| layer,module,loss,samples,damp,time | |
| 0,self_attn.v_proj,0.0000000055,0.05000,4.851 | |
| 0,self_attn.k_proj,0.0000000067,0.05000,4.868 | |
| 0,self_attn.q_proj,0.0000000245,0.05000,4.875 | |
| 0,self_attn.o_proj,0.0000001023,0.05000,1.463 | |
| 0,mlp.gate_proj,0.0000014832,0.05000,3.275 | |
| 0,mlp.up_proj,0.0000012088,0.05000,3.334 | |
| 0,mlp.down_proj,0.0000007021,0.05000,5.384 | |
| 1,self_attn.q_proj,0.0000000516,0.05000,4.584 | |
| 1,self_attn.v_proj,0.0000000140,0.05000,4.683 | |
| 1,self_attn.k_proj,0.0000000128,0.05000,4.723 | |
| 1,self_attn.o_proj,0.0000001024,0.05000,1.334 | |
| 1,mlp.up_proj,0.0000235446,0.05000,3.642 | |
| 1,mlp.gate_proj,0.0000679487,0.05000,3.677 | |
| 1,mlp.down_proj,0.0000010552,0.05000,5.334 | |
| 2,self_attn.k_proj,0.0000000361,0.05000,4.120 | |
| 2,self_attn.v_proj,0.0000000386,0.05000,4.189 | |
| 2,self_attn.q_proj,0.0000001411,0.05000,4.244 | |
| 2,self_attn.o_proj,0.0000002282,0.05000,1.332 | |
| 2,mlp.gate_proj,0.0000977210,0.05000,3.152 | |
| 2,mlp.up_proj,0.0000492783,0.05000,3.153 | |
| 2,mlp.down_proj,0.0000008231,0.05000,5.410 | |
| 3,self_attn.q_proj,0.0000003171,0.05000,4.748 | |
| 3,self_attn.k_proj,0.0000000801,0.05000,4.758 | |
| 3,self_attn.v_proj,0.0000000896,0.05000,4.779 | |
| 3,self_attn.o_proj,0.0000002193,0.05000,1.414 | |
| 3,mlp.gate_proj,0.0001480885,0.05000,3.153 | |
| 3,mlp.up_proj,0.0000745903,0.05000,3.224 | |
| 3,mlp.down_proj,0.0000013200,0.05000,5.582 | |
| 4,self_attn.v_proj,0.0000001349,0.05000,5.075 | |
| 4,self_attn.q_proj,0.0000004968,0.05000,5.101 | |
| 4,self_attn.k_proj,0.0000001217,0.05000,5.105 | |
| 4,self_attn.o_proj,0.0000003706,0.05000,1.499 | |
| 4,mlp.up_proj,0.0001282589,0.05000,3.353 | |
| 4,mlp.gate_proj,0.0002110812,0.05000,3.373 | |
| 4,mlp.down_proj,0.0000014401,0.05000,5.379 | |
| 5,self_attn.q_proj,0.0000005028,0.05000,4.952 | |
| 5,self_attn.v_proj,0.0000001383,0.05000,4.969 | |
| 5,self_attn.k_proj,0.0000001241,0.05000,4.978 | |
| 5,self_attn.o_proj,0.0000003825,0.05000,1.373 | |
| 5,mlp.up_proj,0.0000865598,0.05000,3.568 | |
| 5,mlp.gate_proj,0.0001970606,0.05000,3.592 | |
| 5,mlp.down_proj,0.0000036706,0.05000,5.314 | |
| 6,self_attn.v_proj,0.0000002469,0.05000,4.978 | |
| 6,self_attn.q_proj,0.0000008867,0.05000,5.061 | |
| 6,self_attn.k_proj,0.0000002098,0.05000,5.067 | |
| 6,self_attn.o_proj,0.0000005215,0.05000,1.377 | |
| 6,mlp.up_proj,0.0001740953,0.05000,2.960 | |
| 6,mlp.gate_proj,0.0003093244,0.05000,3.049 | |
| 6,mlp.down_proj,0.0000451711,0.05000,5.501 | |
| 7,self_attn.q_proj,0.0000032589,0.05000,4.941 | |
| 7,self_attn.k_proj,0.0000007494,0.05000,4.953 | |
| 7,self_attn.v_proj,0.0000008548,0.05000,4.969 | |
| 7,self_attn.o_proj,0.0000011350,0.05000,1.371 | |
| 7,mlp.gate_proj,0.0002542938,0.05000,3.356 | |
| 7,mlp.up_proj,0.0001292021,0.05000,3.370 | |
| 7,mlp.down_proj,0.0000053761,0.05000,5.684 | |
