| { | |
| "final_norm.bias": { | |
| "scale": 0.6184078454971313, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "final_norm.weight": { | |
| "scale": 5.523232936859131, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.0.attn.c_attn.bias": { | |
| "scale": 0.8616526126861572, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.0.attn.c_attn.weight": { | |
| "scale": 0.7289496064186096, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.0.attn.c_proj.bias": { | |
| "scale": 0.6265745759010315, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.0.attn.c_proj.weight": { | |
| "scale": 0.8160552382469177, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.0.ln_1.bias": { | |
| "scale": 0.18356449902057648, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.0.ln_1.weight": { | |
| "scale": 0.24023672938346863, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.0.ln_2.bias": { | |
| "scale": 1.4397389888763428, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.0.ln_2.weight": { | |
| "scale": 0.8021197319030762, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.0.mlp.c_fc.bias": { | |
| "scale": 0.3322474956512451, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.0.mlp.c_fc.weight": { | |
| "scale": 1.1929363012313843, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.0.mlp.c_proj.bias": { | |
| "scale": 2.0688436031341553, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.0.mlp.c_proj.weight": { | |
| "scale": 4.6349406242370605, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.1.attn.c_attn.bias": { | |
| "scale": 0.6426463723182678, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.1.attn.c_attn.weight": { | |
| "scale": 0.43571868538856506, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.1.attn.c_proj.bias": { | |
| "scale": 2.0807175636291504, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.1.attn.c_proj.weight": { | |
| "scale": 3.195742607116699, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.1.ln_1.bias": { | |
| "scale": 1.403233289718628, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.1.ln_1.weight": { | |
| "scale": 0.5358895659446716, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.1.ln_2.bias": { | |
| "scale": 1.513456106185913, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.1.ln_2.weight": { | |
| "scale": 1.418876051902771, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.1.mlp.c_fc.bias": { | |
| "scale": 0.185274139046669, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.1.mlp.c_fc.weight": { | |
| "scale": 0.8499930500984192, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.1.mlp.c_proj.bias": { | |
| "scale": 2.412322998046875, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.1.mlp.c_proj.weight": { | |
| "scale": 5.190143585205078, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.10.attn.c_attn.bias": { | |
| "scale": 0.30900076031684875, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.10.attn.c_attn.weight": { | |
| "scale": 0.8287355303764343, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.10.attn.c_proj.bias": { | |
| "scale": 0.3192511796951294, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.10.attn.c_proj.weight": { | |
| "scale": 0.6080719232559204, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.10.ln_1.bias": { | |
| "scale": 1.0192170143127441, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.10.ln_1.weight": { | |
| "scale": 1.0667375326156616, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.10.ln_2.bias": { | |
| "scale": 0.9129897356033325, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.10.ln_2.weight": { | |
| "scale": 1.7987780570983887, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.10.mlp.c_fc.bias": { | |
| "scale": 0.20667202770709991, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.10.mlp.c_fc.weight": { | |
| "scale": 0.6766823530197144, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.10.mlp.c_proj.bias": { | |
| "scale": 0.7202504873275757, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.10.mlp.c_proj.weight": { | |
| "scale": 2.558100461959839, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.11.attn.c_attn.bias": { | |
| "scale": 0.3394349217414856, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.11.attn.c_attn.weight": { | |
| "scale": 0.7863844037055969, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.11.attn.c_proj.bias": { | |
| "scale": 0.3575643002986908, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.11.attn.c_proj.weight": { | |
| "scale": 0.615801990032196, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.11.ln_1.bias": { | |
| "scale": 1.1525065898895264, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.11.ln_1.weight": { | |
| "scale": 1.129403829574585, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.11.ln_2.bias": { | |
| "scale": 0.6890860199928284, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.11.ln_2.weight": { | |
| "scale": 1.56197988986969, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.11.mlp.c_fc.bias": { | |
