| layer,module,loss,samples,damp,time | |
| 0,self_attn.q_proj,0.0000000065,0.05000,3.237 | |
| 0,self_attn.k_proj,0.0000000025,0.05000,3.294 | |
| 0,self_attn.v_proj,0.0000000001,0.05000,3.323 | |
| 0,self_attn.o_proj,0.0000000000,0.05000,1.196 | |
| 0,mlp.up_proj,0.0000000046,0.05000,1.704 | |
| 0,mlp.gate_proj,0.0000000049,0.05000,1.725 | |
| 0,mlp.down_proj,0.0000000100,0.05000,2.482 | |
| 1,self_attn.q_proj,0.0000000252,0.05000,2.524 | |
| 1,self_attn.k_proj,0.0000000101,0.05000,2.555 | |
| 1,self_attn.v_proj,0.0000000020,0.05000,2.586 | |
| 1,self_attn.o_proj,0.0000000000,0.05000,1.217 | |
| 1,mlp.up_proj,0.0000000075,0.05000,1.556 | |
| 1,mlp.gate_proj,0.0000000083,0.05000,1.573 | |
| 1,mlp.down_proj,0.0000000002,0.05000,2.339 | |
| 2,self_attn.q_proj,0.0000000247,0.05000,2.528 | |
| 2,self_attn.v_proj,0.0000000025,0.05000,2.550 | |
| 2,self_attn.k_proj,0.0000000095,0.05000,2.604 | |
| 2,self_attn.o_proj,0.0000000001,0.05000,1.222 | |
| 2,mlp.up_proj,0.0000000174,0.05000,1.522 | |
| 2,mlp.gate_proj,0.0000000207,0.05000,1.526 | |
| 2,mlp.down_proj,0.0000000146,0.05000,2.345 | |
| 3,self_attn.q_proj,0.0000000564,0.05000,2.593 | |
| 3,self_attn.v_proj,0.0000000046,0.05000,2.616 | |
| 3,self_attn.k_proj,0.0000000200,0.05000,2.647 | |
| 3,self_attn.o_proj,0.0000000001,0.05000,1.248 | |
| 3,mlp.up_proj,0.0000000198,0.05000,1.472 | |
| 3,mlp.gate_proj,0.0000000244,0.05000,1.511 | |
| 3,mlp.down_proj,0.0000000007,0.05000,1.934 | |
| 4,self_attn.q_proj,0.0000000371,0.05000,2.624 | |
| 4,self_attn.k_proj,0.0000000134,0.05000,2.656 | |
| 4,self_attn.v_proj,0.0000000043,0.05000,2.673 | |
| 4,self_attn.o_proj,0.0000000002,0.05000,1.202 | |
| 4,mlp.gate_proj,0.0000000294,0.05000,1.394 | |
| 4,mlp.up_proj,0.0000000234,0.05000,1.455 | |
| 4,mlp.down_proj,0.0000000010,0.05000,2.109 | |
| 5,self_attn.q_proj,0.0000000501,0.05000,2.917 | |
| 5,self_attn.k_proj,0.0000000172,0.05000,2.951 | |
| 5,self_attn.v_proj,0.0000000047,0.05000,2.962 | |
| 5,self_attn.o_proj,0.0000000003,0.05000,1.344 | |
| 5,mlp.up_proj,0.0000000274,0.05000,1.585 | |
| 5,mlp.gate_proj,0.0000000334,0.05000,1.606 | |
| 5,mlp.down_proj,0.0000000013,0.05000,2.217 | |
| 6,self_attn.v_proj,0.0000000049,0.05000,3.482 | |
| 6,self_attn.k_proj,0.0000000203,0.05000,3.538 | |
| 6,self_attn.q_proj,0.0000000530,0.05000,3.597 | |
| 6,self_attn.o_proj,0.0000000005,0.05000,1.155 | |
| 6,mlp.up_proj,0.0000000308,0.05000,2.310 | |
| 6,mlp.gate_proj,0.0000000378,0.05000,2.325 | |
