450 lines
20 KiB
CSV
450 lines
20 KiB
CSV
|
|
layer,module,loss,samples,damp,time
|
||
|
|
0,self_attn.k_proj,0.0000022077,0.05000,0.823
|
||
|
|
0,self_attn.v_proj,0.0000006867,0.05000,0.644
|
||
|
|
0,self_attn.q_proj,0.0000063762,0.05000,0.653
|
||
|
|
0,self_attn.o_proj,0.0000078134,0.05000,0.673
|
||
|
|
0,mlp.up_proj,0.0000063333,0.05000,0.756
|
||
|
|
0,mlp.gate_proj,0.0000070154,0.05000,0.732
|
||
|
|
0,mlp.down_proj,0.0000045693,0.05000,5.997
|
||
|
|
1,self_attn.k_proj,0.0000000713,0.05000,0.667
|
||
|
|
1,self_attn.v_proj,0.0000000420,0.05000,0.648
|
||
|
|
1,self_attn.q_proj,0.0000002727,0.05000,0.656
|
||
|
|
1,self_attn.o_proj,0.0000000940,0.05000,0.677
|
||
|
|
1,mlp.up_proj,0.0000328479,0.05000,0.750
|
||
|
|
1,mlp.gate_proj,0.0000974030,0.05000,0.736
|
||
|
|
1,mlp.down_proj,0.0000005106,0.05000,5.997
|
||
|
|
2,self_attn.k_proj,0.0000002637,0.05000,0.669
|
||
|
|
2,self_attn.v_proj,0.0000001370,0.05000,0.654
|
||
|
|
2,self_attn.q_proj,0.0000007469,0.05000,0.660
|
||
|
|
2,self_attn.o_proj,0.0000005078,0.05000,0.670
|
||
|
|
2,mlp.up_proj,0.0000693201,0.05000,0.752
|
||
|
|
2,mlp.gate_proj,0.0001472164,0.05000,0.740
|
||
|
|
2,mlp.down_proj,0.0000019943,0.05000,5.993
|
||
|
|
3,self_attn.k_proj,0.0000016043,0.05000,0.669
|
||
|
|
3,self_attn.v_proj,0.0000006461,0.05000,0.652
|
||
|
|
3,self_attn.q_proj,0.0000045232,0.05000,0.660
|
||
|
|
3,self_attn.o_proj,0.0000016176,0.05000,0.674
|
||
|
|
3,mlp.up_proj,0.0000765738,0.05000,0.750
|
||
|
|
3,mlp.gate_proj,0.0001638541,0.05000,0.736
|
||
|
|
3,mlp.down_proj,0.0000064677,0.05000,6.015
|
||
|
|
4,self_attn.k_proj,0.0000015465,0.05000,0.663
|
||
|
|
4,self_attn.v_proj,0.0000008786,0.05000,0.649
|
||
|
|
4,self_attn.q_proj,0.0000046396,0.05000,0.660
|
||
|
|
4,self_attn.o_proj,0.0000029959,0.05000,0.670
|
||
|
|
4,mlp.up_proj,0.0002003118,0.05000,0.749
|
||
|
|
4,mlp.gate_proj,0.0003645644,0.05000,0.738
|
||
|
|
4,mlp.down_proj,0.0000090716,0.05000,6.006
|
||
|
|
5,self_attn.k_proj,0.0000051957,0.05000,0.669
|
||
|
|
5,self_attn.v_proj,0.0000033450,0.05000,0.648
|
||
|
|
5,self_attn.q_proj,0.0000171433,0.05000,0.658
|
||
|
|
5,self_attn.o_proj,0.0000075782,0.05000,0.671
|
||
|
|
5,mlp.up_proj,0.0002897159,0.05000,0.749
|
||
|
|
5,mlp.gate_proj,0.0004948036,0.05000,0.737
|
||
|
|
5,mlp.down_proj,0.0000162241,0.05000,6.011
|
||
|
|
6,self_attn.k_proj,0.0000047515,0.05000,0.670
|
||
|
|
6,self_attn.v_proj,0.0000038208,0.05000,0.651
|
||
|
|
6,self_attn.q_proj,0.0000166306,0.05000,0.661
|
||
|
|
6,self_attn.o_proj,0.0000077817,0.05000,0.675
|
||
|
|
6,mlp.up_proj,0.0004921902,0.05000,0.758
|
||
|
|
6,mlp.gate_proj,0.0007870232,0.05000,0.738
|
||
|
|
6,mlp.down_proj,0.0039085098,0.05000,5.996
|
||
|
|
7,self_attn.k_proj,0.0000068135,0.05000,0.667
|
||
|
|
7,self_attn.v_proj,0.0000060292,0.05000,0.651
|
||
|
|
7,self_attn.q_proj,0.0000249747,0.05000,0.665
|
||
|
|
7,self_attn.o_proj,0.0000053162,0.05000,0.674
|
||
|
|
7,mlp.up_proj,0.0005382502,0.05000,0.764
|
||
|
|
7,mlp.gate_proj,0.0008569802,0.05000,0.739
|
||
|
|
