| layer,module,loss,samples,damp,time | |
| 0,self_attn.k_proj,0.0000926561,0.05000,1.479 | |
| 0,self_attn.q_proj,0.0004441526,0.05000,1.486 | |
| 0,self_attn.v_proj,0.0000118854,0.05000,1.526 | |
| 0,self_attn.o_proj,0.0000509747,0.05000,0.488 | |
| 0,mlp.gate_proj,0.0005754575,0.05000,0.925 | |
| 0,mlp.up_proj,0.0004449969,0.05000,0.925 | |
| 0,mlp.down_proj,0.0000614704,0.05000,2.638 | |
| 1,self_attn.v_proj,0.0000047415,0.05000,1.379 | |
| 1,self_attn.q_proj,0.0000868213,0.05000,1.386 | |
| 1,self_attn.k_proj,0.0000241810,0.05000,1.393 | |
| 1,self_attn.o_proj,0.0000163117,0.05000,0.436 | |
| 1,mlp.gate_proj,0.0385717081,0.05000,0.929 | |
| 1,mlp.up_proj,0.0290021339,0.05000,0.933 | |
| 1,mlp.down_proj,0.0000056345,0.05000,2.672 | |
| 2,self_attn.q_proj,0.0001292669,0.05000,1.324 | |
| 2,self_attn.v_proj,0.0000073232,0.05000,1.331 | |
| 2,self_attn.k_proj,0.0000284607,0.05000,1.363 | |
| 2,self_attn.o_proj,0.0000094392,0.05000,0.446 | |
| 2,mlp.gate_proj,0.0082810673,0.05000,0.926 | |
| 2,mlp.up_proj,0.0080136446,0.05000,0.929 | |
| 2,mlp.down_proj,0.0269350808,0.05000,2.572 | |
| 3,self_attn.k_proj,0.0001170079,0.05000,1.427 | |
| 3,self_attn.q_proj,0.0005366353,0.05000,1.437 | |
| 3,self_attn.v_proj,0.0000215521,0.05000,1.439 | |
| 3,self_attn.o_proj,0.0000148144,0.05000,0.429 | |
| 3,mlp.gate_proj,0.0101124701,0.05000,0.956 | |
| 3,mlp.up_proj,0.0087271334,0.05000,0.963 | |
| 3,mlp.down_proj,0.0001437532,0.05000,2.713 | |
| 4,self_attn.q_proj,0.0004495623,0.05000,1.299 | |
| 4,self_attn.v_proj,0.0000305555,0.05000,1.323 | |
| 4,self_attn.k_proj,0.0000884943,0.05000,1.351 | |
| 4,self_attn.o_proj,0.0000298310,0.05000,0.433 | |
| 4,mlp.gate_proj,0.0096624994,0.05000,0.973 | |
| 4,mlp.up_proj,0.0069845970,0.05000,0.986 | |
| 4,mlp.down_proj,0.0000694115,0.05000,2.615 | |
| 5,self_attn.q_proj,0.0009599595,0.05000,1.503 | |
| 5,self_attn.v_proj,0.0000786844,0.05000,1.520 | |
| 5,self_attn.k_proj,0.0001727102,0.05000,1.529 | |
| 5,self_attn.o_proj,0.0000174932,0.05000,0.429 | |
| 5,mlp.gate_proj,0.0131947269,0.05000,0.968 | |
| 5,mlp.up_proj,0.0111081641,0.05000,0.971 | |
| 5,mlp.down_proj,0.0001038428,0.05000,2.544 | |
| 6,self_attn.q_proj,0.0006202141,0.05000,1.369 | |
| 6,self_attn.k_proj,0.0000994150,0.05000,1.376 | |
| 6,self_attn.v_proj,0.0000700387,0.05000,1.398 | |
| 6,self_attn.o_proj,0.0000451209,0.05000,0.440 | |
| 6,mlp.up_proj,0.0139774917,0.05000,0.940 | |
| 6,mlp.gate_proj,0.0164863941,0.05000,0.951 | |
| 6,mlp.down_proj,0.0001430597,0.05000,2.530 | |
| 7,self_attn.q_proj,0.0007890122,0.05000,1.466 | |
