| {"current_steps": 5, "total_steps": 588, "loss": 3.0938, "learning_rate": 0.0001999771640780308, "epoch": 0.05089058524173028, "percentage": 0.85, "elapsed_time": "0:01:03", "remaining_time": "2:02:25", "throughput": "3018.59", "total_tokens": 190176} | |
| {"current_steps": 10, "total_steps": 588, "loss": 1.471, "learning_rate": 0.0001998844110196681, "epoch": 0.10178117048346055, "percentage": 1.7, "elapsed_time": "0:02:04", "remaining_time": "1:59:39", "throughput": "3048.56", "total_tokens": 378656} | |
| {"current_steps": 15, "total_steps": 588, "loss": 1.2452, "learning_rate": 0.00019972037971811802, "epoch": 0.15267175572519084, "percentage": 2.55, "elapsed_time": "0:03:07", "remaining_time": "1:59:41", "throughput": "3033.10", "total_tokens": 570208} | |
| {"current_steps": 20, "total_steps": 588, "loss": 1.1957, "learning_rate": 0.00019948518722731206, "epoch": 0.2035623409669211, "percentage": 3.4, "elapsed_time": "0:04:12", "remaining_time": "1:59:28", "throughput": "3022.83", "total_tokens": 763040} | |
| {"current_steps": 25, "total_steps": 588, "loss": 1.1319, "learning_rate": 0.0001991790013823246, "epoch": 0.2544529262086514, "percentage": 4.25, "elapsed_time": "0:05:18", "remaining_time": "1:59:27", "throughput": "3013.74", "total_tokens": 959200} | |
| {"current_steps": 30, "total_steps": 588, "loss": 1.1209, "learning_rate": 0.00019880204067960472, "epoch": 0.3053435114503817, "percentage": 5.1, "elapsed_time": "0:06:23", "remaining_time": "1:58:55", "throughput": "3007.56", "total_tokens": 1153728} | |
| {"current_steps": 35, "total_steps": 588, "loss": 1.123, "learning_rate": 0.00019835457412105528, "epoch": 0.356234096692112, "percentage": 5.95, "elapsed_time": "0:07:29", "remaining_time": "1:58:22", "throughput": "3006.75", "total_tokens": 1351616} | |
| {"current_steps": 40, "total_steps": 588, "loss": 1.0532, "learning_rate": 0.00019783692102207155, "epoch": 0.4071246819338422, "percentage": 6.8, "elapsed_time": "0:08:29", "remaining_time": "1:56:22", "throughput": "3017.07", "total_tokens": 1537696} | |
| {"current_steps": 45, "total_steps": 588, "loss": 1.0805, "learning_rate": 0.00019724945078367513, "epoch": 0.4580152671755725, "percentage": 7.65, "elapsed_time": "0:09:33", "remaining_time": "1:55:18", "throughput": "3017.48", "total_tokens": 1730144} | |
| {"current_steps": 50, "total_steps": 588, "loss": 1.0572, "learning_rate": 0.00019659258262890683, "epoch": 0.5089058524173028, "percentage": 8.5, "elapsed_time": "0:10:34", "remaining_time": "1:53:45", "throughput": "3021.63", "total_tokens": 1916608} | |
| {"current_steps": 55, "total_steps": 588, "loss": 1.0069, "learning_rate": 0.00019586678530366606, "epoch": 0.5597964376590331, "percentage": 9.35, "elapsed_time": "0:11:34", "remaining_time": "1:52:08", "throughput": "3029.77", "total_tokens": 2103712} | |
| {"current_steps": 60, "total_steps": 588, "loss": 1.0293, "learning_rate": 0.00019507257674221027, "epoch": 0.6106870229007634, "percentage": 10.2, "elapsed_time": "0:12:34", "remaining_time": "1:50:43", "throughput": "3033.42", "total_tokens": 2289888} | |
| {"current_steps": 65, "total_steps": 588, "loss": 0.9994, "learning_rate": 0.00019421052369755334, "epoch": 0.6615776081424937, "percentage": 11.05, "elapsed_time": "0:13:36", "remaining_time": "1:49:30", "throughput": "3034.40", "total_tokens": 2477888} | |
| {"current_steps": 70, "total_steps": 588, "loss": 1.0159, "learning_rate": 0.0001932812413370265, "epoch": 0.712468193384224, "percentage": 11.9, "elapsed_time": "0:14:35", "remaining_time": "1:47:57", "throughput": "3042.25", "total_tokens": 2662848} | |
| {"current_steps": 75, "total_steps": 588, "loss": 1.0233, "learning_rate": 0.00019228539280329038, "epoch": 0.7633587786259542, "percentage": 12.76, "elapsed_time": "0:15:41", "remaining_time": "1:47:17", "throughput": "3039.84", "total_tokens": 2861024} | |