7,mlp.down_proj,0.0000286973,0.05000,6.030
|
||
|
|
8,self_attn.k_proj,0.0000081067,0.05000,0.665
|
||
|
|
8,self_attn.v_proj,0.0000059298,0.05000,0.648
|
||
|
|
8,self_attn.q_proj,0.0000282493,0.05000,0.658
|
||
|
|
8,self_attn.o_proj,0.0000042831,0.05000,0.671
|
||
|
|
8,mlp.up_proj,0.0003307477,0.05000,0.754
|
||
|
|
8,mlp.gate_proj,0.0005081947,0.05000,0.739
|
||
|
|
8,mlp.down_proj,0.0000233559,0.05000,5.984
|
||
|
|
9,self_attn.k_proj,0.0000072105,0.05000,0.666
|
||
|
|
9,self_attn.v_proj,0.0000066063,0.05000,0.649
|
||
|
|
9,self_attn.q_proj,0.0000259657,0.05000,0.653
|
||
|
|
9,self_attn.o_proj,0.0000089853,0.05000,0.666
|
||
|
|
9,mlp.up_proj,0.0001224814,0.05000,0.762
|
||
|
|
9,mlp.gate_proj,0.0001323581,0.05000,0.738
|
||
|
|
9,mlp.down_proj,0.0000223179,0.05000,6.019
|
||
|
|
10,self_attn.k_proj,0.0000103246,0.05000,0.671
|
||
|
|
10,self_attn.v_proj,0.0000084822,0.05000,0.646
|
||
|
|
10,self_attn.q_proj,0.0000384352,0.05000,0.654
|
||
|
|
10,self_attn.o_proj,0.0000123051,0.05000,0.670
|
||
|
|
10,mlp.up_proj,0.0001398602,0.05000,0.754
|
||
|
|
10,mlp.gate_proj,0.0001503285,0.05000,0.740
|
||
|
|
10,mlp.down_proj,0.0000270805,0.05000,6.020
|
||
|
|
11,self_attn.k_proj,0.0000079004,0.05000,0.663
|
||
|
|
11,self_attn.v_proj,0.0000062583,0.05000,0.643
|
||
|
|
11,self_attn.q_proj,0.0000287895,0.05000,0.656
|
||
|
|
11,self_attn.o_proj,0.0000124526,0.05000,0.669
|
||
|
|
11,mlp.up_proj,0.0002015323,0.05000,0.751
|
||
|
|
11,mlp.gate_proj,0.0002419696,0.05000,0.738
|
||
|
|
11,mlp.down_proj,0.0000217685,0.05000,6.020
|
||
|
|
12,self_attn.k_proj,0.0000093121,0.05000,0.665
|
||
|
|
12,self_attn.v_proj,0.0000074404,0.05000,0.650
|
||
|
|
12,self_attn.q_proj,0.0000345770,0.05000,0.657
|
||
|
|
12,self_attn.o_proj,0.0000143948,0.05000,0.668
|
||
|
|
12,mlp.up_proj,0.0001704685,0.05000,0.755
|
||
|
|
12,mlp.gate_proj,0.0001819776,0.05000,0.743
|
||
|
|
12,mlp.down_proj,0.0000293608,0.05000,6.014
|
||
|
|
13,self_attn.k_proj,0.0000101114,0.05000,0.664
|
||
|
|
13,self_attn.v_proj,0.0000087253,0.05000,0.644
|
||
|
|
13,self_attn.q_proj,0.0000368231,0.05000,0.654
|
||
|
|
13,self_attn.o_proj,0.0000131411,0.05000,0.670
|
||
|
|
13,mlp.up_proj,0.0001998923,0.05000,0.751
|
||
|
|
13,mlp.gate_proj,0.0002154623,0.05000,0.742
|
||
|
|
13,mlp.down_proj,0.0000334675,0.05000,6.025
|
||
|
|
14,self_attn.k_proj,0.0000128095,0.05000,0.667
|
||
|
|
14,self_attn.v_proj,0.0000097440,0.05000,0.654
|
||
|
|
14,self_attn.q_proj,0.0000471291,0.05000,0.656
|
||
|
|
14,self_attn.o_proj,0.0000115778,0.05000,0.669
|
||
|
|
14,mlp.up_proj,0.0002111189,0.05000,0.752
|
||
|
|
14,mlp.gate_proj,0.0002301692,0.05000,0.739
|
||
|
|
14,mlp.down_proj,0.0000353972,0.05000,6.027
|
||
|
|
15,self_attn.k_proj,0.0000113254,0.05000,0.667
|
||
|
|
15,self_attn.v_proj,0.0000100019,0.05000,0.650
|
||
|
|
15,self_attn.q_proj,0.0000412780,0.05000,0.661
|
||
|
|
15,self_attn.o_proj,0.0000187119,0.05000,0.673
|
||
|
|
15,mlp.up_proj,0.0002209445,0.05000,0.755
|
||
|
|
15,mlp.gate_proj,0.0002505017,0.05000,0.748
|
||
|
|
15,mlp.down_proj,0.0000388736,0.05000,6.027
