8.7 KiB
8.7 KiB
| 1 | layer,module,loss,samples,damp,time |
|---|---|
| 2 | 0,self_attn.q_proj,0.0000000157,0.01000,0.453 |
| 3 | 0,self_attn.k_proj,0.0000000070,0.01000,0.218 |
| 4 | 0,self_attn.v_proj,0.0000000055,0.01000,0.220 |
| 5 | 0,self_attn.o_proj,0.0000000039,0.01000,0.433 |
| 6 | 0,mlp.gate_proj,0.0000001457,0.01000,0.216 |
| 7 | 0,mlp.up_proj,0.0000000745,0.01000,0.215 |
| 8 | 0,mlp.down_proj,0.0000000050,0.01000,0.652 |
| 9 | 1,self_attn.q_proj,0.0000000056,0.01000,0.286 |
| 10 | 1,self_attn.k_proj,0.0000000025,0.01000,0.217 |
| 11 | 1,self_attn.v_proj,0.0000000024,0.01000,0.217 |
| 12 | 1,self_attn.o_proj,0.0000000010,0.01000,0.434 |
| 13 | 1,mlp.gate_proj,0.0000008918,0.01000,0.216 |
| 14 | 1,mlp.up_proj,0.0000002667,0.01000,0.216 |
| 15 | 1,mlp.down_proj,0.0000000066,0.01000,0.659 |
| 16 | 2,self_attn.q_proj,0.0000000110,0.01000,0.255 |
| 17 | 2,self_attn.k_proj,0.0000000047,0.01000,0.219 |
| 18 | 2,self_attn.v_proj,0.0000000046,0.01000,0.218 |
| 19 | 2,self_attn.o_proj,0.0000000016,0.01000,0.435 |
| 20 | 2,mlp.gate_proj,0.0000006115,0.01000,0.219 |
| 21 | 2,mlp.up_proj,0.0000002231,0.01000,0.218 |
| 22 | 2,mlp.down_proj,0.0000078325,0.01000,0.661 |
| 23 | 3,self_attn.q_proj,0.0000000868,0.01000,0.260 |
| 24 | 3,self_attn.k_proj,0.0000000421,0.01000,0.220 |
| 25 | 3,self_attn.v_proj,0.0000000429,0.01000,0.219 |
| 26 | 3,self_attn.o_proj,0.0000000023,0.01000,0.442 |
| 27 | 3,mlp.gate_proj,0.0000008147,0.01000,0.218 |
| 28 | 3,mlp.up_proj,0.0000003186,0.01000,0.218 |
| 29 | 3,mlp.down_proj,0.0000000175,0.01000,0.701 |
| 30 | 4,self_attn.q_proj,0.0000000812,0.01000,0.263 |
| 31 | 4,self_attn.k_proj,0.0000000383,0.01000,0.217 |
| 32 | 4,self_attn.v_proj,0.0000000410,0.01000,0.219 |
| 33 | 4,self_attn.o_proj,0.0000000053,0.01000,0.437 |
| 34 | 4,mlp.gate_proj,0.0000007532,0.01000,0.219 |
| 35 | 4,mlp.up_proj,0.0000003315,0.01000,0.218 |
| 36 | 4,mlp.down_proj,0.0000000219,0.01000,0.659 |
| 37 | 5,self_attn.q_proj,0.0000001455,0.01000,0.262 |
| 38 | 5,self_attn.k_proj,0.0000000597,0.01000,0.223 |
| 39 | 5,self_attn.v_proj,0.0000000636,0.01000,0.226 |
| 40 | 5,self_attn.o_proj,0.0000000076,0.01000,0.452 |
| 41 | 5,mlp.gate_proj,0.0000005187,0.01000,0.223 |
| 42 | 5,mlp.up_proj,0.0000003102,0.01000,0.218 |
| 43 | 5,mlp.down_proj,0.0000000246,0.01000,0.664 |
| 44 | 6,self_attn.q_proj,0.0000000990,0.01000,0.257 |
| 45 | 6,self_attn.k_proj,0.0000000439,0.01000,0.222 |
| 46 | 6,self_attn.v_proj,0.0000000422,0.01000,0.218 |
| 47 | 6,self_attn.o_proj,0.0000000059,0.01000,0.438 |
| 48 | 6,mlp.gate_proj,0.0000006165,0.01000,0.216 |
| 49 | 6,mlp.up_proj,0.0000003886,0.01000,0.216 |
| 50 | 6,mlp.down_proj,0.0000000316,0.01000,0.661 |
