oh_v3-1_only_slim_orca_550k / trainer_log.jsonl
esfrankel17's picture
Training in progress, epoch 2
ef391c5 verified
{"current_steps": 10, "total_steps": 321, "loss": 1.0737, "learning_rate": 5e-06, "epoch": 0.09313154831199069, "percentage": 3.12, "elapsed_time": "0:05:15", "remaining_time": "2:43:19"}
{"current_steps": 20, "total_steps": 321, "loss": 0.853, "learning_rate": 5e-06, "epoch": 0.18626309662398138, "percentage": 6.23, "elapsed_time": "0:10:23", "remaining_time": "2:36:22"}
{"current_steps": 30, "total_steps": 321, "loss": 0.808, "learning_rate": 5e-06, "epoch": 0.27939464493597205, "percentage": 9.35, "elapsed_time": "0:15:32", "remaining_time": "2:30:41"}
{"current_steps": 40, "total_steps": 321, "loss": 0.7705, "learning_rate": 5e-06, "epoch": 0.37252619324796277, "percentage": 12.46, "elapsed_time": "0:20:41", "remaining_time": "2:25:22"}
{"current_steps": 50, "total_steps": 321, "loss": 0.7493, "learning_rate": 5e-06, "epoch": 0.46565774155995343, "percentage": 15.58, "elapsed_time": "0:25:48", "remaining_time": "2:19:54"}
{"current_steps": 60, "total_steps": 321, "loss": 0.723, "learning_rate": 5e-06, "epoch": 0.5587892898719441, "percentage": 18.69, "elapsed_time": "0:30:56", "remaining_time": "2:14:34"}
{"current_steps": 70, "total_steps": 321, "loss": 0.7029, "learning_rate": 5e-06, "epoch": 0.6519208381839348, "percentage": 21.81, "elapsed_time": "0:36:03", "remaining_time": "2:09:18"}
{"current_steps": 80, "total_steps": 321, "loss": 0.7045, "learning_rate": 5e-06, "epoch": 0.7450523864959255, "percentage": 24.92, "elapsed_time": "0:41:11", "remaining_time": "2:04:05"}
{"current_steps": 90, "total_steps": 321, "loss": 0.6913, "learning_rate": 5e-06, "epoch": 0.8381839348079162, "percentage": 28.04, "elapsed_time": "0:46:18", "remaining_time": "1:58:50"}
{"current_steps": 100, "total_steps": 321, "loss": 0.6853, "learning_rate": 5e-06, "epoch": 0.9313154831199069, "percentage": 31.15, "elapsed_time": "0:51:26", "remaining_time": "1:53:41"}
{"current_steps": 107, "total_steps": 321, "eval_loss": 0.6684155464172363, "epoch": 0.9965075669383003, "percentage": 33.33, "elapsed_time": "0:56:25", "remaining_time": "1:52:50"}
{"current_steps": 110, "total_steps": 321, "loss": 0.7062, "learning_rate": 5e-06, "epoch": 1.0244470314318976, "percentage": 34.27, "elapsed_time": "0:58:46", "remaining_time": "1:52:43"}
{"current_steps": 120, "total_steps": 321, "loss": 0.6098, "learning_rate": 5e-06, "epoch": 1.1175785797438882, "percentage": 37.38, "elapsed_time": "1:03:54", "remaining_time": "1:47:03"}
{"current_steps": 130, "total_steps": 321, "loss": 0.6019, "learning_rate": 5e-06, "epoch": 1.210710128055879, "percentage": 40.5, "elapsed_time": "1:09:01", "remaining_time": "1:41:24"}
{"current_steps": 140, "total_steps": 321, "loss": 0.6043, "learning_rate": 5e-06, "epoch": 1.3038416763678695, "percentage": 43.61, "elapsed_time": "1:14:10", "remaining_time": "1:35:53"}
{"current_steps": 150, "total_steps": 321, "loss": 0.6023, "learning_rate": 5e-06, "epoch": 1.3969732246798603, "percentage": 46.73, "elapsed_time": "1:19:18", "remaining_time": "1:30:24"}
{"current_steps": 160, "total_steps": 321, "loss": 0.6021, "learning_rate": 5e-06, "epoch": 1.490104772991851, "percentage": 49.84, "elapsed_time": "1:24:27", "remaining_time": "1:24:59"}
{"current_steps": 170, "total_steps": 321, "loss": 0.6002, "learning_rate": 5e-06, "epoch": 1.5832363213038416, "percentage": 52.96, "elapsed_time": "1:29:34", "remaining_time": "1:19:33"}
