| {"current_steps": 10, "total_steps": 2691, "loss": 0.3679, "learning_rate": 3.7037037037037036e-07, "epoch": 0.011148272017837236, "percentage": 0.37, "elapsed_time": "0:02:27", "remaining_time": "11:00:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 20, "total_steps": 2691, "loss": 0.3385, "learning_rate": 7.407407407407407e-07, "epoch": 0.022296544035674472, "percentage": 0.74, "elapsed_time": "0:04:34", "remaining_time": "10:10:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 30, "total_steps": 2691, "loss": 0.3008, "learning_rate": 1.111111111111111e-06, "epoch": 0.033444816053511704, "percentage": 1.11, "elapsed_time": "0:06:48", "remaining_time": "10:04:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 40, "total_steps": 2691, "loss": 0.2689, "learning_rate": 1.4814814814814815e-06, "epoch": 0.044593088071348944, "percentage": 1.49, "elapsed_time": "0:08:56", "remaining_time": "9:52:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 50, "total_steps": 2691, "loss": 0.2506, "learning_rate": 1.8518518518518519e-06, "epoch": 0.055741360089186176, "percentage": 1.86, "elapsed_time": "0:11:03", "remaining_time": "9:44:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 60, "total_steps": 2691, "loss": 0.24, "learning_rate": 2.222222222222222e-06, "epoch": 0.06688963210702341, "percentage": 2.23, "elapsed_time": "0:13:02", "remaining_time": "9:31:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 70, "total_steps": 2691, "loss": 0.2374, "learning_rate": 2.5925925925925925e-06, "epoch": 0.07803790412486064, "percentage": 2.6, "elapsed_time": "0:15:02", "remaining_time": "9:23:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 80, "total_steps": 2691, "loss": 0.2343, "learning_rate": 2.962962962962963e-06, "epoch": 0.08918617614269789, "percentage": 2.97, "elapsed_time": "0:16:53", "remaining_time": "9:11:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 90, "total_steps": 2691, "loss": 0.2279, "learning_rate": 3.3333333333333333e-06, "epoch": 0.10033444816053512, "percentage": 3.34, "elapsed_time": "0:19:08", "remaining_time": "9:13:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 100, "total_steps": 2691, "loss": 0.2263, "learning_rate": 3.7037037037037037e-06, "epoch": 0.11148272017837235, "percentage": 3.72, "elapsed_time": "0:21:10", "remaining_time": "9:08:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 110, "total_steps": 2691, "loss": 0.2239, "learning_rate": 4.074074074074074e-06, "epoch": 0.12263099219620958, "percentage": 4.09, "elapsed_time": "0:23:47", "remaining_time": "9:18:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 120, "total_steps": 2691, "loss": 0.2232, "learning_rate": 4.444444444444444e-06, "epoch": 0.13377926421404682, "percentage": 4.46, "elapsed_time": "0:25:59", "remaining_time": "9:16:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 130, "total_steps": 2691, "loss": 0.2231, "learning_rate": 4.814814814814815e-06, "epoch": 0.14492753623188406, "percentage": 4.83, "elapsed_time": "0:28:04", "remaining_time": "9:13:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 140, "total_steps": 2691, "loss": 0.2215, "learning_rate": 5.185185185185185e-06, "epoch": 0.15607580824972128, "percentage": 5.2, "elapsed_time": "0:30:23", "remaining_time": "9:13:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 150, "total_steps": 2691, "loss": 0.2203, "learning_rate": 5.555555555555557e-06, "epoch": 0.16722408026755853, "percentage": 5.57, "elapsed_time": "0:32:22", "remaining_time": "9:08:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 160, "total_steps": 2691, "loss": 0.2186, "learning_rate": 5.925925925925926e-06, "epoch": 0.17837235228539577, "percentage": 5.95, "elapsed_time": "0:34:32", "remaining_time": "9:06:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 170, "total_steps": 2691, "loss": 0.2184, "learning_rate": 6.296296296296297e-06, "epoch": 0.189520624303233, "percentage": 6.32, "elapsed_time": "0:36:31", "remaining_time": "9:01:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 180, "total_steps": 2691, "loss": 0.22, "learning_rate": 6.666666666666667e-06, "epoch": 0.20066889632107024, "percentage": 6.69, "elapsed_time": "0:38:35", "remaining_time": "8:58:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 190, "total_steps": 2691, "loss": 0.2181, "learning_rate": 7.0370370370370375e-06, "epoch": 0.21181716833890746, "percentage": 7.06, "elapsed_time": "0:40:45", "remaining_time": "8:56:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 200, "total_steps": 2691, "loss": 0.221, "learning_rate": 7.4074074074074075e-06, "epoch": 0.2229654403567447, "percentage": 7.43, "elapsed_time": "0:42:59", "remaining_time": "8:55:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 210, "total_steps": 2691, "loss": 0.2188, "learning_rate": 7.77777777777778e-06, "epoch": 0.23411371237458195, "percentage": 7.8, "elapsed_time": "0:45:35", "remaining_time": "8:58:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 220, "total_steps": 2691, "loss": 0.2191, "learning_rate": 8.148148148148148e-06, "epoch": 0.24526198439241917, "percentage": 8.18, "elapsed_time": "0:47:42", "remaining_time": "8:55:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 230, "total_steps": 2691, "loss": 0.222, "learning_rate": 8.518518518518519e-06, "epoch": 0.2564102564102564, "percentage": 8.55, "elapsed_time": "0:49:45", "remaining_time": "8:52:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 240, "total_steps": 2691, "loss": 0.2189, "learning_rate": 8.888888888888888e-06, "epoch": 0.26755852842809363, "percentage": 8.92, "elapsed_time": "0:51:55", "remaining_time": "8:50:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 250, "total_steps": 2691, "loss": 0.2177, "learning_rate": 9.25925925925926e-06, "epoch": 0.2787068004459309, "percentage": 9.29, "elapsed_time": "0:53:55", "remaining_time": "8:46:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 260, "total_steps": 2691, "loss": 0.2202, "learning_rate": 9.62962962962963e-06, "epoch": 0.2898550724637681, "percentage": 9.66, "elapsed_time": "0:56:04", "remaining_time": "8:44:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 270, "total_steps": 2691, "loss": 0.217, "learning_rate": 1e-05, "epoch": 0.3010033444816054, "percentage": 10.03, "elapsed_time": "0:58:15", "remaining_time": "8:42:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 280, "total_steps": 2691, "loss": 0.2184, "learning_rate": 9.999579036849509e-06, "epoch": 0.31215161649944256, "percentage": 10.41, "elapsed_time": "1:00:13", "remaining_time": "8:38:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 290, "total_steps": 2691, "loss": 0.2165, "learning_rate": 9.998316218282025e-06, "epoch": 0.3232998885172798, "percentage": 10.78, "elapsed_time": "1:02:08", "remaining_time": "8:34:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 300, "total_steps": 2691, "loss": 0.2153, "learning_rate": 9.996211756937579e-06, "epoch": 0.33444816053511706, "percentage": 11.15, "elapsed_time": "1:04:22", "remaining_time": "8:33:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 310, "total_steps": 2691, "loss": 0.2183, "learning_rate": 9.993266007176446e-06, "epoch": 0.3455964325529543, "percentage": 11.52, "elapsed_time": "1:07:02", "remaining_time": "8:34:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 320, "total_steps": 2691, "loss": 0.2168, "learning_rate": 9.989479465019462e-06, "epoch": 0.35674470457079155, "percentage": 11.89, "elapsed_time": "1:09:15", "remaining_time": "8:33:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 330, "total_steps": 2691, "loss": 0.2164, "learning_rate": 9.984852768064516e-06, "epoch": 0.36789297658862874, "percentage": 12.26, "elapsed_time": "1:11:14", "remaining_time": "8:29:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 340, "total_steps": 2691, "loss": 0.2157, "learning_rate": 9.979386695379179e-06, "epoch": 0.379041248606466, "percentage": 12.63, "elapsed_time": "1:13:24", "remaining_time": "8:27:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 350, "total_steps": 2691, "loss": 0.2144, "learning_rate": 9.973082167369521e-06, "epoch": 