Training in progress, step 200
Browse files- trainer_log.jsonl +82 -119
- training_args.bin +1 -1
trainer_log.jsonl
CHANGED
|
@@ -1,119 +1,82 @@
|
|
| 1 |
-
{"current_steps":
|
| 2 |
-
{"current_steps":
|
| 3 |
-
{"current_steps":
|
| 4 |
-
{"current_steps":
|
| 5 |
-
{"current_steps":
|
| 6 |
-
{"current_steps":
|
| 7 |
-
{"current_steps":
|
| 8 |
-
{"current_steps":
|
| 9 |
-
{"current_steps":
|
| 10 |
-
{"current_steps":
|
| 11 |
-
{"current_steps":
|
| 12 |
-
{"current_steps":
|
| 13 |
-
{"current_steps":
|
| 14 |
-
{"current_steps":
|
| 15 |
-
{"current_steps":
|
| 16 |
-
{"current_steps":
|
| 17 |
-
{"current_steps":
|
| 18 |
-
{"current_steps":
|
| 19 |
-
{"current_steps":
|
| 20 |
-
{"current_steps":
|
| 21 |
-
{"current_steps":
|
| 22 |
-
{"current_steps":
|
| 23 |
-
{"current_steps":
|
| 24 |
-
{"current_steps":
|
| 25 |
-
{"current_steps":
|
| 26 |
-
{"current_steps":
|
| 27 |
-
{"current_steps":
|
| 28 |
-
{"current_steps":
|
| 29 |
-
{"current_steps":
|
| 30 |
-
{"current_steps":
|
| 31 |
-
{"current_steps":
|
| 32 |
-
{"current_steps":
|
| 33 |
-
{"current_steps":
|
| 34 |
-
{"current_steps":
|
| 35 |
-
{"current_steps":
|
| 36 |
-
{"current_steps":
|
| 37 |
-
{"current_steps":
|
| 38 |
-
{"current_steps":
|
| 39 |
-
{"current_steps":
|
| 40 |
-
{"current_steps":
|
| 41 |
-
{"current_steps":
|
| 42 |
-
{"current_steps":
|
| 43 |
-
{"current_steps":
|
| 44 |
-
{"current_steps":
|
| 45 |
-
{"current_steps":
|
| 46 |
-
{"current_steps":
|
| 47 |
-
{"current_steps":
|
| 48 |
-
{"current_steps":
|
| 49 |
-
{"current_steps":
|
| 50 |
-
{"current_steps":
|
| 51 |
-
{"current_steps":
|
| 52 |
-
{"current_steps":
|
| 53 |
-
{"current_steps":
|
| 54 |
-
{"current_steps":
|
| 55 |
-
{"current_steps":
|
| 56 |
-
{"current_steps":
|
| 57 |
-
{"current_steps":
|
| 58 |
-
{"current_steps":
|
| 59 |
-
{"current_steps":
|
| 60 |
-
{"current_steps":
|
| 61 |
-
{"current_steps":
|
| 62 |
-
{"current_steps":
|
| 63 |
-
{"current_steps":
|
| 64 |
-
{"current_steps":
|
| 65 |
-
{"current_steps":
|
| 66 |
-
{"current_steps":
|
| 67 |
-
{"current_steps":
|
| 68 |
-
{"current_steps":
|
| 69 |
-
{"current_steps":
|
| 70 |
-
{"current_steps":
|
| 71 |
-
{"current_steps":
|
| 72 |
-
{"current_steps":
|
| 73 |
-
{"current_steps":
|
| 74 |
-
{"current_steps":
|
| 75 |
-
{"current_steps":
|
| 76 |
-
{"current_steps":
|
| 77 |
-
{"current_steps":
|
| 78 |
-
{"current_steps":
|
| 79 |
-
{"current_steps":
|
| 80 |
-
{"current_steps":
|
| 81 |
-
{"current_steps":
|
| 82 |
-
{"current_steps":
|
| 83 |
-
{"current_steps": 83, "total_steps": 118, "loss": 1.5591, "lr": 1.4744066440523391e-05, "epoch": 0.7000527148128625, "percentage": 70.34, "elapsed_time": "9:05:52", "remaining_time": "3:50:11"}
|
| 84 |
-
{"current_steps": 84, "total_steps": 118, "loss": 1.569, "lr": 1.3985298661056292e-05, "epoch": 0.7084870848708487, "percentage": 71.19, "elapsed_time": "9:12:23", "remaining_time": "3:43:35"}
|
| 85 |
-
{"current_steps": 85, "total_steps": 118, "loss": 1.5631, "lr": 1.324059702589043e-05, "epoch": 0.716921454928835, "percentage": 72.03, "elapsed_time": "9:19:02", "remaining_time": "3:37:02"}
|
| 86 |
-
{"current_steps": 86, "total_steps": 118, "loss": 1.5561, "lr": 1.2510615626565844e-05, "epoch": 0.7253558249868213, "percentage": 72.88, "elapsed_time": "9:25:38", "remaining_time": "3:30:28"}
|
| 87 |
-
