chichi56 commited on
Commit
cb4cb62
·
verified ·
1 Parent(s): ef6e17a

Upload trainer_log.jsonl with huggingface_hub

Browse files
Files changed (1) hide show
  1. trainer_log.jsonl +278 -0
trainer_log.jsonl ADDED
@@ -0,0 +1,278 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"current_steps": 10, "total_steps": 2271, "loss": 0.8143, "lr": 7.894736842105263e-07, "epoch": 0.013205678441729944, "percentage": 0.44, "elapsed_time": "0:00:30", "remaining_time": "1:55:12"}
2
+ {"current_steps": 20, "total_steps": 2271, "loss": 0.7541, "lr": 1.6666666666666667e-06, "epoch": 0.02641135688345989, "percentage": 0.88, "elapsed_time": "0:00:59", "remaining_time": "1:51:08"}
3
+ {"current_steps": 30, "total_steps": 2271, "loss": 0.6724, "lr": 2.5438596491228075e-06, "epoch": 0.03961703532518983, "percentage": 1.32, "elapsed_time": "0:01:27", "remaining_time": "1:49:25"}
4
+ {"current_steps": 40, "total_steps": 2271, "loss": 0.6268, "lr": 3.421052631578948e-06, "epoch": 0.05282271376691978, "percentage": 1.76, "elapsed_time": "0:01:56", "remaining_time": "1:48:15"}
5
+ {"current_steps": 50, "total_steps": 2271, "loss": 0.5768, "lr": 4.298245614035088e-06, "epoch": 0.06602839220864971, "percentage": 2.2, "elapsed_time": "0:02:24", "remaining_time": "1:47:15"}
6
+ {"current_steps": 60, "total_steps": 2271, "loss": 0.5667, "lr": 5.175438596491229e-06, "epoch": 0.07923407065037966, "percentage": 2.64, "elapsed_time": "0:02:53", "remaining_time": "1:46:19"}
7
+ {"current_steps": 70, "total_steps": 2271, "loss": 0.5352, "lr": 6.0526315789473685e-06, "epoch": 0.09243974909210961, "percentage": 3.08, "elapsed_time": "0:03:22", "remaining_time": "1:46:04"}
8
+ {"current_steps": 80, "total_steps": 2271, "loss": 0.5305, "lr": 6.92982456140351e-06, "epoch": 0.10564542753383956, "percentage": 3.52, "elapsed_time": "0:03:51", "remaining_time": "1:45:31"}
9
+ {"current_steps": 90, "total_steps": 2271, "loss": 0.5323, "lr": 7.80701754385965e-06, "epoch": 0.1188511059755695, "percentage": 3.96, "elapsed_time": "0:04:18", "remaining_time": "1:44:26"}
10
+ {"current_steps": 100, "total_steps": 2271, "loss": 0.5373, "lr": 8.68421052631579e-06, "epoch": 0.13205678441729943, "percentage": 4.4, "elapsed_time": "0:04:47", "remaining_time": "1:43:52"}
11
+ {"current_steps": 110, "total_steps": 2271, "loss": 0.5385, "lr": 9.56140350877193e-06, "epoch": 0.14526246285902938, "percentage": 4.84, "elapsed_time": "0:05:15", "remaining_time": "1:43:26"}
12
+ {"current_steps": 120, "total_steps": 2271, "loss": 0.5426, "lr": 1.043859649122807e-05, "epoch": 0.1584681413007593, "percentage": 5.28, "elapsed_time": "0:05:45", "remaining_time": "1:43:04"}
13
+ {"current_steps": 130, "total_steps": 2271, "loss": 0.5321, "lr": 1.1315789473684212e-05, "epoch": 0.17167381974248927, "percentage": 5.72, "elapsed_time": "0:06:13", "remaining_time": "1:42:25"}
14
+ {"current_steps": 140, "total_steps": 2271, "loss": 0.5403, "lr": 1.2192982456140352e-05, "epoch": 0.18487949818421923, "percentage": 6.16, "elapsed_time": "0:06:41", "remaining_time": "1:41:58"}
15
+ {"current_steps": 150, "total_steps": 2271, "loss": 0.5374, "lr": 1.3070175438596493e-05, "epoch": 0.19808517662594916, "percentage": 6.61, "elapsed_time": "0:07:11", "remaining_time": "1:41:38"}
16
+ {"current_steps": 160, "total_steps": 2271, "loss": 0.5469, "lr": 1.3947368421052631e-05, "epoch": 0.2112908550676791, "percentage": 7.05, "elapsed_time": "0:07:40", "remaining_time": "1:41:11"}
17
+ {"current_steps": 170, "total_steps": 2271, "loss": 0.5502, "lr": 1.4824561403508773e-05, "epoch": 0.22449653350940904, "percentage": 7.49, "elapsed_time": "0:08:07", "remaining_time": "1:40:30"}
18
+ {"current_steps": 180, "total_steps": 2271, "loss": 0.5601, "lr": 1.5701754385964915e-05, "epoch": 0.237702211951139, "percentage": 7.93, "elapsed_time": "0:08:37", "remaining_time": "1:40:08"}
19
+ {"current_steps": 190, "total_steps": 2271, "loss": 0.5484, "lr": 1.6578947368421053e-05, "epoch": 0.2509078903928689, "percentage": 8.37, "elapsed_time": "0:09:06", "remaining_time": "1:39:41"}
20
+ {"current_steps": 200, "total_steps": 2271, "loss": 0.5373, "lr": 1.7456140350877195e-05, "epoch": 0.26411356883459886, "percentage": 8.81, "elapsed_time": "0:09:34", "remaining_time": "1:39:10"}
21
+ {"current_steps": 210, "total_steps": 2271, "loss": 0.5427, "lr": 1.8333333333333333e-05, "epoch": 0.27731924727632884, "percentage": 9.25, "elapsed_time": "0:10:01", "remaining_time": "1:38:28"}
22
+ {"current_steps": 220, "total_steps": 2271, "loss": 0.566, "lr": 1.9210526315789474e-05, "epoch": 0.29052492571805877, "percentage": 9.69, "elapsed_time": "0:10:29", "remaining_time": "1:37:52"}
23
+ {"current_steps": 230, "total_steps": 2271, "loss": 0.5662, "lr": 1.9999988176857314e-05, "epoch": 0.3037306041597887, "percentage": 10.13, "elapsed_time": "0:10:57", "remaining_time": "1:37:13"}
24
+ {"current_steps": 240, "total_steps": 2271, "loss": 0.567, "lr": 1.9998569433562935e-05, "epoch": 0.3169362826015186, "percentage": 10.57, "elapsed_time": "0:11:25", "remaining_time": "1:36:41"}
25
+ {"current_steps": 250, "total_steps": 2271, "loss": 0.5715, "lr": 1.999478644612963e-05, "epoch": 0.3301419610432486, "percentage": 11.01, "elapsed_time": "0:11:53", "remaining_time": "1:36:08"}
26
+ {"current_steps": 260, "total_steps": 2271, "loss": 0.552, "lr": 1.9988640109075946e-05, "epoch": 0.34334763948497854, "percentage": 11.45, "elapsed_time": "0:12:22", "remaining_time": "1:35:41"}
27
+ {"current_steps": 270, "total_steps": 2271, "loss": 0.5583, "lr": 1.9980131875753934e-05, "epoch": 0.35655331792670847, "percentage": 11.89, "elapsed_time": "0:12:51", "remaining_time": "1:35:17"}
28
+ {"current_steps": 280, "total_steps": 2271, "loss": 0.5769, "lr": 1.996926375800548e-05, "epoch": 0.36975899636843845, "percentage": 12.33, "elapsed_time": "0:13:19", "remaining_time": "1:34:46"}
29
+ {"current_steps": 290, "total_steps": 2271, "loss": 0.5563, "lr": 1.9956038325686585e-05, "epoch": 0.3829646748101684, "percentage": 12.77, "elapsed_time": "0:13:47", "remaining_time": "1:34:15"}
30
+ {"current_steps": 300, "total_steps": 2271, "loss": 0.5544, "lr": 1.9940458706059707e-05, "epoch": 0.3961703532518983, "percentage": 13.21, "elapsed_time": "0:14:15", "remaining_time": "1:33:42"}
31
+ {"current_steps": 310, "total_steps": 2271, "loss": 0.5476, "lr": 1.99225285830543e-05, "epoch": 0.40937603169362824, "percentage": 13.65, "elapsed_time": "0:14:43", "remaining_time": "1:33:07"}
32
+ {"current_steps": 320, "total_steps": 2271, "loss": 0.5665, "lr": 1.9902252196395708e-05, "epoch": 0.4225817101353582, "percentage": 14.09, "elapsed_time": "0:15:11", "remaining_time": "1:32:34"}
33
+ {"current_steps": 330, "total_steps": 2271, "loss": 0.5632, "lr": 1.9879634340602646e-05, "epoch": 0.43578738857708815, "percentage": 14.53, "elapsed_time": "0:15:39", "remaining_time": "1:32:06"}
34
