mrjackspade's picture
Publish step 3510 merge manifest
e7af8d8 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_00003510/adapters.safetensors",
"adapter_sha256": "3f78038fb2a496d336c5437a1129ce105ca66448e2de752e8f1abe8650c74700",
"adapter_global_step": "3510",
"output": "/workspace/Ideogram/artifacts/huggingface_release/s020_v1/qwen3vl_8b_ideogram4_nl_s020_v1_step_00003510_fp8_scaled.safetensors",
"output_bytes": 10588637800,
"output_sha256": "78c7e664c8ad9c7168e9e875627d7838350e70c44a1d9358a1e80ead03392ef8",
"output_tensors": 1254,
"merged_projections": 252,
"lora_rank": 64,
"lora_alpha": 64.0,
"elapsed_seconds": 70.304,
"projections": [
{
"index": 1,
"total": 252,
"projection": "layers.0.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4534,
"parameters": 50331648,
"reference_delta_l2": 3.129181502262599,
"retained_delta_l2": 2.849571765411123,
"delta_retention_norm_ratio": 0.9106444491477083,
"delta_cosine": 0.725963956941354,
"relative_weight_rmse": 0.012314757333148135
},
{
"index": 2,
"total": 252,
"projection": "layers.0.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4086,
"parameters": 50331648,
"reference_delta_l2": 3.3337197670216088,
"retained_delta_l2": 3.0509625207760105,
"delta_retention_norm_ratio": 0.9151826590096932,
"delta_cosine": 0.7279184627443035,
"relative_weight_rmse": 0.013555924315472518
},
{
"index": 3,
"total": 252,
"projection": "layers.0.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4074,
"parameters": 50331648,
"reference_delta_l2": 2.955960109577813,
"retained_delta_l2": 2.648493023681998,
"delta_retention_norm_ratio": 0.8959840205895981,
"delta_cosine": 0.7091646337618975,
"relative_weight_rmse": 0.013081062046276142
},
{
"index": 4,
"total": 252,
"projection": "layers.0.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0318,
"parameters": 4194304,
"reference_delta_l2": 0.6037935227650949,
"retained_delta_l2": 0.44287902675164964,
"delta_retention_norm_ratio": 0.7334941665546009,
"delta_cosine": 0.5611854527635837,
"relative_weight_rmse": 0.009024314642713551
},
{
"index": 5,
"total": 252,
"projection": "layers.0.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1502,
"parameters": 16777216,
"reference_delta_l2": 1.9745144513507589,
"retained_delta_l2": 1.8626630020097177,
"delta_retention_norm_ratio": 0.9433524281047809,
"delta_cosine": 0.7629244344420502,
"relative_weight_rmse": 0.013629025292775626
},
{
"index": 6,
"total": 252,
"projection": "layers.0.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1338,
"parameters": 16777216,
"reference_delta_l2": 1.228223221402707,
"retained_delta_l2": 0.9566524533281524,
"delta_retention_norm_ratio": 0.7788913583929769,
"delta_cosine": 0.5995317397162899,
"relative_weight_rmse": 0.009822510964249994
},
{
"index": 7,
"total": 252,
"projection": "layers.0.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0295,
"parameters": 4194304,
"reference_delta_l2": 0.7959315234308483,
"retained_delta_l2": 0.6935645814283679,
"delta_retention_norm_ratio": 0.8713872500473036,
"delta_cosine": 0.6841486682247269,
"relative_weight_rmse": 0.011989824017959217
},
{
"index": 8,
"total": 252,
"projection": "layers.1.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4496,
"parameters": 50331648,
"reference_delta_l2": 2.037517401626304,
"retained_delta_l2": 1.9020390502048243,
"delta_retention_norm_ratio": 0.9335081254700728,
"delta_cosine": 0.7565892829532775,
"relative_weight_rmse": 0.011989532799236471
},
{
"index": 9,
"total": 252,
"projection": "layers.1.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4188,
"parameters": 50331648,
"reference_delta_l2": 3.830735627121548,
"retained_delta_l2": 3.7978410797135207,
"delta_retention_norm_ratio": 0.991412994628203,
"delta_cosine": 0.8457848008170453,
"relative_weight_rmse": 0.014362280517668972
},
{
"index": 10,
"total": 252,
"projection": "layers.1.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4193,
"parameters": 50331648,
"reference_delta_l2": 4.237298402759195,
"retained_delta_l2": 4.234414360783846,
"delta_retention_norm_ratio": 0.9993193677430244,
"delta_cosine": 0.9297266764955232,
"relative_weight_rmse": 0.01576342745348578
},
{
"index": 11,
"total": 252,
"projection": "layers.1.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0303,
"parameters": 4194304,
"reference_delta_l2": 0.677620631463665,
"retained_delta_l2": 0.5566885460539878,
"delta_retention_norm_ratio": 0.8215342334715175,
"delta_cosine": 0.6377925021464084,
"relative_weight_rmse": 0.010833166512234186
},
{
"index": 12,
"total": 252,
"projection": "layers.1.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1353,
"parameters": 16777216,
"reference_delta_l2": 1.7142547453382389,
"retained_delta_l2": 1.5337009761988618,
"delta_retention_norm_ratio": 0.8946750652839802,
"delta_cosine": 0.7054103516013294,
"relative_weight_rmse": 0.012663449505014537
},
{
"index": 13,
"total": 252,
"projection": "layers.1.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1369,
"parameters": 16777216,
"reference_delta_l2": 2.0957281429831935,
"retained_delta_l2": 2.0267462888083108,
"delta_retention_norm_ratio": 0.967084540804663,
"delta_cosine": 0.7930254799956314,
"relative_weight_rmse": 0.014648043835512046
},
{
"index": 14,
"total": 252,
"projection": "layers.1.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0299,
"parameters": 4194304,
"reference_delta_l2": 0.9104880969117415,
"retained_delta_l2": 0.8272791245983222,
"delta_retention_norm_ratio": 0.9086105874468284,
"delta_cosine": 0.7211639664261036,
"relative_weight_rmse": 0.0123966860339358
},
{
"index": 15,
"total": 252,
"projection": "layers.10.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4659,
"parameters": 50331648,
"reference_delta_l2": 2.8144663460711623,
"retained_delta_l2": 2.4959902194009564,
"delta_retention_norm_ratio": 0.8868431569221708,
"delta_cosine": 0.700829962481599,
"relative_weight_rmse": 0.011671121533451812
},
{
"index": 16,
"total": 252,
"projection": "layers.10.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4544,
"parameters": 50331648,
"reference_delta_l2": 2.3936277440230165,
"retained_delta_l2": 1.8562866313422006,
"delta_retention_norm_ratio": 0.7755118296808775,
"delta_cosine": 0.5963078160669969,
"relative_weight_rmse": 0.009663354674532077
},
{
"index": 17,
"total": 252,
"projection": "layers.10.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4177,
"parameters": 50331648,
"reference_delta_l2": 2.8563170996760285,
"retained_delta_l2": 2.5276860706959097,
"delta_retention_norm_ratio": 0.8849458874795824,
"delta_cosine": 0.6974442809615463,
"relative_weight_rmse": 0.01190911483390563
},
{
"index": 18,
"total": 252,
"projection": "layers.10.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0333,
"parameters": 4194304,
"reference_delta_l2": 0.8986596075316539,
"retained_delta_l2": 0.8261431601273231,
"delta_retention_norm_ratio": 0.9193059899470597,
"delta_cosine": 0.7372832088369983,
"relative_weight_rmse": 0.01197664420200775
},
{
"index": 19,
"total": 252,
"projection": "layers.10.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1413,
"parameters": 16777216,
"reference_delta_l2": 1.775785806146078,
"retained_delta_l2": 1.587296842741179,
"delta_retention_norm_ratio": 0.893856025455024,
"delta_cosine": 0.7070424533432562,
"relative_weight_rmse": 0.012175673591090182
},
{
"index": 20,
"total": 252,
"projection": "layers.10.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1388,
"parameters": 16777216,
"reference_delta_l2": 2.1216155641654733,
"retained_delta_l2": 2.024516190075094,
"delta_retention_norm_ratio": 0.9542332853649796,
"delta_cosine": 0.7802286121041007,
"relative_weight_rmse": 0.013461967716634627
},
{
"index": 21,
"total": 252,
"projection": "layers.10.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0298,
"parameters": 4194304,
"reference_delta_l2": 0.9111538712390548,
"retained_delta_l2": 0.8138993116014808,
"delta_retention_norm_ratio": 0.8932622000438631,
"delta_cosine": 0.7115583667540949,
"relative_weight_rmse": 0.011215262608775443
},
{
"index": 22,
"total": 252,
"projection": "layers.11.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4562,
"parameters": 50331648,
"reference_delta_l2": 2.6107010148120438,
"retained_delta_l2": 2.2350388407467805,
"delta_retention_norm_ratio": 0.8561067805413526,
"delta_cosine": 0.6707072538209302,
"relative_weight_rmse": 0.011053792660179324
},
{
"index": 23,
"total": 252,
"projection": "layers.11.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4276,
"parameters": 50331648,
"reference_delta_l2": 2.779274623375556,
"retained_delta_l2": 2.3476942828071197,
"delta_retention_norm_ratio": 0.8447147550880516,
"delta_cosine": 0.6600632796961948,
"relative_weight_rmse": 0.01082771749992103
},
{
"index": 24,
"total": 252,
"projection": "layers.11.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4201,
"parameters": 50331648,
"reference_delta_l2": 2.676965416099931,
"retained_delta_l2": 2.2752735327010525,
"delta_retention_norm_ratio": 0.8499450605588685,
"delta_cosine": 0.6635730867698012,
"relative_weight_rmse": 0.011380494790931325
},
{
"index": 25,
"total": 252,
"projection": "layers.11.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0326,
"parameters": 4194304,
"reference_delta_l2": 1.1332802564045652,
"retained_delta_l2": 1.0911203455298104,
"delta_retention_norm_ratio": 0.9627983363898787,
"delta_cosine": 0.7879830061426097,
"relative_weight_rmse": 0.013690291062636377
},
{
"index": 26,
"total": 252,
"projection": "layers.11.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1376,
"parameters": 16777216,
"reference_delta_l2": 1.6713241823170883,
"retained_delta_l2": 1.467968200138001,
"delta_retention_norm_ratio": 0.8783264286302859,
"delta_cosine": 0.6932266854160313,
"relative_weight_rmse": 0.011732857159049927
},
{
"index": 27,
"total": 252,
"projection": "layers.11.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1343,
"parameters": 16777216,
"reference_delta_l2": 1.4109651537848318,
"retained_delta_l2": 1.1746784155143764,
"delta_retention_norm_ratio": 0.832535383572989,
"delta_cosine": 0.6472669805182051,
"relative_weight_rmse": 0.010776480178112764
},
{
"index": 28,
"total": 252,
"projection": "layers.11.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0293,
"parameters": 4194304,
"reference_delta_l2": 0.9999046496863891,
"retained_delta_l2": 0.9098921388024968,
"delta_retention_norm_ratio": 0.9099789055765227,
"delta_cosine": 0.7275145617347912,
"relative_weight_rmse": 0.01190397214223519
},
{
"index": 29,
"total": 252,
"projection": "layers.12.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.5404,
"parameters": 50331648,
"reference_delta_l2": 2.569211503038318,
"retained_delta_l2": 2.189553795757613,
"delta_retention_norm_ratio": 0.8522279279725603,
"delta_cosine": 0.6668454045040122,
"relative_weight_rmse": 0.01083980082357452
},
{
"index": 30,
"total": 252,
"projection": "layers.12.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.5964,
"parameters": 50331648,
"reference_delta_l2": 3.2220693523659687,
"retained_delta_l2": 2.8908806627078163,
"delta_retention_norm_ratio": 0.8972124267235402,
"delta_cosine": 0.7133394290162004,
"relative_weight_rmse": 0.012003467627054025