0.39018952062430323, "percentage": 13.01, "elapsed_time": "1:15:33", "remaining_time": "8:25:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 360, "total_steps": 2691, "loss": 0.2109, "learning_rate": 9.965940245625131e-06, "epoch": 0.4013377926421405, "percentage": 13.38, "elapsed_time": "1:17:30", "remaining_time": "8:21:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 370, "total_steps": 2691, "loss": 0.2159, "learning_rate": 9.95796213274036e-06, "epoch": 0.4124860646599777, "percentage": 13.75, "elapsed_time": "1:19:36", "remaining_time": "8:19:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 380, "total_steps": 2691, "loss": 0.2154, "learning_rate": 9.949149172111825e-06, "epoch": 0.4236343366778149, "percentage": 14.12, "elapsed_time": "1:21:41", "remaining_time": "8:16:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 390, "total_steps": 2691, "loss": 0.215, "learning_rate": 9.93950284771219e-06, "epoch": 0.43478260869565216, "percentage": 14.49, "elapsed_time": "1:23:50", "remaining_time": "8:14:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 400, "total_steps": 2691, "loss": 0.2156, "learning_rate": 9.929024783840302e-06, "epoch": 0.4459308807134894, "percentage": 14.86, "elapsed_time": "1:26:03", "remaining_time": "8:12:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 410, "total_steps": 2691, "loss": 0.2135, "learning_rate": 9.917716744847674e-06, "epoch": 0.45707915273132665, "percentage": 15.24, "elapsed_time": "1:28:32", "remaining_time": "8:12:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 420, "total_steps": 2691, "loss": 0.2127, "learning_rate": 9.905580634841388e-06, "epoch": 0.4682274247491639, "percentage": 15.61, "elapsed_time": "1:30:41", "remaining_time": "8:10:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 430, "total_steps": 2691, "loss": 0.2152, "learning_rate": 9.892618497363491e-06, "epoch": 0.4793756967670011, "percentage": 15.98, "elapsed_time": "1:32:51", "remaining_time": "8:08:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 440, "total_steps": 2691, "loss": 0.2121, "learning_rate": 9.878832515046873e-06, "epoch": 0.49052396878483834, "percentage": 16.35, "elapsed_time": "1:34:53", "remaining_time": "8:05:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 450, "total_steps": 2691, "loss": 0.2124, "learning_rate": 9.864225009247753e-06, "epoch": 0.5016722408026756, "percentage": 16.72, "elapsed_time": "1:37:04", "remaining_time": "8:03:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 460, "total_steps": 2691, "loss": 0.2116, "learning_rate": 9.848798439654797e-06, "epoch": 0.5128205128205128, "percentage": 17.09, "elapsed_time": "1:39:09", "remaining_time": "8:00:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 470, "total_steps": 2691, "loss": 0.2107, "learning_rate": 9.832555403874937e-06, "epoch": 0.5239687848383501, "percentage": 17.47, "elapsed_time": "1:41:18", "remaining_time": "7:58:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 480, "total_steps": 2691, "loss": 0.2092, "learning_rate": 9.815498636995982e-06, "epoch": 0.5351170568561873, "percentage": 17.84, "elapsed_time": "1:43:25", "remaining_time": "7:56:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 490, "total_steps": 2691, "loss": 0.2098, "learning_rate": 9.797631011126061e-06, "epoch": 0.5462653288740246, "percentage": 18.21, "elapsed_time": "1:45:33", "remaining_time": "7:54:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 500, "total_steps": 2691, "loss": 0.2107, "learning_rate": 9.778955534910004e-06, "epoch": 0.5574136008918618, "percentage": 18.58, "elapsed_time": "1:47:44", "remaining_time": "7:52:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 500, "total_steps": 2691, "eval_loss": 0.21421977877616882, "epoch": 0.5574136008918618, "percentage": 18.58, "elapsed_time": "1:47:49", "remaining_time": "7:52:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 510, "total_steps": 2691, "loss": 0.2096, "learning_rate": 9.759475353022735e-06, "epoch": 0.568561872909699, "percentage": 18.95, "elapsed_time": "1:50:12", "remaining_time": "7:51:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 520, "total_steps": 2691, "loss": 0.2088, "learning_rate": 9.739193745639745e-06, "epoch": 0.5797101449275363, "percentage": 19.32, "elapsed_time": "1:52:28", "remaining_time": "7:49:35", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 530, "total_steps": 2691, "loss": 0.2079, "learning_rate": 9.718114127884774e-06, "epoch": 0.5908584169453734, "percentage": 19.7, "elapsed_time": "1:54:44", "remaining_time": "7:47:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 540, "total_steps": 2691, "loss": 0.2095, "learning_rate": 9.696240049254744e-06, "epoch": 0.6020066889632107, "percentage": 20.07, "elapsed_time": "1:56:42", "remaining_time": "7:44:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 550, "total_steps": 2691, "loss": 0.2103, "learning_rate": 9.673575193022073e-06, "epoch": 0.6131549609810479, "percentage": 20.44, "elapsed_time": "1:58:44", "remaining_time": "7:42:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 560, "total_steps": 2691, "loss": 0.209, "learning_rate": 9.650123375614477e-06, "epoch": 0.6243032329988851, "percentage": 20.81, "elapsed_time": "2:00:42", "remaining_time": "7:39:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 570, "total_steps": 2691, "loss": 0.2109, "learning_rate": 9.625888545972333e-06, "epoch": 0.6354515050167224, "percentage": 21.18, "elapsed_time": "2:02:45", "remaining_time": "7:36:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 580, "total_steps": 2691, "loss": 0.2072, "learning_rate": 9.600874784883736e-06, "epoch": 0.6465997770345596, "percentage": 21.55, "elapsed_time": "2:04:53", "remaining_time": "7:34:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 590, "total_steps": 2691, "loss": 0.2068, "learning_rate": 9.575086304297351e-06, "epoch": 0.6577480490523969, "percentage": 21.92, "elapsed_time": "2:07:00", "remaining_time": "7:32:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 600, "total_steps": 2691, "loss": 0.2088, "learning_rate": 9.548527446613198e-06, "epoch": 0.6688963210702341, "percentage": 22.3, "elapsed_time": "2:09:21", "remaining_time": "7:30:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 610, "total_steps": 2691, "loss": 0.2079, "learning_rate": 9.521202683951436e-06, "epoch": 0.6800445930880713, "percentage": 22.67, "elapsed_time": "2:11:56", "remaining_time": "7:30:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 620, "total_steps": 2691, "loss": 0.2086, "learning_rate": 9.493116617399333e-06, "epoch": 0.6911928651059086, "percentage": 23.04, "elapsed_time": "2:14:10", "remaining_time": "7:28:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 630, "total_steps": 2691, "loss": 0.2079, "learning_rate": 9.464273976236518e-06, "epoch": 0.7023411371237458, "percentage": 23.41, "elapsed_time": "2:16:16", "remaining_time": "7:25:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 640, "total_steps": 2691, "loss": 0.207, "learning_rate": 9.434679617138624e-06, "epoch": 0.7134894091415831, "percentage": 23.78, "elapsed_time": "2:18:20", "remaining_time": "7:23:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 650, "total_steps": 2691, "loss": 0.2072, "learning_rate": 9.404338523359511e-06, "epoch": 0.7246376811594203, "percentage": 24.15, "elapsed_time": "2:20:29", "remaining_time": "7:21:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 660, "total_steps": 2691, "loss": 0.2043, "learning_rate": 9.373255803892149e-06, "epoch": 0.7357859531772575, "percentage": 24.53, "elapsed_time": "2:22:45", "remaining_time": "7:19:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 670, "total_steps": 2691, "loss": 0.2059, "learning_rate": 9.341436692608341e-06, "epoch": 0.7469342251950948, "percentage": 24.9, "elapsed_time": "2:24:51", "remaining_time": "7:16:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 680, "total_steps": 2691, "loss": 0.2045, "learning_rate": 9.30888654737742e-06, "epoch": 0.758082497212932, "percentage": 25.27, "elapsed_time": "2:26:58", "remaining_time": "7:14:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 