13 KiB
13 KiB
| 1 | layer,module,loss,samples,damp,time |
|---|---|
| 2 | 0,self_attn.k_proj,0.0000586761,0.01000,1.322 |
| 3 | 0,self_attn.v_proj,0.0000354517,0.01000,0.423 |
| 4 | 0,self_attn.q_proj,0.0001688128,0.01000,0.814 |
| 5 | 0,self_attn.o_proj,0.0000504424,0.01000,0.412 |
| 6 | 0,mlp.up_proj,0.0002153337,0.01000,0.477 |
| 7 | 0,mlp.gate_proj,0.0001274616,0.01000,0.477 |
| 8 | 0,mlp.down_proj,0.0000257240,0.01000,1.993 |
| 9 | 1,self_attn.k_proj,0.0001001337,0.01000,0.431 |
| 10 | 1,self_attn.v_proj,0.0000342787,0.01000,0.584 |
| 11 | 1,self_attn.q_proj,0.0002745207,0.01000,0.443 |
| 12 | 1,self_attn.o_proj,0.0000241509,0.01000,0.438 |
| 13 | 1,mlp.up_proj,0.0003969561,0.01000,0.475 |
| 14 | 1,mlp.gate_proj,0.0002391965,0.01000,0.483 |
| 15 | 1,mlp.down_proj,0.0000294897,0.01000,1.630 |
| 16 | 2,self_attn.k_proj,0.0001192039,0.01000,0.404 |
| 17 | 2,self_attn.v_proj,0.0000332262,0.01000,0.411 |
| 18 | 2,self_attn.q_proj,0.0002993230,0.01000,0.420 |
| 19 | 2,self_attn.o_proj,0.0000098605,0.01000,0.427 |
| 20 | 2,mlp.up_proj,0.0003834351,0.01000,0.476 |
| 21 | 2,mlp.gate_proj,0.0003096603,0.01000,0.617 |
| 22 | 2,mlp.down_proj,0.0000365176,0.01000,2.022 |
| 23 | 3,self_attn.k_proj,0.0002040798,0.01000,0.423 |
| 24 | 3,self_attn.v_proj,0.0000696390,0.01000,0.427 |
| 25 | 3,self_attn.q_proj,0.0006119413,0.01000,0.432 |
| 26 | 3,self_attn.o_proj,0.0000065294,0.01000,0.508 |
| 27 | 3,mlp.up_proj,0.0003998308,0.01000,0.550 |
| 28 | 3,mlp.gate_proj,0.0003683545,0.01000,0.531 |
| 29 | 3,mlp.down_proj,0.0000227664,0.01000,1.659 |
| 30 | 4,self_attn.k_proj,0.0001992193,0.01000,0.414 |
| 31 | 4,self_attn.v_proj,0.0000724991,0.01000,0.399 |
| 32 | 4,self_attn.q_proj,0.0005812223,0.01000,0.418 |
| 33 | 4,self_attn.o_proj,0.0000047541,0.01000,0.415 |
| 34 | 4,mlp.up_proj,0.0003851807,0.01000,0.488 |
| 35 | 4,mlp.gate_proj,0.0003571671,0.01000,0.482 |
| 36 | 4,mlp.down_proj,0.0000273872,0.01000,1.597 |
| 37 | 5,self_attn.k_proj,0.0002788666,0.01000,0.396 |
| 38 | 5,self_attn.v_proj,0.0000720661,0.01000,0.403 |
| 39 | 5,self_attn.q_proj,0.0007805153,0.01000,0.411 |
| 40 | 5,self_attn.o_proj,0.0000036737,0.01000,0.410 |
| 41 | 5,mlp.up_proj,0.0003783789,0.01000,0.472 |
| 42 | 5,mlp.gate_proj,0.0003321028,0.01000,0.480 |
| 43 | 5,mlp.down_proj,0.0010589203,0.01000,1.561 |
| 44 | 6,self_attn.k_proj,0.0002729923,0.01000,0.425 |
| 45 | 6,self_attn.v_proj,0.0001261053,0.01000,0.403 |
| 46 | 6,self_attn.q_proj,0.0008447057,0.01000,0.405 |
| 47 | 6,self_attn.o_proj,0.0000053821,0.01000,0.409 |
| 48 | 6,mlp.up_proj,0.0004143942,0.01000,0.469 |