|
||
|
|
16,self_attn.k_proj,0.0000102344,0.05000,0.666
|
||
|
|
16,self_attn.v_proj,0.0000076844,0.05000,0.647
|
||
|
|
16,self_attn.q_proj,0.0000355875,0.05000,0.658
|
||
|
|
16,self_attn.o_proj,0.0000114868,0.05000,0.668
|
||
|
|
16,mlp.up_proj,0.0002131696,0.05000,0.753
|
||
|
|
16,mlp.gate_proj,0.0002305127,0.05000,0.745
|
||
|
|
16,mlp.down_proj,0.0000348936,0.05000,6.046
|
||
|
|
17,self_attn.k_proj,0.0000132098,0.05000,0.670
|
||
|
|
17,self_attn.v_proj,0.0000090512,0.05000,0.657
|
||
|
|
17,self_attn.q_proj,0.0000458009,0.05000,0.662
|
||
|
|
17,self_attn.o_proj,0.0000097193,0.05000,0.667
|
||
|
|
17,mlp.up_proj,0.0002087758,0.05000,0.753
|
||
|
|
17,mlp.gate_proj,0.0002223910,0.05000,0.739
|
||
|
|
17,mlp.down_proj,0.0000340258,0.05000,6.039
|
||
|
|
18,self_attn.k_proj,0.0000129059,0.05000,0.670
|
||
|
|
18,self_attn.v_proj,0.0000094483,0.05000,0.651
|
||
|
|
18,self_attn.q_proj,0.0000446828,0.05000,0.668
|
||
|
|
18,self_attn.o_proj,0.0000114085,0.05000,0.673
|
||
|
|
18,mlp.up_proj,0.0002044169,0.05000,0.758
|
||
|
|
18,mlp.gate_proj,0.0002160919,0.05000,0.742
|
||
|
|
18,mlp.down_proj,0.0000334648,0.05000,6.014
|
||
|
|
19,self_attn.k_proj,0.0000118550,0.05000,0.670
|
||
|
|
19,self_attn.v_proj,0.0000089792,0.05000,0.648
|
||
|
|
19,self_attn.q_proj,0.0000425266,0.05000,0.664
|
||
|
|
19,self_attn.o_proj,0.0000097391,0.05000,0.674
|
||
|
|
19,mlp.up_proj,0.0002062136,0.05000,0.759
|
||
|
|
19,mlp.gate_proj,0.0002177467,0.05000,0.746
|
||
|
|
19,mlp.down_proj,0.0000371059,0.05000,6.040
|
||
|
|
20,self_attn.k_proj,0.0000133839,0.05000,0.665
|
||
|
|
20,self_attn.v_proj,0.0000093785,0.05000,0.650
|
||
|
|
20,self_attn.q_proj,0.0000457884,0.05000,0.660
|
||
|
|
20,self_attn.o_proj,0.0000117708,0.05000,0.679
|
||
|
|
20,mlp.up_proj,0.0001998364,0.05000,0.756
|
||
|
|
20,mlp.gate_proj,0.0002076299,0.05000,0.741
|
||
|
|
20,mlp.down_proj,0.0000383832,0.05000,6.040
|
||
|
|
21,self_attn.k_proj,0.0000121129,0.05000,0.666
|
||
|
|
21,self_attn.v_proj,0.0000079683,0.05000,0.645
|
||
|
|
21,self_attn.q_proj,0.0000416676,0.05000,0.653
|
||
|
|
21,self_attn.o_proj,0.0000144970,0.05000,0.669
|
||
|
|
21,mlp.up_proj,0.0001920803,0.05000,0.753
|
||
|
|
21,mlp.gate_proj,0.0002002938,0.05000,0.740
|
||
|
|
21,mlp.down_proj,0.0000330328,0.05000,6.046
|
||
|
|
22,self_attn.k_proj,0.0000132585,0.05000,0.668
|
||
|
|
22,self_attn.v_proj,0.0000114245,0.05000,0.652
|
||
|
|
22,self_attn.q_proj,0.0000477719,0.05000,0.656
|
||
|
|
22,self_attn.o_proj,0.0000178106,0.05000,0.673
|
||
|
|
22,mlp.up_proj,0.0002043648,0.05000,0.753
|
||
|
|
22,mlp.gate_proj,0.0002126195,0.05000,0.742
|
||
|
|
22,mlp.down_proj,0.0000375758,0.05000,6.065
|
||
|
|
23,self_attn.k_proj,0.0000126629,0.05000,0.669
|
||
|
|
23,self_attn.v_proj,0.0000113606,0.05000,0.654
|
||
|
|
23,self_attn.q_proj,0.0000466326,0.05000,0.667
|
||
|
|
23,self_attn.o_proj,0.0000183573,0.05000,0.672
|
||
|
|
23,mlp.up_proj,0.0002143453,0.05000,0.756
|
||
|
|
23,mlp.gate_proj,0.0002262560,0.05000,0.741
|
||
|
|
23,mlp.down_proj,0.0000392328,0.05000,6.066
