| layer,module,loss,samples,damp,time | |
| 0,self_attn.q_proj,0.0000156653,0.10000,2.083 | |
| 0,self_attn.k_proj,0.0000032556,0.10000,2.094 | |
| 0,self_attn.v_proj,0.0000005451,0.10000,2.096 | |
| 0,self_attn.o_proj,0.0000008700,0.10000,0.596 | |
| 0,mlp.gate_proj,0.0000435553,0.10000,1.181 | |
| 0,mlp.up_proj,0.0000308079,0.10000,1.186 | |
| 0,mlp.down_proj,0.0000036231,0.10000,3.570 | |
| 1,self_attn.q_proj,0.0000096145,0.10000,2.272 | |
| 1,self_attn.v_proj,0.0000009271,0.10000,2.302 | |
| 1,self_attn.k_proj,0.0000025740,0.10000,2.312 | |
| 1,self_attn.o_proj,0.0000002494,0.10000,0.587 | |
| 1,mlp.gate_proj,0.0008104122,0.10000,1.194 | |
| 1,mlp.up_proj,0.0005772593,0.10000,1.199 | |
| 1,mlp.down_proj,0.0000024426,0.10000,3.619 | |
| 2,self_attn.q_proj,0.0000300082,0.10000,2.124 | |
| 2,self_attn.k_proj,0.0000076323,0.10000,2.135 | |
| 2,self_attn.v_proj,0.0000016731,0.10000,2.138 | |
| 2,self_attn.o_proj,0.0000004566,0.10000,0.589 | |
| 2,mlp.up_proj,0.0005655596,0.10000,1.200 | |
| 2,mlp.gate_proj,0.0008204051,0.10000,1.206 | |
| 2,mlp.down_proj,0.0000034888,0.10000,3.573 | |
| 3,self_attn.v_proj,0.0000023768,0.10000,2.066 | |
| 3,self_attn.q_proj,0.0000321667,0.10000,2.071 | |
| 3,self_attn.k_proj,0.0000077473,0.10000,2.076 | |
| 3,self_attn.o_proj,0.0000013562,0.10000,0.591 | |
| 3,mlp.gate_proj,0.0013856396,0.10000,1.208 | |
| 3,mlp.up_proj,0.0011071415,0.10000,1.213 | |
| 3,mlp.down_proj,0.0000201999,0.10000,3.568 | |
| 4,self_attn.q_proj,0.0000552769,0.10000,2.068 | |
| 4,self_attn.v_proj,0.0000055460,0.10000,2.085 | |
| 4,self_attn.k_proj,0.0000116454,0.10000,2.089 | |
| 4,self_attn.o_proj,0.0000010820,0.10000,0.596 | |
| 4,mlp.up_proj,0.0008047744,0.10000,1.181 | |
| 4,mlp.gate_proj,0.0011026591,0.10000,1.187 | |
| 4,mlp.down_proj,0.0000091144,0.10000,3.550 | |
| 5,self_attn.q_proj,0.0000541105,0.10000,2.220 | |
| 5,self_attn.v_proj,0.0000053406,0.10000,2.220 | |
| 5,self_attn.k_proj,0.0000105450,0.10000,2.224 | |
| 5,self_attn.o_proj,0.0000009863,0.10000,0.591 | |
| 5,mlp.up_proj,0.0011642066,0.10000,1.197 | |
| 5,mlp.gate_proj,0.0013546876,0.10000,1.205 | |
| 5,mlp.down_proj,0.0000043981,0.10000,3.593 | |
| 6,self_attn.q_proj,0.0000342204,0.10000,2.073 | |
| 6,self_attn.v_proj,0.0000037009,0.10000,2.089 | |
| 6,self_attn.k_proj,0.0000067855,0.10000,2.092 | |
| 6,self_attn.o_proj,0.0000017441,0.10000,0.589 | |
| 6,mlp.up_proj,0.0001849451,0.10000,1.194 | |
| 6,mlp.gate_proj,0.0002499141,0.10000,1.194 | |
| 6,mlp.down_proj,0.0000109457,0.10000,3.549 | |
