| layer,module,loss,samples,damp,time | |
| 0,self_attn.k_proj,0.0000472167,0.10000,1.635 | |
| 0,self_attn.v_proj,0.0000078736,0.10000,0.761 | |
| 0,self_attn.q_proj,0.0002322893,0.10000,0.775 | |
| 0,self_attn.o_proj,0.0000110702,0.10000,0.771 | |
| 0,mlp.up_proj,0.0004552265,0.10000,0.827 | |
| 0,mlp.gate_proj,0.0008605841,0.10000,0.821 | |
| 0,mlp.down_proj,0.0000372180,0.10000,5.562 | |
| 1,self_attn.k_proj,0.0000501259,0.10000,0.757 | |
| 1,self_attn.v_proj,0.0000165444,0.10000,0.743 | |
| 1,self_attn.q_proj,0.0001916083,0.10000,0.751 | |
| 1,self_attn.o_proj,0.0000040245,0.10000,0.765 | |
| 1,mlp.up_proj,0.0092325296,0.10000,0.840 | |
| 1,mlp.gate_proj,0.0138325553,0.10000,0.819 | |
| 1,mlp.down_proj,0.0000251277,0.10000,5.547 | |
| 2,self_attn.k_proj,0.0001338489,0.10000,0.768 | |
| 2,self_attn.v_proj,0.0000271386,0.10000,0.753 | |
| 2,self_attn.q_proj,0.0005077048,0.10000,0.756 | |
| 2,self_attn.o_proj,0.0000081085,0.10000,0.756 | |
| 2,mlp.up_proj,0.0088262850,0.10000,0.830 | |
| 2,mlp.gate_proj,0.0135353442,0.10000,0.822 | |
| 2,mlp.down_proj,0.0000469443,0.10000,5.641 | |
| 3,self_attn.k_proj,0.0001345897,0.10000,0.772 | |
| 3,self_attn.v_proj,0.0000399959,0.10000,0.760 | |
| 3,self_attn.q_proj,0.0005372458,0.10000,0.763 | |
| 3,self_attn.o_proj,0.0000187831,0.10000,0.762 | |
| 3,mlp.up_proj,0.0167387179,0.10000,0.836 | |
| 3,mlp.gate_proj,0.0207291428,0.10000,0.825 | |
| 3,mlp.down_proj,0.0006986494,0.10000,5.543 | |
| 4,self_attn.k_proj,0.0002373652,0.10000,0.760 | |
| 4,self_attn.v_proj,0.0001024738,0.10000,0.747 | |
| 4,self_attn.q_proj,0.0010985888,0.10000,0.749 | |
| 4,self_attn.o_proj,0.0000181369,0.10000,0.761 | |
| 4,mlp.up_proj,0.0122986446,0.10000,0.838 | |
| 4,mlp.gate_proj,0.0167416077,0.10000,0.831 | |
| 4,mlp.down_proj,0.0001518217,0.10000,5.565 | |
| 5,self_attn.k_proj,0.0002180186,0.10000,0.763 | |
| 5,self_attn.v_proj,0.0001001975,0.10000,0.768 | |
| 5,self_attn.q_proj,0.0010695679,0.10000,0.756 | |
| 5,self_attn.o_proj,0.0000169917,0.10000,0.769 | |
| 5,mlp.up_proj,0.0181349126,0.10000,0.842 | |
| 5,mlp.gate_proj,0.0210626856,0.10000,0.832 | |
| 5,mlp.down_proj,0.0000545735,0.10000,5.585 | |
| 6,self_attn.k_proj,0.0001318691,0.10000,0.759 | |
| 6,self_attn.v_proj,0.0000696637,0.10000,0.750 | |
| 6,self_attn.q_proj,0.0006738798,0.10000,0.748 | |
| 6,self_attn.o_proj,0.0000257538,0.10000,0.761 | |
| 6,mlp.up_proj,0.0029716475,0.10000,0.837 | |
| 6,mlp.gate_proj,0.0042372570,0.10000,0.827 | |
| 6,mlp.down_proj,0.0001331812,0.10000,5.695 | |
