{"current_steps": 5, "total_steps": 3400, "loss": 3.0444, "lr": 2.9411764705882355e-06, "epoch": 0.0025753283543651817, "percentage": 0.15, "elapsed_time": "0:02:03", "remaining_time": "23:19:52", "throughput": 472.88, "total_tokens": 58496}
{"current_steps": 10, "total_steps": 3400, "loss": 2.9824, "lr": 5.882352941176471e-06, "epoch": 0.0051506567087303634, "percentage": 0.29, "elapsed_time": "0:03:11", "remaining_time": "18:03:18", "throughput": 610.01, "total_tokens": 116960}
{"current_steps": 15, "total_steps": 3400, "loss": 2.8371, "lr": 8.823529411764707e-06, "epoch": 0.007725985063095545, "percentage": 0.44, "elapsed_time": "0:04:20", "remaining_time": "16:19:06", "throughput": 673.96, "total_tokens": 175448}
{"current_steps": 20, "total_steps": 3400, "loss": 2.5198, "lr": 1.1764705882352942e-05, "epoch": 0.010301313417460727, "percentage": 0.59, "elapsed_time": "0:05:28", "remaining_time": "15:24:32", "throughput": 712.72, "total_tokens": 233944}
{"current_steps": 5, "total_steps": 3400, "loss": 3.0444, "lr": 2.9411764705882355e-06, "epoch": 0.0025753283543651817, "percentage": 0.15, "elapsed_time": "0:02:03", "remaining_time": "23:21:43", "throughput": 472.26, "total_tokens": 58496}
{"current_steps": 10, "total_steps": 3400, "loss": 2.9824, "lr": 5.882352941176471e-06, "epoch": 0.0051506567087303634, "percentage": 0.29, "elapsed_time": "0:03:12", "remaining_time": "18:05:24", "throughput": 608.83, "total_tokens": 116960}
{"current_steps": 15, "total_steps": 3400, "loss": 2.8371, "lr": 8.823529411764707e-06, "epoch": 0.007725985063095545, "percentage": 0.44, "elapsed_time": "0:04:20", "remaining_time": "16:19:01", "throughput": 674.02, "total_tokens": 175448}
{"current_steps": 20, "total_steps": 3400, "loss": 2.5198, "lr": 1.1764705882352942e-05, "epoch": 0.010301313417460727, "percentage": 0.59, "elapsed_time": "0:05:28", "remaining_time": "15:25:09", "throughput": 712.24, "total_tokens": 233944}
{"current_steps": 25, "total_steps": 3400, "loss": 1.772, "lr": 1.4705882352941177e-05, "epoch": 0.012876641771825908, "percentage": 0.74, "elapsed_time": "0:06:35", "remaining_time": "14:50:39", "throughput": 738.71, "total_tokens": 292416}
{"current_steps": 30, "total_steps": 3400, "loss": 1.2263, "lr": 1.7647058823529414e-05, "epoch": 0.01545197012619109, "percentage": 0.88, "elapsed_time": "0:07:43", "remaining_time": "14:27:35", "throughput": 757.24, "total_tokens": 350904}
{"current_steps": 35, "total_steps": 3400, "loss": 1.0102, "lr": 2.058823529411765e-05, "epoch": 0.018027298480556272, "percentage": 1.03, "elapsed_time": "0:08:51", "remaining_time": "14:11:22", "throughput": 770.5, "total_tokens": 409384}
{"current_steps": 40, "total_steps": 3400, "loss": 0.9378, "lr": 2.3529411764705884e-05, "epoch": 0.020602626834921454, "percentage": 1.18, "elapsed_time": "0:09:59", "remaining_time": "13:59:28", "throughput": 780.26, "total_tokens": 467864}
{"current_steps": 45, "total_steps": 3400, "loss": 0.9265, "lr": 2.647058823529412e-05, "epoch": 0.023177955189286635, "percentage": 1.32, "elapsed_time": "0:11:07", "remaining_time": "13:49:36", "throughput": 788.43, "total_tokens": 526384}
{"current_steps": 50, "total_steps": 3400, "loss": 0.9157, "lr": 2.9411764705882354e-05, "epoch": 0.025753283543651816, "percentage": 1.47, "elapsed_time": "0:12:15", "remaining_time": "13:41:29", "throughput": 795.0, "total_tokens": 584856}
{"current_steps": 50, "total_steps": 3400, "eval_loss": 0.9191630482673645, "epoch": 0.025753283543651816, "percentage": 1.47, "elapsed_time": "0:12:52", "remaining_time": "14:22:23", "throughput": 757.31, "total_tokens": 584856}
{"current_steps": 55, "total_steps": 3400, "loss": 0.9009, "lr": 3.235294117647059e-05, "epoch": 0.028328611898016998, "percentage": 1.62, "elapsed_time": "0:14:08", "remaining_time": "14:19:34", "throughput": 758.66, "total_tokens": 643344}
{"current_steps": 60, "total_steps": 3400, "loss": 0.9063, "lr": 3.529411764705883e-05, "epoch": 0.03090394025238218, "percentage": 1.76, "elapsed_time": "0:15:15", "remaining_time": "14:09:45", "throughput": 766.24, "total_tokens": 701808}
{"current_steps": 65, "total_steps": 3400, "loss": 0.9031, "lr": 3.8235294117647055e-05, "epoch": 0.03347926860674736, "percentage": 1.91, "elapsed_time": "0:16:23", "remaining_time": "14:01:11", "throughput": 772.9, "total_tokens": 760304}
{"current_steps": 70, "total_steps": 3400, "loss": 0.8991, "lr": 4.11764705882353e-05, "epoch": 0.036054596961112545, "percentage": 2.06, "elapsed_time": "0:17:31", "remaining_time": "13:53:40", "throughput": 778.67, "total_tokens": 818760}
{"current_steps": 75, "total_steps": 3400, "loss": 0.9055, "lr": 4.411764705882353e-05, "epoch": 0.03862992531547772, "percentage": 2.21, "elapsed_time": "0:18:38", "remaining_time": "13:46:47", "throughput": 783.98, "total_tokens": 877256}
{"current_steps": 80, "total_steps": 3400, "loss": 0.9092, "lr": 4.705882352941177e-05, "epoch": 0.04120525366984291, "percentage": 2.35, "elapsed_time": "0:19:46", "remaining_time": "13:40:53", "throughput": 788.45, "total_tokens": 935752}
{"current_steps": 85, "total_steps": 3400, "loss": 0.9069, "lr": 5e-05, "epoch": 0.043780582024208085, "percentage": 2.5, "elapsed_time": "0:20:53", "remaining_time": "13:35:01", "throughput": 792.91, "total_tokens": 994216}
{"current_steps": 90, "total_steps": 3400, "loss": 0.8924, "lr": 5.294117647058824e-05, "epoch": 0.04635591037857327, "percentage": 2.65, "elapsed_time": "0:22:02", "remaining_time": "13:30:22", "throughput": 796.26, "total_tokens": 1052704}
{"current_steps": 95, "total_steps": 3400, "loss": 0.9059, "lr": 5.588235294117647e-05, "epoch": 0.04893123873293845, "percentage": 2.79, "elapsed_time": "0:23:10", "remaining_time": "13:26:06", "throughput": 799.26, "total_tokens": 1111176}
{"current_steps": 100, "total_steps": 3400, "loss": 0.901, "lr": 5.882352941176471e-05, "epoch": 0.05150656708730363, "percentage": 2.94, "elapsed_time": "0:24:20", "remaining_time": "13:23:15", "throughput": 800.88, "total_tokens": 1169664}
{"current_steps": 100, "total_steps": 3400, "eval_loss": 0.9077914953231812, "epoch": 0.05150656708730363, "percentage": 2.94, "elapsed_time": "0:24:37", "remaining_time": "13:32:32", "throughput": 791.73, "total_tokens": 1169664}
{"current_steps": 105, "total_steps": 3400, "loss": 0.9159, "lr": 6.176470588235295e-05, "epoch": 0.05408189544166881, "percentage": 3.09, "elapsed_time": "0:25:51", "remaining_time": "13:31:17", "throughput": 791.74, "total_tokens": 1228112}
{"current_steps": 110, "total_steps": 3400, "loss": 0.91, "lr": 6.470588235294118e-05, "epoch": 0.056657223796033995, "percentage": 3.24, "elapsed_time": "0:27:00", "remaining_time": "13:27:33", "throughput": 794.19, "total_tokens": 1286608}
{"current_steps": 115, "total_steps": 3400, "loss": 0.9047, "lr": 6.764705882352942e-05, "epoch": 0.05923255215039917, "percentage": 3.38, "elapsed_time": "0:28:08", "remaining_time": "13:23:45", "throughput": 796.72, "total_tokens": 1345072}
{"current_steps": 120, "total_steps": 3400, "loss": 0.9022, "lr": 7.058823529411765e-05, "epoch": 0.06180788050476436, "percentage": 3.53, "elapsed_time": "0:29:16", "remaining_time": "13:20:18", "throughput": 798.94, "total_tokens": 1403544}
{"current_steps": 125, "total_steps": 3400, "loss": 0.9081, "lr": 7.352941176470589e-05, "epoch": 0.06438320885912954, "percentage": 3.68, "elapsed_time": "0:30:25", "remaining_time": "13:17:01", "throughput": 801.0, "total_tokens": 1462024}
{"current_steps": 130, "total_steps": 3400, "loss": 0.8939, "lr": 7.647058823529411e-05, "epoch": 0.06695853721349472, "percentage": 3.82, "elapsed_time": "0:31:33", "remaining_time": "13:13:51", "throughput": 802.98, "total_tokens": 1520528}
{"current_steps": 135, "total_steps": 3400, "loss": 0.9029, "lr": 7.941176470588235e-05, "epoch": 0.0695338655678599, "percentage": 3.97, "elapsed_time": "0:32:42", "remaining_time": "13:10:59", "throughput": 804.66, "total_tokens": 1579024}
{"current_steps": 140, "total_steps": 3400, "loss": 0.9014, "lr": 8.23529411764706e-05, "epoch": 0.07210919392222509, "percentage": 4.12, "elapsed_time": "0:33:50", "remaining_time": "13:07:57", "throughput": 806.53, "total_tokens": 1637504}
{"current_steps": 145, "total_steps": 3400, "loss": 0.9053, "lr": 8.529411764705883e-05, "epoch": 0.07468452227659027, "percentage": 4.26, "elapsed_time": "0:34:58", "remaining_time": "13:05:04", "throughput": 808.26, "total_tokens": 1696024}
{"current_steps": 150, "total_steps": 3400, "loss": 0.9032, "lr": 8.823529411764706e-05, "epoch": 0.07725985063095545, "percentage": 4.41, "elapsed_time": "0:36:06", "remaining_time": "13:02:20", "throughput": 809.84, "total_tokens": 1754512}
{"current_steps": 150, "total_steps": 3400, "eval_loss": 0.8962129950523376, "epoch": 0.07725985063095545, "percentage": 4.41, "elapsed_time": "0:36:23", "remaining_time": "13:08:30", "throughput": 803.51, "total_tokens": 1754512}
{"current_steps": 155, "total_steps": 3400, "loss": 0.8985, "lr": 9.11764705882353e-05, "epoch": 0.07983517898532062, "percentage": 4.56, "elapsed_time": "0:37:40", "remaining_time": "13:08:35", "throughput": 802.19, "total_tokens": 1812976}
{"current_steps": 160, "total_steps": 3400, "loss": 0.8949, "lr": 9.411764705882353e-05, "epoch": 0.08241050733968582, "percentage": 4.71, "elapsed_time": "0:38:48", "remaining_time": "13:05:42", "throughput": 803.89, "total_tokens": 1871464}
{"current_steps": 165, "total_steps": 3400, "loss": 0.9069, "lr": 9.705882352941177e-05, "epoch": 0.08498583569405099, "percentage": 4.85, "elapsed_time": "0:39:55", "remaining_time": "13:02:45", "throughput": 805.66, "total_tokens": 1929928}
{"current_steps": 170, "total_steps": 3400, "loss": 0.9049, "lr": 0.0001, "epoch": 0.08756116404841617, "percentage": 5.0, "elapsed_time": "0:41:03", "remaining_time": "13:00:00", "throughput": 807.27, "total_tokens": 1988432}
{"current_steps": 175, "total_steps": 3400, "loss": 0.9026, "lr": 9.999940874631277e-05, "epoch": 0.09013649240278135, "percentage": 5.15, "elapsed_time": "0:42:10", "remaining_time": "12:57:05", "throughput": 809.04, "total_tokens": 2046920}
{"current_steps": 180, "total_steps": 3400, "loss": 0.8984, "lr": 9.999763499923432e-05, "epoch": 0.09271182075714654, "percentage": 5.29, "elapsed_time": "0:43:16", "remaining_time": "12:54:16", "throughput": 810.72, "total_tokens": 2105392}
{"current_steps": 185, "total_steps": 3400, "loss": 0.9057, "lr": 9.999467880071402e-05, "epoch": 0.09528714911151172, "percentage": 5.44, "elapsed_time": "0:44:23", "remaining_time": "12:51:33", "throughput": 812.3, "total_tokens": 2163872}
{"current_steps": 190, "total_steps": 3400, "loss": 0.9078, "lr": 9.999054022066641e-05, "epoch": 0.0978624774658769, "percentage": 5.59, "elapsed_time": "0:45:31", "remaining_time": "12:49:08", "throughput": 813.59, "total_tokens": 2222352}
{"current_steps": 195, "total_steps": 3400, "loss": 0.9028, "lr": 9.998521935696953e-05, "epoch": 0.10043780582024209, "percentage": 5.74, "elapsed_time": "0:46:38", "remaining_time": "12:46:33", "throughput": 815.05, "total_tokens": 2280800}
{"current_steps": 200, "total_steps": 3400, "loss": 0.9053, "lr": 9.997871633546257e-05, "epoch": 0.10301313417460727, "percentage": 5.88, "elapsed_time": "0:47:45", "remaining_time": "12:44:03", "throughput": 816.44, "total_tokens": 2339304}
{"current_steps": 200, "total_steps": 3400, "eval_loss": 0.8982028961181641, "epoch": 0.10301313417460727, "percentage": 5.88, "elapsed_time": "0:48:02", "remaining_time": "12:48:34", "throughput": 811.65, "total_tokens": 2339304}
{"current_steps": 205, "total_steps": 3400, "loss": 0.9003, "lr": 9.997103130994296e-05, "epoch": 0.10558846252897244, "percentage": 6.03, "elapsed_time": "0:49:13", "remaining_time": "12:47:17", "throughput": 811.74, "total_tokens": 2397808}
{"current_steps": 210, "total_steps": 3400, "loss": 0.8969, "lr": 9.996216446216267e-05, "epoch": 0.10816379088333762, "percentage": 6.18, "elapsed_time": "0:50:20", "remaining_time": "12:44:49", "throughput": 813.08, "total_tokens": 2456288}
{"current_steps": 215, "total_steps": 3400, "loss": 0.9114, "lr": 9.995211600182397e-05, "epoch": 0.11073911923770281, "percentage": 6.32, "elapsed_time": "0:51:28", "remaining_time": "12:42:28", "throughput": 814.32, "total_tokens": 2514784}
{"current_steps": 220, "total_steps": 3400, "loss": 0.899, "lr": 9.994088616657444e-05, "epoch": 0.11331444759206799, "percentage": 6.47, "elapsed_time": "0:52:35", "remaining_time": "12:40:13", "throughput": 815.44, "total_tokens": 2573240}
{"current_steps": 225, "total_steps": 3400, "loss": 0.898, "lr": 9.992847522200133e-05, "epoch": 0.11588977594643317, "percentage": 6.62, "elapsed_time": "0:53:42", "remaining_time": "12:37:54", "throughput": 816.63, "total_tokens": 2631672}
{"current_steps": 230, "total_steps": 3400, "loss": 0.9006, "lr": 9.99148834616253e-05, "epoch": 0.11846510430079835, "percentage": 6.76, "elapsed_time": "0:54:49", "remaining_time": "12:35:36", "throughput": 817.82, "total_tokens": 2690112}
{"current_steps": 235, "total_steps": 3400, "loss": 0.8973, "lr": 9.990011120689351e-05, "epoch": 0.12104043265516354, "percentage": 6.91, "elapsed_time": "0:55:56", "remaining_time": "12:33:23", "throughput": 818.92, "total_tokens": 2748608}
{"current_steps": 240, "total_steps": 3400, "loss": 0.8885, "lr": 9.988415880717194e-05, "epoch": 0.12361576100952872, "percentage": 7.06, "elapsed_time": "0:57:03", "remaining_time": "12:31:13", "throughput": 820.0, "total_tokens": 2807080}
{"current_steps": 245, "total_steps": 3400, "loss": 0.9066, "lr": 9.986702663973722e-05, "epoch": 0.1261910893638939, "percentage": 7.21, "elapsed_time": "0:58:10", "remaining_time": "12:29:07", "throughput": 820.98, "total_tokens": 2865520}
{"current_steps": 250, "total_steps": 3400, "loss": 0.9098, "lr": 9.98487151097676e-05, "epoch": 0.12876641771825909, "percentage": 7.35, "elapsed_time": "0:59:17", "remaining_time": "12:26:59", "throughput": 822.02, "total_tokens": 2924016}
{"current_steps": 250, "total_steps": 3400, "eval_loss": 0.8956434726715088, "epoch": 0.12876641771825909, "percentage": 7.35, "elapsed_time": "0:59:34", "remaining_time": "12:30:40", "throughput": 818.0, "total_tokens": 2924016}
{"current_steps": 255, "total_steps": 3400, "loss": 0.8987, "lr": 9.98292246503335e-05, "epoch": 0.13134174607262425, "percentage": 7.5, "elapsed_time": "1:00:46", "remaining_time": "12:29:36", "throughput": 817.86, "total_tokens": 2982520}
{"current_steps": 260, "total_steps": 3400, "loss": 0.9036, "lr": 9.980855572238714e-05, "epoch": 0.13391707442698944, "percentage": 7.65, "elapsed_time": "1:01:53", "remaining_time": "12:27:29", "throughput": 818.88, "total_tokens": 3041008}
{"current_steps": 265, "total_steps": 3400, "loss": 0.8961, "lr": 9.978670881475172e-05, "epoch": 0.13649240278135463, "percentage": 7.79, "elapsed_time": "1:03:00", "remaining_time": "12:25:24", "throughput": 819.86, "total_tokens": 3099464}
{"current_steps": 270, "total_steps": 3400, "loss": 0.9012, "lr": 9.976368444410985e-05, "epoch": 0.1390677311357198, "percentage": 7.94, "elapsed_time": "1:04:07", "remaining_time": "12:23:27", "throughput": 820.69, "total_tokens": 3157944}
{"current_steps": 275, "total_steps": 3400, "loss": 0.8985, "lr": 9.973948315499126e-05, "epoch": 0.141643059490085, "percentage": 8.09, "elapsed_time": "1:05:15", "remaining_time": "12:21:28", "throughput": 821.56, "total_tokens": 3216448}
{"current_steps": 280, "total_steps": 3400, "loss": 0.9114, "lr": 9.971410551976002e-05, "epoch": 0.14421838784445018, "percentage": 8.24, "elapsed_time": "1:06:21", "remaining_time": "12:19:27", "throughput": 822.49, "total_tokens": 3274928}
{"current_steps": 285, "total_steps": 3400, "loss": 0.8886, "lr": 9.968755213860094e-05, "epoch": 0.14679371619881534, "percentage": 8.38, "elapsed_time": "1:07:28", "remaining_time": "12:17:29", "throughput": 823.36, "total_tokens": 3333408}
{"current_steps": 290, "total_steps": 3400, "loss": 0.8929, "lr": 9.96598236395054e-05, "epoch": 0.14936904455318054, "percentage": 8.53, "elapsed_time": "1:08:35", "remaining_time": "12:15:35", "throughput": 824.18, "total_tokens": 3391896}
{"current_steps": 295, "total_steps": 3400, "loss": 0.9091, "lr": 9.96309206782565e-05, "epoch": 0.1519443729075457, "percentage": 8.68, "elapsed_time": "1:09:42", "remaining_time": "12:13:46", "throughput": 824.88, "total_tokens": 3450392}
{"current_steps": 300, "total_steps": 3400, "loss": 0.8893, "lr": 9.960084393841355e-05, "epoch": 0.1545197012619109, "percentage": 8.82, "elapsed_time": "1:10:49", "remaining_time": "12:11:54", "throughput": 825.67, "total_tokens": 3508888}
{"current_steps": 300, "total_steps": 3400, "eval_loss": 0.8908902406692505, "epoch": 0.1545197012619109, "percentage": 8.82, "elapsed_time": "1:11:06", "remaining_time": "12:14:49", "throughput": 822.39, "total_tokens": 3508888}
{"current_steps": 305, "total_steps": 3400, "loss": 0.9056, "lr": 9.956959413129585e-05, "epoch": 0.15709502961627608, "percentage": 8.97, "elapsed_time": "1:12:19", "remaining_time": "12:13:52", "throughput": 822.12, "total_tokens": 3567368}
{"current_steps": 310, "total_steps": 3400, "loss": 0.8982, "lr": 9.953717199596598e-05, "epoch": 0.15967035797064125, "percentage": 9.12, "elapsed_time": "1:13:26", "remaining_time": "12:12:05", "throughput": 822.79, "total_tokens": 3625848}
{"current_steps": 315, "total_steps": 3400, "loss": 0.8968, "lr": 9.95035782992122e-05, "epoch": 0.16224568632500644, "percentage": 9.26, "elapsed_time": "1:14:34", "remaining_time": "12:10:19", "throughput": 823.44, "total_tokens": 3684336}
{"current_steps": 320, "total_steps": 3400, "loss": 0.8975, "lr": 9.94688138355304e-05, "epoch": 0.16482101467937163, "percentage": 9.41, "elapsed_time": "1:15:42", "remaining_time": "12:08:37", "throughput": 824.02, "total_tokens": 3742800}
{"current_steps": 325, "total_steps": 3400, "loss": 0.9061, "lr": 9.943287942710527e-05, "epoch": 0.1673963430337368, "percentage": 9.56, "elapsed_time": "1:16:50", "remaining_time": "12:07:00", "throughput": 824.51, "total_tokens": 3801280}
{"current_steps": 330, "total_steps": 3400, "loss": 0.8948, "lr": 9.939577592379088e-05, "epoch": 0.16997167138810199, "percentage": 9.71, "elapsed_time": "1:17:58", "remaining_time": "12:05:19", "throughput": 825.09, "total_tokens": 3859792}
{"current_steps": 335, "total_steps": 3400, "loss": 0.9063, "lr": 9.935750420309055e-05, "epoch": 0.17254699974246718, "percentage": 9.85, "elapsed_time": "1:19:05", "remaining_time": "12:03:41", "throughput": 825.61, "total_tokens": 3918272}
{"current_steps": 340, "total_steps": 3400, "loss": 0.8952, "lr": 9.931806517013612e-05, "epoch": 0.17512232809683234, "percentage": 10.0, "elapsed_time": "1:20:14", "remaining_time": "12:02:06", "throughput": 826.07, "total_tokens": 3976760}
{"current_steps": 345, "total_steps": 3400, "loss": 0.9136, "lr": 9.927745975766654e-05, "epoch": 0.17769765645119753, "percentage": 10.15, "elapsed_time": "1:21:22", "remaining_time": "12:00:30", "throughput": 826.55, "total_tokens": 4035240}
{"current_steps": 350, "total_steps": 3400, "loss": 0.9075, "lr": 9.923568892600578e-05, "epoch": 0.1802729848055627, "percentage": 10.29, "elapsed_time": "1:22:30", "remaining_time": "11:59:02", "throughput": 826.88, "total_tokens": 4093688}
{"current_steps": 350, "total_steps": 3400, "eval_loss": 0.89204341173172, "epoch": 0.1802729848055627, "percentage": 10.29, "elapsed_time": "1:22:47", "remaining_time": "12:01:26", "throughput": 824.12, "total_tokens": 4093688}
{"current_steps": 355, "total_steps": 3400, "loss": 0.8812, "lr": 9.91927536630402e-05, "epoch": 0.1828483131599279, "percentage": 10.44, "elapsed_time": "1:24:00", "remaining_time": "12:00:35", "throughput": 823.75, "total_tokens": 4152160}
{"current_steps": 360, "total_steps": 3400, "loss": 0.9109, "lr": 9.91486549841951e-05, "epoch": 0.18542364151429308, "percentage": 10.59, "elapsed_time": "1:25:08", "remaining_time": "11:59:02", "throughput": 824.17, "total_tokens": 4210648}
{"current_steps": 365, "total_steps": 3400, "loss": 0.9176, "lr": 9.91033939324107e-05, "epoch": 0.18799896986865824, "percentage": 10.74, "elapsed_time": "1:26:17", "remaining_time": "11:57:30", "throughput": 824.56, "total_tokens": 4269136}
{"current_steps": 370, "total_steps": 3400, "loss": 0.9077, "lr": 9.905697157811761e-05, "epoch": 0.19057429822302344, "percentage": 10.88, "elapsed_time": "1:27:26", "remaining_time": "11:56:02", "throughput": 824.91, "total_tokens": 4327664}
{"current_steps": 375, "total_steps": 3400, "loss": 0.893, "lr": 9.900938901921131e-05, "epoch": 0.19314962657738863, "percentage": 11.03, "elapsed_time": "1:28:34", "remaining_time": "11:54:32", "throughput": 825.27, "total_tokens": 4386120}
