| layer,module,loss,samples,damp,time | |
| 0,self_attn.q_proj,0.0000003104,0.05000,4.905 | |
| 0,self_attn.v_proj,0.0000000719,0.05000,4.994 | |
| 0,self_attn.k_proj,0.0000000848,0.05000,5.018 | |
| 0,self_attn.o_proj,0.0000001642,0.05000,1.457 | |
| 0,mlp.up_proj,0.0000166251,0.05000,1.845 | |
| 0,mlp.gate_proj,0.0000194163,0.05000,1.875 | |
| 0,mlp.down_proj,0.0000009856,0.05000,3.691 | |
| 1,self_attn.k_proj,0.0000001349,0.05000,4.619 | |
| 1,self_attn.q_proj,0.0000004949,0.05000,4.702 | |
| 1,self_attn.v_proj,0.0000001330,0.05000,4.728 | |
| 1,self_attn.o_proj,0.0000001566,0.05000,1.611 | |
| 1,mlp.gate_proj,0.0005942452,0.05000,2.445 | |
| 1,mlp.up_proj,0.0002664099,0.05000,2.507 | |
| 1,mlp.down_proj,0.0000016856,0.05000,3.879 | |
| 2,self_attn.k_proj,0.0000003772,0.05000,4.616 | |
| 2,self_attn.q_proj,0.0000012779,0.05000,4.651 | |
| 2,self_attn.v_proj,0.0000003293,0.05000,4.678 | |
| 2,self_attn.o_proj,0.0000002073,0.05000,1.550 | |
| 2,mlp.gate_proj,0.0010680482,0.05000,2.444 | |
| 2,mlp.up_proj,0.0008312045,0.05000,2.475 | |
| 2,mlp.down_proj,0.0000015937,0.05000,3.758 | |
| 3,self_attn.k_proj,0.0000007141,0.05000,4.042 | |
| 3,self_attn.q_proj,0.0000027178,0.05000,4.098 | |
| 3,self_attn.v_proj,0.0000006610,0.05000,4.133 | |
| 3,self_attn.o_proj,0.0000004343,0.05000,1.443 | |
| 3,mlp.up_proj,0.0004659894,0.05000,1.978 | |
| 3,mlp.gate_proj,0.0009418213,0.05000,1.998 | |
| 3,mlp.down_proj,0.0000026344,0.05000,3.723 | |
| 4,self_attn.v_proj,0.0000011467,0.05000,4.135 | |
| 4,self_attn.k_proj,0.0000012765,0.05000,4.217 | |
| 4,self_attn.q_proj,0.0000045529,0.05000,4.233 | |
| 4,self_attn.o_proj,0.0000005542,0.05000,1.409 | |
| 4,mlp.gate_proj,0.0008452089,0.05000,2.026 | |
| 4,mlp.up_proj,0.0002988149,0.05000,2.043 | |
| 4,mlp.down_proj,0.0000032080,0.05000,3.691 | |
| 5,self_attn.k_proj,0.0000012628,0.05000,4.213 | |
| 5,self_attn.v_proj,0.0000012745,0.05000,4.269 | |
| 5,self_attn.q_proj,0.0000048576,0.05000,4.303 | |
| 5,self_attn.o_proj,0.0000010330,0.05000,1.439 | |
| 5,mlp.up_proj,0.0001115446,0.05000,2.013 | |
| 5,mlp.gate_proj,0.0002728422,0.05000,2.022 | |
| 5,mlp.down_proj,0.0000040210,0.05000,3.732 | |
| 6,self_attn.k_proj,0.0000024965,0.05000,4.120 | |
| 6,self_attn.q_proj,0.0000103244,0.05000,4.136 | |
| 6,self_attn.v_proj,0.0000026516,0.05000,4.181 | |
| 6,self_attn.o_proj,0.0000022328,0.05000,1.425 | |
| 6,mlp.up_proj,0.0001320891,0.05000,1.990 | |
| 6,mlp.gate_proj,0.0002955694,0.05000,1.999 | |
| 6,mlp.down_proj,0.0000908862,0.05000,3.725 | |
| 7,self_attn.q_proj,0.0000219465,0.05000,4.214 | |
