ramblingpolymath's picture
Upload folder using huggingface_hub
ffb8967 verified
layer,module,loss,samples,damp,time
0,self_attn.k_proj,0.0000000081,0.01000,1.085
0,self_attn.v_proj,0.0000000072,0.01000,1.139
0,self_attn.q_proj,0.0000000298,0.01000,1.141
0,self_attn.o_proj,0.0000000182,0.01000,0.474
0,mlp.up_proj,0.0000015007,0.01000,0.528
0,mlp.gate_proj,0.0000016982,0.01000,0.532
0,mlp.down_proj,0.0000001300,0.01000,1.276
1,self_attn.k_proj,0.0000000101,0.01000,1.294
1,self_attn.q_proj,0.0000000377,0.01000,1.381
1,self_attn.v_proj,0.0000000109,0.01000,1.388
1,self_attn.o_proj,0.0000000241,0.01000,0.472
1,mlp.up_proj,0.0000421766,0.01000,0.505
1,mlp.gate_proj,0.0000908424,0.01000,0.509
1,mlp.down_proj,0.0000011421,0.01000,1.296
2,self_attn.k_proj,0.0000000232,0.01000,1.273
2,self_attn.v_proj,0.0000000252,0.01000,1.376
2,self_attn.q_proj,0.0000000853,0.01000,1.382
2,self_attn.o_proj,0.0000000194,0.01000,0.477
2,mlp.up_proj,0.0001395170,0.01000,0.505
2,mlp.gate_proj,0.0001838406,0.01000,0.507
2,mlp.down_proj,0.0000000696,0.01000,1.281
3,self_attn.k_proj,0.0000000418,0.01000,1.309
3,self_attn.q_proj,0.0000001673,0.01000,1.361
3,self_attn.v_proj,0.0000000486,0.01000,1.363
3,self_attn.o_proj,0.0000000222,0.01000,0.473
3,mlp.up_proj,0.0000797915,0.01000,0.518
3,mlp.gate_proj,0.0001503195,0.01000,0.522
3,mlp.down_proj,0.0000001083,0.01000,1.339
4,self_attn.k_proj,0.0000000701,0.01000,1.302
4,self_attn.v_proj,0.0000000774,0.01000,1.368
4,self_attn.q_proj,0.0000002717,0.01000,1.372
4,self_attn.o_proj,0.0000000324,0.01000,0.482
4,mlp.up_proj,0.0000420538,0.01000,0.500
4,mlp.gate_proj,0.0000980137,0.01000,0.501
4,mlp.down_proj,0.0000002608,0.01000,1.294
5,self_attn.k_proj,0.0000000724,0.01000,1.278
5,self_attn.v_proj,0.0000000838,0.01000,1.346
5,self_attn.q_proj,0.0000002967,0.01000,1.348
5,self_attn.o_proj,0.0000000602,0.01000,0.475
5,mlp.gate_proj,0.0000244205,0.01000,0.503
5,mlp.up_proj,0.0000132466,0.01000,0.503
5,mlp.down_proj,0.0000003615,0.01000,1.302
6,self_attn.k_proj,0.0000001312,0.01000,1.256
6,self_attn.q_proj,0.0000005748,0.01000,1.313
6,self_attn.v_proj,0.0000001556,0.01000,1.320
6,self_attn.o_proj,0.0000002136,0.01000,0.475
6,mlp.up_proj,0.0000460187,0.01000,0.515
6,mlp.gate_proj,0.0000727634,0.01000,0.521
6,mlp.down_proj,0.0061423381,0.01000,1.270
7,self_attn.k_proj,0.0000003686,0.01000,1.410
7,self_attn.v_proj,0.0000004743,0.01000,1.464
7,self_attn.q_proj,0.0000014566,0.01000,1.471
7,self_attn.o_proj,0.0000001378,0.01000,0.480
7,mlp.up_proj,0.0000221474,0.01000,0.504
7,mlp.gate_proj,0.0000330759,0.01000,0.507
7,mlp.down_proj,0.0000006566,0.01000,1.303
8,self_attn.k_proj,0.0000005695,0.01000,1.299
8,self_attn.v_proj,0.0000007086,0.01000,1.366
8,self_attn.q_proj,0.0000022442,0.01000,1.371
8,self_attn.o_proj,0.0000002123,0.01000,0.475
8,mlp.gate_proj,0.0000162446,0.01000,0.511
8,mlp.up_proj,0.0000138262,0.01000,0.512
