{ "final_norm.bias": 0.004869353026151657, "final_norm.weight": 0.04349002242088318, "gpt.h.0.attn.c_attn.bias": 0.006784666329622269, "gpt.h.0.attn.c_attn.weight": 0.005739760585129261, "gpt.h.0.attn.c_proj.bias": 0.004933658055961132, "gpt.h.0.attn.c_proj.weight": 0.006425631698220968, "gpt.h.0.ln_1.bias": 0.0014453897019848228, "gpt.h.0.ln_1.weight": 0.001891627791337669, "gpt.h.0.ln_2.bias": 0.01133652776479721, "gpt.h.0.ln_2.weight": 0.006315903272479773, "gpt.h.0.mlp.c_fc.bias": 0.0026161218993365765, "gpt.h.0.mlp.c_fc.weight": 0.009393199346959591, "gpt.h.0.mlp.c_proj.bias": 0.016290107741951942, "gpt.h.0.mlp.c_proj.weight": 0.036495596170425415, "gpt.h.1.attn.c_attn.bias": 0.005060207564383745, "gpt.h.1.attn.c_attn.weight": 0.003430855693295598, "gpt.h.1.attn.c_proj.bias": 0.01638360321521759, "gpt.h.1.attn.c_proj.weight": 0.025163328275084496, "gpt.h.1.ln_1.bias": 0.011049080640077591, "gpt.h.1.ln_1.weight": 0.004219602793455124, "gpt.h.1.ln_2.bias": 0.011916977353394032, "gpt.h.1.ln_2.weight": 0.011172252707183361, "gpt.h.1.mlp.c_fc.bias": 0.0014588515041396022, "gpt.h.1.mlp.c_fc.weight": 0.006692858878523111, "gpt.h.1.mlp.c_proj.bias": 0.018994668498635292, "gpt.h.1.mlp.c_proj.weight": 0.04086727276444435, "gpt.h.10.attn.c_attn.bias": 0.0024330769665539265, "gpt.h.10.attn.c_attn.weight": 0.006525476463139057, "gpt.h.10.attn.c_proj.bias": 0.0025137888733297586, "gpt.h.10.attn.c_proj.weight": 0.004787968005985022, "gpt.h.10.ln_1.bias": 0.008025330491364002, "gpt.h.10.ln_1.weight": 0.008399507962167263, "gpt.h.10.ln_2.bias": 0.007188895717263222, "gpt.h.10.ln_2.weight": 0.01416360680013895, "gpt.h.10.mlp.c_fc.bias": 0.0016273388173431158, "gpt.h.10.mlp.c_fc.weight": 0.005328207276761532, "gpt.h.10.mlp.c_proj.bias": 0.005671263672411442, "gpt.h.10.mlp.c_proj.weight": 0.02014252357184887, "gpt.h.11.attn.c_attn.bias": 0.002672715811058879, "gpt.h.11.attn.c_attn.weight": 0.0061920033767819405, "gpt.h.11.attn.c_proj.bias": 0.0028154668398201466, "gpt.h.11.attn.c_proj.weight": 0.004848834592849016, "gpt.h.11.ln_1.bias": 0.009074854664504528, "gpt.h.11.ln_1.weight": 0.008892943151295185, "gpt.h.11.ln_2.bias": 0.0054258741438388824, "gpt.h.11.ln_2.weight": 0.01229905430227518, "gpt.h.11.mlp.c_fc.bias": 0.00168695580214262, "gpt.h.11.mlp.c_fc.weight": 0.003726626979187131, "gpt.h.11.mlp.c_proj.bias": 0.005866497289389372, "gpt.h.11.mlp.c_proj.weight": 0.02522459626197815, "gpt.h.12.attn.c_attn.bias": 0.0027519443538039923, "gpt.h.12.attn.c_attn.weight": 0.007994277402758598, "gpt.h.12.attn.c_proj.bias": 