mrjackspade's picture
Publish step 510 merge manifest
d1e1f4a 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_00000510/adapters.safetensors",
"adapter_sha256": "85359811e619276f85742072c2b3327915bbb56bcefb874f2f3101a5d1216eef",
"adapter_global_step": "510",
"output": "/workspace/Ideogram/artifacts/huggingface_release/s020_v1/qwen3vl_8b_ideogram4_nl_s020_v1_step_00000510_fp8_scaled.safetensors",
"output_bytes": 10588637800,
"output_sha256": "f0d1009af7407b60bda169967bb5e4c3e8ea264042b70f3d7b667bb0c2404ece",
"output_tensors": 1254,
"merged_projections": 252,
"lora_rank": 64,
"lora_alpha": 64.0,
"elapsed_seconds": 64.338,
"projections": [
{
"index": 1,
"total": 252,
"projection": "layers.0.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4406,
"parameters": 50331648,
"reference_delta_l2": 2.6327089901753045,
"retained_delta_l2": 2.2871515016030406,
"delta_retention_norm_ratio": 0.8687445175817727,
"delta_cosine": 0.6831931704979609,
"relative_weight_rmse": 0.01096298133808025
},
{
"index": 2,
"total": 252,
"projection": "layers.0.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3908,
"parameters": 50331648,
"reference_delta_l2": 2.6772801535866004,
"retained_delta_l2": 2.2880149418926226,
"delta_retention_norm_ratio": 0.8546042291567801,
"delta_cosine": 0.6670250460348754,
"relative_weight_rmse": 0.011768324045049744
},
{
"index": 3,
"total": 252,
"projection": "layers.0.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3978,
"parameters": 50331648,
"reference_delta_l2": 2.390616730980462,
"retained_delta_l2": 1.9866289619366921,
"delta_retention_norm_ratio": 0.8310110676427489,
"delta_cosine": 0.6447662799695758,
"relative_weight_rmse": 0.011412052721242289
},
{
"index": 4,
"total": 252,
"projection": "layers.0.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0309,
"parameters": 4194304,
"reference_delta_l2": 0.5230720972912708,
"retained_delta_l2": 0.3567391071841315,
"delta_retention_norm_ratio": 0.6820075263649222,
"delta_cosine": 0.5185684276018467,
"relative_weight_rmse": 0.008049897840969543
},
{
"index": 5,
"total": 252,
"projection": "layers.0.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1345,
"parameters": 16777216,
"reference_delta_l2": 1.8613124987703769,
"retained_delta_l2": 1.7369299527282984,
"delta_retention_norm_ratio": 0.9331748182402213,
"delta_cosine": 0.7526839574109756,
"relative_weight_rmse": 0.013067631665388612
},
{
"index": 6,
"total": 252,
"projection": "layers.0.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1321,
"parameters": 16777216,
"reference_delta_l2": 1.0563514233516575,
"retained_delta_l2": 0.7668640299186094,
"delta_retention_norm_ratio": 0.725955409313934,
"delta_cosine": 0.5543170897702486,
"relative_weight_rmse": 0.008753209330524843
},
{
"index": 7,
"total": 252,
"projection": "layers.0.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0301,
"parameters": 4194304,
"reference_delta_l2": 0.7022345644480543,
"retained_delta_l2": 0.5874381003072418,
"delta_retention_norm_ratio": 0.8365268957801288,
"delta_cosine": 0.6512962468872779,
"relative_weight_rmse": 0.010973664045303274
},
{
"index": 8,
"total": 252,
"projection": "layers.1.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4453,
"parameters": 50331648,
"reference_delta_l2": 1.8898138296216929,
"retained_delta_l2": 1.737735428436025,
"delta_retention_norm_ratio": 0.9195273106790041,
"delta_cosine": 0.7448556189453459,
"relative_weight_rmse": 0.011322768425377668
},
{
"index": 9,
"total": 252,
"projection": "layers.1.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3905,
"parameters": 50331648,
"reference_delta_l2": 3.657081033192418,
"retained_delta_l2": 3.6015521990413126,
"delta_retention_norm_ratio": 0.984816077727807,
"delta_cosine": 0.8418804183137666,
"relative_weight_rmse": 0.013840827955486369
},
{
"index": 10,
"total": 252,
"projection": "layers.1.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3926,
"parameters": 50331648,
"reference_delta_l2": 3.3758522814584864,
"retained_delta_l2": 3.3480227210234577,
"delta_retention_norm_ratio": 0.9917562860828124,
"delta_cosine": 0.9015911746554313,
"relative_weight_rmse": 0.01481239001519705
},
{
"index": 11,
"total": 252,
"projection": "layers.1.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0311,
"parameters": 4194304,
"reference_delta_l2": 0.6051762921808931,
"retained_delta_l2": 0.4756470663655935,
"delta_retention_norm_ratio": 0.7859644743376992,
"delta_cosine": 0.6060727303822177,
"relative_weight_rmse": 0.009964429077215514
},
{
"index": 12,
"total": 252,
"projection": "layers.1.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.133,
"parameters": 16777216,
"reference_delta_l2": 1.4935938757174607,
"retained_delta_l2": 1.2785725581027818,
"delta_retention_norm_ratio": 0.8560376276908664,
"delta_cosine": 0.6677743891393898,
"relative_weight_rmse": 0.011547649685206055
},
{
"index": 13,
"total": 252,
"projection": "layers.1.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1319,
"parameters": 16777216,
"reference_delta_l2": 1.8196863010939506,
"retained_delta_l2": 1.7101196678521942,
"delta_retention_norm_ratio": 0.9397881749310925,
"delta_cosine": 0.7595529659149882,
"relative_weight_rmse": 0.013550407565098701
},
{
"index": 14,
"total": 252,
"projection": "layers.1.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0301,
"parameters": 4194304,
"reference_delta_l2": 0.7615395905357294,
"retained_delta_l2": 0.6517298088924471,
"delta_retention_norm_ratio": 0.8558055510074886,
"delta_cosine": 0.6681908747332516,
"relative_weight_rmse": 0.011085857994289853
},
{
"index": 15,
"total": 252,
"projection": "layers.10.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4293,
"parameters": 50331648,
"reference_delta_l2": 2.531215055123725,
"retained_delta_l2": 2.1824833997654514,
"delta_retention_norm_ratio": 0.8622275674868615,
"delta_cosine": 0.6764698570780456,
"relative_weight_rmse": 0.01081504682034425
},
{
"index": 16,
"total": 252,
"projection": "layers.10.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3942,
"parameters": 50331648,
"reference_delta_l2": 2.1752664087061495,
"retained_delta_l2": 1.6278962284509517,
"delta_retention_norm_ratio": 0.7483663710962309,
"delta_cosine": 0.5733297280350799,
"relative_weight_rmse": 0.008945254391297452
},
{
"index": 17,
"total": 252,
"projection": "layers.10.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4015,
"parameters": 50331648,
"reference_delta_l2": 2.6665348359398693,
"retained_delta_l2": 2.314993607953368,
"delta_retention_norm_ratio": 0.868165522066921,
"delta_cosine": 0.6808506874174365,
"relative_weight_rmse": 0.011346629690113269
},
{
"index": 18,
"total": 252,
"projection": "layers.10.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0318,
"parameters": 4194304,
"reference_delta_l2": 0.8598177590242261,
"retained_delta_l2": 0.7820980451680212,
"delta_retention_norm_ratio": 0.9096090851340335,
"delta_cosine": 0.7270726588216819,
"relative_weight_rmse": 0.011634985351763842
},
{
"index": 19,
"total": 252,
"projection": "layers.10.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.135,
"parameters": 16777216,
"reference_delta_l2": 1.6733573890753237,
"retained_delta_l2": 1.4703711717179522,
"delta_retention_norm_ratio": 0.8786952394732968,
"delta_cosine": 0.6914507221186205,
"relative_weight_rmse": 0.011706738980892818
},
{
"index": 20,
"total": 252,
"projection": "layers.10.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1341,
"parameters": 16777216,
"reference_delta_l2": 2.027281987832174,
"retained_delta_l2": 1.9186377269978645,
"delta_retention_norm_ratio": 0.9464089053785331,
"delta_cosine": 0.7704910918797055,
"relative_weight_rmse": 0.013102069197817139
},
{
"index": 21,
"total": 252,
"projection": "layers.10.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0308,
"parameters": 4194304,
"reference_delta_l2": 0.8179661631941779,
"retained_delta_l2": 0.7086405890578836,
"delta_retention_norm_ratio": 0.8663446251744996,
"delta_cosine": 0.6851426874031,
"relative_weight_rmse": 0.010413497936023764
},
{
"index": 22,
"total": 252,
"projection": "layers.11.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4364,
"parameters": 50331648,
"reference_delta_l2": 2.34500506277976,
"retained_delta_l2": 1.9435492747038492,
"delta_retention_norm_ratio": 0.8288038714935538,
"delta_cosine": 0.6447956411947068,
"relative_weight_rmse": 0.010210191435923827
},
{
"index": 23,
"total": 252,
"projection": "layers.11.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3924,
"parameters": 50331648,
"reference_delta_l2": 2.5517632567783286,
"retained_delta_l2": 2.1012649758237125,
"delta_retention_norm_ratio": 0.8234560828642926,
"delta_cosine": 0.6404681709972035,
"relative_weight_rmse": 0.010146022133471601
},
{
"index": 24,
"total": 252,
"projection": "layers.11.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3959,
"parameters": 50331648,
"reference_delta_l2": 2.4056315041822263,
"retained_delta_l2": 1.9713382785726346,
"delta_retention_norm_ratio": 0.8194681002245912,
"delta_cosine": 0.6359068240802415,
"relative_weight_rmse": 0.010523245330967618
},
{
"index": 25,
"total": 252,
"projection": "layers.11.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0402,
"parameters": 4194304,
"reference_delta_l2": 0.8617108467783873,
"retained_delta_l2": 0.7790011437447393,
"delta_retention_norm_ratio": 0.9040168713868829,
"delta_cosine": 0.7186376496198529,
"relative_weight_rmse": 0.011705958411568636
},
{
"index": 26,
"total": 252,
"projection": "layers.11.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1368,
"parameters": 16777216,
"reference_delta_l2": 1.5926304671746783,
"retained_delta_l2": 1.3857014199563344,
"delta_retention_norm_ratio": 0.8700708974973741,
"delta_cosine": 0.6847607651679083,
"relative_weight_rmse": 0.011298482092380215
},
{
"index": 27,
"total": 252,
"projection": "layers.11.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1344,
"parameters": 16777216,
"reference_delta_l2": 1.2443009836930687,
"retained_delta_l2": 0.9880804106015009,
"delta_retention_norm_ratio": 0.7940847299412167,
"delta_cosine": 0.6130805571685117,
"relative_weight_rmse": 0.00981917000253394
},
{
"index": 28,
"total": 252,
"projection": "layers.11.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 0.897717225216816,
"retained_delta_l2": 0.7899933152657633,
"delta_retention_norm_ratio": 0.8800024028445759,
"delta_cosine": 0.6969767151975313,
"relative_weight_rmse": 0.011141484998529136
},
{
"index": 29,
"total": 252,
"projection": "layers.12.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4427,
"parameters": 50331648,
"reference_delta_l2": 2.320277198795607,
"retained_delta_l2": 1.9160150920903574,
"delta_retention_norm_ratio": 0.8257699093388106,
"delta_cosine": 0.6425239669217114,
"relative_weight_rmse": 0.010044208576518672
},
{
"index": 30,
"total": 252,
"projection": "layers.12.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3944,
"parameters": 50331648,
"reference_delta_l2": 2.901090413343313,
"retained_delta_l2": 2.5319461880405907,
"delta_retention_norm_ratio": 0.8727567318809247,
