| layer,module,loss,samples,damp,time | |
| 0,self_attn.q_proj,0.0000003401,0.05000,2.456 | |
| 0,self_attn.k_proj,0.0000000715,0.05000,2.511 | |
| 0,self_attn.v_proj,0.0000000085,0.05000,2.522 | |
| 0,self_attn.o_proj,0.0000000277,0.05000,0.727 | |
| 0,mlp.up_proj,0.0000003028,0.05000,1.351 | |
| 0,mlp.gate_proj,0.0000003923,0.05000,1.363 | |
| 0,mlp.down_proj,0.0000000544,0.05000,4.372 | |
| 1,self_attn.k_proj,0.0000000173,0.05000,2.496 | |
| 1,self_attn.v_proj,0.0000000033,0.05000,2.535 | |
| 1,self_attn.q_proj,0.0000000622,0.05000,2.537 | |
| 1,self_attn.o_proj,0.0000000095,0.05000,0.742 | |
| 1,mlp.up_proj,0.0000169856,0.05000,1.392 | |
| 1,mlp.gate_proj,0.0000222209,0.05000,1.398 | |
| 1,mlp.down_proj,0.0000000056,0.05000,4.177 | |
| 2,self_attn.k_proj,0.0000000210,0.05000,2.543 | |
| 2,self_attn.q_proj,0.0000000956,0.05000,2.564 | |
| 2,self_attn.v_proj,0.0000000054,0.05000,2.588 | |
| 2,self_attn.o_proj,0.0000000061,0.05000,0.739 | |
| 2,mlp.gate_proj,0.0000048486,0.05000,1.366 | |
| 2,mlp.up_proj,0.0000049113,0.05000,1.373 | |
| 2,mlp.down_proj,0.0000085108,0.05000,4.199 | |
| 3,self_attn.v_proj,0.0000000161,0.05000,2.525 | |
| 3,self_attn.k_proj,0.0000000878,0.05000,2.538 | |
| 3,self_attn.q_proj,0.0000004073,0.05000,2.542 | |
| 3,self_attn.o_proj,0.0000000094,0.05000,0.753 | |
| 3,mlp.gate_proj,0.0000064317,0.05000,1.357 | |
| 3,mlp.up_proj,0.0000058005,0.05000,1.365 | |
| 3,mlp.down_proj,0.0000001039,0.05000,4.192 | |
| 4,self_attn.q_proj,0.0000003292,0.05000,2.325 | |
| 4,self_attn.v_proj,0.0000000221,0.05000,2.356 | |
| 4,self_attn.k_proj,0.0000000643,0.05000,2.361 | |
| 4,self_attn.o_proj,0.0000000181,0.05000,0.750 | |
| 4,mlp.gate_proj,0.0000060889,0.05000,1.348 | |
| 4,mlp.up_proj,0.0000046043,0.05000,1.352 | |
| 4,mlp.down_proj,0.0000000563,0.05000,4.266 | |
| 5,self_attn.v_proj,0.0000000582,0.05000,2.441 | |
| 5,self_attn.q_proj,0.0000007062,0.05000,2.460 | |
| 5,self_attn.k_proj,0.0000001253,0.05000,2.465 | |
| 5,self_attn.o_proj,0.0000000114,0.05000,0.750 | |
| 5,mlp.gate_proj,0.0000084792,0.05000,1.368 | |
| 5,mlp.up_proj,0.0000071883,0.05000,1.372 | |
| 5,mlp.down_proj,0.0000001014,0.05000,4.254 | |
| 6,self_attn.q_proj,0.0000004542,0.05000,2.507 | |
| 6,self_attn.v_proj,0.0000000503,0.05000,2.529 | |
| 6,self_attn.k_proj,0.0000000723,0.05000,2.543 | |
| 6,self_attn.o_proj,0.0000000305,0.05000,0.750 | |
| 6,mlp.gate_proj,0.0000109521,0.05000,1.369 | |
| 6,mlp.up_proj,0.0000094474,0.05000,1.381 | |
| 6,mlp.down_proj,0.0000001409,0.05000,4.289 | |
| 7,self_attn.k_proj,0.0000000955,0.05000,2.519 | |
