| layer,module,loss,samples,damp,time | |
| 0,self_attn.q_proj,0.0002350870,0.05000,1.389 | |
| 0,self_attn.k_proj,0.0000495884,0.05000,1.401 | |
| 0,self_attn.v_proj,0.0000058943,0.05000,1.411 | |
| 0,self_attn.o_proj,0.0000236358,0.05000,0.468 | |
| 0,mlp.up_proj,0.0002387998,0.05000,0.884 | |
| 0,mlp.gate_proj,0.0003092090,0.05000,0.905 | |
| 0,mlp.down_proj,0.0000352943,0.05000,2.892 | |
| 1,self_attn.q_proj,0.0000474484,0.05000,1.445 | |
| 1,self_attn.k_proj,0.0000132815,0.05000,1.464 | |
| 1,self_attn.v_proj,0.0000025667,0.05000,1.471 | |
| 1,self_attn.o_proj,0.0000079649,0.05000,0.535 | |
| 1,mlp.up_proj,0.0130337269,0.05000,1.058 | |
| 1,mlp.gate_proj,0.0172095274,0.05000,1.066 | |
| 1,mlp.down_proj,0.0000033589,0.05000,3.471 | |
| 2,self_attn.k_proj,0.0000159378,0.05000,1.344 | |
| 2,self_attn.v_proj,0.0000040993,0.05000,1.359 | |
| 2,self_attn.q_proj,0.0000724677,0.05000,1.372 | |
| 2,self_attn.o_proj,0.0000051665,0.05000,0.496 | |
| 2,mlp.gate_proj,0.0038327214,0.05000,0.789 | |
| 2,mlp.up_proj,0.0037921627,0.05000,0.796 | |
| 2,mlp.down_proj,0.0090362954,0.05000,2.958 | |
| 3,self_attn.v_proj,0.0000121165,0.05000,1.371 | |
| 3,self_attn.k_proj,0.0000652644,0.05000,1.377 | |
| 3,self_attn.q_proj,0.0003032616,0.05000,1.398 | |
| 3,self_attn.o_proj,0.0000070980,0.05000,0.840 | |
| 3,mlp.gate_proj,0.0050947977,0.05000,1.129 | |
| 3,mlp.up_proj,0.0043798673,0.05000,1.141 | |
| 3,mlp.down_proj,0.0001146523,0.05000,2.803 | |
| 4,self_attn.v_proj,0.0000165725,0.05000,1.741 | |
| 4,self_attn.q_proj,0.0002473793,0.05000,1.757 | |
| 4,self_attn.k_proj,0.0000484598,0.05000,1.771 | |
| 4,self_attn.o_proj,0.0000154044,0.05000,0.452 | |
| 4,mlp.up_proj,0.0034890500,0.05000,0.945 | |
| 4,mlp.gate_proj,0.0048000967,0.05000,0.968 | |
| 4,mlp.down_proj,0.0000431230,0.05000,2.768 | |
| 5,self_attn.v_proj,0.0000432888,0.05000,1.424 | |
| 5,self_attn.q_proj,0.0005337124,0.05000,1.427 | |
| 5,self_attn.k_proj,0.0000954607,0.05000,1.446 | |
| 5,self_attn.o_proj,0.0000096460,0.05000,0.449 | |
| 5,mlp.gate_proj,0.0064711852,0.05000,0.952 | |
| 5,mlp.up_proj,0.0054382997,0.05000,0.964 | |
| 5,mlp.down_proj,0.0000662426,0.05000,2.780 | |
| 6,self_attn.v_proj,0.0000374054,0.05000,1.393 | |
| 6,self_attn.q_proj,0.0003390273,0.05000,1.413 | |
| 6,self_attn.k_proj,0.0000539993,0.05000,1.431 | |
| 6,self_attn.o_proj,0.0000239811,0.05000,0.453 | |
| 6,mlp.gate_proj,0.0083112013,0.05000,0.931 | |
| 6,mlp.up_proj,0.0070795329,0.05000,0.940 | |
| 6,mlp.down_proj,0.0000900177,0.05000,2.698 | |
| 7,self_attn.k_proj,0.0000704335,0.05000,1.509 | |
