| layer,module,loss,samples,damp,time | |
| 0,self_attn.q_proj,0.0000004658,0.05000,5.010 | |
| 0,self_attn.v_proj,0.0000001105,0.05000,5.075 | |
| 0,self_attn.k_proj,0.0000001265,0.05000,5.110 | |
| 0,self_attn.o_proj,0.0000002945,0.05000,1.782 | |
| 0,mlp.gate_proj,0.0000289746,0.05000,2.652 | |
| 0,mlp.up_proj,0.0000252975,0.05000,2.701 | |
| 0,mlp.down_proj,0.0000037377,0.05000,4.487 | |
| 1,self_attn.v_proj,0.0000002128,0.05000,5.102 | |
| 1,self_attn.k_proj,0.0000002043,0.05000,5.172 | |
| 1,self_attn.q_proj,0.0000007588,0.05000,5.212 | |
| 1,self_attn.o_proj,0.0000003918,0.05000,1.697 | |
| 1,mlp.up_proj,0.0004287361,0.05000,2.695 | |
| 1,mlp.gate_proj,0.0009483043,0.05000,2.728 | |
| 1,mlp.down_proj,0.0000043074,0.05000,4.332 | |
| 2,self_attn.q_proj,0.0000021343,0.05000,5.001 | |
| 2,self_attn.v_proj,0.0000005962,0.05000,5.076 | |
| 2,self_attn.k_proj,0.0000006101,0.05000,5.116 | |
| 2,self_attn.o_proj,0.0000005367,0.05000,1.751 | |
| 2,mlp.gate_proj,0.0015602610,0.05000,2.559 | |
| 2,mlp.up_proj,0.0012285131,0.05000,2.596 | |
| 2,mlp.down_proj,0.0000029970,0.05000,4.283 | |
| 3,self_attn.q_proj,0.0000041543,0.05000,5.029 | |
| 3,self_attn.v_proj,0.0000011004,0.05000,5.103 | |
| 3,self_attn.k_proj,0.0000010634,0.05000,5.118 | |
| 3,self_attn.o_proj,0.0000007820,0.05000,1.644 | |
| 3,mlp.up_proj,0.0008446519,0.05000,2.626 | |
| 3,mlp.gate_proj,0.0016393722,0.05000,2.648 | |
| 3,mlp.down_proj,0.0000086119,0.05000,4.336 | |
| 4,self_attn.k_proj,0.0000021310,0.05000,5.020 | |
| 4,self_attn.q_proj,0.0000078954,0.05000,5.044 | |
| 4,self_attn.v_proj,0.0000021595,0.05000,5.111 | |
| 4,self_attn.o_proj,0.0000012900,0.05000,1.737 | |
| 4,mlp.up_proj,0.0005500840,0.05000,2.736 | |
| 4,mlp.gate_proj,0.0013762149,0.05000,2.738 | |
| 4,mlp.down_proj,0.0000128145,0.05000,4.359 | |
| 5,self_attn.q_proj,0.0000086130,0.05000,4.662 | |
| 5,self_attn.v_proj,0.0000023674,0.05000,4.759 | |
| 5,self_attn.k_proj,0.0000021791,0.05000,4.780 | |
| 5,self_attn.o_proj,0.0000023793,0.05000,1.532 | |
| 5,mlp.up_proj,0.0002447622,0.05000,2.342 | |
| 5,mlp.gate_proj,0.0005030251,0.05000,2.389 | |
| 5,mlp.down_proj,0.0000185442,0.05000,3.948 | |
| 6,self_attn.k_proj,0.0000044010,0.05000,4.673 | |
| 6,self_attn.v_proj,0.0000050325,0.05000,4.750 | |
| 6,self_attn.q_proj,0.0000186530,0.05000,4.770 | |
| 6,self_attn.o_proj,0.0000058681,0.05000,1.525 | |
| 6,mlp.gate_proj,0.0005370448,0.05000,2.268 | |
| 6,mlp.up_proj,0.0003059179,0.05000,2.269 | |
| 6,mlp.down_proj,0.0009229853,0.05000,4.023 | |
| 7,self_attn.q_proj,0.0000376636,0.05000,4.710 | |
