| layer,module,loss,samples,damp,time | |
| 0,self_attn.k_proj,0.0000000099,0.05000,4.817 | |
| 0,self_attn.q_proj,0.0000000326,0.05000,4.873 | |
| 0,self_attn.v_proj,0.0000000008,0.05000,4.883 | |
| 0,self_attn.o_proj,0.0000000000,0.05000,1.506 | |
| 0,mlp.gate_proj,0.0000000322,0.05000,2.852 | |
| 0,mlp.up_proj,0.0000000279,0.05000,2.861 | |
| 0,mlp.down_proj,0.0000000000,0.05000,5.854 | |
| 1,self_attn.v_proj,0.0000000075,0.05000,5.948 | |
| 1,self_attn.q_proj,0.0000000986,0.05000,5.967 | |
| 1,self_attn.k_proj,0.0000000453,0.05000,5.986 | |
| 1,self_attn.o_proj,0.0000000000,0.05000,1.513 | |
| 1,mlp.gate_proj,0.0000001107,0.05000,2.723 | |
| 1,mlp.up_proj,0.0000000970,0.05000,2.728 | |
| 1,mlp.down_proj,0.0000000181,0.05000,5.748 | |
| 2,self_attn.v_proj,0.0000000473,0.05000,5.806 | |
| 2,self_attn.k_proj,0.0000003621,0.05000,5.825 | |
| 2,self_attn.q_proj,0.0000007457,0.05000,5.841 | |
| 2,self_attn.o_proj,0.0000000000,0.05000,1.545 | |
| 2,mlp.gate_proj,0.0000002318,0.05000,2.761 | |
| 2,mlp.up_proj,0.0000002023,0.05000,2.779 | |
| 2,mlp.down_proj,0.0000000001,0.05000,5.801 | |
| 3,self_attn.q_proj,0.0000004040,0.05000,5.748 | |
| 3,self_attn.k_proj,0.0000001964,0.05000,5.757 | |
| 3,self_attn.v_proj,0.0000000324,0.05000,5.836 | |
| 3,self_attn.o_proj,0.0000000000,0.05000,1.507 | |
| 3,mlp.up_proj,0.0000003335,0.05000,2.726 | |
| 3,mlp.gate_proj,0.0000003861,0.05000,2.737 | |
| 3,mlp.down_proj,0.0000000001,0.05000,5.850 | |
| 4,self_attn.k_proj,0.0000002752,0.05000,5.885 | |
| 4,self_attn.q_proj,0.0000006219,0.05000,5.899 | |
| 4,self_attn.v_proj,0.0000000542,0.05000,5.920 | |
| 4,self_attn.o_proj,0.0000000000,0.05000,1.511 | |
| 4,mlp.gate_proj,0.0000005583,0.05000,2.707 | |
| 4,mlp.up_proj,0.0000004521,0.05000,2.712 | |
| 4,mlp.down_proj,0.0000000003,0.05000,5.795 | |
| 5,self_attn.v_proj,0.0000000601,0.05000,5.651 | |
| 5,self_attn.k_proj,0.0000003567,0.05000,5.685 | |
| 5,self_attn.q_proj,0.0000008041,0.05000,5.708 | |
| 5,self_attn.o_proj,0.0000000001,0.05000,1.530 | |
| 5,mlp.gate_proj,0.0000007742,0.05000,2.754 | |
| 5,mlp.up_proj,0.0000005881,0.05000,2.759 | |
| 5,mlp.down_proj,0.0000000004,0.05000,5.792 | |
| 6,self_attn.q_proj,0.0000007546,0.05000,5.715 | |
| 6,self_attn.v_proj,0.0000000610,0.05000,5.753 | |
| 6,self_attn.k_proj,0.0000003538,0.05000,5.763 | |
| 6,self_attn.o_proj,0.0000000001,0.05000,1.491 | |
| 6,mlp.up_proj,0.0000006991,0.05000,2.700 | |
| 6,mlp.gate_proj,0.0000008995,0.05000,2.717 | |
| 6,mlp.down_proj,0.0000000006,0.05000,5.790 | |
| 7,self_attn.k_proj,0.0000004439,0.05000,5.755 | |
| 7,self_attn.q_proj,0.0000009290,0.05000,5.785 | |
| 7,self_attn.v_proj,0.0000000806,0.05000,5.807 | |
| 7,self_attn.o_proj,0.0000000002,0.05000,1.512 | |
| 7,mlp.up_proj,0.0000008049,0.05000,2.738 | |
| 7,mlp.gate_proj,0.0000010599,0.05000,2.751 | |
| 7,mlp.down_proj,0.0000000008,0.05000,5.833 | |