+ {"current_steps": 340, "total_steps": 2271, "loss": 0.561, "lr": 1.9854680363853506e-05, "epoch": 0.4489930670188181, "percentage": 14.97, "elapsed_time": "0:16:06", "remaining_time": "1:31:31"}
35
+ {"current_steps": 10, "total_steps": 2250, "loss": 0.8311, "lr": 8.000000000000001e-07, "epoch": 0.013333333333333334, "percentage": 0.44, "elapsed_time": "0:00:11", "remaining_time": "0:42:38"}
36
+ {"current_steps": 20, "total_steps": 2250, "loss": 0.7559, "lr": 1.688888888888889e-06, "epoch": 0.02666666666666667, "percentage": 0.89, "elapsed_time": "0:00:20", "remaining_time": "0:38:42"}
37
+ {"current_steps": 30, "total_steps": 2250, "loss": 0.6699, "lr": 2.577777777777778e-06, "epoch": 0.04, "percentage": 1.33, "elapsed_time": "0:00:29", "remaining_time": "0:36:37"}
38
+ {"current_steps": 40, "total_steps": 2250, "loss": 0.615, "lr": 3.4666666666666672e-06, "epoch": 0.05333333333333334, "percentage": 1.78, "elapsed_time": "0:00:38", "remaining_time": "0:35:23"}
39
+ {"current_steps": 50, "total_steps": 2250, "loss": 0.572, "lr": 4.3555555555555555e-06, "epoch": 0.06666666666666667, "percentage": 2.22, "elapsed_time": "0:00:47", "remaining_time": "0:35:04"}
40
+ {"current_steps": 60, "total_steps": 2250, "loss": 0.5574, "lr": 5.244444444444445e-06, "epoch": 0.08, "percentage": 2.67, "elapsed_time": "0:00:58", "remaining_time": "0:35:22"}
41
+ {"current_steps": 70, "total_steps": 2250, "loss": 0.5414, "lr": 6.133333333333334e-06, "epoch": 0.09333333333333334, "percentage": 3.11, "elapsed_time": "0:01:07", "remaining_time": "0:35:13"}
42
+ {"current_steps": 80, "total_steps": 2250, "loss": 0.5309, "lr": 7.022222222222222e-06, "epoch": 0.10666666666666667, "percentage": 3.56, "elapsed_time": "0:01:17", "remaining_time": "0:35:00"}
43
+ {"current_steps": 90, "total_steps": 2250, "loss": 0.5359, "lr": 7.911111111111112e-06, "epoch": 0.12, "percentage": 4.0, "elapsed_time": "0:01:27", "remaining_time": "0:34:51"}
44
+ {"current_steps": 100, "total_steps": 2250, "loss": 0.541, "lr": 8.8e-06, "epoch": 0.13333333333333333, "percentage": 4.44, "elapsed_time": "0:01:36", "remaining_time": "0:34:41"}
45
+ {"current_steps": 110, "total_steps": 2250, "loss": 0.5434, "lr": 9.688888888888889e-06, "epoch": 0.14666666666666667, "percentage": 4.89, "elapsed_time": "0:01:45", "remaining_time": "0:34:21"}
46
+ {"current_steps": 120, "total_steps": 2250, "loss": 0.5278, "lr": 1.0577777777777778e-05, "epoch": 0.16, "percentage": 5.33, "elapsed_time": "0:01:55", "remaining_time": "0:34:11"}
47
+ {"current_steps": 130, "total_steps": 2250, "loss": 0.5241, "lr": 1.1466666666666668e-05, "epoch": 0.17333333333333334, "percentage": 5.78, "elapsed_time": "0:02:04", "remaining_time": "0:33:53"}
48
+ {"current_steps": 140, "total_steps": 2250, "loss": 0.53, "lr": 1.2355555555555557e-05, "epoch": 0.18666666666666668, "percentage": 6.22, "elapsed_time": "0:02:14", "remaining_time": "0:33:40"}
49
+ {"current_steps": 150, "total_steps": 2250, "loss": 0.5519, "lr": 1.3244444444444447e-05, "epoch": 0.2, "percentage": 6.67, "elapsed_time": "0:02:24", "remaining_time": "0:33:36"}
50
+ {"current_steps": 160, "total_steps": 2250, "loss": 0.5363, "lr": 1.4133333333333334e-05, "epoch": 0.21333333333333335, "percentage": 7.11, "elapsed_time": "0:02:33", "remaining_time": "0:33:26"}
51
+ {"current_steps": 170, "total_steps": 2250, "loss": 0.5445, "lr": 1.5022222222222223e-05, "epoch": 0.22666666666666666, "percentage": 7.56, "elapsed_time": "0:02:43", "remaining_time": "0:33:22"}
52
+ {"current_steps": 180, "total_steps": 2250, "loss": 0.541, "lr": 1.5911111111111113e-05, "epoch": 0.24, "percentage": 8.0, "elapsed_time": "0:02:53", "remaining_time": "0:33:17"}
53
+ {"current_steps": 10, "total_steps": 2250, "loss": 0.831, "lr": 8.000000000000001e-07, "epoch": 0.013333333333333334, "percentage": 0.44, "elapsed_time": "0:00:11", "remaining_time": "0:44:15"}
54
+ {"current_steps": 20, "total_steps": 2250, "loss": 0.7557, "lr": 1.688888888888889e-06, "epoch": 0.02666666666666667, "percentage": 0.89, "elapsed_time": "0:00:21", "remaining_time": "0:39:10"}
55
+ {"current_steps": 30, "total_steps": 2250, "loss": 0.67, "lr": 2.577777777777778e-06, "epoch": 0.04, "percentage": 1.33, "elapsed_time": "0:00:29", "remaining_time": "0:36:57"}
56
+ {"current_steps": 40, "total_steps": 2250, "loss": 0.6148, "lr": 3.4666666666666672e-06, "epoch": 0.05333333333333334, "percentage": 1.78, "elapsed_time": "0:00:38", "remaining_time": "0:35:36"}
57
+ {"current_steps": 50, "total_steps": 2250, "loss": 0.572, "lr": 4.3555555555555555e-06, "epoch": 0.06666666666666667, "percentage": 2.22, "elapsed_time": "0:00:48", "remaining_time": "0:35:14"}
58
+ {"current_steps": 60, "total_steps": 2250, "loss": 0.5575, "lr": 5.244444444444445e-06, "epoch": 0.08, "percentage": 2.67, "elapsed_time": "0:00:58", "remaining_time": "0:35:30"}
59
+ {"current_steps": 70, "total_steps": 2250, "loss": 0.5413, "lr": 6.133333333333334e-06, "epoch": 0.09333333333333334, "percentage": 3.11, "elapsed_time": "0:01:08", "remaining_time": "0:35:20"}
60
+ {"current_steps": 80, "total_steps": 2250, "loss": 0.5311, "lr": 7.022222222222222e-06, "epoch": 0.10666666666666667, "percentage": 3.56, "elapsed_time": "0:01:18", "remaining_time": "0:35:16"}
61
+ {"current_steps": 90, "total_steps": 2250, "loss": 0.536, "lr": 7.911111111111112e-06, "epoch": 0.12, "percentage": 4.0, "elapsed_time": "0:01:27", "remaining_time": "0:35:06"}
62
+ {"current_steps": 100, "total_steps": 2250, "loss": 0.5414, "lr": 8.8e-06, "epoch": 0.13333333333333333, "percentage": 4.44, "elapsed_time": "0:01:37", "remaining_time": "0:34:55"}
63
+ {"current_steps": 110, "total_steps": 2250, "loss": 0.5432, "lr": 9.688888888888889e-06, "epoch": 0.14666666666666667, "percentage": 4.89, "elapsed_time": "0:01:46", "remaining_time": "0:34:33"}
64
+ {"current_steps": 120, "total_steps": 2250, "loss": 0.5278, "lr": 1.0577777777777778e-05, "epoch": 0.16, "percentage": 5.33, "elapsed_time": "0:01:56", "remaining_time": "0:34:24"}
65
+ {"current_steps": 130, "total_steps": 2250, "loss": 0.5232, "lr": 1.1466666666666668e-05, "epoch": 0.17333333333333334, "percentage": 5.78, "elapsed_time": "0:02:05", "remaining_time": "0:34:04"}
66
+ {"current_steps": 140, "total_steps": 2250, "loss": 0.5309, "lr": 1.2355555555555557e-05, "epoch": 0.18666666666666668, "percentage": 6.22, "elapsed_time": "0:02:14", "remaining_time": "0:33:51"}
67
+ {"current_steps": 150, "total_steps": 2250, "loss": 0.552, "lr": 1.3244444444444447e-05, "epoch": 0.2, "percentage": 6.67, "elapsed_time": "0:02:24", "remaining_time": "0:33:46"}
68
+ {"current_steps": 160, "total_steps": 2250, "loss": 0.5365, "lr": 1.4133333333333334e-05, "epoch": 0.21333333333333335, "percentage": 7.11, "elapsed_time": "0:02:34", "remaining_time": "0:33:37"}
69
