| layer,module,loss,samples,damp,time | |
| 0,self_attn.v_proj,0.0000000085,0.05000,2.499 | |
| 0,self_attn.q_proj,0.0000003401,0.05000,2.506 | |
| 0,self_attn.k_proj,0.0000000715,0.05000,2.508 | |
| 0,self_attn.o_proj,0.0000000277,0.05000,0.743 | |
| 0,mlp.gate_proj,0.0000003923,0.05000,1.322 | |
| 0,mlp.up_proj,0.0000003028,0.05000,1.327 | |
| 0,mlp.down_proj,0.0000000544,0.05000,4.242 | |
| 1,self_attn.v_proj,0.0000000033,0.05000,2.483 | |
| 1,self_attn.k_proj,0.0000000173,0.05000,2.489 | |
| 1,self_attn.q_proj,0.0000000622,0.05000,2.510 | |
| 1,self_attn.o_proj,0.0000000095,0.05000,0.751 | |
| 1,mlp.gate_proj,0.0000222209,0.05000,1.360 | |
| 1,mlp.up_proj,0.0000169856,0.05000,1.362 | |
| 1,mlp.down_proj,0.0000000056,0.05000,4.237 | |
| 2,self_attn.q_proj,0.0000000956,0.05000,2.469 | |
| 2,self_attn.v_proj,0.0000000054,0.05000,2.479 | |
| 2,self_attn.k_proj,0.0000000210,0.05000,2.491 | |
| 2,self_attn.o_proj,0.0000000061,0.05000,0.737 | |
| 2,mlp.gate_proj,0.0000048486,0.05000,1.341 | |
| 2,mlp.up_proj,0.0000049113,0.05000,1.345 | |
| 2,mlp.down_proj,0.0000085108,0.05000,4.254 | |
| 3,self_attn.q_proj,0.0000004073,0.05000,2.481 | |
| 3,self_attn.k_proj,0.0000000878,0.05000,2.510 | |
| 3,self_attn.v_proj,0.0000000161,0.05000,2.516 | |
| 3,self_attn.o_proj,0.0000000094,0.05000,0.749 | |
| 3,mlp.up_proj,0.0000058005,0.05000,1.352 | |
| 3,mlp.gate_proj,0.0000064317,0.05000,1.359 | |
| 3,mlp.down_proj,0.0000001039,0.05000,4.281 | |
| 4,self_attn.k_proj,0.0000187172,0.05000,2.382 | |
| 4,self_attn.q_proj,0.0000959323,0.05000,2.390 | |
| 4,self_attn.v_proj,0.0000064129,0.05000,2.408 | |
| 4,self_attn.o_proj,0.0000053439,0.05000,0.765 | |
| 4,mlp.up_proj,0.0013105773,0.05000,1.337 | |
| 4,mlp.gate_proj,0.0017679583,0.05000,1.337 | |
| 4,mlp.down_proj,0.0000160065,0.05000,4.241 | |
| 5,self_attn.v_proj,0.0000164499,0.05000,2.488 | |
| 5,self_attn.q_proj,0.0002037998,0.05000,2.516 | |
| 5,self_attn.k_proj,0.0000362469,0.05000,2.522 | |
| 5,self_attn.o_proj,0.0000034657,0.05000,0.749 | |
| 5,mlp.gate_proj,0.0024380072,0.05000,1.371 | |
| 5,mlp.up_proj,0.0020617547,0.05000,1.375 | |
| 5,mlp.down_proj,0.0000293610,0.05000,4.193 | |
| 6,self_attn.q_proj,0.0000004554,0.05000,2.422 | |
| 6,self_attn.k_proj,0.0000000725,0.05000,2.462 | |
| 6,self_attn.v_proj,0.0000000506,0.05000,2.476 | |
| 6,self_attn.o_proj,0.0000000303,0.05000,0.751 | |
| 6,mlp.up_proj,0.0000094722,0.05000,1.373 | |
| 6,mlp.gate_proj,0.0000109800,0.05000,1.376 | |
| 6,mlp.down_proj,0.0000001408,0.05000,4.287 | |
| 7,self_attn.q_proj,0.0001760753,0.05000,2.410 | |
