_other_seed": 0.169989040541751, | |
| "energy": 4128.189290364583 | |
| }, | |
| "model.layers.4.mlp.down_proj": { | |
| "singles_other_seed": 0.0034993732902829366, | |
| "joint_other_seed": 0.21224046727205448, | |
| "single_other_seed": 0.009821833337982477, | |
| "energy": 26.269893646240234 | |
| }, | |
| "model.layers.5.self_attn.q_proj": { | |
| "singles_other_seed": 0.060075783794067195, | |
| "joint_other_seed": 0.5740302971480941, | |
| "single_other_seed": 0.3123821327549259, | |
| "energy": 5468.587890625 | |
| }, | |
| "model.layers.5.self_attn.k_proj": { | |
| "singles_other_seed": 0.14280884751901224, | |
| "joint_other_seed": 0.6028128200550344, | |
| "single_other_seed": 0.10059310950435674, | |
| "energy": 1144.3315836588542 | |
| }, | |
| "model.layers.5.self_attn.v_proj": { | |
| "singles_other_seed": 0.011856248645261025, | |
| "joint_other_seed": 0.20850531094499525, | |
| "single_other_seed": 0.20986998801821588, | |
| "energy": 69.08138020833333 | |
| }, | |
| "model.layers.5.self_attn.o_proj": { | |
| "singles_other_seed": 0.02977898726499839, | |
| "joint_other_seed": 0.37079943814399974, | |
| "single_other_seed": 0.17374580439746723, | |
| "energy": 17.967166900634766 | |
| }, | |
| "model.layers.5.mlp.gate_proj": { | |
| "singles_other_seed": 0.02700554829096735, | |
| "joint_other_seed": 0.36496434054926696, | |
| "single_other_seed": 0.1662815712719009, | |
| "energy": 3658.9251302083335 | |
| }, | |
| "model.layers.5.mlp.up_proj": { | |
| "singles_other_seed": 0.02883053333093588, | |
| "joint_other_seed": 0.3755941856782551, | |
| "single_other_seed": 0.2625536565083416, | |
| "energy": 4046.13330078125 | |
| }, | |
| "model.layers.5.mlp.down_proj": { | |
| "singles_other_seed": 0.006100858124274855, | |
| "joint_other_seed": 0.20203651283200885, | |
| "single_other_seed": 0.04128212263723752, | |
| "energy": 36.21170425415039 | |
| }, | |
| "model.layers.6.self_attn.q_proj": { | |
| "singles_other_seed": 0.11678989996299782, | |
| "joint_other_seed": 0.8027456174814469, | |
| "single_other_seed": 0.5410054461993619, | |
| "energy": 12273.160481770834 | |
| }, | |
| "model.layers.6.self_attn.k_proj": { | |
| "singles_other_seed": 0.16014086184560816, | |
| "joint_other_seed": 0.7635882034848652, | |
| "single_other_seed": 0.4146627765952352, | |
| "energy": 1974.9954020182292 | |
| }, | |
| "model.layers.6.self_attn.v_proj": { | |
| "singles_other_seed": 0.011429136623280822, | |
| "joint_other_seed": 0.22095370734578912, | |
| "single_other_seed": 0.13228109444583425, | |
| "energy": 77.70109558105469 | |
| }, | |
| "model.layers.6.self_attn.o_proj": { | |
| "singles_other_seed": 0.14094224450004067, | |
| "joint_other_seed": 0.5650274801080177, | |
| "single_other_seed": 0.5243670516590349, | |
| "energy": 30.193321228027344 | |
| }, | |
| "model.layers.6.mlp.gate_proj": { | |
| "singles_other_seed": 0.017762169006159345, | |
| "joint_other_seed": 0.30937748902331896, | |
| "single_other_seed": 0.25613399472845655, | |
| "energy": 4210.993977864583 | |
| }, | |
| "model.layers.6.mlp.up_proj": { | |
| "singles_other_seed": 0.018632562252564377, | |
| "joint_other_seed": 0.2563178129540855, | |
| "single_other_seed": 0.1416467630350906, | |
| "energy": 3968.0498860677085 | |
| }, | |
| "model.layers.6.mlp.down_proj": { | |
| "singles_other_seed": 0.005286657349553581, | |
| "joint_other_seed": 0.1861546790625429, | |
| "single_other_seed": 0.05353039268358195, | |
| "energy": 50.00858052571615 | |
| }, | |
| "model.layers.7.self_attn.q_proj": { | |
| "singles_other_seed": 0.020066165435583816, | |
| "joint_other_seed": 0.2646282960636378, | |
| "single_other_seed": 0.038957378178193114, | |
| "energy": 4839.559895833333 | |
| }, | |
| "model.layers.7.self_attn.k_proj": { | |
| "singles_other_seed": 0.004621976121465754, | |
| "joint_other_seed": 0.06762252722150908, | |
| "single_other_seed": 0.0079726877072831, | |
| "energy": 285.8127034505208 | |
| }, | |
| "model.layers.7.self_attn.v_proj": { | |
| "singles_other_seed": 0.014183934214532568, | |
| "joint_other_seed": 0.14440902333063943, | |
| "single_other_seed": 0.10130508571115149, | |
| "energy": 56.19954045613607 | |
| }, | |
| "model.layers.7.self_attn.o_proj": { | |
| "singles_other_seed": 0.011646086140135847, | |
| "joint_other_seed": 0.16983889752298853, | |
| "single_other_seed": 0.077842908076036, | |
| "energy": 33.74341074625651 | |
| }, | |
| "model.layers.7.mlp.gate_proj": { | |
| "singles_other_seed": 0.01758050820158419, | |
| "joint_other_seed": 0.24412093308033256, | |
| "single_other_seed": 0.12835020666546726, | |
| "energy": 4009.2550455729165 | |
| }, | |
| "model.layers.7.mlp.up_proj": { | |
| "singles_other_seed": 0.018774974139007194, | |
| "joint_other_seed": 0.2596450309233205, | |
| "single_other_seed": 0.2709930669000443, | |
