| layer,module,loss,samples,damp,time | |
| 0,self_attn.v_proj,0.0000000372,0.05000,4.540 | |
| 0,self_attn.q_proj,0.0000001598,0.05000,4.596 | |
| 0,self_attn.k_proj,0.0000000433,0.05000,4.618 | |
| 0,self_attn.o_proj,0.0000000962,0.05000,1.349 | |
| 0,mlp.gate_proj,0.0000104463,0.05000,1.920 | |
| 0,mlp.up_proj,0.0000091199,0.05000,1.936 | |
| 0,mlp.down_proj,0.0000010827,0.05000,3.513 | |
| 1,self_attn.k_proj,0.0000000738,0.05000,4.644 | |
| 1,self_attn.v_proj,0.0000000748,0.05000,4.684 | |
| 1,self_attn.q_proj,0.0000002716,0.05000,4.696 | |
| 1,self_attn.o_proj,0.0000001209,0.05000,1.631 | |
| 1,mlp.up_proj,0.0001685409,0.05000,2.393 | |
| 1,mlp.gate_proj,0.0003610970,0.05000,2.439 | |
| 1,mlp.down_proj,0.0000016096,0.05000,3.772 | |
| 2,self_attn.q_proj,0.0000007576,0.05000,4.689 | |
| 2,self_attn.k_proj,0.0000002197,0.05000,4.747 | |
| 2,self_attn.v_proj,0.0000002069,0.05000,4.778 | |
| 2,self_attn.o_proj,0.0000001543,0.05000,1.372 | |
| 2,mlp.gate_proj,0.0006330384,0.05000,1.902 | |
| 2,mlp.up_proj,0.0005052607,0.05000,1.925 | |
| 2,mlp.down_proj,0.0000010256,0.05000,3.669 | |
| 3,self_attn.v_proj,0.0000003695,0.05000,3.944 | |
| 3,self_attn.q_proj,0.0000014327,0.05000,3.994 | |
| 3,self_attn.k_proj,0.0000003736,0.05000,4.033 | |
| 3,self_attn.o_proj,0.0000002812,0.05000,1.402 | |
| 3,mlp.gate_proj,0.0005489127,0.05000,1.989 | |
| 3,mlp.up_proj,0.0002831875,0.05000,2.018 | |
| 3,mlp.down_proj,0.0000025971,0.05000,3.661 | |
| 4,self_attn.v_proj,0.0000007139,0.05000,4.002 | |
| 4,self_attn.q_proj,0.0000026597,0.05000,4.038 | |
| 4,self_attn.k_proj,0.0000007332,0.05000,4.080 | |
| 4,self_attn.o_proj,0.0000004123,0.05000,1.373 | |
| 4,mlp.gate_proj,0.0004671095,0.05000,1.857 | |
| 4,mlp.up_proj,0.0001758254,0.05000,1.876 | |
| 4,mlp.down_proj,0.0000032607,0.05000,3.646 | |
| 5,self_attn.v_proj,0.0000007750,0.05000,4.157 | |
| 5,self_attn.q_proj,0.0000028777,0.05000,4.259 | |
| 5,self_attn.k_proj,0.0000007445,0.05000,4.266 | |
| 5,self_attn.o_proj,0.0000007169,0.05000,1.398 | |
| 5,mlp.up_proj,0.0000758844,0.05000,1.903 | |
| 5,mlp.gate_proj,0.0001681334,0.05000,1.902 | |
| 5,mlp.down_proj,0.0000047785,0.05000,3.754 | |
| 6,self_attn.v_proj,0.0000016682,0.05000,4.165 | |
| 6,self_attn.q_proj,0.0000062852,0.05000,4.216 | |
| 6,self_attn.k_proj,0.0000015054,0.05000,4.229 | |
| 6,self_attn.o_proj,0.0000016241,0.05000,1.420 | |
| 6,mlp.up_proj,0.0000998887,0.05000,1.973 | |
| 6,mlp.gate_proj,0.0001912913,0.05000,1.975 | |
| 6,mlp.down_proj,0.0001258073,0.05000,3.691 | |
| 7,self_attn.q_proj,0.0000131708,0.05000,4.055 | |
