mrjackspade's picture
Publish step 2540 merge manifest
47691ad 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_00002540/adapters.safetensors",
"adapter_sha256": "f8fe52d0b625e933521e5d9346da158bd5462738c1af4ea3a046397027f25569",
"adapter_global_step": "2540",
"output": "/workspace/Ideogram/artifacts/huggingface_release/s020_v1/qwen3vl_8b_ideogram4_nl_s020_v1_step_00002540_fp8_scaled.safetensors",
"output_bytes": 10588637800,
"output_sha256": "4b164ff3c6ef27af89cb14c1bdb5b6e6d20c72bf258308b5ee77b811f976dcf7",
"output_tensors": 1254,
"merged_projections": 252,
"lora_rank": 64,
"lora_alpha": 64.0,
"elapsed_seconds": 64.148,
"projections": [
{
"index": 1,
"total": 252,
"projection": "layers.0.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4236,
"parameters": 50331648,
"reference_delta_l2": 3.126840741732905,
"retained_delta_l2": 2.847031594462536,
"delta_retention_norm_ratio": 0.9105137835976584,
"delta_cosine": 0.725848228993972,
"relative_weight_rmse": 0.012307521162095323
},
{
"index": 2,
"total": 252,
"projection": "layers.0.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3782,
"parameters": 50331648,
"reference_delta_l2": 3.3155131653878125,
"retained_delta_l2": 3.0297597793802833,
"delta_retention_norm_ratio": 0.9138132253581008,
"delta_cosine": 0.7264097571772404,
"relative_weight_rmse": 0.013511860068981345
},
{
"index": 3,
"total": 252,
"projection": "layers.0.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.376,
"parameters": 50331648,
"reference_delta_l2": 2.93946837412769,
"retained_delta_l2": 2.629080052603913,
"delta_retention_norm_ratio": 0.8944066470468874,
"delta_cosine": 0.7074603179950618,
"relative_weight_rmse": 0.013038171006044602
},
{
"index": 4,
"total": 252,
"projection": "layers.0.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 0.6028464680738916,
"retained_delta_l2": 0.4418626005952603,
"delta_retention_norm_ratio": 0.7329604202659119,
"delta_cosine": 0.5606693910696365,
"relative_weight_rmse": 0.009013772200868762
},
{
"index": 5,
"total": 252,
"projection": "layers.0.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1262,
"parameters": 16777216,
"reference_delta_l2": 1.973553111162272,
"retained_delta_l2": 1.861623920055665,
"delta_retention_norm_ratio": 0.9432854426498362,
"delta_cosine": 0.7628375185005537,
"relative_weight_rmse": 0.01362450712127664
},
{
"index": 6,
"total": 252,
"projection": "layers.0.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1366,
"parameters": 16777216,
"reference_delta_l2": 1.225185293772824,
"retained_delta_l2": 0.9533197670271886,
"delta_retention_norm_ratio": 0.7781025220206036,
"delta_cosine": 0.5987807928868673,
"relative_weight_rmse": 0.009804670894225527
},
{
"index": 7,
"total": 252,
"projection": "layers.0.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0293,
"parameters": 4194304,
"reference_delta_l2": 0.7953509807620676,
"retained_delta_l2": 0.6926766334641751,
"delta_retention_norm_ratio": 0.8709068703234454,
"delta_cosine": 0.6838780163989387,
"relative_weight_rmse": 0.01198416383607124
},
{
"index": 8,
"total": 252,
"projection": "layers.1.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4275,
"parameters": 50331648,
"reference_delta_l2": 2.0361246515843954,
"retained_delta_l2": 1.900496503757514,
"delta_retention_norm_ratio": 0.933389074327378,
"delta_cosine": 0.756480270633347,
"relative_weight_rmse": 0.011983444253769262
},
{
"index": 9,
"total": 252,
"projection": "layers.1.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3845,
"parameters": 50331648,
"reference_delta_l2": 3.827864663760918,
"retained_delta_l2": 3.7945581449684007,
"delta_retention_norm_ratio": 0.9912989298948219,
"delta_cosine": 0.8456527848261863,
"relative_weight_rmse": 0.014356902032464577
},
{
"index": 10,
"total": 252,
"projection": "layers.1.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3768,
"parameters": 50331648,
"reference_delta_l2": 4.228079577870509,
"retained_delta_l2": 4.225241074129967,
"delta_retention_norm_ratio": 0.9993286541352253,
"delta_cosine": 0.9294420997507352,
"relative_weight_rmse": 0.01576108016664263
},
{
"index": 11,
"total": 252,
"projection": "layers.1.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 0.6761678837629111,
"retained_delta_l2": 0.5551086850164083,
"delta_retention_norm_ratio": 0.8209628087143066,
"delta_cosine": 0.6372181263878011,
"relative_weight_rmse": 0.010816268701898937
},
{
"index": 12,
"total": 252,
"projection": "layers.1.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1266,
"parameters": 16777216,
"reference_delta_l2": 1.7122907077208531,
"retained_delta_l2": 1.5313364953859028,
"delta_retention_norm_ratio": 0.8943203910883745,
"delta_cosine": 0.7050326938351207,
"relative_weight_rmse": 0.012655305524148125
},
{
"index": 13,
"total": 252,
"projection": "layers.1.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1263,
"parameters": 16777216,
"reference_delta_l2": 2.0928366814121127,
"retained_delta_l2": 2.0236042080333014,
"delta_retention_norm_ratio": 0.9669193138701594,
"delta_cosine": 0.7927232891654293,
"relative_weight_rmse": 0.014637441165904277
},
{
"index": 14,
"total": 252,
"projection": "layers.1.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.029,
"parameters": 4194304,
"reference_delta_l2": 0.9145859087024274,
"retained_delta_l2": 0.8318342758362796,
"delta_retention_norm_ratio": 0.9095201095066595,
"delta_cosine": 0.7222397921654805,
"relative_weight_rmse": 0.01243292081589284
},
{
"index": 15,
"total": 252,
"projection": "layers.10.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4172,
"parameters": 50331648,
"reference_delta_l2": 2.81236752270305,
"retained_delta_l2": 2.493766399212893,
"delta_retention_norm_ratio": 0.8867142644344223,
"delta_cosine": 0.7007063524167806,
"relative_weight_rmse": 0.01166425786353078
},
{
"index": 16,
"total": 252,
"projection": "layers.10.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3757,
"parameters": 50331648,
"reference_delta_l2": 2.3905832278522254,
"retained_delta_l2": 1.8532254458493804,
"delta_retention_norm_ratio": 0.7752189609036854,
"delta_cosine": 0.5960260535930171,
"relative_weight_rmse": 0.009653432972177991
},
{
"index": 17,
"total": 252,
"projection": "layers.10.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3774,
"parameters": 50331648,
"reference_delta_l2": 2.8556621352623095,
"retained_delta_l2": 2.5270634634012414,
"delta_retention_norm_ratio": 0.8849308299453695,
"delta_cosine": 0.6974295939624287,
"relative_weight_rmse": 0.011906604700482923
},
{
"index": 18,
"total": 252,
"projection": "layers.10.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.03,
"parameters": 4194304,
"reference_delta_l2": 0.897891377626497,
"retained_delta_l2": 0.8251920079584362,
"delta_retention_norm_ratio": 0.9190332244193773,
"delta_cosine": 0.7371071941855177,
"relative_weight_rmse": 0.011969155540596983
},
{
"index": 19,
"total": 252,
"projection": "layers.10.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1257,
"parameters": 16777216,
"reference_delta_l2": 1.7752074338164896,
"retained_delta_l2": 1.586539641118029,
"delta_retention_norm_ratio": 0.893720706040056,
"delta_cosine": 0.7069774975623085,
"relative_weight_rmse": 0.012172454143964525
},
{
"index": 20,
"total": 252,
"projection": "layers.10.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1253,
"parameters": 16777216,
"reference_delta_l2": 2.1185128595130265,
"retained_delta_l2": 2.020718756582176,
"delta_retention_norm_ratio": 0.9538383246097785,
"delta_cosine": 0.7798847138295467,
"relative_weight_rmse": 0.01345056123715535
},
{
"index": 21,
"total": 252,
"projection": "layers.10.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.029,
"parameters": 4194304,
"reference_delta_l2": 0.9100838136534896,
"retained_delta_l2": 0.8128102757212403,
"delta_retention_norm_ratio": 0.8931158466144463,
"delta_cosine": 0.7113298638535298,
"relative_weight_rmse": 0.011205869215500954
},
{
"index": 22,
"total": 252,
"projection": "layers.11.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4191,
"parameters": 50331648,
"reference_delta_l2": 2.6097737594933634,
"retained_delta_l2": 2.2340911828455416,
"delta_retention_norm_ratio": 0.8560478373724039,
"delta_cosine": 0.6706686430394193,
"relative_weight_rmse": 0.011050285955504799
},
{
"index": 23,
"total": 252,
"projection": "layers.11.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.378,
"parameters": 50331648,
"reference_delta_l2": 2.7751145691860644,
"retained_delta_l2": 2.343027160291451,
"delta_retention_norm_ratio": 0.8442992539146433,
"delta_cosine": 0.6596664167663548,
"relative_weight_rmse": 0.010816182538165395
},
{
"index": 24,
"total": 252,
"projection": "layers.11.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3777,
"parameters": 50331648,
"reference_delta_l2": 2.675125117521029,
"retained_delta_l2": 2.2732189655888804,
"delta_retention_norm_ratio": 0.8497617366381035,
"delta_cosine": 0.6633901224047348,
"relative_weight_rmse": 0.011374993862828652
},
{
"index": 25,
"total": 252,
"projection": "layers.11.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0303,
"parameters": 4194304,
"reference_delta_l2": 1.129744485973673,
"retained_delta_l2": 1.0869780166591216,
"delta_retention_norm_ratio": 0.9621450072600329,
"delta_cosine": 0.7872812397310865,
"relative_weight_rmse": 0.013666281685033765
},
{
"index": 26,
"total": 252,
"projection": "layers.11.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1274,
"parameters": 16777216,
"reference_delta_l2": 1.670846789488529,
"retained_delta_l2": 1.4674471513825977,
"delta_retention_norm_ratio": 0.8782655361427874,
"delta_cosine": 0.6931815425436411,
"relative_weight_rmse": 0.011730107850665127
},
{
"index": 27,
"total": 252,
"projection": "layers.11.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.126,
"parameters": 16777216,
"reference_delta_l2": 1.4076682377658807,
"retained_delta_l2": 1.1708727712056912,
"delta_retention_norm_ratio": 0.8317817649022123,
"delta_cosine": 0.6466330501407813,
"relative_weight_rmse": 0.010758078465344743
},
{
"index": 28,
"total": 252,
"projection": "layers.11.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0292,
"parameters": 4194304,
"reference_delta_l2": 0.9977954436103857,
"retained_delta_l2": 0.9075995021759654,
"delta_retention_norm_ratio": 0.9096047771995645,
"delta_cosine": 0.7270939597590178,
"relative_weight_rmse": 0.01188627268952539
},
{
"index": 29,
"total": 252,
"projection": "layers.12.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4254,
"parameters": 50331648,
"reference_delta_l2": 2.56819137012028,
"retained_delta_l2": 2.188648995444725,
"delta_retention_norm_ratio": 0.8522141382876078,
"delta_cosine": 0.6668181723119647,
"relative_weight_rmse": 0.01083587743189139
},
{
"index": 30,
"total": 252,
"projection": "layers.12.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3844,
"parameters": 50331648,
"reference_delta_l2": 3.215622881801083,
"retained_delta_l2": 2.8836738663452626,
"delta_retention_norm_ratio": 0.8967699174755547,
