File size: 11,414 Bytes

c6ad2b1

layer,module,loss,samples,damp,time
0,self_attn.k_proj,0.0000005085,0.05000,3.240
0,self_attn.v_proj,0.0000000616,0.05000,3.331
0,self_attn.q_proj,0.0000024387,0.05000,3.344
0,self_attn.o_proj,0.0000002079,0.05000,0.973
0,mlp.gate_proj,0.0000028578,0.05000,1.696
0,mlp.up_proj,0.0000022065,0.05000,1.714
0,mlp.down_proj,0.0000005557,0.05000,4.253
1,self_attn.v_proj,0.0000000238,0.05000,3.020
1,self_attn.k_proj,0.0000001244,0.05000,3.038
1,self_attn.q_proj,0.0000004455,0.05000,3.103
1,self_attn.o_proj,0.0000000834,0.05000,0.920
1,mlp.up_proj,0.0000957824,0.05000,1.378
1,mlp.gate_proj,0.0001252921,0.05000,1.436
1,mlp.down_proj,0.0000000248,0.05000,4.110
2,self_attn.v_proj,0.0000000376,0.05000,3.022
2,self_attn.q_proj,0.0000006681,0.05000,3.177
2,self_attn.k_proj,0.0000001465,0.05000,3.203
2,self_attn.o_proj,0.0000000641,0.05000,1.082
2,mlp.gate_proj,0.0000304530,0.05000,1.268
2,mlp.up_proj,0.0000303662,0.05000,1.334
2,mlp.down_proj,0.0000120081,0.05000,4.371
3,self_attn.k_proj,0.0000006040,0.05000,3.001
3,self_attn.q_proj,0.0000028064,0.05000,3.063
3,self_attn.v_proj,0.0000001111,0.05000,3.032
3,self_attn.o_proj,0.0000000855,0.05000,0.928
3,mlp.up_proj,0.0000360368,0.05000,1.576
3,mlp.gate_proj,0.0000382443,0.05000,1.606
3,mlp.down_proj,0.0000003860,0.05000,4.249
4,self_attn.q_proj,0.0000022345,0.05000,3.843
4,self_attn.k_proj,0.0000004367,0.05000,3.907
4,self_attn.v_proj,0.0000001491,0.05000,3.939
4,self_attn.o_proj,0.0000001581,0.05000,0.863
4,mlp.up_proj,0.0000284978,0.05000,1.413
4,mlp.gate_proj,0.0000365522,0.05000,1.449
4,mlp.down_proj,0.0000003562,0.05000,4.254
5,self_attn.v_proj,0.0000004122,0.05000,2.760
5,self_attn.k_proj,0.0000008967,0.05000,2.809
5,self_attn.q_proj,0.0000050007,0.05000,2.831
5,self_attn.o_proj,0.0000001279,0.05000,0.942
5,mlp.gate_proj,0.0000520914,0.05000,1.275
5,mlp.up_proj,0.0000442293,0.05000,1.282
5,mlp.down_proj,0.0000007485,0.05000,5.679
6,self_attn.k_proj,0.0000004771,0.05000,3.021
6,self_attn.v_proj,0.0000003296,0.05000,3.142
6,self_attn.q_proj,0.0000030048,0.05000,3.150
6,self_attn.o_proj,0.0000002717,0.05000,0.938
6,mlp.up_proj,0.0000590103,0.05000,1.374
6,mlp.gate_proj,0.0000687621,0.05000,1.421
6,mlp.down_proj,0.0000011120,0.05000,4.512
7,self_attn.q_proj,0.0000041045,0.05000,2.921
7,self_attn.k_proj,0.0000006536,0.05000,2.968
7,self_attn.v_proj,0.0000004090,0.05000,2.974
7,self_attn.o_proj,0.0000002103,0.05000,0.866
7,mlp.gate_proj,0.0000712782,0.05000,1.340
7,mlp.up_proj,0.0000586724,0.05000,1.357
7,mlp.down_proj,0.0000014121,0.05000,4.585
8,self_attn.k_proj,0.0000006371,0.05000,3.163
8,self_attn.v_proj,0.0000004388,0.05000,3.356
8,self_attn.q_proj,0.0000041995,0.05000,3.369
8,self_attn.o_proj,0.0000002558,0.05000,0.978
8,mlp.up_proj,0.0000598540,0.05000,1.155
8,mlp.gate_proj,0.0000840462,0.05000,1.206
8,mlp.down_proj,0.0000018146,0.05000,4.143