0.001845111371949315, "gpt.h.12.attn.c_proj.weight": 0.006088561844080687, "gpt.h.12.ln_1.bias": 0.010259520262479782, "gpt.h.12.ln_1.weight": 0.00929665844887495, "gpt.h.12.ln_2.bias": 0.007012453395873308, "gpt.h.12.ln_2.weight": 0.012815050780773163, "gpt.h.12.mlp.c_fc.bias": 0.0008164360770024359, "gpt.h.12.mlp.c_fc.weight": 0.0037948056124150753, "gpt.h.12.mlp.c_proj.bias": 0.006633814424276352, "gpt.h.12.mlp.c_proj.weight": 0.01512092724442482, "gpt.h.13.attn.c_attn.bias": 0.0026236390694975853, "gpt.h.13.attn.c_attn.weight": 0.004803134128451347, "gpt.h.13.attn.c_proj.bias": 0.00371875730343163, "gpt.h.13.attn.c_proj.weight": 0.006528449710458517, "gpt.h.13.ln_1.bias": 0.008354649879038334, "gpt.h.13.ln_1.weight": 0.009035688824951649, "gpt.h.13.ln_2.bias": 0.00720137357711792, "gpt.h.13.ln_2.weight": 0.012263036333024502, "gpt.h.13.mlp.c_fc.bias": 0.0010731430957093835, "gpt.h.13.mlp.c_fc.weight": 0.0035445760004222393, "gpt.h.13.mlp.c_proj.bias": 0.006077663507312536, "gpt.h.13.mlp.c_proj.weight": 0.013959192670881748, "gpt.h.14.attn.c_attn.bias": 0.0022155672777444124, "gpt.h.14.attn.c_attn.weight": 0.007756410166621208, "gpt.h.14.attn.c_proj.bias": 0.0013033421710133553, "gpt.h.14.attn.c_proj.weight": 0.005600381642580032, "gpt.h.14.ln_1.bias": 0.008806071244180202, "gpt.h.14.ln_1.weight": 0.008939859457314014, "gpt.h.14.ln_2.bias": 0.007173170801252127, "gpt.h.14.ln_2.weight": 0.011766878888010979, "gpt.h.14.mlp.c_fc.bias": 0.000844161375425756, "gpt.h.14.mlp.c_fc.weight": 0.0045441873371601105, "gpt.h.14.mlp.c_proj.bias": 0.005780238192528486, "gpt.h.14.mlp.c_proj.weight": 0.01581657864153385, "gpt.h.15.attn.c_attn.bias": 0.0022497144527733326, "gpt.h.15.attn.c_attn.weight": 0.004309875890612602, "gpt.h.15.attn.c_proj.bias": 0.004680809564888477, "gpt.h.15.attn.c_proj.weight": 0.006447209510952234, "gpt.h.15.ln_1.bias": 0.007965384982526302, "gpt.h.15.ln_1.weight": 0.00913836620748043, "gpt.h.15.ln_2.bias": 0.0062019070610404015, "gpt.h.15.ln_2.weight": 0.01104413066059351, "gpt.h.15.mlp.c_fc.bias": 0.0011065590661019087, "gpt.h.15.mlp.c_fc.weight": 0.0046559604816138744, "gpt.h.15.mlp.c_proj.bias": 0.005378104280680418, "gpt.h.15.mlp.c_proj.weight": 0.017130374908447266, "gpt.h.16.attn.c_attn.bias": 0.002614511176943779, "gpt.h.16.attn.c_attn.weight": 0.0070121572352945805, "gpt.h.16.attn.c_proj.bias": 0.0052946810610592365, "gpt.h.16.attn.c_proj.weight": 0.004184126853942871, "gpt.h.16.ln_1.bias": 0.00865858979523182, "gpt.h.16.ln_1.weight": 0.008769914507865906, "gpt.h.16.ln_2.bias": 