| 7,self_attn.q_proj,0.0000006046,0.05000,2.523 | |
| 7,self_attn.v_proj,0.0000000611,0.05000,2.539 | |
| 7,self_attn.o_proj,0.0000000201,0.05000,0.749 | |
| 7,mlp.up_proj,0.0000092888,0.05000,1.355 | |
| 7,mlp.gate_proj,0.0000111783,0.05000,1.360 | |
| 7,mlp.down_proj,0.0000001957,0.05000,4.250 | |
| 8,self_attn.v_proj,0.0000000658,0.05000,2.420 | |
| 8,self_attn.k_proj,0.0000000936,0.05000,2.438 | |
| 8,self_attn.q_proj,0.0000006208,0.05000,2.456 | |
| 8,self_attn.o_proj,0.0000000297,0.05000,0.741 | |
| 8,mlp.up_proj,0.0000088363,0.05000,1.376 | |
| 8,mlp.gate_proj,0.0000123492,0.05000,1.387 | |
| 8,mlp.down_proj,0.0000002615,0.05000,4.274 | |
| 9,self_attn.v_proj,0.0000000561,0.05000,2.445 | |
| 9,self_attn.k_proj,0.0000000949,0.05000,2.457 | |
| 9,self_attn.q_proj,0.0000005419,0.05000,2.473 | |
| 9,self_attn.o_proj,0.0000000487,0.05000,0.754 | |
| 9,mlp.up_proj,0.0000078157,0.05000,1.364 | |
| 9,mlp.gate_proj,0.0000116062,0.05000,1.371 | |
| 9,mlp.down_proj,0.0000003051,0.05000,4.189 | |
| 10,self_attn.q_proj,0.0000007660,0.05000,2.538 | |
| 10,self_attn.v_proj,0.0000000931,0.05000,2.554 | |
| 10,self_attn.k_proj,0.0000001141,0.05000,2.582 | |
| 10,self_attn.o_proj,0.0000000494,0.05000,0.754 | |
| 10,mlp.gate_proj,0.0000087138,0.05000,1.354 | |
| 10,mlp.up_proj,0.0000057507,0.05000,1.358 | |
| 10,mlp.down_proj,0.0000003635,0.05000,4.210 | |
| 11,self_attn.q_proj,0.0000005695,0.05000,2.455 | |
| 11,self_attn.k_proj,0.0000000838,0.05000,2.469 | |
| 11,self_attn.v_proj,0.0000000758,0.05000,2.473 | |
| 11,self_attn.o_proj,0.0000000737,0.05000,0.758 | |
| 11,mlp.up_proj,0.0000033895,0.05000,1.429 | |
| 11,mlp.gate_proj,0.0000047200,0.05000,1.434 | |
| 11,mlp.down_proj,0.0000003966,0.05000,4.243 | |
| 12,self_attn.v_proj,0.0000000689,0.05000,2.344 | |
| 12,self_attn.k_proj,0.0000000907,0.05000,2.375 | |
| 12,self_attn.q_proj,0.0000005998,0.05000,2.376 | |
| 12,self_attn.o_proj,0.0000000781,0.05000,0.744 | |
| 12,mlp.up_proj,0.0000036776,0.05000,1.390 | |
| 12,mlp.gate_proj,0.0000052030,0.05000,1.391 | |
| 12,mlp.down_proj,0.0000003800,0.05000,4.251 | |
| 13,self_attn.k_proj,0.0000361445,0.05000,2.470 | |
| 13,self_attn.q_proj,0.0001988869,0.05000,2.479 | |
| 13,self_attn.v_proj,0.0000131335,0.05000,2.485 | |
| 13,self_attn.o_proj,0.0000135128,0.05000,0.745 | |
| 13,mlp.gate_proj,0.0008288599,0.05000,1.382 | |
| 13,mlp.up_proj,0.0007943917,0.05000,1.387 | |
| 13,mlp.down_proj,0.0000983028,0.05000,4.216 | |
| 14,self_attn.v_proj,0.0000000555,0.05000,2.447 | |
| 14,self_attn.k_proj,0.0000000836,0.05000,2.523 | |
| 14,self_attn.q_proj,0.0000005526,0.05000,2.541 | |
