|
layer,module,loss,samples,damp,time
|
|
0,self_attn.k_proj,0.0000614463,0.01000,1.336
|
|
0,self_attn.v_proj,0.0000367260,0.01000,0.409
|
|
0,self_attn.q_proj,0.0001768391,0.01000,0.403
|
|
0,self_attn.o_proj,0.0000525409,0.01000,0.407
|
|
0,mlp.up_proj,0.0002206423,0.01000,0.488
|
|
0,mlp.gate_proj,0.0001312029,0.01000,0.466
|
|
0,mlp.down_proj,0.0000269155,0.01000,1.590
|
|
1,self_attn.k_proj,0.0001003957,0.01000,0.396
|
|
1,self_attn.v_proj,0.0000333207,0.01000,0.391
|
|
1,self_attn.q_proj,0.0002748014,0.01000,0.413
|
|
1,self_attn.o_proj,0.0000217966,0.01000,0.405
|
|
1,mlp.up_proj,0.0003864400,0.01000,0.465
|
|
1,mlp.gate_proj,0.0002352389,0.01000,0.469
|
|
1,mlp.down_proj,0.0000286544,0.01000,1.548
|
|
2,self_attn.k_proj,0.0001131236,0.01000,0.430
|
|
2,self_attn.v_proj,0.0000305789,0.01000,0.426
|
|
2,self_attn.q_proj,0.0002841453,0.01000,0.406
|
|
2,self_attn.o_proj,0.0000085877,0.01000,0.416
|
|
2,mlp.up_proj,0.0003458275,0.01000,0.480
|
|
2,mlp.gate_proj,0.0002811401,0.01000,0.478
|
|
2,mlp.down_proj,0.0000450885,0.01000,1.610
|
|
3,self_attn.k_proj,0.0001865263,0.01000,0.404
|
|
3,self_attn.v_proj,0.0000615749,0.01000,0.402
|
|
3,self_attn.q_proj,0.0005584807,0.01000,0.442
|
|
3,self_attn.o_proj,0.0000052220,0.01000,0.444
|
|
3,mlp.up_proj,0.0003472440,0.01000,0.472
|
|
3,mlp.gate_proj,0.0003216986,0.01000,0.476
|
|
3,mlp.down_proj,0.0000204757,0.01000,1.556
|
|
4,self_attn.k_proj,0.0001783940,0.01000,0.404
|
|
4,self_attn.v_proj,0.0000633529,0.01000,0.393
|
|
4,self_attn.q_proj,0.0005186100,0.01000,0.400
|
|
4,self_attn.o_proj,0.0000038393,0.01000,0.406
|
|
4,mlp.up_proj,0.0003347796,0.01000,0.475
|
|
4,mlp.gate_proj,0.0003114954,0.01000,0.473
|
|
4,mlp.down_proj,0.0000275217,0.01000,1.561
|
|
5,self_attn.k_proj,0.0002603203,0.01000,0.407
|
|
5,self_attn.v_proj,0.0000651183,0.01000,0.400
|
|
5,self_attn.q_proj,0.0007260548,0.01000,0.412
|
|
5,self_attn.o_proj,0.0000032502,0.01000,0.410
|
|
5,mlp.up_proj,0.0003240574,0.01000,0.479
|
|
5,mlp.gate_proj,0.0002856003,0.01000,0.478
|
|
5,mlp.down_proj,0.0012586215,0.01000,1.594
|
|
6,self_attn.k_proj,0.0002431215,0.01000,0.401
|
|
6,self_attn.v_proj,0.0001084795,0.01000,0.398
|
|
6,self_attn.q_proj,0.0007484398,0.01000,0.408
|
|
6,self_attn.o_proj,0.0000047513,0.01000,0.413
|
|
6,mlp.up_proj,0.0003498730,0.01000,0.478
|
|
6,mlp.gate_proj,0.0003110082,0.01000,0.484
|
|
6,mlp.down_proj,0.0000180229,0.01000,1.576
|
|
7,self_attn.k_proj,0.0001716809,0.01000,0.403
|
|
7,self_attn.v_proj,0.0000643876,0.01000,0.411
|
|
7,self_attn.q_proj,0.0005089845,0.01000,0.404
|
|
7,self_attn.o_proj,0.0000057735,0.01000,0.408
|
|
7,mlp.up_proj,0.0003251847,0.01000,0.475
|
|
7,mlp.gate_proj,0.0002875912,0.01000,0.467
|
|
7,mlp.down_proj,0.0000201953,0.01000,1.553
|
|