| 51 | 7,self_attn.q_proj,0.0000001937,0.01000,0.282 |
| 52 | 7,self_attn.k_proj,0.0000000798,0.01000,0.249 |
| 53 | 7,self_attn.v_proj,0.0000000900,0.01000,0.246 |
| 54 | 7,self_attn.o_proj,0.0000000101,0.01000,0.490 |
| 55 | 7,mlp.gate_proj,0.0000007165,0.01000,0.243 |
| 56 | 7,mlp.up_proj,0.0000004434,0.01000,0.243 |
| 57 | 7,mlp.down_proj,0.0000000395,0.01000,0.700 |
| 58 | 8,self_attn.q_proj,0.0000002418,0.01000,0.257 |
| 59 | 8,self_attn.k_proj,0.0000001092,0.01000,0.221 |
| 60 | 8,self_attn.v_proj,0.0000001022,0.01000,0.217 |
| 61 | 8,self_attn.o_proj,0.0000000103,0.01000,0.435 |
| 62 | 8,mlp.gate_proj,0.0000007137,0.01000,0.218 |
| 63 | 8,mlp.up_proj,0.0000004578,0.01000,0.216 |
| 64 | 8,mlp.down_proj,0.0000000417,0.01000,0.659 |
| 65 | 9,self_attn.q_proj,0.0000004561,0.01000,0.291 |
| 66 | 9,self_attn.k_proj,0.0000001857,0.01000,0.219 |
| 67 | 9,self_attn.v_proj,0.0000001929,0.01000,0.221 |
| 68 | 9,self_attn.o_proj,0.0000000167,0.01000,0.439 |
| 69 | 9,mlp.gate_proj,0.0000007954,0.01000,0.220 |
| 70 | 9,mlp.up_proj,0.0000004992,0.01000,0.218 |
| 71 | 9,mlp.down_proj,0.0000000556,0.01000,0.661 |
| 72 | 10,self_attn.q_proj,0.0000003840,0.01000,0.289 |
| 73 | 10,self_attn.k_proj,0.0000001600,0.01000,0.220 |
| 74 | 10,self_attn.v_proj,0.0000001638,0.01000,0.220 |
| 75 | 10,self_attn.o_proj,0.0000000157,0.01000,0.446 |
| 76 | 10,mlp.gate_proj,0.0000008065,0.01000,0.221 |
| 77 | 10,mlp.up_proj,0.0000005084,0.01000,0.222 |
| 78 | 10,mlp.down_proj,0.0000000788,0.01000,0.671 |
| 79 | 11,self_attn.q_proj,0.0000008024,0.01000,0.283 |
| 80 | 11,self_attn.k_proj,0.0000003071,0.01000,0.239 |
| 81 | 11,self_attn.v_proj,0.0000002804,0.01000,0.220 |
| 82 | 11,self_attn.o_proj,0.0000000493,0.01000,0.443 |
| 83 | 11,mlp.gate_proj,0.0000006262,0.01000,0.220 |
| 84 | 11,mlp.up_proj,0.0000004808,0.01000,0.220 |
| 85 | 11,mlp.down_proj,0.0000000880,0.01000,0.668 |
| 86 | 12,self_attn.q_proj,0.0000007135,0.01000,0.266 |
| 87 | 12,self_attn.k_proj,0.0000002560,0.01000,0.218 |
| 88 | 12,self_attn.v_proj,0.0000002706,0.01000,0.215 |
| 89 | 12,self_attn.o_proj,0.0000000163,0.01000,0.443 |
| 90 | 12,mlp.gate_proj,0.0000005636,0.01000,0.215 |
| 91 | 12,mlp.up_proj,0.0000004659,0.01000,0.216 |
| 92 | 12,mlp.down_proj,0.0000000890,0.01000,0.659 |
| 93 | 13,self_attn.q_proj,0.0000007450,0.01000,0.255 |
| 94 | 13,self_attn.k_proj,0.0000002548,0.01000,0.215 |
| 95 | 13,self_attn.v_proj,0.0000002986,0.01000,0.215 |
| 96 | 13,self_attn.o_proj,0.0000000196,0.01000,0.432 |
| 97 | 13,mlp.gate_proj,0.0000006345,0.01000,0.214 |
| 98 | 13,mlp.up_proj,0.0000005136,0.01000,0.218 |
| 99 | 13,mlp.down_proj,0.0000000943,0.01000,0.656 |
| 100 | 14,self_attn.q_proj,0.0000010099,0.01000,0.258 |
