| layer,module,loss,samples,damp,time | |
| 0,self_attn.q_proj,0.0000001497,0.05000,4.382 | |
| 0,self_attn.k_proj,0.0000000404,0.05000,4.441 | |
| 0,self_attn.v_proj,0.0000000348,0.05000,4.350 | |
| 0,self_attn.o_proj,0.0000000850,0.05000,1.280 | |
| 0,mlp.gate_proj,0.0000102160,0.05000,1.724 | |
| 0,mlp.up_proj,0.0000089166,0.05000,1.743 | |
| 0,mlp.down_proj,0.0000010669,0.05000,3.375 | |
| 1,self_attn.q_proj,0.0000002582,0.05000,4.558 | |
| 1,self_attn.k_proj,0.0000000700,0.05000,4.591 | |
| 1,self_attn.v_proj,0.0000000709,0.05000,4.631 | |
| 1,self_attn.o_proj,0.0000001153,0.05000,1.476 | |
| 1,mlp.gate_proj,0.0003654537,0.05000,2.372 | |
| 1,mlp.up_proj,0.0001678912,0.05000,2.411 | |
| 1,mlp.down_proj,0.0000015869,0.05000,3.571 | |
| 2,self_attn.k_proj,0.0000002100,0.05000,4.218 | |
| 2,self_attn.v_proj,0.0000001973,0.05000,4.249 | |
| 2,self_attn.q_proj,0.0000007267,0.05000,4.284 | |
| 2,self_attn.o_proj,0.0000001442,0.05000,1.321 | |
| 2,mlp.gate_proj,0.0006378509,0.05000,1.795 | |
| 2,mlp.up_proj,0.0005036621,0.05000,1.815 | |
| 2,mlp.down_proj,0.0000010006,0.05000,3.541 | |
| 3,self_attn.v_proj,0.0000003527,0.05000,3.926 | |
| 3,self_attn.q_proj,0.0000013755,0.05000,4.000 | |
| 3,self_attn.k_proj,0.0000003573,0.05000,4.022 | |
| 3,self_attn.o_proj,0.0000002609,0.05000,1.337 | |
| 3,mlp.gate_proj,0.0005461829,0.05000,1.862 | |
| 3,mlp.up_proj,0.0002773691,0.05000,1.880 | |
| 3,mlp.down_proj,0.0000025155,0.05000,3.506 | |
| 4,self_attn.v_proj,0.0000006756,0.05000,3.921 | |
| 4,self_attn.k_proj,0.0000006994,0.05000,3.970 | |
| 4,self_attn.q_proj,0.0000025318,0.05000,4.003 | |
| 4,self_attn.o_proj,0.0000003630,0.05000,1.330 | |
| 4,mlp.up_proj,0.0001704913,0.05000,1.772 | |
| 4,mlp.gate_proj,0.0004582719,0.05000,1.796 | |
| 4,mlp.down_proj,0.0000030722,0.05000,3.513 | |
| 5,self_attn.q_proj,0.0000027408,0.05000,3.876 | |
| 5,self_attn.v_proj,0.0000007342,0.05000,3.951 | |
| 5,self_attn.k_proj,0.0000007074,0.05000,3.970 | |
| 5,self_attn.o_proj,0.0000006741,0.05000,1.321 | |
| 5,mlp.gate_proj,0.0001597727,0.05000,1.780 | |
| 5,mlp.up_proj,0.0000717189,0.05000,1.806 | |
| 5,mlp.down_proj,0.0000043840,0.05000,3.511 | |
| 6,self_attn.k_proj,0.0000014229,0.05000,3.949 | |
| 6,self_attn.q_proj,0.0000059482,0.05000,4.004 | |
| 6,self_attn.v_proj,0.0000015715,0.05000,4.024 | |
| 6,self_attn.o_proj,0.0000014801,0.05000,1.334 | |
| 6,mlp.up_proj,0.0000913276,0.05000,1.890 | |
| 6,mlp.gate_proj,0.0001743651,0.05000,1.909 | |
| 6,mlp.down_proj,0.0000798332,0.05000,3.506 | |
| 7,self_attn.k_proj,0.0000034002,0.05000,3.909 | |