},
{
"index": 31,
"total": 252,
"projection": "layers.12.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4809,
"parameters": 50331648,
"reference_delta_l2": 2.6906260303125755,
"retained_delta_l2": 2.2945203306394832,
"delta_retention_norm_ratio": 0.852783071593537,
"delta_cosine": 0.666523303194535,
"relative_weight_rmse": 0.01123184476502964
},
{
"index": 32,
"total": 252,
"projection": "layers.12.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0328,
"parameters": 4194304,
"reference_delta_l2": 0.9327969545169129,
"retained_delta_l2": 0.8660627555335398,
"delta_retention_norm_ratio": 0.9284579579079629,
"delta_cosine": 0.7478191656095213,
"relative_weight_rmse": 0.012294088525294533
},
{
"index": 33,
"total": 252,
"projection": "layers.12.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1486,
"parameters": 16777216,
"reference_delta_l2": 1.774314760768344,
"retained_delta_l2": 1.6073532810355304,
"delta_retention_norm_ratio": 0.905900867521097,
"delta_cosine": 0.7219338515051923,
"relative_weight_rmse": 0.01203635391418437
},
{
"index": 34,
"total": 252,
"projection": "layers.12.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.164,
"parameters": 16777216,
"reference_delta_l2": 1.807578915067375,
"retained_delta_l2": 1.6527096898585498,
"delta_retention_norm_ratio": 0.9143222882730668,
"delta_cosine": 0.7313620655902244,
"relative_weight_rmse": 0.012439736805435798
},
{
"index": 35,
"total": 252,
"projection": "layers.12.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0329,
"parameters": 4194304,
"reference_delta_l2": 1.0872159011698592,
"retained_delta_l2": 1.0241110761752685,
"delta_retention_norm_ratio": 0.9419574116542178,
"delta_cosine": 0.7660379196256691,
"relative_weight_rmse": 0.012192235357308697
},
{
"index": 36,
"total": 252,
"projection": "layers.13.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4715,
"parameters": 50331648,
"reference_delta_l2": 2.346940209245875,
"retained_delta_l2": 1.9226253883381939,
"delta_retention_norm_ratio": 0.8192050997992731,
"delta_cosine": 0.6359904359678231,
"relative_weight_rmse": 0.010198384259348343
},
{
"index": 37,
"total": 252,
"projection": "layers.13.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.407,
"parameters": 50331648,
"reference_delta_l2": 2.9979015501050554,
"retained_delta_l2": 2.6247273954769192,
"delta_retention_norm_ratio": 0.8755215445233487,
"delta_cosine": 0.688348969490412,
"relative_weight_rmse": 0.011908901154116741
},
{
"index": 38,
"total": 252,
"projection": "layers.13.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.413,
"parameters": 50331648,
"reference_delta_l2": 2.4036510391184045,
"retained_delta_l2": 1.9438440179593308,
"delta_retention_norm_ratio": 0.8087047521974244,
"delta_cosine": 0.6253044534969704,
"relative_weight_rmse": 0.010355863628057217
},
{
"index": 39,
"total": 252,
"projection": "layers.13.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0331,
"parameters": 4194304,
"reference_delta_l2": 1.3422833818108673,
"retained_delta_l2": 1.3279172027716049,
"delta_retention_norm_ratio": 0.9892972085969797,
"delta_cosine": 0.8345772586118614,
"relative_weight_rmse": 0.014783332005725245
},
{
"index": 40,
"total": 252,
"projection": "layers.13.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1487,
"parameters": 16777216,
"reference_delta_l2": 1.732001352790981,
"retained_delta_l2": 1.5526660023788106,
"delta_retention_norm_ratio": 0.8964577307499293,
"delta_cosine": 0.7122243081908607,
"relative_weight_rmse": 0.012270951969538693
},
{
"index": 41,
"total": 252,
"projection": "layers.13.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1503,
"parameters": 16777216,
"reference_delta_l2": 1.4115713904007354,
"retained_delta_l2": 1.1756288054530724,
"delta_retention_norm_ratio": 0.832851114331043,
"delta_cosine": 0.6474098618674138,
"relative_weight_rmse": 0.01076085545507543
},
{
"index": 42,
"total": 252,
"projection": "layers.13.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0325,
"parameters": 4194304,
"reference_delta_l2": 1.0324766307200204,
"retained_delta_l2": 0.9516964922553652,
"delta_retention_norm_ratio": 0.9217608069169361,
"delta_cosine": 0.7393890674819148,
"relative_weight_rmse": 0.012348244731211457
},
{
"index": 43,
"total": 252,
"projection": "layers.14.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4912,
"parameters": 50331648,
"reference_delta_l2": 2.4704916026237145,
"retained_delta_l2": 2.0898728084193303,
"delta_retention_norm_ratio": 0.8459339858511727,
"delta_cosine": 0.6615998972243725,
"relative_weight_rmse": 0.010526871142835285
},
{
"index": 44,
"total": 252,
"projection": "layers.14.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.415,
"parameters": 50331648,
"reference_delta_l2": 2.740583113971524,
"retained_delta_l2": 2.3266705898622986,
"delta_retention_norm_ratio": 0.8489691766693391,
"delta_cosine": 0.6622014349646035,
"relative_weight_rmse": 0.011382229634738105
},
{
"index": 45,
"total": 252,
"projection": "layers.14.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4077,
"parameters": 50331648,
"reference_delta_l2": 2.9483577850276887,
"retained_delta_l2": 2.614426607079257,
"delta_retention_norm_ratio": 0.8867399405715966,
"delta_cosine": 0.701177198801095,
"relative_weight_rmse": 0.011793760767552794
},
{
"index": 46,
"total": 252,
"projection": "layers.14.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0293,
"parameters": 4194304,
"reference_delta_l2": 0.7294438784911281,
"retained_delta_l2": 0.6195300460237532,
"delta_retention_norm_ratio": 0.8493183153517796,
"delta_cosine": 0.666293755641315,
"relative_weight_rmse": 0.010725281316582744
},
{
"index": 47,
"total": 252,
"projection": "layers.14.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1351,
"parameters": 16777216,
"reference_delta_l2": 1.5574544323950057,
"retained_delta_l2": 1.356282564492752,
"delta_retention_norm_ratio": 0.8708329028972631,
"delta_cosine": 0.6851090623358104,
"relative_weight_rmse": 0.011451576764233052
},
{
"index": 48,
"total": 252,
"projection": "layers.14.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.143,
"parameters": 16777216,
"reference_delta_l2": 2.3202034320113154,
"retained_delta_l2": 2.261853379734366,
"delta_retention_norm_ratio": 0.9748513205902952,
"delta_cosine": 0.8056851237416697,
"relative_weight_rmse": 0.013919410808683358
},
{
"index": 49,
"total": 252,
"projection": "layers.14.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0328,
"parameters": 4194304,
"reference_delta_l2": 1.0177925604789873,
"retained_delta_l2": 0.9450571933314758,
"delta_retention_norm_ratio": 0.9285361575906182,
"delta_cosine": 0.7459673717009906,
"relative_weight_rmse": 0.012384201248855871
},
{
"index": 50,
"total": 252,
"projection": "layers.15.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4925,
"parameters": 50331648,
"reference_delta_l2": 2.3721677759038755,
"retained_delta_l2": 1.9737182720217867,
"delta_retention_norm_ratio": 0.8320314827941433,
"delta_cosine": 0.6481676122790649,
"relative_weight_rmse": 0.010303816654448563
},
{
"index": 51,
"total": 252,
"projection": "layers.15.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4459,
"parameters": 50331648,
"reference_delta_l2": 2.885887968297399,
"retained_delta_l2": 2.527501575799674,
"delta_retention_norm_ratio": 0.8758141700458442,
"delta_cosine": 0.6876673832554954,
"relative_weight_rmse": 0.012055019327504342
},
{
"index": 52,
"total": 252,
"projection": "layers.15.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4131,
"parameters": 50331648,
"reference_delta_l2": 2.788548419905295,
"retained_delta_l2": 2.439712186071547,
"delta_retention_norm_ratio": 0.8749040069221408,
"delta_cosine": 0.6892488236129848,
"relative_weight_rmse": 0.011425457069043458
},
{
"index": 53,
"total": 252,
"projection": "layers.15.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0297,
"parameters": 4194304,
"reference_delta_l2": 0.8150906297976631,
"retained_delta_l2": 0.7209551611597461,
"delta_retention_norm_ratio": 0.8845092003311521,
"delta_cosine": 0.6993573726320066,
"relative_weight_rmse": 0.01159945529282478
},
{
"index": 54,
"total": 252,
"projection": "layers.15.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1349,
"parameters": 16777216,
"reference_delta_l2": 2.150413839269638,
"retained_delta_l2": 2.0665463942650666,
"delta_retention_norm_ratio": 0.9609993930130881,
"delta_cosine": 0.7886233827455329,
"relative_weight_rmse": 0.013839864673817552
},
{
"index": 55,
"total": 252,
"projection": "layers.15.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1364,
"parameters": 16777216,
"reference_delta_l2": 2.236995179184464,
"retained_delta_l2": 2.161700409143626,
"delta_retention_norm_ratio": 0.966341112067891,
"delta_cosine": 0.8010396883445619,
"relative_weight_rmse": 0.013451670834537348
},
{
"index": 56,
"total": 252,
"projection": "layers.15.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0296,
"parameters": 4194304,
"reference_delta_l2": 0.8594061384724814,
"retained_delta_l2": 0.7642473503487832,
"delta_retention_norm_ratio": 0.889273785857715,
"delta_cosine": 0.7056846810364582,
"relative_weight_rmse": 0.011236711800991986
},
{
"index": 57,
"total": 252,
"projection": "layers.16.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4738,
"parameters": 50331648,
"reference_delta_l2": 2.4930235792014797,
"retained_delta_l2": 2.130485476790213,
"delta_retention_norm_ratio": 0.8545789516650346,
"delta_cosine": 0.6693151771879681,
"relative_weight_rmse": 0.010670956495739464
},
{
"index": 58,
"total": 252,
"projection": "layers.16.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4504,
"parameters": 50331648,
"reference_delta_l2": 2.6448835575655862,
"retained_delta_l2": 2.2627192584541618,
"delta_retention_norm_ratio": 0.8555080816248949,
"delta_cosine": 0.6689483189234104,
"relative_weight_rmse": 0.011086619571348317
},
{
"index": 59,
"total": 252,
"projection": "layers.16.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4331,
"parameters": 50331648,
"reference_delta_l2": 2.5580250090443037,
"retained_delta_l2": 2.156664488486226,
"delta_retention_norm_ratio": 0.8430974993837027,
"delta_cosine": 0.6571866075161675,
"relative_weight_rmse": 0.010892383316172472
},
{
"index": 60,
"total": 252,
"projection": "layers.16.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0295,
"parameters": 4194304,
"reference_delta_l2": 0.7411812188779322,
"retained_delta_l2": 0.6387835350721942,
"delta_retention_norm_ratio": 0.8618452799427959,
"delta_cosine": 0.6757733236841931,
"relative_weight_rmse": 0.01122559452567028
},
{
"index": 61,
"total": 252,
"projection": "layers.16.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1351,
"parameters": 16777216,
"reference_delta_l2": 1.6676004895832297,
"retained_delta_l2": 1.4916106640837614,
"delta_retention_norm_ratio": 0.8944652351694546,
"delta_cosine": 0.7095360431433377,
"relative_weight_rmse": 0.011943781006030541
},
{
"index": 62,
"total": 252,
"projection": "layers.16.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1364,
"parameters": 16777216,
"reference_delta_l2": 1.8119028491426594,
"retained_delta_l2": 1.6801371797614963,
