{"current_steps": 1, "total_steps": 555, "loss": 0.0585, "lr": 1.7857142857142857e-06, "epoch": 0.008978675645342313, "percentage": 0.18, "elapsed_time": "0:00:24", "remaining_time": "3:47:16"}
{"current_steps": 2, "total_steps": 555, "loss": 0.0573, "lr": 3.5714285714285714e-06, "epoch": 0.017957351290684626, "percentage": 0.36, "elapsed_time": "0:00:38", "remaining_time": "2:58:56"}
{"current_steps": 3, "total_steps": 555, "loss": 0.0583, "lr": 5.357142857142857e-06, "epoch": 0.026936026936026935, "percentage": 0.54, "elapsed_time": "0:00:52", "remaining_time": "2:39:31"}
{"current_steps": 4, "total_steps": 555, "loss": 0.0575, "lr": 7.142857142857143e-06, "epoch": 0.03591470258136925, "percentage": 0.72, "elapsed_time": "0:01:09", "remaining_time": "2:38:56"}
{"current_steps": 5, "total_steps": 555, "loss": 0.0612, "lr": 8.92857142857143e-06, "epoch": 0.04489337822671156, "percentage": 0.9, "elapsed_time": "0:01:20", "remaining_time": "2:27:43"}
{"current_steps": 5, "total_steps": 555, "eval_loss": 0.05599946156144142, "epoch": 0.04489337822671156, "percentage": 0.9, "elapsed_time": "0:01:28", "remaining_time": "2:42:39"}
{"current_steps": 6, "total_steps": 555, "loss": 0.05, "lr": 1.0714285714285714e-05, "epoch": 0.05387205387205387, "percentage": 1.08, "elapsed_time": "0:01:40", "remaining_time": "2:33:45"}
{"current_steps": 7, "total_steps": 555, "loss": 0.0515, "lr": 1.25e-05, "epoch": 0.06285072951739619, "percentage": 1.26, "elapsed_time": "0:01:52", "remaining_time": "2:26:45"}
{"current_steps": 8, "total_steps": 555, "loss": 0.0436, "lr": 1.4285714285714285e-05, "epoch": 0.0718294051627385, "percentage": 1.44, "elapsed_time": "0:02:04", "remaining_time": "2:21:52"}
{"current_steps": 9, "total_steps": 555, "loss": 0.044, "lr": 1.6071428571428572e-05, "epoch": 0.08080808080808081, "percentage": 1.62, "elapsed_time": "0:02:16", "remaining_time": "2:18:25"}
{"current_steps": 10, "total_steps": 555, "loss": 0.0411, "lr": 1.785714285714286e-05, "epoch": 0.08978675645342311, "percentage": 1.8, "elapsed_time": "0:02:28", "remaining_time": "2:14:56"}
{"current_steps": 10, "total_steps": 555, "eval_loss": 0.035706527531147, "epoch": 0.08978675645342311, "percentage": 1.8, "elapsed_time": "0:02:34", "remaining_time": "2:20:37"}
{"current_steps": 11, "total_steps": 555, "loss": 0.036, "lr": 1.9642857142857145e-05, "epoch": 0.09876543209876543, "percentage": 1.98, "elapsed_time": "0:02:45", "remaining_time": "2:16:33"}
{"current_steps": 12, "total_steps": 555, "loss": 0.0345, "lr": 2.1428571428571428e-05, "epoch": 0.10774410774410774, "percentage": 2.16, "elapsed_time": "0:03:03", "remaining_time": "2:18:10"}
{"current_steps": 13, "total_steps": 555, "loss": 0.0351, "lr": 2.3214285714285715e-05, "epoch": 0.11672278338945005, "percentage": 2.34, "elapsed_time": "0:03:13", "remaining_time": "2:14:26"}
{"current_steps": 14, "total_steps": 555, "loss": 0.0338, "lr": 2.5e-05, "epoch": 0.12570145903479238, "percentage": 2.52, "elapsed_time": "0:03:24", "remaining_time": "2:11:41"}
{"current_steps": 15, "total_steps": 555, "loss": 0.0353, "lr": 2.6785714285714288e-05, "epoch": 0.13468013468013468, "percentage": 2.7, "elapsed_time": "0:03:36", "remaining_time": "2:10:10"}
{"current_steps": 15, "total_steps": 555, "eval_loss": 0.03013807348906994, "epoch": 0.13468013468013468, "percentage": 2.7, "elapsed_time": "0:03:43", "remaining_time": "2:13:55"}
{"current_steps": 16, "total_steps": 555, "loss": 0.0304, "lr": 2.857142857142857e-05, "epoch": 0.143658810325477, "percentage": 2.88, "elapsed_time": "0:03:54", "remaining_time": "2:11:50"}
{"current_steps": 17, "total_steps": 555, "loss": 0.0303, "lr": 3.0357142857142857e-05, "epoch": 0.1526374859708193, "percentage": 3.06, "elapsed_time": "0:04:07", "remaining_time": "2:10:20"}
{"current_steps": 18, "total_steps": 555, "loss": 0.0327, "lr": 3.2142857142857144e-05, "epoch": 0.16161616161616163, "percentage": 3.24, "elapsed_time": "0:04:19", "remaining_time": "2:09:16"}
{"current_steps": 19, "total_steps": 555, "loss": 0.0271, "lr": 3.392857142857143e-05, "epoch": 0.17059483726150393, "percentage": 3.42, "elapsed_time": "0:04:32", "remaining_time": "2:07:59"}
{"current_steps": 20, "total_steps": 555, "loss": 0.0286, "lr": 3.571428571428572e-05, "epoch": 0.17957351290684623, "percentage": 3.6, "elapsed_time": "0:04:43", "remaining_time": "2:06:34"}
{"current_steps": 20, "total_steps": 555, "eval_loss": 0.026352621614933014, "epoch": 0.17957351290684623, "percentage": 3.6, "elapsed_time": "0:04:50", "remaining_time": "2:09:24"}
{"current_steps": 21, "total_steps": 555, "loss": 0.0293, "lr": 3.7500000000000003e-05, "epoch": 0.18855218855218855, "percentage": 3.78, "elapsed_time": "0:05:01", "remaining_time": "2:07:42"}
{"current_steps": 22, "total_steps": 555, "loss": 0.027, "lr": 3.928571428571429e-05, "epoch": 0.19753086419753085, "percentage": 3.96, "elapsed_time": "0:05:13", "remaining_time": "2:06:30"}
{"current_steps": 23, "total_steps": 555, "loss": 0.027, "lr": 4.107142857142857e-05, "epoch": 0.20650953984287318, "percentage": 4.14, "elapsed_time": "0:05:24", "remaining_time": "2:05:10"}
{"current_steps": 24, "total_steps": 555, "loss": 0.0249, "lr": 4.2857142857142856e-05, "epoch": 0.21548821548821548, "percentage": 4.32, "elapsed_time": "0:05:36", "remaining_time": "2:04:10"}
{"current_steps": 25, "total_steps": 555, "loss": 0.0282, "lr": 4.464285714285715e-05, "epoch": 0.2244668911335578, "percentage": 4.5, "elapsed_time": "0:05:49", "remaining_time": "2:03:21"}
{"current_steps": 25, "total_steps": 555, "eval_loss": 0.023908844217658043, "epoch": 0.2244668911335578, "percentage": 4.5, "elapsed_time": "0:05:55", "remaining_time": "2:05:34"}
{"current_steps": 26, "total_steps": 555, "loss": 0.0249, "lr": 4.642857142857143e-05, "epoch": 0.2334455667789001, "percentage": 4.68, "elapsed_time": "0:06:07", "remaining_time": "2:04:45"}
{"current_steps": 27, "total_steps": 555, "loss": 0.0254, "lr": 4.8214285714285716e-05, "epoch": 0.24242424242424243, "percentage": 4.86, "elapsed_time": "0:06:18", "remaining_time": "2:03:23"}
{"current_steps": 28, "total_steps": 555, "loss": 0.024, "lr": 5e-05, "epoch": 0.25140291806958476, "percentage": 5.05, "elapsed_time": "0:06:29", "remaining_time": "2:02:18"}
{"current_steps": 29, "total_steps": 555, "loss": 0.0251, "lr": 5.1785714285714296e-05, "epoch": 0.26038159371492703, "percentage": 5.23, "elapsed_time": "0:06:40", "remaining_time": "2:01:13"}
{"current_steps": 30, "total_steps": 555, "loss": 0.0223, "lr": 5.3571428571428575e-05, "epoch": 0.26936026936026936, "percentage": 5.41, "elapsed_time": "0:06:53", "remaining_time": "2:00:42"}
{"current_steps": 30, "total_steps": 555, "eval_loss": 0.02239508368074894, "epoch": 0.26936026936026936, "percentage": 5.41, "elapsed_time": "0:07:00", "remaining_time": "2:02:31"}
{"current_steps": 31, "total_steps": 555, "loss": 0.0281, "lr": 5.535714285714286e-05, "epoch": 0.2783389450056117, "percentage": 5.59, "elapsed_time": "0:07:11", "remaining_time": "2:01:38"}
{"current_steps": 32, "total_steps": 555, "loss": 0.0243, "lr": 5.714285714285714e-05, "epoch": 0.287317620650954, "percentage": 5.77, "elapsed_time": "0:07:22", "remaining_time": "2:00:32"}
{"current_steps": 33, "total_steps": 555, "loss": 0.0224, "lr": 5.8928571428571435e-05, "epoch": 0.2962962962962963, "percentage": 5.95, "elapsed_time": "0:07:34", "remaining_time": "1:59:46"}
{"current_steps": 34, "total_steps": 555, "loss": 0.0246, "lr": 6.0714285714285715e-05, "epoch": 0.3052749719416386, "percentage": 6.13, "elapsed_time": "0:07:46", "remaining_time": "1:59:02"}
{"current_steps": 35, "total_steps": 555, "loss": 0.0242, "lr": 6.25e-05, "epoch": 0.31425364758698093, "percentage": 6.31, "elapsed_time": "0:07:57", "remaining_time": "1:58:13"}
{"current_steps": 35, "total_steps": 555, "eval_loss": 0.020946728065609932, "epoch": 0.31425364758698093, "percentage": 6.31, "elapsed_time": "0:08:03", "remaining_time": "1:59:46"}
{"current_steps": 36, "total_steps": 555, "loss": 0.0257, "lr": 6.428571428571429e-05, "epoch": 0.32323232323232326, "percentage": 6.49, "elapsed_time": "0:08:16", "remaining_time": "1:59:16"}
{"current_steps": 37, "total_steps": 555, "loss": 0.0221, "lr": 6.607142857142857e-05, "epoch": 0.33221099887766553, "percentage": 6.67, "elapsed_time": "0:08:27", "remaining_time": "1:58:30"}
{"current_steps": 38, "total_steps": 555, "loss": 0.0201, "lr": 6.785714285714286e-05, "epoch": 0.34118967452300786, "percentage": 6.85, "elapsed_time": "0:08:41", "remaining_time": "1:58:08"}
{"current_steps": 39, "total_steps": 555, "loss": 0.0207, "lr": 6.964285714285715e-05, "epoch": 0.3501683501683502, "percentage": 7.03, "elapsed_time": "0:08:53", "remaining_time": "1:57:32"}
{"current_steps": 40, "total_steps": 555, "loss": 0.0211, "lr": 7.142857142857143e-05, "epoch": 0.35914702581369246, "percentage": 7.21, "elapsed_time": "0:09:05", "remaining_time": "1:57:02"}
{"current_steps": 40, "total_steps": 555, "eval_loss": 0.020273756235837936, "epoch": 0.35914702581369246, "percentage": 7.21, "elapsed_time": "0:09:11", "remaining_time": "1:58:24"}
{"current_steps": 41, "total_steps": 555, "loss": 0.0226, "lr": 7.321428571428571e-05, "epoch": 0.3681257014590348, "percentage": 7.39, "elapsed_time": "0:09:26", "remaining_time": "1:58:26"}
{"current_steps": 42, "total_steps": 555, "loss": 0.0187, "lr": 7.500000000000001e-05, "epoch": 0.3771043771043771, "percentage": 7.57, "elapsed_time": "0:09:38", "remaining_time": "1:57:47"}
{"current_steps": 43, "total_steps": 555, "loss": 0.0213, "lr": 7.67857142857143e-05, "epoch": 0.38608305274971944, "percentage": 7.75, "elapsed_time": "0:09:49", "remaining_time": "1:56:59"}
{"current_steps": 44, "total_steps": 555, "loss": 0.0214, "lr": 7.857142857142858e-05, "epoch": 0.3950617283950617, "percentage": 7.93, "elapsed_time": "0:10:03", "remaining_time": "1:56:48"}
{"current_steps": 45, "total_steps": 555, "loss": 0.0178, "lr": 8.035714285714287e-05, "epoch": 0.40404040404040403, "percentage": 8.11, "elapsed_time": "0:10:15", "remaining_time": "1:56:10"}
{"current_steps": 45, "total_steps": 555, "eval_loss": 0.0200771763920784, "epoch": 0.40404040404040403, "percentage": 8.11, "elapsed_time": "0:10:21", "remaining_time": "1:57:21"}
{"current_steps": 46, "total_steps": 555, "loss": 0.0186, "lr": 8.214285714285714e-05, "epoch": 0.41301907968574636, "percentage": 8.29, "elapsed_time": "0:10:31", "remaining_time": "1:56:28"}
{"current_steps": 47, "total_steps": 555, "loss": 0.0205, "lr": 8.392857142857144e-05, "epoch": 0.4219977553310887, "percentage": 8.47, "elapsed_time": "0:10:43", "remaining_time": "1:55:53"}
{"current_steps": 48, "total_steps": 555, "loss": 0.0207, "lr": 8.571428571428571e-05, "epoch": 0.43097643097643096, "percentage": 8.65, "elapsed_time": "0:10:56", "remaining_time": "1:55:36"}
{"current_steps": 49, "total_steps": 555, "loss": 0.0199, "lr": 8.75e-05, "epoch": 0.4399551066217733, "percentage": 8.83, "elapsed_time": "0:11:08", "remaining_time": "1:54:59"}
{"current_steps": 50, "total_steps": 555, "loss": 0.0206, "lr": 8.92857142857143e-05, "epoch": 0.4489337822671156, "percentage": 9.01, "elapsed_time": "0:11:20", "remaining_time": "1:54:32"}
{"current_steps": 50, "total_steps": 555, "eval_loss": 0.019621608778834343, "epoch": 0.4489337822671156, "percentage": 9.01, "elapsed_time": "0:11:26", "remaining_time": "1:55:35"}
{"current_steps": 51, "total_steps": 555, "loss": 0.0203, "lr": 9.107142857142857e-05, "epoch": 0.45791245791245794, "percentage": 9.19, "elapsed_time": "0:11:43", "remaining_time": "1:55:53"}
{"current_steps": 52, "total_steps": 555, "loss": 0.0209, "lr": 9.285714285714286e-05, "epoch": 0.4668911335578002, "percentage": 9.37, "elapsed_time": "0:11:55", "remaining_time": "1:55:18"}
{"current_steps": 53, "total_steps": 555, "loss": 0.0197, "lr": 9.464285714285715e-05, "epoch": 0.47586980920314254, "percentage": 9.55, "elapsed_time": "0:12:08", "remaining_time": "1:54:57"}
{"current_steps": 54, "total_steps": 555, "loss": 0.018, "lr": 9.642857142857143e-05, "epoch": 0.48484848484848486, "percentage": 9.73, "elapsed_time": "0:12:19", "remaining_time": "1:54:23"}
{"current_steps": 55, "total_steps": 555, "loss": 0.0196, "lr": 9.821428571428572e-05, "epoch": 0.49382716049382713, "percentage": 9.91, "elapsed_time": "0:12:32", "remaining_time": "1:53:59"}
{"current_steps": 55, "total_steps": 555, "eval_loss": 0.01931421272456646, "epoch": 0.49382716049382713, "percentage": 9.91, "elapsed_time": "0:12:38", "remaining_time": "1:54:56"}
{"current_steps": 56, "total_steps": 555, "loss": 0.0195, "lr": 0.0001, "epoch": 0.5028058361391695, "percentage": 10.09, "elapsed_time": "0:12:49", "remaining_time": "1:54:18"}
{"current_steps": 57, "total_steps": 555, "loss": 0.0195, "lr": 9.999900908311602e-05, "epoch": 0.5117845117845118, "percentage": 10.27, "elapsed_time": "0:13:01", "remaining_time": "1:53:46"}
{"current_steps": 58, "total_steps": 555, "loss": 0.0195, "lr": 9.999603637174071e-05, "epoch": 0.5207631874298541, "percentage": 10.45, "elapsed_time": "0:13:12", "remaining_time": "1:53:14"}
{"current_steps": 59, "total_steps": 555, "loss": 0.0195, "lr": 9.999108198370249e-05, "epoch": 0.5297418630751964, "percentage": 10.63, "elapsed_time": "0:13:25", "remaining_time": "1:52:53"}
{"current_steps": 60, "total_steps": 555, "loss": 0.0173, "lr": 9.998414611537681e-05, "epoch": 0.5387205387205387, "percentage": 10.81, "elapsed_time": "0:13:37", "remaining_time": "1:52:25"}
{"current_steps": 60, "total_steps": 555, "eval_loss": 0.019306931644678116, "epoch": 0.5387205387205387, "percentage": 10.81, "elapsed_time": "0:13:44", "remaining_time": "1:53:18"}
{"current_steps": 61, "total_steps": 555, "loss": 0.0199, "lr": 9.997522904167844e-05, "epoch": 0.547699214365881, "percentage": 10.99, "elapsed_time": "0:13:56", "remaining_time": "1:52:54"}
{"current_steps": 62, "total_steps": 555, "loss": 0.0211, "lr": 9.996433111605052e-05, "epoch": 0.5566778900112234, "percentage": 11.17, "elapsed_time": "0:14:09", "remaining_time": "1:52:33"}
{"current_steps": 63, "total_steps": 555, "loss": 0.0181, "lr": 9.995145277045061e-05, "epoch": 0.5656565656565656, "percentage": 11.35, "elapsed_time": "0:14:20", "remaining_time": "1:52:02"}
{"current_steps": 64, "total_steps": 555, "loss": 0.0206, "lr": 9.993659451533353e-05, "epoch": 0.574635241301908, "percentage": 11.53, "elapsed_time": "0:14:32", "remaining_time": "1:51:31"}
{"current_steps": 65, "total_steps": 555, "loss": 0.0184, "lr": 9.991975693963107e-05, "epoch": 0.5836139169472503, "percentage": 11.71, "elapsed_time": "0:14:42", "remaining_time": "1:50:56"}
{"current_steps": 65, "total_steps": 555, "eval_loss": 0.019347084686160088, "epoch": 0.5836139169472503, "percentage": 11.71, "elapsed_time": "0:14:49", "remaining_time": "1:51:43"}
{"current_steps": 66, "total_steps": 555, "loss": 0.0193, "lr": 9.990094071072877e-05, "epoch": 0.5925925925925926, "percentage": 11.89, "elapsed_time": "0:15:00", "remaining_time": "1:51:10"}
{"current_steps": 67, "total_steps": 555, "loss": 0.0193, "lr": 9.988014657443941e-05, "epoch": 0.6015712682379349, "percentage": 12.07, "elapsed_time": "0:15:13", "remaining_time": "1:50:51"}
{"current_steps": 68, "total_steps": 555, "loss": 0.0184, "lr": 9.985737535497337e-05, "epoch": 0.6105499438832772, "percentage": 12.25, "elapsed_time": "0:15:25", "remaining_time": "1:50:30"}
{"current_steps": 69, "total_steps": 555, "loss": 0.0183, "lr": 9.983262795490613e-05, "epoch": 0.6195286195286195, "percentage": 12.43, "elapsed_time": "0:15:36", "remaining_time": "1:49:57"}
{"current_steps": 70, "total_steps": 555, "loss": 0.0194, "lr": 9.980590535514233e-05, "epoch": 0.6285072951739619, "percentage": 12.61, "elapsed_time": "0:15:49", "remaining_time": "1:49:37"}
{"current_steps": 70, "total_steps": 555, "eval_loss": 0.01905701868236065, "epoch": 0.6285072951739619, "percentage": 12.61, "elapsed_time": "0:15:55", "remaining_time": "1:50:20"}
