{"current_steps": 1, "total_steps": 2679, "loss": 1.2427, "learning_rate": 2.0000000000000002e-07, "epoch": 0.0011188811188811189, "percentage": 0.04, "elapsed_time": "0:00:59", "remaining_time": "1 day, 20:26:59", "throughput": "17548.36", "total_tokens": 1048576}
{"current_steps": 2, "total_steps": 2679, "loss": 1.3185, "learning_rate": 4.0000000000000003e-07, "epoch": 0.0022377622377622378, "percentage": 0.07, "elapsed_time": "0:02:00", "remaining_time": "1 day, 20:42:55", "throughput": "17437.69", "total_tokens": 2097152}
{"current_steps": 3, "total_steps": 2679, "loss": 1.2261, "learning_rate": 6.000000000000001e-07, "epoch": 0.0033566433566433566, "percentage": 0.11, "elapsed_time": "0:02:53", "remaining_time": "1 day, 19:00:45", "throughput": "18121.26", "total_tokens": 3145728}
{"current_steps": 4, "total_steps": 2679, "loss": 1.3815, "learning_rate": 8.000000000000001e-07, "epoch": 0.0044755244755244755, "percentage": 0.15, "elapsed_time": "0:03:54", "remaining_time": "1 day, 19:32:07", "throughput": "17896.88", "total_tokens": 4194304}
{"current_steps": 5, "total_steps": 2679, "loss": 1.3404, "learning_rate": 1.0000000000000002e-06, "epoch": 0.005594405594405594, "percentage": 0.19, "elapsed_time": "0:04:46", "remaining_time": "1 day, 18:35:19", "throughput": "18287.94", "total_tokens": 5242880}
{"current_steps": 6, "total_steps": 2679, "loss": 1.236, "learning_rate": 1.2000000000000002e-06, "epoch": 0.006713286713286713, "percentage": 0.22, "elapsed_time": "0:05:51", "remaining_time": "1 day, 19:30:53", "throughput": "17892.01", "total_tokens": 6291456}
{"current_steps": 7, "total_steps": 2679, "loss": 1.2324, "learning_rate": 1.4000000000000001e-06, "epoch": 0.007832167832167832, "percentage": 0.26, "elapsed_time": "0:06:47", "remaining_time": "1 day, 19:14:14", "throughput": "18000.09", "total_tokens": 7340032}
{"current_steps": 8, "total_steps": 2679, "loss": 1.1904, "learning_rate": 1.6000000000000001e-06, "epoch": 0.008951048951048951, "percentage": 0.3, "elapsed_time": "0:07:53", "remaining_time": "1 day, 19:52:06", "throughput": "17734.49", "total_tokens": 8388608}
{"current_steps": 9, "total_steps": 2679, "loss": 1.2405, "learning_rate": 1.8000000000000001e-06, "epoch": 0.01006993006993007, "percentage": 0.34, "elapsed_time": "0:08:52", "remaining_time": "1 day, 19:51:14", "throughput": "17733.74", "total_tokens": 9437184}
{"current_steps": 10, "total_steps": 2679, "loss": 1.0979, "learning_rate": 2.0000000000000003e-06, "epoch": 0.011188811188811189, "percentage": 0.37, "elapsed_time": "0:09:46", "remaining_time": "1 day, 19:27:03", "throughput": "17891.46", "total_tokens": 10485760}
{"current_steps": 11, "total_steps": 2679, "loss": 1.004, "learning_rate": 2.2e-06, "epoch": 0.012307692307692308, "percentage": 0.41, "elapsed_time": "0:10:34", "remaining_time": "1 day, 18:44:31", "throughput": "18181.42", "total_tokens": 11534336}
{"current_steps": 12, "total_steps": 2679, "loss": 0.9888, "learning_rate": 2.4000000000000003e-06, "epoch": 0.013426573426573427, "percentage": 0.45, "elapsed_time": "0:11:28", "remaining_time": "1 day, 18:29:58", "throughput": "18278.26", "total_tokens": 12582912}
{"current_steps": 13, "total_steps": 2679, "loss": 1.0584, "learning_rate": 2.6e-06, "epoch": 0.014545454545454545, "percentage": 0.49, "elapsed_time": "0:12:27", "remaining_time": "1 day, 18:35:54", "throughput": "18229.05", "total_tokens": 13631488}
{"current_steps": 14, "total_steps": 2679, "loss": 0.9863, "learning_rate": 2.8000000000000003e-06, "epoch": 0.015664335664335664, "percentage": 0.52, "elapsed_time": "0:13:26", "remaining_time": "1 day, 18:40:06", "throughput": "18192.32", "total_tokens": 14680064}
{"current_steps": 15, "total_steps": 2679, "loss": 0.9604, "learning_rate": 3e-06, "epoch": 0.016783216783216783, "percentage": 0.56, "elapsed_time": "0:14:20", "remaining_time": "1 day, 18:28:09", "throughput": "18270.75", "total_tokens": 15728640}
{"current_steps": 16, "total_steps": 2679, "loss": 0.9025, "learning_rate": 3.2000000000000003e-06, "epoch": 0.017902097902097902, "percentage": 0.6, "elapsed_time": "0:15:14", "remaining_time": "1 day, 18:16:17", "throughput": "18349.29", "total_tokens": 16777216}
{"current_steps": 17, "total_steps": 2679, "loss": 1.004, "learning_rate": 3.4000000000000005e-06, "epoch": 0.01902097902097902, "percentage": 0.63, "elapsed_time": "0:16:17", "remaining_time": "1 day, 18:31:08", "throughput": "18235.70", "total_tokens": 17825792}
{"current_steps": 18, "total_steps": 2679, "loss": 0.973, "learning_rate": 3.6000000000000003e-06, "epoch": 0.02013986013986014, "percentage": 0.67, "elapsed_time": "0:17:17", "remaining_time": "1 day, 18:36:51", "throughput": "18188.14", "total_tokens": 18874368}
{"current_steps": 19, "total_steps": 2679, "loss": 0.9635, "learning_rate": 3.8000000000000005e-06, "epoch": 0.02125874125874126, "percentage": 0.71, "elapsed_time": "0:18:17", "remaining_time": "1 day, 18:39:58", "throughput": "18159.05", "total_tokens": 19922944}
{"current_steps": 20, "total_steps": 2679, "loss": 0.8557, "learning_rate": 4.000000000000001e-06, "epoch": 0.022377622377622378, "percentage": 0.75, "elapsed_time": "0:19:07", "remaining_time": "1 day, 18:22:29", "throughput": "18277.15", "total_tokens": 20971520}
{"current_steps": 21, "total_steps": 2679, "loss": 1.0787, "learning_rate": 4.2000000000000004e-06, "epoch": 0.023496503496503496, "percentage": 0.78, "elapsed_time": "0:20:13", "remaining_time": "1 day, 18:39:32", "throughput": "18148.55", "total_tokens": 22020096}
{"current_steps": 22, "total_steps": 2679, "loss": 0.8179, "learning_rate": 4.4e-06, "epoch": 0.024615384615384615, "percentage": 0.82, "elapsed_time": "0:21:01", "remaining_time": "1 day, 18:19:10", "throughput": "18287.20", "total_tokens": 23068672}
{"current_steps": 23, "total_steps": 2679, "loss": 0.9493, "learning_rate": 4.600000000000001e-06, "epoch": 0.025734265734265734, "percentage": 0.86, "elapsed_time": "0:22:01", "remaining_time": "1 day, 18:22:52", "throughput": "18253.74", "total_tokens": 24117248}
{"current_steps": 24, "total_steps": 2679, "loss": 0.8281, "learning_rate": 4.800000000000001e-06, "epoch": 0.026853146853146853, "percentage": 0.9, "elapsed_time": "0:22:49", "remaining_time": "1 day, 18:04:46", "throughput": "18377.64", "total_tokens": 25165824}
{"current_steps": 25, "total_steps": 2679, "loss": 1.0281, "learning_rate": 5e-06, "epoch": 0.027972027972027972, "percentage": 0.93, "elapsed_time": "0:23:54", "remaining_time": "1 day, 18:18:48", "throughput": "18269.26", "total_tokens": 26214400}
{"current_steps": 26, "total_steps": 2679, "loss": 0.8749, "learning_rate": 4.999998248509159e-06, "epoch": 0.02909090909090909, "percentage": 0.97, "elapsed_time": "0:24:48", "remaining_time": "1 day, 18:11:21", "throughput": "18316.06", "total_tokens": 27262976}
{"current_steps": 27, "total_steps": 2679, "loss": 0.8868, "learning_rate": 4.99999299403909e-06, "epoch": 0.03020979020979021, "percentage": 1.01, "elapsed_time": "0:25:43", "remaining_time": "1 day, 18:06:22", "throughput": "18345.27", "total_tokens": 28311552}
{"current_steps": 28, "total_steps": 2679, "loss": 0.8974, "learning_rate": 4.999984236597156e-06, "epoch": 0.03132867132867133, "percentage": 1.05, "elapsed_time": "0:26:42", "remaining_time": "1 day, 18:09:29", "throughput": "18315.82", "total_tokens": 29360128}
{"current_steps": 29, "total_steps": 2679, "loss": 0.9162, "learning_rate": 4.999971976195628e-06, "epoch": 0.03244755244755245, "percentage": 1.08, "elapsed_time": "0:27:51", "remaining_time": "1 day, 18:25:56", "throughput": "18190.61", "total_tokens": 30408704}
{"current_steps": 30, "total_steps": 2679, "loss": 0.9994, "learning_rate": 4.9999562128516835e-06, "epoch": 0.033566433566433566, "percentage": 1.12, "elapsed_time": "0:28:55", "remaining_time": "1 day, 18:33:34", "throughput": "18129.38", "total_tokens": 31457280}
{"current_steps": 31, "total_steps": 2679, "loss": 1.0164, "learning_rate": 4.999936946587412e-06, "epoch": 0.03468531468531469, "percentage": 1.16, "elapsed_time": "0:30:00", "remaining_time": "1 day, 18:43:51", "throughput": "18049.84", "total_tokens": 32505856}
{"current_steps": 32, "total_steps": 2679, "loss": 0.9654, "learning_rate": 4.999914177429808e-06, "epoch": 0.035804195804195804, "percentage": 1.19, "elapsed_time": "0:31:01", "remaining_time": "1 day, 18:46:15", "throughput": "18026.15", "total_tokens": 33554432}
{"current_steps": 33, "total_steps": 2679, "loss": 0.9447, "learning_rate": 4.999887905410775e-06, "epoch": 0.036923076923076927, "percentage": 1.23, "elapsed_time": "0:31:55", "remaining_time": "1 day, 18:40:26", "throughput": "18060.20", "total_tokens": 34603008}
{"current_steps": 34, "total_steps": 2679, "loss": 0.7911, "learning_rate": 4.999858130567127e-06, "epoch": 0.03804195804195804, "percentage": 1.27, "elapsed_time": "0:32:44", "remaining_time": "1 day, 18:26:58", "throughput": "18148.88", "total_tokens": 35651584}
{"current_steps": 35, "total_steps": 2679, "loss": 0.8502, "learning_rate": 4.999824852940583e-06, "epoch": 0.039160839160839164, "percentage": 1.31, "elapsed_time": "0:33:38", "remaining_time": "1 day, 18:21:00", "throughput": "18184.63", "total_tokens": 36700160}
{"current_steps": 36, "total_steps": 2679, "loss": 0.8437, "learning_rate": 4.999788072577773e-06, "epoch": 0.04027972027972028, "percentage": 1.34, "elapsed_time": "0:34:32", "remaining_time": "1 day, 18:15:58", "throughput": "18213.87", "total_tokens": 37748736}
{"current_steps": 37, "total_steps": 2679, "loss": 0.7833, "learning_rate": 4.99974778953023e-06, "epoch": 0.0413986013986014, "percentage": 1.38, "elapsed_time": "0:35:21", "remaining_time": "1 day, 18:04:42", "throughput": "18288.20", "total_tokens": 38797312}
{"current_steps": 38, "total_steps": 2679, "loss": 0.9155, "learning_rate": 4.999704003854402e-06, "epoch": 0.04251748251748252, "percentage": 1.42, "elapsed_time": "0:36:21", "remaining_time": "1 day, 18:06:43", "throughput": "18266.66", "total_tokens": 39845888}
{"current_steps": 39, "total_steps": 2679, "loss": 0.9428, "learning_rate": 4.9996567156116395e-06, "epoch": 0.04363636363636364, "percentage": 1.46, "elapsed_time": "0:37:27", "remaining_time": "1 day, 18:15:24", "throughput": "18197.22", "total_tokens": 40894464}
{"current_steps": 40, "total_steps": 2679, "loss": 0.8479, "learning_rate": 4.999605924868201e-06, "epoch": 0.044755244755244755, "percentage": 1.49, "elapsed_time": "0:38:26", "remaining_time": "1 day, 18:16:23", "throughput": "18183.27", "total_tokens": 41943040}
{"current_steps": 41, "total_steps": 2679, "loss": 0.8798, "learning_rate": 4.999551631695257e-06, "epoch": 0.04587412587412588, "percentage": 1.53, "elapsed_time": "0:39:26", "remaining_time": "1 day, 18:18:05", "throughput": "18164.22", "total_tokens": 42991616}
{"current_steps": 42, "total_steps": 2679, "loss": 0.7788, "learning_rate": 4.999493836168882e-06, "epoch": 0.04699300699300699, "percentage": 1.57, "elapsed_time": "0:40:15", "remaining_time": "1 day, 18:07:19", "throughput": "18234.62", "total_tokens": 44040192}
{"current_steps": 43, "total_steps": 2679, "loss": 0.8416, "learning_rate": 4.999432538370057e-06, "epoch": 0.048111888111888115, "percentage": 1.61, "elapsed_time": "0:41:14", "remaining_time": "1 day, 18:08:30", "throughput": "18219.19", "total_tokens": 45088768}
{"current_steps": 44, "total_steps": 2679, "loss": 1.1177, "learning_rate": 4.999367738384673e-06, "epoch": 0.04923076923076923, "percentage": 1.64, "elapsed_time": "0:42:27", "remaining_time": "1 day, 18:22:10", "throughput": "18114.35", "total_tokens": 46137344}
{"current_steps": 45, "total_steps": 2679, "loss": 1.0214, "learning_rate": 4.999299436303527e-06, "epoch": 0.05034965034965035, "percentage": 1.68, "elapsed_time": "0:43:38", "remaining_time": "1 day, 18:34:39", "throughput": "18019.00", "total_tokens": 47185920}
{"current_steps": 46, "total_steps": 2679, "loss": 0.9225, "learning_rate": 4.999227632222324e-06, "epoch": 0.05146853146853147, "percentage": 1.72, "elapsed_time": "0:44:44", "remaining_time": "1 day, 18:40:45", "throughput": "17969.26", "total_tokens": 48234496}
{"current_steps": 47, "total_steps": 2679, "loss": 0.8194, "learning_rate": 4.999152326241675e-06, "epoch": 0.05258741258741259, "percentage": 1.75, "elapsed_time": "0:45:37", "remaining_time": "1 day, 18:35:25", "throughput": "17999.93", "total_tokens": 49283072}
{"current_steps": 48, "total_steps": 2679, "loss": 0.8172, "learning_rate": 4.999073518467098e-06, "epoch": 0.053706293706293706, "percentage": 1.79, "elapsed_time": "0:46:31", "remaining_time": "1 day, 18:30:26", "throughput": "18028.23", "total_tokens": 50331648}
{"current_steps": 49, "total_steps": 2679, "loss": 0.8096, "learning_rate": 4.998991209009019e-06, "epoch": 0.05482517482517483, "percentage": 1.83, "elapsed_time": "0:47:25", "remaining_time": "1 day, 18:25:53", "throughput": "18053.59", "total_tokens": 51380224}
{"current_steps": 50, "total_steps": 2679, "loss": 0.9269, "learning_rate": 4.998905397982767e-06, "epoch": 0.055944055944055944, "percentage": 1.87, "elapsed_time": "0:48:21", "remaining_time": "1 day, 18:22:16", "throughput": "18072.46", "total_tokens": 52428800}
{"current_steps": 51, "total_steps": 2679, "loss": 0.8298, "learning_rate": 4.998816085508582e-06, "epoch": 0.057062937062937066, "percentage": 1.9, "elapsed_time": "0:49:20", "remaining_time": "1 day, 18:22:24", "throughput": "18064.66", "total_tokens": 53477376}
{"current_steps": 52, "total_steps": 2679, "loss": 0.7906, "learning_rate": 4.998723271711607e-06, "epoch": 0.05818181818181818, "percentage": 1.94, "elapsed_time": "0:50:15", "remaining_time": "1 day, 18:18:57", "throughput": "18082.22", "total_tokens": 54525952}
{"current_steps": 53, "total_steps": 2679, "loss": 0.7859, "learning_rate": 4.998626956721894e-06, "epoch": 0.059300699300699304, "percentage": 1.98, "elapsed_time": "0:51:06", "remaining_time": "1 day, 18:12:12", "throughput": "18123.62", "total_tokens": 55574528}
{"current_steps": 54, "total_steps": 2679, "loss": 0.8241, "learning_rate": 4.998527140674395e-06, "epoch": 0.06041958041958042, "percentage": 2.02, "elapsed_time": "0:52:01", "remaining_time": "1 day, 18:08:56", "throughput": "18140.07", "total_tokens": 56623104}
{"current_steps": 55, "total_steps": 2679, "loss": 0.8721, "learning_rate": 4.998423823708974e-06, "epoch": 0.06153846153846154, "percentage": 2.05, "elapsed_time": "0:53:01", "remaining_time": "1 day, 18:09:25", "throughput": "18129.65", "total_tokens": 57671680}
{"current_steps": 56, "total_steps": 2679, "loss": 0.8404, "learning_rate": 4.998317005970398e-06, "epoch": 0.06265734265734266, "percentage": 2.09, "elapsed_time": "0:53:55", "remaining_time": "1 day, 18:05:38", "throughput": "18149.92", "total_tokens": 58720256}
{"current_steps": 57, "total_steps": 2679, "loss": 0.7641, "learning_rate": 4.998206687608339e-06, "epoch": 0.06377622377622377, "percentage": 2.13, "elapsed_time": "0:54:53", "remaining_time": "1 day, 18:04:38", "throughput": "18150.27", "total_tokens": 59768832}
{"current_steps": 58, "total_steps": 2679, "loss": 0.9132, "learning_rate": 4.998092868777374e-06, "epoch": 0.0648951048951049, "percentage": 2.16, "elapsed_time": "0:55:58", "remaining_time": "1 day, 18:09:49", "throughput": "18106.15", "total_tokens": 60817408}
{"current_steps": 59, "total_steps": 2679, "loss": 0.9944, "learning_rate": 4.997975549636985e-06, "epoch": 0.06601398601398602, "percentage": 2.2, "elapsed_time": "0:57:10", "remaining_time": "1 day, 18:18:50", "throughput": "18034.87", "total_tokens": 61865984}
{"current_steps": 60, "total_steps": 2679, "loss": 0.8668, "learning_rate": 4.997854730351559e-06, "epoch": 0.06713286713286713, "percentage": 2.24, "elapsed_time": "0:58:10", "remaining_time": "1 day, 18:19:14", "throughput": "18025.18", "total_tokens": 62914560}
{"current_steps": 61, "total_steps": 2679, "loss": 0.7948, "learning_rate": 4.997730411090387e-06, "epoch": 0.06825174825174825, "percentage": 2.28, "elapsed_time": "0:59:03", "remaining_time": "1 day, 18:14:55", "throughput": "18048.99", "total_tokens": 63963136}
{"current_steps": 62, "total_steps": 2679, "loss": 0.7899, "learning_rate": 4.997602592027664e-06, "epoch": 0.06937062937062938, "percentage": 2.31, "elapsed_time": "0:59:57", "remaining_time": "1 day, 18:10:59", "throughput": "18070.15", "total_tokens": 65011712}
{"current_steps": 63, "total_steps": 2679, "loss": 0.7772, "learning_rate": 4.9974712733424905e-06, "epoch": 0.07048951048951049, "percentage": 2.35, "elapsed_time": "1:00:51", "remaining_time": "1 day, 18:07:14", "throughput": "18090.00", "total_tokens": 66060288}
{"current_steps": 64, "total_steps": 2679, "loss": 0.9086, "learning_rate": 4.997336455218868e-06, "epoch": 0.07160839160839161, "percentage": 2.39, "elapsed_time": "1:01:54", "remaining_time": "1 day, 18:09:17", "throughput": "18068.43", "total_tokens": 67108864}
{"current_steps": 65, "total_steps": 2679, "loss": 0.8423, "learning_rate": 4.997198137845702e-06, "epoch": 0.07272727272727272, "percentage": 2.43, "elapsed_time": "1:02:59", "remaining_time": "1 day, 18:13:01", "throughput": "18034.94", "total_tokens": 68157440}
{"current_steps": 66, "total_steps": 2679, "loss": 0.9755, "learning_rate": 4.997056321416803e-06, "epoch": 0.07384615384615385, "percentage": 2.46, "elapsed_time": "1:04:05", "remaining_time": "1 day, 18:17:37", "throughput": "17995.40", "total_tokens": 69206016}
{"current_steps": 67, "total_steps": 2679, "loss": 0.833, "learning_rate": 4.9969110061308826e-06, "epoch": 0.07496503496503497, "percentage": 2.5, "elapsed_time": "1:05:00", "remaining_time": "1 day, 18:14:07", "throughput": "18013.31", "total_tokens": 70254592}
{"current_steps": 68, "total_steps": 2679, "loss": 0.8858, "learning_rate": 4.996762192191556e-06, "epoch": 0.07608391608391608, "percentage": 2.54, "elapsed_time": "1:06:05", "remaining_time": "1 day, 18:17:28", "throughput": "17982.60", "total_tokens": 71303168}
{"current_steps": 69, "total_steps": 2679, "loss": 0.7604, "learning_rate": 4.996609879807341e-06, "epoch": 0.0772027972027972, "percentage": 2.58, "elapsed_time": "1:06:58", "remaining_time": "1 day, 18:13:37", "throughput": "18003.07", "total_tokens": 72351744}
{"current_steps": 70, "total_steps": 2679, "loss": 0.9583, "learning_rate": 4.996454069191653e-06, "epoch": 0.07832167832167833, "percentage": 2.61, "elapsed_time": "1:07:59", "remaining_time": "1 day, 18:14:17", "throughput": "17991.47", "total_tokens": 73400320}
{"current_steps": 71, "total_steps": 2679, "loss": 0.809, "learning_rate": 4.996294760562817e-06, "epoch": 0.07944055944055944, "percentage": 2.65, "elapsed_time": "1:08:59", "remaining_time": "1 day, 18:14:11", "throughput": "17985.29", "total_tokens": 74448896}
{"current_steps": 72, "total_steps": 2679, "loss": 0.8756, "learning_rate": 4.996131954144053e-06, "epoch": 0.08055944055944056, "percentage": 2.69, "elapsed_time": "1:09:59", "remaining_time": "1 day, 18:14:18", "throughput": "17977.54", "total_tokens": 75497472}
{"current_steps": 73, "total_steps": 2679, "loss": 0.7074, "learning_rate": 4.995965650163485e-06, "epoch": 0.08167832167832167, "percentage": 2.72, "elapsed_time": "1:10:47", "remaining_time": "1 day, 18:07:01", "throughput": "18022.43", "total_tokens": 76546048}
{"current_steps": 74, "total_steps": 2679, "loss": 0.8109, "learning_rate": 4.995795848854134e-06, "epoch": 0.0827972027972028, "percentage": 2.76, "elapsed_time": "1:11:46", "remaining_time": "1 day, 18:06:55", "throughput": "18016.26", "total_tokens": 77594624}
{"current_steps": 75, "total_steps": 2679, "loss": 0.8878, "learning_rate": 4.995622550453929e-06, "epoch": 0.08391608391608392, "percentage": 2.8, "elapsed_time": "1:12:41", "remaining_time": "1 day, 18:03:49", "throughput": "18031.46", "total_tokens": 78643200}
{"current_steps": 76, "total_steps": 2679, "loss": 0.7673, "learning_rate": 4.995445755205692e-06, "epoch": 0.08503496503496503, "percentage": 2.84, "elapsed_time": "1:13:34", "remaining_time": "1 day, 18:00:11", "throughput": "18050.55", "total_tokens": 79691776}
{"current_steps": 77, "total_steps": 2679, "loss": 0.8226, "learning_rate": 4.995265463357147e-06, "epoch": 0.08615384615384615, "percentage": 2.87, "elapsed_time": "1:14:34", "remaining_time": "1 day, 18:00:09", "throughput": "18043.80", "total_tokens": 80740352}
{"current_steps": 78, "total_steps": 2679, "loss": 0.7972, "learning_rate": 4.995081675160918e-06, "epoch": 0.08727272727272728, "percentage": 2.91, "elapsed_time": "1:15:28", "remaining_time": "1 day, 17:56:52", "throughput": "18060.38", "total_tokens": 81788928}
{"current_steps": 79, "total_steps": 2679, "loss": 0.8637, "learning_rate": 4.994894390874527e-06, "epoch": 0.0883916083916084, "percentage": 2.95, "elapsed_time": "1:16:28", "remaining_time": "1 day, 17:57:04", "throughput": "18052.01", "total_tokens": 82837504}
{"current_steps": 80, "total_steps": 2679, "loss": 0.8085, "learning_rate": 4.9947036107603975e-06, "epoch": 0.08951048951048951, "percentage": 2.99, "elapsed_time": "1:17:23", "remaining_time": "1 day, 17:54:25", "throughput": "18064.09", "total_tokens": 83886080}
{"current_steps": 81, "total_steps": 2679, "loss": 0.8668, "learning_rate": 4.994509335085847e-06, "epoch": 0.09062937062937063, "percentage": 3.02, "elapsed_time": "1:18:34", "remaining_time": "1 day, 18:00:24", "throughput": "18014.33", "total_tokens": 84934656}
{"current_steps": 82, "total_steps": 2679, "loss": 0.8147, "learning_rate": 4.994311564123093e-06, "epoch": 0.09174825174825176, "percentage": 3.06, "elapsed_time": "1:19:28", "remaining_time": "1 day, 17:57:07", "throughput": "18030.80", "total_tokens": 85983232}
{"current_steps": 83, "total_steps": 2679, "loss": 0.9393, "learning_rate": 4.994110298149253e-06, "epoch": 0.09286713286713287, "percentage": 3.1, "elapsed_time": "1:20:34", "remaining_time": "1 day, 17:59:56", "throughput": "18003.70", "total_tokens": 87031808}
{"current_steps": 84, "total_steps": 2679, "loss": 0.8956, "learning_rate": 4.993905537446337e-06, "epoch": 0.09398601398601399, "percentage": 3.14, "elapsed_time": "1:21:38", "remaining_time": "1 day, 18:02:21", "throughput": "17979.52", "total_tokens": 88080384}
{"current_steps": 85, "total_steps": 2679, "loss": 0.8612, "learning_rate": 4.993697282301256e-06, "epoch": 0.0951048951048951, "percentage": 3.17, "elapsed_time": "1:22:38", "remaining_time": "1 day, 18:01:59", "throughput": "17975.20", "total_tokens": 89128960}
{"current_steps": 86, "total_steps": 2679, "loss": 0.8561, "learning_rate": 4.9934855330058145e-06, "epoch": 0.09622377622377623, "percentage": 3.21, "elapsed_time": "1:23:38", "remaining_time": "1 day, 18:01:43", "throughput": "17970.21", "total_tokens": 90177536}
{"current_steps": 87, "total_steps": 2679, "loss": 0.8474, "learning_rate": 4.993270289856714e-06, "epoch": 0.09734265734265735, "percentage": 3.25, "elapsed_time": "1:24:43", "remaining_time": "1 day, 18:04:12", "throughput": "17945.67", "total_tokens": 91226112}
{"current_steps": 88, "total_steps": 2679, "loss": 0.9317, "learning_rate": 4.993051553155552e-06, "epoch": 0.09846153846153846, "percentage": 3.28, "elapsed_time": "1:25:48", "remaining_time": "1 day, 18:06:29", "throughput": "17922.43", "total_tokens": 92274688}
{"current_steps": 89, "total_steps": 2679, "loss": 0.7798, "learning_rate": 4.992829323208822e-06, "epoch": 0.09958041958041958, "percentage": 3.32, "elapsed_time": "1:26:42", "remaining_time": "1 day, 18:03:14", "throughput": "17938.62", "total_tokens": 93323264}
{"current_steps": 90, "total_steps": 2679, "loss": 0.83, "learning_rate": 4.992603600327909e-06, "epoch": 0.1006993006993007, "percentage": 3.36, "elapsed_time": "1:27:42", "remaining_time": "1 day, 18:02:53", "throughput": "17934.19", "total_tokens": 94371840}
{"current_steps": 91, "total_steps": 2679, "loss": 0.806, "learning_rate": 4.992374384829094e-06, "epoch": 0.10181818181818182, "percentage": 3.4, "elapsed_time": "1:28:41", "remaining_time": "1 day, 18:02:16", "throughput": "17931.63", "total_tokens": 95420416}
{"current_steps": 92, "total_steps": 2679, "loss": 0.8123, "learning_rate": 4.992141677033554e-06, "epoch": 0.10293706293706294, "percentage": 3.43, "elapsed_time": "1:29:40", "remaining_time": "1 day, 18:01:39", "throughput": "17929.14", "total_tokens": 96468992}
{"current_steps": 93, "total_steps": 2679, "loss": 0.8025, "learning_rate": 4.991905477267356e-06, "epoch": 0.10405594405594405, "percentage": 3.47, "elapsed_time": "1:30:34", "remaining_time": "1 day, 17:58:46", "throughput": "17942.73", "total_tokens": 97517568}
{"current_steps": 94, "total_steps": 2679, "loss": 0.7708, "learning_rate": 4.991665785861463e-06, "epoch": 0.10517482517482518, "percentage": 3.51, "elapsed_time": "1:31:28", "remaining_time": "1 day, 17:55:32", "throughput": "17958.79", "total_tokens": 98566144}
{"current_steps": 95, "total_steps": 2679, "loss": 0.6959, "learning_rate": 4.991422603151727e-06, "epoch": 0.1062937062937063, "percentage": 3.55, "elapsed_time": "1:32:16", "remaining_time": "1 day, 17:49:48", "throughput": "17992.87", "total_tokens": 99614720}
{"current_steps": 96, "total_steps": 2679, "loss": 0.8078, "learning_rate": 4.991175929478894e-06, "epoch": 0.10741258741258741, "percentage": 3.58, "elapsed_time": "1:33:10", "remaining_time": "1 day, 17:46:56", "throughput": "18006.51", "total_tokens": 100663296}
{"current_steps": 97, "total_steps": 2679, "loss": 0.7607, "learning_rate": 4.990925765188602e-06, "epoch": 0.10853146853146853, "percentage": 3.62, "elapsed_time": "1:34:04", "remaining_time": "1 day, 17:44:10", "throughput": "18019.38", "total_tokens": 101711872}
{"current_steps": 98, "total_steps": 2679, "loss": 0.8431, "learning_rate": 4.990672110631379e-06, "epoch": 0.10965034965034966, "percentage": 3.66, "elapsed_time": "1:35:04", "remaining_time": "1 day, 17:43:52", "throughput": "18014.58", "total_tokens": 102760448}
{"current_steps": 99, "total_steps": 2679, "loss": 0.9402, "learning_rate": 4.9904149661626456e-06, "epoch": 0.11076923076923077, "percentage": 3.7, "elapsed_time": "1:36:04", "remaining_time": "1 day, 17:43:43", "throughput": "18008.68", "total_tokens": 103809024}
{"current_steps": 100, "total_steps": 2679, "loss": 0.793, "learning_rate": 4.990154332142708e-06, "epoch": 0.11188811188811189, "percentage": 3.73, "elapsed_time": "1:37:04", "remaining_time": "1 day, 17:43:42", "throughput": "18001.80", "total_tokens": 104857600}
