Abdohaaland's picture
Upload folder using huggingface_hub
217a67f verified
layer,module,loss,samples,damp,time
0,self_attn.q_proj,0.0000001153,0.05000,5.049
0,self_attn.k_proj,0.0000000314,0.05000,5.046
0,self_attn.v_proj,0.0000000264,0.05000,4.904
0,self_attn.o_proj,0.0000000573,0.05000,1.352
0,mlp.up_proj,0.0000071981,0.05000,2.367
0,mlp.gate_proj,0.0000083000,0.05000,2.394
0,mlp.down_proj,0.0000004238,0.05000,3.637
1,self_attn.v_proj,0.0000000536,0.05000,4.725
1,self_attn.q_proj,0.0000002012,0.05000,4.760
1,self_attn.k_proj,0.0000000553,0.05000,4.801
1,self_attn.o_proj,0.0000000548,0.05000,1.421
1,mlp.gate_proj,0.0002712732,0.05000,2.455
1,mlp.up_proj,0.0001199079,0.05000,2.521
1,mlp.down_proj,0.0000006759,0.05000,3.629
2,self_attn.q_proj,0.0000005233,0.05000,4.418
2,self_attn.k_proj,0.0000001536,0.05000,4.511
2,self_attn.v_proj,0.0000001353,0.05000,4.519
2,self_attn.o_proj,0.0000000725,0.05000,1.336
2,mlp.up_proj,0.0003696989,0.05000,1.855
2,mlp.gate_proj,0.0004766503,0.05000,1.883
2,mlp.down_proj,0.0000005235,0.05000,3.528
3,self_attn.k_proj,0.0000002669,0.05000,4.047
3,self_attn.q_proj,0.0000010149,0.05000,4.106
3,self_attn.v_proj,0.0000002467,0.05000,4.133
3,self_attn.o_proj,0.0000001659,0.05000,1.379
3,mlp.up_proj,0.0001861880,0.05000,1.868
3,mlp.gate_proj,0.0003944701,0.05000,1.886
3,mlp.down_proj,0.0000007631,0.05000,3.533
4,self_attn.v_proj,0.0000004171,0.05000,4.027
4,self_attn.k_proj,0.0000004655,0.05000,4.080
4,self_attn.q_proj,0.0000016490,0.05000,4.107
4,self_attn.o_proj,0.0000001532,0.05000,1.332
4,mlp.gate_proj,0.0003126145,0.05000,1.964
4,mlp.up_proj,0.0001026281,0.05000,1.966
4,mlp.down_proj,0.0000008822,0.05000,3.620
5,self_attn.q_proj,0.0000017115,0.05000,3.967
5,self_attn.k_proj,0.0000004532,0.05000,4.103
5,self_attn.v_proj,0.0000004463,0.05000,4.119
5,self_attn.o_proj,0.0000002963,0.05000,1.334
5,mlp.up_proj,0.0000387881,0.05000,1.993
5,mlp.gate_proj,0.0000990348,0.05000,2.011
5,mlp.down_proj,0.0000012321,0.05000,3.554
6,self_attn.k_proj,0.0000009316,0.05000,4.038
6,self_attn.v_proj,0.0000009560,0.05000,4.090
6,self_attn.q_proj,0.0000037971,0.05000,4.108
6,self_attn.o_proj,0.0000006054,0.05000,1.350
6,mlp.gate_proj,0.0001076263,0.05000,1.906
6,mlp.up_proj,0.0000480459,0.05000,1.908
6,mlp.down_proj,0.0000274052,0.05000,3.514
7,self_attn.v_proj,0.0000018685,0.05000,3.939
7,self_attn.k_proj,0.0000022698,0.05000,4.055
7,self_attn.q_proj,0.0000079988,0.05000,4.084
7,self_attn.o_proj,0.0000008302,0.05000,1.393
7,mlp.up_proj,0.0000536222,0.05000,1.848
7,mlp.gate_proj,0.0001139974,0.05000,1.864
7,mlp.down_proj,0.0000020956,0.05000,3.518
8,self_attn.q_proj,0.0000104191,0.05000,4.057
8,self_attn.k_proj,0.0000028003,0.05000,4.107
8,self_attn.v_proj,0.0000025009,0.05000,4.129
8,self_attn.o_proj,0.0000008218,0.05000,1.354
8,mlp.gate_proj,0.0000628367,0.05000,1.945
8,mlp.up_proj,0.0000496221,0.05000,1.967
