mrjackspade's picture
Publish step 1530 merge manifest
9e2d38a 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_00001530/adapters.safetensors",
"adapter_sha256": "eb53557a4d1476bf2b62ac6946476bc4fcbf8282034ab04df1e6a33feceaa96c",
"adapter_global_step": "1530",
"output": "/workspace/Ideogram/artifacts/huggingface_release/s020_v1/qwen3vl_8b_ideogram4_nl_s020_v1_step_00001530_fp8_scaled.safetensors",
"output_bytes": 10588637800,
"output_sha256": "d38a2669c13d2a602ed22be89b2f21bba5622cc6960b355ff0801c4d67489333",
"output_tensors": 1254,
"merged_projections": 252,
"lora_rank": 64,
"lora_alpha": 64.0,
"elapsed_seconds": 62.515,
"projections": [
{
"index": 1,
"total": 252,
"projection": "layers.0.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4227,
"parameters": 50331648,
"reference_delta_l2": 3.1269727785306203,
"retained_delta_l2": 2.847838946061438,
"delta_retention_norm_ratio": 0.9107335265641971,
"delta_cosine": 0.7260631937571582,
"relative_weight_rmse": 0.012304272430821054
},
{
"index": 2,
"total": 252,
"projection": "layers.0.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3802,
"parameters": 50331648,
"reference_delta_l2": 3.2955112311104933,
"retained_delta_l2": 3.0060657864831555,
"delta_retention_norm_ratio": 0.9121697896536062,
"delta_cosine": 0.7246474927129162,
"relative_weight_rmse": 0.013464760980937362
},
{
"index": 3,
"total": 252,
"projection": "layers.0.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3827,
"parameters": 50331648,
"reference_delta_l2": 2.9185604335159008,
"retained_delta_l2": 2.6044050786608244,
"delta_retention_norm_ratio": 0.8923594826931086,
"delta_cosine": 0.7051994882419824,
"relative_weight_rmse": 0.012985052515934122
},
{
"index": 4,
"total": 252,
"projection": "layers.0.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 0.6020524049592196,
"retained_delta_l2": 0.4409459157004102,
"delta_retention_norm_ratio": 0.7324045416449718,
"delta_cosine": 0.5601915050421808,
"relative_weight_rmse": 0.009005100813604077
},
{
"index": 5,
"total": 252,
"projection": "layers.0.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.128,
"parameters": 16777216,
"reference_delta_l2": 1.973039220938326,
"retained_delta_l2": 1.8611340842146211,
"delta_retention_norm_ratio": 0.9432828625320049,
"delta_cosine": 0.7627770951487235,
"relative_weight_rmse": 0.013622675387227352
},
{
"index": 6,
"total": 252,
"projection": "layers.0.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1267,
"parameters": 16777216,
"reference_delta_l2": 1.226169488957416,
"retained_delta_l2": 0.9545936022842115,
"delta_retention_norm_ratio": 0.7785168452494122,
"delta_cosine": 0.5992496607567963,
"relative_weight_rmse": 0.0098083098974002
},
{
"index": 7,
"total": 252,
"projection": "layers.0.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.03,
"parameters": 4194304,
"reference_delta_l2": 0.7963862260294058,
"retained_delta_l2": 0.693827574325755,
"delta_retention_norm_ratio": 0.871219957915415,
"delta_cosine": 0.6842739979149719,
"relative_weight_rmse": 0.01199370525764303
},
{
"index": 8,
"total": 252,
"projection": "layers.1.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4194,
"parameters": 50331648,
"reference_delta_l2": 2.0356480384900277,
"retained_delta_l2": 1.9001821581406886,
"delta_retention_norm_ratio": 0.9334531914221169,
"delta_cosine": 0.7565319603551144,
"relative_weight_rmse": 0.01197967566508188
},
{
"index": 9,
"total": 252,
"projection": "layers.1.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3818,
"parameters": 50331648,
"reference_delta_l2": 3.8434880754811567,
"retained_delta_l2": 3.810271658431912,
"delta_retention_norm_ratio": 0.9913577416146696,
"delta_cosine": 0.846772440107378,
"relative_weight_rmse": 0.014363476876126656
},
{
"index": 10,
"total": 252,
"projection": "layers.1.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3789,
"parameters": 50331648,
"reference_delta_l2": 4.22597100602001,
"retained_delta_l2": 4.222919158295008,
"delta_retention_norm_ratio": 0.9992778351482644,
"delta_cosine": 0.9293241948032982,
"relative_weight_rmse": 0.01576596702506838
},
{
"index": 11,
"total": 252,
"projection": "layers.1.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 0.6774406044671824,
"retained_delta_l2": 0.556723909165314,
"delta_retention_norm_ratio": 0.8218047537956277,
"delta_cosine": 0.6379560394473641,
"relative_weight_rmse": 0.010828829915143327
},
{
"index": 12,
"total": 252,
"projection": "layers.1.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1283,
"parameters": 16777216,
"reference_delta_l2": 1.7138713405131234,
"retained_delta_l2": 1.5333545860507296,
"delta_retention_norm_ratio": 0.8946731004858579,
"delta_cosine": 0.7052913046292778,
"relative_weight_rmse": 0.01266311458670879
},
{
"index": 13,
"total": 252,
"projection": "layers.1.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1283,
"parameters": 16777216,
"reference_delta_l2": 2.0901680883206906,
"retained_delta_l2": 2.0209289132881056,
"delta_retention_norm_ratio": 0.966873872288322,
"delta_cosine": 0.7924993655955894,
"relative_weight_rmse": 0.01462637178641398
},
{
"index": 14,
"total": 252,
"projection": "layers.1.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 0.9185446247120631,
"retained_delta_l2": 0.8362166191359856,
"delta_retention_norm_ratio": 0.910371251040868,
"delta_cosine": 0.7229793123766166,
"relative_weight_rmse": 0.012474231278458162
},
{
"index": 15,
"total": 252,
"projection": "layers.10.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4217,
"parameters": 50331648,
"reference_delta_l2": 2.814992850391522,
"retained_delta_l2": 2.4972698433859355,
"delta_retention_norm_ratio": 0.8871318600466797,
"delta_cosine": 0.7011173536136529,
"relative_weight_rmse": 0.011668980634333866
},
{
"index": 16,
"total": 252,
"projection": "layers.10.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3807,
"parameters": 50331648,
"reference_delta_l2": 2.3901047204142674,
"retained_delta_l2": 1.8529896227829457,
"delta_retention_norm_ratio": 0.7752754960718935,
"delta_cosine": 0.5960595634147681,
"relative_weight_rmse": 0.009651275667908231
},
{
"index": 17,
"total": 252,
"projection": "layers.10.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3777,
"parameters": 50331648,
"reference_delta_l2": 2.858757490165862,
"retained_delta_l2": 2.531018953398552,
"delta_retention_norm_ratio": 0.8853562997579433,
"delta_cosine": 0.6978358040396213,
"relative_weight_rmse": 0.011913428049205213
},
{
"index": 18,
"total": 252,
"projection": "layers.10.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0315,
"parameters": 4194304,
"reference_delta_l2": 0.8993867230288537,
"retained_delta_l2": 0.8269512206391207,
"delta_retention_norm_ratio": 0.9194612278177814,
"delta_cosine": 0.7376595279985013,
"relative_weight_rmse": 0.011978551081292697
},
{
"index": 19,
"total": 252,
"projection": "layers.10.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1286,
"parameters": 16777216,
"reference_delta_l2": 1.7753697362828071,
"retained_delta_l2": 1.5870460932219161,
"delta_retention_norm_ratio": 0.8939242687243306,
"delta_cosine": 0.7071171086611312,
"relative_weight_rmse": 0.01217158891103583
},
{
"index": 20,
"total": 252,
"projection": "layers.10.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1297,
"parameters": 16777216,
"reference_delta_l2": 2.1201857998503106,
"retained_delta_l2": 2.0228005554894857,
"delta_retention_norm_ratio": 0.9540675895632824,
"delta_cosine": 0.7801449625303983,
"relative_weight_rmse": 0.013454529664829615
},
{
"index": 21,
"total": 252,
"projection": "layers.10.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 0.9083682746450251,
"retained_delta_l2": 0.8106920352712357,
"delta_retention_norm_ratio": 0.8924706618447681,
"delta_cosine": 0.7108114019827055,
"relative_weight_rmse": 0.011192079222170918
},
{
"index": 22,
"total": 252,
"projection": "layers.11.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4228,
"parameters": 50331648,
"reference_delta_l2": 2.606977159285387,
"retained_delta_l2": 2.2314210305488467,
"delta_retention_norm_ratio": 0.8559419182485333,
"delta_cosine": 0.6705462804874116,
"relative_weight_rmse": 0.011040055210428465
},
{
"index": 23,
"total": 252,
"projection": "layers.11.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3819,
"parameters": 50331648,
"reference_delta_l2": 2.7680849435848476,
"retained_delta_l2": 2.3351837928526864,
"delta_retention_norm_ratio": 0.8436098748575518,
"delta_cosine": 0.6589891775775347,
"relative_weight_rmse": 0.010796790721620953
},
{
"index": 24,
"total": 252,
"projection": "layers.11.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3802,
"parameters": 50331648,
"reference_delta_l2": 2.6725192607048647,
"retained_delta_l2": 2.270184682610497,
"delta_retention_norm_ratio": 0.8494549378894826,
"delta_cosine": 0.6631062521055323,
"relative_weight_rmse": 0.01136743150604899
},
{
"index": 25,
"total": 252,
"projection": "layers.11.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 1.1311720068959925,
"retained_delta_l2": 1.0887236416233885,
"delta_retention_norm_ratio": 0.9624739959848504,
"delta_cosine": 0.7879357863041313,
"relative_weight_rmse": 0.013664456675574239
},
{
"index": 26,
"total": 252,
"projection": "layers.11.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1286,
"parameters": 16777216,
"reference_delta_l2": 1.6724912657557298,
"retained_delta_l2": 1.4693522691368264,
"delta_retention_norm_ratio": 0.8785410717663071,
"delta_cosine": 0.693464586870248,
"relative_weight_rmse": 0.011737459168859871
},
{
"index": 27,
"total": 252,
"projection": "layers.11.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1291,
"parameters": 16777216,
"reference_delta_l2": 1.4032656533694052,
"retained_delta_l2": 1.1662019028876682,
"delta_retention_norm_ratio": 0.8310628141488968,
"delta_cosine": 0.6459583221479873,
"relative_weight_rmse": 0.010731881370672783
},
{
"index": 28,
"total": 252,
"projection": "layers.11.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 0.9957992771310851,
"retained_delta_l2": 0.9052954064265708,
"delta_retention_norm_ratio": 0.9091143438412032,
"delta_cosine": 0.7266247628209584,
"relative_weight_rmse": 0.011870418153047967
},
{
"index": 29,
"total": 252,
"projection": "layers.12.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4268,
"parameters": 50331648,
"reference_delta_l2": 2.5651561747652067,
"retained_delta_l2": 2.1855692680213097,
"delta_retention_norm_ratio": 0.8520219117736014,
"delta_cosine": 0.6666850001099782,
"relative_weight_rmse": 0.010824502316704015
},
{
"index": 30,
"total": 252,
"projection": "layers.12.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3816,
"parameters": 50331648,
"reference_delta_l2": 3.204162211936139,
"retained_delta_l2": 2.870834955746633,
"delta_retention_norm_ratio": 0.895970542643629,
"delta_cosine": 0.7120944191940738,
"relative_weight_rmse": 0.011956951195376942
},
{
"index": 31,
"total": 252,