| 7,self_attn.v_proj,0.0000788149,0.05000,1.467 | |
| 7,self_attn.k_proj,0.0001257278,0.05000,1.466 | |
| 7,self_attn.o_proj,0.0000292612,0.05000,0.455 | |
| 7,mlp.gate_proj,0.0158752722,0.05000,0.954 | |
| 7,mlp.up_proj,0.0129439407,0.05000,0.957 | |
| 7,mlp.down_proj,0.0001964390,0.05000,2.723 | |
| 8,self_attn.q_proj,0.0007963218,0.05000,1.280 | |
| 8,self_attn.k_proj,0.0001210402,0.05000,1.289 | |
| 8,self_attn.v_proj,0.0000847841,0.05000,1.323 | |
| 8,self_attn.o_proj,0.0000442622,0.05000,0.432 | |
| 8,mlp.gate_proj,0.0170453784,0.05000,0.913 | |
| 8,mlp.up_proj,0.0121827184,0.05000,0.917 | |
| 8,mlp.down_proj,0.0002522487,0.05000,2.543 | |
| 9,self_attn.k_proj,0.0001248283,0.05000,1.415 | |
| 9,self_attn.v_proj,0.0000729903,0.05000,1.415 | |
| 9,self_attn.q_proj,0.0007056552,0.05000,1.434 | |
| 9,self_attn.o_proj,0.0000737344,0.05000,0.427 | |
| 9,mlp.up_proj,0.0105305310,0.05000,0.946 | |
| 9,mlp.gate_proj,0.0156497282,0.05000,0.945 | |
| 9,mlp.down_proj,0.0002750355,0.05000,2.662 | |
| 10,self_attn.q_proj,0.0010030885,0.05000,1.415 | |
| 10,self_attn.v_proj,0.0001205626,0.05000,1.425 | |
| 10,self_attn.k_proj,0.0001500157,0.05000,1.433 | |
| 10,self_attn.o_proj,0.0000719801,0.05000,0.423 | |
| 10,mlp.up_proj,0.0074459834,0.05000,0.876 | |
| 10,mlp.gate_proj,0.0113039666,0.05000,0.877 | |
| 10,mlp.down_proj,0.0003157377,0.05000,2.603 | |
| 11,self_attn.q_proj,0.0007359214,0.05000,1.301 | |
| 11,self_attn.v_proj,0.0000977395,0.05000,1.325 | |
| 11,self_attn.k_proj,0.0001095498,0.05000,1.329 | |
| 11,self_attn.o_proj,0.0001212573,0.05000,0.425 | |
| 11,mlp.up_proj,0.0042120424,0.05000,0.937 | |
| 11,mlp.gate_proj,0.0058166186,0.05000,0.943 | |
| 11,mlp.down_proj,0.0003497882,0.05000,2.562 | |
| 12,self_attn.k_proj,0.0001183058,0.05000,1.428 | |
| 12,self_attn.v_proj,0.0000878876,0.05000,1.438 | |
| 12,self_attn.q_proj,0.0007691479,0.05000,1.445 | |
| 12,self_attn.o_proj,0.0001109989,0.05000,0.416 | |
| 12,mlp.gate_proj,0.0063900089,0.05000,0.951 | |
| 12,mlp.up_proj,0.0045523345,0.05000,0.961 | |
| 12,mlp.down_proj,0.0003277593,0.05000,2.648 | |
| 13,self_attn.q_proj,0.0009233296,0.05000,1.406 | |
| 13,self_attn.v_proj,0.0000616283,0.05000,1.438 | |
| 13,self_attn.k_proj,0.0001702990,0.05000,1.441 | |
| 13,self_attn.o_proj,0.0000684895,0.05000,0.456 | |
| 13,mlp.gate_proj,0.0034981644,0.05000,0.951 | |
| 13,mlp.up_proj,0.0033732841,0.05000,0.956 | |
| 13,mlp.down_proj,0.0002857845,0.05000,2.568 | |
| 14,self_attn.k_proj,0.0001107448,0.05000,1.253 | |
| 14,self_attn.q_proj,0.0007179622,0.05000,1.262 | |
| 14,self_attn.v_proj,0.0000722210,0.05000,1.286 | |