8,mlp.down_proj,0.0000027113,0.05000,3.601
9,self_attn.q_proj,0.0000145289,0.05000,4.056
9,self_attn.v_proj,0.0000031755,0.05000,4.076
9,self_attn.k_proj,0.0000043873,0.05000,4.110
9,self_attn.o_proj,0.0000013182,0.05000,1.376
9,mlp.gate_proj,0.0001058348,0.05000,1.904
9,mlp.up_proj,0.0000626613,0.05000,1.927
9,mlp.down_proj,0.0000029618,0.05000,3.480
10,self_attn.v_proj,0.0000047305,0.05000,4.056
10,self_attn.k_proj,0.0000051334,0.05000,4.069
10,self_attn.q_proj,0.0000193386,0.05000,4.110
10,self_attn.o_proj,0.0000011543,0.05000,1.348
10,mlp.up_proj,0.0000511433,0.05000,1.859
10,mlp.gate_proj,0.0000775700,0.05000,1.882
10,mlp.down_proj,0.0000024109,0.05000,3.514
11,self_attn.q_proj,0.0000095244,0.05000,4.037
11,self_attn.k_proj,0.0000027274,0.05000,4.079
11,self_attn.v_proj,0.0000021945,0.05000,4.111
11,self_attn.o_proj,0.0000007554,0.05000,1.336
11,mlp.gate_proj,0.0000644956,0.05000,1.840
11,mlp.up_proj,0.0000488933,0.05000,1.859
11,mlp.down_proj,0.0000025150,0.05000,3.515
12,self_attn.v_proj,0.0000031391,0.05000,3.958
12,self_attn.k_proj,0.0000035763,0.05000,4.089
12,self_attn.q_proj,0.0000127543,0.05000,4.105
12,self_attn.o_proj,0.0000016265,0.05000,1.372
12,mlp.gate_proj,0.0000602618,0.05000,1.913
12,mlp.up_proj,0.0000506296,0.05000,1.939
12,mlp.down_proj,0.0000027434,0.05000,3.544
13,self_attn.q_proj,0.0000102629,0.05000,3.938
13,self_attn.k_proj,0.0000028292,0.05000,4.006
13,self_attn.v_proj,0.0000023032,0.05000,4.034
13,self_attn.o_proj,0.0000012708,0.05000,1.333
13,mlp.gate_proj,0.0000599222,0.05000,1.867
13,mlp.up_proj,0.0000557118,0.05000,1.906
13,mlp.down_proj,0.0000036174,0.05000,3.530
14,self_attn.v_proj,0.0000038112,0.05000,3.993
14,self_attn.k_proj,0.0000041284,0.05000,4.062
14,self_attn.q_proj,0.0000156123,0.05000,4.090
14,self_attn.o_proj,0.0000016136,0.05000,1.422
14,mlp.up_proj,0.0000593381,0.05000,1.871
14,mlp.gate_proj,0.0000639253,0.05000,1.898
14,mlp.down_proj,0.0000036098,0.05000,3.527
15,self_attn.q_proj,0.0000175619,0.05000,4.078
15,self_attn.k_proj,0.0000045820,0.05000,4.136
15,self_attn.v_proj,0.0000038347,0.05000,4.163
15,self_attn.o_proj,0.0000016670,0.05000,1.349
15,mlp.gate_proj,0.0000615278,0.05000,1.954
15,mlp.up_proj,0.0000598616,0.05000,1.972
15,mlp.down_proj,0.0000040227,0.05000,3.588
16,self_attn.v_proj,0.0000083706,0.05000,4.065
16,self_attn.k_proj,0.0000076236,0.05000,4.114
16,self_attn.q_proj,0.0000314463,0.05000,4.151
16,self_attn.o_proj,0.0000021270,0.05000,1.349
16,mlp.gate_proj,0.0000710921,0.05000,1.947
16,mlp.up_proj,0.0000644939,0.05000,1.966
16,mlp.down_proj,0.0000117242,0.05000,3.546
17,self_attn.k_proj,0.0000067831,0.05000,4.004
17,self_attn.q_proj,0.0000312820,0.05000,4.069
17,self_attn.v_proj,0.0000068218,0.05000,4.087
17,self_attn.o_proj,0.0000022641,0.05000,1.352
17,mlp.up_proj,0.0000593657,0.05000,1.981
17,mlp.gate_proj,0.0000645823,0.05000,1.997
17,mlp.down_proj,0.0000040265,0.05000,3.558