690, "total_steps": 2691, "loss": 0.2037, "learning_rate": 9.275610849164066e-06, "epoch": 0.7692307692307693, "percentage": 25.64, "elapsed_time": "2:29:07", "remaining_time": "7:12:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 700, "total_steps": 2691, "loss": 0.2055, "learning_rate": 9.241615201105372e-06, "epoch": 0.7803790412486065, "percentage": 26.01, "elapsed_time": "2:31:16", "remaining_time": "7:10:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 710, "total_steps": 2691, "loss": 0.2084, "learning_rate": 9.206905327567385e-06, "epoch": 0.7915273132664437, "percentage": 26.38, "elapsed_time": "2:33:50", "remaining_time": "7:09:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 720, "total_steps": 2691, "loss": 0.2054, "learning_rate": 9.171487073181198e-06, "epoch": 0.802675585284281, "percentage": 26.76, "elapsed_time": "2:35:56", "remaining_time": "7:06:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 730, "total_steps": 2691, "loss": 0.2073, "learning_rate": 9.135366401858782e-06, "epoch": 0.8138238573021181, "percentage": 27.13, "elapsed_time": "2:38:03", "remaining_time": "7:04:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 740, "total_steps": 2691, "loss": 0.2073, "learning_rate": 9.098549395788784e-06, "epoch": 0.8249721293199554, "percentage": 27.5, "elapsed_time": "2:40:09", "remaining_time": "7:02:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 750, "total_steps": 2691, "loss": 0.2039, "learning_rate": 9.061042254412346e-06, "epoch": 0.8361204013377926, "percentage": 27.87, "elapsed_time": "2:42:12", "remaining_time": "6:59:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 760, "total_steps": 2691, "loss": 0.2053, "learning_rate": 9.022851293379232e-06, "epoch": 0.8472686733556298, "percentage": 28.24, "elapsed_time": "2:44:12", "remaining_time": "6:57:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 770, "total_steps": 2691, "loss": 0.2023, "learning_rate": 8.983982943484348e-06, "epoch": 0.8584169453734671, "percentage": 28.61, "elapsed_time": "2:46:23", "remaining_time": "6:55:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 780, "total_steps": 2691, "loss": 0.2043, "learning_rate": 8.944443749584908e-06, "epoch": 0.8695652173913043, "percentage": 28.99, "elapsed_time": "2:48:30", "remaining_time": "6:52:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 790, "total_steps": 2691, "loss": 0.2013, "learning_rate": 8.904240369498363e-06, "epoch": 0.8807134894091416, "percentage": 29.36, "elapsed_time": "2:50:34", "remaining_time": "6:50:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 800, "total_steps": 2691, "loss": 0.2038, "learning_rate": 8.863379572881328e-06, "epoch": 0.8918617614269788, "percentage": 29.73, "elapsed_time": "2:52:39", "remaining_time": "6:48:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 810, "total_steps": 2691, "loss": 0.2044, "learning_rate": 8.821868240089676e-06, "epoch": 0.903010033444816, "percentage": 30.1, "elapsed_time": "2:55:20", "remaining_time": "6:47:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 820, "total_steps": 2691, "loss": 0.2068, "learning_rate": 8.779713361019983e-06, "epoch": 0.9141583054626533, "percentage": 30.47, "elapsed_time": "2:57:22", "remaining_time": "6:44:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 830, "total_steps": 2691, "loss": 0.2022, "learning_rate": 8.736922033932522e-06, "epoch": 0.9253065774804905, "percentage": 30.84, "elapsed_time": "2:59:26", "remaining_time": "6:42:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 840, "total_steps": 2691, "loss": 0.2027, "learning_rate": 8.693501464256046e-06, "epoch": 0.9364548494983278, "percentage": 31.22, "elapsed_time": "3:01:36", "remaining_time": "6:40:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 850, "total_steps": 2691, "loss": 0.2032, "learning_rate": 8.649458963374474e-06, "epoch": 0.947603121516165, "percentage": 31.59, "elapsed_time": "3:03:47", "remaining_time": "6:38:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 860, "total_steps": 2691, "loss": 0.2, "learning_rate": 8.604801947395776e-06, "epoch": 0.9587513935340022, "percentage": 31.96, "elapsed_time": "3:06:03", "remaining_time": "6:36:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 870, "total_steps": 2691, "loss": 0.2021, "learning_rate": 8.55953793590321e-06, "epoch": 0.9698996655518395, "percentage": 32.33, "elapsed_time": "3:08:14", "remaining_time": "6:34:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 880, "total_steps": 2691, "loss": 0.2002, "learning_rate": 8.513674550689128e-06, "epoch": 0.9810479375696767, "percentage": 32.7, "elapsed_time": "3:10:18", "remaining_time": "6:31:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 890, "total_steps": 2691, "loss": 0.2024, "learning_rate": 8.46721951447158e-06, "epoch": 0.992196209587514, "percentage": 33.07, "elapsed_time": "3:12:21", "remaining_time": "6:29:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 900, "total_steps": 2691, "loss": 0.1903, "learning_rate": 8.42018064959393e-06, "epoch": 1.0033444816053512, "percentage": 33.44, "elapsed_time": "3:14:33", "remaining_time": "6:27:10", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 910, "total_steps": 2691, "loss": 0.1599, "learning_rate": 8.37256587670768e-06, "epoch": 1.0144927536231885, "percentage": 33.82, "elapsed_time": "3:17:04", "remaining_time": "6:25:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 920, "total_steps": 2691, "loss": 0.1578, "learning_rate": 8.32438321343875e-06, "epoch": 1.0256410256410255, "percentage": 34.19, "elapsed_time": "3:19:14", "remaining_time": "6:23:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 930, "total_steps": 2691, "loss": 0.1584, "learning_rate": 8.275640773037432e-06, "epoch": 1.0367892976588629, "percentage": 34.56, "elapsed_time": "3:21:27", "remaining_time": "6:21:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 940, "total_steps": 2691, "loss": 0.1591, "learning_rate": 8.226346763012234e-06, "epoch": 1.0479375696767002, "percentage": 34.93, "elapsed_time": "3:23:37", "remaining_time": "6:19:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 950, "total_steps": 2691, "loss": 0.1587, "learning_rate": 8.176509483747863e-06, "epoch": 1.0590858416945372, "percentage": 35.3, "elapsed_time": "3:25:46", "remaining_time": "6:17:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 960, "total_steps": 2691, "loss": 0.1559, "learning_rate": 8.126137327107556e-06, "epoch": 1.0702341137123745, "percentage": 35.67, "elapsed_time": "3:27:55", "remaining_time": "6:14:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 970, "total_steps": 2691, "loss": 0.1582, "learning_rate": 8.075238775020011e-06, "epoch": 1.0813823857302118, "percentage": 36.05, "elapsed_time": "3:29:52", "remaining_time": "6:12:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 980, "total_steps": 2691, "loss": 0.1588, "learning_rate": 8.023822398051168e-06, "epoch": 1.0925306577480491, "percentage": 36.42, "elapsed_time": "3:31:59", "remaining_time": "6:10:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 990, "total_steps": 2691, "loss": 0.1576, "learning_rate": 7.971896853961043e-06, "epoch": 1.1036789297658862, "percentage": 36.79, "elapsed_time": "3:34:09", "remaining_time": "6:07:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1000, "total_steps": 2691, "loss": 0.1561, "learning_rate": 7.919470886245886e-06, "epoch": 1.1148272017837235, "percentage": 37.16, "elapsed_time": "3:36:22", "remaining_time": "6:05:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1000, "total_steps": 2691, "eval_loss": 0.20845723152160645, "epoch": 1.1148272017837235, "percentage": 37.16, "elapsed_time": "3:36:27", "remaining_time": "6:06:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1010, "total_steps": 2691, "loss": 0.1586, "learning_rate": 7.86655332266591e-06, "epoch": 1.1259754738015608, "percentage": 37.53, "elapsed_time": "3:38:57", "remaining_time": "6:04:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1020, "total_steps": 2691, "loss": 0.1582, "learning_rate": 7.813153073758833e-06, "epoch": 