{"current_steps": 87, "total_steps": 118, "loss": 1.5471, "lr": 1.1795995625442208e-05, "epoch": 0.7337901950448076, "percentage": 73.73, "elapsed_time": "9:32:16", "remaining_time": "3:23:54"}
|
| 88 |
-
{"current_steps": 88, "total_steps": 118, "loss": 1.5583, "lr": 1.109736469254867e-05, "epoch": 0.7422245651027939, "percentage": 74.58, "elapsed_time": "9:38:49", "remaining_time": "3:17:19"}
|
| 89 |
-
{"current_steps": 89, "total_steps": 118, "loss": 1.5531, "lr": 1.0415336454284356e-05, "epoch": 0.7506589351607802, "percentage": 75.42, "elapsed_time": "9:45:21", "remaining_time": "3:10:44"}
|
| 90 |
-
{"current_steps": 90, "total_steps": 118, "loss": 1.5559, "lr": 9.75050995445369e-06, "epoch": 0.7590933052187665, "percentage": 76.27, "elapsed_time": "9:51:52", "remaining_time": "3:04:08"}
|
| 91 |
-
{"current_steps": 91, "total_steps": 118, "loss": 1.5527, "lr": 9.103469128110098e-06, "epoch": 0.7675276752767528, "percentage": 77.12, "elapsed_time": "9:58:26", "remaining_time": "2:57:33"}
|
| 92 |
-
{"current_steps": 92, "total_steps": 118, "loss": 1.5514, "lr": 8.474782288670058e-06, "epoch": 0.7759620453347391, "percentage": 77.97, "elapsed_time": "10:04:58", "remaining_time": "2:50:58"}
|
| 93 |
-
{"current_steps": 93, "total_steps": 118, "loss": 1.5536, "lr": 7.86500162874818e-06, "epoch": 0.7843964153927253, "percentage": 78.81, "elapsed_time": "10:11:30", "remaining_time": "2:44:23"}
|
| 94 |
-
{"current_steps": 94, "total_steps": 118, "loss": 1.5541, "lr": 7.274662735151396e-06, "epoch": 0.7928307854507116, "percentage": 79.66, "elapsed_time": "10:18:03", "remaining_time": "2:37:48"}
|
| 95 |
-
{"current_steps": 95, "total_steps": 118, "loss": 1.5512, "lr": 6.704284118458731e-06, "epoch": 0.8012651555086979, "percentage": 80.51, "elapsed_time": "10:24:34", "remaining_time": "2:31:12"}
|
| 96 |
-
{"current_steps": 96, "total_steps": 118, "loss": 1.5492, "lr": 6.154366757599399e-06, "epoch": 0.8096995255666842, "percentage": 81.36, "elapsed_time": "10:31:03", "remaining_time": "2:24:37"}
|
| 97 |
-
{"current_steps": 97, "total_steps": 118, "loss": 1.5472, "lr": 5.625393659829561e-06, "epoch": 0.8181338956246705, "percentage": 82.2, "elapsed_time": "10:37:33", "remaining_time": "2:18:01"}
|
| 98 |
-
{"current_steps": 98, "total_steps": 118, "loss": 1.551, "lr": 5.117829436493947e-06, "epoch": 0.8265682656826568, "percentage": 83.05, "elapsed_time": "10:44:05", "remaining_time": "2:11:26"}
|
| 99 |
-
{"current_steps": 99, "total_steps": 118, "loss": 1.5599, "lr": 4.632119894945215e-06, "epoch": 0.8350026357406432, "percentage": 83.9, "elapsed_time": "10:50:40", "remaining_time": "2:04:52"}
|
| 100 |
-
{"current_steps": 100, "total_steps": 118, "loss": 1.5552, "lr": 4.1686916469793335e-06, "epoch": 0.8434370057986295, "percentage": 84.75, "elapsed_time": "10:57:11", "remaining_time": "1:58:17"}
|
| 101 |
-
{"current_steps": 101, "total_steps": 118, "loss": 1.5645, "lr": 3.7279517341308977e-06, "epoch": 0.8518713758566157, "percentage": 85.59, "elapsed_time": "11:10:39", "remaining_time": "1:52:52"}
|
| 102 |
-
{"current_steps": 102, "total_steps": 118, "loss": 1.5466, "lr": 3.3102872701575838e-06, "epoch": 0.860305745914602, "percentage": 86.44, "elapsed_time": "11:17:22", "remaining_time": "1:46:15"}
|
| 103 |
-
{"current_steps": 103, "total_steps": 118, "loss": 1.555, "lr": 2.916065101027694e-06, "epoch": 0.8687401159725883, "percentage": 87.29, "elapsed_time": "11:23:54", "remaining_time": "1:39:35"}