{"current_steps": 380, "total_steps": 3400, "loss": 0.9094, "lr": 9.896064738102635e-05, "epoch": 0.1957249549317538, "percentage": 11.18, "elapsed_time": "1:29:43", "remaining_time": "11:53:03", "throughput": 825.62, "total_tokens": 4444560}
{"current_steps": 385, "total_steps": 3400, "loss": 0.9091, "lr": 9.891074781630966e-05, "epoch": 0.19830028328611898, "percentage": 11.32, "elapsed_time": "1:30:51", "remaining_time": "11:51:29", "throughput": 826.06, "total_tokens": 4503016}
{"current_steps": 390, "total_steps": 3400, "loss": 0.9033, "lr": 9.885969150519331e-05, "epoch": 0.20087561164048418, "percentage": 11.47, "elapsed_time": "1:31:58", "remaining_time": "11:49:54", "throughput": 826.52, "total_tokens": 4561496}
{"current_steps": 395, "total_steps": 3400, "loss": 0.8851, "lr": 9.88074796551666e-05, "epoch": 0.20345093999484934, "percentage": 11.62, "elapsed_time": "1:33:05", "remaining_time": "11:48:14", "throughput": 827.08, "total_tokens": 4619944}
{"current_steps": 400, "total_steps": 3400, "loss": 0.9004, "lr": 9.875411350104744e-05, "epoch": 0.20602626834921453, "percentage": 11.76, "elapsed_time": "1:34:12", "remaining_time": "11:46:36", "throughput": 827.62, "total_tokens": 4678384}
{"current_steps": 400, "total_steps": 3400, "eval_loss": 0.9086406826972961, "epoch": 0.20602626834921453, "percentage": 11.76, "elapsed_time": "1:34:29", "remaining_time": "11:48:42", "throughput": 825.17, "total_tokens": 4678384}
{"current_steps": 405, "total_steps": 3400, "loss": 0.8976, "lr": 9.86995943049533e-05, "epoch": 0.2086015967035797, "percentage": 11.91, "elapsed_time": "1:35:41", "remaining_time": "11:47:40", "throughput": 824.99, "total_tokens": 4736904}
{"current_steps": 410, "total_steps": 3400, "loss": 0.9134, "lr": 9.864392335627117e-05, "epoch": 0.2111769250579449, "percentage": 12.06, "elapsed_time": "1:36:49", "remaining_time": "11:46:04", "throughput": 825.49, "total_tokens": 4795376}
{"current_steps": 415, "total_steps": 3400, "loss": 0.8955, "lr": 9.858710197162721e-05, "epoch": 0.21375225341231008, "percentage": 12.21, "elapsed_time": "1:37:56", "remaining_time": "11:44:26", "throughput": 826.02, "total_tokens": 4853880}
{"current_steps": 420, "total_steps": 3400, "loss": 0.9014, "lr": 9.852913149485556e-05, "epoch": 0.21632758176667524, "percentage": 12.35, "elapsed_time": "1:39:02", "remaining_time": "11:42:46", "throughput": 826.59, "total_tokens": 4912360}
{"current_steps": 425, "total_steps": 3400, "loss": 0.9065, "lr": 9.847001329696653e-05, "epoch": 0.21890291012104043, "percentage": 12.5, "elapsed_time": "1:40:09", "remaining_time": "11:41:08", "throughput": 827.12, "total_tokens": 4970872}
{"current_steps": 430, "total_steps": 3400, "loss": 0.8952, "lr": 9.840974877611422e-05, "epoch": 0.22147823847540563, "percentage": 12.65, "elapsed_time": "1:41:17", "remaining_time": "11:39:33", "throughput": 827.59, "total_tokens": 5029304}
{"current_steps": 435, "total_steps": 3400, "loss": 0.9106, "lr": 9.834833935756344e-05, "epoch": 0.2240535668297708, "percentage": 12.79, "elapsed_time": "1:42:25", "remaining_time": "11:38:11", "throughput": 827.83, "total_tokens": 5087800}
{"current_steps": 440, "total_steps": 3400, "loss": 0.8996, "lr": 9.828578649365601e-05, "epoch": 0.22662889518413598, "percentage": 12.94, "elapsed_time": "1:43:34", "remaining_time": "11:36:48", "throughput": 828.08, "total_tokens": 5146312}
{"current_steps": 445, "total_steps": 3400, "loss": 0.8999, "lr": 9.822209166377635e-05, "epoch": 0.22920422353850115, "percentage": 13.09, "elapsed_time": "1:44:43", "remaining_time": "11:35:24", "throughput": 828.35, "total_tokens": 5204800}
{"current_steps": 450, "total_steps": 3400, "loss": 0.9076, "lr": 9.815725637431662e-05, "epoch": 0.23177955189286634, "percentage": 13.24, "elapsed_time": "1:45:53", "remaining_time": "11:34:08", "throughput": 828.45, "total_tokens": 5263304}
{"current_steps": 450, "total_steps": 3400, "eval_loss": 0.8962157368659973, "epoch": 0.23177955189286634, "percentage": 13.24, "elapsed_time": "1:46:10", "remaining_time": "11:36:01", "throughput": 826.21, "total_tokens": 5263304}
{"current_steps": 455, "total_steps": 3400, "loss": 0.8942, "lr": 9.809128215864097e-05, "epoch": 0.23435488024723153, "percentage": 13.38, "elapsed_time": "1:47:26", "remaining_time": "11:35:24", "throughput": 825.54, "total_tokens": 5321760}
{"current_steps": 460, "total_steps": 3400, "loss": 0.9099, "lr": 9.802417057704931e-05, "epoch": 0.2369302086015967, "percentage": 13.53, "elapsed_time": "1:48:35", "remaining_time": "11:34:01", "throughput": 825.79, "total_tokens": 5380224}
{"current_steps": 465, "total_steps": 3400, "loss": 0.8981, "lr": 9.795592321674045e-05, "epoch": 0.23950553695596188, "percentage": 13.68, "elapsed_time": "1:49:44", "remaining_time": "11:32:37", "throughput": 826.03, "total_tokens": 5438704}
{"current_steps": 470, "total_steps": 3400, "loss": 0.8952, "lr": 9.788654169177453e-05, "epoch": 0.24208086531032708, "percentage": 13.82, "elapsed_time": "1:50:53", "remaining_time": "11:31:17", "throughput": 826.23, "total_tokens": 5497208}
{"current_steps": 475, "total_steps": 3400, "loss": 0.8959, "lr": 9.781602764303487e-05, "epoch": 0.24465619366469224, "percentage": 13.97, "elapsed_time": "1:52:01", "remaining_time": "11:29:52", "throughput": 826.52, "total_tokens": 5555704}
{"current_steps": 480, "total_steps": 3400, "loss": 0.901, "lr": 9.774438273818911e-05, "epoch": 0.24723152201905743, "percentage": 14.12, "elapsed_time": "1:53:09", "remaining_time": "11:28:25", "throughput": 826.83, "total_tokens": 5614160}
{"current_steps": 485, "total_steps": 3400, "loss": 0.9008, "lr": 9.767160867164979e-05, "epoch": 0.24980685037342262, "percentage": 14.26, "elapsed_time": "1:54:18", "remaining_time": "11:27:00", "throughput": 827.13, "total_tokens": 5672640}
{"current_steps": 490, "total_steps": 3400, "loss": 0.9016, "lr": 9.759770716453436e-05, "epoch": 0.2523821787277878, "percentage": 14.41, "elapsed_time": "1:55:26", "remaining_time": "11:25:37", "throughput": 827.37, "total_tokens": 5731072}
{"current_steps": 495, "total_steps": 3400, "loss": 0.9132, "lr": 9.752267996462434e-05, "epoch": 0.254957507082153, "percentage": 14.56, "elapsed_time": "1:56:35", "remaining_time": "11:24:12", "throughput": 827.64, "total_tokens": 5789544}
{"current_steps": 500, "total_steps": 3400, "loss": 0.8962, "lr": 9.744652884632406e-05, "epoch": 0.25753283543651817, "percentage": 14.71, "elapsed_time": "1:57:43", "remaining_time": "11:22:47", "throughput": 827.95, "total_tokens": 5848048}
{"current_steps": 500, "total_steps": 3400, "eval_loss": 0.8987945914268494, "epoch": 0.25753283543651817, "percentage": 14.71, "elapsed_time": "1:58:00", "remaining_time": "11:24:26", "throughput": 825.94, "total_tokens": 5848048}
{"current_steps": 505, "total_steps": 3400, "loss": 0.8954, "lr": 9.736925561061871e-05, "epoch": 0.26010816379088336, "percentage": 14.85, "elapsed_time": "1:59:12", "remaining_time": "11:23:24", "throughput": 825.76, "total_tokens": 5906512}
{"current_steps": 510, "total_steps": 3400, "loss": 0.8927, "lr": 9.729086208503174e-05, "epoch": 0.2626834921452485, "percentage": 15.0, "elapsed_time": "2:00:20", "remaining_time": "11:21:55", "throughput": 826.14, "total_tokens": 5965024}
{"current_steps": 515, "total_steps": 3400, "loss": 0.898, "lr": 9.721135012358156e-05, "epoch": 0.2652588204996137, "percentage": 15.15, "elapsed_time": "2:01:28", "remaining_time": "11:20:27", "throughput": 826.48, "total_tokens": 6023496}
{"current_steps": 520, "total_steps": 3400, "loss": 0.9016, "lr": 9.713072160673777e-05, "epoch": 0.2678341488539789, "percentage": 15.29, "elapsed_time": "2:02:35", "remaining_time": "11:19:00", "throughput": 826.81, "total_tokens": 6082000}
{"current_steps": 525, "total_steps": 3400, "loss": 0.8842, "lr": 9.704897844137673e-05, "epoch": 0.2704094772083441, "percentage": 15.44, "elapsed_time": "2:03:43", "remaining_time": "11:17:34", "throughput": 827.13, "total_tokens": 6140480}
{"current_steps": 530, "total_steps": 3400, "loss": 0.8921, "lr": 9.696612256073633e-05, "epoch": 0.27298480556270927, "percentage": 15.59, "elapsed_time": "2:04:51", "remaining_time": "11:16:05", "throughput": 827.5, "total_tokens": 6198968}
{"current_steps": 535, "total_steps": 3400, "loss": 0.8979, "lr": 9.688215592437039e-05, "epoch": 0.2755601339170744, "percentage": 15.74, "elapsed_time": "2:05:58", "remaining_time": "11:14:37", "throughput": 827.85, "total_tokens": 6257464}
{"current_steps": 540, "total_steps": 3400, "loss": 0.8951, "lr": 9.679708051810221e-05, "epoch": 0.2781354622714396, "percentage": 15.88, "elapsed_time": "2:07:06", "remaining_time": "11:13:12", "throughput": 828.16, "total_tokens": 6315944}
{"current_steps": 545, "total_steps": 3400, "loss": 0.9149, "lr": 9.67108983539777e-05, "epoch": 0.2807107906258048, "percentage": 16.03, "elapsed_time": "2:08:15", "remaining_time": "11:11:51", "throughput": 828.37, "total_tokens": 6374408}
{"current_steps": 550, "total_steps": 3400, "loss": 0.9013, "lr": 9.662361147021779e-05, "epoch": 0.28328611898017, "percentage": 16.18, "elapsed_time": "2:09:23", "remaining_time": "11:10:29", "throughput": 828.6, "total_tokens": 6432936}
{"current_steps": 550, "total_steps": 3400, "eval_loss": 0.9001271724700928, "epoch": 0.28328611898017, "percentage": 16.18, "elapsed_time": "2:09:40", "remaining_time": "11:11:57", "throughput": 826.79, "total_tokens": 6432936}
{"current_steps": 555, "total_steps": 3400, "loss": 0.8981, "lr": 9.653522193117013e-05, "epoch": 0.28586144733453517, "percentage": 16.32, "elapsed_time": "2:10:53", "remaining_time": "11:10:57", "throughput": 826.56, "total_tokens": 6491400}
{"current_steps": 560, "total_steps": 3400, "loss": 0.9041, "lr": 9.644573182726035e-05, "epoch": 0.28843677568890036, "percentage": 16.47, "elapsed_time": "2:12:01", "remaining_time": "11:09:31", "throughput": 826.89, "total_tokens": 6549872}
{"current_steps": 565, "total_steps": 3400, "loss": 0.9024, "lr": 9.63551432749426e-05, "epoch": 0.2910121040432655, "percentage": 16.62, "elapsed_time": "2:13:08", "remaining_time": "11:08:04", "throughput": 827.21, "total_tokens": 6608296}
{"current_steps": 570, "total_steps": 3400, "loss": 0.9002, "lr": 9.626345841664953e-05, "epoch": 0.2935874323976307, "percentage": 16.76, "elapsed_time": "2:14:16", "remaining_time": "11:06:41", "throughput": 827.46, "total_tokens": 6666768}
{"current_steps": 575, "total_steps": 3400, "loss": 0.9035, "lr": 9.617067942074153e-05, "epoch": 0.2961627607519959, "percentage": 16.91, "elapsed_time": "2:15:24", "remaining_time": "11:05:16", "throughput": 827.76, "total_tokens": 6725248}
{"current_steps": 580, "total_steps": 3400, "loss": 0.9019, "lr": 9.607680848145558e-05, "epoch": 0.29873808910636107, "percentage": 17.06, "elapsed_time": "2:16:32", "remaining_time": "11:03:54", "throughput": 827.99, "total_tokens": 6783680}
{"current_steps": 585, "total_steps": 3400, "loss": 0.9001, "lr": 9.598184781885318e-05, "epoch": 0.30131341746072626, "percentage": 17.21, "elapsed_time": "2:17:40", "remaining_time": "11:02:30", "throughput": 828.27, "total_tokens": 6842144}
{"current_steps": 590, "total_steps": 3400, "loss": 0.8961, "lr": 9.588579967876806e-05, "epoch": 0.3038887458150914, "percentage": 17.35, "elapsed_time": "2:18:48", "remaining_time": "11:01:05", "throughput": 828.58, "total_tokens": 6900656}
{"current_steps": 595, "total_steps": 3400, "loss": 0.9, "lr": 9.578866633275288e-05, "epoch": 0.3064640741694566, "percentage": 17.5, "elapsed_time": "2:19:56", "remaining_time": "10:59:41", "throughput": 828.86, "total_tokens": 6959128}
{"current_steps": 600, "total_steps": 3400, "loss": 0.9046, "lr": 9.569045007802559e-05, "epoch": 0.3090394025238218, "percentage": 17.65, "elapsed_time": "2:21:03", "remaining_time": "10:58:14", "throughput": 829.2, "total_tokens": 7017576}
{"current_steps": 600, "total_steps": 3400, "eval_loss": 0.9053278565406799, "epoch": 0.3090394025238218, "percentage": 17.65, "elapsed_time": "2:21:20", "remaining_time": "10:59:34", "throughput": 827.52, "total_tokens": 7017576}
{"current_steps": 605, "total_steps": 3400, "loss": 0.9019, "lr": 9.55911532374151e-05, "epoch": 0.311614730878187, "percentage": 17.79, "elapsed_time": "2:22:31", "remaining_time": "10:58:28", "throughput": 827.42, "total_tokens": 7076032}
{"current_steps": 610, "total_steps": 3400, "loss": 0.8956, "lr": 9.549077815930636e-05, "epoch": 0.31419005923255217, "percentage": 17.94, "elapsed_time": "2:23:39", "remaining_time": "10:57:03", "throughput": 827.73, "total_tokens": 7134536}
{"current_steps": 615, "total_steps": 3400, "loss": 0.898, "lr": 9.538932721758474e-05, "epoch": 0.31676538758691736, "percentage": 18.09, "elapsed_time": "2:24:47", "remaining_time": "10:55:38", "throughput": 828.02, "total_tokens": 7193032}
{"current_steps": 620, "total_steps": 3400, "loss": 0.8991, "lr": 9.528680281157999e-05, "epoch": 0.3193407159412825, "percentage": 18.24, "elapsed_time": "2:25:54", "remaining_time": "10:54:14", "throughput": 828.31, "total_tokens": 7251568}
{"current_steps": 625, "total_steps": 3400, "loss": 0.8961, "lr": 9.518320736600943e-05, "epoch": 0.3219160442956477, "percentage": 18.38, "elapsed_time": "2:27:02", "remaining_time": "10:52:50", "throughput": 828.6, "total_tokens": 7310072}
{"current_steps": 630, "total_steps": 3400, "loss": 0.8994, "lr": 9.507854333092063e-05, "epoch": 0.3244913726500129, "percentage": 18.53, "elapsed_time": "2:28:09", "remaining_time": "10:51:27", "throughput": 828.86, "total_tokens": 7368560}
{"current_steps": 635, "total_steps": 3400, "loss": 0.8925, "lr": 9.497281318163346e-05, "epoch": 0.32706670100437807, "percentage": 18.68, "elapsed_time": "2:29:18", "remaining_time": "10:50:06", "throughput": 829.09, "total_tokens": 7427040}
{"current_steps": 640, "total_steps": 3400, "loss": 0.9087, "lr": 9.486601941868154e-05, "epoch": 0.32964202935874326, "percentage": 18.82, "elapsed_time": "2:30:24", "remaining_time": "10:48:39", "throughput": 829.44, "total_tokens": 7485552}
{"current_steps": 645, "total_steps": 3400, "loss": 0.8924, "lr": 9.475816456775313e-05, "epoch": 0.3322173577131084, "percentage": 18.97, "elapsed_time": "2:31:29", "remaining_time": "10:47:03", "throughput": 829.98, "total_tokens": 7544040}
{"current_steps": 650, "total_steps": 3400, "loss": 0.904, "lr": 9.464925117963133e-05, "epoch": 0.3347926860674736, "percentage": 19.12, "elapsed_time": "2:32:34", "remaining_time": "10:45:29", "throughput": 830.49, "total_tokens": 7602512}
{"current_steps": 650, "total_steps": 3400, "eval_loss": 0.90328449010849, "epoch": 0.3347926860674736, "percentage": 19.12, "elapsed_time": "2:32:50", "remaining_time": "10:46:37", "throughput": 829.03, "total_tokens": 7602512}
{"current_steps": 655, "total_steps": 3400, "loss": 0.8929, "lr": 9.453928183013385e-05, "epoch": 0.3373680144218388, "percentage": 19.26, "elapsed_time": "2:33:59", "remaining_time": "10:45:22", "throughput": 829.14, "total_tokens": 7660968}
{"current_steps": 660, "total_steps": 3400, "loss": 0.9078, "lr": 9.442825912005202e-05, "epoch": 0.33994334277620397, "percentage": 19.41, "elapsed_time": "2:35:03", "remaining_time": "10:43:45", "throughput": 829.7, "total_tokens": 7719448}
{"current_steps": 665, "total_steps": 3400, "loss": 0.8963, "lr": 9.431618567508933e-05, "epoch": 0.34251867113056916, "percentage": 19.56, "elapsed_time": "2:36:07", "remaining_time": "10:42:07", "throughput": 830.29, "total_tokens": 7777928}
{"current_steps": 670, "total_steps": 3400, "loss": 0.9134, "lr": 9.420306414579925e-05, "epoch": 0.34509399948493436, "percentage": 19.71, "elapsed_time": "2:37:11", "remaining_time": "10:40:30", "throughput": 830.86, "total_tokens": 7836424}
{"current_steps": 675, "total_steps": 3400, "loss": 0.8984, "lr": 9.408889720752266e-05, "epoch": 0.3476693278392995, "percentage": 19.85, "elapsed_time": "2:38:16", "remaining_time": "10:38:56", "throughput": 831.38, "total_tokens": 7894904}
{"current_steps": 680, "total_steps": 3400, "loss": 0.8997, "lr": 9.397368756032445e-05, "epoch": 0.3502446561936647, "percentage": 20.0, "elapsed_time": "2:39:20", "remaining_time": "10:37:20", "throughput": 831.93, "total_tokens": 7953432}
{"current_steps": 685, "total_steps": 3400, "loss": 0.8926, "lr": 9.385743792892982e-05, "epoch": 0.3528199845480299, "percentage": 20.15, "elapsed_time": "2:40:24", "remaining_time": "10:35:46", "throughput": 832.45, "total_tokens": 8011888}
{"current_steps": 690, "total_steps": 3400, "loss": 0.9008, "lr": 9.374015106265968e-05, "epoch": 0.35539531290239507, "percentage": 20.29, "elapsed_time": "2:41:28", "remaining_time": "10:34:11", "throughput": 832.99, "total_tokens": 8070344}
{"current_steps": 695, "total_steps": 3400, "loss": 0.8986, "lr": 9.362182973536569e-05, "epoch": 0.35797064125676026, "percentage": 20.44, "elapsed_time": "2:42:32", "remaining_time": "10:32:38", "throughput": 833.49, "total_tokens": 8128816}
{"current_steps": 700, "total_steps": 3400, "loss": 0.8972, "lr": 9.35024767453647e-05, "epoch": 0.3605459696111254, "percentage": 20.59, "elapsed_time": "2:43:36", "remaining_time": "10:31:04", "throughput": 834.03, "total_tokens": 8187320}
{"current_steps": 700, "total_steps": 3400, "eval_loss": 0.9028835892677307, "epoch": 0.3605459696111254, "percentage": 20.59, "elapsed_time": "2:43:52", "remaining_time": "10:32:06", "throughput": 832.66, "total_tokens": 8187320}
{"current_steps": 705, "total_steps": 3400, "loss": 0.8998, "lr": 9.338209491537257e-05, "epoch": 0.3631212979654906, "percentage": 20.74, "elapsed_time": "2:45:01", "remaining_time": "10:30:51", "throughput": 832.76, "total_tokens": 8245776}
{"current_steps": 710, "total_steps": 3400, "loss": 0.8999, "lr": 9.326068709243727e-05, "epoch": 0.3656966263198558, "percentage": 20.88, "elapsed_time": "2:46:05", "remaining_time": "10:29:16", "throughput": 833.31, "total_tokens": 8304280}
{"current_steps": 715, "total_steps": 3400, "loss": 0.8983, "lr": 9.313825614787177e-05, "epoch": 0.36827195467422097, "percentage": 21.03, "elapsed_time": "2:47:09", "remaining_time": "10:27:41", "throughput": 833.84, "total_tokens": 8362728}
{"current_steps": 720, "total_steps": 3400, "loss": 0.892, "lr": 9.301480497718593e-05, "epoch": 0.37084728302858616, "percentage": 21.18, "elapsed_time": "2:48:12", "remaining_time": "10:26:08", "throughput": 834.37, "total_tokens": 8421224}
{"current_steps": 725, "total_steps": 3400, "loss": 0.9034, "lr": 9.289033650001817e-05, "epoch": 0.37342261138295135, "percentage": 21.32, "elapsed_time": "2:49:16", "remaining_time": "10:24:35", "throughput": 834.87, "total_tokens": 8479720}
{"current_steps": 730, "total_steps": 3400, "loss": 0.895, "lr": 9.276485366006634e-05, "epoch": 0.3759979397373165, "percentage": 21.47, "elapsed_time": "2:50:20", "remaining_time": "10:23:02", "throughput": 835.38, "total_tokens": 8538192}
{"current_steps": 735, "total_steps": 3400, "loss": 0.8973, "lr": 9.263835942501807e-05, "epoch": 0.3785732680916817, "percentage": 21.62, "elapsed_time": "2:51:24", "remaining_time": "10:21:29", "throughput": 835.9, "total_tokens": 8596664}
{"current_steps": 740, "total_steps": 3400, "loss": 0.8972, "lr": 9.251085678648072e-05, "epoch": 0.3811485964460469, "percentage": 21.76, "elapsed_time": "2:52:28", "remaining_time": "10:19:58", "throughput": 836.38, "total_tokens": 8655128}
{"current_steps": 745, "total_steps": 3400, "loss": 0.8987, "lr": 9.238234875991046e-05, "epoch": 0.38372392480041206, "percentage": 21.91, "elapsed_time": "2:53:31", "remaining_time": "10:18:25", "throughput": 836.89, "total_tokens": 8713624}
{"current_steps": 750, "total_steps": 3400, "loss": 0.9005, "lr": 9.225283838454111e-05, "epoch": 0.38629925315477726, "percentage": 22.06, "elapsed_time": "2:54:36", "remaining_time": "10:16:57", "throughput": 837.3, "total_tokens": 8772104}
{"current_steps": 750, "total_steps": 3400, "eval_loss": 0.8981761336326599, "epoch": 0.38629925315477726, "percentage": 22.06, "elapsed_time": "2:54:52", "remaining_time": "10:17:54", "throughput": 836.02, "total_tokens": 8772104}
{"current_steps": 755, "total_steps": 3400, "loss": 0.8973, "lr": 9.21223287233121e-05, "epoch": 0.3888745815091424, "percentage": 22.21, "elapsed_time": "2:56:01", "remaining_time": "10:16:40", "throughput": 836.09, "total_tokens": 8830568}
{"current_steps": 760, "total_steps": 3400, "loss": 0.8974, "lr": 9.199082286279622e-05, "epoch": 0.3914499098635076, "percentage": 22.35, "elapsed_time": "2:57:05", "remaining_time": "10:15:10", "throughput": 836.56, "total_tokens": 8889072}
{"current_steps": 765, "total_steps": 3400, "loss": 0.8985, "lr": 9.185832391312644e-05, "epoch": 0.3940252382178728, "percentage": 22.5, "elapsed_time": "2:58:09", "remaining_time": "10:13:39", "throughput": 837.04, "total_tokens": 8947568}
