sft3b / trainer_log.jsonl
LemTenku's picture
Upload folder using huggingface_hub
f476764 verified
{"current_steps": 10, "total_steps": 504, "loss": 2.2836, "lr": 1.9607843137254903e-05, "epoch": 0.05952380952380952, "percentage": 1.98, "elapsed_time": "0:00:03", "remaining_time": "0:02:50"}
{"current_steps": 20, "total_steps": 504, "loss": 2.0531, "lr": 3.9215686274509805e-05, "epoch": 0.11904761904761904, "percentage": 3.97, "elapsed_time": "0:00:06", "remaining_time": "0:02:41"}
{"current_steps": 30, "total_steps": 504, "loss": 2.0231, "lr": 5.882352941176471e-05, "epoch": 0.17857142857142858, "percentage": 5.95, "elapsed_time": "0:00:09", "remaining_time": "0:02:36"}
{"current_steps": 40, "total_steps": 504, "loss": 2.0226, "lr": 7.843137254901961e-05, "epoch": 0.23809523809523808, "percentage": 7.94, "elapsed_time": "0:00:13", "remaining_time": "0:02:32"}
{"current_steps": 50, "total_steps": 504, "loss": 2.1214, "lr": 9.80392156862745e-05, "epoch": 0.2976190476190476, "percentage": 9.92, "elapsed_time": "0:00:16", "remaining_time": "0:02:28"}
{"current_steps": 60, "total_steps": 504, "loss": 2.1241, "lr": 9.990263847374976e-05, "epoch": 0.35714285714285715, "percentage": 11.9, "elapsed_time": "0:00:19", "remaining_time": "0:02:24"}
{"current_steps": 70, "total_steps": 504, "loss": 2.225, "lr": 9.956656688041808e-05, "epoch": 0.4166666666666667, "percentage": 13.89, "elapsed_time": "0:00:22", "remaining_time": "0:02:21"}
{"current_steps": 80, "total_steps": 504, "loss": 2.1955, "lr": 9.899219854168429e-05, "epoch": 0.47619047619047616, "percentage": 15.87, "elapsed_time": "0:00:26", "remaining_time": "0:02:17"}
{"current_steps": 90, "total_steps": 504, "loss": 2.0837, "lr": 9.818229479678158e-05, "epoch": 0.5357142857142857, "percentage": 17.86, "elapsed_time": "0:00:29", "remaining_time": "0:02:14"}
{"current_steps": 100, "total_steps": 504, "loss": 2.0915, "lr": 9.714074934742556e-05, "epoch": 0.5952380952380952, "percentage": 19.84, "elapsed_time": "0:00:32", "remaining_time": "0:02:11"}
{"current_steps": 110, "total_steps": 504, "loss": 2.1907, "lr": 9.587256953841317e-05, "epoch": 0.6547619047619048, "percentage": 21.83, "elapsed_time": "0:00:35", "remaining_time": "0:02:07"}
{"current_steps": 120, "total_steps": 504, "loss": 2.1822, "lr": 9.438385228425938e-05, "epoch": 0.7142857142857143, "percentage": 23.81, "elapsed_time": "0:00:38", "remaining_time": "0:02:04"}
{"current_steps": 130, "total_steps": 504, "loss": 2.1324, "lr": 9.268175475760734e-05, "epoch": 0.7738095238095238, "percentage": 25.79, "elapsed_time": "0:00:42", "remaining_time": "0:02:01"}
{"current_steps": 140, "total_steps": 504, "loss": 2.1216, "lr": 9.077445998033015e-05, "epoch": 0.8333333333333334, "percentage": 27.78, "elapsed_time": "0:00:45", "remaining_time": "0:01:57"}
{"current_steps": 150, "total_steps": 504, "loss": 2.1736, "lr": 8.86711374827494e-05, "epoch": 0.8928571428571429, "percentage": 29.76, "elapsed_time": "0:00:48", "remaining_time": "0:01:54"}
{"current_steps": 160, "total_steps": 504, "loss": 2.161, "lr": 8.638189922010529e-05, "epoch": 0.9523809523809523, "percentage": 31.75, "elapsed_time": "0:00:51", "remaining_time": "0:01:51"}
{"current_steps": 170, "total_steps": 504, "loss": 1.9752, "lr": 8.391775095821482e-05, "epoch": 1.0119047619047619, "percentage": 33.73, "elapsed_time": "0:00:54", "remaining_time": "0:01:47"}