0.0066169509664177895, "gpt.h.16.ln_2.weight": 0.010157736949622631, "gpt.h.16.mlp.c_fc.bias": 0.0012236940674483776, "gpt.h.16.mlp.c_fc.weight": 0.004414326511323452, "gpt.h.16.mlp.c_proj.bias": 0.005387744400650263, "gpt.h.16.mlp.c_proj.weight": 0.008935753256082535, "gpt.h.17.attn.c_attn.bias": 0.0024349491577595472, "gpt.h.17.attn.c_attn.weight": 0.007278294302523136, "gpt.h.17.attn.c_proj.bias": 0.0034934559371322393, "gpt.h.17.attn.c_proj.weight": 0.007657876703888178, "gpt.h.17.ln_1.bias": 0.008721367456018925, "gpt.h.17.ln_1.weight": 0.00882012490183115, "gpt.h.17.ln_2.bias": 0.007266780827194452, "gpt.h.17.ln_2.weight": 0.010302619077265263, "gpt.h.17.mlp.c_fc.bias": 0.0013675907393917441, "gpt.h.17.mlp.c_fc.weight": 0.0037777191027998924, "gpt.h.17.mlp.c_proj.bias": 0.0048697893507778645, "gpt.h.17.mlp.c_proj.weight": 0.01015142910182476, "gpt.h.18.attn.c_attn.bias": 0.002071398077532649, "gpt.h.18.attn.c_attn.weight": 0.006754225119948387, "gpt.h.18.attn.c_proj.bias": 0.006200749427080154, "gpt.h.18.attn.c_proj.weight": 0.005378039088100195, "gpt.h.18.ln_1.bias": 0.008033838123083115, "gpt.h.18.ln_1.weight": 0.00875889416784048, "gpt.h.18.ln_2.bias": 0.006478333845734596, "gpt.h.18.ln_2.weight": 0.009868012741208076, "gpt.h.18.mlp.c_fc.bias": 0.0011885170824825764, "gpt.h.18.mlp.c_fc.weight": 0.003764205379411578, "gpt.h.18.mlp.c_proj.bias": 0.0037589711137115955, "gpt.h.18.mlp.c_proj.weight": 0.00706329382956028, "gpt.h.19.attn.c_attn.bias": 0.002585803624242544, "gpt.h.19.attn.c_attn.weight": 0.006440048571676016, "gpt.h.19.attn.c_proj.bias": 0.003273698966950178, "gpt.h.19.attn.c_proj.weight": 0.006902920547872782, "gpt.h.19.ln_1.bias": 0.008147234097123146, "gpt.h.19.ln_1.weight": 0.008605051785707474, "gpt.h.19.ln_2.bias": 0.0059887804090976715, "gpt.h.19.ln_2.weight": 0.009857536293566227, "gpt.h.19.mlp.c_fc.bias": 0.0012922842288389802, "gpt.h.19.mlp.c_fc.weight": 0.004283702466636896, "gpt.h.19.mlp.c_proj.bias": 0.0032340157777071, "gpt.h.19.mlp.c_proj.weight": 0.008396429941058159, "gpt.h.2.attn.c_attn.bias": 0.0024704847019165754, "gpt.h.2.attn.c_attn.weight": 0.0033952328376471996, "gpt.h.2.attn.c_proj.bias": 0.02068011462688446, "gpt.h.2.attn.c_proj.weight": 0.02372979000210762, "gpt.h.2.ln_1.bias": 0.009298523887991905, "gpt.h.2.ln_1.weight": 0.005359284114092588, "gpt.h.2.ln_2.bias": 0.008068359456956387, "gpt.h.2.ln_2.weight": 0.013421551324427128, "gpt.h.2.mlp.c_fc.bias": 0.0030155687127262354, "gpt.h.2.mlp.c_fc.weight": 0.01074832584708929, "gpt.h.2.mlp.c_proj.bias": 0.015278116799890995, "gpt.h.2.mlp.c_proj.weight": 