"delta_cosine": 0.6884221568080939,
"relative_weight_rmse": 0.011163459806453873
},
{
"index": 31,
"total": 252,
"projection": "layers.12.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4159,
"parameters": 50331648,
"reference_delta_l2": 2.467384856962169,
"retained_delta_l2": 2.0462148511281737,
"delta_retention_norm_ratio": 0.8293051022642096,
"delta_cosine": 0.6449839710058615,
"relative_weight_rmse": 0.010537524426169257
},
{
"index": 32,
"total": 252,
"projection": "layers.12.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0322,
"parameters": 4194304,
"reference_delta_l2": 0.8282236259084472,
"retained_delta_l2": 0.7464401813193698,
"delta_retention_norm_ratio": 0.9012543931002062,
"delta_cosine": 0.7199601632722067,
"relative_weight_rmse": 0.01138058043089329
},
{
"index": 33,
"total": 252,
"projection": "layers.12.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1348,
"parameters": 16777216,
"reference_delta_l2": 1.6453911434628496,
"retained_delta_l2": 1.463020090119954,
"delta_retention_norm_ratio": 0.8891624924155833,
"delta_cosine": 0.703984320068952,
"relative_weight_rmse": 0.011441977611469131
},
{
"index": 34,
"total": 252,
"projection": "layers.12.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1347,
"parameters": 16777216,
"reference_delta_l2": 1.6210399311215336,
"retained_delta_l2": 1.4405291201316213,
"delta_retention_norm_ratio": 0.8886450558531127,
"delta_cosine": 0.7040551826089303,
"relative_weight_rmse": 0.011592997050092375
},
{
"index": 35,
"total": 252,
"projection": "layers.12.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0303,
"parameters": 4194304,
"reference_delta_l2": 1.0237886771272462,
"retained_delta_l2": 0.9535994191334317,
"delta_retention_norm_ratio": 0.9314416543551098,
"delta_cosine": 0.75447831266875,
"relative_weight_rmse": 0.011710835840793916
},
{
"index": 36,
"total": 252,
"projection": "layers.13.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.435,
"parameters": 50331648,
"reference_delta_l2": 2.0846221826380544,
"retained_delta_l2": 1.6348804269860413,
"delta_retention_norm_ratio": 0.7842574259269981,
"delta_cosine": 0.6048871053089986,
"relative_weight_rmse": 0.009322682612643638
},
{
"index": 37,
"total": 252,
"projection": "layers.13.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3907,
"parameters": 50331648,
"reference_delta_l2": 2.5562534815687137,
"retained_delta_l2": 2.1167548474306748,
"delta_retention_norm_ratio": 0.8280692281469956,
"delta_cosine": 0.6432194070144007,
"relative_weight_rmse": 0.01067726899334383
},
{
"index": 38,
"total": 252,
"projection": "layers.13.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.393,
"parameters": 50331648,
"reference_delta_l2": 2.203330089333929,
"retained_delta_l2": 1.726298107952274,
"delta_retention_norm_ratio": 0.7834949998228078,
"delta_cosine": 0.6030605391711824,
"relative_weight_rmse": 0.009684847344538314
},
{
"index": 39,
"total": 252,
"projection": "layers.13.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0311,
"parameters": 4194304,
"reference_delta_l2": 1.2882447085921234,
"retained_delta_l2": 1.2714993722220769,
"delta_retention_norm_ratio": 0.9870014320583961,
"delta_cosine": 0.8276057019002796,
"relative_weight_rmse": 0.014468699445840155
},
{
"index": 40,
"total": 252,
"projection": "layers.13.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1357,
"parameters": 16777216,
"reference_delta_l2": 1.6166214467944622,
"retained_delta_l2": 1.42378514680623,
"delta_retention_norm_ratio": 0.8807164779543163,
"delta_cosine": 0.6955817661003703,
"relative_weight_rmse": 0.011709199355451921
},
{
"index": 41,
"total": 252,
"projection": "layers.13.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.135,
"parameters": 16777216,
"reference_delta_l2": 1.2623715176154386,
"retained_delta_l2": 1.0099243645887293,
"delta_retention_norm_ratio": 0.8000215075324495,
"delta_cosine": 0.6184279126413518,
"relative_weight_rmse": 0.00989565906685087
},
{
"index": 42,
"total": 252,
"projection": "layers.13.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0307,
"parameters": 4194304,
"reference_delta_l2": 0.8825092886380478,
"retained_delta_l2": 0.7817498199472648,
"delta_retention_norm_ratio": 0.8858261663780532,
"delta_cosine": 0.7006507079130241,
"relative_weight_rmse": 0.011154325923963036
},
{
"index": 43,
"total": 252,
"projection": "layers.14.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4362,
"parameters": 50331648,
"reference_delta_l2": 2.13420138255605,
"retained_delta_l2": 1.7189831026670972,
"delta_retention_norm_ratio": 0.8054455951145238,
"delta_cosine": 0.6249848772544617,
"relative_weight_rmse": 0.009436160658905499
},
{
"index": 44,
"total": 252,
"projection": "layers.14.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3913,
"parameters": 50331648,
"reference_delta_l2": 2.4254956227466296,
"retained_delta_l2": 1.9749517522158966,
"delta_retention_norm_ratio": 0.8142466775427376,
"delta_cosine": 0.6301317568289052,
"relative_weight_rmse": 0.010409978211722815
},
{
"index": 45,
"total": 252,
"projection": "layers.14.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3944,
"parameters": 50331648,
"reference_delta_l2": 2.651475706897975,
"retained_delta_l2": 2.2754130914762447,
"delta_retention_norm_ratio": 0.8581685608344891,
"delta_cosine": 0.6730766377166092,
"relative_weight_rmse": 0.0109756270465385
},
{
"index": 46,
"total": 252,
"projection": "layers.14.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 0.6661733648606436,
"retained_delta_l2": 0.5479984811762695,
"delta_retention_norm_ratio": 0.8226064116071422,
"delta_cosine": 0.6406453360097045,
"relative_weight_rmse": 0.010066646015560256
},
{
"index": 47,
"total": 252,
"projection": "layers.14.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1353,
"parameters": 16777216,
"reference_delta_l2": 1.4582571531833697,
"retained_delta_l2": 1.2460505381967344,
"delta_retention_norm_ratio": 0.854479290896404,
"delta_cosine": 0.6694921418838832,
"relative_weight_rmse": 0.010918642035532512
},
{
"index": 48,
"total": 252,
"projection": "layers.14.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1363,
"parameters": 16777216,
"reference_delta_l2": 2.1856064529455286,
"retained_delta_l2": 2.10889967923761,
"delta_retention_norm_ratio": 0.9649036661634387,
"delta_cosine": 0.7919398570664227,
"relative_weight_rmse": 0.01350813377663138
},
{
"index": 49,
"total": 252,
"projection": "layers.14.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 0.9013074414020275,
"retained_delta_l2": 0.8098390973556806,
"delta_retention_norm_ratio": 0.8985159338038269,
"delta_cosine": 0.7149864158653797,
"relative_weight_rmse": 0.011479685591920221
},
{
"index": 50,
"total": 252,
"projection": "layers.15.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4343,
"parameters": 50331648,
"reference_delta_l2": 2.086789207992008,
"retained_delta_l2": 1.659957545345398,
"delta_retention_norm_ratio": 0.7954600967783783,
"delta_cosine": 0.6150906974827606,
"relative_weight_rmse": 0.009358808293002549
},
{
"index": 51,
"total": 252,
"projection": "layers.15.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3912,
"parameters": 50331648,
"reference_delta_l2": 2.4256639917796314,
"retained_delta_l2": 2.004613189385262,
"delta_retention_norm_ratio": 0.8264183317140072,
"delta_cosine": 0.6413103507947648,
"relative_weight_rmse": 0.010663762745827537
},
{
"index": 52,
"total": 252,
"projection": "layers.15.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3935,
"parameters": 50331648,
"reference_delta_l2": 2.595704581856843,
"retained_delta_l2": 2.2266853540778584,
"delta_retention_norm_ratio": 0.8578346587056507,
"delta_cosine": 0.6725383490644409,
"relative_weight_rmse": 0.010848416048602223
},
{
"index": 53,
"total": 252,
"projection": "layers.15.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 0.7174737517442046,
"retained_delta_l2": 0.6089911363066031,
"delta_retention_norm_ratio": 0.8487991857905933,
"delta_cosine": 0.662850053918743,
"relative_weight_rmse": 0.010668703613887207
},
{
"index": 54,
"total": 252,
"projection": "layers.15.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1364,
"parameters": 16777216,
"reference_delta_l2": 2.0240964109023434,
"retained_delta_l2": 1.9264964091557202,
"delta_retention_norm_ratio": 0.9517809521221802,
"delta_cosine": 0.7758733106480629,
"relative_weight_rmse": 0.013361216275558942
},
{
"index": 55,
"total": 252,
"projection": "layers.15.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1344,
"parameters": 16777216,
"reference_delta_l2": 2.1135189974599835,
"retained_delta_l2": 2.02665696980818,
"delta_retention_norm_ratio": 0.9589017048078612,
"delta_cosine": 0.791601174544528,
"relative_weight_rmse": 0.012965520219919587
},
{
"index": 56,
"total": 252,
"projection": "layers.15.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0307,
"parameters": 4194304,
"reference_delta_l2": 0.7616490923373335,
"retained_delta_l2": 0.6541228230333136,
"delta_retention_norm_ratio": 0.8588243977629574,
"delta_cosine": 0.6752154346427254,
"relative_weight_rmse": 0.010342522352985545
},
{
"index": 57,
"total": 252,
"projection": "layers.16.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4408,
"parameters": 50331648,
"reference_delta_l2": 2.2303157381509173,
"retained_delta_l2": 1.8342983474627628,
"delta_retention_norm_ratio": 0.8224388664286251,
"delta_cosine": 0.639472222010357,
"relative_weight_rmse": 0.009852833359225589
},
{
"index": 58,
"total": 252,
"projection": "layers.16.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3928,
"parameters": 50331648,
"reference_delta_l2": 2.4166462281703205,
"retained_delta_l2": 2.0057662695293548,
"delta_retention_norm_ratio": 0.8299792688514244,
"delta_cosine": 0.6450837084917048,
"relative_weight_rmse": 0.010391807503082114
},
{
"index": 59,
"total": 252,
"projection": "layers.16.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3934,
"parameters": 50331648,
"reference_delta_l2": 2.3920914943172247,
"retained_delta_l2": 1.9746787002661033,
"delta_retention_norm_ratio": 0.8255029980906882,
"delta_cosine": 0.640919380175158,
"relative_weight_rmse": 0.010358761010515978
},
{
"index": 60,
"total": 252,
"projection": "layers.16.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 0.6590984356455544,
"retained_delta_l2": 0.5477319763947867,
"delta_retention_norm_ratio": 0.8310321293029778,
"delta_cosine": 0.6469395956324974,
"relative_weight_rmse": 0.010300532914943077
},
{
"index": 61,
"total": 252,
"projection": "layers.16.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.136,
"parameters": 16777216,
"reference_delta_l2": 1.5777249464740992,
"retained_delta_l2": 1.389874264125338,
"delta_retention_norm_ratio": 0.8809357215473014,
"delta_cosine": 0.6947001732960237,
"relative_weight_rmse": 0.011524895889996816
},
{
"index": 62,
"total": 252,
"projection": "layers.16.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1348,
"parameters": 16777216,
"reference_delta_l2": 1.7397412979225706,
"retained_delta_l2": 1.599333859427975,
"delta_retention_norm_ratio": 0.9192940705251657,
"delta_cosine": 0.7386192162456077,