| 7,self_attn.v_proj,0.0000060006,0.10000,2.174 | |
| 7,self_attn.q_proj,0.0000350479,0.10000,2.181 | |
| 7,self_attn.k_proj,0.0000057774,0.10000,2.184 | |
| 7,self_attn.o_proj,0.0000031796,0.10000,0.589 | |
| 7,mlp.up_proj,0.0001335227,0.10000,1.206 | |
| 7,mlp.gate_proj,0.0001452405,0.10000,1.206 | |
| 7,mlp.down_proj,0.0000178218,0.10000,3.530 | |
| 8,self_attn.q_proj,0.0000540192,0.10000,2.123 | |
| 8,self_attn.k_proj,0.0000110568,0.10000,2.137 | |
| 8,self_attn.v_proj,0.0000055582,0.10000,2.139 | |
| 8,self_attn.o_proj,0.0000041908,0.10000,0.589 | |
| 8,mlp.gate_proj,0.0001565939,0.10000,1.199 | |
| 8,mlp.up_proj,0.0001524405,0.10000,1.202 | |
| 8,mlp.down_proj,0.0000192921,0.10000,3.578 | |
| 9,self_attn.q_proj,0.0000465290,0.10000,2.145 | |
| 9,self_attn.v_proj,0.0000078169,0.10000,2.147 | |
| 9,self_attn.k_proj,0.0000080876,0.10000,2.151 | |
| 9,self_attn.o_proj,0.0000069479,0.10000,0.618 | |
| 9,mlp.gate_proj,0.0005196439,0.10000,1.216 | |
| 9,mlp.up_proj,0.0003377999,0.10000,1.219 | |
| 9,mlp.down_proj,0.0000234840,0.10000,3.559 | |
| 10,self_attn.q_proj,0.0000447259,0.10000,2.092 | |
| 10,self_attn.v_proj,0.0000055774,0.10000,2.113 | |
| 10,self_attn.k_proj,0.0000085554,0.10000,2.114 | |
| 10,self_attn.o_proj,0.0000042167,0.10000,0.587 | |
| 10,mlp.gate_proj,0.0001891744,0.10000,1.196 | |
| 10,mlp.up_proj,0.0001727057,0.10000,1.198 | |
| 10,mlp.down_proj,0.0000204545,0.10000,3.594 | |
| 11,self_attn.k_proj,0.0000104406,0.10000,2.091 | |
| 11,self_attn.q_proj,0.0000504052,0.10000,2.101 | |
| 11,self_attn.v_proj,0.0000049300,0.10000,2.104 | |
| 11,self_attn.o_proj,0.0000057882,0.10000,0.592 | |
| 11,mlp.gate_proj,0.0001683877,0.10000,1.228 | |
| 11,mlp.up_proj,0.0001652096,0.10000,1.231 | |
| 11,mlp.down_proj,0.0000196861,0.10000,3.563 | |
| 12,self_attn.q_proj,0.0000550930,0.10000,2.122 | |
| 12,self_attn.v_proj,0.0000062055,0.10000,2.139 | |
| 12,self_attn.k_proj,0.0000111796,0.10000,2.143 | |
| 12,self_attn.o_proj,0.0000068578,0.10000,0.585 | |
| 12,mlp.gate_proj,0.0001639272,0.10000,1.187 | |
| 12,mlp.up_proj,0.0001695412,0.10000,1.190 | |
| 12,mlp.down_proj,0.0000214752,0.10000,3.629 | |
| 13,self_attn.q_proj,0.0000566915,0.10000,2.093 | |
| 13,self_attn.v_proj,0.0000073534,0.10000,2.096 | |
| 13,self_attn.k_proj,0.0000102786,0.10000,2.103 | |
| 13,self_attn.o_proj,0.0000092767,0.10000,0.588 | |
| 13,mlp.gate_proj,0.0001736244,0.10000,1.184 | |
| 13,mlp.up_proj,0.0001677590,0.10000,1.187 | |
| 13,mlp.down_proj,0.0000210207,0.10000,3.562 | |