"delta_retention_norm_ratio": 0.9272777403912628,
"delta_cosine": 0.7482249181790871,
"relative_weight_rmse": 0.012470701690828164
},
{
"index": 63,
"total": 252,
"projection": "layers.16.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0315,
"parameters": 4194304,
"reference_delta_l2": 0.9582364370676199,
"retained_delta_l2": 0.87490151222013,
"delta_retention_norm_ratio": 0.9130330243937393,
"delta_cosine": 0.7299536933319858,
"relative_weight_rmse": 0.01214621210287195
},
{
"index": 64,
"total": 252,
"projection": "layers.17.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4535,
"parameters": 50331648,
"reference_delta_l2": 2.5840454993310598,
"retained_delta_l2": 2.223943997042893,
"delta_retention_norm_ratio": 0.8606442872691724,
"delta_cosine": 0.6751323651503068,
"relative_weight_rmse": 0.01093846525343687
},
{
"index": 65,
"total": 252,
"projection": "layers.17.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.443,
"parameters": 50331648,
"reference_delta_l2": 3.551904640019657,
"retained_delta_l2": 3.3329826092274595,
"delta_retention_norm_ratio": 0.9383648906770803,
"delta_cosine": 0.757284076462378,
"relative_weight_rmse": 0.013327264383487948
},
{
"index": 66,
"total": 252,
"projection": "layers.17.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4495,
"parameters": 50331648,
"reference_delta_l2": 3.1897256979680866,
"retained_delta_l2": 2.906337224320494,
"delta_retention_norm_ratio": 0.9111558483451677,
"delta_cosine": 0.7257837639496658,
"relative_weight_rmse": 0.012391958123551053
},
{
"index": 67,
"total": 252,
"projection": "layers.17.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0349,
"parameters": 4194304,
"reference_delta_l2": 0.7884564575504769,
"retained_delta_l2": 0.705981063949594,
"delta_retention_norm_ratio": 0.8953963877002012,
"delta_cosine": 0.715157686613299,
"relative_weight_rmse": 0.011568193321092414
},
{
"index": 68,
"total": 252,
"projection": "layers.17.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1464,
"parameters": 16777216,
"reference_delta_l2": 1.7305134615854225,
"retained_delta_l2": 1.5706161965637389,
"delta_retention_norm_ratio": 0.9076012590649295,
"delta_cosine": 0.7248156591416853,
"relative_weight_rmse": 0.012203630585911342
},
{
"index": 69,
"total": 252,
"projection": "layers.17.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1388,
"parameters": 16777216,
"reference_delta_l2": 1.6612425258077375,
"retained_delta_l2": 1.4864202187802729,
"delta_retention_norm_ratio": 0.8947641272652459,
"delta_cosine": 0.7077156015337548,
"relative_weight_rmse": 0.011972506647275472
},
{
"index": 70,
"total": 252,
"projection": "layers.17.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0297,
"parameters": 4194304,
"reference_delta_l2": 1.032566441357009,
"retained_delta_l2": 0.968218402599325,
"delta_retention_norm_ratio": 0.9376814544998023,
"delta_cosine": 0.758859138308178,
"relative_weight_rmse": 0.0127529417259923
},
{
"index": 71,
"total": 252,
"projection": "layers.18.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4472,
"parameters": 50331648,
"reference_delta_l2": 2.913347320433776,
"retained_delta_l2": 2.6250599650914257,
"delta_retention_norm_ratio": 0.9010460052873387,
"delta_cosine": 0.7155444618294795,
"relative_weight_rmse": 0.011724101481501378
},
{
"index": 72,
"total": 252,
"projection": "layers.18.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4096,
"parameters": 50331648,
"reference_delta_l2": 4.750153548686225,
"retained_delta_l2": 4.714669840150942,
"delta_retention_norm_ratio": 0.992529987047451,
"delta_cosine": 0.8280802011839735,
"relative_weight_rmse": 0.015469915779854649
},
{
"index": 73,
"total": 252,
"projection": "layers.18.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4504,
"parameters": 50331648,
"reference_delta_l2": 3.26976408192859,
"retained_delta_l2": 3.008785337794287,
"delta_retention_norm_ratio": 0.9201842280986917,
"delta_cosine": 0.7367723514156349,
"relative_weight_rmse": 0.012494491616577707
},
{
"index": 74,
"total": 252,
"projection": "layers.18.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0331,
"parameters": 4194304,
"reference_delta_l2": 1.1195566703162638,
"retained_delta_l2": 1.0864960588675219,
"delta_retention_norm_ratio": 0.9704699080222507,
"delta_cosine": 0.7994249936790332,
"relative_weight_rmse": 0.014009742872822967
},
{
"index": 75,
"total": 252,
"projection": "layers.18.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1486,
"parameters": 16777216,
"reference_delta_l2": 1.6280844154483574,
"retained_delta_l2": 1.4549600446401627,
"delta_retention_norm_ratio": 0.8936637626615214,
"delta_cosine": 0.7106452115577518,
"relative_weight_rmse": 0.011695022312579609
},
{
"index": 76,
"total": 252,
"projection": "layers.18.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1502,
"parameters": 16777216,
"reference_delta_l2": 2.100142714044375,
"retained_delta_l2": 2.0099828567706464,
"delta_retention_norm_ratio": 0.9570696521380195,
"delta_cosine": 0.7837234494328351,
"relative_weight_rmse": 0.01354640255918863
},
{
"index": 77,
"total": 252,
"projection": "layers.18.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0326,
"parameters": 4194304,
"reference_delta_l2": 0.927574166293776,
"retained_delta_l2": 0.8447567768543999,
"delta_retention_norm_ratio": 0.9107161535445925,
"delta_cosine": 0.7277085893672448,
"relative_weight_rmse": 0.011783651543010195
},
{
"index": 78,
"total": 252,
"projection": "layers.19.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4591,
"parameters": 50331648,
"reference_delta_l2": 2.692011605447208,
"retained_delta_l2": 2.355432322570745,
"delta_retention_norm_ratio": 0.8749710877191599,
"delta_cosine": 0.6891561134291521,
"relative_weight_rmse": 0.011205198335274374
},
{
"index": 79,
"total": 252,
"projection": "layers.19.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.406,
"parameters": 50331648,
"reference_delta_l2": 4.628244905524166,
"retained_delta_l2": 4.562739986385441,
"delta_retention_norm_ratio": 0.985846704209507,
"delta_cosine": 0.8190864196359027,
"relative_weight_rmse": 0.015469006883834856
},
{
"index": 80,
"total": 252,
"projection": "layers.19.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4213,
"parameters": 50331648,
"reference_delta_l2": 3.09288783776494,
"retained_delta_l2": 2.7732062936259445,
"delta_retention_norm_ratio": 0.8966397875035741,
"delta_cosine": 0.7109099239438775,
"relative_weight_rmse": 0.012198532314079335
},
{
"index": 81,
"total": 252,
"projection": "layers.19.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0328,
"parameters": 4194304,
"reference_delta_l2": 1.0454850269513585,
"retained_delta_l2": 1.0064490118920915,
"delta_retention_norm_ratio": 0.962662291613017,
"delta_cosine": 0.7954443785124893,
"relative_weight_rmse": 0.01359519753471669
},
{
"index": 82,
"total": 252,
"projection": "layers.19.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1493,
"parameters": 16777216,
"reference_delta_l2": 1.6403320823957444,
"retained_delta_l2": 1.4608275797759906,
"delta_retention_norm_ratio": 0.8905681937540458,
"delta_cosine": 0.7053103295267843,
"relative_weight_rmse": 0.011796377908480253
},
{
"index": 83,
"total": 252,
"projection": "layers.19.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1491,
"parameters": 16777216,
"reference_delta_l2": 1.9374795362107542,
"retained_delta_l2": 1.8280703427592173,
"delta_retention_norm_ratio": 0.9435301424315866,
"delta_cosine": 0.7658346811372263,
"relative_weight_rmse": 0.012827083950192922
},
{
"index": 84,
"total": 252,
"projection": "layers.19.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0326,
"parameters": 4194304,
"reference_delta_l2": 1.2369275511304956,
"retained_delta_l2": 1.2036434323148555,
"delta_retention_norm_ratio": 0.9730912948092878,
"delta_cosine": 0.8058833670591101,
"relative_weight_rmse": 0.013881014319838729
},
{
"index": 85,
"total": 252,
"projection": "layers.2.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.501,
"parameters": 50331648,
"reference_delta_l2": 1.5508441672353184,
"retained_delta_l2": 1.335700467029033,
"delta_retention_norm_ratio": 0.861273166736139,
"delta_cosine": 0.6979921164826142,
"relative_weight_rmse": 0.010106259099975305
},
{
"index": 86,
"total": 252,
"projection": "layers.2.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4053,
"parameters": 50331648,
"reference_delta_l2": 2.5081597069110946,
"retained_delta_l2": 2.36494542927979,
"delta_retention_norm_ratio": 0.9429006545170605,
"delta_cosine": 0.7973663223586619,
"relative_weight_rmse": 0.01265685274849115
},
{
"index": 87,
"total": 252,
"projection": "layers.2.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4141,
"parameters": 50331648,
"reference_delta_l2": 3.371090580531127,
"retained_delta_l2": 3.3512617982924966,
"delta_retention_norm_ratio": 0.9941179918590304,
"delta_cosine": 0.8795441583090302,
"relative_weight_rmse": 0.014953401247200468
},
{
"index": 88,
"total": 252,
"projection": "layers.2.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0298,
"parameters": 4194304,
"reference_delta_l2": 0.5772349439854766,
"retained_delta_l2": 0.4342798550481872,
"delta_retention_norm_ratio": 0.7523450539042796,
"delta_cosine": 0.5761350657444236,
"relative_weight_rmse": 0.009291282473771919
},
{
"index": 89,
"total": 252,
"projection": "layers.2.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1336,
"parameters": 16777216,
"reference_delta_l2": 2.0872068211190125,
"retained_delta_l2": 1.9691445631955444,
"delta_retention_norm_ratio": 0.9434352855074652,
"delta_cosine": 0.7607495118990254,
"relative_weight_rmse": 0.01395110559760262
},
{
"index": 90,
"total": 252,
"projection": "layers.2.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.147,
"parameters": 16777216,
"reference_delta_l2": 1.2301895454431324,
"retained_delta_l2": 0.9719865040390575,
"delta_retention_norm_ratio": 0.790111172411999,
"delta_cosine": 0.6082995164165952,
"relative_weight_rmse": 0.010189354680498844
},
{
"index": 91,
"total": 252,
"projection": "layers.2.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0324,
"parameters": 4194304,
"reference_delta_l2": 1.161429363217796,
"retained_delta_l2": 1.120576061810203,
"delta_retention_norm_ratio": 0.9648249797177445,
"delta_cosine": 0.7933560052433367,
"relative_weight_rmse": 0.013514842713471316
},
{
"index": 92,
"total": 252,
"projection": "layers.20.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4905,
"parameters": 50331648,
"reference_delta_l2": 2.648971079678172,
"retained_delta_l2": 2.2615954100110773,
"delta_retention_norm_ratio": 0.8537637225869005,
"delta_cosine": 0.6661344051556789,
"relative_weight_rmse": 0.011257322513998143
},
{
"index": 93,
"total": 252,
"projection": "layers.20.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4511,
"parameters": 50331648,
"reference_delta_l2": 4.561010493321236,
"retained_delta_l2": 4.494147537265125,
"delta_retention_norm_ratio": 0.9853403196168876,
"delta_cosine": 0.817557901769122,
"relative_weight_rmse": 0.015447725537274087
},
{
"index": 94,
"total": 252,
"projection": "layers.20.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4539,
"parameters": 50331648,
