| layer,module,loss,samples,damp,time | |
| 0,self_attn.q_proj,0.0002747226,0.05000,1.465 | |
| 0,self_attn.v_proj,0.0000084039,0.05000,1.481 | |
| 0,self_attn.k_proj,0.0000548816,0.05000,1.507 | |
| 0,self_attn.o_proj,0.0000240608,0.05000,0.478 | |
| 0,mlp.gate_proj,0.0003001237,0.05000,0.712 | |
| 0,mlp.up_proj,0.0002388472,0.05000,0.717 | |
| 0,mlp.down_proj,0.0000351643,0.05000,2.683 | |
| 1,self_attn.v_proj,0.0000026640,0.05000,1.277 | |
| 1,self_attn.k_proj,0.0000132014,0.05000,1.285 | |
| 1,self_attn.q_proj,0.0000481469,0.05000,1.313 | |
| 1,self_attn.o_proj,0.0000085475,0.05000,0.436 | |
| 1,mlp.gate_proj,0.0200391563,0.05000,0.951 | |
| 1,mlp.up_proj,0.0155865240,0.05000,0.960 | |
| 1,mlp.down_proj,0.0000041996,0.05000,2.711 | |
| 2,self_attn.q_proj,0.0000682623,0.05000,1.249 | |
| 2,self_attn.k_proj,0.0000152074,0.05000,1.276 | |
| 2,self_attn.v_proj,0.0000038336,0.05000,1.299 | |
| 2,self_attn.o_proj,0.0000055928,0.05000,0.440 | |
| 2,mlp.gate_proj,0.0046740029,0.05000,0.720 | |
| 2,mlp.up_proj,0.0049174396,0.05000,0.729 | |
| 2,mlp.down_proj,0.0046229527,0.05000,2.713 | |
| 3,self_attn.q_proj,0.0002987919,0.05000,1.048 | |
| 3,self_attn.v_proj,0.0000140483,0.05000,1.056 | |
| 3,self_attn.k_proj,0.0000659558,0.05000,1.067 | |
| 3,self_attn.o_proj,0.0000085583,0.05000,0.439 | |
| 3,mlp.gate_proj,0.0057774407,0.05000,0.737 | |
| 3,mlp.up_proj,0.0051700106,0.05000,0.740 | |
| 3,mlp.down_proj,0.0000788252,0.05000,2.857 | |
| 4,self_attn.v_proj,0.0000172816,0.05000,1.203 | |
| 4,self_attn.k_proj,0.0000538719,0.05000,1.213 | |
| 4,self_attn.q_proj,0.0002588518,0.05000,1.235 | |
| 4,self_attn.o_proj,0.0000168278,0.05000,0.433 | |
| 4,mlp.up_proj,0.0043241808,0.05000,0.942 | |
| 4,mlp.gate_proj,0.0055869601,0.05000,0.948 | |
| 4,mlp.down_proj,0.0000430156,0.05000,2.642 | |
| 5,self_attn.k_proj,0.0000881785,0.05000,1.192 | |
| 5,self_attn.v_proj,0.0000381224,0.05000,1.215 | |
| 5,self_attn.q_proj,0.0004759409,0.05000,1.223 | |
| 5,self_attn.o_proj,0.0000100381,0.05000,0.437 | |
| 5,mlp.gate_proj,0.0076938863,0.05000,0.935 | |
| 5,mlp.up_proj,0.0066256830,0.05000,0.938 | |
| 5,mlp.down_proj,0.0000689371,0.05000,2.609 | |
| 6,self_attn.k_proj,0.0000567696,0.05000,1.274 | |
| 6,self_attn.v_proj,0.0000389104,0.05000,1.293 | |
| 6,self_attn.q_proj,0.0003420001,0.05000,1.301 | |
| 6,self_attn.o_proj,0.0000270948,0.05000,0.452 | |
| 6,mlp.gate_proj,0.0093043704,0.05000,0.702 | |
| 6,mlp.up_proj,0.0078126297,0.05000,0.709 | |
| 6,mlp.down_proj,0.0000942404,0.05000,2.560 | |
| 7,self_attn.q_proj,0.0004316112,0.05000,1.280 | |