| "scale": 0.21424338221549988, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.11.mlp.c_fc.weight": { | |
| "scale": 0.4732816219329834, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.11.mlp.c_proj.bias": { | |
| "scale": 0.7450451850891113, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.11.mlp.c_proj.weight": { | |
| "scale": 3.203523635864258, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.12.attn.c_attn.bias": { | |
| "scale": 0.3494969308376312, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.12.attn.c_attn.weight": { | |
| "scale": 1.0152732133865356, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.12.attn.c_proj.bias": { | |
| "scale": 0.23432914912700653, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.12.attn.c_proj.weight": { | |
| "scale": 0.7732473611831665, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.12.ln_1.bias": { | |
| "scale": 1.3029590845108032, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.12.ln_1.weight": { | |
| "scale": 1.1806756258010864, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.12.ln_2.bias": { | |
| "scale": 0.8905816078186035, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.12.ln_2.weight": { | |
| "scale": 1.6275115013122559, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.12.mlp.c_fc.bias": { | |
| "scale": 0.10368738323450089, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.12.mlp.c_fc.weight": { | |
| "scale": 0.48194029927253723, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.12.mlp.c_proj.bias": { | |
| "scale": 0.8424944281578064, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.12.mlp.c_proj.weight": { | |
| "scale": 1.9203577041625977, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.13.attn.c_attn.bias": { | |
| "scale": 0.33320215344429016, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.13.attn.c_attn.weight": { | |
| "scale": 0.6099980473518372, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.13.attn.c_proj.bias": { | |
| "scale": 0.47228217124938965, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.13.attn.c_proj.weight": { | |
| "scale": 0.8291131258010864, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.13.ln_1.bias": { | |
| "scale": 1.0610405206680298, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.13.ln_1.weight": { | |
| "scale": 1.1475324630737305, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.13.ln_2.bias": { | |
| "scale": 0.9145744442939758, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.13.ln_2.weight": { | |
| "scale": 1.5574055910110474, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.13.mlp.c_fc.bias": { | |
| "scale": 0.13628917932510376, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.13.mlp.c_fc.weight": { | |
| "scale": 0.4501611590385437, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.13.mlp.c_proj.bias": { | |
| "scale": 0.7718632817268372, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.13.mlp.c_proj.weight": { | |
| "scale": 1.7728174924850464, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.14.attn.c_attn.bias": { | |
| "scale": 0.28137704730033875, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.14.attn.c_attn.weight": { | |
| "scale": 0.9850640892982483, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.14.attn.c_proj.bias": { | |
| "scale": 0.1655244529247284, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.14.attn.c_proj.weight": { | |
| "scale": 0.7112484574317932, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.14.ln_1.bias": { | |
| "scale": 1.1183710098266602, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.14.ln_1.weight": { | |
| "scale": 1.135362148284912, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.14.ln_2.bias": { | |
| "scale": 0.9109926819801331, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.14.ln_2.weight": { | |
| "scale": 1.4943935871124268, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.14.mlp.c_fc.bias": { | |
| "scale": 0.1072084978222847, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.14.mlp.c_fc.weight": { | |
| "scale": 0.5771117806434631, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.14.mlp.c_proj.bias": { | |
| "scale": 0.734090268611908, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.14.mlp.c_proj.weight": { | |
| "scale": 2.0087053775787354, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.15.attn.c_attn.bias": { | |
| "scale": 0.28571373224258423, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.15.attn.c_attn.weight": { | |
| "scale": 0.5473542213439941, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.15.attn.c_proj.bias": { | |
| "scale": 0.5944628119468689, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.15.attn.c_proj.weight": { | |
| "scale": 0.8187956213951111, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.15.ln_1.bias": { | |
| "scale": 1.0116039514541626, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.15.ln_1.weight": { | |
| "scale": 1.1605725288391113, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.15.ln_2.bias": { | |