1.137123745819398, "percentage": 37.9, "elapsed_time": "3:41:09", "remaining_time": "6:02:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1030, "total_steps": 2691, "loss": 0.1579, "learning_rate": 7.759279131339455e-06, "epoch": 1.1482720178372352, "percentage": 38.28, "elapsed_time": "3:43:18", "remaining_time": "6:00:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1040, "total_steps": 2691, "loss": 0.163, "learning_rate": 7.704940566985592e-06, "epoch": 1.1594202898550725, "percentage": 38.65, "elapsed_time": "3:45:24", "remaining_time": "5:57:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1050, "total_steps": 2691, "loss": 0.1617, "learning_rate": 7.650146530510542e-06, "epoch": 1.1705685618729098, "percentage": 39.02, "elapsed_time": "3:47:29", "remaining_time": "5:55:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1060, "total_steps": 2691, "loss": 0.1596, "learning_rate": 7.594906248422392e-06, "epoch": 1.1817168338907469, "percentage": 39.39, "elapsed_time": "3:49:33", "remaining_time": "5:53:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1070, "total_steps": 2691, "loss": 0.1603, "learning_rate": 7.539229022370418e-06, "epoch": 1.1928651059085842, "percentage": 39.76, "elapsed_time": "3:51:37", "remaining_time": "5:50:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1080, "total_steps": 2691, "loss": 0.1579, "learning_rate": 7.483124227578811e-06, "epoch": 1.2040133779264215, "percentage": 40.13, "elapsed_time": "3:53:46", "remaining_time": "5:48:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1090, "total_steps": 2691, "loss": 0.16, "learning_rate": 7.426601311268043e-06, "epoch": 1.2151616499442586, "percentage": 40.51, "elapsed_time": "3:55:57", "remaining_time": "5:46:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1100, "total_steps": 2691, "loss": 0.1596, "learning_rate": 7.369669791064085e-06, "epoch": 1.2263099219620959, "percentage": 40.88, "elapsed_time": "3:58:00", "remaining_time": "5:44:15", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1110, "total_steps": 2691, "loss": 0.1582, "learning_rate": 7.312339253395778e-06, "epoch": 1.2374581939799332, "percentage": 41.25, "elapsed_time": "4:00:34", "remaining_time": "5:42:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1120, "total_steps": 2691, "loss": 0.1581, "learning_rate": 7.254619351880625e-06, "epoch": 1.2486064659977703, "percentage": 41.62, "elapsed_time": "4:02:43", "remaining_time": "5:40:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1130, "total_steps": 2691, "loss": 0.1592, "learning_rate": 7.1965198056992615e-06, "epoch": 1.2597547380156076, "percentage": 41.99, "elapsed_time": "4:04:51", "remaining_time": "5:38:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1140, "total_steps": 2691, "loss": 0.1595, "learning_rate": 7.138050397958889e-06, "epoch": 1.2709030100334449, "percentage": 42.36, "elapsed_time": "4:06:54", "remaining_time": "5:35:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1150, "total_steps": 2691, "loss": 0.1584, "learning_rate": 7.079220974045941e-06, "epoch": 1.282051282051282, "percentage": 42.74, "elapsed_time": "4:08:55", "remaining_time": "5:33:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1160, "total_steps": 2691, "loss": 0.1584, "learning_rate": 7.0200414399682745e-06, "epoch": 1.2931995540691192, "percentage": 43.11, "elapsed_time": "4:11:00", "remaining_time": "5:31:17", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1170, "total_steps": 2691, "loss": 0.1576, "learning_rate": 6.96052176068713e-06, "epoch": 1.3043478260869565, "percentage": 43.48, "elapsed_time": "4:13:10", "remaining_time": "5:29:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1180, "total_steps": 2691, "loss": 0.1599, "learning_rate": 6.900671958439192e-06, "epoch": 1.3154960981047936, "percentage": 43.85, "elapsed_time": "4:15:19", "remaining_time": "5:26:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1190, "total_steps": 2691, "loss": 0.158, "learning_rate": 6.840502111048982e-06, "epoch": 1.326644370122631, "percentage": 44.22, "elapsed_time": "4:17:22", "remaining_time": "5:24:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1200, "total_steps": 2691, "loss": 0.1604, "learning_rate": 6.780022350231912e-06, "epoch": 1.3377926421404682, "percentage": 44.59, "elapsed_time": "4:19:30", "remaining_time": "5:22:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1210, "total_steps": 2691, "loss": 0.1576, "learning_rate": 6.719242859888243e-06, "epoch": 1.3489409141583055, "percentage": 44.96, "elapsed_time": "4:21:58", "remaining_time": "5:20:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1220, "total_steps": 2691, "loss": 0.1588, "learning_rate": 6.65817387438827e-06, "epoch": 1.3600891861761428, "percentage": 45.34, "elapsed_time": "4:24:15", "remaining_time": "5:18:37", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1230, "total_steps": 2691, "loss": 0.1581, "learning_rate": 6.596825676849006e-06, "epoch": 1.37123745819398, "percentage": 45.71, "elapsed_time": "4:26:13", "remaining_time": "5:16:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1240, "total_steps": 2691, "loss": 0.1588, "learning_rate": 6.535208597402658e-06, "epoch": 1.3823857302118172, "percentage": 46.08, "elapsed_time": "4:28:18", "remaining_time": "5:13:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1250, "total_steps": 2691, "loss": 0.1586, "learning_rate": 6.473333011457181e-06, "epoch": 1.3935340022296545, "percentage": 46.45, "elapsed_time": "4:30:23", "remaining_time": "5:11:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1260, "total_steps": 2691, "loss": 0.1594, "learning_rate": 6.411209337949214e-06, "epoch": 1.4046822742474916, "percentage": 46.82, "elapsed_time": "4:32:34", "remaining_time": "5:09:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1270, "total_steps": 2691, "loss": 0.159, "learning_rate": 6.348848037589683e-06, "epoch": 1.415830546265329, "percentage": 47.19, "elapsed_time": "4:34:35", "remaining_time": "5:07:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1280, "total_steps": 2691, "loss": 0.1589, "learning_rate": 6.286259611102376e-06, "epoch": 1.4269788182831662, "percentage": 47.57, "elapsed_time": "4:36:44", "remaining_time": "5:05:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1290, "total_steps": 2691, "loss": 0.1587, "learning_rate": 6.223454597455776e-06, "epoch": 1.4381270903010033, "percentage": 47.94, "elapsed_time": "4:38:58", "remaining_time": "5:02:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1300, "total_steps": 2691, "loss": 0.1595, "learning_rate": 6.160443572088443e-06, "epoch": 1.4492753623188406, "percentage": 48.31, "elapsed_time": "4:41:03", "remaining_time": "5:00:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1310, "total_steps": 2691, "loss": 0.1584, "learning_rate": 6.097237145128282e-06, "epoch": 1.4604236343366779, "percentage": 48.68, "elapsed_time": "4:43:33", "remaining_time": "4:58:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1320, "total_steps": 2691, "loss": 0.1554, "learning_rate": 6.0338459596059395e-06, "epoch": 1.471571906354515, "percentage": 49.05, "elapsed_time": "4:45:47", "remaining_time": "4:56:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1330, "total_steps": 2691, "loss": 0.1585, "learning_rate": 5.9702806896626855e-06, "epoch": 1.4827201783723523, "percentage": 49.42, "elapsed_time": "4:47:50", "remaining_time": "4:54:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1340, "total_steps": 2691, "loss": 0.159, "learning_rate": 5.90655203875304e-06, "epoch": 1.4938684503901896, "percentage": 49.8, "elapsed_time": "4:49:55", "remaining_time": "4:52:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1350, "total_steps": 2691, "loss": 0.1579, "learning_rate": 5.8426707378424675e-06, "epoch": 1.5050167224080266, "percentage": 50.17, "elapsed_time": "4:51:59", "remaining_time": "4:50:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1360, "total_steps": 2691, "loss": 0.1564, "learning_rate": 5.778647543600443e-06, "epoch": 1.516164994425864, "percentage": 