"projection": "layers.12.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3783,
"parameters": 50331648,
"reference_delta_l2": 2.6816222918902204,
"retained_delta_l2": 2.2846585181098966,
"delta_retention_norm_ratio": 0.851968797029759,
"delta_cosine": 0.665714734982328,
"relative_weight_rmse": 0.011204452253251175
},
{
"index": 32,
"total": 252,
"projection": "layers.12.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 0.9333935858291095,
"retained_delta_l2": 0.8668283058369424,
"delta_retention_norm_ratio": 0.92868466100178,
"delta_cosine": 0.7480021997803925,
"relative_weight_rmse": 0.012298593397018543
},
{
"index": 33,
"total": 252,
"projection": "layers.12.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1283,
"parameters": 16777216,
"reference_delta_l2": 1.7763183189653498,
"retained_delta_l2": 1.6100027453249657,
"delta_retention_norm_ratio": 0.906370625205702,
"delta_cosine": 0.722417460316628,
"relative_weight_rmse": 0.012041667196134125
},
{
"index": 34,
"total": 252,
"projection": "layers.12.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1291,
"parameters": 16777216,
"reference_delta_l2": 1.8083228859292346,
"retained_delta_l2": 1.6539196082767817,
"delta_retention_norm_ratio": 0.9146152056948001,
"delta_cosine": 0.7315547943270726,
"relative_weight_rmse": 0.012441787557432443
},
{
"index": 35,
"total": 252,
"projection": "layers.12.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0301,
"parameters": 4194304,
"reference_delta_l2": 1.087080932464243,
"retained_delta_l2": 1.0241777772412444,
"delta_retention_norm_ratio": 0.9421357202168866,
"delta_cosine": 0.7660598065202867,
"relative_weight_rmse": 0.012191010035375172
},
{
"index": 36,
"total": 252,
"projection": "layers.13.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.422,
"parameters": 50331648,
"reference_delta_l2": 2.3448327537056235,
"retained_delta_l2": 1.9207321210034751,
"delta_retention_norm_ratio": 0.8191339522906584,
"delta_cosine": 0.6358730373758618,
"relative_weight_rmse": 0.010190574670263465
},
{
"index": 37,
"total": 252,
"projection": "layers.13.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3786,
"parameters": 50331648,
"reference_delta_l2": 2.9834700673737085,
"retained_delta_l2": 2.6083816703705023,
"delta_retention_norm_ratio": 0.8742778078771243,
"delta_cosine": 0.6871331378438766,
"relative_weight_rmse": 0.011869124678420157
},
{
"index": 38,
"total": 252,
"projection": "layers.13.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3804,
"parameters": 50331648,
"reference_delta_l2": 2.401616101062306,
"retained_delta_l2": 1.9418135391936848,
"delta_retention_norm_ratio": 0.8085445206395657,
"delta_cosine": 0.625121565212576,
"relative_weight_rmse": 0.01034900613300074
},
{
"index": 39,
"total": 252,
"projection": "layers.13.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 1.3429926619871464,
"retained_delta_l2": 1.3290010352693131,
"delta_retention_norm_ratio": 0.9895817549016754,
"delta_cosine": 0.8348059277858658,
"relative_weight_rmse": 0.014782896559259611
},
{
"index": 40,
"total": 252,
"projection": "layers.13.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1291,
"parameters": 16777216,
"reference_delta_l2": 1.7282491063055783,
"retained_delta_l2": 1.548538547686199,
"delta_retention_norm_ratio": 0.8960158243602159,
"delta_cosine": 0.711704845248182,
"relative_weight_rmse": 0.012253301785092546
},
{
"index": 41,
"total": 252,
"projection": "layers.13.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1283,
"parameters": 16777216,
"reference_delta_l2": 1.4115433572363378,
"retained_delta_l2": 1.175948170162018,
"delta_retention_norm_ratio": 0.8330939068456303,
"delta_cosine": 0.6475618078393119,
"relative_weight_rmse": 0.0107592151687993
},
{
"index": 42,
"total": 252,
"projection": "layers.13.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0308,
"parameters": 4194304,
"reference_delta_l2": 1.0309214075337587,
"retained_delta_l2": 0.9500876562599483,
"delta_retention_norm_ratio": 0.9215907724069999,
"delta_cosine": 0.7391921596622596,
"relative_weight_rmse": 0.012333454252584489
},
{
"index": 43,
"total": 252,
"projection": "layers.14.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4185,
"parameters": 50331648,
"reference_delta_l2": 2.4700766817524675,
"retained_delta_l2": 2.0896958078012076,
"delta_retention_norm_ratio": 0.8460044270037043,
"delta_cosine": 0.6616889399153113,
"relative_weight_rmse": 0.010524003264311962
},
{
"index": 44,
"total": 252,
"projection": "layers.14.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3786,
"parameters": 50331648,
"reference_delta_l2": 2.7277497199379104,
"retained_delta_l2": 2.3124647233360016,
"delta_retention_norm_ratio": 0.8477554617395904,
"delta_cosine": 0.6610309572311588,
"relative_weight_rmse": 0.011343492494406103
},
{
"index": 45,
"total": 252,
"projection": "layers.14.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3784,
"parameters": 50331648,
"reference_delta_l2": 2.942592591945144,
"retained_delta_l2": 2.6081569531417084,
"delta_retention_norm_ratio": 0.8863466047869157,
"delta_cosine": 0.7007091347828908,
"relative_weight_rmse": 0.0117781189591218
},
{
"index": 46,
"total": 252,
"projection": "layers.14.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 0.730438284908461,
"retained_delta_l2": 0.6207823208583618,
"delta_retention_norm_ratio": 0.8498764833173533,
"delta_cosine": 0.6669530936392939,
"relative_weight_rmse": 0.010731546950400846
},
{
"index": 47,
"total": 252,
"projection": "layers.14.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1339,
"parameters": 16777216,
"reference_delta_l2": 1.5558795722316472,
"retained_delta_l2": 1.3546514988750578,
"delta_retention_norm_ratio": 0.8706660355030167,
"delta_cosine": 0.6850243635581522,
"relative_weight_rmse": 0.011440865436907418
},
{
"index": 48,
"total": 252,
"projection": "layers.14.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.129,
"parameters": 16777216,
"reference_delta_l2": 2.328584458511773,
"retained_delta_l2": 2.2710392840584004,
"delta_retention_norm_ratio": 0.9752874866775711,
"delta_cosine": 0.806257581943501,
"relative_weight_rmse": 0.013951814180246073
},
{
"index": 49,
"total": 252,
"projection": "layers.14.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 1.019341600551523,
"retained_delta_l2": 0.9467884373671968,
"delta_retention_norm_ratio": 0.9288235041667378,
"delta_cosine": 0.7462463505114401,
"relative_weight_rmse": 0.012397661612345819
},
{
"index": 50,
"total": 252,
"projection": "layers.15.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4207,
"parameters": 50331648,
"reference_delta_l2": 2.3729892154991092,
"retained_delta_l2": 1.9749539619015724,
"delta_retention_norm_ratio": 0.8322641961464547,
"delta_cosine": 0.6484044467838634,
"relative_weight_rmse": 0.010304792086001663
},
{
"index": 51,
"total": 252,
"projection": "layers.15.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3793,
"parameters": 50331648,
"reference_delta_l2": 2.860768519065232,
"retained_delta_l2": 2.499002695722921,
"delta_retention_norm_ratio": 0.8735424341636284,
"delta_cosine": 0.6853882848453808,
"relative_weight_rmse": 0.011983346093962125
},
{
"index": 52,
"total": 252,
"projection": "layers.15.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3779,
"parameters": 50331648,
"reference_delta_l2": 2.784365351837079,
"retained_delta_l2": 2.435335032959766,
"delta_retention_norm_ratio": 0.8746463646923963,
"delta_cosine": 0.6889917605823569,
"relative_weight_rmse": 0.011411931564811607
},
{
"index": 53,
"total": 252,
"projection": "layers.15.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 0.811292672406483,
"retained_delta_l2": 0.7165957721306848,
"delta_retention_norm_ratio": 0.883276524617306,
"delta_cosine": 0.6981370952227917,
"relative_weight_rmse": 0.011563410563958104
},
{
"index": 54,
"total": 252,
"projection": "layers.15.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1281,
"parameters": 16777216,
"reference_delta_l2": 2.1524919824451088,
"retained_delta_l2": 2.069290000539231,
"delta_retention_norm_ratio": 0.9613462058932433,
"delta_cosine": 0.7889175168869766,
"relative_weight_rmse": 0.013845656978810577
},
{
"index": 55,
"total": 252,
"projection": "layers.15.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1284,
"parameters": 16777216,
"reference_delta_l2": 2.237617525101801,
"retained_delta_l2": 2.162694502621995,
"delta_retention_norm_ratio": 0.9665166090096664,
"delta_cosine": 0.8011403103935806,
"relative_weight_rmse": 0.01345303681907185
},
{
"index": 56,
"total": 252,
"projection": "layers.15.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 0.8555601598701333,
"retained_delta_l2": 0.7601176861471938,
"delta_retention_norm_ratio": 0.8884444622369666,
"delta_cosine": 0.7049806055787962,
"relative_weight_rmse": 0.011196324317514623
},
{
"index": 57,
"total": 252,
"projection": "layers.16.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4219,
"parameters": 50331648,
"reference_delta_l2": 2.486670922049404,
"retained_delta_l2": 2.123014037898322,
"delta_retention_norm_ratio": 0.8537575354557281,
"delta_cosine": 0.6684899230202213,
"relative_weight_rmse": 0.010653798959174992
},
{
"index": 58,
"total": 252,
"projection": "layers.16.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3825,
"parameters": 50331648,
"reference_delta_l2": 2.634456337737001,
"retained_delta_l2": 2.250958051927088,
"delta_retention_norm_ratio": 0.8544298190421565,
"delta_cosine": 0.6679123511622501,
"relative_weight_rmse": 0.011055786192365621
},
{
"index": 59,
"total": 252,
"projection": "layers.16.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3819,
"parameters": 50331648,
"reference_delta_l2": 2.555779063907066,
"retained_delta_l2": 2.154211406052585,
"delta_retention_norm_ratio": 0.8428785713423143,
"delta_cosine": 0.6569512870103185,
"relative_weight_rmse": 0.0108856683839722
},
{
"index": 60,
"total": 252,
"projection": "layers.16.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 0.7412285479580177,
"retained_delta_l2": 0.6389193652031682,
"delta_retention_norm_ratio": 0.8619734992173504,
"delta_cosine": 0.6759999689935707,
"relative_weight_rmse": 0.011222976218507576
},
{
"index": 61,
"total": 252,
"projection": "layers.16.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1297,
"parameters": 16777216,
"reference_delta_l2": 1.6690605954864628,
"retained_delta_l2": 1.493513315661502,
"delta_retention_norm_ratio": 0.8948227042806698,
"delta_cosine": 0.7098290927173299,
"relative_weight_rmse": 0.011949817909595841
},
{
"index": 62,
"total": 252,
"projection": "layers.16.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1301,
"parameters": 16777216,
"reference_delta_l2": 1.8097904473131718,
"retained_delta_l2": 1.6776227535572412,
"delta_retention_norm_ratio": 0.9269707197581092,
"delta_cosine": 0.7479046228414,
"relative_weight_rmse": 0.012462548760711567
},
{
"index": 63,
"total": 252,
"projection": "layers.16.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0318,
