| {"current_steps": 10, "total_steps": 2025, "loss": 0.5936, "learning_rate": 4.926108374384237e-07, "epoch": 0.014814814814814815, "percentage": 0.49, "elapsed_time": "0:00:52", "remaining_time": "2:57:46"} | |
| {"current_steps": 20, "total_steps": 2025, "loss": 0.3977, "learning_rate": 9.852216748768474e-07, "epoch": 0.02962962962962963, "percentage": 0.99, "elapsed_time": "0:01:41", "remaining_time": "2:49:57"} | |
| {"current_steps": 30, "total_steps": 2025, "loss": 0.2758, "learning_rate": 1.4778325123152712e-06, "epoch": 0.044444444444444446, "percentage": 1.48, "elapsed_time": "0:02:31", "remaining_time": "2:47:30"} | |
| {"current_steps": 40, "total_steps": 2025, "loss": 0.206, "learning_rate": 1.970443349753695e-06, "epoch": 0.05925925925925926, "percentage": 1.98, "elapsed_time": "0:03:15", "remaining_time": "2:41:25"} | |
| {"current_steps": 50, "total_steps": 2025, "loss": 0.2267, "learning_rate": 2.4630541871921186e-06, "epoch": 0.07407407407407407, "percentage": 2.47, "elapsed_time": "0:04:04", "remaining_time": "2:40:38"} | |
| {"current_steps": 60, "total_steps": 2025, "loss": 0.2088, "learning_rate": 2.9556650246305424e-06, "epoch": 0.08888888888888889, "percentage": 2.96, "elapsed_time": "0:04:53", "remaining_time": "2:40:09"} | |
| {"current_steps": 70, "total_steps": 2025, "loss": 0.197, "learning_rate": 3.448275862068966e-06, "epoch": 0.1037037037037037, "percentage": 3.46, "elapsed_time": "0:05:43", "remaining_time": "2:39:43"} | |
| {"current_steps": 80, "total_steps": 2025, "loss": 0.1934, "learning_rate": 3.94088669950739e-06, "epoch": 0.11851851851851852, "percentage": 3.95, "elapsed_time": "0:06:32", "remaining_time": "2:38:55"} | |
| {"current_steps": 90, "total_steps": 2025, "loss": 0.2117, "learning_rate": 4.4334975369458135e-06, "epoch": 0.13333333333333333, "percentage": 4.44, "elapsed_time": "0:07:21", "remaining_time": "2:38:07"} | |
| {"current_steps": 100, "total_steps": 2025, "loss": 0.2408, "learning_rate": 4.926108374384237e-06, "epoch": 0.14814814814814814, "percentage": 4.94, "elapsed_time": "0:08:09", "remaining_time": "2:36:54"} | |
| {"current_steps": 110, "total_steps": 2025, "loss": 0.2066, "learning_rate": 5.41871921182266e-06, "epoch": 0.16296296296296298, "percentage": 5.43, "elapsed_time": "0:08:58", "remaining_time": "2:36:11"} | |
| {"current_steps": 120, "total_steps": 2025, "loss": 0.2281, "learning_rate": 5.911330049261085e-06, "epoch": 0.17777777777777778, "percentage": 5.93, "elapsed_time": "0:09:47", "remaining_time": "2:35:28"} | |
| {"current_steps": 130, "total_steps": 2025, "loss": 0.234, "learning_rate": 6.403940886699508e-06, "epoch": 0.1925925925925926, "percentage": 6.42, "elapsed_time": "0:10:36", "remaining_time": "2:34:44"} | |
| {"current_steps": 140, "total_steps": 2025, "loss": 0.2239, "learning_rate": 6.896551724137932e-06, "epoch": 0.2074074074074074, "percentage": 6.91, "elapsed_time": "0:11:25", "remaining_time": "2:33:52"} | |
| {"current_steps": 150, "total_steps": 2025, "loss": 0.2339, "learning_rate": 7.3891625615763555e-06, "epoch": 0.2222222222222222, "percentage": 7.41, "elapsed_time": "0:12:14", "remaining_time": "2:33:00"} | |
| {"current_steps": 160, "total_steps": 2025, "loss": 0.2261, "learning_rate": 7.88177339901478e-06, "epoch": 0.23703703703703705, "percentage": 7.9, "elapsed_time": "0:13:03", "remaining_time": "2:32:12"} | |
| {"current_steps": 170, "total_steps": 2025, "loss": 0.253, "learning_rate": 8.374384236453203e-06, "epoch": 0.2518518518518518, "percentage": 8.4, "elapsed_time": "0:13:51", "remaining_time": "2:31:14"} | |
| {"current_steps": 180, "total_steps": 2025, "loss": 0.2592, "learning_rate": 8.866995073891627e-06, "epoch": 0.26666666666666666, "percentage": 8.89, "elapsed_time": "0:14:40", "remaining_time": "2:30:24"} | |
| {"current_steps": 190, "total_steps": 2025, "loss": 0.2558, "learning_rate": 9.359605911330049e-06, "epoch": 0.2814814814814815, "percentage": 9.38, "elapsed_time": "0:15:28", "remaining_time": "2:29:30"} | |
| {"current_steps": 200, "total_steps": 2025, "loss": 0.2791, "learning_rate": 9.852216748768475e-06, "epoch": 0.2962962962962963, "percentage": 9.88, "elapsed_time": "0:16:17", "remaining_time": "2:28:40"} | |
| {"current_steps": 210, "total_steps": 2025, "loss": 0.266, "learning_rate": 9.99963580513638e-06, "epoch": 0.3111111111111111, "percentage": 10.37, "elapsed_time": "0:17:06", "remaining_time": "2:27:55"} | |
| {"current_steps": 220, "total_steps": 2025, "loss": 0.2644, "learning_rate": 9.997852121279563e-06, "epoch": 0.32592592592592595, "percentage": 10.86, "elapsed_time": "0:17:55", "remaining_time": "2:27:04"} | |
| {"current_steps": 230, "total_steps": 2025, "loss": 0.2398, "learning_rate": 9.994582585118449e-06, "epoch": 0.34074074074074073, "percentage": 11.36, "elapsed_time": "0:18:45", "remaining_time": "2:26:19"} | |
| {"current_steps": 240, "total_steps": 2025, "loss": 0.2459, "learning_rate": 9.989828168680164e-06, "epoch": 0.35555555555555557, "percentage": 11.85, "elapsed_time": "0:19:33", "remaining_time": "2:25:27"} | |
| {"current_steps": 250, "total_steps": 2025, "loss": 0.2874, "learning_rate": 9.983590285444025e-06, "epoch": 0.37037037037037035, "percentage": 12.35, "elapsed_time": "0:20:20", "remaining_time": "2:24:25"} | |
| {"current_steps": 260, "total_steps": 2025, "loss": 0.243, "learning_rate": 9.975870789921322e-06, "epoch": 0.3851851851851852, "percentage": 12.84, "elapsed_time": "0:21:09", "remaining_time": "2:23:38"} | |