"reference_delta_l2": 3.12668177773168,
"retained_delta_l2": 2.809176965998698,
"delta_retention_norm_ratio": 0.898453109621113,
"delta_cosine": 0.7122847833548339,
"relative_weight_rmse": 0.012279056278587662
},
{
"index": 95,
"total": 252,
"projection": "layers.20.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0327,
"parameters": 4194304,
"reference_delta_l2": 0.8444102655934065,
"retained_delta_l2": 0.7766709699973873,
"delta_retention_norm_ratio": 0.919779166175324,
"delta_cosine": 0.7402847978305606,
"relative_weight_rmse": 0.012200507561191239
},
{
"index": 96,
"total": 252,
"projection": "layers.20.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1492,
"parameters": 16777216,
"reference_delta_l2": 1.6104533999515782,
"retained_delta_l2": 1.429579822089707,
"delta_retention_norm_ratio": 0.8876877916074382,
"delta_cosine": 0.7036032966464911,
"relative_weight_rmse": 0.01182859582637794
},
{
"index": 97,
"total": 252,
"projection": "layers.20.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1501,
"parameters": 16777216,
"reference_delta_l2": 1.8925005510995239,
"retained_delta_l2": 1.7685062082408078,
"delta_retention_norm_ratio": 0.9344812117562256,
"delta_cosine": 0.7569142446469062,
"relative_weight_rmse": 0.012339049211212694
},
{
"index": 98,
"total": 252,
"projection": "layers.20.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0326,
"parameters": 4194304,
"reference_delta_l2": 0.9444975148597435,
"retained_delta_l2": 0.8667351580621327,
"delta_retention_norm_ratio": 0.9176680133360029,
"delta_cosine": 0.7369691181403826,
"relative_weight_rmse": 0.012025658690495607
},
{
"index": 99,
"total": 252,
"projection": "layers.21.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4553,
"parameters": 50331648,
"reference_delta_l2": 3.107221181231323,
"retained_delta_l2": 2.805385115320624,
"delta_retention_norm_ratio": 0.9028598067836652,
"delta_cosine": 0.7159459913630684,
"relative_weight_rmse": 0.01231006933559084
},
{
"index": 100,
"total": 252,
"projection": "layers.21.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4087,
"parameters": 50331648,
"reference_delta_l2": 4.642512747803093,
"retained_delta_l2": 4.581987771039912,
"delta_retention_norm_ratio": 0.986962884099387,
"delta_cosine": 0.8176793716730177,
"relative_weight_rmse": 0.01594965493365005
},
{
"index": 101,
"total": 252,
"projection": "layers.21.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4381,
"parameters": 50331648,
"reference_delta_l2": 3.345569132582998,
"retained_delta_l2": 3.0728298316599965,
"delta_retention_norm_ratio": 0.9184774577614454,
"delta_cosine": 0.7334441165497892,
"relative_weight_rmse": 0.012769331642977035
},
{
"index": 102,
"total": 252,
"projection": "layers.21.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0328,
"parameters": 4194304,
"reference_delta_l2": 0.8892310195003545,
"retained_delta_l2": 0.8221545518506864,
"delta_retention_norm_ratio": 0.9245680074370805,
"delta_cosine": 0.7420388118138687,
"relative_weight_rmse": 0.012823663977531489
},
{
"index": 103,
"total": 252,
"projection": "layers.21.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1495,
"parameters": 16777216,
"reference_delta_l2": 1.7389603341955504,
"retained_delta_l2": 1.5906396850981026,
"delta_retention_norm_ratio": 0.9147072844729023,
"delta_cosine": 0.7325424592926252,
"relative_weight_rmse": 0.012393883456267903
},
{
"index": 104,
"total": 252,
"projection": "layers.21.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1503,
"parameters": 16777216,
"reference_delta_l2": 1.3476164957119525,
"retained_delta_l2": 1.1147654080362608,
"delta_retention_norm_ratio": 0.8272126466122875,
"delta_cosine": 0.64388309677043,
"relative_weight_rmse": 0.010337739938509697
},
{
"index": 105,
"total": 252,
"projection": "layers.21.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0328,
"parameters": 4194304,
"reference_delta_l2": 1.3434894364701484,
"retained_delta_l2": 1.3268205530828734,
"delta_retention_norm_ratio": 0.987592843728589,
"delta_cosine": 0.8236554435647476,
"relative_weight_rmse": 0.014577137951733116
},
{
"index": 106,
"total": 252,
"projection": "layers.22.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4918,
"parameters": 50331648,
"reference_delta_l2": 3.3609458983357805,
"retained_delta_l2": 3.125339013124354,
"delta_retention_norm_ratio": 0.9298986379613875,
"delta_cosine": 0.7474970101070157,
"relative_weight_rmse": 0.01260720733761587
},
{
"index": 107,
"total": 252,
"projection": "layers.22.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4105,
"parameters": 50331648,
"reference_delta_l2": 4.221243064006641,
"retained_delta_l2": 4.107698603498147,
"delta_retention_norm_ratio": 0.9731016530470239,
"delta_cosine": 0.8024007502803084,
"relative_weight_rmse": 0.014672423527385866
},
{
"index": 108,
"total": 252,
"projection": "layers.22.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4078,
"parameters": 50331648,
"reference_delta_l2": 3.4270118520156996,
"retained_delta_l2": 3.1761527162184167,
"delta_retention_norm_ratio": 0.9267994548516858,
"delta_cosine": 0.7449636622645867,
"relative_weight_rmse": 0.012623632707227259
},
{
"index": 109,
"total": 252,
"projection": "layers.22.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0296,
"parameters": 4194304,
"reference_delta_l2": 1.1248690479543206,
"retained_delta_l2": 1.0996769913538333,
"delta_retention_norm_ratio": 0.9776044539172792,
"delta_cosine": 0.8174457421381668,
"relative_weight_rmse": 0.01441873674870668
},
{
"index": 110,
"total": 252,
"projection": "layers.22.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1472,
"parameters": 16777216,
"reference_delta_l2": 2.0985495045774676,
"retained_delta_l2": 1.9981921338156807,
"delta_retention_norm_ratio": 0.9521777444168547,
"delta_cosine": 0.7728400422923889,
"relative_weight_rmse": 0.013733150895618145
},
{
"index": 111,
"total": 252,
"projection": "layers.22.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1503,
"parameters": 16777216,
"reference_delta_l2": 1.8698448352484587,
"retained_delta_l2": 1.7404118969208338,
"delta_retention_norm_ratio": 0.9307787812722833,
"delta_cosine": 0.7545710882575692,
"relative_weight_rmse": 0.012594991420075953
},
{
"index": 112,
"total": 252,
"projection": "layers.22.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.033,
"parameters": 4194304,
"reference_delta_l2": 1.1046623590021176,
"retained_delta_l2": 1.0524088648410537,
"delta_retention_norm_ratio": 0.9526973163019093,
"delta_cosine": 0.777701212277148,
"relative_weight_rmse": 0.013230322424454686
},
{
"index": 113,
"total": 252,
"projection": "layers.23.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4927,
"parameters": 50331648,
"reference_delta_l2": 3.1620346530790826,
"retained_delta_l2": 2.902533392692175,
"delta_retention_norm_ratio": 0.9179321896000683,
"delta_cosine": 0.7352168014146778,
"relative_weight_rmse": 0.011978278641254893
},
{
"index": 114,
"total": 252,
"projection": "layers.23.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4521,
"parameters": 50331648,
"reference_delta_l2": 5.652564171547626,
"retained_delta_l2": 5.697962059642462,
"delta_retention_norm_ratio": 1.0080313795150435,
"delta_cosine": 0.8538603703413962,
"relative_weight_rmse": 0.016836271684099298
},
{
"index": 115,
"total": 252,
"projection": "layers.23.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.451,
"parameters": 50331648,
"reference_delta_l2": 4.992215728948909,
"retained_delta_l2": 4.956118796350239,
"delta_retention_norm_ratio": 0.9927693564223695,
"delta_cosine": 0.8293166817216231,
"relative_weight_rmse": 0.015429864395519641
},
{
"index": 116,
"total": 252,
"projection": "layers.23.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0326,
"parameters": 4194304,
"reference_delta_l2": 0.7943983009499747,
"retained_delta_l2": 0.7069701702596439,
"delta_retention_norm_ratio": 0.8899442123859272,
"delta_cosine": 0.7056078221082792,
"relative_weight_rmse": 0.011972629556969164
},
{
"index": 117,
"total": 252,
"projection": "layers.23.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1488,
"parameters": 16777216,
"reference_delta_l2": 1.9607945730186995,
"retained_delta_l2": 1.83389000152093,
"delta_retention_norm_ratio": 0.9352790071718752,
"delta_cosine": 0.7547406776303367,
"relative_weight_rmse": 0.012808459672303903
},
{
"index": 118,
"total": 252,
"projection": "layers.23.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1465,
"parameters": 16777216,
"reference_delta_l2": 1.7866254324795166,
"retained_delta_l2": 1.619051896159686,
"delta_retention_norm_ratio": 0.9062066769713065,
"delta_cosine": 0.7239721031076742,
"relative_weight_rmse": 0.012028747476560837
},
{
"index": 119,
"total": 252,
"projection": "layers.23.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0295,
"parameters": 4194304,
"reference_delta_l2": 0.9282231048745307,
"retained_delta_l2": 0.8524835192395849,
"delta_retention_norm_ratio": 0.9184036841604113,
"delta_cosine": 0.7378533914613947,
"relative_weight_rmse": 0.011736658596304246
},
{
"index": 120,
"total": 252,
"projection": "layers.24.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4507,
"parameters": 50331648,
"reference_delta_l2": 3.251424600129174,
"retained_delta_l2": 2.979207623216274,
"delta_retention_norm_ratio": 0.9162776289193095,
"delta_cosine": 0.7315679438953859,
"relative_weight_rmse": 0.012276154510434796
},
{
"index": 121,
"total": 252,
"projection": "layers.24.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4065,
"parameters": 50331648,
"reference_delta_l2": 5.142605585684486,
"retained_delta_l2": 5.113479421584231,
"delta_retention_norm_ratio": 0.994336302169209,
"delta_cosine": 0.8280169487865294,
"relative_weight_rmse": 0.016274551388381405
},
{
"index": 122,
"total": 252,
"projection": "layers.24.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4461,
"parameters": 50331648,
"reference_delta_l2": 4.012708108272533,
"retained_delta_l2": 3.8350650966304483,
"delta_retention_norm_ratio": 0.955729894413237,
"delta_cosine": 0.7757892948718432,
"relative_weight_rmse": 0.013955994131055856
},
{
"index": 123,
"total": 252,
"projection": "layers.24.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0327,
"parameters": 4194304,
"reference_delta_l2": 0.9668094318472028,
"retained_delta_l2": 0.9173263382465839,
"delta_retention_norm_ratio": 0.9488181517777753,
"delta_cosine": 0.7725211199230653,
"relative_weight_rmse": 0.01315692901419039
},
{
"index": 124,
"total": 252,
"projection": "layers.24.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1476,
"parameters": 16777216,
"reference_delta_l2": 1.9194580926397542,
"retained_delta_l2": 1.7805052619422008,
"delta_retention_norm_ratio": 0.9276083019314805,
"delta_cosine": 0.7462979525655402,
"relative_weight_rmse": 0.01258575114768902
},
{
"index": 125,
"total": 252,
"projection": "layers.24.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1518,
"parameters": 16777216,
"reference_delta_l2": 3.03040691525353,
"retained_delta_l2": 3.0277198894603696,
"delta_retention_norm_ratio": 0.9991133118857288,
"delta_cosine": 0.8491828963479011,
"relative_weight_rmse": 0.015582367835321543
},
{
"index": 126,
"total": 252,