| 7,self_attn.q_proj,0.0004461423,0.05000,1.532 | |
| 7,self_attn.v_proj,0.0000449421,0.05000,1.540 | |
| 7,self_attn.o_proj,0.0000158974,0.05000,0.479 | |
| 7,mlp.gate_proj,0.0084789699,0.05000,1.144 | |
| 7,mlp.up_proj,0.0068957208,0.05000,1.156 | |
| 7,mlp.down_proj,0.0001276712,0.05000,3.500 | |
| 8,self_attn.q_proj,0.0004550622,0.05000,1.419 | |
| 8,self_attn.v_proj,0.0000480385,0.05000,1.423 | |
| 8,self_attn.k_proj,0.0000688795,0.05000,1.428 | |
| 8,self_attn.o_proj,0.0000237095,0.05000,0.469 | |
| 8,mlp.up_proj,0.0065720858,0.05000,0.953 | |
| 8,mlp.gate_proj,0.0092143638,0.05000,0.957 | |
| 8,mlp.down_proj,0.0001645900,0.05000,2.690 | |
| 9,self_attn.q_proj,0.0004010807,0.05000,1.298 | |
| 9,self_attn.k_proj,0.0000703355,0.05000,1.310 | |
| 9,self_attn.v_proj,0.0000412660,0.05000,1.319 | |
| 9,self_attn.o_proj,0.0000372614,0.05000,0.451 | |
| 9,mlp.gate_proj,0.0086270275,0.05000,0.724 | |
| 9,mlp.up_proj,0.0057881656,0.05000,0.729 | |
| 9,mlp.down_proj,0.0001873366,0.05000,2.661 | |
| 10,self_attn.v_proj,0.0000675373,0.05000,1.464 | |
| 10,self_attn.k_proj,0.0000842096,0.05000,1.476 | |
| 10,self_attn.q_proj,0.0005640181,0.05000,1.486 | |
| 10,self_attn.o_proj,0.0000375405,0.05000,0.465 | |
| 10,mlp.gate_proj,0.0064309352,0.05000,0.952 | |
| 10,mlp.up_proj,0.0042183016,0.05000,0.952 | |
| 10,mlp.down_proj,0.0002274845,0.05000,2.724 | |
| 11,self_attn.k_proj,0.0000618624,0.05000,1.443 | |
| 11,self_attn.v_proj,0.0000554348,0.05000,1.452 | |
| 11,self_attn.q_proj,0.0004188407,0.05000,1.484 | |
| 11,self_attn.o_proj,0.0000588654,0.05000,0.461 | |
| 11,mlp.up_proj,0.0024528662,0.05000,0.955 | |
| 11,mlp.gate_proj,0.0034099502,0.05000,0.960 | |
| 11,mlp.down_proj,0.0002484049,0.05000,2.826 | |
| 12,self_attn.k_proj,0.0000663459,0.05000,1.707 | |
| 12,self_attn.q_proj,0.0004353355,0.05000,1.714 | |
| 12,self_attn.v_proj,0.0000493771,0.05000,1.720 | |
| 12,self_attn.o_proj,0.0000569917,0.05000,0.534 | |
| 12,mlp.gate_proj,0.0037296684,0.05000,0.927 | |
| 12,mlp.up_proj,0.0026382779,0.05000,0.932 | |
| 12,mlp.down_proj,0.0002238520,0.05000,2.694 | |
| 13,self_attn.k_proj,0.0000918726,0.05000,1.413 | |
| 13,self_attn.v_proj,0.0000332990,0.05000,1.428 | |
| 13,self_attn.q_proj,0.0005053526,0.05000,1.429 | |
| 13,self_attn.o_proj,0.0000357019,0.05000,0.451 | |
| 13,mlp.gate_proj,0.0020309976,0.05000,0.801 | |
| 13,mlp.up_proj,0.0019490962,0.05000,0.810 | |
| 13,mlp.down_proj,0.0001979468,0.05000,2.777 | |
| 14,self_attn.q_proj,0.0003977691,0.05000,1.297 | |
| 14,self_attn.v_proj,0.0000393797,0.05000,1.304 | |
| 14,self_attn.k_proj,0.0000605854,0.05000,1.308 | |
