Hugging Face's logo

HID-APP
/
codebert-rd-keystrokes-v3-1

Safetensors
roberta
Model card Files
xet
 Community
NourFakih commited on
Commit
1fbabad
·
verified ·
1 Parent(s): 78272d2

Update training logs at step 3500

Browse files
Files changed (1) hide show
  1. training_logs.jsonl +21 -129
training_logs.jsonl CHANGED
@@ -1,129 +1,21 @@
1
- {"step": 1, "epoch": 0.0006309148264984228, "time": "2026-07-08 22:01:06", "logs": {"loss": 0.6932622194290161, "grad_norm": 2.571892738342285, "learning_rate": 0.0, "epoch": 0.0006309148264984228}}
2
- {"step": 25, "epoch": 0.015772870662460567, "time": "2026-07-08 22:01:25", "logs": {"loss": 0.7295282681783041, "grad_norm": 2.436145544052124, "learning_rate": 7.570977917981073e-07, "epoch": 0.015772870662460567}}
3
- {"step": 50, "epoch": 0.031545741324921134, "time": "2026-07-08 22:01:45", "logs": {"loss": 0.7090779113769531, "grad_norm": 5.32429838180542, "learning_rate": 1.545741324921136e-06, "epoch": 0.031545741324921134}}
4
- {"step": 75, "epoch": 0.0473186119873817, "time": "2026-07-08 22:02:06", "logs": {"loss": 0.6694438171386718, "grad_norm": 5.425052165985107, "learning_rate": 2.3343848580441643e-06, "epoch": 0.0473186119873817}}
5
- {"step": 100, "epoch": 0.06309148264984227, "time": "2026-07-08 22:02:28", "logs": {"loss": 0.6196812057495117, "grad_norm": 5.74971866607666, "learning_rate": 3.1230283911671926e-06, "epoch": 0.06309148264984227}}
6
- {"step": 125, "epoch": 0.07886435331230283, "time": "2026-07-08 22:02:50", "logs": {"loss": 0.43090095520019533, "grad_norm": 4.759191989898682, "learning_rate": 3.911671924290221e-06, "epoch": 0.07886435331230283}}
7
- {"step": 150, "epoch": 0.0946372239747634, "time": "2026-07-08 22:03:13", "logs": {"loss": 0.29434648513793943, "grad_norm": 5.838165283203125, "learning_rate": 4.70031545741325e-06, "epoch": 0.0946372239747634}}
8
- {"step": 175, "epoch": 0.11041009463722397, "time": "2026-07-08 22:03:36", "logs": {"loss": 0.1876709747314453, "grad_norm": 6.797428131103516, "learning_rate": 5.4889589905362786e-06, "epoch": 0.11041009463722397}}
9
- {"step": 200, "epoch": 0.12618296529968454, "time": "2026-07-08 22:03:59", "logs": {"loss": 0.20350841522216798, "grad_norm": 3.7547245025634766, "learning_rate": 6.2776025236593065e-06, "epoch": 0.12618296529968454}}
10
- {"step": 225, "epoch": 0.14195583596214512, "time": "2026-07-08 22:04:22", "logs": {"loss": 0.1446442985534668, "grad_norm": 11.165263175964355, "learning_rate": 7.066246056782335e-06, "epoch": 0.14195583596214512}}
11
- {"step": 250, "epoch": 0.15772870662460567, "time": "2026-07-08 22:04:45", "logs": {"loss": 0.17608392715454102, "grad_norm": 4.109164237976074, "learning_rate": 7.854889589905364e-06, "epoch": 0.15772870662460567}}
12
- {"step": 275, "epoch": 0.17350157728706625, "time": "2026-07-08 22:05:08", "logs": {"loss": 0.08311139106750488, "grad_norm": 0.37571197748184204, "learning_rate": 8.643533123028391e-06, "epoch": 0.17350157728706625}}
13
- {"step": 300, "epoch": 0.1892744479495268, "time": "2026-07-08 22:05:31", "logs": {"loss": 0.10653317451477051, "grad_norm": 0.21334795653820038, "learning_rate": 9.43217665615142e-06, "epoch": 0.1892744479495268}}
14
- {"step": 325, "epoch": 0.20504731861198738, "time": "2026-07-08 22:05:54", "logs": {"loss": 0.12078428268432617, "grad_norm": 0.21560309827327728, "learning_rate": 1.022082018927445e-05, "epoch": 0.20504731861198738}}
15
- {"step": 350, "epoch": 0.22082018927444794, "time": "2026-07-08 22:06:17", "logs": {"loss": 0.10522660255432129, "grad_norm": 0.07722581177949905, "learning_rate": 1.1009463722397478e-05, "epoch": 0.22082018927444794}}
16
- {"step": 375, "epoch": 0.23659305993690852, "time": "2026-07-08 22:06:39", "logs": {"loss": 0.10268622398376465, "grad_norm": 13.963035583496094, "learning_rate": 1.1798107255520506e-05, "epoch": 0.23659305993690852}}
17
- {"step": 400, "epoch": 0.25236593059936907, "time": "2026-07-08 22:07:02", "logs": {"loss": 0.1168508243560791, "grad_norm": 1.7228655815124512, "learning_rate": 1.2586750788643533e-05, "epoch": 0.25236593059936907}}
18
- {"step": 425, "epoch": 0.26813880126182965, "time": "2026-07-08 22:07:25", "logs": {"loss": 0.1246561050415039, "grad_norm": 4.807567596435547, "learning_rate": 1.3375394321766562e-05, "epoch": 0.26813880126182965}}
19
