{"current_steps": 5, "total_steps": 3400, "loss": 1.9191, "lr": 2.9411764705882355e-06, "epoch": 0.0009188079997549845, "percentage": 0.15, "elapsed_time": "0:02:45", "remaining_time": "1 day, 7:14:13", "throughput": 305.09, "total_tokens": 50528}
{"current_steps": 10, "total_steps": 3400, "loss": 2.0838, "lr": 5.882352941176471e-06, "epoch": 0.001837615999509969, "percentage": 0.29, "elapsed_time": "0:04:14", "remaining_time": "23:59:20", "throughput": 396.0, "total_tokens": 100880}
{"current_steps": 15, "total_steps": 3400, "loss": 1.9921, "lr": 8.823529411764707e-06, "epoch": 0.002756423999264954, "percentage": 0.44, "elapsed_time": "0:05:43", "remaining_time": "21:33:26", "throughput": 440.03, "total_tokens": 151328}
{"current_steps": 20, "total_steps": 3400, "loss": 2.0412, "lr": 1.1764705882352942e-05, "epoch": 0.003675231999019938, "percentage": 0.59, "elapsed_time": "0:07:13", "remaining_time": "20:22:23", "throughput": 462.36, "total_tokens": 200656}
{"current_steps": 25, "total_steps": 3400, "loss": 1.9913, "lr": 1.4705882352941177e-05, "epoch": 0.004594039998774922, "percentage": 0.74, "elapsed_time": "0:08:43", "remaining_time": "19:38:58", "throughput": 478.36, "total_tokens": 250656}
{"current_steps": 30, "total_steps": 3400, "loss": 1.8291, "lr": 1.7647058823529414e-05, "epoch": 0.005512847998529908, "percentage": 0.88, "elapsed_time": "0:10:14", "remaining_time": "19:10:29", "throughput": 489.32, "total_tokens": 300688}
{"current_steps": 35, "total_steps": 3400, "loss": 1.6077, "lr": 2.058823529411765e-05, "epoch": 0.006431655998284892, "percentage": 1.03, "elapsed_time": "0:11:43", "remaining_time": "18:47:30", "throughput": 498.81, "total_tokens": 350984}
{"current_steps": 40, "total_steps": 3400, "loss": 1.4619, "lr": 2.3529411764705884e-05, "epoch": 0.007350463998039876, "percentage": 1.18, "elapsed_time": "0:13:13", "remaining_time": "18:30:40", "throughput": 505.71, "total_tokens": 401200}
{"current_steps": 45, "total_steps": 3400, "loss": 1.1743, "lr": 2.647058823529412e-05, "epoch": 0.00826927199779486, "percentage": 1.32, "elapsed_time": "0:14:41", "remaining_time": "18:15:50", "throughput": 512.82, "total_tokens": 452256}
{"current_steps": 50, "total_steps": 3400, "loss": 1.0351, "lr": 2.9411764705882354e-05, "epoch": 0.009188079997549845, "percentage": 1.47, "elapsed_time": "0:16:10", "remaining_time": "18:04:03", "throughput": 517.48, "total_tokens": 502368}
{"current_steps": 50, "total_steps": 3400, "eval_loss": 0.9696781039237976, "epoch": 0.009188079997549845, "percentage": 1.47, "elapsed_time": "0:16:57", "remaining_time": "18:56:13", "throughput": 493.72, "total_tokens": 502368}
{"current_steps": 55, "total_steps": 3400, "loss": 0.9603, "lr": 3.235294117647059e-05, "epoch": 0.01010688799730483, "percentage": 1.62, "elapsed_time": "0:18:33", "remaining_time": "18:48:42", "throughput": 496.65, "total_tokens": 553032}
{"current_steps": 60, "total_steps": 3400, "loss": 0.8865, "lr": 3.529411764705883e-05, "epoch": 0.011025695997059815, "percentage": 1.76, "elapsed_time": "0:20:02", "remaining_time": "18:36:00", "throughput": 501.7, "total_tokens": 603488}
{"current_steps": 65, "total_steps": 3400, "loss": 0.8452, "lr": 3.8235294117647055e-05, "epoch": 0.0119445039968148, "percentage": 1.91, "elapsed_time": "0:21:30", "remaining_time": "18:23:40", "throughput": 506.74, "total_tokens": 654032}
{"current_steps": 70, "total_steps": 3400, "loss": 0.7973, "lr": 4.11764705882353e-05, "epoch": 0.012863311996569784, "percentage": 2.06, "elapsed_time": "0:23:00", "remaining_time": "18:14:34", "throughput": 509.59, "total_tokens": 703512}
{"current_steps": 75, "total_steps": 3400, "loss": 0.8114, "lr": 4.411764705882353e-05, "epoch": 0.013782119996324768, "percentage": 2.21, "elapsed_time": "0:24:28", "remaining_time": "18:05:06", "throughput": 513.32, "total_tokens": 753856}
{"current_steps": 80, "total_steps": 3400, "loss": 0.8446, "lr": 4.705882352941177e-05, "epoch": 0.014700927996079752, "percentage": 2.35, "elapsed_time": "0:25:58", "remaining_time": "17:57:38", "throughput": 516.14, "total_tokens": 804160}
{"current_steps": 85, "total_steps": 3400, "loss": 0.8256, "lr": 5e-05, "epoch": 0.015619735995834737, "percentage": 2.5, "elapsed_time": "0:27:26", "remaining_time": "17:50:00", "throughput": 519.06, "total_tokens": 854456}
{"current_steps": 55, "total_steps": 3400, "loss": 0.9428, "lr": 3.235294117647059e-05, "epoch": 0.006737925331536553, "percentage": 1.62, "elapsed_time": "0:02:19", "remaining_time": "2:20:56", "throughput": 3854.82, "total_tokens": 535968}
{"current_steps": 60, "total_steps": 3400, "loss": 0.9121, "lr": 3.529411764705883e-05, "epoch": 0.007350463998039876, "percentage": 1.76, "elapsed_time": "0:03:19", "remaining_time": "3:05:15", "throughput": 2851.32, "total_tokens": 569344}
{"current_steps": 65, "total_steps": 3400, "loss": 0.915, "lr": 3.8235294117647055e-05, "epoch": 0.007963002664543199, "percentage": 1.91, "elapsed_time": "0:04:20", "remaining_time": "3:42:25", "throughput": 2321.45, "total_tokens": 603816}
{"current_steps": 70, "total_steps": 3400, "loss": 0.7588, "lr": 4.11764705882353e-05, "epoch": 0.008575541331046523, "percentage": 2.06, "elapsed_time": "0:05:19", "remaining_time": "4:13:25", "throughput": 1993.46, "total_tokens": 637176}
{"current_steps": 75, "total_steps": 3400, "loss": 0.8561, "lr": 4.411764705882353e-05, "epoch": 0.009188079997549845, "percentage": 2.21, "elapsed_time": "0:06:19", "remaining_time": "4:40:41", "throughput": 1765.06, "total_tokens": 670512}
{"current_steps": 80, "total_steps": 3400, "loss": 0.7766, "lr": 4.705882352941177e-05, "epoch": 0.009800618664053169, "percentage": 2.35, "elapsed_time": "0:07:19", "remaining_time": "5:03:53", "throughput": 1602.66, "total_tokens": 704160}
{"current_steps": 85, "total_steps": 3400, "loss": 0.8268, "lr": 5e-05, "epoch": 0.010413157330556491, "percentage": 2.5, "elapsed_time": "0:08:19", "remaining_time": "5:24:32", "throughput": 1478.31, "total_tokens": 738120}
{"current_steps": 90, "total_steps": 3400, "loss": 0.7961, "lr": 5.294117647058824e-05, "epoch": 0.011025695997059815, "percentage": 2.65, "elapsed_time": "0:09:19", "remaining_time": "5:43:13", "throughput": 1378.08, "total_tokens": 771632}
{"current_steps": 95, "total_steps": 3400, "loss": 0.7732, "lr": 5.588235294117647e-05, "epoch": 0.011638234663563138, "percentage": 2.79, "elapsed_time": "0:10:19", "remaining_time": "5:59:11", "throughput": 1300.0, "total_tokens": 805328}
{"current_steps": 100, "total_steps": 3400, "loss": 0.8279, "lr": 5.882352941176471e-05, "epoch": 0.01225077333006646, "percentage": 2.94, "elapsed_time": "0:11:19", "remaining_time": "6:13:43", "throughput": 1234.5, "total_tokens": 838824}
{"current_steps": 100, "total_steps": 3400, "eval_loss": 0.9583126902580261, "epoch": 0.01225077333006646, "percentage": 2.94, "elapsed_time": "0:12:06", "remaining_time": "6:39:42", "throughput": 1154.22, "total_tokens": 838824}
{"current_steps": 105, "total_steps": 3400, "loss": 0.7322, "lr": 6.176470588235295e-05, "epoch": 0.012863311996569784, "percentage": 3.09, "elapsed_time": "0:13:14", "remaining_time": "6:55:47", "throughput": 1096.44, "total_tokens": 871656}
{"current_steps": 110, "total_steps": 3400, "loss": 0.7735, "lr": 6.470588235294118e-05, "epoch": 0.013475850663073106, "percentage": 3.24, "elapsed_time": "0:14:14", "remaining_time": "7:05:56", "throughput": 1059.28, "total_tokens": 905144}
{"current_steps": 115, "total_steps": 3400, "loss": 0.8563, "lr": 6.764705882352942e-05, "epoch": 0.01408838932957643, "percentage": 3.38, "elapsed_time": "0:15:14", "remaining_time": "7:15:22", "throughput": 1026.42, "total_tokens": 938656}
{"current_steps": 120, "total_steps": 3400, "loss": 0.8066, "lr": 7.058823529411765e-05, "epoch": 0.014700927996079752, "percentage": 3.53, "elapsed_time": "0:16:14", "remaining_time": "7:24:08", "throughput": 997.3, "total_tokens": 972304}
{"current_steps": 125, "total_steps": 3400, "loss": 0.7967, "lr": 7.352941176470589e-05, "epoch": 0.015313466662583075, "percentage": 3.68, "elapsed_time": "0:17:14", "remaining_time": "7:31:50", "throughput": 971.81, "total_tokens": 1005576}
{"current_steps": 130, "total_steps": 3400, "loss": 0.7882, "lr": 7.647058823529411e-05, "epoch": 0.015926005329086397, "percentage": 3.82, "elapsed_time": "0:18:14", "remaining_time": "7:38:58", "throughput": 949.4, "total_tokens": 1039400}
{"current_steps": 135, "total_steps": 3400, "loss": 0.7331, "lr": 7.941176470588235e-05, "epoch": 0.01653854399558972, "percentage": 3.97, "elapsed_time": "0:19:14", "remaining_time": "7:45:28", "throughput": 928.77, "total_tokens": 1072528}
{"current_steps": 140, "total_steps": 3400, "loss": 0.751, "lr": 8.23529411764706e-05, "epoch": 0.017151082662093045, "percentage": 4.12, "elapsed_time": "0:20:15", "remaining_time": "7:51:50", "throughput": 909.85, "total_tokens": 1106184}
{"current_steps": 145, "total_steps": 3400, "loss": 0.7722, "lr": 8.529411764705883e-05, "epoch": 0.01776362132859637, "percentage": 4.26, "elapsed_time": "0:21:16", "remaining_time": "7:57:26", "throughput": 892.96, "total_tokens": 1139520}
{"current_steps": 150, "total_steps": 3400, "loss": 0.7354, "lr": 8.823529411764706e-05, "epoch": 0.01837615999509969, "percentage": 4.41, "elapsed_time": "0:22:16", "remaining_time": "8:02:40", "throughput": 877.76, "total_tokens": 1173256}
{"current_steps": 150, "total_steps": 3400, "eval_loss": 0.9015713930130005, "epoch": 0.01837615999509969, "percentage": 4.41, "elapsed_time": "0:22:36", "remaining_time": "8:09:53", "throughput": 864.83, "total_tokens": 1173256}
{"current_steps": 155, "total_steps": 3400, "loss": 0.7864, "lr": 9.11764705882353e-05, "epoch": 0.018988698661603014, "percentage": 4.56, "elapsed_time": "0:23:44", "remaining_time": "8:17:05", "throughput": 847.1, "total_tokens": 1206816}
{"current_steps": 160, "total_steps": 3400, "loss": 0.7358, "lr": 9.411764705882353e-05, "epoch": 0.019601237328106338, "percentage": 4.71, "elapsed_time": "0:24:44", "remaining_time": "8:21:02", "throughput": 835.28, "total_tokens": 1240040}
{"current_steps": 165, "total_steps": 3400, "loss": 0.7654, "lr": 9.705882352941177e-05, "epoch": 0.02021377599460966, "percentage": 4.85, "elapsed_time": "0:25:44", "remaining_time": "8:24:48", "throughput": 824.52, "total_tokens": 1273760}
{"current_steps": 170, "total_steps": 3400, "loss": 0.737, "lr": 0.0001, "epoch": 0.020826314661112982, "percentage": 5.0, "elapsed_time": "0:26:45", "remaining_time": "8:28:18", "throughput": 814.55, "total_tokens": 1307496}
{"current_steps": 175, "total_steps": 3400, "loss": 0.7722, "lr": 9.999940874631277e-05, "epoch": 0.021438853327616306, "percentage": 5.15, "elapsed_time": "0:27:44", "remaining_time": "8:31:19", "throughput": 805.64, "total_tokens": 1341208}
{"current_steps": 180, "total_steps": 3400, "loss": 0.7565, "lr": 9.999763499923432e-05, "epoch": 0.02205139199411963, "percentage": 5.29, "elapsed_time": "0:28:43", "remaining_time": "8:34:00", "throughput": 797.5, "total_tokens": 1374880}
{"current_steps": 185, "total_steps": 3400, "loss": 0.7794, "lr": 9.999467880071402e-05, "epoch": 0.02266393066062295, "percentage": 5.44, "elapsed_time": "0:29:44", "remaining_time": "8:36:50", "throughput": 789.31, "total_tokens": 1408456}
{"current_steps": 190, "total_steps": 3400, "loss": 0.767, "lr": 9.999054022066641e-05, "epoch": 0.023276469327126275, "percentage": 5.59, "elapsed_time": "0:30:43", "remaining_time": "8:39:09", "throughput": 782.22, "total_tokens": 1442192}
{"current_steps": 195, "total_steps": 3400, "loss": 0.7617, "lr": 9.998521935696953e-05, "epoch": 0.0238890079936296, "percentage": 5.74, "elapsed_time": "0:31:43", "remaining_time": "8:41:22", "throughput": 775.59, "total_tokens": 1476184}
{"current_steps": 200, "total_steps": 3400, "loss": 0.7706, "lr": 9.997871633546257e-05, "epoch": 0.02450154666013292, "percentage": 5.88, "elapsed_time": "0:32:43", "remaining_time": "8:43:33", "throughput": 769.03, "total_tokens": 1509888}
{"current_steps": 200, "total_steps": 3400, "eval_loss": 0.9173043966293335, "epoch": 0.02450154666013292, "percentage": 5.88, "elapsed_time": "0:33:03", "remaining_time": "8:48:48", "throughput": 761.4, "total_tokens": 1509888}
{"current_steps": 205, "total_steps": 3400, "loss": 0.7753, "lr": 9.997103130994296e-05, "epoch": 0.025114085326636244, "percentage": 6.03, "elapsed_time": "0:34:07", "remaining_time": "8:51:49", "throughput": 753.88, "total_tokens": 1543512}
{"current_steps": 210, "total_steps": 3400, "loss": 0.7671, "lr": 9.996216446216267e-05, "epoch": 0.025726623993139568, "percentage": 6.18, "elapsed_time": "0:35:06", "remaining_time": "8:53:17", "throughput": 749.43, "total_tokens": 1578600}
{"current_steps": 215, "total_steps": 3400, "loss": 0.7609, "lr": 9.995211600182397e-05, "epoch": 0.02633916265964289, "percentage": 6.32, "elapsed_time": "0:36:06", "remaining_time": "8:54:57", "throughput": 743.88, "total_tokens": 1611752}
{"current_steps": 220, "total_steps": 3400, "loss": 0.7415, "lr": 9.994088616657444e-05, "epoch": 0.026951701326146212, "percentage": 6.47, "elapsed_time": "0:37:05", "remaining_time": "8:56:07", "throughput": 739.21, "total_tokens": 1645048}
{"current_steps": 225, "total_steps": 3400, "loss": 0.7586, "lr": 9.992847522200133e-05, "epoch": 0.027564239992649536, "percentage": 6.62, "elapsed_time": "0:38:04", "remaining_time": "8:57:19", "throughput": 734.62, "total_tokens": 1678384}
{"current_steps": 230, "total_steps": 3400, "loss": 0.7651, "lr": 9.99148834616253e-05, "epoch": 0.02817677865915286, "percentage": 6.76, "elapsed_time": "0:39:05", "remaining_time": "8:58:47", "throughput": 729.71, "total_tokens": 1711528}
{"current_steps": 235, "total_steps": 3400, "loss": 0.7579, "lr": 9.990011120689351e-05, "epoch": 0.02878931732565618, "percentage": 6.91, "elapsed_time": "0:40:04", "remaining_time": "8:59:48", "throughput": 725.47, "total_tokens": 1744672}
{"current_steps": 240, "total_steps": 3400, "loss": 0.6783, "lr": 9.988415880717194e-05, "epoch": 0.029401855992159505, "percentage": 7.06, "elapsed_time": "0:41:04", "remaining_time": "9:00:47", "throughput": 721.6, "total_tokens": 1778296}
{"current_steps": 245, "total_steps": 3400, "loss": 0.7298, "lr": 9.986702663973722e-05, "epoch": 0.03001439465866283, "percentage": 7.21, "elapsed_time": "0:42:04", "remaining_time": "9:01:54", "throughput": 717.72, "total_tokens": 1812184}
{"current_steps": 250, "total_steps": 3400, "loss": 0.7321, "lr": 9.98487151097676e-05, "epoch": 0.03062693332516615, "percentage": 7.35, "elapsed_time": "0:43:04", "remaining_time": "9:02:41", "throughput": 714.33, "total_tokens": 1845992}
{"current_steps": 250, "total_steps": 3400, "eval_loss": 0.9278064966201782, "epoch": 0.03062693332516615, "percentage": 7.35, "elapsed_time": "0:43:24", "remaining_time": "9:06:52", "throughput": 708.85, "total_tokens": 1845992}
{"current_steps": 255, "total_steps": 3400, "loss": 0.6957, "lr": 9.98292246503335e-05, "epoch": 0.031239471991669474, "percentage": 7.5, "elapsed_time": "0:44:29", "remaining_time": "9:08:46", "throughput": 704.03, "total_tokens": 1879552}
{"current_steps": 260, "total_steps": 3400, "loss": 0.7431, "lr": 9.980855572238714e-05, "epoch": 0.031852010658172794, "percentage": 7.65, "elapsed_time": "0:45:30", "remaining_time": "9:09:37", "throughput": 700.57, "total_tokens": 1912984}
{"current_steps": 265, "total_steps": 3400, "loss": 0.7521, "lr": 9.978670881475172e-05, "epoch": 0.03246454932467612, "percentage": 7.79, "elapsed_time": "0:46:31", "remaining_time": "9:10:25", "throughput": 697.33, "total_tokens": 1946696}
{"current_steps": 270, "total_steps": 3400, "loss": 0.6869, "lr": 9.976368444410985e-05, "epoch": 0.03307708799117944, "percentage": 7.94, "elapsed_time": "0:47:32", "remaining_time": "9:11:02", "throughput": 694.39, "total_tokens": 1980440}
{"current_steps": 275, "total_steps": 3400, "loss": 0.7235, "lr": 9.973948315499126e-05, "epoch": 0.03368962665768277, "percentage": 8.09, "elapsed_time": "0:48:32", "remaining_time": "9:11:39", "throughput": 691.49, "total_tokens": 2014144}
{"current_steps": 255, "total_steps": 3400, "loss": 0.6957, "lr": 9.98292246503335e-05, "epoch": 0.031239471991669474, "percentage": 7.5, "elapsed_time": "0:02:18", "remaining_time": "0:28:27", "throughput": 13578.53, "total_tokens": 1879552}
{"current_steps": 260, "total_steps": 3400, "loss": 0.7431, "lr": 9.980855572238714e-05, "epoch": 0.031852010658172794, "percentage": 7.65, "elapsed_time": "0:03:17", "remaining_time": "0:39:44", "throughput": 9687.74, "total_tokens": 1912984}
{"current_steps": 275, "total_steps": 3400, "loss": 0.7235, "lr": 9.973948315499126e-05, "epoch": 0.03368962665768277, "percentage": 8.09, "elapsed_time": "0:06:14", "remaining_time": "1:10:56", "throughput": 5377.47, "total_tokens": 2014144}
{"current_steps": 280, "total_steps": 3400, "loss": 0.7413, "lr": 9.971410551976002e-05, "epoch": 0.03430216532418609, "percentage": 8.24, "elapsed_time": "0:07:13", "remaining_time": "1:20:27", "throughput": 4726.95, "total_tokens": 2047712}
{"current_steps": 285, "total_steps": 3400, "loss": 0.7457, "lr": 9.968755213860094e-05, "epoch": 0.03491470399068941, "percentage": 8.38, "elapsed_time": "0:08:12", "remaining_time": "1:29:38", "throughput": 4228.25, "total_tokens": 2080856}
{"current_steps": 290, "total_steps": 3400, "loss": 0.7467, "lr": 9.96598236395054e-05, "epoch": 0.03552724265719274, "percentage": 8.53, "elapsed_time": "0:09:11", "remaining_time": "1:38:34", "throughput": 3833.96, "total_tokens": 2114608}
{"current_steps": 295, "total_steps": 3400, "loss": 0.7178, "lr": 9.96309206782565e-05, "epoch": 0.03613978132369606, "percentage": 8.68, "elapsed_time": "0:10:10", "remaining_time": "1:47:00", "throughput": 3521.47, "total_tokens": 2148240}
{"current_steps": 300, "total_steps": 3400, "loss": 0.7602, "lr": 9.960084393841355e-05, "epoch": 0.03675231999019938, "percentage": 8.82, "elapsed_time": "0:11:09", "remaining_time": "1:55:14", "throughput": 3260.32, "total_tokens": 2181624}
{"current_steps": 300, "total_steps": 3400, "eval_loss": 0.9220121502876282, "epoch": 0.03675231999019938, "percentage": 8.82, "elapsed_time": "0:11:56", "remaining_time": "2:03:27", "throughput": 3043.16, "total_tokens": 2181624}
{"current_steps": 305, "total_steps": 3400, "loss": 0.7192, "lr": 9.956959413129585e-05, "epoch": 0.03736485865670271, "percentage": 8.97, "elapsed_time": "0:13:05", "remaining_time": "2:12:48", "throughput": 2821.78, "total_tokens": 2215832}
{"current_steps": 310, "total_steps": 3400, "loss": 0.7425, "lr": 9.953717199596598e-05, "epoch": 0.03797739732320603, "percentage": 9.12, "elapsed_time": "0:14:04", "remaining_time": "2:20:19", "throughput": 2662.98, "total_tokens": 2249328}
{"current_steps": 315, "total_steps": 3400, "loss": 0.7005, "lr": 9.95035782992122e-05, "epoch": 0.03858993598970935, "percentage": 9.26, "elapsed_time": "0:15:03", "remaining_time": "2:27:30", "throughput": 2525.83, "total_tokens": 2282584}
{"current_steps": 320, "total_steps": 3400, "loss": 0.6911, "lr": 9.94688138355304e-05, "epoch": 0.039202474656212676, "percentage": 9.41, "elapsed_time": "0:16:03", "remaining_time": "2:34:31", "throughput": 2403.89, "total_tokens": 2315728}
{"current_steps": 325, "total_steps": 3400, "loss": 0.7158, "lr": 9.943287942710527e-05, "epoch": 0.039815013322715996, "percentage": 9.56, "elapsed_time": "0:17:03", "remaining_time": "2:41:19", "throughput": 2296.36, "total_tokens": 2349176}
{"current_steps": 330, "total_steps": 3400, "loss": 0.6953, "lr": 9.939577592379088e-05, "epoch": 0.04042755198921932, "percentage": 9.71, "elapsed_time": "0:18:02", "remaining_time": "2:47:53", "throughput": 2200.1, "total_tokens": 2382408}
{"current_steps": 335, "total_steps": 3400, "loss": 0.7205, "lr": 9.935750420309055e-05, "epoch": 0.041040090655722644, "percentage": 9.85, "elapsed_time": "0:19:02", "remaining_time": "2:54:09", "throughput": 2115.19, "total_tokens": 2415856}
{"current_steps": 340, "total_steps": 3400, "loss": 0.7072, "lr": 9.931806517013612e-05, "epoch": 0.041652629322225965, "percentage": 10.0, "elapsed_time": "0:20:01", "remaining_time": "3:00:17", "throughput": 2038.19, "total_tokens": 2449744}
{"current_steps": 345, "total_steps": 3400, "loss": 0.665, "lr": 9.927745975766654e-05, "epoch": 0.042265167988729285, "percentage": 10.15, "elapsed_time": "0:21:01", "remaining_time": "3:06:06", "throughput": 1968.6, "total_tokens": 2482488}
{"current_steps": 350, "total_steps": 3400, "loss": 0.7052, "lr": 9.923568892600578e-05, "epoch": 0.04287770665523261, "percentage": 10.29, "elapsed_time": "0:22:00", "remaining_time": "3:11:43", "throughput": 1905.98, "total_tokens": 2516008}
{"current_steps": 350, "total_steps": 3400, "eval_loss": 0.8991873264312744, "epoch": 0.04287770665523261, "percentage": 10.29, "elapsed_time": "0:22:19", "remaining_time": "3:14:29", "throughput": 1878.79, "total_tokens": 2516008}
{"current_steps": 355, "total_steps": 3400, "loss": 0.7688, "lr": 9.91927536630402e-05, "epoch": 0.04349024532173593, "percentage": 10.44, "elapsed_time": "0:23:23", "remaining_time": "3:20:42", "throughput": 1816.75, "total_tokens": 2550696}
{"current_steps": 360, "total_steps": 3400, "loss": 0.6736, "lr": 9.91486549841951e-05, "epoch": 0.04410278398823926, "percentage": 10.59, "elapsed_time": "0:24:23", "remaining_time": "3:25:54", "throughput": 1766.49, "total_tokens": 2584440}
{"current_steps": 365, "total_steps": 3400, "loss": 0.6963, "lr": 9.91033939324107e-05, "epoch": 0.04471532265474258, "percentage": 10.74, "elapsed_time": "0:25:21", "remaining_time": "3:30:50", "throughput": 1720.84, "total_tokens": 2617984}
{"current_steps": 370, "total_steps": 3400, "loss": 0.7171, "lr": 9.905697157811761e-05, "epoch": 0.0453278613212459, "percentage": 10.88, "elapsed_time": "0:26:20", "remaining_time": "3:35:46", "throughput": 1677.08, "total_tokens": 2651328}
{"current_steps": 375, "total_steps": 3400, "loss": 0.6854, "lr": 9.900938901921131e-05, "epoch": 0.04594039998774923, "percentage": 11.03, "elapsed_time": "0:27:20", "remaining_time": "3:40:29", "throughput": 1637.0, "total_tokens": 2684728}
{"current_steps": 380, "total_steps": 3400, "loss": 0.7017, "lr": 9.896064738102635e-05, "epoch": 0.04655293865425255, "percentage": 11.18, "elapsed_time": "0:28:18", "remaining_time": "3:45:00", "throughput": 1600.27, "total_tokens": 2718408}
{"current_steps": 385, "total_steps": 3400, "loss": 0.7372, "lr": 9.891074781630966e-05, "epoch": 0.04716547732075587, "percentage": 11.32, "elapsed_time": "0:29:18", "remaining_time": "3:49:32", "throughput": 1565.28, "total_tokens": 2752744}
{"current_steps": 390, "total_steps": 3400, "loss": 0.6381, "lr": 9.885969150519331e-05, "epoch": 0.0477780159872592, "percentage": 11.47, "elapsed_time": "0:30:17", "remaining_time": "3:53:44", "throughput": 1533.01, "total_tokens": 2785664}
{"current_steps": 395, "total_steps": 3400, "loss": 0.674, "lr": 9.88074796551666e-05, "epoch": 0.04839055465376252, "percentage": 11.62, "elapsed_time": "0:31:15", "remaining_time": "3:57:51", "throughput": 1502.36, "total_tokens": 2818336}
{"current_steps": 400, "total_steps": 3400, "loss": 0.709, "lr": 9.875411350104744e-05, "epoch": 0.04900309332026584, "percentage": 11.76, "elapsed_time": "0:32:16", "remaining_time": "4:02:01", "throughput": 1473.03, "total_tokens": 2852080}
{"current_steps": 400, "total_steps": 3400, "eval_loss": 0.8526390790939331, "epoch": 0.04900309332026584, "percentage": 11.76, "elapsed_time": "0:32:35", "remaining_time": "4:04:27", "throughput": 1458.41, "total_tokens": 2852080}
{"current_steps": 405, "total_steps": 3400, "loss": 0.6803, "lr": 9.86995943049533e-05, "epoch": 0.04961563198676917, "percentage": 11.91, "elapsed_time": "0:33:39", "remaining_time": "4:08:57", "throughput": 1428.47, "total_tokens": 2885472}
{"current_steps": 410, "total_steps": 3400, "loss": 0.7016, "lr": 9.864392335627117e-05, "epoch": 0.05022817065327249, "percentage": 12.06, "elapsed_time": "0:34:38", "remaining_time": "4:12:39", "throughput": 1404.19, "total_tokens": 2918920}
{"current_steps": 415, "total_steps": 3400, "loss": 0.6904, "lr": 9.858710197162721e-05, "epoch": 0.05084070931977581, "percentage": 12.21, "elapsed_time": "0:35:38", "remaining_time": "4:16:23", "throughput": 1380.43, "total_tokens": 2952360}
{"current_steps": 420, "total_steps": 3400, "loss": 0.6498, "lr": 9.852913149485556e-05, "epoch": 0.051453247986279135, "percentage": 12.35, "elapsed_time": "0:36:37", "remaining_time": "4:19:49", "throughput": 1358.75, "total_tokens": 2985496}
{"current_steps": 425, "total_steps": 3400, "loss": 0.7565, "lr": 9.847001329696653e-05, "epoch": 0.052065786652782456, "percentage": 12.5, "elapsed_time": "0:37:35", "remaining_time": "4:23:09", "throughput": 1338.52, "total_tokens": 3019296}
{"current_steps": 430, "total_steps": 3400, "loss": 0.6492, "lr": 9.840974877611422e-05, "epoch": 0.05267832531928578, "percentage": 12.65, "elapsed_time": "0:38:35", "remaining_time": "4:26:32", "throughput": 1318.28, "total_tokens": 3052440}
{"current_steps": 435, "total_steps": 3400, "loss": 0.6917, "lr": 9.834833935756344e-05, "epoch": 0.053290863985789104, "percentage": 12.79, "elapsed_time": "0:39:34", "remaining_time": "4:29:43", "throughput": 1299.73, "total_tokens": 3085888}
{"current_steps": 440, "total_steps": 3400, "loss": 0.6842, "lr": 9.828578649365601e-05, "epoch": 0.053903402652292425, "percentage": 12.94, "elapsed_time": "0:40:32", "remaining_time": "4:32:46", "throughput": 1282.7, "total_tokens": 3120584}
{"current_steps": 445, "total_steps": 3400, "loss": 0.6519, "lr": 9.822209166377635e-05, "epoch": 0.05451594131879575, "percentage": 13.09, "elapsed_time": "0:41:32", "remaining_time": "4:35:48", "throughput": 1265.75, "total_tokens": 3154432}
{"current_steps": 450, "total_steps": 3400, "loss": 0.7008, "lr": 9.815725637431662e-05, "epoch": 0.05512847998529907, "percentage": 13.24, "elapsed_time": "0:42:31", "remaining_time": "4:38:45", "throughput": 1249.35, "total_tokens": 3187536}
{"current_steps": 450, "total_steps": 3400, "eval_loss": 0.8276960253715515, "epoch": 0.05512847998529907, "percentage": 13.24, "elapsed_time": "0:42:50", "remaining_time": "4:40:51", "throughput": 1240.0, "total_tokens": 3187536}
{"current_steps": 455, "total_steps": 3400, "loss": 0.6618, "lr": 9.809128215864097e-05, "epoch": 0.05574101865180239, "percentage": 13.38, "elapsed_time": "0:43:55", "remaining_time": "4:44:15", "throughput": 1222.34, "total_tokens": 3220896}
{"current_steps": 460, "total_steps": 3400, "loss": 0.6723, "lr": 9.802417057704931e-05, "epoch": 0.05635355731830572, "percentage": 13.53, "elapsed_time": "0:44:54", "remaining_time": "4:47:01", "throughput": 1207.76, "total_tokens": 3254424}
{"current_steps": 465, "total_steps": 3400, "loss": 0.6156, "lr": 9.795592321674045e-05, "epoch": 0.05696609598480904, "percentage": 13.68, "elapsed_time": "0:45:53", "remaining_time": "4:49:42", "throughput": 1193.8, "total_tokens": 3287624}
{"current_steps": 470, "total_steps": 3400, "loss": 0.664, "lr": 9.788654169177453e-05, "epoch": 0.05757863465131236, "percentage": 13.82, "elapsed_time": "0:46:52", "remaining_time": "4:52:15", "throughput": 1180.64, "total_tokens": 3321048}
{"current_steps": 475, "total_steps": 3400, "loss": 0.6419, "lr": 9.781602764303487e-05, "epoch": 0.05819117331781569, "percentage": 13.97, "elapsed_time": "0:47:51", "remaining_time": "4:54:44", "throughput": 1167.99, "total_tokens": 3354200}
{"current_steps": 480, "total_steps": 3400, "loss": 0.7027, "lr": 9.774438273818911e-05, "epoch": 0.05880371198431901, "percentage": 14.12, "elapsed_time": "0:48:51", "remaining_time": "4:57:15", "throughput": 1155.19, "total_tokens": 3386816}
{"current_steps": 485, "total_steps": 3400, "loss": 0.6849, "lr": 9.767160867164979e-05, "epoch": 0.05941625065082233, "percentage": 14.26, "elapsed_time": "0:49:50", "remaining_time": "4:59:32", "throughput": 1143.85, "total_tokens": 3420400}
{"current_steps": 490, "total_steps": 3400, "loss": 0.6643, "lr": 9.759770716453436e-05, "epoch": 0.06002878931732566, "percentage": 14.41, "elapsed_time": "0:50:48", "remaining_time": "5:01:44", "throughput": 1133.05, "total_tokens": 3454056}
{"current_steps": 495, "total_steps": 3400, "loss": 0.6659, "lr": 9.752267996462434e-05, "epoch": 0.06064132798382898, "percentage": 14.56, "elapsed_time": "0:51:47", "remaining_time": "5:03:59", "throughput": 1122.31, "total_tokens": 3488072}
{"current_steps": 500, "total_steps": 3400, "loss": 0.6225, "lr": 9.744652884632406e-05, "epoch": 0.0612538666503323, "percentage": 14.71, "elapsed_time": "0:52:46", "remaining_time": "5:06:03", "throughput": 1112.3, "total_tokens": 3521680}
{"current_steps": 500, "total_steps": 3400, "eval_loss": 0.8026401996612549, "epoch": 0.0612538666503323, "percentage": 14.71, "elapsed_time": "0:53:05", "remaining_time": "5:07:55", "throughput": 1105.58, "total_tokens": 3521680}
{"current_steps": 505, "total_steps": 3400, "loss": 0.6372, "lr": 9.736925561061871e-05, "epoch": 0.061866405316835626, "percentage": 14.85, "elapsed_time": "0:54:09", "remaining_time": "5:10:28", "throughput": 1094.12, "total_tokens": 3555288}
{"current_steps": 510, "total_steps": 3400, "loss": 0.68, "lr": 9.729086208503174e-05, "epoch": 0.06247894398333895, "percentage": 15.0, "elapsed_time": "0:55:09", "remaining_time": "5:12:31", "throughput": 1084.66, "total_tokens": 3589264}
{"current_steps": 515, "total_steps": 3400, "loss": 0.6723, "lr": 9.721135012358156e-05, "epoch": 0.06309148264984227, "percentage": 15.15, "elapsed_time": "0:56:07", "remaining_time": "5:14:24", "throughput": 1075.84, "total_tokens": 3622936}
{"current_steps": 520, "total_steps": 3400, "loss": 0.6608, "lr": 9.713072160673777e-05, "epoch": 0.06370402131634559, "percentage": 15.29, "elapsed_time": "0:57:06", "remaining_time": "5:16:15", "throughput": 1067.3, "total_tokens": 3656688}
{"current_steps": 525, "total_steps": 3400, "loss": 0.6821, "lr": 9.704897844137673e-05, "epoch": 0.06431655998284892, "percentage": 15.44, "elapsed_time": "0:58:05", "remaining_time": "5:18:08", "throughput": 1058.93, "total_tokens": 3691056}
{"current_steps": 530, "total_steps": 3400, "loss": 0.611, "lr": 9.696612256073633e-05, "epoch": 0.06492909864935224, "percentage": 15.59, "elapsed_time": "0:59:04", "remaining_time": "5:19:51", "throughput": 1051.03, "total_tokens": 3724952}
{"current_steps": 535, "total_steps": 3400, "loss": 0.6804, "lr": 9.688215592437039e-05, "epoch": 0.06554163731585556, "percentage": 15.74, "elapsed_time": "1:00:02", "remaining_time": "5:21:32", "throughput": 1043.22, "total_tokens": 3758240}
{"current_steps": 540, "total_steps": 3400, "loss": 0.634, "lr": 9.679708051810221e-05, "epoch": 0.06615417598235888, "percentage": 15.88, "elapsed_time": "1:01:02", "remaining_time": "5:23:15", "throughput": 1035.36, "total_tokens": 3791608}
{"current_steps": 545, "total_steps": 3400, "loss": 0.6331, "lr": 9.67108983539777e-05, "epoch": 0.0667667146488622, "percentage": 16.03, "elapsed_time": "1:02:00", "remaining_time": "5:24:51", "throughput": 1028.07, "total_tokens": 3825168}
{"current_steps": 550, "total_steps": 3400, "loss": 0.6115, "lr": 9.662361147021779e-05, "epoch": 0.06737925331536554, "percentage": 16.18, "elapsed_time": "1:02:59", "remaining_time": "5:26:25", "throughput": 1020.92, "total_tokens": 3858752}
{"current_steps": 550, "total_steps": 3400, "eval_loss": 0.8088525533676147, "epoch": 0.06737925331536554, "percentage": 16.18, "elapsed_time": "1:03:19", "remaining_time": "5:28:05", "throughput": 1015.72, "total_tokens": 3858752}
{"current_steps": 555, "total_steps": 3400, "loss": 0.6279, "lr": 9.653522193117013e-05, "epoch": 0.06799179198186886, "percentage": 16.32, "elapsed_time": "1:04:24", "remaining_time": "5:30:09", "throughput": 1007.1, "total_tokens": 3891816}
{"current_steps": 560, "total_steps": 3400, "loss": 0.668, "lr": 9.644573182726035e-05, "epoch": 0.06860433064837218, "percentage": 16.47, "elapsed_time": "1:05:22", "remaining_time": "5:31:35", "throughput": 1000.62, "total_tokens": 3925384}
{"current_steps": 565, "total_steps": 3400, "loss": 0.6761, "lr": 9.63551432749426e-05, "epoch": 0.0692168693148755, "percentage": 16.62, "elapsed_time": "1:06:21", "remaining_time": "5:32:59", "throughput": 994.35, "total_tokens": 3959232}
{"current_steps": 570, "total_steps": 3400, "loss": 0.5765, "lr": 9.626345841664953e-05, "epoch": 0.06982940798137882, "percentage": 16.76, "elapsed_time": "1:07:21", "remaining_time": "5:34:24", "throughput": 987.99, "total_tokens": 3992768}
{"current_steps": 575, "total_steps": 3400, "loss": 0.649, "lr": 9.617067942074153e-05, "epoch": 0.07044194664788214, "percentage": 16.91, "elapsed_time": "1:08:20", "remaining_time": "5:35:45", "throughput": 982.05, "total_tokens": 4026720}
{"current_steps": 580, "total_steps": 3400, "loss": 0.634, "lr": 9.607680848145558e-05, "epoch": 0.07105448531438548, "percentage": 17.06, "elapsed_time": "1:09:19", "remaining_time": "5:37:02", "throughput": 976.26, "total_tokens": 4060536}
{"current_steps": 585, "total_steps": 3400, "loss": 0.6074, "lr": 9.598184781885318e-05, "epoch": 0.0716670239808888, "percentage": 17.21, "elapsed_time": "1:10:19", "remaining_time": "5:38:22", "throughput": 970.44, "total_tokens": 4094408}
{"current_steps": 590, "total_steps": 3400, "loss": 0.6965, "lr": 9.588579967876806e-05, "epoch": 0.07227956264739212, "percentage": 17.35, "elapsed_time": "1:11:17", "remaining_time": "5:39:34", "throughput": 965.11, "total_tokens": 4128600}
{"current_steps": 595, "total_steps": 3400, "loss": 0.6751, "lr": 9.578866633275288e-05, "epoch": 0.07289210131389544, "percentage": 17.5, "elapsed_time": "1:12:16", "remaining_time": "5:40:44", "throughput": 959.75, "total_tokens": 4162136}
{"current_steps": 600, "total_steps": 3400, "loss": 0.6367, "lr": 9.569045007802559e-05, "epoch": 0.07350463998039876, "percentage": 17.65, "elapsed_time": "1:13:17", "remaining_time": "5:42:00", "throughput": 954.15, "total_tokens": 4195632}
{"current_steps": 600, "total_steps": 3400, "eval_loss": 0.6892650723457336, "epoch": 0.07350463998039876, "percentage": 17.65, "elapsed_time": "1:13:36", "remaining_time": "5:43:31", "throughput": 949.93, "total_tokens": 4195632}
{"current_steps": 605, "total_steps": 3400, "loss": 0.5934, "lr": 9.55911532374151e-05, "epoch": 0.07411717864690208, "percentage": 17.79, "elapsed_time": "1:14:41", "remaining_time": "5:45:03", "throughput": 943.49, "total_tokens": 4228296}
{"current_steps": 610, "total_steps": 3400, "loss": 0.5862, "lr": 9.549077815930636e-05, "epoch": 0.07472971731340541, "percentage": 17.94, "elapsed_time": "1:15:40", "remaining_time": "5:46:06", "throughput": 938.62, "total_tokens": 4261592}
{"current_steps": 615, "total_steps": 3400, "loss": 0.6037, "lr": 9.538932721758474e-05, "epoch": 0.07534225597990873, "percentage": 18.09, "elapsed_time": "1:16:38", "remaining_time": "5:47:04", "throughput": 934.01, "total_tokens": 4295064}
{"current_steps": 620, "total_steps": 3400, "loss": 0.6898, "lr": 9.528680281157999e-05, "epoch": 0.07595479464641205, "percentage": 18.24, "elapsed_time": "1:17:38", "remaining_time": "5:48:08", "throughput": 929.15, "total_tokens": 4328512}
{"current_steps": 625, "total_steps": 3400, "loss": 0.6429, "lr": 9.518320736600943e-05, "epoch": 0.07656733331291538, "percentage": 18.38, "elapsed_time": "1:18:37", "remaining_time": "5:49:05", "throughput": 924.71, "total_tokens": 4362280}
{"current_steps": 630, "total_steps": 3400, "loss": 0.6793, "lr": 9.507854333092063e-05, "epoch": 0.0771798719794187, "percentage": 18.53, "elapsed_time": "1:19:36", "remaining_time": "5:50:01", "throughput": 920.48, "total_tokens": 4396624}
{"current_steps": 635, "total_steps": 3400, "loss": 0.6721, "lr": 9.497281318163346e-05, "epoch": 0.07779241064592203, "percentage": 18.68, "elapsed_time": "1:20:36", "remaining_time": "5:51:00", "throughput": 916.0, "total_tokens": 4430344}
{"current_steps": 640, "total_steps": 3400, "loss": 0.607, "lr": 9.486601941868154e-05, "epoch": 0.07840494931242535, "percentage": 18.82, "elapsed_time": "1:21:35", "remaining_time": "5:51:52", "throughput": 911.72, "total_tokens": 4463560}
{"current_steps": 645, "total_steps": 3400, "loss": 0.6382, "lr": 9.475816456775313e-05, "epoch": 0.07901748797892867, "percentage": 18.97, "elapsed_time": "1:22:34", "remaining_time": "5:52:43", "throughput": 907.59, "total_tokens": 4496896}
{"current_steps": 650, "total_steps": 3400, "loss": 0.6238, "lr": 9.464925117963133e-05, "epoch": 0.07963002664543199, "percentage": 19.12, "elapsed_time": "1:23:35", "remaining_time": "5:53:38", "throughput": 903.42, "total_tokens": 4530824}
{"current_steps": 650, "total_steps": 3400, "eval_loss": 0.7062045335769653, "epoch": 0.07963002664543199, "percentage": 19.12, "elapsed_time": "1:23:54", "remaining_time": "5:55:00", "throughput": 899.92, "total_tokens": 4530824}
{"current_steps": 655, "total_steps": 3400, "loss": 0.6077, "lr": 9.453928183013385e-05, "epoch": 0.08024256531193531, "percentage": 19.26, "elapsed_time": "1:24:58", "remaining_time": "5:56:09", "throughput": 895.19, "total_tokens": 4564536}
{"current_steps": 660, "total_steps": 3400, "loss": 0.6365, "lr": 9.442825912005202e-05, "epoch": 0.08085510397843863, "percentage": 19.41, "elapsed_time": "1:25:58", "remaining_time": "5:56:54", "throughput": 891.4, "total_tokens": 4598088}
{"current_steps": 665, "total_steps": 3400, "loss": 0.6289, "lr": 9.431618567508933e-05, "epoch": 0.08146764264494197, "percentage": 19.56, "elapsed_time": "1:26:59", "remaining_time": "5:57:44", "throughput": 887.51, "total_tokens": 4631952}
{"current_steps": 670, "total_steps": 3400, "loss": 0.5946, "lr": 9.420306414579925e-05, "epoch": 0.08208018131144529, "percentage": 19.71, "elapsed_time": "1:27:58", "remaining_time": "5:58:28", "throughput": 883.87, "total_tokens": 4665632}
{"current_steps": 675, "total_steps": 3400, "loss": 0.596, "lr": 9.408889720752266e-05, "epoch": 0.08269271997794861, "percentage": 19.85, "elapsed_time": "1:28:58", "remaining_time": "5:59:13", "throughput": 880.19, "total_tokens": 4699240}
{"current_steps": 680, "total_steps": 3400, "loss": 0.6328, "lr": 9.397368756032445e-05, "epoch": 0.08330525864445193, "percentage": 20.0, "elapsed_time": "1:30:00", "remaining_time": "6:00:01", "throughput": 876.32, "total_tokens": 4732400}
{"current_steps": 685, "total_steps": 3400, "loss": 0.6424, "lr": 9.385743792892982e-05, "epoch": 0.08391779731095525, "percentage": 20.15, "elapsed_time": "1:31:00", "remaining_time": "6:00:42", "throughput": 872.9, "total_tokens": 4766448}
{"current_steps": 690, "total_steps": 3400, "loss": 0.6137, "lr": 9.374015106265968e-05, "epoch": 0.08453033597745857, "percentage": 20.29, "elapsed_time": "1:31:59", "remaining_time": "6:01:19", "throughput": 869.64, "total_tokens": 4800344}
{"current_steps": 695, "total_steps": 3400, "loss": 0.6233, "lr": 9.362182973536569e-05, "epoch": 0.0851428746439619, "percentage": 20.44, "elapsed_time": "1:32:59", "remaining_time": "6:01:57", "throughput": 866.17, "total_tokens": 4833096}
{"current_steps": 700, "total_steps": 3400, "loss": 0.6897, "lr": 9.35024767453647e-05, "epoch": 0.08575541331046523, "percentage": 20.59, "elapsed_time": "1:33:59", "remaining_time": "6:02:33", "throughput": 862.97, "total_tokens": 4867048}
{"current_steps": 700, "total_steps": 3400, "eval_loss": 0.6905214190483093, "epoch": 0.08575541331046523, "percentage": 20.59, "elapsed_time": "1:34:19", "remaining_time": "6:03:48", "throughput": 860.02, "total_tokens": 4867048}
{"current_steps": 705, "total_steps": 3400, "loss": 0.6004, "lr": 9.338209491537257e-05, "epoch": 0.08636795197696855, "percentage": 20.74, "elapsed_time": "1:35:23", "remaining_time": "6:04:40", "throughput": 856.19, "total_tokens": 4900696}
{"current_steps": 710, "total_steps": 3400, "loss": 0.6088, "lr": 9.326068709243727e-05, "epoch": 0.08698049064347187, "percentage": 20.88, "elapsed_time": "1:36:24", "remaining_time": "6:05:16", "throughput": 853.01, "total_tokens": 4934312}
{"current_steps": 715, "total_steps": 3400, "loss": 0.6589, "lr": 9.313825614787177e-05, "epoch": 0.08759302930997519, "percentage": 21.03, "elapsed_time": "1:37:24", "remaining_time": "6:05:47", "throughput": 849.89, "total_tokens": 4967176}
{"current_steps": 720, "total_steps": 3400, "loss": 0.6056, "lr": 9.301480497718593e-05, "epoch": 0.08820556797647852, "percentage": 21.18, "elapsed_time": "1:38:23", "remaining_time": "6:06:15", "throughput": 847.12, "total_tokens": 5001336}
{"current_steps": 725, "total_steps": 3400, "loss": 0.6026, "lr": 9.289033650001817e-05, "epoch": 0.08881810664298184, "percentage": 21.32, "elapsed_time": "1:39:24", "remaining_time": "6:06:46", "throughput": 844.24, "total_tokens": 5035240}
{"current_steps": 730, "total_steps": 3400, "loss": 0.5947, "lr": 9.276485366006634e-05, "epoch": 0.08943064530948516, "percentage": 21.47, "elapsed_time": "1:40:23", "remaining_time": "6:07:12", "throughput": 841.41, "total_tokens": 5068640}
{"current_steps": 735, "total_steps": 3400, "loss": 0.6654, "lr": 9.263835942501807e-05, "epoch": 0.09004318397598848, "percentage": 21.62, "elapsed_time": "1:41:23", "remaining_time": "6:07:37", "throughput": 838.84, "total_tokens": 5102912}
{"current_steps": 740, "total_steps": 3400, "loss": 0.6167, "lr": 9.251085678648072e-05, "epoch": 0.0906557226424918, "percentage": 21.76, "elapsed_time": "1:42:23", "remaining_time": "6:08:02", "throughput": 836.06, "total_tokens": 5136064}
{"current_steps": 705, "total_steps": 3400, "loss": 0.6004, "lr": 9.338209491537257e-05, "epoch": 0.08636795197696855, "percentage": 20.74, "elapsed_time": "0:02:20", "remaining_time": "0:08:55", "throughput": 34972.79, "total_tokens": 4900696}
{"current_steps": 710, "total_steps": 3400, "loss": 0.6088, "lr": 9.326068709243727e-05, "epoch": 0.08698049064347187, "percentage": 20.88, "elapsed_time": "0:03:20", "remaining_time": "0:12:41", "throughput": 24555.92, "total_tokens": 4934312}
{"current_steps": 715, "total_steps": 3400, "loss": 0.6589, "lr": 9.313825614787177e-05, "epoch": 0.08759302930997519, "percentage": 21.03, "elapsed_time": "0:04:21", "remaining_time": "0:16:20", "throughput": 19028.17, "total_tokens": 4967176}
{"current_steps": 720, "total_steps": 3400, "loss": 0.6056, "lr": 9.301480497718593e-05, "epoch": 0.08820556797647852, "percentage": 21.18, "elapsed_time": "0:05:21", "remaining_time": "0:19:56", "throughput": 15559.58, "total_tokens": 5001336}
{"current_steps": 725, "total_steps": 3400, "loss": 0.6026, "lr": 9.289033650001817e-05, "epoch": 0.08881810664298184, "percentage": 21.32, "elapsed_time": "0:06:22", "remaining_time": "0:23:30", "throughput": 13169.44, "total_tokens": 5035240}
{"current_steps": 730, "total_steps": 3400, "loss": 0.5947, "lr": 9.276485366006634e-05, "epoch": 0.08943064530948516, "percentage": 21.47, "elapsed_time": "0:07:22", "remaining_time": "0:26:58", "throughput": 11452.89, "total_tokens": 5068640}
{"current_steps": 735, "total_steps": 3400, "loss": 0.6654, "lr": 9.263835942501807e-05, "epoch": 0.09004318397598848, "percentage": 21.62, "elapsed_time": "0:08:21", "remaining_time": "0:30:19", "throughput": 10166.29, "total_tokens": 5102912}
{"current_steps": 740, "total_steps": 3400, "loss": 0.6167, "lr": 9.251085678648072e-05, "epoch": 0.0906557226424918, "percentage": 21.76, "elapsed_time": "0:09:21", "remaining_time": "0:33:39", "throughput": 9144.16, "total_tokens": 5136064}
{"current_steps": 745, "total_steps": 3400, "loss": 0.5852, "lr": 9.238234875991046e-05, "epoch": 0.09126826130899512, "percentage": 21.91, "elapsed_time": "0:10:21", "remaining_time": "0:36:54", "throughput": 8317.85, "total_tokens": 5169640}
{"current_steps": 750, "total_steps": 3400, "loss": 0.6515, "lr": 9.225283838454111e-05, "epoch": 0.09188079997549846, "percentage": 22.06, "elapsed_time": "0:11:21", "remaining_time": "0:40:06", "throughput": 7639.11, "total_tokens": 5203408}
{"current_steps": 750, "total_steps": 3400, "eval_loss": 0.7262604832649231, "epoch": 0.09188079997549846, "percentage": 22.06, "elapsed_time": "0:12:09", "remaining_time": "0:42:56", "throughput": 7136.44, "total_tokens": 5203408}
{"current_steps": 755, "total_steps": 3400, "loss": 0.6032, "lr": 9.21223287233121e-05, "epoch": 0.09249333864200178, "percentage": 22.21, "elapsed_time": "0:13:17", "remaining_time": "0:46:34", "throughput": 6564.88, "total_tokens": 5237296}
{"current_steps": 760, "total_steps": 3400, "loss": 0.5835, "lr": 9.199082286279622e-05, "epoch": 0.0931058773085051, "percentage": 22.35, "elapsed_time": "0:14:17", "remaining_time": "0:49:39", "throughput": 6145.51, "total_tokens": 5270792}
{"current_steps": 765, "total_steps": 3400, "loss": 0.5838, "lr": 9.185832391312644e-05, "epoch": 0.09371841597500842, "percentage": 22.5, "elapsed_time": "0:15:16", "remaining_time": "0:52:38", "throughput": 5785.18, "total_tokens": 5304352}
{"current_steps": 770, "total_steps": 3400, "loss": 0.5733, "lr": 9.172483500792244e-05, "epoch": 0.09433095464151174, "percentage": 22.65, "elapsed_time": "0:16:16", "remaining_time": "0:55:35", "throughput": 5465.97, "total_tokens": 5338544}
{"current_steps": 775, "total_steps": 3400, "loss": 0.6251, "lr": 9.159035930421658e-05, "epoch": 0.09494349330801506, "percentage": 22.79, "elapsed_time": "0:17:16", "remaining_time": "0:58:30", "throughput": 5183.23, "total_tokens": 5371568}
{"current_steps": 780, "total_steps": 3400, "loss": 0.6165, "lr": 9.145489998237902e-05, "epoch": 0.0955560319745184, "percentage": 22.94, "elapsed_time": "0:18:14", "remaining_time": "1:01:17", "throughput": 4936.62, "total_tokens": 5405472}
{"current_steps": 785, "total_steps": 3400, "loss": 0.5865, "lr": 9.131846024604274e-05, "epoch": 0.09616857064102172, "percentage": 23.09, "elapsed_time": "0:19:13", "remaining_time": "1:04:03", "throughput": 4713.29, "total_tokens": 5438600}
{"current_steps": 790, "total_steps": 3400, "loss": 0.5387, "lr": 9.11810433220276e-05, "epoch": 0.09678110930752504, "percentage": 23.24, "elapsed_time": "0:20:13", "remaining_time": "1:06:49", "throughput": 4509.53, "total_tokens": 5472624}
{"current_steps": 795, "total_steps": 3400, "loss": 0.576, "lr": 9.104265246026415e-05, "epoch": 0.09739364797402836, "percentage": 23.38, "elapsed_time": "0:21:12", "remaining_time": "1:09:28", "throughput": 4328.99, "total_tokens": 5506720}
{"current_steps": 800, "total_steps": 3400, "loss": 0.6221, "lr": 9.090329093371666e-05, "epoch": 0.09800618664053168, "percentage": 23.53, "elapsed_time": "0:22:10", "remaining_time": "1:12:03", "throughput": 4164.69, "total_tokens": 5540920}
{"current_steps": 800, "total_steps": 3400, "eval_loss": 0.6957933306694031, "epoch": 0.09800618664053168, "percentage": 23.53, "elapsed_time": "0:22:29", "remaining_time": "1:13:07", "throughput": 4104.75, "total_tokens": 5540920}
{"current_steps": 805, "total_steps": 3400, "loss": 0.5566, "lr": 9.076296203830579e-05, "epoch": 0.09861872530703501, "percentage": 23.68, "elapsed_time": "0:23:34", "remaining_time": "1:16:00", "throughput": 3939.8, "total_tokens": 5574216}
{"current_steps": 810, "total_steps": 3400, "loss": 0.5927, "lr": 9.062166909283062e-05, "epoch": 0.09923126397353833, "percentage": 23.82, "elapsed_time": "0:24:34", "remaining_time": "1:18:33", "throughput": 3804.18, "total_tokens": 5607720}
{"current_steps": 815, "total_steps": 3400, "loss": 0.6153, "lr": 9.047941543889014e-05, "epoch": 0.09984380264004165, "percentage": 23.97, "elapsed_time": "0:25:33", "remaining_time": "1:21:04", "throughput": 3678.58, "total_tokens": 5641192}
{"current_steps": 820, "total_steps": 3400, "loss": 0.5996, "lr": 9.033620444080428e-05, "epoch": 0.10045634130654497, "percentage": 24.12, "elapsed_time": "0:26:34", "remaining_time": "1:23:36", "throughput": 3559.5, "total_tokens": 5675392}
{"current_steps": 825, "total_steps": 3400, "loss": 0.691, "lr": 9.019203948553422e-05, "epoch": 0.1010688799730483, "percentage": 24.26, "elapsed_time": "0:27:33", "remaining_time": "1:26:01", "throughput": 3452.49, "total_tokens": 5709464}
{"current_steps": 830, "total_steps": 3400, "loss": 0.5388, "lr": 9.004692398260244e-05, "epoch": 0.10168141863955162, "percentage": 24.41, "elapsed_time": "0:28:33", "remaining_time": "1:28:24", "throughput": 3352.22, "total_tokens": 5742464}
{"current_steps": 835, "total_steps": 3400, "loss": 0.5515, "lr": 8.9900861364012e-05, "epoch": 0.10229395730605495, "percentage": 24.56, "elapsed_time": "0:29:33", "remaining_time": "1:30:49", "throughput": 3255.89, "total_tokens": 5775832}
{"current_steps": 840, "total_steps": 3400, "loss": 0.5835, "lr": 8.975385508416532e-05, "epoch": 0.10290649597255827, "percentage": 24.71, "elapsed_time": "0:30:33", "remaining_time": "1:33:08", "throughput": 3168.05, "total_tokens": 5809264}
{"current_steps": 845, "total_steps": 3400, "loss": 0.6318, "lr": 8.960590861978265e-05, "epoch": 0.10351903463906159, "percentage": 24.85, "elapsed_time": "0:31:33", "remaining_time": "1:35:24", "throughput": 3086.45, "total_tokens": 5842968}
{"current_steps": 805, "total_steps": 3400, "loss": 0.5566, "lr": 9.076296203830579e-05, "epoch": 0.09861872530703501, "percentage": 23.68, "elapsed_time": "0:02:19", "remaining_time": "0:07:28", "throughput": 40047.9, "total_tokens": 5574216}
{"current_steps": 810, "total_steps": 3400, "loss": 0.5927, "lr": 9.062166909283062e-05, "epoch": 0.09923126397353833, "percentage": 23.82, "elapsed_time": "0:03:19", "remaining_time": "0:10:36", "throughput": 28149.75, "total_tokens": 5607720}
{"current_steps": 815, "total_steps": 3400, "loss": 0.6153, "lr": 9.047941543889014e-05, "epoch": 0.09984380264004165, "percentage": 23.97, "elapsed_time": "0:04:18", "remaining_time": "0:13:40", "throughput": 21794.62, "total_tokens": 5641192}
{"current_steps": 820, "total_steps": 3400, "loss": 0.5996, "lr": 9.033620444080428e-05, "epoch": 0.10045634130654497, "percentage": 24.12, "elapsed_time": "0:05:18", "remaining_time": "0:16:41", "throughput": 17824.57, "total_tokens": 5675392}
{"current_steps": 825, "total_steps": 3400, "loss": 0.691, "lr": 9.019203948553422e-05, "epoch": 0.1010688799730483, "percentage": 24.26, "elapsed_time": "0:06:18", "remaining_time": "0:19:40", "throughput": 15092.68, "total_tokens": 5709464}
{"current_steps": 830, "total_steps": 3400, "loss": 0.5388, "lr": 9.004692398260244e-05, "epoch": 0.10168141863955162, "percentage": 24.41, "elapsed_time": "0:07:17", "remaining_time": "0:22:34", "throughput": 13128.49, "total_tokens": 5742464}
{"current_steps": 835, "total_steps": 3400, "loss": 0.5515, "lr": 8.9900861364012e-05, "epoch": 0.10229395730605495, "percentage": 24.56, "elapsed_time": "0:08:17", "remaining_time": "0:25:26", "throughput": 11620.95, "total_tokens": 5775832}
{"current_steps": 840, "total_steps": 3400, "loss": 0.5835, "lr": 8.975385508416532e-05, "epoch": 0.10290649597255827, "percentage": 24.71, "elapsed_time": "0:09:17", "remaining_time": "0:28:19", "throughput": 10418.89, "total_tokens": 5809264}
{"current_steps": 845, "total_steps": 3400, "loss": 0.6318, "lr": 8.960590861978265e-05, "epoch": 0.10351903463906159, "percentage": 24.85, "elapsed_time": "0:10:16", "remaining_time": "0:31:05", "throughput": 9470.76, "total_tokens": 5842968}
{"current_steps": 850, "total_steps": 3400, "loss": 0.6148, "lr": 8.945702546981969e-05, "epoch": 0.10413157330556491, "percentage": 25.0, "elapsed_time": "0:11:16", "remaining_time": "0:33:48", "throughput": 8691.13, "total_tokens": 5876600}
{"current_steps": 850, "total_steps": 3400, "eval_loss": 0.6981882452964783, "epoch": 0.10413157330556491, "percentage": 25.0, "elapsed_time": "0:12:03", "remaining_time": "0:36:10", "throughput": 8122.54, "total_tokens": 5876600}
{"current_steps": 855, "total_steps": 3400, "loss": 0.6409, "lr": 8.930720915538487e-05, "epoch": 0.10474411197206823, "percentage": 25.15, "elapsed_time": "0:13:11", "remaining_time": "0:39:15", "throughput": 7468.69, "total_tokens": 5910016}
{"current_steps": 860, "total_steps": 3400, "loss": 0.5827, "lr": 8.915646321965614e-05, "epoch": 0.10535665063857157, "percentage": 25.29, "elapsed_time": "0:14:10", "remaining_time": "0:41:52", "throughput": 6986.64, "total_tokens": 5943680}
{"current_steps": 865, "total_steps": 3400, "loss": 0.5722, "lr": 8.900479122779712e-05, "epoch": 0.10596918930507489, "percentage": 25.44, "elapsed_time": "0:15:09", "remaining_time": "0:44:26", "throughput": 6568.94, "total_tokens": 5977216}
{"current_steps": 870, "total_steps": 3400, "loss": 0.5395, "lr": 8.885219676687277e-05, "epoch": 0.10658172797157821, "percentage": 25.59, "elapsed_time": "0:16:09", "remaining_time": "0:47:00", "throughput": 6196.6, "total_tokens": 6010352}
{"current_steps": 875, "total_steps": 3400, "loss": 0.5898, "lr": 8.869868344576459e-05, "epoch": 0.10719426663808153, "percentage": 25.74, "elapsed_time": "0:17:09", "remaining_time": "0:49:32", "throughput": 5868.35, "total_tokens": 6044008}
{"current_steps": 880, "total_steps": 3400, "loss": 0.5175, "lr": 8.854425489508532e-05, "epoch": 0.10780680530458485, "percentage": 25.88, "elapsed_time": "0:18:09", "remaining_time": "0:51:58", "throughput": 5580.23, "total_tokens": 6077272}
{"current_steps": 885, "total_steps": 3400, "loss": 0.613, "lr": 8.838891476709288e-05, "epoch": 0.10841934397108817, "percentage": 26.03, "elapsed_time": "0:19:08", "remaining_time": "0:54:24", "throughput": 5320.02, "total_tokens": 6110864}
{"current_steps": 890, "total_steps": 3400, "loss": 0.61, "lr": 8.823266673560426e-05, "epoch": 0.1090318826375915, "percentage": 26.18, "elapsed_time": "0:20:09", "remaining_time": "0:56:49", "throughput": 5081.39, "total_tokens": 6143976}
{"current_steps": 895, "total_steps": 3400, "loss": 0.6117, "lr": 8.807551449590846e-05, "epoch": 0.10964442130409482, "percentage": 26.32, "elapsed_time": "0:21:08", "remaining_time": "0:59:10", "throughput": 4869.4, "total_tokens": 6176768}
{"current_steps": 900, "total_steps": 3400, "loss": 0.5434, "lr": 8.791746176467907e-05, "epoch": 0.11025695997059815, "percentage": 26.47, "elapsed_time": "0:22:07", "remaining_time": "1:01:28", "throughput": 4676.27, "total_tokens": 6209928}
{"current_steps": 900, "total_steps": 3400, "eval_loss": 0.6621683239936829, "epoch": 0.11025695997059815, "percentage": 26.47, "elapsed_time": "0:22:27", "remaining_time": "1:02:23", "throughput": 4607.83, "total_tokens": 6209928}
{"current_steps": 905, "total_steps": 3400, "loss": 0.5905, "lr": 8.775851227988656e-05, "epoch": 0.11086949863710147, "percentage": 26.62, "elapsed_time": "0:23:33", "remaining_time": "1:04:57", "throughput": 4415.92, "total_tokens": 6242384}
{"current_steps": 910, "total_steps": 3400, "loss": 0.5894, "lr": 8.759866980070963e-05, "epoch": 0.11148203730360479, "percentage": 26.76, "elapsed_time": "0:24:33", "remaining_time": "1:07:10", "throughput": 4260.33, "total_tokens": 6275768}
{"current_steps": 915, "total_steps": 3400, "loss": 0.5514, "lr": 8.743793810744654e-05, "epoch": 0.1120945759701081, "percentage": 26.91, "elapsed_time": "0:25:32", "remaining_time": "1:09:21", "throughput": 4118.21, "total_tokens": 6309936}
{"current_steps": 920, "total_steps": 3400, "loss": 0.5772, "lr": 8.727632100142551e-05, "epoch": 0.11270711463661144, "percentage": 27.06, "elapsed_time": "0:26:32", "remaining_time": "1:11:32", "throughput": 3983.68, "total_tokens": 6343768}
{"current_steps": 925, "total_steps": 3400, "loss": 0.6127, "lr": 8.711382230491493e-05, "epoch": 0.11331965330311476, "percentage": 27.21, "elapsed_time": "0:27:31", "remaining_time": "1:13:38", "throughput": 3862.07, "total_tokens": 6377040}
{"current_steps": 930, "total_steps": 3400, "loss": 0.5769, "lr": 8.695044586103296e-05, "epoch": 0.11393219196961808, "percentage": 27.35, "elapsed_time": "0:28:29", "remaining_time": "1:15:41", "throughput": 3749.02, "total_tokens": 6410232}
{"current_steps": 935, "total_steps": 3400, "loss": 0.5851, "lr": 8.678619553365659e-05, "epoch": 0.1145447306361214, "percentage": 27.5, "elapsed_time": "0:29:29", "remaining_time": "1:17:45", "throughput": 3640.92, "total_tokens": 6443920}
{"current_steps": 940, "total_steps": 3400, "loss": 0.4821, "lr": 8.662107520733027e-05, "epoch": 0.11515726930262472, "percentage": 27.65, "elapsed_time": "0:30:28", "remaining_time": "1:19:44", "throughput": 3542.41, "total_tokens": 6476464}
{"current_steps": 945, "total_steps": 3400, "loss": 0.5848, "lr": 8.64550887871741e-05, "epoch": 0.11576980796912806, "percentage": 27.79, "elapsed_time": "0:31:26", "remaining_time": "1:21:41", "throughput": 3450.77, "total_tokens": 6510856}
{"current_steps": 950, "total_steps": 3400, "loss": 0.5729, "lr": 8.628824019879137e-05, "epoch": 0.11638234663563138, "percentage": 27.94, "elapsed_time": "0:32:26", "remaining_time": "1:23:40", "throughput": 3362.15, "total_tokens": 6544656}
{"current_steps": 950, "total_steps": 3400, "eval_loss": 0.6741500496864319, "epoch": 0.11638234663563138, "percentage": 27.94, "elapsed_time": "0:32:45", "remaining_time": "1:24:29", "throughput": 3329.31, "total_tokens": 6544656}
{"current_steps": 955, "total_steps": 3400, "loss": 0.5924, "lr": 8.612053338817581e-05, "epoch": 0.1169948853021347, "percentage": 28.09, "elapsed_time": "0:33:49", "remaining_time": "1:26:37", "throughput": 3240.67, "total_tokens": 6578488}
{"current_steps": 960, "total_steps": 3400, "loss": 0.5946, "lr": 8.595197232161824e-05, "epoch": 0.11760742396863802, "percentage": 28.24, "elapsed_time": "0:34:48", "remaining_time": "1:28:28", "throughput": 3165.56, "total_tokens": 6611616}
{"current_steps": 965, "total_steps": 3400, "loss": 0.5692, "lr": 8.578256098561275e-05, "epoch": 0.11821996263514134, "percentage": 28.38, "elapsed_time": "0:35:48", "remaining_time": "1:30:22", "throughput": 3092.51, "total_tokens": 6645456}
{"current_steps": 970, "total_steps": 3400, "loss": 0.5347, "lr": 8.561230338676239e-05, "epoch": 0.11883250130164466, "percentage": 28.53, "elapsed_time": "0:36:47", "remaining_time": "1:32:10", "throughput": 3025.17, "total_tokens": 6678648}
{"current_steps": 975, "total_steps": 3400, "loss": 0.5792, "lr": 8.544120355168451e-05, "epoch": 0.119445039968148, "percentage": 28.68, "elapsed_time": "0:37:46", "remaining_time": "1:33:57", "throughput": 2961.76, "total_tokens": 6712928}
{"current_steps": 980, "total_steps": 3400, "loss": 0.5768, "lr": 8.526926552691544e-05, "epoch": 0.12005757863465132, "percentage": 28.82, "elapsed_time": "0:38:46", "remaining_time": "1:35:45", "throughput": 2899.47, "total_tokens": 6746616}
{"current_steps": 985, "total_steps": 3400, "loss": 0.53, "lr": 8.509649337881483e-05, "epoch": 0.12067011730115464, "percentage": 28.97, "elapsed_time": "0:39:45", "remaining_time": "1:37:29", "throughput": 2841.71, "total_tokens": 6779552}
{"current_steps": 990, "total_steps": 3400, "loss": 0.5095, "lr": 8.492289119346943e-05, "epoch": 0.12128265596765796, "percentage": 29.12, "elapsed_time": "0:40:44", "remaining_time": "1:39:11", "throughput": 2786.88, "total_tokens": 6812968}
{"current_steps": 995, "total_steps": 3400, "loss": 0.5278, "lr": 8.474846307659658e-05, "epoch": 0.12189519463416128, "percentage": 29.26, "elapsed_time": "0:41:44", "remaining_time": "1:40:53", "throughput": 2733.92, "total_tokens": 6846816}
{"current_steps": 1000, "total_steps": 3400, "loss": 0.55, "lr": 8.457321315344694e-05, "epoch": 0.1225077333006646, "percentage": 29.41, "elapsed_time": "0:42:44", "remaining_time": "1:42:33", "throughput": 2683.26, "total_tokens": 6880016}
{"current_steps": 1000, "total_steps": 3400, "eval_loss": 0.7234537601470947, "epoch": 0.1225077333006646, "percentage": 29.41, "elapsed_time": "0:43:03", "remaining_time": "1:43:20", "throughput": 2663.13, "total_tokens": 6880016}
{"current_steps": 1005, "total_steps": 3400, "loss": 0.5095, "lr": 8.439714556870704e-05, "epoch": 0.12312027196716793, "percentage": 29.56, "elapsed_time": "0:44:08", "remaining_time": "1:45:10", "throughput": 2610.48, "total_tokens": 6913088}
{"current_steps": 1010, "total_steps": 3400, "loss": 0.5987, "lr": 8.422026448640124e-05, "epoch": 0.12373281063367125, "percentage": 29.71, "elapsed_time": "0:45:07", "remaining_time": "1:46:48", "throughput": 2565.03, "total_tokens": 6946072}
{"current_steps": 1015, "total_steps": 3400, "loss": 0.5267, "lr": 8.40425740897932e-05, "epoch": 0.12434534930017457, "percentage": 29.85, "elapsed_time": "0:46:08", "remaining_time": "1:48:24", "throughput": 2521.14, "total_tokens": 6979400}
{"current_steps": 1020, "total_steps": 3400, "loss": 0.5209, "lr": 8.386407858128706e-05, "epoch": 0.1249578879666779, "percentage": 30.0, "elapsed_time": "0:47:08", "remaining_time": "1:49:58", "throughput": 2479.86, "total_tokens": 7013384}
{"current_steps": 1025, "total_steps": 3400, "loss": 0.5702, "lr": 8.368478218232787e-05, "epoch": 0.12557042663318121, "percentage": 30.15, "elapsed_time": "0:48:08", "remaining_time": "1:51:32", "throughput": 2439.93, "total_tokens": 7047256}
{"current_steps": 1030, "total_steps": 3400, "loss": 0.5883, "lr": 8.350468913330192e-05, "epoch": 0.12618296529968454, "percentage": 30.29, "elapsed_time": "0:49:08", "remaining_time": "1:53:05", "throughput": 2401.04, "total_tokens": 7080464}
{"current_steps": 1035, "total_steps": 3400, "loss": 0.6062, "lr": 8.33238036934364e-05, "epoch": 0.12679550396618786, "percentage": 30.44, "elapsed_time": "0:50:08", "remaining_time": "1:54:34", "throughput": 2364.98, "total_tokens": 7114864}
{"current_steps": 1040, "total_steps": 3400, "loss": 0.6236, "lr": 8.31421301406986e-05, "epoch": 0.12740804263269118, "percentage": 30.59, "elapsed_time": "0:51:07", "remaining_time": "1:56:01", "throughput": 2330.23, "total_tokens": 7148968}
{"current_steps": 1045, "total_steps": 3400, "loss": 0.6023, "lr": 8.29596727716949e-05, "epoch": 0.12802058129919452, "percentage": 30.74, "elapsed_time": "0:52:08", "remaining_time": "1:57:30", "throughput": 2295.64, "total_tokens": 7182000}
{"current_steps": 1050, "total_steps": 3400, "loss": 0.5312, "lr": 8.277643590156894e-05, "epoch": 0.12863311996569785, "percentage": 30.88, "elapsed_time": "0:53:07", "remaining_time": "1:58:54", "throughput": 2263.65, "total_tokens": 7215776}
{"current_steps": 1050, "total_steps": 3400, "eval_loss": 0.6906282901763916, "epoch": 0.12863311996569785, "percentage": 30.88, "elapsed_time": "0:53:26", "remaining_time": "1:59:37", "throughput": 2250.01, "total_tokens": 7215776}
{"current_steps": 1055, "total_steps": 3400, "loss": 0.5912, "lr": 8.259242386389973e-05, "epoch": 0.12924565863220117, "percentage": 31.03, "elapsed_time": "0:54:31", "remaining_time": "2:01:11", "throughput": 2216.1, "total_tokens": 7249440}
{"current_steps": 1060, "total_steps": 3400, "loss": 0.521, "lr": 8.240764101059912e-05, "epoch": 0.1298581972987045, "percentage": 31.18, "elapsed_time": "0:55:30", "remaining_time": "2:02:33", "throughput": 2186.56, "total_tokens": 7283432}
{"current_steps": 1065, "total_steps": 3400, "loss": 0.5359, "lr": 8.222209171180883e-05, "epoch": 0.1304707359652078, "percentage": 31.32, "elapsed_time": "0:56:29", "remaining_time": "2:03:51", "throughput": 2158.88, "total_tokens": 7317360}
{"current_steps": 1070, "total_steps": 3400, "loss": 0.5271, "lr": 8.203578035579715e-05, "epoch": 0.13108327463171113, "percentage": 31.47, "elapsed_time": "0:57:27", "remaining_time": "2:05:08", "throughput": 2132.34, "total_tokens": 7352112}
{"current_steps": 1075, "total_steps": 3400, "loss": 0.5044, "lr": 8.184871134885513e-05, "epoch": 0.13169581329821445, "percentage": 31.62, "elapsed_time": "0:58:27", "remaining_time": "2:06:26", "throughput": 2105.29, "total_tokens": 7384880}
{"current_steps": 1080, "total_steps": 3400, "loss": 0.5953, "lr": 8.166088911519235e-05, "epoch": 0.13230835196471777, "percentage": 31.76, "elapsed_time": "0:59:25", "remaining_time": "2:07:40", "throughput": 2080.34, "total_tokens": 7418368}
{"current_steps": 1085, "total_steps": 3400, "loss": 0.5123, "lr": 8.147231809683236e-05, "epoch": 0.1329208906312211, "percentage": 31.91, "elapsed_time": "1:00:24", "remaining_time": "2:08:53", "throughput": 2056.07, "total_tokens": 7451872}
{"current_steps": 1090, "total_steps": 3400, "loss": 0.5632, "lr": 8.128300275350756e-05, "epoch": 0.1335334292977244, "percentage": 32.06, "elapsed_time": "1:01:23", "remaining_time": "2:10:06", "throughput": 2032.1, "total_tokens": 7485712}
{"current_steps": 1095, "total_steps": 3400, "loss": 0.5144, "lr": 8.109294756255375e-05, "epoch": 0.13414596796422773, "percentage": 32.21, "elapsed_time": "1:02:21", "remaining_time": "2:11:16", "throughput": 2009.48, "total_tokens": 7519168}
{"current_steps": 1100, "total_steps": 3400, "loss": 0.5307, "lr": 8.090215701880419e-05, "epoch": 0.13475850663073108, "percentage": 32.35, "elapsed_time": "1:03:19", "remaining_time": "2:12:25", "throughput": 1987.66, "total_tokens": 7552920}
{"current_steps": 1100, "total_steps": 3400, "eval_loss": 0.6174182295799255, "epoch": 0.13475850663073108, "percentage": 32.35, "elapsed_time": "1:03:39", "remaining_time": "2:13:05", "throughput": 1977.64, "total_tokens": 7552920}
{"current_steps": 1105, "total_steps": 3400, "loss": 0.5067, "lr": 8.07106356344834e-05, "epoch": 0.1353710452972344, "percentage": 32.5, "elapsed_time": "1:04:43", "remaining_time": "2:14:26", "throughput": 1953.37, "total_tokens": 7586336}
{"current_steps": 1110, "total_steps": 3400, "loss": 0.5564, "lr": 8.051838793910038e-05, "epoch": 0.13598358396373772, "percentage": 32.65, "elapsed_time": "1:05:41", "remaining_time": "2:15:32", "throughput": 1933.24, "total_tokens": 7620456}
{"current_steps": 1115, "total_steps": 3400, "loss": 0.5854, "lr": 8.032541847934146e-05, "epoch": 0.13659612263024104, "percentage": 32.79, "elapsed_time": "1:06:39", "remaining_time": "2:16:37", "throughput": 1913.54, "total_tokens": 7654064}
{"current_steps": 1120, "total_steps": 3400, "loss": 0.5729, "lr": 8.013173181896283e-05, "epoch": 0.13720866129674436, "percentage": 32.94, "elapsed_time": "1:07:39", "remaining_time": "2:17:44", "throughput": 1893.75, "total_tokens": 7687768}
{"current_steps": 1125, "total_steps": 3400, "loss": 0.6122, "lr": 7.993733253868256e-05, "epoch": 0.13782119996324768, "percentage": 33.09, "elapsed_time": "1:08:38", "remaining_time": "2:18:48", "throughput": 1874.95, "total_tokens": 7721544}
{"current_steps": 1130, "total_steps": 3400, "loss": 0.5618, "lr": 7.974222523607236e-05, "epoch": 0.138433738629751, "percentage": 33.24, "elapsed_time": "1:09:36", "remaining_time": "2:19:49", "throughput": 1856.94, "total_tokens": 7755464}
{"current_steps": 1135, "total_steps": 3400, "loss": 0.5487, "lr": 7.954641452544865e-05, "epoch": 0.13904627729625432, "percentage": 33.38, "elapsed_time": "1:10:35", "remaining_time": "2:20:52", "throughput": 1838.93, "total_tokens": 7789152}
{"current_steps": 1140, "total_steps": 3400, "loss": 0.5132, "lr": 7.934990503776363e-05, "epoch": 0.13965881596275764, "percentage": 33.53, "elapsed_time": "1:11:34", "remaining_time": "2:21:53", "throughput": 1821.36, "total_tokens": 7822128}
{"current_steps": 1145, "total_steps": 3400, "loss": 0.5576, "lr": 7.915270142049566e-05, "epoch": 0.14027135462926096, "percentage": 33.68, "elapsed_time": "1:12:32", "remaining_time": "2:22:52", "throughput": 1804.77, "total_tokens": 7855584}
{"current_steps": 1150, "total_steps": 3400, "loss": 0.5403, "lr": 7.89548083375394e-05, "epoch": 0.14088389329576428, "percentage": 33.82, "elapsed_time": "1:13:31", "remaining_time": "2:23:50", "throughput": 1788.4, "total_tokens": 7889016}
{"current_steps": 1150, "total_steps": 3400, "eval_loss": 0.6082175970077515, "epoch": 0.14088389329576428, "percentage": 33.82, "elapsed_time": "1:13:50", "remaining_time": "2:24:28", "throughput": 1780.63, "total_tokens": 7889016}
{"current_steps": 1155, "total_steps": 3400, "loss": 0.531, "lr": 7.875623046909544e-05, "epoch": 0.1414964319622676, "percentage": 33.97, "elapsed_time": "1:14:54", "remaining_time": "2:25:36", "throughput": 1762.63, "total_tokens": 7922176}
{"current_steps": 1160, "total_steps": 3400, "loss": 0.5251, "lr": 7.855697251155967e-05, "epoch": 0.14210897062877095, "percentage": 34.12, "elapsed_time": "1:15:52", "remaining_time": "2:26:30", "throughput": 1747.7, "total_tokens": 7955904}
{"current_steps": 1165, "total_steps": 3400, "loss": 0.613, "lr": 7.835703917741212e-05, "epoch": 0.14272150929527427, "percentage": 34.26, "elapsed_time": "1:16:49", "remaining_time": "2:27:23", "throughput": 1733.27, "total_tokens": 7990096}
{"current_steps": 1170, "total_steps": 3400, "loss": 0.5264, "lr": 7.81564351951057e-05, "epoch": 0.1433340479617776, "percentage": 34.41, "elapsed_time": "1:17:48", "remaining_time": "2:28:17", "throughput": 1718.86, "total_tokens": 8024344}
{"current_steps": 1175, "total_steps": 3400, "loss": 0.4993, "lr": 7.795516530895414e-05, "epoch": 0.14394658662828091, "percentage": 34.56, "elapsed_time": "1:18:46", "remaining_time": "2:29:10", "throughput": 1704.98, "total_tokens": 8059128}
{"current_steps": 1180, "total_steps": 3400, "loss": 0.5187, "lr": 7.775323427901993e-05, "epoch": 0.14455912529478424, "percentage": 34.71, "elapsed_time": "1:19:45", "remaining_time": "2:30:02", "throughput": 1691.11, "total_tokens": 8092072}
{"current_steps": 1185, "total_steps": 3400, "loss": 0.5339, "lr": 7.755064688100171e-05, "epoch": 0.14517166396128756, "percentage": 34.85, "elapsed_time": "1:20:43", "remaining_time": "2:30:54", "throughput": 1677.45, "total_tokens": 8125568}
{"current_steps": 1190, "total_steps": 3400, "loss": 0.5192, "lr": 7.734740790612136e-05, "epoch": 0.14578420262779088, "percentage": 35.0, "elapsed_time": "1:21:42", "remaining_time": "2:31:43", "throughput": 1664.46, "total_tokens": 8159296}
{"current_steps": 1195, "total_steps": 3400, "loss": 0.4876, "lr": 7.714352216101055e-05, "epoch": 0.1463967412942942, "percentage": 35.15, "elapsed_time": "1:22:40", "remaining_time": "2:32:32", "throughput": 1651.67, "total_tokens": 8192992}
{"current_steps": 1200, "total_steps": 3400, "loss": 0.4855, "lr": 7.693899446759727e-05, "epoch": 0.14700927996079752, "percentage": 35.29, "elapsed_time": "1:23:40", "remaining_time": "2:33:23", "throughput": 1638.69, "total_tokens": 8226648}
{"current_steps": 1200, "total_steps": 3400, "eval_loss": 0.7277763485908508, "epoch": 0.14700927996079752, "percentage": 35.29, "elapsed_time": "1:23:59", "remaining_time": "2:33:58", "throughput": 1632.45, "total_tokens": 8226648}
{"current_steps": 1205, "total_steps": 3400, "loss": 0.5579, "lr": 7.673382966299163e-05, "epoch": 0.14762181862730084, "percentage": 35.44, "elapsed_time": "1:25:02", "remaining_time": "2:34:53", "throughput": 1618.83, "total_tokens": 8259304}
{"current_steps": 1210, "total_steps": 3400, "loss": 0.5636, "lr": 7.65280325993715e-05, "epoch": 0.14823435729380416, "percentage": 35.59, "elapsed_time": "1:26:00", "remaining_time": "2:35:40", "throughput": 1606.8, "total_tokens": 8292128}
{"current_steps": 1215, "total_steps": 3400, "loss": 0.4929, "lr": 7.63216081438678e-05, "epoch": 0.1488468959603075, "percentage": 35.74, "elapsed_time": "1:27:00", "remaining_time": "2:36:28", "throughput": 1594.68, "total_tokens": 8325272}
{"current_steps": 1220, "total_steps": 3400, "loss": 0.5441, "lr": 7.611456117844934e-05, "epoch": 0.14945943462681083, "percentage": 35.88, "elapsed_time": "1:27:59", "remaining_time": "2:37:13", "throughput": 1583.5, "total_tokens": 8359496}
{"current_steps": 1225, "total_steps": 3400, "loss": 0.4872, "lr": 7.59068965998074e-05, "epoch": 0.15007197329331415, "percentage": 36.03, "elapsed_time": "1:28:57", "remaining_time": "2:37:57", "throughput": 1572.37, "total_tokens": 8392792}
{"current_steps": 1230, "total_steps": 3400, "loss": 0.572, "lr": 7.569861931923989e-05, "epoch": 0.15068451195981747, "percentage": 36.18, "elapsed_time": "1:29:58", "remaining_time": "2:38:43", "throughput": 1560.88, "total_tokens": 8426176}
{"current_steps": 1235, "total_steps": 3400, "loss": 0.5814, "lr": 7.548973426253521e-05, "epoch": 0.1512970506263208, "percentage": 36.32, "elapsed_time": "1:30:56", "remaining_time": "2:39:26", "throughput": 1550.3, "total_tokens": 8459848}
{"current_steps": 1240, "total_steps": 3400, "loss": 0.5583, "lr": 7.528024636985575e-05, "epoch": 0.1519095892928241, "percentage": 36.47, "elapsed_time": "1:31:55", "remaining_time": "2:40:07", "throughput": 1539.87, "total_tokens": 8492616}
{"current_steps": 1245, "total_steps": 3400, "loss": 0.529, "lr": 7.507016059562107e-05, "epoch": 0.15252212795932743, "percentage": 36.62, "elapsed_time": "1:32:55", "remaining_time": "2:40:50", "throughput": 1529.19, "total_tokens": 8526152}
{"current_steps": 1250, "total_steps": 3400, "loss": 0.604, "lr": 7.485948190839077e-05, "epoch": 0.15313466662583075, "percentage": 36.76, "elapsed_time": "1:33:55", "remaining_time": "2:41:32", "throughput": 1519.01, "total_tokens": 8560104}
{"current_steps": 1250, "total_steps": 3400, "eval_loss": 0.6421969532966614, "epoch": 0.15313466662583075, "percentage": 36.76, "elapsed_time": "1:34:14", "remaining_time": "2:42:06", "throughput": 1513.79, "total_tokens": 8560104}
{"current_steps": 1255, "total_steps": 3400, "loss": 0.4929, "lr": 7.464821529074679e-05, "epoch": 0.15374720529233407, "percentage": 36.91, "elapsed_time": "1:35:19", "remaining_time": "2:42:55", "throughput": 1502.67, "total_tokens": 8594200}
{"current_steps": 1260, "total_steps": 3400, "loss": 0.5225, "lr": 7.443636573917585e-05, "epoch": 0.1543597439588374, "percentage": 37.06, "elapsed_time": "1:36:18", "remaining_time": "2:43:34", "throughput": 1493.03, "total_tokens": 8627784}
{"current_steps": 1265, "total_steps": 3400, "loss": 0.4588, "lr": 7.422393826395108e-05, "epoch": 0.1549722826253407, "percentage": 37.21, "elapsed_time": "1:37:17", "remaining_time": "2:44:13", "throughput": 1483.61, "total_tokens": 8661280}
{"current_steps": 1270, "total_steps": 3400, "loss": 0.4687, "lr": 7.40109378890136e-05, "epoch": 0.15558482129184406, "percentage": 37.35, "elapsed_time": "1:38:16", "remaining_time": "2:44:50", "throughput": 1474.41, "total_tokens": 8694496}
{"current_steps": 1275, "total_steps": 3400, "loss": 0.5354, "lr": 7.379736965185368e-05, "epoch": 0.15619735995834738, "percentage": 37.5, "elapsed_time": "1:39:15", "remaining_time": "2:45:26", "throughput": 1465.45, "total_tokens": 8727848}
{"current_steps": 1280, "total_steps": 3400, "loss": 0.4773, "lr": 7.358323860339165e-05, "epoch": 0.1568098986248507, "percentage": 37.65, "elapsed_time": "1:40:14", "remaining_time": "2:46:01", "throughput": 1456.69, "total_tokens": 8761416}
{"current_steps": 1285, "total_steps": 3400, "loss": 0.4848, "lr": 7.336854980785839e-05, "epoch": 0.15742243729135402, "percentage": 37.79, "elapsed_time": "1:41:12", "remaining_time": "2:46:34", "throughput": 1448.34, "total_tokens": 8794672}
{"current_steps": 1290, "total_steps": 3400, "loss": 0.5411, "lr": 7.315330834267553e-05, "epoch": 0.15803497595785734, "percentage": 37.94, "elapsed_time": "1:42:10", "remaining_time": "2:47:07", "throughput": 1440.13, "total_tokens": 8828392}
{"current_steps": 1295, "total_steps": 3400, "loss": 0.5434, "lr": 7.293751929833553e-05, "epoch": 0.15864751462436066, "percentage": 38.09, "elapsed_time": "1:43:09", "remaining_time": "2:47:40", "throughput": 1431.79, "total_tokens": 8861664}
{"current_steps": 1300, "total_steps": 3400, "loss": 0.5032, "lr": 7.272118777828108e-05, "epoch": 0.15926005329086398, "percentage": 38.24, "elapsed_time": "1:44:06", "remaining_time": "2:48:11", "throughput": 1423.94, "total_tokens": 8895088}
{"current_steps": 1300, "total_steps": 3400, "eval_loss": 0.643197774887085, "epoch": 0.15926005329086398, "percentage": 38.24, "elapsed_time": "1:44:26", "remaining_time": "2:48:42", "throughput": 1419.5, "total_tokens": 8895088}
{"current_steps": 1305, "total_steps": 3400, "loss": 0.4661, "lr": 7.250431889878455e-05, "epoch": 0.1598725919573673, "percentage": 38.38, "elapsed_time": "1:45:30", "remaining_time": "2:49:23", "throughput": 1410.31, "total_tokens": 8928264}
{"current_steps": 1310, "total_steps": 3400, "loss": 0.5823, "lr": 7.228691778882693e-05, "epoch": 0.16048513062387063, "percentage": 38.53, "elapsed_time": "1:46:30", "remaining_time": "2:49:54", "throughput": 1402.45, "total_tokens": 8961680}
{"current_steps": 1315, "total_steps": 3400, "loss": 0.5125, "lr": 7.20689895899765e-05, "epoch": 0.16109766929037395, "percentage": 38.68, "elapsed_time": "1:47:28", "remaining_time": "2:50:24", "throughput": 1394.98, "total_tokens": 8995272}
{"current_steps": 1320, "total_steps": 3400, "loss": 0.5206, "lr": 7.185053945626733e-05, "epoch": 0.16171020795687727, "percentage": 38.82, "elapsed_time": "1:48:26", "remaining_time": "2:50:53", "throughput": 1387.64, "total_tokens": 9029064}
{"current_steps": 1325, "total_steps": 3400, "loss": 0.5189, "lr": 7.163157255407732e-05, "epoch": 0.1623227466233806, "percentage": 38.97, "elapsed_time": "1:49:27", "remaining_time": "2:51:24", "throughput": 1380.08, "total_tokens": 9063336}
{"current_steps": 1330, "total_steps": 3400, "loss": 0.5273, "lr": 7.141209406200599e-05, "epoch": 0.16293528528988394, "percentage": 39.12, "elapsed_time": "1:50:25", "remaining_time": "2:51:52", "throughput": 1372.98, "total_tokens": 9097336}
{"current_steps": 1335, "total_steps": 3400, "loss": 0.4906, "lr": 7.1192109170752e-05, "epoch": 0.16354782395638726, "percentage": 39.26, "elapsed_time": "1:51:24", "remaining_time": "2:52:20", "throughput": 1365.79, "total_tokens": 9130160}
{"current_steps": 1340, "total_steps": 3400, "loss": 0.4826, "lr": 7.097162308299054e-05, "epoch": 0.16416036262289058, "percentage": 39.41, "elapsed_time": "1:52:25", "remaining_time": "2:52:50", "throughput": 1358.44, "total_tokens": 9163616}
{"current_steps": 1345, "total_steps": 3400, "loss": 0.5058, "lr": 7.07506410132501e-05, "epoch": 0.1647729012893939, "percentage": 39.56, "elapsed_time": "1:53:24", "remaining_time": "2:53:17", "throughput": 1351.56, "total_tokens": 9197368}
{"current_steps": 1350, "total_steps": 3400, "loss": 0.5053, "lr": 7.052916818778918e-05, "epoch": 0.16538543995589722, "percentage": 39.71, "elapsed_time": "1:54:24", "remaining_time": "2:53:43", "throughput": 1344.88, "total_tokens": 9231680}
{"current_steps": 1350, "total_steps": 3400, "eval_loss": 0.5727524161338806, "epoch": 0.16538543995589722, "percentage": 39.71, "elapsed_time": "1:54:43", "remaining_time": "2:54:13", "throughput": 1341.04, "total_tokens": 9231680}
{"current_steps": 1355, "total_steps": 3400, "loss": 0.5549, "lr": 7.030720984447279e-05, "epoch": 0.16599797862240054, "percentage": 39.85, "elapsed_time": "1:55:50", "remaining_time": "2:54:49", "throughput": 1333.1, "total_tokens": 9265288}
{"current_steps": 1360, "total_steps": 3400, "loss": 0.4997, "lr": 7.008477123264848e-05, "epoch": 0.16661051728890386, "percentage": 40.0, "elapsed_time": "1:56:49", "remaining_time": "2:55:14", "throughput": 1326.65, "total_tokens": 9299312}
{"current_steps": 1365, "total_steps": 3400, "loss": 0.5059, "lr": 6.986185761302224e-05, "epoch": 0.16722305595540718, "percentage": 40.15, "elapsed_time": "1:57:48", "remaining_time": "2:55:38", "throughput": 1320.27, "total_tokens": 9332808}
{"current_steps": 1370, "total_steps": 3400, "loss": 0.4701, "lr": 6.963847425753403e-05, "epoch": 0.1678355946219105, "percentage": 40.29, "elapsed_time": "1:58:49", "remaining_time": "2:56:03", "throughput": 1313.85, "total_tokens": 9366504}
{"current_steps": 1375, "total_steps": 3400, "loss": 0.4739, "lr": 6.941462644923318e-05, "epoch": 0.16844813328841382, "percentage": 40.44, "elapsed_time": "1:59:48", "remaining_time": "2:56:26", "throughput": 1307.6, "total_tokens": 9399752}
{"current_steps": 1380, "total_steps": 3400, "loss": 0.4967, "lr": 6.919031948215335e-05, "epoch": 0.16906067195491714, "percentage": 40.59, "elapsed_time": "2:00:47", "remaining_time": "2:56:48", "throughput": 1301.6, "total_tokens": 9433272}
{"current_steps": 1385, "total_steps": 3400, "loss": 0.4201, "lr": 6.896555866118741e-05, "epoch": 0.1696732106214205, "percentage": 40.74, "elapsed_time": "2:01:47", "remaining_time": "2:57:11", "throughput": 1295.46, "total_tokens": 9466912}
{"current_steps": 1390, "total_steps": 3400, "loss": 0.495, "lr": 6.87403493019619e-05, "epoch": 0.1702857492879238, "percentage": 40.88, "elapsed_time": "2:02:47", "remaining_time": "2:57:33", "throughput": 1289.43, "total_tokens": 9499568}
{"current_steps": 1395, "total_steps": 3400, "loss": 0.5148, "lr": 6.851469673071143e-05, "epoch": 0.17089828795442713, "percentage": 41.03, "elapsed_time": "2:03:45", "remaining_time": "2:57:53", "throughput": 1283.73, "total_tokens": 9532896}
{"current_steps": 1400, "total_steps": 3400, "loss": 0.4987, "lr": 6.828860628415253e-05, "epoch": 0.17151082662093045, "percentage": 41.18, "elapsed_time": "2:04:45", "remaining_time": "2:58:13", "throughput": 1277.95, "total_tokens": 9566000}
{"current_steps": 1400, "total_steps": 3400, "eval_loss": 0.4715874493122101, "epoch": 0.17151082662093045, "percentage": 41.18, "elapsed_time": "2:05:04", "remaining_time": "2:58:41", "throughput": 1274.64, "total_tokens": 9566000}
{"current_steps": 1405, "total_steps": 3400, "loss": 0.499, "lr": 6.806208330935766e-05, "epoch": 0.17212336528743377, "percentage": 41.32, "elapsed_time": "2:06:09", "remaining_time": "2:59:07", "throughput": 1268.25, "total_tokens": 9599640}
{"current_steps": 1410, "total_steps": 3400, "loss": 0.4612, "lr": 6.783513316362855e-05, "epoch": 0.1727359039539371, "percentage": 41.47, "elapsed_time": "2:07:07", "remaining_time": "2:59:25", "throughput": 1262.81, "total_tokens": 9632472}
{"current_steps": 1415, "total_steps": 3400, "loss": 0.5042, "lr": 6.760776121436962e-05, "epoch": 0.1733484426204404, "percentage": 41.62, "elapsed_time": "2:08:06", "remaining_time": "2:59:43", "throughput": 1257.4, "total_tokens": 9665592}
{"current_steps": 1420, "total_steps": 3400, "loss": 0.4785, "lr": 6.737997283896103e-05, "epoch": 0.17396098128694373, "percentage": 41.76, "elapsed_time": "2:09:07", "remaining_time": "3:00:02", "throughput": 1251.98, "total_tokens": 9699384}
{"current_steps": 1425, "total_steps": 3400, "loss": 0.402, "lr": 6.715177342463145e-05, "epoch": 0.17457351995344705, "percentage": 41.91, "elapsed_time": "2:10:06", "remaining_time": "3:00:19", "throughput": 1246.64, "total_tokens": 9732128}
{"current_steps": 1430, "total_steps": 3400, "loss": 0.5755, "lr": 6.692316836833065e-05, "epoch": 0.17518605861995037, "percentage": 42.06, "elapsed_time": "2:11:06", "remaining_time": "3:00:36", "throughput": 1241.48, "total_tokens": 9765864}
{"current_steps": 1435, "total_steps": 3400, "loss": 0.4779, "lr": 6.6694163076602e-05, "epoch": 0.1757985972864537, "percentage": 42.21, "elapsed_time": "2:12:06", "remaining_time": "3:00:54", "throughput": 1236.25, "total_tokens": 9799608}
{"current_steps": 1440, "total_steps": 3400, "loss": 0.5047, "lr": 6.646476296545434e-05, "epoch": 0.17641113595295704, "percentage": 42.35, "elapsed_time": "2:13:05", "remaining_time": "3:01:09", "throughput": 1231.31, "total_tokens": 9832880}
{"current_steps": 1445, "total_steps": 3400, "loss": 0.4869, "lr": 6.623497346023418e-05, "epoch": 0.17702367461946036, "percentage": 42.5, "elapsed_time": "2:14:04", "remaining_time": "3:01:23", "throughput": 1226.38, "total_tokens": 9865744}
{"current_steps": 1450, "total_steps": 3400, "loss": 0.44, "lr": 6.60047999954972e-05, "epoch": 0.17763621328596368, "percentage": 42.65, "elapsed_time": "2:15:06", "remaining_time": "3:01:41", "throughput": 1221.17, "total_tokens": 9899080}
{"current_steps": 1450, "total_steps": 3400, "eval_loss": 0.5081247091293335, "epoch": 0.17763621328596368, "percentage": 42.65, "elapsed_time": "2:15:25", "remaining_time": "3:02:07", "throughput": 1218.23, "total_tokens": 9899080}
{"current_steps": 1455, "total_steps": 3400, "loss": 0.4498, "lr": 6.57742480148798e-05, "epoch": 0.178248751952467, "percentage": 42.79, "elapsed_time": "2:16:29", "remaining_time": "3:02:27", "throughput": 1212.76, "total_tokens": 9932296}
{"current_steps": 1460, "total_steps": 3400, "loss": 0.5155, "lr": 6.554332297097031e-05, "epoch": 0.17886129061897033, "percentage": 42.94, "elapsed_time": "2:17:28", "remaining_time": "3:02:40", "throughput": 1208.14, "total_tokens": 9965544}
{"current_steps": 1465, "total_steps": 3400, "loss": 0.5036, "lr": 6.53120303251801e-05, "epoch": 0.17947382928547365, "percentage": 43.09, "elapsed_time": "2:18:28", "remaining_time": "3:02:54", "throughput": 1203.5, "total_tokens": 9999688}
{"current_steps": 1470, "total_steps": 3400, "loss": 0.4837, "lr": 6.508037554761432e-05, "epoch": 0.18008636795197697, "percentage": 43.24, "elapsed_time": "2:19:28", "remaining_time": "3:03:06", "throughput": 1198.97, "total_tokens": 10033136}
{"current_steps": 1475, "total_steps": 3400, "loss": 0.4418, "lr": 6.484836411694267e-05, "epoch": 0.1806989066184803, "percentage": 43.38, "elapsed_time": "2:20:27", "remaining_time": "3:03:18", "throughput": 1194.49, "total_tokens": 10066832}
{"current_steps": 1480, "total_steps": 3400, "loss": 0.4343, "lr": 6.461600152026965e-05, "epoch": 0.1813114452849836, "percentage": 43.53, "elapsed_time": "2:21:28", "remaining_time": "3:03:31", "throughput": 1189.98, "total_tokens": 10100952}
{"current_steps": 1485, "total_steps": 3400, "loss": 0.4359, "lr": 6.438329325300499e-05, "epoch": 0.18192398395148693, "percentage": 43.68, "elapsed_time": "2:22:27", "remaining_time": "3:03:43", "throughput": 1185.57, "total_tokens": 10134200}
{"current_steps": 1490, "total_steps": 3400, "loss": 0.4746, "lr": 6.415024481873352e-05, "epoch": 0.18253652261799025, "percentage": 43.82, "elapsed_time": "2:23:26", "remaining_time": "3:03:53", "throughput": 1181.38, "total_tokens": 10168056}
{"current_steps": 1495, "total_steps": 3400, "loss": 0.4221, "lr": 6.391686172908506e-05, "epoch": 0.1831490612844936, "percentage": 43.97, "elapsed_time": "2:24:27", "remaining_time": "3:04:04", "throughput": 1176.99, "total_tokens": 10201120}
{"current_steps": 1500, "total_steps": 3400, "loss": 0.4207, "lr": 6.368314950360415e-05, "epoch": 0.18376159995099692, "percentage": 44.12, "elapsed_time": "2:25:26", "remaining_time": "3:04:13", "throughput": 1172.89, "total_tokens": 10234856}
{"current_steps": 1500, "total_steps": 3400, "eval_loss": 0.5098862648010254, "epoch": 0.18376159995099692, "percentage": 44.12, "elapsed_time": "2:25:45", "remaining_time": "3:04:37", "throughput": 1170.28, "total_tokens": 10234856}
{"current_steps": 1505, "total_steps": 3400, "loss": 0.4755, "lr": 6.344911366961934e-05, "epoch": 0.18437413861750024, "percentage": 44.26, "elapsed_time": "2:26:49", "remaining_time": "3:04:52", "throughput": 1165.56, "total_tokens": 10268128}
{"current_steps": 1510, "total_steps": 3400, "loss": 0.4894, "lr": 6.321475976211266e-05, "epoch": 0.18498667728400356, "percentage": 44.41, "elapsed_time": "2:27:49", "remaining_time": "3:05:01", "throughput": 1161.46, "total_tokens": 10301440}
{"current_steps": 1515, "total_steps": 3400, "loss": 0.4294, "lr": 6.298009332358856e-05, "epoch": 0.18559921595050688, "percentage": 44.56, "elapsed_time": "2:28:49", "remaining_time": "3:05:09", "throughput": 1157.45, "total_tokens": 10334896}
{"current_steps": 1520, "total_steps": 3400, "loss": 0.5088, "lr": 6.274511990394294e-05, "epoch": 0.1862117546170102, "percentage": 44.71, "elapsed_time": "2:29:48", "remaining_time": "3:05:16", "throughput": 1153.58, "total_tokens": 10368584}
{"current_steps": 1525, "total_steps": 3400, "loss": 0.4386, "lr": 6.250984506033183e-05, "epoch": 0.18682429328351352, "percentage": 44.85, "elapsed_time": "2:30:47", "remaining_time": "3:05:24", "throughput": 1149.64, "total_tokens": 10401576}
{"current_steps": 1530, "total_steps": 3400, "loss": 0.4836, "lr": 6.227427435703997e-05, "epoch": 0.18743683195001684, "percentage": 45.0, "elapsed_time": "2:31:47", "remaining_time": "3:05:31", "throughput": 1145.72, "total_tokens": 10435032}
{"current_steps": 1535, "total_steps": 3400, "loss": 0.536, "lr": 6.203841336534924e-05, "epoch": 0.18804937061652016, "percentage": 45.15, "elapsed_time": "2:32:47", "remaining_time": "3:05:37", "throughput": 1141.94, "total_tokens": 10468336}
{"current_steps": 1540, "total_steps": 3400, "loss": 0.4544, "lr": 6.180226766340688e-05, "epoch": 0.18866190928302348, "percentage": 45.29, "elapsed_time": "2:33:46", "remaining_time": "3:05:43", "throughput": 1138.26, "total_tokens": 10502128}
{"current_steps": 1545, "total_steps": 3400, "loss": 0.4647, "lr": 6.156584283609359e-05, "epoch": 0.1892744479495268, "percentage": 45.44, "elapsed_time": "2:34:46", "remaining_time": "3:05:49", "throughput": 1134.59, "total_tokens": 10536432}
{"current_steps": 1550, "total_steps": 3400, "loss": 0.4664, "lr": 6.132914447489137e-05, "epoch": 0.18988698661603012, "percentage": 45.59, "elapsed_time": "2:35:45", "remaining_time": "3:05:54", "throughput": 1131.09, "total_tokens": 10570832}
{"current_steps": 1550, "total_steps": 3400, "eval_loss": 0.4264271855354309, "epoch": 0.18988698661603012, "percentage": 45.59, "elapsed_time": "2:36:05", "remaining_time": "3:06:18", "throughput": 1128.69, "total_tokens": 10570832}
{"current_steps": 1555, "total_steps": 3400, "loss": 0.401, "lr": 6.109217817775139e-05, "epoch": 0.19049952528253347, "percentage": 45.74, "elapsed_time": "2:37:10", "remaining_time": "3:06:29", "throughput": 1124.43, "total_tokens": 10604448}
{"current_steps": 1560, "total_steps": 3400, "loss": 0.4273, "lr": 6.085494954896156e-05, "epoch": 0.1911120639490368, "percentage": 45.88, "elapsed_time": "2:38:11", "remaining_time": "3:06:34", "throughput": 1120.84, "total_tokens": 10638216}
{"current_steps": 1565, "total_steps": 3400, "loss": 0.528, "lr": 6.061746419901388e-05, "epoch": 0.1917246026155401, "percentage": 46.03, "elapsed_time": "2:39:10", "remaining_time": "3:06:38", "throughput": 1117.46, "total_tokens": 10672472}
{"current_steps": 1570, "total_steps": 3400, "loss": 0.4963, "lr": 6.0379727744471936e-05, "epoch": 0.19233714128204343, "percentage": 46.18, "elapsed_time": "2:40:10", "remaining_time": "3:06:41", "throughput": 1113.97, "total_tokens": 10705584}
{"current_steps": 1575, "total_steps": 3400, "loss": 0.4823, "lr": 6.014174580783794e-05, "epoch": 0.19294967994854675, "percentage": 46.32, "elapsed_time": "2:41:10", "remaining_time": "3:06:46", "throughput": 1110.52, "total_tokens": 10739776}
{"current_steps": 1580, "total_steps": 3400, "loss": 0.4346, "lr": 5.990352401741981e-05, "epoch": 0.19356221861505007, "percentage": 46.47, "elapsed_time": "2:42:10", "remaining_time": "3:06:48", "throughput": 1107.12, "total_tokens": 10773128}
{"current_steps": 1585, "total_steps": 3400, "loss": 0.4038, "lr": 5.9665068007197976e-05, "epoch": 0.1941747572815534, "percentage": 46.62, "elapsed_time": "2:43:10", "remaining_time": "3:06:51", "throughput": 1103.84, "total_tokens": 10807296}
{"current_steps": 1590, "total_steps": 3400, "loss": 0.5107, "lr": 5.94263834166923e-05, "epoch": 0.19478729594805672, "percentage": 46.76, "elapsed_time": "2:44:10", "remaining_time": "3:06:53", "throughput": 1100.47, "total_tokens": 10840664}
{"current_steps": 1595, "total_steps": 3400, "loss": 0.4815, "lr": 5.918747589082853e-05, "epoch": 0.19539983461456004, "percentage": 46.91, "elapsed_time": "2:45:10", "remaining_time": "3:06:54", "throughput": 1097.23, "total_tokens": 10873584}
{"current_steps": 1600, "total_steps": 3400, "loss": 0.4713, "lr": 5.8948351079804875e-05, "epoch": 0.19601237328106336, "percentage": 47.06, "elapsed_time": "2:46:09", "remaining_time": "3:06:55", "throughput": 1094.08, "total_tokens": 10907608}
{"current_steps": 1600, "total_steps": 3400, "eval_loss": 0.503852128982544, "epoch": 0.19601237328106336, "percentage": 47.06, "elapsed_time": "2:46:29", "remaining_time": "3:07:17", "throughput": 1091.93, "total_tokens": 10907608}
{"current_steps": 1605, "total_steps": 3400, "loss": 0.4977, "lr": 5.8709014638958404e-05, "epoch": 0.19662491194756668, "percentage": 47.21, "elapsed_time": "2:47:34", "remaining_time": "3:07:25", "throughput": 1088.14, "total_tokens": 10941040}
{"current_steps": 1610, "total_steps": 3400, "loss": 0.4665, "lr": 5.846947222863123e-05, "epoch": 0.19723745061407003, "percentage": 47.35, "elapsed_time": "2:48:34", "remaining_time": "3:07:25", "throughput": 1085.05, "total_tokens": 10974800}
{"current_steps": 1615, "total_steps": 3400, "loss": 0.4976, "lr": 5.8229729514036705e-05, "epoch": 0.19784998928057335, "percentage": 47.5, "elapsed_time": "2:49:33", "remaining_time": "3:07:24", "throughput": 1082.06, "total_tokens": 11008560}
{"current_steps": 1620, "total_steps": 3400, "loss": 0.5012, "lr": 5.7989792165125356e-05, "epoch": 0.19846252794707667, "percentage": 47.65, "elapsed_time": "2:50:34", "remaining_time": "3:07:25", "throughput": 1078.9, "total_tokens": 11041792}
{"current_steps": 1625, "total_steps": 3400, "loss": 0.5137, "lr": 5.774966585645092e-05, "epoch": 0.19907506661358, "percentage": 47.79, "elapsed_time": "2:51:33", "remaining_time": "3:07:23", "throughput": 1075.93, "total_tokens": 11075256}
{"current_steps": 1630, "total_steps": 3400, "loss": 0.4566, "lr": 5.7509356267035975e-05, "epoch": 0.1996876052800833, "percentage": 47.94, "elapsed_time": "2:52:32", "remaining_time": "3:07:21", "throughput": 1073.11, "total_tokens": 11109320}
{"current_steps": 1635, "total_steps": 3400, "loss": 0.4311, "lr": 5.726886908023776e-05, "epoch": 0.20030014394658663, "percentage": 48.09, "elapsed_time": "2:53:32", "remaining_time": "3:07:20", "throughput": 1070.16, "total_tokens": 11142816}
{"current_steps": 1640, "total_steps": 3400, "loss": 0.5047, "lr": 5.702820998361373e-05, "epoch": 0.20091268261308995, "percentage": 48.24, "elapsed_time": "2:54:31", "remaining_time": "3:07:17", "throughput": 1067.26, "total_tokens": 11175992}
{"current_steps": 1645, "total_steps": 3400, "loss": 0.3824, "lr": 5.6787384668786994e-05, "epoch": 0.20152522127959327, "percentage": 48.38, "elapsed_time": "2:55:31", "remaining_time": "3:07:15", "throughput": 1064.47, "total_tokens": 11209952}
{"current_steps": 1650, "total_steps": 3400, "loss": 0.4385, "lr": 5.654639883131178e-05, "epoch": 0.2021377599460966, "percentage": 48.53, "elapsed_time": "2:56:31", "remaining_time": "3:07:13", "throughput": 1061.59, "total_tokens": 11243856}
{"current_steps": 1650, "total_steps": 3400, "eval_loss": 0.4518810510635376, "epoch": 0.2021377599460966, "percentage": 48.53, "elapsed_time": "2:56:51", "remaining_time": "3:07:34", "throughput": 1059.61, "total_tokens": 11243856}
{"current_steps": 1655, "total_steps": 3400, "loss": 0.4356, "lr": 5.6305258170538676e-05, "epoch": 0.2027502986125999, "percentage": 48.68, "elapsed_time": "2:57:56", "remaining_time": "3:07:37", "throughput": 1056.24, "total_tokens": 11277312}
{"current_steps": 1660, "total_steps": 3400, "loss": 0.4458, "lr": 5.606396838947988e-05, "epoch": 0.20336283727910323, "percentage": 48.82, "elapsed_time": "2:58:56", "remaining_time": "3:07:34", "throughput": 1053.44, "total_tokens": 11310576}
{"current_steps": 1665, "total_steps": 3400, "loss": 0.4618, "lr": 5.582253519467432e-05, "epoch": 0.20397537594560658, "percentage": 48.97, "elapsed_time": "2:59:56", "remaining_time": "3:07:30", "throughput": 1050.71, "total_tokens": 11344448}
{"current_steps": 1670, "total_steps": 3400, "loss": 0.4647, "lr": 5.558096429605263e-05, "epoch": 0.2045879146121099, "percentage": 49.12, "elapsed_time": "3:00:57", "remaining_time": "3:07:27", "throughput": 1047.92, "total_tokens": 11377768}
{"current_steps": 1675, "total_steps": 3400, "loss": 0.4304, "lr": 5.533926140680221e-05, "epoch": 0.20520045327861322, "percentage": 49.26, "elapsed_time": "3:01:57", "remaining_time": "3:07:23", "throughput": 1045.24, "total_tokens": 11411408}
{"current_steps": 1680, "total_steps": 3400, "loss": 0.5269, "lr": 5.509743224323203e-05, "epoch": 0.20581299194511654, "percentage": 49.41, "elapsed_time": "3:02:58", "remaining_time": "3:07:19", "throughput": 1042.51, "total_tokens": 11445104}
{"current_steps": 1685, "total_steps": 3400, "loss": 0.4352, "lr": 5.485548252463749e-05, "epoch": 0.20642553061161986, "percentage": 49.56, "elapsed_time": "3:04:00", "remaining_time": "3:07:16", "throughput": 1039.65, "total_tokens": 11478128}
{"current_steps": 1690, "total_steps": 3400, "loss": 0.4879, "lr": 5.4613417973165106e-05, "epoch": 0.20703806927812318, "percentage": 49.71, "elapsed_time": "3:05:00", "remaining_time": "3:07:11", "throughput": 1037.08, "total_tokens": 11511768}
{"current_steps": 1695, "total_steps": 3400, "loss": 0.4436, "lr": 5.4371244313677225e-05, "epoch": 0.2076506079446265, "percentage": 49.85, "elapsed_time": "3:05:59", "remaining_time": "3:07:05", "throughput": 1034.55, "total_tokens": 11545128}
{"current_steps": 1700, "total_steps": 3400, "loss": 0.4487, "lr": 5.4128967273616625e-05, "epoch": 0.20826314661112982, "percentage": 50.0, "elapsed_time": "3:07:00", "remaining_time": "3:07:00", "throughput": 1031.96, "total_tokens": 11578832}
{"current_steps": 1700, "total_steps": 3400, "eval_loss": 0.49612876772880554, "epoch": 0.20826314661112982, "percentage": 50.0, "elapsed_time": "3:07:19", "remaining_time": "3:07:19", "throughput": 1030.18, "total_tokens": 11578832}
{"current_steps": 1705, "total_steps": 3400, "loss": 0.4203, "lr": 5.388659258287102e-05, "epoch": 0.20887568527763314, "percentage": 50.15, "elapsed_time": "3:08:23", "remaining_time": "3:07:17", "throughput": 1027.37, "total_tokens": 11612808}
{"current_steps": 1710, "total_steps": 3400, "loss": 0.4455, "lr": 5.364412597363759e-05, "epoch": 0.20948822394413646, "percentage": 50.29, "elapsed_time": "3:09:22", "remaining_time": "3:07:09", "throughput": 1025.03, "total_tokens": 11646424}
{"current_steps": 1715, "total_steps": 3400, "loss": 0.5335, "lr": 5.3401573180287426e-05, "epoch": 0.21010076261063979, "percentage": 50.44, "elapsed_time": "3:10:21", "remaining_time": "3:07:01", "throughput": 1022.65, "total_tokens": 11679824}
{"current_steps": 1720, "total_steps": 3400, "loss": 0.3974, "lr": 5.315893993922986e-05, "epoch": 0.21071330127714313, "percentage": 50.59, "elapsed_time": "3:11:19", "remaining_time": "3:06:52", "throughput": 1020.42, "total_tokens": 11713928}
{"current_steps": 1725, "total_steps": 3400, "loss": 0.479, "lr": 5.29162319887768e-05, "epoch": 0.21132583994364645, "percentage": 50.74, "elapsed_time": "3:12:18", "remaining_time": "3:06:44", "throughput": 1018.08, "total_tokens": 11747400}
{"current_steps": 1730, "total_steps": 3400, "loss": 0.388, "lr": 5.26734550690071e-05, "epoch": 0.21193837861014977, "percentage": 50.88, "elapsed_time": "3:13:17", "remaining_time": "3:06:35", "throughput": 1015.79, "total_tokens": 11781016}
{"current_steps": 1735, "total_steps": 3400, "loss": 0.4914, "lr": 5.243061492163073e-05, "epoch": 0.2125509172766531, "percentage": 51.03, "elapsed_time": "3:14:15", "remaining_time": "3:06:25", "throughput": 1013.67, "total_tokens": 11815064}
{"current_steps": 1790, "total_steps": 3400, "loss": 0.4387, "lr": 4.9756843632322626e-05, "epoch": 0.21928884260818965, "percentage": 52.65, "elapsed_time": "3:25:27", "remaining_time": "3:04:48", "throughput": 988.37, "total_tokens": 12184368}
{"current_steps": 1795, "total_steps": 3400, "loss": 0.3937, "lr": 4.9513693015329197e-05, "epoch": 0.21990138127469297, "percentage": 52.79, "elapsed_time": "3:26:27", "remaining_time": "3:04:36", "throughput": 986.33, "total_tokens": 12218216}
{"current_steps": 1800, "total_steps": 3400, "loss": 0.4561, "lr": 4.9270553899567686e-05, "epoch": 0.2205139199411963, "percentage": 52.94, "elapsed_time": "3:27:26", "remaining_time": "3:04:23", "throughput": 984.36, "total_tokens": 12251696}
{"current_steps": 1800, "total_steps": 3400, "eval_loss": 0.3055322766304016, "epoch": 0.2205139199411963, "percentage": 52.94, "elapsed_time": "3:27:45", "remaining_time": "3:04:40", "throughput": 982.81, "total_tokens": 12251696}
{"current_steps": 1805, "total_steps": 3400, "loss": 0.378, "lr": 4.902743203531405e-05, "epoch": 0.2211264586076996, "percentage": 53.09, "elapsed_time": "3:28:50", "remaining_time": "3:04:32", "throughput": 980.42, "total_tokens": 12284984}
{"current_steps": 1810, "total_steps": 3400, "loss": 0.458, "lr": 4.8784333172436206e-05, "epoch": 0.22173899727420293, "percentage": 53.24, "elapsed_time": "3:29:50", "remaining_time": "3:04:19", "throughput": 978.37, "total_tokens": 12317744}
{"current_steps": 1815, "total_steps": 3400, "loss": 0.4636, "lr": 4.854126306025812e-05, "epoch": 0.22235153594070625, "percentage": 53.38, "elapsed_time": "3:30:47", "remaining_time": "3:04:05", "throughput": 976.58, "total_tokens": 12351680}
{"current_steps": 1820, "total_steps": 3400, "loss": 0.4526, "lr": 4.829822744742383e-05, "epoch": 0.22296407460720957, "percentage": 53.53, "elapsed_time": "3:31:45", "remaining_time": "3:03:50", "throughput": 974.73, "total_tokens": 12384872}
{"current_steps": 1825, "total_steps": 3400, "loss": 0.4075, "lr": 4.8055232081761395e-05, "epoch": 0.2235766132737129, "percentage": 53.68, "elapsed_time": "3:32:44", "remaining_time": "3:03:36", "throughput": 972.88, "total_tokens": 12418488}
{"current_steps": 1830, "total_steps": 3400, "loss": 0.484, "lr": 4.781228271014704e-05, "epoch": 0.2241891519402162, "percentage": 53.82, "elapsed_time": "3:33:42", "remaining_time": "3:03:20", "throughput": 971.12, "total_tokens": 12452200}
{"current_steps": 1835, "total_steps": 3400, "loss": 0.3603, "lr": 4.756938507836929e-05, "epoch": 0.22480169060671956, "percentage": 53.97, "elapsed_time": "3:34:40", "remaining_time": "3:03:05", "throughput": 969.33, "total_tokens": 12485920}
{"current_steps": 1805, "total_steps": 3400, "loss": 0.8177, "lr": 4.902743203531405e-05, "epoch": 0.2211264586076996, "percentage": 53.09, "elapsed_time": "0:02:13", "remaining_time": "0:01:58", "throughput": 91524.53, "total_tokens": 12259384}
{"current_steps": 1810, "total_steps": 3400, "loss": 0.7884, "lr": 4.8784333172436206e-05, "epoch": 0.22173899727420293, "percentage": 53.24, "elapsed_time": "0:03:11", "remaining_time": "0:02:48", "throughput": 64037.18, "total_tokens": 12266544}
{"current_steps": 1815, "total_steps": 3400, "loss": 0.8374, "lr": 4.854126306025812e-05, "epoch": 0.22235153594070625, "percentage": 53.38, "elapsed_time": "0:04:08", "remaining_time": "0:03:37", "throughput": 49327.83, "total_tokens": 12274880}
{"current_steps": 1820, "total_steps": 3400, "loss": 0.8302, "lr": 4.829822744742383e-05, "epoch": 0.22296407460720957, "percentage": 53.53, "elapsed_time": "0:05:06", "remaining_time": "0:04:25", "throughput": 40106.08, "total_tokens": 12282472}
{"current_steps": 1825, "total_steps": 3400, "loss": 0.7611, "lr": 4.8055232081761395e-05, "epoch": 0.2235766132737129, "percentage": 53.68, "elapsed_time": "0:06:03", "remaining_time": "0:05:14", "throughput": 33770.92, "total_tokens": 12290488}
{"current_steps": 1830, "total_steps": 3400, "loss": 0.7718, "lr": 4.781228271014704e-05, "epoch": 0.2241891519402162, "percentage": 53.82, "elapsed_time": "0:07:01", "remaining_time": "0:06:01", "throughput": 29176.75, "total_tokens": 12298600}
{"current_steps": 1835, "total_steps": 3400, "loss": 0.821, "lr": 4.756938507836929e-05, "epoch": 0.22480169060671956, "percentage": 53.97, "elapsed_time": "0:07:58", "remaining_time": "0:06:48", "throughput": 25699.76, "total_tokens": 12306720}
{"current_steps": 1840, "total_steps": 3400, "loss": 0.7713, "lr": 4.732654493099291e-05, "epoch": 0.22541422927322288, "percentage": 54.12, "elapsed_time": "0:08:57", "remaining_time": "0:07:35", "throughput": 22920.87, "total_tokens": 12314240}
{"current_steps": 1845, "total_steps": 3400, "loss": 0.8226, "lr": 4.708376801122321e-05, "epoch": 0.2260267679397262, "percentage": 54.26, "elapsed_time": "0:09:54", "remaining_time": "0:08:21", "throughput": 20718.75, "total_tokens": 12321944}
{"current_steps": 1850, "total_steps": 3400, "loss": 0.7805, "lr": 4.6841060060770154e-05, "epoch": 0.22663930660622952, "percentage": 54.41, "elapsed_time": "0:10:52", "remaining_time": "0:09:06", "throughput": 18898.03, "total_tokens": 12329936}
{"current_steps": 1850, "total_steps": 3400, "eval_loss": 0.9060287475585938, "epoch": 0.22663930660622952, "percentage": 54.41, "elapsed_time": "0:11:38", "remaining_time": "0:09:45", "throughput": 17652.7, "total_tokens": 12329936}
{"current_steps": 1855, "total_steps": 3400, "loss": 0.7034, "lr": 4.659842681971258e-05, "epoch": 0.22725184527273284, "percentage": 54.56, "elapsed_time": "0:12:45", "remaining_time": "0:10:37", "throughput": 16108.93, "total_tokens": 12337664}
{"current_steps": 1860, "total_steps": 3400, "loss": 0.7901, "lr": 4.635587402636241e-05, "epoch": 0.22786438393923616, "percentage": 54.71, "elapsed_time": "0:13:43", "remaining_time": "0:11:21", "throughput": 14991.69, "total_tokens": 12345248}
{"current_steps": 1865, "total_steps": 3400, "loss": 0.7636, "lr": 4.611340741712901e-05, "epoch": 0.22847692260573949, "percentage": 54.85, "elapsed_time": "0:14:40", "remaining_time": "0:12:05", "throughput": 14022.45, "total_tokens": 12353168}
{"current_steps": 1870, "total_steps": 3400, "loss": 0.7921, "lr": 4.5871032726383386e-05, "epoch": 0.2290894612722428, "percentage": 55.0, "elapsed_time": "0:15:39", "remaining_time": "0:12:48", "throughput": 13159.84, "total_tokens": 12361080}
{"current_steps": 1875, "total_steps": 3400, "loss": 0.7547, "lr": 4.562875568632278e-05, "epoch": 0.22970199993874613, "percentage": 55.15, "elapsed_time": "0:16:37", "remaining_time": "0:13:31", "throughput": 12401.45, "total_tokens": 12369008}
{"current_steps": 1880, "total_steps": 3400, "loss": 0.7681, "lr": 4.5386582026834906e-05, "epoch": 0.23031453860524945, "percentage": 55.29, "elapsed_time": "0:17:34", "remaining_time": "0:14:12", "throughput": 11740.26, "total_tokens": 12377544}
{"current_steps": 1885, "total_steps": 3400, "loss": 0.7373, "lr": 4.5144517475362514e-05, "epoch": 0.23092707727175277, "percentage": 55.44, "elapsed_time": "0:18:32", "remaining_time": "0:14:53", "throughput": 11136.35, "total_tokens": 12384824}
{"current_steps": 1890, "total_steps": 3400, "loss": 0.7915, "lr": 4.490256775676797e-05, "epoch": 0.23153961593825612, "percentage": 55.59, "elapsed_time": "0:19:30", "remaining_time": "0:15:35", "throughput": 10583.5, "total_tokens": 12392984}
{"current_steps": 1895, "total_steps": 3400, "loss": 0.8194, "lr": 4.466073859319781e-05, "epoch": 0.23215215460475944, "percentage": 55.74, "elapsed_time": "0:20:28", "remaining_time": "0:16:15", "throughput": 10094.01, "total_tokens": 12401976}
{"current_steps": 1900, "total_steps": 3400, "loss": 0.7445, "lr": 4.441903570394739e-05, "epoch": 0.23276469327126276, "percentage": 55.88, "elapsed_time": "0:21:26", "remaining_time": "0:16:55", "throughput": 9647.99, "total_tokens": 12410504}
{"current_steps": 1900, "total_steps": 3400, "eval_loss": 0.9240804314613342, "epoch": 0.23276469327126276, "percentage": 55.88, "elapsed_time": "0:21:44", "remaining_time": "0:17:09", "throughput": 9515.36, "total_tokens": 12410504}
{"current_steps": 1905, "total_steps": 3400, "loss": 0.7772, "lr": 4.41774648053257e-05, "epoch": 0.23337723193776608, "percentage": 56.03, "elapsed_time": "0:22:48", "remaining_time": "0:17:54", "throughput": 9073.38, "total_tokens": 12418576}
{"current_steps": 1910, "total_steps": 3400, "loss": 0.8106, "lr": 4.3936031610520124e-05, "epoch": 0.2339897706042694, "percentage": 56.18, "elapsed_time": "0:23:46", "remaining_time": "0:18:32", "throughput": 8712.06, "total_tokens": 12426760}
{"current_steps": 1915, "total_steps": 3400, "loss": 0.7826, "lr": 4.3694741829461336e-05, "epoch": 0.23460230927077272, "percentage": 56.32, "elapsed_time": "0:24:44", "remaining_time": "0:19:11", "throughput": 8376.66, "total_tokens": 12434144}
{"current_steps": 1920, "total_steps": 3400, "loss": 0.782, "lr": 4.345360116868823e-05, "epoch": 0.23521484793727604, "percentage": 56.47, "elapsed_time": "0:25:43", "remaining_time": "0:19:49", "throughput": 8062.32, "total_tokens": 12442408}
{"current_steps": 1925, "total_steps": 3400, "loss": 0.7309, "lr": 4.321261533121303e-05, "epoch": 0.23582738660377936, "percentage": 56.62, "elapsed_time": "0:26:41", "remaining_time": "0:20:26", "throughput": 7775.45, "total_tokens": 12449632}
{"current_steps": 1905, "total_steps": 3400, "loss": 0.4769, "lr": 4.41774648053257e-05, "epoch": 0.23337723193776608, "percentage": 56.03, "elapsed_time": "0:02:18", "remaining_time": "0:01:48", "throughput": 89688.02, "total_tokens": 12444176}
{"current_steps": 1910, "total_steps": 3400, "loss": 0.5032, "lr": 4.3936031610520124e-05, "epoch": 0.2339897706042694, "percentage": 56.18, "elapsed_time": "0:03:19", "remaining_time": "0:02:35", "throughput": 62552.31, "total_tokens": 12477960}
{"current_steps": 1915, "total_steps": 3400, "loss": 0.4889, "lr": 4.3694741829461336e-05, "epoch": 0.23460230927077272, "percentage": 56.32, "elapsed_time": "0:04:19", "remaining_time": "0:03:21", "throughput": 48148.96, "total_tokens": 12510944}
{"current_steps": 1920, "total_steps": 3400, "loss": 0.4184, "lr": 4.345360116868823e-05, "epoch": 0.23521484793727604, "percentage": 56.47, "elapsed_time": "0:05:19", "remaining_time": "0:04:06", "throughput": 39244.3, "total_tokens": 12544808}
{"current_steps": 1925, "total_steps": 3400, "loss": 0.4049, "lr": 4.321261533121303e-05, "epoch": 0.23582738660377936, "percentage": 56.62, "elapsed_time": "0:06:20", "remaining_time": "0:04:51", "throughput": 33047.03, "total_tokens": 12577632}
{"current_steps": 1930, "total_steps": 3400, "loss": 0.4695, "lr": 4.2971790016386286e-05, "epoch": 0.23643992527028268, "percentage": 56.76, "elapsed_time": "0:07:20", "remaining_time": "0:05:35", "throughput": 28655.27, "total_tokens": 12610192}
{"current_steps": 1935, "total_steps": 3400, "loss": 0.4962, "lr": 4.273113091976225e-05, "epoch": 0.237052463936786, "percentage": 56.91, "elapsed_time": "0:08:19", "remaining_time": "0:06:18", "throughput": 25301.2, "total_tokens": 12643688}
{"current_steps": 1940, "total_steps": 3400, "loss": 0.4782, "lr": 4.249064373296403e-05, "epoch": 0.23766500260328932, "percentage": 57.06, "elapsed_time": "0:09:19", "remaining_time": "0:07:01", "throughput": 22642.69, "total_tokens": 12677312}
{"current_steps": 1945, "total_steps": 3400, "loss": 0.5615, "lr": 4.225033414354908e-05, "epoch": 0.23827754126979264, "percentage": 57.21, "elapsed_time": "0:10:19", "remaining_time": "0:07:43", "throughput": 20505.95, "total_tokens": 12709896}
{"current_steps": 1950, "total_steps": 3400, "loss": 0.4354, "lr": 4.201020783487464e-05, "epoch": 0.238890079936296, "percentage": 57.35, "elapsed_time": "0:11:19", "remaining_time": "0:08:24", "throughput": 18768.14, "total_tokens": 12744416}
{"current_steps": 1950, "total_steps": 3400, "eval_loss": 0.36006462574005127, "epoch": 0.238890079936296, "percentage": 57.35, "elapsed_time": "0:12:06", "remaining_time": "0:09:00", "throughput": 17538.27, "total_tokens": 12744416}
