| layer,module,loss,samples,damp,time | |
| 0,self_attn.q_proj,0.0000003401,0.05000,2.345 | |
| 0,self_attn.k_proj,0.0000000715,0.05000,2.364 | |
| 0,self_attn.v_proj,0.0000000085,0.05000,2.368 | |
| 0,self_attn.o_proj,0.0000000277,0.05000,0.771 | |
| 0,mlp.gate_proj,0.0000003923,0.05000,1.354 | |
| 0,mlp.up_proj,0.0000003028,0.05000,1.358 | |
| 0,mlp.down_proj,0.0000000544,0.05000,4.481 | |
| 1,self_attn.k_proj,0.0000000173,0.05000,2.555 | |
| 1,self_attn.q_proj,0.0000000622,0.05000,2.587 | |
| 1,self_attn.v_proj,0.0000000033,0.05000,2.622 | |
| 1,self_attn.o_proj,0.0000000095,0.05000,0.750 | |
| 1,mlp.gate_proj,0.0000222209,0.05000,1.368 | |
| 1,mlp.up_proj,0.0000169856,0.05000,1.377 | |
| 1,mlp.down_proj,0.0000000056,0.05000,4.306 | |
| 2,self_attn.q_proj,0.0000000956,0.05000,2.530 | |
| 2,self_attn.k_proj,0.0000000210,0.05000,2.544 | |
| 2,self_attn.v_proj,0.0000000054,0.05000,2.568 | |
| 2,self_attn.o_proj,0.0000000061,0.05000,0.756 | |
| 2,mlp.gate_proj,0.0000048486,0.05000,1.414 | |
| 2,mlp.up_proj,0.0000049113,0.05000,1.432 | |
| 2,mlp.down_proj,0.0000085108,0.05000,4.318 | |
| 3,self_attn.q_proj,0.0000004073,0.05000,2.616 | |
| 3,self_attn.k_proj,0.0000000878,0.05000,2.670 | |
| 3,self_attn.v_proj,0.0000000161,0.05000,2.678 | |
| 3,self_attn.o_proj,0.0000000094,0.05000,0.782 | |
| 3,mlp.up_proj,0.0000058005,0.05000,1.375 | |
| 3,mlp.gate_proj,0.0000064317,0.05000,1.390 | |
| 3,mlp.down_proj,0.0000001039,0.05000,4.202 | |
| 4,self_attn.q_proj,0.0000003292,0.05000,2.495 | |
| 4,self_attn.v_proj,0.0000000221,0.05000,2.539 | |
| 4,self_attn.k_proj,0.0000000643,0.05000,2.549 | |
| 4,self_attn.o_proj,0.0000000181,0.05000,0.765 | |
| 4,mlp.gate_proj,0.0000060889,0.05000,1.356 | |
| 4,mlp.up_proj,0.0000046043,0.05000,1.358 | |
| 4,mlp.down_proj,0.0000000563,0.05000,4.267 | |
| 5,self_attn.k_proj,0.0000001253,0.05000,2.652 | |
| 5,self_attn.v_proj,0.0000000582,0.05000,2.677 | |
| 5,self_attn.q_proj,0.0000007062,0.05000,2.693 | |
| 5,self_attn.o_proj,0.0000000114,0.05000,0.745 | |
| 5,mlp.gate_proj,0.0000084792,0.05000,1.395 | |
| 5,mlp.up_proj,0.0000071883,0.05000,1.400 | |
| 5,mlp.down_proj,0.0000001014,0.05000,4.247 | |
| 6,self_attn.k_proj,0.0000000723,0.05000,2.712 | |
| 6,self_attn.q_proj,0.0000004542,0.05000,2.727 | |
| 6,self_attn.v_proj,0.0000000503,0.05000,2.746 | |
| 6,self_attn.o_proj,0.0000000305,0.05000,0.753 | |
| 6,mlp.gate_proj,0.0000109521,0.05000,1.381 | |
| 6,mlp.up_proj,0.0000094474,0.05000,1.390 | |
| 6,mlp.down_proj,0.0000001409,0.05000,4.240 | |
| 7,self_attn.q_proj,0.0000006046,0.05000,2.634 | |
