File size: 11,667 Bytes

9ddf1f7

layer,module,loss,samples,damp,time
0,self_attn.q_proj,0.0000003892,0.05000,4.898
0,self_attn.v_proj,0.0000000914,0.05000,4.964
0,self_attn.k_proj,0.0000001061,0.05000,4.994
0,self_attn.o_proj,0.0000002403,0.05000,1.686
0,mlp.gate_proj,0.0000248745,0.05000,2.588
0,mlp.up_proj,0.0000215972,0.05000,2.621
0,mlp.down_proj,0.0000025529,0.05000,4.167
1,self_attn.k_proj,0.0000001765,0.05000,4.454
1,self_attn.v_proj,0.0000001794,0.05000,4.541
1,self_attn.q_proj,0.0000006520,0.05000,4.572
1,self_attn.o_proj,0.0000002912,0.05000,1.417
1,mlp.gate_proj,0.0008361811,0.05000,2.121
1,mlp.up_proj,0.0003858884,0.05000,2.142
1,mlp.down_proj,0.0000039680,0.05000,3.713
2,self_attn.v_proj,0.0000004924,0.05000,4.481
2,self_attn.k_proj,0.0000005247,0.05000,4.530
2,self_attn.q_proj,0.0000018126,0.05000,4.564
2,self_attn.o_proj,0.0000004236,0.05000,1.428
2,mlp.gate_proj,0.0014728842,0.05000,2.118
2,mlp.up_proj,0.0011552544,0.05000,2.136
2,mlp.down_proj,0.0000031700,0.05000,3.747
3,self_attn.v_proj,0.0000009231,0.05000,4.443
3,self_attn.k_proj,0.0000009410,0.05000,4.494
3,self_attn.q_proj,0.0000036298,0.05000,4.515
3,self_attn.o_proj,0.0000006802,0.05000,1.470
3,mlp.up_proj,0.0006774332,0.05000,2.075
3,mlp.gate_proj,0.0013199807,0.05000,2.098
3,mlp.down_proj,0.0000069405,0.05000,3.733
4,self_attn.q_proj,0.0000065567,0.05000,4.445
4,self_attn.k_proj,0.0000018016,0.05000,4.494
4,self_attn.v_proj,0.0000017432,0.05000,4.519
4,self_attn.o_proj,0.0000010358,0.05000,1.419
4,mlp.up_proj,0.0004369767,0.05000,2.079
4,mlp.gate_proj,0.0011432846,0.05000,2.112
4,mlp.down_proj,0.0000079352,0.05000,3.826
5,self_attn.v_proj,0.0000019041,0.05000,4.476
5,self_attn.k_proj,0.0000018155,0.05000,4.476
5,self_attn.q_proj,0.0000070605,0.05000,4.539
5,self_attn.o_proj,0.0000018124,0.05000,1.440
5,mlp.up_proj,0.0001843741,0.05000,2.105
5,mlp.gate_proj,0.0004063439,0.05000,2.131
5,mlp.down_proj,0.0000111709,0.05000,3.842
6,self_attn.v_proj,0.0000040331,0.05000,4.427
6,self_attn.k_proj,0.0000036274,0.05000,4.473
6,self_attn.q_proj,0.0000152004,0.05000,4.489
6,self_attn.o_proj,0.0000041757,0.05000,1.448
6,mlp.gate_proj,0.0004484141,0.05000,2.256
6,mlp.up_proj,0.0002336842,0.05000,2.285
6,mlp.down_proj,0.0005620144,0.05000,3.773
7,self_attn.v_proj,0.0000081437,0.05000,4.465
7,self_attn.q_proj,0.0000311665,0.05000,4.572
7,self_attn.k_proj,0.0000085449,0.05000,4.573
7,self_attn.o_proj,0.0000051851,0.05000,1.443
7,mlp.up_proj,0.0002692974,0.05000,2.137
7,mlp.gate_proj,0.0004860758,0.05000,2.168
7,mlp.down_proj,0.0000193441,0.05000,3.857
8,self_attn.v_proj,0.0000124433,0.05000,4.425
8,self_attn.q_proj,0.0000461009,0.05000,4.430
8,self_attn.k_proj,0.0000120985,0.05000,4.472
8,self_attn.o_proj,0.0000070641,0.05000,1.454
8,mlp.gate_proj,0.0003251363,0.05000,1.991
8,mlp.up_proj,0.0002605974,0.05000,1.988
8,mlp.down_proj,0.0000314368,0.05000,3.883