| {"current_steps": 270, "total_steps": 2025, "loss": 0.2418, "learning_rate": 9.966671977103972e-06, "epoch": 0.4, "percentage": 13.33, "elapsed_time": "0:21:59", "remaining_time": "2:22:53"} | |
| {"current_steps": 280, "total_steps": 2025, "loss": 0.3023, "learning_rate": 9.955996581782218e-06, "epoch": 0.4148148148148148, "percentage": 13.83, "elapsed_time": "0:22:48", "remaining_time": "2:22:11"} | |
| {"current_steps": 290, "total_steps": 2025, "loss": 0.2765, "learning_rate": 9.943847777731584e-06, "epoch": 0.42962962962962964, "percentage": 14.32, "elapsed_time": "0:23:38", "remaining_time": "2:21:25"} | |
| {"current_steps": 300, "total_steps": 2025, "loss": 0.2804, "learning_rate": 9.93022917676932e-06, "epoch": 0.4444444444444444, "percentage": 14.81, "elapsed_time": "0:24:27", "remaining_time": "2:20:36"} | |
| {"current_steps": 310, "total_steps": 2025, "loss": 0.2703, "learning_rate": 9.915144827680606e-06, "epoch": 0.45925925925925926, "percentage": 15.31, "elapsed_time": "0:25:16", "remaining_time": "2:19:47"} | |
| {"current_steps": 320, "total_steps": 2025, "loss": 0.2875, "learning_rate": 9.898599215014868e-06, "epoch": 0.4740740740740741, "percentage": 15.8, "elapsed_time": "0:26:05", "remaining_time": "2:18:59"} | |
| {"current_steps": 330, "total_steps": 2025, "loss": 0.2639, "learning_rate": 9.880597257752522e-06, "epoch": 0.4888888888888889, "percentage": 16.3, "elapsed_time": "0:26:52", "remaining_time": "2:18:02"} | |
| {"current_steps": 340, "total_steps": 2025, "loss": 0.2339, "learning_rate": 9.861144307842574e-06, "epoch": 0.5037037037037037, "percentage": 16.79, "elapsed_time": "0:27:39", "remaining_time": "2:17:02"} | |
| {"current_steps": 350, "total_steps": 2025, "loss": 0.2552, "learning_rate": 9.840246148611485e-06, "epoch": 0.5185185185185185, "percentage": 17.28, "elapsed_time": "0:28:27", "remaining_time": "2:16:12"} | |
| {"current_steps": 360, "total_steps": 2025, "loss": 0.2817, "learning_rate": 9.817908993043819e-06, "epoch": 0.5333333333333333, "percentage": 17.78, "elapsed_time": "0:29:16", "remaining_time": "2:15:23"} | |
| {"current_steps": 370, "total_steps": 2025, "loss": 0.2641, "learning_rate": 9.794139481935108e-06, "epoch": 0.5481481481481482, "percentage": 18.27, "elapsed_time": "0:30:05", "remaining_time": "2:14:36"} | |
| {"current_steps": 380, "total_steps": 2025, "loss": 0.2826, "learning_rate": 9.768944681917582e-06, "epoch": 0.562962962962963, "percentage": 18.77, "elapsed_time": "0:30:55", "remaining_time": "2:13:50"} | |
| {"current_steps": 390, "total_steps": 2025, "loss": 0.2559, "learning_rate": 9.742332083359252e-06, "epoch": 0.5777777777777777, "percentage": 19.26, "elapsed_time": "0:31:43", "remaining_time": "2:13:00"} | |
| {"current_steps": 400, "total_steps": 2025, "loss": 0.2306, "learning_rate": 9.714309598137045e-06, "epoch": 0.5925925925925926, "percentage": 19.75, "elapsed_time": "0:32:32", "remaining_time": "2:12:14"} | |
| {"current_steps": 410, "total_steps": 2025, "loss": 0.2682, "learning_rate": 9.68488555728462e-06, "epoch": 0.6074074074074074, "percentage": 20.25, "elapsed_time": "0:33:20", "remaining_time": "2:11:21"} | |
| {"current_steps": 420, "total_steps": 2025, "loss": 0.2607, "learning_rate": 9.654068708515564e-06, "epoch": 0.6222222222222222, "percentage": 20.74, "elapsed_time": "0:34:09", "remaining_time": "2:10:31"} | |
| {"current_steps": 430, "total_steps": 2025, "loss": 0.2383, "learning_rate": 9.621868213622713e-06, "epoch": 0.6370370370370371, "percentage": 21.23, "elapsed_time": "0:34:58", "remaining_time": "2:09:42"} | |
| {"current_steps": 440, "total_steps": 2025, "loss": 0.293, "learning_rate": 9.588293645754363e-06, "epoch": 0.6518518518518519, "percentage": 21.73, "elapsed_time": "0:35:47", "remaining_time": "2:08:55"} | |
| {"current_steps": 450, "total_steps": 2025, "loss": 0.2942, "learning_rate": 9.553354986568201e-06, "epoch": 0.6666666666666666, "percentage": 22.22, "elapsed_time": "0:36:36", "remaining_time": "2:08:08"} | |
| {"current_steps": 460, "total_steps": 2025, "loss": 0.2524, "learning_rate": 9.517062623263768e-06, "epoch": 0.6814814814814815, "percentage": 22.72, "elapsed_time": "0:37:26", "remaining_time": "2:07:22"} | |
| {"current_steps": 470, "total_steps": 2025, "loss": 0.2623, "learning_rate": 9.479427345494366e-06, "epoch": 0.6962962962962963, "percentage": 23.21, "elapsed_time": "0:38:13", "remaining_time": "2:06:29"} | |
| {"current_steps": 480, "total_steps": 2025, "loss": 0.2786, "learning_rate": 9.440460342159314e-06, "epoch": 0.7111111111111111, "percentage": 23.7, "elapsed_time": "0:39:02", "remaining_time": "2:05:41"} | |
| {"current_steps": 490, "total_steps": 2025, "loss": 0.2276, "learning_rate": 9.40017319807751e-06, "epoch": 0.725925925925926, "percentage": 24.2, "elapsed_time": "0:39:52", "remaining_time": "2:04:53"} | |
| {"current_steps": 500, "total_steps": 2025, "loss": 0.2781, "learning_rate": 9.358577890543277e-06, "epoch": 0.7407407407407407, "percentage": 24.69, "elapsed_time": "0:40:41", "remaining_time": "2:04:07"} | |
| {"current_steps": 500, "total_steps": 2025, "eval_loss": 0.2680646777153015, "epoch": 0.7407407407407407, "percentage": 24.69, "elapsed_time": "0:42:32", "remaining_time": "2:09:45"} | |
| {"current_steps": 510, "total_steps": 2025, "loss": 0.2637, "learning_rate": 9.315686785765556e-06, "epoch": 0.7555555555555555, "percentage": 25.19, "elapsed_time": "0:44:44", "remaining_time": "2:12:53"} | |