+ {"current_steps": 170, "total_steps": 2250, "loss": 0.5442, "lr": 1.5022222222222223e-05, "epoch": 0.22666666666666666, "percentage": 7.56, "elapsed_time": "0:02:44", "remaining_time": "0:33:32"}
70
+ {"current_steps": 180, "total_steps": 2250, "loss": 0.5414, "lr": 1.5911111111111113e-05, "epoch": 0.24, "percentage": 8.0, "elapsed_time": "0:02:54", "remaining_time": "0:33:26"}
71
+ {"current_steps": 190, "total_steps": 2250, "loss": 0.5499, "lr": 1.6800000000000002e-05, "epoch": 0.25333333333333335, "percentage": 8.44, "elapsed_time": "0:03:02", "remaining_time": "0:33:01"}
72
+ {"current_steps": 200, "total_steps": 2250, "loss": 0.5573, "lr": 1.768888888888889e-05, "epoch": 0.26666666666666666, "percentage": 8.89, "elapsed_time": "0:03:11", "remaining_time": "0:32:46"}
73
+ {"current_steps": 210, "total_steps": 2250, "loss": 0.5654, "lr": 1.857777777777778e-05, "epoch": 0.28, "percentage": 9.33, "elapsed_time": "0:03:21", "remaining_time": "0:32:38"}
74
+ {"current_steps": 220, "total_steps": 2250, "loss": 0.5797, "lr": 1.9466666666666668e-05, "epoch": 0.29333333333333333, "percentage": 9.78, "elapsed_time": "0:03:30", "remaining_time": "0:32:25"}
75
+ {"current_steps": 230, "total_steps": 2250, "loss": 0.5669, "lr": 1.999980745232294e-05, "epoch": 0.30666666666666664, "percentage": 10.22, "elapsed_time": "0:03:40", "remaining_time": "0:32:15"}
76
+ {"current_steps": 240, "total_steps": 2250, "loss": 0.5642, "lr": 1.9997641376110816e-05, "epoch": 0.32, "percentage": 10.67, "elapsed_time": "0:03:50", "remaining_time": "0:32:07"}
77
+ {"current_steps": 250, "total_steps": 2250, "loss": 0.5738, "lr": 1.999306906216052e-05, "epoch": 0.3333333333333333, "percentage": 11.11, "elapsed_time": "0:03:59", "remaining_time": "0:31:59"}
78
+ {"current_steps": 260, "total_steps": 2250, "loss": 0.564, "lr": 1.998609161093904e-05, "epoch": 0.3466666666666667, "percentage": 11.56, "elapsed_time": "0:04:10", "remaining_time": "0:31:54"}
79
+ {"current_steps": 270, "total_steps": 2250, "loss": 0.5596, "lr": 1.9976710701783116e-05, "epoch": 0.36, "percentage": 12.0, "elapsed_time": "0:04:18", "remaining_time": "0:31:38"}
80
+ {"current_steps": 280, "total_steps": 2250, "loss": 0.552, "lr": 1.9964928592495046e-05, "epoch": 0.37333333333333335, "percentage": 12.44, "elapsed_time": "0:04:27", "remaining_time": "0:31:22"}
81
+ {"current_steps": 290, "total_steps": 2250, "loss": 0.5574, "lr": 1.9950748118799264e-05, "epoch": 0.38666666666666666, "percentage": 12.89, "elapsed_time": "0:04:37", "remaining_time": "0:31:12"}
82
+ {"current_steps": 300, "total_steps": 2250, "loss": 0.5504, "lr": 1.9934172693659854e-05, "epoch": 0.4, "percentage": 13.33, "elapsed_time": "0:04:46", "remaining_time": "0:31:03"}
83
+ {"current_steps": 310, "total_steps": 2250, "loss": 0.5683, "lr": 1.9915206306459117e-05, "epoch": 0.41333333333333333, "percentage": 13.78, "elapsed_time": "0:04:56", "remaining_time": "0:30:52"}
84
+ {"current_steps": 320, "total_steps": 2250, "loss": 0.549, "lr": 1.989385352203739e-05, "epoch": 0.4266666666666667, "percentage": 14.22, "elapsed_time": "0:05:05", "remaining_time": "0:30:43"}
85
+ {"current_steps": 330, "total_steps": 2250, "loss": 0.5514, "lr": 1.9870119479594385e-05, "epoch": 0.44, "percentage": 14.67, "elapsed_time": "0:05:15", "remaining_time": "0:30:37"}
86
+ {"current_steps": 340, "total_steps": 2250, "loss": 0.5585, "lr": 1.984400989145228e-05, "epoch": 0.4533333333333333, "percentage": 15.11, "elapsed_time": "0:05:25", "remaining_time": "0:30:26"}
87
+ {"current_steps": 350, "total_steps": 2250, "loss": 0.5573, "lr": 1.9815531041680894e-05, "epoch": 0.4666666666666667, "percentage": 15.56, "elapsed_time": "0:05:34", "remaining_time": "0:30:16"}
88
+ {"current_steps": 360, "total_steps": 2250, "loss": 0.5587, "lr": 1.9784689784585193e-05, "epoch": 0.48, "percentage": 16.0, "elapsed_time": "0:05:43", "remaining_time": "0:30:02"}
89
+ {"current_steps": 370, "total_steps": 2250, "loss": 0.537, "lr": 1.9751493543055634e-05, "epoch": 0.49333333333333335, "percentage": 16.44, "elapsed_time": "0:05:52", "remaining_time": "0:29:53"}
90
+ {"current_steps": 380, "total_steps": 2250, "loss": 0.5364, "lr": 1.97159503067816e-05, "epoch": 0.5066666666666667, "percentage": 16.89, "elapsed_time": "0:06:02", "remaining_time": "0:29:42"}
91
+ {"current_steps": 390, "total_steps": 2250, "loss": 0.5352, "lr": 1.9678068630328434e-05, "epoch": 0.52, "percentage": 17.33, "elapsed_time": "0:06:12", "remaining_time": "0:29:35"}
92
+ {"current_steps": 400, "total_steps": 2250, "loss": 0.5432, "lr": 1.9637857631078532e-05, "epoch": 0.5333333333333333, "percentage": 17.78, "elapsed_time": "0:06:21", "remaining_time": "0:29:25"}
93
+ {"current_steps": 410, "total_steps": 2250, "loss": 0.5265, "lr": 1.9595326987036963e-05, "epoch": 0.5466666666666666, "percentage": 18.22, "elapsed_time": "0:06:31", "remaining_time": "0:29:17"}
94
+ {"current_steps": 420, "total_steps": 2250, "loss": 0.5322, "lr": 1.9550486934502147e-05, "epoch": 0.56, "percentage": 18.67, "elapsed_time": "0:06:41", "remaining_time": "0:29:08"}
95
+ {"current_steps": 430, "total_steps": 2250, "loss": 0.5381, "lr": 1.9503348265602212e-05, "epoch": 0.5733333333333334, "percentage": 19.11, "elapsed_time": "0:06:50", "remaining_time": "0:28:58"}
96
+ {"current_steps": 440, "total_steps": 2250, "loss": 0.5332, "lr": 1.9453922325697498e-05, "epoch": 0.5866666666666667, "percentage": 19.56, "elapsed_time": "0:07:00", "remaining_time": "0:28:51"}
97
+ {"current_steps": 450, "total_steps": 2250, "loss": 0.547, "lr": 1.940222101064998e-05, "epoch": 0.6, "percentage": 20.0, "elapsed_time": "0:07:09", "remaining_time": "0:28:39"}
98
+ {"current_steps": 460, "total_steps": 2250, "loss": 0.5311, "lr": 1.9348256763960146e-05, "epoch": 0.6133333333333333, "percentage": 20.44, "elapsed_time": "0:07:18", "remaining_time": "0:28:26"}
99
+ {"current_steps": 470, "total_steps": 2250, "loss": 0.538, "lr": 1.92920425737721e-05, "epoch": 0.6266666666666667, "percentage": 20.89, "elapsed_time": "0:07:27", "remaining_time": "0:28:15"}
100
+ {"current_steps": 480, "total_steps": 2250, "loss": 0.5172, "lr": 1.923359196974757e-05, "epoch": 0.64, "percentage": 21.33, "elapsed_time": "0:07:37", "remaining_time": "0:28:06"}
101
+ {"current_steps": 490, "total_steps": 2250, "loss": 0.5241, "lr": 1.9172919019809572e-05, "epoch": 0.6533333333333333, "percentage": 21.78, "elapsed_time": "0:07:46", "remaining_time": "0:27:56"}
102
+ {"current_steps": 500, "total_steps": 2250, "loss": 0.5281, "lr": 1.9110038326756535e-05, "epoch": 0.6666666666666666, "percentage": 22.22, "elapsed_time": "0:07:55", "remaining_time": "0:27:45"}
103
+ {"current_steps": 510, "total_steps": 2250, "loss": 0.5245, "lr": 1.9044965024747703e-05, "epoch": 0.68, "percentage": 22.67, "elapsed_time": "0:08:05", "remaining_time": "0:27:34"}
104
+ {"current_steps": 520, "total_steps": 2250, "loss": 0.5163, "lr": 1.897771477566063e-05, "epoch": 0.6933333333333334, "percentage": 23.11, "elapsed_time": "0:08:14", "remaining_time": "0:27:24"}