| 7,self_attn.q_proj,0.0000122388,0.05000,3.980 | |
| 7,self_attn.v_proj,0.0000031944,0.05000,4.008 | |
| 7,self_attn.o_proj,0.0000018697,0.05000,1.365 | |
| 7,mlp.gate_proj,0.0001922215,0.05000,1.803 | |
| 7,mlp.up_proj,0.0001066429,0.05000,1.820 | |
| 7,mlp.down_proj,0.0000077401,0.05000,3.487 | |
| 8,self_attn.k_proj,0.0000046752,0.05000,3.969 | |
| 8,self_attn.v_proj,0.0000047731,0.05000,4.012 | |
| 8,self_attn.q_proj,0.0000178354,0.05000,4.034 | |
| 8,self_attn.o_proj,0.0000024056,0.05000,1.357 | |
| 8,mlp.gate_proj,0.0001238246,0.05000,1.818 | |
| 8,mlp.up_proj,0.0000997321,0.05000,1.841 | |
| 8,mlp.down_proj,0.0000107304,0.05000,3.535 | |
| 9,self_attn.k_proj,0.0000066289,0.05000,3.922 | |
| 9,self_attn.v_proj,0.0000057406,0.05000,3.985 | |
| 9,self_attn.q_proj,0.0000228142,0.05000,3.999 | |
| 9,self_attn.o_proj,0.0000030817,0.05000,1.339 | |
| 9,mlp.gate_proj,0.0001988376,0.05000,1.867 | |
| 9,mlp.up_proj,0.0001221854,0.05000,1.890 | |
| 9,mlp.down_proj,0.0000100834,0.05000,3.498 | |
| 10,self_attn.v_proj,0.0000088171,0.05000,3.950 | |
| 10,self_attn.q_proj,0.0000320740,0.05000,3.982 | |
| 10,self_attn.k_proj,0.0000083633,0.05000,3.991 | |
| 10,self_attn.o_proj,0.0000037997,0.05000,1.328 | |
| 10,mlp.gate_proj,0.0001482455,0.05000,1.833 | |
| 10,mlp.up_proj,0.0001002380,0.05000,1.847 | |
| 10,mlp.down_proj,0.0000081155,0.05000,3.547 | |
| 11,self_attn.q_proj,0.0000157204,0.05000,3.920 | |
| 11,self_attn.k_proj,0.0000043992,0.05000,3.961 | |
| 11,self_attn.v_proj,0.0000040354,0.05000,3.979 | |
| 11,self_attn.o_proj,0.0000023009,0.05000,1.361 | |
| 11,mlp.gate_proj,0.0001224862,0.05000,1.821 | |
| 11,mlp.up_proj,0.0000944792,0.05000,1.843 | |
| 11,mlp.down_proj,0.0000084048,0.05000,3.499 | |
| 12,self_attn.k_proj,0.0000053237,0.05000,3.876 | |
| 12,self_attn.q_proj,0.0000195351,0.05000,3.972 | |
| 12,self_attn.v_proj,0.0000052086,0.05000,4.002 | |
| 12,self_attn.o_proj,0.0000033271,0.05000,1.326 | |
| 12,mlp.up_proj,0.0000952316,0.05000,1.945 | |
| 12,mlp.gate_proj,0.0001117326,0.05000,1.964 | |
| 12,mlp.down_proj,0.0000084220,0.05000,3.492 | |
| 13,self_attn.k_proj,0.0000040553,0.05000,3.982 | |
| 13,self_attn.v_proj,0.0000035740,0.05000,4.028 | |
| 13,self_attn.q_proj,0.0000149411,0.05000,4.051 | |
| 13,self_attn.o_proj,0.0000025440,0.05000,1.361 | |
| 13,mlp.up_proj,0.0000980827,0.05000,1.869 | |
| 13,mlp.gate_proj,0.0001053022,0.05000,1.887 | |
| 13,mlp.down_proj,0.0000091428,0.05000,3.461 | |
| 14,self_attn.v_proj,0.0000062012,0.05000,3.923 | |
| 14,self_attn.q_proj,0.0000241531,0.05000,4.003 | |
| 14,self_attn.k_proj,0.0000063827,0.05000,4.028 | |