| 7,self_attn.k_proj,0.0000000955,0.05000,2.698 | |
| 7,self_attn.v_proj,0.0000000611,0.05000,2.720 | |
| 7,self_attn.o_proj,0.0000000201,0.05000,0.758 | |
| 7,mlp.gate_proj,0.0000111783,0.05000,1.347 | |
| 7,mlp.up_proj,0.0000092888,0.05000,1.354 | |
| 7,mlp.down_proj,0.0000001957,0.05000,4.267 | |
| 8,self_attn.q_proj,0.0000006208,0.05000,2.460 | |
| 8,self_attn.k_proj,0.0000000936,0.05000,2.473 | |
| 8,self_attn.v_proj,0.0000000658,0.05000,2.481 | |
| 8,self_attn.o_proj,0.0000000297,0.05000,0.755 | |
| 8,mlp.up_proj,0.0000088363,0.05000,1.354 | |
| 8,mlp.gate_proj,0.0000123492,0.05000,1.358 | |
| 8,mlp.down_proj,0.0000002615,0.05000,4.249 | |
| 9,self_attn.q_proj,0.0000005419,0.05000,2.480 | |
| 9,self_attn.k_proj,0.0000000949,0.05000,2.487 | |
| 9,self_attn.v_proj,0.0000000561,0.05000,2.517 | |
| 9,self_attn.o_proj,0.0000000487,0.05000,0.757 | |
| 9,mlp.gate_proj,0.0000116062,0.05000,1.368 | |
| 9,mlp.up_proj,0.0000078157,0.05000,1.375 | |
| 9,mlp.down_proj,0.0000003051,0.05000,4.267 | |
| 10,self_attn.q_proj,0.0000007660,0.05000,2.545 | |
| 10,self_attn.v_proj,0.0000000931,0.05000,2.556 | |
| 10,self_attn.k_proj,0.0000001141,0.05000,2.560 | |
| 10,self_attn.o_proj,0.0000000494,0.05000,0.749 | |
| 10,mlp.gate_proj,0.0000087138,0.05000,1.360 | |
| 10,mlp.up_proj,0.0000057507,0.05000,1.366 | |
| 10,mlp.down_proj,0.0000003635,0.05000,4.293 | |
| 11,self_attn.q_proj,0.0000005695,0.05000,2.428 | |
| 11,self_attn.v_proj,0.0000000758,0.05000,2.446 | |
| 11,self_attn.k_proj,0.0000000838,0.05000,2.454 | |
| 11,self_attn.o_proj,0.0000000737,0.05000,0.767 | |
| 11,mlp.gate_proj,0.0000047200,0.05000,1.349 | |
| 11,mlp.up_proj,0.0000033895,0.05000,1.357 | |
| 11,mlp.down_proj,0.0000003966,0.05000,4.351 | |
| 12,self_attn.k_proj,0.0000000907,0.05000,2.451 | |
| 12,self_attn.v_proj,0.0000000689,0.05000,2.479 | |
| 12,self_attn.q_proj,0.0000005998,0.05000,2.492 | |
| 12,self_attn.o_proj,0.0000000781,0.05000,0.750 | |
| 12,mlp.gate_proj,0.0000052030,0.05000,1.376 | |
| 12,mlp.up_proj,0.0000036776,0.05000,1.383 | |
| 12,mlp.down_proj,0.0000003800,0.05000,4.308 | |
| 13,self_attn.k_proj,0.0000361445,0.05000,2.567 | |
| 13,self_attn.q_proj,0.0001988869,0.05000,2.576 | |
| 13,self_attn.v_proj,0.0000131335,0.05000,2.590 | |
| 13,self_attn.o_proj,0.0000135128,0.05000,0.761 | |
| 13,mlp.up_proj,0.0007943917,0.05000,1.410 | |
| 13,mlp.gate_proj,0.0008288599,0.05000,1.423 | |
| 13,mlp.down_proj,0.0000983028,0.05000,4.245 | |
| 14,self_attn.k_proj,0.0000000836,0.05000,2.513 | |
| 14,self_attn.v_proj,0.0000000555,0.05000,2.535 | |
| 14,self_attn.q_proj,0.0000005526,0.05000,2.542 | |
