coung21 commited on
Commit
ed2bde4
·
verified ·
1 Parent(s): 728cd82

Upload folder using huggingface_hub

Browse files
ssl_dino2/resnet50/checkpoints/epoch=299-step=18900.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84d42e3168e6d8fca3d793b8533c42680218f43967614aa7fb2feced4521ce76
3
+ size 409504589
ssl_dino2/resnet50/checkpoints/last.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b1f7ed50e907ef22be61a0dd76e808a601f2b38d0da71c2b69d9d953db8a701
3
+ size 409504589
ssl_dino2/resnet50/exported_models/exported_last.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:951bf0000677a8bd490e4e2e20c4214a82e0ab9d58e5024796559887b22f07c7
3
+ size 102546083
ssl_dino2/resnet50/metrics.jsonl ADDED
@@ -0,0 +1,756 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"lr-SGD/params": 0.0005952380952380952, "lr-SGD/params_no_weight_decay": 0.0005952380952380952, "step": 49}
2
+ {"train_loss": 7.036804676055908, "profiling/batch_time": 0.23841606080532074, "profiling/data_time": 0.1660670042037964, "epoch": 0, "step": 49}
3
+ {"lr-SGD/params": 0.0011904761904761904, "lr-SGD/params_no_weight_decay": 0.0011904761904761904, "step": 99}
4
+ {"train_loss": 6.965976715087891, "profiling/batch_time": 0.1360819935798645, "profiling/data_time": 0.16384150087833405, "epoch": 1, "step": 99}
5
+ {"lr-SGD/params": 0.0017857142857142854, "lr-SGD/params_no_weight_decay": 0.0017857142857142854, "step": 149}
6
+ {"train_loss": 6.940258979797363, "profiling/batch_time": 0.12056228518486023, "profiling/data_time": 0.15889692306518555, "epoch": 2, "step": 149}
7
+ {"lr-SGD/params": 0.0023809523809523807, "lr-SGD/params_no_weight_decay": 0.0023809523809523807, "step": 199}
8
+ {"train_loss": 6.926732063293457, "profiling/batch_time": 0.1210741251707077, "profiling/data_time": 0.04445579648017883, "epoch": 3, "step": 199}
9
+ {"lr-SGD/params": 0.002976190476190476, "lr-SGD/params_no_weight_decay": 0.002976190476190476, "step": 249}
10
+ {"train_loss": 6.941986083984375, "profiling/batch_time": 0.12421628087759018, "profiling/data_time": 0.050522174686193466, "epoch": 3, "step": 249}
11
+ {"lr-SGD/params": 0.003571428571428571, "lr-SGD/params_no_weight_decay": 0.003571428571428571, "step": 299}
12
+ {"train_loss": 6.950067520141602, "profiling/batch_time": 0.12523356080055237, "profiling/data_time": 0.15994834899902344, "epoch": 4, "step": 299}
13
+ {"lr-SGD/params": 0.004166666666666667, "lr-SGD/params_no_weight_decay": 0.004166666666666667, "step": 349}
14
+ {"train_loss": 6.940849781036377, "profiling/batch_time": 0.24460849165916443, "profiling/data_time": 0.15944863855838776, "epoch": 5, "step": 349}
15
+ {"lr-SGD/params": 0.0047619047619047615, "lr-SGD/params_no_weight_decay": 0.0047619047619047615, "step": 399}
16
+ {"train_loss": 6.932730674743652, "profiling/batch_time": 0.12378929555416107, "profiling/data_time": 0.15882480144500732, "epoch": 6, "step": 399}
17
+ {"lr-SGD/params": 0.005357142857142857, "lr-SGD/params_no_weight_decay": 0.005357142857142857, "step": 449}
18
+ {"train_loss": 6.931278228759766, "profiling/batch_time": 0.12065120041370392, "profiling/data_time": 0.1615215390920639, "epoch": 7, "step": 449}
19
+ {"lr-SGD/params": 0.005952380952380952, "lr-SGD/params_no_weight_decay": 0.005952380952380952, "step": 499}
20
+ {"train_loss": 6.932972431182861, "profiling/batch_time": 0.12038084864616394, "profiling/data_time": 0.15713834762573242, "epoch": 7, "step": 499}
21
+ {"lr-SGD/params": 0.006547619047619048, "lr-SGD/params_no_weight_decay": 0.006547619047619048, "step": 549}
22
+ {"train_loss": 6.931995868682861, "profiling/batch_time": 0.12686997652053833, "profiling/data_time": 0.04593503475189209, "epoch": 8, "step": 549}
23
+ {"lr-SGD/params": 0.007142857142857142, "lr-SGD/params_no_weight_decay": 0.007142857142857142, "step": 599}
24
+ {"train_loss": 6.93213415145874, "profiling/batch_time": 0.1245480626821518, "profiling/data_time": 0.15903733670711517, "epoch": 9, "step": 599}
25
+ {"lr-SGD/params": 0.007499980003970461, "lr-SGD/params_no_weight_decay": 0.007499980003970461, "step": 649}
26
+ {"train_loss": 6.931866645812988, "profiling/batch_time": 0.2394903004169464, "profiling/data_time": 0.15860483050346375, "epoch": 10, "step": 649}
27
+ {"lr-SGD/params": 0.007499736287910266, "lr-SGD/params_no_weight_decay": 0.007499736287910266, "step": 699}
28
+ {"train_loss": 6.929832935333252, "profiling/batch_time": 0.12808623909950256, "profiling/data_time": 0.15933647751808167, "epoch": 11, "step": 699}
29
+ {"lr-SGD/params": 0.007499215639487837, "lr-SGD/params_no_weight_decay": 0.007499215639487837, "step": 749}
30
+ {"train_loss": 6.9300103187561035, "profiling/batch_time": 0.12572212517261505, "profiling/data_time": 0.155305415391922, "epoch": 11, "step": 749}
31
+ {"lr-SGD/params": 0.007498418097193544, "lr-SGD/params_no_weight_decay": 0.007498418097193544, "step": 799}
32
+ {"train_loss": 6.930948734283447, "profiling/batch_time": 0.12079284340143204, "profiling/data_time": 0.16065701842308044, "epoch": 12, "step": 799}
33
+ {"lr-SGD/params": 0.0074973437199878975, "lr-SGD/params_no_weight_decay": 0.0074973437199878975, "step": 849}
34
+ {"train_loss": 6.930294036865234, "profiling/batch_time": 0.1262594610452652, "profiling/data_time": 0.04135080426931381, "epoch": 13, "step": 849}
35
+ {"lr-SGD/params": 0.007495992587297198, "lr-SGD/params_no_weight_decay": 0.007495992587297198, "step": 899}
36
+ {"train_loss": 6.927713871002197, "profiling/batch_time": 0.1269543170928955, "profiling/data_time": 0.16352324187755585, "epoch": 14, "step": 899}
37
+ {"lr-SGD/params": 0.00749436479900765, "lr-SGD/params_no_weight_decay": 0.00749436479900765, "step": 949}
38
+ {"train_loss": 6.927277565002441, "profiling/batch_time": 0.23882454633712769, "profiling/data_time": 0.20522357523441315, "epoch": 15, "step": 949}
39
+ {"lr-SGD/params": 0.00749246047545799, "lr-SGD/params_no_weight_decay": 0.00749246047545799, "step": 999}
40
+ {"train_loss": 6.926417350769043, "profiling/batch_time": 0.23896293342113495, "profiling/data_time": 0.1619165688753128, "epoch": 15, "step": 999}
41
+ {"lr-SGD/params": 0.007490279757430588, "lr-SGD/params_no_weight_decay": 0.007490279757430588, "step": 1049}
42
+ {"train_loss": 6.926244735717773, "profiling/batch_time": 0.1202581375837326, "profiling/data_time": 0.1590297520160675, "epoch": 16, "step": 1049}
43
+ {"lr-SGD/params": 0.007487822806141035, "lr-SGD/params_no_weight_decay": 0.007487822806141035, "step": 1099}
44
+ {"train_loss": 6.922150611877441, "profiling/batch_time": 0.12845197319984436, "profiling/data_time": 0.1568428874015808, "epoch": 17, "step": 1099}
45
+ {"lr-SGD/params": 0.007485089803226227, "lr-SGD/params_no_weight_decay": 0.007485089803226227, "step": 1149}
46
+ {"train_loss": 6.925070762634277, "profiling/batch_time": 0.12096714973449707, "profiling/data_time": 0.04123983904719353, "epoch": 18, "step": 1149}
47
+ {"lr-SGD/params": 0.007482080950730939, "lr-SGD/params_no_weight_decay": 0.007482080950730939, "step": 1199}
48
+ {"train_loss": 6.925851821899414, "profiling/batch_time": 0.13239610195159912, "profiling/data_time": 0.15627901256084442, "epoch": 19, "step": 1199}
49
+ {"lr-SGD/params": 0.007478796471092884, "lr-SGD/params_no_weight_decay": 0.007478796471092884, "step": 1249}
50
+ {"train_loss": 6.916351795196533, "profiling/batch_time": 0.12638841569423676, "profiling/data_time": 0.15642796456813812, "epoch": 19, "step": 1249}
51
+ {"lr-SGD/params": 0.0074752366071262726, "lr-SGD/params_no_weight_decay": 0.0074752366071262726, "step": 1299}
52
+ {"train_loss": 6.913606643676758, "profiling/batch_time": 0.2442007213830948, "profiling/data_time": 0.1591879427433014, "epoch": 20, "step": 1299}
53
+ {"lr-SGD/params": 0.007471401622003865, "lr-SGD/params_no_weight_decay": 0.007471401622003865, "step": 1349}
54
+ {"train_loss": 6.908364772796631, "profiling/batch_time": 0.1238674595952034, "profiling/data_time": 0.15945395827293396, "epoch": 21, "step": 1349}
55
+ {"lr-SGD/params": 0.007467291799237504, "lr-SGD/params_no_weight_decay": 0.007467291799237504, "step": 1399}
56
+ {"train_loss": 6.8991804122924805, "profiling/batch_time": 0.12550008296966553, "profiling/data_time": 0.15993967652320862, "epoch": 22, "step": 1399}
57
+ {"lr-SGD/params": 0.007462907442657168, "lr-SGD/params_no_weight_decay": 0.007462907442657168, "step": 1449}
58
+ {"train_loss": 6.8715410232543945, "profiling/batch_time": 0.23759643733501434, "profiling/data_time": 5.66455602645874, "epoch": 23, "step": 1449}
59
+ {"lr-SGD/params": 0.007458248876388503, "lr-SGD/params_no_weight_decay": 0.007458248876388503, "step": 1499}
60
+ {"train_loss": 6.820511817932129, "profiling/batch_time": 0.12390826642513275, "profiling/data_time": 0.04182536527514458, "epoch": 23, "step": 1499}
61
+ {"lr-SGD/params": 0.007453316444828859, "lr-SGD/params_no_weight_decay": 0.007453316444828859, "step": 1549}
62
+ {"train_loss": 6.750844955444336, "profiling/batch_time": 0.1264069378376007, "profiling/data_time": 0.16497869789600372, "epoch": 24, "step": 1549}
63
+ {"lr-SGD/params": 0.007448110512621841, "lr-SGD/params_no_weight_decay": 0.007448110512621841, "step": 1599}
64
+ {"train_loss": 6.6576151847839355, "profiling/batch_time": 0.24733486771583557, "profiling/data_time": 0.17800572514533997, "epoch": 25, "step": 1599}
65
+ {"lr-SGD/params": 0.007442631464630333, "lr-SGD/params_no_weight_decay": 0.007442631464630333, "step": 1649}
66
+ {"train_loss": 6.609510898590088, "profiling/batch_time": 0.12185362726449966, "profiling/data_time": 0.16850201785564423, "epoch": 26, "step": 1649}
67
+ {"lr-SGD/params": 0.007436879705908062, "lr-SGD/params_no_weight_decay": 0.007436879705908062, "step": 1699}
68
+ {"train_loss": 6.4093804359436035, "profiling/batch_time": 0.12217950075864792, "profiling/data_time": 0.1625451296567917, "epoch": 26, "step": 1699}
69
+ {"lr-SGD/params": 0.007430855661669647, "lr-SGD/params_no_weight_decay": 0.007430855661669647, "step": 1749}
70
+ {"train_loss": 6.012897968292236, "profiling/batch_time": 0.12159913033246994, "profiling/data_time": 0.16415925323963165, "epoch": 27, "step": 1749}
71
+ {"lr-SGD/params": 0.007424559777259163, "lr-SGD/params_no_weight_decay": 0.007424559777259163, "step": 1799}
72
+ {"train_loss": 6.200730323791504, "profiling/batch_time": 0.12452206015586853, "profiling/data_time": 0.053257666528224945, "epoch": 28, "step": 1799}
73
+ {"lr-SGD/params": 0.007417992518117221, "lr-SGD/params_no_weight_decay": 0.007417992518117221, "step": 1849}
74
+ {"train_loss": 5.8828301429748535, "profiling/batch_time": 0.12513107061386108, "profiling/data_time": 0.1645629107952118, "epoch": 29, "step": 1849}
75
+ {"lr-SGD/params": 0.007411154369746556, "lr-SGD/params_no_weight_decay": 0.007411154369746556, "step": 1899}
76
+ {"train_loss": 5.866360187530518, "profiling/batch_time": 0.23916323482990265, "profiling/data_time": 0.1661454290151596, "epoch": 30, "step": 1899}
77
+ {"lr-SGD/params": 0.007404045837676133, "lr-SGD/params_no_weight_decay": 0.007404045837676133, "step": 1949}
78
+ {"train_loss": 5.519585132598877, "profiling/batch_time": 0.24018795788288116, "profiling/data_time": 0.1659889966249466, "epoch": 30, "step": 1949}
79
+ {"lr-SGD/params": 0.007396667447423784, "lr-SGD/params_no_weight_decay": 0.007396667447423784, "step": 1999}
80
+ {"train_loss": 5.702122688293457, "profiling/batch_time": 0.12560497224330902, "profiling/data_time": 0.1669059693813324, "epoch": 31, "step": 1999}
81
+ {"lr-SGD/params": 0.007389019744457344, "lr-SGD/params_no_weight_decay": 0.007389019744457344, "step": 2049}
82
+ {"train_loss": 5.166302680969238, "profiling/batch_time": 0.12627536058425903, "profiling/data_time": 0.15467967092990875, "epoch": 32, "step": 2049}
83
+ {"lr-SGD/params": 0.007381103294154338, "lr-SGD/params_no_weight_decay": 0.007381103294154338, "step": 2099}
84
+ {"train_loss": 5.185615062713623, "profiling/batch_time": 0.12190934270620346, "profiling/data_time": 0.04188748449087143, "epoch": 33, "step": 2099}
85
+ {"lr-SGD/params": 0.007372918681760174, "lr-SGD/params_no_weight_decay": 0.007372918681760174, "step": 2149}
86
+ {"train_loss": 4.717671871185303, "profiling/batch_time": 0.12775127589702606, "profiling/data_time": 0.2718060612678528, "epoch": 34, "step": 2149}
87
+ {"lr-SGD/params": 0.007364466512344889, "lr-SGD/params_no_weight_decay": 0.007364466512344889, "step": 2199}
88
+ {"train_loss": 5.095583438873291, "profiling/batch_time": 0.12620720267295837, "profiling/data_time": 0.1659250557422638, "epoch": 34, "step": 2199}
89
+ {"lr-SGD/params": 0.0073557474107584005, "lr-SGD/params_no_weight_decay": 0.0073557474107584005, "step": 2249}
90
+ {"train_loss": 4.609167575836182, "profiling/batch_time": 0.23927585780620575, "profiling/data_time": 0.1630682647228241, "epoch": 35, "step": 2249}
91
+ {"lr-SGD/params": 0.007346762021584329, "lr-SGD/params_no_weight_decay": 0.007346762021584329, "step": 2299}
92
+ {"train_loss": 5.281973838806152, "profiling/batch_time": 0.12838804721832275, "profiling/data_time": 0.1772765815258026, "epoch": 36, "step": 2299}