| 7,self_attn.k_proj,0.0001290255,0.10000,0.759 | |
| 7,self_attn.v_proj,0.0001260896,0.10000,0.745 | |
| 7,self_attn.q_proj,0.0007587422,0.10000,0.747 | |
| 7,self_attn.o_proj,0.0000583953,0.10000,0.754 | |
| 7,mlp.up_proj,0.0023285143,0.10000,0.835 | |
| 7,mlp.gate_proj,0.0025514315,0.10000,0.818 | |
| 7,mlp.down_proj,0.0002113085,0.10000,5.588 | |
| 8,self_attn.k_proj,0.0002543362,0.10000,0.760 | |
| 8,self_attn.v_proj,0.0001154977,0.10000,0.749 | |
| 8,self_attn.q_proj,0.0011885152,0.10000,0.752 | |
| 8,self_attn.o_proj,0.0000742859,0.10000,0.761 | |
| 8,mlp.up_proj,0.0026835205,0.10000,0.836 | |
| 8,mlp.gate_proj,0.0027838197,0.10000,0.836 | |
| 8,mlp.down_proj,0.0002283458,0.10000,5.535 | |
| 9,self_attn.k_proj,0.0001838648,0.10000,0.759 | |
| 9,self_attn.v_proj,0.0001650378,0.10000,0.744 | |
| 9,self_attn.q_proj,0.0010429632,0.10000,0.751 | |
| 9,self_attn.o_proj,0.0001184023,0.10000,0.758 | |
| 9,mlp.up_proj,0.0057535404,0.10000,0.832 | |
| 9,mlp.gate_proj,0.0091259328,0.10000,0.825 | |
| 9,mlp.down_proj,0.0003028532,0.10000,5.987 | |
| 10,self_attn.k_proj,0.0001916147,0.10000,0.773 | |
| 10,self_attn.v_proj,0.0001171599,0.10000,0.743 | |
| 10,self_attn.q_proj,0.0010033896,0.10000,0.750 | |
| 10,self_attn.o_proj,0.0000735668,0.10000,0.770 | |
| 10,mlp.up_proj,0.0029951623,0.10000,0.831 | |
| 10,mlp.gate_proj,0.0032918856,0.10000,0.816 | |
| 10,mlp.down_proj,0.0002347663,0.10000,5.543 | |
| 11,self_attn.k_proj,0.0002426945,0.10000,0.754 | |
| 11,self_attn.v_proj,0.0001072995,0.10000,0.844 | |
| 11,self_attn.q_proj,0.0011513309,0.10000,0.899 | |
| 11,self_attn.o_proj,0.0000928968,0.10000,0.912 | |
| 11,mlp.up_proj,0.0028736497,0.10000,1.027 | |
| 11,mlp.gate_proj,0.0029332547,0.10000,0.832 | |
| 11,mlp.down_proj,0.0002235528,0.10000,5.550 | |
| 12,self_attn.k_proj,0.0002664533,0.10000,0.766 | |
| 12,self_attn.v_proj,0.0001384261,0.10000,0.748 | |
| 12,self_attn.q_proj,0.0012836385,0.10000,0.755 | |
| 12,self_attn.o_proj,0.0001158217,0.10000,0.800 | |
| 12,mlp.up_proj,0.0030152578,0.10000,0.856 | |
| 12,mlp.gate_proj,0.0029214713,0.10000,0.852 | |
| 12,mlp.down_proj,0.0002485149,0.10000,5.819 | |
| 13,self_attn.k_proj,0.0002370596,0.10000,0.751 | |
| 13,self_attn.v_proj,0.0001593123,0.10000,0.751 | |
| 13,self_attn.q_proj,0.0013060732,0.10000,0.753 | |
| 13,self_attn.o_proj,0.0001333592,0.10000,0.761 | |
| 13,mlp.up_proj,0.0030158192,0.10000,0.844 | |
| 13,mlp.gate_proj,0.0031320744,0.10000,0.832 | |
| 13,mlp.down_proj,0.0002281524,0.10000,5.580 | |
