| layer,module,loss,samples,damp,time | |
| 0,self_attn.k_proj,0.0000000493,0.05000,5.263 | |
| 0,self_attn.v_proj,0.0000000430,0.05000,5.230 | |
| 0,self_attn.q_proj,0.0000001818,0.05000,5.398 | |
| 0,self_attn.o_proj,0.0000001054,0.05000,1.579 | |
| 0,mlp.gate_proj,0.0000114880,0.05000,2.158 | |
| 0,mlp.up_proj,0.0000100735,0.05000,2.185 | |
| 0,mlp.down_proj,0.0000015222,0.05000,4.023 | |
| 1,self_attn.v_proj,0.0000000828,0.05000,5.549 | |
| 1,self_attn.k_proj,0.0000000801,0.05000,5.598 | |
| 1,self_attn.q_proj,0.0000002977,0.05000,5.632 | |
| 1,self_attn.o_proj,0.0000001564,0.05000,1.642 | |
| 1,mlp.up_proj,0.0001828177,0.05000,2.748 | |
| 1,mlp.gate_proj,0.0004096040,0.05000,2.781 | |
| 1,mlp.down_proj,0.0000016779,0.05000,4.167 | |
| 2,self_attn.q_proj,0.0000008498,0.05000,5.458 | |
| 2,self_attn.k_proj,0.0000002434,0.05000,5.462 | |
| 2,self_attn.v_proj,0.0000002356,0.05000,5.488 | |
| 2,self_attn.o_proj,0.0000002095,0.05000,1.587 | |
| 2,mlp.gate_proj,0.0006597485,0.05000,2.158 | |
| 2,mlp.up_proj,0.0005201637,0.05000,2.182 | |
| 2,mlp.down_proj,0.0000012058,0.05000,4.134 | |
| 3,self_attn.v_proj,0.0000004307,0.05000,4.747 | |
| 3,self_attn.q_proj,0.0000016371,0.05000,4.795 | |
| 3,self_attn.k_proj,0.0000004186,0.05000,4.822 | |
| 3,self_attn.o_proj,0.0000003189,0.05000,1.582 | |
| 3,mlp.gate_proj,0.0006873334,0.05000,2.253 | |
| 3,mlp.up_proj,0.0003494455,0.05000,2.278 | |
| 3,mlp.down_proj,0.0000034346,0.05000,4.083 | |
| 4,self_attn.k_proj,0.0000008386,0.05000,4.757 | |
| 4,self_attn.q_proj,0.0000030813,0.05000,4.826 | |
| 4,self_attn.v_proj,0.0000008369,0.05000,4.856 | |
| 4,self_attn.o_proj,0.0000004898,0.05000,1.581 | |
| 4,mlp.gate_proj,0.0005583692,0.05000,2.182 | |
| 4,mlp.up_proj,0.0002213096,0.05000,2.199 | |
| 4,mlp.down_proj,0.0000050593,0.05000,4.103 | |
| 5,self_attn.v_proj,0.0000009181,0.05000,4.694 | |
| 5,self_attn.q_proj,0.0000033667,0.05000,4.788 | |
| 5,self_attn.k_proj,0.0000008563,0.05000,4.810 | |
| 5,self_attn.o_proj,0.0000009128,0.05000,1.617 | |
| 5,mlp.up_proj,0.0000946976,0.05000,2.171 | |
| 5,mlp.gate_proj,0.0001955191,0.05000,2.186 | |
| 5,mlp.down_proj,0.0000073415,0.05000,4.025 | |
| 6,self_attn.k_proj,0.0000017104,0.05000,4.671 | |
| 6,self_attn.q_proj,0.0000072547,0.05000,4.673 | |
| 6,self_attn.v_proj,0.0000019447,0.05000,4.740 | |
| 6,self_attn.o_proj,0.0000021203,0.05000,1.586 | |
| 6,mlp.up_proj,0.0001155930,0.05000,2.146 | |
| 6,mlp.gate_proj,0.0002014494,0.05000,2.173 | |
| 6,mlp.down_proj,0.0001064753,0.05000,4.027 | |
| 7,self_attn.v_proj,0.0000040180,0.05000,4.646 | |