| 14,self_attn.o_proj,0.0001577273,0.05000,0.435 | |
| 14,mlp.up_proj,0.0032660126,0.05000,0.936 | |
| 14,mlp.gate_proj,0.0033751456,0.05000,0.940 | |
| 14,mlp.down_proj,0.0002750711,0.05000,2.512 | |
| 15,self_attn.q_proj,0.0007260779,0.05000,1.240 | |
| 15,self_attn.k_proj,0.0001140344,0.05000,1.259 | |
| 15,self_attn.v_proj,0.0000717924,0.05000,1.279 | |
| 15,self_attn.o_proj,0.0001466886,0.05000,0.435 | |
| 15,mlp.gate_proj,0.0027777215,0.05000,0.956 | |
| 15,mlp.up_proj,0.0029199004,0.05000,0.969 | |
| 15,mlp.down_proj,0.0002450036,0.05000,2.557 | |
| 16,self_attn.k_proj,0.0001236907,0.05000,1.357 | |
| 16,self_attn.q_proj,0.0007415169,0.05000,1.370 | |
| 16,self_attn.v_proj,0.0000728653,0.05000,1.394 | |
| 16,self_attn.o_proj,0.0001548824,0.05000,0.427 | |
| 16,mlp.up_proj,0.0029338877,0.05000,0.930 | |
| 16,mlp.gate_proj,0.0029728621,0.05000,0.936 | |
| 16,mlp.down_proj,0.0002227558,0.05000,2.520 | |
| 17,self_attn.q_proj,0.0011978788,0.05000,1.411 | |
| 17,self_attn.k_proj,0.0002049842,0.05000,1.442 | |
| 17,self_attn.v_proj,0.0001110038,0.05000,1.449 | |
| 17,self_attn.o_proj,0.0001096319,0.05000,0.456 | |
| 17,mlp.gate_proj,0.0025995242,0.05000,0.938 | |
| 17,mlp.up_proj,0.0026509292,0.05000,0.941 | |
| 17,mlp.down_proj,0.0002133363,0.05000,2.545 | |
| 18,self_attn.q_proj,0.0007654526,0.05000,1.373 | |
| 18,self_attn.v_proj,0.0000810104,0.05000,1.386 | |
| 18,self_attn.k_proj,0.0001118958,0.05000,1.415 | |
| 18,self_attn.o_proj,0.0001209933,0.05000,0.437 | |
| 18,mlp.gate_proj,0.0026745959,0.05000,0.988 | |
| 18,mlp.up_proj,0.0026052926,0.05000,0.993 | |
| 18,mlp.down_proj,0.0001979110,0.05000,2.526 | |
| 19,self_attn.q_proj,0.0008727050,0.05000,1.484 | |
| 19,self_attn.v_proj,0.0000879996,0.05000,1.511 | |
| 19,self_attn.k_proj,0.0001525676,0.05000,1.517 | |
| 19,self_attn.o_proj,0.0001613215,0.05000,0.445 | |
| 19,mlp.gate_proj,0.0025457988,0.05000,0.981 | |
| 19,mlp.up_proj,0.0026727604,0.05000,0.990 | |
| 19,mlp.down_proj,0.0001955881,0.05000,2.512 | |
| 20,self_attn.q_proj,0.0013551767,0.05000,1.482 | |
| 20,self_attn.v_proj,0.0001676866,0.05000,1.505 | |
| 20,self_attn.k_proj,0.0001561945,0.05000,1.517 | |
| 20,self_attn.o_proj,0.0001528362,0.05000,0.437 | |
| 20,mlp.gate_proj,0.0026263935,0.05000,0.962 | |
| 20,mlp.up_proj,0.0027138740,0.05000,0.977 | |
| 20,mlp.down_proj,0.0002239240,0.05000,2.573 | |
| 21,self_attn.v_proj,0.0001106188,0.05000,1.444 | |
| 21,self_attn.k_proj,0.0001505467,0.05000,1.450 | |
| 21,self_attn.q_proj,0.0009562430,0.05000,1.456 | |
| 21,self_attn.o_proj,0.0001582452,0.05000,0.433 | |