93
+ {"lr-SGD/params": 0.00733751100909234, "lr-SGD/params_no_weight_decay": 0.00733751100909234, "step": 2349}
94
+ {"train_loss": 4.254548072814941, "profiling/batch_time": 0.12846630811691284, "profiling/data_time": 0.165502667427063, "epoch": 37, "step": 2349}
95
+ {"lr-SGD/params": 0.007327995057189031, "lr-SGD/params_no_weight_decay": 0.007327995057189031, "step": 2399}
96
+ {"train_loss": 4.933053493499756, "profiling/batch_time": 0.12194191664457321, "profiling/data_time": 0.05121346190571785, "epoch": 38, "step": 2399}
97
+ {"lr-SGD/params": 0.0073182148693673775, "lr-SGD/params_no_weight_decay": 0.0073182148693673775, "step": 2449}
98
+ {"train_loss": 4.780442237854004, "profiling/batch_time": 0.12187426537275314, "profiling/data_time": 0.04487227275967598, "epoch": 38, "step": 2449}
99
+ {"lr-SGD/params": 0.007308171168654722, "lr-SGD/params_no_weight_decay": 0.007308171168654722, "step": 2499}
100
+ {"train_loss": 4.392489910125732, "profiling/batch_time": 0.12598717212677002, "profiling/data_time": 0.1631971299648285, "epoch": 39, "step": 2499}
101
+ {"lr-SGD/params": 0.007297864697559329, "lr-SGD/params_no_weight_decay": 0.007297864697559329, "step": 2549}
102
+ {"train_loss": 4.4024977684021, "profiling/batch_time": 0.24110853672027588, "profiling/data_time": 0.16732177138328552, "epoch": 40, "step": 2549}
103
+ {"lr-SGD/params": 0.00728729621801548, "lr-SGD/params_no_weight_decay": 0.00728729621801548, "step": 2599}
104
+ {"train_loss": 4.066000461578369, "profiling/batch_time": 0.12562309205532074, "profiling/data_time": 0.15693388879299164, "epoch": 41, "step": 2599}
105
+ {"lr-SGD/params": 0.00727646651132716, "lr-SGD/params_no_weight_decay": 0.00727646651132716, "step": 2649}
106
+ {"train_loss": 4.202833652496338, "profiling/batch_time": 0.12632256746292114, "profiling/data_time": 0.16750752925872803, "epoch": 42, "step": 2649}
107
+ {"lr-SGD/params": 0.007265376378110289, "lr-SGD/params_no_weight_decay": 0.007265376378110289, "step": 2699}
108
+ {"train_loss": 3.9481911659240723, "profiling/batch_time": 0.12176384031772614, "profiling/data_time": 0.16737797856330872, "epoch": 42, "step": 2699}
109
+ {"lr-SGD/params": 0.007254026638233534, "lr-SGD/params_no_weight_decay": 0.007254026638233534, "step": 2749}
110
+ {"train_loss": 4.257737159729004, "profiling/batch_time": 0.12523332238197327, "profiling/data_time": 0.051237042993307114, "epoch": 43, "step": 2749}
111
+ {"lr-SGD/params": 0.007242418130757702, "lr-SGD/params_no_weight_decay": 0.007242418130757702, "step": 2799}
112
+ {"train_loss": 4.459207534790039, "profiling/batch_time": 0.12506192922592163, "profiling/data_time": 0.15381234884262085, "epoch": 44, "step": 2799}
113
+ {"lr-SGD/params": 0.007230551713873705, "lr-SGD/params_no_weight_decay": 0.007230551713873705, "step": 2849}
114
+ {"train_loss": 3.7037911415100098, "profiling/batch_time": 0.23931457102298737, "profiling/data_time": 0.1548129767179489, "epoch": 45, "step": 2849}
115
+ {"lr-SGD/params": 0.007218428264839124, "lr-SGD/params_no_weight_decay": 0.007218428264839124, "step": 2899}
116
+ {"train_loss": 4.055521488189697, "profiling/batch_time": 0.12473433464765549, "profiling/data_time": 0.15385253727436066, "epoch": 46, "step": 2899}
117
+ {"lr-SGD/params": 0.007206048679913347, "lr-SGD/params_no_weight_decay": 0.007206048679913347, "step": 2949}
118
+ {"train_loss": 3.862396717071533, "profiling/batch_time": 0.1254001259803772, "profiling/data_time": 0.15390180051326752, "epoch": 46, "step": 2949}
119
+ {"lr-SGD/params": 0.007193413874291313, "lr-SGD/params_no_weight_decay": 0.007193413874291313, "step": 2999}
120
+ {"train_loss": 3.8828861713409424, "profiling/batch_time": 0.12618741393089294, "profiling/data_time": 0.16526778042316437, "epoch": 47, "step": 2999}
121
+ {"lr-SGD/params": 0.007180524782035856, "lr-SGD/params_no_weight_decay": 0.007180524782035856, "step": 3049}
122
+ {"train_loss": 3.7288150787353516, "profiling/batch_time": 0.1334228813648224, "profiling/data_time": 0.05101805552840233, "epoch": 48, "step": 3049}
123
+ {"lr-SGD/params": 0.00716738235600865, "lr-SGD/params_no_weight_decay": 0.00716738235600865, "step": 3099}
124
+ {"train_loss": 4.073267459869385, "profiling/batch_time": 0.12185464799404144, "profiling/data_time": 0.1552484929561615, "epoch": 49, "step": 3099}
125
+ {"lr-SGD/params": 0.007153987567799769, "lr-SGD/params_no_weight_decay": 0.007153987567799769, "step": 3149}
126
+ {"train_loss": 3.858922243118286, "profiling/batch_time": 0.23954914510250092, "profiling/data_time": 0.1687803864479065, "epoch": 49, "step": 3149}
127
+ {"lr-SGD/params": 0.007140341407655851, "lr-SGD/params_no_weight_decay": 0.007140341407655851, "step": 3199}
128
+ {"train_loss": 3.2948713302612305, "profiling/batch_time": 0.24121783673763275, "profiling/data_time": 0.15863880515098572, "epoch": 50, "step": 3199}
129
+ {"lr-SGD/params": 0.007126444884406901, "lr-SGD/params_no_weight_decay": 0.007126444884406901, "step": 3249}
130
+ {"train_loss": 3.5028045177459717, "profiling/batch_time": 0.12643703818321228, "profiling/data_time": 0.16260233521461487, "epoch": 51, "step": 3249}
131
+ {"lr-SGD/params": 0.007112299025391709, "lr-SGD/params_no_weight_decay": 0.007112299025391709, "step": 3299}
132
+ {"train_loss": 3.8911869525909424, "profiling/batch_time": 0.12256395816802979, "profiling/data_time": 0.17034468054771423, "epoch": 52, "step": 3299}
133
+ {"lr-SGD/params": 0.007097904876381894, "lr-SGD/params_no_weight_decay": 0.007097904876381894, "step": 3349}
134
+ {"train_loss": 3.9126901626586914, "profiling/batch_time": 0.12197363376617432, "profiling/data_time": 0.04095682129263878, "epoch": 53, "step": 3349}
135
+ {"lr-SGD/params": 0.007083263501504598, "lr-SGD/params_no_weight_decay": 0.007083263501504598, "step": 3399}
136
+ {"train_loss": 3.901865243911743, "profiling/batch_time": 0.12464418262243271, "profiling/data_time": 0.044469207525253296, "epoch": 53, "step": 3399}
137
+ {"lr-SGD/params": 0.007068375983163821, "lr-SGD/params_no_weight_decay": 0.007068375983163821, "step": 3449}
138
+ {"train_loss": 3.745466709136963, "profiling/batch_time": 0.12532757222652435, "profiling/data_time": 0.16516725718975067, "epoch": 54, "step": 3449}
139
+ {"lr-SGD/params": 0.0070532434219603935, "lr-SGD/params_no_weight_decay": 0.0070532434219603935, "step": 3499}
140
+ {"train_loss": 3.8354177474975586, "profiling/batch_time": 0.23973442614078522, "profiling/data_time": 0.16257204115390778, "epoch": 55, "step": 3499}
141
+ {"lr-SGD/params": 0.007037866936610616, "lr-SGD/params_no_weight_decay": 0.007037866936610616, "step": 3549}
142
+ {"train_loss": 3.5368967056274414, "profiling/batch_time": 0.12424282729625702, "profiling/data_time": 0.16577082872390747, "epoch": 56, "step": 3549}
143
+ {"lr-SGD/params": 0.00702224766386355, "lr-SGD/params_no_weight_decay": 0.00702224766386355, "step": 3599}
144
+ {"train_loss": 3.319094181060791, "profiling/batch_time": 0.12782083451747894, "profiling/data_time": 0.15523463487625122, "epoch": 57, "step": 3599}
145
+ {"lr-SGD/params": 0.007006386758416992, "lr-SGD/params_no_weight_decay": 0.007006386758416992, "step": 3649}
146
+ {"train_loss": 3.165205955505371, "profiling/batch_time": 0.12172437459230423, "profiling/data_time": 0.1567973643541336, "epoch": 57, "step": 3649}
147
+ {"lr-SGD/params": 0.006990285392832098, "lr-SGD/params_no_weight_decay": 0.006990285392832098, "step": 3699}
148
+ {"train_loss": 3.2156503200531006, "profiling/batch_time": 0.12273900210857391, "profiling/data_time": 0.04865007847547531, "epoch": 58, "step": 3699}
149
+ {"lr-SGD/params": 0.006973944757446702, "lr-SGD/params_no_weight_decay": 0.006973944757446702, "step": 3749}
150
+ {"train_loss": 3.5502445697784424, "profiling/batch_time": 0.12508974969387054, "profiling/data_time": 0.15490518510341644, "epoch": 59, "step": 3749}
151
+ {"lr-SGD/params": 0.006957366060287316, "lr-SGD/params_no_weight_decay": 0.006957366060287316, "step": 3799}
152
+ {"train_loss": 2.756619691848755, "profiling/batch_time": 0.24003475904464722, "profiling/data_time": 0.15742583572864532, "epoch": 60, "step": 3799}
153
+ {"lr-SGD/params": 0.006940550526979831, "lr-SGD/params_no_weight_decay": 0.006940550526979831, "step": 3849}
154
+ {"train_loss": 3.1995325088500977, "profiling/batch_time": 0.12537963688373566, "profiling/data_time": 0.15595999360084534, "epoch": 61, "step": 3849}
155
+ {"lr-SGD/params": 0.006923499400658897, "lr-SGD/params_no_weight_decay": 0.006923499400658897, "step": 3899}
156
+ {"train_loss": 3.275684118270874, "profiling/batch_time": 0.1292886883020401, "profiling/data_time": 0.17509892582893372, "epoch": 61, "step": 3899}
157
+ {"lr-SGD/params": 0.006906213941876028, "lr-SGD/params_no_weight_decay": 0.006906213941876028, "step": 3949}
158
+ {"train_loss": 3.3230130672454834, "profiling/batch_time": 0.12572208046913147, "profiling/data_time": 0.1650657057762146, "epoch": 62, "step": 3949}
159
+ {"lr-SGD/params": 0.006888695428506416, "lr-SGD/params_no_weight_decay": 0.006888695428506416, "step": 3999}
160
+ {"train_loss": 3.107393980026245, "profiling/batch_time": 0.12234047055244446, "profiling/data_time": 0.042872194200754166, "epoch": 63, "step": 3999}
161
+ {"lr-SGD/params": 0.006870945155654448, "lr-SGD/params_no_weight_decay": 0.006870945155654448, "step": 4049}
162
+ {"train_loss": 2.6827454566955566, "profiling/batch_time": 0.12236487120389938, "profiling/data_time": 0.16012485325336456, "epoch": 64, "step": 4049}
163
+ {"lr-SGD/params": 0.0068529644355579756, "lr-SGD/params_no_weight_decay": 0.0068529644355579756, "step": 4099}
164
+ {"train_loss": 2.5986804962158203, "profiling/batch_time": 0.23998305201530457, "profiling/data_time": 0.21450009942054749, "epoch": 65, "step": 4099}
165
+ {"lr-SGD/params": 0.006834754597491295, "lr-SGD/params_no_weight_decay": 0.006834754597491295, "step": 4149}
166
+ {"train_loss": 2.3342790603637695, "profiling/batch_time": 0.23922359943389893, "profiling/data_time": 0.16149231791496277, "epoch": 65, "step": 4149}
167
+ {"lr-SGD/params": 0.006816316987666884, "lr-SGD/params_no_weight_decay": 0.006816316987666884, "step": 4199}
168
+ {"train_loss": 2.6243972778320312, "profiling/batch_time": 0.1260385364294052, "profiling/data_time": 0.16823925077915192, "epoch": 66, "step": 4199}
169
+ {"lr-SGD/params": 0.006797652969135869, "lr-SGD/params_no_weight_decay": 0.006797652969135869, "step": 4249}
170
+ {"train_loss": 2.66534161567688, "profiling/batch_time": 0.12255872040987015, "profiling/data_time": 0.17120258510112762, "epoch": 67, "step": 4249}
171
+ {"lr-SGD/params": 0.006778763921687271, "lr-SGD/params_no_weight_decay": 0.006778763921687271, "step": 4299}
172
+ {"train_loss": 2.5044054985046387, "profiling/batch_time": 0.12248821556568146, "profiling/data_time": 0.04042670875787735, "epoch": 68, "step": 4299}
173
+ {"lr-SGD/params": 0.006759651241745989, "lr-SGD/params_no_weight_decay": 0.006759651241745989, "step": 4349}
174
+ {"train_loss": 2.429180860519409, "profiling/batch_time": 0.12482256442308426, "profiling/data_time": 0.16238990426063538, "epoch": 69, "step": 4349}
175
+ {"lr-SGD/params": 0.006740316342269573, "lr-SGD/params_no_weight_decay": 0.006740316342269573, "step": 4399}
176
+ {"train_loss": 2.7154386043548584, "profiling/batch_time": 0.12172514200210571, "profiling/data_time": 0.1687832772731781, "epoch": 69, "step": 4399}
177
+ {"lr-SGD/params": 0.006720760652643761, "lr-SGD/params_no_weight_decay": 0.006720760652643761, "step": 4449}
178
+ {"train_loss": 2.3989052772521973, "profiling/batch_time": 0.24026916921138763, "profiling/data_time": 0.17305435240268707, "epoch": 70, "step": 4449}
179
+ {"lr-SGD/params": 0.006700985618576818, "lr-SGD/params_no_weight_decay": 0.006700985618576818, "step": 4499}
180
+ {"train_loss": 2.5755796432495117, "profiling/batch_time": 0.1260344386100769, "profiling/data_time": 0.17132513225078583, "epoch": 71, "step": 4499}
181
+ {"lr-SGD/params": 0.006680992701992646, "lr-SGD/params_no_weight_decay": 0.006680992701992646, "step": 4549}
182
+ {"train_loss": 2.122117280960083, "profiling/batch_time": 0.12197041511535645, "profiling/data_time": 0.15777981281280518, "epoch": 72, "step": 4549}
183
+ {"lr-SGD/params": 0.006660783380922715, "lr-SGD/params_no_weight_decay": 0.006660783380922715, "step": 4599}
184
+ {"train_loss": 2.1432361602783203, "profiling/batch_time": 0.1337031126022339, "profiling/data_time": 5.227355003356934, "epoch": 73, "step": 4599}
185
+ {"lr-SGD/params": 0.006640359149396792, "lr-SGD/params_no_weight_decay": 0.006640359149396792, "step": 4649}
186
+ {"train_loss": 2.16202712059021, "profiling/batch_time": 0.12318161875009537, "profiling/data_time": 0.042044322937726974, "epoch": 73, "step": 4649}
187
+ {"lr-SGD/params": 0.006619721517332492, "lr-SGD/params_no_weight_decay": 0.006619721517332492, "step": 4699}
188
+ {"train_loss": 2.140075206756592, "profiling/batch_time": 0.12178565561771393, "profiling/data_time": 0.17188653349876404, "epoch": 74, "step": 4699}