18,self_attn.k_proj,0.0000076790,0.05000,4.045
18,self_attn.q_proj,0.0000317306,0.05000,4.091
18,self_attn.v_proj,0.0000078188,0.05000,4.120
18,self_attn.o_proj,0.0000024979,0.05000,1.353
18,mlp.gate_proj,0.0000683760,0.05000,1.927
18,mlp.up_proj,0.0000643242,0.05000,1.943
18,mlp.down_proj,0.0000048352,0.05000,3.559
19,self_attn.v_proj,0.0000123465,0.05000,4.063
19,self_attn.q_proj,0.0000572430,0.05000,4.142
19,self_attn.k_proj,0.0000133796,0.05000,4.159
19,self_attn.o_proj,0.0000057088,0.05000,1.345
19,mlp.gate_proj,0.0000665518,0.05000,1.862
19,mlp.up_proj,0.0000632745,0.05000,1.862
19,mlp.down_proj,0.0000050105,0.05000,3.549
20,self_attn.k_proj,0.0000115854,0.05000,4.079
20,self_attn.q_proj,0.0000533304,0.05000,4.150
20,self_attn.v_proj,0.0000104790,0.05000,4.157
20,self_attn.o_proj,0.0000044671,0.05000,1.360
20,mlp.up_proj,0.0000710240,0.05000,1.917
20,mlp.gate_proj,0.0000730992,0.05000,1.937
20,mlp.down_proj,0.0000056272,0.05000,3.524
21,self_attn.k_proj,0.0000134538,0.05000,4.045
21,self_attn.v_proj,0.0000129916,0.05000,4.093
21,self_attn.q_proj,0.0000596051,0.05000,4.095
21,self_attn.o_proj,0.0000049458,0.05000,1.420
21,mlp.up_proj,0.0000742268,0.05000,1.945
21,mlp.gate_proj,0.0000773761,0.05000,1.968
21,mlp.down_proj,0.0000066130,0.05000,3.565
22,self_attn.q_proj,0.0001142109,0.05000,4.051
22,self_attn.k_proj,0.0000264698,0.05000,4.105
22,self_attn.v_proj,0.0000240881,0.05000,4.129
22,self_attn.o_proj,0.0000098465,0.05000,1.356
22,mlp.up_proj,0.0000840972,0.05000,1.850
22,mlp.gate_proj,0.0000945803,0.05000,1.859
22,mlp.down_proj,0.0000115707,0.05000,3.587
23,self_attn.q_proj,0.0001076663,0.05000,4.072
23,self_attn.v_proj,0.0000245380,0.05000,4.140
23,self_attn.k_proj,0.0000248362,0.05000,4.156
23,self_attn.o_proj,0.0000091614,0.05000,1.375
23,mlp.up_proj,0.0001068179,0.05000,1.993
23,mlp.gate_proj,0.0001266348,0.05000,2.014
23,mlp.down_proj,0.0000158050,0.05000,3.526
24,self_attn.v_proj,0.0000519260,0.05000,4.015
24,self_attn.q_proj,0.0001842994,0.05000,4.079
24,self_attn.k_proj,0.0000439712,0.05000,4.112
24,self_attn.o_proj,0.0000171693,0.05000,1.334
24,mlp.gate_proj,0.0001448973,0.05000,1.835
24,mlp.up_proj,0.0001207576,0.05000,1.858
24,mlp.down_proj,0.0000206206,0.05000,3.523
25,self_attn.q_proj,0.0001284047,0.05000,3.922
25,self_attn.k_proj,0.0000323777,0.05000,4.020
25,self_attn.v_proj,0.0000331042,0.05000,4.052
25,self_attn.o_proj,0.0000070419,0.05000,1.324
25,mlp.up_proj,0.0001367776,0.05000,1.950
25,mlp.gate_proj,0.0001674050,0.05000,1.948
25,mlp.down_proj,0.0000251580,0.05000,3.574
26,self_attn.v_proj,0.0000526688,0.05000,4.112
26,self_attn.q_proj,0.0002068482,0.05000,4.155
26,self_attn.k_proj,0.0000496134,0.05000,4.170
26,self_attn.o_proj,0.0000051848,0.05000,1.350
26,mlp.gate_proj,0.0001928277,0.05000,1.937
26,mlp.up_proj,0.0001641577,0.05000,1.944
26,mlp.down_proj,0.0000290895,0.05000,3.588