"parameters": 4194304,
"reference_delta_l2": 0.9607306308156371,
"retained_delta_l2": 0.8776427602291047,
"delta_retention_norm_ratio": 0.913515955543134,
"delta_cosine": 0.7303090433762676,
"relative_weight_rmse": 0.01217207238997773
},
{
"index": 64,
"total": 252,
"projection": "layers.17.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4235,
"parameters": 50331648,
"reference_delta_l2": 2.5743785829242163,
"retained_delta_l2": 2.2126602193637552,
"delta_retention_norm_ratio": 0.8594929409529237,
"delta_cosine": 0.6740881397607131,
"relative_weight_rmse": 0.010910443086156227
},
{
"index": 65,
"total": 252,
"projection": "layers.17.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3837,
"parameters": 50331648,
"reference_delta_l2": 3.5206949864592643,
"retained_delta_l2": 3.2974553312957817,
"delta_retention_norm_ratio": 0.9365921626207123,
"delta_cosine": 0.7552932233186679,
"relative_weight_rmse": 0.013254577723690309
},
{
"index": 66,
"total": 252,
"projection": "layers.17.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3789,
"parameters": 50331648,
"reference_delta_l2": 3.1737660104317835,
"retained_delta_l2": 2.8884029937116837,
"delta_retention_norm_ratio": 0.9100869390553222,
"delta_cosine": 0.7246208492159082,
"relative_weight_rmse": 0.012350866692247381
},
{
"index": 67,
"total": 252,
"projection": "layers.17.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 0.7865736499704759,
"retained_delta_l2": 0.7036271371727426,
"delta_retention_norm_ratio": 0.8945470487082212,
"delta_cosine": 0.7143540152366147,
"relative_weight_rmse": 0.011553108870085779
},
{
"index": 68,
"total": 252,
"projection": "layers.17.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1288,
"parameters": 16777216,
"reference_delta_l2": 1.7300110985450325,
"retained_delta_l2": 1.5699751991256123,
"delta_retention_norm_ratio": 0.907494293213487,
"delta_cosine": 0.7246934984576693,
"relative_weight_rmse": 0.012202281638474382
},
{
"index": 69,
"total": 252,
"projection": "layers.17.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.131,
"parameters": 16777216,
"reference_delta_l2": 1.6625304841109203,
"retained_delta_l2": 1.4882508413095632,
"delta_retention_norm_ratio": 0.8951720618256468,
"delta_cosine": 0.7081489443113639,
"relative_weight_rmse": 0.011974792575036261
},
{
"index": 70,
"total": 252,
"projection": "layers.17.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0307,
"parameters": 4194304,
"reference_delta_l2": 1.0397286647895199,
"retained_delta_l2": 0.9760008642061044,
"delta_retention_norm_ratio": 0.9387072774450089,
"delta_cosine": 0.760020977587067,
"relative_weight_rmse": 0.012815811763191241
},
{
"index": 71,
"total": 252,
"projection": "layers.18.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4283,
"parameters": 50331648,
"reference_delta_l2": 2.907788266384756,
"retained_delta_l2": 2.619001013143209,
"delta_retention_norm_ratio": 0.9006849100465641,
"delta_cosine": 0.7151994622396686,
"relative_weight_rmse": 0.011707196901030538
},
{
"index": 72,
"total": 252,
"projection": "layers.18.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3801,
"parameters": 50331648,
"reference_delta_l2": 4.695436688018443,
"retained_delta_l2": 4.6538394778219105,
"delta_retention_norm_ratio": 0.9911409283181948,
"delta_cosine": 0.8260308652034198,
"relative_weight_rmse": 0.015372444941754954
},
{
"index": 73,
"total": 252,
"projection": "layers.18.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3802,
"parameters": 50331648,
"reference_delta_l2": 3.271060474066289,
"retained_delta_l2": 3.010618542070301,
"delta_retention_norm_ratio": 0.9203799703304689,
"delta_cosine": 0.7369648534543821,
"relative_weight_rmse": 0.01249584639898641
},
{
"index": 74,
"total": 252,
"projection": "layers.18.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 1.1197355469645207,
"retained_delta_l2": 1.0866039993052732,
"delta_retention_norm_ratio": 0.9704112745647279,
"delta_cosine": 0.7993672916496082,
"relative_weight_rmse": 0.014013616857796325
},
{
"index": 75,
"total": 252,
"projection": "layers.18.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1286,
"parameters": 16777216,
"reference_delta_l2": 1.630240547961089,
"retained_delta_l2": 1.4574841013298292,
"delta_retention_norm_ratio": 0.8940300884753952,
"delta_cosine": 0.710921071578677,
"relative_weight_rmse": 0.011706527987302644
},
{
"index": 76,
"total": 252,
"projection": "layers.18.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1308,
"parameters": 16777216,
"reference_delta_l2": 2.1040520683613635,
"retained_delta_l2": 2.01459798744349,
"delta_retention_norm_ratio": 0.9574848539810423,
"delta_cosine": 0.7843100071944246,
"relative_weight_rmse": 0.013555611714405088
},
{
"index": 77,
"total": 252,
"projection": "layers.18.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 0.9279404311746887,
"retained_delta_l2": 0.8448875028231673,
"delta_retention_norm_ratio": 0.9104975647559791,
"delta_cosine": 0.7277625649326424,
"relative_weight_rmse": 0.011786217011799544
},
{
"index": 78,
"total": 252,
"projection": "layers.19.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4259,
"parameters": 50331648,
"reference_delta_l2": 2.6903627105142847,
"retained_delta_l2": 2.353222050442629,
"delta_retention_norm_ratio": 0.8746857965455489,
"delta_cosine": 0.6889555065425643,
"relative_weight_rmse": 0.01120078822653183
},
{
"index": 79,
"total": 252,
"projection": "layers.19.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3795,
"parameters": 50331648,
"reference_delta_l2": 4.609492395721555,
"retained_delta_l2": 4.542064055131354,
"delta_retention_norm_ratio": 0.9853718512145097,
"delta_cosine": 0.8182857698148852,
"relative_weight_rmse": 0.015436989316013602
},
{
"index": 80,
"total": 252,
"projection": "layers.19.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3798,
"parameters": 50331648,
"reference_delta_l2": 3.0876646446042115,
"retained_delta_l2": 2.7671865882324376,
"delta_retention_norm_ratio": 0.8962069741181836,
"delta_cosine": 0.7104741013842515,
"relative_weight_rmse": 0.012185078859367442
},
{
"index": 81,
"total": 252,
"projection": "layers.19.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0322,
"parameters": 4194304,
"reference_delta_l2": 1.043740752861045,
"retained_delta_l2": 1.0044144256684904,
"delta_retention_norm_ratio": 0.9623217479199164,
"delta_cosine": 0.7948396167245384,
"relative_weight_rmse": 0.01359054135439678
},
{
"index": 82,
"total": 252,
"projection": "layers.19.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1291,
"parameters": 16777216,
"reference_delta_l2": 1.6397315405573312,
"retained_delta_l2": 1.4602018074618694,
"delta_retention_norm_ratio": 0.8905127280564227,
"delta_cosine": 0.7051211626469679,
"relative_weight_rmse": 0.011795533568959942
},
{
"index": 83,
"total": 252,
"projection": "layers.19.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.129,
"parameters": 16777216,
"reference_delta_l2": 1.939735345753661,
"retained_delta_l2": 1.8307316810862433,
"delta_retention_norm_ratio": 0.9438048778633429,
"delta_cosine": 0.766157543535145,
"relative_weight_rmse": 0.012834627407222321
},
{
"index": 84,
"total": 252,
"projection": "layers.19.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0308,
"parameters": 4194304,
"reference_delta_l2": 1.240107785319101,
"retained_delta_l2": 1.2070027707450557,
"delta_retention_norm_ratio": 0.9733047280519034,
"delta_cosine": 0.8060440180768825,
"relative_weight_rmse": 0.013912234594691633
},
{
"index": 85,
"total": 252,
"projection": "layers.2.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4212,
"parameters": 50331648,
"reference_delta_l2": 1.5488637789895827,
"retained_delta_l2": 1.333592944930879,
"delta_retention_norm_ratio": 0.8610137076101438,
"delta_cosine": 0.6978094443335896,
"relative_weight_rmse": 0.010095503718382233
},
{
"index": 86,
"total": 252,
"projection": "layers.2.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3797,
"parameters": 50331648,
"reference_delta_l2": 2.5067237100982185,
"retained_delta_l2": 2.3633072007298566,
"delta_retention_norm_ratio": 0.9427872689795788,
"delta_cosine": 0.797306679939323,
"relative_weight_rmse": 0.012650929505998188
},
{
"index": 87,
"total": 252,
"projection": "layers.2.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3872,
"parameters": 50331648,
"reference_delta_l2": 3.388990795860513,
"retained_delta_l2": 3.369597789494612,
"delta_retention_norm_ratio": 0.9942776456077754,
"delta_cosine": 0.8804779353007692,
"relative_weight_rmse": 0.014975526320525897
},
{
"index": 88,
"total": 252,
"projection": "layers.2.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0334,
"parameters": 4194304,
"reference_delta_l2": 0.5767551158284542,
"retained_delta_l2": 0.4338007277480212,
"delta_retention_norm_ratio": 0.7521402339447089,
"delta_cosine": 0.5759507889214573,
"relative_weight_rmse": 0.009284924835653236
},
{
"index": 89,
"total": 252,
"projection": "layers.2.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1323,
"parameters": 16777216,
"reference_delta_l2": 2.0858322921045134,
"retained_delta_l2": 1.9673083687243027,
"delta_retention_norm_ratio": 0.9431766763661401,
"delta_cosine": 0.760528198570256,
"relative_weight_rmse": 0.013946873190318619
},
{
"index": 90,
"total": 252,
"projection": "layers.2.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1292,
"parameters": 16777216,
"reference_delta_l2": 1.22645560287921,
"retained_delta_l2": 0.9678507654384145,
"delta_retention_norm_ratio": 0.7891445586503919,
"delta_cosine": 0.6073761042200562,
"relative_weight_rmse": 0.010166911918841
},
{
"index": 91,
"total": 252,
"projection": "layers.2.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 1.160020130813952,
"retained_delta_l2": 1.1187266586223268,
"delta_retention_norm_ratio": 0.9644027969043514,
"delta_cosine": 0.7930462437441694,
"relative_weight_rmse": 0.013506082946072787
},
{
"index": 92,
"total": 252,
"projection": "layers.20.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4185,
"parameters": 50331648,
"reference_delta_l2": 2.6417271050276083,
"retained_delta_l2": 2.2537007027798666,
"delta_retention_norm_ratio": 0.853116394381059,
"delta_cosine": 0.6655062942081844,
"relative_weight_rmse": 0.011234411585821199
},
{
"index": 93,
"total": 252,
"projection": "layers.20.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3799,
"parameters": 50331648,
"reference_delta_l2": 4.493113904575293,
"retained_delta_l2": 4.418058987551159,
"delta_retention_norm_ratio": 0.9832955677024553,
"delta_cosine": 0.8145858904153334,
"relative_weight_rmse": 0.015326730744241561
},
{
"index": 94,
"total": 252,
"projection": "layers.20.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3827,
"parameters": 50331648,
"reference_delta_l2": 3.1222963302444633,
"retained_delta_l2": 2.8043110500175956,
"delta_retention_norm_ratio": 0.8981565980311771,
"delta_cosine": 0.7120258417330829,
"relative_weight_rmse": 0.012265961759884383
},
{
"index": 95,