| 7,self_attn.k_proj,0.0000278523,0.05000,2.422 | |
| 7,self_attn.v_proj,0.0000177654,0.05000,2.430 | |
| 7,self_attn.o_proj,0.0000057125,0.05000,0.748 | |
| 7,mlp.up_proj,0.0026567668,0.05000,1.384 | |
| 7,mlp.gate_proj,0.0032646559,0.05000,1.394 | |
| 7,mlp.down_proj,0.0000566328,0.05000,4.200 | |
| 8,self_attn.v_proj,0.0000000660,0.05000,2.384 | |
| 8,self_attn.q_proj,0.0000006230,0.05000,2.389 | |
| 8,self_attn.k_proj,0.0000000940,0.05000,2.421 | |
| 8,self_attn.o_proj,0.0000000298,0.05000,0.757 | |
| 8,mlp.gate_proj,0.0000123745,0.05000,1.376 | |
| 8,mlp.up_proj,0.0000088540,0.05000,1.381 | |
| 8,mlp.down_proj,0.0000002613,0.05000,4.221 | |
| 9,self_attn.v_proj,0.0000000562,0.05000,2.470 | |
| 9,self_attn.k_proj,0.0000000951,0.05000,2.477 | |
| 9,self_attn.q_proj,0.0000005431,0.05000,2.501 | |
| 9,self_attn.o_proj,0.0000000489,0.05000,0.751 | |
| 9,mlp.gate_proj,0.0000116203,0.05000,1.342 | |
| 9,mlp.up_proj,0.0000078252,0.05000,1.344 | |
| 9,mlp.down_proj,0.0000003043,0.05000,4.181 | |
| 10,self_attn.v_proj,0.0000000932,0.05000,2.555 | |
| 10,self_attn.k_proj,0.0000001140,0.05000,2.586 | |
| 10,self_attn.q_proj,0.0000007667,0.05000,2.590 | |
| 10,self_attn.o_proj,0.0000000496,0.05000,0.770 | |
| 10,mlp.gate_proj,0.0000087202,0.05000,1.349 | |
| 10,mlp.up_proj,0.0000057541,0.05000,1.354 | |
| 10,mlp.down_proj,0.0000003632,0.05000,4.242 | |
| 11,self_attn.v_proj,0.0000000759,0.05000,2.630 | |
| 11,self_attn.q_proj,0.0000005705,0.05000,2.654 | |
| 11,self_attn.k_proj,0.0000000839,0.05000,2.669 | |
| 11,self_attn.o_proj,0.0000000737,0.05000,0.750 | |
| 11,mlp.gate_proj,0.0000047226,0.05000,1.358 | |
| 11,mlp.up_proj,0.0000033912,0.05000,1.362 | |
| 11,mlp.down_proj,0.0000003967,0.05000,4.218 | |
| 12,self_attn.v_proj,0.0000000690,0.05000,2.519 | |
| 12,self_attn.k_proj,0.0000000907,0.05000,2.537 | |
| 12,self_attn.q_proj,0.0000006007,0.05000,2.552 | |
| 12,self_attn.o_proj,0.0000000782,0.05000,0.745 | |
| 12,mlp.gate_proj,0.0000052067,0.05000,1.358 | |
| 12,mlp.up_proj,0.0000036801,0.05000,1.363 | |
| 12,mlp.down_proj,0.0000003800,0.05000,4.231 | |
| 13,self_attn.k_proj,0.0000361680,0.05000,2.591 | |
| 13,self_attn.q_proj,0.0001995003,0.05000,2.634 | |
| 13,self_attn.v_proj,0.0000131416,0.05000,2.640 | |
| 13,self_attn.o_proj,0.0000135469,0.05000,0.751 | |
| 13,mlp.up_proj,0.0007942385,0.05000,1.379 | |
| 13,mlp.gate_proj,0.0008285486,0.05000,1.388 | |
| 13,mlp.down_proj,0.0000981101,0.05000,4.206 | |
| 14,self_attn.k_proj,0.0000000836,0.05000,2.554 | |
| 14,self_attn.q_proj,0.0000005530,0.05000,2.579 | |
| 14,self_attn.v_proj,0.0000000556,0.05000,2.585 | |