"relative_weight_rmse": 0.01216112369416335
},
{
"index": 63,
"total": 252,
"projection": "layers.16.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0301,
"parameters": 4194304,
"reference_delta_l2": 0.8530860626636955,
"retained_delta_l2": 0.7554267831913181,
"delta_retention_norm_ratio": 0.8855223596462896,
"delta_cosine": 0.7018652535197768,
"relative_weight_rmse": 0.01124175948854005
},
{
"index": 64,
"total": 252,
"projection": "layers.17.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4338,
"parameters": 50331648,
"reference_delta_l2": 2.317107496556227,
"retained_delta_l2": 1.924559504330395,
"delta_retention_norm_ratio": 0.830587060458243,
"delta_cosine": 0.6468727141506057,
"relative_weight_rmse": 0.010114956868444772
},
{
"index": 65,
"total": 252,
"projection": "layers.17.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3988,
"parameters": 50331648,
"reference_delta_l2": 3.086873344951891,
"retained_delta_l2": 2.7992915749567815,
"delta_retention_norm_ratio": 0.9068371980776584,
"delta_cosine": 0.722672586364596,
"relative_weight_rmse": 0.012225240939740489
},
{
"index": 66,
"total": 252,
"projection": "layers.17.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3927,
"parameters": 50331648,
"reference_delta_l2": 2.9036635508281687,
"retained_delta_l2": 2.579993038764108,
"delta_retention_norm_ratio": 0.8885302975367978,
"delta_cosine": 0.7026662603657728,
"relative_weight_rmse": 0.01164404960770755
},
{
"index": 67,
"total": 252,
"projection": "layers.17.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 0.7066499516158611,
"retained_delta_l2": 0.6160339511399299,
"delta_retention_norm_ratio": 0.8717667774989241,
"delta_cosine": 0.6898840952927281,
"relative_weight_rmse": 0.010721400388771996
},
{
"index": 68,
"total": 252,
"projection": "layers.17.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.135,
"parameters": 16777216,
"reference_delta_l2": 1.6344810946602308,
"retained_delta_l2": 1.4619995818789637,
"delta_retention_norm_ratio": 0.89447322863216,
"delta_cosine": 0.7109335062500767,
"relative_weight_rmse": 0.011753573182973773
},
{
"index": 69,
"total": 252,
"projection": "layers.17.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1342,
"parameters": 16777216,
"reference_delta_l2": 1.4864809414200086,
"retained_delta_l2": 1.2848595929969922,
"delta_retention_norm_ratio": 0.864363314183893,
"delta_cosine": 0.6769874056184836,
"relative_weight_rmse": 0.01113307097335904
},
{
"index": 70,
"total": 252,
"projection": "layers.17.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0302,
"parameters": 4194304,
"reference_delta_l2": 0.9355528627118188,
"retained_delta_l2": 0.8580613997950565,
"delta_retention_norm_ratio": 0.9171704069269334,
"delta_cosine": 0.7352963378442325,
"relative_weight_rmse": 0.012006218333944534
},
{
"index": 71,
"total": 252,
"projection": "layers.18.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4359,
"parameters": 50331648,
"reference_delta_l2": 2.623621093517205,
"retained_delta_l2": 2.2971542549158395,
"delta_retention_norm_ratio": 0.8755663158037402,
"delta_cosine": 0.6895082587554824,
"relative_weight_rmse": 0.01092384570817268
},
{
"index": 72,
"total": 252,
"projection": "layers.18.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3935,
"parameters": 50331648,
"reference_delta_l2": 3.9516853977153295,
"retained_delta_l2": 3.814789749658205,
"delta_retention_norm_ratio": 0.9653576552080105,
"delta_cosine": 0.7899677144921917,
"relative_weight_rmse": 0.014049752567654311
},
{
"index": 73,
"total": 252,
"projection": "layers.18.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3941,
"parameters": 50331648,
"reference_delta_l2": 2.975283543456475,
"retained_delta_l2": 2.672285012665405,
"delta_retention_norm_ratio": 0.898161460457289,
"delta_cosine": 0.7131976175914161,
"relative_weight_rmse": 0.01176517984472072
},
{
"index": 74,
"total": 252,
"projection": "layers.18.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 1.1137359991025886,
"retained_delta_l2": 1.0817885328986219,
"delta_retention_norm_ratio": 0.9713150457292312,
"delta_cosine": 0.8014797022907222,
"relative_weight_rmse": 0.013870573203487047
},
{
"index": 75,
"total": 252,
"projection": "layers.18.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1351,
"parameters": 16777216,
"reference_delta_l2": 1.5162794257656762,
"retained_delta_l2": 1.328227386411706,
"delta_retention_norm_ratio": 0.8759779786242172,
"delta_cosine": 0.6919780168741544,
"relative_weight_rmse": 0.01116224109623716
},
{
"index": 76,
"total": 252,
"projection": "layers.18.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1358,
"parameters": 16777216,
"reference_delta_l2": 1.8813425857431623,
"retained_delta_l2": 1.761399415184713,
"delta_retention_norm_ratio": 0.9362459705811265,
"delta_cosine": 0.7578788430641706,
"relative_weight_rmse": 0.012728323169903293
},
{
"index": 77,
"total": 252,
"projection": "layers.18.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 0.8633863260109812,
"retained_delta_l2": 0.7703261769314331,
"delta_retention_norm_ratio": 0.8922149375361262,
"delta_cosine": 0.7084065617538431,
"relative_weight_rmse": 0.011269247563962978
},
{
"index": 78,
"total": 252,
"projection": "layers.19.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4543,
"parameters": 50331648,
"reference_delta_l2": 2.5417637873727275,
"retained_delta_l2": 2.194293303255207,
"delta_retention_norm_ratio": 0.8632955250036511,
"delta_cosine": 0.6779440790016057,
"relative_weight_rmse": 0.010722249439618881
},
{
"index": 79,
"total": 252,
"projection": "layers.19.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3932,
"parameters": 50331648,
"reference_delta_l2": 3.6374279097339968,
"retained_delta_l2": 3.4393685679018664,
"delta_retention_norm_ratio": 0.9455496172715586,
"delta_cosine": 0.7647211066286134,
"relative_weight_rmse": 0.013621033733582259
},
{
"index": 80,
"total": 252,
"projection": "layers.19.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3948,
"parameters": 50331648,
"reference_delta_l2": 2.762804194662718,
"retained_delta_l2": 2.4001113822471773,
"delta_retention_norm_ratio": 0.8687229398608113,
"delta_cosine": 0.683274951546065,
"relative_weight_rmse": 0.011285681848780734
},
{
"index": 81,
"total": 252,
"projection": "layers.19.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0315,
"parameters": 4194304,
"reference_delta_l2": 0.911919532599503,
"retained_delta_l2": 0.8576706193518299,
"delta_retention_norm_ratio": 0.9405112936960216,
"delta_cosine": 0.7652673572184514,
"relative_weight_rmse": 0.012584524885586413
},
{
"index": 82,
"total": 252,
"projection": "layers.19.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1362,
"parameters": 16777216,
"reference_delta_l2": 1.5551248854139976,
"retained_delta_l2": 1.3691271157560179,
"delta_retention_norm_ratio": 0.8803968919779299,
"delta_cosine": 0.695319870840718,
"relative_weight_rmse": 0.011327888696033118
},
{
"index": 83,
"total": 252,
"projection": "layers.19.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1349,
"parameters": 16777216,
"reference_delta_l2": 1.755104352838205,
"retained_delta_l2": 1.6199177558526192,
"delta_retention_norm_ratio": 0.9229751799276359,
"delta_cosine": 0.7423155567664863,
"relative_weight_rmse": 0.01208754545011881
},
{
"index": 84,
"total": 252,
"projection": "layers.19.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0301,
"parameters": 4194304,
"reference_delta_l2": 1.0831238658546205,
"retained_delta_l2": 1.0280764178780222,
"delta_retention_norm_ratio": 0.9491771442658,
"delta_cosine": 0.7753840719743615,
"relative_weight_rmse": 0.012940700707024272
},
{
"index": 85,
"total": 252,
"projection": "layers.2.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4384,
"parameters": 50331648,
"reference_delta_l2": 1.4234882918656107,
"retained_delta_l2": 1.2061008715333281,
"delta_retention_norm_ratio": 0.8472854173971627,
"delta_cosine": 0.6863852764857693,
"relative_weight_rmse": 0.00940710966144368
},
{
"index": 86,
"total": 252,
"projection": "layers.2.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3962,
"parameters": 50331648,
"reference_delta_l2": 2.191820181261131,
"retained_delta_l2": 2.011655969514262,
"delta_retention_norm_ratio": 0.9178015544855482,
"delta_cosine": 0.7659305807977074,
"relative_weight_rmse": 0.011770580379080114
},
{
"index": 87,
"total": 252,
"projection": "layers.2.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3969,
"parameters": 50331648,
"reference_delta_l2": 2.761679938522529,
"retained_delta_l2": 2.697117560202633,
"delta_retention_norm_ratio": 0.9766220634696589,
"delta_cosine": 0.8389325810813095,
"relative_weight_rmse": 0.014053611875972738
},
{
"index": 88,
"total": 252,
"projection": "layers.2.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0315,
"parameters": 4194304,
"reference_delta_l2": 0.530888361819072,
"retained_delta_l2": 0.38520761860965247,
"delta_retention_norm_ratio": 0.7255906256632767,
"delta_cosine": 0.554529133357542,
"relative_weight_rmse": 0.008682646027018927
},
{
"index": 89,
"total": 252,
"projection": "layers.2.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1366,
"parameters": 16777216,
"reference_delta_l2": 1.767860210356245,
"retained_delta_l2": 1.600839620127675,
"delta_retention_norm_ratio": 0.9055238704677258,
"delta_cosine": 0.7176664166780741,
"relative_weight_rmse": 0.01264118198200427
},
{
"index": 90,
"total": 252,
"projection": "layers.2.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.135,
"parameters": 16777216,
"reference_delta_l2": 1.150731244450129,
"retained_delta_l2": 0.8859188775360538,
"delta_retention_norm_ratio": 0.7698747051570547,
"delta_cosine": 0.5907894243743869,
"relative_weight_rmse": 0.009674117713943687
},
{
"index": 91,
"total": 252,
"projection": "layers.2.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 1.0904290529480396,
"retained_delta_l2": 1.0394254308026631,
"delta_retention_norm_ratio": 0.9532260975553751,
"delta_cosine": 0.7800620905527914,
"relative_weight_rmse": 0.013025487112775229
},
{
"index": 92,
"total": 252,
"projection": "layers.20.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4408,
"parameters": 50331648,
"reference_delta_l2": 2.4208288436366288,
"retained_delta_l2": 2.0116675785458877,
"delta_retention_norm_ratio": 0.8309829849531664,
"delta_cosine": 0.6451165140730315,
"relative_weight_rmse": 0.010519303702515534
},
{
"index": 93,
"total": 252,
"projection": "layers.20.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3966,
"parameters": 50331648,
"reference_delta_l2": 3.4874710954715287,
"retained_delta_l2": 3.2772327189895423,
"delta_retention_norm_ratio": 0.9397160949219107,
"delta_cosine": 0.7579808679328548,
"relative_weight_rmse": 0.013336498890557155
},
{
"index": 94,
"total": 252,
"projection": "layers.20.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3944,
"parameters": 50331648,
"reference_delta_l2": 2.727859600528451,
"retained_delta_l2": 2.3557936631685323,
"delta_retention_norm_ratio": 0.8636051733425574,
"delta_cosine": 0.6782957869277911,