{"current_steps": 770, "total_steps": 3400, "loss": 0.8935, "lr": 9.172483500792244e-05, "epoch": 0.39660056657223797, "percentage": 22.65, "elapsed_time": "2:59:13", "remaining_time": "10:12:08", "throughput": 837.52, "total_tokens": 9006056}
{"current_steps": 775, "total_steps": 3400, "loss": 0.8985, "lr": 9.159035930421658e-05, "epoch": 0.39917589492660316, "percentage": 22.79, "elapsed_time": "3:00:17", "remaining_time": "10:10:39", "throughput": 837.96, "total_tokens": 9064592}
{"current_steps": 780, "total_steps": 3400, "loss": 0.9105, "lr": 9.145489998237902e-05, "epoch": 0.40175122328096835, "percentage": 22.94, "elapsed_time": "3:01:21", "remaining_time": "10:09:10", "throughput": 838.42, "total_tokens": 9123096}
{"current_steps": 785, "total_steps": 3400, "loss": 0.8925, "lr": 9.131846024604274e-05, "epoch": 0.4043265516353335, "percentage": 23.09, "elapsed_time": "3:02:25", "remaining_time": "10:07:41", "throughput": 838.85, "total_tokens": 9181576}
{"current_steps": 790, "total_steps": 3400, "loss": 0.8955, "lr": 9.11810433220276e-05, "epoch": 0.4069018799896987, "percentage": 23.24, "elapsed_time": "3:03:29", "remaining_time": "10:06:12", "throughput": 839.29, "total_tokens": 9240048}
{"current_steps": 795, "total_steps": 3400, "loss": 0.8986, "lr": 9.104265246026415e-05, "epoch": 0.40947720834406387, "percentage": 23.38, "elapsed_time": "3:04:33", "remaining_time": "10:04:45", "throughput": 839.7, "total_tokens": 9298528}
{"current_steps": 800, "total_steps": 3400, "loss": 0.8881, "lr": 9.090329093371666e-05, "epoch": 0.41205253669842906, "percentage": 23.53, "elapsed_time": "3:05:37", "remaining_time": "10:03:16", "throughput": 840.14, "total_tokens": 9357016}
{"current_steps": 800, "total_steps": 3400, "eval_loss": 0.8973079919815063, "epoch": 0.41205253669842906, "percentage": 23.53, "elapsed_time": "3:05:53", "remaining_time": "10:04:09", "throughput": 838.92, "total_tokens": 9357016}
{"current_steps": 805, "total_steps": 3400, "loss": 0.8798, "lr": 9.076296203830579e-05, "epoch": 0.41462786505279425, "percentage": 23.68, "elapsed_time": "3:07:02", "remaining_time": "10:02:55", "throughput": 839.0, "total_tokens": 9415480}
{"current_steps": 810, "total_steps": 3400, "loss": 0.9104, "lr": 9.062166909283062e-05, "epoch": 0.4172031934071594, "percentage": 23.82, "elapsed_time": "3:08:05", "remaining_time": "10:01:26", "throughput": 839.45, "total_tokens": 9473928}
{"current_steps": 815, "total_steps": 3400, "loss": 0.9007, "lr": 9.047941543889014e-05, "epoch": 0.4197785217615246, "percentage": 23.97, "elapsed_time": "3:09:09", "remaining_time": "9:59:58", "throughput": 839.88, "total_tokens": 9532408}
{"current_steps": 820, "total_steps": 3400, "loss": 0.8974, "lr": 9.033620444080428e-05, "epoch": 0.4223538501158898, "percentage": 24.12, "elapsed_time": "3:10:13", "remaining_time": "9:58:30", "throughput": 840.31, "total_tokens": 9590920}
{"current_steps": 825, "total_steps": 3400, "loss": 0.8992, "lr": 9.019203948553422e-05, "epoch": 0.42492917847025496, "percentage": 24.26, "elapsed_time": "3:11:17", "remaining_time": "9:57:02", "throughput": 840.75, "total_tokens": 9649400}
{"current_steps": 830, "total_steps": 3400, "loss": 0.8991, "lr": 9.004692398260244e-05, "epoch": 0.42750450682462016, "percentage": 24.41, "elapsed_time": "3:12:20", "remaining_time": "9:55:35", "throughput": 841.17, "total_tokens": 9707888}
{"current_steps": 835, "total_steps": 3400, "loss": 0.8964, "lr": 8.9900861364012e-05, "epoch": 0.43007983517898535, "percentage": 24.56, "elapsed_time": "3:13:24", "remaining_time": "9:54:07", "throughput": 841.59, "total_tokens": 9766384}
{"current_steps": 840, "total_steps": 3400, "loss": 0.8723, "lr": 8.975385508416532e-05, "epoch": 0.4326551635333505, "percentage": 24.71, "elapsed_time": "3:14:28", "remaining_time": "9:52:41", "throughput": 841.99, "total_tokens": 9824896}
{"current_steps": 845, "total_steps": 3400, "loss": 0.874, "lr": 8.960590861978265e-05, "epoch": 0.4352304918877157, "percentage": 24.85, "elapsed_time": "3:15:32", "remaining_time": "9:51:15", "throughput": 842.39, "total_tokens": 9883408}
{"current_steps": 850, "total_steps": 3400, "loss": 0.9035, "lr": 8.945702546981969e-05, "epoch": 0.43780582024208087, "percentage": 25.0, "elapsed_time": "3:16:36", "remaining_time": "9:49:49", "throughput": 842.79, "total_tokens": 9941896}
{"current_steps": 850, "total_steps": 3400, "eval_loss": 0.8779178261756897, "epoch": 0.43780582024208087, "percentage": 25.0, "elapsed_time": "3:16:52", "remaining_time": "9:50:37", "throughput": 841.64, "total_tokens": 9941896}
{"current_steps": 855, "total_steps": 3400, "loss": 0.8516, "lr": 8.930720915538487e-05, "epoch": 0.44038114859644606, "percentage": 25.15, "elapsed_time": "3:18:00", "remaining_time": "9:49:23", "throughput": 841.75, "total_tokens": 10000336}
{"current_steps": 860, "total_steps": 3400, "loss": 0.9206, "lr": 8.915646321965614e-05, "epoch": 0.44295647695081125, "percentage": 25.29, "elapsed_time": "3:19:03", "remaining_time": "9:47:54", "throughput": 842.2, "total_tokens": 10058816}
{"current_steps": 865, "total_steps": 3400, "loss": 0.9028, "lr": 8.900479122779712e-05, "epoch": 0.4455318053051764, "percentage": 25.44, "elapsed_time": "3:20:06", "remaining_time": "9:46:26", "throughput": 842.66, "total_tokens": 10117320}
{"current_steps": 870, "total_steps": 3400, "loss": 0.8991, "lr": 8.885219676687277e-05, "epoch": 0.4481071336595416, "percentage": 25.59, "elapsed_time": "3:21:09", "remaining_time": "9:45:00", "throughput": 843.07, "total_tokens": 10175824}
{"current_steps": 875, "total_steps": 3400, "loss": 0.8934, "lr": 8.869868344576459e-05, "epoch": 0.45068246201390677, "percentage": 25.74, "elapsed_time": "3:22:12", "remaining_time": "9:43:32", "throughput": 843.51, "total_tokens": 10234288}
{"current_steps": 880, "total_steps": 3400, "loss": 0.8908, "lr": 8.854425489508532e-05, "epoch": 0.45325779036827196, "percentage": 25.88, "elapsed_time": "3:23:15", "remaining_time": "9:42:04", "throughput": 843.95, "total_tokens": 10292736}
{"current_steps": 885, "total_steps": 3400, "loss": 0.8988, "lr": 8.838891476709288e-05, "epoch": 0.45583311872263715, "percentage": 26.03, "elapsed_time": "3:24:18", "remaining_time": "9:40:36", "throughput": 844.4, "total_tokens": 10351224}
{"current_steps": 890, "total_steps": 3400, "loss": 0.8965, "lr": 8.823266673560426e-05, "epoch": 0.4584084470770023, "percentage": 26.18, "elapsed_time": "3:25:21", "remaining_time": "9:39:10", "throughput": 844.81, "total_tokens": 10409736}
{"current_steps": 895, "total_steps": 3400, "loss": 0.8989, "lr": 8.807551449590846e-05, "epoch": 0.4609837754313675, "percentage": 26.32, "elapsed_time": "3:26:25", "remaining_time": "9:37:44", "throughput": 845.22, "total_tokens": 10468240}
{"current_steps": 900, "total_steps": 3400, "loss": 0.8961, "lr": 8.791746176467907e-05, "epoch": 0.4635591037857327, "percentage": 26.47, "elapsed_time": "3:27:28", "remaining_time": "9:36:18", "throughput": 845.63, "total_tokens": 10526712}
{"current_steps": 900, "total_steps": 3400, "eval_loss": 0.891426146030426, "epoch": 0.4635591037857327, "percentage": 26.47, "elapsed_time": "3:27:44", "remaining_time": "9:37:03", "throughput": 844.54, "total_tokens": 10526712}
{"current_steps": 905, "total_steps": 3400, "loss": 0.8955, "lr": 8.775851227988656e-05, "epoch": 0.46613443214009787, "percentage": 26.62, "elapsed_time": "3:28:52", "remaining_time": "9:35:50", "throughput": 844.63, "total_tokens": 10585232}
{"current_steps": 910, "total_steps": 3400, "loss": 0.8951, "lr": 8.759866980070963e-05, "epoch": 0.46870976049446306, "percentage": 26.76, "elapsed_time": "3:29:55", "remaining_time": "9:34:23", "throughput": 845.06, "total_tokens": 10643728}
{"current_steps": 915, "total_steps": 3400, "loss": 0.8951, "lr": 8.743793810744654e-05, "epoch": 0.47128508884882825, "percentage": 26.91, "elapsed_time": "3:30:58", "remaining_time": "9:32:58", "throughput": 845.46, "total_tokens": 10702240}
{"current_steps": 920, "total_steps": 3400, "loss": 0.9066, "lr": 8.727632100142551e-05, "epoch": 0.4738604172031934, "percentage": 27.06, "elapsed_time": "3:32:01", "remaining_time": "9:31:32", "throughput": 845.87, "total_tokens": 10760656}
{"current_steps": 925, "total_steps": 3400, "loss": 0.8953, "lr": 8.711382230491493e-05, "epoch": 0.4764357455575586, "percentage": 27.21, "elapsed_time": "3:33:04", "remaining_time": "9:30:06", "throughput": 846.28, "total_tokens": 10819128}
{"current_steps": 930, "total_steps": 3400, "loss": 0.8961, "lr": 8.695044586103296e-05, "epoch": 0.47901107391192377, "percentage": 27.35, "elapsed_time": "3:34:06", "remaining_time": "9:28:40", "throughput": 846.71, "total_tokens": 10877600}
{"current_steps": 935, "total_steps": 3400, "loss": 0.8965, "lr": 8.678619553365659e-05, "epoch": 0.48158640226628896, "percentage": 27.5, "elapsed_time": "3:35:10", "remaining_time": "9:27:15", "throughput": 847.09, "total_tokens": 10936088}
{"current_steps": 940, "total_steps": 3400, "loss": 0.9018, "lr": 8.662107520733027e-05, "epoch": 0.48416173062065415, "percentage": 27.65, "elapsed_time": "3:36:12", "remaining_time": "9:25:50", "throughput": 847.5, "total_tokens": 10994560}
{"current_steps": 945, "total_steps": 3400, "loss": 0.8944, "lr": 8.64550887871741e-05, "epoch": 0.4867370589750193, "percentage": 27.79, "elapsed_time": "3:37:16", "remaining_time": "9:24:26", "throughput": 847.88, "total_tokens": 11053016}
{"current_steps": 950, "total_steps": 3400, "loss": 0.8852, "lr": 8.628824019879137e-05, "epoch": 0.4893123873293845, "percentage": 27.94, "elapsed_time": "3:38:18", "remaining_time": "9:23:01", "throughput": 848.29, "total_tokens": 11111520}
{"current_steps": 950, "total_steps": 3400, "eval_loss": 0.8915690183639526, "epoch": 0.4893123873293845, "percentage": 27.94, "elapsed_time": "3:38:35", "remaining_time": "9:23:43", "throughput": 847.23, "total_tokens": 11111520}
{"current_steps": 955, "total_steps": 3400, "loss": 0.9087, "lr": 8.612053338817581e-05, "epoch": 0.49188771568374967, "percentage": 28.09, "elapsed_time": "3:39:44", "remaining_time": "9:22:36", "throughput": 847.18, "total_tokens": 11170016}
{"current_steps": 960, "total_steps": 3400, "loss": 0.8915, "lr": 8.595197232161824e-05, "epoch": 0.49446304403811486, "percentage": 28.24, "elapsed_time": "3:40:48", "remaining_time": "9:21:13", "throughput": 847.52, "total_tokens": 11228496}
{"current_steps": 965, "total_steps": 3400, "loss": 0.8836, "lr": 8.578256098561275e-05, "epoch": 0.49703837239248005, "percentage": 28.38, "elapsed_time": "3:41:51", "remaining_time": "9:19:49", "throughput": 847.89, "total_tokens": 11286928}
{"current_steps": 970, "total_steps": 3400, "loss": 0.9116, "lr": 8.561230338676239e-05, "epoch": 0.49961370074684525, "percentage": 28.53, "elapsed_time": "3:42:55", "remaining_time": "9:18:27", "throughput": 848.23, "total_tokens": 11345400}
{"current_steps": 975, "total_steps": 3400, "loss": 0.8809, "lr": 8.544120355168451e-05, "epoch": 0.5021890291012104, "percentage": 28.68, "elapsed_time": "3:43:58", "remaining_time": "9:17:03", "throughput": 848.61, "total_tokens": 11403912}
{"current_steps": 980, "total_steps": 3400, "loss": 0.8895, "lr": 8.526926552691544e-05, "epoch": 0.5047643574555756, "percentage": 28.82, "elapsed_time": "3:45:01", "remaining_time": "9:15:40", "throughput": 848.96, "total_tokens": 11462344}
{"current_steps": 985, "total_steps": 3400, "loss": 0.8674, "lr": 8.509649337881483e-05, "epoch": 0.5073396858099408, "percentage": 28.97, "elapsed_time": "3:46:04", "remaining_time": "9:14:17", "throughput": 849.32, "total_tokens": 11520808}
{"current_steps": 990, "total_steps": 3400, "loss": 0.8832, "lr": 8.492289119346943e-05, "epoch": 0.509915014164306, "percentage": 29.12, "elapsed_time": "3:47:08", "remaining_time": "9:12:56", "throughput": 849.64, "total_tokens": 11579248}
{"current_steps": 995, "total_steps": 3400, "loss": 0.8581, "lr": 8.474846307659658e-05, "epoch": 0.5124903425186711, "percentage": 29.26, "elapsed_time": "3:48:11", "remaining_time": "9:11:33", "throughput": 850.0, "total_tokens": 11637712}
{"current_steps": 1000, "total_steps": 3400, "loss": 0.8635, "lr": 8.457321315344694e-05, "epoch": 0.5150656708730363, "percentage": 29.41, "elapsed_time": "3:49:14", "remaining_time": "9:10:11", "throughput": 850.33, "total_tokens": 11696200}
{"current_steps": 1000, "total_steps": 3400, "eval_loss": 0.860200047492981, "epoch": 0.5150656708730363, "percentage": 29.41, "elapsed_time": "3:49:31", "remaining_time": "9:10:50", "throughput": 849.33, "total_tokens": 11696200}
{"current_steps": 1005, "total_steps": 3400, "loss": 0.8499, "lr": 8.439714556870704e-05, "epoch": 0.5176409992274015, "percentage": 29.56, "elapsed_time": "3:50:39", "remaining_time": "9:09:40", "throughput": 849.38, "total_tokens": 11754720}
{"current_steps": 1010, "total_steps": 3400, "loss": 0.8556, "lr": 8.422026448640124e-05, "epoch": 0.5202163275817667, "percentage": 29.71, "elapsed_time": "3:51:43", "remaining_time": "9:08:19", "throughput": 849.68, "total_tokens": 11813216}
{"current_steps": 1015, "total_steps": 3400, "loss": 0.8533, "lr": 8.40425740897932e-05, "epoch": 0.5227916559361319, "percentage": 29.85, "elapsed_time": "3:52:46", "remaining_time": "9:06:57", "throughput": 850.02, "total_tokens": 11871712}
{"current_steps": 1020, "total_steps": 3400, "loss": 0.8921, "lr": 8.386407858128706e-05, "epoch": 0.525366984290497, "percentage": 30.0, "elapsed_time": "3:53:49", "remaining_time": "9:05:35", "throughput": 850.36, "total_tokens": 11930200}
{"current_steps": 1025, "total_steps": 3400, "loss": 0.8815, "lr": 8.368478218232787e-05, "epoch": 0.5279423126448622, "percentage": 30.15, "elapsed_time": "3:54:52", "remaining_time": "9:04:14", "throughput": 850.69, "total_tokens": 11988704}
{"current_steps": 1030, "total_steps": 3400, "loss": 0.854, "lr": 8.350468913330192e-05, "epoch": 0.5305176409992274, "percentage": 30.29, "elapsed_time": "3:55:55", "remaining_time": "9:02:52", "throughput": 851.03, "total_tokens": 12047176}
{"current_steps": 1035, "total_steps": 3400, "loss": 0.8642, "lr": 8.33238036934364e-05, "epoch": 0.5330929693535926, "percentage": 30.44, "elapsed_time": "3:56:59", "remaining_time": "9:01:31", "throughput": 851.36, "total_tokens": 12105680}
{"current_steps": 1040, "total_steps": 3400, "loss": 0.8072, "lr": 8.31421301406986e-05, "epoch": 0.5356682977079578, "percentage": 30.59, "elapsed_time": "3:58:02", "remaining_time": "9:00:09", "throughput": 851.71, "total_tokens": 12164208}
{"current_steps": 1045, "total_steps": 3400, "loss": 0.8532, "lr": 8.29596727716949e-05, "epoch": 0.5382436260623229, "percentage": 30.74, "elapsed_time": "3:59:05", "remaining_time": "8:58:48", "throughput": 852.04, "total_tokens": 12222672}
{"current_steps": 1050, "total_steps": 3400, "loss": 0.8844, "lr": 8.277643590156894e-05, "epoch": 0.5408189544166881, "percentage": 30.88, "elapsed_time": "4:00:08", "remaining_time": "8:57:28", "throughput": 852.34, "total_tokens": 12281072}
{"current_steps": 1050, "total_steps": 3400, "eval_loss": 0.8446129560470581, "epoch": 0.5408189544166881, "percentage": 30.88, "elapsed_time": "4:00:24", "remaining_time": "8:58:03", "throughput": 851.39, "total_tokens": 12281072}
{"current_steps": 1055, "total_steps": 3400, "loss": 0.8602, "lr": 8.259242386389973e-05, "epoch": 0.5433942827710533, "percentage": 31.03, "elapsed_time": "4:01:33", "remaining_time": "8:56:54", "throughput": 851.41, "total_tokens": 12339544}
{"current_steps": 1060, "total_steps": 3400, "loss": 0.8615, "lr": 8.240764101059912e-05, "epoch": 0.5459696111254185, "percentage": 31.18, "elapsed_time": "4:02:36", "remaining_time": "8:55:34", "throughput": 851.71, "total_tokens": 12397992}
{"current_steps": 1065, "total_steps": 3400, "loss": 0.8732, "lr": 8.222209171180883e-05, "epoch": 0.5485449394797837, "percentage": 31.32, "elapsed_time": "4:03:39", "remaining_time": "8:54:13", "throughput": 852.04, "total_tokens": 12456480}
{"current_steps": 1070, "total_steps": 3400, "loss": 0.8691, "lr": 8.203578035579715e-05, "epoch": 0.5511202678341488, "percentage": 31.47, "elapsed_time": "4:04:42", "remaining_time": "8:52:52", "throughput": 852.36, "total_tokens": 12515000}
{"current_steps": 1075, "total_steps": 3400, "loss": 0.8544, "lr": 8.184871134885513e-05, "epoch": 0.553695596188514, "percentage": 31.62, "elapsed_time": "4:05:46", "remaining_time": "8:51:33", "throughput": 852.65, "total_tokens": 12573504}
{"current_steps": 1080, "total_steps": 3400, "loss": 0.8501, "lr": 8.166088911519235e-05, "epoch": 0.5562709245428792, "percentage": 31.76, "elapsed_time": "4:06:49", "remaining_time": "8:50:13", "throughput": 852.96, "total_tokens": 12632008}
{"current_steps": 1085, "total_steps": 3400, "loss": 0.8646, "lr": 8.147231809683236e-05, "epoch": 0.5588462528972444, "percentage": 31.91, "elapsed_time": "4:07:52", "remaining_time": "8:48:52", "throughput": 853.28, "total_tokens": 12690520}
{"current_steps": 1090, "total_steps": 3400, "loss": 0.8327, "lr": 8.128300275350756e-05, "epoch": 0.5614215812516096, "percentage": 32.06, "elapsed_time": "4:08:55", "remaining_time": "8:47:32", "throughput": 853.6, "total_tokens": 12749032}
{"current_steps": 1095, "total_steps": 3400, "loss": 0.8218, "lr": 8.109294756255375e-05, "epoch": 0.5639969096059748, "percentage": 32.21, "elapsed_time": "4:09:58", "remaining_time": "8:46:12", "throughput": 853.9, "total_tokens": 12807504}
{"current_steps": 1100, "total_steps": 3400, "loss": 0.8427, "lr": 8.090215701880419e-05, "epoch": 0.56657223796034, "percentage": 32.35, "elapsed_time": "4:11:01", "remaining_time": "8:44:52", "throughput": 854.21, "total_tokens": 12865992}
{"current_steps": 1100, "total_steps": 3400, "eval_loss": 0.7743102312088013, "epoch": 0.56657223796034, "percentage": 32.35, "elapsed_time": "4:11:17", "remaining_time": "8:45:26", "throughput": 853.3, "total_tokens": 12865992}
{"current_steps": 1105, "total_steps": 3400, "loss": 0.8335, "lr": 8.07106356344834e-05, "epoch": 0.5691475663147051, "percentage": 32.5, "elapsed_time": "4:12:25", "remaining_time": "8:44:16", "throughput": 853.35, "total_tokens": 12924448}
{"current_steps": 1110, "total_steps": 3400, "loss": 0.8267, "lr": 8.051838793910038e-05, "epoch": 0.5717228946690703, "percentage": 32.65, "elapsed_time": "4:13:28", "remaining_time": "8:42:56", "throughput": 853.65, "total_tokens": 12982912}
{"current_steps": 1115, "total_steps": 3400, "loss": 0.8866, "lr": 8.032541847934146e-05, "epoch": 0.5742982230234355, "percentage": 32.79, "elapsed_time": "4:14:31", "remaining_time": "8:41:36", "throughput": 853.96, "total_tokens": 13041424}
{"current_steps": 1120, "total_steps": 3400, "loss": 0.8446, "lr": 8.013173181896283e-05, "epoch": 0.5768735513778007, "percentage": 32.94, "elapsed_time": "4:15:35", "remaining_time": "8:40:18", "throughput": 854.24, "total_tokens": 13099888}
{"current_steps": 1125, "total_steps": 3400, "loss": 0.8176, "lr": 7.993733253868256e-05, "epoch": 0.5794488797321659, "percentage": 33.09, "elapsed_time": "4:16:38", "remaining_time": "8:38:58", "throughput": 854.54, "total_tokens": 13158344}
{"current_steps": 1130, "total_steps": 3400, "loss": 0.8138, "lr": 7.974222523607236e-05, "epoch": 0.582024208086531, "percentage": 33.24, "elapsed_time": "4:17:41", "remaining_time": "8:37:39", "throughput": 854.83, "total_tokens": 13216840}
{"current_steps": 1135, "total_steps": 3400, "loss": 0.8204, "lr": 7.954641452544865e-05, "epoch": 0.5845995364408962, "percentage": 33.38, "elapsed_time": "4:18:44", "remaining_time": "8:36:20", "throughput": 855.14, "total_tokens": 13275328}
{"current_steps": 1140, "total_steps": 3400, "loss": 0.8485, "lr": 7.934990503776363e-05, "epoch": 0.5871748647952614, "percentage": 33.53, "elapsed_time": "4:19:47", "remaining_time": "8:35:01", "throughput": 855.42, "total_tokens": 13333784}
{"current_steps": 1145, "total_steps": 3400, "loss": 0.8191, "lr": 7.915270142049566e-05, "epoch": 0.5897501931496266, "percentage": 33.68, "elapsed_time": "4:20:50", "remaining_time": "8:33:43", "throughput": 855.69, "total_tokens": 13392280}
{"current_steps": 1150, "total_steps": 3400, "loss": 0.8185, "lr": 7.89548083375394e-05, "epoch": 0.5923255215039918, "percentage": 33.82, "elapsed_time": "4:21:54", "remaining_time": "8:32:26", "throughput": 855.93, "total_tokens": 13450720}
{"current_steps": 1150, "total_steps": 3400, "eval_loss": 0.7827339768409729, "epoch": 0.5923255215039918, "percentage": 33.82, "elapsed_time": "4:22:10", "remaining_time": "8:32:57", "throughput": 855.06, "total_tokens": 13450720}
{"current_steps": 1155, "total_steps": 3400, "loss": 0.8168, "lr": 7.875623046909544e-05, "epoch": 0.5949008498583569, "percentage": 33.97, "elapsed_time": "4:23:19", "remaining_time": "8:31:49", "throughput": 855.06, "total_tokens": 13509200}