{"current_steps": 1955, "total_steps": 3400, "loss": 0.4335, "lr": 4.17702704859633e-05, "epoch": 0.2395026186027993, "percentage": 57.5, "elapsed_time": "0:13:15", "remaining_time": "0:09:48", "throughput": 16058.94, "total_tokens": 12777816}
{"current_steps": 1960, "total_steps": 3400, "loss": 0.4991, "lr": 4.153052777136879e-05, "epoch": 0.24011515726930263, "percentage": 57.65, "elapsed_time": "0:14:14", "remaining_time": "0:10:27", "throughput": 14989.66, "total_tokens": 12811792}
{"current_steps": 1965, "total_steps": 3400, "loss": 0.4099, "lr": 4.1290985361041614e-05, "epoch": 0.24072769593580595, "percentage": 57.79, "elapsed_time": "0:15:13", "remaining_time": "0:11:06", "throughput": 14064.24, "total_tokens": 12845480}
{"current_steps": 1970, "total_steps": 3400, "loss": 0.4272, "lr": 4.105164892019514e-05, "epoch": 0.24134023460230927, "percentage": 57.94, "elapsed_time": "0:16:12", "remaining_time": "0:11:46", "throughput": 13241.71, "total_tokens": 12879160}
{"current_steps": 1975, "total_steps": 3400, "loss": 0.4238, "lr": 4.0812524109171476e-05, "epoch": 0.2419527732688126, "percentage": 58.09, "elapsed_time": "0:17:11", "remaining_time": "0:12:24", "throughput": 12515.04, "total_tokens": 12912816}
{"current_steps": 1980, "total_steps": 3400, "loss": 0.368, "lr": 4.0573616583307705e-05, "epoch": 0.2425653119353159, "percentage": 58.24, "elapsed_time": "0:18:10", "remaining_time": "0:13:02", "throughput": 11869.01, "total_tokens": 12946824}
{"current_steps": 1985, "total_steps": 3400, "loss": 0.3969, "lr": 4.033493199280202e-05, "epoch": 0.24317785060181923, "percentage": 58.38, "elapsed_time": "0:19:10", "remaining_time": "0:13:40", "throughput": 11280.77, "total_tokens": 12980496}
{"current_steps": 1990, "total_steps": 3400, "loss": 0.399, "lr": 4.009647598258022e-05, "epoch": 0.24379038926832255, "percentage": 58.53, "elapsed_time": "0:20:11", "remaining_time": "0:14:18", "throughput": 10745.78, "total_tokens": 13013696}
{"current_steps": 1995, "total_steps": 3400, "loss": 0.4231, "lr": 3.985825419216207e-05, "epoch": 0.24440292793482588, "percentage": 58.68, "elapsed_time": "0:21:10", "remaining_time": "0:14:54", "throughput": 10269.27, "total_tokens": 13048176}
{"current_steps": 2000, "total_steps": 3400, "loss": 0.3997, "lr": 3.962027225552807e-05, "epoch": 0.2450154666013292, "percentage": 58.82, "elapsed_time": "0:22:09", "remaining_time": "0:15:30", "throughput": 9835.99, "total_tokens": 13081512}
{"current_steps": 2000, "total_steps": 3400, "eval_loss": 0.3849605619907379, "epoch": 0.2450154666013292, "percentage": 58.82, "elapsed_time": "0:22:29", "remaining_time": "0:15:44", "throughput": 9694.02, "total_tokens": 13081512}
{"current_steps": 2005, "total_steps": 3400, "loss": 0.4169, "lr": 3.938253580098613e-05, "epoch": 0.24562800526783254, "percentage": 58.97, "elapsed_time": "0:23:36", "remaining_time": "0:16:25", "throughput": 9257.72, "total_tokens": 13114744}
{"current_steps": 2010, "total_steps": 3400, "loss": 0.4124, "lr": 3.914505045103845e-05, "epoch": 0.24624054393433586, "percentage": 59.12, "elapsed_time": "0:24:35", "remaining_time": "0:17:00", "throughput": 8908.82, "total_tokens": 13149336}
{"current_steps": 2015, "total_steps": 3400, "loss": 0.3863, "lr": 3.8907821822248605e-05, "epoch": 0.24685308260083919, "percentage": 59.26, "elapsed_time": "0:25:35", "remaining_time": "0:17:35", "throughput": 8586.19, "total_tokens": 13183000}
{"current_steps": 2020, "total_steps": 3400, "loss": 0.4316, "lr": 3.867085552510864e-05, "epoch": 0.2474656212673425, "percentage": 59.41, "elapsed_time": "0:26:36", "remaining_time": "0:18:10", "throughput": 8277.26, "total_tokens": 13216712}
{"current_steps": 2025, "total_steps": 3400, "loss": 0.437, "lr": 3.843415716390644e-05, "epoch": 0.24807815993384583, "percentage": 59.56, "elapsed_time": "0:27:36", "remaining_time": "0:18:45", "throughput": 7997.21, "total_tokens": 13250264}
{"current_steps": 2030, "total_steps": 3400, "loss": 0.429, "lr": 3.819773233659314e-05, "epoch": 0.24869069860034915, "percentage": 59.71, "elapsed_time": "0:28:36", "remaining_time": "0:19:18", "throughput": 7738.49, "total_tokens": 13283480}
{"current_steps": 2035, "total_steps": 3400, "loss": 0.3738, "lr": 3.7961586634650767e-05, "epoch": 0.24930323726685247, "percentage": 59.85, "elapsed_time": "0:29:37", "remaining_time": "0:19:52", "throughput": 7490.86, "total_tokens": 13316784}
{"current_steps": 2040, "total_steps": 3400, "loss": 0.4322, "lr": 3.772572564296005e-05, "epoch": 0.2499157759333558, "percentage": 60.0, "elapsed_time": "0:30:37", "remaining_time": "0:20:25", "throughput": 7264.75, "total_tokens": 13350408}
{"current_steps": 2045, "total_steps": 3400, "loss": 0.488, "lr": 3.749015493966817e-05, "epoch": 0.25052831459985914, "percentage": 60.15, "elapsed_time": "0:31:37", "remaining_time": "0:20:57", "throughput": 7052.84, "total_tokens": 13382992}
{"current_steps": 2005, "total_steps": 3400, "loss": 0.4169, "lr": 3.938253580098613e-05, "epoch": 0.24562800526783254, "percentage": 58.97, "elapsed_time": "0:02:21", "remaining_time": "0:01:38", "throughput": 92501.94, "total_tokens": 13114744}
{"current_steps": 2010, "total_steps": 3400, "loss": 0.4124, "lr": 3.914505045103845e-05, "epoch": 0.24624054393433586, "percentage": 59.12, "elapsed_time": "0:03:24", "remaining_time": "0:02:21", "throughput": 64398.12, "total_tokens": 13149336}
{"current_steps": 2015, "total_steps": 3400, "loss": 0.3863, "lr": 3.8907821822248605e-05, "epoch": 0.24685308260083919, "percentage": 59.26, "elapsed_time": "0:04:25", "remaining_time": "0:03:02", "throughput": 49659.05, "total_tokens": 13183000}
{"current_steps": 2020, "total_steps": 3400, "loss": 0.4316, "lr": 3.867085552510864e-05, "epoch": 0.2474656212673425, "percentage": 59.41, "elapsed_time": "0:05:26", "remaining_time": "0:03:43", "throughput": 40434.22, "total_tokens": 13216712}
{"current_steps": 2025, "total_steps": 3400, "loss": 0.437, "lr": 3.843415716390644e-05, "epoch": 0.24807815993384583, "percentage": 59.56, "elapsed_time": "0:06:28", "remaining_time": "0:04:23", "throughput": 34119.96, "total_tokens": 13250264}
{"current_steps": 2030, "total_steps": 3400, "loss": 0.429, "lr": 3.819773233659314e-05, "epoch": 0.24869069860034915, "percentage": 59.71, "elapsed_time": "0:07:28", "remaining_time": "0:05:02", "throughput": 29592.44, "total_tokens": 13283480}
{"current_steps": 2035, "total_steps": 3400, "loss": 0.3738, "lr": 3.7961586634650767e-05, "epoch": 0.24930323726685247, "percentage": 59.85, "elapsed_time": "0:08:29", "remaining_time": "0:05:41", "throughput": 26122.64, "total_tokens": 13316784}
{"current_steps": 2040, "total_steps": 3400, "loss": 0.4322, "lr": 3.772572564296005e-05, "epoch": 0.2499157759333558, "percentage": 60.0, "elapsed_time": "0:09:31", "remaining_time": "0:06:21", "throughput": 23348.68, "total_tokens": 13350408}
{"current_steps": 2045, "total_steps": 3400, "loss": 0.488, "lr": 3.749015493966817e-05, "epoch": 0.25052831459985914, "percentage": 60.15, "elapsed_time": "0:10:33", "remaining_time": "0:06:59", "throughput": 21141.3, "total_tokens": 13382992}
{"current_steps": 2050, "total_steps": 3400, "loss": 0.4267, "lr": 3.7254880096057073e-05, "epoch": 0.25114085326636243, "percentage": 60.29, "elapsed_time": "0:11:33", "remaining_time": "0:07:37", "throughput": 19333.43, "total_tokens": 13417048}
{"current_steps": 2050, "total_steps": 3400, "eval_loss": 0.494172602891922, "epoch": 0.25114085326636243, "percentage": 60.29, "elapsed_time": "0:12:22", "remaining_time": "0:08:08", "throughput": 18074.6, "total_tokens": 13417048}
{"current_steps": 2055, "total_steps": 3400, "loss": 0.499, "lr": 3.7019906676411446e-05, "epoch": 0.2517533919328658, "percentage": 60.44, "elapsed_time": "0:13:32", "remaining_time": "0:08:51", "throughput": 16554.18, "total_tokens": 13450600}
{"current_steps": 2060, "total_steps": 3400, "loss": 0.3934, "lr": 3.678524023788735e-05, "epoch": 0.25236593059936907, "percentage": 60.59, "elapsed_time": "0:14:33", "remaining_time": "0:09:28", "throughput": 15441.11, "total_tokens": 13483712}
{"current_steps": 2065, "total_steps": 3400, "loss": 0.5314, "lr": 3.6550886330380665e-05, "epoch": 0.2529784692658724, "percentage": 60.74, "elapsed_time": "0:15:34", "remaining_time": "0:10:03", "throughput": 14471.09, "total_tokens": 13517160}
{"current_steps": 2070, "total_steps": 3400, "loss": 0.4361, "lr": 3.631685049639586e-05, "epoch": 0.2535910079323757, "percentage": 60.88, "elapsed_time": "0:16:35", "remaining_time": "0:10:39", "throughput": 13616.71, "total_tokens": 13550536}
{"current_steps": 2075, "total_steps": 3400, "loss": 0.4021, "lr": 3.608313827091493e-05, "epoch": 0.25420354659887906, "percentage": 61.03, "elapsed_time": "0:17:37", "remaining_time": "0:11:15", "throughput": 12847.09, "total_tokens": 13584624}
{"current_steps": 2080, "total_steps": 3400, "loss": 0.4237, "lr": 3.5849755181266474e-05, "epoch": 0.25481608526538235, "percentage": 61.18, "elapsed_time": "0:18:38", "remaining_time": "0:11:50", "throughput": 12171.78, "total_tokens": 13618640}
{"current_steps": 2085, "total_steps": 3400, "loss": 0.4166, "lr": 3.5616706746995026e-05, "epoch": 0.2554286239318857, "percentage": 61.32, "elapsed_time": "0:19:39", "remaining_time": "0:12:24", "throughput": 11570.94, "total_tokens": 13652304}
{"current_steps": 2090, "total_steps": 3400, "loss": 0.3853, "lr": 3.538399847973036e-05, "epoch": 0.25604116259838905, "percentage": 61.47, "elapsed_time": "0:20:42", "remaining_time": "0:12:58", "throughput": 11015.2, "total_tokens": 13685632}
{"current_steps": 2095, "total_steps": 3400, "loss": 0.4844, "lr": 3.515163588305735e-05, "epoch": 0.25665370126489234, "percentage": 61.62, "elapsed_time": "0:21:43", "remaining_time": "0:13:31", "throughput": 10525.85, "total_tokens": 13719416}
{"current_steps": 2100, "total_steps": 3400, "loss": 0.4008, "lr": 3.491962445238569e-05, "epoch": 0.2572662399313957, "percentage": 61.76, "elapsed_time": "0:22:45", "remaining_time": "0:14:05", "throughput": 10073.4, "total_tokens": 13752680}
{"current_steps": 2100, "total_steps": 3400, "eval_loss": 0.35964828729629517, "epoch": 0.2572662399313957, "percentage": 61.76, "elapsed_time": "0:23:05", "remaining_time": "0:14:17", "throughput": 9926.61, "total_tokens": 13752680}
{"current_steps": 2105, "total_steps": 3400, "loss": 0.4508, "lr": 3.4687969674819906e-05, "epoch": 0.257878778597899, "percentage": 61.91, "elapsed_time": "0:24:12", "remaining_time": "0:14:53", "throughput": 9488.96, "total_tokens": 13785680}
{"current_steps": 2110, "total_steps": 3400, "loss": 0.3373, "lr": 3.445667702902969e-05, "epoch": 0.25849131726440233, "percentage": 62.06, "elapsed_time": "0:25:13", "remaining_time": "0:15:25", "throughput": 9132.93, "total_tokens": 13820392}
{"current_steps": 2115, "total_steps": 3400, "loss": 0.4516, "lr": 3.4225751985120215e-05, "epoch": 0.2591038559309056, "percentage": 62.21, "elapsed_time": "0:26:13", "remaining_time": "0:15:56", "throughput": 8802.5, "total_tokens": 13853688}
{"current_steps": 2120, "total_steps": 3400, "loss": 0.3864, "lr": 3.3995200004502816e-05, "epoch": 0.259716394597409, "percentage": 62.35, "elapsed_time": "0:27:14", "remaining_time": "0:16:27", "throughput": 8494.73, "total_tokens": 13887904}
{"current_steps": 2125, "total_steps": 3400, "loss": 0.3819, "lr": 3.3765026539765834e-05, "epoch": 0.26032893326391227, "percentage": 62.5, "elapsed_time": "0:28:14", "remaining_time": "0:16:56", "throughput": 8215.31, "total_tokens": 13921040}
{"current_steps": 2130, "total_steps": 3400, "loss": 0.3935, "lr": 3.3535237034545675e-05, "epoch": 0.2609414719304156, "percentage": 62.65, "elapsed_time": "0:29:13", "remaining_time": "0:17:25", "throughput": 7957.03, "total_tokens": 13954752}
{"current_steps": 2135, "total_steps": 3400, "loss": 0.4752, "lr": 3.330583692339802e-05, "epoch": 0.2615540105969189, "percentage": 62.79, "elapsed_time": "0:30:14", "remaining_time": "0:17:55", "throughput": 7708.81, "total_tokens": 13988376}
{"current_steps": 2140, "total_steps": 3400, "loss": 0.3697, "lr": 3.307683163166934e-05, "epoch": 0.26216654926342225, "percentage": 62.94, "elapsed_time": "0:31:14", "remaining_time": "0:18:23", "throughput": 7481.91, "total_tokens": 14023000}
{"current_steps": 2145, "total_steps": 3400, "loss": 0.448, "lr": 3.284822657536856e-05, "epoch": 0.2627790879299256, "percentage": 63.09, "elapsed_time": "0:32:13", "remaining_time": "0:18:51", "throughput": 7268.76, "total_tokens": 14056712}
{"current_steps": 2150, "total_steps": 3400, "loss": 0.3858, "lr": 3.262002716103897e-05, "epoch": 0.2633916265964289, "percentage": 63.24, "elapsed_time": "0:33:15", "remaining_time": "0:19:19", "throughput": 7062.7, "total_tokens": 14090488}
{"current_steps": 2150, "total_steps": 3400, "eval_loss": 0.31454333662986755, "epoch": 0.2633916265964289, "percentage": 63.24, "elapsed_time": "0:33:34", "remaining_time": "0:19:31", "throughput": 6993.99, "total_tokens": 14090488}
{"current_steps": 2155, "total_steps": 3400, "loss": 0.4647, "lr": 3.2392238785630386e-05, "epoch": 0.26400416526293224, "percentage": 63.38, "elapsed_time": "0:34:39", "remaining_time": "0:20:01", "throughput": 6792.13, "total_tokens": 14124000}
{"current_steps": 2160, "total_steps": 3400, "loss": 0.4581, "lr": 3.216486683637146e-05, "epoch": 0.26461670392943554, "percentage": 63.53, "elapsed_time": "0:35:39", "remaining_time": "0:20:28", "throughput": 6616.74, "total_tokens": 14157144}
{"current_steps": 2165, "total_steps": 3400, "loss": 0.4629, "lr": 3.1937916690642356e-05, "epoch": 0.2652292425959389, "percentage": 63.68, "elapsed_time": "0:36:40", "remaining_time": "0:20:55", "throughput": 6448.1, "total_tokens": 14191304}
{"current_steps": 2170, "total_steps": 3400, "loss": 0.5118, "lr": 3.1711393715847476e-05, "epoch": 0.2658417812624422, "percentage": 63.82, "elapsed_time": "0:37:39", "remaining_time": "0:21:20", "throughput": 6295.69, "total_tokens": 14225392}
{"current_steps": 2175, "total_steps": 3400, "loss": 0.3876, "lr": 3.14853032692886e-05, "epoch": 0.2664543199289455, "percentage": 63.97, "elapsed_time": "0:38:38", "remaining_time": "0:21:45", "throughput": 6149.34, "total_tokens": 14259152}
{"current_steps": 2180, "total_steps": 3400, "loss": 0.3996, "lr": 3.125965069803811e-05, "epoch": 0.2670668585954488, "percentage": 64.12, "elapsed_time": "0:39:39", "remaining_time": "0:22:11", "throughput": 6006.48, "total_tokens": 14293280}
{"current_steps": 2185, "total_steps": 3400, "loss": 0.3859, "lr": 3.103444133881261e-05, "epoch": 0.26767939726195217, "percentage": 64.26, "elapsed_time": "0:40:39", "remaining_time": "0:22:36", "throughput": 5873.66, "total_tokens": 14326560}
{"current_steps": 2190, "total_steps": 3400, "loss": 0.4347, "lr": 3.080968051784666e-05, "epoch": 0.26829193592845546, "percentage": 64.41, "elapsed_time": "0:41:39", "remaining_time": "0:23:00", "throughput": 5746.04, "total_tokens": 14359792}
{"current_steps": 2195, "total_steps": 3400, "loss": 0.5581, "lr": 3.058537355076683e-05, "epoch": 0.2689044745949588, "percentage": 64.56, "elapsed_time": "0:42:39", "remaining_time": "0:23:25", "throughput": 5622.33, "total_tokens": 14392976}
{"current_steps": 2200, "total_steps": 3400, "loss": 0.3598, "lr": 3.0361525742465973e-05, "epoch": 0.26951701326146216, "percentage": 64.71, "elapsed_time": "0:43:40", "remaining_time": "0:23:49", "throughput": 5505.77, "total_tokens": 14427600}
{"current_steps": 2200, "total_steps": 3400, "eval_loss": 0.3024410307407379, "epoch": 0.26951701326146216, "percentage": 64.71, "elapsed_time": "0:43:59", "remaining_time": "0:23:59", "throughput": 5465.02, "total_tokens": 14427600}
{"current_steps": 2205, "total_steps": 3400, "loss": 0.3613, "lr": 3.0138142386977787e-05, "epoch": 0.27012955192796545, "percentage": 64.85, "elapsed_time": "0:45:04", "remaining_time": "0:24:25", "throughput": 5346.54, "total_tokens": 14461704}
{"current_steps": 2210, "total_steps": 3400, "loss": 0.437, "lr": 2.991522876735154e-05, "epoch": 0.2707420905944688, "percentage": 65.0, "elapsed_time": "0:46:04", "remaining_time": "0:24:48", "throughput": 5242.82, "total_tokens": 14494512}
{"current_steps": 2215, "total_steps": 3400, "loss": 0.442, "lr": 2.9692790155527227e-05, "epoch": 0.2713546292609721, "percentage": 65.15, "elapsed_time": "0:47:04", "remaining_time": "0:25:11", "throughput": 5142.85, "total_tokens": 14527576}
{"current_steps": 2220, "total_steps": 3400, "loss": 0.4013, "lr": 2.9470831812210837e-05, "epoch": 0.27196716792747544, "percentage": 65.29, "elapsed_time": "0:48:04", "remaining_time": "0:25:33", "throughput": 5048.57, "total_tokens": 14560696}
{"current_steps": 2225, "total_steps": 3400, "loss": 0.4349, "lr": 2.924935898674992e-05, "epoch": 0.27257970659397873, "percentage": 65.44, "elapsed_time": "0:49:03", "remaining_time": "0:25:54", "throughput": 4958.14, "total_tokens": 14593944}
{"current_steps": 2230, "total_steps": 3400, "loss": 0.4662, "lr": 2.902837691700945e-05, "epoch": 0.2731922452604821, "percentage": 65.59, "elapsed_time": "0:50:04", "remaining_time": "0:26:16", "throughput": 4867.49, "total_tokens": 14626400}
{"current_steps": 2235, "total_steps": 3400, "loss": 0.4348, "lr": 2.880789082924798e-05, "epoch": 0.2738047839269854, "percentage": 65.74, "elapsed_time": "0:51:04", "remaining_time": "0:26:37", "throughput": 4784.08, "total_tokens": 14660184}
{"current_steps": 2240, "total_steps": 3400, "loss": 0.4075, "lr": 2.858790593799405e-05, "epoch": 0.2744173225934887, "percentage": 65.88, "elapsed_time": "0:52:03", "remaining_time": "0:26:57", "throughput": 4703.59, "total_tokens": 14692760}
{"current_steps": 2245, "total_steps": 3400, "loss": 0.3818, "lr": 2.8368427445922696e-05, "epoch": 0.275029861259992, "percentage": 66.03, "elapsed_time": "0:53:05", "remaining_time": "0:27:18", "throughput": 4623.69, "total_tokens": 14727408}
{"current_steps": 2250, "total_steps": 3400, "loss": 0.4352, "lr": 2.8149460543732664e-05, "epoch": 0.27564239992649536, "percentage": 66.18, "elapsed_time": "0:54:05", "remaining_time": "0:27:38", "throughput": 4548.38, "total_tokens": 14761264}
{"current_steps": 2250, "total_steps": 3400, "eval_loss": 0.2756429612636566, "epoch": 0.27564239992649536, "percentage": 66.18, "elapsed_time": "0:54:24", "remaining_time": "0:27:48", "throughput": 4521.17, "total_tokens": 14761264}
{"current_steps": 2255, "total_steps": 3400, "loss": 0.4691, "lr": 2.7931010410023518e-05, "epoch": 0.27625493859299866, "percentage": 66.32, "elapsed_time": "0:55:30", "remaining_time": "0:28:10", "throughput": 4442.42, "total_tokens": 14794600}
{"current_steps": 2260, "total_steps": 3400, "loss": 0.3836, "lr": 2.771308221117309e-05, "epoch": 0.276867477259502, "percentage": 66.47, "elapsed_time": "0:56:31", "remaining_time": "0:28:30", "throughput": 4372.69, "total_tokens": 14828632}
{"current_steps": 2265, "total_steps": 3400, "loss": 0.5366, "lr": 2.749568110121545e-05, "epoch": 0.27748001592600535, "percentage": 66.62, "elapsed_time": "0:57:30", "remaining_time": "0:28:49", "throughput": 4307.17, "total_tokens": 14862480}
{"current_steps": 2270, "total_steps": 3400, "loss": 0.4599, "lr": 2.7278812221718924e-05, "epoch": 0.27809255459250865, "percentage": 66.76, "elapsed_time": "0:58:29", "remaining_time": "0:29:07", "throughput": 4244.29, "total_tokens": 14896168}
{"current_steps": 2275, "total_steps": 3400, "loss": 0.3894, "lr": 2.7062480701664488e-05, "epoch": 0.278705093259012, "percentage": 66.91, "elapsed_time": "0:59:30", "remaining_time": "0:29:25", "throughput": 4181.68, "total_tokens": 14929688}
{"current_steps": 2280, "total_steps": 3400, "loss": 0.3963, "lr": 2.6846691657324473e-05, "epoch": 0.2793176319255153, "percentage": 67.06, "elapsed_time": "1:00:29", "remaining_time": "0:29:43", "throughput": 4122.32, "total_tokens": 14963456}
{"current_steps": 2285, "total_steps": 3400, "loss": 0.4045, "lr": 2.663145019214163e-05, "epoch": 0.27993017059201863, "percentage": 67.21, "elapsed_time": "1:01:29", "remaining_time": "0:30:00", "throughput": 4064.27, "total_tokens": 14996664}
{"current_steps": 2290, "total_steps": 3400, "loss": 0.4594, "lr": 2.6416761396608362e-05, "epoch": 0.2805427092585219, "percentage": 67.35, "elapsed_time": "1:02:31", "remaining_time": "0:30:18", "throughput": 4006.64, "total_tokens": 15030520}
{"current_steps": 2295, "total_steps": 3400, "loss": 0.4254, "lr": 2.6202630348146324e-05, "epoch": 0.2811552479250253, "percentage": 67.5, "elapsed_time": "1:03:30", "remaining_time": "0:30:34", "throughput": 3953.19, "total_tokens": 15064448}
{"current_steps": 2300, "total_steps": 3400, "loss": 0.3788, "lr": 2.598906211098643e-05, "epoch": 0.28176778659152857, "percentage": 67.65, "elapsed_time": "1:04:30", "remaining_time": "0:30:51", "throughput": 3900.94, "total_tokens": 15098096}
{"current_steps": 2300, "total_steps": 3400, "eval_loss": 0.23389142751693726, "epoch": 0.28176778659152857, "percentage": 67.65, "elapsed_time": "1:04:50", "remaining_time": "0:31:00", "throughput": 3880.91, "total_tokens": 15098096}
{"current_steps": 2305, "total_steps": 3400, "loss": 0.4748, "lr": 2.577606173604894e-05, "epoch": 0.2823803252580319, "percentage": 67.79, "elapsed_time": "1:05:56", "remaining_time": "0:31:19", "throughput": 3824.5, "total_tokens": 15132152}
{"current_steps": 2310, "total_steps": 3400, "loss": 0.4236, "lr": 2.5563634260824175e-05, "epoch": 0.2829928639245352, "percentage": 67.94, "elapsed_time": "1:06:56", "remaining_time": "0:31:35", "throughput": 3776.04, "total_tokens": 15165944}
{"current_steps": 2315, "total_steps": 3400, "loss": 0.3802, "lr": 2.535178470925323e-05, "epoch": 0.28360540259103856, "percentage": 68.09, "elapsed_time": "1:07:55", "remaining_time": "0:31:50", "throughput": 3729.22, "total_tokens": 15199864}
{"current_steps": 2320, "total_steps": 3400, "loss": 0.3804, "lr": 2.5140518091609256e-05, "epoch": 0.2842179412575419, "percentage": 68.24, "elapsed_time": "1:08:57", "remaining_time": "0:32:06", "throughput": 3681.79, "total_tokens": 15233216}
{"current_steps": 2325, "total_steps": 3400, "loss": 0.4197, "lr": 2.4929839404378936e-05, "epoch": 0.2848304799240452, "percentage": 68.38, "elapsed_time": "1:09:56", "remaining_time": "0:32:20", "throughput": 3637.68, "total_tokens": 15266328}
{"current_steps": 2330, "total_steps": 3400, "loss": 0.4293, "lr": 2.471975363014428e-05, "epoch": 0.28544301859054855, "percentage": 68.53, "elapsed_time": "1:10:56", "remaining_time": "0:32:34", "throughput": 3594.9, "total_tokens": 15300120}
{"current_steps": 2335, "total_steps": 3400, "loss": 0.3785, "lr": 2.451026573746482e-05, "epoch": 0.28605555725705184, "percentage": 68.68, "elapsed_time": "1:11:55", "remaining_time": "0:32:48", "throughput": 3553.17, "total_tokens": 15333584}
{"current_steps": 2340, "total_steps": 3400, "loss": 0.3369, "lr": 2.430138068076013e-05, "epoch": 0.2866680959235552, "percentage": 68.82, "elapsed_time": "1:12:55", "remaining_time": "0:33:01", "throughput": 3512.23, "total_tokens": 15366848}
{"current_steps": 2345, "total_steps": 3400, "loss": 0.4003, "lr": 2.4093103400192625e-05, "epoch": 0.2872806345900585, "percentage": 68.97, "elapsed_time": "1:13:53", "remaining_time": "0:33:14", "throughput": 3473.33, "total_tokens": 15400264}
{"current_steps": 2350, "total_steps": 3400, "loss": 0.4806, "lr": 2.388543882155067e-05, "epoch": 0.28789317325656183, "percentage": 69.12, "elapsed_time": "1:14:53", "remaining_time": "0:33:27", "throughput": 3434.98, "total_tokens": 15434176}
{"current_steps": 2350, "total_steps": 3400, "eval_loss": 0.28239962458610535, "epoch": 0.28789317325656183, "percentage": 69.12, "elapsed_time": "1:15:12", "remaining_time": "0:33:36", "throughput": 3419.98, "total_tokens": 15434176}
{"current_steps": 2355, "total_steps": 3400, "loss": 0.477, "lr": 2.3678391856132204e-05, "epoch": 0.2885057119230651, "percentage": 69.26, "elapsed_time": "1:16:18", "remaining_time": "0:33:51", "throughput": 3378.57, "total_tokens": 15467792}
{"current_steps": 2360, "total_steps": 3400, "loss": 0.4227, "lr": 2.3471967400628513e-05, "epoch": 0.28911825058956847, "percentage": 69.41, "elapsed_time": "1:17:17", "remaining_time": "0:34:03", "throughput": 3342.8, "total_tokens": 15501216}
{"current_steps": 2365, "total_steps": 3400, "loss": 0.4129, "lr": 2.3266170337008398e-05, "epoch": 0.28973078925607176, "percentage": 69.56, "elapsed_time": "1:18:17", "remaining_time": "0:34:15", "throughput": 3307.3, "total_tokens": 15534752}
{"current_steps": 2370, "total_steps": 3400, "loss": 0.3405, "lr": 2.306100553240274e-05, "epoch": 0.2903433279225751, "percentage": 69.71, "elapsed_time": "1:19:17", "remaining_time": "0:34:27", "throughput": 3272.53, "total_tokens": 15568376}
{"current_steps": 2375, "total_steps": 3400, "loss": 0.4279, "lr": 2.2856477838989456e-05, "epoch": 0.29095586658907846, "percentage": 69.85, "elapsed_time": "1:20:16", "remaining_time": "0:34:38", "throughput": 3239.2, "total_tokens": 15601568}
{"current_steps": 2380, "total_steps": 3400, "loss": 0.3586, "lr": 2.2652592093878666e-05, "epoch": 0.29156840525558175, "percentage": 70.0, "elapsed_time": "1:21:15", "remaining_time": "0:34:49", "throughput": 3207.07, "total_tokens": 15635584}
{"current_steps": 2385, "total_steps": 3400, "loss": 0.4119, "lr": 2.244935311899829e-05, "epoch": 0.2921809439220851, "percentage": 70.15, "elapsed_time": "1:22:15", "remaining_time": "0:35:00", "throughput": 3175.11, "total_tokens": 15669544}
{"current_steps": 2390, "total_steps": 3400, "loss": 0.3929, "lr": 2.224676572098007e-05, "epoch": 0.2927934825885884, "percentage": 70.29, "elapsed_time": "1:23:13", "remaining_time": "0:35:10", "throughput": 3144.54, "total_tokens": 15703472}
{"current_steps": 2395, "total_steps": 3400, "loss": 0.4627, "lr": 2.2044834691045873e-05, "epoch": 0.29340602125509174, "percentage": 70.44, "elapsed_time": "1:24:12", "remaining_time": "0:35:20", "throughput": 3114.67, "total_tokens": 15737312}
{"current_steps": 2400, "total_steps": 3400, "loss": 0.3893, "lr": 2.184356480489432e-05, "epoch": 0.29401855992159504, "percentage": 70.59, "elapsed_time": "1:25:12", "remaining_time": "0:35:30", "throughput": 3085.0, "total_tokens": 15771224}
{"current_steps": 2400, "total_steps": 3400, "eval_loss": 0.22477610409259796, "epoch": 0.29401855992159504, "percentage": 70.59, "elapsed_time": "1:25:31", "remaining_time": "0:35:38", "throughput": 3073.23, "total_tokens": 15771224}
{"current_steps": 2405, "total_steps": 3400, "loss": 0.3744, "lr": 2.1642960822587878e-05, "epoch": 0.2946310985880984, "percentage": 70.74, "elapsed_time": "1:26:36", "remaining_time": "0:35:49", "throughput": 3041.46, "total_tokens": 15804752}
{"current_steps": 2410, "total_steps": 3400, "loss": 0.4134, "lr": 2.1443027488440338e-05, "epoch": 0.2952436372546017, "percentage": 70.88, "elapsed_time": "1:27:35", "remaining_time": "0:35:59", "throughput": 3013.49, "total_tokens": 15838624}
{"current_steps": 2415, "total_steps": 3400, "loss": 0.4031, "lr": 2.124376953090456e-05, "epoch": 0.295856175921105, "percentage": 71.03, "elapsed_time": "1:28:37", "remaining_time": "0:36:08", "throughput": 2984.78, "total_tokens": 15872168}
{"current_steps": 2420, "total_steps": 3400, "loss": 0.3465, "lr": 2.104519166246059e-05, "epoch": 0.2964687145876083, "percentage": 71.18, "elapsed_time": "1:29:37", "remaining_time": "0:36:17", "throughput": 2957.97, "total_tokens": 15905944}
{"current_steps": 2425, "total_steps": 3400, "loss": 0.4785, "lr": 2.0847298579504344e-05, "epoch": 0.29708125325411167, "percentage": 71.32, "elapsed_time": "1:30:37", "remaining_time": "0:36:26", "throughput": 2931.57, "total_tokens": 15939272}
{"current_steps": 2430, "total_steps": 3400, "loss": 0.3268, "lr": 2.065009496223638e-05, "epoch": 0.297693791920615, "percentage": 71.47, "elapsed_time": "1:31:38", "remaining_time": "0:36:34", "throughput": 2904.89, "total_tokens": 15973248}
{"current_steps": 2435, "total_steps": 3400, "loss": 0.3917, "lr": 2.045358547455138e-05, "epoch": 0.2983063305871183, "percentage": 71.62, "elapsed_time": "1:32:38", "remaining_time": "0:36:42", "throughput": 2879.85, "total_tokens": 16006744}
{"current_steps": 2440, "total_steps": 3400, "loss": 0.4549, "lr": 2.0257774763927655e-05, "epoch": 0.29891886925362166, "percentage": 71.76, "elapsed_time": "1:33:37", "remaining_time": "0:36:50", "throughput": 2855.42, "total_tokens": 16039672}
{"current_steps": 2445, "total_steps": 3400, "loss": 0.3082, "lr": 2.0062667461317426e-05, "epoch": 0.29953140792012495, "percentage": 71.91, "elapsed_time": "1:34:37", "remaining_time": "0:36:57", "throughput": 2831.03, "total_tokens": 16073712}
{"current_steps": 2450, "total_steps": 3400, "loss": 0.4445, "lr": 1.9868268181037185e-05, "epoch": 0.3001439465866283, "percentage": 72.06, "elapsed_time": "1:35:37", "remaining_time": "0:37:04", "throughput": 2807.28, "total_tokens": 16106832}
{"current_steps": 2450, "total_steps": 3400, "eval_loss": 0.26860639452934265, "epoch": 0.3001439465866283, "percentage": 72.06, "elapsed_time": "1:35:56", "remaining_time": "0:37:12", "throughput": 2797.8, "total_tokens": 16106832}
{"current_steps": 2455, "total_steps": 3400, "loss": 0.4825, "lr": 1.967458152065857e-05, "epoch": 0.3007564852531316, "percentage": 72.21, "elapsed_time": "1:37:02", "remaining_time": "0:37:21", "throughput": 2772.14, "total_tokens": 16139896}
{"current_steps": 2460, "total_steps": 3400, "loss": 0.482, "lr": 1.9481612060899646e-05, "epoch": 0.30136902391963494, "percentage": 72.35, "elapsed_time": "1:38:02", "remaining_time": "0:37:27", "throughput": 2749.26, "total_tokens": 16172968}
{"current_steps": 2465, "total_steps": 3400, "loss": 0.4084, "lr": 1.928936436551661e-05, "epoch": 0.30198156258613823, "percentage": 72.5, "elapsed_time": "1:39:03", "remaining_time": "0:37:34", "throughput": 2726.7, "total_tokens": 16205904}
{"current_steps": 2470, "total_steps": 3400, "loss": 0.3349, "lr": 1.9097842981195834e-05, "epoch": 0.3025941012526416, "percentage": 72.65, "elapsed_time": "1:40:03", "remaining_time": "0:37:40", "throughput": 2705.25, "total_tokens": 16239760}
{"current_steps": 2475, "total_steps": 3400, "loss": 0.4111, "lr": 1.8907052437446272e-05, "epoch": 0.30320663991914487, "percentage": 72.79, "elapsed_time": "1:41:02", "remaining_time": "0:37:45", "throughput": 2684.14, "total_tokens": 16273064}
{"current_steps": 2480, "total_steps": 3400, "loss": 0.4171, "lr": 1.871699724649244e-05, "epoch": 0.3038191785856482, "percentage": 72.94, "elapsed_time": "1:42:03", "remaining_time": "0:37:51", "throughput": 2663.05, "total_tokens": 16306760}
{"current_steps": 2485, "total_steps": 3400, "loss": 0.4116, "lr": 1.8527681903167644e-05, "epoch": 0.30443171725215157, "percentage": 73.09, "elapsed_time": "1:43:02", "remaining_time": "0:37:56", "throughput": 2642.72, "total_tokens": 16339720}
{"current_steps": 2490, "total_steps": 3400, "loss": 0.4053, "lr": 1.833911088480767e-05, "epoch": 0.30504425591865486, "percentage": 73.24, "elapsed_time": "1:44:02", "remaining_time": "0:38:01", "throughput": 2622.83, "total_tokens": 16373008}
{"current_steps": 2495, "total_steps": 3400, "loss": 0.3783, "lr": 1.8151288651144893e-05, "epoch": 0.3056567945851582, "percentage": 73.38, "elapsed_time": "1:45:03", "remaining_time": "0:38:06", "throughput": 2602.78, "total_tokens": 16406344}
{"current_steps": 2500, "total_steps": 3400, "loss": 0.4099, "lr": 1.796421964420285e-05, "epoch": 0.3062693332516615, "percentage": 73.53, "elapsed_time": "1:46:02", "remaining_time": "0:38:10", "throughput": 2583.66, "total_tokens": 16439616}
{"current_steps": 2500, "total_steps": 3400, "eval_loss": 0.2817166745662689, "epoch": 0.3062693332516615, "percentage": 73.53, "elapsed_time": "1:46:22", "remaining_time": "0:38:17", "throughput": 2575.7, "total_tokens": 16439616}
{"current_steps": 2505, "total_steps": 3400, "loss": 0.4116, "lr": 1.7777908288191176e-05, "epoch": 0.30688187191816485, "percentage": 73.68, "elapsed_time": "1:47:26", "remaining_time": "0:38:23", "throughput": 2555.29, "total_tokens": 16473632}
{"current_steps": 2510, "total_steps": 3400, "loss": 0.3251, "lr": 1.7592358989400883e-05, "epoch": 0.30749441058466814, "percentage": 73.82, "elapsed_time": "1:48:27", "remaining_time": "0:38:27", "throughput": 2536.4, "total_tokens": 16506688}
{"current_steps": 2515, "total_steps": 3400, "loss": 0.3927, "lr": 1.740757613610028e-05, "epoch": 0.3081069492511715, "percentage": 73.97, "elapsed_time": "1:49:27", "remaining_time": "0:38:30", "throughput": 2518.49, "total_tokens": 16539672}
{"current_steps": 2520, "total_steps": 3400, "loss": 0.4169, "lr": 1.7223564098431067e-05, "epoch": 0.3087194879176748, "percentage": 74.12, "elapsed_time": "1:50:26", "remaining_time": "0:38:34", "throughput": 2501.05, "total_tokens": 16573248}
{"current_steps": 2525, "total_steps": 3400, "loss": 0.3538, "lr": 1.704032722830512e-05, "epoch": 0.30933202658417813, "percentage": 74.26, "elapsed_time": "1:51:27", "remaining_time": "0:38:37", "throughput": 2483.17, "total_tokens": 16607280}
{"current_steps": 2530, "total_steps": 3400, "loss": 0.4374, "lr": 1.68578698593014e-05, "epoch": 0.3099445652506814, "percentage": 74.41, "elapsed_time": "1:52:27", "remaining_time": "0:38:40", "throughput": 2466.13, "total_tokens": 16641208}
{"current_steps": 2535, "total_steps": 3400, "loss": 0.3771, "lr": 1.6676196306563613e-05, "epoch": 0.3105571039171848, "percentage": 74.56, "elapsed_time": "1:53:27", "remaining_time": "0:38:43", "throughput": 2449.18, "total_tokens": 16673784}
{"current_steps": 2540, "total_steps": 3400, "loss": 0.419, "lr": 1.6495310866698093e-05, "epoch": 0.3111696425836881, "percentage": 74.71, "elapsed_time": "1:54:28", "remaining_time": "0:38:45", "throughput": 2432.27, "total_tokens": 16707264}
{"current_steps": 2545, "total_steps": 3400, "loss": 0.2976, "lr": 1.631521781767214e-05, "epoch": 0.3117821812501914, "percentage": 74.85, "elapsed_time": "1:55:28", "remaining_time": "0:38:47", "throughput": 2416.32, "total_tokens": 16741208}
{"current_steps": 2550, "total_steps": 3400, "loss": 0.4017, "lr": 1.6135921418712956e-05, "epoch": 0.31239471991669476, "percentage": 75.0, "elapsed_time": "1:56:28", "remaining_time": "0:38:49", "throughput": 2400.45, "total_tokens": 16774544}
{"current_steps": 2550, "total_steps": 3400, "eval_loss": 0.2488662749528885, "epoch": 0.31239471991669476, "percentage": 75.0, "elapsed_time": "1:56:47", "remaining_time": "0:38:55", "throughput": 2393.76, "total_tokens": 16774544}
{"current_steps": 2555, "total_steps": 3400, "loss": 0.3832, "lr": 1.5957425910206785e-05, "epoch": 0.31300725858319806, "percentage": 75.15, "elapsed_time": "1:57:53", "remaining_time": "0:38:59", "throughput": 2376.11, "total_tokens": 16807880}
{"current_steps": 2560, "total_steps": 3400, "loss": 0.452, "lr": 1.577973551359877e-05, "epoch": 0.3136197972497014, "percentage": 75.29, "elapsed_time": "1:58:52", "remaining_time": "0:39:00", "throughput": 2360.99, "total_tokens": 16840456}
{"current_steps": 2565, "total_steps": 3400, "loss": 0.4162, "lr": 1.560285443129296e-05, "epoch": 0.3142323359162047, "percentage": 75.44, "elapsed_time": "1:59:51", "remaining_time": "0:39:01", "throughput": 2346.4, "total_tokens": 16874408}
{"current_steps": 2570, "total_steps": 3400, "loss": 0.38, "lr": 1.542678684655306e-05, "epoch": 0.31484487458270805, "percentage": 75.59, "elapsed_time": "2:00:51", "remaining_time": "0:39:01", "throughput": 2331.77, "total_tokens": 16908120}
{"current_steps": 2575, "total_steps": 3400, "loss": 0.4196, "lr": 1.5251536923403426e-05, "epoch": 0.31545741324921134, "percentage": 75.74, "elapsed_time": "2:01:49", "remaining_time": "0:39:01", "throughput": 2317.68, "total_tokens": 16941248}
{"current_steps": 2580, "total_steps": 3400, "loss": 0.3956, "lr": 1.5077108806530581e-05, "epoch": 0.3160699519157147, "percentage": 75.88, "elapsed_time": "2:02:48", "remaining_time": "0:39:01", "throughput": 2303.9, "total_tokens": 16975272}
{"current_steps": 2585, "total_steps": 3400, "loss": 0.4497, "lr": 1.4903506621185192e-05, "epoch": 0.316682490582218, "percentage": 76.03, "elapsed_time": "2:03:47", "remaining_time": "0:39:01", "throughput": 2289.76, "total_tokens": 17008144}
{"current_steps": 2590, "total_steps": 3400, "loss": 0.4013, "lr": 1.4730734473084568e-05, "epoch": 0.3172950292487213, "percentage": 76.18, "elapsed_time": "2:04:47", "remaining_time": "0:39:01", "throughput": 2276.09, "total_tokens": 17042320}
{"current_steps": 2595, "total_steps": 3400, "loss": 0.3767, "lr": 1.4558796448315504e-05, "epoch": 0.3179075679152247, "percentage": 76.32, "elapsed_time": "2:05:46", "remaining_time": "0:39:00", "throughput": 2262.73, "total_tokens": 17075336}
{"current_steps": 2600, "total_steps": 3400, "loss": 0.4698, "lr": 1.4387696613237612e-05, "epoch": 0.31852010658172797, "percentage": 76.47, "elapsed_time": "2:06:45", "remaining_time": "0:39:00", "throughput": 2249.61, "total_tokens": 17108368}
{"current_steps": 2600, "total_steps": 3400, "eval_loss": 0.20822136104106903, "epoch": 0.31852010658172797, "percentage": 76.47, "elapsed_time": "2:07:04", "remaining_time": "0:39:05", "throughput": 2243.94, "total_tokens": 17108368}
{"current_steps": 2605, "total_steps": 3400, "loss": 0.3813, "lr": 1.4217439014387251e-05, "epoch": 0.3191326452482313, "percentage": 76.62, "elapsed_time": "2:08:09", "remaining_time": "0:39:06", "throughput": 2229.26, "total_tokens": 17141208}
{"current_steps": 2610, "total_steps": 3400, "loss": 0.4327, "lr": 1.404802767838176e-05, "epoch": 0.3197451839147346, "percentage": 76.76, "elapsed_time": "2:09:07", "remaining_time": "0:39:05", "throughput": 2216.7, "total_tokens": 17174376}
{"current_steps": 2615, "total_steps": 3400, "loss": 0.4071, "lr": 1.3879466611824199e-05, "epoch": 0.32035772258123796, "percentage": 76.91, "elapsed_time": "2:10:06", "remaining_time": "0:39:03", "throughput": 2204.21, "total_tokens": 17207208}
{"current_steps": 2620, "total_steps": 3400, "loss": 0.3462, "lr": 1.371175980120864e-05, "epoch": 0.32097026124774125, "percentage": 77.06, "elapsed_time": "2:11:06", "remaining_time": "0:39:01", "throughput": 2191.75, "total_tokens": 17240704}
{"current_steps": 2625, "total_steps": 3400, "loss": 0.3604, "lr": 1.3544911212825906e-05, "epoch": 0.3215827999142446, "percentage": 77.21, "elapsed_time": "2:12:04", "remaining_time": "0:38:59", "throughput": 2179.8, "total_tokens": 17274096}
{"current_steps": 2630, "total_steps": 3400, "loss": 0.4219, "lr": 1.337892479266974e-05, "epoch": 0.3221953385807479, "percentage": 77.35, "elapsed_time": "2:13:03", "remaining_time": "0:38:57", "throughput": 2167.82, "total_tokens": 17306760}
{"current_steps": 2635, "total_steps": 3400, "loss": 0.3395, "lr": 1.3213804466343421e-05, "epoch": 0.32280787724725124, "percentage": 77.5, "elapsed_time": "2:14:03", "remaining_time": "0:38:55", "throughput": 2155.88, "total_tokens": 17340600}
{"current_steps": 2640, "total_steps": 3400, "loss": 0.4356, "lr": 1.3049554138967051e-05, "epoch": 0.32342041591375453, "percentage": 77.65, "elapsed_time": "2:15:02", "remaining_time": "0:38:52", "throughput": 2144.25, "total_tokens": 17373800}
{"current_steps": 2645, "total_steps": 3400, "loss": 0.3651, "lr": 1.2886177695085078e-05, "epoch": 0.3240329545802579, "percentage": 77.79, "elapsed_time": "2:16:01", "remaining_time": "0:38:49", "throughput": 2132.91, "total_tokens": 17407296}
{"current_steps": 2650, "total_steps": 3400, "loss": 0.3722, "lr": 1.2723678998574512e-05, "epoch": 0.3246454932467612, "percentage": 77.94, "elapsed_time": "2:17:02", "remaining_time": "0:38:47", "throughput": 2121.15, "total_tokens": 17440920}
{"current_steps": 2650, "total_steps": 3400, "eval_loss": 0.23050636053085327, "epoch": 0.3246454932467612, "percentage": 77.94, "elapsed_time": "2:17:22", "remaining_time": "0:38:52", "throughput": 2115.97, "total_tokens": 17440920}
{"current_steps": 2655, "total_steps": 3400, "loss": 0.3658, "lr": 1.2562061892553473e-05, "epoch": 0.3252580319132645, "percentage": 78.09, "elapsed_time": "2:18:27", "remaining_time": "0:38:51", "throughput": 2103.48, "total_tokens": 17473920}
{"current_steps": 2660, "total_steps": 3400, "loss": 0.4345, "lr": 1.2401330199290367e-05, "epoch": 0.32587057057976787, "percentage": 78.24, "elapsed_time": "2:19:26", "remaining_time": "0:38:47", "throughput": 2092.59, "total_tokens": 17506776}
{"current_steps": 2665, "total_steps": 3400, "loss": 0.4469, "lr": 1.224148772011346e-05, "epoch": 0.32648310924627116, "percentage": 78.38, "elapsed_time": "2:20:25", "remaining_time": "0:38:43", "throughput": 2081.7, "total_tokens": 17540248}
{"current_steps": 2670, "total_steps": 3400, "loss": 0.4398, "lr": 1.2082538235320929e-05, "epoch": 0.3270956479127745, "percentage": 78.53, "elapsed_time": "2:21:24", "remaining_time": "0:38:39", "throughput": 2071.32, "total_tokens": 17573776}
{"current_steps": 2675, "total_steps": 3400, "loss": 0.387, "lr": 1.1924485504091565e-05, "epoch": 0.3277081865792778, "percentage": 78.68, "elapsed_time": "2:22:22", "remaining_time": "0:38:35", "throughput": 2061.1, "total_tokens": 17607368}
{"current_steps": 2680, "total_steps": 3400, "loss": 0.3381, "lr": 1.1767333264395736e-05, "epoch": 0.32832072524578115, "percentage": 78.82, "elapsed_time": "2:23:22", "remaining_time": "0:38:31", "throughput": 2050.79, "total_tokens": 17641536}
{"current_steps": 2685, "total_steps": 3400, "loss": 0.3718, "lr": 1.1611085232907132e-05, "epoch": 0.32893326391228445, "percentage": 78.97, "elapsed_time": "2:24:20", "remaining_time": "0:38:26", "throughput": 2040.86, "total_tokens": 17674656}
{"current_steps": 2690, "total_steps": 3400, "loss": 0.3983, "lr": 1.14557451049147e-05, "epoch": 0.3295458025787878, "percentage": 79.12, "elapsed_time": "2:25:18", "remaining_time": "0:38:21", "throughput": 2031.12, "total_tokens": 17708072}
{"current_steps": 2695, "total_steps": 3400, "loss": 0.3805, "lr": 1.1301316554235397e-05, "epoch": 0.3301583412452911, "percentage": 79.26, "elapsed_time": "2:26:17", "remaining_time": "0:38:16", "throughput": 2021.29, "total_tokens": 17741152}
{"current_steps": 2700, "total_steps": 3400, "loss": 0.4018, "lr": 1.114780323312724e-05, "epoch": 0.33077087991179444, "percentage": 79.41, "elapsed_time": "2:27:16", "remaining_time": "0:38:10", "throughput": 2011.49, "total_tokens": 17774336}
{"current_steps": 2700, "total_steps": 3400, "eval_loss": 0.20385108888149261, "epoch": 0.33077087991179444, "percentage": 79.41, "elapsed_time": "2:27:35", "remaining_time": "0:38:15", "throughput": 2007.05, "total_tokens": 17774336}
{"current_steps": 2705, "total_steps": 3400, "loss": 0.3834, "lr": 1.0995208772202897e-05, "epoch": 0.33138341857829773, "percentage": 79.56, "elapsed_time": "2:28:40", "remaining_time": "0:38:11", "throughput": 1996.47, "total_tokens": 17808736}
{"current_steps": 2710, "total_steps": 3400, "loss": 0.3883, "lr": 1.0843536780343865e-05, "epoch": 0.3319959572448011, "percentage": 79.71, "elapsed_time": "2:29:39", "remaining_time": "0:38:06", "throughput": 1987.07, "total_tokens": 17842056}
{"current_steps": 2715, "total_steps": 3400, "loss": 0.3865, "lr": 1.069279084461513e-05, "epoch": 0.3326084959113044, "percentage": 79.85, "elapsed_time": "2:30:37", "remaining_time": "0:38:00", "throughput": 1977.93, "total_tokens": 17876008}
{"current_steps": 2720, "total_steps": 3400, "loss": 0.4189, "lr": 1.0542974530180327e-05, "epoch": 0.3332210345778077, "percentage": 80.0, "elapsed_time": "2:31:36", "remaining_time": "0:37:54", "throughput": 1968.88, "total_tokens": 17909480}
{"current_steps": 2725, "total_steps": 3400, "loss": 0.3897, "lr": 1.0394091380217352e-05, "epoch": 0.33383357324431107, "percentage": 80.15, "elapsed_time": "2:32:35", "remaining_time": "0:37:47", "throughput": 1959.76, "total_tokens": 17942952}
{"current_steps": 2730, "total_steps": 3400, "loss": 0.3536, "lr": 1.0246144915834683e-05, "epoch": 0.33444611191081436, "percentage": 80.29, "elapsed_time": "2:33:35", "remaining_time": "0:37:41", "throughput": 1950.77, "total_tokens": 17976368}
{"current_steps": 2735, "total_steps": 3400, "loss": 0.3618, "lr": 1.0099138635988026e-05, "epoch": 0.3350586505773177, "percentage": 80.44, "elapsed_time": "2:34:33", "remaining_time": "0:37:34", "throughput": 1942.01, "total_tokens": 18009520}
{"current_steps": 2740, "total_steps": 3400, "loss": 0.3658, "lr": 9.953076017397578e-06, "epoch": 0.335671189243821, "percentage": 80.59, "elapsed_time": "2:35:32", "remaining_time": "0:37:28", "throughput": 1933.25, "total_tokens": 18042664}
{"current_steps": 2745, "total_steps": 3400, "loss": 0.3484, "lr": 9.807960514465792e-06, "epoch": 0.33628372791032435, "percentage": 80.74, "elapsed_time": "2:36:32", "remaining_time": "0:37:21", "throughput": 1924.61, "total_tokens": 18076976}
{"current_steps": 2750, "total_steps": 3400, "loss": 0.3938, "lr": 9.663795559195733e-06, "epoch": 0.33689626657682764, "percentage": 80.88, "elapsed_time": "2:37:31", "remaining_time": "0:37:13", "throughput": 1916.18, "total_tokens": 18110672}
{"current_steps": 2750, "total_steps": 3400, "eval_loss": 0.18945012986660004, "epoch": 0.33689626657682764, "percentage": 80.88, "elapsed_time": "2:37:50", "remaining_time": "0:37:18", "throughput": 1912.24, "total_tokens": 18110672}
{"current_steps": 2755, "total_steps": 3400, "loss": 0.3582, "lr": 9.520584561109864e-06, "epoch": 0.337508805243331, "percentage": 81.03, "elapsed_time": "2:38:55", "remaining_time": "0:37:12", "throughput": 1902.8, "total_tokens": 18143336}
{"current_steps": 2755, "total_steps": 3400, "loss": 0.3582, "lr": 9.520584561109864e-06, "epoch": 0.337508805243331, "percentage": 81.03, "elapsed_time": "0:02:16", "remaining_time": "0:00:31", "throughput": 132920.88, "total_tokens": 18143336}
{"current_steps": 2760, "total_steps": 3400, "loss": 0.3499, "lr": 9.378330907169386e-06, "epoch": 0.3381213439098343, "percentage": 81.18, "elapsed_time": "0:03:16", "remaining_time": "0:00:45", "throughput": 92468.56, "total_tokens": 18177216}
{"current_steps": 2765, "total_steps": 3400, "loss": 0.3259, "lr": 9.237037961694223e-06, "epoch": 0.33873388257633763, "percentage": 81.32, "elapsed_time": "0:04:15", "remaining_time": "0:00:58", "throughput": 71210.16, "total_tokens": 18210408}
{"current_steps": 2770, "total_steps": 3400, "loss": 0.4411, "lr": 9.096709066283354e-06, "epoch": 0.339346421242841, "percentage": 81.47, "elapsed_time": "0:05:15", "remaining_time": "0:01:11", "throughput": 57900.13, "total_tokens": 18244544}
{"current_steps": 2775, "total_steps": 3400, "loss": 0.3333, "lr": 8.957347539735872e-06, "epoch": 0.33995895990934427, "percentage": 81.62, "elapsed_time": "0:06:14", "remaining_time": "0:01:24", "throughput": 48777.7, "total_tokens": 18278472}
{"current_steps": 2780, "total_steps": 3400, "loss": 0.3915, "lr": 8.818956677972406e-06, "epoch": 0.3405714985758476, "percentage": 81.76, "elapsed_time": "0:07:13", "remaining_time": "0:01:36", "throughput": 42268.65, "total_tokens": 18312176}
{"current_steps": 2785, "total_steps": 3400, "loss": 0.3634, "lr": 8.681539753957269e-06, "epoch": 0.3411840372423509, "percentage": 81.91, "elapsed_time": "0:08:12", "remaining_time": "0:01:48", "throughput": 37258.98, "total_tokens": 18346400}
{"current_steps": 2790, "total_steps": 3400, "loss": 0.4182, "lr": 8.545100017620988e-06, "epoch": 0.34179657590885426, "percentage": 82.06, "elapsed_time": "0:09:12", "remaining_time": "0:02:00", "throughput": 33294.54, "total_tokens": 18379800}
{"current_steps": 2795, "total_steps": 3400, "loss": 0.3936, "lr": 8.409640695783443e-06, "epoch": 0.34240911457535755, "percentage": 82.21, "elapsed_time": "0:10:10", "remaining_time": "0:02:12", "throughput": 30151.19, "total_tokens": 18412976}
{"current_steps": 2800, "total_steps": 3400, "loss": 0.4413, "lr": 8.275164992077556e-06, "epoch": 0.3430216532418609, "percentage": 82.35, "elapsed_time": "0:11:09", "remaining_time": "0:02:23", "throughput": 27552.54, "total_tokens": 18446936}
{"current_steps": 2800, "total_steps": 3400, "eval_loss": 0.1981697380542755, "epoch": 0.3430216532418609, "percentage": 82.35, "elapsed_time": "0:11:56", "remaining_time": "0:02:33", "throughput": 25752.43, "total_tokens": 18446936}
{"current_steps": 2805, "total_steps": 3400, "loss": 0.3995, "lr": 8.141676086873572e-06, "epoch": 0.3436341919083642, "percentage": 82.5, "elapsed_time": "0:13:04", "remaining_time": "0:02:46", "throughput": 23552.9, "total_tokens": 18480624}
{"current_steps": 2810, "total_steps": 3400, "loss": 0.3649, "lr": 8.009177137203794e-06, "epoch": 0.34424673057486754, "percentage": 82.65, "elapsed_time": "0:14:03", "remaining_time": "0:02:57", "throughput": 21951.22, "total_tokens": 18514456}
{"current_steps": 2815, "total_steps": 3400, "loss": 0.3796, "lr": 7.877671276687898e-06, "epoch": 0.34485926924137084, "percentage": 82.79, "elapsed_time": "0:15:02", "remaining_time": "0:03:07", "throughput": 20554.88, "total_tokens": 18548808}
{"current_steps": 2820, "total_steps": 3400, "loss": 0.481, "lr": 7.747161615458902e-06, "epoch": 0.3454718079078742, "percentage": 82.94, "elapsed_time": "0:16:02", "remaining_time": "0:03:17", "throughput": 19312.3, "total_tokens": 18582240}
{"current_steps": 2825, "total_steps": 3400, "loss": 0.3266, "lr": 7.617651240089546e-06, "epoch": 0.34608434657437753, "percentage": 83.09, "elapsed_time": "0:17:02", "remaining_time": "0:03:28", "throughput": 18211.52, "total_tokens": 18616176}
{"current_steps": 2830, "total_steps": 3400, "loss": 0.3436, "lr": 7.489143213519301e-06, "epoch": 0.3466968852408808, "percentage": 83.24, "elapsed_time": "0:18:01", "remaining_time": "0:03:37", "throughput": 17240.98, "total_tokens": 18650088}
{"current_steps": 2835, "total_steps": 3400, "loss": 0.3959, "lr": 7.361640574981937e-06, "epoch": 0.3473094239073842, "percentage": 83.38, "elapsed_time": "0:19:01", "remaining_time": "0:03:47", "throughput": 16370.69, "total_tokens": 18683096}
{"current_steps": 2840, "total_steps": 3400, "loss": 0.348, "lr": 7.2351463399336735e-06, "epoch": 0.34792196257388747, "percentage": 83.53, "elapsed_time": "0:20:01", "remaining_time": "0:03:56", "throughput": 15575.54, "total_tokens": 18716984}
{"current_steps": 2845, "total_steps": 3400, "loss": 0.3712, "lr": 7.109663499981834e-06, "epoch": 0.3485345012403908, "percentage": 83.68, "elapsed_time": "0:21:01", "remaining_time": "0:04:06", "throughput": 14867.11, "total_tokens": 18750952}
{"current_steps": 2850, "total_steps": 3400, "loss": 0.4121, "lr": 6.985195022814067e-06, "epoch": 0.3491470399068941, "percentage": 83.82, "elapsed_time": "0:22:00", "remaining_time": "0:04:14", "throughput": 14220.54, "total_tokens": 18784560}
{"current_steps": 2850, "total_steps": 3400, "eval_loss": 0.19169984757900238, "epoch": 0.3491470399068941, "percentage": 83.82, "elapsed_time": "0:22:20", "remaining_time": "0:04:18", "throughput": 14010.43, "total_tokens": 18784560}
{"current_steps": 2855, "total_steps": 3400, "loss": 0.3724, "lr": 6.861743852128233e-06, "epoch": 0.34975957857339746, "percentage": 83.97, "elapsed_time": "0:23:26", "remaining_time": "0:04:28", "throughput": 13378.49, "total_tokens": 18817608}
{"current_steps": 2860, "total_steps": 3400, "loss": 0.3968, "lr": 6.7393129075627335e-06, "epoch": 0.35037211723990075, "percentage": 84.12, "elapsed_time": "0:24:26", "remaining_time": "0:04:36", "throughput": 12855.09, "total_tokens": 18851504}
{"current_steps": 2865, "total_steps": 3400, "loss": 0.4142, "lr": 6.6179050846274515e-06, "epoch": 0.3509846559064041, "percentage": 84.26, "elapsed_time": "0:25:25", "remaining_time": "0:04:44", "throughput": 12377.01, "total_tokens": 18884968}
{"current_steps": 2870, "total_steps": 3400, "loss": 0.3784, "lr": 6.497523254635296e-06, "epoch": 0.3515971945729074, "percentage": 84.41, "elapsed_time": "0:26:26", "remaining_time": "0:04:52", "throughput": 11925.73, "total_tokens": 18917928}
{"current_steps": 2875, "total_steps": 3400, "loss": 0.339, "lr": 6.37817026463432e-06, "epoch": 0.35220973323941074, "percentage": 84.56, "elapsed_time": "0:27:25", "remaining_time": "0:05:00", "throughput": 11515.82, "total_tokens": 18952048}
{"current_steps": 2880, "total_steps": 3400, "loss": 0.4563, "lr": 6.25984893734034e-06, "epoch": 0.3528222719059141, "percentage": 84.71, "elapsed_time": "0:28:24", "remaining_time": "0:05:07", "throughput": 11137.44, "total_tokens": 18984864}
{"current_steps": 2885, "total_steps": 3400, "loss": 0.3283, "lr": 6.142562071070179e-06, "epoch": 0.3534348105724174, "percentage": 84.85, "elapsed_time": "0:29:24", "remaining_time": "0:05:14", "throughput": 10778.78, "total_tokens": 19018664}
{"current_steps": 2890, "total_steps": 3400, "loss": 0.422, "lr": 6.026312439675552e-06, "epoch": 0.3540473492389207, "percentage": 85.0, "elapsed_time": "0:30:23", "remaining_time": "0:05:21", "throughput": 10446.93, "total_tokens": 19052696}
{"current_steps": 2895, "total_steps": 3400, "loss": 0.3772, "lr": 5.911102792477357e-06, "epoch": 0.354659887905424, "percentage": 85.15, "elapsed_time": "0:31:22", "remaining_time": "0:05:28", "throughput": 10138.03, "total_tokens": 19086368}
{"current_steps": 2900, "total_steps": 3400, "loss": 0.4081, "lr": 5.796935854200763e-06, "epoch": 0.35527242657192737, "percentage": 85.29, "elapsed_time": "0:32:22", "remaining_time": "0:05:34", "throughput": 9842.45, "total_tokens": 19120400}
{"current_steps": 2900, "total_steps": 3400, "eval_loss": 0.16432562470436096, "epoch": 0.35527242657192737, "percentage": 85.29, "elapsed_time": "0:32:42", "remaining_time": "0:05:38", "throughput": 9745.12, "total_tokens": 19120400}
{"current_steps": 2905, "total_steps": 3400, "loss": 0.3497, "lr": 5.683814324910685e-06, "epoch": 0.35588496523843066, "percentage": 85.44, "elapsed_time": "0:33:46", "remaining_time": "0:05:45", "throughput": 9453.58, "total_tokens": 19154080}
{"current_steps": 2910, "total_steps": 3400, "loss": 0.3912, "lr": 5.571740879947979e-06, "epoch": 0.356497503904934, "percentage": 85.59, "elapsed_time": "0:34:44", "remaining_time": "0:05:51", "throughput": 9203.0, "total_tokens": 19187704}
{"current_steps": 2915, "total_steps": 3400, "loss": 0.3573, "lr": 5.4607181698661634e-06, "epoch": 0.3571100425714373, "percentage": 85.74, "elapsed_time": "0:35:45", "remaining_time": "0:05:56", "throughput": 8959.45, "total_tokens": 19221704}
{"current_steps": 2920, "total_steps": 3400, "loss": 0.3441, "lr": 5.35074882036869e-06, "epoch": 0.35772258123794065, "percentage": 85.88, "elapsed_time": "0:36:44", "remaining_time": "0:06:02", "throughput": 8735.33, "total_tokens": 19255184}
{"current_steps": 2925, "total_steps": 3400, "loss": 0.3662, "lr": 5.241835432246889e-06, "epoch": 0.35833511990444394, "percentage": 86.03, "elapsed_time": "0:37:42", "remaining_time": "0:06:07", "throughput": 8523.57, "total_tokens": 19287920}
{"current_steps": 2930, "total_steps": 3400, "loss": 0.3524, "lr": 5.133980581318459e-06, "epoch": 0.3589476585709473, "percentage": 86.18, "elapsed_time": "0:38:42", "remaining_time": "0:06:12", "throughput": 8317.46, "total_tokens": 19321384}
{"current_steps": 2935, "total_steps": 3400, "loss": 0.3397, "lr": 5.027186818366542e-06, "epoch": 0.35956019723745064, "percentage": 86.32, "elapsed_time": "0:39:42", "remaining_time": "0:06:17", "throughput": 8125.1, "total_tokens": 19355128}
{"current_steps": 2940, "total_steps": 3400, "loss": 0.3737, "lr": 4.921456669079366e-06, "epoch": 0.36017273590395393, "percentage": 86.47, "elapsed_time": "0:40:41", "remaining_time": "0:06:21", "throughput": 7942.25, "total_tokens": 19388880}
{"current_steps": 2945, "total_steps": 3400, "loss": 0.3888, "lr": 4.816792633990569e-06, "epoch": 0.3607852745704573, "percentage": 86.62, "elapsed_time": "0:41:41", "remaining_time": "0:06:26", "throughput": 7765.21, "total_tokens": 19422408}
{"current_steps": 2950, "total_steps": 3400, "loss": 0.3425, "lr": 4.713197188420026e-06, "epoch": 0.3613978132369606, "percentage": 86.76, "elapsed_time": "0:42:40", "remaining_time": "0:06:30", "throughput": 7597.25, "total_tokens": 19455392}
{"current_steps": 2950, "total_steps": 3400, "eval_loss": 0.14024704694747925, "epoch": 0.3613978132369606, "percentage": 86.76, "elapsed_time": "0:43:00", "remaining_time": "0:06:33", "throughput": 7538.77, "total_tokens": 19455392}
{"current_steps": 2955, "total_steps": 3400, "loss": 0.367, "lr": 4.610672782415276e-06, "epoch": 0.3620103519034639, "percentage": 86.91, "elapsed_time": "0:44:05", "remaining_time": "0:06:38", "throughput": 7366.61, "total_tokens": 19489208}
{"current_steps": 2960, "total_steps": 3400, "loss": 0.4011, "lr": 4.509221840693656e-06, "epoch": 0.3626228905699672, "percentage": 87.06, "elapsed_time": "0:45:05", "remaining_time": "0:06:42", "throughput": 7215.97, "total_tokens": 19522184}
{"current_steps": 2965, "total_steps": 3400, "loss": 0.3989, "lr": 4.408846762584901e-06, "epoch": 0.36323542923647056, "percentage": 87.21, "elapsed_time": "0:46:05", "remaining_time": "0:06:45", "throughput": 7071.88, "total_tokens": 19556560}
{"current_steps": 2970, "total_steps": 3400, "loss": 0.328, "lr": 4.309549921974421e-06, "epoch": 0.36384796790297386, "percentage": 87.35, "elapsed_time": "0:47:04", "remaining_time": "0:06:48", "throughput": 6935.71, "total_tokens": 19590856}
{"current_steps": 2975, "total_steps": 3400, "loss": 0.4069, "lr": 4.2113336672471245e-06, "epoch": 0.3644605065694772, "percentage": 87.5, "elapsed_time": "0:48:04", "remaining_time": "0:06:52", "throughput": 6804.16, "total_tokens": 19623992}
{"current_steps": 2980, "total_steps": 3400, "loss": 0.4154, "lr": 4.114200321231937e-06, "epoch": 0.3650730452359805, "percentage": 87.65, "elapsed_time": "0:49:04", "remaining_time": "0:06:54", "throughput": 6676.53, "total_tokens": 19657384}
{"current_steps": 2985, "total_steps": 3400, "loss": 0.3801, "lr": 4.018152181146823e-06, "epoch": 0.36568558390248385, "percentage": 87.79, "elapsed_time": "0:50:03", "remaining_time": "0:06:57", "throughput": 6555.23, "total_tokens": 19690344}
{"current_steps": 2990, "total_steps": 3400, "loss": 0.4359, "lr": 3.923191518544434e-06, "epoch": 0.3662981225689872, "percentage": 87.94, "elapsed_time": "0:51:02", "remaining_time": "0:06:59", "throughput": 6440.2, "total_tokens": 19723432}
{"current_steps": 2995, "total_steps": 3400, "loss": 0.366, "lr": 3.829320579258466e-06, "epoch": 0.3669106612354905, "percentage": 88.09, "elapsed_time": "0:52:01", "remaining_time": "0:07:02", "throughput": 6328.74, "total_tokens": 19756768}
{"current_steps": 3000, "total_steps": 3400, "loss": 0.3874, "lr": 3.7365415833504725e-06, "epoch": 0.36752319990199384, "percentage": 88.24, "elapsed_time": "0:53:00", "remaining_time": "0:07:04", "throughput": 6223.01, "total_tokens": 19790096}
{"current_steps": 3000, "total_steps": 3400, "eval_loss": 0.16951163113117218, "epoch": 0.36752319990199384, "percentage": 88.24, "elapsed_time": "0:53:19", "remaining_time": "0:07:06", "throughput": 6185.48, "total_tokens": 19790096}
{"current_steps": 3005, "total_steps": 16324, "loss": 0.374, "lr": 9.516767703953432e-05, "epoch": 0.36813573856849713, "percentage": 18.41, "elapsed_time": "0:02:14", "remaining_time": "0:09:58", "throughput": 146864.94, "total_tokens": 19823560}
{"current_steps": 3010, "total_steps": 16324, "loss": 0.4155, "lr": 9.514593112044912e-05, "epoch": 0.3687482772350005, "percentage": 18.44, "elapsed_time": "0:03:13", "remaining_time": "0:14:17", "throughput": 102404.53, "total_tokens": 19857216}
{"current_steps": 3015, "total_steps": 16324, "loss": 0.4984, "lr": 9.512413887771539e-05, "epoch": 0.36936081590150377, "percentage": 18.47, "elapsed_time": "0:04:12", "remaining_time": "0:18:34", "throughput": 78795.37, "total_tokens": 19891488}
{"current_steps": 3020, "total_steps": 16324, "loss": 0.4179, "lr": 9.510230033369388e-05, "epoch": 0.3699733545680071, "percentage": 18.5, "elapsed_time": "0:05:10", "remaining_time": "0:22:49", "throughput": 64110.62, "total_tokens": 19924968}
{"current_steps": 3025, "total_steps": 16324, "loss": 0.4198, "lr": 9.508041551079284e-05, "epoch": 0.3705858932345104, "percentage": 18.53, "elapsed_time": "0:06:10", "remaining_time": "0:27:09", "throughput": 53851.58, "total_tokens": 19958360}
{"current_steps": 3030, "total_steps": 16324, "loss": 0.3959, "lr": 9.505848443146796e-05, "epoch": 0.37119843190101376, "percentage": 18.56, "elapsed_time": "0:07:09", "remaining_time": "0:31:25", "throughput": 46528.24, "total_tokens": 19991912}
{"current_steps": 3035, "total_steps": 16324, "loss": 0.4709, "lr": 9.503650711822247e-05, "epoch": 0.37181097056751705, "percentage": 18.59, "elapsed_time": "0:08:08", "remaining_time": "0:35:39", "throughput": 40976.42, "total_tokens": 20025264}
{"current_steps": 3040, "total_steps": 16324, "loss": 0.3901, "lr": 9.501448359360698e-05, "epoch": 0.3724235092340204, "percentage": 18.62, "elapsed_time": "0:09:09", "remaining_time": "0:39:59", "throughput": 36530.85, "total_tokens": 20059392}
{"current_steps": 3045, "total_steps": 16324, "loss": 0.4342, "lr": 9.499241388021955e-05, "epoch": 0.3730360479005237, "percentage": 18.65, "elapsed_time": "0:10:09", "remaining_time": "0:44:17", "throughput": 32968.14, "total_tokens": 20092560}
{"current_steps": 3050, "total_steps": 16324, "loss": 0.3706, "lr": 9.497029800070565e-05, "epoch": 0.37364858656702704, "percentage": 18.68, "elapsed_time": "0:11:10", "remaining_time": "0:48:39", "throughput": 30001.47, "total_tokens": 20126152}
{"current_steps": 3050, "total_steps": 16324, "eval_loss": 0.5081549882888794, "epoch": 0.37364858656702704, "percentage": 18.68, "elapsed_time": "0:11:59", "remaining_time": "0:52:12", "throughput": 27966.44, "total_tokens": 20126152}
{"current_steps": 3055, "total_steps": 16324, "loss": 0.5132, "lr": 9.494813597775805e-05, "epoch": 0.3742611252335304, "percentage": 18.71, "elapsed_time": "0:13:10", "remaining_time": "0:57:11", "throughput": 25515.56, "total_tokens": 20159744}
{"current_steps": 3060, "total_steps": 16324, "loss": 0.4203, "lr": 9.492592783411694e-05, "epoch": 0.3748736639000337, "percentage": 18.75, "elapsed_time": "0:14:10", "remaining_time": "1:01:27", "throughput": 23737.94, "total_tokens": 20192704}
{"current_steps": 3065, "total_steps": 16324, "loss": 0.5163, "lr": 9.490367359256979e-05, "epoch": 0.37548620256653703, "percentage": 18.78, "elapsed_time": "0:15:11", "remaining_time": "1:05:40", "throughput": 22202.18, "total_tokens": 20226224}
{"current_steps": 3070, "total_steps": 16324, "loss": 0.5524, "lr": 9.48813732759514e-05, "epoch": 0.3760987412330403, "percentage": 18.81, "elapsed_time": "0:16:11", "remaining_time": "1:09:52", "throughput": 20863.44, "total_tokens": 20260560}
{"current_steps": 3075, "total_steps": 16324, "loss": 0.4073, "lr": 9.485902690714381e-05, "epoch": 0.37671127989954367, "percentage": 18.84, "elapsed_time": "0:17:11", "remaining_time": "1:14:06", "throughput": 19666.3, "total_tokens": 20293408}
{"current_steps": 3080, "total_steps": 16324, "loss": 0.4307, "lr": 9.483663450907635e-05, "epoch": 0.37732381856604696, "percentage": 18.87, "elapsed_time": "0:18:11", "remaining_time": "1:18:14", "throughput": 18621.12, "total_tokens": 20327600}
{"current_steps": 3085, "total_steps": 16324, "loss": 0.5101, "lr": 9.481419610472559e-05, "epoch": 0.3779363572325503, "percentage": 18.9, "elapsed_time": "0:19:11", "remaining_time": "1:22:19", "throughput": 17689.63, "total_tokens": 20361000}
{"current_steps": 3090, "total_steps": 16324, "loss": 0.4317, "lr": 9.479171171711525e-05, "epoch": 0.3785488958990536, "percentage": 18.93, "elapsed_time": "0:20:11", "remaining_time": "1:26:27", "throughput": 16837.43, "total_tokens": 20393928}
{"current_steps": 3095, "total_steps": 16324, "loss": 0.4347, "lr": 9.476918136931627e-05, "epoch": 0.37916143456555695, "percentage": 18.96, "elapsed_time": "0:21:10", "remaining_time": "1:30:29", "throughput": 16081.29, "total_tokens": 20427032}
{"current_steps": 3100, "total_steps": 16324, "loss": 0.372, "lr": 9.474660508444674e-05, "epoch": 0.37977397323206025, "percentage": 18.99, "elapsed_time": "0:22:10", "remaining_time": "1:34:35", "throughput": 15377.35, "total_tokens": 20460696}
{"current_steps": 3100, "total_steps": 16324, "eval_loss": 0.27452927827835083, "epoch": 0.37977397323206025, "percentage": 18.99, "elapsed_time": "0:22:30", "remaining_time": "1:35:59", "throughput": 15153.26, "total_tokens": 20460696}
{"current_steps": 3105, "total_steps": 16324, "loss": 0.4703, "lr": 9.472398288567192e-05, "epoch": 0.3803865118985636, "percentage": 19.02, "elapsed_time": "0:23:37", "remaining_time": "1:40:32", "throughput": 14462.68, "total_tokens": 20494616}
{"current_steps": 3110, "total_steps": 16324, "loss": 0.3934, "lr": 9.470131479620412e-05, "epoch": 0.38099905056506694, "percentage": 19.05, "elapsed_time": "0:24:37", "remaining_time": "1:44:37", "throughput": 13895.6, "total_tokens": 20528808}
{"current_steps": 3115, "total_steps": 16324, "loss": 0.4117, "lr": 9.46786008393028e-05, "epoch": 0.38161158923157024, "percentage": 19.08, "elapsed_time": "0:25:37", "remaining_time": "1:48:41", "throughput": 13371.1, "total_tokens": 20562384}
{"current_steps": 3120, "total_steps": 16324, "loss": 0.3918, "lr": 9.465584103827442e-05, "epoch": 0.3822241278980736, "percentage": 19.11, "elapsed_time": "0:26:39", "remaining_time": "1:52:47", "throughput": 12879.76, "total_tokens": 20595448}
{"current_steps": 3125, "total_steps": 16324, "loss": 0.4222, "lr": 9.463303541647253e-05, "epoch": 0.3828366665645769, "percentage": 19.14, "elapsed_time": "0:27:38", "remaining_time": "1:56:45", "throughput": 12437.54, "total_tokens": 20628992}
{"current_steps": 3130, "total_steps": 16324, "loss": 0.4089, "lr": 9.461018399729768e-05, "epoch": 0.3834492052310802, "percentage": 19.17, "elapsed_time": "0:28:38", "remaining_time": "2:00:44", "throughput": 12022.42, "total_tokens": 20662448}
{"current_steps": 3135, "total_steps": 16324, "loss": 0.4812, "lr": 9.458728680419741e-05, "epoch": 0.3840617438975835, "percentage": 19.2, "elapsed_time": "0:29:40", "remaining_time": "2:04:49", "throughput": 11625.12, "total_tokens": 20696496}
{"current_steps": 3140, "total_steps": 16324, "loss": 0.4318, "lr": 9.456434386066624e-05, "epoch": 0.38467428256408687, "percentage": 19.24, "elapsed_time": "0:30:39", "remaining_time": "2:08:45", "throughput": 11266.38, "total_tokens": 20729976}
{"current_steps": 3145, "total_steps": 16324, "loss": 0.3784, "lr": 9.454135519024561e-05, "epoch": 0.38528682123059016, "percentage": 19.27, "elapsed_time": "0:31:39", "remaining_time": "2:12:39", "throughput": 10931.31, "total_tokens": 20764304}
{"current_steps": 3150, "total_steps": 16324, "loss": 0.3967, "lr": 9.451832081652389e-05, "epoch": 0.3858993598970935, "percentage": 19.3, "elapsed_time": "0:32:39", "remaining_time": "2:16:37", "throughput": 10611.38, "total_tokens": 20798040}
{"current_steps": 3150, "total_steps": 16324, "eval_loss": 0.45261481404304504, "epoch": 0.3858993598970935, "percentage": 19.3, "elapsed_time": "0:32:59", "remaining_time": "2:17:59", "throughput": 10506.21, "total_tokens": 20798040}
{"current_steps": 3155, "total_steps": 16324, "loss": 0.4064, "lr": 9.449524076313637e-05, "epoch": 0.3865118985635968, "percentage": 19.33, "elapsed_time": "0:34:03", "remaining_time": "2:22:11", "throughput": 10192.21, "total_tokens": 20832088}
{"current_steps": 3160, "total_steps": 16324, "loss": 0.4767, "lr": 9.447211505376518e-05, "epoch": 0.38712443723010015, "percentage": 19.36, "elapsed_time": "0:35:03", "remaining_time": "2:26:02", "throughput": 9919.63, "total_tokens": 20865448}
{"current_steps": 3165, "total_steps": 16324, "loss": 0.4363, "lr": 9.444894371213929e-05, "epoch": 0.3877369758966035, "percentage": 19.39, "elapsed_time": "0:36:04", "remaining_time": "2:29:58", "throughput": 9655.67, "total_tokens": 20898416}
{"current_steps": 3170, "total_steps": 16324, "loss": 0.3689, "lr": 9.442572676203454e-05, "epoch": 0.3883495145631068, "percentage": 19.42, "elapsed_time": "0:37:03", "remaining_time": "2:33:47", "throughput": 9413.07, "total_tokens": 20931984}
{"current_steps": 3175, "total_steps": 16324, "loss": 0.4678, "lr": 9.44024642272735e-05, "epoch": 0.38896205322961014, "percentage": 19.45, "elapsed_time": "0:38:03", "remaining_time": "2:37:35", "throughput": 9182.39, "total_tokens": 20965456}
{"current_steps": 3180, "total_steps": 16324, "loss": 0.4174, "lr": 9.437915613172562e-05, "epoch": 0.38957459189611343, "percentage": 19.48, "elapsed_time": "0:39:04", "remaining_time": "2:41:30", "throughput": 8957.19, "total_tokens": 20998864}
{"current_steps": 3185, "total_steps": 16324, "loss": 0.4286, "lr": 9.435580249930696e-05, "epoch": 0.3901871305626168, "percentage": 19.51, "elapsed_time": "0:40:04", "remaining_time": "2:45:17", "throughput": 8748.66, "total_tokens": 21032200}
{"current_steps": 3190, "total_steps": 16324, "loss": 0.4015, "lr": 9.433240335398039e-05, "epoch": 0.3907996692291201, "percentage": 19.54, "elapsed_time": "0:41:04", "remaining_time": "2:49:05", "throughput": 8549.0, "total_tokens": 21066128}
{"current_steps": 3195, "total_steps": 16324, "loss": 0.4048, "lr": 9.430895871975549e-05, "epoch": 0.3914122078956234, "percentage": 19.57, "elapsed_time": "0:42:04", "remaining_time": "2:52:52", "throughput": 8359.36, "total_tokens": 21100016}
{"current_steps": 3200, "total_steps": 16324, "loss": 0.4331, "lr": 9.42854686206885e-05, "epoch": 0.3920247465621267, "percentage": 19.6, "elapsed_time": "0:43:04", "remaining_time": "2:56:38", "throughput": 8178.17, "total_tokens": 21133184}
{"current_steps": 3200, "total_steps": 16324, "eval_loss": 0.2645382583141327, "epoch": 0.3920247465621267, "percentage": 19.6, "elapsed_time": "0:43:23", "remaining_time": "2:57:58", "throughput": 8116.62, "total_tokens": 21133184}
{"current_steps": 3205, "total_steps": 16324, "loss": 0.3808, "lr": 9.426193308088226e-05, "epoch": 0.39263728522863006, "percentage": 19.63, "elapsed_time": "0:44:27", "remaining_time": "3:02:00", "throughput": 7933.89, "total_tokens": 21165984}
{"current_steps": 3210, "total_steps": 16324, "loss": 0.3748, "lr": 9.423835212448632e-05, "epoch": 0.39324982389513335, "percentage": 19.66, "elapsed_time": "0:45:27", "remaining_time": "3:05:41", "throughput": 7774.04, "total_tokens": 21200392}
{"current_steps": 3215, "total_steps": 16324, "loss": 0.4623, "lr": 9.421472577569677e-05, "epoch": 0.3938623625616367, "percentage": 19.69, "elapsed_time": "0:46:26", "remaining_time": "3:09:22", "throughput": 7619.93, "total_tokens": 21234232}
{"current_steps": 3220, "total_steps": 16324, "loss": 0.4058, "lr": 9.419105405875631e-05, "epoch": 0.39447490122814005, "percentage": 19.73, "elapsed_time": "0:47:25", "remaining_time": "3:13:01", "throughput": 7472.9, "total_tokens": 21267600}
{"current_steps": 3225, "total_steps": 16324, "loss": 0.4227, "lr": 9.416733699795417e-05, "epoch": 0.39508743989464334, "percentage": 19.76, "elapsed_time": "0:48:26", "remaining_time": "3:16:45", "throughput": 7328.71, "total_tokens": 21300688}
{"current_steps": 3230, "total_steps": 16324, "loss": 0.4075, "lr": 9.41435746176261e-05, "epoch": 0.3956999785611467, "percentage": 19.79, "elapsed_time": "0:49:26", "remaining_time": "3:20:26", "throughput": 7191.53, "total_tokens": 21334832}
{"current_steps": 3235, "total_steps": 16324, "loss": 0.4461, "lr": 9.41197669421544e-05, "epoch": 0.39631251722765, "percentage": 19.82, "elapsed_time": "0:50:25", "remaining_time": "3:24:01", "throughput": 7062.38, "total_tokens": 21368160}
{"current_steps": 3240, "total_steps": 16324, "loss": 0.3888, "lr": 9.409591399596779e-05, "epoch": 0.39692505589415333, "percentage": 19.85, "elapsed_time": "0:51:24", "remaining_time": "3:27:37", "throughput": 6937.72, "total_tokens": 21401088}
{"current_steps": 3245, "total_steps": 16324, "loss": 0.4325, "lr": 9.407201580354148e-05, "epoch": 0.3975375945606566, "percentage": 19.88, "elapsed_time": "0:52:25", "remaining_time": "3:31:18", "throughput": 6814.65, "total_tokens": 21435856}
{"current_steps": 3250, "total_steps": 16324, "loss": 0.3708, "lr": 9.404807238939707e-05, "epoch": 0.39815013322716, "percentage": 19.91, "elapsed_time": "0:53:25", "remaining_time": "3:34:53", "throughput": 6698.53, "total_tokens": 21468896}
{"current_steps": 3250, "total_steps": 16324, "eval_loss": 0.3505937457084656, "epoch": 0.39815013322716, "percentage": 19.91, "elapsed_time": "0:53:44", "remaining_time": "3:36:11", "throughput": 6657.98, "total_tokens": 21468896}
{"current_steps": 3255, "total_steps": 16324, "loss": 0.4381, "lr": 9.402408377810261e-05, "epoch": 0.39876267189366327, "percentage": 19.94, "elapsed_time": "0:54:49", "remaining_time": "3:40:06", "throughput": 6537.26, "total_tokens": 21502040}
{"current_steps": 3260, "total_steps": 16324, "loss": 0.3854, "lr": 9.400004999427248e-05, "epoch": 0.3993752105601666, "percentage": 19.97, "elapsed_time": "0:55:49", "remaining_time": "3:43:43", "throughput": 6429.47, "total_tokens": 21536112}
{"current_steps": 3265, "total_steps": 16324, "loss": 0.4841, "lr": 9.397597106256744e-05, "epoch": 0.3999877492266699, "percentage": 20.0, "elapsed_time": "0:56:48", "remaining_time": "3:47:14", "throughput": 6327.72, "total_tokens": 21570152}
{"current_steps": 3270, "total_steps": 16324, "loss": 0.3739, "lr": 9.395184700769457e-05, "epoch": 0.40060028789317326, "percentage": 20.03, "elapsed_time": "0:57:48", "remaining_time": "3:50:46", "throughput": 6228.73, "total_tokens": 21603832}
{"current_steps": 3275, "total_steps": 16324, "loss": 0.4677, "lr": 9.392767785440723e-05, "epoch": 0.4012128265596766, "percentage": 20.06, "elapsed_time": "0:58:49", "remaining_time": "3:54:21", "throughput": 6131.15, "total_tokens": 21637048}
{"current_steps": 3280, "total_steps": 16324, "loss": 0.4087, "lr": 9.39034636275051e-05, "epoch": 0.4018253652261799, "percentage": 20.09, "elapsed_time": "0:59:48", "remaining_time": "3:57:52", "throughput": 6038.23, "total_tokens": 21670896}
{"current_steps": 3285, "total_steps": 16324, "loss": 0.4207, "lr": 9.387920435183407e-05, "epoch": 0.40243790389268325, "percentage": 20.12, "elapsed_time": "1:00:48", "remaining_time": "4:01:21", "throughput": 5949.02, "total_tokens": 21704680}
{"current_steps": 3290, "total_steps": 16324, "loss": 0.4417, "lr": 9.385490005228624e-05, "epoch": 0.40305044255918654, "percentage": 20.15, "elapsed_time": "1:01:49", "remaining_time": "4:04:55", "throughput": 5860.28, "total_tokens": 21738520}
{"current_steps": 3295, "total_steps": 16324, "loss": 0.4105, "lr": 9.383055075379996e-05, "epoch": 0.4036629812256899, "percentage": 20.19, "elapsed_time": "1:02:49", "remaining_time": "4:08:25", "throughput": 5775.66, "total_tokens": 21772376}
{"current_steps": 3300, "total_steps": 16324, "loss": 0.3826, "lr": 9.380615648135973e-05, "epoch": 0.4042755198921932, "percentage": 20.22, "elapsed_time": "1:03:49", "remaining_time": "4:11:53", "throughput": 5694.06, "total_tokens": 21805544}
{"current_steps": 3300, "total_steps": 16324, "eval_loss": 0.42393144965171814, "epoch": 0.4042755198921932, "percentage": 20.22, "elapsed_time": "1:04:08", "remaining_time": "4:13:10", "throughput": 5665.28, "total_tokens": 21805544}
{"current_steps": 3305, "total_steps": 16324, "loss": 0.3376, "lr": 9.378171725999618e-05, "epoch": 0.40488805855869653, "percentage": 20.25, "elapsed_time": "1:05:15", "remaining_time": "4:17:03", "throughput": 5577.72, "total_tokens": 21839512}
{"current_steps": 3310, "total_steps": 16324, "loss": 0.5174, "lr": 9.375723311478606e-05, "epoch": 0.4055005972251998, "percentage": 20.28, "elapsed_time": "1:06:14", "remaining_time": "4:20:25", "throughput": 5504.0, "total_tokens": 21873384}
{"current_steps": 3315, "total_steps": 16324, "loss": 0.4144, "lr": 9.373270407085227e-05, "epoch": 0.40611313589170317, "percentage": 20.31, "elapsed_time": "1:07:12", "remaining_time": "4:23:45", "throughput": 5432.25, "total_tokens": 21906552}
{"current_steps": 3320, "total_steps": 16324, "loss": 0.4661, "lr": 9.37081301533637e-05, "epoch": 0.40672567455820646, "percentage": 20.34, "elapsed_time": "1:08:12", "remaining_time": "4:27:08", "throughput": 5361.32, "total_tokens": 21939496}
{"current_steps": 3325, "total_steps": 16324, "loss": 0.3998, "lr": 9.368351138753535e-05, "epoch": 0.4073382132247098, "percentage": 20.37, "elapsed_time": "1:09:11", "remaining_time": "4:30:30", "throughput": 5292.7, "total_tokens": 21972984}
{"current_steps": 3330, "total_steps": 16324, "loss": 0.4144, "lr": 9.36588477986282e-05, "epoch": 0.40795075189121316, "percentage": 20.4, "elapsed_time": "1:10:10", "remaining_time": "4:33:49", "throughput": 5226.67, "total_tokens": 22006464}
{"current_steps": 3335, "total_steps": 16324, "loss": 0.3585, "lr": 9.363413941194926e-05, "epoch": 0.40856329055771645, "percentage": 20.43, "elapsed_time": "1:11:09", "remaining_time": "4:37:10", "throughput": 5161.83, "total_tokens": 22040312}
{"current_steps": 3340, "total_steps": 16324, "loss": 0.5134, "lr": 9.360938625285144e-05, "epoch": 0.4091758292242198, "percentage": 20.46, "elapsed_time": "1:12:09", "remaining_time": "4:40:30", "throughput": 5098.45, "total_tokens": 22073584}
{"current_steps": 3345, "total_steps": 16324, "loss": 0.4455, "lr": 9.35845883467337e-05, "epoch": 0.4097883678907231, "percentage": 20.49, "elapsed_time": "1:13:08", "remaining_time": "4:43:46", "throughput": 5037.88, "total_tokens": 22106912}
{"current_steps": 3350, "total_steps": 16324, "loss": 0.4719, "lr": 9.355974571904076e-05, "epoch": 0.41040090655722644, "percentage": 20.52, "elapsed_time": "1:14:07", "remaining_time": "4:47:02", "throughput": 4978.54, "total_tokens": 22140064}
{"current_steps": 3350, "total_steps": 16324, "eval_loss": 0.3364885747432709, "epoch": 0.41040090655722644, "percentage": 20.52, "elapsed_time": "1:14:26", "remaining_time": "4:48:18", "throughput": 4956.74, "total_tokens": 22140064}
{"current_steps": 3355, "total_steps": 16324, "loss": 0.4072, "lr": 9.353485839526337e-05, "epoch": 0.41101344522372973, "percentage": 20.55, "elapsed_time": "1:15:32", "remaining_time": "4:52:01", "throughput": 4891.9, "total_tokens": 22173000}
{"current_steps": 3360, "total_steps": 16324, "loss": 0.3736, "lr": 9.350992640093807e-05, "epoch": 0.4116259838902331, "percentage": 20.58, "elapsed_time": "1:16:31", "remaining_time": "4:55:17", "throughput": 4836.03, "total_tokens": 22206640}
{"current_steps": 3365, "total_steps": 16324, "loss": 0.4371, "lr": 9.348494976164727e-05, "epoch": 0.4122385225567364, "percentage": 20.61, "elapsed_time": "1:17:31", "remaining_time": "4:58:31", "throughput": 4781.68, "total_tokens": 22239976}
{"current_steps": 3370, "total_steps": 16324, "loss": 0.3755, "lr": 9.345992850301917e-05, "epoch": 0.4128510612232397, "percentage": 20.64, "elapsed_time": "1:18:31", "remaining_time": "5:01:51", "throughput": 4727.35, "total_tokens": 22273640}
{"current_steps": 3375, "total_steps": 16324, "loss": 0.4523, "lr": 9.343486265072772e-05, "epoch": 0.413463599889743, "percentage": 20.68, "elapsed_time": "1:19:30", "remaining_time": "5:05:04", "throughput": 4675.61, "total_tokens": 22306584}
{"current_steps": 3380, "total_steps": 16324, "loss": 0.4513, "lr": 9.340975223049271e-05, "epoch": 0.41407613855624636, "percentage": 20.71, "elapsed_time": "1:20:29", "remaining_time": "5:08:16", "throughput": 4625.51, "total_tokens": 22340328}
{"current_steps": 3385, "total_steps": 16324, "loss": 0.4645, "lr": 9.338459726807957e-05, "epoch": 0.4146886772227497, "percentage": 20.74, "elapsed_time": "1:21:30", "remaining_time": "5:11:34", "throughput": 4575.0, "total_tokens": 22375056}
{"current_steps": 3390, "total_steps": 16324, "loss": 0.4131, "lr": 9.335939778929952e-05, "epoch": 0.415301215889253, "percentage": 20.77, "elapsed_time": "1:22:30", "remaining_time": "5:14:46", "throughput": 4527.16, "total_tokens": 22409792}
{"current_steps": 3395, "total_steps": 16324, "loss": 0.4678, "lr": 9.333415382000938e-05, "epoch": 0.41591375455575635, "percentage": 20.8, "elapsed_time": "1:23:29", "remaining_time": "5:17:56", "throughput": 4480.29, "total_tokens": 22443432}
{"current_steps": 3400, "total_steps": 16324, "loss": 0.452, "lr": 9.330886538611168e-05, "epoch": 0.41652629322225965, "percentage": 20.83, "elapsed_time": "1:24:30", "remaining_time": "5:21:12", "throughput": 4433.2, "total_tokens": 22477472}
{"current_steps": 3400, "total_steps": 16324, "eval_loss": 0.3813478350639343, "epoch": 0.41652629322225965, "percentage": 20.83, "elapsed_time": "1:24:49", "remaining_time": "5:22:27", "throughput": 4416.21, "total_tokens": 22477472}
{"current_steps": 3405, "total_steps": 16324, "loss": 0.473, "lr": 9.328353251355453e-05, "epoch": 0.417138831888763, "percentage": 20.86, "elapsed_time": "1:25:54", "remaining_time": "5:25:56", "throughput": 4367.39, "total_tokens": 22511088}
{"current_steps": 3410, "total_steps": 16324, "loss": 0.4002, "lr": 9.325815522833167e-05, "epoch": 0.4177513705552663, "percentage": 20.89, "elapsed_time": "1:26:54", "remaining_time": "5:29:08", "throughput": 4323.55, "total_tokens": 22545728}
{"current_steps": 3415, "total_steps": 16324, "loss": 0.4302, "lr": 9.323273355648241e-05, "epoch": 0.41836390922176964, "percentage": 20.92, "elapsed_time": "1:27:55", "remaining_time": "5:32:22", "throughput": 4279.89, "total_tokens": 22578928}
{"current_steps": 3420, "total_steps": 16324, "loss": 0.3674, "lr": 9.320726752409157e-05, "epoch": 0.41897644788827293, "percentage": 20.95, "elapsed_time": "1:28:55", "remaining_time": "5:35:30", "throughput": 4238.41, "total_tokens": 22612896}
{"current_steps": 3425, "total_steps": 16324, "loss": 0.3903, "lr": 9.318175715728954e-05, "epoch": 0.4195889865547763, "percentage": 20.98, "elapsed_time": "1:29:54", "remaining_time": "5:38:34", "throughput": 4198.24, "total_tokens": 22645784}
{"current_steps": 3430, "total_steps": 16324, "loss": 0.4181, "lr": 9.315620248225217e-05, "epoch": 0.42020152522127957, "percentage": 21.01, "elapsed_time": "1:30:54", "remaining_time": "5:41:45", "throughput": 4157.55, "total_tokens": 22678528}
{"current_steps": 3435, "total_steps": 16324, "loss": 0.4324, "lr": 9.313060352520073e-05, "epoch": 0.4208140638877829, "percentage": 21.04, "elapsed_time": "1:31:53", "remaining_time": "5:44:48", "throughput": 4119.28, "total_tokens": 22712640}
{"current_steps": 3440, "total_steps": 16324, "loss": 0.3523, "lr": 9.310496031240205e-05, "epoch": 0.42142660255428627, "percentage": 21.07, "elapsed_time": "1:32:52", "remaining_time": "5:47:50", "throughput": 4082.08, "total_tokens": 22746800}
{"current_steps": 3445, "total_steps": 16324, "loss": 0.3652, "lr": 9.307927287016823e-05, "epoch": 0.42203914122078956, "percentage": 21.1, "elapsed_time": "1:33:51", "remaining_time": "5:50:53", "throughput": 4045.03, "total_tokens": 22780448}
{"current_steps": 3450, "total_steps": 16324, "loss": 0.4566, "lr": 9.305354122485686e-05, "epoch": 0.4226516798872929, "percentage": 21.13, "elapsed_time": "1:34:51", "remaining_time": "5:53:59", "throughput": 4008.12, "total_tokens": 22813528}
{"current_steps": 3450, "total_steps": 16324, "eval_loss": 0.4307055175304413, "epoch": 0.4226516798872929, "percentage": 21.13, "elapsed_time": "1:35:10", "remaining_time": "5:55:10", "throughput": 3994.73, "total_tokens": 22813528}
{"current_steps": 3455, "total_steps": 16324, "loss": 0.414, "lr": 9.302776540287082e-05, "epoch": 0.4232642185537962, "percentage": 21.17, "elapsed_time": "1:36:15", "remaining_time": "5:58:31", "throughput": 3955.98, "total_tokens": 22846864}
{"current_steps": 3460, "total_steps": 16324, "loss": 0.5134, "lr": 9.300194543065835e-05, "epoch": 0.42387675722029955, "percentage": 21.2, "elapsed_time": "1:37:14", "remaining_time": "6:01:33", "throughput": 3921.17, "total_tokens": 22879984}
{"current_steps": 3465, "total_steps": 16324, "loss": 0.4776, "lr": 9.2976081334713e-05, "epoch": 0.42448929588680284, "percentage": 21.23, "elapsed_time": "1:38:14", "remaining_time": "6:04:35", "throughput": 3887.33, "total_tokens": 22913752}
{"current_steps": 3470, "total_steps": 16324, "loss": 0.4034, "lr": 9.295017314157354e-05, "epoch": 0.4251018345533062, "percentage": 21.26, "elapsed_time": "1:39:13", "remaining_time": "6:07:34", "throughput": 3854.24, "total_tokens": 22947360}
{"current_steps": 3475, "total_steps": 16324, "loss": 0.4691, "lr": 9.292422087782408e-05, "epoch": 0.4257143732198095, "percentage": 21.29, "elapsed_time": "1:40:13", "remaining_time": "6:10:33", "throughput": 3821.78, "total_tokens": 22980784}
{"current_steps": 3480, "total_steps": 16324, "loss": 0.3737, "lr": 9.289822457009388e-05, "epoch": 0.42632691188631283, "percentage": 21.32, "elapsed_time": "1:41:13", "remaining_time": "6:13:36", "throughput": 3789.42, "total_tokens": 23015304}
{"current_steps": 3485, "total_steps": 16324, "loss": 0.431, "lr": 9.287218424505736e-05, "epoch": 0.4269394505528161, "percentage": 21.35, "elapsed_time": "1:42:12", "remaining_time": "6:16:32", "throughput": 3758.4, "total_tokens": 23048248}
{"current_steps": 3490, "total_steps": 16324, "loss": 0.3861, "lr": 9.284609992943423e-05, "epoch": 0.4275519892193195, "percentage": 21.38, "elapsed_time": "1:43:11", "remaining_time": "6:19:28", "throughput": 3727.97, "total_tokens": 23082144}
{"current_steps": 3495, "total_steps": 16324, "loss": 0.4227, "lr": 9.281997164998923e-05, "epoch": 0.42816452788582277, "percentage": 21.41, "elapsed_time": "1:44:11", "remaining_time": "6:22:28", "throughput": 3697.36, "total_tokens": 23115584}
{"current_steps": 3500, "total_steps": 16324, "loss": 0.4345, "lr": 9.279379943353226e-05, "epoch": 0.4287770665523261, "percentage": 21.44, "elapsed_time": "1:45:11", "remaining_time": "6:25:23", "throughput": 3668.05, "total_tokens": 23149384}
{"current_steps": 3500, "total_steps": 16324, "eval_loss": 0.2862895727157593, "epoch": 0.4287770665523261, "percentage": 21.44, "elapsed_time": "1:45:30", "remaining_time": "6:26:35", "throughput": 3656.67, "total_tokens": 23149384}
{"current_steps": 3505, "total_steps": 16324, "loss": 0.4072, "lr": 9.276758330691826e-05, "epoch": 0.42938960521882946, "percentage": 21.47, "elapsed_time": "1:46:35", "remaining_time": "6:29:50", "throughput": 3624.89, "total_tokens": 23183112}
{"current_steps": 3510, "total_steps": 16324, "loss": 0.5979, "lr": 9.27413232970473e-05, "epoch": 0.43000214388533275, "percentage": 21.5, "elapsed_time": "1:47:36", "remaining_time": "6:32:49", "throughput": 3596.05, "total_tokens": 23216232}
{"current_steps": 3515, "total_steps": 16324, "loss": 0.5025, "lr": 9.271501943086437e-05, "epoch": 0.4306146825518361, "percentage": 21.53, "elapsed_time": "1:48:35", "remaining_time": "6:35:44", "throughput": 3568.12, "total_tokens": 23249256}
{"current_steps": 3520, "total_steps": 16324, "loss": 0.4179, "lr": 9.268867173535957e-05, "epoch": 0.4312272212183394, "percentage": 21.56, "elapsed_time": "1:49:35", "remaining_time": "6:38:38", "throughput": 3540.91, "total_tokens": 23283064}
{"current_steps": 3525, "total_steps": 16324, "loss": 0.3838, "lr": 9.266228023756791e-05, "epoch": 0.43183975988484274, "percentage": 21.59, "elapsed_time": "1:50:36", "remaining_time": "6:41:35", "throughput": 3513.46, "total_tokens": 23316448}
{"current_steps": 3530, "total_steps": 16324, "loss": 0.4015, "lr": 9.263584496456937e-05, "epoch": 0.43245229855134604, "percentage": 21.62, "elapsed_time": "1:51:35", "remaining_time": "6:44:27", "throughput": 3487.26, "total_tokens": 23349592}
{"current_steps": 3535, "total_steps": 16324, "loss": 0.4496, "lr": 9.260936594348882e-05, "epoch": 0.4330648372178494, "percentage": 21.66, "elapsed_time": "1:52:35", "remaining_time": "6:47:18", "throughput": 3461.54, "total_tokens": 23383184}
{"current_steps": 3540, "total_steps": 16324, "loss": 0.4206, "lr": 9.258284320149605e-05, "epoch": 0.4336773758843527, "percentage": 21.69, "elapsed_time": "1:53:35", "remaining_time": "6:50:11", "throughput": 3436.0, "total_tokens": 23416576}
{"current_steps": 3545, "total_steps": 16324, "loss": 0.4029, "lr": 9.25562767658057e-05, "epoch": 0.434289914550856, "percentage": 21.72, "elapsed_time": "1:54:34", "remaining_time": "6:53:01", "throughput": 3411.18, "total_tokens": 23450336}
{"current_steps": 3550, "total_steps": 16324, "loss": 0.4101, "lr": 9.252966666367722e-05, "epoch": 0.4349024532173593, "percentage": 21.75, "elapsed_time": "1:55:33", "remaining_time": "6:55:48", "throughput": 3387.08, "total_tokens": 23484072}
{"current_steps": 3550, "total_steps": 16324, "eval_loss": 0.33307310938835144, "epoch": 0.4349024532173593, "percentage": 21.75, "elapsed_time": "1:55:52", "remaining_time": "6:56:58", "throughput": 3377.61, "total_tokens": 23484072}
{"current_steps": 3555, "total_steps": 16324, "loss": 0.4278, "lr": 9.250301292241492e-05, "epoch": 0.43551499188386267, "percentage": 21.78, "elapsed_time": "1:56:58", "remaining_time": "7:00:09", "throughput": 3350.56, "total_tokens": 23516352}
{"current_steps": 3560, "total_steps": 16324, "loss": 0.4285, "lr": 9.247631556936782e-05, "epoch": 0.436127530550366, "percentage": 21.81, "elapsed_time": "1:57:57", "remaining_time": "7:02:54", "throughput": 3327.58, "total_tokens": 23550240}
{"current_steps": 3565, "total_steps": 16324, "loss": 0.4004, "lr": 9.244957463192976e-05, "epoch": 0.4367400692168693, "percentage": 21.84, "elapsed_time": "1:58:56", "remaining_time": "7:05:40", "throughput": 3304.91, "total_tokens": 23584792}
{"current_steps": 3570, "total_steps": 16324, "loss": 0.3487, "lr": 9.242279013753922e-05, "epoch": 0.43735260788337266, "percentage": 21.87, "elapsed_time": "1:59:55", "remaining_time": "7:08:27", "throughput": 3282.05, "total_tokens": 23617416}
{"current_steps": 3575, "total_steps": 16324, "loss": 0.4213, "lr": 9.239596211367948e-05, "epoch": 0.43796514654987595, "percentage": 21.9, "elapsed_time": "2:00:54", "remaining_time": "7:11:10", "throughput": 3260.2, "total_tokens": 23650840}
{"current_steps": 3580, "total_steps": 16324, "loss": 0.4415, "lr": 9.236909058787837e-05, "epoch": 0.4385776852163793, "percentage": 21.93, "elapsed_time": "2:01:53", "remaining_time": "7:13:53", "throughput": 3238.58, "total_tokens": 23684112}
{"current_steps": 3585, "total_steps": 16324, "loss": 0.5179, "lr": 9.234217558770844e-05, "epoch": 0.4391902238828826, "percentage": 21.96, "elapsed_time": "2:02:52", "remaining_time": "7:16:37", "throughput": 3217.06, "total_tokens": 23718136}
{"current_steps": 3590, "total_steps": 16324, "loss": 0.4178, "lr": 9.231521714078681e-05, "epoch": 0.43980276254938594, "percentage": 21.99, "elapsed_time": "2:03:52", "remaining_time": "7:19:23", "throughput": 3195.75, "total_tokens": 23751840}
{"current_steps": 3595, "total_steps": 16324, "loss": 0.3969, "lr": 9.228821527477519e-05, "epoch": 0.44041530121588923, "percentage": 22.02, "elapsed_time": "2:04:51", "remaining_time": "7:22:04", "throughput": 3175.1, "total_tokens": 23785544}
{"current_steps": 3600, "total_steps": 16324, "loss": 0.3588, "lr": 9.226117001737983e-05, "epoch": 0.4410278398823926, "percentage": 22.05, "elapsed_time": "2:05:50", "remaining_time": "7:24:47", "throughput": 3154.55, "total_tokens": 23819160}
{"current_steps": 3600, "total_steps": 16324, "eval_loss": 0.2863926589488983, "epoch": 0.4410278398823926, "percentage": 22.05, "elapsed_time": "2:06:10", "remaining_time": "7:25:56", "throughput": 3146.46, "total_tokens": 23819160}
{"current_steps": 3605, "total_steps": 16324, "loss": 0.4315, "lr": 9.223408139635153e-05, "epoch": 0.4416403785488959, "percentage": 22.08, "elapsed_time": "2:07:15", "remaining_time": "7:28:58", "throughput": 3123.87, "total_tokens": 23852120}
{"current_steps": 3610, "total_steps": 16324, "loss": 0.3867, "lr": 9.220694943948556e-05, "epoch": 0.4422529172153992, "percentage": 22.11, "elapsed_time": "2:08:14", "remaining_time": "7:31:37", "throughput": 3104.43, "total_tokens": 23885720}
{"current_steps": 3615, "total_steps": 16324, "loss": 0.405, "lr": 9.217977417462167e-05, "epoch": 0.44286545588190257, "percentage": 22.15, "elapsed_time": "2:09:12", "remaining_time": "7:34:16", "throughput": 3085.14, "total_tokens": 23918856}
{"current_steps": 3620, "total_steps": 16324, "loss": 0.402, "lr": 9.215255562964404e-05, "epoch": 0.44347799454840586, "percentage": 22.18, "elapsed_time": "2:10:13", "remaining_time": "7:37:01", "throughput": 3065.33, "total_tokens": 23951984}
{"current_steps": 3625, "total_steps": 16324, "loss": 0.3315, "lr": 9.212529383248127e-05, "epoch": 0.4440905332149092, "percentage": 22.21, "elapsed_time": "2:11:12", "remaining_time": "7:39:40", "throughput": 3046.51, "total_tokens": 23984912}
{"current_steps": 3630, "total_steps": 16324, "loss": 0.4262, "lr": 9.209798881110634e-05, "epoch": 0.4447030718814125, "percentage": 22.24, "elapsed_time": "2:12:11", "remaining_time": "7:42:17", "throughput": 3028.13, "total_tokens": 24018568}
{"current_steps": 3635, "total_steps": 16324, "loss": 0.3885, "lr": 9.207064059353654e-05, "epoch": 0.44531561054791585, "percentage": 22.27, "elapsed_time": "2:13:12", "remaining_time": "7:45:00", "throughput": 3009.34, "total_tokens": 24052536}
{"current_steps": 3640, "total_steps": 16324, "loss": 0.3797, "lr": 9.204324920783355e-05, "epoch": 0.44592814921441914, "percentage": 22.3, "elapsed_time": "2:14:12", "remaining_time": "7:47:38", "throughput": 2991.26, "total_tokens": 24086168}
{"current_steps": 3645, "total_steps": 16324, "loss": 0.3571, "lr": 9.20158146821033e-05, "epoch": 0.4465406878809225, "percentage": 22.33, "elapsed_time": "2:15:11", "remaining_time": "7:50:16", "throughput": 2973.5, "total_tokens": 24120160}
{"current_steps": 3650, "total_steps": 16324, "loss": 0.4439, "lr": 9.1988337044496e-05, "epoch": 0.4471532265474258, "percentage": 22.36, "elapsed_time": "2:16:12", "remaining_time": "7:52:56", "throughput": 2955.62, "total_tokens": 24153672}
{"current_steps": 3650, "total_steps": 16324, "eval_loss": 0.3060784935951233, "epoch": 0.4471532265474258, "percentage": 22.36, "elapsed_time": "2:16:31", "remaining_time": "7:54:03", "throughput": 2948.66, "total_tokens": 24153672}
{"current_steps": 3655, "total_steps": 16324, "loss": 0.3653, "lr": 9.19608163232061e-05, "epoch": 0.44776576521392913, "percentage": 22.39, "elapsed_time": "2:17:35", "remaining_time": "7:56:56", "throughput": 2929.77, "total_tokens": 24187312}
{"current_steps": 3660, "total_steps": 16324, "loss": 0.4087, "lr": 9.193325254647225e-05, "epoch": 0.4483783038804324, "percentage": 22.42, "elapsed_time": "2:18:34", "remaining_time": "7:59:29", "throughput": 2913.08, "total_tokens": 24220960}
{"current_steps": 3665, "total_steps": 16324, "loss": 0.4211, "lr": 9.190564574257727e-05, "epoch": 0.4489908425469358, "percentage": 22.45, "elapsed_time": "2:19:34", "remaining_time": "8:02:06", "throughput": 2896.08, "total_tokens": 24254224}
{"current_steps": 3670, "total_steps": 16324, "loss": 0.2987, "lr": 9.187799593984814e-05, "epoch": 0.4496033812134391, "percentage": 22.48, "elapsed_time": "2:20:33", "remaining_time": "8:04:39", "throughput": 2879.81, "total_tokens": 24288128}
{"current_steps": 3675, "total_steps": 16324, "loss": 0.4294, "lr": 9.185030316665597e-05, "epoch": 0.4502159198799424, "percentage": 22.51, "elapsed_time": "2:21:32", "remaining_time": "8:07:10", "throughput": 2863.83, "total_tokens": 24321200}
{"current_steps": 3680, "total_steps": 16324, "loss": 0.3842, "lr": 9.182256745141595e-05, "epoch": 0.45082845854644576, "percentage": 22.54, "elapsed_time": "2:22:32", "remaining_time": "8:09:45", "throughput": 2847.66, "total_tokens": 24354864}
{"current_steps": 3685, "total_steps": 16324, "loss": 0.463, "lr": 9.179478882258732e-05, "epoch": 0.45144099721294906, "percentage": 22.57, "elapsed_time": "2:23:30", "remaining_time": "8:12:14", "throughput": 2832.26, "total_tokens": 24388472}
{"current_steps": 3690, "total_steps": 16324, "loss": 0.4246, "lr": 9.176696730867339e-05, "epoch": 0.4520535358794524, "percentage": 22.6, "elapsed_time": "2:24:29", "remaining_time": "8:14:42", "throughput": 2816.97, "total_tokens": 24421688}
{"current_steps": 3695, "total_steps": 16324, "loss": 0.541, "lr": 9.173910293822145e-05, "epoch": 0.4526660745459557, "percentage": 22.64, "elapsed_time": "2:25:29", "remaining_time": "8:17:15", "throughput": 2801.52, "total_tokens": 24455024}
{"current_steps": 3700, "total_steps": 16324, "loss": 0.3865, "lr": 9.171119573982274e-05, "epoch": 0.45327861321245905, "percentage": 22.67, "elapsed_time": "2:26:28", "remaining_time": "8:19:44", "throughput": 2786.57, "total_tokens": 24489272}
{"current_steps": 3700, "total_steps": 16324, "eval_loss": 0.3225303292274475, "epoch": 0.45327861321245905, "percentage": 22.67, "elapsed_time": "2:26:47", "remaining_time": "8:20:50", "throughput": 2780.47, "total_tokens": 24489272}
{"current_steps": 3705, "total_steps": 16324, "loss": 0.3733, "lr": 9.168324574211249e-05, "epoch": 0.45389115187896234, "percentage": 22.7, "elapsed_time": "2:27:51", "remaining_time": "8:23:36", "throughput": 2764.18, "total_tokens": 24523128}
{"current_steps": 3710, "total_steps": 16324, "loss": 0.3909, "lr": 9.165525297376982e-05, "epoch": 0.4545036905454657, "percentage": 22.73, "elapsed_time": "2:28:51", "remaining_time": "8:26:07", "throughput": 2749.51, "total_tokens": 24557744}
{"current_steps": 3715, "total_steps": 16324, "loss": 0.4227, "lr": 9.162721746351774e-05, "epoch": 0.455116229211969, "percentage": 22.76, "elapsed_time": "2:29:51", "remaining_time": "8:28:38", "throughput": 2734.83, "total_tokens": 24590912}
{"current_steps": 3720, "total_steps": 16324, "loss": 0.4473, "lr": 9.159913924012314e-05, "epoch": 0.45572876787847233, "percentage": 22.79, "elapsed_time": "2:30:50", "remaining_time": "8:31:06", "throughput": 2720.61, "total_tokens": 24624144}
{"current_steps": 3725, "total_steps": 16324, "loss": 0.3752, "lr": 9.15710183323967e-05, "epoch": 0.4563413065449757, "percentage": 22.82, "elapsed_time": "2:31:50", "remaining_time": "8:33:33", "throughput": 2706.66, "total_tokens": 24658424}
{"current_steps": 3730, "total_steps": 16324, "loss": 0.384, "lr": 9.154285476919291e-05, "epoch": 0.45695384521147897, "percentage": 22.85, "elapsed_time": "2:32:50", "remaining_time": "8:36:04", "throughput": 2692.41, "total_tokens": 24691680}
{"current_steps": 3735, "total_steps": 16324, "loss": 0.3846, "lr": 9.151464857941003e-05, "epoch": 0.4575663838779823, "percentage": 22.88, "elapsed_time": "2:33:50", "remaining_time": "8:38:31", "throughput": 2678.65, "total_tokens": 24724784}
{"current_steps": 3740, "total_steps": 16324, "loss": 0.3126, "lr": 9.148639979199009e-05, "epoch": 0.4581789225444856, "percentage": 22.91, "elapsed_time": "2:34:49", "remaining_time": "8:40:57", "throughput": 2665.1, "total_tokens": 24758064}
{"current_steps": 3745, "total_steps": 16324, "loss": 0.4587, "lr": 9.145810843591879e-05, "epoch": 0.45879146121098896, "percentage": 22.94, "elapsed_time": "2:35:50", "remaining_time": "8:43:27", "throughput": 2651.37, "total_tokens": 24791832}
{"current_steps": 3750, "total_steps": 16324, "loss": 0.4547, "lr": 9.14297745402255e-05, "epoch": 0.45940399987749225, "percentage": 22.97, "elapsed_time": "2:36:50", "remaining_time": "8:45:53", "throughput": 2638.09, "total_tokens": 24825512}
{"current_steps": 3750, "total_steps": 16324, "eval_loss": 0.329493910074234, "epoch": 0.45940399987749225, "percentage": 22.97, "elapsed_time": "2:37:10", "remaining_time": "8:46:59", "throughput": 2632.58, "total_tokens": 24825512}
{"current_steps": 3755, "total_steps": 16324, "loss": 0.3968, "lr": 9.140139813398328e-05, "epoch": 0.4600165385439956, "percentage": 23.0, "elapsed_time": "2:38:14", "remaining_time": "8:49:40", "throughput": 2618.25, "total_tokens": 24859008}
{"current_steps": 3760, "total_steps": 16324, "loss": 0.4446, "lr": 9.137297924630876e-05, "epoch": 0.4606290772104989, "percentage": 23.03, "elapsed_time": "2:39:15", "remaining_time": "8:52:08", "throughput": 2605.17, "total_tokens": 24892584}
{"current_steps": 3765, "total_steps": 16324, "loss": 0.3624, "lr": 9.134451790636223e-05, "epoch": 0.46124161587700224, "percentage": 23.06, "elapsed_time": "2:40:14", "remaining_time": "8:54:29", "throughput": 2592.77, "total_tokens": 24926904}
{"current_steps": 3770, "total_steps": 16324, "loss": 0.3832, "lr": 9.131601414334748e-05, "epoch": 0.46185415454350554, "percentage": 23.09, "elapsed_time": "2:41:12", "remaining_time": "8:56:49", "throughput": 2580.46, "total_tokens": 24959880}
{"current_steps": 3775, "total_steps": 16324, "loss": 0.4412, "lr": 9.128746798651184e-05, "epoch": 0.4624666932100089, "percentage": 23.13, "elapsed_time": "2:42:12", "remaining_time": "8:59:13", "throughput": 2567.94, "total_tokens": 24993064}
{"current_steps": 3780, "total_steps": 16324, "loss": 0.3528, "lr": 9.125887946514615e-05, "epoch": 0.46307923187651223, "percentage": 23.16, "elapsed_time": "2:43:11", "remaining_time": "9:01:32", "throughput": 2556.06, "total_tokens": 25027216}
{"current_steps": 3785, "total_steps": 16324, "loss": 0.4049, "lr": 9.123024860858473e-05, "epoch": 0.4636917705430155, "percentage": 23.19, "elapsed_time": "2:44:09", "remaining_time": "9:03:49", "throughput": 2544.27, "total_tokens": 25059744}
{"current_steps": 3790, "total_steps": 16324, "loss": 0.4077, "lr": 9.120157544620532e-05, "epoch": 0.4643043092095189, "percentage": 23.22, "elapsed_time": "2:45:09", "remaining_time": "9:06:13", "throughput": 2532.13, "total_tokens": 25093136}
{"current_steps": 3795, "total_steps": 16324, "loss": 0.3809, "lr": 9.117286000742907e-05, "epoch": 0.46491684787602217, "percentage": 23.25, "elapsed_time": "2:46:08", "remaining_time": "9:08:30", "throughput": 2520.54, "total_tokens": 25126312}
{"current_steps": 3800, "total_steps": 16324, "loss": 0.3747, "lr": 9.114410232172051e-05, "epoch": 0.4655293865425255, "percentage": 23.28, "elapsed_time": "2:47:07", "remaining_time": "9:10:47", "throughput": 2509.12, "total_tokens": 25159152}
{"current_steps": 3800, "total_steps": 16324, "eval_loss": 0.3077337443828583, "epoch": 0.4655293865425255, "percentage": 23.28, "elapsed_time": "2:47:26", "remaining_time": "9:11:50", "throughput": 2504.31, "total_tokens": 25159152}
{"current_steps": 3805, "total_steps": 16324, "loss": 0.4342, "lr": 9.111530241858752e-05, "epoch": 0.4661419252090288, "percentage": 23.31, "elapsed_time": "2:48:31", "remaining_time": "9:14:29", "throughput": 2491.34, "total_tokens": 25192400}
{"current_steps": 3810, "total_steps": 16324, "loss": 0.3581, "lr": 9.108646032758134e-05, "epoch": 0.46675446387553216, "percentage": 23.34, "elapsed_time": "2:49:30", "remaining_time": "9:16:45", "throughput": 2480.28, "total_tokens": 25225960}
{"current_steps": 3815, "total_steps": 16324, "loss": 0.3421, "lr": 9.105757607829644e-05, "epoch": 0.46736700254203545, "percentage": 23.37, "elapsed_time": "2:50:29", "remaining_time": "9:19:00", "throughput": 2469.37, "total_tokens": 25259928}
{"current_steps": 3820, "total_steps": 16324, "loss": 0.5049, "lr": 9.102864970037055e-05, "epoch": 0.4679795412085388, "percentage": 23.4, "elapsed_time": "2:51:29", "remaining_time": "9:21:20", "throughput": 2458.17, "total_tokens": 25293656}
{"current_steps": 3825, "total_steps": 16324, "loss": 0.5269, "lr": 9.099968122348467e-05, "epoch": 0.4685920798750421, "percentage": 23.43, "elapsed_time": "2:52:28", "remaining_time": "9:23:35", "throughput": 2447.41, "total_tokens": 25326880}
{"current_steps": 3830, "total_steps": 16324, "loss": 0.4591, "lr": 9.097067067736298e-05, "epoch": 0.46920461854154544, "percentage": 23.46, "elapsed_time": "2:53:27", "remaining_time": "9:25:49", "throughput": 2436.8, "total_tokens": 25360360}
{"current_steps": 3835, "total_steps": 16324, "loss": 0.4192, "lr": 9.094161809177281e-05, "epoch": 0.4698171572080488, "percentage": 23.49, "elapsed_time": "2:54:27", "remaining_time": "9:28:08", "throughput": 2426.02, "total_tokens": 25394592}
{"current_steps": 3840, "total_steps": 16324, "loss": 0.3305, "lr": 9.091252349652465e-05, "epoch": 0.4704296958745521, "percentage": 23.52, "elapsed_time": "2:55:26", "remaining_time": "9:30:22", "throughput": 2415.58, "total_tokens": 25427840}
{"current_steps": 3845, "total_steps": 16324, "loss": 0.4276, "lr": 9.088338692147205e-05, "epoch": 0.4710422345410554, "percentage": 23.55, "elapsed_time": "2:56:26", "remaining_time": "9:32:37", "throughput": 2405.22, "total_tokens": 25461840}
{"current_steps": 3850, "total_steps": 16324, "loss": 0.3792, "lr": 9.085420839651171e-05, "epoch": 0.4716547732075587, "percentage": 23.58, "elapsed_time": "2:57:26", "remaining_time": "9:34:53", "throughput": 2394.81, "total_tokens": 25495704}
{"current_steps": 3850, "total_steps": 16324, "eval_loss": 0.22186821699142456, "epoch": 0.4716547732075587, "percentage": 23.58, "elapsed_time": "2:57:45", "remaining_time": "9:35:56", "throughput": 2390.46, "total_tokens": 25495704}
{"current_steps": 3855, "total_steps": 16324, "loss": 0.3757, "lr": 9.08249879515833e-05, "epoch": 0.47226731187406207, "percentage": 23.62, "elapsed_time": "2:58:51", "remaining_time": "9:38:30", "throughput": 2378.97, "total_tokens": 25529528}
{"current_steps": 3860, "total_steps": 16324, "loss": 0.3942, "lr": 9.079572561666956e-05, "epoch": 0.47287985054056536, "percentage": 23.65, "elapsed_time": "2:59:51", "remaining_time": "9:40:45", "throughput": 2368.82, "total_tokens": 25562760}
{"current_steps": 3865, "total_steps": 16324, "loss": 0.3056, "lr": 9.076642142179616e-05, "epoch": 0.4734923892070687, "percentage": 23.68, "elapsed_time": "3:00:50", "remaining_time": "9:42:58", "throughput": 2358.97, "total_tokens": 25596896}
{"current_steps": 3870, "total_steps": 16324, "loss": 0.4779, "lr": 9.07370753970318e-05, "epoch": 0.474104927873572, "percentage": 23.71, "elapsed_time": "3:01:51", "remaining_time": "9:45:14", "throughput": 2348.89, "total_tokens": 25630144}
{"current_steps": 3875, "total_steps": 16324, "loss": 0.3364, "lr": 9.0707687572488e-05, "epoch": 0.47471746654007535, "percentage": 23.74, "elapsed_time": "3:02:50", "remaining_time": "9:47:24", "throughput": 2339.34, "total_tokens": 25664128}
{"current_steps": 3880, "total_steps": 16324, "loss": 0.3854, "lr": 9.067825797831924e-05, "epoch": 0.47533000520657864, "percentage": 23.77, "elapsed_time": "3:03:49", "remaining_time": "9:49:33", "throughput": 2329.99, "total_tokens": 25698152}
{"current_steps": 3885, "total_steps": 16324, "loss": 0.3675, "lr": 9.064878664472282e-05, "epoch": 0.475942543873082, "percentage": 23.8, "elapsed_time": "3:04:49", "remaining_time": "9:51:45", "throughput": 2320.43, "total_tokens": 25732096}
{"current_steps": 3890, "total_steps": 16324, "loss": 0.3702, "lr": 9.061927360193894e-05, "epoch": 0.4765550825395853, "percentage": 23.83, "elapsed_time": "3:05:47", "remaining_time": "9:53:52", "throughput": 2311.29, "total_tokens": 25765792}
{"current_steps": 3895, "total_steps": 16324, "loss": 0.407, "lr": 9.058971888025052e-05, "epoch": 0.47716762120608863, "percentage": 23.86, "elapsed_time": "3:06:46", "remaining_time": "9:56:00", "throughput": 2302.11, "total_tokens": 25798856}
{"current_steps": 3900, "total_steps": 16324, "loss": 0.4111, "lr": 9.056012250998325e-05, "epoch": 0.477780159872592, "percentage": 23.89, "elapsed_time": "3:07:46", "remaining_time": "9:58:09", "throughput": 2292.86, "total_tokens": 25831784}
{"current_steps": 3900, "total_steps": 16324, "eval_loss": 0.4086068868637085, "epoch": 0.477780159872592, "percentage": 23.89, "elapsed_time": "3:08:05", "remaining_time": "9:59:11", "throughput": 2288.93, "total_tokens": 25831784}
{"current_steps": 3905, "total_steps": 16324, "loss": 0.3328, "lr": 9.053048452150562e-05, "epoch": 0.4783926985390953, "percentage": 23.92, "elapsed_time": "3:09:09", "remaining_time": "10:01:34", "throughput": 2278.99, "total_tokens": 25865600}
{"current_steps": 3910, "total_steps": 16324, "loss": 0.4061, "lr": 9.050080494522875e-05, "epoch": 0.4790052372055986, "percentage": 23.95, "elapsed_time": "3:10:08", "remaining_time": "10:03:40", "throughput": 2270.21, "total_tokens": 25899352}
{"current_steps": 3915, "total_steps": 16324, "loss": 0.4168, "lr": 9.047108381160647e-05, "epoch": 0.4796177758721019, "percentage": 23.98, "elapsed_time": "3:11:08", "remaining_time": "10:05:51", "throughput": 2261.14, "total_tokens": 25932640}
{"current_steps": 3920, "total_steps": 16324, "loss": 0.395, "lr": 9.044132115113525e-05, "epoch": 0.48023031453860526, "percentage": 24.01, "elapsed_time": "3:12:07", "remaining_time": "10:07:56", "throughput": 2252.51, "total_tokens": 25966192}
{"current_steps": 3925, "total_steps": 16324, "loss": 0.3331, "lr": 9.041151699435417e-05, "epoch": 0.48084285320510856, "percentage": 24.04, "elapsed_time": "3:13:06", "remaining_time": "10:10:02", "throughput": 2243.92, "total_tokens": 25999896}
{"current_steps": 3930, "total_steps": 16324, "loss": 0.3548, "lr": 9.038167137184488e-05, "epoch": 0.4814553918716119, "percentage": 24.07, "elapsed_time": "3:14:06", "remaining_time": "10:12:10", "throughput": 2235.25, "total_tokens": 26033344}
{"current_steps": 3935, "total_steps": 16324, "loss": 0.3051, "lr": 9.035178431423159e-05, "epoch": 0.4820679305381152, "percentage": 24.11, "elapsed_time": "3:15:05", "remaining_time": "10:14:14", "throughput": 2226.87, "total_tokens": 26066968}
{"current_steps": 3940, "total_steps": 16324, "loss": 0.3944, "lr": 9.0321855852181e-05, "epoch": 0.48268046920461855, "percentage": 24.14, "elapsed_time": "3:16:04", "remaining_time": "10:16:18", "throughput": 2218.5, "total_tokens": 26100048}
{"current_steps": 3945, "total_steps": 16324, "loss": 0.3257, "lr": 9.029188601640234e-05, "epoch": 0.48329300787112184, "percentage": 24.17, "elapsed_time": "3:17:05", "remaining_time": "10:18:27", "throughput": 2210.0, "total_tokens": 26134240}
{"current_steps": 3950, "total_steps": 16324, "loss": 0.3546, "lr": 9.026187483764725e-05, "epoch": 0.4839055465376252, "percentage": 24.2, "elapsed_time": "3:18:04", "remaining_time": "10:20:30", "throughput": 2201.83, "total_tokens": 26167704}
{"current_steps": 3950, "total_steps": 16324, "eval_loss": 0.31338992714881897, "epoch": 0.4839055465376252, "percentage": 24.2, "elapsed_time": "3:18:23", "remaining_time": "10:21:30", "throughput": 2198.25, "total_tokens": 26167704}
{"current_steps": 3955, "total_steps": 16324, "loss": 0.4172, "lr": 9.023182234670981e-05, "epoch": 0.48451808520412853, "percentage": 24.23, "elapsed_time": "3:19:28", "remaining_time": "10:23:49", "throughput": 2189.21, "total_tokens": 26200904}
{"current_steps": 3960, "total_steps": 16324, "loss": 0.3223, "lr": 9.020172857442647e-05, "epoch": 0.4851306238706318, "percentage": 24.26, "elapsed_time": "3:20:28", "remaining_time": "10:25:56", "throughput": 2181.03, "total_tokens": 26234944}