|
| 104 |
-
{"current_steps": 104, "total_steps": 118, "loss": 1.5479, "lr": 2.5456314827094463e-06, "epoch": 0.8771744860305746, "percentage": 88.14, "elapsed_time": "11:30:23", "remaining_time": "1:32:56"}
|
| 105 |
-
{"current_steps": 105, "total_steps": 118, "loss": 1.545, "lr": 2.1993117770449987e-06, "epoch": 0.8856088560885609, "percentage": 88.98, "elapsed_time": "11:36:56", "remaining_time": "1:26:17"}
|
| 106 |
-
{"current_steps": 106, "total_steps": 118, "loss": 1.554, "lr": 1.8774101659763731e-06, "epoch": 0.8940432261465472, "percentage": 89.83, "elapsed_time": "11:43:28", "remaining_time": "1:19:38"}
|
| 107 |
-
{"current_steps": 107, "total_steps": 118, "loss": 1.5467, "lr": 1.5802093843742582e-06, "epoch": 0.9024775962045335, "percentage": 90.68, "elapsed_time": "11:49:59", "remaining_time": "1:12:59"}
|
| 108 |
-
{"current_steps": 108, "total_steps": 118, "loss": 1.55, "lr": 1.3079704717043273e-06, "epoch": 0.9109119662625198, "percentage": 91.53, "elapsed_time": "11:56:29", "remaining_time": "1:06:20"}
|
| 109 |
-
{"current_steps": 109, "total_steps": 118, "loss": 1.5558, "lr": 1.060932542749241e-06, "epoch": 0.9193463363205061, "percentage": 92.37, "elapsed_time": "12:03:07", "remaining_time": "0:59:42"}
|
| 110 |
-
{"current_steps": 110, "total_steps": 118, "loss": 1.5563, "lr": 8.393125775876775e-07, "epoch": 0.9277807063784923, "percentage": 93.22, "elapsed_time": "12:09:40", "remaining_time": "0:53:04"}
|
| 111 |
-
{"current_steps": 111, "total_steps": 118, "loss": 1.5537, "lr": 6.433052310148791e-07, "epoch": 0.9362150764364786, "percentage": 94.07, "elapsed_time": "12:16:16", "remaining_time": "0:46:25"}
|
| 112 |
-
{"current_steps": 112, "total_steps": 118, "loss": 1.5567, "lr": 4.730826615720951e-07, "epoch": 0.9446494464944649, "percentage": 94.92, "elapsed_time": "12:22:49", "remaining_time": "0:39:47"}
|
| 113 |
-
{"current_steps": 113, "total_steps": 118, "loss": 1.5549, "lr": 3.28794380335079e-07, "epoch": 0.9530838165524512, "percentage": 95.76, "elapsed_time": "12:29:23", "remaining_time": "0:33:09"}
|
| 114 |
-
{"current_steps": 114, "total_steps": 118, "loss": 1.5585, "lr": 2.1056711959449247e-07, "epoch": 0.9615181866104375, "percentage": 96.61, "elapsed_time": "12:35:54", "remaining_time": "0:26:31"}
|
| 115 |
-
{"current_steps": 115, "total_steps": 118, "loss": 1.5536, "lr": 1.1850472154349313e-07, "epoch": 0.9699525566684238, "percentage": 97.46, "elapsed_time": "12:42:26", "remaining_time": "0:19:53"}
|
| 116 |
-
{"current_steps": 116, "total_steps": 118, "loss": 1.5705, "lr": 5.268804707035946e-08, "epoch": 0.9783869267264101, "percentage": 98.31, "elapsed_time": "12:48:58", "remaining_time": "0:13:15"}
|
| 117 |
-
{"current_steps": 117, "total_steps": 118, "loss": 1.5566, "lr": 1.3174904736169557e-08, "epoch": 0.9868212967843965, "percentage": 99.15, "elapsed_time": "12:55:36", "remaining_time": "0:06:37"}
|
| 118 |
-
{"current_steps": 118, "total_steps": 118, "loss": 1.5464, "lr": 0.0, "epoch": 0.9952556668423828, "percentage": 100.0, "elapsed_time": "13:02:09", "remaining_time": "0:00:00"}
|
| 119 |
-
{"current_steps": 118, "total_steps": 118, "epoch": 0.9952556668423828, "percentage": 100.0, "elapsed_time": "13:09:19", "remaining_time": "0:00:00"}
|
|
|
|
| 1 |
+