| {"current_steps": 520, "total_steps": 2025, "loss": 0.2707, "learning_rate": 9.271512635191427e-06, "epoch": 0.7703703703703704, "percentage": 25.68, "elapsed_time": "0:45:33", "remaining_time": "2:11:51"} | |
| {"current_steps": 530, "total_steps": 2025, "loss": 0.2376, "learning_rate": 9.22606857171515e-06, "epoch": 0.7851851851851852, "percentage": 26.17, "elapsed_time": "0:46:22", "remaining_time": "2:10:48"} | |
| {"current_steps": 540, "total_steps": 2025, "loss": 0.2495, "learning_rate": 9.179368105773768e-06, "epoch": 0.8, "percentage": 26.67, "elapsed_time": "0:47:11", "remaining_time": "2:09:46"} | |
| {"current_steps": 550, "total_steps": 2025, "loss": 0.2379, "learning_rate": 9.131425121330477e-06, "epoch": 0.8148148148148148, "percentage": 27.16, "elapsed_time": "0:48:00", "remaining_time": "2:08:45"} | |
| {"current_steps": 560, "total_steps": 2025, "loss": 0.2702, "learning_rate": 9.082253871746962e-06, "epoch": 0.8296296296296296, "percentage": 27.65, "elapsed_time": "0:48:49", "remaining_time": "2:07:44"} | |
| {"current_steps": 570, "total_steps": 2025, "loss": 0.2961, "learning_rate": 9.031868975545884e-06, "epoch": 0.8444444444444444, "percentage": 28.15, "elapsed_time": "0:49:39", "remaining_time": "2:06:45"} | |
| {"current_steps": 580, "total_steps": 2025, "loss": 0.2381, "learning_rate": 8.980285412064827e-06, "epoch": 0.8592592592592593, "percentage": 28.64, "elapsed_time": "0:50:28", "remaining_time": "2:05:46"} | |
| {"current_steps": 590, "total_steps": 2025, "loss": 0.2258, "learning_rate": 8.92751851700297e-06, "epoch": 0.8740740740740741, "percentage": 29.14, "elapsed_time": "0:51:14", "remaining_time": "2:04:36"} | |
| {"current_steps": 600, "total_steps": 2025, "loss": 0.2395, "learning_rate": 8.873583977861802e-06, "epoch": 0.8888888888888888, "percentage": 29.63, "elapsed_time": "0:52:00", "remaining_time": "2:03:32"} | |
| {"current_steps": 610, "total_steps": 2025, "loss": 0.2458, "learning_rate": 8.818497829281272e-06, "epoch": 0.9037037037037037, "percentage": 30.12, "elapsed_time": "0:52:50", "remaining_time": "2:02:34"} | |
| {"current_steps": 620, "total_steps": 2025, "loss": 0.2552, "learning_rate": 8.762276448272709e-06, "epoch": 0.9185185185185185, "percentage": 30.62, "elapsed_time": "0:53:39", "remaining_time": "2:01:35"} | |
| {"current_steps": 630, "total_steps": 2025, "loss": 0.2909, "learning_rate": 8.70493654934996e-06, "epoch": 0.9333333333333333, "percentage": 31.11, "elapsed_time": "0:54:29", "remaining_time": "2:00:38"} | |
| {"current_steps": 640, "total_steps": 2025, "loss": 0.2564, "learning_rate": 8.646495179560221e-06, "epoch": 0.9481481481481482, "percentage": 31.6, "elapsed_time": "0:55:18", "remaining_time": "1:59:41"} | |
| {"current_steps": 650, "total_steps": 2025, "loss": 0.2374, "learning_rate": 8.586969713415949e-06, "epoch": 0.9629629629629629, "percentage": 32.1, "elapsed_time": "0:56:07", "remaining_time": "1:58:44"} | |
| {"current_steps": 660, "total_steps": 2025, "loss": 0.2441, "learning_rate": 8.526377847729475e-06, "epoch": 0.9777777777777777, "percentage": 32.59, "elapsed_time": "0:56:56", "remaining_time": "1:57:46"} | |
| {"current_steps": 670, "total_steps": 2025, "loss": 0.2566, "learning_rate": 8.46473759635176e-06, "epoch": 0.9925925925925926, "percentage": 33.09, "elapsed_time": "0:57:43", "remaining_time": "1:56:45"} | |
| {"current_steps": 680, "total_steps": 2025, "loss": 0.1853, "learning_rate": 8.402067284816919e-06, "epoch": 1.0074074074074073, "percentage": 33.58, "elapsed_time": "0:58:33", "remaining_time": "1:55:48"} | |
| {"current_steps": 690, "total_steps": 2025, "loss": 0.127, "learning_rate": 8.338385544894073e-06, "epoch": 1.0222222222222221, "percentage": 34.07, "elapsed_time": "0:59:22", "remaining_time": "1:54:52"} | |
| {"current_steps": 700, "total_steps": 2025, "loss": 0.1254, "learning_rate": 8.273711309048145e-06, "epoch": 1.037037037037037, "percentage": 34.57, "elapsed_time": "1:00:11", "remaining_time": "1:53:56"} | |
| {"current_steps": 710, "total_steps": 2025, "loss": 0.1385, "learning_rate": 8.208063804811293e-06, "epoch": 1.0518518518518518, "percentage": 35.06, "elapsed_time": "1:01:00", "remaining_time": "1:53:00"} | |
| {"current_steps": 720, "total_steps": 2025, "loss": 0.1344, "learning_rate": 8.141462549066581e-06, "epoch": 1.0666666666666667, "percentage": 35.56, "elapsed_time": "1:01:50", "remaining_time": "1:52:05"} | |
| {"current_steps": 730, "total_steps": 2025, "loss": 0.1419, "learning_rate": 8.073927342245663e-06, "epoch": 1.0814814814814815, "percentage": 36.05, "elapsed_time": "1:02:38", "remaining_time": "1:51:06"} | |
| {"current_steps": 740, "total_steps": 2025, "loss": 0.1308, "learning_rate": 8.005478262442132e-06, "epoch": 1.0962962962962963, "percentage": 36.54, "elapsed_time": "1:03:26", "remaining_time": "1:50:10"} | |
| {"current_steps": 750, "total_steps": 2025, "loss": 0.133, "learning_rate": 7.936135659442355e-06, "epoch": 1.1111111111111112, "percentage": 37.04, "elapsed_time": "1:04:15", "remaining_time": "1:49:14"} | |
| {"current_steps": 760, "total_steps": 2025, "loss": 0.1099, "learning_rate": 7.86592014867551e-06, "epoch": 1.125925925925926, "percentage": 37.53, "elapsed_time": "1:05:04", "remaining_time": "1:48:19"} | |
| {"current_steps": 770, "total_steps": 2025, "loss": 0.1429, "learning_rate": 7.794852605084661e-06, "epoch": 1.1407407407407408, "percentage": 38.02, "elapsed_time": "1:05:54", "remaining_time": "1:47:24"} | |