{"current_steps": 1160, "total_steps": 3400, "loss": 0.7749, "lr": 7.855697251155967e-05, "epoch": 0.5974761782127221, "percentage": 34.12, "elapsed_time": "4:24:22", "remaining_time": "8:30:31", "throughput": 855.31, "total_tokens": 13567656}
{"current_steps": 1165, "total_steps": 3400, "loss": 0.9132, "lr": 7.835703917741212e-05, "epoch": 0.6000515065670873, "percentage": 34.26, "elapsed_time": "4:25:28", "remaining_time": "8:29:18", "throughput": 855.43, "total_tokens": 13626136}
{"current_steps": 1170, "total_steps": 3400, "loss": 0.8308, "lr": 7.81564351951057e-05, "epoch": 0.6026268349214525, "percentage": 34.41, "elapsed_time": "4:26:31", "remaining_time": "8:28:00", "throughput": 855.72, "total_tokens": 13684608}
{"current_steps": 1175, "total_steps": 3400, "loss": 0.8011, "lr": 7.795516530895414e-05, "epoch": 0.6052021632758177, "percentage": 34.56, "elapsed_time": "4:27:36", "remaining_time": "8:26:44", "throughput": 855.93, "total_tokens": 13743080}
{"current_steps": 1180, "total_steps": 3400, "loss": 0.8309, "lr": 7.775323427901993e-05, "epoch": 0.6077774916301828, "percentage": 34.71, "elapsed_time": "4:28:39", "remaining_time": "8:25:26", "throughput": 856.19, "total_tokens": 13801552}
{"current_steps": 1185, "total_steps": 3400, "loss": 0.8089, "lr": 7.755064688100171e-05, "epoch": 0.610352819984548, "percentage": 34.85, "elapsed_time": "4:29:43", "remaining_time": "8:24:10", "throughput": 856.42, "total_tokens": 13860064}
{"current_steps": 1190, "total_steps": 3400, "loss": 0.8089, "lr": 7.734740790612136e-05, "epoch": 0.6129281483389132, "percentage": 35.0, "elapsed_time": "4:30:47", "remaining_time": "8:22:53", "throughput": 856.66, "total_tokens": 13918552}
{"current_steps": 1195, "total_steps": 3400, "loss": 0.8511, "lr": 7.714352216101055e-05, "epoch": 0.6155034766932784, "percentage": 35.15, "elapsed_time": "4:31:51", "remaining_time": "8:21:37", "throughput": 856.89, "total_tokens": 13977056}
{"current_steps": 1200, "total_steps": 3400, "loss": 0.8061, "lr": 7.693899446759727e-05, "epoch": 0.6180788050476436, "percentage": 35.29, "elapsed_time": "4:32:54", "remaining_time": "8:20:20", "throughput": 857.15, "total_tokens": 14035544}
{"current_steps": 1200, "total_steps": 3400, "eval_loss": 0.7593821287155151, "epoch": 0.6180788050476436, "percentage": 35.29, "elapsed_time": "4:33:10", "remaining_time": "8:20:49", "throughput": 856.31, "total_tokens": 14035544}
{"current_steps": 1205, "total_steps": 3400, "loss": 0.7871, "lr": 7.673382966299163e-05, "epoch": 0.6206541334020087, "percentage": 35.44, "elapsed_time": "4:34:18", "remaining_time": "8:19:40", "throughput": 856.33, "total_tokens": 14094024}
{"current_steps": 1210, "total_steps": 3400, "loss": 0.7594, "lr": 7.65280325993715e-05, "epoch": 0.623229461756374, "percentage": 35.59, "elapsed_time": "4:35:22", "remaining_time": "8:18:23", "throughput": 856.57, "total_tokens": 14152504}
{"current_steps": 1215, "total_steps": 3400, "loss": 0.7833, "lr": 7.63216081438678e-05, "epoch": 0.6258047901107391, "percentage": 35.74, "elapsed_time": "4:36:26", "remaining_time": "8:17:07", "throughput": 856.81, "total_tokens": 14210992}
{"current_steps": 1220, "total_steps": 3400, "loss": 0.8445, "lr": 7.611456117844934e-05, "epoch": 0.6283801184651043, "percentage": 35.88, "elapsed_time": "4:37:30", "remaining_time": "8:15:51", "throughput": 857.02, "total_tokens": 14269488}
{"current_steps": 1225, "total_steps": 3400, "loss": 0.7857, "lr": 7.59068965998074e-05, "epoch": 0.6309554468194695, "percentage": 36.03, "elapsed_time": "4:38:33", "remaining_time": "8:14:35", "throughput": 857.26, "total_tokens": 14327968}
{"current_steps": 1230, "total_steps": 3400, "loss": 0.8064, "lr": 7.569861931923989e-05, "epoch": 0.6335307751738347, "percentage": 36.18, "elapsed_time": "4:39:37", "remaining_time": "8:13:18", "throughput": 857.51, "total_tokens": 14386448}
{"current_steps": 1235, "total_steps": 3400, "loss": 0.7117, "lr": 7.548973426253521e-05, "epoch": 0.6361061035281999, "percentage": 36.32, "elapsed_time": "4:40:40", "remaining_time": "8:12:01", "throughput": 857.75, "total_tokens": 14444912}
{"current_steps": 1240, "total_steps": 3400, "loss": 0.7449, "lr": 7.528024636985575e-05, "epoch": 0.638681431882565, "percentage": 36.47, "elapsed_time": "4:41:44", "remaining_time": "8:10:45", "throughput": 857.98, "total_tokens": 14503392}
{"current_steps": 1245, "total_steps": 3400, "loss": 0.7507, "lr": 7.507016059562107e-05, "epoch": 0.6412567602369302, "percentage": 36.62, "elapsed_time": "4:42:47", "remaining_time": "8:09:28", "throughput": 858.24, "total_tokens": 14561872}
{"current_steps": 1250, "total_steps": 3400, "loss": 0.7917, "lr": 7.485948190839077e-05, "epoch": 0.6438320885912954, "percentage": 36.76, "elapsed_time": "4:43:50", "remaining_time": "8:08:12", "throughput": 858.48, "total_tokens": 14620336}
{"current_steps": 1250, "total_steps": 3400, "eval_loss": 0.7407085299491882, "epoch": 0.6438320885912954, "percentage": 36.76, "elapsed_time": "4:44:06", "remaining_time": "8:08:40", "throughput": 857.67, "total_tokens": 14620336}
{"current_steps": 1255, "total_steps": 3400, "loss": 0.7763, "lr": 7.464821529074679e-05, "epoch": 0.6464074169456606, "percentage": 36.91, "elapsed_time": "4:45:15", "remaining_time": "8:07:32", "throughput": 857.65, "total_tokens": 14678792}
{"current_steps": 1260, "total_steps": 3400, "loss": 0.7979, "lr": 7.443636573917585e-05, "epoch": 0.6489827453000258, "percentage": 37.06, "elapsed_time": "4:46:18", "remaining_time": "8:06:15", "throughput": 857.9, "total_tokens": 14737272}
{"current_steps": 1265, "total_steps": 3400, "loss": 0.7883, "lr": 7.422393826395108e-05, "epoch": 0.6515580736543909, "percentage": 37.21, "elapsed_time": "4:47:21", "remaining_time": "8:04:59", "throughput": 858.14, "total_tokens": 14795784}
{"current_steps": 1270, "total_steps": 3400, "loss": 0.7183, "lr": 7.40109378890136e-05, "epoch": 0.6541334020087561, "percentage": 37.35, "elapsed_time": "4:48:25", "remaining_time": "8:03:44", "throughput": 858.34, "total_tokens": 14854272}
{"current_steps": 1275, "total_steps": 3400, "loss": 0.762, "lr": 7.379736965185368e-05, "epoch": 0.6567087303631213, "percentage": 37.5, "elapsed_time": "4:49:29", "remaining_time": "8:02:28", "throughput": 858.58, "total_tokens": 14912720}
{"current_steps": 1280, "total_steps": 3400, "loss": 0.7951, "lr": 7.358323860339165e-05, "epoch": 0.6592840587174865, "percentage": 37.65, "elapsed_time": "4:50:32", "remaining_time": "8:01:12", "throughput": 858.82, "total_tokens": 14971192}
{"current_steps": 1285, "total_steps": 3400, "loss": 0.7528, "lr": 7.336854980785839e-05, "epoch": 0.6618593870718517, "percentage": 37.79, "elapsed_time": "4:51:35", "remaining_time": "7:59:56", "throughput": 859.06, "total_tokens": 15029656}
{"current_steps": 1290, "total_steps": 3400, "loss": 0.7633, "lr": 7.315330834267553e-05, "epoch": 0.6644347154262168, "percentage": 37.94, "elapsed_time": "4:52:38", "remaining_time": "7:58:39", "throughput": 859.3, "total_tokens": 15088144}
{"current_steps": 1295, "total_steps": 3400, "loss": 0.7443, "lr": 7.293751929833553e-05, "epoch": 0.667010043780582, "percentage": 38.09, "elapsed_time": "4:53:41", "remaining_time": "7:57:23", "throughput": 859.54, "total_tokens": 15146600}
{"current_steps": 1300, "total_steps": 3400, "loss": 0.7724, "lr": 7.272118777828108e-05, "epoch": 0.6695853721349472, "percentage": 38.24, "elapsed_time": "4:54:44", "remaining_time": "7:56:07", "throughput": 859.78, "total_tokens": 15205064}
{"current_steps": 1300, "total_steps": 3400, "eval_loss": 0.7189856171607971, "epoch": 0.6695853721349472, "percentage": 38.24, "elapsed_time": "4:55:00", "remaining_time": "7:56:33", "throughput": 859.0, "total_tokens": 15205064}
{"current_steps": 1305, "total_steps": 3400, "loss": 0.7524, "lr": 7.250431889878455e-05, "epoch": 0.6721607004893124, "percentage": 38.38, "elapsed_time": "4:56:08", "remaining_time": "7:55:25", "throughput": 859.0, "total_tokens": 15263560}
{"current_steps": 1310, "total_steps": 3400, "loss": 0.6748, "lr": 7.228691778882693e-05, "epoch": 0.6747360288436776, "percentage": 38.53, "elapsed_time": "4:57:11", "remaining_time": "7:54:09", "throughput": 859.25, "total_tokens": 15322016}
{"current_steps": 1315, "total_steps": 3400, "loss": 0.7571, "lr": 7.20689895899765e-05, "epoch": 0.6773113571980427, "percentage": 38.68, "elapsed_time": "4:58:15", "remaining_time": "7:52:53", "throughput": 859.48, "total_tokens": 15380504}
{"current_steps": 1320, "total_steps": 3400, "loss": 0.6615, "lr": 7.185053945626733e-05, "epoch": 0.6798866855524079, "percentage": 38.82, "elapsed_time": "4:59:18", "remaining_time": "7:51:37", "throughput": 859.72, "total_tokens": 15438944}
{"current_steps": 1325, "total_steps": 3400, "loss": 0.7421, "lr": 7.163157255407732e-05, "epoch": 0.6824620139067731, "percentage": 38.97, "elapsed_time": "5:00:21", "remaining_time": "7:50:21", "throughput": 859.95, "total_tokens": 15497384}
{"current_steps": 1330, "total_steps": 3400, "loss": 0.7886, "lr": 7.141209406200599e-05, "epoch": 0.6850373422611383, "percentage": 39.12, "elapsed_time": "5:01:24", "remaining_time": "7:49:06", "throughput": 860.19, "total_tokens": 15555856}
{"current_steps": 1335, "total_steps": 3400, "loss": 0.7484, "lr": 7.1192109170752e-05, "epoch": 0.6876126706155035, "percentage": 39.26, "elapsed_time": "5:02:27", "remaining_time": "7:47:50", "throughput": 860.42, "total_tokens": 15614368}
{"current_steps": 1340, "total_steps": 3400, "loss": 0.7086, "lr": 7.097162308299054e-05, "epoch": 0.6901879989698687, "percentage": 39.41, "elapsed_time": "5:03:30", "remaining_time": "7:46:34", "throughput": 860.66, "total_tokens": 15672864}
{"current_steps": 1345, "total_steps": 3400, "loss": 0.7494, "lr": 7.07506410132501e-05, "epoch": 0.6927633273242338, "percentage": 39.56, "elapsed_time": "5:04:33", "remaining_time": "7:45:19", "throughput": 860.89, "total_tokens": 15731376}
{"current_steps": 1350, "total_steps": 3400, "loss": 0.7278, "lr": 7.052916818778918e-05, "epoch": 0.695338655678599, "percentage": 39.71, "elapsed_time": "5:05:36", "remaining_time": "7:44:04", "throughput": 861.12, "total_tokens": 15789848}
{"current_steps": 1350, "total_steps": 3400, "eval_loss": 0.712917685508728, "epoch": 0.695338655678599, "percentage": 39.71, "elapsed_time": "5:05:52", "remaining_time": "7:44:28", "throughput": 860.37, "total_tokens": 15789848}
{"current_steps": 1355, "total_steps": 3400, "loss": 0.7005, "lr": 7.030720984447279e-05, "epoch": 0.6979139840329642, "percentage": 39.85, "elapsed_time": "5:07:00", "remaining_time": "7:43:20", "throughput": 860.36, "total_tokens": 15848328}
{"current_steps": 1360, "total_steps": 3400, "loss": 0.7406, "lr": 7.008477123264848e-05, "epoch": 0.7004893123873294, "percentage": 40.0, "elapsed_time": "5:08:03", "remaining_time": "7:42:05", "throughput": 860.58, "total_tokens": 15906824}
{"current_steps": 1365, "total_steps": 3400, "loss": 0.73, "lr": 6.986185761302224e-05, "epoch": 0.7030646407416946, "percentage": 40.15, "elapsed_time": "5:09:06", "remaining_time": "7:40:50", "throughput": 860.8, "total_tokens": 15965312}
{"current_steps": 1370, "total_steps": 3400, "loss": 0.7069, "lr": 6.963847425753403e-05, "epoch": 0.7056399690960597, "percentage": 40.29, "elapsed_time": "5:10:10", "remaining_time": "7:39:36", "throughput": 861.0, "total_tokens": 16023824}
{"current_steps": 1375, "total_steps": 3400, "loss": 0.6859, "lr": 6.941462644923318e-05, "epoch": 0.7082152974504249, "percentage": 40.44, "elapsed_time": "5:11:14", "remaining_time": "7:38:21", "throughput": 861.21, "total_tokens": 16082280}
{"current_steps": 1380, "total_steps": 3400, "loss": 0.7254, "lr": 6.919031948215335e-05, "epoch": 0.7107906258047901, "percentage": 40.59, "elapsed_time": "5:12:17", "remaining_time": "7:37:07", "throughput": 861.43, "total_tokens": 16140800}
{"current_steps": 1385, "total_steps": 3400, "loss": 0.717, "lr": 6.896555866118741e-05, "epoch": 0.7133659541591553, "percentage": 40.74, "elapsed_time": "5:13:20", "remaining_time": "7:35:52", "throughput": 861.65, "total_tokens": 16199320}
{"current_steps": 1390, "total_steps": 3400, "loss": 0.7094, "lr": 6.87403493019619e-05, "epoch": 0.7159412825135205, "percentage": 40.88, "elapsed_time": "5:14:23", "remaining_time": "7:34:37", "throughput": 861.86, "total_tokens": 16257768}
{"current_steps": 1395, "total_steps": 3400, "loss": 0.7862, "lr": 6.851469673071143e-05, "epoch": 0.7185166108678857, "percentage": 41.03, "elapsed_time": "5:15:27", "remaining_time": "7:33:23", "throughput": 862.05, "total_tokens": 16316264}
{"current_steps": 1400, "total_steps": 3400, "loss": 0.7359, "lr": 6.828860628415253e-05, "epoch": 0.7210919392222508, "percentage": 41.18, "elapsed_time": "5:16:30", "remaining_time": "7:32:09", "throughput": 862.24, "total_tokens": 16374784}
{"current_steps": 1400, "total_steps": 3400, "eval_loss": 0.6643603444099426, "epoch": 0.7210919392222508, "percentage": 41.18, "elapsed_time": "5:16:47", "remaining_time": "7:32:32", "throughput": 861.51, "total_tokens": 16374784}
{"current_steps": 1405, "total_steps": 3400, "loss": 0.706, "lr": 6.806208330935766e-05, "epoch": 0.723667267576616, "percentage": 41.32, "elapsed_time": "5:17:55", "remaining_time": "7:31:25", "throughput": 861.49, "total_tokens": 16433288}
{"current_steps": 1410, "total_steps": 3400, "loss": 0.6714, "lr": 6.783513316362855e-05, "epoch": 0.7262425959309812, "percentage": 41.47, "elapsed_time": "5:18:58", "remaining_time": "7:30:11", "throughput": 861.7, "total_tokens": 16491784}
{"current_steps": 1415, "total_steps": 3400, "loss": 0.693, "lr": 6.760776121436962e-05, "epoch": 0.7288179242853464, "percentage": 41.62, "elapsed_time": "5:20:02", "remaining_time": "7:28:57", "throughput": 861.9, "total_tokens": 16550272}
{"current_steps": 1420, "total_steps": 3400, "loss": 0.7005, "lr": 6.737997283896103e-05, "epoch": 0.7313932526397116, "percentage": 41.76, "elapsed_time": "5:21:04", "remaining_time": "7:27:42", "throughput": 862.12, "total_tokens": 16608704}
{"current_steps": 1425, "total_steps": 3400, "loss": 0.6573, "lr": 6.715177342463145e-05, "epoch": 0.7339685809940767, "percentage": 41.91, "elapsed_time": "5:22:08", "remaining_time": "7:26:28", "throughput": 862.32, "total_tokens": 16667200}
{"current_steps": 1430, "total_steps": 3400, "loss": 0.6751, "lr": 6.692316836833065e-05, "epoch": 0.7365439093484419, "percentage": 42.06, "elapsed_time": "5:23:11", "remaining_time": "7:25:13", "throughput": 862.54, "total_tokens": 16725704}
{"current_steps": 1435, "total_steps": 3400, "loss": 0.6173, "lr": 6.6694163076602e-05, "epoch": 0.7391192377028071, "percentage": 42.21, "elapsed_time": "5:24:14", "remaining_time": "7:23:59", "throughput": 862.74, "total_tokens": 16784192}
{"current_steps": 1440, "total_steps": 3400, "loss": 0.728, "lr": 6.646476296545434e-05, "epoch": 0.7416945660571723, "percentage": 42.35, "elapsed_time": "5:25:17", "remaining_time": "7:22:45", "throughput": 862.95, "total_tokens": 16842704}
{"current_steps": 1445, "total_steps": 3400, "loss": 0.743, "lr": 6.623497346023418e-05, "epoch": 0.7442698944115375, "percentage": 42.5, "elapsed_time": "5:26:20", "remaining_time": "7:21:31", "throughput": 863.15, "total_tokens": 16901176}
{"current_steps": 1450, "total_steps": 3400, "loss": 0.6291, "lr": 6.60047999954972e-05, "epoch": 0.7468452227659027, "percentage": 42.65, "elapsed_time": "5:27:23", "remaining_time": "7:20:17", "throughput": 863.37, "total_tokens": 16959632}
{"current_steps": 1450, "total_steps": 3400, "eval_loss": 0.753077507019043, "epoch": 0.7468452227659027, "percentage": 42.65, "elapsed_time": "5:27:39", "remaining_time": "7:20:38", "throughput": 862.66, "total_tokens": 16959632}
{"current_steps": 1455, "total_steps": 3400, "loss": 0.6533, "lr": 6.57742480148798e-05, "epoch": 0.7494205511202678, "percentage": 42.79, "elapsed_time": "5:28:47", "remaining_time": "7:19:30", "throughput": 862.67, "total_tokens": 17018072}
{"current_steps": 1460, "total_steps": 3400, "loss": 0.7114, "lr": 6.554332297097031e-05, "epoch": 0.751995879474633, "percentage": 42.94, "elapsed_time": "5:29:50", "remaining_time": "7:18:16", "throughput": 862.87, "total_tokens": 17076560}
{"current_steps": 1465, "total_steps": 3400, "loss": 0.6568, "lr": 6.53120303251801e-05, "epoch": 0.7545712078289982, "percentage": 43.09, "elapsed_time": "5:30:53", "remaining_time": "7:17:02", "throughput": 863.07, "total_tokens": 17135016}
{"current_steps": 1470, "total_steps": 3400, "loss": 0.7016, "lr": 6.508037554761432e-05, "epoch": 0.7571465361833634, "percentage": 43.24, "elapsed_time": "5:31:56", "remaining_time": "7:15:49", "throughput": 863.26, "total_tokens": 17193496}
{"current_steps": 1475, "total_steps": 3400, "loss": 0.6612, "lr": 6.484836411694267e-05, "epoch": 0.7597218645377286, "percentage": 43.38, "elapsed_time": "5:32:59", "remaining_time": "7:14:35", "throughput": 863.46, "total_tokens": 17251944}
{"current_steps": 1480, "total_steps": 3400, "loss": 0.6115, "lr": 6.461600152026965e-05, "epoch": 0.7622971928920937, "percentage": 43.53, "elapsed_time": "5:34:03", "remaining_time": "7:13:22", "throughput": 863.65, "total_tokens": 17310456}
{"current_steps": 1485, "total_steps": 3400, "loss": 0.6458, "lr": 6.438329325300499e-05, "epoch": 0.7648725212464589, "percentage": 43.68, "elapsed_time": "5:35:06", "remaining_time": "7:12:08", "throughput": 863.85, "total_tokens": 17368968}
{"current_steps": 1490, "total_steps": 3400, "loss": 0.6434, "lr": 6.415024481873352e-05, "epoch": 0.7674478496008241, "percentage": 43.82, "elapsed_time": "5:36:09", "remaining_time": "7:10:55", "throughput": 864.04, "total_tokens": 17427424}
{"current_steps": 1495, "total_steps": 3400, "loss": 0.5973, "lr": 6.391686172908506e-05, "epoch": 0.7700231779551893, "percentage": 43.97, "elapsed_time": "5:37:12", "remaining_time": "7:09:41", "throughput": 864.25, "total_tokens": 17485936}
{"current_steps": 1500, "total_steps": 3400, "loss": 0.6021, "lr": 6.368314950360415e-05, "epoch": 0.7725985063095545, "percentage": 44.12, "elapsed_time": "5:38:15", "remaining_time": "7:08:27", "throughput": 864.44, "total_tokens": 17544440}
{"current_steps": 1500, "total_steps": 3400, "eval_loss": 0.632923424243927, "epoch": 0.7725985063095545, "percentage": 44.12, "elapsed_time": "5:38:31", "remaining_time": "7:08:48", "throughput": 863.76, "total_tokens": 17544440}
{"current_steps": 1505, "total_steps": 3400, "loss": 0.5779, "lr": 6.344911366961934e-05, "epoch": 0.7751738346639196, "percentage": 44.26, "elapsed_time": "5:39:39", "remaining_time": "7:07:40", "throughput": 863.76, "total_tokens": 17602952}
{"current_steps": 1510, "total_steps": 3400, "loss": 0.6707, "lr": 6.321475976211266e-05, "epoch": 0.7777491630182848, "percentage": 44.41, "elapsed_time": "5:40:42", "remaining_time": "7:06:27", "throughput": 863.95, "total_tokens": 17661440}
{"current_steps": 1515, "total_steps": 3400, "loss": 0.6326, "lr": 6.298009332358856e-05, "epoch": 0.78032449137265, "percentage": 44.56, "elapsed_time": "5:41:45", "remaining_time": "7:05:13", "throughput": 864.15, "total_tokens": 17719928}
{"current_steps": 1520, "total_steps": 3400, "loss": 0.6472, "lr": 6.274511990394294e-05, "epoch": 0.7828998197270152, "percentage": 44.71, "elapsed_time": "5:42:48", "remaining_time": "7:04:00", "throughput": 864.34, "total_tokens": 17778424}
{"current_steps": 1525, "total_steps": 3400, "loss": 0.6215, "lr": 6.250984506033183e-05, "epoch": 0.7854751480813804, "percentage": 44.85, "elapsed_time": "5:43:51", "remaining_time": "7:02:46", "throughput": 864.55, "total_tokens": 17836936}
{"current_steps": 1530, "total_steps": 3400, "loss": 0.6102, "lr": 6.227427435703997e-05, "epoch": 0.7880504764357456, "percentage": 45.0, "elapsed_time": "5:44:54", "remaining_time": "7:01:33", "throughput": 864.73, "total_tokens": 17895392}
{"current_steps": 1535, "total_steps": 3400, "loss": 0.6161, "lr": 6.203841336534924e-05, "epoch": 0.7906258047901107, "percentage": 45.15, "elapsed_time": "5:45:57", "remaining_time": "7:00:20", "throughput": 864.92, "total_tokens": 17953872}
{"current_steps": 1540, "total_steps": 3400, "loss": 0.6103, "lr": 6.180226766340688e-05, "epoch": 0.7932011331444759, "percentage": 45.29, "elapsed_time": "5:47:00", "remaining_time": "6:59:07", "throughput": 865.11, "total_tokens": 18012320}
{"current_steps": 1545, "total_steps": 3400, "loss": 0.5791, "lr": 6.156584283609359e-05, "epoch": 0.7957764614988411, "percentage": 45.44, "elapsed_time": "5:48:03", "remaining_time": "6:57:54", "throughput": 865.3, "total_tokens": 18070792}
{"current_steps": 1550, "total_steps": 3400, "loss": 0.667, "lr": 6.132914447489137e-05, "epoch": 0.7983517898532063, "percentage": 45.59, "elapsed_time": "5:49:06", "remaining_time": "6:56:40", "throughput": 865.5, "total_tokens": 18129304}