0.04022492840886116, "gpt.h.20.attn.c_attn.bias": 0.001981201348826289, "gpt.h.20.attn.c_attn.weight": 0.006853727623820305, "gpt.h.20.attn.c_proj.bias": 0.004991447553038597, "gpt.h.20.attn.c_proj.weight": 0.005948251578956842, "gpt.h.20.ln_1.bias": 0.007368254475295544, "gpt.h.20.ln_1.weight": 0.00863717496395111, "gpt.h.20.ln_2.bias": 0.006148574873805046, "gpt.h.20.ln_2.weight": 0.009219009429216385, "gpt.h.20.mlp.c_fc.bias": 0.0008407220593653619, "gpt.h.20.mlp.c_fc.weight": 0.003581109456717968, "gpt.h.20.mlp.c_proj.bias": 0.0023647320922464132, "gpt.h.20.mlp.c_proj.weight": 0.014833961613476276, "gpt.h.21.attn.c_attn.bias": 0.0024412847124040127, "gpt.h.21.attn.c_attn.weight": 0.005555002484470606, "gpt.h.21.attn.c_proj.bias": 0.004358917940407991, "gpt.h.21.attn.c_proj.weight": 0.009266942739486694, "gpt.h.21.ln_1.bias": 0.006839239504188299, "gpt.h.21.ln_1.weight": 0.008790756575763226, "gpt.h.21.ln_2.bias": 0.005996923428028822, "gpt.h.21.ln_2.weight": 0.008705759420990944, "gpt.h.21.mlp.c_fc.bias": 0.0011871765600517392, "gpt.h.21.mlp.c_fc.weight": 0.0038676916155964136, "gpt.h.21.mlp.c_proj.bias": 0.0013832782860845327, "gpt.h.21.mlp.c_proj.weight": 0.011286932043731213, "gpt.h.22.attn.c_attn.bias": 0.0022080065682530403, "gpt.h.22.attn.c_attn.weight": 0.00387361366301775, "gpt.h.22.attn.c_proj.bias": 0.0037303422577679157, "gpt.h.22.attn.c_proj.weight": 0.00496130483224988, "gpt.h.22.ln_1.bias": 0.006414500530809164, "gpt.h.22.ln_1.weight": 0.008626559749245644, "gpt.h.22.ln_2.bias": 0.006547602824866772, "gpt.h.22.ln_2.weight": 0.008793600834906101, "gpt.h.22.mlp.c_fc.bias": 0.0009473760146647692, "gpt.h.22.mlp.c_fc.weight": 0.00415464723482728, "gpt.h.22.mlp.c_proj.bias": 0.0013687954051420093, "gpt.h.22.mlp.c_proj.weight": 0.009641721844673157, "gpt.h.23.attn.c_attn.bias": 0.002031621988862753, "gpt.h.23.attn.c_attn.weight": 0.004895014222711325, "gpt.h.23.attn.c_proj.bias": 0.0015391460619866848, "gpt.h.23.attn.c_proj.weight": 0.005748114548623562, "gpt.h.23.ln_1.bias": 0.005657768342643976, "gpt.h.23.ln_1.weight": 0.00802789069712162, "gpt.h.23.ln_2.bias": 0.006499188020825386, "gpt.h.23.ln_2.weight": 0.008431111462414265, "gpt.h.23.mlp.c_fc.bias": 0.0010094406316056848, "gpt.h.23.mlp.c_fc.weight": 0.003703867318108678, "gpt.h.23.mlp.c_proj.bias": 0.0017025587148964405, "gpt.h.23.mlp.c_proj.weight": 0.0074964226223528385, "gpt.h.24.attn.c_attn.bias": 0.002305609406903386, "gpt.h.24.attn.c_attn.weight": 0.0031796761322766542, "gpt.h.24.attn.c_proj.bias": 