| 14,self_attn.k_proj,0.0003049479,0.10000,0.756 | |
| 14,self_attn.v_proj,0.0001570591,0.10000,0.736 | |
| 14,self_attn.q_proj,0.0016287128,0.10000,0.747 | |
| 14,self_attn.o_proj,0.0001430852,0.10000,0.768 | |
| 14,mlp.up_proj,0.0032785469,0.10000,0.844 | |
| 14,mlp.gate_proj,0.0031912735,0.10000,0.828 | |
| 14,mlp.down_proj,0.0002534288,0.10000,5.665 | |
| 15,self_attn.k_proj,0.0002922762,0.10000,0.747 | |
| 15,self_attn.v_proj,0.0001391043,0.10000,0.745 | |
| 15,self_attn.q_proj,0.0013987548,0.10000,0.754 | |
| 15,self_attn.o_proj,0.0001067492,0.10000,0.759 | |
| 15,mlp.up_proj,0.0032067114,0.10000,0.831 | |
| 15,mlp.gate_proj,0.0030483367,0.10000,0.821 | |
| 15,mlp.down_proj,0.0002609062,0.10000,5.764 | |
| 16,self_attn.k_proj,0.0002632407,0.10000,0.754 | |
| 16,self_attn.v_proj,0.0001856220,0.10000,0.744 | |
| 16,self_attn.q_proj,0.0014148347,0.10000,0.748 | |
| 16,self_attn.o_proj,0.0001408646,0.10000,0.762 | |
| 16,mlp.up_proj,0.0033665611,0.10000,0.843 | |
| 16,mlp.gate_proj,0.0032078180,0.10000,0.823 | |
| 16,mlp.down_proj,0.0002613587,0.10000,5.541 | |
| 17,self_attn.k_proj,0.0002712536,0.10000,0.752 | |
| 17,self_attn.v_proj,0.0002099971,0.10000,0.748 | |
| 17,self_attn.q_proj,0.0016259497,0.10000,0.749 | |
| 17,self_attn.o_proj,0.0001131741,0.10000,0.758 | |
| 17,mlp.up_proj,0.0039933213,0.10000,0.833 | |
| 17,mlp.gate_proj,0.0037328810,0.10000,0.824 | |
| 17,mlp.down_proj,0.0003456270,0.10000,5.574 | |
| 18,self_attn.k_proj,0.0002159712,0.10000,0.754 | |
| 18,self_attn.v_proj,0.0002325307,0.10000,0.743 | |
| 18,self_attn.q_proj,0.0013481819,0.10000,0.761 | |
| 18,self_attn.o_proj,0.0001833125,0.10000,0.759 | |
| 18,mlp.up_proj,0.0043035444,0.10000,0.836 | |
| 18,mlp.gate_proj,0.0040026095,0.10000,0.828 | |
| 18,mlp.down_proj,0.0004392644,0.10000,5.576 | |
| 19,self_attn.k_proj,0.0002050852,0.10000,0.763 | |
| 19,self_attn.v_proj,0.0002661283,0.10000,0.746 | |
| 19,self_attn.q_proj,0.0014559043,0.10000,0.756 | |
| 19,self_attn.o_proj,0.0002159337,0.10000,0.758 | |
| 19,mlp.up_proj,0.0047093964,0.10000,0.832 | |
| 19,mlp.gate_proj,0.0046096435,0.10000,0.822 | |
| 19,mlp.down_proj,0.0005018536,0.10000,5.554 | |
| 20,self_attn.k_proj,0.0002232794,0.10000,0.761 | |
| 20,self_attn.v_proj,0.0003003416,0.10000,0.750 | |
| 20,self_attn.q_proj,0.0014417838,0.10000,0.748 | |
| 20,self_attn.o_proj,0.0001322579,0.10000,0.758 | |
| 20,mlp.up_proj,0.0060516974,0.10000,0.834 | |
| 20,mlp.gate_proj,0.0059635491,0.10000,0.820 | |
| 20,mlp.down_proj,0.0009318391,0.10000,5.566 | |