| 7,self_attn.k_proj,0.0000039139,0.05000,4.743 | |
| 7,self_attn.q_proj,0.0000144505,0.05000,4.788 | |
| 7,self_attn.o_proj,0.0000026974,0.05000,1.584 | |
| 7,mlp.gate_proj,0.0002344327,0.05000,2.212 | |
| 7,mlp.up_proj,0.0001410635,0.05000,2.228 | |
| 7,mlp.down_proj,0.0000122430,0.05000,4.045 | |
| 8,self_attn.v_proj,0.0000061553,0.05000,4.724 | |
| 8,self_attn.k_proj,0.0000056424,0.05000,4.805 | |
| 8,self_attn.q_proj,0.0000217994,0.05000,4.822 | |
| 8,self_attn.o_proj,0.0000037739,0.05000,1.553 | |
| 8,mlp.up_proj,0.0001321165,0.05000,2.204 | |
| 8,mlp.gate_proj,0.0001614960,0.05000,2.229 | |
| 8,mlp.down_proj,0.0000182675,0.05000,4.046 | |
| 9,self_attn.v_proj,0.0000074750,0.05000,4.669 | |
| 9,self_attn.k_proj,0.0000076202,0.05000,4.776 | |
| 9,self_attn.q_proj,0.0000272138,0.05000,4.797 | |
| 9,self_attn.o_proj,0.0000046978,0.05000,1.548 | |
| 9,mlp.gate_proj,0.0002562806,0.05000,2.187 | |
| 9,mlp.up_proj,0.0001628826,0.05000,2.215 | |
| 9,mlp.down_proj,0.0000175799,0.05000,4.044 | |
| 10,self_attn.v_proj,0.0000116134,0.05000,4.708 | |
| 10,self_attn.q_proj,0.0000392499,0.05000,4.796 | |
| 10,self_attn.k_proj,0.0000101168,0.05000,4.825 | |
| 10,self_attn.o_proj,0.0000065144,0.05000,1.564 | |
| 10,mlp.gate_proj,0.0001958697,0.05000,2.119 | |
| 10,mlp.up_proj,0.0001360473,0.05000,2.145 | |
| 10,mlp.down_proj,0.0000145277,0.05000,4.086 | |
| 11,self_attn.q_proj,0.0000192507,0.05000,4.803 | |
| 11,self_attn.k_proj,0.0000052222,0.05000,4.898 | |
| 11,self_attn.v_proj,0.0000053556,0.05000,4.912 | |
| 11,self_attn.o_proj,0.0000041119,0.05000,1.587 | |
| 11,mlp.up_proj,0.0001261680,0.05000,2.254 | |
| 11,mlp.gate_proj,0.0001601069,0.05000,2.282 | |
| 11,mlp.down_proj,0.0000135020,0.05000,4.082 | |
| 12,self_attn.v_proj,0.0000064571,0.05000,4.739 | |
| 12,self_attn.k_proj,0.0000060496,0.05000,4.798 | |
| 12,self_attn.q_proj,0.0000229375,0.05000,4.825 | |
| 12,self_attn.o_proj,0.0000045234,0.05000,1.565 | |
| 12,mlp.gate_proj,0.0001424156,0.05000,2.162 | |
| 12,mlp.up_proj,0.0001237598,0.05000,2.199 | |
| 12,mlp.down_proj,0.0000132686,0.05000,4.109 | |
| 13,self_attn.v_proj,0.0000043770,0.05000,4.753 | |
| 13,self_attn.q_proj,0.0000172819,0.05000,4.763 | |
| 13,self_attn.k_proj,0.0000045546,0.05000,4.833 | |
| 13,self_attn.o_proj,0.0000035265,0.05000,1.567 | |
| 13,mlp.up_proj,0.0001254522,0.05000,2.206 | |
| 13,mlp.gate_proj,0.0001320071,0.05000,2.233 | |
| 13,mlp.down_proj,0.0000141580,0.05000,4.072 | |
| 14,self_attn.k_proj,0.0000074098,0.05000,4.615 | |
| 14,self_attn.v_proj,0.0000077529,0.05000,4.679 | |
| 14,self_attn.q_proj,0.0000286967,0.05000,4.749 | |