9,self_attn.v_proj,0.0000003784,0.05000,2.766
9,self_attn.q_proj,0.0000037226,0.05000,2.783
9,self_attn.k_proj,0.0000006510,0.05000,2.768
9,self_attn.o_proj,0.0000004260,0.05000,0.930
9,mlp.up_proj,0.0000524589,0.05000,1.571
9,mlp.gate_proj,0.0000783262,0.05000,1.598
9,mlp.down_proj,0.0000024518,0.05000,5.082
10,self_attn.q_proj,0.0000052897,0.05000,2.622
10,self_attn.k_proj,0.0000007815,0.05000,2.717
10,self_attn.v_proj,0.0000006292,0.05000,2.743
10,self_attn.o_proj,0.0000003979,0.05000,1.276
10,mlp.gate_proj,0.0000586628,0.05000,1.290
10,mlp.up_proj,0.0000382535,0.05000,1.371
10,mlp.down_proj,0.0000026338,0.05000,4.219
11,self_attn.k_proj,0.0000005685,0.05000,2.966
11,self_attn.v_proj,0.0000005079,0.05000,2.980
11,self_attn.q_proj,0.0000038576,0.05000,2.991
11,self_attn.o_proj,0.0000005243,0.05000,1.047
11,mlp.up_proj,0.0000218057,0.05000,1.301
11,mlp.gate_proj,0.0000305343,0.05000,1.333
11,mlp.down_proj,0.0000027184,0.05000,4.218
12,self_attn.q_proj,0.0000040775,0.05000,2.488
12,self_attn.k_proj,0.0000006225,0.05000,2.505
12,self_attn.v_proj,0.0000004578,0.05000,2.521
12,self_attn.o_proj,0.0000004894,0.05000,0.662
12,mlp.gate_proj,0.0000336821,0.05000,1.277
12,mlp.up_proj,0.0000236778,0.05000,1.291
12,mlp.down_proj,0.0000022740,0.05000,4.415
13,self_attn.q_proj,0.0000048006,0.05000,3.943
13,self_attn.v_proj,0.0000003174,0.05000,3.952
13,self_attn.k_proj,0.0000008770,0.05000,3.995
13,self_attn.o_proj,0.0000003654,0.05000,0.734
13,mlp.gate_proj,0.0000177235,0.05000,1.341
13,mlp.up_proj,0.0000169780,0.05000,1.368
13,mlp.down_proj,0.0000021166,0.05000,4.304
14,self_attn.q_proj,0.0000036348,0.05000,2.822
14,self_attn.k_proj,0.0000005575,0.05000,2.855
14,self_attn.v_proj,0.0000003621,0.05000,2.864
14,self_attn.o_proj,0.0000007141,0.05000,1.039
14,mlp.up_proj,0.0000165291,0.05000,1.228
14,mlp.gate_proj,0.0000172143,0.05000,1.244
14,mlp.down_proj,0.0000020680,0.05000,4.865
15,self_attn.q_proj,0.0000037588,0.05000,2.502
15,self_attn.k_proj,0.0000005889,0.05000,2.505
15,self_attn.v_proj,0.0000003618,0.05000,2.549
15,self_attn.o_proj,0.0000008058,0.05000,0.835
15,mlp.up_proj,0.0000149748,0.05000,1.265
15,mlp.gate_proj,0.0000143409,0.05000,1.283
15,mlp.down_proj,0.0000018108,0.05000,4.185
16,self_attn.k_proj,0.0000006456,0.05000,2.311
16,self_attn.q_proj,0.0000038739,0.05000,2.370
16,self_attn.v_proj,0.0000003734,0.05000,2.386
16,self_attn.o_proj,0.0000005621,0.05000,0.878
16,mlp.gate_proj,0.0000152835,0.05000,1.159
16,mlp.up_proj,0.0000149767,0.05000,1.230
16,mlp.down_proj,0.0000016259,0.05000,5.022
17,self_attn.v_proj,0.0000005680,0.05000,2.477
17,self_attn.k_proj,0.0000010732,0.05000,2.463
17,self_attn.q_proj,0.0000062434,0.05000,2.502
17,self_attn.o_proj,0.0000006299,0.05000,0.774
17,mlp.up_proj,0.0000129082,0.05000,1.237
17,mlp.gate_proj,0.0000127357,0.05000,1.305
17,mlp.down_proj,0.0000018334,0.05000,4.142