| 7,self_attn.v_proj,0.0000034307,0.05000,4.095 | |
| 7,self_attn.k_proj,0.0000036401,0.05000,4.149 | |
| 7,self_attn.o_proj,0.0000020525,0.05000,1.392 | |
| 7,mlp.up_proj,0.0001153550,0.05000,1.908 | |
| 7,mlp.gate_proj,0.0002064913,0.05000,1.924 | |
| 7,mlp.down_proj,0.0000085862,0.05000,3.667 | |
| 8,self_attn.v_proj,0.0000052083,0.05000,4.059 | |
| 8,self_attn.q_proj,0.0000193134,0.05000,4.104 | |
| 8,self_attn.k_proj,0.0000050779,0.05000,4.123 | |
| 8,self_attn.o_proj,0.0000026441,0.05000,1.434 | |
| 8,mlp.up_proj,0.0001105051,0.05000,1.911 | |
| 8,mlp.gate_proj,0.0001367360,0.05000,1.933 | |
| 8,mlp.down_proj,0.0000123248,0.05000,3.637 | |
| 9,self_attn.q_proj,0.0000244685,0.05000,3.993 | |
| 9,self_attn.v_proj,0.0000062576,0.05000,4.090 | |
| 9,self_attn.k_proj,0.0000071317,0.05000,4.112 | |
| 9,self_attn.o_proj,0.0000033297,0.05000,1.387 | |
| 9,mlp.up_proj,0.0001365071,0.05000,2.040 | |
| 9,mlp.gate_proj,0.0002204533,0.05000,2.060 | |
| 9,mlp.down_proj,0.0000117399,0.05000,3.677 | |
| 10,self_attn.k_proj,0.0000091113,0.05000,4.005 | |
| 10,self_attn.v_proj,0.0000097468,0.05000,4.016 | |
| 10,self_attn.q_proj,0.0000349093,0.05000,4.072 | |
| 10,self_attn.o_proj,0.0000044036,0.05000,1.447 | |
| 10,mlp.gate_proj,0.0001647009,0.05000,1.997 | |
| 10,mlp.up_proj,0.0001119096,0.05000,2.002 | |
| 10,mlp.down_proj,0.0000092788,0.05000,3.657 | |
| 11,self_attn.k_proj,0.0000046750,0.05000,4.090 | |
| 11,self_attn.v_proj,0.0000043463,0.05000,4.134 | |
| 11,self_attn.q_proj,0.0000167216,0.05000,4.149 | |
| 11,self_attn.o_proj,0.0000024583,0.05000,1.430 | |
| 11,mlp.gate_proj,0.0001339980,0.05000,1.945 | |
| 11,mlp.up_proj,0.0001037463,0.05000,1.967 | |
| 11,mlp.down_proj,0.0000093708,0.05000,3.644 | |
| 12,self_attn.v_proj,0.0000054699,0.05000,4.053 | |
| 12,self_attn.k_proj,0.0000055857,0.05000,4.050 | |
| 12,self_attn.q_proj,0.0000204089,0.05000,4.070 | |
| 12,self_attn.o_proj,0.0000035075,0.05000,1.391 | |
| 12,mlp.up_proj,0.0001045358,0.05000,2.000 | |
| 12,mlp.gate_proj,0.0001222810,0.05000,2.022 | |
| 12,mlp.down_proj,0.0000097573,0.05000,3.707 | |
| 13,self_attn.k_proj,0.0000043581,0.05000,3.889 | |
| 13,self_attn.q_proj,0.0000159966,0.05000,4.022 | |
| 13,self_attn.v_proj,0.0000038494,0.05000,4.046 | |
| 13,self_attn.o_proj,0.0000029048,0.05000,1.381 | |
| 13,mlp.up_proj,0.0001091306,0.05000,1.877 | |
| 13,mlp.gate_proj,0.0001162593,0.05000,1.907 | |
| 13,mlp.down_proj,0.0000108037,0.05000,3.634 | |
| 14,self_attn.q_proj,0.0000260713,0.05000,3.995 | |
| 14,self_attn.k_proj,0.0000068275,0.05000,4.027 | |
| 14,self_attn.v_proj,0.0000067319,0.05000,4.065 | |
