| layer,module,loss,samples,damp,time | |
| 0,self_attn.v_proj,0.0000000010,0.01000,1.829 | |
| 0,self_attn.q_proj,0.0000000062,0.01000,1.845 | |
| 0,self_attn.k_proj,0.0000000016,0.01000,1.847 | |
| 0,self_attn.o_proj,0.0000000027,0.01000,0.577 | |
| 0,mlp.gate_proj,0.0000005435,0.01000,1.155 | |
| 0,mlp.up_proj,0.0000004846,0.01000,1.162 | |
| 0,mlp.down_proj,0.0000002644,0.01000,2.252 | |
| 1,self_attn.v_proj,0.0000000130,0.01000,1.642 | |
| 1,self_attn.q_proj,0.0000000625,0.01000,1.648 | |
| 1,self_attn.k_proj,0.0000000153,0.01000,1.652 | |
| 1,self_attn.o_proj,0.0000000898,0.01000,0.568 | |
| 1,mlp.up_proj,0.0000072422,0.01000,1.143 | |
| 1,mlp.gate_proj,0.0000082405,0.01000,1.145 | |
| 1,mlp.down_proj,0.0000001166,0.01000,2.235 | |
| 2,self_attn.q_proj,0.0000000749,0.01000,1.535 | |
| 2,self_attn.v_proj,0.0000000177,0.01000,1.563 | |
| 2,self_attn.k_proj,0.0000000193,0.01000,1.566 | |
| 2,self_attn.o_proj,0.0000000921,0.01000,0.570 | |
| 2,mlp.up_proj,0.0000127681,0.01000,1.150 | |
| 2,mlp.gate_proj,0.0000122131,0.01000,1.151 | |
| 2,mlp.down_proj,0.0000003517,0.01000,2.241 | |
| 3,self_attn.v_proj,0.0000000529,0.01000,1.574 | |
| 3,self_attn.k_proj,0.0000000404,0.01000,1.573 | |
| 3,self_attn.q_proj,0.0000001661,0.01000,1.543 | |
| 3,self_attn.o_proj,0.0000002439,0.01000,0.574 | |
| 3,mlp.up_proj,0.0000191999,0.01000,1.128 | |
| 3,mlp.gate_proj,0.0000186392,0.01000,1.133 | |
| 3,mlp.down_proj,0.0000030536,0.01000,2.273 | |
| 4,self_attn.v_proj,0.0000000613,0.01000,1.539 | |
| 4,self_attn.q_proj,0.0000002399,0.01000,1.545 | |
| 4,self_attn.k_proj,0.0000000552,0.01000,1.554 | |
| 4,self_attn.o_proj,0.0000003474,0.01000,0.573 | |
| 4,mlp.gate_proj,0.0000143707,0.01000,1.140 | |
| 4,mlp.up_proj,0.0000126579,0.01000,1.146 | |
| 4,mlp.down_proj,0.0000051150,0.01000,2.245 | |
| 5,self_attn.v_proj,0.0000000678,0.01000,1.541 | |
| 5,self_attn.q_proj,0.0000002662,0.01000,1.547 | |
| 5,self_attn.k_proj,0.0000000669,0.01000,1.552 | |
| 5,self_attn.o_proj,0.0000001778,0.01000,0.571 | |
| 5,mlp.gate_proj,0.0000023585,0.01000,1.141 | |
| 5,mlp.up_proj,0.0000018802,0.01000,1.149 | |
| 5,mlp.down_proj,0.0000004271,0.01000,2.224 | |
| 6,self_attn.q_proj,0.0000002649,0.01000,1.560 | |
| 6,self_attn.k_proj,0.0000000600,0.01000,1.563 | |
| 6,self_attn.v_proj,0.0000000583,0.01000,1.566 | |
| 6,self_attn.o_proj,0.0000001879,0.01000,0.583 | |
| 6,mlp.gate_proj,0.0000024982,0.01000,1.135 | |
| 6,mlp.up_proj,0.0000023775,0.01000,1.138 | |
| 6,mlp.down_proj,0.0000003491,0.01000,2.253 | |
| 7,self_attn.v_proj,0.0000000447,0.01000,1.562 | |