| 14,self_attn.q_proj,0.0000750175,0.10000,2.163 | |
| 14,self_attn.v_proj,0.0000076160,0.10000,2.175 | |
| 14,self_attn.k_proj,0.0000138165,0.10000,2.179 | |
| 14,self_attn.o_proj,0.0000093313,0.10000,0.589 | |
| 14,mlp.up_proj,0.0001827195,0.10000,1.219 | |
| 14,mlp.gate_proj,0.0001781230,0.10000,1.222 | |
| 14,mlp.down_proj,0.0000232633,0.10000,3.630 | |
| 15,self_attn.q_proj,0.0000644674,0.10000,2.133 | |
| 15,self_attn.k_proj,0.0000126894,0.10000,2.139 | |
| 15,self_attn.v_proj,0.0000067925,0.10000,2.140 | |
| 15,self_attn.o_proj,0.0000076798,0.10000,0.592 | |
| 15,mlp.up_proj,0.0001769633,0.10000,1.178 | |
| 15,mlp.gate_proj,0.0001678769,0.10000,1.181 | |
| 15,mlp.down_proj,0.0000247494,0.10000,3.561 | |
| 16,self_attn.q_proj,0.0000667001,0.10000,2.092 | |
| 16,self_attn.v_proj,0.0000090089,0.10000,2.111 | |
| 16,self_attn.k_proj,0.0000117023,0.10000,2.113 | |
| 16,self_attn.o_proj,0.0000114549,0.10000,0.589 | |
| 16,mlp.up_proj,0.0001811838,0.10000,1.178 | |
| 16,mlp.gate_proj,0.0001722064,0.10000,1.180 | |
| 16,mlp.down_proj,0.0000223780,0.10000,3.583 | |
| 17,self_attn.v_proj,0.0000100828,0.10000,2.238 | |
| 17,self_attn.k_proj,0.0000123868,0.10000,2.259 | |
| 17,self_attn.q_proj,0.0000762568,0.10000,2.264 | |
| 17,self_attn.o_proj,0.0000079355,0.10000,0.594 | |
| 17,mlp.up_proj,0.0002123742,0.10000,1.371 | |
| 17,mlp.gate_proj,0.0001976952,0.10000,1.374 | |
| 17,mlp.down_proj,0.0000311225,0.10000,3.594 | |
| 18,self_attn.q_proj,0.0000607342,0.10000,2.109 | |
| 18,self_attn.k_proj,0.0000096124,0.10000,2.121 | |
| 18,self_attn.v_proj,0.0000108709,0.10000,2.130 | |
| 18,self_attn.o_proj,0.0000120994,0.10000,0.596 | |
| 18,mlp.gate_proj,0.0002071995,0.10000,1.195 | |
| 18,mlp.up_proj,0.0002252296,0.10000,1.199 | |
| 18,mlp.down_proj,0.0000362348,0.10000,3.553 | |
| 19,self_attn.k_proj,0.0000095172,0.10000,2.110 | |
| 19,self_attn.q_proj,0.0000709936,0.10000,2.114 | |
| 19,self_attn.v_proj,0.0000133349,0.10000,2.115 | |
| 19,self_attn.o_proj,0.0000142899,0.10000,0.590 | |
| 19,mlp.gate_proj,0.0002420128,0.10000,1.201 | |
| 19,mlp.up_proj,0.0002512523,0.10000,1.204 | |
| 19,mlp.down_proj,0.0000400181,0.10000,3.529 | |
| 20,self_attn.k_proj,0.0000105798,0.10000,2.128 | |
| 20,self_attn.v_proj,0.0000149335,0.10000,2.135 | |
| 20,self_attn.q_proj,0.0000714233,0.10000,2.139 | |
| 20,self_attn.o_proj,0.0000085477,0.10000,0.586 | |
| 20,mlp.up_proj,0.0003235113,0.10000,1.247 | |
| 20,mlp.gate_proj,0.0003114748,0.10000,1.252 | |
| 20,mlp.down_proj,0.0000751625,0.10000,3.538 | |