"delta_cosine": 0.7129079018317985,
"relative_weight_rmse": 0.01198643496827087
},
{
"index": 31,
"total": 252,
"projection": "layers.12.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3805,
"parameters": 50331648,
"reference_delta_l2": 2.6879330439167055,
"retained_delta_l2": 2.291630979320779,
"delta_retention_norm_ratio": 0.8525625236488567,
"delta_cosine": 0.6663007010584944,
"relative_weight_rmse": 0.01122343069789282
},
{
"index": 32,
"total": 252,
"projection": "layers.12.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 0.9326365058053449,
"retained_delta_l2": 0.8659338721920582,
"delta_retention_norm_ratio": 0.9284794952823683,
"delta_cosine": 0.7478119556266251,
"relative_weight_rmse": 0.01229225064365473
},
{
"index": 33,
"total": 252,
"projection": "layers.12.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1317,
"parameters": 16777216,
"reference_delta_l2": 1.7740352657937566,
"retained_delta_l2": 1.6069883409975243,
"delta_retention_norm_ratio": 0.9058378781881258,
"delta_cosine": 0.7219368480569792,
"relative_weight_rmse": 0.012034122226082595
},
{
"index": 34,
"total": 252,
"projection": "layers.12.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1301,
"parameters": 16777216,
"reference_delta_l2": 1.8069565591399073,
"retained_delta_l2": 1.6521463119058284,
"delta_retention_norm_ratio": 0.9143254183665782,
"delta_cosine": 0.7313478948888776,
"relative_weight_rmse": 0.012435789153691534
},
{
"index": 35,
"total": 252,
"projection": "layers.12.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0296,
"parameters": 4194304,
"reference_delta_l2": 1.0869529155925972,
"retained_delta_l2": 1.0238572118353657,
"delta_retention_norm_ratio": 0.9419517599593242,
"delta_cosine": 0.7659206070161154,
"relative_weight_rmse": 0.012192291258131012
},
{
"index": 36,
"total": 252,
"projection": "layers.13.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4309,
"parameters": 50331648,
"reference_delta_l2": 2.345717959028678,
"retained_delta_l2": 1.9211998853355021,
"delta_retention_norm_ratio": 0.8190242471141068,
"delta_cosine": 0.6358526600195066,
"relative_weight_rmse": 0.010194365616716888
},
{
"index": 37,
"total": 252,
"projection": "layers.13.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3975,
"parameters": 50331648,
"reference_delta_l2": 2.9930721247000602,
"retained_delta_l2": 2.6193255999400242,
"delta_retention_norm_ratio": 0.8751294625760181,
"delta_cosine": 0.6879834774883976,
"relative_weight_rmse": 0.01189494452769944
},
{
"index": 38,
"total": 252,
"projection": "layers.13.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4029,
"parameters": 50331648,
"reference_delta_l2": 2.4019665807886206,
"retained_delta_l2": 1.9420047828308697,
"delta_retention_norm_ratio": 0.8085061625600407,
"delta_cosine": 0.6251286647554096,
"relative_weight_rmse": 0.010350310001921261
},
{
"index": 39,
"total": 252,
"projection": "layers.13.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0328,
"parameters": 4194304,
"reference_delta_l2": 1.341017407179969,
"retained_delta_l2": 1.3266552896841128,
"delta_retention_norm_ratio": 0.9892901334323031,
"delta_cosine": 0.8343821488449628,
"relative_weight_rmse": 0.014778051436635804
},
{
"index": 40,
"total": 252,
"projection": "layers.13.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1345,
"parameters": 16777216,
"reference_delta_l2": 1.7312984063035275,
"retained_delta_l2": 1.5520233983981218,
"delta_retention_norm_ratio": 0.896450544139197,
"delta_cosine": 0.7121618169131606,
"relative_weight_rmse": 0.012267247947723397
},
{
"index": 41,
"total": 252,
"projection": "layers.13.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1332,
"parameters": 16777216,
"reference_delta_l2": 1.4095608507370043,
"retained_delta_l2": 1.1734218382297974,
"delta_retention_norm_ratio": 0.832473346302333,
"delta_cosine": 0.6470345038595087,
"relative_weight_rmse": 0.01074976588373817
},
{
"index": 42,
"total": 252,
"projection": "layers.13.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 1.02837419663425,
"retained_delta_l2": 0.9472396712269158,
"delta_retention_norm_ratio": 0.9211040828592569,
"delta_cosine": 0.7384602764192114,
"relative_weight_rmse": 0.012317781541514026
},
{
"index": 43,
"total": 252,
"projection": "layers.14.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4305,
"parameters": 50331648,
"reference_delta_l2": 2.4695213210201166,
"retained_delta_l2": 2.088833734616667,
"delta_retention_norm_ratio": 0.8458455964064348,
"delta_cosine": 0.6615175643703667,
"relative_weight_rmse": 0.010523678881525165
},
{
"index": 44,
"total": 252,
"projection": "layers.14.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3935,
"parameters": 50331648,
"reference_delta_l2": 2.73564026062679,
"retained_delta_l2": 2.3210056632930196,
"delta_retention_norm_ratio": 0.8484323383810819,
"delta_cosine": 0.6617345073961669,
"relative_weight_rmse": 0.011367344684873312
},
{
"index": 45,
"total": 252,
"projection": "layers.14.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3944,
"parameters": 50331648,
"reference_delta_l2": 2.9456908148003613,
"retained_delta_l2": 2.611415626037734,
"delta_retention_norm_ratio": 0.8865206127258531,
"delta_cosine": 0.7009710931147657,
"relative_weight_rmse": 0.011786178342384171
},
{
"index": 46,
"total": 252,
"projection": "layers.14.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 0.7290877266076532,
"retained_delta_l2": 0.6191256185610067,
"delta_retention_norm_ratio": 0.8491784952158153,
"delta_cosine": 0.666182717165303,
"relative_weight_rmse": 0.01072129480320891
},
{
"index": 47,
"total": 252,
"projection": "layers.14.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1339,
"parameters": 16777216,
"reference_delta_l2": 1.5562308772486637,
"retained_delta_l2": 1.3548889172329652,
"delta_retention_norm_ratio": 0.8706220503916099,
"delta_cosine": 0.6849544987072025,
"relative_weight_rmse": 0.011444515120076961
},
{
"index": 48,
"total": 252,
"projection": "layers.14.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1336,
"parameters": 16777216,
"reference_delta_l2": 2.3181710795261905,
"retained_delta_l2": 2.259554226479229,
"delta_retention_norm_ratio": 0.9747141815525789,
"delta_cosine": 0.8054773380596841,
"relative_weight_rmse": 0.013913795311187202
},
{
"index": 49,
"total": 252,
"projection": "layers.14.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.03,
"parameters": 4194304,
"reference_delta_l2": 1.0169140246366664,
"retained_delta_l2": 0.9438884227724618,
"delta_retention_norm_ratio": 0.928189010973375,
"delta_cosine": 0.7456435016623865,
"relative_weight_rmse": 0.012379671180818227
},
{
"index": 50,
"total": 252,
"projection": "layers.15.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4343,
"parameters": 50331648,
"reference_delta_l2": 2.37291388097089,
"retained_delta_l2": 1.9747544435249724,
"delta_retention_norm_ratio": 0.8322065370181033,
"delta_cosine": 0.6483283407698353,
"relative_weight_rmse": 0.010305350636473356
},
{
"index": 51,
"total": 252,
"projection": "layers.15.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3924,
"parameters": 50331648,
"reference_delta_l2": 2.877735534294013,
"retained_delta_l2": 2.5181948245020824,
"delta_retention_norm_ratio": 0.875061239816071,
"delta_cosine": 0.6869561824692126,
"relative_weight_rmse": 0.01203124922287954
},
{
"index": 52,
"total": 252,
"projection": "layers.15.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3941,
"parameters": 50331648,
"reference_delta_l2": 2.7887238690235363,
"retained_delta_l2": 2.4400914015325093,
"delta_retention_norm_ratio": 0.8749849451343852,
"delta_cosine": 0.6893492075249347,
"relative_weight_rmse": 0.011424688533661688
},
{
"index": 53,
"total": 252,
"projection": "layers.15.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0317,
"parameters": 4194304,
"reference_delta_l2": 0.8136634946472183,
"retained_delta_l2": 0.7194351998141398,
"delta_retention_norm_ratio": 0.884192549557685,
"delta_cosine": 0.6989657732038983,
"relative_weight_rmse": 0.01158525789051138
},
{
"index": 54,
"total": 252,
"projection": "layers.15.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1354,
"parameters": 16777216,
"reference_delta_l2": 2.14888279095195,
"retained_delta_l2": 2.0648059108914714,
"delta_retention_norm_ratio": 0.9608741433388125,
"delta_cosine": 0.7884679794096899,
"relative_weight_rmse": 0.01383434690156249
},
{
"index": 55,
"total": 252,
"projection": "layers.15.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1339,
"parameters": 16777216,
"reference_delta_l2": 2.2363308934404,
"retained_delta_l2": 2.1612427983529505,
"delta_retention_norm_ratio": 0.9664235309239354,
"delta_cosine": 0.8010344733340968,
"relative_weight_rmse": 0.013448329113463752
},
{
"index": 56,
"total": 252,
"projection": "layers.15.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0318,
"parameters": 4194304,
"reference_delta_l2": 0.8579609120298205,
"retained_delta_l2": 0.7628252507622412,
"delta_retention_norm_ratio": 0.8891142242803333,
"delta_cosine": 0.7053941734827593,
"relative_weight_rmse": 0.011222617076479648
},
{
"index": 57,
"total": 252,
"projection": "layers.16.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4375,
"parameters": 50331648,
"reference_delta_l2": 2.4925761772926727,
"retained_delta_l2": 2.1300505488925414,
"delta_retention_norm_ratio": 0.85455785395739,
"delta_cosine": 0.6692738300226725,
"relative_weight_rmse": 0.010669613300324836
},
{
"index": 58,
"total": 252,
"projection": "layers.16.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3967,
"parameters": 50331648,
"reference_delta_l2": 2.639823835206382,
"retained_delta_l2": 2.2568373816548224,
"delta_retention_norm_ratio": 0.85491969257804,
"delta_cosine": 0.6684142838746178,
"relative_weight_rmse": 0.011071951086947165
},
{
"index": 59,
"total": 252,
"projection": "layers.16.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3994,
"parameters": 50331648,
"reference_delta_l2": 2.5574051385465872,
"retained_delta_l2": 2.155985075223233,
"delta_retention_norm_ratio": 0.8430361864559763,
"delta_cosine": 0.657171055105935,
"relative_weight_rmse": 0.010889774960224613
},
{
"index": 60,
"total": 252,
"projection": "layers.16.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0311,
"parameters": 4194304,
"reference_delta_l2": 0.7403652978861993,
"retained_delta_l2": 0.6378164203841092,
"delta_retention_norm_ratio": 0.8614888112734684,
"delta_cosine": 0.6754536654041666,
"relative_weight_rmse": 0.011217294766259881
},
{
"index": 61,
"total": 252,
"projection": "layers.16.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1357,
"parameters": 16777216,
"reference_delta_l2": 1.6664737578904119,
"retained_delta_l2": 1.4904171603264238,
"delta_retention_norm_ratio": 0.8943538134156651,
"delta_cosine": 0.709376769947244,
"relative_weight_rmse": 0.011938452222249625
},
{
"index": 62,
"total": 252,
"projection": "layers.16.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.134,
"parameters": 16777216,