| 7,self_attn.k_proj,0.0000000380,0.01000,1.567 | |
| 7,self_attn.q_proj,0.0000001416,0.01000,1.576 | |
| 7,self_attn.o_proj,0.0000002175,0.01000,0.576 | |
| 7,mlp.up_proj,0.0000024159,0.01000,1.150 | |
| 7,mlp.gate_proj,0.0000028353,0.01000,1.154 | |
| 7,mlp.down_proj,0.0000003850,0.01000,2.244 | |
| 8,self_attn.q_proj,0.0000001304,0.01000,1.590 | |
| 8,self_attn.v_proj,0.0000000394,0.01000,1.592 | |
| 8,self_attn.k_proj,0.0000000326,0.01000,1.604 | |
| 8,self_attn.o_proj,0.0000001448,0.01000,0.575 | |
| 8,mlp.gate_proj,0.0000018485,0.01000,1.137 | |
| 8,mlp.up_proj,0.0000019048,0.01000,1.141 | |
| 8,mlp.down_proj,0.0000003137,0.01000,2.253 | |
| 9,self_attn.v_proj,0.0000000571,0.01000,1.579 | |
| 9,self_attn.q_proj,0.0000001757,0.01000,1.582 | |
| 9,self_attn.k_proj,0.0000000423,0.01000,1.585 | |
| 9,self_attn.o_proj,0.0000002859,0.01000,0.584 | |
| 9,mlp.gate_proj,0.0000019399,0.01000,1.186 | |
| 9,mlp.up_proj,0.0000018338,0.01000,1.193 | |
| 9,mlp.down_proj,0.0000003901,0.01000,2.240 | |
| 10,self_attn.v_proj,0.0000000418,0.01000,1.544 | |
| 10,self_attn.k_proj,0.0000000374,0.01000,1.549 | |
| 10,self_attn.q_proj,0.0000001466,0.01000,1.563 | |
| 10,self_attn.o_proj,0.0000003586,0.01000,0.568 | |
| 10,mlp.up_proj,0.0000012885,0.01000,1.135 | |
| 10,mlp.gate_proj,0.0000015318,0.01000,1.140 | |
| 10,mlp.down_proj,0.0000001714,0.01000,2.251 | |
| 11,self_attn.q_proj,0.0000000464,0.01000,1.541 | |
| 11,self_attn.v_proj,0.0000000131,0.01000,1.542 | |
| 11,self_attn.k_proj,0.0000000114,0.01000,1.546 | |
| 11,self_attn.o_proj,0.0000001012,0.01000,0.568 | |
| 11,mlp.gate_proj,0.0000006189,0.01000,1.147 | |
| 11,mlp.up_proj,0.0000005842,0.01000,1.161 | |
| 11,mlp.down_proj,0.0000001167,0.01000,2.239 | |
| 12,self_attn.q_proj,0.0000000243,0.01000,1.589 | |
| 12,self_attn.v_proj,0.0000000072,0.01000,1.600 | |
| 12,self_attn.k_proj,0.0000000058,0.01000,1.604 | |
| 12,self_attn.o_proj,0.0000000608,0.01000,0.578 | |
| 12,mlp.gate_proj,0.0000005512,0.01000,1.144 | |
| 12,mlp.up_proj,0.0000005652,0.01000,1.152 | |
| 12,mlp.down_proj,0.0000000815,0.01000,2.234 | |
| 13,self_attn.q_proj,0.0000000319,0.01000,1.542 | |
| 13,self_attn.v_proj,0.0000000089,0.01000,1.546 | |
| 13,self_attn.k_proj,0.0000000074,0.01000,1.549 | |
| 13,self_attn.o_proj,0.0000000500,0.01000,0.566 | |
| 13,mlp.gate_proj,0.0000004800,0.01000,1.142 | |
| 13,mlp.up_proj,0.0000004858,0.01000,1.147 | |
| 13,mlp.down_proj,0.0000000940,0.01000,2.240 | |
| 14,self_attn.k_proj,0.0000000096,0.01000,1.534 | |
| 14,self_attn.q_proj,0.0000000405,0.01000,1.539 | |
| 14,self_attn.v_proj,0.0000000106,0.01000,1.544 | |