| 7,self_attn.k_proj,0.0000061061,0.05000,4.320 | |
| 7,self_attn.v_proj,0.0000053100,0.05000,4.340 | |
| 7,self_attn.o_proj,0.0000029166,0.05000,1.409 | |
| 7,mlp.gate_proj,0.0003294983,0.05000,1.882 | |
| 7,mlp.up_proj,0.0001593253,0.05000,1.897 | |
| 7,mlp.down_proj,0.0000070525,0.05000,3.790 | |
| 8,self_attn.v_proj,0.0000074580,0.05000,4.248 | |
| 8,self_attn.q_proj,0.0000303084,0.05000,4.315 | |
| 8,self_attn.k_proj,0.0000081583,0.05000,4.334 | |
| 8,self_attn.o_proj,0.0000036968,0.05000,1.403 | |
| 8,mlp.up_proj,0.0001478816,0.05000,1.953 | |
| 8,mlp.gate_proj,0.0001890822,0.05000,1.967 | |
| 8,mlp.down_proj,0.0000118579,0.05000,3.753 | |
| 9,self_attn.v_proj,0.0000095992,0.05000,4.190 | |
| 9,self_attn.q_proj,0.0000413045,0.05000,4.249 | |
| 9,self_attn.k_proj,0.0000121445,0.05000,4.268 | |
| 9,self_attn.o_proj,0.0000053125,0.05000,1.515 | |
| 9,mlp.gate_proj,0.0003300729,0.05000,1.967 | |
| 9,mlp.up_proj,0.0001974271,0.05000,1.986 | |
| 9,mlp.down_proj,0.0000125098,0.05000,3.755 | |
| 10,self_attn.k_proj,0.0000153336,0.05000,4.206 | |
| 10,self_attn.q_proj,0.0000581782,0.05000,4.224 | |
| 10,self_attn.v_proj,0.0000151345,0.05000,4.264 | |
| 10,self_attn.o_proj,0.0000061027,0.05000,1.396 | |
| 10,mlp.up_proj,0.0001707550,0.05000,1.915 | |
| 10,mlp.gate_proj,0.0002543201,0.05000,1.933 | |
| 10,mlp.down_proj,0.0000103128,0.05000,3.691 | |
| 11,self_attn.k_proj,0.0000080204,0.05000,4.166 | |
| 11,self_attn.q_proj,0.0000286084,0.05000,4.207 | |
| 11,self_attn.v_proj,0.0000070372,0.05000,4.228 | |
| 11,self_attn.o_proj,0.0000043562,0.05000,1.416 | |
| 11,mlp.up_proj,0.0001597086,0.05000,1.959 | |
| 11,mlp.gate_proj,0.0002079164,0.05000,1.979 | |
| 11,mlp.down_proj,0.0000100665,0.05000,3.695 | |
| 12,self_attn.q_proj,0.0000361063,0.05000,4.147 | |
| 12,self_attn.k_proj,0.0000098947,0.05000,4.237 | |
| 12,self_attn.v_proj,0.0000093602,0.05000,4.269 | |
| 12,self_attn.o_proj,0.0000063775,0.05000,1.406 | |
| 12,mlp.up_proj,0.0001564673,0.05000,1.963 | |
| 12,mlp.gate_proj,0.0001846403,0.05000,1.958 | |
| 12,mlp.down_proj,0.0000100620,0.05000,3.720 | |
| 13,self_attn.q_proj,0.0000280880,0.05000,4.138 | |
| 13,self_attn.k_proj,0.0000076636,0.05000,4.192 | |
| 13,self_attn.v_proj,0.0000065074,0.05000,4.205 | |
| 13,self_attn.o_proj,0.0000054739,0.05000,1.425 | |
| 13,mlp.up_proj,0.0001643997,0.05000,2.025 | |
| 13,mlp.gate_proj,0.0001773453,0.05000,2.042 | |
| 13,mlp.down_proj,0.0000121140,0.05000,3.696 | |
| 14,self_attn.q_proj,0.0000446386,0.05000,4.120 | |
| 14,self_attn.v_proj,0.0000111164,0.05000,4.194 | |
| 14,self_attn.k_proj,0.0000117079,0.05000,4.227 | |