| "scale": 0.7876421809196472, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.15.ln_2.weight": { | |
| "scale": 1.402604579925537, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.15.mlp.c_fc.bias": { | |
| "scale": 0.14053300023078918, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.15.mlp.c_fc.weight": { | |
| "scale": 0.5913069844245911, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.15.mlp.c_proj.bias": { | |
| "scale": 0.68301922082901, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.15.mlp.c_proj.weight": { | |
| "scale": 2.1755576133728027, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.16.attn.c_attn.bias": { | |
| "scale": 0.332042932510376, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.16.attn.c_attn.weight": { | |
| "scale": 0.8905439972877502, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.16.attn.c_proj.bias": { | |
| "scale": 0.6724244952201843, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.16.attn.c_proj.weight": { | |
| "scale": 0.5313841104507446, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.16.ln_1.bias": { | |
| "scale": 1.0996408462524414, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.16.ln_1.weight": { | |
| "scale": 1.1137791872024536, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.16.ln_2.bias": { | |
| "scale": 0.8403527736663818, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.16.ln_2.weight": { | |
| "scale": 1.2900326251983643, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.16.mlp.c_fc.bias": { | |
| "scale": 0.15540914237499237, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.16.mlp.c_fc.weight": { | |
| "scale": 0.5606194734573364, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.16.mlp.c_proj.bias": { | |
| "scale": 0.6842435598373413, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.16.mlp.c_proj.weight": { | |
| "scale": 1.1348406076431274, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.17.attn.c_attn.bias": { | |
| "scale": 0.3092385530471802, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.17.attn.c_attn.weight": { | |
| "scale": 0.9243433475494385, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.17.attn.c_proj.bias": { | |
| "scale": 0.4436689019203186, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.17.attn.c_proj.weight": { | |
| "scale": 0.9725503325462341, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.17.ln_1.bias": { | |
| "scale": 1.1076136827468872, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.17.ln_1.weight": { | |
| "scale": 1.1201558113098145, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.17.ln_2.bias": { | |
| "scale": 0.9228811860084534, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.17.ln_2.weight": { | |
| "scale": 1.3084325790405273, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.17.mlp.c_fc.bias": { | |
| "scale": 0.1736840307712555, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.17.mlp.c_fc.weight": { | |
| "scale": 0.47977033257484436, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.17.mlp.c_proj.bias": { | |
| "scale": 0.6184632182121277, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.17.mlp.c_proj.weight": { | |
| "scale": 1.289231538772583, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.18.attn.c_attn.bias": { | |
| "scale": 0.2630675435066223, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.18.attn.c_attn.weight": { | |
| "scale": 0.8577865958213806, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.18.attn.c_proj.bias": { | |
| "scale": 0.7874951958656311, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.18.attn.c_proj.weight": { | |
| "scale": 0.6830109357833862, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.18.ln_1.bias": { | |
| "scale": 1.0202974081039429, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.18.ln_1.weight": { | |
| "scale": 1.112379550933838, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.18.ln_2.bias": { | |
| "scale": 0.8227484226226807, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.18.ln_2.weight": { | |
| "scale": 1.2532376050949097, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.18.mlp.c_fc.bias": { | |
| "scale": 0.1509416699409485, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.18.mlp.c_fc.weight": { | |
| "scale": 0.47805407643318176, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.18.mlp.c_proj.bias": { | |
| "scale": 0.4773893356323242, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.18.mlp.c_proj.weight": { | |
| "scale": 0.8970383405685425, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.19.attn.c_attn.bias": { | |
| "scale": 0.32839706540107727, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.19.attn.c_attn.weight": { | |
| "scale": 0.8178861737251282, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.19.attn.c_proj.bias": { | |
| "scale": 0.41575977206230164, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.19.attn.c_proj.weight": { | |
| "scale": 0.8766708970069885, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.19.ln_1.bias": { | |