"relative_weight_rmse": 0.011179866765553926
},
{
"index": 95,
"total": 252,
"projection": "layers.20.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0307,
"parameters": 4194304,
"reference_delta_l2": 0.7589017611994083,
"retained_delta_l2": 0.6787277545565447,
"delta_retention_norm_ratio": 0.8943552239012432,
"delta_cosine": 0.7128861863951973,
"relative_weight_rmse": 0.011415075810978939
},
{
"index": 96,
"total": 252,
"projection": "layers.20.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1305,
"parameters": 16777216,
"reference_delta_l2": 1.5289282697500297,
"retained_delta_l2": 1.33978659553641,
"delta_retention_norm_ratio": 0.8762913356003659,
"delta_cosine": 0.6917474952652858,
"relative_weight_rmse": 0.011402773641846947
},
{
"index": 97,
"total": 252,
"projection": "layers.20.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1296,
"parameters": 16777216,
"reference_delta_l2": 1.6938623059639735,
"retained_delta_l2": 1.5412844602292317,
"delta_retention_norm_ratio": 0.9099231116971401,
"delta_cosine": 0.7301091098716266,
"relative_weight_rmse": 0.011523558952480195
},
{
"index": 98,
"total": 252,
"projection": "layers.20.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0284,
"parameters": 4194304,
"reference_delta_l2": 0.9129813777821141,
"retained_delta_l2": 0.831946400145376,
"delta_retention_norm_ratio": 0.9112413685440173,
"delta_cosine": 0.7293886338239729,
"relative_weight_rmse": 0.011760719644457438
},
{
"index": 99,
"total": 252,
"projection": "layers.21.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4237,
"parameters": 50331648,
"reference_delta_l2": 2.831386357743151,
"retained_delta_l2": 2.4949310729952527,
"delta_retention_norm_ratio": 0.8811694194161193,
"delta_cosine": 0.6942913392598284,
"relative_weight_rmse": 0.011540670638753315
},
{
"index": 100,
"total": 252,
"projection": "layers.21.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3817,
"parameters": 50331648,
"reference_delta_l2": 3.581268905576441,
"retained_delta_l2": 3.3809657042356394,
"delta_retention_norm_ratio": 0.9440692093718773,
"delta_cosine": 0.762439197397351,
"relative_weight_rmse": 0.013782512921707265
},
{
"index": 101,
"total": 252,
"projection": "layers.21.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3867,
"parameters": 50331648,
"reference_delta_l2": 2.940294335586719,
"retained_delta_l2": 2.60905610019324,
"delta_retention_norm_ratio": 0.887345211877442,
"delta_cosine": 0.7013493809133563,
"relative_weight_rmse": 0.011735881806793533
},
{
"index": 102,
"total": 252,
"projection": "layers.21.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 0.77410035838613,
"retained_delta_l2": 0.6893916529822197,
"delta_retention_norm_ratio": 0.8905714168890017,
"delta_cosine": 0.7066130616326094,
"relative_weight_rmse": 0.01174789561808868
},
{
"index": 103,
"total": 252,
"projection": "layers.21.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1282,
"parameters": 16777216,
"reference_delta_l2": 1.6399927325322987,
"retained_delta_l2": 1.4793613166097839,
"delta_retention_norm_ratio": 0.9020535806433208,
"delta_cosine": 0.7189615443238899,
"relative_weight_rmse": 0.011922402971183752
},
{
"index": 104,
"total": 252,
"projection": "layers.21.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1295,
"parameters": 16777216,
"reference_delta_l2": 1.2205028435765317,
"retained_delta_l2": 0.9755466710777216,
"delta_retention_norm_ratio": 0.7992989743628973,
"delta_cosine": 0.6191341941737903,
"relative_weight_rmse": 0.009587733282748874
},
{
"index": 105,
"total": 252,
"projection": "layers.21.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0296,
"parameters": 4194304,
"reference_delta_l2": 1.1860379824198768,
"retained_delta_l2": 1.1488803099808462,
"delta_retention_norm_ratio": 0.9686707567634406,
"delta_cosine": 0.7978091745973052,
"relative_weight_rmse": 0.013661968001463982
},
{
"index": 106,
"total": 252,
"projection": "layers.22.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4243,
"parameters": 50331648,
"reference_delta_l2": 2.991059855940698,
"retained_delta_l2": 2.701743807454628,
"delta_retention_norm_ratio": 0.9032730662639719,
"delta_cosine": 0.7181042575501959,
"relative_weight_rmse": 0.011729870238563056
},
{
"index": 107,
"total": 252,
"projection": "layers.22.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3826,
"parameters": 50331648,
"reference_delta_l2": 3.777630486236153,
"retained_delta_l2": 3.6098447655631563,
"delta_retention_norm_ratio": 0.9555844010460192,
"delta_cosine": 0.779378122578379,
"relative_weight_rmse": 0.013768834214939765
},
{
"index": 108,
"total": 252,
"projection": "layers.22.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3824,
"parameters": 50331648,
"reference_delta_l2": 3.1620192849135553,
"retained_delta_l2": 2.876731912335432,
"delta_retention_norm_ratio": 0.9097768397747383,
"delta_cosine": 0.7257917626151396,
"relative_weight_rmse": 0.011995691529653848
},
{
"index": 109,
"total": 252,
"projection": "layers.22.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0306,
"parameters": 4194304,
"reference_delta_l2": 0.9980371973618855,
"retained_delta_l2": 0.9581176806419454,
"delta_retention_norm_ratio": 0.960001975051171,
"delta_cosine": 0.7933572727606913,
"relative_weight_rmse": 0.01350631395620473
},
{
"index": 110,
"total": 252,
"projection": "layers.22.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1301,
"parameters": 16777216,
"reference_delta_l2": 2.0063344121830644,
"retained_delta_l2": 1.893268654854589,
"delta_retention_norm_ratio": 0.9436456073115698,
"delta_cosine": 0.763043309624487,
"relative_weight_rmse": 0.013361870790913288
},
{
"index": 111,
"total": 252,
"projection": "layers.22.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1307,
"parameters": 16777216,
"reference_delta_l2": 1.8013330279960227,
"retained_delta_l2": 1.6650094943185272,
"delta_retention_norm_ratio": 0.9243207493790558,
"delta_cosine": 0.747278912792796,
"relative_weight_rmse": 0.012280580099934449
},
{
"index": 112,
"total": 252,
"projection": "layers.22.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0299,
"parameters": 4194304,
"reference_delta_l2": 1.0591503966182438,
"retained_delta_l2": 1.0033485870682053,
"delta_retention_norm_ratio": 0.9473145553943917,
"delta_cosine": 0.7703772546384692,
"relative_weight_rmse": 0.012863320602021833
},
{
"index": 113,
"total": 252,
"projection": "layers.23.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4257,
"parameters": 50331648,
"reference_delta_l2": 3.023025982853665,
"retained_delta_l2": 2.7483078798014526,
"delta_retention_norm_ratio": 0.9091247959460522,
"delta_cosine": 0.7250317710698787,
"relative_weight_rmse": 0.011629094049505571
},
{
"index": 114,
"total": 252,
"projection": "layers.23.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3865,
"parameters": 50331648,
"reference_delta_l2": 5.000263946159525,
"retained_delta_l2": 4.980002093083772,
"delta_retention_norm_ratio": 0.9959478432950892,
"delta_cosine": 0.831809651737954,
"relative_weight_rmse": 0.015881822890547393
},
{
"index": 115,
"total": 252,
"projection": "layers.23.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3843,
"parameters": 50331648,
"reference_delta_l2": 4.454117936701333,
"retained_delta_l2": 4.354295531700759,
"delta_retention_norm_ratio": 0.9775887377884518,
"delta_cosine": 0.8063541037430465,
"relative_weight_rmse": 0.014560552787199508
},
{
"index": 116,
"total": 252,
"projection": "layers.23.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0309,
"parameters": 4194304,
"reference_delta_l2": 0.7266939802548371,
"retained_delta_l2": 0.6303084490345241,
"delta_retention_norm_ratio": 0.8673643461495132,
"delta_cosine": 0.6826958743765825,
"relative_weight_rmse": 0.011273850499938792
},
{
"index": 117,
"total": 252,
"projection": "layers.23.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1313,
"parameters": 16777216,
"reference_delta_l2": 1.8946229502748384,
"retained_delta_l2": 1.7597239440078793,
"delta_retention_norm_ratio": 0.9287990223873356,
"delta_cosine": 0.7459816596101202,
"relative_weight_rmse": 0.012561695261136124
},
{
"index": 118,
"total": 252,
"projection": "layers.23.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1302,
"parameters": 16777216,
"reference_delta_l2": 1.7125996196835842,
"retained_delta_l2": 1.5415253508809772,
"delta_retention_norm_ratio": 0.9001084276579401,
"delta_cosine": 0.7177602975638901,
"relative_weight_rmse": 0.011632013912543447
},
{
"index": 119,
"total": 252,
"projection": "layers.23.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0297,
"parameters": 4194304,
"reference_delta_l2": 0.8868773646670587,
"retained_delta_l2": 0.8089120451245109,
"delta_retention_norm_ratio": 0.9120900784610545,
"delta_cosine": 0.7316220241406097,
"relative_weight_rmse": 0.011318348079675748
},
{
"index": 120,
"total": 252,
"projection": "layers.24.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4354,
"parameters": 50331648,
"reference_delta_l2": 3.1518128058745902,
"retained_delta_l2": 2.868971739546031,
"delta_retention_norm_ratio": 0.9102608296402063,
"delta_cosine": 0.7251792360856225,
"relative_weight_rmse": 0.012012265245717246
},
{
"index": 121,
"total": 252,
"projection": "layers.24.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3864,
"parameters": 50331648,
"reference_delta_l2": 4.587213119030321,
"retained_delta_l2": 4.488477555342334,
"delta_retention_norm_ratio": 0.9784759153050081,
"delta_cosine": 0.804670399667251,
"relative_weight_rmse": 0.015356809746409184
},
{
"index": 122,
"total": 252,
"projection": "layers.24.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3844,
"parameters": 50331648,
"reference_delta_l2": 3.7073519293553274,
"retained_delta_l2": 3.4827573602492916,
"delta_retention_norm_ratio": 0.9394191397564216,
"delta_cosine": 0.7562527399791867,
"relative_weight_rmse": 0.0133520489083886
},
{
"index": 123,
"total": 252,
"projection": "layers.24.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0311,
"parameters": 4194304,
"reference_delta_l2": 0.8845529687518748,
"retained_delta_l2": 0.8227142190408355,
"delta_retention_norm_ratio": 0.9300903937971117,
"delta_cosine": 0.7504080474234667,
"relative_weight_rmse": 0.012512043117811553
},
{
"index": 124,
"total": 252,
"projection": "layers.24.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1313,
"parameters": 16777216,
"reference_delta_l2": 1.817756095217153,
"retained_delta_l2": 1.6660417920596295,
"delta_retention_norm_ratio": 0.9165375907379922,
"delta_cosine": 0.7337509812864685,
"relative_weight_rmse": 0.012155999048714634
},
{
"index": 125,
"total": 252,
"projection": "layers.24.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.131,
"parameters": 16777216,
"reference_delta_l2": 2.9053735259013806,
"retained_delta_l2": 2.8934678342811617,
"delta_retention_norm_ratio": 0.9959021821070234,
"delta_cosine": 0.8437156082568049,
"relative_weight_rmse": 0.015184226591279349
},
{
"index": 126,
"total": 252,
"projection": "layers.24.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0299,
"parameters": 4194304,