105
+ {"current_steps": 530, "total_steps": 2250, "loss": 0.511, "lr": 1.8908303765321662e-05, "epoch": 0.7066666666666667, "percentage": 23.56, "elapsed_time": "0:08:23", "remaining_time": "0:27:15"}
106
+ {"current_steps": 540, "total_steps": 2250, "loss": 0.5138, "lr": 1.883674869961035e-05, "epoch": 0.72, "percentage": 24.0, "elapsed_time": "0:08:33", "remaining_time": "0:27:06"}
107
+ {"current_steps": 550, "total_steps": 2250, "loss": 0.5164, "lr": 1.8763066800438638e-05, "epoch": 0.7333333333333333, "percentage": 24.44, "elapsed_time": "0:08:43", "remaining_time": "0:26:57"}
108
+ {"current_steps": 560, "total_steps": 2250, "loss": 0.5162, "lr": 1.86872758016059e-05, "epoch": 0.7466666666666667, "percentage": 24.89, "elapsed_time": "0:08:52", "remaining_time": "0:26:48"}
109
+ {"current_steps": 570, "total_steps": 2250, "loss": 0.4943, "lr": 1.8609393944530765e-05, "epoch": 0.76, "percentage": 25.33, "elapsed_time": "0:09:02", "remaining_time": "0:26:38"}
110
+ {"current_steps": 580, "total_steps": 2250, "loss": 0.5266, "lr": 1.852943997386075e-05, "epoch": 0.7733333333333333, "percentage": 25.78, "elapsed_time": "0:09:11", "remaining_time": "0:26:27"}
111
+ {"current_steps": 590, "total_steps": 2250, "loss": 0.5098, "lr": 1.84474331329608e-05, "epoch": 0.7866666666666666, "percentage": 26.22, "elapsed_time": "0:09:21", "remaining_time": "0:26:19"}
112
+ {"current_steps": 600, "total_steps": 2250, "loss": 0.5255, "lr": 1.8363393159281766e-05, "epoch": 0.8, "percentage": 26.67, "elapsed_time": "0:09:30", "remaining_time": "0:26:09"}
113
+ {"current_steps": 610, "total_steps": 2250, "loss": 0.4982, "lr": 1.827734027961001e-05, "epoch": 0.8133333333333334, "percentage": 27.11, "elapsed_time": "0:09:40", "remaining_time": "0:26:00"}
114
+ {"current_steps": 620, "total_steps": 2250, "loss": 0.5194, "lr": 1.8189295205199188e-05, "epoch": 0.8266666666666667, "percentage": 27.56, "elapsed_time": "0:09:49", "remaining_time": "0:25:50"}
115
+ {"current_steps": 630, "total_steps": 2250, "loss": 0.4854, "lr": 1.809927912678547e-05, "epoch": 0.84, "percentage": 28.0, "elapsed_time": "0:09:59", "remaining_time": "0:25:41"}
116
+ {"current_steps": 640, "total_steps": 2250, "loss": 0.5036, "lr": 1.8007313709487334e-05, "epoch": 0.8533333333333334, "percentage": 28.44, "elapsed_time": "0:10:08", "remaining_time": "0:25:31"}
117
+ {"current_steps": 650, "total_steps": 2250, "loss": 0.4899, "lr": 1.7913421087591204e-05, "epoch": 0.8666666666666667, "percentage": 28.89, "elapsed_time": "0:10:17", "remaining_time": "0:25:19"}
118
+ {"current_steps": 660, "total_steps": 2250, "loss": 0.5074, "lr": 1.7817623859224133e-05, "epoch": 0.88, "percentage": 29.33, "elapsed_time": "0:10:26", "remaining_time": "0:25:08"}
119
+ {"current_steps": 670, "total_steps": 2250, "loss": 0.4988, "lr": 1.7719945080914902e-05, "epoch": 0.8933333333333333, "percentage": 29.78, "elapsed_time": "0:10:35", "remaining_time": "0:24:58"}
120
+ {"current_steps": 680, "total_steps": 2250, "loss": 0.5002, "lr": 1.7620408262044753e-05, "epoch": 0.9066666666666666, "percentage": 30.22, "elapsed_time": "0:10:44", "remaining_time": "0:24:47"}
121
+ {"current_steps": 690, "total_steps": 2250, "loss": 0.5032, "lr": 1.7519037359189126e-05, "epoch": 0.92, "percentage": 30.67, "elapsed_time": "0:10:53", "remaining_time": "0:24:38"}
122
+ {"current_steps": 700, "total_steps": 2250, "loss": 0.4847, "lr": 1.7415856770351797e-05, "epoch": 0.9333333333333333, "percentage": 31.11, "elapsed_time": "0:11:03", "remaining_time": "0:24:28"}
123
+ {"current_steps": 710, "total_steps": 2250, "loss": 0.5056, "lr": 1.731089132909274e-05, "epoch": 0.9466666666666667, "percentage": 31.56, "elapsed_time": "0:11:12", "remaining_time": "0:24:18"}
124
+ {"current_steps": 720, "total_steps": 2250, "loss": 0.4978, "lr": 1.720416629855117e-05, "epoch": 0.96, "percentage": 32.0, "elapsed_time": "0:11:22", "remaining_time": "0:24:10"}
125
+ {"current_steps": 730, "total_steps": 2250, "loss": 0.4895, "lr": 1.709570736536521e-05, "epoch": 0.9733333333333334, "percentage": 32.44, "elapsed_time": "0:11:32", "remaining_time": "0:24:01"}
126
+ {"current_steps": 740, "total_steps": 2250, "loss": 0.4892, "lr": 1.6985540633489614e-05, "epoch": 0.9866666666666667, "percentage": 32.89, "elapsed_time": "0:11:41", "remaining_time": "0:23:51"}
127
+ {"current_steps": 750, "total_steps": 2250, "loss": 0.4972, "lr": 1.6873692617913038e-05, "epoch": 1.0, "percentage": 33.33, "elapsed_time": "0:11:50", "remaining_time": "0:23:41"}
128
+ {"current_steps": 760, "total_steps": 2250, "loss": 0.3175, "lr": 1.6760190238276418e-05, "epoch": 1.0133333333333334, "percentage": 33.78, "elapsed_time": "0:12:00", "remaining_time": "0:23:32"}
129
+ {"current_steps": 770, "total_steps": 2250, "loss": 0.3167, "lr": 1.6645060812393926e-05, "epoch": 1.0266666666666666, "percentage": 34.22, "elapsed_time": "0:12:09", "remaining_time": "0:23:22"}
130
+ {"current_steps": 780, "total_steps": 2250, "loss": 0.3213, "lr": 1.6528332049678103e-05, "epoch": 1.04, "percentage": 34.67, "elapsed_time": "0:12:19", "remaining_time": "0:23:12"}
131
+ {"current_steps": 790, "total_steps": 2250, "loss": 0.3154, "lr": 1.6410032044470735e-05, "epoch": 1.0533333333333332, "percentage": 35.11, "elapsed_time": "0:12:28", "remaining_time": "0:23:02"}
132
+ {"current_steps": 800, "total_steps": 2250, "loss": 0.2926, "lr": 1.6290189269281124e-05, "epoch": 1.0666666666666667, "percentage": 35.56, "elapsed_time": "0:12:38", "remaining_time": "0:22:54"}
133
+ {"current_steps": 810, "total_steps": 2250, "loss": 0.3249, "lr": 1.616883256793327e-05, "epoch": 1.08, "percentage": 36.0, "elapsed_time": "0:12:47", "remaining_time": "0:22:44"}
134
+ {"current_steps": 820, "total_steps": 2250, "loss": 0.3133, "lr": 1.6045991148623752e-05, "epoch": 1.0933333333333333, "percentage": 36.44, "elapsed_time": "0:12:57", "remaining_time": "0:22:35"}
135
+ {"current_steps": 830, "total_steps": 2250, "loss": 0.3063, "lr": 1.5921694576891847e-05, "epoch": 1.1066666666666667, "percentage": 36.89, "elapsed_time": "0:13:06", "remaining_time": "0:22:26"}
136
+ {"current_steps": 840, "total_steps": 2250, "loss": 0.2963, "lr": 1.579597276850374e-05, "epoch": 1.12, "percentage": 37.33, "elapsed_time": "0:13:16", "remaining_time": "0:22:16"}
137
+ {"current_steps": 850, "total_steps": 2250, "loss": 0.3153, "lr": 1.566885598225231e-05, "epoch": 1.1333333333333333, "percentage": 37.78, "elapsed_time": "0:13:25", "remaining_time": "0:22:06"}
138
+ {"current_steps": 860, "total_steps": 2250, "loss": 0.3058, "lr": 1.5540374812674496e-05, "epoch": 1.1466666666666667, "percentage": 38.22, "elapsed_time": "0:13:34", "remaining_time": "0:21:57"}
139
+ {"current_steps": 870, "total_steps": 2250, "loss": 0.3083, "lr": 1.5410560182687736e-05, "epoch": 1.16, "percentage": 38.67, "elapsed_time": "0:13:44", "remaining_time": "0:21:47"}