| "scale": 1.034698724746704, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.19.ln_1.weight": { | |
| "scale": 1.092841625213623, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.19.ln_2.bias": { | |
| "scale": 0.7605751156806946, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.19.ln_2.weight": { | |
| "scale": 1.2519071102142334, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.19.mlp.c_fc.bias": { | |
| "scale": 0.16412009298801422, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.19.mlp.c_fc.weight": { | |
| "scale": 0.5440301895141602, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.19.mlp.c_proj.bias": { | |
| "scale": 0.41071999073028564, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.19.mlp.c_proj.weight": { | |
| "scale": 1.0663466453552246, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.2.attn.c_attn.bias": { | |
| "scale": 0.31375154852867126, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.2.attn.c_attn.weight": { | |
| "scale": 0.43119457364082336, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.2.attn.c_proj.bias": { | |
| "scale": 2.6263744831085205, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.2.attn.c_proj.weight": { | |
| "scale": 3.013683319091797, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.2.ln_1.bias": { | |
| "scale": 1.1809124946594238, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.2.ln_1.weight": { | |
| "scale": 0.6806290745735168, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.2.ln_2.bias": { | |
| "scale": 1.0246816873550415, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.2.ln_2.weight": { | |
| "scale": 1.704537034034729, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.2.mlp.c_fc.bias": { | |
| "scale": 0.3829772174358368, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.2.mlp.c_fc.weight": { | |
| "scale": 1.365037441253662, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.2.mlp.c_proj.bias": { | |
| "scale": 1.9403208494186401, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.2.mlp.c_proj.weight": { | |
| "scale": 5.108565807342529, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.20.attn.c_attn.bias": { | |
| "scale": 0.2516125738620758, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.20.attn.c_attn.weight": { | |
| "scale": 0.870423436164856, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.20.attn.c_proj.bias": { | |
| "scale": 0.6339138150215149, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.20.attn.c_proj.weight": { | |
| "scale": 0.7554279565811157, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.20.ln_1.bias": { | |
| "scale": 0.9357683062553406, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.20.ln_1.weight": { | |
| "scale": 1.0969212055206299, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.20.ln_2.bias": { | |
| "scale": 0.7808690071105957, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.20.ln_2.weight": { | |
| "scale": 1.1708141565322876, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.20.mlp.c_fc.bias": { | |
| "scale": 0.10677170008420944, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.20.mlp.c_fc.weight": { | |
| "scale": 0.45480090379714966, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.20.mlp.c_proj.bias": { | |
| "scale": 0.30032098293304443, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.20.mlp.c_proj.weight": { | |
| "scale": 1.8839131593704224, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.21.attn.c_attn.bias": { | |
| "scale": 0.3100431561470032, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.21.attn.c_attn.weight": { | |
| "scale": 0.7054853439331055, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.21.attn.c_proj.bias": { | |
| "scale": 0.5535825490951538, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.21.attn.c_proj.weight": { | |
| "scale": 1.1769016981124878, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.21.ln_1.bias": { | |
| "scale": 0.8685834407806396, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.21.ln_1.weight": { | |
| "scale": 1.1164261102676392, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.21.ln_2.bias": { | |
| "scale": 0.7616092562675476, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.21.ln_2.weight": { | |
| "scale": 1.1056314706802368, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.21.mlp.c_fc.bias": { | |
| "scale": 0.15077142417430878, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.21.mlp.c_fc.weight": { | |
| "scale": 0.4911968410015106, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.21.mlp.c_proj.bias": { | |
| "scale": 0.1756763458251953, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.21.mlp.c_proj.weight": { | |
| "scale": 1.4334403276443481, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.22.attn.c_attn.bias": { | |
| "scale": 0.2804168462753296, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.22.attn.c_attn.weight": { | |
| "scale": 0.4919489324092865, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.22.attn.c_proj.bias": { | |