| 101 | 14,self_attn.k_proj,0.0000003666,0.01000,0.215 |
| 102 | 14,self_attn.v_proj,0.0000003883,0.01000,0.215 |
| 103 | 14,self_attn.o_proj,0.0000000272,0.01000,0.435 |
| 104 | 14,mlp.gate_proj,0.0000006731,0.01000,0.215 |
| 105 | 14,mlp.up_proj,0.0000005531,0.01000,0.217 |
| 106 | 14,mlp.down_proj,0.0000001202,0.01000,0.653 |
| 107 | 15,self_attn.q_proj,0.0000019066,0.01000,0.286 |
| 108 | 15,self_attn.k_proj,0.0000005995,0.01000,0.218 |
| 109 | 15,self_attn.v_proj,0.0000007671,0.01000,0.215 |
| 110 | 15,self_attn.o_proj,0.0000000275,0.01000,0.434 |
| 111 | 15,mlp.gate_proj,0.0000007228,0.01000,0.215 |
| 112 | 15,mlp.up_proj,0.0000006042,0.01000,0.215 |
| 113 | 15,mlp.down_proj,0.0000001389,0.01000,0.656 |
| 114 | 16,self_attn.q_proj,0.0000023607,0.01000,0.288 |
| 115 | 16,self_attn.k_proj,0.0000008254,0.01000,0.216 |
| 116 | 16,self_attn.v_proj,0.0000007666,0.01000,0.215 |
| 117 | 16,self_attn.o_proj,0.0000000499,0.01000,0.436 |
| 118 | 16,mlp.gate_proj,0.0000007419,0.01000,0.213 |
| 119 | 16,mlp.up_proj,0.0000006750,0.01000,0.215 |
| 120 | 16,mlp.down_proj,0.0000002760,0.01000,0.653 |
| 121 | 17,self_attn.q_proj,0.0000051417,0.01000,0.286 |
| 122 | 17,self_attn.k_proj,0.0000016262,0.01000,0.216 |
| 123 | 17,self_attn.v_proj,0.0000019666,0.01000,0.215 |
| 124 | 17,self_attn.o_proj,0.0000001240,0.01000,0.433 |
| 125 | 17,mlp.gate_proj,0.0000010348,0.01000,0.216 |
| 126 | 17,mlp.up_proj,0.0000009428,0.01000,0.213 |
| 127 | 17,mlp.down_proj,0.0000003055,0.01000,0.656 |
| 128 | 18,self_attn.q_proj,0.0000046798,0.01000,0.266 |
| 129 | 18,self_attn.k_proj,0.0000014684,0.01000,0.216 |
| 130 | 18,self_attn.v_proj,0.0000017435,0.01000,0.216 |
| 131 | 18,self_attn.o_proj,0.0000000528,0.01000,0.432 |
| 132 | 18,mlp.gate_proj,0.0000011800,0.01000,0.214 |
| 133 | 18,mlp.up_proj,0.0000010920,0.01000,0.213 |
| 134 | 18,mlp.down_proj,0.0000005320,0.01000,0.651 |
| 135 | 19,self_attn.q_proj,0.0000081104,0.01000,0.262 |
| 136 | 19,self_attn.k_proj,0.0000023922,0.01000,0.216 |
| 137 | 19,self_attn.v_proj,0.0000029877,0.01000,0.215 |
| 138 | 19,self_attn.o_proj,0.0000001101,0.01000,0.433 |
| 139 | 19,mlp.gate_proj,0.0000012617,0.01000,0.214 |
| 140 | 19,mlp.up_proj,0.0000013705,0.01000,0.214 |
| 141 | 19,mlp.down_proj,0.0000010956,0.01000,0.654 |
| 142 | 20,self_attn.q_proj,0.0000104670,0.01000,0.278 |
| 143 | 20,self_attn.k_proj,0.0000034022,0.01000,0.243 |
| 144 | 20,self_attn.v_proj,0.0000042421,0.01000,0.243 |
| 145 | 20,self_attn.o_proj,0.0000001907,0.01000,0.493 |
| 146 | 20,mlp.gate_proj,0.0000013812,0.01000,0.242 |
| 147 | 20,mlp.up_proj,0.0000015767,0.01000,0.242 |
| 148 | 20,mlp.down_proj,0.0000016353,0.01000,0.748 |