| 21,self_attn.k_proj,0.0002299286,0.10000,0.755 | |
| 21,self_attn.v_proj,0.0004667530,0.10000,0.763 | |
| 21,self_attn.q_proj,0.0017117155,0.10000,0.756 | |
| 21,self_attn.o_proj,0.0003423200,0.10000,0.765 | |
| 21,mlp.up_proj,0.0080906422,0.10000,0.843 | |
| 21,mlp.gate_proj,0.0084336273,0.10000,0.825 | |
| 21,mlp.down_proj,0.0014748711,0.10000,5.943 | |
| 22,self_attn.k_proj,0.0003220915,0.10000,1.087 | |
| 22,self_attn.v_proj,0.0008169382,0.10000,0.802 | |
| 22,self_attn.q_proj,0.0025697731,0.10000,0.770 | |
| 22,self_attn.o_proj,0.0002277115,0.10000,0.778 | |
| 22,mlp.up_proj,0.0119962620,0.10000,0.844 | |
| 22,mlp.gate_proj,0.0124196528,0.10000,0.831 | |
| 22,mlp.down_proj,0.0026358763,0.10000,5.724 | |
| 23,self_attn.k_proj,0.0004014596,0.10000,0.764 | |
| 23,self_attn.v_proj,0.0011422982,0.10000,0.754 | |
| 23,self_attn.q_proj,0.0030751132,0.10000,0.769 | |
| 23,self_attn.o_proj,0.0006334905,0.10000,0.921 | |
| 23,mlp.up_proj,0.0183306920,0.10000,0.976 | |
| 23,mlp.gate_proj,0.0189817467,0.10000,0.960 | |
| 23,mlp.down_proj,0.0041346475,0.10000,5.643 | |
| 24,self_attn.k_proj,0.0003703904,0.10000,0.762 | |
| 24,self_attn.v_proj,0.0011029526,0.10000,0.750 | |
| 24,self_attn.q_proj,0.0028345045,0.10000,0.770 | |
| 24,self_attn.o_proj,0.0005504099,0.10000,0.909 | |
| 24,mlp.up_proj,0.0217335806,0.10000,0.969 | |
| 24,mlp.gate_proj,0.0206995038,0.10000,0.960 | |
| 24,mlp.down_proj,0.0059107753,0.10000,5.554 | |
| 25,self_attn.k_proj,0.0004121669,0.10000,0.757 | |
| 25,self_attn.v_proj,0.0018403888,0.10000,0.748 | |
| 25,self_attn.q_proj,0.0034511876,0.10000,0.761 | |
| 25,self_attn.o_proj,0.0008662717,0.10000,0.764 | |
| 25,mlp.up_proj,0.0319784486,0.10000,0.838 | |
| 25,mlp.gate_proj,0.0287693461,0.10000,0.826 | |
| 25,mlp.down_proj,0.0109144369,0.10000,5.618 | |
| 26,self_attn.k_proj,0.0005639863,0.10000,0.755 | |
| 26,self_attn.v_proj,0.0034550298,0.10000,0.754 | |
| 26,self_attn.q_proj,0.0046535053,0.10000,0.780 | |
| 26,self_attn.o_proj,0.0017360485,0.10000,0.763 | |
| 26,mlp.up_proj,0.0365648842,0.10000,0.832 | |
| 26,mlp.gate_proj,0.0322582898,0.10000,0.820 | |
| 26,mlp.down_proj,0.0291283776,0.10000,5.539 | |
| 27,self_attn.k_proj,0.0007934201,0.10000,0.758 | |
| 27,self_attn.v_proj,0.0045992423,0.10000,0.746 | |
| 27,self_attn.q_proj,0.0079554333,0.10000,0.945 | |
| 27,self_attn.o_proj,0.0039928038,0.10000,0.907 | |
| 27,mlp.up_proj,0.0661130137,0.10000,0.876 | |
| 27,mlp.gate_proj,0.0631849785,0.10000,0.850 | |
| 27,mlp.down_proj,0.0570357308,0.10000,5.818 | |