0.001652697566896677, "gpt.h.24.attn.c_proj.weight": 0.0052749463357031345, "gpt.h.24.ln_1.bias": 0.005675694905221462, "gpt.h.24.ln_1.weight": 0.009069245308637619, "gpt.h.24.ln_2.bias": 0.006356207653880119, "gpt.h.24.ln_2.weight": 0.009231998585164547, "gpt.h.24.mlp.c_fc.bias": 0.001341491937637329, "gpt.h.24.mlp.c_fc.weight": 0.003126475727185607, "gpt.h.24.mlp.c_proj.bias": 0.0017264962662011385, "gpt.h.24.mlp.c_proj.weight": 0.004613065160810947, "gpt.h.25.attn.c_attn.bias": 0.002419382333755493, "gpt.h.25.attn.c_attn.weight": 0.003242518287152052, "gpt.h.25.attn.c_proj.bias": 0.0019212045008316636, "gpt.h.25.attn.c_proj.weight": 0.004869922064244747, "gpt.h.25.ln_1.bias": 0.005121177528053522, "gpt.h.25.ln_1.weight": 0.009873942472040653, "gpt.h.25.ln_2.bias": 0.006484670098870993, "gpt.h.25.ln_2.weight": 0.009337698109447956, "gpt.h.25.mlp.c_fc.bias": 0.001216022646985948, "gpt.h.25.mlp.c_fc.weight": 0.0038542866241186857, "gpt.h.25.mlp.c_proj.bias": 0.002242410322651267, "gpt.h.25.mlp.c_proj.weight": 0.005041209515184164, "gpt.h.26.attn.c_attn.bias": 0.00450435234233737, "gpt.h.26.attn.c_attn.weight": 0.0034188455902040005, "gpt.h.26.attn.c_proj.bias": 0.002108540153130889, "gpt.h.26.attn.c_proj.weight": 0.005954994820058346, "gpt.h.26.ln_1.bias": 0.004473517648875713, "gpt.h.26.ln_1.weight": 0.009424147196114063, "gpt.h.26.ln_2.bias": 0.006467711646109819, "gpt.h.26.ln_2.weight": 0.009836550801992416, "gpt.h.26.mlp.c_fc.bias": 0.001742089632898569, "gpt.h.26.mlp.c_fc.weight": 0.0029066475108265877, "gpt.h.26.mlp.c_proj.bias": 0.002795786364004016, "gpt.h.26.mlp.c_proj.weight": 0.00789736956357956, "gpt.h.27.attn.c_attn.bias": 0.0031362660229206085, "gpt.h.27.attn.c_attn.weight": 0.003957696724683046, "gpt.h.27.attn.c_proj.bias": 0.0029156384989619255, "gpt.h.27.attn.c_proj.weight": 0.005406576208770275, "gpt.h.27.ln_1.bias": 0.004558059852570295, "gpt.h.27.ln_1.weight": 0.0094691701233387, "gpt.h.27.ln_2.bias": 0.006574852392077446, "gpt.h.27.ln_2.weight": 0.009961778298020363, "gpt.h.27.mlp.c_fc.bias": 0.0016685383161529899, "gpt.h.27.mlp.c_fc.weight": 0.0029884425457566977, "gpt.h.27.mlp.c_proj.bias": 0.002400637837126851, "gpt.h.27.mlp.c_proj.weight": 0.008324341848492622, "gpt.h.28.attn.c_attn.bias": 0.0024894936941564083, "gpt.h.28.attn.c_attn.weight": 0.003081751521676779, "gpt.h.28.attn.c_proj.bias": 0.004242106340825558, "gpt.h.28.attn.c_proj.weight": 0.009405862540006638, "gpt.h.28.ln_1.bias": 0.004887533839792013, "gpt.h.28.ln_1.weight": 0.009970242157578468, "gpt.h.28.ln_2.bias": 0.005814329255372286, "gpt.h.28.ln_2.weight": 