| 8,self_attn.v_proj,0.0000000749,0.05000,5.644 | |
| 8,self_attn.k_proj,0.0000003480,0.05000,5.679 | |
| 8,self_attn.q_proj,0.0000007657,0.05000,5.711 | |
| 8,self_attn.o_proj,0.0000000003,0.05000,1.486 | |
| 8,mlp.gate_proj,0.0000011395,0.05000,2.710 | |
| 8,mlp.up_proj,0.0000008825,0.05000,2.716 | |
| 8,mlp.down_proj,0.0000000009,0.05000,5.785 | |
| 9,self_attn.k_proj,0.0000004894,0.05000,5.779 | |
| 9,self_attn.q_proj,0.0000010457,0.05000,5.846 | |
| 9,self_attn.v_proj,0.0000000885,0.05000,5.857 | |
| 9,self_attn.o_proj,0.0000000003,0.05000,1.521 | |
| 9,mlp.gate_proj,0.0000011959,0.05000,2.735 | |
| 9,mlp.up_proj,0.0000009581,0.05000,2.754 | |
| 9,mlp.down_proj,0.0000000011,0.05000,5.800 | |
| 10,self_attn.v_proj,0.0000000776,0.05000,5.689 | |
| 10,self_attn.k_proj,0.0000004430,0.05000,5.697 | |
| 10,self_attn.q_proj,0.0000009305,0.05000,5.710 | |
| 10,self_attn.o_proj,0.0000000006,0.05000,1.482 | |
| 10,mlp.gate_proj,0.0000012489,0.05000,2.718 | |
| 10,mlp.up_proj,0.0000010243,0.05000,2.723 | |
| 10,mlp.down_proj,0.0000000013,0.05000,5.783 | |
| 11,self_attn.q_proj,0.0000011214,0.05000,5.811 | |
| 11,self_attn.k_proj,0.0000005155,0.05000,5.824 | |
| 11,self_attn.v_proj,0.0000001141,0.05000,5.853 | |
| 11,self_attn.o_proj,0.0000000008,0.05000,1.511 | |
| 11,mlp.gate_proj,0.0000013628,0.05000,2.734 | |
| 11,mlp.up_proj,0.0000011338,0.05000,2.743 | |
| 11,mlp.down_proj,0.0000000015,0.05000,5.785 | |
| 12,self_attn.q_proj,0.0000014866,0.05000,5.701 | |
| 12,self_attn.v_proj,0.0000001297,0.05000,5.702 | |
| 12,self_attn.k_proj,0.0000006694,0.05000,5.701 | |
| 12,self_attn.o_proj,0.0000000009,0.05000,1.521 | |
| 12,mlp.up_proj,0.0000012766,0.05000,2.776 | |
| 12,mlp.gate_proj,0.0000015000,0.05000,2.783 | |
| 12,mlp.down_proj,0.0000000019,0.05000,5.699 | |
| 13,self_attn.k_proj,0.0000005692,0.05000,5.850 | |
| 13,self_attn.v_proj,0.0000001234,0.05000,5.908 | |
| 13,self_attn.q_proj,0.0000011665,0.05000,5.942 | |
| 13,self_attn.o_proj,0.0000000011,0.05000,1.479 | |
| 13,mlp.up_proj,0.0000014674,0.05000,2.721 | |
| 13,mlp.gate_proj,0.0000016768,0.05000,2.732 | |
| 13,mlp.down_proj,0.0000000024,0.05000,5.763 | |
| 14,self_attn.q_proj,0.0000013253,0.05000,5.565 | |
| 14,self_attn.k_proj,0.0000005664,0.05000,5.643 | |
| 14,self_attn.v_proj,0.0000001981,0.05000,5.653 | |
| 14,self_attn.o_proj,0.0000000014,0.05000,1.508 | |
| 14,mlp.gate_proj,0.0000018725,0.05000,2.713 | |
| 14,mlp.up_proj,0.0000016302,0.05000,2.741 | |
| 14,mlp.down_proj,0.0000000030,0.05000,5.708 | |
| 15,self_attn.k_proj,0.0000007359,0.05000,5.857 | |
| 15,self_attn.q_proj,0.0000016581,0.05000,5.863 | |
| 15,self_attn.v_proj,0.0000002239,0.05000,5.899 | |
| 15,self_attn.o_proj,0.0000000015,0.05000,1.513 | |
| 15,mlp.up_proj,0.0000017916,0.05000,2.713 | |
| 15,mlp.gate_proj,0.0000021294,0.05000,2.719 | |
| 15,mlp.down_proj,0.0000000038,0.05000,5.826 | |