| 14,self_attn.o_proj,0.0000039574,0.05000,1.324 | |
| 14,mlp.gate_proj,0.0001051134,0.05000,1.836 | |
| 14,mlp.up_proj,0.0000985452,0.05000,1.853 | |
| 14,mlp.down_proj,0.0000083144,0.05000,3.479 | |
| 15,self_attn.k_proj,0.0000065399,0.05000,3.894 | |
| 15,self_attn.q_proj,0.0000246789,0.05000,3.961 | |
| 15,self_attn.v_proj,0.0000057165,0.05000,4.000 | |
| 15,self_attn.o_proj,0.0000033902,0.05000,1.349 | |
| 15,mlp.gate_proj,0.0000980051,0.05000,1.929 | |
| 15,mlp.up_proj,0.0000952363,0.05000,1.953 | |
| 15,mlp.down_proj,0.0000081789,0.05000,3.566 | |
| 16,self_attn.v_proj,0.0000111526,0.05000,3.887 | |
| 16,self_attn.k_proj,0.0000106205,0.05000,3.942 | |
| 16,self_attn.q_proj,0.0000424247,0.05000,3.978 | |
| 16,self_attn.o_proj,0.0000038802,0.05000,1.318 | |
| 16,mlp.up_proj,0.0000966814,0.05000,1.790 | |
| 16,mlp.gate_proj,0.0001058810,0.05000,1.820 | |
| 16,mlp.down_proj,0.0000451322,0.05000,3.495 | |
| 17,self_attn.q_proj,0.0000358253,0.05000,3.836 | |
| 17,self_attn.k_proj,0.0000080998,0.05000,3.941 | |
| 17,self_attn.v_proj,0.0000082012,0.05000,3.967 | |
| 17,self_attn.o_proj,0.0000036289,0.05000,1.328 | |
| 17,mlp.gate_proj,0.0000921126,0.05000,1.786 | |
| 17,mlp.up_proj,0.0000849465,0.05000,1.782 | |
| 17,mlp.down_proj,0.0000071025,0.05000,3.466 | |
| 18,self_attn.v_proj,0.0000099717,0.05000,3.973 | |
| 18,self_attn.k_proj,0.0000098200,0.05000,3.997 | |
| 18,self_attn.q_proj,0.0000393778,0.05000,4.048 | |
| 18,self_attn.o_proj,0.0000046202,0.05000,1.349 | |
| 18,mlp.up_proj,0.0000947463,0.05000,1.908 | |
| 18,mlp.gate_proj,0.0001002435,0.05000,1.925 | |
| 18,mlp.down_proj,0.0000094912,0.05000,3.519 | |
| 19,self_attn.v_proj,0.0000176808,0.05000,3.878 | |
| 19,self_attn.q_proj,0.0000786607,0.05000,3.975 | |
| 19,self_attn.k_proj,0.0000189547,0.05000,3.977 | |
| 19,self_attn.o_proj,0.0000072878,0.05000,1.341 | |
| 19,mlp.gate_proj,0.0001130672,0.05000,1.850 | |
| 19,mlp.up_proj,0.0001079505,0.05000,1.874 | |
| 19,mlp.down_proj,0.0000130190,0.05000,3.474 | |
| 20,self_attn.v_proj,0.0000165799,0.05000,3.919 | |
| 20,self_attn.k_proj,0.0000168190,0.05000,3.972 | |
| 20,self_attn.q_proj,0.0000756451,0.05000,3.986 | |
| 20,self_attn.o_proj,0.0000069506,0.05000,1.357 | |
| 20,mlp.gate_proj,0.0001250732,0.05000,1.799 | |
| 20,mlp.up_proj,0.0001220248,0.05000,1.819 | |
| 20,mlp.down_proj,0.0000152876,0.05000,3.501 | |
| 21,self_attn.v_proj,0.0000228874,0.05000,3.940 | |
| 21,self_attn.q_proj,0.0000968350,0.05000,3.979 | |
| 21,self_attn.k_proj,0.0000223913,0.05000,4.006 | |
| 21,self_attn.o_proj,0.0000090330,0.05000,1.340 | |