"reference_delta_l2": 1.810967498456618,
"retained_delta_l2": 1.6792081046389524,
"delta_retention_norm_ratio": 0.9272436452172905,
"delta_cosine": 0.7481180718262056,
"relative_weight_rmse": 0.012466719594668663
},
{
"index": 63,
"total": 252,
"projection": "layers.16.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0308,
"parameters": 4194304,
"reference_delta_l2": 0.9588623100041671,
"retained_delta_l2": 0.8756644620899365,
"delta_retention_norm_ratio": 0.9132327477613871,
"delta_cosine": 0.7301267251224983,
"relative_weight_rmse": 0.012151192168392035
},
{
"index": 64,
"total": 252,
"projection": "layers.17.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4348,
"parameters": 50331648,
"reference_delta_l2": 2.5816971370993724,
"retained_delta_l2": 2.2212677742122406,
"delta_retention_norm_ratio": 0.8603905323720168,
"delta_cosine": 0.6749114085537792,
"relative_weight_rmse": 0.010931229419799657
},
{
"index": 65,
"total": 252,
"projection": "layers.17.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3935,
"parameters": 50331648,
"reference_delta_l2": 3.537919615556736,
"retained_delta_l2": 3.3171539258685923,
"delta_retention_norm_ratio": 0.9376001397212629,
"delta_cosine": 0.7563796909416031,
"relative_weight_rmse": 0.013295309785955068
},
{
"index": 66,
"total": 252,
"projection": "layers.17.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3948,
"parameters": 50331648,
"reference_delta_l2": 3.184727757412606,
"retained_delta_l2": 2.900798539966914,
"delta_retention_norm_ratio": 0.9108466283233053,
"delta_cosine": 0.7254635730513106,
"relative_weight_rmse": 0.012378258470841225
},
{
"index": 67,
"total": 252,
"projection": "layers.17.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0327,
"parameters": 4194304,
"reference_delta_l2": 0.7873391801685035,
"retained_delta_l2": 0.7046799353763415,
"delta_retention_norm_ratio": 0.8950144399336618,
"delta_cosine": 0.7147739387327632,
"relative_weight_rmse": 0.011557893189828878
},
{
"index": 68,
"total": 252,
"projection": "layers.17.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1355,
"parameters": 16777216,
"reference_delta_l2": 1.7297913669458336,
"retained_delta_l2": 1.5698438793182137,
"delta_retention_norm_ratio": 0.9075336536625064,
"delta_cosine": 0.7247548434650827,
"relative_weight_rmse": 0.01219956858762891
},
{
"index": 69,
"total": 252,
"projection": "layers.17.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1356,
"parameters": 16777216,
"reference_delta_l2": 1.6613043369596119,
"retained_delta_l2": 1.486608854445748,
"delta_retention_norm_ratio": 0.8948443830384638,
"delta_cosine": 0.7077833061431172,
"relative_weight_rmse": 0.011971929027731735
},
{
"index": 70,
"total": 252,
"projection": "layers.17.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 1.033477168627319,
"retained_delta_l2": 0.9691635128398266,
"delta_retention_norm_ratio": 0.9377696404528078,
"delta_cosine": 0.7589014017524195,
"relative_weight_rmse": 0.01276351600474143
},
{
"index": 71,
"total": 252,
"projection": "layers.18.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4376,
"parameters": 50331648,
"reference_delta_l2": 2.912291330744063,
"retained_delta_l2": 2.624113222358469,
"delta_retention_norm_ratio": 0.9010476371840288,
"delta_cosine": 0.7155707881308109,
"relative_weight_rmse": 0.011719327332632453
},
{
"index": 72,
"total": 252,
"projection": "layers.18.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3934,
"parameters": 50331648,
"reference_delta_l2": 4.7323895908701195,
"retained_delta_l2": 4.694887208565627,
"delta_retention_norm_ratio": 0.9920753814570036,
"delta_cosine": 0.8274217190394803,
"relative_weight_rmse": 0.01543821079242807
},
{
"index": 73,
"total": 252,
"projection": "layers.18.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.395,
"parameters": 50331648,
"reference_delta_l2": 3.2684049511531974,
"retained_delta_l2": 3.007384330921114,
"delta_retention_norm_ratio": 0.920138225179231,
"delta_cosine": 0.7367400566488568,
"relative_weight_rmse": 0.01248984104763899
},
{
"index": 74,
"total": 252,
"projection": "layers.18.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 1.1176756575337732,
"retained_delta_l2": 1.0843334020518682,
"delta_retention_norm_ratio": 0.9701682189666035,
"delta_cosine": 0.7990453769658297,
"relative_weight_rmse": 0.013997563751336657
},
{
"index": 75,
"total": 252,
"projection": "layers.18.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1349,
"parameters": 16777216,
"reference_delta_l2": 1.6285698839730298,
"retained_delta_l2": 1.4555649563408954,
"delta_retention_norm_ratio": 0.8937688033318689,
"delta_cosine": 0.7107610372529769,
"relative_weight_rmse": 0.01169664296587917
},
{
"index": 76,
"total": 252,
"projection": "layers.18.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1344,
"parameters": 16777216,
"reference_delta_l2": 2.099547953488129,
"retained_delta_l2": 2.009462095351522,
"delta_retention_norm_ratio": 0.9570927360878131,
"delta_cosine": 0.7836973547458617,
"relative_weight_rmse": 0.013543508405450629
},
{
"index": 77,
"total": 252,
"projection": "layers.18.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 0.9280061897807907,
"retained_delta_l2": 0.8453754017599368,
"delta_retention_norm_ratio": 0.9109587964705573,
"delta_cosine": 0.7279125852473827,
"relative_weight_rmse": 0.011785830425009313
},
{
"index": 78,
"total": 252,
"projection": "layers.19.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4367,
"parameters": 50331648,
"reference_delta_l2": 2.6905658774983516,
"retained_delta_l2": 2.353839107682188,
"delta_retention_norm_ratio": 0.8748490893189922,
"delta_cosine": 0.6890581511290075,
"relative_weight_rmse": 0.011200443943313133
},
{
"index": 79,
"total": 252,
"projection": "layers.19.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3929,
"parameters": 50331648,
"reference_delta_l2": 4.618692339064488,
"retained_delta_l2": 4.552187654151666,
"delta_retention_norm_ratio": 0.9856009710042102,
"delta_cosine": 0.8186987969035592,
"relative_weight_rmse": 0.01545184839911245
},
{
"index": 80,
"total": 252,
"projection": "layers.19.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3936,
"parameters": 50331648,
"reference_delta_l2": 3.091580020506551,
"retained_delta_l2": 2.7719047410257023,
"delta_retention_norm_ratio": 0.8965980898568265,
"delta_cosine": 0.7108192278959599,
"relative_weight_rmse": 0.012195071398963545
},
{
"index": 81,
"total": 252,
"projection": "layers.19.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 1.0441969278779464,
"retained_delta_l2": 1.0049150954003139,
"delta_retention_norm_ratio": 0.9623808197200289,
"delta_cosine": 0.7951310937727945,
"relative_weight_rmse": 0.013587193054594958
},
{
"index": 82,
"total": 252,
"projection": "layers.19.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1349,
"parameters": 16777216,
"reference_delta_l2": 1.6398227134396683,
"retained_delta_l2": 1.4603600255517262,
"delta_retention_norm_ratio": 0.890559701108479,
"delta_cosine": 0.705257501501427,
"relative_weight_rmse": 0.011793715018796351
},
{
"index": 83,
"total": 252,
"projection": "layers.19.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1347,
"parameters": 16777216,
"reference_delta_l2": 1.933861545575777,
"retained_delta_l2": 1.8238676760994077,
"delta_retention_norm_ratio": 0.9431221590149462,
"delta_cosine": 0.7653721358101522,
"relative_weight_rmse": 0.012813601727524927
},
{
"index": 84,
"total": 252,
"projection": "layers.19.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0303,
"parameters": 4194304,
"reference_delta_l2": 1.2392430066984266,
"retained_delta_l2": 1.2062007950941982,
"delta_retention_norm_ratio": 0.9733367778348342,
"delta_cosine": 0.806179559805123,
"relative_weight_rmse": 0.013897893803669462
},
{
"index": 85,
"total": 252,
"projection": "layers.2.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.438,
"parameters": 50331648,
"reference_delta_l2": 1.5500714165419396,
"retained_delta_l2": 1.3349876126090936,
"delta_retention_norm_ratio": 0.8612426487982874,
"delta_cosine": 0.697999976458453,
"relative_weight_rmse": 0.01010100203213799
},
{
"index": 86,
"total": 252,
"projection": "layers.2.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4065,
"parameters": 50331648,
"reference_delta_l2": 2.504646258958508,
"retained_delta_l2": 2.3609978273546726,
"delta_retention_norm_ratio": 0.9426472177098703,
"delta_cosine": 0.7970291686137979,
"relative_weight_rmse": 0.012648342970186247
},
{
"index": 87,
"total": 252,
"projection": "layers.2.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3935,
"parameters": 50331648,
"reference_delta_l2": 3.3696103191055706,
"retained_delta_l2": 3.349766906459199,
"delta_retention_norm_ratio": 0.9941110660381528,
"delta_cosine": 0.8793504353476361,
"relative_weight_rmse": 0.014958809364992352
},
{
"index": 88,
"total": 252,
"projection": "layers.2.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0315,
"parameters": 4194304,
"reference_delta_l2": 0.5769582349477705,
"retained_delta_l2": 0.43406403930005427,
"delta_retention_norm_ratio": 0.7523318205855024,
"delta_cosine": 0.5760685043835092,
"relative_weight_rmse": 0.00928746344440853
},
{
"index": 89,
"total": 252,
"projection": "layers.2.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1349,
"parameters": 16777216,
"reference_delta_l2": 2.083701722857339,
"retained_delta_l2": 1.9650056104568143,
"delta_retention_norm_ratio": 0.9430359388301705,
"delta_cosine": 0.7602686010254693,
"relative_weight_rmse": 0.013939342093591264
},
{
"index": 90,
"total": 252,
"projection": "layers.2.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1336,
"parameters": 16777216,
"reference_delta_l2": 1.2288907212918998,
"retained_delta_l2": 0.9707183587780293,
"delta_retention_norm_ratio": 0.7899143039810238,
"delta_cosine": 0.608056427978517,
"relative_weight_rmse": 0.010180997687138508
},
{
"index": 91,
"total": 252,
"projection": "layers.2.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 1.1606500649521356,
"retained_delta_l2": 1.119722100274245,
"delta_retention_norm_ratio": 0.9647370332249295,
"delta_cosine": 0.7932427944826372,
"relative_weight_rmse": 0.013508958069442129
},
{
"index": 92,
"total": 252,
"projection": "layers.20.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4371,
"parameters": 50331648,
"reference_delta_l2": 2.647556093950071,
"retained_delta_l2": 2.2599685058612993,
"delta_retention_norm_ratio": 0.8536055236093212,
"delta_cosine": 0.6660049619390129,
"relative_weight_rmse": 0.011252847750215126
},
{
"index": 93,
"total": 252,
"projection": "layers.20.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3929,
"parameters": 50331648,
"reference_delta_l2": 4.54164391219833,
"retained_delta_l2": 4.472502244452568,
"delta_retention_norm_ratio": 0.9847760702770961,
"delta_cosine": 0.8167110275761873,
"relative_weight_rmse": 0.015413764053976025
},
{
"index": 94,
"total": 252,
"projection": "layers.20.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3911,
"parameters": 50331648,