{"current_steps": 180, "total_steps": 504, "loss": 1.1203, "lr": 8.129053936203687e-05, "epoch": 1.0714285714285714, "percentage": 35.71, "elapsed_time": "0:00:58", "remaining_time": "0:01:44"}
{"current_steps": 190, "total_steps": 504, "loss": 1.1102, "lr": 7.851289504152201e-05, "epoch": 1.130952380952381, "percentage": 37.7, "elapsed_time": "0:01:01", "remaining_time": "0:01:41"}
{"current_steps": 200, "total_steps": 504, "loss": 1.0731, "lr": 7.559817182855977e-05, "epoch": 1.1904761904761905, "percentage": 39.68, "elapsed_time": "0:01:04", "remaining_time": "0:01:38"}
{"current_steps": 210, "total_steps": 504, "loss": 1.0537, "lr": 7.256038257695687e-05, "epoch": 1.25, "percentage": 41.67, "elapsed_time": "0:01:07", "remaining_time": "0:01:34"}
{"current_steps": 220, "total_steps": 504, "loss": 1.1446, "lr": 6.941413179409468e-05, "epoch": 1.3095238095238095, "percentage": 43.65, "elapsed_time": "0:01:11", "remaining_time": "0:01:31"}
{"current_steps": 230, "total_steps": 504, "loss": 1.0612, "lr": 6.61745454281468e-05, "epoch": 1.369047619047619, "percentage": 45.63, "elapsed_time": "0:01:14", "remaining_time": "0:01:28"}
{"current_steps": 240, "total_steps": 504, "loss": 1.0517, "lr": 6.28571981484123e-05, "epoch": 1.4285714285714286, "percentage": 47.62, "elapsed_time": "0:01:17", "remaining_time": "0:01:25"}
{"current_steps": 250, "total_steps": 504, "loss": 1.0755, "lr": 5.94780384683728e-05, "epoch": 1.4880952380952381, "percentage": 49.6, "elapsed_time": "0:01:20", "remaining_time": "0:01:22"}
{"current_steps": 260, "total_steps": 504, "loss": 1.1604, "lr": 5.605331207145219e-05, "epoch": 1.5476190476190477, "percentage": 51.59, "elapsed_time": "0:01:23", "remaining_time": "0:01:18"}
{"current_steps": 270, "total_steps": 504, "loss": 1.0858, "lr": 5.2599483708099016e-05, "epoch": 1.6071428571428572, "percentage": 53.57, "elapsed_time": "0:01:27", "remaining_time": "0:01:15"}
{"current_steps": 280, "total_steps": 504, "loss": 1.3748, "lr": 4.913315803968012e-05, "epoch": 1.6666666666666665, "percentage": 55.56, "elapsed_time": "0:01:30", "remaining_time": "0:01:12"}
{"current_steps": 290, "total_steps": 504, "loss": 1.1309, "lr": 4.567099980973684e-05, "epoch": 1.7261904761904763, "percentage": 57.54, "elapsed_time": "0:01:33", "remaining_time": "0:01:09"}
{"current_steps": 300, "total_steps": 504, "loss": 0.9784, "lr": 4.2229653726389765e-05, "epoch": 1.7857142857142856, "percentage": 59.52, "elapsed_time": "0:01:36", "remaining_time": "0:01:05"}
{"current_steps": 310, "total_steps": 504, "loss": 1.2982, "lr": 3.8825664441065734e-05, "epoch": 1.8452380952380953, "percentage": 61.51, "elapsed_time": "0:01:40", "remaining_time": "0:01:02"}
{"current_steps": 320, "total_steps": 504, "loss": 1.0105, "lr": 3.547539700825874e-05, "epoch": 1.9047619047619047, "percentage": 63.49, "elapsed_time": "0:01:43", "remaining_time": "0:00:59"}
{"current_steps": 330, "total_steps": 504, "loss": 1.0056, "lr": 3.219495820872265e-05, "epoch": 1.9642857142857144, "percentage": 65.48, "elapsed_time": "0:01:46", "remaining_time": "0:00:56"}
{"current_steps": 340, "total_steps": 504, "loss": 0.7399, "lr": 2.9000119114343582e-05, "epoch": 2.0238095238095237, "percentage": 67.46, "elapsed_time": "0:01:49", "remaining_time": "0:00:52"}