| {"current_steps": 780, "total_steps": 2025, "loss": 0.1227, "learning_rate": 7.722954156920675e-06, "epoch": 1.1555555555555554, "percentage": 38.52, "elapsed_time": "1:06:42", "remaining_time": "1:46:28"} | |
| {"current_steps": 790, "total_steps": 2025, "loss": 0.1499, "learning_rate": 7.650246179460826e-06, "epoch": 1.1703703703703703, "percentage": 39.01, "elapsed_time": "1:07:31", "remaining_time": "1:45:33"} | |
| {"current_steps": 800, "total_steps": 2025, "loss": 0.1564, "learning_rate": 7.57675028865397e-06, "epoch": 1.1851851851851851, "percentage": 39.51, "elapsed_time": "1:08:18", "remaining_time": "1:44:35"} | |
| {"current_steps": 810, "total_steps": 2025, "loss": 0.1441, "learning_rate": 7.502488334694167e-06, "epoch": 1.2, "percentage": 40.0, "elapsed_time": "1:09:07", "remaining_time": "1:43:41"} | |
| {"current_steps": 820, "total_steps": 2025, "loss": 0.1317, "learning_rate": 7.427482395524646e-06, "epoch": 1.2148148148148148, "percentage": 40.49, "elapsed_time": "1:09:55", "remaining_time": "1:42:45"} | |
| {"current_steps": 830, "total_steps": 2025, "loss": 0.1211, "learning_rate": 7.35175477027408e-06, "epoch": 1.2296296296296296, "percentage": 40.99, "elapsed_time": "1:10:44", "remaining_time": "1:41:51"} | |
| {"current_steps": 840, "total_steps": 2025, "loss": 0.1284, "learning_rate": 7.2753279726271e-06, "epoch": 1.2444444444444445, "percentage": 41.48, "elapsed_time": "1:11:34", "remaining_time": "1:40:57"} | |
| {"current_steps": 850, "total_steps": 2025, "loss": 0.1436, "learning_rate": 7.198224724131012e-06, "epoch": 1.2592592592592593, "percentage": 41.98, "elapsed_time": "1:12:23", "remaining_time": "1:40:03"} | |
| {"current_steps": 860, "total_steps": 2025, "loss": 0.1221, "learning_rate": 7.120467947440719e-06, "epoch": 1.2740740740740741, "percentage": 42.47, "elapsed_time": "1:13:11", "remaining_time": "1:39:08"} | |
| {"current_steps": 870, "total_steps": 2025, "loss": 0.1162, "learning_rate": 7.042080759503866e-06, "epoch": 1.2888888888888888, "percentage": 42.96, "elapsed_time": "1:13:59", "remaining_time": "1:38:13"} | |
| {"current_steps": 880, "total_steps": 2025, "loss": 0.1291, "learning_rate": 6.963086464688209e-06, "epoch": 1.3037037037037038, "percentage": 43.46, "elapsed_time": "1:14:48", "remaining_time": "1:37:19"} | |
| {"current_steps": 890, "total_steps": 2025, "loss": 0.1319, "learning_rate": 6.883508547853268e-06, "epoch": 1.3185185185185184, "percentage": 43.95, "elapsed_time": "1:15:37", "remaining_time": "1:36:26"} | |
| {"current_steps": 900, "total_steps": 2025, "loss": 0.125, "learning_rate": 6.8033706673683276e-06, "epoch": 1.3333333333333333, "percentage": 44.44, "elapsed_time": "1:16:26", "remaining_time": "1:35:33"} | |
| {"current_steps": 910, "total_steps": 2025, "loss": 0.1382, "learning_rate": 6.722696648078838e-06, "epoch": 1.348148148148148, "percentage": 44.94, "elapsed_time": "1:17:15", "remaining_time": "1:34:40"} | |
| {"current_steps": 920, "total_steps": 2025, "loss": 0.1311, "learning_rate": 6.641510474223338e-06, "epoch": 1.362962962962963, "percentage": 45.43, "elapsed_time": "1:18:05", "remaining_time": "1:33:47"} | |
| {"current_steps": 930, "total_steps": 2025, "loss": 0.1416, "learning_rate": 6.559836282302984e-06, "epoch": 1.3777777777777778, "percentage": 45.93, "elapsed_time": "1:18:52", "remaining_time": "1:32:51"} | |
| {"current_steps": 940, "total_steps": 2025, "loss": 0.1519, "learning_rate": 6.477698353905808e-06, "epoch": 1.3925925925925926, "percentage": 46.42, "elapsed_time": "1:19:41", "remaining_time": "1:31:58"} | |
| {"current_steps": 950, "total_steps": 2025, "loss": 0.1179, "learning_rate": 6.395121108487855e-06, "epoch": 1.4074074074074074, "percentage": 46.91, "elapsed_time": "1:20:29", "remaining_time": "1:31:04"} | |
| {"current_steps": 960, "total_steps": 2025, "loss": 0.1329, "learning_rate": 6.312129096113313e-06, "epoch": 1.4222222222222223, "percentage": 47.41, "elapsed_time": "1:21:18", "remaining_time": "1:30:11"} | |
| {"current_steps": 970, "total_steps": 2025, "loss": 0.0981, "learning_rate": 6.228746990155831e-06, "epoch": 1.4370370370370371, "percentage": 47.9, "elapsed_time": "1:22:07", "remaining_time": "1:29:19"} | |
| {"current_steps": 980, "total_steps": 2025, "loss": 0.116, "learning_rate": 6.144999579963164e-06, "epoch": 1.4518518518518517, "percentage": 48.4, "elapsed_time": "1:22:56", "remaining_time": "1:28:26"} | |
| {"current_steps": 990, "total_steps": 2025, "loss": 0.136, "learning_rate": 6.060911763487353e-06, "epoch": 1.4666666666666668, "percentage": 48.89, "elapsed_time": "1:23:45", "remaining_time": "1:27:33"} | |
| {"current_steps": 1000, "total_steps": 2025, "loss": 0.1417, "learning_rate": 5.976508539882604e-06, "epoch": 1.4814814814814814, "percentage": 49.38, "elapsed_time": "1:24:34", "remaining_time": "1:26:40"} | |
| {"current_steps": 1000, "total_steps": 2025, "eval_loss": 0.23991718888282776, "epoch": 1.4814814814814814, "percentage": 49.38, "elapsed_time": "1:26:24", "remaining_time": "1:28:33"} | |
| {"current_steps": 1010, "total_steps": 2025, "loss": 0.1118, "learning_rate": 5.891815002073081e-06, "epoch": 1.4962962962962962, "percentage": 49.88, "elapsed_time": "1:28:27", "remaining_time": "1:28:54"} | |
| {"current_steps": 1020, "total_steps": 2025, "loss": 0.1202, "learning_rate": 5.806856329292839e-06, "epoch": 1.511111111111111, "percentage": 50.37, "elapsed_time": "1:29:16", "remaining_time": "1:27:58"} | |