- {"step": 450, "epoch": 0.28391167192429023, "time": "2026-07-08 22:07:47", "logs": {"loss": 0.06862223625183106, "grad_norm": 0.10336830466985703, "learning_rate": 1.4164037854889591e-05, "epoch": 0.28391167192429023}}
20
- {"step": 475, "epoch": 0.2996845425867508, "time": "2026-07-08 22:08:10", "logs": {"loss": 0.10785533905029297, "grad_norm": 0.18726640939712524, "learning_rate": 1.495268138801262e-05, "epoch": 0.2996845425867508}}
21
- {"step": 500, "epoch": 0.31545741324921134, "time": "2026-07-08 22:08:33", "logs": {"loss": 0.08045989990234376, "grad_norm": 7.849733829498291, "learning_rate": 1.574132492113565e-05, "epoch": 0.31545741324921134}}
22
- {"step": 500, "epoch": 0.31545741324921134, "time": "2026-07-08 22:09:29", "logs": {"eval_loss": 0.11268972605466843, "eval_accuracy": 0.9766561514195583, "eval_precision": 0.9654959950708565, "eval_recall": 0.9886435331230284, "eval_f1": 0.976932668329177, "eval_runtime": 56.1856, "eval_samples_per_second": 56.42, "eval_steps_per_second": 3.542, "epoch": 0.31545741324921134}}
23
- {"step": 525, "epoch": 0.3312302839116719, "time": "2026-07-08 22:10:07", "logs": {"loss": 0.12939174652099608, "grad_norm": 7.707481861114502, "learning_rate": 1.6529968454258676e-05, "epoch": 0.3312302839116719}}
24
- {"step": 550, "epoch": 0.3470031545741325, "time": "2026-07-08 22:10:31", "logs": {"loss": 0.14456929206848146, "grad_norm": 1.8563039302825928, "learning_rate": 1.7318611987381703e-05, "epoch": 0.3470031545741325}}
25
- {"step": 575, "epoch": 0.3627760252365931, "time": "2026-07-08 22:10:53", "logs": {"loss": 0.09543749809265137, "grad_norm": 9.819541931152344, "learning_rate": 1.8107255520504733e-05, "epoch": 0.3627760252365931}}
26
- {"step": 600, "epoch": 0.3785488958990536, "time": "2026-07-08 22:11:15", "logs": {"loss": 0.11727660179138183, "grad_norm": 2.0916662216186523, "learning_rate": 1.889589905362776e-05, "epoch": 0.3785488958990536}}
27
- {"step": 625, "epoch": 0.3943217665615142, "time": "2026-07-08 22:11:38", "logs": {"loss": 0.1419176959991455, "grad_norm": 4.153768539428711, "learning_rate": 1.968454258675079e-05, "epoch": 0.3943217665615142}}
28
- {"step": 650, "epoch": 0.41009463722397477, "time": "2026-07-08 22:12:01", "logs": {"loss": 0.0911902904510498, "grad_norm": 0.3790145814418793, "learning_rate": 1.999965897496297e-05, "epoch": 0.41009463722397477}}
29
- {"step": 675, "epoch": 0.42586750788643535, "time": "2026-07-08 22:12:23", "logs": {"loss": 0.14416172981262207, "grad_norm": 5.445947170257568, "learning_rate": 1.9997575017301754e-05, "epoch": 0.42586750788643535}}
30
- {"step": 700, "epoch": 0.4416403785488959, "time": "2026-07-08 22:12:46", "logs": {"loss": 0.14161142349243164, "grad_norm": 0.33051055669784546, "learning_rate": 1.999359695467383e-05, "epoch": 0.4416403785488959}}
31
- {"step": 725, "epoch": 0.45741324921135645, "time": "2026-07-08 22:13:09", "logs": {"loss": 0.13508522033691406, "grad_norm": 1.2220453023910522, "learning_rate": 1.998772554074878e-05, "epoch": 0.45741324921135645}}
32
- {"step": 750, "epoch": 0.47318611987381703, "time": "2026-07-08 22:13:31", "logs": {"loss": 0.14012475967407226, "grad_norm": 16.0678768157959, "learning_rate": 1.9979961887903783e-05, "epoch": 0.47318611987381703}}
33
- {"step": 775, "epoch": 0.4889589905362776, "time": "2026-07-08 22:13:54", "logs": {"loss": 0.08824824333190918, "grad_norm": 9.829991340637207, "learning_rate": 1.9970307467012862e-05, "epoch": 0.4889589905362776}}
34
- {"step": 800, "epoch": 0.5047318611987381, "time": "2026-07-08 22:14:16", "logs": {"loss": 0.14100438117980957, "grad_norm": 0.48923227190971375, "learning_rate": 1.9958764107168226e-05, "epoch": 0.5047318611987381}}
35
- {"step": 825, "epoch": 0.5205047318611987, "time": "2026-07-08 22:14:39", "logs": {"loss": 0.09922822952270507, "grad_norm": 10.579341888427734, "learning_rate": 1.994533399533373e-05, "epoch": 0.5205047318611987}}
36
- {"step": 850, "epoch": 0.5362776025236593, "time": "2026-07-08 22:15:02", "logs": {"loss": 0.09808570861816407, "grad_norm": 0.25369012355804443, "learning_rate": 1.9930019675930546e-05, "epoch": 0.5362776025236593}}
37
- {"step": 875, "epoch": 0.5520504731861199, "time": "2026-07-08 22:15:24", "logs": {"loss": 0.08875894546508789, "grad_norm": 0.5847018361091614, "learning_rate": 1.99128240503551e-05, "epoch": 0.5520504731861199}}
38