| "scale": 0.4737534523010254, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.22.attn.c_proj.weight": { | |
| "scale": 0.6300857067108154, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.22.ln_1.bias": { | |
| "scale": 0.8146415948867798, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.22.ln_1.weight": { | |
| "scale": 1.0955730676651, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.22.ln_2.bias": { | |
| "scale": 0.8315455317497253, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.22.ln_2.weight": { | |
| "scale": 1.116787314414978, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.22.mlp.c_fc.bias": { | |
| "scale": 0.1203167513012886, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.22.mlp.c_fc.weight": { | |
| "scale": 0.5276402235031128, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.22.mlp.c_proj.bias": { | |
| "scale": 0.17383702099323273, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.22.mlp.c_proj.weight": { | |
| "scale": 1.2244986295700073, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.23.attn.c_attn.bias": { | |
| "scale": 0.2580159902572632, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.23.attn.c_attn.weight": { | |
| "scale": 0.6216667890548706, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.23.attn.c_proj.bias": { | |
| "scale": 0.19547155499458313, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.23.attn.c_proj.weight": { | |
| "scale": 0.7300105690956116, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.23.ln_1.bias": { | |
| "scale": 0.7185365557670593, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.23.ln_1.weight": { | |
| "scale": 1.0195420980453491, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.23.ln_2.bias": { | |
| "scale": 0.8253968954086304, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.23.ln_2.weight": { | |
| "scale": 1.0707511901855469, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.23.mlp.c_fc.bias": { | |
| "scale": 0.12819896638393402, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.23.mlp.c_fc.weight": { | |
| "scale": 0.4703911542892456, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.23.mlp.c_proj.bias": { | |
| "scale": 0.21622495353221893, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.23.mlp.c_proj.weight": { | |
| "scale": 0.9520456790924072, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.24.attn.c_attn.bias": { | |
| "scale": 0.29281240701675415, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.24.attn.c_attn.weight": { | |
| "scale": 0.40381887555122375, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.24.attn.c_proj.bias": { | |
| "scale": 0.20989258587360382, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.24.attn.c_proj.weight": { | |
| "scale": 0.6699181795120239, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.24.ln_1.bias": { | |
| "scale": 0.7208132743835449, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.24.ln_1.weight": { | |
| "scale": 1.151794195175171, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.24.ln_2.bias": { | |
| "scale": 0.8072383999824524, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.24.ln_2.weight": { | |
| "scale": 1.1724637746810913, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.24.mlp.c_fc.bias": { | |
| "scale": 0.1703694760799408, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.24.mlp.c_fc.weight": { | |
| "scale": 0.39706242084503174, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.24.mlp.c_proj.bias": { | |
| "scale": 0.21926502883434296, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.24.mlp.c_proj.weight": { | |
| "scale": 0.5858592987060547, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.25.attn.c_attn.bias": { | |
| "scale": 0.30726155638694763, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.25.attn.c_attn.weight": { | |
| "scale": 0.411799818277359, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.25.attn.c_proj.bias": { | |
| "scale": 0.24399296939373016, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.25.attn.c_proj.weight": { | |
| "scale": 0.6184800863265991, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.25.ln_1.bias": { | |
| "scale": 0.650389552116394, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.25.ln_1.weight": { | |
| "scale": 1.253990650177002, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.25.ln_2.bias": { | |
| "scale": 0.8235530853271484, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.25.ln_2.weight": { | |
| "scale": 1.1858876943588257, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.25.mlp.c_fc.bias": { | |
| "scale": 0.15443487465381622, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.25.mlp.c_fc.weight": { | |
| "scale": 0.4894944131374359, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.25.mlp.c_proj.bias": { | |
| "scale": 0.2847861051559448, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.25.mlp.c_proj.weight": { | |
| "scale": 0.6402336359024048, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.26.attn.c_attn.bias": { | |