| {"current_steps": 1030, "total_steps": 2025, "loss": 0.1256, "learning_rate": 5.721657779600071e-06, "epoch": 1.525925925925926, "percentage": 50.86, "elapsed_time": "1:30:06", "remaining_time": "1:27:02"} | |
| {"current_steps": 1040, "total_steps": 2025, "loss": 0.1402, "learning_rate": 5.636244682367937e-06, "epoch": 1.5407407407407407, "percentage": 51.36, "elapsed_time": "1:30:55", "remaining_time": "1:26:06"} | |
| {"current_steps": 1050, "total_steps": 2025, "loss": 0.1034, "learning_rate": 5.5506424307541895e-06, "epoch": 1.5555555555555556, "percentage": 51.85, "elapsed_time": "1:31:44", "remaining_time": "1:25:11"} | |
| {"current_steps": 1060, "total_steps": 2025, "loss": 0.1302, "learning_rate": 5.464876474151835e-06, "epoch": 1.5703703703703704, "percentage": 52.35, "elapsed_time": "1:32:33", "remaining_time": "1:24:16"} | |
| {"current_steps": 1070, "total_steps": 2025, "loss": 0.128, "learning_rate": 5.3789723106230675e-06, "epoch": 1.585185185185185, "percentage": 52.84, "elapsed_time": "1:33:22", "remaining_time": "1:23:20"} | |
| {"current_steps": 1080, "total_steps": 2025, "loss": 0.1265, "learning_rate": 5.292955479318756e-06, "epoch": 1.6, "percentage": 53.33, "elapsed_time": "1:34:09", "remaining_time": "1:22:23"} | |
| {"current_steps": 1090, "total_steps": 2025, "loss": 0.1276, "learning_rate": 5.206851552885691e-06, "epoch": 1.6148148148148147, "percentage": 53.83, "elapsed_time": "1:34:57", "remaining_time": "1:21:27"} | |
| {"current_steps": 1100, "total_steps": 2025, "loss": 0.1131, "learning_rate": 5.120686129863882e-06, "epoch": 1.6296296296296298, "percentage": 54.32, "elapsed_time": "1:35:46", "remaining_time": "1:20:32"} | |
| {"current_steps": 1110, "total_steps": 2025, "loss": 0.1353, "learning_rate": 5.0344848270761635e-06, "epoch": 1.6444444444444444, "percentage": 54.81, "elapsed_time": "1:36:35", "remaining_time": "1:19:37"} | |
| {"current_steps": 1120, "total_steps": 2025, "loss": 0.1256, "learning_rate": 4.948273272012363e-06, "epoch": 1.6592592592592592, "percentage": 55.31, "elapsed_time": "1:37:24", "remaining_time": "1:18:42"} | |
| {"current_steps": 1130, "total_steps": 2025, "loss": 0.1223, "learning_rate": 4.862077095210284e-06, "epoch": 1.674074074074074, "percentage": 55.8, "elapsed_time": "1:38:13", "remaining_time": "1:17:48"} | |
| {"current_steps": 1140, "total_steps": 2025, "loss": 0.1101, "learning_rate": 4.775921922635806e-06, "epoch": 1.6888888888888889, "percentage": 56.3, "elapsed_time": "1:39:02", "remaining_time": "1:16:53"} | |
| {"current_steps": 1150, "total_steps": 2025, "loss": 0.1136, "learning_rate": 4.689833368064326e-06, "epoch": 1.7037037037037037, "percentage": 56.79, "elapsed_time": "1:39:48", "remaining_time": "1:15:56"} | |
| {"current_steps": 1160, "total_steps": 2025, "loss": 0.0936, "learning_rate": 4.603837025465829e-06, "epoch": 1.7185185185185186, "percentage": 57.28, "elapsed_time": "1:40:37", "remaining_time": "1:15:02"} | |
| {"current_steps": 1170, "total_steps": 2025, "loss": 0.1035, "learning_rate": 4.517958461395846e-06, "epoch": 1.7333333333333334, "percentage": 57.78, "elapsed_time": "1:41:26", "remaining_time": "1:14:07"} | |
| {"current_steps": 1180, "total_steps": 2025, "loss": 0.0978, "learning_rate": 4.432223207394577e-06, "epoch": 1.748148148148148, "percentage": 58.27, "elapsed_time": "1:42:15", "remaining_time": "1:13:13"} | |
| {"current_steps": 1190, "total_steps": 2025, "loss": 0.1234, "learning_rate": 4.346656752396388e-06, "epoch": 1.762962962962963, "percentage": 58.77, "elapsed_time": "1:43:04", "remaining_time": "1:12:19"} | |
| {"current_steps": 1200, "total_steps": 2025, "loss": 0.1117, "learning_rate": 4.261284535152016e-06, "epoch": 1.7777777777777777, "percentage": 59.26, "elapsed_time": "1:43:52", "remaining_time": "1:11:24"} | |
| {"current_steps": 1210, "total_steps": 2025, "loss": 0.1151, "learning_rate": 4.176131936665669e-06, "epoch": 1.7925925925925927, "percentage": 59.75, "elapsed_time": "1:44:41", "remaining_time": "1:10:30"} | |
| {"current_steps": 1220, "total_steps": 2025, "loss": 0.1087, "learning_rate": 4.0912242726493e-06, "epoch": 1.8074074074074074, "percentage": 60.25, "elapsed_time": "1:45:30", "remaining_time": "1:09:37"} | |
| {"current_steps": 1230, "total_steps": 2025, "loss": 0.1094, "learning_rate": 4.006586785996285e-06, "epoch": 1.8222222222222222, "percentage": 60.74, "elapsed_time": "1:46:19", "remaining_time": "1:08:43"} | |
| {"current_steps": 1240, "total_steps": 2025, "loss": 0.1077, "learning_rate": 3.922244639276773e-06, "epoch": 1.837037037037037, "percentage": 61.23, "elapsed_time": "1:47:08", "remaining_time": "1:07:49"} | |
| {"current_steps": 1250, "total_steps": 2025, "loss": 0.1106, "learning_rate": 3.838222907256884e-06, "epoch": 1.8518518518518519, "percentage": 61.73, "elapsed_time": "1:47:57", "remaining_time": "1:06:56"} | |
| {"current_steps": 1260, "total_steps": 2025, "loss": 0.1075, "learning_rate": 3.7545465694440363e-06, "epoch": 1.8666666666666667, "percentage": 62.22, "elapsed_time": "1:48:46", "remaining_time": "1:06:02"} | |
| {"current_steps": 1270, "total_steps": 2025, "loss": 0.1097, "learning_rate": 3.6712405026605792e-06, "epoch": 1.8814814814814815, "percentage": 62.72, "elapsed_time": "1:49:35", "remaining_time": "1:05:09"} | |
