mrjackspade's picture
Publish step 4020 merge manifest
eb9cb78 verified
Raw
History Blame Contribute Delete
115 kB
{
"schema_version": 1,
"format": "comfy_scaled_fp8_merged_text_encoder",
"scale_policy": "preserve_stock_per_tensor_scale",
"base": "/workspace/Ideogram/checkpoints/text_encoders/qwen3vl_8b_fp8_scaled.safetensors",
"base_sha256": "4ba424cf62e51392e4d1a39933e803706f4e823c1065f36aaf149c6453f66bcd",
"adapter": "/workspace/Ideogram/artifacts/phase2_balanced_4000_8state_debanner_s020_v1_checkpoints/step_00004020/adapters.safetensors",
"adapter_sha256": "fe1e1978fafed74732035f4a372e9deca4fbee6d60b1339bb89c093dcba35ea1",
"adapter_global_step": "4020",
"output": "/workspace/Ideogram/artifacts/huggingface_release/s020_v1/qwen3vl_8b_ideogram4_nl_s020_v1_step_00004020_fp8_scaled.safetensors",
"output_bytes": 10588637800,
"output_sha256": "7091beb3199ef429331551b0fd9160609adbb8406a1b0a1cd78cd4434ce0bc21",
"output_tensors": 1254,
"merged_projections": 252,
"lora_rank": 64,
"lora_alpha": 64.0,
"elapsed_seconds": 72.814,
"projections": [
{
"index": 1,
"total": 252,
"projection": "layers.0.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.426,
"parameters": 50331648,
"reference_delta_l2": 3.1311253190985737,
"retained_delta_l2": 2.851645034670928,
"delta_retention_norm_ratio": 0.9107412652176738,
"delta_cosine": 0.7260173634785715,
"relative_weight_rmse": 0.01232165792227417
},
{
"index": 2,
"total": 252,
"projection": "layers.0.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3897,
"parameters": 50331648,
"reference_delta_l2": 3.344906082655277,
"retained_delta_l2": 3.064115813725918,
"delta_retention_norm_ratio": 0.9160543638623002,
"delta_cosine": 0.7288135473708925,
"relative_weight_rmse": 0.01358371183898543
},
{
"index": 3,
"total": 252,
"projection": "layers.0.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3999,
"parameters": 50331648,
"reference_delta_l2": 2.9641851877393086,
"retained_delta_l2": 2.658110792844558,
"delta_retention_norm_ratio": 0.8967424855367474,
"delta_cosine": 0.7099252460764254,
"relative_weight_rmse": 0.013104123685828459
},
{
"index": 4,
"total": 252,
"projection": "layers.0.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 0.6043952512305132,
"retained_delta_l2": 0.44348190222298217,
"delta_retention_norm_ratio": 0.7337613942533121,
"delta_cosine": 0.5614919159555548,
"relative_weight_rmse": 0.009031046537869064
},
{
"index": 5,
"total": 252,
"projection": "layers.0.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1357,
"parameters": 16777216,
"reference_delta_l2": 1.9750892235880966,
"retained_delta_l2": 1.863235717744188,
"delta_retention_norm_ratio": 0.9433678719381056,
"delta_cosine": 0.7629176793847307,
"relative_weight_rmse": 0.013633266958462929
},
{
"index": 6,
"total": 252,
"projection": "layers.0.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1291,
"parameters": 16777216,
"reference_delta_l2": 1.2288275649963785,
"retained_delta_l2": 0.9573387889537007,
"delta_retention_norm_ratio": 0.7790668245275912,
"delta_cosine": 0.5996510517722599,
"relative_weight_rmse": 0.009826446422821086
},
{
"index": 7,
"total": 252,
"projection": "layers.0.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0303,
"parameters": 4194304,
"reference_delta_l2": 0.796856169873161,
"retained_delta_l2": 0.6945727287203802,
"delta_retention_norm_ratio": 0.8716412760296984,
"delta_cosine": 0.6845171683899881,
"relative_weight_rmse": 0.011997953928031789
},
{
"index": 8,
"total": 252,
"projection": "layers.1.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4285,
"parameters": 50331648,
"reference_delta_l2": 2.0378336344324155,
"retained_delta_l2": 1.902449821386416,
"delta_retention_norm_ratio": 0.9335648353435352,
"delta_cosine": 0.7566387616836463,
"relative_weight_rmse": 0.011990448895243113
},
{
"index": 9,
"total": 252,
"projection": "layers.1.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3896,
"parameters": 50331648,
"reference_delta_l2": 3.8315459685202464,
"retained_delta_l2": 3.7988249423296567,
"delta_retention_norm_ratio": 0.9914600982320391,
"delta_cosine": 0.8458108571812768,
"relative_weight_rmse": 0.01436441994921235
},
{
"index": 10,
"total": 252,
"projection": "layers.1.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4008,
"parameters": 50331648,
"reference_delta_l2": 4.243881420701621,
"retained_delta_l2": 4.240972107520675,
"delta_retention_norm_ratio": 0.999314468786344,
"delta_cosine": 0.9298971059666952,
"relative_weight_rmse": 0.01576868165047045
},
{
"index": 11,
"total": 252,
"projection": "layers.1.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0393,
"parameters": 4194304,
"reference_delta_l2": 0.6778121532949969,
"retained_delta_l2": 0.5569451881081336,
"delta_retention_norm_ratio": 0.8216807347001646,
"delta_cosine": 0.6378824613543993,
"relative_weight_rmse": 0.010835416625909753
},
{
"index": 12,
"total": 252,
"projection": "layers.1.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1418,
"parameters": 16777216,
"reference_delta_l2": 1.7174485232552072,
"retained_delta_l2": 1.5373997116649851,
"delta_retention_norm_ratio": 0.8951649443041458,
"delta_cosine": 0.7058879282980292,
"relative_weight_rmse": 0.012679142249911454
},
{
"index": 13,
"total": 252,
"projection": "layers.1.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.14,
"parameters": 16777216,
"reference_delta_l2": 2.0938452275480617,
"retained_delta_l2": 2.0246605077015576,
"delta_retention_norm_ratio": 0.966958054522721,
"delta_cosine": 0.7927801962088451,
"relative_weight_rmse": 0.014642734296281653
},
{
"index": 14,
"total": 252,
"projection": "layers.1.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0319,
"parameters": 4194304,
"reference_delta_l2": 0.9126597518885261,
"retained_delta_l2": 0.8297511349991242,
"delta_retention_norm_ratio": 0.9091571456746692,
"delta_cosine": 0.7217150860410936,
"relative_weight_rmse": 0.012416623644733855
},
{
"index": 15,
"total": 252,
"projection": "layers.10.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4481,
"parameters": 50331648,
"reference_delta_l2": 2.8154098102976044,
"retained_delta_l2": 2.496899219243625,
"delta_retention_norm_ratio": 0.8868688352619218,
"delta_cosine": 0.7008657006215216,
"relative_weight_rmse": 0.011674455149059019
},
{
"index": 16,
"total": 252,
"projection": "layers.10.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.41,
"parameters": 50331648,
"reference_delta_l2": 2.395607624683916,
"retained_delta_l2": 1.8583388084781514,
"delta_retention_norm_ratio": 0.7757275395729074,
"delta_cosine": 0.5964965099194016,
"relative_weight_rmse": 0.009669806607496357
},
{
"index": 17,
"total": 252,
"projection": "layers.10.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4171,
"parameters": 50331648,
"reference_delta_l2": 2.8577135905386326,
"retained_delta_l2": 2.52926751345058,
"delta_retention_norm_ratio": 0.8850668316882849,
"delta_cosine": 0.6975637132825229,
"relative_weight_rmse": 0.011913133651727727
},
{
"index": 18,
"total": 252,
"projection": "layers.10.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.032,
"parameters": 4194304,
"reference_delta_l2": 0.8998186927776939,
"retained_delta_l2": 0.8275767991465501,
"delta_retention_norm_ratio": 0.9197150556984576,
"delta_cosine": 0.7376539813663522,
"relative_weight_rmse": 0.011985557206987124
},
{
"index": 19,
"total": 252,
"projection": "layers.10.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1402,
"parameters": 16777216,
"reference_delta_l2": 1.7764216169052178,
"retained_delta_l2": 1.5880263761424567,
"delta_retention_norm_ratio": 0.8939467753770229,
"delta_cosine": 0.7071347181534064,
"relative_weight_rmse": 0.012178540141018757
},
{
"index": 20,
"total": 252,
"projection": "layers.10.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1408,
"parameters": 16777216,
"reference_delta_l2": 2.122997374391854,
"retained_delta_l2": 2.025968461198601,
"delta_retention_norm_ratio": 0.9542962632155645,
"delta_cosine": 0.780362016551102,
"relative_weight_rmse": 0.013467013031407668
},
{
"index": 21,
"total": 252,
"projection": "layers.10.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0343,
"parameters": 4194304,
"reference_delta_l2": 0.912184359540502,
"retained_delta_l2": 0.8151538420844168,
"delta_retention_norm_ratio": 0.8936283916280227,
"delta_cosine": 0.7118748333251881,
"relative_weight_rmse": 0.011223332766388696
},
{
"index": 22,
"total": 252,
"projection": "layers.11.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4686,
"parameters": 50331648,
"reference_delta_l2": 2.6104099723148297,
"retained_delta_l2": 2.2346067332217188,
"delta_retention_norm_ratio": 0.8560366980364159,
"delta_cosine": 0.6706123400826592,
"relative_weight_rmse": 0.01105385162026209
},
{
"index": 23,
"total": 252,
"projection": "layers.11.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4194,
"parameters": 50331648,
"reference_delta_l2": 2.7815164311376197,
"retained_delta_l2": 2.3500027721817416,
"delta_retention_norm_ratio": 0.844863882835525,
"delta_cosine": 0.6602484777285531,
"relative_weight_rmse": 0.01083411473290713
},
{
"index": 24,
"total": 252,
"projection": "layers.11.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4375,
"parameters": 50331648,
"reference_delta_l2": 2.678569199184031,
"retained_delta_l2": 2.2772185933427775,
"delta_retention_norm_ratio": 0.8501623157753341,
"delta_cosine": 0.6637258455511463,
"relative_weight_rmse": 0.011385599843358225
},
{
"index": 25,
"total": 252,
"projection": "layers.11.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0465,
"parameters": 4194304,
"reference_delta_l2": 1.1340656916927645,
"retained_delta_l2": 1.0920343839344664,
"delta_retention_norm_ratio": 0.9629375017107165,
"delta_cosine": 0.7881295692080786,
"relative_weight_rmse": 0.013695872593250422
},
{
"index": 26,
"total": 252,
"projection": "layers.11.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1999,
"parameters": 16777216,
"reference_delta_l2": 1.672143527965203,
"retained_delta_l2": 1.4687315594636061,
"delta_retention_norm_ratio": 0.8783525665711694,
"delta_cosine": 0.6933074227588261,
"relative_weight_rmse": 0.011737206507406017
},
{
"index": 27,
"total": 252,
"projection": "layers.11.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1989,
"parameters": 16777216,
"reference_delta_l2": 1.4122816222708314,
"retained_delta_l2": 1.1762013541164018,
"delta_retention_norm_ratio": 0.83283768305727,
"delta_cosine": 0.6475239186534177,
"relative_weight_rmse": 0.010783765175329814
},
{
"index": 28,
"total": 252,
"projection": "layers.11.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0469,
"parameters": 4194304,
"reference_delta_l2": 1.0013409141793463,
"retained_delta_l2": 0.9117468931001278,
"delta_retention_norm_ratio": 0.9105259559351515,
"delta_cosine": 0.7279696889694092,
"relative_weight_rmse": 0.011913626318420027
},
{
"index": 29,
"total": 252,
"projection": "layers.12.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.577,
"parameters": 50331648,
"reference_delta_l2": 2.5691708305434933,
"retained_delta_l2": 2.189496701155307,
"delta_retention_norm_ratio": 0.8522191966083202,
"delta_cosine": 0.6668215305174465,
"relative_weight_rmse": 0.01083997338002989
},
{