140
+ {"current_steps": 880, "total_steps": 2250, "loss": 0.3122, "lr": 1.5279443336147437e-05, "epoch": 1.1733333333333333, "percentage": 39.11, "elapsed_time": "0:13:53", "remaining_time": "0:21:36"}
141
+ {"current_steps": 890, "total_steps": 2250, "loss": 0.3307, "lr": 1.5147055830327206e-05, "epoch": 1.1866666666666668, "percentage": 39.56, "elapsed_time": "0:14:02", "remaining_time": "0:21:26"}
142
+ {"current_steps": 900, "total_steps": 2250, "loss": 0.3046, "lr": 1.5013429528323603e-05, "epoch": 1.2, "percentage": 40.0, "elapsed_time": "0:14:11", "remaining_time": "0:21:17"}
143
+ {"current_steps": 910, "total_steps": 2250, "loss": 0.297, "lr": 1.4878596591387329e-05, "epoch": 1.2133333333333334, "percentage": 40.44, "elapsed_time": "0:14:21", "remaining_time": "0:21:08"}
144
+ {"current_steps": 920, "total_steps": 2250, "loss": 0.3036, "lr": 1.474258947118263e-05, "epoch": 1.2266666666666666, "percentage": 40.89, "elapsed_time": "0:14:30", "remaining_time": "0:20:58"}
145
+ {"current_steps": 930, "total_steps": 2250, "loss": 0.2973, "lr": 1.4605440901976815e-05, "epoch": 1.24, "percentage": 41.33, "elapsed_time": "0:14:40", "remaining_time": "0:20:49"}
146
+ {"current_steps": 940, "total_steps": 2250, "loss": 0.2977, "lr": 1.4467183892761769e-05, "epoch": 1.2533333333333334, "percentage": 41.78, "elapsed_time": "0:14:49", "remaining_time": "0:20:39"}
147
+ {"current_steps": 950, "total_steps": 2250, "loss": 0.3039, "lr": 1.4327851719309298e-05, "epoch": 1.2666666666666666, "percentage": 42.22, "elapsed_time": "0:14:58", "remaining_time": "0:20:30"}
148
+ {"current_steps": 960, "total_steps": 2250, "loss": 0.2982, "lr": 1.4187477916162338e-05, "epoch": 1.28, "percentage": 42.67, "elapsed_time": "0:15:08", "remaining_time": "0:20:20"}
149
+ {"current_steps": 970, "total_steps": 2250, "loss": 0.3155, "lr": 1.4046096268563814e-05, "epoch": 1.2933333333333334, "percentage": 43.11, "elapsed_time": "0:15:17", "remaining_time": "0:20:10"}
150
+ {"current_steps": 980, "total_steps": 2250, "loss": 0.3042, "lr": 1.3903740804325193e-05, "epoch": 1.3066666666666666, "percentage": 43.56, "elapsed_time": "0:15:26", "remaining_time": "0:20:01"}
151
+ {"current_steps": 990, "total_steps": 2250, "loss": 0.2994, "lr": 1.376044578563668e-05, "epoch": 1.32, "percentage": 44.0, "elapsed_time": "0:15:36", "remaining_time": "0:19:51"}
152
+ {"current_steps": 1000, "total_steps": 2250, "loss": 0.2917, "lr": 1.3616245700820922e-05, "epoch": 1.3333333333333333, "percentage": 44.44, "elapsed_time": "0:15:45", "remaining_time": "0:19:41"}
153
+ {"current_steps": 1010, "total_steps": 2250, "loss": 0.3026, "lr": 1.3471175256032406e-05, "epoch": 1.3466666666666667, "percentage": 44.89, "elapsed_time": "0:15:54", "remaining_time": "0:19:31"}
154
+ {"current_steps": 1020, "total_steps": 2250, "loss": 0.3068, "lr": 1.33252693669043e-05, "epoch": 1.3599999999999999, "percentage": 45.33, "elapsed_time": "0:16:03", "remaining_time": "0:19:21"}
155
+ {"current_steps": 1030, "total_steps": 2250, "loss": 0.3015, "lr": 1.317856315014498e-05, "epoch": 1.3733333333333333, "percentage": 45.78, "elapsed_time": "0:16:12", "remaining_time": "0:19:11"}
156
+ {"current_steps": 1040, "total_steps": 2250, "loss": 0.3071, "lr": 1.3031091915086114e-05, "epoch": 1.3866666666666667, "percentage": 46.22, "elapsed_time": "0:16:22", "remaining_time": "0:19:02"}
157
+ {"current_steps": 1050, "total_steps": 2250, "loss": 0.2935, "lr": 1.2882891155184375e-05, "epoch": 1.4, "percentage": 46.67, "elapsed_time": "0:16:31", "remaining_time": "0:18:52"}
158
+ {"current_steps": 1060, "total_steps": 2250, "loss": 0.2918, "lr": 1.2733996539478883e-05, "epoch": 1.4133333333333333, "percentage": 47.11, "elapsed_time": "0:16:40", "remaining_time": "0:18:43"}
159
+ {"current_steps": 1070, "total_steps": 2250, "loss": 0.2885, "lr": 1.2584443904006345e-05, "epoch": 1.4266666666666667, "percentage": 47.56, "elapsed_time": "0:16:50", "remaining_time": "0:18:34"}
160
+ {"current_steps": 1080, "total_steps": 2250, "loss": 0.2965, "lr": 1.2434269243176018e-05, "epoch": 1.44, "percentage": 48.0, "elapsed_time": "0:16:59", "remaining_time": "0:18:24"}
161
+ {"current_steps": 1090, "total_steps": 2250, "loss": 0.2918, "lr": 1.2283508701106559e-05, "epoch": 1.4533333333333334, "percentage": 48.44, "elapsed_time": "0:17:09", "remaining_time": "0:18:15"}
162
+ {"current_steps": 1100, "total_steps": 2250, "loss": 0.3042, "lr": 1.2132198562926845e-05, "epoch": 1.4666666666666668, "percentage": 48.89, "elapsed_time": "0:17:18", "remaining_time": "0:18:06"}
163
+ {"current_steps": 1110, "total_steps": 2250, "loss": 0.2882, "lr": 1.1980375246042832e-05, "epoch": 1.48, "percentage": 49.33, "elapsed_time": "0:17:27", "remaining_time": "0:17:56"}
164
+ {"current_steps": 1120, "total_steps": 2250, "loss": 0.2986, "lr": 1.1828075291372616e-05, "epoch": 1.4933333333333334, "percentage": 49.78, "elapsed_time": "0:17:38", "remaining_time": "0:17:47"}
165
+ {"current_steps": 1130, "total_steps": 2250, "loss": 0.2882, "lr": 1.1675335354551726e-05, "epoch": 1.5066666666666668, "percentage": 50.22, "elapsed_time": "0:17:46", "remaining_time": "0:17:37"}
166
+ {"current_steps": 1140, "total_steps": 2250, "loss": 0.3067, "lr": 1.1522192197110833e-05, "epoch": 1.52, "percentage": 50.67, "elapsed_time": "0:17:56", "remaining_time": "0:17:27"}
167
+ {"current_steps": 1150, "total_steps": 2250, "loss": 0.2927, "lr": 1.1368682677627971e-05, "epoch": 1.5333333333333332, "percentage": 51.11, "elapsed_time": "0:18:05", "remaining_time": "0:17:18"}
168
+ {"current_steps": 1160, "total_steps": 2250, "loss": 0.2796, "lr": 1.1214843742857382e-05, "epoch": 1.5466666666666666, "percentage": 51.56, "elapsed_time": "0:18:15", "remaining_time": "0:17:09"}
169
+ {"current_steps": 1170, "total_steps": 2250, "loss": 0.3046, "lr": 1.1060712418837159e-05, "epoch": 1.56, "percentage": 52.0, "elapsed_time": "0:18:25", "remaining_time": "0:17:00"}
170
+ {"current_steps": 1180, "total_steps": 2250, "loss": 0.2739, "lr": 1.0906325801977804e-05, "epoch": 1.5733333333333333, "percentage": 52.44, "elapsed_time": "0:18:34", "remaining_time": "0:16:50"}
171
+ {"current_steps": 1190, "total_steps": 2250, "loss": 0.2732, "lr": 1.0751721050133825e-05, "epoch": 1.5866666666666667, "percentage": 52.89, "elapsed_time": "0:18:43", "remaining_time": "0:16:41"}
172
+ {"current_steps": 1200, "total_steps": 2250, "loss": 0.2896, "lr": 1.0596935373660595e-05, "epoch": 1.6, "percentage": 53.33, "elapsed_time": "0:18:53", "remaining_time": "0:16:32"}
173
+ {"current_steps": 1210, "total_steps": 2250, "loss": 0.2922, "lr": 1.0442006026458506e-05, "epoch": 1.6133333333333333, "percentage": 53.78, "elapsed_time": "0:19:03", "remaining_time": "0:16:23"}
174