| 14,self_attn.o_proj,0.0000080355,0.05000,1.390 | |
| 14,mlp.up_proj,0.0001662609,0.05000,1.928 | |
| 14,mlp.gate_proj,0.0001788644,0.05000,1.951 | |
| 14,mlp.down_proj,0.0000115574,0.05000,3.707 | |
| 15,self_attn.k_proj,0.0000127820,0.05000,3.983 | |
| 15,self_attn.v_proj,0.0000108480,0.05000,4.041 | |
| 15,self_attn.q_proj,0.0000485437,0.05000,4.066 | |
| 15,self_attn.o_proj,0.0000079289,0.05000,1.396 | |
| 15,mlp.up_proj,0.0001686920,0.05000,1.938 | |
| 15,mlp.gate_proj,0.0001739771,0.05000,1.951 | |
| 15,mlp.down_proj,0.0000120265,0.05000,3.730 | |
| 16,self_attn.v_proj,0.0000226708,0.05000,4.071 | |
| 16,self_attn.q_proj,0.0000885504,0.05000,4.176 | |
| 16,self_attn.k_proj,0.0000220067,0.05000,4.193 | |
| 16,self_attn.o_proj,0.0000097078,0.05000,1.403 | |
| 16,mlp.up_proj,0.0001986332,0.05000,1.905 | |
| 16,mlp.gate_proj,0.0002191926,0.05000,1.925 | |
| 16,mlp.down_proj,0.0001195219,0.05000,3.697 | |
| 17,self_attn.v_proj,0.0000192036,0.05000,4.160 | |
| 17,self_attn.q_proj,0.0000859028,0.05000,4.196 | |
| 17,self_attn.k_proj,0.0000187684,0.05000,4.219 | |
| 17,self_attn.o_proj,0.0000104221,0.05000,1.429 | |
| 17,mlp.up_proj,0.0001920366,0.05000,1.979 | |
| 17,mlp.gate_proj,0.0002083335,0.05000,1.997 | |
| 17,mlp.down_proj,0.0000144278,0.05000,3.702 | |
| 18,self_attn.q_proj,0.0000918104,0.05000,3.970 | |
| 18,self_attn.k_proj,0.0000223279,0.05000,4.084 | |
| 18,self_attn.v_proj,0.0000228247,0.05000,4.105 | |
| 18,self_attn.o_proj,0.0000120030,0.05000,1.408 | |
| 18,mlp.up_proj,0.0002062233,0.05000,1.959 | |
| 18,mlp.gate_proj,0.0002194696,0.05000,1.968 | |
| 18,mlp.down_proj,0.0000170803,0.05000,3.642 | |
| 19,self_attn.q_proj,0.0001786820,0.05000,4.068 | |
| 19,self_attn.v_proj,0.0000392877,0.05000,4.147 | |
| 19,self_attn.k_proj,0.0000417310,0.05000,4.178 | |
| 19,self_attn.o_proj,0.0000172448,0.05000,1.420 | |
| 19,mlp.up_proj,0.0002172751,0.05000,2.036 | |
| 19,mlp.gate_proj,0.0002297170,0.05000,2.057 | |
| 19,mlp.down_proj,0.0000206614,0.05000,3.705 | |
| 20,self_attn.q_proj,0.0001758999,0.05000,4.152 | |
| 20,self_attn.v_proj,0.0000377428,0.05000,4.207 | |
| 20,self_attn.k_proj,0.0000383674,0.05000,4.233 | |
| 20,self_attn.o_proj,0.0000194928,0.05000,1.407 | |
| 20,mlp.gate_proj,0.0002446541,0.05000,1.948 | |
| 20,mlp.up_proj,0.0002372765,0.05000,1.968 | |
| 20,mlp.down_proj,0.0000223872,0.05000,3.734 | |
| 21,self_attn.v_proj,0.0000462733,0.05000,4.032 | |
| 21,self_attn.k_proj,0.0000457615,0.05000,4.107 | |
| 21,self_attn.q_proj,0.0002003734,0.05000,4.130 | |
| 21,self_attn.o_proj,0.0000193219,0.05000,1.382 | |
