| layer,module,loss,samples,damp,time | |
| 0,self_attn.v_proj,0.0000003109,0.05000,1.370 | |
| 0,self_attn.q_proj,0.0000009023,0.05000,1.095 | |
| 0,self_attn.k_proj,0.0000004012,0.05000,1.157 | |
| 0,self_attn.o_proj,0.0000001841,0.05000,0.853 | |
| 0,mlp.up_proj,0.0000039506,0.05000,0.783 | |
| 0,mlp.gate_proj,0.0000078002,0.05000,0.797 | |
| 0,mlp.down_proj,0.0000001560,0.05000,1.132 | |
| 1,self_attn.v_proj,0.0000001268,0.05000,1.047 | |
| 1,self_attn.k_proj,0.0000001332,0.05000,1.117 | |
| 1,self_attn.q_proj,0.0000003011,0.05000,1.129 | |
| 1,self_attn.o_proj,0.0000000526,0.05000,0.745 | |
| 1,mlp.up_proj,0.0000232322,0.05000,0.793 | |
| 1,mlp.gate_proj,0.0000781867,0.05000,0.805 | |
| 1,mlp.down_proj,0.0000001876,0.05000,1.130 | |
| 2,self_attn.k_proj,0.0000002593,0.05000,1.209 | |
| 2,self_attn.v_proj,0.0000002583,0.05000,1.233 | |
| 2,self_attn.q_proj,0.0000006157,0.05000,1.236 | |
| 2,self_attn.o_proj,0.0000001096,0.05000,0.757 | |
| 2,mlp.gate_proj,0.0000407633,0.05000,0.725 | |
| 2,mlp.up_proj,0.0000149177,0.05000,0.737 | |
| 2,mlp.down_proj,0.0005629873,0.05000,1.207 | |
| 3,self_attn.k_proj,0.0000022016,0.05000,1.281 | |
| 3,self_attn.q_proj,0.0000045678,0.05000,1.287 | |
| 3,self_attn.v_proj,0.0000022676,0.05000,1.289 | |
| 3,self_attn.o_proj,0.0000002321,0.05000,0.742 | |
| 3,mlp.up_proj,0.0000176299,0.05000,0.673 | |
| 3,mlp.gate_proj,0.0000446665,0.05000,0.706 | |
| 3,mlp.down_proj,0.0000005399,0.05000,1.150 | |
| 4,self_attn.v_proj,0.0000024637,0.05000,1.190 | |
| 4,self_attn.q_proj,0.0000048056,0.05000,1.245 | |
| 4,self_attn.k_proj,0.0000022371,0.05000,1.264 | |
| 4,self_attn.o_proj,0.0000006409,0.05000,0.750 | |
| 4,mlp.gate_proj,0.0000338714,0.05000,0.737 | |
| 4,mlp.up_proj,0.0000150340,0.05000,0.763 | |
| 4,mlp.down_proj,0.0000004408,0.05000,1.137 | |
| 5,self_attn.k_proj,0.0000031941,0.05000,1.233 | |
| 5,self_attn.v_proj,0.0000034075,0.05000,1.250 | |
| 5,self_attn.q_proj,0.0000078732,0.05000,1.258 | |
| 5,self_attn.o_proj,0.0000002897,0.05000,0.763 | |
| 5,mlp.gate_proj,0.0000207259,0.05000,0.766 | |
| 5,mlp.up_proj,0.0000124314,0.05000,0.786 | |
| 5,mlp.down_proj,0.0000005044,0.05000,1.149 | |
| 6,self_attn.v_proj,0.0000021380,0.05000,1.016 | |
| 6,self_attn.q_proj,0.0000048282,0.05000,1.027 | |
| 6,self_attn.k_proj,0.0000020889,0.05000,1.070 | |
| 6,self_attn.o_proj,0.0000007117,0.05000,0.769 | |
| 6,mlp.gate_proj,0.0000225843,0.05000,0.800 | |
| 6,mlp.up_proj,0.0000143005,0.05000,0.810 | |
| 6,mlp.down_proj,0.0000006335,0.05000,1.209 | |