| 14,self_attn.o_proj,0.0000061177,0.05000,1.550 | |
| 14,mlp.gate_proj,0.0001299273,0.05000,2.266 | |
| 14,mlp.up_proj,0.0001247252,0.05000,2.294 | |
| 14,mlp.down_proj,0.0000125522,0.05000,4.003 | |
| 15,self_attn.k_proj,0.0000074629,0.05000,4.647 | |
| 15,self_attn.v_proj,0.0000070869,0.05000,4.706 | |
| 15,self_attn.q_proj,0.0000287500,0.05000,4.728 | |
| 15,self_attn.o_proj,0.0000053824,0.05000,1.569 | |
| 15,mlp.up_proj,0.0001194172,0.05000,2.237 | |
| 15,mlp.gate_proj,0.0001202382,0.05000,2.265 | |
| 15,mlp.down_proj,0.0000124121,0.05000,4.006 | |
| 16,self_attn.v_proj,0.0000138066,0.05000,4.700 | |
| 16,self_attn.q_proj,0.0000491788,0.05000,4.754 | |
| 16,self_attn.k_proj,0.0000121632,0.05000,4.776 | |
| 16,self_attn.o_proj,0.0000056487,0.05000,1.528 | |
| 16,mlp.gate_proj,0.0001317158,0.05000,2.152 | |
| 16,mlp.up_proj,0.0001234581,0.05000,2.181 | |
| 16,mlp.down_proj,0.0000731754,0.05000,3.985 | |
| 17,self_attn.v_proj,0.0000102906,0.05000,4.538 | |
| 17,self_attn.k_proj,0.0000092378,0.05000,4.595 | |
| 17,self_attn.q_proj,0.0000415329,0.05000,4.611 | |
| 17,self_attn.o_proj,0.0000056563,0.05000,1.586 | |
| 17,mlp.gate_proj,0.0001159002,0.05000,2.097 | |
| 17,mlp.up_proj,0.0001103001,0.05000,2.119 | |
| 17,mlp.down_proj,0.0000116598,0.05000,3.941 | |
| 18,self_attn.v_proj,0.0000126035,0.05000,4.529 | |
| 18,self_attn.k_proj,0.0000112181,0.05000,4.624 | |
| 18,self_attn.q_proj,0.0000459624,0.05000,4.645 | |
| 18,self_attn.o_proj,0.0000067008,0.05000,1.517 | |
| 18,mlp.up_proj,0.0001220867,0.05000,2.127 | |
| 18,mlp.gate_proj,0.0001259595,0.05000,2.151 | |
| 18,mlp.down_proj,0.0000148286,0.05000,3.990 | |
| 19,self_attn.v_proj,0.0000220715,0.05000,4.552 | |
| 19,self_attn.q_proj,0.0000918620,0.05000,4.646 | |
| 19,self_attn.k_proj,0.0000215181,0.05000,4.661 | |
| 19,self_attn.o_proj,0.0000093699,0.05000,1.525 | |
| 19,mlp.up_proj,0.0001385218,0.05000,2.151 | |
| 19,mlp.gate_proj,0.0001418180,0.05000,2.187 | |
| 19,mlp.down_proj,0.0000200673,0.05000,4.031 | |
| 20,self_attn.v_proj,0.0000205235,0.05000,4.611 | |
| 20,self_attn.k_proj,0.0000189061,0.05000,4.666 | |
| 20,self_attn.q_proj,0.0000867353,0.05000,4.707 | |
| 20,self_attn.o_proj,0.0000099226,0.05000,1.529 | |
| 20,mlp.gate_proj,0.0001534587,0.05000,2.104 | |
| 20,mlp.up_proj,0.0001536075,0.05000,2.127 | |
| 20,mlp.down_proj,0.0000229419,0.05000,4.057 | |
| 21,self_attn.q_proj,0.0001135650,0.05000,4.631 | |
| 21,self_attn.v_proj,0.0000283243,0.05000,4.673 | |
| 21,self_attn.k_proj,0.0000256526,0.05000,4.704 | |
| 21,self_attn.o_proj,0.0000121988,0.05000,1.516 | |