| 14,self_attn.o_proj,0.0000000484,0.01000,0.575 | |
| 14,mlp.up_proj,0.0000005409,0.01000,1.146 | |
| 14,mlp.gate_proj,0.0000005316,0.01000,1.150 | |
| 14,mlp.down_proj,0.0000000756,0.01000,2.261 | |
| 15,self_attn.k_proj,0.0000000103,0.01000,1.544 | |
| 15,self_attn.q_proj,0.0000000466,0.01000,1.564 | |
| 15,self_attn.v_proj,0.0000000119,0.01000,1.567 | |
| 15,self_attn.o_proj,0.0000000454,0.01000,0.566 | |
| 15,mlp.up_proj,0.0000005134,0.01000,1.148 | |
| 15,mlp.gate_proj,0.0000005055,0.01000,1.153 | |
| 15,mlp.down_proj,0.0000001016,0.01000,2.260 | |
| 16,self_attn.k_proj,0.0000000187,0.01000,1.558 | |
| 16,self_attn.q_proj,0.0000000835,0.01000,1.570 | |
| 16,self_attn.v_proj,0.0000000236,0.01000,1.574 | |
| 16,self_attn.o_proj,0.0000000851,0.01000,0.576 | |
| 16,mlp.gate_proj,0.0000006101,0.01000,1.139 | |
| 16,mlp.up_proj,0.0000006160,0.01000,1.142 | |
| 16,mlp.down_proj,0.0000004110,0.01000,2.231 | |
| 17,self_attn.k_proj,0.0000000142,0.01000,1.549 | |
| 17,self_attn.v_proj,0.0000000171,0.01000,1.560 | |
| 17,self_attn.q_proj,0.0000000693,0.01000,1.566 | |
| 17,self_attn.o_proj,0.0000000767,0.01000,0.569 | |
| 17,mlp.gate_proj,0.0000005645,0.01000,1.157 | |
| 17,mlp.up_proj,0.0000005859,0.01000,1.164 | |
| 17,mlp.down_proj,0.0000000809,0.01000,2.231 | |
| 18,self_attn.q_proj,0.0000001014,0.01000,1.529 | |
| 18,self_attn.v_proj,0.0000000304,0.01000,1.539 | |
| 18,self_attn.k_proj,0.0000000222,0.01000,1.541 | |
| 18,self_attn.o_proj,0.0000000898,0.01000,0.566 | |
| 18,mlp.gate_proj,0.0000004769,0.01000,1.132 | |
| 18,mlp.up_proj,0.0000005321,0.01000,1.137 | |
| 18,mlp.down_proj,0.0000001979,0.01000,2.247 | |
| 19,self_attn.q_proj,0.0000001481,0.01000,1.541 | |
| 19,self_attn.k_proj,0.0000000314,0.01000,1.548 | |
| 19,self_attn.v_proj,0.0000000349,0.01000,1.551 | |
| 19,self_attn.o_proj,0.0000000803,0.01000,0.573 | |
| 19,mlp.gate_proj,0.0000004752,0.01000,1.147 | |
| 19,mlp.up_proj,0.0000004978,0.01000,1.149 | |
| 19,mlp.down_proj,0.0000000639,0.01000,2.242 | |
| 20,self_attn.q_proj,0.0000001669,0.01000,1.539 | |
| 20,self_attn.k_proj,0.0000000367,0.01000,1.551 | |
| 20,self_attn.v_proj,0.0000000457,0.01000,1.553 | |
| 20,self_attn.o_proj,0.0000000969,0.01000,0.567 | |
| 20,mlp.up_proj,0.0000004723,0.01000,1.139 | |
| 20,mlp.gate_proj,0.0000004650,0.01000,1.143 | |
| 20,mlp.down_proj,0.0000000665,0.01000,2.265 | |
| 21,self_attn.q_proj,0.0000001604,0.01000,1.552 | |
| 21,self_attn.v_proj,0.0000000494,0.01000,1.559 | |
| 21,self_attn.k_proj,0.0000000325,0.01000,1.566 | |
| 21,self_attn.o_proj,0.0000000864,0.01000,0.566 | |