+ {"current_steps": 1220, "total_steps": 2250, "loss": 0.278, "lr": 1.0286970297006709e-05, "epoch": 1.6266666666666667, "percentage": 54.22, "elapsed_time": "0:19:12", "remaining_time": "0:16:13"}
175
+ {"current_steps": 1230, "total_steps": 2250, "loss": 0.2798, "lr": 1.013186549938849e-05, "epoch": 1.6400000000000001, "percentage": 54.67, "elapsed_time": "0:19:22", "remaining_time": "0:16:04"}
176
+ {"current_steps": 1240, "total_steps": 2250, "loss": 0.2841, "lr": 9.976728964310499e-06, "epoch": 1.6533333333333333, "percentage": 55.11, "elapsed_time": "0:19:32", "remaining_time": "0:15:54"}
177
+ {"current_steps": 1250, "total_steps": 2250, "loss": 0.2765, "lr": 9.821598030117973e-06, "epoch": 1.6666666666666665, "percentage": 55.56, "elapsed_time": "0:19:41", "remaining_time": "0:15:45"}
178
+ {"current_steps": 1260, "total_steps": 2250, "loss": 0.2731, "lr": 9.666510033808131e-06, "epoch": 1.6800000000000002, "percentage": 56.0, "elapsed_time": "0:19:51", "remaining_time": "0:15:35"}
179
+ {"current_steps": 1270, "total_steps": 2250, "loss": 0.2658, "lr": 9.511502302043867e-06, "epoch": 1.6933333333333334, "percentage": 56.44, "elapsed_time": "0:20:00", "remaining_time": "0:15:26"}
180
+ {"current_steps": 1280, "total_steps": 2250, "loss": 0.2888, "lr": 9.356612142169978e-06, "epoch": 1.7066666666666666, "percentage": 56.89, "elapsed_time": "0:20:08", "remaining_time": "0:15:15"}
181
+ {"current_steps": 1290, "total_steps": 2250, "loss": 0.2797, "lr": 9.201876833233981e-06, "epoch": 1.72, "percentage": 57.33, "elapsed_time": "0:20:17", "remaining_time": "0:15:06"}
182
+ {"current_steps": 1300, "total_steps": 2250, "loss": 0.2965, "lr": 9.047333617013786e-06, "epoch": 1.7333333333333334, "percentage": 57.78, "elapsed_time": "0:20:27", "remaining_time": "0:14:57"}
183
+ {"current_steps": 1310, "total_steps": 2250, "loss": 0.2648, "lr": 8.893019689054331e-06, "epoch": 1.7466666666666666, "percentage": 58.22, "elapsed_time": "0:20:37", "remaining_time": "0:14:47"}
184
+ {"current_steps": 1320, "total_steps": 2250, "loss": 0.2702, "lr": 8.738972189715333e-06, "epoch": 1.76, "percentage": 58.67, "elapsed_time": "0:20:46", "remaining_time": "0:14:38"}
185
+ {"current_steps": 1330, "total_steps": 2250, "loss": 0.2711, "lr": 8.585228195232311e-06, "epoch": 1.7733333333333334, "percentage": 59.11, "elapsed_time": "0:20:56", "remaining_time": "0:14:29"}
186
+ {"current_steps": 1340, "total_steps": 2250, "loss": 0.2674, "lr": 8.431824708793088e-06, "epoch": 1.7866666666666666, "percentage": 59.56, "elapsed_time": "0:21:06", "remaining_time": "0:14:19"}
187
+ {"current_steps": 1350, "total_steps": 2250, "loss": 0.2661, "lr": 8.27879865163181e-06, "epoch": 1.8, "percentage": 60.0, "elapsed_time": "0:21:15", "remaining_time": "0:14:10"}
188
+ {"current_steps": 1360, "total_steps": 2250, "loss": 0.2695, "lr": 8.126186854142752e-06, "epoch": 1.8133333333333335, "percentage": 60.44, "elapsed_time": "0:21:24", "remaining_time": "0:14:00"}
189
+ {"current_steps": 1370, "total_steps": 2250, "loss": 0.2856, "lr": 7.974026047015951e-06, "epoch": 1.8266666666666667, "percentage": 60.89, "elapsed_time": "0:21:33", "remaining_time": "0:13:51"}
190
+ {"current_steps": 1380, "total_steps": 2250, "loss": 0.2623, "lr": 7.822352852396816e-06, "epoch": 1.8399999999999999, "percentage": 61.33, "elapsed_time": "0:21:43", "remaining_time": "0:13:42"}
191
+ {"current_steps": 1390, "total_steps": 2250, "loss": 0.2695, "lr": 7.671203775071942e-06, "epoch": 1.8533333333333335, "percentage": 61.78, "elapsed_time": "0:21:53", "remaining_time": "0:13:32"}
192
+ {"current_steps": 1400, "total_steps": 2250, "loss": 0.2715, "lr": 7.520615193683073e-06, "epoch": 1.8666666666666667, "percentage": 62.22, "elapsed_time": "0:22:03", "remaining_time": "0:13:23"}
193
+ {"current_steps": 1410, "total_steps": 2250, "loss": 0.2622, "lr": 7.370623351971491e-06, "epoch": 1.88, "percentage": 62.67, "elapsed_time": "0:22:12", "remaining_time": "0:13:13"}
194
+ {"current_steps": 1420, "total_steps": 2250, "loss": 0.2636, "lr": 7.221264350054855e-06, "epoch": 1.8933333333333333, "percentage": 63.11, "elapsed_time": "0:22:23", "remaining_time": "0:13:05"}
195
+ {"current_steps": 1430, "total_steps": 2250, "loss": 0.254, "lr": 7.072574135738585e-06, "epoch": 1.9066666666666667, "percentage": 63.56, "elapsed_time": "0:22:32", "remaining_time": "0:12:55"}
196
+ {"current_steps": 1440, "total_steps": 2250, "loss": 0.2483, "lr": 6.924588495863964e-06, "epoch": 1.92, "percentage": 64.0, "elapsed_time": "0:22:41", "remaining_time": "0:12:45"}
197
+ {"current_steps": 1450, "total_steps": 2250, "loss": 0.2781, "lr": 6.777343047694891e-06, "epoch": 1.9333333333333333, "percentage": 64.44, "elapsed_time": "0:22:51", "remaining_time": "0:12:36"}
198
+ {"current_steps": 1460, "total_steps": 2250, "loss": 0.2547, "lr": 6.630873230345538e-06, "epoch": 1.9466666666666668, "percentage": 64.89, "elapsed_time": "0:23:00", "remaining_time": "0:12:27"}
199
+ {"current_steps": 1470, "total_steps": 2250, "loss": 0.2601, "lr": 6.485214296250836e-06, "epoch": 1.96, "percentage": 65.33, "elapsed_time": "0:23:10", "remaining_time": "0:12:17"}
200
+ {"current_steps": 1480, "total_steps": 2250, "loss": 0.2603, "lr": 6.340401302681879e-06, "epoch": 1.9733333333333334, "percentage": 65.78, "elapsed_time": "0:23:20", "remaining_time": "0:12:08"}
201
+ {"current_steps": 1490, "total_steps": 2250, "loss": 0.2553, "lr": 6.196469103308356e-06, "epoch": 1.9866666666666668, "percentage": 66.22, "elapsed_time": "0:23:29", "remaining_time": "0:11:58"}
202
+ {"current_steps": 1500, "total_steps": 2250, "loss": 0.2472, "lr": 6.053452339809917e-06, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "0:23:37", "remaining_time": "0:11:48"}
203
+ {"current_steps": 1510, "total_steps": 2250, "loss": 0.1039, "lr": 5.911385433538621e-06, "epoch": 2.013333333333333, "percentage": 67.11, "elapsed_time": "0:23:47", "remaining_time": "0:11:39"}
204
+ {"current_steps": 1520, "total_steps": 2250, "loss": 0.1015, "lr": 5.77030257723437e-06, "epoch": 2.026666666666667, "percentage": 67.56, "elapsed_time": "0:23:56", "remaining_time": "0:11:30"}
205
+ {"current_steps": 1530, "total_steps": 2250, "loss": 0.0983, "lr": 5.630237726795388e-06, "epoch": 2.04, "percentage": 68.0, "elapsed_time": "0:24:06", "remaining_time": "0:11:20"}
206
+ {"current_steps": 1540, "total_steps": 2250, "loss": 0.0967, "lr": 5.491224593105695e-06, "epoch": 2.0533333333333332, "percentage": 68.44, "elapsed_time": "0:24:15", "remaining_time": "0:11:11"}
207
+ {"current_steps": 1550, "total_steps": 2250, "loss": 0.102, "lr": 5.353296633921554e-06, "epoch": 2.066666666666667, "percentage": 68.89, "elapsed_time": "0:24:24", "remaining_time": "0:11:01"}
208
+ {"current_steps": 1560, "total_steps": 2250, "loss": 0.0952, "lr": 5.216487045818846e-06, "epoch": 2.08, "percentage": 69.33, "elapsed_time": "0:24:34", "remaining_time": "0:10:52"}
209