8,self_attn.k_proj,0.0002251871,0.01000,0.411
|
|
8,self_attn.v_proj,0.0000774253,0.01000,0.405
|
|
8,self_attn.q_proj,0.0006696486,0.01000,0.412
|
|
8,self_attn.o_proj,0.0000050467,0.01000,0.412
|
|
8,mlp.up_proj,0.0003079109,0.01000,0.472
|
|
8,mlp.gate_proj,0.0002666798,0.01000,0.468
|
|
8,mlp.down_proj,0.0000208221,0.01000,1.566
|
|
9,self_attn.k_proj,0.0002614264,0.01000,0.402
|
|
9,self_attn.v_proj,0.0000689843,0.01000,0.400
|
|
9,self_attn.q_proj,0.0007858098,0.01000,0.409
|
|
9,self_attn.o_proj,0.0000068903,0.01000,0.406
|
|
9,mlp.up_proj,0.0003278151,0.01000,0.467
|
|
9,mlp.gate_proj,0.0002659284,0.01000,0.466
|
|
9,mlp.down_proj,0.0000255585,0.01000,1.563
|
|
10,self_attn.k_proj,0.0002716511,0.01000,0.396
|
|
10,self_attn.v_proj,0.0000627879,0.01000,0.392
|
|
10,self_attn.q_proj,0.0007019225,0.01000,0.396
|
|
10,self_attn.o_proj,0.0000108618,0.01000,0.413
|
|
10,mlp.up_proj,0.0002879963,0.01000,0.466
|
|
10,mlp.gate_proj,0.0002357466,0.01000,0.474
|
|
10,mlp.down_proj,0.0000297172,0.01000,1.563
|
|
11,self_attn.k_proj,0.0002967386,0.01000,0.393
|
|
11,self_attn.v_proj,0.0000832104,0.01000,0.393
|
|
11,self_attn.q_proj,0.0008514731,0.01000,0.397
|
|
11,self_attn.o_proj,0.0000116954,0.01000,0.413
|
|
11,mlp.up_proj,0.0002762500,0.01000,0.466
|
|
11,mlp.gate_proj,0.0002265258,0.01000,0.464
|
|
11,mlp.down_proj,0.0000312174,0.01000,1.572
|
|
12,self_attn.k_proj,0.0003790672,0.01000,0.427
|
|
12,self_attn.v_proj,0.0000921510,0.01000,0.405
|
|
12,self_attn.q_proj,0.0011573786,0.01000,0.407
|
|
12,self_attn.o_proj,0.0000102819,0.01000,0.407
|
|
12,mlp.up_proj,0.0003061092,0.01000,0.482
|
|
12,mlp.gate_proj,0.0002483440,0.01000,0.470
|
|
12,mlp.down_proj,0.0000369618,0.01000,1.566
|
|
13,self_attn.k_proj,0.0003544405,0.01000,0.408
|
|
13,self_attn.v_proj,0.0001215898,0.01000,0.398
|
|
13,self_attn.q_proj,0.0010194392,0.01000,0.399
|
|
13,self_attn.o_proj,0.0000169550,0.01000,0.406
|
|
13,mlp.up_proj,0.0003151845,0.01000,0.486
|
|
13,mlp.gate_proj,0.0002525032,0.01000,0.521
|
|
13,mlp.down_proj,0.0000404234,0.01000,1.607
|
|
14,self_attn.k_proj,0.0003262086,0.01000,0.445
|
|
14,self_attn.v_proj,0.0001572506,0.01000,0.444
|
|
14,self_attn.q_proj,0.0009974413,0.01000,0.453
|
|
14,self_attn.o_proj,0.0000163447,0.01000,0.407
|
|
14,mlp.up_proj,0.0003339299,0.01000,0.468
|
|
14,mlp.gate_proj,0.0002618677,0.01000,0.473
|
|
14,mlp.down_proj,0.0000433021,0.01000,1.563
|
|
15,self_attn.k_proj,0.0003082260,0.01000,0.397
|
|
15,self_attn.v_proj,0.0001545105,0.01000,0.397
|
|
15,self_attn.q_proj,0.0009607012,0.01000,0.406
|
|
15,self_attn.o_proj,0.0000208320,0.01000,0.405
|
|
15,mlp.up_proj,0.0003421122,0.01000,0.469
|
|
15,mlp.gate_proj,0.0002682374,0.01000,0.465
|
|
15,mlp.down_proj,0.0000485211,0.01000,1.563
|
|