| 14,self_attn.o_proj,0.0000001077,0.05000,0.754 | |
| 14,mlp.gate_proj,0.0000027660,0.05000,1.384 | |
| 14,mlp.up_proj,0.0000026624,0.05000,1.388 | |
| 14,mlp.down_proj,0.0000003076,0.05000,4.218 | |
| 15,self_attn.k_proj,0.0000000867,0.05000,2.515 | |
| 15,self_attn.v_proj,0.0000000538,0.05000,2.525 | |
| 15,self_attn.q_proj,0.0000005570,0.05000,2.528 | |
| 15,self_attn.o_proj,0.0000000963,0.05000,0.751 | |
| 15,mlp.gate_proj,0.0000022434,0.05000,1.351 | |
| 15,mlp.up_proj,0.0000023406,0.05000,1.354 | |
| 15,mlp.down_proj,0.0000002715,0.05000,4.256 | |
| 16,self_attn.v_proj,0.0000000568,0.05000,2.429 | |
| 16,self_attn.k_proj,0.0000000957,0.05000,2.439 | |
| 16,self_attn.q_proj,0.0000005776,0.05000,2.444 | |
| 16,self_attn.o_proj,0.0000001003,0.05000,0.751 | |
| 16,mlp.up_proj,0.0000022801,0.05000,1.354 | |
| 16,mlp.gate_proj,0.0000023282,0.05000,1.369 | |
| 16,mlp.down_proj,0.0000002357,0.05000,4.220 | |
| 17,self_attn.q_proj,0.0000009054,0.05000,2.522 | |
| 17,self_attn.v_proj,0.0000000854,0.05000,2.539 | |
| 17,self_attn.k_proj,0.0000001532,0.05000,2.547 | |
| 17,self_attn.o_proj,0.0000000666,0.05000,0.750 | |
| 17,mlp.up_proj,0.0000020422,0.05000,1.365 | |
| 17,mlp.gate_proj,0.0000020144,0.05000,1.378 | |
| 17,mlp.down_proj,0.0000002270,0.05000,4.209 | |
| 18,self_attn.q_proj,0.0000005731,0.05000,2.550 | |
| 18,self_attn.v_proj,0.0000000610,0.05000,2.625 | |
| 18,self_attn.k_proj,0.0000000831,0.05000,2.640 | |
| 18,self_attn.o_proj,0.0000000860,0.05000,0.748 | |
| 18,mlp.up_proj,0.0000019781,0.05000,1.387 | |
| 18,mlp.gate_proj,0.0000020457,0.05000,1.392 | |
| 18,mlp.down_proj,0.0000002135,0.05000,4.196 | |
| 19,self_attn.v_proj,0.0000000631,0.05000,2.677 | |
| 19,self_attn.q_proj,0.0000006335,0.05000,2.683 | |
| 19,self_attn.k_proj,0.0000001105,0.05000,2.701 | |
| 19,self_attn.o_proj,0.0000001114,0.05000,0.759 | |
| 19,mlp.up_proj,0.0000019969,0.05000,1.361 | |
| 19,mlp.gate_proj,0.0000019120,0.05000,1.369 | |
| 19,mlp.down_proj,0.0000002123,0.05000,4.198 | |
| 20,self_attn.q_proj,0.0000009823,0.05000,2.507 | |
| 20,self_attn.k_proj,0.0000001125,0.05000,2.513 | |
| 20,self_attn.v_proj,0.0000001236,0.05000,2.525 | |
| 20,self_attn.o_proj,0.0000000935,0.05000,0.743 | |
| 20,mlp.gate_proj,0.0000019580,0.05000,1.357 | |
| 20,mlp.up_proj,0.0000019916,0.05000,1.363 | |
| 20,mlp.down_proj,0.0000002291,0.05000,4.209 | |
| 21,self_attn.v_proj,0.0000000829,0.05000,2.504 | |
| 21,self_attn.k_proj,0.0000001137,0.05000,2.514 | |
| 21,self_attn.q_proj,0.0000007184,0.05000,2.522 | |
| 21,self_attn.o_proj,0.0000001047,0.05000,0.737 | |
