| layer,module,loss,samples,damp,time | |
| 0,mlp.gate_proj,0.0002078072,0.05000,3.973 | |
| 0,mlp.up_proj,0.0002951170,0.05000,3.987 | |
| 0,mlp.down_proj,0.0001080021,0.05000,7.187 | |
| 0,self_attn.q_proj,0.0002751954,0.05000,5.447 | |
| 0,self_attn.v_proj,0.0000291499,0.05000,5.453 | |
| 0,self_attn.o_proj,0.0001638678,0.05000,5.458 | |
| 0,self_attn.k_proj,0.0000131150,0.05000,5.470 | |
| 1,mlp.gate_proj,0.0002358835,0.05000,3.468 | |
| 1,mlp.up_proj,0.0002490052,0.05000,3.477 | |
| 1,mlp.down_proj,0.0000623201,0.05000,6.719 | |
| 1,self_attn.k_proj,0.0021598425,0.05000,5.184 | |
| 1,self_attn.o_proj,0.0000640641,0.05000,5.191 | |
| 1,self_attn.v_proj,0.0027478211,0.05000,5.198 | |
| 1,self_attn.q_proj,0.0021248550,0.05000,5.201 | |
| 2,mlp.up_proj,0.0002725317,0.05000,3.898 | |
| 2,mlp.gate_proj,0.0002710722,0.05000,3.913 | |
| 2,mlp.down_proj,0.0000934939,0.05000,6.735 | |
| 2,self_attn.q_proj,0.0013359398,0.05000,5.192 | |
| 2,self_attn.k_proj,0.0013283169,0.05000,5.196 | |
| 2,self_attn.v_proj,0.0015758836,0.05000,5.199 | |
| 2,self_attn.o_proj,0.0000310727,0.05000,5.206 | |
| 3,mlp.gate_proj,0.0005309442,0.05000,3.730 | |
| 3,mlp.up_proj,0.0004692136,0.05000,3.741 | |
| 3,mlp.down_proj,0.0001151627,0.05000,6.528 | |
| 3,self_attn.k_proj,0.0015043542,0.05000,5.050 | |
| 3,self_attn.q_proj,0.0014609831,0.05000,5.212 | |
| 3,self_attn.o_proj,0.0000424297,0.05000,5.217 | |
| 3,self_attn.v_proj,0.0015774138,0.05000,5.225 | |
| 4,mlp.gate_proj,0.0006711201,0.05000,3.650 | |
| 4,mlp.up_proj,0.0006097663,0.05000,3.668 | |
| 4,mlp.down_proj,0.0001310055,0.05000,6.790 | |
| 4,self_attn.k_proj,0.0011175845,0.05000,5.169 | |
| 4,self_attn.o_proj,0.0000302304,0.05000,5.176 | |
| 4,self_attn.v_proj,0.0013201939,0.05000,5.178 | |
| 4,self_attn.q_proj,0.0010903492,0.05000,5.183 | |
| 5,mlp.gate_proj,0.0008319342,0.05000,3.374 | |
| 5,mlp.up_proj,0.0007571966,0.05000,3.555 | |
| 5,mlp.down_proj,0.0001388114,0.05000,6.471 | |
| 5,self_attn.o_proj,0.0000473348,0.05000,5.101 | |
| 5,self_attn.k_proj,0.0012779406,0.05000,5.275 | |
| 5,self_attn.q_proj,0.0012068739,0.05000,5.281 | |
| 5,self_attn.v_proj,0.0013603305,0.05000,5.284 | |
| 6,mlp.gate_proj,0.0010110473,0.05000,3.886 | |
| 6,mlp.up_proj,0.0009320581,0.05000,3.909 | |
| 6,mlp.down_proj,0.0001813578,0.05000,7.282 | |
| 6,self_attn.o_proj,0.0000507709,0.05000,5.565 | |
| 6,self_attn.q_proj,0.0011634061,0.05000,5.583 | |
| 6,self_attn.k_proj,0.0014005803,0.05000,5.584 | |
| 6,self_attn.v_proj,0.0012214420,0.05000,5.588 | |
| 7,mlp.gate_proj,0.0010606032,0.05000,3.618 | |
| 7,mlp.up_proj,0.0009992647,0.05000,3.634 | |
| 7,mlp.down_proj,0.0002215556,0.05000,6.559 | |
| 7,self_attn.v_proj,0.0011570467,0.05000,5.247 | |
| 7,self_attn.o_proj,0.0000582520,0.05000,5.254 | |
| 7,self_attn.k_proj,0.0012879103,0.05000,5.258 | |
| 7,self_attn.q_proj,0.0010551224,0.05000,5.269 | |