{"current_steps": 71, "total_steps": 555, "loss": 0.0189, "lr": 9.9777208614877e-05, "epoch": 0.6374859708193041, "percentage": 12.79, "elapsed_time": "0:16:06", "remaining_time": "1:49:48"}
{"current_steps": 72, "total_steps": 555, "loss": 0.0191, "lr": 9.97465388715535e-05, "epoch": 0.6464646464646465, "percentage": 12.97, "elapsed_time": "0:16:18", "remaining_time": "1:49:26"}
{"current_steps": 73, "total_steps": 555, "loss": 0.0187, "lr": 9.971389734081848e-05, "epoch": 0.6554433221099888, "percentage": 13.15, "elapsed_time": "0:16:31", "remaining_time": "1:49:08"}
{"current_steps": 74, "total_steps": 555, "loss": 0.0177, "lr": 9.967928531647374e-05, "epoch": 0.6644219977553311, "percentage": 13.33, "elapsed_time": "0:16:44", "remaining_time": "1:48:48"}
{"current_steps": 75, "total_steps": 555, "loss": 0.0182, "lr": 9.96427041704248e-05, "epoch": 0.6734006734006734, "percentage": 13.51, "elapsed_time": "0:16:57", "remaining_time": "1:48:30"}
{"current_steps": 75, "total_steps": 555, "eval_loss": 0.018457170575857162, "epoch": 0.6734006734006734, "percentage": 13.51, "elapsed_time": "0:17:03", "remaining_time": "1:49:10"}
{"current_steps": 76, "total_steps": 555, "loss": 0.0169, "lr": 9.960415535262671e-05, "epoch": 0.6823793490460157, "percentage": 13.69, "elapsed_time": "0:17:15", "remaining_time": "1:48:48"}
{"current_steps": 77, "total_steps": 555, "loss": 0.0184, "lr": 9.956364039102642e-05, "epoch": 0.691358024691358, "percentage": 13.87, "elapsed_time": "0:17:28", "remaining_time": "1:48:29"}
{"current_steps": 78, "total_steps": 555, "loss": 0.0187, "lr": 9.952116089150232e-05, "epoch": 0.7003367003367004, "percentage": 14.05, "elapsed_time": "0:17:40", "remaining_time": "1:48:03"}
{"current_steps": 79, "total_steps": 555, "loss": 0.0166, "lr": 9.947671853780054e-05, "epoch": 0.7093153759820426, "percentage": 14.23, "elapsed_time": "0:17:52", "remaining_time": "1:47:43"}
{"current_steps": 80, "total_steps": 555, "loss": 0.0169, "lr": 9.943031509146825e-05, "epoch": 0.7182940516273849, "percentage": 14.41, "elapsed_time": "0:18:04", "remaining_time": "1:47:19"}
{"current_steps": 80, "total_steps": 555, "eval_loss": 0.01826069876551628, "epoch": 0.7182940516273849, "percentage": 14.41, "elapsed_time": "0:18:10", "remaining_time": "1:47:57"}
{"current_steps": 81, "total_steps": 555, "loss": 0.0172, "lr": 9.938195239178374e-05, "epoch": 0.7272727272727273, "percentage": 14.59, "elapsed_time": "0:18:23", "remaining_time": "1:47:35"}
{"current_steps": 82, "total_steps": 555, "loss": 0.0183, "lr": 9.933163235568367e-05, "epoch": 0.7362514029180696, "percentage": 14.77, "elapsed_time": "0:18:34", "remaining_time": "1:47:08"}
{"current_steps": 83, "total_steps": 555, "loss": 0.0171, "lr": 9.927935697768698e-05, "epoch": 0.745230078563412, "percentage": 14.95, "elapsed_time": "0:18:46", "remaining_time": "1:46:44"}
{"current_steps": 84, "total_steps": 555, "loss": 0.0214, "lr": 9.922512832981584e-05, "epoch": 0.7542087542087542, "percentage": 15.14, "elapsed_time": "0:18:58", "remaining_time": "1:46:24"}
{"current_steps": 85, "total_steps": 555, "loss": 0.0176, "lr": 9.916894856151357e-05, "epoch": 0.7631874298540965, "percentage": 15.32, "elapsed_time": "0:19:10", "remaining_time": "1:46:02"}
{"current_steps": 85, "total_steps": 555, "eval_loss": 0.017831869423389435, "epoch": 0.7631874298540965, "percentage": 15.32, "elapsed_time": "0:19:16", "remaining_time": "1:46:36"}
{"current_steps": 86, "total_steps": 555, "loss": 0.0156, "lr": 9.91108198995594e-05, "epoch": 0.7721661054994389, "percentage": 15.5, "elapsed_time": "0:19:27", "remaining_time": "1:46:08"}
{"current_steps": 87, "total_steps": 555, "loss": 0.0186, "lr": 9.905074464798024e-05, "epoch": 0.7811447811447811, "percentage": 15.68, "elapsed_time": "0:19:40", "remaining_time": "1:45:47"}
{"current_steps": 88, "total_steps": 555, "loss": 0.0163, "lr": 9.898872518795932e-05, "epoch": 0.7901234567901234, "percentage": 15.86, "elapsed_time": "0:19:51", "remaining_time": "1:45:22"}
{"current_steps": 89, "total_steps": 555, "loss": 0.0173, "lr": 9.892476397774186e-05, "epoch": 0.7991021324354658, "percentage": 16.04, "elapsed_time": "0:20:03", "remaining_time": "1:44:59"}
{"current_steps": 90, "total_steps": 555, "loss": 0.0158, "lr": 9.885886355253758e-05, "epoch": 0.8080808080808081, "percentage": 16.22, "elapsed_time": "0:20:15", "remaining_time": "1:44:38"}
{"current_steps": 90, "total_steps": 555, "eval_loss": 0.017583945766091347, "epoch": 0.8080808080808081, "percentage": 16.22, "elapsed_time": "0:20:21", "remaining_time": "1:45:11"}
{"current_steps": 91, "total_steps": 555, "loss": 0.0178, "lr": 9.879102652442024e-05, "epoch": 0.8170594837261503, "percentage": 16.4, "elapsed_time": "0:20:33", "remaining_time": "1:44:50"}
{"current_steps": 92, "total_steps": 555, "loss": 0.0158, "lr": 9.872125558222409e-05, "epoch": 0.8260381593714927, "percentage": 16.58, "elapsed_time": "0:20:43", "remaining_time": "1:44:19"}
{"current_steps": 93, "total_steps": 555, "loss": 0.0162, "lr": 9.864955349143734e-05, "epoch": 0.835016835016835, "percentage": 16.76, "elapsed_time": "0:20:54", "remaining_time": "1:43:51"}
{"current_steps": 94, "total_steps": 555, "loss": 0.0154, "lr": 9.857592309409247e-05, "epoch": 0.8439955106621774, "percentage": 16.94, "elapsed_time": "0:21:05", "remaining_time": "1:43:25"}
{"current_steps": 95, "total_steps": 555, "loss": 0.02, "lr": 9.850036730865364e-05, "epoch": 0.8529741863075196, "percentage": 17.12, "elapsed_time": "0:21:16", "remaining_time": "1:43:00"}
{"current_steps": 95, "total_steps": 555, "eval_loss": 0.017212502658367157, "epoch": 0.8529741863075196, "percentage": 17.12, "elapsed_time": "0:21:22", "remaining_time": "1:43:31"}
{"current_steps": 96, "total_steps": 555, "loss": 0.0194, "lr": 9.842288912990096e-05, "epoch": 0.8619528619528619, "percentage": 17.3, "elapsed_time": "0:21:34", "remaining_time": "1:43:09"}
{"current_steps": 97, "total_steps": 555, "loss": 0.0152, "lr": 9.83434916288119e-05, "epoch": 0.8709315375982043, "percentage": 17.48, "elapsed_time": "0:21:45", "remaining_time": "1:42:44"}
{"current_steps": 98, "total_steps": 555, "loss": 0.017, "lr": 9.82621779524394e-05, "epoch": 0.8799102132435466, "percentage": 17.66, "elapsed_time": "0:21:58", "remaining_time": "1:42:29"}
{"current_steps": 99, "total_steps": 555, "loss": 0.0158, "lr": 9.817895132378725e-05, "epoch": 0.8888888888888888, "percentage": 17.84, "elapsed_time": "0:22:12", "remaining_time": "1:42:15"}
{"current_steps": 100, "total_steps": 555, "loss": 0.0165, "lr": 9.809381504168234e-05, "epoch": 0.8978675645342312, "percentage": 18.02, "elapsed_time": "0:22:25", "remaining_time": "1:42:00"}
{"current_steps": 100, "total_steps": 555, "eval_loss": 0.017287207767367363, "epoch": 0.8978675645342312, "percentage": 18.02, "elapsed_time": "0:22:31", "remaining_time": "1:42:29"}
{"current_steps": 101, "total_steps": 555, "loss": 0.0169, "lr": 9.800677248064382e-05, "epoch": 0.9068462401795735, "percentage": 18.2, "elapsed_time": "0:22:50", "remaining_time": "1:42:41"}
{"current_steps": 102, "total_steps": 555, "loss": 0.0152, "lr": 9.791782709074944e-05, "epoch": 0.9158249158249159, "percentage": 18.38, "elapsed_time": "0:23:02", "remaining_time": "1:42:18"}
{"current_steps": 103, "total_steps": 555, "loss": 0.017, "lr": 9.782698239749873e-05, "epoch": 0.9248035914702581, "percentage": 18.56, "elapsed_time": "0:23:13", "remaining_time": "1:41:56"}
{"current_steps": 104, "total_steps": 555, "loss": 0.0172, "lr": 9.77342420016733e-05, "epoch": 0.9337822671156004, "percentage": 18.74, "elapsed_time": "0:23:25", "remaining_time": "1:41:35"}
{"current_steps": 105, "total_steps": 555, "loss": 0.0181, "lr": 9.763960957919413e-05, "epoch": 0.9427609427609428, "percentage": 18.92, "elapsed_time": "0:23:38", "remaining_time": "1:41:18"}
{"current_steps": 105, "total_steps": 555, "eval_loss": 0.016786355525255203, "epoch": 0.9427609427609428, "percentage": 18.92, "elapsed_time": "0:23:44", "remaining_time": "1:41:45"}
{"current_steps": 106, "total_steps": 555, "loss": 0.0165, "lr": 9.754308888097583e-05, "epoch": 0.9517396184062851, "percentage": 19.1, "elapsed_time": "0:23:57", "remaining_time": "1:41:30"}
{"current_steps": 107, "total_steps": 555, "loss": 0.0162, "lr": 9.744468373277797e-05, "epoch": 0.9607182940516273, "percentage": 19.28, "elapsed_time": "0:24:09", "remaining_time": "1:41:08"}
{"current_steps": 108, "total_steps": 555, "loss": 0.0158, "lr": 9.734439803505345e-05, "epoch": 0.9696969696969697, "percentage": 19.46, "elapsed_time": "0:24:21", "remaining_time": "1:40:50"}
{"current_steps": 109, "total_steps": 555, "loss": 0.0163, "lr": 9.724223576279395e-05, "epoch": 0.978675645342312, "percentage": 19.64, "elapsed_time": "0:24:33", "remaining_time": "1:40:29"}
{"current_steps": 110, "total_steps": 555, "loss": 0.0176, "lr": 9.713820096537225e-05, "epoch": 0.9876543209876543, "percentage": 19.82, "elapsed_time": "0:24:44", "remaining_time": "1:40:04"}
{"current_steps": 110, "total_steps": 555, "eval_loss": 0.016797909513115883, "epoch": 0.9876543209876543, "percentage": 19.82, "elapsed_time": "0:24:50", "remaining_time": "1:40:29"}
{"current_steps": 111, "total_steps": 555, "loss": 0.0166, "lr": 9.703229776638185e-05, "epoch": 0.9966329966329966, "percentage": 20.0, "elapsed_time": "0:25:02", "remaining_time": "1:40:08"}
{"current_steps": 112, "total_steps": 555, "loss": 0.0323, "lr": 9.692453036347351e-05, "epoch": 1.0078563411896746, "percentage": 20.18, "elapsed_time": "0:25:12", "remaining_time": "1:39:43"}
{"current_steps": 113, "total_steps": 555, "loss": 0.0159, "lr": 9.681490302818874e-05, "epoch": 1.0168350168350169, "percentage": 20.36, "elapsed_time": "0:25:24", "remaining_time": "1:39:21"}
{"current_steps": 114, "total_steps": 555, "loss": 0.0141, "lr": 9.670342010579065e-05, "epoch": 1.0258136924803591, "percentage": 20.54, "elapsed_time": "0:25:35", "remaining_time": "1:39:00"}
{"current_steps": 115, "total_steps": 555, "loss": 0.0184, "lr": 9.659008601509168e-05, "epoch": 1.0347923681257014, "percentage": 20.72, "elapsed_time": "0:25:48", "remaining_time": "1:38:43"}
{"current_steps": 115, "total_steps": 555, "eval_loss": 0.0183156318962574, "epoch": 1.0347923681257014, "percentage": 20.72, "elapsed_time": "0:25:54", "remaining_time": "1:39:06"}
{"current_steps": 116, "total_steps": 555, "loss": 0.0157, "lr": 9.647490524827834e-05, "epoch": 1.0437710437710437, "percentage": 20.9, "elapsed_time": "0:26:06", "remaining_time": "1:38:47"}
{"current_steps": 117, "total_steps": 555, "loss": 0.0152, "lr": 9.635788237073334e-05, "epoch": 1.0527497194163862, "percentage": 21.08, "elapsed_time": "0:26:17", "remaining_time": "1:38:24"}
{"current_steps": 118, "total_steps": 555, "loss": 0.0176, "lr": 9.623902202085444e-05, "epoch": 1.0617283950617284, "percentage": 21.26, "elapsed_time": "0:26:27", "remaining_time": "1:37:57"}
{"current_steps": 119, "total_steps": 555, "loss": 0.0156, "lr": 9.611832890987076e-05, "epoch": 1.0707070707070707, "percentage": 21.44, "elapsed_time": "0:26:39", "remaining_time": "1:37:38"}
{"current_steps": 120, "total_steps": 555, "loss": 0.0162, "lr": 9.599580782165598e-05, "epoch": 1.079685746352413, "percentage": 21.62, "elapsed_time": "0:26:51", "remaining_time": "1:37:20"}
{"current_steps": 120, "total_steps": 555, "eval_loss": 0.017910869792103767, "epoch": 1.079685746352413, "percentage": 21.62, "elapsed_time": "0:26:57", "remaining_time": "1:37:43"}
{"current_steps": 121, "total_steps": 555, "loss": 0.0161, "lr": 9.587146361253868e-05, "epoch": 1.0886644219977553, "percentage": 21.8, "elapsed_time": "0:27:09", "remaining_time": "1:37:25"}
{"current_steps": 122, "total_steps": 555, "loss": 0.0178, "lr": 9.57453012111099e-05, "epoch": 1.0976430976430978, "percentage": 21.98, "elapsed_time": "0:27:22", "remaining_time": "1:37:09"}
{"current_steps": 123, "total_steps": 555, "loss": 0.017, "lr": 9.561732561802778e-05, "epoch": 1.10662177328844, "percentage": 22.16, "elapsed_time": "0:27:35", "remaining_time": "1:36:52"}
{"current_steps": 124, "total_steps": 555, "loss": 0.017, "lr": 9.548754190581939e-05, "epoch": 1.1156004489337823, "percentage": 22.34, "elapsed_time": "0:27:47", "remaining_time": "1:36:34"}
{"current_steps": 125, "total_steps": 555, "loss": 0.017, "lr": 9.53559552186796e-05, "epoch": 1.1245791245791246, "percentage": 22.52, "elapsed_time": "0:28:00", "remaining_time": "1:36:19"}
{"current_steps": 125, "total_steps": 555, "eval_loss": 0.016845999285578728, "epoch": 1.1245791245791246, "percentage": 22.52, "elapsed_time": "0:28:06", "remaining_time": "1:36:41"}
{"current_steps": 126, "total_steps": 555, "loss": 0.0123, "lr": 9.522257077226717e-05, "epoch": 1.1335578002244668, "percentage": 22.7, "elapsed_time": "0:28:18", "remaining_time": "1:36:24"}
{"current_steps": 127, "total_steps": 555, "loss": 0.017, "lr": 9.508739385349812e-05, "epoch": 1.142536475869809, "percentage": 22.88, "elapsed_time": "0:28:29", "remaining_time": "1:36:02"}
{"current_steps": 128, "total_steps": 555, "loss": 0.0156, "lr": 9.49504298203361e-05, "epoch": 1.1515151515151516, "percentage": 23.06, "elapsed_time": "0:28:42", "remaining_time": "1:35:46"}
{"current_steps": 129, "total_steps": 555, "loss": 0.0154, "lr": 9.481168410158003e-05, "epoch": 1.1604938271604939, "percentage": 23.24, "elapsed_time": "0:28:55", "remaining_time": "1:35:31"}
{"current_steps": 130, "total_steps": 555, "loss": 0.0143, "lr": 9.467116219664894e-05, "epoch": 1.1694725028058361, "percentage": 23.42, "elapsed_time": "0:29:08", "remaining_time": "1:35:16"}
{"current_steps": 130, "total_steps": 555, "eval_loss": 0.016655858606100082, "epoch": 1.1694725028058361, "percentage": 23.42, "elapsed_time": "0:29:14", "remaining_time": "1:35:36"}
{"current_steps": 131, "total_steps": 555, "loss": 0.015, "lr": 9.45288696753639e-05, "epoch": 1.1784511784511784, "percentage": 23.6, "elapsed_time": "0:29:27", "remaining_time": "1:35:21"}
{"current_steps": 132, "total_steps": 555, "loss": 0.016, "lr": 9.438481217772744e-05, "epoch": 1.1874298540965207, "percentage": 23.78, "elapsed_time": "0:29:40", "remaining_time": "1:35:06"}
{"current_steps": 133, "total_steps": 555, "loss": 0.0131, "lr": 9.423899541369978e-05, "epoch": 1.1964085297418632, "percentage": 23.96, "elapsed_time": "0:29:54", "remaining_time": "1:34:54"}
{"current_steps": 134, "total_steps": 555, "loss": 0.016, "lr": 9.409142516297269e-05, "epoch": 1.2053872053872055, "percentage": 24.14, "elapsed_time": "0:30:07", "remaining_time": "1:34:37"}
{"current_steps": 135, "total_steps": 555, "loss": 0.0177, "lr": 9.394210727474028e-05, "epoch": 1.2143658810325477, "percentage": 24.32, "elapsed_time": "0:30:17", "remaining_time": "1:34:15"}
{"current_steps": 135, "total_steps": 555, "eval_loss": 0.016584472730755806, "epoch": 1.2143658810325477, "percentage": 24.32, "elapsed_time": "0:30:23", "remaining_time": "1:34:34"}
{"current_steps": 136, "total_steps": 555, "loss": 0.0163, "lr": 9.379104766746722e-05, "epoch": 1.22334455667789, "percentage": 24.5, "elapsed_time": "0:30:35", "remaining_time": "1:34:14"}
{"current_steps": 137, "total_steps": 555, "loss": 0.0138, "lr": 9.363825232865413e-05, "epoch": 1.2323232323232323, "percentage": 24.68, "elapsed_time": "0:30:47", "remaining_time": "1:33:55"}
{"current_steps": 138, "total_steps": 555, "loss": 0.0119, "lr": 9.348372731460023e-05, "epoch": 1.2413019079685745, "percentage": 24.86, "elapsed_time": "0:30:58", "remaining_time": "1:33:35"}
{"current_steps": 139, "total_steps": 555, "loss": 0.0166, "lr": 9.332747875016332e-05, "epoch": 1.250280583613917, "percentage": 25.05, "elapsed_time": "0:31:10", "remaining_time": "1:33:19"}
{"current_steps": 140, "total_steps": 555, "loss": 0.0138, "lr": 9.316951282851707e-05, "epoch": 1.2592592592592593, "percentage": 25.23, "elapsed_time": "0:31:22", "remaining_time": "1:32:58"}