| "energy": 3740.4822591145835 | |
| }, | |
| "model.layers.7.mlp.down_proj": { | |
| "singles_other_seed": 0.006439864022703468, | |
| "joint_other_seed": 0.140210739932364, | |
| "single_other_seed": 0.06911898293569085, | |
| "energy": 68.08351643880208 | |
| }, | |
| "model.layers.8.self_attn.q_proj": { | |
| "singles_other_seed": 0.1340140169609506, | |
| "joint_other_seed": 0.4105733964588506, | |
| "single_other_seed": 0.3999837686855341, | |
| "energy": 4504.122884114583 | |
| }, | |
| "model.layers.8.self_attn.k_proj": { | |
| "singles_other_seed": 0.012152025745798989, | |
| "joint_other_seed": 0.20887552429176637, | |
| "single_other_seed": 0.08840016119708094, | |
| "energy": 510.070068359375 | |
| }, | |
| "model.layers.8.self_attn.v_proj": { | |
| "singles_other_seed": 0.01101502816390143, | |
| "joint_other_seed": 0.16975911687826373, | |
| "single_other_seed": 0.02941813421287487, | |
| "energy": 86.75473276774089 | |
| }, | |
| "model.layers.8.self_attn.o_proj": { | |
| "singles_other_seed": 0.012248817284683744, | |
| "joint_other_seed": 0.20789418766995527, | |
| "single_other_seed": 0.07232331464741404, | |
| "energy": 48.27656682332357 | |
| }, | |
| "model.layers.8.mlp.gate_proj": { | |
| "singles_other_seed": 0.026416785364622553, | |
| "joint_other_seed": 0.31678225175354535, | |
| "single_other_seed": 0.2830446572341965, | |
| "energy": 4703.153157552083 | |
| }, | |
| "model.layers.8.mlp.up_proj": { | |
| "singles_other_seed": 0.013727728836817832, | |
| "joint_other_seed": 0.26698502904621746, | |
| "single_other_seed": 0.15169360345131822, | |
| "energy": 4176.9033203125 | |
| }, | |
| "model.layers.8.mlp.down_proj": { | |
| "singles_other_seed": 0.004712482390009098, | |
| "joint_other_seed": 0.1859152726144605, | |
| "single_other_seed": 0.06414332605032513, | |
| "energy": 107.3213602701823 | |
| }, | |
| "model.layers.9.self_attn.q_proj": { | |
| "singles_other_seed": 0.1265974912351343, | |
| "joint_other_seed": 0.41774911409215076, | |
| "single_other_seed": 0.4233531035832384, | |
| "energy": 3973.4391276041665 | |
| }, | |
| "model.layers.9.self_attn.k_proj": { | |
| "singles_other_seed": 0.029303054498701075, | |
| "joint_other_seed": 0.31458328350378184, | |
| "single_other_seed": 0.17966342118076561, | |
| "energy": 555.4484659830729 | |
| }, | |
| "model.layers.9.self_attn.v_proj": { | |
| "singles_other_seed": 0.018466814451073454, | |
| "joint_other_seed": 0.24383908604956794, | |
| "single_other_seed": 0.04517116300473545, | |
| "energy": 100.79182688395183 | |
| }, | |
| "model.layers.9.self_attn.o_proj": { | |
| "singles_other_seed": 0.07012197636914075, | |
| "joint_other_seed": 0.3702629759084949, | |
| "single_other_seed": 0.07614318612348714, | |
| "energy": 70.29076131184895 | |
| }, | |
| "model.layers.9.mlp.gate_proj": { | |
| "singles_other_seed": 0.04362410422182228, | |
| "joint_other_seed": 0.2983596500749679, | |
| "single_other_seed": 0.4028023229884472, | |
| "energy": 4131.000162760417 | |
| }, | |
| "model.layers.9.mlp.up_proj": { | |
| "singles_other_seed": 0.03285880089548465, | |
| "joint_other_seed": 0.2717764801358986, | |
| "single_other_seed": 0.391338549547512, | |
| "energy": 3649.8943684895835 | |
| }, | |
| "model.layers.9.mlp.down_proj": { | |
| "singles_other_seed": 0.023775684125992402, | |
| "joint_other_seed": 0.26884334937737286, | |
| "single_other_seed": 0.047692506238434795, | |
| "energy": 149.96471150716147 | |
| }, | |
| "model.layers.10.self_attn.q_proj": { | |
| "singles_other_seed": 0.05916557664176081, | |
| "joint_other_seed": 0.4828915917971506, | |
| "single_other_seed": 0.1978118973016012, | |
| "energy": 6108.22900390625 | |
| }, | |
| "model.layers.10.self_attn.k_proj": { | |
| "singles_other_seed": 0.032275718075477466, | |
| "joint_other_seed": 0.24358231013318088, | |
| "single_other_seed": 0.03538017866274212, | |
| "energy": 667.8701782226562 | |
| }, | |
| "model.layers.10.self_attn.v_proj": { | |
| "singles_other_seed": 0.041534015743290964, | |
| "joint_other_seed": 0.4017617221259949, | |
| "single_other_seed": 0.15209762286262196, | |
| "energy": 151.9014434814453 | |
| }, | |
| "model.layers.10.self_attn.o_proj": { | |
| "singles_other_seed": 0.14073649699882862, | |
| "joint_other_seed": 0.5122703504383821, | |
| "single_other_seed": 0.31794007114593115, | |
| "energy": 95.79160817464192 | |
| }, | |
| "model.layers.10.mlp.gate_proj": { | |
| "singles_other_seed": 0.04330076349288625, | |
| "joint_other_seed": 0.33434412049570794, | |
| "single_other_seed": 0.29167323031406245, | |
| "energy": 4254.279296875 | |
| }, | |
| "model.layers.10.mlp.up_proj": { | |
| "singles_other_seed": 0.028326859675440352, | |
| "joint_other_seed": 0.31886315918766783, | |
| "single_other_seed": 0.2725821146826102, | |
| "energy": 4150.224446614583 | |