9,self_attn.k_proj,0.0000169086,0.05000,4.294
9,self_attn.q_proj,0.0000589728,0.05000,4.408
9,self_attn.v_proj,0.0000152795,0.05000,4.425
9,self_attn.o_proj,0.0000086252,0.05000,1.540
9,mlp.gate_proj,0.0005424109,0.05000,2.123
9,mlp.up_proj,0.0003351726,0.05000,2.146
9,mlp.down_proj,0.0000309148,0.05000,3.826
10,self_attn.v_proj,0.0000244346,0.05000,4.451
10,self_attn.q_proj,0.0000857773,0.05000,4.490
10,self_attn.k_proj,0.0000222903,0.05000,4.521
10,self_attn.o_proj,0.0000124682,0.05000,1.459
10,mlp.gate_proj,0.0004176983,0.05000,1.985
10,mlp.up_proj,0.0002851436,0.05000,2.013
10,mlp.down_proj,0.0000257911,0.05000,3.751
11,self_attn.q_proj,0.0000414936,0.05000,4.344
11,self_attn.v_proj,0.0000111498,0.05000,4.412
11,self_attn.k_proj,0.0000114142,0.05000,4.420
11,self_attn.o_proj,0.0000084636,0.05000,1.479
11,mlp.up_proj,0.0002661766,0.05000,2.072
11,mlp.gate_proj,0.0003413026,0.05000,2.094
11,mlp.down_proj,0.0000248982,0.05000,3.852
12,self_attn.v_proj,0.0000136642,0.05000,4.438
12,self_attn.k_proj,0.0000133012,0.05000,4.487
12,self_attn.q_proj,0.0000495074,0.05000,4.490
12,self_attn.o_proj,0.0000102142,0.05000,1.436
12,mlp.gate_proj,0.0003048110,0.05000,2.085
12,mlp.up_proj,0.0002618897,0.05000,2.117
12,mlp.down_proj,0.0000251344,0.05000,3.762
13,self_attn.q_proj,0.0000380912,0.05000,4.432
13,self_attn.v_proj,0.0000094012,0.05000,4.432
13,self_attn.k_proj,0.0000102629,0.05000,4.455
13,self_attn.o_proj,0.0000085788,0.05000,1.441
13,mlp.gate_proj,0.0002850975,0.05000,2.084
13,mlp.up_proj,0.0002683511,0.05000,2.110
13,mlp.down_proj,0.0000268963,0.05000,3.765
14,self_attn.q_proj,0.0000627420,0.05000,4.459
14,self_attn.k_proj,0.0000162856,0.05000,4.523
14,self_attn.v_proj,0.0000165909,0.05000,4.555
14,self_attn.o_proj,0.0000132363,0.05000,1.429
14,mlp.up_proj,0.0002667329,0.05000,2.120
14,mlp.gate_proj,0.0002825767,0.05000,2.155
14,mlp.down_proj,0.0000251557,0.05000,3.783
15,self_attn.k_proj,0.0000167620,0.05000,4.314
15,self_attn.v_proj,0.0000153608,0.05000,4.378
15,self_attn.q_proj,0.0000641311,0.05000,4.410
15,self_attn.o_proj,0.0000114332,0.05000,1.439
15,mlp.up_proj,0.0002600501,0.05000,2.165
15,mlp.gate_proj,0.0002651350,0.05000,2.185
15,mlp.down_proj,0.0000244735,0.05000,3.813
16,self_attn.v_proj,0.0000304125,0.05000,4.336
16,self_attn.q_proj,0.0001121167,0.05000,4.425
16,self_attn.k_proj,0.0000278084,0.05000,4.450
16,self_attn.o_proj,0.0000125121,0.05000,1.439
16,mlp.gate_proj,0.0003042360,0.05000,2.117
16,mlp.up_proj,0.0002817502,0.05000,2.139
16,mlp.down_proj,0.0003102970,0.05000,3.794
17,self_attn.k_proj,0.0000216207,0.05000,4.342
17,self_attn.q_proj,0.0000981481,0.05000,4.380
17,self_attn.v_proj,0.0000235827,0.05000,4.421
17,self_attn.o_proj,0.0000135337,0.05000,1.455
17,mlp.gate_proj,0.0002716687,0.05000,2.068
17,mlp.up_proj,0.0002553188,0.05000,2.086
17,mlp.down_proj,0.0000246212,0.05000,3.851