+ {"current_steps": 1570, "total_steps": 2250, "loss": 0.0988, "lr": 5.080828756203294e-06, "epoch": 2.0933333333333333, "percentage": 69.78, "elapsed_time": "0:24:44", "remaining_time": "0:10:43"}
210
+ {"current_steps": 1580, "total_steps": 2250, "loss": 0.092, "lr": 4.946354415385473e-06, "epoch": 2.1066666666666665, "percentage": 70.22, "elapsed_time": "0:24:53", "remaining_time": "0:10:33"}
211
+ {"current_steps": 1590, "total_steps": 2250, "loss": 0.0952, "lr": 4.8130963887225205e-06, "epoch": 2.12, "percentage": 70.67, "elapsed_time": "0:25:02", "remaining_time": "0:10:23"}
212
+ {"current_steps": 1600, "total_steps": 2250, "loss": 0.0932, "lr": 4.681086748828424e-06, "epoch": 2.1333333333333333, "percentage": 71.11, "elapsed_time": "0:25:11", "remaining_time": "0:10:14"}
213
+ {"current_steps": 1610, "total_steps": 2250, "loss": 0.0952, "lr": 4.5503572678547754e-06, "epoch": 2.1466666666666665, "percentage": 71.56, "elapsed_time": "0:25:20", "remaining_time": "0:10:04"}
214
+ {"current_steps": 1620, "total_steps": 2250, "loss": 0.0981, "lr": 4.420939409843823e-06, "epoch": 2.16, "percentage": 72.0, "elapsed_time": "0:25:29", "remaining_time": "0:09:54"}
215
+ {"current_steps": 1630, "total_steps": 2250, "loss": 0.0968, "lr": 4.292864323155684e-06, "epoch": 2.1733333333333333, "percentage": 72.44, "elapsed_time": "0:25:39", "remaining_time": "0:09:45"}
216
+ {"current_steps": 1640, "total_steps": 2250, "loss": 0.0894, "lr": 4.166162832971551e-06, "epoch": 2.1866666666666665, "percentage": 72.89, "elapsed_time": "0:25:48", "remaining_time": "0:09:35"}
217
+ {"current_steps": 1650, "total_steps": 2250, "loss": 0.0925, "lr": 4.04086543387466e-06, "epoch": 2.2, "percentage": 73.33, "elapsed_time": "0:25:57", "remaining_time": "0:09:26"}
218
+ {"current_steps": 1660, "total_steps": 2250, "loss": 0.0945, "lr": 3.917002282510854e-06, "epoch": 2.2133333333333334, "percentage": 73.78, "elapsed_time": "0:26:07", "remaining_time": "0:09:17"}
219
+ {"current_steps": 1670, "total_steps": 2250, "loss": 0.0974, "lr": 3.79460319033046e-06, "epoch": 2.2266666666666666, "percentage": 74.22, "elapsed_time": "0:26:16", "remaining_time": "0:09:07"}
220
+ {"current_steps": 1680, "total_steps": 2250, "loss": 0.0907, "lr": 3.6736976164132245e-06, "epoch": 2.24, "percentage": 74.67, "elapsed_time": "0:26:25", "remaining_time": "0:08:57"}
221
+ {"current_steps": 1690, "total_steps": 2250, "loss": 0.0904, "lr": 3.554314660378133e-06, "epoch": 2.2533333333333334, "percentage": 75.11, "elapsed_time": "0:26:34", "remaining_time": "0:08:48"}
222
+ {"current_steps": 1700, "total_steps": 2250, "loss": 0.0838, "lr": 3.4364830553796614e-06, "epoch": 2.2666666666666666, "percentage": 75.56, "elapsed_time": "0:26:43", "remaining_time": "0:08:38"}
223
+ {"current_steps": 1710, "total_steps": 2250, "loss": 0.0896, "lr": 3.3202311611922878e-06, "epoch": 2.2800000000000002, "percentage": 76.0, "elapsed_time": "0:26:52", "remaining_time": "0:08:29"}
224
+ {"current_steps": 1720, "total_steps": 2250, "loss": 0.0862, "lr": 3.2055869573848374e-06, "epoch": 2.2933333333333334, "percentage": 76.44, "elapsed_time": "0:27:02", "remaining_time": "0:08:19"}
225
+ {"current_steps": 1730, "total_steps": 2250, "loss": 0.0964, "lr": 3.092578036586329e-06, "epoch": 2.3066666666666666, "percentage": 76.89, "elapsed_time": "0:27:11", "remaining_time": "0:08:10"}
226
+ {"current_steps": 1740, "total_steps": 2250, "loss": 0.0885, "lr": 2.9812315978450014e-06, "epoch": 2.32, "percentage": 77.33, "elapsed_time": "0:27:22", "remaining_time": "0:08:01"}
227
+ {"current_steps": 1750, "total_steps": 2250, "loss": 0.0937, "lr": 2.8715744400819976e-06, "epoch": 2.3333333333333335, "percentage": 77.78, "elapsed_time": "0:27:31", "remaining_time": "0:07:51"}
228
+ {"current_steps": 1760, "total_steps": 2250, "loss": 0.087, "lr": 2.7636329556414076e-06, "epoch": 2.3466666666666667, "percentage": 78.22, "elapsed_time": "0:27:40", "remaining_time": "0:07:42"}
229
+ {"current_steps": 1770, "total_steps": 2250, "loss": 0.0883, "lr": 2.6574331239381446e-06, "epoch": 2.36, "percentage": 78.67, "elapsed_time": "0:27:51", "remaining_time": "0:07:33"}
230
+ {"current_steps": 1780, "total_steps": 2250, "loss": 0.0907, "lr": 2.5530005052051742e-06, "epoch": 2.3733333333333335, "percentage": 79.11, "elapsed_time": "0:28:00", "remaining_time": "0:07:23"}
231
+ {"current_steps": 1790, "total_steps": 2250, "loss": 0.0871, "lr": 2.4503602343417145e-06, "epoch": 2.3866666666666667, "percentage": 79.56, "elapsed_time": "0:28:09", "remaining_time": "0:07:14"}
232
+ {"current_steps": 1800, "total_steps": 2250, "loss": 0.0863, "lr": 2.3495370148637097e-06, "epoch": 2.4, "percentage": 80.0, "elapsed_time": "0:28:19", "remaining_time": "0:07:04"}
233
+ {"current_steps": 1810, "total_steps": 2250, "loss": 0.0892, "lr": 2.2505551129582047e-06, "epoch": 2.413333333333333, "percentage": 80.44, "elapsed_time": "0:28:28", "remaining_time": "0:06:55"}
234
+ {"current_steps": 1820, "total_steps": 2250, "loss": 0.0864, "lr": 2.1534383516429292e-06, "epoch": 2.4266666666666667, "percentage": 80.89, "elapsed_time": "0:28:38", "remaining_time": "0:06:45"}
235
+ {"current_steps": 1830, "total_steps": 2250, "loss": 0.0891, "lr": 2.058210105032562e-06, "epoch": 2.44, "percentage": 81.33, "elapsed_time": "0:28:46", "remaining_time": "0:06:36"}
236
+ {"current_steps": 1840, "total_steps": 2250, "loss": 0.082, "lr": 1.964893292713049e-06, "epoch": 2.453333333333333, "percentage": 81.78, "elapsed_time": "0:28:56", "remaining_time": "0:06:26"}
237
+ {"current_steps": 1850, "total_steps": 2250, "loss": 0.0816, "lr": 1.873510374225289e-06, "epoch": 2.466666666666667, "percentage": 82.22, "elapsed_time": "0:29:05", "remaining_time": "0:06:17"}
238
+ {"current_steps": 1860, "total_steps": 2250, "loss": 0.0864, "lr": 1.784083343659575e-06, "epoch": 2.48, "percentage": 82.67, "elapsed_time": "0:29:14", "remaining_time": "0:06:07"}
239
+ {"current_steps": 1870, "total_steps": 2250, "loss": 0.0835, "lr": 1.6966337243620267e-06, "epoch": 2.493333333333333, "percentage": 83.11, "elapsed_time": "0:29:25", "remaining_time": "0:05:58"}
240
+ {"current_steps": 1880, "total_steps": 2250, "loss": 0.0856, "lr": 1.6111825637543522e-06, "epoch": 2.506666666666667, "percentage": 83.56, "elapsed_time": "0:29:34", "remaining_time": "0:05:49"}
241
+ {"current_steps": 1890, "total_steps": 2250, "loss": 0.0809, "lr": 1.5277504282681454e-06, "epoch": 2.52, "percentage": 84.0, "elapsed_time": "0:29:43", "remaining_time": "0:05:39"}
242
+ {"current_steps": 1900, "total_steps": 2250, "loss": 0.0799, "lr": 1.446357398394934e-06, "epoch": 2.533333333333333, "percentage": 84.44, "elapsed_time": "0:29:52", "remaining_time": "0:05:30"}
243
+ {"current_steps": 1910, "total_steps": 2250, "loss": 0.0851, "lr": 1.36702306385321e-06, "epoch": 2.546666666666667, "percentage": 84.89, "elapsed_time": "0:30:02", "remaining_time": "0:05:20"}