{"current_steps": 140, "total_steps": 555, "eval_loss": 0.01608719676733017, "epoch": 1.2592592592592593, "percentage": 25.23, "elapsed_time": "0:31:28", "remaining_time": "1:33:17"}
{"current_steps": 141, "total_steps": 555, "loss": 0.0157, "lr": 9.300983581090541e-05, "epoch": 1.2682379349046016, "percentage": 25.41, "elapsed_time": "0:31:38", "remaining_time": "1:32:55"}
{"current_steps": 142, "total_steps": 555, "loss": 0.0142, "lr": 9.284845402639446e-05, "epoch": 1.2772166105499438, "percentage": 25.59, "elapsed_time": "0:31:51", "remaining_time": "1:32:38"}
{"current_steps": 143, "total_steps": 555, "loss": 0.0148, "lr": 9.26853738716216e-05, "epoch": 1.2861952861952861, "percentage": 25.77, "elapsed_time": "0:32:02", "remaining_time": "1:32:18"}
{"current_steps": 144, "total_steps": 555, "loss": 0.0153, "lr": 9.2520601810542e-05, "epoch": 1.2951739618406286, "percentage": 25.95, "elapsed_time": "0:32:13", "remaining_time": "1:31:58"}
{"current_steps": 145, "total_steps": 555, "loss": 0.0149, "lr": 9.235414437417234e-05, "epoch": 1.3041526374859709, "percentage": 26.13, "elapsed_time": "0:32:25", "remaining_time": "1:31:40"}
{"current_steps": 145, "total_steps": 555, "eval_loss": 0.01574772223830223, "epoch": 1.3041526374859709, "percentage": 26.13, "elapsed_time": "0:32:31", "remaining_time": "1:31:57"}
{"current_steps": 146, "total_steps": 555, "loss": 0.0124, "lr": 9.2186008160332e-05, "epoch": 1.3131313131313131, "percentage": 26.31, "elapsed_time": "0:32:43", "remaining_time": "1:31:39"}
{"current_steps": 147, "total_steps": 555, "loss": 0.0188, "lr": 9.201619983338153e-05, "epoch": 1.3221099887766554, "percentage": 26.49, "elapsed_time": "0:32:55", "remaining_time": "1:31:23"}
{"current_steps": 148, "total_steps": 555, "loss": 0.0128, "lr": 9.18447261239584e-05, "epoch": 1.3310886644219977, "percentage": 26.67, "elapsed_time": "0:33:05", "remaining_time": "1:31:01"}
{"current_steps": 149, "total_steps": 555, "loss": 0.0142, "lr": 9.167159382871039e-05, "epoch": 1.34006734006734, "percentage": 26.85, "elapsed_time": "0:33:17", "remaining_time": "1:30:43"}
{"current_steps": 150, "total_steps": 555, "loss": 0.0162, "lr": 9.149680981002609e-05, "epoch": 1.3490460157126825, "percentage": 27.03, "elapsed_time": "0:33:28", "remaining_time": "1:30:23"}
{"current_steps": 150, "total_steps": 555, "eval_loss": 0.0159834623336792, "epoch": 1.3490460157126825, "percentage": 27.03, "elapsed_time": "0:33:35", "remaining_time": "1:30:40"}
{"current_steps": 151, "total_steps": 555, "loss": 0.0149, "lr": 9.13203809957629e-05, "epoch": 1.3580246913580247, "percentage": 27.21, "elapsed_time": "0:33:53", "remaining_time": "1:30:40"}
{"current_steps": 152, "total_steps": 555, "loss": 0.0166, "lr": 9.114231437897244e-05, "epoch": 1.367003367003367, "percentage": 27.39, "elapsed_time": "0:34:05", "remaining_time": "1:30:22"}
{"current_steps": 153, "total_steps": 555, "loss": 0.0146, "lr": 9.096261701762342e-05, "epoch": 1.3759820426487093, "percentage": 27.57, "elapsed_time": "0:34:17", "remaining_time": "1:30:04"}
{"current_steps": 154, "total_steps": 555, "loss": 0.0141, "lr": 9.078129603432181e-05, "epoch": 1.3849607182940518, "percentage": 27.75, "elapsed_time": "0:34:28", "remaining_time": "1:29:45"}
{"current_steps": 155, "total_steps": 555, "loss": 0.0148, "lr": 9.059835861602853e-05, "epoch": 1.393939393939394, "percentage": 27.93, "elapsed_time": "0:34:41", "remaining_time": "1:29:30"}
{"current_steps": 155, "total_steps": 555, "eval_loss": 0.015588033944368362, "epoch": 1.393939393939394, "percentage": 27.93, "elapsed_time": "0:34:47", "remaining_time": "1:29:46"}
{"current_steps": 156, "total_steps": 555, "loss": 0.0152, "lr": 9.041381201377468e-05, "epoch": 1.4029180695847363, "percentage": 28.11, "elapsed_time": "0:34:58", "remaining_time": "1:29:27"}
{"current_steps": 157, "total_steps": 555, "loss": 0.0146, "lr": 9.0227663542374e-05, "epoch": 1.4118967452300786, "percentage": 28.29, "elapsed_time": "0:35:11", "remaining_time": "1:29:11"}
{"current_steps": 158, "total_steps": 555, "loss": 0.015, "lr": 9.003992058013302e-05, "epoch": 1.4208754208754208, "percentage": 28.47, "elapsed_time": "0:35:21", "remaining_time": "1:28:51"}
{"current_steps": 159, "total_steps": 555, "loss": 0.0144, "lr": 8.985059056855858e-05, "epoch": 1.4298540965207631, "percentage": 28.65, "elapsed_time": "0:35:34", "remaining_time": "1:28:37"}
{"current_steps": 160, "total_steps": 555, "loss": 0.0168, "lr": 8.965968101206291e-05, "epoch": 1.4388327721661054, "percentage": 28.83, "elapsed_time": "0:35:47", "remaining_time": "1:28:22"}
{"current_steps": 160, "total_steps": 555, "eval_loss": 0.015444611199200153, "epoch": 1.4388327721661054, "percentage": 28.83, "elapsed_time": "0:35:54", "remaining_time": "1:28:38"}
{"current_steps": 161, "total_steps": 555, "loss": 0.0147, "lr": 8.94671994776661e-05, "epoch": 1.4478114478114479, "percentage": 29.01, "elapsed_time": "0:36:05", "remaining_time": "1:28:18"}
{"current_steps": 162, "total_steps": 555, "loss": 0.0129, "lr": 8.927315359469626e-05, "epoch": 1.4567901234567902, "percentage": 29.19, "elapsed_time": "0:36:17", "remaining_time": "1:28:02"}
{"current_steps": 163, "total_steps": 555, "loss": 0.019, "lr": 8.907755105448704e-05, "epoch": 1.4657687991021324, "percentage": 29.37, "elapsed_time": "0:36:29", "remaining_time": "1:27:45"}
{"current_steps": 164, "total_steps": 555, "loss": 0.0157, "lr": 8.888039961007282e-05, "epoch": 1.4747474747474747, "percentage": 29.55, "elapsed_time": "0:36:40", "remaining_time": "1:27:26"}
{"current_steps": 165, "total_steps": 555, "loss": 0.0148, "lr": 8.868170707588142e-05, "epoch": 1.4837261503928172, "percentage": 29.73, "elapsed_time": "0:36:51", "remaining_time": "1:27:08"}
{"current_steps": 165, "total_steps": 555, "eval_loss": 0.015270690433681011, "epoch": 1.4837261503928172, "percentage": 29.73, "elapsed_time": "0:36:58", "remaining_time": "1:27:22"}
{"current_steps": 166, "total_steps": 555, "loss": 0.0133, "lr": 8.848148132742431e-05, "epoch": 1.4927048260381595, "percentage": 29.91, "elapsed_time": "0:37:09", "remaining_time": "1:27:03"}
{"current_steps": 167, "total_steps": 555, "loss": 0.0135, "lr": 8.827973030098448e-05, "epoch": 1.5016835016835017, "percentage": 30.09, "elapsed_time": "0:37:20", "remaining_time": "1:26:44"}
{"current_steps": 168, "total_steps": 555, "loss": 0.0159, "lr": 8.807646199330187e-05, "epoch": 1.510662177328844, "percentage": 30.27, "elapsed_time": "0:37:30", "remaining_time": "1:26:24"}
{"current_steps": 169, "total_steps": 555, "loss": 0.0129, "lr": 8.787168446125638e-05, "epoch": 1.5196408529741863, "percentage": 30.45, "elapsed_time": "0:37:44", "remaining_time": "1:26:11"}
{"current_steps": 170, "total_steps": 555, "loss": 0.0146, "lr": 8.766540582154859e-05, "epoch": 1.5286195286195285, "percentage": 30.63, "elapsed_time": "0:37:56", "remaining_time": "1:25:55"}
{"current_steps": 170, "total_steps": 555, "eval_loss": 0.015371887013316154, "epoch": 1.5286195286195285, "percentage": 30.63, "elapsed_time": "0:38:02", "remaining_time": "1:26:09"}
{"current_steps": 171, "total_steps": 555, "loss": 0.015, "lr": 8.745763425037797e-05, "epoch": 1.5375982042648708, "percentage": 30.81, "elapsed_time": "0:38:14", "remaining_time": "1:25:52"}
{"current_steps": 172, "total_steps": 555, "loss": 0.0153, "lr": 8.724837798311882e-05, "epoch": 1.546576879910213, "percentage": 30.99, "elapsed_time": "0:38:26", "remaining_time": "1:25:36"}
{"current_steps": 173, "total_steps": 555, "loss": 0.0122, "lr": 8.703764531399392e-05, "epoch": 1.5555555555555556, "percentage": 31.17, "elapsed_time": "0:38:39", "remaining_time": "1:25:20"}
{"current_steps": 174, "total_steps": 555, "loss": 0.0144, "lr": 8.682544459574562e-05, "epoch": 1.5645342312008979, "percentage": 31.35, "elapsed_time": "0:38:51", "remaining_time": "1:25:04"}
{"current_steps": 175, "total_steps": 555, "loss": 0.0137, "lr": 8.661178423930491e-05, "epoch": 1.5735129068462403, "percentage": 31.53, "elapsed_time": "0:39:04", "remaining_time": "1:24:49"}
{"current_steps": 175, "total_steps": 555, "eval_loss": 0.014978926628828049, "epoch": 1.5735129068462403, "percentage": 31.53, "elapsed_time": "0:39:10", "remaining_time": "1:25:03"}
{"current_steps": 176, "total_steps": 555, "loss": 0.0177, "lr": 8.639667271345798e-05, "epoch": 1.5824915824915826, "percentage": 31.71, "elapsed_time": "0:39:23", "remaining_time": "1:24:49"}
{"current_steps": 177, "total_steps": 555, "loss": 0.0115, "lr": 8.618011854451056e-05, "epoch": 1.5914702581369249, "percentage": 31.89, "elapsed_time": "0:39:36", "remaining_time": "1:24:34"}
{"current_steps": 178, "total_steps": 555, "loss": 0.0167, "lr": 8.596213031594991e-05, "epoch": 1.6004489337822672, "percentage": 32.07, "elapsed_time": "0:39:48", "remaining_time": "1:24:18"}
{"current_steps": 179, "total_steps": 555, "loss": 0.0153, "lr": 8.57427166681047e-05, "epoch": 1.6094276094276094, "percentage": 32.25, "elapsed_time": "0:39:59", "remaining_time": "1:24:00"}
{"current_steps": 180, "total_steps": 555, "loss": 0.0144, "lr": 8.552188629780244e-05, "epoch": 1.6184062850729517, "percentage": 32.43, "elapsed_time": "0:40:11", "remaining_time": "1:23:44"}
{"current_steps": 180, "total_steps": 555, "eval_loss": 0.014994567260146141, "epoch": 1.6184062850729517, "percentage": 32.43, "elapsed_time": "0:40:18", "remaining_time": "1:23:57"}
{"current_steps": 181, "total_steps": 555, "loss": 0.0125, "lr": 8.529964795802485e-05, "epoch": 1.627384960718294, "percentage": 32.61, "elapsed_time": "0:40:27", "remaining_time": "1:23:36"}
{"current_steps": 182, "total_steps": 555, "loss": 0.0152, "lr": 8.507601045756085e-05, "epoch": 1.6363636363636362, "percentage": 32.79, "elapsed_time": "0:40:39", "remaining_time": "1:23:19"}
{"current_steps": 183, "total_steps": 555, "loss": 0.0125, "lr": 8.485098266065744e-05, "epoch": 1.6453423120089785, "percentage": 32.97, "elapsed_time": "0:40:51", "remaining_time": "1:23:03"}
{"current_steps": 184, "total_steps": 555, "loss": 0.0163, "lr": 8.462457348666835e-05, "epoch": 1.654320987654321, "percentage": 33.15, "elapsed_time": "0:41:05", "remaining_time": "1:22:50"}
{"current_steps": 185, "total_steps": 555, "loss": 0.0129, "lr": 8.439679190970052e-05, "epoch": 1.6632996632996633, "percentage": 33.33, "elapsed_time": "0:41:15", "remaining_time": "1:22:31"}
{"current_steps": 185, "total_steps": 555, "eval_loss": 0.014757846482098103, "epoch": 1.6632996632996633, "percentage": 33.33, "elapsed_time": "0:41:22", "remaining_time": "1:22:44"}
{"current_steps": 186, "total_steps": 555, "loss": 0.015, "lr": 8.416764695825835e-05, "epoch": 1.6722783389450058, "percentage": 33.51, "elapsed_time": "0:41:34", "remaining_time": "1:22:29"}
{"current_steps": 187, "total_steps": 555, "loss": 0.0166, "lr": 8.39371477148859e-05, "epoch": 1.681257014590348, "percentage": 33.69, "elapsed_time": "0:41:46", "remaining_time": "1:22:12"}
{"current_steps": 188, "total_steps": 555, "loss": 0.0131, "lr": 8.370530331580686e-05, "epoch": 1.6902356902356903, "percentage": 33.87, "elapsed_time": "0:41:58", "remaining_time": "1:21:56"}
{"current_steps": 189, "total_steps": 555, "loss": 0.0158, "lr": 8.347212295056239e-05, "epoch": 1.6992143658810326, "percentage": 34.05, "elapsed_time": "0:42:10", "remaining_time": "1:21:40"}
{"current_steps": 190, "total_steps": 555, "loss": 0.0139, "lr": 8.323761586164695e-05, "epoch": 1.7081930415263749, "percentage": 34.23, "elapsed_time": "0:42:22", "remaining_time": "1:21:23"}
{"current_steps": 190, "total_steps": 555, "eval_loss": 0.014534353278577328, "epoch": 1.7081930415263749, "percentage": 34.23, "elapsed_time": "0:42:28", "remaining_time": "1:21:35"}
{"current_steps": 191, "total_steps": 555, "loss": 0.0116, "lr": 8.300179134414188e-05, "epoch": 1.7171717171717171, "percentage": 34.41, "elapsed_time": "0:42:39", "remaining_time": "1:21:17"}
{"current_steps": 192, "total_steps": 555, "loss": 0.0127, "lr": 8.276465874534702e-05, "epoch": 1.7261503928170594, "percentage": 34.59, "elapsed_time": "0:42:49", "remaining_time": "1:20:57"}
{"current_steps": 193, "total_steps": 555, "loss": 0.0142, "lr": 8.252622746441021e-05, "epoch": 1.7351290684624017, "percentage": 34.77, "elapsed_time": "0:43:01", "remaining_time": "1:20:41"}
{"current_steps": 194, "total_steps": 555, "loss": 0.0151, "lr": 8.228650695195472e-05, "epoch": 1.7441077441077442, "percentage": 34.95, "elapsed_time": "0:43:13", "remaining_time": "1:20:25"}
{"current_steps": 195, "total_steps": 555, "loss": 0.013, "lr": 8.204550670970469e-05, "epoch": 1.7530864197530864, "percentage": 35.14, "elapsed_time": "0:43:25", "remaining_time": "1:20:10"}
{"current_steps": 195, "total_steps": 555, "eval_loss": 0.014544461853802204, "epoch": 1.7530864197530864, "percentage": 35.14, "elapsed_time": "0:43:31", "remaining_time": "1:20:21"}
{"current_steps": 196, "total_steps": 555, "loss": 0.0146, "lr": 8.180323629010848e-05, "epoch": 1.7620650953984287, "percentage": 35.32, "elapsed_time": "0:43:44", "remaining_time": "1:20:07"}
{"current_steps": 197, "total_steps": 555, "loss": 0.0147, "lr": 8.155970529596006e-05, "epoch": 1.7710437710437712, "percentage": 35.5, "elapsed_time": "0:43:56", "remaining_time": "1:19:51"}
{"current_steps": 198, "total_steps": 555, "loss": 0.0132, "lr": 8.131492338001839e-05, "epoch": 1.7800224466891135, "percentage": 35.68, "elapsed_time": "0:44:08", "remaining_time": "1:19:35"}
{"current_steps": 199, "total_steps": 555, "loss": 0.0149, "lr": 8.106890024462481e-05, "epoch": 1.7890011223344557, "percentage": 35.86, "elapsed_time": "0:44:20", "remaining_time": "1:19:19"}
{"current_steps": 200, "total_steps": 555, "loss": 0.013, "lr": 8.082164564131845e-05, "epoch": 1.797979797979798, "percentage": 36.04, "elapsed_time": "0:44:33", "remaining_time": "1:19:05"}
{"current_steps": 200, "total_steps": 555, "eval_loss": 0.01437403354793787, "epoch": 1.797979797979798, "percentage": 36.04, "elapsed_time": "0:44:40", "remaining_time": "1:19:17"}
{"current_steps": 201, "total_steps": 555, "loss": 0.018, "lr": 8.057316937044977e-05, "epoch": 1.8069584736251403, "percentage": 36.22, "elapsed_time": "0:44:59", "remaining_time": "1:19:14"}
{"current_steps": 202, "total_steps": 555, "loss": 0.0151, "lr": 8.032348128079203e-05, "epoch": 1.8159371492704826, "percentage": 36.4, "elapsed_time": "0:45:10", "remaining_time": "1:18:57"}
{"current_steps": 203, "total_steps": 555, "loss": 0.0135, "lr": 8.0072591269151e-05, "epoch": 1.8249158249158248, "percentage": 36.58, "elapsed_time": "0:45:22", "remaining_time": "1:18:41"}
{"current_steps": 204, "total_steps": 555, "loss": 0.0141, "lr": 7.982050927997264e-05, "epoch": 1.833894500561167, "percentage": 36.76, "elapsed_time": "0:45:35", "remaining_time": "1:18:26"}
{"current_steps": 205, "total_steps": 555, "loss": 0.0124, "lr": 7.956724530494887e-05, "epoch": 1.8428731762065096, "percentage": 36.94, "elapsed_time": "0:45:46", "remaining_time": "1:18:09"}
{"current_steps": 205, "total_steps": 555, "eval_loss": 0.014373213052749634, "epoch": 1.8428731762065096, "percentage": 36.94, "elapsed_time": "0:45:52", "remaining_time": "1:18:19"}
{"current_steps": 206, "total_steps": 555, "loss": 0.0166, "lr": 7.931280938262169e-05, "epoch": 1.8518518518518519, "percentage": 37.12, "elapsed_time": "0:46:04", "remaining_time": "1:18:02"}
{"current_steps": 207, "total_steps": 555, "loss": 0.0129, "lr": 7.905721159798513e-05, "epoch": 1.8608305274971941, "percentage": 37.3, "elapsed_time": "0:46:16", "remaining_time": "1:17:48"}
{"current_steps": 208, "total_steps": 555, "loss": 0.0149, "lr": 7.880046208208563e-05, "epoch": 1.8698092031425366, "percentage": 37.48, "elapsed_time": "0:46:29", "remaining_time": "1:17:33"}
{"current_steps": 209, "total_steps": 555, "loss": 0.0134, "lr": 7.854257101162037e-05, "epoch": 1.878787878787879, "percentage": 37.66, "elapsed_time": "0:46:39", "remaining_time": "1:17:15"}
{"current_steps": 210, "total_steps": 555, "loss": 0.0135, "lr": 7.828354860853399e-05, "epoch": 1.8877665544332212, "percentage": 37.84, "elapsed_time": "0:46:52", "remaining_time": "1:17:00"}