| 7,self_attn.k_proj,0.0000753655,0.05000,1.290 | |
| 7,self_attn.v_proj,0.0000414025,0.05000,1.322 | |
| 7,self_attn.o_proj,0.0000166787,0.05000,0.451 | |
| 7,mlp.up_proj,0.0072872794,0.05000,0.719 | |
| 7,mlp.gate_proj,0.0087963101,0.05000,0.723 | |
| 7,mlp.down_proj,0.0001376350,0.05000,2.620 | |
| 8,self_attn.q_proj,0.0004245544,0.05000,1.229 | |
| 8,self_attn.v_proj,0.0000451166,0.05000,1.255 | |
| 8,self_attn.k_proj,0.0000665214,0.05000,1.281 | |
| 8,self_attn.o_proj,0.0000254294,0.05000,0.442 | |
| 8,mlp.gate_proj,0.0092212750,0.05000,0.939 | |
| 8,mlp.up_proj,0.0067839634,0.05000,0.942 | |
| 8,mlp.down_proj,0.0001715169,0.05000,2.676 | |
| 9,self_attn.k_proj,0.0000700954,0.05000,1.225 | |
| 9,self_attn.v_proj,0.0000408858,0.05000,1.232 | |
| 9,self_attn.q_proj,0.0003906921,0.05000,1.263 | |
| 9,self_attn.o_proj,0.0000408029,0.05000,0.444 | |
| 9,mlp.up_proj,0.0057729447,0.05000,0.829 | |
| 9,mlp.gate_proj,0.0083765922,0.05000,0.830 | |
| 9,mlp.down_proj,0.0001908187,0.05000,2.612 | |
| 10,self_attn.q_proj,0.0005441840,0.05000,1.265 | |
| 10,self_attn.k_proj,0.0000848859,0.05000,1.293 | |
| 10,self_attn.v_proj,0.0000628567,0.05000,1.302 | |
| 10,self_attn.o_proj,0.0000383119,0.05000,0.452 | |
| 10,mlp.gate_proj,0.0060825786,0.05000,0.837 | |
| 10,mlp.up_proj,0.0040899490,0.05000,0.842 | |
| 10,mlp.down_proj,0.0002219601,0.05000,2.567 | |
| 11,self_attn.q_proj,0.0003964540,0.05000,1.292 | |
| 11,self_attn.k_proj,0.0000630133,0.05000,1.299 | |
| 11,self_attn.v_proj,0.0000498622,0.05000,1.325 | |
| 11,self_attn.o_proj,0.0000546958,0.05000,0.435 | |
| 11,mlp.up_proj,0.0023352402,0.05000,0.803 | |
| 11,mlp.gate_proj,0.0031954204,0.05000,0.808 | |
| 11,mlp.down_proj,0.0002431936,0.05000,2.605 | |
| 12,self_attn.v_proj,0.0000462074,0.05000,1.295 | |
| 12,self_attn.k_proj,0.0000678828,0.05000,1.317 | |
| 12,self_attn.q_proj,0.0004207567,0.05000,1.331 | |
| 12,self_attn.o_proj,0.0000542564,0.05000,0.464 | |
| 12,mlp.up_proj,0.0025469633,0.05000,0.953 | |
| 12,mlp.gate_proj,0.0035938459,0.05000,0.959 | |
| 12,mlp.down_proj,0.0002270444,0.05000,2.567 | |
| 13,self_attn.v_proj,0.0000348742,0.05000,1.205 | |
| 13,self_attn.q_proj,0.0004997598,0.05000,1.238 | |
| 13,self_attn.k_proj,0.0000908744,0.05000,1.253 | |
| 13,self_attn.o_proj,0.0000351865,0.05000,0.449 | |
| 13,mlp.gate_proj,0.0019457909,0.05000,0.776 | |
| 13,mlp.up_proj,0.0018662845,0.05000,0.784 | |
| 13,mlp.down_proj,0.0001949961,0.05000,2.744 | |
| 14,self_attn.k_proj,0.0000637346,0.05000,1.221 | |
| 14,self_attn.v_proj,0.0000362549,0.05000,1.238 | |
| 14,self_attn.q_proj,0.0003836463,0.05000,1.266 | |