| 21,mlp.up_proj,0.0000019637,0.05000,1.380 | |
| 21,mlp.gate_proj,0.0000020983,0.05000,1.383 | |
| 21,mlp.down_proj,0.0000002016,0.05000,4.206 | |
| 22,self_attn.v_proj,0.0000001125,0.05000,2.485 | |
| 22,self_attn.q_proj,0.0000008492,0.05000,2.484 | |
| 22,self_attn.k_proj,0.0000001309,0.05000,2.501 | |
| 22,self_attn.o_proj,0.0000000925,0.05000,0.743 | |
| 22,mlp.gate_proj,0.0000021090,0.05000,1.340 | |
| 22,mlp.up_proj,0.0000020592,0.05000,1.343 | |
| 22,mlp.down_proj,0.0000002234,0.05000,4.213 | |
| 23,self_attn.q_proj,0.0000008715,0.05000,2.384 | |
| 23,self_attn.k_proj,0.0000001256,0.05000,2.441 | |
| 23,self_attn.v_proj,0.0000000670,0.05000,2.447 | |
| 23,self_attn.o_proj,0.0000000901,0.05000,0.747 | |
| 23,mlp.gate_proj,0.0000021141,0.05000,1.354 | |
| 23,mlp.up_proj,0.0000022042,0.05000,1.364 | |
| 23,mlp.down_proj,0.0000002550,0.05000,4.255 | |
| 24,self_attn.q_proj,0.0001880032,0.05000,2.437 | |
| 24,self_attn.k_proj,0.0000255162,0.05000,2.487 | |
| 24,self_attn.v_proj,0.0000224003,0.05000,2.494 | |
| 24,self_attn.o_proj,0.0000325736,0.05000,0.754 | |
| 24,mlp.gate_proj,0.0006041626,0.05000,1.362 | |
| 24,mlp.up_proj,0.0006203335,0.05000,1.367 | |
| 24,mlp.down_proj,0.0000791926,0.05000,4.315 | |
| 25,self_attn.k_proj,0.0000000866,0.05000,2.411 | |
| 25,self_attn.v_proj,0.0000001260,0.05000,2.444 | |
| 25,self_attn.q_proj,0.0000009027,0.05000,2.451 | |
| 25,self_attn.o_proj,0.0000000853,0.05000,0.754 | |
| 25,mlp.gate_proj,0.0000023311,0.05000,1.374 | |
| 25,mlp.up_proj,0.0000023624,0.05000,1.379 | |
| 25,mlp.down_proj,0.0000003523,0.05000,4.301 | |
| 26,self_attn.k_proj,0.0000000916,0.05000,2.474 | |
| 26,self_attn.v_proj,0.0000001193,0.05000,2.494 | |
| 26,self_attn.q_proj,0.0000007168,0.05000,2.504 | |
| 26,self_attn.o_proj,0.0000001270,0.05000,0.745 | |
| 26,mlp.up_proj,0.0000027133,0.05000,1.352 | |
| 26,mlp.gate_proj,0.0000025628,0.05000,1.356 | |
| 26,mlp.down_proj,0.0000004770,0.05000,4.249 | |
| 27,self_attn.v_proj,0.0000001746,0.05000,2.479 | |
| 27,self_attn.k_proj,0.0000001062,0.05000,2.514 | |
| 27,self_attn.q_proj,0.0000012390,0.05000,2.537 | |
| 27,self_attn.o_proj,0.0000001945,0.05000,0.754 | |
| 27,mlp.up_proj,0.0000029269,0.05000,1.367 | |
| 27,mlp.gate_proj,0.0000028995,0.05000,1.372 | |
| 27,mlp.down_proj,0.0000006006,0.05000,4.223 | |
| 28,self_attn.v_proj,0.0000001665,0.05000,2.396 | |
| 28,self_attn.k_proj,0.0000000975,0.05000,2.404 | |
| 28,self_attn.q_proj,0.0000009054,0.05000,2.413 | |
| 28,self_attn.o_proj,0.0000001827,0.05000,0.765 | |
| 28,mlp.gate_proj,0.0000033200,0.05000,1.357 | |
| 28,mlp.up_proj,0.0000033334,0.05000,1.360 | |