{"current_steps": 101, "total_steps": 2679, "loss": 0.8518, "learning_rate": 4.989890208936767e-06, "epoch": 0.113006993006993, "percentage": 3.77, "elapsed_time": "1:38:04", "remaining_time": "1 day, 17:43:18", "throughput": "17997.77", "total_tokens": 105906176}
{"current_steps": 102, "total_steps": 2679, "loss": 0.9362, "learning_rate": 4.989622596914908e-06, "epoch": 0.11412587412587413, "percentage": 3.81, "elapsed_time": "1:39:09", "remaining_time": "1 day, 17:45:17", "throughput": "17976.50", "total_tokens": 106954752}
{"current_steps": 103, "total_steps": 2679, "loss": 0.8008, "learning_rate": 4.989351496452109e-06, "epoch": 0.11524475524475525, "percentage": 3.84, "elapsed_time": "1:40:03", "remaining_time": "1 day, 17:42:31", "throughput": "17989.37", "total_tokens": 108003328}
{"current_steps": 104, "total_steps": 2679, "loss": 0.7556, "learning_rate": 4.989076907928233e-06, "epoch": 0.11636363636363636, "percentage": 3.88, "elapsed_time": "1:40:51", "remaining_time": "1 day, 17:37:10", "throughput": "18020.96", "total_tokens": 109051904}
{"current_steps": 105, "total_steps": 2679, "loss": 0.8277, "learning_rate": 4.988798831728031e-06, "epoch": 0.11748251748251748, "percentage": 3.92, "elapsed_time": "1:41:45", "remaining_time": "1 day, 17:34:35", "throughput": "18032.61", "total_tokens": 110100480}
{"current_steps": 106, "total_steps": 2679, "loss": 0.95, "learning_rate": 4.988517268241142e-06, "epoch": 0.11860139860139861, "percentage": 3.96, "elapsed_time": "1:42:51", "remaining_time": "1 day, 17:36:50", "throughput": "18009.37", "total_tokens": 111149056}
{"current_steps": 107, "total_steps": 2679, "loss": 0.8442, "learning_rate": 4.988232217862091e-06, "epoch": 0.11972027972027972, "percentage": 3.99, "elapsed_time": "1:43:50", "remaining_time": "1 day, 17:36:16", "throughput": "18006.46", "total_tokens": 112197632}
{"current_steps": 108, "total_steps": 2679, "loss": 1.0015, "learning_rate": 4.987943680990288e-06, "epoch": 0.12083916083916084, "percentage": 4.03, "elapsed_time": "1:45:03", "remaining_time": "1 day, 17:40:51", "throughput": "17966.38", "total_tokens": 113246208}
{"current_steps": 109, "total_steps": 2679, "loss": 0.8104, "learning_rate": 4.9876516580300285e-06, "epoch": 0.12195804195804195, "percentage": 4.07, "elapsed_time": "1:46:02", "remaining_time": "1 day, 17:40:19", "throughput": "17963.30", "total_tokens": 114294784}
{"current_steps": 110, "total_steps": 2679, "loss": 1.0469, "learning_rate": 4.987356149390493e-06, "epoch": 0.12307692307692308, "percentage": 4.11, "elapsed_time": "1:47:09", "remaining_time": "1 day, 17:42:29", "throughput": "17940.72", "total_tokens": 115343360}
{"current_steps": 111, "total_steps": 2679, "loss": 0.7177, "learning_rate": 4.987057155485746e-06, "epoch": 0.1241958041958042, "percentage": 4.14, "elapsed_time": "1:47:59", "remaining_time": "1 day, 17:38:24", "throughput": "17963.10", "total_tokens": 116391936}
{"current_steps": 112, "total_steps": 2679, "loss": 0.8348, "learning_rate": 4.986754676734737e-06, "epoch": 0.12531468531468531, "percentage": 4.18, "elapsed_time": "1:48:59", "remaining_time": "1 day, 17:38:04", "throughput": "17958.50", "total_tokens": 117440512}
{"current_steps": 113, "total_steps": 2679, "loss": 0.7816, "learning_rate": 4.986448713561295e-06, "epoch": 0.12643356643356643, "percentage": 4.22, "elapsed_time": "1:49:53", "remaining_time": "1 day, 17:35:27", "throughput": "17970.32", "total_tokens": 118489088}
{"current_steps": 114, "total_steps": 2679, "loss": 0.9002, "learning_rate": 4.986139266394134e-06, "epoch": 0.12755244755244755, "percentage": 4.26, "elapsed_time": "1:50:59", "remaining_time": "1 day, 17:37:13", "throughput": "17950.63", "total_tokens": 119537664}
{"current_steps": 115, "total_steps": 2679, "loss": 0.8056, "learning_rate": 4.9858263356668505e-06, "epoch": 0.12867132867132866, "percentage": 4.29, "elapsed_time": "1:51:53", "remaining_time": "1 day, 17:34:40", "throughput": "17961.86", "total_tokens": 120586240}
{"current_steps": 116, "total_steps": 2679, "loss": 1.0116, "learning_rate": 4.9855099218179186e-06, "epoch": 0.1297902097902098, "percentage": 4.33, "elapsed_time": "1:52:59", "remaining_time": "1 day, 17:36:36", "throughput": "17940.97", "total_tokens": 121634816}
{"current_steps": 117, "total_steps": 2679, "loss": 0.8787, "learning_rate": 4.985190025290696e-06, "epoch": 0.13090909090909092, "percentage": 4.37, "elapsed_time": "1:54:05", "remaining_time": "1 day, 17:38:14", "throughput": "17922.24", "total_tokens": 122683392}
{"current_steps": 118, "total_steps": 2679, "loss": 0.8072, "learning_rate": 4.98486664653342e-06, "epoch": 0.13202797202797203, "percentage": 4.4, "elapsed_time": "1:55:00", "remaining_time": "1 day, 17:35:54", "throughput": "17932.05", "total_tokens": 123731968}
{"current_steps": 119, "total_steps": 2679, "loss": 0.8445, "learning_rate": 4.984539785999205e-06, "epoch": 0.13314685314685315, "percentage": 4.44, "elapsed_time": "1:56:00", "remaining_time": "1 day, 17:35:29", "throughput": "17927.99", "total_tokens": 124780544}
{"current_steps": 120, "total_steps": 2679, "loss": 0.9251, "learning_rate": 4.9842094441460476e-06, "epoch": 0.13426573426573427, "percentage": 4.48, "elapsed_time": "1:57:02", "remaining_time": "1 day, 17:35:53", "throughput": "17918.14", "total_tokens": 125829120}
{"current_steps": 121, "total_steps": 2679, "loss": 0.8878, "learning_rate": 4.9838756214368185e-06, "epoch": 0.13538461538461538, "percentage": 4.52, "elapsed_time": "1:58:01", "remaining_time": "1 day, 17:35:16", "throughput": "17915.52", "total_tokens": 126877696}
{"current_steps": 122, "total_steps": 2679, "loss": 0.779, "learning_rate": 4.983538318339268e-06, "epoch": 0.1365034965034965, "percentage": 4.55, "elapsed_time": "1:58:55", "remaining_time": "1 day, 17:32:34", "throughput": "17927.97", "total_tokens": 127926272}
{"current_steps": 123, "total_steps": 2679, "loss": 0.7563, "learning_rate": 4.983197535326024e-06, "epoch": 0.1376223776223776, "percentage": 4.59, "elapsed_time": "1:59:49", "remaining_time": "1 day, 17:30:00", "throughput": "17939.40", "total_tokens": 128974848}
{"current_steps": 124, "total_steps": 2679, "loss": 0.7584, "learning_rate": 4.982853272874589e-06, "epoch": 0.13874125874125876, "percentage": 4.63, "elapsed_time": "2:00:37", "remaining_time": "1 day, 17:25:31", "throughput": "17964.75", "total_tokens": 130023424}
{"current_steps": 125, "total_steps": 2679, "loss": 0.7233, "learning_rate": 4.982505531467339e-06, "epoch": 0.13986013986013987, "percentage": 4.67, "elapsed_time": "2:01:26", "remaining_time": "1 day, 17:21:11", "throughput": "17989.14", "total_tokens": 131072000}
{"current_steps": 126, "total_steps": 2679, "loss": 0.865, "learning_rate": 4.982154311591529e-06, "epoch": 0.14097902097902099, "percentage": 4.7, "elapsed_time": "2:02:25", "remaining_time": "1 day, 17:20:43", "throughput": "17985.39", "total_tokens": 132120576}
{"current_steps": 127, "total_steps": 2679, "loss": 0.7565, "learning_rate": 4.981799613739284e-06, "epoch": 0.1420979020979021, "percentage": 4.74, "elapsed_time": "2:03:19", "remaining_time": "1 day, 17:18:09", "throughput": "17997.03", "total_tokens": 133169152}
{"current_steps": 128, "total_steps": 2679, "loss": 0.7785, "learning_rate": 4.981441438407605e-06, "epoch": 0.14321678321678322, "percentage": 4.78, "elapsed_time": "2:04:09", "remaining_time": "1 day, 17:14:21", "throughput": "18017.58", "total_tokens": 134217728}
{"current_steps": 129, "total_steps": 2679, "loss": 0.8922, "learning_rate": 4.981079786098365e-06, "epoch": 0.14433566433566433, "percentage": 4.82, "elapsed_time": "2:05:14", "remaining_time": "1 day, 17:15:39", "throughput": "18001.06", "total_tokens": 135266304}
{"current_steps": 130, "total_steps": 2679, "loss": 0.8254, "learning_rate": 4.980714657318307e-06, "epoch": 0.14545454545454545, "percentage": 4.85, "elapsed_time": "2:06:08", "remaining_time": "1 day, 17:13:20", "throughput": "18010.87", "total_tokens": 136314880}
{"current_steps": 131, "total_steps": 2679, "loss": 0.8365, "learning_rate": 4.980346052579049e-06, "epoch": 0.14657342657342656, "percentage": 4.89, "elapsed_time": "2:07:13", "remaining_time": "1 day, 17:14:37", "throughput": "17994.43", "total_tokens": 137363456}
{"current_steps": 132, "total_steps": 2679, "loss": 0.8213, "learning_rate": 4.979973972397075e-06, "epoch": 0.1476923076923077, "percentage": 4.93, "elapsed_time": "2:08:07", "remaining_time": "1 day, 17:12:15", "throughput": "18004.65", "total_tokens": 138412032}
{"current_steps": 133, "total_steps": 2679, "loss": 0.8138, "learning_rate": 4.979598417293743e-06, "epoch": 0.14881118881118882, "percentage": 4.96, "elapsed_time": "2:09:07", "remaining_time": "1 day, 17:11:51", "throughput": "18000.42", "total_tokens": 139460608}
{"current_steps": 134, "total_steps": 2679, "loss": 0.8871, "learning_rate": 4.9792193877952765e-06, "epoch": 0.14993006993006994, "percentage": 5.0, "elapsed_time": "2:10:08", "remaining_time": "1 day, 17:11:35", "throughput": "17995.37", "total_tokens": 140509184}
{"current_steps": 135, "total_steps": 2679, "loss": 0.7556, "learning_rate": 4.97883688443277e-06, "epoch": 0.15104895104895105, "percentage": 5.04, "elapsed_time": "2:10:58", "remaining_time": "1 day, 17:08:12", "throughput": "18012.90", "total_tokens": 141557760}
{"current_steps": 136, "total_steps": 2679, "loss": 0.831, "learning_rate": 4.9784509077421836e-06, "epoch": 0.15216783216783217, "percentage": 5.08, "elapsed_time": "2:11:58", "remaining_time": "1 day, 17:07:37", "throughput": "18010.14", "total_tokens": 142606336}
{"current_steps": 137, "total_steps": 2679, "loss": 0.8148, "learning_rate": 4.978061458264346e-06, "epoch": 0.15328671328671328, "percentage": 5.11, "elapsed_time": "2:12:57", "remaining_time": "1 day, 17:07:04", "throughput": "18007.04", "total_tokens": 143654912}
{"current_steps": 138, "total_steps": 2679, "loss": 0.8116, "learning_rate": 4.97766853654495e-06, "epoch": 0.1544055944055944, "percentage": 5.15, "elapsed_time": "2:13:54", "remaining_time": "1 day, 17:05:32", "throughput": "18011.16", "total_tokens": 144703488}
{"current_steps": 139, "total_steps": 2679, "loss": 0.8231, "learning_rate": 4.977272143134554e-06, "epoch": 0.15552447552447551, "percentage": 5.19, "elapsed_time": "2:14:54", "remaining_time": "1 day, 17:05:05", "throughput": "18007.31", "total_tokens": 145752064}
{"current_steps": 140, "total_steps": 2679, "loss": 0.9111, "learning_rate": 4.976872278588582e-06, "epoch": 0.15664335664335666, "percentage": 5.23, "elapsed_time": "2:15:59", "remaining_time": "1 day, 17:06:20", "throughput": "17991.11", "total_tokens": 146800640}
{"current_steps": 141, "total_steps": 2679, "loss": 0.8487, "learning_rate": 4.976468943467323e-06, "epoch": 0.15776223776223777, "percentage": 5.26, "elapsed_time": "2:16:59", "remaining_time": "1 day, 17:05:57", "throughput": "17986.81", "total_tokens": 147849216}
{"current_steps": 142, "total_steps": 2679, "loss": 0.76, "learning_rate": 4.976062138335926e-06, "epoch": 0.1588811188811189, "percentage": 5.3, "elapsed_time": "2:17:53", "remaining_time": "1 day, 17:03:34", "throughput": "17997.12", "total_tokens": 148897792}
{"current_steps": 143, "total_steps": 2679, "loss": 0.7741, "learning_rate": 4.975651863764403e-06, "epoch": 0.16, "percentage": 5.34, "elapsed_time": "2:18:47", "remaining_time": "1 day, 17:01:24", "throughput": "18005.91", "total_tokens": 149946368}
{"current_steps": 144, "total_steps": 2679, "loss": 0.8565, "learning_rate": 4.975238120327628e-06, "epoch": 0.16111888111888112, "percentage": 5.38, "elapsed_time": "2:19:52", "remaining_time": "1 day, 17:02:25", "throughput": "17991.31", "total_tokens": 150994944}
{"current_steps": 145, "total_steps": 2679, "loss": 0.8921, "learning_rate": 4.974820908605336e-06, "epoch": 0.16223776223776223, "percentage": 5.41, "elapsed_time": "2:20:52", "remaining_time": "1 day, 17:01:54", "throughput": "17988.00", "total_tokens": 152043520}
{"current_steps": 146, "total_steps": 2679, "loss": 0.9724, "learning_rate": 4.974400229182119e-06, "epoch": 0.16335664335664335, "percentage": 5.45, "elapsed_time": "2:21:58", "remaining_time": "1 day, 17:03:11", "throughput": "17971.54", "total_tokens": 153092096}
{"current_steps": 147, "total_steps": 2679, "loss": 0.7849, "learning_rate": 4.973976082647432e-06, "epoch": 0.16447552447552446, "percentage": 5.49, "elapsed_time": "2:22:53", "remaining_time": "1 day, 17:01:07", "throughput": "17979.58", "total_tokens": 154140672}
{"current_steps": 148, "total_steps": 2679, "loss": 0.774, "learning_rate": 4.973548469595585e-06, "epoch": 0.1655944055944056, "percentage": 5.52, "elapsed_time": "2:23:44", "remaining_time": "1 day, 16:58:05", "throughput": "17994.60", "total_tokens": 155189248}
{"current_steps": 149, "total_steps": 2679, "loss": 0.7858, "learning_rate": 4.973117390625746e-06, "epoch": 0.16671328671328672, "percentage": 5.56, "elapsed_time": "2:24:43", "remaining_time": "1 day, 16:57:28", "throughput": "17992.06", "total_tokens": 156237824}
{"current_steps": 150, "total_steps": 2679, "loss": 0.8224, "learning_rate": 4.972682846341941e-06, "epoch": 0.16783216783216784, "percentage": 5.6, "elapsed_time": "2:25:43", "remaining_time": "1 day, 16:57:01", "throughput": "17988.23", "total_tokens": 157286400}
{"current_steps": 151, "total_steps": 2679, "loss": 0.8042, "learning_rate": 4.97224483735305e-06, "epoch": 0.16895104895104895, "percentage": 5.64, "elapsed_time": "2:26:43", "remaining_time": "1 day, 16:56:28", "throughput": "17985.10", "total_tokens": 158334976}
{"current_steps": 152, "total_steps": 2679, "loss": 0.867, "learning_rate": 4.971803364272806e-06, "epoch": 0.17006993006993007, "percentage": 5.67, "elapsed_time": "2:27:43", "remaining_time": "1 day, 16:55:57", "throughput": "17981.76", "total_tokens": 159383552}
{"current_steps": 153, "total_steps": 2679, "loss": 0.8237, "learning_rate": 4.9713584277198e-06, "epoch": 0.17118881118881119, "percentage": 5.71, "elapsed_time": "2:28:48", "remaining_time": "1 day, 16:56:48", "throughput": "17968.40", "total_tokens": 160432128}
{"current_steps": 154, "total_steps": 2679, "loss": 0.7817, "learning_rate": 4.9709100283174735e-06, "epoch": 0.1723076923076923, "percentage": 5.75, "elapsed_time": "2:29:47", "remaining_time": "1 day, 16:56:05", "throughput": "17966.62", "total_tokens": 161480704}
{"current_steps": 155, "total_steps": 2679, "loss": 0.8944, "learning_rate": 4.97045816669412e-06, "epoch": 0.17342657342657342, "percentage": 5.79, "elapsed_time": "2:30:52", "remaining_time": "1 day, 16:56:56", "throughput": "17953.28", "total_tokens": 162529280}
{"current_steps": 156, "total_steps": 2679, "loss": 0.75, "learning_rate": 4.970002843482885e-06, "epoch": 0.17454545454545456, "percentage": 5.82, "elapsed_time": "2:31:46", "remaining_time": "1 day, 16:54:43", "throughput": "17962.31", "total_tokens": 163577856}
{"current_steps": 157, "total_steps": 2679, "loss": 0.9354, "learning_rate": 4.9695440593217635e-06, "epoch": 0.17566433566433567, "percentage": 5.86, "elapsed_time": "2:32:52", "remaining_time": "1 day, 16:55:46", "throughput": "17947.60", "total_tokens": 164626432}
{"current_steps": 158, "total_steps": 2679, "loss": 0.7858, "learning_rate": 4.969081814853601e-06, "epoch": 0.1767832167832168, "percentage": 5.9, "elapsed_time": "2:33:52", "remaining_time": "1 day, 16:55:09", "throughput": "17944.97", "total_tokens": 165675008}
{"current_steps": 159, "total_steps": 2679, "loss": 0.8166, "learning_rate": 4.9686161107260906e-06, "epoch": 0.1779020979020979, "percentage": 5.94, "elapsed_time": "2:34:51", "remaining_time": "1 day, 16:54:21", "throughput": "17943.69", "total_tokens": 166723584}
{"current_steps": 160, "total_steps": 2679, "loss": 0.7839, "learning_rate": 4.9681469475917746e-06, "epoch": 0.17902097902097902, "percentage": 5.97, "elapsed_time": "2:35:52", "remaining_time": "1 day, 16:54:09", "throughput": "17938.05", "total_tokens": 167772160}
{"current_steps": 161, "total_steps": 2679, "loss": 0.694, "learning_rate": 4.967674326108039e-06, "epoch": 0.18013986013986014, "percentage": 6.01, "elapsed_time": "2:36:40", "remaining_time": "1 day, 16:50:22", "throughput": "17958.56", "total_tokens": 168820736}
{"current_steps": 162, "total_steps": 2679, "loss": 0.7829, "learning_rate": 4.967198246937119e-06, "epoch": 0.18125874125874125, "percentage": 6.05, "elapsed_time": "2:37:40", "remaining_time": "1 day, 16:49:46", "throughput": "17955.90", "total_tokens": 169869312}
{"current_steps": 163, "total_steps": 2679, "loss": 0.9115, "learning_rate": 4.9667187107460934e-06, "epoch": 0.18237762237762237, "percentage": 6.08, "elapsed_time": "2:38:41", "remaining_time": "1 day, 16:49:23", "throughput": "17951.50", "total_tokens": 170917888}
{"current_steps": 164, "total_steps": 2679, "loss": 0.7645, "learning_rate": 4.966235718206885e-06, "epoch": 0.1834965034965035, "percentage": 6.12, "elapsed_time": "2:39:35", "remaining_time": "1 day, 16:47:19", "throughput": "17959.55", "total_tokens": 171966464}
{"current_steps": 165, "total_steps": 2679, "loss": 0.8021, "learning_rate": 4.965749269996258e-06, "epoch": 0.18461538461538463, "percentage": 6.16, "elapsed_time": "2:40:34", "remaining_time": "1 day, 16:46:29", "throughput": "17958.45", "total_tokens": 173015040}
{"current_steps": 166, "total_steps": 2679, "loss": 0.9263, "learning_rate": 4.965259366795821e-06, "epoch": 0.18573426573426574, "percentage": 6.2, "elapsed_time": "2:41:39", "remaining_time": "1 day, 16:47:18", "throughput": "17945.42", "total_tokens": 174063616}
{"current_steps": 167, "total_steps": 2679, "loss": 0.7471, "learning_rate": 4.964766009292022e-06, "epoch": 0.18685314685314686, "percentage": 6.23, "elapsed_time": "2:42:33", "remaining_time": "1 day, 16:45:07", "throughput": "17954.29", "total_tokens": 175112192}
{"current_steps": 168, "total_steps": 2679, "loss": 0.7309, "learning_rate": 4.964269198176152e-06, "epoch": 0.18797202797202797, "percentage": 6.27, "elapsed_time": "2:43:27", "remaining_time": "1 day, 16:43:01", "throughput": "17962.51", "total_tokens": 176160768}
{"current_steps": 169, "total_steps": 2679, "loss": 0.7829, "learning_rate": 4.963768934144336e-06, "epoch": 0.1890909090909091, "percentage": 6.31, "elapsed_time": "2:44:21", "remaining_time": "1 day, 16:41:08", "throughput": "17969.21", "total_tokens": 177209344}
{"current_steps": 170, "total_steps": 2679, "loss": 0.7963, "learning_rate": 4.963265217897543e-06, "epoch": 0.1902097902097902, "percentage": 6.35, "elapsed_time": "2:45:22", "remaining_time": "1 day, 16:40:37", "throughput": "17965.82", "total_tokens": 178257920}
{"current_steps": 171, "total_steps": 2679, "loss": 0.7968, "learning_rate": 4.962758050141576e-06, "epoch": 0.19132867132867132, "percentage": 6.38, "elapsed_time": "2:46:21", "remaining_time": "1 day, 16:39:54", "throughput": "17963.95", "total_tokens": 179306496}
{"current_steps": 172, "total_steps": 2679, "loss": 0.7048, "learning_rate": 4.962247431587073e-06, "epoch": 0.19244755244755246, "percentage": 6.42, "elapsed_time": "2:47:11", "remaining_time": "1 day, 16:36:59", "throughput": "17978.29", "total_tokens": 180355072}
{"current_steps": 173, "total_steps": 2679, "loss": 0.8132, "learning_rate": 4.96173336294951e-06, "epoch": 0.19356643356643358, "percentage": 6.46, "elapsed_time": "2:48:05", "remaining_time": "1 day, 16:35:00", "throughput": "17985.74", "total_tokens": 181403648}
{"current_steps": 174, "total_steps": 2679, "loss": 0.7951, "learning_rate": 4.961215844949197e-06, "epoch": 0.1946853146853147, "percentage": 6.49, "elapsed_time": "2:49:05", "remaining_time": "1 day, 16:34:23", "throughput": "17983.16", "total_tokens": 182452224}
{"current_steps": 175, "total_steps": 2679, "loss": 0.865, "learning_rate": 4.960694878311276e-06, "epoch": 0.1958041958041958, "percentage": 6.53, "elapsed_time": "2:50:05", "remaining_time": "1 day, 16:33:47", "throughput": "17980.46", "total_tokens": 183500800}
{"current_steps": 176, "total_steps": 2679, "loss": 0.7015, "learning_rate": 4.9601704637657225e-06, "epoch": 0.19692307692307692, "percentage": 6.57, "elapsed_time": "2:50:56", "remaining_time": "1 day, 16:31:06", "throughput": "17993.12", "total_tokens": 184549376}
{"current_steps": 177, "total_steps": 2679, "loss": 0.7767, "learning_rate": 4.959642602047339e-06, "epoch": 0.19804195804195804, "percentage": 6.61, "elapsed_time": "2:51:55", "remaining_time": "1 day, 16:30:20", "throughput": "17991.50", "total_tokens": 185597952}
{"current_steps": 178, "total_steps": 2679, "loss": 0.8576, "learning_rate": 4.959111293895765e-06, "epoch": 0.19916083916083915, "percentage": 6.64, "elapsed_time": "2:53:01", "remaining_time": "1 day, 16:31:00", "throughput": "17979.43", "total_tokens": 186646528}
{"current_steps": 179, "total_steps": 2679, "loss": 0.8765, "learning_rate": 4.958576540055464e-06, "epoch": 0.20027972027972027, "percentage": 6.68, "elapsed_time": "2:54:11", "remaining_time": "1 day, 16:32:57", "throughput": "17957.85", "total_tokens": 187695104}
{"current_steps": 180, "total_steps": 2679, "loss": 0.9279, "learning_rate": 4.95803834127573e-06, "epoch": 0.2013986013986014, "percentage": 6.72, "elapsed_time": "2:55:17", "remaining_time": "1 day, 16:33:38", "throughput": "17945.64", "total_tokens": 188743680}
{"current_steps": 181, "total_steps": 2679, "loss": 0.7802, "learning_rate": 4.9574966983106824e-06, "epoch": 0.20251748251748253, "percentage": 6.76, "elapsed_time": "2:56:16", "remaining_time": "1 day, 16:32:46", "throughput": "17944.76", "total_tokens": 189792256}
{"current_steps": 182, "total_steps": 2679, "loss": 0.717, "learning_rate": 4.956951611919267e-06, "epoch": 0.20363636363636364, "percentage": 6.79, "elapsed_time": "2:57:10", "remaining_time": "1 day, 16:30:45", "throughput": "17952.51", "total_tokens": 190840832}
{"current_steps": 183, "total_steps": 2679, "loss": 0.7341, "learning_rate": 4.9564030828652565e-06, "epoch": 0.20475524475524476, "percentage": 6.83, "elapsed_time": "2:58:04", "remaining_time": "1 day, 16:28:46", "throughput": "17959.99", "total_tokens": 191889408}
{"current_steps": 184, "total_steps": 2679, "loss": 0.6868, "learning_rate": 4.955851111917245e-06, "epoch": 0.20587412587412587, "percentage": 6.87, "elapsed_time": "2:58:52", "remaining_time": "1 day, 16:25:25", "throughput": "17977.61", "total_tokens": 192937984}
{"current_steps": 185, "total_steps": 2679, "loss": 0.8103, "learning_rate": 4.955295699848649e-06, "epoch": 0.206993006993007, "percentage": 6.91, "elapsed_time": "2:59:52", "remaining_time": "1 day, 16:24:48", "throughput": "17975.00", "total_tokens": 193986560}
{"current_steps": 186, "total_steps": 2679, "loss": 0.7368, "learning_rate": 4.954736847437709e-06, "epoch": 0.2081118881118881, "percentage": 6.94, "elapsed_time": "3:00:40", "remaining_time": "1 day, 16:21:34", "throughput": "17991.68", "total_tokens": 195035136}
{"current_steps": 187, "total_steps": 2679, "loss": 0.9349, "learning_rate": 4.954174555467484e-06, "epoch": 0.20923076923076922, "percentage": 6.98, "elapsed_time": "3:01:40", "remaining_time": "1 day, 16:21:04", "throughput": "17988.20", "total_tokens": 196083712}
{"current_steps": 188, "total_steps": 2679, "loss": 0.8829, "learning_rate": 4.953608824725855e-06, "epoch": 0.21034965034965036, "percentage": 7.02, "elapsed_time": "3:02:36", "remaining_time": "1 day, 16:19:32", "throughput": "17992.42", "total_tokens": 197132288}
{"current_steps": 189, "total_steps": 2679, "loss": 1.0609, "learning_rate": 4.953039656005519e-06, "epoch": 0.21146853146853148, "percentage": 7.05, "elapsed_time": "3:03:45", "remaining_time": "1 day, 16:20:57", "throughput": "17974.60", "total_tokens": 198180864}
{"current_steps": 190, "total_steps": 2679, "loss": 0.7973, "learning_rate": 4.95246705010399e-06, "epoch": 0.2125874125874126, "percentage": 7.09, "elapsed_time": "3:04:42", "remaining_time": "1 day, 16:19:37", "throughput": "17977.38", "total_tokens": 199229440}
{"current_steps": 191, "total_steps": 2679, "loss": 0.7003, "learning_rate": 4.951891007823601e-06, "epoch": 0.2137062937062937, "percentage": 7.13, "elapsed_time": "3:05:30", "remaining_time": "1 day, 16:16:25", "throughput": "17993.89", "total_tokens": 200278016}
{"current_steps": 192, "total_steps": 2679, "loss": 0.8175, "learning_rate": 4.951311529971496e-06, "epoch": 0.21482517482517482, "percentage": 7.17, "elapsed_time": "3:06:24", "remaining_time": "1 day, 16:14:39", "throughput": "17999.80", "total_tokens": 201326592}
{"current_steps": 193, "total_steps": 2679, "loss": 0.7979, "learning_rate": 4.950728617359637e-06, "epoch": 0.21594405594405594, "percentage": 7.2, "elapsed_time": "3:07:24", "remaining_time": "1 day, 16:13:52", "throughput": "17998.46", "total_tokens": 202375168}
{"current_steps": 194, "total_steps": 2679, "loss": 0.7616, "learning_rate": 4.950142270804797e-06, "epoch": 0.21706293706293706, "percentage": 7.24, "elapsed_time": "3:08:17", "remaining_time": "1 day, 16:11:50", "throughput": "18006.38", "total_tokens": 203423744}
{"current_steps": 195, "total_steps": 2679, "loss": 0.6991, "learning_rate": 4.949552491128559e-06, "epoch": 0.21818181818181817, "percentage": 7.28, "elapsed_time": "3:09:05", "remaining_time": "1 day, 16:08:41", "throughput": "18022.61", "total_tokens": 204472320}
{"current_steps": 196, "total_steps": 2679, "loss": 0.8588, "learning_rate": 4.948959279157319e-06, "epoch": 0.21930069930069931, "percentage": 7.32, "elapsed_time": "3:10:10", "remaining_time": "1 day, 16:09:15", "throughput": "18011.12", "total_tokens": 205520896}
{"current_steps": 197, "total_steps": 2679, "loss": 0.7721, "learning_rate": 4.948362635722281e-06, "epoch": 0.22041958041958043, "percentage": 7.35, "elapsed_time": "3:11:10", "remaining_time": "1 day, 16:08:30", "throughput": "18009.47", "total_tokens": 206569472}
{"current_steps": 198, "total_steps": 2679, "loss": 0.8775, "learning_rate": 4.947762561659457e-06, "epoch": 0.22153846153846155, "percentage": 7.39, "elapsed_time": "3:12:10", "remaining_time": "1 day, 16:07:56", "throughput": "18006.47", "total_tokens": 207618048}
{"current_steps": 199, "total_steps": 2679, "loss": 0.9032, "learning_rate": 4.947159057809668e-06, "epoch": 0.22265734265734266, "percentage": 7.43, "elapsed_time": "3:13:15", "remaining_time": "1 day, 16:08:30", "throughput": "17994.95", "total_tokens": 208666624}
{"current_steps": 200, "total_steps": 2679, "loss": 0.8398, "learning_rate": 4.9465521250185365e-06, "epoch": 0.22377622377622378, "percentage": 7.47, "elapsed_time": "3:14:20", "remaining_time": "1 day, 16:08:57", "throughput": "17984.43", "total_tokens": 209715200}