{"current_steps": 119, "total_steps": 355, "loss": 1.5345, "lr": 5.052348727216575e-05, "epoch": 0.33444581064465134, "percentage": 33.52, "elapsed_time": "0:06:51", "remaining_time": "0:13:35"}
|
| 2 |
+
{"current_steps": 120, "total_steps": 355, "loss": 1.528, "lr": 5.030700400656054e-05, "epoch": 0.3372562796416652, "percentage": 33.8, "elapsed_time": "0:13:28", "remaining_time": "0:26:23"}
|
| 3 |
+
{"current_steps": 121, "total_steps": 355, "loss": 1.5315, "lr": 5.0088551219466324e-05, "epoch": 0.3400667486386791, "percentage": 34.08, "elapsed_time": "0:20:02", "remaining_time": "0:38:44"}
|
| 4 |
+
{"current_steps": 122, "total_steps": 355, "loss": 1.5237, "lr": 4.9868150098029096e-05, "epoch": 0.34287721763569295, "percentage": 34.37, "elapsed_time": "0:26:34", "remaining_time": "0:50:44"}
|
| 5 |
+
{"current_steps": 123, "total_steps": 355, "loss": 1.5347, "lr": 4.964582201835856e-05, "epoch": 0.3456876866327068, "percentage": 34.65, "elapsed_time": "0:33:09", "remaining_time": "1:02:31"}
|
| 6 |
+
{"current_steps": 124, "total_steps": 355, "loss": 1.5237, "lr": 4.942158854345483e-05, "epoch": 0.3484981556297207, "percentage": 34.93, "elapsed_time": "0:39:40", "remaining_time": "1:13:54"}
|
| 7 |
+
{"current_steps": 125, "total_steps": 355, "loss": 1.5245, "lr": 4.919547142111723e-05, "epoch": 0.35130862462673457, "percentage": 35.21, "elapsed_time": "0:46:14", "remaining_time": "1:25:04"}
|
| 8 |
+
{"current_steps": 126, "total_steps": 355, "loss": 1.534, "lr": 4.896749258183492e-05, "epoch": 0.35411909362374844, "percentage": 35.49, "elapsed_time": "0:52:50", "remaining_time": "1:36:01"}
|
| 9 |
+
{"current_steps": 127, "total_steps": 355, "loss": 1.5198, "lr": 4.8737674136659965e-05, "epoch": 0.35692956262076236, "percentage": 35.77, "elapsed_time": "0:59:25", "remaining_time": "1:46:41"}
|
| 10 |
+
{"current_steps": 128, "total_steps": 355, "loss": 1.525, "lr": 4.850603837506285e-05, "epoch": 0.35974003161777623, "percentage": 36.06, "elapsed_time": "1:06:01", "remaining_time": "1:57:06"}
|
| 11 |
+
{"current_steps": 129, "total_steps": 355, "loss": 1.531, "lr": 4.8272607762770674e-05, "epoch": 0.3625505006147901, "percentage": 36.34, "elapsed_time": "1:12:36", "remaining_time": "2:07:13"}
|
| 12 |
+
{"current_steps": 130, "total_steps": 355, "loss": 1.5211, "lr": 4.803740493958826e-05, "epoch": 0.365360969611804, "percentage": 36.62, "elapsed_time": "1:19:10", "remaining_time": "2:17:01"}
|
| 13 |
+
{"current_steps": 131, "total_steps": 355, "loss": 1.5194, "lr": 4.780045271720234e-05, "epoch": 0.36817143860881785, "percentage": 36.9, "elapsed_time": "1:25:47", "remaining_time": "2:26:42"}
|
| 14 |
+
{"current_steps": 132, "total_steps": 355, "loss": 1.5334, "lr": 4.756177407696921e-05, "epoch": 0.3709819076058317, "percentage": 37.18, "elapsed_time": "1:32:23", "remaining_time": "2:36:04"}
|
| 15 |
+
{"current_steps": 133, "total_steps": 355, "loss": 1.5262, "lr": 4.7321392167685746e-05, "epoch": 0.3737923766028456, "percentage": 37.46, "elapsed_time": "1:38:57", "remaining_time": "2:45:11"}
|
| 16 |
+
{"current_steps": 134, "total_steps": 355, "loss": 1.5346, "lr": 4.707933030334428e-05, "epoch": 0.37660284559985946, "percentage": 37.75, "elapsed_time": "1:45:34", "remaining_time": "2:54:06"}
|
| 17 |
+
{"current_steps": 135, "total_steps": 355, "loss": 1.5282, "lr": 4.683561196087148e-05, "epoch": 0.37941331459687333, "percentage": 38.03, "elapsed_time": "1:52:10", "remaining_time": "3:02:48"}