189
+ {"lr-SGD/params": 0.006598872010423656, "lr-SGD/params_no_weight_decay": 0.006598872010423656, "step": 4749}
190
+ {"train_loss": 2.32617449760437, "profiling/batch_time": 0.2398069202899933, "profiling/data_time": 0.15444783866405487, "epoch": 75, "step": 4749}
191
+ {"lr-SGD/params": 0.0065778121700275525, "lr-SGD/params_no_weight_decay": 0.0065778121700275525, "step": 4799}
192
+ {"train_loss": 2.062258720397949, "profiling/batch_time": 0.12545515596866608, "profiling/data_time": 0.1673913449048996, "epoch": 76, "step": 4799}
193
+ {"lr-SGD/params": 0.006556543553050934, "lr-SGD/params_no_weight_decay": 0.006556543553050934, "step": 4849}
194
+ {"train_loss": 2.129628896713257, "profiling/batch_time": 0.12125606834888458, "profiling/data_time": 0.16046760976314545, "epoch": 76, "step": 4849}
195
+ {"lr-SGD/params": 0.0065350677318349384, "lr-SGD/params_no_weight_decay": 0.0065350677318349384, "step": 4899}
196
+ {"train_loss": 1.8616584539413452, "profiling/batch_time": 0.12113004922866821, "profiling/data_time": 0.16874299943447113, "epoch": 77, "step": 4899}
197
+ {"lr-SGD/params": 0.006513386294038849, "lr-SGD/params_no_weight_decay": 0.006513386294038849, "step": 4949}
198
+ {"train_loss": 1.9525800943374634, "profiling/batch_time": 0.12064872682094574, "profiling/data_time": 0.04737721383571625, "epoch": 78, "step": 4949}
199
+ {"lr-SGD/params": 0.00649150084252272, "lr-SGD/params_no_weight_decay": 0.00649150084252272, "step": 4999}
200
+ {"train_loss": 2.091287136077881, "profiling/batch_time": 0.12185035645961761, "profiling/data_time": 0.1545131504535675, "epoch": 79, "step": 4999}
201
+ {"lr-SGD/params": 0.006469412995228881, "lr-SGD/params_no_weight_decay": 0.006469412995228881, "step": 5049}
202
+ {"train_loss": 1.947624683380127, "profiling/batch_time": 0.24289089441299438, "profiling/data_time": 0.15485644340515137, "epoch": 80, "step": 5049}
203
+ {"lr-SGD/params": 0.00644712438506233, "lr-SGD/params_no_weight_decay": 0.00644712438506233, "step": 5099}
204
+ {"train_loss": 2.0774946212768555, "profiling/batch_time": 0.23965950310230255, "profiling/data_time": 0.16540731489658356, "epoch": 80, "step": 5099}
205
+ {"lr-SGD/params": 0.006424636659770012, "lr-SGD/params_no_weight_decay": 0.006424636659770012, "step": 5149}
206
+ {"train_loss": 2.032754421234131, "profiling/batch_time": 0.12398013472557068, "profiling/data_time": 0.16268084943294525, "epoch": 81, "step": 5149}
207
+ {"lr-SGD/params": 0.006401951481819008, "lr-SGD/params_no_weight_decay": 0.006401951481819008, "step": 5199}
208
+ {"train_loss": 2.2514262199401855, "profiling/batch_time": 0.12503038346767426, "profiling/data_time": 0.15654432773590088, "epoch": 82, "step": 5199}
209
+ {"lr-SGD/params": 0.006379070528273629, "lr-SGD/params_no_weight_decay": 0.006379070528273629, "step": 5249}
210
+ {"train_loss": 1.964591383934021, "profiling/batch_time": 0.12446270883083344, "profiling/data_time": 0.051656197756528854, "epoch": 83, "step": 5249}
211
+ {"lr-SGD/params": 0.0063559954906714415, "lr-SGD/params_no_weight_decay": 0.0063559954906714415, "step": 5299}
212
+ {"train_loss": 1.9732400178909302, "profiling/batch_time": 0.12622319161891937, "profiling/data_time": 0.15375757217407227, "epoch": 84, "step": 5299}
213
+ {"lr-SGD/params": 0.006332728074898201, "lr-SGD/params_no_weight_decay": 0.006332728074898201, "step": 5349}
214
+ {"train_loss": 1.8740417957305908, "profiling/batch_time": 0.12153571099042892, "profiling/data_time": 0.16127116978168488, "epoch": 84, "step": 5349}
215
+ {"lr-SGD/params": 0.006309270001061761, "lr-SGD/params_no_weight_decay": 0.006309270001061761, "step": 5399}
216
+ {"train_loss": 1.7343707084655762, "profiling/batch_time": 0.23947995901107788, "profiling/data_time": 0.16633133590221405, "epoch": 85, "step": 5399}
217
+ {"lr-SGD/params": 0.0062856230033648906, "lr-SGD/params_no_weight_decay": 0.0062856230033648906, "step": 5449}
218
+ {"train_loss": 1.835343599319458, "profiling/batch_time": 0.12444368004798889, "profiling/data_time": 0.1633024662733078, "epoch": 86, "step": 5449}
219
+ {"lr-SGD/params": 0.006261788829977077, "lr-SGD/params_no_weight_decay": 0.006261788829977077, "step": 5499}
220
+ {"train_loss": 1.7862977981567383, "profiling/batch_time": 0.12116380780935287, "profiling/data_time": 0.1565333902835846, "epoch": 87, "step": 5499}
221
+ {"lr-SGD/params": 0.006237769242905288, "lr-SGD/params_no_weight_decay": 0.006237769242905288, "step": 5549}
222
+ {"train_loss": 1.9258463382720947, "profiling/batch_time": 0.13193269073963165, "profiling/data_time": 0.046793967485427856, "epoch": 88, "step": 5549}
223
+ {"lr-SGD/params": 0.006213566017863711, "lr-SGD/params_no_weight_decay": 0.006213566017863711, "step": 5599}
224
+ {"train_loss": 1.8566666841506958, "profiling/batch_time": 0.12357773631811142, "profiling/data_time": 0.041973695158958435, "epoch": 88, "step": 5599}
225
+ {"lr-SGD/params": 0.00618918094414247, "lr-SGD/params_no_weight_decay": 0.00618918094414247, "step": 5649}
226
+ {"train_loss": 1.920528531074524, "profiling/batch_time": 0.12205547094345093, "profiling/data_time": 0.1584828644990921, "epoch": 89, "step": 5649}
227
+ {"lr-SGD/params": 0.00616461582447536, "lr-SGD/params_no_weight_decay": 0.00616461582447536, "step": 5699}
228
+ {"train_loss": 1.7513959407806396, "profiling/batch_time": 0.23963358998298645, "profiling/data_time": 0.1715700924396515, "epoch": 90, "step": 5699}
229
+ {"lr-SGD/params": 0.0061398724749065665, "lr-SGD/params_no_weight_decay": 0.0061398724749065665, "step": 5749}
230
+ {"train_loss": 1.7335903644561768, "profiling/batch_time": 0.13615304231643677, "profiling/data_time": 0.1596328616142273, "epoch": 91, "step": 5749}
231
+ {"lr-SGD/params": 0.006114952724656413, "lr-SGD/params_no_weight_decay": 0.006114952724656413, "step": 5799}
232
+ {"train_loss": 1.740799069404602, "profiling/batch_time": 0.12463821470737457, "profiling/data_time": 0.16478104889392853, "epoch": 92, "step": 5799}
233
+ {"lr-SGD/params": 0.006089858415986126, "lr-SGD/params_no_weight_decay": 0.006089858415986126, "step": 5849}
234
+ {"train_loss": 1.9332733154296875, "profiling/batch_time": 0.12445712834596634, "profiling/data_time": 0.16190436482429504, "epoch": 92, "step": 5849}
235
+ {"lr-SGD/params": 0.006064591404061649, "lr-SGD/params_no_weight_decay": 0.006064591404061649, "step": 5899}
236
+ {"train_loss": 1.757057785987854, "profiling/batch_time": 0.1331460326910019, "profiling/data_time": 0.048784028738737106, "epoch": 93, "step": 5899}
237
+ {"lr-SGD/params": 0.006039153556816488, "lr-SGD/params_no_weight_decay": 0.006039153556816488, "step": 5949}
238
+ {"train_loss": 1.6725032329559326, "profiling/batch_time": 0.12199380248785019, "profiling/data_time": 0.16915298998355865, "epoch": 94, "step": 5949}
239
+ {"lr-SGD/params": 0.00601354675481362, "lr-SGD/params_no_weight_decay": 0.00601354675481362, "step": 5999}
240
+ {"train_loss": 1.892687201499939, "profiling/batch_time": 0.23888127505779266, "profiling/data_time": 0.15934455394744873, "epoch": 95, "step": 5999}
241
+ {"lr-SGD/params": 0.005987772891106467, "lr-SGD/params_no_weight_decay": 0.005987772891106467, "step": 6049}
242
+ {"train_loss": 1.804595947265625, "profiling/batch_time": 0.1262298822402954, "profiling/data_time": 0.19894912838935852, "epoch": 96, "step": 6049}
243
+ {"lr-SGD/params": 0.005961833871098953, "lr-SGD/params_no_weight_decay": 0.005961833871098953, "step": 6099}
244
+ {"train_loss": 1.8115068674087524, "profiling/batch_time": 0.12218761444091797, "profiling/data_time": 0.1717042326927185, "epoch": 96, "step": 6099}
245
+ {"lr-SGD/params": 0.005935731612404632, "lr-SGD/params_no_weight_decay": 0.005935731612404632, "step": 6149}
246
+ {"train_loss": 1.7011313438415527, "profiling/batch_time": 0.12352977693080902, "profiling/data_time": 0.15455465018749237, "epoch": 97, "step": 6149}
247
+ {"lr-SGD/params": 0.005909468044704928, "lr-SGD/params_no_weight_decay": 0.005909468044704928, "step": 6199}
248
+ {"train_loss": 1.7410836219787598, "profiling/batch_time": 0.12477803230285645, "profiling/data_time": 0.04145502671599388, "epoch": 98, "step": 6199}
249
+ {"lr-SGD/params": 0.0058830451096064836, "lr-SGD/params_no_weight_decay": 0.0058830451096064836, "step": 6249}
250
+ {"train_loss": 1.5777922868728638, "profiling/batch_time": 0.12288759648799896, "profiling/data_time": 0.16047824919223785, "epoch": 99, "step": 6249}
251
+ {"lr-SGD/params": 0.005856464760497606, "lr-SGD/params_no_weight_decay": 0.005856464760497606, "step": 6299}
252
+ {"train_loss": 1.847308874130249, "profiling/batch_time": 0.23947814106941223, "profiling/data_time": 0.15654970705509186, "epoch": 99, "step": 6299}
253
+ {"lr-SGD/params": 0.005829728962403877, "lr-SGD/params_no_weight_decay": 0.005829728962403877, "step": 6349}
254
+ {"train_loss": 1.6915602684020996, "profiling/batch_time": 0.23991261422634125, "profiling/data_time": 0.15230335295200348, "epoch": 100, "step": 6349}
255
+ {"lr-SGD/params": 0.005802839691842867, "lr-SGD/params_no_weight_decay": 0.005802839691842867, "step": 6399}
256
+ {"train_loss": 1.7861181497573853, "profiling/batch_time": 0.12486816942691803, "profiling/data_time": 0.16950908303260803, "epoch": 101, "step": 6399}
257
+ {"lr-SGD/params": 0.005775798936678021, "lr-SGD/params_no_weight_decay": 0.005775798936678021, "step": 6449}
258
+ {"train_loss": 1.8448026180267334, "profiling/batch_time": 0.12095407396554947, "profiling/data_time": 0.15393991768360138, "epoch": 102, "step": 6449}
259
+ {"lr-SGD/params": 0.005748608695971705, "lr-SGD/params_no_weight_decay": 0.005748608695971705, "step": 6499}
260
+ {"train_loss": 1.822641372680664, "profiling/batch_time": 0.12155602127313614, "profiling/data_time": 0.04159587621688843, "epoch": 103, "step": 6499}
261
+ {"lr-SGD/params": 0.005721270979837415, "lr-SGD/params_no_weight_decay": 0.005721270979837415, "step": 6549}
262
+ {"train_loss": 1.7197613716125488, "profiling/batch_time": 0.1281646490097046, "profiling/data_time": 0.06978949159383774, "epoch": 103, "step": 6549}
263
+ {"lr-SGD/params": 0.005693787809291172, "lr-SGD/params_no_weight_decay": 0.005693787809291172, "step": 6599}
264
+ {"train_loss": 1.5870652198791504, "profiling/batch_time": 0.12409145385026932, "profiling/data_time": 0.16529318690299988, "epoch": 104, "step": 6599}
265
+ {"lr-SGD/params": 0.005666161216102115, "lr-SGD/params_no_weight_decay": 0.005666161216102115, "step": 6649}
266
+ {"train_loss": 1.5961320400238037, "profiling/batch_time": 0.2398846447467804, "profiling/data_time": 0.16727274656295776, "epoch": 105, "step": 6649}
267
+ {"lr-SGD/params": 0.005638393242642302, "lr-SGD/params_no_weight_decay": 0.005638393242642302, "step": 6699}
268
+ {"train_loss": 1.535551905632019, "profiling/batch_time": 0.12464461475610733, "profiling/data_time": 0.15617045760154724, "epoch": 106, "step": 6699}
269
+ {"lr-SGD/params": 0.005610485941735712, "lr-SGD/params_no_weight_decay": 0.005610485941735712, "step": 6749}
270
+ {"train_loss": 1.5164239406585693, "profiling/batch_time": 0.12378984689712524, "profiling/data_time": 0.15385523438453674, "epoch": 107, "step": 6749}
271
+ {"lr-SGD/params": 0.005582441376506492, "lr-SGD/params_no_weight_decay": 0.005582441376506492, "step": 6799}
272
+ {"train_loss": 1.4952605962753296, "profiling/batch_time": 0.12731939554214478, "profiling/data_time": 0.1671607494354248, "epoch": 107, "step": 6799}
273
+ {"lr-SGD/params": 0.005554261620226433, "lr-SGD/params_no_weight_decay": 0.005554261620226433, "step": 6849}
274
+ {"train_loss": 1.6715298891067505, "profiling/batch_time": 0.1251714676618576, "profiling/data_time": 0.05284680798649788, "epoch": 108, "step": 6849}
275
+ {"lr-SGD/params": 0.005525948756161695, "lr-SGD/params_no_weight_decay": 0.005525948756161695, "step": 6899}
276
+ {"train_loss": 1.4934823513031006, "profiling/batch_time": 0.12339124083518982, "profiling/data_time": 0.1594378650188446, "epoch": 109, "step": 6899}
277
+ {"lr-SGD/params": 0.005497504877418799, "lr-SGD/params_no_weight_decay": 0.005497504877418799, "step": 6949}
278
+ {"train_loss": 1.6301673650741577, "profiling/batch_time": 0.23957334458827972, "profiling/data_time": 0.15228965878486633, "epoch": 110, "step": 6949}
279
+ {"lr-SGD/params": 0.005468932086789889, "lr-SGD/params_no_weight_decay": 0.005468932086789889, "step": 6999}
280
+ {"train_loss": 1.6586252450942993, "profiling/batch_time": 0.12420568615198135, "profiling/data_time": 0.15781699120998383, "epoch": 111, "step": 6999}
281
+ {"lr-SGD/params": 0.005440232496597272, "lr-SGD/params_no_weight_decay": 0.005440232496597272, "step": 7049}
282
+ {"train_loss": 1.5173407793045044, "profiling/batch_time": 0.12420881539583206, "profiling/data_time": 0.15578599274158478, "epoch": 111, "step": 7049}
283
+ {"lr-SGD/params": 0.005411408228537265, "lr-SGD/params_no_weight_decay": 0.005411408228537265, "step": 7099}
284
+ {"train_loss": 1.2476837635040283, "profiling/batch_time": 0.12452506273984909, "profiling/data_time": 0.16678817570209503, "epoch": 112, "step": 7099}