{"current_steps": 1550, "total_steps": 3400, "eval_loss": 0.6617516279220581, "epoch": 0.7983517898532063, "percentage": 45.59, "elapsed_time": "5:49:22", "remaining_time": "6:57:00", "throughput": 864.83, "total_tokens": 18129304}
{"current_steps": 1555, "total_steps": 3400, "loss": 0.5681, "lr": 6.109217817775139e-05, "epoch": 0.8009271182075715, "percentage": 45.74, "elapsed_time": "5:50:30", "remaining_time": "6:55:52", "throughput": 864.83, "total_tokens": 18187728}
{"current_steps": 1560, "total_steps": 3400, "loss": 0.6292, "lr": 6.085494954896156e-05, "epoch": 0.8035024465619367, "percentage": 45.88, "elapsed_time": "5:51:33", "remaining_time": "6:54:39", "throughput": 865.0, "total_tokens": 18246192}
{"current_steps": 1565, "total_steps": 3400, "loss": 0.6512, "lr": 6.061746419901388e-05, "epoch": 0.8060777749163018, "percentage": 46.03, "elapsed_time": "5:52:37", "remaining_time": "6:53:27", "throughput": 865.16, "total_tokens": 18304632}
{"current_steps": 1570, "total_steps": 3400, "loss": 0.5476, "lr": 6.0379727744471936e-05, "epoch": 0.808653103270667, "percentage": 46.18, "elapsed_time": "5:53:40", "remaining_time": "6:52:15", "throughput": 865.34, "total_tokens": 18363136}
{"current_steps": 1575, "total_steps": 3400, "loss": 0.5632, "lr": 6.014174580783794e-05, "epoch": 0.8112284316250322, "percentage": 46.32, "elapsed_time": "5:54:44", "remaining_time": "6:51:02", "throughput": 865.51, "total_tokens": 18421592}
{"current_steps": 1580, "total_steps": 3400, "loss": 0.6225, "lr": 5.990352401741981e-05, "epoch": 0.8138037599793974, "percentage": 46.47, "elapsed_time": "5:55:47", "remaining_time": "6:49:49", "throughput": 865.69, "total_tokens": 18480104}
{"current_steps": 1585, "total_steps": 3400, "loss": 0.5801, "lr": 5.9665068007197976e-05, "epoch": 0.8163790883337626, "percentage": 46.62, "elapsed_time": "5:56:50", "remaining_time": "6:48:37", "throughput": 865.85, "total_tokens": 18538600}
{"current_steps": 1590, "total_steps": 3400, "loss": 0.6364, "lr": 5.94263834166923e-05, "epoch": 0.8189544166881277, "percentage": 46.76, "elapsed_time": "5:57:53", "remaining_time": "6:47:25", "throughput": 866.03, "total_tokens": 18597104}
{"current_steps": 1595, "total_steps": 3400, "loss": 0.6088, "lr": 5.918747589082853e-05, "epoch": 0.8215297450424929, "percentage": 46.91, "elapsed_time": "5:58:57", "remaining_time": "6:46:12", "throughput": 866.21, "total_tokens": 18655584}
{"current_steps": 1600, "total_steps": 3400, "loss": 0.6564, "lr": 5.8948351079804875e-05, "epoch": 0.8241050733968581, "percentage": 47.06, "elapsed_time": "6:00:00", "remaining_time": "6:45:00", "throughput": 866.39, "total_tokens": 18714072}
{"current_steps": 1600, "total_steps": 3400, "eval_loss": 0.6319106221199036, "epoch": 0.8241050733968581, "percentage": 47.06, "elapsed_time": "6:00:16", "remaining_time": "6:45:18", "throughput": 865.75, "total_tokens": 18714072}
{"current_steps": 1605, "total_steps": 3400, "loss": 0.6095, "lr": 5.8709014638958404e-05, "epoch": 0.8266804017512233, "percentage": 47.21, "elapsed_time": "6:01:23", "remaining_time": "6:44:10", "throughput": 865.74, "total_tokens": 18772552}
{"current_steps": 1610, "total_steps": 3400, "loss": 0.5896, "lr": 5.846947222863123e-05, "epoch": 0.8292557301055885, "percentage": 47.35, "elapsed_time": "6:02:26", "remaining_time": "6:42:58", "throughput": 865.92, "total_tokens": 18830992}
{"current_steps": 1615, "total_steps": 3400, "loss": 0.545, "lr": 5.8229729514036705e-05, "epoch": 0.8318310584599536, "percentage": 47.5, "elapsed_time": "6:03:29", "remaining_time": "6:41:45", "throughput": 866.1, "total_tokens": 18889480}
{"current_steps": 1620, "total_steps": 3400, "loss": 0.6021, "lr": 5.7989792165125356e-05, "epoch": 0.8344063868143188, "percentage": 47.65, "elapsed_time": "6:04:32", "remaining_time": "6:40:33", "throughput": 866.27, "total_tokens": 18947936}
{"current_steps": 1625, "total_steps": 3400, "loss": 0.5741, "lr": 5.774966585645092e-05, "epoch": 0.836981715168684, "percentage": 47.79, "elapsed_time": "6:05:35", "remaining_time": "6:39:20", "throughput": 866.46, "total_tokens": 19006432}
{"current_steps": 1630, "total_steps": 3400, "loss": 0.5796, "lr": 5.7509356267035975e-05, "epoch": 0.8395570435230492, "percentage": 47.94, "elapsed_time": "6:06:39", "remaining_time": "6:38:08", "throughput": 866.62, "total_tokens": 19064920}
{"current_steps": 1635, "total_steps": 3400, "loss": 0.5088, "lr": 5.726886908023776e-05, "epoch": 0.8421323718774144, "percentage": 48.09, "elapsed_time": "6:07:42", "remaining_time": "6:36:56", "throughput": 866.8, "total_tokens": 19123376}
{"current_steps": 1640, "total_steps": 3400, "loss": 0.5431, "lr": 5.702820998361373e-05, "epoch": 0.8447077002317795, "percentage": 48.24, "elapsed_time": "6:08:45", "remaining_time": "6:35:44", "throughput": 866.97, "total_tokens": 19181864}
{"current_steps": 1645, "total_steps": 3400, "loss": 0.5849, "lr": 5.6787384668786994e-05, "epoch": 0.8472830285861447, "percentage": 48.38, "elapsed_time": "6:09:48", "remaining_time": "6:34:31", "throughput": 867.14, "total_tokens": 19240352}
{"current_steps": 1650, "total_steps": 3400, "loss": 0.5668, "lr": 5.654639883131178e-05, "epoch": 0.8498583569405099, "percentage": 48.53, "elapsed_time": "6:10:51", "remaining_time": "6:33:19", "throughput": 867.31, "total_tokens": 19298848}
{"current_steps": 1650, "total_steps": 3400, "eval_loss": 0.6634677648544312, "epoch": 0.8498583569405099, "percentage": 48.53, "elapsed_time": "6:11:07", "remaining_time": "6:33:36", "throughput": 866.69, "total_tokens": 19298848}
{"current_steps": 1655, "total_steps": 3400, "loss": 0.584, "lr": 5.6305258170538676e-05, "epoch": 0.8524336852948751, "percentage": 48.68, "elapsed_time": "6:12:15", "remaining_time": "6:32:29", "throughput": 866.67, "total_tokens": 19357304}
{"current_steps": 1660, "total_steps": 3400, "loss": 0.5544, "lr": 5.606396838947988e-05, "epoch": 0.8550090136492403, "percentage": 48.82, "elapsed_time": "6:13:18", "remaining_time": "6:31:17", "throughput": 866.84, "total_tokens": 19415800}
{"current_steps": 1665, "total_steps": 3400, "loss": 0.6269, "lr": 5.582253519467432e-05, "epoch": 0.8575843420036054, "percentage": 48.97, "elapsed_time": "6:14:21", "remaining_time": "6:30:05", "throughput": 867.02, "total_tokens": 19474256}
{"current_steps": 1670, "total_steps": 3400, "loss": 0.5073, "lr": 5.558096429605263e-05, "epoch": 0.8601596703579707, "percentage": 49.12, "elapsed_time": "6:15:24", "remaining_time": "6:28:53", "throughput": 867.18, "total_tokens": 19532736}
{"current_steps": 1675, "total_steps": 3400, "loss": 0.5319, "lr": 5.533926140680221e-05, "epoch": 0.8627349987123358, "percentage": 49.26, "elapsed_time": "6:16:28", "remaining_time": "6:27:42", "throughput": 867.31, "total_tokens": 19591184}
{"current_steps": 1680, "total_steps": 3400, "loss": 0.4525, "lr": 5.509743224323203e-05, "epoch": 0.865310327066701, "percentage": 49.41, "elapsed_time": "6:17:31", "remaining_time": "6:26:30", "throughput": 867.48, "total_tokens": 19649656}
{"current_steps": 1685, "total_steps": 3400, "loss": 0.5276, "lr": 5.485548252463749e-05, "epoch": 0.8678856554210662, "percentage": 49.56, "elapsed_time": "6:18:34", "remaining_time": "6:25:19", "throughput": 867.64, "total_tokens": 19708144}
{"current_steps": 1690, "total_steps": 3400, "loss": 0.5482, "lr": 5.4613417973165106e-05, "epoch": 0.8704609837754314, "percentage": 49.71, "elapsed_time": "6:19:37", "remaining_time": "6:24:07", "throughput": 867.79, "total_tokens": 19766592}
{"current_steps": 1695, "total_steps": 3400, "loss": 0.4656, "lr": 5.4371244313677225e-05, "epoch": 0.8730363121297966, "percentage": 49.85, "elapsed_time": "6:20:41", "remaining_time": "6:22:56", "throughput": 867.93, "total_tokens": 19825064}
{"current_steps": 1700, "total_steps": 3400, "loss": 0.5701, "lr": 5.4128967273616625e-05, "epoch": 0.8756116404841617, "percentage": 50.0, "elapsed_time": "6:21:45", "remaining_time": "6:21:45", "throughput": 868.08, "total_tokens": 19883504}
{"current_steps": 1700, "total_steps": 3400, "eval_loss": 0.7144017815589905, "epoch": 0.8756116404841617, "percentage": 50.0, "elapsed_time": "6:22:01", "remaining_time": "6:22:01", "throughput": 867.47, "total_tokens": 19883504}
{"current_steps": 1705, "total_steps": 3400, "loss": 0.5823, "lr": 5.388659258287102e-05, "epoch": 0.8781869688385269, "percentage": 50.15, "elapsed_time": "6:23:09", "remaining_time": "6:20:54", "throughput": 867.44, "total_tokens": 19942000}
{"current_steps": 1710, "total_steps": 3400, "loss": 0.5446, "lr": 5.364412597363759e-05, "epoch": 0.8807622971928921, "percentage": 50.29, "elapsed_time": "6:24:12", "remaining_time": "6:19:42", "throughput": 867.6, "total_tokens": 20000440}
{"current_steps": 1715, "total_steps": 3400, "loss": 0.5769, "lr": 5.3401573180287426e-05, "epoch": 0.8833376255472573, "percentage": 50.44, "elapsed_time": "6:25:15", "remaining_time": "6:18:31", "throughput": 867.76, "total_tokens": 20058920}
{"current_steps": 1720, "total_steps": 3400, "loss": 0.5614, "lr": 5.315893993922986e-05, "epoch": 0.8859129539016225, "percentage": 50.59, "elapsed_time": "6:26:18", "remaining_time": "6:17:19", "throughput": 867.92, "total_tokens": 20117416}
{"current_steps": 1725, "total_steps": 3400, "loss": 0.5215, "lr": 5.29162319887768e-05, "epoch": 0.8884882822559876, "percentage": 50.74, "elapsed_time": "6:27:21", "remaining_time": "6:16:08", "throughput": 868.08, "total_tokens": 20175936}
{"current_steps": 1730, "total_steps": 3400, "loss": 0.4968, "lr": 5.26734550690071e-05, "epoch": 0.8910636106103528, "percentage": 50.88, "elapsed_time": "6:28:24", "remaining_time": "6:14:56", "throughput": 868.24, "total_tokens": 20234368}
{"current_steps": 1735, "total_steps": 3400, "loss": 0.5029, "lr": 5.243061492163073e-05, "epoch": 0.893638938964718, "percentage": 51.03, "elapsed_time": "6:29:27", "remaining_time": "6:13:45", "throughput": 868.41, "total_tokens": 20292856}
{"current_steps": 1740, "total_steps": 3400, "loss": 0.5249, "lr": 5.2187717289852955e-05, "epoch": 0.8962142673190832, "percentage": 51.18, "elapsed_time": "6:30:30", "remaining_time": "6:12:33", "throughput": 868.57, "total_tokens": 20351272}
{"current_steps": 1745, "total_steps": 3400, "loss": 0.5801, "lr": 5.1944767918238624e-05, "epoch": 0.8987895956734484, "percentage": 51.32, "elapsed_time": "6:31:33", "remaining_time": "6:11:22", "throughput": 868.73, "total_tokens": 20409744}
{"current_steps": 1750, "total_steps": 3400, "loss": 0.546, "lr": 5.170177255257618e-05, "epoch": 0.9013649240278135, "percentage": 51.47, "elapsed_time": "6:32:36", "remaining_time": "6:10:10", "throughput": 868.88, "total_tokens": 20468200}
{"current_steps": 1750, "total_steps": 3400, "eval_loss": 0.672294020652771, "epoch": 0.9013649240278135, "percentage": 51.47, "elapsed_time": "6:32:52", "remaining_time": "6:10:25", "throughput": 868.29, "total_tokens": 20468200}
{"current_steps": 1755, "total_steps": 3400, "loss": 0.5248, "lr": 5.145873693974188e-05, "epoch": 0.9039402523821787, "percentage": 51.62, "elapsed_time": "6:34:00", "remaining_time": "6:09:19", "throughput": 868.27, "total_tokens": 20526696}
{"current_steps": 1760, "total_steps": 3400, "loss": 0.4756, "lr": 5.12156668275638e-05, "epoch": 0.9065155807365439, "percentage": 51.76, "elapsed_time": "6:35:04", "remaining_time": "6:08:07", "throughput": 868.42, "total_tokens": 20585160}
{"current_steps": 1765, "total_steps": 3400, "loss": 0.4405, "lr": 5.097256796468598e-05, "epoch": 0.9090909090909091, "percentage": 51.91, "elapsed_time": "6:36:06", "remaining_time": "6:06:56", "throughput": 868.6, "total_tokens": 20643672}
{"current_steps": 1770, "total_steps": 3400, "loss": 0.5201, "lr": 5.072944610043232e-05, "epoch": 0.9116662374452743, "percentage": 52.06, "elapsed_time": "6:37:09", "remaining_time": "6:05:45", "throughput": 868.75, "total_tokens": 20702152}
{"current_steps": 1775, "total_steps": 3400, "loss": 0.4662, "lr": 5.048630698467081e-05, "epoch": 0.9142415657996394, "percentage": 52.21, "elapsed_time": "6:38:13", "remaining_time": "6:04:34", "throughput": 868.88, "total_tokens": 20760664}
{"current_steps": 1780, "total_steps": 3400, "loss": 0.5376, "lr": 5.024315636767738e-05, "epoch": 0.9168168941540046, "percentage": 52.35, "elapsed_time": "6:39:16", "remaining_time": "6:03:23", "throughput": 869.03, "total_tokens": 20819144}
{"current_steps": 1785, "total_steps": 3400, "loss": 0.5174, "lr": 5e-05, "epoch": 0.9193922225083698, "percentage": 52.5, "elapsed_time": "6:40:20", "remaining_time": "6:02:12", "throughput": 869.17, "total_tokens": 20877624}
{"current_steps": 1790, "total_steps": 3400, "loss": 0.4273, "lr": 4.9756843632322626e-05, "epoch": 0.921967550862735, "percentage": 52.65, "elapsed_time": "6:41:23", "remaining_time": "6:01:01", "throughput": 869.32, "total_tokens": 20936112}
{"current_steps": 1795, "total_steps": 3400, "loss": 0.4646, "lr": 4.9513693015329197e-05, "epoch": 0.9245428792171002, "percentage": 52.79, "elapsed_time": "6:42:26", "remaining_time": "5:59:50", "throughput": 869.47, "total_tokens": 20994608}
{"current_steps": 1800, "total_steps": 3400, "loss": 0.412, "lr": 4.9270553899567686e-05, "epoch": 0.9271182075714653, "percentage": 52.94, "elapsed_time": "6:43:29", "remaining_time": "5:58:39", "throughput": 869.61, "total_tokens": 21053080}
{"current_steps": 1800, "total_steps": 3400, "eval_loss": 0.6768696904182434, "epoch": 0.9271182075714653, "percentage": 52.94, "elapsed_time": "6:43:45", "remaining_time": "5:58:54", "throughput": 869.03, "total_tokens": 21053080}
{"current_steps": 1805, "total_steps": 3400, "loss": 0.4845, "lr": 4.902743203531405e-05, "epoch": 0.9296935359258306, "percentage": 53.09, "elapsed_time": "6:44:53", "remaining_time": "5:57:47", "throughput": 869.02, "total_tokens": 21111592}
{"current_steps": 1810, "total_steps": 3400, "loss": 0.441, "lr": 4.8784333172436206e-05, "epoch": 0.9322688642801957, "percentage": 53.24, "elapsed_time": "6:45:56", "remaining_time": "5:56:35", "throughput": 869.18, "total_tokens": 21170024}
{"current_steps": 1815, "total_steps": 3400, "loss": 0.545, "lr": 4.854126306025812e-05, "epoch": 0.9348441926345609, "percentage": 53.38, "elapsed_time": "6:46:59", "remaining_time": "5:55:25", "throughput": 869.32, "total_tokens": 21228480}
{"current_steps": 1820, "total_steps": 3400, "loss": 0.4697, "lr": 4.829822744742383e-05, "epoch": 0.9374195209889261, "percentage": 53.53, "elapsed_time": "6:48:02", "remaining_time": "5:54:13", "throughput": 869.48, "total_tokens": 21286944}
{"current_steps": 1825, "total_steps": 3400, "loss": 0.423, "lr": 4.8055232081761395e-05, "epoch": 0.9399948493432912, "percentage": 53.68, "elapsed_time": "6:49:05", "remaining_time": "5:53:03", "throughput": 869.61, "total_tokens": 21345456}
{"current_steps": 1830, "total_steps": 3400, "loss": 0.4715, "lr": 4.781228271014704e-05, "epoch": 0.9425701776976565, "percentage": 53.82, "elapsed_time": "6:50:09", "remaining_time": "5:51:52", "throughput": 869.75, "total_tokens": 21403896}
{"current_steps": 1835, "total_steps": 3400, "loss": 0.5149, "lr": 4.756938507836929e-05, "epoch": 0.9451455060520216, "percentage": 53.97, "elapsed_time": "6:51:13", "remaining_time": "5:50:42", "throughput": 869.87, "total_tokens": 21462360}
{"current_steps": 1840, "total_steps": 3400, "loss": 0.5403, "lr": 4.732654493099291e-05, "epoch": 0.9477208344063868, "percentage": 54.12, "elapsed_time": "6:52:16", "remaining_time": "5:49:32", "throughput": 870.0, "total_tokens": 21520864}
{"current_steps": 1845, "total_steps": 3400, "loss": 0.4757, "lr": 4.708376801122321e-05, "epoch": 0.950296162760752, "percentage": 54.26, "elapsed_time": "6:53:19", "remaining_time": "5:48:21", "throughput": 870.14, "total_tokens": 21579376}
{"current_steps": 1850, "total_steps": 3400, "loss": 0.4347, "lr": 4.6841060060770154e-05, "epoch": 0.9528714911151172, "percentage": 54.41, "elapsed_time": "6:54:22", "remaining_time": "5:47:11", "throughput": 870.29, "total_tokens": 21637848}
{"current_steps": 1850, "total_steps": 3400, "eval_loss": 0.6808218359947205, "epoch": 0.9528714911151172, "percentage": 54.41, "elapsed_time": "6:54:39", "remaining_time": "5:47:24", "throughput": 869.72, "total_tokens": 21637848}
{"current_steps": 1855, "total_steps": 3400, "loss": 0.5132, "lr": 4.659842681971258e-05, "epoch": 0.9554468194694824, "percentage": 54.56, "elapsed_time": "6:55:47", "remaining_time": "5:46:18", "throughput": 869.69, "total_tokens": 21696328}
{"current_steps": 1860, "total_steps": 3400, "loss": 0.4347, "lr": 4.635587402636241e-05, "epoch": 0.9580221478238475, "percentage": 54.71, "elapsed_time": "6:56:50", "remaining_time": "5:45:07", "throughput": 869.84, "total_tokens": 21754816}
{"current_steps": 1865, "total_steps": 3400, "loss": 0.4015, "lr": 4.611340741712901e-05, "epoch": 0.9605974761782127, "percentage": 54.85, "elapsed_time": "6:57:53", "remaining_time": "5:43:56", "throughput": 869.99, "total_tokens": 21813296}
{"current_steps": 1870, "total_steps": 3400, "loss": 0.5023, "lr": 4.5871032726383386e-05, "epoch": 0.9631728045325779, "percentage": 55.0, "elapsed_time": "6:58:56", "remaining_time": "5:42:45", "throughput": 870.14, "total_tokens": 21871800}
{"current_steps": 1875, "total_steps": 3400, "loss": 0.5334, "lr": 4.562875568632278e-05, "epoch": 0.9657481328869431, "percentage": 55.15, "elapsed_time": "6:59:59", "remaining_time": "5:41:35", "throughput": 870.27, "total_tokens": 21930272}
{"current_steps": 1880, "total_steps": 3400, "loss": 0.4386, "lr": 4.5386582026834906e-05, "epoch": 0.9683234612413083, "percentage": 55.29, "elapsed_time": "7:01:02", "remaining_time": "5:40:25", "throughput": 870.4, "total_tokens": 21988736}
{"current_steps": 1885, "total_steps": 3400, "loss": 0.3807, "lr": 4.5144517475362514e-05, "epoch": 0.9708987895956734, "percentage": 55.44, "elapsed_time": "7:02:05", "remaining_time": "5:39:14", "throughput": 870.55, "total_tokens": 22047200}
{"current_steps": 1890, "total_steps": 3400, "loss": 0.4177, "lr": 4.490256775676797e-05, "epoch": 0.9734741179500386, "percentage": 55.59, "elapsed_time": "7:03:08", "remaining_time": "5:38:04", "throughput": 870.69, "total_tokens": 22105664}
{"current_steps": 1895, "total_steps": 3400, "loss": 0.5239, "lr": 4.466073859319781e-05, "epoch": 0.9760494463044038, "percentage": 55.74, "elapsed_time": "7:04:11", "remaining_time": "5:36:53", "throughput": 870.85, "total_tokens": 22164184}
{"current_steps": 1900, "total_steps": 3400, "loss": 0.3737, "lr": 4.441903570394739e-05, "epoch": 0.978624774658769, "percentage": 55.88, "elapsed_time": "7:05:14", "remaining_time": "5:35:43", "throughput": 870.98, "total_tokens": 22222632}
{"current_steps": 1900, "total_steps": 3400, "eval_loss": 0.773033082485199, "epoch": 0.978624774658769, "percentage": 55.88, "elapsed_time": "7:05:30", "remaining_time": "5:35:55", "throughput": 870.43, "total_tokens": 22222632}
{"current_steps": 1905, "total_steps": 3400, "loss": 0.4437, "lr": 4.41774648053257e-05, "epoch": 0.9812001030131342, "percentage": 56.03, "elapsed_time": "7:06:39", "remaining_time": "5:34:49", "throughput": 870.38, "total_tokens": 22281080}
{"current_steps": 1910, "total_steps": 3400, "loss": 0.465, "lr": 4.3936031610520124e-05, "epoch": 0.9837754313674993, "percentage": 56.18, "elapsed_time": "7:07:42", "remaining_time": "5:33:39", "throughput": 870.51, "total_tokens": 22339552}
{"current_steps": 1915, "total_steps": 3400, "loss": 0.4975, "lr": 4.3694741829461336e-05, "epoch": 0.9863507597218646, "percentage": 56.32, "elapsed_time": "7:08:45", "remaining_time": "5:32:29", "throughput": 870.66, "total_tokens": 22398056}
{"current_steps": 1920, "total_steps": 3400, "loss": 0.4504, "lr": 4.345360116868823e-05, "epoch": 0.9889260880762297, "percentage": 56.47, "elapsed_time": "7:09:48", "remaining_time": "5:31:18", "throughput": 870.8, "total_tokens": 22456520}
{"current_steps": 1925, "total_steps": 3400, "loss": 0.4528, "lr": 4.321261533121303e-05, "epoch": 0.9915014164305949, "percentage": 56.62, "elapsed_time": "7:10:51", "remaining_time": "5:30:08", "throughput": 870.94, "total_tokens": 22515024}
{"current_steps": 1930, "total_steps": 3400, "loss": 0.441, "lr": 4.2971790016386286e-05, "epoch": 0.9940767447849601, "percentage": 56.76, "elapsed_time": "7:11:54", "remaining_time": "5:28:57", "throughput": 871.07, "total_tokens": 22573480}
{"current_steps": 1935, "total_steps": 3400, "loss": 0.4678, "lr": 4.273113091976225e-05, "epoch": 0.9966520731393252, "percentage": 56.91, "elapsed_time": "7:12:57", "remaining_time": "5:27:47", "throughput": 871.21, "total_tokens": 22631960}