| 14,self_attn.o_proj,0.0000001077,0.05000,0.758 | |
| 14,mlp.gate_proj,0.0000027660,0.05000,1.381 | |
| 14,mlp.up_proj,0.0000026624,0.05000,1.387 | |
| 14,mlp.down_proj,0.0000003076,0.05000,4.313 | |
| 15,self_attn.v_proj,0.0000000538,0.05000,2.525 | |
| 15,self_attn.k_proj,0.0000000867,0.05000,2.572 | |
| 15,self_attn.q_proj,0.0000005570,0.05000,2.578 | |
| 15,self_attn.o_proj,0.0000000963,0.05000,0.750 | |
| 15,mlp.up_proj,0.0000023406,0.05000,1.377 | |
| 15,mlp.gate_proj,0.0000022434,0.05000,1.380 | |
| 15,mlp.down_proj,0.0000002715,0.05000,4.259 | |
| 16,self_attn.q_proj,0.0001674280,0.05000,2.427 | |
| 16,self_attn.v_proj,0.0000162792,0.05000,2.520 | |
| 16,self_attn.k_proj,0.0000277001,0.05000,2.537 | |
| 16,self_attn.o_proj,0.0000293571,0.05000,0.758 | |
| 16,mlp.up_proj,0.0006605779,0.05000,1.385 | |
| 16,mlp.gate_proj,0.0006744062,0.05000,1.399 | |
| 16,mlp.down_proj,0.0000691259,0.05000,4.257 | |
| 17,self_attn.k_proj,0.0000001538,0.05000,2.463 | |
| 17,self_attn.q_proj,0.0000009075,0.05000,2.471 | |
| 17,self_attn.v_proj,0.0000000858,0.05000,2.469 | |
| 17,self_attn.o_proj,0.0000000672,0.05000,0.773 | |
| 17,mlp.up_proj,0.0000020540,0.05000,1.401 | |
| 17,mlp.gate_proj,0.0000020261,0.05000,1.408 | |
| 17,mlp.down_proj,0.0000002283,0.05000,4.257 | |
| 18,self_attn.v_proj,0.0000000613,0.05000,2.509 | |
| 18,self_attn.k_proj,0.0000000836,0.05000,2.535 | |
| 18,self_attn.q_proj,0.0000005755,0.05000,2.543 | |
| 18,self_attn.o_proj,0.0000000861,0.05000,0.763 | |
| 18,mlp.gate_proj,0.0000020542,0.05000,1.371 | |
| 18,mlp.up_proj,0.0000019867,0.05000,1.384 | |
| 18,mlp.down_proj,0.0000002142,0.05000,4.281 | |
| 19,self_attn.v_proj,0.0000000633,0.05000,2.534 | |
| 19,self_attn.k_proj,0.0000001105,0.05000,2.549 | |
| 19,self_attn.q_proj,0.0000006358,0.05000,2.559 | |
| 19,self_attn.o_proj,0.0000001115,0.05000,0.778 | |
| 19,mlp.gate_proj,0.0000019188,0.05000,1.345 | |
| 19,mlp.up_proj,0.0000020043,0.05000,1.350 | |
| 19,mlp.down_proj,0.0000002130,0.05000,4.248 | |
| 20,self_attn.v_proj,0.0000001238,0.05000,2.480 | |
| 20,self_attn.q_proj,0.0000009841,0.05000,2.509 | |
| 20,self_attn.k_proj,0.0000001128,0.05000,2.515 | |
| 20,self_attn.o_proj,0.0000000938,0.05000,0.753 | |
| 20,mlp.gate_proj,0.0000019638,0.05000,1.350 | |
| 20,mlp.up_proj,0.0000019977,0.05000,1.360 | |
| 20,mlp.down_proj,0.0000002298,0.05000,4.285 | |
| 21,self_attn.v_proj,0.0000000831,0.05000,2.684 | |
| 21,self_attn.k_proj,0.0000001138,0.05000,2.703 | |
| 21,self_attn.q_proj,0.0000007205,0.05000,2.715 | |
| 21,self_attn.o_proj,0.0000001049,0.05000,0.765 | |