| 8,mlp.gate_proj,0.0011260961,0.05000,3.717 | |
| 8,mlp.up_proj,0.0010257109,0.05000,3.724 | |
| 8,mlp.down_proj,0.0002213782,0.05000,6.778 | |
| 8,self_attn.q_proj,0.0007443796,0.05000,5.545 | |
| 8,self_attn.o_proj,0.0000838215,0.05000,5.553 | |
| 8,self_attn.k_proj,0.0009184965,0.05000,5.581 | |
| 8,self_attn.v_proj,0.0008096808,0.05000,5.590 | |
| 9,mlp.up_proj,0.0011578900,0.05000,4.170 | |
| 9,mlp.gate_proj,0.0012150619,0.05000,4.172 | |
| 9,mlp.down_proj,0.0002476013,0.05000,7.254 | |
| 9,self_attn.v_proj,0.0000515016,0.05000,5.568 | |
| 9,self_attn.o_proj,0.0000509674,0.05000,5.584 | |
| 9,self_attn.k_proj,0.0000803375,0.05000,5.585 | |
| 9,self_attn.q_proj,0.0008326407,0.05000,5.593 | |
| 10,mlp.up_proj,0.0011782159,0.05000,4.128 | |
| 10,mlp.gate_proj,0.0012158477,0.05000,4.132 | |
| 10,mlp.down_proj,0.0002546134,0.05000,7.238 | |
| 10,self_attn.v_proj,0.0011085082,0.05000,5.713 | |
| 10,self_attn.q_proj,0.0010387903,0.05000,5.720 | |
| 10,self_attn.o_proj,0.0000718013,0.05000,5.722 | |
| 10,self_attn.k_proj,0.0013175182,0.05000,5.734 | |
| 11,mlp.gate_proj,0.0011684080,0.05000,3.892 | |
| 11,mlp.up_proj,0.0011704125,0.05000,3.898 | |
| 11,mlp.down_proj,0.0002768170,0.05000,6.982 | |
| 11,self_attn.o_proj,0.0000842501,0.05000,5.694 | |
| 11,self_attn.q_proj,0.0007722052,0.05000,5.701 | |
| 11,self_attn.k_proj,0.0009454727,0.05000,5.707 | |
| 11,self_attn.v_proj,0.0008207958,0.05000,5.712 | |
| 12,mlp.gate_proj,0.0010926098,0.05000,4.550 | |
| 12,mlp.up_proj,0.0011160669,0.05000,4.566 | |
| 12,mlp.down_proj,0.0002585554,0.05000,7.980 | |
| 12,self_attn.k_proj,0.0010605023,0.05000,5.557 | |
| 12,self_attn.o_proj,0.0000954078,0.05000,5.570 | |
| 12,self_attn.v_proj,0.0011184107,0.05000,5.594 | |
| 12,self_attn.q_proj,0.0009598745,0.05000,5.598 | |
| 13,mlp.gate_proj,0.0010929265,0.05000,4.198 | |
| 13,mlp.up_proj,0.0011373865,0.05000,4.201 | |
| 13,mlp.down_proj,0.0002754575,0.05000,7.376 | |
| 13,self_attn.o_proj,0.0001053596,0.05000,5.857 | |
| 13,self_attn.v_proj,0.0007298129,0.05000,5.866 | |
| 13,self_attn.k_proj,0.0009108803,0.05000,5.884 | |
| 13,self_attn.q_proj,0.0007128777,0.05000,5.896 | |
| 14,mlp.gate_proj,0.0010234204,0.05000,4.505 | |
| 14,mlp.up_proj,0.0011231728,0.05000,4.527 | |
| 14,mlp.down_proj,0.0002988453,0.05000,8.039 | |
| 14,self_attn.q_proj,0.0007725519,0.05000,5.838 | |
| 14,self_attn.k_proj,0.0008887160,0.05000,5.839 | |
| 14,self_attn.o_proj,0.0001257012,0.05000,5.849 | |
| 14,self_attn.v_proj,0.0007643019,0.05000,5.862 | |
| 15,mlp.gate_proj,0.0010563693,0.05000,4.093 | |
| 15,mlp.up_proj,0.0010843390,0.05000,4.112 | |
| 15,mlp.down_proj,0.0002757835,0.05000,7.433 | |
| 15,self_attn.o_proj,0.0001515282,0.05000,6.018 | |
| 15,self_attn.k_proj,0.0006521820,0.05000,6.034 | |
| 15,self_attn.v_proj,0.0005368739,0.05000,6.046 | |
| 15,self_attn.q_proj,0.0005058027,0.05000,6.051 | |