| 7,self_attn.v_proj,0.0000042041,0.05000,1.054 | |
| 7,self_attn.k_proj,0.0000036301,0.05000,1.109 | |
| 7,self_attn.q_proj,0.0000089036,0.05000,1.143 | |
| 7,self_attn.o_proj,0.0000004541,0.05000,0.743 | |
| 7,mlp.up_proj,0.0000154759,0.05000,0.645 | |
| 7,mlp.gate_proj,0.0000249142,0.05000,0.658 | |
| 7,mlp.down_proj,0.0000007252,0.05000,1.215 | |
| 8,self_attn.v_proj,0.0000045567,0.05000,1.027 | |
| 8,self_attn.k_proj,0.0000046925,0.05000,1.111 | |
| 8,self_attn.q_proj,0.0000105670,0.05000,1.117 | |
| 8,self_attn.o_proj,0.0000007483,0.05000,0.817 | |
| 8,mlp.gate_proj,0.0000256787,0.05000,0.817 | |
| 8,mlp.up_proj,0.0000164744,0.05000,0.816 | |
| 8,mlp.down_proj,0.0000007462,0.05000,1.163 | |
| 9,self_attn.q_proj,0.0000227217,0.05000,1.112 | |
| 9,self_attn.k_proj,0.0000091256,0.05000,1.116 | |
| 9,self_attn.v_proj,0.0000097113,0.05000,1.127 | |
| 9,self_attn.o_proj,0.0000012330,0.05000,0.789 | |
| 9,mlp.up_proj,0.0000192108,0.05000,0.686 | |
| 9,mlp.gate_proj,0.0000304332,0.05000,0.727 | |
| 9,mlp.down_proj,0.0000010080,0.05000,1.143 | |
| 10,self_attn.q_proj,0.0000212646,0.05000,1.196 | |
| 10,self_attn.v_proj,0.0000092167,0.05000,1.230 | |
| 10,self_attn.k_proj,0.0000088879,0.05000,1.235 | |
| 10,self_attn.o_proj,0.0000008453,0.05000,0.772 | |
| 10,mlp.gate_proj,0.0000305974,0.05000,0.678 | |
| 10,mlp.up_proj,0.0000193792,0.05000,0.706 | |
| 10,mlp.down_proj,0.0000017456,0.05000,1.147 | |
| 11,self_attn.q_proj,0.0000482304,0.05000,1.152 | |
| 11,self_attn.k_proj,0.0000176641,0.05000,1.166 | |
| 11,self_attn.v_proj,0.0000170822,0.05000,1.187 | |
| 11,self_attn.o_proj,0.0000039876,0.05000,0.812 | |
| 11,mlp.up_proj,0.0000166498,0.05000,0.748 | |
| 11,mlp.gate_proj,0.0000215405,0.05000,0.766 | |
| 11,mlp.down_proj,0.0000015616,0.05000,1.239 | |
| 12,self_attn.k_proj,0.0000123774,0.05000,1.164 | |
| 12,self_attn.v_proj,0.0000131811,0.05000,1.165 | |
| 12,self_attn.q_proj,0.0000345661,0.05000,1.189 | |
| 12,self_attn.o_proj,0.0000016069,0.05000,0.812 | |
| 12,mlp.gate_proj,0.0000186182,0.05000,0.789 | |
| 12,mlp.up_proj,0.0000152392,0.05000,0.804 | |
| 12,mlp.down_proj,0.0000019535,0.05000,1.240 | |
| 13,self_attn.v_proj,0.0000176766,0.05000,1.229 | |
| 13,self_attn.q_proj,0.0000432059,0.05000,1.238 | |
| 13,self_attn.k_proj,0.0000150400,0.05000,1.240 | |
| 13,self_attn.o_proj,0.0000044218,0.05000,0.829 | |
| 13,mlp.gate_proj,0.0000209368,0.05000,0.720 | |
| 13,mlp.up_proj,0.0000169902,0.05000,0.761 | |
| 13,mlp.down_proj,0.0000019087,0.05000,1.201 | |