"projection": "layers.24.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0329,
"parameters": 4194304,
"reference_delta_l2": 1.0881418374460758,
"retained_delta_l2": 1.0367986053596372,
"delta_retention_norm_ratio": 0.9528156805302664,
"delta_cosine": 0.7790750324231591,
"relative_weight_rmse": 0.01305306389348348
},
{
"index": 127,
"total": 252,
"projection": "layers.25.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4742,
"parameters": 50331648,
"reference_delta_l2": 3.639019471804408,
"retained_delta_l2": 3.4060651741134307,
"delta_retention_norm_ratio": 0.9359843222890294,
"delta_cosine": 0.7515604382084587,
"relative_weight_rmse": 0.013267491353305256
},
{
"index": 128,
"total": 252,
"projection": "layers.25.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4063,
"parameters": 50331648,
"reference_delta_l2": 5.615149002693838,
"retained_delta_l2": 5.630635217384503,
"delta_retention_norm_ratio": 1.0027579347730997,
"delta_cosine": 0.8411372051553513,
"relative_weight_rmse": 0.01697377661460169
},
{
"index": 129,
"total": 252,
"projection": "layers.25.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4111,
"parameters": 50331648,
"reference_delta_l2": 3.7478251188093408,
"retained_delta_l2": 3.498347787523113,
"delta_retention_norm_ratio": 0.9334341055471965,
"delta_cosine": 0.7484383798647123,
"relative_weight_rmse": 0.01362706679316093
},
{
"index": 130,
"total": 252,
"projection": "layers.25.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0326,
"parameters": 4194304,
"reference_delta_l2": 1.0284858292340313,
"retained_delta_l2": 0.9820655007569365,
"delta_retention_norm_ratio": 0.954865368916491,
"delta_cosine": 0.7799315686869575,
"relative_weight_rmse": 0.013543057075537882
},
{
"index": 131,
"total": 252,
"projection": "layers.25.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1493,
"parameters": 16777216,
"reference_delta_l2": 1.955958374402271,
"retained_delta_l2": 1.8090541262395137,
"delta_retention_norm_ratio": 0.9248939803191617,
"delta_cosine": 0.740241619895826,
"relative_weight_rmse": 0.013198429915754848
},
{
"index": 132,
"total": 252,
"projection": "layers.25.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.151,
"parameters": 16777216,
"reference_delta_l2": 2.4489920917333023,
"retained_delta_l2": 2.4017239487341233,
"delta_retention_norm_ratio": 0.9806989401236799,
"delta_cosine": 0.8151070374507796,
"relative_weight_rmse": 0.01459235557382001
},
{
"index": 133,
"total": 252,
"projection": "layers.25.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0333,
"parameters": 4194304,
"reference_delta_l2": 1.1884885257253468,
"retained_delta_l2": 1.145197026679363,
"delta_retention_norm_ratio": 0.9635743230928017,
"delta_cosine": 0.787303539246883,
"relative_weight_rmse": 0.013460904970252758
},
{
"index": 134,
"total": 252,
"projection": "layers.26.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.491,
"parameters": 50331648,
"reference_delta_l2": 3.2985177852649774,
"retained_delta_l2": 3.0201852227440886,
"delta_retention_norm_ratio": 0.915618898959937,
"delta_cosine": 0.7285283299851003,
"relative_weight_rmse": 0.012344082698543684
},
{
"index": 135,
"total": 252,
"projection": "layers.26.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4149,
"parameters": 50331648,
"reference_delta_l2": 5.262324969323531,
"retained_delta_l2": 5.225001115373795,
"delta_retention_norm_ratio": 0.9929073452956034,
"delta_cosine": 0.8222423912762875,
"relative_weight_rmse": 0.016659485781630477
},
{
"index": 136,
"total": 252,
"projection": "layers.26.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4182,
"parameters": 50331648,
"reference_delta_l2": 3.601875439552374,
"retained_delta_l2": 3.3299843645601643,
"delta_retention_norm_ratio": 0.9245140262190745,
"delta_cosine": 0.738296741300208,
"relative_weight_rmse": 0.01317555515652475
},
{
"index": 137,
"total": 252,
"projection": "layers.26.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0299,
"parameters": 4194304,
"reference_delta_l2": 1.10276341844642,
"retained_delta_l2": 1.0644083817619203,
"delta_retention_norm_ratio": 0.965219161206368,
"delta_cosine": 0.7914008934281718,
"relative_weight_rmse": 0.014444454066023087
},
{
"index": 138,
"total": 252,
"projection": "layers.26.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1357,
"parameters": 16777216,
"reference_delta_l2": 2.145819257120039,
"retained_delta_l2": 2.021478643463708,
"delta_retention_norm_ratio": 0.9420544795449307,
"delta_cosine": 0.7604333020441107,
"relative_weight_rmse": 0.013812336892443362
},
{
"index": 139,
"total": 252,
"projection": "layers.26.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1447,
"parameters": 16777216,
"reference_delta_l2": 2.7138340521627686,
"retained_delta_l2": 2.689784374224837,
"delta_retention_norm_ratio": 0.9911381177051832,
"delta_cosine": 0.8319858180666069,
"relative_weight_rmse": 0.01536621409453747
},
{
"index": 140,
"total": 252,
"projection": "layers.26.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0329,
"parameters": 4194304,
"reference_delta_l2": 1.3017342452096916,
"retained_delta_l2": 1.266469377332241,
"delta_retention_norm_ratio": 0.9729093184670962,
"delta_cosine": 0.8070487335221931,
"relative_weight_rmse": 0.014261161588877415
},
{
"index": 141,
"total": 252,
"projection": "layers.27.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4925,
"parameters": 50331648,
"reference_delta_l2": 3.124715513558961,
"retained_delta_l2": 2.7984383043710457,
"delta_retention_norm_ratio": 0.8955817872788377,
"delta_cosine": 0.7080091778414741,
"relative_weight_rmse": 0.011915628934201969
},
{
"index": 142,
"total": 252,
"projection": "layers.27.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4508,
"parameters": 50331648,
"reference_delta_l2": 5.237959926379879,
"retained_delta_l2": 5.206038113975948,
"delta_retention_norm_ratio": 0.9939056783838375,
"delta_cosine": 0.8252827314307719,
"relative_weight_rmse": 0.016469591622974865
},
{
"index": 143,
"total": 252,
"projection": "layers.27.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4563,
"parameters": 50331648,
"reference_delta_l2": 3.146552071698491,
"retained_delta_l2": 2.7636743761780713,
"delta_retention_norm_ratio": 0.8783183348643125,
"delta_cosine": 0.6900722094746743,
"relative_weight_rmse": 0.012191736525387142
},
{
"index": 144,
"total": 252,
"projection": "layers.27.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0334,
"parameters": 4194304,
"reference_delta_l2": 0.9581121990064987,
"retained_delta_l2": 0.9019702495484059,
"delta_retention_norm_ratio": 0.941403575159247,
"delta_cosine": 0.7665746823434738,
"relative_weight_rmse": 0.01350311211150427
},
{
"index": 145,
"total": 252,
"projection": "layers.27.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1498,
"parameters": 16777216,
"reference_delta_l2": 1.8355717950441208,
"retained_delta_l2": 1.670861751372117,
"delta_retention_norm_ratio": 0.9102677192378383,
"delta_cosine": 0.7241152563856859,
"relative_weight_rmse": 0.012750278893097801
},
{
"index": 146,
"total": 252,
"projection": "layers.27.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1508,
"parameters": 16777216,
"reference_delta_l2": 2.5050098143817583,
"retained_delta_l2": 2.45045837460127,
"delta_retention_norm_ratio": 0.9782230634517687,
"delta_cosine": 0.813365238440355,
"relative_weight_rmse": 0.014895028071551018
},
{
"index": 147,
"total": 252,
"projection": "layers.27.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0327,
"parameters": 4194304,
"reference_delta_l2": 1.3040544961619782,
"retained_delta_l2": 1.28054596555992,
"delta_retention_norm_ratio": 0.9819727391215265,
"delta_cosine": 0.8139855038447851,
"relative_weight_rmse": 0.01430637569344675
},
{
"index": 148,
"total": 252,
"projection": "layers.28.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4956,
"parameters": 50331648,
"reference_delta_l2": 3.764845141674202,
"retained_delta_l2": 3.5504303943634237,
"delta_retention_norm_ratio": 0.9430481894361717,
"delta_cosine": 0.7611944338026825,
"relative_weight_rmse": 0.013106892823588578
},
{
"index": 149,
"total": 252,
"projection": "layers.28.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4596,
"parameters": 50331648,
"reference_delta_l2": 6.442421034070186,
"retained_delta_l2": 6.5579984602763775,
"delta_retention_norm_ratio": 1.0179400609794005,
"delta_cosine": 0.8673769877438279,
"relative_weight_rmse": 0.01796578548094511
},
{
"index": 150,
"total": 252,
"projection": "layers.28.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.5438,
"parameters": 50331648,
"reference_delta_l2": 3.2466748504913374,
"retained_delta_l2": 2.879382216819249,
"delta_retention_norm_ratio": 0.88687113721397,
"delta_cosine": 0.6981195835720675,
"relative_weight_rmse": 0.012332533245406866
},
{
"index": 151,
"total": 252,
"projection": "layers.28.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0469,
"parameters": 4194304,
"reference_delta_l2": 1.0322561434007247,
"retained_delta_l2": 0.9824991241737554,
"delta_retention_norm_ratio": 0.9517977979156929,
"delta_cosine": 0.7711953547626353,
"relative_weight_rmse": 0.014134794851969812
},
{
"index": 152,
"total": 252,
"projection": "layers.28.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1937,
"parameters": 16777216,
"reference_delta_l2": 1.8686280100518486,
"retained_delta_l2": 1.7172346752833605,
"delta_retention_norm_ratio": 0.9189815554759412,
"delta_cosine": 0.7339688363690187,
"relative_weight_rmse": 0.01283825487100208
},
{
"index": 153,
"total": 252,
"projection": "layers.28.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1953,
"parameters": 16777216,
"reference_delta_l2": 2.109136925707816,
"retained_delta_l2": 2.009608517801217,
"delta_retention_norm_ratio": 0.9528108361797338,
"delta_cosine": 0.7772529210335036,
"relative_weight_rmse": 0.013257235238862845
},
{
"index": 154,
"total": 252,
"projection": "layers.28.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0438,
"parameters": 4194304,
"reference_delta_l2": 1.0576402598291665,
"retained_delta_l2": 0.9884129021297963,
"delta_retention_norm_ratio": 0.9345454590480963,
"delta_cosine": 0.7520063025712015,
"relative_weight_rmse": 0.012589367209166029
},
{
"index": 155,
"total": 252,
"projection": "layers.29.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.6475,
"parameters": 50331648,
"reference_delta_l2": 3.744595072360419,
"retained_delta_l2": 3.5146251679139366,
"delta_retention_norm_ratio": 0.9385861755403316,
"delta_cosine": 0.7561840841964639,
"relative_weight_rmse": 0.013006904610922267
},
{
"index": 156,
"total": 252,
"projection": "layers.29.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.5817,
"parameters": 50331648,
"reference_delta_l2": 6.146902145327938,
"retained_delta_l2": 6.232910373596634,
"delta_retention_norm_ratio": 1.0139921258278153,
"delta_cosine": 0.8574046352520125,
"relative_weight_rmse": 0.017737338865605447
},
{
"index": 157,
"total": 252,
"projection": "layers.29.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4498,
"parameters": 50331648,
"reference_delta_l2": 3.585418074060757,
"retained_delta_l2": 3.285923219441303,
"delta_retention_norm_ratio": 0.9164686381244647,