| 21,mlp.gate_proj,0.0027915318,0.05000,0.984 | |
| 21,mlp.up_proj,0.0026419421,0.05000,0.997 | |
| 21,mlp.down_proj,0.0002030738,0.05000,2.577 | |
| 22,self_attn.q_proj,0.0011346004,0.05000,1.464 | |
| 22,self_attn.v_proj,0.0001487743,0.05000,1.496 | |
| 22,self_attn.k_proj,0.0001729702,0.05000,1.499 | |
| 22,self_attn.o_proj,0.0001293588,0.05000,0.436 | |
| 22,mlp.gate_proj,0.0027813923,0.05000,0.973 | |
| 22,mlp.up_proj,0.0027358021,0.05000,0.979 | |
| 22,mlp.down_proj,0.0002278643,0.05000,2.698 | |
| 23,self_attn.q_proj,0.0011565852,0.05000,1.424 | |
| 23,self_attn.k_proj,0.0001676210,0.05000,1.443 | |
| 23,self_attn.v_proj,0.0000908016,0.05000,1.451 | |
| 23,self_attn.o_proj,0.0001457047,0.05000,0.434 | |
| 23,mlp.gate_proj,0.0028181392,0.05000,0.982 | |
| 23,mlp.up_proj,0.0029544076,0.05000,0.990 | |
| 23,mlp.down_proj,0.0002631970,0.05000,2.507 | |
| 24,self_attn.k_proj,0.0001165944,0.05000,1.458 | |
| 24,self_attn.v_proj,0.0001046166,0.05000,1.466 | |
| 24,self_attn.q_proj,0.0008574808,0.05000,1.481 | |
| 24,self_attn.o_proj,0.0001873872,0.05000,0.441 | |
| 24,mlp.up_proj,0.0029398206,0.05000,0.966 | |
| 24,mlp.gate_proj,0.0028506532,0.05000,0.975 | |
| 24,mlp.down_proj,0.0002885577,0.05000,2.585 | |
| 25,self_attn.q_proj,0.0012196636,0.05000,1.466 | |
| 25,self_attn.k_proj,0.0001167471,0.05000,1.479 | |
| 25,self_attn.v_proj,0.0001735481,0.05000,1.490 | |
| 25,self_attn.o_proj,0.0001489266,0.05000,0.451 | |
| 25,mlp.gate_proj,0.0032135869,0.05000,0.950 | |
| 25,mlp.up_proj,0.0032790664,0.05000,0.953 | |
| 25,mlp.down_proj,0.0003838621,0.05000,2.757 | |
| 26,self_attn.v_proj,0.0001645723,0.05000,1.358 | |
| 26,self_attn.q_proj,0.0009637461,0.05000,1.371 | |
| 26,self_attn.k_proj,0.0001212483,0.05000,1.376 | |
| 26,self_attn.o_proj,0.0002090202,0.05000,0.427 | |
| 26,mlp.up_proj,0.0038802780,0.05000,0.938 | |
| 26,mlp.gate_proj,0.0036459990,0.05000,0.942 | |
| 26,mlp.down_proj,0.0005390834,0.05000,2.532 | |
| 27,self_attn.k_proj,0.0001402346,0.05000,1.335 | |
| 27,self_attn.v_proj,0.0002450933,0.05000,1.359 | |
| 27,self_attn.q_proj,0.0016638996,0.05000,1.369 | |
| 27,self_attn.o_proj,0.0003133276,0.05000,0.422 | |
| 27,mlp.up_proj,0.0040160986,0.05000,0.946 | |
| 27,mlp.gate_proj,0.0039604067,0.05000,0.953 | |
| 27,mlp.down_proj,0.0006700770,0.05000,2.505 | |
| 28,self_attn.v_proj,0.0002193391,0.05000,1.531 | |
| 28,self_attn.q_proj,0.0011876772,0.05000,1.540 | |
| 28,self_attn.k_proj,0.0001271059,0.05000,1.549 | |
| 28,self_attn.o_proj,0.0002864604,0.05000,0.440 | |
| 28,mlp.up_proj,0.0045377837,0.05000,0.981 | |
| 28,mlp.gate_proj,0.0045106539,0.05000,0.991 | |