| 14,self_attn.o_proj,0.0000044881,0.05000,1.401 | |
| 14,mlp.up_proj,0.0001112787,0.05000,1.968 | |
| 14,mlp.gate_proj,0.0001179207,0.05000,1.966 | |
| 14,mlp.down_proj,0.0000101188,0.05000,3.620 | |
| 15,self_attn.v_proj,0.0000062567,0.05000,4.044 | |
| 15,self_attn.q_proj,0.0000269484,0.05000,4.059 | |
| 15,self_attn.k_proj,0.0000071068,0.05000,4.097 | |
| 15,self_attn.o_proj,0.0000038186,0.05000,1.407 | |
| 15,mlp.gate_proj,0.0001098568,0.05000,1.904 | |
| 15,mlp.up_proj,0.0001071991,0.05000,1.925 | |
| 15,mlp.down_proj,0.0000096914,0.05000,3.681 | |
| 16,self_attn.v_proj,0.0000123046,0.05000,3.948 | |
| 16,self_attn.k_proj,0.0000116075,0.05000,4.038 | |
| 16,self_attn.q_proj,0.0000466369,0.05000,4.046 | |
| 16,self_attn.o_proj,0.0000044731,0.05000,1.404 | |
| 16,mlp.up_proj,0.0001113207,0.05000,1.917 | |
| 16,mlp.gate_proj,0.0001211038,0.05000,1.924 | |
| 16,mlp.down_proj,0.0000872585,0.05000,3.648 | |
| 17,self_attn.k_proj,0.0000090312,0.05000,4.024 | |
| 17,self_attn.v_proj,0.0000093076,0.05000,4.073 | |
| 17,self_attn.q_proj,0.0000403532,0.05000,4.105 | |
| 17,self_attn.o_proj,0.0000043949,0.05000,1.497 | |
| 17,mlp.up_proj,0.0000990942,0.05000,1.850 | |
| 17,mlp.gate_proj,0.0001068454,0.05000,1.870 | |
| 17,mlp.down_proj,0.0000089522,0.05000,3.610 | |
| 18,self_attn.k_proj,0.0000108056,0.05000,4.046 | |
| 18,self_attn.q_proj,0.0000436995,0.05000,4.096 | |
| 18,self_attn.v_proj,0.0000111328,0.05000,4.135 | |
| 18,self_attn.o_proj,0.0000057354,0.05000,1.378 | |
| 18,mlp.up_proj,0.0001083605,0.05000,1.924 | |
| 18,mlp.gate_proj,0.0001144615,0.05000,1.945 | |
| 18,mlp.down_proj,0.0000113469,0.05000,3.680 | |
| 19,self_attn.v_proj,0.0000194956,0.05000,4.021 | |
| 19,self_attn.k_proj,0.0000207048,0.05000,4.103 | |
| 19,self_attn.q_proj,0.0000867269,0.05000,4.125 | |
| 19,self_attn.o_proj,0.0000082300,0.05000,1.408 | |
| 19,mlp.up_proj,0.0001204733,0.05000,1.885 | |
| 19,mlp.gate_proj,0.0001256971,0.05000,1.883 | |
| 19,mlp.down_proj,0.0000149575,0.05000,3.666 | |
| 20,self_attn.v_proj,0.0000181873,0.05000,3.973 | |
| 20,self_attn.k_proj,0.0000184160,0.05000,4.048 | |
| 20,self_attn.q_proj,0.0000832375,0.05000,4.065 | |
| 20,self_attn.o_proj,0.0000077851,0.05000,1.456 | |
| 20,mlp.gate_proj,0.0001373728,0.05000,1.933 | |
| 20,mlp.up_proj,0.0001342345,0.05000,1.953 | |
| 20,mlp.down_proj,0.0000174853,0.05000,3.660 | |
| 21,self_attn.v_proj,0.0000244685,0.05000,3.980 | |
| 21,self_attn.k_proj,0.0000240417,0.05000,4.051 | |
| 21,self_attn.q_proj,0.0001041426,0.05000,4.056 | |
| 21,self_attn.o_proj,0.0000098543,0.05000,1.411 | |