{"current_steps": 3965, "total_steps": 16324, "loss": 0.3836, "lr": 9.017159355167609e-05, "epoch": 0.4857431625371352, "percentage": 24.29, "elapsed_time": "3:21:28", "remaining_time": "10:27:59", "throughput": 2173.04, "total_tokens": 26268272}
{"current_steps": 3970, "total_steps": 16324, "loss": 0.3538, "lr": 9.014141730937978e-05, "epoch": 0.48635570120363847, "percentage": 24.32, "elapsed_time": "3:22:27", "remaining_time": "10:30:01", "throughput": 2165.16, "total_tokens": 26301736}
{"current_steps": 3975, "total_steps": 16324, "loss": 0.4015, "lr": 9.011119987850103e-05, "epoch": 0.4869682398701418, "percentage": 24.35, "elapsed_time": "3:23:27", "remaining_time": "10:32:06", "throughput": 2157.18, "total_tokens": 26334800}
{"current_steps": 3980, "total_steps": 16324, "loss": 0.4309, "lr": 9.008094129004552e-05, "epoch": 0.4875807785366451, "percentage": 24.38, "elapsed_time": "3:24:26", "remaining_time": "10:34:05", "throughput": 2149.55, "total_tokens": 26368176}
{"current_steps": 3985, "total_steps": 16324, "loss": 0.3684, "lr": 9.00506415750612e-05, "epoch": 0.48819331720314846, "percentage": 24.41, "elapsed_time": "3:25:25", "remaining_time": "10:36:05", "throughput": 2141.93, "total_tokens": 26401184}
{"current_steps": 3990, "total_steps": 16324, "loss": 0.3302, "lr": 9.00203007646382e-05, "epoch": 0.48880585586965175, "percentage": 24.44, "elapsed_time": "3:26:25", "remaining_time": "10:38:07", "throughput": 2134.33, "total_tokens": 26435504}
{"current_steps": 3995, "total_steps": 16324, "loss": 0.4337, "lr": 8.998991888990886e-05, "epoch": 0.4894183945361551, "percentage": 24.47, "elapsed_time": "3:27:24", "remaining_time": "10:40:06", "throughput": 2126.9, "total_tokens": 26468848}
{"current_steps": 4000, "total_steps": 16324, "loss": 0.3654, "lr": 8.99594959820476e-05, "epoch": 0.4900309332026584, "percentage": 24.5, "elapsed_time": "3:28:23", "remaining_time": "10:42:02", "throughput": 2119.66, "total_tokens": 26502544}
{"current_steps": 4000, "total_steps": 16324, "eval_loss": 0.21694067120552063, "epoch": 0.4900309332026584, "percentage": 24.5, "elapsed_time": "3:28:42", "remaining_time": "10:43:01", "throughput": 2116.41, "total_tokens": 26502544}
{"current_steps": 4005, "total_steps": 16324, "loss": 0.3121, "lr": 8.992903207227097e-05, "epoch": 0.49064347186916174, "percentage": 24.53, "elapsed_time": "3:29:47", "remaining_time": "10:45:17", "throughput": 2108.11, "total_tokens": 26535712}
{"current_steps": 4010, "total_steps": 16324, "loss": 0.4167, "lr": 8.98985271918376e-05, "epoch": 0.4912560105356651, "percentage": 24.57, "elapsed_time": "3:30:47", "remaining_time": "10:47:17", "throughput": 2100.79, "total_tokens": 26568952}
{"current_steps": 4015, "total_steps": 16324, "loss": 0.3339, "lr": 8.986798137204813e-05, "epoch": 0.4918685492021684, "percentage": 24.6, "elapsed_time": "3:31:45", "remaining_time": "10:49:12", "throughput": 2093.73, "total_tokens": 26602464}
{"current_steps": 4020, "total_steps": 16324, "loss": 0.3833, "lr": 8.983739464424522e-05, "epoch": 0.49248108786867173, "percentage": 24.63, "elapsed_time": "3:32:44", "remaining_time": "10:51:09", "throughput": 2086.65, "total_tokens": 26635984}
{"current_steps": 4025, "total_steps": 16324, "loss": 0.3197, "lr": 8.980676703981354e-05, "epoch": 0.493093626535175, "percentage": 24.66, "elapsed_time": "3:33:45", "remaining_time": "10:53:09", "throughput": 2079.5, "total_tokens": 26670048}
{"current_steps": 4030, "total_steps": 16324, "loss": 0.367, "lr": 8.977609859017964e-05, "epoch": 0.49370616520167837, "percentage": 24.69, "elapsed_time": "3:34:43", "remaining_time": "10:55:03", "throughput": 2072.65, "total_tokens": 26703248}
{"current_steps": 4035, "total_steps": 16324, "loss": 0.474, "lr": 8.974538932681204e-05, "epoch": 0.49431870386818166, "percentage": 24.72, "elapsed_time": "3:35:42", "remaining_time": "10:56:57", "throughput": 2065.82, "total_tokens": 26736920}
{"current_steps": 4040, "total_steps": 16324, "loss": 0.392, "lr": 8.971463928122113e-05, "epoch": 0.494931242534685, "percentage": 24.75, "elapsed_time": "3:36:42", "remaining_time": "10:58:55", "throughput": 2058.82, "total_tokens": 26770000}
{"current_steps": 4045, "total_steps": 16324, "loss": 0.3828, "lr": 8.968384848495908e-05, "epoch": 0.4955437812011883, "percentage": 24.78, "elapsed_time": "3:37:41", "remaining_time": "11:00:48", "throughput": 2052.12, "total_tokens": 26803344}
{"current_steps": 4050, "total_steps": 16324, "loss": 0.4304, "lr": 8.965301696961994e-05, "epoch": 0.49615631986769165, "percentage": 24.81, "elapsed_time": "3:38:40", "remaining_time": "11:02:41", "throughput": 2045.5, "total_tokens": 26837152}
{"current_steps": 4050, "total_steps": 16324, "eval_loss": 0.5118626356124878, "epoch": 0.49615631986769165, "percentage": 24.81, "elapsed_time": "3:38:59", "remaining_time": "11:03:40", "throughput": 2042.49, "total_tokens": 26837152}
{"current_steps": 4055, "total_steps": 16324, "loss": 0.4811, "lr": 8.962214476683954e-05, "epoch": 0.49676885853419495, "percentage": 24.84, "elapsed_time": "3:40:05", "remaining_time": "11:05:55", "throughput": 2034.8, "total_tokens": 26870528}
{"current_steps": 4060, "total_steps": 16324, "loss": 0.4701, "lr": 8.959123190829544e-05, "epoch": 0.4973813972006983, "percentage": 24.87, "elapsed_time": "3:41:04", "remaining_time": "11:07:47", "throughput": 2028.21, "total_tokens": 26903176}
{"current_steps": 4065, "total_steps": 16324, "loss": 0.4449, "lr": 8.956027842570692e-05, "epoch": 0.49799393586720164, "percentage": 24.9, "elapsed_time": "3:42:03", "remaining_time": "11:09:40", "throughput": 2021.75, "total_tokens": 26936832}
{"current_steps": 4070, "total_steps": 16324, "loss": 0.3572, "lr": 8.952928435083491e-05, "epoch": 0.49860647453370494, "percentage": 24.93, "elapsed_time": "3:43:04", "remaining_time": "11:11:38", "throughput": 2015.02, "total_tokens": 26970504}
{"current_steps": 4075, "total_steps": 16324, "loss": 0.3907, "lr": 8.949824971548206e-05, "epoch": 0.4992190132002083, "percentage": 24.96, "elapsed_time": "3:44:03", "remaining_time": "11:13:30", "throughput": 2008.62, "total_tokens": 27003736}
{"current_steps": 4080, "total_steps": 16324, "loss": 0.361, "lr": 8.946717455149259e-05, "epoch": 0.4998315518667116, "percentage": 24.99, "elapsed_time": "3:45:03", "remaining_time": "11:15:22", "throughput": 2002.34, "total_tokens": 27037648}
{"current_steps": 4085, "total_steps": 16324, "loss": 0.3149, "lr": 8.943605889075233e-05, "epoch": 0.5004440905332149, "percentage": 25.02, "elapsed_time": "3:46:03", "remaining_time": "11:17:17", "throughput": 1995.9, "total_tokens": 27071560}
{"current_steps": 4090, "total_steps": 16324, "loss": 0.3322, "lr": 8.940490276518864e-05, "epoch": 0.5010566291997183, "percentage": 25.06, "elapsed_time": "3:47:03", "remaining_time": "11:19:10", "throughput": 1989.64, "total_tokens": 27106032}
{"current_steps": 4095, "total_steps": 16324, "loss": 0.3581, "lr": 8.937370620677042e-05, "epoch": 0.5016691678662215, "percentage": 25.09, "elapsed_time": "3:48:03", "remaining_time": "11:21:01", "throughput": 1983.46, "total_tokens": 27139656}
{"current_steps": 4100, "total_steps": 16324, "loss": 0.3471, "lr": 8.934246924750804e-05, "epoch": 0.5022817065327249, "percentage": 25.12, "elapsed_time": "3:49:03", "remaining_time": "11:22:55", "throughput": 1977.2, "total_tokens": 27173784}
{"current_steps": 4100, "total_steps": 16324, "eval_loss": 0.20771612226963043, "epoch": 0.5022817065327249, "percentage": 25.12, "elapsed_time": "3:49:22", "remaining_time": "11:23:53", "throughput": 1974.44, "total_tokens": 27173784}
{"current_steps": 4105, "total_steps": 16324, "loss": 0.3476, "lr": 8.931119191945334e-05, "epoch": 0.5028942451992282, "percentage": 25.15, "elapsed_time": "3:50:27", "remaining_time": "11:25:58", "throughput": 1967.75, "total_tokens": 27208344}
{"current_steps": 4110, "total_steps": 16324, "loss": 0.3914, "lr": 8.927987425469957e-05, "epoch": 0.5035067838657316, "percentage": 25.18, "elapsed_time": "3:51:25", "remaining_time": "11:27:45", "throughput": 1961.9, "total_tokens": 27242560}
{"current_steps": 4115, "total_steps": 16324, "loss": 0.3901, "lr": 8.92485162853814e-05, "epoch": 0.5041193225322348, "percentage": 25.21, "elapsed_time": "3:52:25", "remaining_time": "11:29:34", "throughput": 1955.99, "total_tokens": 27276256}
{"current_steps": 4120, "total_steps": 16324, "loss": 0.3934, "lr": 8.921711804367481e-05, "epoch": 0.5047318611987381, "percentage": 25.24, "elapsed_time": "3:53:24", "remaining_time": "11:31:23", "throughput": 1950.1, "total_tokens": 27310360}
{"current_steps": 4125, "total_steps": 16324, "loss": 0.3704, "lr": 8.918567956179711e-05, "epoch": 0.5053443998652415, "percentage": 25.27, "elapsed_time": "3:54:23", "remaining_time": "11:33:09", "throughput": 1944.36, "total_tokens": 27343672}
{"current_steps": 4130, "total_steps": 16324, "loss": 0.3748, "lr": 8.915420087200694e-05, "epoch": 0.5059569385317448, "percentage": 25.3, "elapsed_time": "3:55:22", "remaining_time": "11:34:57", "throughput": 1938.57, "total_tokens": 27377504}
{"current_steps": 4135, "total_steps": 16324, "loss": 0.3259, "lr": 8.912268200660418e-05, "epoch": 0.5065694771982482, "percentage": 25.33, "elapsed_time": "3:56:22", "remaining_time": "11:36:46", "throughput": 1932.75, "total_tokens": 27411072}
{"current_steps": 4140, "total_steps": 16324, "loss": 0.4131, "lr": 8.90911229979299e-05, "epoch": 0.5071820158647514, "percentage": 25.36, "elapsed_time": "3:57:20", "remaining_time": "11:38:31", "throughput": 1927.12, "total_tokens": 27444096}
{"current_steps": 4145, "total_steps": 16324, "loss": 0.3484, "lr": 8.905952387836639e-05, "epoch": 0.5077945545312548, "percentage": 25.39, "elapsed_time": "3:58:19", "remaining_time": "11:40:16", "throughput": 1921.53, "total_tokens": 27477864}
{"current_steps": 4150, "total_steps": 16324, "loss": 0.3778, "lr": 8.90278846803371e-05, "epoch": 0.5084070931977581, "percentage": 25.42, "elapsed_time": "3:59:19", "remaining_time": "11:42:04", "throughput": 1915.91, "total_tokens": 27512048}
{"current_steps": 4150, "total_steps": 16324, "eval_loss": 0.1999550759792328, "epoch": 0.5084070931977581, "percentage": 25.42, "elapsed_time": "3:59:39", "remaining_time": "11:43:01", "throughput": 1913.32, "total_tokens": 27512048}
{"current_steps": 4155, "total_steps": 16324, "loss": 0.3974, "lr": 8.89962054363066e-05, "epoch": 0.5090196318642615, "percentage": 25.45, "elapsed_time": "4:00:43", "remaining_time": "11:45:02", "throughput": 1907.02, "total_tokens": 27545032}
{"current_steps": 4160, "total_steps": 16324, "loss": 0.3533, "lr": 8.896448617878053e-05, "epoch": 0.5096321705307647, "percentage": 25.48, "elapsed_time": "4:01:43", "remaining_time": "11:46:49", "throughput": 1901.48, "total_tokens": 27578320}
{"current_steps": 4165, "total_steps": 16324, "loss": 0.2897, "lr": 8.893272694030562e-05, "epoch": 0.510244709197268, "percentage": 25.51, "elapsed_time": "4:02:44", "remaining_time": "11:48:37", "throughput": 1895.92, "total_tokens": 27612376}
{"current_steps": 4170, "total_steps": 16324, "loss": 0.3093, "lr": 8.890092775346961e-05, "epoch": 0.5108572478637714, "percentage": 25.55, "elapsed_time": "4:03:43", "remaining_time": "11:50:21", "throughput": 1890.58, "total_tokens": 27646232}
{"current_steps": 4175, "total_steps": 16324, "loss": 0.3161, "lr": 8.886908865090121e-05, "epoch": 0.5114697865302748, "percentage": 25.58, "elapsed_time": "4:04:42", "remaining_time": "11:52:05", "throughput": 1885.2, "total_tokens": 27679704}
{"current_steps": 4180, "total_steps": 16324, "loss": 0.3491, "lr": 8.883720966527012e-05, "epoch": 0.5120823251967781, "percentage": 25.61, "elapsed_time": "4:05:43", "remaining_time": "11:53:54", "throughput": 1879.7, "total_tokens": 27713616}
{"current_steps": 4185, "total_steps": 16324, "loss": 0.348, "lr": 8.880529082928696e-05, "epoch": 0.5126948638632813, "percentage": 25.64, "elapsed_time": "4:06:42", "remaining_time": "11:55:37", "throughput": 1874.45, "total_tokens": 27747272}
{"current_steps": 4190, "total_steps": 16324, "loss": 0.417, "lr": 8.877333217570323e-05, "epoch": 0.5133074025297847, "percentage": 25.67, "elapsed_time": "4:07:42", "remaining_time": "11:57:20", "throughput": 1869.2, "total_tokens": 27780888}
{"current_steps": 4195, "total_steps": 16324, "loss": 0.3867, "lr": 8.874133373731123e-05, "epoch": 0.513919941196288, "percentage": 25.7, "elapsed_time": "4:08:43", "remaining_time": "11:59:07", "throughput": 1863.83, "total_tokens": 27814576}
{"current_steps": 4200, "total_steps": 16324, "loss": 0.3802, "lr": 8.870929554694423e-05, "epoch": 0.5145324798627914, "percentage": 25.73, "elapsed_time": "4:09:43", "remaining_time": "12:00:51", "throughput": 1858.62, "total_tokens": 27847792}
{"current_steps": 4200, "total_steps": 16324, "eval_loss": 0.354165256023407, "epoch": 0.5145324798627914, "percentage": 25.73, "elapsed_time": "4:10:03", "remaining_time": "12:01:48", "throughput": 1856.14, "total_tokens": 27847792}
{"current_steps": 4205, "total_steps": 16324, "loss": 0.428, "lr": 8.867721763747613e-05, "epoch": 0.5151450185292946, "percentage": 25.76, "elapsed_time": "4:11:08", "remaining_time": "12:03:46", "throughput": 1850.34, "total_tokens": 27880936}
{"current_steps": 4210, "total_steps": 16324, "loss": 0.3322, "lr": 8.864510004182167e-05, "epoch": 0.515757557195798, "percentage": 25.79, "elapsed_time": "4:12:08", "remaining_time": "12:05:32", "throughput": 1845.14, "total_tokens": 27914960}
{"current_steps": 4215, "total_steps": 16324, "loss": 0.3358, "lr": 8.86129427929363e-05, "epoch": 0.5163700958623013, "percentage": 25.82, "elapsed_time": "4:13:07", "remaining_time": "12:07:12", "throughput": 1840.19, "total_tokens": 27948640}
{"current_steps": 4220, "total_steps": 16324, "loss": 0.3476, "lr": 8.858074592381616e-05, "epoch": 0.5169826345288047, "percentage": 25.85, "elapsed_time": "4:14:06", "remaining_time": "12:08:50", "throughput": 1835.35, "total_tokens": 27982736}
{"current_steps": 4225, "total_steps": 16324, "loss": 0.3419, "lr": 8.854850946749801e-05, "epoch": 0.5175951731953079, "percentage": 25.88, "elapsed_time": "4:15:06", "remaining_time": "12:10:32", "throughput": 1830.36, "total_tokens": 28016472}
{"current_steps": 4230, "total_steps": 16324, "loss": 0.2932, "lr": 8.851623345705927e-05, "epoch": 0.5182077118618112, "percentage": 25.91, "elapsed_time": "4:16:05", "remaining_time": "12:12:10", "throughput": 1825.58, "total_tokens": 28050552}
{"current_steps": 4235, "total_steps": 16324, "loss": 0.3589, "lr": 8.848391792561793e-05, "epoch": 0.5188202505283146, "percentage": 25.94, "elapsed_time": "4:17:04", "remaining_time": "12:13:48", "throughput": 1820.81, "total_tokens": 28084200}
{"current_steps": 4240, "total_steps": 16324, "loss": 0.3527, "lr": 8.845156290633255e-05, "epoch": 0.519432789194818, "percentage": 25.97, "elapsed_time": "4:18:03", "remaining_time": "12:15:27", "throughput": 1815.99, "total_tokens": 28117840}
{"current_steps": 4245, "total_steps": 16324, "loss": 0.4401, "lr": 8.841916843240216e-05, "epoch": 0.5200453278613213, "percentage": 26.0, "elapsed_time": "4:19:02", "remaining_time": "12:17:05", "throughput": 1811.25, "total_tokens": 28151088}
{"current_steps": 4250, "total_steps": 16324, "loss": 0.3934, "lr": 8.838673453706632e-05, "epoch": 0.5206578665278245, "percentage": 26.04, "elapsed_time": "4:20:01", "remaining_time": "12:18:42", "throughput": 1806.55, "total_tokens": 28184288}
{"current_steps": 4250, "total_steps": 16324, "eval_loss": 0.27066436409950256, "epoch": 0.5206578665278245, "percentage": 26.04, "elapsed_time": "4:20:20", "remaining_time": "12:19:36", "throughput": 1804.33, "total_tokens": 28184288}
{"current_steps": 4255, "total_steps": 16324, "loss": 0.3608, "lr": 8.835426125360504e-05, "epoch": 0.5212704051943279, "percentage": 26.07, "elapsed_time": "4:21:25", "remaining_time": "12:21:30", "throughput": 1799.04, "total_tokens": 28218704}
{"current_steps": 4260, "total_steps": 16324, "loss": 0.3643, "lr": 8.83217486153387e-05, "epoch": 0.5218829438608312, "percentage": 26.1, "elapsed_time": "4:22:25", "remaining_time": "12:23:08", "throughput": 1794.32, "total_tokens": 28251552}
{"current_steps": 4265, "total_steps": 16324, "loss": 0.4165, "lr": 8.828919665562814e-05, "epoch": 0.5224954825273346, "percentage": 26.13, "elapsed_time": "4:23:23", "remaining_time": "12:24:43", "throughput": 1789.77, "total_tokens": 28284760}
{"current_steps": 4270, "total_steps": 16324, "loss": 0.3577, "lr": 8.825660540787444e-05, "epoch": 0.5231080211938378, "percentage": 26.16, "elapsed_time": "4:24:23", "remaining_time": "12:26:20", "throughput": 1785.15, "total_tokens": 28318008}
{"current_steps": 4275, "total_steps": 16324, "loss": 0.3533, "lr": 8.82239749055191e-05, "epoch": 0.5237205598603412, "percentage": 26.19, "elapsed_time": "4:25:22", "remaining_time": "12:27:58", "throughput": 1780.57, "total_tokens": 28351584}
{"current_steps": 4280, "total_steps": 16324, "loss": 0.3971, "lr": 8.819130518204383e-05, "epoch": 0.5243330985268445, "percentage": 26.22, "elapsed_time": "4:26:21", "remaining_time": "12:29:31", "throughput": 1776.18, "total_tokens": 28385808}
{"current_steps": 4285, "total_steps": 16324, "loss": 0.3658, "lr": 8.815859627097059e-05, "epoch": 0.5249456371933479, "percentage": 26.25, "elapsed_time": "4:27:21", "remaining_time": "12:31:08", "throughput": 1771.7, "total_tokens": 28419816}
{"current_steps": 4290, "total_steps": 16324, "loss": 0.3209, "lr": 8.81258482058616e-05, "epoch": 0.5255581758598512, "percentage": 26.28, "elapsed_time": "4:28:21", "remaining_time": "12:32:45", "throughput": 1767.19, "total_tokens": 28453888}
{"current_steps": 4295, "total_steps": 16324, "loss": 0.3307, "lr": 8.80930610203192e-05, "epoch": 0.5261707145263544, "percentage": 26.31, "elapsed_time": "4:29:19", "remaining_time": "12:34:18", "throughput": 1762.85, "total_tokens": 28487400}
{"current_steps": 4300, "total_steps": 16324, "loss": 0.3213, "lr": 8.806023474798591e-05, "epoch": 0.5267832531928578, "percentage": 26.34, "elapsed_time": "4:30:19", "remaining_time": "12:35:53", "throughput": 1758.45, "total_tokens": 28520768}
{"current_steps": 4300, "total_steps": 16324, "eval_loss": 0.1924230009317398, "epoch": 0.5267832531928578, "percentage": 26.34, "elapsed_time": "4:30:38", "remaining_time": "12:36:47", "throughput": 1756.35, "total_tokens": 28520768}
{"current_steps": 4305, "total_steps": 16324, "loss": 0.3658, "lr": 8.80273694225443e-05, "epoch": 0.5273957918593611, "percentage": 26.37, "elapsed_time": "4:31:43", "remaining_time": "12:38:37", "throughput": 1751.38, "total_tokens": 28554056}
{"current_steps": 4310, "total_steps": 16324, "loss": 0.3763, "lr": 8.799446507771709e-05, "epoch": 0.5280083305258645, "percentage": 26.4, "elapsed_time": "4:32:43", "remaining_time": "12:40:11", "throughput": 1747.04, "total_tokens": 28587184}
{"current_steps": 4315, "total_steps": 16324, "loss": 0.3491, "lr": 8.796152174726697e-05, "epoch": 0.5286208691923677, "percentage": 26.43, "elapsed_time": "4:33:42", "remaining_time": "12:41:44", "throughput": 1742.82, "total_tokens": 28621016}
{"current_steps": 4320, "total_steps": 16324, "loss": 0.2954, "lr": 8.792853946499666e-05, "epoch": 0.5292334078588711, "percentage": 26.46, "elapsed_time": "4:34:43", "remaining_time": "12:43:22", "throughput": 1738.42, "total_tokens": 28655368}
{"current_steps": 4325, "total_steps": 16324, "loss": 0.3704, "lr": 8.789551826474886e-05, "epoch": 0.5298459465253744, "percentage": 26.49, "elapsed_time": "4:35:42", "remaining_time": "12:44:55", "throughput": 1734.22, "total_tokens": 28688848}
{"current_steps": 4330, "total_steps": 16324, "loss": 0.2967, "lr": 8.786245818040622e-05, "epoch": 0.5304584851918778, "percentage": 26.53, "elapsed_time": "4:36:41", "remaining_time": "12:46:26", "throughput": 1730.16, "total_tokens": 28723488}
{"current_steps": 4335, "total_steps": 16324, "loss": 0.3078, "lr": 8.78293592458912e-05, "epoch": 0.531071023858381, "percentage": 26.56, "elapsed_time": "4:37:41", "remaining_time": "12:48:00", "throughput": 1725.91, "total_tokens": 28756968}
{"current_steps": 4340, "total_steps": 16324, "loss": 0.3285, "lr": 8.779622149516621e-05, "epoch": 0.5316835625248844, "percentage": 26.59, "elapsed_time": "4:38:40", "remaining_time": "12:49:30", "throughput": 1721.85, "total_tokens": 28790376}
{"current_steps": 4345, "total_steps": 16324, "loss": 0.4012, "lr": 8.776304496223349e-05, "epoch": 0.5322961011913877, "percentage": 26.62, "elapsed_time": "4:39:39", "remaining_time": "12:51:00", "throughput": 1717.78, "total_tokens": 28823400}
{"current_steps": 4350, "total_steps": 16324, "loss": 0.33, "lr": 8.772982968113499e-05, "epoch": 0.532908639857891, "percentage": 26.65, "elapsed_time": "4:40:39", "remaining_time": "12:52:32", "throughput": 1713.68, "total_tokens": 28857320}
{"current_steps": 4350, "total_steps": 16324, "eval_loss": 0.34251755475997925, "epoch": 0.532908639857891, "percentage": 26.65, "elapsed_time": "4:40:58", "remaining_time": "12:53:25", "throughput": 1711.72, "total_tokens": 28857320}
{"current_steps": 4355, "total_steps": 16324, "loss": 0.3825, "lr": 8.769657568595252e-05, "epoch": 0.5335211785243944, "percentage": 26.68, "elapsed_time": "4:42:02", "remaining_time": "12:55:08", "throughput": 1707.28, "total_tokens": 28891544}
{"current_steps": 4360, "total_steps": 16324, "loss": 0.3988, "lr": 8.766328301080756e-05, "epoch": 0.5341337171908976, "percentage": 26.71, "elapsed_time": "4:43:01", "remaining_time": "12:56:37", "throughput": 1703.36, "total_tokens": 28925232}
{"current_steps": 4365, "total_steps": 16324, "loss": 0.328, "lr": 8.762995168986126e-05, "epoch": 0.534746255857401, "percentage": 26.74, "elapsed_time": "4:44:01", "remaining_time": "12:58:08", "throughput": 1699.35, "total_tokens": 28958664}
{"current_steps": 4370, "total_steps": 16324, "loss": 0.3096, "lr": 8.759658175731449e-05, "epoch": 0.5353587945239043, "percentage": 26.77, "elapsed_time": "4:44:59", "remaining_time": "12:59:36", "throughput": 1695.47, "total_tokens": 28992488}
{"current_steps": 4375, "total_steps": 16324, "loss": 0.2844, "lr": 8.756317324740766e-05, "epoch": 0.5359713331904077, "percentage": 26.8, "elapsed_time": "4:45:58", "remaining_time": "13:01:03", "throughput": 1691.65, "total_tokens": 29026336}
{"current_steps": 4380, "total_steps": 16324, "loss": 0.3573, "lr": 8.752972619442086e-05, "epoch": 0.5365838718569109, "percentage": 26.83, "elapsed_time": "4:46:58", "remaining_time": "13:02:33", "throughput": 1687.71, "total_tokens": 29059736}
{"current_steps": 4385, "total_steps": 16324, "loss": 0.3899, "lr": 8.749624063267362e-05, "epoch": 0.5371964105234143, "percentage": 26.86, "elapsed_time": "4:47:57", "remaining_time": "13:04:01", "throughput": 1683.86, "total_tokens": 29092848}
{"current_steps": 4390, "total_steps": 16324, "loss": 0.3047, "lr": 8.746271659652506e-05, "epoch": 0.5378089491899176, "percentage": 26.89, "elapsed_time": "4:48:56", "remaining_time": "13:05:28", "throughput": 1680.05, "total_tokens": 29126336}
{"current_steps": 4395, "total_steps": 16324, "loss": 0.3134, "lr": 8.742915412037376e-05, "epoch": 0.538421487856421, "percentage": 26.92, "elapsed_time": "4:49:56", "remaining_time": "13:06:57", "throughput": 1676.21, "total_tokens": 29160048}
{"current_steps": 4400, "total_steps": 16324, "loss": 0.4625, "lr": 8.739555323865771e-05, "epoch": 0.5390340265229243, "percentage": 26.95, "elapsed_time": "4:50:55", "remaining_time": "13:08:25", "throughput": 1672.41, "total_tokens": 29193456}
{"current_steps": 4400, "total_steps": 16324, "eval_loss": 0.2848778963088989, "epoch": 0.5390340265229243, "percentage": 26.95, "elapsed_time": "4:51:15", "remaining_time": "13:09:18", "throughput": 1670.54, "total_tokens": 29193456}
{"current_steps": 4405, "total_steps": 16324, "loss": 0.3406, "lr": 8.736191398585436e-05, "epoch": 0.5396465651894276, "percentage": 26.98, "elapsed_time": "4:52:20", "remaining_time": "13:11:00", "throughput": 1666.26, "total_tokens": 29226768}
{"current_steps": 4410, "total_steps": 16324, "loss": 0.4326, "lr": 8.732823639648052e-05, "epoch": 0.5402591038559309, "percentage": 27.02, "elapsed_time": "4:53:20", "remaining_time": "13:12:28", "throughput": 1662.51, "total_tokens": 29260664}
{"current_steps": 4415, "total_steps": 16324, "loss": 0.3559, "lr": 8.729452050509228e-05, "epoch": 0.5408716425224342, "percentage": 27.05, "elapsed_time": "4:54:20", "remaining_time": "13:13:57", "throughput": 1658.71, "total_tokens": 29293776}
{"current_steps": 4420, "total_steps": 16324, "loss": 0.3238, "lr": 8.726076634628511e-05, "epoch": 0.5414841811889376, "percentage": 27.08, "elapsed_time": "4:55:19", "remaining_time": "13:15:23", "throughput": 1655.04, "total_tokens": 29327128}
{"current_steps": 4425, "total_steps": 16324, "loss": 0.4011, "lr": 8.722697395469368e-05, "epoch": 0.5420967198554408, "percentage": 27.11, "elapsed_time": "4:56:19", "remaining_time": "13:16:49", "throughput": 1651.38, "total_tokens": 29360832}
{"current_steps": 4430, "total_steps": 16324, "loss": 0.3679, "lr": 8.719314336499196e-05, "epoch": 0.5427092585219442, "percentage": 27.14, "elapsed_time": "4:57:20", "remaining_time": "13:18:19", "throughput": 1647.62, "total_tokens": 29394440}
{"current_steps": 4435, "total_steps": 16324, "loss": 0.3443, "lr": 8.715927461189304e-05, "epoch": 0.5433217971884475, "percentage": 27.17, "elapsed_time": "4:58:20", "remaining_time": "13:19:45", "throughput": 1644.02, "total_tokens": 29428240}
{"current_steps": 4440, "total_steps": 16324, "loss": 0.3434, "lr": 8.71253677301492e-05, "epoch": 0.5439343358549509, "percentage": 27.2, "elapsed_time": "4:59:20", "remaining_time": "13:21:11", "throughput": 1640.39, "total_tokens": 29461536}
{"current_steps": 4445, "total_steps": 16324, "loss": 0.4073, "lr": 8.709142275455189e-05, "epoch": 0.5445468745214541, "percentage": 27.23, "elapsed_time": "5:00:20", "remaining_time": "13:22:37", "throughput": 1636.79, "total_tokens": 29495024}
{"current_steps": 4450, "total_steps": 16324, "loss": 0.2876, "lr": 8.705743971993157e-05, "epoch": 0.5451594131879575, "percentage": 27.26, "elapsed_time": "5:01:18", "remaining_time": "13:24:00", "throughput": 1633.33, "total_tokens": 29528840}
{"current_steps": 4450, "total_steps": 16324, "eval_loss": 0.2150648534297943, "epoch": 0.5451594131879575, "percentage": 27.26, "elapsed_time": "5:01:37", "remaining_time": "13:24:51", "throughput": 1631.61, "total_tokens": 29528840}
{"current_steps": 4455, "total_steps": 16324, "loss": 0.3817, "lr": 8.70234186611578e-05, "epoch": 0.5457719518544608, "percentage": 27.29, "elapsed_time": "5:02:41", "remaining_time": "13:26:26", "throughput": 1627.71, "total_tokens": 29561992}
{"current_steps": 4460, "total_steps": 16324, "loss": 0.3339, "lr": 8.698935961313913e-05, "epoch": 0.5463844905209642, "percentage": 27.32, "elapsed_time": "5:03:41", "remaining_time": "13:27:49", "throughput": 1624.22, "total_tokens": 29595288}
{"current_steps": 4465, "total_steps": 16324, "loss": 0.3188, "lr": 8.695526261082311e-05, "epoch": 0.5469970291874675, "percentage": 27.35, "elapsed_time": "5:04:39", "remaining_time": "13:29:10", "throughput": 1620.86, "total_tokens": 29629048}
{"current_steps": 4470, "total_steps": 16324, "loss": 0.3273, "lr": 8.692112768919624e-05, "epoch": 0.5476095678539707, "percentage": 27.38, "elapsed_time": "5:05:38", "remaining_time": "13:30:30", "throughput": 1617.52, "total_tokens": 29662200}
{"current_steps": 4475, "total_steps": 16324, "loss": 0.3223, "lr": 8.68869548832839e-05, "epoch": 0.5482221065204741, "percentage": 27.41, "elapsed_time": "5:06:37", "remaining_time": "13:31:54", "throughput": 1614.11, "total_tokens": 29696128}
{"current_steps": 4480, "total_steps": 16324, "loss": 0.4173, "lr": 8.685274422815035e-05, "epoch": 0.5488346451869774, "percentage": 27.44, "elapsed_time": "5:07:35", "remaining_time": "13:33:12", "throughput": 1610.86, "total_tokens": 29729768}
{"current_steps": 4485, "total_steps": 16324, "loss": 0.3312, "lr": 8.681849575889873e-05, "epoch": 0.5494471838534808, "percentage": 27.47, "elapsed_time": "5:08:34", "remaining_time": "13:34:32", "throughput": 1607.6, "total_tokens": 29763736}
{"current_steps": 4490, "total_steps": 16324, "loss": 0.3364, "lr": 8.678420951067091e-05, "epoch": 0.550059722519984, "percentage": 27.51, "elapsed_time": "5:09:34", "remaining_time": "13:35:54", "throughput": 1604.26, "total_tokens": 29797848}
{"current_steps": 4495, "total_steps": 16324, "loss": 0.4096, "lr": 8.674988551864758e-05, "epoch": 0.5506722611864874, "percentage": 27.54, "elapsed_time": "5:10:32", "remaining_time": "13:37:14", "throughput": 1601.05, "total_tokens": 29832160}
{"current_steps": 4500, "total_steps": 16324, "loss": 0.4329, "lr": 8.671552381804813e-05, "epoch": 0.5512847998529907, "percentage": 27.57, "elapsed_time": "5:11:31", "remaining_time": "13:38:33", "throughput": 1597.83, "total_tokens": 29865992}
{"current_steps": 4500, "total_steps": 16324, "eval_loss": 0.19009599089622498, "epoch": 0.5512847998529907, "percentage": 27.57, "elapsed_time": "5:11:50", "remaining_time": "13:39:23", "throughput": 1596.19, "total_tokens": 29865992}
{"current_steps": 4505, "total_steps": 16324, "loss": 0.2654, "lr": 8.668112444413065e-05, "epoch": 0.5518973385194941, "percentage": 27.6, "elapsed_time": "5:12:55", "remaining_time": "13:40:59", "throughput": 1592.48, "total_tokens": 29900248}
{"current_steps": 4510, "total_steps": 16324, "loss": 0.2877, "lr": 8.664668743219188e-05, "epoch": 0.5525098771859973, "percentage": 27.63, "elapsed_time": "5:13:54", "remaining_time": "13:42:17", "throughput": 1589.34, "total_tokens": 29934856}
{"current_steps": 4515, "total_steps": 16324, "loss": 0.3991, "lr": 8.661221281756722e-05, "epoch": 0.5531224158525007, "percentage": 27.66, "elapsed_time": "5:14:53", "remaining_time": "13:43:35", "throughput": 1586.15, "total_tokens": 29967480}
{"current_steps": 4520, "total_steps": 16324, "loss": 0.3031, "lr": 8.657770063563058e-05, "epoch": 0.553734954519004, "percentage": 27.69, "elapsed_time": "5:15:52", "remaining_time": "13:44:54", "throughput": 1582.95, "total_tokens": 30001000}
{"current_steps": 4525, "total_steps": 16324, "loss": 0.348, "lr": 8.654315092179449e-05, "epoch": 0.5543474931855074, "percentage": 27.72, "elapsed_time": "5:16:51", "remaining_time": "13:46:13", "throughput": 1579.82, "total_tokens": 30035464}
{"current_steps": 4530, "total_steps": 16324, "loss": 0.3051, "lr": 8.650856371150993e-05, "epoch": 0.5549600318520107, "percentage": 27.75, "elapsed_time": "5:17:50", "remaining_time": "13:47:30", "throughput": 1576.72, "total_tokens": 30069096}
{"current_steps": 4535, "total_steps": 16324, "loss": 0.4479, "lr": 8.64739390402664e-05, "epoch": 0.5555725705185139, "percentage": 27.78, "elapsed_time": "5:18:50", "remaining_time": "13:48:50", "throughput": 1573.55, "total_tokens": 30102376}
{"current_steps": 4540, "total_steps": 16324, "loss": 0.3024, "lr": 8.643927694359182e-05, "epoch": 0.5561851091850173, "percentage": 27.81, "elapsed_time": "5:19:50", "remaining_time": "13:50:09", "throughput": 1570.34, "total_tokens": 30135120}
{"current_steps": 4545, "total_steps": 16324, "loss": 0.3314, "lr": 8.640457745705252e-05, "epoch": 0.5567976478515206, "percentage": 27.84, "elapsed_time": "5:20:49", "remaining_time": "13:51:27", "throughput": 1567.3, "total_tokens": 30169416}
{"current_steps": 4550, "total_steps": 16324, "loss": 0.2804, "lr": 8.636984061625317e-05, "epoch": 0.557410186518024, "percentage": 27.87, "elapsed_time": "5:21:48", "remaining_time": "13:52:44", "throughput": 1564.25, "total_tokens": 30203424}
{"current_steps": 4550, "total_steps": 16324, "eval_loss": 0.2320593297481537, "epoch": 0.557410186518024, "percentage": 27.87, "elapsed_time": "5:22:07", "remaining_time": "13:53:34", "throughput": 1562.68, "total_tokens": 30203424}
{"current_steps": 4555, "total_steps": 16324, "loss": 0.2574, "lr": 8.63350664568368e-05, "epoch": 0.5580227251845272, "percentage": 27.9, "elapsed_time": "5:23:13", "remaining_time": "13:55:08", "throughput": 1559.18, "total_tokens": 30238032}
{"current_steps": 4560, "total_steps": 16324, "loss": 0.4057, "lr": 8.630025501448473e-05, "epoch": 0.5586352638510306, "percentage": 27.93, "elapsed_time": "5:24:12", "remaining_time": "13:56:22", "throughput": 1556.24, "total_tokens": 30272112}
{"current_steps": 4565, "total_steps": 16324, "loss": 0.3657, "lr": 8.62654063249165e-05, "epoch": 0.5592478025175339, "percentage": 27.96, "elapsed_time": "5:25:11", "remaining_time": "13:57:38", "throughput": 1553.26, "total_tokens": 30305896}
{"current_steps": 4570, "total_steps": 16324, "loss": 0.3327, "lr": 8.62305204238899e-05, "epoch": 0.5598603411840373, "percentage": 28.0, "elapsed_time": "5:26:10", "remaining_time": "13:58:56", "throughput": 1550.22, "total_tokens": 30339184}
{"current_steps": 4575, "total_steps": 16324, "loss": 0.3472, "lr": 8.619559734720092e-05, "epoch": 0.5604728798505406, "percentage": 28.03, "elapsed_time": "5:27:09", "remaining_time": "14:00:10", "throughput": 1547.27, "total_tokens": 30372520}
{"current_steps": 4580, "total_steps": 16324, "loss": 0.3242, "lr": 8.616063713068365e-05, "epoch": 0.5610854185170439, "percentage": 28.06, "elapsed_time": "5:28:07", "remaining_time": "14:01:23", "throughput": 1544.41, "total_tokens": 30406224}
{"current_steps": 4585, "total_steps": 16324, "loss": 0.3828, "lr": 8.612563981021031e-05, "epoch": 0.5616979571835472, "percentage": 28.09, "elapsed_time": "5:29:07", "remaining_time": "14:02:40", "throughput": 1541.43, "total_tokens": 30439864}
{"current_steps": 4590, "total_steps": 16324, "loss": 0.3547, "lr": 8.60906054216912e-05, "epoch": 0.5623104958500506, "percentage": 28.12, "elapsed_time": "5:30:06", "remaining_time": "14:03:52", "throughput": 1538.6, "total_tokens": 30473688}
{"current_steps": 4595, "total_steps": 16324, "loss": 0.3963, "lr": 8.605553400107463e-05, "epoch": 0.5629230345165539, "percentage": 28.15, "elapsed_time": "5:31:04", "remaining_time": "14:05:05", "throughput": 1535.73, "total_tokens": 30506632}
{"current_steps": 4600, "total_steps": 16324, "loss": 0.3363, "lr": 8.602042558434696e-05, "epoch": 0.5635355731830571, "percentage": 28.18, "elapsed_time": "5:32:04", "remaining_time": "14:06:20", "throughput": 1532.8, "total_tokens": 30539672}
{"current_steps": 4600, "total_steps": 16324, "eval_loss": 0.3674183487892151, "epoch": 0.5635355731830571, "percentage": 28.18, "elapsed_time": "5:32:23", "remaining_time": "14:07:09", "throughput": 1531.34, "total_tokens": 30539672}
{"current_steps": 4605, "total_steps": 16324, "loss": 0.3515, "lr": 8.598528020753244e-05, "epoch": 0.5641481118495605, "percentage": 28.21, "elapsed_time": "5:33:27", "remaining_time": "14:08:35", "throughput": 1528.15, "total_tokens": 30573928}
{"current_steps": 4610, "total_steps": 16324, "loss": 0.4061, "lr": 8.595009790669331e-05, "epoch": 0.5647606505160638, "percentage": 28.24, "elapsed_time": "5:34:25", "remaining_time": "14:09:45", "throughput": 1525.39, "total_tokens": 30607080}
{"current_steps": 4615, "total_steps": 16324, "loss": 0.267, "lr": 8.591487871792963e-05, "epoch": 0.5653731891825672, "percentage": 28.27, "elapsed_time": "5:35:25", "remaining_time": "14:11:00", "throughput": 1522.52, "total_tokens": 30640832}
{"current_steps": 4620, "total_steps": 16324, "loss": 0.3081, "lr": 8.58796226773794e-05, "epoch": 0.5659857278490704, "percentage": 28.3, "elapsed_time": "5:36:23", "remaining_time": "14:12:11", "throughput": 1519.81, "total_tokens": 30675144}
{"current_steps": 4625, "total_steps": 16324, "loss": 0.3699, "lr": 8.584432982121832e-05, "epoch": 0.5665982665155738, "percentage": 28.33, "elapsed_time": "5:37:21", "remaining_time": "14:13:21", "throughput": 1517.11, "total_tokens": 30708688}
{"current_steps": 4630, "total_steps": 16324, "loss": 0.3889, "lr": 8.580900018565998e-05, "epoch": 0.5672108051820771, "percentage": 28.36, "elapsed_time": "5:38:21", "remaining_time": "14:14:34", "throughput": 1514.29, "total_tokens": 30741960}
{"current_steps": 4635, "total_steps": 16324, "loss": 0.3255, "lr": 8.577363380695564e-05, "epoch": 0.5678233438485805, "percentage": 28.39, "elapsed_time": "5:39:20", "remaining_time": "14:15:45", "throughput": 1511.57, "total_tokens": 30775624}
{"current_steps": 4640, "total_steps": 16324, "loss": 0.3598, "lr": 8.573823072139429e-05, "epoch": 0.5684358825150838, "percentage": 28.42, "elapsed_time": "5:40:18", "remaining_time": "14:16:56", "throughput": 1508.83, "total_tokens": 30808448}
{"current_steps": 4645, "total_steps": 16324, "loss": 0.3538, "lr": 8.570279096530253e-05, "epoch": 0.569048421181587, "percentage": 28.46, "elapsed_time": "5:41:18", "remaining_time": "14:18:08", "throughput": 1506.09, "total_tokens": 30842072}
{"current_steps": 4650, "total_steps": 16324, "loss": 0.2994, "lr": 8.566731457504466e-05, "epoch": 0.5696609598480904, "percentage": 28.49, "elapsed_time": "5:42:18", "remaining_time": "14:19:21", "throughput": 1503.35, "total_tokens": 30875840}
{"current_steps": 4650, "total_steps": 16324, "eval_loss": 0.27580884099006653, "epoch": 0.5696609598480904, "percentage": 28.49, "elapsed_time": "5:42:37", "remaining_time": "14:20:10", "throughput": 1501.92, "total_tokens": 30875840}
{"current_steps": 4655, "total_steps": 16324, "loss": 0.3418, "lr": 8.563180158702256e-05, "epoch": 0.5702734985145937, "percentage": 28.52, "elapsed_time": "5:43:42", "remaining_time": "14:21:36", "throughput": 1498.77, "total_tokens": 30908664}
{"current_steps": 4660, "total_steps": 16324, "loss": 0.3987, "lr": 8.559625203767559e-05, "epoch": 0.5708860371810971, "percentage": 28.55, "elapsed_time": "5:44:42", "remaining_time": "14:22:48", "throughput": 1496.06, "total_tokens": 30942328}
{"current_steps": 4665, "total_steps": 16324, "loss": 0.3358, "lr": 8.556066596348072e-05, "epoch": 0.5714985758476003, "percentage": 28.58, "elapsed_time": "5:45:42", "remaining_time": "14:24:00", "throughput": 1493.38, "total_tokens": 30976456}
{"current_steps": 4670, "total_steps": 16324, "loss": 0.3643, "lr": 8.552504340095233e-05, "epoch": 0.5721111145141037, "percentage": 28.61, "elapsed_time": "5:46:40", "remaining_time": "14:25:08", "throughput": 1490.76, "total_tokens": 31009328}
{"current_steps": 4675, "total_steps": 16324, "loss": 0.3039, "lr": 8.548938438664229e-05, "epoch": 0.572723653180607, "percentage": 28.64, "elapsed_time": "5:47:40", "remaining_time": "14:26:19", "throughput": 1488.12, "total_tokens": 31043200}
{"current_steps": 4680, "total_steps": 16324, "loss": 0.3185, "lr": 8.545368895713982e-05, "epoch": 0.5733361918471104, "percentage": 28.67, "elapsed_time": "5:48:40", "remaining_time": "14:27:31", "throughput": 1485.49, "total_tokens": 31077480}
{"current_steps": 4685, "total_steps": 16324, "loss": 0.4018, "lr": 8.541795714907155e-05, "epoch": 0.5739487305136137, "percentage": 28.7, "elapsed_time": "5:49:39", "remaining_time": "14:28:38", "throughput": 1482.9, "total_tokens": 31110000}
{"current_steps": 4690, "total_steps": 16324, "loss": 0.3636, "lr": 8.53821889991014e-05, "epoch": 0.574561269180117, "percentage": 28.73, "elapsed_time": "5:50:38", "remaining_time": "14:29:47", "throughput": 1480.35, "total_tokens": 31143728}
{"current_steps": 4695, "total_steps": 16324, "loss": 0.351, "lr": 8.534638454393063e-05, "epoch": 0.5751738078466203, "percentage": 28.76, "elapsed_time": "5:51:37", "remaining_time": "14:30:57", "throughput": 1477.77, "total_tokens": 31177720}
{"current_steps": 4700, "total_steps": 16324, "loss": 0.3163, "lr": 8.53105438202977e-05, "epoch": 0.5757863465131237, "percentage": 28.79, "elapsed_time": "5:52:36", "remaining_time": "14:32:03", "throughput": 1475.26, "total_tokens": 31211360}
{"current_steps": 4700, "total_steps": 16324, "eval_loss": 0.18082624673843384, "epoch": 0.5757863465131237, "percentage": 28.79, "elapsed_time": "5:52:55", "remaining_time": "14:32:51", "throughput": 1473.91, "total_tokens": 31211360}
{"current_steps": 4705, "total_steps": 16324, "loss": 0.3304, "lr": 8.527466686497831e-05, "epoch": 0.576398885179627, "percentage": 28.82, "elapsed_time": "5:54:00", "remaining_time": "14:34:13", "throughput": 1471.0, "total_tokens": 31244648}
{"current_steps": 4710, "total_steps": 16324, "loss": 0.3744, "lr": 8.523875371478534e-05, "epoch": 0.5770114238461302, "percentage": 28.85, "elapsed_time": "5:55:00", "remaining_time": "14:35:22", "throughput": 1468.47, "total_tokens": 31278528}
{"current_steps": 4715, "total_steps": 16324, "loss": 0.4074, "lr": 8.520280440656882e-05, "epoch": 0.5776239625126336, "percentage": 28.88, "elapsed_time": "5:55:58", "remaining_time": "14:36:27", "throughput": 1466.02, "total_tokens": 31312144}
{"current_steps": 4720, "total_steps": 16324, "loss": 0.4165, "lr": 8.516681897721583e-05, "epoch": 0.5782365011791369, "percentage": 28.91, "elapsed_time": "5:56:57", "remaining_time": "14:37:34", "throughput": 1463.59, "total_tokens": 31346248}
{"current_steps": 4725, "total_steps": 16324, "loss": 0.3424, "lr": 8.513079746365055e-05, "epoch": 0.5788490398456403, "percentage": 28.95, "elapsed_time": "5:57:56", "remaining_time": "14:38:41", "throughput": 1461.1, "total_tokens": 31379752}
{"current_steps": 4730, "total_steps": 16324, "loss": 0.4001, "lr": 8.509473990283421e-05, "epoch": 0.5794615785121435, "percentage": 28.98, "elapsed_time": "5:58:55", "remaining_time": "14:39:46", "throughput": 1458.67, "total_tokens": 31412984}
{"current_steps": 4735, "total_steps": 16324, "loss": 0.31, "lr": 8.505864633176499e-05, "epoch": 0.5800741171786469, "percentage": 29.01, "elapsed_time": "5:59:54", "remaining_time": "14:40:51", "throughput": 1456.27, "total_tokens": 31446768}
{"current_steps": 4740, "total_steps": 16324, "loss": 0.3824, "lr": 8.502251678747802e-05, "epoch": 0.5806866558451502, "percentage": 29.04, "elapsed_time": "6:00:54", "remaining_time": "14:42:01", "throughput": 1453.76, "total_tokens": 31480608}
{"current_steps": 4745, "total_steps": 16324, "loss": 0.3319, "lr": 8.498635130704536e-05, "epoch": 0.5812991945116536, "percentage": 29.07, "elapsed_time": "6:01:53", "remaining_time": "14:43:05", "throughput": 1451.35, "total_tokens": 31513536}
{"current_steps": 4750, "total_steps": 16324, "loss": 0.307, "lr": 8.495014992757595e-05, "epoch": 0.5819117331781569, "percentage": 29.1, "elapsed_time": "6:02:52", "remaining_time": "14:44:11", "throughput": 1448.97, "total_tokens": 31547504}
{"current_steps": 4750, "total_steps": 16324, "eval_loss": 0.22180898487567902, "epoch": 0.5819117331781569, "percentage": 29.1, "elapsed_time": "6:03:11", "remaining_time": "14:44:58", "throughput": 1447.67, "total_tokens": 31547504}
{"current_steps": 4755, "total_steps": 16324, "loss": 0.3643, "lr": 8.491391268621552e-05, "epoch": 0.5825242718446602, "percentage": 29.13, "elapsed_time": "6:04:17", "remaining_time": "14:46:20", "throughput": 1444.84, "total_tokens": 31580736}
{"current_steps": 4760, "total_steps": 16324, "loss": 0.4018, "lr": 8.487763962014665e-05, "epoch": 0.5831368105111635, "percentage": 29.16, "elapsed_time": "6:05:17", "remaining_time": "14:47:26", "throughput": 1442.39, "total_tokens": 31613464}
{"current_steps": 4765, "total_steps": 16324, "loss": 0.3828, "lr": 8.484133076658865e-05, "epoch": 0.5837493491776669, "percentage": 29.19, "elapsed_time": "6:06:17", "remaining_time": "14:48:33", "throughput": 1439.98, "total_tokens": 31646976}
{"current_steps": 4770, "total_steps": 16324, "loss": 0.3694, "lr": 8.480498616279756e-05, "epoch": 0.5843618878441702, "percentage": 29.22, "elapsed_time": "6:07:17", "remaining_time": "14:49:40", "throughput": 1437.52, "total_tokens": 31679736}
{"current_steps": 4775, "total_steps": 16324, "loss": 0.3272, "lr": 8.47686058460661e-05, "epoch": 0.5849744265106734, "percentage": 29.25, "elapsed_time": "6:08:18", "remaining_time": "14:50:47", "throughput": 1435.13, "total_tokens": 31713832}
{"current_steps": 4780, "total_steps": 16324, "loss": 0.4287, "lr": 8.473218985372366e-05, "epoch": 0.5855869651771768, "percentage": 29.28, "elapsed_time": "6:09:17", "remaining_time": "14:51:51", "throughput": 1432.86, "total_tokens": 31748352}
{"current_steps": 4785, "total_steps": 16324, "loss": 0.3326, "lr": 8.469573822313617e-05, "epoch": 0.5861995038436801, "percentage": 29.31, "elapsed_time": "6:10:17", "remaining_time": "14:52:57", "throughput": 1430.51, "total_tokens": 31782192}
{"current_steps": 4790, "total_steps": 16324, "loss": 0.3272, "lr": 8.465925099170621e-05, "epoch": 0.5868120425101835, "percentage": 29.34, "elapsed_time": "6:11:17", "remaining_time": "14:54:02", "throughput": 1428.16, "total_tokens": 31815816}
{"current_steps": 4795, "total_steps": 16324, "loss": 0.3232, "lr": 8.462272819687283e-05, "epoch": 0.5874245811766868, "percentage": 29.37, "elapsed_time": "6:12:17", "remaining_time": "14:55:06", "throughput": 1425.88, "total_tokens": 31849808}
{"current_steps": 4800, "total_steps": 16324, "loss": 0.2896, "lr": 8.458616987611158e-05, "epoch": 0.5880371198431901, "percentage": 29.4, "elapsed_time": "6:13:17", "remaining_time": "14:56:11", "throughput": 1423.57, "total_tokens": 31883816}
{"current_steps": 4800, "total_steps": 16324, "eval_loss": 0.19881321489810944, "epoch": 0.5880371198431901, "percentage": 29.4, "elapsed_time": "6:13:36", "remaining_time": "14:56:58", "throughput": 1422.33, "total_tokens": 31883816}
{"current_steps": 4805, "total_steps": 16324, "loss": 0.254, "lr": 8.45495760669345e-05, "epoch": 0.5886496585096934, "percentage": 29.44, "elapsed_time": "6:14:41", "remaining_time": "14:58:14", "throughput": 1419.74, "total_tokens": 31917488}
{"current_steps": 4810, "total_steps": 16324, "loss": 0.3292, "lr": 8.451294680689002e-05, "epoch": 0.5892621971761968, "percentage": 29.47, "elapsed_time": "6:15:40", "remaining_time": "14:59:15", "throughput": 1417.51, "total_tokens": 31950880}
{"current_steps": 4815, "total_steps": 16324, "loss": 0.3146, "lr": 8.447628213356291e-05, "epoch": 0.5898747358427001, "percentage": 29.5, "elapsed_time": "6:16:38", "remaining_time": "15:00:16", "throughput": 1415.33, "total_tokens": 31984952}
{"current_steps": 4820, "total_steps": 16324, "loss": 0.2873, "lr": 8.443958208457434e-05, "epoch": 0.5904872745092034, "percentage": 29.53, "elapsed_time": "6:17:38", "remaining_time": "15:01:20", "throughput": 1413.06, "total_tokens": 32018408}
{"current_steps": 4825, "total_steps": 16324, "loss": 0.3149, "lr": 8.440284669758177e-05, "epoch": 0.5910998131757067, "percentage": 29.56, "elapsed_time": "6:18:37", "remaining_time": "15:02:20", "throughput": 1410.91, "total_tokens": 32052368}
{"current_steps": 4830, "total_steps": 16324, "loss": 0.3625, "lr": 8.436607601027883e-05, "epoch": 0.59171235184221, "percentage": 29.59, "elapsed_time": "6:19:37", "remaining_time": "15:03:22", "throughput": 1408.69, "total_tokens": 32085792}
{"current_steps": 4835, "total_steps": 16324, "loss": 0.2981, "lr": 8.43292700603955e-05, "epoch": 0.5923248905087134, "percentage": 29.62, "elapsed_time": "6:20:36", "remaining_time": "15:04:25", "throughput": 1406.46, "total_tokens": 32119160}
{"current_steps": 4840, "total_steps": 16324, "loss": 0.358, "lr": 8.429242888569786e-05, "epoch": 0.5929374291752166, "percentage": 29.65, "elapsed_time": "6:21:35", "remaining_time": "15:05:25", "throughput": 1404.29, "total_tokens": 32152400}
{"current_steps": 4845, "total_steps": 16324, "loss": 0.3486, "lr": 8.425555252398816e-05, "epoch": 0.59354996784172, "percentage": 29.68, "elapsed_time": "6:22:34", "remaining_time": "15:06:24", "throughput": 1402.17, "total_tokens": 32185912}
{"current_steps": 4850, "total_steps": 16324, "loss": 0.3107, "lr": 8.421864101310475e-05, "epoch": 0.5941625065082233, "percentage": 29.71, "elapsed_time": "6:23:34", "remaining_time": "15:07:27", "throughput": 1399.98, "total_tokens": 32220000}
{"current_steps": 4850, "total_steps": 16324, "eval_loss": 0.3321171998977661, "epoch": 0.5941625065082233, "percentage": 29.71, "elapsed_time": "6:23:54", "remaining_time": "15:08:13", "throughput": 1398.8, "total_tokens": 32220000}
{"current_steps": 4855, "total_steps": 16324, "loss": 0.3067, "lr": 8.418169439092207e-05, "epoch": 0.5947750451747267, "percentage": 29.74, "elapsed_time": "6:24:58", "remaining_time": "15:09:25", "throughput": 1396.4, "total_tokens": 32254304}
{"current_steps": 4860, "total_steps": 16324, "loss": 0.3232, "lr": 8.414471269535053e-05, "epoch": 0.59538758384123, "percentage": 29.77, "elapsed_time": "6:25:57", "remaining_time": "15:10:24", "throughput": 1394.3, "total_tokens": 32288320}
{"current_steps": 4865, "total_steps": 16324, "loss": 0.335, "lr": 8.410769596433659e-05, "epoch": 0.5960001225077333, "percentage": 29.8, "elapsed_time": "6:26:57", "remaining_time": "15:11:26", "throughput": 1392.12, "total_tokens": 32321632}
{"current_steps": 4870, "total_steps": 16324, "loss": 0.3497, "lr": 8.40706442358626e-05, "epoch": 0.5966126611742366, "percentage": 29.83, "elapsed_time": "6:27:57", "remaining_time": "15:12:26", "throughput": 1389.99, "total_tokens": 32355120}
{"current_steps": 4875, "total_steps": 16324, "loss": 0.3302, "lr": 8.403355754794688e-05, "epoch": 0.59722519984074, "percentage": 29.86, "elapsed_time": "6:28:56", "remaining_time": "15:13:26", "throughput": 1387.9, "total_tokens": 32388696}
{"current_steps": 4880, "total_steps": 16324, "loss": 0.3205, "lr": 8.399643593864361e-05, "epoch": 0.5978377385072433, "percentage": 29.89, "elapsed_time": "6:29:57", "remaining_time": "15:14:28", "throughput": 1385.75, "total_tokens": 32422728}
{"current_steps": 4885, "total_steps": 16324, "loss": 0.3074, "lr": 8.395927944604276e-05, "epoch": 0.5984502771737465, "percentage": 29.93, "elapsed_time": "6:30:56", "remaining_time": "15:15:27", "throughput": 1383.66, "total_tokens": 32456296}
{"current_steps": 4890, "total_steps": 16324, "loss": 0.3231, "lr": 8.392208810827011e-05, "epoch": 0.5990628158402499, "percentage": 29.96, "elapsed_time": "6:31:56", "remaining_time": "15:16:26", "throughput": 1381.56, "total_tokens": 32489280}
{"current_steps": 4895, "total_steps": 16324, "loss": 0.2702, "lr": 8.388486196348724e-05, "epoch": 0.5996753545067532, "percentage": 29.99, "elapsed_time": "6:32:56", "remaining_time": "15:17:27", "throughput": 1379.45, "total_tokens": 32523128}
{"current_steps": 4900, "total_steps": 16324, "loss": 0.3079, "lr": 8.384760104989139e-05, "epoch": 0.6002878931732566, "percentage": 30.02, "elapsed_time": "6:33:57", "remaining_time": "15:18:28", "throughput": 1377.32, "total_tokens": 32556072}
{"current_steps": 4900, "total_steps": 16324, "eval_loss": 0.2282978892326355, "epoch": 0.6002878931732566, "percentage": 30.02, "elapsed_time": "6:34:16", "remaining_time": "15:19:14", "throughput": 1376.17, "total_tokens": 32556072}
{"current_steps": 4905, "total_steps": 16324, "loss": 0.3373, "lr": 8.38103054057155e-05, "epoch": 0.6009004318397598, "percentage": 30.05, "elapsed_time": "6:35:22", "remaining_time": "15:20:26", "throughput": 1373.84, "total_tokens": 32590584}
{"current_steps": 4910, "total_steps": 16324, "loss": 0.3009, "lr": 8.377297506922812e-05, "epoch": 0.6015129705062632, "percentage": 30.08, "elapsed_time": "6:36:22", "remaining_time": "15:21:25", "throughput": 1371.79, "total_tokens": 32624472}
{"current_steps": 4915, "total_steps": 16324, "loss": 0.3028, "lr": 8.373561007873345e-05, "epoch": 0.6021255091727665, "percentage": 30.11, "elapsed_time": "6:37:22", "remaining_time": "15:22:24", "throughput": 1369.76, "total_tokens": 32658640}
{"current_steps": 4920, "total_steps": 16324, "loss": 0.3556, "lr": 8.369821047257121e-05, "epoch": 0.6027380478392699, "percentage": 30.14, "elapsed_time": "6:38:21", "remaining_time": "15:23:21", "throughput": 1367.78, "total_tokens": 32692552}
{"current_steps": 4925, "total_steps": 16324, "loss": 0.324, "lr": 8.366077628911664e-05, "epoch": 0.6033505865057732, "percentage": 30.17, "elapsed_time": "6:39:21", "remaining_time": "15:24:20", "throughput": 1365.75, "total_tokens": 32725792}
{"current_steps": 4930, "total_steps": 16324, "loss": 0.336, "lr": 8.362330756678046e-05, "epoch": 0.6039631251722765, "percentage": 30.2, "elapsed_time": "6:40:21", "remaining_time": "15:25:18", "throughput": 1363.74, "total_tokens": 32759344}
{"current_steps": 4935, "total_steps": 16324, "loss": 0.2681, "lr": 8.358580434400884e-05, "epoch": 0.6045756638387798, "percentage": 30.23, "elapsed_time": "6:41:20", "remaining_time": "15:26:13", "throughput": 1361.8, "total_tokens": 32792880}
{"current_steps": 4940, "total_steps": 16324, "loss": 0.3509, "lr": 8.354826665928335e-05, "epoch": 0.6051882025052832, "percentage": 30.26, "elapsed_time": "6:42:19", "remaining_time": "15:27:08", "throughput": 1359.89, "total_tokens": 32826952}
{"current_steps": 4945, "total_steps": 16324, "loss": 0.3395, "lr": 8.351069455112091e-05, "epoch": 0.6058007411717865, "percentage": 30.29, "elapsed_time": "6:43:19", "remaining_time": "15:28:06", "throughput": 1357.87, "total_tokens": 32860128}
{"current_steps": 4950, "total_steps": 16324, "loss": 0.3462, "lr": 8.347308805807378e-05, "epoch": 0.6064132798382897, "percentage": 30.32, "elapsed_time": "6:44:19", "remaining_time": "15:29:02", "throughput": 1355.92, "total_tokens": 32893512}
{"current_steps": 4950, "total_steps": 16324, "eval_loss": 0.1940447837114334, "epoch": 0.6064132798382897, "percentage": 30.32, "elapsed_time": "6:44:38", "remaining_time": "15:29:46", "throughput": 1354.83, "total_tokens": 32893512}
{"current_steps": 4955, "total_steps": 16324, "loss": 0.3274, "lr": 8.343544721872947e-05, "epoch": 0.6070258185047931, "percentage": 30.35, "elapsed_time": "6:45:43", "remaining_time": "15:30:55", "throughput": 1352.61, "total_tokens": 32927352}
{"current_steps": 4960, "total_steps": 16324, "loss": 0.2968, "lr": 8.339777207171078e-05, "epoch": 0.6076383571712964, "percentage": 30.38, "elapsed_time": "6:46:44", "remaining_time": "15:31:52", "throughput": 1350.66, "total_tokens": 32961520}
{"current_steps": 4965, "total_steps": 16324, "loss": 0.3236, "lr": 8.336006265567567e-05, "epoch": 0.6082508958377998, "percentage": 30.42, "elapsed_time": "6:47:43", "remaining_time": "15:32:47", "throughput": 1348.73, "total_tokens": 32994296}
{"current_steps": 4970, "total_steps": 16324, "loss": 0.3188, "lr": 8.332231900931727e-05, "epoch": 0.6088634345043031, "percentage": 30.45, "elapsed_time": "6:48:43", "remaining_time": "15:33:43", "throughput": 1346.81, "total_tokens": 33028048}
{"current_steps": 4975, "total_steps": 16324, "loss": 0.3119, "lr": 8.328454117136382e-05, "epoch": 0.6094759731708064, "percentage": 30.48, "elapsed_time": "6:49:43", "remaining_time": "15:34:39", "throughput": 1344.87, "total_tokens": 33061592}
{"current_steps": 4980, "total_steps": 16324, "loss": 0.3797, "lr": 8.324672918057872e-05, "epoch": 0.6100885118373097, "percentage": 30.51, "elapsed_time": "6:50:42", "remaining_time": "15:35:33", "throughput": 1342.98, "total_tokens": 33094608}
{"current_steps": 4985, "total_steps": 16324, "loss": 0.3213, "lr": 8.320888307576034e-05, "epoch": 0.6107010505038131, "percentage": 30.54, "elapsed_time": "6:51:41", "remaining_time": "15:36:27", "throughput": 1341.11, "total_tokens": 33127896}
{"current_steps": 4990, "total_steps": 16324, "loss": 0.3225, "lr": 8.317100289574204e-05, "epoch": 0.6113135891703164, "percentage": 30.57, "elapsed_time": "6:52:42", "remaining_time": "15:37:23", "throughput": 1339.21, "total_tokens": 33161720}
{"current_steps": 4995, "total_steps": 16324, "loss": 0.3007, "lr": 8.31330886793922e-05, "epoch": 0.6119261278368197, "percentage": 30.6, "elapsed_time": "6:53:41", "remaining_time": "15:38:17", "throughput": 1337.35, "total_tokens": 33195408}
{"current_steps": 5000, "total_steps": 16324, "loss": 0.3169, "lr": 8.309514046561412e-05, "epoch": 0.612538666503323, "percentage": 30.63, "elapsed_time": "6:54:41", "remaining_time": "15:39:11", "throughput": 1335.51, "total_tokens": 33229224}
{"current_steps": 5000, "total_steps": 16324, "eval_loss": 0.16666926443576813, "epoch": 0.612538666503323, "percentage": 30.63, "elapsed_time": "6:55:01", "remaining_time": "15:39:56", "throughput": 1334.43, "total_tokens": 33229224}
{"current_steps": 5005, "total_steps": 16324, "loss": 0.2759, "lr": 8.305715829334592e-05, "epoch": 0.6131512051698264, "percentage": 30.66, "elapsed_time": "6:56:07", "remaining_time": "15:41:05", "throughput": 1332.24, "total_tokens": 33262888}
{"current_steps": 5010, "total_steps": 16324, "loss": 0.2963, "lr": 8.301914220156067e-05, "epoch": 0.6137637438363297, "percentage": 30.69, "elapsed_time": "6:57:07", "remaining_time": "15:42:00", "throughput": 1330.38, "total_tokens": 33296648}
{"current_steps": 5015, "total_steps": 16324, "loss": 0.301, "lr": 8.29810922292661e-05, "epoch": 0.6143762825028329, "percentage": 30.72, "elapsed_time": "6:58:07", "remaining_time": "15:42:53", "throughput": 1328.56, "total_tokens": 33330264}
{"current_steps": 5020, "total_steps": 16324, "loss": 0.3329, "lr": 8.294300841550486e-05, "epoch": 0.6149888211693363, "percentage": 30.75, "elapsed_time": "6:59:07", "remaining_time": "15:43:47", "throughput": 1326.7, "total_tokens": 33363344}
{"current_steps": 5025, "total_steps": 16324, "loss": 0.2672, "lr": 8.290489079935423e-05, "epoch": 0.6156013598358396, "percentage": 30.78, "elapsed_time": "7:00:08", "remaining_time": "15:44:42", "throughput": 1324.9, "total_tokens": 33398208}
{"current_steps": 5030, "total_steps": 16324, "loss": 0.3415, "lr": 8.286673941992621e-05, "epoch": 0.616213898502343, "percentage": 30.81, "elapsed_time": "7:01:07", "remaining_time": "15:45:34", "throughput": 1323.12, "total_tokens": 33432144}
{"current_steps": 5035, "total_steps": 16324, "loss": 0.312, "lr": 8.28285543163674e-05, "epoch": 0.6168264371688463, "percentage": 30.84, "elapsed_time": "7:02:07", "remaining_time": "15:46:27", "throughput": 1321.29, "total_tokens": 33465232}
{"current_steps": 5040, "total_steps": 16324, "loss": 0.2891, "lr": 8.279033552785906e-05, "epoch": 0.6174389758353496, "percentage": 30.87, "elapsed_time": "7:03:08", "remaining_time": "15:47:20", "throughput": 1319.5, "total_tokens": 33499568}
{"current_steps": 5045, "total_steps": 16324, "loss": 0.3449, "lr": 8.275208309361697e-05, "epoch": 0.6180515145018529, "percentage": 30.91, "elapsed_time": "7:04:07", "remaining_time": "15:48:12", "throughput": 1317.73, "total_tokens": 33532984}
{"current_steps": 5050, "total_steps": 16324, "loss": 0.3488, "lr": 8.271379705289147e-05, "epoch": 0.6186640531683563, "percentage": 30.94, "elapsed_time": "7:05:07", "remaining_time": "15:49:04", "throughput": 1315.93, "total_tokens": 33566056}
{"current_steps": 5050, "total_steps": 16324, "eval_loss": 0.29742637276649475, "epoch": 0.6186640531683563, "percentage": 30.94, "elapsed_time": "7:05:27", "remaining_time": "15:49:48", "throughput": 1314.91, "total_tokens": 33566056}
{"current_steps": 5055, "total_steps": 16324, "loss": 0.29, "lr": 8.267547744496734e-05, "epoch": 0.6192765918348596, "percentage": 30.97, "elapsed_time": "7:06:32", "remaining_time": "15:50:52", "throughput": 1312.88, "total_tokens": 33599808}
{"current_steps": 5060, "total_steps": 16324, "loss": 0.3881, "lr": 8.263712430916385e-05, "epoch": 0.6198891305013629, "percentage": 31.0, "elapsed_time": "7:07:32", "remaining_time": "15:51:43", "throughput": 1311.14, "total_tokens": 33633440}
{"current_steps": 5065, "total_steps": 16324, "loss": 0.3024, "lr": 8.259873768483464e-05, "epoch": 0.6205016691678662, "percentage": 31.03, "elapsed_time": "7:08:31", "remaining_time": "15:52:35", "throughput": 1309.39, "total_tokens": 33667064}
{"current_steps": 5070, "total_steps": 16324, "loss": 0.3578, "lr": 8.256031761136773e-05, "epoch": 0.6211142078343695, "percentage": 31.06, "elapsed_time": "7:09:32", "remaining_time": "15:53:27", "throughput": 1307.63, "total_tokens": 33700856}
{"current_steps": 5075, "total_steps": 16324, "loss": 0.3412, "lr": 8.252186412818546e-05, "epoch": 0.6217267465008729, "percentage": 31.09, "elapsed_time": "7:10:31", "remaining_time": "15:54:17", "throughput": 1305.9, "total_tokens": 33733776}
{"current_steps": 5080, "total_steps": 16324, "loss": 0.3171, "lr": 8.248337727474442e-05, "epoch": 0.6223392851673762, "percentage": 31.12, "elapsed_time": "7:11:31", "remaining_time": "15:55:07", "throughput": 1304.17, "total_tokens": 33766872}
{"current_steps": 5085, "total_steps": 16324, "loss": 0.3452, "lr": 8.244485709053549e-05, "epoch": 0.6229518238338795, "percentage": 31.15, "elapsed_time": "7:12:32", "remaining_time": "15:55:59", "throughput": 1302.41, "total_tokens": 33800032}
{"current_steps": 5090, "total_steps": 16324, "loss": 0.2639, "lr": 8.240630361508373e-05, "epoch": 0.6235643625003828, "percentage": 31.18, "elapsed_time": "7:13:31", "remaining_time": "15:56:49", "throughput": 1300.72, "total_tokens": 33833960}
{"current_steps": 5095, "total_steps": 16324, "loss": 0.3078, "lr": 8.236771688794832e-05, "epoch": 0.6241769011668862, "percentage": 31.21, "elapsed_time": "7:14:32", "remaining_time": "15:57:40", "throughput": 1299.02, "total_tokens": 33867984}
{"current_steps": 5100, "total_steps": 16324, "loss": 0.3996, "lr": 8.232909694872261e-05, "epoch": 0.6247894398333895, "percentage": 31.24, "elapsed_time": "7:15:33", "remaining_time": "15:58:33", "throughput": 1297.24, "total_tokens": 33901184}
{"current_steps": 5100, "total_steps": 16324, "eval_loss": 0.31558582186698914, "epoch": 0.6247894398333895, "percentage": 31.24, "elapsed_time": "7:15:52", "remaining_time": "15:59:16", "throughput": 1296.28, "total_tokens": 33901184}
{"current_steps": 5105, "total_steps": 16324, "loss": 0.3072, "lr": 8.229044383703406e-05, "epoch": 0.6254019784998928, "percentage": 31.27, "elapsed_time": "7:16:57", "remaining_time": "16:00:16", "throughput": 1294.34, "total_tokens": 33934000}
{"current_steps": 5110, "total_steps": 16324, "loss": 0.3554, "lr": 8.225175759254406e-05, "epoch": 0.6260145171663961, "percentage": 31.3, "elapsed_time": "7:17:57", "remaining_time": "16:01:06", "throughput": 1292.66, "total_tokens": 33967616}
{"current_steps": 5115, "total_steps": 16324, "loss": 0.3061, "lr": 8.221303825494809e-05, "epoch": 0.6266270558328995, "percentage": 31.33, "elapsed_time": "7:18:58", "remaining_time": "16:01:57", "throughput": 1290.97, "total_tokens": 34001784}
{"current_steps": 5120, "total_steps": 16324, "loss": 0.3074, "lr": 8.217428586397555e-05, "epoch": 0.6272395944994028, "percentage": 31.36, "elapsed_time": "7:19:58", "remaining_time": "16:02:47", "throughput": 1289.3, "total_tokens": 34035480}
{"current_steps": 5125, "total_steps": 16324, "loss": 0.3414, "lr": 8.213550045938976e-05, "epoch": 0.627852133165906, "percentage": 31.4, "elapsed_time": "7:20:58", "remaining_time": "16:03:36", "throughput": 1287.65, "total_tokens": 34069440}
{"current_steps": 5130, "total_steps": 16324, "loss": 0.2914, "lr": 8.20966820809879e-05, "epoch": 0.6284646718324094, "percentage": 31.43, "elapsed_time": "7:21:59", "remaining_time": "16:04:26", "throughput": 1285.98, "total_tokens": 34103032}
{"current_steps": 5135, "total_steps": 16324, "loss": 0.2738, "lr": 8.205783076860102e-05, "epoch": 0.6290772104989127, "percentage": 31.46, "elapsed_time": "7:22:59", "remaining_time": "16:05:16", "throughput": 1284.33, "total_tokens": 34137288}
{"current_steps": 5140, "total_steps": 16324, "loss": 0.3311, "lr": 8.201894656209395e-05, "epoch": 0.6296897491654161, "percentage": 31.49, "elapsed_time": "7:23:59", "remaining_time": "16:06:04", "throughput": 1282.71, "total_tokens": 34171000}
{"current_steps": 5145, "total_steps": 16324, "loss": 0.3003, "lr": 8.198002950136524e-05, "epoch": 0.6303022878319194, "percentage": 31.52, "elapsed_time": "7:25:00", "remaining_time": "16:06:53", "throughput": 1281.08, "total_tokens": 34204896}
{"current_steps": 5150, "total_steps": 16324, "loss": 0.3088, "lr": 8.194107962634719e-05, "epoch": 0.6309148264984227, "percentage": 31.55, "elapsed_time": "7:26:00", "remaining_time": "16:07:42", "throughput": 1279.45, "total_tokens": 34238496}
{"current_steps": 5150, "total_steps": 16324, "eval_loss": 0.2916165292263031, "epoch": 0.6309148264984227, "percentage": 31.55, "elapsed_time": "7:26:19", "remaining_time": "16:08:23", "throughput": 1278.54, "total_tokens": 34238496}
{"current_steps": 5155, "total_steps": 16324, "loss": 0.2772, "lr": 8.190209697700575e-05, "epoch": 0.631527365164926, "percentage": 31.58, "elapsed_time": "7:27:24", "remaining_time": "16:09:22", "throughput": 1276.68, "total_tokens": 34271792}
{"current_steps": 5160, "total_steps": 16324, "loss": 0.3122, "lr": 8.186308159334051e-05, "epoch": 0.6321399038314294, "percentage": 31.61, "elapsed_time": "7:28:24", "remaining_time": "16:10:09", "throughput": 1275.07, "total_tokens": 34305288}
{"current_steps": 5165, "total_steps": 16324, "loss": 0.2928, "lr": 8.182403351538466e-05, "epoch": 0.6327524424979327, "percentage": 31.64, "elapsed_time": "7:29:24", "remaining_time": "16:10:56", "throughput": 1273.48, "total_tokens": 34338808}
{"current_steps": 5170, "total_steps": 16324, "loss": 0.3602, "lr": 8.178495278320489e-05, "epoch": 0.633364981164436, "percentage": 31.67, "elapsed_time": "7:30:23", "remaining_time": "16:11:42", "throughput": 1271.95, "total_tokens": 34372672}
{"current_steps": 5175, "total_steps": 16324, "loss": 0.4253, "lr": 8.174583943690146e-05, "epoch": 0.6339775198309393, "percentage": 31.7, "elapsed_time": "7:31:23", "remaining_time": "16:12:28", "throughput": 1270.35, "total_tokens": 34405760}
{"current_steps": 5180, "total_steps": 16324, "loss": 0.27, "lr": 8.170669351660803e-05, "epoch": 0.6345900584974427, "percentage": 31.73, "elapsed_time": "7:32:23", "remaining_time": "16:13:14", "throughput": 1268.83, "total_tokens": 34439920}
{"current_steps": 5185, "total_steps": 16324, "loss": 0.3228, "lr": 8.166751506249172e-05, "epoch": 0.635202597163946, "percentage": 31.76, "elapsed_time": "7:33:22", "remaining_time": "16:13:59", "throughput": 1267.28, "total_tokens": 34473336}
{"current_steps": 5190, "total_steps": 16324, "loss": 0.3353, "lr": 8.162830411475306e-05, "epoch": 0.6358151358304494, "percentage": 31.79, "elapsed_time": "7:34:22", "remaining_time": "16:14:45", "throughput": 1265.73, "total_tokens": 34507024}
{"current_steps": 5195, "total_steps": 16324, "loss": 0.3101, "lr": 8.158906071362587e-05, "epoch": 0.6364276744969526, "percentage": 31.82, "elapsed_time": "7:35:22", "remaining_time": "16:15:30", "throughput": 1264.19, "total_tokens": 34540456}
{"current_steps": 5200, "total_steps": 16324, "loss": 0.2997, "lr": 8.154978489937728e-05, "epoch": 0.6370402131634559, "percentage": 31.85, "elapsed_time": "7:36:21", "remaining_time": "16:16:15", "throughput": 1262.69, "total_tokens": 34574624}
{"current_steps": 5200, "total_steps": 16324, "eval_loss": 0.18659941852092743, "epoch": 0.6370402131634559, "percentage": 31.85, "elapsed_time": "7:36:40", "remaining_time": "16:16:56", "throughput": 1261.81, "total_tokens": 34574624}
{"current_steps": 5205, "total_steps": 16324, "loss": 0.2585, "lr": 8.15104767123077e-05, "epoch": 0.6376527518299593, "percentage": 31.89, "elapsed_time": "7:37:46", "remaining_time": "16:17:53", "throughput": 1260.04, "total_tokens": 34608464}
{"current_steps": 5210, "total_steps": 16324, "loss": 0.3456, "lr": 8.147113619275075e-05, "epoch": 0.6382652904964626, "percentage": 31.92, "elapsed_time": "7:38:46", "remaining_time": "16:18:39", "throughput": 1258.49, "total_tokens": 34641632}
{"current_steps": 5215, "total_steps": 16324, "loss": 0.3748, "lr": 8.143176338107321e-05, "epoch": 0.6388778291629659, "percentage": 31.95, "elapsed_time": "7:39:45", "remaining_time": "16:19:23", "throughput": 1257.0, "total_tokens": 34675528}
{"current_steps": 5220, "total_steps": 16324, "loss": 0.3116, "lr": 8.1392358317675e-05, "epoch": 0.6394903678294692, "percentage": 31.98, "elapsed_time": "7:40:45", "remaining_time": "16:20:08", "throughput": 1255.48, "total_tokens": 34708832}
{"current_steps": 5225, "total_steps": 16324, "loss": 0.4067, "lr": 8.135292104298917e-05, "epoch": 0.6401029064959726, "percentage": 32.01, "elapsed_time": "7:41:47", "remaining_time": "16:20:55", "throughput": 1253.93, "total_tokens": 34742912}
{"current_steps": 5230, "total_steps": 16324, "loss": 0.2883, "lr": 8.131345159748177e-05, "epoch": 0.6407154451624759, "percentage": 32.04, "elapsed_time": "7:42:47", "remaining_time": "16:21:40", "throughput": 1252.45, "total_tokens": 34777040}
{"current_steps": 5235, "total_steps": 16324, "loss": 0.3315, "lr": 8.127395002165187e-05, "epoch": 0.6413279838289792, "percentage": 32.07, "elapsed_time": "7:43:47", "remaining_time": "16:22:24", "throughput": 1250.94, "total_tokens": 34810352}
{"current_steps": 5240, "total_steps": 16324, "loss": 0.4601, "lr": 8.123441635603155e-05, "epoch": 0.6419405224954825, "percentage": 32.1, "elapsed_time": "7:44:48", "remaining_time": "16:23:11", "throughput": 1249.39, "total_tokens": 34843504}
{"current_steps": 5245, "total_steps": 16324, "loss": 0.3493, "lr": 8.119485064118577e-05, "epoch": 0.6425530611619859, "percentage": 32.13, "elapsed_time": "7:45:47", "remaining_time": "16:23:54", "throughput": 1247.94, "total_tokens": 34877344}
{"current_steps": 5250, "total_steps": 16324, "loss": 0.2817, "lr": 8.115525291771237e-05, "epoch": 0.6431655998284892, "percentage": 32.16, "elapsed_time": "7:46:48", "remaining_time": "16:24:39", "throughput": 1246.49, "total_tokens": 34912360}
{"current_steps": 5250, "total_steps": 16324, "eval_loss": 0.15959948301315308, "epoch": 0.6431655998284892, "percentage": 32.16, "elapsed_time": "7:47:08", "remaining_time": "16:25:20", "throughput": 1245.62, "total_tokens": 34912360}
{"current_steps": 5255, "total_steps": 16324, "loss": 0.3115, "lr": 8.111562322624208e-05, "epoch": 0.6437781384949925, "percentage": 32.19, "elapsed_time": "7:48:14", "remaining_time": "16:26:17", "throughput": 1243.88, "total_tokens": 34945928}
{"current_steps": 5260, "total_steps": 16324, "loss": 0.2517, "lr": 8.107596160743842e-05, "epoch": 0.6443906771614958, "percentage": 32.22, "elapsed_time": "7:49:14", "remaining_time": "16:27:00", "throughput": 1242.44, "total_tokens": 34980184}
{"current_steps": 5265, "total_steps": 16324, "loss": 0.2597, "lr": 8.103626810199762e-05, "epoch": 0.6450032158279991, "percentage": 32.25, "elapsed_time": "7:50:14", "remaining_time": "16:27:43", "throughput": 1241.0, "total_tokens": 35014128}
{"current_steps": 5270, "total_steps": 16324, "loss": 0.3011, "lr": 8.099654275064871e-05, "epoch": 0.6456157544945025, "percentage": 32.28, "elapsed_time": "7:51:14", "remaining_time": "16:28:26", "throughput": 1239.56, "total_tokens": 35047936}
{"current_steps": 5275, "total_steps": 16324, "loss": 0.3512, "lr": 8.095678559415332e-05, "epoch": 0.6462282931610058, "percentage": 32.31, "elapsed_time": "7:52:14", "remaining_time": "16:29:09", "throughput": 1238.13, "total_tokens": 35081808}
{"current_steps": 5280, "total_steps": 16324, "loss": 0.2854, "lr": 8.091699667330574e-05, "epoch": 0.6468408318275091, "percentage": 32.35, "elapsed_time": "7:53:13", "remaining_time": "16:29:49", "throughput": 1236.74, "total_tokens": 35115352}
{"current_steps": 5285, "total_steps": 16324, "loss": 0.3199, "lr": 8.08771760289329e-05, "epoch": 0.6474533704940124, "percentage": 32.38, "elapsed_time": "7:54:13", "remaining_time": "16:30:30", "throughput": 1235.33, "total_tokens": 35148976}
{"current_steps": 5290, "total_steps": 16324, "loss": 0.246, "lr": 8.083732370189421e-05, "epoch": 0.6480659091605158, "percentage": 32.41, "elapsed_time": "7:55:13", "remaining_time": "16:31:13", "throughput": 1233.93, "total_tokens": 35183240}
{"current_steps": 5295, "total_steps": 16324, "loss": 0.2896, "lr": 8.079743973308163e-05, "epoch": 0.6486784478270191, "percentage": 32.44, "elapsed_time": "7:56:12", "remaining_time": "16:31:53", "throughput": 1232.56, "total_tokens": 35216928}
{"current_steps": 5300, "total_steps": 16324, "loss": 0.2997, "lr": 8.075752416341959e-05, "epoch": 0.6492909864935223, "percentage": 32.47, "elapsed_time": "7:57:12", "remaining_time": "16:32:34", "throughput": 1231.17, "total_tokens": 35251096}
{"current_steps": 5300, "total_steps": 16324, "eval_loss": 0.11087673157453537, "epoch": 0.6492909864935223, "percentage": 32.47, "elapsed_time": "7:57:31", "remaining_time": "16:33:15", "throughput": 1230.34, "total_tokens": 35251096}
{"current_steps": 5305, "total_steps": 16324, "loss": 0.2363, "lr": 8.07175770338649e-05, "epoch": 0.6499035251600257, "percentage": 32.5, "elapsed_time": "7:58:36", "remaining_time": "16:34:07", "throughput": 1228.73, "total_tokens": 35285272}
{"current_steps": 5310, "total_steps": 16324, "loss": 0.2582, "lr": 8.067759838540682e-05, "epoch": 0.650516063826529, "percentage": 32.53, "elapsed_time": "7:59:36", "remaining_time": "16:34:47", "throughput": 1227.38, "total_tokens": 35319216}
{"current_steps": 5315, "total_steps": 16324, "loss": 0.3376, "lr": 8.06375882590669e-05, "epoch": 0.6511286024930324, "percentage": 32.56, "elapsed_time": "8:00:36", "remaining_time": "16:35:28", "throughput": 1226.01, "total_tokens": 35353264}
{"current_steps": 5320, "total_steps": 16324, "loss": 0.2647, "lr": 8.0597546695899e-05, "epoch": 0.6517411411595357, "percentage": 32.59, "elapsed_time": "8:01:36", "remaining_time": "16:36:09", "throughput": 1224.62, "total_tokens": 35386600}
{"current_steps": 5325, "total_steps": 16324, "loss": 0.3179, "lr": 8.055747373698927e-05, "epoch": 0.652353679826039, "percentage": 32.62, "elapsed_time": "8:02:35", "remaining_time": "16:36:48", "throughput": 1223.25, "total_tokens": 35419584}
{"current_steps": 5330, "total_steps": 16324, "loss": 0.3219, "lr": 8.051736942345603e-05, "epoch": 0.6529662184925423, "percentage": 32.65, "elapsed_time": "8:03:35", "remaining_time": "16:37:29", "throughput": 1221.87, "total_tokens": 35452936}
{"current_steps": 5335, "total_steps": 16324, "loss": 0.3395, "lr": 8.04772337964498e-05, "epoch": 0.6535787571590457, "percentage": 32.68, "elapsed_time": "8:04:35", "remaining_time": "16:38:10", "throughput": 1220.47, "total_tokens": 35486136}
{"current_steps": 5340, "total_steps": 16324, "loss": 0.2821, "lr": 8.04370668971532e-05, "epoch": 0.654191295825549, "percentage": 32.71, "elapsed_time": "8:05:35", "remaining_time": "16:38:49", "throughput": 1219.11, "total_tokens": 35519144}
{"current_steps": 5345, "total_steps": 16324, "loss": 0.2706, "lr": 8.039686876678099e-05, "epoch": 0.6548038344920523, "percentage": 32.74, "elapsed_time": "8:06:34", "remaining_time": "16:39:28", "throughput": 1217.75, "total_tokens": 35552264}
{"current_steps": 5350, "total_steps": 16324, "loss": 0.3163, "lr": 8.035663944657992e-05, "epoch": 0.6554163731585556, "percentage": 32.77, "elapsed_time": "8:07:35", "remaining_time": "16:40:09", "throughput": 1216.39, "total_tokens": 35586128}
{"current_steps": 5350, "total_steps": 16324, "eval_loss": 0.18712441623210907, "epoch": 0.6554163731585556, "percentage": 32.77, "elapsed_time": "8:07:55", "remaining_time": "16:40:49", "throughput": 1215.57, "total_tokens": 35586128}
{"current_steps": 5355, "total_steps": 16324, "loss": 0.2562, "lr": 8.031637897782877e-05, "epoch": 0.656028911825059, "percentage": 32.8, "elapsed_time": "8:09:00", "remaining_time": "16:41:40", "throughput": 1214.03, "total_tokens": 35620280}
{"current_steps": 5360, "total_steps": 16324, "loss": 0.3284, "lr": 8.027608740183828e-05, "epoch": 0.6566414504915623, "percentage": 32.84, "elapsed_time": "8:10:00", "remaining_time": "16:42:18", "throughput": 1212.7, "total_tokens": 35653840}
{"current_steps": 5365, "total_steps": 16324, "loss": 0.3455, "lr": 8.023576475995113e-05, "epoch": 0.6572539891580657, "percentage": 32.87, "elapsed_time": "8:11:01", "remaining_time": "16:43:00", "throughput": 1211.33, "total_tokens": 35687320}
{"current_steps": 5370, "total_steps": 16324, "loss": 0.3068, "lr": 8.019541109354184e-05, "epoch": 0.6578665278245689, "percentage": 32.9, "elapsed_time": "8:12:00", "remaining_time": "16:43:38", "throughput": 1210.03, "total_tokens": 35721288}
{"current_steps": 5375, "total_steps": 16324, "loss": 0.2947, "lr": 8.015502644401678e-05, "epoch": 0.6584790664910722, "percentage": 32.93, "elapsed_time": "8:13:01", "remaining_time": "16:44:17", "throughput": 1208.7, "total_tokens": 35754632}
{"current_steps": 5380, "total_steps": 16324, "loss": 0.3521, "lr": 8.01146108528141e-05, "epoch": 0.6590916051575756, "percentage": 32.96, "elapsed_time": "8:14:01", "remaining_time": "16:44:55", "throughput": 1207.37, "total_tokens": 35787736}
{"current_steps": 5385, "total_steps": 16324, "loss": 0.2652, "lr": 8.007416436140373e-05, "epoch": 0.6597041438240789, "percentage": 32.99, "elapsed_time": "8:15:01", "remaining_time": "16:45:35", "throughput": 1206.04, "total_tokens": 35821344}
{"current_steps": 5390, "total_steps": 16324, "loss": 0.2908, "lr": 8.003368701128727e-05, "epoch": 0.6603166824905822, "percentage": 33.02, "elapsed_time": "8:16:00", "remaining_time": "16:46:11", "throughput": 1204.79, "total_tokens": 35855384}
{"current_steps": 5395, "total_steps": 16324, "loss": 0.3399, "lr": 7.999317884399798e-05, "epoch": 0.6609292211570855, "percentage": 33.05, "elapsed_time": "8:16:59", "remaining_time": "16:46:48", "throughput": 1203.5, "total_tokens": 35888456}
{"current_steps": 5400, "total_steps": 16324, "loss": 0.3341, "lr": 7.995263990110079e-05, "epoch": 0.6615417598235889, "percentage": 33.08, "elapsed_time": "8:18:00", "remaining_time": "16:47:26", "throughput": 1202.22, "total_tokens": 35922304}
{"current_steps": 5400, "total_steps": 16324, "eval_loss": 0.20246051251888275, "epoch": 0.6615417598235889, "percentage": 33.08, "elapsed_time": "8:18:19", "remaining_time": "16:48:05", "throughput": 1201.44, "total_tokens": 35922304}
{"current_steps": 5405, "total_steps": 16324, "loss": 0.3332, "lr": 7.991207022419219e-05, "epoch": 0.6621542984900922, "percentage": 33.11, "elapsed_time": "8:19:24", "remaining_time": "16:48:53", "throughput": 1199.96, "total_tokens": 35956112}
{"current_steps": 5410, "total_steps": 16324, "loss": 0.329, "lr": 7.987146985490014e-05, "epoch": 0.6627668371565955, "percentage": 33.14, "elapsed_time": "8:20:23", "remaining_time": "16:49:29", "throughput": 1198.68, "total_tokens": 35989040}
{"current_steps": 5415, "total_steps": 16324, "loss": 0.3065, "lr": 7.983083883488419e-05, "epoch": 0.6633793758230988, "percentage": 33.17, "elapsed_time": "8:21:23", "remaining_time": "16:50:06", "throughput": 1197.42, "total_tokens": 36023016}
{"current_steps": 5420, "total_steps": 16324, "loss": 0.3266, "lr": 7.97901772058353e-05, "epoch": 0.6639919144896022, "percentage": 33.2, "elapsed_time": "8:22:22", "remaining_time": "16:50:41", "throughput": 1196.19, "total_tokens": 36056648}
{"current_steps": 5425, "total_steps": 16324, "loss": 0.3161, "lr": 7.974948500947582e-05, "epoch": 0.6646044531561055, "percentage": 33.23, "elapsed_time": "8:23:23", "remaining_time": "16:51:19", "throughput": 1194.9, "total_tokens": 36089784}
{"current_steps": 5430, "total_steps": 16324, "loss": 0.3045, "lr": 7.970876228755945e-05, "epoch": 0.6652169918226088, "percentage": 33.26, "elapsed_time": "8:24:23", "remaining_time": "16:51:56", "throughput": 1193.65, "total_tokens": 36123648}
{"current_steps": 5435, "total_steps": 16324, "loss": 0.2872, "lr": 7.96680090818713e-05, "epoch": 0.6658295304891121, "percentage": 33.29, "elapsed_time": "8:25:22", "remaining_time": "16:52:31", "throughput": 1192.41, "total_tokens": 36157176}
{"current_steps": 5440, "total_steps": 16324, "loss": 0.28, "lr": 7.962722543422767e-05, "epoch": 0.6664420691556154, "percentage": 33.33, "elapsed_time": "8:26:22", "remaining_time": "16:53:07", "throughput": 1191.18, "total_tokens": 36190944}
{"current_steps": 5445, "total_steps": 16324, "loss": 0.3365, "lr": 7.958641138647615e-05, "epoch": 0.6670546078221188, "percentage": 33.36, "elapsed_time": "8:27:22", "remaining_time": "16:53:43", "throughput": 1189.92, "total_tokens": 36224464}
{"current_steps": 5450, "total_steps": 16324, "loss": 0.3126, "lr": 7.954556698049544e-05, "epoch": 0.6676671464886221, "percentage": 33.39, "elapsed_time": "8:28:21", "remaining_time": "16:54:18", "throughput": 1188.67, "total_tokens": 36256800}
{"current_steps": 5450, "total_steps": 16324, "eval_loss": 0.20314465463161469, "epoch": 0.6676671464886221, "percentage": 33.39, "elapsed_time": "8:28:41", "remaining_time": "16:54:57", "throughput": 1187.92, "total_tokens": 36256800}
{"current_steps": 5455, "total_steps": 16324, "loss": 0.3205, "lr": 7.950469225819553e-05, "epoch": 0.6682796851551254, "percentage": 33.42, "elapsed_time": "8:29:46", "remaining_time": "16:55:43", "throughput": 1186.46, "total_tokens": 36289856}
{"current_steps": 5460, "total_steps": 16324, "loss": 0.3886, "lr": 7.946378726151736e-05, "epoch": 0.6688922238216287, "percentage": 33.45, "elapsed_time": "8:30:46", "remaining_time": "16:56:19", "throughput": 1185.2, "total_tokens": 36322792}
{"current_steps": 5465, "total_steps": 16324, "loss": 0.3497, "lr": 7.942285203243306e-05, "epoch": 0.6695047624881321, "percentage": 33.48, "elapsed_time": "8:31:46", "remaining_time": "16:56:53", "throughput": 1184.0, "total_tokens": 36356144}
{"current_steps": 5470, "total_steps": 16324, "loss": 0.3647, "lr": 7.938188661294574e-05, "epoch": 0.6701173011546354, "percentage": 33.51, "elapsed_time": "8:32:45", "remaining_time": "16:57:27", "throughput": 1182.78, "total_tokens": 36389240}
{"current_steps": 5475, "total_steps": 16324, "loss": 0.405, "lr": 7.934089104508943e-05, "epoch": 0.6707298398211388, "percentage": 33.54, "elapsed_time": "8:33:46", "remaining_time": "16:58:04", "throughput": 1181.53, "total_tokens": 36422536}
{"current_steps": 5480, "total_steps": 16324, "loss": 0.3243, "lr": 7.92998653709292e-05, "epoch": 0.671342378487642, "percentage": 33.57, "elapsed_time": "8:34:46", "remaining_time": "16:58:38", "throughput": 1180.33, "total_tokens": 36455680}
{"current_steps": 5485, "total_steps": 16324, "loss": 0.2469, "lr": 7.925880963256094e-05, "epoch": 0.6719549171541453, "percentage": 33.6, "elapsed_time": "8:35:46", "remaining_time": "16:59:13", "throughput": 1179.11, "total_tokens": 36489320}
{"current_steps": 5490, "total_steps": 16324, "loss": 0.309, "lr": 7.921772387211138e-05, "epoch": 0.6725674558206487, "percentage": 33.63, "elapsed_time": "8:36:47", "remaining_time": "16:59:49", "throughput": 1177.88, "total_tokens": 36522792}
{"current_steps": 5495, "total_steps": 16324, "loss": 0.3236, "lr": 7.917660813173811e-05, "epoch": 0.673179994487152, "percentage": 33.66, "elapsed_time": "8:37:46", "remaining_time": "17:00:22", "throughput": 1176.7, "total_tokens": 36556056}
{"current_steps": 5500, "total_steps": 16324, "loss": 0.3145, "lr": 7.913546245362942e-05, "epoch": 0.6737925331536553, "percentage": 33.69, "elapsed_time": "8:38:46", "remaining_time": "17:00:56", "throughput": 1175.52, "total_tokens": 36589616}
{"current_steps": 5500, "total_steps": 16324, "eval_loss": 0.1359092891216278, "epoch": 0.6737925331536553, "percentage": 33.69, "elapsed_time": "8:39:05", "remaining_time": "17:01:35", "throughput": 1174.78, "total_tokens": 36589616}
{"current_steps": 5505, "total_steps": 16324, "loss": 0.3184, "lr": 7.909428688000438e-05, "epoch": 0.6744050718201586, "percentage": 33.72, "elapsed_time": "8:40:10", "remaining_time": "17:02:17", "throughput": 1173.43, "total_tokens": 36623040}
{"current_steps": 5510, "total_steps": 16324, "loss": 0.2705, "lr": 7.905308145311272e-05, "epoch": 0.675017610486662, "percentage": 33.75, "elapsed_time": "8:41:09", "remaining_time": "17:02:49", "throughput": 1172.29, "total_tokens": 36656560}
{"current_steps": 5515, "total_steps": 16324, "loss": 0.3864, "lr": 7.901184621523475e-05, "epoch": 0.6756301491531653, "percentage": 33.78, "elapsed_time": "8:42:08", "remaining_time": "17:03:21", "throughput": 1171.12, "total_tokens": 36689320}
{"current_steps": 5520, "total_steps": 16324, "loss": 0.3058, "lr": 7.897058120868143e-05, "epoch": 0.6762426878196686, "percentage": 33.82, "elapsed_time": "8:43:07", "remaining_time": "17:03:53", "throughput": 1169.97, "total_tokens": 36722560}
{"current_steps": 5525, "total_steps": 16324, "loss": 0.3803, "lr": 7.892928647579423e-05, "epoch": 0.6768552264861719, "percentage": 33.85, "elapsed_time": "8:44:07", "remaining_time": "17:04:25", "throughput": 1168.8, "total_tokens": 36755400}
{"current_steps": 5530, "total_steps": 16324, "loss": 0.3232, "lr": 7.888796205894516e-05, "epoch": 0.6774677651526753, "percentage": 33.88, "elapsed_time": "8:45:06", "remaining_time": "17:04:57", "throughput": 1167.68, "total_tokens": 36789136}
{"current_steps": 5535, "total_steps": 16324, "loss": 0.3131, "lr": 7.884660800053662e-05, "epoch": 0.6780803038191786, "percentage": 33.91, "elapsed_time": "8:46:05", "remaining_time": "17:05:28", "throughput": 1166.55, "total_tokens": 36822600}
{"current_steps": 5540, "total_steps": 16324, "loss": 0.3068, "lr": 7.880522434300152e-05, "epoch": 0.678692842485682, "percentage": 33.94, "elapsed_time": "8:47:05", "remaining_time": "17:06:00", "throughput": 1165.39, "total_tokens": 36855528}
{"current_steps": 5545, "total_steps": 16324, "loss": 0.3238, "lr": 7.876381112880305e-05, "epoch": 0.6793053811521852, "percentage": 33.97, "elapsed_time": "8:48:04", "remaining_time": "17:06:31", "throughput": 1164.25, "total_tokens": 36888496}
{"current_steps": 5550, "total_steps": 16324, "loss": 0.337, "lr": 7.872236840043477e-05, "epoch": 0.6799179198186885, "percentage": 34.0, "elapsed_time": "8:49:04", "remaining_time": "17:07:03", "throughput": 1163.12, "total_tokens": 36922056}
{"current_steps": 5550, "total_steps": 16324, "eval_loss": 0.09095880389213562, "epoch": 0.6799179198186885, "percentage": 34.0, "elapsed_time": "8:49:23", "remaining_time": "17:07:42", "throughput": 1162.39, "total_tokens": 36922056}
{"current_steps": 5555, "total_steps": 16324, "loss": 0.2545, "lr": 7.868089620042054e-05, "epoch": 0.6805304584851919, "percentage": 34.03, "elapsed_time": "8:50:30", "remaining_time": "17:08:26", "throughput": 1161.01, "total_tokens": 36955456}
{"current_steps": 5560, "total_steps": 16324, "loss": 0.3144, "lr": 7.863939457131443e-05, "epoch": 0.6811429971516952, "percentage": 34.06, "elapsed_time": "8:51:29", "remaining_time": "17:08:57", "throughput": 1159.89, "total_tokens": 36988504}
{"current_steps": 5565, "total_steps": 16324, "loss": 0.3051, "lr": 7.859786355570071e-05, "epoch": 0.6817555358181985, "percentage": 34.09, "elapsed_time": "8:52:29", "remaining_time": "17:09:28", "throughput": 1158.75, "total_tokens": 37021504}
{"current_steps": 5570, "total_steps": 16324, "loss": 0.2666, "lr": 7.855630319619383e-05, "epoch": 0.6823680744847018, "percentage": 34.12, "elapsed_time": "8:53:28", "remaining_time": "17:09:59", "throughput": 1157.67, "total_tokens": 37055704}
{"current_steps": 5575, "total_steps": 16324, "loss": 0.2894, "lr": 7.851471353543831e-05, "epoch": 0.6829806131512052, "percentage": 34.15, "elapsed_time": "8:54:27", "remaining_time": "17:10:28", "throughput": 1156.6, "total_tokens": 37089240}
{"current_steps": 5580, "total_steps": 16324, "loss": 0.3601, "lr": 7.847309461610878e-05, "epoch": 0.6835931518177085, "percentage": 34.18, "elapsed_time": "8:55:27", "remaining_time": "17:10:59", "throughput": 1155.49, "total_tokens": 37122600}
{"current_steps": 5585, "total_steps": 16324, "loss": 0.3257, "lr": 7.843144648090984e-05, "epoch": 0.6842056904842119, "percentage": 34.21, "elapsed_time": "8:56:27", "remaining_time": "17:11:30", "throughput": 1154.38, "total_tokens": 37155976}
{"current_steps": 5590, "total_steps": 16324, "loss": 0.277, "lr": 7.838976917257612e-05, "epoch": 0.6848182291507151, "percentage": 34.24, "elapsed_time": "8:57:26", "remaining_time": "17:11:59", "throughput": 1153.33, "total_tokens": 37190544}
{"current_steps": 5595, "total_steps": 16324, "loss": 0.287, "lr": 7.834806273387214e-05, "epoch": 0.6854307678172185, "percentage": 34.27, "elapsed_time": "8:58:26", "remaining_time": "17:12:30", "throughput": 1152.24, "total_tokens": 37224600}
{"current_steps": 5600, "total_steps": 16324, "loss": 0.2861, "lr": 7.830632720759234e-05, "epoch": 0.6860433064837218, "percentage": 34.31, "elapsed_time": "8:59:26", "remaining_time": "17:13:02", "throughput": 1151.12, "total_tokens": 37258136}
{"current_steps": 5600, "total_steps": 16324, "eval_loss": 0.16602179408073425, "epoch": 0.6860433064837218, "percentage": 34.31, "elapsed_time": "8:59:46", "remaining_time": "17:13:40", "throughput": 1150.42, "total_tokens": 37258136}
{"current_steps": 5605, "total_steps": 16324, "loss": 0.2793, "lr": 7.8264562636561e-05, "epoch": 0.6866558451502252, "percentage": 34.34, "elapsed_time": "9:00:51", "remaining_time": "17:14:21", "throughput": 1149.14, "total_tokens": 37291664}
{"current_steps": 5610, "total_steps": 16324, "loss": 0.3119, "lr": 7.822276906363222e-05, "epoch": 0.6872683838167284, "percentage": 34.37, "elapsed_time": "9:01:51", "remaining_time": "17:14:50", "throughput": 1148.05, "total_tokens": 37324896}
{"current_steps": 5615, "total_steps": 16324, "loss": 0.3234, "lr": 7.818094653168978e-05, "epoch": 0.6878809224832317, "percentage": 34.4, "elapsed_time": "9:02:52", "remaining_time": "17:15:21", "throughput": 1146.96, "total_tokens": 37358664}
{"current_steps": 5620, "total_steps": 16324, "loss": 0.3019, "lr": 7.813909508364727e-05, "epoch": 0.6884934611497351, "percentage": 34.43, "elapsed_time": "9:03:50", "remaining_time": "17:15:49", "throughput": 1145.93, "total_tokens": 37392600}
{"current_steps": 5625, "total_steps": 16324, "loss": 0.3575, "lr": 7.809721476244792e-05, "epoch": 0.6891059998162384, "percentage": 34.46, "elapsed_time": "9:04:49", "remaining_time": "17:16:17", "throughput": 1144.89, "total_tokens": 37426144}
{"current_steps": 5630, "total_steps": 16324, "loss": 0.4029, "lr": 7.805530561106458e-05, "epoch": 0.6897185384827417, "percentage": 34.49, "elapsed_time": "9:05:48", "remaining_time": "17:16:45", "throughput": 1143.86, "total_tokens": 37460184}
{"current_steps": 5635, "total_steps": 16324, "loss": 0.3128, "lr": 7.801336767249965e-05, "epoch": 0.690331077149245, "percentage": 34.52, "elapsed_time": "9:06:47", "remaining_time": "17:17:12", "throughput": 1142.82, "total_tokens": 37493384}
{"current_steps": 5640, "total_steps": 16324, "loss": 0.257, "lr": 7.797140098978512e-05, "epoch": 0.6909436158157484, "percentage": 34.55, "elapsed_time": "9:07:46", "remaining_time": "17:17:39", "throughput": 1141.83, "total_tokens": 37527904}
{"current_steps": 5645, "total_steps": 16324, "loss": 0.2803, "lr": 7.792940560598245e-05, "epoch": 0.6915561544822517, "percentage": 34.58, "elapsed_time": "9:08:45", "remaining_time": "17:18:07", "throughput": 1140.81, "total_tokens": 37561984}
{"current_steps": 5650, "total_steps": 16324, "loss": 0.383, "lr": 7.788738156418256e-05, "epoch": 0.6921686931487551, "percentage": 34.61, "elapsed_time": "9:09:45", "remaining_time": "17:18:36", "throughput": 1139.74, "total_tokens": 37594912}
{"current_steps": 5650, "total_steps": 16324, "eval_loss": 0.15634669363498688, "epoch": 0.6921686931487551, "percentage": 34.61, "elapsed_time": "9:10:04", "remaining_time": "17:19:12", "throughput": 1139.08, "total_tokens": 37594912}
{"current_steps": 5655, "total_steps": 16324, "loss": 0.3315, "lr": 7.784532890750576e-05, "epoch": 0.6927812318152583, "percentage": 34.64, "elapsed_time": "9:11:09", "remaining_time": "17:19:49", "throughput": 1137.85, "total_tokens": 37627672}
{"current_steps": 5660, "total_steps": 16324, "loss": 0.284, "lr": 7.780324767910174e-05, "epoch": 0.6933937704817617, "percentage": 34.67, "elapsed_time": "9:12:08", "remaining_time": "17:20:18", "throughput": 1136.81, "total_tokens": 37661200}
{"current_steps": 5665, "total_steps": 16324, "loss": 0.3313, "lr": 7.77611379221495e-05, "epoch": 0.694006309148265, "percentage": 34.7, "elapsed_time": "9:13:08", "remaining_time": "17:20:46", "throughput": 1135.76, "total_tokens": 37694472}
{"current_steps": 5670, "total_steps": 16324, "loss": 0.3156, "lr": 7.771899967985728e-05, "epoch": 0.6946188478147683, "percentage": 34.73, "elapsed_time": "9:14:08", "remaining_time": "17:21:15", "throughput": 1134.71, "total_tokens": 37727712}
{"current_steps": 5675, "total_steps": 16324, "loss": 0.3487, "lr": 7.76768329954626e-05, "epoch": 0.6952313864812716, "percentage": 34.76, "elapsed_time": "9:15:08", "remaining_time": "17:21:43", "throughput": 1133.64, "total_tokens": 37760528}
{"current_steps": 5680, "total_steps": 16324, "loss": 0.2953, "lr": 7.763463791223216e-05, "epoch": 0.6958439251477749, "percentage": 34.8, "elapsed_time": "9:16:09", "remaining_time": "17:22:11", "throughput": 1132.6, "total_tokens": 37793968}
{"current_steps": 5685, "total_steps": 16324, "loss": 0.3258, "lr": 7.759241447346175e-05, "epoch": 0.6964564638142783, "percentage": 34.83, "elapsed_time": "9:17:08", "remaining_time": "17:22:38", "throughput": 1131.59, "total_tokens": 37827352}
{"current_steps": 5690, "total_steps": 16324, "loss": 0.3724, "lr": 7.755016272247631e-05, "epoch": 0.6970690024807816, "percentage": 34.86, "elapsed_time": "9:18:08", "remaining_time": "17:23:06", "throughput": 1130.55, "total_tokens": 37860544}
{"current_steps": 5695, "total_steps": 16324, "loss": 0.2496, "lr": 7.75078827026298e-05, "epoch": 0.6976815411472849, "percentage": 34.89, "elapsed_time": "9:19:08", "remaining_time": "17:23:34", "throughput": 1129.53, "total_tokens": 37894560}
{"current_steps": 5700, "total_steps": 16324, "loss": 0.3001, "lr": 7.746557445730517e-05, "epoch": 0.6982940798137882, "percentage": 34.92, "elapsed_time": "9:20:08", "remaining_time": "17:24:01", "throughput": 1128.53, "total_tokens": 37928296}
{"current_steps": 5700, "total_steps": 16324, "eval_loss": 0.1474122703075409, "epoch": 0.6982940798137882, "percentage": 34.92, "elapsed_time": "9:20:28", "remaining_time": "17:24:38", "throughput": 1127.86, "total_tokens": 37928296}
{"current_steps": 5705, "total_steps": 16324, "loss": 0.262, "lr": 7.742323802991442e-05, "epoch": 0.6989066184802916, "percentage": 34.95, "elapsed_time": "9:21:34", "remaining_time": "17:25:16", "throughput": 1126.68, "total_tokens": 37962624}
{"current_steps": 5710, "total_steps": 16324, "loss": 0.2939, "lr": 7.738087346389835e-05, "epoch": 0.6995191571467949, "percentage": 34.98, "elapsed_time": "9:22:35", "remaining_time": "17:25:45", "throughput": 1125.64, "total_tokens": 37996520}
{"current_steps": 5715, "total_steps": 16324, "loss": 0.2748, "lr": 7.733848080272668e-05, "epoch": 0.7001316958132983, "percentage": 35.01, "elapsed_time": "9:23:34", "remaining_time": "17:26:12", "throughput": 1124.63, "total_tokens": 38029384}
{"current_steps": 5720, "total_steps": 16324, "loss": 0.3109, "lr": 7.729606008989801e-05, "epoch": 0.7007442344798015, "percentage": 35.04, "elapsed_time": "9:24:35", "remaining_time": "17:26:39", "throughput": 1123.62, "total_tokens": 38062808}
{"current_steps": 5725, "total_steps": 16324, "loss": 0.2942, "lr": 7.725361136893963e-05, "epoch": 0.7013567731463048, "percentage": 35.07, "elapsed_time": "9:25:35", "remaining_time": "17:27:06", "throughput": 1122.62, "total_tokens": 38096552}
{"current_steps": 5730, "total_steps": 16324, "loss": 0.2759, "lr": 7.721113468340766e-05, "epoch": 0.7019693118128082, "percentage": 35.1, "elapsed_time": "9:26:34", "remaining_time": "17:27:32", "throughput": 1121.66, "total_tokens": 38130696}
{"current_steps": 5735, "total_steps": 16324, "loss": 0.2912, "lr": 7.716863007688684e-05, "epoch": 0.7025818504793115, "percentage": 35.13, "elapsed_time": "9:27:34", "remaining_time": "17:27:58", "throughput": 1120.67, "total_tokens": 38164136}
{"current_steps": 5740, "total_steps": 16324, "loss": 0.2909, "lr": 7.71260975929906e-05, "epoch": 0.7031943891458148, "percentage": 35.16, "elapsed_time": "9:28:34", "remaining_time": "17:28:24", "throughput": 1119.68, "total_tokens": 38197608}
{"current_steps": 5745, "total_steps": 16324, "loss": 0.334, "lr": 7.708353727536097e-05, "epoch": 0.7038069278123181, "percentage": 35.19, "elapsed_time": "9:29:34", "remaining_time": "17:28:49", "throughput": 1118.7, "total_tokens": 38230808}
{"current_steps": 5750, "total_steps": 16324, "loss": 0.3218, "lr": 7.70409491676685e-05, "epoch": 0.7044194664788215, "percentage": 35.22, "elapsed_time": "9:30:33", "remaining_time": "17:29:13", "throughput": 1117.74, "total_tokens": 38264144}
{"current_steps": 5750, "total_steps": 16324, "eval_loss": 0.07439830899238586, "epoch": 0.7044194664788215, "percentage": 35.22, "elapsed_time": "9:30:52", "remaining_time": "17:29:49", "throughput": 1117.1, "total_tokens": 38264144}
{"current_steps": 5755, "total_steps": 16324, "loss": 0.3302, "lr": 7.699833331361234e-05, "epoch": 0.7050320051453248, "percentage": 35.25, "elapsed_time": "9:31:58", "remaining_time": "17:30:25", "throughput": 1115.95, "total_tokens": 38297728}
{"current_steps": 5760, "total_steps": 16324, "loss": 0.3029, "lr": 7.695568975692005e-05, "epoch": 0.7056445438118282, "percentage": 35.29, "elapsed_time": "9:32:57", "remaining_time": "17:30:49", "throughput": 1115.03, "total_tokens": 38332192}
{"current_steps": 5765, "total_steps": 16324, "loss": 0.2635, "lr": 7.69130185413476e-05, "epoch": 0.7062570824783314, "percentage": 35.32, "elapsed_time": "9:33:57", "remaining_time": "17:31:14", "throughput": 1114.07, "total_tokens": 38365608}
{"current_steps": 5770, "total_steps": 16324, "loss": 0.2867, "lr": 7.687031971067937e-05, "epoch": 0.7068696211448348, "percentage": 35.35, "elapsed_time": "9:34:57", "remaining_time": "17:31:39", "throughput": 1113.11, "total_tokens": 38398936}
{"current_steps": 5775, "total_steps": 16324, "loss": 0.3174, "lr": 7.682759330872809e-05, "epoch": 0.7074821598113381, "percentage": 35.38, "elapsed_time": "9:35:56", "remaining_time": "17:32:03", "throughput": 1112.15, "total_tokens": 38432144}
{"current_steps": 5780, "total_steps": 16324, "loss": 0.3089, "lr": 7.678483937933474e-05, "epoch": 0.7080946984778415, "percentage": 35.41, "elapsed_time": "9:36:56", "remaining_time": "17:32:27", "throughput": 1111.22, "total_tokens": 38466128}
{"current_steps": 5785, "total_steps": 16324, "loss": 0.3194, "lr": 7.674205796636858e-05, "epoch": 0.7087072371443447, "percentage": 35.44, "elapsed_time": "9:37:55", "remaining_time": "17:32:51", "throughput": 1110.27, "total_tokens": 38499408}
{"current_steps": 5790, "total_steps": 16324, "loss": 0.248, "lr": 7.669924911372706e-05, "epoch": 0.709319775810848, "percentage": 35.47, "elapsed_time": "9:38:55", "remaining_time": "17:33:16", "throughput": 1109.32, "total_tokens": 38533184}
{"current_steps": 5795, "total_steps": 16324, "loss": 0.2899, "lr": 7.665641286533578e-05, "epoch": 0.7099323144773514, "percentage": 35.5, "elapsed_time": "9:39:55", "remaining_time": "17:33:40", "throughput": 1108.38, "total_tokens": 38567016}
{"current_steps": 5800, "total_steps": 16324, "loss": 0.2762, "lr": 7.661354926514843e-05, "epoch": 0.7105448531438547, "percentage": 35.53, "elapsed_time": "9:40:55", "remaining_time": "17:34:05", "throughput": 1107.46, "total_tokens": 38601128}
{"current_steps": 5800, "total_steps": 16324, "eval_loss": 0.2444901168346405, "epoch": 0.7105448531438547, "percentage": 35.53, "elapsed_time": "9:41:15", "remaining_time": "17:34:40", "throughput": 1106.84, "total_tokens": 38601128}
{"current_steps": 5805, "total_steps": 16324, "loss": 0.2591, "lr": 7.657065835714683e-05, "epoch": 0.711157391810358, "percentage": 35.56, "elapsed_time": "9:42:19", "remaining_time": "17:35:13", "throughput": 1105.76, "total_tokens": 38635176}
{"current_steps": 5810, "total_steps": 16324, "loss": 0.2323, "lr": 7.652774018534076e-05, "epoch": 0.7117699304768613, "percentage": 35.59, "elapsed_time": "9:43:19", "remaining_time": "17:35:36", "throughput": 1104.86, "total_tokens": 38669304}
{"current_steps": 5815, "total_steps": 16324, "loss": 0.3004, "lr": 7.648479479376797e-05, "epoch": 0.7123824691433647, "percentage": 35.62, "elapsed_time": "9:44:19", "remaining_time": "17:35:59", "throughput": 1103.94, "total_tokens": 38703192}
{"current_steps": 5820, "total_steps": 16324, "loss": 0.2882, "lr": 7.64418222264942e-05, "epoch": 0.712995007809868, "percentage": 35.65, "elapsed_time": "9:45:18", "remaining_time": "17:36:22", "throughput": 1103.02, "total_tokens": 38736688}
{"current_steps": 5825, "total_steps": 16324, "loss": 0.2939, "lr": 7.639882252761302e-05, "epoch": 0.7136075464763714, "percentage": 35.68, "elapsed_time": "9:46:18", "remaining_time": "17:36:45", "throughput": 1102.1, "total_tokens": 38770184}
{"current_steps": 5830, "total_steps": 16324, "loss": 0.3033, "lr": 7.635579574124586e-05, "epoch": 0.7142200851428746, "percentage": 35.71, "elapsed_time": "9:47:17", "remaining_time": "17:37:08", "throughput": 1101.2, "total_tokens": 38803864}
{"current_steps": 5835, "total_steps": 16324, "loss": 0.312, "lr": 7.631274191154194e-05, "epoch": 0.714832623809378, "percentage": 35.74, "elapsed_time": "9:48:18", "remaining_time": "17:37:32", "throughput": 1100.26, "total_tokens": 38837440}
{"current_steps": 5840, "total_steps": 16324, "loss": 0.3425, "lr": 7.626966108267823e-05, "epoch": 0.7154451624758813, "percentage": 35.78, "elapsed_time": "9:49:18", "remaining_time": "17:37:56", "throughput": 1099.33, "total_tokens": 38871192}
{"current_steps": 5845, "total_steps": 16324, "loss": 0.338, "lr": 7.622655329885943e-05, "epoch": 0.7160577011423847, "percentage": 35.81, "elapsed_time": "9:50:18", "remaining_time": "17:38:19", "throughput": 1098.42, "total_tokens": 38904744}
{"current_steps": 5850, "total_steps": 16324, "loss": 0.2624, "lr": 7.618341860431783e-05, "epoch": 0.7166702398088879, "percentage": 35.84, "elapsed_time": "9:51:19", "remaining_time": "17:38:43", "throughput": 1097.52, "total_tokens": 38939296}
{"current_steps": 5850, "total_steps": 16324, "eval_loss": 0.1622983068227768, "epoch": 0.7166702398088879, "percentage": 35.84, "elapsed_time": "9:51:39", "remaining_time": "17:39:18", "throughput": 1096.91, "total_tokens": 38939296}
{"current_steps": 5855, "total_steps": 16324, "loss": 0.3184, "lr": 7.614025704331342e-05, "epoch": 0.7172827784753912, "percentage": 35.87, "elapsed_time": "9:52:43", "remaining_time": "17:39:49", "throughput": 1095.85, "total_tokens": 38972184}
{"current_steps": 5860, "total_steps": 16324, "loss": 0.3129, "lr": 7.609706866013368e-05, "epoch": 0.7178953171418946, "percentage": 35.9, "elapsed_time": "9:53:42", "remaining_time": "17:40:09", "throughput": 1094.96, "total_tokens": 39005072}
{"current_steps": 5865, "total_steps": 16324, "loss": 0.329, "lr": 7.605385349909362e-05, "epoch": 0.7185078558083979, "percentage": 35.93, "elapsed_time": "9:54:42", "remaining_time": "17:40:32", "throughput": 1094.06, "total_tokens": 39039112}
{"current_steps": 5870, "total_steps": 16324, "loss": 0.3301, "lr": 7.601061160453581e-05, "epoch": 0.7191203944749013, "percentage": 35.96, "elapsed_time": "9:55:41", "remaining_time": "17:40:53", "throughput": 1093.19, "total_tokens": 39072768}
{"current_steps": 5875, "total_steps": 16324, "loss": 0.3299, "lr": 7.596734302083016e-05, "epoch": 0.7197329331414045, "percentage": 35.99, "elapsed_time": "9:56:41", "remaining_time": "17:41:14", "throughput": 1092.32, "total_tokens": 39106416}
{"current_steps": 5880, "total_steps": 16324, "loss": 0.2706, "lr": 7.592404779237401e-05, "epoch": 0.7203454718079079, "percentage": 36.02, "elapsed_time": "9:57:41", "remaining_time": "17:41:36", "throughput": 1091.44, "total_tokens": 39140168}
{"current_steps": 5885, "total_steps": 16324, "loss": 0.2823, "lr": 7.588072596359201e-05, "epoch": 0.7209580104744112, "percentage": 36.05, "elapsed_time": "9:58:40", "remaining_time": "17:41:56", "throughput": 1090.58, "total_tokens": 39174160}
{"current_steps": 5890, "total_steps": 16324, "loss": 0.2834, "lr": 7.583737757893612e-05, "epoch": 0.7215705491409146, "percentage": 36.08, "elapsed_time": "9:59:39", "remaining_time": "17:42:17", "throughput": 1089.72, "total_tokens": 39207944}
{"current_steps": 5895, "total_steps": 16324, "loss": 0.2995, "lr": 7.579400268288557e-05, "epoch": 0.7221830878074178, "percentage": 36.11, "elapsed_time": "10:00:39", "remaining_time": "17:42:39", "throughput": 1088.84, "total_tokens": 39241768}
{"current_steps": 5900, "total_steps": 16324, "loss": 0.3096, "lr": 7.575060131994676e-05, "epoch": 0.7227956264739211, "percentage": 36.14, "elapsed_time": "10:01:39", "remaining_time": "17:43:00", "throughput": 1087.97, "total_tokens": 39275432}
{"current_steps": 5900, "total_steps": 16324, "eval_loss": 0.20010535418987274, "epoch": 0.7227956264739211, "percentage": 36.14, "elapsed_time": "10:01:59", "remaining_time": "17:43:35", "throughput": 1087.37, "total_tokens": 39275432}
{"current_steps": 5905, "total_steps": 16324, "loss": 0.2899, "lr": 7.570717353465327e-05, "epoch": 0.7234081651404245, "percentage": 36.17, "elapsed_time": "10:03:04", "remaining_time": "17:44:04", "throughput": 1086.35, "total_tokens": 39308840}
{"current_steps": 5910, "total_steps": 16324, "loss": 0.3417, "lr": 7.566371937156574e-05, "epoch": 0.7240207038069278, "percentage": 36.2, "elapsed_time": "10:04:03", "remaining_time": "17:44:25", "throughput": 1085.49, "total_tokens": 39342200}
{"current_steps": 5915, "total_steps": 16324, "loss": 0.3779, "lr": 7.562023887527198e-05, "epoch": 0.7246332424734311, "percentage": 36.23, "elapsed_time": "10:05:03", "remaining_time": "17:44:46", "throughput": 1084.61, "total_tokens": 39375392}
{"current_steps": 5920, "total_steps": 16324, "loss": 0.2938, "lr": 7.557673209038672e-05, "epoch": 0.7252457811399344, "percentage": 36.27, "elapsed_time": "10:06:03", "remaining_time": "17:45:06", "throughput": 1083.76, "total_tokens": 39409016}
{"current_steps": 5925, "total_steps": 16324, "loss": 0.2468, "lr": 7.553319906155169e-05, "epoch": 0.7258583198064378, "percentage": 36.3, "elapsed_time": "10:07:03", "remaining_time": "17:45:26", "throughput": 1082.93, "total_tokens": 39443520}
{"current_steps": 5930, "total_steps": 16324, "loss": 0.2442, "lr": 7.548963983343562e-05, "epoch": 0.7264708584729411, "percentage": 36.33, "elapsed_time": "10:08:03", "remaining_time": "17:45:47", "throughput": 1082.05, "total_tokens": 39477320}
{"current_steps": 5935, "total_steps": 16324, "loss": 0.2802, "lr": 7.544605445073401e-05, "epoch": 0.7270833971394445, "percentage": 36.36, "elapsed_time": "10:09:03", "remaining_time": "17:46:08", "throughput": 1081.21, "total_tokens": 39511200}
{"current_steps": 5940, "total_steps": 16324, "loss": 0.3173, "lr": 7.540244295816926e-05, "epoch": 0.7276959358059477, "percentage": 36.39, "elapsed_time": "10:10:03", "remaining_time": "17:46:29", "throughput": 1080.32, "total_tokens": 39544072}
{"current_steps": 5945, "total_steps": 16324, "loss": 0.2781, "lr": 7.535880540049054e-05, "epoch": 0.7283084744724511, "percentage": 36.42, "elapsed_time": "10:11:04", "remaining_time": "17:46:50", "throughput": 1079.46, "total_tokens": 39577952}
{"current_steps": 5950, "total_steps": 16324, "loss": 0.2848, "lr": 7.531514182247382e-05, "epoch": 0.7289210131389544, "percentage": 36.45, "elapsed_time": "10:12:05", "remaining_time": "17:47:11", "throughput": 1078.6, "total_tokens": 39611784}
{"current_steps": 5950, "total_steps": 16324, "eval_loss": 0.30868977308273315, "epoch": 0.7289210131389544, "percentage": 36.45, "elapsed_time": "10:12:24", "remaining_time": "17:47:45", "throughput": 1078.02, "total_tokens": 39611784}
{"current_steps": 5960, "total_steps": 16324, "loss": 0.2649, "lr": 7.522773678466341e-05, "epoch": 0.730146090471961, "percentage": 36.51, "elapsed_time": "10:14:31", "remaining_time": "17:48:37", "throughput": 1076.13, "total_tokens": 39678616}
{"current_steps": 5965, "total_steps": 16324, "loss": 0.2404, "lr": 7.51839954145549e-05, "epoch": 0.7307586291384643, "percentage": 36.54, "elapsed_time": "10:15:31", "remaining_time": "17:48:57", "throughput": 1075.3, "total_tokens": 39713016}
{"current_steps": 5970, "total_steps": 16324, "loss": 0.3474, "lr": 7.514022820347862e-05, "epoch": 0.7313711678049677, "percentage": 36.57, "elapsed_time": "10:16:31", "remaining_time": "17:49:16", "throughput": 1074.47, "total_tokens": 39746720}
{"current_steps": 5975, "total_steps": 16324, "loss": 0.3241, "lr": 7.509643519634348e-05, "epoch": 0.731983706471471, "percentage": 36.6, "elapsed_time": "10:17:32", "remaining_time": "17:49:35", "throughput": 1073.63, "total_tokens": 39780176}
{"current_steps": 5980, "total_steps": 16324, "loss": 0.3367, "lr": 7.505261643808497e-05, "epoch": 0.7325962451379744, "percentage": 36.63, "elapsed_time": "10:18:31", "remaining_time": "17:49:53", "throughput": 1072.82, "total_tokens": 39813608}
{"current_steps": 5985, "total_steps": 16324, "loss": 0.2756, "lr": 7.500877197366492e-05, "epoch": 0.7332087838044776, "percentage": 36.66, "elapsed_time": "10:19:30", "remaining_time": "17:50:11", "throughput": 1072.02, "total_tokens": 39847256}
{"current_steps": 5990, "total_steps": 16324, "loss": 0.2507, "lr": 7.496490184807152e-05, "epoch": 0.733821322470981, "percentage": 36.69, "elapsed_time": "10:20:30", "remaining_time": "17:50:30", "throughput": 1071.19, "total_tokens": 39880848}
{"current_steps": 5995, "total_steps": 16324, "loss": 0.3189, "lr": 7.492100610631939e-05, "epoch": 0.7344338611374843, "percentage": 36.73, "elapsed_time": "10:21:29", "remaining_time": "17:50:47", "throughput": 1070.38, "total_tokens": 39913792}
{"current_steps": 6000, "total_steps": 16324, "loss": 0.3025, "lr": 7.487708479344934e-05, "epoch": 0.7350463998039877, "percentage": 36.76, "elapsed_time": "10:22:28", "remaining_time": "17:51:05", "throughput": 1069.57, "total_tokens": 39947544}
{"current_steps": 6000, "total_steps": 16324, "eval_loss": 0.14401155710220337, "epoch": 0.7350463998039877, "percentage": 36.76, "elapsed_time": "10:22:48", "remaining_time": "17:51:38", "throughput": 1069.02, "total_tokens": 39947544}
{"current_steps": 6005, "total_steps": 16324, "loss": 0.2161, "lr": 7.483313795452848e-05, "epoch": 0.7356589384704909, "percentage": 36.79, "elapsed_time": "10:23:53", "remaining_time": "17:52:06", "throughput": 1068.08, "total_tokens": 39982296}
{"current_steps": 6010, "total_steps": 16324, "loss": 0.2991, "lr": 7.478916563465007e-05, "epoch": 0.7362714771369943, "percentage": 36.82, "elapsed_time": "10:24:52", "remaining_time": "17:52:23", "throughput": 1067.27, "total_tokens": 40015240}
{"current_steps": 6015, "total_steps": 16324, "loss": 0.2634, "lr": 7.474516787893354e-05, "epoch": 0.7368840158034976, "percentage": 36.85, "elapsed_time": "10:25:52", "remaining_time": "17:52:40", "throughput": 1066.47, "total_tokens": 40048800}
{"current_steps": 6020, "total_steps": 16324, "loss": 0.3008, "lr": 7.470114473252439e-05, "epoch": 0.737496554470001, "percentage": 36.88, "elapsed_time": "10:26:52", "remaining_time": "17:52:58", "throughput": 1065.66, "total_tokens": 40081872}