| 49 | 6,mlp.gate_proj,0.0003667180,0.01000,0.467 |
| 50 | 6,mlp.down_proj,0.0000214279,0.01000,1.551 |
| 51 | 7,self_attn.k_proj,0.0001874382,0.01000,0.405 |
| 52 | 7,self_attn.v_proj,0.0000727971,0.01000,0.416 |
| 53 | 7,self_attn.q_proj,0.0005578802,0.01000,0.413 |
| 54 | 7,self_attn.o_proj,0.0000066064,0.01000,0.410 |
| 55 | 7,mlp.up_proj,0.0003853983,0.01000,0.475 |
| 56 | 7,mlp.gate_proj,0.0003386188,0.01000,0.473 |
| 57 | 7,mlp.down_proj,0.0000226514,0.01000,1.591 |
| 58 | 8,self_attn.k_proj,0.0002412666,0.01000,0.454 |
| 59 | 8,self_attn.v_proj,0.0000851461,0.01000,0.396 |
| 60 | 8,self_attn.q_proj,0.0007171566,0.01000,0.405 |
| 61 | 8,self_attn.o_proj,0.0000049167,0.01000,0.404 |
| 62 | 8,mlp.up_proj,0.0003648815,0.01000,0.469 |
| 63 | 8,mlp.gate_proj,0.0003146806,0.01000,0.469 |
| 64 | 8,mlp.down_proj,0.0000242255,0.01000,1.566 |
| 65 | 9,self_attn.k_proj,0.0002791988,0.01000,0.402 |
| 66 | 9,self_attn.v_proj,0.0000772140,0.01000,0.397 |
| 67 | 9,self_attn.q_proj,0.0008410137,0.01000,0.408 |
| 68 | 9,self_attn.o_proj,0.0000074889,0.01000,0.410 |
| 69 | 9,mlp.up_proj,0.0003793764,0.01000,0.480 |
| 70 | 9,mlp.gate_proj,0.0003070541,0.01000,0.466 |
| 71 | 9,mlp.down_proj,0.0000285720,0.01000,1.561 |
| 72 | 10,self_attn.k_proj,0.0002847525,0.01000,0.401 |
| 73 | 10,self_attn.v_proj,0.0000690638,0.01000,0.401 |
| 74 | 10,self_attn.q_proj,0.0007431567,0.01000,0.408 |
| 75 | 10,self_attn.o_proj,0.0000112762,0.01000,0.401 |
| 76 | 10,mlp.up_proj,0.0003249209,0.01000,0.473 |
| 77 | 10,mlp.gate_proj,0.0002650708,0.01000,0.468 |
| 78 | 10,mlp.down_proj,0.0000318138,0.01000,1.563 |
| 79 | 11,self_attn.k_proj,0.0003096352,0.01000,0.401 |
| 80 | 11,self_attn.v_proj,0.0000897225,0.01000,0.400 |
| 81 | 11,self_attn.q_proj,0.0008893770,0.01000,0.407 |
| 82 | 11,self_attn.o_proj,0.0000122147,0.01000,0.406 |
| 83 | 11,mlp.up_proj,0.0003072199,0.01000,0.479 |
| 84 | 11,mlp.gate_proj,0.0002517828,0.01000,0.483 |
| 85 | 11,mlp.down_proj,0.0000332955,0.01000,1.573 |
| 86 | 12,self_attn.k_proj,0.0003948360,0.01000,0.405 |
| 87 | 12,self_attn.v_proj,0.0000998961,0.01000,0.402 |
| 88 | 12,self_attn.q_proj,0.0012023712,0.01000,0.416 |
| 89 | 12,self_attn.o_proj,0.0000106507,0.01000,0.414 |
| 90 | 12,mlp.up_proj,0.0003385118,0.01000,0.472 |
| 91 | 12,mlp.gate_proj,0.0002744211,0.01000,0.484 |
| 92 | 12,mlp.down_proj,0.0000393885,0.01000,1.570 |
| 93 | 13,self_attn.k_proj,0.0003714790,0.01000,0.401 |
| 94 | 13,self_attn.v_proj,0.0001324728,0.01000,0.396 |
| 95 | 13,self_attn.q_proj,0.0010769964,0.01000,0.406 |