{"current_steps": 210, "total_steps": 555, "eval_loss": 0.014275193214416504, "epoch": 1.8877665544332212, "percentage": 37.84, "elapsed_time": "0:46:58", "remaining_time": "1:17:10"}
{"current_steps": 211, "total_steps": 555, "loss": 0.012, "lr": 7.802340513961342e-05, "epoch": 1.8967452300785634, "percentage": 38.02, "elapsed_time": "0:47:10", "remaining_time": "1:16:54"}
{"current_steps": 212, "total_steps": 555, "loss": 0.0132, "lr": 7.776215091608085e-05, "epoch": 1.9057239057239057, "percentage": 38.2, "elapsed_time": "0:47:23", "remaining_time": "1:16:39"}
{"current_steps": 213, "total_steps": 555, "loss": 0.0126, "lr": 7.749979629318516e-05, "epoch": 1.914702581369248, "percentage": 38.38, "elapsed_time": "0:47:34", "remaining_time": "1:16:23"}
{"current_steps": 214, "total_steps": 555, "loss": 0.0135, "lr": 7.723635166979133e-05, "epoch": 1.9236812570145903, "percentage": 38.56, "elapsed_time": "0:47:46", "remaining_time": "1:16:07"}
{"current_steps": 215, "total_steps": 555, "loss": 0.0128, "lr": 7.697182748796841e-05, "epoch": 1.9326599326599325, "percentage": 38.74, "elapsed_time": "0:47:56", "remaining_time": "1:15:48"}
{"current_steps": 215, "total_steps": 555, "eval_loss": 0.014707864262163639, "epoch": 1.9326599326599325, "percentage": 38.74, "elapsed_time": "0:48:02", "remaining_time": "1:15:58"}
{"current_steps": 216, "total_steps": 555, "loss": 0.0131, "lr": 7.670623423257548e-05, "epoch": 1.941638608305275, "percentage": 38.92, "elapsed_time": "0:48:15", "remaining_time": "1:15:44"}
{"current_steps": 217, "total_steps": 555, "loss": 0.0137, "lr": 7.64395824308462e-05, "epoch": 1.9506172839506173, "percentage": 39.1, "elapsed_time": "0:48:27", "remaining_time": "1:15:28"}
{"current_steps": 218, "total_steps": 555, "loss": 0.0128, "lr": 7.617188265197148e-05, "epoch": 1.9595959595959596, "percentage": 39.28, "elapsed_time": "0:48:38", "remaining_time": "1:15:10"}
{"current_steps": 219, "total_steps": 555, "loss": 0.0142, "lr": 7.590314550668054e-05, "epoch": 1.968574635241302, "percentage": 39.46, "elapsed_time": "0:48:48", "remaining_time": "1:14:53"}
{"current_steps": 220, "total_steps": 555, "loss": 0.0149, "lr": 7.563338164682036e-05, "epoch": 1.9775533108866443, "percentage": 39.64, "elapsed_time": "0:49:00", "remaining_time": "1:14:38"}
{"current_steps": 220, "total_steps": 555, "eval_loss": 0.01433955691754818, "epoch": 1.9775533108866443, "percentage": 39.64, "elapsed_time": "0:49:07", "remaining_time": "1:14:47"}
{"current_steps": 221, "total_steps": 555, "loss": 0.0143, "lr": 7.536260176493348e-05, "epoch": 1.9865319865319866, "percentage": 39.82, "elapsed_time": "0:49:18", "remaining_time": "1:14:30"}
{"current_steps": 222, "total_steps": 555, "loss": 0.0134, "lr": 7.509081659383417e-05, "epoch": 1.9955106621773289, "percentage": 40.0, "elapsed_time": "0:49:31", "remaining_time": "1:14:16"}
{"current_steps": 223, "total_steps": 555, "loss": 0.0255, "lr": 7.481803690618304e-05, "epoch": 2.006734006734007, "percentage": 40.18, "elapsed_time": "0:49:43", "remaining_time": "1:14:01"}
{"current_steps": 224, "total_steps": 555, "loss": 0.0154, "lr": 7.454427351405999e-05, "epoch": 2.015712682379349, "percentage": 40.36, "elapsed_time": "0:49:56", "remaining_time": "1:13:47"}
{"current_steps": 225, "total_steps": 555, "loss": 0.0138, "lr": 7.426953726853574e-05, "epoch": 2.0246913580246915, "percentage": 40.54, "elapsed_time": "0:50:08", "remaining_time": "1:13:31"}
{"current_steps": 225, "total_steps": 555, "eval_loss": 0.014397691935300827, "epoch": 2.0246913580246915, "percentage": 40.54, "elapsed_time": "0:50:14", "remaining_time": "1:13:41"}
{"current_steps": 226, "total_steps": 555, "loss": 0.0113, "lr": 7.399383905924165e-05, "epoch": 2.0336700336700337, "percentage": 40.72, "elapsed_time": "0:50:24", "remaining_time": "1:13:22"}
{"current_steps": 227, "total_steps": 555, "loss": 0.0108, "lr": 7.371718981393815e-05, "epoch": 2.042648709315376, "percentage": 40.9, "elapsed_time": "0:50:35", "remaining_time": "1:13:06"}
{"current_steps": 228, "total_steps": 555, "loss": 0.0136, "lr": 7.343960049808156e-05, "epoch": 2.0516273849607183, "percentage": 41.08, "elapsed_time": "0:50:47", "remaining_time": "1:12:51"}
{"current_steps": 229, "total_steps": 555, "loss": 0.0129, "lr": 7.316108211438945e-05, "epoch": 2.0606060606060606, "percentage": 41.26, "elapsed_time": "0:50:57", "remaining_time": "1:12:33"}
{"current_steps": 230, "total_steps": 555, "loss": 0.0127, "lr": 7.288164570240463e-05, "epoch": 2.069584736251403, "percentage": 41.44, "elapsed_time": "0:51:08", "remaining_time": "1:12:15"}
{"current_steps": 230, "total_steps": 555, "eval_loss": 0.014279232360422611, "epoch": 2.069584736251403, "percentage": 41.44, "elapsed_time": "0:51:14", "remaining_time": "1:12:24"}
{"current_steps": 231, "total_steps": 555, "loss": 0.0121, "lr": 7.26013023380574e-05, "epoch": 2.078563411896745, "percentage": 41.62, "elapsed_time": "0:51:26", "remaining_time": "1:12:08"}
{"current_steps": 232, "total_steps": 555, "loss": 0.013, "lr": 7.232006313322667e-05, "epoch": 2.0875420875420874, "percentage": 41.8, "elapsed_time": "0:51:37", "remaining_time": "1:11:52"}
{"current_steps": 233, "total_steps": 555, "loss": 0.0127, "lr": 7.203793923529956e-05, "epoch": 2.0965207631874296, "percentage": 41.98, "elapsed_time": "0:51:48", "remaining_time": "1:11:35"}
{"current_steps": 234, "total_steps": 555, "loss": 0.0141, "lr": 7.175494182672939e-05, "epoch": 2.1054994388327724, "percentage": 42.16, "elapsed_time": "0:51:59", "remaining_time": "1:11:19"}
{"current_steps": 235, "total_steps": 555, "loss": 0.0116, "lr": 7.147108212459257e-05, "epoch": 2.1144781144781146, "percentage": 42.34, "elapsed_time": "0:52:12", "remaining_time": "1:11:04"}
{"current_steps": 235, "total_steps": 555, "eval_loss": 0.0142152588814497, "epoch": 2.1144781144781146, "percentage": 42.34, "elapsed_time": "0:52:18", "remaining_time": "1:11:13"}
{"current_steps": 236, "total_steps": 555, "loss": 0.011, "lr": 7.118637138014396e-05, "epoch": 2.123456790123457, "percentage": 42.52, "elapsed_time": "0:52:29", "remaining_time": "1:10:57"}
{"current_steps": 237, "total_steps": 555, "loss": 0.0137, "lr": 7.090082087837092e-05, "epoch": 2.132435465768799, "percentage": 42.7, "elapsed_time": "0:52:40", "remaining_time": "1:10:40"}
{"current_steps": 238, "total_steps": 555, "loss": 0.012, "lr": 7.061444193754596e-05, "epoch": 2.1414141414141414, "percentage": 42.88, "elapsed_time": "0:52:52", "remaining_time": "1:10:25"}
{"current_steps": 239, "total_steps": 555, "loss": 0.0119, "lr": 7.032724590877821e-05, "epoch": 2.1503928170594837, "percentage": 43.06, "elapsed_time": "0:53:03", "remaining_time": "1:10:09"}
{"current_steps": 240, "total_steps": 555, "loss": 0.0128, "lr": 7.003924417556343e-05, "epoch": 2.159371492704826, "percentage": 43.24, "elapsed_time": "0:53:15", "remaining_time": "1:09:53"}
{"current_steps": 240, "total_steps": 555, "eval_loss": 0.014257782138884068, "epoch": 2.159371492704826, "percentage": 43.24, "elapsed_time": "0:53:21", "remaining_time": "1:10:02"}
{"current_steps": 241, "total_steps": 555, "loss": 0.0109, "lr": 6.975044815333282e-05, "epoch": 2.1683501683501682, "percentage": 43.42, "elapsed_time": "0:53:33", "remaining_time": "1:09:46"}
{"current_steps": 242, "total_steps": 555, "loss": 0.0132, "lr": 6.946086928900054e-05, "epoch": 2.1773288439955105, "percentage": 43.6, "elapsed_time": "0:53:45", "remaining_time": "1:09:31"}
{"current_steps": 243, "total_steps": 555, "loss": 0.0106, "lr": 6.917051906051006e-05, "epoch": 2.186307519640853, "percentage": 43.78, "elapsed_time": "0:53:56", "remaining_time": "1:09:15"}
{"current_steps": 244, "total_steps": 555, "loss": 0.0103, "lr": 6.887940897637908e-05, "epoch": 2.1952861952861955, "percentage": 43.96, "elapsed_time": "0:54:10", "remaining_time": "1:09:02"}
{"current_steps": 245, "total_steps": 555, "loss": 0.0145, "lr": 6.858755057524354e-05, "epoch": 2.204264870931538, "percentage": 44.14, "elapsed_time": "0:54:22", "remaining_time": "1:08:47"}
{"current_steps": 245, "total_steps": 555, "eval_loss": 0.014115707948803902, "epoch": 2.204264870931538, "percentage": 44.14, "elapsed_time": "0:54:28", "remaining_time": "1:08:55"}
{"current_steps": 246, "total_steps": 555, "loss": 0.0118, "lr": 6.829495542540013e-05, "epoch": 2.21324354657688, "percentage": 44.32, "elapsed_time": "0:54:41", "remaining_time": "1:08:41"}
{"current_steps": 247, "total_steps": 555, "loss": 0.0142, "lr": 6.80016351243478e-05, "epoch": 2.2222222222222223, "percentage": 44.5, "elapsed_time": "0:54:53", "remaining_time": "1:08:27"}
{"current_steps": 248, "total_steps": 555, "loss": 0.0119, "lr": 6.77076012983281e-05, "epoch": 2.2312008978675646, "percentage": 44.68, "elapsed_time": "0:55:06", "remaining_time": "1:08:12"}
{"current_steps": 249, "total_steps": 555, "loss": 0.0096, "lr": 6.741286560186437e-05, "epoch": 2.240179573512907, "percentage": 44.86, "elapsed_time": "0:55:18", "remaining_time": "1:07:58"}
{"current_steps": 250, "total_steps": 555, "loss": 0.0147, "lr": 6.711743971729967e-05, "epoch": 2.249158249158249, "percentage": 45.05, "elapsed_time": "0:55:31", "remaining_time": "1:07:44"}
{"current_steps": 250, "total_steps": 555, "eval_loss": 0.013896584510803223, "epoch": 2.249158249158249, "percentage": 45.05, "elapsed_time": "0:55:37", "remaining_time": "1:07:51"}
{"current_steps": 251, "total_steps": 555, "loss": 0.0102, "lr": 6.682133535433393e-05, "epoch": 2.2581369248035914, "percentage": 45.23, "elapsed_time": "0:55:56", "remaining_time": "1:07:45"}
{"current_steps": 252, "total_steps": 555, "loss": 0.0147, "lr": 6.652456424955963e-05, "epoch": 2.2671156004489337, "percentage": 45.41, "elapsed_time": "0:56:10", "remaining_time": "1:07:32"}
{"current_steps": 253, "total_steps": 555, "loss": 0.0112, "lr": 6.622713816599673e-05, "epoch": 2.276094276094276, "percentage": 45.59, "elapsed_time": "0:56:22", "remaining_time": "1:07:18"}
{"current_steps": 254, "total_steps": 555, "loss": 0.013, "lr": 6.592906889262632e-05, "epoch": 2.285072951739618, "percentage": 45.77, "elapsed_time": "0:56:36", "remaining_time": "1:07:04"}
{"current_steps": 255, "total_steps": 555, "loss": 0.0114, "lr": 6.563036824392344e-05, "epoch": 2.2940516273849605, "percentage": 45.95, "elapsed_time": "0:56:49", "remaining_time": "1:06:50"}
{"current_steps": 255, "total_steps": 555, "eval_loss": 0.013858611695468426, "epoch": 2.2940516273849605, "percentage": 45.95, "elapsed_time": "0:56:55", "remaining_time": "1:06:58"}
{"current_steps": 256, "total_steps": 555, "loss": 0.0114, "lr": 6.533104805938873e-05, "epoch": 2.303030303030303, "percentage": 46.13, "elapsed_time": "0:57:07", "remaining_time": "1:06:43"}
{"current_steps": 257, "total_steps": 555, "loss": 0.0116, "lr": 6.503112020307916e-05, "epoch": 2.3120089786756455, "percentage": 46.31, "elapsed_time": "0:57:20", "remaining_time": "1:06:29"}
{"current_steps": 258, "total_steps": 555, "loss": 0.0133, "lr": 6.473059656313782e-05, "epoch": 2.3209876543209877, "percentage": 46.49, "elapsed_time": "0:57:31", "remaining_time": "1:06:13"}
{"current_steps": 259, "total_steps": 555, "loss": 0.0127, "lr": 6.442948905132266e-05, "epoch": 2.32996632996633, "percentage": 46.67, "elapsed_time": "0:57:42", "remaining_time": "1:05:57"}
{"current_steps": 260, "total_steps": 555, "loss": 0.0114, "lr": 6.412780960253436e-05, "epoch": 2.3389450056116723, "percentage": 46.85, "elapsed_time": "0:57:53", "remaining_time": "1:05:40"}
{"current_steps": 260, "total_steps": 555, "eval_loss": 0.01390094868838787, "epoch": 2.3389450056116723, "percentage": 46.85, "elapsed_time": "0:57:59", "remaining_time": "1:05:48"}
{"current_steps": 261, "total_steps": 555, "loss": 0.0122, "lr": 6.382557017434332e-05, "epoch": 2.3479236812570146, "percentage": 47.03, "elapsed_time": "0:58:12", "remaining_time": "1:05:33"}
{"current_steps": 262, "total_steps": 555, "loss": 0.0091, "lr": 6.352278274651561e-05, "epoch": 2.356902356902357, "percentage": 47.21, "elapsed_time": "0:58:23", "remaining_time": "1:05:17"}
{"current_steps": 263, "total_steps": 555, "loss": 0.0125, "lr": 6.321945932053822e-05, "epoch": 2.365881032547699, "percentage": 47.39, "elapsed_time": "0:58:37", "remaining_time": "1:05:05"}
{"current_steps": 264, "total_steps": 555, "loss": 0.0125, "lr": 6.291561191914333e-05, "epoch": 2.3748597081930414, "percentage": 47.57, "elapsed_time": "0:58:47", "remaining_time": "1:04:48"}
{"current_steps": 265, "total_steps": 555, "loss": 0.0112, "lr": 6.261125258583171e-05, "epoch": 2.3838383838383836, "percentage": 47.75, "elapsed_time": "0:59:00", "remaining_time": "1:04:34"}
{"current_steps": 265, "total_steps": 555, "eval_loss": 0.013734661974012852, "epoch": 2.3838383838383836, "percentage": 47.75, "elapsed_time": "0:59:06", "remaining_time": "1:04:41"}
{"current_steps": 266, "total_steps": 555, "loss": 0.0134, "lr": 6.230639338439549e-05, "epoch": 2.3928170594837264, "percentage": 47.93, "elapsed_time": "0:59:18", "remaining_time": "1:04:26"}
{"current_steps": 267, "total_steps": 555, "loss": 0.0125, "lr": 6.200104639843985e-05, "epoch": 2.4017957351290686, "percentage": 48.11, "elapsed_time": "0:59:31", "remaining_time": "1:04:12"}
{"current_steps": 268, "total_steps": 555, "loss": 0.0117, "lr": 6.169522373090412e-05, "epoch": 2.410774410774411, "percentage": 48.29, "elapsed_time": "0:59:46", "remaining_time": "1:04:00"}
{"current_steps": 269, "total_steps": 555, "loss": 0.012, "lr": 6.138893750358212e-05, "epoch": 2.419753086419753, "percentage": 48.47, "elapsed_time": "0:59:57", "remaining_time": "1:03:45"}
{"current_steps": 270, "total_steps": 555, "loss": 0.0105, "lr": 6.108219985664161e-05, "epoch": 2.4287317620650954, "percentage": 48.65, "elapsed_time": "1:00:08", "remaining_time": "1:03:29"}
{"current_steps": 270, "total_steps": 555, "eval_loss": 0.013788803480565548, "epoch": 2.4287317620650954, "percentage": 48.65, "elapsed_time": "1:00:15", "remaining_time": "1:03:36"}
{"current_steps": 271, "total_steps": 555, "loss": 0.0137, "lr": 6.0775022948143115e-05, "epoch": 2.4377104377104377, "percentage": 48.83, "elapsed_time": "1:00:26", "remaining_time": "1:03:20"}
{"current_steps": 272, "total_steps": 555, "loss": 0.0117, "lr": 6.046741895355802e-05, "epoch": 2.44668911335578, "percentage": 49.01, "elapsed_time": "1:00:38", "remaining_time": "1:03:05"}
{"current_steps": 273, "total_steps": 555, "loss": 0.0108, "lr": 6.015940006528602e-05, "epoch": 2.4556677890011223, "percentage": 49.19, "elapsed_time": "1:00:49", "remaining_time": "1:02:49"}
{"current_steps": 274, "total_steps": 555, "loss": 0.011, "lr": 5.9850978492171794e-05, "epoch": 2.4646464646464645, "percentage": 49.37, "elapsed_time": "1:01:00", "remaining_time": "1:02:33"}
{"current_steps": 275, "total_steps": 555, "loss": 0.0129, "lr": 5.954216645902109e-05, "epoch": 2.473625140291807, "percentage": 49.55, "elapsed_time": "1:01:11", "remaining_time": "1:02:18"}
{"current_steps": 275, "total_steps": 555, "eval_loss": 0.01363787055015564, "epoch": 2.473625140291807, "percentage": 49.55, "elapsed_time": "1:01:17", "remaining_time": "1:02:24"}
{"current_steps": 276, "total_steps": 555, "loss": 0.0107, "lr": 5.923297620611623e-05, "epoch": 2.482603815937149, "percentage": 49.73, "elapsed_time": "1:01:29", "remaining_time": "1:02:09"}
{"current_steps": 277, "total_steps": 555, "loss": 0.0137, "lr": 5.892341998873089e-05, "epoch": 2.4915824915824913, "percentage": 49.91, "elapsed_time": "1:01:41", "remaining_time": "1:01:55"}
{"current_steps": 278, "total_steps": 555, "loss": 0.0127, "lr": 5.861351007664434e-05, "epoch": 2.500561167227834, "percentage": 50.09, "elapsed_time": "1:01:52", "remaining_time": "1:01:38"}
{"current_steps": 279, "total_steps": 555, "loss": 0.0127, "lr": 5.83032587536552e-05, "epoch": 2.5095398428731763, "percentage": 50.27, "elapsed_time": "1:02:05", "remaining_time": "1:01:25"}