18,self_attn.q_proj,0.0000035806,0.05000,2.411
18,self_attn.k_proj,0.0000005222,0.05000,2.460
18,self_attn.v_proj,0.0000003745,0.05000,2.489
18,self_attn.o_proj,0.0000006935,0.05000,1.018
18,mlp.up_proj,0.0000123926,0.05000,1.250
18,mlp.gate_proj,0.0000127913,0.05000,1.322
18,mlp.down_proj,0.0000014052,0.05000,4.173
19,self_attn.v_proj,0.0000003944,0.05000,3.463
19,self_attn.k_proj,0.0000006968,0.05000,3.586
19,self_attn.q_proj,0.0000039632,0.05000,3.608
19,self_attn.o_proj,0.0000009026,0.05000,0.818
19,mlp.gate_proj,0.0000114576,0.05000,1.112
19,mlp.up_proj,0.0000119524,0.05000,1.161
19,mlp.down_proj,0.0000014570,0.05000,4.321
20,self_attn.k_proj,0.0000007010,0.05000,2.711
20,self_attn.v_proj,0.0000007374,0.05000,2.718
20,self_attn.q_proj,0.0000060873,0.05000,2.745
20,self_attn.o_proj,0.0000007991,0.05000,0.912
20,mlp.gate_proj,0.0000118486,0.05000,1.426
20,mlp.up_proj,0.0000121822,0.05000,1.442
20,mlp.down_proj,0.0000014912,0.05000,4.249
21,self_attn.k_proj,0.0000007058,0.05000,2.460
21,self_attn.q_proj,0.0000043860,0.05000,2.542
21,self_attn.v_proj,0.0000004948,0.05000,2.576
21,self_attn.o_proj,0.0000006633,0.05000,0.855
21,mlp.up_proj,0.0000111701,0.05000,1.315
21,mlp.gate_proj,0.0000118580,0.05000,1.357
21,mlp.down_proj,0.0000010636,0.05000,4.237
22,self_attn.q_proj,0.0000050432,0.05000,2.470
22,self_attn.k_proj,0.0000007732,0.05000,2.483
22,self_attn.v_proj,0.0000006412,0.05000,2.494
22,self_attn.o_proj,0.0000006260,0.05000,0.899
22,mlp.gate_proj,0.0000112563,0.05000,1.363
22,mlp.up_proj,0.0000109995,0.05000,1.422
22,mlp.down_proj,0.0000010489,0.05000,4.304
23,self_attn.k_proj,0.0000007014,0.05000,3.266
23,self_attn.q_proj,0.0000048249,0.05000,3.317
23,self_attn.v_proj,0.0000003740,0.05000,3.373
23,self_attn.o_proj,0.0000007296,0.05000,0.939
23,mlp.gate_proj,0.0000105611,0.05000,1.372
23,mlp.up_proj,0.0000110128,0.05000,1.381
23,mlp.down_proj,0.0000012117,0.05000,4.254
24,self_attn.v_proj,0.0000004095,0.05000,2.385
24,self_attn.k_proj,0.0000004718,0.05000,2.412
24,self_attn.q_proj,0.0000034988,0.05000,2.413
24,self_attn.o_proj,0.0000007943,0.05000,0.955
24,mlp.up_proj,0.0000105276,0.05000,1.198
24,mlp.gate_proj,0.0000102487,0.05000,1.258
24,mlp.down_proj,0.0000014155,0.05000,4.128
25,self_attn.k_proj,0.0000004508,0.05000,2.690
25,self_attn.v_proj,0.0000006476,0.05000,2.825
25,self_attn.q_proj,0.0000047023,0.05000,2.832
25,self_attn.o_proj,0.0000006867,0.05000,0.852
25,mlp.up_proj,0.0000116300,0.05000,1.634
25,mlp.gate_proj,0.0000114514,0.05000,1.708
25,mlp.down_proj,0.0000016359,0.05000,4.386
26,self_attn.k_proj,0.0000004960,0.05000,3.098
26,self_attn.q_proj,0.0000038692,0.05000,3.103
26,self_attn.v_proj,0.0000006341,0.05000,3.127
26,self_attn.o_proj,0.0000009083,0.05000,0.835
26,mlp.up_proj,0.0000133298,0.05000,1.431
26,mlp.gate_proj,0.0000125842,0.05000,1.454
26,mlp.down_proj,0.0000022201,0.05000,4.144