|
||
|
|
24,self_attn.k_proj,0.0000154229,0.05000,0.663
|
||
|
|
24,self_attn.v_proj,0.0000119528,0.05000,0.650
|
||
|
|
24,self_attn.q_proj,0.0000559860,0.05000,0.661
|
||
|
|
24,self_attn.o_proj,0.0000161610,0.05000,0.670
|
||
|
|
24,mlp.up_proj,0.0002105366,0.05000,0.759
|
||
|
|
24,mlp.gate_proj,0.0002192322,0.05000,0.739
|
||
|
|
24,mlp.down_proj,0.0000415121,0.05000,6.073
|
||
|
|
25,self_attn.k_proj,0.0000171919,0.05000,0.671
|
||
|
|
25,self_attn.v_proj,0.0000144000,0.05000,0.652
|
||
|
|
25,self_attn.q_proj,0.0000627893,0.05000,0.658
|
||
|
|
25,self_attn.o_proj,0.0000090962,0.05000,0.677
|
||
|
|
25,mlp.up_proj,0.0002131117,0.05000,0.765
|
||
|
|
25,mlp.gate_proj,0.0002183458,0.05000,0.748
|
||
|
|
25,mlp.down_proj,0.0000412529,0.05000,6.035
|
||
|
|
26,self_attn.k_proj,0.0000153086,0.05000,0.674
|
||
|
|
26,self_attn.v_proj,0.0000095222,0.05000,0.655
|
||
|
|
26,self_attn.q_proj,0.0000556917,0.05000,0.665
|
||
|
|
26,self_attn.o_proj,0.0000178903,0.05000,0.679
|
||
|
|
26,mlp.up_proj,0.0002247247,0.05000,0.761
|
||
|
|
26,mlp.gate_proj,0.0002281609,0.05000,0.749
|
||
|
|
26,mlp.down_proj,0.0000503585,0.05000,6.042
|
||
|
|
27,self_attn.k_proj,0.0000152973,0.05000,0.674
|
||
|
|
27,self_attn.v_proj,0.0000102050,0.05000,0.651
|
||
|
|
27,self_attn.q_proj,0.0000513762,0.05000,0.665
|
||
|
|
27,self_attn.o_proj,0.0000228966,0.05000,0.673
|
||
|
|
27,mlp.up_proj,0.0002341506,0.05000,0.765
|
||
|
|
27,mlp.gate_proj,0.0002378677,0.05000,0.751
|
||
|
|
27,mlp.down_proj,0.0000502993,0.05000,6.039
|
||
|
|
28,self_attn.k_proj,0.0000143219,0.05000,0.673
|
||
|
|
28,self_attn.v_proj,0.0000147895,0.05000,0.651
|
||
|
|
28,self_attn.q_proj,0.0000542909,0.05000,0.666
|
||
|
|
28,self_attn.o_proj,0.0000263406,0.05000,0.681
|
||
|
|
28,mlp.up_proj,0.0002539306,0.05000,0.761
|
||
|
|
28,mlp.gate_proj,0.0002540842,0.05000,0.747
|
||
|
|
28,mlp.down_proj,0.0000555785,0.05000,6.035
|
||
|
|
29,self_attn.k_proj,0.0000189423,0.05000,0.680
|
||
|
|
29,self_attn.v_proj,0.0000181694,0.05000,0.654
|
||
|
|
29,self_attn.q_proj,0.0000703678,0.05000,0.665
|
||
|
|
29,self_attn.o_proj,0.0000225502,0.05000,0.677
|
||
|
|
29,mlp.up_proj,0.0002634465,0.05000,0.766
|
||
|
|
29,mlp.gate_proj,0.0002623749,0.05000,0.750
|
||
|
|
29,mlp.down_proj,0.0000548736,0.05000,6.054
|
||
|
|
30,self_attn.k_proj,0.0000163911,0.05000,0.674
|
||
|
|
30,self_attn.v_proj,0.0000183102,0.05000,0.659
|
||
|
|
30,self_attn.q_proj,0.0000643131,0.05000,0.668
|
||
|
|
30,self_attn.o_proj,0.0000396940,0.05000,0.677
|
||
|
|
30,mlp.up_proj,0.0002883857,0.05000,0.760
|
||
|
|
30,mlp.gate_proj,0.0002864540,0.05000,0.756
|
||
|
|
30,mlp.down_proj,0.0000628212,0.05000,6.067
|
||
|
|
31,self_attn.k_proj,0.0000178963,0.05000,0.675
|
||
|
|
31,self_attn.v_proj,0.0000182467,0.05000,0.661
|
||
|
|
31,self_attn.q_proj,0.0000671255,0.05000,0.660
|
||
|
|
31,self_attn.o_proj,0.0000456650,0.05000,0.677
|
||
|
|
31,mlp.up_proj,0.0003307039,0.05000,0.773
|
||
|
|
31,mlp.gate_proj,0.0003285395,0.05000,0.750
|
||
|
|
31,mlp.down_proj,0.0000792020,0.05000,6.055