| {"current_steps": 80, "total_steps": 588, "loss": 1.0109, "learning_rate": 0.00019122368874111172, "epoch": 0.8142493638676844, "percentage": 13.61, "elapsed_time": "0:16:39", "remaining_time": "1:45:46", "throughput": "3045.32", "total_tokens": 3043712} | |
| {"current_steps": 85, "total_steps": 588, "loss": 1.0132, "learning_rate": 0.0001900968867902419, "epoch": 0.8651399491094147, "percentage": 14.46, "elapsed_time": "0:17:42", "remaining_time": "1:44:49", "throughput": "3040.96", "total_tokens": 3231840} | |
| {"current_steps": 90, "total_steps": 588, "loss": 1.0079, "learning_rate": 0.00018890579104475995, "epoch": 0.916030534351145, "percentage": 15.31, "elapsed_time": "0:18:48", "remaining_time": "1:44:06", "throughput": "3037.90", "total_tokens": 3429632} | |
| {"current_steps": 95, "total_steps": 588, "loss": 0.9515, "learning_rate": 0.00018765125147926476, "epoch": 0.9669211195928753, "percentage": 16.16, "elapsed_time": "0:19:53", "remaining_time": "1:43:14", "throughput": "3032.44", "total_tokens": 3620000} | |
| {"current_steps": 100, "total_steps": 588, "loss": 0.93, "learning_rate": 0.00018633416334232753, "epoch": 1.0178117048346056, "percentage": 17.01, "elapsed_time": "0:20:55", "remaining_time": "1:42:05", "throughput": "3035.21", "total_tokens": 3809824} | |
| {"current_steps": 105, "total_steps": 588, "loss": 0.8925, "learning_rate": 0.0001849554665176354, "epoch": 1.0687022900763359, "percentage": 17.86, "elapsed_time": "0:22:02", "remaining_time": "1:41:22", "throughput": "3029.35", "total_tokens": 4005472} | |
| {"current_steps": 110, "total_steps": 588, "loss": 0.8715, "learning_rate": 0.00018351614485328388, "epoch": 1.1195928753180662, "percentage": 18.71, "elapsed_time": "0:23:03", "remaining_time": "1:40:12", "throughput": "3030.17", "total_tokens": 4192480} | |
| {"current_steps": 115, "total_steps": 588, "loss": 0.8569, "learning_rate": 0.0001820172254596956, "epoch": 1.1704834605597965, "percentage": 19.56, "elapsed_time": "0:24:03", "remaining_time": "1:38:58", "throughput": "3033.38", "total_tokens": 4379296} | |
| {"current_steps": 120, "total_steps": 588, "loss": 0.8269, "learning_rate": 0.00018045977797666684, "epoch": 1.2213740458015268, "percentage": 20.41, "elapsed_time": "0:25:05", "remaining_time": "1:37:49", "throughput": "3035.24", "total_tokens": 4568352} | |
| {"current_steps": 125, "total_steps": 588, "loss": 0.84, "learning_rate": 0.00017884491381006478, "epoch": 1.272264631043257, "percentage": 21.26, "elapsed_time": "0:26:06", "remaining_time": "1:36:41", "throughput": "3034.50", "total_tokens": 4752928} | |
| {"current_steps": 130, "total_steps": 588, "loss": 0.8243, "learning_rate": 0.00017717378533872017, "epoch": 1.3231552162849873, "percentage": 22.11, "elapsed_time": "0:27:07", "remaining_time": "1:35:34", "throughput": "3034.43", "total_tokens": 4939008} | |
| {"current_steps": 135, "total_steps": 588, "loss": 0.8624, "learning_rate": 0.00017544758509208146, "epoch": 1.3740458015267176, "percentage": 22.96, "elapsed_time": "0:28:11", "remaining_time": "1:34:35", "throughput": "3035.04", "total_tokens": 5133664} | |
| {"current_steps": 140, "total_steps": 588, "loss": 0.8543, "learning_rate": 0.00017366754489921694, "epoch": 1.424936386768448, "percentage": 23.81, "elapsed_time": "0:29:19", "remaining_time": "1:33:49", "throughput": "3030.74", "total_tokens": 5332000} | |
| {"current_steps": 145, "total_steps": 588, "loss": 0.842, "learning_rate": 0.00017183493500977278, "epoch": 1.4758269720101782, "percentage": 24.66, "elapsed_time": "0:30:19", "remaining_time": "1:32:39", "throughput": "3033.80", "total_tokens": 5520640} | |
| {"current_steps": 150, "total_steps": 588, "loss": 0.8268, "learning_rate": 0.0001699510631875134, "epoch": 1.5267175572519083, "percentage": 25.51, "elapsed_time": "0:31:21", "remaining_time": "1:31:34", "throughput": "3034.62", "total_tokens": 5710528} | |