{"current_steps": 180, "total_steps": 321, "loss": 0.5935, "learning_rate": 5e-06, "epoch": 1.6763678696158324, "percentage": 56.07, "elapsed_time": "1:34:41", "remaining_time": "1:14:10"}
{"current_steps": 190, "total_steps": 321, "loss": 0.589, "learning_rate": 5e-06, "epoch": 1.7694994179278232, "percentage": 59.19, "elapsed_time": "1:39:50", "remaining_time": "1:08:50"}
{"current_steps": 200, "total_steps": 321, "loss": 0.5954, "learning_rate": 5e-06, "epoch": 1.8626309662398137, "percentage": 62.31, "elapsed_time": "1:44:57", "remaining_time": "1:03:30"}
{"current_steps": 210, "total_steps": 321, "loss": 0.591, "learning_rate": 5e-06, "epoch": 1.9557625145518043, "percentage": 65.42, "elapsed_time": "1:50:05", "remaining_time": "0:58:11"}
{"current_steps": 214, "total_steps": 321, "eval_loss": 0.6565061807632446, "epoch": 1.9930151338766007, "percentage": 66.67, "elapsed_time": "1:53:44", "remaining_time": "0:56:52"}
{"current_steps": 220, "total_steps": 321, "loss": 0.5986, "learning_rate": 5e-06, "epoch": 2.0488940628637953, "percentage": 68.54, "elapsed_time": "1:57:39", "remaining_time": "0:54:00"}
{"current_steps": 230, "total_steps": 321, "loss": 0.5179, "learning_rate": 5e-06, "epoch": 2.142025611175786, "percentage": 71.65, "elapsed_time": "2:02:46", "remaining_time": "0:48:34"}
{"current_steps": 240, "total_steps": 321, "loss": 0.5178, "learning_rate": 5e-06, "epoch": 2.2351571594877764, "percentage": 74.77, "elapsed_time": "2:07:53", "remaining_time": "0:43:09"}
{"current_steps": 250, "total_steps": 321, "loss": 0.513, "learning_rate": 5e-06, "epoch": 2.328288707799767, "percentage": 77.88, "elapsed_time": "2:13:01", "remaining_time": "0:37:46"}
{"current_steps": 260, "total_steps": 321, "loss": 0.5137, "learning_rate": 5e-06, "epoch": 2.421420256111758, "percentage": 81.0, "elapsed_time": "2:18:07", "remaining_time": "0:32:24"}
{"current_steps": 270, "total_steps": 321, "loss": 0.5263, "learning_rate": 5e-06, "epoch": 2.5145518044237485, "percentage": 84.11, "elapsed_time": "2:23:14", "remaining_time": "0:27:03"}
{"current_steps": 280, "total_steps": 321, "loss": 0.5201, "learning_rate": 5e-06, "epoch": 2.607683352735739, "percentage": 87.23, "elapsed_time": "2:28:23", "remaining_time": "0:21:43"}
{"current_steps": 290, "total_steps": 321, "loss": 0.5181, "learning_rate": 5e-06, "epoch": 2.70081490104773, "percentage": 90.34, "elapsed_time": "2:33:32", "remaining_time": "0:16:24"}
{"current_steps": 300, "total_steps": 321, "loss": 0.5197, "learning_rate": 5e-06, "epoch": 2.7939464493597206, "percentage": 93.46, "elapsed_time": "2:38:39", "remaining_time": "0:11:06"}
{"current_steps": 310, "total_steps": 321, "loss": 0.5274, "learning_rate": 5e-06, "epoch": 2.887077997671711, "percentage": 96.57, "elapsed_time": "2:43:49", "remaining_time": "0:05:48"}
{"current_steps": 320, "total_steps": 321, "loss": 0.5134, "learning_rate": 5e-06, "epoch": 2.980209545983702, "percentage": 99.69, "elapsed_time": "2:48:58", "remaining_time": "0:00:31"}
{"current_steps": 321, "total_steps": 321, "eval_loss": 0.6689090132713318, "epoch": 2.9895227008149012, "percentage": 100.0, "elapsed_time": "2:51:46", "remaining_time": "0:00:00"}
{"current_steps": 321, "total_steps": 321, "epoch": 2.9895227008149012, "percentage": 100.0, "elapsed_time": "2:52:41", "remaining_time": "0:00:00"}