|
| 18 |
+
{"current_steps": 136, "total_steps": 355, "loss": 1.525, "lr": 4.65902607778513e-05, "epoch": 0.38222378359388726, "percentage": 38.31, "elapsed_time": "1:58:47", "remaining_time": "3:11:16"}
|
| 19 |
+
{"current_steps": 137, "total_steps": 355, "loss": 1.5233, "lr": 4.634330055023255e-05, "epoch": 0.3850342525909011, "percentage": 38.59, "elapsed_time": "2:05:19", "remaining_time": "3:19:26"}
|
| 20 |
+
{"current_steps": 138, "total_steps": 355, "loss": 1.5313, "lr": 4.609475523002089e-05, "epoch": 0.387844721587915, "percentage": 38.87, "elapsed_time": "2:11:55", "remaining_time": "3:27:26"}
|
| 21 |
+
{"current_steps": 139, "total_steps": 355, "loss": 1.5168, "lr": 4.584464892295588e-05, "epoch": 0.39065519058492887, "percentage": 39.15, "elapsed_time": "2:18:29", "remaining_time": "3:35:11"}
|
| 22 |
+
{"current_steps": 140, "total_steps": 355, "loss": 1.5342, "lr": 4.5593005886172906e-05, "epoch": 0.39346565958194274, "percentage": 39.44, "elapsed_time": "2:25:04", "remaining_time": "3:42:47"}
|
| 23 |
+
{"current_steps": 141, "total_steps": 355, "loss": 1.5266, "lr": 4.533985052585068e-05, "epoch": 0.3962761285789566, "percentage": 39.72, "elapsed_time": "2:31:42", "remaining_time": "3:50:14"}
|
| 24 |
+
{"current_steps": 142, "total_steps": 355, "loss": 1.5364, "lr": 4.508520739484407e-05, "epoch": 0.3990865975759705, "percentage": 40.0, "elapsed_time": "2:38:13", "remaining_time": "3:57:20"}
|
| 25 |
+
{"current_steps": 143, "total_steps": 355, "loss": 1.5219, "lr": 4.48291011903028e-05, "epoch": 0.40189706657298435, "percentage": 40.28, "elapsed_time": "2:44:49", "remaining_time": "4:04:21"}
|
| 26 |
+
{"current_steps": 144, "total_steps": 355, "loss": 1.524, "lr": 4.457155675127616e-05, "epoch": 0.4047075355699982, "percentage": 40.56, "elapsed_time": "2:51:25", "remaining_time": "4:11:10"}
|
| 27 |
+
{"current_steps": 145, "total_steps": 355, "loss": 1.5275, "lr": 4.431259905630392e-05, "epoch": 0.4075180045670121, "percentage": 40.85, "elapsed_time": "2:58:00", "remaining_time": "4:17:48"}
|
| 28 |
+
{"current_steps": 146, "total_steps": 355, "loss": 1.527, "lr": 4.4052253220993704e-05, "epoch": 0.410328473564026, "percentage": 41.13, "elapsed_time": "3:04:37", "remaining_time": "4:24:17"}
|
| 29 |
+
{"current_steps": 147, "total_steps": 355, "loss": 1.5331, "lr": 4.3790544495585154e-05, "epoch": 0.4131389425610399, "percentage": 41.41, "elapsed_time": "3:11:15", "remaining_time": "4:30:37"}
|
| 30 |
+
{"current_steps": 148, "total_steps": 355, "loss": 1.5374, "lr": 4.352749826250091e-05, "epoch": 0.41594941155805376, "percentage": 41.69, "elapsed_time": "3:17:55", "remaining_time": "4:36:49"}
|
| 31 |
+
{"current_steps": 149, "total_steps": 355, "loss": 1.527, "lr": 4.326314003388488e-05, "epoch": 0.41875988055506763, "percentage": 41.97, "elapsed_time": "3:24:32", "remaining_time": "4:42:46"}
|
| 32 |
+
{"current_steps": 150, "total_steps": 355, "loss": 1.5258, "lr": 4.299749544912791e-05, "epoch": 0.4215703495520815, "percentage": 42.25, "elapsed_time": "3:31:11", "remaining_time": "4:48:37"}
|
| 33 |
+
{"current_steps": 151, "total_steps": 355, "loss": 1.5192, "lr": 4.2730590272381e-05, "epoch": 0.4243808185490954, "percentage": 42.54, "elapsed_time": "3:37:50", "remaining_time": "4:54:17"}
|
| 34 |
+