| 16,mlp.gate_proj,0.0012604057,0.05000,4.138 | |
| 16,mlp.up_proj,0.0012323827,0.05000,4.160 | |
| 16,mlp.down_proj,0.0002975532,0.05000,7.626 | |
| 16,self_attn.k_proj,0.0000700808,0.05000,5.996 | |
| 16,self_attn.q_proj,0.0007158842,0.05000,6.021 | |
| 16,self_attn.v_proj,0.0000399206,0.05000,6.019 | |
| 16,self_attn.o_proj,0.0001237841,0.05000,6.025 | |
| 17,mlp.gate_proj,0.0015615467,0.05000,4.348 | |
| 17,mlp.up_proj,0.0014232672,0.05000,4.384 | |
| 17,mlp.down_proj,0.0003209173,0.05000,8.060 | |
| 17,self_attn.k_proj,0.0000623059,0.05000,6.262 | |
| 17,self_attn.v_proj,0.0000560221,0.05000,6.262 | |
| 17,self_attn.o_proj,0.0001281610,0.05000,6.272 | |
| 17,self_attn.q_proj,0.0007087458,0.05000,6.279 | |
| 18,mlp.gate_proj,0.0014200928,0.05000,4.596 | |
| 18,mlp.up_proj,0.0013662458,0.05000,4.619 | |
| 18,mlp.down_proj,0.0003235204,0.05000,8.186 | |
| 18,self_attn.k_proj,0.0009590061,0.05000,6.615 | |
| 18,self_attn.o_proj,0.0001038908,0.05000,6.643 | |
| 18,self_attn.v_proj,0.0006829601,0.05000,6.646 | |
| 18,self_attn.q_proj,0.0006769612,0.05000,6.649 | |
| 19,mlp.gate_proj,0.0013718377,0.05000,4.395 | |
| 19,mlp.up_proj,0.0014014285,0.05000,4.661 | |
| 19,mlp.down_proj,0.0003881076,0.05000,8.269 | |
| 19,self_attn.k_proj,0.0011457390,0.05000,6.411 | |
| 19,self_attn.q_proj,0.0008744106,0.05000,6.428 | |
| 19,self_attn.o_proj,0.0001004489,0.05000,6.426 | |
| 19,self_attn.v_proj,0.0009188803,0.05000,6.433 | |
| 20,mlp.gate_proj,0.0013132144,0.05000,4.429 | |
| 20,mlp.up_proj,0.0013869754,0.05000,4.452 | |
| 20,mlp.down_proj,0.0004866157,0.05000,8.146 | |
| 20,self_attn.k_proj,0.0009195520,0.05000,6.902 | |
| 20,self_attn.o_proj,0.0001352315,0.05000,6.916 | |
| 20,self_attn.q_proj,0.0006837781,0.05000,6.921 | |
| 20,self_attn.v_proj,0.0006840548,0.05000,6.932 | |
| 21,mlp.gate_proj,0.0011899443,0.05000,4.188 | |
| 21,mlp.up_proj,0.0012786226,0.05000,4.204 | |
| 21,mlp.down_proj,0.0005073890,0.05000,7.768 | |
| 21,self_attn.o_proj,0.0002255129,0.05000,6.370 | |
| 21,self_attn.k_proj,0.0007886602,0.05000,6.380 | |
| 21,self_attn.q_proj,0.0006903726,0.05000,6.391 | |
| 21,self_attn.v_proj,0.0007393014,0.05000,6.388 | |
| 22,mlp.gate_proj,0.0013304233,0.05000,4.497 | |
| 22,mlp.up_proj,0.0014272174,0.05000,4.509 | |
| 22,mlp.down_proj,0.0006732390,0.05000,8.299 | |
| 22,self_attn.k_proj,0.0000527559,0.05000,6.658 | |
| 22,self_attn.o_proj,0.0002934376,0.05000,6.665 | |
| 22,self_attn.v_proj,0.0000915629,0.05000,6.692 | |
| 22,self_attn.q_proj,0.0008205932,0.05000,6.703 | |
| 23,mlp.gate_proj,0.0016920954,0.05000,4.612 | |
| 23,mlp.up_proj,0.0018172881,0.05000,4.613 | |
| 23,mlp.down_proj,0.0009602727,0.05000,8.565 | |
| 23,self_attn.o_proj,0.0002255482,0.05000,6.927 | |
| 23,self_attn.k_proj,0.0010880575,0.05000,6.935 | |
| 23,self_attn.q_proj,0.0008466790,0.05000,6.943 | |
| 23,self_attn.v_proj,0.0008170104,0.05000,6.953 | |