| 21,mlp.up_proj,0.0002598240,0.05000,1.994 | |
| 21,mlp.gate_proj,0.0002718615,0.05000,2.013 | |
| 21,mlp.down_proj,0.0000280902,0.05000,3.820 | |
| 22,self_attn.v_proj,0.0000893241,0.05000,4.115 | |
| 22,self_attn.k_proj,0.0000910953,0.05000,4.191 | |
| 22,self_attn.q_proj,0.0003968653,0.05000,4.210 | |
| 22,self_attn.o_proj,0.0000352733,0.05000,1.394 | |
| 22,mlp.up_proj,0.0002868396,0.05000,1.933 | |
| 22,mlp.gate_proj,0.0003197921,0.05000,1.949 | |
| 22,mlp.down_proj,0.0000435962,0.05000,3.718 | |
| 23,self_attn.k_proj,0.0000845952,0.05000,4.110 | |
| 23,self_attn.q_proj,0.0003595412,0.05000,4.206 | |
| 23,self_attn.v_proj,0.0000884456,0.05000,4.234 | |
| 23,self_attn.o_proj,0.0000378814,0.05000,1.389 | |
| 23,mlp.gate_proj,0.0003853287,0.05000,1.959 | |
| 23,mlp.up_proj,0.0003278120,0.05000,1.955 | |
| 23,mlp.down_proj,0.0000534777,0.05000,3.675 | |
| 24,self_attn.k_proj,0.0001351002,0.05000,4.143 | |
| 24,self_attn.v_proj,0.0001678118,0.05000,4.186 | |
| 24,self_attn.q_proj,0.0005597780,0.05000,4.222 | |
| 24,self_attn.o_proj,0.0000466474,0.05000,1.423 | |
| 24,mlp.up_proj,0.0003403691,0.05000,1.974 | |
| 24,mlp.gate_proj,0.0004069859,0.05000,1.992 | |
| 24,mlp.down_proj,0.0000601208,0.05000,3.689 | |
| 25,self_attn.q_proj,0.0003528415,0.05000,4.221 | |
| 25,self_attn.k_proj,0.0000905896,0.05000,4.284 | |
| 25,self_attn.v_proj,0.0000948952,0.05000,4.304 | |
| 25,self_attn.o_proj,0.0000231992,0.05000,1.395 | |
| 25,mlp.up_proj,0.0003582715,0.05000,2.032 | |
| 25,mlp.gate_proj,0.0004356605,0.05000,2.049 | |
| 25,mlp.down_proj,0.0000689074,0.05000,3.674 | |
| 26,self_attn.k_proj,0.0001325942,0.05000,4.171 | |
| 26,self_attn.v_proj,0.0001480613,0.05000,4.257 | |
| 26,self_attn.q_proj,0.0005470764,0.05000,4.309 | |
| 26,self_attn.o_proj,0.0000218077,0.05000,1.398 | |
| 26,mlp.gate_proj,0.0004783157,0.05000,1.944 | |
| 26,mlp.up_proj,0.0004074046,0.05000,1.974 | |
| 26,mlp.down_proj,0.0000757961,0.05000,3.774 | |
| 27,self_attn.v_proj,0.0001919626,0.05000,4.095 | |
| 27,self_attn.q_proj,0.0006371495,0.05000,4.127 | |
| 27,self_attn.k_proj,0.0001516362,0.05000,4.139 | |
| 27,self_attn.o_proj,0.0000294259,0.05000,1.432 | |
| 27,mlp.gate_proj,0.0005036255,0.05000,1.895 | |
| 27,mlp.up_proj,0.0004426054,0.05000,1.907 | |
| 27,mlp.down_proj,0.0001096823,0.05000,3.705 | |
| 28,self_attn.v_proj,0.0002048029,0.05000,4.179 | |
| 28,self_attn.q_proj,0.0007217175,0.05000,4.261 | |
| 28,self_attn.k_proj,0.0001894458,0.05000,4.277 | |
| 28,self_attn.o_proj,0.0000451533,0.05000,1.456 | |
| 28,mlp.up_proj,0.0004963349,0.05000,1.984 | |
| 28,mlp.gate_proj,0.0005422109,0.05000,1.999 | |