|
||
|
|
32,self_attn.k_proj,0.0000166342,0.05000,0.672
|
||
|
|
32,self_attn.v_proj,0.0000146921,0.05000,0.658
|
||
|
|
32,self_attn.q_proj,0.0000611444,0.05000,0.675
|
||
|
|
32,self_attn.o_proj,0.0000394544,0.05000,0.687
|
||
|
|
32,mlp.up_proj,0.0003476900,0.05000,0.765
|
||
|
|
32,mlp.gate_proj,0.0003627944,0.05000,0.749
|
||
|
|
32,mlp.down_proj,0.0000767992,0.05000,6.070
|
||
|
|
33,self_attn.k_proj,0.0000194685,0.05000,0.680
|
||
|
|
33,self_attn.v_proj,0.0000154845,0.05000,0.654
|
||
|
|
33,self_attn.q_proj,0.0000719577,0.05000,0.672
|
||
|
|
33,self_attn.o_proj,0.0000336899,0.05000,0.687
|
||
|
|
33,mlp.up_proj,0.0003242552,0.05000,0.764
|
||
|
|
33,mlp.gate_proj,0.0003319873,0.05000,0.747
|
||
|
|
33,mlp.down_proj,0.0000720865,0.05000,6.091
|
||
|
|
34,self_attn.k_proj,0.0000189671,0.05000,0.679
|
||
|
|
34,self_attn.v_proj,0.0000177655,0.05000,0.666
|
||
|
|
34,self_attn.q_proj,0.0000712278,0.05000,0.665
|
||
|
|
34,self_attn.o_proj,0.0000411208,0.05000,0.684
|
||
|
|
34,mlp.up_proj,0.0003175667,0.05000,0.758
|
||
|
|
34,mlp.gate_proj,0.0003184013,0.05000,0.740
|
||
|
|
34,mlp.down_proj,0.0000682010,0.05000,6.073
|
||
|
|
35,self_attn.k_proj,0.0000197936,0.05000,0.678
|
||
|
|
35,self_attn.v_proj,0.0000176204,0.05000,0.662
|
||
|
|
35,self_attn.q_proj,0.0000792466,0.05000,0.669
|
||
|
|
35,self_attn.o_proj,0.0000301874,0.05000,0.689
|
||
|
|
35,mlp.up_proj,0.0003235649,0.05000,0.758
|
||
|
|
35,mlp.gate_proj,0.0003235762,0.05000,0.746
|
||
|
|
35,mlp.down_proj,0.0000702756,0.05000,6.072
|
||
|
|
36,self_attn.k_proj,0.0000204304,0.05000,0.676
|
||
|
|
36,self_attn.v_proj,0.0000174813,0.05000,0.660
|
||
|
|
36,self_attn.q_proj,0.0000773615,0.05000,0.672
|
||
|
|
36,self_attn.o_proj,0.0000297189,0.05000,0.677
|
||
|
|
36,mlp.up_proj,0.0002965451,0.05000,0.757
|
||
|
|
36,mlp.gate_proj,0.0002855328,0.05000,0.744
|
||
|
|
36,mlp.down_proj,0.0000625354,0.05000,6.121
|
||
|
|
37,self_attn.k_proj,0.0000187126,0.05000,0.672
|
||
|
|
37,self_attn.v_proj,0.0000154525,0.05000,0.652
|
||
|
|
37,self_attn.q_proj,0.0000695368,0.05000,0.662
|
||
|
|
37,self_attn.o_proj,0.0000321507,0.05000,0.671
|
||
|
|
37,mlp.up_proj,0.0002908931,0.05000,0.759
|
||
|
|
37,mlp.gate_proj,0.0002796531,0.05000,0.745
|
||
|
|
37,mlp.down_proj,0.0000561437,0.05000,6.092
|
||
|
|
38,self_attn.k_proj,0.0000203654,0.05000,0.666
|
||
|
|
38,self_attn.v_proj,0.0000232785,0.05000,0.647
|
||
|
|
38,self_attn.q_proj,0.0000777056,0.05000,0.661
|
||
|
|
38,self_attn.o_proj,0.0000360887,0.05000,0.669
|
||
|
|
38,mlp.up_proj,0.0002907469,0.05000,0.759
|
||
|
|
38,mlp.gate_proj,0.0002802605,0.05000,0.743
|
||
|
|
38,mlp.down_proj,0.0000611777,0.05000,6.094
|
||
|
|
39,self_attn.k_proj,0.0000177757,0.05000,0.668
|
||
|
|
39,self_attn.v_proj,0.0000218557,0.05000,0.650
|
||
|
|
39,self_attn.q_proj,0.0000729561,0.05000,0.666
|
||
|
|
39,self_attn.o_proj,0.0000351012,0.05000,0.673
|
||
|
|
39,mlp.up_proj,0.0002971604,0.05000,0.768
|
||
|
|
39,mlp.gate_proj,0.0002967976,0.05000,0.756
|
||
|
|
39,mlp.down_proj,0.0000579369,0.05000,6.042