| 16,self_attn.v_proj,0.0000002110,0.05000,5.704 | |
| 16,self_attn.k_proj,0.0000006766,0.05000,5.723 | |
| 16,self_attn.q_proj,0.0000014709,0.05000,5.736 | |
| 16,self_attn.o_proj,0.0000000017,0.05000,1.504 | |
| 16,mlp.up_proj,0.0000020874,0.05000,2.717 | |
| 16,mlp.gate_proj,0.0000026221,0.05000,2.732 | |
| 16,mlp.down_proj,0.0000000054,0.05000,5.778 | |
| 17,self_attn.v_proj,0.0000002017,0.05000,5.752 | |
| 17,self_attn.q_proj,0.0000013624,0.05000,5.840 | |
| 17,self_attn.k_proj,0.0000005692,0.05000,5.847 | |
| 17,self_attn.o_proj,0.0000000022,0.05000,1.490 | |
| 17,mlp.gate_proj,0.0000030491,0.05000,2.742 | |
| 17,mlp.up_proj,0.0000024313,0.05000,2.746 | |
| 17,mlp.down_proj,0.0000000074,0.05000,5.748 | |
| 18,self_attn.k_proj,0.0000006751,0.05000,5.822 | |
| 18,self_attn.v_proj,0.0000002480,0.05000,5.845 | |
| 18,self_attn.q_proj,0.0000017297,0.05000,5.888 | |
| 18,self_attn.o_proj,0.0000000020,0.05000,1.496 | |
| 18,mlp.gate_proj,0.0000034365,0.05000,2.732 | |
| 18,mlp.up_proj,0.0000027417,0.05000,2.739 | |
| 18,mlp.down_proj,0.0000000107,0.05000,5.821 | |
| 19,self_attn.v_proj,0.0000002874,0.05000,5.735 | |
| 19,self_attn.q_proj,0.0000015817,0.05000,5.766 | |
| 19,self_attn.k_proj,0.0000006702,0.05000,5.773 | |
| 19,self_attn.o_proj,0.0000000025,0.05000,1.524 | |
| 19,mlp.up_proj,0.0000030297,0.05000,2.756 | |
| 19,mlp.gate_proj,0.0000038671,0.05000,2.773 | |
| 19,mlp.down_proj,0.0000000143,0.05000,5.762 | |
| 20,self_attn.q_proj,0.0000016975,0.05000,5.790 | |
| 20,self_attn.k_proj,0.0000006890,0.05000,5.835 | |
| 20,self_attn.v_proj,0.0000003146,0.05000,5.849 | |
| 20,self_attn.o_proj,0.0000000023,0.05000,1.505 | |
| 20,mlp.up_proj,0.0000033502,0.05000,2.737 | |
| 20,mlp.gate_proj,0.0000043862,0.05000,2.745 | |
| 20,mlp.down_proj,0.0000000149,0.05000,5.762 | |
| 21,self_attn.q_proj,0.0000016742,0.05000,5.741 | |
| 21,self_attn.v_proj,0.0000003297,0.05000,5.744 | |
| 21,self_attn.k_proj,0.0000006758,0.05000,5.762 | |
| 21,self_attn.o_proj,0.0000000024,0.05000,1.527 | |
| 21,mlp.up_proj,0.0000036733,0.05000,2.724 | |
| 21,mlp.gate_proj,0.0000051593,0.05000,2.730 | |
| 21,mlp.down_proj,0.0000000152,0.05000,5.811 | |
| 22,self_attn.k_proj,0.0000006403,0.05000,5.796 | |
| 22,self_attn.v_proj,0.0000003419,0.05000,5.830 | |
| 22,self_attn.q_proj,0.0000016118,0.05000,5.901 | |
| 22,self_attn.o_proj,0.0000000014,0.05000,1.528 | |
| 22,mlp.up_proj,0.0000039479,0.05000,2.729 | |
| 22,mlp.gate_proj,0.0000054861,0.05000,2.733 | |
| 22,mlp.down_proj,0.0000000157,0.05000,5.797 | |
| 23,self_attn.v_proj,0.0000003579,0.05000,5.894 | |
| 23,self_attn.k_proj,0.0000006380,0.05000,5.895 | |
| 23,self_attn.q_proj,0.0000016239,0.05000,5.922 | |
| 23,self_attn.o_proj,0.0000000023,0.05000,1.531 | |
| 23,mlp.gate_proj,0.0000060618,0.05000,2.698 | |
| 23,mlp.up_proj,0.0000043452,0.05000,2.699 | |
| 23,mlp.down_proj,0.0000000177,0.05000,5.782 | |