| 14,self_attn.o_proj,0.0000767360,0.05000,0.446 | |
| 14,mlp.up_proj,0.0017634257,0.05000,0.954 | |
| 14,mlp.gate_proj,0.0018520612,0.05000,0.959 | |
| 14,mlp.down_proj,0.0001823228,0.05000,2.635 | |
| 15,self_attn.k_proj,0.0000624404,0.05000,1.271 | |
| 15,self_attn.q_proj,0.0003831870,0.05000,1.277 | |
| 15,self_attn.v_proj,0.0000364707,0.05000,1.304 | |
| 15,self_attn.o_proj,0.0000753158,0.05000,0.433 | |
| 15,mlp.up_proj,0.0015657423,0.05000,0.810 | |
| 15,mlp.gate_proj,0.0015058519,0.05000,0.814 | |
| 15,mlp.down_proj,0.0001640118,0.05000,2.627 | |
| 16,self_attn.k_proj,0.0000659427,0.05000,1.392 | |
| 16,self_attn.q_proj,0.0003814801,0.05000,1.402 | |
| 16,self_attn.v_proj,0.0000370104,0.05000,1.408 | |
| 16,self_attn.o_proj,0.0000775972,0.05000,0.454 | |
| 16,mlp.up_proj,0.0015405270,0.05000,0.958 | |
| 16,mlp.gate_proj,0.0015773342,0.05000,0.976 | |
| 16,mlp.down_proj,0.0001440141,0.05000,2.691 | |
| 17,self_attn.k_proj,0.0001108535,0.05000,1.435 | |
| 17,self_attn.q_proj,0.0006258364,0.05000,1.435 | |
| 17,self_attn.v_proj,0.0000534320,0.05000,1.444 | |
| 17,self_attn.o_proj,0.0000479042,0.05000,0.432 | |
| 17,mlp.up_proj,0.0013993403,0.05000,0.941 | |
| 17,mlp.gate_proj,0.0013784876,0.05000,0.942 | |
| 17,mlp.down_proj,0.0001402667,0.05000,2.566 | |
| 18,self_attn.v_proj,0.0000375932,0.05000,1.408 | |
| 18,self_attn.k_proj,0.0000597105,0.05000,1.420 | |
| 18,self_attn.q_proj,0.0003847960,0.05000,1.429 | |
| 18,self_attn.o_proj,0.0000639632,0.05000,0.448 | |
| 18,mlp.gate_proj,0.0014018252,0.05000,0.949 | |
| 18,mlp.up_proj,0.0013598988,0.05000,0.954 | |
| 18,mlp.down_proj,0.0001361705,0.05000,2.647 | |
| 19,self_attn.k_proj,0.0000805691,0.05000,1.253 | |
| 19,self_attn.v_proj,0.0000390765,0.05000,1.271 | |
| 19,self_attn.q_proj,0.0004623365,0.05000,1.301 | |
| 19,self_attn.o_proj,0.0000811320,0.05000,0.444 | |
| 19,mlp.up_proj,0.0013748722,0.05000,0.825 | |
| 19,mlp.gate_proj,0.0013197439,0.05000,0.829 | |
| 19,mlp.down_proj,0.0001283883,0.05000,2.622 | |
| 20,self_attn.q_proj,0.0006775008,0.05000,1.225 | |
| 20,self_attn.v_proj,0.0000753612,0.05000,1.241 | |
| 20,self_attn.k_proj,0.0000815596,0.05000,1.272 | |
| 20,self_attn.o_proj,0.0000648474,0.05000,0.447 | |
| 20,mlp.gate_proj,0.0013668196,0.05000,0.938 | |
| 20,mlp.up_proj,0.0013983431,0.05000,0.948 | |
| 20,mlp.down_proj,0.0001474679,0.05000,2.651 | |
| 21,self_attn.v_proj,0.0000508111,0.05000,1.314 | |
| 21,self_attn.q_proj,0.0004851653,0.05000,1.327 | |
| 21,self_attn.k_proj,0.0000785037,0.05000,1.354 | |
| 21,self_attn.o_proj,0.0000685299,0.05000,0.438 | |