0.010871784761548042, "gpt.h.28.mlp.c_fc.bias": 0.0019840362947434187, "gpt.h.28.mlp.c_fc.weight": 0.008523846976459026, "gpt.h.28.mlp.c_proj.bias": 0.005565112456679344, "gpt.h.28.mlp.c_proj.weight": 0.032014038413763046, "gpt.h.29.attn.c_attn.bias": 0.00337958219461143, "gpt.h.29.attn.c_attn.weight": 0.004022927954792976, "gpt.h.29.attn.c_proj.bias": 0.0018416948150843382, "gpt.h.29.attn.c_proj.weight": 0.018323034048080444, "gpt.h.29.ln_1.bias": 0.004602092783898115, "gpt.h.29.ln_1.weight": 0.009674133732914925, "gpt.h.29.ln_2.bias": 0.003327032318338752, "gpt.h.29.ln_2.weight": 0.01144911628216505, "gpt.h.29.mlp.c_fc.bias": 0.002207099460065365, "gpt.h.29.mlp.c_fc.weight": 0.014188914559781551, "gpt.h.29.mlp.c_proj.bias": 0.00442220363765955, "gpt.h.29.mlp.c_proj.weight": 0.0867144837975502, "gpt.h.3.attn.c_attn.bias": 0.002024676650762558, "gpt.h.3.attn.c_attn.weight": 0.004979935009032488, "gpt.h.3.attn.c_proj.bias": 0.021005922928452492, "gpt.h.3.attn.c_proj.weight": 0.028328293934464455, "gpt.h.3.ln_1.bias": 0.008041512221097946, "gpt.h.3.ln_1.weight": 0.006430978886783123, "gpt.h.3.ln_2.bias": 0.006599007174372673, "gpt.h.3.ln_2.weight": 0.013134118169546127, "gpt.h.3.mlp.c_fc.bias": 0.002585547510534525, "gpt.h.3.mlp.c_fc.weight": 0.009125065989792347, "gpt.h.3.mlp.c_proj.bias": 0.01302343886345625, "gpt.h.3.mlp.c_proj.weight": 0.05233204737305641, "gpt.h.4.attn.c_attn.bias": 0.002189196879044175, "gpt.h.4.attn.c_attn.weight": 0.003678702749311924, "gpt.h.4.attn.c_proj.bias": 0.017389798536896706, "gpt.h.4.attn.c_proj.weight": 0.01729857549071312, "gpt.h.4.ln_1.bias": 0.007602763827890158, "gpt.h.4.ln_1.weight": 0.0074179875664412975, "gpt.h.4.ln_2.bias": 0.004727063234895468, "gpt.h.4.ln_2.weight": 0.013179998844861984, "gpt.h.4.mlp.c_fc.bias": 0.0009291492169722915, "gpt.h.4.mlp.c_fc.weight": 0.005233710631728172, "gpt.h.4.mlp.c_proj.bias": 0.007813659496605396, "gpt.h.4.mlp.c_proj.weight": 0.056338295340538025, "gpt.h.5.attn.c_attn.bias": 0.0016779249999672174, "gpt.h.5.attn.c_attn.weight": 0.0030687786638736725, "gpt.h.5.attn.c_proj.bias": 0.009757072664797306, "gpt.h.5.attn.c_proj.weight": 0.007529787719249725, "gpt.h.5.ln_1.bias": 0.006044740788638592, "gpt.h.5.ln_1.weight": 0.009458239190280437, "gpt.h.5.ln_2.bias": 0.0036309966817498207, "gpt.h.5.ln_2.weight": 0.013853815384209156, "gpt.h.5.mlp.c_fc.bias": 0.0011706397635862231, "gpt.h.5.mlp.c_fc.weight": 0.004535002168267965, "gpt.h.5.mlp.c_proj.bias": 0.00503140315413475, "gpt.h.5.mlp.c_proj.weight": 