"reference_delta_l2": 1.0337579943004718,
"retained_delta_l2": 0.9772839036433504,
"delta_retention_norm_ratio": 0.9453701050260449,
"delta_cosine": 0.7691690872666292,
"relative_weight_rmse": 0.012635955369569987
},
{
"index": 127,
"total": 252,
"projection": "layers.25.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4263,
"parameters": 50331648,
"reference_delta_l2": 3.445096597809559,
"retained_delta_l2": 3.186482018700214,
"delta_retention_norm_ratio": 0.9249325608826829,
"delta_cosine": 0.7393372155827732,
"relative_weight_rmse": 0.012808033611830865
},
{
"index": 128,
"total": 252,
"projection": "layers.25.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3849,
"parameters": 50331648,
"reference_delta_l2": 4.836393393002488,
"retained_delta_l2": 4.760006423474026,
"delta_retention_norm_ratio": 0.9842057989660268,
"delta_cosine": 0.8104606899433433,
"relative_weight_rmse": 0.01582758491989961
},
{
"index": 129,
"total": 252,
"projection": "layers.25.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3909,
"parameters": 50331648,
"reference_delta_l2": 3.5694317531546527,
"retained_delta_l2": 3.294683142876742,
"delta_retention_norm_ratio": 0.9230273530135185,
"delta_cosine": 0.7363368990986016,
"relative_weight_rmse": 0.013230779177507392
},
{
"index": 130,
"total": 252,
"projection": "layers.25.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0311,
"parameters": 4194304,
"reference_delta_l2": 0.9684747142564072,
"retained_delta_l2": 0.9139151451869285,
"delta_retention_norm_ratio": 0.9436644361836857,
"delta_cosine": 0.7667776638829132,
"relative_weight_rmse": 0.013066873828265414
},
{
"index": 131,
"total": 252,
"projection": "layers.25.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1295,
"parameters": 16777216,
"reference_delta_l2": 1.8723731810900412,
"retained_delta_l2": 1.7128064476755267,
"delta_retention_norm_ratio": 0.9147783491955276,
"delta_cosine": 0.728811149111852,
"relative_weight_rmse": 0.012857156453864262
},
{
"index": 132,
"total": 252,
"projection": "layers.25.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1307,
"parameters": 16777216,
"reference_delta_l2": 2.2995768828199923,
"retained_delta_l2": 2.237261999690431,
"delta_retention_norm_ratio": 0.9729015874202285,
"delta_cosine": 0.8042533755324931,
"relative_weight_rmse": 0.014049139264545053
},
{
"index": 133,
"total": 252,
"projection": "layers.25.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.03,
"parameters": 4194304,
"reference_delta_l2": 1.1434754687882178,
"retained_delta_l2": 1.0940195160007935,
"delta_retention_norm_ratio": 0.956749441385188,
"delta_cosine": 0.7792480631184642,
"relative_weight_rmse": 0.013155256927629555
},
{
"index": 134,
"total": 252,
"projection": "layers.26.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4295,
"parameters": 50331648,
"reference_delta_l2": 3.100600291359713,
"retained_delta_l2": 2.78999051565817,
"delta_retention_norm_ratio": 0.8998226967316285,
"delta_cosine": 0.7122871122785358,
"relative_weight_rmse": 0.011871667884522654
},
{
"index": 135,
"total": 252,
"projection": "layers.26.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3889,
"parameters": 50331648,
"reference_delta_l2": 4.893807320286038,
"retained_delta_l2": 4.8117371373593025,
"delta_retention_norm_ratio": 0.983229788678738,
"delta_cosine": 0.8084951726018041,
"relative_weight_rmse": 0.016007899478833562
},
{
"index": 136,
"total": 252,
"projection": "layers.26.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4059,
"parameters": 50331648,
"reference_delta_l2": 3.3465387865727174,
"retained_delta_l2": 3.0393184470170573,
"delta_retention_norm_ratio": 0.9081975858793817,
"delta_cosine": 0.7211040808517322,
"relative_weight_rmse": 0.012555611507868866
},
{
"index": 137,
"total": 252,
"projection": "layers.26.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.032,
"parameters": 4194304,
"reference_delta_l2": 1.029310458939815,
"retained_delta_l2": 0.9828256739356831,
"delta_retention_norm_ratio": 0.9548389073477297,
"delta_cosine": 0.7789737204154658,
"relative_weight_rmse": 0.013816204935721046
},
{
"index": 138,
"total": 252,
"projection": "layers.26.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1356,
"parameters": 16777216,
"reference_delta_l2": 2.0150367336082815,
"retained_delta_l2": 1.8735956356598982,
"delta_retention_norm_ratio": 0.9298071863458749,
"delta_cosine": 0.7467704867104442,
"relative_weight_rmse": 0.013268451660151346
},
{
"index": 139,
"total": 252,
"projection": "layers.26.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1345,
"parameters": 16777216,
"reference_delta_l2": 2.4415805033686278,
"retained_delta_l2": 2.3860146046965705,
"delta_retention_norm_ratio": 0.9772418322494819,
"delta_cosine": 0.8109300150687073,
"relative_weight_rmse": 0.014571772704839753
},
{
"index": 140,
"total": 252,
"projection": "layers.26.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 1.2568343213357946,
"retained_delta_l2": 1.2165891361066137,
"delta_retention_norm_ratio": 0.9679789256658687,
"delta_cosine": 0.7980823330124914,
"relative_weight_rmse": 0.014054609773679536
},
{
"index": 141,
"total": 252,
"projection": "layers.27.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4345,
"parameters": 50331648,
"reference_delta_l2": 2.9353829255202477,
"retained_delta_l2": 2.583391304076503,
"delta_retention_norm_ratio": 0.8800866427396827,
"delta_cosine": 0.6926504154991127,
"relative_weight_rmse": 0.011416576102284396
},
{
"index": 142,
"total": 252,
"projection": "layers.27.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.396,
"parameters": 50331648,
"reference_delta_l2": 4.621100690329421,
"retained_delta_l2": 4.51226220412324,
"delta_retention_norm_ratio": 0.9764474973605427,
"delta_cosine": 0.7987724775225853,
"relative_weight_rmse": 0.015467352890931468
},
{
"index": 143,
"total": 252,
"projection": "layers.27.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3952,
"parameters": 50331648,
"reference_delta_l2": 2.9433083517247884,
"retained_delta_l2": 2.5303005007309234,
"delta_retention_norm_ratio": 0.8596790408480848,
"delta_cosine": 0.6717519690332472,
"relative_weight_rmse": 0.01165486483635682
},
{
"index": 144,
"total": 252,
"projection": "layers.27.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 0.9194693899536039,
"retained_delta_l2": 0.8600552773735223,
"delta_retention_norm_ratio": 0.9353821745136293,
"delta_cosine": 0.7599026923746851,
"relative_weight_rmse": 0.013110172409810413
},
{
"index": 145,
"total": 252,
"projection": "layers.27.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.136,
"parameters": 16777216,
"reference_delta_l2": 1.7790627812408335,
"retained_delta_l2": 1.6067912785954226,
"delta_retention_norm_ratio": 0.9031672718568944,
"delta_cosine": 0.7163036853094985,
"relative_weight_rmse": 0.012496548305428564
},
{
"index": 146,
"total": 252,
"projection": "layers.27.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1364,
"parameters": 16777216,
"reference_delta_l2": 2.4260204123321056,
"retained_delta_l2": 2.3651331226871024,
"delta_retention_norm_ratio": 0.9749024001053341,
"delta_cosine": 0.808750994766525,
"relative_weight_rmse": 0.014580906870193738
},
{
"index": 147,
"total": 252,
"projection": "layers.27.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 1.242049958017081,
"retained_delta_l2": 1.211006410844666,
"delta_retention_norm_ratio": 0.9750062008600883,
"delta_cosine": 0.8028187484830746,
"relative_weight_rmse": 0.013984726316701689
},
{
"index": 148,
"total": 252,
"projection": "layers.28.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4407,
"parameters": 50331648,
"reference_delta_l2": 3.638692805640952,
"retained_delta_l2": 3.4087694478197332,
"delta_retention_norm_ratio": 0.9368115501630763,
"delta_cosine": 0.7541776733335835,
"relative_weight_rmse": 0.012819379374375835
},
{
"index": 149,
"total": 252,
"projection": "layers.28.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3981,
"parameters": 50331648,
"reference_delta_l2": 6.1861094377317905,
"retained_delta_l2": 6.282238923939749,
"delta_retention_norm_ratio": 1.0155395709008352,
"delta_cosine": 0.861449919033094,
"relative_weight_rmse": 0.0176093265485177
},
{
"index": 150,
"total": 252,
"projection": "layers.28.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.398,
"parameters": 50331648,
"reference_delta_l2": 3.1102966714154165,
"retained_delta_l2": 2.7260694584834826,
"delta_retention_norm_ratio": 0.8764660566102582,
"delta_cosine": 0.6877677824189267,
"relative_weight_rmse": 0.011968009684836535
},
{
"index": 151,
"total": 252,
"projection": "layers.28.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 0.9862083874382753,
"retained_delta_l2": 0.9299796439745039,
"delta_retention_norm_ratio": 0.9429849267355874,
"delta_cosine": 0.7629343145104028,
"relative_weight_rmse": 0.013695552985927896
},
{
"index": 152,
"total": 252,
"projection": "layers.28.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1485,
"parameters": 16777216,
"reference_delta_l2": 1.7954228131743857,
"retained_delta_l2": 1.6349143503376387,
"delta_retention_norm_ratio": 0.9106013014544685,
"delta_cosine": 0.7244370935225608,
"relative_weight_rmse": 0.012512480231727003
},
{
"index": 153,
"total": 252,
"projection": "layers.28.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1344,
"parameters": 16777216,
"reference_delta_l2": 2.0244268892946304,
"retained_delta_l2": 1.9141484041752483,
"delta_retention_norm_ratio": 0.9455260717477398,
"delta_cosine": 0.7690629313154679,
"relative_weight_rmse": 0.012917130709178035
},
{
"index": 154,
"total": 252,
"projection": "layers.28.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 1.024426379202668,
"retained_delta_l2": 0.9516176701958848,
"delta_retention_norm_ratio": 0.9289273387674264,
"delta_cosine": 0.7461564435433649,
"relative_weight_rmse": 0.012308951982824896
},
{
"index": 155,
"total": 252,
"projection": "layers.29.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4357,
"parameters": 50331648,
"reference_delta_l2": 3.5565162810659685,
"retained_delta_l2": 3.305066392719017,
"delta_retention_norm_ratio": 0.9292988226468666,
"delta_cosine": 0.745816707250538,
"relative_weight_rmse": 0.012565746413120983
},
{
"index": 156,
"total": 252,
"projection": "layers.29.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3937,
"parameters": 50331648,
"reference_delta_l2": 5.41711384810868,
"retained_delta_l2": 5.42151814200812,
"delta_retention_norm_ratio": 1.0008130332909615,
"delta_cosine": 0.8324762974622172,
"relative_weight_rmse": 0.016828774825352794
},
{
"index": 157,
"total": 252,
"projection": "layers.29.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3924,
"parameters": 50331648,
"reference_delta_l2": 3.424315115658444,
"retained_delta_l2": 3.101676017804758,
"delta_retention_norm_ratio": 0.905779962720619,
"delta_cosine": 0.7182981034319839,
"relative_weight_rmse": 0.012533934178832876
},
{
"index": 158,
"total": 252,
"projection": "layers.29.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0315,
"parameters": 4194304,
"reference_delta_l2": 0.9605145594104075,