| 21,mlp.up_proj,0.0013653357,0.05000,0.944 | |
| 21,mlp.gate_proj,0.0014496203,0.05000,0.949 | |
| 21,mlp.down_proj,0.0001247860,0.05000,2.537 | |
| 22,self_attn.q_proj,0.0005651744,0.05000,1.239 | |
| 22,self_attn.v_proj,0.0000697534,0.05000,1.258 | |
| 22,self_attn.k_proj,0.0000869150,0.05000,1.286 | |
| 22,self_attn.o_proj,0.0000606621,0.05000,0.433 | |
| 22,mlp.gate_proj,0.0014293856,0.05000,0.712 | |
| 22,mlp.up_proj,0.0014016083,0.05000,0.713 | |
| 22,mlp.down_proj,0.0001302481,0.05000,2.710 | |
| 23,self_attn.v_proj,0.0000417556,0.05000,1.330 | |
| 23,self_attn.k_proj,0.0000860245,0.05000,1.337 | |
| 23,self_attn.q_proj,0.0005656183,0.05000,1.362 | |
| 23,self_attn.o_proj,0.0000591576,0.05000,0.425 | |
| 23,mlp.up_proj,0.0014252818,0.05000,0.936 | |
| 23,mlp.gate_proj,0.0013683091,0.05000,0.940 | |
| 23,mlp.down_proj,0.0001475577,0.05000,2.642 | |
| 24,self_attn.k_proj,0.0000600232,0.05000,1.390 | |
| 24,self_attn.q_proj,0.0004127045,0.05000,1.401 | |
| 24,self_attn.v_proj,0.0000478212,0.05000,1.408 | |
| 24,self_attn.o_proj,0.0000761375,0.05000,0.436 | |
| 24,mlp.up_proj,0.0013949384,0.05000,0.916 | |
| 24,mlp.gate_proj,0.0013531580,0.05000,0.917 | |
| 24,mlp.down_proj,0.0001648138,0.05000,2.615 | |
| 25,self_attn.q_proj,0.0005688818,0.05000,1.224 | |
| 25,self_attn.k_proj,0.0000611430,0.05000,1.229 | |
| 25,self_attn.v_proj,0.0000785096,0.05000,1.257 | |
| 25,self_attn.o_proj,0.0000547303,0.05000,0.440 | |
| 25,mlp.gate_proj,0.0015417385,0.05000,0.869 | |
| 25,mlp.up_proj,0.0015539572,0.05000,0.875 | |
| 25,mlp.down_proj,0.0002121821,0.05000,2.650 | |
| 26,self_attn.q_proj,0.0004628822,0.05000,1.317 | |
| 26,self_attn.k_proj,0.0000623467,0.05000,1.326 | |
| 26,self_attn.v_proj,0.0000757300,0.05000,1.352 | |
| 26,self_attn.o_proj,0.0000830591,0.05000,0.454 | |
| 26,mlp.up_proj,0.0017814280,0.05000,0.862 | |
| 26,mlp.gate_proj,0.0016955417,0.05000,0.866 | |
| 26,mlp.down_proj,0.0002887561,0.05000,2.571 | |
| 27,self_attn.q_proj,0.0007779023,0.05000,1.206 | |
| 27,self_attn.k_proj,0.0000669245,0.05000,1.211 | |
| 27,self_attn.v_proj,0.0001101871,0.05000,1.227 | |
| 27,self_attn.o_proj,0.0001293488,0.05000,0.434 | |
| 27,mlp.up_proj,0.0019002785,0.05000,0.948 | |
| 27,mlp.gate_proj,0.0018908290,0.05000,0.953 | |
| 27,mlp.down_proj,0.0003484332,0.05000,2.648 | |
| 28,self_attn.q_proj,0.0005814589,0.05000,1.416 | |
| 28,self_attn.k_proj,0.0000659744,0.05000,1.428 | |
| 28,self_attn.v_proj,0.0001026726,0.05000,1.440 | |
| 28,self_attn.o_proj,0.0001264897,0.05000,0.440 | |
| 28,mlp.gate_proj,0.0021317722,0.05000,0.947 | |
| 28,mlp.up_proj,0.0021244350,0.05000,0.947 | |