50.54, "elapsed_time": "4:54:10", "remaining_time": "4:47:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1370, "total_steps": 2691, "loss": 0.1545, "learning_rate": 5.714493236589187e-06, "epoch": 1.5273132664437012, "percentage": 50.91, "elapsed_time": "4:56:15", "remaining_time": "4:45:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1380, "total_steps": 2691, "loss": 0.1552, "learning_rate": 5.650218619448379e-06, "epoch": 1.5384615384615383, "percentage": 51.28, "elapsed_time": "4:58:27", "remaining_time": "4:43:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1390, "total_steps": 2691, "loss": 0.1556, "learning_rate": 5.5858345150761515e-06, "epoch": 1.5496098104793758, "percentage": 51.65, "elapsed_time": "5:00:43", "remaining_time": "4:41:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1400, "total_steps": 2691, "loss": 0.1568, "learning_rate": 5.521351764806672e-06, "epoch": 1.560758082497213, "percentage": 52.03, "elapsed_time": "5:02:57", "remaining_time": "4:39:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1410, "total_steps": 2691, "loss": 0.158, "learning_rate": 5.456781226584621e-06, "epoch": 1.57190635451505, "percentage": 52.4, "elapsed_time": "5:05:35", "remaining_time": "4:37:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1420, "total_steps": 2691, "loss": 0.1602, "learning_rate": 5.392133773136877e-06, "epoch": 1.5830546265328875, "percentage": 52.77, "elapsed_time": "5:07:47", "remaining_time": "4:35:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1430, "total_steps": 2691, "loss": 0.1571, "learning_rate": 5.327420290141712e-06, "epoch": 1.5942028985507246, "percentage": 53.14, "elapsed_time": "5:09:52", "remaining_time": "4:33:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1440, "total_steps": 2691, "loss": 0.1567, "learning_rate": 5.262651674395799e-06, "epoch": 1.605351170568562, "percentage": 53.51, "elapsed_time": "5:11:56", "remaining_time": "4:30:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1450, "total_steps": 2691, "loss": 0.1574, "learning_rate": 5.197838831979352e-06, "epoch": 1.6164994425863992, "percentage": 53.88, "elapsed_time": "5:14:01", "remaining_time": "4:28:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1460, "total_steps": 2691, "loss": 0.1569, "learning_rate": 5.132992676419704e-06, "epoch": 1.6276477146042363, "percentage": 54.25, "elapsed_time": "5:16:06", "remaining_time": "4:26:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1470, "total_steps": 2691, "loss": 0.157, "learning_rate": 5.068124126853633e-06, "epoch": 1.6387959866220736, "percentage": 54.63, "elapsed_time": "5:18:19", "remaining_time": "4:24:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1480, "total_steps": 2691, "loss": 0.1569, "learning_rate": 5.0032441061887345e-06, "epoch": 1.649944258639911, "percentage": 55.0, "elapsed_time": "5:20:19", "remaining_time": "4:22:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1490, "total_steps": 2691, "loss": 0.1571, "learning_rate": 4.938363539264175e-06, "epoch": 1.661092530657748, "percentage": 55.37, "elapsed_time": "5:22:25", "remaining_time": "4:19:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1500, "total_steps": 2691, "loss": 0.1547, "learning_rate": 4.87349335101109e-06, "epoch": 1.6722408026755853, "percentage": 55.74, "elapsed_time": "5:24:34", "remaining_time": "4:17:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1500, "total_steps": 2691, "eval_loss": 0.19943761825561523, "epoch": 1.6722408026755853, "percentage": 55.74, "elapsed_time": "5:24:40", "remaining_time": "4:17:47", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1510, "total_steps": 2691, "loss": 0.1562, "learning_rate": 4.808644464613015e-06, "epoch": 1.6833890746934226, "percentage": 56.11, "elapsed_time": "5:27:13", "remaining_time": "4:15:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1520, "total_steps": 2691, "loss": 0.1547, "learning_rate": 4.743827799666555e-06, "epoch": 1.6945373467112597, "percentage": 56.48, "elapsed_time": "5:29:16", "remaining_time": "4:13:39", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1530, "total_steps": 2691, "loss": 0.1563, "learning_rate": 4.679054270342703e-06, "epoch": 1.705685618729097, "percentage": 56.86, "elapsed_time": "5:31:25", "remaining_time": "4:11:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1540, "total_steps": 2691, "loss": 0.1538, "learning_rate": 4.614334783549049e-06, "epoch": 1.7168338907469343, "percentage": 57.23, "elapsed_time": "5:33:38", "remaining_time": "4:09:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1550, "total_steps": 2691, "loss": 0.1571, "learning_rate": 4.549680237093215e-06, "epoch": 1.7279821627647713, "percentage": 57.6, "elapsed_time": "5:35:42", "remaining_time": "4:07:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1560, "total_steps": 2691, "loss": 0.1561, "learning_rate": 4.485101517847831e-06, "epoch": 1.7391304347826086, "percentage": 57.97, "elapsed_time": "5:37:43", "remaining_time": "4:04:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1570, "total_steps": 2691, "loss": 0.1555, "learning_rate": 4.420609499917337e-06, "epoch": 1.750278706800446, "percentage": 58.34, "elapsed_time": "5:39:46", "remaining_time": "4:02:36", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1580, "total_steps": 2691, "loss": 0.1545, "learning_rate": 4.3562150428069565e-06, "epoch": 1.761426978818283, "percentage": 58.71, "elapsed_time": "5:41:45", "remaining_time": "4:00:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1590, "total_steps": 2691, "loss": 0.1547, "learning_rate": 4.291928989594102e-06, "epoch": 1.7725752508361206, "percentage": 59.09, "elapsed_time": "5:43:51", "remaining_time": "3:58:06", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1600, "total_steps": 2691, "loss": 0.1548, "learning_rate": 4.2277621651025734e-06, "epoch": 1.7837235228539576, "percentage": 59.46, "elapsed_time": "5:45:59", "remaining_time": "3:55:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1610, "total_steps": 2691, "loss": 0.1559, "learning_rate": 4.163725374079808e-06, "epoch": 1.7948717948717947, "percentage": 59.83, "elapsed_time": "5:48:29", "remaining_time": "3:53:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1620, "total_steps": 2691, "loss": 0.1527, "learning_rate": 4.099829399377524e-06, "epoch": 1.8060200668896322, "percentage": 60.2, "elapsed_time": "5:50:32", "remaining_time": "3:51:45", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1630, "total_steps": 2691, "loss": 0.1562, "learning_rate": 4.036085000136046e-06, "epoch": 1.8171683389074693, "percentage": 60.57, "elapsed_time": "5:52:31", "remaining_time": "3:49:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1640, "total_steps": 2691, "loss": 0.1546, "learning_rate": 3.972502909972628e-06, "epoch": 1.8283166109253066, "percentage": 60.94, "elapsed_time": "5:54:44", "remaining_time": "3:47:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1650, "total_steps": 2691, "loss": 0.1547, "learning_rate": 3.909093835174066e-06, "epoch": 1.839464882943144, "percentage": 61.32, "elapsed_time": "5:56:56", "remaining_time": "3:45:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1660, "total_steps": 2691, "loss": 0.1546, "learning_rate": 3.845868452893919e-06, "epoch": 1.850613154960981, "percentage": 61.69, "elapsed_time": "5:59:09", "remaining_time": "3:43:03", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1670, "total_steps": 2691, "loss": 0.1526, "learning_rate": 3.7828374093546303e-06, "epoch": 1.8617614269788183, "percentage": 62.06, "elapsed_time": "6:01:20", "remaining_time": "3:40:55", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1680, "total_steps": 2691, "loss": 0.1542, "learning_rate": 3.720011318054871e-06, "epoch": 1.8729096989966556, "percentage": 62.43, "elapsed_time": "6:03:29", "remaining_time": "3:38:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1690, "total_steps": 2691, "loss": 0.1543, "learning_rate": 3.657400757982367e-06, "epoch": 1.8840579710144927, "percentage": 62.8, "elapsed_time": "6:05:32", "remaining_time": "3:36:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1700, "total_steps": 