| }, | |
| "model.layers.10.mlp.down_proj": { | |
| "singles_other_seed": 0.07078240612630915, | |
| "joint_other_seed": 0.4497988421608027, | |
| "single_other_seed": 0.10344464143675647, | |
| "energy": 260.5283915201823 | |
| }, | |
| "model.layers.11.self_attn.q_proj": { | |
| "singles_other_seed": 0.04414873726449231, | |
| "joint_other_seed": 0.34034239254272924, | |
| "single_other_seed": 0.2044090064329085, | |
| "energy": 5809.34716796875 | |
| }, | |
| "model.layers.11.self_attn.k_proj": { | |
| "singles_other_seed": 0.02724279589236822, | |
| "joint_other_seed": 0.21957741785987672, | |
| "single_other_seed": 0.07191660332068683, | |
| "energy": 695.90673828125 | |
| }, | |
| "model.layers.11.self_attn.v_proj": { | |
| "singles_other_seed": 0.20818930209712472, | |
| "joint_other_seed": 0.6525536772992506, | |
| "single_other_seed": 0.3921588775154543, | |
| "energy": 224.63429260253906 | |
| }, | |
| "model.layers.11.self_attn.o_proj": { | |
| "singles_other_seed": 0.44377762111850405, | |
| "joint_other_seed": 0.7562994376604327, | |
| "single_other_seed": 0.652889282105622, | |
| "energy": 255.89800008138022 | |
| }, | |
| "model.layers.11.mlp.gate_proj": { | |
| "singles_other_seed": 0.07543906940076728, | |
| "joint_other_seed": 0.45595465716231753, | |
| "single_other_seed": 0.43420584420781094, | |
| "energy": 4424.37255859375 | |
| }, | |
| "model.layers.11.mlp.up_proj": { | |
| "singles_other_seed": 0.09594251728465975, | |
| "joint_other_seed": 0.4648319188261397, | |
| "single_other_seed": 0.46562197736008193, | |
| "energy": 3866.2833658854165 | |
| }, | |
| "model.layers.11.mlp.down_proj": { | |
| "singles_other_seed": 0.23243650000346694, | |
| "joint_other_seed": 0.6429051314558771, | |
| "single_other_seed": 0.44801295216340764, | |
| "energy": 557.4945068359375 | |
| } | |
| } | |
| }, | |
| "groups": { | |
| "trained/contrast/r1": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.15118580011499372, | |
| "joint_other_seed": 0.3143820837102935, | |
| "single_other_seed": 0.47735655007467603, | |
| "corrected": 0.3902659536105742 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.8383709890880477, | |
| "singles_other_seed": 0.14656295213312098, | |
| "joint_other_seed": 0.3116221348409083, | |
| "single_other_seed": 0.5029426955018235, | |
| "corrected": 0.3702122647581829 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.8743877906613601, | |
| "singles_other_seed": 0.15115624028400798, | |
| "joint_other_seed": 0.3160984287151336, | |
| "single_other_seed": 0.49429344616358994, | |
| "corrected": 0.382404052952363 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj" | |
| ], | |
| "energy_share": 0.8297188628978601, | |
| "singles_other_seed": 0.14587427590724822, | |
| "joint_other_seed": 0.3090586842247457, | |
| "single_other_seed": 0.5008961566982497, | |
| "corrected": 0.3707527558962239 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.12561220933863987, | |
| "singles_other_seed": 0.1513915663812192, | |
| "joint_other_seed": 0.3024345897151125, | |
| "single_other_seed": 0.37246397053338154, | |
| "corrected": 0.45106986935710575 | |
| } | |
| }, | |
| "trained/contrast/r3": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.18252321686433703, | |
| "joint_other_seed": 0.5866823451817111, | |
| "single_other_seed": 0.47735655007467603, | |
| "corrected": 0.34490156231232977 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.7914125695651522, | |
| "singles_other_seed": 0.19010509529625053, | |
| "joint_other_seed": 0.6147162079492696, | |
| "single_other_seed": 0.5029426955018235, | |
| "corrected": 0.3418984820434473 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.8364275928193164, | |
| "singles_other_seed": 0.19305102739205515, | |
| "joint_other_seed": 0.6110109078225722, | |
| "single_other_seed": 0.49429344616358994, | |
| "corrected": 0.3512814440830644 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.3.self_attn.q_proj" | |
| ], | |
| "energy_share": 0.8043517059743768, | |
| "singles_other_seed": 0.18820177648839087, | |
| "joint_other_seed": 0.6138242599039903, | |
| "single_other_seed": 0.49914316763391264, | |
| "corrected": 0.34000799459492426 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.16357240718068353, | |
| "singles_other_seed": 0.1286892520219035, | |
| "joint_other_seed": 0.46227822692494724, | |
| "single_other_seed": 0.37246397053338154, | |
| "corrected": 0.3101333161261369 | |
| } | |
| }, | |
| "trained/contrast/r16": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.19712998680099625, | |
| "joint_other_seed": 0.7293115931667272, | |
| "single_other_seed": 0.47735655007467603, | |
| "corrected": 0.33409862722459294 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.7644106552409223, | |
| "singles_other_seed": 0.21122640709975382, | |