16,self_attn.k_proj,0.0003386762,0.01000,0.411
|
|
16,self_attn.v_proj,0.0001445804,0.01000,0.401
|
|
16,self_attn.q_proj,0.0009659624,0.01000,0.427
|
|
16,self_attn.o_proj,0.0000269696,0.01000,0.409
|
|
16,mlp.up_proj,0.0003653567,0.01000,0.470
|
|
16,mlp.gate_proj,0.0002830245,0.01000,0.474
|
|
16,mlp.down_proj,0.0000531954,0.01000,1.548
|
|
17,self_attn.k_proj,0.0003605710,0.01000,0.394
|
|
17,self_attn.v_proj,0.0001402870,0.01000,0.390
|
|
17,self_attn.q_proj,0.0010452676,0.01000,0.402
|
|
17,self_attn.o_proj,0.0000135152,0.01000,0.484
|
|
17,mlp.up_proj,0.0004317064,0.01000,0.470
|
|
17,mlp.gate_proj,0.0003206919,0.01000,0.507
|
|
17,mlp.down_proj,0.0000544806,0.01000,1.556
|
|
18,self_attn.k_proj,0.0003156988,0.01000,0.395
|
|
18,self_attn.v_proj,0.0001398579,0.01000,0.394
|
|
18,self_attn.q_proj,0.0009304984,0.01000,0.403
|
|
18,self_attn.o_proj,0.0000195992,0.01000,0.408
|
|
18,mlp.up_proj,0.0004443839,0.01000,0.470
|
|
18,mlp.gate_proj,0.0003426512,0.01000,0.494
|
|
18,mlp.down_proj,0.0000620541,0.01000,1.566
|
|
19,self_attn.k_proj,0.0003820088,0.01000,0.398
|
|
19,self_attn.v_proj,0.0001864415,0.01000,0.393
|
|
19,self_attn.q_proj,0.0011555548,0.01000,0.399
|
|
19,self_attn.o_proj,0.0000171530,0.01000,0.402
|
|
19,mlp.up_proj,0.0004638179,0.01000,0.465
|
|
19,mlp.gate_proj,0.0003661488,0.01000,0.464
|
|
19,mlp.down_proj,0.0000766811,0.01000,1.584
|
|
20,self_attn.k_proj,0.0000000000,0.01000,0.403
|
|
20,self_attn.v_proj,0.0000001070,0.01000,0.393
|
|
20,self_attn.q_proj,0.0000000000,0.01000,0.446
|
|
20,self_attn.o_proj,0.0000000018,0.01000,0.450
|
|
20,mlp.up_proj,0.0005112301,0.01000,0.467
|
|
20,mlp.gate_proj,0.0004270064,0.01000,0.466
|
|
20,mlp.down_proj,0.0000762542,0.01000,1.551
|
|
21,self_attn.k_proj,0.0004718139,0.01000,0.407
|
|
21,self_attn.v_proj,0.0002004363,0.01000,0.404
|
|
21,self_attn.q_proj,0.0013466007,0.01000,0.409
|
|
21,self_attn.o_proj,0.0000155370,0.01000,0.410
|
|
21,mlp.up_proj,0.0005676214,0.01000,0.510
|
|
21,mlp.gate_proj,0.0004834948,0.01000,0.469
|
|
21,mlp.down_proj,0.0000928768,0.01000,1.560
|
|
22,self_attn.k_proj,0.0004762067,0.01000,0.406
|
|
22,self_attn.v_proj,0.0002690355,0.01000,0.439
|
|
22,self_attn.q_proj,0.0014745827,0.01000,0.442
|
|
22,self_attn.o_proj,0.0000086187,0.01000,0.406
|
|
22,mlp.up_proj,0.0006510282,0.01000,0.468
|
|
22,mlp.gate_proj,0.0005652614,0.01000,0.468
|
|
22,mlp.down_proj,0.0001044421,0.01000,1.542
|
|
23,self_attn.k_proj,0.0000000000,0.01000,0.402
|
|
23,self_attn.v_proj,0.0000000519,0.01000,0.387
|
|
23,self_attn.q_proj,0.0000000000,0.01000,0.411
|
|
23,self_attn.o_proj,0.0000000004,0.01000,0.406
|
|
23,mlp.up_proj,0.0007391464,0.01000,0.468
|
|
23,mlp.gate_proj,0.0006502947,0.01000,0.465
|
|
23,mlp.down_proj,0.0001170578,0.01000,1.550
|
|