{"current_steps": 6025, "total_steps": 16324, "loss": 0.2665, "lr": 7.46570962405942e-05, "epoch": 0.7381090931365042, "percentage": 36.91, "elapsed_time": "10:27:51", "remaining_time": "17:53:15", "throughput": 1064.86, "total_tokens": 40115376}
{"current_steps": 6030, "total_steps": 16324, "loss": 0.2847, "lr": 7.461302244834058e-05, "epoch": 0.7387216318030075, "percentage": 36.94, "elapsed_time": "10:28:51", "remaining_time": "17:53:32", "throughput": 1064.07, "total_tokens": 40148936}
{"current_steps": 6035, "total_steps": 16324, "loss": 0.2553, "lr": 7.456892340098703e-05, "epoch": 0.7393341704695109, "percentage": 36.97, "elapsed_time": "10:29:50", "remaining_time": "17:53:49", "throughput": 1063.3, "total_tokens": 40182880}
{"current_steps": 6040, "total_steps": 16324, "loss": 0.3273, "lr": 7.452479914378304e-05, "epoch": 0.7399467091360142, "percentage": 37.0, "elapsed_time": "10:30:50", "remaining_time": "17:54:06", "throughput": 1062.5, "total_tokens": 40216432}
{"current_steps": 6045, "total_steps": 16324, "loss": 0.3363, "lr": 7.448064972200392e-05, "epoch": 0.7405592478025176, "percentage": 37.03, "elapsed_time": "10:31:50", "remaining_time": "17:54:23", "throughput": 1061.71, "total_tokens": 40250064}
{"current_steps": 6050, "total_steps": 16324, "loss": 0.2845, "lr": 7.443647518095078e-05, "epoch": 0.7411717864690208, "percentage": 37.06, "elapsed_time": "10:32:50", "remaining_time": "17:54:40", "throughput": 1060.94, "total_tokens": 40284288}
{"current_steps": 6050, "total_steps": 16324, "eval_loss": 0.26300135254859924, "epoch": 0.7411717864690208, "percentage": 37.06, "elapsed_time": "10:33:10", "remaining_time": "17:55:14", "throughput": 1060.39, "total_tokens": 40284288}
{"current_steps": 6055, "total_steps": 16324, "loss": 0.3481, "lr": 7.43922755659506e-05, "epoch": 0.7417843251355242, "percentage": 37.09, "elapsed_time": "10:34:15", "remaining_time": "17:55:40", "throughput": 1059.44, "total_tokens": 40317688}
{"current_steps": 6060, "total_steps": 16324, "loss": 0.2952, "lr": 7.434805092235599e-05, "epoch": 0.7423968638020275, "percentage": 37.12, "elapsed_time": "10:35:15", "remaining_time": "17:55:57", "throughput": 1058.64, "total_tokens": 40350808}
{"current_steps": 6065, "total_steps": 16324, "loss": 0.2918, "lr": 7.430380129554527e-05, "epoch": 0.7430094024685309, "percentage": 37.15, "elapsed_time": "10:36:16", "remaining_time": "17:56:15", "throughput": 1057.85, "total_tokens": 40384432}
{"current_steps": 6070, "total_steps": 16324, "loss": 0.3835, "lr": 7.42595267309224e-05, "epoch": 0.7436219411350341, "percentage": 37.18, "elapsed_time": "10:37:16", "remaining_time": "17:56:33", "throughput": 1057.05, "total_tokens": 40418328}
{"current_steps": 6075, "total_steps": 16324, "loss": 0.3326, "lr": 7.421522727391696e-05, "epoch": 0.7442344798015375, "percentage": 37.22, "elapsed_time": "10:38:17", "remaining_time": "17:56:50", "throughput": 1056.25, "total_tokens": 40451488}
{"current_steps": 6080, "total_steps": 16324, "loss": 0.2993, "lr": 7.417090296998404e-05, "epoch": 0.7448470184680408, "percentage": 37.25, "elapsed_time": "10:39:17", "remaining_time": "17:57:07", "throughput": 1055.48, "total_tokens": 40485496}
{"current_steps": 6085, "total_steps": 16324, "loss": 0.3238, "lr": 7.412655386460419e-05, "epoch": 0.7454595571345441, "percentage": 37.28, "elapsed_time": "10:40:18", "remaining_time": "17:57:25", "throughput": 1054.66, "total_tokens": 40518408}
{"current_steps": 6090, "total_steps": 16324, "loss": 0.2752, "lr": 7.408218000328349e-05, "epoch": 0.7460720958010474, "percentage": 37.31, "elapsed_time": "10:41:18", "remaining_time": "17:57:40", "throughput": 1053.93, "total_tokens": 40553264}
{"current_steps": 6095, "total_steps": 16324, "loss": 0.3469, "lr": 7.403778143155336e-05, "epoch": 0.7466846344675507, "percentage": 37.34, "elapsed_time": "10:42:17", "remaining_time": "17:57:55", "throughput": 1053.2, "total_tokens": 40587264}
{"current_steps": 6100, "total_steps": 16324, "loss": 0.3172, "lr": 7.399335819497057e-05, "epoch": 0.7472971731340541, "percentage": 37.37, "elapsed_time": "10:43:17", "remaining_time": "17:58:11", "throughput": 1052.4, "total_tokens": 40619872}
{"current_steps": 6100, "total_steps": 16324, "eval_loss": 0.16394594311714172, "epoch": 0.7472971731340541, "percentage": 37.37, "elapsed_time": "10:43:37", "remaining_time": "17:58:45", "throughput": 1051.86, "total_tokens": 40619872}
{"current_steps": 6105, "total_steps": 16324, "loss": 0.2465, "lr": 7.394891033911724e-05, "epoch": 0.7479097118005574, "percentage": 37.4, "elapsed_time": "10:44:41", "remaining_time": "17:59:08", "throughput": 1050.98, "total_tokens": 40653952}
{"current_steps": 6110, "total_steps": 16324, "loss": 0.3163, "lr": 7.390443790960077e-05, "epoch": 0.7485222504670608, "percentage": 37.43, "elapsed_time": "10:45:41", "remaining_time": "17:59:23", "throughput": 1050.23, "total_tokens": 40687256}
{"current_steps": 6115, "total_steps": 16324, "loss": 0.286, "lr": 7.385994095205366e-05, "epoch": 0.749134789133564, "percentage": 37.46, "elapsed_time": "10:46:41", "remaining_time": "17:59:39", "throughput": 1049.48, "total_tokens": 40721512}
{"current_steps": 6120, "total_steps": 16324, "loss": 0.2814, "lr": 7.381541951213373e-05, "epoch": 0.7497473278000674, "percentage": 37.49, "elapsed_time": "10:47:40", "remaining_time": "17:59:52", "throughput": 1048.75, "total_tokens": 40754824}
{"current_steps": 6125, "total_steps": 16324, "loss": 0.2902, "lr": 7.37708736355238e-05, "epoch": 0.7503598664665707, "percentage": 37.52, "elapsed_time": "10:48:39", "remaining_time": "18:00:06", "throughput": 1048.02, "total_tokens": 40788144}
{"current_steps": 6130, "total_steps": 16324, "loss": 0.2628, "lr": 7.372630336793183e-05, "epoch": 0.7509724051330741, "percentage": 37.55, "elapsed_time": "10:49:39", "remaining_time": "18:00:20", "throughput": 1047.28, "total_tokens": 40821984}
{"current_steps": 6135, "total_steps": 16324, "loss": 0.2101, "lr": 7.368170875509078e-05, "epoch": 0.7515849437995773, "percentage": 37.58, "elapsed_time": "10:50:38", "remaining_time": "18:00:34", "throughput": 1046.57, "total_tokens": 40856448}
{"current_steps": 6140, "total_steps": 16324, "loss": 0.2821, "lr": 7.36370898427586e-05, "epoch": 0.7521974824660806, "percentage": 37.61, "elapsed_time": "10:51:37", "remaining_time": "18:00:48", "throughput": 1045.85, "total_tokens": 40889960}
{"current_steps": 6145, "total_steps": 16324, "loss": 0.2582, "lr": 7.359244667671818e-05, "epoch": 0.752810021132584, "percentage": 37.64, "elapsed_time": "10:52:37", "remaining_time": "18:01:03", "throughput": 1045.12, "total_tokens": 40924176}
{"current_steps": 6150, "total_steps": 16324, "loss": 0.2963, "lr": 7.354777930277725e-05, "epoch": 0.7534225597990873, "percentage": 37.67, "elapsed_time": "10:53:36", "remaining_time": "18:01:16", "throughput": 1044.36, "total_tokens": 40956544}
{"current_steps": 6150, "total_steps": 16324, "eval_loss": 0.15747645497322083, "epoch": 0.7534225597990873, "percentage": 37.67, "elapsed_time": "10:53:56", "remaining_time": "18:01:49", "throughput": 1043.83, "total_tokens": 40956544}
{"current_steps": 6155, "total_steps": 16324, "loss": 0.2898, "lr": 7.350308776676848e-05, "epoch": 0.7540350984655907, "percentage": 37.71, "elapsed_time": "10:55:01", "remaining_time": "18:02:12", "throughput": 1042.98, "total_tokens": 40991016}
{"current_steps": 6160, "total_steps": 16324, "loss": 0.3118, "lr": 7.345837211454922e-05, "epoch": 0.7546476371320939, "percentage": 37.74, "elapsed_time": "10:56:01", "remaining_time": "18:02:26", "throughput": 1042.26, "total_tokens": 41024800}
{"current_steps": 6165, "total_steps": 16324, "loss": 0.2695, "lr": 7.341363239200164e-05, "epoch": 0.7552601757985973, "percentage": 37.77, "elapsed_time": "10:57:01", "remaining_time": "18:02:40", "throughput": 1041.52, "total_tokens": 41058208}
{"current_steps": 6170, "total_steps": 16324, "loss": 0.3063, "lr": 7.336886864503256e-05, "epoch": 0.7558727144651006, "percentage": 37.8, "elapsed_time": "10:58:01", "remaining_time": "18:02:55", "throughput": 1040.79, "total_tokens": 41092048}
{"current_steps": 6175, "total_steps": 16324, "loss": 0.2352, "lr": 7.33240809195735e-05, "epoch": 0.756485253131604, "percentage": 37.83, "elapsed_time": "10:59:01", "remaining_time": "18:03:08", "throughput": 1040.06, "total_tokens": 41125720}
{"current_steps": 6180, "total_steps": 16324, "loss": 0.2885, "lr": 7.327926926158054e-05, "epoch": 0.7570977917981072, "percentage": 37.86, "elapsed_time": "11:00:01", "remaining_time": "18:03:22", "throughput": 1039.35, "total_tokens": 41159528}
{"current_steps": 6185, "total_steps": 16324, "loss": 0.3619, "lr": 7.323443371703434e-05, "epoch": 0.7577103304646106, "percentage": 37.89, "elapsed_time": "11:01:00", "remaining_time": "18:03:35", "throughput": 1038.62, "total_tokens": 41192496}
{"current_steps": 6190, "total_steps": 16324, "loss": 0.2902, "lr": 7.318957433194009e-05, "epoch": 0.7583228691311139, "percentage": 37.92, "elapsed_time": "11:02:00", "remaining_time": "18:03:49", "throughput": 1037.9, "total_tokens": 41226168}
{"current_steps": 6195, "total_steps": 16324, "loss": 0.3258, "lr": 7.314469115232736e-05, "epoch": 0.7589354077976173, "percentage": 37.95, "elapsed_time": "11:03:01", "remaining_time": "18:04:03", "throughput": 1037.14, "total_tokens": 41258896}
{"current_steps": 6200, "total_steps": 16324, "loss": 0.307, "lr": 7.309978422425026e-05, "epoch": 0.7595479464641205, "percentage": 37.98, "elapsed_time": "11:04:01", "remaining_time": "18:04:17", "throughput": 1036.4, "total_tokens": 41291792}
{"current_steps": 6200, "total_steps": 16324, "eval_loss": 0.1483473926782608, "epoch": 0.7595479464641205, "percentage": 37.98, "elapsed_time": "11:04:21", "remaining_time": "18:04:49", "throughput": 1035.88, "total_tokens": 41291792}
{"current_steps": 6205, "total_steps": 16324, "loss": 0.3005, "lr": 7.305485359378715e-05, "epoch": 0.7601604851306238, "percentage": 38.01, "elapsed_time": "11:05:26", "remaining_time": "18:05:10", "throughput": 1035.04, "total_tokens": 41325280}
{"current_steps": 6210, "total_steps": 16324, "loss": 0.261, "lr": 7.300989930704077e-05, "epoch": 0.7607730237971272, "percentage": 38.04, "elapsed_time": "11:06:26", "remaining_time": "18:05:23", "throughput": 1034.33, "total_tokens": 41358808}
{"current_steps": 6215, "total_steps": 16324, "loss": 0.3242, "lr": 7.296492141013816e-05, "epoch": 0.7613855624636305, "percentage": 38.07, "elapsed_time": "11:07:25", "remaining_time": "18:05:35", "throughput": 1033.64, "total_tokens": 41392496}
{"current_steps": 6220, "total_steps": 16324, "loss": 0.3319, "lr": 7.29199199492305e-05, "epoch": 0.7619981011301339, "percentage": 38.1, "elapsed_time": "11:08:24", "remaining_time": "18:05:47", "throughput": 1032.97, "total_tokens": 41426512}
{"current_steps": 6225, "total_steps": 16324, "loss": 0.2954, "lr": 7.287489497049324e-05, "epoch": 0.7626106397966371, "percentage": 38.13, "elapsed_time": "11:09:24", "remaining_time": "18:05:59", "throughput": 1032.26, "total_tokens": 41460136}
{"current_steps": 6230, "total_steps": 16324, "loss": 0.3331, "lr": 7.28298465201259e-05, "epoch": 0.7632231784631405, "percentage": 38.16, "elapsed_time": "11:10:23", "remaining_time": "18:06:11", "throughput": 1031.56, "total_tokens": 41493096}
{"current_steps": 6235, "total_steps": 16324, "loss": 0.3114, "lr": 7.278477464435212e-05, "epoch": 0.7638357171296438, "percentage": 38.2, "elapsed_time": "11:11:22", "remaining_time": "18:06:22", "throughput": 1030.87, "total_tokens": 41526376}
{"current_steps": 6240, "total_steps": 16324, "loss": 0.2628, "lr": 7.273967938941954e-05, "epoch": 0.7644482557961472, "percentage": 38.23, "elapsed_time": "11:12:23", "remaining_time": "18:06:35", "throughput": 1030.17, "total_tokens": 41560264}
{"current_steps": 6245, "total_steps": 16324, "loss": 0.2767, "lr": 7.269456080159983e-05, "epoch": 0.7650607944626504, "percentage": 38.26, "elapsed_time": "11:13:22", "remaining_time": "18:06:46", "throughput": 1029.47, "total_tokens": 41593080}
{"current_steps": 6250, "total_steps": 16324, "loss": 0.2854, "lr": 7.264941892718857e-05, "epoch": 0.7656733331291538, "percentage": 38.29, "elapsed_time": "11:14:21", "remaining_time": "18:06:58", "throughput": 1028.77, "total_tokens": 41626144}
{"current_steps": 6250, "total_steps": 16324, "eval_loss": 0.08869815617799759, "epoch": 0.7656733331291538, "percentage": 38.29, "elapsed_time": "11:14:41", "remaining_time": "18:07:29", "throughput": 1028.28, "total_tokens": 41626144}
{"current_steps": 6255, "total_steps": 16324, "loss": 0.2942, "lr": 7.260425381250526e-05, "epoch": 0.7662858717956571, "percentage": 38.32, "elapsed_time": "11:15:46", "remaining_time": "18:07:50", "throughput": 1027.44, "total_tokens": 41659576}
{"current_steps": 6260, "total_steps": 16324, "loss": 0.2992, "lr": 7.255906550389321e-05, "epoch": 0.7668984104621605, "percentage": 38.35, "elapsed_time": "11:16:46", "remaining_time": "18:08:01", "throughput": 1026.76, "total_tokens": 41693032}
{"current_steps": 6265, "total_steps": 16324, "loss": 0.265, "lr": 7.251385404771956e-05, "epoch": 0.7675109491286638, "percentage": 38.38, "elapsed_time": "11:17:45", "remaining_time": "18:08:11", "throughput": 1026.11, "total_tokens": 41726952}
{"current_steps": 6270, "total_steps": 16324, "loss": 0.33, "lr": 7.24686194903752e-05, "epoch": 0.768123487795167, "percentage": 38.41, "elapsed_time": "11:18:45", "remaining_time": "18:08:23", "throughput": 1025.41, "total_tokens": 41760288}
{"current_steps": 6275, "total_steps": 16324, "loss": 0.2968, "lr": 7.242336187827472e-05, "epoch": 0.7687360264616704, "percentage": 38.44, "elapsed_time": "11:19:44", "remaining_time": "18:08:34", "throughput": 1024.74, "total_tokens": 41793904}
{"current_steps": 6280, "total_steps": 16324, "loss": 0.2806, "lr": 7.237808125785635e-05, "epoch": 0.7693485651281737, "percentage": 38.47, "elapsed_time": "11:20:43", "remaining_time": "18:08:43", "throughput": 1024.09, "total_tokens": 41827464}
{"current_steps": 6285, "total_steps": 16324, "loss": 0.3241, "lr": 7.233277767558194e-05, "epoch": 0.7699611037946771, "percentage": 38.5, "elapsed_time": "11:21:43", "remaining_time": "18:08:54", "throughput": 1023.42, "total_tokens": 41861128}
{"current_steps": 6290, "total_steps": 16324, "loss": 0.2861, "lr": 7.228745117793692e-05, "epoch": 0.7705736424611803, "percentage": 38.53, "elapsed_time": "11:22:43", "remaining_time": "18:09:05", "throughput": 1022.74, "total_tokens": 41894512}
{"current_steps": 6295, "total_steps": 16324, "loss": 0.3877, "lr": 7.224210181143022e-05, "epoch": 0.7711861811276837, "percentage": 38.56, "elapsed_time": "11:23:42", "remaining_time": "18:09:15", "throughput": 1022.07, "total_tokens": 41927648}
{"current_steps": 6300, "total_steps": 16324, "loss": 0.2569, "lr": 7.21967296225942e-05, "epoch": 0.771798719794187, "percentage": 38.59, "elapsed_time": "11:24:42", "remaining_time": "18:09:26", "throughput": 1021.4, "total_tokens": 41961432}
{"current_steps": 6300, "total_steps": 16324, "eval_loss": 0.08859598636627197, "epoch": 0.771798719794187, "percentage": 38.59, "elapsed_time": "11:25:01", "remaining_time": "18:09:57", "throughput": 1020.92, "total_tokens": 41961432}
{"current_steps": 6305, "total_steps": 16324, "loss": 0.3425, "lr": 7.21513346579847e-05, "epoch": 0.7724112584606904, "percentage": 38.62, "elapsed_time": "11:26:07", "remaining_time": "18:10:18", "throughput": 1020.07, "total_tokens": 41994280}
{"current_steps": 6310, "total_steps": 16324, "loss": 0.3086, "lr": 7.210591696418086e-05, "epoch": 0.7730237971271936, "percentage": 38.65, "elapsed_time": "11:27:07", "remaining_time": "18:10:28", "throughput": 1019.41, "total_tokens": 42027712}
{"current_steps": 6315, "total_steps": 16324, "loss": 0.2838, "lr": 7.206047658778521e-05, "epoch": 0.773636335793697, "percentage": 38.69, "elapsed_time": "11:28:07", "remaining_time": "18:10:38", "throughput": 1018.73, "total_tokens": 42060632}
{"current_steps": 6320, "total_steps": 16324, "loss": 0.2823, "lr": 7.201501357542352e-05, "epoch": 0.7742488744602003, "percentage": 38.72, "elapsed_time": "11:29:07", "remaining_time": "18:10:48", "throughput": 1018.06, "total_tokens": 42093888}
{"current_steps": 6325, "total_steps": 16324, "loss": 0.3035, "lr": 7.19695279737448e-05, "epoch": 0.7748614131267036, "percentage": 38.75, "elapsed_time": "11:30:06", "remaining_time": "18:10:58", "throughput": 1017.42, "total_tokens": 42127968}
{"current_steps": 6330, "total_steps": 16324, "loss": 0.3127, "lr": 7.192401982942122e-05, "epoch": 0.775473951793207, "percentage": 38.78, "elapsed_time": "11:31:05", "remaining_time": "18:11:07", "throughput": 1016.78, "total_tokens": 42161584}
{"current_steps": 6335, "total_steps": 16324, "loss": 0.2813, "lr": 7.187848918914808e-05, "epoch": 0.7760864904597102, "percentage": 38.81, "elapsed_time": "11:32:05", "remaining_time": "18:11:16", "throughput": 1016.14, "total_tokens": 42195320}
{"current_steps": 6340, "total_steps": 16324, "loss": 0.3069, "lr": 7.18329360996438e-05, "epoch": 0.7766990291262136, "percentage": 38.84, "elapsed_time": "11:33:04", "remaining_time": "18:11:25", "throughput": 1015.49, "total_tokens": 42228600}
{"current_steps": 6345, "total_steps": 16324, "loss": 0.3452, "lr": 7.178736060764977e-05, "epoch": 0.7773115677927169, "percentage": 38.87, "elapsed_time": "11:34:03", "remaining_time": "18:11:33", "throughput": 1014.87, "total_tokens": 42262592}
{"current_steps": 6350, "total_steps": 16324, "loss": 0.2468, "lr": 7.174176275993046e-05, "epoch": 0.7779241064592203, "percentage": 38.9, "elapsed_time": "11:35:03", "remaining_time": "18:11:44", "throughput": 1014.21, "total_tokens": 42296360}
{"current_steps": 6350, "total_steps": 16324, "eval_loss": 0.1457025557756424, "epoch": 0.7779241064592203, "percentage": 38.9, "elapsed_time": "11:35:23", "remaining_time": "18:12:14", "throughput": 1013.74, "total_tokens": 42296360}
{"current_steps": 6355, "total_steps": 16324, "loss": 0.2664, "lr": 7.169614260327317e-05, "epoch": 0.7785366451257235, "percentage": 38.93, "elapsed_time": "11:36:27", "remaining_time": "18:12:31", "throughput": 1012.99, "total_tokens": 42330296}
{"current_steps": 6360, "total_steps": 16324, "loss": 0.2861, "lr": 7.165050018448818e-05, "epoch": 0.7791491837922269, "percentage": 38.96, "elapsed_time": "11:37:26", "remaining_time": "18:12:39", "throughput": 1012.37, "total_tokens": 42364048}
{"current_steps": 6365, "total_steps": 16324, "loss": 0.3407, "lr": 7.160483555040855e-05, "epoch": 0.7797617224587302, "percentage": 38.99, "elapsed_time": "11:38:26", "remaining_time": "18:12:49", "throughput": 1011.71, "total_tokens": 42397600}
{"current_steps": 6370, "total_steps": 16324, "loss": 0.2991, "lr": 7.15591487478902e-05, "epoch": 0.7803742611252336, "percentage": 39.02, "elapsed_time": "11:39:26", "remaining_time": "18:12:59", "throughput": 1011.04, "total_tokens": 42430264}
{"current_steps": 6375, "total_steps": 16324, "loss": 0.3304, "lr": 7.151343982381172e-05, "epoch": 0.7809867997917369, "percentage": 39.05, "elapsed_time": "11:40:26", "remaining_time": "18:13:07", "throughput": 1010.39, "total_tokens": 42463064}
{"current_steps": 6380, "total_steps": 16324, "loss": 0.294, "lr": 7.146770882507448e-05, "epoch": 0.7815993384582401, "percentage": 39.08, "elapsed_time": "11:41:26", "remaining_time": "18:13:17", "throughput": 1009.74, "total_tokens": 42497032}
{"current_steps": 6385, "total_steps": 16324, "loss": 0.3075, "lr": 7.142195579860242e-05, "epoch": 0.7822118771247435, "percentage": 39.11, "elapsed_time": "11:42:26", "remaining_time": "18:13:26", "throughput": 1009.1, "total_tokens": 42530448}
{"current_steps": 6390, "total_steps": 16324, "loss": 0.2973, "lr": 7.137618079134213e-05, "epoch": 0.7828244157912468, "percentage": 39.14, "elapsed_time": "11:43:26", "remaining_time": "18:13:35", "throughput": 1008.46, "total_tokens": 42563872}
{"current_steps": 6395, "total_steps": 16324, "loss": 0.2532, "lr": 7.133038385026276e-05, "epoch": 0.7834369544577502, "percentage": 39.18, "elapsed_time": "11:44:27", "remaining_time": "18:13:45", "throughput": 1007.79, "total_tokens": 42597152}
{"current_steps": 6400, "total_steps": 16324, "loss": 0.2472, "lr": 7.128456502235595e-05, "epoch": 0.7840494931242534, "percentage": 39.21, "elapsed_time": "11:45:27", "remaining_time": "18:13:54", "throughput": 1007.17, "total_tokens": 42631184}
{"current_steps": 6400, "total_steps": 16324, "eval_loss": 0.12907207012176514, "epoch": 0.7840494931242534, "percentage": 39.21, "elapsed_time": "11:45:47", "remaining_time": "18:14:24", "throughput": 1006.7, "total_tokens": 42631184}
{"current_steps": 6405, "total_steps": 16324, "loss": 0.2744, "lr": 7.123872435463577e-05, "epoch": 0.7846620317907568, "percentage": 39.24, "elapsed_time": "11:46:52", "remaining_time": "18:14:41", "throughput": 1005.95, "total_tokens": 42664864}
{"current_steps": 6410, "total_steps": 16324, "loss": 0.3276, "lr": 7.119286189413877e-05, "epoch": 0.7852745704572601, "percentage": 39.27, "elapsed_time": "11:47:52", "remaining_time": "18:14:49", "throughput": 1005.31, "total_tokens": 42697720}
{"current_steps": 6415, "total_steps": 16324, "loss": 0.3092, "lr": 7.114697768792378e-05, "epoch": 0.7858871091237635, "percentage": 39.3, "elapsed_time": "11:48:52", "remaining_time": "18:14:58", "throughput": 1004.67, "total_tokens": 42731136}
{"current_steps": 6420, "total_steps": 16324, "loss": 0.2763, "lr": 7.110107178307205e-05, "epoch": 0.7864996477902667, "percentage": 39.33, "elapsed_time": "11:49:52", "remaining_time": "18:15:06", "throughput": 1004.04, "total_tokens": 42764792}
{"current_steps": 6425, "total_steps": 16324, "loss": 0.2399, "lr": 7.105514422668694e-05, "epoch": 0.7871121864567701, "percentage": 39.36, "elapsed_time": "11:50:52", "remaining_time": "18:15:14", "throughput": 1003.45, "total_tokens": 42799336}
{"current_steps": 6430, "total_steps": 16324, "loss": 0.2724, "lr": 7.100919506589418e-05, "epoch": 0.7877247251232734, "percentage": 39.39, "elapsed_time": "11:51:52", "remaining_time": "18:15:23", "throughput": 1002.82, "total_tokens": 42833272}
{"current_steps": 6435, "total_steps": 16324, "loss": 0.2987, "lr": 7.096322434784155e-05, "epoch": 0.7883372637897768, "percentage": 39.42, "elapsed_time": "11:52:52", "remaining_time": "18:15:31", "throughput": 1002.2, "total_tokens": 42866792}
{"current_steps": 6440, "total_steps": 16324, "loss": 0.2642, "lr": 7.091723211969906e-05, "epoch": 0.7889498024562801, "percentage": 39.45, "elapsed_time": "11:53:52", "remaining_time": "18:15:38", "throughput": 1001.57, "total_tokens": 42899928}
{"current_steps": 6445, "total_steps": 16324, "loss": 0.24, "lr": 7.087121842865866e-05, "epoch": 0.7895623411227833, "percentage": 39.48, "elapsed_time": "11:54:53", "remaining_time": "18:15:47", "throughput": 1000.95, "total_tokens": 42934056}
{"current_steps": 6450, "total_steps": 16324, "loss": 0.2823, "lr": 7.082518332193447e-05, "epoch": 0.7901748797892867, "percentage": 39.51, "elapsed_time": "11:55:52", "remaining_time": "18:15:54", "throughput": 1000.35, "total_tokens": 42967536}
{"current_steps": 6450, "total_steps": 16324, "eval_loss": 0.08559667319059372, "epoch": 0.7901748797892867, "percentage": 39.51, "elapsed_time": "11:56:11", "remaining_time": "18:16:22", "throughput": 999.91, "total_tokens": 42967536}
{"current_steps": 6455, "total_steps": 16324, "loss": 0.2871, "lr": 7.077912684676247e-05, "epoch": 0.79078741845579, "percentage": 39.54, "elapsed_time": "11:57:15", "remaining_time": "18:16:36", "throughput": 999.19, "total_tokens": 43000680}
{"current_steps": 6460, "total_steps": 16324, "loss": 0.3005, "lr": 7.073304905040061e-05, "epoch": 0.7913999571222934, "percentage": 39.57, "elapsed_time": "11:58:14", "remaining_time": "18:16:42", "throughput": 998.59, "total_tokens": 43033896}
{"current_steps": 6465, "total_steps": 16324, "loss": 0.2582, "lr": 7.068694998012872e-05, "epoch": 0.7920124957887966, "percentage": 39.6, "elapsed_time": "11:59:13", "remaining_time": "18:16:48", "throughput": 998.01, "total_tokens": 43067696}
{"current_steps": 6470, "total_steps": 16324, "loss": 0.2367, "lr": 7.064082968324846e-05, "epoch": 0.7926250344553, "percentage": 39.63, "elapsed_time": "12:00:12", "remaining_time": "18:16:53", "throughput": 997.43, "total_tokens": 43101232}
{"current_steps": 6475, "total_steps": 16324, "loss": 0.2661, "lr": 7.059468820708324e-05, "epoch": 0.7932375731218033, "percentage": 39.67, "elapsed_time": "12:01:11", "remaining_time": "18:16:59", "throughput": 996.85, "total_tokens": 43135256}
{"current_steps": 6480, "total_steps": 16324, "loss": 0.2757, "lr": 7.054852559897822e-05, "epoch": 0.7938501117883067, "percentage": 39.7, "elapsed_time": "12:02:10", "remaining_time": "18:17:05", "throughput": 996.26, "total_tokens": 43168880}
{"current_steps": 6485, "total_steps": 16324, "loss": 0.2786, "lr": 7.050234190630028e-05, "epoch": 0.7944626504548099, "percentage": 39.73, "elapsed_time": "12:03:09", "remaining_time": "18:17:10", "throughput": 995.68, "total_tokens": 43202464}
{"current_steps": 6490, "total_steps": 16324, "loss": 0.2872, "lr": 7.045613717643788e-05, "epoch": 0.7950751891213133, "percentage": 39.76, "elapsed_time": "12:04:09", "remaining_time": "18:17:17", "throughput": 995.08, "total_tokens": 43235952}
{"current_steps": 6495, "total_steps": 16324, "loss": 0.3142, "lr": 7.040991145680109e-05, "epoch": 0.7956877277878166, "percentage": 39.79, "elapsed_time": "12:05:08", "remaining_time": "18:17:22", "throughput": 994.49, "total_tokens": 43268736}
{"current_steps": 6500, "total_steps": 16324, "loss": 0.2839, "lr": 7.036366479482153e-05, "epoch": 0.79630026645432, "percentage": 39.82, "elapsed_time": "12:06:07", "remaining_time": "18:17:26", "throughput": 993.92, "total_tokens": 43302272}
{"current_steps": 6500, "total_steps": 16324, "eval_loss": 0.2148544043302536, "epoch": 0.79630026645432, "percentage": 39.82, "elapsed_time": "12:06:26", "remaining_time": "18:17:55", "throughput": 993.48, "total_tokens": 43302272}
{"current_steps": 6505, "total_steps": 16324, "loss": 0.2997, "lr": 7.031739723795227e-05, "epoch": 0.7969128051208233, "percentage": 39.85, "elapsed_time": "12:07:31", "remaining_time": "18:18:09", "throughput": 992.77, "total_tokens": 43335592}
{"current_steps": 6510, "total_steps": 16324, "loss": 0.2643, "lr": 7.027110883366787e-05, "epoch": 0.7975253437873265, "percentage": 39.88, "elapsed_time": "12:08:30", "remaining_time": "18:18:14", "throughput": 992.19, "total_tokens": 43369112}
{"current_steps": 6515, "total_steps": 16324, "loss": 0.2879, "lr": 7.022479962946422e-05, "epoch": 0.7981378824538299, "percentage": 39.91, "elapsed_time": "12:09:29", "remaining_time": "18:18:19", "throughput": 991.61, "total_tokens": 43402568}
{"current_steps": 6520, "total_steps": 16324, "loss": 0.3314, "lr": 7.017846967285862e-05, "epoch": 0.7987504211203332, "percentage": 39.94, "elapsed_time": "12:10:29", "remaining_time": "18:18:26", "throughput": 991.02, "total_tokens": 43436112}
{"current_steps": 6525, "total_steps": 16324, "loss": 0.2422, "lr": 7.013211901138964e-05, "epoch": 0.7993629597868366, "percentage": 39.97, "elapsed_time": "12:11:30", "remaining_time": "18:18:33", "throughput": 990.43, "total_tokens": 43470496}
{"current_steps": 6530, "total_steps": 16324, "loss": 0.4454, "lr": 7.008574769261704e-05, "epoch": 0.7999754984533398, "percentage": 40.0, "elapsed_time": "12:12:31", "remaining_time": "18:18:39", "throughput": 989.82, "total_tokens": 43503760}
{"current_steps": 6535, "total_steps": 16324, "loss": 0.318, "lr": 7.003935576412185e-05, "epoch": 0.8005880371198432, "percentage": 40.03, "elapsed_time": "12:13:32", "remaining_time": "18:18:48", "throughput": 989.18, "total_tokens": 43536552}
{"current_steps": 6540, "total_steps": 16324, "loss": 0.2918, "lr": 6.999294327350625e-05, "epoch": 0.8012005757863465, "percentage": 40.06, "elapsed_time": "12:14:34", "remaining_time": "18:18:56", "throughput": 988.55, "total_tokens": 43569608}
{"current_steps": 6545, "total_steps": 16324, "loss": 0.2728, "lr": 6.994651026839344e-05, "epoch": 0.8018131144528499, "percentage": 40.09, "elapsed_time": "12:15:35", "remaining_time": "18:19:03", "throughput": 987.94, "total_tokens": 43603048}
{"current_steps": 6550, "total_steps": 16324, "loss": 0.2764, "lr": 6.990005679642773e-05, "epoch": 0.8024256531193532, "percentage": 40.12, "elapsed_time": "12:16:36", "remaining_time": "18:19:10", "throughput": 987.34, "total_tokens": 43636808}
{"current_steps": 6550, "total_steps": 16324, "eval_loss": 0.15134289860725403, "epoch": 0.8024256531193532, "percentage": 40.12, "elapsed_time": "12:16:56", "remaining_time": "18:19:40", "throughput": 986.9, "total_tokens": 43636808}
{"current_steps": 6555, "total_steps": 16324, "loss": 0.2778, "lr": 6.985358290527443e-05, "epoch": 0.8030381917858564, "percentage": 40.16, "elapsed_time": "12:18:03", "remaining_time": "18:19:56", "throughput": 986.15, "total_tokens": 43670288}
{"current_steps": 6560, "total_steps": 16324, "loss": 0.2343, "lr": 6.980708864261979e-05, "epoch": 0.8036507304523598, "percentage": 40.19, "elapsed_time": "12:19:04", "remaining_time": "18:20:02", "throughput": 985.55, "total_tokens": 43703216}
{"current_steps": 6565, "total_steps": 16324, "loss": 0.2788, "lr": 6.976057405617096e-05, "epoch": 0.8042632691188631, "percentage": 40.22, "elapsed_time": "12:20:04", "remaining_time": "18:20:07", "throughput": 984.98, "total_tokens": 43737192}
{"current_steps": 6570, "total_steps": 16324, "loss": 0.2827, "lr": 6.971403919365596e-05, "epoch": 0.8048758077853665, "percentage": 40.25, "elapsed_time": "12:21:05", "remaining_time": "18:20:14", "throughput": 984.37, "total_tokens": 43770280}
{"current_steps": 6575, "total_steps": 16324, "loss": 0.2085, "lr": 6.966748410282358e-05, "epoch": 0.8054883464518697, "percentage": 40.28, "elapsed_time": "12:22:04", "remaining_time": "18:20:17", "throughput": 983.83, "total_tokens": 43804504}
{"current_steps": 6580, "total_steps": 16324, "loss": 0.3103, "lr": 6.96209088314434e-05, "epoch": 0.8061008851183731, "percentage": 40.31, "elapsed_time": "12:23:03", "remaining_time": "18:20:21", "throughput": 983.26, "total_tokens": 43837480}
{"current_steps": 6585, "total_steps": 16324, "loss": 0.258, "lr": 6.95743134273057e-05, "epoch": 0.8067134237848764, "percentage": 40.34, "elapsed_time": "12:24:03", "remaining_time": "18:20:26", "throughput": 982.7, "total_tokens": 43871144}
{"current_steps": 6590, "total_steps": 16324, "loss": 0.2682, "lr": 6.95276979382214e-05, "epoch": 0.8073259624513798, "percentage": 40.37, "elapsed_time": "12:25:02", "remaining_time": "18:20:30", "throughput": 982.14, "total_tokens": 43904576}
{"current_steps": 6595, "total_steps": 16324, "loss": 0.3714, "lr": 6.948106241202206e-05, "epoch": 0.807938501117883, "percentage": 40.4, "elapsed_time": "12:26:02", "remaining_time": "18:20:34", "throughput": 981.58, "total_tokens": 43938152}
{"current_steps": 6600, "total_steps": 16324, "loss": 0.308, "lr": 6.943440689655978e-05, "epoch": 0.8085510397843864, "percentage": 40.43, "elapsed_time": "12:27:02", "remaining_time": "18:20:38", "throughput": 981.01, "total_tokens": 43971808}
{"current_steps": 6600, "total_steps": 16324, "eval_loss": 0.19454015791416168, "epoch": 0.8085510397843864, "percentage": 40.43, "elapsed_time": "12:27:22", "remaining_time": "18:21:07", "throughput": 980.59, "total_tokens": 43971808}
{"current_steps": 6605, "total_steps": 16324, "loss": 0.322, "lr": 6.938773143970716e-05, "epoch": 0.8091635784508897, "percentage": 40.46, "elapsed_time": "12:28:27", "remaining_time": "18:21:19", "throughput": 979.91, "total_tokens": 44005224}
{"current_steps": 6610, "total_steps": 16324, "loss": 0.296, "lr": 6.934103608935729e-05, "epoch": 0.8097761171173931, "percentage": 40.49, "elapsed_time": "12:29:26", "remaining_time": "18:21:23", "throughput": 979.34, "total_tokens": 44037992}
{"current_steps": 6615, "total_steps": 16324, "loss": 0.2729, "lr": 6.929432089342365e-05, "epoch": 0.8103886557838964, "percentage": 40.52, "elapsed_time": "12:30:26", "remaining_time": "18:21:26", "throughput": 978.8, "total_tokens": 44071872}
{"current_steps": 6620, "total_steps": 16324, "loss": 0.2848, "lr": 6.92475858998401e-05, "epoch": 0.8110011944503996, "percentage": 40.55, "elapsed_time": "12:31:26", "remaining_time": "18:21:30", "throughput": 978.23, "total_tokens": 44104968}
{"current_steps": 6625, "total_steps": 16324, "loss": 0.2987, "lr": 6.92008311565608e-05, "epoch": 0.811613733116903, "percentage": 40.58, "elapsed_time": "12:32:25", "remaining_time": "18:21:33", "throughput": 977.7, "total_tokens": 44139184}
{"current_steps": 6630, "total_steps": 16324, "loss": 0.3034, "lr": 6.915405671156014e-05, "epoch": 0.8122262717834063, "percentage": 40.62, "elapsed_time": "12:33:25", "remaining_time": "18:21:37", "throughput": 977.14, "total_tokens": 44172288}
{"current_steps": 6635, "total_steps": 16324, "loss": 0.3226, "lr": 6.910726261283283e-05, "epoch": 0.8128388104499097, "percentage": 40.65, "elapsed_time": "12:34:25", "remaining_time": "18:21:40", "throughput": 976.58, "total_tokens": 44205384}
{"current_steps": 6640, "total_steps": 16324, "loss": 0.2152, "lr": 6.906044890839365e-05, "epoch": 0.8134513491164129, "percentage": 40.68, "elapsed_time": "12:35:24", "remaining_time": "18:21:43", "throughput": 976.05, "total_tokens": 44239384}
{"current_steps": 6645, "total_steps": 16324, "loss": 0.2887, "lr": 6.901361564627753e-05, "epoch": 0.8140638877829163, "percentage": 40.71, "elapsed_time": "12:36:25", "remaining_time": "18:21:48", "throughput": 975.47, "total_tokens": 44272696}
{"current_steps": 6650, "total_steps": 16324, "loss": 0.2873, "lr": 6.896676287453948e-05, "epoch": 0.8146764264494196, "percentage": 40.74, "elapsed_time": "12:37:26", "remaining_time": "18:21:53", "throughput": 974.9, "total_tokens": 44306360}
{"current_steps": 6650, "total_steps": 16324, "eval_loss": 0.13924284279346466, "epoch": 0.8146764264494196, "percentage": 40.74, "elapsed_time": "12:37:46", "remaining_time": "18:22:21", "throughput": 974.48, "total_tokens": 44306360}
{"current_steps": 6655, "total_steps": 16324, "loss": 0.251, "lr": 6.891989064125447e-05, "epoch": 0.815288965115923, "percentage": 40.77, "elapsed_time": "12:38:52", "remaining_time": "18:22:34", "throughput": 973.82, "total_tokens": 44340600}
{"current_steps": 6660, "total_steps": 16324, "loss": 0.2395, "lr": 6.887299899451754e-05, "epoch": 0.8159015037824263, "percentage": 40.8, "elapsed_time": "12:39:54", "remaining_time": "18:22:39", "throughput": 973.24, "total_tokens": 44374144}
{"current_steps": 6665, "total_steps": 16324, "loss": 0.2767, "lr": 6.882608798244358e-05, "epoch": 0.8165140424489296, "percentage": 40.83, "elapsed_time": "12:40:55", "remaining_time": "18:22:44", "throughput": 972.66, "total_tokens": 44407408}
{"current_steps": 6670, "total_steps": 16324, "loss": 0.287, "lr": 6.877915765316735e-05, "epoch": 0.8171265811154329, "percentage": 40.86, "elapsed_time": "12:41:57", "remaining_time": "18:22:50", "throughput": 972.09, "total_tokens": 44441208}
{"current_steps": 6675, "total_steps": 16324, "loss": 0.2508, "lr": 6.873220805484343e-05, "epoch": 0.8177391197819363, "percentage": 40.89, "elapsed_time": "12:42:59", "remaining_time": "18:22:55", "throughput": 971.5, "total_tokens": 44474368}
{"current_steps": 6680, "total_steps": 16324, "loss": 0.2729, "lr": 6.868523923564625e-05, "epoch": 0.8183516584484396, "percentage": 40.92, "elapsed_time": "12:44:01", "remaining_time": "18:23:02", "throughput": 970.9, "total_tokens": 44507824}
{"current_steps": 6685, "total_steps": 16324, "loss": 0.2981, "lr": 6.863825124376984e-05, "epoch": 0.8189641971149428, "percentage": 40.95, "elapsed_time": "12:45:02", "remaining_time": "18:23:05", "throughput": 970.35, "total_tokens": 44541288}
{"current_steps": 6690, "total_steps": 16324, "loss": 0.228, "lr": 6.859124412742798e-05, "epoch": 0.8195767357814462, "percentage": 40.98, "elapsed_time": "12:46:01", "remaining_time": "18:23:08", "throughput": 969.82, "total_tokens": 44575072}
{"current_steps": 6695, "total_steps": 16324, "loss": 0.2697, "lr": 6.854421793485408e-05, "epoch": 0.8201892744479495, "percentage": 41.01, "elapsed_time": "12:47:02", "remaining_time": "18:23:10", "throughput": 969.29, "total_tokens": 44608856}
{"current_steps": 6700, "total_steps": 16324, "loss": 0.2768, "lr": 6.849717271430108e-05, "epoch": 0.8208018131144529, "percentage": 41.04, "elapsed_time": "12:48:01", "remaining_time": "18:23:12", "throughput": 968.76, "total_tokens": 44641896}
{"current_steps": 6700, "total_steps": 16324, "eval_loss": 0.09934918582439423, "epoch": 0.8208018131144529, "percentage": 41.04, "elapsed_time": "12:48:21", "remaining_time": "18:23:40", "throughput": 968.34, "total_tokens": 44641896}
{"current_steps": 6705, "total_steps": 16324, "loss": 0.2711, "lr": 6.845010851404146e-05, "epoch": 0.8214143517809561, "percentage": 41.07, "elapsed_time": "12:49:26", "remaining_time": "18:23:50", "throughput": 967.7, "total_tokens": 44675200}
{"current_steps": 6710, "total_steps": 16324, "loss": 0.2986, "lr": 6.840302538236718e-05, "epoch": 0.8220268904474595, "percentage": 41.11, "elapsed_time": "12:50:26", "remaining_time": "18:23:52", "throughput": 967.17, "total_tokens": 44708840}
{"current_steps": 6715, "total_steps": 16324, "loss": 0.3317, "lr": 6.835592336758963e-05, "epoch": 0.8226394291139628, "percentage": 41.14, "elapsed_time": "12:51:25", "remaining_time": "18:23:53", "throughput": 966.65, "total_tokens": 44742240}
{"current_steps": 6720, "total_steps": 16324, "loss": 0.2862, "lr": 6.830880251803955e-05, "epoch": 0.8232519677804662, "percentage": 41.17, "elapsed_time": "12:52:25", "remaining_time": "18:23:55", "throughput": 966.12, "total_tokens": 44775232}
{"current_steps": 6725, "total_steps": 16324, "loss": 0.288, "lr": 6.826166288206708e-05, "epoch": 0.8238645064469695, "percentage": 41.2, "elapsed_time": "12:53:25", "remaining_time": "18:23:57", "throughput": 965.61, "total_tokens": 44809216}
{"current_steps": 6730, "total_steps": 16324, "loss": 0.2734, "lr": 6.821450450804154e-05, "epoch": 0.8244770451134728, "percentage": 41.23, "elapsed_time": "12:54:24", "remaining_time": "18:23:58", "throughput": 965.09, "total_tokens": 44842656}
{"current_steps": 6735, "total_steps": 16324, "loss": 0.2868, "lr": 6.816732744435153e-05, "epoch": 0.8250895837799761, "percentage": 41.26, "elapsed_time": "12:55:23", "remaining_time": "18:23:58", "throughput": 964.58, "total_tokens": 44876144}
{"current_steps": 6740, "total_steps": 16324, "loss": 0.2909, "lr": 6.812013173940482e-05, "epoch": 0.8257021224464794, "percentage": 41.29, "elapsed_time": "12:56:23", "remaining_time": "18:24:00", "throughput": 964.05, "total_tokens": 44908984}
{"current_steps": 6745, "total_steps": 16324, "loss": 0.3024, "lr": 6.807291744162829e-05, "epoch": 0.8263146611129828, "percentage": 41.32, "elapsed_time": "12:57:23", "remaining_time": "18:24:00", "throughput": 963.54, "total_tokens": 44942704}
{"current_steps": 6750, "total_steps": 16324, "loss": 0.2599, "lr": 6.802568459946792e-05, "epoch": 0.826927199779486, "percentage": 41.35, "elapsed_time": "12:58:22", "remaining_time": "18:24:00", "throughput": 963.04, "total_tokens": 44976072}
{"current_steps": 6750, "total_steps": 16324, "eval_loss": 0.1032017320394516, "epoch": 0.826927199779486, "percentage": 41.35, "elapsed_time": "12:58:41", "remaining_time": "18:24:28", "throughput": 962.64, "total_tokens": 44976072}
{"current_steps": 6755, "total_steps": 16324, "loss": 0.2842, "lr": 6.797843326138871e-05, "epoch": 0.8275397384459894, "percentage": 41.38, "elapsed_time": "12:59:47", "remaining_time": "18:24:38", "throughput": 962.0, "total_tokens": 45009792}
{"current_steps": 6760, "total_steps": 16324, "loss": 0.2743, "lr": 6.793116347587467e-05, "epoch": 0.8281522771124927, "percentage": 41.41, "elapsed_time": "13:00:47", "remaining_time": "18:24:39", "throughput": 961.49, "total_tokens": 45043328}
{"current_steps": 6765, "total_steps": 16324, "loss": 0.3115, "lr": 6.788387529142865e-05, "epoch": 0.8287648157789961, "percentage": 41.44, "elapsed_time": "13:01:46", "remaining_time": "18:24:39", "throughput": 960.99, "total_tokens": 45076920}
{"current_steps": 6770, "total_steps": 16324, "loss": 0.2929, "lr": 6.783656875657248e-05, "epoch": 0.8293773544454994, "percentage": 41.47, "elapsed_time": "13:02:47", "remaining_time": "18:24:41", "throughput": 960.46, "total_tokens": 45110256}
{"current_steps": 6775, "total_steps": 16324, "loss": 0.3656, "lr": 6.778924391984676e-05, "epoch": 0.8299898931120027, "percentage": 41.5, "elapsed_time": "13:03:47", "remaining_time": "18:24:42", "throughput": 959.95, "total_tokens": 45143792}
{"current_steps": 6780, "total_steps": 16324, "loss": 0.261, "lr": 6.774190082981089e-05, "epoch": 0.830602431778506, "percentage": 41.53, "elapsed_time": "13:04:46", "remaining_time": "18:24:42", "throughput": 959.48, "total_tokens": 45178448}
{"current_steps": 6785, "total_steps": 16324, "loss": 0.3413, "lr": 6.7694539535043e-05, "epoch": 0.8312149704450094, "percentage": 41.56, "elapsed_time": "13:05:46", "remaining_time": "18:24:43", "throughput": 958.96, "total_tokens": 45211664}
{"current_steps": 6790, "total_steps": 16324, "loss": 0.2459, "lr": 6.764716008413988e-05, "epoch": 0.8318275091115127, "percentage": 41.6, "elapsed_time": "13:06:46", "remaining_time": "18:24:43", "throughput": 958.45, "total_tokens": 45245336}
{"current_steps": 6795, "total_steps": 16324, "loss": 0.2311, "lr": 6.759976252571698e-05, "epoch": 0.832440047778016, "percentage": 41.63, "elapsed_time": "13:07:45", "remaining_time": "18:24:43", "throughput": 957.97, "total_tokens": 45279552}
{"current_steps": 6800, "total_steps": 16324, "loss": 0.2864, "lr": 6.755234690840828e-05, "epoch": 0.8330525864445193, "percentage": 41.66, "elapsed_time": "13:08:45", "remaining_time": "18:24:43", "throughput": 957.49, "total_tokens": 45313600}
{"current_steps": 6800, "total_steps": 16324, "eval_loss": 0.11257417500019073, "epoch": 0.8330525864445193, "percentage": 41.66, "elapsed_time": "13:09:04", "remaining_time": "18:25:10", "throughput": 957.1, "total_tokens": 45313600}
{"current_steps": 6805, "total_steps": 16324, "loss": 0.2353, "lr": 6.750491328086636e-05, "epoch": 0.8336651251110226, "percentage": 41.69, "elapsed_time": "13:10:10", "remaining_time": "18:25:19", "throughput": 956.47, "total_tokens": 45346792}
{"current_steps": 6810, "total_steps": 16324, "loss": 0.2854, "lr": 6.745746169176224e-05, "epoch": 0.834277663777526, "percentage": 41.72, "elapsed_time": "13:11:08", "remaining_time": "18:25:17", "throughput": 956.0, "total_tokens": 45380272}
{"current_steps": 6815, "total_steps": 16324, "loss": 0.3025, "lr": 6.740999218978536e-05, "epoch": 0.8348902024440292, "percentage": 41.75, "elapsed_time": "13:12:07", "remaining_time": "18:25:15", "throughput": 955.51, "total_tokens": 45413232}
{"current_steps": 6820, "total_steps": 16324, "loss": 0.3557, "lr": 6.736250482364356e-05, "epoch": 0.8355027411105326, "percentage": 41.78, "elapsed_time": "13:13:07", "remaining_time": "18:25:15", "throughput": 955.01, "total_tokens": 45446816}
{"current_steps": 6825, "total_steps": 16324, "loss": 0.287, "lr": 6.731499964206301e-05, "epoch": 0.8361152797770359, "percentage": 41.81, "elapsed_time": "13:14:06", "remaining_time": "18:25:14", "throughput": 954.54, "total_tokens": 45480432}
{"current_steps": 6830, "total_steps": 16324, "loss": 0.2686, "lr": 6.726747669378815e-05, "epoch": 0.8367278184435393, "percentage": 41.84, "elapsed_time": "13:15:05", "remaining_time": "18:25:12", "throughput": 954.07, "total_tokens": 45514408}
{"current_steps": 6835, "total_steps": 16324, "loss": 0.3115, "lr": 6.721993602758164e-05, "epoch": 0.8373403571100426, "percentage": 41.87, "elapsed_time": "13:16:05", "remaining_time": "18:25:12", "throughput": 953.56, "total_tokens": 45547120}
{"current_steps": 6840, "total_steps": 16324, "loss": 0.349, "lr": 6.717237769222437e-05, "epoch": 0.8379528957765459, "percentage": 41.9, "elapsed_time": "13:17:04", "remaining_time": "18:25:10", "throughput": 953.1, "total_tokens": 45580944}
{"current_steps": 6845, "total_steps": 16324, "loss": 0.3055, "lr": 6.712480173651527e-05, "epoch": 0.8385654344430492, "percentage": 41.93, "elapsed_time": "13:18:03", "remaining_time": "18:25:09", "throughput": 952.61, "total_tokens": 45614296}
{"current_steps": 6850, "total_steps": 16324, "loss": 0.3468, "lr": 6.707720820927146e-05, "epoch": 0.8391779731095526, "percentage": 41.96, "elapsed_time": "13:19:02", "remaining_time": "18:25:08", "throughput": 952.13, "total_tokens": 45647976}
{"current_steps": 6850, "total_steps": 16324, "eval_loss": 0.11305823177099228, "epoch": 0.8391779731095526, "percentage": 41.96, "elapsed_time": "13:19:22", "remaining_time": "18:25:34", "throughput": 951.75, "total_tokens": 45647976}
{"current_steps": 6855, "total_steps": 16324, "loss": 0.3153, "lr": 6.702959715932802e-05, "epoch": 0.8397905117760559, "percentage": 41.99, "elapsed_time": "13:20:27", "remaining_time": "18:25:41", "throughput": 951.15, "total_tokens": 45680992}
{"current_steps": 6860, "total_steps": 16324, "loss": 0.3125, "lr": 6.698196863553799e-05, "epoch": 0.8404030504425591, "percentage": 42.02, "elapsed_time": "13:21:27", "remaining_time": "18:25:40", "throughput": 950.66, "total_tokens": 45714544}
{"current_steps": 6865, "total_steps": 16324, "loss": 0.2774, "lr": 6.693432268677244e-05, "epoch": 0.8410155891090625, "percentage": 42.05, "elapsed_time": "13:22:27", "remaining_time": "18:25:40", "throughput": 950.17, "total_tokens": 45748488}
{"current_steps": 6870, "total_steps": 16324, "loss": 0.2185, "lr": 6.688665936192019e-05, "epoch": 0.8416281277755658, "percentage": 42.09, "elapsed_time": "13:23:28", "remaining_time": "18:25:40", "throughput": 949.68, "total_tokens": 45782288}
{"current_steps": 6875, "total_steps": 16324, "loss": 0.2745, "lr": 6.6838978709888e-05, "epoch": 0.8422406664420692, "percentage": 42.12, "elapsed_time": "13:24:27", "remaining_time": "18:25:39", "throughput": 949.2, "total_tokens": 45815728}
{"current_steps": 6880, "total_steps": 16324, "loss": 0.2921, "lr": 6.679128077960031e-05, "epoch": 0.8428532051085725, "percentage": 42.15, "elapsed_time": "13:25:27", "remaining_time": "18:25:38", "throughput": 948.72, "total_tokens": 45849616}
{"current_steps": 6885, "total_steps": 16324, "loss": 0.2403, "lr": 6.67435656199994e-05, "epoch": 0.8434657437750758, "percentage": 42.18, "elapsed_time": "13:26:27", "remaining_time": "18:25:37", "throughput": 948.24, "total_tokens": 45883224}
{"current_steps": 6890, "total_steps": 16324, "loss": 0.3004, "lr": 6.669583328004511e-05, "epoch": 0.8440782824415791, "percentage": 42.21, "elapsed_time": "13:27:27", "remaining_time": "18:25:35", "throughput": 947.76, "total_tokens": 45916512}
{"current_steps": 6895, "total_steps": 16324, "loss": 0.2926, "lr": 6.664808380871503e-05, "epoch": 0.8446908211080825, "percentage": 42.24, "elapsed_time": "13:28:27", "remaining_time": "18:25:34", "throughput": 947.27, "total_tokens": 45950112}
{"current_steps": 6900, "total_steps": 16324, "loss": 0.2933, "lr": 6.660031725500424e-05, "epoch": 0.8453033597745858, "percentage": 42.27, "elapsed_time": "13:29:27", "remaining_time": "18:25:33", "throughput": 946.8, "total_tokens": 45983624}
{"current_steps": 6900, "total_steps": 16324, "eval_loss": 0.048528868705034256, "epoch": 0.8453033597745858, "percentage": 42.27, "elapsed_time": "13:29:47", "remaining_time": "18:26:00", "throughput": 946.41, "total_tokens": 45983624}
{"current_steps": 6905, "total_steps": 16324, "loss": 0.3146, "lr": 6.655253366792537e-05, "epoch": 0.845915898441089, "percentage": 42.3, "elapsed_time": "13:30:52", "remaining_time": "18:26:06", "throughput": 945.83, "total_tokens": 46017008}
{"current_steps": 6910, "total_steps": 16324, "loss": 0.2708, "lr": 6.650473309650855e-05, "epoch": 0.8465284371075924, "percentage": 42.33, "elapsed_time": "13:31:52", "remaining_time": "18:26:04", "throughput": 945.34, "total_tokens": 46050160}
{"current_steps": 6915, "total_steps": 16324, "loss": 0.2725, "lr": 6.645691558980127e-05, "epoch": 0.8471409757740958, "percentage": 42.36, "elapsed_time": "13:32:52", "remaining_time": "18:26:02", "throughput": 944.88, "total_tokens": 46083872}
{"current_steps": 6920, "total_steps": 16324, "loss": 0.3157, "lr": 6.640908119686852e-05, "epoch": 0.8477535144405991, "percentage": 42.39, "elapsed_time": "13:33:51", "remaining_time": "18:25:59", "throughput": 944.42, "total_tokens": 46117064}
{"current_steps": 6925, "total_steps": 16324, "loss": 0.3113, "lr": 6.636122996679249e-05, "epoch": 0.8483660531071023, "percentage": 42.42, "elapsed_time": "13:34:51", "remaining_time": "18:25:57", "throughput": 943.95, "total_tokens": 46150520}
{"current_steps": 6930, "total_steps": 16324, "loss": 0.2832, "lr": 6.631336194867271e-05, "epoch": 0.8489785917736057, "percentage": 42.45, "elapsed_time": "13:35:50", "remaining_time": "18:25:55", "throughput": 943.47, "total_tokens": 46183832}
{"current_steps": 6935, "total_steps": 16324, "loss": 0.29, "lr": 6.626547719162595e-05, "epoch": 0.849591130440109, "percentage": 42.48, "elapsed_time": "13:36:49", "remaining_time": "18:25:52", "throughput": 943.01, "total_tokens": 46217064}
{"current_steps": 6940, "total_steps": 16324, "loss": 0.2726, "lr": 6.621757574478611e-05, "epoch": 0.8502036691066124, "percentage": 42.51, "elapsed_time": "13:37:49", "remaining_time": "18:25:49", "throughput": 942.55, "total_tokens": 46250464}
{"current_steps": 6945, "total_steps": 16324, "loss": 0.2943, "lr": 6.616965765730426e-05, "epoch": 0.8508162077731157, "percentage": 42.54, "elapsed_time": "13:38:50", "remaining_time": "18:25:49", "throughput": 942.06, "total_tokens": 46284400}
{"current_steps": 6950, "total_steps": 16324, "loss": 0.3528, "lr": 6.612172297834849e-05, "epoch": 0.851428746439619, "percentage": 42.58, "elapsed_time": "13:39:50", "remaining_time": "18:25:47", "throughput": 941.6, "total_tokens": 46317728}
{"current_steps": 6950, "total_steps": 16324, "eval_loss": 0.05446135997772217, "epoch": 0.851428746439619, "percentage": 42.58, "elapsed_time": "13:40:10", "remaining_time": "18:26:14", "throughput": 941.22, "total_tokens": 46317728}
{"current_steps": 6955, "total_steps": 16324, "loss": 0.2848, "lr": 6.607377175710396e-05, "epoch": 0.8520412851061223, "percentage": 42.61, "elapsed_time": "13:41:15", "remaining_time": "18:26:18", "throughput": 940.66, "total_tokens": 46351912}
{"current_steps": 6960, "total_steps": 16324, "loss": 0.2825, "lr": 6.60258040427728e-05, "epoch": 0.8526538237726257, "percentage": 42.64, "elapsed_time": "13:42:16", "remaining_time": "18:26:16", "throughput": 940.2, "total_tokens": 46385576}
{"current_steps": 6965, "total_steps": 16324, "loss": 0.2624, "lr": 6.597781988457405e-05, "epoch": 0.853266362439129, "percentage": 42.67, "elapsed_time": "13:43:15", "remaining_time": "18:26:13", "throughput": 939.74, "total_tokens": 46418776}
{"current_steps": 6970, "total_steps": 16324, "loss": 0.3017, "lr": 6.59298193317436e-05, "epoch": 0.8538789011056322, "percentage": 42.7, "elapsed_time": "13:44:15", "remaining_time": "18:26:11", "throughput": 939.28, "total_tokens": 46452688}
{"current_steps": 6975, "total_steps": 16324, "loss": 0.2531, "lr": 6.588180243353424e-05, "epoch": 0.8544914397721356, "percentage": 42.73, "elapsed_time": "13:45:16", "remaining_time": "18:26:09", "throughput": 938.8, "total_tokens": 46485920}
{"current_steps": 6980, "total_steps": 16324, "loss": 0.2594, "lr": 6.583376923921542e-05, "epoch": 0.855103978438639, "percentage": 42.76, "elapsed_time": "13:46:16", "remaining_time": "18:26:07", "throughput": 938.35, "total_tokens": 46520152}
{"current_steps": 6985, "total_steps": 16324, "loss": 0.2598, "lr": 6.578571979807342e-05, "epoch": 0.8557165171051423, "percentage": 42.79, "elapsed_time": "13:47:16", "remaining_time": "18:26:04", "throughput": 937.89, "total_tokens": 46553776}
{"current_steps": 6990, "total_steps": 16324, "loss": 0.2809, "lr": 6.573765415941113e-05, "epoch": 0.8563290557716455, "percentage": 42.82, "elapsed_time": "13:48:17", "remaining_time": "18:26:02", "throughput": 937.41, "total_tokens": 46586840}
{"current_steps": 6995, "total_steps": 16324, "loss": 0.3272, "lr": 6.568957237254805e-05, "epoch": 0.8569415944381489, "percentage": 42.85, "elapsed_time": "13:49:17", "remaining_time": "18:25:59", "throughput": 936.95, "total_tokens": 46620216}
{"current_steps": 7000, "total_steps": 16324, "loss": 0.2581, "lr": 6.564147448682028e-05, "epoch": 0.8575541331046522, "percentage": 42.88, "elapsed_time": "13:50:17", "remaining_time": "18:25:56", "throughput": 936.5, "total_tokens": 46653728}
{"current_steps": 7000, "total_steps": 16324, "eval_loss": 0.1109340488910675, "epoch": 0.8575541331046522, "percentage": 42.88, "elapsed_time": "13:50:37", "remaining_time": "18:26:22", "throughput": 936.12, "total_tokens": 46653728}
{"current_steps": 7005, "total_steps": 16324, "loss": 0.2989, "lr": 6.559336055158045e-05, "epoch": 0.8581666717711556, "percentage": 42.91, "elapsed_time": "13:51:43", "remaining_time": "18:26:28", "throughput": 935.56, "total_tokens": 46687312}
{"current_steps": 7010, "total_steps": 16324, "loss": 0.2734, "lr": 6.554523061619764e-05, "epoch": 0.8587792104376589, "percentage": 42.94, "elapsed_time": "13:52:43", "remaining_time": "18:26:25", "throughput": 935.12, "total_tokens": 46721712}
{"current_steps": 7015, "total_steps": 16324, "loss": 0.2504, "lr": 6.549708473005731e-05, "epoch": 0.8593917491041622, "percentage": 42.97, "elapsed_time": "13:53:43", "remaining_time": "18:26:21", "throughput": 934.66, "total_tokens": 46754840}
{"current_steps": 7020, "total_steps": 16324, "loss": 0.2449, "lr": 6.544892294256135e-05, "epoch": 0.8600042877706655, "percentage": 43.0, "elapsed_time": "13:54:44", "remaining_time": "18:26:19", "throughput": 934.2, "total_tokens": 46788624}
{"current_steps": 7025, "total_steps": 16324, "loss": 0.288, "lr": 6.540074530312795e-05, "epoch": 0.8606168264371689, "percentage": 43.03, "elapsed_time": "13:55:44", "remaining_time": "18:26:16", "throughput": 933.73, "total_tokens": 46821304}
{"current_steps": 7030, "total_steps": 16324, "loss": 0.2344, "lr": 6.535255186119153e-05, "epoch": 0.8612293651036722, "percentage": 43.07, "elapsed_time": "13:56:43", "remaining_time": "18:26:11", "throughput": 933.3, "total_tokens": 46855192}
{"current_steps": 7035, "total_steps": 16324, "loss": 0.2126, "lr": 6.530434266620277e-05, "epoch": 0.8618419037701754, "percentage": 43.1, "elapsed_time": "13:57:43", "remaining_time": "18:26:08", "throughput": 932.87, "total_tokens": 46889632}
{"current_steps": 7040, "total_steps": 16324, "loss": 0.3162, "lr": 6.525611776762845e-05, "epoch": 0.8624544424366788, "percentage": 43.13, "elapsed_time": "13:58:42", "remaining_time": "18:26:02", "throughput": 932.44, "total_tokens": 46923080}
{"current_steps": 7045, "total_steps": 16324, "loss": 0.3463, "lr": 6.520787721495157e-05, "epoch": 0.8630669811031821, "percentage": 43.16, "elapsed_time": "13:59:41", "remaining_time": "18:25:57", "throughput": 932.02, "total_tokens": 46956576}
{"current_steps": 7050, "total_steps": 16324, "loss": 0.2281, "lr": 6.515962105767109e-05, "epoch": 0.8636795197696855, "percentage": 43.19, "elapsed_time": "14:00:40", "remaining_time": "18:25:53", "throughput": 931.59, "total_tokens": 46990264}
{"current_steps": 7050, "total_steps": 16324, "eval_loss": 0.09453262388706207, "epoch": 0.8636795197696855, "percentage": 43.19, "elapsed_time": "14:01:00", "remaining_time": "18:26:18", "throughput": 931.23, "total_tokens": 46990264}
{"current_steps": 7055, "total_steps": 16324, "loss": 0.2473, "lr": 6.511134934530203e-05, "epoch": 0.8642920584361888, "percentage": 43.22, "elapsed_time": "14:02:05", "remaining_time": "18:26:21", "throughput": 930.7, "total_tokens": 47024144}
{"current_steps": 7060, "total_steps": 16324, "loss": 0.25, "lr": 6.506306212737536e-05, "epoch": 0.8649045971026921, "percentage": 43.25, "elapsed_time": "14:03:04", "remaining_time": "18:26:16", "throughput": 930.28, "total_tokens": 47058016}
{"current_steps": 7065, "total_steps": 16324, "loss": 0.296, "lr": 6.501475945343796e-05, "epoch": 0.8655171357691954, "percentage": 43.28, "elapsed_time": "14:04:04", "remaining_time": "18:26:11", "throughput": 929.85, "total_tokens": 47091352}
{"current_steps": 7080, "total_steps": 16324, "loss": 0.2516, "lr": 6.486975919126783e-05, "epoch": 0.8673547517687054, "percentage": 43.37, "elapsed_time": "14:07:03", "remaining_time": "18:25:58", "throughput": 928.53, "total_tokens": 47191296}
{"current_steps": 7085, "total_steps": 16324, "loss": 0.2734, "lr": 6.482139518907276e-05, "epoch": 0.8679672904352087, "percentage": 43.4, "elapsed_time": "14:08:04", "remaining_time": "18:25:54", "throughput": 928.08, "total_tokens": 47225208}
{"current_steps": 7090, "total_steps": 16324, "loss": 0.2692, "lr": 6.477301597883823e-05, "epoch": 0.868579829101712, "percentage": 43.43, "elapsed_time": "14:09:04", "remaining_time": "18:25:50", "throughput": 927.65, "total_tokens": 47258616}
{"current_steps": 7095, "total_steps": 16324, "loss": 0.2194, "lr": 6.472462161020556e-05, "epoch": 0.8691923677682154, "percentage": 43.46, "elapsed_time": "14:10:04", "remaining_time": "18:25:45", "throughput": 927.22, "total_tokens": 47292712}
{"current_steps": 7100, "total_steps": 16324, "loss": 0.3067, "lr": 6.467621213283151e-05, "epoch": 0.8698049064347186, "percentage": 43.49, "elapsed_time": "14:11:05", "remaining_time": "18:25:41", "throughput": 926.78, "total_tokens": 47326520}
{"current_steps": 7100, "total_steps": 16324, "eval_loss": 0.1614220291376114, "epoch": 0.8698049064347186, "percentage": 43.49, "elapsed_time": "14:11:24", "remaining_time": "18:26:07", "throughput": 926.43, "total_tokens": 47326520}
{"current_steps": 7105, "total_steps": 16324, "loss": 0.3365, "lr": 6.462778759638846e-05, "epoch": 0.870417445101222, "percentage": 43.52, "elapsed_time": "14:12:30", "remaining_time": "18:26:09", "throughput": 925.89, "total_tokens": 47359416}
{"current_steps": 7110, "total_steps": 16324, "loss": 0.2462, "lr": 6.457934805056419e-05, "epoch": 0.8710299837677253, "percentage": 43.56, "elapsed_time": "14:13:30", "remaining_time": "18:26:04", "throughput": 925.47, "total_tokens": 47393568}
{"current_steps": 7115, "total_steps": 16324, "loss": 0.2549, "lr": 6.453089354506187e-05, "epoch": 0.8716425224342287, "percentage": 43.59, "elapsed_time": "14:14:31", "remaining_time": "18:26:01", "throughput": 925.01, "total_tokens": 47426776}
{"current_steps": 7120, "total_steps": 16324, "loss": 0.2853, "lr": 6.448242412960002e-05, "epoch": 0.872255061100732, "percentage": 43.62, "elapsed_time": "14:15:32", "remaining_time": "18:25:56", "throughput": 924.57, "total_tokens": 47460136}
{"current_steps": 7125, "total_steps": 16324, "loss": 0.2791, "lr": 6.443393985391248e-05, "epoch": 0.8728675997672353, "percentage": 43.65, "elapsed_time": "14:16:32", "remaining_time": "18:25:52", "throughput": 924.14, "total_tokens": 47493496}
{"current_steps": 7130, "total_steps": 16324, "loss": 0.2899, "lr": 6.438544076774835e-05, "epoch": 0.8734801384337386, "percentage": 43.68, "elapsed_time": "14:17:33", "remaining_time": "18:25:48", "throughput": 923.68, "total_tokens": 47526760}
{"current_steps": 7135, "total_steps": 16324, "loss": 0.2632, "lr": 6.433692692087189e-05, "epoch": 0.874092677100242, "percentage": 43.71, "elapsed_time": "14:18:34", "remaining_time": "18:25:44", "throughput": 923.24, "total_tokens": 47560312}
{"current_steps": 7140, "total_steps": 16324, "loss": 0.3208, "lr": 6.428839836306253e-05, "epoch": 0.8747052157667453, "percentage": 43.74, "elapsed_time": "14:19:34", "remaining_time": "18:25:38", "throughput": 922.82, "total_tokens": 47593960}
{"current_steps": 7145, "total_steps": 16324, "loss": 0.2891, "lr": 6.423985514411477e-05, "epoch": 0.8753177544332486, "percentage": 43.77, "elapsed_time": "14:20:35", "remaining_time": "18:25:35", "throughput": 922.38, "total_tokens": 47627816}
{"current_steps": 7150, "total_steps": 16324, "loss": 0.3548, "lr": 6.419129731383822e-05, "epoch": 0.8759302930997519, "percentage": 43.8, "elapsed_time": "14:21:35", "remaining_time": "18:25:29", "throughput": 921.96, "total_tokens": 47661416}
{"current_steps": 7150, "total_steps": 16324, "eval_loss": 0.1315048336982727, "epoch": 0.8759302930997519, "percentage": 43.8, "elapsed_time": "14:21:54", "remaining_time": "18:25:53", "throughput": 921.62, "total_tokens": 47661416}
{"current_steps": 7155, "total_steps": 16324, "loss": 0.3215, "lr": 6.414272492205736e-05, "epoch": 0.8765428317662552, "percentage": 43.83, "elapsed_time": "14:22:59", "remaining_time": "18:25:54", "throughput": 921.11, "total_tokens": 47694568}
{"current_steps": 7160, "total_steps": 16324, "loss": 0.2936, "lr": 6.409413801861174e-05, "epoch": 0.8771553704327586, "percentage": 43.86, "elapsed_time": "14:24:00", "remaining_time": "18:25:49", "throughput": 920.67, "total_tokens": 47727576}
{"current_steps": 7165, "total_steps": 16324, "loss": 0.2344, "lr": 6.40455366533557e-05, "epoch": 0.877767909099262, "percentage": 43.89, "elapsed_time": "14:24:59", "remaining_time": "18:25:43", "throughput": 920.25, "total_tokens": 47760752}
{"current_steps": 7170, "total_steps": 16324, "loss": 0.2927, "lr": 6.399692087615853e-05, "epoch": 0.8783804477657652, "percentage": 43.92, "elapsed_time": "14:25:59", "remaining_time": "18:25:36", "throughput": 919.85, "total_tokens": 47794544}
{"current_steps": 7175, "total_steps": 16324, "loss": 0.2445, "lr": 6.394829073690419e-05, "epoch": 0.8789929864322685, "percentage": 43.95, "elapsed_time": "14:26:59", "remaining_time": "18:25:30", "throughput": 919.45, "total_tokens": 47828872}
{"current_steps": 7180, "total_steps": 16324, "loss": 0.2414, "lr": 6.389964628549144e-05, "epoch": 0.8796055250987719, "percentage": 43.98, "elapsed_time": "14:27:59", "remaining_time": "18:25:25", "throughput": 919.03, "total_tokens": 47862896}
{"current_steps": 7185, "total_steps": 16324, "loss": 0.3247, "lr": 6.385098757183373e-05, "epoch": 0.8802180637652752, "percentage": 44.01, "elapsed_time": "14:28:58", "remaining_time": "18:25:18", "throughput": 918.62, "total_tokens": 47895992}
{"current_steps": 7190, "total_steps": 16324, "loss": 0.2542, "lr": 6.38023146458591e-05, "epoch": 0.8808306024317785, "percentage": 44.05, "elapsed_time": "14:29:58", "remaining_time": "18:25:11", "throughput": 918.22, "total_tokens": 47929568}
{"current_steps": 7195, "total_steps": 16324, "loss": 0.2198, "lr": 6.375362755751024e-05, "epoch": 0.8814431410982818, "percentage": 44.08, "elapsed_time": "14:30:58", "remaining_time": "18:25:05", "throughput": 917.82, "total_tokens": 47964048}
{"current_steps": 7200, "total_steps": 16324, "loss": 0.2941, "lr": 6.37049263567443e-05, "epoch": 0.8820556797647852, "percentage": 44.11, "elapsed_time": "14:31:58", "remaining_time": "18:24:58", "throughput": 917.41, "total_tokens": 47997384}
{"current_steps": 7200, "total_steps": 16324, "eval_loss": 0.16184844076633453, "epoch": 0.8820556797647852, "percentage": 44.11, "elapsed_time": "14:32:17", "remaining_time": "18:25:23", "throughput": 917.07, "total_tokens": 47997384}
{"current_steps": 7205, "total_steps": 16324, "loss": 0.3223, "lr": 6.365621109353294e-05, "epoch": 0.8826682184312885, "percentage": 44.14, "elapsed_time": "14:33:22", "remaining_time": "18:25:23", "throughput": 916.57, "total_tokens": 48030784}
{"current_steps": 7210, "total_steps": 16324, "loss": 0.3319, "lr": 6.360748181786229e-05, "epoch": 0.8832807570977917, "percentage": 44.17, "elapsed_time": "14:34:23", "remaining_time": "18:25:18", "throughput": 916.15, "total_tokens": 48064664}
{"current_steps": 7215, "total_steps": 16324, "loss": 0.3325, "lr": 6.35587385797328e-05, "epoch": 0.8838932957642951, "percentage": 44.2, "elapsed_time": "14:35:23", "remaining_time": "18:25:11", "throughput": 915.74, "total_tokens": 48097784}
{"current_steps": 7220, "total_steps": 16324, "loss": 0.2594, "lr": 6.350998142915927e-05, "epoch": 0.8845058344307984, "percentage": 44.23, "elapsed_time": "14:36:23", "remaining_time": "18:25:04", "throughput": 915.35, "total_tokens": 48131872}
{"current_steps": 7225, "total_steps": 16324, "loss": 0.2343, "lr": 6.346121041617076e-05, "epoch": 0.8851183730973018, "percentage": 44.26, "elapsed_time": "14:37:24", "remaining_time": "18:24:59", "throughput": 914.93, "total_tokens": 48165824}
{"current_steps": 7230, "total_steps": 16324, "loss": 0.2769, "lr": 6.34124255908106e-05, "epoch": 0.8857309117638051, "percentage": 44.29, "elapsed_time": "14:38:24", "remaining_time": "18:24:51", "throughput": 914.51, "total_tokens": 48198496}
{"current_steps": 7235, "total_steps": 16324, "loss": 0.2866, "lr": 6.336362700313622e-05, "epoch": 0.8863434504303084, "percentage": 44.32, "elapsed_time": "14:39:23", "remaining_time": "18:24:44", "throughput": 914.12, "total_tokens": 48232376}
{"current_steps": 7240, "total_steps": 16324, "loss": 0.2083, "lr": 6.331481470321923e-05, "epoch": 0.8869559890968117, "percentage": 44.35, "elapsed_time": "14:40:24", "remaining_time": "18:24:38", "throughput": 913.72, "total_tokens": 48266576}
{"current_steps": 7245, "total_steps": 16324, "loss": 0.2569, "lr": 6.32659887411453e-05, "epoch": 0.8875685277633151, "percentage": 44.38, "elapsed_time": "14:41:23", "remaining_time": "18:24:30", "throughput": 913.34, "total_tokens": 48300576}
{"current_steps": 7250, "total_steps": 16324, "loss": 0.3044, "lr": 6.321714916701411e-05, "epoch": 0.8881810664298184, "percentage": 44.41, "elapsed_time": "14:42:23", "remaining_time": "18:24:23", "throughput": 912.93, "total_tokens": 48333536}
{"current_steps": 7250, "total_steps": 16324, "eval_loss": 0.10158420354127884, "epoch": 0.8881810664298184, "percentage": 44.41, "elapsed_time": "14:42:43", "remaining_time": "18:24:47", "throughput": 912.59, "total_tokens": 48333536}
{"current_steps": 7255, "total_steps": 16324, "loss": 0.3441, "lr": 6.316829603093931e-05, "epoch": 0.8887936050963217, "percentage": 44.44, "elapsed_time": "14:43:48", "remaining_time": "18:24:47", "throughput": 912.08, "total_tokens": 48366624}
{"current_steps": 7260, "total_steps": 16324, "loss": 0.2336, "lr": 6.311942938304842e-05, "epoch": 0.889406143762825, "percentage": 44.47, "elapsed_time": "14:44:47", "remaining_time": "18:24:38", "throughput": 911.71, "total_tokens": 48400456}
{"current_steps": 7265, "total_steps": 16324, "loss": 0.2426, "lr": 6.307054927348293e-05, "epoch": 0.8900186824293284, "percentage": 44.51, "elapsed_time": "14:45:46", "remaining_time": "18:24:29", "throughput": 911.35, "total_tokens": 48434648}
{"current_steps": 7270, "total_steps": 16324, "loss": 0.3274, "lr": 6.302165575239804e-05, "epoch": 0.8906312210958317, "percentage": 44.54, "elapsed_time": "14:46:46", "remaining_time": "18:24:23", "throughput": 910.95, "total_tokens": 48468824}
{"current_steps": 7275, "total_steps": 16324, "loss": 0.3144, "lr": 6.297274886996278e-05, "epoch": 0.891243759762335, "percentage": 44.57, "elapsed_time": "14:47:45", "remaining_time": "18:24:14", "throughput": 910.58, "total_tokens": 48502576}
{"current_steps": 7280, "total_steps": 16324, "loss": 0.2943, "lr": 6.292382867635981e-05, "epoch": 0.8918562984288383, "percentage": 44.6, "elapsed_time": "14:48:44", "remaining_time": "18:24:05", "throughput": 910.21, "total_tokens": 48536224}
{"current_steps": 7285, "total_steps": 16324, "loss": 0.2551, "lr": 6.287489522178556e-05, "epoch": 0.8924688370953416, "percentage": 44.63, "elapsed_time": "14:49:44", "remaining_time": "18:23:58", "throughput": 909.81, "total_tokens": 48570024}
{"current_steps": 7290, "total_steps": 16324, "loss": 0.2862, "lr": 6.282594855644995e-05, "epoch": 0.893081375761845, "percentage": 44.66, "elapsed_time": "14:50:44", "remaining_time": "18:23:50", "throughput": 909.42, "total_tokens": 48603472}
{"current_steps": 7295, "total_steps": 16324, "loss": 0.2366, "lr": 6.277698873057658e-05, "epoch": 0.8936939144283483, "percentage": 44.69, "elapsed_time": "14:51:44", "remaining_time": "18:23:42", "throughput": 909.03, "total_tokens": 48637176}
{"current_steps": 7300, "total_steps": 16324, "loss": 0.3151, "lr": 6.272801579440243e-05, "epoch": 0.8943064530948516, "percentage": 44.72, "elapsed_time": "14:52:44", "remaining_time": "18:23:34", "throughput": 908.63, "total_tokens": 48670648}
{"current_steps": 7300, "total_steps": 16324, "eval_loss": 0.07981760799884796, "epoch": 0.8943064530948516, "percentage": 44.72, "elapsed_time": "14:53:04", "remaining_time": "18:23:58", "throughput": 908.3, "total_tokens": 48670648}
{"current_steps": 7305, "total_steps": 16324, "loss": 0.2829, "lr": 6.2679029798178e-05, "epoch": 0.8949189917613549, "percentage": 44.75, "elapsed_time": "14:54:09", "remaining_time": "18:23:57", "throughput": 907.83, "total_tokens": 48704528}
{"current_steps": 7310, "total_steps": 16324, "loss": 0.2833, "lr": 6.263003079216721e-05, "epoch": 0.8955315304278583, "percentage": 44.78, "elapsed_time": "14:55:09", "remaining_time": "18:23:49", "throughput": 907.44, "total_tokens": 48738256}
{"current_steps": 7315, "total_steps": 16324, "loss": 0.317, "lr": 6.258101882664725e-05, "epoch": 0.8961440690943616, "percentage": 44.81, "elapsed_time": "14:56:09", "remaining_time": "18:23:41", "throughput": 907.05, "total_tokens": 48771552}
{"current_steps": 7320, "total_steps": 16324, "loss": 0.2923, "lr": 6.25319939519087e-05, "epoch": 0.8967566077608649, "percentage": 44.84, "elapsed_time": "14:57:10", "remaining_time": "18:23:33", "throughput": 906.65, "total_tokens": 48805200}
{"current_steps": 7325, "total_steps": 16324, "loss": 0.3079, "lr": 6.24829562182553e-05, "epoch": 0.8973691464273682, "percentage": 44.87, "elapsed_time": "14:58:09", "remaining_time": "18:23:25", "throughput": 906.28, "total_tokens": 48839192}
{"current_steps": 7330, "total_steps": 16324, "loss": 0.3042, "lr": 6.243390567600406e-05, "epoch": 0.8979816850938716, "percentage": 44.9, "elapsed_time": "14:59:09", "remaining_time": "18:23:17", "throughput": 905.88, "total_tokens": 48872152}
{"current_steps": 7335, "total_steps": 16324, "loss": 0.2434, "lr": 6.238484237548509e-05, "epoch": 0.8985942237603749, "percentage": 44.93, "elapsed_time": "15:00:10", "remaining_time": "18:23:09", "throughput": 905.48, "total_tokens": 48905568}
{"current_steps": 7340, "total_steps": 16324, "loss": 0.238, "lr": 6.233576636704158e-05, "epoch": 0.8992067624268782, "percentage": 44.96, "elapsed_time": "15:01:10", "remaining_time": "18:23:01", "throughput": 905.11, "total_tokens": 48939632}
{"current_steps": 7345, "total_steps": 16324, "loss": 0.2288, "lr": 6.228667770102982e-05, "epoch": 0.8998193010933815, "percentage": 45.0, "elapsed_time": "15:02:10", "remaining_time": "18:22:53", "throughput": 904.73, "total_tokens": 48973952}
{"current_steps": 7350, "total_steps": 16324, "loss": 0.2531, "lr": 6.2237576427819e-05, "epoch": 0.9004318397598848, "percentage": 45.03, "elapsed_time": "15:03:11", "remaining_time": "18:22:45", "throughput": 904.33, "total_tokens": 49007448}
{"current_steps": 7350, "total_steps": 16324, "eval_loss": 0.19394417107105255, "epoch": 0.9004318397598848, "percentage": 45.03, "elapsed_time": "15:03:31", "remaining_time": "18:23:09", "throughput": 904.0, "total_tokens": 49007448}
{"current_steps": 7355, "total_steps": 16324, "loss": 0.2663, "lr": 6.218846259779133e-05, "epoch": 0.9010443784263882, "percentage": 45.06, "elapsed_time": "15:04:36", "remaining_time": "18:23:07", "throughput": 903.54, "total_tokens": 49041288}
{"current_steps": 7360, "total_steps": 16324, "loss": 0.2916, "lr": 6.213933626134185e-05, "epoch": 0.9016569170928915, "percentage": 45.09, "elapsed_time": "15:05:36", "remaining_time": "18:22:58", "throughput": 903.16, "total_tokens": 49074616}
{"current_steps": 7365, "total_steps": 16324, "loss": 0.2928, "lr": 6.209019746887846e-05, "epoch": 0.9022694557593948, "percentage": 45.12, "elapsed_time": "15:06:36", "remaining_time": "18:22:49", "throughput": 902.77, "total_tokens": 49107784}
{"current_steps": 7370, "total_steps": 16324, "loss": 0.2488, "lr": 6.204104627082181e-05, "epoch": 0.9028819944258981, "percentage": 45.15, "elapsed_time": "15:07:35", "remaining_time": "18:22:39", "throughput": 902.41, "total_tokens": 49141360}
{"current_steps": 7375, "total_steps": 16324, "loss": 0.2659, "lr": 6.199188271760532e-05, "epoch": 0.9034945330924015, "percentage": 45.18, "elapsed_time": "15:08:34", "remaining_time": "18:22:29", "throughput": 902.05, "total_tokens": 49174824}
{"current_steps": 7380, "total_steps": 16324, "loss": 0.3333, "lr": 6.194270685967507e-05, "epoch": 0.9041070717589048, "percentage": 45.21, "elapsed_time": "15:09:35", "remaining_time": "18:22:20", "throughput": 901.67, "total_tokens": 49208608}
{"current_steps": 7385, "total_steps": 16324, "loss": 0.2677, "lr": 6.189351874748976e-05, "epoch": 0.904719610425408, "percentage": 45.24, "elapsed_time": "15:10:34", "remaining_time": "18:22:11", "throughput": 901.31, "total_tokens": 49242616}
{"current_steps": 7390, "total_steps": 16324, "loss": 0.2366, "lr": 6.184431843152065e-05, "epoch": 0.9053321490919114, "percentage": 45.27, "elapsed_time": "15:11:34", "remaining_time": "18:22:01", "throughput": 900.96, "total_tokens": 49277400}
{"current_steps": 7395, "total_steps": 16324, "loss": 0.2633, "lr": 6.179510596225157e-05, "epoch": 0.9059446877584147, "percentage": 45.3, "elapsed_time": "15:12:35", "remaining_time": "18:21:53", "throughput": 900.56, "total_tokens": 49310528}
{"current_steps": 7400, "total_steps": 16324, "loss": 0.2813, "lr": 6.174588139017878e-05, "epoch": 0.9065572264249181, "percentage": 45.33, "elapsed_time": "15:13:36", "remaining_time": "18:21:45", "throughput": 900.16, "total_tokens": 49343416}
{"current_steps": 7400, "total_steps": 16324, "eval_loss": 0.12092752754688263, "epoch": 0.9065572264249181, "percentage": 45.33, "elapsed_time": "15:13:55", "remaining_time": "18:22:08", "throughput": 899.84, "total_tokens": 49343416}
{"current_steps": 7405, "total_steps": 16324, "loss": 0.2468, "lr": 6.169664476581095e-05, "epoch": 0.9071697650914214, "percentage": 45.36, "elapsed_time": "15:15:01", "remaining_time": "18:22:05", "throughput": 899.39, "total_tokens": 49377224}
{"current_steps": 7410, "total_steps": 16324, "loss": 0.2945, "lr": 6.164739613966915e-05, "epoch": 0.9077823037579247, "percentage": 45.39, "elapsed_time": "15:16:02", "remaining_time": "18:21:58", "throughput": 898.98, "total_tokens": 49410368}
{"current_steps": 7415, "total_steps": 16324, "loss": 0.261, "lr": 6.159813556228674e-05, "epoch": 0.908394842424428, "percentage": 45.42, "elapsed_time": "15:17:02", "remaining_time": "18:21:48", "throughput": 898.62, "total_tokens": 49444360}
{"current_steps": 7420, "total_steps": 16324, "loss": 0.2423, "lr": 6.154886308420937e-05, "epoch": 0.9090073810909314, "percentage": 45.45, "elapsed_time": "15:18:02", "remaining_time": "18:21:39", "throughput": 898.25, "total_tokens": 49478040}
{"current_steps": 7425, "total_steps": 16324, "loss": 0.3007, "lr": 6.149957875599489e-05, "epoch": 0.9096199197574347, "percentage": 45.49, "elapsed_time": "15:19:03", "remaining_time": "18:21:30", "throughput": 897.88, "total_tokens": 49512288}
{"current_steps": 7430, "total_steps": 16324, "loss": 0.3359, "lr": 6.145028262821328e-05, "epoch": 0.910232458423938, "percentage": 45.52, "elapsed_time": "15:20:04", "remaining_time": "18:21:21", "throughput": 897.49, "total_tokens": 49545200}
{"current_steps": 7435, "total_steps": 16324, "loss": 0.2323, "lr": 6.140097475144666e-05, "epoch": 0.9108449970904413, "percentage": 45.55, "elapsed_time": "15:21:04", "remaining_time": "18:21:11", "throughput": 897.14, "total_tokens": 49579520}
{"current_steps": 7440, "total_steps": 16324, "loss": 0.2624, "lr": 6.135165517628919e-05, "epoch": 0.9114575357569447, "percentage": 45.58, "elapsed_time": "15:22:05", "remaining_time": "18:21:02", "throughput": 896.76, "total_tokens": 49613224}
{"current_steps": 7445, "total_steps": 16324, "loss": 0.2297, "lr": 6.130232395334705e-05, "epoch": 0.912070074423448, "percentage": 45.61, "elapsed_time": "15:23:06", "remaining_time": "18:20:54", "throughput": 896.38, "total_tokens": 49647256}
{"current_steps": 7450, "total_steps": 16324, "loss": 0.2411, "lr": 6.125298113323836e-05, "epoch": 0.9126826130899514, "percentage": 45.64, "elapsed_time": "15:24:06", "remaining_time": "18:20:45", "throughput": 896.01, "total_tokens": 49681144}
{"current_steps": 7450, "total_steps": 16324, "eval_loss": 0.13680709898471832, "epoch": 0.9126826130899514, "percentage": 45.64, "elapsed_time": "15:24:26", "remaining_time": "18:21:08", "throughput": 895.7, "total_tokens": 49681144}
{"current_steps": 7455, "total_steps": 16324, "loss": 0.2216, "lr": 6.120362676659315e-05, "epoch": 0.9132951517564546, "percentage": 45.67, "elapsed_time": "15:25:33", "remaining_time": "18:21:06", "throughput": 895.22, "total_tokens": 49714752}
{"current_steps": 7460, "total_steps": 16324, "loss": 0.256, "lr": 6.115426090405328e-05, "epoch": 0.9139076904229579, "percentage": 45.7, "elapsed_time": "15:26:34", "remaining_time": "18:20:57", "throughput": 894.84, "total_tokens": 49748256}
{"current_steps": 7465, "total_steps": 16324, "loss": 0.2347, "lr": 6.110488359627239e-05, "epoch": 0.9145202290894613, "percentage": 45.73, "elapsed_time": "15:27:35", "remaining_time": "18:20:48", "throughput": 894.46, "total_tokens": 49781400}
{"current_steps": 7470, "total_steps": 16324, "loss": 0.3048, "lr": 6.105549489391593e-05, "epoch": 0.9151327677559646, "percentage": 45.76, "elapsed_time": "15:28:36", "remaining_time": "18:20:39", "throughput": 894.08, "total_tokens": 49814944}
{"current_steps": 7475, "total_steps": 16324, "loss": 0.2603, "lr": 6.100609484766098e-05, "epoch": 0.9157453064224679, "percentage": 45.79, "elapsed_time": "15:29:38", "remaining_time": "18:20:31", "throughput": 893.69, "total_tokens": 49848608}
{"current_steps": 7480, "total_steps": 16324, "loss": 0.2747, "lr": 6.095668350819629e-05, "epoch": 0.9163578450889712, "percentage": 45.82, "elapsed_time": "15:30:38", "remaining_time": "18:20:20", "throughput": 893.33, "total_tokens": 49882168}
{"current_steps": 7485, "total_steps": 16324, "loss": 0.2553, "lr": 6.090726092622216e-05, "epoch": 0.9169703837554746, "percentage": 45.85, "elapsed_time": "15:31:38", "remaining_time": "18:20:10", "throughput": 892.97, "total_tokens": 49915840}
{"current_steps": 7490, "total_steps": 16324, "loss": 0.2541, "lr": 6.08578271524505e-05, "epoch": 0.9175829224219779, "percentage": 45.88, "elapsed_time": "15:32:40", "remaining_time": "18:20:01", "throughput": 892.59, "total_tokens": 49949480}
{"current_steps": 7495, "total_steps": 16324, "loss": 0.2757, "lr": 6.080838223760462e-05, "epoch": 0.9181954610884812, "percentage": 45.91, "elapsed_time": "15:33:40", "remaining_time": "18:19:51", "throughput": 892.24, "total_tokens": 49983888}
{"current_steps": 7500, "total_steps": 16324, "loss": 0.2807, "lr": 6.0758926232419325e-05, "epoch": 0.9188079997549845, "percentage": 45.94, "elapsed_time": "15:34:42", "remaining_time": "18:19:42", "throughput": 891.85, "total_tokens": 50016856}
{"current_steps": 7500, "total_steps": 16324, "eval_loss": 0.1768772155046463, "epoch": 0.9188079997549845, "percentage": 45.94, "elapsed_time": "15:35:01", "remaining_time": "18:20:05", "throughput": 891.54, "total_tokens": 50016856}
{"current_steps": 7505, "total_steps": 16324, "loss": 0.2769, "lr": 6.070945918764075e-05, "epoch": 0.9194205384214879, "percentage": 45.98, "elapsed_time": "15:36:07", "remaining_time": "18:20:01", "throughput": 891.09, "total_tokens": 50050272}
{"current_steps": 7510, "total_steps": 16324, "loss": 0.2663, "lr": 6.065998115402639e-05, "epoch": 0.9200330770879912, "percentage": 46.01, "elapsed_time": "15:37:07", "remaining_time": "18:19:50", "throughput": 890.73, "total_tokens": 50083664}
{"current_steps": 7515, "total_steps": 16324, "loss": 0.2737, "lr": 6.0610492182345035e-05, "epoch": 0.9206456157544946, "percentage": 46.04, "elapsed_time": "15:38:07", "remaining_time": "18:19:39", "throughput": 890.38, "total_tokens": 50117640}
{"current_steps": 7520, "total_steps": 16324, "loss": 0.2674, "lr": 6.0560992323376617e-05, "epoch": 0.9212581544209978, "percentage": 46.07, "elapsed_time": "15:39:08", "remaining_time": "18:19:29", "throughput": 890.01, "total_tokens": 50151144}
{"current_steps": 7525, "total_steps": 16324, "loss": 0.2974, "lr": 6.0511481627912334e-05, "epoch": 0.9218706930875011, "percentage": 46.1, "elapsed_time": "15:40:08", "remaining_time": "18:19:18", "throughput": 889.66, "total_tokens": 50184120}
{"current_steps": 7530, "total_steps": 16324, "loss": 0.2534, "lr": 6.046196014675445e-05, "epoch": 0.9224832317540045, "percentage": 46.13, "elapsed_time": "15:41:07", "remaining_time": "18:19:06", "throughput": 889.31, "total_tokens": 50217472}
{"current_steps": 7535, "total_steps": 16324, "loss": 0.2732, "lr": 6.041242793071631e-05, "epoch": 0.9230957704205078, "percentage": 46.16, "elapsed_time": "15:42:08", "remaining_time": "18:18:56", "throughput": 888.95, "total_tokens": 50251224}
{"current_steps": 7540, "total_steps": 16324, "loss": 0.269, "lr": 6.036288503062227e-05, "epoch": 0.9237083090870111, "percentage": 46.19, "elapsed_time": "15:43:08", "remaining_time": "18:18:44", "throughput": 888.59, "total_tokens": 50284144}
{"current_steps": 7545, "total_steps": 16324, "loss": 0.2523, "lr": 6.031333149730765e-05, "epoch": 0.9243208477535144, "percentage": 46.22, "elapsed_time": "15:44:08", "remaining_time": "18:18:33", "throughput": 888.25, "total_tokens": 50317784}
{"current_steps": 7550, "total_steps": 16324, "loss": 0.2508, "lr": 6.0263767381618674e-05, "epoch": 0.9249333864200178, "percentage": 46.25, "elapsed_time": "15:45:08", "remaining_time": "18:18:22", "throughput": 887.89, "total_tokens": 50351504}
{"current_steps": 7550, "total_steps": 16324, "eval_loss": 0.32131484150886536, "epoch": 0.9249333864200178, "percentage": 46.25, "elapsed_time": "15:45:28", "remaining_time": "18:18:45", "throughput": 887.58, "total_tokens": 50351504}
{"current_steps": 7555, "total_steps": 16324, "loss": 0.2933, "lr": 6.021419273441246e-05, "epoch": 0.9255459250865211, "percentage": 46.28, "elapsed_time": "15:46:34", "remaining_time": "18:18:40", "throughput": 887.15, "total_tokens": 50385360}
{"current_steps": 7560, "total_steps": 16324, "loss": 0.2925, "lr": 6.016460760655687e-05, "epoch": 0.9261584637530245, "percentage": 46.31, "elapsed_time": "15:47:34", "remaining_time": "18:18:29", "throughput": 886.8, "total_tokens": 50418744}
{"current_steps": 7565, "total_steps": 16324, "loss": 0.2047, "lr": 6.011501204893056e-05, "epoch": 0.9267710024195277, "percentage": 46.34, "elapsed_time": "15:48:36", "remaining_time": "18:18:19", "throughput": 886.45, "total_tokens": 50453160}
{"current_steps": 7570, "total_steps": 16324, "loss": 0.2542, "lr": 6.006540611242291e-05, "epoch": 0.927383541086031, "percentage": 46.37, "elapsed_time": "15:49:37", "remaining_time": "18:18:09", "throughput": 886.07, "total_tokens": 50486256}
{"current_steps": 7575, "total_steps": 16324, "loss": 0.2613, "lr": 6.001578984793388e-05, "epoch": 0.9279960797525344, "percentage": 46.4, "elapsed_time": "15:50:38", "remaining_time": "18:17:58", "throughput": 885.72, "total_tokens": 50519696}
{"current_steps": 7580, "total_steps": 16324, "loss": 0.2945, "lr": 5.9966163306374115e-05, "epoch": 0.9286086184190377, "percentage": 46.43, "elapsed_time": "15:51:39", "remaining_time": "18:17:47", "throughput": 885.35, "total_tokens": 50552912}
{"current_steps": 7585, "total_steps": 16324, "loss": 0.2545, "lr": 5.991652653866472e-05, "epoch": 0.929221157085541, "percentage": 46.47, "elapsed_time": "15:52:39", "remaining_time": "18:17:35", "throughput": 885.02, "total_tokens": 50587184}
{"current_steps": 7590, "total_steps": 16324, "loss": 0.2947, "lr": 5.9866879595737344e-05, "epoch": 0.9298336957520443, "percentage": 46.5, "elapsed_time": "15:53:39", "remaining_time": "18:17:23", "throughput": 884.68, "total_tokens": 50620448}
{"current_steps": 7595, "total_steps": 16324, "loss": 0.2863, "lr": 5.981722252853409e-05, "epoch": 0.9304462344185477, "percentage": 46.53, "elapsed_time": "15:54:39", "remaining_time": "18:17:11", "throughput": 884.33, "total_tokens": 50653968}
{"current_steps": 7600, "total_steps": 16324, "loss": 0.3394, "lr": 5.976755538800738e-05, "epoch": 0.931058773085051, "percentage": 46.56, "elapsed_time": "15:55:41", "remaining_time": "18:17:01", "throughput": 883.96, "total_tokens": 50687088}
{"current_steps": 7600, "total_steps": 16324, "eval_loss": 0.2350298911333084, "epoch": 0.931058773085051, "percentage": 46.56, "elapsed_time": "15:56:00", "remaining_time": "18:17:24", "throughput": 883.65, "total_tokens": 50687088}
{"current_steps": 7605, "total_steps": 16324, "loss": 0.264, "lr": 5.971787822512005e-05, "epoch": 0.9316713117515543, "percentage": 46.59, "elapsed_time": "15:57:06", "remaining_time": "18:17:18", "throughput": 883.22, "total_tokens": 50720400}
{"current_steps": 7610, "total_steps": 16324, "loss": 0.2592, "lr": 5.966819109084516e-05, "epoch": 0.9322838504180576, "percentage": 46.62, "elapsed_time": "15:58:08", "remaining_time": "18:17:07", "throughput": 882.86, "total_tokens": 50753984}
{"current_steps": 7615, "total_steps": 16324, "loss": 0.2627, "lr": 5.961849403616606e-05, "epoch": 0.932896389084561, "percentage": 46.65, "elapsed_time": "15:59:09", "remaining_time": "18:16:57", "throughput": 882.51, "total_tokens": 50787920}
{"current_steps": 7620, "total_steps": 16324, "loss": 0.2497, "lr": 5.956878711207623e-05, "epoch": 0.9335089277510643, "percentage": 46.68, "elapsed_time": "16:00:10", "remaining_time": "18:16:46", "throughput": 882.16, "total_tokens": 50821736}
{"current_steps": 7625, "total_steps": 16324, "loss": 0.287, "lr": 5.951907036957927e-05, "epoch": 0.9341214664175677, "percentage": 46.71, "elapsed_time": "16:01:12", "remaining_time": "18:16:35", "throughput": 881.79, "total_tokens": 50855040}
{"current_steps": 7630, "total_steps": 16324, "loss": 0.2704, "lr": 5.946934385968892e-05, "epoch": 0.9347340050840709, "percentage": 46.74, "elapsed_time": "16:02:14", "remaining_time": "18:16:25", "throughput": 881.42, "total_tokens": 50888688}
{"current_steps": 7635, "total_steps": 16324, "loss": 0.2227, "lr": 5.9419607633428885e-05, "epoch": 0.9353465437505742, "percentage": 46.77, "elapsed_time": "16:03:15", "remaining_time": "18:16:14", "throughput": 881.07, "total_tokens": 50922464}
{"current_steps": 7640, "total_steps": 16324, "loss": 0.2602, "lr": 5.9369861741832847e-05, "epoch": 0.9359590824170776, "percentage": 46.8, "elapsed_time": "16:04:17", "remaining_time": "18:16:03", "throughput": 880.72, "total_tokens": 50955944}
{"current_steps": 7645, "total_steps": 16324, "loss": 0.2673, "lr": 5.932010623594441e-05, "epoch": 0.9365716210835809, "percentage": 46.83, "elapsed_time": "16:05:18", "remaining_time": "18:15:52", "throughput": 880.36, "total_tokens": 50989432}
{"current_steps": 7650, "total_steps": 16324, "loss": 0.2424, "lr": 5.927034116681705e-05, "epoch": 0.9371841597500842, "percentage": 46.86, "elapsed_time": "16:06:19", "remaining_time": "18:15:40", "throughput": 880.02, "total_tokens": 51023288}
{"current_steps": 7650, "total_steps": 16324, "eval_loss": 0.1906452476978302, "epoch": 0.9371841597500842, "percentage": 46.86, "elapsed_time": "16:06:39", "remaining_time": "18:16:03", "throughput": 879.72, "total_tokens": 51023288}
{"current_steps": 7655, "total_steps": 16324, "loss": 0.26, "lr": 5.922056658551407e-05, "epoch": 0.9377966984165875, "percentage": 46.89, "elapsed_time": "16:07:45", "remaining_time": "18:15:57", "throughput": 879.28, "total_tokens": 51056376}
{"current_steps": 7660, "total_steps": 16324, "loss": 0.2123, "lr": 5.91707825431085e-05, "epoch": 0.9384092370830909, "percentage": 46.92, "elapsed_time": "16:08:47", "remaining_time": "18:15:46", "throughput": 878.93, "total_tokens": 51090288}
{"current_steps": 7665, "total_steps": 16324, "loss": 0.2648, "lr": 5.91209890906831e-05, "epoch": 0.9390217757495942, "percentage": 46.96, "elapsed_time": "16:09:48", "remaining_time": "18:15:34", "throughput": 878.6, "total_tokens": 51124128}
{"current_steps": 7670, "total_steps": 16324, "loss": 0.264, "lr": 5.907118627933027e-05, "epoch": 0.9396343144160976, "percentage": 46.99, "elapsed_time": "16:10:49", "remaining_time": "18:15:22", "throughput": 878.26, "total_tokens": 51158152}
{"current_steps": 7675, "total_steps": 16324, "loss": 0.2314, "lr": 5.902137416015204e-05, "epoch": 0.9402468530826008, "percentage": 47.02, "elapsed_time": "16:11:50", "remaining_time": "18:15:10", "throughput": 877.92, "total_tokens": 51191840}
{"current_steps": 7680, "total_steps": 16324, "loss": 0.2395, "lr": 5.897155278425995e-05, "epoch": 0.9408593917491042, "percentage": 47.05, "elapsed_time": "16:12:51", "remaining_time": "18:14:58", "throughput": 877.57, "total_tokens": 51225560}
{"current_steps": 7685, "total_steps": 16324, "loss": 0.239, "lr": 5.892172220277509e-05, "epoch": 0.9414719304156075, "percentage": 47.08, "elapsed_time": "16:13:51", "remaining_time": "18:14:45", "throughput": 877.24, "total_tokens": 51258880}
{"current_steps": 7690, "total_steps": 16324, "loss": 0.2422, "lr": 5.887188246682792e-05, "epoch": 0.9420844690821109, "percentage": 47.11, "elapsed_time": "16:14:53", "remaining_time": "18:14:34", "throughput": 876.89, "total_tokens": 51292504}
{"current_steps": 7695, "total_steps": 16324, "loss": 0.2685, "lr": 5.882203362755839e-05, "epoch": 0.9426970077486141, "percentage": 47.14, "elapsed_time": "16:15:54", "remaining_time": "18:14:21", "throughput": 876.55, "total_tokens": 51326264}
{"current_steps": 7700, "total_steps": 16324, "loss": 0.2568, "lr": 5.877217573611572e-05, "epoch": 0.9433095464151174, "percentage": 47.17, "elapsed_time": "16:16:54", "remaining_time": "18:14:08", "throughput": 876.22, "total_tokens": 51359824}
{"current_steps": 7700, "total_steps": 16324, "eval_loss": 0.2279118001461029, "epoch": 0.9433095464151174, "percentage": 47.17, "elapsed_time": "16:17:14", "remaining_time": "18:14:30", "throughput": 875.93, "total_tokens": 51359824}
{"current_steps": 7705, "total_steps": 16324, "loss": 0.2776, "lr": 5.8722308843658436e-05, "epoch": 0.9439220850816208, "percentage": 47.2, "elapsed_time": "16:18:20", "remaining_time": "18:14:23", "throughput": 875.52, "total_tokens": 51393152}
{"current_steps": 7710, "total_steps": 16324, "loss": 0.3103, "lr": 5.867243300135431e-05, "epoch": 0.9445346237481241, "percentage": 47.23, "elapsed_time": "16:19:20", "remaining_time": "18:14:10", "throughput": 875.18, "total_tokens": 51426424}
{"current_steps": 7715, "total_steps": 16324, "loss": 0.2722, "lr": 5.8622548260380294e-05, "epoch": 0.9451471624146274, "percentage": 47.26, "elapsed_time": "16:20:20", "remaining_time": "18:13:56", "throughput": 874.87, "total_tokens": 51460504}
{"current_steps": 7720, "total_steps": 16324, "loss": 0.3227, "lr": 5.8572654671922455e-05, "epoch": 0.9457597010811307, "percentage": 47.29, "elapsed_time": "16:21:20", "remaining_time": "18:13:42", "throughput": 874.54, "total_tokens": 51493200}
{"current_steps": 7725, "total_steps": 16324, "loss": 0.2477, "lr": 5.852275228717595e-05, "epoch": 0.9463722397476341, "percentage": 47.32, "elapsed_time": "16:22:20", "remaining_time": "18:13:28", "throughput": 874.21, "total_tokens": 51526208}
{"current_steps": 7730, "total_steps": 16324, "loss": 0.2477, "lr": 5.847284115734497e-05, "epoch": 0.9469847784141374, "percentage": 47.35, "elapsed_time": "16:23:19", "remaining_time": "18:13:14", "throughput": 873.9, "total_tokens": 51559472}
{"current_steps": 7735, "total_steps": 16324, "loss": 0.2826, "lr": 5.8422921333642676e-05, "epoch": 0.9475973170806408, "percentage": 47.38, "elapsed_time": "16:24:19", "remaining_time": "18:12:59", "throughput": 873.57, "total_tokens": 51592584}
{"current_steps": 7740, "total_steps": 16324, "loss": 0.2885, "lr": 5.8372992867291146e-05, "epoch": 0.948209855747144, "percentage": 47.41, "elapsed_time": "16:25:19", "remaining_time": "18:12:46", "throughput": 873.25, "total_tokens": 51626208}
{"current_steps": 7745, "total_steps": 16324, "loss": 0.2608, "lr": 5.832305580952131e-05, "epoch": 0.9488223944136474, "percentage": 47.45, "elapsed_time": "16:26:18", "remaining_time": "18:12:30", "throughput": 872.95, "total_tokens": 51659776}
{"current_steps": 7750, "total_steps": 16324, "loss": 0.2863, "lr": 5.827311021157292e-05, "epoch": 0.9494349330801507, "percentage": 47.48, "elapsed_time": "16:27:17", "remaining_time": "18:12:16", "throughput": 872.63, "total_tokens": 51692760}
{"current_steps": 7750, "total_steps": 16324, "eval_loss": 0.20178191363811493, "epoch": 0.9494349330801507, "percentage": 47.48, "elapsed_time": "16:27:37", "remaining_time": "18:12:37", "throughput": 872.34, "total_tokens": 51692760}
{"current_steps": 7755, "total_steps": 16324, "loss": 0.2742, "lr": 5.822315612469455e-05, "epoch": 0.950047471746654, "percentage": 47.51, "elapsed_time": "16:28:42", "remaining_time": "18:12:29", "throughput": 871.96, "total_tokens": 51726896}
{"current_steps": 7760, "total_steps": 16324, "loss": 0.2665, "lr": 5.8173193600143383e-05, "epoch": 0.9506600104131573, "percentage": 47.54, "elapsed_time": "16:29:42", "remaining_time": "18:12:14", "throughput": 871.64, "total_tokens": 51760104}
{"current_steps": 7765, "total_steps": 16324, "loss": 0.301, "lr": 5.812322268918534e-05, "epoch": 0.9512725490796606, "percentage": 47.57, "elapsed_time": "16:30:42", "remaining_time": "18:12:00", "throughput": 871.33, "total_tokens": 51793832}
{"current_steps": 7770, "total_steps": 16324, "loss": 0.267, "lr": 5.807324344309489e-05, "epoch": 0.951885087746164, "percentage": 47.6, "elapsed_time": "16:31:42", "remaining_time": "18:11:46", "throughput": 871.01, "total_tokens": 51827408}
{"current_steps": 7775, "total_steps": 16324, "loss": 0.2462, "lr": 5.802325591315512e-05, "epoch": 0.9524976264126673, "percentage": 47.63, "elapsed_time": "16:32:42", "remaining_time": "18:11:31", "throughput": 870.7, "total_tokens": 51860984}
{"current_steps": 7780, "total_steps": 16324, "loss": 0.2373, "lr": 5.797326015065755e-05, "epoch": 0.9531101650791706, "percentage": 47.66, "elapsed_time": "16:33:42", "remaining_time": "18:11:17", "throughput": 870.39, "total_tokens": 51894936}
{"current_steps": 7785, "total_steps": 16324, "loss": 0.2417, "lr": 5.792325620690218e-05, "epoch": 0.9537227037456739, "percentage": 47.69, "elapsed_time": "16:34:42", "remaining_time": "18:11:03", "throughput": 870.09, "total_tokens": 51929072}
{"current_steps": 7790, "total_steps": 16324, "loss": 0.2802, "lr": 5.787324413319742e-05, "epoch": 0.9543352424121773, "percentage": 47.72, "elapsed_time": "16:35:42", "remaining_time": "18:10:47", "throughput": 869.79, "total_tokens": 51963016}
{"current_steps": 7795, "total_steps": 16324, "loss": 0.2768, "lr": 5.782322398085999e-05, "epoch": 0.9549477810786806, "percentage": 47.75, "elapsed_time": "16:36:42", "remaining_time": "18:10:33", "throughput": 869.47, "total_tokens": 51996552}
{"current_steps": 7800, "total_steps": 16324, "loss": 0.3015, "lr": 5.777319580121492e-05, "epoch": 0.955560319745184, "percentage": 47.78, "elapsed_time": "16:37:42", "remaining_time": "18:10:19", "throughput": 869.15, "total_tokens": 52029952}
{"current_steps": 7800, "total_steps": 16324, "eval_loss": 0.17758481204509735, "epoch": 0.955560319745184, "percentage": 47.78, "elapsed_time": "16:38:02", "remaining_time": "18:10:40", "throughput": 868.87, "total_tokens": 52029952}
{"current_steps": 7805, "total_steps": 16324, "loss": 0.2423, "lr": 5.7723159645595445e-05, "epoch": 0.9561728584116872, "percentage": 47.81, "elapsed_time": "16:39:07", "remaining_time": "18:10:31", "throughput": 868.48, "total_tokens": 52063720}
{"current_steps": 7810, "total_steps": 16324, "loss": 0.3008, "lr": 5.767311556534305e-05, "epoch": 0.9567853970781905, "percentage": 47.84, "elapsed_time": "16:40:07", "remaining_time": "18:10:16", "throughput": 868.18, "total_tokens": 52097360}
{"current_steps": 7815, "total_steps": 16324, "loss": 0.2982, "lr": 5.762306361180727e-05, "epoch": 0.9573979357446939, "percentage": 47.87, "elapsed_time": "16:41:08", "remaining_time": "18:10:02", "throughput": 867.87, "total_tokens": 52131184}
{"current_steps": 7820, "total_steps": 16324, "loss": 0.2885, "lr": 5.75730038363458e-05, "epoch": 0.9580104744111972, "percentage": 47.9, "elapsed_time": "16:42:09", "remaining_time": "18:09:49", "throughput": 867.53, "total_tokens": 52164512}
{"current_steps": 7825, "total_steps": 16324, "loss": 0.3219, "lr": 5.752293629032429e-05, "epoch": 0.9586230130777005, "percentage": 47.94, "elapsed_time": "16:43:09", "remaining_time": "18:09:34", "throughput": 867.21, "total_tokens": 52197296}
{"current_steps": 7830, "total_steps": 16324, "loss": 0.3073, "lr": 5.747286102511639e-05, "epoch": 0.9592355517442038, "percentage": 47.97, "elapsed_time": "16:44:11", "remaining_time": "18:09:20", "throughput": 866.89, "total_tokens": 52231016}
{"current_steps": 7835, "total_steps": 16324, "loss": 0.281, "lr": 5.74227780921037e-05, "epoch": 0.9598480904107072, "percentage": 48.0, "elapsed_time": "16:45:13", "remaining_time": "18:09:07", "throughput": 866.54, "total_tokens": 52263688}
{"current_steps": 7840, "total_steps": 16324, "loss": 0.2759, "lr": 5.7372687542675644e-05, "epoch": 0.9604606290772105, "percentage": 48.03, "elapsed_time": "16:46:14", "remaining_time": "18:08:53", "throughput": 866.21, "total_tokens": 52296880}
{"current_steps": 7845, "total_steps": 16324, "loss": 0.2422, "lr": 5.7322589428229476e-05, "epoch": 0.9610731677437139, "percentage": 48.06, "elapsed_time": "16:47:16", "remaining_time": "18:08:40", "throughput": 865.88, "total_tokens": 52330584}
{"current_steps": 7850, "total_steps": 16324, "loss": 0.2331, "lr": 5.727248380017021e-05, "epoch": 0.9616857064102171, "percentage": 48.09, "elapsed_time": "16:48:18", "remaining_time": "18:08:27", "throughput": 865.56, "total_tokens": 52365160}
{"current_steps": 7850, "total_steps": 16324, "eval_loss": 0.1693449765443802, "epoch": 0.9616857064102171, "percentage": 48.09, "elapsed_time": "16:48:38", "remaining_time": "18:08:49", "throughput": 865.27, "total_tokens": 52365160}
{"current_steps": 7855, "total_steps": 16324, "loss": 0.2671, "lr": 5.7222370709910586e-05, "epoch": 0.9622982450767205, "percentage": 48.12, "elapsed_time": "16:49:44", "remaining_time": "18:08:40", "throughput": 864.88, "total_tokens": 52398664}
{"current_steps": 7860, "total_steps": 16324, "loss": 0.2717, "lr": 5.7172250208871004e-05, "epoch": 0.9629107837432238, "percentage": 48.15, "elapsed_time": "16:50:45", "remaining_time": "18:08:25", "throughput": 864.57, "total_tokens": 52432176}
{"current_steps": 7865, "total_steps": 16324, "loss": 0.2688, "lr": 5.7122122348479424e-05, "epoch": 0.9635233224097272, "percentage": 48.18, "elapsed_time": "16:51:47", "remaining_time": "18:08:12", "throughput": 864.23, "total_tokens": 52465688}
{"current_steps": 7870, "total_steps": 16324, "loss": 0.2322, "lr": 5.7071987180171405e-05, "epoch": 0.9641358610762304, "percentage": 48.21, "elapsed_time": "16:52:48", "remaining_time": "18:07:58", "throughput": 863.91, "total_tokens": 52498856}
{"current_steps": 7875, "total_steps": 16324, "loss": 0.2777, "lr": 5.7021844755389976e-05, "epoch": 0.9647483997427337, "percentage": 48.24, "elapsed_time": "16:53:49", "remaining_time": "18:07:43", "throughput": 863.6, "total_tokens": 52532640}
{"current_steps": 7880, "total_steps": 16324, "loss": 0.2561, "lr": 5.6971695125585647e-05, "epoch": 0.9653609384092371, "percentage": 48.27, "elapsed_time": "16:54:51", "remaining_time": "18:07:29", "throughput": 863.28, "total_tokens": 52566704}
{"current_steps": 7885, "total_steps": 16324, "loss": 0.2266, "lr": 5.692153834221625e-05, "epoch": 0.9659734770757404, "percentage": 48.3, "elapsed_time": "16:55:52", "remaining_time": "18:07:15", "throughput": 862.97, "total_tokens": 52600248}
{"current_steps": 7890, "total_steps": 16324, "loss": 0.2087, "lr": 5.687137445674704e-05, "epoch": 0.9665860157422437, "percentage": 48.33, "elapsed_time": "16:56:54", "remaining_time": "18:07:00", "throughput": 862.67, "total_tokens": 52634744}
{"current_steps": 7895, "total_steps": 16324, "loss": 0.2681, "lr": 5.682120352065051e-05, "epoch": 0.967198554408747, "percentage": 48.36, "elapsed_time": "16:57:56", "remaining_time": "18:06:47", "throughput": 862.33, "total_tokens": 52667912}
{"current_steps": 7900, "total_steps": 16324, "loss": 0.2669, "lr": 5.677102558540641e-05, "epoch": 0.9678110930752504, "percentage": 48.4, "elapsed_time": "16:58:56", "remaining_time": "18:06:31", "throughput": 862.04, "total_tokens": 52702264}
{"current_steps": 7900, "total_steps": 16324, "eval_loss": 0.1096976026892662, "epoch": 0.9678110930752504, "percentage": 48.4, "elapsed_time": "16:59:16", "remaining_time": "18:06:53", "throughput": 861.76, "total_tokens": 52702264}
{"current_steps": 7905, "total_steps": 16324, "loss": 0.2886, "lr": 5.672084070250165e-05, "epoch": 0.9684236317417537, "percentage": 48.43, "elapsed_time": "17:00:22", "remaining_time": "18:06:43", "throughput": 861.38, "total_tokens": 52735976}
{"current_steps": 7910, "total_steps": 16324, "loss": 0.2443, "lr": 5.6670648923430295e-05, "epoch": 0.9690361704082571, "percentage": 48.46, "elapsed_time": "17:01:24", "remaining_time": "18:06:29", "throughput": 861.06, "total_tokens": 52769864}
{"current_steps": 7915, "total_steps": 16324, "loss": 0.2523, "lr": 5.662045029969345e-05, "epoch": 0.9696487090747603, "percentage": 48.49, "elapsed_time": "17:02:25", "remaining_time": "18:06:14", "throughput": 860.76, "total_tokens": 52803392}
{"current_steps": 7920, "total_steps": 16324, "loss": 0.2446, "lr": 5.6570244882799296e-05, "epoch": 0.9702612477412637, "percentage": 48.52, "elapsed_time": "17:03:25", "remaining_time": "18:05:57", "throughput": 860.46, "total_tokens": 52837136}
{"current_steps": 7925, "total_steps": 16324, "loss": 0.2661, "lr": 5.652003272426293e-05, "epoch": 0.970873786407767, "percentage": 48.55, "elapsed_time": "17:04:26", "remaining_time": "18:05:43", "throughput": 860.15, "total_tokens": 52870832}
{"current_steps": 7930, "total_steps": 16324, "loss": 0.1969, "lr": 5.646981387560638e-05, "epoch": 0.9714863250742704, "percentage": 48.58, "elapsed_time": "17:05:27", "remaining_time": "18:05:27", "throughput": 859.86, "total_tokens": 52905064}
{"current_steps": 7935, "total_steps": 16324, "loss": 0.2645, "lr": 5.641958838835858e-05, "epoch": 0.9720988637407736, "percentage": 48.61, "elapsed_time": "17:06:27", "remaining_time": "18:05:10", "throughput": 859.57, "total_tokens": 52938840}
{"current_steps": 7950, "total_steps": 16324, "eval_loss": 0.19213582575321198, "epoch": 0.9739364797402836, "percentage": 48.7, "elapsed_time": "17:09:46", "remaining_time": "18:04:42", "throughput": 858.44, "total_tokens": 53040192}
{"current_steps": 7955, "total_steps": 16324, "loss": 0.2846, "lr": 5.6218621084270076e-05, "epoch": 0.974549018406787, "percentage": 48.73, "elapsed_time": "17:10:52", "remaining_time": "18:04:31", "throughput": 858.06, "total_tokens": 53073472}
{"current_steps": 7960, "total_steps": 16324, "loss": 0.3231, "lr": 5.616836317723606e-05, "epoch": 0.9751615570732902, "percentage": 48.76, "elapsed_time": "17:11:52", "remaining_time": "18:04:14", "throughput": 857.76, "total_tokens": 53106392}
{"current_steps": 7965, "total_steps": 16324, "loss": 0.2599, "lr": 5.6118098940925365e-05, "epoch": 0.9757740957397936, "percentage": 48.79, "elapsed_time": "17:12:52", "remaining_time": "18:03:57", "throughput": 857.48, "total_tokens": 53140200}
{"current_steps": 7970, "total_steps": 16324, "loss": 0.2592, "lr": 5.606782842691352e-05, "epoch": 0.9763866344062969, "percentage": 48.82, "elapsed_time": "17:13:52", "remaining_time": "18:03:41", "throughput": 857.18, "total_tokens": 53173384}
{"current_steps": 7975, "total_steps": 16324, "loss": 0.2737, "lr": 5.60175516867824e-05, "epoch": 0.9769991730728003, "percentage": 48.85, "elapsed_time": "17:14:53", "remaining_time": "18:03:25", "throughput": 856.88, "total_tokens": 53206768}
{"current_steps": 7980, "total_steps": 16324, "loss": 0.2233, "lr": 5.596726877212032e-05, "epoch": 0.9776117117393035, "percentage": 48.89, "elapsed_time": "17:15:53", "remaining_time": "18:03:08", "throughput": 856.59, "total_tokens": 53240024}
{"current_steps": 7985, "total_steps": 16324, "loss": 0.2648, "lr": 5.5916979734521936e-05, "epoch": 0.9782242504058068, "percentage": 48.92, "elapsed_time": "17:16:53", "remaining_time": "18:02:51", "throughput": 856.29, "total_tokens": 53273088}
{"current_steps": 7990, "total_steps": 16324, "loss": 0.2603, "lr": 5.586668462558814e-05, "epoch": 0.9788367890723102, "percentage": 48.95, "elapsed_time": "17:17:55", "remaining_time": "18:02:36", "throughput": 855.99, "total_tokens": 53306896}
{"current_steps": 7995, "total_steps": 16324, "loss": 0.2012, "lr": 5.5816383496926105e-05, "epoch": 0.9794493277388135, "percentage": 48.98, "elapsed_time": "17:18:55", "remaining_time": "18:02:19", "throughput": 855.71, "total_tokens": 53341104}
{"current_steps": 8000, "total_steps": 16324, "loss": 0.2426, "lr": 5.576607640014917e-05, "epoch": 0.9800618664053168, "percentage": 49.01, "elapsed_time": "17:19:56", "remaining_time": "18:02:03", "throughput": 855.43, "total_tokens": 53375240}
{"current_steps": 8000, "total_steps": 16324, "eval_loss": 0.2315770536661148, "epoch": 0.9800618664053168, "percentage": 49.01, "elapsed_time": "17:20:15", "remaining_time": "18:02:23", "throughput": 855.16, "total_tokens": 53375240}
{"current_steps": 8005, "total_steps": 16324, "loss": 0.3245, "lr": 5.5715763386876774e-05, "epoch": 0.9806744050718201, "percentage": 49.04, "elapsed_time": "17:21:22", "remaining_time": "18:02:13", "throughput": 854.76, "total_tokens": 53407776}
{"current_steps": 8010, "total_steps": 16324, "loss": 0.2672, "lr": 5.566544450873442e-05, "epoch": 0.9812869437383235, "percentage": 49.07, "elapsed_time": "17:22:22", "remaining_time": "18:01:56", "throughput": 854.48, "total_tokens": 53441320}
{"current_steps": 8015, "total_steps": 16324, "loss": 0.2563, "lr": 5.5615119817353686e-05, "epoch": 0.9818994824048268, "percentage": 49.1, "elapsed_time": "17:23:23", "remaining_time": "18:01:39", "throughput": 854.19, "total_tokens": 53475008}
{"current_steps": 8020, "total_steps": 16324, "loss": 0.2955, "lr": 5.556478936437207e-05, "epoch": 0.9825120210713302, "percentage": 49.13, "elapsed_time": "17:24:24", "remaining_time": "18:01:23", "throughput": 853.9, "total_tokens": 53508784}
{"current_steps": 8025, "total_steps": 16324, "loss": 0.3579, "lr": 5.551445320143297e-05, "epoch": 0.9831245597378334, "percentage": 49.16, "elapsed_time": "17:25:24", "remaining_time": "18:01:05", "throughput": 853.61, "total_tokens": 53542048}
{"current_steps": 8030, "total_steps": 16324, "loss": 0.2346, "lr": 5.546411138018569e-05, "epoch": 0.9837370984043368, "percentage": 49.19, "elapsed_time": "17:26:24", "remaining_time": "18:00:48", "throughput": 853.33, "total_tokens": 53575856}
{"current_steps": 8035, "total_steps": 16324, "loss": 0.2352, "lr": 5.541376395228529e-05, "epoch": 0.9843496370708401, "percentage": 49.22, "elapsed_time": "17:27:25", "remaining_time": "18:00:32", "throughput": 853.03, "total_tokens": 53609232}
{"current_steps": 8040, "total_steps": 16324, "loss": 0.2512, "lr": 5.5363410969392637e-05, "epoch": 0.9849621757373435, "percentage": 49.25, "elapsed_time": "17:28:25", "remaining_time": "18:00:14", "throughput": 852.75, "total_tokens": 53642416}
{"current_steps": 8045, "total_steps": 16324, "loss": 0.2727, "lr": 5.531305248317422e-05, "epoch": 0.9855747144038467, "percentage": 49.28, "elapsed_time": "17:29:25", "remaining_time": "17:59:56", "throughput": 852.47, "total_tokens": 53675592}
{"current_steps": 8050, "total_steps": 16324, "loss": 0.2674, "lr": 5.526268854530228e-05, "epoch": 0.98618725307035, "percentage": 49.31, "elapsed_time": "17:30:25", "remaining_time": "17:59:39", "throughput": 852.17, "total_tokens": 53709056}
{"current_steps": 8050, "total_steps": 16324, "eval_loss": 0.1892443746328354, "epoch": 0.98618725307035, "percentage": 49.31, "elapsed_time": "17:30:45", "remaining_time": "17:59:59", "throughput": 851.91, "total_tokens": 53709056}
{"current_steps": 8055, "total_steps": 16324, "loss": 0.2445, "lr": 5.5212319207454535e-05, "epoch": 0.9867997917368534, "percentage": 49.34, "elapsed_time": "17:31:50", "remaining_time": "17:59:47", "throughput": 851.56, "total_tokens": 53742352}
{"current_steps": 8060, "total_steps": 16324, "loss": 0.2281, "lr": 5.516194452131435e-05, "epoch": 0.9874123304033567, "percentage": 49.38, "elapsed_time": "17:32:50", "remaining_time": "17:59:28", "throughput": 851.29, "total_tokens": 53775736}
{"current_steps": 8065, "total_steps": 16324, "loss": 0.251, "lr": 5.51115645385705e-05, "epoch": 0.9880248690698601, "percentage": 49.41, "elapsed_time": "17:33:51", "remaining_time": "17:59:12", "throughput": 851.0, "total_tokens": 53809504}
{"current_steps": 8070, "total_steps": 16324, "loss": 0.2299, "lr": 5.506117931091723e-05, "epoch": 0.9886374077363633, "percentage": 49.44, "elapsed_time": "17:34:50", "remaining_time": "17:58:53", "throughput": 850.74, "total_tokens": 53843688}
{"current_steps": 8075, "total_steps": 16324, "loss": 0.2477, "lr": 5.501078889005419e-05, "epoch": 0.9892499464028667, "percentage": 49.47, "elapsed_time": "17:35:49", "remaining_time": "17:58:34", "throughput": 850.48, "total_tokens": 53877968}
{"current_steps": 8080, "total_steps": 16324, "loss": 0.2732, "lr": 5.49603933276863e-05, "epoch": 0.98986248506937, "percentage": 49.5, "elapsed_time": "17:36:50", "remaining_time": "17:58:17", "throughput": 850.21, "total_tokens": 53911840}
{"current_steps": 8085, "total_steps": 16324, "loss": 0.2639, "lr": 5.4909992675523804e-05, "epoch": 0.9904750237358734, "percentage": 49.53, "elapsed_time": "17:37:50", "remaining_time": "17:57:59", "throughput": 849.93, "total_tokens": 53945424}
{"current_steps": 8090, "total_steps": 16324, "loss": 0.323, "lr": 5.485958698528213e-05, "epoch": 0.9910875624023766, "percentage": 49.56, "elapsed_time": "17:38:50", "remaining_time": "17:57:40", "throughput": 849.65, "total_tokens": 53978136}
{"current_steps": 8095, "total_steps": 16324, "loss": 0.2943, "lr": 5.480917630868192e-05, "epoch": 0.99170010106888, "percentage": 49.59, "elapsed_time": "17:39:50", "remaining_time": "17:57:23", "throughput": 849.35, "total_tokens": 54010768}
{"current_steps": 8100, "total_steps": 16324, "loss": 0.2282, "lr": 5.4758760697448886e-05, "epoch": 0.9923126397353833, "percentage": 49.62, "elapsed_time": "17:40:51", "remaining_time": "17:57:05", "throughput": 849.07, "total_tokens": 54044368}