27,self_attn.v_proj,0.0000689536,0.05000,4.199
27,self_attn.q_proj,0.0002489993,0.05000,4.245
27,self_attn.k_proj,0.0000588823,0.05000,4.284
27,self_attn.o_proj,0.0000074859,0.05000,1.371
27,mlp.gate_proj,0.0002114797,0.05000,1.897
27,mlp.up_proj,0.0001867100,0.05000,1.921
27,mlp.down_proj,0.0000425167,0.05000,3.522
28,self_attn.v_proj,0.0000753414,0.05000,4.083
28,self_attn.k_proj,0.0000747432,0.05000,4.138
28,self_attn.q_proj,0.0002869514,0.05000,4.171
28,self_attn.o_proj,0.0000164335,0.05000,1.460
28,mlp.gate_proj,0.0002405842,0.05000,1.938
28,mlp.up_proj,0.0002226133,0.05000,1.956
28,mlp.down_proj,0.0000668795,0.05000,3.585
29,self_attn.q_proj,0.0007534960,0.05000,3.975
29,self_attn.k_proj,0.0001845237,0.05000,4.050
29,self_attn.v_proj,0.0002504832,0.05000,4.091
29,self_attn.o_proj,0.0000320753,0.05000,1.371
29,mlp.gate_proj,0.0002801606,0.05000,1.924
29,mlp.up_proj,0.0002734993,0.05000,1.944
29,mlp.down_proj,0.0000491190,0.05000,3.589
30,self_attn.k_proj,0.0002038838,0.05000,4.011
30,self_attn.v_proj,0.0002371485,0.05000,4.099
30,self_attn.q_proj,0.0007788552,0.05000,4.139
30,self_attn.o_proj,0.0000364104,0.05000,1.333
30,mlp.up_proj,0.0002901775,0.05000,1.967
30,mlp.gate_proj,0.0002870724,0.05000,1.969
30,mlp.down_proj,0.0000643464,0.05000,3.535
31,self_attn.k_proj,0.0003022571,0.05000,4.016
31,self_attn.q_proj,0.0010716498,0.05000,4.085
31,self_attn.v_proj,0.0003914933,0.05000,4.107
31,self_attn.o_proj,0.0000348235,0.05000,1.387
31,mlp.gate_proj,0.0002821420,0.05000,1.868
31,mlp.up_proj,0.0003007668,0.05000,1.885
31,mlp.down_proj,0.0000811069,0.05000,3.580
32,self_attn.v_proj,0.0006320391,0.05000,4.059
32,self_attn.k_proj,0.0004232653,0.05000,4.098
32,self_attn.q_proj,0.0016716093,0.05000,4.124
32,self_attn.o_proj,0.0000694807,0.05000,1.362
32,mlp.up_proj,0.0003274428,0.05000,1.839
32,mlp.gate_proj,0.0002995185,0.05000,1.859
32,mlp.down_proj,0.0001096588,0.05000,3.546
33,self_attn.q_proj,0.0040160338,0.05000,4.085
33,self_attn.v_proj,0.0015813031,0.05000,4.130
33,self_attn.k_proj,0.0008651940,0.05000,4.169
33,self_attn.o_proj,0.0001325059,0.05000,1.345
33,mlp.up_proj,0.0003519209,0.05000,1.878
33,mlp.gate_proj,0.0003186744,0.05000,1.903
33,mlp.down_proj,0.0001409312,0.05000,3.574
34,self_attn.v_proj,0.0009027103,0.05000,4.124
34,self_attn.q_proj,0.0026703046,0.05000,4.195
34,self_attn.k_proj,0.0005990280,0.05000,4.223
34,self_attn.o_proj,0.0001643766,0.05000,1.372
34,mlp.up_proj,0.0003952665,0.05000,1.916
34,mlp.gate_proj,0.0003853188,0.05000,1.932
34,mlp.down_proj,0.0002338718,0.05000,3.562
35,self_attn.v_proj,0.0003357541,0.05000,3.973
35,self_attn.q_proj,0.0010748606,0.05000,4.043
35,self_attn.k_proj,0.0002876785,0.05000,4.094
35,self_attn.o_proj,0.0001355600,0.05000,1.385
35,mlp.gate_proj,0.0008457840,0.05000,1.880
35,mlp.up_proj,0.0008631636,0.05000,1.895
35,mlp.down_proj,0.0006807271,0.05000,3.547