| {"current_steps": 155, "total_steps": 588, "loss": 0.8238, "learning_rate": 0.00016801727377709194, "epoch": 1.5776081424936388, "percentage": 26.36, "elapsed_time": "0:32:28", "remaining_time": "1:30:43", "throughput": "3031.78", "total_tokens": 5907264} | |
| {"current_steps": 160, "total_steps": 588, "loss": 0.8324, "learning_rate": 0.00016603494674471593, "epoch": 1.6284987277353689, "percentage": 27.21, "elapsed_time": "0:33:30", "remaining_time": "1:29:37", "throughput": "3032.30", "total_tokens": 6095776} | |
| {"current_steps": 165, "total_steps": 588, "loss": 0.8431, "learning_rate": 0.0001640054966933935, "epoch": 1.6793893129770994, "percentage": 28.06, "elapsed_time": "0:34:31", "remaining_time": "1:28:30", "throughput": "3032.36", "total_tokens": 6281280} | |
| {"current_steps": 170, "total_steps": 588, "loss": 0.8054, "learning_rate": 0.00016193037185346224, "epoch": 1.7302798982188294, "percentage": 28.91, "elapsed_time": "0:35:34", "remaining_time": "1:27:27", "throughput": "3032.03", "total_tokens": 6470368} | |
| {"current_steps": 175, "total_steps": 588, "loss": 0.8255, "learning_rate": 0.00015981105304912162, "epoch": 1.78117048346056, "percentage": 29.76, "elapsed_time": "0:36:35", "remaining_time": "1:26:21", "throughput": "3032.84", "total_tokens": 6658944} | |
| {"current_steps": 180, "total_steps": 588, "loss": 0.8334, "learning_rate": 0.0001576490526417059, "epoch": 1.83206106870229, "percentage": 30.61, "elapsed_time": "0:37:32", "remaining_time": "1:25:05", "throughput": "3037.84", "total_tokens": 6842880} | |
| {"current_steps": 185, "total_steps": 588, "loss": 0.8354, "learning_rate": 0.0001554459134504523, "epoch": 1.8829516539440203, "percentage": 31.46, "elapsed_time": "0:38:36", "remaining_time": "1:24:06", "throughput": "3036.53", "total_tokens": 7034432} | |
| {"current_steps": 190, "total_steps": 588, "loss": 0.8294, "learning_rate": 0.00015320320765153367, "epoch": 1.9338422391857506, "percentage": 32.31, "elapsed_time": "0:39:41", "remaining_time": "1:23:09", "throughput": "3035.06", "total_tokens": 7228672} | |
| {"current_steps": 195, "total_steps": 588, "loss": 0.8561, "learning_rate": 0.00015092253565614233, "epoch": 1.984732824427481, "percentage": 33.16, "elapsed_time": "0:40:39", "remaining_time": "1:21:55", "throughput": "3039.76", "total_tokens": 7414144} | |
| {"current_steps": 200, "total_steps": 588, "loss": 0.6968, "learning_rate": 0.00014860552496842494, "epoch": 2.035623409669211, "percentage": 34.01, "elapsed_time": "0:41:39", "remaining_time": "1:20:48", "throughput": "3041.96", "total_tokens": 7603104} | |
| {"current_steps": 205, "total_steps": 588, "loss": 0.6587, "learning_rate": 0.00014625382902408356, "epoch": 2.0865139949109412, "percentage": 34.86, "elapsed_time": "0:42:43", "remaining_time": "1:19:50", "throughput": "3039.80", "total_tokens": 7794016} | |
| {"current_steps": 210, "total_steps": 588, "loss": 0.6276, "learning_rate": 0.00014386912601047213, "epoch": 2.1374045801526718, "percentage": 35.71, "elapsed_time": "0:43:48", "remaining_time": "1:18:51", "throughput": "3038.85", "total_tokens": 7987168} | |
| {"current_steps": 215, "total_steps": 588, "loss": 0.6451, "learning_rate": 0.00014145311766902957, "epoch": 2.188295165394402, "percentage": 36.56, "elapsed_time": "0:44:52", "remaining_time": "1:17:51", "throughput": "3038.06", "total_tokens": 8180224} | |
| {"current_steps": 220, "total_steps": 588, "loss": 0.6776, "learning_rate": 0.00013900752808090468, "epoch": 2.2391857506361323, "percentage": 37.41, "elapsed_time": "0:45:55", "remaining_time": "1:16:48", "throughput": "3038.52", "total_tokens": 8371136} | |