| 28,mlp.down_proj,0.0003927946,0.05000,2.611 | |
| 29,self_attn.k_proj,0.0000676989,0.05000,1.290 | |
| 29,self_attn.q_proj,0.0005627562,0.05000,1.295 | |
| 29,self_attn.v_proj,0.0000897833,0.05000,1.325 | |
| 29,self_attn.o_proj,0.0000908731,0.05000,0.433 | |
| 29,mlp.up_proj,0.0025176517,0.05000,0.769 | |
| 29,mlp.gate_proj,0.0024979140,0.05000,0.776 | |
| 29,mlp.down_proj,0.0005486268,0.05000,2.560 | |
| 30,self_attn.q_proj,0.0007684063,0.05000,1.438 | |
| 30,self_attn.k_proj,0.0000728211,0.05000,1.447 | |
| 30,self_attn.v_proj,0.0002139594,0.05000,1.449 | |
| 30,self_attn.o_proj,0.0000971499,0.05000,0.450 | |
| 30,mlp.gate_proj,0.0031557634,0.05000,0.920 | |
| 30,mlp.up_proj,0.0033465312,0.05000,0.928 | |
| 30,mlp.down_proj,0.0075672583,0.05000,2.609 | |
| 31,self_attn.k_proj,0.0000989171,0.05000,1.305 | |
| 31,self_attn.v_proj,0.0002433076,0.05000,1.312 | |
| 31,self_attn.q_proj,0.0009695329,0.05000,1.337 | |
| 31,self_attn.o_proj,0.0002389346,0.05000,0.426 | |
| 31,mlp.gate_proj,0.0035235909,0.05000,0.964 | |
| 31,mlp.up_proj,0.0039159792,0.05000,0.970 | |
| 31,mlp.down_proj,0.0016639339,0.05000,2.541 | |
| 32,self_attn.k_proj,0.0001418183,0.05000,1.133 | |
| 32,self_attn.v_proj,0.0006315781,0.05000,1.142 | |
| 32,self_attn.q_proj,0.0015098560,0.05000,1.166 | |
| 32,self_attn.o_proj,0.0002902363,0.05000,0.430 | |
| 32,mlp.up_proj,0.0039817701,0.05000,0.963 | |
| 32,mlp.gate_proj,0.0035569574,0.05000,0.974 | |
| 32,mlp.down_proj,0.0015996505,0.05000,2.623 | |
| 33,self_attn.v_proj,0.0011238059,0.05000,1.317 | |
| 33,self_attn.k_proj,0.0001341973,0.05000,1.330 | |
| 33,self_attn.q_proj,0.0015063443,0.05000,1.355 | |
| 33,self_attn.o_proj,0.0003465505,0.05000,0.429 | |
| 33,mlp.gate_proj,0.0033355778,0.05000,0.955 | |
| 33,mlp.up_proj,0.0041252049,0.05000,0.965 | |
| 33,mlp.down_proj,0.0032116803,0.05000,2.632 | |
| 34,self_attn.k_proj,0.0000561139,0.05000,1.279 | |
| 34,self_attn.v_proj,0.0001796343,0.05000,1.286 | |
| 34,self_attn.q_proj,0.0005739081,0.05000,1.320 | |
| 34,self_attn.o_proj,0.0002795816,0.05000,0.443 | |
| 34,mlp.up_proj,0.0042357320,0.05000,0.884 | |
| 34,mlp.gate_proj,0.0036401345,0.05000,0.887 | |
| 34,mlp.down_proj,0.0023706419,0.05000,2.606 | |
| 35,self_attn.v_proj,0.0001593449,0.05000,1.279 | |
| 35,self_attn.k_proj,0.0000554864,0.05000,1.290 | |
| 35,self_attn.q_proj,0.0005304807,0.05000,1.314 | |
| 35,self_attn.o_proj,0.0002627382,0.05000,0.438 | |
| 35,mlp.gate_proj,0.0052445997,0.05000,0.849 | |
| 35,mlp.up_proj,0.0057744002,0.05000,0.854 | |
| 35,mlp.down_proj,0.0144989756,0.05000,2.584 | |