285
+ {"lr-SGD/params": 0.0053824614135233395, "lr-SGD/params_no_weight_decay": 0.0053824614135233395, "step": 7149}
286
+ {"train_loss": 1.421595811843872, "profiling/batch_time": 0.12126931548118591, "profiling/data_time": 0.05181608721613884, "epoch": 113, "step": 7149}
287
+ {"lr-SGD/params": 0.005353394191528586, "lr-SGD/params_no_weight_decay": 0.005353394191528586, "step": 7199}
288
+ {"train_loss": 1.379381537437439, "profiling/batch_time": 0.12802299857139587, "profiling/data_time": 0.16948504745960236, "epoch": 114, "step": 7199}
289
+ {"lr-SGD/params": 0.005324208711427517, "lr-SGD/params_no_weight_decay": 0.005324208711427517, "step": 7249}
290
+ {"train_loss": 1.525693416595459, "profiling/batch_time": 0.23959757387638092, "profiling/data_time": 0.1590050458908081, "epoch": 115, "step": 7249}
291
+ {"lr-SGD/params": 0.005294907130837198, "lr-SGD/params_no_weight_decay": 0.005294907130837198, "step": 7299}
292
+ {"train_loss": 1.4469754695892334, "profiling/batch_time": 0.23960289359092712, "profiling/data_time": 0.1656365990638733, "epoch": 115, "step": 7299}
293
+ {"lr-SGD/params": 0.005265491615957747, "lr-SGD/params_no_weight_decay": 0.005265491615957747, "step": 7349}
294
+ {"train_loss": 1.7286646366119385, "profiling/batch_time": 0.1261339783668518, "profiling/data_time": 0.15661267936229706, "epoch": 116, "step": 7349}
295
+ {"lr-SGD/params": 0.005235964341412183, "lr-SGD/params_no_weight_decay": 0.005235964341412183, "step": 7399}
296
+ {"train_loss": 1.6178288459777832, "profiling/batch_time": 0.12085632234811783, "profiling/data_time": 0.17120996117591858, "epoch": 117, "step": 7399}
297
+ {"lr-SGD/params": 0.005206327490085672, "lr-SGD/params_no_weight_decay": 0.005206327490085672, "step": 7449}
298
+ {"train_loss": 1.5336967706680298, "profiling/batch_time": 0.12082191556692123, "profiling/data_time": 0.048440925776958466, "epoch": 118, "step": 7449}
299
+ {"lr-SGD/params": 0.005176583252964141, "lr-SGD/params_no_weight_decay": 0.005176583252964141, "step": 7499}
300
+ {"train_loss": 1.3102774620056152, "profiling/batch_time": 0.12395087629556656, "profiling/data_time": 0.1557862013578415, "epoch": 119, "step": 7499}
301
+ {"lr-SGD/params": 0.005146733828972312, "lr-SGD/params_no_weight_decay": 0.005146733828972312, "step": 7549}
302
+ {"train_loss": 1.7302124500274658, "profiling/batch_time": 0.12202747166156769, "profiling/data_time": 0.16167087852954865, "epoch": 119, "step": 7549}
303
+ {"lr-SGD/params": 0.0051167814248111315, "lr-SGD/params_no_weight_decay": 0.0051167814248111315, "step": 7599}
304
+ {"train_loss": 1.5831239223480225, "profiling/batch_time": 0.23971426486968994, "profiling/data_time": 0.1552162915468216, "epoch": 120, "step": 7599}
305
+ {"lr-SGD/params": 0.005086728254794644, "lr-SGD/params_no_weight_decay": 0.005086728254794644, "step": 7649}
306
+ {"train_loss": 1.4333643913269043, "profiling/batch_time": 0.12594307959079742, "profiling/data_time": 0.17010359466075897, "epoch": 121, "step": 7649}
307
+ {"lr-SGD/params": 0.005056576540686285, "lr-SGD/params_no_weight_decay": 0.005056576540686285, "step": 7699}
308
+ {"train_loss": 1.6398762464523315, "profiling/batch_time": 0.12704293429851532, "profiling/data_time": 0.15588811039924622, "epoch": 122, "step": 7699}
309
+ {"lr-SGD/params": 0.0050263285115346315, "lr-SGD/params_no_weight_decay": 0.0050263285115346315, "step": 7749}
310
+ {"train_loss": 1.5773073434829712, "profiling/batch_time": 0.13133206963539124, "profiling/data_time": 5.1879777908325195, "epoch": 123, "step": 7749}
311
+ {"lr-SGD/params": 0.0049959864035086215, "lr-SGD/params_no_weight_decay": 0.0049959864035086215, "step": 7799}
312
+ {"train_loss": 1.4757336378097534, "profiling/batch_time": 0.1218353882431984, "profiling/data_time": 0.050299763679504395, "epoch": 123, "step": 7799}
313
+ {"lr-SGD/params": 0.004965552459732227, "lr-SGD/params_no_weight_decay": 0.004965552459732227, "step": 7849}
314
+ {"train_loss": 1.454671025276184, "profiling/batch_time": 0.12159032374620438, "profiling/data_time": 0.1629103273153305, "epoch": 124, "step": 7849}
315
+ {"lr-SGD/params": 0.004935028930118634, "lr-SGD/params_no_weight_decay": 0.004935028930118634, "step": 7899}
316
+ {"train_loss": 1.3838614225387573, "profiling/batch_time": 0.23955205082893372, "profiling/data_time": 0.167230486869812, "epoch": 125, "step": 7899}
317
+ {"lr-SGD/params": 0.004904418071203909, "lr-SGD/params_no_weight_decay": 0.004904418071203909, "step": 7949}
318
+ {"train_loss": 1.4470468759536743, "profiling/batch_time": 0.1253996640443802, "profiling/data_time": 0.16124014556407928, "epoch": 126, "step": 7949}
319
+ {"lr-SGD/params": 0.004873722145980177, "lr-SGD/params_no_weight_decay": 0.004873722145980177, "step": 7999}
320
+ {"train_loss": 1.4847220182418823, "profiling/batch_time": 0.12445023655891418, "profiling/data_time": 0.1553075760602951, "epoch": 126, "step": 7999}
321
+ {"lr-SGD/params": 0.004842943423728322, "lr-SGD/params_no_weight_decay": 0.004842943423728322, "step": 8049}
322
+ {"train_loss": 1.5280784368515015, "profiling/batch_time": 0.12555496394634247, "profiling/data_time": 0.1673879474401474, "epoch": 127, "step": 8049}
323
+ {"lr-SGD/params": 0.00481208417985023, "lr-SGD/params_no_weight_decay": 0.00481208417985023, "step": 8099}
324
+ {"train_loss": 1.4109729528427124, "profiling/batch_time": 0.12124310433864594, "profiling/data_time": 0.052055392414331436, "epoch": 128, "step": 8099}
325
+ {"lr-SGD/params": 0.004781146695700571, "lr-SGD/params_no_weight_decay": 0.004781146695700571, "step": 8149}
326
+ {"train_loss": 1.3554214239120483, "profiling/batch_time": 0.12277916073799133, "profiling/data_time": 0.15650632977485657, "epoch": 129, "step": 8149}
327
+ {"lr-SGD/params": 0.004750133258418137, "lr-SGD/params_no_weight_decay": 0.004750133258418137, "step": 8199}
328
+ {"train_loss": 1.5885064601898193, "profiling/batch_time": 0.24480178952217102, "profiling/data_time": 0.18648438155651093, "epoch": 130, "step": 8199}
329
+ {"lr-SGD/params": 0.004719046160756771, "lr-SGD/params_no_weight_decay": 0.004719046160756771, "step": 8249}
330
+ {"train_loss": 1.3516334295272827, "profiling/batch_time": 0.23952403664588928, "profiling/data_time": 0.15481053292751312, "epoch": 130, "step": 8249}
331
+ {"lr-SGD/params": 0.004687887700915859, "lr-SGD/params_no_weight_decay": 0.004687887700915859, "step": 8299}
332
+ {"train_loss": 1.439639687538147, "profiling/batch_time": 0.1236371174454689, "profiling/data_time": 0.1686955690383911, "epoch": 131, "step": 8299}
333
+ {"lr-SGD/params": 0.004656660182370435, "lr-SGD/params_no_weight_decay": 0.004656660182370435, "step": 8349}
334
+ {"train_loss": 1.2816903591156006, "profiling/batch_time": 0.12371913343667984, "profiling/data_time": 0.17110076546669006, "epoch": 132, "step": 8349}
335
+ {"lr-SGD/params": 0.004625365913700883, "lr-SGD/params_no_weight_decay": 0.004625365913700883, "step": 8399}
336
+ {"train_loss": 1.5233423709869385, "profiling/batch_time": 0.12435373663902283, "profiling/data_time": 0.04852976277470589, "epoch": 133, "step": 8399}
337
+ {"lr-SGD/params": 0.004594007208422283, "lr-SGD/params_no_weight_decay": 0.004594007208422283, "step": 8449}
338
+ {"train_loss": 1.5310571193695068, "profiling/batch_time": 0.1207302138209343, "profiling/data_time": 0.15801748633384705, "epoch": 134, "step": 8449}
339
+ {"lr-SGD/params": 0.004562586384813361, "lr-SGD/params_no_weight_decay": 0.004562586384813361, "step": 8499}
340
+ {"train_loss": 1.668411135673523, "profiling/batch_time": 0.1223101019859314, "profiling/data_time": 0.16663850843906403, "epoch": 134, "step": 8499}
341
+ {"lr-SGD/params": 0.004531105765745115, "lr-SGD/params_no_weight_decay": 0.004531105765745115, "step": 8549}
342
+ {"train_loss": 1.4525620937347412, "profiling/batch_time": 0.2403763383626938, "profiling/data_time": 0.16669276356697083, "epoch": 135, "step": 8549}
343
+ {"lr-SGD/params": 0.004499567678509085, "lr-SGD/params_no_weight_decay": 0.004499567678509085, "step": 8599}
344
+ {"train_loss": 1.23919677734375, "profiling/batch_time": 0.1249888464808464, "profiling/data_time": 0.17010498046875, "epoch": 136, "step": 8599}
345
+ {"lr-SGD/params": 0.004467974454645308, "lr-SGD/params_no_weight_decay": 0.004467974454645308, "step": 8649}
346
+ {"train_loss": 1.3331170082092285, "profiling/batch_time": 0.12641313672065735, "profiling/data_time": 0.15374232828617096, "epoch": 137, "step": 8649}
347
+ {"lr-SGD/params": 0.004436328429769944, "lr-SGD/params_no_weight_decay": 0.004436328429769944, "step": 8699}
348
+ {"train_loss": 1.3022950887680054, "profiling/batch_time": 0.1255551427602768, "profiling/data_time": 0.04483673721551895, "epoch": 138, "step": 8699}
349
+ {"lr-SGD/params": 0.004404631943402617, "lr-SGD/params_no_weight_decay": 0.004404631943402617, "step": 8749}
350
+ {"train_loss": 1.2949864864349365, "profiling/batch_time": 0.12489961087703705, "profiling/data_time": 0.05263261869549751, "epoch": 138, "step": 8749}
351
+ {"lr-SGD/params": 0.004372887338793451, "lr-SGD/params_no_weight_decay": 0.004372887338793451, "step": 8799}
352
+ {"train_loss": 1.5909953117370605, "profiling/batch_time": 0.12548956274986267, "profiling/data_time": 0.1735186129808426, "epoch": 139, "step": 8799}
353
+ {"lr-SGD/params": 0.004341096962749849, "lr-SGD/params_no_weight_decay": 0.004341096962749849, "step": 8849}
354
+ {"train_loss": 1.3221890926361084, "profiling/batch_time": 0.24088837206363678, "profiling/data_time": 0.1671207994222641, "epoch": 140, "step": 8849}
355
+ {"lr-SGD/params": 0.004309263165462988, "lr-SGD/params_no_weight_decay": 0.004309263165462988, "step": 8899}
356
+ {"train_loss": 1.6528103351593018, "profiling/batch_time": 0.12443418800830841, "profiling/data_time": 0.1681574434041977, "epoch": 141, "step": 8899}
357
+ {"lr-SGD/params": 0.0042773883003340875, "lr-SGD/params_no_weight_decay": 0.0042773883003340875, "step": 8949}
358
+ {"train_loss": 1.4124027490615845, "profiling/batch_time": 0.12423167377710342, "profiling/data_time": 0.17086508870124817, "epoch": 142, "step": 8949}
359
+ {"lr-SGD/params": 0.004245474723800412, "lr-SGD/params_no_weight_decay": 0.004245474723800412, "step": 8999}
360
+ {"train_loss": 1.3733762502670288, "profiling/batch_time": 0.12414517998695374, "profiling/data_time": 0.16950982809066772, "epoch": 142, "step": 8999}
361
+ {"lr-SGD/params": 0.004213524795161077, "lr-SGD/params_no_weight_decay": 0.004213524795161077, "step": 9049}
362
+ {"train_loss": 1.5901827812194824, "profiling/batch_time": 0.12582403421401978, "profiling/data_time": 0.03982541710138321, "epoch": 143, "step": 9049}
363
+ {"lr-SGD/params": 0.004181540876402631, "lr-SGD/params_no_weight_decay": 0.004181540876402631, "step": 9099}
364
+ {"train_loss": 1.7906770706176758, "profiling/batch_time": 0.12117200344800949, "profiling/data_time": 0.156166210770607, "epoch": 144, "step": 9099}
365
+ {"lr-SGD/params": 0.004149525332024431, "lr-SGD/params_no_weight_decay": 0.004149525332024431, "step": 9149}
366
+ {"train_loss": 1.2943155765533447, "profiling/batch_time": 0.24480104446411133, "profiling/data_time": 0.1600130796432495, "epoch": 145, "step": 9149}
367
+ {"lr-SGD/params": 0.0041174805288638466, "lr-SGD/params_no_weight_decay": 0.0041174805288638466, "step": 9199}
368
+ {"train_loss": 1.2253749370574951, "profiling/batch_time": 0.12112107127904892, "profiling/data_time": 0.1686205118894577, "epoch": 146, "step": 9199}
369
+ {"lr-SGD/params": 0.004085408835921283, "lr-SGD/params_no_weight_decay": 0.004085408835921283, "step": 9249}
370
+ {"train_loss": 1.5660347938537598, "profiling/batch_time": 0.12424401938915253, "profiling/data_time": 0.1588873714208603, "epoch": 146, "step": 9249}
371
+ {"lr-SGD/params": 0.0040533126241850525, "lr-SGD/params_no_weight_decay": 0.0040533126241850525, "step": 9299}
372
+ {"train_loss": 1.4353022575378418, "profiling/batch_time": 0.12157467752695084, "profiling/data_time": 0.1717182844877243, "epoch": 147, "step": 9299}
373
+ {"lr-SGD/params": 0.004021194266456074, "lr-SGD/params_no_weight_decay": 0.004021194266456074, "step": 9349}
374
+ {"train_loss": 1.3797662258148193, "profiling/batch_time": 0.12549935281276703, "profiling/data_time": 0.05552687868475914, "epoch": 148, "step": 9349}
375
+ {"lr-SGD/params": 0.003989056137172481, "lr-SGD/params_no_weight_decay": 0.003989056137172481, "step": 9399}
376
+ {"train_loss": 1.5214821100234985, "profiling/batch_time": 0.12131572514772415, "profiling/data_time": 0.16433222591876984, "epoch": 149, "step": 9399}
377
+ {"lr-SGD/params": 0.003956900612234073, "lr-SGD/params_no_weight_decay": 0.003956900612234073, "step": 9449}
378
+ {"train_loss": 1.1702593564987183, "profiling/batch_time": 0.24044103920459747, "profiling/data_time": 0.1628003865480423, "epoch": 149, "step": 9449}
379
+ {"lr-SGD/params": 0.003924730068826664, "lr-SGD/params_no_weight_decay": 0.003924730068826664, "step": 9499}
380