{"current_steps": 152, "total_steps": 355, "loss": 1.529, "lr": 4.24624503900566e-05, "epoch": 0.42719128754610924, "percentage": 42.82, "elapsed_time": "3:44:29", "remaining_time": "4:59:49"}
|
| 35 |
+
{"current_steps": 153, "total_steps": 355, "loss": 1.5253, "lr": 4.21931018083179e-05, "epoch": 0.4300017565431231, "percentage": 43.1, "elapsed_time": "3:51:02", "remaining_time": "5:05:02"}
|
| 36 |
+
{"current_steps": 154, "total_steps": 355, "loss": 1.5241, "lr": 4.192257065055661e-05, "epoch": 0.432812225540137, "percentage": 43.38, "elapsed_time": "3:57:37", "remaining_time": "5:10:08"}
|
| 37 |
+
{"current_steps": 155, "total_steps": 355, "loss": 1.5284, "lr": 4.165088315485926e-05, "epoch": 0.4356226945371509, "percentage": 43.66, "elapsed_time": "4:04:13", "remaining_time": "5:15:07"}
|
| 38 |
+
{"current_steps": 156, "total_steps": 355, "loss": 1.5377, "lr": 4.137806567146249e-05, "epoch": 0.4384331635341648, "percentage": 43.94, "elapsed_time": "4:10:49", "remaining_time": "5:19:58"}
|
| 39 |
+
{"current_steps": 157, "total_steps": 355, "loss": 1.5325, "lr": 4.110414466019744e-05, "epoch": 0.44124363253117865, "percentage": 44.23, "elapsed_time": "4:17:26", "remaining_time": "5:24:39"}
|
| 40 |
+
{"current_steps": 158, "total_steps": 355, "loss": 1.5189, "lr": 4.082914668792334e-05, "epoch": 0.4440541015281925, "percentage": 44.51, "elapsed_time": "4:24:03", "remaining_time": "5:29:13"}
|
| 41 |
+
{"current_steps": 159, "total_steps": 355, "loss": 1.5259, "lr": 4.055309842595105e-05, "epoch": 0.4468645705252064, "percentage": 44.79, "elapsed_time": "4:30:41", "remaining_time": "5:33:41"}
|
| 42 |
+
{"current_steps": 160, "total_steps": 355, "loss": 1.5287, "lr": 4.027602664745615e-05, "epoch": 0.44967503952222027, "percentage": 45.07, "elapsed_time": "4:37:14", "remaining_time": "5:37:53"}
|
| 43 |
+
{"current_steps": 161, "total_steps": 355, "loss": 1.53, "lr": 3.999795822488233e-05, "epoch": 0.45248550851923414, "percentage": 45.35, "elapsed_time": "4:43:47", "remaining_time": "5:41:57"}
|
| 44 |
+
{"current_steps": 162, "total_steps": 355, "loss": 1.5387, "lr": 3.9718920127335105e-05, "epoch": 0.455295977516248, "percentage": 45.63, "elapsed_time": "4:50:23", "remaining_time": "5:45:57"}
|
| 45 |
+
{"current_steps": 163, "total_steps": 355, "loss": 1.5281, "lr": 3.943893941796615e-05, "epoch": 0.4581064465132619, "percentage": 45.92, "elapsed_time": "4:56:59", "remaining_time": "5:49:50"}
|
| 46 |
+
{"current_steps": 164, "total_steps": 355, "loss": 1.5341, "lr": 3.915804325134851e-05, "epoch": 0.4609169155102758, "percentage": 46.2, "elapsed_time": "5:03:38", "remaining_time": "5:53:37"}
|
| 47 |
+
{"current_steps": 165, "total_steps": 355, "loss": 1.5133, "lr": 3.887625887084293e-05, "epoch": 0.4637273845072897, "percentage": 46.48, "elapsed_time": "5:10:12", "remaining_time": "5:57:12"}
|
| 48 |
+
{"current_steps": 166, "total_steps": 355, "loss": 1.5226, "lr": 3.859361360595568e-05, "epoch": 0.46653785350430355, "percentage": 46.76, "elapsed_time": "5:16:49", "remaining_time": "6:00:43"}
|
| 49 |
+
{"current_steps": 167, "total_steps": 355, "loss": 1.5262, "lr": 3.831013486968784e-05, "epoch": 0.4693483225013174, "percentage": 47.04, "elapsed_time": "5:23:21", "remaining_time": "6:04:00"}
|
| 50 |
+
{"current_steps": 168, "total_steps": 355, "loss": 1.5289, "lr": 3.8025850155876635e-05, "epoch": 0.4721587914983313, "percentage": 47.32, "elapsed_time": "5:29:58", "remaining_time": "6:07:17"}