2691, "loss": 0.1531, "learning_rate": 3.595016271832572e-06, "epoch": 1.89520624303233, "percentage": 63.17, "elapsed_time": "6:07:37", "remaining_time": "3:34:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1710, "total_steps": 2691, "loss": 0.1541, "learning_rate": 3.532868364233416e-06, "epoch": 1.9063545150501673, "percentage": 63.55, "elapsed_time": "6:10:03", "remaining_time": "3:32:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1720, "total_steps": 2691, "loss": 0.1552, "learning_rate": 3.470967499976492e-06, "epoch": 1.9175027870680044, "percentage": 63.92, "elapsed_time": "6:12:15", "remaining_time": "3:30:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1730, "total_steps": 2691, "loss": 0.1521, "learning_rate": 3.4093241022549316e-06, "epoch": 1.9286510590858417, "percentage": 64.29, "elapsed_time": "6:14:29", "remaining_time": "3:28:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1740, "total_steps": 2691, "loss": 0.1536, "learning_rate": 3.347948550908303e-06, "epoch": 1.939799331103679, "percentage": 64.66, "elapsed_time": "6:16:35", "remaining_time": "3:25:49", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1750, "total_steps": 2691, "loss": 0.1524, "learning_rate": 3.286851180674788e-06, "epoch": 1.950947603121516, "percentage": 65.03, "elapsed_time": "6:18:42", "remaining_time": "3:23:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1760, "total_steps": 2691, "loss": 0.1527, "learning_rate": 3.2260422794509704e-06, "epoch": 1.9620958751393534, "percentage": 65.4, "elapsed_time": "6:20:56", "remaining_time": "3:21:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1770, "total_steps": 2691, "loss": 0.1523, "learning_rate": 3.165532086559504e-06, "epoch": 1.9732441471571907, "percentage": 65.77, "elapsed_time": "6:23:16", "remaining_time": "3:19:26", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1780, "total_steps": 2691, "loss": 0.1528, "learning_rate": 3.105330791024965e-06, "epoch": 1.9843924191750277, "percentage": 66.15, "elapsed_time": "6:25:19", "remaining_time": "3:17:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1790, "total_steps": 2691, "loss": 0.1523, "learning_rate": 3.045448529858165e-06, "epoch": 1.9955406911928653, "percentage": 66.52, "elapsed_time": "6:27:25", "remaining_time": "3:15:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1800, "total_steps": 2691, "loss": 0.1257, "learning_rate": 2.9858953863492334e-06, "epoch": 2.0066889632107023, "percentage": 66.89, "elapsed_time": "6:29:38", "remaining_time": "3:12:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1810, "total_steps": 2691, "loss": 0.111, "learning_rate": 2.9266813883697342e-06, "epoch": 2.0178372352285394, "percentage": 67.26, "elapsed_time": "6:32:13", "remaining_time": "3:10:54", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1820, "total_steps": 2691, "loss": 0.1125, "learning_rate": 2.867816506684126e-06, "epoch": 2.028985507246377, "percentage": 67.63, "elapsed_time": "6:34:21", "remaining_time": "3:08:43", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1830, "total_steps": 2691, "loss": 0.1113, "learning_rate": 2.809310653270825e-06, "epoch": 2.040133779264214, "percentage": 68.0, "elapsed_time": "6:36:24", "remaining_time": "3:06:30", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1840, "total_steps": 2691, "loss": 0.1101, "learning_rate": 2.751173679653184e-06, "epoch": 2.051282051282051, "percentage": 68.38, "elapsed_time": "6:38:31", "remaining_time": "3:04:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1850, "total_steps": 2691, "loss": 0.1099, "learning_rate": 2.693415375240629e-06, "epoch": 2.0624303232998886, "percentage": 68.75, "elapsed_time": "6:40:38", "remaining_time": "3:02:07", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1860, "total_steps": 2691, "loss": 0.1098, "learning_rate": 2.636045465680282e-06, "epoch": 2.0735785953177257, "percentage": 69.12, "elapsed_time": "6:42:38", "remaining_time": "2:59:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1870, "total_steps": 2691, "loss": 0.1094, "learning_rate": 2.5790736112192893e-06, "epoch": 2.084726867335563, "percentage": 69.49, "elapsed_time": "6:44:50", "remaining_time": "2:57:44", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1880, "total_steps": 2691, "loss": 0.1105, "learning_rate": 2.522509405078187e-06, "epoch": 2.0958751393534003, "percentage": 69.86, "elapsed_time": "6:46:58", "remaining_time": "2:55:33", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1890, "total_steps": 2691, "loss": 0.1102, "learning_rate": 2.4663623718355444e-06, "epoch": 2.1070234113712374, "percentage": 70.23, "elapsed_time": "6:49:07", "remaining_time": "2:53:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1900, "total_steps": 2691, "loss": 0.1089, "learning_rate": 2.410641965824166e-06, "epoch": 2.1181716833890745, "percentage": 70.61, "elapsed_time": "6:51:15", "remaining_time": "2:51:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1910, "total_steps": 2691, "loss": 0.1119, "learning_rate": 2.355357569539114e-06, "epoch": 2.129319955406912, "percentage": 70.98, "elapsed_time": "6:53:49", "remaining_time": "2:49:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1920, "total_steps": 2691, "loss": 0.114, "learning_rate": 2.300518492057842e-06, "epoch": 2.140468227424749, "percentage": 71.35, "elapsed_time": "6:55:59", "remaining_time": "2:47:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1930, "total_steps": 2691, "loss": 0.1102, "learning_rate": 2.2461339674726804e-06, "epoch": 2.1516164994425866, "percentage": 71.72, "elapsed_time": "6:58:03", "remaining_time": "2:44:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1940, "total_steps": 2691, "loss": 0.1109, "learning_rate": 2.192213153335953e-06, "epoch": 2.1627647714604237, "percentage": 72.09, "elapsed_time": "7:00:09", "remaining_time": "2:42:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1950, "total_steps": 2691, "loss": 0.1106, "learning_rate": 2.1387651291179775e-06, "epoch": 2.1739130434782608, "percentage": 72.46, "elapsed_time": "7:02:11", "remaining_time": "2:40:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1960, "total_steps": 2691, "loss": 0.1089, "learning_rate": 2.085798894678217e-06, "epoch": 2.1850613154960983, "percentage": 72.84, "elapsed_time": "7:04:23", "remaining_time": "2:38:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1970, "total_steps": 2691, "loss": 0.1094, "learning_rate": 2.0333233687498433e-06, "epoch": 2.1962095875139354, "percentage": 73.21, "elapsed_time": "7:06:30", "remaining_time": "2:36:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1980, "total_steps": 2691, "loss": 0.1105, "learning_rate": 1.98134738743794e-06, "epoch": 2.2073578595317724, "percentage": 73.58, "elapsed_time": "7:08:33", "remaining_time": "2:33:53", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 1990, "total_steps": 2691, "loss": 0.1088, "learning_rate": 1.9298797027316474e-06, "epoch": 2.21850613154961, "percentage": 73.95, "elapsed_time": "7:10:29", "remaining_time": "2:31:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2000, "total_steps": 2691, "loss": 0.1092, "learning_rate": 1.878928981030445e-06, "epoch": 2.229654403567447, "percentage": 74.32, "elapsed_time": "7:12:28", "remaining_time": "2:29:25", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2000, "total_steps": 2691, "eval_loss": 0.20725835859775543, "epoch": 2.229654403567447, "percentage": 74.32, "elapsed_time": "7:12:33", "remaining_time": "2:29:27", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2010, "total_steps": 2691, "loss": 0.1098, "learning_rate": 1.828503801684864e-06, "epoch": 2.240802675585284, "percentage": 74.69, "elapsed_time": "7:15:00", "remaining_time": "2:27:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2020, "total_steps": 2691, "loss": 0.1093, "learning_rate": 1.7786126555518484e-06, "epoch": 2.2519509476031216, "percentage": 75.07, "elapsed_time": "7:17:11", "remaining_time": "2:25:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2030, "total_steps": 2691, "loss": 0.1088, "learning_rate": 1.729263943565022e-06, "epoch": 