"reference_delta_l2": 3.1240951518173383,
"retained_delta_l2": 2.8064881224754443,
"delta_retention_norm_ratio": 0.898336313746034,
"delta_cosine": 0.7121239647325682,
"relative_weight_rmse": 0.01227175571033149
},
{
"index": 95,
"total": 252,
"projection": "layers.20.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 0.8435434130294448,
"retained_delta_l2": 0.7755878258928495,
"delta_retention_norm_ratio": 0.9194403203356847,
"delta_cosine": 0.7399829493130454,
"relative_weight_rmse": 0.01219344130971694
},
{
"index": 96,
"total": 252,
"projection": "layers.20.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1345,
"parameters": 16777216,
"reference_delta_l2": 1.60998213071129,
"retained_delta_l2": 1.4290075961452402,
"delta_retention_norm_ratio": 0.887592209184275,
"delta_cosine": 0.703519942844837,
"relative_weight_rmse": 0.011826373126148216
},
{
"index": 97,
"total": 252,
"projection": "layers.20.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1354,
"parameters": 16777216,
"reference_delta_l2": 1.893648789853392,
"retained_delta_l2": 1.7698240053263536,
"delta_retention_norm_ratio": 0.9346104804700216,
"delta_cosine": 0.757096633151059,
"relative_weight_rmse": 0.012342558871086262
},
{
"index": 98,
"total": 252,
"projection": "layers.20.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 0.9448415236568111,
"retained_delta_l2": 0.8670623666086616,
"delta_retention_norm_ratio": 0.9176802086903193,
"delta_cosine": 0.7371462243551566,
"relative_weight_rmse": 0.012026095186860188
},
{
"index": 99,
"total": 252,
"projection": "layers.21.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4349,
"parameters": 50331648,
"reference_delta_l2": 3.105276205896329,
"retained_delta_l2": 2.8032148298004578,
"delta_retention_norm_ratio": 0.9027264062622469,
"delta_cosine": 0.7157920387866503,
"relative_weight_rmse": 0.012305052366142202
},
{
"index": 100,
"total": 252,
"projection": "layers.21.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3908,
"parameters": 50331648,
"reference_delta_l2": 4.619770502917215,
"retained_delta_l2": 4.556478246695928,
"delta_retention_norm_ratio": 0.986299696882925,
"delta_cosine": 0.8167262470911177,
"relative_weight_rmse": 0.01590804096656067
},
{
"index": 101,
"total": 252,
"projection": "layers.21.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3915,
"parameters": 50331648,
"reference_delta_l2": 3.3382291691770045,
"retained_delta_l2": 3.0642641788203435,
"delta_retention_norm_ratio": 0.9179310417372564,
"delta_cosine": 0.7328829388695662,
"relative_weight_rmse": 0.012751954415997446
},
{
"index": 102,
"total": 252,
"projection": "layers.21.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.031,
"parameters": 4194304,
"reference_delta_l2": 0.889886028991794,
"retained_delta_l2": 0.8231915337286541,
"delta_retention_norm_ratio": 0.9250527673316749,
"delta_cosine": 0.7423830456244178,
"relative_weight_rmse": 0.012826996818041645
},
{
"index": 103,
"total": 252,
"projection": "layers.21.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1337,
"parameters": 16777216,
"reference_delta_l2": 1.7388208776579452,
"retained_delta_l2": 1.590756945940496,
"delta_retention_norm_ratio": 0.9148480826174115,
"delta_cosine": 0.7325991137758434,
"relative_weight_rmse": 0.012392239183371758
},
{
"index": 104,
"total": 252,
"projection": "layers.21.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1335,
"parameters": 16777216,
"reference_delta_l2": 1.346426603628278,
"retained_delta_l2": 1.1134869039593562,
"delta_retention_norm_ratio": 0.8269941346663766,
"delta_cosine": 0.6436933610520533,
"relative_weight_rmse": 0.010330564164912478
},
{
"index": 105,
"total": 252,
"projection": "layers.21.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0309,
"parameters": 4194304,
"reference_delta_l2": 1.341526986386814,
"retained_delta_l2": 1.3248957647642399,
"delta_retention_norm_ratio": 0.9876027677479917,
"delta_cosine": 0.8234711888276329,
"relative_weight_rmse": 0.014563529675619849
},
{
"index": 106,
"total": 252,
"projection": "layers.22.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4313,
"parameters": 50331648,
"reference_delta_l2": 3.360717108118368,
"retained_delta_l2": 3.125093900153933,
"delta_retention_norm_ratio": 0.9298890086894703,
"delta_cosine": 0.7474962910334413,
"relative_weight_rmse": 0.01260631904480232
},
{
"index": 107,
"total": 252,
"projection": "layers.22.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.389,
"parameters": 50331648,
"reference_delta_l2": 4.210747269890564,
"retained_delta_l2": 4.096031547804069,
"delta_retention_norm_ratio": 0.9727564456535345,
"delta_cosine": 0.8019330786138767,
"relative_weight_rmse": 0.014651000649681535
},
{
"index": 108,
"total": 252,
"projection": "layers.22.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3913,
"parameters": 50331648,
"reference_delta_l2": 3.4252916600455015,
"retained_delta_l2": 3.174494875191197,
"delta_retention_norm_ratio": 0.9267808964183293,
"delta_cosine": 0.7449150154482405,
"relative_weight_rmse": 0.012618399403942023
},
{
"index": 109,
"total": 252,
"projection": "layers.22.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 1.1256697925354247,
"retained_delta_l2": 1.1006366382431394,
"delta_retention_norm_ratio": 0.9777615474286636,
"delta_cosine": 0.817683777268333,
"relative_weight_rmse": 0.014420613901313934
},
{
"index": 110,
"total": 252,
"projection": "layers.22.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1345,
"parameters": 16777216,
"reference_delta_l2": 2.099753726538612,
"retained_delta_l2": 1.999529306975156,
"delta_retention_norm_ratio": 0.9522684883009239,
"delta_cosine": 0.7730058820038571,
"relative_weight_rmse": 0.01373655270061581
},
{
"index": 111,
"total": 252,
"projection": "layers.22.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1346,
"parameters": 16777216,
"reference_delta_l2": 1.8696581267064718,
"retained_delta_l2": 1.740177726303158,
"delta_retention_norm_ratio": 0.9307464832453608,
"delta_cosine": 0.7545627126669409,
"relative_weight_rmse": 0.01259379142506784
},
{
"index": 112,
"total": 252,
"projection": "layers.22.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 1.10357555374315,
"retained_delta_l2": 1.0512816270257077,
"delta_retention_norm_ratio": 0.9526140946670395,
"delta_cosine": 0.7775259204534436,
"relative_weight_rmse": 0.013222044058572085
},
{
"index": 113,
"total": 252,
"projection": "layers.23.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4311,
"parameters": 50331648,
"reference_delta_l2": 3.160872318390849,
"retained_delta_l2": 2.9013703143071194,
"delta_retention_norm_ratio": 0.9179017758566602,
"delta_cosine": 0.73519129883951,
"relative_weight_rmse": 0.011974311913947348
},
{
"index": 114,
"total": 252,
"projection": "layers.23.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.39,
"parameters": 50331648,
"reference_delta_l2": 5.633661482659359,
"retained_delta_l2": 5.677116194607454,
"delta_retention_norm_ratio": 1.0077134048756482,
"delta_cosine": 0.8533048383623517,
"relative_weight_rmse": 0.016809090155428346
},
{
"index": 115,
"total": 252,
"projection": "layers.23.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3924,
"parameters": 50331648,
"reference_delta_l2": 4.980404555262815,
"retained_delta_l2": 4.942861482350325,
"delta_retention_norm_ratio": 0.992461842708577,
"delta_cosine": 0.828900774170685,
"relative_weight_rmse": 0.015409840480461122
},
{
"index": 116,
"total": 252,
"projection": "layers.23.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0316,
"parameters": 4194304,
"reference_delta_l2": 0.7947115673087248,
"retained_delta_l2": 0.7073865528627679,
"delta_retention_norm_ratio": 0.8901173481824589,
"delta_cosine": 0.7057249262457971,
"relative_weight_rmse": 0.011975734018331215
},
{
"index": 117,
"total": 252,
"projection": "layers.23.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1365,
"parameters": 16777216,
"reference_delta_l2": 1.961022080361704,
"retained_delta_l2": 1.8342382011607155,
"delta_retention_norm_ratio": 0.9353480613652226,
"delta_cosine": 0.7548158970571475,
"relative_weight_rmse": 0.012808345048691697
},
{
"index": 118,
"total": 252,
"projection": "layers.23.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1343,
"parameters": 16777216,
"reference_delta_l2": 1.7869979322140728,
"retained_delta_l2": 1.6195382621873493,
"delta_retention_norm_ratio": 0.906289947510324,
"delta_cosine": 0.7240350295402113,
"relative_weight_rmse": 0.012030264827546133
},
{
"index": 119,
"total": 252,
"projection": "layers.23.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0301,
"parameters": 4194304,
"reference_delta_l2": 0.9269590107602388,
"retained_delta_l2": 0.8512280843157533,
"delta_retention_norm_ratio": 0.9183017527577888,
"delta_cosine": 0.7376277684845678,
"relative_weight_rmse": 0.011725215372115768
},
{
"index": 120,
"total": 252,
"projection": "layers.24.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4305,
"parameters": 50331648,
"reference_delta_l2": 3.2503102452109283,
"retained_delta_l2": 2.978091014440522,
"delta_retention_norm_ratio": 0.9162482316352727,
"delta_cosine": 0.7315355094947241,
"relative_weight_rmse": 0.01227254580627398
},
{
"index": 121,
"total": 252,
"projection": "layers.24.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3947,
"parameters": 50331648,
"reference_delta_l2": 5.122346784643699,
"retained_delta_l2": 5.090735605570083,
"delta_retention_norm_ratio": 0.9938287702097051,
"delta_cosine": 0.8272340568179678,
"relative_weight_rmse": 0.01624333933890823
},
{
"index": 122,
"total": 252,
"projection": "layers.24.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3933,
"parameters": 50331648,
"reference_delta_l2": 4.007395782761482,
"retained_delta_l2": 3.8291209124834666,
"delta_retention_norm_ratio": 0.9555135354873368,
"delta_cosine": 0.7755254983957733,
"relative_weight_rmse": 0.013944425880539327
},
{
"index": 123,
"total": 252,
"projection": "layers.24.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 0.9660765277735609,
"retained_delta_l2": 0.9167020992533299,
"delta_retention_norm_ratio": 0.9488918040126486,
"delta_cosine": 0.7724223783247411,
"relative_weight_rmse": 0.013150187824250323
},
{
"index": 124,
"total": 252,
"projection": "layers.24.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.135,
"parameters": 16777216,
"reference_delta_l2": 1.918328876575902,
"retained_delta_l2": 1.779257488401996,
"delta_retention_norm_ratio": 0.9275038863919206,
"delta_cosine": 0.7461881577671706,
"relative_weight_rmse": 0.012580540298291065
},
{
"index": 125,
"total": 252,
"projection": "layers.24.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1348,
"parameters": 16777216,
"reference_delta_l2": 3.0209953608095272,
"retained_delta_l2": 3.0175153247508884,
"delta_retention_norm_ratio": 0.9988480498501309,
"delta_cosine": 0.8487291450530009,
"relative_weight_rmse": 0.015555315896263225
},
{
"index": 126,
"total": 252,
"projection": "layers.24.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 1.0885553070106562,