"retained_delta_l2": 0.9092781367913338,
"delta_retention_norm_ratio": 0.9466573180832113,
"delta_cosine": 0.7835077347108719,
"relative_weight_rmse": 0.013053425356086912
},
{
"index": 159,
"total": 252,
"projection": "layers.29.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1386,
"parameters": 16777216,
"reference_delta_l2": 1.89051543914091,
"retained_delta_l2": 1.73202122115128,
"delta_retention_norm_ratio": 0.9161634892219378,
"delta_cosine": 0.7317391462712602,
"relative_weight_rmse": 0.012673621636119101
},
{
"index": 160,
"total": 252,
"projection": "layers.29.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1351,
"parameters": 16777216,
"reference_delta_l2": 2.874744189154029,
"retained_delta_l2": 2.8536079671021835,
"delta_retention_norm_ratio": 0.9926476163925858,
"delta_cosine": 0.8440291239415602,
"relative_weight_rmse": 0.015081184443721588
},
{
"index": 161,
"total": 252,
"projection": "layers.29.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0302,
"parameters": 4194304,
"reference_delta_l2": 1.0001311038921359,
"retained_delta_l2": 0.9315136588648855,
"delta_retention_norm_ratio": 0.9313915498076032,
"delta_cosine": 0.7529484221709883,
"relative_weight_rmse": 0.012423154016305886
},
{
"index": 162,
"total": 252,
"projection": "layers.3.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4353,
"parameters": 50331648,
"reference_delta_l2": 1.6328367078875345,
"retained_delta_l2": 1.3508772476926554,
"delta_retention_norm_ratio": 0.8273192543792935,
"delta_cosine": 0.6584088554881028,
"relative_weight_rmse": 0.009621364425890839
},
{
"index": 163,
"total": 252,
"projection": "layers.3.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.392,
"parameters": 50331648,
"reference_delta_l2": 2.50532351776761,
"retained_delta_l2": 2.224602902980253,
"delta_retention_norm_ratio": 0.8879503533988716,
"delta_cosine": 0.7076553575863995,
"relative_weight_rmse": 0.011101486563117465
},
{
"index": 164,
"total": 252,
"projection": "layers.3.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3897,
"parameters": 50331648,
"reference_delta_l2": 2.7206147817330204,
"retained_delta_l2": 2.607509862283104,
"delta_retention_norm_ratio": 0.9584267055338613,
"delta_cosine": 0.8175185854428151,
"relative_weight_rmse": 0.013394276951894962
},
{
"index": 165,
"total": 252,
"projection": "layers.3.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0312,
"parameters": 4194304,
"reference_delta_l2": 0.5357802240764118,
"retained_delta_l2": 0.39576030078112745,
"delta_retention_norm_ratio": 0.738661643332108,
"delta_cosine": 0.5649501893590434,
"relative_weight_rmse": 0.008909499721650315
},
{
"index": 166,
"total": 252,
"projection": "layers.3.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1348,
"parameters": 16777216,
"reference_delta_l2": 1.6799251397544268,
"retained_delta_l2": 1.5044440246540813,
"delta_retention_norm_ratio": 0.8955423006968053,
"delta_cosine": 0.7088951660126287,
"relative_weight_rmse": 0.011989444728619451
},
{
"index": 167,
"total": 252,
"projection": "layers.3.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1337,
"parameters": 16777216,
"reference_delta_l2": 1.4293373412915793,
"retained_delta_l2": 1.2224811667208624,
"delta_retention_norm_ratio": 0.8552782687508903,
"delta_cosine": 0.6705742277797081,
"relative_weight_rmse": 0.011213731400939248
},
{
"index": 168,
"total": 252,
"projection": "layers.3.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0323,
"parameters": 4194304,
"reference_delta_l2": 0.8352270904275945,
"retained_delta_l2": 0.7284911564395894,
"delta_retention_norm_ratio": 0.8722072892375154,
"delta_cosine": 0.6869123726427542,
"relative_weight_rmse": 0.01133900766299753
},
{
"index": 169,
"total": 252,
"projection": "layers.30.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4362,
"parameters": 50331648,
"reference_delta_l2": 3.6032802907362416,
"retained_delta_l2": 3.3594063182533183,
"delta_retention_norm_ratio": 0.932318900333703,
"delta_cosine": 0.74834901740462,
"relative_weight_rmse": 0.01256070797211164
},
{
"index": 170,
"total": 252,
"projection": "layers.30.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3912,
"parameters": 50331648,
"reference_delta_l2": 4.974662185288301,
"retained_delta_l2": 4.915490479267123,
"delta_retention_norm_ratio": 0.9881053820707328,
"delta_cosine": 0.8134401687949316,
"relative_weight_rmse": 0.016286688583266058
},
{
"index": 171,
"total": 252,
"projection": "layers.30.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3948,
"parameters": 50331648,
"reference_delta_l2": 3.425780991411554,
"retained_delta_l2": 3.0838601650457007,
"delta_retention_norm_ratio": 0.9001918607105796,
"delta_cosine": 0.7118132612098147,
"relative_weight_rmse": 0.012553027922354673
},
{
"index": 172,
"total": 252,
"projection": "layers.30.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.031,
"parameters": 4194304,
"reference_delta_l2": 1.454615150624287,
"retained_delta_l2": 1.4618871442415353,
"delta_retention_norm_ratio": 1.0049992560672334,
"delta_cosine": 0.8472728064664286,
"relative_weight_rmse": 0.016429652274220923
},
{
"index": 173,
"total": 252,
"projection": "layers.30.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1387,
"parameters": 16777216,
"reference_delta_l2": 1.8134549324581692,
"retained_delta_l2": 1.6305396279163396,
"delta_retention_norm_ratio": 0.8991343532900017,
"delta_cosine": 0.7111653079940985,
"relative_weight_rmse": 0.012591365368045006
},
{
"index": 174,
"total": 252,
"projection": "layers.30.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1437,
"parameters": 16777216,
"reference_delta_l2": 2.9316027309299133,
"retained_delta_l2": 2.934824171507668,
"delta_retention_norm_ratio": 1.001098866686051,
"delta_cosine": 0.8501326455787492,
"relative_weight_rmse": 0.016070274349163405
},
{
"index": 175,
"total": 252,
"projection": "layers.30.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0302,
"parameters": 4194304,
"reference_delta_l2": 1.2023008133072424,
"retained_delta_l2": 1.152090852601463,
"delta_retention_norm_ratio": 0.9582384373776945,
"delta_cosine": 0.7856136728802574,
"relative_weight_rmse": 0.013207984268747731
},
{
"index": 176,
"total": 252,
"projection": "layers.31.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4331,
"parameters": 50331648,
"reference_delta_l2": 3.8810682441090334,
"retained_delta_l2": 3.6820452560203676,
"delta_retention_norm_ratio": 0.9487195340121222,
"delta_cosine": 0.7672569290578198,
"relative_weight_rmse": 0.012947203877242058
},
{
"index": 177,
"total": 252,
"projection": "layers.31.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3909,
"parameters": 50331648,
"reference_delta_l2": 3.9223406788622532,
"retained_delta_l2": 3.7336836843750696,
"delta_retention_norm_ratio": 0.9519019356212813,
"delta_cosine": 0.7693485087735656,
"relative_weight_rmse": 0.01417219648585819
},
{
"index": 178,
"total": 252,
"projection": "layers.31.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3899,
"parameters": 50331648,
"reference_delta_l2": 3.623399834624329,
"retained_delta_l2": 3.3226520060300264,
"delta_retention_norm_ratio": 0.9169984428104154,
"delta_cosine": 0.7302631770450294,
"relative_weight_rmse": 0.01281987193963721
},
{
"index": 179,
"total": 252,
"projection": "layers.31.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0313,
"parameters": 4194304,
"reference_delta_l2": 1.058268080045612,
"retained_delta_l2": 1.0259746177602465,
"delta_retention_norm_ratio": 0.9694846108521258,
"delta_cosine": 0.7989432352726251,
"relative_weight_rmse": 0.013939990322789822
},
{
"index": 180,
"total": 252,
"projection": "layers.31.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1345,
"parameters": 16777216,
"reference_delta_l2": 1.8867381401875651,
"retained_delta_l2": 1.735412364638458,
"delta_retention_norm_ratio": 0.9197950302027268,
"delta_cosine": 0.7354900241339846,
"relative_weight_rmse": 0.01234021488469923
},
{
"index": 181,
"total": 252,
"projection": "layers.31.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1345,
"parameters": 16777216,
"reference_delta_l2": 2.9886940841019882,
"retained_delta_l2": 3.005044412553113,
"delta_retention_norm_ratio": 1.0054707266756064,
"delta_cosine": 0.8543355348065301,
"relative_weight_rmse": 0.016057610254095186
},
{
"index": 182,
"total": 252,
"projection": "layers.31.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 0.8857199160991623,
"retained_delta_l2": 0.77501534643497,
"delta_retention_norm_ratio": 0.8750117642699612,
"delta_cosine": 0.6921349673342735,
"relative_weight_rmse": 0.011009704772931905
},
{
"index": 183,
"total": 252,
"projection": "layers.32.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4328,
"parameters": 50331648,
"reference_delta_l2": 2.896536184285432,
"retained_delta_l2": 2.5290553029548155,
"delta_retention_norm_ratio": 0.8731309198468469,
"delta_cosine": 0.686871665027865,
"relative_weight_rmse": 0.010863571786767239
},
{
"index": 184,
"total": 252,
"projection": "layers.32.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.392,
"parameters": 50331648,
"reference_delta_l2": 5.160132814296683,
"retained_delta_l2": 5.147276654109073,
"delta_retention_norm_ratio": 0.9975085602153514,
"delta_cosine": 0.8283919781893334,
"relative_weight_rmse": 0.016636527728732507
},
{
"index": 185,
"total": 252,
"projection": "layers.32.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3902,
"parameters": 50331648,
"reference_delta_l2": 3.9571520455669007,
"retained_delta_l2": 3.7115304062341608,
"delta_retention_norm_ratio": 0.9379296937533891,
"delta_cosine": 0.7552081979176619,
"relative_weight_rmse": 0.013379703382347076
},
{
"index": 186,
"total": 252,
"projection": "layers.32.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0308,
"parameters": 4194304,
"reference_delta_l2": 1.0581180770275143,
"retained_delta_l2": 1.0206194721872686,
"delta_retention_norm_ratio": 0.9645610394015878,
"delta_cosine": 0.7934770464785277,
"relative_weight_rmse": 0.014015358346925523
},
{
"index": 187,
"total": 252,
"projection": "layers.32.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1353,
"parameters": 16777216,
"reference_delta_l2": 1.9059274659853356,
"retained_delta_l2": 1.7442278343465185,
"delta_retention_norm_ratio": 0.9151596088914009,
"delta_cosine": 0.7290675782804628,
"relative_weight_rmse": 0.012663736807892926
},
{
"index": 188,
"total": 252,
"projection": "layers.32.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1347,
"parameters": 16777216,
"reference_delta_l2": 3.386354161601392,
"retained_delta_l2": 3.437371326120008,
"delta_retention_norm_ratio": 1.0150655135535176,
"delta_cosine": 0.8787210072049341,
"relative_weight_rmse": 0.016779092018824694
},
{
"index": 189,
"total": 252,
"projection": "layers.32.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0302,
"parameters": 4194304,
"reference_delta_l2": 1.035264008716314,
"retained_delta_l2": 0.9559619705252885,
"delta_retention_norm_ratio": 0.923399212642042,
"delta_cosine": 0.742351791872357,
"relative_weight_rmse": 0.01211994296760845
},
{
"index": 190,
"total": 252,