24,self_attn.k_proj,0.0000000000,0.01000,0.405
|
|
24,self_attn.v_proj,0.0000000586,0.01000,0.427
|
|
24,self_attn.q_proj,0.0000000000,0.01000,0.405
|
|
24,self_attn.o_proj,0.0000000004,0.01000,0.408
|
|
24,mlp.up_proj,0.0007253281,0.01000,0.469
|
|
24,mlp.gate_proj,0.0006449697,0.01000,0.463
|
|
24,mlp.down_proj,0.0001037727,0.01000,1.722
|
|
25,self_attn.k_proj,0.0005589081,0.01000,0.437
|
|
25,self_attn.v_proj,0.0003744282,0.01000,0.392
|
|
25,self_attn.q_proj,0.0017770897,0.01000,0.401
|
|
25,self_attn.o_proj,0.0000130858,0.01000,0.414
|
|
25,mlp.up_proj,0.0008171453,0.01000,0.467
|
|
25,mlp.gate_proj,0.0007300299,0.01000,0.465
|
|
25,mlp.down_proj,0.0001281814,0.01000,1.545
|
|
26,self_attn.k_proj,0.0000000000,0.01000,0.401
|
|
26,self_attn.v_proj,0.0000000966,0.01000,0.399
|
|
26,self_attn.q_proj,0.0000000000,0.01000,0.405
|
|
26,self_attn.o_proj,0.0000000006,0.01000,0.407
|
|
26,mlp.up_proj,0.0009040351,0.01000,0.465
|
|
26,mlp.gate_proj,0.0008104020,0.01000,0.462
|
|
26,mlp.down_proj,0.0001466325,0.01000,1.568
|
|
27,self_attn.k_proj,0.0000000000,0.01000,0.399
|
|
27,self_attn.v_proj,0.0000001440,0.01000,0.391
|
|
27,self_attn.q_proj,0.0000000000,0.01000,0.401
|
|
27,self_attn.o_proj,0.0000000011,0.01000,0.414
|
|
27,mlp.up_proj,0.0010109719,0.01000,0.466
|
|
27,mlp.gate_proj,0.0009065109,0.01000,0.462
|
|
27,mlp.down_proj,0.0001677109,0.01000,1.534
|
|
28,self_attn.k_proj,0.0005801313,0.01000,0.395
|
|
28,self_attn.v_proj,0.0004394035,0.01000,0.395
|
|
28,self_attn.q_proj,0.0019255374,0.01000,0.447
|
|
28,self_attn.o_proj,0.0000255029,0.01000,0.442
|
|
28,mlp.up_proj,0.0011190363,0.01000,0.511
|
|
28,mlp.gate_proj,0.0009941222,0.01000,0.510
|
|
28,mlp.down_proj,0.0002010200,0.01000,1.683
|
|
29,self_attn.k_proj,0.0000000000,0.01000,0.404
|
|
29,self_attn.v_proj,0.0000006639,0.01000,0.414
|
|
29,self_attn.q_proj,0.0000000000,0.01000,0.455
|
|
29,self_attn.o_proj,0.0000000083,0.01000,0.403
|
|
29,mlp.up_proj,0.0012196709,0.01000,0.465
|
|
29,mlp.gate_proj,0.0010653910,0.01000,0.476
|
|
29,mlp.down_proj,0.0002349926,0.01000,1.555
|
|
30,self_attn.k_proj,0.0005828383,0.01000,0.394
|
|
30,self_attn.v_proj,0.0005253269,0.01000,0.398
|
|
30,self_attn.q_proj,0.0020789497,0.01000,0.403
|
|
30,self_attn.o_proj,0.0000383137,0.01000,0.401
|
|
30,mlp.up_proj,0.0013405440,0.01000,0.543
|
|
30,mlp.gate_proj,0.0011444824,0.01000,0.465
|
|
30,mlp.down_proj,0.0002900253,0.01000,1.660
|
|
31,self_attn.k_proj,0.0006085422,0.01000,0.404
|
|
31,self_attn.v_proj,0.0005152642,0.01000,0.405
|
|
31,self_attn.q_proj,0.0020584779,0.01000,0.395
|
|
31,self_attn.o_proj,0.0000598492,0.01000,0.404
|
|
31,mlp.up_proj,0.0014893149,0.01000,0.471
|
|
31,mlp.gate_proj,0.0012367403,0.01000,0.475
|
|
31,mlp.down_proj,0.0003556967,0.01000,1.581
|
|