| 24,mlp.gate_proj,0.0019866878,0.05000,4.705 | |
| 24,mlp.up_proj,0.0021496204,0.05000,4.735 | |
| 24,mlp.down_proj,0.0016623680,0.05000,8.530 | |
| 24,self_attn.k_proj,0.0009090479,0.05000,7.095 | |
| 24,self_attn.o_proj,0.0003707870,0.05000,7.118 | |
| 24,self_attn.q_proj,0.0007286620,0.05000,7.141 | |
| 24,self_attn.v_proj,0.0007401827,0.05000,7.147 | |
| 25,mlp.gate_proj,0.0024015805,0.05000,5.336 | |
| 25,mlp.up_proj,0.0026404986,0.05000,5.350 | |
| 25,mlp.down_proj,0.0034734985,0.05000,9.255 | |
| 25,self_attn.o_proj,0.0006143702,0.05000,7.027 | |
| 25,self_attn.v_proj,0.0007805540,0.05000,7.053 | |
| 25,self_attn.q_proj,0.0007903687,0.05000,7.057 | |
| 25,self_attn.k_proj,0.0010052844,0.05000,7.063 | |
| 26,mlp.gate_proj,0.0027684612,0.05000,4.732 | |
| 26,mlp.up_proj,0.0031322542,0.05000,4.748 | |
| 26,mlp.down_proj,0.0031763896,0.05000,8.622 | |
| 26,self_attn.k_proj,0.0019508587,0.05000,6.850 | |
| 26,self_attn.q_proj,0.0013604624,0.05000,6.857 | |
| 26,self_attn.v_proj,0.0013274833,0.05000,6.865 | |
| 26,self_attn.o_proj,0.0008038559,0.05000,6.879 | |
| 27,mlp.up_proj,0.0037468435,0.05000,5.283 | |
| 27,mlp.gate_proj,0.0032518848,0.05000,5.293 | |
| 27,mlp.down_proj,0.0044657355,0.05000,9.521 | |
| 27,self_attn.o_proj,0.0006732306,0.05000,6.937 | |
| 27,self_attn.v_proj,0.0011655967,0.05000,6.945 | |
| 27,self_attn.q_proj,0.0012446561,0.05000,6.955 | |
| 27,self_attn.k_proj,0.0013658012,0.05000,6.962 | |
| 28,mlp.gate_proj,0.0037523353,0.05000,4.865 | |
| 28,mlp.up_proj,0.0044277377,0.05000,4.887 | |
| 28,mlp.down_proj,0.0062658060,0.05000,8.965 | |
| 28,self_attn.k_proj,0.0021060843,0.05000,8.730 | |
| 28,self_attn.v_proj,0.0015762776,0.05000,8.754 | |
| 28,self_attn.o_proj,0.0016443902,0.05000,8.777 | |
| 28,self_attn.q_proj,0.0015356449,0.05000,8.784 | |
| 29,mlp.up_proj,0.0052878110,0.05000,5.066 | |
| 29,mlp.gate_proj,0.0043512575,0.05000,5.077 | |
| 29,mlp.down_proj,0.0094577564,0.05000,9.052 | |
| 29,self_attn.v_proj,0.0008256575,0.05000,7.410 | |
| 29,self_attn.o_proj,0.0014812295,0.05000,7.426 | |
| 29,self_attn.k_proj,0.0000835427,0.05000,7.436 | |
| 29,self_attn.q_proj,0.0016312229,0.05000,7.449 | |
| 30,mlp.gate_proj,0.0052317987,0.05000,5.402 | |
| 30,mlp.up_proj,0.0063166269,0.05000,5.429 | |
| 30,mlp.down_proj,0.0177200560,0.05000,9.493 | |
| 30,self_attn.q_proj,0.0023054092,0.05000,7.067 | |
| 30,self_attn.k_proj,0.0000905518,0.05000,7.075 | |
| 30,self_attn.v_proj,0.0023724051,0.05000,7.086 | |
| 30,self_attn.o_proj,0.0039247042,0.05000,7.094 | |
| 31,mlp.up_proj,0.0094467415,0.05000,4.988 | |
| 31,mlp.gate_proj,0.0084637072,0.05000,5.021 | |
| 31,mlp.down_proj,0.0731122261,0.05000,8.989 | |
| 31,self_attn.o_proj,0.0030262335,0.05000,6.952 | |
| 31,self_attn.v_proj,0.0004133850,0.05000,6.961 | |
| 31,self_attn.q_proj,0.0017336883,0.05000,6.967 | |
| 31,self_attn.k_proj,0.0000637945,0.05000,6.973 | |