"index": 30,
"total": 252,
"projection": "layers.12.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4148,
"parameters": 50331648,
"reference_delta_l2": 3.2247269569166295,
"retained_delta_l2": 2.8937501222543474,
"delta_retention_norm_ratio": 0.8973628344091648,
"delta_cosine": 0.713492442612776,
"relative_weight_rmse": 0.01201085600684952
},
{
"index": 31,
"total": 252,
"projection": "layers.12.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4063,
"parameters": 50331648,
"reference_delta_l2": 2.690858587917568,
"retained_delta_l2": 2.294693909145447,
"delta_retention_norm_ratio": 0.8527738765050789,
"delta_cosine": 0.6665205460872305,
"relative_weight_rmse": 0.01123282761356742
},
{
"index": 32,
"total": 252,
"projection": "layers.12.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0323,
"parameters": 4194304,
"reference_delta_l2": 0.9326008986662857,
"retained_delta_l2": 0.8657912033264226,
"delta_retention_norm_ratio": 0.9283619655145006,
"delta_cosine": 0.7477186684337965,
"relative_weight_rmse": 0.012293476345715487
},
{
"index": 33,
"total": 252,
"projection": "layers.12.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1434,
"parameters": 16777216,
"reference_delta_l2": 1.775347398911911,
"retained_delta_l2": 1.608620445441323,
"delta_retention_norm_ratio": 0.9060877022870154,
"delta_cosine": 0.7221078916643545,
"relative_weight_rmse": 0.012040461606114156
},
{
"index": 34,
"total": 252,
"projection": "layers.12.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1499,
"parameters": 16777216,
"reference_delta_l2": 1.8080238692898043,
"retained_delta_l2": 1.6531886810346097,
"delta_retention_norm_ratio": 0.9143621990366675,
"delta_cosine": 0.731378765364209,
"relative_weight_rmse": 0.012442599820811065
},
{
"index": 35,
"total": 252,
"projection": "layers.12.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0373,
"parameters": 4194304,
"reference_delta_l2": 1.0883769590790944,
"retained_delta_l2": 1.0253875091173987,
"delta_retention_norm_ratio": 0.9421253367812998,
"delta_cosine": 0.7663484953141674,
"relative_weight_rmse": 0.012198017874787075
},
{
"index": 36,
"total": 252,
"projection": "layers.13.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4742,
"parameters": 50331648,
"reference_delta_l2": 2.3478968953192507,
"retained_delta_l2": 1.9235928393847987,
"delta_retention_norm_ratio": 0.8192833523565957,
"delta_cosine": 0.6361034701796185,
"relative_weight_rmse": 0.010201271050695015
},
{
"index": 37,
"total": 252,
"projection": "layers.13.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4218,
"parameters": 50331648,
"reference_delta_l2": 3.000471438626002,
"retained_delta_l2": 2.627753262768171,
"delta_retention_norm_ratio": 0.875780128729201,
"delta_cosine": 0.6885754794884557,
"relative_weight_rmse": 0.011915921757060728
},
{
"index": 38,
"total": 252,
"projection": "layers.13.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4051,
"parameters": 50331648,
"reference_delta_l2": 2.4051219611132675,
"retained_delta_l2": 1.945671262513034,
"delta_retention_norm_ratio": 0.8089698959018419,
"delta_cosine": 0.6255035112948868,
"relative_weight_rmse": 0.010360387584441638
},
{
"index": 39,
"total": 252,
"projection": "layers.13.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0319,
"parameters": 4194304,
"reference_delta_l2": 1.3429069331806525,
"retained_delta_l2": 1.3287285477304247,
"delta_retention_norm_ratio": 0.9894420193239702,
"delta_cosine": 0.8346571726075079,
"relative_weight_rmse": 0.014787635515364923
},
{
"index": 40,
"total": 252,
"projection": "layers.13.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1378,
"parameters": 16777216,
"reference_delta_l2": 1.7328648694224957,
"retained_delta_l2": 1.5537014169525183,
"delta_retention_norm_ratio": 0.8966085263591924,
"delta_cosine": 0.7123589803493854,
"relative_weight_rmse": 0.012274901358271239
},
{
"index": 41,
"total": 252,
"projection": "layers.13.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1369,
"parameters": 16777216,
"reference_delta_l2": 1.4123517420363745,
"retained_delta_l2": 1.1765043274465432,
"delta_retention_norm_ratio": 0.8330108516382904,
"delta_cosine": 0.6475889102758927,
"relative_weight_rmse": 0.01076471145912194
},
{
"index": 42,
"total": 252,
"projection": "layers.13.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 1.0333335552344707,
"retained_delta_l2": 0.9529341362641146,
"delta_retention_norm_ratio": 0.9221941273821184,
"delta_cosine": 0.73967487511522,
"relative_weight_rmse": 0.0123538070895839
},
{
"index": 43,
"total": 252,
"projection": "layers.14.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4555,
"parameters": 50331648,
"reference_delta_l2": 2.469573129616929,
"retained_delta_l2": 2.0888005907238436,
"delta_retention_norm_ratio": 0.8458144307100759,
"delta_cosine": 0.6614629304359669,
"relative_weight_rmse": 0.010524613952879095
},
{
"index": 44,
"total": 252,
"projection": "layers.14.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4081,
"parameters": 50331648,
"reference_delta_l2": 2.742309509862897,
"retained_delta_l2": 2.328518937993237,
"delta_retention_norm_ratio": 0.8491087273768936,
"delta_cosine": 0.6623423703861038,
"relative_weight_rmse": 0.011387610089471041
},
{
"index": 45,
"total": 252,
"projection": "layers.14.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4052,
"parameters": 50331648,
"reference_delta_l2": 2.947539534787834,
"retained_delta_l2": 2.6134287856822414,
"delta_retention_norm_ratio": 0.8866475766780032,
"delta_cosine": 0.7010843497097912,
"relative_weight_rmse": 0.011791903699862993
},
{
"index": 46,
"total": 252,
"projection": "layers.14.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0323,
"parameters": 4194304,
"reference_delta_l2": 0.7297764978434413,
"retained_delta_l2": 0.619903107980693,
"delta_retention_norm_ratio": 0.8494424112211963,
"delta_cosine": 0.6664616509602604,
"relative_weight_rmse": 0.010727987313735468
},
{
"index": 47,
"total": 252,
"projection": "layers.14.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1377,
"parameters": 16777216,
"reference_delta_l2": 1.558674985596514,
"retained_delta_l2": 1.3576325874039534,
"delta_retention_norm_ratio": 0.8710171138625026,
"delta_cosine": 0.6852993618997607,
"relative_weight_rmse": 0.011457881608973835
},
{
"index": 48,
"total": 252,
"projection": "layers.14.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1439,
"parameters": 16777216,
"reference_delta_l2": 2.3179808306560723,
"retained_delta_l2": 2.259398847505989,
"delta_retention_norm_ratio": 0.9747271494331112,
"delta_cosine": 0.80549694770149,
"relative_weight_rmse": 0.01391203550972023
},
{
"index": 49,
"total": 252,
"projection": "layers.14.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0311,
"parameters": 4194304,
"reference_delta_l2": 1.0188578856172599,
"retained_delta_l2": 0.9462917556027628,
"delta_retention_norm_ratio": 0.9287769854472551,
"delta_cosine": 0.746251700282814,
"relative_weight_rmse": 0.01239143536617792
},
{
"index": 50,
"total": 252,
"projection": "layers.15.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4492,
"parameters": 50331648,
"reference_delta_l2": 2.3708678589434378,
"retained_delta_l2": 1.9723246740883256,
"delta_retention_norm_ratio": 0.8318998744060243,
"delta_cosine": 0.6480306353655926,
"relative_weight_rmse": 0.010299677630315413
},
{
"index": 51,
"total": 252,
"projection": "layers.15.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4045,
"parameters": 50331648,
"reference_delta_l2": 2.8883940009536784,
"retained_delta_l2": 2.530292392394288,
"delta_retention_norm_ratio": 0.8760205122842818,
"delta_cosine": 0.6878748286799004,
"relative_weight_rmse": 0.012062419373950086
},
{
"index": 52,
"total": 252,
"projection": "layers.15.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4102,
"parameters": 50331648,
"reference_delta_l2": 2.788342252793093,
"retained_delta_l2": 2.4395352636314054,
"delta_retention_norm_ratio": 0.8749052456482751,
"delta_cosine": 0.6892251782824516,
"relative_weight_rmse": 0.011425039353465094
},
{
"index": 53,
"total": 252,
"projection": "layers.15.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.032,
"parameters": 4194304,
"reference_delta_l2": 0.8145269180918248,
"retained_delta_l2": 0.720379517527383,
"delta_retention_norm_ratio": 0.8844146234172361,
"delta_cosine": 0.6991369929850383,
"relative_weight_rmse": 0.011595209140609975
},
{
"index": 54,
"total": 252,
"projection": "layers.15.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1405,
"parameters": 16777216,
"reference_delta_l2": 2.1497846497272506,
"retained_delta_l2": 2.0659256007681135,
"delta_retention_norm_ratio": 0.9609918840150913,
"delta_cosine": 0.7885974868360888,
"relative_weight_rmse": 0.013836619128579801
},
{
"index": 55,
"total": 252,
"projection": "layers.15.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1406,
"parameters": 16777216,
"reference_delta_l2": 2.2377648457258323,
"retained_delta_l2": 2.162654914482189,
"delta_retention_norm_ratio": 0.9664352885929438,
"delta_cosine": 0.8011023199522198,
"relative_weight_rmse": 0.013454728084133871
},
{
"index": 56,
"total": 252,
"projection": "layers.15.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0322,
"parameters": 4194304,
"reference_delta_l2": 0.8605210346924138,
"retained_delta_l2": 0.7654756680176747,
"delta_retention_norm_ratio": 0.8895490489564707,
"delta_cosine": 0.7059715266274591,
"relative_weight_rmse": 0.011246988194474008
},
{
"index": 57,
"total": 252,
"projection": "layers.16.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4979,
"parameters": 50331648,
"reference_delta_l2": 2.490685078994634,
"retained_delta_l2": 2.1279313342383284,
"delta_retention_norm_ratio": 0.8543558365464929,
"delta_cosine": 0.669094258774527,
"relative_weight_rmse": 0.010663628776454517
},
{
"index": 58,
"total": 252,
"projection": "layers.16.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4092,
"parameters": 50331648,
"reference_delta_l2": 2.6473584438715316,
"retained_delta_l2": 2.2656418589367266,
"delta_retention_norm_ratio": 0.8558122773973222,
"delta_cosine": 0.6691989170714746,
"relative_weight_rmse": 0.011094012154423467
},
{
"index": 59,
"total": 252,
"projection": "layers.16.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4396,
"parameters": 50331648,
"reference_delta_l2": 2.55736669406065,
"retained_delta_l2": 2.1559497890563675,
"delta_retention_norm_ratio": 0.8430350618327234,
"delta_cosine": 0.6571284538780846,
"relative_weight_rmse": 0.010890256567761988
},
{
"index": 60,
"total": 252,
"projection": "layers.16.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0328,
"parameters": 4194304,
"reference_delta_l2": 0.7410805911644441,
"retained_delta_l2": 0.6386088520802584,
"delta_retention_norm_ratio": 0.8617265918094359,
"delta_cosine": 0.6756666042764502,
"relative_weight_rmse": 0.011225430131309012
},
{
"index": 61,
"total": 252,
"projection": "layers.16.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1381,
"parameters": 16777216,
"reference_delta_l2": 1.6683307894897468,
"retained_delta_l2": 1.4925542040303772,
"delta_retention_norm_ratio": 0.8946392486629523,
"delta_cosine": 0.7096960070588256,
"relative_weight_rmse": 0.011946513807621155