| 7,self_attn.k_proj,0.0000101153,0.05000,4.744 | |
| 7,self_attn.v_proj,0.0000104857,0.05000,4.764 | |
| 7,self_attn.o_proj,0.0000067601,0.05000,1.529 | |
| 7,mlp.up_proj,0.0003681960,0.05000,2.167 | |
| 7,mlp.gate_proj,0.0006117455,0.05000,2.191 | |
| 7,mlp.down_proj,0.0000305151,0.05000,3.961 | |
| 8,self_attn.k_proj,0.0000149234,0.05000,4.718 | |
| 8,self_attn.v_proj,0.0000164171,0.05000,4.838 | |
| 8,self_attn.q_proj,0.0000576840,0.05000,4.853 | |
| 8,self_attn.o_proj,0.0000096945,0.05000,1.528 | |
| 8,mlp.gate_proj,0.0004354699,0.05000,2.192 | |
| 8,mlp.up_proj,0.0003553537,0.05000,2.220 | |
| 8,mlp.down_proj,0.0000482470,0.05000,3.933 | |
| 9,self_attn.q_proj,0.0000708480,0.05000,4.591 | |
| 9,self_attn.v_proj,0.0000198840,0.05000,4.660 | |
| 9,self_attn.k_proj,0.0000197770,0.05000,4.679 | |
| 9,self_attn.o_proj,0.0000112224,0.05000,1.536 | |
| 9,mlp.up_proj,0.0004471922,0.05000,2.262 | |
| 9,mlp.gate_proj,0.0007011541,0.05000,2.280 | |
| 9,mlp.down_proj,0.0000475860,0.05000,4.021 | |
| 10,self_attn.v_proj,0.0000316658,0.05000,4.767 | |
| 10,self_attn.k_proj,0.0000271843,0.05000,4.818 | |
| 10,self_attn.q_proj,0.0001055211,0.05000,4.854 | |
| 10,self_attn.o_proj,0.0000176322,0.05000,1.520 | |
| 10,mlp.up_proj,0.0003787521,0.05000,2.269 | |
| 10,mlp.gate_proj,0.0005428473,0.05000,2.288 | |
| 10,mlp.down_proj,0.0000400349,0.05000,4.012 | |
| 11,self_attn.v_proj,0.0000144231,0.05000,4.735 | |
| 11,self_attn.k_proj,0.0000135963,0.05000,4.801 | |
| 11,self_attn.q_proj,0.0000505602,0.05000,4.845 | |
| 11,self_attn.o_proj,0.0000114296,0.05000,1.576 | |
| 11,mlp.up_proj,0.0003493419,0.05000,2.223 | |
| 11,mlp.gate_proj,0.0004413690,0.05000,2.253 | |
| 11,mlp.down_proj,0.0000366131,0.05000,4.017 | |
| 12,self_attn.k_proj,0.0000153139,0.05000,4.535 | |
| 12,self_attn.v_proj,0.0000167666,0.05000,4.586 | |
| 12,self_attn.q_proj,0.0000584962,0.05000,4.616 | |
| 12,self_attn.o_proj,0.0000125013,0.05000,1.622 | |
| 12,mlp.gate_proj,0.0003892739,0.05000,2.209 | |
| 12,mlp.up_proj,0.0003392469,0.05000,2.254 | |
| 12,mlp.down_proj,0.0000370598,0.05000,4.058 | |
| 13,self_attn.k_proj,0.0000116586,0.05000,4.696 | |
| 13,self_attn.q_proj,0.0000443353,0.05000,4.778 | |
| 13,self_attn.v_proj,0.0000114623,0.05000,4.786 | |
| 13,self_attn.o_proj,0.0000097119,0.05000,1.567 | |
| 13,mlp.gate_proj,0.0003625256,0.05000,2.177 | |
| 13,mlp.up_proj,0.0003459045,0.05000,2.200 | |
| 13,mlp.down_proj,0.0000401894,0.05000,3.992 | |
| 14,self_attn.k_proj,0.0000190907,0.05000,4.754 | |
| 14,self_attn.v_proj,0.0000204954,0.05000,4.793 | |
| 14,self_attn.q_proj,0.0000746915,0.05000,4.818 | |