| 21,mlp.gate_proj,0.0001505156,0.05000,1.844 | |
| 21,mlp.up_proj,0.0001460762,0.05000,1.855 | |
| 21,mlp.down_proj,0.0000205126,0.05000,3.539 | |
| 22,self_attn.v_proj,0.0000470790,0.05000,3.991 | |
| 22,self_attn.q_proj,0.0002022357,0.05000,4.041 | |
| 22,self_attn.k_proj,0.0000466329,0.05000,4.060 | |
| 22,self_attn.o_proj,0.0000183918,0.05000,1.325 | |
| 22,mlp.gate_proj,0.0001807079,0.05000,1.864 | |
| 22,mlp.up_proj,0.0001656352,0.05000,1.887 | |
| 22,mlp.down_proj,0.0000324256,0.05000,3.467 | |
| 23,self_attn.k_proj,0.0000448676,0.05000,3.970 | |
| 23,self_attn.v_proj,0.0000475164,0.05000,4.023 | |
| 23,self_attn.q_proj,0.0001911028,0.05000,4.047 | |
| 23,self_attn.o_proj,0.0000203201,0.05000,1.321 | |
| 23,mlp.up_proj,0.0001968862,0.05000,1.826 | |
| 23,mlp.gate_proj,0.0002260412,0.05000,1.850 | |
| 23,mlp.down_proj,0.0000465797,0.05000,3.430 | |
| 24,self_attn.v_proj,0.0000918311,0.05000,3.915 | |
| 24,self_attn.q_proj,0.0003130213,0.05000,3.980 | |
| 24,self_attn.k_proj,0.0000751774,0.05000,4.005 | |
| 24,self_attn.o_proj,0.0000240284,0.05000,1.329 | |
| 24,mlp.up_proj,0.0002207510,0.05000,1.803 | |
| 24,mlp.gate_proj,0.0002575789,0.05000,1.812 | |
| 24,mlp.down_proj,0.0000600394,0.05000,3.719 | |
| 25,self_attn.q_proj,0.0002166719,0.05000,3.930 | |
| 25,self_attn.v_proj,0.0000578186,0.05000,3.994 | |
| 25,self_attn.k_proj,0.0000542144,0.05000,4.005 | |
| 25,self_attn.o_proj,0.0000134359,0.05000,1.332 | |
| 25,mlp.up_proj,0.0002558766,0.05000,1.811 | |
| 25,mlp.gate_proj,0.0003033456,0.05000,1.834 | |
| 25,mlp.down_proj,0.0000730246,0.05000,3.487 | |
| 26,self_attn.q_proj,0.0003547980,0.05000,3.916 | |
| 26,self_attn.k_proj,0.0000841941,0.05000,3.954 | |
| 26,self_attn.v_proj,0.0000925788,0.05000,3.989 | |
| 26,self_attn.o_proj,0.0000134945,0.05000,1.351 | |
| 26,mlp.gate_proj,0.0003733178,0.05000,1.903 | |
| 26,mlp.up_proj,0.0003256705,0.05000,1.927 | |
| 26,mlp.down_proj,0.0000952209,0.05000,3.523 | |
| 27,self_attn.q_proj,0.0004478611,0.05000,3.904 | |
| 27,self_attn.v_proj,0.0001275532,0.05000,3.981 | |
| 27,self_attn.k_proj,0.0001037366,0.05000,4.014 | |
| 27,self_attn.o_proj,0.0000206886,0.05000,1.319 | |
| 27,mlp.gate_proj,0.0004283897,0.05000,1.854 | |
| 27,mlp.up_proj,0.0003877850,0.05000,1.870 | |
| 27,mlp.down_proj,0.0001451184,0.05000,3.689 | |
| 28,self_attn.k_proj,0.0001290215,0.05000,3.955 | |
| 28,self_attn.v_proj,0.0001380381,0.05000,4.015 | |
| 28,self_attn.q_proj,0.0005028249,0.05000,4.041 | |
| 28,self_attn.o_proj,0.0000354018,0.05000,1.331 | |
| 28,mlp.gate_proj,0.0004989572,0.05000,1.777 | |
| 28,mlp.up_proj,0.0004736893,0.05000,1.800 | |