{"current_steps": 8100, "total_steps": 16324, "eval_loss": 0.14283877611160278, "epoch": 0.9923126397353833, "percentage": 49.62, "elapsed_time": "17:41:10", "remaining_time": "17:57:25", "throughput": 848.81, "total_tokens": 54044368}
{"current_steps": 8105, "total_steps": 16324, "loss": 0.2701, "lr": 5.470834020331385e-05, "epoch": 0.9929251784018867, "percentage": 49.65, "elapsed_time": "17:42:16", "remaining_time": "17:57:12", "throughput": 848.46, "total_tokens": 54077728}
{"current_steps": 8110, "total_steps": 16324, "loss": 0.3021, "lr": 5.4657914878012604e-05, "epoch": 0.9935377170683899, "percentage": 49.68, "elapsed_time": "17:43:18", "remaining_time": "17:56:56", "throughput": 848.17, "total_tokens": 54111328}
{"current_steps": 8115, "total_steps": 16324, "loss": 0.262, "lr": 5.460748477328592e-05, "epoch": 0.9941502557348932, "percentage": 49.71, "elapsed_time": "17:44:19", "remaining_time": "17:56:39", "throughput": 847.88, "total_tokens": 54144840}
{"current_steps": 8120, "total_steps": 16324, "loss": 0.2603, "lr": 5.455704994087947e-05, "epoch": 0.9947627944013966, "percentage": 49.74, "elapsed_time": "17:45:19", "remaining_time": "17:56:21", "throughput": 847.6, "total_tokens": 54178336}
{"current_steps": 8125, "total_steps": 16324, "loss": 0.2087, "lr": 5.450661043254377e-05, "epoch": 0.9953753330678999, "percentage": 49.77, "elapsed_time": "17:46:21", "remaining_time": "17:56:03", "throughput": 847.33, "total_tokens": 54212800}
{"current_steps": 8130, "total_steps": 16324, "loss": 0.2833, "lr": 5.445616630003415e-05, "epoch": 0.9959878717344033, "percentage": 49.8, "elapsed_time": "17:47:24", "remaining_time": "17:55:48", "throughput": 847.02, "total_tokens": 54246408}
{"current_steps": 8135, "total_steps": 16324, "loss": 0.2866, "lr": 5.440571759511064e-05, "epoch": 0.9966004104009065, "percentage": 49.83, "elapsed_time": "17:48:25", "remaining_time": "17:55:30", "throughput": 846.73, "total_tokens": 54279864}
{"current_steps": 8140, "total_steps": 16324, "loss": 0.2228, "lr": 5.435526436953804e-05, "epoch": 0.9972129490674099, "percentage": 49.87, "elapsed_time": "17:49:25", "remaining_time": "17:55:12", "throughput": 846.46, "total_tokens": 54313656}
{"current_steps": 8145, "total_steps": 16324, "loss": 0.2519, "lr": 5.430480667508572e-05, "epoch": 0.9978254877339132, "percentage": 49.9, "elapsed_time": "17:50:28", "remaining_time": "17:54:56", "throughput": 846.15, "total_tokens": 54347224}
{"current_steps": 8150, "total_steps": 16324, "loss": 0.2573, "lr": 5.4254344563527646e-05, "epoch": 0.9984380264004166, "percentage": 49.93, "elapsed_time": "17:51:28", "remaining_time": "17:54:38", "throughput": 845.89, "total_tokens": 54381008}
{"current_steps": 8150, "total_steps": 16324, "eval_loss": 0.1260269433259964, "epoch": 0.9984380264004166, "percentage": 49.93, "elapsed_time": "17:51:48", "remaining_time": "17:54:57", "throughput": 845.63, "total_tokens": 54381008}
{"current_steps": 8155, "total_steps": 16324, "loss": 0.2837, "lr": 5.420387808664237e-05, "epoch": 0.9990505650669198, "percentage": 49.96, "elapsed_time": "0:02:17", "remaining_time": "0:02:17", "throughput": 396750.13, "total_tokens": 54414464}
{"current_steps": 8160, "total_steps": 16324, "loss": 0.3385, "lr": 5.415340729621283e-05, "epoch": 0.9996631037334232, "percentage": 49.99, "elapsed_time": "0:03:17", "remaining_time": "0:03:17", "throughput": 275344.1, "total_tokens": 54447944}
{"current_steps": 8165, "total_steps": 16324, "loss": 0.2689, "lr": 5.4102932244026513e-05, "epoch": 1.000367523199902, "percentage": 50.02, "elapsed_time": "0:04:26", "remaining_time": "0:04:26", "throughput": 204376.87, "total_tokens": 54487096}
{"current_steps": 8170, "total_steps": 16324, "loss": 0.2541, "lr": 5.405245298187514e-05, "epoch": 1.0009800618664053, "percentage": 50.05, "elapsed_time": "0:05:26", "remaining_time": "0:05:26", "throughput": 166866.66, "total_tokens": 54520352}
{"current_steps": 8175, "total_steps": 16324, "loss": 0.1983, "lr": 5.4001969561554856e-05, "epoch": 1.0015926005329086, "percentage": 50.08, "elapsed_time": "0:06:28", "remaining_time": "0:06:26", "throughput": 140602.86, "total_tokens": 54554976}
{"current_steps": 8180, "total_steps": 16324, "loss": 0.2281, "lr": 5.395148203486602e-05, "epoch": 1.002205139199412, "percentage": 50.11, "elapsed_time": "0:07:27", "remaining_time": "0:07:25", "throughput": 121939.08, "total_tokens": 54588840}
{"current_steps": 8185, "total_steps": 16324, "loss": 0.2676, "lr": 5.390099045361323e-05, "epoch": 1.0028176778659152, "percentage": 50.14, "elapsed_time": "0:08:27", "remaining_time": "0:08:24", "throughput": 107606.34, "total_tokens": 54622384}
{"current_steps": 8190, "total_steps": 16324, "loss": 0.2328, "lr": 5.385049486960525e-05, "epoch": 1.0034302165324187, "percentage": 50.17, "elapsed_time": "0:09:28", "remaining_time": "0:09:24", "throughput": 96164.16, "total_tokens": 54656104}
{"current_steps": 8195, "total_steps": 16324, "loss": 0.2279, "lr": 5.379999533465492e-05, "epoch": 1.004042755198922, "percentage": 50.2, "elapsed_time": "0:10:28", "remaining_time": "0:10:23", "throughput": 87007.9, "total_tokens": 54690048}
{"current_steps": 8200, "total_steps": 16324, "loss": 0.2249, "lr": 5.374949190057915e-05, "epoch": 1.0046552938654252, "percentage": 50.23, "elapsed_time": "0:11:28", "remaining_time": "0:11:21", "throughput": 79525.23, "total_tokens": 54724208}
{"current_steps": 8200, "total_steps": 16324, "eval_loss": 0.2748353183269501, "epoch": 1.0046552938654252, "percentage": 50.23, "elapsed_time": "0:12:14", "remaining_time": "0:12:07", "throughput": 74480.17, "total_tokens": 54724208}
{"current_steps": 8205, "total_steps": 16324, "loss": 0.2527, "lr": 5.369898461919888e-05, "epoch": 1.0052678325319286, "percentage": 50.26, "elapsed_time": "0:13:24", "remaining_time": "0:13:15", "throughput": 68071.26, "total_tokens": 54757616}
{"current_steps": 8210, "total_steps": 16324, "loss": 0.2637, "lr": 5.3648473542338926e-05, "epoch": 1.0058803711984319, "percentage": 50.29, "elapsed_time": "0:14:24", "remaining_time": "0:14:14", "throughput": 63401.15, "total_tokens": 54791024}
{"current_steps": 8215, "total_steps": 16324, "loss": 0.2377, "lr": 5.359795872182807e-05, "epoch": 1.0064929098649351, "percentage": 50.32, "elapsed_time": "0:15:23", "remaining_time": "0:15:11", "throughput": 59341.49, "total_tokens": 54824208}
{"current_steps": 8220, "total_steps": 16324, "loss": 0.2291, "lr": 5.35474402094989e-05, "epoch": 1.0071054485314386, "percentage": 50.36, "elapsed_time": "0:16:25", "remaining_time": "0:16:11", "throughput": 55678.24, "total_tokens": 54858152}
{"current_steps": 8225, "total_steps": 16324, "loss": 0.2697, "lr": 5.3496918057187804e-05, "epoch": 1.0077179871979418, "percentage": 50.39, "elapsed_time": "0:17:24", "remaining_time": "0:17:08", "throughput": 52530.01, "total_tokens": 54891448}
{"current_steps": 8230, "total_steps": 16324, "loss": 0.2626, "lr": 5.3446392316734894e-05, "epoch": 1.0083305258644453, "percentage": 50.42, "elapsed_time": "0:18:24", "remaining_time": "0:18:06", "throughput": 49718.98, "total_tokens": 54925096}
{"current_steps": 8235, "total_steps": 16324, "loss": 0.2661, "lr": 5.339586303998399e-05, "epoch": 1.0089430645309485, "percentage": 50.45, "elapsed_time": "0:19:25", "remaining_time": "0:19:05", "throughput": 47136.76, "total_tokens": 54958216}
{"current_steps": 8240, "total_steps": 16324, "loss": 0.2984, "lr": 5.334533027878248e-05, "epoch": 1.0095556031974517, "percentage": 50.48, "elapsed_time": "0:20:25", "remaining_time": "0:20:02", "throughput": 44854.47, "total_tokens": 54991184}
{"current_steps": 8245, "total_steps": 16324, "loss": 0.2597, "lr": 5.329479408498141e-05, "epoch": 1.0101681418639552, "percentage": 50.51, "elapsed_time": "0:21:26", "remaining_time": "0:21:00", "throughput": 42786.44, "total_tokens": 55025552}
{"current_steps": 8250, "total_steps": 16324, "loss": 0.266, "lr": 5.324425451043529e-05, "epoch": 1.0107806805304584, "percentage": 50.54, "elapsed_time": "0:22:26", "remaining_time": "0:21:57", "throughput": 40897.27, "total_tokens": 55059128}
{"current_steps": 8250, "total_steps": 16324, "eval_loss": 0.15544439852237701, "epoch": 1.0107806805304584, "percentage": 50.54, "elapsed_time": "0:22:46", "remaining_time": "0:22:17", "throughput": 40301.0, "total_tokens": 55059128}
{"current_steps": 8255, "total_steps": 16324, "loss": 0.2317, "lr": 5.3193711607002115e-05, "epoch": 1.011393219196962, "percentage": 50.57, "elapsed_time": "0:23:53", "remaining_time": "0:23:21", "throughput": 38427.24, "total_tokens": 55092984}
{"current_steps": 8260, "total_steps": 16324, "loss": 0.2701, "lr": 5.314316542654328e-05, "epoch": 1.0120057578634651, "percentage": 50.6, "elapsed_time": "0:24:53", "remaining_time": "0:24:17", "throughput": 36917.35, "total_tokens": 55126376}
{"current_steps": 8265, "total_steps": 16324, "loss": 0.2832, "lr": 5.30926160209236e-05, "epoch": 1.0126182965299684, "percentage": 50.63, "elapsed_time": "0:25:52", "remaining_time": "0:25:13", "throughput": 35525.24, "total_tokens": 55159832}
{"current_steps": 8270, "total_steps": 16324, "loss": 0.2376, "lr": 5.304206344201112e-05, "epoch": 1.0132308351964718, "percentage": 50.66, "elapsed_time": "0:26:53", "remaining_time": "0:26:11", "throughput": 34202.22, "total_tokens": 55193328}
{"current_steps": 8275, "total_steps": 16324, "loss": 0.2542, "lr": 5.2991507741677195e-05, "epoch": 1.013843373862975, "percentage": 50.69, "elapsed_time": "0:27:53", "remaining_time": "0:27:07", "throughput": 33007.21, "total_tokens": 55225936}
{"current_steps": 8280, "total_steps": 16324, "loss": 0.2407, "lr": 5.2940948971796365e-05, "epoch": 1.0144559125294785, "percentage": 50.72, "elapsed_time": "0:28:52", "remaining_time": "0:28:03", "throughput": 31894.58, "total_tokens": 55258992}
{"current_steps": 8285, "total_steps": 16324, "loss": 0.2304, "lr": 5.2890387184246324e-05, "epoch": 1.0150684511959818, "percentage": 50.75, "elapsed_time": "0:29:53", "remaining_time": "0:28:59", "throughput": 30834.77, "total_tokens": 55292976}
{"current_steps": 8290, "total_steps": 16324, "loss": 0.1923, "lr": 5.283982243090786e-05, "epoch": 1.015680989862485, "percentage": 50.78, "elapsed_time": "0:30:52", "remaining_time": "0:29:55", "throughput": 29865.31, "total_tokens": 55327720}
{"current_steps": 8295, "total_steps": 16324, "loss": 0.2865, "lr": 5.278925476366479e-05, "epoch": 1.0162935285289885, "percentage": 50.81, "elapsed_time": "0:31:51", "remaining_time": "0:30:50", "throughput": 28957.79, "total_tokens": 55361392}
{"current_steps": 8300, "total_steps": 16324, "loss": 0.2599, "lr": 5.273868423440395e-05, "epoch": 1.0169060671954917, "percentage": 50.85, "elapsed_time": "0:32:52", "remaining_time": "0:31:46", "throughput": 28087.46, "total_tokens": 55394968}
{"current_steps": 8300, "total_steps": 16324, "eval_loss": 0.2275010645389557, "epoch": 1.0169060671954917, "percentage": 50.85, "elapsed_time": "0:33:11", "remaining_time": "0:32:05", "throughput": 27814.62, "total_tokens": 55394968}
{"current_steps": 8305, "total_steps": 16324, "loss": 0.2679, "lr": 5.2688110895015096e-05, "epoch": 1.017518605861995, "percentage": 50.88, "elapsed_time": "0:34:16", "remaining_time": "0:33:05", "throughput": 26952.29, "total_tokens": 55427424}
{"current_steps": 8310, "total_steps": 16324, "loss": 0.2744, "lr": 5.263753479739085e-05, "epoch": 1.0181311445284984, "percentage": 50.91, "elapsed_time": "0:35:15", "remaining_time": "0:34:00", "throughput": 26215.53, "total_tokens": 55461776}
{"current_steps": 8315, "total_steps": 16324, "loss": 0.2556, "lr": 5.258695599342671e-05, "epoch": 1.0187436831950016, "percentage": 50.94, "elapsed_time": "0:36:16", "remaining_time": "0:34:56", "throughput": 25497.81, "total_tokens": 55494896}
{"current_steps": 8320, "total_steps": 16324, "loss": 0.2656, "lr": 5.253637453502092e-05, "epoch": 1.019356221861505, "percentage": 50.97, "elapsed_time": "0:37:15", "remaining_time": "0:35:50", "throughput": 24839.54, "total_tokens": 55528760}
{"current_steps": 8325, "total_steps": 16324, "loss": 0.2604, "lr": 5.2485790474074445e-05, "epoch": 1.0199687605280083, "percentage": 51.0, "elapsed_time": "0:38:14", "remaining_time": "0:36:44", "throughput": 24216.22, "total_tokens": 55562648}
{"current_steps": 8330, "total_steps": 16324, "loss": 0.2444, "lr": 5.243520386249092e-05, "epoch": 1.0205812991945116, "percentage": 51.03, "elapsed_time": "0:39:15", "remaining_time": "0:37:40", "throughput": 23603.92, "total_tokens": 55595648}
{"current_steps": 8335, "total_steps": 16324, "loss": 0.255, "lr": 5.238461475217662e-05, "epoch": 1.021193837861015, "percentage": 51.06, "elapsed_time": "0:40:14", "remaining_time": "0:38:34", "throughput": 23035.09, "total_tokens": 55629152}
{"current_steps": 8340, "total_steps": 16324, "loss": 0.3054, "lr": 5.233402319504035e-05, "epoch": 1.0218063765275183, "percentage": 51.09, "elapsed_time": "0:41:14", "remaining_time": "0:39:28", "throughput": 22493.75, "total_tokens": 55661832}
{"current_steps": 8345, "total_steps": 16324, "loss": 0.2577, "lr": 5.2283429242993465e-05, "epoch": 1.0224189151940217, "percentage": 51.12, "elapsed_time": "0:42:15", "remaining_time": "0:40:24", "throughput": 21966.63, "total_tokens": 55695464}
{"current_steps": 8350, "total_steps": 16324, "loss": 0.2485, "lr": 5.223283294794974e-05, "epoch": 1.023031453860525, "percentage": 51.15, "elapsed_time": "0:43:15", "remaining_time": "0:41:18", "throughput": 21473.7, "total_tokens": 55728792}
{"current_steps": 8350, "total_steps": 16324, "eval_loss": 0.22839440405368805, "epoch": 1.023031453860525, "percentage": 51.15, "elapsed_time": "0:43:34", "remaining_time": "0:41:37", "throughput": 21312.21, "total_tokens": 55728792}
{"current_steps": 8355, "total_steps": 16324, "loss": 0.2523, "lr": 5.2182234361825366e-05, "epoch": 1.0236439925270282, "percentage": 51.18, "elapsed_time": "0:44:40", "remaining_time": "0:42:36", "throughput": 20802.79, "total_tokens": 55762040}
{"current_steps": 8360, "total_steps": 16324, "loss": 0.2023, "lr": 5.213163353653892e-05, "epoch": 1.0242565311935317, "percentage": 51.21, "elapsed_time": "0:45:42", "remaining_time": "0:43:32", "throughput": 20347.4, "total_tokens": 55795784}
{"current_steps": 8365, "total_steps": 16324, "loss": 0.2392, "lr": 5.208103052401122e-05, "epoch": 1.024869069860035, "percentage": 51.24, "elapsed_time": "0:46:41", "remaining_time": "0:44:25", "throughput": 19927.35, "total_tokens": 55829712}
{"current_steps": 8370, "total_steps": 16324, "loss": 0.2684, "lr": 5.203042537616536e-05, "epoch": 1.0254816085265381, "percentage": 51.27, "elapsed_time": "0:47:41", "remaining_time": "0:45:19", "throughput": 19521.4, "total_tokens": 55863320}
{"current_steps": 8375, "total_steps": 16324, "loss": 0.3133, "lr": 5.1979818144926615e-05, "epoch": 1.0260941471930416, "percentage": 51.3, "elapsed_time": "0:48:42", "remaining_time": "0:46:13", "throughput": 19126.53, "total_tokens": 55896184}
{"current_steps": 8380, "total_steps": 16324, "loss": 0.2341, "lr": 5.1929208882222415e-05, "epoch": 1.0267066858595448, "percentage": 51.34, "elapsed_time": "0:49:41", "remaining_time": "0:47:06", "throughput": 18759.09, "total_tokens": 55930272}
{"current_steps": 8385, "total_steps": 16324, "loss": 0.2398, "lr": 5.187859763998225e-05, "epoch": 1.0273192245260483, "percentage": 51.37, "elapsed_time": "0:50:40", "remaining_time": "0:47:58", "throughput": 18405.97, "total_tokens": 55964048}
{"current_steps": 8390, "total_steps": 16324, "loss": 0.1992, "lr": 5.182798447013765e-05, "epoch": 1.0279317631925515, "percentage": 51.4, "elapsed_time": "0:51:40", "remaining_time": "0:48:51", "throughput": 18062.93, "total_tokens": 55998064}
{"current_steps": 8395, "total_steps": 16324, "loss": 0.286, "lr": 5.1777369424622135e-05, "epoch": 1.0285443018590548, "percentage": 51.43, "elapsed_time": "0:52:40", "remaining_time": "0:49:44", "throughput": 17728.94, "total_tokens": 56030752}
{"current_steps": 8400, "total_steps": 16324, "loss": 0.2513, "lr": 5.172675255537113e-05, "epoch": 1.0291568405255582, "percentage": 51.46, "elapsed_time": "0:53:39", "remaining_time": "0:50:37", "throughput": 17413.84, "total_tokens": 56064312}
{"current_steps": 8400, "total_steps": 16324, "eval_loss": 0.16583839058876038, "epoch": 1.0291568405255582, "percentage": 51.46, "elapsed_time": "0:53:59", "remaining_time": "0:50:55", "throughput": 17309.06, "total_tokens": 56064312}
{"current_steps": 8405, "total_steps": 16324, "loss": 0.2424, "lr": 5.1676133914321965e-05, "epoch": 1.0297693791920615, "percentage": 51.49, "elapsed_time": "0:55:04", "remaining_time": "0:51:53", "throughput": 16978.07, "total_tokens": 56098088}
{"current_steps": 8410, "total_steps": 16324, "loss": 0.2252, "lr": 5.162551355341373e-05, "epoch": 1.030381917858565, "percentage": 51.52, "elapsed_time": "0:56:03", "remaining_time": "0:52:45", "throughput": 16686.65, "total_tokens": 56132272}
{"current_steps": 8415, "total_steps": 16324, "loss": 0.2803, "lr": 5.1574891524587333e-05, "epoch": 1.0309944565250682, "percentage": 51.55, "elapsed_time": "0:57:02", "remaining_time": "0:53:37", "throughput": 16408.74, "total_tokens": 56165256}
{"current_steps": 8420, "total_steps": 16324, "loss": 0.229, "lr": 5.1524267879785374e-05, "epoch": 1.0316069951915714, "percentage": 51.58, "elapsed_time": "0:58:01", "remaining_time": "0:54:28", "throughput": 16141.15, "total_tokens": 56198512}
{"current_steps": 8425, "total_steps": 16324, "loss": 0.2367, "lr": 5.1473642670952114e-05, "epoch": 1.0322195338580749, "percentage": 51.61, "elapsed_time": "0:59:02", "remaining_time": "0:55:21", "throughput": 15874.58, "total_tokens": 56231456}
{"current_steps": 8430, "total_steps": 16324, "loss": 0.2478, "lr": 5.142301595003343e-05, "epoch": 1.032832072524578, "percentage": 51.64, "elapsed_time": "1:00:01", "remaining_time": "0:56:12", "throughput": 15624.73, "total_tokens": 56265120}
{"current_steps": 8435, "total_steps": 16324, "loss": 0.2885, "lr": 5.1372387768976694e-05, "epoch": 1.0334446111910813, "percentage": 51.67, "elapsed_time": "1:01:00", "remaining_time": "0:57:03", "throughput": 15382.16, "total_tokens": 56298848}
{"current_steps": 8440, "total_steps": 16324, "loss": 0.2662, "lr": 5.132175817973086e-05, "epoch": 1.0340571498575848, "percentage": 51.7, "elapsed_time": "1:02:00", "remaining_time": "0:57:55", "throughput": 15140.27, "total_tokens": 56331816}
{"current_steps": 8445, "total_steps": 16324, "loss": 0.2736, "lr": 5.1271127234246275e-05, "epoch": 1.034669688524088, "percentage": 51.73, "elapsed_time": "1:02:59", "remaining_time": "0:58:46", "throughput": 14913.5, "total_tokens": 56366040}
{"current_steps": 8450, "total_steps": 16324, "loss": 0.2644, "lr": 5.1220494984474674e-05, "epoch": 1.0352822271905915, "percentage": 51.76, "elapsed_time": "1:03:58", "remaining_time": "0:59:36", "throughput": 14692.67, "total_tokens": 56398904}
{"current_steps": 8450, "total_steps": 16324, "eval_loss": 0.17673595249652863, "epoch": 1.0352822271905915, "percentage": 51.76, "elapsed_time": "1:04:17", "remaining_time": "0:59:54", "throughput": 14619.27, "total_tokens": 56398904}
{"current_steps": 8455, "total_steps": 16324, "loss": 0.2463, "lr": 5.116986148236916e-05, "epoch": 1.0358947658570947, "percentage": 51.79, "elapsed_time": "1:05:24", "remaining_time": "1:00:52", "throughput": 14379.92, "total_tokens": 56432392}
{"current_steps": 8460, "total_steps": 16324, "loss": 0.2756, "lr": 5.1119226779884113e-05, "epoch": 1.036507304523598, "percentage": 51.83, "elapsed_time": "1:06:23", "remaining_time": "1:01:42", "throughput": 14174.93, "total_tokens": 56465048}
{"current_steps": 8465, "total_steps": 16324, "loss": 0.3336, "lr": 5.106859092897514e-05, "epoch": 1.0371198431901014, "percentage": 51.86, "elapsed_time": "1:07:22", "remaining_time": "1:02:33", "throughput": 13975.26, "total_tokens": 56498760}
{"current_steps": 8470, "total_steps": 16324, "loss": 0.2393, "lr": 5.1017953981598974e-05, "epoch": 1.0377323818566047, "percentage": 51.89, "elapsed_time": "1:08:22", "remaining_time": "1:03:24", "throughput": 13778.46, "total_tokens": 56532000}
{"current_steps": 8475, "total_steps": 16324, "loss": 0.3168, "lr": 5.096731598971359e-05, "epoch": 1.0383449205231081, "percentage": 51.92, "elapsed_time": "1:09:21", "remaining_time": "1:04:14", "throughput": 13591.16, "total_tokens": 56565360}
{"current_steps": 8480, "total_steps": 16324, "loss": 0.203, "lr": 5.0916677005277934e-05, "epoch": 1.0389574591896114, "percentage": 51.95, "elapsed_time": "1:10:21", "remaining_time": "1:05:04", "throughput": 13407.7, "total_tokens": 56599504}
{"current_steps": 8485, "total_steps": 16324, "loss": 0.3049, "lr": 5.086603708025201e-05, "epoch": 1.0395699978561146, "percentage": 51.98, "elapsed_time": "1:11:22", "remaining_time": "1:05:56", "throughput": 13225.03, "total_tokens": 56631704}
{"current_steps": 8490, "total_steps": 16324, "loss": 0.2823, "lr": 5.0815396266596794e-05, "epoch": 1.040182536522618, "percentage": 52.01, "elapsed_time": "1:12:21", "remaining_time": "1:06:46", "throughput": 13051.21, "total_tokens": 56665032}
{"current_steps": 8495, "total_steps": 16324, "loss": 0.2552, "lr": 5.076475461627415e-05, "epoch": 1.0407950751891213, "percentage": 52.04, "elapsed_time": "1:13:20", "remaining_time": "1:07:35", "throughput": 12883.66, "total_tokens": 56698112}
{"current_steps": 8500, "total_steps": 16324, "loss": 0.206, "lr": 5.07141121812468e-05, "epoch": 1.0414076138556245, "percentage": 52.07, "elapsed_time": "1:14:21", "remaining_time": "1:08:26", "throughput": 12716.5, "total_tokens": 56732240}
{"current_steps": 8500, "total_steps": 16324, "eval_loss": 0.16079044342041016, "epoch": 1.0414076138556245, "percentage": 52.07, "elapsed_time": "1:14:40", "remaining_time": "1:08:44", "throughput": 12662.36, "total_tokens": 56732240}
{"current_steps": 8505, "total_steps": 16324, "loss": 0.2222, "lr": 5.06634690134783e-05, "epoch": 1.042020152522128, "percentage": 52.1, "elapsed_time": "1:15:43", "remaining_time": "1:09:37", "throughput": 12493.31, "total_tokens": 56766312}
{"current_steps": 8510, "total_steps": 16324, "loss": 0.2588, "lr": 5.061282516493294e-05, "epoch": 1.0426326911886312, "percentage": 52.13, "elapsed_time": "1:16:42", "remaining_time": "1:10:26", "throughput": 12341.12, "total_tokens": 56799880}
{"current_steps": 8515, "total_steps": 16324, "loss": 0.2503, "lr": 5.05621806875757e-05, "epoch": 1.0432452298551347, "percentage": 52.16, "elapsed_time": "1:17:42", "remaining_time": "1:11:16", "throughput": 12188.28, "total_tokens": 56833000}
{"current_steps": 8520, "total_steps": 16324, "loss": 0.2596, "lr": 5.051153563337222e-05, "epoch": 1.043857768521638, "percentage": 52.19, "elapsed_time": "1:18:41", "remaining_time": "1:12:05", "throughput": 12043.01, "total_tokens": 56867064}
{"current_steps": 8525, "total_steps": 16324, "loss": 0.1845, "lr": 5.046089005428872e-05, "epoch": 1.0444703071881412, "percentage": 52.22, "elapsed_time": "1:19:40", "remaining_time": "1:12:53", "throughput": 11902.44, "total_tokens": 56900952}
{"current_steps": 8530, "total_steps": 16324, "loss": 0.2324, "lr": 5.0410244002292004e-05, "epoch": 1.0450828458546446, "percentage": 52.25, "elapsed_time": "1:20:40", "remaining_time": "1:13:42", "throughput": 11762.05, "total_tokens": 56934104}
{"current_steps": 8535, "total_steps": 16324, "loss": 0.2302, "lr": 5.0359597529349266e-05, "epoch": 1.0456953845211479, "percentage": 52.28, "elapsed_time": "1:21:40", "remaining_time": "1:14:31", "throughput": 11625.63, "total_tokens": 56967848}
{"current_steps": 8540, "total_steps": 16324, "loss": 0.2654, "lr": 5.030895068742822e-05, "epoch": 1.0463079231876513, "percentage": 52.32, "elapsed_time": "1:22:39", "remaining_time": "1:15:20", "throughput": 11493.81, "total_tokens": 57001056}
{"current_steps": 8545, "total_steps": 16324, "loss": 0.2233, "lr": 5.025830352849693e-05, "epoch": 1.0469204618541545, "percentage": 52.35, "elapsed_time": "1:23:39", "remaining_time": "1:16:09", "throughput": 11363.75, "total_tokens": 57035056}
{"current_steps": 8550, "total_steps": 16324, "loss": 0.2414, "lr": 5.020765610452375e-05, "epoch": 1.0475330005206578, "percentage": 52.38, "elapsed_time": "1:24:39", "remaining_time": "1:16:58", "throughput": 11235.96, "total_tokens": 57068112}
{"current_steps": 8550, "total_steps": 16324, "eval_loss": 0.21543771028518677, "epoch": 1.0475330005206578, "percentage": 52.38, "elapsed_time": "1:24:58", "remaining_time": "1:17:15", "throughput": 11193.02, "total_tokens": 57068112}
{"current_steps": 8555, "total_steps": 16324, "loss": 0.2635, "lr": 5.0157008467477385e-05, "epoch": 1.0481455391871612, "percentage": 52.41, "elapsed_time": "1:26:03", "remaining_time": "1:18:09", "throughput": 11058.77, "total_tokens": 57101496}
{"current_steps": 8560, "total_steps": 16324, "loss": 0.233, "lr": 5.010636066932667e-05, "epoch": 1.0487580778536645, "percentage": 52.44, "elapsed_time": "1:27:03", "remaining_time": "1:18:57", "throughput": 10938.51, "total_tokens": 57135456}
{"current_steps": 8565, "total_steps": 16324, "loss": 0.2612, "lr": 5.0055712762040674e-05, "epoch": 1.049370616520168, "percentage": 52.47, "elapsed_time": "1:28:03", "remaining_time": "1:19:46", "throughput": 10820.59, "total_tokens": 57168912}
{"current_steps": 8570, "total_steps": 16324, "loss": 0.2731, "lr": 5.000506479758854e-05, "epoch": 1.0499831551866712, "percentage": 52.5, "elapsed_time": "1:29:02", "remaining_time": "1:20:34", "throughput": 10706.29, "total_tokens": 57202432}
{"current_steps": 8575, "total_steps": 16324, "loss": 0.2406, "lr": 4.995441682793949e-05, "epoch": 1.0505956938531744, "percentage": 52.53, "elapsed_time": "1:30:02", "remaining_time": "1:21:22", "throughput": 10593.77, "total_tokens": 57235744}
{"current_steps": 8580, "total_steps": 16324, "loss": 0.2107, "lr": 4.990376890506273e-05, "epoch": 1.0512082325196779, "percentage": 52.56, "elapsed_time": "1:31:03", "remaining_time": "1:22:11", "throughput": 10481.38, "total_tokens": 57269840}
{"current_steps": 8585, "total_steps": 16324, "loss": 0.2, "lr": 4.9853121080927435e-05, "epoch": 1.0518207711861811, "percentage": 52.59, "elapsed_time": "1:32:03", "remaining_time": "1:22:59", "throughput": 10374.78, "total_tokens": 57304016}
{"current_steps": 8590, "total_steps": 16324, "loss": 0.2134, "lr": 4.980247340750268e-05, "epoch": 1.0524333098526844, "percentage": 52.62, "elapsed_time": "1:33:03", "remaining_time": "1:23:46", "throughput": 10269.48, "total_tokens": 57337792}
{"current_steps": 8595, "total_steps": 16324, "loss": 0.2785, "lr": 4.975182593675736e-05, "epoch": 1.0530458485191878, "percentage": 52.65, "elapsed_time": "1:34:04", "remaining_time": "1:24:35", "throughput": 10164.0, "total_tokens": 57371248}
{"current_steps": 8600, "total_steps": 16324, "loss": 0.2406, "lr": 4.970117872066021e-05, "epoch": 1.053658387185691, "percentage": 52.68, "elapsed_time": "1:35:05", "remaining_time": "1:25:24", "throughput": 10061.56, "total_tokens": 57404752}
{"current_steps": 8600, "total_steps": 16324, "eval_loss": 0.15721498429775238, "epoch": 1.053658387185691, "percentage": 52.68, "elapsed_time": "1:35:25", "remaining_time": "1:25:42", "throughput": 10025.94, "total_tokens": 57404752}
{"current_steps": 8605, "total_steps": 16324, "loss": 0.2482, "lr": 4.965053181117965e-05, "epoch": 1.0542709258521945, "percentage": 52.71, "elapsed_time": "1:36:31", "remaining_time": "1:26:35", "throughput": 9917.03, "total_tokens": 57438376}
{"current_steps": 8610, "total_steps": 16324, "loss": 0.2265, "lr": 4.959988526028384e-05, "epoch": 1.0548834645186977, "percentage": 52.74, "elapsed_time": "1:37:33", "remaining_time": "1:27:23", "throughput": 9819.19, "total_tokens": 57471696}
{"current_steps": 8615, "total_steps": 16324, "loss": 0.2478, "lr": 4.954923911994051e-05, "epoch": 1.055496003185201, "percentage": 52.78, "elapsed_time": "1:38:32", "remaining_time": "1:28:10", "throughput": 9726.79, "total_tokens": 57505344}
{"current_steps": 8620, "total_steps": 16324, "loss": 0.2196, "lr": 4.9498593442117e-05, "epoch": 1.0561085418517044, "percentage": 52.81, "elapsed_time": "1:39:30", "remaining_time": "1:28:56", "throughput": 9636.78, "total_tokens": 57538944}
{"current_steps": 8625, "total_steps": 16324, "loss": 0.2548, "lr": 4.9447948278780216e-05, "epoch": 1.0567210805182077, "percentage": 52.84, "elapsed_time": "1:40:30", "remaining_time": "1:29:43", "throughput": 9546.77, "total_tokens": 57572168}
{"current_steps": 8630, "total_steps": 16324, "loss": 0.2524, "lr": 4.939730368189647e-05, "epoch": 1.0573336191847111, "percentage": 52.87, "elapsed_time": "1:41:29", "remaining_time": "1:30:28", "throughput": 9459.86, "total_tokens": 57605056}
{"current_steps": 8635, "total_steps": 16324, "loss": 0.3168, "lr": 4.934665970343151e-05, "epoch": 1.0579461578512144, "percentage": 52.9, "elapsed_time": "1:42:28", "remaining_time": "1:31:14", "throughput": 9374.41, "total_tokens": 57638896}
{"current_steps": 8640, "total_steps": 16324, "loss": 0.2335, "lr": 4.929601639535047e-05, "epoch": 1.0585586965177176, "percentage": 52.93, "elapsed_time": "1:43:28", "remaining_time": "1:32:01", "throughput": 9289.27, "total_tokens": 57672136}
{"current_steps": 8645, "total_steps": 16324, "loss": 0.2285, "lr": 4.924537380961779e-05, "epoch": 1.059171235184221, "percentage": 52.96, "elapsed_time": "1:44:27", "remaining_time": "1:32:47", "throughput": 9207.25, "total_tokens": 57705352}
{"current_steps": 8650, "total_steps": 16324, "loss": 0.2487, "lr": 4.919473199819715e-05, "epoch": 1.0597837738507243, "percentage": 52.99, "elapsed_time": "1:45:26", "remaining_time": "1:33:32", "throughput": 9126.74, "total_tokens": 57738424}
{"current_steps": 8650, "total_steps": 16324, "eval_loss": 0.1030297502875328, "epoch": 1.0597837738507243, "percentage": 52.99, "elapsed_time": "1:45:45", "remaining_time": "1:33:49", "throughput": 9098.63, "total_tokens": 57738424}
{"current_steps": 8655, "total_steps": 16324, "loss": 0.2628, "lr": 4.914409101305146e-05, "epoch": 1.0603963125172275, "percentage": 53.02, "elapsed_time": "1:46:50", "remaining_time": "1:34:40", "throughput": 9011.66, "total_tokens": 57771856}
{"current_steps": 8660, "total_steps": 16324, "loss": 0.2704, "lr": 4.9093450906142765e-05, "epoch": 1.061008851183731, "percentage": 53.05, "elapsed_time": "1:47:51", "remaining_time": "1:35:27", "throughput": 8932.5, "total_tokens": 57804920}
{"current_steps": 8665, "total_steps": 16324, "loss": 0.2324, "lr": 4.904281172943223e-05, "epoch": 1.0616213898502342, "percentage": 53.08, "elapsed_time": "1:48:50", "remaining_time": "1:36:12", "throughput": 8856.3, "total_tokens": 57838728}
{"current_steps": 8670, "total_steps": 16324, "loss": 0.246, "lr": 4.899217353488004e-05, "epoch": 1.0622339285167377, "percentage": 53.11, "elapsed_time": "1:49:50", "remaining_time": "1:36:58", "throughput": 8781.24, "total_tokens": 57871664}
{"current_steps": 8675, "total_steps": 16324, "loss": 0.2308, "lr": 4.894153637444537e-05, "epoch": 1.062846467183241, "percentage": 53.14, "elapsed_time": "1:50:51", "remaining_time": "1:37:44", "throughput": 8705.63, "total_tokens": 57905416}
{"current_steps": 8680, "total_steps": 16324, "loss": 0.2462, "lr": 4.889090030008638e-05, "epoch": 1.0634590058497442, "percentage": 53.17, "elapsed_time": "1:51:50", "remaining_time": "1:38:29", "throughput": 8633.6, "total_tokens": 57938832}
{"current_steps": 8685, "total_steps": 16324, "loss": 0.2348, "lr": 4.8840265363760046e-05, "epoch": 1.0640715445162476, "percentage": 53.2, "elapsed_time": "1:52:50", "remaining_time": "1:39:15", "throughput": 8561.99, "total_tokens": 57972576}
{"current_steps": 8690, "total_steps": 16324, "loss": 0.2228, "lr": 4.878963161742224e-05, "epoch": 1.0646840831827509, "percentage": 53.23, "elapsed_time": "1:53:51", "remaining_time": "1:40:01", "throughput": 8491.12, "total_tokens": 58006624}
{"current_steps": 8695, "total_steps": 16324, "loss": 0.2831, "lr": 4.873899911302758e-05, "epoch": 1.0652966218492543, "percentage": 53.27, "elapsed_time": "1:54:50", "remaining_time": "1:40:45", "throughput": 8422.9, "total_tokens": 58040088}
{"current_steps": 8700, "total_steps": 16324, "loss": 0.2314, "lr": 4.86883679025294e-05, "epoch": 1.0659091605157576, "percentage": 53.3, "elapsed_time": "1:55:51", "remaining_time": "1:41:31", "throughput": 8354.63, "total_tokens": 58073704}
{"current_steps": 8700, "total_steps": 16324, "eval_loss": 0.1421510875225067, "epoch": 1.0659091605157576, "percentage": 53.3, "elapsed_time": "1:56:10", "remaining_time": "1:41:48", "throughput": 8331.08, "total_tokens": 58073704}
{"current_steps": 8705, "total_steps": 16324, "loss": 0.2336, "lr": 4.8637738037879734e-05, "epoch": 1.0665216991822608, "percentage": 53.33, "elapsed_time": "1:57:16", "remaining_time": "1:42:38", "throughput": 8258.21, "total_tokens": 58106888}
{"current_steps": 8710, "total_steps": 16324, "loss": 0.2186, "lr": 4.8587109571029246e-05, "epoch": 1.0671342378487643, "percentage": 53.36, "elapsed_time": "1:58:15", "remaining_time": "1:43:22", "throughput": 8193.59, "total_tokens": 58140584}
{"current_steps": 8715, "total_steps": 16324, "loss": 0.2173, "lr": 4.853648255392709e-05, "epoch": 1.0677467765152675, "percentage": 53.39, "elapsed_time": "1:59:16", "remaining_time": "1:44:07", "throughput": 8129.37, "total_tokens": 58174072}
{"current_steps": 8720, "total_steps": 16324, "loss": 0.2116, "lr": 4.848585703852103e-05, "epoch": 1.068359315181771, "percentage": 53.42, "elapsed_time": "2:00:16", "remaining_time": "1:44:52", "throughput": 8066.17, "total_tokens": 58207808}
{"current_steps": 8725, "total_steps": 16324, "loss": 0.2611, "lr": 4.843523307675723e-05, "epoch": 1.0689718538482742, "percentage": 53.45, "elapsed_time": "2:01:15", "remaining_time": "1:45:36", "throughput": 8005.01, "total_tokens": 58241136}
{"current_steps": 8730, "total_steps": 16324, "loss": 0.2031, "lr": 4.838461072058025e-05, "epoch": 1.0695843925147774, "percentage": 53.48, "elapsed_time": "2:02:14", "remaining_time": "1:46:20", "throughput": 7945.33, "total_tokens": 58275600}
{"current_steps": 8735, "total_steps": 16324, "loss": 0.2438, "lr": 4.8333990021933065e-05, "epoch": 1.070196931181281, "percentage": 53.51, "elapsed_time": "2:03:14", "remaining_time": "1:47:04", "throughput": 7885.42, "total_tokens": 58309360}
{"current_steps": 8740, "total_steps": 16324, "loss": 0.2834, "lr": 4.8283371032756885e-05, "epoch": 1.0708094698477841, "percentage": 53.54, "elapsed_time": "2:04:13", "remaining_time": "1:47:47", "throughput": 7827.56, "total_tokens": 58342896}
{"current_steps": 8745, "total_steps": 16324, "loss": 0.3, "lr": 4.823275380499121e-05, "epoch": 1.0714220085142874, "percentage": 53.57, "elapsed_time": "2:05:12", "remaining_time": "1:48:30", "throughput": 7770.32, "total_tokens": 58375272}
{"current_steps": 8750, "total_steps": 16324, "loss": 0.2452, "lr": 4.818213839057371e-05, "epoch": 1.0720345471807908, "percentage": 53.6, "elapsed_time": "2:06:12", "remaining_time": "1:49:14", "throughput": 7713.22, "total_tokens": 58409256}
{"current_steps": 8750, "total_steps": 16324, "eval_loss": 0.21566660702228546, "epoch": 1.0720345471807908, "percentage": 53.6, "elapsed_time": "2:06:31", "remaining_time": "1:49:31", "throughput": 7693.67, "total_tokens": 58409256}
{"current_steps": 8755, "total_steps": 16324, "loss": 0.2479, "lr": 4.8131524841440176e-05, "epoch": 1.072647085847294, "percentage": 53.63, "elapsed_time": "2:07:36", "remaining_time": "1:50:18", "throughput": 7633.51, "total_tokens": 58442384}
{"current_steps": 8760, "total_steps": 16324, "loss": 0.2227, "lr": 4.8080913209524534e-05, "epoch": 1.0732596245137975, "percentage": 53.66, "elapsed_time": "2:08:35", "remaining_time": "1:51:01", "throughput": 7579.35, "total_tokens": 58475920}
{"current_steps": 8765, "total_steps": 16324, "loss": 0.2618, "lr": 4.8030303546758705e-05, "epoch": 1.0738721631803008, "percentage": 53.69, "elapsed_time": "2:09:34", "remaining_time": "1:51:45", "throughput": 7525.34, "total_tokens": 58509144}
{"current_steps": 8770, "total_steps": 16324, "loss": 0.2381, "lr": 4.797969590507261e-05, "epoch": 1.074484701846804, "percentage": 53.72, "elapsed_time": "2:10:34", "remaining_time": "1:52:28", "throughput": 7472.53, "total_tokens": 58542680}
{"current_steps": 8775, "total_steps": 16324, "loss": 0.2139, "lr": 4.792909033639409e-05, "epoch": 1.0750972405133075, "percentage": 53.76, "elapsed_time": "2:11:33", "remaining_time": "1:53:10", "throughput": 7421.24, "total_tokens": 58576552}
{"current_steps": 8780, "total_steps": 16324, "loss": 0.2487, "lr": 4.787848689264883e-05, "epoch": 1.0757097791798107, "percentage": 53.79, "elapsed_time": "2:12:32", "remaining_time": "1:53:53", "throughput": 7369.75, "total_tokens": 58609576}
{"current_steps": 8785, "total_steps": 16324, "loss": 0.21, "lr": 4.7827885625760385e-05, "epoch": 1.076322317846314, "percentage": 53.82, "elapsed_time": "2:13:32", "remaining_time": "1:54:36", "throughput": 7318.8, "total_tokens": 58643352}
{"current_steps": 8790, "total_steps": 16324, "loss": 0.2483, "lr": 4.777728658765007e-05, "epoch": 1.0769348565128174, "percentage": 53.85, "elapsed_time": "2:14:31", "remaining_time": "1:55:18", "throughput": 7269.39, "total_tokens": 58677304}
{"current_steps": 8795, "total_steps": 16324, "loss": 0.2495, "lr": 4.7726689830236834e-05, "epoch": 1.0775473951793206, "percentage": 53.88, "elapsed_time": "2:15:31", "remaining_time": "1:56:00", "throughput": 7220.35, "total_tokens": 58710320}
{"current_steps": 8800, "total_steps": 16324, "loss": 0.2578, "lr": 4.76760954054374e-05, "epoch": 1.078159933845824, "percentage": 53.91, "elapsed_time": "2:16:31", "remaining_time": "1:56:43", "throughput": 7171.14, "total_tokens": 58743680}
{"current_steps": 8800, "total_steps": 16324, "eval_loss": 0.13818946480751038, "epoch": 1.078159933845824, "percentage": 53.91, "elapsed_time": "2:16:51", "remaining_time": "1:57:00", "throughput": 7154.01, "total_tokens": 58743680}
{"current_steps": 8805, "total_steps": 16324, "loss": 0.2521, "lr": 4.762550336516603e-05, "epoch": 1.0787724725123273, "percentage": 53.94, "elapsed_time": "2:17:56", "remaining_time": "1:57:47", "throughput": 7101.88, "total_tokens": 58777032}
{"current_steps": 8810, "total_steps": 16324, "loss": 0.2287, "lr": 4.757491376133454e-05, "epoch": 1.0793850111788306, "percentage": 53.97, "elapsed_time": "2:18:56", "remaining_time": "1:58:30", "throughput": 7054.59, "total_tokens": 58811392}
{"current_steps": 8815, "total_steps": 16324, "loss": 0.2349, "lr": 4.752432664585228e-05, "epoch": 1.079997549845334, "percentage": 54.0, "elapsed_time": "2:19:57", "remaining_time": "1:59:13", "throughput": 7007.73, "total_tokens": 58844584}
{"current_steps": 8820, "total_steps": 16324, "loss": 0.2907, "lr": 4.747374207062601e-05, "epoch": 1.0806100885118373, "percentage": 54.03, "elapsed_time": "2:20:56", "remaining_time": "1:59:54", "throughput": 6962.19, "total_tokens": 58877152}
{"current_steps": 8825, "total_steps": 16324, "loss": 0.2687, "lr": 4.742316008755992e-05, "epoch": 1.0812226271783407, "percentage": 54.06, "elapsed_time": "2:21:56", "remaining_time": "2:00:37", "throughput": 6916.85, "total_tokens": 58910592}
{"current_steps": 8830, "total_steps": 16324, "loss": 0.2837, "lr": 4.73725807485555e-05, "epoch": 1.081835165844844, "percentage": 54.09, "elapsed_time": "2:22:57", "remaining_time": "2:01:19", "throughput": 6871.66, "total_tokens": 58943496}
{"current_steps": 8835, "total_steps": 16324, "loss": 0.2229, "lr": 4.732200410551155e-05, "epoch": 1.0824477045113472, "percentage": 54.12, "elapsed_time": "2:23:57", "remaining_time": "2:02:01", "throughput": 6827.79, "total_tokens": 58976320}
{"current_steps": 8840, "total_steps": 16324, "loss": 0.2708, "lr": 4.727143021032412e-05, "epoch": 1.0830602431778507, "percentage": 54.15, "elapsed_time": "2:24:57", "remaining_time": "2:02:43", "throughput": 6784.58, "total_tokens": 59010264}
{"current_steps": 8845, "total_steps": 16324, "loss": 0.2271, "lr": 4.7220859114886394e-05, "epoch": 1.083672781844354, "percentage": 54.18, "elapsed_time": "2:25:57", "remaining_time": "2:03:25", "throughput": 6742.1, "total_tokens": 59043984}
{"current_steps": 8850, "total_steps": 16324, "loss": 0.2266, "lr": 4.7170290871088734e-05, "epoch": 1.0842853205108574, "percentage": 54.21, "elapsed_time": "2:26:56", "remaining_time": "2:04:05", "throughput": 6700.94, "total_tokens": 59077880}
{"current_steps": 8850, "total_steps": 16324, "eval_loss": 0.1759234070777893, "epoch": 1.0842853205108574, "percentage": 54.21, "elapsed_time": "2:27:15", "remaining_time": "2:04:21", "throughput": 6686.55, "total_tokens": 59077880}
{"current_steps": 8855, "total_steps": 16324, "loss": 0.2391, "lr": 4.711972553081855e-05, "epoch": 1.0848978591773606, "percentage": 54.25, "elapsed_time": "2:28:19", "remaining_time": "2:05:06", "throughput": 6641.75, "total_tokens": 59111400}
{"current_steps": 8860, "total_steps": 16324, "loss": 0.2489, "lr": 4.706916314596025e-05, "epoch": 1.0855103978438638, "percentage": 54.28, "elapsed_time": "2:29:19", "remaining_time": "2:05:48", "throughput": 6601.11, "total_tokens": 59145032}
{"current_steps": 8865, "total_steps": 16324, "loss": 0.27, "lr": 4.701860376839526e-05, "epoch": 1.0861229365103673, "percentage": 54.31, "elapsed_time": "2:30:18", "remaining_time": "2:06:28", "throughput": 6561.76, "total_tokens": 59178760}
{"current_steps": 8870, "total_steps": 16324, "loss": 0.2261, "lr": 4.696804745000188e-05, "epoch": 1.0867354751768705, "percentage": 54.34, "elapsed_time": "2:31:17", "remaining_time": "2:07:08", "throughput": 6522.72, "total_tokens": 59212424}
{"current_steps": 8875, "total_steps": 16324, "loss": 0.2284, "lr": 4.6917494242655325e-05, "epoch": 1.0873480138433738, "percentage": 54.37, "elapsed_time": "2:32:18", "remaining_time": "2:07:50", "throughput": 6482.95, "total_tokens": 59245624}
{"current_steps": 8880, "total_steps": 16324, "loss": 0.2783, "lr": 4.6866944198227526e-05, "epoch": 1.0879605525098772, "percentage": 54.4, "elapsed_time": "2:33:17", "remaining_time": "2:08:30", "throughput": 6445.15, "total_tokens": 59278912}
{"current_steps": 8885, "total_steps": 16324, "loss": 0.2173, "lr": 4.681639736858725e-05, "epoch": 1.0885730911763805, "percentage": 54.43, "elapsed_time": "2:34:16", "remaining_time": "2:09:09", "throughput": 6407.89, "total_tokens": 59312872}
{"current_steps": 8890, "total_steps": 16324, "loss": 0.249, "lr": 4.6765853805599914e-05, "epoch": 1.089185629842884, "percentage": 54.46, "elapsed_time": "2:35:16", "remaining_time": "2:09:50", "throughput": 6370.27, "total_tokens": 59345944}
{"current_steps": 8895, "total_steps": 16324, "loss": 0.2639, "lr": 4.6715313561127634e-05, "epoch": 1.0897981685093872, "percentage": 54.49, "elapsed_time": "2:36:14", "remaining_time": "2:10:29", "throughput": 6333.87, "total_tokens": 59379368}
{"current_steps": 8900, "total_steps": 16324, "loss": 0.2321, "lr": 4.666477668702907e-05, "epoch": 1.0904107071758904, "percentage": 54.52, "elapsed_time": "2:37:13", "remaining_time": "2:11:09", "throughput": 6297.91, "total_tokens": 59412496}
{"current_steps": 8900, "total_steps": 16324, "eval_loss": 0.15350688993930817, "epoch": 1.0904107071758904, "percentage": 54.52, "elapsed_time": "2:37:32", "remaining_time": "2:11:25", "throughput": 6285.05, "total_tokens": 59412496}
{"current_steps": 8905, "total_steps": 16324, "loss": 0.236, "lr": 4.661424323515946e-05, "epoch": 1.0910232458423939, "percentage": 54.55, "elapsed_time": "2:38:37", "remaining_time": "2:12:09", "throughput": 6245.86, "total_tokens": 59445248}
{"current_steps": 8910, "total_steps": 16324, "loss": 0.3239, "lr": 4.6563713257370526e-05, "epoch": 1.091635784508897, "percentage": 54.58, "elapsed_time": "2:39:37", "remaining_time": "2:12:49", "throughput": 6210.04, "total_tokens": 59478080}
{"current_steps": 8915, "total_steps": 16324, "loss": 0.2349, "lr": 4.651318680551039e-05, "epoch": 1.0922483231754005, "percentage": 54.61, "elapsed_time": "2:40:36", "remaining_time": "2:13:28", "throughput": 6175.63, "total_tokens": 59511552}
{"current_steps": 8920, "total_steps": 16324, "loss": 0.2413, "lr": 4.646266393142361e-05, "epoch": 1.0928608618419038, "percentage": 54.64, "elapsed_time": "2:41:36", "remaining_time": "2:14:08", "throughput": 6140.96, "total_tokens": 59545072}
{"current_steps": 8925, "total_steps": 16324, "loss": 0.2575, "lr": 4.6412144686951014e-05, "epoch": 1.093473400508407, "percentage": 54.67, "elapsed_time": "2:42:36", "remaining_time": "2:14:48", "throughput": 6106.39, "total_tokens": 59578112}
{"current_steps": 8930, "total_steps": 16324, "loss": 0.2172, "lr": 4.6361629123929776e-05, "epoch": 1.0940859391749105, "percentage": 54.7, "elapsed_time": "2:43:35", "remaining_time": "2:15:27", "throughput": 6073.08, "total_tokens": 59611888}
{"current_steps": 8935, "total_steps": 16324, "loss": 0.2573, "lr": 4.6311117294193216e-05, "epoch": 1.0946984778414137, "percentage": 54.74, "elapsed_time": "2:44:35", "remaining_time": "2:16:06", "throughput": 6039.94, "total_tokens": 59645488}
{"current_steps": 8940, "total_steps": 16324, "loss": 0.2463, "lr": 4.626060924957088e-05, "epoch": 1.095311016507917, "percentage": 54.77, "elapsed_time": "2:45:35", "remaining_time": "2:16:46", "throughput": 6006.8, "total_tokens": 59679080}
{"current_steps": 8945, "total_steps": 16324, "loss": 0.2196, "lr": 4.6210105041888415e-05, "epoch": 1.0959235551744204, "percentage": 54.8, "elapsed_time": "2:46:34", "remaining_time": "2:17:24", "throughput": 5974.45, "total_tokens": 59713224}
{"current_steps": 8950, "total_steps": 16324, "loss": 0.2312, "lr": 4.615960472296751e-05, "epoch": 1.0965360938409237, "percentage": 54.83, "elapsed_time": "2:47:34", "remaining_time": "2:18:04", "throughput": 5942.03, "total_tokens": 59746528}
{"current_steps": 8950, "total_steps": 16324, "eval_loss": 0.14111146330833435, "epoch": 1.0965360938409237, "percentage": 54.83, "elapsed_time": "2:47:54", "remaining_time": "2:18:20", "throughput": 5930.51, "total_tokens": 59746528}
{"current_steps": 8955, "total_steps": 16324, "loss": 0.2199, "lr": 4.610910834462591e-05, "epoch": 1.0971486325074271, "percentage": 54.86, "elapsed_time": "2:49:00", "remaining_time": "2:19:04", "throughput": 5895.25, "total_tokens": 59780704}
{"current_steps": 8960, "total_steps": 16324, "loss": 0.2347, "lr": 4.605861595867726e-05, "epoch": 1.0977611711739304, "percentage": 54.89, "elapsed_time": "2:50:02", "remaining_time": "2:19:45", "throughput": 5862.82, "total_tokens": 59814448}
{"current_steps": 8965, "total_steps": 16324, "loss": 0.2494, "lr": 4.600812761693114e-05, "epoch": 1.0983737098404336, "percentage": 54.92, "elapsed_time": "2:51:02", "remaining_time": "2:20:24", "throughput": 5831.62, "total_tokens": 59847496}
{"current_steps": 8970, "total_steps": 16324, "loss": 0.2494, "lr": 4.595764337119297e-05, "epoch": 1.098986248506937, "percentage": 54.95, "elapsed_time": "2:52:02", "remaining_time": "2:21:02", "throughput": 5800.98, "total_tokens": 59880232}
{"current_steps": 8975, "total_steps": 16324, "loss": 0.2589, "lr": 4.5907163273263995e-05, "epoch": 1.0995987871734403, "percentage": 54.98, "elapsed_time": "2:53:01", "remaining_time": "2:21:40", "throughput": 5771.41, "total_tokens": 59913928}
{"current_steps": 8980, "total_steps": 16324, "loss": 0.2584, "lr": 4.585668737494116e-05, "epoch": 1.1002113258399437, "percentage": 55.01, "elapsed_time": "2:54:00", "remaining_time": "2:22:18", "throughput": 5741.99, "total_tokens": 59947160}
{"current_steps": 8985, "total_steps": 16324, "loss": 0.2586, "lr": 4.5806215728017144e-05, "epoch": 1.100823864506447, "percentage": 55.04, "elapsed_time": "2:55:00", "remaining_time": "2:22:56", "throughput": 5712.38, "total_tokens": 59980048}
{"current_steps": 8990, "total_steps": 16324, "loss": 0.2098, "lr": 4.575574838428023e-05, "epoch": 1.1014364031729502, "percentage": 55.07, "elapsed_time": "2:55:58", "remaining_time": "2:23:33", "throughput": 5683.98, "total_tokens": 60014696}
{"current_steps": 8995, "total_steps": 16324, "loss": 0.2502, "lr": 4.57052853955143e-05, "epoch": 1.1020489418394537, "percentage": 55.1, "elapsed_time": "2:56:58", "remaining_time": "2:24:11", "throughput": 5655.14, "total_tokens": 60048400}
{"current_steps": 9000, "total_steps": 16324, "loss": 0.1656, "lr": 4.5654826813498764e-05, "epoch": 1.102661480505957, "percentage": 55.13, "elapsed_time": "2:57:57", "remaining_time": "2:24:49", "throughput": 5626.86, "total_tokens": 60083320}
{"current_steps": 9000, "total_steps": 16324, "eval_loss": 0.2304692566394806, "epoch": 1.102661480505957, "percentage": 55.13, "elapsed_time": "2:58:17", "remaining_time": "2:25:05", "throughput": 5616.82, "total_tokens": 60083320}
{"current_steps": 9005, "total_steps": 16324, "loss": 0.208, "lr": 4.5604372690008506e-05, "epoch": 1.1032740191724604, "percentage": 55.16, "elapsed_time": "2:59:21", "remaining_time": "2:25:46", "throughput": 5586.51, "total_tokens": 60117224}
{"current_steps": 9010, "total_steps": 16324, "loss": 0.2704, "lr": 4.555392307681387e-05, "epoch": 1.1038865578389636, "percentage": 55.19, "elapsed_time": "3:00:20", "remaining_time": "2:26:23", "throughput": 5559.0, "total_tokens": 60151072}
{"current_steps": 9015, "total_steps": 16324, "loss": 0.2445, "lr": 4.5503478025680496e-05, "epoch": 1.1044990965054668, "percentage": 55.23, "elapsed_time": "3:01:20", "remaining_time": "2:27:01", "throughput": 5531.34, "total_tokens": 60184968}
{"current_steps": 9020, "total_steps": 16324, "loss": 0.2017, "lr": 4.5453037588369436e-05, "epoch": 1.1051116351719703, "percentage": 55.26, "elapsed_time": "3:02:19", "remaining_time": "2:27:38", "throughput": 5504.62, "total_tokens": 60219128}
{"current_steps": 9025, "total_steps": 16324, "loss": 0.2748, "lr": 4.540260181663694e-05, "epoch": 1.1057241738384735, "percentage": 55.29, "elapsed_time": "3:03:18", "remaining_time": "2:28:15", "throughput": 5478.22, "total_tokens": 60252200}
{"current_steps": 9030, "total_steps": 16324, "loss": 0.2797, "lr": 4.535217076223449e-05, "epoch": 1.1063367125049768, "percentage": 55.32, "elapsed_time": "3:04:18", "remaining_time": "2:28:52", "throughput": 5451.55, "total_tokens": 60285264}
{"current_steps": 9035, "total_steps": 16324, "loss": 0.2408, "lr": 4.530174447690875e-05, "epoch": 1.1069492511714802, "percentage": 55.35, "elapsed_time": "3:05:18", "remaining_time": "2:29:29", "throughput": 5425.26, "total_tokens": 60318840}
{"current_steps": 9040, "total_steps": 16324, "loss": 0.212, "lr": 4.5251323012401444e-05, "epoch": 1.1075617898379835, "percentage": 55.38, "elapsed_time": "3:06:17", "remaining_time": "2:30:06", "throughput": 5399.6, "total_tokens": 60352632}
{"current_steps": 9045, "total_steps": 16324, "loss": 0.2171, "lr": 4.5200906420449394e-05, "epoch": 1.108174328504487, "percentage": 55.41, "elapsed_time": "3:07:16", "remaining_time": "2:30:42", "throughput": 5373.96, "total_tokens": 60386072}
{"current_steps": 9050, "total_steps": 16324, "loss": 0.2287, "lr": 4.5150494752784395e-05, "epoch": 1.1087868671709902, "percentage": 55.44, "elapsed_time": "3:08:17", "remaining_time": "2:31:20", "throughput": 5348.06, "total_tokens": 60419128}
{"current_steps": 9050, "total_steps": 16324, "eval_loss": 0.04092838987708092, "epoch": 1.1087868671709902, "percentage": 55.44, "elapsed_time": "3:08:36", "remaining_time": "2:31:35", "throughput": 5338.88, "total_tokens": 60419128}
{"current_steps": 9055, "total_steps": 16324, "loss": 0.2298, "lr": 4.5100088061133206e-05, "epoch": 1.1093994058374934, "percentage": 55.47, "elapsed_time": "3:09:42", "remaining_time": "2:32:17", "throughput": 5310.89, "total_tokens": 60453344}
{"current_steps": 9060, "total_steps": 16324, "loss": 0.2425, "lr": 4.504968639721746e-05, "epoch": 1.1100119445039969, "percentage": 55.5, "elapsed_time": "3:10:42", "remaining_time": "2:32:54", "throughput": 5286.0, "total_tokens": 60486888}
{"current_steps": 9065, "total_steps": 16324, "loss": 0.2285, "lr": 4.4999289812753664e-05, "epoch": 1.1106244831705, "percentage": 55.53, "elapsed_time": "3:11:42", "remaining_time": "2:33:31", "throughput": 5261.39, "total_tokens": 60521120}
{"current_steps": 9070, "total_steps": 16324, "loss": 0.2351, "lr": 4.494889835945308e-05, "epoch": 1.1112370218370033, "percentage": 55.56, "elapsed_time": "3:12:42", "remaining_time": "2:34:07", "throughput": 5237.16, "total_tokens": 60554328}
{"current_steps": 9075, "total_steps": 16324, "loss": 0.2191, "lr": 4.489851208902172e-05, "epoch": 1.1118495605035068, "percentage": 55.59, "elapsed_time": "3:13:42", "remaining_time": "2:34:43", "throughput": 5213.23, "total_tokens": 60588328}
{"current_steps": 9080, "total_steps": 16324, "loss": 0.2143, "lr": 4.484813105316028e-05, "epoch": 1.11246209917001, "percentage": 55.62, "elapsed_time": "3:14:42", "remaining_time": "2:35:20", "throughput": 5189.24, "total_tokens": 60621936}
{"current_steps": 9085, "total_steps": 16324, "loss": 0.2731, "lr": 4.4797755303564076e-05, "epoch": 1.1130746378365135, "percentage": 55.65, "elapsed_time": "3:15:41", "remaining_time": "2:35:55", "throughput": 5165.93, "total_tokens": 60654744}
{"current_steps": 9090, "total_steps": 16324, "loss": 0.2238, "lr": 4.474738489192301e-05, "epoch": 1.1136871765030167, "percentage": 55.68, "elapsed_time": "3:16:40", "remaining_time": "2:36:31", "throughput": 5142.82, "total_tokens": 60688768}
{"current_steps": 9095, "total_steps": 16324, "loss": 0.2107, "lr": 4.469701986992149e-05, "epoch": 1.11429971516952, "percentage": 55.72, "elapsed_time": "3:17:40", "remaining_time": "2:37:06", "throughput": 5119.8, "total_tokens": 60722144}
{"current_steps": 9100, "total_steps": 16324, "loss": 0.2446, "lr": 4.464666028923843e-05, "epoch": 1.1149122538360234, "percentage": 55.75, "elapsed_time": "3:18:39", "remaining_time": "2:37:42", "throughput": 5097.25, "total_tokens": 60755592}
{"current_steps": 9100, "total_steps": 16324, "eval_loss": 0.0737871304154396, "epoch": 1.1149122538360234, "percentage": 55.75, "elapsed_time": "3:18:58", "remaining_time": "2:37:57", "throughput": 5089.04, "total_tokens": 60755592}
{"current_steps": 9105, "total_steps": 16324, "loss": 0.1852, "lr": 4.4596306201547114e-05, "epoch": 1.1155247925025267, "percentage": 55.78, "elapsed_time": "3:20:02", "remaining_time": "2:38:36", "throughput": 5064.69, "total_tokens": 60789632}
{"current_steps": 9110, "total_steps": 16324, "loss": 0.2752, "lr": 4.45459576585152e-05, "epoch": 1.1161373311690301, "percentage": 55.81, "elapsed_time": "3:21:01", "remaining_time": "2:39:11", "throughput": 5042.56, "total_tokens": 60822720}
{"current_steps": 9115, "total_steps": 16324, "loss": 0.2468, "lr": 4.449561471180472e-05, "epoch": 1.1167498698355334, "percentage": 55.84, "elapsed_time": "3:22:00", "remaining_time": "2:39:45", "throughput": 5021.0, "total_tokens": 60856216}
{"current_steps": 9120, "total_steps": 16324, "loss": 0.2363, "lr": 4.444527741307187e-05, "epoch": 1.1173624085020366, "percentage": 55.87, "elapsed_time": "3:22:58", "remaining_time": "2:40:20", "throughput": 4999.67, "total_tokens": 60889440}
{"current_steps": 9125, "total_steps": 16324, "loss": 0.2657, "lr": 4.439494581396709e-05, "epoch": 1.11797494716854, "percentage": 55.9, "elapsed_time": "3:23:57", "remaining_time": "2:40:54", "throughput": 4978.41, "total_tokens": 60922448}
{"current_steps": 9130, "total_steps": 16324, "loss": 0.2729, "lr": 4.434461996613498e-05, "epoch": 1.1185874858350433, "percentage": 55.93, "elapsed_time": "3:24:55", "remaining_time": "2:41:28", "throughput": 4957.58, "total_tokens": 60955232}
{"current_steps": 9135, "total_steps": 16324, "loss": 0.1933, "lr": 4.429429992121425e-05, "epoch": 1.1192000245015468, "percentage": 55.96, "elapsed_time": "3:25:54", "remaining_time": "2:42:02", "throughput": 4936.74, "total_tokens": 60989688}
{"current_steps": 9140, "total_steps": 16324, "loss": 0.2801, "lr": 4.4243985730837616e-05, "epoch": 1.11981256316805, "percentage": 55.99, "elapsed_time": "3:26:54", "remaining_time": "2:42:38", "throughput": 4915.32, "total_tokens": 61022928}
{"current_steps": 9145, "total_steps": 16324, "loss": 0.2302, "lr": 4.4193677446631834e-05, "epoch": 1.1204251018345532, "percentage": 56.02, "elapsed_time": "3:27:53", "remaining_time": "2:43:11", "throughput": 4894.85, "total_tokens": 61056264}
{"current_steps": 9150, "total_steps": 16324, "loss": 0.2335, "lr": 4.4143375120217556e-05, "epoch": 1.1210376405010567, "percentage": 56.05, "elapsed_time": "3:28:52", "remaining_time": "2:43:45", "throughput": 4874.69, "total_tokens": 61089840}
{"current_steps": 9150, "total_steps": 16324, "eval_loss": 0.16165664792060852, "epoch": 1.1210376405010567, "percentage": 56.05, "elapsed_time": "3:29:11", "remaining_time": "2:44:00", "throughput": 4867.23, "total_tokens": 61089840}
{"current_steps": 9155, "total_steps": 16324, "loss": 0.2204, "lr": 4.409307880320934e-05, "epoch": 1.12165017916756, "percentage": 56.08, "elapsed_time": "3:30:16", "remaining_time": "2:44:39", "throughput": 4844.72, "total_tokens": 61123296}
{"current_steps": 9160, "total_steps": 16324, "loss": 0.1992, "lr": 4.4042788547215605e-05, "epoch": 1.1222627178340632, "percentage": 56.11, "elapsed_time": "3:31:16", "remaining_time": "2:45:13", "throughput": 4824.63, "total_tokens": 61157128}
{"current_steps": 9165, "total_steps": 16324, "loss": 0.2861, "lr": 4.399250440383848e-05, "epoch": 1.1228752565005666, "percentage": 56.14, "elapsed_time": "3:32:15", "remaining_time": "2:45:47", "throughput": 4804.81, "total_tokens": 61189824}
{"current_steps": 9170, "total_steps": 16324, "loss": 0.2244, "lr": 4.394222642467391e-05, "epoch": 1.1234877951670699, "percentage": 56.17, "elapsed_time": "3:33:14", "remaining_time": "2:46:21", "throughput": 4785.06, "total_tokens": 61223240}
{"current_steps": 9175, "total_steps": 16324, "loss": 0.2098, "lr": 4.389195466131143e-05, "epoch": 1.1241003338335733, "percentage": 56.21, "elapsed_time": "3:34:14", "remaining_time": "2:46:56", "throughput": 4765.29, "total_tokens": 61257016}
{"current_steps": 9180, "total_steps": 16324, "loss": 0.2895, "lr": 4.384168916533428e-05, "epoch": 1.1247128725000766, "percentage": 56.24, "elapsed_time": "3:35:14", "remaining_time": "2:47:29", "throughput": 4746.04, "total_tokens": 61290784}
{"current_steps": 9185, "total_steps": 16324, "loss": 0.2716, "lr": 4.3791429988319196e-05, "epoch": 1.1253254111665798, "percentage": 56.27, "elapsed_time": "3:36:13", "remaining_time": "2:48:03", "throughput": 4726.84, "total_tokens": 61324656}
{"current_steps": 9190, "total_steps": 16324, "loss": 0.2216, "lr": 4.374117718183647e-05, "epoch": 1.1259379498330833, "percentage": 56.3, "elapsed_time": "3:37:14", "remaining_time": "2:48:38", "throughput": 4707.56, "total_tokens": 61358896}
{"current_steps": 9195, "total_steps": 16324, "loss": 0.246, "lr": 4.369093079744986e-05, "epoch": 1.1265504884995865, "percentage": 56.33, "elapsed_time": "3:38:13", "remaining_time": "2:49:11", "throughput": 4688.89, "total_tokens": 61392976}
{"current_steps": 9200, "total_steps": 16324, "loss": 0.2946, "lr": 4.3640690886716505e-05, "epoch": 1.1271630271660897, "percentage": 56.36, "elapsed_time": "3:39:13", "remaining_time": "2:49:45", "throughput": 4670.07, "total_tokens": 61425936}
{"current_steps": 9200, "total_steps": 16324, "eval_loss": 0.11595644801855087, "epoch": 1.1271630271660897, "percentage": 56.36, "elapsed_time": "3:39:32", "remaining_time": "2:50:00", "throughput": 4663.12, "total_tokens": 61425936}
{"current_steps": 9205, "total_steps": 16324, "loss": 0.2455, "lr": 4.359045750118693e-05, "epoch": 1.1277755658325932, "percentage": 56.39, "elapsed_time": "3:40:38", "remaining_time": "2:50:38", "throughput": 4642.44, "total_tokens": 61458944}
{"current_steps": 9210, "total_steps": 16324, "loss": 0.2201, "lr": 4.354023069240494e-05, "epoch": 1.1283881044990964, "percentage": 56.42, "elapsed_time": "3:41:37", "remaining_time": "2:51:11", "throughput": 4624.29, "total_tokens": 61492624}
{"current_steps": 9215, "total_steps": 16324, "loss": 0.2575, "lr": 4.349001051190762e-05, "epoch": 1.1290006431656, "percentage": 56.45, "elapsed_time": "3:42:37", "remaining_time": "2:51:45", "throughput": 4605.99, "total_tokens": 61526360}
{"current_steps": 9220, "total_steps": 16324, "loss": 0.2324, "lr": 4.343979701122523e-05, "epoch": 1.1296131818321031, "percentage": 56.48, "elapsed_time": "3:43:37", "remaining_time": "2:52:18", "throughput": 4587.87, "total_tokens": 61559968}
{"current_steps": 9225, "total_steps": 16324, "loss": 0.2494, "lr": 4.338959024188119e-05, "epoch": 1.1302257204986064, "percentage": 56.51, "elapsed_time": "3:44:36", "remaining_time": "2:52:50", "throughput": 4570.41, "total_tokens": 61593088}
{"current_steps": 9230, "total_steps": 16324, "loss": 0.202, "lr": 4.3339390255392014e-05, "epoch": 1.1308382591651098, "percentage": 56.54, "elapsed_time": "3:45:35", "remaining_time": "2:53:23", "throughput": 4552.91, "total_tokens": 61626536}
{"current_steps": 9235, "total_steps": 16324, "loss": 0.2349, "lr": 4.328919710326723e-05, "epoch": 1.131450797831613, "percentage": 56.57, "elapsed_time": "3:46:35", "remaining_time": "2:53:55", "throughput": 4535.41, "total_tokens": 61659896}
{"current_steps": 9240, "total_steps": 16324, "loss": 0.2678, "lr": 4.323901083700941e-05, "epoch": 1.1320633364981165, "percentage": 56.6, "elapsed_time": "3:47:34", "remaining_time": "2:54:28", "throughput": 4518.25, "total_tokens": 61693536}
{"current_steps": 9245, "total_steps": 16324, "loss": 0.2432, "lr": 4.3188831508113984e-05, "epoch": 1.1326758751646198, "percentage": 56.63, "elapsed_time": "3:48:33", "remaining_time": "2:55:00", "throughput": 4501.19, "total_tokens": 61726448}
{"current_steps": 9250, "total_steps": 16324, "loss": 0.2601, "lr": 4.3138659168069337e-05, "epoch": 1.133288413831123, "percentage": 56.67, "elapsed_time": "3:49:32", "remaining_time": "2:55:32", "throughput": 4484.18, "total_tokens": 61760040}
{"current_steps": 9250, "total_steps": 16324, "eval_loss": 0.19667980074882507, "epoch": 1.133288413831123, "percentage": 56.67, "elapsed_time": "3:49:51", "remaining_time": "2:55:47", "throughput": 4477.98, "total_tokens": 61760040}
{"current_steps": 9255, "total_steps": 16324, "loss": 0.2285, "lr": 4.308849386835663e-05, "epoch": 1.1339009524976265, "percentage": 56.7, "elapsed_time": "3:50:56", "remaining_time": "2:56:23", "throughput": 4459.53, "total_tokens": 61793816}
{"current_steps": 9260, "total_steps": 16324, "loss": 0.2498, "lr": 4.3038335660449824e-05, "epoch": 1.1345134911641297, "percentage": 56.73, "elapsed_time": "3:51:55", "remaining_time": "2:56:55", "throughput": 4443.02, "total_tokens": 61826856}
{"current_steps": 9265, "total_steps": 16324, "loss": 0.2659, "lr": 4.298818459581561e-05, "epoch": 1.1351260298306332, "percentage": 56.76, "elapsed_time": "3:52:55", "remaining_time": "2:57:28", "throughput": 4426.25, "total_tokens": 61859600}
{"current_steps": 9270, "total_steps": 16324, "loss": 0.2753, "lr": 4.293804072591332e-05, "epoch": 1.1357385684971364, "percentage": 56.79, "elapsed_time": "3:53:54", "remaining_time": "2:57:59", "throughput": 4409.98, "total_tokens": 61892344}
{"current_steps": 9275, "total_steps": 16324, "loss": 0.2441, "lr": 4.2887904102194936e-05, "epoch": 1.1363511071636396, "percentage": 56.82, "elapsed_time": "3:54:53", "remaining_time": "2:58:31", "throughput": 4393.95, "total_tokens": 61926304}
{"current_steps": 9280, "total_steps": 16324, "loss": 0.2048, "lr": 4.2837774776105004e-05, "epoch": 1.136963645830143, "percentage": 56.85, "elapsed_time": "3:55:52", "remaining_time": "2:59:02", "throughput": 4377.87, "total_tokens": 61959928}
{"current_steps": 9285, "total_steps": 16324, "loss": 0.2422, "lr": 4.2787652799080534e-05, "epoch": 1.1375761844966463, "percentage": 56.88, "elapsed_time": "3:56:53", "remaining_time": "2:59:34", "throughput": 4361.77, "total_tokens": 61994152}
{"current_steps": 9290, "total_steps": 16324, "loss": 0.2168, "lr": 4.273753822255104e-05, "epoch": 1.1381887231631498, "percentage": 56.91, "elapsed_time": "3:57:51", "remaining_time": "3:00:05", "throughput": 4346.27, "total_tokens": 62028112}
{"current_steps": 9295, "total_steps": 16324, "loss": 0.2453, "lr": 4.2687431097938454e-05, "epoch": 1.138801261829653, "percentage": 56.94, "elapsed_time": "3:58:51", "remaining_time": "3:00:37", "throughput": 4330.51, "total_tokens": 62061552}
{"current_steps": 9300, "total_steps": 16324, "loss": 0.244, "lr": 4.263733147665701e-05, "epoch": 1.1394138004961563, "percentage": 56.97, "elapsed_time": "3:59:51", "remaining_time": "3:01:09", "throughput": 4314.71, "total_tokens": 62094712}
{"current_steps": 9300, "total_steps": 16324, "eval_loss": 0.20997019112110138, "epoch": 1.1394138004961563, "percentage": 56.97, "elapsed_time": "4:00:10", "remaining_time": "3:01:23", "throughput": 4308.93, "total_tokens": 62094712}
{"current_steps": 9305, "total_steps": 16324, "loss": 0.2508, "lr": 4.25872394101133e-05, "epoch": 1.1400263391626597, "percentage": 57.0, "elapsed_time": "4:01:15", "remaining_time": "3:01:58", "throughput": 4292.07, "total_tokens": 62128328}
{"current_steps": 9310, "total_steps": 16324, "loss": 0.2413, "lr": 4.2537154949706125e-05, "epoch": 1.140638877829163, "percentage": 57.03, "elapsed_time": "4:02:14", "remaining_time": "3:02:30", "throughput": 4276.83, "total_tokens": 62162296}
{"current_steps": 9315, "total_steps": 16324, "loss": 0.2265, "lr": 4.248707814682649e-05, "epoch": 1.1412514164956662, "percentage": 57.06, "elapsed_time": "4:03:15", "remaining_time": "3:03:02", "throughput": 4261.36, "total_tokens": 62195464}
{"current_steps": 9320, "total_steps": 16324, "loss": 0.2592, "lr": 4.243700905285756e-05, "epoch": 1.1418639551621697, "percentage": 57.09, "elapsed_time": "4:04:14", "remaining_time": "3:03:32", "throughput": 4246.48, "total_tokens": 62229200}
{"current_steps": 9325, "total_steps": 16324, "loss": 0.2213, "lr": 4.238694771917455e-05, "epoch": 1.142476493828673, "percentage": 57.12, "elapsed_time": "4:05:13", "remaining_time": "3:04:03", "throughput": 4231.59, "total_tokens": 62262880}
{"current_steps": 9330, "total_steps": 16324, "loss": 0.2364, "lr": 4.233689419714477e-05, "epoch": 1.1430890324951763, "percentage": 57.16, "elapsed_time": "4:06:14", "remaining_time": "3:04:35", "throughput": 4216.57, "total_tokens": 62296248}
{"current_steps": 9335, "total_steps": 16324, "loss": 0.246, "lr": 4.2286848538127466e-05, "epoch": 1.1437015711616796, "percentage": 57.19, "elapsed_time": "4:07:13", "remaining_time": "3:05:05", "throughput": 4202.02, "total_tokens": 62329936}
{"current_steps": 9340, "total_steps": 16324, "loss": 0.2451, "lr": 4.2236810793473844e-05, "epoch": 1.1443141098281828, "percentage": 57.22, "elapsed_time": "4:08:13", "remaining_time": "3:05:36", "throughput": 4187.45, "total_tokens": 62363832}
{"current_steps": 9345, "total_steps": 16324, "loss": 0.2524, "lr": 4.218678101452696e-05, "epoch": 1.1449266484946863, "percentage": 57.25, "elapsed_time": "4:09:12", "remaining_time": "3:06:07", "throughput": 4172.91, "total_tokens": 62396696}
{"current_steps": 9350, "total_steps": 16324, "loss": 0.2379, "lr": 4.213675925262173e-05, "epoch": 1.1455391871611895, "percentage": 57.28, "elapsed_time": "4:10:11", "remaining_time": "3:06:36", "throughput": 4158.83, "total_tokens": 62430368}
{"current_steps": 9350, "total_steps": 16324, "eval_loss": 0.18152250349521637, "epoch": 1.1455391871611895, "percentage": 57.28, "elapsed_time": "4:10:30", "remaining_time": "3:06:51", "throughput": 4153.56, "total_tokens": 62430368}
{"current_steps": 9355, "total_steps": 16324, "loss": 0.2233, "lr": 4.208674555908484e-05, "epoch": 1.1461517258276928, "percentage": 57.31, "elapsed_time": "4:11:34", "remaining_time": "3:07:24", "throughput": 4138.2, "total_tokens": 62463696}
{"current_steps": 9360, "total_steps": 16324, "loss": 0.231, "lr": 4.203673998523467e-05, "epoch": 1.1467642644941962, "percentage": 57.34, "elapsed_time": "4:12:34", "remaining_time": "3:07:55", "throughput": 4124.05, "total_tokens": 62497344}
{"current_steps": 9365, "total_steps": 16324, "loss": 0.2353, "lr": 4.1986742582381274e-05, "epoch": 1.1473768031606995, "percentage": 57.37, "elapsed_time": "4:13:33", "remaining_time": "3:08:24", "throughput": 4110.3, "total_tokens": 62530120}
{"current_steps": 9370, "total_steps": 16324, "loss": 0.2508, "lr": 4.193675340182633e-05, "epoch": 1.147989341827203, "percentage": 57.4, "elapsed_time": "4:14:31", "remaining_time": "3:08:54", "throughput": 4096.7, "total_tokens": 62563856}
{"current_steps": 9375, "total_steps": 16324, "loss": 0.2152, "lr": 4.188677249486311e-05, "epoch": 1.1486018804937062, "percentage": 57.43, "elapsed_time": "4:15:31", "remaining_time": "3:09:24", "throughput": 4082.91, "total_tokens": 62597400}
{"current_steps": 9380, "total_steps": 16324, "loss": 0.2277, "lr": 4.183679991277635e-05, "epoch": 1.1492144191602094, "percentage": 57.46, "elapsed_time": "4:16:30", "remaining_time": "3:09:53", "throughput": 4069.46, "total_tokens": 62630920}
{"current_steps": 9385, "total_steps": 16324, "loss": 0.2302, "lr": 4.178683570684226e-05, "epoch": 1.1498269578267128, "percentage": 57.49, "elapsed_time": "4:17:29", "remaining_time": "3:10:22", "throughput": 4056.11, "total_tokens": 62664904}
{"current_steps": 9390, "total_steps": 16324, "loss": 0.2507, "lr": 4.1736879928328465e-05, "epoch": 1.150439496493216, "percentage": 57.52, "elapsed_time": "4:18:29", "remaining_time": "3:10:52", "throughput": 4042.61, "total_tokens": 62697600}
{"current_steps": 9395, "total_steps": 16324, "loss": 0.2246, "lr": 4.1686932628493924e-05, "epoch": 1.1510520351597195, "percentage": 57.55, "elapsed_time": "4:19:28", "remaining_time": "3:11:21", "throughput": 4029.47, "total_tokens": 62730904}
{"current_steps": 9400, "total_steps": 16324, "loss": 0.2501, "lr": 4.1636993858588925e-05, "epoch": 1.1516645738262228, "percentage": 57.58, "elapsed_time": "4:20:27", "remaining_time": "3:11:50", "throughput": 4016.33, "total_tokens": 62764152}
{"current_steps": 9400, "total_steps": 16324, "eval_loss": 0.16211895644664764, "epoch": 1.1516645738262228, "percentage": 57.58, "elapsed_time": "4:20:46", "remaining_time": "3:12:05", "throughput": 4011.43, "total_tokens": 62764152}
{"current_steps": 9405, "total_steps": 16324, "loss": 0.2507, "lr": 4.158706366985496e-05, "epoch": 1.152277112492726, "percentage": 57.61, "elapsed_time": "4:21:51", "remaining_time": "3:12:38", "throughput": 3996.91, "total_tokens": 62797240}
{"current_steps": 9410, "total_steps": 16324, "loss": 0.253, "lr": 4.153714211352477e-05, "epoch": 1.1528896511592295, "percentage": 57.65, "elapsed_time": "4:22:50", "remaining_time": "3:13:07", "throughput": 3984.04, "total_tokens": 62830784}
{"current_steps": 9415, "total_steps": 16324, "loss": 0.2106, "lr": 4.1487229240822184e-05, "epoch": 1.1535021898257327, "percentage": 57.68, "elapsed_time": "4:23:50", "remaining_time": "3:13:36", "throughput": 3971.19, "total_tokens": 62864352}
{"current_steps": 9420, "total_steps": 16324, "loss": 0.2186, "lr": 4.1437325102962166e-05, "epoch": 1.1541147284922362, "percentage": 57.71, "elapsed_time": "4:24:48", "remaining_time": "3:14:05", "throughput": 3958.6, "total_tokens": 62898080}
{"current_steps": 9425, "total_steps": 16324, "loss": 0.2311, "lr": 4.13874297511507e-05, "epoch": 1.1547272671587394, "percentage": 57.74, "elapsed_time": "4:25:49", "remaining_time": "3:14:34", "throughput": 3945.74, "total_tokens": 62931672}
{"current_steps": 9430, "total_steps": 16324, "loss": 0.2374, "lr": 4.1337543236584725e-05, "epoch": 1.1553398058252426, "percentage": 57.77, "elapsed_time": "4:26:48", "remaining_time": "3:15:03", "throughput": 3933.26, "total_tokens": 62965192}
{"current_steps": 9435, "total_steps": 16324, "loss": 0.2119, "lr": 4.128766561045218e-05, "epoch": 1.155952344491746, "percentage": 57.8, "elapsed_time": "4:27:48", "remaining_time": "3:15:32", "throughput": 3920.75, "total_tokens": 62999184}
{"current_steps": 9440, "total_steps": 16324, "loss": 0.212, "lr": 4.123779692393182e-05, "epoch": 1.1565648831582493, "percentage": 57.83, "elapsed_time": "4:28:48", "remaining_time": "3:16:01", "throughput": 3908.07, "total_tokens": 63033216}
{"current_steps": 9445, "total_steps": 16324, "loss": 0.2037, "lr": 4.118793722819325e-05, "epoch": 1.1571774218247528, "percentage": 57.86, "elapsed_time": "4:29:48", "remaining_time": "3:16:30", "throughput": 3895.78, "total_tokens": 63067312}
{"current_steps": 9450, "total_steps": 16324, "loss": 0.2499, "lr": 4.1138086574396853e-05, "epoch": 1.157789960491256, "percentage": 57.89, "elapsed_time": "4:30:48", "remaining_time": "3:16:59", "throughput": 3883.42, "total_tokens": 63100624}
{"current_steps": 9450, "total_steps": 16324, "eval_loss": 0.26281774044036865, "epoch": 1.157789960491256, "percentage": 57.89, "elapsed_time": "4:31:08", "remaining_time": "3:17:13", "throughput": 3878.76, "total_tokens": 63100624}
{"current_steps": 9455, "total_steps": 16324, "loss": 0.3113, "lr": 4.108824501369373e-05, "epoch": 1.1584024991577593, "percentage": 57.92, "elapsed_time": "4:32:13", "remaining_time": "3:17:46", "throughput": 3865.27, "total_tokens": 63133768}
{"current_steps": 9460, "total_steps": 16324, "loss": 0.237, "lr": 4.103841259722565e-05, "epoch": 1.1590150378242627, "percentage": 57.95, "elapsed_time": "4:33:13", "remaining_time": "3:18:14", "throughput": 3853.23, "total_tokens": 63167024}
{"current_steps": 9465, "total_steps": 16324, "loss": 0.2199, "lr": 4.098858937612503e-05, "epoch": 1.159627576490766, "percentage": 57.98, "elapsed_time": "4:34:13", "remaining_time": "3:18:43", "throughput": 3841.2, "total_tokens": 63200352}
{"current_steps": 9470, "total_steps": 16324, "loss": 0.2444, "lr": 4.0938775401514786e-05, "epoch": 1.1602401151572692, "percentage": 58.01, "elapsed_time": "4:35:13", "remaining_time": "3:19:11", "throughput": 3829.28, "total_tokens": 63234024}
{"current_steps": 9475, "total_steps": 16324, "loss": 0.2142, "lr": 4.088897072450841e-05, "epoch": 1.1608526538237727, "percentage": 58.04, "elapsed_time": "4:36:12", "remaining_time": "3:19:39", "throughput": 3817.52, "total_tokens": 63267560}
{"current_steps": 9480, "total_steps": 16324, "loss": 0.2666, "lr": 4.083917539620983e-05, "epoch": 1.161465192490276, "percentage": 58.07, "elapsed_time": "4:37:12", "remaining_time": "3:20:07", "throughput": 3805.8, "total_tokens": 63300920}
{"current_steps": 9485, "total_steps": 16324, "loss": 0.2489, "lr": 4.0789389467713355e-05, "epoch": 1.1620777311567791, "percentage": 58.1, "elapsed_time": "4:38:12", "remaining_time": "3:20:35", "throughput": 3794.19, "total_tokens": 63334752}
{"current_steps": 9490, "total_steps": 16324, "loss": 0.3074, "lr": 4.0739612990103695e-05, "epoch": 1.1626902698232826, "percentage": 58.14, "elapsed_time": "4:39:11", "remaining_time": "3:21:03", "throughput": 3782.76, "total_tokens": 63368400}
{"current_steps": 9495, "total_steps": 16324, "loss": 0.2356, "lr": 4.068984601445584e-05, "epoch": 1.1633028084897858, "percentage": 58.17, "elapsed_time": "4:40:11", "remaining_time": "3:21:31", "throughput": 3771.31, "total_tokens": 63401416}
{"current_steps": 9500, "total_steps": 16324, "loss": 0.2128, "lr": 4.064008859183504e-05, "epoch": 1.1639153471562893, "percentage": 58.2, "elapsed_time": "4:41:11", "remaining_time": "3:21:58", "throughput": 3760.0, "total_tokens": 63435392}
{"current_steps": 9500, "total_steps": 16324, "eval_loss": 0.18865041434764862, "epoch": 1.1639153471562893, "percentage": 58.2, "elapsed_time": "4:41:30", "remaining_time": "3:22:12", "throughput": 3755.7, "total_tokens": 63435392}
{"current_steps": 9505, "total_steps": 16324, "loss": 0.251, "lr": 4.059034077329671e-05, "epoch": 1.1645278858227925, "percentage": 58.23, "elapsed_time": "4:42:34", "remaining_time": "3:22:43", "throughput": 3743.44, "total_tokens": 63468256}
{"current_steps": 9510, "total_steps": 16324, "loss": 0.2268, "lr": 4.0540602609886444e-05, "epoch": 1.1651404244892958, "percentage": 58.26, "elapsed_time": "4:43:34", "remaining_time": "3:23:10", "throughput": 3732.38, "total_tokens": 63502664}
{"current_steps": 9515, "total_steps": 16324, "loss": 0.2266, "lr": 4.049087415263994e-05, "epoch": 1.1657529631557992, "percentage": 58.29, "elapsed_time": "4:44:34", "remaining_time": "3:23:38", "throughput": 3721.09, "total_tokens": 63536472}
{"current_steps": 9520, "total_steps": 16324, "loss": 0.2444, "lr": 4.0441155452582916e-05, "epoch": 1.1663655018223025, "percentage": 58.32, "elapsed_time": "4:45:33", "remaining_time": "3:24:05", "throughput": 3710.22, "total_tokens": 63569768}
{"current_steps": 9525, "total_steps": 16324, "loss": 0.2204, "lr": 4.039144656073105e-05, "epoch": 1.166978040488806, "percentage": 58.35, "elapsed_time": "4:46:33", "remaining_time": "3:24:32", "throughput": 3699.27, "total_tokens": 63603296}
{"current_steps": 9530, "total_steps": 16324, "loss": 0.1856, "lr": 4.034174752808998e-05, "epoch": 1.1675905791553092, "percentage": 58.38, "elapsed_time": "4:47:33", "remaining_time": "3:25:00", "throughput": 3688.32, "total_tokens": 63637232}
{"current_steps": 9535, "total_steps": 16324, "loss": 0.2501, "lr": 4.0292058405655264e-05, "epoch": 1.1682031178218124, "percentage": 58.41, "elapsed_time": "4:48:33", "remaining_time": "3:25:27", "throughput": 3677.44, "total_tokens": 63670416}
{"current_steps": 9540, "total_steps": 16324, "loss": 0.2348, "lr": 4.024237924441224e-05, "epoch": 1.1688156564883159, "percentage": 58.44, "elapsed_time": "4:49:33", "remaining_time": "3:25:54", "throughput": 3666.73, "total_tokens": 63704184}
{"current_steps": 9545, "total_steps": 16324, "loss": 0.2098, "lr": 4.019271009533606e-05, "epoch": 1.169428195154819, "percentage": 58.47, "elapsed_time": "4:50:33", "remaining_time": "3:26:21", "throughput": 3656.0, "total_tokens": 63737640}
{"current_steps": 9550, "total_steps": 16324, "loss": 0.2689, "lr": 4.014305100939157e-05, "epoch": 1.1700407338213226, "percentage": 58.5, "elapsed_time": "4:51:33", "remaining_time": "3:26:48", "throughput": 3645.39, "total_tokens": 63771112}
{"current_steps": 9550, "total_steps": 16324, "eval_loss": 0.09276539832353592, "epoch": 1.1700407338213226, "percentage": 58.5, "elapsed_time": "4:51:53", "remaining_time": "3:27:02", "throughput": 3641.32, "total_tokens": 63771112}
{"current_steps": 9555, "total_steps": 16324, "loss": 0.2242, "lr": 4.009340203753332e-05, "epoch": 1.1706532724878258, "percentage": 58.53, "elapsed_time": "4:52:57", "remaining_time": "3:27:32", "throughput": 3629.85, "total_tokens": 63805152}
{"current_steps": 9560, "total_steps": 16324, "loss": 0.2551, "lr": 4.004376323070547e-05, "epoch": 1.171265811154329, "percentage": 58.56, "elapsed_time": "4:53:57", "remaining_time": "3:27:59", "throughput": 3619.44, "total_tokens": 63838496}
{"current_steps": 9565, "total_steps": 16324, "loss": 0.2105, "lr": 3.999413463984174e-05, "epoch": 1.1718783498208325, "percentage": 58.59, "elapsed_time": "4:54:58", "remaining_time": "3:28:26", "throughput": 3608.86, "total_tokens": 63872464}
{"current_steps": 9570, "total_steps": 16324, "loss": 0.2316, "lr": 3.994451631586539e-05, "epoch": 1.1724908884873357, "percentage": 58.63, "elapsed_time": "4:55:58", "remaining_time": "3:28:52", "throughput": 3598.68, "total_tokens": 63906352}
{"current_steps": 9575, "total_steps": 16324, "loss": 0.1975, "lr": 3.989490830968911e-05, "epoch": 1.1731034271538392, "percentage": 58.66, "elapsed_time": "4:56:58", "remaining_time": "3:29:19", "throughput": 3588.48, "total_tokens": 63940936}
{"current_steps": 9580, "total_steps": 16324, "loss": 0.2207, "lr": 3.9845310672215054e-05, "epoch": 1.1737159658203424, "percentage": 58.69, "elapsed_time": "4:57:59", "remaining_time": "3:29:46", "throughput": 3578.14, "total_tokens": 63974536}
{"current_steps": 9585, "total_steps": 16324, "loss": 0.2777, "lr": 3.979572345433469e-05, "epoch": 1.1743285044868457, "percentage": 58.72, "elapsed_time": "4:58:58", "remaining_time": "3:30:12", "throughput": 3568.13, "total_tokens": 64007880}
{"current_steps": 9590, "total_steps": 16324, "loss": 0.2191, "lr": 3.974614670692881e-05, "epoch": 1.1749410431533491, "percentage": 58.75, "elapsed_time": "4:59:58", "remaining_time": "3:30:38", "throughput": 3558.11, "total_tokens": 64041728}
{"current_steps": 9595, "total_steps": 16324, "loss": 0.2531, "lr": 3.9696580480867476e-05, "epoch": 1.1755535818198524, "percentage": 58.78, "elapsed_time": "5:00:58", "remaining_time": "3:31:04", "throughput": 3548.08, "total_tokens": 64074776}
{"current_steps": 9600, "total_steps": 16324, "loss": 0.2556, "lr": 3.964702482700993e-05, "epoch": 1.1761661204863558, "percentage": 58.81, "elapsed_time": "5:01:58", "remaining_time": "3:31:30", "throughput": 3538.29, "total_tokens": 64107872}
{"current_steps": 9600, "total_steps": 16324, "eval_loss": 0.11478094756603241, "epoch": 1.1761661204863558, "percentage": 58.81, "elapsed_time": "5:02:17", "remaining_time": "3:31:43", "throughput": 3534.54, "total_tokens": 64107872}
{"current_steps": 9605, "total_steps": 16324, "loss": 0.2505, "lr": 3.9597479796204625e-05, "epoch": 1.176778659152859, "percentage": 58.84, "elapsed_time": "5:03:22", "remaining_time": "3:32:13", "throughput": 3523.78, "total_tokens": 64141168}
{"current_steps": 9610, "total_steps": 16324, "loss": 0.2437, "lr": 3.954794543928899e-05, "epoch": 1.1773911978193623, "percentage": 58.87, "elapsed_time": "5:04:22", "remaining_time": "3:32:39", "throughput": 3514.0, "total_tokens": 64174704}
{"current_steps": 9615, "total_steps": 16324, "loss": 0.2566, "lr": 3.9498421807089644e-05, "epoch": 1.1780037364858658, "percentage": 58.9, "elapsed_time": "5:05:21", "remaining_time": "3:33:04", "throughput": 3504.45, "total_tokens": 64208160}
{"current_steps": 9620, "total_steps": 16324, "loss": 0.1975, "lr": 3.944890895042209e-05, "epoch": 1.178616275152369, "percentage": 58.93, "elapsed_time": "5:06:21", "remaining_time": "3:33:29", "throughput": 3495.0, "total_tokens": 64242328}
{"current_steps": 9625, "total_steps": 16324, "loss": 0.1948, "lr": 3.9399406920090864e-05, "epoch": 1.1792288138188722, "percentage": 58.96, "elapsed_time": "5:07:21", "remaining_time": "3:33:55", "throughput": 3485.44, "total_tokens": 64276064}
{"current_steps": 9630, "total_steps": 16324, "loss": 0.2698, "lr": 3.934991576688932e-05, "epoch": 1.1798413524853757, "percentage": 58.99, "elapsed_time": "5:08:20", "remaining_time": "3:34:19", "throughput": 3476.17, "total_tokens": 64309672}
{"current_steps": 9635, "total_steps": 16324, "loss": 0.2303, "lr": 3.930043554159971e-05, "epoch": 1.180453891151879, "percentage": 59.02, "elapsed_time": "5:09:19", "remaining_time": "3:34:45", "throughput": 3466.77, "total_tokens": 64342944}
{"current_steps": 9640, "total_steps": 16324, "loss": 0.228, "lr": 3.925096629499302e-05, "epoch": 1.1810664298183822, "percentage": 59.05, "elapsed_time": "5:10:20", "remaining_time": "3:35:10", "throughput": 3457.28, "total_tokens": 64377048}
{"current_steps": 9645, "total_steps": 16324, "loss": 0.2396, "lr": 3.9201508077829e-05, "epoch": 1.1816789684848856, "percentage": 59.08, "elapsed_time": "5:11:19", "remaining_time": "3:35:35", "throughput": 3448.2, "total_tokens": 64410448}
{"current_steps": 9650, "total_steps": 16324, "loss": 0.2527, "lr": 3.91520609408561e-05, "epoch": 1.1822915071513889, "percentage": 59.12, "elapsed_time": "5:12:19", "remaining_time": "3:36:00", "throughput": 3438.98, "total_tokens": 64443680}
{"current_steps": 9650, "total_steps": 16324, "eval_loss": 0.12790514528751373, "epoch": 1.1822915071513889, "percentage": 59.12, "elapsed_time": "5:12:38", "remaining_time": "3:36:13", "throughput": 3435.42, "total_tokens": 64443680}
{"current_steps": 9655, "total_steps": 16324, "loss": 0.2157, "lr": 3.910262493481135e-05, "epoch": 1.1829040458178923, "percentage": 59.15, "elapsed_time": "5:13:44", "remaining_time": "3:36:42", "throughput": 3425.19, "total_tokens": 64476992}
{"current_steps": 9660, "total_steps": 16324, "loss": 0.2082, "lr": 3.905320011042041e-05, "epoch": 1.1835165844843956, "percentage": 59.18, "elapsed_time": "5:14:43", "remaining_time": "3:37:07", "throughput": 3416.18, "total_tokens": 64510904}
{"current_steps": 9665, "total_steps": 16324, "loss": 0.2468, "lr": 3.9003786518397434e-05, "epoch": 1.1841291231508988, "percentage": 59.21, "elapsed_time": "5:15:43", "remaining_time": "3:37:31", "throughput": 3407.14, "total_tokens": 64544304}
{"current_steps": 9670, "total_steps": 16324, "loss": 0.1972, "lr": 3.895438420944506e-05, "epoch": 1.1847416618174023, "percentage": 59.24, "elapsed_time": "5:16:43", "remaining_time": "3:37:56", "throughput": 3398.18, "total_tokens": 64578968}
{"current_steps": 9675, "total_steps": 16324, "loss": 0.1995, "lr": 3.890499323425436e-05, "epoch": 1.1853542004839055, "percentage": 59.27, "elapsed_time": "5:17:44", "remaining_time": "3:38:21", "throughput": 3389.21, "total_tokens": 64612840}
{"current_steps": 9680, "total_steps": 16324, "loss": 0.2309, "lr": 3.885561364350475e-05, "epoch": 1.185966739150409, "percentage": 59.3, "elapsed_time": "5:18:44", "remaining_time": "3:38:46", "throughput": 3380.35, "total_tokens": 64646224}
{"current_steps": 9685, "total_steps": 16324, "loss": 0.2277, "lr": 3.8806245487864026e-05, "epoch": 1.1865792778169122, "percentage": 59.33, "elapsed_time": "5:19:44", "remaining_time": "3:39:10", "throughput": 3371.55, "total_tokens": 64680384}
{"current_steps": 9690, "total_steps": 16324, "loss": 0.2545, "lr": 3.875688881798814e-05, "epoch": 1.1871918164834154, "percentage": 59.36, "elapsed_time": "5:20:44", "remaining_time": "3:39:35", "throughput": 3362.68, "total_tokens": 64714000}
{"current_steps": 9695, "total_steps": 16324, "loss": 0.2513, "lr": 3.870754368452137e-05, "epoch": 1.187804355149919, "percentage": 59.39, "elapsed_time": "5:21:44", "remaining_time": "3:39:59", "throughput": 3353.96, "total_tokens": 64747416}
{"current_steps": 9700, "total_steps": 16324, "loss": 0.2621, "lr": 3.865821013809609e-05, "epoch": 1.1884168938164221, "percentage": 59.42, "elapsed_time": "5:22:44", "remaining_time": "3:40:23", "throughput": 3345.35, "total_tokens": 64781416}
{"current_steps": 9700, "total_steps": 16324, "eval_loss": 0.13076919317245483, "epoch": 1.1884168938164221, "percentage": 59.42, "elapsed_time": "5:23:04", "remaining_time": "3:40:37", "throughput": 3341.93, "total_tokens": 64781416}
{"current_steps": 9705, "total_steps": 16324, "loss": 0.1975, "lr": 3.8608888229332804e-05, "epoch": 1.1890294324829256, "percentage": 59.45, "elapsed_time": "5:24:10", "remaining_time": "3:41:05", "throughput": 3332.33, "total_tokens": 64815384}
{"current_steps": 9710, "total_steps": 16324, "loss": 0.2391, "lr": 3.855957800884009e-05, "epoch": 1.1896419711494288, "percentage": 59.48, "elapsed_time": "5:25:10", "remaining_time": "3:41:29", "throughput": 3323.84, "total_tokens": 64849224}
{"current_steps": 9715, "total_steps": 16324, "loss": 0.2046, "lr": 3.851027952721453e-05, "epoch": 1.190254509815932, "percentage": 59.51, "elapsed_time": "5:26:09", "remaining_time": "3:41:53", "throughput": 3315.47, "total_tokens": 64883384}
{"current_steps": 9720, "total_steps": 16324, "loss": 0.1976, "lr": 3.846099283504063e-05, "epoch": 1.1908670484824355, "percentage": 59.54, "elapsed_time": "5:27:10", "remaining_time": "3:42:17", "throughput": 3306.95, "total_tokens": 64916936}
{"current_steps": 9725, "total_steps": 16324, "loss": 0.208, "lr": 3.841171798289081e-05, "epoch": 1.1914795871489388, "percentage": 59.57, "elapsed_time": "5:28:09", "remaining_time": "3:42:40", "throughput": 3298.76, "total_tokens": 64950880}
{"current_steps": 9730, "total_steps": 16324, "loss": 0.2598, "lr": 3.836245502132538e-05, "epoch": 1.1920921258154422, "percentage": 59.61, "elapsed_time": "5:29:08", "remaining_time": "3:43:03", "throughput": 3290.6, "total_tokens": 64984160}
{"current_steps": 9735, "total_steps": 16324, "loss": 0.217, "lr": 3.831320400089239e-05, "epoch": 1.1927046644819455, "percentage": 59.64, "elapsed_time": "5:30:07", "remaining_time": "3:43:26", "throughput": 3282.41, "total_tokens": 65017744}
{"current_steps": 9740, "total_steps": 16324, "loss": 0.2476, "lr": 3.826396497212769e-05, "epoch": 1.1933172031484487, "percentage": 59.67, "elapsed_time": "5:31:07", "remaining_time": "3:43:49", "throughput": 3274.29, "total_tokens": 65050920}
{"current_steps": 9745, "total_steps": 16324, "loss": 0.2501, "lr": 3.82147379855548e-05, "epoch": 1.1939297418149522, "percentage": 59.7, "elapsed_time": "5:32:06", "remaining_time": "3:44:12", "throughput": 3266.3, "total_tokens": 65084512}
{"current_steps": 9750, "total_steps": 16324, "loss": 0.2257, "lr": 3.816552309168486e-05, "epoch": 1.1945422804814554, "percentage": 59.73, "elapsed_time": "5:33:05", "remaining_time": "3:44:35", "throughput": 3258.31, "total_tokens": 65118432}
{"current_steps": 9750, "total_steps": 16324, "eval_loss": 0.12043237686157227, "epoch": 1.1945422804814554, "percentage": 59.73, "elapsed_time": "5:33:24", "remaining_time": "3:44:48", "throughput": 3255.16, "total_tokens": 65118432}
{"current_steps": 9755, "total_steps": 16324, "loss": 0.2107, "lr": 3.8116320341016665e-05, "epoch": 1.1951548191479586, "percentage": 59.76, "elapsed_time": "5:34:29", "remaining_time": "3:45:14", "throughput": 3246.4, "total_tokens": 65152912}
{"current_steps": 9760, "total_steps": 16324, "loss": 0.2184, "lr": 3.8067129784036484e-05, "epoch": 1.195767357814462, "percentage": 59.79, "elapsed_time": "5:35:27", "remaining_time": "3:45:36", "throughput": 3238.68, "total_tokens": 65186848}
{"current_steps": 9765, "total_steps": 16324, "loss": 0.2131, "lr": 3.801795147121815e-05, "epoch": 1.1963798964809653, "percentage": 59.82, "elapsed_time": "5:36:26", "remaining_time": "3:45:59", "throughput": 3230.83, "total_tokens": 65220376}
{"current_steps": 9770, "total_steps": 16324, "loss": 0.2544, "lr": 3.796878545302281e-05, "epoch": 1.1969924351474688, "percentage": 59.85, "elapsed_time": "5:37:24", "remaining_time": "3:46:20", "throughput": 3223.23, "total_tokens": 65253824}
{"current_steps": 9775, "total_steps": 16324, "loss": 0.2142, "lr": 3.791963177989912e-05, "epoch": 1.197604973813972, "percentage": 59.88, "elapsed_time": "5:38:23", "remaining_time": "3:46:43", "throughput": 3215.56, "total_tokens": 65288104}
{"current_steps": 9780, "total_steps": 16324, "loss": 0.1976, "lr": 3.7870490502283005e-05, "epoch": 1.1982175124804753, "percentage": 59.91, "elapsed_time": "5:39:23", "remaining_time": "3:47:05", "throughput": 3207.79, "total_tokens": 65321912}
{"current_steps": 9785, "total_steps": 16324, "loss": 0.2692, "lr": 3.782136167059771e-05, "epoch": 1.1988300511469787, "percentage": 59.94, "elapsed_time": "5:40:22", "remaining_time": "3:47:27", "throughput": 3200.2, "total_tokens": 65355040}
{"current_steps": 9790, "total_steps": 16324, "loss": 0.2428, "lr": 3.777224533525365e-05, "epoch": 1.199442589813482, "percentage": 59.97, "elapsed_time": "5:41:21", "remaining_time": "3:47:49", "throughput": 3192.59, "total_tokens": 65388392}
{"current_steps": 9795, "total_steps": 16324, "loss": 0.1856, "lr": 3.7723141546648484e-05, "epoch": 1.2000551284799852, "percentage": 60.0, "elapsed_time": "5:42:20", "remaining_time": "3:48:11", "throughput": 3185.05, "total_tokens": 65422864}
{"current_steps": 9800, "total_steps": 16324, "loss": 0.1952, "lr": 3.767405035516695e-05, "epoch": 1.2006676671464886, "percentage": 60.03, "elapsed_time": "5:43:19", "remaining_time": "3:48:33", "throughput": 3177.57, "total_tokens": 65456984}
{"current_steps": 9800, "total_steps": 16324, "eval_loss": 0.15540121495723724, "epoch": 1.2006676671464886, "percentage": 60.03, "elapsed_time": "5:43:39", "remaining_time": "3:48:46", "throughput": 3174.58, "total_tokens": 65456984}
{"current_steps": 9805, "total_steps": 16324, "loss": 0.1921, "lr": 3.7624971811180876e-05, "epoch": 1.2012802058129919, "percentage": 60.06, "elapsed_time": "5:44:43", "remaining_time": "3:49:11", "throughput": 3166.34, "total_tokens": 65491160}
{"current_steps": 9810, "total_steps": 16324, "loss": 0.2193, "lr": 3.757590596504913e-05, "epoch": 1.2018927444794953, "percentage": 60.1, "elapsed_time": "5:45:42", "remaining_time": "3:49:33", "throughput": 3158.92, "total_tokens": 65525176}
{"current_steps": 9815, "total_steps": 16324, "loss": 0.2093, "lr": 3.75268528671175e-05, "epoch": 1.2025052831459986, "percentage": 60.13, "elapsed_time": "5:46:42", "remaining_time": "3:49:55", "throughput": 3151.42, "total_tokens": 65558616}
{"current_steps": 9820, "total_steps": 16324, "loss": 0.25, "lr": 3.747781256771876e-05, "epoch": 1.2031178218125018, "percentage": 60.16, "elapsed_time": "5:47:42", "remaining_time": "3:50:17", "throughput": 3143.96, "total_tokens": 65592224}
{"current_steps": 9825, "total_steps": 16324, "loss": 0.2461, "lr": 3.7428785117172516e-05, "epoch": 1.2037303604790053, "percentage": 60.19, "elapsed_time": "5:48:42", "remaining_time": "3:50:39", "throughput": 3136.57, "total_tokens": 65625296}
{"current_steps": 9830, "total_steps": 16324, "loss": 0.2281, "lr": 3.7379770565785166e-05, "epoch": 1.2043428991455085, "percentage": 60.22, "elapsed_time": "5:49:43", "remaining_time": "3:51:02", "throughput": 3129.03, "total_tokens": 65658896}
{"current_steps": 9835, "total_steps": 16324, "loss": 0.2641, "lr": 3.733076896384993e-05, "epoch": 1.204955437812012, "percentage": 60.25, "elapsed_time": "5:50:43", "remaining_time": "3:51:24", "throughput": 3121.78, "total_tokens": 65692240}
{"current_steps": 9840, "total_steps": 16324, "loss": 0.2389, "lr": 3.7281780361646685e-05, "epoch": 1.2055679764785152, "percentage": 60.28, "elapsed_time": "5:51:41", "remaining_time": "3:51:45", "throughput": 3114.66, "total_tokens": 65725568}
{"current_steps": 9845, "total_steps": 16324, "loss": 0.1951, "lr": 3.723280480944202e-05, "epoch": 1.2061805151450185, "percentage": 60.31, "elapsed_time": "5:52:41", "remaining_time": "3:52:06", "throughput": 3107.5, "total_tokens": 65759352}
{"current_steps": 9850, "total_steps": 16324, "loss": 0.2196, "lr": 3.718384235748909e-05, "epoch": 1.206793053811522, "percentage": 60.34, "elapsed_time": "5:53:40", "remaining_time": "3:52:27", "throughput": 3100.52, "total_tokens": 65793256}
{"current_steps": 9850, "total_steps": 16324, "eval_loss": 0.08569300174713135, "epoch": 1.206793053811522, "percentage": 60.34, "elapsed_time": "5:53:59", "remaining_time": "3:52:39", "throughput": 3097.71, "total_tokens": 65793256}
{"current_steps": 9855, "total_steps": 16324, "loss": 0.241, "lr": 3.713489305602764e-05, "epoch": 1.2074055924780251, "percentage": 60.37, "elapsed_time": "5:55:03", "remaining_time": "3:53:04", "throughput": 3089.94, "total_tokens": 65826720}
{"current_steps": 9860, "total_steps": 16324, "loss": 0.2296, "lr": 3.7085956955283876e-05, "epoch": 1.2080181311445286, "percentage": 60.4, "elapsed_time": "5:56:03", "remaining_time": "3:53:25", "throughput": 3082.87, "total_tokens": 65860616}
{"current_steps": 9865, "total_steps": 16324, "loss": 0.2382, "lr": 3.703703410547052e-05, "epoch": 1.2086306698110318, "percentage": 60.43, "elapsed_time": "5:57:01", "remaining_time": "3:53:45", "throughput": 3076.02, "total_tokens": 65894088}
{"current_steps": 9870, "total_steps": 16324, "loss": 0.2488, "lr": 3.6988124556786643e-05, "epoch": 1.209243208477535, "percentage": 60.46, "elapsed_time": "5:58:00", "remaining_time": "3:54:06", "throughput": 3069.17, "total_tokens": 65927088}
{"current_steps": 9875, "total_steps": 16324, "loss": 0.2449, "lr": 3.693922835941772e-05, "epoch": 1.2098557471440385, "percentage": 60.49, "elapsed_time": "5:59:01", "remaining_time": "3:54:27", "throughput": 3062.04, "total_tokens": 65960688}
{"current_steps": 9880, "total_steps": 16324, "loss": 0.2294, "lr": 3.689034556353549e-05, "epoch": 1.2104682858105418, "percentage": 60.52, "elapsed_time": "6:00:00", "remaining_time": "3:54:48", "throughput": 3055.22, "total_tokens": 65994224}
{"current_steps": 9885, "total_steps": 16324, "loss": 0.205, "lr": 3.6841476219297925e-05, "epoch": 1.2110808244770452, "percentage": 60.56, "elapsed_time": "6:01:00", "remaining_time": "3:55:09", "throughput": 3048.37, "total_tokens": 66028432}
{"current_steps": 9890, "total_steps": 16324, "loss": 0.2232, "lr": 3.679262037684926e-05, "epoch": 1.2116933631435485, "percentage": 60.59, "elapsed_time": "6:02:01", "remaining_time": "3:55:30", "throughput": 3041.36, "total_tokens": 66061560}
{"current_steps": 9895, "total_steps": 16324, "loss": 0.2588, "lr": 3.674377808631979e-05, "epoch": 1.2123059018100517, "percentage": 60.62, "elapsed_time": "6:03:00", "remaining_time": "3:55:51", "throughput": 3034.58, "total_tokens": 66094600}
{"current_steps": 9900, "total_steps": 16324, "loss": 0.2593, "lr": 3.669494939782599e-05, "epoch": 1.2129184404765552, "percentage": 60.65, "elapsed_time": "6:04:00", "remaining_time": "3:56:12", "throughput": 3027.76, "total_tokens": 66127864}
{"current_steps": 9900, "total_steps": 16324, "eval_loss": 0.12373431026935577, "epoch": 1.2129184404765552, "percentage": 60.65, "elapsed_time": "6:04:20", "remaining_time": "3:56:24", "throughput": 3025.06, "total_tokens": 66127864}
{"current_steps": 9905, "total_steps": 16324, "loss": 0.2379, "lr": 3.664613436147032e-05, "epoch": 1.2135309791430584, "percentage": 60.68, "elapsed_time": "6:05:25", "remaining_time": "3:56:49", "throughput": 3017.5, "total_tokens": 66160792}
{"current_steps": 9910, "total_steps": 16324, "loss": 0.2287, "lr": 3.659733302734124e-05, "epoch": 1.2141435178095616, "percentage": 60.71, "elapsed_time": "6:06:25", "remaining_time": "3:57:09", "throughput": 3010.8, "total_tokens": 66194432}
{"current_steps": 9915, "total_steps": 16324, "loss": 0.2561, "lr": 3.654854544551318e-05, "epoch": 1.214756056476065, "percentage": 60.74, "elapsed_time": "6:07:25", "remaining_time": "3:57:30", "throughput": 3004.04, "total_tokens": 66226944}
{"current_steps": 9920, "total_steps": 16324, "loss": 0.2355, "lr": 3.649977166604642e-05, "epoch": 1.2153685951425683, "percentage": 60.77, "elapsed_time": "6:08:26", "remaining_time": "3:57:51", "throughput": 2997.36, "total_tokens": 66260416}
{"current_steps": 9925, "total_steps": 16324, "loss": 0.252, "lr": 3.645101173898712e-05, "epoch": 1.2159811338090716, "percentage": 60.8, "elapsed_time": "6:09:26", "remaining_time": "3:58:11", "throughput": 2990.79, "total_tokens": 66294488}
{"current_steps": 9930, "total_steps": 16324, "loss": 0.2395, "lr": 3.640226571436718e-05, "epoch": 1.216593672475575, "percentage": 60.83, "elapsed_time": "6:10:26", "remaining_time": "3:58:31", "throughput": 2984.26, "total_tokens": 66328552}
{"current_steps": 9935, "total_steps": 16324, "loss": 0.2348, "lr": 3.635353364220426e-05, "epoch": 1.2172062111420783, "percentage": 60.86, "elapsed_time": "6:11:26", "remaining_time": "3:58:52", "throughput": 2977.66, "total_tokens": 66362368}
{"current_steps": 9940, "total_steps": 16324, "loss": 0.2409, "lr": 3.630481557250171e-05, "epoch": 1.2178187498085817, "percentage": 60.89, "elapsed_time": "6:12:27", "remaining_time": "3:59:12", "throughput": 2971.06, "total_tokens": 66395760}
{"current_steps": 9945, "total_steps": 16324, "loss": 0.226, "lr": 3.625611155524849e-05, "epoch": 1.218431288475085, "percentage": 60.92, "elapsed_time": "6:13:28", "remaining_time": "3:59:33", "throughput": 2964.47, "total_tokens": 66429176}
{"current_steps": 9950, "total_steps": 16324, "loss": 0.2589, "lr": 3.620742164041916e-05, "epoch": 1.2190438271415882, "percentage": 60.95, "elapsed_time": "6:14:29", "remaining_time": "3:59:53", "throughput": 2957.9, "total_tokens": 66462256}
{"current_steps": 9950, "total_steps": 16324, "eval_loss": 0.16011233627796173, "epoch": 1.2190438271415882, "percentage": 60.95, "elapsed_time": "6:14:49", "remaining_time": "4:00:07", "throughput": 2955.22, "total_tokens": 66462256}
{"current_steps": 9955, "total_steps": 16324, "loss": 0.2162, "lr": 3.615874587797381e-05, "epoch": 1.2196563658080917, "percentage": 60.98, "elapsed_time": "6:15:56", "remaining_time": "4:00:30", "throughput": 2948.03, "total_tokens": 66495928}
{"current_steps": 9960, "total_steps": 16324, "loss": 0.2371, "lr": 3.6110084317858005e-05, "epoch": 1.220268904474595, "percentage": 61.01, "elapsed_time": "6:16:55", "remaining_time": "4:00:50", "throughput": 2941.75, "total_tokens": 66529384}
{"current_steps": 9965, "total_steps": 16324, "loss": 0.2425, "lr": 3.606143701000271e-05, "epoch": 1.2208814431410984, "percentage": 61.05, "elapsed_time": "6:17:54", "remaining_time": "4:01:09", "throughput": 2935.58, "total_tokens": 66563312}
{"current_steps": 9970, "total_steps": 16324, "loss": 0.228, "lr": 3.6012804004324326e-05, "epoch": 1.2214939818076016, "percentage": 61.08, "elapsed_time": "6:18:55", "remaining_time": "4:01:29", "throughput": 2929.25, "total_tokens": 66596840}
{"current_steps": 9975, "total_steps": 16324, "loss": 0.2407, "lr": 3.596418535072451e-05, "epoch": 1.2221065204741048, "percentage": 61.11, "elapsed_time": "6:19:54", "remaining_time": "4:01:48", "throughput": 2923.05, "total_tokens": 66630168}
{"current_steps": 9980, "total_steps": 16324, "loss": 0.2049, "lr": 3.591558109909025e-05, "epoch": 1.2227190591406083, "percentage": 61.14, "elapsed_time": "6:20:53", "remaining_time": "4:02:07", "throughput": 2917.02, "total_tokens": 66663848}
{"current_steps": 9985, "total_steps": 16324, "loss": 0.1661, "lr": 3.586699129929373e-05, "epoch": 1.2233315978071115, "percentage": 61.17, "elapsed_time": "6:21:53", "remaining_time": "4:02:26", "throughput": 2910.85, "total_tokens": 66698912}
{"current_steps": 9990, "total_steps": 16324, "loss": 0.2329, "lr": 3.5818416001192276e-05, "epoch": 1.223944136473615, "percentage": 61.2, "elapsed_time": "6:22:53", "remaining_time": "4:02:46", "throughput": 2904.73, "total_tokens": 66732112}
{"current_steps": 9995, "total_steps": 16324, "loss": 0.2364, "lr": 3.576985525462841e-05, "epoch": 1.2245566751401182, "percentage": 61.23, "elapsed_time": "6:23:52", "remaining_time": "4:03:04", "throughput": 2898.72, "total_tokens": 66765832}
{"current_steps": 10000, "total_steps": 16324, "loss": 0.2736, "lr": 3.572130910942964e-05, "epoch": 1.2251692138066215, "percentage": 61.26, "elapsed_time": "6:24:54", "remaining_time": "4:03:24", "throughput": 2892.42, "total_tokens": 66798760}
{"current_steps": 10000, "total_steps": 16324, "eval_loss": 0.15618817508220673, "epoch": 1.2251692138066215, "percentage": 61.26, "elapsed_time": "6:25:14", "remaining_time": "4:03:37", "throughput": 2889.87, "total_tokens": 66798760}
{"current_steps": 10005, "total_steps": 16324, "loss": 0.2139, "lr": 3.567277761540856e-05, "epoch": 1.225781752473125, "percentage": 61.29, "elapsed_time": "6:26:20", "remaining_time": "4:04:00", "throughput": 2883.13, "total_tokens": 66832616}
{"current_steps": 10010, "total_steps": 16324, "loss": 0.2183, "lr": 3.56242608223627e-05, "epoch": 1.2263942911396282, "percentage": 61.32, "elapsed_time": "6:27:21", "remaining_time": "4:04:19", "throughput": 2877.09, "total_tokens": 66866584}
{"current_steps": 10015, "total_steps": 16324, "loss": 0.2262, "lr": 3.5575758780074475e-05, "epoch": 1.2270068298061316, "percentage": 61.35, "elapsed_time": "6:28:22", "remaining_time": "4:04:39", "throughput": 2870.9, "total_tokens": 66900296}
{"current_steps": 10020, "total_steps": 16324, "loss": 0.2384, "lr": 3.5527271538311205e-05, "epoch": 1.2276193684726349, "percentage": 61.38, "elapsed_time": "6:29:23", "remaining_time": "4:04:58", "throughput": 2864.95, "total_tokens": 66933888}
{"current_steps": 10025, "total_steps": 16324, "loss": 0.257, "lr": 3.5478799146825024e-05, "epoch": 1.228231907139138, "percentage": 61.41, "elapsed_time": "6:30:22", "remaining_time": "4:05:17", "throughput": 2859.03, "total_tokens": 66966912}
{"current_steps": 10030, "total_steps": 16324, "loss": 0.2248, "lr": 3.543034165535282e-05, "epoch": 1.2288444458056416, "percentage": 61.44, "elapsed_time": "6:31:24", "remaining_time": "4:05:36", "throughput": 2852.97, "total_tokens": 67000680}
{"current_steps": 10035, "total_steps": 16324, "loss": 0.2604, "lr": 3.538189911361618e-05, "epoch": 1.2294569844721448, "percentage": 61.47, "elapsed_time": "6:32:23", "remaining_time": "4:05:55", "throughput": 2847.18, "total_tokens": 67033808}
{"current_steps": 10040, "total_steps": 16324, "loss": 0.1984, "lr": 3.5333471571321375e-05, "epoch": 1.230069523138648, "percentage": 61.5, "elapsed_time": "6:33:24", "remaining_time": "4:06:13", "throughput": 2841.36, "total_tokens": 67067608}
{"current_steps": 10045, "total_steps": 16324, "loss": 0.1937, "lr": 3.528505907815925e-05, "epoch": 1.2306820618051515, "percentage": 61.54, "elapsed_time": "6:34:25", "remaining_time": "4:06:32", "throughput": 2835.44, "total_tokens": 67101040}
{"current_steps": 10050, "total_steps": 16324, "loss": 0.2419, "lr": 3.523666168380525e-05, "epoch": 1.2312946004716547, "percentage": 61.57, "elapsed_time": "6:35:24", "remaining_time": "4:06:50", "throughput": 2829.7, "total_tokens": 67134040}
{"current_steps": 10050, "total_steps": 16324, "eval_loss": 0.12978222966194153, "epoch": 1.2312946004716547, "percentage": 61.57, "elapsed_time": "6:35:44", "remaining_time": "4:07:03", "throughput": 2827.32, "total_tokens": 67134040}
{"current_steps": 10055, "total_steps": 16324, "loss": 0.2511, "lr": 3.51882794379193e-05, "epoch": 1.2319071391381582, "percentage": 61.6, "elapsed_time": "6:36:49", "remaining_time": "4:07:24", "throughput": 2821.02, "total_tokens": 67167360}
{"current_steps": 10060, "total_steps": 16324, "loss": 0.2128, "lr": 3.513991239014579e-05, "epoch": 1.2325196778046614, "percentage": 61.63, "elapsed_time": "6:37:50", "remaining_time": "4:07:43", "throughput": 2815.29, "total_tokens": 67201216}
{"current_steps": 10065, "total_steps": 16324, "loss": 0.2481, "lr": 3.509156059011352e-05, "epoch": 1.2331322164711647, "percentage": 61.66, "elapsed_time": "6:38:50", "remaining_time": "4:08:01", "throughput": 2809.54, "total_tokens": 67234824}
{"current_steps": 10070, "total_steps": 16324, "loss": 0.2275, "lr": 3.504322408743562e-05, "epoch": 1.2337447551376681, "percentage": 61.69, "elapsed_time": "6:39:52", "remaining_time": "4:08:20", "throughput": 2803.72, "total_tokens": 67268160}
{"current_steps": 10075, "total_steps": 16324, "loss": 0.2403, "lr": 3.499490293170956e-05, "epoch": 1.2343572938041714, "percentage": 61.72, "elapsed_time": "6:40:53", "remaining_time": "4:08:39", "throughput": 2797.95, "total_tokens": 67301776}
{"current_steps": 10080, "total_steps": 16324, "loss": 0.2145, "lr": 3.494659717251704e-05, "epoch": 1.2349698324706746, "percentage": 61.75, "elapsed_time": "6:41:54", "remaining_time": "4:08:57", "throughput": 2792.32, "total_tokens": 67336184}
{"current_steps": 10085, "total_steps": 16324, "loss": 0.2269, "lr": 3.489830685942397e-05, "epoch": 1.235582371137178, "percentage": 61.78, "elapsed_time": "6:42:54", "remaining_time": "4:09:15", "throughput": 2786.85, "total_tokens": 67370240}
{"current_steps": 10090, "total_steps": 16324, "loss": 0.2157, "lr": 3.485003204198041e-05, "epoch": 1.2361949098036813, "percentage": 61.81, "elapsed_time": "6:43:53", "remaining_time": "4:09:32", "throughput": 2781.49, "total_tokens": 67404432}
{"current_steps": 10095, "total_steps": 16324, "loss": 0.2186, "lr": 3.480177276972051e-05, "epoch": 1.2368074484701848, "percentage": 61.84, "elapsed_time": "6:44:53", "remaining_time": "4:09:49", "throughput": 2775.98, "total_tokens": 67437912}
{"current_steps": 10100, "total_steps": 16324, "loss": 0.2574, "lr": 3.475352909216246e-05, "epoch": 1.237419987136688, "percentage": 61.87, "elapsed_time": "6:45:52", "remaining_time": "4:10:07", "throughput": 2770.56, "total_tokens": 67471000}
{"current_steps": 10100, "total_steps": 16324, "eval_loss": 0.21491877734661102, "epoch": 1.237419987136688, "percentage": 61.87, "elapsed_time": "6:46:12", "remaining_time": "4:10:19", "throughput": 2768.35, "total_tokens": 67471000}
{"current_steps": 10105, "total_steps": 16324, "loss": 0.2594, "lr": 3.47053010588085e-05, "epoch": 1.2380325258031912, "percentage": 61.9, "elapsed_time": "6:47:15", "remaining_time": "4:10:38", "throughput": 2762.49, "total_tokens": 67504136}
{"current_steps": 10110, "total_steps": 16324, "loss": 0.2491, "lr": 3.465708871914475e-05, "epoch": 1.2386450644696947, "percentage": 61.93, "elapsed_time": "6:48:16", "remaining_time": "4:10:56", "throughput": 2757.07, "total_tokens": 67537536}
{"current_steps": 10115, "total_steps": 16324, "loss": 0.2029, "lr": 3.4608892122641295e-05, "epoch": 1.239257603136198, "percentage": 61.96, "elapsed_time": "6:49:15", "remaining_time": "4:11:13", "throughput": 2751.75, "total_tokens": 67571048}
{"current_steps": 10120, "total_steps": 16324, "loss": 0.2015, "lr": 3.4560711318752016e-05, "epoch": 1.2398701418027014, "percentage": 61.99, "elapsed_time": "6:50:14", "remaining_time": "4:11:29", "throughput": 2746.54, "total_tokens": 67604768}
{"current_steps": 10125, "total_steps": 16324, "loss": 0.2179, "lr": 3.451254635691458e-05, "epoch": 1.2404826804692046, "percentage": 62.03, "elapsed_time": "6:51:14", "remaining_time": "4:11:46", "throughput": 2741.23, "total_tokens": 67638424}
{"current_steps": 10130, "total_steps": 16324, "loss": 0.2452, "lr": 3.446439728655047e-05, "epoch": 1.2410952191357079, "percentage": 62.06, "elapsed_time": "6:52:13", "remaining_time": "4:12:03", "throughput": 2736.03, "total_tokens": 67671792}
{"current_steps": 10135, "total_steps": 16324, "loss": 0.2498, "lr": 3.441626415706477e-05, "epoch": 1.2417077578022113, "percentage": 62.09, "elapsed_time": "6:53:12", "remaining_time": "4:12:19", "throughput": 2730.9, "total_tokens": 67705816}
{"current_steps": 10140, "total_steps": 16324, "loss": 0.1976, "lr": 3.4368147017846295e-05, "epoch": 1.2423202964687146, "percentage": 62.12, "elapsed_time": "6:54:12", "remaining_time": "4:12:36", "throughput": 2725.66, "total_tokens": 67740120}
{"current_steps": 10145, "total_steps": 16324, "loss": 0.2408, "lr": 3.432004591826739e-05, "epoch": 1.242932835135218, "percentage": 62.15, "elapsed_time": "6:55:12", "remaining_time": "4:12:53", "throughput": 2720.45, "total_tokens": 67773216}
{"current_steps": 10150, "total_steps": 16324, "loss": 0.2124, "lr": 3.427196090768395e-05, "epoch": 1.2435453738017213, "percentage": 62.18, "elapsed_time": "6:56:11", "remaining_time": "4:13:09", "throughput": 2715.41, "total_tokens": 67807536}
{"current_steps": 10150, "total_steps": 16324, "eval_loss": 0.1516389101743698, "epoch": 1.2435453738017213, "percentage": 62.18, "elapsed_time": "6:56:30", "remaining_time": "4:13:21", "throughput": 2713.33, "total_tokens": 67807536}