18,self_attn.v_proj,0.0000281725,0.05000,4.252
18,self_attn.q_proj,0.0001061965,0.05000,4.310
18,self_attn.k_proj,0.0000259232,0.05000,4.339
18,self_attn.o_proj,0.0000173272,0.05000,1.469
18,mlp.up_proj,0.0002783262,0.05000,2.040
18,mlp.gate_proj,0.0002907026,0.05000,2.060
18,mlp.down_proj,0.0000303231,0.05000,3.794
19,self_attn.k_proj,0.0000492216,0.05000,4.381
19,self_attn.v_proj,0.0000490726,0.05000,4.433
19,self_attn.q_proj,0.0002095862,0.05000,4.480
19,self_attn.o_proj,0.0000210620,0.05000,1.460
19,mlp.up_proj,0.0003063424,0.05000,1.990
19,mlp.gate_proj,0.0003174873,0.05000,2.012
19,mlp.down_proj,0.0000387613,0.05000,3.854
20,self_attn.k_proj,0.0000430631,0.05000,4.405
20,self_attn.q_proj,0.0001980007,0.05000,4.477
20,self_attn.v_proj,0.0000455319,0.05000,4.506
20,self_attn.o_proj,0.0000226272,0.05000,1.474
20,mlp.gate_proj,0.0003353285,0.05000,2.040
20,mlp.up_proj,0.0003304923,0.05000,2.068
20,mlp.down_proj,0.0000424498,0.05000,3.794
21,self_attn.q_proj,0.0002467767,0.05000,4.283
21,self_attn.k_proj,0.0000562617,0.05000,4.328
21,self_attn.v_proj,0.0000601845,0.05000,4.356
21,self_attn.o_proj,0.0000256164,0.05000,1.463
21,mlp.up_proj,0.0003876656,0.05000,1.998
21,mlp.gate_proj,0.0003971141,0.05000,2.022
21,mlp.down_proj,0.0000551685,0.05000,3.871
22,self_attn.v_proj,0.0001205271,0.05000,4.378
22,self_attn.k_proj,0.0001126014,0.05000,4.436
22,self_attn.q_proj,0.0004939758,0.05000,4.459
22,self_attn.o_proj,0.0000473019,0.05000,1.485
22,mlp.up_proj,0.0004406472,0.05000,2.101
22,mlp.gate_proj,0.0004752020,0.05000,2.119
22,mlp.down_proj,0.0000884866,0.05000,3.858
23,self_attn.q_proj,0.0004799678,0.05000,4.516
23,self_attn.v_proj,0.0001249926,0.05000,4.563
23,self_attn.k_proj,0.0001114300,0.05000,4.571
23,self_attn.o_proj,0.0000583897,0.05000,1.460
23,mlp.up_proj,0.0005192255,0.05000,2.057
23,mlp.gate_proj,0.0005905081,0.05000,2.082
23,mlp.down_proj,0.0001205872,0.05000,3.755
24,self_attn.k_proj,0.0001827744,0.05000,4.333
24,self_attn.v_proj,0.0002347561,0.05000,4.400
24,self_attn.q_proj,0.0007712197,0.05000,4.434
24,self_attn.o_proj,0.0000619258,0.05000,1.433
24,mlp.gate_proj,0.0006641536,0.05000,2.010
24,mlp.up_proj,0.0005745464,0.05000,2.016
24,mlp.down_proj,0.0001467679,0.05000,3.859
25,self_attn.q_proj,0.0005373499,0.05000,4.433
25,self_attn.k_proj,0.0001349687,0.05000,4.473
25,self_attn.v_proj,0.0001483015,0.05000,4.499
25,self_attn.o_proj,0.0000400881,0.05000,1.459
25,mlp.gate_proj,0.0007548458,0.05000,2.121
25,mlp.up_proj,0.0006440500,0.05000,2.149
25,mlp.down_proj,0.0001719484,0.05000,3.805
26,self_attn.k_proj,0.0002035127,0.05000,4.441
26,self_attn.q_proj,0.0008599364,0.05000,4.477
26,self_attn.v_proj,0.0002357316,0.05000,4.492
26,self_attn.o_proj,0.0000370940,0.05000,1.434
26,mlp.up_proj,0.0007832906,0.05000,2.080
26,mlp.gate_proj,0.0008894922,0.05000,2.103
26,mlp.down_proj,0.0002024437,0.05000,3.804