| "joint_other_seed": 0.7547286654756791, | |
| "single_other_seed": 0.5029426955018235, | |
| "corrected": 0.34284161852096495 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.8076551578567643, | |
| "singles_other_seed": 0.21291946808448392, | |
| "joint_other_seed": 0.7522569392001052, | |
| "single_other_seed": 0.49429344616358994, | |
| "corrected": 0.3491723494807317 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.3.mlp.up_proj" | |
| ], | |
| "energy_share": 0.8112439097746312, | |
| "singles_other_seed": 0.21153255191741716, | |
| "joint_other_seed": 0.7515730843093564, | |
| "single_other_seed": 0.4936999941996034, | |
| "corrected": 0.3472642198849651 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.1923448421432357, | |
| "singles_other_seed": 0.130830023379075, | |
| "joint_other_seed": 0.6329641846988381, | |
| "single_other_seed": 0.37246397053338154, | |
| "corrected": 0.2694484665128006 | |
| } | |
| }, | |
| "trained/trigger/r1": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.13954749424936996, | |
| "joint_other_seed": 0.2791917246230651, | |
| "single_other_seed": 0.3922550950079986, | |
| "corrected": 0.4216833711517425 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.49347091395988835, | |
| "singles_other_seed": 0.1658072677785266, | |
| "joint_other_seed": 0.2890938742609621, | |
| "single_other_seed": 0.47448754105361524, | |
| "corrected": 0.4476841426723804 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.5432581121373365, | |
| "singles_other_seed": 0.17478828587775164, | |
| "joint_other_seed": 0.3101909168841498, | |
| "single_other_seed": 0.46140150789968803, | |
| "corrected": 0.46201734524514715 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.4.mlp.up_proj" | |
| ], | |
| "energy_share": 0.8055442947366691, | |
| "singles_other_seed": 0.14571931310330588, | |
| "joint_other_seed": 0.28494705738923176, | |
| "single_other_seed": 0.43358342280779194, | |
| "corrected": 0.414570783600845 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.4567418878626634, | |
| "singles_other_seed": 0.097631378300713, | |
| "joint_other_seed": 0.2423206534271537, | |
| "single_other_seed": 0.30428717135635786, | |
| "corrected": 0.35954427941403555 | |
| } | |
| }, | |
| "trained/trigger/r3": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.1437155031488253, | |
| "joint_other_seed": 0.44263358131634, | |
| "single_other_seed": 0.3922550950079986, | |
| "corrected": 0.34490313607917916 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.3946855417605277, | |
| "singles_other_seed": 0.2132772433955743, | |
| "joint_other_seed": 0.5988273538273052, | |
| "single_other_seed": 0.47448754105361524, | |
| "corrected": 0.40011174231126473 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.4530211567056298, | |
| "singles_other_seed": 0.21171468729824927, | |
| "joint_other_seed": 0.5838880554936026, | |
| "single_other_seed": 0.46140150789968803, | |
| "corrected": 0.407893408403436 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.mlp.gate_proj" | |
| ], | |
| "energy_share": 0.8015952699155042, | |
| "singles_other_seed": 0.15942473143472313, | |
| "joint_other_seed": 0.4864404446846313, | |
| "single_other_seed": 0.41428718749778826, | |
| "corrected": 0.35513224313690317 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.5469788432943702, | |
| "singles_other_seed": 0.08739692805938137, | |
| "joint_other_seed": 0.3256431966495047, | |
| "single_other_seed": 0.30428717135635786, | |
| "corrected": 0.2776408348205996 | |
| } | |
| }, | |
| "trained/trigger/r16": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.13180656972957439, | |
| "joint_other_seed": 0.5576826654447682, | |
| "single_other_seed": 0.3922550950079986, | |
| "corrected": 0.2818118093144182 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.307457924798905, | |
| "singles_other_seed": 0.23705516638440716, | |
| "joint_other_seed": 0.7303617741369218, | |
| "single_other_seed": 0.47448754105361524, | |
| "corrected": 0.4026873627389081 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.363349408352467, | |
| "singles_other_seed": 0.23350419460437966, | |
| "joint_other_seed": 0.7182881301016888, | |
| "single_other_seed": 0.46140150789968803, | |
| "corrected": 0.4056073693335827 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.9.mlp.gate_proj" | |
| ], | |
| "energy_share": 0.8148785145934205, | |
| "singles_other_seed": 0.143842394499755, | |
| "joint_other_seed": 0.5853879245744831, | |
| "single_other_seed": 0.40824180388484943, | |
| "corrected": 0.29424309976507323 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.636650591647533, | |
| "singles_other_seed": 0.07376567207881235, | |
| "joint_other_seed": 0.46602186856990635, | |
| "single_other_seed": 0.30428717135635786, | |