{"current_steps": 10155, "total_steps": 16324, "loss": 0.2761, "lr": 3.4223892035435414e-05, "epoch": 1.2441579124682245, "percentage": 62.21, "elapsed_time": "6:57:36", "remaining_time": "4:13:41", "throughput": 2707.53, "total_tokens": 67840480}
{"current_steps": 10160, "total_steps": 16324, "loss": 0.2171, "lr": 3.4175839350844576e-05, "epoch": 1.244770451134728, "percentage": 62.24, "elapsed_time": "6:58:36", "remaining_time": "4:13:57", "throughput": 2702.39, "total_tokens": 67873680}
{"current_steps": 10165, "total_steps": 16324, "loss": 0.2261, "lr": 3.41278029032177e-05, "epoch": 1.2453829898012312, "percentage": 62.27, "elapsed_time": "6:59:35", "remaining_time": "4:14:13", "throughput": 2697.36, "total_tokens": 67907360}
{"current_steps": 10170, "total_steps": 16324, "loss": 0.2116, "lr": 3.407978274184435e-05, "epoch": 1.2459955284677346, "percentage": 62.3, "elapsed_time": "7:00:35", "remaining_time": "4:14:30", "throughput": 2692.31, "total_tokens": 67941136}
{"current_steps": 10175, "total_steps": 16324, "loss": 0.1925, "lr": 3.4031778915997375e-05, "epoch": 1.2466080671342379, "percentage": 62.33, "elapsed_time": "7:01:34", "remaining_time": "4:14:46", "throughput": 2687.34, "total_tokens": 67975016}
{"current_steps": 10180, "total_steps": 16324, "loss": 0.194, "lr": 3.3983791474932864e-05, "epoch": 1.2472206058007411, "percentage": 62.36, "elapsed_time": "7:02:34", "remaining_time": "4:15:02", "throughput": 2682.38, "total_tokens": 68009336}
{"current_steps": 10185, "total_steps": 16324, "loss": 0.243, "lr": 3.393582046789012e-05, "epoch": 1.2478331444672446, "percentage": 62.39, "elapsed_time": "7:03:34", "remaining_time": "4:15:18", "throughput": 2677.34, "total_tokens": 68042592}
{"current_steps": 10190, "total_steps": 16324, "loss": 0.2092, "lr": 3.388786594409155e-05, "epoch": 1.2484456831337478, "percentage": 62.42, "elapsed_time": "7:04:32", "remaining_time": "4:15:33", "throughput": 2672.51, "total_tokens": 68076744}
{"current_steps": 10195, "total_steps": 16324, "loss": 0.2521, "lr": 3.383992795274268e-05, "epoch": 1.249058221800251, "percentage": 62.45, "elapsed_time": "7:05:31", "remaining_time": "4:15:49", "throughput": 2667.67, "total_tokens": 68110272}
{"current_steps": 10200, "total_steps": 16324, "loss": 0.2047, "lr": 3.379200654303204e-05, "epoch": 1.2496707604667545, "percentage": 62.48, "elapsed_time": "7:06:30", "remaining_time": "4:16:04", "throughput": 2662.87, "total_tokens": 68143912}
{"current_steps": 10200, "total_steps": 16324, "eval_loss": 0.16254977881908417, "epoch": 1.2496707604667545, "percentage": 62.48, "elapsed_time": "7:06:49", "remaining_time": "4:16:15", "throughput": 2660.85, "total_tokens": 68143912}
{"current_steps": 10205, "total_steps": 16324, "loss": 0.1799, "lr": 3.374410176413117e-05, "epoch": 1.2502832991332578, "percentage": 62.52, "elapsed_time": "7:07:54", "remaining_time": "4:16:34", "throughput": 2655.51, "total_tokens": 68177808}
{"current_steps": 10210, "total_steps": 16324, "loss": 0.2216, "lr": 3.369621366519453e-05, "epoch": 1.250895837799761, "percentage": 62.55, "elapsed_time": "7:08:52", "remaining_time": "4:16:49", "throughput": 2650.78, "total_tokens": 68211568}
{"current_steps": 10215, "total_steps": 16324, "loss": 0.2399, "lr": 3.364834229535946e-05, "epoch": 1.2515083764662644, "percentage": 62.58, "elapsed_time": "7:09:50", "remaining_time": "4:17:03", "throughput": 2646.09, "total_tokens": 68244728}
{"current_steps": 10220, "total_steps": 16324, "loss": 0.2358, "lr": 3.360048770374618e-05, "epoch": 1.2521209151327677, "percentage": 62.61, "elapsed_time": "7:10:50", "remaining_time": "4:17:19", "throughput": 2641.29, "total_tokens": 68278040}
{"current_steps": 10225, "total_steps": 16324, "loss": 0.2189, "lr": 3.3552649939457634e-05, "epoch": 1.2527334537992711, "percentage": 62.64, "elapsed_time": "7:11:48", "remaining_time": "4:17:34", "throughput": 2636.62, "total_tokens": 68312056}
{"current_steps": 10230, "total_steps": 16324, "loss": 0.2465, "lr": 3.350482905157952e-05, "epoch": 1.2533459924657744, "percentage": 62.67, "elapsed_time": "7:12:47", "remaining_time": "4:17:48", "throughput": 2631.95, "total_tokens": 68345568}
{"current_steps": 10235, "total_steps": 16324, "loss": 0.2393, "lr": 3.345702508918024e-05, "epoch": 1.2539585311322776, "percentage": 62.7, "elapsed_time": "7:13:47", "remaining_time": "4:18:04", "throughput": 2627.23, "total_tokens": 68379152}
{"current_steps": 10240, "total_steps": 16324, "loss": 0.27, "lr": 3.34092381013108e-05, "epoch": 1.254571069798781, "percentage": 62.73, "elapsed_time": "7:14:45", "remaining_time": "4:18:18", "throughput": 2622.58, "total_tokens": 68412584}
{"current_steps": 10245, "total_steps": 16324, "loss": 0.2215, "lr": 3.336146813700482e-05, "epoch": 1.2551836084652843, "percentage": 62.76, "elapsed_time": "7:15:44", "remaining_time": "4:18:33", "throughput": 2617.99, "total_tokens": 68446232}
{"current_steps": 10250, "total_steps": 16324, "loss": 0.2368, "lr": 3.3313715245278434e-05, "epoch": 1.2557961471317878, "percentage": 62.79, "elapsed_time": "7:16:44", "remaining_time": "4:18:48", "throughput": 2613.28, "total_tokens": 68479264}
{"current_steps": 10250, "total_steps": 16324, "eval_loss": 0.1966872662305832, "epoch": 1.2557961471317878, "percentage": 62.79, "elapsed_time": "7:17:03", "remaining_time": "4:18:59", "throughput": 2611.37, "total_tokens": 68479264}
{"current_steps": 10255, "total_steps": 16324, "loss": 0.1938, "lr": 3.326597947513025e-05, "epoch": 1.256408685798291, "percentage": 62.82, "elapsed_time": "7:18:07", "remaining_time": "4:19:17", "throughput": 2606.27, "total_tokens": 68513176}
{"current_steps": 10260, "total_steps": 16324, "loss": 0.2257, "lr": 3.321826087554129e-05, "epoch": 1.2570212244647943, "percentage": 62.85, "elapsed_time": "7:19:06", "remaining_time": "4:19:31", "throughput": 2601.75, "total_tokens": 68547256}
{"current_steps": 10265, "total_steps": 16324, "loss": 0.2599, "lr": 3.317055949547503e-05, "epoch": 1.2576337631312977, "percentage": 62.88, "elapsed_time": "7:20:07", "remaining_time": "4:19:47", "throughput": 2597.01, "total_tokens": 68580448}
{"current_steps": 10270, "total_steps": 16324, "loss": 0.2444, "lr": 3.3122875383877194e-05, "epoch": 1.258246301797801, "percentage": 62.91, "elapsed_time": "7:21:07", "remaining_time": "4:20:02", "throughput": 2592.41, "total_tokens": 68614144}
{"current_steps": 10275, "total_steps": 16324, "loss": 0.2425, "lr": 3.307520858967586e-05, "epoch": 1.2588588404643044, "percentage": 62.94, "elapsed_time": "7:22:06", "remaining_time": "4:20:16", "throughput": 2587.89, "total_tokens": 68647664}
{"current_steps": 10280, "total_steps": 16324, "loss": 0.2108, "lr": 3.302755916178128e-05, "epoch": 1.2594713791308076, "percentage": 62.97, "elapsed_time": "7:23:06", "remaining_time": "4:20:31", "throughput": 2583.28, "total_tokens": 68681192}
{"current_steps": 10285, "total_steps": 16324, "loss": 0.2493, "lr": 3.297992714908589e-05, "epoch": 1.2600839177973109, "percentage": 63.01, "elapsed_time": "7:24:06", "remaining_time": "4:20:46", "throughput": 2578.71, "total_tokens": 68714600}
{"current_steps": 10290, "total_steps": 16324, "loss": 0.2286, "lr": 3.293231260046431e-05, "epoch": 1.2606964564638143, "percentage": 63.04, "elapsed_time": "7:25:06", "remaining_time": "4:21:00", "throughput": 2574.21, "total_tokens": 68748432}
{"current_steps": 10295, "total_steps": 16324, "loss": 0.2285, "lr": 3.288471556477317e-05, "epoch": 1.2613089951303176, "percentage": 63.07, "elapsed_time": "7:26:07", "remaining_time": "4:21:15", "throughput": 2569.61, "total_tokens": 68781728}
{"current_steps": 10300, "total_steps": 16324, "loss": 0.2253, "lr": 3.2837136090851205e-05, "epoch": 1.261921533796821, "percentage": 63.1, "elapsed_time": "7:27:07", "remaining_time": "4:21:30", "throughput": 2565.11, "total_tokens": 68815336}
{"current_steps": 10300, "total_steps": 16324, "eval_loss": 0.2143474966287613, "epoch": 1.261921533796821, "percentage": 63.1, "elapsed_time": "7:27:27", "remaining_time": "4:21:41", "throughput": 2563.24, "total_tokens": 68815336}
{"current_steps": 10305, "total_steps": 16324, "loss": 0.2482, "lr": 3.278957422751907e-05, "epoch": 1.2625340724633243, "percentage": 63.13, "elapsed_time": "7:28:31", "remaining_time": "4:21:58", "throughput": 2558.32, "total_tokens": 68848592}
{"current_steps": 10310, "total_steps": 16324, "loss": 0.228, "lr": 3.274203002357935e-05, "epoch": 1.2631466111298275, "percentage": 63.16, "elapsed_time": "7:29:32", "remaining_time": "4:22:13", "throughput": 2553.8, "total_tokens": 68882344}
{"current_steps": 10315, "total_steps": 16324, "loss": 0.2314, "lr": 3.2694503527816565e-05, "epoch": 1.263759149796331, "percentage": 63.19, "elapsed_time": "7:30:31", "remaining_time": "4:22:27", "throughput": 2549.45, "total_tokens": 68916040}
{"current_steps": 10320, "total_steps": 16324, "loss": 0.2507, "lr": 3.2646994788997006e-05, "epoch": 1.2643716884628342, "percentage": 63.22, "elapsed_time": "7:31:30", "remaining_time": "4:22:40", "throughput": 2545.14, "total_tokens": 68949568}
{"current_steps": 10325, "total_steps": 16324, "loss": 0.2489, "lr": 3.2599503855868775e-05, "epoch": 1.2649842271293377, "percentage": 63.25, "elapsed_time": "7:32:29", "remaining_time": "4:22:54", "throughput": 2540.8, "total_tokens": 68982152}
{"current_steps": 10330, "total_steps": 16324, "loss": 0.2389, "lr": 3.255203077716169e-05, "epoch": 1.265596765795841, "percentage": 63.28, "elapsed_time": "7:33:29", "remaining_time": "4:23:08", "throughput": 2536.49, "total_tokens": 69015864}
{"current_steps": 10335, "total_steps": 16324, "loss": 0.2492, "lr": 3.2504575601587294e-05, "epoch": 1.2662093044623441, "percentage": 63.31, "elapsed_time": "7:34:28", "remaining_time": "4:23:21", "throughput": 2532.25, "total_tokens": 69049896}
{"current_steps": 10340, "total_steps": 16324, "loss": 0.2217, "lr": 3.245713837783866e-05, "epoch": 1.2668218431288474, "percentage": 63.34, "elapsed_time": "7:35:27", "remaining_time": "4:23:35", "throughput": 2527.99, "total_tokens": 69083632}
{"current_steps": 10345, "total_steps": 16324, "loss": 0.2422, "lr": 3.2409719154590534e-05, "epoch": 1.2674343817953508, "percentage": 63.37, "elapsed_time": "7:36:26", "remaining_time": "4:23:48", "throughput": 2523.71, "total_tokens": 69116488}
{"current_steps": 10350, "total_steps": 16324, "loss": 0.2333, "lr": 3.236231798049915e-05, "epoch": 1.268046920461854, "percentage": 63.4, "elapsed_time": "7:37:25", "remaining_time": "4:24:01", "throughput": 2519.5, "total_tokens": 69150112}
{"current_steps": 10350, "total_steps": 16324, "eval_loss": 0.1691417545080185, "epoch": 1.268046920461854, "percentage": 63.4, "elapsed_time": "7:37:45", "remaining_time": "4:24:12", "throughput": 2517.72, "total_tokens": 69150112}
{"current_steps": 10355, "total_steps": 16324, "loss": 0.217, "lr": 3.231493490420224e-05, "epoch": 1.2686594591283575, "percentage": 63.43, "elapsed_time": "7:38:50", "remaining_time": "4:24:29", "throughput": 2513.04, "total_tokens": 69184024}
{"current_steps": 10360, "total_steps": 16324, "loss": 0.2475, "lr": 3.2267569974318964e-05, "epoch": 1.2692719977948608, "percentage": 63.46, "elapsed_time": "7:39:50", "remaining_time": "4:24:42", "throughput": 2508.81, "total_tokens": 69218568}
{"current_steps": 10365, "total_steps": 16324, "loss": 0.1892, "lr": 3.2220223239449834e-05, "epoch": 1.269884536461364, "percentage": 63.5, "elapsed_time": "7:40:49", "remaining_time": "4:24:56", "throughput": 2504.68, "total_tokens": 69252568}
{"current_steps": 10370, "total_steps": 16324, "loss": 0.2745, "lr": 3.2172894748176743e-05, "epoch": 1.2704970751278675, "percentage": 63.53, "elapsed_time": "7:41:48", "remaining_time": "4:25:08", "throughput": 2500.59, "total_tokens": 69286336}
{"current_steps": 10375, "total_steps": 16324, "loss": 0.2482, "lr": 3.21255845490628e-05, "epoch": 1.2711096137943707, "percentage": 63.56, "elapsed_time": "7:42:48", "remaining_time": "4:25:22", "throughput": 2496.32, "total_tokens": 69319784}
{"current_steps": 10380, "total_steps": 16324, "loss": 0.2778, "lr": 3.207829269065242e-05, "epoch": 1.2717221524608742, "percentage": 63.59, "elapsed_time": "7:43:48", "remaining_time": "4:25:35", "throughput": 2492.18, "total_tokens": 69352992}
{"current_steps": 10385, "total_steps": 16324, "loss": 0.2333, "lr": 3.203101922147113e-05, "epoch": 1.2723346911273774, "percentage": 63.62, "elapsed_time": "7:44:47", "remaining_time": "4:25:48", "throughput": 2488.06, "total_tokens": 69386192}
{"current_steps": 10390, "total_steps": 16324, "loss": 0.1913, "lr": 3.198376419002563e-05, "epoch": 1.2729472297938806, "percentage": 63.65, "elapsed_time": "7:45:48", "remaining_time": "4:26:02", "throughput": 2483.82, "total_tokens": 69420224}
{"current_steps": 10395, "total_steps": 16324, "loss": 0.2391, "lr": 3.19365276448037e-05, "epoch": 1.273559768460384, "percentage": 63.68, "elapsed_time": "7:46:49", "remaining_time": "4:26:15", "throughput": 2479.66, "total_tokens": 69453232}
{"current_steps": 10400, "total_steps": 16324, "loss": 0.2009, "lr": 3.188930963427411e-05, "epoch": 1.2741723071268873, "percentage": 63.71, "elapsed_time": "7:47:48", "remaining_time": "4:26:28", "throughput": 2475.65, "total_tokens": 69486944}
{"current_steps": 10400, "total_steps": 16324, "eval_loss": 0.3153703510761261, "epoch": 1.2741723071268873, "percentage": 63.71, "elapsed_time": "7:48:07", "remaining_time": "4:26:39", "throughput": 2473.92, "total_tokens": 69486944}
{"current_steps": 10405, "total_steps": 16324, "loss": 0.2063, "lr": 3.184211020688667e-05, "epoch": 1.2747848457933908, "percentage": 63.74, "elapsed_time": "7:49:13", "remaining_time": "4:26:55", "throughput": 2469.35, "total_tokens": 69521024}
{"current_steps": 10410, "total_steps": 16324, "loss": 0.1923, "lr": 3.179492941107207e-05, "epoch": 1.275397384459894, "percentage": 63.77, "elapsed_time": "7:50:13", "remaining_time": "4:27:08", "throughput": 2465.29, "total_tokens": 69555328}
{"current_steps": 10415, "total_steps": 16324, "loss": 0.2565, "lr": 3.174776729524196e-05, "epoch": 1.2760099231263973, "percentage": 63.8, "elapsed_time": "7:51:13", "remaining_time": "4:27:21", "throughput": 2461.25, "total_tokens": 69589040}
{"current_steps": 10420, "total_steps": 16324, "loss": 0.2575, "lr": 3.17006239077887e-05, "epoch": 1.2766224617929007, "percentage": 63.83, "elapsed_time": "7:52:14", "remaining_time": "4:27:34", "throughput": 2457.13, "total_tokens": 69622376}
{"current_steps": 10425, "total_steps": 16324, "loss": 0.2472, "lr": 3.165349929708553e-05, "epoch": 1.277235000459404, "percentage": 63.86, "elapsed_time": "7:53:14", "remaining_time": "4:27:47", "throughput": 2453.07, "total_tokens": 69654960}
{"current_steps": 10430, "total_steps": 16324, "loss": 0.2499, "lr": 3.160639351148639e-05, "epoch": 1.2778475391259074, "percentage": 63.89, "elapsed_time": "7:54:14", "remaining_time": "4:27:59", "throughput": 2449.13, "total_tokens": 69688264}
{"current_steps": 10435, "total_steps": 16324, "loss": 0.1853, "lr": 3.155930659932593e-05, "epoch": 1.2784600777924107, "percentage": 63.92, "elapsed_time": "7:55:14", "remaining_time": "4:28:12", "throughput": 2445.16, "total_tokens": 69722104}
{"current_steps": 10440, "total_steps": 16324, "loss": 0.192, "lr": 3.15122386089194e-05, "epoch": 1.279072616458914, "percentage": 63.95, "elapsed_time": "7:56:14", "remaining_time": "4:28:24", "throughput": 2441.23, "total_tokens": 69755784}
{"current_steps": 10445, "total_steps": 16324, "loss": 0.2412, "lr": 3.146518958856264e-05, "epoch": 1.2796851551254174, "percentage": 63.99, "elapsed_time": "7:57:12", "remaining_time": "4:28:35", "throughput": 2437.43, "total_tokens": 69789632}
{"current_steps": 10450, "total_steps": 16324, "loss": 0.2546, "lr": 3.1418159586532055e-05, "epoch": 1.2802976937919206, "percentage": 64.02, "elapsed_time": "7:58:12", "remaining_time": "4:28:48", "throughput": 2433.5, "total_tokens": 69822376}
{"current_steps": 10450, "total_steps": 16324, "eval_loss": 0.18921419978141785, "epoch": 1.2802976937919206, "percentage": 64.02, "elapsed_time": "7:58:31", "remaining_time": "4:28:58", "throughput": 2431.86, "total_tokens": 69822376}
{"current_steps": 10455, "total_steps": 16324, "loss": 0.2298, "lr": 3.137114865108448e-05, "epoch": 1.280910232458424, "percentage": 64.05, "elapsed_time": "7:59:36", "remaining_time": "4:29:13", "throughput": 2427.56, "total_tokens": 69856272}
{"current_steps": 10460, "total_steps": 16324, "loss": 0.2138, "lr": 3.1324156830457265e-05, "epoch": 1.2815227711249273, "percentage": 64.08, "elapsed_time": "8:00:35", "remaining_time": "4:29:25", "throughput": 2423.75, "total_tokens": 69890232}
{"current_steps": 10465, "total_steps": 16324, "loss": 0.1803, "lr": 3.127718417286806e-05, "epoch": 1.2821353097914305, "percentage": 64.11, "elapsed_time": "8:01:34", "remaining_time": "4:29:36", "throughput": 2420.01, "total_tokens": 69924584}
{"current_steps": 10470, "total_steps": 16324, "loss": 0.2116, "lr": 3.1230230726514924e-05, "epoch": 1.2827478484579338, "percentage": 64.14, "elapsed_time": "8:02:34", "remaining_time": "4:29:49", "throughput": 2416.13, "total_tokens": 69957760}
{"current_steps": 10475, "total_steps": 16324, "loss": 0.23, "lr": 3.118329653957617e-05, "epoch": 1.2833603871244372, "percentage": 64.17, "elapsed_time": "8:03:34", "remaining_time": "4:30:01", "throughput": 2412.3, "total_tokens": 69991520}
{"current_steps": 10480, "total_steps": 16324, "loss": 0.2256, "lr": 3.113638166021032e-05, "epoch": 1.2839729257909407, "percentage": 64.2, "elapsed_time": "8:04:34", "remaining_time": "4:30:13", "throughput": 2408.44, "total_tokens": 70025056}
{"current_steps": 10485, "total_steps": 16324, "loss": 0.233, "lr": 3.1089486136556154e-05, "epoch": 1.284585464457444, "percentage": 64.23, "elapsed_time": "8:05:36", "remaining_time": "4:30:25", "throughput": 2404.49, "total_tokens": 70058384}
{"current_steps": 10490, "total_steps": 16324, "loss": 0.2252, "lr": 3.104261001673253e-05, "epoch": 1.2851980031239472, "percentage": 64.26, "elapsed_time": "8:06:37", "remaining_time": "4:30:38", "throughput": 2400.59, "total_tokens": 70091384}
{"current_steps": 10495, "total_steps": 16324, "loss": 0.2096, "lr": 3.099575334883846e-05, "epoch": 1.2858105417904504, "percentage": 64.29, "elapsed_time": "8:07:37", "remaining_time": "4:30:49", "throughput": 2396.82, "total_tokens": 70125232}
{"current_steps": 10500, "total_steps": 16324, "loss": 0.2345, "lr": 3.0948916180952904e-05, "epoch": 1.2864230804569539, "percentage": 64.32, "elapsed_time": "8:08:38", "remaining_time": "4:31:02", "throughput": 2392.97, "total_tokens": 70158672}
{"current_steps": 10500, "total_steps": 16324, "eval_loss": 0.16857217252254486, "epoch": 1.2864230804569539, "percentage": 64.32, "elapsed_time": "8:08:58", "remaining_time": "4:31:13", "throughput": 2391.36, "total_tokens": 70158672}
{"current_steps": 10505, "total_steps": 16324, "loss": 0.2116, "lr": 3.0902098561134904e-05, "epoch": 1.287035619123457, "percentage": 64.35, "elapsed_time": "8:10:04", "remaining_time": "4:31:27", "throughput": 2387.17, "total_tokens": 70192984}
{"current_steps": 10510, "total_steps": 16324, "loss": 0.2022, "lr": 3.085530053742337e-05, "epoch": 1.2876481577899606, "percentage": 64.38, "elapsed_time": "8:11:04", "remaining_time": "4:31:39", "throughput": 2383.47, "total_tokens": 70226672}
{"current_steps": 10515, "total_steps": 16324, "loss": 0.2485, "lr": 3.080852215783719e-05, "epoch": 1.2882606964564638, "percentage": 64.41, "elapsed_time": "8:12:05", "remaining_time": "4:31:51", "throughput": 2379.67, "total_tokens": 70260368}
{"current_steps": 10520, "total_steps": 16324, "loss": 0.2304, "lr": 3.0761763470375013e-05, "epoch": 1.288873235122967, "percentage": 64.44, "elapsed_time": "8:13:06", "remaining_time": "4:32:03", "throughput": 2375.89, "total_tokens": 70293664}
{"current_steps": 10525, "total_steps": 16324, "loss": 0.2487, "lr": 3.071502452301531e-05, "epoch": 1.2894857737894705, "percentage": 64.48, "elapsed_time": "8:14:07", "remaining_time": "4:32:14", "throughput": 2372.15, "total_tokens": 70327240}
{"current_steps": 10530, "total_steps": 16324, "loss": 0.2326, "lr": 3.066830536371633e-05, "epoch": 1.2900983124559737, "percentage": 64.51, "elapsed_time": "8:15:08", "remaining_time": "4:32:26", "throughput": 2368.33, "total_tokens": 70360248}
{"current_steps": 10535, "total_steps": 16324, "loss": 0.2181, "lr": 3.0621606040415975e-05, "epoch": 1.2907108511224772, "percentage": 64.54, "elapsed_time": "8:16:10", "remaining_time": "4:32:38", "throughput": 2364.56, "total_tokens": 70393960}
{"current_steps": 10540, "total_steps": 16324, "loss": 0.2285, "lr": 3.0574926601031806e-05, "epoch": 1.2913233897889804, "percentage": 64.57, "elapsed_time": "8:17:12", "remaining_time": "4:32:51", "throughput": 2360.74, "total_tokens": 70427088}
{"current_steps": 10545, "total_steps": 16324, "loss": 0.2164, "lr": 3.0528267093460985e-05, "epoch": 1.2919359284554837, "percentage": 64.6, "elapsed_time": "8:18:14", "remaining_time": "4:33:03", "throughput": 2356.95, "total_tokens": 70460952}
{"current_steps": 10550, "total_steps": 16324, "loss": 0.2319, "lr": 3.048162756558024e-05, "epoch": 1.2925484671219871, "percentage": 64.63, "elapsed_time": "8:19:16", "remaining_time": "4:33:15", "throughput": 2353.22, "total_tokens": 70494392}
{"current_steps": 10550, "total_steps": 16324, "eval_loss": 0.196364626288414, "epoch": 1.2925484671219871, "percentage": 64.63, "elapsed_time": "8:19:36", "remaining_time": "4:33:25", "throughput": 2351.68, "total_tokens": 70494392}
{"current_steps": 10555, "total_steps": 16324, "loss": 0.2058, "lr": 3.0435008065245756e-05, "epoch": 1.2931610057884904, "percentage": 64.66, "elapsed_time": "8:20:42", "remaining_time": "4:33:39", "throughput": 2347.66, "total_tokens": 70528424}
{"current_steps": 10560, "total_steps": 16324, "loss": 0.2432, "lr": 3.0388408640293198e-05, "epoch": 1.2937735444549938, "percentage": 64.69, "elapsed_time": "8:21:43", "remaining_time": "4:33:51", "throughput": 2343.98, "total_tokens": 70562120}
{"current_steps": 10565, "total_steps": 16324, "loss": 0.2223, "lr": 3.034182933853763e-05, "epoch": 1.294386083121497, "percentage": 64.72, "elapsed_time": "8:22:43", "remaining_time": "4:34:02", "throughput": 2340.42, "total_tokens": 70596224}
{"current_steps": 10570, "total_steps": 16324, "loss": 0.2498, "lr": 3.0295270207773464e-05, "epoch": 1.2949986217880003, "percentage": 64.75, "elapsed_time": "8:23:44", "remaining_time": "4:34:13", "throughput": 2336.83, "total_tokens": 70629656}
{"current_steps": 10575, "total_steps": 16324, "loss": 0.2135, "lr": 3.024873129577443e-05, "epoch": 1.2956111604545038, "percentage": 64.78, "elapsed_time": "8:24:45", "remaining_time": "4:34:24", "throughput": 2333.26, "total_tokens": 70663696}
{"current_steps": 10580, "total_steps": 16324, "loss": 0.2202, "lr": 3.0202212650293448e-05, "epoch": 1.296223699121007, "percentage": 64.81, "elapsed_time": "8:25:46", "remaining_time": "4:34:35", "throughput": 2329.7, "total_tokens": 70697176}
{"current_steps": 10585, "total_steps": 16324, "loss": 0.2005, "lr": 3.0155714319062734e-05, "epoch": 1.2968362377875104, "percentage": 64.84, "elapsed_time": "8:26:46", "remaining_time": "4:34:45", "throughput": 2326.2, "total_tokens": 70731408}
{"current_steps": 10590, "total_steps": 16324, "loss": 0.1899, "lr": 3.010923634979359e-05, "epoch": 1.2974487764540137, "percentage": 64.87, "elapsed_time": "8:27:46", "remaining_time": "4:34:56", "throughput": 2322.72, "total_tokens": 70765400}
{"current_steps": 10595, "total_steps": 16324, "loss": 0.2298, "lr": 3.0062778790176483e-05, "epoch": 1.298061315120517, "percentage": 64.9, "elapsed_time": "8:28:46", "remaining_time": "4:35:06", "throughput": 2319.23, "total_tokens": 70798848}
{"current_steps": 10600, "total_steps": 16324, "loss": 0.2834, "lr": 3.001634168788088e-05, "epoch": 1.2986738537870204, "percentage": 64.94, "elapsed_time": "8:29:46", "remaining_time": "4:35:16", "throughput": 2315.78, "total_tokens": 70832136}
{"current_steps": 10600, "total_steps": 16324, "eval_loss": 0.18608230352401733, "epoch": 1.2986738537870204, "percentage": 64.94, "elapsed_time": "8:30:06", "remaining_time": "4:35:27", "throughput": 2314.29, "total_tokens": 70832136}
{"current_steps": 10605, "total_steps": 16324, "loss": 0.218, "lr": 2.996992509055528e-05, "epoch": 1.2992863924535236, "percentage": 64.97, "elapsed_time": "8:31:11", "remaining_time": "4:35:40", "throughput": 2310.47, "total_tokens": 70866120}
{"current_steps": 10610, "total_steps": 16324, "loss": 0.215, "lr": 2.992352904582717e-05, "epoch": 1.299898931120027, "percentage": 65.0, "elapsed_time": "8:32:12", "remaining_time": "4:35:50", "throughput": 2307.01, "total_tokens": 70899936}
{"current_steps": 10615, "total_steps": 16324, "loss": 0.222, "lr": 2.9877153601302893e-05, "epoch": 1.3005114697865303, "percentage": 65.03, "elapsed_time": "8:33:12", "remaining_time": "4:36:00", "throughput": 2303.59, "total_tokens": 70933288}
{"current_steps": 10620, "total_steps": 16324, "loss": 0.2343, "lr": 2.9830798804567716e-05, "epoch": 1.3011240084530336, "percentage": 65.06, "elapsed_time": "8:34:12", "remaining_time": "4:36:10", "throughput": 2300.2, "total_tokens": 70967296}
{"current_steps": 10625, "total_steps": 16324, "loss": 0.2402, "lr": 2.9784464703185666e-05, "epoch": 1.3017365471195368, "percentage": 65.09, "elapsed_time": "8:35:13", "remaining_time": "4:36:21", "throughput": 2296.73, "total_tokens": 71000784}
{"current_steps": 10630, "total_steps": 16324, "loss": 0.2326, "lr": 2.973815134469958e-05, "epoch": 1.3023490857860403, "percentage": 65.12, "elapsed_time": "8:36:14", "remaining_time": "4:36:31", "throughput": 2293.35, "total_tokens": 71034616}
{"current_steps": 10635, "total_steps": 16324, "loss": 0.2387, "lr": 2.9691858776630965e-05, "epoch": 1.3029616244525435, "percentage": 65.15, "elapsed_time": "8:37:14", "remaining_time": "4:36:41", "throughput": 2289.95, "total_tokens": 71067488}
{"current_steps": 10640, "total_steps": 16324, "loss": 0.2606, "lr": 2.964558704648003e-05, "epoch": 1.303574163119047, "percentage": 65.18, "elapsed_time": "8:38:16", "remaining_time": "4:36:52", "throughput": 2286.46, "total_tokens": 71100848}
{"current_steps": 10645, "total_steps": 16324, "loss": 0.2349, "lr": 2.959933620172559e-05, "epoch": 1.3041867017855502, "percentage": 65.21, "elapsed_time": "8:39:18", "remaining_time": "4:37:02", "throughput": 2283.0, "total_tokens": 71134040}
{"current_steps": 10650, "total_steps": 16324, "loss": 0.2106, "lr": 2.9553106289825028e-05, "epoch": 1.3047992404520534, "percentage": 65.24, "elapsed_time": "8:40:18", "remaining_time": "4:37:12", "throughput": 2279.63, "total_tokens": 71167464}
{"current_steps": 10650, "total_steps": 16324, "eval_loss": 0.14994314312934875, "epoch": 1.3047992404520534, "percentage": 65.24, "elapsed_time": "8:40:38", "remaining_time": "4:37:22", "throughput": 2278.2, "total_tokens": 71167464}
{"current_steps": 10655, "total_steps": 16324, "loss": 0.2511, "lr": 2.950689735821428e-05, "epoch": 1.3054117791185569, "percentage": 65.27, "elapsed_time": "8:41:44", "remaining_time": "4:37:35", "throughput": 2274.45, "total_tokens": 71200776}
{"current_steps": 10660, "total_steps": 16324, "loss": 0.2303, "lr": 2.9460709454307668e-05, "epoch": 1.3060243177850601, "percentage": 65.3, "elapsed_time": "8:42:44", "remaining_time": "4:37:44", "throughput": 2271.19, "total_tokens": 71234168}
{"current_steps": 10665, "total_steps": 16324, "loss": 0.2423, "lr": 2.9414542625498032e-05, "epoch": 1.3066368564515636, "percentage": 65.33, "elapsed_time": "8:43:43", "remaining_time": "4:37:53", "throughput": 2267.98, "total_tokens": 71268008}
{"current_steps": 10670, "total_steps": 16324, "loss": 0.2461, "lr": 2.9368396919156543e-05, "epoch": 1.3072493951180668, "percentage": 65.36, "elapsed_time": "8:44:44", "remaining_time": "4:38:03", "throughput": 2264.67, "total_tokens": 71301256}
{"current_steps": 10675, "total_steps": 16324, "loss": 0.2585, "lr": 2.932227238263271e-05, "epoch": 1.30786193378457, "percentage": 65.39, "elapsed_time": "8:45:44", "remaining_time": "4:38:12", "throughput": 2261.41, "total_tokens": 71334304}
{"current_steps": 10680, "total_steps": 16324, "loss": 0.2558, "lr": 2.927616906325432e-05, "epoch": 1.3084744724510735, "percentage": 65.43, "elapsed_time": "8:46:43", "remaining_time": "4:38:21", "throughput": 2258.2, "total_tokens": 71367784}
{"current_steps": 10685, "total_steps": 16324, "loss": 0.2396, "lr": 2.9230087008327368e-05, "epoch": 1.3090870111175767, "percentage": 65.46, "elapsed_time": "8:47:44", "remaining_time": "4:38:31", "throughput": 2254.9, "total_tokens": 71400960}
{"current_steps": 10690, "total_steps": 16324, "loss": 0.2212, "lr": 2.9184026265136056e-05, "epoch": 1.3096995497840802, "percentage": 65.49, "elapsed_time": "8:48:44", "remaining_time": "4:38:40", "throughput": 2251.66, "total_tokens": 71433712}
{"current_steps": 10695, "total_steps": 16324, "loss": 0.2074, "lr": 2.9137986880942715e-05, "epoch": 1.3103120884505834, "percentage": 65.52, "elapsed_time": "8:49:45", "remaining_time": "4:38:49", "throughput": 2248.45, "total_tokens": 71467256}
{"current_steps": 10700, "total_steps": 16324, "loss": 0.206, "lr": 2.909196890298773e-05, "epoch": 1.3109246271170867, "percentage": 65.55, "elapsed_time": "8:50:46", "remaining_time": "4:38:58", "throughput": 2245.21, "total_tokens": 71501128}
{"current_steps": 10700, "total_steps": 16324, "eval_loss": 0.19016502797603607, "epoch": 1.3109246271170867, "percentage": 65.55, "elapsed_time": "8:51:05", "remaining_time": "4:39:08", "throughput": 2243.82, "total_tokens": 71501128}
{"current_steps": 10705, "total_steps": 16324, "loss": 0.2671, "lr": 2.9045972378489604e-05, "epoch": 1.3115371657835901, "percentage": 65.58, "elapsed_time": "8:52:11", "remaining_time": "4:39:20", "throughput": 2240.28, "total_tokens": 71534760}
{"current_steps": 10710, "total_steps": 16324, "loss": 0.2431, "lr": 2.8999997354644698e-05, "epoch": 1.3121497044500934, "percentage": 65.61, "elapsed_time": "8:53:11", "remaining_time": "4:39:29", "throughput": 2237.12, "total_tokens": 71568000}
{"current_steps": 10715, "total_steps": 16324, "loss": 0.2767, "lr": 2.8954043878627414e-05, "epoch": 1.3127622431165968, "percentage": 65.64, "elapsed_time": "8:54:11", "remaining_time": "4:39:37", "throughput": 2233.97, "total_tokens": 71601664}
{"current_steps": 10720, "total_steps": 16324, "loss": 0.2077, "lr": 2.890811199759003e-05, "epoch": 1.3133747817831, "percentage": 65.67, "elapsed_time": "8:55:12", "remaining_time": "4:39:47", "throughput": 2230.75, "total_tokens": 71635832}
{"current_steps": 10725, "total_steps": 16324, "loss": 0.2503, "lr": 2.886220175866261e-05, "epoch": 1.3139873204496033, "percentage": 65.7, "elapsed_time": "8:56:13", "remaining_time": "4:39:55", "throughput": 2227.63, "total_tokens": 71669720}
{"current_steps": 10730, "total_steps": 16324, "loss": 0.2579, "lr": 2.881631320895306e-05, "epoch": 1.3145998591161068, "percentage": 65.73, "elapsed_time": "8:57:14", "remaining_time": "4:40:05", "throughput": 2224.41, "total_tokens": 71703304}
{"current_steps": 10735, "total_steps": 16324, "loss": 0.2089, "lr": 2.8770446395547036e-05, "epoch": 1.31521239778261, "percentage": 65.76, "elapsed_time": "8:58:17", "remaining_time": "4:40:15", "throughput": 2221.14, "total_tokens": 71737192}
{"current_steps": 10740, "total_steps": 16324, "loss": 0.2317, "lr": 2.87246013655079e-05, "epoch": 1.3158249364491135, "percentage": 65.79, "elapsed_time": "8:59:19", "remaining_time": "4:40:24", "throughput": 2217.93, "total_tokens": 71770632}
{"current_steps": 10745, "total_steps": 16324, "loss": 0.2344, "lr": 2.8678778165876556e-05, "epoch": 1.3164374751156167, "percentage": 65.82, "elapsed_time": "9:00:20", "remaining_time": "4:40:33", "throughput": 2214.77, "total_tokens": 71804112}
{"current_steps": 10750, "total_steps": 16324, "loss": 0.184, "lr": 2.863297684367164e-05, "epoch": 1.31705001378212, "percentage": 65.85, "elapsed_time": "9:01:21", "remaining_time": "4:40:42", "throughput": 2211.63, "total_tokens": 71837520}
{"current_steps": 10750, "total_steps": 16324, "eval_loss": 0.2411387860774994, "epoch": 1.31705001378212, "percentage": 65.85, "elapsed_time": "9:01:41", "remaining_time": "4:40:52", "throughput": 2210.26, "total_tokens": 71837520}
{"current_steps": 10755, "total_steps": 16324, "loss": 0.1921, "lr": 2.858719744588928e-05, "epoch": 1.3176625524486232, "percentage": 65.88, "elapsed_time": "9:02:47", "remaining_time": "4:41:03", "throughput": 2206.81, "total_tokens": 71870856}
{"current_steps": 10760, "total_steps": 16324, "loss": 0.1655, "lr": 2.854144001950314e-05, "epoch": 1.3182750911151266, "percentage": 65.92, "elapsed_time": "9:03:48", "remaining_time": "4:41:12", "throughput": 2203.75, "total_tokens": 71904984}
{"current_steps": 10765, "total_steps": 16324, "loss": 0.1703, "lr": 2.8495704611464275e-05, "epoch": 1.31888762978163, "percentage": 65.95, "elapsed_time": "9:04:49", "remaining_time": "4:41:20", "throughput": 2200.68, "total_tokens": 71939480}
{"current_steps": 10770, "total_steps": 16324, "loss": 0.2253, "lr": 2.8449991268701208e-05, "epoch": 1.3195001684481333, "percentage": 65.98, "elapsed_time": "9:05:49", "remaining_time": "4:41:28", "throughput": 2197.67, "total_tokens": 71973360}
{"current_steps": 10775, "total_steps": 16324, "loss": 0.2047, "lr": 2.8404300038119825e-05, "epoch": 1.3201127071146366, "percentage": 66.01, "elapsed_time": "9:06:49", "remaining_time": "4:41:36", "throughput": 2194.73, "total_tokens": 72007440}
{"current_steps": 10780, "total_steps": 16324, "loss": 0.1839, "lr": 2.835863096660325e-05, "epoch": 1.3207252457811398, "percentage": 66.04, "elapsed_time": "9:07:50", "remaining_time": "4:41:44", "throughput": 2191.7, "total_tokens": 72041328}
{"current_steps": 10785, "total_steps": 16324, "loss": 0.2406, "lr": 2.8312984101011946e-05, "epoch": 1.3213377844476433, "percentage": 66.07, "elapsed_time": "9:08:50", "remaining_time": "4:41:52", "throughput": 2188.72, "total_tokens": 72074904}
{"current_steps": 10790, "total_steps": 16324, "loss": 0.2341, "lr": 2.826735948818358e-05, "epoch": 1.3219503231141465, "percentage": 66.1, "elapsed_time": "9:09:49", "remaining_time": "4:41:59", "throughput": 2185.82, "total_tokens": 72108648}
{"current_steps": 10795, "total_steps": 16324, "loss": 0.2294, "lr": 2.8221757174932938e-05, "epoch": 1.32256286178065, "percentage": 66.13, "elapsed_time": "9:10:49", "remaining_time": "4:42:07", "throughput": 2182.85, "total_tokens": 72141880}
{"current_steps": 10800, "total_steps": 16324, "loss": 0.2369, "lr": 2.817617720805197e-05, "epoch": 1.3231754004471532, "percentage": 66.16, "elapsed_time": "9:11:50", "remaining_time": "4:42:15", "throughput": 2179.85, "total_tokens": 72175680}
{"current_steps": 10800, "total_steps": 16324, "eval_loss": 0.17310434579849243, "epoch": 1.3231754004471532, "percentage": 66.16, "elapsed_time": "9:12:10", "remaining_time": "4:42:25", "throughput": 2178.55, "total_tokens": 72175680}
{"current_steps": 10805, "total_steps": 16324, "loss": 0.2378, "lr": 2.8130619634309676e-05, "epoch": 1.3237879391136564, "percentage": 66.19, "elapsed_time": "9:13:15", "remaining_time": "4:42:35", "throughput": 2175.25, "total_tokens": 72209280}
{"current_steps": 10810, "total_steps": 16324, "loss": 0.1945, "lr": 2.8085084500452113e-05, "epoch": 1.32440047778016, "percentage": 66.22, "elapsed_time": "9:14:16", "remaining_time": "4:42:43", "throughput": 2172.3, "total_tokens": 72243368}
{"current_steps": 10815, "total_steps": 16324, "loss": 0.2723, "lr": 2.803957185320225e-05, "epoch": 1.3250130164466631, "percentage": 66.25, "elapsed_time": "9:15:16", "remaining_time": "4:42:51", "throughput": 2169.37, "total_tokens": 72276136}
{"current_steps": 10820, "total_steps": 16324, "loss": 0.2144, "lr": 2.7994081739260048e-05, "epoch": 1.3256255551131666, "percentage": 66.28, "elapsed_time": "9:16:16", "remaining_time": "4:42:58", "throughput": 2166.5, "total_tokens": 72310256}
{"current_steps": 10825, "total_steps": 16324, "loss": 0.27, "lr": 2.7948614205302288e-05, "epoch": 1.3262380937796698, "percentage": 66.31, "elapsed_time": "9:17:17", "remaining_time": "4:43:05", "throughput": 2163.58, "total_tokens": 72343800}
{"current_steps": 10830, "total_steps": 16324, "loss": 0.2218, "lr": 2.790316929798264e-05, "epoch": 1.326850632446173, "percentage": 66.34, "elapsed_time": "9:18:18", "remaining_time": "4:43:13", "throughput": 2160.62, "total_tokens": 72377088}
{"current_steps": 10835, "total_steps": 16324, "loss": 0.2287, "lr": 2.7857747063931493e-05, "epoch": 1.3274631711126765, "percentage": 66.37, "elapsed_time": "9:19:18", "remaining_time": "4:43:20", "throughput": 2157.71, "total_tokens": 72410240}
{"current_steps": 10840, "total_steps": 16324, "loss": 0.2552, "lr": 2.7812347549756035e-05, "epoch": 1.3280757097791798, "percentage": 66.41, "elapsed_time": "9:20:19", "remaining_time": "4:43:28", "throughput": 2154.84, "total_tokens": 72443824}
{"current_steps": 10845, "total_steps": 16324, "loss": 0.2357, "lr": 2.7766970802040114e-05, "epoch": 1.3286882484456832, "percentage": 66.44, "elapsed_time": "9:21:20", "remaining_time": "4:43:35", "throughput": 2151.89, "total_tokens": 72477288}
{"current_steps": 10850, "total_steps": 16324, "loss": 0.2175, "lr": 2.772161686734419e-05, "epoch": 1.3293007871121865, "percentage": 66.47, "elapsed_time": "9:22:21", "remaining_time": "4:43:42", "throughput": 2149.05, "total_tokens": 72511424}
{"current_steps": 10850, "total_steps": 16324, "eval_loss": 0.1683052033185959, "epoch": 1.3293007871121865, "percentage": 66.47, "elapsed_time": "9:22:41", "remaining_time": "4:43:53", "throughput": 2147.78, "total_tokens": 72511424}
{"current_steps": 10855, "total_steps": 16324, "loss": 0.2181, "lr": 2.767628579220536e-05, "epoch": 1.3299133257786897, "percentage": 66.5, "elapsed_time": "9:23:47", "remaining_time": "4:44:02", "throughput": 2144.59, "total_tokens": 72545080}
{"current_steps": 10860, "total_steps": 16324, "loss": 0.2648, "lr": 2.7630977623137244e-05, "epoch": 1.3305258644451932, "percentage": 66.53, "elapsed_time": "9:24:48", "remaining_time": "4:44:10", "throughput": 2141.67, "total_tokens": 72578272}
{"current_steps": 10865, "total_steps": 16324, "loss": 0.1966, "lr": 2.758569240662998e-05, "epoch": 1.3311384031116964, "percentage": 66.56, "elapsed_time": "9:25:49", "remaining_time": "4:44:17", "throughput": 2138.86, "total_tokens": 72612536}
{"current_steps": 10870, "total_steps": 16324, "loss": 0.247, "lr": 2.7540430189150098e-05, "epoch": 1.3317509417781999, "percentage": 66.59, "elapsed_time": "9:26:50", "remaining_time": "4:44:24", "throughput": 2135.98, "total_tokens": 72645920}
{"current_steps": 10875, "total_steps": 16324, "loss": 0.1914, "lr": 2.7495191017140582e-05, "epoch": 1.332363480444703, "percentage": 66.62, "elapsed_time": "9:27:51", "remaining_time": "4:44:31", "throughput": 2133.15, "total_tokens": 72680416}
{"current_steps": 10880, "total_steps": 16324, "loss": 0.2471, "lr": 2.7449974937020778e-05, "epoch": 1.3329760191112063, "percentage": 66.65, "elapsed_time": "9:28:51", "remaining_time": "4:44:38", "throughput": 2130.4, "total_tokens": 72714560}
{"current_steps": 10885, "total_steps": 16324, "loss": 0.231, "lr": 2.7404781995186267e-05, "epoch": 1.3335885577777098, "percentage": 66.68, "elapsed_time": "9:29:52", "remaining_time": "4:44:45", "throughput": 2127.64, "total_tokens": 72748336}
{"current_steps": 10890, "total_steps": 16324, "loss": 0.2507, "lr": 2.7359612238008968e-05, "epoch": 1.334201096444213, "percentage": 66.71, "elapsed_time": "9:30:53", "remaining_time": "4:44:52", "throughput": 2124.82, "total_tokens": 72782040}
{"current_steps": 10895, "total_steps": 16324, "loss": 0.2298, "lr": 2.7314465711836967e-05, "epoch": 1.3348136351107165, "percentage": 66.74, "elapsed_time": "9:31:52", "remaining_time": "4:44:58", "throughput": 2122.12, "total_tokens": 72815992}
{"current_steps": 10900, "total_steps": 16324, "loss": 0.197, "lr": 2.7269342462994563e-05, "epoch": 1.3354261737772197, "percentage": 66.77, "elapsed_time": "9:32:52", "remaining_time": "4:45:04", "throughput": 2119.47, "total_tokens": 72850512}
{"current_steps": 10900, "total_steps": 16324, "eval_loss": 0.22737418115139008, "epoch": 1.3354261737772197, "percentage": 66.77, "elapsed_time": "9:33:11", "remaining_time": "4:45:13", "throughput": 2118.26, "total_tokens": 72850512}
{"current_steps": 10905, "total_steps": 16324, "loss": 0.1994, "lr": 2.7224242537782056e-05, "epoch": 1.336038712443723, "percentage": 66.8, "elapsed_time": "9:34:17", "remaining_time": "4:45:22", "throughput": 2115.19, "total_tokens": 72884432}
{"current_steps": 10910, "total_steps": 16324, "loss": 0.2839, "lr": 2.7179165982475924e-05, "epoch": 1.3366512511102262, "percentage": 66.83, "elapsed_time": "9:35:17", "remaining_time": "4:45:28", "throughput": 2112.49, "total_tokens": 72917608}
{"current_steps": 10915, "total_steps": 16324, "loss": 0.186, "lr": 2.713411284332863e-05, "epoch": 1.3372637897767297, "percentage": 66.86, "elapsed_time": "9:36:17", "remaining_time": "4:45:34", "throughput": 2109.84, "total_tokens": 72952000}
{"current_steps": 10920, "total_steps": 16324, "loss": 0.2018, "lr": 2.708908316656863e-05, "epoch": 1.337876328443233, "percentage": 66.9, "elapsed_time": "9:37:18", "remaining_time": "4:45:41", "throughput": 2107.1, "total_tokens": 72985840}
{"current_steps": 10925, "total_steps": 16324, "loss": 0.2494, "lr": 2.7044076998400247e-05, "epoch": 1.3384888671097364, "percentage": 66.93, "elapsed_time": "9:38:18", "remaining_time": "4:45:47", "throughput": 2104.41, "total_tokens": 73019136}
{"current_steps": 10930, "total_steps": 16324, "loss": 0.199, "lr": 2.6999094385003743e-05, "epoch": 1.3391014057762396, "percentage": 66.96, "elapsed_time": "9:39:17", "remaining_time": "4:45:53", "throughput": 2101.77, "total_tokens": 73053088}
{"current_steps": 10935, "total_steps": 16324, "loss": 0.288, "lr": 2.69541353725352e-05, "epoch": 1.3397139444427428, "percentage": 66.99, "elapsed_time": "9:40:18", "remaining_time": "4:45:59", "throughput": 2099.08, "total_tokens": 73086296}
{"current_steps": 10940, "total_steps": 16324, "loss": 0.1766, "lr": 2.690920000712644e-05, "epoch": 1.3403264831092463, "percentage": 67.02, "elapsed_time": "9:41:18", "remaining_time": "4:46:05", "throughput": 2096.43, "total_tokens": 73120432}
{"current_steps": 10945, "total_steps": 16324, "loss": 0.2227, "lr": 2.6864288334885067e-05, "epoch": 1.3409390217757495, "percentage": 67.05, "elapsed_time": "9:42:18", "remaining_time": "4:46:10", "throughput": 2093.8, "total_tokens": 73153912}
{"current_steps": 10950, "total_steps": 16324, "loss": 0.2192, "lr": 2.6819400401894385e-05, "epoch": 1.341551560442253, "percentage": 67.08, "elapsed_time": "9:43:18", "remaining_time": "4:46:16", "throughput": 2091.13, "total_tokens": 73187184}
{"current_steps": 10950, "total_steps": 16324, "eval_loss": 0.21316958963871002, "epoch": 1.341551560442253, "percentage": 67.08, "elapsed_time": "9:43:38", "remaining_time": "4:46:26", "throughput": 2089.94, "total_tokens": 73187184}
{"current_steps": 10955, "total_steps": 16324, "loss": 0.2332, "lr": 2.6774536254213274e-05, "epoch": 1.3421640991087562, "percentage": 67.11, "elapsed_time": "9:44:45", "remaining_time": "4:46:35", "throughput": 2086.92, "total_tokens": 73220296}
{"current_steps": 10960, "total_steps": 16324, "loss": 0.1974, "lr": 2.672969593787626e-05, "epoch": 1.3427766377752595, "percentage": 67.14, "elapsed_time": "9:45:44", "remaining_time": "4:46:40", "throughput": 2084.35, "total_tokens": 73254424}
{"current_steps": 10965, "total_steps": 16324, "loss": 0.2102, "lr": 2.6684879498893412e-05, "epoch": 1.343389176441763, "percentage": 67.17, "elapsed_time": "9:46:45", "remaining_time": "4:46:46", "throughput": 2081.71, "total_tokens": 73287672}
{"current_steps": 10970, "total_steps": 16324, "loss": 0.2269, "lr": 2.6640086983250313e-05, "epoch": 1.3440017151082662, "percentage": 67.2, "elapsed_time": "9:47:46", "remaining_time": "4:46:51", "throughput": 2079.09, "total_tokens": 73321264}
{"current_steps": 10975, "total_steps": 16324, "loss": 0.2103, "lr": 2.6595318436907924e-05, "epoch": 1.3446142537747696, "percentage": 67.23, "elapsed_time": "9:48:46", "remaining_time": "4:46:57", "throughput": 2076.52, "total_tokens": 73355680}
{"current_steps": 10980, "total_steps": 16324, "loss": 0.2326, "lr": 2.6550573905802718e-05, "epoch": 1.3452267924412729, "percentage": 67.26, "elapsed_time": "9:49:47", "remaining_time": "4:47:03", "throughput": 2073.9, "total_tokens": 73389408}
{"current_steps": 10985, "total_steps": 16324, "loss": 0.2272, "lr": 2.6505853435846422e-05, "epoch": 1.345839331107776, "percentage": 67.29, "elapsed_time": "9:50:48", "remaining_time": "4:47:08", "throughput": 2071.24, "total_tokens": 73422048}
{"current_steps": 10990, "total_steps": 16324, "loss": 0.2807, "lr": 2.6461157072926175e-05, "epoch": 1.3464518697742796, "percentage": 67.32, "elapsed_time": "9:51:48", "remaining_time": "4:47:13", "throughput": 2068.68, "total_tokens": 73455432}
{"current_steps": 10995, "total_steps": 16324, "loss": 0.2443, "lr": 2.6416484862904278e-05, "epoch": 1.3470644084407828, "percentage": 67.35, "elapsed_time": "9:52:48", "remaining_time": "4:47:19", "throughput": 2066.1, "total_tokens": 73488176}
{"current_steps": 11000, "total_steps": 16324, "loss": 0.2141, "lr": 2.6371836851618325e-05, "epoch": 1.3476769471072862, "percentage": 67.39, "elapsed_time": "9:53:49", "remaining_time": "4:47:24", "throughput": 2063.52, "total_tokens": 73522280}
{"current_steps": 11000, "total_steps": 16324, "eval_loss": 0.18228411674499512, "epoch": 1.3476769471072862, "percentage": 67.39, "elapsed_time": "9:54:09", "remaining_time": "4:47:34", "throughput": 2062.37, "total_tokens": 73522280}
{"current_steps": 11005, "total_steps": 16324, "loss": 0.2179, "lr": 2.6327213084881087e-05, "epoch": 1.3482894857737895, "percentage": 67.42, "elapsed_time": "9:55:15", "remaining_time": "4:47:42", "throughput": 2059.49, "total_tokens": 73555864}
{"current_steps": 11010, "total_steps": 16324, "loss": 0.219, "lr": 2.6282613608480365e-05, "epoch": 1.3489020244402927, "percentage": 67.45, "elapsed_time": "9:56:16", "remaining_time": "4:47:47", "throughput": 2056.92, "total_tokens": 73589336}
{"current_steps": 11015, "total_steps": 16324, "loss": 0.2664, "lr": 2.623803846817913e-05, "epoch": 1.3495145631067962, "percentage": 67.48, "elapsed_time": "9:57:17", "remaining_time": "4:47:52", "throughput": 2054.36, "total_tokens": 73622416}
{"current_steps": 11020, "total_steps": 16324, "loss": 0.1887, "lr": 2.6193487709715347e-05, "epoch": 1.3501271017732994, "percentage": 67.51, "elapsed_time": "9:58:16", "remaining_time": "4:47:57", "throughput": 2051.88, "total_tokens": 73656424}
{"current_steps": 11025, "total_steps": 16324, "loss": 0.1939, "lr": 2.6148961378801994e-05, "epoch": 1.3507396404398029, "percentage": 67.54, "elapsed_time": "9:59:17", "remaining_time": "4:48:02", "throughput": 2049.4, "total_tokens": 73690536}
{"current_steps": 11030, "total_steps": 16324, "loss": 0.2455, "lr": 2.6104459521126902e-05, "epoch": 1.3513521791063061, "percentage": 67.57, "elapsed_time": "10:00:19", "remaining_time": "4:48:07", "throughput": 2046.8, "total_tokens": 73724000}
{"current_steps": 11035, "total_steps": 16324, "loss": 0.2239, "lr": 2.605998218235287e-05, "epoch": 1.3519647177728094, "percentage": 67.6, "elapsed_time": "10:01:18", "remaining_time": "4:48:12", "throughput": 2044.35, "total_tokens": 73757704}
{"current_steps": 11040, "total_steps": 16324, "loss": 0.1928, "lr": 2.6015529408117534e-05, "epoch": 1.3525772564393128, "percentage": 67.63, "elapsed_time": "10:02:18", "remaining_time": "4:48:16", "throughput": 2041.9, "total_tokens": 73790936}
{"current_steps": 11045, "total_steps": 16324, "loss": 0.2083, "lr": 2.597110124403327e-05, "epoch": 1.353189795105816, "percentage": 67.66, "elapsed_time": "10:03:19", "remaining_time": "4:48:21", "throughput": 2039.38, "total_tokens": 73824584}
{"current_steps": 11050, "total_steps": 16324, "loss": 0.2257, "lr": 2.592669773568724e-05, "epoch": 1.3538023337723195, "percentage": 67.69, "elapsed_time": "10:04:19", "remaining_time": "4:48:25", "throughput": 2036.95, "total_tokens": 73858104}
{"current_steps": 11050, "total_steps": 16324, "eval_loss": 0.16352061927318573, "epoch": 1.3538023337723195, "percentage": 67.69, "elapsed_time": "10:04:38", "remaining_time": "4:48:35", "throughput": 2035.85, "total_tokens": 73858104}
{"current_steps": 11055, "total_steps": 16324, "loss": 0.2284, "lr": 2.5882318928641313e-05, "epoch": 1.3544148724388227, "percentage": 67.72, "elapsed_time": "10:05:43", "remaining_time": "4:48:42", "throughput": 2033.11, "total_tokens": 73891168}
{"current_steps": 11060, "total_steps": 16324, "loss": 0.2511, "lr": 2.583796486843202e-05, "epoch": 1.355027411105326, "percentage": 67.75, "elapsed_time": "10:06:44", "remaining_time": "4:48:46", "throughput": 2030.63, "total_tokens": 73923784}
{"current_steps": 11065, "total_steps": 16324, "loss": 0.2252, "lr": 2.5793635600570455e-05, "epoch": 1.3556399497718292, "percentage": 67.78, "elapsed_time": "10:07:44", "remaining_time": "4:48:50", "throughput": 2028.21, "total_tokens": 73956896}
{"current_steps": 11070, "total_steps": 16324, "loss": 0.2297, "lr": 2.574933117054228e-05, "epoch": 1.3562524884383327, "percentage": 67.81, "elapsed_time": "10:08:43", "remaining_time": "4:48:54", "throughput": 2025.84, "total_tokens": 73990912}
{"current_steps": 11075, "total_steps": 16324, "loss": 0.2242, "lr": 2.5705051623807726e-05, "epoch": 1.356865027104836, "percentage": 67.84, "elapsed_time": "10:09:43", "remaining_time": "4:48:58", "throughput": 2023.43, "total_tokens": 74024656}
{"current_steps": 11080, "total_steps": 16324, "loss": 0.21, "lr": 2.5660797005801452e-05, "epoch": 1.3574775657713394, "percentage": 67.88, "elapsed_time": "10:10:44", "remaining_time": "4:49:03", "throughput": 2021.01, "total_tokens": 74058600}
{"current_steps": 11085, "total_steps": 16324, "loss": 0.2284, "lr": 2.561656736193252e-05, "epoch": 1.3580901044378426, "percentage": 67.91, "elapsed_time": "10:11:43", "remaining_time": "4:49:07", "throughput": 2018.64, "total_tokens": 74091880}
{"current_steps": 11090, "total_steps": 16324, "loss": 0.2295, "lr": 2.5572362737584404e-05, "epoch": 1.3587026431043459, "percentage": 67.94, "elapsed_time": "10:12:44", "remaining_time": "4:49:11", "throughput": 2016.22, "total_tokens": 74125064}
{"current_steps": 11095, "total_steps": 16324, "loss": 0.1969, "lr": 2.5528183178114906e-05, "epoch": 1.3593151817708493, "percentage": 67.97, "elapsed_time": "10:13:44", "remaining_time": "4:49:15", "throughput": 2013.82, "total_tokens": 74158416}
{"current_steps": 11100, "total_steps": 16324, "loss": 0.2696, "lr": 2.5484028728856058e-05, "epoch": 1.3599277204373525, "percentage": 68.0, "elapsed_time": "10:14:44", "remaining_time": "4:49:18", "throughput": 2011.47, "total_tokens": 74191944}
{"current_steps": 11100, "total_steps": 16324, "eval_loss": 0.13369590044021606, "epoch": 1.3599277204373525, "percentage": 68.0, "elapsed_time": "10:15:03", "remaining_time": "4:49:28", "throughput": 2010.4, "total_tokens": 74191944}
{"current_steps": 11105, "total_steps": 16324, "loss": 0.2247, "lr": 2.5439899435114168e-05, "epoch": 1.360540259103856, "percentage": 68.03, "elapsed_time": "10:16:09", "remaining_time": "4:49:34", "throughput": 2007.73, "total_tokens": 74225480}
{"current_steps": 11110, "total_steps": 16324, "loss": 0.2181, "lr": 2.5395795342169748e-05, "epoch": 1.3611527977703592, "percentage": 68.06, "elapsed_time": "10:17:10", "remaining_time": "4:49:38", "throughput": 2005.36, "total_tokens": 74259160}
{"current_steps": 11115, "total_steps": 16324, "loss": 0.2141, "lr": 2.5351716495277434e-05, "epoch": 1.3617653364368625, "percentage": 68.09, "elapsed_time": "10:18:09", "remaining_time": "4:49:41", "throughput": 2003.06, "total_tokens": 74292656}
{"current_steps": 11120, "total_steps": 16324, "loss": 0.2328, "lr": 2.530766293966592e-05, "epoch": 1.362377875103366, "percentage": 68.12, "elapsed_time": "10:19:08", "remaining_time": "4:49:45", "throughput": 2000.77, "total_tokens": 74325968}
{"current_steps": 11125, "total_steps": 16324, "loss": 0.1933, "lr": 2.5263634720538005e-05, "epoch": 1.3629904137698692, "percentage": 68.15, "elapsed_time": "10:20:08", "remaining_time": "4:49:48", "throughput": 1998.46, "total_tokens": 74360072}
{"current_steps": 11130, "total_steps": 16324, "loss": 0.2065, "lr": 2.5219631883070482e-05, "epoch": 1.3636029524363726, "percentage": 68.18, "elapsed_time": "10:21:07", "remaining_time": "4:49:51", "throughput": 1996.19, "total_tokens": 74393752}
{"current_steps": 11135, "total_steps": 16324, "loss": 0.2419, "lr": 2.5175654472414047e-05, "epoch": 1.3642154911028759, "percentage": 68.21, "elapsed_time": "10:22:07", "remaining_time": "4:49:54", "throughput": 1993.92, "total_tokens": 74427864}
{"current_steps": 11140, "total_steps": 16324, "loss": 0.2281, "lr": 2.513170253369337e-05, "epoch": 1.3648280297693791, "percentage": 68.24, "elapsed_time": "10:23:07", "remaining_time": "4:49:58", "throughput": 1991.6, "total_tokens": 74461312}
{"current_steps": 11145, "total_steps": 16324, "loss": 0.2062, "lr": 2.5087776112006966e-05, "epoch": 1.3654405684358826, "percentage": 68.27, "elapsed_time": "10:24:06", "remaining_time": "4:50:01", "throughput": 1989.36, "total_tokens": 74494648}
{"current_steps": 11150, "total_steps": 16324, "loss": 0.236, "lr": 2.5043875252427156e-05, "epoch": 1.3660531071023858, "percentage": 68.3, "elapsed_time": "10:25:05", "remaining_time": "4:50:03", "throughput": 1987.13, "total_tokens": 74528296}
{"current_steps": 11150, "total_steps": 16324, "eval_loss": 0.14952169358730316, "epoch": 1.3660531071023858, "percentage": 68.3, "elapsed_time": "10:25:24", "remaining_time": "4:50:12", "throughput": 1986.1, "total_tokens": 74528296}
{"current_steps": 11155, "total_steps": 16324, "loss": 0.1971, "lr": 2.500000000000001e-05, "epoch": 1.3666656457688893, "percentage": 68.33, "elapsed_time": "10:26:30", "remaining_time": "4:50:18", "throughput": 1983.55, "total_tokens": 74562368}
{"current_steps": 11160, "total_steps": 16324, "loss": 0.2395, "lr": 2.4956150399745375e-05, "epoch": 1.3672781844353925, "percentage": 68.37, "elapsed_time": "10:27:29", "remaining_time": "4:50:21", "throughput": 1981.34, "total_tokens": 74596328}
{"current_steps": 11165, "total_steps": 16324, "loss": 0.2587, "lr": 2.4912326496656762e-05, "epoch": 1.3678907231018957, "percentage": 68.4, "elapsed_time": "10:28:29", "remaining_time": "4:50:24", "throughput": 1979.07, "total_tokens": 74629408}
{"current_steps": 11170, "total_steps": 16324, "loss": 0.2276, "lr": 2.4868528335701275e-05, "epoch": 1.3685032617683992, "percentage": 68.43, "elapsed_time": "10:29:30", "remaining_time": "4:50:27", "throughput": 1976.76, "total_tokens": 74662920}
{"current_steps": 11175, "total_steps": 16324, "loss": 0.2238, "lr": 2.4824755961819647e-05, "epoch": 1.3691158004349024, "percentage": 68.46, "elapsed_time": "10:30:30", "remaining_time": "4:50:30", "throughput": 1974.53, "total_tokens": 74696728}
{"current_steps": 11180, "total_steps": 16324, "loss": 0.2288, "lr": 2.478100941992613e-05, "epoch": 1.369728339101406, "percentage": 68.49, "elapsed_time": "10:31:29", "remaining_time": "4:50:33", "throughput": 1972.3, "total_tokens": 74730448}
{"current_steps": 11185, "total_steps": 16324, "loss": 0.2066, "lr": 2.4737288754908517e-05, "epoch": 1.3703408777679091, "percentage": 68.52, "elapsed_time": "10:32:31", "remaining_time": "4:50:36", "throughput": 1970.01, "total_tokens": 74764088}
{"current_steps": 11190, "total_steps": 16324, "loss": 0.2158, "lr": 2.469359401162795e-05, "epoch": 1.3709534164344124, "percentage": 68.55, "elapsed_time": "10:33:30", "remaining_time": "4:50:39", "throughput": 1967.79, "total_tokens": 74797488}
{"current_steps": 11195, "total_steps": 16324, "loss": 0.1934, "lr": 2.4649925234919063e-05, "epoch": 1.3715659551009156, "percentage": 68.58, "elapsed_time": "10:34:30", "remaining_time": "4:50:42", "throughput": 1965.57, "total_tokens": 74830920}
{"current_steps": 11200, "total_steps": 16324, "loss": 0.1875, "lr": 2.4606282469589837e-05, "epoch": 1.372178493767419, "percentage": 68.61, "elapsed_time": "10:35:31", "remaining_time": "4:50:45", "throughput": 1963.34, "total_tokens": 74864760}
{"current_steps": 11200, "total_steps": 16324, "eval_loss": 0.15537530183792114, "epoch": 1.372178493767419, "percentage": 68.61, "elapsed_time": "10:35:51", "remaining_time": "4:50:54", "throughput": 1962.33, "total_tokens": 74864760}
{"current_steps": 11205, "total_steps": 16324, "loss": 0.2228, "lr": 2.4562665760421493e-05, "epoch": 1.3727910324339223, "percentage": 68.64, "elapsed_time": "10:36:57", "remaining_time": "4:50:59", "throughput": 1959.78, "total_tokens": 74898192}
{"current_steps": 11210, "total_steps": 16324, "loss": 0.2745, "lr": 2.4519075152168597e-05, "epoch": 1.3734035711004258, "percentage": 68.67, "elapsed_time": "10:37:57", "remaining_time": "4:51:02", "throughput": 1957.57, "total_tokens": 74930960}
{"current_steps": 11215, "total_steps": 16324, "loss": 0.204, "lr": 2.447551068955889e-05, "epoch": 1.374016109766929, "percentage": 68.7, "elapsed_time": "10:38:58", "remaining_time": "4:51:04", "throughput": 1955.35, "total_tokens": 74964680}
{"current_steps": 11220, "total_steps": 16324, "loss": 0.2171, "lr": 2.4431972417293325e-05, "epoch": 1.3746286484334322, "percentage": 68.73, "elapsed_time": "10:39:58", "remaining_time": "4:51:07", "throughput": 1953.15, "total_tokens": 74998424}
{"current_steps": 11225, "total_steps": 16324, "loss": 0.2694, "lr": 2.4388460380045925e-05, "epoch": 1.3752411870999357, "percentage": 68.76, "elapsed_time": "10:40:57", "remaining_time": "4:51:09", "throughput": 1951.02, "total_tokens": 75031816}
{"current_steps": 11230, "total_steps": 16324, "loss": 0.2386, "lr": 2.4344974622463805e-05, "epoch": 1.375853725766439, "percentage": 68.79, "elapsed_time": "10:41:57", "remaining_time": "4:51:11", "throughput": 1948.88, "total_tokens": 75065480}
{"current_steps": 11235, "total_steps": 16324, "loss": 0.215, "lr": 2.4301515189167152e-05, "epoch": 1.3764662644329424, "percentage": 68.83, "elapsed_time": "10:42:56", "remaining_time": "4:51:13", "throughput": 1946.73, "total_tokens": 75098816}
{"current_steps": 11240, "total_steps": 16324, "loss": 0.1899, "lr": 2.4258082124749137e-05, "epoch": 1.3770788030994456, "percentage": 68.86, "elapsed_time": "10:43:55", "remaining_time": "4:51:15", "throughput": 1944.62, "total_tokens": 75131904}
{"current_steps": 11245, "total_steps": 16324, "loss": 0.2442, "lr": 2.4214675473775826e-05, "epoch": 1.3776913417659489, "percentage": 68.89, "elapsed_time": "10:44:54", "remaining_time": "4:51:17", "throughput": 1942.5, "total_tokens": 75164672}
{"current_steps": 11250, "total_steps": 16324, "loss": 0.2163, "lr": 2.417129528078623e-05, "epoch": 1.3783038804324523, "percentage": 68.92, "elapsed_time": "10:45:55", "remaining_time": "4:51:19", "throughput": 1940.37, "total_tokens": 75199264}
{"current_steps": 11250, "total_steps": 16324, "eval_loss": 0.14667704701423645, "epoch": 1.3783038804324523, "percentage": 68.92, "elapsed_time": "10:46:15", "remaining_time": "4:51:28", "throughput": 1939.35, "total_tokens": 75199264}
{"current_steps": 11255, "total_steps": 16324, "loss": 0.2058, "lr": 2.4127941590292225e-05, "epoch": 1.3789164190989556, "percentage": 68.95, "elapsed_time": "10:47:23", "remaining_time": "4:51:34", "throughput": 1936.8, "total_tokens": 75232800}
{"current_steps": 11260, "total_steps": 16324, "loss": 0.2428, "lr": 2.408461444677841e-05, "epoch": 1.379528957765459, "percentage": 68.98, "elapsed_time": "10:48:24", "remaining_time": "4:51:36", "throughput": 1934.64, "total_tokens": 75266240}
{"current_steps": 11265, "total_steps": 16324, "loss": 0.1717, "lr": 2.404131389470223e-05, "epoch": 1.3801414964319623, "percentage": 69.01, "elapsed_time": "10:49:27", "remaining_time": "4:51:39", "throughput": 1932.39, "total_tokens": 75300144}
{"current_steps": 11270, "total_steps": 16324, "loss": 0.2426, "lr": 2.399803997849381e-05, "epoch": 1.3807540350984655, "percentage": 69.04, "elapsed_time": "10:50:29", "remaining_time": "4:51:42", "throughput": 1930.16, "total_tokens": 75333520}
{"current_steps": 11275, "total_steps": 16324, "loss": 0.2372, "lr": 2.3954792742555964e-05, "epoch": 1.381366573764969, "percentage": 69.07, "elapsed_time": "10:51:31", "remaining_time": "4:51:45", "throughput": 1927.98, "total_tokens": 75367128}
{"current_steps": 11280, "total_steps": 16324, "loss": 0.1993, "lr": 2.391157223126409e-05, "epoch": 1.3819791124314722, "percentage": 69.1, "elapsed_time": "10:52:33", "remaining_time": "4:51:47", "throughput": 1925.8, "total_tokens": 75400848}
{"current_steps": 11285, "total_steps": 16324, "loss": 0.2131, "lr": 2.386837848896619e-05, "epoch": 1.3825916510979757, "percentage": 69.13, "elapsed_time": "10:53:33", "remaining_time": "4:51:49", "throughput": 1923.69, "total_tokens": 75433880}
{"current_steps": 11290, "total_steps": 16324, "loss": 0.1944, "lr": 2.382521155998283e-05, "epoch": 1.383204189764479, "percentage": 69.16, "elapsed_time": "10:54:32", "remaining_time": "4:51:51", "throughput": 1921.62, "total_tokens": 75467816}
{"current_steps": 11295, "total_steps": 16324, "loss": 0.2091, "lr": 2.378207148860699e-05, "epoch": 1.3838167284309821, "percentage": 69.19, "elapsed_time": "10:55:34", "remaining_time": "4:51:53", "throughput": 1919.49, "total_tokens": 75501792}
{"current_steps": 11300, "total_steps": 16324, "loss": 0.1914, "lr": 2.3738958319104145e-05, "epoch": 1.3844292670974856, "percentage": 69.22, "elapsed_time": "10:56:33", "remaining_time": "4:51:54", "throughput": 1917.44, "total_tokens": 75535760}
{"current_steps": 11300, "total_steps": 16324, "eval_loss": 0.12553976476192474, "epoch": 1.3844292670974856, "percentage": 69.22, "elapsed_time": "10:56:53", "remaining_time": "4:52:03", "throughput": 1916.48, "total_tokens": 75535760}
{"current_steps": 11305, "total_steps": 16324, "loss": 0.236, "lr": 2.369587209571217e-05, "epoch": 1.3850418057639888, "percentage": 69.25, "elapsed_time": "10:57:59", "remaining_time": "4:52:07", "throughput": 1914.13, "total_tokens": 75568840}
{"current_steps": 11310, "total_steps": 16324, "loss": 0.261, "lr": 2.3652812862641265e-05, "epoch": 1.3856543444304923, "percentage": 69.28, "elapsed_time": "10:59:00", "remaining_time": "4:52:09", "throughput": 1911.99, "total_tokens": 75601944}
{"current_steps": 11315, "total_steps": 16324, "loss": 0.2013, "lr": 2.3609780664073923e-05, "epoch": 1.3862668830969955, "percentage": 69.32, "elapsed_time": "11:00:00", "remaining_time": "4:52:10", "throughput": 1909.95, "total_tokens": 75635768}
{"current_steps": 11320, "total_steps": 16324, "loss": 0.218, "lr": 2.3566775544164937e-05, "epoch": 1.3868794217634988, "percentage": 69.35, "elapsed_time": "11:01:00", "remaining_time": "4:52:12", "throughput": 1907.91, "total_tokens": 75669384}
{"current_steps": 11325, "total_steps": 16324, "loss": 0.1893, "lr": 2.3523797547041316e-05, "epoch": 1.3874919604300022, "percentage": 69.38, "elapsed_time": "11:02:01", "remaining_time": "4:52:13", "throughput": 1905.86, "total_tokens": 75703016}
{"current_steps": 11330, "total_steps": 16324, "loss": 0.2366, "lr": 2.3480846716802184e-05, "epoch": 1.3881044990965055, "percentage": 69.41, "elapsed_time": "11:03:01", "remaining_time": "4:52:14", "throughput": 1903.81, "total_tokens": 75736336}
{"current_steps": 11335, "total_steps": 16324, "loss": 0.1948, "lr": 2.3437923097518843e-05, "epoch": 1.388717037763009, "percentage": 69.44, "elapsed_time": "11:04:00", "remaining_time": "4:52:15", "throughput": 1901.82, "total_tokens": 75770120}
{"current_steps": 11340, "total_steps": 16324, "loss": 0.2454, "lr": 2.3395026733234653e-05, "epoch": 1.3893295764295122, "percentage": 69.47, "elapsed_time": "11:05:00", "remaining_time": "4:52:16", "throughput": 1899.81, "total_tokens": 75803520}
{"current_steps": 11345, "total_steps": 16324, "loss": 0.1971, "lr": 2.3352157667965034e-05, "epoch": 1.3899421150960154, "percentage": 69.5, "elapsed_time": "11:06:00", "remaining_time": "4:52:17", "throughput": 1897.81, "total_tokens": 75837704}
{"current_steps": 11350, "total_steps": 16324, "loss": 0.2302, "lr": 2.3309315945697335e-05, "epoch": 1.3905546537625186, "percentage": 69.53, "elapsed_time": "11:06:59", "remaining_time": "4:52:18", "throughput": 1895.84, "total_tokens": 75870800}
{"current_steps": 11350, "total_steps": 16324, "eval_loss": 0.1493845283985138, "epoch": 1.3905546537625186, "percentage": 69.53, "elapsed_time": "11:07:19", "remaining_time": "4:52:26", "throughput": 1894.91, "total_tokens": 75870800}
{"current_steps": 11355, "total_steps": 16324, "loss": 0.1958, "lr": 2.3266501610390906e-05, "epoch": 1.391167192429022, "percentage": 69.56, "elapsed_time": "11:08:24", "remaining_time": "4:52:29", "throughput": 1892.69, "total_tokens": 75905072}
{"current_steps": 11360, "total_steps": 16324, "loss": 0.2484, "lr": 2.322371470597699e-05, "epoch": 1.3917797310955253, "percentage": 69.59, "elapsed_time": "11:09:24", "remaining_time": "4:52:30", "throughput": 1890.7, "total_tokens": 75938048}
{"current_steps": 11365, "total_steps": 16324, "loss": 0.2457, "lr": 2.3180955276358628e-05, "epoch": 1.3923922697620288, "percentage": 69.62, "elapsed_time": "11:10:22", "remaining_time": "4:52:30", "throughput": 1888.75, "total_tokens": 75971152}
{"current_steps": 11370, "total_steps": 16324, "loss": 0.219, "lr": 2.3138223365410743e-05, "epoch": 1.393004808428532, "percentage": 69.65, "elapsed_time": "11:11:22", "remaining_time": "4:52:31", "throughput": 1886.8, "total_tokens": 76004648}
{"current_steps": 11375, "total_steps": 16324, "loss": 0.2099, "lr": 2.3095519016979976e-05, "epoch": 1.3936173470950353, "percentage": 69.68, "elapsed_time": "11:12:22", "remaining_time": "4:52:32", "throughput": 1884.8, "total_tokens": 76037936}
{"current_steps": 11380, "total_steps": 16324, "loss": 0.202, "lr": 2.3052842274884728e-05, "epoch": 1.3942298857615387, "percentage": 69.71, "elapsed_time": "11:13:21", "remaining_time": "4:52:32", "throughput": 1882.89, "total_tokens": 76071568}
{"current_steps": 11385, "total_steps": 16324, "loss": 0.2149, "lr": 2.301019318291502e-05, "epoch": 1.394842424428042, "percentage": 69.74, "elapsed_time": "11:14:21", "remaining_time": "4:52:32", "throughput": 1880.93, "total_tokens": 76105328}
{"current_steps": 11390, "total_steps": 16324, "loss": 0.2472, "lr": 2.296757178483251e-05, "epoch": 1.3954549630945454, "percentage": 69.77, "elapsed_time": "11:15:21", "remaining_time": "4:52:33", "throughput": 1878.95, "total_tokens": 76138608}
{"current_steps": 11395, "total_steps": 16324, "loss": 0.1948, "lr": 2.2924978124370472e-05, "epoch": 1.3960675017610487, "percentage": 69.81, "elapsed_time": "11:16:21", "remaining_time": "4:52:33", "throughput": 1877.02, "total_tokens": 76171936}
{"current_steps": 11400, "total_steps": 16324, "loss": 0.2228, "lr": 2.2882412245233727e-05, "epoch": 1.396680040427552, "percentage": 69.84, "elapsed_time": "11:17:20", "remaining_time": "4:52:33", "throughput": 1875.08, "total_tokens": 76204736}
{"current_steps": 11400, "total_steps": 16324, "eval_loss": 0.13607601821422577, "epoch": 1.396680040427552, "percentage": 69.84, "elapsed_time": "11:17:40", "remaining_time": "4:52:42", "throughput": 1874.18, "total_tokens": 76204736}
{"current_steps": 11405, "total_steps": 16324, "loss": 0.246, "lr": 2.2839874191098527e-05, "epoch": 1.3972925790940554, "percentage": 69.87, "elapsed_time": "11:18:47", "remaining_time": "4:52:45", "throughput": 1871.93, "total_tokens": 76238512}
{"current_steps": 11410, "total_steps": 16324, "loss": 0.207, "lr": 2.2797364005612624e-05, "epoch": 1.3979051177605586, "percentage": 69.9, "elapsed_time": "11:19:46", "remaining_time": "4:52:45", "throughput": 1870.01, "total_tokens": 76271944}
{"current_steps": 11415, "total_steps": 16324, "loss": 0.2436, "lr": 2.275488173239519e-05, "epoch": 1.398517656427062, "percentage": 69.93, "elapsed_time": "11:20:46", "remaining_time": "4:52:45", "throughput": 1868.1, "total_tokens": 76305032}
{"current_steps": 11420, "total_steps": 16324, "loss": 0.2002, "lr": 2.271242741503669e-05, "epoch": 1.3991301950935653, "percentage": 69.96, "elapsed_time": "11:21:47", "remaining_time": "4:52:46", "throughput": 1866.12, "total_tokens": 76338824}
{"current_steps": 11425, "total_steps": 16324, "loss": 0.2215, "lr": 2.267000109709896e-05, "epoch": 1.3997427337600685, "percentage": 69.99, "elapsed_time": "11:22:47", "remaining_time": "4:52:46", "throughput": 1864.2, "total_tokens": 76372360}
{"current_steps": 11430, "total_steps": 16324, "loss": 0.2445, "lr": 2.2627602822115083e-05, "epoch": 1.400355272426572, "percentage": 70.02, "elapsed_time": "11:23:47", "remaining_time": "4:52:46", "throughput": 1862.32, "total_tokens": 76405752}
{"current_steps": 11435, "total_steps": 16324, "loss": 0.2329, "lr": 2.2585232633589402e-05, "epoch": 1.4009678110930752, "percentage": 70.05, "elapsed_time": "11:24:47", "remaining_time": "4:52:46", "throughput": 1860.41, "total_tokens": 76439832}
{"current_steps": 11440, "total_steps": 16324, "loss": 0.1915, "lr": 2.2542890574997372e-05, "epoch": 1.4015803497595787, "percentage": 70.08, "elapsed_time": "11:25:46", "remaining_time": "4:52:46", "throughput": 1858.56, "total_tokens": 76473504}
{"current_steps": 11445, "total_steps": 16324, "loss": 0.1851, "lr": 2.2500576689785645e-05, "epoch": 1.402192888426082, "percentage": 70.11, "elapsed_time": "11:26:46", "remaining_time": "4:52:46", "throughput": 1856.7, "total_tokens": 76507384}
{"current_steps": 11450, "total_steps": 16324, "loss": 0.2203, "lr": 2.245829102137196e-05, "epoch": 1.4028054270925852, "percentage": 70.14, "elapsed_time": "11:27:46", "remaining_time": "4:52:46", "throughput": 1854.79, "total_tokens": 76541376}
{"current_steps": 11450, "total_steps": 16324, "eval_loss": 0.13985607028007507, "epoch": 1.4028054270925852, "percentage": 70.14, "elapsed_time": "11:28:06", "remaining_time": "4:52:54", "throughput": 1853.92, "total_tokens": 76541376}
{"current_steps": 11455, "total_steps": 16324, "loss": 0.2098, "lr": 2.241603361314505e-05, "epoch": 1.4034179657590886, "percentage": 70.17, "elapsed_time": "11:29:11", "remaining_time": "4:52:56", "throughput": 1851.82, "total_tokens": 76575016}
{"current_steps": 11460, "total_steps": 16324, "loss": 0.1934, "lr": 2.2373804508464707e-05, "epoch": 1.4040305044255919, "percentage": 70.2, "elapsed_time": "11:30:10", "remaining_time": "4:52:55", "throughput": 1849.98, "total_tokens": 76608464}
{"current_steps": 11465, "total_steps": 16324, "loss": 0.2108, "lr": 2.233160375066168e-05, "epoch": 1.4046430430920953, "percentage": 70.23, "elapsed_time": "11:31:10", "remaining_time": "4:52:55", "throughput": 1848.09, "total_tokens": 76641856}
{"current_steps": 11470, "total_steps": 16324, "loss": 0.218, "lr": 2.2289431383037594e-05, "epoch": 1.4052555817585985, "percentage": 70.26, "elapsed_time": "11:32:10", "remaining_time": "4:52:55", "throughput": 1846.27, "total_tokens": 76675760}
{"current_steps": 11475, "total_steps": 16324, "loss": 0.2457, "lr": 2.2247287448864952e-05, "epoch": 1.4058681204251018, "percentage": 70.3, "elapsed_time": "11:33:09", "remaining_time": "4:52:54", "throughput": 1844.46, "total_tokens": 76709472}
{"current_steps": 11480, "total_steps": 16324, "loss": 0.198, "lr": 2.2205171991387104e-05, "epoch": 1.406480659091605, "percentage": 70.33, "elapsed_time": "11:34:09", "remaining_time": "4:52:53", "throughput": 1842.61, "total_tokens": 76743120}
{"current_steps": 11485, "total_steps": 16324, "loss": 0.2278, "lr": 2.2163085053818173e-05, "epoch": 1.4070931977581085, "percentage": 70.36, "elapsed_time": "11:35:09", "remaining_time": "4:52:53", "throughput": 1840.75, "total_tokens": 76776432}
{"current_steps": 11490, "total_steps": 16324, "loss": 0.2212, "lr": 2.2121026679343028e-05, "epoch": 1.407705736424612, "percentage": 70.39, "elapsed_time": "11:36:08", "remaining_time": "4:52:52", "throughput": 1838.94, "total_tokens": 76809752}
{"current_steps": 11495, "total_steps": 16324, "loss": 0.2273, "lr": 2.2078996911117182e-05, "epoch": 1.4083182750911152, "percentage": 70.42, "elapsed_time": "11:37:08", "remaining_time": "4:52:51", "throughput": 1837.12, "total_tokens": 76843064}
{"current_steps": 11500, "total_steps": 16324, "loss": 0.2274, "lr": 2.2036995792266852e-05, "epoch": 1.4089308137576184, "percentage": 70.45, "elapsed_time": "11:38:08", "remaining_time": "4:52:51", "throughput": 1835.26, "total_tokens": 76876640}
{"current_steps": 11500, "total_steps": 16324, "eval_loss": 0.14584578573703766, "epoch": 1.4089308137576184, "percentage": 70.45, "elapsed_time": "11:38:28", "remaining_time": "4:52:59", "throughput": 1834.4, "total_tokens": 76876640}
{"current_steps": 11505, "total_steps": 16324, "loss": 0.2269, "lr": 2.199502336588885e-05, "epoch": 1.4095433524241217, "percentage": 70.48, "elapsed_time": "11:39:32", "remaining_time": "4:53:00", "throughput": 1832.38, "total_tokens": 76910112}
{"current_steps": 11510, "total_steps": 16324, "loss": 0.2547, "lr": 2.1953079675050498e-05, "epoch": 1.4101558910906251, "percentage": 70.51, "elapsed_time": "11:40:32", "remaining_time": "4:52:59", "throughput": 1830.57, "total_tokens": 76943248}
{"current_steps": 11515, "total_steps": 16324, "loss": 0.2192, "lr": 2.1911164762789682e-05, "epoch": 1.4107684297571284, "percentage": 70.54, "elapsed_time": "11:41:33", "remaining_time": "4:52:59", "throughput": 1828.71, "total_tokens": 76976416}
{"current_steps": 11520, "total_steps": 16324, "loss": 0.1972, "lr": 2.1869278672114762e-05, "epoch": 1.4113809684236318, "percentage": 70.57, "elapsed_time": "11:42:33", "remaining_time": "4:52:58", "throughput": 1826.92, "total_tokens": 77010144}
{"current_steps": 11525, "total_steps": 16324, "loss": 0.1968, "lr": 2.182742144600448e-05, "epoch": 1.411993507090135, "percentage": 70.6, "elapsed_time": "11:43:32", "remaining_time": "4:52:57", "throughput": 1825.13, "total_tokens": 77044016}
{"current_steps": 11530, "total_steps": 16324, "loss": 0.1695, "lr": 2.1785593127407993e-05, "epoch": 1.4126060457566383, "percentage": 70.63, "elapsed_time": "11:44:33", "remaining_time": "4:52:56", "throughput": 1823.3, "total_tokens": 77077688}
{"current_steps": 11535, "total_steps": 16324, "loss": 0.2303, "lr": 2.1743793759244807e-05, "epoch": 1.4132185844231417, "percentage": 70.66, "elapsed_time": "11:45:33", "remaining_time": "4:52:55", "throughput": 1821.52, "total_tokens": 77111176}
{"current_steps": 11540, "total_steps": 16324, "loss": 0.2219, "lr": 2.1702023384404706e-05, "epoch": 1.413831123089645, "percentage": 70.69, "elapsed_time": "11:46:32", "remaining_time": "4:52:54", "throughput": 1819.75, "total_tokens": 77144408}
{"current_steps": 11545, "total_steps": 16324, "loss": 0.2047, "lr": 2.1660282045747698e-05, "epoch": 1.4144436617561484, "percentage": 70.72, "elapsed_time": "11:47:33", "remaining_time": "4:52:53", "throughput": 1817.95, "total_tokens": 77177960}
{"current_steps": 11550, "total_steps": 16324, "loss": 0.2299, "lr": 2.1618569786104063e-05, "epoch": 1.4150562004226517, "percentage": 70.75, "elapsed_time": "11:48:32", "remaining_time": "4:52:51", "throughput": 1816.2, "total_tokens": 77211216}
{"current_steps": 11550, "total_steps": 16324, "eval_loss": 0.18886250257492065, "epoch": 1.4150562004226517, "percentage": 70.75, "elapsed_time": "11:48:51", "remaining_time": "4:52:59", "throughput": 1815.37, "total_tokens": 77211216}
{"current_steps": 11555, "total_steps": 16324, "loss": 0.2107, "lr": 2.1576886648274157e-05, "epoch": 1.415668739089155, "percentage": 70.79, "elapsed_time": "11:49:56", "remaining_time": "4:53:00", "throughput": 1813.42, "total_tokens": 77245192}
{"current_steps": 11560, "total_steps": 16324, "loss": 0.2082, "lr": 2.1535232675028556e-05, "epoch": 1.4162812777556584, "percentage": 70.82, "elapsed_time": "11:50:56", "remaining_time": "4:52:59", "throughput": 1811.64, "total_tokens": 77278776}
{"current_steps": 11565, "total_steps": 16324, "loss": 0.1563, "lr": 2.14936079091078e-05, "epoch": 1.4168938164221616, "percentage": 70.85, "elapsed_time": "11:51:55", "remaining_time": "4:52:57", "throughput": 1809.97, "total_tokens": 77313896}
{"current_steps": 11570, "total_steps": 16324, "loss": 0.2277, "lr": 2.1452012393222544e-05, "epoch": 1.417506355088665, "percentage": 70.88, "elapsed_time": "11:52:54", "remaining_time": "4:52:55", "throughput": 1808.28, "total_tokens": 77348024}
{"current_steps": 11575, "total_steps": 16324, "loss": 0.2206, "lr": 2.1410446170053418e-05, "epoch": 1.4181188937551683, "percentage": 70.91, "elapsed_time": "11:53:54", "remaining_time": "4:52:54", "throughput": 1806.54, "total_tokens": 77381360}
{"current_steps": 11580, "total_steps": 16324, "loss": 0.2265, "lr": 2.1368909282250944e-05, "epoch": 1.4187314324216715, "percentage": 70.94, "elapsed_time": "11:54:53", "remaining_time": "4:52:52", "throughput": 1804.81, "total_tokens": 77414776}
{"current_steps": 11585, "total_steps": 16324, "loss": 0.2596, "lr": 2.13274017724356e-05, "epoch": 1.419343971088175, "percentage": 70.97, "elapsed_time": "11:55:52", "remaining_time": "4:52:50", "throughput": 1803.11, "total_tokens": 77448456}
{"current_steps": 11590, "total_steps": 16324, "loss": 0.1953, "lr": 2.128592368319769e-05, "epoch": 1.4199565097546782, "percentage": 71.0, "elapsed_time": "11:56:52", "remaining_time": "4:52:48", "throughput": 1801.39, "total_tokens": 77482664}
{"current_steps": 11595, "total_steps": 16324, "loss": 0.2097, "lr": 2.1244475057097364e-05, "epoch": 1.4205690484211817, "percentage": 71.03, "elapsed_time": "11:57:52", "remaining_time": "4:52:46", "throughput": 1799.69, "total_tokens": 77516336}
{"current_steps": 11600, "total_steps": 16324, "loss": 0.2541, "lr": 2.1203055936664473e-05, "epoch": 1.421181587087685, "percentage": 71.06, "elapsed_time": "11:58:51", "remaining_time": "4:52:44", "throughput": 1798.0, "total_tokens": 77549560}
{"current_steps": 11600, "total_steps": 16324, "eval_loss": 0.15744981169700623, "epoch": 1.421181587087685, "percentage": 71.06, "elapsed_time": "11:59:10", "remaining_time": "4:52:52", "throughput": 1797.18, "total_tokens": 77549560}
{"current_steps": 11605, "total_steps": 16324, "loss": 0.2383, "lr": 2.116166636439865e-05, "epoch": 1.4217941257541882, "percentage": 71.09, "elapsed_time": "12:00:16", "remaining_time": "4:52:53", "throughput": 1795.21, "total_tokens": 77582480}
{"current_steps": 11610, "total_steps": 16324, "loss": 0.2478, "lr": 2.1120306382769223e-05, "epoch": 1.4224066644206916, "percentage": 71.12, "elapsed_time": "12:01:16", "remaining_time": "4:52:51", "throughput": 1793.5, "total_tokens": 77616096}
{"current_steps": 11615, "total_steps": 16324, "loss": 0.2231, "lr": 2.1078976034215076e-05, "epoch": 1.4230192030871949, "percentage": 71.15, "elapsed_time": "12:02:16", "remaining_time": "4:52:49", "throughput": 1791.8, "total_tokens": 77649232}
{"current_steps": 11620, "total_steps": 16324, "loss": 0.2292, "lr": 2.1037675361144764e-05, "epoch": 1.4236317417536983, "percentage": 71.18, "elapsed_time": "12:03:16", "remaining_time": "4:52:47", "throughput": 1790.07, "total_tokens": 77683000}
{"current_steps": 11625, "total_steps": 16324, "loss": 0.2083, "lr": 2.0996404405936388e-05, "epoch": 1.4242442804202016, "percentage": 71.21, "elapsed_time": "12:04:16", "remaining_time": "4:52:45", "throughput": 1788.35, "total_tokens": 77716136}
{"current_steps": 11630, "total_steps": 16324, "loss": 0.204, "lr": 2.0955163210937507e-05, "epoch": 1.4248568190867048, "percentage": 71.24, "elapsed_time": "12:05:16", "remaining_time": "4:52:43", "throughput": 1786.66, "total_tokens": 77750064}
{"current_steps": 11635, "total_steps": 16324, "loss": 0.2225, "lr": 2.091395181846516e-05, "epoch": 1.425469357753208, "percentage": 71.28, "elapsed_time": "12:06:17", "remaining_time": "4:52:41", "throughput": 1784.97, "total_tokens": 77783792}
{"current_steps": 11640, "total_steps": 16324, "loss": 0.207, "lr": 2.0872770270805834e-05, "epoch": 1.4260818964197115, "percentage": 71.31, "elapsed_time": "12:07:17", "remaining_time": "4:52:40", "throughput": 1783.27, "total_tokens": 77817472}
{"current_steps": 11645, "total_steps": 16324, "loss": 0.2444, "lr": 2.0831618610215374e-05, "epoch": 1.4266944350862147, "percentage": 71.34, "elapsed_time": "12:08:17", "remaining_time": "4:52:37", "throughput": 1781.58, "total_tokens": 77850496}
{"current_steps": 11650, "total_steps": 16324, "loss": 0.23, "lr": 2.079049687891898e-05, "epoch": 1.4273069737527182, "percentage": 71.37, "elapsed_time": "12:09:17", "remaining_time": "4:52:35", "throughput": 1779.9, "total_tokens": 77884056}
{"current_steps": 11650, "total_steps": 16324, "eval_loss": 0.17718727886676788, "epoch": 1.4273069737527182, "percentage": 71.37, "elapsed_time": "12:09:37", "remaining_time": "4:52:43", "throughput": 1779.09, "total_tokens": 77884056}
{"current_steps": 11655, "total_steps": 16324, "loss": 0.2166, "lr": 2.0749405119111086e-05, "epoch": 1.4279195124192214, "percentage": 71.4, "elapsed_time": "12:10:43", "remaining_time": "4:52:43", "throughput": 1777.17, "total_tokens": 77917736}
{"current_steps": 11660, "total_steps": 16324, "loss": 0.29, "lr": 2.0708343372955417e-05, "epoch": 1.4285320510857247, "percentage": 71.43, "elapsed_time": "12:11:43", "remaining_time": "4:52:41", "throughput": 1775.52, "total_tokens": 77951440}
{"current_steps": 11665, "total_steps": 16324, "loss": 0.2354, "lr": 2.0667311682584927e-05, "epoch": 1.4291445897522281, "percentage": 71.46, "elapsed_time": "12:12:43", "remaining_time": "4:52:38", "throughput": 1773.86, "total_tokens": 77984592}
{"current_steps": 11670, "total_steps": 16324, "loss": 0.199, "lr": 2.0626310090101646e-05, "epoch": 1.4297571284187314, "percentage": 71.49, "elapsed_time": "12:13:44", "remaining_time": "4:52:36", "throughput": 1772.17, "total_tokens": 78018336}
{"current_steps": 11675, "total_steps": 16324, "loss": 0.2172, "lr": 2.0585338637576802e-05, "epoch": 1.4303696670852348, "percentage": 71.52, "elapsed_time": "12:14:43", "remaining_time": "4:52:34", "throughput": 1770.53, "total_tokens": 78051872}
{"current_steps": 11680, "total_steps": 16324, "loss": 0.2054, "lr": 2.0544397367050673e-05, "epoch": 1.430982205751738, "percentage": 71.55, "elapsed_time": "12:15:43", "remaining_time": "4:52:31", "throughput": 1768.91, "total_tokens": 78085520}
{"current_steps": 11685, "total_steps": 16324, "loss": 0.2326, "lr": 2.0503486320532523e-05, "epoch": 1.4315947444182413, "percentage": 71.58, "elapsed_time": "12:16:44", "remaining_time": "4:52:29", "throughput": 1767.22, "total_tokens": 78118664}
{"current_steps": 11690, "total_steps": 16324, "loss": 0.2424, "lr": 2.0462605540000668e-05, "epoch": 1.4322072830847448, "percentage": 71.61, "elapsed_time": "12:17:43", "remaining_time": "4:52:26", "throughput": 1765.59, "total_tokens": 78151408}
{"current_steps": 11695, "total_steps": 16324, "loss": 0.2131, "lr": 2.042175506740233e-05, "epoch": 1.432819821751248, "percentage": 71.64, "elapsed_time": "12:18:43", "remaining_time": "4:52:23", "throughput": 1763.97, "total_tokens": 78184680}
{"current_steps": 11700, "total_steps": 16324, "loss": 0.2303, "lr": 2.038093494465368e-05, "epoch": 1.4334323604177515, "percentage": 71.67, "elapsed_time": "12:19:44", "remaining_time": "4:52:21", "throughput": 1762.29, "total_tokens": 78217776}
{"current_steps": 11700, "total_steps": 16324, "eval_loss": 0.1379169523715973, "epoch": 1.4334323604177515, "percentage": 71.67, "elapsed_time": "12:20:03", "remaining_time": "4:52:29", "throughput": 1761.5, "total_tokens": 78217776}
{"current_steps": 11705, "total_steps": 16324, "loss": 0.2127, "lr": 2.0340145213639655e-05, "epoch": 1.4340448990842547, "percentage": 71.7, "elapsed_time": "12:21:09", "remaining_time": "4:52:28", "throughput": 1759.67, "total_tokens": 78251376}
{"current_steps": 11710, "total_steps": 16324, "loss": 0.2821, "lr": 2.0299385916214116e-05, "epoch": 1.434657437750758, "percentage": 71.73, "elapsed_time": "12:22:09", "remaining_time": "4:52:25", "throughput": 1758.05, "total_tokens": 78284456}
{"current_steps": 11715, "total_steps": 16324, "loss": 0.2111, "lr": 2.0258657094199597e-05, "epoch": 1.4352699764172614, "percentage": 71.77, "elapsed_time": "12:23:10", "remaining_time": "4:52:23", "throughput": 1756.4, "total_tokens": 78318256}
{"current_steps": 11720, "total_steps": 16324, "loss": 0.1973, "lr": 2.0217958789387446e-05, "epoch": 1.4358825150837646, "percentage": 71.8, "elapsed_time": "12:24:10", "remaining_time": "4:52:20", "throughput": 1754.78, "total_tokens": 78351968}
{"current_steps": 11725, "total_steps": 16324, "loss": 0.2047, "lr": 2.017729104353763e-05, "epoch": 1.436495053750268, "percentage": 71.83, "elapsed_time": "12:25:10", "remaining_time": "4:52:17", "throughput": 1753.18, "total_tokens": 78385944}
{"current_steps": 11730, "total_steps": 16324, "loss": 0.2477, "lr": 2.0136653898378805e-05, "epoch": 1.4371075924167713, "percentage": 71.86, "elapsed_time": "12:26:11", "remaining_time": "4:52:14", "throughput": 1751.53, "total_tokens": 78418960}
{"current_steps": 11735, "total_steps": 16324, "loss": 0.215, "lr": 2.009604739560823e-05, "epoch": 1.4377201310832746, "percentage": 71.89, "elapsed_time": "12:27:12", "remaining_time": "4:52:11", "throughput": 1749.92, "total_tokens": 78452624}
{"current_steps": 11740, "total_steps": 16324, "loss": 0.2267, "lr": 2.0055471576891672e-05, "epoch": 1.438332669749778, "percentage": 71.92, "elapsed_time": "12:28:12", "remaining_time": "4:52:08", "throughput": 1748.31, "total_tokens": 78485592}
{"current_steps": 11745, "total_steps": 16324, "loss": 0.2245, "lr": 2.0014926483863466e-05, "epoch": 1.4389452084162813, "percentage": 71.95, "elapsed_time": "12:29:12", "remaining_time": "4:52:05", "throughput": 1746.7, "total_tokens": 78519176}
{"current_steps": 11750, "total_steps": 16324, "loss": 0.2207, "lr": 1.99744121581264e-05, "epoch": 1.4395577470827847, "percentage": 71.98, "elapsed_time": "12:30:14", "remaining_time": "4:52:02", "throughput": 1745.07, "total_tokens": 78552600}
{"current_steps": 11750, "total_steps": 16324, "eval_loss": 0.07964655011892319, "epoch": 1.4395577470827847, "percentage": 71.98, "elapsed_time": "12:30:33", "remaining_time": "4:52:10", "throughput": 1744.32, "total_tokens": 78552600}
{"current_steps": 11755, "total_steps": 16324, "loss": 0.2158, "lr": 1.9933928641251702e-05, "epoch": 1.440170285749288, "percentage": 72.01, "elapsed_time": "12:31:39", "remaining_time": "4:52:09", "throughput": 1742.5, "total_tokens": 78586312}
{"current_steps": 11760, "total_steps": 16324, "loss": 0.1939, "lr": 1.9893475974778957e-05, "epoch": 1.4407828244157912, "percentage": 72.04, "elapsed_time": "12:32:40", "remaining_time": "4:52:06", "throughput": 1740.92, "total_tokens": 78620064}
{"current_steps": 11765, "total_steps": 16324, "loss": 0.212, "lr": 1.9853054200216124e-05, "epoch": 1.4413953630822944, "percentage": 72.07, "elapsed_time": "12:33:40", "remaining_time": "4:52:03", "throughput": 1739.33, "total_tokens": 78653936}
{"current_steps": 11770, "total_steps": 16324, "loss": 0.2009, "lr": 1.9812663359039485e-05, "epoch": 1.442007901748798, "percentage": 72.1, "elapsed_time": "12:34:40", "remaining_time": "4:51:59", "throughput": 1737.79, "total_tokens": 78687776}
{"current_steps": 11775, "total_steps": 16324, "loss": 0.2234, "lr": 1.97723034926935e-05, "epoch": 1.4426204404153014, "percentage": 72.13, "elapsed_time": "12:35:40", "remaining_time": "4:51:56", "throughput": 1736.22, "total_tokens": 78721528}
{"current_steps": 11780, "total_steps": 16324, "loss": 0.2094, "lr": 1.9731974642590933e-05, "epoch": 1.4432329790818046, "percentage": 72.16, "elapsed_time": "12:36:40", "remaining_time": "4:51:52", "throughput": 1734.68, "total_tokens": 78755520}
{"current_steps": 11785, "total_steps": 16324, "loss": 0.2597, "lr": 1.96916768501127e-05, "epoch": 1.4438455177483078, "percentage": 72.19, "elapsed_time": "12:37:40", "remaining_time": "4:51:49", "throughput": 1733.13, "total_tokens": 78788352}
{"current_steps": 11790, "total_steps": 16324, "loss": 0.1969, "lr": 1.9651410156607803e-05, "epoch": 1.444458056414811, "percentage": 72.22, "elapsed_time": "12:38:39", "remaining_time": "4:51:45", "throughput": 1731.6, "total_tokens": 78821984}
{"current_steps": 11795, "total_steps": 16324, "loss": 0.2171, "lr": 1.9611174603393402e-05, "epoch": 1.4450705950813145, "percentage": 72.26, "elapsed_time": "12:39:40", "remaining_time": "4:51:41", "throughput": 1730.02, "total_tokens": 78855176}
{"current_steps": 11800, "total_steps": 16324, "loss": 0.2276, "lr": 1.9570970231754636e-05, "epoch": 1.4456831337478178, "percentage": 72.29, "elapsed_time": "12:40:40", "remaining_time": "4:51:37", "throughput": 1728.49, "total_tokens": 78888720}
{"current_steps": 11800, "total_steps": 16324, "eval_loss": 0.07458024471998215, "epoch": 1.4456831337478178, "percentage": 72.29, "elapsed_time": "12:40:59", "remaining_time": "4:51:45", "throughput": 1727.75, "total_tokens": 78888720}
{"current_steps": 11805, "total_steps": 16324, "loss": 0.2215, "lr": 1.9530797082944713e-05, "epoch": 1.4462956724143212, "percentage": 72.32, "elapsed_time": "12:42:05", "remaining_time": "4:51:43", "throughput": 1726.0, "total_tokens": 78921912}
{"current_steps": 11810, "total_steps": 16324, "loss": 0.2606, "lr": 1.9490655198184778e-05, "epoch": 1.4469082110808245, "percentage": 72.35, "elapsed_time": "12:43:05", "remaining_time": "4:51:40", "throughput": 1724.44, "total_tokens": 78954768}
{"current_steps": 11815, "total_steps": 16324, "loss": 0.2183, "lr": 1.945054461866386e-05, "epoch": 1.4475207497473277, "percentage": 72.38, "elapsed_time": "12:44:05", "remaining_time": "4:51:36", "throughput": 1722.94, "total_tokens": 78988256}
{"current_steps": 11820, "total_steps": 16324, "loss": 0.1923, "lr": 1.941046538553892e-05, "epoch": 1.4481332884138312, "percentage": 72.41, "elapsed_time": "12:45:04", "remaining_time": "4:51:31", "throughput": 1721.45, "total_tokens": 79022296}
{"current_steps": 11825, "total_steps": 16324, "loss": 0.2035, "lr": 1.937041753993475e-05, "epoch": 1.4487458270803344, "percentage": 72.44, "elapsed_time": "12:46:05", "remaining_time": "4:51:28", "throughput": 1719.88, "total_tokens": 79055696}
{"current_steps": 11830, "total_steps": 16324, "loss": 0.2214, "lr": 1.933040112294389e-05, "epoch": 1.4493583657468379, "percentage": 72.47, "elapsed_time": "12:47:05", "remaining_time": "4:51:24", "throughput": 1718.39, "total_tokens": 79088904}
{"current_steps": 11835, "total_steps": 16324, "loss": 0.2164, "lr": 1.9290416175626676e-05, "epoch": 1.449970904413341, "percentage": 72.5, "elapsed_time": "12:48:04", "remaining_time": "4:51:19", "throughput": 1716.89, "total_tokens": 79122224}
{"current_steps": 11840, "total_steps": 16324, "loss": 0.2127, "lr": 1.9250462739011148e-05, "epoch": 1.4505834430798443, "percentage": 72.53, "elapsed_time": "12:49:05", "remaining_time": "4:51:16", "throughput": 1715.36, "total_tokens": 79155792}
{"current_steps": 11845, "total_steps": 16324, "loss": 0.2558, "lr": 1.9210540854092978e-05, "epoch": 1.4511959817463478, "percentage": 72.56, "elapsed_time": "12:50:05", "remaining_time": "4:51:11", "throughput": 1713.87, "total_tokens": 79189896}
{"current_steps": 11850, "total_steps": 16324, "loss": 0.2157, "lr": 1.9170650561835506e-05, "epoch": 1.451808520412851, "percentage": 72.59, "elapsed_time": "12:51:04", "remaining_time": "4:51:07", "throughput": 1712.4, "total_tokens": 79223936}
{"current_steps": 11850, "total_steps": 16324, "eval_loss": 0.09504926949739456, "epoch": 1.451808520412851, "percentage": 72.59, "elapsed_time": "12:51:24", "remaining_time": "4:51:14", "throughput": 1711.66, "total_tokens": 79223936}
{"current_steps": 11855, "total_steps": 16324, "loss": 0.2199, "lr": 1.9130791903169638e-05, "epoch": 1.4524210590793545, "percentage": 72.62, "elapsed_time": "12:52:30", "remaining_time": "4:51:12", "throughput": 1709.94, "total_tokens": 79257208}
{"current_steps": 11860, "total_steps": 16324, "loss": 0.1793, "lr": 1.9090964918993836e-05, "epoch": 1.4530335977458577, "percentage": 72.65, "elapsed_time": "12:53:30", "remaining_time": "4:51:08", "throughput": 1708.47, "total_tokens": 79291584}
{"current_steps": 11865, "total_steps": 16324, "loss": 0.2024, "lr": 1.9051169650174017e-05, "epoch": 1.453646136412361, "percentage": 72.68, "elapsed_time": "12:54:30", "remaining_time": "4:51:04", "throughput": 1707.01, "total_tokens": 79325208}
{"current_steps": 11870, "total_steps": 16324, "loss": 0.235, "lr": 1.9011406137543597e-05, "epoch": 1.4542586750788644, "percentage": 72.72, "elapsed_time": "12:55:30", "remaining_time": "4:50:59", "throughput": 1705.52, "total_tokens": 79358368}
{"current_steps": 11875, "total_steps": 16324, "loss": 0.2488, "lr": 1.8971674421903413e-05, "epoch": 1.4548712137453677, "percentage": 72.75, "elapsed_time": "12:56:30", "remaining_time": "4:50:55", "throughput": 1704.03, "total_tokens": 79391560}
{"current_steps": 11880, "total_steps": 16324, "loss": 0.2149, "lr": 1.8931974544021645e-05, "epoch": 1.4554837524118711, "percentage": 72.78, "elapsed_time": "12:57:29", "remaining_time": "4:50:50", "throughput": 1702.6, "total_tokens": 79425088}
{"current_steps": 11885, "total_steps": 16324, "loss": 0.184, "lr": 1.8892306544633798e-05, "epoch": 1.4560962910783743, "percentage": 72.81, "elapsed_time": "12:58:28", "remaining_time": "4:50:45", "throughput": 1701.17, "total_tokens": 79459464}
{"current_steps": 11890, "total_steps": 16324, "loss": 0.2283, "lr": 1.88526704644427e-05, "epoch": 1.4567088297448776, "percentage": 72.84, "elapsed_time": "12:59:28", "remaining_time": "4:50:40", "throughput": 1699.72, "total_tokens": 79493000}
{"current_steps": 11895, "total_steps": 16324, "loss": 0.2184, "lr": 1.881306634411844e-05, "epoch": 1.457321368411381, "percentage": 72.87, "elapsed_time": "13:00:26", "remaining_time": "4:50:35", "throughput": 1698.32, "total_tokens": 79526712}
{"current_steps": 11900, "total_steps": 16324, "loss": 0.2073, "lr": 1.8773494224298244e-05, "epoch": 1.4579339070778843, "percentage": 72.9, "elapsed_time": "13:01:26", "remaining_time": "4:50:30", "throughput": 1696.89, "total_tokens": 79560856}
{"current_steps": 11900, "total_steps": 16324, "eval_loss": 0.09442640841007233, "epoch": 1.4579339070778843, "percentage": 72.9, "elapsed_time": "13:01:45", "remaining_time": "4:50:37", "throughput": 1696.19, "total_tokens": 79560856}
{"current_steps": 11905, "total_steps": 16324, "loss": 0.1984, "lr": 1.8733954145586564e-05, "epoch": 1.4585464457443877, "percentage": 72.93, "elapsed_time": "13:02:51", "remaining_time": "4:50:35", "throughput": 1694.54, "total_tokens": 79594856}
{"current_steps": 11910, "total_steps": 16324, "loss": 0.2249, "lr": 1.869444614855496e-05, "epoch": 1.459158984410891, "percentage": 72.96, "elapsed_time": "13:03:50", "remaining_time": "4:50:30", "throughput": 1693.13, "total_tokens": 79628416}
{"current_steps": 11915, "total_steps": 16324, "loss": 0.2391, "lr": 1.865497027374209e-05, "epoch": 1.4597715230773942, "percentage": 72.99, "elapsed_time": "13:04:49", "remaining_time": "4:50:24", "throughput": 1691.71, "total_tokens": 79661656}
{"current_steps": 11920, "total_steps": 16324, "loss": 0.1946, "lr": 1.861552656165359e-05, "epoch": 1.4603840617438975, "percentage": 73.02, "elapsed_time": "13:05:49", "remaining_time": "4:50:19", "throughput": 1690.28, "total_tokens": 79695424}
{"current_steps": 11925, "total_steps": 16324, "loss": 0.2245, "lr": 1.857611505276216e-05, "epoch": 1.460996600410401, "percentage": 73.05, "elapsed_time": "13:06:48", "remaining_time": "4:50:14", "throughput": 1688.88, "total_tokens": 79728696}
{"current_steps": 11930, "total_steps": 16324, "loss": 0.2071, "lr": 1.8536735787507447e-05, "epoch": 1.4616091390769042, "percentage": 73.08, "elapsed_time": "13:07:47", "remaining_time": "4:50:09", "throughput": 1687.49, "total_tokens": 79762584}
{"current_steps": 11935, "total_steps": 16324, "loss": 0.1768, "lr": 1.849738880629597e-05, "epoch": 1.4622216777434076, "percentage": 73.11, "elapsed_time": "13:08:46", "remaining_time": "4:50:03", "throughput": 1686.09, "total_tokens": 79796928}
{"current_steps": 11940, "total_steps": 16324, "loss": 0.223, "lr": 1.8458074149501153e-05, "epoch": 1.4628342164099108, "percentage": 73.14, "elapsed_time": "13:09:45", "remaining_time": "4:49:58", "throughput": 1684.7, "total_tokens": 79830296}
{"current_steps": 11945, "total_steps": 16324, "loss": 0.1694, "lr": 1.8418791857463287e-05, "epoch": 1.463446755076414, "percentage": 73.17, "elapsed_time": "13:10:44", "remaining_time": "4:49:53", "throughput": 1683.32, "total_tokens": 79864528}
{"current_steps": 11950, "total_steps": 16324, "loss": 0.1818, "lr": 1.8379541970489366e-05, "epoch": 1.4640592937429175, "percentage": 73.21, "elapsed_time": "13:11:45", "remaining_time": "4:49:48", "throughput": 1681.87, "total_tokens": 79898064}
{"current_steps": 11950, "total_steps": 16324, "eval_loss": 0.1088213175535202, "epoch": 1.4640592937429175, "percentage": 73.21, "elapsed_time": "13:12:04", "remaining_time": "4:49:55", "throughput": 1681.18, "total_tokens": 79898064}
{"current_steps": 11955, "total_steps": 16324, "loss": 0.2375, "lr": 1.8340324528853237e-05, "epoch": 1.4646718324094208, "percentage": 73.24, "elapsed_time": "13:13:09", "remaining_time": "4:49:51", "throughput": 1679.6, "total_tokens": 79931680}
{"current_steps": 11960, "total_steps": 16324, "loss": 0.2129, "lr": 1.830113957279535e-05, "epoch": 1.4652843710759242, "percentage": 73.27, "elapsed_time": "13:14:08", "remaining_time": "4:49:46", "throughput": 1678.23, "total_tokens": 79966176}
{"current_steps": 11965, "total_steps": 16324, "loss": 0.2137, "lr": 1.8261987142522906e-05, "epoch": 1.4658969097424275, "percentage": 73.3, "elapsed_time": "13:15:09", "remaining_time": "4:49:41", "throughput": 1676.84, "total_tokens": 80000440}
{"current_steps": 11970, "total_steps": 16324, "loss": 0.1713, "lr": 1.8222867278209714e-05, "epoch": 1.4665094484089307, "percentage": 73.33, "elapsed_time": "13:16:09", "remaining_time": "4:49:35", "throughput": 1675.43, "total_tokens": 80034544}
{"current_steps": 11975, "total_steps": 16324, "loss": 0.1994, "lr": 1.8183780019996123e-05, "epoch": 1.4671219870754342, "percentage": 73.36, "elapsed_time": "13:17:08", "remaining_time": "4:49:30", "throughput": 1674.08, "total_tokens": 80068616}
{"current_steps": 11980, "total_steps": 16324, "loss": 0.2176, "lr": 1.814472540798906e-05, "epoch": 1.4677345257419374, "percentage": 73.39, "elapsed_time": "13:18:08", "remaining_time": "4:49:24", "throughput": 1672.67, "total_tokens": 80102032}
{"current_steps": 11985, "total_steps": 16324, "loss": 0.2363, "lr": 1.8105703482261983e-05, "epoch": 1.4683470644084409, "percentage": 73.42, "elapsed_time": "13:19:08", "remaining_time": "4:49:18", "throughput": 1671.31, "total_tokens": 80136208}
{"current_steps": 11990, "total_steps": 16324, "loss": 0.2082, "lr": 1.806671428285474e-05, "epoch": 1.468959603074944, "percentage": 73.45, "elapsed_time": "13:20:06", "remaining_time": "4:49:12", "throughput": 1669.98, "total_tokens": 80170208}
{"current_steps": 11995, "total_steps": 16324, "loss": 0.1776, "lr": 1.8027757849773657e-05, "epoch": 1.4695721417414473, "percentage": 73.48, "elapsed_time": "13:21:06", "remaining_time": "4:49:07", "throughput": 1668.63, "total_tokens": 80205136}
{"current_steps": 12000, "total_steps": 16324, "loss": 0.2189, "lr": 1.798883422299143e-05, "epoch": 1.4701846804079508, "percentage": 73.51, "elapsed_time": "13:22:05", "remaining_time": "4:49:01", "throughput": 1667.27, "total_tokens": 80238176}
{"current_steps": 12000, "total_steps": 16324, "eval_loss": 0.07542052865028381, "epoch": 1.4701846804079508, "percentage": 73.51, "elapsed_time": "13:22:24", "remaining_time": "4:49:08", "throughput": 1666.6, "total_tokens": 80238176}
{"current_steps": 12005, "total_steps": 16324, "loss": 0.2321, "lr": 1.7949943442447054e-05, "epoch": 1.470797219074454, "percentage": 73.54, "elapsed_time": "13:23:28", "remaining_time": "4:49:03", "throughput": 1665.09, "total_tokens": 80271304}
{"current_steps": 12010, "total_steps": 16324, "loss": 0.2077, "lr": 1.7911085548045863e-05, "epoch": 1.4714097577409575, "percentage": 73.57, "elapsed_time": "13:24:28", "remaining_time": "4:48:58", "throughput": 1663.73, "total_tokens": 80305440}
{"current_steps": 12015, "total_steps": 16324, "loss": 0.2186, "lr": 1.7872260579659437e-05, "epoch": 1.4720222964074607, "percentage": 73.6, "elapsed_time": "13:25:27", "remaining_time": "4:48:51", "throughput": 1662.39, "total_tokens": 80339392}
{"current_steps": 12020, "total_steps": 16324, "loss": 0.2285, "lr": 1.7833468577125584e-05, "epoch": 1.472634835073964, "percentage": 73.63, "elapsed_time": "13:26:26", "remaining_time": "4:48:45", "throughput": 1661.05, "total_tokens": 80372552}
{"current_steps": 12025, "total_steps": 16324, "loss": 0.2289, "lr": 1.7794709580248237e-05, "epoch": 1.4732473737404674, "percentage": 73.66, "elapsed_time": "13:27:27", "remaining_time": "4:48:40", "throughput": 1659.68, "total_tokens": 80406488}
{"current_steps": 12030, "total_steps": 16324, "loss": 0.2, "lr": 1.7755983628797508e-05, "epoch": 1.4738599124069707, "percentage": 73.7, "elapsed_time": "13:28:27", "remaining_time": "4:48:34", "throughput": 1658.31, "total_tokens": 80441080}
{"current_steps": 12035, "total_steps": 16324, "loss": 0.1899, "lr": 1.7717290762509603e-05, "epoch": 1.4744724510734741, "percentage": 73.73, "elapsed_time": "13:29:27", "remaining_time": "4:48:28", "throughput": 1656.96, "total_tokens": 80474952}
{"current_steps": 12040, "total_steps": 16324, "loss": 0.2064, "lr": 1.7678631021086755e-05, "epoch": 1.4750849897399774, "percentage": 73.76, "elapsed_time": "13:30:28", "remaining_time": "4:48:22", "throughput": 1655.58, "total_tokens": 80508448}
{"current_steps": 12045, "total_steps": 16324, "loss": 0.2092, "lr": 1.76400044441972e-05, "epoch": 1.4756975284064806, "percentage": 73.79, "elapsed_time": "13:31:29", "remaining_time": "4:48:17", "throughput": 1654.2, "total_tokens": 80542456}
{"current_steps": 12050, "total_steps": 16324, "loss": 0.2262, "lr": 1.7601411071475175e-05, "epoch": 1.4763100670729838, "percentage": 73.82, "elapsed_time": "13:32:30", "remaining_time": "4:48:11", "throughput": 1652.84, "total_tokens": 80575888}
{"current_steps": 12050, "total_steps": 16324, "eval_loss": 0.10674486309289932, "epoch": 1.4763100670729838, "percentage": 73.82, "elapsed_time": "13:32:49", "remaining_time": "4:48:18", "throughput": 1652.17, "total_tokens": 80575888}
{"current_steps": 12055, "total_steps": 16324, "loss": 0.2615, "lr": 1.756285094252086e-05, "epoch": 1.4769226057394873, "percentage": 73.85, "elapsed_time": "13:33:56", "remaining_time": "4:48:14", "throughput": 1650.6, "total_tokens": 80609304}
{"current_steps": 12060, "total_steps": 16324, "loss": 0.2337, "lr": 1.752432409690025e-05, "epoch": 1.4775351444059908, "percentage": 73.88, "elapsed_time": "13:34:57", "remaining_time": "4:48:08", "throughput": 1649.23, "total_tokens": 80642480}
{"current_steps": 12065, "total_steps": 16324, "loss": 0.2164, "lr": 1.7485830574145268e-05, "epoch": 1.478147683072494, "percentage": 73.91, "elapsed_time": "13:35:57", "remaining_time": "4:48:02", "throughput": 1647.88, "total_tokens": 80676336}
{"current_steps": 12070, "total_steps": 16324, "loss": 0.2001, "lr": 1.7447370413753612e-05, "epoch": 1.4787602217389972, "percentage": 73.94, "elapsed_time": "13:36:57", "remaining_time": "4:47:55", "throughput": 1646.55, "total_tokens": 80710200}
{"current_steps": 12075, "total_steps": 16324, "loss": 0.1961, "lr": 1.7408943655188753e-05, "epoch": 1.4793727604055005, "percentage": 73.97, "elapsed_time": "13:37:59", "remaining_time": "4:47:50", "throughput": 1645.18, "total_tokens": 80743864}
{"current_steps": 12080, "total_steps": 16324, "loss": 0.2163, "lr": 1.737055033787986e-05, "epoch": 1.479985299072004, "percentage": 74.0, "elapsed_time": "13:38:59", "remaining_time": "4:47:43", "throughput": 1643.85, "total_tokens": 80777296}
{"current_steps": 12085, "total_steps": 16324, "loss": 0.2491, "lr": 1.733219050122183e-05, "epoch": 1.4805978377385072, "percentage": 74.03, "elapsed_time": "13:39:59", "remaining_time": "4:47:37", "throughput": 1642.52, "total_tokens": 80810832}
{"current_steps": 12090, "total_steps": 16324, "loss": 0.1939, "lr": 1.7293864184575203e-05, "epoch": 1.4812103764050106, "percentage": 74.06, "elapsed_time": "13:41:00", "remaining_time": "4:47:31", "throughput": 1641.16, "total_tokens": 80845120}
{"current_steps": 12095, "total_steps": 16324, "loss": 0.1617, "lr": 1.725557142726608e-05, "epoch": 1.4818229150715139, "percentage": 74.09, "elapsed_time": "13:42:00", "remaining_time": "4:47:24", "throughput": 1639.87, "total_tokens": 80879472}
{"current_steps": 12100, "total_steps": 16324, "loss": 0.2088, "lr": 1.7217312268586177e-05, "epoch": 1.482435453738017, "percentage": 74.12, "elapsed_time": "13:43:00", "remaining_time": "4:47:18", "throughput": 1638.56, "total_tokens": 80912808}
{"current_steps": 12100, "total_steps": 16324, "eval_loss": 0.12759126722812653, "epoch": 1.482435453738017, "percentage": 74.12, "elapsed_time": "13:43:19", "remaining_time": "4:47:25", "throughput": 1637.91, "total_tokens": 80912808}
{"current_steps": 12105, "total_steps": 16324, "loss": 0.2205, "lr": 1.7179086747792705e-05, "epoch": 1.4830479924045206, "percentage": 74.15, "elapsed_time": "13:44:26", "remaining_time": "4:47:20", "throughput": 1636.38, "total_tokens": 80945688}
{"current_steps": 12110, "total_steps": 16324, "loss": 0.2271, "lr": 1.7140894904108396e-05, "epoch": 1.4836605310710238, "percentage": 74.19, "elapsed_time": "13:45:25", "remaining_time": "4:47:13", "throughput": 1635.08, "total_tokens": 80979064}
{"current_steps": 12115, "total_steps": 16324, "loss": 0.219, "lr": 1.7102736776721383e-05, "epoch": 1.4842730697375273, "percentage": 74.22, "elapsed_time": "13:46:25", "remaining_time": "4:47:07", "throughput": 1633.79, "total_tokens": 81012720}
{"current_steps": 12120, "total_steps": 16324, "loss": 0.1997, "lr": 1.7064612404785196e-05, "epoch": 1.4848856084040305, "percentage": 74.25, "elapsed_time": "13:47:27", "remaining_time": "4:47:00", "throughput": 1632.45, "total_tokens": 81046296}
{"current_steps": 12125, "total_steps": 16324, "loss": 0.2113, "lr": 1.7026521827418786e-05, "epoch": 1.4854981470705337, "percentage": 74.28, "elapsed_time": "13:48:27", "remaining_time": "4:46:54", "throughput": 1631.13, "total_tokens": 81079272}
{"current_steps": 12130, "total_steps": 16324, "loss": 0.1786, "lr": 1.698846508370639e-05, "epoch": 1.4861106857370372, "percentage": 74.31, "elapsed_time": "13:49:27", "remaining_time": "4:46:47", "throughput": 1629.86, "total_tokens": 81113592}
{"current_steps": 12135, "total_steps": 16324, "loss": 0.2348, "lr": 1.695044221269752e-05, "epoch": 1.4867232244035404, "percentage": 74.34, "elapsed_time": "13:50:28", "remaining_time": "4:46:40", "throughput": 1628.53, "total_tokens": 81146704}
{"current_steps": 12140, "total_steps": 16324, "loss": 0.224, "lr": 1.6912453253406958e-05, "epoch": 1.487335763070044, "percentage": 74.37, "elapsed_time": "13:51:28", "remaining_time": "4:46:33", "throughput": 1627.23, "total_tokens": 81179856}
{"current_steps": 12145, "total_steps": 16324, "loss": 0.1995, "lr": 1.687449824481469e-05, "epoch": 1.4879483017365471, "percentage": 74.4, "elapsed_time": "13:52:28", "remaining_time": "4:46:26", "throughput": 1625.95, "total_tokens": 81213312}
{"current_steps": 12150, "total_steps": 16324, "loss": 0.2078, "lr": 1.683657722586582e-05, "epoch": 1.4885608404030504, "percentage": 74.43, "elapsed_time": "13:53:28", "remaining_time": "4:46:19", "throughput": 1624.65, "total_tokens": 81247096}
{"current_steps": 12150, "total_steps": 16324, "eval_loss": 0.1698664128780365, "epoch": 1.4885608404030504, "percentage": 74.43, "elapsed_time": "13:53:48", "remaining_time": "4:46:26", "throughput": 1624.02, "total_tokens": 81247096}
{"current_steps": 12155, "total_steps": 16324, "loss": 0.2268, "lr": 1.6798690235470628e-05, "epoch": 1.4891733790695538, "percentage": 74.46, "elapsed_time": "13:54:54", "remaining_time": "4:46:21", "throughput": 1622.56, "total_tokens": 81280840}
{"current_steps": 12160, "total_steps": 16324, "loss": 0.2036, "lr": 1.6760837312504473e-05, "epoch": 1.489785917736057, "percentage": 74.49, "elapsed_time": "13:55:54", "remaining_time": "4:46:14", "throughput": 1621.29, "total_tokens": 81314448}
{"current_steps": 12165, "total_steps": 16324, "loss": 0.182, "lr": 1.672301849580771e-05, "epoch": 1.4903984564025605, "percentage": 74.52, "elapsed_time": "13:56:55", "remaining_time": "4:46:07", "throughput": 1619.98, "total_tokens": 81348608}
{"current_steps": 12170, "total_steps": 16324, "loss": 0.2224, "lr": 1.6685233824185752e-05, "epoch": 1.4910109950690638, "percentage": 74.55, "elapsed_time": "13:57:57", "remaining_time": "4:46:01", "throughput": 1618.67, "total_tokens": 81382256}
{"current_steps": 12175, "total_steps": 16324, "loss": 0.2027, "lr": 1.664748333640896e-05, "epoch": 1.491623533735567, "percentage": 74.58, "elapsed_time": "13:58:58", "remaining_time": "4:45:54", "throughput": 1617.39, "total_tokens": 81416208}
{"current_steps": 12180, "total_steps": 16324, "loss": 0.2203, "lr": 1.660976707121262e-05, "epoch": 1.4922360724020705, "percentage": 74.61, "elapsed_time": "13:59:59", "remaining_time": "4:45:47", "throughput": 1616.09, "total_tokens": 81449672}
{"current_steps": 12185, "total_steps": 16324, "loss": 0.2567, "lr": 1.657208506729686e-05, "epoch": 1.4928486110685737, "percentage": 74.64, "elapsed_time": "14:01:00", "remaining_time": "4:45:40", "throughput": 1614.79, "total_tokens": 81483056}
{"current_steps": 12190, "total_steps": 16324, "loss": 0.2132, "lr": 1.6534437363326726e-05, "epoch": 1.4934611497350772, "percentage": 74.68, "elapsed_time": "14:02:00", "remaining_time": "4:45:33", "throughput": 1613.54, "total_tokens": 81516992}
{"current_steps": 12195, "total_steps": 16324, "loss": 0.1981, "lr": 1.6496823997932037e-05, "epoch": 1.4940736884015804, "percentage": 74.71, "elapsed_time": "14:03:01", "remaining_time": "4:45:26", "throughput": 1612.26, "total_tokens": 81551128}
{"current_steps": 12200, "total_steps": 16324, "loss": 0.2199, "lr": 1.6459245009707363e-05, "epoch": 1.4946862270680836, "percentage": 74.74, "elapsed_time": "14:04:03", "remaining_time": "4:45:19", "throughput": 1610.97, "total_tokens": 81584816}