| 96 | 13,self_attn.o_proj,0.0000178290,0.01000,0.406 |
| 97 | 13,mlp.up_proj,0.0003482320,0.01000,0.472 |
| 98 | 13,mlp.gate_proj,0.0002787499,0.01000,0.470 |
| 99 | 13,mlp.down_proj,0.0000429903,0.01000,1.559 |
| 100 | 14,self_attn.k_proj,0.0003436909,0.01000,0.404 |
| 101 | 14,self_attn.v_proj,0.0001705214,0.01000,0.404 |
| 102 | 14,self_attn.q_proj,0.0010559997,0.01000,0.409 |
| 103 | 14,self_attn.o_proj,0.0000169769,0.01000,0.433 |
| 104 | 14,mlp.up_proj,0.0003698721,0.01000,0.479 |
| 105 | 14,mlp.gate_proj,0.0002898470,0.01000,0.481 |
| 106 | 14,mlp.down_proj,0.0000458820,0.01000,1.589 |
| 107 | 15,self_attn.k_proj,0.0003220908,0.01000,0.407 |
| 108 | 15,self_attn.v_proj,0.0001672515,0.01000,0.403 |
| 109 | 15,self_attn.q_proj,0.0010098481,0.01000,0.410 |
| 110 | 15,self_attn.o_proj,0.0000213154,0.01000,0.405 |
| 111 | 15,mlp.up_proj,0.0003779694,0.01000,0.489 |
| 112 | 15,mlp.gate_proj,0.0002958367,0.01000,0.477 |
| 113 | 15,mlp.down_proj,0.0000516704,0.01000,1.591 |
| 114 | 16,self_attn.k_proj,0.0003507759,0.01000,0.402 |
| 115 | 16,self_attn.v_proj,0.0001544250,0.01000,0.396 |
| 116 | 16,self_attn.q_proj,0.0010080066,0.01000,0.404 |
| 117 | 16,self_attn.o_proj,0.0000288016,0.01000,0.401 |
| 118 | 16,mlp.up_proj,0.0003919088,0.01000,0.468 |
| 119 | 16,mlp.gate_proj,0.0003031051,0.01000,0.468 |
| 120 | 16,mlp.down_proj,0.0000538878,0.01000,1.553 |
| 121 | 17,self_attn.k_proj,0.0003736942,0.01000,0.400 |
| 122 | 17,self_attn.v_proj,0.0001494565,0.01000,0.395 |
| 123 | 17,self_attn.q_proj,0.0010871407,0.01000,0.413 |
| 124 | 17,self_attn.o_proj,0.0000141816,0.01000,0.422 |
| 125 | 17,mlp.up_proj,0.0004607816,0.01000,0.472 |
| 126 | 17,mlp.gate_proj,0.0003417494,0.01000,0.476 |
| 127 | 17,mlp.down_proj,0.0000554428,0.01000,1.569 |
| 128 | 18,self_attn.k_proj,0.0003214898,0.01000,0.407 |
| 129 | 18,self_attn.v_proj,0.0001460328,0.01000,0.403 |
| 130 | 18,self_attn.q_proj,0.0009515478,0.01000,0.406 |
| 131 | 18,self_attn.o_proj,0.0000206779,0.01000,0.407 |
| 132 | 18,mlp.up_proj,0.0004685978,0.01000,0.472 |
| 133 | 18,mlp.gate_proj,0.0003603305,0.01000,0.471 |
| 134 | 18,mlp.down_proj,0.0000611692,0.01000,1.573 |
| 135 | 19,self_attn.k_proj,0.0003957018,0.01000,0.398 |
| 136 | 19,self_attn.v_proj,0.0001980347,0.01000,0.397 |
| 137 | 19,self_attn.q_proj,0.0012038962,0.01000,0.403 |
| 138 | 19,self_attn.o_proj,0.0000182705,0.01000,0.405 |
| 139 | 19,mlp.up_proj,0.0004899677,0.01000,0.469 |
| 140 | 19,mlp.gate_proj,0.0003855893,0.01000,0.470 |
| 141 | 19,mlp.down_proj,0.0000754451,0.01000,1.578 |
| 142 | 20,self_attn.k_proj,0.0000000000,0.01000,0.412 |