{"current_steps": 201, "total_steps": 2679, "loss": 0.7901, "learning_rate": 4.945941764136494e-06, "epoch": 0.2248951048951049, "percentage": 7.5, "elapsed_time": "3:15:20", "remaining_time": "1 day, 16:08:13", "throughput": "17982.66", "total_tokens": 210763776}
{"current_steps": 202, "total_steps": 2679, "loss": 0.9251, "learning_rate": 4.945327976018774e-06, "epoch": 0.226013986013986, "percentage": 7.54, "elapsed_time": "3:16:26", "remaining_time": "1 day, 16:08:49", "throughput": "17970.83", "total_tokens": 211812352}
{"current_steps": 203, "total_steps": 2679, "loss": 0.7021, "learning_rate": 4.944710761525411e-06, "epoch": 0.22713286713286712, "percentage": 7.58, "elapsed_time": "3:17:14", "remaining_time": "1 day, 16:05:47", "throughput": "17986.23", "total_tokens": 212860928}
{"current_steps": 204, "total_steps": 2679, "loss": 0.767, "learning_rate": 4.944090121521242e-06, "epoch": 0.22825174825174827, "percentage": 7.61, "elapsed_time": "3:18:14", "remaining_time": "1 day, 16:05:07", "throughput": "17984.06", "total_tokens": 213909504}
{"current_steps": 205, "total_steps": 2679, "loss": 0.7803, "learning_rate": 4.943466056875903e-06, "epoch": 0.22937062937062938, "percentage": 7.65, "elapsed_time": "3:19:08", "remaining_time": "1 day, 16:03:14", "throughput": "17990.76", "total_tokens": 214958080}
{"current_steps": 206, "total_steps": 2679, "loss": 0.6921, "learning_rate": 4.942838568463829e-06, "epoch": 0.2304895104895105, "percentage": 7.69, "elapsed_time": "3:19:56", "remaining_time": "1 day, 16:00:20", "throughput": "18005.31", "total_tokens": 216006656}
{"current_steps": 207, "total_steps": 2679, "loss": 0.8301, "learning_rate": 4.9422076571642516e-06, "epoch": 0.2316083916083916, "percentage": 7.73, "elapsed_time": "3:20:55", "remaining_time": "1 day, 15:59:23", "throughput": "18005.18", "total_tokens": 217055232}
{"current_steps": 208, "total_steps": 2679, "loss": 0.7935, "learning_rate": 4.9415733238612e-06, "epoch": 0.23272727272727273, "percentage": 7.76, "elapsed_time": "3:21:57", "remaining_time": "1 day, 15:59:13", "throughput": "17999.11", "total_tokens": 218103808}
{"current_steps": 209, "total_steps": 2679, "loss": 0.7251, "learning_rate": 4.940935569443496e-06, "epoch": 0.23384615384615384, "percentage": 7.8, "elapsed_time": "3:22:51", "remaining_time": "1 day, 15:57:20", "throughput": "18005.88", "total_tokens": 219152384}
{"current_steps": 210, "total_steps": 2679, "loss": 0.8056, "learning_rate": 4.940294394804757e-06, "epoch": 0.23496503496503496, "percentage": 7.84, "elapsed_time": "3:23:50", "remaining_time": "1 day, 15:56:41", "throughput": "18003.56", "total_tokens": 220200960}
{"current_steps": 211, "total_steps": 2679, "loss": 0.709, "learning_rate": 4.939649800843394e-06, "epoch": 0.23608391608391607, "percentage": 7.88, "elapsed_time": "3:24:42", "remaining_time": "1 day, 15:54:25", "throughput": "18013.31", "total_tokens": 221249536}
{"current_steps": 212, "total_steps": 2679, "loss": 0.8409, "learning_rate": 4.939001788462604e-06, "epoch": 0.23720279720279722, "percentage": 7.91, "elapsed_time": "3:25:47", "remaining_time": "1 day, 15:54:47", "throughput": "18003.19", "total_tokens": 222298112}
{"current_steps": 213, "total_steps": 2679, "loss": 0.8815, "learning_rate": 4.93835035857038e-06, "epoch": 0.23832167832167833, "percentage": 7.95, "elapsed_time": "3:26:53", "remaining_time": "1 day, 15:55:12", "throughput": "17992.85", "total_tokens": 223346688}
{"current_steps": 214, "total_steps": 2679, "loss": 0.8565, "learning_rate": 4.9376955120795e-06, "epoch": 0.23944055944055945, "percentage": 7.99, "elapsed_time": "3:27:53", "remaining_time": "1 day, 15:54:32", "throughput": "17990.45", "total_tokens": 224395264}
{"current_steps": 215, "total_steps": 2679, "loss": 0.8288, "learning_rate": 4.937037249907529e-06, "epoch": 0.24055944055944056, "percentage": 8.03, "elapsed_time": "3:28:52", "remaining_time": "1 day, 15:53:50", "throughput": "17988.46", "total_tokens": 225443840}
{"current_steps": 216, "total_steps": 2679, "loss": 0.8191, "learning_rate": 4.936375572976822e-06, "epoch": 0.24167832167832168, "percentage": 8.06, "elapsed_time": "3:29:52", "remaining_time": "1 day, 15:53:12", "throughput": "17985.94", "total_tokens": 226492416}
{"current_steps": 217, "total_steps": 2679, "loss": 0.7696, "learning_rate": 4.935710482214512e-06, "epoch": 0.2427972027972028, "percentage": 8.1, "elapsed_time": "3:30:52", "remaining_time": "1 day, 15:52:26", "throughput": "17984.32", "total_tokens": 227540992}
{"current_steps": 218, "total_steps": 2679, "loss": 0.8006, "learning_rate": 4.935041978552522e-06, "epoch": 0.2439160839160839, "percentage": 8.14, "elapsed_time": "3:31:49", "remaining_time": "1 day, 15:51:19", "throughput": "17985.46", "total_tokens": 228589568}
{"current_steps": 219, "total_steps": 2679, "loss": 0.8424, "learning_rate": 4.9343700629275525e-06, "epoch": 0.24503496503496502, "percentage": 8.17, "elapsed_time": "3:32:55", "remaining_time": "1 day, 15:51:46", "throughput": "17974.71", "total_tokens": 229638144}
{"current_steps": 220, "total_steps": 2679, "loss": 0.8801, "learning_rate": 4.933694736281089e-06, "epoch": 0.24615384615384617, "percentage": 8.21, "elapsed_time": "3:34:01", "remaining_time": "1 day, 15:52:10", "throughput": "17964.52", "total_tokens": 230686720}
{"current_steps": 221, "total_steps": 2679, "loss": 0.8984, "learning_rate": 4.9330159995593926e-06, "epoch": 0.24727272727272728, "percentage": 8.25, "elapsed_time": "3:35:01", "remaining_time": "1 day, 15:51:28", "throughput": "17962.36", "total_tokens": 231735296}
{"current_steps": 222, "total_steps": 2679, "loss": 0.9508, "learning_rate": 4.932333853713505e-06, "epoch": 0.2483916083916084, "percentage": 8.29, "elapsed_time": "3:36:06", "remaining_time": "1 day, 15:51:52", "throughput": "17952.11", "total_tokens": 232783872}
{"current_steps": 223, "total_steps": 2679, "loss": 0.7423, "learning_rate": 4.931648299699245e-06, "epoch": 0.2495104895104895, "percentage": 8.32, "elapsed_time": "3:37:00", "remaining_time": "1 day, 15:49:57", "throughput": "17959.19", "total_tokens": 233832448}
{"current_steps": 224, "total_steps": 2679, "loss": 0.7948, "learning_rate": 4.930959338477203e-06, "epoch": 0.25062937062937063, "percentage": 8.36, "elapsed_time": "3:37:55", "remaining_time": "1 day, 15:48:23", "throughput": "17963.59", "total_tokens": 234881024}
{"current_steps": 225, "total_steps": 2679, "loss": 0.7943, "learning_rate": 4.930266971012748e-06, "epoch": 0.2517482517482518, "percentage": 8.4, "elapsed_time": "3:38:55", "remaining_time": "1 day, 15:47:41", "throughput": "17961.63", "total_tokens": 235929600}
{"current_steps": 226, "total_steps": 2679, "loss": 0.7972, "learning_rate": 4.92957119827602e-06, "epoch": 0.25286713286713286, "percentage": 8.44, "elapsed_time": "3:39:54", "remaining_time": "1 day, 15:46:54", "throughput": "17960.20", "total_tokens": 236978176}
{"current_steps": 227, "total_steps": 2679, "loss": 0.7005, "learning_rate": 4.928872021241932e-06, "epoch": 0.253986013986014, "percentage": 8.47, "elapsed_time": "3:40:43", "remaining_time": "1 day, 15:44:10", "throughput": "17973.42", "total_tokens": 238026752}
{"current_steps": 228, "total_steps": 2679, "loss": 0.7865, "learning_rate": 4.928169440890164e-06, "epoch": 0.2551048951048951, "percentage": 8.51, "elapsed_time": "3:41:43", "remaining_time": "1 day, 15:43:29", "throughput": "17971.20", "total_tokens": 239075328}
{"current_steps": 229, "total_steps": 2679, "loss": 0.8001, "learning_rate": 4.927463458205167e-06, "epoch": 0.25622377622377623, "percentage": 8.55, "elapsed_time": "3:42:43", "remaining_time": "1 day, 15:42:53", "throughput": "17968.41", "total_tokens": 240123904}
{"current_steps": 230, "total_steps": 2679, "loss": 0.7972, "learning_rate": 4.926754074176159e-06, "epoch": 0.2573426573426573, "percentage": 8.59, "elapsed_time": "3:43:48", "remaining_time": "1 day, 15:43:03", "throughput": "17959.91", "total_tokens": 241172480}
{"current_steps": 231, "total_steps": 2679, "loss": 0.756, "learning_rate": 4.9260412897971225e-06, "epoch": 0.25846153846153846, "percentage": 8.62, "elapsed_time": "3:44:43", "remaining_time": "1 day, 15:41:25", "throughput": "17964.79", "total_tokens": 242221056}
{"current_steps": 232, "total_steps": 2679, "loss": 0.7881, "learning_rate": 4.925325106066808e-06, "epoch": 0.2595804195804196, "percentage": 8.66, "elapsed_time": "3:45:37", "remaining_time": "1 day, 15:39:44", "throughput": "17970.15", "total_tokens": 243269632}
{"current_steps": 233, "total_steps": 2679, "loss": 0.7635, "learning_rate": 4.9246055239887255e-06, "epoch": 0.2606993006993007, "percentage": 8.7, "elapsed_time": "3:46:32", "remaining_time": "1 day, 15:38:14", "throughput": "17974.18", "total_tokens": 244318208}
{"current_steps": 234, "total_steps": 2679, "loss": 0.8651, "learning_rate": 4.923882544571148e-06, "epoch": 0.26181818181818184, "percentage": 8.73, "elapsed_time": "3:47:32", "remaining_time": "1 day, 15:37:30", "throughput": "17972.36", "total_tokens": 245366784}
{"current_steps": 235, "total_steps": 2679, "loss": 0.8524, "learning_rate": 4.923156168827109e-06, "epoch": 0.2629370629370629, "percentage": 8.77, "elapsed_time": "3:48:32", "remaining_time": "1 day, 15:36:50", "throughput": "17970.01", "total_tokens": 246415360}
{"current_steps": 236, "total_steps": 2679, "loss": 0.9272, "learning_rate": 4.922426397774402e-06, "epoch": 0.26405594405594407, "percentage": 8.81, "elapsed_time": "3:49:33", "remaining_time": "1 day, 15:36:17", "throughput": "17966.90", "total_tokens": 247463936}
{"current_steps": 237, "total_steps": 2679, "loss": 0.8336, "learning_rate": 4.9216932324355755e-06, "epoch": 0.26517482517482516, "percentage": 8.85, "elapsed_time": "3:50:33", "remaining_time": "1 day, 15:35:36", "throughput": "17964.74", "total_tokens": 248512512}
{"current_steps": 238, "total_steps": 2679, "loss": 0.8021, "learning_rate": 4.920956673837936e-06, "epoch": 0.2662937062937063, "percentage": 8.88, "elapsed_time": "3:51:28", "remaining_time": "1 day, 15:34:08", "throughput": "17968.45", "total_tokens": 249561088}
{"current_steps": 239, "total_steps": 2679, "loss": 0.7572, "learning_rate": 4.920216723013544e-06, "epoch": 0.2674125874125874, "percentage": 8.92, "elapsed_time": "3:52:22", "remaining_time": "1 day, 15:32:23", "throughput": "17974.27", "total_tokens": 250609664}
{"current_steps": 240, "total_steps": 2679, "loss": 0.773, "learning_rate": 4.919473380999212e-06, "epoch": 0.26853146853146853, "percentage": 8.96, "elapsed_time": "3:53:22", "remaining_time": "1 day, 15:31:35", "throughput": "17972.96", "total_tokens": 251658240}
{"current_steps": 241, "total_steps": 2679, "loss": 0.8538, "learning_rate": 4.918726648836507e-06, "epoch": 0.2696503496503497, "percentage": 9.0, "elapsed_time": "3:54:21", "remaining_time": "1 day, 15:30:52", "throughput": "17971.12", "total_tokens": 252706816}
{"current_steps": 242, "total_steps": 2679, "loss": 0.7968, "learning_rate": 4.917976527571745e-06, "epoch": 0.27076923076923076, "percentage": 9.03, "elapsed_time": "3:55:21", "remaining_time": "1 day, 15:30:07", "throughput": "17969.39", "total_tokens": 253755392}
{"current_steps": 243, "total_steps": 2679, "loss": 0.6936, "learning_rate": 4.917223018255989e-06, "epoch": 0.2718881118881119, "percentage": 9.07, "elapsed_time": "3:56:09", "remaining_time": "1 day, 15:27:25", "throughput": "17982.48", "total_tokens": 254803968}
{"current_steps": 244, "total_steps": 2679, "loss": 0.8573, "learning_rate": 4.9164661219450504e-06, "epoch": 0.273006993006993, "percentage": 9.11, "elapsed_time": "3:57:15", "remaining_time": "1 day, 15:27:40", "throughput": "17973.26", "total_tokens": 255852544}
{"current_steps": 245, "total_steps": 2679, "loss": 0.7018, "learning_rate": 4.915705839699488e-06, "epoch": 0.27412587412587414, "percentage": 9.15, "elapsed_time": "3:58:03", "remaining_time": "1 day, 15:24:57", "throughput": "17986.43", "total_tokens": 256901120}
{"current_steps": 246, "total_steps": 2679, "loss": 0.9167, "learning_rate": 4.914942172584605e-06, "epoch": 0.2752447552447552, "percentage": 9.18, "elapsed_time": "3:58:57", "remaining_time": "1 day, 15:23:20", "throughput": "17991.37", "total_tokens": 257949696}
{"current_steps": 247, "total_steps": 2679, "loss": 0.7525, "learning_rate": 4.914175121670443e-06, "epoch": 0.27636363636363637, "percentage": 9.22, "elapsed_time": "3:59:52", "remaining_time": "1 day, 15:21:48", "throughput": "17995.61", "total_tokens": 258998272}
{"current_steps": 248, "total_steps": 2679, "loss": 0.859, "learning_rate": 4.9134046880317895e-06, "epoch": 0.2774825174825175, "percentage": 9.26, "elapsed_time": "4:00:57", "remaining_time": "1 day, 15:22:03", "throughput": "17986.38", "total_tokens": 260046848}
{"current_steps": 249, "total_steps": 2679, "loss": 0.7317, "learning_rate": 4.912630872748171e-06, "epoch": 0.2786013986013986, "percentage": 9.29, "elapsed_time": "4:01:51", "remaining_time": "1 day, 15:20:19", "throughput": "17992.15", "total_tokens": 261095424}
{"current_steps": 250, "total_steps": 2679, "loss": 0.7855, "learning_rate": 4.911853676903851e-06, "epoch": 0.27972027972027974, "percentage": 9.33, "elapsed_time": "4:02:51", "remaining_time": "1 day, 15:19:38", "throughput": "17990.00", "total_tokens": 262144000}
{"current_steps": 250, "total_steps": 2679, "eval_loss": 0.7918664813041687, "epoch": 0.27972027972027974, "percentage": 9.33, "elapsed_time": "4:07:00", "remaining_time": "1 day, 15:59:51", "throughput": "17688.50", "total_tokens": 262144000}
{"current_steps": 251, "total_steps": 2679, "loss": 0.7646, "learning_rate": 4.911073101587831e-06, "epoch": 0.28083916083916083, "percentage": 9.37, "elapsed_time": "4:08:21", "remaining_time": "1 day, 16:02:31", "throughput": "17661.60", "total_tokens": 263192576}
{"current_steps": 252, "total_steps": 2679, "loss": 0.7942, "learning_rate": 4.9102891478938475e-06, "epoch": 0.28195804195804197, "percentage": 9.41, "elapsed_time": "4:09:21", "remaining_time": "1 day, 16:01:34", "throughput": "17661.24", "total_tokens": 264241152}
{"current_steps": 253, "total_steps": 2679, "loss": 0.7701, "learning_rate": 4.90950181692037e-06, "epoch": 0.28307692307692306, "percentage": 9.44, "elapsed_time": "4:10:21", "remaining_time": "1 day, 16:00:37", "throughput": "17660.98", "total_tokens": 265289728}
{"current_steps": 254, "total_steps": 2679, "loss": 0.8457, "learning_rate": 4.908711109770602e-06, "epoch": 0.2841958041958042, "percentage": 9.48, "elapsed_time": "4:11:21", "remaining_time": "1 day, 15:59:42", "throughput": "17660.48", "total_tokens": 266338304}
{"current_steps": 255, "total_steps": 2679, "loss": 0.7976, "learning_rate": 4.9079170275524765e-06, "epoch": 0.2853146853146853, "percentage": 9.52, "elapsed_time": "4:12:15", "remaining_time": "1 day, 15:57:54", "throughput": "17666.47", "total_tokens": 267386880}
{"current_steps": 256, "total_steps": 2679, "loss": 0.8408, "learning_rate": 4.907119571378655e-06, "epoch": 0.28643356643356643, "percentage": 9.56, "elapsed_time": "4:13:20", "remaining_time": "1 day, 15:57:53", "throughput": "17659.28", "total_tokens": 268435456}
{"current_steps": 257, "total_steps": 2679, "loss": 0.8099, "learning_rate": 4.906318742366527e-06, "epoch": 0.2875524475524476, "percentage": 9.59, "elapsed_time": "4:14:19", "remaining_time": "1 day, 15:56:46", "throughput": "17660.26", "total_tokens": 269484032}
{"current_steps": 258, "total_steps": 2679, "loss": 0.8456, "learning_rate": 4.90551454163821e-06, "epoch": 0.28867132867132866, "percentage": 9.63, "elapsed_time": "4:15:18", "remaining_time": "1 day, 15:55:47", "throughput": "17660.10", "total_tokens": 270532608}
{"current_steps": 259, "total_steps": 2679, "loss": 0.7634, "learning_rate": 4.904706970320542e-06, "epoch": 0.2897902097902098, "percentage": 9.67, "elapsed_time": "4:16:18", "remaining_time": "1 day, 15:54:46", "throughput": "17660.32", "total_tokens": 271581184}
{"current_steps": 260, "total_steps": 2679, "loss": 0.8899, "learning_rate": 4.9038960295450865e-06, "epoch": 0.2909090909090909, "percentage": 9.71, "elapsed_time": "4:17:17", "remaining_time": "1 day, 15:53:52", "throughput": "17659.69", "total_tokens": 272629760}
{"current_steps": 261, "total_steps": 2679, "loss": 0.7128, "learning_rate": 4.903081720448128e-06, "epoch": 0.29202797202797204, "percentage": 9.74, "elapsed_time": "4:18:05", "remaining_time": "1 day, 15:51:06", "throughput": "17672.81", "total_tokens": 273678336}
{"current_steps": 262, "total_steps": 2679, "loss": 0.7855, "learning_rate": 4.902264044170671e-06, "epoch": 0.2931468531468531, "percentage": 9.78, "elapsed_time": "4:19:00", "remaining_time": "1 day, 15:49:22", "throughput": "17678.30", "total_tokens": 274726912}
{"current_steps": 263, "total_steps": 2679, "loss": 0.788, "learning_rate": 4.901443001858438e-06, "epoch": 0.29426573426573427, "percentage": 9.82, "elapsed_time": "4:19:57", "remaining_time": "1 day, 15:47:59", "throughput": "17681.19", "total_tokens": 275775488}
{"current_steps": 264, "total_steps": 2679, "loss": 0.7231, "learning_rate": 4.900618594661865e-06, "epoch": 0.2953846153846154, "percentage": 9.85, "elapsed_time": "4:20:51", "remaining_time": "1 day, 15:46:18", "throughput": "17686.40", "total_tokens": 276824064}
{"current_steps": 265, "total_steps": 2679, "loss": 0.7654, "learning_rate": 4.899790823736108e-06, "epoch": 0.2965034965034965, "percentage": 9.89, "elapsed_time": "4:21:45", "remaining_time": "1 day, 15:44:26", "throughput": "17692.97", "total_tokens": 277872640}
{"current_steps": 266, "total_steps": 2679, "loss": 0.8424, "learning_rate": 4.898959690241033e-06, "epoch": 0.29762237762237764, "percentage": 9.93, "elapsed_time": "4:22:44", "remaining_time": "1 day, 15:43:25", "throughput": "17693.17", "total_tokens": 278921216}
{"current_steps": 267, "total_steps": 2679, "loss": 0.9365, "learning_rate": 4.898125195341217e-06, "epoch": 0.29874125874125873, "percentage": 9.97, "elapsed_time": "4:23:50", "remaining_time": "1 day, 15:43:26", "throughput": "17685.62", "total_tokens": 279969792}
{"current_steps": 268, "total_steps": 2679, "loss": 0.7048, "learning_rate": 4.897287340205948e-06, "epoch": 0.2998601398601399, "percentage": 10.0, "elapsed_time": "4:24:39", "remaining_time": "1 day, 15:40:54", "throughput": "17697.13", "total_tokens": 281018368}
{"current_steps": 269, "total_steps": 2679, "loss": 0.7451, "learning_rate": 4.896446126009224e-06, "epoch": 0.30097902097902096, "percentage": 10.04, "elapsed_time": "4:25:32", "remaining_time": "1 day, 15:38:59", "throughput": "17704.03", "total_tokens": 282066944}
{"current_steps": 270, "total_steps": 2679, "loss": 0.8328, "learning_rate": 4.895601553929748e-06, "epoch": 0.3020979020979021, "percentage": 10.08, "elapsed_time": "4:26:32", "remaining_time": "1 day, 15:38:08", "throughput": "17703.05", "total_tokens": 283115520}
{"current_steps": 271, "total_steps": 2679, "loss": 0.8177, "learning_rate": 4.894753625150927e-06, "epoch": 0.3032167832167832, "percentage": 10.12, "elapsed_time": "4:27:32", "remaining_time": "1 day, 15:37:15", "throughput": "17702.22", "total_tokens": 284164096}
{"current_steps": 272, "total_steps": 2679, "loss": 0.7319, "learning_rate": 4.893902340860872e-06, "epoch": 0.30433566433566434, "percentage": 10.15, "elapsed_time": "4:28:26", "remaining_time": "1 day, 15:35:27", "throughput": "17708.27", "total_tokens": 285212672}
{"current_steps": 273, "total_steps": 2679, "loss": 0.8058, "learning_rate": 4.893047702252399e-06, "epoch": 0.3054545454545455, "percentage": 10.19, "elapsed_time": "4:29:26", "remaining_time": "1 day, 15:34:37", "throughput": "17707.17", "total_tokens": 286261248}
{"current_steps": 274, "total_steps": 2679, "loss": 0.657, "learning_rate": 4.89218971052302e-06, "epoch": 0.30657342657342657, "percentage": 10.23, "elapsed_time": "4:30:14", "remaining_time": "1 day, 15:32:01", "throughput": "17719.27", "total_tokens": 287309824}
{"current_steps": 275, "total_steps": 2679, "loss": 0.68, "learning_rate": 4.891328366874946e-06, "epoch": 0.3076923076923077, "percentage": 10.27, "elapsed_time": "4:31:02", "remaining_time": "1 day, 15:29:27", "throughput": "17731.02", "total_tokens": 288358400}
{"current_steps": 276, "total_steps": 2679, "loss": 0.6954, "learning_rate": 4.890463672515086e-06, "epoch": 0.3088111888111888, "percentage": 10.3, "elapsed_time": "4:31:51", "remaining_time": "1 day, 15:26:54", "throughput": "17742.73", "total_tokens": 289406976}
{"current_steps": 277, "total_steps": 2679, "loss": 0.7821, "learning_rate": 4.889595628655044e-06, "epoch": 0.30993006993006994, "percentage": 10.34, "elapsed_time": "4:32:51", "remaining_time": "1 day, 15:26:02", "throughput": "17741.81", "total_tokens": 290455552}
{"current_steps": 278, "total_steps": 2679, "loss": 0.8368, "learning_rate": 4.8887242365111155e-06, "epoch": 0.31104895104895103, "percentage": 10.38, "elapsed_time": "4:33:54", "remaining_time": "1 day, 15:25:41", "throughput": "17737.15", "total_tokens": 291504128}
{"current_steps": 279, "total_steps": 2679, "loss": 0.8302, "learning_rate": 4.887849497304289e-06, "epoch": 0.31216783216783217, "percentage": 10.41, "elapsed_time": "4:34:59", "remaining_time": "1 day, 15:25:32", "throughput": "17730.88", "total_tokens": 292552704}
{"current_steps": 280, "total_steps": 2679, "loss": 0.7727, "learning_rate": 4.886971412260244e-06, "epoch": 0.3132867132867133, "percentage": 10.45, "elapsed_time": "4:35:59", "remaining_time": "1 day, 15:24:35", "throughput": "17730.55", "total_tokens": 293601280}
{"current_steps": 281, "total_steps": 2679, "loss": 0.7328, "learning_rate": 4.886089982609345e-06, "epoch": 0.3144055944055944, "percentage": 10.49, "elapsed_time": "4:36:52", "remaining_time": "1 day, 15:22:52", "throughput": "17736.10", "total_tokens": 294649856}
{"current_steps": 282, "total_steps": 2679, "loss": 0.7126, "learning_rate": 4.885205209586647e-06, "epoch": 0.31552447552447555, "percentage": 10.53, "elapsed_time": "4:37:40", "remaining_time": "1 day, 15:20:17", "throughput": "17748.07", "total_tokens": 295698432}
{"current_steps": 283, "total_steps": 2679, "loss": 0.8727, "learning_rate": 4.8843170944318855e-06, "epoch": 0.31664335664335663, "percentage": 10.56, "elapsed_time": "4:38:41", "remaining_time": "1 day, 15:19:31", "throughput": "17746.45", "total_tokens": 296747008}
{"current_steps": 284, "total_steps": 2679, "loss": 0.9965, "learning_rate": 4.883425638389482e-06, "epoch": 0.3177622377622378, "percentage": 10.6, "elapsed_time": "4:39:53", "remaining_time": "1 day, 15:20:20", "throughput": "17732.91", "total_tokens": 297795584}
{"current_steps": 285, "total_steps": 2679, "loss": 0.7599, "learning_rate": 4.882530842708537e-06, "epoch": 0.31888111888111886, "percentage": 10.64, "elapsed_time": "4:40:47", "remaining_time": "1 day, 15:18:38", "throughput": "17738.26", "total_tokens": 298844160}
{"current_steps": 286, "total_steps": 2679, "loss": 0.6943, "learning_rate": 4.881632708642832e-06, "epoch": 0.32, "percentage": 10.68, "elapsed_time": "4:41:35", "remaining_time": "1 day, 15:16:07", "throughput": "17749.83", "total_tokens": 299892736}
{"current_steps": 287, "total_steps": 2679, "loss": 0.7503, "learning_rate": 4.880731237450828e-06, "epoch": 0.3211188811188811, "percentage": 10.71, "elapsed_time": "4:42:35", "remaining_time": "1 day, 15:15:11", "throughput": "17749.41", "total_tokens": 300941312}
{"current_steps": 288, "total_steps": 2679, "loss": 0.7757, "learning_rate": 4.8798264303956565e-06, "epoch": 0.32223776223776224, "percentage": 10.75, "elapsed_time": "4:43:34", "remaining_time": "1 day, 15:14:18", "throughput": "17748.58", "total_tokens": 301989888}
{"current_steps": 289, "total_steps": 2679, "loss": 0.757, "learning_rate": 4.878918288745128e-06, "epoch": 0.3233566433566434, "percentage": 10.79, "elapsed_time": "4:44:28", "remaining_time": "1 day, 15:12:38", "throughput": "17753.76", "total_tokens": 303038464}
{"current_steps": 290, "total_steps": 2679, "loss": 0.7132, "learning_rate": 4.8780068137717255e-06, "epoch": 0.32447552447552447, "percentage": 10.82, "elapsed_time": "4:45:22", "remaining_time": "1 day, 15:10:57", "throughput": "17759.05", "total_tokens": 304087040}
{"current_steps": 291, "total_steps": 2679, "loss": 0.7994, "learning_rate": 4.877092006752599e-06, "epoch": 0.3255944055944056, "percentage": 10.86, "elapsed_time": "4:46:23", "remaining_time": "1 day, 15:10:07", "throughput": "17757.98", "total_tokens": 305135616}
{"current_steps": 292, "total_steps": 2679, "loss": 0.9409, "learning_rate": 4.8761738689695695e-06, "epoch": 0.3267132867132867, "percentage": 10.9, "elapsed_time": "4:47:28", "remaining_time": "1 day, 15:10:03", "throughput": "17750.99", "total_tokens": 306184192}
{"current_steps": 293, "total_steps": 2679, "loss": 0.7061, "learning_rate": 4.875252401709126e-06, "epoch": 0.32783216783216784, "percentage": 10.94, "elapsed_time": "4:48:22", "remaining_time": "1 day, 15:08:18", "throughput": "17756.71", "total_tokens": 307232768}
{"current_steps": 294, "total_steps": 2679, "loss": 0.8946, "learning_rate": 4.8743276062624214e-06, "epoch": 0.32895104895104893, "percentage": 10.97, "elapsed_time": "4:49:33", "remaining_time": "1 day, 15:08:59", "throughput": "17744.14", "total_tokens": 308281344}
{"current_steps": 295, "total_steps": 2679, "loss": 0.7836, "learning_rate": 4.873399483925272e-06, "epoch": 0.3300699300699301, "percentage": 11.01, "elapsed_time": "4:50:33", "remaining_time": "1 day, 15:08:02", "throughput": "17743.89", "total_tokens": 309329920}
{"current_steps": 296, "total_steps": 2679, "loss": 0.753, "learning_rate": 4.872468035998155e-06, "epoch": 0.3311888111888112, "percentage": 11.05, "elapsed_time": "4:51:31", "remaining_time": "1 day, 15:06:58", "throughput": "17744.53", "total_tokens": 310378496}
{"current_steps": 297, "total_steps": 2679, "loss": 0.8541, "learning_rate": 4.87153326378621e-06, "epoch": 0.3323076923076923, "percentage": 11.09, "elapsed_time": "4:52:37", "remaining_time": "1 day, 15:06:51", "throughput": "17737.94", "total_tokens": 311427072}
{"current_steps": 298, "total_steps": 2679, "loss": 0.7654, "learning_rate": 4.8705951685992325e-06, "epoch": 0.33342657342657345, "percentage": 11.12, "elapsed_time": "4:53:31", "remaining_time": "1 day, 15:05:13", "throughput": "17742.90", "total_tokens": 312475648}