},
{
"index": 62,
"total": 252,
"projection": "layers.16.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.138,
"parameters": 16777216,
"reference_delta_l2": 1.8123964496879301,
"retained_delta_l2": 1.6808503386928568,
"delta_retention_norm_ratio": 0.9274186886551651,
"delta_cosine": 0.7482999852666631,
"relative_weight_rmse": 0.012472925655758988
},
{
"index": 63,
"total": 252,
"projection": "layers.16.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0315,
"parameters": 4194304,
"reference_delta_l2": 0.9577929367868838,
"retained_delta_l2": 0.8744181947321424,
"delta_retention_norm_ratio": 0.9129511830245487,
"delta_cosine": 0.7297937432743306,
"relative_weight_rmse": 0.012143767704550699
},
{
"index": 64,
"total": 252,
"projection": "layers.17.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.6283,
"parameters": 50331648,
"reference_delta_l2": 2.5799456264058795,
"retained_delta_l2": 2.2192014783790848,
"delta_retention_norm_ratio": 0.8601737399677887,
"delta_cosine": 0.6746896043620583,
"relative_weight_rmse": 0.01092665647466412
},
{
"index": 65,
"total": 252,
"projection": "layers.17.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.5892,
"parameters": 50331648,
"reference_delta_l2": 3.553925118055408,
"retained_delta_l2": 3.335465464267366,
"delta_retention_norm_ratio": 0.9385300346711931,
"delta_cosine": 0.7574282628963688,
"relative_weight_rmse": 0.013331781964907223
},
{
"index": 66,
"total": 252,
"projection": "layers.17.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4603,
"parameters": 50331648,
"reference_delta_l2": 3.185910889753096,
"retained_delta_l2": 2.9020136708320443,
"delta_retention_norm_ratio": 0.9108897804285251,
"delta_cosine": 0.7254698345383945,
"relative_weight_rmse": 0.012382911774040701
},
{
"index": 67,
"total": 252,
"projection": "layers.17.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0356,
"parameters": 4194304,
"reference_delta_l2": 0.7875389681256384,
"retained_delta_l2": 0.7048170134748019,
"delta_retention_norm_ratio": 0.8949614457203092,
"delta_cosine": 0.7148004125434958,
"relative_weight_rmse": 0.011560088995021325
},
{
"index": 68,
"total": 252,
"projection": "layers.17.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1534,
"parameters": 16777216,
"reference_delta_l2": 1.7301552499646233,
"retained_delta_l2": 1.57031997815632,
"delta_retention_norm_ratio": 0.9076179598266851,
"delta_cosine": 0.7248082791642272,
"relative_weight_rmse": 0.012201341600099743
},
{
"index": 69,
"total": 252,
"projection": "layers.17.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1562,
"parameters": 16777216,
"reference_delta_l2": 1.6622515265341566,
"retained_delta_l2": 1.4875158863932494,
"delta_retention_norm_ratio": 0.8948801445800225,
"delta_cosine": 0.7079099484224681,
"relative_weight_rmse": 0.011976361528570949
},
{
"index": 70,
"total": 252,
"projection": "layers.17.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0349,
"parameters": 4194304,
"reference_delta_l2": 1.0327487637903503,
"retained_delta_l2": 0.9684890309610055,
"delta_retention_norm_ratio": 0.9377779619958088,
"delta_cosine": 0.7589254804851308,
"relative_weight_rmse": 0.01275393762219772
},
{
"index": 71,
"total": 252,
"projection": "layers.18.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4716,
"parameters": 50331648,
"reference_delta_l2": 2.910942824426125,
"retained_delta_l2": 2.6223137638455887,
"delta_retention_norm_ratio": 0.9008468808941866,
"delta_cosine": 0.715347308907642,
"relative_weight_rmse": 0.011717583497849349
},
{
"index": 72,
"total": 252,
"projection": "layers.18.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4071,
"parameters": 50331648,
"reference_delta_l2": 4.7566559139668145,
"retained_delta_l2": 4.721819911363684,
"delta_retention_norm_ratio": 0.9926763669197003,
"delta_cosine": 0.828302544944248,
"relative_weight_rmse": 0.015482149915699557
},
{
"index": 73,
"total": 252,
"projection": "layers.18.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4384,
"parameters": 50331648,
"reference_delta_l2": 3.269018312458689,
"retained_delta_l2": 3.008014333209573,
"delta_retention_norm_ratio": 0.9201583000454929,
"delta_cosine": 0.7367324745005237,
"relative_weight_rmse": 0.012492455714722801
},
{
"index": 74,
"total": 252,
"projection": "layers.18.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0359,
"parameters": 4194304,
"reference_delta_l2": 1.1195674639366135,
"retained_delta_l2": 1.0864407095655124,
"delta_retention_norm_ratio": 0.9704111137218823,
"delta_cosine": 0.7993208457886833,
"relative_weight_rmse": 0.014013148599041037
},
{
"index": 75,
"total": 252,
"projection": "layers.18.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1538,
"parameters": 16777216,
"reference_delta_l2": 1.6276674612074848,
"retained_delta_l2": 1.4546150543877026,
"delta_retention_norm_ratio": 0.8936807358111076,
"delta_cosine": 0.7106156353594916,
"relative_weight_rmse": 0.011692682871370855
},
{
"index": 76,
"total": 252,
"projection": "layers.18.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1513,
"parameters": 16777216,
"reference_delta_l2": 2.10005497505043,
"retained_delta_l2": 2.0097741177809665,
"delta_retention_norm_ratio": 0.9570102409974789,
"delta_cosine": 0.7836979023090542,
"relative_weight_rmse": 0.013546298349590657
},
{
"index": 77,
"total": 252,
"projection": "layers.18.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0349,
"parameters": 4194304,
"reference_delta_l2": 0.9280940235763724,
"retained_delta_l2": 0.8454209077890051,
"delta_retention_norm_ratio": 0.9109216160354209,
"delta_cosine": 0.7278410211198819,
"relative_weight_rmse": 0.01178831388535288
},
{
"index": 78,
"total": 252,
"projection": "layers.19.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4856,
"parameters": 50331648,
"reference_delta_l2": 2.6909513491215695,
"retained_delta_l2": 2.35436383498776,
"delta_retention_norm_ratio": 0.8749187664638811,
"delta_cosine": 0.6891278949986944,
"relative_weight_rmse": 0.011201087239634143
},
{
"index": 79,
"total": 252,
"projection": "layers.19.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4043,
"parameters": 50331648,
"reference_delta_l2": 4.63653378419394,
"retained_delta_l2": 4.572165535335737,
"delta_retention_norm_ratio": 0.9861171616871127,
"delta_cosine": 0.8194393635845368,
"relative_weight_rmse": 0.015483533554919179
},
{
"index": 80,
"total": 252,
"projection": "layers.19.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4053,
"parameters": 50331648,
"reference_delta_l2": 3.0936210245432267,
"retained_delta_l2": 2.7739735520783784,
"delta_retention_norm_ratio": 0.8966752973525436,
"delta_cosine": 0.7109372653041035,
"relative_weight_rmse": 0.012201010045957531
},
{
"index": 81,
"total": 252,
"projection": "layers.19.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0357,
"parameters": 4194304,
"reference_delta_l2": 1.0461313388783178,
"retained_delta_l2": 1.007070701509642,
"delta_retention_norm_ratio": 0.9626618227396215,
"delta_cosine": 0.7954978221092615,
"relative_weight_rmse": 0.013601824359116688
},
{
"index": 82,
"total": 252,
"projection": "layers.19.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1535,
"parameters": 16777216,
"reference_delta_l2": 1.6404153005963256,
"retained_delta_l2": 1.4610182459530188,
"delta_retention_norm_ratio": 0.8906392456970548,
"delta_cosine": 0.7053464552792614,
"relative_weight_rmse": 0.011796560151505405
},
{
"index": 83,
"total": 252,
"projection": "layers.19.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1515,
"parameters": 16777216,
"reference_delta_l2": 1.9394315274438587,
"retained_delta_l2": 1.8302753962433107,
"delta_retention_norm_ratio": 0.943717460680649,
"delta_cosine": 0.7660351373048885,
"relative_weight_rmse": 0.012835505145347845
},
{
"index": 84,
"total": 252,
"projection": "layers.19.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0347,
"parameters": 4194304,
"reference_delta_l2": 1.2337371802814954,
"retained_delta_l2": 1.2002451732587,
"delta_retention_norm_ratio": 0.972853207670086,
"delta_cosine": 0.8055223257719838,
"relative_weight_rmse": 0.013856626774007605
},
{
"index": 85,
"total": 252,
"projection": "layers.2.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4935,
"parameters": 50331648,
"reference_delta_l2": 1.5518511196740064,
"retained_delta_l2": 1.3366759420320173,
"delta_retention_norm_ratio": 0.8613428988682946,
"delta_cosine": 0.6980578700863995,
"relative_weight_rmse": 0.010111972469663685
},
{
"index": 86,
"total": 252,
"projection": "layers.2.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4504,
"parameters": 50331648,
"reference_delta_l2": 2.5094241436849516,
"retained_delta_l2": 2.366347044034418,
"delta_retention_norm_ratio": 0.9429840905888341,
"delta_cosine": 0.7973101390863121,
"relative_weight_rmse": 0.012665371307374164
},
{
"index": 87,
"total": 252,
"projection": "layers.2.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4497,
"parameters": 50331648,
"reference_delta_l2": 3.3647272098404217,
"retained_delta_l2": 3.3447345474638537,
"delta_retention_norm_ratio": 0.9940581624810184,
"delta_cosine": 0.8791971267764601,
"relative_weight_rmse": 0.014946247063864092
},
{
"index": 88,
"total": 252,
"projection": "layers.2.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0357,
"parameters": 4194304,
"reference_delta_l2": 0.5776991345012915,
"retained_delta_l2": 0.43478786114566365,
"delta_retention_norm_ratio": 0.7526198936077714,
"delta_cosine": 0.5763481613386853,
"relative_weight_rmse": 0.00929726448531777
},
{
"index": 89,
"total": 252,
"projection": "layers.2.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1533,
"parameters": 16777216,
"reference_delta_l2": 2.09035577820529,
"retained_delta_l2": 1.9727312208670051,
"delta_retention_norm_ratio": 0.9437298862879345,
"delta_cosine": 0.7610889829487967,
"relative_weight_rmse": 0.013963950827943881
},
{
"index": 90,
"total": 252,
"projection": "layers.2.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1414,
"parameters": 16777216,
"reference_delta_l2": 1.231077115157188,
"retained_delta_l2": 0.9730372890945638,
"delta_retention_norm_ratio": 0.7903950752673385,
"delta_cosine": 0.6085251089247281,
"relative_weight_rmse": 0.010194756420213015
},
{
"index": 91,
"total": 252,
"projection": "layers.2.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 1.1638421245105781,
"retained_delta_l2": 1.1231926668431822,
"delta_retention_norm_ratio": 0.9650730483015556,
"delta_cosine": 0.7938501587218837,
"relative_weight_rmse": 0.013528191351573087
},
{
"index": 92,
"total": 252,
"projection": "layers.20.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4508,
"parameters": 50331648,
"reference_delta_l2": 2.648327819083784,
"retained_delta_l2": 2.2607752431035935,
"delta_retention_norm_ratio": 0.853661403551518,
"delta_cosine": 0.6660629663101232,
"relative_weight_rmse": 0.01125538558597506
},
{
"index": 93,
"total": 252,
"projection": "layers.20.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4129,
"parameters": 50331648,
"reference_delta_l2": 4.569251145168252,
"retained_delta_l2": 4.503283948830417,
"delta_retention_norm_ratio": 0.9855627991891862,
"delta_cosine": 0.8179012740137436,