{"current_steps": 1940, "total_steps": 3400, "loss": 0.4352, "lr": 4.249064373296403e-05, "epoch": 0.9992274014936905, "percentage": 57.06, "elapsed_time": "7:14:00", "remaining_time": "5:26:37", "throughput": 871.35, "total_tokens": 22690432}
{"current_steps": 1945, "total_steps": 3400, "loss": 0.3792, "lr": 4.225033414354908e-05, "epoch": 1.001545197012619, "percentage": 57.21, "elapsed_time": "7:14:57", "remaining_time": "5:25:23", "throughput": 871.45, "total_tokens": 22743048}
{"current_steps": 1950, "total_steps": 3400, "loss": 0.3783, "lr": 4.201020783487464e-05, "epoch": 1.0041205253669843, "percentage": 57.35, "elapsed_time": "7:16:03", "remaining_time": "5:24:14", "throughput": 871.5, "total_tokens": 22801512}
{"current_steps": 1950, "total_steps": 3400, "eval_loss": 0.6983156204223633, "epoch": 1.0041205253669843, "percentage": 57.35, "elapsed_time": "7:16:19", "remaining_time": "5:24:27", "throughput": 870.95, "total_tokens": 22801512}
{"current_steps": 1955, "total_steps": 3400, "loss": 0.3784, "lr": 4.17702704859633e-05, "epoch": 1.0066958537213495, "percentage": 57.5, "elapsed_time": "7:17:30", "remaining_time": "5:23:22", "throughput": 870.83, "total_tokens": 22859952}
{"current_steps": 1960, "total_steps": 3400, "loss": 0.5587, "lr": 4.153052777136879e-05, "epoch": 1.0092711820757148, "percentage": 57.65, "elapsed_time": "7:18:36", "remaining_time": "5:22:14", "throughput": 870.87, "total_tokens": 22918440}
{"current_steps": 1965, "total_steps": 3400, "loss": 0.3803, "lr": 4.1290985361041614e-05, "epoch": 1.0118465104300798, "percentage": 57.79, "elapsed_time": "7:19:40", "remaining_time": "5:21:05", "throughput": 870.98, "total_tokens": 22976944}
{"current_steps": 1970, "total_steps": 3400, "loss": 0.4038, "lr": 4.105164892019514e-05, "epoch": 1.014421838784445, "percentage": 57.94, "elapsed_time": "7:20:44", "remaining_time": "5:19:55", "throughput": 871.09, "total_tokens": 23035408}
{"current_steps": 1975, "total_steps": 3400, "loss": 0.3226, "lr": 4.0812524109171476e-05, "epoch": 1.0169971671388103, "percentage": 58.09, "elapsed_time": "7:21:48", "remaining_time": "5:18:46", "throughput": 871.2, "total_tokens": 23093912}
{"current_steps": 1980, "total_steps": 3400, "loss": 0.4026, "lr": 4.0573616583307705e-05, "epoch": 1.0195724954931753, "percentage": 58.24, "elapsed_time": "7:22:52", "remaining_time": "5:17:36", "throughput": 871.3, "total_tokens": 23152344}
{"current_steps": 1985, "total_steps": 3400, "loss": 0.4225, "lr": 4.033493199280202e-05, "epoch": 1.0221478238475405, "percentage": 58.38, "elapsed_time": "7:23:55", "remaining_time": "5:16:26", "throughput": 871.43, "total_tokens": 23210800}
{"current_steps": 1990, "total_steps": 3400, "loss": 0.3058, "lr": 4.009647598258022e-05, "epoch": 1.0247231522019058, "percentage": 58.53, "elapsed_time": "7:24:58", "remaining_time": "5:15:17", "throughput": 871.55, "total_tokens": 23269304}
{"current_steps": 1995, "total_steps": 3400, "loss": 0.3821, "lr": 3.985825419216207e-05, "epoch": 1.0272984805562708, "percentage": 58.68, "elapsed_time": "7:26:01", "remaining_time": "5:14:07", "throughput": 871.69, "total_tokens": 23327800}
{"current_steps": 2000, "total_steps": 3400, "loss": 0.3328, "lr": 3.962027225552807e-05, "epoch": 1.029873808910636, "percentage": 58.82, "elapsed_time": "7:27:05", "remaining_time": "5:12:57", "throughput": 871.8, "total_tokens": 23386232}
{"current_steps": 2000, "total_steps": 3400, "eval_loss": 0.7484827041625977, "epoch": 1.029873808910636, "percentage": 58.82, "elapsed_time": "7:27:21", "remaining_time": "5:13:08", "throughput": 871.28, "total_tokens": 23386232}
{"current_steps": 2005, "total_steps": 3400, "loss": 0.362, "lr": 3.938253580098613e-05, "epoch": 1.0324491372650013, "percentage": 58.97, "elapsed_time": "7:28:30", "remaining_time": "5:12:03", "throughput": 871.22, "total_tokens": 23444712}
{"current_steps": 2010, "total_steps": 3400, "loss": 0.3903, "lr": 3.914505045103845e-05, "epoch": 1.0350244656193666, "percentage": 59.12, "elapsed_time": "7:29:34", "remaining_time": "5:10:54", "throughput": 871.31, "total_tokens": 23503192}
{"current_steps": 2015, "total_steps": 3400, "loss": 0.3341, "lr": 3.8907821822248605e-05, "epoch": 1.0375997939737316, "percentage": 59.26, "elapsed_time": "7:30:38", "remaining_time": "5:09:44", "throughput": 871.41, "total_tokens": 23561688}
{"current_steps": 2020, "total_steps": 3400, "loss": 0.4794, "lr": 3.867085552510864e-05, "epoch": 1.0401751223280968, "percentage": 59.41, "elapsed_time": "7:31:43", "remaining_time": "5:08:35", "throughput": 871.49, "total_tokens": 23620160}
{"current_steps": 2025, "total_steps": 3400, "loss": 0.4104, "lr": 3.843415716390644e-05, "epoch": 1.042750450682462, "percentage": 59.56, "elapsed_time": "7:32:47", "remaining_time": "5:07:26", "throughput": 871.59, "total_tokens": 23678624}
{"current_steps": 2030, "total_steps": 3400, "loss": 0.3639, "lr": 3.819773233659314e-05, "epoch": 1.045325779036827, "percentage": 59.71, "elapsed_time": "7:33:51", "remaining_time": "5:06:17", "throughput": 871.68, "total_tokens": 23737064}
{"current_steps": 2035, "total_steps": 3400, "loss": 0.4294, "lr": 3.7961586634650767e-05, "epoch": 1.0479011073911924, "percentage": 59.85, "elapsed_time": "7:34:55", "remaining_time": "5:05:08", "throughput": 871.77, "total_tokens": 23795568}
{"current_steps": 2040, "total_steps": 3400, "loss": 0.4713, "lr": 3.772572564296005e-05, "epoch": 1.0504764357455576, "percentage": 60.0, "elapsed_time": "7:36:00", "remaining_time": "5:04:00", "throughput": 871.85, "total_tokens": 23854040}
{"current_steps": 2045, "total_steps": 3400, "loss": 0.3864, "lr": 3.749015493966817e-05, "epoch": 1.0530517640999228, "percentage": 60.15, "elapsed_time": "7:37:04", "remaining_time": "5:02:51", "throughput": 871.93, "total_tokens": 23912520}
{"current_steps": 2050, "total_steps": 3400, "loss": 0.3602, "lr": 3.7254880096057073e-05, "epoch": 1.0556270924542879, "percentage": 60.29, "elapsed_time": "7:38:09", "remaining_time": "5:01:42", "throughput": 872.01, "total_tokens": 23971048}
{"current_steps": 2050, "total_steps": 3400, "eval_loss": 0.7190810441970825, "epoch": 1.0556270924542879, "percentage": 60.29, "elapsed_time": "7:38:25", "remaining_time": "5:01:53", "throughput": 871.5, "total_tokens": 23971048}
{"current_steps": 2055, "total_steps": 3400, "loss": 0.3203, "lr": 3.7019906676411446e-05, "epoch": 1.0582024208086531, "percentage": 60.44, "elapsed_time": "7:39:34", "remaining_time": "5:00:47", "throughput": 871.43, "total_tokens": 24029544}
{"current_steps": 2060, "total_steps": 3400, "loss": 0.3906, "lr": 3.678524023788735e-05, "epoch": 1.0607777491630184, "percentage": 60.59, "elapsed_time": "7:40:39", "remaining_time": "4:59:38", "throughput": 871.52, "total_tokens": 24088008}
{"current_steps": 2065, "total_steps": 3400, "loss": 0.3604, "lr": 3.6550886330380665e-05, "epoch": 1.0633530775173834, "percentage": 60.74, "elapsed_time": "7:41:42", "remaining_time": "4:58:29", "throughput": 871.63, "total_tokens": 24146480}
{"current_steps": 2070, "total_steps": 3400, "loss": 0.3271, "lr": 3.631685049639586e-05, "epoch": 1.0659284058717486, "percentage": 60.88, "elapsed_time": "7:42:46", "remaining_time": "4:57:20", "throughput": 871.73, "total_tokens": 24204984}
{"current_steps": 2075, "total_steps": 3400, "loss": 0.2996, "lr": 3.608313827091493e-05, "epoch": 1.0685037342261139, "percentage": 61.03, "elapsed_time": "7:43:51", "remaining_time": "4:56:11", "throughput": 871.8, "total_tokens": 24263456}
{"current_steps": 2080, "total_steps": 3400, "loss": 0.3767, "lr": 3.5849755181266474e-05, "epoch": 1.071079062580479, "percentage": 61.18, "elapsed_time": "7:44:55", "remaining_time": "4:55:03", "throughput": 871.88, "total_tokens": 24321960}
{"current_steps": 2085, "total_steps": 3400, "loss": 0.4208, "lr": 3.5616706746995026e-05, "epoch": 1.0736543909348442, "percentage": 61.32, "elapsed_time": "7:46:00", "remaining_time": "4:53:54", "throughput": 871.96, "total_tokens": 24380464}
{"current_steps": 2090, "total_steps": 3400, "loss": 0.3479, "lr": 3.538399847973036e-05, "epoch": 1.0762297192892094, "percentage": 61.47, "elapsed_time": "7:47:04", "remaining_time": "4:52:45", "throughput": 872.05, "total_tokens": 24438976}
{"current_steps": 2095, "total_steps": 3400, "loss": 0.3654, "lr": 3.515163588305735e-05, "epoch": 1.0788050476435747, "percentage": 61.62, "elapsed_time": "7:48:09", "remaining_time": "4:51:37", "throughput": 872.12, "total_tokens": 24497448}
{"current_steps": 2100, "total_steps": 3400, "loss": 0.3351, "lr": 3.491962445238569e-05, "epoch": 1.0813803759979397, "percentage": 61.76, "elapsed_time": "7:49:14", "remaining_time": "4:50:28", "throughput": 872.19, "total_tokens": 24555904}
{"current_steps": 2100, "total_steps": 3400, "eval_loss": 0.8075026869773865, "epoch": 1.0813803759979397, "percentage": 61.76, "elapsed_time": "7:49:30", "remaining_time": "4:50:38", "throughput": 871.69, "total_tokens": 24555904}
{"current_steps": 2105, "total_steps": 3400, "loss": 0.3827, "lr": 3.4687969674819906e-05, "epoch": 1.083955704352305, "percentage": 61.91, "elapsed_time": "7:50:39", "remaining_time": "4:49:32", "throughput": 871.63, "total_tokens": 24614392}
{"current_steps": 2110, "total_steps": 3400, "loss": 0.3676, "lr": 3.445667702902969e-05, "epoch": 1.0865310327066702, "percentage": 62.06, "elapsed_time": "7:51:43", "remaining_time": "4:48:24", "throughput": 871.72, "total_tokens": 24672848}
{"current_steps": 2115, "total_steps": 3400, "loss": 0.3253, "lr": 3.4225751985120215e-05, "epoch": 1.0891063610610352, "percentage": 62.21, "elapsed_time": "7:52:47", "remaining_time": "4:47:15", "throughput": 871.81, "total_tokens": 24731344}
{"current_steps": 2120, "total_steps": 3400, "loss": 0.4297, "lr": 3.3995200004502816e-05, "epoch": 1.0916816894154004, "percentage": 62.35, "elapsed_time": "7:53:52", "remaining_time": "4:46:06", "throughput": 871.89, "total_tokens": 24789832}
{"current_steps": 2125, "total_steps": 3400, "loss": 0.3536, "lr": 3.3765026539765834e-05, "epoch": 1.0942570177697657, "percentage": 62.5, "elapsed_time": "7:54:56", "remaining_time": "4:44:57", "throughput": 871.98, "total_tokens": 24848264}
{"current_steps": 2130, "total_steps": 3400, "loss": 0.3588, "lr": 3.3535237034545675e-05, "epoch": 1.0968323461241307, "percentage": 62.65, "elapsed_time": "7:56:00", "remaining_time": "4:43:49", "throughput": 872.06, "total_tokens": 24906744}
{"current_steps": 2135, "total_steps": 3400, "loss": 0.3666, "lr": 3.330583692339802e-05, "epoch": 1.099407674478496, "percentage": 62.79, "elapsed_time": "7:57:05", "remaining_time": "4:42:40", "throughput": 872.15, "total_tokens": 24965256}
{"current_steps": 2140, "total_steps": 3400, "loss": 0.3334, "lr": 3.307683163166934e-05, "epoch": 1.1019830028328612, "percentage": 62.94, "elapsed_time": "7:58:08", "remaining_time": "4:41:31", "throughput": 872.25, "total_tokens": 25023768}
{"current_steps": 2145, "total_steps": 3400, "loss": 0.3848, "lr": 3.284822657536856e-05, "epoch": 1.1045583311872265, "percentage": 63.09, "elapsed_time": "7:59:11", "remaining_time": "4:40:22", "throughput": 872.37, "total_tokens": 25082248}
{"current_steps": 2150, "total_steps": 3400, "loss": 0.3699, "lr": 3.262002716103897e-05, "epoch": 1.1071336595415915, "percentage": 63.24, "elapsed_time": "8:00:15", "remaining_time": "4:39:12", "throughput": 872.49, "total_tokens": 25140752}
{"current_steps": 2150, "total_steps": 3400, "eval_loss": 0.8523861169815063, "epoch": 1.1071336595415915, "percentage": 63.24, "elapsed_time": "8:00:31", "remaining_time": "4:39:22", "throughput": 872.0, "total_tokens": 25140752}
{"current_steps": 2155, "total_steps": 3400, "loss": 0.3154, "lr": 3.2392238785630386e-05, "epoch": 1.1097089878959567, "percentage": 63.38, "elapsed_time": "8:01:38", "remaining_time": "4:38:15", "throughput": 871.98, "total_tokens": 25199208}
{"current_steps": 2160, "total_steps": 3400, "loss": 0.3915, "lr": 3.216486683637146e-05, "epoch": 1.112284316250322, "percentage": 63.53, "elapsed_time": "8:02:41", "remaining_time": "4:37:06", "throughput": 872.11, "total_tokens": 25257680}
{"current_steps": 2165, "total_steps": 3400, "loss": 0.3675, "lr": 3.1937916690642356e-05, "epoch": 1.114859644604687, "percentage": 63.68, "elapsed_time": "8:03:44", "remaining_time": "4:35:56", "throughput": 872.23, "total_tokens": 25316200}
{"current_steps": 2170, "total_steps": 3400, "loss": 0.4047, "lr": 3.1711393715847476e-05, "epoch": 1.1174349729590523, "percentage": 63.82, "elapsed_time": "8:04:47", "remaining_time": "4:34:47", "throughput": 872.35, "total_tokens": 25374656}
{"current_steps": 2175, "total_steps": 3400, "loss": 0.4155, "lr": 3.14853032692886e-05, "epoch": 1.1200103013134175, "percentage": 63.97, "elapsed_time": "8:05:50", "remaining_time": "4:33:38", "throughput": 872.47, "total_tokens": 25433168}
{"current_steps": 2180, "total_steps": 3400, "loss": 0.3966, "lr": 3.125965069803811e-05, "epoch": 1.1225856296677827, "percentage": 64.12, "elapsed_time": "8:06:53", "remaining_time": "4:32:29", "throughput": 872.59, "total_tokens": 25491664}
{"current_steps": 2185, "total_steps": 3400, "loss": 0.3068, "lr": 3.103444133881261e-05, "epoch": 1.1251609580221478, "percentage": 64.26, "elapsed_time": "8:07:57", "remaining_time": "4:31:19", "throughput": 872.7, "total_tokens": 25550128}
{"current_steps": 2190, "total_steps": 3400, "loss": 0.386, "lr": 3.080968051784666e-05, "epoch": 1.127736286376513, "percentage": 64.41, "elapsed_time": "8:09:00", "remaining_time": "4:30:10", "throughput": 872.82, "total_tokens": 25608624}
{"current_steps": 2195, "total_steps": 3400, "loss": 0.3898, "lr": 3.058537355076683e-05, "epoch": 1.1303116147308783, "percentage": 64.56, "elapsed_time": "8:10:03", "remaining_time": "4:29:01", "throughput": 872.93, "total_tokens": 25667128}
{"current_steps": 2200, "total_steps": 3400, "loss": 0.4016, "lr": 3.0361525742465973e-05, "epoch": 1.1328869430852433, "percentage": 64.71, "elapsed_time": "8:11:06", "remaining_time": "4:27:52", "throughput": 873.04, "total_tokens": 25725560}
{"current_steps": 2200, "total_steps": 3400, "eval_loss": 0.7534744143486023, "epoch": 1.1328869430852433, "percentage": 64.71, "elapsed_time": "8:11:22", "remaining_time": "4:28:01", "throughput": 872.57, "total_tokens": 25725560}
{"current_steps": 2205, "total_steps": 3400, "loss": 0.3465, "lr": 3.0138142386977787e-05, "epoch": 1.1354622714396085, "percentage": 64.85, "elapsed_time": "8:12:30", "remaining_time": "4:26:55", "throughput": 872.53, "total_tokens": 25784048}
{"current_steps": 2210, "total_steps": 3400, "loss": 0.3077, "lr": 2.991522876735154e-05, "epoch": 1.1380375997939738, "percentage": 65.0, "elapsed_time": "8:13:33", "remaining_time": "4:25:45", "throughput": 872.66, "total_tokens": 25842512}
{"current_steps": 2215, "total_steps": 3400, "loss": 0.4497, "lr": 2.9692790155527227e-05, "epoch": 1.140612928148339, "percentage": 65.15, "elapsed_time": "8:14:36", "remaining_time": "4:24:36", "throughput": 872.77, "total_tokens": 25900992}
{"current_steps": 2220, "total_steps": 3400, "loss": 0.3811, "lr": 2.9470831812210837e-05, "epoch": 1.143188256502704, "percentage": 65.29, "elapsed_time": "8:15:39", "remaining_time": "4:23:27", "throughput": 872.89, "total_tokens": 25959448}
{"current_steps": 2225, "total_steps": 3400, "loss": 0.4053, "lr": 2.924935898674992e-05, "epoch": 1.1457635848570693, "percentage": 65.44, "elapsed_time": "8:16:42", "remaining_time": "4:22:18", "throughput": 873.0, "total_tokens": 26017936}
{"current_steps": 2230, "total_steps": 3400, "loss": 0.3421, "lr": 2.902837691700945e-05, "epoch": 1.1483389132114346, "percentage": 65.59, "elapsed_time": "8:17:46", "remaining_time": "4:21:09", "throughput": 873.11, "total_tokens": 26076440}
{"current_steps": 2235, "total_steps": 3400, "loss": 0.3228, "lr": 2.880789082924798e-05, "epoch": 1.1509142415657996, "percentage": 65.74, "elapsed_time": "8:18:49", "remaining_time": "4:20:00", "throughput": 873.22, "total_tokens": 26134896}
{"current_steps": 2240, "total_steps": 3400, "loss": 0.3695, "lr": 2.858790593799405e-05, "epoch": 1.1534895699201648, "percentage": 65.88, "elapsed_time": "8:19:52", "remaining_time": "4:18:51", "throughput": 873.33, "total_tokens": 26193368}
{"current_steps": 2245, "total_steps": 3400, "loss": 0.3463, "lr": 2.8368427445922696e-05, "epoch": 1.15606489827453, "percentage": 66.03, "elapsed_time": "8:20:55", "remaining_time": "4:17:42", "throughput": 873.45, "total_tokens": 26251848}
{"current_steps": 2250, "total_steps": 3400, "loss": 0.3442, "lr": 2.8149460543732664e-05, "epoch": 1.158640226628895, "percentage": 66.18, "elapsed_time": "8:21:58", "remaining_time": "4:16:33", "throughput": 873.56, "total_tokens": 26310336}
{"current_steps": 2250, "total_steps": 3400, "eval_loss": 0.7066138386726379, "epoch": 1.158640226628895, "percentage": 66.18, "elapsed_time": "8:22:14", "remaining_time": "4:16:42", "throughput": 873.09, "total_tokens": 26310336}
{"current_steps": 2255, "total_steps": 3400, "loss": 0.3547, "lr": 2.7931010410023518e-05, "epoch": 1.1612155549832603, "percentage": 66.32, "elapsed_time": "8:23:22", "remaining_time": "4:15:35", "throughput": 873.08, "total_tokens": 26368840}
{"current_steps": 2260, "total_steps": 3400, "loss": 0.3125, "lr": 2.771308221117309e-05, "epoch": 1.1637908833376256, "percentage": 66.47, "elapsed_time": "8:24:25", "remaining_time": "4:14:26", "throughput": 873.19, "total_tokens": 26427280}
{"current_steps": 2265, "total_steps": 3400, "loss": 0.3521, "lr": 2.749568110121545e-05, "epoch": 1.1663662116919906, "percentage": 66.62, "elapsed_time": "8:25:28", "remaining_time": "4:13:17", "throughput": 873.29, "total_tokens": 26485760}
{"current_steps": 2270, "total_steps": 3400, "loss": 0.281, "lr": 2.7278812221718924e-05, "epoch": 1.1689415400463559, "percentage": 66.76, "elapsed_time": "8:26:35", "remaining_time": "4:12:10", "throughput": 873.31, "total_tokens": 26544224}
{"current_steps": 2275, "total_steps": 3400, "loss": 0.3653, "lr": 2.7062480701664488e-05, "epoch": 1.1715168684007211, "percentage": 66.91, "elapsed_time": "8:27:38", "remaining_time": "4:11:01", "throughput": 873.42, "total_tokens": 26602712}
{"current_steps": 2280, "total_steps": 3400, "loss": 0.3964, "lr": 2.6846691657324473e-05, "epoch": 1.1740921967550864, "percentage": 67.06, "elapsed_time": "8:28:41", "remaining_time": "4:09:52", "throughput": 873.52, "total_tokens": 26661160}
{"current_steps": 2285, "total_steps": 3400, "loss": 0.3119, "lr": 2.663145019214163e-05, "epoch": 1.1766675251094514, "percentage": 67.21, "elapsed_time": "8:29:44", "remaining_time": "4:08:44", "throughput": 873.63, "total_tokens": 26719648}
{"current_steps": 2290, "total_steps": 3400, "loss": 0.3832, "lr": 2.6416761396608362e-05, "epoch": 1.1792428534638166, "percentage": 67.35, "elapsed_time": "8:30:47", "remaining_time": "4:07:35", "throughput": 873.74, "total_tokens": 26778112}
{"current_steps": 2295, "total_steps": 3400, "loss": 0.3277, "lr": 2.6202630348146324e-05, "epoch": 1.1818181818181819, "percentage": 67.5, "elapsed_time": "8:31:50", "remaining_time": "4:06:26", "throughput": 873.86, "total_tokens": 26836592}
{"current_steps": 2300, "total_steps": 3400, "loss": 0.3877, "lr": 2.598906211098643e-05, "epoch": 1.184393510172547, "percentage": 67.65, "elapsed_time": "8:32:53", "remaining_time": "4:05:17", "throughput": 873.97, "total_tokens": 26895096}
{"current_steps": 2300, "total_steps": 3400, "eval_loss": 0.727741539478302, "epoch": 1.184393510172547, "percentage": 67.65, "elapsed_time": "8:33:09", "remaining_time": "4:05:25", "throughput": 873.52, "total_tokens": 26895096}
{"current_steps": 2305, "total_steps": 3400, "loss": 0.3033, "lr": 2.577606173604894e-05, "epoch": 1.1869688385269122, "percentage": 67.79, "elapsed_time": "8:34:17", "remaining_time": "4:04:19", "throughput": 873.48, "total_tokens": 26953560}
{"current_steps": 2310, "total_steps": 3400, "loss": 0.4104, "lr": 2.5563634260824175e-05, "epoch": 1.1895441668812774, "percentage": 67.94, "elapsed_time": "8:35:20", "remaining_time": "4:03:10", "throughput": 873.58, "total_tokens": 27012024}
{"current_steps": 2315, "total_steps": 3400, "loss": 0.3447, "lr": 2.535178470925323e-05, "epoch": 1.1921194952356426, "percentage": 68.09, "elapsed_time": "8:36:24", "remaining_time": "4:02:01", "throughput": 873.69, "total_tokens": 27070520}
{"current_steps": 2320, "total_steps": 3400, "loss": 0.2882, "lr": 2.5140518091609256e-05, "epoch": 1.1946948235900077, "percentage": 68.24, "elapsed_time": "8:37:26", "remaining_time": "4:00:52", "throughput": 873.8, "total_tokens": 27128984}
{"current_steps": 2325, "total_steps": 3400, "loss": 0.2817, "lr": 2.4929839404378936e-05, "epoch": 1.197270151944373, "percentage": 68.38, "elapsed_time": "8:38:30", "remaining_time": "3:59:44", "throughput": 873.91, "total_tokens": 27187432}