{"current_steps": 12200, "total_steps": 16324, "eval_loss": 0.16632874310016632, "epoch": 1.4946862270680836, "percentage": 74.74, "elapsed_time": "14:04:23", "remaining_time": "4:45:25", "throughput": 1610.34, "total_tokens": 81584816}
{"current_steps": 12205, "total_steps": 16324, "loss": 0.1831, "lr": 1.6421700437211996e-05, "epoch": 1.4952987657345869, "percentage": 74.77, "elapsed_time": "14:05:28", "remaining_time": "4:45:20", "throughput": 1608.94, "total_tokens": 81619184}
{"current_steps": 12210, "total_steps": 16324, "loss": 0.2632, "lr": 1.638419031896994e-05, "epoch": 1.4959113044010903, "percentage": 74.8, "elapsed_time": "14:06:28", "remaining_time": "4:45:12", "throughput": 1607.7, "total_tokens": 81652656}
{"current_steps": 12215, "total_steps": 16324, "loss": 0.2505, "lr": 1.634671469346986e-05, "epoch": 1.4965238430675936, "percentage": 74.83, "elapsed_time": "14:07:30", "remaining_time": "4:45:05", "throughput": 1606.42, "total_tokens": 81686304}
{"current_steps": 12220, "total_steps": 16324, "loss": 0.1972, "lr": 1.6309273599164963e-05, "epoch": 1.497136381734097, "percentage": 74.86, "elapsed_time": "14:08:29", "remaining_time": "4:44:57", "throughput": 1605.2, "total_tokens": 81719752}
{"current_steps": 12225, "total_steps": 16324, "loss": 0.2257, "lr": 1.627186707447309e-05, "epoch": 1.4977489204006003, "percentage": 74.89, "elapsed_time": "14:09:29", "remaining_time": "4:44:49", "throughput": 1603.97, "total_tokens": 81753160}
{"current_steps": 12230, "total_steps": 16324, "loss": 0.2207, "lr": 1.6234495157776585e-05, "epoch": 1.4983614590671035, "percentage": 74.92, "elapsed_time": "14:10:30", "remaining_time": "4:44:42", "throughput": 1602.7, "total_tokens": 81786432}
{"current_steps": 12235, "total_steps": 16324, "loss": 0.2238, "lr": 1.6197157887422298e-05, "epoch": 1.498973997733607, "percentage": 74.95, "elapsed_time": "14:11:29", "remaining_time": "4:44:34", "throughput": 1601.5, "total_tokens": 81820360}
{"current_steps": 12240, "total_steps": 16324, "loss": 0.2336, "lr": 1.615985530172148e-05, "epoch": 1.4995865364001102, "percentage": 74.98, "elapsed_time": "14:12:29", "remaining_time": "4:44:26", "throughput": 1600.28, "total_tokens": 81853752}
{"current_steps": 12245, "total_steps": 16324, "loss": 0.2765, "lr": 1.6122587438949832e-05, "epoch": 1.5001990750666137, "percentage": 75.01, "elapsed_time": "14:13:30", "remaining_time": "4:44:19", "throughput": 1599.02, "total_tokens": 81887064}
{"current_steps": 12250, "total_steps": 16324, "loss": 0.2553, "lr": 1.6085354337347447e-05, "epoch": 1.500811613733117, "percentage": 75.04, "elapsed_time": "14:14:31", "remaining_time": "4:44:11", "throughput": 1597.78, "total_tokens": 81919896}
{"current_steps": 12250, "total_steps": 16324, "eval_loss": 0.15666794776916504, "epoch": 1.500811613733117, "percentage": 75.04, "elapsed_time": "14:14:50", "remaining_time": "4:44:17", "throughput": 1597.16, "total_tokens": 81919896}
{"current_steps": 12255, "total_steps": 16324, "loss": 0.2293, "lr": 1.604815603511869e-05, "epoch": 1.5014241523996201, "percentage": 75.07, "elapsed_time": "14:15:56", "remaining_time": "4:44:11", "throughput": 1595.79, "total_tokens": 81953488}
{"current_steps": 12260, "total_steps": 16324, "loss": 0.1958, "lr": 1.6010992570432266e-05, "epoch": 1.5020366910661236, "percentage": 75.1, "elapsed_time": "14:16:56", "remaining_time": "4:44:03", "throughput": 1594.57, "total_tokens": 81987512}
{"current_steps": 12265, "total_steps": 16324, "loss": 0.2044, "lr": 1.5973863981421123e-05, "epoch": 1.5026492297326268, "percentage": 75.13, "elapsed_time": "14:17:56", "remaining_time": "4:43:55", "throughput": 1593.36, "total_tokens": 82021296}
{"current_steps": 12270, "total_steps": 16324, "loss": 0.2212, "lr": 1.5936770306182425e-05, "epoch": 1.5032617683991303, "percentage": 75.17, "elapsed_time": "14:18:56", "remaining_time": "4:43:47", "throughput": 1592.16, "total_tokens": 82054576}
{"current_steps": 12275, "total_steps": 16324, "loss": 0.2066, "lr": 1.589971158277749e-05, "epoch": 1.5038743070656335, "percentage": 75.2, "elapsed_time": "14:19:57", "remaining_time": "4:43:39", "throughput": 1590.95, "total_tokens": 82088824}
{"current_steps": 12280, "total_steps": 16324, "loss": 0.1794, "lr": 1.5862687849231805e-05, "epoch": 1.5044868457321368, "percentage": 75.23, "elapsed_time": "14:20:57", "remaining_time": "4:43:31", "throughput": 1589.76, "total_tokens": 82123328}
{"current_steps": 12285, "total_steps": 16324, "loss": 0.1734, "lr": 1.582569914353491e-05, "epoch": 1.5050993843986402, "percentage": 75.26, "elapsed_time": "14:21:57", "remaining_time": "4:43:23", "throughput": 1588.56, "total_tokens": 82156640}
{"current_steps": 12290, "total_steps": 16324, "loss": 0.2372, "lr": 1.5788745503640474e-05, "epoch": 1.5057119230651435, "percentage": 75.29, "elapsed_time": "14:22:58", "remaining_time": "4:43:15", "throughput": 1587.33, "total_tokens": 82189512}
{"current_steps": 12295, "total_steps": 16324, "loss": 0.2114, "lr": 1.57518269674661e-05, "epoch": 1.506324461731647, "percentage": 75.32, "elapsed_time": "14:23:59", "remaining_time": "4:43:07", "throughput": 1586.11, "total_tokens": 82222648}
{"current_steps": 12300, "total_steps": 16324, "loss": 0.19, "lr": 1.5714943572893433e-05, "epoch": 1.5069370003981502, "percentage": 75.35, "elapsed_time": "14:24:59", "remaining_time": "4:42:59", "throughput": 1584.93, "total_tokens": 82256360}
{"current_steps": 12300, "total_steps": 16324, "eval_loss": 0.13799481093883514, "epoch": 1.5069370003981502, "percentage": 75.35, "elapsed_time": "14:25:18", "remaining_time": "4:43:05", "throughput": 1584.33, "total_tokens": 82256360}
{"current_steps": 12305, "total_steps": 16324, "loss": 0.2132, "lr": 1.5678095357768056e-05, "epoch": 1.5075495390646534, "percentage": 75.38, "elapsed_time": "14:26:24", "remaining_time": "4:42:58", "throughput": 1582.96, "total_tokens": 82289496}
{"current_steps": 12310, "total_steps": 16324, "loss": 0.2384, "lr": 1.5641282359899413e-05, "epoch": 1.5081620777311566, "percentage": 75.41, "elapsed_time": "14:27:24", "remaining_time": "4:42:50", "throughput": 1581.76, "total_tokens": 82322464}
{"current_steps": 12315, "total_steps": 16324, "loss": 0.2558, "lr": 1.5604504617060843e-05, "epoch": 1.50877461639766, "percentage": 75.44, "elapsed_time": "14:28:24", "remaining_time": "4:42:42", "throughput": 1580.57, "total_tokens": 82355032}
{"current_steps": 12320, "total_steps": 16324, "loss": 0.237, "lr": 1.556776216698954e-05, "epoch": 1.5093871550641635, "percentage": 75.47, "elapsed_time": "14:29:24", "remaining_time": "4:42:33", "throughput": 1579.39, "total_tokens": 82388648}
{"current_steps": 12325, "total_steps": 16324, "loss": 0.1899, "lr": 1.5531055047386406e-05, "epoch": 1.5099996937306668, "percentage": 75.5, "elapsed_time": "14:30:25", "remaining_time": "4:42:25", "throughput": 1578.19, "total_tokens": 82422312}
{"current_steps": 12330, "total_steps": 16324, "loss": 0.1856, "lr": 1.5494383295916153e-05, "epoch": 1.51061223239717, "percentage": 75.53, "elapsed_time": "14:31:25", "remaining_time": "4:42:16", "throughput": 1577.03, "total_tokens": 82456152}
{"current_steps": 12335, "total_steps": 16324, "loss": 0.1833, "lr": 1.5457746950207186e-05, "epoch": 1.5112247710636733, "percentage": 75.56, "elapsed_time": "14:32:25", "remaining_time": "4:42:08", "throughput": 1575.86, "total_tokens": 82489776}
{"current_steps": 12340, "total_steps": 16324, "loss": 0.2428, "lr": 1.5421146047851598e-05, "epoch": 1.5118373097301767, "percentage": 75.59, "elapsed_time": "14:33:26", "remaining_time": "4:41:59", "throughput": 1574.66, "total_tokens": 82522720}
{"current_steps": 12345, "total_steps": 16324, "loss": 0.2209, "lr": 1.538458062640506e-05, "epoch": 1.5124498483966802, "percentage": 75.62, "elapsed_time": "14:34:26", "remaining_time": "4:41:50", "throughput": 1573.51, "total_tokens": 82555976}
{"current_steps": 12350, "total_steps": 16324, "loss": 0.2405, "lr": 1.5348050723386885e-05, "epoch": 1.5130623870631834, "percentage": 75.66, "elapsed_time": "14:35:26", "remaining_time": "4:41:41", "throughput": 1572.34, "total_tokens": 82589056}
{"current_steps": 12350, "total_steps": 16324, "eval_loss": 0.12636159360408783, "epoch": 1.5130623870631834, "percentage": 75.66, "elapsed_time": "14:35:45", "remaining_time": "4:41:48", "throughput": 1571.76, "total_tokens": 82589056}
{"current_steps": 12355, "total_steps": 16324, "loss": 0.2444, "lr": 1.5311556376279957e-05, "epoch": 1.5136749257296866, "percentage": 75.69, "elapsed_time": "14:36:52", "remaining_time": "4:41:41", "throughput": 1570.39, "total_tokens": 82621792}
{"current_steps": 12360, "total_steps": 16324, "loss": 0.2305, "lr": 1.5275097622530616e-05, "epoch": 1.5142874643961899, "percentage": 75.72, "elapsed_time": "14:37:52", "remaining_time": "4:41:32", "throughput": 1569.24, "total_tokens": 82655856}
{"current_steps": 12365, "total_steps": 16324, "loss": 0.21, "lr": 1.52386744995487e-05, "epoch": 1.5149000030626933, "percentage": 75.75, "elapsed_time": "14:38:53", "remaining_time": "4:41:23", "throughput": 1568.08, "total_tokens": 82689712}
{"current_steps": 12370, "total_steps": 16324, "loss": 0.2243, "lr": 1.5202287044707525e-05, "epoch": 1.5155125417291968, "percentage": 75.78, "elapsed_time": "14:39:54", "remaining_time": "4:41:15", "throughput": 1566.89, "total_tokens": 82723480}
{"current_steps": 12375, "total_steps": 16324, "loss": 0.2285, "lr": 1.516593529534378e-05, "epoch": 1.5161250803957, "percentage": 75.81, "elapsed_time": "14:40:54", "remaining_time": "4:41:06", "throughput": 1565.75, "total_tokens": 82756648}
{"current_steps": 12380, "total_steps": 16324, "loss": 0.2065, "lr": 1.5129619288757501e-05, "epoch": 1.5167376190622033, "percentage": 75.84, "elapsed_time": "14:41:54", "remaining_time": "4:40:57", "throughput": 1564.62, "total_tokens": 82790296}
{"current_steps": 12385, "total_steps": 16324, "loss": 0.2169, "lr": 1.5093339062212081e-05, "epoch": 1.5173501577287065, "percentage": 75.87, "elapsed_time": "14:42:55", "remaining_time": "4:40:48", "throughput": 1563.44, "total_tokens": 82823520}
{"current_steps": 12390, "total_steps": 16324, "loss": 0.2081, "lr": 1.5057094652934179e-05, "epoch": 1.51796269639521, "percentage": 75.9, "elapsed_time": "14:43:55", "remaining_time": "4:40:39", "throughput": 1562.29, "total_tokens": 82857288}
{"current_steps": 12395, "total_steps": 16324, "loss": 0.22, "lr": 1.5020886098113724e-05, "epoch": 1.5185752350617134, "percentage": 75.93, "elapsed_time": "14:44:56", "remaining_time": "4:40:30", "throughput": 1561.13, "total_tokens": 82890528}
{"current_steps": 12400, "total_steps": 16324, "loss": 0.2239, "lr": 1.4984713434903813e-05, "epoch": 1.5191877737282167, "percentage": 75.96, "elapsed_time": "14:45:57", "remaining_time": "4:40:21", "throughput": 1559.96, "total_tokens": 82923840}
{"current_steps": 12400, "total_steps": 16324, "eval_loss": 0.12071842700242996, "epoch": 1.5191877737282167, "percentage": 75.96, "elapsed_time": "14:46:17", "remaining_time": "4:40:28", "throughput": 1559.38, "total_tokens": 82923840}
{"current_steps": 12405, "total_steps": 16324, "loss": 0.1957, "lr": 1.4948576700420757e-05, "epoch": 1.51980031239472, "percentage": 75.99, "elapsed_time": "14:47:22", "remaining_time": "4:40:20", "throughput": 1558.1, "total_tokens": 82957936}
{"current_steps": 12410, "total_steps": 16324, "loss": 0.2119, "lr": 1.4912475931744002e-05, "epoch": 1.5204128510612231, "percentage": 76.02, "elapsed_time": "14:48:23", "remaining_time": "4:40:11", "throughput": 1556.98, "total_tokens": 82991576}
{"current_steps": 12415, "total_steps": 16324, "loss": 0.2164, "lr": 1.4876411165916043e-05, "epoch": 1.5210253897277266, "percentage": 76.05, "elapsed_time": "14:49:23", "remaining_time": "4:40:02", "throughput": 1555.82, "total_tokens": 83024640}
{"current_steps": 12420, "total_steps": 16324, "loss": 0.1959, "lr": 1.4840382439942474e-05, "epoch": 1.5216379283942298, "percentage": 76.08, "elapsed_time": "14:50:25", "remaining_time": "4:39:53", "throughput": 1554.66, "total_tokens": 83058376}
{"current_steps": 12425, "total_steps": 16324, "loss": 0.207, "lr": 1.4804389790791906e-05, "epoch": 1.5222504670607333, "percentage": 76.11, "elapsed_time": "14:51:25", "remaining_time": "4:39:43", "throughput": 1553.54, "total_tokens": 83091944}
{"current_steps": 12430, "total_steps": 16324, "loss": 0.2022, "lr": 1.4768433255395935e-05, "epoch": 1.5228630057272365, "percentage": 76.15, "elapsed_time": "14:52:26", "remaining_time": "4:39:34", "throughput": 1552.42, "total_tokens": 83126048}
{"current_steps": 12435, "total_steps": 16324, "loss": 0.2235, "lr": 1.4732512870649057e-05, "epoch": 1.5234755443937398, "percentage": 76.18, "elapsed_time": "14:53:27", "remaining_time": "4:39:25", "throughput": 1551.28, "total_tokens": 83159512}
{"current_steps": 12440, "total_steps": 16324, "loss": 0.1937, "lr": 1.4696628673408753e-05, "epoch": 1.524088083060243, "percentage": 76.21, "elapsed_time": "14:54:26", "remaining_time": "4:39:15", "throughput": 1550.19, "total_tokens": 83193544}
{"current_steps": 12445, "total_steps": 16324, "loss": 0.1757, "lr": 1.466078070049528e-05, "epoch": 1.5247006217267465, "percentage": 76.24, "elapsed_time": "14:55:27", "remaining_time": "4:39:06", "throughput": 1549.07, "total_tokens": 83227368}
{"current_steps": 12450, "total_steps": 16324, "loss": 0.2518, "lr": 1.4624968988691817e-05, "epoch": 1.52531316039325, "percentage": 76.27, "elapsed_time": "14:56:27", "remaining_time": "4:38:56", "throughput": 1547.94, "total_tokens": 83260088}
{"current_steps": 12450, "total_steps": 16324, "eval_loss": 0.09943088889122009, "epoch": 1.52531316039325, "percentage": 76.27, "elapsed_time": "14:56:47", "remaining_time": "4:39:02", "throughput": 1547.38, "total_tokens": 83260088}
{"current_steps": 12455, "total_steps": 16324, "loss": 0.1817, "lr": 1.4589193574744254e-05, "epoch": 1.5259256990597532, "percentage": 76.3, "elapsed_time": "14:57:52", "remaining_time": "4:38:54", "throughput": 1546.13, "total_tokens": 83293904}
{"current_steps": 12460, "total_steps": 16324, "loss": 0.214, "lr": 1.4553454495361291e-05, "epoch": 1.5265382377262564, "percentage": 76.33, "elapsed_time": "14:58:52", "remaining_time": "4:38:45", "throughput": 1545.02, "total_tokens": 83327216}
{"current_steps": 12465, "total_steps": 16324, "loss": 0.1842, "lr": 1.4517751787214345e-05, "epoch": 1.5271507763927596, "percentage": 76.36, "elapsed_time": "14:59:53", "remaining_time": "4:38:35", "throughput": 1543.91, "total_tokens": 83361192}
{"current_steps": 12470, "total_steps": 16324, "loss": 0.2159, "lr": 1.4482085486937475e-05, "epoch": 1.527763315059263, "percentage": 76.39, "elapsed_time": "15:00:53", "remaining_time": "4:38:25", "throughput": 1542.8, "total_tokens": 83394488}
{"current_steps": 12475, "total_steps": 16324, "loss": 0.2541, "lr": 1.4446455631127403e-05, "epoch": 1.5283758537257666, "percentage": 76.42, "elapsed_time": "15:01:53", "remaining_time": "4:38:16", "throughput": 1541.71, "total_tokens": 83427656}
{"current_steps": 12480, "total_steps": 16324, "loss": 0.2185, "lr": 1.4410862256343472e-05, "epoch": 1.5289883923922698, "percentage": 76.45, "elapsed_time": "15:02:54", "remaining_time": "4:38:06", "throughput": 1540.59, "total_tokens": 83461144}
{"current_steps": 12485, "total_steps": 16324, "loss": 0.268, "lr": 1.4375305399107586e-05, "epoch": 1.529600931058773, "percentage": 76.48, "elapsed_time": "15:03:54", "remaining_time": "4:37:56", "throughput": 1539.51, "total_tokens": 83494432}
{"current_steps": 12490, "total_steps": 16324, "loss": 0.1842, "lr": 1.433978509590414e-05, "epoch": 1.5302134697252763, "percentage": 76.51, "elapsed_time": "15:04:54", "remaining_time": "4:37:46", "throughput": 1538.43, "total_tokens": 83528280}
{"current_steps": 12495, "total_steps": 16324, "loss": 0.2401, "lr": 1.4304301383180074e-05, "epoch": 1.5308260083917797, "percentage": 76.54, "elapsed_time": "15:05:55", "remaining_time": "4:37:36", "throughput": 1537.32, "total_tokens": 83561440}
{"current_steps": 12500, "total_steps": 16324, "loss": 0.2383, "lr": 1.4268854297344764e-05, "epoch": 1.5314385470582832, "percentage": 76.57, "elapsed_time": "15:06:54", "remaining_time": "4:37:26", "throughput": 1536.25, "total_tokens": 83595024}
{"current_steps": 12500, "total_steps": 16324, "eval_loss": 0.07955755293369293, "epoch": 1.5314385470582832, "percentage": 76.57, "elapsed_time": "15:07:14", "remaining_time": "4:37:32", "throughput": 1535.7, "total_tokens": 83595024}
{"current_steps": 12505, "total_steps": 16324, "loss": 0.2208, "lr": 1.423344387476998e-05, "epoch": 1.5320510857247864, "percentage": 76.6, "elapsed_time": "15:08:19", "remaining_time": "4:37:24", "throughput": 1534.47, "total_tokens": 83628592}
{"current_steps": 12510, "total_steps": 16324, "loss": 0.2076, "lr": 1.4198070151789905e-05, "epoch": 1.5326636243912897, "percentage": 76.64, "elapsed_time": "15:09:20", "remaining_time": "4:37:14", "throughput": 1533.4, "total_tokens": 83662424}
{"current_steps": 12515, "total_steps": 16324, "loss": 0.1989, "lr": 1.4162733164701063e-05, "epoch": 1.533276163057793, "percentage": 76.67, "elapsed_time": "15:10:19", "remaining_time": "4:37:03", "throughput": 1532.34, "total_tokens": 83696256}
{"current_steps": 12520, "total_steps": 16324, "loss": 0.2344, "lr": 1.4127432949762254e-05, "epoch": 1.5338887017242964, "percentage": 76.7, "elapsed_time": "15:11:19", "remaining_time": "4:36:53", "throughput": 1531.28, "total_tokens": 83730416}
{"current_steps": 12525, "total_steps": 16324, "loss": 0.2036, "lr": 1.409216954319459e-05, "epoch": 1.5345012403907998, "percentage": 76.73, "elapsed_time": "15:12:20", "remaining_time": "4:36:43", "throughput": 1530.21, "total_tokens": 83764872}
{"current_steps": 12530, "total_steps": 16324, "loss": 0.1949, "lr": 1.4056942981181354e-05, "epoch": 1.535113779057303, "percentage": 76.76, "elapsed_time": "15:13:20", "remaining_time": "4:36:33", "throughput": 1529.15, "total_tokens": 83798928}
{"current_steps": 12535, "total_steps": 16324, "loss": 0.2276, "lr": 1.4021753299868101e-05, "epoch": 1.5357263177238063, "percentage": 76.79, "elapsed_time": "15:14:20", "remaining_time": "4:36:22", "throughput": 1528.08, "total_tokens": 83831792}
{"current_steps": 12540, "total_steps": 16324, "loss": 0.2339, "lr": 1.3986600535362466e-05, "epoch": 1.5363388563903095, "percentage": 76.82, "elapsed_time": "15:15:21", "remaining_time": "4:36:12", "throughput": 1527.0, "total_tokens": 83864704}
{"current_steps": 12545, "total_steps": 16324, "loss": 0.1957, "lr": 1.3951484723734254e-05, "epoch": 1.536951395056813, "percentage": 76.85, "elapsed_time": "15:16:22", "remaining_time": "4:36:02", "throughput": 1525.92, "total_tokens": 83898496}
{"current_steps": 12550, "total_steps": 16324, "loss": 0.2197, "lr": 1.3916405901015344e-05, "epoch": 1.5375639337233162, "percentage": 76.88, "elapsed_time": "15:17:22", "remaining_time": "4:35:52", "throughput": 1524.87, "total_tokens": 83932352}
{"current_steps": 12550, "total_steps": 16324, "eval_loss": 0.13997627794742584, "epoch": 1.5375639337233162, "percentage": 76.88, "elapsed_time": "15:17:41", "remaining_time": "4:35:58", "throughput": 1524.33, "total_tokens": 83932352}
{"current_steps": 12555, "total_steps": 16324, "loss": 0.2191, "lr": 1.3881364103199667e-05, "epoch": 1.5381764723898197, "percentage": 76.91, "elapsed_time": "15:18:48", "remaining_time": "4:35:49", "throughput": 1523.09, "total_tokens": 83965416}
{"current_steps": 12560, "total_steps": 16324, "loss": 0.2528, "lr": 1.3846359366243128e-05, "epoch": 1.538789011056323, "percentage": 76.94, "elapsed_time": "15:19:48", "remaining_time": "4:35:39", "throughput": 1522.01, "total_tokens": 83998008}
{"current_steps": 12565, "total_steps": 16324, "loss": 0.179, "lr": 1.3811391726063639e-05, "epoch": 1.5394015497228262, "percentage": 76.97, "elapsed_time": "15:20:48", "remaining_time": "4:35:28", "throughput": 1520.96, "total_tokens": 84031432}
{"current_steps": 12570, "total_steps": 16324, "loss": 0.2365, "lr": 1.3776461218541054e-05, "epoch": 1.5400140883893296, "percentage": 77.0, "elapsed_time": "15:21:49", "remaining_time": "4:35:18", "throughput": 1519.88, "total_tokens": 84064192}
{"current_steps": 12575, "total_steps": 16324, "loss": 0.2428, "lr": 1.3741567879517081e-05, "epoch": 1.5406266270558329, "percentage": 77.03, "elapsed_time": "15:22:50", "remaining_time": "4:35:07", "throughput": 1518.82, "total_tokens": 84097448}
{"current_steps": 12580, "total_steps": 16324, "loss": 0.2054, "lr": 1.370671174479533e-05, "epoch": 1.5412391657223363, "percentage": 77.06, "elapsed_time": "15:23:49", "remaining_time": "4:34:56", "throughput": 1517.81, "total_tokens": 84131400}
{"current_steps": 12585, "total_steps": 16324, "loss": 0.2132, "lr": 1.3671892850141226e-05, "epoch": 1.5418517043888396, "percentage": 77.1, "elapsed_time": "15:24:49", "remaining_time": "4:34:46", "throughput": 1516.76, "total_tokens": 84164672}
{"current_steps": 12590, "total_steps": 16324, "loss": 0.2051, "lr": 1.3637111231282001e-05, "epoch": 1.5424642430553428, "percentage": 77.13, "elapsed_time": "15:25:50", "remaining_time": "4:34:35", "throughput": 1515.71, "total_tokens": 84198624}
{"current_steps": 12595, "total_steps": 16324, "loss": 0.2517, "lr": 1.3602366923906578e-05, "epoch": 1.543076781721846, "percentage": 77.16, "elapsed_time": "15:26:50", "remaining_time": "4:34:24", "throughput": 1514.68, "total_tokens": 84232112}
{"current_steps": 12600, "total_steps": 16324, "loss": 0.2038, "lr": 1.3567659963665657e-05, "epoch": 1.5436893203883495, "percentage": 77.19, "elapsed_time": "15:27:50", "remaining_time": "4:34:13", "throughput": 1513.65, "total_tokens": 84265664}
{"current_steps": 12600, "total_steps": 16324, "eval_loss": 0.0905894786119461, "epoch": 1.5436893203883495, "percentage": 77.19, "elapsed_time": "15:28:10", "remaining_time": "4:34:19", "throughput": 1513.11, "total_tokens": 84265664}
{"current_steps": 12605, "total_steps": 16324, "loss": 0.2199, "lr": 1.3532990386171608e-05, "epoch": 1.544301859054853, "percentage": 77.22, "elapsed_time": "15:29:16", "remaining_time": "4:34:10", "throughput": 1511.9, "total_tokens": 84298840}
{"current_steps": 12610, "total_steps": 16324, "loss": 0.2262, "lr": 1.3498358226998414e-05, "epoch": 1.5449143977213562, "percentage": 77.25, "elapsed_time": "15:30:16", "remaining_time": "4:33:59", "throughput": 1510.88, "total_tokens": 84332296}
{"current_steps": 12615, "total_steps": 16324, "loss": 0.2246, "lr": 1.3463763521681672e-05, "epoch": 1.5455269363878594, "percentage": 77.28, "elapsed_time": "15:31:16", "remaining_time": "4:33:48", "throughput": 1509.86, "total_tokens": 84366136}
{"current_steps": 12620, "total_steps": 16324, "loss": 0.252, "lr": 1.3429206305718566e-05, "epoch": 1.5461394750543627, "percentage": 77.31, "elapsed_time": "15:32:18", "remaining_time": "4:33:37", "throughput": 1508.81, "total_tokens": 84399600}
{"current_steps": 12625, "total_steps": 16324, "loss": 0.2128, "lr": 1.3394686614567813e-05, "epoch": 1.5467520137208661, "percentage": 77.34, "elapsed_time": "15:33:17", "remaining_time": "4:33:26", "throughput": 1507.79, "total_tokens": 84432544}
{"current_steps": 12630, "total_steps": 16324, "loss": 0.2009, "lr": 1.3360204483649596e-05, "epoch": 1.5473645523873696, "percentage": 77.37, "elapsed_time": "15:34:18", "remaining_time": "4:33:15", "throughput": 1506.75, "total_tokens": 84465648}
{"current_steps": 12635, "total_steps": 16324, "loss": 0.2205, "lr": 1.3325759948345584e-05, "epoch": 1.5479770910538728, "percentage": 77.4, "elapsed_time": "15:35:18", "remaining_time": "4:33:04", "throughput": 1505.72, "total_tokens": 84499192}
{"current_steps": 12640, "total_steps": 16324, "loss": 0.209, "lr": 1.3291353043998872e-05, "epoch": 1.548589629720376, "percentage": 77.43, "elapsed_time": "15:36:19", "remaining_time": "4:32:53", "throughput": 1504.7, "total_tokens": 84533112}
{"current_steps": 12645, "total_steps": 16324, "loss": 0.2236, "lr": 1.3256983805913937e-05, "epoch": 1.5492021683868793, "percentage": 77.46, "elapsed_time": "15:37:19", "remaining_time": "4:32:42", "throughput": 1503.68, "total_tokens": 84566368}
{"current_steps": 12650, "total_steps": 16324, "loss": 0.218, "lr": 1.3222652269356584e-05, "epoch": 1.5498147070533828, "percentage": 77.49, "elapsed_time": "15:38:20", "remaining_time": "4:32:31", "throughput": 1502.66, "total_tokens": 84599960}
{"current_steps": 12650, "total_steps": 16324, "eval_loss": 0.08059627562761307, "epoch": 1.5498147070533828, "percentage": 77.49, "elapsed_time": "15:38:40", "remaining_time": "4:32:37", "throughput": 1502.12, "total_tokens": 84599960}
{"current_steps": 12655, "total_steps": 16324, "loss": 0.2095, "lr": 1.3188358469553964e-05, "epoch": 1.5504272457198862, "percentage": 77.52, "elapsed_time": "15:39:46", "remaining_time": "4:32:27", "throughput": 1500.95, "total_tokens": 84633040}
{"current_steps": 12660, "total_steps": 16324, "loss": 0.2304, "lr": 1.3154102441694511e-05, "epoch": 1.5510397843863895, "percentage": 77.55, "elapsed_time": "15:40:46", "remaining_time": "4:32:16", "throughput": 1499.96, "total_tokens": 84666880}
{"current_steps": 12665, "total_steps": 16324, "loss": 0.2087, "lr": 1.3119884220927859e-05, "epoch": 1.5516523230528927, "percentage": 77.59, "elapsed_time": "15:41:46", "remaining_time": "4:32:05", "throughput": 1498.96, "total_tokens": 84700520}
{"current_steps": 12670, "total_steps": 16324, "loss": 0.2182, "lr": 1.3085703842364888e-05, "epoch": 1.552264861719396, "percentage": 77.62, "elapsed_time": "15:42:46", "remaining_time": "4:31:53", "throughput": 1497.96, "total_tokens": 84733512}
{"current_steps": 12675, "total_steps": 16324, "loss": 0.2096, "lr": 1.3051561341077651e-05, "epoch": 1.5528774003858994, "percentage": 77.65, "elapsed_time": "15:43:45", "remaining_time": "4:31:41", "throughput": 1496.98, "total_tokens": 84767024}
{"current_steps": 12680, "total_steps": 16324, "loss": 0.1873, "lr": 1.3017456752099294e-05, "epoch": 1.5534899390524028, "percentage": 77.68, "elapsed_time": "15:44:45", "remaining_time": "4:31:30", "throughput": 1495.99, "total_tokens": 84800840}
{"current_steps": 12685, "total_steps": 16324, "loss": 0.2297, "lr": 1.298339011042412e-05, "epoch": 1.554102477718906, "percentage": 77.71, "elapsed_time": "15:45:45", "remaining_time": "4:31:18", "throughput": 1494.99, "total_tokens": 84833904}
{"current_steps": 12690, "total_steps": 16324, "loss": 0.1763, "lr": 1.294936145100743e-05, "epoch": 1.5547150163854093, "percentage": 77.74, "elapsed_time": "15:46:44", "remaining_time": "4:31:07", "throughput": 1494.03, "total_tokens": 84868168}
{"current_steps": 12695, "total_steps": 16324, "loss": 0.2232, "lr": 1.2915370808765614e-05, "epoch": 1.5553275550519126, "percentage": 77.77, "elapsed_time": "15:47:45", "remaining_time": "4:30:55", "throughput": 1493.03, "total_tokens": 84901960}
{"current_steps": 12700, "total_steps": 16324, "loss": 0.1887, "lr": 1.2881418218576003e-05, "epoch": 1.555940093718416, "percentage": 77.8, "elapsed_time": "15:48:45", "remaining_time": "4:30:44", "throughput": 1492.04, "total_tokens": 84935752}
{"current_steps": 12700, "total_steps": 16324, "eval_loss": 0.0622766874730587, "epoch": 1.555940093718416, "percentage": 77.8, "elapsed_time": "15:49:05", "remaining_time": "4:30:49", "throughput": 1491.52, "total_tokens": 84935752}
{"current_steps": 12705, "total_steps": 16324, "loss": 0.1813, "lr": 1.284750371527691e-05, "epoch": 1.5565526323849193, "percentage": 77.83, "elapsed_time": "15:50:10", "remaining_time": "4:30:39", "throughput": 1490.43, "total_tokens": 84969792}
{"current_steps": 12710, "total_steps": 16324, "loss": 0.2106, "lr": 1.2813627333667572e-05, "epoch": 1.5571651710514227, "percentage": 77.86, "elapsed_time": "15:51:10", "remaining_time": "4:30:27", "throughput": 1489.46, "total_tokens": 85003840}
{"current_steps": 12715, "total_steps": 16324, "loss": 0.2457, "lr": 1.2779789108508111e-05, "epoch": 1.557777709717926, "percentage": 77.89, "elapsed_time": "15:52:10", "remaining_time": "4:30:15", "throughput": 1488.46, "total_tokens": 85036840}
{"current_steps": 12720, "total_steps": 16324, "loss": 0.1992, "lr": 1.2745989074519454e-05, "epoch": 1.5583902483844292, "percentage": 77.92, "elapsed_time": "15:53:11", "remaining_time": "4:30:04", "throughput": 1487.49, "total_tokens": 85070776}
{"current_steps": 12725, "total_steps": 16324, "loss": 0.1745, "lr": 1.2712227266383392e-05, "epoch": 1.5590027870509324, "percentage": 77.95, "elapsed_time": "15:54:11", "remaining_time": "4:29:52", "throughput": 1486.51, "total_tokens": 85105048}
{"current_steps": 12730, "total_steps": 16324, "loss": 0.2094, "lr": 1.2678503718742491e-05, "epoch": 1.5596153257174359, "percentage": 77.98, "elapsed_time": "15:55:12", "remaining_time": "4:29:40", "throughput": 1485.5, "total_tokens": 85137864}
{"current_steps": 12735, "total_steps": 16324, "loss": 0.2083, "lr": 1.2644818466200004e-05, "epoch": 1.5602278643839393, "percentage": 78.01, "elapsed_time": "15:56:12", "remaining_time": "4:29:28", "throughput": 1484.52, "total_tokens": 85171312}
{"current_steps": 12740, "total_steps": 16324, "loss": 0.1899, "lr": 1.2611171543319944e-05, "epoch": 1.5608404030504426, "percentage": 78.04, "elapsed_time": "15:57:13", "remaining_time": "4:29:17", "throughput": 1483.55, "total_tokens": 85205072}
{"current_steps": 12745, "total_steps": 16324, "loss": 0.234, "lr": 1.2577562984626984e-05, "epoch": 1.5614529417169458, "percentage": 78.08, "elapsed_time": "15:58:14", "remaining_time": "4:29:05", "throughput": 1482.55, "total_tokens": 85238256}
{"current_steps": 12750, "total_steps": 16324, "loss": 0.2372, "lr": 1.2543992824606437e-05, "epoch": 1.562065480383449, "percentage": 78.11, "elapsed_time": "15:59:15", "remaining_time": "4:28:53", "throughput": 1481.57, "total_tokens": 85271648}
{"current_steps": 12750, "total_steps": 16324, "eval_loss": 0.11330395191907883, "epoch": 1.562065480383449, "percentage": 78.11, "elapsed_time": "15:59:34", "remaining_time": "4:28:59", "throughput": 1481.05, "total_tokens": 85271648}
{"current_steps": 12755, "total_steps": 16324, "loss": 0.2537, "lr": 1.251046109770418e-05, "epoch": 1.5626780190499525, "percentage": 78.14, "elapsed_time": "16:00:40", "remaining_time": "4:28:48", "throughput": 1479.95, "total_tokens": 85305136}
{"current_steps": 12760, "total_steps": 16324, "loss": 0.2563, "lr": 1.2476967838326681e-05, "epoch": 1.563290557716456, "percentage": 78.17, "elapsed_time": "16:01:41", "remaining_time": "4:28:36", "throughput": 1478.95, "total_tokens": 85338344}
{"current_steps": 12765, "total_steps": 16324, "loss": 0.2139, "lr": 1.2443513080840958e-05, "epoch": 1.5639030963829592, "percentage": 78.2, "elapsed_time": "16:02:41", "remaining_time": "4:28:24", "throughput": 1478.01, "total_tokens": 85372272}
{"current_steps": 12770, "total_steps": 16324, "loss": 0.224, "lr": 1.2410096859574489e-05, "epoch": 1.5645156350494624, "percentage": 78.23, "elapsed_time": "16:03:41", "remaining_time": "4:28:12", "throughput": 1477.06, "total_tokens": 85405312}
{"current_steps": 12775, "total_steps": 16324, "loss": 0.2125, "lr": 1.237671920881519e-05, "epoch": 1.5651281737159657, "percentage": 78.26, "elapsed_time": "16:04:41", "remaining_time": "4:27:59", "throughput": 1476.11, "total_tokens": 85439168}
{"current_steps": 12780, "total_steps": 16324, "loss": 0.2469, "lr": 1.2343380162811469e-05, "epoch": 1.5657407123824691, "percentage": 78.29, "elapsed_time": "16:05:41", "remaining_time": "4:27:47", "throughput": 1475.16, "total_tokens": 85472104}
{"current_steps": 12785, "total_steps": 16324, "loss": 0.2053, "lr": 1.2310079755772081e-05, "epoch": 1.5663532510489726, "percentage": 78.32, "elapsed_time": "16:06:40", "remaining_time": "4:27:35", "throughput": 1474.23, "total_tokens": 85506432}
{"current_steps": 12790, "total_steps": 16324, "loss": 0.2374, "lr": 1.227681802186611e-05, "epoch": 1.5669657897154758, "percentage": 78.35, "elapsed_time": "16:07:40", "remaining_time": "4:27:22", "throughput": 1473.28, "total_tokens": 85539792}
{"current_steps": 12795, "total_steps": 16324, "loss": 0.234, "lr": 1.2243594995223007e-05, "epoch": 1.567578328381979, "percentage": 78.38, "elapsed_time": "16:08:40", "remaining_time": "4:27:10", "throughput": 1472.35, "total_tokens": 85573104}
{"current_steps": 12800, "total_steps": 16324, "loss": 0.1879, "lr": 1.2210410709932479e-05, "epoch": 1.5681908670484823, "percentage": 78.41, "elapsed_time": "16:09:39", "remaining_time": "4:26:57", "throughput": 1471.43, "total_tokens": 85607024}
{"current_steps": 12800, "total_steps": 16324, "eval_loss": 0.09048807621002197, "epoch": 1.5681908670484823, "percentage": 78.41, "elapsed_time": "16:09:59", "remaining_time": "4:27:03", "throughput": 1470.93, "total_tokens": 85607024}
{"current_steps": 12805, "total_steps": 16324, "loss": 0.2375, "lr": 1.2177265200044507e-05, "epoch": 1.5688034057149858, "percentage": 78.44, "elapsed_time": "16:11:04", "remaining_time": "4:26:51", "throughput": 1469.86, "total_tokens": 85640488}
{"current_steps": 12810, "total_steps": 16324, "loss": 0.2367, "lr": 1.214415849956923e-05, "epoch": 1.5694159443814892, "percentage": 78.47, "elapsed_time": "16:12:04", "remaining_time": "4:26:39", "throughput": 1468.92, "total_tokens": 85673952}
{"current_steps": 12815, "total_steps": 16324, "loss": 0.2035, "lr": 1.2111090642477019e-05, "epoch": 1.5700284830479925, "percentage": 78.5, "elapsed_time": "16:13:03", "remaining_time": "4:26:26", "throughput": 1468.01, "total_tokens": 85707752}
{"current_steps": 12820, "total_steps": 16324, "loss": 0.2531, "lr": 1.2078061662698375e-05, "epoch": 1.5706410217144957, "percentage": 78.53, "elapsed_time": "16:14:03", "remaining_time": "4:26:14", "throughput": 1467.06, "total_tokens": 85740768}
{"current_steps": 12825, "total_steps": 16324, "loss": 0.1939, "lr": 1.204507159412388e-05, "epoch": 1.571253560380999, "percentage": 78.57, "elapsed_time": "16:15:04", "remaining_time": "4:26:01", "throughput": 1466.12, "total_tokens": 85774488}
{"current_steps": 12830, "total_steps": 16324, "loss": 0.2063, "lr": 1.2012120470604227e-05, "epoch": 1.5718660990475024, "percentage": 78.6, "elapsed_time": "16:16:04", "remaining_time": "4:25:48", "throughput": 1465.2, "total_tokens": 85808448}
{"current_steps": 12835, "total_steps": 16324, "loss": 0.1732, "lr": 1.1979208325950142e-05, "epoch": 1.5724786377140056, "percentage": 78.63, "elapsed_time": "16:17:04", "remaining_time": "4:25:36", "throughput": 1464.27, "total_tokens": 85842184}
{"current_steps": 12840, "total_steps": 16324, "loss": 0.1844, "lr": 1.1946335193932312e-05, "epoch": 1.573091176380509, "percentage": 78.66, "elapsed_time": "16:18:04", "remaining_time": "4:25:23", "throughput": 1463.34, "total_tokens": 85875920}
{"current_steps": 12845, "total_steps": 16324, "loss": 0.2313, "lr": 1.1913501108281466e-05, "epoch": 1.5737037150470123, "percentage": 78.69, "elapsed_time": "16:19:04", "remaining_time": "4:25:10", "throughput": 1462.42, "total_tokens": 85909184}
{"current_steps": 12850, "total_steps": 16324, "loss": 0.2275, "lr": 1.1880706102688199e-05, "epoch": 1.5743162537135156, "percentage": 78.72, "elapsed_time": "16:20:05", "remaining_time": "4:24:57", "throughput": 1461.47, "total_tokens": 85942072}
{"current_steps": 12850, "total_steps": 16324, "eval_loss": 0.15477542579174042, "epoch": 1.5743162537135156, "percentage": 78.72, "elapsed_time": "16:20:24", "remaining_time": "4:25:03", "throughput": 1460.98, "total_tokens": 85942072}
{"current_steps": 12855, "total_steps": 16324, "loss": 0.1994, "lr": 1.1847950210803043e-05, "epoch": 1.574928792380019, "percentage": 78.75, "elapsed_time": "16:21:31", "remaining_time": "4:24:52", "throughput": 1459.91, "total_tokens": 85975824}
{"current_steps": 12860, "total_steps": 16324, "loss": 0.1829, "lr": 1.1815233466236415e-05, "epoch": 1.5755413310465223, "percentage": 78.78, "elapsed_time": "16:22:31", "remaining_time": "4:24:39", "throughput": 1458.99, "total_tokens": 86009872}
{"current_steps": 12865, "total_steps": 16324, "loss": 0.2332, "lr": 1.1782555902558495e-05, "epoch": 1.5761538697130257, "percentage": 78.81, "elapsed_time": "16:23:31", "remaining_time": "4:24:26", "throughput": 1458.08, "total_tokens": 86043408}
{"current_steps": 12870, "total_steps": 16324, "loss": 0.1934, "lr": 1.1749917553299333e-05, "epoch": 1.576766408379529, "percentage": 78.84, "elapsed_time": "16:24:31", "remaining_time": "4:24:13", "throughput": 1457.15, "total_tokens": 86076872}
{"current_steps": 12875, "total_steps": 16324, "loss": 0.2071, "lr": 1.1717318451948716e-05, "epoch": 1.5773789470460322, "percentage": 78.87, "elapsed_time": "16:25:31", "remaining_time": "4:24:00", "throughput": 1456.26, "total_tokens": 86110216}
{"current_steps": 12880, "total_steps": 16324, "loss": 0.2189, "lr": 1.1684758631956127e-05, "epoch": 1.5779914857125354, "percentage": 78.9, "elapsed_time": "16:26:30", "remaining_time": "4:23:47", "throughput": 1455.37, "total_tokens": 86144000}
{"current_steps": 12885, "total_steps": 16324, "loss": 0.1942, "lr": 1.1652238126730792e-05, "epoch": 1.578604024379039, "percentage": 78.93, "elapsed_time": "16:27:31", "remaining_time": "4:23:34", "throughput": 1454.45, "total_tokens": 86177784}
{"current_steps": 12890, "total_steps": 16324, "loss": 0.1607, "lr": 1.1619756969641583e-05, "epoch": 1.5792165630455424, "percentage": 78.96, "elapsed_time": "16:28:30", "remaining_time": "4:23:20", "throughput": 1453.58, "total_tokens": 86212072}
{"current_steps": 12895, "total_steps": 16324, "loss": 0.1805, "lr": 1.1587315194016957e-05, "epoch": 1.5798291017120456, "percentage": 78.99, "elapsed_time": "16:29:29", "remaining_time": "4:23:07", "throughput": 1452.7, "total_tokens": 86246240}
{"current_steps": 12900, "total_steps": 16324, "loss": 0.2114, "lr": 1.155491283314502e-05, "epoch": 1.5804416403785488, "percentage": 79.02, "elapsed_time": "16:30:28", "remaining_time": "4:22:53", "throughput": 1451.83, "total_tokens": 86280424}
{"current_steps": 12900, "total_steps": 16324, "eval_loss": 0.1280047595500946, "epoch": 1.5804416403785488, "percentage": 79.02, "elapsed_time": "16:30:48", "remaining_time": "4:22:59", "throughput": 1451.35, "total_tokens": 86280424}
{"current_steps": 12905, "total_steps": 16324, "loss": 0.185, "lr": 1.1522549920273407e-05, "epoch": 1.581054179045052, "percentage": 79.06, "elapsed_time": "16:31:52", "remaining_time": "4:22:47", "throughput": 1450.35, "total_tokens": 86314360}
{"current_steps": 12910, "total_steps": 16324, "loss": 0.2093, "lr": 1.149022648860929e-05, "epoch": 1.5816667177115555, "percentage": 79.09, "elapsed_time": "16:32:52", "remaining_time": "4:22:33", "throughput": 1449.46, "total_tokens": 86347904}
{"current_steps": 12915, "total_steps": 16324, "loss": 0.2327, "lr": 1.1457942571319292e-05, "epoch": 1.582279256378059, "percentage": 79.12, "elapsed_time": "16:33:52", "remaining_time": "4:22:20", "throughput": 1448.57, "total_tokens": 86381240}
{"current_steps": 12920, "total_steps": 16324, "loss": 0.1798, "lr": 1.1425698201529527e-05, "epoch": 1.5828917950445622, "percentage": 79.15, "elapsed_time": "16:34:51", "remaining_time": "4:22:06", "throughput": 1447.69, "total_tokens": 86415248}
{"current_steps": 12925, "total_steps": 16324, "loss": 0.207, "lr": 1.1393493412325545e-05, "epoch": 1.5835043337110655, "percentage": 79.18, "elapsed_time": "16:35:51", "remaining_time": "4:21:53", "throughput": 1446.81, "total_tokens": 86449200}
{"current_steps": 12930, "total_steps": 16324, "loss": 0.1878, "lr": 1.1361328236752238e-05, "epoch": 1.5841168723775687, "percentage": 79.21, "elapsed_time": "16:36:50", "remaining_time": "4:21:39", "throughput": 1445.94, "total_tokens": 86483376}
{"current_steps": 12935, "total_steps": 16324, "loss": 0.1721, "lr": 1.1329202707813857e-05, "epoch": 1.5847294110440722, "percentage": 79.24, "elapsed_time": "16:37:51", "remaining_time": "4:21:26", "throughput": 1445.05, "total_tokens": 86517304}
{"current_steps": 12940, "total_steps": 16324, "loss": 0.214, "lr": 1.1297116858474e-05, "epoch": 1.5853419497105756, "percentage": 79.27, "elapsed_time": "16:38:50", "remaining_time": "4:21:12", "throughput": 1444.16, "total_tokens": 86550072}
{"current_steps": 12945, "total_steps": 16324, "loss": 0.2054, "lr": 1.1265070721655552e-05, "epoch": 1.5859544883770789, "percentage": 79.3, "elapsed_time": "16:39:50", "remaining_time": "4:20:59", "throughput": 1443.28, "total_tokens": 86583616}
{"current_steps": 12950, "total_steps": 16324, "loss": 0.2206, "lr": 1.1233064330240607e-05, "epoch": 1.586567027043582, "percentage": 79.33, "elapsed_time": "16:40:51", "remaining_time": "4:20:45", "throughput": 1442.37, "total_tokens": 86616808}
{"current_steps": 12950, "total_steps": 16324, "eval_loss": 0.1442350596189499, "epoch": 1.586567027043582, "percentage": 79.33, "elapsed_time": "16:41:11", "remaining_time": "4:20:51", "throughput": 1441.89, "total_tokens": 86616808}
{"current_steps": 12955, "total_steps": 16324, "loss": 0.1998, "lr": 1.1201097717070514e-05, "epoch": 1.5871795657100853, "percentage": 79.36, "elapsed_time": "16:42:16", "remaining_time": "4:20:38", "throughput": 1440.89, "total_tokens": 86650088}
{"current_steps": 12960, "total_steps": 16324, "loss": 0.2073, "lr": 1.1169170914945799e-05, "epoch": 1.5877921043765888, "percentage": 79.39, "elapsed_time": "16:43:16", "remaining_time": "4:20:25", "throughput": 1440.01, "total_tokens": 86683688}
{"current_steps": 12965, "total_steps": 16324, "loss": 0.2252, "lr": 1.1137283956626154e-05, "epoch": 1.5884046430430923, "percentage": 79.42, "elapsed_time": "16:44:16", "remaining_time": "4:20:11", "throughput": 1439.14, "total_tokens": 86717248}
{"current_steps": 12970, "total_steps": 16324, "loss": 0.1592, "lr": 1.1105436874830333e-05, "epoch": 1.5890171817095955, "percentage": 79.45, "elapsed_time": "16:45:16", "remaining_time": "4:19:57", "throughput": 1438.28, "total_tokens": 86751584}
{"current_steps": 12975, "total_steps": 16324, "loss": 0.2505, "lr": 1.1073629702236227e-05, "epoch": 1.5896297203760987, "percentage": 79.48, "elapsed_time": "16:46:16", "remaining_time": "4:19:43", "throughput": 1437.38, "total_tokens": 86784104}
{"current_steps": 12980, "total_steps": 16324, "loss": 0.2184, "lr": 1.1041862471480774e-05, "epoch": 1.590242259042602, "percentage": 79.51, "elapsed_time": "16:47:16", "remaining_time": "4:19:30", "throughput": 1436.5, "total_tokens": 86817544}
{"current_steps": 12985, "total_steps": 16324, "loss": 0.2064, "lr": 1.1010135215159883e-05, "epoch": 1.5908547977091054, "percentage": 79.55, "elapsed_time": "16:48:16", "remaining_time": "4:19:16", "throughput": 1435.65, "total_tokens": 86851848}
{"current_steps": 12990, "total_steps": 16324, "loss": 0.2336, "lr": 1.0978447965828498e-05, "epoch": 1.5914673363756087, "percentage": 79.58, "elapsed_time": "16:49:16", "remaining_time": "4:19:02", "throughput": 1434.78, "total_tokens": 86885200}
{"current_steps": 12995, "total_steps": 16324, "loss": 0.2132, "lr": 1.0946800756000492e-05, "epoch": 1.5920798750421121, "percentage": 79.61, "elapsed_time": "16:50:16", "remaining_time": "4:18:48", "throughput": 1433.92, "total_tokens": 86918344}
{"current_steps": 13000, "total_steps": 16324, "loss": 0.1829, "lr": 1.0915193618148628e-05, "epoch": 1.5926924137086154, "percentage": 79.64, "elapsed_time": "16:51:15", "remaining_time": "4:18:34", "throughput": 1433.07, "total_tokens": 86952144}
{"current_steps": 13000, "total_steps": 16324, "eval_loss": 0.17378541827201843, "epoch": 1.5926924137086154, "percentage": 79.64, "elapsed_time": "16:51:35", "remaining_time": "4:18:39", "throughput": 1432.6, "total_tokens": 86952144}
{"current_steps": 13005, "total_steps": 16324, "loss": 0.2113, "lr": 1.0883626584704599e-05, "epoch": 1.5933049523751186, "percentage": 79.67, "elapsed_time": "16:52:41", "remaining_time": "4:18:26", "throughput": 1431.61, "total_tokens": 86986432}
{"current_steps": 13010, "total_steps": 16324, "loss": 0.2341, "lr": 1.085209968805893e-05, "epoch": 1.5939174910416218, "percentage": 79.7, "elapsed_time": "16:53:40", "remaining_time": "4:18:12", "throughput": 1430.76, "total_tokens": 87019888}
{"current_steps": 13015, "total_steps": 16324, "loss": 0.1944, "lr": 1.0820612960560928e-05, "epoch": 1.5945300297081253, "percentage": 79.73, "elapsed_time": "16:54:40", "remaining_time": "4:17:58", "throughput": 1429.92, "total_tokens": 87054224}
{"current_steps": 13020, "total_steps": 16324, "loss": 0.1609, "lr": 1.078916643451875e-05, "epoch": 1.5951425683746288, "percentage": 79.76, "elapsed_time": "16:55:39", "remaining_time": "4:17:44", "throughput": 1429.1, "total_tokens": 87088584}
{"current_steps": 13025, "total_steps": 16324, "loss": 0.2363, "lr": 1.075776014219922e-05, "epoch": 1.595755107041132, "percentage": 79.79, "elapsed_time": "16:56:39", "remaining_time": "4:17:30", "throughput": 1428.25, "total_tokens": 87122448}
{"current_steps": 13030, "total_steps": 16324, "loss": 0.1856, "lr": 1.0726394115827949e-05, "epoch": 1.5963676457076352, "percentage": 79.82, "elapsed_time": "16:57:38", "remaining_time": "4:17:15", "throughput": 1427.42, "total_tokens": 87156416}
{"current_steps": 13035, "total_steps": 16324, "loss": 0.2147, "lr": 1.0695068387589218e-05, "epoch": 1.5969801843741385, "percentage": 79.85, "elapsed_time": "16:58:38", "remaining_time": "4:17:01", "throughput": 1426.58, "total_tokens": 87190064}
{"current_steps": 13040, "total_steps": 16324, "loss": 0.2193, "lr": 1.0663782989625914e-05, "epoch": 1.597592723040642, "percentage": 79.88, "elapsed_time": "16:59:37", "remaining_time": "4:16:47", "throughput": 1425.73, "total_tokens": 87223272}
{"current_steps": 13045, "total_steps": 16324, "loss": 0.1845, "lr": 1.0632537954039595e-05, "epoch": 1.5982052617071454, "percentage": 79.91, "elapsed_time": "17:00:37", "remaining_time": "4:16:32", "throughput": 1424.91, "total_tokens": 87257696}
{"current_steps": 13050, "total_steps": 16324, "loss": 0.1889, "lr": 1.0601333312890393e-05, "epoch": 1.5988178003736486, "percentage": 79.94, "elapsed_time": "17:01:36", "remaining_time": "4:16:18", "throughput": 1424.07, "total_tokens": 87291504}
{"current_steps": 13050, "total_steps": 16324, "eval_loss": 0.12491484731435776, "epoch": 1.5988178003736486, "percentage": 79.94, "elapsed_time": "17:01:56", "remaining_time": "4:16:23", "throughput": 1423.62, "total_tokens": 87291504}
{"current_steps": 13055, "total_steps": 16324, "loss": 0.1994, "lr": 1.0570169098196952e-05, "epoch": 1.5994303390401519, "percentage": 79.97, "elapsed_time": "17:03:00", "remaining_time": "4:16:09", "throughput": 1422.67, "total_tokens": 87324920}
{"current_steps": 13060, "total_steps": 16324, "loss": 0.2388, "lr": 1.0539045341936477e-05, "epoch": 1.600042877706655, "percentage": 80.0, "elapsed_time": "17:04:00", "remaining_time": "4:15:55", "throughput": 1421.82, "total_tokens": 87358072}
{"current_steps": 13065, "total_steps": 16324, "loss": 0.1867, "lr": 1.0507962076044652e-05, "epoch": 1.6006554163731586, "percentage": 80.04, "elapsed_time": "17:05:00", "remaining_time": "4:15:41", "throughput": 1420.99, "total_tokens": 87391824}
{"current_steps": 13070, "total_steps": 16324, "loss": 0.2139, "lr": 1.0476919332415619e-05, "epoch": 1.601267955039662, "percentage": 80.07, "elapsed_time": "17:06:00", "remaining_time": "4:15:26", "throughput": 1420.17, "total_tokens": 87425544}
{"current_steps": 13075, "total_steps": 16324, "loss": 0.1886, "lr": 1.0445917142901906e-05, "epoch": 1.6018804937061653, "percentage": 80.1, "elapsed_time": "17:07:00", "remaining_time": "4:15:11", "throughput": 1419.33, "total_tokens": 87459408}
{"current_steps": 13080, "total_steps": 16324, "loss": 0.2091, "lr": 1.0414955539314464e-05, "epoch": 1.6024930323726685, "percentage": 80.13, "elapsed_time": "17:07:59", "remaining_time": "4:14:57", "throughput": 1418.51, "total_tokens": 87493464}
{"current_steps": 13085, "total_steps": 16324, "loss": 0.18, "lr": 1.0384034553422605e-05, "epoch": 1.6031055710391717, "percentage": 80.16, "elapsed_time": "17:08:59", "remaining_time": "4:14:42", "throughput": 1417.69, "total_tokens": 87527488}
{"current_steps": 13090, "total_steps": 16324, "loss": 0.1927, "lr": 1.0353154216953937e-05, "epoch": 1.6037181097056752, "percentage": 80.19, "elapsed_time": "17:09:59", "remaining_time": "4:14:28", "throughput": 1416.85, "total_tokens": 87561072}
{"current_steps": 13095, "total_steps": 16324, "loss": 0.2034, "lr": 1.0322314561594343e-05, "epoch": 1.6043306483721786, "percentage": 80.22, "elapsed_time": "17:10:59", "remaining_time": "4:14:13", "throughput": 1416.03, "total_tokens": 87594680}
{"current_steps": 13100, "total_steps": 16324, "loss": 0.2406, "lr": 1.0291515618988012e-05, "epoch": 1.6049431870386819, "percentage": 80.25, "elapsed_time": "17:11:58", "remaining_time": "4:13:58", "throughput": 1415.21, "total_tokens": 87627616}
{"current_steps": 13100, "total_steps": 16324, "eval_loss": 0.08859918266534805, "epoch": 1.6049431870386819, "percentage": 80.25, "elapsed_time": "17:12:17", "remaining_time": "4:14:03", "throughput": 1414.77, "total_tokens": 87627616}
{"current_steps": 13105, "total_steps": 16324, "loss": 0.1987, "lr": 1.0260757420737354e-05, "epoch": 1.6055557257051851, "percentage": 80.28, "elapsed_time": "17:13:22", "remaining_time": "4:13:49", "throughput": 1413.83, "total_tokens": 87660704}
{"current_steps": 13110, "total_steps": 16324, "loss": 0.2083, "lr": 1.0230039998402919e-05, "epoch": 1.6061682643716884, "percentage": 80.31, "elapsed_time": "17:14:20", "remaining_time": "4:13:34", "throughput": 1413.04, "total_tokens": 87694136}
{"current_steps": 13115, "total_steps": 16324, "loss": 0.2125, "lr": 1.0199363383503464e-05, "epoch": 1.6067808030381918, "percentage": 80.34, "elapsed_time": "17:15:20", "remaining_time": "4:13:19", "throughput": 1412.22, "total_tokens": 87727504}
{"current_steps": 13120, "total_steps": 16324, "loss": 0.2167, "lr": 1.016872760751586e-05, "epoch": 1.607393341704695, "percentage": 80.37, "elapsed_time": "17:16:19", "remaining_time": "4:13:04", "throughput": 1411.41, "total_tokens": 87760824}
{"current_steps": 13125, "total_steps": 16324, "loss": 0.2319, "lr": 1.0138132701875103e-05, "epoch": 1.6080058803711985, "percentage": 80.4, "elapsed_time": "17:17:17", "remaining_time": "4:12:49", "throughput": 1410.62, "total_tokens": 87794008}
{"current_steps": 13130, "total_steps": 16324, "loss": 0.2051, "lr": 1.0107578697974179e-05, "epoch": 1.6086184190377018, "percentage": 80.43, "elapsed_time": "17:18:16", "remaining_time": "4:12:34", "throughput": 1409.83, "total_tokens": 87827672}
{"current_steps": 13135, "total_steps": 16324, "loss": 0.2217, "lr": 1.0077065627164178e-05, "epoch": 1.609230957704205, "percentage": 80.46, "elapsed_time": "17:19:15", "remaining_time": "4:12:19", "throughput": 1409.03, "total_tokens": 87861320}
{"current_steps": 13140, "total_steps": 16324, "loss": 0.2425, "lr": 1.0046593520754171e-05, "epoch": 1.6098434963707084, "percentage": 80.49, "elapsed_time": "17:20:14", "remaining_time": "4:12:03", "throughput": 1408.25, "total_tokens": 87894304}
{"current_steps": 13145, "total_steps": 16324, "loss": 0.189, "lr": 1.001616241001116e-05, "epoch": 1.6104560350372117, "percentage": 80.53, "elapsed_time": "17:21:13", "remaining_time": "4:11:48", "throughput": 1407.46, "total_tokens": 87928496}
{"current_steps": 13150, "total_steps": 16324, "loss": 0.2467, "lr": 9.985772326160125e-06, "epoch": 1.6110685737037151, "percentage": 80.56, "elapsed_time": "17:22:12", "remaining_time": "4:11:33", "throughput": 1406.65, "total_tokens": 87961872}
{"current_steps": 13150, "total_steps": 16324, "eval_loss": 0.0699838325381279, "epoch": 1.6110685737037151, "percentage": 80.56, "elapsed_time": "17:22:32", "remaining_time": "4:11:38", "throughput": 1406.22, "total_tokens": 87961872}
{"current_steps": 13155, "total_steps": 16324, "loss": 0.2211, "lr": 9.955423300383942e-06, "epoch": 1.6116811123702184, "percentage": 80.59, "elapsed_time": "17:23:36", "remaining_time": "4:11:24", "throughput": 1405.3, "total_tokens": 87994856}
{"current_steps": 13160, "total_steps": 16324, "loss": 0.176, "lr": 9.925115363823328e-06, "epoch": 1.6122936510367216, "percentage": 80.62, "elapsed_time": "17:24:35", "remaining_time": "4:11:08", "throughput": 1404.52, "total_tokens": 88028832}
{"current_steps": 13165, "total_steps": 16324, "loss": 0.2368, "lr": 9.894848547576868e-06, "epoch": 1.6129061897032249, "percentage": 80.65, "elapsed_time": "17:25:34", "remaining_time": "4:10:53", "throughput": 1403.72, "total_tokens": 88062144}
{"current_steps": 13170, "total_steps": 16324, "loss": 0.155, "lr": 9.864622882700958e-06, "epoch": 1.6135187283697283, "percentage": 80.68, "elapsed_time": "17:26:33", "remaining_time": "4:10:38", "throughput": 1402.95, "total_tokens": 88096560}
{"current_steps": 13175, "total_steps": 16324, "loss": 0.2433, "lr": 9.834438400209733e-06, "epoch": 1.6141312670362318, "percentage": 80.71, "elapsed_time": "17:27:32", "remaining_time": "4:10:22", "throughput": 1402.17, "total_tokens": 88129968}
{"current_steps": 13180, "total_steps": 16324, "loss": 0.2006, "lr": 9.804295131075125e-06, "epoch": 1.614743805702735, "percentage": 80.74, "elapsed_time": "17:28:32", "remaining_time": "4:10:07", "throughput": 1401.37, "total_tokens": 88163752}
{"current_steps": 13185, "total_steps": 16324, "loss": 0.2332, "lr": 9.774193106226715e-06, "epoch": 1.6153563443692383, "percentage": 80.77, "elapsed_time": "17:29:31", "remaining_time": "4:09:51", "throughput": 1400.59, "total_tokens": 88197712}
{"current_steps": 13190, "total_steps": 16324, "loss": 0.1907, "lr": 9.744132356551815e-06, "epoch": 1.6159688830357415, "percentage": 80.8, "elapsed_time": "17:30:31", "remaining_time": "4:09:36", "throughput": 1399.79, "total_tokens": 88231264}
{"current_steps": 13195, "total_steps": 16324, "loss": 0.2256, "lr": 9.714112912895373e-06, "epoch": 1.616581421702245, "percentage": 80.83, "elapsed_time": "17:31:31", "remaining_time": "4:09:21", "throughput": 1398.99, "total_tokens": 88264088}
{"current_steps": 13200, "total_steps": 16324, "loss": 0.2259, "lr": 9.684134806059925e-06, "epoch": 1.6171939603687484, "percentage": 80.86, "elapsed_time": "17:32:31", "remaining_time": "4:09:05", "throughput": 1398.18, "total_tokens": 88297128}
{"current_steps": 13200, "total_steps": 16324, "eval_loss": 0.08684820681810379, "epoch": 1.6171939603687484, "percentage": 80.86, "elapsed_time": "17:32:51", "remaining_time": "4:09:10", "throughput": 1397.75, "total_tokens": 88297128}
{"current_steps": 13205, "total_steps": 16324, "loss": 0.2348, "lr": 9.65419806680563e-06, "epoch": 1.6178064990352516, "percentage": 80.89, "elapsed_time": "17:33:56", "remaining_time": "4:08:56", "throughput": 1396.83, "total_tokens": 88330432}
{"current_steps": 13210, "total_steps": 16324, "loss": 0.1765, "lr": 9.624302725850187e-06, "epoch": 1.6184190377017549, "percentage": 80.92, "elapsed_time": "17:34:55", "remaining_time": "4:08:40", "throughput": 1396.06, "total_tokens": 88364576}
{"current_steps": 13215, "total_steps": 16324, "loss": 0.2151, "lr": 9.5944488138688e-06, "epoch": 1.6190315763682581, "percentage": 80.95, "elapsed_time": "17:35:55", "remaining_time": "4:08:25", "throughput": 1395.27, "total_tokens": 88397304}
{"current_steps": 13220, "total_steps": 16324, "loss": 0.1647, "lr": 9.564636361494178e-06, "epoch": 1.6196441150347616, "percentage": 80.99, "elapsed_time": "17:36:53", "remaining_time": "4:08:09", "throughput": 1394.52, "total_tokens": 88431768}
{"current_steps": 13225, "total_steps": 16324, "loss": 0.1876, "lr": 9.5348653993165e-06, "epoch": 1.620256653701265, "percentage": 81.02, "elapsed_time": "17:37:52", "remaining_time": "4:07:53", "throughput": 1393.77, "total_tokens": 88465312}
{"current_steps": 13230, "total_steps": 16324, "loss": 0.2052, "lr": 9.505135957883365e-06, "epoch": 1.6208691923677683, "percentage": 81.05, "elapsed_time": "17:38:51", "remaining_time": "4:07:37", "throughput": 1392.99, "total_tokens": 88498456}
{"current_steps": 13235, "total_steps": 16324, "loss": 0.1931, "lr": 9.475448067699739e-06, "epoch": 1.6214817310342715, "percentage": 81.08, "elapsed_time": "17:39:49", "remaining_time": "4:07:21", "throughput": 1392.24, "total_tokens": 88532112}
{"current_steps": 13240, "total_steps": 16324, "loss": 0.2242, "lr": 9.44580175922799e-06, "epoch": 1.6220942697007747, "percentage": 81.11, "elapsed_time": "17:40:48", "remaining_time": "4:07:05", "throughput": 1391.47, "total_tokens": 88565560}
{"current_steps": 13245, "total_steps": 16324, "loss": 0.1775, "lr": 9.416197062887815e-06, "epoch": 1.6227068083672782, "percentage": 81.14, "elapsed_time": "17:41:47", "remaining_time": "4:06:49", "throughput": 1390.72, "total_tokens": 88599824}
{"current_steps": 13250, "total_steps": 16324, "loss": 0.2092, "lr": 9.386634009056194e-06, "epoch": 1.6233193470337817, "percentage": 81.17, "elapsed_time": "17:42:46", "remaining_time": "4:06:33", "throughput": 1389.96, "total_tokens": 88633080}
{"current_steps": 13250, "total_steps": 16324, "eval_loss": 0.08336101472377777, "epoch": 1.6233193470337817, "percentage": 81.17, "elapsed_time": "17:43:05", "remaining_time": "4:06:38", "throughput": 1389.55, "total_tokens": 88633080}
{"current_steps": 13255, "total_steps": 16324, "loss": 0.202, "lr": 9.357112628067399e-06, "epoch": 1.623931885700285, "percentage": 81.2, "elapsed_time": "17:44:10", "remaining_time": "4:06:23", "throughput": 1388.67, "total_tokens": 88666784}
{"current_steps": 13260, "total_steps": 16324, "loss": 0.1991, "lr": 9.327632950212917e-06, "epoch": 1.6245444243667881, "percentage": 81.23, "elapsed_time": "17:45:09", "remaining_time": "4:06:07", "throughput": 1387.9, "total_tokens": 88700064}
{"current_steps": 13265, "total_steps": 16324, "loss": 0.1668, "lr": 9.298195005741484e-06, "epoch": 1.6251569630332914, "percentage": 81.26, "elapsed_time": "17:46:07", "remaining_time": "4:05:51", "throughput": 1387.18, "total_tokens": 88734648}
{"current_steps": 13270, "total_steps": 16324, "loss": 0.2213, "lr": 9.268798824858954e-06, "epoch": 1.6257695016997948, "percentage": 81.29, "elapsed_time": "17:47:06", "remaining_time": "4:05:35", "throughput": 1386.41, "total_tokens": 88767416}
{"current_steps": 13275, "total_steps": 16324, "loss": 0.2311, "lr": 9.239444437728384e-06, "epoch": 1.626382040366298, "percentage": 81.32, "elapsed_time": "17:48:06", "remaining_time": "4:05:19", "throughput": 1385.64, "total_tokens": 88800680}
{"current_steps": 13280, "total_steps": 16324, "loss": 0.2014, "lr": 9.21013187446993e-06, "epoch": 1.6269945790328015, "percentage": 81.35, "elapsed_time": "17:49:05", "remaining_time": "4:05:03", "throughput": 1384.89, "total_tokens": 88834152}
{"current_steps": 13285, "total_steps": 16324, "loss": 0.1985, "lr": 9.18086116516083e-06, "epoch": 1.6276071176993048, "percentage": 81.38, "elapsed_time": "17:50:04", "remaining_time": "4:04:47", "throughput": 1384.14, "total_tokens": 88868144}
{"current_steps": 13290, "total_steps": 16324, "loss": 0.2323, "lr": 9.15163233983536e-06, "epoch": 1.628219656365808, "percentage": 81.41, "elapsed_time": "17:51:04", "remaining_time": "4:04:30", "throughput": 1383.37, "total_tokens": 88901064}
{"current_steps": 13295, "total_steps": 16324, "loss": 0.1814, "lr": 9.122445428484844e-06, "epoch": 1.6288321950323115, "percentage": 81.44, "elapsed_time": "17:52:03", "remaining_time": "4:04:14", "throughput": 1382.62, "total_tokens": 88934848}
{"current_steps": 13300, "total_steps": 16324, "loss": 0.2285, "lr": 9.093300461057602e-06, "epoch": 1.6294447336988147, "percentage": 81.48, "elapsed_time": "17:53:02", "remaining_time": "4:03:58", "throughput": 1381.86, "total_tokens": 88968232}
{"current_steps": 13300, "total_steps": 16324, "eval_loss": 0.13456711173057556, "epoch": 1.6294447336988147, "percentage": 81.48, "elapsed_time": "17:53:22", "remaining_time": "4:04:03", "throughput": 1381.44, "total_tokens": 88968232}
{"current_steps": 13305, "total_steps": 16324, "loss": 0.2156, "lr": 9.06419746745888e-06, "epoch": 1.6300572723653182, "percentage": 81.51, "elapsed_time": "17:54:27", "remaining_time": "4:03:48", "throughput": 1380.58, "total_tokens": 89002208}
{"current_steps": 13310, "total_steps": 16324, "loss": 0.1972, "lr": 9.035136477550887e-06, "epoch": 1.6306698110318214, "percentage": 81.54, "elapsed_time": "17:55:26", "remaining_time": "4:03:31", "throughput": 1379.83, "total_tokens": 89036088}
{"current_steps": 13315, "total_steps": 16324, "loss": 0.2018, "lr": 9.00611752115273e-06, "epoch": 1.6312823496983246, "percentage": 81.57, "elapsed_time": "17:56:26", "remaining_time": "4:03:15", "throughput": 1379.08, "total_tokens": 89070144}
{"current_steps": 13320, "total_steps": 16324, "loss": 0.2123, "lr": 8.977140628040382e-06, "epoch": 1.6318948883648279, "percentage": 81.6, "elapsed_time": "17:57:26", "remaining_time": "4:02:59", "throughput": 1378.32, "total_tokens": 89103352}
{"current_steps": 13325, "total_steps": 16324, "loss": 0.2224, "lr": 8.948205827946637e-06, "epoch": 1.6325074270313313, "percentage": 81.63, "elapsed_time": "17:58:26", "remaining_time": "4:02:43", "throughput": 1377.57, "total_tokens": 89137296}
{"current_steps": 13330, "total_steps": 16324, "loss": 0.2518, "lr": 8.919313150561131e-06, "epoch": 1.6331199656978348, "percentage": 81.66, "elapsed_time": "17:59:24", "remaining_time": "4:02:26", "throughput": 1376.83, "total_tokens": 89169968}
{"current_steps": 13335, "total_steps": 16324, "loss": 0.2301, "lr": 8.890462625530266e-06, "epoch": 1.633732504364338, "percentage": 81.69, "elapsed_time": "18:00:23", "remaining_time": "4:02:10", "throughput": 1376.08, "total_tokens": 89202856}
{"current_steps": 13340, "total_steps": 16324, "loss": 0.1833, "lr": 8.861654282457176e-06, "epoch": 1.6343450430308413, "percentage": 81.72, "elapsed_time": "18:01:23", "remaining_time": "4:01:53", "throughput": 1375.34, "total_tokens": 89236560}
{"current_steps": 13345, "total_steps": 16324, "loss": 0.1836, "lr": 8.832888150901713e-06, "epoch": 1.6349575816973445, "percentage": 81.75, "elapsed_time": "18:02:22", "remaining_time": "4:01:37", "throughput": 1374.62, "total_tokens": 89270712}
{"current_steps": 13350, "total_steps": 16324, "loss": 0.1994, "lr": 8.804164260380443e-06, "epoch": 1.635570120363848, "percentage": 81.78, "elapsed_time": "18:03:20", "remaining_time": "4:01:20", "throughput": 1373.89, "total_tokens": 89304224}
{"current_steps": 13350, "total_steps": 16324, "eval_loss": 0.14407266676425934, "epoch": 1.635570120363848, "percentage": 81.78, "elapsed_time": "18:03:40", "remaining_time": "4:01:24", "throughput": 1373.48, "total_tokens": 89304224}
{"current_steps": 13355, "total_steps": 16324, "loss": 0.1972, "lr": 8.775482640366583e-06, "epoch": 1.6361826590303514, "percentage": 81.81, "elapsed_time": "18:04:44", "remaining_time": "4:01:09", "throughput": 1372.65, "total_tokens": 89338352}
{"current_steps": 13360, "total_steps": 16324, "loss": 0.2029, "lr": 8.746843320289944e-06, "epoch": 1.6367951976968547, "percentage": 81.84, "elapsed_time": "18:05:43", "remaining_time": "4:00:52", "throughput": 1371.93, "total_tokens": 89372176}
{"current_steps": 13365, "total_steps": 16324, "loss": 0.2527, "lr": 8.718246329536967e-06, "epoch": 1.637407736363358, "percentage": 81.87, "elapsed_time": "18:06:42", "remaining_time": "4:00:35", "throughput": 1371.2, "total_tokens": 89405592}
{"current_steps": 13370, "total_steps": 16324, "loss": 0.2507, "lr": 8.689691697450669e-06, "epoch": 1.6380202750298611, "percentage": 81.9, "elapsed_time": "18:07:42", "remaining_time": "4:00:19", "throughput": 1370.46, "total_tokens": 89439272}
{"current_steps": 13375, "total_steps": 16324, "loss": 0.2086, "lr": 8.661179453330553e-06, "epoch": 1.6386328136963646, "percentage": 81.93, "elapsed_time": "18:08:42", "remaining_time": "4:00:02", "throughput": 1369.72, "total_tokens": 89473016}
{"current_steps": 13380, "total_steps": 16324, "loss": 0.2259, "lr": 8.632709626432672e-06, "epoch": 1.639245352362868, "percentage": 81.97, "elapsed_time": "18:09:42", "remaining_time": "3:59:46", "throughput": 1368.95, "total_tokens": 89506200}
{"current_steps": 13385, "total_steps": 16324, "loss": 0.1939, "lr": 8.604282245969548e-06, "epoch": 1.6398578910293713, "percentage": 82.0, "elapsed_time": "18:10:44", "remaining_time": "3:59:29", "throughput": 1368.18, "total_tokens": 89539440}
{"current_steps": 13390, "total_steps": 16324, "loss": 0.2199, "lr": 8.575897341110145e-06, "epoch": 1.6404704296958745, "percentage": 82.03, "elapsed_time": "18:11:44", "remaining_time": "3:59:13", "throughput": 1367.42, "total_tokens": 89572368}
{"current_steps": 13395, "total_steps": 16324, "loss": 0.1917, "lr": 8.54755494097983e-06, "epoch": 1.6410829683623778, "percentage": 82.06, "elapsed_time": "18:12:46", "remaining_time": "3:58:57", "throughput": 1366.65, "total_tokens": 89606416}
{"current_steps": 13400, "total_steps": 16324, "loss": 0.2202, "lr": 8.51925507466037e-06, "epoch": 1.6416955070288812, "percentage": 82.09, "elapsed_time": "18:13:48", "remaining_time": "3:58:40", "throughput": 1365.87, "total_tokens": 89640800}
{"current_steps": 13400, "total_steps": 16324, "eval_loss": 0.12518566846847534, "epoch": 1.6416955070288812, "percentage": 82.09, "elapsed_time": "18:14:09", "remaining_time": "3:58:45", "throughput": 1365.45, "total_tokens": 89640800}
{"current_steps": 13405, "total_steps": 16324, "loss": 0.2142, "lr": 8.490997771189907e-06, "epoch": 1.6423080456953845, "percentage": 82.12, "elapsed_time": "18:15:16", "remaining_time": "3:58:29", "throughput": 1364.58, "total_tokens": 89674616}
{"current_steps": 13410, "total_steps": 16324, "loss": 0.2115, "lr": 8.462783059562862e-06, "epoch": 1.642920584361888, "percentage": 82.15, "elapsed_time": "18:16:18", "remaining_time": "3:58:13", "throughput": 1363.8, "total_tokens": 89708056}
{"current_steps": 13415, "total_steps": 16324, "loss": 0.1682, "lr": 8.434610968730006e-06, "epoch": 1.6435331230283912, "percentage": 82.18, "elapsed_time": "18:17:18", "remaining_time": "3:57:56", "throughput": 1363.07, "total_tokens": 89742072}
{"current_steps": 13420, "total_steps": 16324, "loss": 0.2267, "lr": 8.406481527598325e-06, "epoch": 1.6441456616948944, "percentage": 82.21, "elapsed_time": "18:18:18", "remaining_time": "3:57:39", "throughput": 1362.33, "total_tokens": 89774920}
{"current_steps": 13425, "total_steps": 16324, "loss": 0.1916, "lr": 8.378394765031106e-06, "epoch": 1.6447582003613979, "percentage": 82.24, "elapsed_time": "18:19:18", "remaining_time": "3:57:23", "throughput": 1361.59, "total_tokens": 89808840}
{"current_steps": 13430, "total_steps": 16324, "loss": 0.2105, "lr": 8.350350709847764e-06, "epoch": 1.645370739027901, "percentage": 82.27, "elapsed_time": "18:20:18", "remaining_time": "3:57:06", "throughput": 1360.87, "total_tokens": 89842896}
{"current_steps": 13435, "total_steps": 16324, "loss": 0.1993, "lr": 8.322349390823969e-06, "epoch": 1.6459832776944046, "percentage": 82.3, "elapsed_time": "18:21:17", "remaining_time": "3:56:49", "throughput": 1360.17, "total_tokens": 89876832}
{"current_steps": 13440, "total_steps": 16324, "loss": 0.2305, "lr": 8.294390836691496e-06, "epoch": 1.6465958163609078, "percentage": 82.33, "elapsed_time": "18:22:16", "remaining_time": "3:56:31", "throughput": 1359.46, "total_tokens": 89910688}
{"current_steps": 13445, "total_steps": 16324, "loss": 0.1883, "lr": 8.266475076138263e-06, "epoch": 1.647208355027411, "percentage": 82.36, "elapsed_time": "18:23:15", "remaining_time": "3:56:14", "throughput": 1358.77, "total_tokens": 89944992}
{"current_steps": 13450, "total_steps": 16324, "loss": 0.1682, "lr": 8.238602137808249e-06, "epoch": 1.6478208936939143, "percentage": 82.39, "elapsed_time": "18:24:15", "remaining_time": "3:55:57", "throughput": 1358.05, "total_tokens": 89978672}
{"current_steps": 13450, "total_steps": 16324, "eval_loss": 0.12995800375938416, "epoch": 1.6478208936939143, "percentage": 82.39, "elapsed_time": "18:24:35", "remaining_time": "3:56:01", "throughput": 1357.66, "total_tokens": 89978672}
{"current_steps": 13455, "total_steps": 16324, "loss": 0.1763, "lr": 8.210772050301519e-06, "epoch": 1.6484334323604177, "percentage": 82.42, "elapsed_time": "18:25:40", "remaining_time": "3:55:45", "throughput": 1356.83, "total_tokens": 90012432}
{"current_steps": 13460, "total_steps": 16324, "loss": 0.1827, "lr": 8.182984842174174e-06, "epoch": 1.6490459710269212, "percentage": 82.46, "elapsed_time": "18:26:39", "remaining_time": "3:55:28", "throughput": 1356.13, "total_tokens": 90046184}
{"current_steps": 13465, "total_steps": 16324, "loss": 0.2052, "lr": 8.155240541938275e-06, "epoch": 1.6496585096934244, "percentage": 82.49, "elapsed_time": "18:27:39", "remaining_time": "3:55:11", "throughput": 1355.42, "total_tokens": 90080280}
{"current_steps": 13470, "total_steps": 16324, "loss": 0.229, "lr": 8.127539178061906e-06, "epoch": 1.6502710483599277, "percentage": 82.52, "elapsed_time": "18:28:38", "remaining_time": "3:54:53", "throughput": 1354.7, "total_tokens": 90113024}
{"current_steps": 13475, "total_steps": 16324, "loss": 0.2731, "lr": 8.099880778969066e-06, "epoch": 1.650883587026431, "percentage": 82.55, "elapsed_time": "18:29:38", "remaining_time": "3:54:36", "throughput": 1353.99, "total_tokens": 90146384}
{"current_steps": 13480, "total_steps": 16324, "loss": 0.2132, "lr": 8.07226537303969e-06, "epoch": 1.6514961256929344, "percentage": 82.58, "elapsed_time": "18:30:38", "remaining_time": "3:54:19", "throughput": 1353.27, "total_tokens": 90179920}
{"current_steps": 13485, "total_steps": 16324, "loss": 0.2227, "lr": 8.044692988609565e-06, "epoch": 1.6521086643594378, "percentage": 82.61, "elapsed_time": "18:31:38", "remaining_time": "3:54:01", "throughput": 1352.56, "total_tokens": 90213008}
{"current_steps": 13490, "total_steps": 16324, "loss": 0.2084, "lr": 8.017163653970361e-06, "epoch": 1.652721203025941, "percentage": 82.64, "elapsed_time": "18:32:37", "remaining_time": "3:53:44", "throughput": 1351.85, "total_tokens": 90246872}
{"current_steps": 13495, "total_steps": 16324, "loss": 0.2222, "lr": 7.989677397369577e-06, "epoch": 1.6533337416924443, "percentage": 82.67, "elapsed_time": "18:33:37", "remaining_time": "3:53:27", "throughput": 1351.14, "total_tokens": 90280160}
{"current_steps": 13500, "total_steps": 16324, "loss": 0.1801, "lr": 7.962234247010497e-06, "epoch": 1.6539462803589475, "percentage": 82.7, "elapsed_time": "18:34:37", "remaining_time": "3:53:09", "throughput": 1350.43, "total_tokens": 90313896}
{"current_steps": 13500, "total_steps": 16324, "eval_loss": 0.11713992059230804, "epoch": 1.6539462803589475, "percentage": 82.7, "elapsed_time": "18:34:57", "remaining_time": "3:53:13", "throughput": 1350.03, "total_tokens": 90313896}
{"current_steps": 13505, "total_steps": 16324, "loss": 0.1951, "lr": 7.934834231052157e-06, "epoch": 1.654558819025451, "percentage": 82.73, "elapsed_time": "18:36:02", "remaining_time": "3:52:57", "throughput": 1349.22, "total_tokens": 90347776}
{"current_steps": 13510, "total_steps": 16324, "loss": 0.2118, "lr": 7.907477377609379e-06, "epoch": 1.6551713576919544, "percentage": 82.76, "elapsed_time": "18:37:03", "remaining_time": "3:52:40", "throughput": 1348.5, "total_tokens": 90381544}
{"current_steps": 13515, "total_steps": 16324, "loss": 0.1658, "lr": 7.880163714752669e-06, "epoch": 1.6557838963584577, "percentage": 82.79, "elapsed_time": "18:38:03", "remaining_time": "3:52:22", "throughput": 1347.81, "total_tokens": 90415952}
{"current_steps": 13520, "total_steps": 16324, "loss": 0.21, "lr": 7.8528932705082e-06, "epoch": 1.656396435024961, "percentage": 82.82, "elapsed_time": "18:39:04", "remaining_time": "3:52:05", "throughput": 1347.09, "total_tokens": 90449864}
{"current_steps": 13525, "total_steps": 16324, "loss": 0.2054, "lr": 7.825666072857834e-06, "epoch": 1.6570089736914642, "percentage": 82.85, "elapsed_time": "18:40:05", "remaining_time": "3:51:48", "throughput": 1346.38, "total_tokens": 90483936}
{"current_steps": 13530, "total_steps": 16324, "loss": 0.2316, "lr": 7.798482149739051e-06, "epoch": 1.6576215123579676, "percentage": 82.88, "elapsed_time": "18:41:05", "remaining_time": "3:51:30", "throughput": 1345.66, "total_tokens": 90516976}
{"current_steps": 13535, "total_steps": 16324, "loss": 0.2072, "lr": 7.771341529044895e-06, "epoch": 1.658234051024471, "percentage": 82.91, "elapsed_time": "18:42:06", "remaining_time": "3:51:13", "throughput": 1344.94, "total_tokens": 90550136}
{"current_steps": 13540, "total_steps": 16324, "loss": 0.2168, "lr": 7.74424423862401e-06, "epoch": 1.6588465896909743, "percentage": 82.95, "elapsed_time": "18:43:06", "remaining_time": "3:50:55", "throughput": 1344.23, "total_tokens": 90583032}
{"current_steps": 13545, "total_steps": 16324, "loss": 0.2199, "lr": 7.717190306280575e-06, "epoch": 1.6594591283574776, "percentage": 82.98, "elapsed_time": "18:44:06", "remaining_time": "3:50:37", "throughput": 1343.53, "total_tokens": 90617000}
{"current_steps": 13550, "total_steps": 16324, "loss": 0.2621, "lr": 7.690179759774285e-06, "epoch": 1.6600716670239808, "percentage": 83.01, "elapsed_time": "18:45:06", "remaining_time": "3:50:20", "throughput": 1342.83, "total_tokens": 90650056}
{"current_steps": 13550, "total_steps": 16324, "eval_loss": 0.09069520980119705, "epoch": 1.6600716670239808, "percentage": 83.01, "elapsed_time": "18:45:26", "remaining_time": "3:50:24", "throughput": 1342.44, "total_tokens": 90650056}
{"current_steps": 13555, "total_steps": 16324, "loss": 0.1932, "lr": 7.663212626820287e-06, "epoch": 1.6606842056904842, "percentage": 83.04, "elapsed_time": "18:46:31", "remaining_time": "3:50:07", "throughput": 1341.66, "total_tokens": 90684496}
{"current_steps": 13560, "total_steps": 16324, "loss": 0.218, "lr": 7.636288935089214e-06, "epoch": 1.6612967443569875, "percentage": 83.07, "elapsed_time": "18:47:31", "remaining_time": "3:49:49", "throughput": 1340.95, "total_tokens": 90717064}
{"current_steps": 13565, "total_steps": 16324, "loss": 0.2356, "lr": 7.609408712207122e-06, "epoch": 1.661909283023491, "percentage": 83.1, "elapsed_time": "18:48:31", "remaining_time": "3:49:31", "throughput": 1340.25, "total_tokens": 90750248}
{"current_steps": 13570, "total_steps": 16324, "loss": 0.2164, "lr": 7.582571985755433e-06, "epoch": 1.6625218216899942, "percentage": 83.13, "elapsed_time": "18:49:32", "remaining_time": "3:49:14", "throughput": 1339.55, "total_tokens": 90783952}
{"current_steps": 13575, "total_steps": 16324, "loss": 0.2227, "lr": 7.555778783270984e-06, "epoch": 1.6631343603564974, "percentage": 83.16, "elapsed_time": "18:50:32", "remaining_time": "3:48:56", "throughput": 1338.85, "total_tokens": 90817168}
{"current_steps": 13580, "total_steps": 16324, "loss": 0.2205, "lr": 7.529029132245907e-06, "epoch": 1.6637468990230009, "percentage": 83.19, "elapsed_time": "18:51:31", "remaining_time": "3:48:38", "throughput": 1338.18, "total_tokens": 90851120}
{"current_steps": 13585, "total_steps": 16324, "loss": 0.2159, "lr": 7.502323060127697e-06, "epoch": 1.6643594376895041, "percentage": 83.22, "elapsed_time": "18:52:31", "remaining_time": "3:48:20", "throughput": 1337.49, "total_tokens": 90884568}
{"current_steps": 13590, "total_steps": 16324, "loss": 0.2444, "lr": 7.4756605943190725e-06, "epoch": 1.6649719763560076, "percentage": 83.25, "elapsed_time": "18:53:32", "remaining_time": "3:48:02", "throughput": 1336.79, "total_tokens": 90917968}
{"current_steps": 13595, "total_steps": 16324, "loss": 0.2165, "lr": 7.4490417621780495e-06, "epoch": 1.6655845150225108, "percentage": 83.28, "elapsed_time": "18:54:32", "remaining_time": "3:47:44", "throughput": 1336.1, "total_tokens": 90951304}
{"current_steps": 13600, "total_steps": 16324, "loss": 0.2048, "lr": 7.422466591017879e-06, "epoch": 1.666197053689014, "percentage": 83.31, "elapsed_time": "18:55:32", "remaining_time": "3:47:26", "throughput": 1335.42, "total_tokens": 90985192}
{"current_steps": 13600, "total_steps": 16324, "eval_loss": 0.09095155447721481, "epoch": 1.666197053689014, "percentage": 83.31, "elapsed_time": "18:55:51", "remaining_time": "3:47:30", "throughput": 1335.03, "total_tokens": 90985192}
{"current_steps": 13605, "total_steps": 16324, "loss": 0.1976, "lr": 7.395935108106988e-06, "epoch": 1.6668095923555173, "percentage": 83.34, "elapsed_time": "18:56:57", "remaining_time": "3:47:13", "throughput": 1334.24, "total_tokens": 91018688}
{"current_steps": 13610, "total_steps": 16324, "loss": 0.2054, "lr": 7.369447340668978e-06, "epoch": 1.6674221310220207, "percentage": 83.37, "elapsed_time": "18:57:57", "remaining_time": "3:46:55", "throughput": 1333.56, "total_tokens": 91052632}
{"current_steps": 13615, "total_steps": 16324, "loss": 0.2193, "lr": 7.3430033158825966e-06, "epoch": 1.6680346696885242, "percentage": 83.4, "elapsed_time": "18:58:57", "remaining_time": "3:46:37", "throughput": 1332.87, "total_tokens": 91085408}
{"current_steps": 13620, "total_steps": 16324, "loss": 0.2208, "lr": 7.3166030608817385e-06, "epoch": 1.6686472083550274, "percentage": 83.44, "elapsed_time": "18:59:58", "remaining_time": "3:46:19", "throughput": 1332.17, "total_tokens": 91118408}
{"current_steps": 13625, "total_steps": 16324, "loss": 0.2142, "lr": 7.290246602755318e-06, "epoch": 1.6692597470215307, "percentage": 83.47, "elapsed_time": "19:00:58", "remaining_time": "3:46:01", "throughput": 1331.48, "total_tokens": 91151832}
{"current_steps": 13630, "total_steps": 16324, "loss": 0.2356, "lr": 7.263933968547382e-06, "epoch": 1.669872285688034, "percentage": 83.5, "elapsed_time": "19:01:59", "remaining_time": "3:45:42", "throughput": 1330.8, "total_tokens": 91185240}
{"current_steps": 13635, "total_steps": 16324, "loss": 0.2047, "lr": 7.237665185256964e-06, "epoch": 1.6704848243545374, "percentage": 83.53, "elapsed_time": "19:03:01", "remaining_time": "3:45:25", "throughput": 1330.09, "total_tokens": 91218944}
{"current_steps": 13640, "total_steps": 16324, "loss": 0.1918, "lr": 7.2114402798381385e-06, "epoch": 1.6710973630210408, "percentage": 83.56, "elapsed_time": "19:04:02", "remaining_time": "3:45:07", "throughput": 1329.39, "total_tokens": 91252776}
{"current_steps": 13645, "total_steps": 16324, "loss": 0.2431, "lr": 7.185259279199913e-06, "epoch": 1.671709901687544, "percentage": 83.59, "elapsed_time": "19:05:03", "remaining_time": "3:44:48", "throughput": 1328.7, "total_tokens": 91286232}
{"current_steps": 13650, "total_steps": 16324, "loss": 0.205, "lr": 7.159122210206276e-06, "epoch": 1.6723224403540473, "percentage": 83.62, "elapsed_time": "19:06:05", "remaining_time": "3:44:30", "throughput": 1327.98, "total_tokens": 91319000}
{"current_steps": 13650, "total_steps": 16324, "eval_loss": 0.07880138605833054, "epoch": 1.6723224403540473, "percentage": 83.62, "elapsed_time": "19:06:25", "remaining_time": "3:44:34", "throughput": 1327.6, "total_tokens": 91319000}
{"current_steps": 13655, "total_steps": 16324, "loss": 0.2391, "lr": 7.1330290996761524e-06, "epoch": 1.6729349790205505, "percentage": 83.65, "elapsed_time": "19:07:30", "remaining_time": "3:44:17", "throughput": 1326.81, "total_tokens": 91352368}
{"current_steps": 13660, "total_steps": 16324, "loss": 0.2211, "lr": 7.106979974383327e-06, "epoch": 1.673547517687054, "percentage": 83.68, "elapsed_time": "19:08:31", "remaining_time": "3:43:59", "throughput": 1326.12, "total_tokens": 91385488}
{"current_steps": 13665, "total_steps": 16324, "loss": 0.2213, "lr": 7.080974861056444e-06, "epoch": 1.6741600563535575, "percentage": 83.71, "elapsed_time": "19:09:32", "remaining_time": "3:43:40", "throughput": 1325.44, "total_tokens": 91418536}
{"current_steps": 13670, "total_steps": 16324, "loss": 0.1736, "lr": 7.055013786379034e-06, "epoch": 1.6747725950200607, "percentage": 83.74, "elapsed_time": "19:10:32", "remaining_time": "3:43:22", "throughput": 1324.76, "total_tokens": 91452184}
{"current_steps": 13675, "total_steps": 16324, "loss": 0.2208, "lr": 7.029096776989425e-06, "epoch": 1.675385133686564, "percentage": 83.77, "elapsed_time": "19:11:32", "remaining_time": "3:43:03", "throughput": 1324.1, "total_tokens": 91485488}
{"current_steps": 13680, "total_steps": 16324, "loss": 0.1727, "lr": 7.00322385948069e-06, "epoch": 1.6759976723530672, "percentage": 83.8, "elapsed_time": "19:12:32", "remaining_time": "3:42:45", "throughput": 1323.44, "total_tokens": 91519048}
{"current_steps": 13685, "total_steps": 16324, "loss": 0.2349, "lr": 6.97739506040071e-06, "epoch": 1.6766102110195706, "percentage": 83.83, "elapsed_time": "19:13:32", "remaining_time": "3:42:26", "throughput": 1322.77, "total_tokens": 91552104}
{"current_steps": 13690, "total_steps": 16324, "loss": 0.195, "lr": 6.951610406252085e-06, "epoch": 1.677222749686074, "percentage": 83.86, "elapsed_time": "19:14:32", "remaining_time": "3:42:08", "throughput": 1322.1, "total_tokens": 91585608}
{"current_steps": 13695, "total_steps": 16324, "loss": 0.2421, "lr": 6.925869923492084e-06, "epoch": 1.6778352883525773, "percentage": 83.89, "elapsed_time": "19:15:32", "remaining_time": "3:41:49", "throughput": 1321.44, "total_tokens": 91618816}
{"current_steps": 13700, "total_steps": 16324, "loss": 0.2343, "lr": 6.900173638532703e-06, "epoch": 1.6784478270190806, "percentage": 83.93, "elapsed_time": "19:16:32", "remaining_time": "3:41:31", "throughput": 1320.78, "total_tokens": 91652896}
{"current_steps": 13700, "total_steps": 16324, "eval_loss": 0.10617480427026749, "epoch": 1.6784478270190806, "percentage": 83.93, "elapsed_time": "19:16:52", "remaining_time": "3:41:34", "throughput": 1320.41, "total_tokens": 91652896}
{"current_steps": 13705, "total_steps": 16324, "loss": 0.2294, "lr": 6.874521577740556e-06, "epoch": 1.6790603656855838, "percentage": 83.96, "elapsed_time": "19:17:57", "remaining_time": "3:41:17", "throughput": 1319.64, "total_tokens": 91686192}
{"current_steps": 13710, "total_steps": 16324, "loss": 0.2198, "lr": 6.848913767436893e-06, "epoch": 1.6796729043520873, "percentage": 83.99, "elapsed_time": "19:18:57", "remaining_time": "3:40:58", "throughput": 1318.99, "total_tokens": 91719400}
{"current_steps": 13715, "total_steps": 16324, "loss": 0.1973, "lr": 6.823350233897541e-06, "epoch": 1.6802854430185905, "percentage": 84.02, "elapsed_time": "19:19:58", "remaining_time": "3:40:39", "throughput": 1318.32, "total_tokens": 91753776}
{"current_steps": 13720, "total_steps": 16324, "loss": 0.2014, "lr": 6.7978310033529056e-06, "epoch": 1.680897981685094, "percentage": 84.05, "elapsed_time": "19:20:58", "remaining_time": "3:40:20", "throughput": 1317.68, "total_tokens": 91787632}
{"current_steps": 13725, "total_steps": 16324, "loss": 0.1972, "lr": 6.7723561019879585e-06, "epoch": 1.6815105203515972, "percentage": 84.08, "elapsed_time": "19:21:58", "remaining_time": "3:40:02", "throughput": 1317.03, "total_tokens": 91821520}
{"current_steps": 13730, "total_steps": 16324, "loss": 0.2423, "lr": 6.7469255559421295e-06, "epoch": 1.6821230590181004, "percentage": 84.11, "elapsed_time": "19:22:59", "remaining_time": "3:39:43", "throughput": 1316.36, "total_tokens": 91854952}
{"current_steps": 13735, "total_steps": 16324, "loss": 0.1949, "lr": 6.721539391309389e-06, "epoch": 1.6827355976846037, "percentage": 84.14, "elapsed_time": "19:24:00", "remaining_time": "3:39:24", "throughput": 1315.69, "total_tokens": 91888520}
{"current_steps": 13740, "total_steps": 16324, "loss": 0.2018, "lr": 6.696197634138152e-06, "epoch": 1.6833481363511071, "percentage": 84.17, "elapsed_time": "19:25:00", "remaining_time": "3:39:05", "throughput": 1315.03, "total_tokens": 91921920}
{"current_steps": 13745, "total_steps": 16324, "loss": 0.177, "lr": 6.670900310431255e-06, "epoch": 1.6839606750176106, "percentage": 84.2, "elapsed_time": "19:26:01", "remaining_time": "3:38:47", "throughput": 1314.37, "total_tokens": 91955424}
{"current_steps": 13750, "total_steps": 16324, "loss": 0.2717, "lr": 6.645647446145942e-06, "epoch": 1.6845732136841138, "percentage": 84.23, "elapsed_time": "19:27:02", "remaining_time": "3:38:28", "throughput": 1313.7, "total_tokens": 91988016}
{"current_steps": 13750, "total_steps": 16324, "eval_loss": 0.05082106962800026, "epoch": 1.6845732136841138, "percentage": 84.23, "elapsed_time": "19:27:22", "remaining_time": "3:38:31", "throughput": 1313.32, "total_tokens": 91988016}
{"current_steps": 13755, "total_steps": 16324, "loss": 0.2036, "lr": 6.620439067193857e-06, "epoch": 1.685185752350617, "percentage": 84.26, "elapsed_time": "19:28:28", "remaining_time": "3:38:14", "throughput": 1312.55, "total_tokens": 92021256}
{"current_steps": 13760, "total_steps": 16324, "loss": 0.2085, "lr": 6.595275199440981e-06, "epoch": 1.6857982910171203, "percentage": 84.29, "elapsed_time": "19:29:29", "remaining_time": "3:37:55", "throughput": 1311.88, "total_tokens": 92054640}
{"current_steps": 13765, "total_steps": 16324, "loss": 0.1851, "lr": 6.570155868707645e-06, "epoch": 1.6864108296836238, "percentage": 84.32, "elapsed_time": "19:30:29", "remaining_time": "3:37:36", "throughput": 1311.24, "total_tokens": 92088368}
{"current_steps": 13770, "total_steps": 16324, "loss": 0.2536, "lr": 6.545081100768441e-06, "epoch": 1.6870233683501272, "percentage": 84.35, "elapsed_time": "19:31:29", "remaining_time": "3:37:17", "throughput": 1310.6, "total_tokens": 92121792}
{"current_steps": 13775, "total_steps": 16324, "loss": 0.1916, "lr": 6.520050921352272e-06, "epoch": 1.6876359070166305, "percentage": 84.38, "elapsed_time": "19:32:29", "remaining_time": "3:36:57", "throughput": 1309.95, "total_tokens": 92154800}
{"current_steps": 13780, "total_steps": 16324, "loss": 0.1581, "lr": 6.495065356142294e-06, "epoch": 1.6882484456831337, "percentage": 84.42, "elapsed_time": "19:33:29", "remaining_time": "3:36:38", "throughput": 1309.33, "total_tokens": 92189080}
{"current_steps": 13785, "total_steps": 16324, "loss": 0.1999, "lr": 6.4701244307758364e-06, "epoch": 1.688860984349637, "percentage": 84.45, "elapsed_time": "19:34:28", "remaining_time": "3:36:19", "throughput": 1308.69, "total_tokens": 92221832}
{"current_steps": 13790, "total_steps": 16324, "loss": 0.1753, "lr": 6.44522817084447e-06, "epoch": 1.6894735230161404, "percentage": 84.48, "elapsed_time": "19:35:28", "remaining_time": "3:36:00", "throughput": 1308.07, "total_tokens": 92256128}
{"current_steps": 13795, "total_steps": 16324, "loss": 0.2252, "lr": 6.4203766018939235e-06, "epoch": 1.6900860616826439, "percentage": 84.51, "elapsed_time": "19:36:27", "remaining_time": "3:35:40", "throughput": 1307.45, "total_tokens": 92290008}
{"current_steps": 13800, "total_steps": 16324, "loss": 0.2281, "lr": 6.395569749424079e-06, "epoch": 1.690698600349147, "percentage": 84.54, "elapsed_time": "19:37:27", "remaining_time": "3:35:21", "throughput": 1306.82, "total_tokens": 92323496}
{"current_steps": 13800, "total_steps": 16324, "eval_loss": 0.05405490845441818, "epoch": 1.690698600349147, "percentage": 84.54, "elapsed_time": "19:37:47", "remaining_time": "3:35:24", "throughput": 1306.45, "total_tokens": 92323496}
{"current_steps": 13805, "total_steps": 16324, "loss": 0.2022, "lr": 6.370807638888898e-06, "epoch": 1.6913111390156503, "percentage": 84.57, "elapsed_time": "19:38:52", "remaining_time": "3:35:06", "throughput": 1305.72, "total_tokens": 92356752}
{"current_steps": 13810, "total_steps": 16324, "loss": 0.2124, "lr": 6.346090295696466e-06, "epoch": 1.6919236776821536, "percentage": 84.6, "elapsed_time": "19:39:52", "remaining_time": "3:34:47", "throughput": 1305.08, "total_tokens": 92390136}
{"current_steps": 13815, "total_steps": 16324, "loss": 0.2198, "lr": 6.321417745208941e-06, "epoch": 1.692536216348657, "percentage": 84.63, "elapsed_time": "19:40:53", "remaining_time": "3:34:27", "throughput": 1304.44, "total_tokens": 92423496}
{"current_steps": 13820, "total_steps": 16324, "loss": 0.1615, "lr": 6.296790012742493e-06, "epoch": 1.6931487550151605, "percentage": 84.66, "elapsed_time": "19:41:52", "remaining_time": "3:34:08", "throughput": 1303.83, "total_tokens": 92457784}
{"current_steps": 13825, "total_steps": 16324, "loss": 0.2166, "lr": 6.272207123567297e-06, "epoch": 1.6937612936816637, "percentage": 84.69, "elapsed_time": "19:42:52", "remaining_time": "3:33:48", "throughput": 1303.19, "total_tokens": 92490568}
{"current_steps": 13830, "total_steps": 16324, "loss": 0.2214, "lr": 6.2476691029075495e-06, "epoch": 1.694373832348167, "percentage": 84.72, "elapsed_time": "19:43:52", "remaining_time": "3:33:29", "throughput": 1302.56, "total_tokens": 92523880}
{"current_steps": 13835, "total_steps": 16324, "loss": 0.237, "lr": 6.223175975941409e-06, "epoch": 1.6949863710146702, "percentage": 84.75, "elapsed_time": "19:44:52", "remaining_time": "3:33:09", "throughput": 1301.93, "total_tokens": 92556928}
{"current_steps": 13840, "total_steps": 16324, "loss": 0.2038, "lr": 6.198727767800921e-06, "epoch": 1.6955989096811737, "percentage": 84.78, "elapsed_time": "19:45:52", "remaining_time": "3:32:50", "throughput": 1301.29, "total_tokens": 92590288}
{"current_steps": 13845, "total_steps": 16324, "loss": 0.2175, "lr": 6.174324503572088e-06, "epoch": 1.696211448347677, "percentage": 84.81, "elapsed_time": "19:46:52", "remaining_time": "3:32:30", "throughput": 1300.67, "total_tokens": 92624344}
{"current_steps": 13850, "total_steps": 16324, "loss": 0.228, "lr": 6.1499662082947774e-06, "epoch": 1.6968239870141804, "percentage": 84.84, "elapsed_time": "19:47:52", "remaining_time": "3:32:11", "throughput": 1300.04, "total_tokens": 92657456}
{"current_steps": 13850, "total_steps": 16324, "eval_loss": 0.09495604783296585, "epoch": 1.6968239870141804, "percentage": 84.84, "elapsed_time": "19:48:12", "remaining_time": "3:32:14", "throughput": 1299.69, "total_tokens": 92657456}
{"current_steps": 13855, "total_steps": 16324, "loss": 0.2283, "lr": 6.1256529069627255e-06, "epoch": 1.6974365256806836, "percentage": 84.88, "elapsed_time": "19:49:17", "remaining_time": "3:31:56", "throughput": 1298.96, "total_tokens": 92690832}
{"current_steps": 13860, "total_steps": 16324, "loss": 0.1816, "lr": 6.101384624523476e-06, "epoch": 1.6980490643471868, "percentage": 84.91, "elapsed_time": "19:50:16", "remaining_time": "3:31:36", "throughput": 1298.36, "total_tokens": 92725184}
{"current_steps": 13865, "total_steps": 16324, "loss": 0.1779, "lr": 6.0771613858784045e-06, "epoch": 1.6986616030136903, "percentage": 84.94, "elapsed_time": "19:51:16", "remaining_time": "3:31:16", "throughput": 1297.76, "total_tokens": 92759376}
{"current_steps": 13870, "total_steps": 16324, "loss": 0.1943, "lr": 6.0529832158826686e-06, "epoch": 1.6992741416801935, "percentage": 84.97, "elapsed_time": "19:52:15", "remaining_time": "3:30:56", "throughput": 1297.15, "total_tokens": 92793144}
{"current_steps": 13875, "total_steps": 16324, "loss": 0.203, "lr": 6.0288501393451546e-06, "epoch": 1.699886680346697, "percentage": 85.0, "elapsed_time": "19:53:15", "remaining_time": "3:30:36", "throughput": 1296.55, "total_tokens": 92826632}
{"current_steps": 13880, "total_steps": 16324, "loss": 0.2273, "lr": 6.004762181028495e-06, "epoch": 1.7004992190132002, "percentage": 85.03, "elapsed_time": "19:54:13", "remaining_time": "3:30:16", "throughput": 1295.96, "total_tokens": 92860552}
{"current_steps": 13885, "total_steps": 16324, "loss": 0.2133, "lr": 5.980719365649051e-06, "epoch": 1.7011117576797035, "percentage": 85.06, "elapsed_time": "19:55:13", "remaining_time": "3:29:56", "throughput": 1295.35, "total_tokens": 92894088}
{"current_steps": 13890, "total_steps": 16324, "loss": 0.1898, "lr": 5.956721717876817e-06, "epoch": 1.7017242963462067, "percentage": 85.09, "elapsed_time": "19:56:13", "remaining_time": "3:29:37", "throughput": 1294.74, "total_tokens": 92927200}
{"current_steps": 13895, "total_steps": 16324, "loss": 0.227, "lr": 5.932769262335475e-06, "epoch": 1.7023368350127102, "percentage": 85.12, "elapsed_time": "19:57:13", "remaining_time": "3:29:17", "throughput": 1294.13, "total_tokens": 92960976}
{"current_steps": 13900, "total_steps": 16324, "loss": 0.1996, "lr": 5.90886202360233e-06, "epoch": 1.7029493736792136, "percentage": 85.15, "elapsed_time": "19:58:12", "remaining_time": "3:28:57", "throughput": 1293.52, "total_tokens": 92994920}
{"current_steps": 13900, "total_steps": 16324, "eval_loss": 0.08521660417318344, "epoch": 1.7029493736792136, "percentage": 85.15, "elapsed_time": "19:58:32", "remaining_time": "3:29:00", "throughput": 1293.16, "total_tokens": 92994920}
{"current_steps": 13905, "total_steps": 16324, "loss": 0.2174, "lr": 5.885000026208287e-06, "epoch": 1.7035619123457169, "percentage": 85.18, "elapsed_time": "19:59:38", "remaining_time": "3:28:41", "throughput": 1292.43, "total_tokens": 93027872}
{"current_steps": 13910, "total_steps": 16324, "loss": 0.2163, "lr": 5.861183294637806e-06, "epoch": 1.70417445101222, "percentage": 85.21, "elapsed_time": "20:00:39", "remaining_time": "3:28:21", "throughput": 1291.82, "total_tokens": 93061384}
{"current_steps": 13915, "total_steps": 16324, "loss": 0.234, "lr": 5.837411853328944e-06, "epoch": 1.7047869896787233, "percentage": 85.24, "elapsed_time": "20:01:39", "remaining_time": "3:28:02", "throughput": 1291.19, "total_tokens": 93094288}
{"current_steps": 13920, "total_steps": 16324, "loss": 0.2196, "lr": 5.8136857266732625e-06, "epoch": 1.7053995283452268, "percentage": 85.27, "elapsed_time": "20:02:39", "remaining_time": "3:27:41", "throughput": 1290.58, "total_tokens": 93127288}
{"current_steps": 13925, "total_steps": 16324, "loss": 0.1984, "lr": 5.790004939015842e-06, "epoch": 1.7060120670117302, "percentage": 85.3, "elapsed_time": "20:03:39", "remaining_time": "3:27:21", "throughput": 1289.97, "total_tokens": 93161032}
{"current_steps": 13930, "total_steps": 16324, "loss": 0.2261, "lr": 5.766369514655201e-06, "epoch": 1.7066246056782335, "percentage": 85.33, "elapsed_time": "20:04:38", "remaining_time": "3:27:01", "throughput": 1289.37, "total_tokens": 93194120}
{"current_steps": 13935, "total_steps": 16324, "loss": 0.1541, "lr": 5.742779477843352e-06, "epoch": 1.7072371443447367, "percentage": 85.37, "elapsed_time": "20:05:38", "remaining_time": "3:26:41", "throughput": 1288.78, "total_tokens": 93228512}
{"current_steps": 13940, "total_steps": 16324, "loss": 0.2392, "lr": 5.7192348527857374e-06, "epoch": 1.70784968301124, "percentage": 85.4, "elapsed_time": "20:06:39", "remaining_time": "3:26:21", "throughput": 1288.16, "total_tokens": 93261680}
{"current_steps": 13945, "total_steps": 16324, "loss": 0.1866, "lr": 5.6957356636411606e-06, "epoch": 1.7084622216777434, "percentage": 85.43, "elapsed_time": "20:07:39", "remaining_time": "3:26:01", "throughput": 1287.56, "total_tokens": 93295648}
{"current_steps": 13950, "total_steps": 16324, "loss": 0.2211, "lr": 5.67228193452185e-06, "epoch": 1.7090747603442469, "percentage": 85.46, "elapsed_time": "20:08:39", "remaining_time": "3:25:41", "throughput": 1286.94, "total_tokens": 93328616}
{"current_steps": 13950, "total_steps": 16324, "eval_loss": 0.10547046363353729, "epoch": 1.7090747603442469, "percentage": 85.46, "elapsed_time": "20:08:59", "remaining_time": "3:25:44", "throughput": 1286.59, "total_tokens": 93328616}
{"current_steps": 13955, "total_steps": 16324, "loss": 0.195, "lr": 5.648873689493367e-06, "epoch": 1.7096872990107501, "percentage": 85.49, "elapsed_time": "20:10:04", "remaining_time": "3:25:25", "throughput": 1285.9, "total_tokens": 93362448}
{"current_steps": 13960, "total_steps": 16324, "loss": 0.19, "lr": 5.625510952574614e-06, "epoch": 1.7102998376772534, "percentage": 85.52, "elapsed_time": "20:11:04", "remaining_time": "3:25:05", "throughput": 1285.31, "total_tokens": 93396232}
{"current_steps": 13965, "total_steps": 16324, "loss": 0.2068, "lr": 5.602193747737766e-06, "epoch": 1.7109123763437566, "percentage": 85.55, "elapsed_time": "20:12:05", "remaining_time": "3:24:44", "throughput": 1284.69, "total_tokens": 93429472}
{"current_steps": 13970, "total_steps": 16324, "loss": 0.2117, "lr": 5.578922098908313e-06, "epoch": 1.71152491501026, "percentage": 85.58, "elapsed_time": "20:13:04", "remaining_time": "3:24:24", "throughput": 1284.1, "total_tokens": 93463024}
{"current_steps": 13975, "total_steps": 16324, "loss": 0.2151, "lr": 5.555696029965008e-06, "epoch": 1.7121374536767635, "percentage": 85.61, "elapsed_time": "20:14:04", "remaining_time": "3:24:04", "throughput": 1283.52, "total_tokens": 93496712}
{"current_steps": 13955, "total_steps": 16324, "loss": 0.1962, "lr": 5.648873689493367e-06, "epoch": 1.7096872990107501, "percentage": 85.49, "elapsed_time": "0:02:15", "remaining_time": "0:00:23", "throughput": 687931.8, "total_tokens": 93362448}
{"current_steps": 13960, "total_steps": 16324, "loss": 0.1886, "lr": 5.625510952574614e-06, "epoch": 1.7102998376772534, "percentage": 85.52, "elapsed_time": "0:03:14", "remaining_time": "0:00:32", "throughput": 479315.28, "total_tokens": 93396232}
{"current_steps": 13965, "total_steps": 16324, "loss": 0.2069, "lr": 5.602193747737766e-06, "epoch": 1.7109123763437566, "percentage": 85.55, "elapsed_time": "0:04:13", "remaining_time": "0:00:42", "throughput": 368796.47, "total_tokens": 93429472}
{"current_steps": 13970, "total_steps": 16324, "loss": 0.2115, "lr": 5.578922098908313e-06, "epoch": 1.71152491501026, "percentage": 85.58, "elapsed_time": "0:05:12", "remaining_time": "0:00:52", "throughput": 299499.76, "total_tokens": 93463024}
{"current_steps": 14040, "total_steps": 16324, "loss": 0.1793, "lr": 5.257915682937914e-06, "epoch": 1.7201004563413065, "percentage": 86.01, "elapsed_time": "0:19:43", "remaining_time": "0:03:12", "throughput": 79399.37, "total_tokens": 93934912}
{"current_steps": 14045, "total_steps": 16324, "loss": 0.217, "lr": 5.235330213432615e-06, "epoch": 1.7207129950078097, "percentage": 86.04, "elapsed_time": "0:20:41", "remaining_time": "0:03:21", "throughput": 75709.78, "total_tokens": 93968800}
{"current_steps": 14050, "total_steps": 16324, "loss": 0.2246, "lr": 5.212790676368568e-06, "epoch": 1.7213255336743132, "percentage": 86.07, "elapsed_time": "0:21:38", "remaining_time": "0:03:30", "throughput": 72368.52, "total_tokens": 94002480}
{"current_steps": 14050, "total_steps": 16324, "eval_loss": 0.10589968413114548, "epoch": 1.7213255336743132, "percentage": 86.07, "elapsed_time": "0:21:58", "remaining_time": "0:03:33", "throughput": 71316.2, "total_tokens": 94002480}
{"current_steps": 14055, "total_steps": 16324, "loss": 0.2239, "lr": 5.190297094873309e-06, "epoch": 1.7219380723408166, "percentage": 86.1, "elapsed_time": "0:23:03", "remaining_time": "0:03:43", "throughput": 67991.19, "total_tokens": 94036472}
{"current_steps": 14060, "total_steps": 16324, "loss": 0.1969, "lr": 5.16784949202721e-06, "epoch": 1.7225506110073199, "percentage": 86.13, "elapsed_time": "0:24:02", "remaining_time": "0:03:52", "throughput": 65223.81, "total_tokens": 94070568}
{"current_steps": 14065, "total_steps": 16324, "loss": 0.2337, "lr": 5.145447890863475e-06, "epoch": 1.7231631496738231, "percentage": 86.16, "elapsed_time": "0:25:01", "remaining_time": "0:04:01", "throughput": 62688.41, "total_tokens": 94103832}
{"current_steps": 14070, "total_steps": 16324, "loss": 0.2054, "lr": 5.123092314368061e-06, "epoch": 1.7237756883403264, "percentage": 86.19, "elapsed_time": "0:26:01", "remaining_time": "0:04:10", "throughput": 60298.22, "total_tokens": 94137064}
{"current_steps": 14075, "total_steps": 16324, "loss": 0.2282, "lr": 5.100782785479746e-06, "epoch": 1.7243882270068298, "percentage": 86.22, "elapsed_time": "0:26:59", "remaining_time": "0:04:18", "throughput": 58131.72, "total_tokens": 94170640}
{"current_steps": 14080, "total_steps": 16324, "loss": 0.211, "lr": 5.078519327090048e-06, "epoch": 1.7250007656733333, "percentage": 86.25, "elapsed_time": "0:27:58", "remaining_time": "0:04:27", "throughput": 56131.22, "total_tokens": 94204568}
{"current_steps": 14085, "total_steps": 16324, "loss": 0.2282, "lr": 5.056301962043219e-06, "epoch": 1.7256133043398365, "percentage": 86.28, "elapsed_time": "0:28:57", "remaining_time": "0:04:36", "throughput": 54240.39, "total_tokens": 94237824}
{"current_steps": 14090, "total_steps": 16324, "loss": 0.2472, "lr": 5.034130713136187e-06, "epoch": 1.7262258430063397, "percentage": 86.31, "elapsed_time": "0:29:55", "remaining_time": "0:04:44", "throughput": 52506.4, "total_tokens": 94271016}
{"current_steps": 14095, "total_steps": 16324, "loss": 0.1795, "lr": 5.012005603118597e-06, "epoch": 1.726838381672843, "percentage": 86.35, "elapsed_time": "0:30:53", "remaining_time": "0:04:53", "throughput": 50885.36, "total_tokens": 94305600}
{"current_steps": 14100, "total_steps": 16324, "loss": 0.2114, "lr": 4.989926654692728e-06, "epoch": 1.7274509203393464, "percentage": 86.38, "elapsed_time": "0:31:52", "remaining_time": "0:05:01", "throughput": 49340.12, "total_tokens": 94338656}
{"current_steps": 14100, "total_steps": 16324, "eval_loss": 0.08303268253803253, "epoch": 1.7274509203393464, "percentage": 86.38, "elapsed_time": "0:32:10", "remaining_time": "0:05:04", "throughput": 48857.68, "total_tokens": 94338656}
{"current_steps": 14105, "total_steps": 16324, "loss": 0.2181, "lr": 4.967893890513498e-06, "epoch": 1.72806345900585, "percentage": 86.41, "elapsed_time": "0:33:14", "remaining_time": "0:05:13", "throughput": 47321.14, "total_tokens": 94371872}
{"current_steps": 14110, "total_steps": 16324, "loss": 0.2158, "lr": 4.945907333188432e-06, "epoch": 1.7286759976723531, "percentage": 86.44, "elapsed_time": "0:34:12", "remaining_time": "0:05:22", "throughput": 45999.26, "total_tokens": 94405304}
{"current_steps": 14115, "total_steps": 16324, "loss": 0.1915, "lr": 4.923967005277658e-06, "epoch": 1.7292885363388564, "percentage": 86.47, "elapsed_time": "0:35:10", "remaining_time": "0:05:30", "throughput": 44745.81, "total_tokens": 94439304}
{"current_steps": 14120, "total_steps": 16324, "loss": 0.1906, "lr": 4.902072929293866e-06, "epoch": 1.7299010750053596, "percentage": 86.5, "elapsed_time": "0:36:08", "remaining_time": "0:05:38", "throughput": 43568.53, "total_tokens": 94473400}
{"current_steps": 14125, "total_steps": 16324, "loss": 0.2117, "lr": 4.8802251277022645e-06, "epoch": 1.730513613671863, "percentage": 86.53, "elapsed_time": "0:37:06", "remaining_time": "0:05:46", "throughput": 42440.99, "total_tokens": 94506920}
{"current_steps": 14130, "total_steps": 16324, "loss": 0.2001, "lr": 4.858423622920605e-06, "epoch": 1.7311261523383663, "percentage": 86.56, "elapsed_time": "0:38:06", "remaining_time": "0:05:55", "throughput": 41349.09, "total_tokens": 94540784}
{"current_steps": 14135, "total_steps": 16324, "loss": 0.1964, "lr": 4.836668437319125e-06, "epoch": 1.7317386910048698, "percentage": 86.59, "elapsed_time": "0:39:04", "remaining_time": "0:06:03", "throughput": 40332.77, "total_tokens": 94574920}
{"current_steps": 14140, "total_steps": 16324, "loss": 0.1883, "lr": 4.814959593220531e-06, "epoch": 1.732351229671373, "percentage": 86.62, "elapsed_time": "0:40:03", "remaining_time": "0:06:11", "throughput": 39354.89, "total_tokens": 94608760}
{"current_steps": 14145, "total_steps": 16324, "loss": 0.1621, "lr": 4.793297112899997e-06, "epoch": 1.7329637683378762, "percentage": 86.65, "elapsed_time": "0:41:04", "remaining_time": "0:06:19", "throughput": 38407.59, "total_tokens": 94642800}
{"current_steps": 14150, "total_steps": 16324, "loss": 0.2018, "lr": 4.771681018585089e-06, "epoch": 1.7335763070043797, "percentage": 86.68, "elapsed_time": "0:42:02", "remaining_time": "0:06:27", "throughput": 37527.68, "total_tokens": 94676200}
{"current_steps": 14150, "total_steps": 16324, "eval_loss": 0.06674149632453918, "epoch": 1.7335763070043797, "percentage": 86.68, "elapsed_time": "0:42:22", "remaining_time": "0:06:30", "throughput": 37243.59, "total_tokens": 94676200}
{"current_steps": 14155, "total_steps": 16324, "loss": 0.2392, "lr": 4.750111332455825e-06, "epoch": 1.734188845670883, "percentage": 86.71, "elapsed_time": "0:43:26", "remaining_time": "0:06:39", "throughput": 36340.27, "total_tokens": 94709600}
{"current_steps": 14160, "total_steps": 16324, "loss": 0.2236, "lr": 4.728588076644547e-06, "epoch": 1.7348013843373864, "percentage": 86.74, "elapsed_time": "0:44:26", "remaining_time": "0:06:47", "throughput": 35528.76, "total_tokens": 94743392}
{"current_steps": 14165, "total_steps": 16324, "loss": 0.2583, "lr": 4.707111273236004e-06, "epoch": 1.7354139230038896, "percentage": 86.77, "elapsed_time": "0:45:26", "remaining_time": "0:06:55", "throughput": 34763.8, "total_tokens": 94776848}
{"current_steps": 14170, "total_steps": 16324, "loss": 0.2111, "lr": 4.685680944267257e-06, "epoch": 1.7360264616703929, "percentage": 86.8, "elapsed_time": "0:46:24", "remaining_time": "0:07:03", "throughput": 34046.65, "total_tokens": 94810360}
{"current_steps": 14175, "total_steps": 16324, "loss": 0.1759, "lr": 4.6642971117277e-06, "epoch": 1.736639000336896, "percentage": 86.84, "elapsed_time": "0:47:22", "remaining_time": "0:07:11", "throughput": 33361.08, "total_tokens": 94844688}
{"current_steps": 14180, "total_steps": 16324, "loss": 0.204, "lr": 4.6429597975589765e-06, "epoch": 1.7372515390033996, "percentage": 86.87, "elapsed_time": "0:48:22", "remaining_time": "0:07:18", "throughput": 32686.86, "total_tokens": 94878096}
{"current_steps": 14185, "total_steps": 16324, "loss": 0.1906, "lr": 4.6216690236550454e-06, "epoch": 1.737864077669903, "percentage": 86.9, "elapsed_time": "0:49:20", "remaining_time": "0:07:26", "throughput": 32054.21, "total_tokens": 94911424}
{"current_steps": 14190, "total_steps": 16324, "loss": 0.1868, "lr": 4.600424811862098e-06, "epoch": 1.7384766163364063, "percentage": 86.93, "elapsed_time": "0:50:19", "remaining_time": "0:07:34", "throughput": 31442.77, "total_tokens": 94945448}
{"current_steps": 14195, "total_steps": 16324, "loss": 0.2381, "lr": 4.579227183978518e-06, "epoch": 1.7390891550029095, "percentage": 86.96, "elapsed_time": "0:51:19", "remaining_time": "0:07:41", "throughput": 30843.45, "total_tokens": 94979056}
{"current_steps": 14200, "total_steps": 16324, "loss": 0.2196, "lr": 4.5580761617549205e-06, "epoch": 1.7397016936694127, "percentage": 86.99, "elapsed_time": "0:52:17", "remaining_time": "0:07:49", "throughput": 30283.79, "total_tokens": 95012776}
{"current_steps": 14200, "total_steps": 16324, "eval_loss": 0.09127970039844513, "epoch": 1.7397016936694127, "percentage": 86.99, "elapsed_time": "0:52:36", "remaining_time": "0:07:52", "throughput": 30100.77, "total_tokens": 95012776}
{"current_steps": 14205, "total_steps": 16324, "loss": 0.1893, "lr": 4.5369717668941155e-06, "epoch": 1.7403142323359162, "percentage": 87.02, "elapsed_time": "0:53:40", "remaining_time": "0:08:00", "throughput": 29511.55, "total_tokens": 95045912}
{"current_steps": 14210, "total_steps": 16324, "loss": 0.2287, "lr": 4.51591402105101e-06, "epoch": 1.7409267710024197, "percentage": 87.05, "elapsed_time": "0:54:39", "remaining_time": "0:08:07", "throughput": 28988.26, "total_tokens": 95078840}
{"current_steps": 14215, "total_steps": 16324, "loss": 0.1898, "lr": 4.494902945832702e-06, "epoch": 1.741539309668923, "percentage": 87.08, "elapsed_time": "0:55:37", "remaining_time": "0:08:15", "throughput": 28497.35, "total_tokens": 95111880}
{"current_steps": 14220, "total_steps": 16324, "loss": 0.2223, "lr": 4.47393856279838e-06, "epoch": 1.7421518483354261, "percentage": 87.11, "elapsed_time": "0:56:35", "remaining_time": "0:08:22", "throughput": 28024.97, "total_tokens": 95145128}
{"current_steps": 14225, "total_steps": 16324, "loss": 0.2186, "lr": 4.4530208934593255e-06, "epoch": 1.7427643870019294, "percentage": 87.14, "elapsed_time": "0:57:33", "remaining_time": "0:08:29", "throughput": 27559.36, "total_tokens": 95178440}
{"current_steps": 14230, "total_steps": 16324, "loss": 0.1852, "lr": 4.432149959278886e-06, "epoch": 1.7433769256684328, "percentage": 87.17, "elapsed_time": "0:58:31", "remaining_time": "0:08:36", "throughput": 27114.88, "total_tokens": 95212352}
{"current_steps": 14235, "total_steps": 16324, "loss": 0.2282, "lr": 4.411325781672432e-06, "epoch": 1.7439894643349363, "percentage": 87.2, "elapsed_time": "0:59:29", "remaining_time": "0:08:43", "throughput": 26684.3, "total_tokens": 95245744}
{"current_steps": 14240, "total_steps": 16324, "loss": 0.1665, "lr": 4.390548382007398e-06, "epoch": 1.7446020030014395, "percentage": 87.23, "elapsed_time": "1:00:28", "remaining_time": "0:08:50", "throughput": 26261.39, "total_tokens": 95280344}
{"current_steps": 14245, "total_steps": 16324, "loss": 0.2084, "lr": 4.369817781603208e-06, "epoch": 1.7452145416679428, "percentage": 87.26, "elapsed_time": "1:01:26", "remaining_time": "0:08:57", "throughput": 25856.19, "total_tokens": 95313328}
{"current_steps": 14250, "total_steps": 16324, "loss": 0.2155, "lr": 4.349134001731236e-06, "epoch": 1.745827080334446, "percentage": 87.29, "elapsed_time": "1:02:24", "remaining_time": "0:09:04", "throughput": 25463.04, "total_tokens": 95347128}
{"current_steps": 14250, "total_steps": 16324, "eval_loss": 0.09919190406799316, "epoch": 1.745827080334446, "percentage": 87.29, "elapsed_time": "1:02:43", "remaining_time": "0:09:07", "throughput": 25334.43, "total_tokens": 95347128}
{"current_steps": 14255, "total_steps": 16324, "loss": 0.1812, "lr": 4.3284970636148555e-06, "epoch": 1.7464396190009495, "percentage": 87.33, "elapsed_time": "1:03:47", "remaining_time": "0:09:15", "throughput": 24922.01, "total_tokens": 95380888}
{"current_steps": 14260, "total_steps": 16324, "loss": 0.2123, "lr": 4.307906988429366e-06, "epoch": 1.747052157667453, "percentage": 87.36, "elapsed_time": "1:04:44", "remaining_time": "0:09:22", "throughput": 24562.7, "total_tokens": 95414592}
{"current_steps": 14265, "total_steps": 16324, "loss": 0.2305, "lr": 4.287363797301947e-06, "epoch": 1.7476646963339562, "percentage": 87.39, "elapsed_time": "1:05:42", "remaining_time": "0:09:28", "throughput": 24212.55, "total_tokens": 95448016}
{"current_steps": 14270, "total_steps": 16324, "loss": 0.1886, "lr": 4.266867511311718e-06, "epoch": 1.7482772350004594, "percentage": 87.42, "elapsed_time": "1:06:40", "remaining_time": "0:09:35", "throughput": 23864.84, "total_tokens": 95481752}
{"current_steps": 14275, "total_steps": 16324, "loss": 0.1991, "lr": 4.246418151489639e-06, "epoch": 1.7488897736669626, "percentage": 87.45, "elapsed_time": "1:07:38", "remaining_time": "0:09:42", "throughput": 23532.62, "total_tokens": 95515320}
{"current_steps": 14280, "total_steps": 16324, "loss": 0.2442, "lr": 4.226015738818551e-06, "epoch": 1.749502312333466, "percentage": 87.48, "elapsed_time": "1:08:36", "remaining_time": "0:09:49", "throughput": 23209.39, "total_tokens": 95548384}