| "corrected": 0.19588867725446754 | |
| } | |
| }, | |
| "trained/all/r1": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.12395210508511545, | |
| "joint_other_seed": 0.26660045325494397, | |
| "single_other_seed": 0.3587347579194623, | |
| "corrected": 0.400808327582073 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.326815391008406, | |
| "singles_other_seed": 0.16023578481986267, | |
| "joint_other_seed": 0.2757705340725496, | |
| "single_other_seed": 0.45238485233995085, | |
| "corrected": 0.45366111354145383 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.39169174822535013, | |
| "singles_other_seed": 0.1806273595857436, | |
| "joint_other_seed": 0.3175221558962993, | |
| "single_other_seed": 0.44456814096027364, | |
| "corrected": 0.48075891656211894 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.2.mlp.gate_proj" | |
| ], | |
| "energy_share": 0.8085824071815759, | |
| "singles_other_seed": 0.13073988264759095, | |
| "joint_other_seed": 0.27645452945851867, | |
| "single_other_seed": 0.3984747441656814, | |
| "corrected": 0.393908938320103 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.6083082517746499, | |
| "singles_other_seed": 0.08745871632751645, | |
| "joint_other_seed": 0.23381179607001268, | |
| "single_other_seed": 0.29681030676713727, | |
| "corrected": 0.3319941267964208 | |
| } | |
| }, | |
| "trained/all/r3": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.11983588865585959, | |
| "joint_other_seed": 0.3841497199036973, | |
| "single_other_seed": 0.3587347579194623, | |
| "corrected": 0.3228121791395964 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.24921923231802676, | |
| "singles_other_seed": 0.2015189481875017, | |
| "joint_other_seed": 0.569102749098143, | |
| "single_other_seed": 0.45238485233995085, | |
| "corrected": 0.39716092973494466 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.3152585860867118, | |
| "singles_other_seed": 0.20590416750746898, | |
| "joint_other_seed": 0.5555516179643105, | |
| "single_other_seed": 0.44456814096027364, | |
| "corrected": 0.41431796044961683 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.4.mlp.gate_proj" | |
| ], | |
| "energy_share": 0.8089777210304151, | |
| "singles_other_seed": 0.13125875532225972, | |
| "joint_other_seed": 0.4157271679227319, | |
| "single_other_seed": 0.3887350317896131, | |
| "corrected": 0.3265105649570514 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.6847414139132881, | |
| "singles_other_seed": 0.08020959565481833, | |
| "joint_other_seed": 0.3052353751054865, | |
| "single_other_seed": 0.29681030676713727, | |
| "corrected": 0.266482936775873 | |
| } | |
| }, | |
| "trained/all/r16": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.10437008868330633, | |
| "joint_other_seed": 0.5010189390246579, | |
| "single_other_seed": 0.3587347579194623, | |
| "corrected": 0.24618537654559672 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.18280790575127404, | |
| "singles_other_seed": 0.22000687065872548, | |
| "joint_other_seed": 0.6993153859128058, | |
| "single_other_seed": 0.45238485233995085, | |
| "corrected": 0.3911521076842514 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.24255829009139795, | |
| "singles_other_seed": 0.2245219016848794, | |
| "joint_other_seed": 0.6879037357057902, | |
| "single_other_seed": 0.44456814096027364, | |
| "corrected": 0.4059996972309857 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj" | |
| ], | |
| "energy_share": 0.8096582904797117, | |
| "singles_other_seed": 0.10568579975833914, | |
| "joint_other_seed": 0.4990572551663759, | |
| "single_other_seed": 0.37053694694312844, | |
| "corrected": 0.24576820183963535 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.757441709908602, | |
| "singles_other_seed": 0.06589344033955502, | |
| "joint_other_seed": 0.4411721466787176, | |
| "single_other_seed": 0.29681030676713727, | |
| "corrected": 0.18209526426223474 | |
| } | |
| }, | |
| "learned/contrast/r1": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.15125384481176485, | |
| "joint_other_seed": 0.3145630138667902, | |
| "single_other_seed": 0.47768719633479495, | |
| "corrected": 0.3901941860274961 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.8384993539805384, | |
| "singles_other_seed": 0.1466177189131913, | |
| "joint_other_seed": 0.31186371677706715, | |
| "single_other_seed": 0.5033329585503437, | |
| "corrected": 0.3700635823197259 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.8744337907421467, | |
| "singles_other_seed": 0.15122345323743713, | |
| "joint_other_seed": 0.316334138596542, | |
| "single_other_seed": 0.4946617851687032, | |
| "corrected": 0.3822891209850408 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj" | |
| ], | |
| "energy_share": 0.8298677048717683, | |