{"current_steps": 299, "total_steps": 2679, "loss": 0.8081, "learning_rate": 4.8696537517516754e-06, "epoch": 0.33454545454545453, "percentage": 11.16, "elapsed_time": "4:54:36", "remaining_time": "1 day, 15:05:04", "throughput": "17736.51", "total_tokens": 313524224}
{"current_steps": 300, "total_steps": 2679, "loss": 0.6635, "learning_rate": 4.868709014562643e-06, "epoch": 0.3356643356643357, "percentage": 11.2, "elapsed_time": "4:55:24", "remaining_time": "1 day, 15:02:34", "throughput": "17747.96", "total_tokens": 314572800}
{"current_steps": 301, "total_steps": 2679, "loss": 0.8295, "learning_rate": 4.8677609583558956e-06, "epoch": 0.33678321678321677, "percentage": 11.24, "elapsed_time": "4:56:29", "remaining_time": "1 day, 15:02:25", "throughput": "17741.72", "total_tokens": 315621376}
{"current_steps": 302, "total_steps": 2679, "loss": 0.824, "learning_rate": 4.866809584459842e-06, "epoch": 0.3379020979020979, "percentage": 11.27, "elapsed_time": "4:57:35", "remaining_time": "1 day, 15:02:15", "throughput": "17735.54", "total_tokens": 316669952}
{"current_steps": 303, "total_steps": 2679, "loss": 0.6988, "learning_rate": 4.865854894207541e-06, "epoch": 0.339020979020979, "percentage": 11.31, "elapsed_time": "4:58:29", "remaining_time": "1 day, 15:00:39", "throughput": "17740.18", "total_tokens": 317718528}
{"current_steps": 304, "total_steps": 2679, "loss": 0.7389, "learning_rate": 4.864896888936698e-06, "epoch": 0.34013986013986014, "percentage": 11.35, "elapsed_time": "4:59:23", "remaining_time": "1 day, 14:58:57", "throughput": "17745.55", "total_tokens": 318767104}
{"current_steps": 305, "total_steps": 2679, "loss": 0.8982, "learning_rate": 4.863935569989662e-06, "epoch": 0.3412587412587413, "percentage": 11.38, "elapsed_time": "5:00:29", "remaining_time": "1 day, 14:58:51", "throughput": "17738.87", "total_tokens": 319815680}
{"current_steps": 306, "total_steps": 2679, "loss": 0.8236, "learning_rate": 4.8629709387134255e-06, "epoch": 0.34237762237762237, "percentage": 11.42, "elapsed_time": "5:01:29", "remaining_time": "1 day, 14:58:00", "throughput": "17737.83", "total_tokens": 320864256}
{"current_steps": 307, "total_steps": 2679, "loss": 0.7898, "learning_rate": 4.8620029964596234e-06, "epoch": 0.3434965034965035, "percentage": 11.46, "elapsed_time": "5:02:29", "remaining_time": "1 day, 14:57:07", "throughput": "17737.08", "total_tokens": 321912832}
{"current_steps": 308, "total_steps": 2679, "loss": 0.7371, "learning_rate": 4.86103174458453e-06, "epoch": 0.3446153846153846, "percentage": 11.5, "elapsed_time": "5:03:23", "remaining_time": "1 day, 14:55:30", "throughput": "17741.89", "total_tokens": 322961408}
{"current_steps": 309, "total_steps": 2679, "loss": 0.8417, "learning_rate": 4.860057184449057e-06, "epoch": 0.34573426573426574, "percentage": 11.53, "elapsed_time": "5:04:24", "remaining_time": "1 day, 14:54:49", "throughput": "17739.60", "total_tokens": 324009984}
{"current_steps": 310, "total_steps": 2679, "loss": 0.7701, "learning_rate": 4.8590793174187486e-06, "epoch": 0.34685314685314683, "percentage": 11.57, "elapsed_time": "5:05:24", "remaining_time": "1 day, 14:53:51", "throughput": "17739.41", "total_tokens": 325058560}
{"current_steps": 311, "total_steps": 2679, "loss": 0.6911, "learning_rate": 4.858098144863786e-06, "epoch": 0.347972027972028, "percentage": 11.61, "elapsed_time": "5:06:17", "remaining_time": "1 day, 14:52:09", "throughput": "17744.79", "total_tokens": 326107136}
{"current_steps": 312, "total_steps": 2679, "loss": 0.8692, "learning_rate": 4.85711366815898e-06, "epoch": 0.3490909090909091, "percentage": 11.65, "elapsed_time": "5:07:23", "remaining_time": "1 day, 14:52:05", "throughput": "17737.88", "total_tokens": 327155712}
{"current_steps": 313, "total_steps": 2679, "loss": 0.8684, "learning_rate": 4.856125888683775e-06, "epoch": 0.3502097902097902, "percentage": 11.68, "elapsed_time": "5:08:23", "remaining_time": "1 day, 14:51:12", "throughput": "17737.13", "total_tokens": 328204288}
{"current_steps": 314, "total_steps": 2679, "loss": 0.6737, "learning_rate": 4.855134807822238e-06, "epoch": 0.35132867132867135, "percentage": 11.72, "elapsed_time": "5:09:12", "remaining_time": "1 day, 14:48:52", "throughput": "17747.41", "total_tokens": 329252864}
{"current_steps": 315, "total_steps": 2679, "loss": 0.7885, "learning_rate": 4.854140426963064e-06, "epoch": 0.35244755244755244, "percentage": 11.76, "elapsed_time": "5:10:08", "remaining_time": "1 day, 14:47:32", "throughput": "17750.04", "total_tokens": 330301440}
{"current_steps": 316, "total_steps": 2679, "loss": 0.9624, "learning_rate": 4.853142747499574e-06, "epoch": 0.3535664335664336, "percentage": 11.8, "elapsed_time": "5:11:17", "remaining_time": "1 day, 14:47:47", "throughput": "17740.66", "total_tokens": 331350016}
{"current_steps": 317, "total_steps": 2679, "loss": 0.6914, "learning_rate": 4.852141770829707e-06, "epoch": 0.35468531468531467, "percentage": 11.83, "elapsed_time": "5:12:05", "remaining_time": "1 day, 14:45:27", "throughput": "17750.84", "total_tokens": 332398592}
{"current_steps": 318, "total_steps": 2679, "loss": 0.9217, "learning_rate": 4.851137498356025e-06, "epoch": 0.3558041958041958, "percentage": 11.87, "elapsed_time": "5:13:04", "remaining_time": "1 day, 14:44:23", "throughput": "17751.51", "total_tokens": 333447168}
{"current_steps": 319, "total_steps": 2679, "loss": 0.8165, "learning_rate": 4.850129931485709e-06, "epoch": 0.3569230769230769, "percentage": 11.91, "elapsed_time": "5:14:04", "remaining_time": "1 day, 14:43:32", "throughput": "17750.55", "total_tokens": 334495744}
{"current_steps": 320, "total_steps": 2679, "loss": 0.7721, "learning_rate": 4.849119071630553e-06, "epoch": 0.35804195804195804, "percentage": 11.94, "elapsed_time": "5:14:58", "remaining_time": "1 day, 14:42:00", "throughput": "17754.65", "total_tokens": 335544320}
{"current_steps": 321, "total_steps": 2679, "loss": 0.9471, "learning_rate": 4.848104920206964e-06, "epoch": 0.3591608391608392, "percentage": 11.98, "elapsed_time": "5:16:10", "remaining_time": "1 day, 14:42:30", "throughput": "17743.39", "total_tokens": 336592896}
{"current_steps": 322, "total_steps": 2679, "loss": 0.8085, "learning_rate": 4.847087478635968e-06, "epoch": 0.3602797202797203, "percentage": 12.02, "elapsed_time": "5:17:08", "remaining_time": "1 day, 14:41:25", "throughput": "17744.14", "total_tokens": 337641472}
{"current_steps": 323, "total_steps": 2679, "loss": 0.7591, "learning_rate": 4.846066748343193e-06, "epoch": 0.3613986013986014, "percentage": 12.06, "elapsed_time": "5:18:08", "remaining_time": "1 day, 14:40:32", "throughput": "17743.32", "total_tokens": 338690048}
{"current_steps": 324, "total_steps": 2679, "loss": 0.7275, "learning_rate": 4.845042730758881e-06, "epoch": 0.3625174825174825, "percentage": 12.09, "elapsed_time": "5:19:01", "remaining_time": "1 day, 14:38:53", "throughput": "17748.40", "total_tokens": 339738624}
{"current_steps": 325, "total_steps": 2679, "loss": 0.8182, "learning_rate": 4.844015427317878e-06, "epoch": 0.36363636363636365, "percentage": 12.13, "elapsed_time": "5:20:03", "remaining_time": "1 day, 14:38:14", "throughput": "17745.79", "total_tokens": 340787200}
{"current_steps": 326, "total_steps": 2679, "loss": 0.7841, "learning_rate": 4.842984839459631e-06, "epoch": 0.36475524475524473, "percentage": 12.17, "elapsed_time": "5:21:03", "remaining_time": "1 day, 14:37:20", "throughput": "17745.12", "total_tokens": 341835776}
{"current_steps": 327, "total_steps": 2679, "loss": 0.7658, "learning_rate": 4.8419509686281965e-06, "epoch": 0.3658741258741259, "percentage": 12.21, "elapsed_time": "5:22:02", "remaining_time": "1 day, 14:36:23", "throughput": "17744.93", "total_tokens": 342884352}
{"current_steps": 328, "total_steps": 2679, "loss": 0.7905, "learning_rate": 4.8409138162722235e-06, "epoch": 0.366993006993007, "percentage": 12.24, "elapsed_time": "5:23:01", "remaining_time": "1 day, 14:35:22", "throughput": "17745.11", "total_tokens": 343932928}
{"current_steps": 329, "total_steps": 2679, "loss": 0.7325, "learning_rate": 4.839873383844964e-06, "epoch": 0.3681118881118881, "percentage": 12.28, "elapsed_time": "5:23:55", "remaining_time": "1 day, 14:33:44", "throughput": "17750.10", "total_tokens": 344981504}
{"current_steps": 330, "total_steps": 2679, "loss": 0.8508, "learning_rate": 4.838829672804264e-06, "epoch": 0.36923076923076925, "percentage": 12.32, "elapsed_time": "5:24:57", "remaining_time": "1 day, 14:33:05", "throughput": "17747.58", "total_tokens": 346030080}
{"current_steps": 331, "total_steps": 2679, "loss": 0.7284, "learning_rate": 4.837782684612562e-06, "epoch": 0.37034965034965034, "percentage": 12.36, "elapsed_time": "5:25:51", "remaining_time": "1 day, 14:31:32", "throughput": "17751.93", "total_tokens": 347078656}
{"current_steps": 332, "total_steps": 2679, "loss": 0.7731, "learning_rate": 4.836732420736893e-06, "epoch": 0.3714685314685315, "percentage": 12.39, "elapsed_time": "5:26:45", "remaining_time": "1 day, 14:29:53", "throughput": "17757.02", "total_tokens": 348127232}
{"current_steps": 333, "total_steps": 2679, "loss": 0.862, "learning_rate": 4.835678882648878e-06, "epoch": 0.37258741258741257, "percentage": 12.43, "elapsed_time": "5:27:45", "remaining_time": "1 day, 14:29:02", "throughput": "17756.01", "total_tokens": 349175808}
{"current_steps": 334, "total_steps": 2679, "loss": 0.8556, "learning_rate": 4.834622071824726e-06, "epoch": 0.3737062937062937, "percentage": 12.47, "elapsed_time": "5:28:51", "remaining_time": "1 day, 14:28:51", "throughput": "17749.81", "total_tokens": 350224384}
{"current_steps": 335, "total_steps": 2679, "loss": 0.7893, "learning_rate": 4.833561989745232e-06, "epoch": 0.3748251748251748, "percentage": 12.5, "elapsed_time": "5:29:50", "remaining_time": "1 day, 14:27:57", "throughput": "17749.14", "total_tokens": 351272960}
{"current_steps": 336, "total_steps": 2679, "loss": 0.8358, "learning_rate": 4.832498637895778e-06, "epoch": 0.37594405594405594, "percentage": 12.54, "elapsed_time": "5:30:51", "remaining_time": "1 day, 14:27:06", "throughput": "17748.12", "total_tokens": 352321536}
{"current_steps": 337, "total_steps": 2679, "loss": 0.7147, "learning_rate": 4.831432017766323e-06, "epoch": 0.3770629370629371, "percentage": 12.58, "elapsed_time": "5:31:45", "remaining_time": "1 day, 14:25:35", "throughput": "17752.24", "total_tokens": 353370112}
{"current_steps": 338, "total_steps": 2679, "loss": 0.829, "learning_rate": 4.830362130851407e-06, "epoch": 0.3781818181818182, "percentage": 12.62, "elapsed_time": "5:32:49", "remaining_time": "1 day, 14:25:07", "throughput": "17748.25", "total_tokens": 354418688}
{"current_steps": 339, "total_steps": 2679, "loss": 0.7823, "learning_rate": 4.829288978650149e-06, "epoch": 0.3793006993006993, "percentage": 12.65, "elapsed_time": "5:33:48", "remaining_time": "1 day, 14:24:10", "throughput": "17747.96", "total_tokens": 355467264}
{"current_steps": 340, "total_steps": 2679, "loss": 0.722, "learning_rate": 4.82821256266624e-06, "epoch": 0.3804195804195804, "percentage": 12.69, "elapsed_time": "5:34:48", "remaining_time": "1 day, 14:23:15", "throughput": "17747.46", "total_tokens": 356515840}
{"current_steps": 341, "total_steps": 2679, "loss": 0.7389, "learning_rate": 4.827132884407948e-06, "epoch": 0.38153846153846155, "percentage": 12.73, "elapsed_time": "5:35:42", "remaining_time": "1 day, 14:21:45", "throughput": "17751.43", "total_tokens": 357564416}
{"current_steps": 342, "total_steps": 2679, "loss": 0.7596, "learning_rate": 4.826049945388109e-06, "epoch": 0.38265734265734264, "percentage": 12.77, "elapsed_time": "5:36:39", "remaining_time": "1 day, 14:20:29", "throughput": "17753.58", "total_tokens": 358612992}
{"current_steps": 343, "total_steps": 2679, "loss": 0.8015, "learning_rate": 4.824963747124132e-06, "epoch": 0.3837762237762238, "percentage": 12.8, "elapsed_time": "5:37:42", "remaining_time": "1 day, 14:19:55", "throughput": "17750.41", "total_tokens": 359661568}
{"current_steps": 344, "total_steps": 2679, "loss": 0.7084, "learning_rate": 4.823874291137986e-06, "epoch": 0.3848951048951049, "percentage": 12.84, "elapsed_time": "5:38:41", "remaining_time": "1 day, 14:19:00", "throughput": "17749.91", "total_tokens": 360710144}
{"current_steps": 345, "total_steps": 2679, "loss": 0.7906, "learning_rate": 4.822781578956212e-06, "epoch": 0.386013986013986, "percentage": 12.88, "elapsed_time": "5:39:46", "remaining_time": "1 day, 14:18:41", "throughput": "17744.73", "total_tokens": 361758720}
{"current_steps": 346, "total_steps": 2679, "loss": 0.7454, "learning_rate": 4.8216856121099074e-06, "epoch": 0.38713286713286715, "percentage": 12.92, "elapsed_time": "5:40:40", "remaining_time": "1 day, 14:17:07", "throughput": "17749.20", "total_tokens": 362807296}
{"current_steps": 347, "total_steps": 2679, "loss": 0.6856, "learning_rate": 4.820586392134735e-06, "epoch": 0.38825174825174824, "percentage": 12.95, "elapsed_time": "5:41:28", "remaining_time": "1 day, 14:14:54", "throughput": "17758.74", "total_tokens": 363855872}
{"current_steps": 348, "total_steps": 2679, "loss": 0.7819, "learning_rate": 4.819483920570914e-06, "epoch": 0.3893706293706294, "percentage": 12.99, "elapsed_time": "5:42:28", "remaining_time": "1 day, 14:14:01", "throughput": "17757.92", "total_tokens": 364904448}
{"current_steps": 349, "total_steps": 2679, "loss": 0.8421, "learning_rate": 4.818378198963218e-06, "epoch": 0.39048951048951047, "percentage": 13.03, "elapsed_time": "5:43:34", "remaining_time": "1 day, 14:13:46", "throughput": "17752.25", "total_tokens": 365953024}
{"current_steps": 350, "total_steps": 2679, "loss": 0.8554, "learning_rate": 4.817269228860978e-06, "epoch": 0.3916083916083916, "percentage": 13.06, "elapsed_time": "5:44:33", "remaining_time": "1 day, 14:12:46", "throughput": "17752.37", "total_tokens": 367001600}
{"current_steps": 351, "total_steps": 2679, "loss": 0.8218, "learning_rate": 4.816157011818073e-06, "epoch": 0.3927272727272727, "percentage": 13.1, "elapsed_time": "5:45:33", "remaining_time": "1 day, 14:11:53", "throughput": "17751.55", "total_tokens": 368050176}
{"current_steps": 352, "total_steps": 2679, "loss": 0.8468, "learning_rate": 4.815041549392934e-06, "epoch": 0.39384615384615385, "percentage": 13.14, "elapsed_time": "5:46:33", "remaining_time": "1 day, 14:11:01", "throughput": "17750.62", "total_tokens": 369098752}
{"current_steps": 353, "total_steps": 2679, "loss": 0.9159, "learning_rate": 4.813922843148537e-06, "epoch": 0.394965034965035, "percentage": 13.18, "elapsed_time": "5:47:33", "remaining_time": "1 day, 14:10:11", "throughput": "17749.55", "total_tokens": 370147328}
{"current_steps": 354, "total_steps": 2679, "loss": 0.7099, "learning_rate": 4.8128008946524085e-06, "epoch": 0.3960839160839161, "percentage": 13.21, "elapsed_time": "5:48:27", "remaining_time": "1 day, 14:08:38", "throughput": "17753.88", "total_tokens": 371195904}
{"current_steps": 355, "total_steps": 2679, "loss": 0.9111, "learning_rate": 4.811675705476613e-06, "epoch": 0.3972027972027972, "percentage": 13.25, "elapsed_time": "5:49:33", "remaining_time": "1 day, 14:08:21", "throughput": "17748.48", "total_tokens": 372244480}
{"current_steps": 356, "total_steps": 2679, "loss": 0.8061, "learning_rate": 4.810547277197755e-06, "epoch": 0.3983216783216783, "percentage": 13.29, "elapsed_time": "5:50:33", "remaining_time": "1 day, 14:07:27", "throughput": "17747.80", "total_tokens": 373293056}
{"current_steps": 357, "total_steps": 2679, "loss": 0.7518, "learning_rate": 4.809415611396984e-06, "epoch": 0.39944055944055945, "percentage": 13.33, "elapsed_time": "5:51:27", "remaining_time": "1 day, 14:05:57", "throughput": "17751.76", "total_tokens": 374341632}
{"current_steps": 358, "total_steps": 2679, "loss": 0.7067, "learning_rate": 4.80828070965998e-06, "epoch": 0.40055944055944054, "percentage": 13.36, "elapsed_time": "5:52:19", "remaining_time": "1 day, 14:04:15", "throughput": "17757.39", "total_tokens": 375390208}
{"current_steps": 359, "total_steps": 2679, "loss": 0.8054, "learning_rate": 4.807142573576958e-06, "epoch": 0.4016783216783217, "percentage": 13.4, "elapsed_time": "5:53:14", "remaining_time": "1 day, 14:02:47", "throughput": "17761.08", "total_tokens": 376438784}
{"current_steps": 360, "total_steps": 2679, "loss": 0.7113, "learning_rate": 4.8060012047426666e-06, "epoch": 0.4027972027972028, "percentage": 13.44, "elapsed_time": "5:54:09", "remaining_time": "1 day, 14:01:21", "throughput": "17764.68", "total_tokens": 377487360}
{"current_steps": 361, "total_steps": 2679, "loss": 0.7598, "learning_rate": 4.8048566047563835e-06, "epoch": 0.4039160839160839, "percentage": 13.48, "elapsed_time": "5:55:08", "remaining_time": "1 day, 14:00:25", "throughput": "17764.29", "total_tokens": 378535936}
{"current_steps": 362, "total_steps": 2679, "loss": 0.821, "learning_rate": 4.803708775221914e-06, "epoch": 0.40503496503496506, "percentage": 13.51, "elapsed_time": "5:56:08", "remaining_time": "1 day, 13:59:32", "throughput": "17763.41", "total_tokens": 379584512}
{"current_steps": 363, "total_steps": 2679, "loss": 0.8442, "learning_rate": 4.802557717747588e-06, "epoch": 0.40615384615384614, "percentage": 13.55, "elapsed_time": "5:57:02", "remaining_time": "1 day, 13:58:02", "throughput": "17767.51", "total_tokens": 380633088}
{"current_steps": 364, "total_steps": 2679, "loss": 0.7244, "learning_rate": 4.80140343394626e-06, "epoch": 0.4072727272727273, "percentage": 13.59, "elapsed_time": "5:57:57", "remaining_time": "1 day, 13:56:33", "throughput": "17771.34", "total_tokens": 381681664}
{"current_steps": 365, "total_steps": 2679, "loss": 0.7735, "learning_rate": 4.800245925435302e-06, "epoch": 0.4083916083916084, "percentage": 13.62, "elapsed_time": "5:58:57", "remaining_time": "1 day, 13:55:40", "throughput": "17770.54", "total_tokens": 382730240}
{"current_steps": 366, "total_steps": 2679, "loss": 0.7388, "learning_rate": 4.799085193836609e-06, "epoch": 0.4095104895104895, "percentage": 13.66, "elapsed_time": "5:59:51", "remaining_time": "1 day, 13:54:13", "throughput": "17774.24", "total_tokens": 383778816}
{"current_steps": 367, "total_steps": 2679, "loss": 0.7508, "learning_rate": 4.797921240776587e-06, "epoch": 0.4106293706293706, "percentage": 13.7, "elapsed_time": "6:00:50", "remaining_time": "1 day, 13:53:15", "throughput": "17774.13", "total_tokens": 384827392}
{"current_steps": 368, "total_steps": 2679, "loss": 0.6736, "learning_rate": 4.79675406788616e-06, "epoch": 0.41174825174825175, "percentage": 13.74, "elapsed_time": "6:01:39", "remaining_time": "1 day, 13:51:12", "throughput": "17782.39", "total_tokens": 385875968}
{"current_steps": 369, "total_steps": 2679, "loss": 0.8284, "learning_rate": 4.795583676800762e-06, "epoch": 0.4128671328671329, "percentage": 13.77, "elapsed_time": "6:02:45", "remaining_time": "1 day, 13:50:54", "throughput": "17777.16", "total_tokens": 386924544}
{"current_steps": 370, "total_steps": 2679, "loss": 0.7609, "learning_rate": 4.794410069160337e-06, "epoch": 0.413986013986014, "percentage": 13.81, "elapsed_time": "6:03:44", "remaining_time": "1 day, 13:49:59", "throughput": "17776.61", "total_tokens": 387973120}
{"current_steps": 371, "total_steps": 2679, "loss": 0.7684, "learning_rate": 4.793233246609333e-06, "epoch": 0.4151048951048951, "percentage": 13.85, "elapsed_time": "6:04:39", "remaining_time": "1 day, 13:48:32", "throughput": "17780.19", "total_tokens": 389021696}
{"current_steps": 372, "total_steps": 2679, "loss": 0.6941, "learning_rate": 4.792053210796708e-06, "epoch": 0.4162237762237762, "percentage": 13.89, "elapsed_time": "6:05:33", "remaining_time": "1 day, 13:47:02", "throughput": "17784.27", "total_tokens": 390070272}
{"current_steps": 373, "total_steps": 2679, "loss": 0.8273, "learning_rate": 4.790869963375918e-06, "epoch": 0.41734265734265735, "percentage": 13.92, "elapsed_time": "6:06:33", "remaining_time": "1 day, 13:46:10", "throughput": "17783.40", "total_tokens": 391118848}
{"current_steps": 374, "total_steps": 2679, "loss": 0.8277, "learning_rate": 4.789683506004921e-06, "epoch": 0.41846153846153844, "percentage": 13.96, "elapsed_time": "6:07:38", "remaining_time": "1 day, 13:45:48", "throughput": "17778.55", "total_tokens": 392167424}
{"current_steps": 375, "total_steps": 2679, "loss": 0.8717, "learning_rate": 4.788493840346172e-06, "epoch": 0.4195804195804196, "percentage": 14.0, "elapsed_time": "6:08:41", "remaining_time": "1 day, 13:45:16", "throughput": "17775.02", "total_tokens": 393216000}
{"current_steps": 376, "total_steps": 2679, "loss": 0.8423, "learning_rate": 4.7873009680666225e-06, "epoch": 0.4206993006993007, "percentage": 14.04, "elapsed_time": "6:09:42", "remaining_time": "1 day, 13:44:25", "throughput": "17774.03", "total_tokens": 394264576}
{"current_steps": 377, "total_steps": 2679, "loss": 0.6853, "learning_rate": 4.786104890837715e-06, "epoch": 0.4218181818181818, "percentage": 14.07, "elapsed_time": "6:10:36", "remaining_time": "1 day, 13:42:55", "throughput": "17778.00", "total_tokens": 395313152}
{"current_steps": 378, "total_steps": 2679, "loss": 0.6717, "learning_rate": 4.7849056103353864e-06, "epoch": 0.42293706293706296, "percentage": 14.11, "elapsed_time": "6:11:24", "remaining_time": "1 day, 13:40:50", "throughput": "17786.75", "total_tokens": 396361728}
{"current_steps": 379, "total_steps": 2679, "loss": 0.9347, "learning_rate": 4.783703128240058e-06, "epoch": 0.42405594405594405, "percentage": 14.15, "elapsed_time": "6:12:31", "remaining_time": "1 day, 13:40:40", "throughput": "17780.29", "total_tokens": 397410304}
{"current_steps": 380, "total_steps": 2679, "loss": 0.7345, "learning_rate": 4.782497446236639e-06, "epoch": 0.4251748251748252, "percentage": 14.18, "elapsed_time": "6:13:31", "remaining_time": "1 day, 13:39:48", "throughput": "17779.32", "total_tokens": 398458880}
{"current_steps": 381, "total_steps": 2679, "loss": 0.6862, "learning_rate": 4.781288566014524e-06, "epoch": 0.4262937062937063, "percentage": 14.22, "elapsed_time": "6:14:19", "remaining_time": "1 day, 13:37:43", "throughput": "17787.97", "total_tokens": 399507456}
{"current_steps": 382, "total_steps": 2679, "loss": 0.6942, "learning_rate": 4.7800764892675836e-06, "epoch": 0.4274125874125874, "percentage": 14.26, "elapsed_time": "6:15:12", "remaining_time": "1 day, 13:36:12", "throughput": "17792.27", "total_tokens": 400556032}
{"current_steps": 383, "total_steps": 2679, "loss": 0.8405, "learning_rate": 4.778861217694174e-06, "epoch": 0.4285314685314685, "percentage": 14.3, "elapsed_time": "6:16:12", "remaining_time": "1 day, 13:35:16", "throughput": "17791.82", "total_tokens": 401604608}
{"current_steps": 384, "total_steps": 2679, "loss": 0.7666, "learning_rate": 4.7776427529971245e-06, "epoch": 0.42965034965034965, "percentage": 14.33, "elapsed_time": "6:17:07", "remaining_time": "1 day, 13:33:55", "throughput": "17794.77", "total_tokens": 402653184}
{"current_steps": 385, "total_steps": 2679, "loss": 0.7281, "learning_rate": 4.776421096883737e-06, "epoch": 0.4307692307692308, "percentage": 14.37, "elapsed_time": "6:18:07", "remaining_time": "1 day, 13:33:01", "throughput": "17794.09", "total_tokens": 403701760}
{"current_steps": 386, "total_steps": 2679, "loss": 0.8681, "learning_rate": 4.775196251065789e-06, "epoch": 0.4318881118881119, "percentage": 14.41, "elapsed_time": "6:19:12", "remaining_time": "1 day, 13:32:41", "throughput": "17788.96", "total_tokens": 404750336}
{"current_steps": 387, "total_steps": 2679, "loss": 0.797, "learning_rate": 4.773968217259525e-06, "epoch": 0.433006993006993, "percentage": 14.45, "elapsed_time": "6:20:13", "remaining_time": "1 day, 13:31:49", "throughput": "17788.05", "total_tokens": 405798912}
{"current_steps": 388, "total_steps": 2679, "loss": 0.7469, "learning_rate": 4.772736997185656e-06, "epoch": 0.4341258741258741, "percentage": 14.48, "elapsed_time": "6:21:06", "remaining_time": "1 day, 13:30:19", "throughput": "17792.14", "total_tokens": 406847488}
{"current_steps": 389, "total_steps": 2679, "loss": 0.6814, "learning_rate": 4.7715025925693595e-06, "epoch": 0.43524475524475525, "percentage": 14.52, "elapsed_time": "6:21:54", "remaining_time": "1 day, 13:28:16", "throughput": "17800.64", "total_tokens": 407896064}
{"current_steps": 390, "total_steps": 2679, "loss": 0.6619, "learning_rate": 4.7702650051402745e-06, "epoch": 0.43636363636363634, "percentage": 14.56, "elapsed_time": "6:22:43", "remaining_time": "1 day, 13:26:16", "throughput": "17808.66", "total_tokens": 408944640}
{"current_steps": 391, "total_steps": 2679, "loss": 0.6681, "learning_rate": 4.769024236632498e-06, "epoch": 0.4374825174825175, "percentage": 14.59, "elapsed_time": "6:23:31", "remaining_time": "1 day, 13:24:14", "throughput": "17817.04", "total_tokens": 409993216}
{"current_steps": 392, "total_steps": 2679, "loss": 0.72, "learning_rate": 4.767780288784588e-06, "epoch": 0.43860139860139863, "percentage": 14.63, "elapsed_time": "6:24:25", "remaining_time": "1 day, 13:22:45", "throughput": "17821.01", "total_tokens": 411041792}
{"current_steps": 393, "total_steps": 2679, "loss": 0.8121, "learning_rate": 4.766533163339553e-06, "epoch": 0.4397202797202797, "percentage": 14.67, "elapsed_time": "6:25:24", "remaining_time": "1 day, 13:21:49", "throughput": "17820.61", "total_tokens": 412090368}
{"current_steps": 394, "total_steps": 2679, "loss": 0.8041, "learning_rate": 4.765282862044857e-06, "epoch": 0.44083916083916086, "percentage": 14.71, "elapsed_time": "6:26:18", "remaining_time": "1 day, 13:20:24", "throughput": "17824.05", "total_tokens": 413138944}
{"current_steps": 395, "total_steps": 2679, "loss": 0.8315, "learning_rate": 4.764029386652412e-06, "epoch": 0.44195804195804195, "percentage": 14.74, "elapsed_time": "6:27:14", "remaining_time": "1 day, 13:19:06", "throughput": "17826.62", "total_tokens": 414187520}
{"current_steps": 396, "total_steps": 2679, "loss": 0.7477, "learning_rate": 4.76277273891858e-06, "epoch": 0.4430769230769231, "percentage": 14.78, "elapsed_time": "6:28:13", "remaining_time": "1 day, 13:18:11", "throughput": "17826.07", "total_tokens": 415236096}
{"current_steps": 397, "total_steps": 2679, "loss": 0.6838, "learning_rate": 4.761512920604165e-06, "epoch": 0.4441958041958042, "percentage": 14.82, "elapsed_time": "6:29:01", "remaining_time": "1 day, 13:16:11", "throughput": "17834.28", "total_tokens": 416284672}