| {"current_steps": 1280, "total_steps": 2025, "loss": 0.1004, "learning_rate": 3.5883294736479612e-06, "epoch": 1.8962962962962964, "percentage": 63.21, "elapsed_time": "1:50:21", "remaining_time": "1:04:13"} | |
| {"current_steps": 1290, "total_steps": 2025, "loss": 0.0906, "learning_rate": 3.5058381317036285e-06, "epoch": 1.911111111111111, "percentage": 63.7, "elapsed_time": "1:51:08", "remaining_time": "1:03:19"} | |
| {"current_steps": 1300, "total_steps": 2025, "loss": 0.1209, "learning_rate": 3.423791001352823e-06, "epoch": 1.925925925925926, "percentage": 64.2, "elapsed_time": "1:51:56", "remaining_time": "1:02:25"} | |
| {"current_steps": 1310, "total_steps": 2025, "loss": 0.1084, "learning_rate": 3.3422124750574902e-06, "epoch": 1.9407407407407407, "percentage": 64.69, "elapsed_time": "1:52:45", "remaining_time": "1:01:32"} | |
| {"current_steps": 1320, "total_steps": 2025, "loss": 0.0974, "learning_rate": 3.2611268059644535e-06, "epoch": 1.9555555555555557, "percentage": 65.19, "elapsed_time": "1:53:33", "remaining_time": "1:00:39"} | |
| {"current_steps": 1330, "total_steps": 2025, "loss": 0.1291, "learning_rate": 3.1805581006949856e-06, "epoch": 1.9703703703703703, "percentage": 65.68, "elapsed_time": "1:54:22", "remaining_time": "0:59:45"} | |
| {"current_steps": 1340, "total_steps": 2025, "loss": 0.1028, "learning_rate": 3.100530312177956e-06, "epoch": 1.9851851851851852, "percentage": 66.17, "elapsed_time": "1:55:10", "remaining_time": "0:58:52"} | |
| {"current_steps": 1350, "total_steps": 2025, "loss": 0.1026, "learning_rate": 3.0210672325286806e-06, "epoch": 2.0, "percentage": 66.67, "elapsed_time": "1:55:59", "remaining_time": "0:57:59"} | |
| {"current_steps": 1360, "total_steps": 2025, "loss": 0.0428, "learning_rate": 2.9421924859755525e-06, "epoch": 2.0148148148148146, "percentage": 67.16, "elapsed_time": "1:56:45", "remaining_time": "0:57:05"} | |
| {"current_steps": 1370, "total_steps": 2025, "loss": 0.0373, "learning_rate": 2.8639295218366115e-06, "epoch": 2.0296296296296297, "percentage": 67.65, "elapsed_time": "1:57:32", "remaining_time": "0:56:11"} | |
| {"current_steps": 1380, "total_steps": 2025, "loss": 0.0354, "learning_rate": 2.78630160754811e-06, "epoch": 2.0444444444444443, "percentage": 68.15, "elapsed_time": "1:58:21", "remaining_time": "0:55:18"} | |
| {"current_steps": 1390, "total_steps": 2025, "loss": 0.0265, "learning_rate": 2.709331821747133e-06, "epoch": 2.0592592592592593, "percentage": 68.64, "elapsed_time": "1:59:09", "remaining_time": "0:54:26"} | |
| {"current_steps": 1400, "total_steps": 2025, "loss": 0.0278, "learning_rate": 2.63304304741037e-06, "epoch": 2.074074074074074, "percentage": 69.14, "elapsed_time": "1:59:58", "remaining_time": "0:53:33"} | |
| {"current_steps": 1410, "total_steps": 2025, "loss": 0.0487, "learning_rate": 2.55745796505105e-06, "epoch": 2.088888888888889, "percentage": 69.63, "elapsed_time": "2:00:47", "remaining_time": "0:52:41"} | |
| {"current_steps": 1420, "total_steps": 2025, "loss": 0.0364, "learning_rate": 2.482599045976059e-06, "epoch": 2.1037037037037036, "percentage": 70.12, "elapsed_time": "2:01:36", "remaining_time": "0:51:48"} | |
| {"current_steps": 1430, "total_steps": 2025, "loss": 0.045, "learning_rate": 2.408488545605265e-06, "epoch": 2.1185185185185187, "percentage": 70.62, "elapsed_time": "2:02:25", "remaining_time": "0:50:56"} | |
| {"current_steps": 1440, "total_steps": 2025, "loss": 0.0368, "learning_rate": 2.3351484968550264e-06, "epoch": 2.1333333333333333, "percentage": 71.11, "elapsed_time": "2:03:11", "remaining_time": "0:50:02"} | |
| {"current_steps": 1450, "total_steps": 2025, "loss": 0.0421, "learning_rate": 2.2626007035878377e-06, "epoch": 2.148148148148148, "percentage": 71.6, "elapsed_time": "2:04:00", "remaining_time": "0:49:10"} | |
| {"current_steps": 1460, "total_steps": 2025, "loss": 0.0315, "learning_rate": 2.1908667341300923e-06, "epoch": 2.162962962962963, "percentage": 72.1, "elapsed_time": "2:04:49", "remaining_time": "0:48:18"} | |
| {"current_steps": 1470, "total_steps": 2025, "loss": 0.0348, "learning_rate": 2.1199679148598434e-06, "epoch": 2.1777777777777776, "percentage": 72.59, "elapsed_time": "2:05:38", "remaining_time": "0:47:26"} | |
| {"current_steps": 1480, "total_steps": 2025, "loss": 0.0496, "learning_rate": 2.0499253238665284e-06, "epoch": 2.1925925925925926, "percentage": 73.09, "elapsed_time": "2:06:26", "remaining_time": "0:46:33"} | |
| {"current_steps": 1490, "total_steps": 2025, "loss": 0.0287, "learning_rate": 1.9807597846844737e-06, "epoch": 2.2074074074074073, "percentage": 73.58, "elapsed_time": "2:07:16", "remaining_time": "0:45:41"} | |
| {"current_steps": 1500, "total_steps": 2025, "loss": 0.0362, "learning_rate": 1.9124918601021124e-06, "epoch": 2.2222222222222223, "percentage": 74.07, "elapsed_time": "2:08:05", "remaining_time": "0:44:49"} | |
| {"current_steps": 1500, "total_steps": 2025, "eval_loss": 0.25548920035362244, "epoch": 2.2222222222222223, "percentage": 74.07, "elapsed_time": "2:09:55", "remaining_time": "0:45:28"} | |
| {"current_steps": 1510, "total_steps": 2025, "loss": 0.0327, "learning_rate": 1.845141846048691e-06, "epoch": 2.237037037037037, "percentage": 74.57, "elapsed_time": "2:11:58", "remaining_time": "0:45:00"} | |
| {"current_steps": 1520, "total_steps": 2025, "loss": 0.0275, "learning_rate": 1.778729765560337e-06, "epoch": 2.251851851851852, "percentage": 75.06, "elapsed_time": "2:12:47", "remaining_time": "0:44:07"} | |