8,mlp.down_proj,0.0000010385,0.01000,1.279
9,self_attn.k_proj,0.0000005919,0.01000,1.223
9,self_attn.q_proj,0.0000022763,0.01000,1.303
9,self_attn.v_proj,0.0000007881,0.01000,1.309
9,self_attn.o_proj,0.0000001922,0.01000,0.474
9,mlp.up_proj,0.0000176753,0.01000,0.500
9,mlp.gate_proj,0.0000271661,0.01000,0.501
9,mlp.down_proj,0.0000009237,0.01000,1.317
10,self_attn.k_proj,0.0000010477,0.01000,1.301
10,self_attn.q_proj,0.0000040611,0.01000,1.373
10,self_attn.v_proj,0.0000013349,0.01000,1.378
10,self_attn.o_proj,0.0000004386,0.01000,0.477
10,mlp.gate_proj,0.0000211780,0.01000,0.539
10,mlp.up_proj,0.0000149951,0.01000,0.542
10,mlp.down_proj,0.0000010175,0.01000,1.288
11,self_attn.k_proj,0.0000004850,0.01000,1.266
11,self_attn.v_proj,0.0000006466,0.01000,1.328
11,self_attn.q_proj,0.0000019295,0.01000,1.335
11,self_attn.o_proj,0.0000002865,0.01000,0.482
11,mlp.up_proj,0.0000130007,0.01000,0.508
11,mlp.gate_proj,0.0000161569,0.01000,0.510
11,mlp.down_proj,0.0000008605,0.01000,1.301
12,self_attn.k_proj,0.0000005923,0.01000,1.277
12,self_attn.v_proj,0.0000007873,0.01000,1.344
12,self_attn.q_proj,0.0000024283,0.01000,1.345
12,self_attn.o_proj,0.0000004031,0.01000,0.477
12,mlp.up_proj,0.0000132932,0.01000,0.504
12,mlp.gate_proj,0.0000148126,0.01000,0.505
12,mlp.down_proj,0.0000018859,0.01000,1.273
13,self_attn.k_proj,0.0000005838,0.01000,1.329
13,self_attn.v_proj,0.0000006637,0.01000,1.399
13,self_attn.q_proj,0.0000022540,0.01000,1.404
13,self_attn.o_proj,0.0000004204,0.01000,0.472
13,mlp.gate_proj,0.0000148458,0.01000,0.505
13,mlp.up_proj,0.0000145850,0.01000,0.507
13,mlp.down_proj,0.0000010664,0.01000,1.312
14,self_attn.k_proj,0.0000011471,0.01000,1.241
14,self_attn.q_proj,0.0000045900,0.01000,1.314
14,self_attn.v_proj,0.0000013485,0.01000,1.317
14,self_attn.o_proj,0.0000005681,0.01000,0.477
14,mlp.gate_proj,0.0000192607,0.01000,0.546
14,mlp.up_proj,0.0000185098,0.01000,0.550
14,mlp.down_proj,0.0000026846,0.01000,1.282
15,self_attn.k_proj,0.0000012532,0.01000,1.228
15,self_attn.v_proj,0.0000014174,0.01000,1.299
15,self_attn.q_proj,0.0000050999,0.01000,1.302
15,self_attn.o_proj,0.0000008368,0.01000,0.474
15,mlp.gate_proj,0.0000203562,0.01000,0.501
15,mlp.up_proj,0.0000204011,0.01000,0.501
15,mlp.down_proj,0.0000012258,0.01000,1.299
16,self_attn.k_proj,0.0000017525,0.01000,1.244
16,self_attn.v_proj,0.0000022914,0.01000,1.303
16,self_attn.q_proj,0.0000075159,0.01000,1.313
16,self_attn.o_proj,0.0000006825,0.01000,0.474
16,mlp.gate_proj,0.0000191996,0.01000,0.505
16,mlp.up_proj,0.0000183455,0.01000,0.504
16,mlp.down_proj,0.0000039744,0.01000,1.277
17,self_attn.k_proj,0.0000015515,0.01000,1.283
17,self_attn.q_proj,0.0000069198,0.01000,1.362
17,self_attn.v_proj,0.0000018403,0.01000,1.368
17,self_attn.o_proj,0.0000004020,0.01000,0.489
17,mlp.gate_proj,0.0000176530,0.01000,0.587
17,mlp.up_proj,0.0000170819,0.01000,0.591