| 6,mlp.down_proj,0.0000000016,0.05000,2.162 | |
| 7,self_attn.q_proj,0.0000000523,0.05000,2.692 | |
| 7,self_attn.k_proj,0.0000000185,0.05000,2.735 | |
| 7,self_attn.v_proj,0.0000000059,0.05000,2.742 | |
| 7,self_attn.o_proj,0.0000000007,0.05000,1.292 | |
| 7,mlp.up_proj,0.0000000322,0.05000,1.648 | |
| 7,mlp.gate_proj,0.0000000383,0.05000,1.686 | |
| 7,mlp.down_proj,0.0000000017,0.05000,2.494 | |
| 8,self_attn.q_proj,0.0000000548,0.05000,2.815 | |
| 8,self_attn.v_proj,0.0000000058,0.05000,2.891 | |
| 8,self_attn.k_proj,0.0000000204,0.05000,2.901 | |
| 8,self_attn.o_proj,0.0000000008,0.05000,1.266 | |
| 8,mlp.up_proj,0.0000000345,0.05000,1.495 | |
| 8,mlp.gate_proj,0.0000000389,0.05000,1.512 | |
| 8,mlp.down_proj,0.0000000018,0.05000,2.806 | |
| 9,self_attn.q_proj,0.0000000562,0.05000,3.370 | |
| 9,self_attn.v_proj,0.0000000069,0.05000,3.427 | |
| 9,self_attn.k_proj,0.0000000202,0.05000,3.466 | |
| 9,self_attn.o_proj,0.0000000011,0.05000,1.213 | |
| 9,mlp.gate_proj,0.0000000390,0.05000,2.619 | |
| 9,mlp.up_proj,0.0000000360,0.05000,2.656 | |
| 9,mlp.down_proj,0.0000000022,0.05000,2.583 | |
| 10,self_attn.q_proj,0.0000000571,0.05000,3.761 | |
| 10,self_attn.k_proj,0.0000000206,0.05000,3.794 | |
| 10,self_attn.v_proj,0.0000000067,0.05000,3.890 | |
| 10,self_attn.o_proj,0.0000000012,0.05000,1.331 | |
| 10,mlp.up_proj,0.0000000367,0.05000,1.505 | |
| 10,mlp.gate_proj,0.0000000398,0.05000,1.521 | |
| 10,mlp.down_proj,0.0000000022,0.05000,2.692 | |
| 11,self_attn.v_proj,0.0000000071,0.05000,4.067 | |
| 11,self_attn.k_proj,0.0000000236,0.05000,4.114 | |
| 11,self_attn.q_proj,0.0000000628,0.05000,4.207 | |
| 11,self_attn.o_proj,0.0000000013,0.05000,1.199 | |
| 11,mlp.up_proj,0.0000000377,0.05000,2.797 | |
| 11,mlp.gate_proj,0.0000000459,0.05000,2.812 | |
| 11,mlp.down_proj,0.0000000023,0.05000,2.457 | |
| 12,self_attn.q_proj,0.0000000592,0.05000,3.017 | |
| 12,self_attn.k_proj,0.0000000211,0.05000,3.025 | |
| 12,self_attn.v_proj,0.0000000078,0.05000,3.054 | |
| 12,self_attn.o_proj,0.0000000012,0.05000,1.142 | |
| 12,mlp.up_proj,0.0000000408,0.05000,1.467 | |
| 12,mlp.gate_proj,0.0000000459,0.05000,1.494 | |
| 12,mlp.down_proj,0.0000000025,0.05000,1.930 | |
| 13,self_attn.k_proj,0.0000000262,0.05000,3.431 | |
| 13,self_attn.v_proj,0.0000000082,0.05000,3.440 | |
| 13,self_attn.q_proj,0.0000000735,0.05000,3.470 | |
| 13,self_attn.o_proj,0.0000000014,0.05000,1.137 | |