| "scale": 0.5720527768135071, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.26.attn.c_attn.weight": { | |
| "scale": 0.4341934025287628, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.26.attn.c_proj.bias": { | |
| "scale": 0.26778459548950195, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.26.attn.c_proj.weight": { | |
| "scale": 0.7562843561172485, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.26.ln_1.bias": { | |
| "scale": 0.5681367516517639, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.26.ln_1.weight": { | |
| "scale": 1.196866750717163, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.26.ln_2.bias": { | |
| "scale": 0.8213993906974792, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.26.ln_2.weight": { | |
| "scale": 1.2492419481277466, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.26.mlp.c_fc.bias": { | |
| "scale": 0.22124537825584412, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.26.mlp.c_fc.weight": { | |
| "scale": 0.3691442310810089, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.26.mlp.c_proj.bias": { | |
| "scale": 0.35506486892700195, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.26.mlp.c_proj.weight": { | |
| "scale": 1.0029659271240234, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.27.attn.c_attn.bias": { | |
| "scale": 0.3983057737350464, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.27.attn.c_attn.weight": { | |
| "scale": 0.5026274919509888, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.27.attn.c_proj.bias": { | |
| "scale": 0.37028607726097107, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.27.attn.c_proj.weight": { | |
| "scale": 0.6866351962089539, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.27.ln_1.bias": { | |
| "scale": 0.5788735747337341, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.27.ln_1.weight": { | |
| "scale": 1.2025846242904663, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.27.ln_2.bias": { | |
| "scale": 0.8350062370300293, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.27.ln_2.weight": { | |
| "scale": 1.2651458978652954, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.27.mlp.c_fc.bias": { | |
| "scale": 0.2119043618440628, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.27.mlp.c_fc.weight": { | |
| "scale": 0.37953221797943115, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.27.mlp.c_proj.bias": { | |
| "scale": 0.3048810064792633, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.27.mlp.c_proj.weight": { | |
| "scale": 1.0571913719177246, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.28.attn.c_attn.bias": { | |
| "scale": 0.3161656856536865, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.28.attn.c_attn.weight": { | |
| "scale": 0.39138245582580566, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.28.attn.c_proj.bias": { | |
| "scale": 0.5387474894523621, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.28.attn.c_proj.weight": { | |
| "scale": 1.1945445537567139, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.28.ln_1.bias": { | |
| "scale": 0.6207168102264404, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.28.ln_1.weight": { | |
| "scale": 1.2662208080291748, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.28.ln_2.bias": { | |
| "scale": 0.7384198307991028, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.28.ln_2.weight": { | |
| "scale": 1.3807166814804077, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.28.mlp.c_fc.bias": { | |
| "scale": 0.25197261571884155, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.28.mlp.c_fc.weight": { | |
| "scale": 1.0825285911560059, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.28.mlp.c_proj.bias": { | |
| "scale": 0.7067692875862122, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.28.mlp.c_proj.weight": { | |
| "scale": 4.0657830238342285, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.29.attn.c_attn.bias": { | |
| "scale": 0.4292069375514984, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.29.attn.c_attn.weight": { | |
| "scale": 0.5109118223190308, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.29.attn.c_proj.bias": { | |
| "scale": 0.23389524221420288, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.29.attn.c_proj.weight": { | |
| "scale": 2.3270254135131836, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.29.ln_1.bias": { | |
| "scale": 0.5844658017158508, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.29.ln_1.weight": { | |
| "scale": 1.2286149263381958, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.29.ln_2.bias": { | |
| "scale": 0.4225330948829651, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.29.ln_2.weight": { | |
| "scale": 1.4540377855300903, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.29.mlp.c_fc.bias": { | |
| "scale": 0.28030163049697876, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.29.mlp.c_fc.weight": { | |
| "scale": 1.8019921779632568, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.29.mlp.c_proj.bias": { | |