| 21,mlp.gate_proj,0.0001884211,0.05000,2.185 | |
| 21,mlp.up_proj,0.0001869941,0.05000,2.210 | |
| 21,mlp.down_proj,0.0000308657,0.05000,4.063 | |
| 22,self_attn.v_proj,0.0000583108,0.05000,4.585 | |
| 22,self_attn.k_proj,0.0000517917,0.05000,4.706 | |
| 22,self_attn.q_proj,0.0002313806,0.05000,4.729 | |
| 22,self_attn.o_proj,0.0000236768,0.05000,1.536 | |
| 22,mlp.gate_proj,0.0002260857,0.05000,2.162 | |
| 22,mlp.up_proj,0.0002135566,0.05000,2.166 | |
| 22,mlp.down_proj,0.0000494211,0.05000,4.036 | |
| 23,self_attn.q_proj,0.0002298174,0.05000,4.589 | |
| 23,self_attn.v_proj,0.0000604328,0.05000,4.642 | |
| 23,self_attn.k_proj,0.0000521171,0.05000,4.669 | |
| 23,self_attn.o_proj,0.0000286128,0.05000,1.542 | |
| 23,mlp.up_proj,0.0002542144,0.05000,2.091 | |
| 23,mlp.gate_proj,0.0002834950,0.05000,2.117 | |
| 23,mlp.down_proj,0.0000714043,0.05000,3.947 | |
| 24,self_attn.q_proj,0.0003790222,0.05000,4.545 | |
| 24,self_attn.k_proj,0.0000869552,0.05000,4.611 | |
| 24,self_attn.v_proj,0.0001125668,0.05000,4.636 | |
| 24,self_attn.o_proj,0.0000321837,0.05000,1.519 | |
| 24,mlp.gate_proj,0.0003256180,0.05000,2.167 | |
| 24,mlp.up_proj,0.0002876031,0.05000,2.184 | |
| 24,mlp.down_proj,0.0000910746,0.05000,4.001 | |
| 25,self_attn.v_proj,0.0000745178,0.05000,4.557 | |
| 25,self_attn.k_proj,0.0000651236,0.05000,4.615 | |
| 25,self_attn.q_proj,0.0002660532,0.05000,4.618 | |
| 25,self_attn.o_proj,0.0000205836,0.05000,1.570 | |
| 25,mlp.gate_proj,0.0003814693,0.05000,2.173 | |
| 25,mlp.up_proj,0.0003324611,0.05000,2.198 | |
| 25,mlp.down_proj,0.0001092646,0.05000,3.999 | |
| 26,self_attn.q_proj,0.0004336856,0.05000,4.503 | |
| 26,self_attn.k_proj,0.0001006330,0.05000,4.610 | |
| 26,self_attn.v_proj,0.0001189561,0.05000,4.639 | |
| 26,self_attn.o_proj,0.0000175224,0.05000,1.559 | |
| 26,mlp.gate_proj,0.0004622448,0.05000,2.077 | |
| 26,mlp.up_proj,0.0004160738,0.05000,2.105 | |
| 26,mlp.down_proj,0.0001339366,0.05000,4.006 | |
| 27,self_attn.q_proj,0.0005402514,0.05000,4.494 | |
| 27,self_attn.v_proj,0.0001606098,0.05000,4.572 | |
| 27,self_attn.k_proj,0.0001219699,0.05000,4.587 | |
| 27,self_attn.o_proj,0.0000265767,0.05000,1.516 | |
| 27,mlp.up_proj,0.0004859503,0.05000,2.107 | |
| 27,mlp.gate_proj,0.0005210892,0.05000,2.127 | |
| 27,mlp.down_proj,0.0001976302,0.05000,3.981 | |
| 28,self_attn.v_proj,0.0001727967,0.05000,4.584 | |
| 28,self_attn.q_proj,0.0006006611,0.05000,4.654 | |
| 28,self_attn.k_proj,0.0001497753,0.05000,4.672 | |
| 28,self_attn.o_proj,0.0000468070,0.05000,1.525 | |
| 28,mlp.gate_proj,0.0005994609,0.05000,2.076 | |
| 28,mlp.up_proj,0.0005836146,0.05000,2.096 | |