|
||
|
|
40,self_attn.k_proj,0.0000215583,0.05000,0.669
|
||
|
|
40,self_attn.v_proj,0.0000192123,0.05000,0.654
|
||
|
|
40,self_attn.q_proj,0.0000795933,0.05000,0.672
|
||
|
|
40,self_attn.o_proj,0.0000300833,0.05000,0.685
|
||
|
|
40,mlp.up_proj,0.0002780093,0.05000,0.760
|
||
|
|
40,mlp.gate_proj,0.0002736469,0.05000,0.763
|
||
|
|
40,mlp.down_proj,0.0000523938,0.05000,6.046
|
||
|
|
41,self_attn.k_proj,0.0000219502,0.05000,0.684
|
||
|
|
41,self_attn.v_proj,0.0000261218,0.05000,0.658
|
||
|
|
41,self_attn.q_proj,0.0000877220,0.05000,0.671
|
||
|
|
41,self_attn.o_proj,0.0000206927,0.05000,0.685
|
||
|
|
41,mlp.up_proj,0.0002734285,0.05000,0.778
|
||
|
|
41,mlp.gate_proj,0.0002620355,0.05000,0.753
|
||
|
|
41,mlp.down_proj,0.0000507733,0.05000,6.084
|
||
|
|
42,self_attn.k_proj,0.0000202025,0.05000,0.686
|
||
|
|
42,self_attn.v_proj,0.0000156105,0.05000,0.664
|
||
|
|
42,self_attn.q_proj,0.0000796620,0.05000,0.680
|
||
|
|
42,self_attn.o_proj,0.0000293642,0.05000,0.689
|
||
|
|
42,mlp.up_proj,0.0003010015,0.05000,0.760
|
||
|
|
42,mlp.gate_proj,0.0002825469,0.05000,0.764
|
||
|
|
42,mlp.down_proj,0.0000717976,0.05000,6.109
|
||
|
|
43,self_attn.k_proj,0.0000209122,0.05000,0.679
|
||
|
|
43,self_attn.v_proj,0.0000188210,0.05000,0.665
|
||
|
|
43,self_attn.q_proj,0.0000762192,0.05000,0.676
|
||
|
|
43,self_attn.o_proj,0.0000369718,0.05000,0.685
|
||
|
|
43,mlp.up_proj,0.0003245416,0.05000,0.770
|
||
|
|
43,mlp.gate_proj,0.0003047796,0.05000,0.751
|
||
|
|
43,mlp.down_proj,0.0000943858,0.05000,6.128
|
||
|
|
44,self_attn.k_proj,0.0000164873,0.05000,0.682
|
||
|
|
44,self_attn.v_proj,0.0000256224,0.05000,0.661
|
||
|
|
44,self_attn.q_proj,0.0000715989,0.05000,0.672
|
||
|
|
44,self_attn.o_proj,0.0000435539,0.05000,0.680
|
||
|
|
44,mlp.up_proj,0.0003300955,0.05000,0.762
|
||
|
|
44,mlp.gate_proj,0.0003045399,0.05000,0.749
|
||
|
|
44,mlp.down_proj,0.0000742722,0.05000,6.116
|
||
|
|
45,self_attn.k_proj,0.0000202093,0.05000,0.683
|
||
|
|
45,self_attn.v_proj,0.0000279675,0.05000,0.666
|
||
|
|
45,self_attn.q_proj,0.0000850428,0.05000,0.667
|
||
|
|
45,self_attn.o_proj,0.0000346460,0.05000,0.679
|
||
|
|
45,mlp.up_proj,0.0003242656,0.05000,0.760
|
||
|
|
45,mlp.gate_proj,0.0002998104,0.05000,0.745
|
||
|
|
45,mlp.down_proj,0.0000709695,0.05000,6.125
|
||
|
|
46,self_attn.k_proj,0.0000168016,0.05000,0.681
|
||
|
|
46,self_attn.v_proj,0.0000287407,0.05000,0.656
|
||
|
|
46,self_attn.q_proj,0.0000765311,0.05000,0.663
|
||
|
|
46,self_attn.o_proj,0.0000442687,0.05000,0.671
|
||
|
|
46,mlp.up_proj,0.0003310285,0.05000,0.760
|
||
|
|
46,mlp.gate_proj,0.0003074441,0.05000,0.744
|
||
|
|
46,mlp.down_proj,0.0000749102,0.05000,6.104
|
||
|
|
47,self_attn.k_proj,0.0000169921,0.05000,0.671
|
||
|
|
47,self_attn.v_proj,0.0000251264,0.05000,0.647
|
||
|
|
47,self_attn.q_proj,0.0000766527,0.05000,0.662
|
||
|
|
47,self_attn.o_proj,0.0000405218,0.05000,0.676
|
||
|
|
47,mlp.up_proj,0.0003626132,0.05000,0.761
|
||
|
|
47,mlp.gate_proj,0.0003408490,0.05000,0.748
|
||
|
|
47,mlp.down_proj,0.0000886861,0.05000,6.121