| 14,self_attn.o_proj,0.0000171502,0.05000,1.543 | |
| 14,mlp.up_proj,0.0003421074,0.05000,2.281 | |
| 14,mlp.gate_proj,0.0003563725,0.05000,2.298 | |
| 14,mlp.down_proj,0.0000358082,0.05000,4.040 | |
| 15,self_attn.v_proj,0.0000185207,0.05000,4.646 | |
| 15,self_attn.k_proj,0.0000190547,0.05000,4.731 | |
| 15,self_attn.q_proj,0.0000740901,0.05000,4.774 | |
| 15,self_attn.o_proj,0.0000141826,0.05000,1.478 | |
| 15,mlp.gate_proj,0.0003278327,0.05000,2.248 | |
| 15,mlp.up_proj,0.0003265059,0.05000,2.267 | |
| 15,mlp.down_proj,0.0000344754,0.05000,4.042 | |
| 16,self_attn.k_proj,0.0000315280,0.05000,4.783 | |
| 16,self_attn.v_proj,0.0000365092,0.05000,4.807 | |
| 16,self_attn.q_proj,0.0001286270,0.05000,4.856 | |
| 16,self_attn.o_proj,0.0000152998,0.05000,1.529 | |
| 16,mlp.up_proj,0.0003464136,0.05000,2.168 | |
| 16,mlp.gate_proj,0.0003686556,0.05000,2.192 | |
| 16,mlp.down_proj,0.0003761285,0.05000,4.012 | |
| 17,self_attn.v_proj,0.0000274548,0.05000,4.671 | |
| 17,self_attn.q_proj,0.0001093356,0.05000,4.719 | |
| 17,self_attn.k_proj,0.0000240166,0.05000,4.734 | |
| 17,self_attn.o_proj,0.0000160825,0.05000,1.532 | |
| 17,mlp.up_proj,0.0003082692,0.05000,2.155 | |
| 17,mlp.gate_proj,0.0003231052,0.05000,2.188 | |
| 17,mlp.down_proj,0.0000337791,0.05000,4.057 | |
| 18,self_attn.v_proj,0.0000330313,0.05000,4.643 | |
| 18,self_attn.q_proj,0.0001190752,0.05000,4.730 | |
| 18,self_attn.k_proj,0.0000287658,0.05000,4.754 | |
| 18,self_attn.o_proj,0.0000195600,0.05000,1.516 | |
| 18,mlp.up_proj,0.0003365247,0.05000,2.106 | |
| 18,mlp.gate_proj,0.0003462859,0.05000,2.140 | |
| 18,mlp.down_proj,0.0000416005,0.05000,3.978 | |
| 19,self_attn.q_proj,0.0002362754,0.05000,4.728 | |
| 19,self_attn.v_proj,0.0000574272,0.05000,4.794 | |
| 19,self_attn.k_proj,0.0000544455,0.05000,4.836 | |
| 19,self_attn.o_proj,0.0000249074,0.05000,1.527 | |
| 19,mlp.up_proj,0.0003698838,0.05000,2.291 | |
| 19,mlp.gate_proj,0.0003773468,0.05000,2.316 | |
| 19,mlp.down_proj,0.0000530812,0.05000,3.979 | |
| 20,self_attn.q_proj,0.0002188809,0.05000,4.637 | |
| 20,self_attn.k_proj,0.0000471045,0.05000,4.642 | |
| 20,self_attn.v_proj,0.0000521005,0.05000,4.717 | |
| 20,self_attn.o_proj,0.0000248449,0.05000,1.551 | |
| 20,mlp.up_proj,0.0004029941,0.05000,2.137 | |
| 20,mlp.gate_proj,0.0004013012,0.05000,2.158 | |
| 20,mlp.down_proj,0.0000585520,0.05000,4.006 | |
| 21,self_attn.v_proj,0.0000706983,0.05000,4.767 | |
| 21,self_attn.k_proj,0.0000634547,0.05000,4.821 | |
| 21,self_attn.q_proj,0.0002828724,0.05000,4.833 | |
| 21,self_attn.o_proj,0.0000280543,0.05000,1.523 | |