| {"current_steps": 225, "total_steps": 588, "loss": 0.6307, "learning_rate": 0.00013653410243663952, "epoch": 2.2900763358778624, "percentage": 38.27, "elapsed_time": "0:46:53", "remaining_time": "1:15:38", "throughput": "3040.02", "total_tokens": 8552064} | |
| {"current_steps": 230, "total_steps": 588, "loss": 0.6623, "learning_rate": 0.00013403460579078833, "epoch": 2.340966921119593, "percentage": 39.12, "elapsed_time": "0:47:53", "remaining_time": "1:14:32", "throughput": "3040.73", "total_tokens": 8737120} | |
| {"current_steps": 235, "total_steps": 588, "loss": 0.6754, "learning_rate": 0.0001315108218023621, "epoch": 2.391857506361323, "percentage": 39.97, "elapsed_time": "0:48:55", "remaining_time": "1:13:29", "throughput": "3041.73", "total_tokens": 8928608} | |
| {"current_steps": 240, "total_steps": 588, "loss": 0.6518, "learning_rate": 0.0001289645514619963, "epoch": 2.4427480916030535, "percentage": 40.82, "elapsed_time": "0:49:57", "remaining_time": "1:12:26", "throughput": "3042.45", "total_tokens": 9120032} | |
| {"current_steps": 245, "total_steps": 588, "loss": 0.6331, "learning_rate": 0.00012639761180675098, "epoch": 2.4936386768447836, "percentage": 41.67, "elapsed_time": "0:50:57", "remaining_time": "1:11:20", "throughput": "3044.26", "total_tokens": 9307840} | |
| {"current_steps": 250, "total_steps": 588, "loss": 0.6588, "learning_rate": 0.00012381183462345982, "epoch": 2.544529262086514, "percentage": 42.52, "elapsed_time": "0:52:01", "remaining_time": "1:10:20", "throughput": "3042.72", "total_tokens": 9499104} | |
| {"current_steps": 255, "total_steps": 588, "loss": 0.6657, "learning_rate": 0.0001212090651415537, "epoch": 2.595419847328244, "percentage": 43.37, "elapsed_time": "0:53:07", "remaining_time": "1:09:21", "throughput": "3041.93", "total_tokens": 9694752} | |
| {"current_steps": 260, "total_steps": 588, "loss": 0.6707, "learning_rate": 0.00011859116071629149, "epoch": 2.6463104325699747, "percentage": 44.22, "elapsed_time": "0:54:13", "remaining_time": "1:08:24", "throughput": "3039.94", "total_tokens": 9889824} | |
| {"current_steps": 265, "total_steps": 588, "loss": 0.6624, "learning_rate": 0.00011595998950333793, "epoch": 2.6972010178117047, "percentage": 45.07, "elapsed_time": "0:55:18", "remaining_time": "1:07:24", "throughput": "3037.65", "total_tokens": 10080896} | |
| {"current_steps": 270, "total_steps": 588, "loss": 0.6671, "learning_rate": 0.00011331742912563413, "epoch": 2.7480916030534353, "percentage": 45.92, "elapsed_time": "0:56:24", "remaining_time": "1:06:25", "throughput": "3037.08", "total_tokens": 10278144} | |
| {"current_steps": 275, "total_steps": 588, "loss": 0.6654, "learning_rate": 0.00011066536533351202, "epoch": 2.7989821882951653, "percentage": 46.77, "elapsed_time": "0:57:25", "remaining_time": "1:05:21", "throughput": "3037.65", "total_tokens": 10466528} | |
| {"current_steps": 280, "total_steps": 588, "loss": 0.6681, "learning_rate": 0.00010800569065900933, "epoch": 2.849872773536896, "percentage": 47.62, "elapsed_time": "0:58:25", "remaining_time": "1:04:15", "throughput": "3037.96", "total_tokens": 10649024} | |
| {"current_steps": 285, "total_steps": 588, "loss": 0.6718, "learning_rate": 0.0001053403030653449, "epoch": 2.900763358778626, "percentage": 48.47, "elapsed_time": "0:59:25", "remaining_time": "1:03:10", "throughput": "3038.96", "total_tokens": 10834816} | |
| {"current_steps": 290, "total_steps": 588, "loss": 0.637, "learning_rate": 0.00010267110459251823, "epoch": 2.9516539440203564, "percentage": 49.32, "elapsed_time": "1:00:29", "remaining_time": "1:02:09", "throughput": "3038.68", "total_tokens": 11027616} | |
| {"current_steps": 295, "total_steps": 588, "loss": 0.6567, "learning_rate": 0.0001, "epoch": 3.0025445292620865, "percentage": 50.17, "elapsed_time": "1:01:31", "remaining_time": "1:01:06", "throughput": "3038.89", "total_tokens": 11217664} | |