| 28,mlp.down_proj,0.0007840230,0.05000,2.636 | |
| 29,self_attn.q_proj,0.0011711047,0.05000,1.465 | |
| 29,self_attn.k_proj,0.0001328981,0.05000,1.484 | |
| 29,self_attn.v_proj,0.0002036485,0.05000,1.487 | |
| 29,self_attn.o_proj,0.0002445588,0.05000,0.433 | |
| 29,mlp.gate_proj,0.0053584442,0.05000,0.863 | |
| 29,mlp.up_proj,0.0054983478,0.05000,0.867 | |
| 29,mlp.down_proj,0.0010751190,0.05000,2.710 | |
| 30,self_attn.q_proj,0.0016243049,0.05000,1.236 | |
| 30,self_attn.k_proj,0.0001383619,0.05000,1.274 | |
| 30,self_attn.v_proj,0.0004685805,0.05000,1.281 | |
| 30,self_attn.o_proj,0.0002138880,0.05000,0.422 | |
| 30,mlp.up_proj,0.0074140683,0.05000,0.931 | |
| 30,mlp.gate_proj,0.0067897284,0.05000,0.931 | |
| 30,mlp.down_proj,0.0253920822,0.05000,2.541 | |
| 31,self_attn.q_proj,0.0018998680,0.05000,1.392 | |
| 31,self_attn.k_proj,0.0001888821,0.05000,1.397 | |
| 31,self_attn.v_proj,0.0005093985,0.05000,1.399 | |
| 31,self_attn.o_proj,0.0004122503,0.05000,0.442 | |
| 31,mlp.gate_proj,0.0078921620,0.05000,1.006 | |
| 31,mlp.up_proj,0.0090096026,0.05000,1.026 | |
| 31,mlp.down_proj,0.0036010571,0.05000,2.663 | |
| 32,self_attn.v_proj,0.0013319228,0.05000,1.439 | |
| 32,self_attn.k_proj,0.0002597295,0.05000,1.439 | |
| 32,self_attn.q_proj,0.0029558921,0.05000,1.455 | |
| 32,self_attn.o_proj,0.0003896669,0.05000,0.489 | |
| 32,mlp.gate_proj,0.0077764657,0.05000,0.945 | |
| 32,mlp.up_proj,0.0088612360,0.05000,0.960 | |
| 32,mlp.down_proj,0.0033728618,0.05000,2.690 | |
| 33,self_attn.q_proj,0.0028402272,0.05000,1.341 | |
| 33,self_attn.k_proj,0.0002332372,0.05000,1.346 | |
| 33,self_attn.v_proj,0.0027896581,0.05000,1.362 | |
| 33,self_attn.o_proj,0.0004359603,0.05000,0.474 | |
| 33,mlp.gate_proj,0.0072181967,0.05000,0.921 | |
| 33,mlp.up_proj,0.0090274869,0.05000,0.927 | |
| 33,mlp.down_proj,0.0057725686,0.05000,2.693 | |
| 34,self_attn.q_proj,0.0011860414,0.05000,1.332 | |
| 34,self_attn.v_proj,0.0003992968,0.05000,1.340 | |
| 34,self_attn.k_proj,0.0001209296,0.05000,1.356 | |
| 34,self_attn.o_proj,0.0005546022,0.05000,0.431 | |
| 34,mlp.gate_proj,0.0077909217,0.05000,1.025 | |
| 34,mlp.up_proj,0.0090878079,0.05000,1.049 | |
| 34,mlp.down_proj,0.0041094714,0.05000,2.748 | |
| 35,self_attn.q_proj,0.0010943316,0.05000,1.456 | |
| 35,self_attn.v_proj,0.0003437671,0.05000,1.478 | |
| 35,self_attn.k_proj,0.0001138098,0.05000,1.483 | |
| 35,self_attn.o_proj,0.0005312666,0.05000,0.440 | |
| 35,mlp.up_proj,0.0124718128,0.05000,0.965 | |
| 35,mlp.gate_proj,0.0117675475,0.05000,0.969 | |
| 35,mlp.down_proj,0.0163082102,0.05000,2.571 | |