| 21,mlp.up_proj,0.0004820233,0.05000,2.173 | |
| 21,mlp.gate_proj,0.0004849778,0.05000,2.193 | |
| 21,mlp.down_proj,0.0000768389,0.05000,3.981 | |
| 22,self_attn.v_proj,0.0001434762,0.05000,4.732 | |
| 22,self_attn.k_proj,0.0001253538,0.05000,4.819 | |
| 22,self_attn.q_proj,0.0005631076,0.05000,4.846 | |
| 22,self_attn.o_proj,0.0000553381,0.05000,1.557 | |
| 22,mlp.up_proj,0.0005509696,0.05000,2.289 | |
| 22,mlp.gate_proj,0.0005793622,0.05000,2.321 | |
| 22,mlp.down_proj,0.0001229682,0.05000,4.013 | |
| 23,self_attn.k_proj,0.0001267572,0.05000,4.744 | |
| 23,self_attn.q_proj,0.0005634705,0.05000,4.804 | |
| 23,self_attn.v_proj,0.0001490576,0.05000,4.816 | |
| 23,self_attn.o_proj,0.0000699550,0.05000,1.576 | |
| 23,mlp.up_proj,0.0006518718,0.05000,2.161 | |
| 23,mlp.gate_proj,0.0007231190,0.05000,2.181 | |
| 23,mlp.down_proj,0.0001715308,0.05000,4.017 | |
| 24,self_attn.k_proj,0.0002084854,0.05000,4.662 | |
| 24,self_attn.q_proj,0.0009156117,0.05000,4.669 | |
| 24,self_attn.v_proj,0.0002740377,0.05000,4.713 | |
| 24,self_attn.o_proj,0.0000720820,0.05000,1.559 | |
| 24,mlp.up_proj,0.0007225850,0.05000,2.197 | |
| 24,mlp.gate_proj,0.0008156972,0.05000,2.225 | |
| 24,mlp.down_proj,0.0002084210,0.05000,3.929 | |
| 25,self_attn.k_proj,0.0001581930,0.05000,4.615 | |
| 25,self_attn.v_proj,0.0001814060,0.05000,4.679 | |
| 25,self_attn.q_proj,0.0006456851,0.05000,4.706 | |
| 25,self_attn.o_proj,0.0000473329,0.05000,1.560 | |
| 25,mlp.up_proj,0.0008242684,0.05000,2.142 | |
| 25,mlp.gate_proj,0.0009428286,0.05000,2.172 | |
| 25,mlp.down_proj,0.0002486016,0.05000,3.990 | |
| 26,self_attn.k_proj,0.0002404927,0.05000,4.621 | |
| 26,self_attn.v_proj,0.0002888501,0.05000,4.642 | |
| 26,self_attn.q_proj,0.0010402773,0.05000,4.699 | |
| 26,self_attn.o_proj,0.0000435772,0.05000,1.576 | |
| 26,mlp.gate_proj,0.0011219122,0.05000,2.293 | |
| 26,mlp.up_proj,0.0010119856,0.05000,2.308 | |
| 26,mlp.down_proj,0.0002938876,0.05000,3.967 | |
| 27,self_attn.q_proj,0.0012800641,0.05000,4.697 | |
| 27,self_attn.v_proj,0.0003850639,0.05000,4.761 | |
| 27,self_attn.k_proj,0.0002874724,0.05000,4.777 | |
| 27,self_attn.o_proj,0.0000617831,0.05000,1.501 | |
| 27,mlp.gate_proj,0.0012472775,0.05000,2.290 | |
| 27,mlp.up_proj,0.0011642600,0.05000,2.332 | |
| 27,mlp.down_proj,0.0004163454,0.05000,3.981 | |
| 28,self_attn.k_proj,0.0003459948,0.05000,4.645 | |
| 28,self_attn.q_proj,0.0013993718,0.05000,4.743 | |
| 28,self_attn.v_proj,0.0004060790,0.05000,4.769 | |
| 28,self_attn.o_proj,0.0000979789,0.05000,1.566 | |
| 28,mlp.up_proj,0.0013726200,0.05000,2.287 | |
| 28,mlp.gate_proj,0.0014115914,0.05000,2.297 | |