| {"current_steps": 300, "total_steps": 588, "loss": 0.5179, "learning_rate": 9.73288954074818e-05, "epoch": 3.053435114503817, "percentage": 51.02, "elapsed_time": "1:02:33", "remaining_time": "1:00:03", "throughput": "3038.68", "total_tokens": 11406624} | |
| {"current_steps": 305, "total_steps": 588, "loss": 0.5329, "learning_rate": 9.46596969346551e-05, "epoch": 3.104325699745547, "percentage": 51.87, "elapsed_time": "1:03:37", "remaining_time": "0:59:02", "throughput": "3037.66", "total_tokens": 11596096} | |
| {"current_steps": 310, "total_steps": 588, "loss": 0.5039, "learning_rate": 9.199430934099068e-05, "epoch": 3.1552162849872776, "percentage": 52.72, "elapsed_time": "1:04:39", "remaining_time": "0:57:59", "throughput": "3038.49", "total_tokens": 11788992} | |
| {"current_steps": 315, "total_steps": 588, "loss": 0.5085, "learning_rate": 8.933463466648798e-05, "epoch": 3.2061068702290076, "percentage": 53.57, "elapsed_time": "1:05:45", "remaining_time": "0:56:59", "throughput": "3038.15", "total_tokens": 11986976} | |
| {"current_steps": 320, "total_steps": 588, "loss": 0.5065, "learning_rate": 8.66825708743659e-05, "epoch": 3.2569974554707377, "percentage": 54.42, "elapsed_time": "1:06:49", "remaining_time": "0:55:57", "throughput": "3037.65", "total_tokens": 12179136} | |
| {"current_steps": 325, "total_steps": 588, "loss": 0.5059, "learning_rate": 8.404001049666211e-05, "epoch": 3.3078880407124682, "percentage": 55.27, "elapsed_time": "1:07:50", "remaining_time": "0:54:54", "throughput": "3038.27", "total_tokens": 12368608} | |
| {"current_steps": 330, "total_steps": 588, "loss": 0.5089, "learning_rate": 8.140883928370855e-05, "epoch": 3.3587786259541983, "percentage": 56.12, "elapsed_time": "1:08:53", "remaining_time": "0:53:51", "throughput": "3037.77", "total_tokens": 12556256} | |
| {"current_steps": 335, "total_steps": 588, "loss": 0.4853, "learning_rate": 7.879093485844635e-05, "epoch": 3.409669211195929, "percentage": 56.97, "elapsed_time": "1:09:56", "remaining_time": "0:52:49", "throughput": "3037.63", "total_tokens": 12746784} | |
| {"current_steps": 340, "total_steps": 588, "loss": 0.5119, "learning_rate": 7.618816537654018e-05, "epoch": 3.460559796437659, "percentage": 57.82, "elapsed_time": "1:11:03", "remaining_time": "0:51:49", "throughput": "3036.23", "total_tokens": 12944096} | |
| {"current_steps": 345, "total_steps": 588, "loss": 0.5024, "learning_rate": 7.360238819324903e-05, "epoch": 3.5114503816793894, "percentage": 58.67, "elapsed_time": "1:12:07", "remaining_time": "0:50:48", "throughput": "3036.34", "total_tokens": 13139904} | |
| {"current_steps": 350, "total_steps": 588, "loss": 0.5074, "learning_rate": 7.10354485380037e-05, "epoch": 3.5623409669211195, "percentage": 59.52, "elapsed_time": "1:13:11", "remaining_time": "0:49:45", "throughput": "3035.31", "total_tokens": 13328480} | |
| {"current_steps": 355, "total_steps": 588, "loss": 0.5078, "learning_rate": 6.848917819763793e-05, "epoch": 3.61323155216285, "percentage": 60.37, "elapsed_time": "1:14:11", "remaining_time": "0:48:41", "throughput": "3036.58", "total_tokens": 13518112} | |
| {"current_steps": 360, "total_steps": 588, "loss": 0.5211, "learning_rate": 6.596539420921171e-05, "epoch": 3.66412213740458, "percentage": 61.22, "elapsed_time": "1:15:12", "remaining_time": "0:47:37", "throughput": "3037.21", "total_tokens": 13705632} | |
| {"current_steps": 365, "total_steps": 588, "loss": 0.5042, "learning_rate": 6.34658975633605e-05, "epoch": 3.7150127226463106, "percentage": 62.07, "elapsed_time": "1:16:16", "remaining_time": "0:46:35", "throughput": "3036.89", "total_tokens": 13896896} | |
| {"current_steps": 370, "total_steps": 588, "loss": 0.5038, "learning_rate": 6.0992471919095315e-05, "epoch": 3.7659033078880406, "percentage": 62.93, "elapsed_time": "1:17:15", "remaining_time": "0:45:30", "throughput": "3038.14", "total_tokens": 14081888} | |