| 28,mlp.down_proj,0.0003001203,0.05000,4.036 | |
| 29,self_attn.q_proj,0.0014331605,0.05000,4.599 | |
| 29,self_attn.v_proj,0.0004406140,0.05000,4.658 | |
| 29,self_attn.k_proj,0.0003258389,0.05000,4.691 | |
| 29,self_attn.o_proj,0.0000410523,0.05000,1.552 | |
| 29,mlp.up_proj,0.0007288250,0.05000,2.198 | |
| 29,mlp.gate_proj,0.0007126805,0.05000,2.224 | |
| 29,mlp.down_proj,0.0003305236,0.05000,4.009 | |
| 30,self_attn.k_proj,0.0004136996,0.05000,4.614 | |
| 30,self_attn.v_proj,0.0005229475,0.05000,4.680 | |
| 30,self_attn.q_proj,0.0016422393,0.05000,4.692 | |
| 30,self_attn.o_proj,0.0000856037,0.05000,1.541 | |
| 30,mlp.up_proj,0.0007942698,0.05000,2.189 | |
| 30,mlp.gate_proj,0.0007508333,0.05000,2.222 | |
| 30,mlp.down_proj,0.0004645914,0.05000,4.039 | |
| 31,self_attn.k_proj,0.0005456899,0.05000,4.581 | |
| 31,self_attn.q_proj,0.0020693638,0.05000,4.670 | |
| 31,self_attn.v_proj,0.0007509262,0.05000,4.698 | |
| 31,self_attn.o_proj,0.0000844962,0.05000,1.550 | |
| 31,mlp.gate_proj,0.0007564633,0.05000,2.158 | |
| 31,mlp.up_proj,0.0008417759,0.05000,2.180 | |
| 31,mlp.down_proj,0.0005862378,0.05000,4.025 | |
| 32,self_attn.k_proj,0.0007358906,0.05000,4.489 | |
| 32,self_attn.v_proj,0.0011125570,0.05000,4.616 | |
| 32,self_attn.q_proj,0.0030247277,0.05000,4.656 | |
| 32,self_attn.o_proj,0.0001102863,0.05000,1.526 | |
| 32,mlp.gate_proj,0.0007851195,0.05000,2.050 | |
| 32,mlp.up_proj,0.0008939442,0.05000,2.082 | |
| 32,mlp.down_proj,0.0007211865,0.05000,4.016 | |
| 33,self_attn.k_proj,0.0012740580,0.05000,4.581 | |
| 33,self_attn.q_proj,0.0062236731,0.05000,4.623 | |
| 33,self_attn.v_proj,0.0023895111,0.05000,4.652 | |
| 33,self_attn.o_proj,0.0001302601,0.05000,1.533 | |
| 33,mlp.gate_proj,0.0008421095,0.05000,2.077 | |
| 33,mlp.up_proj,0.0009716948,0.05000,2.091 | |
| 33,mlp.down_proj,0.0009164421,0.05000,4.036 | |
| 34,self_attn.k_proj,0.0010975073,0.05000,4.594 | |
| 34,self_attn.v_proj,0.0018084718,0.05000,4.686 | |
| 34,self_attn.q_proj,0.0049361077,0.05000,4.708 | |
| 34,self_attn.o_proj,0.0002369204,0.05000,1.572 | |
| 34,mlp.up_proj,0.0010833157,0.05000,2.200 | |
| 34,mlp.gate_proj,0.0010048764,0.05000,2.219 | |
| 34,mlp.down_proj,0.0010852498,0.05000,4.017 | |
| 35,self_attn.k_proj,0.0006271339,0.05000,4.627 | |
| 35,self_attn.q_proj,0.0024465988,0.05000,4.708 | |
| 35,self_attn.v_proj,0.0008356234,0.05000,4.728 | |
| 35,self_attn.o_proj,0.0003305586,0.05000,1.565 | |
| 35,mlp.gate_proj,0.0016273851,0.05000,2.216 | |
| 35,mlp.up_proj,0.0016999776,0.05000,2.243 | |
| 35,mlp.down_proj,0.0020161150,0.05000,4.022 | |