| {"current_steps": 1530, "total_steps": 2025, "loss": 0.0465, "learning_rate": 1.7132753628272403e-06, "epoch": 2.2666666666666666, "percentage": 75.56, "elapsed_time": "2:13:36", "remaining_time": "0:43:13"} | |
| {"current_steps": 1540, "total_steps": 2025, "loss": 0.0308, "learning_rate": 1.6487980973237434e-06, "epoch": 2.2814814814814817, "percentage": 76.05, "elapsed_time": "2:14:25", "remaining_time": "0:42:20"} | |
| {"current_steps": 1550, "total_steps": 2025, "loss": 0.0345, "learning_rate": 1.5853171380230791e-06, "epoch": 2.2962962962962963, "percentage": 76.54, "elapsed_time": "2:15:14", "remaining_time": "0:41:26"} | |
| {"current_steps": 1560, "total_steps": 2025, "loss": 0.0281, "learning_rate": 1.5228513576984633e-06, "epoch": 2.311111111111111, "percentage": 77.04, "elapsed_time": "2:16:03", "remaining_time": "0:40:33"} | |
| {"current_steps": 1570, "total_steps": 2025, "loss": 0.0445, "learning_rate": 1.4614193273122562e-06, "epoch": 2.325925925925926, "percentage": 77.53, "elapsed_time": "2:16:52", "remaining_time": "0:39:40"} | |
| {"current_steps": 1580, "total_steps": 2025, "loss": 0.0435, "learning_rate": 1.401039310494855e-06, "epoch": 2.3407407407407406, "percentage": 78.02, "elapsed_time": "2:17:40", "remaining_time": "0:38:46"} | |
| {"current_steps": 1590, "total_steps": 2025, "loss": 0.0353, "learning_rate": 1.3417292581149388e-06, "epoch": 2.3555555555555556, "percentage": 78.52, "elapsed_time": "2:18:30", "remaining_time": "0:37:53"} | |
| {"current_steps": 1600, "total_steps": 2025, "loss": 0.0299, "learning_rate": 1.2835068029427188e-06, "epoch": 2.3703703703703702, "percentage": 79.01, "elapsed_time": "2:19:17", "remaining_time": "0:36:59"} | |
| {"current_steps": 1610, "total_steps": 2025, "loss": 0.0351, "learning_rate": 1.2263892544077439e-06, "epoch": 2.3851851851851853, "percentage": 79.51, "elapsed_time": "2:20:00", "remaining_time": "0:36:05"} | |
| {"current_steps": 1620, "total_steps": 2025, "loss": 0.0449, "learning_rate": 1.1703935934528327e-06, "epoch": 2.4, "percentage": 80.0, "elapsed_time": "2:20:47", "remaining_time": "0:35:11"} | |
| {"current_steps": 1630, "total_steps": 2025, "loss": 0.0279, "learning_rate": 1.1155364674856834e-06, "epoch": 2.414814814814815, "percentage": 80.49, "elapsed_time": "2:21:36", "remaining_time": "0:34:18"} | |
| {"current_steps": 1640, "total_steps": 2025, "loss": 0.0443, "learning_rate": 1.0618341854296176e-06, "epoch": 2.4296296296296296, "percentage": 80.99, "elapsed_time": "2:22:25", "remaining_time": "0:33:26"} | |
| {"current_steps": 1650, "total_steps": 2025, "loss": 0.0398, "learning_rate": 1.0093027128749722e-06, "epoch": 2.4444444444444446, "percentage": 81.48, "elapsed_time": "2:23:13", "remaining_time": "0:32:33"} | |
| {"current_steps": 1660, "total_steps": 2025, "loss": 0.0392, "learning_rate": 9.57957667332562e-07, "epoch": 2.4592592592592593, "percentage": 81.98, "elapsed_time": "2:24:02", "remaining_time": "0:31:40"} | |
| {"current_steps": 1670, "total_steps": 2025, "loss": 0.0361, "learning_rate": 9.078143135906154e-07, "epoch": 2.474074074074074, "percentage": 82.47, "elapsed_time": "2:24:51", "remaining_time": "0:30:47"} | |
| {"current_steps": 1680, "total_steps": 2025, "loss": 0.0424, "learning_rate": 8.588875591765838e-07, "epoch": 2.488888888888889, "percentage": 82.96, "elapsed_time": "2:25:38", "remaining_time": "0:29:54"} | |
| {"current_steps": 1690, "total_steps": 2025, "loss": 0.0265, "learning_rate": 8.111919499251653e-07, "epoch": 2.5037037037037035, "percentage": 83.46, "elapsed_time": "2:26:27", "remaining_time": "0:29:01"} | |
| {"current_steps": 1700, "total_steps": 2025, "loss": 0.0379, "learning_rate": 7.647416656538464e-07, "epoch": 2.5185185185185186, "percentage": 83.95, "elapsed_time": "2:27:17", "remaining_time": "0:28:09"} | |
| {"current_steps": 1710, "total_steps": 2025, "loss": 0.0269, "learning_rate": 7.195505159472726e-07, "epoch": 2.533333333333333, "percentage": 84.44, "elapsed_time": "2:28:06", "remaining_time": "0:27:16"} | |
| {"current_steps": 1720, "total_steps": 2025, "loss": 0.0276, "learning_rate": 6.756319360516856e-07, "epoch": 2.5481481481481483, "percentage": 84.94, "elapsed_time": "2:28:54", "remaining_time": "0:26:24"} | |
| {"current_steps": 1730, "total_steps": 2025, "loss": 0.0283, "learning_rate": 6.329989828806482e-07, "epoch": 2.562962962962963, "percentage": 85.43, "elapsed_time": "2:29:43", "remaining_time": "0:25:31"} | |
| {"current_steps": 1740, "total_steps": 2025, "loss": 0.0391, "learning_rate": 5.916643311332438e-07, "epoch": 2.5777777777777775, "percentage": 85.93, "elapsed_time": "2:30:30", "remaining_time": "0:24:39"} | |
| {"current_steps": 1750, "total_steps": 2025, "loss": 0.0324, "learning_rate": 5.516402695259165e-07, "epoch": 2.5925925925925926, "percentage": 86.42, "elapsed_time": "2:31:19", "remaining_time": "0:23:46"} | |
| {"current_steps": 1760, "total_steps": 2025, "loss": 0.0298, "learning_rate": 5.12938697139056e-07, "epoch": 2.6074074074074076, "percentage": 86.91, "elapsed_time": "2:32:07", "remaining_time": "0:22:54"} | |
| {"current_steps": 1770, "total_steps": 2025, "loss": 0.0323, "learning_rate": 4.755711198794233e-07, "epoch": 2.6222222222222222, "percentage": 87.41, "elapsed_time": "2:32:55", "remaining_time": "0:22:01"} | |
| {"current_steps": 1780, "total_steps": 2025, "loss": 0.0256, "learning_rate": 4.395486470594645e-07, "epoch": 2.637037037037037, "percentage": 87.9, "elapsed_time": "2:33:45", "remaining_time": "0:21:09"} | |