| 14,self_attn.k_proj,0.0000196122,0.05000,1.051 | |
| 14,self_attn.v_proj,0.0000204509,0.05000,1.111 | |
| 14,self_attn.q_proj,0.0000537520,0.05000,1.123 | |
| 14,self_attn.o_proj,0.0000023196,0.05000,0.754 | |
| 14,mlp.up_proj,0.0000169194,0.05000,0.659 | |
| 14,mlp.gate_proj,0.0000204248,0.05000,0.669 | |
| 14,mlp.down_proj,0.0000024035,0.05000,1.131 | |
| 15,self_attn.q_proj,0.0001016872,0.05000,1.187 | |
| 15,self_attn.k_proj,0.0000315203,0.05000,1.216 | |
| 15,self_attn.v_proj,0.0000414071,0.05000,1.222 | |
| 15,self_attn.o_proj,0.0000034049,0.05000,0.748 | |
| 15,mlp.gate_proj,0.0000248279,0.05000,0.737 | |
| 15,mlp.up_proj,0.0000212035,0.05000,0.762 | |
| 15,mlp.down_proj,0.0000031727,0.05000,1.125 | |
| 16,self_attn.v_proj,0.0000503639,0.05000,1.213 | |
| 16,self_attn.q_proj,0.0001481956,0.05000,1.217 | |
| 16,self_attn.k_proj,0.0000496516,0.05000,1.234 | |
| 16,self_attn.o_proj,0.0000070520,0.05000,0.751 | |
| 16,mlp.up_proj,0.0000282283,0.05000,0.766 | |
| 16,mlp.gate_proj,0.0000303624,0.05000,0.782 | |
| 16,mlp.down_proj,0.0000080868,0.05000,1.145 | |
| 17,self_attn.q_proj,0.0003941290,0.05000,1.144 | |
| 17,self_attn.v_proj,0.0001566996,0.05000,1.149 | |
| 17,self_attn.k_proj,0.0001197921,0.05000,1.183 | |
| 17,self_attn.o_proj,0.0000113098,0.05000,0.778 | |
| 17,mlp.up_proj,0.0000453949,0.05000,0.750 | |
| 17,mlp.gate_proj,0.0000473724,0.05000,0.790 | |
| 17,mlp.down_proj,0.0000078071,0.05000,1.186 | |
| 18,self_attn.k_proj,0.0001169356,0.05000,1.116 | |
| 18,self_attn.q_proj,0.0003876292,0.05000,1.148 | |
| 18,self_attn.v_proj,0.0001425182,0.05000,1.172 | |
| 18,self_attn.o_proj,0.0000107651,0.05000,0.735 | |
| 18,mlp.up_proj,0.0000571764,0.05000,0.780 | |
| 18,mlp.gate_proj,0.0000587824,0.05000,0.787 | |
| 18,mlp.down_proj,0.0000141827,0.05000,1.229 | |
| 19,self_attn.k_proj,0.0001639422,0.05000,1.208 | |
| 19,self_attn.v_proj,0.0002111269,0.05000,1.250 | |
| 19,self_attn.q_proj,0.0005905121,0.05000,1.258 | |
| 19,self_attn.o_proj,0.0000155351,0.05000,0.778 | |
| 19,mlp.up_proj,0.0000690915,0.05000,0.722 | |
| 19,mlp.gate_proj,0.0000618267,0.05000,0.744 | |
| 19,mlp.down_proj,0.0000299620,0.05000,1.155 | |
| 20,self_attn.q_proj,0.0009659973,0.05000,1.117 | |
| 20,self_attn.k_proj,0.0002813039,0.05000,1.147 | |
| 20,self_attn.v_proj,0.0003481653,0.05000,1.176 | |
| 20,self_attn.o_proj,0.0000241482,0.05000,0.748 | |
| 20,mlp.gate_proj,0.0000734136,0.05000,0.758 | |
| 20,mlp.up_proj,0.0000858579,0.05000,0.783 | |
| 20,mlp.down_proj,0.0000282660,0.05000,1.241 | |