"relative_weight_rmse": 0.015462660043373235
},
{
"index": 94,
"total": 252,
"projection": "layers.20.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4066,
"parameters": 50331648,
"reference_delta_l2": 3.127333039964154,
"retained_delta_l2": 2.809774051086796,
"delta_retention_norm_ratio": 0.8984569328500434,
"delta_cosine": 0.7123398664772798,
"relative_weight_rmse": 0.012280477657352691
},
{
"index": 95,
"total": 252,
"projection": "layers.20.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0325,
"parameters": 4194304,
"reference_delta_l2": 0.8444471215421555,
"retained_delta_l2": 0.7766856328214965,
"delta_retention_norm_ratio": 0.9197563861702662,
"delta_cosine": 0.7402785963194388,
"relative_weight_rmse": 0.012201081714308734
},
{
"index": 96,
"total": 252,
"projection": "layers.20.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1385,
"parameters": 16777216,
"reference_delta_l2": 1.6102583747320622,
"retained_delta_l2": 1.4294648241491053,
"delta_retention_norm_ratio": 0.8877238874084166,
"delta_cosine": 0.7036153111364795,
"relative_weight_rmse": 0.011827076546477997
},
{
"index": 97,
"total": 252,
"projection": "layers.20.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.137,
"parameters": 16777216,
"reference_delta_l2": 1.8950289561585765,
"retained_delta_l2": 1.771394157481164,
"delta_retention_norm_ratio": 0.9347583590870119,
"delta_cosine": 0.7573024527449137,
"relative_weight_rmse": 0.012347075017562319
},
{
"index": 98,
"total": 252,
"projection": "layers.20.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0307,
"parameters": 4194304,
"reference_delta_l2": 0.9456752961407162,
"retained_delta_l2": 0.8681683904444714,
"delta_retention_norm_ratio": 0.9180406784310121,
"delta_cosine": 0.7373077284451539,
"relative_weight_rmse": 0.012034663979451769
},
{
"index": 99,
"total": 252,
"projection": "layers.21.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4561,
"parameters": 50331648,
"reference_delta_l2": 3.1059472178434917,
"retained_delta_l2": 2.8040450977659654,
"delta_retention_norm_ratio": 0.9027986958879675,
"delta_cosine": 0.7158659485307688,
"relative_weight_rmse": 0.012306458853181658
},
{
"index": 100,
"total": 252,
"projection": "layers.21.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4063,
"parameters": 50331648,
"reference_delta_l2": 4.647257827016896,
"retained_delta_l2": 4.587235764830932,
"delta_retention_norm_ratio": 0.9870844131270218,
"delta_cosine": 0.8178677648554424,
"relative_weight_rmse": 0.015958610880059444
},
{
"index": 101,
"total": 252,
"projection": "layers.21.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4075,
"parameters": 50331648,
"reference_delta_l2": 3.343274641208289,
"retained_delta_l2": 3.070131448345034,
"delta_retention_norm_ratio": 0.9183007015048759,
"delta_cosine": 0.733268621811139,
"relative_weight_rmse": 0.012763880102308084
},
{
"index": 102,
"total": 252,
"projection": "layers.21.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0326,
"parameters": 4194304,
"reference_delta_l2": 0.8895788270860215,
"retained_delta_l2": 0.8227176449246952,
"delta_retention_norm_ratio": 0.9248395081745119,
"delta_cosine": 0.7421453803773878,
"relative_weight_rmse": 0.012827377562577503
},
{
"index": 103,
"total": 252,
"projection": "layers.21.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1382,
"parameters": 16777216,
"reference_delta_l2": 1.7387482516144692,
"retained_delta_l2": 1.5904826546311277,
"delta_retention_norm_ratio": 0.9147285428780886,
"delta_cosine": 0.7325426554372927,
"relative_weight_rmse": 0.01239246548091918
},
{
"index": 104,
"total": 252,
"projection": "layers.21.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1373,
"parameters": 16777216,
"reference_delta_l2": 1.3482105333013195,
"retained_delta_l2": 1.1153298159841214,
"delta_retention_norm_ratio": 0.8272668017605895,
"delta_cosine": 0.6439534939573677,
"relative_weight_rmse": 0.010341488774834423
},
{
"index": 105,
"total": 252,
"projection": "layers.21.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.031,
"parameters": 4194304,
"reference_delta_l2": 1.3409023100049466,
"retained_delta_l2": 1.3237128538538998,
"delta_retention_norm_ratio": 0.9871806797387175,
"delta_cosine": 0.8233526974986747,
"relative_weight_rmse": 0.014558751804341042
},
{
"index": 106,
"total": 252,
"projection": "layers.22.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4503,
"parameters": 50331648,
"reference_delta_l2": 3.3613694006325487,
"retained_delta_l2": 3.1257459687149756,
"delta_retention_norm_ratio": 0.9299025474935206,
"delta_cosine": 0.7475027723164964,
"relative_weight_rmse": 0.012608671896082564
},
{
"index": 107,
"total": 252,
"projection": "layers.22.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4042,
"parameters": 50331648,
"reference_delta_l2": 4.227090543335068,
"retained_delta_l2": 4.114101999060572,
"delta_retention_norm_ratio": 0.9732703751868653,
"delta_cosine": 0.8026350646972197,
"relative_weight_rmse": 0.014685136811504023
},
{
"index": 108,
"total": 252,
"projection": "layers.22.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4044,
"parameters": 50331648,
"reference_delta_l2": 3.42858856230666,
"retained_delta_l2": 3.1778742711030548,
"delta_retention_norm_ratio": 0.9268753638275772,
"delta_cosine": 0.74505347679932,
"relative_weight_rmse": 0.012627603517297475
},
{
"index": 109,
"total": 252,
"projection": "layers.22.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0321,
"parameters": 4194304,
"reference_delta_l2": 1.1247322733485896,
"retained_delta_l2": 1.0995433623215611,
"delta_retention_norm_ratio": 0.9776045272071412,
"delta_cosine": 0.8173192492799993,
"relative_weight_rmse": 0.014421970360981725
},
{
"index": 110,
"total": 252,
"projection": "layers.22.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1382,
"parameters": 16777216,
"reference_delta_l2": 2.0979872197349207,
"retained_delta_l2": 1.997669141812822,
"delta_retention_norm_ratio": 0.9521836563261936,
"delta_cosine": 0.7727792638598913,
"relative_weight_rmse": 0.013731333527131968
},
{
"index": 111,
"total": 252,
"projection": "layers.22.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1374,
"parameters": 16777216,
"reference_delta_l2": 1.8707772426064833,
"retained_delta_l2": 1.7414875110702557,
"delta_retention_norm_ratio": 0.9308898309260524,
"delta_cosine": 0.7546670902460179,
"relative_weight_rmse": 0.012599364135796138
},
{
"index": 112,
"total": 252,
"projection": "layers.22.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 1.104802659576068,
"retained_delta_l2": 1.05249396885162,
"delta_retention_norm_ratio": 0.9526533627783627,
"delta_cosine": 0.7776846089237898,
"relative_weight_rmse": 0.013232255619878642
},
{
"index": 113,
"total": 252,
"projection": "layers.23.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.6258,
"parameters": 50331648,
"reference_delta_l2": 3.1633513202201087,
"retained_delta_l2": 2.904218525122913,
"delta_retention_norm_ratio": 0.9180828277147652,
"delta_cosine": 0.7353431817072045,
"relative_weight_rmse": 0.011981116038754442
},
{
"index": 114,
"total": 252,
"projection": "layers.23.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.5469,
"parameters": 50331648,
"reference_delta_l2": 5.656505743819756,
"retained_delta_l2": 5.702276957599591,
"delta_retention_norm_ratio": 1.0080917824276665,
"delta_cosine": 0.853967621844258,
"relative_weight_rmse": 0.01684234859200357
},
{
"index": 115,
"total": 252,
"projection": "layers.23.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4509,
"parameters": 50331648,
"reference_delta_l2": 4.986715878731079,
"retained_delta_l2": 4.949932608725162,
"delta_retention_norm_ratio": 0.9926237485951822,
"delta_cosine": 0.8291316907956199,
"relative_weight_rmse": 0.0154201441844509
},
{
"index": 116,
"total": 252,
"projection": "layers.23.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0356,
"parameters": 4194304,
"reference_delta_l2": 0.7939559279294958,
"retained_delta_l2": 0.7065204724238692,
"delta_retention_norm_ratio": 0.889873666245375,
"delta_cosine": 0.7055871174972871,
"relative_weight_rmse": 0.011966085222068048
},
{
"index": 117,
"total": 252,
"projection": "layers.23.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1568,
"parameters": 16777216,
"reference_delta_l2": 1.9604635366469738,
"retained_delta_l2": 1.83341823016883,
"delta_retention_norm_ratio": 0.9351962920486486,
"delta_cosine": 0.7546900769753205,
"relative_weight_rmse": 0.012807196115946568
},
{
"index": 118,
"total": 252,
"projection": "layers.23.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1514,
"parameters": 16777216,
"reference_delta_l2": 1.786934764451321,
"retained_delta_l2": 1.6193395711722547,
"delta_retention_norm_ratio": 0.9062107936936767,
"delta_cosine": 0.7239861199376085,
"relative_weight_rmse": 0.012030546679848587
},
{
"index": 119,
"total": 252,
"projection": "layers.23.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0344,
"parameters": 4194304,
"reference_delta_l2": 0.9284739219207063,
"retained_delta_l2": 0.8527710710862602,
"delta_retention_norm_ratio": 0.9184652912191201,
"delta_cosine": 0.7379057501024063,
"relative_weight_rmse": 0.011738941352338165
},
{
"index": 120,
"total": 252,
"projection": "layers.24.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4538,
"parameters": 50331648,
"reference_delta_l2": 3.2512199635465575,
"retained_delta_l2": 2.9791611152369986,
"delta_retention_norm_ratio": 0.916320996007669,
"delta_cosine": 0.7315979684198591,
"relative_weight_rmse": 0.012274904687060372
},
{
"index": 121,
"total": 252,
"projection": "layers.24.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4141,
"parameters": 50331648,
"reference_delta_l2": 5.154227232527777,
"retained_delta_l2": 5.126463049363616,
"delta_retention_norm_ratio": 0.9946133179792803,
"delta_cosine": 0.8284564309620176,
"relative_weight_rmse": 0.01629264417549896
},
{
"index": 122,
"total": 252,
"projection": "layers.24.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4327,
"parameters": 50331648,
"reference_delta_l2": 4.012577164607281,
"retained_delta_l2": 3.8350333758980173,
"delta_retention_norm_ratio": 0.9557531777144925,
"delta_cosine": 0.7758107049347756,
"relative_weight_rmse": 0.013955012864349846
},
{
"index": 123,
"total": 252,
"projection": "layers.24.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0367,
"parameters": 4194304,
"reference_delta_l2": 0.967778332282462,
"retained_delta_l2": 0.918504467005979,
"delta_retention_norm_ratio": 0.9490855874399742,
"delta_cosine": 0.7728183425993649,
"relative_weight_rmse": 0.013162984015967556
},
{
"index": 124,
"total": 252,
"projection": "layers.24.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1527,
"parameters": 16777216,
"reference_delta_l2": 1.9183039308063368,
"retained_delta_l2": 1.779208986243734,
"delta_retention_norm_ratio": 0.9274906638469244,
"delta_cosine": 0.746153369911424,
"relative_weight_rmse": 0.012581166671527361
},
{
"index": 125,
"total": 252,
"projection": "layers.24.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1554,
"parameters": 16777216,
"reference_delta_l2": 3.030138108840141,
"retained_delta_l2": 3.0274390199969954,
"delta_retention_norm_ratio": 0.9991092522036301,