"total": 252,
"projection": "layers.20.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 0.8398008136310411,
"retained_delta_l2": 0.7713855239799262,
"delta_retention_norm_ratio": 0.9185339088261797,
"delta_cosine": 0.7387832950710411,
"relative_weight_rmse": 0.01216287387936285
},
{
"index": 96,
"total": 252,
"projection": "layers.20.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1305,
"parameters": 16777216,
"reference_delta_l2": 1.610349253641708,
"retained_delta_l2": 1.4294397436336252,
"delta_retention_norm_ratio": 0.8876582147636815,
"delta_cosine": 0.7035696401886516,
"relative_weight_rmse": 0.011828366034993853
},
{
"index": 97,
"total": 252,
"projection": "layers.20.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1305,
"parameters": 16777216,
"reference_delta_l2": 1.8896686050855038,
"retained_delta_l2": 1.7652411736320934,
"delta_retention_norm_ratio": 0.9341538346361107,
"delta_cosine": 0.7566278054862103,
"relative_weight_rmse": 0.012326217639228183
},
{
"index": 98,
"total": 252,
"projection": "layers.20.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0316,
"parameters": 4194304,
"reference_delta_l2": 0.943199112391037,
"retained_delta_l2": 0.8653885918846724,
"delta_retention_norm_ratio": 0.9175036113964181,
"delta_cosine": 0.7368046799191541,
"relative_weight_rmse": 0.012012097415746282
},
{
"index": 99,
"total": 252,
"projection": "layers.21.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4227,
"parameters": 50331648,
"reference_delta_l2": 3.096352728706979,
"retained_delta_l2": 2.7926921415794785,
"delta_retention_norm_ratio": 0.9019295882177133,
"delta_cosine": 0.7149317138374539,
"relative_weight_rmse": 0.012284424447190342
},
{
"index": 100,
"total": 252,
"projection": "layers.21.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3831,
"parameters": 50331648,
"reference_delta_l2": 4.601478745121303,
"retained_delta_l2": 4.536254819602865,
"delta_retention_norm_ratio": 0.9858254423998828,
"delta_cosine": 0.8160791678831575,
"relative_weight_rmse": 0.01586951141282905
},
{
"index": 101,
"total": 252,
"projection": "layers.21.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3808,
"parameters": 50331648,
"reference_delta_l2": 3.3244217417021664,
"retained_delta_l2": 3.048031819044838,
"delta_retention_norm_ratio": 0.9168607522955822,
"delta_cosine": 0.7318288858858512,
"relative_weight_rmse": 0.0127188568650319
},
{
"index": 102,
"total": 252,
"projection": "layers.21.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 0.8883500110394597,
"retained_delta_l2": 0.821323965944344,
"delta_retention_norm_ratio": 0.9245499586174504,
"delta_cosine": 0.7418620870457034,
"relative_weight_rmse": 0.012815217993680855
},
{
"index": 103,
"total": 252,
"projection": "layers.21.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1284,
"parameters": 16777216,
"reference_delta_l2": 1.7363350952651777,
"retained_delta_l2": 1.5879440262573505,
"delta_retention_norm_ratio": 0.9145377701501999,
"delta_cosine": 0.7322787696329526,
"relative_weight_rmse": 0.01238042253526102
},
{
"index": 104,
"total": 252,
"projection": "layers.21.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1298,
"parameters": 16777216,
"reference_delta_l2": 1.3431818757953342,
"retained_delta_l2": 1.1096977110615218,
"delta_retention_norm_ratio": 0.8261708492786503,
"delta_cosine": 0.6430382021709762,
"relative_weight_rmse": 0.010312170814682213
},
{
"index": 105,
"total": 252,
"projection": "layers.21.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0302,
"parameters": 4194304,
"reference_delta_l2": 1.3401222621873143,
"retained_delta_l2": 1.3232840437090565,
"delta_retention_norm_ratio": 0.9874353117223984,
"delta_cosine": 0.8233491223793983,
"relative_weight_rmse": 0.0145521763231229
},
{
"index": 106,
"total": 252,
"projection": "layers.22.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4182,
"parameters": 50331648,
"reference_delta_l2": 3.348253695373735,
"retained_delta_l2": 3.1108121498557035,
"delta_retention_norm_ratio": 0.9290849597669067,
"delta_cosine": 0.7466446635867791,
"relative_weight_rmse": 0.012576642698515664
},
{
"index": 107,
"total": 252,
"projection": "layers.22.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3841,
"parameters": 50331648,
"reference_delta_l2": 4.191437704163066,
"retained_delta_l2": 4.074424099330737,
"delta_retention_norm_ratio": 0.9720827045297352,
"delta_cosine": 0.8009936044647307,
"relative_weight_rmse": 0.014613979432748203
},
{
"index": 108,
"total": 252,
"projection": "layers.22.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3823,
"parameters": 50331648,
"reference_delta_l2": 3.4193685344857125,
"retained_delta_l2": 3.1681149826609203,
"delta_retention_norm_ratio": 0.9265204819864257,
"delta_cosine": 0.7446025566787715,
"relative_weight_rmse": 0.01260296600000183
},
{
"index": 109,
"total": 252,
"projection": "layers.22.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 1.1233588468588183,
"retained_delta_l2": 1.0978930305204808,
"delta_retention_norm_ratio": 0.9773306486973901,
"delta_cosine": 0.8174398193523686,
"relative_weight_rmse": 0.014397857497982244
},
{
"index": 110,
"total": 252,
"projection": "layers.22.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1292,
"parameters": 16777216,
"reference_delta_l2": 2.096172040574513,
"retained_delta_l2": 1.9955246542014025,
"delta_retention_norm_ratio": 0.951985149870845,
"delta_cosine": 0.7726040268200726,
"relative_weight_rmse": 0.01372359958810335
},
{
"index": 111,
"total": 252,
"projection": "layers.22.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1295,
"parameters": 16777216,
"reference_delta_l2": 1.8735820710146767,
"retained_delta_l2": 1.7449586608398842,
"delta_retention_norm_ratio": 0.9313489319925368,
"delta_cosine": 0.7552588343752609,
"relative_weight_rmse": 0.012605390978871518
},
{
"index": 112,
"total": 252,
"projection": "layers.22.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0298,
"parameters": 4194304,
"reference_delta_l2": 1.104789789987148,
"retained_delta_l2": 1.0530537624706828,
"delta_retention_norm_ratio": 0.9531711570967116,
"delta_cosine": 0.7782066073504973,
"relative_weight_rmse": 0.013219467528028158
},
{
"index": 113,
"total": 252,
"projection": "layers.23.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4151,
"parameters": 50331648,
"reference_delta_l2": 3.1573986783340944,
"retained_delta_l2": 2.8972070137585906,
"delta_retention_norm_ratio": 0.9175930279692186,
"delta_cosine": 0.7349246598047527,
"relative_weight_rmse": 0.011965726104044899
},
{
"index": 114,
"total": 252,
"projection": "layers.23.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3787,
"parameters": 50331648,
"reference_delta_l2": 5.5870157652651,
"retained_delta_l2": 5.626138092141853,
"delta_retention_norm_ratio": 1.007002365577699,
"delta_cosine": 0.851880484960273,
"relative_weight_rmse": 0.016744569548766984
},
{
"index": 115,
"total": 252,
"projection": "layers.23.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.382,
"parameters": 50331648,
"reference_delta_l2": 4.955648869435568,
"retained_delta_l2": 4.915349353509262,
"delta_retention_norm_ratio": 0.9918679638150198,
"delta_cosine": 0.8280099828088893,
"relative_weight_rmse": 0.015368761690623919
},
{
"index": 116,
"total": 252,
"projection": "layers.23.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 0.7949727976848021,
"retained_delta_l2": 0.7075521488332062,
"delta_retention_norm_ratio": 0.8900331569756965,
"delta_cosine": 0.7056548515418034,
"relative_weight_rmse": 0.011980711537248771
},
{
"index": 117,
"total": 252,
"projection": "layers.23.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1286,
"parameters": 16777216,
"reference_delta_l2": 1.9611178356180183,
"retained_delta_l2": 1.8341140758076107,
"delta_retention_norm_ratio": 0.9352390980777633,
"delta_cosine": 0.7547164028337787,
"relative_weight_rmse": 0.012811001876643429
},
{
"index": 118,
"total": 252,
"projection": "layers.23.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.129,
"parameters": 16777216,
"reference_delta_l2": 1.7866810382619926,
"retained_delta_l2": 1.6194285043128502,
"delta_retention_norm_ratio": 0.9063892601044008,
"delta_cosine": 0.7240963533143272,
"relative_weight_rmse": 0.012027247097238233
},
{
"index": 119,
"total": 252,
"projection": "layers.23.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 0.9268902828092771,
"retained_delta_l2": 0.8511800459627746,
"delta_retention_norm_ratio": 0.9183180164355212,
"delta_cosine": 0.7376135916253438,
"relative_weight_rmse": 0.011724739325830702
},
{
"index": 120,
"total": 252,
"projection": "layers.24.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4194,
"parameters": 50331648,
"reference_delta_l2": 3.248001361732338,
"retained_delta_l2": 2.9759227463417086,
"delta_retention_norm_ratio": 0.9162319885095382,
"delta_cosine": 0.731467528974081,
"relative_weight_rmse": 0.012265286006596474
},
{
"index": 121,
"total": 252,
"projection": "layers.24.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3798,
"parameters": 50331648,
"reference_delta_l2": 5.079159790012967,
"retained_delta_l2": 5.041899184536804,
"delta_retention_norm_ratio": 0.9926640218034828,
"delta_cosine": 0.8255122463800576,
"relative_weight_rmse": 0.016177434591443086
},
{
"index": 122,
"total": 252,
"projection": "layers.24.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3785,
"parameters": 50331648,
"reference_delta_l2": 4.002127200102585,
"retained_delta_l2": 3.8233247575193747,
"delta_retention_norm_ratio": 0.9553231485049684,
"delta_cosine": 0.7753053899270538,
"relative_weight_rmse": 0.013931788832164253
},
{
"index": 123,
"total": 252,
"projection": "layers.24.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0311,
"parameters": 4194304,
"reference_delta_l2": 0.9638034365683473,
"retained_delta_l2": 0.9140020321799122,
"delta_retention_norm_ratio": 0.9483282560542068,
"delta_cosine": 0.7718683986406603,
"relative_weight_rmse": 0.013132119891595104
},
{
"index": 124,
"total": 252,
"projection": "layers.24.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.128,
"parameters": 16777216,
"reference_delta_l2": 1.914972691650025,
"retained_delta_l2": 1.7753847481146188,
"delta_retention_norm_ratio": 0.9271070840100958,
"delta_cosine": 0.7456926850988511,
"relative_weight_rmse": 0.012568758413584789
},
{
"index": 125,
"total": 252,
"projection": "layers.24.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1284,
"parameters": 16777216,
"reference_delta_l2": 3.0094068047413,
"retained_delta_l2": 3.004943364043855,
"delta_retention_norm_ratio": 0.9985168370422991,
"delta_cosine": 0.8481736759550794,
"relative_weight_rmse": 0.015521565684129443
},
{
"index": 126,
"total": 252,
"projection": "layers.24.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0294,
"parameters": 4194304,
"reference_delta_l2": 1.0899244622607374,
"retained_delta_l2": 1.039031573136485,
"delta_retention_norm_ratio": 0.9533060401097068,