| 21,mlp.gate_proj,0.0000004808,0.01000,1.142 | |
| 21,mlp.up_proj,0.0000005138,0.01000,1.148 | |
| 21,mlp.down_proj,0.0000002167,0.01000,2.243 | |
| 22,self_attn.q_proj,0.0000005798,0.01000,1.572 | |
| 22,self_attn.k_proj,0.0000001053,0.01000,1.580 | |
| 22,self_attn.v_proj,0.0000001031,0.01000,1.582 | |
| 22,self_attn.o_proj,0.0000001522,0.01000,0.563 | |
| 22,mlp.up_proj,0.0000006384,0.01000,1.142 | |
| 22,mlp.gate_proj,0.0000005433,0.01000,1.146 | |
| 22,mlp.down_proj,0.0000000918,0.01000,2.259 | |
| 23,self_attn.q_proj,0.0000006371,0.01000,1.548 | |
| 23,self_attn.k_proj,0.0000001273,0.01000,1.554 | |
| 23,self_attn.v_proj,0.0000001431,0.01000,1.557 | |
| 23,self_attn.o_proj,0.0000002528,0.01000,0.573 | |
| 23,mlp.gate_proj,0.0000006702,0.01000,1.130 | |
| 23,mlp.up_proj,0.0000007339,0.01000,1.137 | |
| 23,mlp.down_proj,0.0000000830,0.01000,2.247 | |
| 24,self_attn.k_proj,0.0000001151,0.01000,1.554 | |
| 24,self_attn.v_proj,0.0000001562,0.01000,1.558 | |
| 24,self_attn.q_proj,0.0000008313,0.01000,1.564 | |
| 24,self_attn.o_proj,0.0000002746,0.01000,0.572 | |
| 24,mlp.up_proj,0.0000009086,0.01000,1.142 | |
| 24,mlp.gate_proj,0.0000008630,0.01000,1.148 | |
| 24,mlp.down_proj,0.0000002657,0.01000,2.239 | |
| 25,self_attn.q_proj,0.0000007934,0.01000,1.552 | |
| 25,self_attn.v_proj,0.0000001541,0.01000,1.554 | |
| 25,self_attn.k_proj,0.0000001487,0.01000,1.561 | |
| 25,self_attn.o_proj,0.0000003166,0.01000,0.572 | |
| 25,mlp.gate_proj,0.0000008579,0.01000,1.151 | |
| 25,mlp.up_proj,0.0000009756,0.01000,1.156 | |
| 25,mlp.down_proj,0.0000002044,0.01000,2.242 | |
| 26,self_attn.q_proj,0.0000014293,0.01000,1.558 | |
| 26,self_attn.k_proj,0.0000002228,0.01000,1.562 | |
| 26,self_attn.v_proj,0.0000002375,0.01000,1.563 | |
| 26,self_attn.o_proj,0.0000004076,0.01000,0.567 | |
| 26,mlp.gate_proj,0.0000009329,0.01000,1.151 | |
| 26,mlp.up_proj,0.0000011207,0.01000,1.154 | |
| 26,mlp.down_proj,0.0000003374,0.01000,2.246 | |
| 27,self_attn.k_proj,0.0000004542,0.01000,1.571 | |
| 27,self_attn.q_proj,0.0000023284,0.01000,1.599 | |
| 27,self_attn.v_proj,0.0000008071,0.01000,1.606 | |
| 27,self_attn.o_proj,0.0000011174,0.01000,0.568 | |
| 27,mlp.up_proj,0.0000015285,0.01000,1.143 | |
| 27,mlp.gate_proj,0.0000011756,0.01000,1.147 | |
| 27,mlp.down_proj,0.0000007143,0.01000,2.246 | |
| 28,self_attn.v_proj,0.0000006405,0.01000,1.542 | |
| 28,self_attn.k_proj,0.0000006730,0.01000,1.546 | |
| 28,self_attn.q_proj,0.0000030264,0.01000,1.549 | |
| 28,self_attn.o_proj,0.0000012090,0.01000,0.572 | |
| 28,mlp.up_proj,0.0000021163,0.01000,1.138 | |
| 28,mlp.gate_proj,0.0000018347,0.01000,1.146 | |