0.04938670992851257, "gpt.h.6.attn.c_attn.bias": 0.0022753833327442408, "gpt.h.6.attn.c_attn.weight": 0.0051308972761034966, "gpt.h.6.attn.c_proj.bias": 0.005487145856022835, "gpt.h.6.attn.c_proj.weight": 0.005496611353009939, "gpt.h.6.ln_1.bias": 0.0058417185209691525, "gpt.h.6.ln_1.weight": 0.011210721917450428, "gpt.h.6.ln_2.bias": 0.004622064530849457, "gpt.h.6.ln_2.weight": 0.01402156613767147, "gpt.h.6.mlp.c_fc.bias": 0.0014565958408638835, "gpt.h.6.mlp.c_fc.weight": 0.006464360281825066, "gpt.h.6.mlp.c_proj.bias": 0.004647988360375166, "gpt.h.6.mlp.c_proj.weight": 0.048505980521440506, "gpt.h.7.attn.c_attn.bias": 0.002493516309186816, "gpt.h.7.attn.c_attn.weight": 0.005425372160971165, "gpt.h.7.attn.c_proj.bias": 0.005013930611312389, "gpt.h.7.attn.c_proj.weight": 0.0051925997249782085, "gpt.h.7.ln_1.bias": 0.008432278409600258, "gpt.h.7.ln_1.weight": 0.0083107715472579, "gpt.h.7.ln_2.bias": 0.007790034171193838, "gpt.h.7.ln_2.weight": 0.01420080941170454, "gpt.h.7.mlp.c_fc.bias": 0.0018345315475016832, "gpt.h.7.mlp.c_fc.weight": 0.0038694925606250763, "gpt.h.7.mlp.c_proj.bias": 0.005904932040721178, "gpt.h.7.mlp.c_proj.weight": 0.023479172959923744, "gpt.h.8.attn.c_attn.bias": 0.00208841310814023, "gpt.h.8.attn.c_attn.weight": 0.004913092590868473, "gpt.h.8.attn.c_proj.bias": 0.00487521244212985, "gpt.h.8.attn.c_proj.weight": 0.0052477335557341576, "gpt.h.8.ln_1.bias": 0.006899502594023943, "gpt.h.8.ln_1.weight": 0.009084747172892094, "gpt.h.8.ln_2.bias": 0.007770676631480455, "gpt.h.8.ln_2.weight": 0.013765494339168072, "gpt.h.8.mlp.c_fc.bias": 0.0014608169440180063, "gpt.h.8.mlp.c_fc.weight": 0.0037384077440947294, "gpt.h.8.mlp.c_proj.bias": 0.005353797692805529, "gpt.h.8.mlp.c_proj.weight": 0.02410462312400341, "gpt.h.9.attn.c_attn.bias": 0.0024391484912484884, "gpt.h.9.attn.c_attn.weight": 0.0065130204893648624, "gpt.h.9.attn.c_proj.bias": 0.003902623662725091, "gpt.h.9.attn.c_proj.weight": 0.0059540290385484695, "gpt.h.9.ln_1.bias": 0.008465644903481007, "gpt.h.9.ln_1.weight": 0.008637243881821632, "gpt.h.9.ln_2.bias": 0.007591988891363144, "gpt.h.9.ln_2.weight": 0.014198592863976955, "gpt.h.9.mlp.c_fc.bias": 0.0017329183174297214, "gpt.h.9.mlp.c_fc.weight": 0.004214098211377859, "gpt.h.9.mlp.c_proj.bias": 0.005952149163931608, "gpt.h.9.mlp.c_proj.weight": 0.024082494899630547, "gpt.ln_f.bias": 0.011065124534070492, "gpt.ln_f.weight": 0.022843239828944206, "gpt.wpe.emb.weight": 0.004260858986526728, "gpt.wte.weight": 0.0044210017658770084, "mel_head.bias": 0.0015680863289162517, "mel_head.weight": 0.004180184565484524 }