"delta_cosine": 0.7795839023899411,
"relative_weight_rmse": 0.013062099484726879
},
{
"index": 127,
"total": 252,
"projection": "layers.25.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4275,
"parameters": 50331648,
"reference_delta_l2": 3.633931408786495,
"retained_delta_l2": 3.400346578673652,
"delta_retention_norm_ratio": 0.9357211780200205,
"delta_cosine": 0.7512748559045608,
"relative_weight_rmse": 0.013255124553773066
},
{
"index": 128,
"total": 252,
"projection": "layers.25.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.382,
"parameters": 50331648,
"reference_delta_l2": 5.584503030567997,
"retained_delta_l2": 5.596796432252929,
"delta_retention_norm_ratio": 1.0022013421100573,
"delta_cosine": 0.8400906306425101,
"relative_weight_rmse": 0.016931961692208255
},
{
"index": 129,
"total": 252,
"projection": "layers.25.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3794,
"parameters": 50331648,
"reference_delta_l2": 3.737413926592672,
"retained_delta_l2": 3.4868664782660126,
"delta_retention_norm_ratio": 0.9329623495690565,
"delta_cosine": 0.7478532825383876,
"relative_weight_rmse": 0.013602370535550844
},
{
"index": 130,
"total": 252,
"projection": "layers.25.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 1.0272466552293884,
"retained_delta_l2": 0.9806886758883855,
"delta_retention_norm_ratio": 0.9546769229142866,
"delta_cosine": 0.7797080882376601,
"relative_weight_rmse": 0.013532519417488998
},
{
"index": 131,
"total": 252,
"projection": "layers.25.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1281,
"parameters": 16777216,
"reference_delta_l2": 1.9527008465824647,
"retained_delta_l2": 1.8050283587839864,
"delta_retention_norm_ratio": 0.9243752630839852,
"delta_cosine": 0.7397487044160916,
"relative_weight_rmse": 0.01318621053304837
},
{
"index": 132,
"total": 252,
"projection": "layers.25.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1289,
"parameters": 16777216,
"reference_delta_l2": 2.450536633916322,
"retained_delta_l2": 2.4029435739095537,
"delta_retention_norm_ratio": 0.9805785151921163,
"delta_cosine": 0.8153367447804893,
"relative_weight_rmse": 0.014591689306671735
},
{
"index": 133,
"total": 252,
"projection": "layers.25.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0299,
"parameters": 4194304,
"reference_delta_l2": 1.185708349706415,
"retained_delta_l2": 1.1420416359025602,
"delta_retention_norm_ratio": 0.9631724666401593,
"delta_cosine": 0.7868584096539307,
"relative_weight_rmse": 0.01344112341912221
},
{
"index": 134,
"total": 252,
"projection": "layers.26.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4184,
"parameters": 50331648,
"reference_delta_l2": 3.2898090617020626,
"retained_delta_l2": 3.0098549097248215,
"delta_retention_norm_ratio": 0.914902613882278,
"delta_cosine": 0.7278429398237305,
"relative_weight_rmse": 0.01232353824091038
},
{
"index": 135,
"total": 252,
"projection": "layers.26.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3781,
"parameters": 50331648,
"reference_delta_l2": 5.26015590203999,
"retained_delta_l2": 5.222621517173403,
"delta_retention_norm_ratio": 0.992864396879942,
"delta_cosine": 0.8222124191353445,
"relative_weight_rmse": 0.016653677789210033
},
{
"index": 136,
"total": 252,
"projection": "layers.26.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3846,
"parameters": 50331648,
"reference_delta_l2": 3.590069458954333,
"retained_delta_l2": 3.316763413324665,
"delta_retention_norm_ratio": 0.9238716551993196,
"delta_cosine": 0.7375929700677254,
"relative_weight_rmse": 0.01314661682087519
},
{
"index": 137,
"total": 252,
"projection": "layers.26.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.032,
"parameters": 4194304,
"reference_delta_l2": 1.1020804593860358,
"retained_delta_l2": 1.0637554874409914,
"delta_retention_norm_ratio": 0.9652248875128454,
"delta_cosine": 0.7913003888851425,
"relative_weight_rmse": 0.014439016768790466
},
{
"index": 138,
"total": 252,
"projection": "layers.26.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1279,
"parameters": 16777216,
"reference_delta_l2": 2.1431077000425294,
"retained_delta_l2": 2.0182462766227127,
"delta_retention_norm_ratio": 0.9417381481027113,
"delta_cosine": 0.7600793909340787,
"relative_weight_rmse": 0.013803256785379104
},
{
"index": 139,
"total": 252,
"projection": "layers.26.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1277,
"parameters": 16777216,
"reference_delta_l2": 2.719952969878672,
"retained_delta_l2": 2.696589776484013,
"delta_retention_norm_ratio": 0.9914104421461003,
"delta_cosine": 0.8326224887262295,
"relative_weight_rmse": 0.015373628702985906
},
{
"index": 140,
"total": 252,
"projection": "layers.26.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.03,
"parameters": 4194304,
"reference_delta_l2": 1.300620124236764,
"retained_delta_l2": 1.2651561449286852,
"delta_retention_norm_ratio": 0.972733022773356,
"delta_cosine": 0.8066927122654023,
"relative_weight_rmse": 0.01426096093375654
},
{
"index": 141,
"total": 252,
"projection": "layers.27.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4175,
"parameters": 50331648,
"reference_delta_l2": 3.1127612376333307,
"retained_delta_l2": 2.784529759790215,
"delta_retention_norm_ratio": 0.8945529538614166,
"delta_cosine": 0.706960318732661,
"relative_weight_rmse": 0.011886623254813496
},
{
"index": 142,
"total": 252,
"projection": "layers.27.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3814,
"parameters": 50331648,
"reference_delta_l2": 5.218309481668899,
"retained_delta_l2": 5.184642749566007,
"delta_retention_norm_ratio": 0.9935483450682336,
"delta_cosine": 0.8246295446084377,
"relative_weight_rmse": 0.016435645092332085
},
{
"index": 143,
"total": 252,
"projection": "layers.27.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3802,
"parameters": 50331648,
"reference_delta_l2": 3.1371583611558593,
"retained_delta_l2": 2.753390771113731,
"delta_retention_norm_ratio": 0.8776703162983676,
"delta_cosine": 0.6894176588930889,
"relative_weight_rmse": 0.012165186836253738
},
{
"index": 144,
"total": 252,
"projection": "layers.27.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0318,
"parameters": 4194304,
"reference_delta_l2": 0.9566006062182262,
"retained_delta_l2": 0.9004468825838453,
"delta_retention_norm_ratio": 0.9412986744213178,
"delta_cosine": 0.7663775376414128,
"relative_weight_rmse": 0.01348690912064083
},
{
"index": 145,
"total": 252,
"projection": "layers.27.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.129,
"parameters": 16777216,
"reference_delta_l2": 1.8313920593686184,
"retained_delta_l2": 1.6663060630742128,
"delta_retention_norm_ratio": 0.9098576432883957,
"delta_cosine": 0.723637594500945,
"relative_weight_rmse": 0.01273018359232453
},
{
"index": 146,
"total": 252,
"projection": "layers.27.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1278,
"parameters": 16777216,
"reference_delta_l2": 2.505769969980535,
"retained_delta_l2": 2.4511339349255747,
"delta_retention_norm_ratio": 0.9781959095569397,
"delta_cosine": 0.8134119135818588,
"relative_weight_rmse": 0.014897494871373889
},
{
"index": 147,
"total": 252,
"projection": "layers.27.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0301,
"parameters": 4194304,
"reference_delta_l2": 1.3077336076957404,
"retained_delta_l2": 1.2847687144409412,
"delta_retention_norm_ratio": 0.9824391656529621,
"delta_cosine": 0.8145542778588396,
"relative_weight_rmse": 0.014327868666712023
},
{
"index": 148,
"total": 252,
"projection": "layers.28.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4227,
"parameters": 50331648,
"reference_delta_l2": 3.753788447028095,
"retained_delta_l2": 3.538348373928671,
"delta_retention_norm_ratio": 0.9426072949662387,
"delta_cosine": 0.7606475857732973,
"relative_weight_rmse": 0.01308095261202953
},
{
"index": 149,
"total": 252,
"projection": "layers.28.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3851,
"parameters": 50331648,
"reference_delta_l2": 6.384628728464536,
"retained_delta_l2": 6.495170579397263,
"delta_retention_norm_ratio": 1.0173137476952887,
"delta_cosine": 0.8657629496281001,
"relative_weight_rmse": 0.01790649059899249
},
{
"index": 150,
"total": 252,
"projection": "layers.28.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3786,
"parameters": 50331648,
"reference_delta_l2": 3.236737353783521,
"retained_delta_l2": 2.867916332999228,
"delta_retention_norm_ratio": 0.8860516067659409,
"delta_cosine": 0.6973045633320863,
"relative_weight_rmse": 0.012307524760496754
},
{
"index": 151,
"total": 252,
"projection": "layers.28.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0308,
"parameters": 4194304,
"reference_delta_l2": 1.032288943294731,
"retained_delta_l2": 0.9823748789222145,
"delta_retention_norm_ratio": 0.9516471965560273,
"delta_cosine": 0.7710451408030818,
"relative_weight_rmse": 0.014138985070877371
},
{
"index": 152,
"total": 252,
"projection": "layers.28.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1289,
"parameters": 16777216,
"reference_delta_l2": 1.8651147118672111,
"retained_delta_l2": 1.7130419963641268,
"delta_retention_norm_ratio": 0.918464685021523,
"delta_cosine": 0.733442060333026,
"relative_weight_rmse": 0.012824160612533261
},
{
"index": 153,
"total": 252,
"projection": "layers.28.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1271,
"parameters": 16777216,
"reference_delta_l2": 2.105314988338305,
"retained_delta_l2": 2.0052706387196078,
"delta_retention_norm_ratio": 0.9524801038453345,
"delta_cosine": 0.7768201327216553,
"relative_weight_rmse": 0.013244203073642227
},
{
"index": 154,
"total": 252,
"projection": "layers.28.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 1.057093412456726,
"retained_delta_l2": 0.9882311455489668,
"delta_retention_norm_ratio": 0.9348569709201757,
"delta_cosine": 0.7521307203524212,
"relative_weight_rmse": 0.012581262930721628
},
{
"index": 155,
"total": 252,
"projection": "layers.29.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4232,
"parameters": 50331648,
"reference_delta_l2": 3.7318718982283765,
"retained_delta_l2": 3.50050716333633,
"delta_retention_norm_ratio": 0.93800303408006,
"delta_cosine": 0.7553819010153049,
"relative_weight_rmse": 0.012980868513796524
},
{
"index": 156,
"total": 252,
"projection": "layers.29.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.382,
"parameters": 50331648,
"reference_delta_l2": 6.111839640398802,
"retained_delta_l2": 6.194424768364596,
"delta_retention_norm_ratio": 1.013512319174723,
"delta_cosine": 0.8563557745027518,
"relative_weight_rmse": 0.017696393880623638
},
{
"index": 157,
"total": 252,
"projection": "layers.29.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3792,
"parameters": 50331648,
"reference_delta_l2": 3.574412188208577,
"retained_delta_l2": 3.273411412573307,
"delta_retention_norm_ratio": 0.9157901328145018,
"delta_cosine": 0.7289595334981157,
"relative_weight_rmse": 0.012884148207696703
},
{
"index": 158,
"total": 252,