| 143 | 20,self_attn.v_proj,0.0000001169,0.01000,0.398 |
| 144 | 20,self_attn.q_proj,0.0000000000,0.01000,0.406 |
| 145 | 20,self_attn.o_proj,0.0000000019,0.01000,0.405 |
| 146 | 20,mlp.up_proj,0.0005356293,0.01000,0.506 |
| 147 | 20,mlp.gate_proj,0.0004464875,0.01000,0.497 |
| 148 | 20,mlp.down_proj,0.0000753395,0.01000,1.704 |
| 149 | 21,self_attn.k_proj,0.0004827192,0.01000,0.455 |
| 150 | 21,self_attn.v_proj,0.0002100711,0.01000,0.423 |
| 151 | 21,self_attn.q_proj,0.0013876994,0.01000,0.446 |
| 152 | 21,self_attn.o_proj,0.0000157326,0.01000,0.429 |
| 153 | 21,mlp.up_proj,0.0005929679,0.01000,0.534 |
| 154 | 21,mlp.gate_proj,0.0005037194,0.01000,0.550 |
| 155 | 21,mlp.down_proj,0.0000908208,0.01000,1.741 |
| 156 | 22,self_attn.k_proj,0.0004898814,0.01000,0.461 |
| 157 | 22,self_attn.v_proj,0.0002796991,0.01000,0.430 |
| 158 | 22,self_attn.q_proj,0.0015202498,0.01000,0.463 |
| 159 | 22,self_attn.o_proj,0.0000083112,0.01000,0.420 |
| 160 | 22,mlp.up_proj,0.0006740890,0.01000,0.525 |
| 161 | 22,mlp.gate_proj,0.0005844082,0.01000,0.522 |
| 162 | 22,mlp.down_proj,0.0001015973,0.01000,1.680 |
| 163 | 23,self_attn.k_proj,0.0000000000,0.01000,0.419 |
| 164 | 23,self_attn.v_proj,0.0000000540,0.01000,0.454 |
| 165 | 23,self_attn.q_proj,0.0000000000,0.01000,0.431 |
| 166 | 23,self_attn.o_proj,0.0000000004,0.01000,0.448 |
| 167 | 23,mlp.up_proj,0.0007605743,0.01000,0.530 |
| 168 | 23,mlp.gate_proj,0.0006684575,0.01000,0.485 |
| 169 | 23,mlp.down_proj,0.0001126684,0.01000,1.681 |
| 170 | 24,self_attn.k_proj,0.0000000000,0.01000,0.443 |
| 171 | 24,self_attn.v_proj,0.0000000667,0.01000,0.424 |
| 172 | 24,self_attn.q_proj,0.0000000000,0.01000,0.442 |
| 173 | 24,self_attn.o_proj,0.0000000005,0.01000,0.454 |
| 174 | 24,mlp.up_proj,0.0008429098,0.01000,0.508 |
| 175 | 24,mlp.gate_proj,0.0007483406,0.01000,0.533 |
| 176 | 24,mlp.down_proj,0.0001261322,0.01000,1.699 |
| 177 | 25,self_attn.k_proj,0.0006377540,0.01000,0.455 |
| 178 | 25,self_attn.v_proj,0.0004328548,0.01000,0.426 |
| 179 | 25,self_attn.q_proj,0.0020318676,0.01000,0.461 |
| 180 | 25,self_attn.o_proj,0.0000157571,0.01000,0.429 |
| 181 | 25,mlp.up_proj,0.0009439098,0.01000,0.526 |
| 182 | 25,mlp.gate_proj,0.0008411989,0.01000,0.562 |
| 183 | 25,mlp.down_proj,0.0001479721,0.01000,1.709 |
| 184 | 26,self_attn.k_proj,0.0000000000,0.01000,0.449 |
| 185 | 26,self_attn.v_proj,0.0000001092,0.01000,0.453 |
| 186 | 26,self_attn.q_proj,0.0000000000,0.01000,0.464 |
| 187 | 26,self_attn.o_proj,0.0000000007,0.01000,0.428 |
| 188 | 26,mlp.up_proj,0.0010536772,0.01000,0.558 |