| 8,self_attn.v_proj,0.0000008264,0.05000,4.899 | |
| 8,self_attn.q_proj,0.0000030318,0.05000,4.924 | |
| 8,self_attn.k_proj,0.0000007100,0.05000,4.975 | |
| 8,self_attn.o_proj,0.0000013708,0.05000,1.409 | |
| 8,mlp.up_proj,0.0000463701,0.05000,3.645 | |
| 8,mlp.gate_proj,0.0000779276,0.05000,3.699 | |
| 8,mlp.down_proj,0.0000070508,0.05000,5.465 | |
| 9,self_attn.v_proj,0.0000006791,0.05000,5.006 | |
| 9,self_attn.q_proj,0.0000025509,0.05000,5.023 | |
| 9,self_attn.k_proj,0.0000006144,0.05000,5.032 | |
| 9,self_attn.o_proj,0.0000013103,0.05000,1.424 | |
| 9,mlp.gate_proj,0.0000426071,0.05000,3.332 | |
| 9,mlp.up_proj,0.0000384853,0.05000,3.368 | |
| 9,mlp.down_proj,0.0000094974,0.05000,5.585 | |
| 10,self_attn.k_proj,0.0000008871,0.05000,5.058 | |
| 10,self_attn.v_proj,0.0000010250,0.05000,5.070 | |
| 10,self_attn.q_proj,0.0000039080,0.05000,5.100 | |
| 10,self_attn.o_proj,0.0000023923,0.05000,1.394 | |
| 10,mlp.gate_proj,0.0000476698,0.05000,2.910 | |
| 10,mlp.up_proj,0.0000438941,0.05000,2.918 | |
| 10,mlp.down_proj,0.0000099568,0.05000,5.422 | |
| 11,self_attn.v_proj,0.0000016619,0.05000,5.104 | |
| 11,self_attn.q_proj,0.0000061273,0.05000,5.127 | |
| 11,self_attn.k_proj,0.0000014110,0.05000,5.151 | |
| 11,self_attn.o_proj,0.0000034863,0.05000,1.411 | |
| 11,mlp.up_proj,0.0000549275,0.05000,3.648 | |
| 11,mlp.gate_proj,0.0000592392,0.05000,3.697 | |
| 11,mlp.down_proj,0.0000136331,0.05000,5.316 | |
| 12,self_attn.k_proj,0.0000015385,0.05000,4.950 | |
| 12,self_attn.v_proj,0.0000017779,0.05000,4.997 | |
| 12,self_attn.q_proj,0.0000064274,0.05000,5.019 | |
| 12,self_attn.o_proj,0.0000026011,0.05000,1.423 | |
| 12,mlp.gate_proj,0.0000721614,0.05000,3.257 | |
| 12,mlp.up_proj,0.0000664742,0.05000,3.284 | |
| 12,mlp.down_proj,0.0000165714,0.05000,5.587 | |
| 13,self_attn.v_proj,0.0000035299,0.05000,4.510 | |
| 13,self_attn.q_proj,0.0000127921,0.05000,4.531 | |
| 13,self_attn.k_proj,0.0000029702,0.05000,4.564 | |
| 13,self_attn.o_proj,0.0000040227,0.05000,1.394 | |
| 13,mlp.up_proj,0.0000723846,0.05000,2.925 | |
| 13,mlp.gate_proj,0.0000838850,0.05000,2.940 | |
| 13,mlp.down_proj,0.0000187251,0.05000,5.496 | |
| 14,self_attn.v_proj,0.0000024263,0.05000,4.790 | |
| 14,self_attn.q_proj,0.0000085204,0.05000,4.817 | |
| 14,self_attn.k_proj,0.0000020249,0.05000,4.848 | |
| 14,self_attn.o_proj,0.0000040741,0.05000,1.353 | |
| 14,mlp.gate_proj,0.0000798361,0.05000,2.961 | |
| 14,mlp.up_proj,0.0000716579,0.05000,2.974 | |
| 14,mlp.down_proj,0.0000192025,0.05000,5.455 | |
| 15,self_attn.v_proj,0.0000024046,0.05000,4.509 | |
| 15,self_attn.k_proj,0.0000021407,0.05000,4.558 | |
| 15,self_attn.q_proj,0.0000092248,0.05000,4.569 | |
| 15,self_attn.o_proj,0.0000045808,0.05000,1.348 | |
| 15,mlp.up_proj,0.0000751190,0.05000,3.045 | |
| 15,mlp.gate_proj,0.0000782121,0.05000,3.079 | |
| 15,mlp.down_proj,0.0000209370,0.05000,5.471 | |