"projection": "layers.29.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 0.9613778159833816,
"retained_delta_l2": 0.9108547548183007,
"delta_retention_norm_ratio": 0.9474472363257088,
"delta_cosine": 0.7814741530933297,
"relative_weight_rmse": 0.013130078542218879
},
{
"index": 159,
"total": 252,
"projection": "layers.29.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1272,
"parameters": 16777216,
"reference_delta_l2": 1.9429901539976733,
"retained_delta_l2": 1.7908867120716394,
"delta_retention_norm_ratio": 0.9217168231073722,
"delta_cosine": 0.737270406241436,
"relative_weight_rmse": 0.012918803561957477
},
{
"index": 160,
"total": 252,
"projection": "layers.29.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1305,
"parameters": 16777216,
"reference_delta_l2": 2.940179549700112,
"retained_delta_l2": 2.9265865240914684,
"delta_retention_norm_ratio": 0.9953768042465193,
"delta_cosine": 0.8475101752498819,
"relative_weight_rmse": 0.015271158765365453
},
{
"index": 161,
"total": 252,
"projection": "layers.29.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0298,
"parameters": 4194304,
"reference_delta_l2": 1.0698147715634443,
"retained_delta_l2": 1.0104317699671235,
"delta_retention_norm_ratio": 0.9444922586836808,
"delta_cosine": 0.7676017515607246,
"relative_weight_rmse": 0.012957966998483238
},
{
"index": 162,
"total": 252,
"projection": "layers.3.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4196,
"parameters": 50331648,
"reference_delta_l2": 2.0469881021138128,
"retained_delta_l2": 1.8067919142813011,
"delta_retention_norm_ratio": 0.8826587279210494,
"delta_cosine": 0.7129162638492356,
"relative_weight_rmse": 0.011281224627702802
},
{
"index": 163,
"total": 252,
"projection": "layers.3.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3811,
"parameters": 50331648,
"reference_delta_l2": 2.922833700925656,
"retained_delta_l2": 2.701297923257113,
"delta_retention_norm_ratio": 0.9242051377748987,
"delta_cosine": 0.7456390004606592,
"relative_weight_rmse": 0.01225232535469193
},
{
"index": 164,
"total": 252,
"projection": "layers.3.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3803,
"parameters": 50331648,
"reference_delta_l2": 3.170575036365232,
"retained_delta_l2": 3.1039285920713064,
"delta_retention_norm_ratio": 0.9789796981526954,
"delta_cosine": 0.8417859713352379,
"relative_weight_rmse": 0.014662548426467735
},
{
"index": 165,
"total": 252,
"projection": "layers.3.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 0.5928153003252811,
"retained_delta_l2": 0.45925093149985696,
"delta_retention_norm_ratio": 0.7746948016487823,
"delta_cosine": 0.5960206976001503,
"relative_weight_rmse": 0.009616935234874856
},
{
"index": 166,
"total": 252,
"projection": "layers.3.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1286,
"parameters": 16777216,
"reference_delta_l2": 1.8492563116636087,
"retained_delta_l2": 1.6954448559515825,
"delta_retention_norm_ratio": 0.9168252368576988,
"delta_cosine": 0.7302789316434527,
"relative_weight_rmse": 0.012809784802966738
},
{
"index": 167,
"total": 252,
"projection": "layers.3.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1293,
"parameters": 16777216,
"reference_delta_l2": 1.506985172201776,
"retained_delta_l2": 1.3041041112765286,
"delta_retention_norm_ratio": 0.8653728884214378,
"delta_cosine": 0.6787827891036176,
"relative_weight_rmse": 0.011717372717298042
},
{
"index": 168,
"total": 252,
"projection": "layers.3.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0303,
"parameters": 4194304,
"reference_delta_l2": 0.9484510766660493,
"retained_delta_l2": 0.8550643812206361,
"delta_retention_norm_ratio": 0.9015376778592716,
"delta_cosine": 0.7159253627656402,
"relative_weight_rmse": 0.012402602443049677
},
{
"index": 169,
"total": 252,
"projection": "layers.30.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4241,
"parameters": 50331648,
"reference_delta_l2": 3.881081942741242,
"retained_delta_l2": 3.679213410037991,
"delta_retention_norm_ratio": 0.9479865316730031,
"delta_cosine": 0.7660717988774276,
"relative_weight_rmse": 0.013128896619984549
},
{
"index": 170,
"total": 252,
"projection": "layers.30.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3852,
"parameters": 50331648,
"reference_delta_l2": 5.385688107680878,
"retained_delta_l2": 5.381089580955163,
"delta_retention_norm_ratio": 0.9991461579961981,
"delta_cosine": 0.8302506979760844,
"relative_weight_rmse": 0.016908597417800836
},
{
"index": 171,
"total": 252,
"projection": "layers.30.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3793,
"parameters": 50331648,
"reference_delta_l2": 3.7886798447697156,
"retained_delta_l2": 3.5044195098186384,
"delta_retention_norm_ratio": 0.924971138602936,
"delta_cosine": 0.7381686108121072,
"relative_weight_rmse": 0.01336324497183794
},
{
"index": 172,
"total": 252,
"projection": "layers.30.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0321,
"parameters": 4194304,
"reference_delta_l2": 1.4967965365370788,
"retained_delta_l2": 1.5084521827425899,
"delta_retention_norm_ratio": 1.007787061180992,
"delta_cosine": 0.8529388802167588,
"relative_weight_rmse": 0.01661322374914887
},
{
"index": 173,
"total": 252,
"projection": "layers.30.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1293,
"parameters": 16777216,
"reference_delta_l2": 1.8641895953134608,
"retained_delta_l2": 1.687691231712539,
"delta_retention_norm_ratio": 0.9053216668279688,
"delta_cosine": 0.7170319774495255,
"relative_weight_rmse": 0.01284219142237025
},
{
"index": 174,
"total": 252,
"projection": "layers.30.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1265,
"parameters": 16777216,
"reference_delta_l2": 3.01325671683292,
"retained_delta_l2": 3.0238985865702173,
"delta_retention_norm_ratio": 1.003531683735358,
"delta_cosine": 0.854792895207731,
"relative_weight_rmse": 0.01627868799924678
},
{
"index": 175,
"total": 252,
"projection": "layers.30.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 1.3092454472265989,
"retained_delta_l2": 1.2711618368957232,
"delta_retention_norm_ratio": 0.9709117870819799,
"delta_cosine": 0.8013761680772169,
"relative_weight_rmse": 0.013920487605929862
},
{
"index": 176,
"total": 252,
"projection": "layers.31.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4169,
"parameters": 50331648,
"reference_delta_l2": 4.1084550327590295,
"retained_delta_l2": 3.9426855677034225,
"delta_retention_norm_ratio": 0.9596516296919807,
"delta_cosine": 0.7804069254305991,
"relative_weight_rmse": 0.013375184816653753
},
{
"index": 177,
"total": 252,
"projection": "layers.31.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.384,
"parameters": 50331648,
"reference_delta_l2": 4.204636558660181,
"retained_delta_l2": 4.054791818793385,
"delta_retention_norm_ratio": 0.9643620232625897,
"delta_cosine": 0.7852028074197616,
"relative_weight_rmse": 0.014739784107740565
},
{
"index": 178,
"total": 252,
"projection": "layers.31.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3806,
"parameters": 50331648,
"reference_delta_l2": 3.7410814813654816,
"retained_delta_l2": 3.452811685816432,
"delta_retention_norm_ratio": 0.922944796314933,
"delta_cosine": 0.7364122265675088,
"relative_weight_rmse": 0.01311554283213831
},
{
"index": 179,
"total": 252,
"projection": "layers.31.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0308,
"parameters": 4194304,
"reference_delta_l2": 1.1323445033739201,
"retained_delta_l2": 1.1086924409108212,
"delta_retention_norm_ratio": 0.9791123086722939,
"delta_cosine": 0.8130617153573887,
"relative_weight_rmse": 0.014444677927331962
},
{
"index": 180,
"total": 252,
"projection": "layers.31.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1322,
"parameters": 16777216,
"reference_delta_l2": 1.9231706062594072,
"retained_delta_l2": 1.7740289106067064,
"delta_retention_norm_ratio": 0.9224500961239297,
"delta_cosine": 0.7383528409683014,
"relative_weight_rmse": 0.01252349737431234
},
{
"index": 181,
"total": 252,
"projection": "layers.31.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1269,
"parameters": 16777216,
"reference_delta_l2": 3.0864807513581716,
"retained_delta_l2": 3.1091758585622844,
"delta_retention_norm_ratio": 1.007353069412186,
"delta_cosine": 0.8588801820594556,
"relative_weight_rmse": 0.01633773267765145
},
{
"index": 182,
"total": 252,
"projection": "layers.31.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 0.911606139216967,
"retained_delta_l2": 0.8037103391056234,
"delta_retention_norm_ratio": 0.8816420870047872,
"delta_cosine": 0.698835520254118,
"relative_weight_rmse": 0.011235475360537445
},
{
"index": 183,
"total": 252,
"projection": "layers.32.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4267,
"parameters": 50331648,
"reference_delta_l2": 3.0495582576332665,
"retained_delta_l2": 2.6967207646394122,
"delta_retention_norm_ratio": 0.8842988186532668,
"delta_cosine": 0.6980275932606044,
"relative_weight_rmse": 0.01127934914427212
},
{
"index": 184,
"total": 252,
"projection": "layers.32.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.382,
"parameters": 50331648,
"reference_delta_l2": 5.692511022736389,
"retained_delta_l2": 5.739681946755911,
"delta_retention_norm_ratio": 1.0082864879542819,
"delta_cosine": 0.8468962010552334,
"relative_weight_rmse": 0.017428843321927585
},
{
"index": 185,
"total": 252,
"projection": "layers.32.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3842,
"parameters": 50331648,
"reference_delta_l2": 4.1613233155376745,
"retained_delta_l2": 3.9427416482145152,
"delta_retention_norm_ratio": 0.9474730390433705,
"delta_cosine": 0.7663476507769479,
"relative_weight_rmse": 0.013800835891177412
},
{
"index": 186,
"total": 252,
"projection": "layers.32.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0325,
"parameters": 4194304,
"reference_delta_l2": 1.0749308514731222,
"retained_delta_l2": 1.0404449056170206,
"delta_retention_norm_ratio": 0.9679179867162238,
"delta_cosine": 0.7986748314670193,
"relative_weight_rmse": 0.014078454901238931
},
{
"index": 187,
"total": 252,
"projection": "layers.32.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1296,
"parameters": 16777216,
"reference_delta_l2": 1.9521065727725724,
"retained_delta_l2": 1.794087986749361,
"delta_retention_norm_ratio": 0.9190522750001411,
"delta_cosine": 0.7326878381179381,
"relative_weight_rmse": 0.012903346914620015
},
{
"index": 188,
"total": 252,
"projection": "layers.32.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1304,
"parameters": 16777216,
"reference_delta_l2": 3.489609887071843,
"retained_delta_l2": 3.54785274581623,
"delta_retention_norm_ratio": 1.0166903638599154,
"delta_cosine": 0.8830956089103636,
"relative_weight_rmse": 0.016991135952278535
},
{
"index": 189,
"total": 252,
"projection": "layers.32.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 1.1042808294327686,
"retained_delta_l2": 1.034127379772928,
"delta_retention_norm_ratio": 0.9364713687043936,
"delta_cosine": 0.7578150313742703,
"relative_weight_rmse": 0.012600668825086671