+ {"train_loss": 1.4966179132461548, "profiling/batch_time": 0.24683506786823273, "profiling/data_time": 0.15999463200569153, "epoch": 150, "step": 9499}
381
+ {"lr-SGD/params": 0.0038925468852463578, "lr-SGD/params_no_weight_decay": 0.0038925468852463578, "step": 9549}
382
+ {"train_loss": 1.4371534585952759, "profiling/batch_time": 0.12446322292089462, "profiling/data_time": 0.1666041910648346, "epoch": 151, "step": 9549}
383
+ {"lr-SGD/params": 0.0038603534407237127, "lr-SGD/params_no_weight_decay": 0.0038603534407237127, "step": 9599}
384
+ {"train_loss": 1.3884294033050537, "profiling/batch_time": 0.12490009516477585, "profiling/data_time": 0.16184668242931366, "epoch": 152, "step": 9599}
385
+ {"lr-SGD/params": 0.0038281521152478564, "lr-SGD/params_no_weight_decay": 0.0038281521152478564, "step": 9649}
386
+ {"train_loss": 1.3335005044937134, "profiling/batch_time": 0.12156173586845398, "profiling/data_time": 0.04207899793982506, "epoch": 153, "step": 9649}
387
+ {"lr-SGD/params": 0.0037959452893905387, "lr-SGD/params_no_weight_decay": 0.0037959452893905387, "step": 9699}
388
+ {"train_loss": 1.6325663328170776, "profiling/batch_time": 0.12049484252929688, "profiling/data_time": 0.043138276785612106, "epoch": 153, "step": 9699}
389
+ {"lr-SGD/params": 0.003763735344130142, "lr-SGD/params_no_weight_decay": 0.003763735344130142, "step": 9749}
390
+ {"train_loss": 1.3882794380187988, "profiling/batch_time": 0.12542037665843964, "profiling/data_time": 0.16687074303627014, "epoch": 154, "step": 9749}
391
+ {"lr-SGD/params": 0.0037315246606756543, "lr-SGD/params_no_weight_decay": 0.0037315246606756543, "step": 9799}
392
+ {"train_loss": 1.31875479221344, "profiling/batch_time": 0.23977509140968323, "profiling/data_time": 0.16366425156593323, "epoch": 155, "step": 9799}
393
+ {"lr-SGD/params": 0.0036993156202906417, "lr-SGD/params_no_weight_decay": 0.0036993156202906417, "step": 9849}
394
+ {"train_loss": 1.4348055124282837, "profiling/batch_time": 0.12520147860050201, "profiling/data_time": 0.15336839854717255, "epoch": 156, "step": 9849}
395
+ {"lr-SGD/params": 0.003667110604117197, "lr-SGD/params_no_weight_decay": 0.003667110604117197, "step": 9899}
396
+ {"train_loss": 1.4103819131851196, "profiling/batch_time": 0.12457533180713654, "profiling/data_time": 0.15792714059352875, "epoch": 157, "step": 9899}
397
+ {"lr-SGD/params": 0.0036349119929999195, "lr-SGD/params_no_weight_decay": 0.0036349119929999195, "step": 9949}
398
+ {"train_loss": 1.4677464962005615, "profiling/batch_time": 0.12462619692087173, "profiling/data_time": 0.1675359308719635, "epoch": 157, "step": 9949}
399
+ {"lr-SGD/params": 0.003602722167309887, "lr-SGD/params_no_weight_decay": 0.003602722167309887, "step": 9999}
400
+ {"train_loss": 1.426251769065857, "profiling/batch_time": 0.1246633380651474, "profiling/data_time": 0.0520380362868309, "epoch": 158, "step": 9999}
401
+ {"lr-SGD/params": 0.0035705435067686984, "lr-SGD/params_no_weight_decay": 0.0035705435067686984, "step": 10049}
402
+ {"train_loss": 1.5633331537246704, "profiling/batch_time": 0.1260138303041458, "profiling/data_time": 0.17042559385299683, "epoch": 159, "step": 10049}
403
+ {"lr-SGD/params": 0.003538378390272536, "lr-SGD/params_no_weight_decay": 0.003538378390272536, "step": 10099}
404
+ {"train_loss": 1.2404468059539795, "profiling/batch_time": 0.2400442510843277, "profiling/data_time": 0.161017507314682, "epoch": 160, "step": 10099}
405
+ {"lr-SGD/params": 0.003506229195716299, "lr-SGD/params_no_weight_decay": 0.003506229195716299, "step": 10149}
406
+ {"train_loss": 1.4475696086883545, "profiling/batch_time": 0.12118644267320633, "profiling/data_time": 0.15813159942626953, "epoch": 161, "step": 10149}
407
+ {"lr-SGD/params": 0.0034740982998178157, "lr-SGD/params_no_weight_decay": 0.0034740982998178157, "step": 10199}
408
+ {"train_loss": 1.5314382314682007, "profiling/batch_time": 0.12625037133693695, "profiling/data_time": 0.15974745154380798, "epoch": 161, "step": 10199}
409
+ {"lr-SGD/params": 0.003441988077942135, "lr-SGD/params_no_weight_decay": 0.003441988077942135, "step": 10249}
410
+ {"train_loss": 1.4051748514175415, "profiling/batch_time": 0.12089817970991135, "profiling/data_time": 0.1701837033033371, "epoch": 162, "step": 10249}
411
+ {"lr-SGD/params": 0.003409900903925921, "lr-SGD/params_no_weight_decay": 0.003409900903925921, "step": 10299}
412
+ {"train_loss": 1.341823935508728, "profiling/batch_time": 0.12453179806470871, "profiling/data_time": 0.05189014598727226, "epoch": 163, "step": 10299}
413
+ {"lr-SGD/params": 0.003377839149901963, "lr-SGD/params_no_weight_decay": 0.003377839149901963, "step": 10349}
414
+ {"train_loss": 1.504239797592163, "profiling/batch_time": 0.13037019968032837, "profiling/data_time": 0.1643347442150116, "epoch": 164, "step": 10349}
415
+ {"lr-SGD/params": 0.0033458051861238063, "lr-SGD/params_no_weight_decay": 0.0033458051861238063, "step": 10399}
416
+ {"train_loss": 1.469572901725769, "profiling/batch_time": 0.23976914584636688, "profiling/data_time": 0.16748224198818207, "epoch": 165, "step": 10399}
417
+ {"lr-SGD/params": 0.003313801380790524, "lr-SGD/params_no_weight_decay": 0.003313801380790524, "step": 10449}
418
+ {"train_loss": 1.4122974872589111, "profiling/batch_time": 0.23990386724472046, "profiling/data_time": 0.15537576377391815, "epoch": 165, "step": 10449}
419
+ {"lr-SGD/params": 0.0032818300998716464, "lr-SGD/params_no_weight_decay": 0.0032818300998716464, "step": 10499}
420
+ {"train_loss": 1.3634172677993774, "profiling/batch_time": 0.12435917556285858, "profiling/data_time": 0.1548335999250412, "epoch": 166, "step": 10499}
421
+ {"lr-SGD/params": 0.0032498937069322443, "lr-SGD/params_no_weight_decay": 0.0032498937069322443, "step": 10549}
422
+ {"train_loss": 1.4184635877609253, "profiling/batch_time": 0.12394599616527557, "profiling/data_time": 0.1653163731098175, "epoch": 167, "step": 10549}
423
+ {"lr-SGD/params": 0.003217994562958199, "lr-SGD/params_no_weight_decay": 0.003217994562958199, "step": 10599}
424
+ {"train_loss": 1.5268136262893677, "profiling/batch_time": 0.12883783876895905, "profiling/data_time": 0.07485253363847733, "epoch": 168, "step": 10599}
425
+ {"lr-SGD/params": 0.0031861350261816585, "lr-SGD/params_no_weight_decay": 0.0031861350261816585, "step": 10649}
426
+ {"train_loss": 1.402324914932251, "profiling/batch_time": 0.12246331572532654, "profiling/data_time": 0.22515921294689178, "epoch": 169, "step": 10649}
427
+ {"lr-SGD/params": 0.0031543174519067033, "lr-SGD/params_no_weight_decay": 0.0031543174519067033, "step": 10699}
428
+ {"train_loss": 1.3565876483917236, "profiling/batch_time": 0.12136048078536987, "profiling/data_time": 0.15473496913909912, "epoch": 169, "step": 10699}
429
+ {"lr-SGD/params": 0.003122544192335216, "lr-SGD/params_no_weight_decay": 0.003122544192335216, "step": 10749}
430
+ {"train_loss": 1.5093883275985718, "profiling/batch_time": 0.23957349359989166, "profiling/data_time": 0.15880118310451508, "epoch": 170, "step": 10749}
431
+ {"lr-SGD/params": 0.003090817596392993, "lr-SGD/params_no_weight_decay": 0.003090817596392993, "step": 10799}
432
+ {"train_loss": 1.309105396270752, "profiling/batch_time": 0.12980060279369354, "profiling/data_time": 0.15921278297901154, "epoch": 171, "step": 10799}
433
+ {"lr-SGD/params": 0.0030591400095560965, "lr-SGD/params_no_weight_decay": 0.0030591400095560965, "step": 10849}
434
+ {"train_loss": 1.483024001121521, "profiling/batch_time": 0.12446194142103195, "profiling/data_time": 0.1623600274324417, "epoch": 172, "step": 10849}
435
+ {"lr-SGD/params": 0.0030275137736774523, "lr-SGD/params_no_weight_decay": 0.0030275137736774523, "step": 10899}
436
+ {"train_loss": 1.2277487516403198, "profiling/batch_time": 0.1340670883655548, "profiling/data_time": 4.95751428604126, "epoch": 173, "step": 10899}
437
+ {"lr-SGD/params": 0.0029959412268137273, "lr-SGD/params_no_weight_decay": 0.0029959412268137273, "step": 10949}
438
+ {"train_loss": 1.3945610523223877, "profiling/batch_time": 0.12474183738231659, "profiling/data_time": 0.052699729800224304, "epoch": 173, "step": 10949}
439
+ {"lr-SGD/params": 0.0029644247030524763, "lr-SGD/params_no_weight_decay": 0.0029644247030524763, "step": 10999}
440
+ {"train_loss": 1.337837815284729, "profiling/batch_time": 0.12153173238039017, "profiling/data_time": 0.16836152970790863, "epoch": 174, "step": 10999}
441
+ {"lr-SGD/params": 0.002932966532339598, "lr-SGD/params_no_weight_decay": 0.002932966532339598, "step": 11049}
442
+ {"train_loss": 1.3587164878845215, "profiling/batch_time": 0.24041058123111725, "profiling/data_time": 0.16311052441596985, "epoch": 175, "step": 11049}
443
+ {"lr-SGD/params": 0.002901569040307075, "lr-SGD/params_no_weight_decay": 0.002901569040307075, "step": 11099}
444
+ {"train_loss": 1.555832028388977, "profiling/batch_time": 0.12458068877458572, "profiling/data_time": 0.15355542302131653, "epoch": 176, "step": 11099}
445
+ {"lr-SGD/params": 0.002870234548101053, "lr-SGD/params_no_weight_decay": 0.002870234548101053, "step": 11149}
446
+ {"train_loss": 1.2937721014022827, "profiling/batch_time": 0.12455607205629349, "profiling/data_time": 0.15754468739032745, "epoch": 176, "step": 11149}
447
+ {"lr-SGD/params": 0.0028389653722102467, "lr-SGD/params_no_weight_decay": 0.0028389653722102467, "step": 11199}
448
+ {"train_loss": 1.4576936960220337, "profiling/batch_time": 0.12434675544500351, "profiling/data_time": 0.1601092368364334, "epoch": 177, "step": 11199}
449
+ {"lr-SGD/params": 0.0028077638242946746, "lr-SGD/params_no_weight_decay": 0.0028077638242946746, "step": 11249}
450
+ {"train_loss": 1.3093504905700684, "profiling/batch_time": 0.12426628172397614, "profiling/data_time": 0.03887128829956055, "epoch": 178, "step": 11249}
451
+ {"lr-SGD/params": 0.002776632211014775, "lr-SGD/params_no_weight_decay": 0.002776632211014775, "step": 11299}
452
+ {"train_loss": 1.3257654905319214, "profiling/batch_time": 0.12092655152082443, "profiling/data_time": 0.16483572125434875, "epoch": 179, "step": 11299}
453
+ {"lr-SGD/params": 0.002745572833860874, "lr-SGD/params_no_weight_decay": 0.002745572833860874, "step": 11349}
454
+ {"train_loss": 1.365466594696045, "profiling/batch_time": 0.2425481230020523, "profiling/data_time": 0.17034539580345154, "epoch": 180, "step": 11349}
455
+ {"lr-SGD/params": 0.002714587988983044, "lr-SGD/params_no_weight_decay": 0.002714587988983044, "step": 11399}
456
+ {"train_loss": 1.426277995109558, "profiling/batch_time": 0.2409752458333969, "profiling/data_time": 0.1656266301870346, "epoch": 180, "step": 11399}
457
+ {"lr-SGD/params": 0.002683679967021354, "lr-SGD/params_no_weight_decay": 0.002683679967021354, "step": 11449}
458
+ {"train_loss": 1.407314419746399, "profiling/batch_time": 0.12542183697223663, "profiling/data_time": 0.16499967873096466, "epoch": 181, "step": 11449}
459
+ {"lr-SGD/params": 0.002652851052936526, "lr-SGD/params_no_weight_decay": 0.002652851052936526, "step": 11499}
460
+ {"train_loss": 1.6236127614974976, "profiling/batch_time": 0.12145162373781204, "profiling/data_time": 0.15734754502773285, "epoch": 182, "step": 11499}
461
+ {"lr-SGD/params": 0.002622103525841014, "lr-SGD/params_no_weight_decay": 0.002622103525841014, "step": 11549}
462
+ {"train_loss": 1.283078670501709, "profiling/batch_time": 0.12496781349182129, "profiling/data_time": 0.04022728651762009, "epoch": 183, "step": 11549}
463
+ {"lr-SGD/params": 0.002591439658830513, "lr-SGD/params_no_weight_decay": 0.002591439658830513, "step": 11599}
464
+ {"train_loss": 1.5339621305465698, "profiling/batch_time": 0.12461557984352112, "profiling/data_time": 0.16351929306983948, "epoch": 184, "step": 11599}
465
+ {"lr-SGD/params": 0.00256086171881592, "lr-SGD/params_no_weight_decay": 0.00256086171881592, "step": 11649}
466
+ {"train_loss": 1.3604543209075928, "profiling/batch_time": 0.1251538246870041, "profiling/data_time": 0.16670870780944824, "epoch": 184, "step": 11649}
467
+ {"lr-SGD/params": 0.0025303719663557402, "lr-SGD/params_no_weight_decay": 0.0025303719663557402, "step": 11699}
468
+ {"train_loss": 1.488834261894226, "profiling/batch_time": 0.23994365334510803, "profiling/data_time": 0.1568785160779953, "epoch": 185, "step": 11699}
469
+ {"lr-SGD/params": 0.002499972655488967, "lr-SGD/params_no_weight_decay": 0.002499972655488967, "step": 11749}
470
+ {"train_loss": 1.3641523122787476, "profiling/batch_time": 0.12669262290000916, "profiling/data_time": 0.16210727393627167, "epoch": 186, "step": 11749}
471
+ {"lr-SGD/params": 0.002469666033568452, "lr-SGD/params_no_weight_decay": 0.002469666033568452, "step": 11799}
472
+ {"train_loss": 1.375364899635315, "profiling/batch_time": 0.12199296057224274, "profiling/data_time": 0.1568789780139923, "epoch": 187, "step": 11799}
473
+ {"lr-SGD/params": 0.0024394543410947668, "lr-SGD/params_no_weight_decay": 0.0024394543410947668, "step": 11849}
474
+ {"train_loss": 1.228126049041748, "profiling/batch_time": 0.12110016494989395, "profiling/data_time": 0.04182791709899902, "epoch": 188, "step": 11849}
475