| 14,self_attn.o_proj,0.0000810008,0.05000,0.458 | |
| 14,mlp.up_proj,0.0018685926,0.05000,0.907 | |
| 14,mlp.gate_proj,0.0019428260,0.05000,0.913 | |
| 14,mlp.down_proj,0.0001839744,0.05000,2.739 | |
| 15,self_attn.k_proj,0.0000632234,0.05000,1.297 | |
| 15,self_attn.v_proj,0.0000390386,0.05000,1.308 | |
| 15,self_attn.q_proj,0.0004023011,0.05000,1.323 | |
| 15,self_attn.o_proj,0.0000734222,0.05000,0.459 | |
| 15,mlp.up_proj,0.0016407124,0.05000,0.922 | |
| 15,mlp.gate_proj,0.0015691990,0.05000,0.930 | |
| 15,mlp.down_proj,0.0001619132,0.05000,2.651 | |
| 16,self_attn.v_proj,0.0000397472,0.05000,1.758 | |
| 16,self_attn.q_proj,0.0004093029,0.05000,1.780 | |
| 16,self_attn.k_proj,0.0000679986,0.05000,1.785 | |
| 16,self_attn.o_proj,0.0000735932,0.05000,0.714 | |
| 16,mlp.gate_proj,0.0016409640,0.05000,1.356 | |
| 16,mlp.up_proj,0.0016108529,0.05000,1.364 | |
| 16,mlp.down_proj,0.0001433920,0.05000,2.709 | |
| 17,self_attn.q_proj,0.0006384926,0.05000,1.454 | |
| 17,self_attn.v_proj,0.0000570526,0.05000,1.470 | |
| 17,self_attn.k_proj,0.0001091180,0.05000,1.481 | |
| 17,self_attn.o_proj,0.0000506870,0.05000,0.467 | |
| 17,mlp.up_proj,0.0014302930,0.05000,0.949 | |
| 17,mlp.gate_proj,0.0014094597,0.05000,0.954 | |
| 17,mlp.down_proj,0.0001370377,0.05000,2.728 | |
| 18,self_attn.q_proj,0.0004070264,0.05000,1.559 | |
| 18,self_attn.k_proj,0.0000593264,0.05000,1.642 | |
| 18,self_attn.v_proj,0.0000428977,0.05000,1.651 | |
| 18,self_attn.o_proj,0.0000611151,0.05000,0.643 | |
| 18,mlp.up_proj,0.0013951592,0.05000,1.321 | |
| 18,mlp.gate_proj,0.0014391315,0.05000,1.320 | |
| 18,mlp.down_proj,0.0001301621,0.05000,4.631 | |
| 19,self_attn.v_proj,0.0000454103,0.05000,2.170 | |
| 19,self_attn.k_proj,0.0000802021,0.05000,2.260 | |
| 19,self_attn.q_proj,0.0004584398,0.05000,2.283 | |
| 19,self_attn.o_proj,0.0000832018,0.05000,0.855 | |
| 19,mlp.gate_proj,0.0013592220,0.05000,1.760 | |
| 19,mlp.up_proj,0.0014207353,0.05000,1.786 | |
| 19,mlp.down_proj,0.0001298065,0.05000,4.033 | |
| 20,self_attn.k_proj,0.0000823069,0.05000,1.879 | |
| 20,self_attn.q_proj,0.0007093750,0.05000,1.935 | |
| 20,self_attn.v_proj,0.0000864942,0.05000,1.993 | |
| 20,self_attn.o_proj,0.0000718808,0.05000,0.821 | |
| 20,mlp.gate_proj,0.0013735232,0.05000,1.072 | |
| 20,mlp.up_proj,0.0014128388,0.05000,1.215 | |
| 20,mlp.down_proj,0.0001408066,0.05000,4.534 | |
| 21,self_attn.k_proj,0.0000800281,0.05000,2.156 | |
| 21,self_attn.q_proj,0.0005046818,0.05000,2.172 | |
| 21,self_attn.v_proj,0.0000577273,0.05000,2.274 | |
| 21,self_attn.o_proj,0.0000763357,0.05000,0.748 | |