"delta_cosine": 0.7297235008349721,
"relative_weight_rmse": 0.012909084892325087
},
{
"index": 158,
"total": 252,
"projection": "layers.29.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0328,
"parameters": 4194304,
"reference_delta_l2": 0.961986646946259,
"retained_delta_l2": 0.9117233578831268,
"delta_retention_norm_ratio": 0.9477505335207213,
"delta_cosine": 0.7818085754795856,
"relative_weight_rmse": 0.013129984425673723
},
{
"index": 159,
"total": 252,
"projection": "layers.29.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.148,
"parameters": 16777216,
"reference_delta_l2": 1.9464019495769305,
"retained_delta_l2": 1.794594545398929,
"delta_retention_norm_ratio": 0.9220061384489477,
"delta_cosine": 0.7376071256868779,
"relative_weight_rmse": 0.01293469417484533
},
{
"index": 160,
"total": 252,
"projection": "layers.29.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1504,
"parameters": 16777216,
"reference_delta_l2": 2.9449799373554164,
"retained_delta_l2": 2.9323393211048,
"delta_retention_norm_ratio": 0.9957077411325364,
"delta_cosine": 0.8479247667602094,
"relative_weight_rmse": 0.015277724302880269
},
{
"index": 161,
"total": 252,
"projection": "layers.29.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0327,
"parameters": 4194304,
"reference_delta_l2": 1.0703222881461318,
"retained_delta_l2": 1.0110120744623348,
"delta_retention_norm_ratio": 0.9445865844889336,
"delta_cosine": 0.7677460449425708,
"relative_weight_rmse": 0.012960603686888467
},
{
"index": 162,
"total": 252,
"projection": "layers.3.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.459,
"parameters": 50331648,
"reference_delta_l2": 2.0816467687208644,
"retained_delta_l2": 1.846851025558511,
"delta_retention_norm_ratio": 0.8872067313770859,
"delta_cosine": 0.718505065436174,
"relative_weight_rmse": 0.01137977605185741
},
{
"index": 163,
"total": 252,
"projection": "layers.3.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4077,
"parameters": 50331648,
"reference_delta_l2": 2.928159352528277,
"retained_delta_l2": 2.70725959595232,
"delta_retention_norm_ratio": 0.9245601997769608,
"delta_cosine": 0.746213924942928,
"relative_weight_rmse": 0.012262553419182597
},
{
"index": 164,
"total": 252,
"projection": "layers.3.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4251,
"parameters": 50331648,
"reference_delta_l2": 3.1827874854332223,
"retained_delta_l2": 3.1170175872157957,
"delta_retention_norm_ratio": 0.9793357556800641,
"delta_cosine": 0.8427649198473434,
"relative_weight_rmse": 0.014675753770537979
},
{
"index": 165,
"total": 252,
"projection": "layers.3.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0329,
"parameters": 4194304,
"reference_delta_l2": 0.5952192957808071,
"retained_delta_l2": 0.4621196895078905,
"delta_retention_norm_ratio": 0.7763855990281416,
"delta_cosine": 0.5973615971078495,
"relative_weight_rmse": 0.009645398247060876
},
{
"index": 166,
"total": 252,
"projection": "layers.3.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1493,
"parameters": 16777216,
"reference_delta_l2": 1.8549823933345801,
"retained_delta_l2": 1.7019374827442797,
"delta_retention_norm_ratio": 0.9174952219814972,
"delta_cosine": 0.7310230556559157,
"relative_weight_rmse": 0.012835139449262132
},
{
"index": 167,
"total": 252,
"projection": "layers.3.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1503,
"parameters": 16777216,
"reference_delta_l2": 1.5109060405438957,
"retained_delta_l2": 1.3087133697306923,
"delta_retention_norm_ratio": 0.866177865871515,
"delta_cosine": 0.6794745953678916,
"relative_weight_rmse": 0.011738667585957624
},
{
"index": 168,
"total": 252,
"projection": "layers.3.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0329,
"parameters": 4194304,
"reference_delta_l2": 0.9483394800318138,
"retained_delta_l2": 0.8550751614686678,
"delta_retention_norm_ratio": 0.9016551345515877,
"delta_cosine": 0.7160606498127464,
"relative_weight_rmse": 0.012398764910472593
},
{
"index": 169,
"total": 252,
"projection": "layers.30.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4893,
"parameters": 50331648,
"reference_delta_l2": 3.890440954504071,
"retained_delta_l2": 3.690426575760282,
"delta_retention_norm_ratio": 0.9485882497426862,
"delta_cosine": 0.7668105183568503,
"relative_weight_rmse": 0.013143095688071201
},
{
"index": 170,
"total": 252,
"projection": "layers.30.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.407,
"parameters": 50331648,
"reference_delta_l2": 5.412388467981441,
"retained_delta_l2": 5.410813537592956,
"delta_retention_norm_ratio": 0.9997090137934848,
"delta_cosine": 0.8312169075321302,
"relative_weight_rmse": 0.016948679764562272
},
{
"index": 171,
"total": 252,
"projection": "layers.30.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4059,
"parameters": 50331648,
"reference_delta_l2": 3.820850325456324,
"retained_delta_l2": 3.5424926268605526,
"delta_retention_norm_ratio": 0.9271477093093073,
"delta_cosine": 0.7405393291006691,
"relative_weight_rmse": 0.01342737372482948
},
{
"index": 172,
"total": 252,
"projection": "layers.30.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0295,
"parameters": 4194304,
"reference_delta_l2": 1.4964166465326538,
"retained_delta_l2": 1.5085010902675233,
"delta_retention_norm_ratio": 1.0080755876131628,
"delta_cosine": 0.8529538137770175,
"relative_weight_rmse": 0.016610669367884554
},
{
"index": 173,
"total": 252,
"projection": "layers.30.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1393,
"parameters": 16777216,
"reference_delta_l2": 1.8708535812690759,
"retained_delta_l2": 1.6951518586696184,
"delta_retention_norm_ratio": 0.9060847281911437,
"delta_cosine": 0.7178714125470955,
"relative_weight_rmse": 0.012872821014317882
},
{
"index": 174,
"total": 252,
"projection": "layers.30.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1502,
"parameters": 16777216,
"reference_delta_l2": 3.0265541942692145,
"retained_delta_l2": 3.0382014095244996,
"delta_retention_norm_ratio": 1.0038483418791375,
"delta_cosine": 0.8555119585920778,
"relative_weight_rmse": 0.01631255720029809
},
{
"index": 175,
"total": 252,
"projection": "layers.30.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0345,
"parameters": 4194304,
"reference_delta_l2": 1.3106068679861238,
"retained_delta_l2": 1.2727688000592756,
"delta_retention_norm_ratio": 0.971129353240007,
"delta_cosine": 0.8016201059609241,
"relative_weight_rmse": 0.013927737576126836
},
{
"index": 176,
"total": 252,
"projection": "layers.31.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4924,
"parameters": 50331648,
"reference_delta_l2": 4.113798576311484,
"retained_delta_l2": 3.9484840660209626,
"delta_retention_norm_ratio": 0.9598146318484203,
"delta_cosine": 0.7806503092168832,
"relative_weight_rmse": 0.013386101092529485
},
{
"index": 177,
"total": 252,
"projection": "layers.31.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4347,
"parameters": 50331648,
"reference_delta_l2": 4.220864313283881,
"retained_delta_l2": 4.073609670229595,
"delta_retention_norm_ratio": 0.9651126802179243,
"delta_cosine": 0.786110534496433,
"relative_weight_rmse": 0.014770223849838454
},
{
"index": 178,
"total": 252,
"projection": "layers.31.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4032,
"parameters": 50331648,
"reference_delta_l2": 3.7486175185665345,
"retained_delta_l2": 3.461453822165797,
"delta_retention_norm_ratio": 0.9233947728786829,
"delta_cosine": 0.7369009849212661,
"relative_weight_rmse": 0.0131321468733143
},
{
"index": 179,
"total": 252,
"projection": "layers.31.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 1.1311475844923473,
"retained_delta_l2": 1.1077071383582915,
"delta_retention_norm_ratio": 0.9792772875481358,
"delta_cosine": 0.8130569202455515,
"relative_weight_rmse": 0.014430682084851742
},
{
"index": 180,
"total": 252,
"projection": "layers.31.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1338,
"parameters": 16777216,
"reference_delta_l2": 1.928581700638491,
"retained_delta_l2": 1.780116048227859,
"delta_retention_norm_ratio": 0.9230182198858986,
"delta_cosine": 0.7389795095759798,
"relative_weight_rmse": 0.01254654298808731
},
{
"index": 181,
"total": 252,
"projection": "layers.31.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1358,
"parameters": 16777216,
"reference_delta_l2": 3.0957887647369557,
"retained_delta_l2": 3.1192606103564167,
"delta_retention_norm_ratio": 1.0075818627830235,
"delta_cosine": 0.8593482849876289,
"relative_weight_rmse": 0.01636171596434865
},
{
"index": 182,
"total": 252,
"projection": "layers.31.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0296,
"parameters": 4194304,
"reference_delta_l2": 0.9148601426527398,
"retained_delta_l2": 0.8074669819466506,
"delta_retention_norm_ratio": 0.8826124828274945,
"delta_cosine": 0.6998511333234301,
"relative_weight_rmse": 0.011260691148171707
},
{
"index": 183,
"total": 252,
"projection": "layers.32.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.488,
"parameters": 50331648,
"reference_delta_l2": 3.055807821872593,
"retained_delta_l2": 2.7039779326575353,
"delta_retention_norm_ratio": 0.8848651781382453,
"delta_cosine": 0.6985091473251286,
"relative_weight_rmse": 0.011295839653652632
},
{
"index": 184,
"total": 252,
"projection": "layers.32.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4491,
"parameters": 50331648,
"reference_delta_l2": 5.754306238584233,
"retained_delta_l2": 5.807574715978076,
"delta_retention_norm_ratio": 1.0092571502428325,
"delta_cosine": 0.848981480575959,
"relative_weight_rmse": 0.017506381182025764
},
{
"index": 185,
"total": 252,
"projection": "layers.32.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4304,
"parameters": 50331648,
"reference_delta_l2": 4.185474965701407,
"retained_delta_l2": 3.9698100260580733,
"delta_retention_norm_ratio": 0.9484730069082633,
"delta_cosine": 0.7675864010193029,
"relative_weight_rmse": 0.013849912318123502
},
{
"index": 186,
"total": 252,
"projection": "layers.32.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0298,
"parameters": 4194304,
"reference_delta_l2": 1.0760679670383597,
"retained_delta_l2": 1.041675143509792,
"delta_retention_norm_ratio": 0.9680384282572536,
"delta_cosine": 0.7988605713507398,
"relative_weight_rmse": 0.014087585177892058
},
{
"index": 187,
"total": 252,
"projection": "layers.32.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1365,
"parameters": 16777216,
"reference_delta_l2": 1.9537101376652952,
"retained_delta_l2": 1.7957463648408594,
"delta_retention_norm_ratio": 0.9191467711718974,
"delta_cosine": 0.7327887015073427,
"relative_weight_rmse": 0.01291199405923289
},
{
"index": 188,
"total": 252,
"projection": "layers.32.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1374,
"parameters": 16777216,
"reference_delta_l2": 3.508291642540976,
"retained_delta_l2": 3.5679730543291797,
"delta_retention_norm_ratio": 1.0170115309299024,
"delta_cosine": 0.8839663164565524,
"relative_weight_rmse": 0.017021398109748054
},
{
"index": 189,
"total": 252,
"projection": "layers.32.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0298,
"parameters": 4194304,
"reference_delta_l2": 1.1061199086674303,
"retained_delta_l2": 1.0365557342178422,