{"current_steps": 350, "total_steps": 504, "loss": 0.3847, "lr": 2.5906239266969805e-05, "epoch": 2.0833333333333335, "percentage": 69.44, "elapsed_time": "0:01:52", "remaining_time": "0:00:49"}
{"current_steps": 360, "total_steps": 504, "loss": 0.4477, "lr": 2.2928192835717644e-05, "epoch": 2.142857142857143, "percentage": 71.43, "elapsed_time": "0:01:56", "remaining_time": "0:00:46"}
{"current_steps": 370, "total_steps": 504, "loss": 0.3674, "lr": 2.008029710776041e-05, "epoch": 2.2023809523809526, "percentage": 73.41, "elapsed_time": "0:01:59", "remaining_time": "0:00:43"}
{"current_steps": 380, "total_steps": 504, "loss": 0.4358, "lr": 1.7376243656388923e-05, "epoch": 2.261904761904762, "percentage": 75.4, "elapsed_time": "0:02:02", "remaining_time": "0:00:40"}
{"current_steps": 390, "total_steps": 504, "loss": 0.4029, "lr": 1.4829032517260489e-05, "epoch": 2.3214285714285716, "percentage": 77.38, "elapsed_time": "0:02:05", "remaining_time": "0:00:36"}
{"current_steps": 400, "total_steps": 504, "loss": 0.3537, "lr": 1.245090968929148e-05, "epoch": 2.380952380952381, "percentage": 79.37, "elapsed_time": "0:02:09", "remaining_time": "0:00:33"}
{"current_steps": 410, "total_steps": 504, "loss": 0.3583, "lr": 1.0253308260664902e-05, "epoch": 2.4404761904761907, "percentage": 81.35, "elapsed_time": "0:02:12", "remaining_time": "0:00:30"}
{"current_steps": 420, "total_steps": 504, "loss": 0.3662, "lr": 8.246793442995954e-06, "epoch": 2.5, "percentage": 83.33, "elapsed_time": "0:02:15", "remaining_time": "0:00:27"}
{"current_steps": 430, "total_steps": 504, "loss": 0.354, "lr": 6.4410117779103e-06, "epoch": 2.5595238095238093, "percentage": 85.32, "elapsed_time": "0:02:18", "remaining_time": "0:00:23"}
{"current_steps": 440, "total_steps": 504, "loss": 0.4411, "lr": 4.844644760229544e-06, "epoch": 2.619047619047619, "percentage": 87.3, "elapsed_time": "0:02:22", "remaining_time": "0:00:20"}
{"current_steps": 450, "total_steps": 504, "loss": 0.4508, "lr": 3.465367100725908e-06, "epoch": 2.678571428571429, "percentage": 89.29, "elapsed_time": "0:02:25", "remaining_time": "0:00:17"}
{"current_steps": 460, "total_steps": 504, "loss": 0.3551, "lr": 2.3098098291025273e-06, "epoch": 2.738095238095238, "percentage": 91.27, "elapsed_time": "0:02:28", "remaining_time": "0:00:14"}
{"current_steps": 470, "total_steps": 504, "loss": 0.3454, "lr": 1.3835284145856275e-06, "epoch": 2.7976190476190474, "percentage": 93.25, "elapsed_time": "0:02:31", "remaining_time": "0:00:10"}
{"current_steps": 480, "total_steps": 504, "loss": 0.3835, "lr": 6.909760573925561e-07, "epoch": 2.857142857142857, "percentage": 95.24, "elapsed_time": "0:02:34", "remaining_time": "0:00:07"}
{"current_steps": 490, "total_steps": 504, "loss": 0.3762, "lr": 2.3548227947988167e-07, "epoch": 2.9166666666666665, "percentage": 97.22, "elapsed_time": "0:02:38", "remaining_time": "0:00:04"}
{"current_steps": 500, "total_steps": 504, "loss": 0.3555, "lr": 1.9236917498782758e-08, "epoch": 2.9761904761904763, "percentage": 99.21, "elapsed_time": "0:02:41", "remaining_time": "0:00:01"}
{"current_steps": 504, "total_steps": 504, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "0:02:59", "remaining_time": "0:00:00"}