| 14,self_attn.o_proj,0.0000001078,0.05000,0.753 | |
| 14,mlp.up_proj,0.0000026623,0.05000,1.337 | |
| 14,mlp.gate_proj,0.0000027652,0.05000,1.350 | |
| 14,mlp.down_proj,0.0000003071,0.05000,4.240 | |
| 15,self_attn.v_proj,0.0000000539,0.05000,2.473 | |
| 15,self_attn.q_proj,0.0000005573,0.05000,2.478 | |
| 15,self_attn.k_proj,0.0000000866,0.05000,2.480 | |
| 15,self_attn.o_proj,0.0000000966,0.05000,0.766 | |
| 15,mlp.gate_proj,0.0000022427,0.05000,1.350 | |
| 15,mlp.up_proj,0.0000023399,0.05000,1.355 | |
| 15,mlp.down_proj,0.0000002715,0.05000,4.260 | |
| 16,self_attn.q_proj,0.0001676437,0.05000,2.371 | |
| 16,self_attn.v_proj,0.0000162746,0.05000,2.390 | |
| 16,self_attn.k_proj,0.0000276929,0.05000,2.397 | |
| 16,self_attn.o_proj,0.0000294759,0.05000,0.758 | |
| 16,mlp.gate_proj,0.0006744311,0.05000,1.361 | |
| 16,mlp.up_proj,0.0006608038,0.05000,1.371 | |
| 16,mlp.down_proj,0.0000690996,0.05000,4.304 | |
| 17,self_attn.k_proj,0.0000001542,0.05000,2.555 | |
| 17,self_attn.q_proj,0.0000009098,0.05000,2.564 | |
| 17,self_attn.v_proj,0.0000000860,0.05000,2.593 | |
| 17,self_attn.o_proj,0.0000000673,0.05000,0.742 | |
| 17,mlp.up_proj,0.0000020564,0.05000,1.364 | |
| 17,mlp.gate_proj,0.0000020284,0.05000,1.365 | |
| 17,mlp.down_proj,0.0000002287,0.05000,4.246 | |
| 18,self_attn.v_proj,0.0000000613,0.05000,2.463 | |
| 18,self_attn.q_proj,0.0000005767,0.05000,2.462 | |
| 18,self_attn.k_proj,0.0000000836,0.05000,2.475 | |
| 18,self_attn.o_proj,0.0000000865,0.05000,0.766 | |
| 18,mlp.up_proj,0.0000019892,0.05000,1.361 | |
| 18,mlp.gate_proj,0.0000020570,0.05000,1.369 | |
| 18,mlp.down_proj,0.0000002146,0.05000,4.306 | |
| 19,self_attn.q_proj,0.0000006371,0.05000,2.512 | |
| 19,self_attn.v_proj,0.0000000634,0.05000,2.527 | |
| 19,self_attn.k_proj,0.0000001108,0.05000,2.533 | |
| 19,self_attn.o_proj,0.0000001121,0.05000,0.746 | |
| 19,mlp.gate_proj,0.0000019210,0.05000,1.364 | |
| 19,mlp.up_proj,0.0000020055,0.05000,1.369 | |
| 19,mlp.down_proj,0.0000002133,0.05000,4.201 | |
| 20,self_attn.k_proj,0.0000001130,0.05000,2.380 | |
| 20,self_attn.q_proj,0.0000009866,0.05000,2.417 | |
| 20,self_attn.v_proj,0.0000001241,0.05000,2.422 | |
| 20,self_attn.o_proj,0.0000000941,0.05000,0.751 | |
| 20,mlp.gate_proj,0.0000019665,0.05000,1.350 | |
| 20,mlp.up_proj,0.0000019998,0.05000,1.355 | |
| 20,mlp.down_proj,0.0000002300,0.05000,4.280 | |
| 21,self_attn.v_proj,0.0000000834,0.05000,2.548 | |
| 21,self_attn.k_proj,0.0000001145,0.05000,2.595 | |
| 21,self_attn.q_proj,0.0000007214,0.05000,2.606 | |
| 21,self_attn.o_proj,0.0000001060,0.05000,0.770 | |