27,self_attn.q_proj,0.0000063679,0.05000,3.371
27,self_attn.k_proj,0.0000005433,0.05000,3.429
27,self_attn.v_proj,0.0000008838,0.05000,3.431
27,self_attn.o_proj,0.0000011006,0.05000,1.076
27,mlp.gate_proj,0.0000139996,0.05000,1.350
27,mlp.up_proj,0.0000141233,0.05000,1.353
27,mlp.down_proj,0.0000025846,0.05000,4.317
28,self_attn.v_proj,0.0000008987,0.05000,4.054
28,self_attn.k_proj,0.0000005489,0.05000,4.102
28,self_attn.q_proj,0.0000051061,0.05000,4.118
28,self_attn.o_proj,0.0000011722,0.05000,1.007
28,mlp.up_proj,0.0000153893,0.05000,1.213
28,mlp.gate_proj,0.0000153670,0.05000,1.274
28,mlp.down_proj,0.0000028280,0.05000,4.456
29,self_attn.v_proj,0.0000007629,0.05000,3.439
29,self_attn.k_proj,0.0000005342,0.05000,3.521
29,self_attn.q_proj,0.0000046298,0.05000,3.534
29,self_attn.o_proj,0.0000009201,0.05000,0.812
29,mlp.gate_proj,0.0000180554,0.05000,1.402
29,mlp.up_proj,0.0000184227,0.05000,1.406
29,mlp.down_proj,0.0000038424,0.05000,4.300
30,self_attn.v_proj,0.0000020394,0.05000,3.110
30,self_attn.k_proj,0.0000006122,0.05000,3.143
30,self_attn.q_proj,0.0000070738,0.05000,3.167
30,self_attn.o_proj,0.0000011594,0.05000,1.005
30,mlp.gate_proj,0.0000228827,0.05000,1.470
30,mlp.up_proj,0.0000249021,0.05000,1.495
30,mlp.down_proj,0.0000158357,0.05000,4.416
31,self_attn.q_proj,0.0000093229,0.05000,2.831
31,self_attn.k_proj,0.0000009378,0.05000,2.882
31,self_attn.v_proj,0.0000024443,0.05000,2.891
31,self_attn.o_proj,0.0000030832,0.05000,0.684
31,mlp.gate_proj,0.0000264529,0.05000,1.345
31,mlp.up_proj,0.0000300895,0.05000,1.400
31,mlp.down_proj,0.0000134932,0.05000,4.460
32,self_attn.q_proj,0.0000164472,0.05000,3.071
32,self_attn.k_proj,0.0000014786,0.05000,3.103
32,self_attn.v_proj,0.0000073760,0.05000,3.108
32,self_attn.o_proj,0.0000065663,0.05000,0.774
32,mlp.up_proj,0.0000336128,0.05000,1.444
32,mlp.gate_proj,0.0000295613,0.05000,1.467
32,mlp.down_proj,0.0000133721,0.05000,4.501
33,self_attn.v_proj,0.0000151532,0.05000,2.874
33,self_attn.q_proj,0.0000151740,0.05000,2.961
33,self_attn.k_proj,0.0000012399,0.05000,2.964
33,self_attn.o_proj,0.0000136846,0.05000,0.722
33,mlp.gate_proj,0.0000268336,0.05000,1.449
33,mlp.up_proj,0.0000334955,0.05000,1.466
33,mlp.down_proj,0.0000303056,0.05000,4.839
34,self_attn.q_proj,0.0000051019,0.05000,3.116
34,self_attn.k_proj,0.0000005239,0.05000,3.161
34,self_attn.v_proj,0.0000017698,0.05000,3.193
34,self_attn.o_proj,0.0000030042,0.05000,0.749
34,mlp.up_proj,0.0000342593,0.05000,1.270
34,mlp.gate_proj,0.0000294181,0.05000,1.299
34,mlp.down_proj,0.0000177885,0.05000,4.654
35,self_attn.v_proj,0.0000014588,0.05000,3.398
35,self_attn.q_proj,0.0000045896,0.05000,3.432
35,self_attn.k_proj,0.0000004904,0.05000,3.471
35,self_attn.o_proj,0.0000032275,0.05000,0.855
35,mlp.gate_proj,0.0000420900,0.05000,1.395
35,mlp.up_proj,0.0000445427,0.05000,1.458
35,mlp.down_proj,0.0000312161,0.05000,4.283