{"current_steps": 2330, "total_steps": 3400, "loss": 0.3693, "lr": 2.471975363014428e-05, "epoch": 1.1998454802987382, "percentage": 68.53, "elapsed_time": "8:39:32", "remaining_time": "3:58:35", "throughput": 874.03, "total_tokens": 27245920}
{"current_steps": 2335, "total_steps": 3400, "loss": 0.3587, "lr": 2.451026573746482e-05, "epoch": 1.2024208086531032, "percentage": 68.68, "elapsed_time": "8:40:35", "remaining_time": "3:57:26", "throughput": 874.13, "total_tokens": 27304384}
{"current_steps": 2340, "total_steps": 3400, "loss": 0.354, "lr": 2.430138068076013e-05, "epoch": 1.2049961370074684, "percentage": 68.82, "elapsed_time": "8:41:38", "remaining_time": "3:56:18", "throughput": 874.24, "total_tokens": 27362864}
{"current_steps": 2345, "total_steps": 3400, "loss": 0.3209, "lr": 2.4093103400192625e-05, "epoch": 1.2075714653618337, "percentage": 68.97, "elapsed_time": "8:42:41", "remaining_time": "3:55:09", "throughput": 874.35, "total_tokens": 27421360}
{"current_steps": 2350, "total_steps": 3400, "loss": 0.3871, "lr": 2.388543882155067e-05, "epoch": 1.210146793716199, "percentage": 69.12, "elapsed_time": "8:43:44", "remaining_time": "3:54:00", "throughput": 874.46, "total_tokens": 27479840}
{"current_steps": 2350, "total_steps": 3400, "eval_loss": 0.7659633755683899, "epoch": 1.210146793716199, "percentage": 69.12, "elapsed_time": "8:44:00", "remaining_time": "3:54:08", "throughput": 874.02, "total_tokens": 27479840}
{"current_steps": 2355, "total_steps": 3400, "loss": 0.352, "lr": 2.3678391856132204e-05, "epoch": 1.212722122070564, "percentage": 69.26, "elapsed_time": "8:45:08", "remaining_time": "3:53:01", "throughput": 874.0, "total_tokens": 27538344}
{"current_steps": 2360, "total_steps": 3400, "loss": 0.347, "lr": 2.3471967400628513e-05, "epoch": 1.2152974504249292, "percentage": 69.41, "elapsed_time": "8:46:11", "remaining_time": "3:51:52", "throughput": 874.1, "total_tokens": 27596808}
{"current_steps": 2365, "total_steps": 3400, "loss": 0.3667, "lr": 2.3266170337008398e-05, "epoch": 1.2178727787792945, "percentage": 69.56, "elapsed_time": "8:47:14", "remaining_time": "3:50:44", "throughput": 874.2, "total_tokens": 27655272}
{"current_steps": 2370, "total_steps": 3400, "loss": 0.3311, "lr": 2.306100553240274e-05, "epoch": 1.2204481071336595, "percentage": 69.71, "elapsed_time": "8:48:18", "remaining_time": "3:49:36", "throughput": 874.29, "total_tokens": 27713784}
{"current_steps": 2375, "total_steps": 3400, "loss": 0.2964, "lr": 2.2856477838989456e-05, "epoch": 1.2230234354880247, "percentage": 69.85, "elapsed_time": "8:49:21", "remaining_time": "3:48:27", "throughput": 874.4, "total_tokens": 27772248}
{"current_steps": 2380, "total_steps": 3400, "loss": 0.3683, "lr": 2.2652592093878666e-05, "epoch": 1.22559876384239, "percentage": 70.0, "elapsed_time": "8:50:24", "remaining_time": "3:47:19", "throughput": 874.5, "total_tokens": 27830704}
{"current_steps": 2385, "total_steps": 3400, "loss": 0.3819, "lr": 2.244935311899829e-05, "epoch": 1.228174092196755, "percentage": 70.15, "elapsed_time": "8:51:27", "remaining_time": "3:46:10", "throughput": 874.6, "total_tokens": 27889160}
{"current_steps": 2390, "total_steps": 3400, "loss": 0.3084, "lr": 2.224676572098007e-05, "epoch": 1.2307494205511202, "percentage": 70.29, "elapsed_time": "8:52:31", "remaining_time": "3:45:02", "throughput": 874.7, "total_tokens": 27947608}
{"current_steps": 2395, "total_steps": 3400, "loss": 0.4267, "lr": 2.2044834691045873e-05, "epoch": 1.2333247489054855, "percentage": 70.44, "elapsed_time": "8:53:34", "remaining_time": "3:43:53", "throughput": 874.81, "total_tokens": 28006112}
{"current_steps": 2400, "total_steps": 3400, "loss": 0.3486, "lr": 2.184356480489432e-05, "epoch": 1.2359000772598505, "percentage": 70.59, "elapsed_time": "8:54:37", "remaining_time": "3:42:45", "throughput": 874.91, "total_tokens": 28064552}
{"current_steps": 2400, "total_steps": 3400, "eval_loss": 0.7410638928413391, "epoch": 1.2359000772598505, "percentage": 70.59, "elapsed_time": "8:54:53", "remaining_time": "3:42:52", "throughput": 874.47, "total_tokens": 28064552}
{"current_steps": 2405, "total_steps": 3400, "loss": 0.2416, "lr": 2.1642960822587878e-05, "epoch": 1.2384754056142158, "percentage": 70.74, "elapsed_time": "8:56:01", "remaining_time": "3:41:45", "throughput": 874.43, "total_tokens": 28123016}
{"current_steps": 2410, "total_steps": 3400, "loss": 0.3007, "lr": 2.1443027488440338e-05, "epoch": 1.241050733968581, "percentage": 70.88, "elapsed_time": "8:57:04", "remaining_time": "3:40:37", "throughput": 874.53, "total_tokens": 28181464}
{"current_steps": 2415, "total_steps": 3400, "loss": 0.2655, "lr": 2.124376953090456e-05, "epoch": 1.2436260623229463, "percentage": 71.03, "elapsed_time": "8:58:08", "remaining_time": "3:39:29", "throughput": 874.63, "total_tokens": 28239920}
{"current_steps": 2420, "total_steps": 3400, "loss": 0.3075, "lr": 2.104519166246059e-05, "epoch": 1.2462013906773113, "percentage": 71.18, "elapsed_time": "8:59:11", "remaining_time": "3:38:20", "throughput": 874.73, "total_tokens": 28298432}
{"current_steps": 2425, "total_steps": 3400, "loss": 0.3537, "lr": 2.0847298579504344e-05, "epoch": 1.2487767190316765, "percentage": 71.32, "elapsed_time": "9:00:14", "remaining_time": "3:37:12", "throughput": 874.83, "total_tokens": 28356904}
{"current_steps": 2430, "total_steps": 3400, "loss": 0.2993, "lr": 2.065009496223638e-05, "epoch": 1.2513520473860418, "percentage": 71.47, "elapsed_time": "9:01:17", "remaining_time": "3:36:04", "throughput": 874.93, "total_tokens": 28415384}
{"current_steps": 2435, "total_steps": 3400, "loss": 0.2752, "lr": 2.045358547455138e-05, "epoch": 1.2539273757404068, "percentage": 71.62, "elapsed_time": "9:02:20", "remaining_time": "3:34:56", "throughput": 875.02, "total_tokens": 28473848}
{"current_steps": 2440, "total_steps": 3400, "loss": 0.2975, "lr": 2.0257774763927655e-05, "epoch": 1.256502704094772, "percentage": 71.76, "elapsed_time": "9:03:23", "remaining_time": "3:33:47", "throughput": 875.12, "total_tokens": 28532312}
{"current_steps": 2445, "total_steps": 3400, "loss": 0.4051, "lr": 2.0062667461317426e-05, "epoch": 1.2590780324491373, "percentage": 71.91, "elapsed_time": "9:04:27", "remaining_time": "3:32:39", "throughput": 875.22, "total_tokens": 28590784}
{"current_steps": 2450, "total_steps": 3400, "loss": 0.2966, "lr": 1.9868268181037185e-05, "epoch": 1.2616533608035025, "percentage": 72.06, "elapsed_time": "9:05:30", "remaining_time": "3:31:31", "throughput": 875.32, "total_tokens": 28649256}
{"current_steps": 2450, "total_steps": 3400, "eval_loss": 0.7485548853874207, "epoch": 1.2616533608035025, "percentage": 72.06, "elapsed_time": "9:05:46", "remaining_time": "3:31:37", "throughput": 874.89, "total_tokens": 28649256}
{"current_steps": 2455, "total_steps": 3400, "loss": 0.2664, "lr": 1.967458152065857e-05, "epoch": 1.2642286891578676, "percentage": 72.21, "elapsed_time": "9:06:54", "remaining_time": "3:30:31", "throughput": 874.86, "total_tokens": 28707736}
{"current_steps": 2460, "total_steps": 3400, "loss": 0.3692, "lr": 1.9481612060899646e-05, "epoch": 1.2668040175122328, "percentage": 72.35, "elapsed_time": "9:07:57", "remaining_time": "3:29:22", "throughput": 874.96, "total_tokens": 28766232}
{"current_steps": 2465, "total_steps": 3400, "loss": 0.315, "lr": 1.928936436551661e-05, "epoch": 1.269379345866598, "percentage": 72.5, "elapsed_time": "9:09:00", "remaining_time": "3:28:14", "throughput": 875.06, "total_tokens": 28824688}
{"current_steps": 2470, "total_steps": 3400, "loss": 0.3536, "lr": 1.9097842981195834e-05, "epoch": 1.271954674220963, "percentage": 72.65, "elapsed_time": "9:10:03", "remaining_time": "3:27:06", "throughput": 875.15, "total_tokens": 28883176}
{"current_steps": 2475, "total_steps": 3400, "loss": 0.3143, "lr": 1.8907052437446272e-05, "epoch": 1.2745300025753283, "percentage": 72.79, "elapsed_time": "9:11:06", "remaining_time": "3:25:58", "throughput": 875.25, "total_tokens": 28941592}
{"current_steps": 2480, "total_steps": 3400, "loss": 0.3114, "lr": 1.871699724649244e-05, "epoch": 1.2771053309296936, "percentage": 72.94, "elapsed_time": "9:12:09", "remaining_time": "3:24:50", "throughput": 875.34, "total_tokens": 29000064}
{"current_steps": 2485, "total_steps": 3400, "loss": 0.3327, "lr": 1.8527681903167644e-05, "epoch": 1.2796806592840588, "percentage": 73.09, "elapsed_time": "9:13:13", "remaining_time": "3:23:42", "throughput": 875.44, "total_tokens": 29058496}
{"current_steps": 2490, "total_steps": 3400, "loss": 0.2543, "lr": 1.833911088480767e-05, "epoch": 1.2822559876384239, "percentage": 73.24, "elapsed_time": "9:14:16", "remaining_time": "3:22:33", "throughput": 875.53, "total_tokens": 29116992}
{"current_steps": 2495, "total_steps": 3400, "loss": 0.2854, "lr": 1.8151288651144893e-05, "epoch": 1.284831315992789, "percentage": 73.38, "elapsed_time": "9:15:19", "remaining_time": "3:21:25", "throughput": 875.63, "total_tokens": 29175496}
{"current_steps": 2500, "total_steps": 3400, "loss": 0.3221, "lr": 1.796421964420285e-05, "epoch": 1.2874066443471541, "percentage": 73.53, "elapsed_time": "9:16:22", "remaining_time": "3:20:17", "throughput": 875.72, "total_tokens": 29233968}
{"current_steps": 2500, "total_steps": 3400, "eval_loss": 0.7222262620925903, "epoch": 1.2874066443471541, "percentage": 73.53, "elapsed_time": "9:16:38", "remaining_time": "3:20:23", "throughput": 875.3, "total_tokens": 29233968}
{"current_steps": 2505, "total_steps": 3400, "loss": 0.2344, "lr": 1.7777908288191176e-05, "epoch": 1.2899819727015194, "percentage": 73.68, "elapsed_time": "9:17:46", "remaining_time": "3:19:17", "throughput": 875.27, "total_tokens": 29292464}
{"current_steps": 2510, "total_steps": 3400, "loss": 0.2727, "lr": 1.7592358989400883e-05, "epoch": 1.2925573010558846, "percentage": 73.82, "elapsed_time": "9:18:50", "remaining_time": "3:18:09", "throughput": 875.36, "total_tokens": 29350952}
{"current_steps": 2515, "total_steps": 3400, "loss": 0.2687, "lr": 1.740757613610028e-05, "epoch": 1.2951326294102499, "percentage": 73.97, "elapsed_time": "9:19:53", "remaining_time": "3:17:01", "throughput": 875.46, "total_tokens": 29409432}
{"current_steps": 2520, "total_steps": 3400, "loss": 0.2632, "lr": 1.7223564098431067e-05, "epoch": 1.2977079577646151, "percentage": 74.12, "elapsed_time": "9:20:56", "remaining_time": "3:15:52", "throughput": 875.56, "total_tokens": 29467880}
{"current_steps": 2525, "total_steps": 3400, "loss": 0.3057, "lr": 1.704032722830512e-05, "epoch": 1.3002832861189801, "percentage": 74.26, "elapsed_time": "9:21:59", "remaining_time": "3:14:44", "throughput": 875.65, "total_tokens": 29526384}
{"current_steps": 2530, "total_steps": 3400, "loss": 0.3054, "lr": 1.68578698593014e-05, "epoch": 1.3028586144733454, "percentage": 74.41, "elapsed_time": "9:23:03", "remaining_time": "3:13:37", "throughput": 875.72, "total_tokens": 29584880}
{"current_steps": 2535, "total_steps": 3400, "loss": 0.28, "lr": 1.6676196306563613e-05, "epoch": 1.3054339428277104, "percentage": 74.56, "elapsed_time": "9:24:07", "remaining_time": "3:12:29", "throughput": 875.8, "total_tokens": 29643344}
{"current_steps": 2540, "total_steps": 3400, "loss": 0.3169, "lr": 1.6495310866698093e-05, "epoch": 1.3080092711820757, "percentage": 74.71, "elapsed_time": "9:25:11", "remaining_time": "3:11:21", "throughput": 875.87, "total_tokens": 29701864}
{"current_steps": 2545, "total_steps": 3400, "loss": 0.2985, "lr": 1.631521781767214e-05, "epoch": 1.310584599536441, "percentage": 74.85, "elapsed_time": "9:26:15", "remaining_time": "3:10:14", "throughput": 875.94, "total_tokens": 29760376}
{"current_steps": 2550, "total_steps": 3400, "loss": 0.3231, "lr": 1.6135921418712956e-05, "epoch": 1.3131599278908062, "percentage": 75.0, "elapsed_time": "9:27:19", "remaining_time": "3:09:06", "throughput": 876.02, "total_tokens": 29818856}
{"current_steps": 2550, "total_steps": 3400, "eval_loss": 0.7146337628364563, "epoch": 1.3131599278908062, "percentage": 75.0, "elapsed_time": "9:27:35", "remaining_time": "3:09:11", "throughput": 875.6, "total_tokens": 29818856}
{"current_steps": 2555, "total_steps": 3400, "loss": 0.2689, "lr": 1.5957425910206785e-05, "epoch": 1.3157352562451712, "percentage": 75.15, "elapsed_time": "9:28:44", "remaining_time": "3:08:05", "throughput": 875.55, "total_tokens": 29877288}
{"current_steps": 2560, "total_steps": 3400, "loss": 0.3889, "lr": 1.577973551359877e-05, "epoch": 1.3183105845995364, "percentage": 75.29, "elapsed_time": "9:29:47", "remaining_time": "3:06:57", "throughput": 875.63, "total_tokens": 29935776}
{"current_steps": 2565, "total_steps": 3400, "loss": 0.2489, "lr": 1.560285443129296e-05, "epoch": 1.3208859129539017, "percentage": 75.44, "elapsed_time": "9:30:52", "remaining_time": "3:05:50", "throughput": 875.68, "total_tokens": 29994232}
{"current_steps": 2570, "total_steps": 3400, "loss": 0.3016, "lr": 1.542678684655306e-05, "epoch": 1.3234612413082667, "percentage": 75.59, "elapsed_time": "9:31:56", "remaining_time": "3:04:42", "throughput": 875.76, "total_tokens": 30052760}
{"current_steps": 2575, "total_steps": 3400, "loss": 0.3063, "lr": 1.5251536923403426e-05, "epoch": 1.326036569662632, "percentage": 75.74, "elapsed_time": "9:32:59", "remaining_time": "3:03:34", "throughput": 875.84, "total_tokens": 30111200}
{"current_steps": 2580, "total_steps": 3400, "loss": 0.3159, "lr": 1.5077108806530581e-05, "epoch": 1.3286118980169972, "percentage": 75.88, "elapsed_time": "9:34:03", "remaining_time": "3:02:27", "throughput": 875.91, "total_tokens": 30169680}
{"current_steps": 2585, "total_steps": 3400, "loss": 0.3752, "lr": 1.4903506621185192e-05, "epoch": 1.3311872263713624, "percentage": 76.03, "elapsed_time": "9:35:07", "remaining_time": "3:01:19", "throughput": 875.98, "total_tokens": 30228176}
{"current_steps": 2590, "total_steps": 3400, "loss": 0.3207, "lr": 1.4730734473084568e-05, "epoch": 1.3337625547257275, "percentage": 76.18, "elapsed_time": "9:36:11", "remaining_time": "3:00:11", "throughput": 876.06, "total_tokens": 30286656}
{"current_steps": 2595, "total_steps": 3400, "loss": 0.2928, "lr": 1.4558796448315504e-05, "epoch": 1.3363378830800927, "percentage": 76.32, "elapsed_time": "9:37:15", "remaining_time": "2:59:04", "throughput": 876.13, "total_tokens": 30345160}
{"current_steps": 2600, "total_steps": 3400, "loss": 0.2779, "lr": 1.4387696613237612e-05, "epoch": 1.338913211434458, "percentage": 76.47, "elapsed_time": "9:38:19", "remaining_time": "2:57:56", "throughput": 876.2, "total_tokens": 30403640}
{"current_steps": 2600, "total_steps": 3400, "eval_loss": 0.695651650428772, "epoch": 1.338913211434458, "percentage": 76.47, "elapsed_time": "9:38:35", "remaining_time": "2:58:01", "throughput": 875.79, "total_tokens": 30403640}
{"current_steps": 2605, "total_steps": 3400, "loss": 0.3037, "lr": 1.4217439014387251e-05, "epoch": 1.341488539788823, "percentage": 76.62, "elapsed_time": "9:39:44", "remaining_time": "2:56:55", "throughput": 875.74, "total_tokens": 30462128}
{"current_steps": 2610, "total_steps": 3400, "loss": 0.2905, "lr": 1.404802767838176e-05, "epoch": 1.3440638681431882, "percentage": 76.76, "elapsed_time": "9:40:48", "remaining_time": "2:55:47", "throughput": 875.81, "total_tokens": 30520616}
{"current_steps": 2615, "total_steps": 3400, "loss": 0.317, "lr": 1.3879466611824199e-05, "epoch": 1.3466391964975535, "percentage": 76.91, "elapsed_time": "9:41:51", "remaining_time": "2:54:40", "throughput": 875.89, "total_tokens": 30579024}
{"current_steps": 2620, "total_steps": 3400, "loss": 0.2794, "lr": 1.371175980120864e-05, "epoch": 1.3492145248519187, "percentage": 77.06, "elapsed_time": "9:42:55", "remaining_time": "2:53:32", "throughput": 875.96, "total_tokens": 30637464}
{"current_steps": 2625, "total_steps": 3400, "loss": 0.3056, "lr": 1.3544911212825906e-05, "epoch": 1.3517898532062838, "percentage": 77.21, "elapsed_time": "9:43:59", "remaining_time": "2:52:25", "throughput": 876.03, "total_tokens": 30695936}
{"current_steps": 2630, "total_steps": 3400, "loss": 0.2712, "lr": 1.337892479266974e-05, "epoch": 1.354365181560649, "percentage": 77.35, "elapsed_time": "9:45:03", "remaining_time": "2:51:17", "throughput": 876.11, "total_tokens": 30754408}
{"current_steps": 2635, "total_steps": 3400, "loss": 0.2615, "lr": 1.3213804466343421e-05, "epoch": 1.356940509915014, "percentage": 77.5, "elapsed_time": "9:46:06", "remaining_time": "2:50:09", "throughput": 876.2, "total_tokens": 30812848}
{"current_steps": 2640, "total_steps": 3400, "loss": 0.2661, "lr": 1.3049554138967051e-05, "epoch": 1.3595158382693793, "percentage": 77.65, "elapsed_time": "9:47:09", "remaining_time": "2:49:01", "throughput": 876.28, "total_tokens": 30871344}
{"current_steps": 2645, "total_steps": 3400, "loss": 0.3272, "lr": 1.2886177695085078e-05, "epoch": 1.3620911666237445, "percentage": 77.79, "elapsed_time": "9:48:12", "remaining_time": "2:47:54", "throughput": 876.38, "total_tokens": 30929824}
{"current_steps": 2650, "total_steps": 3400, "loss": 0.2962, "lr": 1.2723678998574512e-05, "epoch": 1.3646664949781098, "percentage": 77.94, "elapsed_time": "9:49:16", "remaining_time": "2:46:46", "throughput": 876.46, "total_tokens": 30988344}
{"current_steps": 2650, "total_steps": 3400, "eval_loss": 0.7657458186149597, "epoch": 1.3646664949781098, "percentage": 77.94, "elapsed_time": "9:49:32", "remaining_time": "2:46:51", "throughput": 876.06, "total_tokens": 30988344}
{"current_steps": 2655, "total_steps": 3400, "loss": 0.3207, "lr": 1.2562061892553473e-05, "epoch": 1.367241823332475, "percentage": 78.09, "elapsed_time": "9:50:40", "remaining_time": "2:45:44", "throughput": 876.02, "total_tokens": 31046848}
{"current_steps": 2660, "total_steps": 3400, "loss": 0.3001, "lr": 1.2401330199290367e-05, "epoch": 1.36981715168684, "percentage": 78.24, "elapsed_time": "9:51:44", "remaining_time": "2:44:37", "throughput": 876.11, "total_tokens": 31105352}
{"current_steps": 2665, "total_steps": 3400, "loss": 0.2858, "lr": 1.224148772011346e-05, "epoch": 1.3723924800412053, "percentage": 78.38, "elapsed_time": "9:52:48", "remaining_time": "2:43:29", "throughput": 876.18, "total_tokens": 31163848}
{"current_steps": 2670, "total_steps": 3400, "loss": 0.2338, "lr": 1.2082538235320929e-05, "epoch": 1.3749678083955703, "percentage": 78.53, "elapsed_time": "9:53:51", "remaining_time": "2:42:22", "throughput": 876.25, "total_tokens": 31222360}
{"current_steps": 2675, "total_steps": 3400, "loss": 0.2212, "lr": 1.1924485504091565e-05, "epoch": 1.3775431367499356, "percentage": 78.68, "elapsed_time": "9:54:55", "remaining_time": "2:41:14", "throughput": 876.32, "total_tokens": 31280840}
{"current_steps": 2680, "total_steps": 3400, "loss": 0.3131, "lr": 1.1767333264395736e-05, "epoch": 1.3801184651043008, "percentage": 78.82, "elapsed_time": "9:56:00", "remaining_time": "2:40:07", "throughput": 876.37, "total_tokens": 31339264}
{"current_steps": 2685, "total_steps": 3400, "loss": 0.3616, "lr": 1.1611085232907132e-05, "epoch": 1.382693793458666, "percentage": 78.97, "elapsed_time": "9:57:04", "remaining_time": "2:38:59", "throughput": 876.42, "total_tokens": 31397744}
{"current_steps": 2690, "total_steps": 3400, "loss": 0.3153, "lr": 1.14557451049147e-05, "epoch": 1.385269121813031, "percentage": 79.12, "elapsed_time": "9:58:09", "remaining_time": "2:37:52", "throughput": 876.48, "total_tokens": 31456240}
{"current_steps": 2695, "total_steps": 3400, "loss": 0.3044, "lr": 1.1301316554235397e-05, "epoch": 1.3878444501673963, "percentage": 79.26, "elapsed_time": "9:59:13", "remaining_time": "2:36:45", "throughput": 876.53, "total_tokens": 31514744}
{"current_steps": 2700, "total_steps": 3400, "loss": 0.3163, "lr": 1.114780323312724e-05, "epoch": 1.3904197785217616, "percentage": 79.41, "elapsed_time": "10:00:18", "remaining_time": "2:35:38", "throughput": 876.58, "total_tokens": 31573240}
{"current_steps": 2700, "total_steps": 3400, "eval_loss": 0.7473158240318298, "epoch": 1.3904197785217616, "percentage": 79.41, "elapsed_time": "10:00:34", "remaining_time": "2:35:42", "throughput": 876.19, "total_tokens": 31573240}
{"current_steps": 2705, "total_steps": 3400, "loss": 0.2798, "lr": 1.0995208772202897e-05, "epoch": 1.3929951068761266, "percentage": 79.56, "elapsed_time": "10:01:44", "remaining_time": "2:34:36", "throughput": 876.12, "total_tokens": 31631688}
{"current_steps": 2710, "total_steps": 3400, "loss": 0.289, "lr": 1.0843536780343865e-05, "epoch": 1.3955704352304918, "percentage": 79.71, "elapsed_time": "10:02:48", "remaining_time": "2:33:29", "throughput": 876.17, "total_tokens": 31690200}
{"current_steps": 2715, "total_steps": 3400, "loss": 0.2844, "lr": 1.069279084461513e-05, "epoch": 1.398145763584857, "percentage": 79.85, "elapsed_time": "10:03:53", "remaining_time": "2:32:21", "throughput": 876.23, "total_tokens": 31748664}