{"current_steps": 280, "total_steps": 555, "loss": 0.014, "lr": 5.799267831709442e-05, "epoch": 2.5185185185185186, "percentage": 50.45, "elapsed_time": "1:02:18", "remaining_time": "1:01:11"}
{"current_steps": 280, "total_steps": 555, "eval_loss": 0.013543778099119663, "epoch": 2.5185185185185186, "percentage": 50.45, "elapsed_time": "1:02:24", "remaining_time": "1:01:17"}
{"current_steps": 281, "total_steps": 555, "loss": 0.0096, "lr": 5.7681781077337905e-05, "epoch": 2.527497194163861, "percentage": 50.63, "elapsed_time": "1:02:36", "remaining_time": "1:01:02"}
{"current_steps": 282, "total_steps": 555, "loss": 0.0092, "lr": 5.737057935731868e-05, "epoch": 2.536475869809203, "percentage": 50.81, "elapsed_time": "1:02:47", "remaining_time": "1:00:47"}
{"current_steps": 283, "total_steps": 555, "loss": 0.0126, "lr": 5.705908549203823e-05, "epoch": 2.5454545454545454, "percentage": 50.99, "elapsed_time": "1:02:58", "remaining_time": "1:00:31"}
{"current_steps": 284, "total_steps": 555, "loss": 0.0122, "lr": 5.674731182807781e-05, "epoch": 2.5544332210998877, "percentage": 51.17, "elapsed_time": "1:03:10", "remaining_time": "1:00:16"}
{"current_steps": 285, "total_steps": 555, "loss": 0.0124, "lr": 5.643527072310891e-05, "epoch": 2.56341189674523, "percentage": 51.35, "elapsed_time": "1:03:22", "remaining_time": "1:00:02"}
{"current_steps": 285, "total_steps": 555, "eval_loss": 0.013581929728388786, "epoch": 2.56341189674523, "percentage": 51.35, "elapsed_time": "1:03:28", "remaining_time": "1:00:08"}
{"current_steps": 286, "total_steps": 555, "loss": 0.0134, "lr": 5.612297454540352e-05, "epoch": 2.5723905723905722, "percentage": 51.53, "elapsed_time": "1:03:39", "remaining_time": "0:59:52"}
{"current_steps": 287, "total_steps": 555, "loss": 0.0105, "lr": 5.581043567334383e-05, "epoch": 2.581369248035915, "percentage": 51.71, "elapsed_time": "1:03:53", "remaining_time": "0:59:39"}
{"current_steps": 288, "total_steps": 555, "loss": 0.0116, "lr": 5.5497666494931654e-05, "epoch": 2.590347923681257, "percentage": 51.89, "elapsed_time": "1:04:04", "remaining_time": "0:59:24"}
{"current_steps": 289, "total_steps": 555, "loss": 0.0127, "lr": 5.518467940729739e-05, "epoch": 2.5993265993265995, "percentage": 52.07, "elapsed_time": "1:04:15", "remaining_time": "0:59:09"}
{"current_steps": 290, "total_steps": 555, "loss": 0.0128, "lr": 5.487148681620862e-05, "epoch": 2.6083052749719418, "percentage": 52.25, "elapsed_time": "1:04:25", "remaining_time": "0:58:52"}
{"current_steps": 290, "total_steps": 555, "eval_loss": 0.01327795721590519, "epoch": 2.6083052749719418, "percentage": 52.25, "elapsed_time": "1:04:32", "remaining_time": "0:58:58"}
{"current_steps": 291, "total_steps": 555, "loss": 0.0129, "lr": 5.455810113557839e-05, "epoch": 2.617283950617284, "percentage": 52.43, "elapsed_time": "1:04:46", "remaining_time": "0:58:45"}
{"current_steps": 292, "total_steps": 555, "loss": 0.0113, "lr": 5.4244534786973214e-05, "epoch": 2.6262626262626263, "percentage": 52.61, "elapsed_time": "1:04:57", "remaining_time": "0:58:30"}
{"current_steps": 293, "total_steps": 555, "loss": 0.0132, "lr": 5.3930800199120616e-05, "epoch": 2.6352413019079686, "percentage": 52.79, "elapsed_time": "1:05:09", "remaining_time": "0:58:15"}
{"current_steps": 294, "total_steps": 555, "loss": 0.0111, "lr": 5.361690980741663e-05, "epoch": 2.644219977553311, "percentage": 52.97, "elapsed_time": "1:05:20", "remaining_time": "0:58:00"}
{"current_steps": 295, "total_steps": 555, "loss": 0.0106, "lr": 5.330287605343279e-05, "epoch": 2.653198653198653, "percentage": 53.15, "elapsed_time": "1:05:32", "remaining_time": "0:57:45"}
{"current_steps": 295, "total_steps": 555, "eval_loss": 0.01291807834059, "epoch": 2.653198653198653, "percentage": 53.15, "elapsed_time": "1:05:38", "remaining_time": "0:57:51"}
{"current_steps": 296, "total_steps": 555, "loss": 0.0127, "lr": 5.298871138442307e-05, "epoch": 2.6621773288439954, "percentage": 53.33, "elapsed_time": "1:05:49", "remaining_time": "0:57:36"}
{"current_steps": 297, "total_steps": 555, "loss": 0.0123, "lr": 5.267442825283048e-05, "epoch": 2.6711560044893377, "percentage": 53.51, "elapsed_time": "1:06:00", "remaining_time": "0:57:20"}
{"current_steps": 298, "total_steps": 555, "loss": 0.0138, "lr": 5.236003911579345e-05, "epoch": 2.68013468013468, "percentage": 53.69, "elapsed_time": "1:06:12", "remaining_time": "0:57:06"}
{"current_steps": 299, "total_steps": 555, "loss": 0.011, "lr": 5.204555643465215e-05, "epoch": 2.689113355780022, "percentage": 53.87, "elapsed_time": "1:06:24", "remaining_time": "0:56:51"}
{"current_steps": 300, "total_steps": 555, "loss": 0.0099, "lr": 5.173099267445451e-05, "epoch": 2.698092031425365, "percentage": 54.05, "elapsed_time": "1:06:36", "remaining_time": "0:56:37"}
{"current_steps": 300, "total_steps": 555, "eval_loss": 0.012918239459395409, "epoch": 2.698092031425365, "percentage": 54.05, "elapsed_time": "1:06:43", "remaining_time": "0:56:42"}
{"current_steps": 301, "total_steps": 555, "loss": 0.0116, "lr": 5.1416360303462206e-05, "epoch": 2.707070707070707, "percentage": 54.23, "elapsed_time": "1:07:02", "remaining_time": "0:56:34"}
{"current_steps": 302, "total_steps": 555, "loss": 0.0134, "lr": 5.110167179265636e-05, "epoch": 2.7160493827160495, "percentage": 54.41, "elapsed_time": "1:07:15", "remaining_time": "0:56:20"}
{"current_steps": 303, "total_steps": 555, "loss": 0.0109, "lr": 5.078693961524329e-05, "epoch": 2.7250280583613917, "percentage": 54.59, "elapsed_time": "1:07:26", "remaining_time": "0:56:05"}
{"current_steps": 304, "total_steps": 555, "loss": 0.0121, "lr": 5.0472176246160184e-05, "epoch": 2.734006734006734, "percentage": 54.77, "elapsed_time": "1:07:39", "remaining_time": "0:55:51"}
{"current_steps": 305, "total_steps": 555, "loss": 0.0111, "lr": 5.01573941615805e-05, "epoch": 2.7429854096520763, "percentage": 54.95, "elapsed_time": "1:07:49", "remaining_time": "0:55:35"}
{"current_steps": 305, "total_steps": 555, "eval_loss": 0.012916718609631062, "epoch": 2.7429854096520763, "percentage": 54.95, "elapsed_time": "1:07:55", "remaining_time": "0:55:40"}
{"current_steps": 306, "total_steps": 555, "loss": 0.0115, "lr": 4.984260583841953e-05, "epoch": 2.7519640852974185, "percentage": 55.14, "elapsed_time": "1:08:06", "remaining_time": "0:55:25"}
{"current_steps": 307, "total_steps": 555, "loss": 0.0135, "lr": 4.9527823753839834e-05, "epoch": 2.760942760942761, "percentage": 55.32, "elapsed_time": "1:08:20", "remaining_time": "0:55:12"}
{"current_steps": 308, "total_steps": 555, "loss": 0.009, "lr": 4.9213060384756716e-05, "epoch": 2.7699214365881035, "percentage": 55.5, "elapsed_time": "1:08:31", "remaining_time": "0:54:57"}
{"current_steps": 309, "total_steps": 555, "loss": 0.0123, "lr": 4.8898328207343666e-05, "epoch": 2.778900112233446, "percentage": 55.68, "elapsed_time": "1:08:43", "remaining_time": "0:54:42"}
{"current_steps": 310, "total_steps": 555, "loss": 0.0129, "lr": 4.858363969653781e-05, "epoch": 2.787878787878788, "percentage": 55.86, "elapsed_time": "1:08:55", "remaining_time": "0:54:28"}
{"current_steps": 310, "total_steps": 555, "eval_loss": 0.012887900695204735, "epoch": 2.787878787878788, "percentage": 55.86, "elapsed_time": "1:09:01", "remaining_time": "0:54:33"}
{"current_steps": 311, "total_steps": 555, "loss": 0.0127, "lr": 4.8269007325545506e-05, "epoch": 2.7968574635241303, "percentage": 56.04, "elapsed_time": "1:09:13", "remaining_time": "0:54:18"}
{"current_steps": 312, "total_steps": 555, "loss": 0.0104, "lr": 4.7954443565347865e-05, "epoch": 2.8058361391694726, "percentage": 56.22, "elapsed_time": "1:09:24", "remaining_time": "0:54:03"}
{"current_steps": 313, "total_steps": 555, "loss": 0.0132, "lr": 4.7639960884206576e-05, "epoch": 2.814814814814815, "percentage": 56.4, "elapsed_time": "1:09:37", "remaining_time": "0:53:49"}
{"current_steps": 314, "total_steps": 555, "loss": 0.0106, "lr": 4.7325571747169545e-05, "epoch": 2.823793490460157, "percentage": 56.58, "elapsed_time": "1:09:48", "remaining_time": "0:53:34"}
{"current_steps": 315, "total_steps": 555, "loss": 0.0088, "lr": 4.7011288615576934e-05, "epoch": 2.8327721661054994, "percentage": 56.76, "elapsed_time": "1:10:01", "remaining_time": "0:53:21"}
{"current_steps": 315, "total_steps": 555, "eval_loss": 0.012852279469370842, "epoch": 2.8327721661054994, "percentage": 56.76, "elapsed_time": "1:10:07", "remaining_time": "0:53:25"}
{"current_steps": 316, "total_steps": 555, "loss": 0.0159, "lr": 4.6697123946567227e-05, "epoch": 2.8417508417508417, "percentage": 56.94, "elapsed_time": "1:10:18", "remaining_time": "0:53:10"}
{"current_steps": 317, "total_steps": 555, "loss": 0.0098, "lr": 4.63830901925834e-05, "epoch": 2.850729517396184, "percentage": 57.12, "elapsed_time": "1:10:31", "remaining_time": "0:52:56"}
{"current_steps": 318, "total_steps": 555, "loss": 0.011, "lr": 4.60691998008794e-05, "epoch": 2.8597081930415262, "percentage": 57.3, "elapsed_time": "1:10:43", "remaining_time": "0:52:42"}
{"current_steps": 319, "total_steps": 555, "loss": 0.0115, "lr": 4.575546521302681e-05, "epoch": 2.8686868686868685, "percentage": 57.48, "elapsed_time": "1:10:54", "remaining_time": "0:52:27"}
{"current_steps": 320, "total_steps": 555, "loss": 0.0092, "lr": 4.544189886442162e-05, "epoch": 2.877665544332211, "percentage": 57.66, "elapsed_time": "1:11:04", "remaining_time": "0:52:11"}
{"current_steps": 320, "total_steps": 555, "eval_loss": 0.012964904308319092, "epoch": 2.877665544332211, "percentage": 57.66, "elapsed_time": "1:11:10", "remaining_time": "0:52:16"}
{"current_steps": 321, "total_steps": 555, "loss": 0.0141, "lr": 4.5128513183791386e-05, "epoch": 2.886644219977553, "percentage": 57.84, "elapsed_time": "1:11:23", "remaining_time": "0:52:02"}
{"current_steps": 322, "total_steps": 555, "loss": 0.0115, "lr": 4.481532059270262e-05, "epoch": 2.8956228956228958, "percentage": 58.02, "elapsed_time": "1:11:36", "remaining_time": "0:51:48"}
{"current_steps": 323, "total_steps": 555, "loss": 0.0135, "lr": 4.450233350506836e-05, "epoch": 2.904601571268238, "percentage": 58.2, "elapsed_time": "1:11:47", "remaining_time": "0:51:33"}
{"current_steps": 324, "total_steps": 555, "loss": 0.0108, "lr": 4.418956432665618e-05, "epoch": 2.9135802469135803, "percentage": 58.38, "elapsed_time": "1:11:58", "remaining_time": "0:51:18"}
{"current_steps": 325, "total_steps": 555, "loss": 0.0086, "lr": 4.387702545459649e-05, "epoch": 2.9225589225589226, "percentage": 58.56, "elapsed_time": "1:12:08", "remaining_time": "0:51:03"}
{"current_steps": 325, "total_steps": 555, "eval_loss": 0.012879169546067715, "epoch": 2.9225589225589226, "percentage": 58.56, "elapsed_time": "1:12:14", "remaining_time": "0:51:07"}
{"current_steps": 326, "total_steps": 555, "loss": 0.0112, "lr": 4.356472927689109e-05, "epoch": 2.931537598204265, "percentage": 58.74, "elapsed_time": "1:12:25", "remaining_time": "0:50:52"}
{"current_steps": 327, "total_steps": 555, "loss": 0.0129, "lr": 4.32526881719222e-05, "epoch": 2.940516273849607, "percentage": 58.92, "elapsed_time": "1:12:37", "remaining_time": "0:50:38"}
{"current_steps": 328, "total_steps": 555, "loss": 0.0104, "lr": 4.2940914507961775e-05, "epoch": 2.9494949494949494, "percentage": 59.1, "elapsed_time": "1:12:48", "remaining_time": "0:50:23"}
{"current_steps": 329, "total_steps": 555, "loss": 0.0123, "lr": 4.262942064268134e-05, "epoch": 2.9584736251402917, "percentage": 59.28, "elapsed_time": "1:12:59", "remaining_time": "0:50:08"}
{"current_steps": 330, "total_steps": 555, "loss": 0.0132, "lr": 4.23182189226621e-05, "epoch": 2.9674523007856344, "percentage": 59.46, "elapsed_time": "1:13:12", "remaining_time": "0:49:55"}
{"current_steps": 330, "total_steps": 555, "eval_loss": 0.012610589154064655, "epoch": 2.9674523007856344, "percentage": 59.46, "elapsed_time": "1:13:19", "remaining_time": "0:49:59"}
{"current_steps": 331, "total_steps": 555, "loss": 0.0124, "lr": 4.20073216829056e-05, "epoch": 2.9764309764309766, "percentage": 59.64, "elapsed_time": "1:13:30", "remaining_time": "0:49:44"}
{"current_steps": 332, "total_steps": 555, "loss": 0.009, "lr": 4.169674124634481e-05, "epoch": 2.985409652076319, "percentage": 59.82, "elapsed_time": "1:13:42", "remaining_time": "0:49:30"}
{"current_steps": 333, "total_steps": 555, "loss": 0.0131, "lr": 4.138648992335566e-05, "epoch": 2.994388327721661, "percentage": 60.0, "elapsed_time": "1:13:54", "remaining_time": "0:49:16"}
{"current_steps": 334, "total_steps": 555, "loss": 0.0173, "lr": 4.107658001126913e-05, "epoch": 3.005611672278339, "percentage": 60.18, "elapsed_time": "1:14:04", "remaining_time": "0:49:01"}
{"current_steps": 335, "total_steps": 555, "loss": 0.0126, "lr": 4.0767023793883785e-05, "epoch": 3.014590347923681, "percentage": 60.36, "elapsed_time": "1:14:15", "remaining_time": "0:48:46"}
{"current_steps": 335, "total_steps": 555, "eval_loss": 0.012994157150387764, "epoch": 3.014590347923681, "percentage": 60.36, "elapsed_time": "1:14:22", "remaining_time": "0:48:50"}
{"current_steps": 336, "total_steps": 555, "loss": 0.0092, "lr": 4.045783354097893e-05, "epoch": 3.0235690235690234, "percentage": 60.54, "elapsed_time": "1:14:35", "remaining_time": "0:48:36"}
{"current_steps": 337, "total_steps": 555, "loss": 0.0131, "lr": 4.0149021507828224e-05, "epoch": 3.032547699214366, "percentage": 60.72, "elapsed_time": "1:14:46", "remaining_time": "0:48:22"}
{"current_steps": 338, "total_steps": 555, "loss": 0.0086, "lr": 3.984059993471399e-05, "epoch": 3.0415263748597083, "percentage": 60.9, "elapsed_time": "1:14:58", "remaining_time": "0:48:07"}
{"current_steps": 339, "total_steps": 555, "loss": 0.0104, "lr": 3.9532581046442e-05, "epoch": 3.0505050505050506, "percentage": 61.08, "elapsed_time": "1:15:11", "remaining_time": "0:47:54"}
{"current_steps": 340, "total_steps": 555, "loss": 0.0117, "lr": 3.9224977051856904e-05, "epoch": 3.059483726150393, "percentage": 61.26, "elapsed_time": "1:15:24", "remaining_time": "0:47:40"}
{"current_steps": 340, "total_steps": 555, "eval_loss": 0.013331728056073189, "epoch": 3.059483726150393, "percentage": 61.26, "elapsed_time": "1:15:30", "remaining_time": "0:47:44"}
{"current_steps": 341, "total_steps": 555, "loss": 0.0077, "lr": 3.8917800143358404e-05, "epoch": 3.068462401795735, "percentage": 61.44, "elapsed_time": "1:15:40", "remaining_time": "0:47:29"}
{"current_steps": 342, "total_steps": 555, "loss": 0.0097, "lr": 3.861106249641789e-05, "epoch": 3.0774410774410774, "percentage": 61.62, "elapsed_time": "1:15:52", "remaining_time": "0:47:15"}
{"current_steps": 343, "total_steps": 555, "loss": 0.0125, "lr": 3.830477626909589e-05, "epoch": 3.0864197530864197, "percentage": 61.8, "elapsed_time": "1:16:03", "remaining_time": "0:47:00"}
{"current_steps": 344, "total_steps": 555, "loss": 0.0111, "lr": 3.7998953601560175e-05, "epoch": 3.095398428731762, "percentage": 61.98, "elapsed_time": "1:16:14", "remaining_time": "0:46:45"}
{"current_steps": 345, "total_steps": 555, "loss": 0.0102, "lr": 3.769360661560453e-05, "epoch": 3.1043771043771042, "percentage": 62.16, "elapsed_time": "1:16:26", "remaining_time": "0:46:31"}
{"current_steps": 345, "total_steps": 555, "eval_loss": 0.013248049654066563, "epoch": 3.1043771043771042, "percentage": 62.16, "elapsed_time": "1:16:32", "remaining_time": "0:46:35"}
{"current_steps": 346, "total_steps": 555, "loss": 0.0088, "lr": 3.73887474141683e-05, "epoch": 3.1133557800224465, "percentage": 62.34, "elapsed_time": "1:16:43", "remaining_time": "0:46:20"}
{"current_steps": 347, "total_steps": 555, "loss": 0.01, "lr": 3.708438808085668e-05, "epoch": 3.122334455667789, "percentage": 62.52, "elapsed_time": "1:16:56", "remaining_time": "0:46:07"}
{"current_steps": 348, "total_steps": 555, "loss": 0.0091, "lr": 3.6780540679461784e-05, "epoch": 3.1313131313131315, "percentage": 62.7, "elapsed_time": "1:17:08", "remaining_time": "0:45:52"}