+ {"lr-SGD/params": 0.0024093398115505524, "lr-SGD/params_no_weight_decay": 0.0024093398115505524, "step": 11899}
476
+ {"train_loss": 1.3116447925567627, "profiling/batch_time": 0.12105496227741241, "profiling/data_time": 0.05214134603738785, "epoch": 188, "step": 11899}
477
+ {"lr-SGD/params": 0.0023793246712354174, "lr-SGD/params_no_weight_decay": 0.0023793246712354174, "step": 11949}
478
+ {"train_loss": 1.190457820892334, "profiling/batch_time": 0.12471407651901245, "profiling/data_time": 0.16692839562892914, "epoch": 189, "step": 11949}
479
+ {"lr-SGD/params": 0.002349411139101348, "lr-SGD/params_no_weight_decay": 0.002349411139101348, "step": 11999}
480
+ {"train_loss": 1.2798465490341187, "profiling/batch_time": 0.2496642917394638, "profiling/data_time": 0.1558738350868225, "epoch": 190, "step": 11999}
481
+ {"lr-SGD/params": 0.0023196014265886643, "lr-SGD/params_no_weight_decay": 0.0023196014265886643, "step": 12049}
482
+ {"train_loss": 1.427772879600525, "profiling/batch_time": 0.12219149619340897, "profiling/data_time": 0.1601627767086029, "epoch": 191, "step": 12049}
483
+ {"lr-SGD/params": 0.0022898977374625346, "lr-SGD/params_no_weight_decay": 0.0022898977374625346, "step": 12099}
484
+ {"train_loss": 1.1536015272140503, "profiling/batch_time": 0.12287797778844833, "profiling/data_time": 0.15979520976543427, "epoch": 192, "step": 12099}
485
+ {"lr-SGD/params": 0.0022603022676500563, "lr-SGD/params_no_weight_decay": 0.0022603022676500563, "step": 12149}
486
+ {"train_loss": 1.5801770687103271, "profiling/batch_time": 0.12417637556791306, "profiling/data_time": 0.16669847071170807, "epoch": 192, "step": 12149}
487
+ {"lr-SGD/params": 0.002230817205077918, "lr-SGD/params_no_weight_decay": 0.002230817205077918, "step": 12199}
488
+ {"train_loss": 1.2333910465240479, "profiling/batch_time": 0.13087832927703857, "profiling/data_time": 0.05228947103023529, "epoch": 193, "step": 12199}
489
+ {"lr-SGD/params": 0.002201444729510642, "lr-SGD/params_no_weight_decay": 0.002201444729510642, "step": 12249}
490
+ {"train_loss": 1.629618525505066, "profiling/batch_time": 0.12294512987136841, "profiling/data_time": 0.1747852861881256, "epoch": 194, "step": 12249}
491
+ {"lr-SGD/params": 0.002172187012389456, "lr-SGD/params_no_weight_decay": 0.002172187012389456, "step": 12299}
492
+ {"train_loss": 1.2529501914978027, "profiling/batch_time": 0.24503520131111145, "profiling/data_time": 0.15909893810749054, "epoch": 195, "step": 12299}
493
+ {"lr-SGD/params": 0.0021430462166717452, "lr-SGD/params_no_weight_decay": 0.0021430462166717452, "step": 12349}
494
+ {"train_loss": 1.2073687314987183, "profiling/batch_time": 0.12204619497060776, "profiling/data_time": 0.1567506045103073, "epoch": 196, "step": 12349}
495
+ {"lr-SGD/params": 0.002114024496671159, "lr-SGD/params_no_weight_decay": 0.002114024496671159, "step": 12399}
496
+ {"train_loss": 1.3392233848571777, "profiling/batch_time": 0.1265222132205963, "profiling/data_time": 0.16691236197948456, "epoch": 196, "step": 12399}
497
+ {"lr-SGD/params": 0.002085123997898348, "lr-SGD/params_no_weight_decay": 0.002085123997898348, "step": 12449}
498
+ {"train_loss": 1.5158703327178955, "profiling/batch_time": 0.12236557900905609, "profiling/data_time": 0.15720301866531372, "epoch": 197, "step": 12449}
499
+ {"lr-SGD/params": 0.0020563468569023517, "lr-SGD/params_no_weight_decay": 0.0020563468569023517, "step": 12499}
500
+ {"train_loss": 1.1624990701675415, "profiling/batch_time": 0.12469322234392166, "profiling/data_time": 0.04587329551577568, "epoch": 198, "step": 12499}
501
+ {"lr-SGD/params": 0.002027695201112641, "lr-SGD/params_no_weight_decay": 0.002027695201112641, "step": 12549}
502
+ {"train_loss": 1.2600271701812744, "profiling/batch_time": 0.12267505377531052, "profiling/data_time": 0.1607874184846878, "epoch": 199, "step": 12549}
503
+ {"lr-SGD/params": 0.0019991711486818504, "lr-SGD/params_no_weight_decay": 0.0019991711486818504, "step": 12599}
504
+ {"train_loss": 1.3109623193740845, "profiling/batch_time": 0.24049881100654602, "profiling/data_time": 0.15959031879901886, "epoch": 199, "step": 12599}
505
+ {"lr-SGD/params": 0.001970776808329181, "lr-SGD/params_no_weight_decay": 0.001970776808329181, "step": 12649}
506
+ {"train_loss": 1.2842530012130737, "profiling/batch_time": 0.2456439882516861, "profiling/data_time": 0.1610061675310135, "epoch": 200, "step": 12649}
507
+ {"lr-SGD/params": 0.0019425142791845123, "lr-SGD/params_no_weight_decay": 0.0019425142791845123, "step": 12699}
508
+ {"train_loss": 1.2957546710968018, "profiling/batch_time": 0.12557388842105865, "profiling/data_time": 0.16178540885448456, "epoch": 201, "step": 12699}
509
+ {"lr-SGD/params": 0.001914385650633221, "lr-SGD/params_no_weight_decay": 0.001914385650633221, "step": 12749}
510
+ {"train_loss": 1.387406826019287, "profiling/batch_time": 0.12293396890163422, "profiling/data_time": 0.16444948315620422, "epoch": 202, "step": 12749}
511
+ {"lr-SGD/params": 0.0018863930021617084, "lr-SGD/params_no_weight_decay": 0.0018863930021617084, "step": 12799}
512
+ {"train_loss": 1.5621638298034668, "profiling/batch_time": 0.12215671688318253, "profiling/data_time": 0.04666706919670105, "epoch": 203, "step": 12799}
513
+ {"lr-SGD/params": 0.001858538403203675, "lr-SGD/params_no_weight_decay": 0.001858538403203675, "step": 12849}
514
+ {"train_loss": 1.2427736520767212, "profiling/batch_time": 0.12627637386322021, "profiling/data_time": 0.03970533236861229, "epoch": 203, "step": 12849}
515
+ {"lr-SGD/params": 0.001830823912987131, "lr-SGD/params_no_weight_decay": 0.001830823912987131, "step": 12899}
516
+ {"train_loss": 1.338883638381958, "profiling/batch_time": 0.12146633863449097, "profiling/data_time": 0.1613941639661789, "epoch": 204, "step": 12899}
517
+ {"lr-SGD/params": 0.001803251580382161, "lr-SGD/params_no_weight_decay": 0.001803251580382161, "step": 12949}
518
+ {"train_loss": 1.2927241325378418, "profiling/batch_time": 0.24026064574718475, "profiling/data_time": 0.1569375842809677, "epoch": 205, "step": 12949}
519
+ {"lr-SGD/params": 0.0017758234437494566, "lr-SGD/params_no_weight_decay": 0.0017758234437494566, "step": 12999}
520
+ {"train_loss": 1.3915648460388184, "profiling/batch_time": 0.12505504488945007, "profiling/data_time": 0.169476717710495, "epoch": 206, "step": 12999}
521
+ {"lr-SGD/params": 0.0017485415307896249, "lr-SGD/params_no_weight_decay": 0.0017485415307896249, "step": 13049}
522
+ {"train_loss": 1.4519730806350708, "profiling/batch_time": 0.121795155107975, "profiling/data_time": 0.16161991655826569, "epoch": 207, "step": 13049}
523
+ {"lr-SGD/params": 0.0017214078583932836, "lr-SGD/params_no_weight_decay": 0.0017214078583932836, "step": 13099}
524
+ {"train_loss": 1.4465008974075317, "profiling/batch_time": 0.1216198280453682, "profiling/data_time": 0.1618436872959137, "epoch": 207, "step": 13099}
525
+ {"lr-SGD/params": 0.001694424432491961, "lr-SGD/params_no_weight_decay": 0.001694424432491961, "step": 13149}
526
+ {"train_loss": 1.370766043663025, "profiling/batch_time": 0.12908360362052917, "profiling/data_time": 0.04274208843708038, "epoch": 208, "step": 13149}
527
+ {"lr-SGD/params": 0.0016675932479097968, "lr-SGD/params_no_weight_decay": 0.0016675932479097968, "step": 13199}
528
+ {"train_loss": 1.5045667886734009, "profiling/batch_time": 0.13013333082199097, "profiling/data_time": 0.15545456111431122, "epoch": 209, "step": 13199}
529
+ {"lr-SGD/params": 0.0016409162882160733, "lr-SGD/params_no_weight_decay": 0.0016409162882160733, "step": 13249}
530
+ {"train_loss": 1.3988953828811646, "profiling/batch_time": 0.23948484659194946, "profiling/data_time": 0.15935677289962769, "epoch": 210, "step": 13249}
531
+ {"lr-SGD/params": 0.0016143955255785688, "lr-SGD/params_no_weight_decay": 0.0016143955255785688, "step": 13299}
532
+ {"train_loss": 1.3435606956481934, "profiling/batch_time": 0.1245865598320961, "profiling/data_time": 0.17229638993740082, "epoch": 211, "step": 13299}
533
+ {"lr-SGD/params": 0.001588032920617766, "lr-SGD/params_no_weight_decay": 0.001588032920617766, "step": 13349}
534
+ {"train_loss": 1.218510389328003, "profiling/batch_time": 0.13099028170108795, "profiling/data_time": 0.16359716653823853, "epoch": 211, "step": 13349}
535
+ {"lr-SGD/params": 0.0015618304222619083, "lr-SGD/params_no_weight_decay": 0.0015618304222619083, "step": 13399}
536
+ {"train_loss": 1.2311089038848877, "profiling/batch_time": 0.12816275656223297, "profiling/data_time": 0.15801772475242615, "epoch": 212, "step": 13399}
537
+ {"lr-SGD/params": 0.0015357899676029143, "lr-SGD/params_no_weight_decay": 0.0015357899676029143, "step": 13449}
538
+ {"train_loss": 1.2234795093536377, "profiling/batch_time": 0.1299135386943817, "profiling/data_time": 0.0453638955950737, "epoch": 213, "step": 13449}
539
+ {"lr-SGD/params": 0.0015099134817531746, "lr-SGD/params_no_weight_decay": 0.0015099134817531746, "step": 13499}
540
+ {"train_loss": 1.4845740795135498, "profiling/batch_time": 0.1245635375380516, "profiling/data_time": 0.16265960037708282, "epoch": 214, "step": 13499}
541
+ {"lr-SGD/params": 0.0014842028777032354, "lr-SGD/params_no_weight_decay": 0.0014842028777032354, "step": 13549}
542
+ {"train_loss": 1.1692842245101929, "profiling/batch_time": 0.24062767624855042, "profiling/data_time": 0.1567971110343933, "epoch": 215, "step": 13549}
543
+ {"lr-SGD/params": 0.0014586600561803725, "lr-SGD/params_no_weight_decay": 0.0014586600561803725, "step": 13599}
544
+ {"train_loss": 1.195277452468872, "profiling/batch_time": 0.23927569389343262, "profiling/data_time": 0.1630610227584839, "epoch": 215, "step": 13599}
545
+ {"lr-SGD/params": 0.0014332869055080771, "lr-SGD/params_no_weight_decay": 0.0014332869055080771, "step": 13649}
546
+ {"train_loss": 1.4452255964279175, "profiling/batch_time": 0.13328449428081512, "profiling/data_time": 0.16541002690792084, "epoch": 216, "step": 13649}
547
+ {"lr-SGD/params": 0.001408085301466454, "lr-SGD/params_no_weight_decay": 0.001408085301466454, "step": 13699}
548
+ {"train_loss": 1.3850980997085571, "profiling/batch_time": 0.12304413318634033, "profiling/data_time": 0.162007674574852, "epoch": 217, "step": 13699}
549
+ {"lr-SGD/params": 0.001383057107153549, "lr-SGD/params_no_weight_decay": 0.001383057107153549, "step": 13749}
550
+ {"train_loss": 1.2614392042160034, "profiling/batch_time": 0.12564042210578918, "profiling/data_time": 0.04210083186626434, "epoch": 218, "step": 13749}
551
+ {"lr-SGD/params": 0.0013582041728476212, "lr-SGD/params_no_weight_decay": 0.0013582041728476212, "step": 13799}
552
+ {"train_loss": 1.246803879737854, "profiling/batch_time": 0.12218530476093292, "profiling/data_time": 0.1589435636997223, "epoch": 219, "step": 13799}
553
+ {"lr-SGD/params": 0.0013335283358703446, "lr-SGD/params_no_weight_decay": 0.0013335283358703446, "step": 13849}
554
+ {"train_loss": 1.316601276397705, "profiling/batch_time": 0.1242116242647171, "profiling/data_time": 0.16013287007808685, "epoch": 219, "step": 13849}
555
+ {"lr-SGD/params": 0.0013090314204509887, "lr-SGD/params_no_weight_decay": 0.0013090314204509887, "step": 13899}
556
+ {"train_loss": 1.0827982425689697, "profiling/batch_time": 0.24042244255542755, "profiling/data_time": 0.16553214192390442, "epoch": 220, "step": 13899}
557
+ {"lr-SGD/params": 0.00128471523759155, "lr-SGD/params_no_weight_decay": 0.00128471523759155, "step": 13949}
558
+ {"train_loss": 1.073810338973999, "profiling/batch_time": 0.12782150506973267, "profiling/data_time": 0.15932218730449677, "epoch": 221, "step": 13949}
559
+ {"lr-SGD/params": 0.0012605815849328745, "lr-SGD/params_no_weight_decay": 0.0012605815849328745, "step": 13999}
560
+ {"train_loss": 1.4364591836929321, "profiling/batch_time": 0.12384553253650665, "profiling/data_time": 0.15785391628742218, "epoch": 222, "step": 13999}
561
+ {"lr-SGD/params": 0.0012366322466217622, "lr-SGD/params_no_weight_decay": 0.0012366322466217622, "step": 14049}
562
+ {"train_loss": 1.1891626119613647, "profiling/batch_time": 0.1345280557870865, "profiling/data_time": 5.042775630950928, "epoch": 223, "step": 14049}
563
+ {"lr-SGD/params": 0.0012128689931790635, "lr-SGD/params_no_weight_decay": 0.0012128689931790635, "step": 14099}
564
+ {"train_loss": 1.354917287826538, "profiling/batch_time": 0.1280432790517807, "profiling/data_time": 0.04412165656685829, "epoch": 223, "step": 14099}
565
+ {"lr-SGD/params": 0.0011892935813687937, "lr-SGD/params_no_weight_decay": 0.0011892935813687937, "step": 14149}
566
+ {"train_loss": 1.3033455610275269, "profiling/batch_time": 0.12733298540115356, "profiling/data_time": 0.16120794415473938, "epoch": 224, "step": 14149}
567
+ {"lr-SGD/params": 0.0011659077540682559, "lr-SGD/params_no_weight_decay": 0.0011659077540682559, "step": 14199}
568
+ {"train_loss": 1.4572583436965942, "profiling/batch_time": 0.24311387538909912, "profiling/data_time": 0.1572999507188797, "epoch": 225, "step": 14199}
569
+ {"lr-SGD/params": 0.001142713240139198, "lr-SGD/params_no_weight_decay": 0.001142713240139198, "step": 14249}