- {"step": 900, "epoch": 0.5678233438485805, "time": "2026-07-08 22:15:47", "logs": {"loss": 0.11056282043457032, "grad_norm": 0.4339558482170105, "learning_rate": 1.9893750376429396e-05, "epoch": 0.5678233438485805}}
39
- {"step": 925, "epoch": 0.583596214511041, "time": "2026-07-08 22:16:10", "logs": {"loss": 0.09730767250061036, "grad_norm": 2.8855783939361572, "learning_rate": 1.9872802267783777e-05, "epoch": 0.583596214511041}}
40
- {"step": 950, "epoch": 0.5993690851735016, "time": "2026-07-08 22:16:33", "logs": {"loss": 0.10335501670837402, "grad_norm": 6.300772666931152, "learning_rate": 1.9849983693172324e-05, "epoch": 0.5993690851735016}}
41
- {"step": 975, "epoch": 0.6151419558359621, "time": "2026-07-08 22:16:55", "logs": {"loss": 0.06983193874359131, "grad_norm": 0.1946100890636444, "learning_rate": 1.9825298975720938e-05, "epoch": 0.6151419558359621}}
42
- {"step": 1000, "epoch": 0.6309148264984227, "time": "2026-07-08 22:17:18", "logs": {"loss": 0.06307486057281494, "grad_norm": 0.05659753456711769, "learning_rate": 1.97987527921083e-05, "epoch": 0.6309148264984227}}
43
- {"step": 1000, "epoch": 0.6309148264984227, "time": "2026-07-08 22:18:14", "logs": {"eval_loss": 0.10372021794319153, "eval_accuracy": 0.9788643533123028, "eval_precision": 0.9645042839657283, "eval_recall": 0.9943217665615142, "eval_f1": 0.9791860826343585, "eval_runtime": 56.2004, "eval_samples_per_second": 56.405, "eval_steps_per_second": 3.541, "epoch": 0.6309148264984227}}
44
- {"step": 1025, "epoch": 0.6466876971608833, "time": "2026-07-08 22:18:51", "logs": {"loss": 0.07478074073791503, "grad_norm": 0.0960870087146759, "learning_rate": 1.977035017167984e-05, "epoch": 0.6466876971608833}}
45
- {"step": 1050, "epoch": 0.6624605678233438, "time": "2026-07-08 22:19:14", "logs": {"loss": 0.07338234424591064, "grad_norm": 2.863891124725342, "learning_rate": 1.974009649549489e-05, "epoch": 0.6624605678233438}}
46
- {"step": 1075, "epoch": 0.6782334384858044, "time": "2026-07-08 22:19:37", "logs": {"loss": 0.10608007431030274, "grad_norm": 0.3583734631538391, "learning_rate": 1.9707997495307232e-05, "epoch": 0.6782334384858044}}
47
- {"step": 1100, "epoch": 0.694006309148265, "time": "2026-07-08 22:19:59", "logs": {"loss": 0.07668295860290528, "grad_norm": 0.8535360097885132, "learning_rate": 1.9674059252479156e-05, "epoch": 0.694006309148265}}
48
- {"step": 1125, "epoch": 0.7097791798107256, "time": "2026-07-08 22:20:22", "logs": {"loss": 0.07370511531829833, "grad_norm": 0.4096319377422333, "learning_rate": 1.9638288196829303e-05, "epoch": 0.7097791798107256}}
49
- {"step": 1150, "epoch": 0.7255520504731862, "time": "2026-07-08 22:20:44", "logs": {"loss": 0.04484860897064209, "grad_norm": 0.10992075502872467, "learning_rate": 1.9600691105414523e-05, "epoch": 0.7255520504731862}}
50
- {"step": 1175, "epoch": 0.7413249211356467, "time": "2026-07-08 22:21:06", "logs": {"loss": 0.15258366584777833, "grad_norm": 4.087841510772705, "learning_rate": 1.9561275101245886e-05, "epoch": 0.7413249211356467}}
51
- {"step": 1200, "epoch": 0.7570977917981072, "time": "2026-07-08 22:21:29", "logs": {"loss": 0.12657273292541504, "grad_norm": 2.7859838008880615, "learning_rate": 1.95200476519392e-05, "epoch": 0.7570977917981072}}
52
- {"step": 1225, "epoch": 0.7728706624605678, "time": "2026-07-08 22:21:52", "logs": {"loss": 0.058558349609375, "grad_norm": 0.08890348672866821, "learning_rate": 1.9477016568300233e-05, "epoch": 0.7728706624605678}}
53
- {"step": 1250, "epoch": 0.7886435331230284, "time": "2026-07-08 22:22:14", "logs": {"loss": 0.08944664001464844, "grad_norm": 2.4231088161468506, "learning_rate": 1.9432190002844868e-05, "epoch": 0.7886435331230284}}
54
- {"step": 1275, "epoch": 0.804416403785489, "time": "2026-07-08 22:22:36", "logs": {"loss": 0.07989122390747071, "grad_norm": 0.15423132479190826, "learning_rate": 1.93855764482546e-05, "epoch": 0.804416403785489}}
55
- {"step": 1300, "epoch": 0.8201892744479495, "time": "2026-07-08 22:22:59", "logs": {"loss": 0.0904345703125, "grad_norm": 0.1543116569519043, "learning_rate": 1.933718473576751e-05, "epoch": 0.8201892744479495}}
56
- {"step": 1325, "epoch": 0.8359621451104101, "time": "2026-07-08 22:23:21", "logs": {"loss": 0.08953811645507813, "grad_norm": 0.2740912139415741, "learning_rate": 1.9287024033505156e-05, "epoch": 0.8359621451104101}}
57