| 21,mlp.up_proj,0.0000019678,0.05000,1.388 | |
| 21,mlp.gate_proj,0.0000021021,0.05000,1.392 | |
| 21,mlp.down_proj,0.0000002017,0.05000,4.230 | |
| 22,self_attn.v_proj,0.0000320336,0.05000,2.473 | |
| 22,self_attn.q_proj,0.0002469760,0.05000,2.488 | |
| 22,self_attn.k_proj,0.0000380821,0.05000,2.497 | |
| 22,self_attn.o_proj,0.0000279594,0.05000,0.758 | |
| 22,mlp.gate_proj,0.0006105048,0.05000,1.384 | |
| 22,mlp.up_proj,0.0005966521,0.05000,1.390 | |
| 22,mlp.down_proj,0.0000650289,0.05000,4.249 | |
| 23,self_attn.v_proj,0.0000000674,0.05000,2.445 | |
| 23,self_attn.k_proj,0.0000001263,0.05000,2.491 | |
| 23,self_attn.q_proj,0.0000008772,0.05000,2.498 | |
| 23,self_attn.o_proj,0.0000000902,0.05000,0.752 | |
| 23,mlp.up_proj,0.0000022178,0.05000,1.404 | |
| 23,mlp.gate_proj,0.0000021279,0.05000,1.410 | |
| 23,mlp.down_proj,0.0000002565,0.05000,4.249 | |
| 24,self_attn.v_proj,0.0000224701,0.05000,2.336 | |
| 24,self_attn.k_proj,0.0000256526,0.05000,2.340 | |
| 24,self_attn.q_proj,0.0001890992,0.05000,2.348 | |
| 24,self_attn.o_proj,0.0000328178,0.05000,0.760 | |
| 24,mlp.gate_proj,0.0006073164,0.05000,1.381 | |
| 24,mlp.up_proj,0.0006237665,0.05000,1.390 | |
| 24,mlp.down_proj,0.0000798095,0.05000,4.219 | |
| 25,self_attn.v_proj,0.0000001261,0.05000,2.476 | |
| 25,self_attn.k_proj,0.0000000869,0.05000,2.492 | |
| 25,self_attn.q_proj,0.0000009063,0.05000,2.513 | |
| 25,self_attn.o_proj,0.0000000863,0.05000,0.786 | |
| 25,mlp.up_proj,0.0000023707,0.05000,1.362 | |
| 25,mlp.gate_proj,0.0000023394,0.05000,1.371 | |
| 25,mlp.down_proj,0.0000003536,0.05000,4.257 | |
| 26,self_attn.k_proj,0.0000267902,0.05000,2.723 | |
| 26,self_attn.v_proj,0.0000346260,0.05000,2.743 | |
| 26,self_attn.q_proj,0.0002087079,0.05000,2.770 | |
| 26,self_attn.o_proj,0.0000374781,0.05000,0.758 | |
| 26,mlp.gate_proj,0.0007423891,0.05000,1.343 | |
| 26,mlp.up_proj,0.0007881456,0.05000,1.347 | |
| 26,mlp.down_proj,0.0001396396,0.05000,4.212 | |
| 27,self_attn.k_proj,0.0000001067,0.05000,2.623 | |
| 27,self_attn.v_proj,0.0000001766,0.05000,2.673 | |
| 27,self_attn.q_proj,0.0000012480,0.05000,2.698 | |
| 27,self_attn.o_proj,0.0000001959,0.05000,0.756 | |
| 27,mlp.up_proj,0.0000029441,0.05000,1.377 | |
| 27,mlp.gate_proj,0.0000029157,0.05000,1.382 | |
| 27,mlp.down_proj,0.0000006032,0.05000,4.258 | |
| 28,self_attn.v_proj,0.0000473291,0.05000,2.467 | |
| 28,self_attn.k_proj,0.0000284646,0.05000,2.475 | |
| 28,self_attn.q_proj,0.0002644169,0.05000,2.497 | |
| 28,self_attn.o_proj,0.0000534566,0.05000,0.749 | |
| 28,mlp.gate_proj,0.0009634670,0.05000,1.345 | |
| 28,mlp.up_proj,0.0009686252,0.05000,1.350 | |