| {"current_steps": 375, "total_steps": 588, "loss": 0.5046, "learning_rate": 5.854688233097045e-05, "epoch": 3.816793893129771, "percentage": 63.78, "elapsed_time": "1:18:17", "remaining_time": "0:44:28", "throughput": "3038.07", "total_tokens": 14270560} | |
| {"current_steps": 380, "total_steps": 588, "loss": 0.5196, "learning_rate": 5.613087398952792e-05, "epoch": 3.867684478371501, "percentage": 64.63, "elapsed_time": "1:19:18", "remaining_time": "0:43:24", "throughput": "3038.83", "total_tokens": 14461120} | |
| {"current_steps": 385, "total_steps": 588, "loss": 0.4898, "learning_rate": 5.37461709759165e-05, "epoch": 3.9185750636132317, "percentage": 65.48, "elapsed_time": "1:20:19", "remaining_time": "0:42:21", "throughput": "3039.49", "total_tokens": 14648096} | |
| {"current_steps": 390, "total_steps": 588, "loss": 0.5283, "learning_rate": 5.139447503157513e-05, "epoch": 3.969465648854962, "percentage": 66.33, "elapsed_time": "1:21:23", "remaining_time": "0:41:19", "throughput": "3038.89", "total_tokens": 14839840} | |
| {"current_steps": 395, "total_steps": 588, "loss": 0.4485, "learning_rate": 4.9077464343857694e-05, "epoch": 4.020356234096692, "percentage": 67.18, "elapsed_time": "1:22:22", "remaining_time": "0:40:14", "throughput": "3039.20", "total_tokens": 15020448} | |
| {"current_steps": 400, "total_steps": 588, "loss": 0.4041, "learning_rate": 4.6796792348466356e-05, "epoch": 4.071246819338422, "percentage": 68.03, "elapsed_time": "1:23:25", "remaining_time": "0:39:12", "throughput": "3038.92", "total_tokens": 15210368} | |
| {"current_steps": 405, "total_steps": 588, "loss": 0.4028, "learning_rate": 4.4554086549547715e-05, "epoch": 4.122137404580153, "percentage": 68.88, "elapsed_time": "1:24:28", "remaining_time": "0:38:10", "throughput": "3038.51", "total_tokens": 15399296} | |
| {"current_steps": 410, "total_steps": 588, "loss": 0.4001, "learning_rate": 4.23509473582941e-05, "epoch": 4.1730279898218825, "percentage": 69.73, "elapsed_time": "1:25:24", "remaining_time": "0:37:04", "throughput": "3040.50", "total_tokens": 15580416} | |
| {"current_steps": 415, "total_steps": 588, "loss": 0.3725, "learning_rate": 4.0188946950878404e-05, "epoch": 4.223918575063613, "percentage": 70.58, "elapsed_time": "1:26:23", "remaining_time": "0:36:00", "throughput": "3041.84", "total_tokens": 15768480} | |
| {"current_steps": 420, "total_steps": 588, "loss": 0.3941, "learning_rate": 3.806962814653779e-05, "epoch": 4.2748091603053435, "percentage": 71.43, "elapsed_time": "1:27:22", "remaining_time": "0:34:56", "throughput": "3042.81", "total_tokens": 15951072} | |
| {"current_steps": 425, "total_steps": 588, "loss": 0.4274, "learning_rate": 3.5994503306606497e-05, "epoch": 4.325699745547074, "percentage": 72.28, "elapsed_time": "1:28:27", "remaining_time": "0:33:55", "throughput": "3042.49", "total_tokens": 16148512} | |
| {"current_steps": 430, "total_steps": 588, "loss": 0.394, "learning_rate": 3.3965053255284084e-05, "epoch": 4.376590330788804, "percentage": 73.13, "elapsed_time": "1:29:30", "remaining_time": "0:32:53", "throughput": "3043.08", "total_tokens": 16341600} | |
| {"current_steps": 435, "total_steps": 588, "loss": 0.4072, "learning_rate": 3.198272622290804e-05, "epoch": 4.427480916030534, "percentage": 73.98, "elapsed_time": "1:30:33", "remaining_time": "0:31:51", "throughput": "3042.89", "total_tokens": 16533792} | |
| {"current_steps": 440, "total_steps": 588, "loss": 0.4025, "learning_rate": 3.0048936812486615e-05, "epoch": 4.478371501272265, "percentage": 74.83, "elapsed_time": "1:31:39", "remaining_time": "0:30:49", "throughput": "3041.84", "total_tokens": 16728128} | |