- {"step": 1350, "epoch": 0.8517350157728707, "time": "2026-07-08 22:23:44", "logs": {"loss": 0.09112126350402833, "grad_norm": 0.06053502485156059, "learning_rate": 1.9235103844735587e-05, "epoch": 0.8517350157728707}}
58
- {"step": 1375, "epoch": 0.8675078864353313, "time": "2026-07-08 22:24:06", "logs": {"loss": 0.10066482543945313, "grad_norm": 3.397080898284912, "learning_rate": 1.9181434006072906e-05, "epoch": 0.8675078864353313}}
59
- {"step": 1400, "epoch": 0.8832807570977917, "time": "2026-07-08 22:24:29", "logs": {"loss": 0.06111194610595703, "grad_norm": 0.24578048288822174, "learning_rate": 1.9126024685613664e-05, "epoch": 0.8832807570977917}}
60
- {"step": 1425, "epoch": 0.8990536277602523, "time": "2026-07-08 22:24:51", "logs": {"loss": 0.07008316040039063, "grad_norm": 0.09262054413557053, "learning_rate": 1.9068886381010424e-05, "epoch": 0.8990536277602523}}
61
- {"step": 1450, "epoch": 0.9148264984227129, "time": "2026-07-08 22:25:14", "logs": {"loss": 0.14111000061035156, "grad_norm": 1.9535330533981323, "learning_rate": 1.9010029917482943e-05, "epoch": 0.9148264984227129}}
62
- {"step": 1475, "epoch": 0.9305993690851735, "time": "2026-07-08 22:25:36", "logs": {"loss": 0.07531814575195313, "grad_norm": 0.1983080357313156, "learning_rate": 1.894946644576724e-05, "epoch": 0.9305993690851735}}
63
- {"step": 1500, "epoch": 0.9463722397476341, "time": "2026-07-08 22:25:59", "logs": {"loss": 0.09356441497802734, "grad_norm": 0.09421352297067642, "learning_rate": 1.8887207440003028e-05, "epoch": 0.9463722397476341}}
64
- {"step": 1500, "epoch": 0.9463722397476341, "time": "2026-07-08 22:26:54", "logs": {"eval_loss": 0.07328534126281738, "eval_accuracy": 0.9782334384858045, "eval_precision": 0.9767295597484277, "eval_recall": 0.9798107255520505, "eval_f1": 0.9782677165354331, "eval_runtime": 55.7759, "eval_samples_per_second": 56.835, "eval_steps_per_second": 3.568, "epoch": 0.9463722397476341}}
65
- {"step": 1525, "epoch": 0.9621451104100947, "time": "2026-07-08 22:27:31", "logs": {"loss": 0.04988880157470703, "grad_norm": 0.23979413509368896, "learning_rate": 1.8823264695559876e-05, "epoch": 0.9621451104100947}}
66
- {"step": 1550, "epoch": 0.9779179810725552, "time": "2026-07-08 22:27:55", "logs": {"loss": 0.03976966857910156, "grad_norm": 0.1378251314163208, "learning_rate": 1.8757650326802474e-05, "epoch": 0.9779179810725552}}
67
- {"step": 1575, "epoch": 0.9936908517350158, "time": "2026-07-08 22:28:17", "logs": {"loss": 0.08686065673828125, "grad_norm": 0.06288843601942062, "learning_rate": 1.8690376764795512e-05, "epoch": 0.9936908517350158}}
68
- {"step": 1600, "epoch": 1.0094637223974763, "time": "2026-07-08 22:28:40", "logs": {"loss": 0.05250164031982422, "grad_norm": 5.365837097167969, "learning_rate": 1.8621456754948524e-05, "epoch": 1.0094637223974763}}
69
- {"step": 1625, "epoch": 1.025236593059937, "time": "2026-07-08 22:29:03", "logs": {"loss": 0.05328141212463379, "grad_norm": 0.18591798841953278, "learning_rate": 1.8550903354601182e-05, "epoch": 1.025236593059937}}
70
- {"step": 1650, "epoch": 1.0410094637223974, "time": "2026-07-08 22:29:25", "logs": {"loss": 0.0559916353225708, "grad_norm": 0.14107061922550201, "learning_rate": 1.847872993054952e-05, "epoch": 1.0410094637223974}}
71
- {"step": 1675, "epoch": 1.0567823343848581, "time": "2026-07-08 22:29:48", "logs": {"loss": 0.1134580421447754, "grad_norm": 6.363674163818359, "learning_rate": 1.840495015651348e-05, "epoch": 1.0567823343848581}}
72
- {"step": 1700, "epoch": 1.0725552050473186, "time": "2026-07-08 22:30:10", "logs": {"loss": 0.04161137580871582, "grad_norm": 0.072722889482975, "learning_rate": 1.8329578010546364e-05, "epoch": 1.0725552050473186}}
73
- {"step": 1725, "epoch": 1.088328075709779, "time": "2026-07-08 22:30:33", "logs": {"loss": 0.07552608013153077, "grad_norm": 7.499703884124756, "learning_rate": 1.825262777238659e-05, "epoch": 1.088328075709779}}
74
- {"step": 1750, "epoch": 1.1041009463722398, "time": "2026-07-08 22:30:55", "logs": {"loss": 0.07142114162445068, "grad_norm": 0.034939683973789215, "learning_rate": 1.817411402075229e-05, "epoch": 1.1041009463722398}}
75
- {"step": 1775, "epoch": 1.1198738170347002, "time": "2026-07-08 22:31:18", "logs": {"loss": 0.09205500602722168, "grad_norm": 0.19647593796253204, "learning_rate": 1.809405163057931e-05, "epoch": 1.1198738170347002}}
76