| 21,mlp.up_proj,0.0000019671,0.05000,1.369 | |
| 21,mlp.gate_proj,0.0000021018,0.05000,1.374 | |
| 21,mlp.down_proj,0.0000002018,0.05000,4.270 | |
| 22,self_attn.v_proj,0.0000001127,0.05000,2.530 | |
| 22,self_attn.q_proj,0.0000008507,0.05000,2.575 | |
| 22,self_attn.k_proj,0.0000001312,0.05000,2.581 | |
| 22,self_attn.o_proj,0.0000000927,0.05000,0.761 | |
| 22,mlp.gate_proj,0.0000021110,0.05000,1.372 | |
| 22,mlp.up_proj,0.0000020616,0.05000,1.376 | |
| 22,mlp.down_proj,0.0000002235,0.05000,4.254 | |
| 23,self_attn.v_proj,0.0000000670,0.05000,2.514 | |
| 23,self_attn.k_proj,0.0000001256,0.05000,2.541 | |
| 23,self_attn.q_proj,0.0000008720,0.05000,2.550 | |
| 23,self_attn.o_proj,0.0000000902,0.05000,0.759 | |
| 23,mlp.gate_proj,0.0000021169,0.05000,1.350 | |
| 23,mlp.up_proj,0.0000022067,0.05000,1.356 | |
| 23,mlp.down_proj,0.0000002553,0.05000,4.296 | |
| 24,self_attn.k_proj,0.0000255149,0.05000,2.405 | |
| 24,self_attn.v_proj,0.0000224293,0.05000,2.434 | |
| 24,self_attn.q_proj,0.0001882081,0.05000,2.442 | |
| 24,self_attn.o_proj,0.0000325786,0.05000,0.769 | |
| 24,mlp.up_proj,0.0006212120,0.05000,1.378 | |
| 24,mlp.gate_proj,0.0006047986,0.05000,1.390 | |
| 24,mlp.down_proj,0.0000792528,0.05000,4.284 | |
| 25,self_attn.v_proj,0.0000001261,0.05000,2.402 | |
| 25,self_attn.k_proj,0.0000000865,0.05000,2.413 | |
| 25,self_attn.q_proj,0.0000009028,0.05000,2.425 | |
| 25,self_attn.o_proj,0.0000000856,0.05000,0.752 | |
| 25,mlp.gate_proj,0.0000023320,0.05000,1.363 | |
| 25,mlp.up_proj,0.0000023642,0.05000,1.368 | |
| 25,mlp.down_proj,0.0000003521,0.05000,4.316 | |
| 26,self_attn.v_proj,0.0000345133,0.05000,2.484 | |
| 26,self_attn.q_proj,0.0002079831,0.05000,2.532 | |
| 26,self_attn.k_proj,0.0000266343,0.05000,2.539 | |
| 26,self_attn.o_proj,0.0000372935,0.05000,0.759 | |
| 26,mlp.up_proj,0.0007854277,0.05000,1.365 | |
| 26,mlp.gate_proj,0.0007399612,0.05000,1.366 | |
| 26,mlp.down_proj,0.0001390844,0.05000,4.290 | |
| 27,self_attn.q_proj,0.0000012426,0.05000,2.442 | |
| 27,self_attn.k_proj,0.0000001063,0.05000,2.455 | |
| 27,self_attn.v_proj,0.0000001755,0.05000,2.464 | |
| 27,self_attn.o_proj,0.0000001954,0.05000,0.743 | |
| 27,mlp.gate_proj,0.0000029087,0.05000,1.346 | |
| 27,mlp.up_proj,0.0000029363,0.05000,1.359 | |
| 27,mlp.down_proj,0.0000006022,0.05000,4.340 | |
| 28,self_attn.k_proj,0.0000283726,0.05000,2.394 | |
| 28,self_attn.v_proj,0.0000471470,0.05000,2.405 | |
| 28,self_attn.q_proj,0.0002638888,0.05000,2.417 | |
| 28,self_attn.o_proj,0.0000527611,0.05000,0.737 | |
| 28,mlp.gate_proj,0.0009632122,0.05000,1.353 | |
| 28,mlp.up_proj,0.0009681229,0.05000,1.360 | |