| 13,mlp.gate_proj,0.0000000465,0.05000,2.825 | |
| 13,mlp.up_proj,0.0000000430,0.05000,2.842 | |
| 13,mlp.down_proj,0.0000000043,0.05000,2.422 | |
| 14,self_attn.q_proj,0.0000000784,0.05000,2.812 | |
| 14,self_attn.v_proj,0.0000000085,0.05000,2.849 | |
| 14,self_attn.k_proj,0.0000000266,0.05000,2.859 | |
| 14,self_attn.o_proj,0.0000000014,0.05000,1.253 | |
| 14,mlp.up_proj,0.0000000472,0.05000,1.679 | |
| 14,mlp.gate_proj,0.0000000482,0.05000,1.697 | |
| 14,mlp.down_proj,0.0000000032,0.05000,2.484 | |
| 15,self_attn.q_proj,0.0000000665,0.05000,3.227 | |
| 15,self_attn.v_proj,0.0000000077,0.05000,3.285 | |
| 15,self_attn.k_proj,0.0000000255,0.05000,3.301 | |
| 15,self_attn.o_proj,0.0000000014,0.05000,1.267 | |
| 15,mlp.up_proj,0.0000000480,0.05000,1.881 | |
| 15,mlp.gate_proj,0.0000000479,0.05000,1.896 | |
| 15,mlp.down_proj,0.0000000033,0.05000,2.738 | |
| 16,self_attn.k_proj,0.0000000255,0.05000,3.339 | |
| 16,self_attn.v_proj,0.0000000098,0.05000,3.377 | |
| 16,self_attn.q_proj,0.0000000680,0.05000,3.406 | |
| 16,self_attn.o_proj,0.0000000019,0.05000,0.877 | |
| 16,mlp.up_proj,0.0000000505,0.05000,2.809 | |
| 16,mlp.gate_proj,0.0000000501,0.05000,2.840 | |
| 16,mlp.down_proj,0.0000000038,0.05000,2.236 | |
| 17,self_attn.v_proj,0.0000000107,0.05000,3.385 | |
| 17,self_attn.k_proj,0.0000000242,0.05000,3.440 | |
| 17,self_attn.q_proj,0.0000000681,0.05000,3.506 | |
| 17,self_attn.o_proj,0.0000000023,0.05000,1.141 | |
| 17,mlp.up_proj,0.0000000530,0.05000,2.651 | |
| 17,mlp.gate_proj,0.0000000523,0.05000,2.683 | |
| 17,mlp.down_proj,0.0000000048,0.05000,2.129 | |
| 18,self_attn.q_proj,0.0000000738,0.05000,2.996 | |
| 18,self_attn.v_proj,0.0000000118,0.05000,3.027 | |
| 18,self_attn.k_proj,0.0000000268,0.05000,3.048 | |
| 18,self_attn.o_proj,0.0000000025,0.05000,1.195 | |
| 18,mlp.up_proj,0.0000000553,0.05000,1.710 | |
| 18,mlp.gate_proj,0.0000000574,0.05000,1.729 | |
| 18,mlp.down_proj,0.0000000055,0.05000,2.685 | |
| 19,self_attn.q_proj,0.0000000821,0.05000,2.524 | |
| 19,self_attn.v_proj,0.0000000151,0.05000,2.541 | |
| 19,self_attn.k_proj,0.0000000250,0.05000,2.562 | |
| 19,self_attn.o_proj,0.0000000021,0.05000,1.216 | |
| 19,mlp.up_proj,0.0000000566,0.05000,1.476 | |
| 19,mlp.gate_proj,0.0000000588,0.05000,1.517 | |
| 19,mlp.down_proj,0.0000000064,0.05000,2.657 | |
| 20,self_attn.k_proj,0.0000000254,0.05000,2.863 | |
| 20,self_attn.q_proj,0.0000000785,0.05000,2.882 | |
| 20,self_attn.v_proj,0.0000000142,0.05000,2.892 | |