"delta_cosine": 0.8491407163819161,
"relative_weight_rmse": 0.015583133273445198
},
{
"index": 126,
"total": 252,
"projection": "layers.24.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0398,
"parameters": 4194304,
"reference_delta_l2": 1.0877776464250843,
"retained_delta_l2": 1.0365651068283683,
"delta_retention_norm_ratio": 0.9529200294149978,
"delta_cosine": 0.7790764719100599,
"relative_weight_rmse": 0.013049217585105998
},
{
"index": 127,
"total": 252,
"projection": "layers.25.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4886,
"parameters": 50331648,
"reference_delta_l2": 3.6398335349873516,
"retained_delta_l2": 3.4069360092034424,
"delta_retention_norm_ratio": 0.9360142370399039,
"delta_cosine": 0.7516619597313153,
"relative_weight_rmse": 0.013267928950284245
},
{
"index": 128,
"total": 252,
"projection": "layers.25.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4171,
"parameters": 50331648,
"reference_delta_l2": 5.610837782782727,
"retained_delta_l2": 5.626149005812841,
"delta_retention_norm_ratio": 1.0027288657456996,
"delta_cosine": 0.8409889119700277,
"relative_weight_rmse": 0.016968421605840087
},
{
"index": 129,
"total": 252,
"projection": "layers.25.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4115,
"parameters": 50331648,
"reference_delta_l2": 3.746807110370269,
"retained_delta_l2": 3.497193070805163,
"delta_retention_norm_ratio": 0.9333795329697562,
"delta_cosine": 0.74840306974875,
"relative_weight_rmse": 0.013624024235054728
},
{
"index": 130,
"total": 252,
"projection": "layers.25.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0358,
"parameters": 4194304,
"reference_delta_l2": 1.028118545551861,
"retained_delta_l2": 0.9815052098251581,
"delta_retention_norm_ratio": 0.9546615164872039,
"delta_cosine": 0.7798073063133772,
"relative_weight_rmse": 0.013540882267493162
},
{
"index": 131,
"total": 252,
"projection": "layers.25.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1537,
"parameters": 16777216,
"reference_delta_l2": 1.9557405090667663,
"retained_delta_l2": 1.8088519203064648,
"delta_retention_norm_ratio": 0.9248936205599212,
"delta_cosine": 0.7402053610074313,
"relative_weight_rmse": 0.013197869629674524
},
{
"index": 132,
"total": 252,
"projection": "layers.25.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1527,
"parameters": 16777216,
"reference_delta_l2": 2.450287342227927,
"retained_delta_l2": 2.403087385404849,
"delta_retention_norm_ratio": 0.9807369707178256,
"delta_cosine": 0.8152029253546118,
"relative_weight_rmse": 0.01459653992312539
},
{
"index": 133,
"total": 252,
"projection": "layers.25.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0343,
"parameters": 4194304,
"reference_delta_l2": 1.1892145406651282,
"retained_delta_l2": 1.1459863763592617,
"delta_retention_norm_ratio": 0.96364981857547,
"delta_cosine": 0.7874501262899521,
"relative_weight_rmse": 0.013464933980101157
},
{
"index": 134,
"total": 252,
"projection": "layers.26.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4945,
"parameters": 50331648,
"reference_delta_l2": 3.30003946904166,
"retained_delta_l2": 3.022262544036764,
"delta_retention_norm_ratio": 0.9158261809863859,
"delta_cosine": 0.7287034973659762,
"relative_weight_rmse": 0.012346798780409918
},
{
"index": 135,
"total": 252,
"projection": "layers.26.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4621,
"parameters": 50331648,
"reference_delta_l2": 5.261299580581659,
"retained_delta_l2": 5.224002267062033,
"delta_retention_norm_ratio": 0.9929110074519836,
"delta_cosine": 0.8222294814243173,
"relative_weight_rmse": 0.016656878845297127
},
{
"index": 136,
"total": 252,
"projection": "layers.26.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.5217,
"parameters": 50331648,
"reference_delta_l2": 3.6014725942155867,
"retained_delta_l2": 3.3296092416057674,
"delta_retention_norm_ratio": 0.9245132801936449,
"delta_cosine": 0.7382738382693826,
"relative_weight_rmse": 0.01317464821742268
},
{
"index": 137,
"total": 252,
"projection": "layers.26.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0441,
"parameters": 4194304,
"reference_delta_l2": 1.1031921973196641,
"retained_delta_l2": 1.0648616560180328,
"delta_retention_norm_ratio": 0.9652548836052687,
"delta_cosine": 0.7914792581057428,
"relative_weight_rmse": 0.014447584081242486
},
{
"index": 138,
"total": 252,
"projection": "layers.26.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1742,
"parameters": 16777216,
"reference_delta_l2": 2.146077180004444,
"retained_delta_l2": 2.021769037377566,
"delta_retention_norm_ratio": 0.9420765740463163,
"delta_cosine": 0.7604286350107957,
"relative_weight_rmse": 0.013814252954227363
},
{
"index": 139,
"total": 252,
"projection": "layers.26.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1543,
"parameters": 16777216,
"reference_delta_l2": 2.7172992998817547,
"retained_delta_l2": 2.6935380442306704,
"delta_retention_norm_ratio": 0.9912555618543314,
"delta_cosine": 0.8322062734385259,
"relative_weight_rmse": 0.015376587207712187
},
{
"index": 140,
"total": 252,
"projection": "layers.26.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0479,
"parameters": 4194304,
"reference_delta_l2": 1.3026609614803737,
"retained_delta_l2": 1.2674092692591252,
"delta_retention_norm_ratio": 0.9729387052628126,
"delta_cosine": 0.8072340171450033,
"relative_weight_rmse": 0.014264655200412561
},
{
"index": 141,
"total": 252,
"projection": "layers.27.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4973,
"parameters": 50331648,
"reference_delta_l2": 3.125244636407231,
"retained_delta_l2": 2.799171625202859,
"delta_retention_norm_ratio": 0.895664804154588,
"delta_cosine": 0.708104414605446,
"relative_weight_rmse": 0.011916091086381152
},
{
"index": 142,
"total": 252,
"projection": "layers.27.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4268,
"parameters": 50331648,
"reference_delta_l2": 5.234706562326785,
"retained_delta_l2": 5.202401334124844,
"delta_retention_norm_ratio": 0.9938286458243074,
"delta_cosine": 0.8251230091485036,
"relative_weight_rmse": 0.01646627650655071
},
{
"index": 143,
"total": 252,
"projection": "layers.27.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.451,
"parameters": 50331648,
"reference_delta_l2": 3.1466878793120747,
"retained_delta_l2": 2.7638350222987,
"delta_retention_norm_ratio": 0.8783314800522657,
"delta_cosine": 0.6900926775153142,
"relative_weight_rmse": 0.012191925146307
},
{
"index": 144,
"total": 252,
"projection": "layers.27.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0353,
"parameters": 4194304,
"reference_delta_l2": 0.9588440948085311,
"retained_delta_l2": 0.902925356600317,
"delta_retention_norm_ratio": 0.9416810944438466,
"delta_cosine": 0.7667588223432246,
"relative_weight_rmse": 0.013509612115864522
},
{
"index": 145,
"total": 252,
"projection": "layers.27.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1524,
"parameters": 16777216,
"reference_delta_l2": 1.835792245036977,
"retained_delta_l2": 1.6710422386350765,
"delta_retention_norm_ratio": 0.9102567260280686,
"delta_cosine": 0.7241382091019168,
"relative_weight_rmse": 0.012751238601696262
},
{
"index": 146,
"total": 252,
"projection": "layers.27.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1519,
"parameters": 16777216,
"reference_delta_l2": 2.5048628458225077,
"retained_delta_l2": 2.450570651459129,
"delta_retention_norm_ratio": 0.9783252825782757,
"delta_cosine": 0.8133870694379497,
"relative_weight_rmse": 0.014893972338647133
},
{
"index": 147,
"total": 252,
"projection": "layers.27.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0319,
"parameters": 4194304,
"reference_delta_l2": 1.3030739739734214,
"retained_delta_l2": 1.2793698934282252,
"delta_retention_norm_ratio": 0.9818091059919521,
"delta_cosine": 0.8138158125114515,
"relative_weight_rmse": 0.014301067828013556
},
{
"index": 148,
"total": 252,
"projection": "layers.28.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.6694,
"parameters": 50331648,
"reference_delta_l2": 3.764876726849901,
"retained_delta_l2": 3.5504729834230586,
"delta_retention_norm_ratio": 0.943051590003523,
"delta_cosine": 0.7612047837688289,
"relative_weight_rmse": 0.013106740899625642
},
{
"index": 149,
"total": 252,
"projection": "layers.28.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.6168,
"parameters": 50331648,
"reference_delta_l2": 6.449671079728101,
"retained_delta_l2": 6.565925261839936,
"delta_retention_norm_ratio": 1.0180248233863014,
"delta_cosine": 0.8675629330702189,
"relative_weight_rmse": 0.017974227113179638
},
{
"index": 150,
"total": 252,
"projection": "layers.28.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.6184,
"parameters": 50331648,
"reference_delta_l2": 3.2479959876819815,
"retained_delta_l2": 2.8809992800601387,
"delta_retention_norm_ratio": 0.8870082632448818,
"delta_cosine": 0.6982356355923812,
"relative_weight_rmse": 0.012335816337365748
},
{
"index": 151,
"total": 252,
"projection": "layers.28.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0759,
"parameters": 4194304,
"reference_delta_l2": 1.0321389974726225,
"retained_delta_l2": 0.9822582185272546,
"delta_retention_norm_ratio": 0.9516724209941588,
"delta_cosine": 0.7711618405799451,
"relative_weight_rmse": 0.014133489233940942
},
{
"index": 152,
"total": 252,
"projection": "layers.28.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.2248,
"parameters": 16777216,
"reference_delta_l2": 1.8688173101850416,
"retained_delta_l2": 1.7173842126559218,
"delta_retention_norm_ratio": 0.918968485199805,
"delta_cosine": 0.7339589456875333,
"relative_weight_rmse": 0.012839728307364809
},
{
"index": 153,
"total": 252,
"projection": "layers.28.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.208,
"parameters": 16777216,
"reference_delta_l2": 2.109512276053977,
"retained_delta_l2": 2.0100535882124766,
"delta_retention_norm_ratio": 0.9528522829800513,
"delta_cosine": 0.7773094450468354,
"relative_weight_rmse": 0.013258147079906166
},
{
"index": 154,
"total": 252,
"projection": "layers.28.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0529,
"parameters": 4194304,
"reference_delta_l2": 1.057163102098395,
"retained_delta_l2": 0.9878139310429103,
"delta_retention_norm_ratio": 0.9344006890537218,
"delta_cosine": 0.7519254733182212,
"relative_weight_rmse": 0.012585012211511898
},
{
"index": 155,
"total": 252,
"projection": "layers.29.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.6574,
"parameters": 50331648,
"reference_delta_l2": 3.744975467124991,
"retained_delta_l2": 3.5152066847470396,
"delta_retention_norm_ratio": 0.9386461181401693,
"delta_cosine": 0.7562522943054943,
"relative_weight_rmse": 0.013006729822357985
},
{
"index": 156,
"total": 252,
"projection": "layers.29.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4129,
"parameters": 50331648,
"reference_delta_l2": 6.148255092640315,
"retained_delta_l2": 6.234482157492212,
"delta_retention_norm_ratio": 1.0140246400893669,
"delta_cosine": 0.8574262187978022,
"relative_weight_rmse": 0.017740207715586282
},
{
"index": 157,
"total": 252,
"projection": "layers.29.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4096,
"parameters": 50331648,
"reference_delta_l2": 3.585335358597078,