570
+ {"train_loss": 1.481059193611145, "profiling/batch_time": 0.12483620643615723, "profiling/data_time": 0.1604817807674408, "epoch": 226, "step": 14249}
571
+ {"lr-SGD/params": 0.0011197117542999977, "lr-SGD/params_no_weight_decay": 0.0011197117542999977, "step": 14299}
572
+ {"train_loss": 1.43338143825531, "profiling/batch_time": 0.12461964786052704, "profiling/data_time": 0.15766923129558563, "epoch": 226, "step": 14299}
573
+ {"lr-SGD/params": 0.0010969049969989004, "lr-SGD/params_no_weight_decay": 0.0010969049969989004, "step": 14349}
574
+ {"train_loss": 1.1746259927749634, "profiling/batch_time": 0.12195838242769241, "profiling/data_time": 0.16916102170944214, "epoch": 227, "step": 14349}
575
+ {"lr-SGD/params": 0.0010742946542883064, "lr-SGD/params_no_weight_decay": 0.0010742946542883064, "step": 14399}
576
+ {"train_loss": 1.1699808835983276, "profiling/batch_time": 0.12183889746665955, "profiling/data_time": 0.05246110260486603, "epoch": 228, "step": 14399}
577
+ {"lr-SGD/params": 0.001051882397700126, "lr-SGD/params_no_weight_decay": 0.001051882397700126, "step": 14449}
578
+ {"train_loss": 1.3991624116897583, "profiling/batch_time": 0.12147320806980133, "profiling/data_time": 0.15722769498825073, "epoch": 229, "step": 14449}
579
+ {"lr-SGD/params": 0.0010296698841222099, "lr-SGD/params_no_weight_decay": 0.0010296698841222099, "step": 14499}
580
+ {"train_loss": 1.2136558294296265, "profiling/batch_time": 0.23919279873371124, "profiling/data_time": 0.15997157990932465, "epoch": 230, "step": 14499}
581
+ {"lr-SGD/params": 0.0010076587556758518, "lr-SGD/params_no_weight_decay": 0.0010076587556758518, "step": 14549}
582
+ {"train_loss": 1.125064492225647, "profiling/batch_time": 0.2399231493473053, "profiling/data_time": 0.15998458862304688, "epoch": 230, "step": 14549}
583
+ {"lr-SGD/params": 0.0009858506395943956, "lr-SGD/params_no_weight_decay": 0.0009858506395943956, "step": 14599}
584
+ {"train_loss": 1.3101248741149902, "profiling/batch_time": 0.12358653545379639, "profiling/data_time": 0.16574132442474365, "epoch": 231, "step": 14599}
585
+ {"lr-SGD/params": 0.000964247148102937, "lr-SGD/params_no_weight_decay": 0.000964247148102937, "step": 14649}
586
+ {"train_loss": 1.1805051565170288, "profiling/batch_time": 0.12160111963748932, "profiling/data_time": 0.17000378668308258, "epoch": 232, "step": 14649}
587
+ {"lr-SGD/params": 0.0009428498782991378, "lr-SGD/params_no_weight_decay": 0.0009428498782991378, "step": 14699}
588
+ {"train_loss": 1.2722949981689453, "profiling/batch_time": 0.12469946593046188, "profiling/data_time": 0.041021451354026794, "epoch": 233, "step": 14699}
589
+ {"lr-SGD/params": 0.0009216604120351479, "lr-SGD/params_no_weight_decay": 0.0009216604120351479, "step": 14749}
590
+ {"train_loss": 1.3341981172561646, "profiling/batch_time": 0.12413238734006882, "profiling/data_time": 0.16376745700836182, "epoch": 234, "step": 14749}
591
+ {"lr-SGD/params": 0.0009006803158006716, "lr-SGD/params_no_weight_decay": 0.0009006803158006716, "step": 14799}
592
+ {"train_loss": 1.3178791999816895, "profiling/batch_time": 0.1257179081439972, "profiling/data_time": 0.17173683643341064, "epoch": 234, "step": 14799}
593
+ {"lr-SGD/params": 0.0008799111406071539, "lr-SGD/params_no_weight_decay": 0.0008799111406071539, "step": 14849}
594
+ {"train_loss": 1.2662849426269531, "profiling/batch_time": 0.23973838984966278, "profiling/data_time": 0.1673736870288849, "epoch": 235, "step": 14849}
595
+ {"lr-SGD/params": 0.0008593544218731226, "lr-SGD/params_no_weight_decay": 0.0008593544218731226, "step": 14899}
596
+ {"train_loss": 1.2497490644454956, "profiling/batch_time": 0.12413648515939713, "profiling/data_time": 0.16750456392765045, "epoch": 236, "step": 14899}
597
+ {"lr-SGD/params": 0.0008390116793106748, "lr-SGD/params_no_weight_decay": 0.0008390116793106748, "step": 14949}
598
+ {"train_loss": 1.1964623928070068, "profiling/batch_time": 0.12183575332164764, "profiling/data_time": 0.15560340881347656, "epoch": 237, "step": 14949}
599
+ {"lr-SGD/params": 0.0008188844168131298, "lr-SGD/params_no_weight_decay": 0.0008188844168131298, "step": 14999}
600
+ {"train_loss": 1.2281486988067627, "profiling/batch_time": 0.12189087271690369, "profiling/data_time": 0.04312790185213089, "epoch": 238, "step": 14999}
601
+ {"lr-SGD/params": 0.0007989741223438485, "lr-SGD/params_no_weight_decay": 0.0007989741223438485, "step": 15049}
602
+ {"train_loss": 1.2038967609405518, "profiling/batch_time": 0.12611253559589386, "profiling/data_time": 0.04702070727944374, "epoch": 238, "step": 15049}
603
+ {"lr-SGD/params": 0.0007792822678262328, "lr-SGD/params_no_weight_decay": 0.0007792822678262328, "step": 15099}
604
+ {"train_loss": 1.2569550275802612, "profiling/batch_time": 0.13133051991462708, "profiling/data_time": 0.16320738196372986, "epoch": 239, "step": 15099}
605
+ {"lr-SGD/params": 0.000759810309034911, "lr-SGD/params_no_weight_decay": 0.000759810309034911, "step": 15149}
606
+ {"train_loss": 1.3018567562103271, "profiling/batch_time": 0.2406652718782425, "profiling/data_time": 0.15794041752815247, "epoch": 240, "step": 15149}
607
+ {"lr-SGD/params": 0.0007405596854881096, "lr-SGD/params_no_weight_decay": 0.0007405596854881096, "step": 15199}
608
+ {"train_loss": 1.3627266883850098, "profiling/batch_time": 0.12275678664445877, "profiling/data_time": 0.16495169699192047, "epoch": 241, "step": 15199}
609
+ {"lr-SGD/params": 0.0007215318203412389, "lr-SGD/params_no_weight_decay": 0.0007215318203412389, "step": 15249}
610
+ {"train_loss": 1.2822508811950684, "profiling/batch_time": 0.12495072185993195, "profiling/data_time": 0.16146646440029144, "epoch": 242, "step": 15249}
611
+ {"lr-SGD/params": 0.0007027281202816806, "lr-SGD/params_no_weight_decay": 0.0007027281202816806, "step": 15299}
612
+ {"train_loss": 1.3508813381195068, "profiling/batch_time": 0.1245127022266388, "profiling/data_time": 0.16305774450302124, "epoch": 242, "step": 15299}
613
+ {"lr-SGD/params": 0.0006841499754247961, "lr-SGD/params_no_weight_decay": 0.0006841499754247961, "step": 15349}
614
+ {"train_loss": 1.1135174036026, "profiling/batch_time": 0.12141955643892288, "profiling/data_time": 0.04844117537140846, "epoch": 243, "step": 15349}
615
+ {"lr-SGD/params": 0.0006657987592111542, "lr-SGD/params_no_weight_decay": 0.0006657987592111542, "step": 15399}
616
+ {"train_loss": 1.2243914604187012, "profiling/batch_time": 0.12110701948404312, "profiling/data_time": 0.16732677817344666, "epoch": 244, "step": 15399}
617
+ {"lr-SGD/params": 0.0006476758283049989, "lr-SGD/params_no_weight_decay": 0.0006476758283049989, "step": 15449}
618
+ {"train_loss": 1.0785493850708008, "profiling/batch_time": 0.23956842720508575, "profiling/data_time": 0.15902861952781677, "epoch": 245, "step": 15449}
619
+ {"lr-SGD/params": 0.0006297825224939537, "lr-SGD/params_no_weight_decay": 0.0006297825224939537, "step": 15499}
620
+ {"train_loss": 1.302217960357666, "profiling/batch_time": 0.12117218226194382, "profiling/data_time": 0.15317858755588531, "epoch": 246, "step": 15499}
621
+ {"lr-SGD/params": 0.0006121201645899744, "lr-SGD/params_no_weight_decay": 0.0006121201645899744, "step": 15549}
622
+ {"train_loss": 1.3693379163742065, "profiling/batch_time": 0.1274128556251526, "profiling/data_time": 0.16298100352287292, "epoch": 246, "step": 15549}
623
+ {"lr-SGD/params": 0.0005946900603315549, "lr-SGD/params_no_weight_decay": 0.0005946900603315549, "step": 15599}
624
+ {"train_loss": 1.3637317419052124, "profiling/batch_time": 0.12542419135570526, "profiling/data_time": 0.15479405224323273, "epoch": 247, "step": 15599}
625
+ {"lr-SGD/params": 0.0005774934982872001, "lr-SGD/params_no_weight_decay": 0.0005774934982872001, "step": 15649}
626
+ {"train_loss": 1.406272292137146, "profiling/batch_time": 0.12102124094963074, "profiling/data_time": 0.048932552337646484, "epoch": 248, "step": 15649}
627
+ {"lr-SGD/params": 0.0005605317497601609, "lr-SGD/params_no_weight_decay": 0.0005605317497601609, "step": 15699}
628
+ {"train_loss": 1.3298139572143555, "profiling/batch_time": 0.12104339152574539, "profiling/data_time": 0.16354341804981232, "epoch": 249, "step": 15699}
629
+ {"lr-SGD/params": 0.0005438060686944523, "lr-SGD/params_no_weight_decay": 0.0005438060686944523, "step": 15749}
630
+ {"train_loss": 1.201293706893921, "profiling/batch_time": 0.23965606093406677, "profiling/data_time": 0.15436120331287384, "epoch": 249, "step": 15749}
631
+ {"lr-SGD/params": 0.0005273176915821546, "lr-SGD/params_no_weight_decay": 0.0005273176915821546, "step": 15799}
632
+ {"train_loss": 1.132200837135315, "profiling/batch_time": 0.24043546617031097, "profiling/data_time": 0.16002660989761353, "epoch": 250, "step": 15799}
633
+ {"lr-SGD/params": 0.0005110678373719932, "lr-SGD/params_no_weight_decay": 0.0005110678373719932, "step": 15849}
634
+ {"train_loss": 1.2759621143341064, "profiling/batch_time": 0.12462792545557022, "profiling/data_time": 0.1546960026025772, "epoch": 251, "step": 15849}
635
+ {"lr-SGD/params": 0.000495057707379235, "lr-SGD/params_no_weight_decay": 0.000495057707379235, "step": 15899}
636
+ {"train_loss": 1.126311182975769, "profiling/batch_time": 0.12563671171665192, "profiling/data_time": 0.15990442037582397, "epoch": 252, "step": 15899}
637
+ {"lr-SGD/params": 0.0004792884851968711, "lr-SGD/params_no_weight_decay": 0.0004792884851968711, "step": 15949}
638
+ {"train_loss": 1.3084428310394287, "profiling/batch_time": 0.12174136191606522, "profiling/data_time": 0.05157836899161339, "epoch": 253, "step": 15949}
639
+ {"lr-SGD/params": 0.0004637613366081214, "lr-SGD/params_no_weight_decay": 0.0004637613366081214, "step": 15999}
640
+ {"train_loss": 1.285117506980896, "profiling/batch_time": 0.12181182205677032, "profiling/data_time": 0.039113178849220276, "epoch": 253, "step": 15999}
641
+ {"lr-SGD/params": 0.00044847740950024504, "lr-SGD/params_no_weight_decay": 0.00044847740950024504, "step": 16049}
642
+ {"train_loss": 1.2118972539901733, "profiling/batch_time": 0.1272580921649933, "profiling/data_time": 0.1657620072364807, "epoch": 254, "step": 16049}
643
+ {"lr-SGD/params": 0.00043343783377968413, "lr-SGD/params_no_weight_decay": 0.00043343783377968413, "step": 16099}
644
+ {"train_loss": 1.2424561977386475, "profiling/batch_time": 0.2396050989627838, "profiling/data_time": 0.1547819823026657, "epoch": 255, "step": 16099}
645
+ {"lr-SGD/params": 0.000418643721288531, "lr-SGD/params_no_weight_decay": 0.000418643721288531, "step": 16149}
646
+ {"train_loss": 1.2282755374908447, "profiling/batch_time": 0.12479904294013977, "profiling/data_time": 0.155418261885643, "epoch": 256, "step": 16149}
647
+ {"lr-SGD/params": 0.0004040961657223319, "lr-SGD/params_no_weight_decay": 0.0004040961657223319, "step": 16199}
648
+ {"train_loss": 1.2549504041671753, "profiling/batch_time": 0.12425987422466278, "profiling/data_time": 0.1566580981016159, "epoch": 257, "step": 16199}
649
+ {"lr-SGD/params": 0.0003897962425492323, "lr-SGD/params_no_weight_decay": 0.0003897962425492323, "step": 16249}
650
+ {"train_loss": 1.2613255977630615, "profiling/batch_time": 0.12462827563285828, "profiling/data_time": 0.15640953183174133, "epoch": 257, "step": 16249}
651
+ {"lr-SGD/params": 0.000375745008930471, "lr-SGD/params_no_weight_decay": 0.000375745008930471, "step": 16299}
652
+ {"train_loss": 1.5046718120574951, "profiling/batch_time": 0.12128961086273193, "profiling/data_time": 0.05501759052276611, "epoch": 258, "step": 16299}
653
+ {"lr-SGD/params": 0.0003619435036422261, "lr-SGD/params_no_weight_decay": 0.0003619435036422261, "step": 16349}
654
+ {"train_loss": 1.2921538352966309, "profiling/batch_time": 0.12409891933202744, "profiling/data_time": 0.1590263843536377, "epoch": 259, "step": 16349}
655
+ {"lr-SGD/params": 0.0003483927469988198, "lr-SGD/params_no_weight_decay": 0.0003483927469988198, "step": 16399}
656
+ {"train_loss": 1.263568639755249, "profiling/batch_time": 0.239770770072937, "profiling/data_time": 0.1662643998861313, "epoch": 260, "step": 16399}
657
+ {"lr-SGD/params": 0.00033509374077729087, "lr-SGD/params_no_weight_decay": 0.00033509374077729087, "step": 16449}
658
+ {"train_loss": 1.3893256187438965, "profiling/batch_time": 0.12475691735744476, "profiling/data_time": 0.16885170340538025, "epoch": 261, "step": 16449}
659
+ {"lr-SGD/params": 0.0003220474681433325, "lr-SGD/params_no_weight_decay": 0.0003220474681433325, "step": 16499}
660
+ {"train_loss": 1.2907871007919312, "profiling/batch_time": 0.12439969927072525, "profiling/data_time": 0.16552035510540009, "epoch": 261, "step": 16499}
661
+ {"lr-SGD/params": 0.0003092548935786109, "lr-SGD/params_no_weight_decay": 0.0003092548935786109, "step": 16549}
662
+ {"train_loss": 1.390498399734497, "profiling/batch_time": 0.1248864009976387, "profiling/data_time": 0.16792477667331696, "epoch": 262, "step": 16549}
663
+ {"lr-SGD/params": 0.0002967169628094645, "lr-SGD/params_no_weight_decay": 0.0002967169628094645, "step": 16599}
664