| 16,self_attn.q_proj,0.0000100945,0.05000,5.142 | |
| 16,self_attn.v_proj,0.0000027845,0.05000,5.142 | |
| 16,self_attn.k_proj,0.0000023761,0.05000,5.241 | |
| 16,self_attn.o_proj,0.0000062379,0.05000,1.380 | |
| 16,mlp.up_proj,0.0000748809,0.05000,2.870 | |
| 16,mlp.gate_proj,0.0000727667,0.05000,2.884 | |
| 16,mlp.down_proj,0.0000219730,0.05000,5.353 | |
| 17,self_attn.k_proj,0.0000028254,0.05000,4.584 | |
| 17,self_attn.q_proj,0.0000126655,0.05000,4.615 | |
| 17,self_attn.v_proj,0.0000032815,0.05000,4.642 | |
| 17,self_attn.o_proj,0.0000061028,0.05000,1.371 | |
| 17,mlp.gate_proj,0.0000801209,0.05000,3.312 | |
| 17,mlp.up_proj,0.0000837841,0.05000,3.340 | |
| 17,mlp.down_proj,0.0000245772,0.05000,5.519 | |
| 18,self_attn.v_proj,0.0000047651,0.05000,4.578 | |
| 18,self_attn.k_proj,0.0000041778,0.05000,4.587 | |
| 18,self_attn.q_proj,0.0000184517,0.05000,4.621 | |
| 18,self_attn.o_proj,0.0000063895,0.05000,1.366 | |
| 18,mlp.up_proj,0.0000915147,0.05000,3.025 | |
| 18,mlp.gate_proj,0.0000842209,0.05000,3.032 | |
| 18,mlp.down_proj,0.0000275443,0.05000,5.559 | |
| 19,self_attn.k_proj,0.0000049127,0.05000,4.855 | |
| 19,self_attn.v_proj,0.0000055556,0.05000,4.940 | |
| 19,self_attn.q_proj,0.0000216752,0.05000,4.962 | |
| 19,self_attn.o_proj,0.0000083279,0.05000,1.398 | |
| 19,mlp.up_proj,0.0001011631,0.05000,2.898 | |
| 19,mlp.gate_proj,0.0000937019,0.05000,2.930 | |
| 19,mlp.down_proj,0.0000327259,0.05000,5.399 | |
| 20,self_attn.v_proj,0.0000109982,0.05000,4.754 | |
| 20,self_attn.k_proj,0.0000086623,0.05000,4.805 | |
| 20,self_attn.q_proj,0.0000405637,0.05000,4.808 | |
| 20,self_attn.o_proj,0.0000105337,0.05000,1.392 | |
| 20,mlp.gate_proj,0.0000998617,0.05000,3.189 | |
| 20,mlp.up_proj,0.0001093770,0.05000,3.217 | |
| 20,mlp.down_proj,0.0000415528,0.05000,5.438 | |
| 21,self_attn.k_proj,0.0000117670,0.05000,4.439 | |
| 21,self_attn.q_proj,0.0000514724,0.05000,4.453 | |
| 21,self_attn.v_proj,0.0000140596,0.05000,4.516 | |
| 21,self_attn.o_proj,0.0000151081,0.05000,1.453 | |
| 21,mlp.gate_proj,0.0001088591,0.05000,2.926 | |
| 21,mlp.up_proj,0.0001167877,0.05000,2.938 | |
| 21,mlp.down_proj,0.0000477426,0.05000,5.551 | |
| 22,self_attn.v_proj,0.0000117180,0.05000,4.538 | |
| 22,self_attn.k_proj,0.0000094837,0.05000,4.573 | |
| 22,self_attn.q_proj,0.0000441959,0.05000,4.617 | |
| 22,self_attn.o_proj,0.0000190037,0.05000,1.410 | |
| 22,mlp.up_proj,0.0001366769,0.05000,3.587 | |
| 22,mlp.gate_proj,0.0001259929,0.05000,3.622 | |
| 22,mlp.down_proj,0.0000632469,0.05000,5.323 | |
| 23,self_attn.v_proj,0.0000206423,0.05000,5.066 | |
| 23,self_attn.q_proj,0.0000767674,0.05000,5.099 | |
| 23,self_attn.k_proj,0.0000157870,0.05000,5.111 | |
| 23,self_attn.o_proj,0.0000160790,0.05000,1.398 | |
| 23,mlp.up_proj,0.0001518091,0.05000,3.378 | |
| 23,mlp.gate_proj,0.0001383609,0.05000,3.395 | |
| 23,mlp.down_proj,0.0000856998,0.05000,5.398 | |