2.2630992196209587, "percentage": 75.44, "elapsed_time": "7:19:24", "remaining_time": "2:23:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2040, "total_steps": 2691, "loss": 0.1101, "learning_rate": 1.6804659753200925e-06, "epoch": 2.274247491638796, "percentage": 75.81, "elapsed_time": "7:21:24", "remaining_time": "2:20:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2050, "total_steps": 2691, "loss": 0.1096, "learning_rate": 1.6322269676756402e-06, "epoch": 2.2853957636566333, "percentage": 76.18, "elapsed_time": "7:23:28", "remaining_time": "2:18:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2060, "total_steps": 2691, "loss": 0.109, "learning_rate": 1.5845550433695172e-06, "epoch": 2.2965440356744704, "percentage": 76.55, "elapsed_time": "7:25:36", "remaining_time": "2:16:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2070, "total_steps": 2691, "loss": 0.1097, "learning_rate": 1.5374582296511054e-06, "epoch": 2.3076923076923075, "percentage": 76.92, "elapsed_time": "7:27:40", "remaining_time": "2:14:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2080, "total_steps": 2691, "loss": 0.1084, "learning_rate": 1.4909444569296334e-06, "epoch": 2.318840579710145, "percentage": 77.29, "elapsed_time": "7:29:37", "remaining_time": "2:12:04", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2090, "total_steps": 2691, "loss": 0.1094, "learning_rate": 1.4450215574388265e-06, "epoch": 2.329988851727982, "percentage": 77.67, "elapsed_time": "7:31:37", "remaining_time": "2:09:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2100, "total_steps": 2691, "loss": 0.1086, "learning_rate": 1.3996972639180645e-06, "epoch": 2.3411371237458196, "percentage": 78.04, "elapsed_time": "7:33:43", "remaining_time": "2:07:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2110, "total_steps": 2691, "loss": 0.1079, "learning_rate": 1.3549792083103037e-06, "epoch": 2.3522853957636567, "percentage": 78.41, "elapsed_time": "7:36:27", "remaining_time": "2:05:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2120, "total_steps": 2691, "loss": 0.1093, "learning_rate": 1.3108749204769732e-06, "epoch": 2.3634336677814938, "percentage": 78.78, "elapsed_time": "7:38:36", "remaining_time": "2:03:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2130, "total_steps": 2691, "loss": 0.109, "learning_rate": 1.2673918269300557e-06, "epoch": 2.374581939799331, "percentage": 79.15, "elapsed_time": "7:40:36", "remaining_time": "2:01:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2140, "total_steps": 2691, "loss": 0.1086, "learning_rate": 1.2245372495815726e-06, "epoch": 2.3857302118171684, "percentage": 79.52, "elapsed_time": "7:42:41", "remaining_time": "1:59:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2150, "total_steps": 2691, "loss": 0.1088, "learning_rate": 1.1823184045106816e-06, "epoch": 2.3968784838350055, "percentage": 79.9, "elapsed_time": "7:44:55", "remaining_time": "1:56:59", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2160, "total_steps": 2691, "loss": 0.1085, "learning_rate": 1.140742400748593e-06, "epoch": 2.408026755852843, "percentage": 80.27, "elapsed_time": "7:47:02", "remaining_time": "1:54:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2170, "total_steps": 2691, "loss": 0.1084, "learning_rate": 1.099816239081521e-06, "epoch": 2.41917502787068, "percentage": 80.64, "elapsed_time": "7:49:08", "remaining_time": "1:52:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2180, "total_steps": 2691, "loss": 0.1089, "learning_rate": 1.0595468108718448e-06, "epoch": 2.430323299888517, "percentage": 81.01, "elapsed_time": "7:51:17", "remaining_time": "1:50:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2190, "total_steps": 2691, "loss": 0.1087, "learning_rate": 1.0199408968977138e-06, "epoch": 2.4414715719063547, "percentage": 81.38, "elapsed_time": "7:53:19", "remaining_time": "1:48:16", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2200, "total_steps": 2691, "loss": 0.108, "learning_rate": 9.810051662112558e-07, "epoch": 2.4526198439241917, "percentage": 81.75, "elapsed_time": "7:55:21", "remaining_time": "1:46:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2210, "total_steps": 2691, "loss": 0.1096, "learning_rate": 9.427461750156142e-07, "epoch": 2.463768115942029, "percentage": 82.13, "elapsed_time": "7:57:54", "remaining_time": "1:44:00", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2220, "total_steps": 2691, "loss": 0.1089, "learning_rate": 9.051703655609762e-07, "epoch": 2.4749163879598663, "percentage": 82.5, "elapsed_time": "8:00:02", "remaining_time": "1:41:50", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2230, "total_steps": 2691, "loss": 0.1095, "learning_rate": 8.68284065059794e-07, "epoch": 2.4860646599777034, "percentage": 82.87, "elapsed_time": "8:02:13", "remaining_time": "1:39:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2240, "total_steps": 2691, "loss": 0.1082, "learning_rate": 8.320934846213746e-07, "epoch": 2.4972129319955405, "percentage": 83.24, "elapsed_time": "8:04:22", "remaining_time": "1:37:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2250, "total_steps": 2691, "loss": 0.1083, "learning_rate": 7.966047182060226e-07, "epoch": 2.508361204013378, "percentage": 83.61, "elapsed_time": "8:06:27", "remaining_time": "1:35:20", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2260, "total_steps": 2691, "loss": 0.1083, "learning_rate": 7.618237415989032e-07, "epoch": 2.519509476031215, "percentage": 83.98, "elapsed_time": "8:08:30", "remaining_time": "1:33:09", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2270, "total_steps": 2691, "loss": 0.1082, "learning_rate": 7.277564114038149e-07, "epoch": 2.5306577480490526, "percentage": 84.36, "elapsed_time": "8:10:26", "remaining_time": "1:30:57", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2280, "total_steps": 2691, "loss": 0.1072, "learning_rate": 6.944084640570142e-07, "epoch": 2.5418060200668897, "percentage": 84.73, "elapsed_time": "8:12:40", "remaining_time": "1:28:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2290, "total_steps": 2691, "loss": 0.1071, "learning_rate": 6.617855148612945e-07, "epoch": 2.552954292084727, "percentage": 85.1, "elapsed_time": "8:14:58", "remaining_time": "1:26:40", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2300, "total_steps": 2691, "loss": 0.1087, "learning_rate": 6.298930570404432e-07, "epoch": 2.564102564102564, "percentage": 85.47, "elapsed_time": "8:17:00", "remaining_time": "1:24:29", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2310, "total_steps": 2691, "loss": 0.1076, "learning_rate": 5.987364608142693e-07, "epoch": 2.5752508361204014, "percentage": 85.84, "elapsed_time": "8:19:34", "remaining_time": "1:22:23", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2320, "total_steps": 2691, "loss": 0.109, "learning_rate": 5.683209724943344e-07, "epoch": 2.5863991081382385, "percentage": 86.21, "elapsed_time": "8:21:39", "remaining_time": "1:20:13", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2330, "total_steps": 2691, "loss": 0.1081, "learning_rate": 5.386517136005543e-07, "epoch": 2.597547380156076, "percentage": 86.58, "elapsed_time": "8:23:38", "remaining_time": "1:18:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2340, "total_steps": 2691, "loss": 0.1079, "learning_rate": 5.097336799988067e-07, "epoch": 2.608695652173913, "percentage": 86.96, "elapsed_time": "8:25:47", "remaining_time": "1:15:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2350, "total_steps": 2691, "loss": 0.1081, "learning_rate": 4.815717410597042e-07, "epoch": 2.61984392419175, "percentage": 87.33, "elapsed_time": "8:27:59", "remaining_time": "1:13:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2360, "total_steps": 2691, "loss": 0.1086, "learning_rate": 4.541706388386624e-07, "epoch": 2.6309921962095872, "percentage": 87.7, "elapsed_time": "8:30:19", "remaining_time": "1:11:34", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2370, "total_steps": 2691, "loss": 0.1073, "learning_rate": 4.275349872774098e-07, "epoch": 2.6421404682274248, "percentage": 