"delta_retention_norm_ratio": 0.9371097347543508,
"delta_cosine": 0.7583887313346791,
"relative_weight_rmse": 0.01260996238822116
},
{
"index": 190,
"total": 252,
"projection": "layers.33.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4526,
"parameters": 50331648,
"reference_delta_l2": 4.682668165881411,
"retained_delta_l2": 4.596757594662856,
"delta_retention_norm_ratio": 0.9816535000612445,
"delta_cosine": 0.8099869817584112,
"relative_weight_rmse": 0.014409651842740531
},
{
"index": 191,
"total": 252,
"projection": "layers.33.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4485,
"parameters": 50331648,
"reference_delta_l2": 4.014875652524574,
"retained_delta_l2": 3.8318803510402013,
"delta_retention_norm_ratio": 0.9544206801599685,
"delta_cosine": 0.7715879258349685,
"relative_weight_rmse": 0.014752805565232565
},
{
"index": 192,
"total": 252,
"projection": "layers.33.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4507,
"parameters": 50331648,
"reference_delta_l2": 4.556334045148909,
"retained_delta_l2": 4.420249311874855,
"delta_retention_norm_ratio": 0.9701328454135311,
"delta_cosine": 0.7957782034107042,
"relative_weight_rmse": 0.014235354318674601
},
{
"index": 193,
"total": 252,
"projection": "layers.33.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0329,
"parameters": 4194304,
"reference_delta_l2": 1.1715448335136518,
"retained_delta_l2": 1.162213151149902,
"delta_retention_norm_ratio": 0.9920347202285358,
"delta_cosine": 0.8277241129643732,
"relative_weight_rmse": 0.015549852935554914
},
{
"index": 194,
"total": 252,
"projection": "layers.33.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1358,
"parameters": 16777216,
"reference_delta_l2": 1.958514518580026,
"retained_delta_l2": 1.8105376501951116,
"delta_retention_norm_ratio": 0.9244443342231635,
"delta_cosine": 0.7414636046442074,
"relative_weight_rmse": 0.012306379666645365
},
{
"index": 195,
"total": 252,
"projection": "layers.33.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1384,
"parameters": 16777216,
"reference_delta_l2": 1.69608482977684,
"retained_delta_l2": 1.5639059770081984,
"delta_retention_norm_ratio": 0.9220682536344407,
"delta_cosine": 0.7428706605467192,
"relative_weight_rmse": 0.011926121485927121
},
{
"index": 196,
"total": 252,
"projection": "layers.33.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.03,
"parameters": 4194304,
"reference_delta_l2": 1.2353120393304602,
"retained_delta_l2": 1.1859367950398427,
"delta_retention_norm_ratio": 0.960030144029537,
"delta_cosine": 0.7865256282621401,
"relative_weight_rmse": 0.012735783508330498
},
{
"index": 197,
"total": 252,
"projection": "layers.34.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4558,
"parameters": 50331648,
"reference_delta_l2": 3.2502062861549645,
"retained_delta_l2": 3.0183480362894173,
"delta_retention_norm_ratio": 0.9286635279572244,
"delta_cosine": 0.7450418333486442,
"relative_weight_rmse": 0.011829379613228915
},
{
"index": 198,
"total": 252,
"projection": "layers.34.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4254,
"parameters": 50331648,
"reference_delta_l2": 3.514521711869592,
"retained_delta_l2": 3.310944444722435,
"delta_retention_norm_ratio": 0.9420753992045019,
"delta_cosine": 0.7571595644355864,
"relative_weight_rmse": 0.013111643557186574
},
{
"index": 199,
"total": 252,
"projection": "layers.34.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4488,
"parameters": 50331648,
"reference_delta_l2": 3.1211155033737468,
"retained_delta_l2": 2.797477929416452,
"delta_retention_norm_ratio": 0.8963070820008228,
"delta_cosine": 0.7098218767994207,
"relative_weight_rmse": 0.011416485843684192
},
{
"index": 200,
"total": 252,
"projection": "layers.34.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0327,
"parameters": 4194304,
"reference_delta_l2": 0.7907094830985183,
"retained_delta_l2": 0.729662410455242,
"delta_retention_norm_ratio": 0.9227945611527842,
"delta_cosine": 0.7531973398853903,
"relative_weight_rmse": 0.011647258114937617
},
{
"index": 201,
"total": 252,
"projection": "layers.34.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1498,
"parameters": 16777216,
"reference_delta_l2": 2.3709792314018685,
"retained_delta_l2": 2.3010744192113624,
"delta_retention_norm_ratio": 0.9705164805896785,
"delta_cosine": 0.7934258079256239,
"relative_weight_rmse": 0.014059371687087978
},
{
"index": 202,
"total": 252,
"projection": "layers.34.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1504,
"parameters": 16777216,
"reference_delta_l2": 2.79968777926931,
"retained_delta_l2": 2.813808867337706,
"delta_retention_norm_ratio": 1.0050438081606663,
"delta_cosine": 0.8602580880686758,
"relative_weight_rmse": 0.015301044294419657
},
{
"index": 203,
"total": 252,
"projection": "layers.34.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0329,
"parameters": 4194304,
"reference_delta_l2": 0.982734313621265,
"retained_delta_l2": 0.8782993706129937,
"delta_retention_norm_ratio": 0.893730236585064,
"delta_cosine": 0.7157103209327442,
"relative_weight_rmse": 0.011415744786938194
},
{
"index": 204,
"total": 252,
"projection": "layers.35.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4528,
"parameters": 50331648,
"reference_delta_l2": 2.328893146712811,
"retained_delta_l2": 2.0819414533345597,
"delta_retention_norm_ratio": 0.8939617759076585,
"delta_cosine": 0.7143709682990358,
"relative_weight_rmse": 0.009801664114430475
},
{
"index": 205,
"total": 252,
"projection": "layers.35.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4085,
"parameters": 50331648,
"reference_delta_l2": 2.62207187548266,
"retained_delta_l2": 2.287770005806219,
"delta_retention_norm_ratio": 0.8725046888293617,
"delta_cosine": 0.6872223936163249,
"relative_weight_rmse": 0.010519143397147825
},
{
"index": 206,
"total": 252,
"projection": "layers.35.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4384,
"parameters": 50331648,
"reference_delta_l2": 2.661133717717755,
"retained_delta_l2": 2.288405518147219,
"delta_retention_norm_ratio": 0.8599363132003018,
"delta_cosine": 0.6745585987490511,
"relative_weight_rmse": 0.010343928392183153
},
{
"index": 207,
"total": 252,
"projection": "layers.35.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0329,
"parameters": 4194304,
"reference_delta_l2": 0.8324495774848747,
"retained_delta_l2": 0.7623977761279171,
"delta_retention_norm_ratio": 0.9158485952162906,
"delta_cosine": 0.7340647630068782,
"relative_weight_rmse": 0.012227252216187886
},
{
"index": 208,
"total": 252,
"projection": "layers.35.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1491,
"parameters": 16777216,
"reference_delta_l2": 1.9766753038638798,
"retained_delta_l2": 1.875041401772476,
"delta_retention_norm_ratio": 0.9485834107945111,
"delta_cosine": 0.7673942637881171,
"relative_weight_rmse": 0.012977993516845263
},
{
"index": 209,
"total": 252,
"projection": "layers.35.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1515,
"parameters": 16777216,
"reference_delta_l2": 1.9213753179971833,
"retained_delta_l2": 1.8334755477137652,
"delta_retention_norm_ratio": 0.9542516397186555,
"delta_cosine": 0.774517248750545,
"relative_weight_rmse": 0.013083061903593797
},
{
"index": 210,
"total": 252,
"projection": "layers.35.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.033,
"parameters": 4194304,
"reference_delta_l2": 0.9019032027159725,
"retained_delta_l2": 0.7953044652105381,
"delta_retention_norm_ratio": 0.8818068976976409,
"delta_cosine": 0.6986239148344,
"relative_weight_rmse": 0.011399535881684587
},
{
"index": 211,
"total": 252,
"projection": "layers.4.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4966,
"parameters": 50331648,
"reference_delta_l2": 2.227049220283431,
"retained_delta_l2": 1.9107794283904227,
"delta_retention_norm_ratio": 0.8579870669168429,
"delta_cosine": 0.6732765069415915,
"relative_weight_rmse": 0.010843631659398535
},
{
"index": 212,
"total": 252,
"projection": "layers.4.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4563,
"parameters": 50331648,
"reference_delta_l2": 3.319017164372891,
"retained_delta_l2": 2.9283370279102168,
"delta_retention_norm_ratio": 0.8822904139646138,
"delta_cosine": 0.709533551101388,
"relative_weight_rmse": 0.010838908238713913
},
{
"index": 213,
"total": 252,
"projection": "layers.4.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4488,
"parameters": 50331648,
"reference_delta_l2": 3.305167090618733,
"retained_delta_l2": 3.1833585538819222,
"delta_retention_norm_ratio": 0.9631460275994677,
"delta_cosine": 0.7955641669252091,
"relative_weight_rmse": 0.014028221678350019
},
{
"index": 214,
"total": 252,
"projection": "layers.4.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0294,
"parameters": 4194304,
"reference_delta_l2": 0.5938600546036626,
"retained_delta_l2": 0.448688235407731,
"delta_retention_norm_ratio": 0.7555454048970878,
"delta_cosine": 0.5784878446307598,
"relative_weight_rmse": 0.009472304471816687
},
{
"index": 215,
"total": 252,
"projection": "layers.4.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1322,
"parameters": 16777216,
"reference_delta_l2": 1.6305222661589405,
"retained_delta_l2": 1.4171410386828274,
"delta_retention_norm_ratio": 0.8691332023457856,
"delta_cosine": 0.6813334731567672,
"relative_weight_rmse": 0.01181406308888834
},
{
"index": 216,
"total": 252,
"projection": "layers.4.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1333,
"parameters": 16777216,
"reference_delta_l2": 1.26308637687989,
"retained_delta_l2": 0.9989223703184803,
"delta_retention_norm_ratio": 0.790858320225134,
"delta_cosine": 0.6086203757277258,
"relative_weight_rmse": 0.010250808841646246
},
{
"index": 217,
"total": 252,
"projection": "layers.4.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0295,
"parameters": 4194304,
"reference_delta_l2": 1.1353646180752022,
"retained_delta_l2": 1.0777128938919571,
"delta_retention_norm_ratio": 0.9492218418070991,
"delta_cosine": 0.7692008360631348,
"relative_weight_rmse": 0.013532222269163355
},
{
"index": 218,
"total": 252,
"projection": "layers.5.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4515,
"parameters": 50331648,
"reference_delta_l2": 1.808160475405009,
"retained_delta_l2": 1.3468812294137011,
"delta_retention_norm_ratio": 0.7448903168354091,
"delta_cosine": 0.5699993278789788,
"relative_weight_rmse": 0.009060283604047972
},
{
"index": 219,
"total": 252,
"projection": "layers.5.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4355,
"parameters": 50331648,
"reference_delta_l2": 2.9558560508595235,
"retained_delta_l2": 2.4923560930472966,
"delta_retention_norm_ratio": 0.843192648817439,
"delta_cosine": 0.6606220321851646,
"relative_weight_rmse": 0.010000901949344193
},
{
"index": 220,
"total": 252,
"projection": "layers.5.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4502,
"parameters": 50331648,
"reference_delta_l2": 3.1816025520962876,
"retained_delta_l2": 2.9681974110479925,
"delta_retention_norm_ratio": 0.9329252672029423,
"delta_cosine": 0.7497088358760183,
"relative_weight_rmse": 0.013138840574027187
},
{
"index": 221,
"total": 252,
"projection": "layers.5.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0327,