27,self_attn.q_proj,0.0010485745,0.05000,4.313
27,self_attn.v_proj,0.0003151497,0.05000,4.392
27,self_attn.k_proj,0.0002427360,0.05000,4.415
27,self_attn.o_proj,0.0000494488,0.05000,1.487
27,mlp.gate_proj,0.0009806582,0.05000,2.119
27,mlp.up_proj,0.0008931541,0.05000,2.144
27,mlp.down_proj,0.0002912259,0.05000,3.804
28,self_attn.v_proj,0.0003319008,0.05000,4.377
28,self_attn.q_proj,0.0011587325,0.05000,4.435
28,self_attn.k_proj,0.0002960693,0.05000,4.458
28,self_attn.o_proj,0.0000796063,0.05000,1.483
28,mlp.up_proj,0.0010471307,0.05000,2.063
28,mlp.gate_proj,0.0011029143,0.05000,2.087
28,mlp.down_proj,0.0004399204,0.05000,3.750
29,self_attn.q_proj,0.0027975505,0.05000,4.514
29,self_attn.v_proj,0.0009092517,0.05000,4.587
29,self_attn.k_proj,0.0006667851,0.05000,4.612
29,self_attn.o_proj,0.0000638855,0.05000,1.417
29,mlp.up_proj,0.0012708087,0.05000,2.009
29,mlp.gate_proj,0.0012681824,0.05000,2.030
29,mlp.down_proj,0.0004382937,0.05000,3.804
30,self_attn.q_proj,0.0030858449,0.05000,4.423
30,self_attn.k_proj,0.0007915465,0.05000,4.480
30,self_attn.v_proj,0.0009827334,0.05000,4.496
30,self_attn.o_proj,0.0001427864,0.05000,1.448
30,mlp.gate_proj,0.0013024663,0.05000,2.031
30,mlp.up_proj,0.0013509312,0.05000,2.066
30,mlp.down_proj,0.0005863545,0.05000,3.799
31,self_attn.q_proj,0.0039200108,0.05000,4.377
31,self_attn.k_proj,0.0010847513,0.05000,4.434
31,self_attn.v_proj,0.0014725985,0.05000,4.466
31,self_attn.o_proj,0.0001331001,0.05000,1.455
31,mlp.gate_proj,0.0012615698,0.05000,2.082
31,mlp.up_proj,0.0013760641,0.05000,2.110
31,mlp.down_proj,0.0007124285,0.05000,3.789
32,self_attn.k_proj,0.0014174075,0.05000,4.457
32,self_attn.q_proj,0.0056738988,0.05000,4.468
32,self_attn.v_proj,0.0021258397,0.05000,4.533
32,self_attn.o_proj,0.0001640287,0.05000,1.476
32,mlp.up_proj,0.0014403572,0.05000,1.993
32,mlp.gate_proj,0.0012891649,0.05000,2.017
32,mlp.down_proj,0.0008746092,0.05000,3.744
33,self_attn.k_proj,0.0026198407,0.05000,4.279
33,self_attn.q_proj,0.0125964452,0.05000,4.312
33,self_attn.v_proj,0.0049467264,0.05000,4.353
33,self_attn.o_proj,0.0002326117,0.05000,1.468
33,mlp.gate_proj,0.0014007832,0.05000,2.023
33,mlp.up_proj,0.0015872569,0.05000,2.036
33,mlp.down_proj,0.0011593697,0.05000,3.832
34,self_attn.q_proj,0.0094372779,0.05000,4.304
34,self_attn.k_proj,0.0021265713,0.05000,4.424
34,self_attn.v_proj,0.0034324985,0.05000,4.441
34,self_attn.o_proj,0.0004474311,0.05000,1.478
34,mlp.gate_proj,0.0016823788,0.05000,2.039
34,mlp.up_proj,0.0017845972,0.05000,2.038
34,mlp.down_proj,0.0014755854,0.05000,3.751
35,self_attn.v_proj,0.0013864043,0.05000,4.366
35,self_attn.k_proj,0.0010910282,0.05000,4.469
35,self_attn.q_proj,0.0041450113,0.05000,4.472
35,self_attn.o_proj,0.0005922779,0.05000,1.439
35,mlp.gate_proj,0.0033476329,0.05000,2.023
35,mlp.up_proj,0.0038470728,0.05000,2.051
35,mlp.down_proj,0.0026691593,0.05000,3.786