| {"current_steps": 445, "total_steps": 588, "loss": 0.3966, "learning_rate": 2.8165064990227252e-05, "epoch": 4.529262086513995, "percentage": 75.68, "elapsed_time": "1:32:45", "remaining_time": "0:29:48", "throughput": "3040.36", "total_tokens": 16922560} | |
| {"current_steps": 450, "total_steps": 588, "loss": 0.394, "learning_rate": 2.6332455100783083e-05, "epoch": 4.580152671755725, "percentage": 76.53, "elapsed_time": "1:33:43", "remaining_time": "0:28:44", "throughput": "3041.81", "total_tokens": 17106272} | |
| {"current_steps": 455, "total_steps": 588, "loss": 0.4152, "learning_rate": 2.4552414907918564e-05, "epoch": 4.631043256997455, "percentage": 77.38, "elapsed_time": "1:34:48", "remaining_time": "0:27:42", "throughput": "3041.01", "total_tokens": 17298240} | |
| {"current_steps": 460, "total_steps": 588, "loss": 0.408, "learning_rate": 2.282621466127982e-05, "epoch": 4.681933842239186, "percentage": 78.23, "elapsed_time": "1:35:52", "remaining_time": "0:26:40", "throughput": "3040.55", "total_tokens": 17490624} | |
| {"current_steps": 465, "total_steps": 588, "loss": 0.3924, "learning_rate": 2.1155086189935224e-05, "epoch": 4.732824427480916, "percentage": 79.08, "elapsed_time": "1:36:52", "remaining_time": "0:25:37", "throughput": "3041.51", "total_tokens": 17677632} | |
| {"current_steps": 470, "total_steps": 588, "loss": 0.4161, "learning_rate": 1.9540222023333166e-05, "epoch": 4.783715012722646, "percentage": 79.93, "elapsed_time": "1:37:54", "remaining_time": "0:24:34", "throughput": "3041.80", "total_tokens": 17867872} | |
| {"current_steps": 475, "total_steps": 588, "loss": 0.3904, "learning_rate": 1.7982774540304403e-05, "epoch": 4.8346055979643765, "percentage": 80.78, "elapsed_time": "1:38:54", "remaining_time": "0:23:31", "throughput": "3042.41", "total_tokens": 18055552} | |
| {"current_steps": 480, "total_steps": 588, "loss": 0.4057, "learning_rate": 1.6483855146716152e-05, "epoch": 4.885496183206107, "percentage": 81.63, "elapsed_time": "1:39:58", "remaining_time": "0:22:29", "throughput": "3042.12", "total_tokens": 18247104} | |
| {"current_steps": 485, "total_steps": 588, "loss": 0.3934, "learning_rate": 1.504453348236461e-05, "epoch": 4.9363867684478375, "percentage": 82.48, "elapsed_time": "1:40:58", "remaining_time": "0:21:26", "throughput": "3042.62", "total_tokens": 18434752} | |
| {"current_steps": 490, "total_steps": 588, "loss": 0.4161, "learning_rate": 1.3665836657672493e-05, "epoch": 4.987277353689567, "percentage": 83.33, "elapsed_time": "1:42:06", "remaining_time": "0:20:25", "throughput": "3041.53", "total_tokens": 18633824} | |
| {"current_steps": 495, "total_steps": 588, "loss": 0.3599, "learning_rate": 1.2348748520735221e-05, "epoch": 5.038167938931298, "percentage": 84.18, "elapsed_time": "1:43:08", "remaining_time": "0:19:22", "throughput": "3042.11", "total_tokens": 18825024} | |
| {"current_steps": 500, "total_steps": 588, "loss": 0.356, "learning_rate": 1.1094208955240081e-05, "epoch": 5.089058524173028, "percentage": 85.03, "elapsed_time": "1:44:09", "remaining_time": "0:18:19", "throughput": "3042.53", "total_tokens": 19013664} | |
| {"current_steps": 505, "total_steps": 588, "loss": 0.3565, "learning_rate": 9.903113209758096e-06, "epoch": 5.139949109414759, "percentage": 85.88, "elapsed_time": "1:45:13", "remaining_time": "0:17:17", "throughput": "3042.17", "total_tokens": 19207456} | |
| {"current_steps": 510, "total_steps": 588, "loss": 0.3469, "learning_rate": 8.776311258888303e-06, "epoch": 5.190839694656488, "percentage": 86.73, "elapsed_time": "1:46:18", "remaining_time": "0:16:15", "throughput": "3041.33", "total_tokens": 19398208} | |
| {"current_steps": 515, "total_steps": 588, "loss": 0.3525, "learning_rate": 7.714607196709634e-06, "epoch": 5.241730279898219, "percentage": 87.59, "elapsed_time": "1:47:19", "remaining_time": "0:15:12", "throughput": "3041.38", "total_tokens": 19585056} | |