| {"current_steps": 1790, "total_steps": 2025, "loss": 0.0291, "learning_rate": 4.048819880945337e-07, "epoch": 2.651851851851852, "percentage": 88.4, "elapsed_time": "2:34:33", "remaining_time": "0:20:17"} | |
| {"current_steps": 1800, "total_steps": 2025, "loss": 0.0262, "learning_rate": 3.7158144931900395e-07, "epoch": 2.6666666666666665, "percentage": 88.89, "elapsed_time": "2:35:22", "remaining_time": "0:19:25"} | |
| {"current_steps": 1810, "total_steps": 2025, "loss": 0.0282, "learning_rate": 3.396569309222114e-07, "epoch": 2.6814814814814816, "percentage": 89.38, "elapsed_time": "2:36:11", "remaining_time": "0:18:33"} | |
| {"current_steps": 1820, "total_steps": 2025, "loss": 0.0282, "learning_rate": 3.091179240051462e-07, "epoch": 2.696296296296296, "percentage": 89.88, "elapsed_time": "2:36:58", "remaining_time": "0:17:40"} | |
| {"current_steps": 1830, "total_steps": 2025, "loss": 0.036, "learning_rate": 2.799735077587695e-07, "epoch": 2.7111111111111112, "percentage": 90.37, "elapsed_time": "2:37:46", "remaining_time": "0:16:48"} | |
| {"current_steps": 1840, "total_steps": 2025, "loss": 0.0236, "learning_rate": 2.5223234676478193e-07, "epoch": 2.725925925925926, "percentage": 90.86, "elapsed_time": "2:38:35", "remaining_time": "0:15:56"} | |
| {"current_steps": 1850, "total_steps": 2025, "loss": 0.0306, "learning_rate": 2.2590268841966357e-07, "epoch": 2.7407407407407405, "percentage": 91.36, "elapsed_time": "2:39:24", "remaining_time": "0:15:04"} | |
| {"current_steps": 1860, "total_steps": 2025, "loss": 0.0306, "learning_rate": 2.0099236048273407e-07, "epoch": 2.7555555555555555, "percentage": 91.85, "elapsed_time": "2:40:13", "remaining_time": "0:14:12"} | |
| {"current_steps": 1870, "total_steps": 2025, "loss": 0.0317, "learning_rate": 1.7750876874897627e-07, "epoch": 2.7703703703703706, "percentage": 92.35, "elapsed_time": "2:41:02", "remaining_time": "0:13:20"} | |
| {"current_steps": 1880, "total_steps": 2025, "loss": 0.0303, "learning_rate": 1.554588948473068e-07, "epoch": 2.785185185185185, "percentage": 92.84, "elapsed_time": "2:41:51", "remaining_time": "0:12:29"} | |
| {"current_steps": 1890, "total_steps": 2025, "loss": 0.0318, "learning_rate": 1.3484929416495096e-07, "epoch": 2.8, "percentage": 93.33, "elapsed_time": "2:42:40", "remaining_time": "0:11:37"} | |
| {"current_steps": 1900, "total_steps": 2025, "loss": 0.0403, "learning_rate": 1.1568609389853546e-07, "epoch": 2.814814814814815, "percentage": 93.83, "elapsed_time": "2:43:27", "remaining_time": "0:10:45"} | |
| {"current_steps": 1910, "total_steps": 2025, "loss": 0.0284, "learning_rate": 9.7974991232489e-08, "epoch": 2.8296296296296295, "percentage": 94.32, "elapsed_time": "2:44:16", "remaining_time": "0:09:53"} | |
| {"current_steps": 1920, "total_steps": 2025, "loss": 0.0408, "learning_rate": 8.172125164527312e-08, "epoch": 2.8444444444444446, "percentage": 94.81, "elapsed_time": "2:45:05", "remaining_time": "0:09:01"} | |
| {"current_steps": 1930, "total_steps": 2025, "loss": 0.038, "learning_rate": 6.692970734397176e-08, "epoch": 2.859259259259259, "percentage": 95.31, "elapsed_time": "2:45:53", "remaining_time": "0:08:09"} | |
| {"current_steps": 1940, "total_steps": 2025, "loss": 0.0362, "learning_rate": 5.360475582768088e-08, "epoch": 2.8740740740740742, "percentage": 95.8, "elapsed_time": "2:46:42", "remaining_time": "0:07:18"} | |
| {"current_steps": 1950, "total_steps": 2025, "loss": 0.0294, "learning_rate": 4.175035858013987e-08, "epoch": 2.888888888888889, "percentage": 96.3, "elapsed_time": "2:47:28", "remaining_time": "0:06:26"} | |
| {"current_steps": 1960, "total_steps": 2025, "loss": 0.0303, "learning_rate": 3.13700398919925e-08, "epoch": 2.9037037037037035, "percentage": 96.79, "elapsed_time": "2:48:17", "remaining_time": "0:05:34"} | |
| {"current_steps": 1970, "total_steps": 2025, "loss": 0.0284, "learning_rate": 2.2466885813018925e-08, "epoch": 2.9185185185185185, "percentage": 97.28, "elapsed_time": "2:49:06", "remaining_time": "0:04:43"} | |
| {"current_steps": 1980, "total_steps": 2025, "loss": 0.0235, "learning_rate": 1.504354323466073e-08, "epoch": 2.9333333333333336, "percentage": 97.78, "elapsed_time": "2:49:55", "remaining_time": "0:03:51"} | |
| {"current_steps": 1990, "total_steps": 2025, "loss": 0.0383, "learning_rate": 9.102219103103161e-09, "epoch": 2.948148148148148, "percentage": 98.27, "elapsed_time": "2:50:44", "remaining_time": "0:03:00"} | |
| {"current_steps": 2000, "total_steps": 2025, "loss": 0.0244, "learning_rate": 4.644679763155524e-09, "epoch": 2.962962962962963, "percentage": 98.77, "elapsed_time": "2:51:33", "remaining_time": "0:02:08"} | |
| {"current_steps": 2000, "total_steps": 2025, "eval_loss": 0.2506597936153412, "epoch": 2.962962962962963, "percentage": 98.77, "elapsed_time": "2:53:23", "remaining_time": "0:02:10"} | |
| {"current_steps": 2010, "total_steps": 2025, "loss": 0.024, "learning_rate": 1.6722504331195822e-09, "epoch": 2.977777777777778, "percentage": 99.26, "elapsed_time": "2:55:34", "remaining_time": "0:01:18"} | |
| {"current_steps": 2020, "total_steps": 2025, "loss": 0.0264, "learning_rate": 1.8581481080415242e-10, "epoch": 2.9925925925925925, "percentage": 99.75, "elapsed_time": "2:56:22", "remaining_time": "0:00:26"} | |
| {"current_steps": 2025, "total_steps": 2025, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "2:58:06", "remaining_time": "0:00:00"} | |