- {"step": 1800, "epoch": 1.135646687697161, "time": "2026-07-08 22:31:41", "logs": {"loss": 0.06862086296081543, "grad_norm": 0.1461659073829651, "learning_rate": 1.8012455770203038e-05, "epoch": 1.135646687697161}}
77
- {"step": 1825, "epoch": 1.1514195583596214, "time": "2026-07-08 22:32:03", "logs": {"loss": 0.03869293212890625, "grad_norm": 0.09000778943300247, "learning_rate": 1.7929341898484653e-05, "epoch": 1.1514195583596214}}
78
- {"step": 1850, "epoch": 1.167192429022082, "time": "2026-07-08 22:32:26", "logs": {"loss": 0.10784174919128418, "grad_norm": 3.4395689964294434, "learning_rate": 1.784472576188237e-05, "epoch": 1.167192429022082}}
79
- {"step": 1875, "epoch": 1.1829652996845426, "time": "2026-07-08 22:32:49", "logs": {"loss": 0.08741586685180663, "grad_norm": 6.185035228729248, "learning_rate": 1.775862339146815e-05, "epoch": 1.1829652996845426}}
80
- {"step": 1900, "epoch": 1.1987381703470033, "time": "2026-07-08 22:33:12", "logs": {"loss": 0.03821196556091309, "grad_norm": 0.03306306153535843, "learning_rate": 1.7671051099890538e-05, "epoch": 1.1987381703470033}}
81
- {"step": 1925, "epoch": 1.2145110410094637, "time": "2026-07-08 22:33:35", "logs": {"loss": 0.05004812240600586, "grad_norm": 0.11795313656330109, "learning_rate": 1.758202547828409e-05, "epoch": 1.2145110410094637}}
82
- {"step": 1950, "epoch": 1.2302839116719242, "time": "2026-07-08 22:33:57", "logs": {"loss": 0.09024031639099121, "grad_norm": 0.27102428674697876, "learning_rate": 1.74915633931261e-05, "epoch": 1.2302839116719242}}
83
- {"step": 1975, "epoch": 1.2460567823343849, "time": "2026-07-08 22:34:20", "logs": {"loss": 0.0506083345413208, "grad_norm": 0.1362110674381256, "learning_rate": 1.739968198304114e-05, "epoch": 1.2460567823343849}}
84
- {"step": 2000, "epoch": 1.2618296529968454, "time": "2026-07-08 22:34:43", "logs": {"loss": 0.12149024963378906, "grad_norm": 0.05871753394603729, "learning_rate": 1.7306398655554023e-05, "epoch": 1.2618296529968454}}
85
- {"step": 2000, "epoch": 1.2618296529968454, "time": "2026-07-08 22:35:39", "logs": {"eval_loss": 0.0678657591342926, "eval_accuracy": 0.9832807570977918, "eval_precision": 0.974009900990099, "eval_recall": 0.9930599369085173, "eval_f1": 0.9834426741643236, "eval_runtime": 56.2139, "eval_samples_per_second": 56.392, "eval_steps_per_second": 3.54, "epoch": 1.2618296529968454}}
86
- {"step": 2025, "epoch": 1.277602523659306, "time": "2026-07-08 22:36:16", "logs": {"loss": 0.09227321624755859, "grad_norm": 14.435188293457031, "learning_rate": 1.721173108379184e-05, "epoch": 1.277602523659306}}
87
- {"step": 2050, "epoch": 1.2933753943217665, "time": "2026-07-08 22:36:39", "logs": {"loss": 0.07053103446960449, "grad_norm": 0.14022010564804077, "learning_rate": 1.711569720313568e-05, "epoch": 1.2933753943217665}}
88
- {"step": 2075, "epoch": 1.3091482649842272, "time": "2026-07-08 22:37:02", "logs": {"loss": 0.07794711112976074, "grad_norm": 0.17620061337947845, "learning_rate": 1.701831520782264e-05, "epoch": 1.3091482649842272}}
89
- {"step": 2100, "epoch": 1.3249211356466877, "time": "2026-07-08 22:37:24", "logs": {"loss": 0.05158097267150879, "grad_norm": 6.461155891418457, "learning_rate": 1.6919603547498832e-05, "epoch": 1.3249211356466877}}
90
- {"step": 2125, "epoch": 1.3406940063091484, "time": "2026-07-08 22:37:47", "logs": {"loss": 0.04181491374969482, "grad_norm": 8.352033615112305, "learning_rate": 1.6819580923723957e-05, "epoch": 1.3406940063091484}}
91
- {"step": 2150, "epoch": 1.3564668769716088, "time": "2026-07-08 22:38:10", "logs": {"loss": 0.049018054008483886, "grad_norm": 0.052411582320928574, "learning_rate": 1.671826628642819e-05, "epoch": 1.3564668769716088}}
92
- {"step": 2175, "epoch": 1.3722397476340693, "time": "2026-07-08 22:38:32", "logs": {"loss": 0.06123344898223877, "grad_norm": 0.0428645983338356, "learning_rate": 1.6615678830321987e-05, "epoch": 1.3722397476340693}}
93
- {"step": 2200, "epoch": 1.38801261829653, "time": "2026-07-08 22:38:55", "logs": {"loss": 0.06224549293518066, "grad_norm": 0.12796339392662048, "learning_rate": 1.651183799125952e-05, "epoch": 1.38801261829653}}
94
- {"step": 2225, "epoch": 1.4037854889589905, "time": "2026-07-08 22:39:18", "logs": {"loss": 0.04998977661132813, "grad_norm": 0.3184094727039337, "learning_rate": 1.640676344255643e-05, "epoch": 1.4037854889589905}}
95