32,self_attn.k_proj,0.0007163941,0.01000,0.396
|
|
32,self_attn.v_proj,0.0008300889,0.01000,0.441
|
|
32,self_attn.q_proj,0.0026545515,0.01000,0.445
|
|
32,self_attn.o_proj,0.0000486150,0.01000,0.408
|
|
32,mlp.up_proj,0.0017586935,0.01000,0.473
|
|
32,mlp.gate_proj,0.0014310041,0.01000,0.514
|
|
32,mlp.down_proj,0.0005898569,0.01000,1.556
|
|
33,self_attn.k_proj,0.0005947711,0.01000,0.398
|
|
33,self_attn.v_proj,0.0006349163,0.01000,0.404
|
|
33,self_attn.q_proj,0.0021236858,0.01000,0.417
|
|
33,self_attn.o_proj,0.0000946185,0.01000,0.453
|
|
33,mlp.up_proj,0.0019140079,0.01000,0.519
|
|
33,mlp.gate_proj,0.0015256776,0.01000,0.467
|
|
33,mlp.down_proj,0.0007962697,0.01000,1.644
|
|
34,self_attn.k_proj,0.0005864225,0.01000,0.394
|
|
34,self_attn.v_proj,0.0009883775,0.01000,0.391
|
|
34,self_attn.q_proj,0.0023040078,0.01000,0.401
|
|
34,self_attn.o_proj,0.0001355599,0.01000,0.402
|
|
34,mlp.up_proj,0.0023079843,0.01000,0.468
|
|
34,mlp.gate_proj,0.0017628265,0.01000,0.476
|
|
34,mlp.down_proj,0.0010113104,0.01000,1.545
|
|
35,self_attn.k_proj,0.0005291416,0.01000,0.397
|
|
35,self_attn.v_proj,0.0010603956,0.01000,0.398
|
|
35,self_attn.q_proj,0.0022376764,0.01000,0.405
|
|
35,self_attn.o_proj,0.0001632905,0.01000,0.409
|
|
35,mlp.up_proj,0.0026185449,0.01000,0.477
|
|
35,mlp.gate_proj,0.0019186655,0.01000,0.467
|
|
35,mlp.down_proj,0.0015119576,0.01000,1.563
|
|
36,self_attn.k_proj,0.0004839136,0.01000,0.399
|
|
36,self_attn.v_proj,0.0016906889,0.01000,0.395
|
|
36,self_attn.q_proj,0.0021772609,0.01000,0.400
|
|
36,self_attn.o_proj,0.0001481065,0.01000,0.404
|
|
36,mlp.up_proj,0.0033432513,0.01000,0.522
|
|
36,mlp.gate_proj,0.0024389136,0.01000,0.518
|
|
36,mlp.down_proj,0.0021110890,0.01000,1.551
|
|
37,self_attn.k_proj,0.0004868450,0.01000,0.402
|
|
37,self_attn.v_proj,0.0021354857,0.01000,0.392
|
|
37,self_attn.q_proj,0.0023403343,0.01000,0.447
|
|
37,self_attn.o_proj,0.0002652918,0.01000,0.402
|
|
37,mlp.up_proj,0.0034354774,0.01000,0.466
|
|
37,mlp.gate_proj,0.0025300565,0.01000,0.521
|
|
37,mlp.down_proj,0.0028967630,0.01000,1.635
|
|
38,self_attn.k_proj,0.0004139596,0.01000,0.444
|
|
38,self_attn.v_proj,0.0033904954,0.01000,0.430
|
|
38,self_attn.q_proj,0.0023008679,0.01000,0.399
|
|
38,self_attn.o_proj,0.0004398844,0.01000,0.407
|
|
38,mlp.up_proj,0.0035764058,0.01000,0.478
|
|
38,mlp.gate_proj,0.0028133283,0.01000,0.466
|
|
38,mlp.down_proj,0.0049499186,0.01000,1.550
|
|
39,self_attn.k_proj,0.0004715901,0.01000,0.422
|
|
39,self_attn.v_proj,0.0029149932,0.01000,0.401
|
|
39,self_attn.q_proj,0.0029376479,0.01000,0.404
|
|
39,self_attn.o_proj,0.0013820319,0.01000,0.416
|
|
39,mlp.up_proj,0.0037004514,0.01000,0.523
|
|
39,mlp.gate_proj,0.0030838024,0.01000,0.521
|
|
39,mlp.down_proj,0.0127753246,0.01000,1.587
|
|
|