| 21,mlp.gate_proj,0.0014464086,0.05000,1.168 | |
| 21,mlp.up_proj,0.0013651836,0.05000,1.212 | |
| 21,mlp.down_proj,0.0001177456,0.05000,2.956 | |
| 22,self_attn.k_proj,0.0000903085,0.05000,1.766 | |
| 22,self_attn.q_proj,0.0005909076,0.05000,1.806 | |
| 22,self_attn.v_proj,0.0000757894,0.05000,1.810 | |
| 22,self_attn.o_proj,0.0000623799,0.05000,0.526 | |
| 22,mlp.up_proj,0.0014004772,0.05000,1.257 | |
| 22,mlp.gate_proj,0.0014311959,0.05000,1.275 | |
| 22,mlp.down_proj,0.0001264588,0.05000,4.373 | |
| 23,self_attn.q_proj,0.0006088087,0.05000,2.115 | |
| 23,self_attn.k_proj,0.0000888351,0.05000,2.155 | |
| 23,self_attn.v_proj,0.0000476617,0.05000,2.223 | |
| 23,self_attn.o_proj,0.0000646709,0.05000,0.644 | |
| 23,mlp.up_proj,0.0014787837,0.05000,1.319 | |
| 23,mlp.gate_proj,0.0014157394,0.05000,1.447 | |
| 23,mlp.down_proj,0.0001453594,0.05000,4.296 | |
| 24,self_attn.v_proj,0.0000535015,0.05000,1.915 | |
| 24,self_attn.k_proj,0.0000606524,0.05000,1.972 | |
| 24,self_attn.q_proj,0.0004475276,0.05000,1.990 | |
| 24,self_attn.o_proj,0.0000783631,0.05000,0.887 | |
| 24,mlp.gate_proj,0.0013949360,0.05000,1.192 | |
| 24,mlp.up_proj,0.0014352342,0.05000,1.383 | |
| 24,mlp.down_proj,0.0001567249,0.05000,4.689 | |
| 25,self_attn.k_proj,0.0000588555,0.05000,2.278 | |
| 25,self_attn.v_proj,0.0000853784,0.05000,2.371 | |
| 25,self_attn.q_proj,0.0006110521,0.05000,2.532 | |
| 25,self_attn.o_proj,0.0000598715,0.05000,0.982 | |
| 25,mlp.up_proj,0.0015592789,0.05000,1.336 | |
| 25,mlp.gate_proj,0.0015332821,0.05000,1.365 | |
| 25,mlp.down_proj,0.0001992039,0.05000,3.003 | |
| 26,self_attn.q_proj,0.0004816819,0.05000,1.612 | |
| 26,self_attn.v_proj,0.0000801456,0.05000,1.647 | |
| 26,self_attn.k_proj,0.0000613980,0.05000,1.656 | |
| 26,self_attn.o_proj,0.0000884467,0.05000,0.513 | |
| 26,mlp.gate_proj,0.0016698752,0.05000,1.040 | |
| 26,mlp.up_proj,0.0017734253,0.05000,1.054 | |
| 26,mlp.down_proj,0.0002655405,0.05000,3.050 | |
| 27,self_attn.v_proj,0.0001156350,0.05000,1.534 | |
| 27,self_attn.k_proj,0.0000694705,0.05000,1.571 | |
| 27,self_attn.q_proj,0.0008163267,0.05000,1.579 | |
| 27,self_attn.o_proj,0.0001265619,0.05000,0.511 | |
| 27,mlp.up_proj,0.0018934087,0.05000,0.999 | |
| 27,mlp.gate_proj,0.0018705411,0.05000,1.017 | |
| 27,mlp.down_proj,0.0003228687,0.05000,2.862 | |
| 28,self_attn.k_proj,0.0000650905,0.05000,2.290 | |
| 28,self_attn.q_proj,0.0006038837,0.05000,2.303 | |
| 28,self_attn.v_proj,0.0001092675,0.05000,2.350 | |
| 28,self_attn.o_proj,0.0001192229,0.05000,0.755 | |
| 28,mlp.up_proj,0.0021129013,0.05000,1.005 | |
| 28,mlp.gate_proj,0.0021004773,0.05000,1.026 | |