|
||
|
|
48,self_attn.k_proj,0.0000188492,0.05000,0.668
|
||
|
|
48,self_attn.v_proj,0.0000322942,0.05000,0.653
|
||
|
|
48,self_attn.q_proj,0.0000862149,0.05000,0.667
|
||
|
|
48,self_attn.o_proj,0.0000377576,0.05000,0.675
|
||
|
|
48,mlp.up_proj,0.0003911038,0.05000,0.766
|
||
|
|
48,mlp.gate_proj,0.0003737860,0.05000,0.757
|
||
|
|
48,mlp.down_proj,0.0000992985,0.05000,6.089
|
||
|
|
49,self_attn.k_proj,0.0000184366,0.05000,0.671
|
||
|
|
49,self_attn.v_proj,0.0000308705,0.05000,0.655
|
||
|
|
49,self_attn.q_proj,0.0000872545,0.05000,0.669
|
||
|
|
49,self_attn.o_proj,0.0000557681,0.05000,0.695
|
||
|
|
49,mlp.up_proj,0.0004658231,0.05000,0.775
|
||
|
|
49,mlp.gate_proj,0.0004469727,0.05000,0.756
|
||
|
|
49,mlp.down_proj,0.0001539159,0.05000,6.061
|
||
|
|
50,self_attn.k_proj,0.0000197978,0.05000,0.682
|
||
|
|
50,self_attn.v_proj,0.0000376188,0.05000,0.669
|
||
|
|
50,self_attn.q_proj,0.0000987396,0.05000,0.668
|
||
|
|
50,self_attn.o_proj,0.0000552519,0.05000,0.691
|
||
|
|
50,mlp.up_proj,0.0005598184,0.05000,0.767
|
||
|
|
50,mlp.gate_proj,0.0005505634,0.05000,0.759
|
||
|
|
50,mlp.down_proj,0.0002117901,0.05000,6.059
|
||
|
|
51,self_attn.k_proj,0.0000222420,0.05000,0.682
|
||
|
|
51,self_attn.v_proj,0.0000352546,0.05000,0.662
|
||
|
|
51,self_attn.q_proj,0.0001003685,0.05000,0.685
|
||
|
|
51,self_attn.o_proj,0.0000740499,0.05000,0.691
|
||
|
|
51,mlp.up_proj,0.0006666129,0.05000,0.779
|
||
|
|
51,mlp.gate_proj,0.0006646402,0.05000,0.765
|
||
|
|
51,mlp.down_proj,0.0002415502,0.05000,6.111
|
||
|
|
52,self_attn.k_proj,0.0000238645,0.05000,0.685
|
||
|
|
52,self_attn.v_proj,0.0000568164,0.05000,0.665
|
||
|
|
52,self_attn.q_proj,0.0001193794,0.05000,0.675
|
||
|
|
52,self_attn.o_proj,0.0000718841,0.05000,0.686
|
||
|
|
52,mlp.up_proj,0.0007413024,0.05000,0.775
|
||
|
|
52,mlp.gate_proj,0.0007344707,0.05000,0.763
|
||
|
|
52,mlp.down_proj,0.0002956851,0.05000,6.141
|
||
|
|
53,self_attn.k_proj,0.0000288697,0.05000,0.685
|
||
|
|
53,self_attn.v_proj,0.0000568506,0.05000,0.655
|
||
|
|
53,self_attn.q_proj,0.0001322156,0.05000,0.681
|
||
|
|
53,self_attn.o_proj,0.0000779313,0.05000,0.689
|
||
|
|
53,mlp.up_proj,0.0008568734,0.05000,0.764
|
||
|
|
53,mlp.gate_proj,0.0008507893,0.05000,0.746
|
||
|
|
53,mlp.down_proj,0.0003484684,0.05000,6.115
|
||
|
|
54,self_attn.k_proj,0.0000266966,0.05000,0.684
|
||
|
|
54,self_attn.v_proj,0.0000531413,0.05000,0.656
|
||
|
|
54,self_attn.q_proj,0.0001269680,0.05000,0.677
|
||
|
|
54,self_attn.o_proj,0.0000721886,0.05000,0.692
|
||
|
|
54,mlp.up_proj,0.0009590754,0.05000,0.756
|
||
|
|
54,mlp.gate_proj,0.0009435314,0.05000,0.739
|
||
|
|
54,mlp.down_proj,0.0003761099,0.05000,6.120
|
||
|
|
55,self_attn.k_proj,0.0000258792,0.05000,0.682
|
||
|
|
55,self_attn.v_proj,0.0000580578,0.05000,0.667
|
||
|
|
55,self_attn.q_proj,0.0001275678,0.05000,0.670
|
||
|
|
55,self_attn.o_proj,0.0000933949,0.05000,0.679
|
||
|
|
55,mlp.up_proj,0.0010799354,0.05000,0.758
|
||
|
|
55,mlp.gate_proj,0.0010489707,0.05000,0.749
|
||
|
|
55,mlp.down_proj,0.0004415629,0.05000,6.139