- {"step": 2250, "epoch": 1.4195583596214512, "time": "2026-07-08 22:39:40", "logs": {"loss": 0.07321974277496338, "grad_norm": 0.08195166289806366, "learning_rate": 1.6300475091262615e-05, "epoch": 1.4195583596214512}}
96
- {"step": 2275, "epoch": 1.4353312302839116, "time": "2026-07-08 22:40:03", "logs": {"loss": 0.08428791999816894, "grad_norm": 3.03761625289917, "learning_rate": 1.6192993074390658e-05, "epoch": 1.4353312302839116}}
97
- {"step": 2300, "epoch": 1.4511041009463723, "time": "2026-07-08 22:40:26", "logs": {"loss": 0.06157721996307373, "grad_norm": 0.41116461157798767, "learning_rate": 1.6084337755100795e-05, "epoch": 1.4511041009463723}}
98
- {"step": 2325, "epoch": 1.4668769716088328, "time": "2026-07-08 22:40:48", "logs": {"loss": 0.09335646629333497, "grad_norm": 4.645056247711182, "learning_rate": 1.597452971884296e-05, "epoch": 1.4668769716088328}}
99
- {"step": 2350, "epoch": 1.4826498422712935, "time": "2026-07-08 22:41:11", "logs": {"loss": 0.035682902336120606, "grad_norm": 0.07679587602615356, "learning_rate": 1.5863589769456725e-05, "epoch": 1.4826498422712935}}
100
- {"step": 2375, "epoch": 1.498422712933754, "time": "2026-07-08 22:41:33", "logs": {"loss": 0.08860483169555664, "grad_norm": 16.212730407714844, "learning_rate": 1.575153892522991e-05, "epoch": 1.498422712933754}}
101
- {"step": 2400, "epoch": 1.5141955835962144, "time": "2026-07-08 22:41:56", "logs": {"loss": 0.06739212989807129, "grad_norm": 0.035901762545108795, "learning_rate": 1.5638398414916504e-05, "epoch": 1.5141955835962144}}
102
- {"step": 2425, "epoch": 1.5299684542586751, "time": "2026-07-08 22:42:18", "logs": {"loss": 0.03637696027755737, "grad_norm": 14.2339448928833, "learning_rate": 1.552418967371476e-05, "epoch": 1.5299684542586751}}
103
- {"step": 2450, "epoch": 1.5457413249211358, "time": "2026-07-08 22:42:40", "logs": {"loss": 0.03717231750488281, "grad_norm": 0.021874595433473587, "learning_rate": 1.540893433920614e-05, "epoch": 1.5457413249211358}}
104
- {"step": 2475, "epoch": 1.5615141955835963, "time": "2026-07-08 22:43:03", "logs": {"loss": 0.02598174810409546, "grad_norm": 0.048296067863702774, "learning_rate": 1.5292654247255976e-05, "epoch": 1.5615141955835963}}
105
- {"step": 2500, "epoch": 1.5772870662460567, "time": "2026-07-08 22:43:26", "logs": {"loss": 0.06133613586425781, "grad_norm": 9.524810791015625, "learning_rate": 1.5175371427876462e-05, "epoch": 1.5772870662460567}}
106
- {"step": 2500, "epoch": 1.5772870662460567, "time": "2026-07-08 22:44:22", "logs": {"eval_loss": 0.07992753386497498, "eval_accuracy": 0.9826498422712934, "eval_precision": 0.9745657568238213, "eval_recall": 0.9911671924290221, "eval_f1": 0.9827963715983735, "eval_runtime": 55.9735, "eval_samples_per_second": 56.634, "eval_steps_per_second": 3.555, "epoch": 1.5772870662460567}}
107
- {"step": 2525, "epoch": 1.5930599369085172, "time": "2026-07-08 22:45:00", "logs": {"loss": 0.05316779613494873, "grad_norm": 0.015798641368746758, "learning_rate": 1.5057108101052978e-05, "epoch": 1.5930599369085172}}
108
- {"step": 2550, "epoch": 1.608832807570978, "time": "2026-07-08 22:45:23", "logs": {"loss": 0.04096637725830078, "grad_norm": 1.9981679916381836, "learning_rate": 1.4937886672534355e-05, "epoch": 1.608832807570978}}
109
- {"step": 2575, "epoch": 1.6246056782334386, "time": "2026-07-08 22:45:46", "logs": {"loss": 0.11973735809326172, "grad_norm": 14.948141098022461, "learning_rate": 1.4817729729587951e-05, "epoch": 1.6246056782334386}}
110
- {"step": 2600, "epoch": 1.640378548895899, "time": "2026-07-08 22:46:08", "logs": {"loss": 0.07024142742156983, "grad_norm": 10.006237030029297, "learning_rate": 1.4696660036720366e-05, "epoch": 1.640378548895899}}
111
- {"step": 2625, "epoch": 1.6561514195583595, "time": "2026-07-08 22:46:31", "logs": {"loss": 0.05831857681274414, "grad_norm": 2.659205436706543, "learning_rate": 1.4574700531364538e-05, "epoch": 1.6561514195583595}}
112
- {"step": 2650, "epoch": 1.6719242902208202, "time": "2026-07-08 22:46:53", "logs": {"loss": 0.012276544570922851, "grad_norm": 0.023404600098729134, "learning_rate": 1.445187431953413e-05, "epoch": 1.6719242902208202}}
113
- {"step": 2675, "epoch": 1.687697160883281, "time": "2026-07-08 22:47:15", "logs": {"loss": 0.04452509880065918, "grad_norm": 0.05033299699425697, "learning_rate": 1.4328204671445906e-05, "epoch": 1.687697160883281}}
114