88.07, "elapsed_time": "8:32:30", "remaining_time": "1:09:24", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2380, "total_steps": 2691, "loss": 0.1076, "learning_rate": 4.0166927142706313e-07, "epoch": 2.653288740245262, "percentage": 88.44, "elapsed_time": "8:34:38", "remaining_time": "1:07:14", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2390, "total_steps": 2691, "loss": 0.1066, "learning_rate": 3.765778466929182e-07, "epoch": 2.6644370122630994, "percentage": 88.81, "elapsed_time": "8:36:50", "remaining_time": "1:05:05", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2400, "total_steps": 2691, "loss": 0.1081, "learning_rate": 3.5226493810105624e-07, "epoch": 2.6755852842809364, "percentage": 89.19, "elapsed_time": "8:39:04", "remaining_time": "1:02:56", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2410, "total_steps": 2691, "loss": 0.108, "learning_rate": 3.2873463958691675e-07, "epoch": 2.6867335562987735, "percentage": 89.56, "elapsed_time": "8:41:33", "remaining_time": "1:00:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2420, "total_steps": 2691, "loss": 0.1077, "learning_rate": 3.05990913305938e-07, "epoch": 2.697881828316611, "percentage": 89.93, "elapsed_time": "8:43:41", "remaining_time": "0:58:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2430, "total_steps": 2691, "loss": 0.106, "learning_rate": 2.840375889663871e-07, "epoch": 2.709030100334448, "percentage": 90.3, "elapsed_time": "8:45:46", "remaining_time": "0:56:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2440, "total_steps": 2691, "loss": 0.107, "learning_rate": 2.62878363184495e-07, "epoch": 2.7201783723522857, "percentage": 90.67, "elapsed_time": "8:47:55", "remaining_time": "0:54:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2450, "total_steps": 2691, "loss": 0.1078, "learning_rate": 2.425167988620014e-07, "epoch": 2.7313266443701227, "percentage": 91.04, "elapsed_time": "8:50:04", "remaining_time": "0:52:08", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2460, "total_steps": 2691, "loss": 0.1076, "learning_rate": 2.2295632458621097e-07, "epoch": 2.74247491638796, "percentage": 91.42, "elapsed_time": "8:52:08", "remaining_time": "0:49:58", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2470, "total_steps": 2691, "loss": 0.1076, "learning_rate": 2.0420023405267663e-07, "epoch": 2.753623188405797, "percentage": 91.79, "elapsed_time": "8:54:15", "remaining_time": "0:47:48", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2480, "total_steps": 2691, "loss": 0.1071, "learning_rate": 1.8625168551058115e-07, "epoch": 2.7647714604236344, "percentage": 92.16, "elapsed_time": "8:56:28", "remaining_time": "0:45:38", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2490, "total_steps": 2691, "loss": 0.1081, "learning_rate": 1.6911370123094238e-07, "epoch": 2.7759197324414715, "percentage": 92.53, "elapsed_time": "8:58:31", "remaining_time": "0:43:28", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2500, "total_steps": 2691, "loss": 0.1073, "learning_rate": 1.5278916699770163e-07, "epoch": 2.787068004459309, "percentage": 92.9, "elapsed_time": "9:00:45", "remaining_time": "0:41:18", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2500, "total_steps": 2691, "eval_loss": 0.20636913180351257, "epoch": 2.787068004459309, "percentage": 92.9, "elapsed_time": "9:00:51", "remaining_time": "0:41:19", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2510, "total_steps": 2691, "loss": 0.1075, "learning_rate": 1.3728083162180384e-07, "epoch": 2.798216276477146, "percentage": 93.27, "elapsed_time": "9:03:27", "remaining_time": "0:39:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2520, "total_steps": 2691, "loss": 0.1065, "learning_rate": 1.2259130647833627e-07, "epoch": 2.809364548494983, "percentage": 93.65, "elapsed_time": "9:05:33", "remaining_time": "0:37:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2530, "total_steps": 2691, "loss": 0.1079, "learning_rate": 1.0872306506681251e-07, "epoch": 2.8205128205128203, "percentage": 94.02, "elapsed_time": "9:07:42", "remaining_time": "0:34:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2540, "total_steps": 2691, "loss": 0.1083, "learning_rate": 9.567844259467051e-08, "epoch": 2.831661092530658, "percentage": 94.39, "elapsed_time": "9:09:48", "remaining_time": "0:32:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2550, "total_steps": 2691, "loss": 0.1079, "learning_rate": 8.345963558406001e-08, "epoch": 2.842809364548495, "percentage": 94.76, "elapsed_time": "9:12:00", "remaining_time": "0:30:31", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2560, "total_steps": 2691, "loss": 0.107, "learning_rate": 7.206870150197831e-08, "epoch": 2.8539576365663324, "percentage": 95.13, "elapsed_time": "9:14:02", "remaining_time": "0:28:21", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2570, "total_steps": 2691, "loss": 0.1067, "learning_rate": 6.150755841382527e-08, "epoch": 2.8651059085841695, "percentage": 95.5, "elapsed_time": "9:16:09", "remaining_time": "0:26:11", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2580, "total_steps": 2691, "loss": 0.1078, "learning_rate": 5.177798466042716e-08, "epoch": 2.8762541806020065, "percentage": 95.88, "elapsed_time": "9:18:13", "remaining_time": "0:24:01", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2590, "total_steps": 2691, "loss": 0.109, "learning_rate": 4.2881618558592855e-08, "epoch": 2.887402452619844, "percentage": 96.25, "elapsed_time": "9:20:23", "remaining_time": "0:21:51", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2600, "total_steps": 2691, "loss": 0.1064, "learning_rate": 3.481995812524286e-08, "epoch": 2.898550724637681, "percentage": 96.62, "elapsed_time": "9:22:27", "remaining_time": "0:19:41", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2610, "total_steps": 2691, "loss": 0.107, "learning_rate": 2.7594360825166644e-08, "epoch": 2.9096989966555182, "percentage": 96.99, "elapsed_time": "9:24:59", "remaining_time": "0:17:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2620, "total_steps": 2691, "loss": 0.1067, "learning_rate": 2.120604334244547e-08, "epoch": 2.9208472686733558, "percentage": 97.36, "elapsed_time": "9:27:09", "remaining_time": "0:15:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2630, "total_steps": 2691, "loss": 0.1063, "learning_rate": 1.565608137558128e-08, "epoch": 2.931995540691193, "percentage": 97.73, "elapsed_time": "9:29:16", "remaining_time": "0:13:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2640, "total_steps": 2691, "loss": 0.1075, "learning_rate": 1.0945409456364353e-08, "epoch": 2.94314381270903, "percentage": 98.1, "elapsed_time": "9:31:17", "remaining_time": "0:11:02", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2650, "total_steps": 2691, "loss": 0.1075, "learning_rate": 7.0748207925103176e-09, "epoch": 2.9542920847268674, "percentage": 98.48, "elapsed_time": "9:33:26", "remaining_time": "0:08:52", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2660, "total_steps": 2691, "loss": 0.1079, "learning_rate": 4.044967134099187e-09, "epoch": 2.9654403567447045, "percentage": 98.85, "elapsed_time": "9:35:38", "remaining_time": "0:06:42", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2670, "total_steps": 2691, "loss": 0.1064, "learning_rate": 1.8563586638281617e-09, "epoch": 2.976588628762542, "percentage": 99.22, "elapsed_time": "9:37:46", "remaining_time": "0:04:32", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2680, "total_steps": 2691, "loss": 0.1071, "learning_rate": 5.093639111025673e-10, "epoch": 2.987736900780379, "percentage": 99.59, "elapsed_time": "9:39:55", "remaining_time": "0:02:22", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2690, "total_steps": 2691, "loss": 0.1079, "learning_rate": 4.2096899854904285e-12, "epoch": 2.998885172798216, "percentage": 99.96, "elapsed_time": "9:42:00", "remaining_time": "0:00:12", "throughput": "0.00", "total_tokens": 0} | |
| {"current_steps": 2691, "total_steps": 2691, "epoch": 3.0, "percentage": 100.0, "elapsed_time": "9:42:37", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0} | |