"projection": "layers.33.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4322,
"parameters": 50331648,
"reference_delta_l2": 4.158056237987143,
"retained_delta_l2": 4.004873229183985,
"delta_retention_norm_ratio": 0.9631599478131849,
"delta_cosine": 0.7858069270485946,
"relative_weight_rmse": 0.013473337260780828
},
{
"index": 191,
"total": 252,
"projection": "layers.33.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.393,
"parameters": 50331648,
"reference_delta_l2": 3.5037537092766597,
"retained_delta_l2": 3.244679998070068,
"delta_retention_norm_ratio": 0.9260582413311018,
"delta_cosine": 0.7390390031138095,
"relative_weight_rmse": 0.013600405401520944
},
{
"index": 192,
"total": 252,
"projection": "layers.33.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3919,
"parameters": 50331648,
"reference_delta_l2": 4.153804678041435,
"retained_delta_l2": 3.9647059325031586,
"delta_retention_norm_ratio": 0.954475773370394,
"delta_cosine": 0.7760354082093616,
"relative_weight_rmse": 0.013498489161895274
},
{
"index": 193,
"total": 252,
"projection": "layers.33.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0311,
"parameters": 4194304,
"reference_delta_l2": 1.163209272843663,
"retained_delta_l2": 1.1531137400696825,
"delta_retention_norm_ratio": 0.9913209660465478,
"delta_cosine": 0.8259170052841797,
"relative_weight_rmse": 0.015514670628493776
},
{
"index": 194,
"total": 252,
"projection": "layers.33.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1356,
"parameters": 16777216,
"reference_delta_l2": 1.917218370680728,
"retained_delta_l2": 1.7672382025223023,
"delta_retention_norm_ratio": 0.9217719950674301,
"delta_cosine": 0.7383447960236257,
"relative_weight_rmse": 0.012106365343981027
},
{
"index": 195,
"total": 252,
"projection": "layers.33.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1346,
"parameters": 16777216,
"reference_delta_l2": 1.6553096736770674,
"retained_delta_l2": 1.5189603787286248,
"delta_retention_norm_ratio": 0.9176291318073679,
"delta_cosine": 0.7383709968763957,
"relative_weight_rmse": 0.011720228551450676
},
{
"index": 196,
"total": 252,
"projection": "layers.33.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 1.185793149834154,
"retained_delta_l2": 1.129896173369846,
"delta_retention_norm_ratio": 0.9528611069543403,
"delta_cosine": 0.7771329614959223,
"relative_weight_rmse": 0.012453042571658742
},
{
"index": 197,
"total": 252,
"projection": "layers.34.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4397,
"parameters": 50331648,
"reference_delta_l2": 2.8218776755036408,
"retained_delta_l2": 2.5233471752540573,
"delta_retention_norm_ratio": 0.8942085609021649,
"delta_cosine": 0.7087053800668754,
"relative_weight_rmse": 0.010829783070142294
},
{
"index": 198,
"total": 252,
"projection": "layers.34.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4388,
"parameters": 50331648,
"reference_delta_l2": 3.2385337740934514,
"retained_delta_l2": 2.9915522313131575,
"delta_retention_norm_ratio": 0.9237366166269393,
"delta_cosine": 0.7353802799688657,
"relative_weight_rmse": 0.012517406687409704
},
{
"index": 199,
"total": 252,
"projection": "layers.34.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.4274,
"parameters": 50331648,
"reference_delta_l2": 3.0235831105993602,
"retained_delta_l2": 2.6943013028485963,
"delta_retention_norm_ratio": 0.8910954997081292,
"delta_cosine": 0.7038879602808116,
"relative_weight_rmse": 0.011149633146402372
},
{
"index": 200,
"total": 252,
"projection": "layers.34.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0409,
"parameters": 4194304,
"reference_delta_l2": 0.7928188424322914,
"retained_delta_l2": 0.7317704908256062,
"delta_retention_norm_ratio": 0.9229983593485307,
"delta_cosine": 0.7535575556119191,
"relative_weight_rmse": 0.011670621769581675
},
{
"index": 201,
"total": 252,
"projection": "layers.34.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1509,
"parameters": 16777216,
"reference_delta_l2": 2.4645978849554875,
"retained_delta_l2": 2.4001892009258414,
"delta_retention_norm_ratio": 0.9738664532568122,
"delta_cosine": 0.7983828446517862,
"relative_weight_rmse": 0.014459432274182423
},
{
"index": 202,
"total": 252,
"projection": "layers.34.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1423,
"parameters": 16777216,
"reference_delta_l2": 2.695511834193117,
"retained_delta_l2": 2.699277898890852,
"delta_retention_norm_ratio": 1.0013971612552248,
"delta_cosine": 0.8541317814560678,
"relative_weight_rmse": 0.015023564746444064
},
{
"index": 203,
"total": 252,
"projection": "layers.34.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0346,
"parameters": 4194304,
"reference_delta_l2": 1.0207525569136708,
"retained_delta_l2": 0.9269824099360373,
"delta_retention_norm_ratio": 0.9081362605045485,
"delta_cosine": 0.7288602731730222,
"relative_weight_rmse": 0.01164327693543864
},
{
"index": 204,
"total": 252,
"projection": "layers.35.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4324,
"parameters": 50331648,
"reference_delta_l2": 1.9823095213349184,
"retained_delta_l2": 1.6833332043325306,
"delta_retention_norm_ratio": 0.8491777828918198,
"delta_cosine": 0.6694214090375227,
"relative_weight_rmse": 0.008826801286047611
},
{
"index": 205,
"total": 252,
"projection": "layers.35.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3937,
"parameters": 50331648,
"reference_delta_l2": 2.4543541404296376,
"retained_delta_l2": 2.095939159235461,
"delta_retention_norm_ratio": 0.8539676995710831,
"delta_cosine": 0.6683413065179096,
"relative_weight_rmse": 0.010069223796027413
},
{
"index": 206,
"total": 252,
"projection": "layers.35.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3909,
"parameters": 50331648,
"reference_delta_l2": 2.499227317560434,
"retained_delta_l2": 2.1108465532485687,
"delta_retention_norm_ratio": 0.8445996642310335,
"delta_cosine": 0.6601572203554983,
"relative_weight_rmse": 0.009871612295636547
},
{
"index": 207,
"total": 252,
"projection": "layers.35.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0331,
"parameters": 4194304,
"reference_delta_l2": 0.7804828082506937,
"retained_delta_l2": 0.7046762074304083,
"delta_retention_norm_ratio": 0.902872170893563,
"delta_cosine": 0.7212488537699842,
"relative_weight_rmse": 0.011677823781018355
},
{
"index": 208,
"total": 252,
"projection": "layers.35.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1351,
"parameters": 16777216,
"reference_delta_l2": 1.9358674579050277,
"retained_delta_l2": 1.826476966807879,
"delta_retention_norm_ratio": 0.9434927785730074,
"delta_cosine": 0.7613448331168875,
"relative_weight_rmse": 0.012846772474266906
},
{
"index": 209,
"total": 252,
"projection": "layers.35.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1376,
"parameters": 16777216,
"reference_delta_l2": 1.789558247625012,
"retained_delta_l2": 1.6807697827683608,
"delta_retention_norm_ratio": 0.9392093188355123,
"delta_cosine": 0.7543470277448979,
"relative_weight_rmse": 0.012638227777340855
},
{
"index": 210,
"total": 252,
"projection": "layers.35.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0314,
"parameters": 4194304,
"reference_delta_l2": 0.8107990141519036,
"retained_delta_l2": 0.6923380793625497,
"delta_retention_norm_ratio": 0.8538960547290944,
"delta_cosine": 0.6701267360411922,
"relative_weight_rmse": 0.010610346900305285
},
{
"index": 211,
"total": 252,
"projection": "layers.4.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4251,
"parameters": 50331648,
"reference_delta_l2": 2.043343408116534,
"retained_delta_l2": 1.7174886693961613,
"delta_retention_norm_ratio": 0.8405286466161204,
"delta_cosine": 0.6572729898384961,
"relative_weight_rmse": 0.01012551010778355
},
{
"index": 212,
"total": 252,
"projection": "layers.4.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4127,
"parameters": 50331648,
"reference_delta_l2": 3.0372171514712547,
"retained_delta_l2": 2.61912948708616,
"delta_retention_norm_ratio": 0.8623451523107034,
"delta_cosine": 0.6878635130036593,
"relative_weight_rmse": 0.01020563041830835
},
{
"index": 213,
"total": 252,
"projection": "layers.4.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3853,
"parameters": 50331648,
"reference_delta_l2": 3.1333670723021627,
"retained_delta_l2": 2.991480228291325,
"delta_retention_norm_ratio": 0.9547174522688177,
"delta_cosine": 0.7852368918844294,
"relative_weight_rmse": 0.013581849235775443
},
{
"index": 214,
"total": 252,
"projection": "layers.4.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0358,
"parameters": 4194304,
"reference_delta_l2": 0.5244352595642214,
"retained_delta_l2": 0.37542370971843775,
"delta_retention_norm_ratio": 0.7158628312489808,
"delta_cosine": 0.5453996124130298,
"relative_weight_rmse": 0.00857201591110074
},
{
"index": 215,
"total": 252,
"projection": "layers.4.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1337,
"parameters": 16777216,
"reference_delta_l2": 1.4702355284702315,
"retained_delta_l2": 1.236151778187749,
"delta_retention_norm_ratio": 0.8407848635476487,
"delta_cosine": 0.65436025809255,
"relative_weight_rmse": 0.01097918446677673
},
{
"index": 216,
"total": 252,
"projection": "layers.4.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.132,
"parameters": 16777216,
"reference_delta_l2": 1.1330369850333857,
"retained_delta_l2": 0.8538898964738426,
"delta_retention_norm_ratio": 0.753629323449386,
"delta_cosine": 0.5765922595156237,
"relative_weight_rmse": 0.009442583482949792
},
{
"index": 217,
"total": 252,
"projection": "layers.4.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0302,
"parameters": 4194304,
"reference_delta_l2": 0.9219868617247438,
"retained_delta_l2": 0.8307413349834236,
"delta_retention_norm_ratio": 0.9010338102100188,
"delta_cosine": 0.7170312757451748,
"relative_weight_rmse": 0.011933906834989327
},
{
"index": 218,
"total": 252,
"projection": "layers.5.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4244,
"parameters": 50331648,
"reference_delta_l2": 1.60459054270334,
"retained_delta_l2": 1.1396238620246248,
"delta_retention_norm_ratio": 0.7102272085591625,
"delta_cosine": 0.5413391457434584,
"relative_weight_rmse": 0.0082082789800158
},
{
"index": 219,
"total": 252,
"projection": "layers.5.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.387,
"parameters": 50331648,
"reference_delta_l2": 2.7167476668565116,
"retained_delta_l2": 2.226031133887409,
"delta_retention_norm_ratio": 0.8193735329358358,
"delta_cosine": 0.6379289789806206,
"relative_weight_rmse": 0.009413097609736909
},
{
"index": 220,
"total": 252,
"projection": "layers.5.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3857,
"parameters": 50331648,
"reference_delta_l2": 2.9329646658700654,
"retained_delta_l2": 2.6871599963746737,
"delta_retention_norm_ratio": 0.9161924204693842,
"delta_cosine": 0.7311448007579044,
"relative_weight_rmse": 0.012468887222045642
},
{
"index": 221,
"total": 252,
"projection": "layers.5.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0326,
"parameters": 4194304,
"reference_delta_l2": 0.5437987065128471,
"retained_delta_l2": 0.4047357132618228,