| 21,mlp.up_proj,0.0001608356,0.05000,1.880 | |
| 21,mlp.gate_proj,0.0001651234,0.05000,1.899 | |
| 21,mlp.down_proj,0.0000237775,0.05000,3.748 | |
| 22,self_attn.q_proj,0.0002165873,0.05000,4.104 | |
| 22,self_attn.v_proj,0.0000505528,0.05000,4.155 | |
| 22,self_attn.k_proj,0.0000500092,0.05000,4.182 | |
| 22,self_attn.o_proj,0.0000204702,0.05000,1.381 | |
| 22,mlp.up_proj,0.0001821093,0.05000,1.927 | |
| 22,mlp.gate_proj,0.0001975882,0.05000,1.943 | |
| 22,mlp.down_proj,0.0000375069,0.05000,3.653 | |
| 23,self_attn.q_proj,0.0002052678,0.05000,4.067 | |
| 23,self_attn.k_proj,0.0000481465,0.05000,4.102 | |
| 23,self_attn.v_proj,0.0000507366,0.05000,4.123 | |
| 23,self_attn.o_proj,0.0000210706,0.05000,1.465 | |
| 23,mlp.up_proj,0.0002144172,0.05000,1.990 | |
| 23,mlp.gate_proj,0.0002453108,0.05000,2.009 | |
| 23,mlp.down_proj,0.0000515478,0.05000,3.654 | |
| 24,self_attn.k_proj,0.0000796690,0.05000,4.027 | |
| 24,self_attn.v_proj,0.0000970791,0.05000,4.076 | |
| 24,self_attn.q_proj,0.0003320909,0.05000,4.100 | |
| 24,self_attn.o_proj,0.0000288356,0.05000,1.429 | |
| 24,mlp.gate_proj,0.0002778602,0.05000,1.944 | |
| 24,mlp.up_proj,0.0002385447,0.05000,1.965 | |
| 24,mlp.down_proj,0.0000658266,0.05000,3.663 | |
| 25,self_attn.q_proj,0.0002312621,0.05000,3.981 | |
| 25,self_attn.v_proj,0.0000614569,0.05000,4.023 | |
| 25,self_attn.k_proj,0.0000580671,0.05000,4.076 | |
| 25,self_attn.o_proj,0.0000149890,0.05000,1.393 | |
| 25,mlp.up_proj,0.0002772560,0.05000,1.910 | |
| 25,mlp.gate_proj,0.0003277600,0.05000,1.930 | |
| 25,mlp.down_proj,0.0000812657,0.05000,3.608 | |
| 26,self_attn.k_proj,0.0000911344,0.05000,3.968 | |
| 26,self_attn.v_proj,0.0001004724,0.05000,3.987 | |
| 26,self_attn.q_proj,0.0003837158,0.05000,4.036 | |
| 26,self_attn.o_proj,0.0000154729,0.05000,1.449 | |
| 26,mlp.up_proj,0.0003514904,0.05000,1.952 | |
| 26,mlp.gate_proj,0.0004019208,0.05000,1.982 | |
| 26,mlp.down_proj,0.0001047065,0.05000,3.641 | |
| 27,self_attn.k_proj,0.0001113520,0.05000,4.056 | |
| 27,self_attn.v_proj,0.0001370804,0.05000,4.102 | |
| 27,self_attn.q_proj,0.0004825537,0.05000,4.124 | |
| 27,self_attn.o_proj,0.0000248997,0.05000,1.391 | |
| 27,mlp.gate_proj,0.0004625157,0.05000,2.020 | |
| 27,mlp.up_proj,0.0004192917,0.05000,2.036 | |
| 27,mlp.down_proj,0.0001594466,0.05000,3.677 | |
| 28,self_attn.k_proj,0.0001383105,0.05000,4.090 | |
| 28,self_attn.q_proj,0.0005395452,0.05000,4.121 | |
| 28,self_attn.v_proj,0.0001479625,0.05000,4.168 | |
| 28,self_attn.o_proj,0.0000397905,0.05000,1.428 | |
| 28,mlp.up_proj,0.0005127904,0.05000,1.930 | |
| 28,mlp.gate_proj,0.0005403935,0.05000,1.953 | |