},
{
"index": 190,
"total": 252,
"projection": "layers.33.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4231,
"parameters": 50331648,
"reference_delta_l2": 4.666259945277719,
"retained_delta_l2": 4.578391719038913,
"delta_retention_norm_ratio": 0.9811694532089389,
"delta_cosine": 0.8093000623021638,
"relative_weight_rmse": 0.014381902234994248
},
{
"index": 191,
"total": 252,
"projection": "layers.33.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3827,
"parameters": 50331648,
"reference_delta_l2": 3.9832469507356367,
"retained_delta_l2": 3.7956235365409667,
"delta_retention_norm_ratio": 0.9528968661709465,
"delta_cosine": 0.7697364410652019,
"relative_weight_rmse": 0.014686245571898645
},
{
"index": 192,
"total": 252,
"projection": "layers.33.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3804,
"parameters": 50331648,
"reference_delta_l2": 4.536695712695747,
"retained_delta_l2": 4.397943606116111,
"delta_retention_norm_ratio": 0.969415602154814,
"delta_cosine": 0.7947863676826681,
"relative_weight_rmse": 0.014203867204328913
},
{
"index": 193,
"total": 252,
"projection": "layers.33.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 1.171700695902796,
"retained_delta_l2": 1.1623095407540411,
"delta_retention_norm_ratio": 0.9919850221292913,
"delta_cosine": 0.827751229147002,
"relative_weight_rmse": 0.01555032560869696
},
{
"index": 194,
"total": 252,
"projection": "layers.33.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1288,
"parameters": 16777216,
"reference_delta_l2": 1.9538396201038633,
"retained_delta_l2": 1.8052516114707444,
"delta_retention_norm_ratio": 0.9239507648917366,
"delta_cosine": 0.7410008753100303,
"relative_weight_rmse": 0.012285579486915223
},
{
"index": 195,
"total": 252,
"projection": "layers.33.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1299,
"parameters": 16777216,
"reference_delta_l2": 1.6938910069946165,
"retained_delta_l2": 1.5612559026371533,
"delta_retention_norm_ratio": 0.9216979700525179,
"delta_cosine": 0.7425285502146433,
"relative_weight_rmse": 0.011916875261304336
},
{
"index": 196,
"total": 252,
"projection": "layers.33.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 1.2353249119776444,
"retained_delta_l2": 1.186066026793788,
"delta_retention_norm_ratio": 0.9601247536528691,
"delta_cosine": 0.7866217684692136,
"relative_weight_rmse": 0.012733576318062864
},
{
"index": 197,
"total": 252,
"projection": "layers.34.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4203,
"parameters": 50331648,
"reference_delta_l2": 3.250888771198972,
"retained_delta_l2": 3.0194243808532835,
"delta_retention_norm_ratio": 0.9287996586052616,
"delta_cosine": 0.7451810336261245,
"relative_weight_rmse": 0.011829284789644647
},
{
"index": 198,
"total": 252,
"projection": "layers.34.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3786,
"parameters": 50331648,
"reference_delta_l2": 3.5072204020963986,
"retained_delta_l2": 3.3023841588143523,
"delta_retention_norm_ratio": 0.9415958451999173,
"delta_cosine": 0.7566511228221033,
"relative_weight_rmse": 0.013095487324174874
},
{
"index": 199,
"total": 252,
"projection": "layers.34.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3817,
"parameters": 50331648,
"reference_delta_l2": 3.1199293402159056,
"retained_delta_l2": 2.7957443046500536,
"delta_retention_norm_ratio": 0.896092186644388,
"delta_cosine": 0.7096322687115229,
"relative_weight_rmse": 0.011414938266865215
},
{
"index": 200,
"total": 252,
"projection": "layers.34.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0316,
"parameters": 4194304,
"reference_delta_l2": 0.788029415327778,
"retained_delta_l2": 0.7265886893910003,
"delta_retention_norm_ratio": 0.9220324460715446,
"delta_cosine": 0.7525129228802909,
"relative_weight_rmse": 0.011620429007118772
},
{
"index": 201,
"total": 252,
"projection": "layers.34.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1281,
"parameters": 16777216,
"reference_delta_l2": 2.3613394109958232,
"retained_delta_l2": 2.289951534807643,
"delta_retention_norm_ratio": 0.9697680579692377,
"delta_cosine": 0.7925439024667102,
"relative_weight_rmse": 0.01402741762220263
},
{
"index": 202,
"total": 252,
"projection": "layers.34.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1288,
"parameters": 16777216,
"reference_delta_l2": 2.785871038243405,
"retained_delta_l2": 2.7988799538483624,
"delta_retention_norm_ratio": 1.004669604380955,
"delta_cosine": 0.8594569487442454,
"relative_weight_rmse": 0.015266239440760419
},
{
"index": 203,
"total": 252,
"projection": "layers.34.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 0.9814467233633697,
"retained_delta_l2": 0.8768093993968407,
"delta_retention_norm_ratio": 0.8933846112319351,
"delta_cosine": 0.7153985705871384,
"relative_weight_rmse": 0.011405559396125359
},
{
"index": 204,
"total": 252,
"projection": "layers.35.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4213,
"parameters": 50331648,
"reference_delta_l2": 2.3183374458505197,
"retained_delta_l2": 2.0695365530536636,
"delta_retention_norm_ratio": 0.8926813293542867,
"delta_cosine": 0.7130042803163568,
"relative_weight_rmse": 0.00977575248991696
},
{
"index": 205,
"total": 252,
"projection": "layers.35.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3849,
"parameters": 50331648,
"reference_delta_l2": 2.614786567527581,
"retained_delta_l2": 2.279114457992027,
"delta_retention_norm_ratio": 0.871625426830554,
"delta_cosine": 0.686332682020759,
"relative_weight_rmse": 0.010501353670140995
},
{
"index": 206,
"total": 252,
"projection": "layers.35.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3791,
"parameters": 50331648,
"reference_delta_l2": 2.6518225242333027,
"retained_delta_l2": 2.27767724625173,
"delta_retention_norm_ratio": 0.8589101364957498,
"delta_cosine": 0.6736036409544919,
"relative_weight_rmse": 0.010318953190741935
},
{
"index": 207,
"total": 252,
"projection": "layers.35.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0309,
"parameters": 4194304,
"reference_delta_l2": 0.8334287809981532,
"retained_delta_l2": 0.7635522219713372,
"delta_retention_norm_ratio": 0.9161577322262274,
"delta_cosine": 0.73447000352539,
"relative_weight_rmse": 0.012233826186733552
},
{
"index": 208,
"total": 252,
"projection": "layers.35.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.13,
"parameters": 16777216,
"reference_delta_l2": 1.9820359827690113,
"retained_delta_l2": 1.8810726256763255,
"delta_retention_norm_ratio": 0.9490607849855306,
"delta_cosine": 0.7679138582696287,
"relative_weight_rmse": 0.013001248772620494
},
{
"index": 209,
"total": 252,
"projection": "layers.35.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1264,
"parameters": 16777216,
"reference_delta_l2": 1.9167783776293494,
"retained_delta_l2": 1.8284949611298134,
"delta_retention_norm_ratio": 0.9539417715006134,
"delta_cosine": 0.7741266142645553,
"relative_weight_rmse": 0.01306133912536974
},
{
"index": 210,
"total": 252,
"projection": "layers.35.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 0.9045634157104502,
"retained_delta_l2": 0.7983614933900623,
"delta_retention_norm_ratio": 0.8825931709420548,
"delta_cosine": 0.6994521063502559,
"relative_weight_rmse": 0.011420838616268927
},
{
"index": 211,
"total": 252,
"projection": "layers.4.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4191,
"parameters": 50331648,
"reference_delta_l2": 2.221336647346869,
"retained_delta_l2": 1.9045813778894223,
"delta_retention_norm_ratio": 0.8574033027205603,
"delta_cosine": 0.6727181068599302,
"relative_weight_rmse": 0.010822728427872375
},
{
"index": 212,
"total": 252,
"projection": "layers.4.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.381,
"parameters": 50331648,
"reference_delta_l2": 3.309274307029437,
"retained_delta_l2": 2.9183820142372894,
"delta_retention_norm_ratio": 0.8818797547360071,
"delta_cosine": 0.7088330041186263,
"relative_weight_rmse": 0.010818316989641936
},
{
"index": 213,
"total": 252,
"projection": "layers.4.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3791,
"parameters": 50331648,
"reference_delta_l2": 3.3002149367491143,
"retained_delta_l2": 3.1780020825172097,
"delta_retention_norm_ratio": 0.9629682137151072,
"delta_cosine": 0.7953073969949832,
"relative_weight_rmse": 0.014014920542321037
},
{
"index": 214,
"total": 252,
"projection": "layers.4.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0321,
"parameters": 4194304,
"reference_delta_l2": 0.5923128812842459,
"retained_delta_l2": 0.44706547078012393,
"delta_retention_norm_ratio": 0.7547792474322047,
"delta_cosine": 0.5777734174137593,
"relative_weight_rmse": 0.00945310317345178
},
{
"index": 215,
"total": 252,
"projection": "layers.4.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1283,
"parameters": 16777216,
"reference_delta_l2": 1.6233424860329353,
"retained_delta_l2": 1.4087840092320407,
"delta_retention_norm_ratio": 0.8678291989232508,
"delta_cosine": 0.6799719691456307,
"relative_weight_rmse": 0.011781348708998187
},
{
"index": 216,
"total": 252,
"projection": "layers.4.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1281,
"parameters": 16777216,
"reference_delta_l2": 1.2599063228139533,
"retained_delta_l2": 0.9953410590893825,
"delta_retention_norm_ratio": 0.7900119564971512,
"delta_cosine": 0.607803419298993,
"relative_weight_rmse": 0.010232583667853115
},
{
"index": 217,
"total": 252,
"projection": "layers.4.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0301,
"parameters": 4194304,
"reference_delta_l2": 1.1362098663278084,
"retained_delta_l2": 1.0785328669695842,
"delta_retention_norm_ratio": 0.949237371486102,
"delta_cosine": 0.7692033108834336,
"relative_weight_rmse": 0.013542298508869052
},
{
"index": 218,
"total": 252,
"projection": "layers.5.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4233,
"parameters": 50331648,
"reference_delta_l2": 1.803962899947611,
"retained_delta_l2": 1.342634588105705,
"delta_retention_norm_ratio": 0.744269512496458,
"delta_cosine": 0.5694831159509246,
"relative_weight_rmse": 0.009042785321309456
},
{
"index": 219,
"total": 252,
"projection": "layers.5.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3847,
"parameters": 50331648,
"reference_delta_l2": 2.9502308395883783,
"retained_delta_l2": 2.4860241230127524,
"delta_retention_norm_ratio": 0.8426541034191098,
"delta_cosine": 0.66015731037967,
"relative_weight_rmse": 0.009986779270096871
},
{
"index": 220,
"total": 252,
"projection": "layers.5.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3806,
"parameters": 50331648,
"reference_delta_l2": 3.173490918626583,
"retained_delta_l2": 2.9589975451196926,
"delta_retention_norm_ratio": 0.9324109067878731,
"delta_cosine": 0.7491525974044665,
"relative_weight_rmse": 0.013117147848501721
},
{
"index": 221,
"total": 252,
"projection": "layers.5.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 0.6193610286474956,
"retained_delta_l2": 0.487339283051826,
"delta_retention_norm_ratio": 0.7868420202608376,