- {"step": 2700, "epoch": 1.7034700315457414, "time": "2026-07-08 22:47:39", "logs": {"loss": 0.03908012628555298, "grad_norm": 0.008091382682323456, "learning_rate": 1.4203715017111077e-05, "epoch": 1.7034700315457414}}
115
- {"step": 2725, "epoch": 1.7192429022082019, "time": "2026-07-08 22:48:01", "logs": {"loss": 0.07871007442474365, "grad_norm": 0.8330066800117493, "learning_rate": 1.4078428941896308e-05, "epoch": 1.7192429022082019}}
116
- {"step": 2750, "epoch": 1.7350157728706623, "time": "2026-07-08 22:48:23", "logs": {"loss": 0.05369951725006104, "grad_norm": 0.20014899969100952, "learning_rate": 1.3952370182055332e-05, "epoch": 1.7350157728706623}}
117
- {"step": 2775, "epoch": 1.750788643533123, "time": "2026-07-08 22:48:46", "logs": {"loss": 0.06861288547515869, "grad_norm": 5.281189918518066, "learning_rate": 1.3825562620231962e-05, "epoch": 1.750788643533123}}
118
- {"step": 2800, "epoch": 1.7665615141955837, "time": "2026-07-08 22:49:09", "logs": {"loss": 0.0692290210723877, "grad_norm": 13.993217468261719, "learning_rate": 1.3698030280935365e-05, "epoch": 1.7665615141955837}}
119
- {"step": 2825, "epoch": 1.7823343848580442, "time": "2026-07-08 22:49:31", "logs": {"loss": 0.0943898868560791, "grad_norm": 0.19194826483726501, "learning_rate": 1.356979732598847e-05, "epoch": 1.7823343848580442}}
120
- {"step": 2850, "epoch": 1.7981072555205047, "time": "2026-07-08 22:49:54", "logs": {"loss": 0.05583916664123535, "grad_norm": 0.24530774354934692, "learning_rate": 1.3440888049950352e-05, "epoch": 1.7981072555205047}}
121
- {"step": 2875, "epoch": 1.8138801261829653, "time": "2026-07-08 22:50:16", "logs": {"loss": 0.06628149509429931, "grad_norm": 23.30782699584961, "learning_rate": 1.3311326875513473e-05, "epoch": 1.8138801261829653}}
122
- {"step": 2900, "epoch": 1.8296529968454258, "time": "2026-07-08 22:50:39", "logs": {"loss": 0.08753268241882324, "grad_norm": 5.725240230560303, "learning_rate": 1.3181138348876637e-05, "epoch": 1.8296529968454258}}
123
- {"step": 2925, "epoch": 1.8454258675078865, "time": "2026-07-08 22:51:01", "logs": {"loss": 0.04335375308990479, "grad_norm": 3.2451798915863037, "learning_rate": 1.305034713509457e-05, "epoch": 1.8454258675078865}}
124
- {"step": 2950, "epoch": 1.861198738170347, "time": "2026-07-08 22:51:24", "logs": {"loss": 0.08735511779785156, "grad_norm": 8.147355079650879, "learning_rate": 1.2918978013404961e-05, "epoch": 1.861198738170347}}
125
- {"step": 2975, "epoch": 1.8769716088328074, "time": "2026-07-08 22:51:46", "logs": {"loss": 0.04033580780029297, "grad_norm": 0.31234487891197205, "learning_rate": 1.2787055872533867e-05, "epoch": 1.8769716088328074}}
126
- {"step": 3000, "epoch": 1.8927444794952681, "time": "2026-07-08 22:52:09", "logs": {"loss": 0.07418310165405273, "grad_norm": 2.878502368927002, "learning_rate": 1.2654605705980394e-05, "epoch": 1.8927444794952681}}
127
- {"step": 3000, "epoch": 1.8927444794952681, "time": "2026-07-08 22:53:05", "logs": {"eval_loss": 0.09431762993335724, "eval_accuracy": 0.9823343848580441, "eval_precision": 0.9716224552745218, "eval_recall": 0.9936908517350158, "eval_f1": 0.982532751091703, "eval_runtime": 55.9884, "eval_samples_per_second": 56.619, "eval_steps_per_second": 3.554, "epoch": 1.8927444794952681}}
128
- {"step": 3000, "epoch": 1.8927444794952681, "time": "2026-07-08 22:53:20", "logs": {"train_runtime": 3136.8433, "train_samples_per_second": 32.331, "train_steps_per_second": 2.021, "total_flos": 1.257414311341824e+16, "train_loss": 0.10879098949829738, "epoch": 1.8927444794952681}}
129
- {"step": 3000, "epoch": 1.8927444794952681, "time": "2026-07-08 22:54:17", "logs": {"eval_loss": 0.0678657591342926, "eval_accuracy": 0.9832807570977918, "eval_precision": 0.974009900990099, "eval_recall": 0.9930599369085173, "eval_f1": 0.9834426741643236, "eval_runtime": 56.4471, "eval_samples_per_second": 56.159, "eval_steps_per_second": 3.525, "epoch": 1.8927444794952681}}
 
1
+ {"step": 3025, "epoch": 1.9085173501577288, "time": "2026-07-08 23:43:48", "logs": {"loss": 0.05197489261627197, "grad_norm": 32.33955001831055, "learning_rate": 1.2521652607281511e-05, "epoch": 1.9085173501577288}}
2
+ {"step": 3050, "epoch": 1.9242902208201893, "time": "2026-07-08 23:44:08", "logs": {"loss": 0.08039166450500489, "grad_norm": 5.458424091339111, "learning_rate": 1.2388221765257922e-05, "epoch": 1.9242902208201893}}
3