+ {"train_loss": 1.2892175912857056, "profiling/batch_time": 0.1209157258272171, "profiling/data_time": 0.05343443527817726, "epoch": 263, "step": 16599}
665
+ {"lr-SGD/params": 0.00028443460273698713, "lr-SGD/params_no_weight_decay": 0.00028443460273698713, "step": 16649}
666
+ {"train_loss": 1.2578648328781128, "profiling/batch_time": 0.12485397607088089, "profiling/data_time": 0.16587898135185242, "epoch": 264, "step": 16649}
667
+ {"lr-SGD/params": 0.0002724087213685032, "lr-SGD/params_no_weight_decay": 0.0002724087213685032, "step": 16699}
668
+ {"train_loss": 1.1645264625549316, "profiling/batch_time": 0.23984968662261963, "profiling/data_time": 0.1657443642616272, "epoch": 265, "step": 16699}
669
+ {"lr-SGD/params": 0.0002606402077504421, "lr-SGD/params_no_weight_decay": 0.0002606402077504421, "step": 16749}
670
+ {"train_loss": 1.3515645265579224, "profiling/batch_time": 0.2394360452890396, "profiling/data_time": 0.15925784409046173, "epoch": 265, "step": 16749}
671
+ {"lr-SGD/params": 0.00024912993190261395, "lr-SGD/params_no_weight_decay": 0.00024912993190261395, "step": 16799}
672
+ {"train_loss": 1.341690182685852, "profiling/batch_time": 0.1257871687412262, "profiling/data_time": 0.16710366308689117, "epoch": 266, "step": 16799}
673
+ {"lr-SGD/params": 0.00023787874475388878, "lr-SGD/params_no_weight_decay": 0.00023787874475388878, "step": 16849}
674
+ {"train_loss": 1.4610373973846436, "profiling/batch_time": 0.1252395212650299, "profiling/data_time": 0.16570742428302765, "epoch": 267, "step": 16849}
675
+ {"lr-SGD/params": 0.00022688747807929048, "lr-SGD/params_no_weight_decay": 0.00022688747807929048, "step": 16899}
676
+ {"train_loss": 1.3133550882339478, "profiling/batch_time": 0.12360819429159164, "profiling/data_time": 0.04106632620096207, "epoch": 268, "step": 16899}
677
+ {"lr-SGD/params": 0.00021615694443850617, "lr-SGD/params_no_weight_decay": 0.00021615694443850617, "step": 16949}
678
+ {"train_loss": 1.4594700336456299, "profiling/batch_time": 0.1256301999092102, "profiling/data_time": 0.15541191399097443, "epoch": 269, "step": 16949}
679
+ {"lr-SGD/params": 0.00020568793711581387, "lr-SGD/params_no_weight_decay": 0.00020568793711581387, "step": 16999}
680
+ {"train_loss": 1.2803843021392822, "profiling/batch_time": 0.12288035452365875, "profiling/data_time": 0.1636442095041275, "epoch": 269, "step": 16999}
681
+ {"lr-SGD/params": 0.00019548123006143787, "lr-SGD/params_no_weight_decay": 0.00019548123006143787, "step": 17049}
682
+ {"train_loss": 1.1363598108291626, "profiling/batch_time": 0.23977535963058472, "profiling/data_time": 0.16447538137435913, "epoch": 270, "step": 17049}
683
+ {"lr-SGD/params": 0.00018553757783433336, "lr-SGD/params_no_weight_decay": 0.00018553757783433336, "step": 17099}
684
+ {"train_loss": 1.5121523141860962, "profiling/batch_time": 0.12352284044027328, "profiling/data_time": 0.18396513164043427, "epoch": 271, "step": 17099}
685
+ {"lr-SGD/params": 0.00017585771554639944, "lr-SGD/params_no_weight_decay": 0.00017585771554639944, "step": 17149}
686
+ {"train_loss": 1.3830809593200684, "profiling/batch_time": 0.12414563447237015, "profiling/data_time": 0.15381762385368347, "epoch": 272, "step": 17149}
687
+ {"lr-SGD/params": 0.00016644235880813925, "lr-SGD/params_no_weight_decay": 0.00016644235880813925, "step": 17199}
688
+ {"train_loss": 1.2220230102539062, "profiling/batch_time": 0.13853298127651215, "profiling/data_time": 5.428259372711182, "epoch": 273, "step": 17199}
689
+ {"lr-SGD/params": 0.00015729220367575311, "lr-SGD/params_no_weight_decay": 0.00015729220367575311, "step": 17249}
690
+ {"train_loss": 1.4591877460479736, "profiling/batch_time": 0.1207488477230072, "profiling/data_time": 0.03834666684269905, "epoch": 273, "step": 17249}
691
+ {"lr-SGD/params": 0.0001484079265996824, "lr-SGD/params_no_weight_decay": 0.0001484079265996824, "step": 17299}
692
+ {"train_loss": 1.2534630298614502, "profiling/batch_time": 0.12428693473339081, "profiling/data_time": 0.16401976346969604, "epoch": 274, "step": 17299}
693
+ {"lr-SGD/params": 0.00013979018437459948, "lr-SGD/params_no_weight_decay": 0.00013979018437459948, "step": 17349}
694
+ {"train_loss": 1.2808306217193604, "profiling/batch_time": 0.23952603340148926, "profiling/data_time": 0.1670539528131485, "epoch": 275, "step": 17349}
695
+ {"lr-SGD/params": 0.00013143961409085324, "lr-SGD/params_no_weight_decay": 0.00013143961409085324, "step": 17399}
696
+ {"train_loss": 1.162952184677124, "profiling/batch_time": 0.12271013855934143, "profiling/data_time": 0.15975309908390045, "epoch": 276, "step": 17399}
697
+ {"lr-SGD/params": 0.00012335683308737082, "lr-SGD/params_no_weight_decay": 0.00012335683308737082, "step": 17449}
698
+ {"train_loss": 1.1876853704452515, "profiling/batch_time": 0.12550266087055206, "profiling/data_time": 0.15567152202129364, "epoch": 276, "step": 17449}
699
+ {"lr-SGD/params": 0.0001155424389060168, "lr-SGD/params_no_weight_decay": 0.0001155424389060168, "step": 17499}
700
+ {"train_loss": 1.0537056922912598, "profiling/batch_time": 0.12493333965539932, "profiling/data_time": 0.15378816425800323, "epoch": 277, "step": 17499}
701
+ {"lr-SGD/params": 0.00010799700924742328, "lr-SGD/params_no_weight_decay": 0.00010799700924742328, "step": 17549}
702
+ {"train_loss": 1.2269837856292725, "profiling/batch_time": 0.13066366314888, "profiling/data_time": 0.05148928612470627, "epoch": 278, "step": 17549}
703
+ {"lr-SGD/params": 0.00010072110192827642, "lr-SGD/params_no_weight_decay": 0.00010072110192827642, "step": 17599}
704
+ {"train_loss": 1.2975815534591675, "profiling/batch_time": 0.1239573284983635, "profiling/data_time": 0.1635693609714508, "epoch": 279, "step": 17599}
705
+ {"lr-SGD/params": 9.371525484008105e-05, "lr-SGD/params_no_weight_decay": 9.371525484008105e-05, "step": 17649}
706
+ {"train_loss": 1.2603715658187866, "profiling/batch_time": 0.2399410903453827, "profiling/data_time": 0.15964321792125702, "epoch": 280, "step": 17649}
707
+ {"lr-SGD/params": 8.697998590939408e-05, "lr-SGD/params_no_weight_decay": 8.697998590939408e-05, "step": 17699}
708
+ {"train_loss": 1.3075510263442993, "profiling/batch_time": 0.24008263647556305, "profiling/data_time": 0.16474418342113495, "epoch": 280, "step": 17699}
709
+ {"lr-SGD/params": 8.051579305953856e-05, "lr-SGD/params_no_weight_decay": 8.051579305953856e-05, "step": 17749}
710
+ {"train_loss": 1.2019222974777222, "profiling/batch_time": 0.1214505136013031, "profiling/data_time": 0.16902762651443481, "epoch": 281, "step": 17749}
711
+ {"lr-SGD/params": 7.432315417378889e-05, "lr-SGD/params_no_weight_decay": 7.432315417378889e-05, "step": 17799}
712
+ {"train_loss": 1.1720890998840332, "profiling/batch_time": 0.12398785352706909, "profiling/data_time": 0.16354987025260925, "epoch": 282, "step": 17799}
713
+ {"lr-SGD/params": 6.84025270600448e-05, "lr-SGD/params_no_weight_decay": 6.84025270600448e-05, "step": 17849}
714
+ {"train_loss": 1.363693356513977, "profiling/batch_time": 0.12102341651916504, "profiling/data_time": 0.04085170850157738, "epoch": 283, "step": 17849}
715
+ {"lr-SGD/params": 6.275434941698549e-05, "lr-SGD/params_no_weight_decay": 6.275434941698549e-05, "step": 17899}
716
+ {"train_loss": 1.3540793657302856, "profiling/batch_time": 0.1282563954591751, "profiling/data_time": 0.21279045939445496, "epoch": 284, "step": 17899}
717
+ {"lr-SGD/params": 5.7379038801713735e-05, "lr-SGD/params_no_weight_decay": 5.7379038801713735e-05, "step": 17949}
718
+ {"train_loss": 1.1996692419052124, "profiling/batch_time": 0.12476526945829391, "profiling/data_time": 0.1565343290567398, "epoch": 284, "step": 17949}
719
+ {"lr-SGD/params": 5.227699259888367e-05, "lr-SGD/params_no_weight_decay": 5.227699259888367e-05, "step": 17999}
720
+ {"train_loss": 1.3223966360092163, "profiling/batch_time": 0.2396056354045868, "profiling/data_time": 0.16918402910232544, "epoch": 285, "step": 17999}
721
+ {"lr-SGD/params": 4.744858799132411e-05, "lr-SGD/params_no_weight_decay": 4.744858799132411e-05, "step": 18049}
722
+ {"train_loss": 1.3816845417022705, "profiling/batch_time": 0.12447280436754227, "profiling/data_time": 0.15442921221256256, "epoch": 286, "step": 18049}
723
+ {"lr-SGD/params": 4.2894181932155884e-05, "lr-SGD/params_no_weight_decay": 4.2894181932155884e-05, "step": 18099}
724
+ {"train_loss": 1.17496657371521, "profiling/batch_time": 0.12461427599191666, "profiling/data_time": 0.16633084416389465, "epoch": 287, "step": 18099}
725
+ {"lr-SGD/params": 3.8614111118401e-05, "lr-SGD/params_no_weight_decay": 3.8614111118401e-05, "step": 18149}
726
+ {"train_loss": 1.3116765022277832, "profiling/batch_time": 0.12453584372997284, "profiling/data_time": 0.04149496927857399, "epoch": 288, "step": 18149}
727
+ {"lr-SGD/params": 3.460869196609213e-05, "lr-SGD/params_no_weight_decay": 3.460869196609213e-05, "step": 18199}
728
+ {"train_loss": 1.2979635000228882, "profiling/batch_time": 0.12177031487226486, "profiling/data_time": 0.05626557394862175, "epoch": 288, "step": 18199}
729
+ {"lr-SGD/params": 3.08782205868818e-05, "lr-SGD/params_no_weight_decay": 3.08782205868818e-05, "step": 18249}
730
+ {"train_loss": 1.2177317142486572, "profiling/batch_time": 0.1210298091173172, "profiling/data_time": 0.16186575591564178, "epoch": 289, "step": 18249}
731
+ {"lr-SGD/params": 2.7422972766149376e-05, "lr-SGD/params_no_weight_decay": 2.7422972766149376e-05, "step": 18299}
732
+ {"train_loss": 1.2439314126968384, "profiling/batch_time": 0.2398872822523117, "profiling/data_time": 0.15380191802978516, "epoch": 290, "step": 18299}
733
+ {"lr-SGD/params": 2.4243203942614825e-05, "lr-SGD/params_no_weight_decay": 2.4243203942614825e-05, "step": 18349}
734
+ {"train_loss": 1.2223570346832275, "profiling/batch_time": 0.12542758882045746, "profiling/data_time": 0.15843647718429565, "epoch": 291, "step": 18349}
735
+ {"lr-SGD/params": 2.133914918945361e-05, "lr-SGD/params_no_weight_decay": 2.133914918945361e-05, "step": 18399}
736
+ {"train_loss": 1.1647089719772339, "profiling/batch_time": 0.12161565572023392, "profiling/data_time": 0.16476093232631683, "epoch": 292, "step": 18399}
737
+ {"lr-SGD/params": 1.8711023196917583e-05, "lr-SGD/params_no_weight_decay": 1.8711023196917583e-05, "step": 18449}
738
+ {"train_loss": 1.3165111541748047, "profiling/batch_time": 0.12498448044061661, "profiling/data_time": 0.16687889397144318, "epoch": 292, "step": 18449}
739
+ {"lr-SGD/params": 1.635902025646648e-05, "lr-SGD/params_no_weight_decay": 1.635902025646648e-05, "step": 18499}
740
+ {"train_loss": 1.1857216358184814, "profiling/batch_time": 0.12152261286973953, "profiling/data_time": 0.051583148539066315, "epoch": 293, "step": 18499}
741
+ {"lr-SGD/params": 1.4283314246400057e-05, "lr-SGD/params_no_weight_decay": 1.4283314246400057e-05, "step": 18549}
742
+ {"train_loss": 1.1741279363632202, "profiling/batch_time": 0.12083910405635834, "profiling/data_time": 0.16734792292118073, "epoch": 294, "step": 18549}
743
+ {"lr-SGD/params": 1.2484058619007093e-05, "lr-SGD/params_no_weight_decay": 1.2484058619007093e-05, "step": 18599}
744
+ {"train_loss": 1.2412506341934204, "profiling/batch_time": 0.23971183598041534, "profiling/data_time": 0.1564503014087677, "epoch": 295, "step": 18599}
745
+ {"lr-SGD/params": 1.0961386389219987e-05, "lr-SGD/params_no_weight_decay": 1.0961386389219987e-05, "step": 18649}
746
+ {"train_loss": 1.1873950958251953, "profiling/batch_time": 0.12422172725200653, "profiling/data_time": 0.16630367934703827, "epoch": 296, "step": 18649}
747
+ {"lr-SGD/params": 9.715410124779975e-06, "lr-SGD/params_no_weight_decay": 9.715410124779975e-06, "step": 18699}
748
+ {"train_loss": 1.2892200946807861, "profiling/batch_time": 0.1252584308385849, "profiling/data_time": 0.16207486391067505, "epoch": 296, "step": 18699}
749
+ {"lr-SGD/params": 8.746221937917545e-06, "lr-SGD/params_no_weight_decay": 8.746221937917545e-06, "step": 18749}
750
+ {"train_loss": 1.1057746410369873, "profiling/batch_time": 0.12293604016304016, "profiling/data_time": 0.1698208451271057, "epoch": 297, "step": 18749}
751
+ {"lr-SGD/params": 8.053893478540532e-06, "lr-SGD/params_no_weight_decay": 8.053893478540532e-06, "step": 18799}
752
+ {"train_loss": 1.2065918445587158, "profiling/batch_time": 0.12072161585092545, "profiling/data_time": 0.04154318571090698, "epoch": 298, "step": 18799}
753
+ {"lr-SGD/params": 7.638475928938675e-06, "lr-SGD/params_no_weight_decay": 7.638475928938675e-06, "step": 18849}
754
+ {"train_loss": 1.3170959949493408, "profiling/batch_time": 0.12701506912708282, "profiling/data_time": 0.23405618965625763, "epoch": 299, "step": 18849}
755
+ {"lr-SGD/params": 7.5e-06, "lr-SGD/params_no_weight_decay": 7.5e-06, "step": 18899}
756
+ {"train_loss": 1.2580989599227905, "profiling/batch_time": 0.2382694035768509, "profiling/data_time": 0.16797509789466858, "epoch": 299, "step": 18899}
ssl_dino2/resnet50/resnet50_dino.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:808c822989223f23efba3606b284464f99ce13041c02911c6e0dee5a96883934
3
+ size 102546083