"retained_delta_l2": 1.0372228366876877,
"delta_retention_norm_ratio": 0.9528434889873116,
"delta_cosine": 0.7791866473600985,
"relative_weight_rmse": 0.013054892107370254
},
{
"index": 127,
"total": 252,
"projection": "layers.25.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4331,
"parameters": 50331648,
"reference_delta_l2": 3.6373147216160677,
"retained_delta_l2": 3.404055364663721,
"delta_retention_norm_ratio": 0.9358704498221949,
"delta_cosine": 0.7514459847745487,
"relative_weight_rmse": 0.013263712821630426
},
{
"index": 128,
"total": 252,
"projection": "layers.25.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3927,
"parameters": 50331648,
"reference_delta_l2": 5.600910492578175,
"retained_delta_l2": 5.614913656800555,
"delta_retention_norm_ratio": 1.0025001585440323,
"delta_cosine": 0.8406864774637491,
"relative_weight_rmse": 0.016952561312630426
},
{
"index": 129,
"total": 252,
"projection": "layers.25.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3946,
"parameters": 50331648,
"reference_delta_l2": 3.745221022962041,
"retained_delta_l2": 3.49549093622834,
"delta_retention_norm_ratio": 0.9333203340463487,
"delta_cosine": 0.7482711929833467,
"relative_weight_rmse": 0.013621479587259277
},
{
"index": 130,
"total": 252,
"projection": "layers.25.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 1.0289437223960307,
"retained_delta_l2": 0.982520413562484,
"delta_retention_norm_ratio": 0.9548825578862138,
"delta_cosine": 0.7800469586103691,
"relative_weight_rmse": 0.013545644703222918
},
{
"index": 131,
"total": 252,
"projection": "layers.25.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1346,
"parameters": 16777216,
"reference_delta_l2": 1.9553584720962225,
"retained_delta_l2": 1.8082254096930057,
"delta_retention_norm_ratio": 0.9247539187812022,
"delta_cosine": 0.7401622947458671,
"relative_weight_rmse": 0.013195673526209827
},
{
"index": 132,
"total": 252,
"projection": "layers.25.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1314,
"parameters": 16777216,
"reference_delta_l2": 2.450296786092303,
"retained_delta_l2": 2.402918629425209,
"delta_retention_norm_ratio": 0.980664319140437,
"delta_cosine": 0.8152132178531424,
"relative_weight_rmse": 0.014595703215884949
},
{
"index": 133,
"total": 252,
"projection": "layers.25.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0302,
"parameters": 4194304,
"reference_delta_l2": 1.1880491684330374,
"retained_delta_l2": 1.1446621680107196,
"delta_retention_norm_ratio": 0.9634804673281809,
"delta_cosine": 0.7872418634950002,
"relative_weight_rmse": 0.01345733935618882
},
{
"index": 134,
"total": 252,
"projection": "layers.26.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4608,
"parameters": 50331648,
"reference_delta_l2": 3.2957611727198537,
"retained_delta_l2": 3.016838300593981,
"delta_retention_norm_ratio": 0.9153692098703593,
"delta_cosine": 0.7283018254489741,
"relative_weight_rmse": 0.012337697827002952
},
{
"index": 135,
"total": 252,
"projection": "layers.26.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4003,
"parameters": 50331648,
"reference_delta_l2": 5.260130561195044,
"retained_delta_l2": 5.222515082979985,
"delta_retention_norm_ratio": 0.9928489459002109,
"delta_cosine": 0.8221569371622288,
"relative_weight_rmse": 0.016656073965757832
},
{
"index": 136,
"total": 252,
"projection": "layers.26.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4084,
"parameters": 50331648,
"reference_delta_l2": 3.598925250679574,
"retained_delta_l2": 3.3267956511150807,
"delta_retention_norm_ratio": 0.9243858706114811,
"delta_cosine": 0.7381639234093447,
"relative_weight_rmse": 0.013167427550091431
},
{
"index": 137,
"total": 252,
"projection": "layers.26.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0321,
"parameters": 4194304,
"reference_delta_l2": 1.102208846410291,
"retained_delta_l2": 1.0639393919165272,
"delta_retention_norm_ratio": 0.9652793074394194,
"delta_cosine": 0.7913675307064206,
"relative_weight_rmse": 0.014438720269170098
},
{
"index": 138,
"total": 252,
"projection": "layers.26.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1349,
"parameters": 16777216,
"reference_delta_l2": 2.145429788307753,
"retained_delta_l2": 2.021116465396612,
"delta_retention_norm_ratio": 0.9420566808624414,
"delta_cosine": 0.7604249328209329,
"relative_weight_rmse": 0.013810083511670299
},
{
"index": 139,
"total": 252,
"projection": "layers.26.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1332,
"parameters": 16777216,
"reference_delta_l2": 2.7116328068071667,
"retained_delta_l2": 2.6872061495663733,
"delta_retention_norm_ratio": 0.9909919008283593,
"delta_cosine": 0.8318779690607394,
"relative_weight_rmse": 0.015357608678216816
},
{
"index": 140,
"total": 252,
"projection": "layers.26.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0303,
"parameters": 4194304,
"reference_delta_l2": 1.300912100078543,
"retained_delta_l2": 1.2655509068920476,
"delta_retention_norm_ratio": 0.9728181533676561,
"delta_cosine": 0.8069024380143299,
"relative_weight_rmse": 0.014256978492757717
},
{
"index": 141,
"total": 252,
"projection": "layers.27.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4441,
"parameters": 50331648,
"reference_delta_l2": 3.1221671631429038,
"retained_delta_l2": 2.7953025124195756,
"delta_retention_norm_ratio": 0.8953084080244145,
"delta_cosine": 0.7077534475689679,
"relative_weight_rmse": 0.011909876215115344
},
{
"index": 142,
"total": 252,
"projection": "layers.27.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3914,
"parameters": 50331648,
"reference_delta_l2": 5.226479588987116,
"retained_delta_l2": 5.1933838720565655,
"delta_retention_norm_ratio": 0.9936676846494746,
"delta_cosine": 0.8248796468505778,
"relative_weight_rmse": 0.01645056821096251
},
{
"index": 143,
"total": 252,
"projection": "layers.27.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3904,
"parameters": 50331648,
"reference_delta_l2": 3.1439875550600127,
"retained_delta_l2": 2.7609244012810183,
"delta_retention_norm_ratio": 0.8781600922171328,
"delta_cosine": 0.6898881540050043,
"relative_weight_rmse": 0.012184674639886598
},
{
"index": 144,
"total": 252,
"projection": "layers.27.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.032,
"parameters": 4194304,
"reference_delta_l2": 0.9578251695993613,
"retained_delta_l2": 0.9016073643405902,
"delta_retention_norm_ratio": 0.9413068198215278,
"delta_cosine": 0.7665743632754864,
"relative_weight_rmse": 0.013498560985661111
},
{
"index": 145,
"total": 252,
"projection": "layers.27.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1343,
"parameters": 16777216,
"reference_delta_l2": 1.834174655123338,
"retained_delta_l2": 1.6691012027622767,
"delta_retention_norm_ratio": 0.9100012357602002,
"delta_cosine": 0.7239421816255456,
"relative_weight_rmse": 0.012743272136643499
},
{
"index": 146,
"total": 252,
"projection": "layers.27.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1336,
"parameters": 16777216,
"reference_delta_l2": 2.5051869458196565,
"retained_delta_l2": 2.450649274322715,
"delta_retention_norm_ratio": 0.9782300991197695,
"delta_cosine": 0.8133656660897193,
"relative_weight_rmse": 0.014896108715641143
},
{
"index": 147,
"total": 252,
"projection": "layers.27.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 1.305157478619902,
"retained_delta_l2": 1.2820106051062443,
"delta_retention_norm_ratio": 0.9822650723051953,
"delta_cosine": 0.814217592620622,
"relative_weight_rmse": 0.014311472366894704
},
{
"index": 148,
"total": 252,
"projection": "layers.28.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4307,
"parameters": 50331648,
"reference_delta_l2": 3.7617323107123806,
"retained_delta_l2": 3.5468414125548615,
"delta_retention_norm_ratio": 0.9428744843045932,
"delta_cosine": 0.7610179061584897,
"relative_weight_rmse": 0.013099953206764646
},
{
"index": 149,
"total": 252,
"projection": "layers.28.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3892,
"parameters": 50331648,
"reference_delta_l2": 6.423387439542335,
"retained_delta_l2": 6.537355606301714,
"delta_retention_norm_ratio": 1.0177426891701709,
"delta_cosine": 0.8668797945436078,
"relative_weight_rmse": 0.017944310239398756
},
{
"index": 150,
"total": 252,
"projection": "layers.28.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3942,
"parameters": 50331648,
"reference_delta_l2": 3.243137758540773,
"retained_delta_l2": 2.875504229529378,
"delta_retention_norm_ratio": 0.8866426416690949,
"delta_cosine": 0.6978610398465285,
"relative_weight_rmse": 0.01232328346889927
},
{
"index": 151,
"total": 252,
"projection": "layers.28.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 1.0316368625723313,
"retained_delta_l2": 0.9816314502213522,
"delta_retention_norm_ratio": 0.9515280869023105,
"delta_cosine": 0.7709917053597658,
"relative_weight_rmse": 0.014130999330449982
},
{
"index": 152,
"total": 252,
"projection": "layers.28.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1354,
"parameters": 16777216,
"reference_delta_l2": 1.8674884937261969,
"retained_delta_l2": 1.715844398314334,
"delta_retention_norm_ratio": 0.9187978421707502,
"delta_cosine": 0.7338063460605813,
"relative_weight_rmse": 0.012833414630404159
},
{
"index": 153,
"total": 252,
"projection": "layers.28.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1345,
"parameters": 16777216,
"reference_delta_l2": 2.1062141931540292,
"retained_delta_l2": 2.0063063710901976,
"delta_retention_norm_ratio": 0.9525652127933762,
"delta_cosine": 0.7769483273433703,
"relative_weight_rmse": 0.013246534383223457
},
{
"index": 154,
"total": 252,
"projection": "layers.28.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 1.0568340113396937,
"retained_delta_l2": 0.9874382825117356,
"delta_retention_norm_ratio": 0.9343362078781051,
"delta_cosine": 0.7518618664441322,
"relative_weight_rmse": 0.012582377681010531
},
{
"index": 155,
"total": 252,
"projection": "layers.29.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.6148,
"parameters": 50331648,
"reference_delta_l2": 3.7401416658072963,
"retained_delta_l2": 3.5095521416446185,
"delta_retention_norm_ratio": 0.9383473823275873,
"delta_cosine": 0.7558945279178694,
"relative_weight_rmse": 0.012997864369369908
},
{
"index": 156,
"total": 252,
"projection": "layers.29.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.516,
"parameters": 50331648,
"reference_delta_l2": 6.127935986595271,
"retained_delta_l2": 6.212127194514415,
"delta_retention_norm_ratio": 1.0137389176556855,
"delta_cosine": 0.8568512009900341,
"relative_weight_rmse": 0.017714506498618523
},
{
"index": 157,
"total": 252,
"projection": "layers.29.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3926,
"parameters": 50331648,
"reference_delta_l2": 3.580009790840115,
"retained_delta_l2": 3.2798771313174777,
"delta_retention_norm_ratio": 0.9161642908657504,
"delta_cosine": 0.7293571040606517,
"relative_weight_rmse": 0.01289677374486609
},
{
"index": 158,
"total": 252,
