Jiazheng Li
init push
7d35748
{"current_steps": 20, "total_steps": 632, "loss": 1.3016, "learning_rate": 4.9876553763060684e-05, "epoch": 0.13, "percentage": 3.16, "elapsed_time": "0:22:10", "remaining_time": "11:18:44"}
{"current_steps": 40, "total_steps": 632, "loss": 0.9953, "learning_rate": 4.950743417011591e-05, "epoch": 0.25, "percentage": 6.33, "elapsed_time": "0:44:06", "remaining_time": "10:52:50"}
{"current_steps": 60, "total_steps": 632, "loss": 0.925, "learning_rate": 4.889628653514402e-05, "epoch": 0.38, "percentage": 9.49, "elapsed_time": "1:09:28", "remaining_time": "11:02:22"}
{"current_steps": 80, "total_steps": 632, "loss": 0.8788, "learning_rate": 4.804914636820517e-05, "epoch": 0.51, "percentage": 12.66, "elapsed_time": "1:31:29", "remaining_time": "10:31:14"}
{"current_steps": 100, "total_steps": 632, "loss": 0.8668, "learning_rate": 4.6974379770560846e-05, "epoch": 0.63, "percentage": 15.82, "elapsed_time": "1:53:23", "remaining_time": "10:03:12"}
{"current_steps": 100, "total_steps": 632, "eval_loss": 0.8571113348007202, "epoch": 0.63, "percentage": 15.82, "elapsed_time": "2:07:51", "remaining_time": "11:20:11"}
{"current_steps": 120, "total_steps": 632, "loss": 0.8488, "learning_rate": 4.5682600813576435e-05, "epoch": 0.76, "percentage": 18.99, "elapsed_time": "2:29:50", "remaining_time": "10:39:21"}
{"current_steps": 140, "total_steps": 632, "loss": 0.8462, "learning_rate": 4.41865667173477e-05, "epoch": 0.88, "percentage": 22.15, "elapsed_time": "2:51:49", "remaining_time": "10:03:51"}
{"current_steps": 160, "total_steps": 632, "loss": 0.827, "learning_rate": 4.2501051864235636e-05, "epoch": 1.01, "percentage": 25.32, "elapsed_time": "3:14:53", "remaining_time": "9:34:54"}
{"current_steps": 180, "total_steps": 632, "loss": 0.7885, "learning_rate": 4.0642701891514e-05, "epoch": 1.14, "percentage": 28.48, "elapsed_time": "3:37:46", "remaining_time": "9:06:50"}
{"current_steps": 200, "total_steps": 632, "loss": 0.7837, "learning_rate": 3.862986930406669e-05, "epoch": 1.26, "percentage": 31.65, "elapsed_time": "3:59:48", "remaining_time": "8:37:58"}
{"current_steps": 200, "total_steps": 632, "eval_loss": 0.8230095505714417, "epoch": 1.26, "percentage": 31.65, "elapsed_time": "4:14:17", "remaining_time": "9:09:16"}
{"current_steps": 220, "total_steps": 632, "loss": 0.7938, "learning_rate": 3.6482432230574446e-05, "epoch": 1.39, "percentage": 34.81, "elapsed_time": "4:36:16", "remaining_time": "8:37:23"}
{"current_steps": 240, "total_steps": 632, "loss": 0.7819, "learning_rate": 3.4221598113100195e-05, "epoch": 1.52, "percentage": 37.97, "elapsed_time": "4:58:13", "remaining_time": "8:07:06"}
{"current_steps": 260, "total_steps": 632, "loss": 0.7877, "learning_rate": 3.186969426877563e-05, "epoch": 1.64, "percentage": 41.14, "elapsed_time": "5:20:11", "remaining_time": "7:38:07"}
{"current_steps": 280, "total_steps": 632, "loss": 0.7918, "learning_rate": 2.9449947391938766e-05, "epoch": 1.77, "percentage": 44.3, "elapsed_time": "5:42:00", "remaining_time": "7:09:57"}
{"current_steps": 300, "total_steps": 632, "loss": 0.7824, "learning_rate": 2.6986254174292862e-05, "epoch": 1.9, "percentage": 47.47, "elapsed_time": "6:04:00", "remaining_time": "6:42:49"}
{"current_steps": 300, "total_steps": 632, "eval_loss": 0.8058096766471863, "epoch": 1.9, "percentage": 47.47, "elapsed_time": "6:18:27", "remaining_time": "6:58:50"}
{"current_steps": 320, "total_steps": 632, "loss": 0.7741, "learning_rate": 2.4502945308373246e-05, "epoch": 2.02, "percentage": 50.63, "elapsed_time": "6:40:24", "remaining_time": "6:30:24"}