{"current_steps": 2720, "total_steps": 3400, "loss": 0.3254, "lr": 1.0542974530180327e-05, "epoch": 1.4007210919392223, "percentage": 80.0, "elapsed_time": "10:04:57", "remaining_time": "2:31:14", "throughput": 876.3, "total_tokens": 31807176}
{"current_steps": 2725, "total_steps": 3400, "loss": 0.3683, "lr": 1.0394091380217352e-05, "epoch": 1.4032964202935874, "percentage": 80.15, "elapsed_time": "10:06:01", "remaining_time": "2:30:06", "throughput": 876.36, "total_tokens": 31865696}
{"current_steps": 2730, "total_steps": 3400, "loss": 0.2968, "lr": 1.0246144915834683e-05, "epoch": 1.4058717486479526, "percentage": 80.29, "elapsed_time": "10:07:05", "remaining_time": "2:28:59", "throughput": 876.42, "total_tokens": 31924200}
{"current_steps": 2735, "total_steps": 3400, "loss": 0.2943, "lr": 1.0099138635988026e-05, "epoch": 1.4084470770023179, "percentage": 80.44, "elapsed_time": "10:08:09", "remaining_time": "2:27:52", "throughput": 876.48, "total_tokens": 31982712}
{"current_steps": 2740, "total_steps": 3400, "loss": 0.3037, "lr": 9.953076017397578e-06, "epoch": 1.4110224053566829, "percentage": 80.59, "elapsed_time": "10:09:14", "remaining_time": "2:26:45", "throughput": 876.54, "total_tokens": 32041176}
{"current_steps": 2745, "total_steps": 3400, "loss": 0.3019, "lr": 9.807960514465792e-06, "epoch": 1.4135977337110481, "percentage": 80.74, "elapsed_time": "10:10:18", "remaining_time": "2:25:37", "throughput": 876.6, "total_tokens": 32099656}
{"current_steps": 2750, "total_steps": 3400, "loss": 0.164, "lr": 9.663795559195733e-06, "epoch": 1.4161730620654134, "percentage": 80.88, "elapsed_time": "10:11:22", "remaining_time": "2:24:30", "throughput": 876.65, "total_tokens": 32158144}
{"current_steps": 2750, "total_steps": 3400, "eval_loss": 0.7807286381721497, "epoch": 1.4161730620654134, "percentage": 80.88, "elapsed_time": "10:11:39", "remaining_time": "2:24:34", "throughput": 876.27, "total_tokens": 32158144}
{"current_steps": 2755, "total_steps": 3400, "loss": 0.3333, "lr": 9.520584561109864e-06, "epoch": 1.4187483904197786, "percentage": 81.03, "elapsed_time": "10:12:47", "remaining_time": "2:23:28", "throughput": 876.22, "total_tokens": 32216656}
{"current_steps": 2760, "total_steps": 3400, "loss": 0.2993, "lr": 9.378330907169386e-06, "epoch": 1.4213237187741437, "percentage": 81.18, "elapsed_time": "10:13:51", "remaining_time": "2:22:20", "throughput": 876.28, "total_tokens": 32275168}
{"current_steps": 2765, "total_steps": 3400, "loss": 0.2683, "lr": 9.237037961694223e-06, "epoch": 1.423899047128509, "percentage": 81.32, "elapsed_time": "10:14:56", "remaining_time": "2:21:13", "throughput": 876.34, "total_tokens": 32333664}
{"current_steps": 2770, "total_steps": 3400, "loss": 0.3145, "lr": 9.096709066283354e-06, "epoch": 1.4264743754828741, "percentage": 81.47, "elapsed_time": "10:16:00", "remaining_time": "2:20:06", "throughput": 876.39, "total_tokens": 32392088}
{"current_steps": 2775, "total_steps": 3400, "loss": 0.3092, "lr": 8.957347539735872e-06, "epoch": 1.4290497038372392, "percentage": 81.62, "elapsed_time": "10:17:05", "remaining_time": "2:18:58", "throughput": 876.45, "total_tokens": 32450584}
{"current_steps": 2780, "total_steps": 3400, "loss": 0.2993, "lr": 8.818956677972406e-06, "epoch": 1.4316250321916044, "percentage": 81.76, "elapsed_time": "10:18:09", "remaining_time": "2:17:51", "throughput": 876.51, "total_tokens": 32509096}
{"current_steps": 2785, "total_steps": 3400, "loss": 0.326, "lr": 8.681539753957269e-06, "epoch": 1.4342003605459697, "percentage": 81.91, "elapsed_time": "10:19:13", "remaining_time": "2:16:44", "throughput": 876.57, "total_tokens": 32567560}
{"current_steps": 2790, "total_steps": 3400, "loss": 0.2494, "lr": 8.545100017620988e-06, "epoch": 1.436775688900335, "percentage": 82.06, "elapsed_time": "10:20:17", "remaining_time": "2:15:37", "throughput": 876.63, "total_tokens": 32626056}
{"current_steps": 2795, "total_steps": 3400, "loss": 0.2691, "lr": 8.409640695783443e-06, "epoch": 1.4393510172547, "percentage": 82.21, "elapsed_time": "10:21:21", "remaining_time": "2:14:29", "throughput": 876.69, "total_tokens": 32684520}
{"current_steps": 2800, "total_steps": 3400, "loss": 0.2939, "lr": 8.275164992077556e-06, "epoch": 1.4419263456090652, "percentage": 82.35, "elapsed_time": "10:22:25", "remaining_time": "2:13:22", "throughput": 876.75, "total_tokens": 32743032}
{"current_steps": 2800, "total_steps": 3400, "eval_loss": 0.791334331035614, "epoch": 1.4419263456090652, "percentage": 82.35, "elapsed_time": "10:22:41", "remaining_time": "2:13:26", "throughput": 876.37, "total_tokens": 32743032}
{"current_steps": 2805, "total_steps": 3400, "loss": 0.2974, "lr": 8.141676086873572e-06, "epoch": 1.4445016739634302, "percentage": 82.5, "elapsed_time": "10:23:50", "remaining_time": "2:12:19", "throughput": 876.33, "total_tokens": 32801504}
{"current_steps": 2810, "total_steps": 3400, "loss": 0.2849, "lr": 8.009177137203794e-06, "epoch": 1.4470770023177955, "percentage": 82.65, "elapsed_time": "10:24:53", "remaining_time": "2:11:12", "throughput": 876.41, "total_tokens": 32860032}
{"current_steps": 2815, "total_steps": 3400, "loss": 0.3024, "lr": 7.877671276687898e-06, "epoch": 1.4496523306721607, "percentage": 82.79, "elapsed_time": "10:25:56", "remaining_time": "2:10:04", "throughput": 876.49, "total_tokens": 32918472}
{"current_steps": 2820, "total_steps": 3400, "loss": 0.2565, "lr": 7.747161615458902e-06, "epoch": 1.452227659026526, "percentage": 82.94, "elapsed_time": "10:27:00", "remaining_time": "2:08:57", "throughput": 876.57, "total_tokens": 32976944}
{"current_steps": 2825, "total_steps": 3400, "loss": 0.2473, "lr": 7.617651240089546e-06, "epoch": 1.4548029873808912, "percentage": 83.09, "elapsed_time": "10:28:03", "remaining_time": "2:07:50", "throughput": 876.66, "total_tokens": 33035424}
{"current_steps": 2830, "total_steps": 3400, "loss": 0.3118, "lr": 7.489143213519301e-06, "epoch": 1.4573783157352562, "percentage": 83.24, "elapsed_time": "10:29:06", "remaining_time": "2:06:42", "throughput": 876.74, "total_tokens": 33093880}
{"current_steps": 2835, "total_steps": 3400, "loss": 0.2593, "lr": 7.361640574981937e-06, "epoch": 1.4599536440896215, "percentage": 83.38, "elapsed_time": "10:30:09", "remaining_time": "2:05:35", "throughput": 876.83, "total_tokens": 33152328}
{"current_steps": 2840, "total_steps": 3400, "loss": 0.284, "lr": 7.2351463399336735e-06, "epoch": 1.4625289724439865, "percentage": 83.53, "elapsed_time": "10:31:12", "remaining_time": "2:04:27", "throughput": 876.91, "total_tokens": 33210816}
{"current_steps": 2845, "total_steps": 3400, "loss": 0.2671, "lr": 7.109663499981834e-06, "epoch": 1.4651043007983517, "percentage": 83.68, "elapsed_time": "10:32:15", "remaining_time": "2:03:20", "throughput": 877.0, "total_tokens": 33269320}
{"current_steps": 2850, "total_steps": 3400, "loss": 0.2848, "lr": 6.985195022814067e-06, "epoch": 1.467679629152717, "percentage": 83.82, "elapsed_time": "10:33:18", "remaining_time": "2:02:13", "throughput": 877.07, "total_tokens": 33327720}
{"current_steps": 2850, "total_steps": 3400, "eval_loss": 0.8045337796211243, "epoch": 1.467679629152717, "percentage": 83.82, "elapsed_time": "10:33:34", "remaining_time": "2:02:16", "throughput": 876.71, "total_tokens": 33327720}
{"current_steps": 2855, "total_steps": 3400, "loss": 0.2811, "lr": 6.861743852128233e-06, "epoch": 1.4702549575070822, "percentage": 83.97, "elapsed_time": "10:34:42", "remaining_time": "2:01:09", "throughput": 876.68, "total_tokens": 33386160}
{"current_steps": 2860, "total_steps": 3400, "loss": 0.2394, "lr": 6.7393129075627335e-06, "epoch": 1.4728302858614473, "percentage": 84.12, "elapsed_time": "10:35:45", "remaining_time": "2:00:02", "throughput": 876.76, "total_tokens": 33444648}
{"current_steps": 2865, "total_steps": 3400, "loss": 0.243, "lr": 6.6179050846274515e-06, "epoch": 1.4754056142158125, "percentage": 84.26, "elapsed_time": "10:36:49", "remaining_time": "1:58:55", "throughput": 876.84, "total_tokens": 33503144}
{"current_steps": 2870, "total_steps": 3400, "loss": 0.242, "lr": 6.497523254635296e-06, "epoch": 1.4779809425701778, "percentage": 84.41, "elapsed_time": "10:37:52", "remaining_time": "1:57:47", "throughput": 876.92, "total_tokens": 33561600}
{"current_steps": 2875, "total_steps": 3400, "loss": 0.1864, "lr": 6.37817026463432e-06, "epoch": 1.4805562709245428, "percentage": 84.56, "elapsed_time": "10:38:55", "remaining_time": "1:56:40", "throughput": 877.0, "total_tokens": 33620056}
{"current_steps": 2880, "total_steps": 3400, "loss": 0.2406, "lr": 6.25984893734034e-06, "epoch": 1.483131599278908, "percentage": 84.71, "elapsed_time": "10:39:58", "remaining_time": "1:55:32", "throughput": 877.09, "total_tokens": 33678512}
{"current_steps": 2885, "total_steps": 3400, "loss": 0.2287, "lr": 6.142562071070179e-06, "epoch": 1.4857069276332733, "percentage": 84.85, "elapsed_time": "10:41:01", "remaining_time": "1:54:25", "throughput": 877.17, "total_tokens": 33736960}
{"current_steps": 2890, "total_steps": 3400, "loss": 0.2643, "lr": 6.026312439675552e-06, "epoch": 1.4882822559876385, "percentage": 85.0, "elapsed_time": "10:42:04", "remaining_time": "1:53:18", "throughput": 877.25, "total_tokens": 33795416}
{"current_steps": 2895, "total_steps": 3400, "loss": 0.2956, "lr": 5.911102792477357e-06, "epoch": 1.4908575843420036, "percentage": 85.15, "elapsed_time": "10:43:07", "remaining_time": "1:52:11", "throughput": 877.32, "total_tokens": 33853936}
{"current_steps": 2900, "total_steps": 3400, "loss": 0.29, "lr": 5.796935854200763e-06, "epoch": 1.4934329126963688, "percentage": 85.29, "elapsed_time": "10:44:10", "remaining_time": "1:51:03", "throughput": 877.41, "total_tokens": 33912440}
{"current_steps": 2900, "total_steps": 3400, "eval_loss": 0.8113046884536743, "epoch": 1.4934329126963688, "percentage": 85.29, "elapsed_time": "10:44:26", "remaining_time": "1:51:06", "throughput": 877.04, "total_tokens": 33912440}
{"current_steps": 2905, "total_steps": 3400, "loss": 0.2815, "lr": 5.683814324910685e-06, "epoch": 1.496008241050734, "percentage": 85.44, "elapsed_time": "10:45:35", "remaining_time": "1:50:00", "throughput": 877.0, "total_tokens": 33970888}
{"current_steps": 2910, "total_steps": 3400, "loss": 0.2737, "lr": 5.571740879947979e-06, "epoch": 1.498583569405099, "percentage": 85.59, "elapsed_time": "10:46:39", "remaining_time": "1:48:53", "throughput": 877.06, "total_tokens": 34029376}
{"current_steps": 2915, "total_steps": 3400, "loss": 0.2445, "lr": 5.4607181698661634e-06, "epoch": 1.5011588977594643, "percentage": 85.74, "elapsed_time": "10:47:43", "remaining_time": "1:47:46", "throughput": 877.12, "total_tokens": 34087864}
{"current_steps": 2920, "total_steps": 3400, "loss": 0.2802, "lr": 5.35074882036869e-06, "epoch": 1.5037342261138296, "percentage": 85.88, "elapsed_time": "10:48:47", "remaining_time": "1:46:39", "throughput": 877.17, "total_tokens": 34146296}
{"current_steps": 2925, "total_steps": 3400, "loss": 0.2379, "lr": 5.241835432246889e-06, "epoch": 1.5063095544681948, "percentage": 86.03, "elapsed_time": "10:49:52", "remaining_time": "1:45:32", "throughput": 877.22, "total_tokens": 34204800}
{"current_steps": 2930, "total_steps": 3400, "loss": 0.2783, "lr": 5.133980581318459e-06, "epoch": 1.5088848828225598, "percentage": 86.18, "elapsed_time": "10:50:56", "remaining_time": "1:44:25", "throughput": 877.27, "total_tokens": 34263296}
{"current_steps": 2935, "total_steps": 3400, "loss": 0.2609, "lr": 5.027186818366542e-06, "epoch": 1.511460211176925, "percentage": 86.32, "elapsed_time": "10:52:00", "remaining_time": "1:43:18", "throughput": 877.33, "total_tokens": 34321792}
{"current_steps": 2940, "total_steps": 3400, "loss": 0.2367, "lr": 4.921456669079366e-06, "epoch": 1.51403553953129, "percentage": 86.47, "elapsed_time": "10:53:05", "remaining_time": "1:42:11", "throughput": 877.38, "total_tokens": 34380264}
{"current_steps": 2945, "total_steps": 3400, "loss": 0.3644, "lr": 4.816792633990569e-06, "epoch": 1.5166108678856554, "percentage": 86.62, "elapsed_time": "10:54:09", "remaining_time": "1:41:04", "throughput": 877.43, "total_tokens": 34438752}
{"current_steps": 2950, "total_steps": 3400, "loss": 0.2494, "lr": 4.713197188420026e-06, "epoch": 1.5191861962400206, "percentage": 86.76, "elapsed_time": "10:55:14", "remaining_time": "1:39:57", "throughput": 877.48, "total_tokens": 34497216}
{"current_steps": 2950, "total_steps": 3400, "eval_loss": 0.8177086710929871, "epoch": 1.5191861962400206, "percentage": 86.76, "elapsed_time": "10:55:30", "remaining_time": "1:39:59", "throughput": 877.12, "total_tokens": 34497216}
{"current_steps": 2955, "total_steps": 3400, "loss": 0.2892, "lr": 4.610672782415276e-06, "epoch": 1.5217615245943859, "percentage": 86.91, "elapsed_time": "10:56:39", "remaining_time": "1:38:53", "throughput": 877.06, "total_tokens": 34555704}
{"current_steps": 2960, "total_steps": 3400, "loss": 0.3006, "lr": 4.509221840693656e-06, "epoch": 1.524336852948751, "percentage": 87.06, "elapsed_time": "10:57:43", "remaining_time": "1:37:46", "throughput": 877.12, "total_tokens": 34614168}
{"current_steps": 2965, "total_steps": 3400, "loss": 0.2931, "lr": 4.408846762584901e-06, "epoch": 1.5269121813031161, "percentage": 87.21, "elapsed_time": "10:58:48", "remaining_time": "1:36:39", "throughput": 877.16, "total_tokens": 34672624}
{"current_steps": 2970, "total_steps": 3400, "loss": 0.2255, "lr": 4.309549921974421e-06, "epoch": 1.5294875096574814, "percentage": 87.35, "elapsed_time": "10:59:52", "remaining_time": "1:35:32", "throughput": 877.21, "total_tokens": 34731056}
{"current_steps": 2975, "total_steps": 3400, "loss": 0.2725, "lr": 4.2113336672471245e-06, "epoch": 1.5320628380118464, "percentage": 87.5, "elapsed_time": "11:00:56", "remaining_time": "1:34:25", "throughput": 877.26, "total_tokens": 34789552}
{"current_steps": 2980, "total_steps": 3400, "loss": 0.3158, "lr": 4.114200321231937e-06, "epoch": 1.5346381663662116, "percentage": 87.65, "elapsed_time": "11:02:01", "remaining_time": "1:33:18", "throughput": 877.31, "total_tokens": 34848064}
{"current_steps": 2985, "total_steps": 3400, "loss": 0.2562, "lr": 4.018152181146823e-06, "epoch": 1.537213494720577, "percentage": 87.79, "elapsed_time": "11:03:05", "remaining_time": "1:32:11", "throughput": 877.37, "total_tokens": 34906592}
{"current_steps": 2990, "total_steps": 3400, "loss": 0.2814, "lr": 3.923191518544434e-06, "epoch": 1.5397888230749421, "percentage": 87.94, "elapsed_time": "11:04:09", "remaining_time": "1:31:04", "throughput": 877.42, "total_tokens": 34965064}
{"current_steps": 2995, "total_steps": 3400, "loss": 0.2555, "lr": 3.829320579258466e-06, "epoch": 1.5423641514293074, "percentage": 88.09, "elapsed_time": "11:05:14", "remaining_time": "1:29:57", "throughput": 877.47, "total_tokens": 35023552}
{"current_steps": 3000, "total_steps": 3400, "loss": 0.2259, "lr": 3.7365415833504725e-06, "epoch": 1.5449394797836724, "percentage": 88.24, "elapsed_time": "11:06:19", "remaining_time": "1:28:50", "throughput": 877.5, "total_tokens": 35082056}
{"current_steps": 3000, "total_steps": 3400, "eval_loss": 0.8405727744102478, "epoch": 1.5449394797836724, "percentage": 88.24, "elapsed_time": "11:06:35", "remaining_time": "1:28:52", "throughput": 877.15, "total_tokens": 35082056}
{"current_steps": 3005, "total_steps": 3400, "loss": 0.2157, "lr": 3.644856725057405e-06, "epoch": 1.5475148081380374, "percentage": 88.38, "elapsed_time": "11:07:45", "remaining_time": "1:27:46", "throughput": 877.08, "total_tokens": 35140568}
{"current_steps": 3010, "total_steps": 3400, "loss": 0.2233, "lr": 3.554268172739661e-06, "epoch": 1.5500901364924027, "percentage": 88.53, "elapsed_time": "11:08:50", "remaining_time": "1:26:39", "throughput": 877.12, "total_tokens": 35199064}
{"current_steps": 3015, "total_steps": 3400, "loss": 0.2951, "lr": 3.4647780688298826e-06, "epoch": 1.552665464846768, "percentage": 88.68, "elapsed_time": "11:09:55", "remaining_time": "1:25:32", "throughput": 877.16, "total_tokens": 35257576}
{"current_steps": 3020, "total_steps": 3400, "loss": 0.2274, "lr": 3.376388529782215e-06, "epoch": 1.5552407932011332, "percentage": 88.82, "elapsed_time": "11:10:59", "remaining_time": "1:24:25", "throughput": 877.21, "total_tokens": 35316064}
{"current_steps": 3025, "total_steps": 3400, "loss": 0.2479, "lr": 3.2891016460222967e-06, "epoch": 1.5578161215554984, "percentage": 88.97, "elapsed_time": "11:12:04", "remaining_time": "1:23:18", "throughput": 877.25, "total_tokens": 35374504}
{"current_steps": 3030, "total_steps": 3400, "loss": 0.292, "lr": 3.2029194818977983e-06, "epoch": 1.5603914499098637, "percentage": 89.12, "elapsed_time": "11:13:08", "remaining_time": "1:22:11", "throughput": 877.31, "total_tokens": 35432984}
{"current_steps": 3035, "total_steps": 3400, "loss": 0.241, "lr": 3.117844075629617e-06, "epoch": 1.5629667782642287, "percentage": 89.26, "elapsed_time": "11:14:12", "remaining_time": "1:21:05", "throughput": 877.35, "total_tokens": 35491488}
{"current_steps": 3040, "total_steps": 3400, "loss": 0.228, "lr": 3.033877439263666e-06, "epoch": 1.5655421066185937, "percentage": 89.41, "elapsed_time": "11:15:17", "remaining_time": "1:19:58", "throughput": 877.41, "total_tokens": 35549984}
{"current_steps": 3045, "total_steps": 3400, "loss": 0.2485, "lr": 2.951021558623274e-06, "epoch": 1.568117434972959, "percentage": 89.56, "elapsed_time": "11:16:21", "remaining_time": "1:18:51", "throughput": 877.45, "total_tokens": 35608488}
{"current_steps": 3050, "total_steps": 3400, "loss": 0.2851, "lr": 2.869278393262226e-06, "epoch": 1.5706927633273242, "percentage": 89.71, "elapsed_time": "11:17:25", "remaining_time": "1:17:44", "throughput": 877.5, "total_tokens": 35666976}
{"current_steps": 3050, "total_steps": 3400, "eval_loss": 0.8473746180534363, "epoch": 1.5706927633273242, "percentage": 89.71, "elapsed_time": "11:17:41", "remaining_time": "1:17:46", "throughput": 877.16, "total_tokens": 35666976}
{"current_steps": 3055, "total_steps": 3400, "loss": 0.2514, "lr": 2.7886498764184588e-06, "epoch": 1.5732680916816895, "percentage": 89.85, "elapsed_time": "11:18:51", "remaining_time": "1:16:39", "throughput": 877.1, "total_tokens": 35725456}
{"current_steps": 3060, "total_steps": 3400, "loss": 0.3091, "lr": 2.7091379149682685e-06, "epoch": 1.5758434200360547, "percentage": 90.0, "elapsed_time": "11:19:55", "remaining_time": "1:15:32", "throughput": 877.15, "total_tokens": 35783912}
{"current_steps": 3065, "total_steps": 3400, "loss": 0.2629, "lr": 2.6307443893812843e-06, "epoch": 1.5784187483904197, "percentage": 90.15, "elapsed_time": "11:20:59", "remaining_time": "1:14:25", "throughput": 877.21, "total_tokens": 35842376}
{"current_steps": 3070, "total_steps": 3400, "loss": 0.3065, "lr": 2.5534711536759404e-06, "epoch": 1.580994076744785, "percentage": 90.29, "elapsed_time": "11:22:04", "remaining_time": "1:13:19", "throughput": 877.25, "total_tokens": 35900824}
{"current_steps": 3075, "total_steps": 3400, "loss": 0.2577, "lr": 2.4773200353756798e-06, "epoch": 1.58356940509915, "percentage": 90.44, "elapsed_time": "11:23:08", "remaining_time": "1:12:12", "throughput": 877.3, "total_tokens": 35959264}
{"current_steps": 3080, "total_steps": 3400, "loss": 0.2359, "lr": 2.4022928354656473e-06, "epoch": 1.5861447334535153, "percentage": 90.59, "elapsed_time": "11:24:12", "remaining_time": "1:11:05", "throughput": 877.35, "total_tokens": 36017760}
{"current_steps": 3085, "total_steps": 3400, "loss": 0.1897, "lr": 2.3283913283502044e-06, "epoch": 1.5887200618078805, "percentage": 90.74, "elapsed_time": "11:25:17", "remaining_time": "1:09:58", "throughput": 877.4, "total_tokens": 36076280}
{"current_steps": 3090, "total_steps": 3400, "loss": 0.286, "lr": 2.2556172618108997e-06, "epoch": 1.5912953901622457, "percentage": 90.88, "elapsed_time": "11:26:21", "remaining_time": "1:08:51", "throughput": 877.44, "total_tokens": 36134784}
{"current_steps": 3095, "total_steps": 3400, "loss": 0.2733, "lr": 2.183972356965125e-06, "epoch": 1.593870718516611, "percentage": 91.03, "elapsed_time": "11:27:26", "remaining_time": "1:07:44", "throughput": 877.49, "total_tokens": 36193288}
{"current_steps": 3100, "total_steps": 3400, "loss": 0.2351, "lr": 2.113458308225458e-06, "epoch": 1.596446046870976, "percentage": 91.18, "elapsed_time": "11:28:30", "remaining_time": "1:06:37", "throughput": 877.54, "total_tokens": 36251744}
{"current_steps": 3100, "total_steps": 3400, "eval_loss": 0.8650907874107361, "epoch": 1.596446046870976, "percentage": 91.18, "elapsed_time": "11:28:46", "remaining_time": "1:06:39", "throughput": 877.2, "total_tokens": 36251744}