"projection": "layers.29.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0309,
"parameters": 4194304,
"reference_delta_l2": 0.9620634543526848,
"retained_delta_l2": 0.9117803875080858,
"delta_retention_norm_ratio": 0.9477341472466266,
"delta_cosine": 0.7817947687473311,
"relative_weight_rmse": 0.013131358653178712
},
{
"index": 159,
"total": 252,
"projection": "layers.29.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1346,
"parameters": 16777216,
"reference_delta_l2": 1.9458139633919214,
"retained_delta_l2": 1.79409683508142,
"delta_retention_norm_ratio": 0.9220289651709407,
"delta_cosine": 0.7376047556088177,
"relative_weight_rmse": 0.01293095805509277
},
{
"index": 160,
"total": 252,
"projection": "layers.29.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.134,
"parameters": 16777216,
"reference_delta_l2": 2.9427336589969983,
"retained_delta_l2": 2.929736370121851,
"delta_retention_norm_ratio": 0.9955832601990976,
"delta_cosine": 0.8477660174783292,
"relative_weight_rmse": 0.015273119054114149
},
{
"index": 161,
"total": 252,
"projection": "layers.29.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 1.0690905177656662,
"retained_delta_l2": 1.0095726237992413,
"delta_retention_norm_ratio": 0.9443284801638558,
"delta_cosine": 0.7674733040578655,
"relative_weight_rmse": 0.012951905130350124
},
{
"index": 162,
"total": 252,
"projection": "layers.3.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4443,
"parameters": 50331648,
"reference_delta_l2": 2.0679400082790367,
"retained_delta_l2": 1.8308734297139297,
"delta_retention_norm_ratio": 0.885360998086982,
"delta_cosine": 0.7163025161839787,
"relative_weight_rmse": 0.011341012968361944
},
{
"index": 163,
"total": 252,
"projection": "layers.3.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3974,
"parameters": 50331648,
"reference_delta_l2": 2.9244244667822037,
"retained_delta_l2": 2.702986678376405,
"delta_retention_norm_ratio": 0.9242798742381435,
"delta_cosine": 0.7458795583880982,
"relative_weight_rmse": 0.012253597562551328
},
{
"index": 164,
"total": 252,
"projection": "layers.3.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3976,
"parameters": 50331648,
"reference_delta_l2": 3.1770019858721863,
"retained_delta_l2": 3.1111471145462324,
"delta_retention_norm_ratio": 0.979271378608259,
"delta_cosine": 0.8422329906912724,
"relative_weight_rmse": 0.014673401123906079
},
{
"index": 165,
"total": 252,
"projection": "layers.3.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0331,
"parameters": 4194304,
"reference_delta_l2": 0.5944881406399625,
"retained_delta_l2": 0.4613283113924287,
"delta_retention_norm_ratio": 0.7760092756363681,
"delta_cosine": 0.5969925389093143,
"relative_weight_rmse": 0.009636678973232336
},
{
"index": 166,
"total": 252,
"projection": "layers.3.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.138,
"parameters": 16777216,
"reference_delta_l2": 1.8542744034159462,
"retained_delta_l2": 1.7012318629959648,
"delta_retention_norm_ratio": 0.9174649986334027,
"delta_cosine": 0.7309382652239491,
"relative_weight_rmse": 0.01283209216131865
},
{
"index": 167,
"total": 252,
"projection": "layers.3.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1356,
"parameters": 16777216,
"reference_delta_l2": 1.509164806161759,
"retained_delta_l2": 1.3064463511188427,
"delta_retention_norm_ratio": 0.8656750712611118,
"delta_cosine": 0.6791005299483913,
"relative_weight_rmse": 0.01172984899046623
},
{
"index": 168,
"total": 252,
"projection": "layers.3.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0307,
"parameters": 4194304,
"reference_delta_l2": 0.9490041062383832,
"retained_delta_l2": 0.8558240584115787,
"delta_retention_norm_ratio": 0.9018128085913694,
"delta_cosine": 0.7160810428968206,
"relative_weight_rmse": 0.012407708147450661
},
{
"index": 169,
"total": 252,
"projection": "layers.30.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4365,
"parameters": 50331648,
"reference_delta_l2": 3.887870797219381,
"retained_delta_l2": 3.6874548470093127,
"delta_retention_norm_ratio": 0.9484509746688581,
"delta_cosine": 0.7666196798290923,
"relative_weight_rmse": 0.013139021782059881
},
{
"index": 170,
"total": 252,
"projection": "layers.30.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3966,
"parameters": 50331648,
"reference_delta_l2": 5.399243984840606,
"retained_delta_l2": 5.396158107286254,
"delta_retention_norm_ratio": 0.9994284611765988,
"delta_cosine": 0.8307007253136764,
"relative_weight_rmse": 0.016931017999366662
},
{
"index": 171,
"total": 252,
"projection": "layers.30.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3967,
"parameters": 50331648,
"reference_delta_l2": 3.809190933435205,
"retained_delta_l2": 3.5287988984779015,
"delta_retention_norm_ratio": 0.9263906588414458,
"delta_cosine": 0.7397053682401827,
"relative_weight_rmse": 0.013403786510319985
},
{
"index": 172,
"total": 252,
"projection": "layers.30.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 1.495041637306258,
"retained_delta_l2": 1.5068418452467638,
"delta_retention_norm_ratio": 1.0078928958539024,
"delta_cosine": 0.8527559337476618,
"relative_weight_rmse": 0.016604996721110866
},
{
"index": 173,
"total": 252,
"projection": "layers.30.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1353,
"parameters": 16777216,
"reference_delta_l2": 1.8684826155757526,
"retained_delta_l2": 1.6923697735085421,
"delta_retention_norm_ratio": 0.9057455281632667,
"delta_cosine": 0.717530868631098,
"relative_weight_rmse": 0.012862563146356202
},
{
"index": 174,
"total": 252,
"projection": "layers.30.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1337,
"parameters": 16777216,
"reference_delta_l2": 3.018528780814635,
"retained_delta_l2": 3.0295039027640356,
"delta_retention_norm_ratio": 1.0036359176096503,
"delta_cosine": 0.8550451128261393,
"relative_weight_rmse": 0.016293838818625218
},
{
"index": 175,
"total": 252,
"projection": "layers.30.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 1.3092329397348734,
"retained_delta_l2": 1.2710158221445458,
"delta_retention_norm_ratio": 0.9708095355452431,
"delta_cosine": 0.8014254919574851,
"relative_weight_rmse": 0.013918002500388403
},
{
"index": 176,
"total": 252,
"projection": "layers.31.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4388,
"parameters": 50331648,
"reference_delta_l2": 4.110364027266316,
"retained_delta_l2": 3.9447523499479558,
"delta_retention_norm_ratio": 0.959708756640588,
"delta_cosine": 0.7805102658039987,
"relative_weight_rmse": 0.013378582867462649
},
{
"index": 177,
"total": 252,
"projection": "layers.31.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4003,
"parameters": 50331648,
"reference_delta_l2": 4.213688960532953,
"retained_delta_l2": 4.065268359369619,
"delta_retention_norm_ratio": 0.9647765645367994,
"delta_cosine": 0.7857099884712581,
"relative_weight_rmse": 0.014756743435253107
},
{
"index": 178,
"total": 252,
"projection": "layers.31.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3951,
"parameters": 50331648,
"reference_delta_l2": 3.7452434781817034,
"retained_delta_l2": 3.457631741618175,
"delta_retention_norm_ratio": 0.92320613112631,
"delta_cosine": 0.7366745284849822,
"relative_weight_rmse": 0.013124970861274277
},
{
"index": 179,
"total": 252,
"projection": "layers.31.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0322,
"parameters": 4194304,
"reference_delta_l2": 1.131255621313341,
"retained_delta_l2": 1.1076962069633571,
"delta_retention_norm_ratio": 0.9791741018509749,
"delta_cosine": 0.8129665759244713,
"relative_weight_rmse": 0.014434867123085856
},
{
"index": 180,
"total": 252,
"projection": "layers.31.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1346,
"parameters": 16777216,
"reference_delta_l2": 1.9271778268048494,
"retained_delta_l2": 1.7785369541374052,
"delta_retention_norm_ratio": 0.9228712210155082,
"delta_cosine": 0.738814993615326,
"relative_weight_rmse": 0.012540620968805942
},
{
"index": 181,
"total": 252,
"projection": "layers.31.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1351,
"parameters": 16777216,
"reference_delta_l2": 3.0951075436016566,
"retained_delta_l2": 3.1186091997514307,
"delta_retention_norm_ratio": 1.0075931630221888,
"delta_cosine": 0.8593239195692756,
"relative_weight_rmse": 0.016359630296747883
},
{
"index": 182,
"total": 252,
"projection": "layers.31.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0311,
"parameters": 4194304,
"reference_delta_l2": 0.913180736908778,
"retained_delta_l2": 0.8056474687920849,
"delta_retention_norm_ratio": 0.8822431707432797,
"delta_cosine": 0.6994518824150215,
"relative_weight_rmse": 0.011245913328725037
},
{
"index": 183,
"total": 252,
"projection": "layers.32.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4373,
"parameters": 50331648,
"reference_delta_l2": 3.053233543999877,
"retained_delta_l2": 2.7010398933001922,
"delta_retention_norm_ratio": 0.8846489645733767,
"delta_cosine": 0.6983303428179822,
"relative_weight_rmse": 0.011288760006122339
},
{
"index": 184,
"total": 252,
"projection": "layers.32.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3952,
"parameters": 50331648,
"reference_delta_l2": 5.735749275322441,
"retained_delta_l2": 5.787299683307666,
"delta_retention_norm_ratio": 1.008987563003672,
"delta_cosine": 0.8483716330269929,
"relative_weight_rmse": 0.01748269997937928
},
{
"index": 185,
"total": 252,
"projection": "layers.32.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.393,
"parameters": 50331648,
"reference_delta_l2": 4.1791007274280325,
"retained_delta_l2": 3.9628442817535,
"delta_retention_norm_ratio": 0.9482528754917988,
"delta_cosine": 0.7672786556742622,
"relative_weight_rmse": 0.013836685178664192
},
{
"index": 186,
"total": 252,
"projection": "layers.32.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 1.0751946475751073,
"retained_delta_l2": 1.040603625981545,
"delta_retention_norm_ratio": 0.9678281307746688,
"delta_cosine": 0.7986439823197274,
"relative_weight_rmse": 0.014082435322431848
},
{
"index": 187,
"total": 252,
"projection": "layers.32.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1355,
"parameters": 16777216,
"reference_delta_l2": 1.9526543310434923,
"retained_delta_l2": 1.7946643532432096,
"delta_retention_norm_ratio": 0.9190896333833682,
"delta_cosine": 0.7327058975081081,
"relative_weight_rmse": 0.012906716155433212
},
{
"index": 188,
"total": 252,
"projection": "layers.32.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.135,
"parameters": 16777216,
"reference_delta_l2": 3.5037919178289854,
"retained_delta_l2": 3.5634283880741373,
"delta_retention_norm_ratio": 1.0170205513465833,
"delta_cosine": 0.8837426455755633,
"relative_weight_rmse": 0.017016034747910174
},
{
"index": 189,
"total": 252,
"projection": "layers.32.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0307,
"parameters": 4194304,
"reference_delta_l2": 1.1042248941807926,
"retained_delta_l2": 1.0345746225272665,
"delta_retention_norm_ratio": 0.9369238349718617,
"delta_cosine": 0.7579939795578423,