"retained_delta_l2": 3.285876741858305,
"delta_retention_norm_ratio": 0.9164768182645125,
"delta_cosine": 0.7297086222222687,
"relative_weight_rmse": 0.012909177383043459
},
{
"index": 158,
"total": 252,
"projection": "layers.29.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0354,
"parameters": 4194304,
"reference_delta_l2": 0.9626150691825459,
"retained_delta_l2": 0.9122830564587375,
"delta_retention_norm_ratio": 0.9477132507736967,
"delta_cosine": 0.7818845706182568,
"relative_weight_rmse": 0.01313609049197388
},
{
"index": 159,
"total": 252,
"projection": "layers.29.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1541,
"parameters": 16777216,
"reference_delta_l2": 1.9459053646947755,
"retained_delta_l2": 1.7941218014153966,
"delta_retention_norm_ratio": 0.9219984866513861,
"delta_cosine": 0.7375401582118324,
"relative_weight_rmse": 0.012932987990345222
},
{
"index": 160,
"total": 252,
"projection": "layers.29.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1538,
"parameters": 16777216,
"reference_delta_l2": 2.947751788228365,
"retained_delta_l2": 2.935180959658215,
"delta_retention_norm_ratio": 0.9957354521434435,
"delta_cosine": 0.8480855551489582,
"relative_weight_rmse": 0.015284212821969631
},
{
"index": 161,
"total": 252,
"projection": "layers.29.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0351,
"parameters": 4194304,
"reference_delta_l2": 1.0691175046208725,
"retained_delta_l2": 1.0096838531622976,
"delta_retention_norm_ratio": 0.9444086817382612,
"delta_cosine": 0.7674711374560128,
"relative_weight_rmse": 0.012952705969364688
},
{
"index": 162,
"total": 252,
"projection": "layers.3.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.493,
"parameters": 50331648,
"reference_delta_l2": 2.089716575315149,
"retained_delta_l2": 1.8561741241206091,
"delta_retention_norm_ratio": 0.8882420449006012,
"delta_cosine": 0.7197316256945676,
"relative_weight_rmse": 0.011403469895362428
},
{
"index": 163,
"total": 252,
"projection": "layers.3.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4547,
"parameters": 50331648,
"reference_delta_l2": 2.9318396387227974,
"retained_delta_l2": 2.7113679609099473,
"delta_retention_norm_ratio": 0.9248009083099461,
"delta_cosine": 0.7465708736142087,
"relative_weight_rmse": 0.012270531671797895
},
{
"index": 164,
"total": 252,
"projection": "layers.3.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4518,
"parameters": 50331648,
"reference_delta_l2": 3.188902082085761,
"retained_delta_l2": 3.1234115725582,
"delta_retention_norm_ratio": 0.9794629913864504,
"delta_cosine": 0.843247925504227,
"relative_weight_rmse": 0.014682183398788486
},
{
"index": 165,
"total": 252,
"projection": "layers.3.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0361,
"parameters": 4194304,
"reference_delta_l2": 0.5957548225170014,
"retained_delta_l2": 0.462647700915074,
"delta_retention_norm_ratio": 0.7765739922346515,
"delta_cosine": 0.5976545679020129,
"relative_weight_rmse": 0.009651304860544914
},
{
"index": 166,
"total": 252,
"projection": "layers.3.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1558,
"parameters": 16777216,
"reference_delta_l2": 1.8564977617974963,
"retained_delta_l2": 1.703664558130756,
"delta_retention_norm_ratio": 0.9176766022498382,
"delta_cosine": 0.7312300528278413,
"relative_weight_rmse": 0.012841612285808502
},
{
"index": 167,
"total": 252,
"projection": "layers.3.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1416,
"parameters": 16777216,
"reference_delta_l2": 1.5114591927222754,
"retained_delta_l2": 1.3092610042901212,
"delta_retention_norm_ratio": 0.8662231905394834,
"delta_cosine": 0.6795542546803904,
"relative_weight_rmse": 0.01174172346514312
},
{
"index": 168,
"total": 252,
"projection": "layers.3.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0316,
"parameters": 4194304,
"reference_delta_l2": 0.9495305863978187,
"retained_delta_l2": 0.8565283992640378,
"delta_retention_norm_ratio": 0.9020545641540647,
"delta_cosine": 0.7163760029101252,
"relative_weight_rmse": 0.012409328467697229
},
{
"index": 169,
"total": 252,
"projection": "layers.30.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4665,
"parameters": 50331648,
"reference_delta_l2": 3.889297551407395,
"retained_delta_l2": 3.689156905368003,
"delta_retention_norm_ratio": 0.9485406700325697,
"delta_cosine": 0.7668341102234857,
"relative_weight_rmse": 0.013138320072956873
},
{
"index": 170,
"total": 252,
"projection": "layers.30.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4462,
"parameters": 50331648,
"reference_delta_l2": 5.417024655242779,
"retained_delta_l2": 5.416135364193542,
"delta_retention_norm_ratio": 0.9998358340406709,
"delta_cosine": 0.8313714569311573,
"relative_weight_rmse": 0.0169564909913298
},
{
"index": 171,
"total": 252,
"projection": "layers.30.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4666,
"parameters": 50331648,
"reference_delta_l2": 3.819126871634718,
"retained_delta_l2": 3.5405898546090633,
"delta_retention_norm_ratio": 0.9270678805948043,
"delta_cosine": 0.7404093729714868,
"relative_weight_rmse": 0.013424233084753504
},
{
"index": 172,
"total": 252,
"projection": "layers.30.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0368,
"parameters": 4194304,
"reference_delta_l2": 1.4963148357683524,
"retained_delta_l2": 1.5083287477569252,
"delta_retention_norm_ratio": 1.0080290001150751,
"delta_cosine": 0.8529398641780789,
"relative_weight_rmse": 0.01660992040173974
},
{
"index": 173,
"total": 252,
"projection": "layers.30.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1584,
"parameters": 16777216,
"reference_delta_l2": 1.8699612129395147,
"retained_delta_l2": 1.6939845275148544,
"delta_retention_norm_ratio": 0.9058928686825375,
"delta_cosine": 0.7177428471689205,
"relative_weight_rmse": 0.012868674227738065
},
{
"index": 174,
"total": 252,
"projection": "layers.30.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1696,
"parameters": 16777216,
"reference_delta_l2": 3.026867626466198,
"retained_delta_l2": 3.038492448510906,
"delta_retention_norm_ratio": 1.0038405452366215,
"delta_cosine": 0.8555437465099579,
"relative_weight_rmse": 0.016312385117527632
},
{
"index": 175,
"total": 252,
"projection": "layers.30.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0381,
"parameters": 4194304,
"reference_delta_l2": 1.3103580973978421,
"retained_delta_l2": 1.2727454662392619,
"delta_retention_norm_ratio": 0.971295914274676,
"delta_cosine": 0.8016130016801397,
"relative_weight_rmse": 0.013926367269040557
},
{
"index": 176,
"total": 252,
"projection": "layers.31.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4622,
"parameters": 50331648,
"reference_delta_l2": 4.112557465011257,
"retained_delta_l2": 3.9471197082600393,
"delta_retention_norm_ratio": 0.9597725361508681,
"delta_cosine": 0.7806248768891962,
"relative_weight_rmse": 0.013382597606866855
},
{
"index": 177,
"total": 252,
"projection": "layers.31.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4048,
"parameters": 50331648,
"reference_delta_l2": 4.223828711060859,
"retained_delta_l2": 4.077127989569533,
"delta_retention_norm_ratio": 0.9652683071386953,
"delta_cosine": 0.7862870963489712,
"relative_weight_rmse": 0.014775501836585539
},
{
"index": 178,
"total": 252,
"projection": "layers.31.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4298,
"parameters": 50331648,
"reference_delta_l2": 3.749679731544607,
"retained_delta_l2": 3.46268146614541,
"delta_retention_norm_ratio": 0.9234605923847866,
"delta_cosine": 0.7369913069010527,
"relative_weight_rmse": 0.013133966561587464
},
{
"index": 179,
"total": 252,
"projection": "layers.31.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.036,
"parameters": 4194304,
"reference_delta_l2": 1.1312737989394437,
"retained_delta_l2": 1.1077378579281763,
"delta_retention_norm_ratio": 0.9791951859635288,
"delta_cosine": 0.8130418253379471,
"relative_weight_rmse": 0.014432338472134166
},
{
"index": 180,
"total": 252,
"projection": "layers.31.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1567,
"parameters": 16777216,
"reference_delta_l2": 1.928063724419499,
"retained_delta_l2": 1.7795399112745443,
"delta_retention_norm_ratio": 0.922967373295884,
"delta_cosine": 0.738932673971109,
"relative_weight_rmse": 0.012544046940459179
},
{
"index": 181,
"total": 252,
"projection": "layers.31.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1575,
"parameters": 16777216,
"reference_delta_l2": 3.0953250374266994,
"retained_delta_l2": 3.118746367727834,
"delta_retention_norm_ratio": 1.0075666787875066,
"delta_cosine": 0.8592927365671965,
"relative_weight_rmse": 0.016362367411292777
},
{
"index": 182,
"total": 252,
"projection": "layers.31.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0355,
"parameters": 4194304,
"reference_delta_l2": 0.9145042079702408,
"retained_delta_l2": 0.8072281483043932,
"delta_retention_norm_ratio": 0.8826948430298107,
"delta_cosine": 0.699848976733929,
"relative_weight_rmse": 0.01125666329561259
},
{
"index": 183,
"total": 252,
"projection": "layers.32.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.5698,
"parameters": 50331648,
"reference_delta_l2": 3.0550519151629563,
"retained_delta_l2": 2.7030221189348613,
"delta_retention_norm_ratio": 0.8847712556107848,
"delta_cosine": 0.6984499045656849,
"relative_weight_rmse": 0.011293767506202028
},
{
"index": 184,
"total": 252,
"projection": "layers.32.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.483,
"parameters": 50331648,
"reference_delta_l2": 5.754062997691919,
"retained_delta_l2": 5.806999408392191,
"delta_retention_norm_ratio": 1.0091998316183723,
"delta_cosine": 0.8489266209799153,
"relative_weight_rmse": 0.017508291949826876
},
{
"index": 185,
"total": 252,
"projection": "layers.32.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.5024,
"parameters": 50331648,
"reference_delta_l2": 4.183795657381237,
"retained_delta_l2": 3.9679568447305917,
"delta_retention_norm_ratio": 0.9484107661257658,
"delta_cosine": 0.7674900854696458,
"relative_weight_rmse": 0.0138468582380262
},
{
"index": 186,
"total": 252,
"projection": "layers.32.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.054,
"parameters": 4194304,
"reference_delta_l2": 1.0770161828840301,
"retained_delta_l2": 1.0427000998279297,
"delta_retention_norm_ratio": 0.9681378203954103,
"delta_cosine": 0.7990333694275377,
"relative_weight_rmse": 0.014094558298269931
},
{
"index": 187,
"total": 252,
"projection": "layers.32.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1914,
"parameters": 16777216,
"reference_delta_l2": 1.953978887850504,
"retained_delta_l2": 1.795990008493363,
"delta_retention_norm_ratio": 0.9191450427947365,
"delta_cosine": 0.7328364849623904,
"relative_weight_rmse": 0.012912622199988916
},
{
"index": 188,
"total": 252,
"projection": "layers.32.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1796,
"parameters": 16777216,
"reference_delta_l2": 3.5136802373394898,
"retained_delta_l2": 3.5736288757330885,
"delta_retention_norm_ratio": 1.0170614951686643,
"delta_cosine": 0.8841806139891618,
"relative_weight_rmse": 0.01703225774601036
},
{
"index": 189,
"total": 252,
"projection": "layers.32.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.035,