| 28,mlp.down_proj,0.0002143012,0.05000,4.288 | |
| 29,self_attn.v_proj,0.0000444782,0.05000,2.417 | |
| 29,self_attn.k_proj,0.0000306777,0.05000,2.451 | |
| 29,self_attn.q_proj,0.0002660261,0.05000,2.460 | |
| 29,self_attn.o_proj,0.0000419030,0.05000,0.753 | |
| 29,mlp.gate_proj,0.0011661545,0.05000,1.349 | |
| 29,mlp.up_proj,0.0011946059,0.05000,1.355 | |
| 29,mlp.down_proj,0.0002983677,0.05000,4.321 | |
| 30,self_attn.v_proj,0.0000003594,0.05000,2.520 | |
| 30,self_attn.k_proj,0.0000001052,0.05000,2.556 | |
| 30,self_attn.q_proj,0.0000012281,0.05000,2.563 | |
| 30,self_attn.o_proj,0.0000001564,0.05000,0.763 | |
| 30,mlp.gate_proj,0.0000051266,0.05000,1.384 | |
| 30,mlp.up_proj,0.0000055988,0.05000,1.389 | |
| 30,mlp.down_proj,0.0000080552,0.05000,4.288 | |
| 31,self_attn.q_proj,0.0000014235,0.05000,2.486 | |
| 31,self_attn.k_proj,0.0000001413,0.05000,2.555 | |
| 31,self_attn.v_proj,0.0000003790,0.05000,2.578 | |
| 31,self_attn.o_proj,0.0000003545,0.05000,0.753 | |
| 31,mlp.gate_proj,0.0000059219,0.05000,1.385 | |
| 31,mlp.up_proj,0.0000067765,0.05000,1.389 | |
| 31,mlp.down_proj,0.0000033691,0.05000,4.295 | |
| 32,self_attn.q_proj,0.0000022495,0.05000,2.460 | |
| 32,self_attn.k_proj,0.0000001965,0.05000,2.467 | |
| 32,self_attn.v_proj,0.0000009772,0.05000,2.481 | |
| 32,self_attn.o_proj,0.0000003497,0.05000,0.767 | |
| 32,mlp.gate_proj,0.0000061569,0.05000,1.398 | |
| 32,mlp.up_proj,0.0000070279,0.05000,1.404 | |
| 32,mlp.down_proj,0.0000036677,0.05000,4.273 | |
| 33,self_attn.k_proj,0.0000001776,0.05000,2.578 | |
| 33,self_attn.v_proj,0.0000021733,0.05000,2.592 | |
| 33,self_attn.q_proj,0.0000022470,0.05000,2.597 | |
| 33,self_attn.o_proj,0.0000004441,0.05000,0.758 | |
| 33,mlp.gate_proj,0.0000059196,0.05000,1.358 | |
| 33,mlp.up_proj,0.0000074468,0.05000,1.363 | |
| 33,mlp.down_proj,0.0000068785,0.05000,4.258 | |
| 34,self_attn.k_proj,0.0000001017,0.05000,2.520 | |
| 34,self_attn.q_proj,0.0000009964,0.05000,2.541 | |
| 34,self_attn.v_proj,0.0000003421,0.05000,2.558 | |
| 34,self_attn.o_proj,0.0000004992,0.05000,0.785 | |
| 34,mlp.up_proj,0.0000077734,0.05000,1.347 | |
| 34,mlp.gate_proj,0.0000066457,0.05000,1.352 | |
| 34,mlp.down_proj,0.0000052847,0.05000,4.240 | |
| 35,self_attn.k_proj,0.0000001003,0.05000,2.693 | |
| 35,self_attn.v_proj,0.0000003007,0.05000,2.721 | |
| 35,self_attn.q_proj,0.0000009509,0.05000,2.724 | |
| 35,self_attn.o_proj,0.0000005282,0.05000,0.765 | |
| 35,mlp.gate_proj,0.0000101811,0.05000,1.361 | |
| 35,mlp.up_proj,0.0000107798,0.05000,1.373 | |
| 35,mlp.down_proj,0.0000105316,0.05000,4.297 | |