"delta_retention_norm_ratio": 0.7442748730632757,
"delta_cosine": 0.5708455477077047,
"relative_weight_rmse": 0.008915155872740532
},
{
"index": 222,
"total": 252,
"projection": "layers.5.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1322,
"parameters": 16777216,
"reference_delta_l2": 1.352433111519197,
"retained_delta_l2": 1.1069321443420719,
"delta_retention_norm_ratio": 0.8184745958331705,
"delta_cosine": 0.6329988325220858,
"relative_weight_rmse": 0.010657983726530096
},
{
"index": 223,
"total": 252,
"projection": "layers.5.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1327,
"parameters": 16777216,
"reference_delta_l2": 1.2645682741168707,
"retained_delta_l2": 1.0160588001837498,
"delta_retention_norm_ratio": 0.8034827545340159,
"delta_cosine": 0.6220893589407358,
"relative_weight_rmse": 0.010130382111453774
},
{
"index": 224,
"total": 252,
"projection": "layers.5.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0305,
"parameters": 4194304,
"reference_delta_l2": 0.9997299186419492,
"retained_delta_l2": 0.9316298816862428,
"delta_retention_norm_ratio": 0.9318815655250023,
"delta_cosine": 0.7544580375427529,
"relative_weight_rmse": 0.012434136650850247
},
{
"index": 225,
"total": 252,
"projection": "layers.6.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.428,
"parameters": 50331648,
"reference_delta_l2": 4.376686711230453,
"retained_delta_l2": 4.350106184769022,
"delta_retention_norm_ratio": 0.9939267925224744,
"delta_cosine": 0.8288901954627105,
"relative_weight_rmse": 0.014787761817631922
},
{
"index": 226,
"total": 252,
"projection": "layers.6.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3997,
"parameters": 50331648,
"reference_delta_l2": 2.9074557058505683,
"retained_delta_l2": 2.5005355483480867,
"delta_retention_norm_ratio": 0.8600425256062711,
"delta_cosine": 0.6776434227254817,
"relative_weight_rmse": 0.01017618471205037
},
{
"index": 227,
"total": 252,
"projection": "layers.6.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3888,
"parameters": 50331648,
"reference_delta_l2": 3.1183011367968954,
"retained_delta_l2": 2.886978332990382,
"delta_retention_norm_ratio": 0.9258176828796826,
"delta_cosine": 0.7412652873621648,
"relative_weight_rmse": 0.01262435683416308
},
{
"index": 228,
"total": 252,
"projection": "layers.6.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0318,
"parameters": 4194304,
"reference_delta_l2": 0.5506121946790155,
"retained_delta_l2": 0.4057511541629519,
"delta_retention_norm_ratio": 0.7369091314795313,
"delta_cosine": 0.5636920049824873,
"relative_weight_rmse": 0.00899741538593721
},
{
"index": 229,
"total": 252,
"projection": "layers.6.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1399,
"parameters": 16777216,
"reference_delta_l2": 1.2370675869345922,
"retained_delta_l2": 0.9687512486064946,
"delta_retention_norm_ratio": 0.7831029273081388,
"delta_cosine": 0.6026394837415746,
"relative_weight_rmse": 0.00962555464796606
},
{
"index": 230,
"total": 252,
"projection": "layers.6.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1372,
"parameters": 16777216,
"reference_delta_l2": 1.2723583985596465,
"retained_delta_l2": 1.0190259907621784,
"delta_retention_norm_ratio": 0.8008954017325234,
"delta_cosine": 0.6186974562229031,
"relative_weight_rmse": 0.009970888524534273
},
{
"index": 231,
"total": 252,
"projection": "layers.6.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.036,
"parameters": 4194304,
"reference_delta_l2": 0.9237469929926623,
"retained_delta_l2": 0.8381914238174702,
"delta_retention_norm_ratio": 0.9073820322835175,
"delta_cosine": 0.7221583677612461,
"relative_weight_rmse": 0.011920943866063252
},
{
"index": 232,
"total": 252,
"projection": "layers.7.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4406,
"parameters": 50331648,
"reference_delta_l2": 1.6011354845798527,
"retained_delta_l2": 1.114551863096439,
"delta_retention_norm_ratio": 0.696100907031552,
"delta_cosine": 0.5296439743814515,
"relative_weight_rmse": 0.00797270964724083
},
{
"index": 233,
"total": 252,
"projection": "layers.7.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.4062,
"parameters": 50331648,
"reference_delta_l2": 2.604621866268664,
"retained_delta_l2": 2.156010126213745,
"delta_retention_norm_ratio": 0.8277631982343785,
"delta_cosine": 0.6468744511766974,
"relative_weight_rmse": 0.009582833654521025
},
{
"index": 234,
"total": 252,
"projection": "layers.7.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3931,
"parameters": 50331648,
"reference_delta_l2": 2.714862796861589,
"retained_delta_l2": 2.3940188386459096,
"delta_retention_norm_ratio": 0.8818194574743968,
"delta_cosine": 0.6931481733288807,
"relative_weight_rmse": 0.011652355285720086
},
{
"index": 235,
"total": 252,
"projection": "layers.7.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0375,
"parameters": 4194304,
"reference_delta_l2": 0.6375207609376344,
"retained_delta_l2": 0.5155370261913551,
"delta_retention_norm_ratio": 0.8086591963422939,
"delta_cosine": 0.6255068307050059,
"relative_weight_rmse": 0.00986875980171051
},
{
"index": 236,
"total": 252,
"projection": "layers.7.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1368,
"parameters": 16777216,
"reference_delta_l2": 1.2552230316150195,
"retained_delta_l2": 0.982903345138241,
"delta_retention_norm_ratio": 0.7830507570225178,
"delta_cosine": 0.6021611665254776,
"relative_weight_rmse": 0.009895264289633533
},
{
"index": 237,
"total": 252,
"projection": "layers.7.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1363,
"parameters": 16777216,
"reference_delta_l2": 1.5359069696221528,
"retained_delta_l2": 1.3581859768630977,
"delta_retention_norm_ratio": 0.8842892204579448,
"delta_cosine": 0.6978932431003803,
"relative_weight_rmse": 0.011338741686984611
},
{
"index": 238,
"total": 252,
"projection": "layers.7.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0335,
"parameters": 4194304,
"reference_delta_l2": 0.7051042187633481,
"retained_delta_l2": 0.5652459705552851,
"delta_retention_norm_ratio": 0.8016488279514846,
"delta_cosine": 0.6220085894138233,
"relative_weight_rmse": 0.009523562848348033
},
{
"index": 239,
"total": 252,
"projection": "layers.8.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4275,
"parameters": 50331648,
"reference_delta_l2": 2.241443869715866,
"retained_delta_l2": 1.8272421294903742,
"delta_retention_norm_ratio": 0.8152076231656886,
"delta_cosine": 0.6324387939860261,
"relative_weight_rmse": 0.009842683384753201
},
{
"index": 240,
"total": 252,
"projection": "layers.8.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.389,
"parameters": 50331648,
"reference_delta_l2": 2.0020562075991806,
"retained_delta_l2": 1.45355364340143,
"delta_retention_norm_ratio": 0.7260303871011182,
"delta_cosine": 0.55432970546666,
"relative_weight_rmse": 0.008751578472771991
},
{
"index": 241,
"total": 252,
"projection": "layers.8.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3914,
"parameters": 50331648,
"reference_delta_l2": 2.3050198953747403,
"retained_delta_l2": 1.8581929482664734,
"delta_retention_norm_ratio": 0.8061505030803113,
"delta_cosine": 0.6234078341037631,
"relative_weight_rmse": 0.010262403171030403
},
{
"index": 242,
"total": 252,
"projection": "layers.8.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0328,
"parameters": 4194304,
"reference_delta_l2": 0.8142274162935644,
"retained_delta_l2": 0.7196864411222547,
"delta_retention_norm_ratio": 0.8838887351623841,
"delta_cosine": 0.6959786563698768,
"relative_weight_rmse": 0.01141390426469825
},
{
"index": 243,
"total": 252,
"projection": "layers.8.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1324,
"parameters": 16777216,
"reference_delta_l2": 1.3767169569898148,
"retained_delta_l2": 1.125127602026786,
"delta_retention_norm_ratio": 0.817254117714125,
"delta_cosine": 0.6336929939668535,
"relative_weight_rmse": 0.01029017423529436
},
{
"index": 244,
"total": 252,
"projection": "layers.8.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1346,
"parameters": 16777216,
"reference_delta_l2": 1.4916791552785456,
"retained_delta_l2": 1.264670684597102,
"delta_retention_norm_ratio": 0.8478168244973207,
"delta_cosine": 0.6609397318833818,
"relative_weight_rmse": 0.011302634113662864
},
{
"index": 245,
"total": 252,
"projection": "layers.8.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0343,
"parameters": 4194304,
"reference_delta_l2": 0.7376420382648309,
"retained_delta_l2": 0.5951106031452117,
"delta_retention_norm_ratio": 0.8067742512955219,
"delta_cosine": 0.6237091618733237,
"relative_weight_rmse": 0.01003165832577633
},
{
"index": 246,
"total": 252,
"projection": "layers.9.mlp.down_proj",
"shape": [
4096,
12288
],
"seconds": 0.4352,
"parameters": 50331648,
"reference_delta_l2": 2.198025661201102,
"retained_delta_l2": 1.7953132726162053,
"delta_retention_norm_ratio": 0.8167844917857618,
"delta_cosine": 0.6336958049607184,
"relative_weight_rmse": 0.009999010507091
},
{
"index": 247,
"total": 252,
"projection": "layers.9.mlp.gate_proj",
"shape": [
12288,
4096
],
"seconds": 0.3882,
"parameters": 50331648,
"reference_delta_l2": 2.367346213681673,
"retained_delta_l2": 1.8376129596715194,
"delta_retention_norm_ratio": 0.7762332983031163,
"delta_cosine": 0.597617687706116,
"relative_weight_rmse": 0.009410607398687657
},
{
"index": 248,
"total": 252,
"projection": "layers.9.mlp.up_proj",
"shape": [
12288,
4096
],
"seconds": 0.3829,
"parameters": 50331648,
"reference_delta_l2": 2.7911405036465617,
"retained_delta_l2": 2.4785721061854744,
"delta_retention_norm_ratio": 0.8880140942196483,
"delta_cosine": 0.7027169076037411,
"relative_weight_rmse": 0.011728165449823363
},
{
"index": 249,
"total": 252,
"projection": "layers.9.self_attn.k_proj",
"shape": [
1024,
4096
],
"seconds": 0.0327,
"parameters": 4194304,
"reference_delta_l2": 0.7479720000175083,
"retained_delta_l2": 0.6531119478001653,
"delta_retention_norm_ratio": 0.8731770009905151,
"delta_cosine": 0.6897070446896675,
"relative_weight_rmse": 0.010911995385218835
},
{
"index": 250,
"total": 252,
"projection": "layers.9.self_attn.o_proj",
"shape": [
4096,
4096
],
"seconds": 0.1338,
"parameters": 16777216,
"reference_delta_l2": 1.3714384751859114,
"retained_delta_l2": 1.127018845066739,
"delta_retention_norm_ratio": 0.8217786400618234,
"delta_cosine": 0.6383344346502468,
"relative_weight_rmse": 0.010335860800388192
},
{
"index": 251,
"total": 252,
"projection": "layers.9.self_attn.q_proj",
"shape": [
4096,
4096
],
"seconds": 0.1312,
"parameters": 16777216,
"reference_delta_l2": 1.6757371128988487,
"retained_delta_l2": 1.5104975488994226,
"delta_retention_norm_ratio": 0.9013929077971072,
"delta_cosine": 0.7160380408555146,
"relative_weight_rmse": 0.012124181065717473
},
{
"index": 252,
"total": 252,
"projection": "layers.9.self_attn.v_proj",
"shape": [
1024,
4096
],
"seconds": 0.0304,
"parameters": 4194304,
"reference_delta_l2": 0.6723826521492818,
"retained_delta_l2": 0.5284218432859524,
"delta_retention_norm_ratio": 0.7858945224075066,
"delta_cosine": 0.6078980628877407,
"relative_weight_rmse": 0.009142790598844487
}
]
}