| 28,mlp.down_proj,0.0002141817,0.05000,4.200 | |
| 29,self_attn.v_proj,0.0000444986,0.05000,2.577 | |
| 29,self_attn.k_proj,0.0000307635,0.05000,2.588 | |
| 29,self_attn.q_proj,0.0002663389,0.05000,2.599 | |
| 29,self_attn.o_proj,0.0000424056,0.05000,0.752 | |
| 29,mlp.gate_proj,0.0011668274,0.05000,1.362 | |
| 29,mlp.up_proj,0.0011953835,0.05000,1.365 | |
| 29,mlp.down_proj,0.0002982608,0.05000,4.231 | |
| 30,self_attn.q_proj,0.0000012274,0.05000,2.437 | |
| 30,self_attn.k_proj,0.0000001052,0.05000,2.452 | |
| 30,self_attn.v_proj,0.0000003586,0.05000,2.453 | |
| 30,self_attn.o_proj,0.0000001560,0.05000,0.756 | |
| 30,mlp.gate_proj,0.0000051271,0.05000,1.339 | |
| 30,mlp.up_proj,0.0000056004,0.05000,1.346 | |
| 30,mlp.down_proj,0.0000080519,0.05000,4.242 | |
| 31,self_attn.v_proj,0.0000003798,0.05000,2.458 | |
| 31,self_attn.k_proj,0.0000001412,0.05000,2.466 | |
| 31,self_attn.q_proj,0.0000014251,0.05000,2.478 | |
| 31,self_attn.o_proj,0.0000003600,0.05000,0.751 | |
| 31,mlp.up_proj,0.0000067604,0.05000,1.359 | |
| 31,mlp.gate_proj,0.0000059077,0.05000,1.364 | |
| 31,mlp.down_proj,0.0000033549,0.05000,4.242 | |
| 32,self_attn.v_proj,0.0000009778,0.05000,2.392 | |
| 32,self_attn.q_proj,0.0000022483,0.05000,2.408 | |
| 32,self_attn.k_proj,0.0000001966,0.05000,2.422 | |
| 32,self_attn.o_proj,0.0000003576,0.05000,0.755 | |
| 32,mlp.gate_proj,0.0000061421,0.05000,1.355 | |
| 32,mlp.up_proj,0.0000070083,0.05000,1.360 | |
| 32,mlp.down_proj,0.0000036502,0.05000,4.249 | |
| 33,self_attn.k_proj,0.0000001775,0.05000,2.490 | |
| 33,self_attn.v_proj,0.0000021774,0.05000,2.549 | |
| 33,self_attn.q_proj,0.0000022492,0.05000,2.551 | |
| 33,self_attn.o_proj,0.0000004517,0.05000,0.749 | |
| 33,mlp.up_proj,0.0000074347,0.05000,1.371 | |
| 33,mlp.gate_proj,0.0000059079,0.05000,1.378 | |
| 33,mlp.down_proj,0.0000068711,0.05000,4.238 | |
| 34,self_attn.v_proj,0.0000003419,0.05000,2.529 | |
| 34,self_attn.k_proj,0.0000001017,0.05000,2.561 | |
| 34,self_attn.q_proj,0.0000009959,0.05000,2.571 | |
| 34,self_attn.o_proj,0.0000005017,0.05000,0.751 | |
| 34,mlp.up_proj,0.0000077549,0.05000,1.376 | |
| 34,mlp.gate_proj,0.0000066303,0.05000,1.379 | |
| 34,mlp.down_proj,0.0000052670,0.05000,4.235 | |
| 35,self_attn.k_proj,0.0000001000,0.05000,2.504 | |
| 35,self_attn.q_proj,0.0000009493,0.05000,2.536 | |
| 35,self_attn.v_proj,0.0000002995,0.05000,2.552 | |
| 35,self_attn.o_proj,0.0000005288,0.05000,0.770 | |
| 35,mlp.gate_proj,0.0000101551,0.05000,1.356 | |
| 35,mlp.up_proj,0.0000107530,0.05000,1.361 | |
| 35,mlp.down_proj,0.0000105331,0.05000,4.218 | |