| 21,self_attn.k_proj,0.0004253489,0.05000,1.104 | |
| 21,self_attn.v_proj,0.0005330068,0.05000,1.110 | |
| 21,self_attn.q_proj,0.0013775340,0.05000,1.124 | |
| 21,self_attn.o_proj,0.0000278336,0.05000,0.774 | |
| 21,mlp.gate_proj,0.0000810407,0.05000,0.657 | |
| 21,mlp.up_proj,0.0001050367,0.05000,0.682 | |
| 21,mlp.down_proj,0.0000429905,0.05000,1.113 | |
| 22,self_attn.q_proj,0.0019330996,0.05000,1.083 | |
| 22,self_attn.k_proj,0.0006078479,0.05000,1.106 | |
| 22,self_attn.v_proj,0.0008003848,0.05000,1.111 | |
| 22,self_attn.o_proj,0.0000184063,0.05000,0.738 | |
| 22,mlp.gate_proj,0.0000879157,0.05000,0.658 | |
| 22,mlp.up_proj,0.0001162049,0.05000,0.696 | |
| 22,mlp.down_proj,0.0000539103,0.05000,1.239 | |
| 23,self_attn.v_proj,0.0008948659,0.05000,1.241 | |
| 23,self_attn.q_proj,0.0019309018,0.05000,1.258 | |
| 23,self_attn.k_proj,0.0006971491,0.05000,1.264 | |
| 23,self_attn.o_proj,0.0000309773,0.05000,0.752 | |
| 23,mlp.gate_proj,0.0000932382,0.05000,0.682 | |
| 23,mlp.up_proj,0.0001253555,0.05000,0.705 | |
| 23,mlp.down_proj,0.0000351105,0.05000,1.210 | |
| 24,self_attn.k_proj,0.0009955094,0.05000,1.084 | |
| 24,self_attn.v_proj,0.0012167826,0.05000,1.156 | |
| 24,self_attn.q_proj,0.0032639090,0.05000,1.168 | |
| 24,self_attn.o_proj,0.0000507511,0.05000,0.748 | |
| 24,mlp.gate_proj,0.0000872461,0.05000,0.720 | |
| 24,mlp.up_proj,0.0001224611,0.05000,0.734 | |
| 24,mlp.down_proj,0.0000617926,0.05000,1.213 | |
| 25,self_attn.v_proj,0.0016807246,0.05000,1.137 | |
| 25,self_attn.k_proj,0.0009900553,0.05000,1.162 | |
| 25,self_attn.q_proj,0.0038033783,0.05000,1.175 | |
| 25,self_attn.o_proj,0.0000621992,0.05000,0.769 | |
| 25,mlp.up_proj,0.0001295126,0.05000,0.625 | |
| 25,mlp.gate_proj,0.0000873002,0.05000,0.629 | |
| 25,mlp.down_proj,0.0000822279,0.05000,1.196 | |
| 26,self_attn.v_proj,0.0016285805,0.05000,1.178 | |
| 26,self_attn.k_proj,0.0010357921,0.05000,1.217 | |
| 26,self_attn.q_proj,0.0048038255,0.05000,1.224 | |
| 26,self_attn.o_proj,0.0000872400,0.05000,0.750 | |
| 26,mlp.gate_proj,0.0001130300,0.05000,0.808 | |
| 26,mlp.up_proj,0.0001643867,0.05000,0.817 | |
| 26,mlp.down_proj,0.0002912967,0.05000,1.126 | |
| 27,self_attn.q_proj,0.0021020752,0.05000,1.244 | |
| 27,self_attn.v_proj,0.0011721282,0.05000,1.264 | |
| 27,self_attn.k_proj,0.0009209189,0.05000,1.266 | |
| 27,self_attn.o_proj,0.0000383862,0.05000,0.744 | |
| 27,mlp.up_proj,0.0006354488,0.05000,0.832 | |
| 27,mlp.gate_proj,0.0005814799,0.05000,0.837 | |
| 27,mlp.down_proj,0.0010889856,0.05000,1.237 | |