17,mlp.down_proj,0.0000007446,0.01000,1.336
18,self_attn.k_proj,0.0000021546,0.01000,1.175
18,self_attn.q_proj,0.0000092629,0.01000,1.240
18,self_attn.v_proj,0.0000027618,0.01000,1.242
18,self_attn.o_proj,0.0000004712,0.01000,0.476
18,mlp.gate_proj,0.0000180721,0.01000,0.505
18,mlp.up_proj,0.0000179228,0.01000,0.506
18,mlp.down_proj,0.0000027112,0.01000,1.285
19,self_attn.k_proj,0.0000033295,0.01000,1.355
19,self_attn.q_proj,0.0000157591,0.01000,1.420
19,self_attn.v_proj,0.0000043254,0.01000,1.424
19,self_attn.o_proj,0.0000005817,0.01000,0.477
19,mlp.up_proj,0.0000173141,0.01000,0.498
19,mlp.gate_proj,0.0000172161,0.01000,0.503
19,mlp.down_proj,0.0000012885,0.01000,1.290
20,self_attn.k_proj,0.0000029203,0.01000,1.259
20,self_attn.q_proj,0.0000147554,0.01000,1.330
20,self_attn.v_proj,0.0000037685,0.01000,1.332
20,self_attn.o_proj,0.0000005304,0.01000,0.466
20,mlp.up_proj,0.0000180820,0.01000,0.531
20,mlp.gate_proj,0.0000174652,0.01000,0.539
20,mlp.down_proj,0.0000010553,0.01000,1.288
21,self_attn.k_proj,0.0000044806,0.01000,1.325
21,self_attn.q_proj,0.0000222443,0.01000,1.415
21,self_attn.v_proj,0.0000057858,0.01000,1.419
21,self_attn.o_proj,0.0000007092,0.01000,0.480
21,mlp.gate_proj,0.0000200361,0.01000,0.510
21,mlp.up_proj,0.0000210307,0.01000,0.510
21,mlp.down_proj,0.0000015563,0.01000,1.322
22,self_attn.k_proj,0.0000072332,0.01000,1.285
22,self_attn.q_proj,0.0000346988,0.01000,1.374
22,self_attn.v_proj,0.0000096312,0.01000,1.383
22,self_attn.o_proj,0.0000009259,0.01000,0.481
22,mlp.gate_proj,0.0000242100,0.01000,0.514
22,mlp.up_proj,0.0000250956,0.01000,0.517
22,mlp.down_proj,0.0000028355,0.01000,1.300
23,self_attn.k_proj,0.0000067722,0.01000,1.339
23,self_attn.v_proj,0.0000088958,0.01000,1.411
23,self_attn.q_proj,0.0000359667,0.01000,1.414
23,self_attn.o_proj,0.0000011439,0.01000,0.474
23,mlp.up_proj,0.0000285567,0.01000,0.553
23,mlp.gate_proj,0.0000291968,0.01000,0.553
23,mlp.down_proj,0.0000035951,0.01000,1.333
24,self_attn.k_proj,0.0000096078,0.01000,1.313
24,self_attn.q_proj,0.0000521621,0.01000,1.385
24,self_attn.v_proj,0.0000127284,0.01000,1.388
24,self_attn.o_proj,0.0000012113,0.01000,0.482
24,mlp.up_proj,0.0000323534,0.01000,0.524
24,mlp.gate_proj,0.0000342747,0.01000,0.528
24,mlp.down_proj,0.0000046040,0.01000,1.282
25,self_attn.k_proj,0.0000088128,0.01000,1.378
25,self_attn.q_proj,0.0000400565,0.01000,1.457
25,self_attn.v_proj,0.0000114145,0.01000,1.463
25,self_attn.o_proj,0.0000006758,0.01000,0.485
25,mlp.up_proj,0.0000371785,0.01000,0.513
25,mlp.gate_proj,0.0000400409,0.01000,0.516
25,mlp.down_proj,0.0000052616,0.01000,1.327
26,self_attn.k_proj,0.0000113154,0.01000,1.336
26,self_attn.v_proj,0.0000149217,0.01000,1.407
26,self_attn.q_proj,0.0000558885,0.01000,1.412
26,self_attn.o_proj,0.0000007748,0.01000,0.480
26,mlp.gate_proj,0.0000467025,0.01000,0.510
26,mlp.up_proj,0.0000444803,0.01000,0.510
26,mlp.down_proj,0.0000055660,0.01000,1.301