|
||
|
|
56,self_attn.k_proj,0.0000300139,0.05000,0.691
|
||
|
|
56,self_attn.v_proj,0.0000785556,0.05000,0.653
|
||
|
|
56,self_attn.q_proj,0.0001425975,0.05000,0.672
|
||
|
|
56,self_attn.o_proj,0.0000923193,0.05000,0.676
|
||
|
|
56,mlp.up_proj,0.0012177298,0.05000,0.760
|
||
|
|
56,mlp.gate_proj,0.0011750842,0.05000,0.746
|
||
|
|
56,mlp.down_proj,0.0005107334,0.05000,6.124
|
||
|
|
57,self_attn.k_proj,0.0000303419,0.05000,0.667
|
||
|
|
57,self_attn.v_proj,0.0000819483,0.05000,0.649
|
||
|
|
57,self_attn.q_proj,0.0001509253,0.05000,0.657
|
||
|
|
57,self_attn.o_proj,0.0000811847,0.05000,0.676
|
||
|
|
57,mlp.up_proj,0.0013414129,0.05000,0.760
|
||
|
|
57,mlp.gate_proj,0.0012758095,0.05000,0.747
|
||
|
|
57,mlp.down_proj,0.0005883840,0.05000,6.111
|
||
|
|
58,self_attn.k_proj,0.0000317013,0.05000,0.673
|
||
|
|
58,self_attn.v_proj,0.0000916080,0.05000,0.653
|
||
|
|
58,self_attn.q_proj,0.0001544135,0.05000,0.656
|
||
|
|
58,self_attn.o_proj,0.0000810032,0.05000,0.680
|
||
|
|
58,mlp.up_proj,0.0015102160,0.05000,0.758
|
||
|
|
58,mlp.gate_proj,0.0014192124,0.05000,0.749
|
||
|
|
58,mlp.down_proj,0.0007239864,0.05000,6.140
|
||
|
|
59,self_attn.k_proj,0.0000355026,0.05000,0.671
|
||
|
|
59,self_attn.v_proj,0.0001328919,0.05000,0.659
|
||
|
|
59,self_attn.q_proj,0.0001763182,0.05000,0.664
|
||
|
|
59,self_attn.o_proj,0.0001949633,0.05000,0.677
|
||
|
|
59,mlp.up_proj,0.0017215954,0.05000,0.766
|
||
|
|
59,mlp.gate_proj,0.0015890894,0.05000,0.760
|
||
|
|
59,mlp.down_proj,0.0009149372,0.05000,6.103
|
||
|
|
60,self_attn.k_proj,0.0000302485,0.05000,0.690
|
||
|
|
60,self_attn.v_proj,0.0001508339,0.05000,0.661
|
||
|
|
60,self_attn.q_proj,0.0001744292,0.05000,0.684
|
||
|
|
60,self_attn.o_proj,0.0004136908,0.05000,0.691
|
||
|
|
60,mlp.up_proj,0.0019182535,0.05000,0.772
|
||
|
|
60,mlp.gate_proj,0.0017420459,0.05000,0.762
|
||
|
|
60,mlp.down_proj,0.0018365022,0.05000,6.075
|
||
|
|
61,self_attn.k_proj,0.0000342876,0.05000,0.680
|
||
|
|
61,self_attn.v_proj,0.0001912820,0.05000,0.664
|
||
|
|
61,self_attn.q_proj,0.0001930302,0.05000,0.670
|
||
|
|
61,self_attn.o_proj,0.0005102926,0.05000,0.691
|
||
|
|
61,mlp.up_proj,0.0022156300,0.05000,0.776
|
||
|
|
61,mlp.gate_proj,0.0020273371,0.05000,0.750
|
||
|
|
61,mlp.down_proj,0.0021458070,0.05000,6.108
|
||
|
|
62,self_attn.k_proj,0.0000345618,0.05000,0.689
|
||
|
|
62,self_attn.v_proj,0.0002236197,0.05000,0.660
|
||
|
|
62,self_attn.q_proj,0.0001984577,0.05000,0.684
|
||
|
|
62,self_attn.o_proj,0.0006962981,0.05000,0.695
|
||
|
|
62,mlp.up_proj,0.0025008655,0.05000,0.758
|
||
|
|
62,mlp.gate_proj,0.0023534444,0.05000,0.741
|
||
|
|
62,mlp.down_proj,0.0041401221,0.05000,6.124
|
||
|
|
63,self_attn.k_proj,0.0000385404,0.05000,0.668
|
||
|
|
63,self_attn.v_proj,0.0001727117,0.05000,0.653
|
||
|
|
63,self_attn.q_proj,0.0001995420,0.05000,0.664
|
||
|
|
63,self_attn.o_proj,0.0006111872,0.05000,0.674
|
||
|
|
63,mlp.up_proj,0.0040122621,0.05000,0.760
|
||
|
|
63,mlp.gate_proj,0.0038694526,0.05000,0.756
|
||
|
|
63,mlp.down_proj,0.0146558697,0.05000,6.063
|