| 28,mlp.down_proj,0.0006263638,0.05000,4.001 | |
| 29,self_attn.k_proj,0.0007600260,0.05000,4.717 | |
| 29,self_attn.v_proj,0.0010328331,0.05000,4.782 | |
| 29,self_attn.q_proj,0.0033356990,0.05000,4.817 | |
| 29,self_attn.o_proj,0.0000782954,0.05000,1.573 | |
| 29,mlp.up_proj,0.0016839991,0.05000,2.209 | |
| 29,mlp.gate_proj,0.0016476954,0.05000,2.240 | |
| 29,mlp.down_proj,0.0006718912,0.05000,4.071 | |
| 30,self_attn.k_proj,0.0009500644,0.05000,4.648 | |
| 30,self_attn.q_proj,0.0037828581,0.05000,4.691 | |
| 30,self_attn.v_proj,0.0012155685,0.05000,4.745 | |
| 30,self_attn.o_proj,0.0001820555,0.05000,1.580 | |
| 30,mlp.gate_proj,0.0017180550,0.05000,2.131 | |
| 30,mlp.up_proj,0.0018161121,0.05000,2.136 | |
| 30,mlp.down_proj,0.0009175093,0.05000,4.057 | |
| 31,self_attn.q_proj,0.0047508813,0.05000,4.746 | |
| 31,self_attn.k_proj,0.0012552110,0.05000,4.801 | |
| 31,self_attn.v_proj,0.0017514399,0.05000,4.843 | |
| 31,self_attn.o_proj,0.0001683568,0.05000,1.558 | |
| 31,mlp.gate_proj,0.0016933631,0.05000,2.218 | |
| 31,mlp.up_proj,0.0018819273,0.05000,2.249 | |
| 31,mlp.down_proj,0.0011198272,0.05000,4.011 | |
| 32,self_attn.q_proj,0.0068592699,0.05000,4.781 | |
| 32,self_attn.v_proj,0.0025475807,0.05000,4.837 | |
| 32,self_attn.k_proj,0.0016688757,0.05000,4.866 | |
| 32,self_attn.o_proj,0.0001884213,0.05000,1.525 | |
| 32,mlp.up_proj,0.0019701249,0.05000,2.197 | |
| 32,mlp.gate_proj,0.0017300433,0.05000,2.227 | |
| 32,mlp.down_proj,0.0013363534,0.05000,4.113 | |
| 33,self_attn.q_proj,0.0142963547,0.05000,4.643 | |
| 33,self_attn.k_proj,0.0029161619,0.05000,4.714 | |
| 33,self_attn.v_proj,0.0055924580,0.05000,4.730 | |
| 33,self_attn.o_proj,0.0002554161,0.05000,1.524 | |
| 33,mlp.up_proj,0.0021203640,0.05000,2.182 | |
| 33,mlp.gate_proj,0.0018360139,0.05000,2.221 | |
| 33,mlp.down_proj,0.0017152019,0.05000,4.001 | |
| 34,self_attn.q_proj,0.0111581404,0.05000,4.838 | |
| 34,self_attn.v_proj,0.0041854484,0.05000,4.842 | |
| 34,self_attn.k_proj,0.0024777448,0.05000,4.885 | |
| 34,self_attn.o_proj,0.0005260598,0.05000,1.526 | |
| 34,mlp.gate_proj,0.0021799901,0.05000,2.246 | |
| 34,mlp.up_proj,0.0023547660,0.05000,2.265 | |
| 34,mlp.down_proj,0.0020657629,0.05000,4.042 | |
| 35,self_attn.q_proj,0.0052947048,0.05000,4.801 | |
| 35,self_attn.k_proj,0.0013647166,0.05000,4.871 | |
| 35,self_attn.v_proj,0.0018166249,0.05000,4.910 | |
| 35,self_attn.o_proj,0.0007476909,0.05000,1.543 | |
| 35,mlp.up_proj,0.0045160744,0.05000,2.199 | |
| 35,mlp.gate_proj,0.0039883456,0.05000,2.221 | |
| 35,mlp.down_proj,0.0036397912,0.05000,4.071 | |