27,self_attn.k_proj,0.0000138321,0.01000,1.244
27,self_attn.q_proj,0.0000710704,0.01000,1.311
27,self_attn.v_proj,0.0000190896,0.01000,1.314
27,self_attn.o_proj,0.0000010194,0.01000,0.483
27,mlp.up_proj,0.0000496334,0.01000,0.536
27,mlp.gate_proj,0.0000504245,0.01000,0.540
27,mlp.down_proj,0.0000083369,0.01000,1.327
28,self_attn.k_proj,0.0000168529,0.01000,1.339
28,self_attn.v_proj,0.0000234338,0.01000,1.420
28,self_attn.q_proj,0.0000835909,0.01000,1.419
28,self_attn.o_proj,0.0000015521,0.01000,0.476
28,mlp.up_proj,0.0000584004,0.01000,0.510
28,mlp.gate_proj,0.0000570477,0.01000,0.516
28,mlp.down_proj,0.0000118570,0.01000,1.301
29,self_attn.k_proj,0.0000240786,0.01000,1.390
29,self_attn.v_proj,0.0000324057,0.01000,1.459
29,self_attn.q_proj,0.0001361566,0.01000,1.465
29,self_attn.o_proj,0.0000016135,0.01000,0.481
29,mlp.up_proj,0.0000688286,0.01000,0.527
29,mlp.gate_proj,0.0000649982,0.01000,0.534
29,mlp.down_proj,0.0000111695,0.01000,1.342
30,self_attn.k_proj,0.0000363515,0.01000,1.292
30,self_attn.q_proj,0.0001620006,0.01000,1.357
30,self_attn.v_proj,0.0000510523,0.01000,1.359
30,self_attn.o_proj,0.0000039036,0.01000,0.477
30,mlp.up_proj,0.0000738163,0.01000,0.556
30,mlp.gate_proj,0.0000673011,0.01000,0.567
30,mlp.down_proj,0.0000149511,0.01000,1.299
31,self_attn.k_proj,0.0000414222,0.01000,1.407
31,self_attn.q_proj,0.0002041749,0.01000,1.467
31,self_attn.v_proj,0.0000668156,0.01000,1.470
31,self_attn.o_proj,0.0000028319,0.01000,0.483
31,mlp.gate_proj,0.0000658788,0.01000,0.517
31,mlp.up_proj,0.0000759223,0.01000,0.519
31,mlp.down_proj,0.0000170370,0.01000,1.328
32,self_attn.k_proj,0.0000618467,0.01000,1.325
32,self_attn.v_proj,0.0001005293,0.01000,1.400
32,self_attn.q_proj,0.0002950472,0.01000,1.405
32,self_attn.o_proj,0.0000056080,0.01000,0.489
32,mlp.up_proj,0.0000818480,0.01000,0.529
32,mlp.gate_proj,0.0000691801,0.01000,0.529
32,mlp.down_proj,0.0000200854,0.01000,1.302
33,self_attn.k_proj,0.0000819142,0.01000,1.282
33,self_attn.v_proj,0.0001815281,0.01000,1.354
33,self_attn.q_proj,0.0004891983,0.01000,1.357
33,self_attn.o_proj,0.0000050695,0.01000,0.483
33,mlp.up_proj,0.0000987050,0.01000,0.509
33,mlp.gate_proj,0.0000820996,0.01000,0.511
33,mlp.down_proj,0.0000356632,0.01000,1.337
34,self_attn.k_proj,0.0001026835,0.01000,1.424
34,self_attn.q_proj,0.0005004918,0.01000,1.484
34,self_attn.v_proj,0.0002004345,0.01000,1.488
34,self_attn.o_proj,0.0000194029,0.01000,0.487
34,mlp.up_proj,0.0001359247,0.01000,0.509
34,mlp.gate_proj,0.0001231596,0.01000,0.511
34,mlp.down_proj,0.0001338438,0.01000,1.293
35,self_attn.k_proj,0.0000897737,0.01000,1.337
35,self_attn.q_proj,0.0003517778,0.01000,1.400
35,self_attn.v_proj,0.0001144365,0.01000,1.405
35,self_attn.o_proj,0.0000256265,0.01000,0.481
35,mlp.gate_proj,0.0009960180,0.01000,0.504
35,mlp.up_proj,0.0012875651,0.01000,0.506
35,mlp.down_proj,0.0012820768,0.01000,1.317