| 24,self_attn.k_proj,0.0000007371,0.05000,5.596 | |
| 24,self_attn.q_proj,0.0000018533,0.05000,5.663 | |
| 24,self_attn.v_proj,0.0000004165,0.05000,5.676 | |
| 24,self_attn.o_proj,0.0000000021,0.05000,1.492 | |
| 24,mlp.gate_proj,0.0000068435,0.05000,2.771 | |
| 24,mlp.up_proj,0.0000048163,0.05000,2.777 | |
| 24,mlp.down_proj,0.0000000192,0.05000,5.771 | |
| 25,self_attn.k_proj,0.0000007104,0.05000,5.831 | |
| 25,self_attn.v_proj,0.0000004764,0.05000,5.842 | |
| 25,self_attn.q_proj,0.0000018583,0.05000,5.841 | |
| 25,self_attn.o_proj,0.0000000024,0.05000,1.509 | |
| 25,mlp.up_proj,0.0000052669,0.05000,2.721 | |
| 25,mlp.gate_proj,0.0000074224,0.05000,2.730 | |
| 25,mlp.down_proj,0.0000000218,0.05000,5.820 | |
| 26,self_attn.v_proj,0.0000005038,0.05000,5.621 | |
| 26,self_attn.k_proj,0.0000006691,0.05000,5.630 | |
| 26,self_attn.q_proj,0.0000017644,0.05000,5.648 | |
| 26,self_attn.o_proj,0.0000000033,0.05000,1.515 | |
| 26,mlp.up_proj,0.0000057272,0.05000,2.755 | |
| 26,mlp.gate_proj,0.0000077866,0.05000,2.775 | |
| 26,mlp.down_proj,0.0000000250,0.05000,5.808 | |
| 27,self_attn.q_proj,0.0000018230,0.05000,5.916 | |
| 27,self_attn.v_proj,0.0000004564,0.05000,5.989 | |
| 27,self_attn.k_proj,0.0000006762,0.05000,6.016 | |
| 27,self_attn.o_proj,0.0000000037,0.05000,1.485 | |
| 27,mlp.up_proj,0.0000063079,0.05000,2.725 | |
| 27,mlp.gate_proj,0.0000085269,0.05000,2.739 | |
| 27,mlp.down_proj,0.0000000300,0.05000,5.801 | |
| 28,self_attn.v_proj,0.0000006713,0.05000,5.844 | |
| 28,self_attn.k_proj,0.0000006783,0.05000,5.868 | |
| 28,self_attn.q_proj,0.0000017780,0.05000,5.884 | |
| 28,self_attn.o_proj,0.0000000056,0.05000,1.507 | |
| 28,mlp.up_proj,0.0000070183,0.05000,2.716 | |
| 28,mlp.gate_proj,0.0000091368,0.05000,2.723 | |
| 28,mlp.down_proj,0.0000000409,0.05000,5.721 | |
| 29,self_attn.v_proj,0.0000009821,0.05000,5.830 | |
| 29,self_attn.k_proj,0.0000006843,0.05000,5.854 | |
| 29,self_attn.q_proj,0.0000020354,0.05000,5.860 | |
| 29,self_attn.o_proj,0.0000000119,0.05000,1.516 | |
| 29,mlp.gate_proj,0.0000088944,0.05000,2.812 | |
| 29,mlp.up_proj,0.0000071220,0.05000,2.825 | |
| 29,mlp.down_proj,0.0000000519,0.05000,5.835 | |
| 30,self_attn.q_proj,0.0000018947,0.05000,5.718 | |
| 30,self_attn.k_proj,0.0000006325,0.05000,5.845 | |
| 30,self_attn.v_proj,0.0000010680,0.05000,5.874 | |
| 30,self_attn.o_proj,0.0000000106,0.05000,1.507 | |
| 30,mlp.gate_proj,0.0000089889,0.05000,2.782 | |
| 30,mlp.up_proj,0.0000073173,0.05000,2.796 | |
| 30,mlp.down_proj,0.0000000681,0.05000,5.731 | |
| 31,self_attn.v_proj,0.0000010866,0.05000,5.793 | |
| 31,self_attn.k_proj,0.0000006014,0.05000,5.846 | |
| 31,self_attn.q_proj,0.0000017606,0.05000,5.861 | |
| 31,self_attn.o_proj,0.0000000142,0.05000,1.512 | |
| 31,mlp.gate_proj,0.0000075029,0.05000,2.765 | |
| 31,mlp.up_proj,0.0000059300,0.05000,2.774 | |
| 31,mlp.down_proj,0.0000001083,0.05000,5.806 | |