| 28,mlp.down_proj,0.0002259568,0.05000,3.485 | |
| 29,self_attn.q_proj,0.0012338733,0.05000,3.883 | |
| 29,self_attn.k_proj,0.0002937443,0.05000,3.952 | |
| 29,self_attn.v_proj,0.0003944995,0.05000,3.980 | |
| 29,self_attn.o_proj,0.0000378399,0.05000,1.396 | |
| 29,mlp.up_proj,0.0005982406,0.05000,1.797 | |
| 29,mlp.gate_proj,0.0005981093,0.05000,1.801 | |
| 29,mlp.down_proj,0.0002386291,0.05000,3.496 | |
| 30,self_attn.v_proj,0.0004350560,0.05000,3.897 | |
| 30,self_attn.q_proj,0.0014054731,0.05000,3.941 | |
| 30,self_attn.k_proj,0.0003603588,0.05000,3.986 | |
| 30,self_attn.o_proj,0.0000687038,0.05000,1.319 | |
| 30,mlp.gate_proj,0.0006366458,0.05000,1.815 | |
| 30,mlp.up_proj,0.0006604899,0.05000,1.813 | |
| 30,mlp.down_proj,0.0003442335,0.05000,3.540 | |
| 31,self_attn.k_proj,0.0005021534,0.05000,4.020 | |
| 31,self_attn.q_proj,0.0018248984,0.05000,4.098 | |
| 31,self_attn.v_proj,0.0006612717,0.05000,4.119 | |
| 31,self_attn.o_proj,0.0000640828,0.05000,1.338 | |
| 31,mlp.up_proj,0.0007156640,0.05000,1.771 | |
| 31,mlp.gate_proj,0.0006541759,0.05000,1.792 | |
| 31,mlp.down_proj,0.0004586547,0.05000,3.540 | |
| 32,self_attn.q_proj,0.0027697828,0.05000,3.898 | |
| 32,self_attn.k_proj,0.0006917736,0.05000,3.967 | |
| 32,self_attn.v_proj,0.0010175645,0.05000,4.006 | |
| 32,self_attn.o_proj,0.0001164538,0.05000,1.352 | |
| 32,mlp.up_proj,0.0007774354,0.05000,1.754 | |
| 32,mlp.gate_proj,0.0006933136,0.05000,1.779 | |
| 32,mlp.down_proj,0.0005796508,0.05000,3.510 | |
| 33,self_attn.q_proj,0.0060845944,0.05000,3.802 | |
| 33,self_attn.k_proj,0.0012694908,0.05000,3.924 | |
| 33,self_attn.v_proj,0.0023390502,0.05000,3.946 | |
| 33,self_attn.o_proj,0.0001119355,0.05000,1.334 | |
| 33,mlp.gate_proj,0.0007650941,0.05000,1.843 | |
| 33,mlp.up_proj,0.0008669923,0.05000,1.857 | |
| 33,mlp.down_proj,0.0007560028,0.05000,3.585 | |
| 34,self_attn.v_proj,0.0016800856,0.05000,3.933 | |
| 34,self_attn.q_proj,0.0047758973,0.05000,3.971 | |
| 34,self_attn.k_proj,0.0010792233,0.05000,4.010 | |
| 34,self_attn.o_proj,0.0002744614,0.05000,1.330 | |
| 34,mlp.gate_proj,0.0009299143,0.05000,1.851 | |
| 34,mlp.up_proj,0.0009779911,0.05000,1.870 | |
| 34,mlp.down_proj,0.0009601268,0.05000,3.484 | |
| 35,self_attn.q_proj,0.0022361877,0.05000,3.879 | |
| 35,self_attn.k_proj,0.0005822161,0.05000,3.916 | |
| 35,self_attn.v_proj,0.0007247474,0.05000,3.974 | |
| 35,self_attn.o_proj,0.0003144698,0.05000,1.357 | |
| 35,mlp.gate_proj,0.0015252715,0.05000,1.929 | |
| 35,mlp.up_proj,0.0015549625,0.05000,1.927 | |
| 35,mlp.down_proj,0.0016945056,0.05000,3.497 | |