| 24,self_attn.v_proj,0.0000329967,0.05000,5.030 | |
| 24,self_attn.k_proj,0.0000237478,0.05000,5.054 | |
| 24,self_attn.q_proj,0.0001217306,0.05000,5.056 | |
| 24,self_attn.o_proj,0.0000220392,0.05000,1.360 | |
| 24,mlp.gate_proj,0.0001632924,0.05000,3.625 | |
| 24,mlp.up_proj,0.0001751961,0.05000,3.657 | |
| 24,mlp.down_proj,0.0000941419,0.05000,5.674 | |
| 25,self_attn.k_proj,0.0000201400,0.05000,4.789 | |
| 25,self_attn.q_proj,0.0000949212,0.05000,4.936 | |
| 25,self_attn.v_proj,0.0000253956,0.05000,4.953 | |
| 25,self_attn.o_proj,0.0000235002,0.05000,1.350 | |
| 25,mlp.up_proj,0.0002028768,0.05000,3.754 | |
| 25,mlp.gate_proj,0.0001922596,0.05000,3.767 | |
| 25,mlp.down_proj,0.0001267621,0.05000,5.595 | |
| 26,self_attn.k_proj,0.0000225750,0.05000,4.881 | |
| 26,self_attn.v_proj,0.0000295418,0.05000,4.921 | |
| 26,self_attn.q_proj,0.0001108223,0.05000,4.949 | |
| 26,self_attn.o_proj,0.0000265471,0.05000,1.385 | |
| 26,mlp.up_proj,0.0002334800,0.05000,3.388 | |
| 26,mlp.gate_proj,0.0002268478,0.05000,3.422 | |
| 26,mlp.down_proj,0.0001893703,0.05000,5.634 | |
| 27,self_attn.q_proj,0.0001638606,0.05000,5.063 | |
| 27,self_attn.v_proj,0.0000431965,0.05000,5.074 | |
| 27,self_attn.k_proj,0.0000341267,0.05000,5.090 | |
| 27,self_attn.o_proj,0.0000498931,0.05000,1.395 | |
| 27,mlp.up_proj,0.0002877421,0.05000,3.456 | |
| 27,mlp.gate_proj,0.0002798097,0.05000,3.483 | |
| 27,mlp.down_proj,0.0002758613,0.05000,5.413 | |
| 28,self_attn.q_proj,0.0002983511,0.05000,5.019 | |
| 28,self_attn.k_proj,0.0000565795,0.05000,5.021 | |
| 28,self_attn.v_proj,0.0000813798,0.05000,5.037 | |
| 28,self_attn.o_proj,0.0000459752,0.05000,1.567 | |
| 28,mlp.gate_proj,0.0003511385,0.05000,3.554 | |
| 28,mlp.up_proj,0.0003660403,0.05000,3.605 | |
| 28,mlp.down_proj,0.0003470357,0.05000,5.451 | |
| 29,self_attn.v_proj,0.0000988175,0.05000,4.698 | |
| 29,self_attn.q_proj,0.0003714425,0.05000,4.765 | |
| 29,self_attn.k_proj,0.0000788545,0.05000,4.812 | |
| 29,self_attn.o_proj,0.0000517854,0.05000,1.332 | |
| 29,mlp.up_proj,0.0004072312,0.05000,3.735 | |
| 29,mlp.gate_proj,0.0003876266,0.05000,3.765 | |
| 29,mlp.down_proj,0.0004905408,0.05000,5.507 | |
| 30,self_attn.q_proj,0.0005447937,0.05000,4.696 | |
| 30,self_attn.v_proj,0.0001534902,0.05000,4.769 | |
| 30,self_attn.k_proj,0.0001148308,0.05000,4.786 | |
| 30,self_attn.o_proj,0.0000775836,0.05000,1.395 | |
| 30,mlp.gate_proj,0.0004827204,0.05000,3.138 | |
| 30,mlp.up_proj,0.0005053384,0.05000,3.150 | |
| 30,mlp.down_proj,0.0005774746,0.05000,5.607 | |
| 31,self_attn.k_proj,0.0001142838,0.05000,4.573 | |
| 31,self_attn.v_proj,0.0001417302,0.05000,4.617 | |
| 31,self_attn.q_proj,0.0005521921,0.05000,4.641 | |
| 31,self_attn.o_proj,0.0000683407,0.05000,1.471 | |
| 31,mlp.gate_proj,0.0005245574,0.05000,3.371 | |
| 31,mlp.up_proj,0.0005602224,0.05000,3.391 | |
| 31,mlp.down_proj,0.0007208518,0.05000,5.466 | |