| "scale": 0.5616198778152466, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.29.mlp.c_proj.weight": { | |
| "scale": 11.012739181518555, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.3.attn.c_attn.bias": { | |
| "scale": 0.25713393092155457, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.3.attn.c_attn.weight": { | |
| "scale": 0.6324517726898193, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.3.attn.c_proj.bias": { | |
| "scale": 2.667752265930176, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.3.attn.c_proj.weight": { | |
| "scale": 3.59769344329834, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.3.ln_1.bias": { | |
| "scale": 1.02127206325531, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.3.ln_1.weight": { | |
| "scale": 0.8167343139648438, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.3.ln_2.bias": { | |
| "scale": 0.8380739092826843, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.3.ln_2.weight": { | |
| "scale": 1.6680330038070679, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.3.mlp.c_fc.bias": { | |
| "scale": 0.3283645212650299, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.3.mlp.c_fc.weight": { | |
| "scale": 1.1588833332061768, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.3.mlp.c_proj.bias": { | |
| "scale": 1.6539766788482666, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.3.mlp.c_proj.weight": { | |
| "scale": 6.646170139312744, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.4.attn.c_attn.bias": { | |
| "scale": 0.2780280113220215, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.4.attn.c_attn.weight": { | |
| "scale": 0.4671952426433563, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.4.attn.c_proj.bias": { | |
| "scale": 2.2085044384002686, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.4.attn.c_proj.weight": { | |
| "scale": 2.1969192028045654, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.4.ln_1.bias": { | |
| "scale": 0.9655510187149048, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.4.ln_1.weight": { | |
| "scale": 0.9420844316482544, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.4.ln_2.bias": { | |
| "scale": 0.600337028503418, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.4.ln_2.weight": { | |
| "scale": 1.6738598346710205, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.4.mlp.c_fc.bias": { | |
| "scale": 0.11800195276737213, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.4.mlp.c_fc.weight": { | |
| "scale": 0.6646812558174133, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.4.mlp.c_proj.bias": { | |
| "scale": 0.99233478307724, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.4.mlp.c_proj.weight": { | |
| "scale": 7.154963493347168, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.5.attn.c_attn.bias": { | |
| "scale": 0.2130964696407318, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.5.attn.c_attn.weight": { | |
| "scale": 0.3897348940372467, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.5.attn.c_proj.bias": { | |
| "scale": 1.2391482591629028, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.5.attn.c_proj.weight": { | |
| "scale": 0.9562830328941345, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.5.ln_1.bias": { | |
| "scale": 0.7676820755004883, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.5.ln_1.weight": { | |
| "scale": 1.2011964321136475, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.5.ln_2.bias": { | |
| "scale": 0.4611365795135498, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.5.ln_2.weight": { | |
| "scale": 1.7594345808029175, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.5.mlp.c_fc.bias": { | |
| "scale": 0.1486712545156479, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.5.mlp.c_fc.weight": { | |
| "scale": 0.575945258140564, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.5.mlp.c_proj.bias": { | |
| "scale": 0.638988196849823, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.5.mlp.c_proj.weight": { | |
| "scale": 6.2721123695373535, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.6.attn.c_attn.bias": { | |
| "scale": 0.28897368907928467, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.6.attn.c_attn.weight": { | |
| "scale": 0.6516239643096924, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.6.attn.c_proj.bias": { | |
| "scale": 0.6968675255775452, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.6.attn.c_proj.weight": { | |
| "scale": 0.6980696320533752, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.6.ln_1.bias": { | |
| "scale": 0.741898238658905, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.6.ln_1.weight": { | |
| "scale": 1.4237617254257202, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.6.ln_2.bias": { | |
| "scale": 0.5870022177696228, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.6.ln_2.weight": { | |
| "scale": 1.7807389497756958, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.6.mlp.c_fc.bias": { | |
| "scale": 0.18498767912387848, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.6.mlp.c_fc.weight": { | |