| "singles_other_seed": 0.14594201376351568, | |
| "joint_other_seed": 0.30930009674759595, | |
| "single_other_seed": 0.5012952867644282, | |
| "corrected": 0.37063249733814435 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.12556620925785328, | |
| "singles_other_seed": 0.15146548948673766, | |
| "joint_other_seed": 0.3022290322927055, | |
| "single_other_seed": 0.3725355643353748, | |
| "corrected": 0.4514001850415748 | |
| } | |
| }, | |
| "learned/contrast/r3": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.18253792956046877, | |
| "joint_other_seed": 0.5867736225720839, | |
| "single_other_seed": 0.47768719633479495, | |
| "corrected": 0.34478314631266815 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.7914620026313561, | |
| "singles_other_seed": 0.1901347523949124, | |
| "joint_other_seed": 0.6148093925520464, | |
| "single_other_seed": 0.5033329585503437, | |
| "corrected": 0.34179332111001154 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.8364243120648378, | |
| "singles_other_seed": 0.1930461549422091, | |
| "joint_other_seed": 0.611100816860525, | |
| "single_other_seed": 0.4946617851687032, | |
| "corrected": 0.351115937921699 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.3.self_attn.q_proj" | |
| ], | |
| "energy_share": 0.804428140118467, | |
| "singles_other_seed": 0.18822558911384824, | |
| "joint_other_seed": 0.6138638112103185, | |
| "single_other_seed": 0.49952528268664287, | |
| "corrected": 0.33990997709593684 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.16357568793516222, | |
| "singles_other_seed": 0.12880540184260703, | |
| "joint_other_seed": 0.4623794842742065, | |
| "single_other_seed": 0.3725355643353748, | |
| "corrected": 0.3103494135829539 | |
| } | |
| }, | |
| "learned/contrast/r16": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.19713641808807436, | |
| "joint_other_seed": 0.7293671093306947, | |
| "single_other_seed": 0.47768719633479495, | |
| "corrected": 0.3339811635162396 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.7644381961268416, | |
| "singles_other_seed": 0.21122543038544717, | |
| "joint_other_seed": 0.7547947220725734, | |
| "single_other_seed": 0.5033329585503437, | |
| "corrected": 0.34269209912378107 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.8076620310783911, | |
| "singles_other_seed": 0.2129054744787655, | |
| "joint_other_seed": 0.7523195267969996, | |
| "single_other_seed": 0.4946617851687032, | |
| "corrected": 0.3490048653181454 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.3.mlp.up_proj" | |
| ], | |
| "energy_share": 0.8112776084516392, | |
| "singles_other_seed": 0.2115297346318411, | |
| "joint_other_seed": 0.7516165096502119, | |
| "single_other_seed": 0.4940745399841985, | |
| "corrected": 0.3471179174562325 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.1923379689216089, | |
| "singles_other_seed": 0.13091928902040595, | |
| "joint_other_seed": 0.6329857436911812, | |
| "single_other_seed": 0.3725355643353748, | |
| "corrected": 0.26960181034009784 | |
| } | |
| }, | |
| "learned/trigger/r1": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.1395954580607608, | |
| "joint_other_seed": 0.279262468079983, | |
| "single_other_seed": 0.3924589814436419, | |
| "corrected": 0.4216653026875701 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.49357112665993297, | |
| "singles_other_seed": 0.1658955885421517, | |
| "joint_other_seed": 0.2893146979553539, | |
| "single_other_seed": 0.47483789362737994, | |
| "corrected": 0.4475864225956977 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.5433399150804941, | |
| "singles_other_seed": 0.17486276476741625, | |
| "joint_other_seed": 0.3103570515311225, | |
| "single_other_seed": 0.46167870069119443, | |
| "corrected": 0.4619517456454605 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.4.mlp.up_proj" | |
| ], | |
| "energy_share": 0.805630481980365, | |
| "singles_other_seed": 0.14578507547260955, | |
| "joint_other_seed": 0.28504373292489915, | |
| "single_other_seed": 0.43383353458809515, | |
| "corrected": 0.41456798242105036 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.45666008491950594, | |
| "singles_other_seed": 0.097633972781202, | |
| "joint_other_seed": 0.24226574143420354, | |
| "single_other_seed": 0.3043758447324084, | |
| "corrected": 0.35954219608913895 | |
| } | |
| }, | |
| "learned/trigger/r3": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.1437138295151933, | |
| "joint_other_seed": 0.4426919751500006, | |
| "single_other_seed": 0.3924589814436419, | |
| "corrected": 0.344786776568601 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.39471930429002783, | |
| "singles_other_seed": 0.2132528766990518, | |
| "joint_other_seed": 0.5988311355516739, | |
| "single_other_seed": 0.47483789362737994, | |