|
| 51 |
+
{"current_steps": 169, "total_steps": 355, "loss": 1.5147, "lr": 3.7740787036528866e-05, "epoch": 0.47496926049534516, "percentage": 47.61, "elapsed_time": "5:36:34", "remaining_time": "6:10:26"}
|
| 52 |
+
{"current_steps": 170, "total_steps": 355, "loss": 1.5285, "lr": 3.745497315914682e-05, "epoch": 0.47777972949235903, "percentage": 47.89, "elapsed_time": "5:43:11", "remaining_time": "6:13:28"}
|
| 53 |
+
{"current_steps": 171, "total_steps": 355, "loss": 1.5255, "lr": 3.716843624404675e-05, "epoch": 0.4805901984893729, "percentage": 48.17, "elapsed_time": "5:49:43", "remaining_time": "6:16:18"}
|
| 54 |
+
{"current_steps": 172, "total_steps": 355, "loss": 1.5225, "lr": 3.6881204081670394e-05, "epoch": 0.48340066748638677, "percentage": 48.45, "elapsed_time": "5:56:15", "remaining_time": "6:19:02"}
|
| 55 |
+
{"current_steps": 173, "total_steps": 355, "loss": 1.5249, "lr": 3.659330452988963e-05, "epoch": 0.48621113648340064, "percentage": 48.73, "elapsed_time": "6:02:49", "remaining_time": "6:21:42"}
|
| 56 |
+
{"current_steps": 174, "total_steps": 355, "loss": 1.5278, "lr": 3.630476551130467e-05, "epoch": 0.48902160548041457, "percentage": 49.01, "elapsed_time": "6:09:22", "remaining_time": "6:24:14"}
|
| 57 |
+
{"current_steps": 175, "total_steps": 355, "loss": 1.5206, "lr": 3.6015615010535845e-05, "epoch": 0.49183207447742844, "percentage": 49.3, "elapsed_time": "6:15:55", "remaining_time": "6:26:40"}
|
| 58 |
+
{"current_steps": 176, "total_steps": 355, "loss": 1.5132, "lr": 3.572588107150951e-05, "epoch": 0.4946425434744423, "percentage": 49.58, "elapsed_time": "6:22:31", "remaining_time": "6:29:02"}
|
| 59 |
+
{"current_steps": 177, "total_steps": 355, "loss": 1.531, "lr": 3.543559179473813e-05, "epoch": 0.4974530124714562, "percentage": 49.86, "elapsed_time": "6:29:09", "remaining_time": "6:31:21"}
|
| 60 |
+
{"current_steps": 178, "total_steps": 355, "loss": 1.5389, "lr": 3.5144775334594854e-05, "epoch": 0.5002634814684701, "percentage": 50.14, "elapsed_time": "6:35:46", "remaining_time": "6:33:32"}
|
| 61 |
+
{"current_steps": 179, "total_steps": 355, "loss": 1.5272, "lr": 3.485345989658295e-05, "epoch": 0.503073950465484, "percentage": 50.42, "elapsed_time": "6:42:20", "remaining_time": "6:35:35"}
|
| 62 |
+
{"current_steps": 180, "total_steps": 355, "loss": 1.5317, "lr": 3.456167373460017e-05, "epoch": 0.5058844194624978, "percentage": 50.7, "elapsed_time": "6:48:55", "remaining_time": "6:37:34"}
|
| 63 |
+
{"current_steps": 181, "total_steps": 355, "loss": 1.5255, "lr": 3.426944514819856e-05, "epoch": 0.5086948884595117, "percentage": 50.99, "elapsed_time": "6:55:28", "remaining_time": "6:39:24"}
|
| 64 |
+
{"current_steps": 182, "total_steps": 355, "loss": 1.5316, "lr": 3.397680247983966e-05, "epoch": 0.5115053574565256, "percentage": 51.27, "elapsed_time": "7:02:06", "remaining_time": "6:41:14"}
|
| 65 |
+
{"current_steps": 183, "total_steps": 355, "loss": 1.5256, "lr": 3.368377411214575e-05, "epoch": 0.5143158264535395, "percentage": 51.55, "elapsed_time": "7:08:42", "remaining_time": "6:42:56"}
|
| 66 |
+
{"current_steps": 184, "total_steps": 355, "loss": 1.5314, "lr": 3.3390388465147023e-05, "epoch": 0.5171262954505533, "percentage": 51.83, "elapsed_time": "7:15:15", "remaining_time": "6:44:30"}
|
| 67 |
+