244
+ {"current_steps": 1920, "total_steps": 2250, "loss": 0.0857, "lr": 1.2897665188735531e-06, "epoch": 2.56, "percentage": 85.33, "elapsed_time": "0:30:10", "remaining_time": "0:05:11"}
245
+ {"current_steps": 1930, "total_steps": 2250, "loss": 0.0858, "lr": 1.2146063576030265e-06, "epoch": 2.5733333333333333, "percentage": 85.78, "elapsed_time": "0:30:20", "remaining_time": "0:05:01"}
246
+ {"current_steps": 1940, "total_steps": 2250, "loss": 0.0849, "lr": 1.141560669629933e-06, "epoch": 2.586666666666667, "percentage": 86.22, "elapsed_time": "0:30:30", "remaining_time": "0:04:52"}
247
+ {"current_steps": 1950, "total_steps": 2250, "loss": 0.0803, "lr": 1.0706470356299914e-06, "epoch": 2.6, "percentage": 86.67, "elapsed_time": "0:30:40", "remaining_time": "0:04:43"}
248
+ {"current_steps": 1960, "total_steps": 2250, "loss": 0.0781, "lr": 1.0018825231350203e-06, "epoch": 2.6133333333333333, "percentage": 87.11, "elapsed_time": "0:30:50", "remaining_time": "0:04:33"}
249
+ {"current_steps": 1970, "total_steps": 2250, "loss": 0.0847, "lr": 9.352836824251089e-07, "epoch": 2.626666666666667, "percentage": 87.56, "elapsed_time": "0:31:00", "remaining_time": "0:04:24"}
250
+ {"current_steps": 1980, "total_steps": 2250, "loss": 0.078, "lr": 8.708665425452878e-07, "epoch": 2.64, "percentage": 88.0, "elapsed_time": "0:31:10", "remaining_time": "0:04:15"}
251
+ {"current_steps": 1990, "total_steps": 2250, "loss": 0.0781, "lr": 8.086466074476562e-07, "epoch": 2.6533333333333333, "percentage": 88.44, "elapsed_time": "0:31:19", "remaining_time": "0:04:05"}
252
+ {"current_steps": 2000, "total_steps": 2250, "loss": 0.0848, "lr": 7.486388522598697e-07, "epoch": 2.6666666666666665, "percentage": 88.89, "elapsed_time": "0:31:29", "remaining_time": "0:03:56"}
253
+ {"current_steps": 2010, "total_steps": 2250, "loss": 0.0826, "lr": 6.908577196809341e-07, "epoch": 2.68, "percentage": 89.33, "elapsed_time": "0:31:38", "remaining_time": "0:03:46"}
254
+ {"current_steps": 2020, "total_steps": 2250, "loss": 0.0776, "lr": 6.353171165051109e-07, "epoch": 2.6933333333333334, "percentage": 89.78, "elapsed_time": "0:31:47", "remaining_time": "0:03:37"}
255
+ {"current_steps": 2030, "total_steps": 2250, "loss": 0.0782, "lr": 5.820304102748387e-07, "epoch": 2.7066666666666666, "percentage": 90.22, "elapsed_time": "0:31:57", "remaining_time": "0:03:27"}
256
+ {"current_steps": 2040, "total_steps": 2250, "loss": 0.0871, "lr": 5.310104260634175e-07, "epoch": 2.7199999999999998, "percentage": 90.67, "elapsed_time": "0:32:06", "remaining_time": "0:03:18"}
257
+ {"current_steps": 2050, "total_steps": 2250, "loss": 0.0834, "lr": 4.822694433882635e-07, "epoch": 2.7333333333333334, "percentage": 91.11, "elapsed_time": "0:32:16", "remaining_time": "0:03:08"}
258
+ {"current_steps": 2060, "total_steps": 2250, "loss": 0.0776, "lr": 4.358191932554745e-07, "epoch": 2.7466666666666666, "percentage": 91.56, "elapsed_time": "0:32:25", "remaining_time": "0:02:59"}
259
+ {"current_steps": 2070, "total_steps": 2250, "loss": 0.0826, "lr": 3.9167085533639235e-07, "epoch": 2.76, "percentage": 92.0, "elapsed_time": "0:32:35", "remaining_time": "0:02:50"}
260
+ {"current_steps": 2080, "total_steps": 2250, "loss": 0.0741, "lr": 3.498350552768859e-07, "epoch": 2.7733333333333334, "percentage": 92.44, "elapsed_time": "0:32:44", "remaining_time": "0:02:40"}
261
+ {"current_steps": 2090, "total_steps": 2250, "loss": 0.0869, "lr": 3.103218621399595e-07, "epoch": 2.7866666666666666, "percentage": 92.89, "elapsed_time": "0:32:54", "remaining_time": "0:02:31"}
262
+ {"current_steps": 2100, "total_steps": 2250, "loss": 0.0823, "lr": 2.731407859823276e-07, "epoch": 2.8, "percentage": 93.33, "elapsed_time": "0:33:03", "remaining_time": "0:02:21"}
263
+ {"current_steps": 2110, "total_steps": 2250, "loss": 0.079, "lr": 2.3830077556552424e-07, "epoch": 2.8133333333333335, "percentage": 93.78, "elapsed_time": "0:33:13", "remaining_time": "0:02:12"}
264
+ {"current_steps": 2120, "total_steps": 2250, "loss": 0.0815, "lr": 2.0581021620211694e-07, "epoch": 2.8266666666666667, "percentage": 94.22, "elapsed_time": "0:33:22", "remaining_time": "0:02:02"}
265
+ {"current_steps": 2130, "total_steps": 2250, "loss": 0.0762, "lr": 1.756769277375192e-07, "epoch": 2.84, "percentage": 94.67, "elapsed_time": "0:33:32", "remaining_time": "0:01:53"}
266
+ {"current_steps": 2140, "total_steps": 2250, "loss": 0.0747, "lr": 1.4790816266791018e-07, "epoch": 2.8533333333333335, "percentage": 95.11, "elapsed_time": "0:33:41", "remaining_time": "0:01:43"}
267
+ {"current_steps": 2150, "total_steps": 2250, "loss": 0.0788, "lr": 1.2251060439469996e-07, "epoch": 2.8666666666666667, "percentage": 95.56, "elapsed_time": "0:33:51", "remaining_time": "0:01:34"}
268
+ {"current_steps": 2160, "total_steps": 2250, "loss": 0.0679, "lr": 9.94903656159607e-08, "epoch": 2.88, "percentage": 96.0, "elapsed_time": "0:34:00", "remaining_time": "0:01:25"}
269
+ {"current_steps": 2170, "total_steps": 2250, "loss": 0.0767, "lr": 7.885298685522235e-08, "epoch": 2.8933333333333335, "percentage": 96.44, "elapsed_time": "0:34:09", "remaining_time": "0:01:15"}
270
+ {"current_steps": 2180, "total_steps": 2250, "loss": 0.0759, "lr": 6.060343512797496e-08, "epoch": 2.9066666666666667, "percentage": 96.89, "elapsed_time": "0:34:19", "remaining_time": "0:01:06"}
271
+ {"current_steps": 2190, "total_steps": 2250, "loss": 0.0794, "lr": 4.474610274620816e-08, "epoch": 2.92, "percentage": 97.33, "elapsed_time": "0:34:27", "remaining_time": "0:00:56"}
272
+ {"current_steps": 2200, "total_steps": 2250, "loss": 0.0724, "lr": 3.1284806261264735e-08, "epoch": 2.9333333333333336, "percentage": 97.78, "elapsed_time": "0:34:37", "remaining_time": "0:00:47"}
273
+ {"current_steps": 2210, "total_steps": 2250, "loss": 0.0779, "lr": 2.0222785545265243e-08, "epoch": 2.9466666666666668, "percentage": 98.22, "elapsed_time": "0:34:46", "remaining_time": "0:00:37"}
274
+ {"current_steps": 2220, "total_steps": 2250, "loss": 0.0782, "lr": 1.1562703011347431e-08, "epoch": 2.96, "percentage": 98.67, "elapsed_time": "0:34:56", "remaining_time": "0:00:28"}
275
+ {"current_steps": 2230, "total_steps": 2250, "loss": 0.0741, "lr": 5.306642972862097e-09, "epoch": 2.9733333333333336, "percentage": 99.11, "elapsed_time": "0:35:04", "remaining_time": "0:00:18"}
276
+ {"current_steps": 2240, "total_steps": 2250, "loss": 0.0847, "lr": 1.4561111417255024e-09, "epoch": 2.986666666666667, "percentage": 99.56, "elapsed_time": "0:35:13", "remaining_time": "0:00:09"}
277
+ {"current_steps": 2250, "total_steps": 2250, "loss": 0.0796, "lr": 1.2034266022586948e-11, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "0:35:23", "remaining_time": "0:00:00"}
278
+ {"current_steps": 2250, "total_steps": 2250, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "0:35:57", "remaining_time": "0:00:00"}