{"current_steps": 14285, "total_steps": 16324, "loss": 0.2071, "lr": 4.205660294233077e-06, "epoch": 1.7501148509999693, "percentage": 87.51, "elapsed_time": "1:09:35", "remaining_time": "0:09:56", "throughput": 22889.91, "total_tokens": 95582680}
{"current_steps": 14290, "total_steps": 16324, "loss": 0.2228, "lr": 4.185351838619683e-06, "epoch": 1.7507273896664728, "percentage": 87.54, "elapsed_time": "1:10:34", "remaining_time": "0:10:02", "throughput": 22582.92, "total_tokens": 95616160}
{"current_steps": 14295, "total_steps": 16324, "loss": 0.2195, "lr": 4.165090392816612e-06, "epoch": 1.751339928332976, "percentage": 87.57, "elapsed_time": "1:11:32", "remaining_time": "0:10:09", "throughput": 22283.92, "total_tokens": 95649280}
{"current_steps": 14300, "total_steps": 16324, "loss": 0.2333, "lr": 4.14487597761386e-06, "epoch": 1.7519524669994793, "percentage": 87.6, "elapsed_time": "1:12:30", "remaining_time": "0:10:15", "throughput": 21993.94, "total_tokens": 95683048}
{"current_steps": 14300, "total_steps": 16324, "eval_loss": 0.12027577310800552, "epoch": 1.7519524669994793, "percentage": 87.6, "elapsed_time": "1:12:49", "remaining_time": "0:10:18", "throughput": 21896.66, "total_tokens": 95683048}
{"current_steps": 14305, "total_steps": 16324, "loss": 0.2071, "lr": 4.124708613753186e-06, "epoch": 1.7525650056659825, "percentage": 87.63, "elapsed_time": "1:13:54", "remaining_time": "0:10:25", "throughput": 21585.01, "total_tokens": 95716536}
{"current_steps": 14310, "total_steps": 16324, "loss": 0.2135, "lr": 4.104588321928043e-06, "epoch": 1.753177544332486, "percentage": 87.66, "elapsed_time": "1:14:52", "remaining_time": "0:10:32", "throughput": 21311.65, "total_tokens": 95750128}
{"current_steps": 14315, "total_steps": 16324, "loss": 0.2196, "lr": 4.0845151227836165e-06, "epoch": 1.7537900829989894, "percentage": 87.69, "elapsed_time": "1:15:51", "remaining_time": "0:10:38", "throughput": 21045.75, "total_tokens": 95783912}
{"current_steps": 14320, "total_steps": 16324, "loss": 0.1735, "lr": 4.06448903691673e-06, "epoch": 1.7544026216654927, "percentage": 87.72, "elapsed_time": "1:16:50", "remaining_time": "0:10:45", "throughput": 20782.98, "total_tokens": 95817888}
{"current_steps": 14325, "total_steps": 16324, "loss": 0.2017, "lr": 4.044510084875902e-06, "epoch": 1.755015160331996, "percentage": 87.75, "elapsed_time": "1:17:48", "remaining_time": "0:10:51", "throughput": 20532.37, "total_tokens": 95851648}
{"current_steps": 14330, "total_steps": 16324, "loss": 0.2393, "lr": 4.0245782871612735e-06, "epoch": 1.7556276989984991, "percentage": 87.78, "elapsed_time": "1:18:46", "remaining_time": "0:10:57", "throughput": 20287.7, "total_tokens": 95885200}
{"current_steps": 14335, "total_steps": 16324, "loss": 0.2201, "lr": 4.004693664224607e-06, "epoch": 1.7562402376650026, "percentage": 87.82, "elapsed_time": "1:19:44", "remaining_time": "0:11:03", "throughput": 20047.55, "total_tokens": 95918408}
{"current_steps": 14340, "total_steps": 16324, "loss": 0.2115, "lr": 3.984856236469237e-06, "epoch": 1.756852776331506, "percentage": 87.85, "elapsed_time": "1:20:42", "remaining_time": "0:11:09", "throughput": 19814.12, "total_tokens": 95951624}
{"current_steps": 14345, "total_steps": 16324, "loss": 0.2459, "lr": 3.965066024250097e-06, "epoch": 1.7574653149980093, "percentage": 87.88, "elapsed_time": "1:21:40", "remaining_time": "0:11:16", "throughput": 19587.71, "total_tokens": 95984584}
{"current_steps": 14350, "total_steps": 16324, "loss": 0.201, "lr": 3.945323047873678e-06, "epoch": 1.7580778536645125, "percentage": 87.91, "elapsed_time": "1:22:38", "remaining_time": "0:11:22", "throughput": 19362.69, "total_tokens": 96018648}
{"current_steps": 14350, "total_steps": 16324, "eval_loss": 0.11044047772884369, "epoch": 1.7580778536645125, "percentage": 87.91, "elapsed_time": "1:22:58", "remaining_time": "0:11:24", "throughput": 19288.55, "total_tokens": 96018648}
{"current_steps": 14355, "total_steps": 16324, "loss": 0.2159, "lr": 3.925627327597964e-06, "epoch": 1.7586903923310158, "percentage": 87.94, "elapsed_time": "1:24:00", "remaining_time": "0:11:31", "throughput": 19054.17, "total_tokens": 96051648}
{"current_steps": 14360, "total_steps": 16324, "loss": 0.1974, "lr": 3.905978883632488e-06, "epoch": 1.7593029309975192, "percentage": 87.97, "elapsed_time": "1:24:58", "remaining_time": "0:11:37", "throughput": 18845.19, "total_tokens": 96084984}
{"current_steps": 14365, "total_steps": 16324, "loss": 0.2076, "lr": 3.886377736138269e-06, "epoch": 1.7599154696640227, "percentage": 88.0, "elapsed_time": "1:25:57", "remaining_time": "0:11:43", "throughput": 18636.96, "total_tokens": 96118432}
{"current_steps": 14370, "total_steps": 16324, "loss": 0.1962, "lr": 3.866823905227773e-06, "epoch": 1.760528008330526, "percentage": 88.03, "elapsed_time": "1:26:55", "remaining_time": "0:11:49", "throughput": 18435.6, "total_tokens": 96151920}
{"current_steps": 14375, "total_steps": 16324, "loss": 0.1959, "lr": 3.847317410964935e-06, "epoch": 1.7611405469970292, "percentage": 88.06, "elapsed_time": "1:27:53", "remaining_time": "0:11:55", "throughput": 18237.96, "total_tokens": 96185448}
{"current_steps": 14380, "total_steps": 16324, "loss": 0.2026, "lr": 3.827858273365114e-06, "epoch": 1.7617530856635324, "percentage": 88.09, "elapsed_time": "1:28:53", "remaining_time": "0:12:00", "throughput": 18041.71, "total_tokens": 96219072}
{"current_steps": 14385, "total_steps": 16324, "loss": 0.2392, "lr": 3.8084465123950862e-06, "epoch": 1.7623656243300359, "percentage": 88.12, "elapsed_time": "1:29:51", "remaining_time": "0:12:06", "throughput": 17853.66, "total_tokens": 96252472}
{"current_steps": 14390, "total_steps": 16324, "loss": 0.2442, "lr": 3.789082147973e-06, "epoch": 1.7629781629965393, "percentage": 88.15, "elapsed_time": "1:30:49", "remaining_time": "0:12:12", "throughput": 17668.83, "total_tokens": 96286104}
{"current_steps": 14395, "total_steps": 16324, "loss": 0.1803, "lr": 3.7697651999683615e-06, "epoch": 1.7635907016630425, "percentage": 88.18, "elapsed_time": "1:31:48", "remaining_time": "0:12:18", "throughput": 17484.38, "total_tokens": 96319824}
{"current_steps": 14400, "total_steps": 16324, "loss": 0.1886, "lr": 3.7504956882020436e-06, "epoch": 1.7642032403295458, "percentage": 88.21, "elapsed_time": "1:32:47", "remaining_time": "0:12:23", "throughput": 17307.98, "total_tokens": 96353568}
{"current_steps": 14400, "total_steps": 16324, "eval_loss": 0.12810835242271423, "epoch": 1.7642032403295458, "percentage": 88.21, "elapsed_time": "1:33:06", "remaining_time": "0:12:26", "throughput": 17248.62, "total_tokens": 96353568}
{"current_steps": 14405, "total_steps": 16324, "loss": 0.1917, "lr": 3.7312736324462583e-06, "epoch": 1.764815778996049, "percentage": 88.24, "elapsed_time": "1:34:09", "remaining_time": "0:12:32", "throughput": 17062.36, "total_tokens": 96387800}
{"current_steps": 14410, "total_steps": 16324, "loss": 0.1684, "lr": 3.7120990524244782e-06, "epoch": 1.7654283176625525, "percentage": 88.27, "elapsed_time": "1:35:08", "remaining_time": "0:12:38", "throughput": 16891.18, "total_tokens": 96422160}
{"current_steps": 14415, "total_steps": 16324, "loss": 0.195, "lr": 3.6929719678115093e-06, "epoch": 1.7660408563290557, "percentage": 88.31, "elapsed_time": "1:36:06", "remaining_time": "0:12:43", "throughput": 16727.23, "total_tokens": 96456552}
{"current_steps": 14420, "total_steps": 16324, "loss": 0.1954, "lr": 3.6738923982333972e-06, "epoch": 1.7666533949955592, "percentage": 88.34, "elapsed_time": "1:37:04", "remaining_time": "0:12:49", "throughput": 16566.22, "total_tokens": 96489688}
{"current_steps": 14425, "total_steps": 16324, "loss": 0.2134, "lr": 3.654860363267432e-06, "epoch": 1.7672659336620624, "percentage": 88.37, "elapsed_time": "1:38:02", "remaining_time": "0:12:54", "throughput": 16407.96, "total_tokens": 96522744}
{"current_steps": 14430, "total_steps": 16324, "loss": 0.2143, "lr": 3.6358758824421357e-06, "epoch": 1.7678784723285657, "percentage": 88.4, "elapsed_time": "1:39:01", "remaining_time": "0:12:59", "throughput": 16251.23, "total_tokens": 96556248}
{"current_steps": 14435, "total_steps": 16324, "loss": 0.1869, "lr": 3.616938975237244e-06, "epoch": 1.7684910109950691, "percentage": 88.43, "elapsed_time": "1:39:59", "remaining_time": "0:13:05", "throughput": 16098.41, "total_tokens": 96589768}
{"current_steps": 14440, "total_steps": 16324, "loss": 0.2192, "lr": 3.598049661083669e-06, "epoch": 1.7691035496615723, "percentage": 88.46, "elapsed_time": "1:40:58", "remaining_time": "0:13:10", "throughput": 15949.43, "total_tokens": 96622984}
{"current_steps": 14445, "total_steps": 16324, "loss": 0.2049, "lr": 3.5792079593634785e-06, "epoch": 1.7697160883280758, "percentage": 88.49, "elapsed_time": "1:41:57", "remaining_time": "0:13:15", "throughput": 15801.06, "total_tokens": 96656136}
{"current_steps": 14450, "total_steps": 16324, "loss": 0.1834, "lr": 3.5604138894099027e-06, "epoch": 1.770328626994579, "percentage": 88.52, "elapsed_time": "1:42:55", "remaining_time": "0:13:20", "throughput": 15657.51, "total_tokens": 96689680}
{"current_steps": 14450, "total_steps": 16324, "eval_loss": 0.12905412912368774, "epoch": 1.770328626994579, "percentage": 88.52, "elapsed_time": "1:43:14", "remaining_time": "0:13:23", "throughput": 15609.59, "total_tokens": 96689680}
{"current_steps": 14455, "total_steps": 16324, "loss": 0.2021, "lr": 3.5416674705072985e-06, "epoch": 1.7709411656610823, "percentage": 88.55, "elapsed_time": "1:44:17", "remaining_time": "0:13:29", "throughput": 15457.29, "total_tokens": 96723296}
{"current_steps": 14460, "total_steps": 16324, "loss": 0.1756, "lr": 3.522968721891101e-06, "epoch": 1.7715537043275855, "percentage": 88.58, "elapsed_time": "1:45:17", "remaining_time": "0:13:34", "throughput": 15316.66, "total_tokens": 96757880}
{"current_steps": 14465, "total_steps": 16324, "loss": 0.2119, "lr": 3.5043176627478636e-06, "epoch": 1.772166242994089, "percentage": 88.61, "elapsed_time": "1:46:14", "remaining_time": "0:13:39", "throughput": 15183.26, "total_tokens": 96791424}
{"current_steps": 14470, "total_steps": 16324, "loss": 0.1666, "lr": 3.4857143122152046e-06, "epoch": 1.7727787816605924, "percentage": 88.64, "elapsed_time": "1:47:12", "remaining_time": "0:13:44", "throughput": 15052.29, "total_tokens": 96825400}
{"current_steps": 14475, "total_steps": 16324, "loss": 0.2343, "lr": 3.467158689381761e-06, "epoch": 1.7733913203270957, "percentage": 88.67, "elapsed_time": "1:48:12", "remaining_time": "0:13:49", "throughput": 14919.38, "total_tokens": 96858304}
{"current_steps": 14480, "total_steps": 16324, "loss": 0.1987, "lr": 3.4486508132872076e-06, "epoch": 1.774003858993599, "percentage": 88.7, "elapsed_time": "1:49:10", "remaining_time": "0:13:54", "throughput": 14792.54, "total_tokens": 96892408}
{"current_steps": 14485, "total_steps": 16324, "loss": 0.2021, "lr": 3.4301907029222534e-06, "epoch": 1.7746163976601022, "percentage": 88.73, "elapsed_time": "1:50:08", "remaining_time": "0:13:58", "throughput": 14667.61, "total_tokens": 96926456}
{"current_steps": 14490, "total_steps": 16324, "loss": 0.2406, "lr": 3.4117783772285637e-06, "epoch": 1.7752289363266056, "percentage": 88.77, "elapsed_time": "1:51:07", "remaining_time": "0:14:03", "throughput": 14541.95, "total_tokens": 96959368}
{"current_steps": 14495, "total_steps": 16324, "loss": 0.23, "lr": 3.3934138550987983e-06, "epoch": 1.775841474993109, "percentage": 88.8, "elapsed_time": "1:52:05", "remaining_time": "0:14:08", "throughput": 14422.3, "total_tokens": 96993128}
{"current_steps": 14500, "total_steps": 16324, "loss": 0.2138, "lr": 3.3750971553765333e-06, "epoch": 1.7764540136596123, "percentage": 88.83, "elapsed_time": "1:53:03", "remaining_time": "0:14:13", "throughput": 14302.94, "total_tokens": 97026536}
{"current_steps": 14500, "total_steps": 16324, "eval_loss": 0.12095509469509125, "epoch": 1.7764540136596123, "percentage": 88.83, "elapsed_time": "1:53:22", "remaining_time": "0:14:15", "throughput": 14263.41, "total_tokens": 97026536}
{"current_steps": 14505, "total_steps": 16324, "loss": 0.2085, "lr": 3.3568282968563015e-06, "epoch": 1.7770665523261155, "percentage": 88.86, "elapsed_time": "1:54:27", "remaining_time": "0:14:21", "throughput": 14133.32, "total_tokens": 97060344}
{"current_steps": 14510, "total_steps": 16324, "loss": 0.2155, "lr": 3.3386072982835524e-06, "epoch": 1.7776790909926188, "percentage": 88.89, "elapsed_time": "1:55:25", "remaining_time": "0:14:25", "throughput": 14019.31, "total_tokens": 97094160}
{"current_steps": 14515, "total_steps": 16324, "loss": 0.1951, "lr": 3.3204341783545977e-06, "epoch": 1.7782916296591222, "percentage": 88.92, "elapsed_time": "1:56:23", "remaining_time": "0:14:30", "throughput": 13907.51, "total_tokens": 97127784}
{"current_steps": 14520, "total_steps": 16324, "loss": 0.2075, "lr": 3.302308955716643e-06, "epoch": 1.7789041683256257, "percentage": 88.95, "elapsed_time": "1:57:22", "remaining_time": "0:14:35", "throughput": 13795.49, "total_tokens": 97161064}
{"current_steps": 14525, "total_steps": 16324, "loss": 0.1805, "lr": 3.2842316489677507e-06, "epoch": 1.779516706992129, "percentage": 88.98, "elapsed_time": "1:58:21", "remaining_time": "0:14:39", "throughput": 13687.41, "total_tokens": 97194760}
{"current_steps": 14530, "total_steps": 16324, "loss": 0.2141, "lr": 3.2662022766567935e-06, "epoch": 1.7801292456586322, "percentage": 89.01, "elapsed_time": "1:59:19", "remaining_time": "0:14:43", "throughput": 13580.79, "total_tokens": 97228744}
{"current_steps": 14535, "total_steps": 16324, "loss": 0.1904, "lr": 3.248220857283474e-06, "epoch": 1.7807417843251354, "percentage": 89.04, "elapsed_time": "2:00:19", "remaining_time": "0:14:48", "throughput": 13472.47, "total_tokens": 97262720}
{"current_steps": 14540, "total_steps": 16324, "loss": 0.2037, "lr": 3.230287409298305e-06, "epoch": 1.7813543229916389, "percentage": 89.07, "elapsed_time": "2:01:17", "remaining_time": "0:14:52", "throughput": 13368.98, "total_tokens": 97296152}
{"current_steps": 14545, "total_steps": 16324, "loss": 0.2289, "lr": 3.2124019511025616e-06, "epoch": 1.7819668616581423, "percentage": 89.1, "elapsed_time": "2:02:16", "remaining_time": "0:14:57", "throughput": 13267.09, "total_tokens": 97329360}
{"current_steps": 14550, "total_steps": 16324, "loss": 0.1703, "lr": 3.1945645010482696e-06, "epoch": 1.7825794003246456, "percentage": 89.13, "elapsed_time": "2:03:15", "remaining_time": "0:15:01", "throughput": 13165.34, "total_tokens": 97363272}
{"current_steps": 14550, "total_steps": 16324, "eval_loss": 0.152376189827919, "epoch": 1.7825794003246456, "percentage": 89.13, "elapsed_time": "2:03:35", "remaining_time": "0:15:04", "throughput": 13130.44, "total_tokens": 97363272}
{"current_steps": 14555, "total_steps": 16324, "loss": 0.1921, "lr": 3.176775077438199e-06, "epoch": 1.7831919389911488, "percentage": 89.16, "elapsed_time": "2:04:39", "remaining_time": "0:15:09", "throughput": 13021.38, "total_tokens": 97396880}
{"current_steps": 14560, "total_steps": 16324, "loss": 0.2139, "lr": 3.159033698525854e-06, "epoch": 1.783804477657652, "percentage": 89.19, "elapsed_time": "2:05:38", "remaining_time": "0:15:13", "throughput": 12924.49, "total_tokens": 97430608}
{"current_steps": 14565, "total_steps": 16324, "loss": 0.2144, "lr": 3.1413403825154285e-06, "epoch": 1.7844170163241555, "percentage": 89.22, "elapsed_time": "2:06:36", "remaining_time": "0:15:17", "throughput": 12829.72, "total_tokens": 97463904}
{"current_steps": 14570, "total_steps": 16324, "loss": 0.2079, "lr": 3.1236951475618002e-06, "epoch": 1.7850295549906587, "percentage": 89.26, "elapsed_time": "2:07:36", "remaining_time": "0:15:21", "throughput": 12734.16, "total_tokens": 97497480}
{"current_steps": 14575, "total_steps": 16324, "loss": 0.2057, "lr": 3.1060980117705085e-06, "epoch": 1.7856420936571622, "percentage": 89.29, "elapsed_time": "2:08:34", "remaining_time": "0:15:25", "throughput": 12642.87, "total_tokens": 97531168}
{"current_steps": 14580, "total_steps": 16324, "loss": 0.2223, "lr": 3.0885489931977596e-06, "epoch": 1.7862546323236654, "percentage": 89.32, "elapsed_time": "2:09:32", "remaining_time": "0:15:29", "throughput": 12552.81, "total_tokens": 97564792}
{"current_steps": 14585, "total_steps": 16324, "loss": 0.1863, "lr": 3.0710481098503608e-06, "epoch": 1.7868671709901687, "percentage": 89.35, "elapsed_time": "2:10:32", "remaining_time": "0:15:33", "throughput": 12461.44, "total_tokens": 97598536}
{"current_steps": 14590, "total_steps": 16324, "loss": 0.2025, "lr": 3.0535953796857364e-06, "epoch": 1.787479709656672, "percentage": 89.38, "elapsed_time": "2:11:30", "remaining_time": "0:15:37", "throughput": 12373.33, "total_tokens": 97632352}
{"current_steps": 14595, "total_steps": 16324, "loss": 0.2024, "lr": 3.036190820611906e-06, "epoch": 1.7880922483231754, "percentage": 89.41, "elapsed_time": "2:12:28", "remaining_time": "0:15:41", "throughput": 12286.72, "total_tokens": 97665488}
{"current_steps": 14600, "total_steps": 16324, "loss": 0.2117, "lr": 3.0188344504874734e-06, "epoch": 1.7887047869896788, "percentage": 89.44, "elapsed_time": "2:13:28", "remaining_time": "0:15:45", "throughput": 12199.52, "total_tokens": 97698864}
{"current_steps": 14600, "total_steps": 16324, "eval_loss": 0.13145245611667633, "epoch": 1.7887047869896788, "percentage": 89.44, "elapsed_time": "2:13:47", "remaining_time": "0:15:47", "throughput": 12170.35, "total_tokens": 97698864}
{"current_steps": 14605, "total_steps": 16324, "loss": 0.1954, "lr": 3.0015262871215587e-06, "epoch": 1.789317325656182, "percentage": 89.47, "elapsed_time": "2:14:51", "remaining_time": "0:15:52", "throughput": 12078.51, "total_tokens": 97732008}
{"current_steps": 14610, "total_steps": 16324, "loss": 0.2309, "lr": 2.9842663482738566e-06, "epoch": 1.7899298643226853, "percentage": 89.5, "elapsed_time": "2:15:49", "remaining_time": "0:15:56", "throughput": 11995.93, "total_tokens": 97764984}
{"current_steps": 14615, "total_steps": 16324, "loss": 0.2307, "lr": 2.9670546516545717e-06, "epoch": 1.7905424029891885, "percentage": 89.53, "elapsed_time": "2:16:49", "remaining_time": "0:15:59", "throughput": 11912.71, "total_tokens": 97797792}
{"current_steps": 14620, "total_steps": 16324, "loss": 0.2041, "lr": 2.9498912149243827e-06, "epoch": 1.791154941655692, "percentage": 89.56, "elapsed_time": "2:17:47", "remaining_time": "0:16:03", "throughput": 11833.04, "total_tokens": 97831144}
{"current_steps": 14625, "total_steps": 16324, "loss": 0.2428, "lr": 2.9327760556944694e-06, "epoch": 1.7917674803221955, "percentage": 89.59, "elapsed_time": "2:18:45", "remaining_time": "0:16:07", "throughput": 11754.78, "total_tokens": 97864296}
{"current_steps": 14630, "total_steps": 16324, "loss": 0.2153, "lr": 2.9157091915264944e-06, "epoch": 1.7923800189886987, "percentage": 89.62, "elapsed_time": "2:19:44", "remaining_time": "0:16:10", "throughput": 11676.36, "total_tokens": 97897536}
{"current_steps": 14635, "total_steps": 16324, "loss": 0.1993, "lr": 2.898690639932522e-06, "epoch": 1.792992557655202, "percentage": 89.65, "elapsed_time": "2:20:42", "remaining_time": "0:16:14", "throughput": 11599.69, "total_tokens": 97931168}
{"current_steps": 14640, "total_steps": 16324, "loss": 0.2056, "lr": 2.881720418375061e-06, "epoch": 1.7936050963217052, "percentage": 89.68, "elapsed_time": "2:21:40", "remaining_time": "0:16:17", "throughput": 11524.31, "total_tokens": 97965328}
{"current_steps": 14645, "total_steps": 16324, "loss": 0.1727, "lr": 2.8647985442670444e-06, "epoch": 1.7942176349882086, "percentage": 89.71, "elapsed_time": "2:22:40", "remaining_time": "0:16:21", "throughput": 11448.17, "total_tokens": 97999096}
{"current_steps": 14650, "total_steps": 16324, "loss": 0.1927, "lr": 2.847925034971788e-06, "epoch": 1.794830173654712, "percentage": 89.75, "elapsed_time": "2:23:38", "remaining_time": "0:16:24", "throughput": 11375.07, "total_tokens": 98032608}
{"current_steps": 14650, "total_steps": 16324, "eval_loss": 0.1303679496049881, "epoch": 1.794830173654712, "percentage": 89.75, "elapsed_time": "2:23:57", "remaining_time": "0:16:26", "throughput": 11350.11, "total_tokens": 98032608}
{"current_steps": 14655, "total_steps": 16324, "loss": 0.1957, "lr": 2.8310999078029754e-06, "epoch": 1.7954427123212153, "percentage": 89.78, "elapsed_time": "2:25:01", "remaining_time": "0:16:30", "throughput": 11270.24, "total_tokens": 98066712}
{"current_steps": 14660, "total_steps": 16324, "loss": 0.1854, "lr": 2.814323180024647e-06, "epoch": 1.7960552509877186, "percentage": 89.81, "elapsed_time": "2:26:01", "remaining_time": "0:16:34", "throughput": 11197.21, "total_tokens": 98100856}
{"current_steps": 14665, "total_steps": 16324, "loss": 0.1604, "lr": 2.797594868851183e-06, "epoch": 1.7966677896542218, "percentage": 89.84, "elapsed_time": "2:26:59", "remaining_time": "0:16:37", "throughput": 11126.87, "total_tokens": 98135624}
{"current_steps": 14670, "total_steps": 16324, "loss": 0.2322, "lr": 2.780914991447292e-06, "epoch": 1.7972803283207253, "percentage": 89.87, "elapsed_time": "2:27:57", "remaining_time": "0:16:40", "throughput": 11057.93, "total_tokens": 98169432}
{"current_steps": 14675, "total_steps": 16324, "loss": 0.1974, "lr": 2.7642835649279606e-06, "epoch": 1.7978928669872287, "percentage": 89.9, "elapsed_time": "2:28:57", "remaining_time": "0:16:44", "throughput": 10988.37, "total_tokens": 98203464}
{"current_steps": 14680, "total_steps": 16324, "loss": 0.1307, "lr": 2.7477006063584942e-06, "epoch": 1.798505405653732, "percentage": 89.93, "elapsed_time": "2:29:55", "remaining_time": "0:16:47", "throughput": 10920.64, "total_tokens": 98238568}
{"current_steps": 14685, "total_steps": 16324, "loss": 0.1698, "lr": 2.7311661327544423e-06, "epoch": 1.7991179443202352, "percentage": 89.96, "elapsed_time": "2:30:53", "remaining_time": "0:16:50", "throughput": 10854.48, "total_tokens": 98272600}
{"current_steps": 14690, "total_steps": 16324, "loss": 0.2237, "lr": 2.714680161081623e-06, "epoch": 1.7997304829867384, "percentage": 89.99, "elapsed_time": "2:31:52", "remaining_time": "0:16:53", "throughput": 10787.94, "total_tokens": 98306000}
{"current_steps": 14695, "total_steps": 16324, "loss": 0.2262, "lr": 2.6982427082560544e-06, "epoch": 1.800343021653242, "percentage": 90.02, "elapsed_time": "2:32:51", "remaining_time": "0:16:56", "throughput": 10722.36, "total_tokens": 98339152}
{"current_steps": 14700, "total_steps": 16324, "loss": 0.184, "lr": 2.6818537911440065e-06, "epoch": 1.8009555603197451, "percentage": 90.05, "elapsed_time": "2:33:49", "remaining_time": "0:16:59", "throughput": 10658.83, "total_tokens": 98372656}
{"current_steps": 14700, "total_steps": 16324, "eval_loss": 0.13143287599086761, "epoch": 1.8009555603197451, "percentage": 90.05, "elapsed_time": "2:34:08", "remaining_time": "0:17:01", "throughput": 10637.12, "total_tokens": 98372656}
{"current_steps": 14705, "total_steps": 16324, "loss": 0.1738, "lr": 2.6655134265619385e-06, "epoch": 1.8015680989862486, "percentage": 90.08, "elapsed_time": "2:35:11", "remaining_time": "0:17:05", "throughput": 10568.34, "total_tokens": 98406728}
{"current_steps": 14710, "total_steps": 16324, "loss": 0.1778, "lr": 2.6492216312764662e-06, "epoch": 1.8021806376527518, "percentage": 90.11, "elapsed_time": "2:36:11", "remaining_time": "0:17:08", "throughput": 10504.89, "total_tokens": 98441344}
{"current_steps": 14715, "total_steps": 16324, "loss": 0.21, "lr": 2.6329784220044007e-06, "epoch": 1.802793176319255, "percentage": 90.14, "elapsed_time": "2:37:09", "remaining_time": "0:17:11", "throughput": 10443.72, "total_tokens": 98474960}
{"current_steps": 14720, "total_steps": 16324, "loss": 0.1878, "lr": 2.6167838154126765e-06, "epoch": 1.8034057149857585, "percentage": 90.17, "elapsed_time": "2:38:07", "remaining_time": "0:17:13", "throughput": 10383.37, "total_tokens": 98509216}
{"current_steps": 14725, "total_steps": 16324, "loss": 0.19, "lr": 2.6006378281183798e-06, "epoch": 1.8040182536522618, "percentage": 90.2, "elapsed_time": "2:39:06", "remaining_time": "0:17:16", "throughput": 10322.33, "total_tokens": 98543568}
{"current_steps": 14730, "total_steps": 16324, "loss": 0.2289, "lr": 2.584540476688685e-06, "epoch": 1.8046307923187652, "percentage": 90.24, "elapsed_time": "2:40:04", "remaining_time": "0:17:19", "throughput": 10263.18, "total_tokens": 98577256}
{"current_steps": 14735, "total_steps": 16324, "loss": 0.2327, "lr": 2.5684917776408746e-06, "epoch": 1.8052433309852685, "percentage": 90.27, "elapsed_time": "2:41:03", "remaining_time": "0:17:22", "throughput": 10204.8, "total_tokens": 98610824}
{"current_steps": 14740, "total_steps": 16324, "loss": 0.2092, "lr": 2.5524917474423205e-06, "epoch": 1.8058558696517717, "percentage": 90.3, "elapsed_time": "2:42:02", "remaining_time": "0:17:24", "throughput": 10145.45, "total_tokens": 98644120}
{"current_steps": 14745, "total_steps": 16324, "loss": 0.2086, "lr": 2.536540402510429e-06, "epoch": 1.806468408318275, "percentage": 90.33, "elapsed_time": "2:43:01", "remaining_time": "0:17:27", "throughput": 10088.34, "total_tokens": 98677256}
{"current_steps": 14750, "total_steps": 16324, "loss": 0.2065, "lr": 2.5206377592126683e-06, "epoch": 1.8070809469847784, "percentage": 90.36, "elapsed_time": "2:43:59", "remaining_time": "0:17:30", "throughput": 10031.95, "total_tokens": 98710624}
{"current_steps": 14750, "total_steps": 16324, "eval_loss": 0.11075767874717712, "epoch": 1.8070809469847784, "percentage": 90.36, "elapsed_time": "2:44:18", "remaining_time": "0:17:32", "throughput": 10012.4, "total_tokens": 98710624}
{"current_steps": 14755, "total_steps": 16324, "loss": 0.2144, "lr": 2.504783833866542e-06, "epoch": 1.8076934856512819, "percentage": 90.39, "elapsed_time": "2:45:23", "remaining_time": "0:17:35", "throughput": 9950.59, "total_tokens": 98743648}
{"current_steps": 14760, "total_steps": 16324, "loss": 0.2127, "lr": 2.4889786427395534e-06, "epoch": 1.808306024317785, "percentage": 90.42, "elapsed_time": "2:46:21", "remaining_time": "0:17:37", "throughput": 9895.55, "total_tokens": 98776688}
{"current_steps": 14765, "total_steps": 16324, "loss": 0.1938, "lr": 2.4732222020491967e-06, "epoch": 1.8089185629842883, "percentage": 90.45, "elapsed_time": "2:47:20", "remaining_time": "0:17:40", "throughput": 9840.9, "total_tokens": 98810616}
{"current_steps": 14770, "total_steps": 16324, "loss": 0.2509, "lr": 2.4575145279629452e-06, "epoch": 1.8095311016507916, "percentage": 90.48, "elapsed_time": "2:48:20", "remaining_time": "0:17:42", "throughput": 9785.57, "total_tokens": 98843768}
{"current_steps": 14775, "total_steps": 16324, "loss": 0.1735, "lr": 2.441855636598256e-06, "epoch": 1.810143640317295, "percentage": 90.51, "elapsed_time": "2:49:19", "remaining_time": "0:17:45", "throughput": 9732.48, "total_tokens": 98877968}
{"current_steps": 14780, "total_steps": 16324, "loss": 0.1706, "lr": 2.4262455440224872e-06, "epoch": 1.8107561789837985, "percentage": 90.54, "elapsed_time": "2:50:18", "remaining_time": "0:17:47", "throughput": 9679.51, "total_tokens": 98912168}
{"current_steps": 14785, "total_steps": 16324, "loss": 0.233, "lr": 2.410684266252966e-06, "epoch": 1.8113687176503017, "percentage": 90.57, "elapsed_time": "2:51:18", "remaining_time": "0:17:49", "throughput": 9626.32, "total_tokens": 98946184}
{"current_steps": 14790, "total_steps": 16324, "loss": 0.2107, "lr": 2.395171819256914e-06, "epoch": 1.811981256316805, "percentage": 90.6, "elapsed_time": "2:52:17", "remaining_time": "0:17:52", "throughput": 9574.75, "total_tokens": 98979480}
{"current_steps": 14795, "total_steps": 16324, "loss": 0.2219, "lr": 2.3797082189514596e-06, "epoch": 1.8125937949833082, "percentage": 90.63, "elapsed_time": "2:53:16", "remaining_time": "0:17:54", "throughput": 9523.43, "total_tokens": 99012760}
{"current_steps": 14800, "total_steps": 16324, "loss": 0.2039, "lr": 2.364293481203578e-06, "epoch": 1.8132063336498117, "percentage": 90.66, "elapsed_time": "2:54:16", "remaining_time": "0:17:56", "throughput": 9472.08, "total_tokens": 99045480}
{"current_steps": 14800, "total_steps": 16324, "eval_loss": 0.08662518113851547, "epoch": 1.8132063336498117, "percentage": 90.66, "elapsed_time": "2:54:36", "remaining_time": "0:17:58", "throughput": 9454.39, "total_tokens": 99045480}
{"current_steps": 14805, "total_steps": 16324, "loss": 0.2164, "lr": 2.3489276218301437e-06, "epoch": 1.8138188723163151, "percentage": 90.69, "elapsed_time": "2:55:41", "remaining_time": "0:18:01", "throughput": 9399.27, "total_tokens": 99079336}
{"current_steps": 14810, "total_steps": 16324, "loss": 0.1915, "lr": 2.3336106565978567e-06, "epoch": 1.8144314109828183, "percentage": 90.73, "elapsed_time": "2:56:40", "remaining_time": "0:18:03", "throughput": 9350.25, "total_tokens": 99113184}
{"current_steps": 14815, "total_steps": 16324, "loss": 0.2385, "lr": 2.3183426012232666e-06, "epoch": 1.8150439496493216, "percentage": 90.76, "elapsed_time": "2:57:40", "remaining_time": "0:18:05", "throughput": 9300.52, "total_tokens": 99145968}
{"current_steps": 14820, "total_steps": 16324, "loss": 0.1638, "lr": 2.3031234713727145e-06, "epoch": 1.8156564883158248, "percentage": 90.79, "elapsed_time": "2:58:39", "remaining_time": "0:18:07", "throughput": 9252.67, "total_tokens": 99180864}
{"current_steps": 14825, "total_steps": 16324, "loss": 0.2222, "lr": 2.2879532826623473e-06, "epoch": 1.8162690269823283, "percentage": 90.82, "elapsed_time": "2:59:37", "remaining_time": "0:18:09", "throughput": 9205.33, "total_tokens": 99214104}
{"current_steps": 14830, "total_steps": 16324, "loss": 0.1742, "lr": 2.2728320506581138e-06, "epoch": 1.8168815656488317, "percentage": 90.85, "elapsed_time": "3:00:36", "remaining_time": "0:18:11", "throughput": 9158.44, "total_tokens": 99248104}
{"current_steps": 14835, "total_steps": 16324, "loss": 0.2108, "lr": 2.257759790875691e-06, "epoch": 1.817494104315335, "percentage": 90.88, "elapsed_time": "3:01:35", "remaining_time": "0:18:13", "throughput": 9112.26, "total_tokens": 99282424}
{"current_steps": 14840, "total_steps": 16324, "loss": 0.2084, "lr": 2.242736518780536e-06, "epoch": 1.8181066429818382, "percentage": 90.91, "elapsed_time": "3:02:33", "remaining_time": "0:18:15", "throughput": 9067.08, "total_tokens": 99315864}
{"current_steps": 14845, "total_steps": 16324, "loss": 0.1833, "lr": 2.227762249787835e-06, "epoch": 1.8187191816483415, "percentage": 90.94, "elapsed_time": "3:03:31", "remaining_time": "0:18:17", "throughput": 9022.68, "total_tokens": 99349368}
{"current_steps": 14850, "total_steps": 16324, "loss": 0.1985, "lr": 2.2128369992624962e-06, "epoch": 1.819331720314845, "percentage": 90.97, "elapsed_time": "3:04:30", "remaining_time": "0:18:18", "throughput": 8976.95, "total_tokens": 99382520}
{"current_steps": 14850, "total_steps": 16324, "eval_loss": 0.09910637885332108, "epoch": 1.819331720314845, "percentage": 90.97, "elapsed_time": "3:04:49", "remaining_time": "0:18:20", "throughput": 8961.53, "total_tokens": 99382520}
{"current_steps": 14855, "total_steps": 16324, "loss": 0.1815, "lr": 2.197960782519104e-06, "epoch": 1.8199442589813482, "percentage": 91.0, "elapsed_time": "3:05:53", "remaining_time": "0:18:22", "throughput": 8913.76, "total_tokens": 99415992}
{"current_steps": 14860, "total_steps": 16324, "loss": 0.2273, "lr": 2.1831336148219583e-06, "epoch": 1.8205567976478516, "percentage": 91.03, "elapsed_time": "3:06:50", "remaining_time": "0:18:24", "throughput": 8870.85, "total_tokens": 99449616}
{"current_steps": 14865, "total_steps": 16324, "loss": 0.1987, "lr": 2.168355511385034e-06, "epoch": 1.8211693363143548, "percentage": 91.06, "elapsed_time": "3:07:50", "remaining_time": "0:18:26", "throughput": 8827.17, "total_tokens": 99483296}
{"current_steps": 14870, "total_steps": 16324, "loss": 0.2172, "lr": 2.1536264873719293e-06, "epoch": 1.821781874980858, "percentage": 91.09, "elapsed_time": "3:08:48", "remaining_time": "0:18:27", "throughput": 8784.79, "total_tokens": 99516888}
{"current_steps": 14875, "total_steps": 16324, "loss": 0.1972, "lr": 2.1389465578959154e-06, "epoch": 1.8223944136473615, "percentage": 91.12, "elapsed_time": "3:09:46", "remaining_time": "0:18:29", "throughput": 8742.9, "total_tokens": 99550560}
{"current_steps": 14880, "total_steps": 16324, "loss": 0.2309, "lr": 2.124315738019872e-06, "epoch": 1.8230069523138648, "percentage": 91.15, "elapsed_time": "3:10:46", "remaining_time": "0:18:30", "throughput": 8700.32, "total_tokens": 99584216}
{"current_steps": 14885, "total_steps": 16324, "loss": 0.214, "lr": 2.109734042756295e-06, "epoch": 1.8236194909803682, "percentage": 91.18, "elapsed_time": "3:11:44", "remaining_time": "0:18:32", "throughput": 8659.13, "total_tokens": 99617792}
{"current_steps": 14890, "total_steps": 16324, "loss": 0.1892, "lr": 2.095201487067261e-06, "epoch": 1.8242320296468715, "percentage": 91.22, "elapsed_time": "3:12:43", "remaining_time": "0:18:33", "throughput": 8618.08, "total_tokens": 99651752}
{"current_steps": 14895, "total_steps": 16324, "loss": 0.2182, "lr": 2.0807180858644417e-06, "epoch": 1.8248445683133747, "percentage": 91.25, "elapsed_time": "3:13:43", "remaining_time": "0:18:35", "throughput": 8575.86, "total_tokens": 99685120}
{"current_steps": 14900, "total_steps": 16324, "loss": 0.1867, "lr": 2.066283854009077e-06, "epoch": 1.825457106979878, "percentage": 91.28, "elapsed_time": "3:14:42", "remaining_time": "0:18:36", "throughput": 8535.7, "total_tokens": 99719032}
{"current_steps": 14900, "total_steps": 16324, "eval_loss": 0.10013201087713242, "epoch": 1.825457106979878, "percentage": 91.28, "elapsed_time": "3:15:01", "remaining_time": "0:18:38", "throughput": 8521.79, "total_tokens": 99719032}
{"current_steps": 14905, "total_steps": 16324, "loss": 0.2255, "lr": 2.0518988063119194e-06, "epoch": 1.8260696456463814, "percentage": 91.31, "elapsed_time": "3:16:06", "remaining_time": "0:18:40", "throughput": 8477.52, "total_tokens": 99752200}
{"current_steps": 14910, "total_steps": 16324, "loss": 0.199, "lr": 2.0375629575332957e-06, "epoch": 1.8266821843128849, "percentage": 91.34, "elapsed_time": "3:17:06", "remaining_time": "0:18:41", "throughput": 8437.64, "total_tokens": 99785928}
{"current_steps": 14915, "total_steps": 16324, "loss": 0.2024, "lr": 2.023276322383022e-06, "epoch": 1.827294722979388, "percentage": 91.37, "elapsed_time": "3:18:04", "remaining_time": "0:18:42", "throughput": 8398.89, "total_tokens": 99819448}
{"current_steps": 14920, "total_steps": 16324, "loss": 0.2321, "lr": 2.009038915520439e-06, "epoch": 1.8279072616458913, "percentage": 91.4, "elapsed_time": "3:19:03", "remaining_time": "0:18:43", "throughput": 8360.53, "total_tokens": 99852736}
{"current_steps": 14925, "total_steps": 16324, "loss": 0.2397, "lr": 1.994850751554356e-06, "epoch": 1.8285198003123946, "percentage": 91.43, "elapsed_time": "3:20:03", "remaining_time": "0:18:45", "throughput": 8321.44, "total_tokens": 99886184}
{"current_steps": 14930, "total_steps": 16324, "loss": 0.1966, "lr": 1.9807118450430594e-06, "epoch": 1.829132338978898, "percentage": 91.46, "elapsed_time": "3:21:01", "remaining_time": "0:18:46", "throughput": 8283.89, "total_tokens": 99920056}
{"current_steps": 14935, "total_steps": 16324, "loss": 0.2183, "lr": 1.966622210494301e-06, "epoch": 1.8297448776454015, "percentage": 91.49, "elapsed_time": "3:22:00", "remaining_time": "0:18:47", "throughput": 8246.6, "total_tokens": 99953608}
{"current_steps": 14940, "total_steps": 16324, "loss": 0.2188, "lr": 1.9525818623652713e-06, "epoch": 1.8303574163119047, "percentage": 91.52, "elapsed_time": "3:23:00", "remaining_time": "0:18:48", "throughput": 8208.62, "total_tokens": 99987368}
{"current_steps": 14945, "total_steps": 16324, "loss": 0.214, "lr": 1.93859081506258e-06, "epoch": 1.830969954978408, "percentage": 91.55, "elapsed_time": "3:23:59", "remaining_time": "0:18:49", "throughput": 8172.13, "total_tokens": 100020472}
{"current_steps": 14950, "total_steps": 16324, "loss": 0.1983, "lr": 1.9246490829422646e-06, "epoch": 1.8315824936449112, "percentage": 91.58, "elapsed_time": "3:24:57", "remaining_time": "0:18:50", "throughput": 8136.21, "total_tokens": 100053904}
{"current_steps": 14950, "total_steps": 16324, "eval_loss": 0.07142162322998047, "epoch": 1.8315824936449112, "percentage": 91.58, "elapsed_time": "3:25:16", "remaining_time": "0:18:51", "throughput": 8123.68, "total_tokens": 100053904}
{"current_steps": 14955, "total_steps": 16324, "loss": 0.209, "lr": 1.9107566803097608e-06, "epoch": 1.8321950323114147, "percentage": 91.61, "elapsed_time": "3:26:20", "remaining_time": "0:18:53", "throughput": 8084.55, "total_tokens": 100087184}
{"current_steps": 14960, "total_steps": 16324, "loss": 0.1727, "lr": 1.8969136214198657e-06, "epoch": 1.8328075709779181, "percentage": 91.64, "elapsed_time": "3:27:17", "remaining_time": "0:18:54", "throughput": 8049.81, "total_tokens": 100120816}
{"current_steps": 14965, "total_steps": 16324, "loss": 0.1774, "lr": 1.8831199204767612e-06, "epoch": 1.8334201096444214, "percentage": 91.67, "elapsed_time": "3:28:15", "remaining_time": "0:18:54", "throughput": 8015.38, "total_tokens": 100154856}
{"current_steps": 14970, "total_steps": 16324, "loss": 0.1541, "lr": 1.8693755916339929e-06, "epoch": 1.8340326483109246, "percentage": 91.71, "elapsed_time": "3:29:13", "remaining_time": "0:18:55", "throughput": 7980.75, "total_tokens": 100189688}
{"current_steps": 14975, "total_steps": 16324, "loss": 0.2346, "lr": 1.8556806489944344e-06, "epoch": 1.8346451869774278, "percentage": 91.74, "elapsed_time": "3:30:12", "remaining_time": "0:18:56", "throughput": 7946.59, "total_tokens": 100222896}
{"current_steps": 14980, "total_steps": 16324, "loss": 0.1726, "lr": 1.8420351066102847e-06, "epoch": 1.8352577256439313, "percentage": 91.77, "elapsed_time": "3:31:10", "remaining_time": "0:18:56", "throughput": 7912.88, "total_tokens": 100256920}
{"current_steps": 14985, "total_steps": 16324, "loss": 0.1766, "lr": 1.8284389784830547e-06, "epoch": 1.8358702643104345, "percentage": 91.8, "elapsed_time": "3:32:08", "remaining_time": "0:18:57", "throughput": 7879.32, "total_tokens": 100291000}
{"current_steps": 14990, "total_steps": 16324, "loss": 0.1408, "lr": 1.814892278563557e-06, "epoch": 1.836482802976938, "percentage": 91.83, "elapsed_time": "3:33:06", "remaining_time": "0:18:57", "throughput": 7846.04, "total_tokens": 100325272}
{"current_steps": 14995, "total_steps": 16324, "loss": 0.231, "lr": 1.801395020751878e-06, "epoch": 1.8370953416434412, "percentage": 91.86, "elapsed_time": "3:34:04", "remaining_time": "0:18:58", "throughput": 7813.49, "total_tokens": 100359016}
{"current_steps": 15000, "total_steps": 16324, "loss": 0.1842, "lr": 1.7879472188973844e-06, "epoch": 1.8377078803099445, "percentage": 91.89, "elapsed_time": "3:35:02", "remaining_time": "0:18:58", "throughput": 7781.15, "total_tokens": 100392776}
{"current_steps": 15000, "total_steps": 16324, "eval_loss": 0.08897951990365982, "epoch": 1.8377078803099445, "percentage": 91.89, "elapsed_time": "3:35:20", "remaining_time": "0:19:00", "throughput": 7769.87, "total_tokens": 100392776}
{"current_steps": 15005, "total_steps": 16324, "loss": 0.2254, "lr": 1.7745488867986771e-06, "epoch": 1.838320418976448, "percentage": 91.92, "elapsed_time": "3:36:25", "remaining_time": "0:19:01", "throughput": 7733.84, "total_tokens": 100426336}
{"current_steps": 15010, "total_steps": 16324, "loss": 0.2226, "lr": 1.7612000382036308e-06, "epoch": 1.8389329576429512, "percentage": 91.95, "elapsed_time": "3:37:23", "remaining_time": "0:19:01", "throughput": 7701.89, "total_tokens": 100459744}
{"current_steps": 15015, "total_steps": 16324, "loss": 0.1888, "lr": 1.7479006868093006e-06, "epoch": 1.8395454963094546, "percentage": 91.98, "elapsed_time": "3:38:21", "remaining_time": "0:19:02", "throughput": 7670.43, "total_tokens": 100493288}
{"current_steps": 15020, "total_steps": 16324, "loss": 0.1688, "lr": 1.734650846261987e-06, "epoch": 1.8401580349759579, "percentage": 92.01, "elapsed_time": "3:39:21", "remaining_time": "0:19:02", "throughput": 7638.25, "total_tokens": 100527512}
{"current_steps": 15025, "total_steps": 16324, "loss": 0.2857, "lr": 1.7214505301571871e-06, "epoch": 1.840770573642461, "percentage": 92.04, "elapsed_time": "3:40:18", "remaining_time": "0:19:02", "throughput": 7607.35, "total_tokens": 100560864}
{"current_steps": 15030, "total_steps": 16324, "loss": 0.2345, "lr": 1.708299752039555e-06, "epoch": 1.8413831123089643, "percentage": 92.07, "elapsed_time": "3:41:16", "remaining_time": "0:19:03", "throughput": 7576.62, "total_tokens": 100593464}
{"current_steps": 15035, "total_steps": 16324, "loss": 0.1801, "lr": 1.6951985254029411e-06, "epoch": 1.8419956509754678, "percentage": 92.1, "elapsed_time": "3:42:16", "remaining_time": "0:19:03", "throughput": 7545.15, "total_tokens": 100627640}
{"current_steps": 15040, "total_steps": 16324, "loss": 0.1927, "lr": 1.6821468636903427e-06, "epoch": 1.8426081896419713, "percentage": 92.13, "elapsed_time": "3:43:15", "remaining_time": "0:19:03", "throughput": 7514.64, "total_tokens": 100661184}
{"current_steps": 15045, "total_steps": 16324, "loss": 0.2323, "lr": 1.6691447802938963e-06, "epoch": 1.8432207283084745, "percentage": 92.16, "elapsed_time": "3:44:13", "remaining_time": "0:19:03", "throughput": 7484.58, "total_tokens": 100694104}
{"current_steps": 15050, "total_steps": 16324, "loss": 0.1583, "lr": 1.6561922885548698e-06, "epoch": 1.8438332669749777, "percentage": 92.2, "elapsed_time": "3:45:13", "remaining_time": "0:19:03", "throughput": 7454.21, "total_tokens": 100729000}
{"current_steps": 15050, "total_steps": 16324, "eval_loss": 0.08737240731716156, "epoch": 1.8438332669749777, "percentage": 92.2, "elapsed_time": "3:45:32", "remaining_time": "0:19:05", "throughput": 7443.32, "total_tokens": 100729000}
{"current_steps": 15055, "total_steps": 16324, "loss": 0.2174, "lr": 1.643289401763648e-06, "epoch": 1.844445805641481, "percentage": 92.23, "elapsed_time": "3:46:37", "remaining_time": "0:19:06", "throughput": 7410.66, "total_tokens": 100762816}
{"current_steps": 15060, "total_steps": 16324, "loss": 0.2164, "lr": 1.6304361331597074e-06, "epoch": 1.8450583443079844, "percentage": 92.26, "elapsed_time": "3:47:34", "remaining_time": "0:19:06", "throughput": 7381.93, "total_tokens": 100796032}
{"current_steps": 15065, "total_steps": 16324, "loss": 0.2185, "lr": 1.6176324959316314e-06, "epoch": 1.845670882974488, "percentage": 92.29, "elapsed_time": "3:48:32", "remaining_time": "0:19:05", "throughput": 7352.98, "total_tokens": 100829824}
{"current_steps": 15070, "total_steps": 16324, "loss": 0.1937, "lr": 1.6048785032170443e-06, "epoch": 1.8462834216409911, "percentage": 92.32, "elapsed_time": "3:49:30", "remaining_time": "0:19:05", "throughput": 7324.46, "total_tokens": 100863496}
{"current_steps": 15075, "total_steps": 16324, "loss": 0.2065, "lr": 1.592174168102667e-06, "epoch": 1.8468959603074944, "percentage": 92.35, "elapsed_time": "3:50:28", "remaining_time": "0:19:05", "throughput": 7296.3, "total_tokens": 100897264}
{"current_steps": 15080, "total_steps": 16324, "loss": 0.2037, "lr": 1.579519503624255e-06, "epoch": 1.8475084989739976, "percentage": 92.38, "elapsed_time": "3:51:27", "remaining_time": "0:19:05", "throughput": 7267.99, "total_tokens": 100931256}
{"current_steps": 15085, "total_steps": 16324, "loss": 0.2094, "lr": 1.566914522766577e-06, "epoch": 1.848121037640501, "percentage": 92.41, "elapsed_time": "3:52:25", "remaining_time": "0:19:05", "throughput": 7239.92, "total_tokens": 100965144}
{"current_steps": 15090, "total_steps": 16324, "loss": 0.1951, "lr": 1.5543592384634486e-06, "epoch": 1.8487335763070045, "percentage": 92.44, "elapsed_time": "3:53:23", "remaining_time": "0:19:05", "throughput": 7212.3, "total_tokens": 100998808}
{"current_steps": 15095, "total_steps": 16324, "loss": 0.2197, "lr": 1.541853663597692e-06, "epoch": 1.8493461149735078, "percentage": 92.47, "elapsed_time": "3:54:21", "remaining_time": "0:19:04", "throughput": 7184.88, "total_tokens": 101032848}
{"current_steps": 15100, "total_steps": 16324, "loss": 0.2087, "lr": 1.5293978110010932e-06, "epoch": 1.849958653640011, "percentage": 92.5, "elapsed_time": "3:55:20", "remaining_time": "0:19:04", "throughput": 7157.35, "total_tokens": 101066288}
{"current_steps": 15100, "total_steps": 16324, "eval_loss": 0.07200353592634201, "epoch": 1.849958653640011, "percentage": 92.5, "elapsed_time": "3:55:39", "remaining_time": "0:19:06", "throughput": 7147.67, "total_tokens": 101066288}
{"current_steps": 15105, "total_steps": 16324, "loss": 0.1724, "lr": 1.516991693454445e-06, "epoch": 1.8505711923065142, "percentage": 92.53, "elapsed_time": "3:56:43", "remaining_time": "0:19:06", "throughput": 7117.86, "total_tokens": 101100024}
{"current_steps": 15110, "total_steps": 16324, "loss": 0.211, "lr": 1.5046353236875143e-06, "epoch": 1.8511837309730177, "percentage": 92.56, "elapsed_time": "3:57:41", "remaining_time": "0:19:05", "throughput": 7091.11, "total_tokens": 101133448}
{"current_steps": 15115, "total_steps": 16324, "loss": 0.2188, "lr": 1.4923287143790043e-06, "epoch": 1.8517962696395212, "percentage": 92.59, "elapsed_time": "3:58:41", "remaining_time": "0:19:05", "throughput": 7064.05, "total_tokens": 101167424}
{"current_steps": 15120, "total_steps": 16324, "loss": 0.2137, "lr": 1.4800718781565637e-06, "epoch": 1.8524088083060244, "percentage": 92.62, "elapsed_time": "3:59:39", "remaining_time": "0:19:05", "throughput": 7037.87, "total_tokens": 101201784}
{"current_steps": 15125, "total_steps": 16324, "loss": 0.2083, "lr": 1.4678648275967654e-06, "epoch": 1.8530213469725276, "percentage": 92.65, "elapsed_time": "4:00:37", "remaining_time": "0:19:04", "throughput": 7011.74, "total_tokens": 101234824}
{"current_steps": 15130, "total_steps": 16324, "loss": 0.1964, "lr": 1.4557075752251125e-06, "epoch": 1.8536338856390309, "percentage": 92.69, "elapsed_time": "4:01:37", "remaining_time": "0:19:04", "throughput": 6985.43, "total_tokens": 101268296}
{"current_steps": 15135, "total_steps": 16324, "loss": 0.2402, "lr": 1.4436001335160044e-06, "epoch": 1.8542464243055343, "percentage": 92.72, "elapsed_time": "4:02:35", "remaining_time": "0:19:03", "throughput": 6959.73, "total_tokens": 101300984}
{"current_steps": 15140, "total_steps": 16324, "loss": 0.2117, "lr": 1.4315425148927198e-06, "epoch": 1.8548589629720376, "percentage": 92.75, "elapsed_time": "4:03:33", "remaining_time": "0:19:02", "throughput": 6934.2, "total_tokens": 101334096}
{"current_steps": 15145, "total_steps": 16324, "loss": 0.2101, "lr": 1.4195347317274287e-06, "epoch": 1.855471501638541, "percentage": 92.78, "elapsed_time": "4:04:33", "remaining_time": "0:19:02", "throughput": 6908.24, "total_tokens": 101367920}
{"current_steps": 15150, "total_steps": 16324, "loss": 0.1811, "lr": 1.4075767963411645e-06, "epoch": 1.8560840403050443, "percentage": 92.81, "elapsed_time": "4:05:31", "remaining_time": "0:19:01", "throughput": 6883.12, "total_tokens": 101401720}
{"current_steps": 15150, "total_steps": 16324, "eval_loss": 0.09131244570016861, "epoch": 1.8560840403050443, "percentage": 92.81, "elapsed_time": "4:05:51", "remaining_time": "0:19:03", "throughput": 6874.2, "total_tokens": 101401720}
{"current_steps": 15155, "total_steps": 16324, "loss": 0.188, "lr": 1.3956687210037955e-06, "epoch": 1.8566965789715475, "percentage": 92.84, "elapsed_time": "4:06:54", "remaining_time": "0:19:02", "throughput": 6846.97, "total_tokens": 101435536}
{"current_steps": 15160, "total_steps": 16324, "loss": 0.2047, "lr": 1.3838105179340477e-06, "epoch": 1.857309117638051, "percentage": 92.87, "elapsed_time": "4:07:54", "remaining_time": "0:19:02", "throughput": 6821.78, "total_tokens": 101469400}
{"current_steps": 15165, "total_steps": 16324, "loss": 0.2097, "lr": 1.3720021992994714e-06, "epoch": 1.8579216563045542, "percentage": 92.9, "elapsed_time": "4:08:52", "remaining_time": "0:19:01", "throughput": 6797.43, "total_tokens": 101502592}
{"current_steps": 15170, "total_steps": 16324, "loss": 0.2079, "lr": 1.36024377721643e-06, "epoch": 1.8585341949710577, "percentage": 92.93, "elapsed_time": "4:09:50", "remaining_time": "0:19:00", "throughput": 6773.28, "total_tokens": 101535872}
{"current_steps": 15175, "total_steps": 16324, "loss": 0.2412, "lr": 1.3485352637500782e-06, "epoch": 1.8591467336375609, "percentage": 92.96, "elapsed_time": "4:10:49", "remaining_time": "0:18:59", "throughput": 6748.95, "total_tokens": 101569360}
{"current_steps": 15180, "total_steps": 16324, "loss": 0.19, "lr": 1.336876670914372e-06, "epoch": 1.8597592723040641, "percentage": 92.99, "elapsed_time": "4:11:47", "remaining_time": "0:18:58", "throughput": 6725.46, "total_tokens": 101603224}
{"current_steps": 15185, "total_steps": 16324, "loss": 0.1842, "lr": 1.3252680106720428e-06, "epoch": 1.8603718109705674, "percentage": 93.02, "elapsed_time": "4:12:44", "remaining_time": "0:18:57", "throughput": 6702.18, "total_tokens": 101636856}
{"current_steps": 15190, "total_steps": 16324, "loss": 0.2077, "lr": 1.3137092949345785e-06, "epoch": 1.8609843496370708, "percentage": 93.05, "elapsed_time": "4:13:43", "remaining_time": "0:18:56", "throughput": 6678.63, "total_tokens": 101670752}
{"current_steps": 15195, "total_steps": 16324, "loss": 0.2215, "lr": 1.3022005355622314e-06, "epoch": 1.8615968883035743, "percentage": 93.08, "elapsed_time": "4:14:41", "remaining_time": "0:18:55", "throughput": 6655.38, "total_tokens": 101703784}
{"current_steps": 15200, "total_steps": 16324, "loss": 0.1989, "lr": 1.2907417443639936e-06, "epoch": 1.8622094269700775, "percentage": 93.11, "elapsed_time": "4:15:39", "remaining_time": "0:18:54", "throughput": 6632.46, "total_tokens": 101737120}
{"current_steps": 15200, "total_steps": 16324, "eval_loss": 0.0902518779039383, "epoch": 1.8622094269700775, "percentage": 93.11, "elapsed_time": "4:15:58", "remaining_time": "0:18:55", "throughput": 6624.25, "total_tokens": 101737120}
{"current_steps": 15205, "total_steps": 16324, "loss": 0.2134, "lr": 1.2793329330975711e-06, "epoch": 1.8628219656365808, "percentage": 93.15, "elapsed_time": "4:17:01", "remaining_time": "0:18:54", "throughput": 6599.18, "total_tokens": 101770368}
{"current_steps": 15210, "total_steps": 16324, "loss": 0.2126, "lr": 1.2679741134694e-06, "epoch": 1.863434504303084, "percentage": 93.18, "elapsed_time": "4:18:00", "remaining_time": "0:18:53", "throughput": 6576.4, "total_tokens": 101803752}
{"current_steps": 15215, "total_steps": 16324, "loss": 0.2103, "lr": 1.2566652971346126e-06, "epoch": 1.8640470429695875, "percentage": 93.21, "elapsed_time": "4:18:57", "remaining_time": "0:18:52", "throughput": 6554.19, "total_tokens": 101837112}
{"current_steps": 15220, "total_steps": 16324, "loss": 0.2127, "lr": 1.245406495697038e-06, "epoch": 1.864659581636091, "percentage": 93.24, "elapsed_time": "4:19:55", "remaining_time": "0:18:51", "throughput": 6532.04, "total_tokens": 101870472}
{"current_steps": 15225, "total_steps": 16324, "loss": 0.1788, "lr": 1.2341977207091904e-06, "epoch": 1.8652721203025941, "percentage": 93.27, "elapsed_time": "4:20:54", "remaining_time": "0:18:49", "throughput": 6509.72, "total_tokens": 101904464}
{"current_steps": 15230, "total_steps": 16324, "loss": 0.1633, "lr": 1.2230389836722422e-06, "epoch": 1.8658846589690974, "percentage": 93.3, "elapsed_time": "4:21:51", "remaining_time": "0:18:48", "throughput": 6488.08, "total_tokens": 101938384}
{"current_steps": 15235, "total_steps": 16324, "loss": 0.2051, "lr": 1.211930296036029e-06, "epoch": 1.8664971976356006, "percentage": 93.33, "elapsed_time": "4:22:49", "remaining_time": "0:18:47", "throughput": 6466.46, "total_tokens": 101972312}
{"current_steps": 15240, "total_steps": 16324, "loss": 0.2214, "lr": 1.2008716691990329e-06, "epoch": 1.867109736302104, "percentage": 93.36, "elapsed_time": "4:23:48", "remaining_time": "0:18:45", "throughput": 6444.52, "total_tokens": 102006056}
{"current_steps": 15245, "total_steps": 16324, "loss": 0.1999, "lr": 1.189863114508355e-06, "epoch": 1.8677222749686075, "percentage": 93.39, "elapsed_time": "4:24:46", "remaining_time": "0:18:44", "throughput": 6423.08, "total_tokens": 102038824}
{"current_steps": 15250, "total_steps": 16324, "loss": 0.217, "lr": 1.1789046432597427e-06, "epoch": 1.8683348136351108, "percentage": 93.42, "elapsed_time": "4:25:44", "remaining_time": "0:18:42", "throughput": 6401.63, "total_tokens": 102072288}
{"current_steps": 15250, "total_steps": 16324, "eval_loss": 0.08231386542320251, "epoch": 1.8683348136351108, "percentage": 93.42, "elapsed_time": "4:26:03", "remaining_time": "0:18:44", "throughput": 6394.03, "total_tokens": 102072288}
{"current_steps": 15255, "total_steps": 16324, "loss": 0.1987, "lr": 1.1679962666975296e-06, "epoch": 1.868947352301614, "percentage": 93.45, "elapsed_time": "4:27:07", "remaining_time": "0:18:43", "throughput": 6370.58, "total_tokens": 102106056}
{"current_steps": 15260, "total_steps": 16324, "loss": 0.2353, "lr": 1.1571379960146622e-06, "epoch": 1.8695598909681173, "percentage": 93.48, "elapsed_time": "4:28:05", "remaining_time": "0:18:41", "throughput": 6349.73, "total_tokens": 102139288}
{"current_steps": 15265, "total_steps": 16324, "loss": 0.2193, "lr": 1.1463298423526725e-06, "epoch": 1.8701724296346207, "percentage": 93.51, "elapsed_time": "4:29:03", "remaining_time": "0:18:39", "throughput": 6328.84, "total_tokens": 102172456}
{"current_steps": 15270, "total_steps": 16324, "loss": 0.1866, "lr": 1.135571816801656e-06, "epoch": 1.8707849683011242, "percentage": 93.54, "elapsed_time": "4:30:03", "remaining_time": "0:18:38", "throughput": 6307.81, "total_tokens": 102205864}
{"current_steps": 15275, "total_steps": 16324, "loss": 0.2008, "lr": 1.1248639304002996e-06, "epoch": 1.8713975069676274, "percentage": 93.57, "elapsed_time": "4:31:01", "remaining_time": "0:18:36", "throughput": 6287.13, "total_tokens": 102239224}
{"current_steps": 15280, "total_steps": 16324, "loss": 0.199, "lr": 1.1142061941358085e-06, "epoch": 1.8720100456341306, "percentage": 93.6, "elapsed_time": "4:31:59", "remaining_time": "0:18:35", "throughput": 6266.71, "total_tokens": 102272672}
{"current_steps": 15285, "total_steps": 16324, "loss": 0.2167, "lr": 1.1035986189439517e-06, "epoch": 1.8726225843006339, "percentage": 93.64, "elapsed_time": "4:32:59", "remaining_time": "0:18:33", "throughput": 6245.94, "total_tokens": 102305384}
{"current_steps": 15290, "total_steps": 16324, "loss": 0.2107, "lr": 1.0930412157090175e-06, "epoch": 1.8732351229671373, "percentage": 93.67, "elapsed_time": "4:33:57", "remaining_time": "0:18:31", "throughput": 6226.11, "total_tokens": 102339072}
{"current_steps": 15295, "total_steps": 16324, "loss": 0.2095, "lr": 1.0825339952638346e-06, "epoch": 1.8738476616336406, "percentage": 93.7, "elapsed_time": "4:34:54", "remaining_time": "0:18:29", "throughput": 6206.51, "total_tokens": 102372232}
{"current_steps": 15300, "total_steps": 16324, "loss": 0.1836, "lr": 1.072076968389707e-06, "epoch": 1.874460200300144, "percentage": 93.73, "elapsed_time": "4:35:53", "remaining_time": "0:18:27", "throughput": 6186.38, "total_tokens": 102406408}
{"current_steps": 15300, "total_steps": 16324, "eval_loss": 0.09559149295091629, "epoch": 1.874460200300144, "percentage": 93.73, "elapsed_time": "4:36:12", "remaining_time": "0:18:29", "throughput": 6179.41, "total_tokens": 102406408}
{"current_steps": 15305, "total_steps": 16324, "loss": 0.1698, "lr": 1.061670145816457e-06, "epoch": 1.8750727389666473, "percentage": 93.76, "elapsed_time": "4:37:15", "remaining_time": "0:18:27", "throughput": 6158.01, "total_tokens": 102440880}
{"current_steps": 15310, "total_steps": 16324, "loss": 0.1908, "lr": 1.0513135382224037e-06, "epoch": 1.8756852776331505, "percentage": 93.79, "elapsed_time": "4:38:12", "remaining_time": "0:18:25", "throughput": 6138.94, "total_tokens": 102475112}
{"current_steps": 15315, "total_steps": 16324, "loss": 0.1934, "lr": 1.0410071562343026e-06, "epoch": 1.8762978162996538, "percentage": 93.82, "elapsed_time": "4:39:11", "remaining_time": "0:18:23", "throughput": 6119.32, "total_tokens": 102508888}
{"current_steps": 15320, "total_steps": 16324, "loss": 0.2157, "lr": 1.0307510104274165e-06, "epoch": 1.8769103549661572, "percentage": 93.85, "elapsed_time": "4:40:09", "remaining_time": "0:18:21", "throughput": 6100.24, "total_tokens": 102542336}
{"current_steps": 15325, "total_steps": 16324, "loss": 0.2349, "lr": 1.0205451113254327e-06, "epoch": 1.8775228936326607, "percentage": 93.88, "elapsed_time": "4:41:07", "remaining_time": "0:18:19", "throughput": 6081.36, "total_tokens": 102575480}
{"current_steps": 15330, "total_steps": 16324, "loss": 0.1854, "lr": 1.0103894694004968e-06, "epoch": 1.878135432299164, "percentage": 93.91, "elapsed_time": "4:42:05", "remaining_time": "0:18:17", "throughput": 6062.41, "total_tokens": 102609328}
{"current_steps": 15335, "total_steps": 16324, "loss": 0.167, "lr": 1.0002840950731728e-06, "epoch": 1.8787479709656671, "percentage": 93.94, "elapsed_time": "4:43:03", "remaining_time": "0:18:15", "throughput": 6043.73, "total_tokens": 102643024}
{"current_steps": 15340, "total_steps": 16324, "loss": 0.1919, "lr": 9.902289987124612e-07, "epoch": 1.8793605096321704, "percentage": 93.97, "elapsed_time": "4:44:00", "remaining_time": "0:18:13", "throughput": 6025.33, "total_tokens": 102677064}
{"current_steps": 15345, "total_steps": 16324, "loss": 0.1832, "lr": 9.80224190635759e-07, "epoch": 1.8799730482986738, "percentage": 94.0, "elapsed_time": "4:44:59", "remaining_time": "0:18:10", "throughput": 6006.6, "total_tokens": 102710648}
{"current_steps": 15350, "total_steps": 16324, "loss": 0.1769, "lr": 9.702696811088652e-07, "epoch": 1.8805855869651773, "percentage": 94.03, "elapsed_time": "4:45:58", "remaining_time": "0:18:08", "throughput": 5988.11, "total_tokens": 102744360}
{"current_steps": 15350, "total_steps": 16324, "eval_loss": 0.08247771114110947, "epoch": 1.8805855869651773, "percentage": 94.03, "elapsed_time": "4:46:17", "remaining_time": "0:18:09", "throughput": 5981.45, "total_tokens": 102744360}
{"current_steps": 15355, "total_steps": 16324, "loss": 0.1963, "lr": 9.60365480345976e-07, "epoch": 1.8811981256316805, "percentage": 94.06, "elapsed_time": "4:47:20", "remaining_time": "0:18:07", "throughput": 5961.56, "total_tokens": 102778024}
{"current_steps": 15360, "total_steps": 16324, "loss": 0.1971, "lr": 9.50511598509668e-07, "epoch": 1.8818106642981838, "percentage": 94.09, "elapsed_time": "4:48:18", "remaining_time": "0:18:05", "throughput": 5943.47, "total_tokens": 102812400}
{"current_steps": 15365, "total_steps": 16324, "loss": 0.223, "lr": 9.407080457108753e-07, "epoch": 1.882423202964687, "percentage": 94.13, "elapsed_time": "4:49:17", "remaining_time": "0:18:03", "throughput": 5925.18, "total_tokens": 102845576}
{"current_steps": 15370, "total_steps": 16324, "loss": 0.2699, "lr": 9.309548320088902e-07, "epoch": 1.8830357416311905, "percentage": 94.16, "elapsed_time": "4:50:15", "remaining_time": "0:18:00", "throughput": 5907.28, "total_tokens": 102878696}
{"current_steps": 15375, "total_steps": 16324, "loss": 0.2078, "lr": 9.212519674113573e-07, "epoch": 1.883648280297694, "percentage": 94.19, "elapsed_time": "4:51:14", "remaining_time": "0:17:58", "throughput": 5889.4, "total_tokens": 102911816}
{"current_steps": 15380, "total_steps": 16324, "loss": 0.2417, "lr": 9.115994618742685e-07, "epoch": 1.8842608189641972, "percentage": 94.22, "elapsed_time": "4:52:13", "remaining_time": "0:17:56", "throughput": 5871.45, "total_tokens": 102944840}
{"current_steps": 15385, "total_steps": 16324, "loss": 0.2054, "lr": 9.019973253019287e-07, "epoch": 1.8848733576307004, "percentage": 94.25, "elapsed_time": "4:53:11", "remaining_time": "0:17:53", "throughput": 5853.97, "total_tokens": 102978360}
{"current_steps": 15390, "total_steps": 16324, "loss": 0.1522, "lr": 8.924455675469622e-07, "epoch": 1.8854858962972036, "percentage": 94.28, "elapsed_time": "4:54:09", "remaining_time": "0:17:51", "throughput": 5836.5, "total_tokens": 103013184}
{"current_steps": 15395, "total_steps": 16324, "loss": 0.2204, "lr": 8.829441984103016e-07, "epoch": 1.886098434963707, "percentage": 94.31, "elapsed_time": "4:55:09", "remaining_time": "0:17:48", "throughput": 5818.84, "total_tokens": 103046528}
{"current_steps": 15400, "total_steps": 16324, "loss": 0.1906, "lr": 8.734932276411812e-07, "epoch": 1.8867109736302106, "percentage": 94.34, "elapsed_time": "4:56:07", "remaining_time": "0:17:46", "throughput": 5801.71, "total_tokens": 103079768}
{"current_steps": 15400, "total_steps": 16324, "eval_loss": 0.10278834402561188, "epoch": 1.8867109736302106, "percentage": 94.34, "elapsed_time": "4:56:26", "remaining_time": "0:17:47", "throughput": 5795.45, "total_tokens": 103079768}
{"current_steps": 15405, "total_steps": 16324, "loss": 0.2313, "lr": 8.640926649371162e-07, "epoch": 1.8873235122967138, "percentage": 94.37, "elapsed_time": "4:57:29", "remaining_time": "0:17:44", "throughput": 5776.71, "total_tokens": 103112608}
{"current_steps": 15410, "total_steps": 16324, "loss": 0.2245, "lr": 8.547425199439019e-07, "epoch": 1.887936050963217, "percentage": 94.4, "elapsed_time": "4:58:27", "remaining_time": "0:17:42", "throughput": 5759.92, "total_tokens": 103146728}
{"current_steps": 15415, "total_steps": 16324, "loss": 0.2345, "lr": 8.454428022556027e-07, "epoch": 1.8885485896297203, "percentage": 94.43, "elapsed_time": "4:59:24", "remaining_time": "0:17:39", "throughput": 5743.43, "total_tokens": 103179448}
{"current_steps": 15420, "total_steps": 16324, "loss": 0.2319, "lr": 8.3619352141453e-07, "epoch": 1.8891611282962237, "percentage": 94.46, "elapsed_time": "5:00:21", "remaining_time": "0:17:36", "throughput": 5727.08, "total_tokens": 103212696}
{"current_steps": 15425, "total_steps": 16324, "loss": 0.1937, "lr": 8.269946869112478e-07, "epoch": 1.889773666962727, "percentage": 94.49, "elapsed_time": "5:01:19", "remaining_time": "0:17:33", "throughput": 5710.57, "total_tokens": 103246288}
{"current_steps": 15430, "total_steps": 16324, "loss": 0.2249, "lr": 8.178463081845722e-07, "epoch": 1.8903862056292304, "percentage": 94.52, "elapsed_time": "5:02:17", "remaining_time": "0:17:30", "throughput": 5694.34, "total_tokens": 103279592}
{"current_steps": 15435, "total_steps": 16324, "loss": 0.2035, "lr": 8.08748394621528e-07, "epoch": 1.8909987442957337, "percentage": 94.55, "elapsed_time": "5:03:14", "remaining_time": "0:17:27", "throughput": 5678.2, "total_tokens": 103312920}
{"current_steps": 15440, "total_steps": 16324, "loss": 0.1801, "lr": 7.997009555573642e-07, "epoch": 1.891611282962237, "percentage": 94.58, "elapsed_time": "5:04:13", "remaining_time": "0:17:25", "throughput": 5661.85, "total_tokens": 103346728}
{"current_steps": 15445, "total_steps": 16324, "loss": 0.2356, "lr": 7.907040002755383e-07, "epoch": 1.8922238216287404, "percentage": 94.62, "elapsed_time": "5:05:10", "remaining_time": "0:17:22", "throughput": 5645.86, "total_tokens": 103379768}
{"current_steps": 15450, "total_steps": 16324, "loss": 0.2168, "lr": 7.817575380077047e-07, "epoch": 1.8928363602952436, "percentage": 94.65, "elapsed_time": "5:06:07", "remaining_time": "0:17:19", "throughput": 5630.1, "total_tokens": 103412640}
{"current_steps": 15450, "total_steps": 16324, "eval_loss": 0.07886143028736115, "epoch": 1.8928363602952436, "percentage": 94.65, "elapsed_time": "5:06:26", "remaining_time": "0:17:20", "throughput": 5624.31, "total_tokens": 103412640}
{"current_steps": 15455, "total_steps": 16324, "loss": 0.2313, "lr": 7.728615779337256e-07, "epoch": 1.893448898961747, "percentage": 94.68, "elapsed_time": "5:07:30", "remaining_time": "0:17:17", "throughput": 5606.62, "total_tokens": 103445512}
{"current_steps": 15460, "total_steps": 16324, "loss": 0.1936, "lr": 7.640161291816161e-07, "epoch": 1.8940614376282503, "percentage": 94.71, "elapsed_time": "5:08:28", "remaining_time": "0:17:14", "throughput": 5590.98, "total_tokens": 103479464}
{"current_steps": 15465, "total_steps": 16324, "loss": 0.2039, "lr": 7.552212008275772e-07, "epoch": 1.8946739762947535, "percentage": 94.74, "elapsed_time": "5:09:25", "remaining_time": "0:17:11", "throughput": 5575.49, "total_tokens": 103513600}
{"current_steps": 15470, "total_steps": 16324, "loss": 0.2087, "lr": 7.464768018959733e-07, "epoch": 1.8952865149612568, "percentage": 94.77, "elapsed_time": "5:10:24", "remaining_time": "0:17:08", "throughput": 5559.78, "total_tokens": 103547448}
{"current_steps": 15475, "total_steps": 16324, "loss": 0.2327, "lr": 7.377829413593163e-07, "epoch": 1.8958990536277602, "percentage": 94.8, "elapsed_time": "5:11:23", "remaining_time": "0:17:05", "throughput": 5544.09, "total_tokens": 103580760}
{"current_steps": 15480, "total_steps": 16324, "loss": 0.156, "lr": 7.29139628138259e-07, "epoch": 1.8965115922942637, "percentage": 94.83, "elapsed_time": "5:12:20", "remaining_time": "0:17:01", "throughput": 5528.78, "total_tokens": 103614760}
{"current_steps": 15485, "total_steps": 16324, "loss": 0.1996, "lr": 7.205468711015961e-07, "epoch": 1.897124130960767, "percentage": 94.86, "elapsed_time": "5:13:19", "remaining_time": "0:16:58", "throughput": 5513.42, "total_tokens": 103648480}
{"current_steps": 15490, "total_steps": 16324, "loss": 0.178, "lr": 7.120046790662471e-07, "epoch": 1.8977366696272702, "percentage": 94.89, "elapsed_time": "5:14:18", "remaining_time": "0:16:55", "throughput": 5497.91, "total_tokens": 103682216}
{"current_steps": 15495, "total_steps": 16324, "loss": 0.2103, "lr": 7.03513060797234e-07, "epoch": 1.8983492082937734, "percentage": 94.92, "elapsed_time": "5:15:16", "remaining_time": "0:16:52", "throughput": 5482.87, "total_tokens": 103715888}
{"current_steps": 15500, "total_steps": 16324, "loss": 0.2276, "lr": 6.950720250076981e-07, "epoch": 1.8989617469602769, "percentage": 94.95, "elapsed_time": "5:16:14", "remaining_time": "0:16:48", "throughput": 5467.71, "total_tokens": 103749376}
{"current_steps": 15500, "total_steps": 16324, "eval_loss": 0.11778271198272705, "epoch": 1.8989617469602769, "percentage": 94.95, "elapsed_time": "5:16:34", "remaining_time": "0:16:49", "throughput": 5462.16, "total_tokens": 103749376}
{"current_steps": 15505, "total_steps": 16324, "loss": 0.156, "lr": 6.866815803588833e-07, "epoch": 1.8995742856267803, "percentage": 94.98, "elapsed_time": "5:17:38", "remaining_time": "0:16:46", "throughput": 5445.42, "total_tokens": 103783288}
{"current_steps": 15510, "total_steps": 16324, "loss": 0.1895, "lr": 6.783417354601029e-07, "epoch": 1.9001868242932836, "percentage": 95.01, "elapsed_time": "5:18:37", "remaining_time": "0:16:43", "throughput": 5430.54, "total_tokens": 103816976}
{"current_steps": 15515, "total_steps": 16324, "loss": 0.217, "lr": 6.700524988687673e-07, "epoch": 1.9007993629597868, "percentage": 95.04, "elapsed_time": "5:19:35", "remaining_time": "0:16:39", "throughput": 5415.75, "total_tokens": 103850120}
{"current_steps": 15520, "total_steps": 16324, "loss": 0.2099, "lr": 6.618138790903561e-07, "epoch": 1.90141190162629, "percentage": 95.07, "elapsed_time": "5:20:33", "remaining_time": "0:16:36", "throughput": 5401.12, "total_tokens": 103883680}
{"current_steps": 15525, "total_steps": 16324, "loss": 0.1916, "lr": 6.536258845784127e-07, "epoch": 1.9020244402927935, "percentage": 95.11, "elapsed_time": "5:21:31", "remaining_time": "0:16:32", "throughput": 5386.79, "total_tokens": 103917296}
{"current_steps": 15530, "total_steps": 16324, "loss": 0.1995, "lr": 6.454885237345166e-07, "epoch": 1.902636978959297, "percentage": 95.14, "elapsed_time": "5:22:28", "remaining_time": "0:16:29", "throughput": 5372.4, "total_tokens": 103950536}
{"current_steps": 15535, "total_steps": 16324, "loss": 0.2049, "lr": 6.374018049083219e-07, "epoch": 1.9032495176258002, "percentage": 95.17, "elapsed_time": "5:23:27", "remaining_time": "0:16:25", "throughput": 5357.95, "total_tokens": 103983136}
{"current_steps": 15540, "total_steps": 16324, "loss": 0.2146, "lr": 6.29365736397497e-07, "epoch": 1.9038620562923034, "percentage": 95.2, "elapsed_time": "5:24:24", "remaining_time": "0:16:21", "throughput": 5343.95, "total_tokens": 104016528}
{"current_steps": 15545, "total_steps": 16324, "loss": 0.1904, "lr": 6.213803264477513e-07, "epoch": 1.9044745949588067, "percentage": 95.23, "elapsed_time": "5:25:21", "remaining_time": "0:16:18", "throughput": 5330.06, "total_tokens": 104050032}
{"current_steps": 15550, "total_steps": 16324, "loss": 0.1944, "lr": 6.13445583252803e-07, "epoch": 1.9050871336253101, "percentage": 95.26, "elapsed_time": "5:26:19", "remaining_time": "0:16:14", "throughput": 5315.82, "total_tokens": 104083336}
{"current_steps": 15550, "total_steps": 16324, "eval_loss": 0.11403975635766983, "epoch": 1.9050871336253101, "percentage": 95.26, "elapsed_time": "5:26:38", "remaining_time": "0:16:15", "throughput": 5310.76, "total_tokens": 104083336}
{"current_steps": 15555, "total_steps": 16324, "loss": 0.2264, "lr": 6.055615149543892e-07, "epoch": 1.9056996722918136, "percentage": 95.29, "elapsed_time": "5:27:41", "remaining_time": "0:16:12", "throughput": 5295.45, "total_tokens": 104116376}
{"current_steps": 15560, "total_steps": 16324, "loss": 0.1904, "lr": 5.97728129642261e-07, "epoch": 1.9063122109583168, "percentage": 95.32, "elapsed_time": "5:28:39", "remaining_time": "0:16:08", "throughput": 5281.67, "total_tokens": 104150184}
{"current_steps": 15565, "total_steps": 16324, "loss": 0.2119, "lr": 5.899454353541389e-07, "epoch": 1.90692474962482, "percentage": 95.35, "elapsed_time": "5:29:38", "remaining_time": "0:16:04", "throughput": 5267.65, "total_tokens": 104183632}
{"current_steps": 15570, "total_steps": 16324, "loss": 0.1957, "lr": 5.822134400757462e-07, "epoch": 1.9075372882913233, "percentage": 95.38, "elapsed_time": "5:30:35", "remaining_time": "0:16:00", "throughput": 5254.01, "total_tokens": 104217184}
{"current_steps": 15575, "total_steps": 16324, "loss": 0.2181, "lr": 5.745321517407864e-07, "epoch": 1.9081498269578268, "percentage": 95.41, "elapsed_time": "5:31:33", "remaining_time": "0:15:56", "throughput": 5240.56, "total_tokens": 104250656}
{"current_steps": 15580, "total_steps": 16324, "loss": 0.2237, "lr": 5.669015782309273e-07, "epoch": 1.90876236562433, "percentage": 95.44, "elapsed_time": "5:32:31", "remaining_time": "0:15:52", "throughput": 5226.86, "total_tokens": 104283904}
{"current_steps": 15585, "total_steps": 16324, "loss": 0.2018, "lr": 5.593217273757945e-07, "epoch": 1.9093749042908335, "percentage": 95.47, "elapsed_time": "5:33:29", "remaining_time": "0:15:48", "throughput": 5213.41, "total_tokens": 104317416}
{"current_steps": 15590, "total_steps": 16324, "loss": 0.2093, "lr": 5.517926069529777e-07, "epoch": 1.9099874429573367, "percentage": 95.5, "elapsed_time": "5:34:27", "remaining_time": "0:15:44", "throughput": 5200.11, "total_tokens": 104350752}
{"current_steps": 15595, "total_steps": 16324, "loss": 0.2168, "lr": 5.44314224688014e-07, "epoch": 1.91059998162384, "percentage": 95.53, "elapsed_time": "5:35:25", "remaining_time": "0:15:40", "throughput": 5186.61, "total_tokens": 104384184}
{"current_steps": 15600, "total_steps": 16324, "loss": 0.1814, "lr": 5.368865882543705e-07, "epoch": 1.9112125202903432, "percentage": 95.56, "elapsed_time": "5:36:23", "remaining_time": "0:15:36", "throughput": 5173.32, "total_tokens": 104417928}
{"current_steps": 15600, "total_steps": 16324, "eval_loss": 0.11014119535684586, "epoch": 1.9112125202903432, "percentage": 95.56, "elapsed_time": "5:36:43", "remaining_time": "0:15:37", "throughput": 5168.41, "total_tokens": 104417928}
{"current_steps": 15605, "total_steps": 16324, "loss": 0.2189, "lr": 5.295097052734454e-07, "epoch": 1.9118250589568466, "percentage": 95.6, "elapsed_time": "5:37:47", "remaining_time": "0:15:33", "throughput": 5153.71, "total_tokens": 104451328}
{"current_steps": 15610, "total_steps": 16324, "loss": 0.1891, "lr": 5.221835833145561e-07, "epoch": 1.91243759762335, "percentage": 95.63, "elapsed_time": "5:38:46", "remaining_time": "0:15:29", "throughput": 5140.45, "total_tokens": 104484896}
{"current_steps": 15615, "total_steps": 16324, "loss": 0.1933, "lr": 5.149082298949559e-07, "epoch": 1.9130501362898533, "percentage": 95.66, "elapsed_time": "5:39:44", "remaining_time": "0:15:25", "throughput": 5127.28, "total_tokens": 104518728}
{"current_steps": 15620, "total_steps": 16324, "loss": 0.2248, "lr": 5.076836524797735e-07, "epoch": 1.9136626749563566, "percentage": 95.69, "elapsed_time": "5:40:43", "remaining_time": "0:15:21", "throughput": 5114.26, "total_tokens": 104552224}
{"current_steps": 15625, "total_steps": 16324, "loss": 0.1867, "lr": 5.005098584820622e-07, "epoch": 1.9142752136228598, "percentage": 95.72, "elapsed_time": "5:41:42", "remaining_time": "0:15:17", "throughput": 5101.19, "total_tokens": 104585848}
{"current_steps": 15630, "total_steps": 16324, "loss": 0.2197, "lr": 4.933868552627563e-07, "epoch": 1.9148877522893633, "percentage": 95.75, "elapsed_time": "5:42:41", "remaining_time": "0:15:12", "throughput": 5088.04, "total_tokens": 104619440}
{"current_steps": 15635, "total_steps": 16324, "loss": 0.2361, "lr": 4.863146501306759e-07, "epoch": 1.9155002909558667, "percentage": 95.78, "elapsed_time": "5:43:40", "remaining_time": "0:15:08", "throughput": 5075.25, "total_tokens": 104652296}
{"current_steps": 15640, "total_steps": 16324, "loss": 0.2343, "lr": 4.792932503425163e-07, "epoch": 1.91611282962237, "percentage": 95.81, "elapsed_time": "5:44:39", "remaining_time": "0:15:04", "throughput": 5062.39, "total_tokens": 104685712}
{"current_steps": 15645, "total_steps": 16324, "loss": 0.2473, "lr": 4.7232266310284214e-07, "epoch": 1.9167253682888732, "percentage": 95.84, "elapsed_time": "5:45:38", "remaining_time": "0:15:00", "throughput": 5049.63, "total_tokens": 104719336}
{"current_steps": 15650, "total_steps": 16324, "loss": 0.1527, "lr": 4.6540289556409323e-07, "epoch": 1.9173379069553764, "percentage": 95.87, "elapsed_time": "5:46:35", "remaining_time": "0:14:55", "throughput": 5037.27, "total_tokens": 104753392}
{"current_steps": 15650, "total_steps": 16324, "eval_loss": 0.09997235238552094, "epoch": 1.9173379069553764, "percentage": 95.87, "elapsed_time": "5:46:54", "remaining_time": "0:14:56", "throughput": 5032.68, "total_tokens": 104753392}
{"current_steps": 15655, "total_steps": 16324, "loss": 0.1524, "lr": 4.5853395482653995e-07, "epoch": 1.9179504456218799, "percentage": 95.9, "elapsed_time": "5:47:58", "remaining_time": "0:14:52", "throughput": 5019.02, "total_tokens": 104787432}
{"current_steps": 15660, "total_steps": 16324, "loss": 0.2254, "lr": 4.5171584793831657e-07, "epoch": 1.9185629842883833, "percentage": 95.93, "elapsed_time": "5:48:56", "remaining_time": "0:14:47", "throughput": 5006.62, "total_tokens": 104820256}
{"current_steps": 15665, "total_steps": 16324, "loss": 0.2048, "lr": 4.4494858189539356e-07, "epoch": 1.9191755229548866, "percentage": 95.96, "elapsed_time": "5:49:53", "remaining_time": "0:14:43", "throughput": 4994.53, "total_tokens": 104853472}
{"current_steps": 15670, "total_steps": 16324, "loss": 0.1861, "lr": 4.3823216364157207e-07, "epoch": 1.9197880616213898, "percentage": 95.99, "elapsed_time": "5:50:51", "remaining_time": "0:14:38", "throughput": 4982.36, "total_tokens": 104887064}
{"current_steps": 15675, "total_steps": 16324, "loss": 0.2233, "lr": 4.315666000684837e-07, "epoch": 1.920400600287893, "percentage": 96.02, "elapsed_time": "5:51:50", "remaining_time": "0:14:34", "throughput": 4970.16, "total_tokens": 104920344}
{"current_steps": 15680, "total_steps": 16324, "loss": 0.1858, "lr": 4.2495189801557424e-07, "epoch": 1.9210131389543965, "percentage": 96.05, "elapsed_time": "5:52:47", "remaining_time": "0:14:29", "throughput": 4958.33, "total_tokens": 104954688}
{"current_steps": 15685, "total_steps": 16324, "loss": 0.1971, "lr": 4.183880642701088e-07, "epoch": 1.9216256776209, "percentage": 96.09, "elapsed_time": "5:53:45", "remaining_time": "0:14:24", "throughput": 4946.39, "total_tokens": 104987976}
{"current_steps": 15690, "total_steps": 16324, "loss": 0.1947, "lr": 4.1187510556713884e-07, "epoch": 1.9222382162874032, "percentage": 96.12, "elapsed_time": "5:54:43", "remaining_time": "0:14:20", "throughput": 4934.42, "total_tokens": 105021296}
{"current_steps": 15695, "total_steps": 16324, "loss": 0.2465, "lr": 4.0541302858953525e-07, "epoch": 1.9228507549539064, "percentage": 96.15, "elapsed_time": "5:55:41", "remaining_time": "0:14:15", "throughput": 4922.62, "total_tokens": 105054192}
{"current_steps": 15700, "total_steps": 16324, "loss": 0.2245, "lr": 3.9900183996794405e-07, "epoch": 1.9234632936204097, "percentage": 96.18, "elapsed_time": "5:56:38", "remaining_time": "0:14:10", "throughput": 4910.98, "total_tokens": 105087584}
{"current_steps": 15700, "total_steps": 16324, "eval_loss": 0.11582653969526291, "epoch": 1.9234632936204097, "percentage": 96.18, "elapsed_time": "5:56:57", "remaining_time": "0:14:11", "throughput": 4906.6, "total_tokens": 105087584}
{"current_steps": 15705, "total_steps": 16324, "loss": 0.1691, "lr": 3.926415462808086e-07, "epoch": 1.9240758322869131, "percentage": 96.21, "elapsed_time": "5:58:01", "remaining_time": "0:14:06", "throughput": 4893.5, "total_tokens": 105122072}
{"current_steps": 15710, "total_steps": 16324, "loss": 0.2275, "lr": 3.8633215405434185e-07, "epoch": 1.9246883709534164, "percentage": 96.24, "elapsed_time": "5:59:00", "remaining_time": "0:14:01", "throughput": 4881.85, "total_tokens": 105155088}
{"current_steps": 15715, "total_steps": 16324, "loss": 0.2168, "lr": 3.8007366976252623e-07, "epoch": 1.9253009096199198, "percentage": 96.27, "elapsed_time": "5:59:57", "remaining_time": "0:13:56", "throughput": 4870.37, "total_tokens": 105188144}
{"current_steps": 15720, "total_steps": 16324, "loss": 0.1978, "lr": 3.7386609982711373e-07, "epoch": 1.925913448286423, "percentage": 96.3, "elapsed_time": "6:00:55", "remaining_time": "0:13:52", "throughput": 4858.79, "total_tokens": 105221752}
{"current_steps": 15725, "total_steps": 16324, "loss": 0.1928, "lr": 3.6770945061760377e-07, "epoch": 1.9265259869529263, "percentage": 96.33, "elapsed_time": "6:01:54", "remaining_time": "0:13:47", "throughput": 4847.24, "total_tokens": 105255528}
{"current_steps": 15730, "total_steps": 16324, "loss": 0.2134, "lr": 3.6160372845125414e-07, "epoch": 1.9271385256194298, "percentage": 96.36, "elapsed_time": "6:02:52", "remaining_time": "0:13:42", "throughput": 4835.91, "total_tokens": 105288824}
{"current_steps": 15735, "total_steps": 16324, "loss": 0.2446, "lr": 3.5554893959308113e-07, "epoch": 1.927751064285933, "percentage": 96.39, "elapsed_time": "6:03:51", "remaining_time": "0:13:37", "throughput": 4824.43, "total_tokens": 105322376}
{"current_steps": 15740, "total_steps": 16324, "loss": 0.2499, "lr": 3.4954509025580393e-07, "epoch": 1.9283636029524365, "percentage": 96.42, "elapsed_time": "6:04:49", "remaining_time": "0:13:32", "throughput": 4813.06, "total_tokens": 105355488}
{"current_steps": 15745, "total_steps": 16324, "loss": 0.1947, "lr": 3.435921865999059e-07, "epoch": 1.9289761416189397, "percentage": 96.45, "elapsed_time": "6:05:47", "remaining_time": "0:13:27", "throughput": 4801.88, "total_tokens": 105389224}
{"current_steps": 15750, "total_steps": 16324, "loss": 0.209, "lr": 3.3769023473357866e-07, "epoch": 1.929588680285443, "percentage": 96.48, "elapsed_time": "6:06:46", "remaining_time": "0:13:22", "throughput": 4790.51, "total_tokens": 105422368}
{"current_steps": 15750, "total_steps": 16324, "eval_loss": 0.11289756000041962, "epoch": 1.929588680285443, "percentage": 96.48, "elapsed_time": "6:07:05", "remaining_time": "0:13:22", "throughput": 4786.34, "total_tokens": 105422368}
{"current_steps": 15755, "total_steps": 16324, "loss": 0.2341, "lr": 3.3183924071274464e-07, "epoch": 1.9302012189519462, "percentage": 96.51, "elapsed_time": "6:08:09", "remaining_time": "0:13:17", "throughput": 4773.96, "total_tokens": 105455608}
{"current_steps": 15760, "total_steps": 16324, "loss": 0.2335, "lr": 3.260392105410182e-07, "epoch": 1.9308137576184496, "percentage": 96.54, "elapsed_time": "6:09:07", "remaining_time": "0:13:12", "throughput": 4762.99, "total_tokens": 105488952}
{"current_steps": 15765, "total_steps": 16324, "loss": 0.186, "lr": 3.202901501697497e-07, "epoch": 1.931426296284953, "percentage": 96.58, "elapsed_time": "6:10:05", "remaining_time": "0:13:07", "throughput": 4752.03, "total_tokens": 105521848}
{"current_steps": 15770, "total_steps": 16324, "loss": 0.1425, "lr": 3.145920654979706e-07, "epoch": 1.9320388349514563, "percentage": 96.61, "elapsed_time": "6:11:04", "remaining_time": "0:13:02", "throughput": 4740.95, "total_tokens": 105556056}
{"current_steps": 15775, "total_steps": 16324, "loss": 0.1663, "lr": 3.0894496237240387e-07, "epoch": 1.9326513736179596, "percentage": 96.64, "elapsed_time": "6:12:02", "remaining_time": "0:12:56", "throughput": 4730.29, "total_tokens": 105589936}
{"current_steps": 15780, "total_steps": 16324, "loss": 0.1754, "lr": 3.033488465874701e-07, "epoch": 1.9332639122844628, "percentage": 96.67, "elapsed_time": "6:13:00", "remaining_time": "0:12:51", "throughput": 4719.52, "total_tokens": 105623640}
{"current_steps": 15785, "total_steps": 16324, "loss": 0.199, "lr": 2.97803723885276e-07, "epoch": 1.9338764509509663, "percentage": 96.7, "elapsed_time": "6:13:59", "remaining_time": "0:12:46", "throughput": 4708.62, "total_tokens": 105657280}
{"current_steps": 15790, "total_steps": 16324, "loss": 0.2091, "lr": 2.9230959995559804e-07, "epoch": 1.9344889896174697, "percentage": 96.73, "elapsed_time": "6:14:56", "remaining_time": "0:12:40", "throughput": 4698.02, "total_tokens": 105690576}
{"current_steps": 15795, "total_steps": 16324, "loss": 0.2042, "lr": 2.8686648043588227e-07, "epoch": 1.935101528283973, "percentage": 96.76, "elapsed_time": "6:15:54", "remaining_time": "0:12:35", "throughput": 4687.48, "total_tokens": 105724272}
{"current_steps": 15800, "total_steps": 16324, "loss": 0.2102, "lr": 2.8147437091123885e-07, "epoch": 1.9357140669504762, "percentage": 96.79, "elapsed_time": "6:16:53", "remaining_time": "0:12:29", "throughput": 4676.75, "total_tokens": 105757576}
{"current_steps": 15800, "total_steps": 16324, "eval_loss": 0.13141392171382904, "epoch": 1.9357140669504762, "percentage": 96.79, "elapsed_time": "6:17:12", "remaining_time": "0:12:30", "throughput": 4672.87, "total_tokens": 105757576}
{"current_steps": 15805, "total_steps": 16324, "loss": 0.1666, "lr": 2.7613327691444756e-07, "epoch": 1.9363266056169794, "percentage": 96.82, "elapsed_time": "6:18:14", "remaining_time": "0:12:25", "throughput": 4661.48, "total_tokens": 105791544}
{"current_steps": 15810, "total_steps": 16324, "loss": 0.1876, "lr": 2.7084320392593564e-07, "epoch": 1.936939144283483, "percentage": 96.85, "elapsed_time": "6:19:13", "remaining_time": "0:12:19", "throughput": 4651.04, "total_tokens": 105825424}
{"current_steps": 15815, "total_steps": 16324, "loss": 0.1917, "lr": 2.656041573737722e-07, "epoch": 1.9375516829499864, "percentage": 96.88, "elapsed_time": "6:20:11", "remaining_time": "0:12:14", "throughput": 4640.55, "total_tokens": 105858992}
{"current_steps": 15820, "total_steps": 16324, "loss": 0.2277, "lr": 2.6041614263367375e-07, "epoch": 1.9381642216164896, "percentage": 96.91, "elapsed_time": "6:21:09", "remaining_time": "0:12:08", "throughput": 4630.28, "total_tokens": 105892432}
{"current_steps": 15825, "total_steps": 16324, "loss": 0.2092, "lr": 2.5527916502899874e-07, "epoch": 1.9387767602829928, "percentage": 96.94, "elapsed_time": "6:22:07", "remaining_time": "0:12:02", "throughput": 4620.05, "total_tokens": 105925512}
{"current_steps": 15830, "total_steps": 16324, "loss": 0.2118, "lr": 2.501932298307308e-07, "epoch": 1.939389298949496, "percentage": 96.97, "elapsed_time": "6:23:06", "remaining_time": "0:11:57", "throughput": 4609.7, "total_tokens": 105959104}
{"current_steps": 15835, "total_steps": 16324, "loss": 0.1792, "lr": 2.4515834225748433e-07, "epoch": 1.9400018376159995, "percentage": 97.0, "elapsed_time": "6:24:04", "remaining_time": "0:11:51", "throughput": 4599.51, "total_tokens": 105993424}
{"current_steps": 15840, "total_steps": 16324, "loss": 0.1992, "lr": 2.401745074754824e-07, "epoch": 1.940614376282503, "percentage": 97.04, "elapsed_time": "6:25:02", "remaining_time": "0:11:45", "throughput": 4589.5, "total_tokens": 106027048}
{"current_steps": 15845, "total_steps": 16324, "loss": 0.1595, "lr": 2.3524173059858435e-07, "epoch": 1.9412269149490062, "percentage": 97.07, "elapsed_time": "6:26:01", "remaining_time": "0:11:40", "throughput": 4579.3, "total_tokens": 106061376}
{"current_steps": 15850, "total_steps": 16324, "loss": 0.2344, "lr": 2.3036001668824136e-07, "epoch": 1.9418394536155095, "percentage": 97.1, "elapsed_time": "6:26:59", "remaining_time": "0:11:34", "throughput": 4569.16, "total_tokens": 106094464}
{"current_steps": 15850, "total_steps": 16324, "eval_loss": 0.1082676351070404, "epoch": 1.9418394536155095, "percentage": 97.1, "elapsed_time": "6:27:18", "remaining_time": "0:11:34", "throughput": 4565.37, "total_tokens": 106094464}
{"current_steps": 15855, "total_steps": 16324, "loss": 0.2072, "lr": 2.2552937075351887e-07, "epoch": 1.9424519922820127, "percentage": 97.13, "elapsed_time": "6:28:22", "remaining_time": "0:11:29", "throughput": 4554.31, "total_tokens": 106127944}
{"current_steps": 15860, "total_steps": 16324, "loss": 0.2336, "lr": 2.207497977510742e-07, "epoch": 1.9430645309485162, "percentage": 97.16, "elapsed_time": "6:29:21", "remaining_time": "0:11:23", "throughput": 4544.19, "total_tokens": 106161040}
{"current_steps": 15865, "total_steps": 16324, "loss": 0.2168, "lr": 2.160213025851676e-07, "epoch": 1.9436770696150194, "percentage": 97.19, "elapsed_time": "6:30:20", "remaining_time": "0:11:17", "throughput": 4534.22, "total_tokens": 106193968}
{"current_steps": 15870, "total_steps": 16324, "loss": 0.2315, "lr": 2.1134389010765144e-07, "epoch": 1.9442896082815229, "percentage": 97.22, "elapsed_time": "6:31:18", "remaining_time": "0:11:11", "throughput": 4524.35, "total_tokens": 106226568}
{"current_steps": 15875, "total_steps": 16324, "loss": 0.1941, "lr": 2.0671756511794759e-07, "epoch": 1.944902146948026, "percentage": 97.25, "elapsed_time": "6:32:17", "remaining_time": "0:11:05", "throughput": 4514.42, "total_tokens": 106260320}
{"current_steps": 15880, "total_steps": 16324, "loss": 0.2008, "lr": 2.0214233236307556e-07, "epoch": 1.9455146856145293, "percentage": 97.28, "elapsed_time": "6:33:16", "remaining_time": "0:10:59", "throughput": 4504.56, "total_tokens": 106293088}
{"current_steps": 15885, "total_steps": 16324, "loss": 0.19, "lr": 1.9761819653761339e-07, "epoch": 1.9461272242810326, "percentage": 97.31, "elapsed_time": "6:34:14", "remaining_time": "0:10:53", "throughput": 4495.02, "total_tokens": 106326440}
{"current_steps": 15890, "total_steps": 16324, "loss": 0.1893, "lr": 1.9314516228372547e-07, "epoch": 1.946739762947536, "percentage": 97.34, "elapsed_time": "6:35:12", "remaining_time": "0:10:47", "throughput": 4485.45, "total_tokens": 106360320}
{"current_steps": 15895, "total_steps": 16324, "loss": 0.2165, "lr": 1.887232341911349e-07, "epoch": 1.9473523016140395, "percentage": 97.37, "elapsed_time": "6:36:10", "remaining_time": "0:10:41", "throughput": 4475.82, "total_tokens": 106393496}
{"current_steps": 15900, "total_steps": 16324, "loss": 0.1837, "lr": 1.8435241679711778e-07, "epoch": 1.9479648402805427, "percentage": 97.4, "elapsed_time": "6:37:08", "remaining_time": "0:10:35", "throughput": 4466.43, "total_tokens": 106427400}
{"current_steps": 15900, "total_steps": 16324, "eval_loss": 0.1029721349477768, "epoch": 1.9479648402805427, "percentage": 97.4, "elapsed_time": "6:37:27", "remaining_time": "0:10:35", "throughput": 4462.91, "total_tokens": 106427400}
{"current_steps": 15905, "total_steps": 16324, "loss": 0.2007, "lr": 1.800327145865144e-07, "epoch": 1.948577378947046, "percentage": 97.43, "elapsed_time": "6:38:30", "remaining_time": "0:10:29", "throughput": 4452.39, "total_tokens": 106460416}
{"current_steps": 15910, "total_steps": 16324, "loss": 0.2198, "lr": 1.7576413199171248e-07, "epoch": 1.9491899176135492, "percentage": 97.46, "elapsed_time": "6:39:30", "remaining_time": "0:10:23", "throughput": 4442.8, "total_tokens": 106493920}
{"current_steps": 15915, "total_steps": 16324, "loss": 0.197, "lr": 1.7154667339265295e-07, "epoch": 1.9498024562800527, "percentage": 97.49, "elapsed_time": "6:40:27", "remaining_time": "0:10:17", "throughput": 4433.49, "total_tokens": 106527576}
{"current_steps": 15920, "total_steps": 16324, "loss": 0.1965, "lr": 1.673803431168186e-07, "epoch": 1.9504149949465561, "percentage": 97.53, "elapsed_time": "6:41:26", "remaining_time": "0:10:11", "throughput": 4424.17, "total_tokens": 106561096}
{"current_steps": 15925, "total_steps": 16324, "loss": 0.2261, "lr": 1.6326514543921202e-07, "epoch": 1.9510275336130594, "percentage": 97.56, "elapsed_time": "6:42:24", "remaining_time": "0:10:04", "throughput": 4414.81, "total_tokens": 106594784}
{"current_steps": 15930, "total_steps": 16324, "loss": 0.1829, "lr": 1.592010845823999e-07, "epoch": 1.9516400722795626, "percentage": 97.59, "elapsed_time": "6:43:22", "remaining_time": "0:09:58", "throughput": 4405.62, "total_tokens": 106629144}
{"current_steps": 15935, "total_steps": 16324, "loss": 0.2445, "lr": 1.5518816471645214e-07, "epoch": 1.9522526109460658, "percentage": 97.62, "elapsed_time": "6:44:21", "remaining_time": "0:09:52", "throughput": 4396.33, "total_tokens": 106662152}
{"current_steps": 15940, "total_steps": 16324, "loss": 0.21, "lr": 1.5122638995898054e-07, "epoch": 1.9528651496125693, "percentage": 97.65, "elapsed_time": "6:45:20", "remaining_time": "0:09:45", "throughput": 4387.02, "total_tokens": 106695312}
{"current_steps": 15945, "total_steps": 16324, "loss": 0.1825, "lr": 1.4731576437509998e-07, "epoch": 1.9534776882790728, "percentage": 97.68, "elapsed_time": "6:46:18", "remaining_time": "0:09:39", "throughput": 4377.95, "total_tokens": 106729024}
{"current_steps": 15950, "total_steps": 16324, "loss": 0.2315, "lr": 1.4345629197745626e-07, "epoch": 1.954090226945576, "percentage": 97.71, "elapsed_time": "6:47:16", "remaining_time": "0:09:32", "throughput": 4368.92, "total_tokens": 106761720}
{"current_steps": 15950, "total_steps": 16324, "eval_loss": 0.1217665895819664, "epoch": 1.954090226945576, "percentage": 97.71, "elapsed_time": "6:47:36", "remaining_time": "0:09:33", "throughput": 4365.43, "total_tokens": 106761720}
{"current_steps": 15955, "total_steps": 16324, "loss": 0.1974, "lr": 1.3964797672620934e-07, "epoch": 1.9547027656120792, "percentage": 97.74, "elapsed_time": "6:48:40", "remaining_time": "0:09:27", "throughput": 4355.3, "total_tokens": 106795624}
{"current_steps": 15960, "total_steps": 16324, "loss": 0.1681, "lr": 1.358908225290112e-07, "epoch": 1.9553153042785825, "percentage": 97.77, "elapsed_time": "6:49:39", "remaining_time": "0:09:20", "throughput": 4346.29, "total_tokens": 106829912}
{"current_steps": 15965, "total_steps": 16324, "loss": 0.21, "lr": 1.3218483324103915e-07, "epoch": 1.955927842945086, "percentage": 97.8, "elapsed_time": "6:50:37", "remaining_time": "0:09:14", "throughput": 4337.36, "total_tokens": 106863336}
{"current_steps": 15970, "total_steps": 16324, "loss": 0.2095, "lr": 1.285300126649569e-07, "epoch": 1.9565403816115894, "percentage": 97.83, "elapsed_time": "6:51:37", "remaining_time": "0:09:07", "throughput": 4328.25, "total_tokens": 106896896}
{"current_steps": 15975, "total_steps": 16324, "loss": 0.1936, "lr": 1.2492636455092578e-07, "epoch": 1.9571529202780926, "percentage": 97.86, "elapsed_time": "6:52:36", "remaining_time": "0:09:00", "throughput": 4319.34, "total_tokens": 106930200}
{"current_steps": 15980, "total_steps": 16324, "loss": 0.1897, "lr": 1.2137389259659904e-07, "epoch": 1.9577654589445959, "percentage": 97.89, "elapsed_time": "6:53:34", "remaining_time": "0:08:54", "throughput": 4310.47, "total_tokens": 106963352}
{"current_steps": 15985, "total_steps": 16324, "loss": 0.2566, "lr": 1.1787260044712756e-07, "epoch": 1.958377997611099, "percentage": 97.92, "elapsed_time": "6:54:33", "remaining_time": "0:08:47", "throughput": 4301.59, "total_tokens": 106996336}
{"current_steps": 15990, "total_steps": 16324, "loss": 0.2063, "lr": 1.1442249169513752e-07, "epoch": 1.9589905362776026, "percentage": 97.95, "elapsed_time": "6:55:33", "remaining_time": "0:08:40", "throughput": 4292.69, "total_tokens": 107030304}
{"current_steps": 15995, "total_steps": 16324, "loss": 0.2248, "lr": 1.1102356988074714e-07, "epoch": 1.9596030749441058, "percentage": 97.98, "elapsed_time": "6:56:31", "remaining_time": "0:08:34", "throughput": 4283.97, "total_tokens": 107063136}
{"current_steps": 16000, "total_steps": 16324, "loss": 0.2309, "lr": 1.0767583849153328e-07, "epoch": 1.9602156136106093, "percentage": 98.02, "elapsed_time": "6:57:30", "remaining_time": "0:08:27", "throughput": 4275.21, "total_tokens": 107096336}
{"current_steps": 16000, "total_steps": 16324, "eval_loss": 0.12916378676891327, "epoch": 1.9602156136106093, "percentage": 98.02, "elapsed_time": "6:57:49", "remaining_time": "0:08:27", "throughput": 4271.95, "total_tokens": 107096336}
{"current_steps": 16005, "total_steps": 16324, "loss": 0.2443, "lr": 1.0437930096257043e-07, "epoch": 1.9608281522771125, "percentage": 98.05, "elapsed_time": "6:58:54", "remaining_time": "0:08:20", "throughput": 4262.25, "total_tokens": 107129472}
{"current_steps": 16010, "total_steps": 16324, "loss": 0.2061, "lr": 1.0113396067639169e-07, "epoch": 1.9614406909436157, "percentage": 98.08, "elapsed_time": "6:59:52", "remaining_time": "0:08:14", "throughput": 4253.78, "total_tokens": 107163088}
{"current_steps": 16015, "total_steps": 16324, "loss": 0.2235, "lr": 9.793982096298892e-08, "epoch": 1.9620532296101192, "percentage": 98.11, "elapsed_time": "7:00:50", "remaining_time": "0:08:07", "throughput": 4245.33, "total_tokens": 107196360}
{"current_steps": 16020, "total_steps": 16324, "loss": 0.204, "lr": 9.479688509983487e-08, "epoch": 1.9626657682766224, "percentage": 98.14, "elapsed_time": "7:01:49", "remaining_time": "0:08:00", "throughput": 4236.79, "total_tokens": 107230128}
{"current_steps": 16025, "total_steps": 16324, "loss": 0.2279, "lr": 9.170515631185539e-08, "epoch": 1.9632783069431259, "percentage": 98.17, "elapsed_time": "7:02:47", "remaining_time": "0:07:53", "throughput": 4228.45, "total_tokens": 107264040}
{"current_steps": 16030, "total_steps": 16324, "loss": 0.2044, "lr": 8.866463777142953e-08, "epoch": 1.9638908456096291, "percentage": 98.2, "elapsed_time": "7:03:45", "remaining_time": "0:07:46", "throughput": 4220.17, "total_tokens": 107298064}
{"current_steps": 16035, "total_steps": 16324, "loss": 0.1939, "lr": 8.567533259838945e-08, "epoch": 1.9645033842761324, "percentage": 98.23, "elapsed_time": "7:04:43", "remaining_time": "0:07:39", "throughput": 4211.75, "total_tokens": 107331904}
{"current_steps": 16040, "total_steps": 16324, "loss": 0.2031, "lr": 8.273724386003712e-08, "epoch": 1.9651159229426356, "percentage": 98.26, "elapsed_time": "7:05:41", "remaining_time": "0:07:32", "throughput": 4203.57, "total_tokens": 107366240}
{"current_steps": 16045, "total_steps": 16324, "loss": 0.1778, "lr": 7.985037457108879e-08, "epoch": 1.965728461609139, "percentage": 98.29, "elapsed_time": "7:06:39", "remaining_time": "0:07:25", "throughput": 4195.33, "total_tokens": 107399448}
{"current_steps": 16050, "total_steps": 16324, "loss": 0.2081, "lr": 7.701472769373607e-08, "epoch": 1.9663410002756425, "percentage": 98.32, "elapsed_time": "7:07:38", "remaining_time": "0:07:18", "throughput": 4186.98, "total_tokens": 107433008}
{"current_steps": 16050, "total_steps": 16324, "eval_loss": 0.10997606813907623, "epoch": 1.9663410002756425, "percentage": 98.32, "elapsed_time": "7:07:57", "remaining_time": "0:07:18", "throughput": 4183.88, "total_tokens": 107433008}
{"current_steps": 16055, "total_steps": 16324, "loss": 0.1744, "lr": 7.423030613759041e-08, "epoch": 1.9669535389421458, "percentage": 98.35, "elapsed_time": "7:09:01", "remaining_time": "0:07:11", "throughput": 4174.9, "total_tokens": 107467192}
{"current_steps": 16060, "total_steps": 16324, "loss": 0.221, "lr": 7.149711275971637e-08, "epoch": 1.967566077608649, "percentage": 98.38, "elapsed_time": "7:09:59", "remaining_time": "0:07:04", "throughput": 4166.8, "total_tokens": 107500864}
{"current_steps": 16065, "total_steps": 16324, "loss": 0.2284, "lr": 6.881515036460396e-08, "epoch": 1.9681786162751522, "percentage": 98.41, "elapsed_time": "7:10:58", "remaining_time": "0:06:56", "throughput": 4158.63, "total_tokens": 107534680}
{"current_steps": 16070, "total_steps": 16324, "loss": 0.1996, "lr": 6.618442170418515e-08, "epoch": 1.9687911549416557, "percentage": 98.44, "elapsed_time": "7:11:56", "remaining_time": "0:06:49", "throughput": 4150.63, "total_tokens": 107568272}
{"current_steps": 16075, "total_steps": 16324, "loss": 0.2096, "lr": 6.360492947781183e-08, "epoch": 1.9694036936081591, "percentage": 98.47, "elapsed_time": "7:12:54", "remaining_time": "0:06:42", "throughput": 4142.63, "total_tokens": 107601304}
{"current_steps": 16080, "total_steps": 16324, "loss": 0.1764, "lr": 6.107667633226677e-08, "epoch": 1.9700162322746624, "percentage": 98.51, "elapsed_time": "7:13:52", "remaining_time": "0:06:35", "throughput": 4134.58, "total_tokens": 107635416}
{"current_steps": 16085, "total_steps": 16324, "loss": 0.202, "lr": 5.8599664861758166e-08, "epoch": 1.9706287709411656, "percentage": 98.54, "elapsed_time": "7:14:51", "remaining_time": "0:06:27", "throughput": 4126.62, "total_tokens": 107669080}
{"current_steps": 16090, "total_steps": 16324, "loss": 0.1872, "lr": 5.6173897607914025e-08, "epoch": 1.9712413096076689, "percentage": 98.57, "elapsed_time": "7:15:49", "remaining_time": "0:06:20", "throughput": 4118.66, "total_tokens": 107702840}
{"current_steps": 16095, "total_steps": 16324, "loss": 0.2251, "lr": 5.379937705978777e-08, "epoch": 1.9718538482741723, "percentage": 98.6, "elapsed_time": "7:16:49", "remaining_time": "0:06:12", "throughput": 4110.67, "total_tokens": 107736744}
{"current_steps": 16100, "total_steps": 16324, "loss": 0.1759, "lr": 5.147610565383598e-08, "epoch": 1.9724663869406758, "percentage": 98.63, "elapsed_time": "7:17:47", "remaining_time": "0:06:05", "throughput": 4102.79, "total_tokens": 107770392}
{"current_steps": 16100, "total_steps": 16324, "eval_loss": 0.12338043004274368, "epoch": 1.9724663869406758, "percentage": 98.63, "elapsed_time": "7:18:06", "remaining_time": "0:06:05", "throughput": 4099.78, "total_tokens": 107770392}
{"current_steps": 16105, "total_steps": 16324, "loss": 0.1794, "lr": 4.920408577394065e-08, "epoch": 1.973078925607179, "percentage": 98.66, "elapsed_time": "7:19:10", "remaining_time": "0:05:58", "throughput": 4091.16, "total_tokens": 107804232}
{"current_steps": 16110, "total_steps": 16324, "loss": 0.2013, "lr": 4.6983319751392476e-08, "epoch": 1.9736914642736823, "percentage": 98.69, "elapsed_time": "7:20:10", "remaining_time": "0:05:50", "throughput": 4083.19, "total_tokens": 107837344}
{"current_steps": 16115, "total_steps": 16324, "loss": 0.2222, "lr": 4.481380986489092e-08, "epoch": 1.9743040029401855, "percentage": 98.72, "elapsed_time": "7:21:08", "remaining_time": "0:05:43", "throughput": 4075.36, "total_tokens": 107870296}
{"current_steps": 16120, "total_steps": 16324, "loss": 0.1532, "lr": 4.2695558340538623e-08, "epoch": 1.974916541606689, "percentage": 98.75, "elapsed_time": "7:22:07", "remaining_time": "0:05:35", "throughput": 4067.62, "total_tokens": 107904576}
{"current_steps": 16125, "total_steps": 16324, "loss": 0.2123, "lr": 4.062856735184695e-08, "epoch": 1.9755290802731924, "percentage": 98.78, "elapsed_time": "7:23:05", "remaining_time": "0:05:28", "throughput": 4060.02, "total_tokens": 107938600}
{"current_steps": 16130, "total_steps": 16324, "loss": 0.2383, "lr": 3.861283901973045e-08, "epoch": 1.9761416189396956, "percentage": 98.81, "elapsed_time": "7:24:04", "remaining_time": "0:05:20", "throughput": 4052.37, "total_tokens": 107972008}
{"current_steps": 16135, "total_steps": 16324, "loss": 0.2278, "lr": 3.664837541250132e-08, "epoch": 1.9767541576061989, "percentage": 98.84, "elapsed_time": "7:25:02", "remaining_time": "0:05:12", "throughput": 4044.75, "total_tokens": 108005224}
{"current_steps": 16140, "total_steps": 16324, "loss": 0.2028, "lr": 3.473517854586939e-08, "epoch": 1.9773666962727021, "percentage": 98.87, "elapsed_time": "7:26:01", "remaining_time": "0:05:05", "throughput": 4037.15, "total_tokens": 108038648}
{"current_steps": 16145, "total_steps": 16324, "loss": 0.1958, "lr": 3.287325038293654e-08, "epoch": 1.9779792349392056, "percentage": 98.9, "elapsed_time": "7:27:00", "remaining_time": "0:04:57", "throughput": 4029.5, "total_tokens": 108072496}
{"current_steps": 16150, "total_steps": 16324, "loss": 0.1927, "lr": 3.106259283420787e-08, "epoch": 1.9785917736057088, "percentage": 98.93, "elapsed_time": "7:27:57", "remaining_time": "0:04:49", "throughput": 4022.15, "total_tokens": 108105848}
{"current_steps": 16150, "total_steps": 16324, "eval_loss": 0.12711873650550842, "epoch": 1.9785917736057088, "percentage": 98.93, "elapsed_time": "7:28:16", "remaining_time": "0:04:49", "throughput": 4019.36, "total_tokens": 108105848}
{"current_steps": 16155, "total_steps": 16324, "loss": 0.1994, "lr": 2.930320775756945e-08, "epoch": 1.9792043122722123, "percentage": 98.96, "elapsed_time": "7:29:19", "remaining_time": "0:04:42", "throughput": 4011.17, "total_tokens": 108139704}
{"current_steps": 16160, "total_steps": 16324, "loss": 0.2619, "lr": 2.759509695831608e-08, "epoch": 1.9798168509387155, "percentage": 99.0, "elapsed_time": "7:30:18", "remaining_time": "0:04:34", "throughput": 4003.68, "total_tokens": 108172816}
{"current_steps": 16165, "total_steps": 16324, "loss": 0.2145, "lr": 2.5938262189101338e-08, "epoch": 1.9804293896052187, "percentage": 99.03, "elapsed_time": "7:31:16", "remaining_time": "0:04:26", "throughput": 3996.37, "total_tokens": 108206544}
{"current_steps": 16170, "total_steps": 16324, "loss": 0.2193, "lr": 2.4332705149993086e-08, "epoch": 1.9810419282717222, "percentage": 99.06, "elapsed_time": "7:32:14", "remaining_time": "0:04:18", "throughput": 3989.01, "total_tokens": 108239608}
{"current_steps": 16175, "total_steps": 16324, "loss": 0.2354, "lr": 2.2778427488434618e-08, "epoch": 1.9816544669382254, "percentage": 99.09, "elapsed_time": "7:33:13", "remaining_time": "0:04:10", "throughput": 3981.57, "total_tokens": 108273024}
{"current_steps": 16180, "total_steps": 16324, "loss": 0.1882, "lr": 2.1275430799239105e-08, "epoch": 1.982267005604729, "percentage": 99.12, "elapsed_time": "7:34:11", "remaining_time": "0:04:02", "throughput": 3974.39, "total_tokens": 108307408}
{"current_steps": 16185, "total_steps": 16324, "loss": 0.2061, "lr": 1.982371662461735e-08, "epoch": 1.9828795442712321, "percentage": 99.15, "elapsed_time": "7:35:09", "remaining_time": "0:03:54", "throughput": 3967.13, "total_tokens": 108340504}
{"current_steps": 16190, "total_steps": 16324, "loss": 0.2093, "lr": 1.8423286454150036e-08, "epoch": 1.9834920829377354, "percentage": 99.18, "elapsed_time": "7:36:08", "remaining_time": "0:03:46", "throughput": 3959.78, "total_tokens": 108374192}
{"current_steps": 16195, "total_steps": 16324, "loss": 0.1632, "lr": 1.707414172480992e-08, "epoch": 1.9841046216042386, "percentage": 99.21, "elapsed_time": "7:37:06", "remaining_time": "0:03:38", "throughput": 3952.65, "total_tokens": 108408208}
{"current_steps": 16200, "total_steps": 16324, "loss": 0.1823, "lr": 1.577628382092855e-08, "epoch": 1.984717160270742, "percentage": 99.24, "elapsed_time": "7:38:04", "remaining_time": "0:03:30", "throughput": 3945.5, "total_tokens": 108441584}
{"current_steps": 16200, "total_steps": 16324, "eval_loss": 0.10478620231151581, "epoch": 1.984717160270742, "percentage": 99.24, "elapsed_time": "7:38:23", "remaining_time": "0:03:30", "throughput": 3942.79, "total_tokens": 108441584}
{"current_steps": 16205, "total_steps": 16324, "loss": 0.1985, "lr": 1.4529714074223988e-08, "epoch": 1.9853296989372455, "percentage": 99.27, "elapsed_time": "7:39:28", "remaining_time": "0:03:22", "throughput": 3934.79, "total_tokens": 108474648}
{"current_steps": 16210, "total_steps": 16324, "loss": 0.189, "lr": 1.3334433763784181e-08, "epoch": 1.9859422376037488, "percentage": 99.3, "elapsed_time": "7:40:26", "remaining_time": "0:03:14", "throughput": 3927.7, "total_tokens": 108508000}
{"current_steps": 16215, "total_steps": 16324, "loss": 0.164, "lr": 1.2190444116066957e-08, "epoch": 1.986554776270252, "percentage": 99.33, "elapsed_time": "7:41:25", "remaining_time": "0:03:06", "throughput": 3920.56, "total_tokens": 108542400}
{"current_steps": 16220, "total_steps": 16324, "loss": 0.183, "lr": 1.1097746304916667e-08, "epoch": 1.9871673149367552, "percentage": 99.36, "elapsed_time": "7:42:24", "remaining_time": "0:02:57", "throughput": 3913.38, "total_tokens": 108576552}
{"current_steps": 16225, "total_steps": 16324, "loss": 0.1981, "lr": 1.0056341451525342e-08, "epoch": 1.9877798536032587, "percentage": 99.39, "elapsed_time": "7:43:23", "remaining_time": "0:02:49", "throughput": 3906.3, "total_tokens": 108610120}
{"current_steps": 16230, "total_steps": 16324, "loss": 0.2291, "lr": 9.066230624465988e-09, "epoch": 1.9883923922697622, "percentage": 99.42, "elapsed_time": "7:44:22", "remaining_time": "0:02:41", "throughput": 3899.26, "total_tokens": 108643208}
{"current_steps": 16235, "total_steps": 16324, "loss": 0.2264, "lr": 8.12741483968149e-09, "epoch": 1.9890049309362654, "percentage": 99.45, "elapsed_time": "7:45:21", "remaining_time": "0:02:33", "throughput": 3892.24, "total_tokens": 108676224}
{"current_steps": 16240, "total_steps": 16324, "loss": 0.2312, "lr": 7.239895060479063e-09, "epoch": 1.9896174696027686, "percentage": 99.49, "elapsed_time": "7:46:19", "remaining_time": "0:02:24", "throughput": 3885.28, "total_tokens": 108708432}
{"current_steps": 16245, "total_steps": 16324, "loss": 0.1797, "lr": 6.4036721975246905e-09, "epoch": 1.9902300082692719, "percentage": 99.52, "elapsed_time": "7:47:17", "remaining_time": "0:02:16", "throughput": 3878.52, "total_tokens": 108742768}
{"current_steps": 16250, "total_steps": 16324, "loss": 0.2257, "lr": 5.618747108865341e-09, "epoch": 1.9908425469357753, "percentage": 99.55, "elapsed_time": "7:48:15", "remaining_time": "0:02:07", "throughput": 3871.69, "total_tokens": 108775784}
{"current_steps": 16250, "total_steps": 16324, "eval_loss": 0.13182586431503296, "epoch": 1.9908425469357753, "percentage": 99.55, "elapsed_time": "7:48:33", "remaining_time": "0:02:08", "throughput": 3869.12, "total_tokens": 108775784}
{"current_steps": 16255, "total_steps": 16324, "loss": 0.2066, "lr": 4.885120599895654e-09, "epoch": 1.9914550856022788, "percentage": 99.58, "elapsed_time": "7:49:37", "remaining_time": "0:01:59", "throughput": 3861.65, "total_tokens": 108809944}
{"current_steps": 16260, "total_steps": 16324, "loss": 0.2017, "lr": 4.202793423380147e-09, "epoch": 1.992067624268782, "percentage": 99.61, "elapsed_time": "7:50:34", "remaining_time": "0:01:51", "throughput": 3854.95, "total_tokens": 108842592}
{"current_steps": 16265, "total_steps": 16324, "loss": 0.1999, "lr": 3.571766279453215e-09, "epoch": 1.9926801629352853, "percentage": 99.64, "elapsed_time": "7:51:32", "remaining_time": "0:01:42", "throughput": 3848.28, "total_tokens": 108876200}
{"current_steps": 16270, "total_steps": 16324, "loss": 0.1854, "lr": 2.9920398155913742e-09, "epoch": 1.9932927016017885, "percentage": 99.67, "elapsed_time": "7:52:30", "remaining_time": "0:01:34", "throughput": 3841.55, "total_tokens": 108910256}
{"current_steps": 16275, "total_steps": 16324, "loss": 0.2155, "lr": 2.4636146266521222e-09, "epoch": 1.993905240268292, "percentage": 99.7, "elapsed_time": "7:53:27", "remaining_time": "0:01:25", "throughput": 3835.0, "total_tokens": 108944032}
{"current_steps": 16280, "total_steps": 16324, "loss": 0.2018, "lr": 1.986491254846179e-09, "epoch": 1.9945177789347952, "percentage": 99.73, "elapsed_time": "7:54:25", "remaining_time": "0:01:16", "throughput": 3828.4, "total_tokens": 108977664}
{"current_steps": 16285, "total_steps": 16324, "loss": 0.2066, "lr": 1.5606701897430409e-09, "epoch": 1.9951303176012987, "percentage": 99.76, "elapsed_time": "7:55:24", "remaining_time": "0:01:08", "throughput": 3821.69, "total_tokens": 109010600}
{"current_steps": 16290, "total_steps": 16324, "loss": 0.1816, "lr": 1.1861518682709793e-09, "epoch": 1.995742856267802, "percentage": 99.79, "elapsed_time": "7:56:21", "remaining_time": "0:00:59", "throughput": 3815.22, "total_tokens": 109044464}
{"current_steps": 16295, "total_steps": 16324, "loss": 0.1816, "lr": 8.629366747170408e-10, "epoch": 1.9963553949343051, "percentage": 99.82, "elapsed_time": "7:57:19", "remaining_time": "0:00:50", "throughput": 3808.74, "total_tokens": 109078888}
{"current_steps": 16300, "total_steps": 16324, "loss": 0.2158, "lr": 5.910249407270474e-10, "epoch": 1.9969679336008086, "percentage": 99.85, "elapsed_time": "7:58:17", "remaining_time": "0:00:42", "throughput": 3802.14, "total_tokens": 109112120}
{"current_steps": 16300, "total_steps": 16324, "eval_loss": 0.11464700102806091, "epoch": 1.9969679336008086, "percentage": 99.85, "elapsed_time": "7:58:36", "remaining_time": "0:00:42", "throughput": 3799.61, "total_tokens": 109112120}
{"current_steps": 16305, "total_steps": 16324, "loss": 0.2376, "lr": 3.7041694531114723e-10, "epoch": 1.9975804722673118, "percentage": 99.88, "elapsed_time": "7:59:40", "remaining_time": "0:00:33", "throughput": 3792.37, "total_tokens": 109145312}
{"current_steps": 16310, "total_steps": 16324, "loss": 0.2031, "lr": 2.0111291483271288e-10, "epoch": 1.9981930109338153, "percentage": 99.91, "elapsed_time": "8:00:38", "remaining_time": "0:00:24", "throughput": 3785.87, "total_tokens": 109179312}
{"current_steps": 16315, "total_steps": 16324, "loss": 0.2135, "lr": 8.311302300278989e-11, "epoch": 1.9988055496003185, "percentage": 99.94, "elapsed_time": "8:01:37", "remaining_time": "0:00:15", "throughput": 3779.33, "total_tokens": 109212488}
{"current_steps": 16320, "total_steps": 16324, "loss": 0.2092, "lr": 1.6417390907852438e-11, "epoch": 1.9994180882668218, "percentage": 99.98, "elapsed_time": "8:02:35", "remaining_time": "0:00:07", "throughput": 3772.91, "total_tokens": 109246064}
{"current_steps": 16324, "total_steps": 16324, "epoch": 1.9999081192000245, "percentage": 100.0, "elapsed_time": "8:03:27", "remaining_time": "0:00:00", "throughput": 3767.11, "total_tokens": 109273408}