"parameters": 4194304,
"reference_delta_l2": 0.6209149901094878,
"retained_delta_l2": 0.4890781833757427,
"delta_retention_norm_ratio": 0.7876733388084286,
"delta_cosine": 0.6083395403547178,
"relative_weight_rmse": 0.009870073710516089
},
{
"index": 222,
"total": 252,
"projection": "layers.5.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.149,
"parameters": 16777216,
"reference_delta_l2": 1.469867212365177,
"retained_delta_l2": 1.2353633578744059,
"delta_retention_norm_ratio": 0.8404591567741492,
"delta_cosine": 0.6524398624650307,
"relative_weight_rmse": 0.011361423810286084
},
{
"index": 223,
"total": 252,
"projection": "layers.5.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1437,
"parameters": 16777216,
"reference_delta_l2": 1.4705658193966193,
"retained_delta_l2": 1.2563663047180098,
"delta_retention_norm_ratio": 0.8543421097829563,
"delta_cosine": 0.6708542762096589,
"relative_weight_rmse": 0.01119795778669845
},
{
"index": 224,
"total": 252,
"projection": "layers.5.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0297,
"parameters": 4194304,
"reference_delta_l2": 1.0468097647814418,
"retained_delta_l2": 0.9824996366568349,
"delta_retention_norm_ratio": 0.938565601613361,
"delta_cosine": 0.7610281819227241,
"relative_weight_rmse": 0.01287908952462893
},
{
"index": 225,
"total": 252,
"projection": "layers.6.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4536,
"parameters": 50331648,
"reference_delta_l2": 4.766346157763611,
"retained_delta_l2": 4.779398771166934,
"delta_retention_norm_ratio": 1.0027384946395603,
"delta_cosine": 0.8441933980723504,
"relative_weight_rmse": 0.015433725989502949
},
{
"index": 226,
"total": 252,
"projection": "layers.6.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4073,
"parameters": 50331648,
"reference_delta_l2": 3.1107591056769075,
"retained_delta_l2": 2.7209549756998825,
"delta_retention_norm_ratio": 0.874691637399482,
"delta_cosine": 0.6923863276695883,
"relative_weight_rmse": 0.010694026948477185
},
{
"index": 227,
"total": 252,
"projection": "layers.6.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4476,
"parameters": 50331648,
"reference_delta_l2": 3.3879989311350243,
"retained_delta_l2": 3.200031219321589,
"delta_retention_norm_ratio": 0.9445195480771702,
"delta_cosine": 0.7631754116426505,
"relative_weight_rmse": 0.013223850199498026
},
{
"index": 228,
"total": 252,
"projection": "layers.6.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0328,
"parameters": 4194304,
"reference_delta_l2": 0.5929994796202283,
"retained_delta_l2": 0.4478840275474995,
"delta_retention_norm_ratio": 0.7552857008143341,
"delta_cosine": 0.578375646961317,
"relative_weight_rmse": 0.009584032159648776
},
{
"index": 229,
"total": 252,
"projection": "layers.6.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1491,
"parameters": 16777216,
"reference_delta_l2": 1.3652270565043076,
"retained_delta_l2": 1.1068695571620093,
"delta_retention_norm_ratio": 0.810758585459163,
"delta_cosine": 0.6266250445827917,
"relative_weight_rmse": 0.01039687480533049
},
{
"index": 230,
"total": 252,
"projection": "layers.6.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1528,
"parameters": 16777216,
"reference_delta_l2": 1.4030166677155094,
"retained_delta_l2": 1.1638888719771145,
"delta_retention_norm_ratio": 0.8295616857298214,
"delta_cosine": 0.6446211265825159,
"relative_weight_rmse": 0.010722959891243085
},
{
"index": 231,
"total": 252,
"projection": "layers.6.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0337,
"parameters": 4194304,
"reference_delta_l2": 0.9864648004210607,
"retained_delta_l2": 0.9106563553347464,
"delta_retention_norm_ratio": 0.9231513936899154,
"delta_cosine": 0.7396376108548454,
"relative_weight_rmse": 0.012400840574320192
},
{
"index": 232,
"total": 252,
"projection": "layers.7.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4925,
"parameters": 50331648,
"reference_delta_l2": 1.8483545328860906,
"retained_delta_l2": 1.3684473384899813,
"delta_retention_norm_ratio": 0.7403597708894278,
"delta_cosine": 0.5659163043326431,
"relative_weight_rmse": 0.008972718015620999
},
{
"index": 233,
"total": 252,
"projection": "layers.7.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4528,
"parameters": 50331648,
"reference_delta_l2": 2.900322939192617,
"retained_delta_l2": 2.4857437715697834,
"delta_retention_norm_ratio": 0.8570575841674228,
"delta_cosine": 0.6740358648736131,
"relative_weight_rmse": 0.010361223144676333
},
{
"index": 234,
"total": 252,
"projection": "layers.7.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4374,
"parameters": 50331648,
"reference_delta_l2": 2.9776668668124984,
"retained_delta_l2": 2.6946860110047024,
"delta_retention_norm_ratio": 0.9049655759138971,
"delta_cosine": 0.7167067872256636,
"relative_weight_rmse": 0.01238997303612827
},
{
"index": 235,
"total": 252,
"projection": "layers.7.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0299,
"parameters": 4194304,
"reference_delta_l2": 0.7222155374904952,
"retained_delta_l2": 0.61291940064473,
"delta_retention_norm_ratio": 0.848665486725002,
"delta_cosine": 0.6623839656715682,
"relative_weight_rmse": 0.010768835737077919
},
{
"index": 236,
"total": 252,
"projection": "layers.7.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1372,
"parameters": 16777216,
"reference_delta_l2": 1.3670254756742946,
"retained_delta_l2": 1.107747944638482,
"delta_retention_norm_ratio": 0.8103345287636851,
"delta_cosine": 0.6254622332605992,
"relative_weight_rmse": 0.01055596827212504
},
{
"index": 237,
"total": 252,
"projection": "layers.7.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1377,
"parameters": 16777216,
"reference_delta_l2": 1.6871151206750943,
"retained_delta_l2": 1.5254978814500704,
"delta_retention_norm_ratio": 0.9042049725922953,
"delta_cosine": 0.7180807252217836,
"relative_weight_rmse": 0.012124167719558517
},
{
"index": 238,
"total": 252,
"projection": "layers.7.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0302,
"parameters": 4194304,
"reference_delta_l2": 0.9908706514359342,
"retained_delta_l2": 0.8947807194620843,
"delta_retention_norm_ratio": 0.9030247471407091,
"delta_cosine": 0.7181343340910804,
"relative_weight_rmse": 0.011994887642077554
},
{
"index": 239,
"total": 252,
"projection": "layers.8.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4506,
"parameters": 50331648,
"reference_delta_l2": 2.6446095772017633,
"retained_delta_l2": 2.2832568056495055,
"delta_retention_norm_ratio": 0.8633625262997793,
"delta_cosine": 0.6776312558021473,
"relative_weight_rmse": 0.011067212445544225
},
{
"index": 240,
"total": 252,
"projection": "layers.8.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4559,
"parameters": 50331648,
"reference_delta_l2": 2.2077595267625174,
"retained_delta_l2": 1.665300374400343,
"delta_retention_norm_ratio": 0.7542942762622147,
"delta_cosine": 0.5776268691871583,
"relative_weight_rmse": 0.009484578090261434
},
{
"index": 241,
"total": 252,
"projection": "layers.8.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4535,
"parameters": 50331648,
"reference_delta_l2": 2.681689242676618,
"retained_delta_l2": 2.2867682544544476,
"delta_retention_norm_ratio": 0.852734246035161,
"delta_cosine": 0.6659738637942503,
"relative_weight_rmse": 0.011433970182999973
},
{
"index": 242,
"total": 252,
"projection": "layers.8.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.033,
"parameters": 4194304,
"reference_delta_l2": 0.8934593974940912,
"retained_delta_l2": 0.8101411189595243,
"delta_retention_norm_ratio": 0.9067464299236744,
"delta_cosine": 0.7198668298992376,
"relative_weight_rmse": 0.012129230318956664
},
{
"index": 243,
"total": 252,
"projection": "layers.8.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1441,
"parameters": 16777216,
"reference_delta_l2": 1.5739691609273754,
"retained_delta_l2": 1.3486393208494216,
"delta_retention_norm_ratio": 0.8568397363356277,
"delta_cosine": 0.6707955703139571,
"relative_weight_rmse": 0.01131379714469802
},
{
"index": 244,
"total": 252,
"projection": "layers.8.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1378,
"parameters": 16777216,
"reference_delta_l2": 1.5962491855603174,
"retained_delta_l2": 1.3857519264014981,
"delta_retention_norm_ratio": 0.868130075765752,
"delta_cosine": 0.6795047651979869,
"relative_weight_rmse": 0.011847211121479321
},
{
"index": 245,
"total": 252,
"projection": "layers.8.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0297,
"parameters": 4194304,
"reference_delta_l2": 0.7964105169524149,
"retained_delta_l2": 0.6603781066385155,
"delta_retention_norm_ratio": 0.8291931015245152,
"delta_cosine": 0.6433784794827893,
"relative_weight_rmse": 0.010628078572564035
},
{
"index": 246,
"total": 252,
"projection": "layers.9.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4506,
"parameters": 50331648,
"reference_delta_l2": 2.4700772804854574,
"retained_delta_l2": 2.099192188365903,
"delta_retention_norm_ratio": 0.8498487901371804,
"delta_cosine": 0.6642743964441243,
"relative_weight_rmse": 0.010886223242880886
},
{
"index": 247,
"total": 252,
"projection": "layers.9.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4206,
"parameters": 50331648,
"reference_delta_l2": 2.707595817057222,
"retained_delta_l2": 2.220126704130334,
"delta_retention_norm_ratio": 0.8199623777463586,
"delta_cosine": 0.6360014827122122,
"relative_weight_rmse": 0.010394406152054227
},
{
"index": 248,
"total": 252,
"projection": "layers.9.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4546,
"parameters": 50331648,
"reference_delta_l2": 3.028492982219278,
"retained_delta_l2": 2.747441128761651,
"delta_retention_norm_ratio": 0.9071974559268511,
"delta_cosine": 0.7221532518239889,
"relative_weight_rmse": 0.01239379122802997
},
{
"index": 249,
"total": 252,
"projection": "layers.9.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0331,
"parameters": 4194304,
"reference_delta_l2": 0.8388941457915762,
"retained_delta_l2": 0.7569475472799214,
"delta_retention_norm_ratio": 0.9023159251703557,
"delta_cosine": 0.7182517899736679,
"relative_weight_rmse": 0.011791685681330627
},
{
"index": 250,
"total": 252,
"projection": "layers.9.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1493,
"parameters": 16777216,
"reference_delta_l2": 1.5261561238923318,
"retained_delta_l2": 1.3008078622212615,
"delta_retention_norm_ratio": 0.8523425892389447,
"delta_cosine": 0.6668637418111044,
"relative_weight_rmse": 0.011161535692909856
},
{
"index": 251,
"total": 252,
"projection": "layers.9.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1522,
"parameters": 16777216,
"reference_delta_l2": 1.8057444961044307,
"retained_delta_l2": 1.6591710515428146,
"delta_retention_norm_ratio": 0.9188293554941899,
"delta_cosine": 0.7350772510464526,
"relative_weight_rmse": 0.01270620772084897
},
{
"index": 252,
"total": 252,
"projection": "layers.9.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0331,
"parameters": 4194304,
"reference_delta_l2": 0.8016387043846219,
"retained_delta_l2": 0.6712410582430005,
"delta_retention_norm_ratio": 0.8373361397991366,
"delta_cosine": 0.6535679524295183,
"relative_weight_rmse": 0.010433106285083201
}
]
}