"delta_cosine": 0.6077146773250078,
"relative_weight_rmse": 0.009850471449688411
},
{
"index": 222,
"total": 252,
"projection": "layers.5.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1345,
"parameters": 16777216,
"reference_delta_l2": 1.4660458596862542,
"retained_delta_l2": 1.2308176050578241,
"delta_retention_norm_ratio": 0.8395491838988101,
"delta_cosine": 0.6516145241740764,
"relative_weight_rmse": 0.011341592423666287
},
{
"index": 223,
"total": 252,
"projection": "layers.5.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1283,
"parameters": 16777216,
"reference_delta_l2": 1.4671626671458942,
"retained_delta_l2": 1.252449986018564,
"delta_retention_norm_ratio": 0.8536544815817759,
"delta_cosine": 0.6701923660259413,
"relative_weight_rmse": 0.011180446979722173
},
{
"index": 224,
"total": 252,
"projection": "layers.5.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0296,
"parameters": 4194304,
"reference_delta_l2": 1.044090832035962,
"retained_delta_l2": 0.9792640619326973,
"delta_retention_norm_ratio": 0.9379107946222902,
"delta_cosine": 0.760232127059494,
"relative_weight_rmse": 0.012863542624411998
},
{
"index": 225,
"total": 252,
"projection": "layers.6.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4213,
"parameters": 50331648,
"reference_delta_l2": 4.766108258771339,
"retained_delta_l2": 4.77873869719586,
"delta_retention_norm_ratio": 1.0026500527765556,
"delta_cosine": 0.8442033313899234,
"relative_weight_rmse": 0.01543176975270308
},
{
"index": 226,
"total": 252,
"projection": "layers.6.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3839,
"parameters": 50331648,
"reference_delta_l2": 3.1029703699387263,
"retained_delta_l2": 2.712238549129781,
"delta_retention_norm_ratio": 0.874078133457439,
"delta_cosine": 0.6918195489369741,
"relative_weight_rmse": 0.010674646842871805
},
{
"index": 227,
"total": 252,
"projection": "layers.6.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3821,
"parameters": 50331648,
"reference_delta_l2": 3.3821355337004557,
"retained_delta_l2": 3.1934349534654216,
"delta_retention_norm_ratio": 0.9442066770078333,
"delta_cosine": 0.7628082158548107,
"relative_weight_rmse": 0.013209469970077926
},
{
"index": 228,
"total": 252,
"projection": "layers.6.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0315,
"parameters": 4194304,
"reference_delta_l2": 0.5910609699867302,
"retained_delta_l2": 0.4457745114424727,
"delta_retention_norm_ratio": 0.7541937872373483,
"delta_cosine": 0.5775529443999605,
"relative_weight_rmse": 0.009558571249523508
},
{
"index": 229,
"total": 252,
"projection": "layers.6.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1303,
"parameters": 16777216,
"reference_delta_l2": 1.3625651734345294,
"retained_delta_l2": 1.1040539958059539,
"delta_retention_norm_ratio": 0.8102761007923289,
"delta_cosine": 0.626134758602679,
"relative_weight_rmse": 0.01038159614631551
},
{
"index": 230,
"total": 252,
"projection": "layers.6.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1292,
"parameters": 16777216,
"reference_delta_l2": 1.4025439238449944,
"retained_delta_l2": 1.1632560404983912,
"delta_retention_norm_ratio": 0.8293900966105867,
"delta_cosine": 0.6446595803979036,
"relative_weight_rmse": 0.010718242315551555
},
{
"index": 231,
"total": 252,
"projection": "layers.6.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.03,
"parameters": 4194304,
"reference_delta_l2": 0.9850880555955308,
"retained_delta_l2": 0.9089428205266216,
"delta_retention_norm_ratio": 0.9227021029882695,
"delta_cosine": 0.7391486772354445,
"relative_weight_rmse": 0.012392920894060367
},
{
"index": 232,
"total": 252,
"projection": "layers.7.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4368,
"parameters": 50331648,
"reference_delta_l2": 1.8440352305241692,
"retained_delta_l2": 1.3640679381239535,
"delta_retention_norm_ratio": 0.7397190224702028,
"delta_cosine": 0.5653753919522522,
"relative_weight_rmse": 0.008955389570752136
},
{
"index": 233,
"total": 252,
"projection": "layers.7.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4066,
"parameters": 50331648,
"reference_delta_l2": 2.893335151708075,
"retained_delta_l2": 2.477994315015933,
"delta_retention_norm_ratio": 0.8564491097939531,
"delta_cosine": 0.673437015493994,
"relative_weight_rmse": 0.010343431056423151
},
{
"index": 234,
"total": 252,
"projection": "layers.7.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3857,
"parameters": 50331648,
"reference_delta_l2": 2.967222221507067,
"retained_delta_l2": 2.682568633972963,
"delta_retention_norm_ratio": 0.9040673174153006,
"delta_cosine": 0.7157974740017237,
"relative_weight_rmse": 0.012361975496102632
},
{
"index": 235,
"total": 252,
"projection": "layers.7.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0318,
"parameters": 4194304,
"reference_delta_l2": 0.7208766702180713,
"retained_delta_l2": 0.6114166942831658,
"delta_retention_norm_ratio": 0.8481571391375546,
"delta_cosine": 0.6619191498041295,
"relative_weight_rmse": 0.01075427974342996
},
{
"index": 236,
"total": 252,
"projection": "layers.7.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1302,
"parameters": 16777216,
"reference_delta_l2": 1.3658873172509274,
"retained_delta_l2": 1.106406551160283,
"delta_retention_norm_ratio": 0.8100276920259483,
"delta_cosine": 0.6252429909572943,
"relative_weight_rmse": 0.010549164315745227
},
{
"index": 237,
"total": 252,
"projection": "layers.7.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1301,
"parameters": 16777216,
"reference_delta_l2": 1.6847457418984,
"retained_delta_l2": 1.5226516861692236,
"delta_retention_norm_ratio": 0.9037872293141835,
"delta_cosine": 0.7177071108487745,
"relative_weight_rmse": 0.012113206962301877
},
{
"index": 238,
"total": 252,
"projection": "layers.7.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 0.9868820964566372,
"retained_delta_l2": 0.8897068211702872,
"delta_retention_norm_ratio": 0.9015330446917073,
"delta_cosine": 0.7170913768002884,
"relative_weight_rmse": 0.01196194767053509
},
{
"index": 239,
"total": 252,
"projection": "layers.8.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4308,
"parameters": 50331648,
"reference_delta_l2": 2.6355924662208907,
"retained_delta_l2": 2.2731708175513594,
"delta_retention_norm_ratio": 0.8624894958858346,
"delta_cosine": 0.6768140251766139,
"relative_weight_rmse": 0.011039891760100158
},
{
"index": 240,
"total": 252,
"projection": "layers.8.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.389,
"parameters": 50331648,
"reference_delta_l2": 2.202283141318376,
"retained_delta_l2": 1.659652546325219,
"delta_retention_norm_ratio": 0.7536054357350632,
"delta_cosine": 0.5770296667376316,
"relative_weight_rmse": 0.009465511365822644
},
{
"index": 241,
"total": 252,
"projection": "layers.8.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3955,
"parameters": 50331648,
"reference_delta_l2": 2.6806339499510345,
"retained_delta_l2": 2.2858577300024723,
"delta_retention_norm_ratio": 0.8527302767482395,
"delta_cosine": 0.666012255909814,
"relative_weight_rmse": 0.011428825233552205
},
{
"index": 242,
"total": 252,
"projection": "layers.8.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0348,
"parameters": 4194304,
"reference_delta_l2": 0.892108098567984,
"retained_delta_l2": 0.8088306720344752,
"delta_retention_norm_ratio": 0.9066509690168869,
"delta_cosine": 0.7194814890534805,
"relative_weight_rmse": 0.0121186655265901
},
{
"index": 243,
"total": 252,
"projection": "layers.8.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1428,
"parameters": 16777216,
"reference_delta_l2": 1.5711709173138488,
"retained_delta_l2": 1.3450190219279747,
"delta_retention_norm_ratio": 0.8560615570885729,
"delta_cosine": 0.6701946040661244,
"relative_weight_rmse": 0.01130083083234018
},
{
"index": 244,
"total": 252,
"projection": "layers.8.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.147,
"parameters": 16777216,
"reference_delta_l2": 1.5935161523801298,
"retained_delta_l2": 1.382567306616549,
"delta_retention_norm_ratio": 0.8676205161469493,
"delta_cosine": 0.6790418645088696,
"relative_weight_rmse": 0.011833231993008326
},
{
"index": 245,
"total": 252,
"projection": "layers.8.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.035,
"parameters": 4194304,
"reference_delta_l2": 0.7958179709276099,
"retained_delta_l2": 0.6599420021002363,
"delta_retention_norm_ratio": 0.8292625024928807,
"delta_cosine": 0.6433285511923893,
"relative_weight_rmse": 0.01062109515860514
},
{
"index": 246,
"total": 252,
"projection": "layers.9.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4556,
"parameters": 50331648,
"reference_delta_l2": 2.4632649867432574,
"retained_delta_l2": 2.0916648368795094,
"delta_retention_norm_ratio": 0.8491432501725892,
"delta_cosine": 0.6636211036150158,
"relative_weight_rmse": 0.010863964650101629
},
{
"index": 247,
"total": 252,
"projection": "layers.9.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4064,
"parameters": 50331648,
"reference_delta_l2": 2.6912322423445403,
"retained_delta_l2": 2.2009723269453074,
"delta_retention_norm_ratio": 0.8178306919464781,
"delta_cosine": 0.634134962968425,
"relative_weight_rmse": 0.010350243159792981
},
{
"index": 248,
"total": 252,
"projection": "layers.9.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4039,
"parameters": 50331648,
"reference_delta_l2": 3.019301808316451,
"retained_delta_l2": 2.736703036249453,
"delta_retention_norm_ratio": 0.9064026089446905,
"delta_cosine": 0.7213458216286287,
"relative_weight_rmse": 0.01237028304932083
},
{
"index": 249,
"total": 252,
"projection": "layers.9.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0322,
"parameters": 4194304,
"reference_delta_l2": 0.8372366151629234,
"retained_delta_l2": 0.7554391472146549,
"delta_retention_norm_ratio": 0.9023006561504109,
"delta_cosine": 0.7181386352936268,
"relative_weight_rmse": 0.01177065462703115
},
{
"index": 250,
"total": 252,
"projection": "layers.9.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1401,
"parameters": 16777216,
"reference_delta_l2": 1.5248485846368605,
"retained_delta_l2": 1.2996361778554804,
"delta_retention_norm_ratio": 0.8523050688111345,
"delta_cosine": 0.6667510008390406,
"relative_weight_rmse": 0.011153659988921252
},
{
"index": 251,
"total": 252,
"projection": "layers.9.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1403,
"parameters": 16777216,
"reference_delta_l2": 1.803779278376794,
"retained_delta_l2": 1.6567698226437095,
"delta_retention_norm_ratio": 0.918499199156242,
"delta_cosine": 0.7350062150712456,
"relative_weight_rmse": 0.0126925054191312
},
{
"index": 252,
"total": 252,
"projection": "layers.9.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0309,
"parameters": 4194304,
"reference_delta_l2": 0.7978971622820221,
"retained_delta_l2": 0.6671771373215317,
"delta_retention_norm_ratio": 0.8361693321647802,
"delta_cosine": 0.6525337153106847,
"relative_weight_rmse": 0.010395554903376797
}
]
}