| 189 | 26,mlp.gate_proj,0.0009431117,0.01000,0.549 |
| 190 | 26,mlp.down_proj,0.0001719375,0.01000,1.696 |
| 191 | 27,self_attn.k_proj,0.0000000000,0.01000,0.421 |
| 192 | 27,self_attn.v_proj,0.0000001668,0.01000,0.445 |
| 193 | 27,self_attn.q_proj,0.0000000000,0.01000,0.424 |
| 194 | 27,self_attn.o_proj,0.0000000011,0.01000,0.447 |
| 195 | 27,mlp.up_proj,0.0011833173,0.01000,0.506 |
| 196 | 27,mlp.gate_proj,0.0010602921,0.01000,0.482 |
| 197 | 27,mlp.down_proj,0.0001971719,0.01000,1.692 |
| 198 | 28,self_attn.k_proj,0.0006555448,0.01000,0.455 |
| 199 | 28,self_attn.v_proj,0.0005032099,0.01000,0.427 |
| 200 | 28,self_attn.q_proj,0.0021870123,0.01000,0.441 |
| 201 | 28,self_attn.o_proj,0.0000279065,0.01000,0.437 |
| 202 | 28,mlp.up_proj,0.0013163162,0.01000,0.497 |
| 203 | 28,mlp.gate_proj,0.0011671635,0.01000,0.510 |
| 204 | 28,mlp.down_proj,0.0002391058,0.01000,1.691 |
| 205 | 29,self_attn.k_proj,0.0000000000,0.01000,0.438 |
| 206 | 29,self_attn.v_proj,0.0000007706,0.01000,0.452 |
| 207 | 29,self_attn.q_proj,0.0000000000,0.01000,0.475 |
| 208 | 29,self_attn.o_proj,0.0000000089,0.01000,0.445 |
| 209 | 29,mlp.up_proj,0.0014455931,0.01000,0.548 |
| 210 | 29,mlp.gate_proj,0.0012611508,0.01000,0.525 |
| 211 | 29,mlp.down_proj,0.0002838329,0.01000,1.793 |
| 212 | 30,self_attn.k_proj,0.0006660341,0.01000,0.440 |
| 213 | 30,self_attn.v_proj,0.0006107283,0.01000,0.455 |
| 214 | 30,self_attn.q_proj,0.0023818917,0.01000,0.443 |
| 215 | 30,self_attn.o_proj,0.0000448454,0.01000,0.435 |
| 216 | 30,mlp.up_proj,0.0015934884,0.01000,0.517 |
| 217 | 30,mlp.gate_proj,0.0013594907,0.01000,0.494 |
| 218 | 30,mlp.down_proj,0.0003501898,0.01000,1.681 |
| 219 | 31,self_attn.k_proj,0.0007014952,0.01000,0.448 |
| 220 | 31,self_attn.v_proj,0.0006060585,0.01000,0.434 |
| 221 | 31,self_attn.q_proj,0.0023846514,0.01000,0.451 |
| 222 | 31,self_attn.o_proj,0.0000715734,0.01000,0.464 |
| 223 | 31,mlp.up_proj,0.0018080392,0.01000,0.497 |
| 224 | 31,mlp.gate_proj,0.0014998679,0.01000,0.518 |
| 225 | 31,mlp.down_proj,0.0004398297,0.01000,1.681 |
| 226 | 32,self_attn.k_proj,0.0007654972,0.01000,0.516 |
| 227 | 32,self_attn.v_proj,0.0008954310,0.01000,0.581 |
| 228 | 32,self_attn.q_proj,0.0028394967,0.01000,0.455 |
| 229 | 32,self_attn.o_proj,0.0000516377,0.01000,0.435 |
| 230 | 32,mlp.up_proj,0.0019804822,0.01000,0.535 |
| 231 | 32,mlp.gate_proj,0.0016116200,0.01000,0.539 |
| 232 | 32,mlp.down_proj,0.0006111872,0.01000,1.698 |
| 233 | 33,self_attn.k_proj,0.0006394647,0.01000,0.466 |
| 234 | 33,self_attn.v_proj,0.0006978745,0.01000,0.449 |
| 235 | 33,self_attn.q_proj,0.0023009892,0.01000,0.460 |