"relative_weight_rmse": 0.01259764580426598
},
{
"index": 190,
"total": 252,
"projection": "layers.33.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4356,
"parameters": 50331648,
"reference_delta_l2": 4.677230833346037,
"retained_delta_l2": 4.590480414565143,
"delta_retention_norm_ratio": 0.9814526111984015,
"delta_cosine": 0.8097453894324198,
"relative_weight_rmse": 0.01440073902379026
},
{
"index": 191,
"total": 252,
"projection": "layers.33.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3933,
"parameters": 50331648,
"reference_delta_l2": 4.002110728928395,
"retained_delta_l2": 3.8174978532167745,
"delta_retention_norm_ratio": 0.9538711224611587,
"delta_cosine": 0.7708522234085368,
"relative_weight_rmse": 0.014726103170449745
},
{
"index": 192,
"total": 252,
"projection": "layers.33.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3936,
"parameters": 50331648,
"reference_delta_l2": 4.549819665593091,
"retained_delta_l2": 4.4131762772195,
"delta_retention_norm_ratio": 0.9699672957574728,
"delta_cosine": 0.7954954853276326,
"relative_weight_rmse": 0.014223793742498481
},
{
"index": 193,
"total": 252,
"projection": "layers.33.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 1.1715315410735023,
"retained_delta_l2": 1.1622101639761016,
"delta_retention_norm_ratio": 0.9920434262581961,
"delta_cosine": 0.8277438557448046,
"relative_weight_rmse": 0.015548853767619908
},
{
"index": 194,
"total": 252,
"projection": "layers.33.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1386,
"parameters": 16777216,
"reference_delta_l2": 1.9572263469894886,
"retained_delta_l2": 1.809083490347337,
"delta_retention_norm_ratio": 0.9243097984706685,
"delta_cosine": 0.7413618217608011,
"relative_weight_rmse": 0.012300056453069263
},
{
"index": 195,
"total": 252,
"projection": "layers.33.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1413,
"parameters": 16777216,
"reference_delta_l2": 1.695420696327585,
"retained_delta_l2": 1.5631995893456434,
"delta_retention_norm_ratio": 0.9220128035075053,
"delta_cosine": 0.7427885960279792,
"relative_weight_rmse": 0.011923084861626338
},
{
"index": 196,
"total": 252,
"projection": "layers.33.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0307,
"parameters": 4194304,
"reference_delta_l2": 1.2334249104591484,
"retained_delta_l2": 1.1838953351223982,
"delta_retention_norm_ratio": 0.9598438665242195,
"delta_cosine": 0.7863020889219451,
"relative_weight_rmse": 0.01272197288675906
},
{
"index": 197,
"total": 252,
"projection": "layers.34.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4345,
"parameters": 50331648,
"reference_delta_l2": 3.2554531412284744,
"retained_delta_l2": 3.0244708407371674,
"delta_retention_norm_ratio": 0.9290475732652862,
"delta_cosine": 0.7454650945192635,
"relative_weight_rmse": 0.011840481594971156
},
{
"index": 198,
"total": 252,
"projection": "layers.34.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3928,
"parameters": 50331648,
"reference_delta_l2": 3.5119099908536424,
"retained_delta_l2": 3.307978801417105,
"delta_retention_norm_ratio": 0.9419315443825006,
"delta_cosine": 0.7569929254989856,
"relative_weight_rmse": 0.013105608965876779
},
{
"index": 199,
"total": 252,
"projection": "layers.34.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3944,
"parameters": 50331648,
"reference_delta_l2": 3.1197603574733135,
"retained_delta_l2": 2.795634123462441,
"delta_retention_norm_ratio": 0.8961054065468729,
"delta_cosine": 0.709619591445098,
"relative_weight_rmse": 0.011414617605965967
},
{
"index": 200,
"total": 252,
"projection": "layers.34.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 0.7884807301773037,
"retained_delta_l2": 0.7270568299201107,
"delta_retention_norm_ratio": 0.9220984129271228,
"delta_cosine": 0.7525313403172603,
"relative_weight_rmse": 0.011626929601153246
},
{
"index": 201,
"total": 252,
"projection": "layers.34.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1354,
"parameters": 16777216,
"reference_delta_l2": 2.3657537786912455,
"retained_delta_l2": 2.295153793011888,
"delta_retention_norm_ratio": 0.970157509071627,
"delta_cosine": 0.7929635793398593,
"relative_weight_rmse": 0.014041838832518524
},
{
"index": 202,
"total": 252,
"projection": "layers.34.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1308,
"parameters": 16777216,
"reference_delta_l2": 2.7912707575141327,
"retained_delta_l2": 2.8046778609760277,
"delta_retention_norm_ratio": 1.0048032257085069,
"delta_cosine": 0.859762626141908,
"relative_weight_rmse": 0.015280208881268444
},
{
"index": 203,
"total": 252,
"projection": "layers.34.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0303,
"parameters": 4194304,
"reference_delta_l2": 0.98146189599056,
"retained_delta_l2": 0.8769177949999751,
"delta_retention_norm_ratio": 0.8934812432172197,
"delta_cosine": 0.7154066928079399,
"relative_weight_rmse": 0.011405952287331642
},
{
"index": 204,
"total": 252,
"projection": "layers.35.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4307,
"parameters": 50331648,
"reference_delta_l2": 2.323442105920221,
"retained_delta_l2": 2.0756301760951703,
"delta_retention_norm_ratio": 0.8933427567686683,
"delta_cosine": 0.7136688211460611,
"relative_weight_rmse": 0.009788394551676757
},
{
"index": 205,
"total": 252,
"projection": "layers.35.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3882,
"parameters": 50331648,
"reference_delta_l2": 2.619391209776972,
"retained_delta_l2": 2.284627070690156,
"delta_retention_norm_ratio": 0.8721977313517366,
"delta_cosine": 0.686894707285102,
"relative_weight_rmse": 0.010512670834223626
},
{
"index": 206,
"total": 252,
"projection": "layers.35.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3896,
"parameters": 50331648,
"reference_delta_l2": 2.658687967607658,
"retained_delta_l2": 2.2856054411666,
"delta_retention_norm_ratio": 0.8596741960747031,
"delta_cosine": 0.6743087239399935,
"relative_weight_rmse": 0.01033738841891372
},
{
"index": 207,
"total": 252,
"projection": "layers.35.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 0.8329213911977184,
"retained_delta_l2": 0.7629292702052147,
"delta_retention_norm_ratio": 0.9159679151812192,
"delta_cosine": 0.7342860229477369,
"relative_weight_rmse": 0.01222969875625576
},
{
"index": 208,
"total": 252,
"projection": "layers.35.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1319,
"parameters": 16777216,
"reference_delta_l2": 1.978736408751587,
"retained_delta_l2": 1.877442477532796,
"delta_retention_norm_ratio": 0.9488087797996809,
"delta_cosine": 0.767618956994911,
"relative_weight_rmse": 0.012986472755049936
},
{
"index": 209,
"total": 252,
"projection": "layers.35.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1323,
"parameters": 16777216,
"reference_delta_l2": 1.9212411836563124,
"retained_delta_l2": 1.8332294176962938,
"delta_retention_norm_ratio": 0.9541901523303161,
"delta_cosine": 0.7745190405033069,
"relative_weight_rmse": 0.013081762644918027
},
{
"index": 210,
"total": 252,
"projection": "layers.35.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0298,
"parameters": 4194304,
"reference_delta_l2": 0.9030694749944501,
"retained_delta_l2": 0.7966562129589475,
"delta_retention_norm_ratio": 0.8821649219888021,
"delta_cosine": 0.6990966561194207,
"relative_weight_rmse": 0.011406911676881168
},
{
"index": 211,
"total": 252,
"projection": "layers.4.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4323,
"parameters": 50331648,
"reference_delta_l2": 2.2245974764067546,
"retained_delta_l2": 1.9080686344231745,
"delta_retention_norm_ratio": 0.8577141054323012,
"delta_cosine": 0.6729869101107243,
"relative_weight_rmse": 0.010835389092090777
},
{
"index": 212,
"total": 252,
"projection": "layers.4.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3893,
"parameters": 50331648,
"reference_delta_l2": 3.314203046594274,
"retained_delta_l2": 2.92327997457648,
"delta_retention_norm_ratio": 0.882046130993841,
"delta_cosine": 0.7090416043478848,
"relative_weight_rmse": 0.010831238116930547
},
{
"index": 213,
"total": 252,
"projection": "layers.4.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3878,
"parameters": 50331648,
"reference_delta_l2": 3.2992289651260265,
"retained_delta_l2": 3.176838706079354,
"delta_retention_norm_ratio": 0.9629033752005759,
"delta_cosine": 0.7950672894455574,
"relative_weight_rmse": 0.01401853766560735
},
{
"index": 214,
"total": 252,
"projection": "layers.4.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0309,
"parameters": 4194304,
"reference_delta_l2": 0.5934329624541549,
"retained_delta_l2": 0.4482821586618158,
"delta_retention_norm_ratio": 0.7554048848381041,
"delta_cosine": 0.5783625575635638,
"relative_weight_rmse": 0.009466441979177278
},
{
"index": 215,
"total": 252,
"projection": "layers.4.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1334,
"parameters": 16777216,
"reference_delta_l2": 1.6294614994627534,
"retained_delta_l2": 1.4159643905718167,
"delta_retention_norm_ratio": 0.8689768926965574,
"delta_cosine": 0.6811809510270403,
"relative_weight_rmse": 0.011808513059008886
},
{
"index": 216,
"total": 252,
"projection": "layers.4.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1339,
"parameters": 16777216,
"reference_delta_l2": 1.2620082729612303,
"retained_delta_l2": 0.9977812321989201,
"delta_retention_norm_ratio": 0.7906297078843101,
"delta_cosine": 0.6084140135379782,
"relative_weight_rmse": 0.010243938736757576
},
{
"index": 217,
"total": 252,
"projection": "layers.4.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0308,
"parameters": 4194304,
"reference_delta_l2": 1.1348753309421484,
"retained_delta_l2": 1.077055790992353,
"delta_retention_norm_ratio": 0.9490520779037512,
"delta_cosine": 0.7689872270233765,
"relative_weight_rmse": 0.013531675501347078
},
{
"index": 218,
"total": 252,
"projection": "layers.5.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4312,
"parameters": 50331648,
"reference_delta_l2": 1.8065839643892028,
"retained_delta_l2": 1.3452533951368082,
"delta_retention_norm_ratio": 0.7446392869935784,
"delta_cosine": 0.5698075119000315,
"relative_weight_rmse": 0.009053654516748385
},
{
"index": 219,
"total": 252,
"projection": "layers.5.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3906,
"parameters": 50331648,
"reference_delta_l2": 2.9535589675034295,
"retained_delta_l2": 2.48962526367793,
"delta_retention_norm_ratio": 0.8429238390260916,
"delta_cosine": 0.6604189599227185,
"relative_weight_rmse": 0.00999517624475973
},
{
"index": 220,
"total": 252,
"projection": "layers.5.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3916,
"parameters": 50331648,
"reference_delta_l2": 3.1775253397334686,
"retained_delta_l2": 2.9633965747586823,
"delta_retention_norm_ratio": 0.9326114689638486,
"delta_cosine": 0.7494165594063157,
"relative_weight_rmse": 0.013127994811945903
},
{
"index": 221,
"total": 252,
"projection": "layers.5.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0311,
"parameters": 4194304,
"reference_delta_l2": 0.6207896422405921,