| 20,self_attn.o_proj,0.0000000018,0.05000,1.273 | |
| 20,mlp.up_proj,0.0000000645,0.05000,1.523 | |
| 20,mlp.gate_proj,0.0000000688,0.05000,1.529 | |
| 20,mlp.down_proj,0.0000000074,0.05000,2.146 | |
| 21,self_attn.q_proj,0.0000000758,0.05000,2.626 | |
| 21,self_attn.v_proj,0.0000000129,0.05000,2.643 | |
| 21,self_attn.k_proj,0.0000000244,0.05000,2.664 | |
| 21,self_attn.o_proj,0.0000000014,0.05000,1.328 | |
| 21,mlp.up_proj,0.0000000716,0.05000,1.519 | |
| 21,mlp.gate_proj,0.0000000772,0.05000,1.541 | |
| 21,mlp.down_proj,0.0000000084,0.05000,2.648 | |
| 22,self_attn.q_proj,0.0000000760,0.05000,3.406 | |
| 22,self_attn.v_proj,0.0000000171,0.05000,3.436 | |
| 22,self_attn.k_proj,0.0000000222,0.05000,3.509 | |
| 22,self_attn.o_proj,0.0000000020,0.05000,1.197 | |
| 22,mlp.up_proj,0.0000000826,0.05000,1.531 | |
| 22,mlp.gate_proj,0.0000000891,0.05000,1.551 | |
| 22,mlp.down_proj,0.0000000106,0.05000,1.926 | |
| 23,self_attn.k_proj,0.0000000248,0.05000,3.158 | |
| 23,self_attn.v_proj,0.0000000174,0.05000,3.208 | |
| 23,self_attn.q_proj,0.0000000820,0.05000,3.223 | |
| 23,self_attn.o_proj,0.0000000017,0.05000,1.352 | |
| 23,mlp.up_proj,0.0000000922,0.05000,1.857 | |
| 23,mlp.gate_proj,0.0000000959,0.05000,1.868 | |
| 23,mlp.down_proj,0.0000000136,0.05000,2.161 | |
| 24,self_attn.v_proj,0.0000000204,0.05000,3.519 | |
| 24,self_attn.k_proj,0.0000000271,0.05000,3.567 | |
| 24,self_attn.q_proj,0.0000000894,0.05000,3.585 | |
| 24,self_attn.o_proj,0.0000000019,0.05000,1.231 | |
| 24,mlp.gate_proj,0.0000001053,0.05000,1.510 | |
| 24,mlp.up_proj,0.0000001039,0.05000,1.517 | |
| 24,mlp.down_proj,0.0000000152,0.05000,2.719 | |
| 25,self_attn.v_proj,0.0000000223,0.05000,3.591 | |
| 25,self_attn.k_proj,0.0000000271,0.05000,3.614 | |
| 25,self_attn.q_proj,0.0000000925,0.05000,3.645 | |
| 25,self_attn.o_proj,0.0000000020,0.05000,1.242 | |
| 25,mlp.up_proj,0.0000001170,0.05000,2.637 | |
| 25,mlp.gate_proj,0.0000001152,0.05000,2.675 | |
| 25,mlp.down_proj,0.0000000275,0.05000,2.711 | |
| 26,self_attn.q_proj,0.0000000935,0.05000,3.617 | |
| 26,self_attn.k_proj,0.0000000267,0.05000,3.679 | |
| 26,self_attn.v_proj,0.0000000270,0.05000,3.773 | |
| 26,self_attn.o_proj,0.0000000023,0.05000,1.309 | |
| 26,mlp.gate_proj,0.0000001292,0.05000,1.537 | |
| 26,mlp.up_proj,0.0000001352,0.05000,1.546 | |
| 26,mlp.down_proj,0.0000000254,0.05000,2.422 | |
| 27,self_attn.k_proj,0.0000000290,0.05000,3.722 | |