| 28,mlp.down_proj,0.0003793565,0.05000,2.973 | |
| 29,self_attn.q_proj,0.0005846746,0.05000,1.838 | |
| 29,self_attn.k_proj,0.0000670443,0.05000,1.869 | |
| 29,self_attn.v_proj,0.0000986339,0.05000,1.879 | |
| 29,self_attn.o_proj,0.0000909046,0.05000,0.470 | |
| 29,mlp.up_proj,0.0025685690,0.05000,1.072 | |
| 29,mlp.gate_proj,0.0025046702,0.05000,1.101 | |
| 29,mlp.down_proj,0.0005272932,0.05000,3.049 | |
| 30,self_attn.v_proj,0.0002365673,0.05000,1.805 | |
| 30,self_attn.q_proj,0.0008247248,0.05000,1.814 | |
| 30,self_attn.k_proj,0.0000707403,0.05000,1.832 | |
| 30,self_attn.o_proj,0.0000990036,0.05000,0.519 | |
| 30,mlp.up_proj,0.0035084721,0.05000,0.990 | |
| 30,mlp.gate_proj,0.0032103522,0.05000,1.040 | |
| 30,mlp.down_proj,0.0053239912,0.05000,2.915 | |
| 31,self_attn.q_proj,0.0009805325,0.05000,1.691 | |
| 31,self_attn.v_proj,0.0002592109,0.05000,1.704 | |
| 31,self_attn.k_proj,0.0000985458,0.05000,1.714 | |
| 31,self_attn.o_proj,0.0002562517,0.05000,0.501 | |
| 31,mlp.up_proj,0.0041799812,0.05000,1.191 | |
| 31,mlp.gate_proj,0.0036576450,0.05000,1.214 | |
| 31,mlp.down_proj,0.0017657173,0.05000,4.082 | |
| 32,self_attn.v_proj,0.0007043787,0.05000,2.017 | |
| 32,self_attn.k_proj,0.0001399740,0.05000,2.077 | |
| 32,self_attn.q_proj,0.0015637881,0.05000,2.086 | |
| 32,self_attn.o_proj,0.0002418083,0.05000,0.756 | |
| 32,mlp.up_proj,0.0043462683,0.05000,1.288 | |
| 32,mlp.gate_proj,0.0038090997,0.05000,1.308 | |
| 32,mlp.down_proj,0.0018351650,0.05000,4.499 | |
| 33,self_attn.v_proj,0.0015008834,0.05000,2.409 | |
| 33,self_attn.k_proj,0.0001224033,0.05000,2.454 | |
| 33,self_attn.q_proj,0.0014767564,0.05000,2.478 | |
| 33,self_attn.o_proj,0.0002690905,0.05000,0.725 | |
| 33,mlp.gate_proj,0.0035885438,0.05000,1.361 | |
| 33,mlp.up_proj,0.0044905547,0.05000,1.399 | |
| 33,mlp.down_proj,0.0034433876,0.05000,4.359 | |
| 34,self_attn.v_proj,0.0002083482,0.05000,2.172 | |
| 34,self_attn.q_proj,0.0006153963,0.05000,2.220 | |
| 34,self_attn.k_proj,0.0000626373,0.05000,2.271 | |
| 34,self_attn.o_proj,0.0002969657,0.05000,0.690 | |
| 34,mlp.gate_proj,0.0039339582,0.05000,1.359 | |
| 34,mlp.up_proj,0.0045975191,0.05000,1.368 | |
| 34,mlp.down_proj,0.0025222073,0.05000,3.361 | |
| 35,self_attn.v_proj,0.0001811058,0.05000,1.607 | |
| 35,self_attn.k_proj,0.0000604469,0.05000,1.616 | |
| 35,self_attn.q_proj,0.0005729022,0.05000,1.642 | |
| 35,self_attn.o_proj,0.0002992628,0.05000,0.602 | |
| 35,mlp.up_proj,0.0064264687,0.05000,1.194 | |
| 35,mlp.gate_proj,0.0060764383,0.05000,1.285 | |
| 35,mlp.down_proj,0.0061064307,0.05000,3.911 | |