{"current_steps": 349, "total_steps": 555, "loss": 0.0118, "lr": 3.64772172534844e-05, "epoch": 3.1402918069584738, "percentage": 62.88, "elapsed_time": "1:17:18", "remaining_time": "0:45:38"}
{"current_steps": 350, "total_steps": 555, "loss": 0.0074, "lr": 3.6174429825656685e-05, "epoch": 3.149270482603816, "percentage": 63.06, "elapsed_time": "1:17:29", "remaining_time": "0:45:23"}
{"current_steps": 350, "total_steps": 555, "eval_loss": 0.013174821622669697, "epoch": 3.149270482603816, "percentage": 63.06, "elapsed_time": "1:17:35", "remaining_time": "0:45:26"}
{"current_steps": 351, "total_steps": 555, "loss": 0.0124, "lr": 3.587219039746564e-05, "epoch": 3.1582491582491583, "percentage": 63.24, "elapsed_time": "1:17:53", "remaining_time": "0:45:16"}
{"current_steps": 352, "total_steps": 555, "loss": 0.0122, "lr": 3.557051094867735e-05, "epoch": 3.1672278338945006, "percentage": 63.42, "elapsed_time": "1:18:05", "remaining_time": "0:45:02"}
{"current_steps": 353, "total_steps": 555, "loss": 0.0089, "lr": 3.5269403436862175e-05, "epoch": 3.176206509539843, "percentage": 63.6, "elapsed_time": "1:18:15", "remaining_time": "0:44:47"}
{"current_steps": 354, "total_steps": 555, "loss": 0.0111, "lr": 3.496887979692084e-05, "epoch": 3.185185185185185, "percentage": 63.78, "elapsed_time": "1:18:27", "remaining_time": "0:44:32"}
{"current_steps": 355, "total_steps": 555, "loss": 0.0105, "lr": 3.466895194061128e-05, "epoch": 3.1941638608305274, "percentage": 63.96, "elapsed_time": "1:18:39", "remaining_time": "0:44:18"}
{"current_steps": 355, "total_steps": 555, "eval_loss": 0.01294049434363842, "epoch": 3.1941638608305274, "percentage": 63.96, "elapsed_time": "1:18:45", "remaining_time": "0:44:22"}
{"current_steps": 356, "total_steps": 555, "loss": 0.0096, "lr": 3.436963175607656e-05, "epoch": 3.2031425364758697, "percentage": 64.14, "elapsed_time": "1:18:57", "remaining_time": "0:44:08"}
{"current_steps": 357, "total_steps": 555, "loss": 0.0092, "lr": 3.4070931107373675e-05, "epoch": 3.212121212121212, "percentage": 64.32, "elapsed_time": "1:19:07", "remaining_time": "0:43:53"}
{"current_steps": 358, "total_steps": 555, "loss": 0.011, "lr": 3.377286183400328e-05, "epoch": 3.221099887766554, "percentage": 64.5, "elapsed_time": "1:19:18", "remaining_time": "0:43:38"}
{"current_steps": 359, "total_steps": 555, "loss": 0.0101, "lr": 3.3475435750440356e-05, "epoch": 3.230078563411897, "percentage": 64.68, "elapsed_time": "1:19:31", "remaining_time": "0:43:25"}
{"current_steps": 360, "total_steps": 555, "loss": 0.0117, "lr": 3.3178664645666066e-05, "epoch": 3.239057239057239, "percentage": 64.86, "elapsed_time": "1:19:44", "remaining_time": "0:43:11"}
{"current_steps": 360, "total_steps": 555, "eval_loss": 0.012872601859271526, "epoch": 3.239057239057239, "percentage": 64.86, "elapsed_time": "1:19:51", "remaining_time": "0:43:15"}
{"current_steps": 361, "total_steps": 555, "loss": 0.0096, "lr": 3.2882560282700336e-05, "epoch": 3.2480359147025815, "percentage": 65.05, "elapsed_time": "1:20:03", "remaining_time": "0:43:01"}
{"current_steps": 362, "total_steps": 555, "loss": 0.0105, "lr": 3.258713439813566e-05, "epoch": 3.2570145903479237, "percentage": 65.23, "elapsed_time": "1:20:16", "remaining_time": "0:42:48"}
{"current_steps": 363, "total_steps": 555, "loss": 0.0103, "lr": 3.229239870167191e-05, "epoch": 3.265993265993266, "percentage": 65.41, "elapsed_time": "1:20:30", "remaining_time": "0:42:34"}
{"current_steps": 364, "total_steps": 555, "loss": 0.0109, "lr": 3.199836487565222e-05, "epoch": 3.2749719416386083, "percentage": 65.59, "elapsed_time": "1:20:42", "remaining_time": "0:42:20"}
{"current_steps": 365, "total_steps": 555, "loss": 0.0107, "lr": 3.170504457459989e-05, "epoch": 3.2839506172839505, "percentage": 65.77, "elapsed_time": "1:20:55", "remaining_time": "0:42:07"}
{"current_steps": 365, "total_steps": 555, "eval_loss": 0.012733125127851963, "epoch": 3.2839506172839505, "percentage": 65.77, "elapsed_time": "1:21:01", "remaining_time": "0:42:10"}
{"current_steps": 366, "total_steps": 555, "loss": 0.0097, "lr": 3.1412449424756474e-05, "epoch": 3.292929292929293, "percentage": 65.95, "elapsed_time": "1:21:12", "remaining_time": "0:41:56"}
{"current_steps": 367, "total_steps": 555, "loss": 0.0106, "lr": 3.112059102362093e-05, "epoch": 3.301907968574635, "percentage": 66.13, "elapsed_time": "1:21:24", "remaining_time": "0:41:42"}
{"current_steps": 368, "total_steps": 555, "loss": 0.0094, "lr": 3.082948093948997e-05, "epoch": 3.3108866442199774, "percentage": 66.31, "elapsed_time": "1:21:37", "remaining_time": "0:41:28"}
{"current_steps": 369, "total_steps": 555, "loss": 0.0096, "lr": 3.053913071099947e-05, "epoch": 3.31986531986532, "percentage": 66.49, "elapsed_time": "1:21:49", "remaining_time": "0:41:14"}
{"current_steps": 370, "total_steps": 555, "loss": 0.0098, "lr": 3.0249551846667207e-05, "epoch": 3.3288439955106623, "percentage": 66.67, "elapsed_time": "1:22:00", "remaining_time": "0:41:00"}
{"current_steps": 370, "total_steps": 555, "eval_loss": 0.012761359103024006, "epoch": 3.3288439955106623, "percentage": 66.67, "elapsed_time": "1:22:07", "remaining_time": "0:41:03"}
{"current_steps": 371, "total_steps": 555, "loss": 0.0115, "lr": 2.996075582443658e-05, "epoch": 3.3378226711560046, "percentage": 66.85, "elapsed_time": "1:22:17", "remaining_time": "0:40:48"}
{"current_steps": 372, "total_steps": 555, "loss": 0.0086, "lr": 2.9672754091221805e-05, "epoch": 3.346801346801347, "percentage": 67.03, "elapsed_time": "1:22:28", "remaining_time": "0:40:34"}
{"current_steps": 373, "total_steps": 555, "loss": 0.0106, "lr": 2.938555806245406e-05, "epoch": 3.355780022446689, "percentage": 67.21, "elapsed_time": "1:22:41", "remaining_time": "0:40:20"}
{"current_steps": 374, "total_steps": 555, "loss": 0.0115, "lr": 2.9099179121629117e-05, "epoch": 3.3647586980920314, "percentage": 67.39, "elapsed_time": "1:22:53", "remaining_time": "0:40:07"}
{"current_steps": 375, "total_steps": 555, "loss": 0.0092, "lr": 2.881362861985606e-05, "epoch": 3.3737373737373737, "percentage": 67.57, "elapsed_time": "1:23:07", "remaining_time": "0:39:53"}
{"current_steps": 375, "total_steps": 555, "eval_loss": 0.012749603018164635, "epoch": 3.3737373737373737, "percentage": 67.57, "elapsed_time": "1:23:13", "remaining_time": "0:39:56"}
{"current_steps": 376, "total_steps": 555, "loss": 0.0094, "lr": 2.8528917875407433e-05, "epoch": 3.382716049382716, "percentage": 67.75, "elapsed_time": "1:23:24", "remaining_time": "0:39:42"}
{"current_steps": 377, "total_steps": 555, "loss": 0.0102, "lr": 2.8245058173270622e-05, "epoch": 3.3916947250280582, "percentage": 67.93, "elapsed_time": "1:23:37", "remaining_time": "0:39:28"}
{"current_steps": 378, "total_steps": 555, "loss": 0.0082, "lr": 2.796206076470044e-05, "epoch": 3.4006734006734005, "percentage": 68.11, "elapsed_time": "1:23:48", "remaining_time": "0:39:14"}
{"current_steps": 379, "total_steps": 555, "loss": 0.0095, "lr": 2.7679936866773315e-05, "epoch": 3.409652076318743, "percentage": 68.29, "elapsed_time": "1:24:00", "remaining_time": "0:39:00"}
{"current_steps": 380, "total_steps": 555, "loss": 0.0114, "lr": 2.739869766194263e-05, "epoch": 3.418630751964085, "percentage": 68.47, "elapsed_time": "1:24:13", "remaining_time": "0:38:47"}
{"current_steps": 380, "total_steps": 555, "eval_loss": 0.012560264207422733, "epoch": 3.418630751964085, "percentage": 68.47, "elapsed_time": "1:24:19", "remaining_time": "0:38:50"}
{"current_steps": 381, "total_steps": 555, "loss": 0.01, "lr": 2.7118354297595396e-05, "epoch": 3.4276094276094278, "percentage": 68.65, "elapsed_time": "1:24:32", "remaining_time": "0:38:36"}
{"current_steps": 382, "total_steps": 555, "loss": 0.0109, "lr": 2.683891788561055e-05, "epoch": 3.43658810325477, "percentage": 68.83, "elapsed_time": "1:24:44", "remaining_time": "0:38:22"}
{"current_steps": 383, "total_steps": 555, "loss": 0.0083, "lr": 2.6560399501918465e-05, "epoch": 3.4455667789001123, "percentage": 69.01, "elapsed_time": "1:24:55", "remaining_time": "0:38:08"}
{"current_steps": 384, "total_steps": 555, "loss": 0.0095, "lr": 2.6282810186061862e-05, "epoch": 3.4545454545454546, "percentage": 69.19, "elapsed_time": "1:25:07", "remaining_time": "0:37:54"}
{"current_steps": 385, "total_steps": 555, "loss": 0.0118, "lr": 2.600616094075835e-05, "epoch": 3.463524130190797, "percentage": 69.37, "elapsed_time": "1:25:17", "remaining_time": "0:37:39"}
{"current_steps": 385, "total_steps": 555, "eval_loss": 0.012463411316275597, "epoch": 3.463524130190797, "percentage": 69.37, "elapsed_time": "1:25:23", "remaining_time": "0:37:42"}
{"current_steps": 386, "total_steps": 555, "loss": 0.0106, "lr": 2.5730462731464273e-05, "epoch": 3.472502805836139, "percentage": 69.55, "elapsed_time": "1:25:35", "remaining_time": "0:37:28"}
{"current_steps": 387, "total_steps": 555, "loss": 0.0088, "lr": 2.5455726485940012e-05, "epoch": 3.4814814814814814, "percentage": 69.73, "elapsed_time": "1:25:46", "remaining_time": "0:37:14"}
{"current_steps": 388, "total_steps": 555, "loss": 0.0086, "lr": 2.5181963093816962e-05, "epoch": 3.4904601571268237, "percentage": 69.91, "elapsed_time": "1:25:59", "remaining_time": "0:37:00"}
{"current_steps": 389, "total_steps": 555, "loss": 0.0095, "lr": 2.4909183406165836e-05, "epoch": 3.499438832772166, "percentage": 70.09, "elapsed_time": "1:26:09", "remaining_time": "0:36:46"}
{"current_steps": 390, "total_steps": 555, "loss": 0.0108, "lr": 2.4637398235066527e-05, "epoch": 3.5084175084175087, "percentage": 70.27, "elapsed_time": "1:26:22", "remaining_time": "0:36:32"}
{"current_steps": 390, "total_steps": 555, "eval_loss": 0.012311117723584175, "epoch": 3.5084175084175087, "percentage": 70.27, "elapsed_time": "1:26:29", "remaining_time": "0:36:35"}
{"current_steps": 391, "total_steps": 555, "loss": 0.0128, "lr": 2.4366618353179644e-05, "epoch": 3.517396184062851, "percentage": 70.45, "elapsed_time": "1:26:41", "remaining_time": "0:36:21"}
{"current_steps": 392, "total_steps": 555, "loss": 0.0089, "lr": 2.4096854493319477e-05, "epoch": 3.526374859708193, "percentage": 70.63, "elapsed_time": "1:26:53", "remaining_time": "0:36:07"}
{"current_steps": 393, "total_steps": 555, "loss": 0.0088, "lr": 2.3828117348028528e-05, "epoch": 3.5353535353535355, "percentage": 70.81, "elapsed_time": "1:27:04", "remaining_time": "0:35:53"}
{"current_steps": 394, "total_steps": 555, "loss": 0.0096, "lr": 2.3560417569153796e-05, "epoch": 3.5443322109988777, "percentage": 70.99, "elapsed_time": "1:27:14", "remaining_time": "0:35:39"}
{"current_steps": 395, "total_steps": 555, "loss": 0.0092, "lr": 2.3293765767424537e-05, "epoch": 3.55331088664422, "percentage": 71.17, "elapsed_time": "1:27:25", "remaining_time": "0:35:24"}
{"current_steps": 395, "total_steps": 555, "eval_loss": 0.012257438153028488, "epoch": 3.55331088664422, "percentage": 71.17, "elapsed_time": "1:27:31", "remaining_time": "0:35:27"}
{"current_steps": 396, "total_steps": 555, "loss": 0.0101, "lr": 2.3028172512031604e-05, "epoch": 3.5622895622895623, "percentage": 71.35, "elapsed_time": "1:27:43", "remaining_time": "0:35:13"}
{"current_steps": 397, "total_steps": 555, "loss": 0.0103, "lr": 2.276364833020868e-05, "epoch": 3.5712682379349046, "percentage": 71.53, "elapsed_time": "1:27:55", "remaining_time": "0:34:59"}
{"current_steps": 398, "total_steps": 555, "loss": 0.0089, "lr": 2.2500203706814856e-05, "epoch": 3.580246913580247, "percentage": 71.71, "elapsed_time": "1:28:06", "remaining_time": "0:34:45"}
{"current_steps": 399, "total_steps": 555, "loss": 0.0097, "lr": 2.2237849083919142e-05, "epoch": 3.589225589225589, "percentage": 71.89, "elapsed_time": "1:28:19", "remaining_time": "0:34:31"}
{"current_steps": 400, "total_steps": 555, "loss": 0.0085, "lr": 2.1976594860386597e-05, "epoch": 3.5982042648709314, "percentage": 72.07, "elapsed_time": "1:28:29", "remaining_time": "0:34:17"}
{"current_steps": 400, "total_steps": 555, "eval_loss": 0.01230713166296482, "epoch": 3.5982042648709314, "percentage": 72.07, "elapsed_time": "1:28:36", "remaining_time": "0:34:20"}
{"current_steps": 401, "total_steps": 555, "loss": 0.01, "lr": 2.1716451391466008e-05, "epoch": 3.6071829405162736, "percentage": 72.25, "elapsed_time": "1:28:56", "remaining_time": "0:34:09"}
{"current_steps": 402, "total_steps": 555, "loss": 0.01, "lr": 2.1457428988379635e-05, "epoch": 3.616161616161616, "percentage": 72.43, "elapsed_time": "1:29:07", "remaining_time": "0:33:55"}
{"current_steps": 403, "total_steps": 555, "loss": 0.01, "lr": 2.1199537917914386e-05, "epoch": 3.6251402918069586, "percentage": 72.61, "elapsed_time": "1:29:19", "remaining_time": "0:33:41"}
{"current_steps": 404, "total_steps": 555, "loss": 0.0097, "lr": 2.0942788402014867e-05, "epoch": 3.634118967452301, "percentage": 72.79, "elapsed_time": "1:29:31", "remaining_time": "0:33:27"}
{"current_steps": 405, "total_steps": 555, "loss": 0.0088, "lr": 2.068719061737831e-05, "epoch": 3.643097643097643, "percentage": 72.97, "elapsed_time": "1:29:42", "remaining_time": "0:33:13"}
{"current_steps": 405, "total_steps": 555, "eval_loss": 0.012639960274100304, "epoch": 3.643097643097643, "percentage": 72.97, "elapsed_time": "1:29:49", "remaining_time": "0:33:15"}
{"current_steps": 406, "total_steps": 555, "loss": 0.0112, "lr": 2.0432754695051136e-05, "epoch": 3.6520763187429854, "percentage": 73.15, "elapsed_time": "1:30:00", "remaining_time": "0:33:02"}
{"current_steps": 407, "total_steps": 555, "loss": 0.0104, "lr": 2.0179490720027372e-05, "epoch": 3.6610549943883277, "percentage": 73.33, "elapsed_time": "1:30:11", "remaining_time": "0:32:47"}
{"current_steps": 408, "total_steps": 555, "loss": 0.0084, "lr": 1.992740873084899e-05, "epoch": 3.67003367003367, "percentage": 73.51, "elapsed_time": "1:30:23", "remaining_time": "0:32:34"}
{"current_steps": 409, "total_steps": 555, "loss": 0.0093, "lr": 1.9676518719207977e-05, "epoch": 3.6790123456790123, "percentage": 73.69, "elapsed_time": "1:30:36", "remaining_time": "0:32:20"}
{"current_steps": 410, "total_steps": 555, "loss": 0.0095, "lr": 1.9426830629550242e-05, "epoch": 3.6879910213243545, "percentage": 73.87, "elapsed_time": "1:30:48", "remaining_time": "0:32:07"}
{"current_steps": 410, "total_steps": 555, "eval_loss": 0.012408388778567314, "epoch": 3.6879910213243545, "percentage": 73.87, "elapsed_time": "1:30:55", "remaining_time": "0:32:09"}
{"current_steps": 411, "total_steps": 555, "loss": 0.0101, "lr": 1.917835435868155e-05, "epoch": 3.6969696969696972, "percentage": 74.05, "elapsed_time": "1:31:07", "remaining_time": "0:31:55"}
{"current_steps": 412, "total_steps": 555, "loss": 0.0085, "lr": 1.8931099755375203e-05, "epoch": 3.7059483726150395, "percentage": 74.23, "elapsed_time": "1:31:19", "remaining_time": "0:31:41"}
{"current_steps": 413, "total_steps": 555, "loss": 0.0102, "lr": 1.8685076619981608e-05, "epoch": 3.714927048260382, "percentage": 74.41, "elapsed_time": "1:31:31", "remaining_time": "0:31:28"}
{"current_steps": 414, "total_steps": 555, "loss": 0.0097, "lr": 1.844029470403993e-05, "epoch": 3.723905723905724, "percentage": 74.59, "elapsed_time": "1:31:44", "remaining_time": "0:31:14"}
{"current_steps": 415, "total_steps": 555, "loss": 0.0072, "lr": 1.8196763709891524e-05, "epoch": 3.7328843995510663, "percentage": 74.77, "elapsed_time": "1:31:56", "remaining_time": "0:31:01"}
{"current_steps": 415, "total_steps": 555, "eval_loss": 0.012383312918245792, "epoch": 3.7328843995510663, "percentage": 74.77, "elapsed_time": "1:32:03", "remaining_time": "0:31:03"}
{"current_steps": 416, "total_steps": 555, "loss": 0.0103, "lr": 1.795449329029531e-05, "epoch": 3.7418630751964086, "percentage": 74.95, "elapsed_time": "1:32:14", "remaining_time": "0:30:49"}
{"current_steps": 417, "total_steps": 555, "loss": 0.0108, "lr": 1.7713493048045294e-05, "epoch": 3.750841750841751, "percentage": 75.14, "elapsed_time": "1:32:27", "remaining_time": "0:30:35"}