{"current_steps": 398, "total_steps": 2679, "loss": 0.7911, "learning_rate": 4.760249933474418e-06, "epoch": 0.4453146853146853, "percentage": 14.86, "elapsed_time": "6:29:56", "remaining_time": "1 day, 13:14:47", "throughput": "17837.67", "total_tokens": 417333248}
{"current_steps": 399, "total_steps": 2679, "loss": 0.7992, "learning_rate": 4.758983779299025e-06, "epoch": 0.4464335664335664, "percentage": 14.89, "elapsed_time": "6:30:51", "remaining_time": "1 day, 13:13:26", "throughput": "17840.62", "total_tokens": 418381824}
{"current_steps": 400, "total_steps": 2679, "loss": 0.7182, "learning_rate": 4.757714459852111e-06, "epoch": 0.44755244755244755, "percentage": 14.93, "elapsed_time": "6:31:44", "remaining_time": "1 day, 13:11:58", "throughput": "17844.53", "total_tokens": 419430400}
{"current_steps": 401, "total_steps": 2679, "loss": 0.6905, "learning_rate": 4.75644197691224e-06, "epoch": 0.4486713286713287, "percentage": 14.97, "elapsed_time": "6:32:38", "remaining_time": "1 day, 13:10:33", "throughput": "17847.99", "total_tokens": 420478976}
{"current_steps": 402, "total_steps": 2679, "loss": 0.6494, "learning_rate": 4.755166332262403e-06, "epoch": 0.4497902097902098, "percentage": 15.01, "elapsed_time": "6:33:26", "remaining_time": "1 day, 13:08:34", "throughput": "17856.06", "total_tokens": 421527552}
{"current_steps": 403, "total_steps": 2679, "loss": 0.8166, "learning_rate": 4.753887527690027e-06, "epoch": 0.4509090909090909, "percentage": 15.04, "elapsed_time": "6:34:26", "remaining_time": "1 day, 13:07:41", "throughput": "17855.21", "total_tokens": 422576128}
{"current_steps": 404, "total_steps": 2679, "loss": 0.7885, "learning_rate": 4.7526055649869606e-06, "epoch": 0.452027972027972, "percentage": 15.08, "elapsed_time": "6:35:23", "remaining_time": "1 day, 13:06:33", "throughput": "17856.55", "total_tokens": 423624704}
{"current_steps": 405, "total_steps": 2679, "loss": 0.9796, "learning_rate": 4.7513204459494825e-06, "epoch": 0.45314685314685316, "percentage": 15.12, "elapsed_time": "6:36:29", "remaining_time": "1 day, 13:06:15", "throughput": "17851.09", "total_tokens": 424673280}
{"current_steps": 406, "total_steps": 2679, "loss": 0.6518, "learning_rate": 4.7500321723782905e-06, "epoch": 0.45426573426573424, "percentage": 15.15, "elapsed_time": "6:37:19", "remaining_time": "1 day, 13:04:25", "throughput": "17857.92", "total_tokens": 425721856}
{"current_steps": 407, "total_steps": 2679, "loss": 0.7536, "learning_rate": 4.748740746078505e-06, "epoch": 0.4553846153846154, "percentage": 15.19, "elapsed_time": "6:38:18", "remaining_time": "1 day, 13:03:30", "throughput": "17857.36", "total_tokens": 426770432}
{"current_steps": 408, "total_steps": 2679, "loss": 0.714, "learning_rate": 4.747446168859664e-06, "epoch": 0.45650349650349653, "percentage": 15.23, "elapsed_time": "6:39:13", "remaining_time": "1 day, 13:02:07", "throughput": "17860.71", "total_tokens": 427819008}
{"current_steps": 409, "total_steps": 2679, "loss": 0.9065, "learning_rate": 4.746148442535717e-06, "epoch": 0.4576223776223776, "percentage": 15.27, "elapsed_time": "6:40:14", "remaining_time": "1 day, 13:01:24", "throughput": "17858.57", "total_tokens": 428867584}
{"current_steps": 410, "total_steps": 2679, "loss": 0.7311, "learning_rate": 4.744847568925032e-06, "epoch": 0.45874125874125876, "percentage": 15.3, "elapsed_time": "6:41:10", "remaining_time": "1 day, 13:00:07", "throughput": "17861.00", "total_tokens": 429916160}
{"current_steps": 411, "total_steps": 2679, "loss": 0.8334, "learning_rate": 4.743543549850381e-06, "epoch": 0.45986013986013985, "percentage": 15.34, "elapsed_time": "6:42:15", "remaining_time": "1 day, 12:59:43", "throughput": "17856.41", "total_tokens": 430964736}
{"current_steps": 412, "total_steps": 2679, "loss": 0.6762, "learning_rate": 4.7422363871389465e-06, "epoch": 0.460979020979021, "percentage": 15.38, "elapsed_time": "6:43:03", "remaining_time": "1 day, 12:57:45", "throughput": "17864.31", "total_tokens": 432013312}
{"current_steps": 413, "total_steps": 2679, "loss": 0.8831, "learning_rate": 4.740926082622316e-06, "epoch": 0.4620979020979021, "percentage": 15.42, "elapsed_time": "6:44:06", "remaining_time": "1 day, 12:57:14", "throughput": "17860.55", "total_tokens": 433061888}
{"current_steps": 414, "total_steps": 2679, "loss": 0.8917, "learning_rate": 4.739612638136478e-06, "epoch": 0.4632167832167832, "percentage": 15.45, "elapsed_time": "6:45:14", "remaining_time": "1 day, 12:57:02", "throughput": "17854.30", "total_tokens": 434110464}
{"current_steps": 415, "total_steps": 2679, "loss": 0.721, "learning_rate": 4.738296055521821e-06, "epoch": 0.4643356643356643, "percentage": 15.49, "elapsed_time": "6:46:08", "remaining_time": "1 day, 12:55:40", "throughput": "17857.42", "total_tokens": 435159040}
{"current_steps": 416, "total_steps": 2679, "loss": 0.7055, "learning_rate": 4.736976336623133e-06, "epoch": 0.46545454545454545, "percentage": 15.53, "elapsed_time": "6:47:02", "remaining_time": "1 day, 12:54:14", "throughput": "17861.05", "total_tokens": 436207616}
{"current_steps": 417, "total_steps": 2679, "loss": 0.783, "learning_rate": 4.735653483289591e-06, "epoch": 0.4665734265734266, "percentage": 15.57, "elapsed_time": "6:47:56", "remaining_time": "1 day, 12:52:54", "throughput": "17864.02", "total_tokens": 437256192}
{"current_steps": 418, "total_steps": 2679, "loss": 0.7741, "learning_rate": 4.734327497374771e-06, "epoch": 0.4676923076923077, "percentage": 15.6, "elapsed_time": "6:48:57", "remaining_time": "1 day, 12:52:07", "throughput": "17862.40", "total_tokens": 438304768}
{"current_steps": 419, "total_steps": 2679, "loss": 0.7771, "learning_rate": 4.732998380736632e-06, "epoch": 0.46881118881118883, "percentage": 15.64, "elapsed_time": "6:49:57", "remaining_time": "1 day, 12:51:12", "throughput": "17861.88", "total_tokens": 439353344}
{"current_steps": 420, "total_steps": 2679, "loss": 0.944, "learning_rate": 4.731666135237524e-06, "epoch": 0.4699300699300699, "percentage": 15.68, "elapsed_time": "6:51:03", "remaining_time": "1 day, 12:50:54", "throughput": "17856.47", "total_tokens": 440401920}
{"current_steps": 421, "total_steps": 2679, "loss": 0.6851, "learning_rate": 4.730330762744178e-06, "epoch": 0.47104895104895106, "percentage": 15.71, "elapsed_time": "6:51:51", "remaining_time": "1 day, 12:48:58", "throughput": "17864.15", "total_tokens": 441450496}
{"current_steps": 422, "total_steps": 2679, "loss": 0.7333, "learning_rate": 4.72899226512771e-06, "epoch": 0.47216783216783215, "percentage": 15.75, "elapsed_time": "6:52:51", "remaining_time": "1 day, 12:48:03", "throughput": "17863.59", "total_tokens": 442499072}
{"current_steps": 423, "total_steps": 2679, "loss": 0.7371, "learning_rate": 4.7276506442636125e-06, "epoch": 0.4732867132867133, "percentage": 15.79, "elapsed_time": "6:53:46", "remaining_time": "1 day, 12:46:48", "throughput": "17865.80", "total_tokens": 443547648}
{"current_steps": 424, "total_steps": 2679, "loss": 0.7004, "learning_rate": 4.726305902031754e-06, "epoch": 0.47440559440559443, "percentage": 15.83, "elapsed_time": "6:54:35", "remaining_time": "1 day, 12:44:57", "throughput": "17872.89", "total_tokens": 444596224}
{"current_steps": 425, "total_steps": 2679, "loss": 0.76, "learning_rate": 4.7249580403163786e-06, "epoch": 0.4755244755244755, "percentage": 15.86, "elapsed_time": "6:55:28", "remaining_time": "1 day, 12:43:30", "throughput": "17876.71", "total_tokens": 445644800}
{"current_steps": 426, "total_steps": 2679, "loss": 0.7535, "learning_rate": 4.7236070610061e-06, "epoch": 0.47664335664335666, "percentage": 15.9, "elapsed_time": "6:56:28", "remaining_time": "1 day, 12:42:37", "throughput": "17875.95", "total_tokens": 446693376}
{"current_steps": 427, "total_steps": 2679, "loss": 0.8175, "learning_rate": 4.7222529659939e-06, "epoch": 0.47776223776223775, "percentage": 15.94, "elapsed_time": "6:57:28", "remaining_time": "1 day, 12:41:45", "throughput": "17875.11", "total_tokens": 447741952}
{"current_steps": 428, "total_steps": 2679, "loss": 0.836, "learning_rate": 4.720895757177126e-06, "epoch": 0.4788811188811189, "percentage": 15.98, "elapsed_time": "6:58:34", "remaining_time": "1 day, 12:41:23", "throughput": "17870.12", "total_tokens": 448790528}
{"current_steps": 429, "total_steps": 2679, "loss": 0.7785, "learning_rate": 4.7195354364574915e-06, "epoch": 0.48, "percentage": 16.01, "elapsed_time": "6:59:27", "remaining_time": "1 day, 12:39:59", "throughput": "17873.51", "total_tokens": 449839104}
{"current_steps": 430, "total_steps": 2679, "loss": 0.903, "learning_rate": 4.718172005741066e-06, "epoch": 0.4811188811188811, "percentage": 16.05, "elapsed_time": "7:00:39", "remaining_time": "1 day, 12:40:06", "throughput": "17864.59", "total_tokens": 450887680}
{"current_steps": 431, "total_steps": 2679, "loss": 0.8167, "learning_rate": 4.716805466938278e-06, "epoch": 0.4822377622377622, "percentage": 16.09, "elapsed_time": "7:01:41", "remaining_time": "1 day, 12:39:25", "throughput": "17862.24", "total_tokens": 451936256}
{"current_steps": 432, "total_steps": 2679, "loss": 0.7245, "learning_rate": 4.715435821963913e-06, "epoch": 0.48335664335664336, "percentage": 16.13, "elapsed_time": "7:02:35", "remaining_time": "1 day, 12:38:05", "throughput": "17865.18", "total_tokens": 452984832}
{"current_steps": 433, "total_steps": 2679, "loss": 0.6369, "learning_rate": 4.714063072737108e-06, "epoch": 0.4844755244755245, "percentage": 16.16, "elapsed_time": "7:03:23", "remaining_time": "1 day, 12:36:08", "throughput": "17872.97", "total_tokens": 454033408}
{"current_steps": 434, "total_steps": 2679, "loss": 0.6817, "learning_rate": 4.712687221181348e-06, "epoch": 0.4855944055944056, "percentage": 16.2, "elapsed_time": "7:04:17", "remaining_time": "1 day, 12:34:49", "throughput": "17875.83", "total_tokens": 455081984}
{"current_steps": 435, "total_steps": 2679, "loss": 0.7328, "learning_rate": 4.711308269224466e-06, "epoch": 0.48671328671328673, "percentage": 16.24, "elapsed_time": "7:05:17", "remaining_time": "1 day, 12:33:54", "throughput": "17875.27", "total_tokens": 456130560}
{"current_steps": 436, "total_steps": 2679, "loss": 0.6596, "learning_rate": 4.70992621879864e-06, "epoch": 0.4878321678321678, "percentage": 16.27, "elapsed_time": "7:06:13", "remaining_time": "1 day, 12:32:41", "throughput": "17877.26", "total_tokens": 457179136}
{"current_steps": 437, "total_steps": 2679, "loss": 0.684, "learning_rate": 4.708541071840388e-06, "epoch": 0.48895104895104896, "percentage": 16.31, "elapsed_time": "7:07:06", "remaining_time": "1 day, 12:31:15", "throughput": "17881.00", "total_tokens": 458227712}
{"current_steps": 438, "total_steps": 2679, "loss": 0.7575, "learning_rate": 4.70715283029057e-06, "epoch": 0.49006993006993005, "percentage": 16.35, "elapsed_time": "7:08:05", "remaining_time": "1 day, 12:30:20", "throughput": "17880.42", "total_tokens": 459276288}
{"current_steps": 439, "total_steps": 2679, "loss": 0.7695, "learning_rate": 4.705761496094377e-06, "epoch": 0.4911888111888112, "percentage": 16.39, "elapsed_time": "7:09:05", "remaining_time": "1 day, 12:29:27", "throughput": "17879.63", "total_tokens": 460324864}
{"current_steps": 440, "total_steps": 2679, "loss": 0.9081, "learning_rate": 4.704367071201339e-06, "epoch": 0.49230769230769234, "percentage": 16.42, "elapsed_time": "7:10:06", "remaining_time": "1 day, 12:28:38", "throughput": "17878.44", "total_tokens": 461373440}
{"current_steps": 441, "total_steps": 2679, "loss": 0.7648, "learning_rate": 4.702969557565312e-06, "epoch": 0.4934265734265734, "percentage": 16.46, "elapsed_time": "7:11:05", "remaining_time": "1 day, 12:27:43", "throughput": "17877.88", "total_tokens": 462422016}
{"current_steps": 442, "total_steps": 2679, "loss": 0.7444, "learning_rate": 4.701568957144483e-06, "epoch": 0.49454545454545457, "percentage": 16.5, "elapsed_time": "7:12:05", "remaining_time": "1 day, 12:26:49", "throughput": "17877.21", "total_tokens": 463470592}
{"current_steps": 443, "total_steps": 2679, "loss": 0.7005, "learning_rate": 4.700165271901361e-06, "epoch": 0.49566433566433565, "percentage": 16.54, "elapsed_time": "7:12:58", "remaining_time": "1 day, 12:25:25", "throughput": "17880.66", "total_tokens": 464519168}
{"current_steps": 444, "total_steps": 2679, "loss": 0.7749, "learning_rate": 4.698758503802782e-06, "epoch": 0.4967832167832168, "percentage": 16.57, "elapsed_time": "7:13:58", "remaining_time": "1 day, 12:24:33", "throughput": "17879.76", "total_tokens": 465567744}
{"current_steps": 445, "total_steps": 2679, "loss": 0.7667, "learning_rate": 4.697348654819898e-06, "epoch": 0.4979020979020979, "percentage": 16.61, "elapsed_time": "7:14:58", "remaining_time": "1 day, 12:23:38", "throughput": "17879.33", "total_tokens": 466616320}
{"current_steps": 446, "total_steps": 2679, "loss": 0.797, "learning_rate": 4.695935726928179e-06, "epoch": 0.499020979020979, "percentage": 16.65, "elapsed_time": "7:16:03", "remaining_time": "1 day, 12:23:13", "throughput": "17874.67", "total_tokens": 467664896}
{"current_steps": 447, "total_steps": 2679, "loss": 0.8127, "learning_rate": 4.6945197221074104e-06, "epoch": 0.5001398601398601, "percentage": 16.69, "elapsed_time": "7:17:08", "remaining_time": "1 day, 12:22:47", "throughput": "17870.29", "total_tokens": 468713472}
{"current_steps": 448, "total_steps": 2679, "loss": 0.6833, "learning_rate": 4.693100642341686e-06, "epoch": 0.5012587412587413, "percentage": 16.72, "elapsed_time": "7:18:02", "remaining_time": "1 day, 12:21:23", "throughput": "17873.72", "total_tokens": 469762048}
{"current_steps": 449, "total_steps": 2679, "loss": 0.7943, "learning_rate": 4.691678489619411e-06, "epoch": 0.5023776223776224, "percentage": 16.76, "elapsed_time": "7:18:56", "remaining_time": "1 day, 12:20:01", "throughput": "17876.89", "total_tokens": 470810624}
{"current_steps": 450, "total_steps": 2679, "loss": 0.7282, "learning_rate": 4.690253265933295e-06, "epoch": 0.5034965034965035, "percentage": 16.8, "elapsed_time": "7:19:50", "remaining_time": "1 day, 12:18:41", "throughput": "17879.83", "total_tokens": 471859200}
{"current_steps": 451, "total_steps": 2679, "loss": 0.8112, "learning_rate": 4.6888249732803516e-06, "epoch": 0.5046153846153846, "percentage": 16.83, "elapsed_time": "7:20:50", "remaining_time": "1 day, 12:17:48", "throughput": "17879.00", "total_tokens": 472907776}
{"current_steps": 452, "total_steps": 2679, "loss": 0.791, "learning_rate": 4.6873936136618925e-06, "epoch": 0.5057342657342657, "percentage": 16.87, "elapsed_time": "7:21:50", "remaining_time": "1 day, 12:16:56", "throughput": "17878.10", "total_tokens": 473956352}
{"current_steps": 453, "total_steps": 2679, "loss": 0.7884, "learning_rate": 4.685959189083531e-06, "epoch": 0.5068531468531469, "percentage": 16.91, "elapsed_time": "7:22:49", "remaining_time": "1 day, 12:16:01", "throughput": "17877.67", "total_tokens": 475004928}
{"current_steps": 454, "total_steps": 2679, "loss": 0.9747, "learning_rate": 4.68452170155517e-06, "epoch": 0.507972027972028, "percentage": 16.95, "elapsed_time": "7:24:01", "remaining_time": "1 day, 12:16:06", "throughput": "17868.98", "total_tokens": 476053504}
{"current_steps": 455, "total_steps": 2679, "loss": 0.7799, "learning_rate": 4.683081153091006e-06, "epoch": 0.509090909090909, "percentage": 16.98, "elapsed_time": "7:24:55", "remaining_time": "1 day, 12:14:46", "throughput": "17871.82", "total_tokens": 477102080}
{"current_steps": 456, "total_steps": 2679, "loss": 0.7367, "learning_rate": 4.681637545709527e-06, "epoch": 0.5102097902097902, "percentage": 17.02, "elapsed_time": "7:25:55", "remaining_time": "1 day, 12:13:51", "throughput": "17871.27", "total_tokens": 478150656}
{"current_steps": 457, "total_steps": 2679, "loss": 0.7124, "learning_rate": 4.680190881433504e-06, "epoch": 0.5113286713286713, "percentage": 17.06, "elapsed_time": "7:26:49", "remaining_time": "1 day, 12:12:31", "throughput": "17874.20", "total_tokens": 479199232}
{"current_steps": 458, "total_steps": 2679, "loss": 0.8018, "learning_rate": 4.678741162289993e-06, "epoch": 0.5124475524475525, "percentage": 17.1, "elapsed_time": "7:27:54", "remaining_time": "1 day, 12:12:04", "throughput": "17869.95", "total_tokens": 480247808}
{"current_steps": 459, "total_steps": 2679, "loss": 0.7131, "learning_rate": 4.67728839031033e-06, "epoch": 0.5135664335664336, "percentage": 17.13, "elapsed_time": "7:28:43", "remaining_time": "1 day, 12:10:16", "throughput": "17876.73", "total_tokens": 481296384}
{"current_steps": 460, "total_steps": 2679, "loss": 0.6567, "learning_rate": 4.675832567530126e-06, "epoch": 0.5146853146853146, "percentage": 17.17, "elapsed_time": "7:29:32", "remaining_time": "1 day, 12:08:34", "throughput": "17882.63", "total_tokens": 482344960}
{"current_steps": 461, "total_steps": 2679, "loss": 0.8276, "learning_rate": 4.674373695989272e-06, "epoch": 0.5158041958041958, "percentage": 17.21, "elapsed_time": "7:30:34", "remaining_time": "1 day, 12:07:52", "throughput": "17880.37", "total_tokens": 483393536}
{"current_steps": 462, "total_steps": 2679, "loss": 0.6733, "learning_rate": 4.6729117777319275e-06, "epoch": 0.5169230769230769, "percentage": 17.25, "elapsed_time": "7:31:22", "remaining_time": "1 day, 12:06:02", "throughput": "17887.41", "total_tokens": 484442112}
{"current_steps": 463, "total_steps": 2679, "loss": 0.8719, "learning_rate": 4.6714468148065215e-06, "epoch": 0.5180419580419581, "percentage": 17.28, "elapsed_time": "7:32:28", "remaining_time": "1 day, 12:05:37", "throughput": "17882.77", "total_tokens": 485490688}
{"current_steps": 464, "total_steps": 2679, "loss": 0.8106, "learning_rate": 4.669978809265749e-06, "epoch": 0.5191608391608392, "percentage": 17.32, "elapsed_time": "7:33:33", "remaining_time": "1 day, 12:05:10", "throughput": "17878.49", "total_tokens": 486539264}
{"current_steps": 465, "total_steps": 2679, "loss": 0.6892, "learning_rate": 4.668507763166568e-06, "epoch": 0.5202797202797202, "percentage": 17.36, "elapsed_time": "7:34:28", "remaining_time": "1 day, 12:03:52", "throughput": "17881.03", "total_tokens": 487587840}
{"current_steps": 466, "total_steps": 2679, "loss": 0.7149, "learning_rate": 4.667033678570199e-06, "epoch": 0.5213986013986014, "percentage": 17.39, "elapsed_time": "7:35:22", "remaining_time": "1 day, 12:02:31", "throughput": "17884.16", "total_tokens": 488636416}
{"current_steps": 467, "total_steps": 2679, "loss": 0.7681, "learning_rate": 4.665556557542118e-06, "epoch": 0.5225174825174825, "percentage": 17.43, "elapsed_time": "7:36:16", "remaining_time": "1 day, 12:01:11", "throughput": "17887.18", "total_tokens": 489684992}
{"current_steps": 468, "total_steps": 2679, "loss": 0.7663, "learning_rate": 4.664076402152056e-06, "epoch": 0.5236363636363637, "percentage": 17.47, "elapsed_time": "7:37:15", "remaining_time": "1 day, 12:00:16", "throughput": "17886.63", "total_tokens": 490733568}
{"current_steps": 469, "total_steps": 2679, "loss": 0.7281, "learning_rate": 4.662593214473995e-06, "epoch": 0.5247552447552447, "percentage": 17.51, "elapsed_time": "7:38:15", "remaining_time": "1 day, 11:59:22", "throughput": "17886.04", "total_tokens": 491782144}
{"current_steps": 470, "total_steps": 2679, "loss": 0.7265, "learning_rate": 4.6611069965861685e-06, "epoch": 0.5258741258741259, "percentage": 17.54, "elapsed_time": "7:39:14", "remaining_time": "1 day, 11:58:28", "throughput": "17885.36", "total_tokens": 492830720}
{"current_steps": 471, "total_steps": 2679, "loss": 0.8549, "learning_rate": 4.659617750571052e-06, "epoch": 0.526993006993007, "percentage": 17.58, "elapsed_time": "7:40:18", "remaining_time": "1 day, 11:57:51", "throughput": "17882.39", "total_tokens": 493879296}
{"current_steps": 472, "total_steps": 2679, "loss": 0.7059, "learning_rate": 4.658125478515369e-06, "epoch": 0.5281118881118881, "percentage": 17.62, "elapsed_time": "7:41:12", "remaining_time": "1 day, 11:56:31", "throughput": "17885.31", "total_tokens": 494927872}
{"current_steps": 473, "total_steps": 2679, "loss": 0.8532, "learning_rate": 4.656630182510078e-06, "epoch": 0.5292307692307693, "percentage": 17.66, "elapsed_time": "7:42:23", "remaining_time": "1 day, 11:56:31", "throughput": "17877.24", "total_tokens": 495976448}
{"current_steps": 474, "total_steps": 2679, "loss": 0.6441, "learning_rate": 4.6551318646503765e-06, "epoch": 0.5303496503496503, "percentage": 17.69, "elapsed_time": "7:43:13", "remaining_time": "1 day, 11:54:52", "throughput": "17882.82", "total_tokens": 497025024}
{"current_steps": 475, "total_steps": 2679, "loss": 0.828, "learning_rate": 4.6536305270356975e-06, "epoch": 0.5314685314685315, "percentage": 17.73, "elapsed_time": "7:44:19", "remaining_time": "1 day, 11:54:27", "throughput": "17878.17", "total_tokens": 498073600}
{"current_steps": 476, "total_steps": 2679, "loss": 0.6788, "learning_rate": 4.652126171769705e-06, "epoch": 0.5325874125874126, "percentage": 17.77, "elapsed_time": "7:45:12", "remaining_time": "1 day, 11:53:05", "throughput": "17881.37", "total_tokens": 499122176}
{"current_steps": 477, "total_steps": 2679, "loss": 0.752, "learning_rate": 4.6506188009602885e-06, "epoch": 0.5337062937062937, "percentage": 17.81, "elapsed_time": "7:46:07", "remaining_time": "1 day, 11:51:46", "throughput": "17884.24", "total_tokens": 500170752}
{"current_steps": 478, "total_steps": 2679, "loss": 0.7119, "learning_rate": 4.6491084167195665e-06, "epoch": 0.5348251748251748, "percentage": 17.84, "elapsed_time": "7:47:04", "remaining_time": "1 day, 11:50:40", "throughput": "17885.23", "total_tokens": 501219328}
{"current_steps": 479, "total_steps": 2679, "loss": 0.7091, "learning_rate": 4.647595021163878e-06, "epoch": 0.5359440559440559, "percentage": 17.88, "elapsed_time": "7:48:03", "remaining_time": "1 day, 11:49:46", "throughput": "17884.55", "total_tokens": 502267904}
{"current_steps": 480, "total_steps": 2679, "loss": 0.7446, "learning_rate": 4.646078616413781e-06, "epoch": 0.5370629370629371, "percentage": 17.92, "elapsed_time": "7:49:00", "remaining_time": "1 day, 11:48:37", "throughput": "17886.03", "total_tokens": 503316480}
{"current_steps": 481, "total_steps": 2679, "loss": 0.8345, "learning_rate": 4.6445592045940515e-06, "epoch": 0.5381818181818182, "percentage": 17.95, "elapsed_time": "7:50:00", "remaining_time": "1 day, 11:47:45", "throughput": "17885.04", "total_tokens": 504365056}
{"current_steps": 482, "total_steps": 2679, "loss": 0.8465, "learning_rate": 4.6430367878336795e-06, "epoch": 0.5393006993006993, "percentage": 17.99, "elapsed_time": "7:51:00", "remaining_time": "1 day, 11:46:53", "throughput": "17884.09", "total_tokens": 505413632}
{"current_steps": 483, "total_steps": 2679, "loss": 0.7268, "learning_rate": 4.641511368265861e-06, "epoch": 0.5404195804195804, "percentage": 18.03, "elapsed_time": "7:51:54", "remaining_time": "1 day, 11:45:33", "throughput": "17887.11", "total_tokens": 506462208}
{"current_steps": 484, "total_steps": 2679, "loss": 0.7069, "learning_rate": 4.6399829480280055e-06, "epoch": 0.5415384615384615, "percentage": 18.07, "elapsed_time": "7:52:48", "remaining_time": "1 day, 11:44:14", "throughput": "17890.01", "total_tokens": 507510784}
{"current_steps": 485, "total_steps": 2679, "loss": 0.7307, "learning_rate": 4.6384515292617226e-06, "epoch": 0.5426573426573427, "percentage": 18.1, "elapsed_time": "7:53:41", "remaining_time": "1 day, 11:42:52", "throughput": "17893.25", "total_tokens": 508559360}
{"current_steps": 486, "total_steps": 2679, "loss": 0.6262, "learning_rate": 4.636917114112827e-06, "epoch": 0.5437762237762238, "percentage": 18.14, "elapsed_time": "7:54:29", "remaining_time": "1 day, 11:41:06", "throughput": "17899.82", "total_tokens": 509607936}
{"current_steps": 487, "total_steps": 2679, "loss": 0.7048, "learning_rate": 4.635379704731327e-06, "epoch": 0.5448951048951048, "percentage": 18.18, "elapsed_time": "7:55:27", "remaining_time": "1 day, 11:40:02", "throughput": "17900.54", "total_tokens": 510656512}
{"current_steps": 488, "total_steps": 2679, "loss": 0.7502, "learning_rate": 4.633839303271432e-06, "epoch": 0.546013986013986, "percentage": 18.22, "elapsed_time": "7:56:27", "remaining_time": "1 day, 11:39:10", "throughput": "17899.71", "total_tokens": 511705088}
{"current_steps": 489, "total_steps": 2679, "loss": 0.802, "learning_rate": 4.63229591189154e-06, "epoch": 0.5471328671328671, "percentage": 18.25, "elapsed_time": "7:57:32", "remaining_time": "1 day, 11:38:42", "throughput": "17895.34", "total_tokens": 512753664}
{"current_steps": 490, "total_steps": 2679, "loss": 0.7881, "learning_rate": 4.63074953275424e-06, "epoch": 0.5482517482517483, "percentage": 18.29, "elapsed_time": "7:58:37", "remaining_time": "1 day, 11:38:12", "throughput": "17891.37", "total_tokens": 513802240}
{"current_steps": 491, "total_steps": 2679, "loss": 0.6901, "learning_rate": 4.629200168026307e-06, "epoch": 0.5493706293706294, "percentage": 18.33, "elapsed_time": "7:59:26", "remaining_time": "1 day, 11:36:29", "throughput": "17897.60", "total_tokens": 514850816}
{"current_steps": 492, "total_steps": 2679, "loss": 0.7506, "learning_rate": 4.6276478198787004e-06, "epoch": 0.5504895104895104, "percentage": 18.37, "elapsed_time": "8:00:26", "remaining_time": "1 day, 11:35:35", "throughput": "17896.93", "total_tokens": 515899392}
{"current_steps": 493, "total_steps": 2679, "loss": 0.8355, "learning_rate": 4.626092490486557e-06, "epoch": 0.5516083916083916, "percentage": 18.4, "elapsed_time": "8:01:32", "remaining_time": "1 day, 11:35:10", "throughput": "17892.26", "total_tokens": 516947968}
{"current_steps": 494, "total_steps": 2679, "loss": 0.8243, "learning_rate": 4.624534182029195e-06, "epoch": 0.5527272727272727, "percentage": 18.44, "elapsed_time": "8:02:37", "remaining_time": "1 day, 11:34:42", "throughput": "17888.05", "total_tokens": 517996544}
{"current_steps": 495, "total_steps": 2679, "loss": 0.8418, "learning_rate": 4.6229728966901036e-06, "epoch": 0.5538461538461539, "percentage": 18.48, "elapsed_time": "8:03:44", "remaining_time": "1 day, 11:34:21", "throughput": "17882.71", "total_tokens": 519045120}
{"current_steps": 496, "total_steps": 2679, "loss": 0.8343, "learning_rate": 4.621408636656944e-06, "epoch": 0.554965034965035, "percentage": 18.51, "elapsed_time": "8:04:44", "remaining_time": "1 day, 11:33:26", "throughput": "17882.21", "total_tokens": 520093696}