"parameters": 4194304,
"reference_delta_l2": 1.1058690941584695,
"retained_delta_l2": 1.0362385630081135,
"delta_retention_norm_ratio": 0.9370354669298877,
"delta_cosine": 0.7584594640024848,
"relative_weight_rmse": 0.012604906964990689
},
{
"index": 190,
"total": 252,
"projection": "layers.33.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.5121,
"parameters": 50331648,
"reference_delta_l2": 4.682289042544063,
"retained_delta_l2": 4.596022338556956,
"delta_retention_norm_ratio": 0.9815759550076312,
"delta_cosine": 0.8099761594481375,
"relative_weight_rmse": 0.014408382780754198
},
{
"index": 191,
"total": 252,
"projection": "layers.33.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4865,
"parameters": 50331648,
"reference_delta_l2": 4.014127092269512,
"retained_delta_l2": 3.8309685730944287,
"delta_retention_norm_ratio": 0.9543715196442549,
"delta_cosine": 0.7715101511205649,
"relative_weight_rmse": 0.014752253148085384
},
{
"index": 192,
"total": 252,
"projection": "layers.33.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4507,
"parameters": 50331648,
"reference_delta_l2": 4.552880855051247,
"retained_delta_l2": 4.416455711552382,
"delta_retention_norm_ratio": 0.9700354241979546,
"delta_cosine": 0.7956230998996326,
"relative_weight_rmse": 0.014229351268283607
},
{
"index": 193,
"total": 252,
"projection": "layers.33.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.044,
"parameters": 4194304,
"reference_delta_l2": 1.1714884759216244,
"retained_delta_l2": 1.162239862543118,
"delta_retention_norm_ratio": 0.9921052459596494,
"delta_cosine": 0.8277283945417799,
"relative_weight_rmse": 0.015549435727050668
},
{
"index": 194,
"total": 252,
"projection": "layers.33.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1469,
"parameters": 16777216,
"reference_delta_l2": 1.956410571334175,
"retained_delta_l2": 1.8080818840710324,
"delta_retention_norm_ratio": 0.92418325200421,
"delta_cosine": 0.741240418934167,
"relative_weight_rmse": 0.012297194196049766
},
{
"index": 195,
"total": 252,
"projection": "layers.33.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1398,
"parameters": 16777216,
"reference_delta_l2": 1.6953487201611341,
"retained_delta_l2": 1.5631024817956607,
"delta_retention_norm_ratio": 0.9219946688296056,
"delta_cosine": 0.7427254714025582,
"relative_weight_rmse": 0.011923941274140972
},
{
"index": 196,
"total": 252,
"projection": "layers.33.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 1.2341457244056702,
"retained_delta_l2": 1.1847387794147948,
"delta_retention_norm_ratio": 0.9599666846355049,
"delta_cosine": 0.7863607664004597,
"relative_weight_rmse": 0.0127283179428275
},
{
"index": 197,
"total": 252,
"projection": "layers.34.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4646,
"parameters": 50331648,
"reference_delta_l2": 3.254529718921971,
"retained_delta_l2": 3.0231779689918046,
"delta_retention_norm_ratio": 0.9289139230823188,
"delta_cosine": 0.745341223879758,
"relative_weight_rmse": 0.011839365452368134
},
{
"index": 198,
"total": 252,
"projection": "layers.34.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.5288,
"parameters": 50331648,
"reference_delta_l2": 3.5153039041836927,
"retained_delta_l2": 3.3118616785388983,
"delta_retention_norm_ratio": 0.9421267033548165,
"delta_cosine": 0.7571959815236355,
"relative_weight_rmse": 0.013113854354279319
},
{
"index": 199,
"total": 252,
"projection": "layers.34.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4858,
"parameters": 50331648,
"reference_delta_l2": 3.1220471900841384,
"retained_delta_l2": 2.79862654980446,
"delta_retention_norm_ratio": 0.896407510652982,
"delta_cosine": 0.7099098019678689,
"relative_weight_rmse": 0.011418600434844762
},
{
"index": 200,
"total": 252,
"projection": "layers.34.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0374,
"parameters": 4194304,
"reference_delta_l2": 0.7903795316389047,
"retained_delta_l2": 0.7292891800891133,
"delta_retention_norm_ratio": 0.9227075738878048,
"delta_cosine": 0.753058670486461,
"relative_weight_rmse": 0.011645255779586767
},
{
"index": 201,
"total": 252,
"projection": "layers.34.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1537,
"parameters": 16777216,
"reference_delta_l2": 2.37039100209066,
"retained_delta_l2": 2.3003068185121704,
"delta_retention_norm_ratio": 0.9704334923999138,
"delta_cosine": 0.7933461271137942,
"relative_weight_rmse": 0.014058076907602178
},
{
"index": 202,
"total": 252,
"projection": "layers.34.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1521,
"parameters": 16777216,
"reference_delta_l2": 2.7976161115782117,
"retained_delta_l2": 2.811384210003641,
"delta_retention_norm_ratio": 1.0049213680062996,
"delta_cosine": 0.8600720864190587,
"relative_weight_rmse": 0.015298939500820763
},
{
"index": 203,
"total": 252,
"projection": "layers.34.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0346,
"parameters": 4194304,
"reference_delta_l2": 0.9822597916177848,
"retained_delta_l2": 0.8778726541723202,
"delta_retention_norm_ratio": 0.8937275674559185,
"delta_cosine": 0.7156215745230402,
"relative_weight_rmse": 0.01141196556432579
},
{
"index": 204,
"total": 252,
"projection": "layers.35.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.485,
"parameters": 50331648,
"reference_delta_l2": 2.3278731184321253,
"retained_delta_l2": 2.080644748742452,
"delta_retention_norm_ratio": 0.8937964583498489,
"delta_cosine": 0.7142308098011656,
"relative_weight_rmse": 0.009799166249988927
},
{
"index": 205,
"total": 252,
"projection": "layers.35.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4459,
"parameters": 50331648,
"reference_delta_l2": 2.624490457214364,
"retained_delta_l2": 2.2907097766541056,
"delta_retention_norm_ratio": 0.8728207680684297,
"delta_cosine": 0.6874999700162999,
"relative_weight_rmse": 0.010525403547088693
},
{
"index": 206,
"total": 252,
"projection": "layers.35.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4468,
"parameters": 50331648,
"reference_delta_l2": 2.6619026667954984,
"retained_delta_l2": 2.2893727982491834,
"delta_retention_norm_ratio": 0.8600512809152482,
"delta_cosine": 0.6746867834542335,
"relative_weight_rmse": 0.010345328694455945
},
{
"index": 207,
"total": 252,
"projection": "layers.35.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0353,
"parameters": 4194304,
"reference_delta_l2": 0.8326687018400024,
"retained_delta_l2": 0.7626197331487948,
"delta_retention_norm_ratio": 0.9158741423372636,
"delta_cosine": 0.7341666872680872,
"relative_weight_rmse": 0.012228272670653023
},
{
"index": 208,
"total": 252,
"projection": "layers.35.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1528,
"parameters": 16777216,
"reference_delta_l2": 1.9770264378206903,
"retained_delta_l2": 1.8754233496711648,
"delta_retention_norm_ratio": 0.9486081287504053,
"delta_cosine": 0.7674469102145189,
"relative_weight_rmse": 0.012978968041747977
},
{
"index": 209,
"total": 252,
"projection": "layers.35.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1493,
"parameters": 16777216,
"reference_delta_l2": 1.9240252066305126,
"retained_delta_l2": 1.8365858347119743,
"delta_retention_norm_ratio": 0.9545539363948,
"delta_cosine": 0.7748208172971213,
"relative_weight_rmse": 0.013093962966165837
},
{
"index": 210,
"total": 252,
"projection": "layers.35.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0315,
"parameters": 4194304,
"reference_delta_l2": 0.9011555191613387,
"retained_delta_l2": 0.7946283892491797,
"delta_retention_norm_ratio": 0.8817882955305,
"delta_cosine": 0.6985563372860758,
"relative_weight_rmse": 0.01139125996910996
},
{
"index": 211,
"total": 252,
"projection": "layers.4.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4602,
"parameters": 50331648,
"reference_delta_l2": 2.2284041107493677,
"retained_delta_l2": 1.912308090646361,
"delta_retention_norm_ratio": 0.8581513924793874,
"delta_cosine": 0.6733987038950043,
"relative_weight_rmse": 0.010848835221468896
},
{
"index": 212,
"total": 252,
"projection": "layers.4.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.599,
"parameters": 50331648,
"reference_delta_l2": 3.3223631680557593,
"retained_delta_l2": 2.9321477421189353,
"delta_retention_norm_ratio": 0.8825488346100413,
"delta_cosine": 0.7099485187697749,
"relative_weight_rmse": 0.010843206512101423
},
{
"index": 213,
"total": 252,
"projection": "layers.4.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.6199,
"parameters": 50331648,
"reference_delta_l2": 3.3068778367400546,
"retained_delta_l2": 3.1849764378593615,
"delta_retention_norm_ratio": 0.9631370117376744,
"delta_cosine": 0.7957763627454438,
"relative_weight_rmse": 0.014028163445487966
},
{
"index": 214,
"total": 252,
"projection": "layers.4.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0515,
"parameters": 4194304,
"reference_delta_l2": 0.5943265881958918,
"retained_delta_l2": 0.44923539580877975,
"delta_retention_norm_ratio": 0.7558729572783482,
"delta_cosine": 0.5787146182002962,
"relative_weight_rmse": 0.009478202685571882
},
{
"index": 215,
"total": 252,
"projection": "layers.4.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.2143,
"parameters": 16777216,
"reference_delta_l2": 1.6315352222274537,
"retained_delta_l2": 1.4182049594601929,
"delta_retention_norm_ratio": 0.8692456896664408,
"delta_cosine": 0.6814565809601563,
"relative_weight_rmse": 0.01181962232864801
},
{
"index": 216,
"total": 252,
"projection": "layers.4.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.2219,
"parameters": 16777216,
"reference_delta_l2": 1.2635314101224357,
"retained_delta_l2": 0.9992829602515754,
"delta_retention_norm_ratio": 0.7908651516266978,
"delta_cosine": 0.6086767805152506,
"relative_weight_rmse": 0.010253748835538172
},
{
"index": 217,
"total": 252,
"projection": "layers.4.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0519,
"parameters": 4194304,
"reference_delta_l2": 1.1348673716260338,
"retained_delta_l2": 1.0771443670257408,
"delta_retention_norm_ratio": 0.9491367836951838,
"delta_cosine": 0.769077303214833,
"relative_weight_rmse": 0.013529424722191423
},
{
"index": 218,
"total": 252,
"projection": "layers.5.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.6594,
"parameters": 50331648,
"reference_delta_l2": 1.8096449488022504,
"retained_delta_l2": 1.348441732456797,
"delta_retention_norm_ratio": 0.7451415999305776,
"delta_cosine": 0.5702021105966943,
"relative_weight_rmse": 0.009066344475002007
},
{
"index": 219,
"total": 252,
"projection": "layers.5.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4398,
"parameters": 50331648,
"reference_delta_l2": 2.9581444466825837,
"retained_delta_l2": 2.494926968208721,
"delta_retention_norm_ratio": 0.8434094457444975,
"delta_cosine": 0.6608254790586037,
"relative_weight_rmse": 0.010006429742271778
},
{
"index": 220,
"total": 252,
"projection": "layers.5.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4416,
"parameters": 50331648,
"reference_delta_l2": 3.1847576094724808,
"retained_delta_l2": 2.9718679926436264,
"delta_retention_norm_ratio": 0.9331535887705699,
"delta_cosine": 0.7499739613494094,
"relative_weight_rmse": 0.013146130478080225
},
{
"index": 221,
"total": 252,