{"current_steps": 418, "total_steps": 555, "loss": 0.0102, "lr": 1.747377253558982e-05, "epoch": 3.759820426487093, "percentage": 75.32, "elapsed_time": "1:32:40", "remaining_time": "0:30:22"}
{"current_steps": 419, "total_steps": 555, "loss": 0.0103, "lr": 1.7235341254653005e-05, "epoch": 3.7687991021324354, "percentage": 75.5, "elapsed_time": "1:32:52", "remaining_time": "0:30:08"}
{"current_steps": 420, "total_steps": 555, "loss": 0.0105, "lr": 1.6998208655858137e-05, "epoch": 3.7777777777777777, "percentage": 75.68, "elapsed_time": "1:33:04", "remaining_time": "0:29:55"}
{"current_steps": 420, "total_steps": 555, "eval_loss": 0.012317318469285965, "epoch": 3.7777777777777777, "percentage": 75.68, "elapsed_time": "1:33:11", "remaining_time": "0:29:57"}
{"current_steps": 421, "total_steps": 555, "loss": 0.011, "lr": 1.6762384138353078e-05, "epoch": 3.78675645342312, "percentage": 75.86, "elapsed_time": "1:33:23", "remaining_time": "0:29:43"}
{"current_steps": 422, "total_steps": 555, "loss": 0.0082, "lr": 1.6527877049437622e-05, "epoch": 3.795735129068462, "percentage": 76.04, "elapsed_time": "1:33:34", "remaining_time": "0:29:29"}
{"current_steps": 423, "total_steps": 555, "loss": 0.0095, "lr": 1.6294696684193154e-05, "epoch": 3.8047138047138045, "percentage": 76.22, "elapsed_time": "1:33:45", "remaining_time": "0:29:15"}
{"current_steps": 424, "total_steps": 555, "loss": 0.0091, "lr": 1.6062852285114123e-05, "epoch": 3.8136924803591468, "percentage": 76.4, "elapsed_time": "1:33:57", "remaining_time": "0:29:01"}
{"current_steps": 425, "total_steps": 555, "loss": 0.0115, "lr": 1.583235304174167e-05, "epoch": 3.8226711560044895, "percentage": 76.58, "elapsed_time": "1:34:09", "remaining_time": "0:28:48"}
{"current_steps": 425, "total_steps": 555, "eval_loss": 0.01215137168765068, "epoch": 3.8226711560044895, "percentage": 76.58, "elapsed_time": "1:34:15", "remaining_time": "0:28:50"}
{"current_steps": 426, "total_steps": 555, "loss": 0.0082, "lr": 1.5603208090299498e-05, "epoch": 3.8316498316498318, "percentage": 76.76, "elapsed_time": "1:34:29", "remaining_time": "0:28:36"}
{"current_steps": 427, "total_steps": 555, "loss": 0.0098, "lr": 1.537542651333167e-05, "epoch": 3.840628507295174, "percentage": 76.94, "elapsed_time": "1:34:40", "remaining_time": "0:28:22"}
{"current_steps": 428, "total_steps": 555, "loss": 0.0105, "lr": 1.5149017339342574e-05, "epoch": 3.8496071829405163, "percentage": 77.12, "elapsed_time": "1:34:52", "remaining_time": "0:28:09"}
{"current_steps": 429, "total_steps": 555, "loss": 0.0108, "lr": 1.4923989542439159e-05, "epoch": 3.8585858585858586, "percentage": 77.3, "elapsed_time": "1:35:02", "remaining_time": "0:27:55"}
{"current_steps": 430, "total_steps": 555, "loss": 0.007, "lr": 1.4700352041975168e-05, "epoch": 3.867564534231201, "percentage": 77.48, "elapsed_time": "1:35:14", "remaining_time": "0:27:41"}
{"current_steps": 430, "total_steps": 555, "eval_loss": 0.012098519131541252, "epoch": 3.867564534231201, "percentage": 77.48, "elapsed_time": "1:35:21", "remaining_time": "0:27:43"}
{"current_steps": 431, "total_steps": 555, "loss": 0.01, "lr": 1.447811370219757e-05, "epoch": 3.876543209876543, "percentage": 77.66, "elapsed_time": "1:35:32", "remaining_time": "0:27:29"}
{"current_steps": 432, "total_steps": 555, "loss": 0.0082, "lr": 1.4257283331895315e-05, "epoch": 3.8855218855218854, "percentage": 77.84, "elapsed_time": "1:35:45", "remaining_time": "0:27:15"}
{"current_steps": 433, "total_steps": 555, "loss": 0.0117, "lr": 1.4037869684050115e-05, "epoch": 3.894500561167228, "percentage": 78.02, "elapsed_time": "1:35:55", "remaining_time": "0:27:01"}
{"current_steps": 434, "total_steps": 555, "loss": 0.0085, "lr": 1.3819881455489458e-05, "epoch": 3.9034792368125704, "percentage": 78.2, "elapsed_time": "1:36:10", "remaining_time": "0:26:48"}
{"current_steps": 435, "total_steps": 555, "loss": 0.0112, "lr": 1.3603327286542023e-05, "epoch": 3.9124579124579126, "percentage": 78.38, "elapsed_time": "1:36:21", "remaining_time": "0:26:35"}
{"current_steps": 435, "total_steps": 555, "eval_loss": 0.012134820222854614, "epoch": 3.9124579124579126, "percentage": 78.38, "elapsed_time": "1:36:28", "remaining_time": "0:26:36"}
{"current_steps": 436, "total_steps": 555, "loss": 0.0092, "lr": 1.33882157606951e-05, "epoch": 3.921436588103255, "percentage": 78.56, "elapsed_time": "1:36:39", "remaining_time": "0:26:22"}
{"current_steps": 437, "total_steps": 555, "loss": 0.0074, "lr": 1.317455540425439e-05, "epoch": 3.930415263748597, "percentage": 78.74, "elapsed_time": "1:36:50", "remaining_time": "0:26:09"}
{"current_steps": 438, "total_steps": 555, "loss": 0.0115, "lr": 1.2962354686006084e-05, "epoch": 3.9393939393939394, "percentage": 78.92, "elapsed_time": "1:37:01", "remaining_time": "0:25:55"}
{"current_steps": 439, "total_steps": 555, "loss": 0.0085, "lr": 1.2751622016881182e-05, "epoch": 3.9483726150392817, "percentage": 79.1, "elapsed_time": "1:37:13", "remaining_time": "0:25:41"}
{"current_steps": 440, "total_steps": 555, "loss": 0.0103, "lr": 1.2542365749622049e-05, "epoch": 3.957351290684624, "percentage": 79.28, "elapsed_time": "1:37:25", "remaining_time": "0:25:27"}
{"current_steps": 440, "total_steps": 555, "eval_loss": 0.012064680457115173, "epoch": 3.957351290684624, "percentage": 79.28, "elapsed_time": "1:37:31", "remaining_time": "0:25:29"}
{"current_steps": 441, "total_steps": 555, "loss": 0.0088, "lr": 1.2334594178451425e-05, "epoch": 3.9663299663299663, "percentage": 79.46, "elapsed_time": "1:37:42", "remaining_time": "0:25:15"}
{"current_steps": 442, "total_steps": 555, "loss": 0.0074, "lr": 1.2128315538743646e-05, "epoch": 3.9753086419753085, "percentage": 79.64, "elapsed_time": "1:37:54", "remaining_time": "0:25:01"}
{"current_steps": 443, "total_steps": 555, "loss": 0.0099, "lr": 1.1923538006698154e-05, "epoch": 3.984287317620651, "percentage": 79.82, "elapsed_time": "1:38:06", "remaining_time": "0:24:48"}
{"current_steps": 444, "total_steps": 555, "loss": 0.0089, "lr": 1.172026969901553e-05, "epoch": 3.993265993265993, "percentage": 80.0, "elapsed_time": "1:38:18", "remaining_time": "0:24:34"}
{"current_steps": 445, "total_steps": 555, "loss": 0.0162, "lr": 1.1518518672575701e-05, "epoch": 4.004489337822672, "percentage": 80.18, "elapsed_time": "1:38:31", "remaining_time": "0:24:21"}
{"current_steps": 445, "total_steps": 555, "eval_loss": 0.0121712451800704, "epoch": 4.004489337822672, "percentage": 80.18, "elapsed_time": "1:38:37", "remaining_time": "0:24:22"}
{"current_steps": 446, "total_steps": 555, "loss": 0.0092, "lr": 1.1318292924118584e-05, "epoch": 4.013468013468014, "percentage": 80.36, "elapsed_time": "1:38:47", "remaining_time": "0:24:08"}
{"current_steps": 447, "total_steps": 555, "loss": 0.0089, "lr": 1.1119600389927182e-05, "epoch": 4.022446689113356, "percentage": 80.54, "elapsed_time": "1:38:59", "remaining_time": "0:23:55"}
{"current_steps": 448, "total_steps": 555, "loss": 0.0076, "lr": 1.092244894551298e-05, "epoch": 4.031425364758698, "percentage": 80.72, "elapsed_time": "1:39:11", "remaining_time": "0:23:41"}
{"current_steps": 449, "total_steps": 555, "loss": 0.0091, "lr": 1.0726846405303754e-05, "epoch": 4.040404040404041, "percentage": 80.9, "elapsed_time": "1:39:23", "remaining_time": "0:23:27"}
{"current_steps": 450, "total_steps": 555, "loss": 0.0079, "lr": 1.0532800522333902e-05, "epoch": 4.049382716049383, "percentage": 81.08, "elapsed_time": "1:39:34", "remaining_time": "0:23:14"}
{"current_steps": 450, "total_steps": 555, "eval_loss": 0.0124581940472126, "epoch": 4.049382716049383, "percentage": 81.08, "elapsed_time": "1:39:40", "remaining_time": "0:23:15"}
{"current_steps": 451, "total_steps": 555, "loss": 0.0086, "lr": 1.0340318987937097e-05, "epoch": 4.058361391694725, "percentage": 81.26, "elapsed_time": "1:39:58", "remaining_time": "0:23:03"}
{"current_steps": 452, "total_steps": 555, "loss": 0.0099, "lr": 1.014940943144142e-05, "epoch": 4.0673400673400675, "percentage": 81.44, "elapsed_time": "1:40:11", "remaining_time": "0:22:49"}
{"current_steps": 453, "total_steps": 555, "loss": 0.0067, "lr": 9.960079419866985e-06, "epoch": 4.07631874298541, "percentage": 81.62, "elapsed_time": "1:40:22", "remaining_time": "0:22:36"}
{"current_steps": 454, "total_steps": 555, "loss": 0.0065, "lr": 9.772336457626014e-06, "epoch": 4.085297418630752, "percentage": 81.8, "elapsed_time": "1:40:33", "remaining_time": "0:22:22"}
{"current_steps": 455, "total_steps": 555, "loss": 0.0102, "lr": 9.586187986225325e-06, "epoch": 4.094276094276094, "percentage": 81.98, "elapsed_time": "1:40:45", "remaining_time": "0:22:08"}
{"current_steps": 455, "total_steps": 555, "eval_loss": 0.012580028735101223, "epoch": 4.094276094276094, "percentage": 81.98, "elapsed_time": "1:40:51", "remaining_time": "0:22:10"}
{"current_steps": 456, "total_steps": 555, "loss": 0.01, "lr": 9.401641383971477e-06, "epoch": 4.103254769921437, "percentage": 82.16, "elapsed_time": "1:41:02", "remaining_time": "0:21:56"}
{"current_steps": 457, "total_steps": 555, "loss": 0.0101, "lr": 9.218703965678204e-06, "epoch": 4.112233445566779, "percentage": 82.34, "elapsed_time": "1:41:15", "remaining_time": "0:21:42"}
{"current_steps": 458, "total_steps": 555, "loss": 0.0063, "lr": 9.03738298237658e-06, "epoch": 4.121212121212121, "percentage": 82.52, "elapsed_time": "1:41:27", "remaining_time": "0:21:29"}
{"current_steps": 459, "total_steps": 555, "loss": 0.0078, "lr": 8.857685621027568e-06, "epoch": 4.130190796857463, "percentage": 82.7, "elapsed_time": "1:41:38", "remaining_time": "0:21:15"}
{"current_steps": 460, "total_steps": 555, "loss": 0.0087, "lr": 8.67961900423711e-06, "epoch": 4.139169472502806, "percentage": 82.88, "elapsed_time": "1:41:50", "remaining_time": "0:21:01"}
{"current_steps": 460, "total_steps": 555, "eval_loss": 0.012577519752085209, "epoch": 4.139169472502806, "percentage": 82.88, "elapsed_time": "1:41:56", "remaining_time": "0:21:03"}
{"current_steps": 461, "total_steps": 555, "loss": 0.0099, "lr": 8.503190189973914e-06, "epoch": 4.148148148148148, "percentage": 83.06, "elapsed_time": "1:42:09", "remaining_time": "0:20:49"}
{"current_steps": 462, "total_steps": 555, "loss": 0.008, "lr": 8.328406171289621e-06, "epoch": 4.15712682379349, "percentage": 83.24, "elapsed_time": "1:42:21", "remaining_time": "0:20:36"}
{"current_steps": 463, "total_steps": 555, "loss": 0.0076, "lr": 8.155273876041614e-06, "epoch": 4.1661054994388325, "percentage": 83.42, "elapsed_time": "1:42:32", "remaining_time": "0:20:22"}
{"current_steps": 464, "total_steps": 555, "loss": 0.0087, "lr": 7.983800166618482e-06, "epoch": 4.175084175084175, "percentage": 83.6, "elapsed_time": "1:42:44", "remaining_time": "0:20:09"}
{"current_steps": 465, "total_steps": 555, "loss": 0.0107, "lr": 7.813991839667995e-06, "epoch": 4.184062850729517, "percentage": 83.78, "elapsed_time": "1:42:56", "remaining_time": "0:19:55"}
{"current_steps": 465, "total_steps": 555, "eval_loss": 0.012637750245630741, "epoch": 4.184062850729517, "percentage": 83.78, "elapsed_time": "1:43:03", "remaining_time": "0:19:56"}
{"current_steps": 466, "total_steps": 555, "loss": 0.007, "lr": 7.645855625827658e-06, "epoch": 4.193041526374859, "percentage": 83.96, "elapsed_time": "1:43:15", "remaining_time": "0:19:43"}
{"current_steps": 467, "total_steps": 555, "loss": 0.0071, "lr": 7.4793981894580034e-06, "epoch": 4.202020202020202, "percentage": 84.14, "elapsed_time": "1:43:27", "remaining_time": "0:19:29"}
{"current_steps": 468, "total_steps": 555, "loss": 0.0086, "lr": 7.3146261283784104e-06, "epoch": 4.210998877665545, "percentage": 84.32, "elapsed_time": "1:43:39", "remaining_time": "0:19:16"}
{"current_steps": 469, "total_steps": 555, "loss": 0.0071, "lr": 7.1515459736055505e-06, "epoch": 4.219977553310887, "percentage": 84.5, "elapsed_time": "1:43:51", "remaining_time": "0:19:02"}
{"current_steps": 470, "total_steps": 555, "loss": 0.0105, "lr": 6.990164189094589e-06, "epoch": 4.228956228956229, "percentage": 84.68, "elapsed_time": "1:44:04", "remaining_time": "0:18:49"}
{"current_steps": 470, "total_steps": 555, "eval_loss": 0.012454288080334663, "epoch": 4.228956228956229, "percentage": 84.68, "elapsed_time": "1:44:10", "remaining_time": "0:18:50"}
{"current_steps": 471, "total_steps": 555, "loss": 0.0085, "lr": 6.830487171482935e-06, "epoch": 4.2379349046015715, "percentage": 84.86, "elapsed_time": "1:44:20", "remaining_time": "0:18:36"}
{"current_steps": 472, "total_steps": 555, "loss": 0.0087, "lr": 6.6725212498366885e-06, "epoch": 4.246913580246914, "percentage": 85.05, "elapsed_time": "1:44:32", "remaining_time": "0:18:22"}
{"current_steps": 473, "total_steps": 555, "loss": 0.009, "lr": 6.516272685399793e-06, "epoch": 4.255892255892256, "percentage": 85.23, "elapsed_time": "1:44:42", "remaining_time": "0:18:09"}
{"current_steps": 474, "total_steps": 555, "loss": 0.0056, "lr": 6.36174767134588e-06, "epoch": 4.264870931537598, "percentage": 85.41, "elapsed_time": "1:44:52", "remaining_time": "0:17:55"}
{"current_steps": 475, "total_steps": 555, "loss": 0.0089, "lr": 6.208952332532786e-06, "epoch": 4.273849607182941, "percentage": 85.59, "elapsed_time": "1:45:04", "remaining_time": "0:17:41"}
{"current_steps": 475, "total_steps": 555, "eval_loss": 0.012438948266208172, "epoch": 4.273849607182941, "percentage": 85.59, "elapsed_time": "1:45:10", "remaining_time": "0:17:42"}
{"current_steps": 476, "total_steps": 555, "loss": 0.0079, "lr": 6.057892725259717e-06, "epoch": 4.282828282828283, "percentage": 85.77, "elapsed_time": "1:45:22", "remaining_time": "0:17:29"}
{"current_steps": 477, "total_steps": 555, "loss": 0.0086, "lr": 5.908574837027309e-06, "epoch": 4.291806958473625, "percentage": 85.95, "elapsed_time": "1:45:35", "remaining_time": "0:17:15"}
{"current_steps": 478, "total_steps": 555, "loss": 0.0092, "lr": 5.761004586300234e-06, "epoch": 4.300785634118967, "percentage": 86.13, "elapsed_time": "1:45:45", "remaining_time": "0:17:02"}
{"current_steps": 479, "total_steps": 555, "loss": 0.0073, "lr": 5.615187822272583e-06, "epoch": 4.30976430976431, "percentage": 86.31, "elapsed_time": "1:45:58", "remaining_time": "0:16:48"}
{"current_steps": 480, "total_steps": 555, "loss": 0.0061, "lr": 5.4711303246361144e-06, "epoch": 4.318742985409652, "percentage": 86.49, "elapsed_time": "1:46:11", "remaining_time": "0:16:35"}
{"current_steps": 480, "total_steps": 555, "eval_loss": 0.012455189600586891, "epoch": 4.318742985409652, "percentage": 86.49, "elapsed_time": "1:46:17", "remaining_time": "0:16:36"}
{"current_steps": 481, "total_steps": 555, "loss": 0.008, "lr": 5.328837803351083e-06, "epoch": 4.327721661054994, "percentage": 86.67, "elapsed_time": "1:46:29", "remaining_time": "0:16:23"}
{"current_steps": 482, "total_steps": 555, "loss": 0.0085, "lr": 5.188315898419971e-06, "epoch": 4.3367003367003365, "percentage": 86.85, "elapsed_time": "1:46:41", "remaining_time": "0:16:09"}
{"current_steps": 483, "total_steps": 555, "loss": 0.0079, "lr": 5.04957017966391e-06, "epoch": 4.345679012345679, "percentage": 87.03, "elapsed_time": "1:46:54", "remaining_time": "0:15:56"}
{"current_steps": 484, "total_steps": 555, "loss": 0.0103, "lr": 4.912606146501886e-06, "epoch": 4.354657687991021, "percentage": 87.21, "elapsed_time": "1:47:06", "remaining_time": "0:15:42"}
{"current_steps": 485, "total_steps": 555, "loss": 0.0074, "lr": 4.777429227732844e-06, "epoch": 4.363636363636363, "percentage": 87.39, "elapsed_time": "1:47:17", "remaining_time": "0:15:29"}
{"current_steps": 485, "total_steps": 555, "eval_loss": 0.012579456903040409, "epoch": 4.363636363636363, "percentage": 87.39, "elapsed_time": "1:47:24", "remaining_time": "0:15:30"}
{"current_steps": 486, "total_steps": 555, "loss": 0.0082, "lr": 4.644044781320422e-06, "epoch": 4.372615039281706, "percentage": 87.57, "elapsed_time": "1:47:35", "remaining_time": "0:15:16"}