| 28,mlp.down_proj,0.0001637474,0.05000,3.662 | |
| 29,self_attn.q_proj,0.0017744177,0.05000,4.176 | |
| 29,self_attn.v_proj,0.0006037154,0.05000,4.219 | |
| 29,self_attn.k_proj,0.0004375837,0.05000,4.246 | |
| 29,self_attn.o_proj,0.0000526574,0.05000,1.403 | |
| 29,mlp.gate_proj,0.0005959541,0.05000,2.040 | |
| 29,mlp.up_proj,0.0005792906,0.05000,2.057 | |
| 29,mlp.down_proj,0.0001176740,0.05000,3.716 | |
| 30,self_attn.q_proj,0.0017855356,0.05000,4.072 | |
| 30,self_attn.k_proj,0.0004658168,0.05000,4.109 | |
| 30,self_attn.v_proj,0.0005561837,0.05000,4.118 | |
| 30,self_attn.o_proj,0.0000770569,0.05000,1.385 | |
| 30,mlp.up_proj,0.0005726860,0.05000,1.973 | |
| 30,mlp.gate_proj,0.0005691606,0.05000,1.992 | |
| 30,mlp.down_proj,0.0001377208,0.05000,3.679 | |
| 31,self_attn.k_proj,0.0006738857,0.05000,4.074 | |
| 31,self_attn.q_proj,0.0023534875,0.05000,4.180 | |
| 31,self_attn.v_proj,0.0009032022,0.05000,4.202 | |
| 31,self_attn.o_proj,0.0000924383,0.05000,1.398 | |
| 31,mlp.gate_proj,0.0005124867,0.05000,2.077 | |
| 31,mlp.up_proj,0.0005424235,0.05000,2.092 | |
| 31,mlp.down_proj,0.0001507833,0.05000,3.702 | |
| 32,self_attn.q_proj,0.0033660829,0.05000,4.254 | |
| 32,self_attn.v_proj,0.0012981689,0.05000,4.251 | |
| 32,self_attn.k_proj,0.0008497750,0.05000,4.311 | |
| 32,self_attn.o_proj,0.0001307035,0.05000,1.413 | |
| 32,mlp.up_proj,0.0005475206,0.05000,1.964 | |
| 32,mlp.gate_proj,0.0005065469,0.05000,1.982 | |
| 32,mlp.down_proj,0.0001925587,0.05000,3.766 | |
| 33,self_attn.q_proj,0.0081515312,0.05000,4.143 | |
| 33,self_attn.k_proj,0.0017276332,0.05000,4.244 | |
| 33,self_attn.v_proj,0.0032664668,0.05000,4.253 | |
| 33,self_attn.o_proj,0.0002568730,0.05000,1.422 | |
| 33,mlp.up_proj,0.0006197004,0.05000,1.991 | |
| 33,mlp.gate_proj,0.0005619661,0.05000,2.013 | |
| 33,mlp.down_proj,0.0002740641,0.05000,3.756 | |
| 34,self_attn.k_proj,0.0012690087,0.05000,4.134 | |
| 34,self_attn.q_proj,0.0056529627,0.05000,4.188 | |
| 34,self_attn.v_proj,0.0020060930,0.05000,4.221 | |
| 34,self_attn.o_proj,0.0002881256,0.05000,1.424 | |
| 34,mlp.up_proj,0.0007620189,0.05000,1.904 | |
| 34,mlp.gate_proj,0.0007313601,0.05000,1.906 | |
| 34,mlp.down_proj,0.0005008702,0.05000,3.746 | |
| 35,self_attn.k_proj,0.0005734052,0.05000,4.103 | |
| 35,self_attn.v_proj,0.0006820257,0.05000,4.200 | |
| 35,self_attn.q_proj,0.0021040146,0.05000,4.217 | |
| 35,self_attn.o_proj,0.0002867113,0.05000,1.422 | |
| 35,mlp.up_proj,0.0022627204,0.05000,1.957 | |
| 35,mlp.gate_proj,0.0019303147,0.05000,1.969 | |
| 35,mlp.down_proj,0.0015434768,0.05000,3.715 | |