{"current_steps": 497, "total_steps": 2679, "loss": 0.6521, "learning_rate": 4.6198414041215484e-06, "epoch": 0.556083916083916, "percentage": 18.55, "elapsed_time": "8:05:32", "remaining_time": "1 day, 11:31:41", "throughput": "17888.75", "total_tokens": 521142272}
{"current_steps": 498, "total_steps": 2679, "loss": 0.7437, "learning_rate": 4.618271201279908e-06, "epoch": 0.5572027972027972, "percentage": 18.59, "elapsed_time": "8:06:27", "remaining_time": "1 day, 11:30:27", "throughput": "17890.87", "total_tokens": 522190848}
{"current_steps": 499, "total_steps": 2679, "loss": 0.7586, "learning_rate": 4.616698030332183e-06, "epoch": 0.5583216783216783, "percentage": 18.63, "elapsed_time": "8:07:21", "remaining_time": "1 day, 11:29:06", "throughput": "17893.98", "total_tokens": 523239424}
{"current_steps": 500, "total_steps": 2679, "loss": 0.6871, "learning_rate": 4.6151218934826866e-06, "epoch": 0.5594405594405595, "percentage": 18.66, "elapsed_time": "8:08:09", "remaining_time": "1 day, 11:27:24", "throughput": "17900.13", "total_tokens": 524288000}
{"current_steps": 500, "total_steps": 2679, "eval_loss": 0.7598036527633667, "epoch": 0.5594405594405595, "percentage": 18.66, "elapsed_time": "8:12:16", "remaining_time": "1 day, 11:45:20", "throughput": "17750.46", "total_tokens": 524288000}
{"current_steps": 501, "total_steps": 2679, "loss": 0.8157, "learning_rate": 4.613542792939891e-06, "epoch": 0.5605594405594405, "percentage": 18.7, "elapsed_time": "8:13:50", "remaining_time": "1 day, 11:46:51", "throughput": "17729.79", "total_tokens": 525336576}
{"current_steps": 502, "total_steps": 2679, "loss": 0.6763, "learning_rate": 4.611960730916419e-06, "epoch": 0.5616783216783217, "percentage": 18.74, "elapsed_time": "8:14:39", "remaining_time": "1 day, 11:45:07", "throughput": "17735.93", "total_tokens": 526385152}
{"current_steps": 503, "total_steps": 2679, "loss": 0.7936, "learning_rate": 4.610375709629047e-06, "epoch": 0.5627972027972028, "percentage": 18.78, "elapsed_time": "8:15:44", "remaining_time": "1 day, 11:44:36", "throughput": "17732.02", "total_tokens": 527433728}
{"current_steps": 504, "total_steps": 2679, "loss": 0.6723, "learning_rate": 4.608787731298694e-06, "epoch": 0.5639160839160839, "percentage": 18.81, "elapsed_time": "8:16:38", "remaining_time": "1 day, 11:43:16", "throughput": "17734.97", "total_tokens": 528482304}
{"current_steps": 505, "total_steps": 2679, "loss": 0.7054, "learning_rate": 4.607196798150423e-06, "epoch": 0.5650349650349651, "percentage": 18.85, "elapsed_time": "8:17:32", "remaining_time": "1 day, 11:41:54", "throughput": "17738.12", "total_tokens": 529530880}
{"current_steps": 506, "total_steps": 2679, "loss": 0.7401, "learning_rate": 4.605602912413437e-06, "epoch": 0.5661538461538461, "percentage": 18.89, "elapsed_time": "8:18:26", "remaining_time": "1 day, 11:40:34", "throughput": "17741.05", "total_tokens": 530579456}
{"current_steps": 507, "total_steps": 2679, "loss": 0.7786, "learning_rate": 4.60400607632108e-06, "epoch": 0.5672727272727273, "percentage": 18.92, "elapsed_time": "8:19:26", "remaining_time": "1 day, 11:39:36", "throughput": "17740.82", "total_tokens": 531628032}
{"current_steps": 508, "total_steps": 2679, "loss": 0.6818, "learning_rate": 4.602406292110825e-06, "epoch": 0.5683916083916084, "percentage": 18.96, "elapsed_time": "8:20:15", "remaining_time": "1 day, 11:37:52", "throughput": "17747.00", "total_tokens": 532676608}
{"current_steps": 509, "total_steps": 2679, "loss": 0.8581, "learning_rate": 4.600803562024277e-06, "epoch": 0.5695104895104895, "percentage": 19.0, "elapsed_time": "8:21:20", "remaining_time": "1 day, 11:37:22", "throughput": "17743.02", "total_tokens": 533725184}
{"current_steps": 510, "total_steps": 2679, "loss": 0.759, "learning_rate": 4.599197888307172e-06, "epoch": 0.5706293706293706, "percentage": 19.04, "elapsed_time": "8:22:20", "remaining_time": "1 day, 11:36:25", "throughput": "17742.71", "total_tokens": 534773760}
{"current_steps": 511, "total_steps": 2679, "loss": 0.7531, "learning_rate": 4.597589273209366e-06, "epoch": 0.5717482517482517, "percentage": 19.07, "elapsed_time": "8:23:13", "remaining_time": "1 day, 11:35:02", "throughput": "17746.06", "total_tokens": 535822336}
{"current_steps": 512, "total_steps": 2679, "loss": 0.8437, "learning_rate": 4.5959777189848405e-06, "epoch": 0.5728671328671329, "percentage": 19.11, "elapsed_time": "8:24:19", "remaining_time": "1 day, 11:34:29", "throughput": "17742.44", "total_tokens": 536870912}
{"current_steps": 513, "total_steps": 2679, "loss": 0.7465, "learning_rate": 4.594363227891693e-06, "epoch": 0.573986013986014, "percentage": 19.15, "elapsed_time": "8:25:18", "remaining_time": "1 day, 11:33:33", "throughput": "17742.03", "total_tokens": 537919488}
{"current_steps": 514, "total_steps": 2679, "loss": 0.6948, "learning_rate": 4.592745802192136e-06, "epoch": 0.5751048951048952, "percentage": 19.19, "elapsed_time": "8:26:12", "remaining_time": "1 day, 11:32:12", "throughput": "17745.03", "total_tokens": 538968064}
{"current_steps": 515, "total_steps": 2679, "loss": 0.7117, "learning_rate": 4.591125444152495e-06, "epoch": 0.5762237762237762, "percentage": 19.22, "elapsed_time": "8:27:07", "remaining_time": "1 day, 11:30:54", "throughput": "17747.71", "total_tokens": 540016640}
{"current_steps": 516, "total_steps": 2679, "loss": 0.8269, "learning_rate": 4.589502156043203e-06, "epoch": 0.5773426573426573, "percentage": 19.26, "elapsed_time": "8:28:11", "remaining_time": "1 day, 11:30:17", "throughput": "17744.53", "total_tokens": 541065216}
{"current_steps": 517, "total_steps": 2679, "loss": 0.6603, "learning_rate": 4.587875940138801e-06, "epoch": 0.5784615384615385, "percentage": 19.3, "elapsed_time": "8:29:00", "remaining_time": "1 day, 11:28:33", "throughput": "17750.82", "total_tokens": 542113792}
{"current_steps": 518, "total_steps": 2679, "loss": 0.6775, "learning_rate": 4.58624679871793e-06, "epoch": 0.5795804195804196, "percentage": 19.34, "elapsed_time": "8:29:54", "remaining_time": "1 day, 11:27:12", "throughput": "17753.85", "total_tokens": 543162368}
{"current_steps": 519, "total_steps": 2679, "loss": 0.7769, "learning_rate": 4.5846147340633305e-06, "epoch": 0.5806993006993006, "percentage": 19.37, "elapsed_time": "8:30:48", "remaining_time": "1 day, 11:25:53", "throughput": "17756.72", "total_tokens": 544210944}
{"current_steps": 520, "total_steps": 2679, "loss": 0.6541, "learning_rate": 4.58297974846184e-06, "epoch": 0.5818181818181818, "percentage": 19.41, "elapsed_time": "8:31:41", "remaining_time": "1 day, 11:24:31", "throughput": "17759.81", "total_tokens": 545259520}
{"current_steps": 521, "total_steps": 2679, "loss": 0.7022, "learning_rate": 4.5813418442043885e-06, "epoch": 0.5829370629370629, "percentage": 19.45, "elapsed_time": "8:32:35", "remaining_time": "1 day, 11:23:09", "throughput": "17763.05", "total_tokens": 546308096}
{"current_steps": 522, "total_steps": 2679, "loss": 0.7934, "learning_rate": 4.579701023585995e-06, "epoch": 0.5840559440559441, "percentage": 19.48, "elapsed_time": "8:33:40", "remaining_time": "1 day, 11:22:36", "throughput": "17759.42", "total_tokens": 547356672}
{"current_steps": 523, "total_steps": 2679, "loss": 0.7407, "learning_rate": 4.578057288905766e-06, "epoch": 0.5851748251748252, "percentage": 19.52, "elapsed_time": "8:34:34", "remaining_time": "1 day, 11:21:15", "throughput": "17762.53", "total_tokens": 548405248}
{"current_steps": 524, "total_steps": 2679, "loss": 0.805, "learning_rate": 4.57641064246689e-06, "epoch": 0.5862937062937063, "percentage": 19.56, "elapsed_time": "8:35:34", "remaining_time": "1 day, 11:20:21", "throughput": "17761.84", "total_tokens": 549453824}
{"current_steps": 525, "total_steps": 2679, "loss": 0.7296, "learning_rate": 4.574761086576635e-06, "epoch": 0.5874125874125874, "percentage": 19.6, "elapsed_time": "8:36:28", "remaining_time": "1 day, 11:19:02", "throughput": "17764.64", "total_tokens": 550502400}
{"current_steps": 526, "total_steps": 2679, "loss": 0.6807, "learning_rate": 4.573108623546348e-06, "epoch": 0.5885314685314685, "percentage": 19.63, "elapsed_time": "8:37:23", "remaining_time": "1 day, 11:17:47", "throughput": "17766.82", "total_tokens": 551550976}
{"current_steps": 527, "total_steps": 2679, "loss": 0.7452, "learning_rate": 4.571453255691449e-06, "epoch": 0.5896503496503497, "percentage": 19.67, "elapsed_time": "8:38:17", "remaining_time": "1 day, 11:16:28", "throughput": "17769.64", "total_tokens": 552599552}
{"current_steps": 528, "total_steps": 2679, "loss": 0.7424, "learning_rate": 4.569794985331425e-06, "epoch": 0.5907692307692308, "percentage": 19.71, "elapsed_time": "8:39:17", "remaining_time": "1 day, 11:15:32", "throughput": "17769.21", "total_tokens": 553648128}
{"current_steps": 529, "total_steps": 2679, "loss": 0.6451, "learning_rate": 4.568133814789833e-06, "epoch": 0.5918881118881119, "percentage": 19.75, "elapsed_time": "8:40:06", "remaining_time": "1 day, 11:13:51", "throughput": "17775.12", "total_tokens": 554696704}
{"current_steps": 530, "total_steps": 2679, "loss": 0.7206, "learning_rate": 4.566469746394294e-06, "epoch": 0.593006993006993, "percentage": 19.78, "elapsed_time": "8:41:00", "remaining_time": "1 day, 11:12:33", "throughput": "17777.68", "total_tokens": 555745280}
{"current_steps": 531, "total_steps": 2679, "loss": 0.8261, "learning_rate": 4.564802782476487e-06, "epoch": 0.5941258741258741, "percentage": 19.82, "elapsed_time": "8:42:01", "remaining_time": "1 day, 11:11:40", "throughput": "17776.95", "total_tokens": 556793856}
{"current_steps": 532, "total_steps": 2679, "loss": 0.8197, "learning_rate": 4.563132925372151e-06, "epoch": 0.5952447552447553, "percentage": 19.86, "elapsed_time": "8:43:00", "remaining_time": "1 day, 11:10:44", "throughput": "17776.52", "total_tokens": 557842432}
{"current_steps": 533, "total_steps": 2679, "loss": 0.7082, "learning_rate": 4.561460177421078e-06, "epoch": 0.5963636363636363, "percentage": 19.9, "elapsed_time": "8:43:55", "remaining_time": "1 day, 11:09:25", "throughput": "17779.23", "total_tokens": 558891008}
{"current_steps": 534, "total_steps": 2679, "loss": 0.6756, "learning_rate": 4.559784540967107e-06, "epoch": 0.5974825174825175, "percentage": 19.93, "elapsed_time": "8:44:49", "remaining_time": "1 day, 11:08:07", "throughput": "17781.95", "total_tokens": 559939584}
{"current_steps": 535, "total_steps": 2679, "loss": 0.6532, "learning_rate": 4.558106018358131e-06, "epoch": 0.5986013986013986, "percentage": 19.97, "elapsed_time": "8:45:37", "remaining_time": "1 day, 11:06:23", "throughput": "17788.25", "total_tokens": 560988160}
{"current_steps": 536, "total_steps": 2679, "loss": 0.7283, "learning_rate": 4.5564246119460805e-06, "epoch": 0.5997202797202797, "percentage": 20.01, "elapsed_time": "8:46:34", "remaining_time": "1 day, 11:05:20", "throughput": "17788.90", "total_tokens": 562036736}
{"current_steps": 537, "total_steps": 2679, "loss": 0.739, "learning_rate": 4.5547403240869335e-06, "epoch": 0.6008391608391609, "percentage": 20.04, "elapsed_time": "8:47:28", "remaining_time": "1 day, 11:04:01", "throughput": "17791.71", "total_tokens": 563085312}
{"current_steps": 538, "total_steps": 2679, "loss": 0.654, "learning_rate": 4.553053157140699e-06, "epoch": 0.6019580419580419, "percentage": 20.08, "elapsed_time": "8:48:17", "remaining_time": "1 day, 11:02:21", "throughput": "17797.42", "total_tokens": 564133888}
{"current_steps": 539, "total_steps": 2679, "loss": 0.7559, "learning_rate": 4.5513631134714235e-06, "epoch": 0.6030769230769231, "percentage": 20.12, "elapsed_time": "8:49:17", "remaining_time": "1 day, 11:01:25", "throughput": "17797.04", "total_tokens": 565182464}
{"current_steps": 540, "total_steps": 2679, "loss": 0.8768, "learning_rate": 4.5496701954471845e-06, "epoch": 0.6041958041958042, "percentage": 20.16, "elapsed_time": "8:50:22", "remaining_time": "1 day, 11:00:52", "throughput": "17793.37", "total_tokens": 566231040}
{"current_steps": 541, "total_steps": 2679, "loss": 0.7983, "learning_rate": 4.547974405440085e-06, "epoch": 0.6053146853146854, "percentage": 20.19, "elapsed_time": "8:51:16", "remaining_time": "1 day, 10:59:34", "throughput": "17796.11", "total_tokens": 567279616}
{"current_steps": 542, "total_steps": 2679, "loss": 0.7015, "learning_rate": 4.5462757458262565e-06, "epoch": 0.6064335664335664, "percentage": 20.23, "elapsed_time": "8:52:10", "remaining_time": "1 day, 10:58:17", "throughput": "17798.66", "total_tokens": 568328192}
{"current_steps": 543, "total_steps": 2679, "loss": 0.949, "learning_rate": 4.544574218985845e-06, "epoch": 0.6075524475524475, "percentage": 20.27, "elapsed_time": "8:53:22", "remaining_time": "1 day, 10:58:07", "throughput": "17791.70", "total_tokens": 569376768}
{"current_steps": 544, "total_steps": 2679, "loss": 0.6954, "learning_rate": 4.5428698273030185e-06, "epoch": 0.6086713286713287, "percentage": 20.31, "elapsed_time": "8:54:16", "remaining_time": "1 day, 10:56:50", "throughput": "17794.26", "total_tokens": 570425344}
{"current_steps": 545, "total_steps": 2679, "loss": 0.6441, "learning_rate": 4.5411625731659595e-06, "epoch": 0.6097902097902098, "percentage": 20.34, "elapsed_time": "8:55:04", "remaining_time": "1 day, 10:55:09", "throughput": "17800.30", "total_tokens": 571473920}
{"current_steps": 546, "total_steps": 2679, "loss": 0.8815, "learning_rate": 4.539452458966857e-06, "epoch": 0.610909090909091, "percentage": 20.38, "elapsed_time": "8:56:05", "remaining_time": "1 day, 10:54:16", "throughput": "17799.46", "total_tokens": 572522496}
{"current_steps": 547, "total_steps": 2679, "loss": 0.8271, "learning_rate": 4.5377394871019145e-06, "epoch": 0.612027972027972, "percentage": 20.42, "elapsed_time": "8:57:11", "remaining_time": "1 day, 10:53:44", "throughput": "17795.54", "total_tokens": 573571072}
{"current_steps": 548, "total_steps": 2679, "loss": 0.6533, "learning_rate": 4.536023659971332e-06, "epoch": 0.6131468531468531, "percentage": 20.46, "elapsed_time": "8:57:59", "remaining_time": "1 day, 10:52:03", "throughput": "17801.50", "total_tokens": 574619648}
{"current_steps": 549, "total_steps": 2679, "loss": 0.7459, "learning_rate": 4.534304979979317e-06, "epoch": 0.6142657342657343, "percentage": 20.49, "elapsed_time": "8:59:01", "remaining_time": "1 day, 10:51:16", "throughput": "17799.89", "total_tokens": 575668224}
{"current_steps": 550, "total_steps": 2679, "loss": 0.7696, "learning_rate": 4.5325834495340695e-06, "epoch": 0.6153846153846154, "percentage": 20.53, "elapsed_time": "9:00:00", "remaining_time": "1 day, 10:50:19", "throughput": "17799.59", "total_tokens": 576716800}
{"current_steps": 551, "total_steps": 2679, "loss": 0.9027, "learning_rate": 4.530859071047785e-06, "epoch": 0.6165034965034965, "percentage": 20.57, "elapsed_time": "9:01:01", "remaining_time": "1 day, 10:49:28", "throughput": "17798.50", "total_tokens": 577765376}
{"current_steps": 552, "total_steps": 2679, "loss": 0.7433, "learning_rate": 4.529131846936651e-06, "epoch": 0.6176223776223776, "percentage": 20.6, "elapsed_time": "9:01:56", "remaining_time": "1 day, 10:48:12", "throughput": "17800.86", "total_tokens": 578813952}
{"current_steps": 553, "total_steps": 2679, "loss": 0.8272, "learning_rate": 4.52740177962084e-06, "epoch": 0.6187412587412587, "percentage": 20.64, "elapsed_time": "9:02:55", "remaining_time": "1 day, 10:47:17", "throughput": "17800.31", "total_tokens": 579862528}
{"current_steps": 554, "total_steps": 2679, "loss": 0.7529, "learning_rate": 4.525668871524512e-06, "epoch": 0.6198601398601399, "percentage": 20.68, "elapsed_time": "9:03:56", "remaining_time": "1 day, 10:46:25", "throughput": "17799.36", "total_tokens": 580911104}
{"current_steps": 555, "total_steps": 2679, "loss": 0.8129, "learning_rate": 4.5239331250758025e-06, "epoch": 0.620979020979021, "percentage": 20.72, "elapsed_time": "9:04:56", "remaining_time": "1 day, 10:45:31", "throughput": "17798.66", "total_tokens": 581959680}
{"current_steps": 556, "total_steps": 2679, "loss": 0.7073, "learning_rate": 4.522194542706828e-06, "epoch": 0.6220979020979021, "percentage": 20.75, "elapsed_time": "9:05:51", "remaining_time": "1 day, 10:44:15", "throughput": "17801.16", "total_tokens": 583008256}
{"current_steps": 557, "total_steps": 2679, "loss": 0.8055, "learning_rate": 4.520453126853677e-06, "epoch": 0.6232167832167832, "percentage": 20.79, "elapsed_time": "9:06:56", "remaining_time": "1 day, 10:43:41", "throughput": "17797.56", "total_tokens": 584056832}
{"current_steps": 558, "total_steps": 2679, "loss": 0.6958, "learning_rate": 4.51870887995641e-06, "epoch": 0.6243356643356643, "percentage": 20.83, "elapsed_time": "9:07:49", "remaining_time": "1 day, 10:42:20", "throughput": "17800.69", "total_tokens": 585105408}
{"current_steps": 559, "total_steps": 2679, "loss": 0.8523, "learning_rate": 4.51696180445905e-06, "epoch": 0.6254545454545455, "percentage": 20.87, "elapsed_time": "9:08:53", "remaining_time": "1 day, 10:41:41", "throughput": "17797.88", "total_tokens": 586153984}
{"current_steps": 560, "total_steps": 2679, "loss": 0.782, "learning_rate": 4.51521190280959e-06, "epoch": 0.6265734265734266, "percentage": 20.9, "elapsed_time": "9:09:52", "remaining_time": "1 day, 10:40:39", "throughput": "17798.33", "total_tokens": 587202560}
{"current_steps": 561, "total_steps": 2679, "loss": 0.7491, "learning_rate": 4.513459177459977e-06, "epoch": 0.6276923076923077, "percentage": 20.94, "elapsed_time": "9:10:51", "remaining_time": "1 day, 10:39:41", "throughput": "17798.11", "total_tokens": 588251136}
{"current_steps": 562, "total_steps": 2679, "loss": 0.7499, "learning_rate": 4.511703630866118e-06, "epoch": 0.6288111888111888, "percentage": 20.98, "elapsed_time": "9:11:47", "remaining_time": "1 day, 10:38:34", "throughput": "17799.35", "total_tokens": 589299712}
{"current_steps": 563, "total_steps": 2679, "loss": 0.8599, "learning_rate": 4.509945265487871e-06, "epoch": 0.62993006993007, "percentage": 21.02, "elapsed_time": "9:12:48", "remaining_time": "1 day, 10:37:40", "throughput": "17798.58", "total_tokens": 590348288}
{"current_steps": 564, "total_steps": 2679, "loss": 0.6938, "learning_rate": 4.5081840837890445e-06, "epoch": 0.6310489510489511, "percentage": 21.05, "elapsed_time": "9:13:42", "remaining_time": "1 day, 10:36:22", "throughput": "17801.33", "total_tokens": 591396864}
{"current_steps": 565, "total_steps": 2679, "loss": 0.783, "learning_rate": 4.506420088237395e-06, "epoch": 0.6321678321678321, "percentage": 21.09, "elapsed_time": "9:14:41", "remaining_time": "1 day, 10:35:25", "throughput": "17801.07", "total_tokens": 592445440}
{"current_steps": 566, "total_steps": 2679, "loss": 0.7826, "learning_rate": 4.504653281304619e-06, "epoch": 0.6332867132867133, "percentage": 21.13, "elapsed_time": "9:15:44", "remaining_time": "1 day, 10:34:41", "throughput": "17798.94", "total_tokens": 593494016}
{"current_steps": 567, "total_steps": 2679, "loss": 0.7171, "learning_rate": 4.5028836654663535e-06, "epoch": 0.6344055944055944, "percentage": 21.16, "elapsed_time": "9:16:38", "remaining_time": "1 day, 10:33:23", "throughput": "17801.66", "total_tokens": 594542592}
{"current_steps": 568, "total_steps": 2679, "loss": 0.6844, "learning_rate": 4.501111243202172e-06, "epoch": 0.6355244755244756, "percentage": 21.2, "elapsed_time": "9:17:32", "remaining_time": "1 day, 10:32:06", "throughput": "17804.34", "total_tokens": 595591168}
{"current_steps": 569, "total_steps": 2679, "loss": 0.7739, "learning_rate": 4.4993360169955784e-06, "epoch": 0.6366433566433567, "percentage": 21.24, "elapsed_time": "9:18:31", "remaining_time": "1 day, 10:31:10", "throughput": "17803.83", "total_tokens": 596639744}
{"current_steps": 570, "total_steps": 2679, "loss": 0.7509, "learning_rate": 4.497557989334008e-06, "epoch": 0.6377622377622377, "percentage": 21.28, "elapsed_time": "9:19:28", "remaining_time": "1 day, 10:30:04", "throughput": "17804.86", "total_tokens": 597688320}
{"current_steps": 571, "total_steps": 2679, "loss": 0.7863, "learning_rate": 4.4957771627088185e-06, "epoch": 0.6388811188811189, "percentage": 21.31, "elapsed_time": "9:20:34", "remaining_time": "1 day, 10:29:30", "throughput": "17801.34", "total_tokens": 598736896}
{"current_steps": 572, "total_steps": 2679, "loss": 0.7671, "learning_rate": 4.493993539615293e-06, "epoch": 0.64, "percentage": 21.35, "elapsed_time": "9:21:29", "remaining_time": "1 day, 10:28:15", "throughput": "17803.58", "total_tokens": 599785472}
{"current_steps": 573, "total_steps": 2679, "loss": 0.7251, "learning_rate": 4.492207122552629e-06, "epoch": 0.6411188811188812, "percentage": 21.39, "elapsed_time": "9:22:22", "remaining_time": "1 day, 10:26:58", "throughput": "17806.19", "total_tokens": 600834048}
{"current_steps": 574, "total_steps": 2679, "loss": 0.7206, "learning_rate": 4.490417914023944e-06, "epoch": 0.6422377622377622, "percentage": 21.43, "elapsed_time": "9:23:12", "remaining_time": "1 day, 10:25:24", "throughput": "17811.25", "total_tokens": 601882624}
{"current_steps": 575, "total_steps": 2679, "loss": 0.6965, "learning_rate": 4.48862591653626e-06, "epoch": 0.6433566433566433, "percentage": 21.46, "elapsed_time": "9:24:06", "remaining_time": "1 day, 10:24:09", "throughput": "17813.55", "total_tokens": 602931200}
{"current_steps": 576, "total_steps": 2679, "loss": 0.7772, "learning_rate": 4.486831132600513e-06, "epoch": 0.6444755244755245, "percentage": 21.5, "elapsed_time": "9:25:07", "remaining_time": "1 day, 10:23:16", "throughput": "17812.74", "total_tokens": 603979776}
{"current_steps": 577, "total_steps": 2679, "loss": 0.6837, "learning_rate": 4.485033564731542e-06, "epoch": 0.6455944055944056, "percentage": 21.54, "elapsed_time": "9:26:00", "remaining_time": "1 day, 10:21:57", "throughput": "17815.63", "total_tokens": 605028352}
{"current_steps": 578, "total_steps": 2679, "loss": 0.7357, "learning_rate": 4.483233215448084e-06, "epoch": 0.6467132867132868, "percentage": 21.58, "elapsed_time": "9:26:54", "remaining_time": "1 day, 10:20:40", "throughput": "17818.26", "total_tokens": 606076928}
{"current_steps": 579, "total_steps": 2679, "loss": 0.6863, "learning_rate": 4.481430087272776e-06, "epoch": 0.6478321678321678, "percentage": 21.61, "elapsed_time": "9:27:49", "remaining_time": "1 day, 10:19:26", "throughput": "17820.41", "total_tokens": 607125504}
{"current_steps": 580, "total_steps": 2679, "loss": 0.7333, "learning_rate": 4.479624182732148e-06, "epoch": 0.6489510489510489, "percentage": 21.65, "elapsed_time": "9:28:48", "remaining_time": "1 day, 10:18:31", "throughput": "17819.96", "total_tokens": 608174080}
{"current_steps": 581, "total_steps": 2679, "loss": 0.8355, "learning_rate": 4.47781550435662e-06, "epoch": 0.6500699300699301, "percentage": 21.69, "elapsed_time": "9:29:48", "remaining_time": "1 day, 10:17:36", "throughput": "17819.34", "total_tokens": 609222656}
{"current_steps": 582, "total_steps": 2679, "loss": 0.8488, "learning_rate": 4.476004054680501e-06, "epoch": 0.6511888111888112, "percentage": 21.72, "elapsed_time": "9:30:49", "remaining_time": "1 day, 10:16:43", "throughput": "17818.51", "total_tokens": 610271232}
{"current_steps": 583, "total_steps": 2679, "loss": 0.7343, "learning_rate": 4.474189836241976e-06, "epoch": 0.6523076923076923, "percentage": 21.76, "elapsed_time": "9:31:49", "remaining_time": "1 day, 10:15:48", "throughput": "17817.92", "total_tokens": 611319808}
{"current_steps": 584, "total_steps": 2679, "loss": 0.8272, "learning_rate": 4.472372851583121e-06, "epoch": 0.6534265734265734, "percentage": 21.8, "elapsed_time": "9:32:55", "remaining_time": "1 day, 10:15:16", "throughput": "17814.10", "total_tokens": 612368384}
{"current_steps": 585, "total_steps": 2679, "loss": 0.7532, "learning_rate": 4.470553103249876e-06, "epoch": 0.6545454545454545, "percentage": 21.84, "elapsed_time": "9:33:56", "remaining_time": "1 day, 10:14:26", "throughput": "17812.84", "total_tokens": 613416960}
{"current_steps": 586, "total_steps": 2679, "loss": 0.7231, "learning_rate": 4.468730593792062e-06, "epoch": 0.6556643356643357, "percentage": 21.87, "elapsed_time": "9:34:56", "remaining_time": "1 day, 10:13:30", "throughput": "17812.38", "total_tokens": 614465536}
{"current_steps": 587, "total_steps": 2679, "loss": 0.8676, "learning_rate": 4.466905325763365e-06, "epoch": 0.6567832167832168, "percentage": 21.91, "elapsed_time": "9:36:02", "remaining_time": "1 day, 10:12:55", "throughput": "17808.96", "total_tokens": 615514112}
{"current_steps": 588, "total_steps": 2679, "loss": 0.6988, "learning_rate": 4.465077301721338e-06, "epoch": 0.6579020979020979, "percentage": 21.95, "elapsed_time": "9:36:56", "remaining_time": "1 day, 10:11:38", "throughput": "17811.49", "total_tokens": 616562688}
{"current_steps": 589, "total_steps": 2679, "loss": 0.7208, "learning_rate": 4.463246524227393e-06, "epoch": 0.659020979020979, "percentage": 21.99, "elapsed_time": "9:37:50", "remaining_time": "1 day, 10:10:24", "throughput": "17813.71", "total_tokens": 617611264}
{"current_steps": 590, "total_steps": 2679, "loss": 0.7239, "learning_rate": 4.461412995846804e-06, "epoch": 0.6601398601398601, "percentage": 22.02, "elapsed_time": "9:38:45", "remaining_time": "1 day, 10:09:11", "throughput": "17815.71", "total_tokens": 618659840}
{"current_steps": 591, "total_steps": 2679, "loss": 0.6923, "learning_rate": 4.459576719148697e-06, "epoch": 0.6612587412587413, "percentage": 22.06, "elapsed_time": "9:39:39", "remaining_time": "1 day, 10:07:56", "throughput": "17818.07", "total_tokens": 619708416}
{"current_steps": 592, "total_steps": 2679, "loss": 0.7647, "learning_rate": 4.4577376967060495e-06, "epoch": 0.6623776223776224, "percentage": 22.1, "elapsed_time": "9:40:39", "remaining_time": "1 day, 10:07:00", "throughput": "17817.76", "total_tokens": 620756992}
{"current_steps": 593, "total_steps": 2679, "loss": 0.844, "learning_rate": 4.4558959310956865e-06, "epoch": 0.6634965034965035, "percentage": 22.14, "elapsed_time": "9:41:47", "remaining_time": "1 day, 10:06:33", "throughput": "17813.05", "total_tokens": 621805568}
{"current_steps": 594, "total_steps": 2679, "loss": 0.7201, "learning_rate": 4.4540514248982744e-06, "epoch": 0.6646153846153846, "percentage": 22.17, "elapsed_time": "9:42:42", "remaining_time": "1 day, 10:05:22", "throughput": "17814.82", "total_tokens": 622854144}
{"current_steps": 595, "total_steps": 2679, "loss": 0.7064, "learning_rate": 4.452204180698325e-06, "epoch": 0.6657342657342658, "percentage": 22.21, "elapsed_time": "9:43:37", "remaining_time": "1 day, 10:04:10", "throughput": "17816.77", "total_tokens": 623902720}