| 27,self_attn.v_proj,0.0000000276,0.05000,3.737 | |
| 27,self_attn.q_proj,0.0000001083,0.05000,3.835 | |
| 27,self_attn.o_proj,0.0000000035,0.05000,1.329 | |
| 27,mlp.gate_proj,0.0000001528,0.05000,2.441 | |
| 27,mlp.up_proj,0.0000001548,0.05000,2.449 | |
| 27,mlp.down_proj,0.0000000359,0.05000,2.385 | |
| 28,self_attn.q_proj,0.0000001191,0.05000,4.144 | |
| 28,self_attn.k_proj,0.0000000272,0.05000,4.234 | |
| 28,self_attn.v_proj,0.0000000615,0.05000,4.276 | |
| 28,self_attn.o_proj,0.0000000068,0.05000,1.309 | |
| 28,mlp.gate_proj,0.0000001769,0.05000,2.691 | |
| 28,mlp.up_proj,0.0000001749,0.05000,2.721 | |
| 28,mlp.down_proj,0.0000000435,0.05000,2.449 | |
| 29,self_attn.k_proj,0.0000000269,0.05000,3.207 | |
| 29,self_attn.q_proj,0.0000001154,0.05000,3.232 | |
| 29,self_attn.v_proj,0.0000000527,0.05000,3.249 | |
| 29,self_attn.o_proj,0.0000000058,0.05000,1.255 | |
| 29,mlp.gate_proj,0.0000002023,0.05000,1.783 | |
| 29,mlp.up_proj,0.0000001973,0.05000,1.790 | |
| 29,mlp.down_proj,0.0000000607,0.05000,2.164 | |
| 30,self_attn.q_proj,0.0000001173,0.05000,2.626 | |
| 30,self_attn.k_proj,0.0000000259,0.05000,2.679 | |
| 30,self_attn.v_proj,0.0000000572,0.05000,2.690 | |
| 30,self_attn.o_proj,0.0000000086,0.05000,1.171 | |
| 30,mlp.gate_proj,0.0000002210,0.05000,1.537 | |
| 30,mlp.up_proj,0.0000002175,0.05000,1.571 | |
| 30,mlp.down_proj,0.0000000779,0.05000,2.725 | |
| 31,self_attn.q_proj,0.0000001154,0.05000,3.332 | |
| 31,self_attn.v_proj,0.0000000605,0.05000,3.378 | |
| 31,self_attn.k_proj,0.0000000247,0.05000,3.403 | |
| 31,self_attn.o_proj,0.0000000200,0.05000,1.127 | |
| 31,mlp.up_proj,0.0000002282,0.05000,1.718 | |
| 31,mlp.gate_proj,0.0000002381,0.05000,1.728 | |
| 31,mlp.down_proj,0.0000001104,0.05000,1.935 | |
| 32,self_attn.k_proj,0.0000000261,0.05000,2.511 | |
| 32,self_attn.v_proj,0.0000001036,0.05000,2.530 | |
| 32,self_attn.q_proj,0.0000001396,0.05000,2.548 | |
| 32,self_attn.o_proj,0.0000000206,0.05000,1.102 | |
| 32,mlp.up_proj,0.0000002448,0.05000,2.389 | |
| 32,mlp.gate_proj,0.0000002711,0.05000,2.400 | |
| 32,mlp.down_proj,0.0000001743,0.05000,2.373 | |
| 33,self_attn.k_proj,0.0000000231,0.05000,3.245 | |
| 33,self_attn.q_proj,0.0000001061,0.05000,3.269 | |
| 33,self_attn.v_proj,0.0000000586,0.05000,3.274 | |
| 33,self_attn.o_proj,0.0000000248,0.05000,1.331 | |
| 33,mlp.gate_proj,0.0000002696,0.05000,1.556 | |
| 33,mlp.up_proj,0.0000002391,0.05000,1.580 | |
| 33,mlp.down_proj,0.0000005028,0.05000,2.148 | |