| 32,self_attn.k_proj,0.0002052982,0.05000,4.842 | |
| 32,self_attn.q_proj,0.0009621031,0.05000,4.916 | |
| 32,self_attn.v_proj,0.0002829138,0.05000,4.930 | |
| 32,self_attn.o_proj,0.0000929007,0.05000,1.420 | |
| 32,mlp.gate_proj,0.0005781620,0.05000,3.086 | |
| 32,mlp.up_proj,0.0006268456,0.05000,3.146 | |
| 32,mlp.down_proj,0.0007982043,0.05000,5.426 | |
| 33,self_attn.v_proj,0.0003873876,0.05000,4.661 | |
| 33,self_attn.q_proj,0.0013329935,0.05000,4.679 | |
| 33,self_attn.k_proj,0.0002684569,0.05000,4.699 | |
| 33,self_attn.o_proj,0.0001004743,0.05000,1.397 | |
| 33,mlp.gate_proj,0.0006186912,0.05000,3.072 | |
| 33,mlp.up_proj,0.0006873148,0.05000,3.080 | |
| 33,mlp.down_proj,0.0008831521,0.05000,5.923 | |
| 34,self_attn.k_proj,0.0004681831,0.05000,4.314 | |
| 34,self_attn.q_proj,0.0022188602,0.05000,4.330 | |
| 34,self_attn.v_proj,0.0006818156,0.05000,4.369 | |
| 34,self_attn.o_proj,0.0001390811,0.05000,1.400 | |
| 34,mlp.gate_proj,0.0006815079,0.05000,2.850 | |
| 34,mlp.up_proj,0.0007714760,0.05000,2.856 | |
| 34,mlp.down_proj,0.0010444433,0.05000,5.671 | |
| 35,self_attn.v_proj,0.0009464067,0.05000,4.743 | |
| 35,self_attn.q_proj,0.0028209405,0.05000,4.756 | |
| 35,self_attn.k_proj,0.0005987322,0.05000,4.770 | |
| 35,self_attn.o_proj,0.0001368956,0.05000,1.364 | |
| 35,mlp.up_proj,0.0008114978,0.05000,3.085 | |
| 35,mlp.gate_proj,0.0007047556,0.05000,3.087 | |
| 35,mlp.down_proj,0.0012339309,0.05000,5.406 | |
| 36,self_attn.k_proj,0.0005470300,0.05000,4.830 | |
| 36,self_attn.q_proj,0.0024462599,0.05000,4.882 | |
| 36,self_attn.v_proj,0.0008094013,0.05000,4.888 | |
| 36,self_attn.o_proj,0.0002597887,0.05000,1.454 | |
| 36,mlp.up_proj,0.0008480489,0.05000,3.371 | |
| 36,mlp.gate_proj,0.0007133146,0.05000,3.388 | |
| 36,mlp.down_proj,0.0015971187,0.05000,5.952 | |
| 37,self_attn.q_proj,0.0035302981,0.05000,4.597 | |
| 37,self_attn.k_proj,0.0007176153,0.05000,4.631 | |
| 37,self_attn.v_proj,0.0012357151,0.05000,4.644 | |
| 37,self_attn.o_proj,0.0002945466,0.05000,1.404 | |
| 37,mlp.gate_proj,0.0007103522,0.05000,3.077 | |
| 37,mlp.up_proj,0.0008592182,0.05000,3.084 | |
| 37,mlp.down_proj,0.0022288549,0.05000,5.345 | |
| 38,self_attn.v_proj,0.0011880093,0.05000,4.358 | |
| 38,self_attn.q_proj,0.0031908930,0.05000,4.372 | |
| 38,self_attn.k_proj,0.0006751707,0.05000,4.399 | |
| 38,self_attn.o_proj,0.0003943942,0.05000,1.340 | |
| 38,mlp.gate_proj,0.0008062866,0.05000,3.148 | |
| 38,mlp.up_proj,0.0009217460,0.05000,3.174 | |
| 38,mlp.down_proj,0.0034524150,0.05000,5.362 | |
| 39,self_attn.k_proj,0.0002554454,0.05000,4.257 | |
| 39,self_attn.v_proj,0.0003735272,0.05000,4.294 | |
| 39,self_attn.q_proj,0.0012166777,0.05000,4.295 | |
| 39,self_attn.o_proj,0.0003847522,0.05000,1.344 | |
| 39,mlp.gate_proj,0.0009165438,0.05000,3.561 | |
| 39,mlp.up_proj,0.0010139998,0.05000,3.578 | |
| 39,mlp.down_proj,0.0066875947,0.05000,5.467 | |