{"current_steps": 487, "total_steps": 555, "loss": 0.0066, "lr": 4.5124580941806165e-06, "epoch": 4.381593714927048, "percentage": 87.75, "elapsed_time": "1:47:46", "remaining_time": "0:15:02"}
{"current_steps": 488, "total_steps": 555, "loss": 0.0094, "lr": 4.382674381972224e-06, "epoch": 4.390572390572391, "percentage": 87.93, "elapsed_time": "1:47:58", "remaining_time": "0:14:49"}
{"current_steps": 489, "total_steps": 555, "loss": 0.0084, "lr": 4.254698788890127e-06, "epoch": 4.399551066217733, "percentage": 88.11, "elapsed_time": "1:48:10", "remaining_time": "0:14:36"}
{"current_steps": 490, "total_steps": 555, "loss": 0.008, "lr": 4.12853638746134e-06, "epoch": 4.408529741863076, "percentage": 88.29, "elapsed_time": "1:48:22", "remaining_time": "0:14:22"}
{"current_steps": 490, "total_steps": 555, "eval_loss": 0.01255893800407648, "epoch": 4.408529741863076, "percentage": 88.29, "elapsed_time": "1:48:29", "remaining_time": "0:14:23"}
{"current_steps": 491, "total_steps": 555, "loss": 0.0089, "lr": 4.004192178344029e-06, "epoch": 4.417508417508418, "percentage": 88.47, "elapsed_time": "1:48:41", "remaining_time": "0:14:10"}
{"current_steps": 492, "total_steps": 555, "loss": 0.0074, "lr": 3.881671090129247e-06, "epoch": 4.42648709315376, "percentage": 88.65, "elapsed_time": "1:48:51", "remaining_time": "0:13:56"}
{"current_steps": 493, "total_steps": 555, "loss": 0.0077, "lr": 3.7609779791455744e-06, "epoch": 4.435465768799102, "percentage": 88.83, "elapsed_time": "1:49:03", "remaining_time": "0:13:42"}
{"current_steps": 494, "total_steps": 555, "loss": 0.0094, "lr": 3.6421176292666783e-06, "epoch": 4.444444444444445, "percentage": 89.01, "elapsed_time": "1:49:16", "remaining_time": "0:13:29"}
{"current_steps": 495, "total_steps": 555, "loss": 0.0092, "lr": 3.5250947517216637e-06, "epoch": 4.453423120089787, "percentage": 89.19, "elapsed_time": "1:49:28", "remaining_time": "0:13:16"}
{"current_steps": 495, "total_steps": 555, "eval_loss": 0.012510711327195168, "epoch": 4.453423120089787, "percentage": 89.19, "elapsed_time": "1:49:34", "remaining_time": "0:13:16"}
{"current_steps": 496, "total_steps": 555, "loss": 0.0068, "lr": 3.4099139849083307e-06, "epoch": 4.462401795735129, "percentage": 89.37, "elapsed_time": "1:49:47", "remaining_time": "0:13:03"}
{"current_steps": 497, "total_steps": 555, "loss": 0.006, "lr": 3.296579894209345e-06, "epoch": 4.4713804713804715, "percentage": 89.55, "elapsed_time": "1:49:59", "remaining_time": "0:12:50"}
{"current_steps": 498, "total_steps": 555, "loss": 0.0101, "lr": 3.1850969718112745e-06, "epoch": 4.480359147025814, "percentage": 89.73, "elapsed_time": "1:50:11", "remaining_time": "0:12:36"}
{"current_steps": 499, "total_steps": 555, "loss": 0.0071, "lr": 3.0754696365265068e-06, "epoch": 4.489337822671156, "percentage": 89.91, "elapsed_time": "1:50:22", "remaining_time": "0:12:23"}
{"current_steps": 500, "total_steps": 555, "loss": 0.0092, "lr": 2.9677022336181413e-06, "epoch": 4.498316498316498, "percentage": 90.09, "elapsed_time": "1:50:32", "remaining_time": "0:12:09"}
{"current_steps": 500, "total_steps": 555, "eval_loss": 0.012468446046113968, "epoch": 4.498316498316498, "percentage": 90.09, "elapsed_time": "1:50:39", "remaining_time": "0:12:10"}
{"current_steps": 501, "total_steps": 555, "loss": 0.0057, "lr": 2.8617990346277657e-06, "epoch": 4.5072951739618405, "percentage": 90.27, "elapsed_time": "1:50:58", "remaining_time": "0:11:57"}
{"current_steps": 502, "total_steps": 555, "loss": 0.0086, "lr": 2.7577642372060673e-06, "epoch": 4.516273849607183, "percentage": 90.45, "elapsed_time": "1:51:08", "remaining_time": "0:11:44"}
{"current_steps": 503, "total_steps": 555, "loss": 0.0103, "lr": 2.6556019649465525e-06, "epoch": 4.525252525252525, "percentage": 90.63, "elapsed_time": "1:51:19", "remaining_time": "0:11:30"}
{"current_steps": 504, "total_steps": 555, "loss": 0.008, "lr": 2.5553162672220465e-06, "epoch": 4.534231200897867, "percentage": 90.81, "elapsed_time": "1:51:30", "remaining_time": "0:11:17"}
{"current_steps": 505, "total_steps": 555, "loss": 0.0061, "lr": 2.45691111902418e-06, "epoch": 4.54320987654321, "percentage": 90.99, "elapsed_time": "1:51:41", "remaining_time": "0:11:03"}
{"current_steps": 505, "total_steps": 555, "eval_loss": 0.012444855645298958, "epoch": 4.54320987654321, "percentage": 90.99, "elapsed_time": "1:51:47", "remaining_time": "0:11:04"}
{"current_steps": 506, "total_steps": 555, "loss": 0.0099, "lr": 2.360390420805869e-06, "epoch": 4.552188552188552, "percentage": 91.17, "elapsed_time": "1:51:59", "remaining_time": "0:10:50"}
{"current_steps": 507, "total_steps": 555, "loss": 0.0059, "lr": 2.2657579983267064e-06, "epoch": 4.561167227833894, "percentage": 91.35, "elapsed_time": "1:52:10", "remaining_time": "0:10:37"}
{"current_steps": 508, "total_steps": 555, "loss": 0.0081, "lr": 2.1730176025012816e-06, "epoch": 4.570145903479236, "percentage": 91.53, "elapsed_time": "1:52:23", "remaining_time": "0:10:23"}
{"current_steps": 509, "total_steps": 555, "loss": 0.0082, "lr": 2.082172909250568e-06, "epoch": 4.57912457912458, "percentage": 91.71, "elapsed_time": "1:52:36", "remaining_time": "0:10:10"}
{"current_steps": 510, "total_steps": 555, "loss": 0.0089, "lr": 1.993227519356189e-06, "epoch": 4.588103254769921, "percentage": 91.89, "elapsed_time": "1:52:48", "remaining_time": "0:09:57"}
{"current_steps": 510, "total_steps": 555, "eval_loss": 0.012445243075489998, "epoch": 4.588103254769921, "percentage": 91.89, "elapsed_time": "1:52:54", "remaining_time": "0:09:57"}
{"current_steps": 511, "total_steps": 555, "loss": 0.0069, "lr": 1.906184958317664e-06, "epoch": 4.597081930415264, "percentage": 92.07, "elapsed_time": "1:53:06", "remaining_time": "0:09:44"}
{"current_steps": 512, "total_steps": 555, "loss": 0.0084, "lr": 1.8210486762127499e-06, "epoch": 4.606060606060606, "percentage": 92.25, "elapsed_time": "1:53:18", "remaining_time": "0:09:30"}
{"current_steps": 513, "total_steps": 555, "loss": 0.0076, "lr": 1.737822047560611e-06, "epoch": 4.615039281705949, "percentage": 92.43, "elapsed_time": "1:53:30", "remaining_time": "0:09:17"}
{"current_steps": 514, "total_steps": 555, "loss": 0.0088, "lr": 1.656508371188109e-06, "epoch": 4.624017957351291, "percentage": 92.61, "elapsed_time": "1:53:42", "remaining_time": "0:09:04"}
{"current_steps": 515, "total_steps": 555, "loss": 0.01, "lr": 1.5771108700990412e-06, "epoch": 4.632996632996633, "percentage": 92.79, "elapsed_time": "1:53:55", "remaining_time": "0:08:50"}
{"current_steps": 515, "total_steps": 555, "eval_loss": 0.012414975091814995, "epoch": 4.632996632996633, "percentage": 92.79, "elapsed_time": "1:54:01", "remaining_time": "0:08:51"}
{"current_steps": 516, "total_steps": 555, "loss": 0.0058, "lr": 1.4996326913463754e-06, "epoch": 4.6419753086419755, "percentage": 92.97, "elapsed_time": "1:54:12", "remaining_time": "0:08:37"}
{"current_steps": 517, "total_steps": 555, "loss": 0.0089, "lr": 1.4240769059075342e-06, "epoch": 4.650953984287318, "percentage": 93.15, "elapsed_time": "1:54:24", "remaining_time": "0:08:24"}
{"current_steps": 518, "total_steps": 555, "loss": 0.0059, "lr": 1.3504465085626638e-06, "epoch": 4.65993265993266, "percentage": 93.33, "elapsed_time": "1:54:34", "remaining_time": "0:08:11"}
{"current_steps": 519, "total_steps": 555, "loss": 0.0075, "lr": 1.2787444177759068e-06, "epoch": 4.668911335578002, "percentage": 93.51, "elapsed_time": "1:54:45", "remaining_time": "0:07:57"}
{"current_steps": 520, "total_steps": 555, "loss": 0.0081, "lr": 1.208973475579761e-06, "epoch": 4.677890011223345, "percentage": 93.69, "elapsed_time": "1:54:56", "remaining_time": "0:07:44"}
{"current_steps": 520, "total_steps": 555, "eval_loss": 0.012423361651599407, "epoch": 4.677890011223345, "percentage": 93.69, "elapsed_time": "1:55:02", "remaining_time": "0:07:44"}
{"current_steps": 521, "total_steps": 555, "loss": 0.007, "lr": 1.1411364474624264e-06, "epoch": 4.686868686868687, "percentage": 93.87, "elapsed_time": "1:55:14", "remaining_time": "0:07:31"}
{"current_steps": 522, "total_steps": 555, "loss": 0.009, "lr": 1.075236022258147e-06, "epoch": 4.695847362514029, "percentage": 94.05, "elapsed_time": "1:55:26", "remaining_time": "0:07:17"}
{"current_steps": 523, "total_steps": 555, "loss": 0.0092, "lr": 1.0112748120406856e-06, "epoch": 4.704826038159371, "percentage": 94.23, "elapsed_time": "1:55:38", "remaining_time": "0:07:04"}
{"current_steps": 524, "total_steps": 555, "loss": 0.0072, "lr": 9.492553520197733e-07, "epoch": 4.713804713804714, "percentage": 94.41, "elapsed_time": "1:55:49", "remaining_time": "0:06:51"}
{"current_steps": 525, "total_steps": 555, "loss": 0.0072, "lr": 8.891801004406119e-07, "epoch": 4.722783389450056, "percentage": 94.59, "elapsed_time": "1:56:02", "remaining_time": "0:06:37"}
{"current_steps": 525, "total_steps": 555, "eval_loss": 0.01243152841925621, "epoch": 4.722783389450056, "percentage": 94.59, "elapsed_time": "1:56:08", "remaining_time": "0:06:38"}
{"current_steps": 526, "total_steps": 555, "loss": 0.0089, "lr": 8.31051438486441e-07, "epoch": 4.731762065095398, "percentage": 94.77, "elapsed_time": "1:56:20", "remaining_time": "0:06:24"}
{"current_steps": 527, "total_steps": 555, "loss": 0.0097, "lr": 7.748716701841685e-07, "epoch": 4.7407407407407405, "percentage": 94.95, "elapsed_time": "1:56:33", "remaining_time": "0:06:11"}
{"current_steps": 528, "total_steps": 555, "loss": 0.0054, "lr": 7.206430223130278e-07, "epoch": 4.749719416386083, "percentage": 95.14, "elapsed_time": "1:56:43", "remaining_time": "0:05:58"}
{"current_steps": 529, "total_steps": 555, "loss": 0.0114, "lr": 6.683676443163311e-07, "epoch": 4.758698092031425, "percentage": 95.32, "elapsed_time": "1:56:55", "remaining_time": "0:05:44"}
{"current_steps": 530, "total_steps": 555, "loss": 0.0078, "lr": 6.180476082162656e-07, "epoch": 4.767676767676767, "percentage": 95.5, "elapsed_time": "1:57:08", "remaining_time": "0:05:31"}
{"current_steps": 530, "total_steps": 555, "eval_loss": 0.012432167306542397, "epoch": 4.767676767676767, "percentage": 95.5, "elapsed_time": "1:57:15", "remaining_time": "0:05:31"}
{"current_steps": 531, "total_steps": 555, "loss": 0.0068, "lr": 5.696849085317646e-07, "epoch": 4.77665544332211, "percentage": 95.68, "elapsed_time": "1:57:27", "remaining_time": "0:05:18"}
{"current_steps": 532, "total_steps": 555, "loss": 0.0084, "lr": 5.232814621994598e-07, "epoch": 4.785634118967453, "percentage": 95.86, "elapsed_time": "1:57:40", "remaining_time": "0:05:05"}
{"current_steps": 533, "total_steps": 555, "loss": 0.0097, "lr": 4.788391084976862e-07, "epoch": 4.794612794612795, "percentage": 96.04, "elapsed_time": "1:57:52", "remaining_time": "0:04:51"}
{"current_steps": 534, "total_steps": 555, "loss": 0.0075, "lr": 4.363596089735911e-07, "epoch": 4.803591470258137, "percentage": 96.22, "elapsed_time": "1:58:04", "remaining_time": "0:04:38"}
{"current_steps": 535, "total_steps": 555, "loss": 0.009, "lr": 3.958446473733002e-07, "epoch": 4.8125701459034795, "percentage": 96.4, "elapsed_time": "1:58:14", "remaining_time": "0:04:25"}
{"current_steps": 535, "total_steps": 555, "eval_loss": 0.012407775036990643, "epoch": 4.8125701459034795, "percentage": 96.4, "elapsed_time": "1:58:21", "remaining_time": "0:04:25"}
{"current_steps": 536, "total_steps": 555, "loss": 0.0066, "lr": 3.572958295752049e-07, "epoch": 4.821548821548822, "percentage": 96.58, "elapsed_time": "1:58:32", "remaining_time": "0:04:12"}
{"current_steps": 537, "total_steps": 555, "loss": 0.0063, "lr": 3.207146835262742e-07, "epoch": 4.830527497194164, "percentage": 96.76, "elapsed_time": "1:58:44", "remaining_time": "0:03:58"}
{"current_steps": 538, "total_steps": 555, "loss": 0.0093, "lr": 2.8610265918151414e-07, "epoch": 4.839506172839506, "percentage": 96.94, "elapsed_time": "1:58:55", "remaining_time": "0:03:45"}
{"current_steps": 539, "total_steps": 555, "loss": 0.0076, "lr": 2.534611284465083e-07, "epoch": 4.848484848484849, "percentage": 97.12, "elapsed_time": "1:59:09", "remaining_time": "0:03:32"}
{"current_steps": 540, "total_steps": 555, "loss": 0.0106, "lr": 2.2279138512300567e-07, "epoch": 4.857463524130191, "percentage": 97.3, "elapsed_time": "1:59:19", "remaining_time": "0:03:18"}
{"current_steps": 540, "total_steps": 555, "eval_loss": 0.01241106167435646, "epoch": 4.857463524130191, "percentage": 97.3, "elapsed_time": "1:59:26", "remaining_time": "0:03:19"}
{"current_steps": 541, "total_steps": 555, "loss": 0.0063, "lr": 1.940946448576675e-07, "epoch": 4.866442199775533, "percentage": 97.48, "elapsed_time": "1:59:38", "remaining_time": "0:03:05"}
{"current_steps": 542, "total_steps": 555, "loss": 0.0075, "lr": 1.6737204509387206e-07, "epoch": 4.875420875420875, "percentage": 97.66, "elapsed_time": "1:59:52", "remaining_time": "0:02:52"}
{"current_steps": 543, "total_steps": 555, "loss": 0.0099, "lr": 1.4262464502663443e-07, "epoch": 4.884399551066218, "percentage": 97.84, "elapsed_time": "2:00:06", "remaining_time": "0:02:39"}
{"current_steps": 544, "total_steps": 555, "loss": 0.0074, "lr": 1.1985342556060652e-07, "epoch": 4.89337822671156, "percentage": 98.02, "elapsed_time": "2:00:18", "remaining_time": "0:02:25"}
{"current_steps": 545, "total_steps": 555, "loss": 0.0079, "lr": 9.905928927123609e-08, "epoch": 4.902356902356902, "percentage": 98.2, "elapsed_time": "2:00:30", "remaining_time": "0:02:12"}
{"current_steps": 545, "total_steps": 555, "eval_loss": 0.012396564707159996, "epoch": 4.902356902356902, "percentage": 98.2, "elapsed_time": "2:00:37", "remaining_time": "0:02:12"}
{"current_steps": 546, "total_steps": 555, "loss": 0.0079, "lr": 8.02430603689397e-08, "epoch": 4.9113355780022445, "percentage": 98.38, "elapsed_time": "2:00:48", "remaining_time": "0:01:59"}
{"current_steps": 547, "total_steps": 555, "loss": 0.0084, "lr": 6.340548466648443e-08, "epoch": 4.920314253647587, "percentage": 98.56, "elapsed_time": "2:01:01", "remaining_time": "0:01:46"}
{"current_steps": 548, "total_steps": 555, "loss": 0.0093, "lr": 4.8547229549383844e-08, "epoch": 4.929292929292929, "percentage": 98.74, "elapsed_time": "2:01:14", "remaining_time": "0:01:32"}
{"current_steps": 549, "total_steps": 555, "loss": 0.0074, "lr": 3.566888394948009e-08, "epoch": 4.938271604938271, "percentage": 98.92, "elapsed_time": "2:01:26", "remaining_time": "0:01:19"}
{"current_steps": 550, "total_steps": 555, "loss": 0.0082, "lr": 2.4770958321568283e-08, "epoch": 4.947250280583614, "percentage": 99.1, "elapsed_time": "2:01:37", "remaining_time": "0:01:06"}
{"current_steps": 550, "total_steps": 555, "eval_loss": 0.01241134200245142, "epoch": 4.947250280583614, "percentage": 99.1, "elapsed_time": "2:01:43", "remaining_time": "0:01:06"}
{"current_steps": 551, "total_steps": 555, "loss": 0.0063, "lr": 1.5853884623195925e-08, "epoch": 4.956228956228956, "percentage": 99.28, "elapsed_time": "2:02:02", "remaining_time": "0:00:53"}
{"current_steps": 552, "total_steps": 555, "loss": 0.0098, "lr": 8.918016297515541e-09, "epoch": 4.965207631874298, "percentage": 99.46, "elapsed_time": "2:02:13", "remaining_time": "0:00:39"}
{"current_steps": 553, "total_steps": 555, "loss": 0.0088, "lr": 3.963628259290308e-09, "epoch": 4.974186307519641, "percentage": 99.64, "elapsed_time": "2:02:24", "remaining_time": "0:00:26"}
{"current_steps": 554, "total_steps": 555, "loss": 0.0077, "lr": 9.90916883986115e-10, "epoch": 4.983164983164983, "percentage": 99.82, "elapsed_time": "2:02:36", "remaining_time": "0:00:13"}
{"current_steps": 555, "total_steps": 555, "loss": 0.0082, "lr": 0.0, "epoch": 4.992143658810326, "percentage": 100.0, "elapsed_time": "2:02:47", "remaining_time": "0:00:00"}
{"current_steps": 555, "total_steps": 555, "eval_loss": 0.012434104457497597, "epoch": 4.992143658810326, "percentage": 100.0, "elapsed_time": "2:02:53", "remaining_time": "0:00:00"}
{"current_steps": 555, "total_steps": 555, "epoch": 4.992143658810326, "percentage": 100.0, "elapsed_time": "2:03:00", "remaining_time": "0:00:00"}