| 149 | 21,self_attn.q_proj,0.0000185475,0.01000,0.276 |
| 150 | 21,self_attn.k_proj,0.0000059908,0.01000,0.242 |
| 151 | 21,self_attn.v_proj,0.0000075683,0.01000,0.242 |
| 152 | 21,self_attn.o_proj,0.0000003463,0.01000,0.487 |
| 153 | 21,mlp.gate_proj,0.0000014603,0.01000,0.239 |
| 154 | 21,mlp.up_proj,0.0000018684,0.01000,0.239 |
| 155 | 21,mlp.down_proj,0.0000023298,0.01000,0.744 |
| 156 | 22,self_attn.q_proj,0.0000192655,0.01000,0.280 |
| 157 | 22,self_attn.k_proj,0.0000066621,0.01000,0.248 |
| 158 | 22,self_attn.v_proj,0.0000090980,0.01000,0.250 |
| 159 | 22,self_attn.o_proj,0.0000002642,0.01000,0.503 |
| 160 | 22,mlp.gate_proj,0.0000016697,0.01000,0.243 |
| 161 | 22,mlp.up_proj,0.0000021728,0.01000,0.240 |
| 162 | 22,mlp.down_proj,0.0000026377,0.01000,0.750 |
| 163 | 23,self_attn.q_proj,0.0000223355,0.01000,0.282 |
| 164 | 23,self_attn.k_proj,0.0000091246,0.01000,0.251 |
| 165 | 23,self_attn.v_proj,0.0000111566,0.01000,0.250 |
| 166 | 23,self_attn.o_proj,0.0000002412,0.01000,0.487 |
| 167 | 23,mlp.gate_proj,0.0000018961,0.01000,0.243 |
| 168 | 23,mlp.up_proj,0.0000025225,0.01000,0.239 |
| 169 | 23,mlp.down_proj,0.0000023748,0.01000,0.733 |
| 170 | 24,self_attn.q_proj,0.0000478594,0.01000,0.277 |
| 171 | 24,self_attn.k_proj,0.0000162195,0.01000,0.244 |
| 172 | 24,self_attn.v_proj,0.0000187105,0.01000,0.244 |
| 173 | 24,self_attn.o_proj,0.0000003140,0.01000,0.490 |
| 174 | 24,mlp.gate_proj,0.0000018491,0.01000,0.242 |
| 175 | 24,mlp.up_proj,0.0000025646,0.01000,0.241 |
| 176 | 24,mlp.down_proj,0.0000022875,0.01000,0.743 |
| 177 | 25,self_attn.q_proj,0.0000736474,0.01000,0.261 |
| 178 | 25,self_attn.k_proj,0.0000219495,0.01000,0.218 |
| 179 | 25,self_attn.v_proj,0.0000348672,0.01000,0.220 |
| 180 | 25,self_attn.o_proj,0.0000005398,0.01000,0.441 |
| 181 | 25,mlp.gate_proj,0.0000018458,0.01000,0.217 |
| 182 | 25,mlp.up_proj,0.0000026985,0.01000,0.217 |
| 183 | 25,mlp.down_proj,0.0000027734,0.01000,0.647 |
| 184 | 26,self_attn.q_proj,0.0000973767,0.01000,0.251 |
| 185 | 26,self_attn.k_proj,0.0000243580,0.01000,0.241 |
| 186 | 26,self_attn.v_proj,0.0000364561,0.01000,0.243 |
| 187 | 26,self_attn.o_proj,0.0000044389,0.01000,0.485 |
| 188 | 26,mlp.gate_proj,0.0000018377,0.01000,0.240 |
| 189 | 26,mlp.up_proj,0.0000026512,0.01000,0.246 |
| 190 | 26,mlp.down_proj,0.0000062773,0.01000,0.744 |
| 191 | 27,self_attn.q_proj,0.0000334389,0.01000,0.290 |
| 192 | 27,self_attn.k_proj,0.0000149378,0.01000,0.243 |
| 193 | 27,self_attn.v_proj,0.0000184760,0.01000,0.243 |
| 194 | 27,self_attn.o_proj,0.0000022290,0.01000,0.489 |
| 195 | 27,mlp.gate_proj,0.0000068754,0.01000,0.242 |
| 196 | 27,mlp.up_proj,0.0000081715,0.01000,0.240 |
| 197 | 27,mlp.down_proj,0.0000067090,0.01000,0.735 |