| "corrected": 0.3999171483650462 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.4530365741708649, | |
| "singles_other_seed": 0.21166154547748595, | |
| "joint_other_seed": 0.5838978015947548, | |
| "single_other_seed": 0.46167870069119443, | |
| "corrected": 0.40766518444695454 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.mlp.gate_proj" | |
| ], | |
| "energy_share": 0.8016388513558518, | |
| "singles_other_seed": 0.15941771109840142, | |
| "joint_other_seed": 0.4864608874865288, | |
| "single_other_seed": 0.41450159344056003, | |
| "corrected": 0.35501728891126827 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.546963425829135, | |
| "singles_other_seed": 0.08743438008833482, | |
| "joint_other_seed": 0.3257346049775367, | |
| "single_other_seed": 0.3043758447324084, | |
| "corrected": 0.2776803792292405 | |
| } | |
| }, | |
| "learned/trigger/r16": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.1317981821522234, | |
| "joint_other_seed": 0.5577253448124586, | |
| "single_other_seed": 0.3924589814436419, | |
| "corrected": 0.28170988977475614 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.307488838018419, | |
| "singles_other_seed": 0.23695563833849165, | |
| "joint_other_seed": 0.7304192084037109, | |
| "single_other_seed": 0.47483789362737994, | |
| "corrected": 0.4023539501664715 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.36337579353797517, | |
| "singles_other_seed": 0.23341513377023548, | |
| "joint_other_seed": 0.7183178836828132, | |
| "single_other_seed": 0.46167870069119443, | |
| "corrected": 0.40532253633567006 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.9.mlp.gate_proj" | |
| ], | |
| "energy_share": 0.814938370373833, | |
| "singles_other_seed": 0.1438256523599334, | |
| "joint_other_seed": 0.5854284654307824, | |
| "single_other_seed": 0.4084434838267871, | |
| "corrected": 0.2941260217919656 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.6366242064620249, | |
| "singles_other_seed": 0.07379671118034037, | |
| "joint_other_seed": 0.4660614704954613, | |
| "single_other_seed": 0.3043758447324084, | |
| "corrected": 0.19593423023574458 | |
| } | |
| }, | |
| "learned/all/r1": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.12397603322869855, | |
| "joint_other_seed": 0.2666200372884032, | |
| "single_other_seed": 0.3588921907515229, | |
| "corrected": 0.40078304427966466 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.32689611788256445, | |
| "singles_other_seed": 0.16032413526414901, | |
| "joint_other_seed": 0.2759810171312006, | |
| "single_other_seed": 0.45272279652600234, | |
| "corrected": 0.4535687437436016 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.39176049934402235, | |
| "singles_other_seed": 0.18068858577620628, | |
| "joint_other_seed": 0.31763097684377706, | |
| "single_other_seed": 0.4447939935495705, | |
| "corrected": 0.48071739397886876 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.2.mlp.gate_proj" | |
| ], | |
| "energy_share": 0.8086528810981997, | |
| "singles_other_seed": 0.13077996039111317, | |
| "joint_other_seed": 0.2764757402864926, | |
| "single_other_seed": 0.39866384933995924, | |
| "corrected": 0.3939211134191288 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.6082395006559776, | |
| "singles_other_seed": 0.08744809007287879, | |
| "joint_other_seed": 0.2337644415392477, | |
| "single_other_seed": 0.29690371112192215, | |
| "corrected": 0.3319351854256884 | |
| } | |
| }, | |
| "learned/all/r3": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.11983278988434098, | |
| "joint_other_seed": 0.38421088001617765, | |
| "single_other_seed": 0.3588921907515229, | |
| "corrected": 0.3227073349089904 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.24924836028634048, | |
| "singles_other_seed": 0.20150051810058445, | |
| "joint_other_seed": 0.5690790548088691, | |
| "single_other_seed": 0.45272279652600234, | |
| "corrected": 0.39698462269013635 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.3152735922006733, | |
| "singles_other_seed": 0.20582909924296555, | |
| "joint_other_seed": 0.5555495279149192, | |
| "single_other_seed": 0.4447939935495705, | |
| "corrected": 0.4140625239365675 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.4.mlp.gate_proj" | |
| ], | |
| "energy_share": 0.8090345702900374, | |
| "singles_other_seed": 0.13124737911113937, | |
| "joint_other_seed": 0.4157391315387, | |
| "single_other_seed": 0.3888771636841787, | |
| "corrected": 0.3264179005773182 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.6847264077993267, | |
| "singles_other_seed": 0.08023687966578874, | |
| "joint_other_seed": 0.3053201721501626, | |
| "single_other_seed": 0.29690371112192215, | |
| "corrected": 0.26649463400881157 | |
| } | |
| }, | |
| "learned/all/r16": { | |
| "all": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 1.0, | |