| 21,self_attn.q_proj,0.0000870349,0.10000,2.168 | |
| 21,self_attn.v_proj,0.0000235002,0.10000,2.177 | |
| 21,self_attn.k_proj,0.0000107961,0.10000,2.181 | |
| 21,self_attn.o_proj,0.0000312681,0.10000,0.588 | |
| 21,mlp.gate_proj,0.0004483905,0.10000,1.200 | |
| 21,mlp.up_proj,0.0004428911,0.10000,1.202 | |
| 21,mlp.down_proj,0.0001190182,0.10000,3.628 | |
| 22,self_attn.q_proj,0.0001302130,0.10000,2.126 | |
| 22,self_attn.k_proj,0.0000150554,0.10000,2.130 | |
| 22,self_attn.v_proj,0.0000373550,0.10000,2.138 | |
| 22,self_attn.o_proj,0.0000171763,0.10000,0.592 | |
| 22,mlp.up_proj,0.0006391320,0.10000,1.196 | |
| 22,mlp.gate_proj,0.0006453207,0.10000,1.199 | |
| 22,mlp.down_proj,0.0002066607,0.10000,3.603 | |
| 23,self_attn.q_proj,0.0001625592,0.10000,2.178 | |
| 23,self_attn.v_proj,0.0000508609,0.10000,2.185 | |
| 23,self_attn.k_proj,0.0000197490,0.10000,2.188 | |
| 23,self_attn.o_proj,0.0000418422,0.10000,0.590 | |
| 23,mlp.up_proj,0.0009001281,0.10000,1.205 | |
| 23,mlp.gate_proj,0.0009187021,0.10000,1.205 | |
| 23,mlp.down_proj,0.0002839736,0.10000,3.559 | |
| 24,self_attn.k_proj,0.0000170952,0.10000,2.132 | |
| 24,self_attn.q_proj,0.0001363194,0.10000,2.140 | |
| 24,self_attn.v_proj,0.0000470417,0.10000,2.146 | |
| 24,self_attn.o_proj,0.0000316222,0.10000,0.598 | |
| 24,mlp.gate_proj,0.0009379816,0.10000,1.190 | |
| 24,mlp.up_proj,0.0009908391,0.10000,1.193 | |
| 24,mlp.down_proj,0.0003710463,0.10000,3.578 | |
| 25,self_attn.q_proj,0.0001556033,0.10000,2.113 | |
| 25,self_attn.v_proj,0.0000708320,0.10000,2.118 | |
| 25,self_attn.k_proj,0.0000174687,0.10000,2.121 | |
| 25,self_attn.o_proj,0.0000477942,0.10000,0.607 | |
| 25,mlp.up_proj,0.0012794713,0.10000,1.372 | |
| 25,mlp.gate_proj,0.0011451249,0.10000,1.385 | |
| 25,mlp.down_proj,0.0005603151,0.10000,3.566 | |
| 26,self_attn.q_proj,0.0002087234,0.10000,2.704 | |
| 26,self_attn.k_proj,0.0000243788,0.10000,2.723 | |
| 26,self_attn.v_proj,0.0001307992,0.10000,2.778 | |
| 26,self_attn.o_proj,0.0000752641,0.10000,0.586 | |
| 26,mlp.gate_proj,0.0011720247,0.10000,1.198 | |
| 26,mlp.up_proj,0.0013400149,0.10000,1.200 | |
| 26,mlp.down_proj,0.0010762026,0.10000,3.589 | |
| 27,self_attn.q_proj,0.0003179918,0.10000,2.117 | |
| 27,self_attn.k_proj,0.0000315897,0.10000,2.125 | |
| 27,self_attn.v_proj,0.0001504091,0.10000,2.142 | |
| 27,self_attn.o_proj,0.0001204303,0.10000,0.590 | |
| 27,mlp.up_proj,0.0021019895,0.10000,1.184 | |
| 27,mlp.gate_proj,0.0019921324,0.10000,1.188 | |
| 27,mlp.down_proj,0.0029983620,0.10000,3.615 | |