"projection": "layers.5.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0365,
"parameters": 4194304,
"reference_delta_l2": 0.621163379722661,
"retained_delta_l2": 0.4892789656444176,
"delta_retention_norm_ratio": 0.7876816013572346,
"delta_cosine": 0.6083900925778365,
"relative_weight_rmse": 0.0098734492791527
},
{
"index": 222,
"total": 252,
"projection": "layers.5.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1603,
"parameters": 16777216,
"reference_delta_l2": 1.4706815754227913,
"retained_delta_l2": 1.2361778289034484,
"delta_retention_norm_ratio": 0.840547573017682,
"delta_cosine": 0.6525468273811887,
"relative_weight_rmse": 0.011366349618374402
},
{
"index": 223,
"total": 252,
"projection": "layers.5.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.2023,
"parameters": 16777216,
"reference_delta_l2": 1.471888310223771,
"retained_delta_l2": 1.2577901637049853,
"delta_retention_norm_ratio": 0.8545418527807749,
"delta_cosine": 0.6710499482341943,
"relative_weight_rmse": 0.011205519981466199
},
{
"index": 224,
"total": 252,
"projection": "layers.5.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0475,
"parameters": 4194304,
"reference_delta_l2": 1.0478447746472717,
"retained_delta_l2": 0.9835083975713196,
"delta_retention_norm_ratio": 0.9386012330904554,
"delta_cosine": 0.761201483781111,
"relative_weight_rmse": 0.012887362532565526
},
{
"index": 225,
"total": 252,
"projection": "layers.6.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.6356,
"parameters": 50331648,
"reference_delta_l2": 4.767733159016775,
"retained_delta_l2": 4.780823858164376,
"delta_retention_norm_ratio": 1.0027456862015953,
"delta_cosine": 0.8442620247623968,
"relative_weight_rmse": 0.015434869473842374
},
{
"index": 226,
"total": 252,
"projection": "layers.6.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.5937,
"parameters": 50331648,
"reference_delta_l2": 3.1127853658519617,
"retained_delta_l2": 2.7231327704813753,
"delta_retention_norm_ratio": 0.8748218879318911,
"delta_cosine": 0.6925269462917778,
"relative_weight_rmse": 0.01069907289296045
},
{
"index": 227,
"total": 252,
"projection": "layers.6.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.5698,
"parameters": 50331648,
"reference_delta_l2": 3.3891323795442707,
"retained_delta_l2": 3.2012610817942666,
"delta_retention_norm_ratio": 0.9445665507538342,
"delta_cosine": 0.7632462715161046,
"relative_weight_rmse": 0.013226556935270523
},
{
"index": 228,
"total": 252,
"projection": "layers.6.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0366,
"parameters": 4194304,
"reference_delta_l2": 0.5936267916366719,
"retained_delta_l2": 0.44849650488225523,
"delta_retention_norm_ratio": 0.7555193114611927,
"delta_cosine": 0.5786583185869822,
"relative_weight_rmse": 0.009591797724909461
},
{
"index": 229,
"total": 252,
"projection": "layers.6.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1548,
"parameters": 16777216,
"reference_delta_l2": 1.3662937933533277,
"retained_delta_l2": 1.1081746998525444,
"delta_retention_norm_ratio": 0.8110808270106568,
"delta_cosine": 0.6268541267404001,
"relative_weight_rmse": 0.010402945447451687
},
{
"index": 230,
"total": 252,
"projection": "layers.6.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1523,
"parameters": 16777216,
"reference_delta_l2": 1.404094947081925,
"retained_delta_l2": 1.1650367131920587,
"delta_retention_norm_ratio": 0.8297421165237496,
"delta_cosine": 0.644835786205775,
"relative_weight_rmse": 0.010728688531155947
},
{
"index": 231,
"total": 252,
"projection": "layers.6.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0351,
"parameters": 4194304,
"reference_delta_l2": 0.9868940636844633,
"retained_delta_l2": 0.9111959620082837,
"delta_retention_norm_ratio": 0.9232966288260273,
"delta_cosine": 0.7397924912744128,
"relative_weight_rmse": 0.012403267907682316
},
{
"index": 232,
"total": 252,
"projection": "layers.7.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4885,
"parameters": 50331648,
"reference_delta_l2": 1.8496785554608322,
"retained_delta_l2": 1.3698546632482855,
"delta_retention_norm_ratio": 0.7405906605793986,
"delta_cosine": 0.5660935642501336,
"relative_weight_rmse": 0.00897799409859664
},
{
"index": 233,
"total": 252,
"projection": "layers.7.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4093,
"parameters": 50331648,
"reference_delta_l2": 2.9020701264005404,
"retained_delta_l2": 2.487640078236522,
"delta_retention_norm_ratio": 0.8571950262697342,
"delta_cosine": 0.6741874333990765,
"relative_weight_rmse": 0.010365587764481267
},
{
"index": 234,
"total": 252,
"projection": "layers.7.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4031,
"parameters": 50331648,
"reference_delta_l2": 2.9786677730928384,
"retained_delta_l2": 2.695959894362702,
"delta_retention_norm_ratio": 0.9050891538546467,
"delta_cosine": 0.7168065799801924,
"relative_weight_rmse": 0.012392544160721922
},
{
"index": 235,
"total": 252,
"projection": "layers.7.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0331,
"parameters": 4194304,
"reference_delta_l2": 0.7232052892145597,
"retained_delta_l2": 0.6140335767614647,
"delta_retention_norm_ratio": 0.8490446432275665,
"delta_cosine": 0.662784549584905,
"relative_weight_rmse": 0.010778714108441558
},
{
"index": 236,
"total": 252,
"projection": "layers.7.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1386,
"parameters": 16777216,
"reference_delta_l2": 1.3679511683806942,
"retained_delta_l2": 1.1087497139751692,
"delta_retention_norm_ratio": 0.8105184889659814,
"delta_cosine": 0.6256539219338475,
"relative_weight_rmse": 0.010561121610774705
},
{
"index": 237,
"total": 252,
"projection": "layers.7.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1474,
"parameters": 16777216,
"reference_delta_l2": 1.6880094197968922,
"retained_delta_l2": 1.5265190547660956,
"delta_retention_norm_ratio": 0.9043308863464591,
"delta_cosine": 0.7181951299018756,
"relative_weight_rmse": 0.012128721524007775
},
{
"index": 238,
"total": 252,
"projection": "layers.7.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0347,
"parameters": 4194304,
"reference_delta_l2": 0.9924121578930758,
"retained_delta_l2": 0.8964162423979289,
"delta_retention_norm_ratio": 0.9032701133982988,
"delta_cosine": 0.7184582477278942,
"relative_weight_rmse": 0.012007815299748084
},
{
"index": 239,
"total": 252,
"projection": "layers.8.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4828,
"parameters": 50331648,
"reference_delta_l2": 2.64691221760978,
"retained_delta_l2": 2.2857947357503132,
"delta_retention_norm_ratio": 0.8635702841004816,
"delta_cosine": 0.677817519256314,
"relative_weight_rmse": 0.011074492759919266
},
{
"index": 240,
"total": 252,
"projection": "layers.8.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4374,
"parameters": 50331648,
"reference_delta_l2": 2.2096077673770287,
"retained_delta_l2": 1.6672304623673437,
"delta_retention_norm_ratio": 0.7545368399688748,
"delta_cosine": 0.577845816218415,
"relative_weight_rmse": 0.009490851973574195
},
{
"index": 241,
"total": 252,
"projection": "layers.8.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4032,
"parameters": 50331648,
"reference_delta_l2": 2.6833324735021264,
"retained_delta_l2": 2.2885269179734777,
"delta_retention_norm_ratio": 0.852867447687773,
"delta_cosine": 0.6661050884361409,
"relative_weight_rmse": 0.01143929142860376
},
{
"index": 242,
"total": 252,
"projection": "layers.8.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0325,
"parameters": 4194304,
"reference_delta_l2": 0.894970914476173,
"retained_delta_l2": 0.8119598485870462,
"delta_retention_norm_ratio": 0.9072471914489945,
"delta_cosine": 0.7203591771205954,
"relative_weight_rmse": 0.012141446313081472
},
{
"index": 243,
"total": 252,
"projection": "layers.8.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1396,
"parameters": 16777216,
"reference_delta_l2": 1.5742749243899192,
"retained_delta_l2": 1.3489051259474591,
"delta_retention_norm_ratio": 0.8568421595549469,
"delta_cosine": 0.6708010472652016,
"relative_weight_rmse": 0.01131591184137918
},
{
"index": 244,
"total": 252,
"projection": "layers.8.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1385,
"parameters": 16777216,
"reference_delta_l2": 1.5968418540652398,
"retained_delta_l2": 1.3863617481042254,
"delta_retention_norm_ratio": 0.8681897612934091,
"delta_cosine": 0.6795901804160726,
"relative_weight_rmse": 0.011850308263669078
},
{
"index": 245,
"total": 252,
"projection": "layers.8.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 0.7965422909630013,
"retained_delta_l2": 0.6606204273462776,
"delta_retention_norm_ratio": 0.8293601417541845,
"delta_cosine": 0.6434455112758906,
"relative_weight_rmse": 0.010629416686343467
},
{
"index": 246,
"total": 252,
"projection": "layers.9.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4723,
"parameters": 50331648,
"reference_delta_l2": 2.47232075829251,
"retained_delta_l2": 2.101686287322884,
"delta_retention_norm_ratio": 0.8500864138577222,
"delta_cosine": 0.6644696748878688,
"relative_weight_rmse": 0.010893869061863947
},
{
"index": 247,
"total": 252,
"projection": "layers.9.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4518,
"parameters": 50331648,
"reference_delta_l2": 2.7120365927867636,
"retained_delta_l2": 2.2251890219404666,
"delta_retention_norm_ratio": 0.8204863562161472,
"delta_cosine": 0.6364763574864426,
"relative_weight_rmse": 0.010406600532479828
},
{
"index": 248,
"total": 252,
"projection": "layers.9.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4461,
"parameters": 50331648,
"reference_delta_l2": 3.0321729076256774,
"retained_delta_l2": 2.751587013465937,
"delta_retention_norm_ratio": 0.9074637552976981,
"delta_cosine": 0.7224490158947353,
"relative_weight_rmse": 0.01240354235029758
},
{
"index": 249,
"total": 252,
"projection": "layers.9.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0352,
"parameters": 4194304,
"reference_delta_l2": 0.8394663585466136,
"retained_delta_l2": 0.7573787040711727,
"delta_retention_norm_ratio": 0.9022144799017783,
"delta_cosine": 0.7183590459382686,
"relative_weight_rmse": 0.011797098425141166
},
{
"index": 250,
"total": 252,
"projection": "layers.9.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1532,
"parameters": 16777216,
"reference_delta_l2": 1.5269071510640992,
"retained_delta_l2": 1.301707145134331,
"delta_retention_norm_ratio": 0.8525123117192641,
"delta_cosine": 0.6669816972555566,
"relative_weight_rmse": 0.01116571897517731
},
{
"index": 251,
"total": 252,
"projection": "layers.9.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1532,
"parameters": 16777216,
"reference_delta_l2": 1.8062023392131712,
"retained_delta_l2": 1.659700464068496,
"delta_retention_norm_ratio": 0.9188895551931933,
"delta_cosine": 0.7351277573769627,
"relative_weight_rmse": 0.012708519539599299
},
{
"index": 252,
"total": 252,
"projection": "layers.9.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.035,
"parameters": 4194304,
"reference_delta_l2": 0.801285046068713,
"retained_delta_l2": 0.6708872938731084,
"delta_retention_norm_ratio": 0.8372642134838711,
"delta_cosine": 0.6533930656662315,
"relative_weight_rmse": 0.01043079318737646
}
]
}