| 28,mlp.down_proj,0.0000017199,0.01000,2.240 | |
| 29,self_attn.v_proj,0.0000015738,0.01000,1.545 | |
| 29,self_attn.q_proj,0.0000077582,0.01000,1.558 | |
| 29,self_attn.k_proj,0.0000012017,0.01000,1.566 | |
| 29,self_attn.o_proj,0.0000032131,0.01000,0.565 | |
| 29,mlp.up_proj,0.0000028664,0.01000,1.148 | |
| 29,mlp.gate_proj,0.0000025585,0.01000,1.151 | |
| 29,mlp.down_proj,0.0000256958,0.01000,2.225 | |
| 30,self_attn.k_proj,0.0000013016,0.01000,1.539 | |
| 30,self_attn.q_proj,0.0000079752,0.01000,1.542 | |
| 30,self_attn.v_proj,0.0000017943,0.01000,1.547 | |
| 30,self_attn.o_proj,0.0000069710,0.01000,0.563 | |
| 30,mlp.up_proj,0.0000044374,0.01000,1.152 | |
| 30,mlp.gate_proj,0.0000036016,0.01000,1.156 | |
| 30,mlp.down_proj,0.0000099136,0.01000,2.248 | |
| 31,self_attn.k_proj,0.0000022095,0.01000,1.536 | |
| 31,self_attn.q_proj,0.0000119859,0.01000,1.540 | |
| 31,self_attn.v_proj,0.0000024236,0.01000,1.543 | |
| 31,self_attn.o_proj,0.0000090073,0.01000,0.565 | |
| 31,mlp.gate_proj,0.0000042423,0.01000,1.140 | |
| 31,mlp.up_proj,0.0000057331,0.01000,1.150 | |
| 31,mlp.down_proj,0.0000113583,0.01000,2.248 | |
| 32,self_attn.k_proj,0.0000032344,0.01000,1.553 | |
| 32,self_attn.q_proj,0.0000177639,0.01000,1.556 | |
| 32,self_attn.v_proj,0.0000057728,0.01000,1.588 | |
| 32,self_attn.o_proj,0.0000294938,0.01000,0.575 | |
| 32,mlp.gate_proj,0.0000079738,0.01000,1.134 | |
| 32,mlp.up_proj,0.0000090901,0.01000,1.134 | |
| 32,mlp.down_proj,0.0001718242,0.01000,2.271 | |
| 33,self_attn.k_proj,0.0000067906,0.01000,1.549 | |
| 33,self_attn.q_proj,0.0000625890,0.01000,1.553 | |
| 33,self_attn.v_proj,0.0000118461,0.01000,1.557 | |
| 33,self_attn.o_proj,0.0000358326,0.01000,0.568 | |
| 33,mlp.up_proj,0.0000198184,0.01000,1.151 | |
| 33,mlp.gate_proj,0.0000165273,0.01000,1.159 | |
| 33,mlp.down_proj,0.0001491802,0.01000,2.239 | |
| 34,self_attn.q_proj,0.0000543061,0.01000,1.567 | |
| 34,self_attn.v_proj,0.0000131741,0.01000,1.573 | |
| 34,self_attn.k_proj,0.0000087816,0.01000,1.579 | |
| 34,self_attn.o_proj,0.0000818334,0.01000,0.569 | |
| 34,mlp.gate_proj,0.0000169936,0.01000,1.145 | |
| 34,mlp.up_proj,0.0000183243,0.01000,1.151 | |
| 34,mlp.down_proj,0.0000622494,0.01000,2.241 | |
| 35,self_attn.q_proj,0.0000360077,0.01000,1.567 | |
| 35,self_attn.k_proj,0.0000084222,0.01000,1.570 | |
| 35,self_attn.v_proj,0.0000127596,0.01000,1.587 | |
| 35,self_attn.o_proj,0.0000860495,0.01000,0.571 | |
| 35,mlp.gate_proj,0.0000566747,0.01000,1.142 | |
| 35,mlp.up_proj,0.0000673418,0.01000,1.145 | |
| 35,mlp.down_proj,0.0008014260,0.01000,2.267 | |