{"current_steps": 596, "total_steps": 2679, "loss": 0.7925, "learning_rate": 4.450354201084181e-06, "epoch": 0.6668531468531469, "percentage": 22.25, "elapsed_time": "9:44:37", "remaining_time": "1 day, 10:03:14", "throughput": "17816.32", "total_tokens": 624951296}
{"current_steps": 597, "total_steps": 2679, "loss": 0.7738, "learning_rate": 4.448501488648021e-06, "epoch": 0.6679720279720279, "percentage": 22.28, "elapsed_time": "9:45:37", "remaining_time": "1 day, 10:02:19", "throughput": "17815.79", "total_tokens": 625999872}
{"current_steps": 598, "total_steps": 2679, "loss": 0.7432, "learning_rate": 4.446646045985852e-06, "epoch": 0.6690909090909091, "percentage": 22.32, "elapsed_time": "9:46:35", "remaining_time": "1 day, 10:01:18", "throughput": "17816.12", "total_tokens": 627048448}
{"current_steps": 599, "total_steps": 2679, "loss": 0.6673, "learning_rate": 4.4447878756975074e-06, "epoch": 0.6702097902097902, "percentage": 22.36, "elapsed_time": "9:47:23", "remaining_time": "1 day, 9:59:41", "throughput": "17821.63", "total_tokens": 628097024}
{"current_steps": 600, "total_steps": 2679, "loss": 0.7206, "learning_rate": 4.44292698038664e-06, "epoch": 0.6713286713286714, "percentage": 22.4, "elapsed_time": "9:48:24", "remaining_time": "1 day, 9:58:48", "throughput": "17820.78", "total_tokens": 629145600}
{"current_steps": 601, "total_steps": 2679, "loss": 0.7508, "learning_rate": 4.441063362660726e-06, "epoch": 0.6724475524475525, "percentage": 22.43, "elapsed_time": "9:49:17", "remaining_time": "1 day, 9:57:32", "throughput": "17823.33", "total_tokens": 630194176}
{"current_steps": 602, "total_steps": 2679, "loss": 0.7696, "learning_rate": 4.439197025131049e-06, "epoch": 0.6735664335664335, "percentage": 22.47, "elapsed_time": "9:50:22", "remaining_time": "1 day, 9:56:54", "throughput": "17820.21", "total_tokens": 631242752}
{"current_steps": 603, "total_steps": 2679, "loss": 0.6821, "learning_rate": 4.4373279704127095e-06, "epoch": 0.6746853146853147, "percentage": 22.51, "elapsed_time": "9:51:16", "remaining_time": "1 day, 9:55:36", "throughput": "17823.01", "total_tokens": 632291328}
{"current_steps": 604, "total_steps": 2679, "loss": 0.7345, "learning_rate": 4.435456201124613e-06, "epoch": 0.6758041958041958, "percentage": 22.55, "elapsed_time": "9:52:15", "remaining_time": "1 day, 9:54:38", "throughput": "17822.87", "total_tokens": 633339904}
{"current_steps": 605, "total_steps": 2679, "loss": 0.6975, "learning_rate": 4.433581719889469e-06, "epoch": 0.676923076923077, "percentage": 22.58, "elapsed_time": "9:53:06", "remaining_time": "1 day, 9:53:14", "throughput": "17826.59", "total_tokens": 634388480}
{"current_steps": 606, "total_steps": 2679, "loss": 0.7357, "learning_rate": 4.431704529333787e-06, "epoch": 0.678041958041958, "percentage": 22.62, "elapsed_time": "9:54:06", "remaining_time": "1 day, 9:52:20", "throughput": "17825.88", "total_tokens": 635437056}
{"current_steps": 607, "total_steps": 2679, "loss": 0.8389, "learning_rate": 4.429824632087873e-06, "epoch": 0.6791608391608391, "percentage": 22.66, "elapsed_time": "9:55:07", "remaining_time": "1 day, 9:51:28", "throughput": "17824.86", "total_tokens": 636485632}
{"current_steps": 608, "total_steps": 2679, "loss": 0.738, "learning_rate": 4.427942030785824e-06, "epoch": 0.6802797202797203, "percentage": 22.7, "elapsed_time": "9:56:01", "remaining_time": "1 day, 9:50:12", "throughput": "17827.45", "total_tokens": 637534208}
{"current_steps": 609, "total_steps": 2679, "loss": 0.8876, "learning_rate": 4.426056728065527e-06, "epoch": 0.6813986013986014, "percentage": 22.73, "elapsed_time": "9:57:12", "remaining_time": "1 day, 9:49:55", "throughput": "17821.26", "total_tokens": 638582784}
{"current_steps": 610, "total_steps": 2679, "loss": 0.6831, "learning_rate": 4.424168726568656e-06, "epoch": 0.6825174825174826, "percentage": 22.77, "elapsed_time": "9:58:06", "remaining_time": "1 day, 9:48:41", "throughput": "17823.51", "total_tokens": 639631360}
{"current_steps": 611, "total_steps": 2679, "loss": 0.7425, "learning_rate": 4.422278028940664e-06, "epoch": 0.6836363636363636, "percentage": 22.81, "elapsed_time": "9:59:07", "remaining_time": "1 day, 9:47:47", "throughput": "17822.79", "total_tokens": 640679936}
{"current_steps": 612, "total_steps": 2679, "loss": 0.8177, "learning_rate": 4.420384637830783e-06, "epoch": 0.6847552447552447, "percentage": 22.84, "elapsed_time": "10:00:12", "remaining_time": "1 day, 9:47:10", "throughput": "17819.60", "total_tokens": 641728512}
{"current_steps": 613, "total_steps": 2679, "loss": 0.8866, "learning_rate": 4.418488555892018e-06, "epoch": 0.6858741258741259, "percentage": 22.88, "elapsed_time": "10:01:16", "remaining_time": "1 day, 9:46:27", "throughput": "17817.30", "total_tokens": 642777088}
{"current_steps": 614, "total_steps": 2679, "loss": 0.8667, "learning_rate": 4.4165897857811455e-06, "epoch": 0.686993006993007, "percentage": 22.92, "elapsed_time": "10:02:22", "remaining_time": "1 day, 9:45:54", "throughput": "17813.50", "total_tokens": 643825664}
{"current_steps": 615, "total_steps": 2679, "loss": 0.706, "learning_rate": 4.414688330158709e-06, "epoch": 0.6881118881118881, "percentage": 22.96, "elapsed_time": "10:03:16", "remaining_time": "1 day, 9:44:39", "throughput": "17815.87", "total_tokens": 644874240}
{"current_steps": 616, "total_steps": 2679, "loss": 0.8243, "learning_rate": 4.412784191689013e-06, "epoch": 0.6892307692307692, "percentage": 22.99, "elapsed_time": "10:04:17", "remaining_time": "1 day, 9:43:45", "throughput": "17815.10", "total_tokens": 645922816}
{"current_steps": 617, "total_steps": 2679, "loss": 0.8379, "learning_rate": 4.4108773730401235e-06, "epoch": 0.6903496503496503, "percentage": 23.03, "elapsed_time": "10:05:22", "remaining_time": "1 day, 9:43:08", "throughput": "17811.93", "total_tokens": 646971392}
{"current_steps": 618, "total_steps": 2679, "loss": 0.7371, "learning_rate": 4.40896787688386e-06, "epoch": 0.6914685314685315, "percentage": 23.07, "elapsed_time": "10:06:16", "remaining_time": "1 day, 9:41:54", "throughput": "17814.21", "total_tokens": 648019968}
{"current_steps": 619, "total_steps": 2679, "loss": 0.7052, "learning_rate": 4.407055705895794e-06, "epoch": 0.6925874125874126, "percentage": 23.11, "elapsed_time": "10:07:09", "remaining_time": "1 day, 9:40:35", "throughput": "17817.05", "total_tokens": 649068544}
{"current_steps": 620, "total_steps": 2679, "loss": 0.894, "learning_rate": 4.405140862755247e-06, "epoch": 0.6937062937062937, "percentage": 23.14, "elapsed_time": "10:08:15", "remaining_time": "1 day, 9:39:58", "throughput": "17813.87", "total_tokens": 650117120}
{"current_steps": 621, "total_steps": 2679, "loss": 0.7827, "learning_rate": 4.403223350145283e-06, "epoch": 0.6948251748251748, "percentage": 23.18, "elapsed_time": "10:09:20", "remaining_time": "1 day, 9:39:21", "throughput": "17810.64", "total_tokens": 651165696}
{"current_steps": 622, "total_steps": 2679, "loss": 0.7743, "learning_rate": 4.401303170752705e-06, "epoch": 0.695944055944056, "percentage": 23.22, "elapsed_time": "10:10:20", "remaining_time": "1 day, 9:38:25", "throughput": "17810.25", "total_tokens": 652214272}
{"current_steps": 623, "total_steps": 2679, "loss": 0.7744, "learning_rate": 4.3993803272680555e-06, "epoch": 0.6970629370629371, "percentage": 23.25, "elapsed_time": "10:11:14", "remaining_time": "1 day, 9:37:12", "throughput": "17812.33", "total_tokens": 653262848}
{"current_steps": 624, "total_steps": 2679, "loss": 0.7421, "learning_rate": 4.397454822385608e-06, "epoch": 0.6981818181818182, "percentage": 23.29, "elapsed_time": "10:12:13", "remaining_time": "1 day, 9:36:14", "throughput": "17812.16", "total_tokens": 654311424}
{"current_steps": 625, "total_steps": 2679, "loss": 0.7657, "learning_rate": 4.395526658803367e-06, "epoch": 0.6993006993006993, "percentage": 23.33, "elapsed_time": "10:13:14", "remaining_time": "1 day, 9:35:22", "throughput": "17811.26", "total_tokens": 655360000}
{"current_steps": 626, "total_steps": 2679, "loss": 0.7183, "learning_rate": 4.393595839223061e-06, "epoch": 0.7004195804195804, "percentage": 23.37, "elapsed_time": "10:14:08", "remaining_time": "1 day, 9:34:05", "throughput": "17813.94", "total_tokens": 656408576}
{"current_steps": 627, "total_steps": 2679, "loss": 0.8945, "learning_rate": 4.391662366350139e-06, "epoch": 0.7015384615384616, "percentage": 23.4, "elapsed_time": "10:15:13", "remaining_time": "1 day, 9:33:28", "throughput": "17810.69", "total_tokens": 657457152}
{"current_steps": 628, "total_steps": 2679, "loss": 0.7571, "learning_rate": 4.38972624289377e-06, "epoch": 0.7026573426573427, "percentage": 23.44, "elapsed_time": "10:16:08", "remaining_time": "1 day, 9:32:16", "throughput": "17812.53", "total_tokens": 658505728}
{"current_steps": 629, "total_steps": 2679, "loss": 0.7475, "learning_rate": 4.387787471566837e-06, "epoch": 0.7037762237762237, "percentage": 23.48, "elapsed_time": "10:17:07", "remaining_time": "1 day, 9:31:19", "throughput": "17812.32", "total_tokens": 659554304}
{"current_steps": 630, "total_steps": 2679, "loss": 0.7916, "learning_rate": 4.385846055085929e-06, "epoch": 0.7048951048951049, "percentage": 23.52, "elapsed_time": "10:18:07", "remaining_time": "1 day, 9:30:23", "throughput": "17811.83", "total_tokens": 660602880}
{"current_steps": 631, "total_steps": 2679, "loss": 0.871, "learning_rate": 4.383901996171348e-06, "epoch": 0.706013986013986, "percentage": 23.55, "elapsed_time": "10:19:07", "remaining_time": "1 day, 9:29:28", "throughput": "17811.30", "total_tokens": 661651456}
{"current_steps": 632, "total_steps": 2679, "loss": 0.6305, "learning_rate": 4.381955297547093e-06, "epoch": 0.7071328671328672, "percentage": 23.59, "elapsed_time": "10:19:56", "remaining_time": "1 day, 9:27:57", "throughput": "17816.11", "total_tokens": 662700032}
{"current_steps": 633, "total_steps": 2679, "loss": 0.8249, "learning_rate": 4.380005961940864e-06, "epoch": 0.7082517482517483, "percentage": 23.63, "elapsed_time": "10:20:53", "remaining_time": "1 day, 9:26:51", "throughput": "17817.09", "total_tokens": 663748608}
{"current_steps": 634, "total_steps": 2679, "loss": 0.7488, "learning_rate": 4.378053992084057e-06, "epoch": 0.7093706293706293, "percentage": 23.67, "elapsed_time": "10:21:53", "remaining_time": "1 day, 9:25:55", "throughput": "17816.74", "total_tokens": 664797184}
{"current_steps": 635, "total_steps": 2679, "loss": 0.883, "learning_rate": 4.376099390711758e-06, "epoch": 0.7104895104895105, "percentage": 23.7, "elapsed_time": "10:22:58", "remaining_time": "1 day, 9:25:18", "throughput": "17813.47", "total_tokens": 665845760}
{"current_steps": 636, "total_steps": 2679, "loss": 0.7111, "learning_rate": 4.374142160562738e-06, "epoch": 0.7116083916083916, "percentage": 23.74, "elapsed_time": "10:23:58", "remaining_time": "1 day, 9:24:22", "throughput": "17813.09", "total_tokens": 666894336}
{"current_steps": 637, "total_steps": 2679, "loss": 0.7723, "learning_rate": 4.372182304379455e-06, "epoch": 0.7127272727272728, "percentage": 23.78, "elapsed_time": "10:25:01", "remaining_time": "1 day, 9:23:36", "throughput": "17811.08", "total_tokens": 667942912}
{"current_steps": 638, "total_steps": 2679, "loss": 0.7532, "learning_rate": 4.370219824908045e-06, "epoch": 0.7138461538461538, "percentage": 23.81, "elapsed_time": "10:26:01", "remaining_time": "1 day, 9:22:40", "throughput": "17810.66", "total_tokens": 668991488}
{"current_steps": 639, "total_steps": 2679, "loss": 0.8121, "learning_rate": 4.368254724898319e-06, "epoch": 0.7149650349650349, "percentage": 23.85, "elapsed_time": "10:27:01", "remaining_time": "1 day, 9:21:47", "throughput": "17809.88", "total_tokens": 670040064}
{"current_steps": 640, "total_steps": 2679, "loss": 0.692, "learning_rate": 4.366287007103762e-06, "epoch": 0.7160839160839161, "percentage": 23.89, "elapsed_time": "10:27:56", "remaining_time": "1 day, 9:20:36", "throughput": "17811.70", "total_tokens": 671088640}
{"current_steps": 641, "total_steps": 2679, "loss": 0.7969, "learning_rate": 4.364316674281526e-06, "epoch": 0.7172027972027972, "percentage": 23.93, "elapsed_time": "10:29:02", "remaining_time": "1 day, 9:19:58", "throughput": "17808.60", "total_tokens": 672137216}
{"current_steps": 642, "total_steps": 2679, "loss": 0.6253, "learning_rate": 4.362343729192425e-06, "epoch": 0.7183216783216784, "percentage": 23.96, "elapsed_time": "10:29:52", "remaining_time": "1 day, 9:18:30", "throughput": "17812.89", "total_tokens": 673185792}
{"current_steps": 643, "total_steps": 2679, "loss": 0.7665, "learning_rate": 4.3603681746009374e-06, "epoch": 0.7194405594405594, "percentage": 24.0, "elapsed_time": "10:30:51", "remaining_time": "1 day, 9:17:33", "throughput": "17812.65", "total_tokens": 674234368}
{"current_steps": 644, "total_steps": 2679, "loss": 0.7127, "learning_rate": 4.358390013275195e-06, "epoch": 0.7205594405594405, "percentage": 24.04, "elapsed_time": "10:31:45", "remaining_time": "1 day, 9:16:17", "throughput": "17815.09", "total_tokens": 675282944}
{"current_steps": 645, "total_steps": 2679, "loss": 0.7536, "learning_rate": 4.356409247986982e-06, "epoch": 0.7216783216783217, "percentage": 24.08, "elapsed_time": "10:32:50", "remaining_time": "1 day, 9:15:39", "throughput": "17811.99", "total_tokens": 676331520}
{"current_steps": 646, "total_steps": 2679, "loss": 0.8728, "learning_rate": 4.354425881511733e-06, "epoch": 0.7227972027972028, "percentage": 24.11, "elapsed_time": "10:33:50", "remaining_time": "1 day, 9:14:44", "throughput": "17811.53", "total_tokens": 677380096}
{"current_steps": 647, "total_steps": 2679, "loss": 0.688, "learning_rate": 4.352439916628527e-06, "epoch": 0.7239160839160839, "percentage": 24.15, "elapsed_time": "10:34:44", "remaining_time": "1 day, 9:13:29", "throughput": "17813.86", "total_tokens": 678428672}
{"current_steps": 648, "total_steps": 2679, "loss": 0.6268, "learning_rate": 4.350451356120082e-06, "epoch": 0.725034965034965, "percentage": 24.19, "elapsed_time": "10:35:32", "remaining_time": "1 day, 9:11:57", "throughput": "17818.85", "total_tokens": 679477248}
{"current_steps": 649, "total_steps": 2679, "loss": 0.773, "learning_rate": 4.348460202772756e-06, "epoch": 0.7261538461538461, "percentage": 24.23, "elapsed_time": "10:36:32", "remaining_time": "1 day, 9:11:02", "throughput": "17818.19", "total_tokens": 680525824}
{"current_steps": 650, "total_steps": 2679, "loss": 0.7218, "learning_rate": 4.3464664593765385e-06, "epoch": 0.7272727272727273, "percentage": 24.26, "elapsed_time": "10:37:26", "remaining_time": "1 day, 9:09:48", "throughput": "17820.49", "total_tokens": 681574400}
{"current_steps": 651, "total_steps": 2679, "loss": 0.7529, "learning_rate": 4.344470128725047e-06, "epoch": 0.7283916083916084, "percentage": 24.3, "elapsed_time": "10:38:21", "remaining_time": "1 day, 9:08:35", "throughput": "17822.57", "total_tokens": 682622976}
{"current_steps": 652, "total_steps": 2679, "loss": 0.6315, "learning_rate": 4.342471213615525e-06, "epoch": 0.7295104895104895, "percentage": 24.34, "elapsed_time": "10:39:08", "remaining_time": "1 day, 9:07:02", "throughput": "17827.63", "total_tokens": 683671552}
{"current_steps": 653, "total_steps": 2679, "loss": 0.7196, "learning_rate": 4.34046971684884e-06, "epoch": 0.7306293706293706, "percentage": 24.37, "elapsed_time": "10:40:03", "remaining_time": "1 day, 9:05:49", "throughput": "17829.85", "total_tokens": 684720128}
{"current_steps": 654, "total_steps": 2679, "loss": 1.0307, "learning_rate": 4.3384656412294725e-06, "epoch": 0.7317482517482518, "percentage": 24.41, "elapsed_time": "10:41:15", "remaining_time": "1 day, 9:05:31", "throughput": "17823.74", "total_tokens": 685768704}
{"current_steps": 655, "total_steps": 2679, "loss": 0.8165, "learning_rate": 4.336458989565519e-06, "epoch": 0.7328671328671329, "percentage": 24.45, "elapsed_time": "10:42:19", "remaining_time": "1 day, 9:04:50", "throughput": "17820.98", "total_tokens": 686817280}
{"current_steps": 656, "total_steps": 2679, "loss": 0.6931, "learning_rate": 4.334449764668686e-06, "epoch": 0.733986013986014, "percentage": 24.49, "elapsed_time": "10:43:15", "remaining_time": "1 day, 9:03:41", "throughput": "17822.56", "total_tokens": 687865856}
{"current_steps": 657, "total_steps": 2679, "loss": 0.9317, "learning_rate": 4.332437969354284e-06, "epoch": 0.7351048951048951, "percentage": 24.52, "elapsed_time": "10:44:18", "remaining_time": "1 day, 9:02:57", "throughput": "17820.31", "total_tokens": 688914432}
{"current_steps": 658, "total_steps": 2679, "loss": 0.7599, "learning_rate": 4.330423606441225e-06, "epoch": 0.7362237762237762, "percentage": 24.56, "elapsed_time": "10:45:18", "remaining_time": "1 day, 9:02:00", "throughput": "17820.13", "total_tokens": 689963008}
{"current_steps": 659, "total_steps": 2679, "loss": 0.8251, "learning_rate": 4.328406678752022e-06, "epoch": 0.7373426573426574, "percentage": 24.6, "elapsed_time": "10:46:24", "remaining_time": "1 day, 9:01:23", "throughput": "17816.84", "total_tokens": 691011584}
{"current_steps": 660, "total_steps": 2679, "loss": 0.7429, "learning_rate": 4.326387189112776e-06, "epoch": 0.7384615384615385, "percentage": 24.64, "elapsed_time": "10:47:24", "remaining_time": "1 day, 9:00:27", "throughput": "17816.32", "total_tokens": 692060160}
{"current_steps": 661, "total_steps": 2679, "loss": 0.6885, "learning_rate": 4.324365140353185e-06, "epoch": 0.7395804195804195, "percentage": 24.67, "elapsed_time": "10:48:17", "remaining_time": "1 day, 8:59:13", "throughput": "17818.65", "total_tokens": 693108736}
{"current_steps": 662, "total_steps": 2679, "loss": 0.7502, "learning_rate": 4.322340535306525e-06, "epoch": 0.7406993006993007, "percentage": 24.71, "elapsed_time": "10:49:17", "remaining_time": "1 day, 8:58:16", "throughput": "17818.30", "total_tokens": 694157312}
{"current_steps": 663, "total_steps": 2679, "loss": 0.7426, "learning_rate": 4.32031337680966e-06, "epoch": 0.7418181818181818, "percentage": 24.75, "elapsed_time": "10:50:15", "remaining_time": "1 day, 8:57:14", "throughput": "17818.82", "total_tokens": 695205888}
{"current_steps": 664, "total_steps": 2679, "loss": 0.6304, "learning_rate": 4.31828366770303e-06, "epoch": 0.742937062937063, "percentage": 24.79, "elapsed_time": "10:51:03", "remaining_time": "1 day, 8:55:42", "throughput": "17823.78", "total_tokens": 696254464}
{"current_steps": 665, "total_steps": 2679, "loss": 0.7407, "learning_rate": 4.3162514108306465e-06, "epoch": 0.7440559440559441, "percentage": 24.82, "elapsed_time": "10:51:57", "remaining_time": "1 day, 8:54:29", "throughput": "17825.95", "total_tokens": 697303040}
{"current_steps": 666, "total_steps": 2679, "loss": 0.6271, "learning_rate": 4.314216609040095e-06, "epoch": 0.7451748251748251, "percentage": 24.86, "elapsed_time": "10:52:45", "remaining_time": "1 day, 8:52:58", "throughput": "17830.81", "total_tokens": 698351616}
{"current_steps": 667, "total_steps": 2679, "loss": 0.6549, "learning_rate": 4.312179265182523e-06, "epoch": 0.7462937062937063, "percentage": 24.9, "elapsed_time": "10:53:33", "remaining_time": "1 day, 8:51:27", "throughput": "17835.71", "total_tokens": 699400192}
{"current_steps": 668, "total_steps": 2679, "loss": 0.7184, "learning_rate": 4.310139382112644e-06, "epoch": 0.7474125874125874, "percentage": 24.93, "elapsed_time": "10:54:32", "remaining_time": "1 day, 8:50:29", "throughput": "17835.48", "total_tokens": 700448768}
{"current_steps": 669, "total_steps": 2679, "loss": 0.6622, "learning_rate": 4.308096962688726e-06, "epoch": 0.7485314685314686, "percentage": 24.97, "elapsed_time": "10:55:26", "remaining_time": "1 day, 8:49:15", "throughput": "17837.86", "total_tokens": 701497344}
{"current_steps": 670, "total_steps": 2679, "loss": 0.7929, "learning_rate": 4.3060520097725915e-06, "epoch": 0.7496503496503496, "percentage": 25.01, "elapsed_time": "10:56:30", "remaining_time": "1 day, 8:48:32", "throughput": "17835.43", "total_tokens": 702545920}
{"current_steps": 671, "total_steps": 2679, "loss": 0.8827, "learning_rate": 4.304004526229614e-06, "epoch": 0.7507692307692307, "percentage": 25.05, "elapsed_time": "10:57:36", "remaining_time": "1 day, 8:47:55", "throughput": "17832.18", "total_tokens": 703594496}
{"current_steps": 672, "total_steps": 2679, "loss": 0.6351, "learning_rate": 4.301954514928713e-06, "epoch": 0.7518881118881119, "percentage": 25.08, "elapsed_time": "10:58:24", "remaining_time": "1 day, 8:46:24", "throughput": "17837.05", "total_tokens": 704643072}
{"current_steps": 673, "total_steps": 2679, "loss": 0.7272, "learning_rate": 4.299901978742349e-06, "epoch": 0.753006993006993, "percentage": 25.12, "elapsed_time": "10:59:23", "remaining_time": "1 day, 8:45:26", "throughput": "17836.87", "total_tokens": 705691648}
{"current_steps": 674, "total_steps": 2679, "loss": 0.7258, "learning_rate": 4.29784692054652e-06, "epoch": 0.7541258741258742, "percentage": 25.16, "elapsed_time": "11:00:17", "remaining_time": "1 day, 8:44:13", "throughput": "17839.12", "total_tokens": 706740224}
{"current_steps": 675, "total_steps": 2679, "loss": 0.7522, "learning_rate": 4.29578934322076e-06, "epoch": 0.7552447552447552, "percentage": 25.2, "elapsed_time": "11:01:16", "remaining_time": "1 day, 8:43:14", "throughput": "17839.09", "total_tokens": 707788800}
{"current_steps": 676, "total_steps": 2679, "loss": 0.7302, "learning_rate": 4.29372924964813e-06, "epoch": 0.7563636363636363, "percentage": 25.23, "elapsed_time": "11:02:16", "remaining_time": "1 day, 8:42:19", "throughput": "17838.58", "total_tokens": 708837376}
{"current_steps": 677, "total_steps": 2679, "loss": 0.8214, "learning_rate": 4.2916666427152175e-06, "epoch": 0.7574825174825175, "percentage": 25.27, "elapsed_time": "11:03:21", "remaining_time": "1 day, 8:41:39", "throughput": "17835.68", "total_tokens": 709885952}
{"current_steps": 678, "total_steps": 2679, "loss": 0.7277, "learning_rate": 4.289601525312134e-06, "epoch": 0.7586013986013986, "percentage": 25.31, "elapsed_time": "11:04:15", "remaining_time": "1 day, 8:40:28", "throughput": "17837.59", "total_tokens": 710934528}
{"current_steps": 679, "total_steps": 2679, "loss": 0.755, "learning_rate": 4.287533900332506e-06, "epoch": 0.7597202797202797, "percentage": 25.35, "elapsed_time": "11:05:15", "remaining_time": "1 day, 8:39:30", "throughput": "17837.43", "total_tokens": 711983104}
{"current_steps": 680, "total_steps": 2679, "loss": 0.7632, "learning_rate": 4.285463770673474e-06, "epoch": 0.7608391608391608, "percentage": 25.38, "elapsed_time": "11:06:18", "remaining_time": "1 day, 8:38:46", "throughput": "17835.23", "total_tokens": 713031680}
{"current_steps": 681, "total_steps": 2679, "loss": 0.6914, "learning_rate": 4.283391139235688e-06, "epoch": 0.761958041958042, "percentage": 25.42, "elapsed_time": "11:07:12", "remaining_time": "1 day, 8:37:32", "throughput": "17837.43", "total_tokens": 714080256}
{"current_steps": 682, "total_steps": 2679, "loss": 0.6956, "learning_rate": 4.281316008923306e-06, "epoch": 0.7630769230769231, "percentage": 25.46, "elapsed_time": "11:08:06", "remaining_time": "1 day, 8:36:18", "throughput": "17839.72", "total_tokens": 715128832}
{"current_steps": 683, "total_steps": 2679, "loss": 0.8227, "learning_rate": 4.279238382643985e-06, "epoch": 0.7641958041958042, "percentage": 25.49, "elapsed_time": "11:09:06", "remaining_time": "1 day, 8:35:24", "throughput": "17839.09", "total_tokens": 716177408}
{"current_steps": 684, "total_steps": 2679, "loss": 0.8398, "learning_rate": 4.277158263308878e-06, "epoch": 0.7653146853146853, "percentage": 25.53, "elapsed_time": "11:10:06", "remaining_time": "1 day, 8:34:28", "throughput": "17838.56", "total_tokens": 717225984}
{"current_steps": 685, "total_steps": 2679, "loss": 0.744, "learning_rate": 4.275075653832635e-06, "epoch": 0.7664335664335664, "percentage": 25.57, "elapsed_time": "11:11:09", "remaining_time": "1 day, 8:33:42", "throughput": "17836.69", "total_tokens": 718274560}
{"current_steps": 686, "total_steps": 2679, "loss": 0.8331, "learning_rate": 4.272990557133391e-06, "epoch": 0.7675524475524476, "percentage": 25.61, "elapsed_time": "11:12:09", "remaining_time": "1 day, 8:32:47", "throughput": "17836.17", "total_tokens": 719323136}
{"current_steps": 687, "total_steps": 2679, "loss": 0.8187, "learning_rate": 4.27090297613277e-06, "epoch": 0.7686713286713287, "percentage": 25.64, "elapsed_time": "11:13:09", "remaining_time": "1 day, 8:31:51", "throughput": "17835.72", "total_tokens": 720371712}
{"current_steps": 688, "total_steps": 2679, "loss": 0.7071, "learning_rate": 4.268812913755875e-06, "epoch": 0.7697902097902098, "percentage": 25.68, "elapsed_time": "11:14:05", "remaining_time": "1 day, 8:30:45", "throughput": "17836.73", "total_tokens": 721420288}
{"current_steps": 689, "total_steps": 2679, "loss": 0.7088, "learning_rate": 4.266720372931285e-06, "epoch": 0.7709090909090909, "percentage": 25.72, "elapsed_time": "11:15:03", "remaining_time": "1 day, 8:29:42", "throughput": "17837.40", "total_tokens": 722468864}
{"current_steps": 690, "total_steps": 2679, "loss": 0.6332, "learning_rate": 4.2646253565910535e-06, "epoch": 0.772027972027972, "percentage": 25.76, "elapsed_time": "11:15:51", "remaining_time": "1 day, 8:28:13", "throughput": "17841.97", "total_tokens": 723517440}
{"current_steps": 691, "total_steps": 2679, "loss": 0.808, "learning_rate": 4.262527867670702e-06, "epoch": 0.7731468531468532, "percentage": 25.79, "elapsed_time": "11:16:50", "remaining_time": "1 day, 8:27:16", "throughput": "17841.69", "total_tokens": 724566016}
{"current_steps": 692, "total_steps": 2679, "loss": 0.7215, "learning_rate": 4.260427909109216e-06, "epoch": 0.7742657342657343, "percentage": 25.83, "elapsed_time": "11:17:48", "remaining_time": "1 day, 8:26:14", "throughput": "17842.24", "total_tokens": 725614592}
{"current_steps": 693, "total_steps": 2679, "loss": 0.8065, "learning_rate": 4.258325483849044e-06, "epoch": 0.7753846153846153, "percentage": 25.87, "elapsed_time": "11:18:45", "remaining_time": "1 day, 8:25:11", "throughput": "17842.98", "total_tokens": 726663168}
{"current_steps": 694, "total_steps": 2679, "loss": 0.7724, "learning_rate": 4.256220594836087e-06, "epoch": 0.7765034965034965, "percentage": 25.91, "elapsed_time": "11:19:42", "remaining_time": "1 day, 8:24:07", "throughput": "17843.72", "total_tokens": 727711744}