{"current_steps": 185, "total_steps": 355, "loss": 1.5285, "lr": 3.309667399352522e-05, "epoch": 0.5199367644475672, "percentage": 52.11, "elapsed_time": "7:21:49", "remaining_time": "6:45:59"}
|
| 68 |
+
{"current_steps": 186, "total_steps": 355, "loss": 1.5319, "lr": 3.28026591838539e-05, "epoch": 0.5227472334445811, "percentage": 52.39, "elapsed_time": "7:28:21", "remaining_time": "6:47:23"}
|
| 69 |
+
{"current_steps": 187, "total_steps": 355, "loss": 1.5309, "lr": 3.250837255183563e-05, "epoch": 0.5255577024415949, "percentage": 52.68, "elapsed_time": "7:34:57", "remaining_time": "6:48:43"}
|
| 70 |
+
{"current_steps": 188, "total_steps": 355, "loss": 1.5219, "lr": 3.2213842639536236e-05, "epoch": 0.5283681714386088, "percentage": 52.96, "elapsed_time": "7:41:30", "remaining_time": "6:49:57"}
|
| 71 |
+
{"current_steps": 189, "total_steps": 355, "loss": 1.5295, "lr": 3.191909801261669e-05, "epoch": 0.5311786404356227, "percentage": 53.24, "elapsed_time": "7:48:08", "remaining_time": "6:51:10"}
|
| 72 |
+
{"current_steps": 190, "total_steps": 355, "loss": 1.5217, "lr": 3.162416725756253e-05, "epoch": 0.5339891094326366, "percentage": 53.52, "elapsed_time": "7:54:42", "remaining_time": "6:52:14"}
|
| 73 |
+
{"current_steps": 191, "total_steps": 355, "loss": 1.5218, "lr": 3.132907897891137e-05, "epoch": 0.5367995784296504, "percentage": 53.8, "elapsed_time": "8:01:15", "remaining_time": "6:53:13"}
|
| 74 |
+
{"current_steps": 192, "total_steps": 355, "loss": 1.5239, "lr": 3.103386179647859e-05, "epoch": 0.5396100474266643, "percentage": 54.08, "elapsed_time": "8:07:49", "remaining_time": "6:54:08"}
|
| 75 |
+
{"current_steps": 193, "total_steps": 355, "loss": 1.5259, "lr": 3.073854434258163e-05, "epoch": 0.5424205164236782, "percentage": 54.37, "elapsed_time": "8:14:24", "remaining_time": "6:54:59"}
|
| 76 |
+
{"current_steps": 194, "total_steps": 355, "loss": 1.5298, "lr": 3.0443155259263006e-05, "epoch": 0.545230985420692, "percentage": 54.65, "elapsed_time": "8:20:57", "remaining_time": "6:55:44"}
|
| 77 |
+
{"current_steps": 195, "total_steps": 355, "loss": 1.5147, "lr": 3.0147723195512293e-05, "epoch": 0.5480414544177059, "percentage": 54.93, "elapsed_time": "8:27:34", "remaining_time": "6:56:28"}
|
| 78 |
+
{"current_steps": 196, "total_steps": 355, "loss": 1.5398, "lr": 2.9852276804487722e-05, "epoch": 0.5508519234147198, "percentage": 55.21, "elapsed_time": "8:34:09", "remaining_time": "6:57:05"}
|
| 79 |
+
{"current_steps": 197, "total_steps": 355, "loss": 1.5235, "lr": 2.955684474073701e-05, "epoch": 0.5536623924117338, "percentage": 55.49, "elapsed_time": "8:40:44", "remaining_time": "6:57:39"}
|
| 80 |
+
{"current_steps": 198, "total_steps": 355, "loss": 1.5266, "lr": 2.926145565741837e-05, "epoch": 0.5564728614087476, "percentage": 55.77, "elapsed_time": "8:47:23", "remaining_time": "6:58:11"}
|
| 81 |
+
{"current_steps": 199, "total_steps": 355, "loss": 1.5165, "lr": 2.896613820352141e-05, "epoch": 0.5592833304057615, "percentage": 56.06, "elapsed_time": "8:53:56", "remaining_time": "6:58:34"}
|
| 82 |
+
{"current_steps": 200, "total_steps": 355, "loss": 1.5259, "lr": 2.867092102108864e-05, "epoch": 0.5620937994027754, "percentage": 56.34, "elapsed_time": "9:00:31", "remaining_time": "6:58:54"}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7672
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d3e3dd1535863e2ec2e072bfe27af39531aebea5677850ae85725c8ed0e328b5
|
| 3 |
size 7672
|