"retained_delta_l2": 0.48910098674505714,
"delta_retention_norm_ratio": 0.7878691161465965,
"delta_cosine": 0.6084948994396171,
"relative_weight_rmse": 0.00986678113185319
},
{
"index": 222,
"total": 252,
"projection": "layers.5.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.135,
"parameters": 16777216,
"reference_delta_l2": 1.468287392795666,
"retained_delta_l2": 1.2335401666266865,
"delta_retention_norm_ratio": 0.8401217450202216,
"delta_cosine": 0.6521262975815546,
"relative_weight_rmse": 0.01135293861593865
},
{
"index": 223,
"total": 252,
"projection": "layers.5.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1335,
"parameters": 16777216,
"reference_delta_l2": 1.4697062134668988,
"retained_delta_l2": 1.255421403051614,
"delta_retention_norm_ratio": 0.8541988810744653,
"delta_cosine": 0.6707564193730499,
"relative_weight_rmse": 0.011192511327309623
},
{
"index": 224,
"total": 252,
"projection": "layers.5.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0303,
"parameters": 4194304,
"reference_delta_l2": 1.046720925079732,
"retained_delta_l2": 0.9822764761169425,
"delta_retention_norm_ratio": 0.9384320620533304,
"delta_cosine": 0.7609794098905587,
"relative_weight_rmse": 0.012878622273056339
},
{
"index": 225,
"total": 252,
"projection": "layers.6.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4324,
"parameters": 50331648,
"reference_delta_l2": 4.763581604307177,
"retained_delta_l2": 4.776047230176001,
"delta_retention_norm_ratio": 1.0026168599394945,
"delta_cosine": 0.8440799355337425,
"relative_weight_rmse": 0.015429444299361368
},
{
"index": 226,
"total": 252,
"projection": "layers.6.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3916,
"parameters": 50331648,
"reference_delta_l2": 3.105994770214036,
"retained_delta_l2": 2.7156819801891334,
"delta_retention_norm_ratio": 0.874335657687535,
"delta_cosine": 0.6920577735011633,
"relative_weight_rmse": 0.010681939796870762
},
{
"index": 227,
"total": 252,
"projection": "layers.6.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3917,
"parameters": 50331648,
"reference_delta_l2": 3.385041898713471,
"retained_delta_l2": 3.1967324013083513,
"delta_retention_norm_ratio": 0.944370113269,
"delta_cosine": 0.762999101289802,
"relative_weight_rmse": 0.013216400552743588
},
{
"index": 228,
"total": 252,
"projection": "layers.6.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0311,
"parameters": 4194304,
"reference_delta_l2": 0.5923543402760572,
"retained_delta_l2": 0.44722429559841054,
"delta_retention_norm_ratio": 0.754994544971155,
"delta_cosine": 0.5781239453280363,
"relative_weight_rmse": 0.009575510514011771
},
{
"index": 229,
"total": 252,
"projection": "layers.6.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1367,
"parameters": 16777216,
"reference_delta_l2": 1.3637666672894704,
"retained_delta_l2": 1.105231379185307,
"delta_retention_norm_ratio": 0.810425570366806,
"delta_cosine": 0.6263504572146829,
"relative_weight_rmse": 0.01038836662554142
},
{
"index": 230,
"total": 252,
"projection": "layers.6.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1335,
"parameters": 16777216,
"reference_delta_l2": 1.4025034551702102,
"retained_delta_l2": 1.1632515080468524,
"delta_retention_norm_ratio": 0.8294107966426921,
"delta_cosine": 0.6445231369465194,
"relative_weight_rmse": 0.010719962886095127
},
{
"index": 231,
"total": 252,
"projection": "layers.6.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 0.9849644445899759,
"retained_delta_l2": 0.9089375611654249,
"delta_retention_norm_ratio": 0.9228125605526809,
"delta_cosine": 0.739246623026045,
"relative_weight_rmse": 0.012389586886707311
},
{
"index": 232,
"total": 252,
"projection": "layers.7.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4338,
"parameters": 50331648,
"reference_delta_l2": 1.845972451357094,
"retained_delta_l2": 1.3659906541197822,
"delta_retention_norm_ratio": 0.7399843118544667,
"delta_cosine": 0.5655852668367456,
"relative_weight_rmse": 0.00896342088611992
},
{
"index": 233,
"total": 252,
"projection": "layers.7.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4004,
"parameters": 50331648,
"reference_delta_l2": 2.8972880598888335,
"retained_delta_l2": 2.4823971140636263,
"delta_retention_norm_ratio": 0.8568002431069536,
"delta_cosine": 0.6737961212272754,
"relative_weight_rmse": 0.010353212347255036
},
{
"index": 234,
"total": 252,
"projection": "layers.7.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3954,
"parameters": 50331648,
"reference_delta_l2": 2.976464301108795,
"retained_delta_l2": 2.693361096432054,
"delta_retention_norm_ratio": 0.9048860742017706,
"delta_cosine": 0.71663898129702,
"relative_weight_rmse": 0.01238607089131992
},
{
"index": 235,
"total": 252,
"projection": "layers.7.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.032,
"parameters": 4194304,
"reference_delta_l2": 0.7227176135675857,
"retained_delta_l2": 0.6135642538998713,
"delta_retention_norm_ratio": 0.8489681756489987,
"delta_cosine": 0.6626721699461946,
"relative_weight_rmse": 0.010772914353297651
},
{
"index": 236,
"total": 252,
"projection": "layers.7.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1364,
"parameters": 16777216,
"reference_delta_l2": 1.3660572752775417,
"retained_delta_l2": 1.1066998580108307,
"delta_retention_norm_ratio": 0.8101416229315733,
"delta_cosine": 0.6252957723577145,
"relative_weight_rmse": 0.010550120700751548
},
{
"index": 237,
"total": 252,
"projection": "layers.7.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1352,
"parameters": 16777216,
"reference_delta_l2": 1.6868798978306623,
"retained_delta_l2": 1.5252113523148896,
"delta_retention_norm_ratio": 0.9041611997844782,
"delta_cosine": 0.7181098490516066,
"relative_weight_rmse": 0.012121673317143276
},
{
"index": 238,
"total": 252,
"projection": "layers.7.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0311,
"parameters": 4194304,
"reference_delta_l2": 0.9902755720656208,
"retained_delta_l2": 0.8939089233714745,
"delta_retention_norm_ratio": 0.9026870384239263,
"delta_cosine": 0.7179580569495855,
"relative_weight_rmse": 0.011989921320523676
},
{
"index": 239,
"total": 252,
"projection": "layers.8.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4369,
"parameters": 50331648,
"reference_delta_l2": 2.641258801378591,
"retained_delta_l2": 2.2794388272454036,
"delta_retention_norm_ratio": 0.8630122977936363,
"delta_cosine": 0.677297685340227,
"relative_weight_rmse": 0.011057474749445264
},
{
"index": 240,
"total": 252,
"projection": "layers.8.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3945,
"parameters": 50331648,
"reference_delta_l2": 2.2050936875452227,
"retained_delta_l2": 1.6625676195063588,
"delta_retention_norm_ratio": 0.7539668853513337,
"delta_cosine": 0.5773198349692906,
"relative_weight_rmse": 0.009475486385452227
},
{
"index": 241,
"total": 252,
"projection": "layers.8.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3945,
"parameters": 50331648,
"reference_delta_l2": 2.679677747498109,
"retained_delta_l2": 2.284396234756145,
"delta_retention_norm_ratio": 0.8524891610153421,
"delta_cosine": 0.6657822834602609,
"relative_weight_rmse": 0.011427666874516884
},
{
"index": 242,
"total": 252,
"projection": "layers.8.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0333,
"parameters": 4194304,
"reference_delta_l2": 0.8932447143308502,
"retained_delta_l2": 0.8098993999824166,
"delta_retention_norm_ratio": 0.9066937503113361,
"delta_cosine": 0.7197835568554013,
"relative_weight_rmse": 0.012127856844388323
},
{
"index": 243,
"total": 252,
"projection": "layers.8.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1357,
"parameters": 16777216,
"reference_delta_l2": 1.5718337251376286,
"retained_delta_l2": 1.3461363236464665,
"delta_retention_norm_ratio": 0.8564114016122156,
"delta_cosine": 0.6703790087422654,
"relative_weight_rmse": 0.011303805629343487
},
{
"index": 244,
"total": 252,
"projection": "layers.8.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1355,
"parameters": 16777216,
"reference_delta_l2": 1.5937350677877007,
"retained_delta_l2": 1.382939622920773,
"delta_retention_norm_ratio": 0.867734952234227,
"delta_cosine": 0.679088737874005,
"relative_weight_rmse": 0.01183446250094784
},
{
"index": 245,
"total": 252,
"projection": "layers.8.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 0.7963146426204377,
"retained_delta_l2": 0.6603613999230062,
"delta_retention_norm_ratio": 0.829271954299309,
"delta_cosine": 0.6434056559316997,
"relative_weight_rmse": 0.01062666469671945
},
{
"index": 246,
"total": 252,
"projection": "layers.9.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4357,
"parameters": 50331648,
"reference_delta_l2": 2.467327752515527,
"retained_delta_l2": 2.096186873024144,
"delta_retention_norm_ratio": 0.8495777956078224,
"delta_cosine": 0.6640106694351272,
"relative_weight_rmse": 0.010877294060438782
},
{
"index": 247,
"total": 252,
"projection": "layers.9.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3917,
"parameters": 50331648,
"reference_delta_l2": 2.698803219715491,
"retained_delta_l2": 2.210001490704963,
"delta_retention_norm_ratio": 0.8188820416991878,
"delta_cosine": 0.6350180124903053,
"relative_weight_rmse": 0.010370650234180913
},
{
"index": 248,
"total": 252,
"projection": "layers.9.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3925,
"parameters": 50331648,
"reference_delta_l2": 3.0233243335706885,
"retained_delta_l2": 2.741480983449664,
"delta_retention_norm_ratio": 0.9067770047058915,
"delta_cosine": 0.7217231268446289,
"relative_weight_rmse": 0.01238018059960264
},
{
"index": 249,
"total": 252,
"projection": "layers.9.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0321,
"parameters": 4194304,
"reference_delta_l2": 0.8387394521719984,
"retained_delta_l2": 0.7567867999749869,
"delta_retention_norm_ratio": 0.9022906911260856,
"delta_cosine": 0.7183405611602802,
"relative_weight_rmse": 0.011787582836933876
},
{
"index": 250,
"total": 252,
"projection": "layers.9.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.135,
"parameters": 16777216,
"reference_delta_l2": 1.5247160233711252,
"retained_delta_l2": 1.299308342528706,
"delta_retention_norm_ratio": 0.852164155562525,
"delta_cosine": 0.6666228983984552,
"relative_weight_rmse": 0.011154261210944667
},
{
"index": 251,
"total": 252,
"projection": "layers.9.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1379,
"parameters": 16777216,
"reference_delta_l2": 1.8040588235553003,
"retained_delta_l2": 1.6572263164999084,
"delta_retention_norm_ratio": 0.9186099116402281,
"delta_cosine": 0.7348702177287972,
"relative_weight_rmse": 0.012698203899833869
},
{
"index": 252,
"total": 252,
"projection": "layers.9.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0308,
"parameters": 4194304,
"reference_delta_l2": 0.8010147536640196,
"retained_delta_l2": 0.6707311777466941,
"delta_retention_norm_ratio": 0.8373518398738855,
"delta_cosine": 0.6534294135001347,
"relative_weight_rmse": 0.01042703006403757
}
]
}