| "scale": 0.8209737539291382, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.6.mlp.c_proj.bias": { | |
| "scale": 0.5902945399284363, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.6.mlp.c_proj.weight": { | |
| "scale": 6.16025972366333, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.7.attn.c_attn.bias": { | |
| "scale": 0.3166765570640564, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.7.attn.c_attn.weight": { | |
| "scale": 0.6890222430229187, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.7.attn.c_proj.bias": { | |
| "scale": 0.63676917552948, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.7.attn.c_proj.weight": { | |
| "scale": 0.659460186958313, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.7.ln_1.bias": { | |
| "scale": 1.0708993673324585, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.7.ln_1.weight": { | |
| "scale": 1.055467963218689, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.7.ln_2.bias": { | |
| "scale": 0.9893343448638916, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.7.ln_2.weight": { | |
| "scale": 1.8035027980804443, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.7.mlp.c_fc.bias": { | |
| "scale": 0.2329855114221573, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.7.mlp.c_fc.weight": { | |
| "scale": 0.4914255440235138, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.7.mlp.c_proj.bias": { | |
| "scale": 0.7499263882637024, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.7.mlp.c_proj.weight": { | |
| "scale": 2.9818549156188965, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.8.attn.c_attn.bias": { | |
| "scale": 0.2652284502983093, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.8.attn.c_attn.weight": { | |
| "scale": 0.6239627599716187, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.8.attn.c_proj.bias": { | |
| "scale": 0.6191520094871521, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.8.attn.c_proj.weight": { | |
| "scale": 0.6664621829986572, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.8.ln_1.bias": { | |
| "scale": 0.8762368559837341, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.8.ln_1.weight": { | |
| "scale": 1.153762936592102, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.8.ln_2.bias": { | |
| "scale": 0.9868759512901306, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.8.ln_2.weight": { | |
| "scale": 1.7482178211212158, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.8.mlp.c_fc.bias": { | |
| "scale": 0.18552374839782715, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.8.mlp.c_fc.weight": { | |
| "scale": 0.47477778792381287, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.8.mlp.c_proj.bias": { | |
| "scale": 0.6799322962760925, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.8.mlp.c_proj.weight": { | |
| "scale": 3.0612871646881104, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.9.attn.c_attn.bias": { | |
| "scale": 0.309771865606308, | |
| "shape": [ | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.9.attn.c_attn.weight": { | |
| "scale": 0.8271536231040955, | |
| "shape": [ | |
| 1024, | |
| 3072 | |
| ] | |
| }, | |
| "gpt.h.9.attn.c_proj.bias": { | |
| "scale": 0.49563321471214294, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.9.attn.c_proj.weight": { | |
| "scale": 0.7561616897583008, | |
| "shape": [ | |
| 1024, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.9.ln_1.bias": { | |
| "scale": 1.0751368999481201, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.9.ln_1.weight": { | |
| "scale": 1.0969300270080566, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.9.ln_2.bias": { | |
| "scale": 0.9641826152801514, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.9.ln_2.weight": { | |
| "scale": 1.803221344947815, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.9.mlp.c_fc.bias": { | |
| "scale": 0.22008062899112701, | |
| "shape": [ | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.9.mlp.c_fc.weight": { | |
| "scale": 0.5351904630661011, | |
| "shape": [ | |
| 1024, | |
| 4096 | |
| ] | |
| }, | |
| "gpt.h.9.mlp.c_proj.bias": { | |
| "scale": 0.7559229731559753, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.h.9.mlp.c_proj.weight": { | |
| "scale": 3.0584769248962402, | |
| "shape": [ | |
| 4096, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.ln_f.bias": { | |
| "scale": 1.4052708148956299, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.ln_f.weight": { | |
| "scale": 2.9010915756225586, | |
| "shape": [ | |
| 1024 | |
| ] | |
| }, | |
| "gpt.wpe.emb.weight": { | |
| "scale": 0.5411291122436523, | |
| "shape": [ | |
| 608, | |
| 1024 | |
| ] | |
| }, | |
| "gpt.wte.weight": { | |
| "scale": 0.5614672303199768, | |
| "shape": [ | |
| 1026, | |
| 1024 | |
| ] | |
| }, | |
| "mel_head.bias": { | |
| "scale": 0.19914697110652924, | |
| "shape": [ | |
| 1026 | |
| ] | |
| }, | |
| "mel_head.weight": { | |
| "scale": 0.5308834314346313, | |
| "shape": [ | |
| 1026, | |
| 1024 | |
| ] | |
| } | |
| } |