| 28,mlp.down_proj,0.0000007344,0.05000,4.210 | |
| 29,self_attn.k_proj,0.0000305678,0.05000,2.542 | |
| 29,self_attn.q_proj,0.0002649824,0.05000,2.574 | |
| 29,self_attn.v_proj,0.0000443468,0.05000,2.591 | |
| 29,self_attn.o_proj,0.0000417074,0.05000,0.756 | |
| 29,mlp.up_proj,0.0011906463,0.05000,1.378 | |
| 29,mlp.gate_proj,0.0011616761,0.05000,1.384 | |
| 29,mlp.down_proj,0.0002980117,0.05000,4.286 | |
| 30,self_attn.q_proj,0.0000012240,0.05000,2.520 | |
| 30,self_attn.v_proj,0.0000003584,0.05000,2.549 | |
| 30,self_attn.k_proj,0.0000001048,0.05000,2.563 | |
| 30,self_attn.o_proj,0.0000001559,0.05000,0.753 | |
| 30,mlp.up_proj,0.0000055833,0.05000,1.423 | |
| 30,mlp.gate_proj,0.0000051110,0.05000,1.428 | |
| 30,mlp.down_proj,0.0000081476,0.05000,4.274 | |
| 31,self_attn.k_proj,0.0000001408,0.05000,2.504 | |
| 31,self_attn.q_proj,0.0000014188,0.05000,2.516 | |
| 31,self_attn.v_proj,0.0000003782,0.05000,2.525 | |
| 31,self_attn.o_proj,0.0000003638,0.05000,0.748 | |
| 31,mlp.up_proj,0.0000067382,0.05000,1.376 | |
| 31,mlp.gate_proj,0.0000058890,0.05000,1.390 | |
| 31,mlp.down_proj,0.0000033553,0.05000,4.233 | |
| 32,self_attn.q_proj,0.0000022478,0.05000,2.452 | |
| 32,self_attn.k_proj,0.0000001967,0.05000,2.457 | |
| 32,self_attn.v_proj,0.0000009761,0.05000,2.464 | |
| 32,self_attn.o_proj,0.0000003565,0.05000,0.755 | |
| 32,mlp.gate_proj,0.0000061484,0.05000,1.382 | |
| 32,mlp.up_proj,0.0000070169,0.05000,1.385 | |
| 32,mlp.down_proj,0.0000036669,0.05000,4.195 | |
| 33,self_attn.k_proj,0.0000001779,0.05000,2.507 | |
| 33,self_attn.v_proj,0.0000021782,0.05000,2.514 | |
| 33,self_attn.q_proj,0.0000022511,0.05000,2.517 | |
| 33,self_attn.o_proj,0.0000004571,0.05000,0.752 | |
| 33,mlp.gate_proj,0.0000059207,0.05000,1.364 | |
| 33,mlp.up_proj,0.0000074530,0.05000,1.369 | |
| 33,mlp.down_proj,0.0000068917,0.05000,4.231 | |
| 34,self_attn.v_proj,0.0000003427,0.05000,2.650 | |
| 34,self_attn.q_proj,0.0000009977,0.05000,2.765 | |
| 34,self_attn.k_proj,0.0000001021,0.05000,2.776 | |
| 34,self_attn.o_proj,0.0000005056,0.05000,0.760 | |
| 34,mlp.gate_proj,0.0000066525,0.05000,1.370 | |
| 34,mlp.up_proj,0.0000077802,0.05000,1.373 | |
| 34,mlp.down_proj,0.0000052944,0.05000,4.182 | |
| 35,self_attn.k_proj,0.0000001002,0.05000,2.618 | |
| 35,self_attn.q_proj,0.0000009525,0.05000,2.639 | |
| 35,self_attn.v_proj,0.0000003007,0.05000,2.666 | |
| 35,self_attn.o_proj,0.0000005326,0.05000,0.757 | |
| 35,mlp.up_proj,0.0000107934,0.05000,1.366 | |
| 35,mlp.gate_proj,0.0000101982,0.05000,1.372 | |
| 35,mlp.down_proj,0.0000105537,0.05000,4.231 | |