| 236 | 33,self_attn.o_proj,0.0000961008,0.01000,0.434 |
| 237 | 33,mlp.up_proj,0.0021798046,0.01000,0.520 |
| 238 | 33,mlp.gate_proj,0.0017339357,0.01000,0.526 |
| 239 | 33,mlp.down_proj,0.0008359131,0.01000,1.690 |
| 240 | 34,self_attn.k_proj,0.0006541725,0.01000,0.440 |
| 241 | 34,self_attn.v_proj,0.0011296207,0.01000,0.479 |
| 242 | 34,self_attn.q_proj,0.0025815739,0.01000,0.459 |
| 243 | 34,self_attn.o_proj,0.0001475956,0.01000,0.440 |
| 244 | 34,mlp.up_proj,0.0025971491,0.01000,0.569 |
| 245 | 34,mlp.gate_proj,0.0019881812,0.01000,0.502 |
| 246 | 34,mlp.down_proj,0.0010410425,0.01000,1.719 |
| 247 | 35,self_attn.k_proj,0.0005942605,0.01000,0.422 |
| 248 | 35,self_attn.v_proj,0.0012088416,0.01000,0.440 |
| 249 | 35,self_attn.q_proj,0.0025064127,0.01000,0.422 |
| 250 | 35,self_attn.o_proj,0.0001772174,0.01000,0.449 |
| 251 | 35,mlp.up_proj,0.0030033083,0.01000,0.513 |
| 252 | 35,mlp.gate_proj,0.0022048278,0.01000,0.495 |
| 253 | 35,mlp.down_proj,0.0015544999,0.01000,1.684 |
| 254 | 36,self_attn.k_proj,0.0005515935,0.01000,0.448 |
| 255 | 36,self_attn.v_proj,0.0019261715,0.01000,0.425 |
| 256 | 36,self_attn.q_proj,0.0024816982,0.01000,0.459 |
| 257 | 36,self_attn.o_proj,0.0001544181,0.01000,0.439 |
| 258 | 36,mlp.up_proj,0.0039325399,0.01000,0.496 |
| 259 | 36,mlp.gate_proj,0.0028782809,0.01000,0.522 |
| 260 | 36,mlp.down_proj,0.0022360407,0.01000,1.667 |
| 261 | 37,self_attn.k_proj,0.0005540850,0.01000,0.428 |
| 262 | 37,self_attn.v_proj,0.0023931593,0.01000,0.449 |
| 263 | 37,self_attn.q_proj,0.0026355596,0.01000,0.450 |
| 264 | 37,self_attn.o_proj,0.0002773240,0.01000,0.435 |
| 265 | 37,mlp.up_proj,0.0040397442,0.01000,0.527 |
| 266 | 37,mlp.gate_proj,0.0029825685,0.01000,0.515 |
| 267 | 37,mlp.down_proj,0.0030968198,0.01000,1.671 |
| 268 | 38,self_attn.k_proj,0.0004877125,0.01000,0.436 |
| 269 | 38,self_attn.v_proj,0.0039252046,0.01000,0.444 |
| 270 | 38,self_attn.q_proj,0.0026660137,0.01000,0.424 |
| 271 | 38,self_attn.o_proj,0.0005240002,0.01000,0.449 |
| 272 | 38,mlp.up_proj,0.0042532832,0.01000,0.500 |
| 273 | 38,mlp.gate_proj,0.0033421522,0.01000,0.502 |
| 274 | 38,mlp.down_proj,0.0052362741,0.01000,1.683 |
| 275 | 39,self_attn.k_proj,0.0005696468,0.01000,0.438 |
| 276 | 39,self_attn.v_proj,0.0035458919,0.01000,0.418 |
| 277 | 39,self_attn.q_proj,0.0035293297,0.01000,0.447 |
| 278 | 39,self_attn.o_proj,0.0016652486,0.01000,0.427 |
| 279 | 39,mlp.up_proj,0.0042222063,0.01000,0.524 |
| 280 | 39,mlp.gate_proj,0.0035428723,0.01000,0.524 |
| 281 | 39,mlp.down_proj,0.0119190907,0.01000,1.665 |