| {"current_steps": 520, "total_steps": 588, "loss": 0.3587, "learning_rate": 6.718758662973523e-06, "epoch": 5.292620865139949, "percentage": 88.44, "elapsed_time": "1:48:21", "remaining_time": "0:14:10", "throughput": "3041.73", "total_tokens": 19774784} | |
| {"current_steps": 525, "total_steps": 588, "loss": 0.3501, "learning_rate": 5.789476302446662e-06, "epoch": 5.34351145038168, "percentage": 89.29, "elapsed_time": "1:49:29", "remaining_time": "0:13:08", "throughput": "3040.12", "total_tokens": 19970656} | |
| {"current_steps": 530, "total_steps": 588, "loss": 0.367, "learning_rate": 4.927423257789721e-06, "epoch": 5.3944020356234095, "percentage": 90.14, "elapsed_time": "1:50:33", "remaining_time": "0:12:05", "throughput": "3039.92", "total_tokens": 20164416} | |
| {"current_steps": 535, "total_steps": 588, "loss": 0.3383, "learning_rate": 4.133214696333942e-06, "epoch": 5.44529262086514, "percentage": 90.99, "elapsed_time": "1:51:34", "remaining_time": "0:11:03", "throughput": "3040.17", "total_tokens": 20352544} | |
| {"current_steps": 540, "total_steps": 588, "loss": 0.3545, "learning_rate": 3.40741737109318e-06, "epoch": 5.4961832061068705, "percentage": 91.84, "elapsed_time": "1:52:35", "remaining_time": "0:10:00", "throughput": "3040.50", "total_tokens": 20538848} | |
| {"current_steps": 545, "total_steps": 588, "loss": 0.3318, "learning_rate": 2.7505492163248934e-06, "epoch": 5.5470737913486, "percentage": 92.69, "elapsed_time": "1:53:35", "remaining_time": "0:08:57", "throughput": "3040.71", "total_tokens": 20723936} | |
| {"current_steps": 550, "total_steps": 588, "loss": 0.3671, "learning_rate": 2.1630789779284675e-06, "epoch": 5.597964376590331, "percentage": 93.54, "elapsed_time": "1:54:36", "remaining_time": "0:07:55", "throughput": "3041.42", "total_tokens": 20913376} | |
| {"current_steps": 555, "total_steps": 588, "loss": 0.3548, "learning_rate": 1.6454258789447285e-06, "epoch": 5.648854961832061, "percentage": 94.39, "elapsed_time": "1:55:42", "remaining_time": "0:06:52", "throughput": "3041.41", "total_tokens": 21114240} | |
| {"current_steps": 560, "total_steps": 588, "loss": 0.3575, "learning_rate": 1.1979593203953033e-06, "epoch": 5.699745547073792, "percentage": 95.24, "elapsed_time": "1:56:46", "remaining_time": "0:05:50", "throughput": "3041.17", "total_tokens": 21306752} | |
| {"current_steps": 565, "total_steps": 588, "loss": 0.3632, "learning_rate": 8.209986176753948e-07, "epoch": 5.750636132315521, "percentage": 96.09, "elapsed_time": "1:57:47", "remaining_time": "0:04:47", "throughput": "3041.65", "total_tokens": 21495328} | |
| {"current_steps": 570, "total_steps": 588, "loss": 0.3394, "learning_rate": 5.148127726879448e-07, "epoch": 5.801526717557252, "percentage": 96.94, "elapsed_time": "1:58:46", "remaining_time": "0:03:45", "throughput": "3042.03", "total_tokens": 21680448} | |
| {"current_steps": 575, "total_steps": 588, "loss": 0.346, "learning_rate": 2.7962028188198706e-07, "epoch": 5.852417302798982, "percentage": 97.79, "elapsed_time": "1:59:46", "remaining_time": "0:02:42", "throughput": "3042.67", "total_tokens": 21865760} | |
| {"current_steps": 580, "total_steps": 588, "loss": 0.3537, "learning_rate": 1.1558898033191546e-07, "epoch": 5.903307888040713, "percentage": 98.64, "elapsed_time": "2:00:48", "remaining_time": "0:01:39", "throughput": "3042.91", "total_tokens": 22057632} | |
| {"current_steps": 585, "total_steps": 588, "loss": 0.3421, "learning_rate": 2.2835921969210917e-08, "epoch": 5.9541984732824424, "percentage": 99.49, "elapsed_time": "2:01:49", "remaining_time": "0:00:37", "throughput": "3043.71", "total_tokens": 22248480} | |
| {"current_steps": 588, "total_steps": 588, "epoch": 5.984732824427481, "percentage": 100.0, "elapsed_time": "2:02:26", "remaining_time": "0:00:00", "throughput": "3043.96", "total_tokens": 22361536} | |