+ {"step": 3075, "epoch": 1.9400630914826498, "time": "2026-07-08 23:44:30", "logs": {"loss": 0.03573692560195923, "grad_norm": 0.032594233751297, "learning_rate": 1.2254338459241878e-05, "epoch": 1.9400630914826498}}
4
+ {"step": 3100, "epoch": 1.9558359621451105, "time": "2026-07-08 23:44:53", "logs": {"loss": 0.07872509479522705, "grad_norm": 13.031876564025879, "learning_rate": 1.2120028054287853e-05, "epoch": 1.9558359621451105}}
5
+ {"step": 3125, "epoch": 1.971608832807571, "time": "2026-07-08 23:45:17", "logs": {"loss": 0.07527520656585693, "grad_norm": 3.845304012298584, "learning_rate": 1.198531599636696e-05, "epoch": 1.971608832807571}}
6
+ {"step": 3150, "epoch": 1.9873817034700316, "time": "2026-07-08 23:45:41", "logs": {"loss": 0.08369281768798828, "grad_norm": 0.061868082731962204, "learning_rate": 1.1850227807546056e-05, "epoch": 1.9873817034700316}}
7
+ {"step": 3175, "epoch": 2.003154574132492, "time": "2026-07-08 23:46:04", "logs": {"loss": 0.06914769172668457, "grad_norm": 0.2873733639717102, "learning_rate": 1.1714789081152414e-05, "epoch": 2.003154574132492}}
8
+ {"step": 3200, "epoch": 2.0189274447949526, "time": "2026-07-08 23:46:28", "logs": {"loss": 0.04823586940765381, "grad_norm": 0.2408512681722641, "learning_rate": 1.1579025476924912e-05, "epoch": 2.0189274447949526}}
9
+ {"step": 3225, "epoch": 2.034700315457413, "time": "2026-07-08 23:46:52", "logs": {"loss": 0.052933053970336916, "grad_norm": 0.06181686371564865, "learning_rate": 1.1442962716152626e-05, "epoch": 2.034700315457413}}
10
+ {"step": 3250, "epoch": 2.050473186119874, "time": "2026-07-08 23:47:16", "logs": {"loss": 0.042997312545776364, "grad_norm": 0.23159997165203094, "learning_rate": 1.1306626576801762e-05, "epoch": 2.050473186119874}}
11
+ {"step": 3275, "epoch": 2.0662460567823344, "time": "2026-07-08 23:47:40", "logs": {"loss": 0.030156188011169434, "grad_norm": 0.11183202266693115, "learning_rate": 1.1170042888631865e-05, "epoch": 2.0662460567823344}}
12
+ {"step": 3300, "epoch": 2.082018927444795, "time": "2026-07-08 23:48:04", "logs": {"loss": 0.06210995674133301, "grad_norm": 0.020436838269233704, "learning_rate": 1.103323752830219e-05, "epoch": 2.082018927444795}}
13
+ {"step": 3325, "epoch": 2.0977917981072554, "time": "2026-07-08 23:48:28", "logs": {"loss": 0.06024267673492432, "grad_norm": 0.09942680597305298, "learning_rate": 1.0896236414469217e-05, "epoch": 2.0977917981072554}}
14
+ {"step": 3350, "epoch": 2.1135646687697163, "time": "2026-07-08 23:48:52", "logs": {"loss": 0.09793594360351562, "grad_norm": 0.06319867074489594, "learning_rate": 1.075906550287618e-05, "epoch": 2.1135646687697163}}
15
+ {"step": 3375, "epoch": 2.1293375394321767, "time": "2026-07-08 23:49:16", "logs": {"loss": 0.039036962985992435, "grad_norm": 0.14570729434490204, "learning_rate": 1.0621750781435597e-05, "epoch": 2.1293375394321767}}
16
+ {"step": 3400, "epoch": 2.145110410094637, "time": "2026-07-08 23:49:39", "logs": {"loss": 0.024112117290496827, "grad_norm": 30.197425842285156, "learning_rate": 1.048431826530569e-05, "epoch": 2.145110410094637}}
17
+ {"step": 3425, "epoch": 2.1608832807570977, "time": "2026-07-08 23:50:03", "logs": {"loss": 0.09378963470458984, "grad_norm": 0.0806874930858612, "learning_rate": 1.0346793991961636e-05, "epoch": 2.1608832807570977}}
18
+ {"step": 3450, "epoch": 2.176656151419558, "time": "2026-07-08 23:50:27", "logs": {"loss": 0.028721425533294678, "grad_norm": 0.041268810629844666, "learning_rate": 1.0209204016262617e-05, "epoch": 2.176656151419558}}
19
+ {"step": 3475, "epoch": 2.192429022082019, "time": "2026-07-08 23:50:51", "logs": {"loss": 0.019525480270385743, "grad_norm": 0.027517961338162422, "learning_rate": 1.0071574405515545e-05, "epoch": 2.192429022082019}}
20
+ {"step": 3500, "epoch": 2.2082018927444795, "time": "2026-07-08 23:51:15", "logs": {"loss": 0.07426059722900391, "grad_norm": 0.06598785519599915, "learning_rate": 9.93393123453644e-06, "epoch": 2.2082018927444795}}
21
+ {"step": 3500, "epoch": 2.2082018927444795, "time": "2026-07-08 23:52:15", "logs": {"eval_loss": 0.07981470227241516, "eval_accuracy": 0.9820189274447949, "eval_precision": 0.9769038701622972, "eval_recall": 0.9873817034700315, "eval_f1": 0.9821148415437716, "eval_runtime": 59.5362, "eval_samples_per_second": 53.245, "eval_steps_per_second": 3.343, "epoch": 2.2082018927444795}}