| "singles_other_seed": 0.10437337243687561, | |
| "joint_other_seed": 0.5010556729497015, | |
| "single_other_seed": 0.3588921907515229, | |
| "corrected": 0.2461300955378225 | |
| }, | |
| "layer 0": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj" | |
| ], | |
| "energy_share": 0.18282807831487902, | |
| "singles_other_seed": 0.21991863259393096, | |
| "joint_other_seed": 0.6993578381368716, | |
| "single_other_seed": 0.45272279652600234, | |
| "corrected": 0.39083740524395033 | |
| }, | |
| "layers 0-1": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.0.self_attn.k_proj", | |
| "model.layers.0.self_attn.v_proj", | |
| "model.layers.0.self_attn.o_proj", | |
| "model.layers.0.mlp.gate_proj", | |
| "model.layers.0.mlp.up_proj", | |
| "model.layers.0.mlp.down_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.1.self_attn.k_proj", | |
| "model.layers.1.self_attn.v_proj", | |
| "model.layers.1.self_attn.o_proj", | |
| "model.layers.1.mlp.gate_proj", | |
| "model.layers.1.mlp.up_proj", | |
| "model.layers.1.mlp.down_proj" | |
| ], | |
| "energy_share": 0.24257712036626516, | |
| "singles_other_seed": 0.224450583977701, | |
| "joint_other_seed": 0.6879076205661936, | |
| "single_other_seed": 0.4447939935495705, | |
| "corrected": 0.4057665313296206 | |
| }, | |
| "top 80%": { | |
| "modules": [ | |
| "model.layers.0.self_attn.q_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.1.self_attn.q_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj" | |
| ], | |
| "energy_share": 0.809720536017201, | |
| "singles_other_seed": 0.10567591291698597, | |
| "joint_other_seed": 0.49909809180926734, | |
| "single_other_seed": 0.37063484135826485, | |
| "corrected": 0.2457027019222427 | |
| }, | |
| "layers 2-11": { | |
| "modules": [ | |
| "model.layers.2.self_attn.q_proj", | |
| "model.layers.2.self_attn.k_proj", | |
| "model.layers.2.self_attn.v_proj", | |
| "model.layers.2.self_attn.o_proj", | |
| "model.layers.2.mlp.gate_proj", | |
| "model.layers.2.mlp.up_proj", | |
| "model.layers.2.mlp.down_proj", | |
| "model.layers.3.self_attn.q_proj", | |
| "model.layers.3.self_attn.k_proj", | |
| "model.layers.3.self_attn.v_proj", | |
| "model.layers.3.self_attn.o_proj", | |
| "model.layers.3.mlp.gate_proj", | |
| "model.layers.3.mlp.up_proj", | |
| "model.layers.3.mlp.down_proj", | |
| "model.layers.4.self_attn.q_proj", | |
| "model.layers.4.self_attn.k_proj", | |
| "model.layers.4.self_attn.v_proj", | |
| "model.layers.4.self_attn.o_proj", | |
| "model.layers.4.mlp.gate_proj", | |
| "model.layers.4.mlp.up_proj", | |
| "model.layers.4.mlp.down_proj", | |
| "model.layers.5.self_attn.q_proj", | |
| "model.layers.5.self_attn.k_proj", | |
| "model.layers.5.self_attn.v_proj", | |
| "model.layers.5.self_attn.o_proj", | |
| "model.layers.5.mlp.gate_proj", | |
| "model.layers.5.mlp.up_proj", | |
| "model.layers.5.mlp.down_proj", | |
| "model.layers.6.self_attn.q_proj", | |
| "model.layers.6.self_attn.k_proj", | |
| "model.layers.6.self_attn.v_proj", | |
| "model.layers.6.self_attn.o_proj", | |
| "model.layers.6.mlp.gate_proj", | |
| "model.layers.6.mlp.up_proj", | |
| "model.layers.6.mlp.down_proj", | |
| "model.layers.7.self_attn.q_proj", | |
| "model.layers.7.self_attn.k_proj", | |
| "model.layers.7.self_attn.v_proj", | |
| "model.layers.7.self_attn.o_proj", | |
| "model.layers.7.mlp.gate_proj", | |
| "model.layers.7.mlp.up_proj", | |
| "model.layers.7.mlp.down_proj", | |
| "model.layers.8.self_attn.q_proj", | |
| "model.layers.8.self_attn.k_proj", | |
| "model.layers.8.self_attn.v_proj", | |
| "model.layers.8.self_attn.o_proj", | |
| "model.layers.8.mlp.gate_proj", | |
| "model.layers.8.mlp.up_proj", | |
| "model.layers.8.mlp.down_proj", | |
| "model.layers.9.self_attn.q_proj", | |
| "model.layers.9.self_attn.k_proj", | |
| "model.layers.9.self_attn.v_proj", | |
| "model.layers.9.self_attn.o_proj", | |
| "model.layers.9.mlp.gate_proj", | |
| "model.layers.9.mlp.up_proj", | |
| "model.layers.9.mlp.down_proj", | |
| "model.layers.10.self_attn.q_proj", | |
| "model.layers.10.self_attn.k_proj", | |
| "model.layers.10.self_attn.v_proj", | |
| "model.layers.10.self_attn.o_proj", | |
| "model.layers.10.mlp.gate_proj", | |
| "model.layers.10.mlp.up_proj", | |
| "model.layers.10.mlp.down_proj", | |
| "model.layers.11.self_attn.q_proj", | |
| "model.layers.11.self_attn.k_proj", | |
| "model.layers.11.self_attn.v_proj", | |
| "model.layers.11.self_attn.o_proj", | |
| "model.layers.11.mlp.gate_proj", | |
| "model.layers.11.mlp.up_proj", | |
| "model.layers.11.mlp.down_proj" | |
| ], | |
| "energy_share": 0.7574228796337349, | |
| "singles_other_seed": 0.0659166727775391, | |
| "joint_other_seed": 0.4412132670673298, | |
| "single_other_seed": 0.29690371112192215, | |
| "corrected": 0.1821223238477676 | |
| } | |
| } | |
| } | |
| } | |