{"current_steps": 340, "total_steps": 632, "loss": 0.7369, "learning_rate": 2.2024545204952383e-05, "epoch": 2.15, "percentage": 53.8, "elapsed_time": "7:02:25", "remaining_time": "6:02:47"}
{"current_steps": 360, "total_steps": 632, "loss": 0.7365, "learning_rate": 1.957552979734205e-05, "epoch": 2.27, "percentage": 56.96, "elapsed_time": "7:24:19", "remaining_time": "5:35:42"}
{"current_steps": 380, "total_steps": 632, "loss": 0.7463, "learning_rate": 1.7180084824444325e-05, "epoch": 2.4, "percentage": 60.13, "elapsed_time": "7:46:14", "remaining_time": "5:09:11"}
{"current_steps": 400, "total_steps": 632, "loss": 0.7401, "learning_rate": 1.4861866979675154e-05, "epoch": 2.53, "percentage": 63.29, "elapsed_time": "8:08:15", "remaining_time": "4:43:11"}
{"current_steps": 400, "total_steps": 632, "eval_loss": 0.8059037923812866, "epoch": 2.53, "percentage": 63.29, "elapsed_time": "8:22:49", "remaining_time": "4:51:38"}
{"current_steps": 420, "total_steps": 632, "loss": 0.7332, "learning_rate": 1.2643770284581929e-05, "epoch": 2.65, "percentage": 66.46, "elapsed_time": "8:44:49", "remaining_time": "4:24:54"}
{"current_steps": 440, "total_steps": 632, "loss": 0.7364, "learning_rate": 1.0547699994378787e-05, "epoch": 2.78, "percentage": 69.62, "elapsed_time": "9:06:48", "remaining_time": "3:58:36"}
{"current_steps": 460, "total_steps": 632, "loss": 0.7318, "learning_rate": 8.594356268240616e-06, "epoch": 2.91, "percentage": 72.78, "elapsed_time": "9:28:46", "remaining_time": "3:32:40"}
{"current_steps": 480, "total_steps": 632, "loss": 0.7222, "learning_rate": 6.803029740762648e-06, "epoch": 3.03, "percentage": 75.95, "elapsed_time": "9:50:39", "remaining_time": "3:07:02"}
{"current_steps": 500, "total_steps": 632, "loss": 0.7101, "learning_rate": 5.191411013460645e-06, "epoch": 3.16, "percentage": 79.11, "elapsed_time": "10:12:38", "remaining_time": "2:41:44"}
{"current_steps": 500, "total_steps": 632, "eval_loss": 0.807178795337677, "epoch": 3.16, "percentage": 79.11, "elapsed_time": "10:27:05", "remaining_time": "2:45:33"}
{"current_steps": 520, "total_steps": 632, "loss": 0.7135, "learning_rate": 3.775415947715899e-06, "epoch": 3.29, "percentage": 82.28, "elapsed_time": "10:49:03", "remaining_time": "2:19:47"}
{"current_steps": 540, "total_steps": 632, "loss": 0.7045, "learning_rate": 2.5690284845196923e-06, "epoch": 3.41, "percentage": 85.44, "elapsed_time": "11:10:58", "remaining_time": "1:54:18"}
{"current_steps": 560, "total_steps": 632, "loss": 0.7057, "learning_rate": 1.5841625432818057e-06, "epoch": 3.54, "percentage": 88.61, "elapsed_time": "11:32:56", "remaining_time": "1:29:05"}
{"current_steps": 580, "total_steps": 632, "loss": 0.7044, "learning_rate": 8.305443635490711e-07, "epoch": 3.67, "percentage": 91.77, "elapsed_time": "11:54:45", "remaining_time": "1:04:04"}
{"current_steps": 600, "total_steps": 632, "loss": 0.7037, "learning_rate": 3.1561645159166597e-07, "epoch": 3.79, "percentage": 94.94, "elapsed_time": "12:16:39", "remaining_time": "0:39:17"}
{"current_steps": 600, "total_steps": 632, "eval_loss": 0.8061766028404236, "epoch": 3.79, "percentage": 94.94, "elapsed_time": "12:31:07", "remaining_time": "0:40:03"}
{"current_steps": 620, "total_steps": 632, "loss": 0.7027, "learning_rate": 4.4464080451675494e-08, "epoch": 3.92, "percentage": 98.1, "elapsed_time": "12:53:10", "remaining_time": "0:14:57"}
{"current_steps": 632, "total_steps": 632, "epoch": 3.99, "percentage": 100.0, "elapsed_time": "13:06:25", "remaining_time": "0:00:00"}