| 28,mlp.down_proj,0.0002484342,0.05000,3.655 | |
| 29,self_attn.q_proj,0.0013496861,0.05000,3.944 | |
| 29,self_attn.v_proj,0.0004323819,0.05000,4.024 | |
| 29,self_attn.k_proj,0.0003228422,0.05000,4.047 | |
| 29,self_attn.o_proj,0.0000484061,0.05000,1.422 | |
| 29,mlp.gate_proj,0.0006483419,0.05000,1.925 | |
| 29,mlp.up_proj,0.0006496695,0.05000,1.925 | |
| 29,mlp.down_proj,0.0002600003,0.05000,3.624 | |
| 30,self_attn.q_proj,0.0015428930,0.05000,3.853 | |
| 30,self_attn.v_proj,0.0004768748,0.05000,3.943 | |
| 30,self_attn.k_proj,0.0003945851,0.05000,3.970 | |
| 30,self_attn.o_proj,0.0000822501,0.05000,1.387 | |
| 30,mlp.gate_proj,0.0006952679,0.05000,1.906 | |
| 30,mlp.up_proj,0.0007226467,0.05000,1.926 | |
| 30,mlp.down_proj,0.0003748471,0.05000,3.639 | |
| 31,self_attn.k_proj,0.0005569797,0.05000,4.087 | |
| 31,self_attn.q_proj,0.0020152309,0.05000,4.146 | |
| 31,self_attn.v_proj,0.0007386387,0.05000,4.160 | |
| 31,self_attn.o_proj,0.0000730647,0.05000,1.423 | |
| 31,mlp.gate_proj,0.0007217523,0.05000,2.030 | |
| 31,mlp.up_proj,0.0007903871,0.05000,2.049 | |
| 31,mlp.down_proj,0.0005056261,0.05000,3.588 | |
| 32,self_attn.v_proj,0.0011295374,0.05000,3.986 | |
| 32,self_attn.k_proj,0.0007691019,0.05000,4.011 | |
| 32,self_attn.q_proj,0.0030725903,0.05000,4.044 | |
| 32,self_attn.o_proj,0.0001355138,0.05000,1.374 | |
| 32,mlp.up_proj,0.0008563334,0.05000,1.920 | |
| 32,mlp.gate_proj,0.0007635706,0.05000,1.950 | |
| 32,mlp.down_proj,0.0006399382,0.05000,3.592 | |
| 33,self_attn.q_proj,0.0068798436,0.05000,3.950 | |
| 33,self_attn.k_proj,0.0014373954,0.05000,4.008 | |
| 33,self_attn.v_proj,0.0026777900,0.05000,4.049 | |
| 33,self_attn.o_proj,0.0001293710,0.05000,1.427 | |
| 33,mlp.gate_proj,0.0008459430,0.05000,1.951 | |
| 33,mlp.up_proj,0.0009587818,0.05000,1.976 | |
| 33,mlp.down_proj,0.0008440783,0.05000,3.667 | |
| 34,self_attn.k_proj,0.0012213234,0.05000,4.055 | |
| 34,self_attn.v_proj,0.0019338977,0.05000,4.098 | |
| 34,self_attn.q_proj,0.0053902740,0.05000,4.110 | |
| 34,self_attn.o_proj,0.0003464893,0.05000,1.511 | |
| 34,mlp.up_proj,0.0010697745,0.05000,2.085 | |
| 34,mlp.gate_proj,0.0010178187,0.05000,2.113 | |
| 34,mlp.down_proj,0.0010601362,0.05000,3.797 | |
| 35,self_attn.v_proj,0.0008049977,0.05000,4.341 | |
| 35,self_attn.q_proj,0.0024685818,0.05000,4.384 | |
| 35,self_attn.k_proj,0.0006466726,0.05000,4.404 | |
| 35,self_attn.o_proj,0.0003548001,0.05000,1.465 | |
| 35,mlp.up_proj,0.0016807885,0.05000,2.054 | |
| 35,mlp.gate_proj,0.0016346245,0.05000,2.080 | |
| 35,mlp.down_proj,0.0016113244,0.05000,3.722 | |