{"current_steps": 695, "total_steps": 2679, "loss": 0.726, "learning_rate": 4.254113245019701e-06, "epoch": 0.7776223776223776, "percentage": 25.94, "elapsed_time": "11:20:36", "remaining_time": "1 day, 8:22:55", "throughput": "17845.70", "total_tokens": 728760320}
{"current_steps": 696, "total_steps": 2679, "loss": 0.8208, "learning_rate": 4.252003437352688e-06, "epoch": 0.7787412587412588, "percentage": 25.98, "elapsed_time": "11:21:42", "remaining_time": "1 day, 8:22:16", "throughput": "17842.65", "total_tokens": 729808896}
{"current_steps": 697, "total_steps": 2679, "loss": 0.6953, "learning_rate": 4.249891174791297e-06, "epoch": 0.7798601398601399, "percentage": 26.02, "elapsed_time": "11:22:36", "remaining_time": "1 day, 8:21:03", "throughput": "17844.94", "total_tokens": 730857472}
{"current_steps": 698, "total_steps": 2679, "loss": 0.6712, "learning_rate": 4.247776460295213e-06, "epoch": 0.7809790209790209, "percentage": 26.05, "elapsed_time": "11:23:24", "remaining_time": "1 day, 8:19:34", "throughput": "17849.56", "total_tokens": 731906048}
{"current_steps": 699, "total_steps": 2679, "loss": 0.7831, "learning_rate": 4.245659296827559e-06, "epoch": 0.7820979020979021, "percentage": 26.09, "elapsed_time": "11:24:29", "remaining_time": "1 day, 8:18:53", "throughput": "17846.74", "total_tokens": 732954624}
{"current_steps": 700, "total_steps": 2679, "loss": 0.7666, "learning_rate": 4.243539687354889e-06, "epoch": 0.7832167832167832, "percentage": 26.13, "elapsed_time": "11:25:34", "remaining_time": "1 day, 8:18:12", "throughput": "17844.09", "total_tokens": 734003200}
{"current_steps": 701, "total_steps": 2679, "loss": 0.7889, "learning_rate": 4.2414176348471845e-06, "epoch": 0.7843356643356644, "percentage": 26.17, "elapsed_time": "11:26:37", "remaining_time": "1 day, 8:17:25", "throughput": "17842.21", "total_tokens": 735051776}
{"current_steps": 702, "total_steps": 2679, "loss": 0.6987, "learning_rate": 4.23929314227785e-06, "epoch": 0.7854545454545454, "percentage": 26.2, "elapsed_time": "11:27:36", "remaining_time": "1 day, 8:16:28", "throughput": "17842.01", "total_tokens": 736100352}
{"current_steps": 703, "total_steps": 2679, "loss": 0.7479, "learning_rate": 4.237166212623708e-06, "epoch": 0.7865734265734265, "percentage": 26.24, "elapsed_time": "11:28:33", "remaining_time": "1 day, 8:15:24", "throughput": "17842.85", "total_tokens": 737148928}
{"current_steps": 704, "total_steps": 2679, "loss": 0.727, "learning_rate": 4.235036848864998e-06, "epoch": 0.7876923076923077, "percentage": 26.28, "elapsed_time": "11:29:27", "remaining_time": "1 day, 8:14:13", "throughput": "17844.75", "total_tokens": 738197504}
{"current_steps": 705, "total_steps": 2679, "loss": 0.6466, "learning_rate": 4.232905053985368e-06, "epoch": 0.7888111888111888, "percentage": 26.32, "elapsed_time": "11:30:15", "remaining_time": "1 day, 8:12:44", "throughput": "17849.34", "total_tokens": 739246080}
{"current_steps": 706, "total_steps": 2679, "loss": 0.9148, "learning_rate": 4.230770830971873e-06, "epoch": 0.78993006993007, "percentage": 26.35, "elapsed_time": "11:31:27", "remaining_time": "1 day, 8:12:20", "throughput": "17843.93", "total_tokens": 740294656}
{"current_steps": 707, "total_steps": 2679, "loss": 0.7618, "learning_rate": 4.228634182814972e-06, "epoch": 0.791048951048951, "percentage": 26.39, "elapsed_time": "11:32:23", "remaining_time": "1 day, 8:11:15", "throughput": "17844.94", "total_tokens": 741343232}
{"current_steps": 708, "total_steps": 2679, "loss": 0.7249, "learning_rate": 4.226495112508521e-06, "epoch": 0.7921678321678322, "percentage": 26.43, "elapsed_time": "11:33:22", "remaining_time": "1 day, 8:10:18", "throughput": "17844.69", "total_tokens": 742391808}
{"current_steps": 709, "total_steps": 2679, "loss": 0.7519, "learning_rate": 4.224353623049767e-06, "epoch": 0.7932867132867133, "percentage": 26.47, "elapsed_time": "11:34:16", "remaining_time": "1 day, 8:09:05", "throughput": "17846.83", "total_tokens": 743440384}
{"current_steps": 710, "total_steps": 2679, "loss": 0.7133, "learning_rate": 4.222209717439351e-06, "epoch": 0.7944055944055944, "percentage": 26.5, "elapsed_time": "11:35:11", "remaining_time": "1 day, 8:07:55", "throughput": "17848.64", "total_tokens": 744488960}
{"current_steps": 711, "total_steps": 2679, "loss": 0.7734, "learning_rate": 4.220063398681299e-06, "epoch": 0.7955244755244755, "percentage": 26.54, "elapsed_time": "11:36:11", "remaining_time": "1 day, 8:07:00", "throughput": "17848.09", "total_tokens": 745537536}
{"current_steps": 712, "total_steps": 2679, "loss": 0.7867, "learning_rate": 4.2179146697830155e-06, "epoch": 0.7966433566433566, "percentage": 26.58, "elapsed_time": "11:37:11", "remaining_time": "1 day, 8:06:04", "throughput": "17847.57", "total_tokens": 746586112}
{"current_steps": 713, "total_steps": 2679, "loss": 0.7775, "learning_rate": 4.215763533755285e-06, "epoch": 0.7977622377622378, "percentage": 26.61, "elapsed_time": "11:38:18", "remaining_time": "1 day, 8:05:29", "throughput": "17844.00", "total_tokens": 747634688}
{"current_steps": 714, "total_steps": 2679, "loss": 0.791, "learning_rate": 4.213609993612262e-06, "epoch": 0.7988811188811189, "percentage": 26.65, "elapsed_time": "11:39:18", "remaining_time": "1 day, 8:04:33", "throughput": "17843.45", "total_tokens": 748683264}
{"current_steps": 715, "total_steps": 2679, "loss": 0.7399, "learning_rate": 4.211454052371471e-06, "epoch": 0.8, "percentage": 26.69, "elapsed_time": "11:40:19", "remaining_time": "1 day, 8:03:42", "throughput": "17842.33", "total_tokens": 749731840}
{"current_steps": 716, "total_steps": 2679, "loss": 0.8503, "learning_rate": 4.209295713053802e-06, "epoch": 0.8011188811188811, "percentage": 26.73, "elapsed_time": "11:41:20", "remaining_time": "1 day, 8:02:47", "throughput": "17841.71", "total_tokens": 750780416}
{"current_steps": 717, "total_steps": 2679, "loss": 0.881, "learning_rate": 4.207134978683506e-06, "epoch": 0.8022377622377622, "percentage": 26.76, "elapsed_time": "11:42:31", "remaining_time": "1 day, 8:02:22", "throughput": "17836.55", "total_tokens": 751828992}
{"current_steps": 718, "total_steps": 2679, "loss": 0.7961, "learning_rate": 4.204971852288185e-06, "epoch": 0.8033566433566434, "percentage": 26.8, "elapsed_time": "11:43:31", "remaining_time": "1 day, 8:01:28", "throughput": "17835.71", "total_tokens": 752877568}
{"current_steps": 719, "total_steps": 2679, "loss": 0.7213, "learning_rate": 4.202806336898798e-06, "epoch": 0.8044755244755245, "percentage": 26.84, "elapsed_time": "11:44:26", "remaining_time": "1 day, 8:00:17", "throughput": "17837.63", "total_tokens": 753926144}
{"current_steps": 720, "total_steps": 2679, "loss": 0.7755, "learning_rate": 4.200638435549648e-06, "epoch": 0.8055944055944056, "percentage": 26.88, "elapsed_time": "11:45:31", "remaining_time": "1 day, 7:59:37", "throughput": "17834.71", "total_tokens": 754974720}
{"current_steps": 721, "total_steps": 2679, "loss": 0.7661, "learning_rate": 4.198468151278382e-06, "epoch": 0.8067132867132867, "percentage": 26.91, "elapsed_time": "11:46:32", "remaining_time": "1 day, 7:58:44", "throughput": "17833.87", "total_tokens": 756023296}
{"current_steps": 722, "total_steps": 2679, "loss": 0.642, "learning_rate": 4.196295487125989e-06, "epoch": 0.8078321678321678, "percentage": 26.95, "elapsed_time": "11:47:20", "remaining_time": "1 day, 7:57:17", "throughput": "17838.26", "total_tokens": 757071872}
{"current_steps": 723, "total_steps": 2679, "loss": 0.7341, "learning_rate": 4.194120446136788e-06, "epoch": 0.808951048951049, "percentage": 26.99, "elapsed_time": "11:48:19", "remaining_time": "1 day, 7:56:18", "throughput": "17838.17", "total_tokens": 758120448}
{"current_steps": 724, "total_steps": 2679, "loss": 0.8549, "learning_rate": 4.19194303135843e-06, "epoch": 0.8100699300699301, "percentage": 27.03, "elapsed_time": "11:49:25", "remaining_time": "1 day, 7:55:37", "throughput": "17835.50", "total_tokens": 759169024}
{"current_steps": 725, "total_steps": 2679, "loss": 0.709, "learning_rate": 4.189763245841895e-06, "epoch": 0.8111888111888111, "percentage": 27.06, "elapsed_time": "11:50:18", "remaining_time": "1 day, 7:54:25", "throughput": "17837.57", "total_tokens": 760217600}
{"current_steps": 726, "total_steps": 2679, "loss": 0.7409, "learning_rate": 4.187581092641481e-06, "epoch": 0.8123076923076923, "percentage": 27.1, "elapsed_time": "11:51:14", "remaining_time": "1 day, 7:53:18", "throughput": "17838.82", "total_tokens": 761266176}
{"current_steps": 727, "total_steps": 2679, "loss": 0.7986, "learning_rate": 4.185396574814804e-06, "epoch": 0.8134265734265734, "percentage": 27.14, "elapsed_time": "11:52:15", "remaining_time": "1 day, 7:52:23", "throughput": "17838.17", "total_tokens": 762314752}
{"current_steps": 728, "total_steps": 2679, "loss": 0.7582, "learning_rate": 4.183209695422797e-06, "epoch": 0.8145454545454546, "percentage": 27.17, "elapsed_time": "11:53:09", "remaining_time": "1 day, 7:51:14", "throughput": "17839.77", "total_tokens": 763363328}
{"current_steps": 729, "total_steps": 2679, "loss": 0.6475, "learning_rate": 4.1810204575296966e-06, "epoch": 0.8156643356643357, "percentage": 27.21, "elapsed_time": "11:53:58", "remaining_time": "1 day, 7:49:47", "throughput": "17844.22", "total_tokens": 764411904}
{"current_steps": 730, "total_steps": 2679, "loss": 0.701, "learning_rate": 4.178828864203049e-06, "epoch": 0.8167832167832167, "percentage": 27.25, "elapsed_time": "11:54:52", "remaining_time": "1 day, 7:48:36", "throughput": "17846.17", "total_tokens": 765460480}
{"current_steps": 731, "total_steps": 2679, "loss": 0.6757, "learning_rate": 4.176634918513698e-06, "epoch": 0.8179020979020979, "percentage": 27.29, "elapsed_time": "11:55:45", "remaining_time": "1 day, 7:47:24", "throughput": "17848.25", "total_tokens": 766509056}
{"current_steps": 732, "total_steps": 2679, "loss": 0.7554, "learning_rate": 4.174438623535784e-06, "epoch": 0.819020979020979, "percentage": 27.32, "elapsed_time": "11:56:45", "remaining_time": "1 day, 7:46:28", "throughput": "17847.76", "total_tokens": 767557632}
{"current_steps": 733, "total_steps": 2679, "loss": 0.7702, "learning_rate": 4.17223998234674e-06, "epoch": 0.8201398601398602, "percentage": 27.36, "elapsed_time": "11:57:45", "remaining_time": "1 day, 7:45:31", "throughput": "17847.55", "total_tokens": 768606208}
{"current_steps": 734, "total_steps": 2679, "loss": 0.7197, "learning_rate": 4.170038998027285e-06, "epoch": 0.8212587412587412, "percentage": 27.4, "elapsed_time": "11:58:40", "remaining_time": "1 day, 7:44:22", "throughput": "17849.00", "total_tokens": 769654784}
{"current_steps": 735, "total_steps": 2679, "loss": 0.6775, "learning_rate": 4.167835673661422e-06, "epoch": 0.8223776223776224, "percentage": 27.44, "elapsed_time": "11:59:33", "remaining_time": "1 day, 7:43:10", "throughput": "17851.17", "total_tokens": 770703360}
{"current_steps": 736, "total_steps": 2679, "loss": 0.7137, "learning_rate": 4.165630012336435e-06, "epoch": 0.8234965034965035, "percentage": 27.47, "elapsed_time": "12:00:32", "remaining_time": "1 day, 7:42:12", "throughput": "17851.03", "total_tokens": 771751936}
{"current_steps": 737, "total_steps": 2679, "loss": 0.6855, "learning_rate": 4.163422017142879e-06, "epoch": 0.8246153846153846, "percentage": 27.51, "elapsed_time": "12:01:26", "remaining_time": "1 day, 7:41:00", "throughput": "17853.05", "total_tokens": 772800512}
{"current_steps": 738, "total_steps": 2679, "loss": 0.779, "learning_rate": 4.1612116911745805e-06, "epoch": 0.8257342657342658, "percentage": 27.55, "elapsed_time": "12:02:26", "remaining_time": "1 day, 7:40:05", "throughput": "17852.59", "total_tokens": 773849088}
{"current_steps": 739, "total_steps": 2679, "loss": 0.7513, "learning_rate": 4.158999037528632e-06, "epoch": 0.8268531468531468, "percentage": 27.58, "elapsed_time": "12:03:26", "remaining_time": "1 day, 7:39:08", "throughput": "17852.22", "total_tokens": 774897664}
{"current_steps": 740, "total_steps": 2679, "loss": 0.7909, "learning_rate": 4.156784059305388e-06, "epoch": 0.827972027972028, "percentage": 27.62, "elapsed_time": "12:04:25", "remaining_time": "1 day, 7:38:12", "throughput": "17851.90", "total_tokens": 775946240}
{"current_steps": 741, "total_steps": 2679, "loss": 0.8066, "learning_rate": 4.1545667596084596e-06, "epoch": 0.8290909090909091, "percentage": 27.66, "elapsed_time": "12:05:31", "remaining_time": "1 day, 7:37:30", "throughput": "17849.21", "total_tokens": 776994816}
{"current_steps": 742, "total_steps": 2679, "loss": 0.7322, "learning_rate": 4.152347141544711e-06, "epoch": 0.8302097902097902, "percentage": 27.7, "elapsed_time": "12:06:30", "remaining_time": "1 day, 7:36:33", "throughput": "17848.95", "total_tokens": 778043392}
{"current_steps": 743, "total_steps": 2679, "loss": 0.8437, "learning_rate": 4.150125208224255e-06, "epoch": 0.8313286713286713, "percentage": 27.73, "elapsed_time": "12:07:26", "remaining_time": "1 day, 7:35:27", "throughput": "17850.07", "total_tokens": 779091968}
{"current_steps": 744, "total_steps": 2679, "loss": 0.8597, "learning_rate": 4.147900962760447e-06, "epoch": 0.8324475524475524, "percentage": 27.77, "elapsed_time": "12:08:37", "remaining_time": "1 day, 7:35:01", "throughput": "17844.93", "total_tokens": 780140544}
{"current_steps": 745, "total_steps": 2679, "loss": 0.7272, "learning_rate": 4.145674408269885e-06, "epoch": 0.8335664335664336, "percentage": 27.81, "elapsed_time": "12:09:32", "remaining_time": "1 day, 7:33:51", "throughput": "17846.75", "total_tokens": 781189120}
{"current_steps": 746, "total_steps": 2679, "loss": 0.6691, "learning_rate": 4.1434455478724e-06, "epoch": 0.8346853146853147, "percentage": 27.85, "elapsed_time": "12:10:20", "remaining_time": "1 day, 7:32:26", "throughput": "17850.78", "total_tokens": 782237696}
{"current_steps": 747, "total_steps": 2679, "loss": 0.7349, "learning_rate": 4.141214384691056e-06, "epoch": 0.8358041958041958, "percentage": 27.88, "elapsed_time": "12:11:20", "remaining_time": "1 day, 7:31:30", "throughput": "17850.45", "total_tokens": 783286272}
{"current_steps": 748, "total_steps": 2679, "loss": 0.8242, "learning_rate": 4.138980921852141e-06, "epoch": 0.8369230769230769, "percentage": 27.92, "elapsed_time": "12:12:20", "remaining_time": "1 day, 7:30:34", "throughput": "17849.87", "total_tokens": 784334848}
{"current_steps": 749, "total_steps": 2679, "loss": 0.6431, "learning_rate": 4.136745162485168e-06, "epoch": 0.838041958041958, "percentage": 27.96, "elapsed_time": "12:13:08", "remaining_time": "1 day, 7:29:08", "throughput": "17854.28", "total_tokens": 785383424}
{"current_steps": 750, "total_steps": 2679, "loss": 0.7689, "learning_rate": 4.134507109722865e-06, "epoch": 0.8391608391608392, "percentage": 28.0, "elapsed_time": "12:14:13", "remaining_time": "1 day, 7:28:26", "throughput": "17851.65", "total_tokens": 786432000}
{"current_steps": 750, "total_steps": 2679, "eval_loss": 0.7424774169921875, "epoch": 0.8391608391608392, "percentage": 28.0, "elapsed_time": "12:18:20", "remaining_time": "1 day, 7:39:00", "throughput": "17752.25", "total_tokens": 786432000}
{"current_steps": 751, "total_steps": 2679, "loss": 0.7584, "learning_rate": 4.1322667667011774e-06, "epoch": 0.8402797202797203, "percentage": 28.03, "elapsed_time": "12:19:56", "remaining_time": "1 day, 7:39:36", "throughput": "17737.48", "total_tokens": 787480576}
{"current_steps": 752, "total_steps": 2679, "loss": 0.6891, "learning_rate": 4.130024136559255e-06, "epoch": 0.8413986013986015, "percentage": 28.07, "elapsed_time": "12:20:52", "remaining_time": "1 day, 7:38:29", "throughput": "17738.66", "total_tokens": 788529152}
{"current_steps": 753, "total_steps": 2679, "loss": 0.7819, "learning_rate": 4.127779222439457e-06, "epoch": 0.8425174825174825, "percentage": 28.11, "elapsed_time": "12:21:57", "remaining_time": "1 day, 7:37:46", "throughput": "17736.24", "total_tokens": 789577728}
{"current_steps": 754, "total_steps": 2679, "loss": 0.8617, "learning_rate": 4.125532027487339e-06, "epoch": 0.8436363636363636, "percentage": 28.14, "elapsed_time": "12:23:04", "remaining_time": "1 day, 7:37:07", "throughput": "17733.09", "total_tokens": 790626304}
{"current_steps": 755, "total_steps": 2679, "loss": 0.8794, "learning_rate": 4.123282554851654e-06, "epoch": 0.8447552447552448, "percentage": 28.18, "elapsed_time": "12:24:10", "remaining_time": "1 day, 7:36:24", "throughput": "17730.52", "total_tokens": 791674880}
{"current_steps": 756, "total_steps": 2679, "loss": 0.6896, "learning_rate": 4.121030807684349e-06, "epoch": 0.8458741258741259, "percentage": 28.22, "elapsed_time": "12:25:04", "remaining_time": "1 day, 7:35:11", "throughput": "17732.67", "total_tokens": 792723456}
{"current_steps": 757, "total_steps": 2679, "loss": 0.6846, "learning_rate": 4.118776789140551e-06, "epoch": 0.846993006993007, "percentage": 28.26, "elapsed_time": "12:25:57", "remaining_time": "1 day, 7:33:58", "throughput": "17734.79", "total_tokens": 793772032}
{"current_steps": 758, "total_steps": 2679, "loss": 0.715, "learning_rate": 4.11652050237858e-06, "epoch": 0.8481118881118881, "percentage": 28.29, "elapsed_time": "12:26:57", "remaining_time": "1 day, 7:33:00", "throughput": "17734.63", "total_tokens": 794820608}
{"current_steps": 759, "total_steps": 2679, "loss": 0.7033, "learning_rate": 4.114261950559924e-06, "epoch": 0.8492307692307692, "percentage": 28.33, "elapsed_time": "12:27:51", "remaining_time": "1 day, 7:31:49", "throughput": "17736.55", "total_tokens": 795869184}
{"current_steps": 760, "total_steps": 2679, "loss": 0.6531, "learning_rate": 4.112001136849252e-06, "epoch": 0.8503496503496504, "percentage": 28.37, "elapsed_time": "12:28:40", "remaining_time": "1 day, 7:30:23", "throughput": "17740.72", "total_tokens": 796917760}
{"current_steps": 761, "total_steps": 2679, "loss": 0.7427, "learning_rate": 4.109738064414397e-06, "epoch": 0.8514685314685315, "percentage": 28.41, "elapsed_time": "12:29:36", "remaining_time": "1 day, 7:29:18", "throughput": "17741.72", "total_tokens": 797966336}
{"current_steps": 762, "total_steps": 2679, "loss": 0.8499, "learning_rate": 4.107472736426362e-06, "epoch": 0.8525874125874126, "percentage": 28.44, "elapsed_time": "12:30:42", "remaining_time": "1 day, 7:28:36", "throughput": "17739.01", "total_tokens": 799014912}
{"current_steps": 763, "total_steps": 2679, "loss": 0.7645, "learning_rate": 4.105205156059307e-06, "epoch": 0.8537062937062937, "percentage": 28.48, "elapsed_time": "12:31:39", "remaining_time": "1 day, 7:27:31", "throughput": "17739.97", "total_tokens": 800063488}
{"current_steps": 764, "total_steps": 2679, "loss": 0.6868, "learning_rate": 4.102935326490549e-06, "epoch": 0.8548251748251748, "percentage": 28.52, "elapsed_time": "12:32:33", "remaining_time": "1 day, 7:26:18", "throughput": "17742.13", "total_tokens": 801112064}
{"current_steps": 765, "total_steps": 2679, "loss": 0.6391, "learning_rate": 4.100663250900556e-06, "epoch": 0.855944055944056, "percentage": 28.56, "elapsed_time": "12:33:21", "remaining_time": "1 day, 7:24:51", "throughput": "17746.50", "total_tokens": 802160640}
{"current_steps": 766, "total_steps": 2679, "loss": 0.7374, "learning_rate": 4.098388932472944e-06, "epoch": 0.857062937062937, "percentage": 28.59, "elapsed_time": "12:34:20", "remaining_time": "1 day, 7:23:52", "throughput": "17746.45", "total_tokens": 803209216}
{"current_steps": 767, "total_steps": 2679, "loss": 0.7492, "learning_rate": 4.0961123743944715e-06, "epoch": 0.8581818181818182, "percentage": 28.63, "elapsed_time": "12:35:19", "remaining_time": "1 day, 7:22:52", "throughput": "17746.54", "total_tokens": 804257792}
{"current_steps": 768, "total_steps": 2679, "loss": 0.6784, "learning_rate": 4.093833579855036e-06, "epoch": 0.8593006993006993, "percentage": 28.67, "elapsed_time": "12:36:12", "remaining_time": "1 day, 7:21:39", "throughput": "17748.78", "total_tokens": 805306368}
{"current_steps": 769, "total_steps": 2679, "loss": 0.6226, "learning_rate": 4.0915525520476665e-06, "epoch": 0.8604195804195804, "percentage": 28.7, "elapsed_time": "12:37:00", "remaining_time": "1 day, 7:20:13", "throughput": "17753.05", "total_tokens": 806354944}
{"current_steps": 770, "total_steps": 2679, "loss": 0.8488, "learning_rate": 4.089269294168522e-06, "epoch": 0.8615384615384616, "percentage": 28.74, "elapsed_time": "12:38:01", "remaining_time": "1 day, 7:19:19", "throughput": "17752.20", "total_tokens": 807403520}
{"current_steps": 771, "total_steps": 2679, "loss": 0.8694, "learning_rate": 4.086983809416887e-06, "epoch": 0.8626573426573426, "percentage": 28.78, "elapsed_time": "12:39:08", "remaining_time": "1 day, 7:18:38", "throughput": "17749.39", "total_tokens": 808452096}
{"current_steps": 772, "total_steps": 2679, "loss": 0.7307, "learning_rate": 4.084696100995167e-06, "epoch": 0.8637762237762238, "percentage": 28.82, "elapsed_time": "12:40:07", "remaining_time": "1 day, 7:17:40", "throughput": "17749.14", "total_tokens": 809500672}
{"current_steps": 773, "total_steps": 2679, "loss": 0.8257, "learning_rate": 4.082406172108882e-06, "epoch": 0.8648951048951049, "percentage": 28.85, "elapsed_time": "12:41:13", "remaining_time": "1 day, 7:16:57", "throughput": "17746.63", "total_tokens": 810549248}
{"current_steps": 774, "total_steps": 2679, "loss": 0.7092, "learning_rate": 4.0801140259666624e-06, "epoch": 0.866013986013986, "percentage": 28.89, "elapsed_time": "12:42:07", "remaining_time": "1 day, 7:15:46", "throughput": "17748.52", "total_tokens": 811597824}
{"current_steps": 775, "total_steps": 2679, "loss": 0.7616, "learning_rate": 4.0778196657802484e-06, "epoch": 0.8671328671328671, "percentage": 28.93, "elapsed_time": "12:43:07", "remaining_time": "1 day, 7:14:49", "throughput": "17748.21", "total_tokens": 812646400}
{"current_steps": 776, "total_steps": 2679, "loss": 0.7918, "learning_rate": 4.075523094764479e-06, "epoch": 0.8682517482517482, "percentage": 28.97, "elapsed_time": "12:44:12", "remaining_time": "1 day, 7:14:04", "throughput": "17745.94", "total_tokens": 813694976}
{"current_steps": 777, "total_steps": 2679, "loss": 0.6629, "learning_rate": 4.073224316137293e-06, "epoch": 0.8693706293706294, "percentage": 29.0, "elapsed_time": "12:45:06", "remaining_time": "1 day, 7:12:53", "throughput": "17747.89", "total_tokens": 814743552}
{"current_steps": 778, "total_steps": 2679, "loss": 0.8073, "learning_rate": 4.070923333119723e-06, "epoch": 0.8704895104895105, "percentage": 29.04, "elapsed_time": "12:46:11", "remaining_time": "1 day, 7:12:08", "throughput": "17745.60", "total_tokens": 815792128}
{"current_steps": 779, "total_steps": 2679, "loss": 0.6477, "learning_rate": 4.068620148935889e-06, "epoch": 0.8716083916083917, "percentage": 29.08, "elapsed_time": "12:46:59", "remaining_time": "1 day, 7:10:43", "throughput": "17749.76", "total_tokens": 816840704}
{"current_steps": 780, "total_steps": 2679, "loss": 0.84, "learning_rate": 4.066314766812996e-06, "epoch": 0.8727272727272727, "percentage": 29.12, "elapsed_time": "12:47:59", "remaining_time": "1 day, 7:09:46", "throughput": "17749.38", "total_tokens": 817889280}
{"current_steps": 781, "total_steps": 2679, "loss": 0.7025, "learning_rate": 4.0640071899813284e-06, "epoch": 0.8738461538461538, "percentage": 29.15, "elapsed_time": "12:48:54", "remaining_time": "1 day, 7:08:36", "throughput": "17751.23", "total_tokens": 818937856}
{"current_steps": 782, "total_steps": 2679, "loss": 0.9056, "learning_rate": 4.061697421674245e-06, "epoch": 0.874965034965035, "percentage": 29.19, "elapsed_time": "12:50:00", "remaining_time": "1 day, 7:07:54", "throughput": "17748.50", "total_tokens": 819986432}
{"current_steps": 783, "total_steps": 2679, "loss": 0.9119, "learning_rate": 4.059385465128179e-06, "epoch": 0.8760839160839161, "percentage": 29.23, "elapsed_time": "12:51:05", "remaining_time": "1 day, 7:07:10", "throughput": "17746.00", "total_tokens": 821035008}
{"current_steps": 784, "total_steps": 2679, "loss": 0.7266, "learning_rate": 4.057071323582623e-06, "epoch": 0.8772027972027973, "percentage": 29.26, "elapsed_time": "12:52:05", "remaining_time": "1 day, 7:06:12", "throughput": "17745.84", "total_tokens": 822083584}
{"current_steps": 785, "total_steps": 2679, "loss": 0.8663, "learning_rate": 4.054755000280139e-06, "epoch": 0.8783216783216783, "percentage": 29.3, "elapsed_time": "12:53:10", "remaining_time": "1 day, 7:05:28", "throughput": "17743.42", "total_tokens": 823132160}
{"current_steps": 786, "total_steps": 2679, "loss": 0.7008, "learning_rate": 4.05243649846634e-06, "epoch": 0.8794405594405594, "percentage": 29.34, "elapsed_time": "12:54:04", "remaining_time": "1 day, 7:04:16", "throughput": "17745.50", "total_tokens": 824180736}
{"current_steps": 787, "total_steps": 2679, "loss": 0.7324, "learning_rate": 4.050115821389894e-06, "epoch": 0.8805594405594406, "percentage": 29.38, "elapsed_time": "12:55:04", "remaining_time": "1 day, 7:03:19", "throughput": "17745.14", "total_tokens": 825229312}
{"current_steps": 788, "total_steps": 2679, "loss": 0.6976, "learning_rate": 4.047792972302518e-06, "epoch": 0.8816783216783217, "percentage": 29.41, "elapsed_time": "12:55:58", "remaining_time": "1 day, 7:02:08", "throughput": "17747.17", "total_tokens": 826277888}
{"current_steps": 789, "total_steps": 2679, "loss": 0.8645, "learning_rate": 4.045467954458969e-06, "epoch": 0.8827972027972028, "percentage": 29.45, "elapsed_time": "12:57:04", "remaining_time": "1 day, 7:01:25", "throughput": "17744.58", "total_tokens": 827326464}
{"current_steps": 790, "total_steps": 2679, "loss": 0.684, "learning_rate": 4.0431407711170465e-06, "epoch": 0.8839160839160839, "percentage": 29.49, "elapsed_time": "12:57:57", "remaining_time": "1 day, 7:00:12", "throughput": "17746.83", "total_tokens": 828375040}
{"current_steps": 791, "total_steps": 2679, "loss": 0.8267, "learning_rate": 4.040811425537583e-06, "epoch": 0.885034965034965, "percentage": 29.53, "elapsed_time": "12:59:07", "remaining_time": "1 day, 6:59:40", "throughput": "17742.44", "total_tokens": 829423616}
{"current_steps": 792, "total_steps": 2679, "loss": 0.7346, "learning_rate": 4.038479920984439e-06, "epoch": 0.8861538461538462, "percentage": 29.56, "elapsed_time": "13:00:02", "remaining_time": "1 day, 6:58:29", "throughput": "17744.31", "total_tokens": 830472192}
{"current_steps": 793, "total_steps": 2679, "loss": 0.783, "learning_rate": 4.036146260724503e-06, "epoch": 0.8872727272727273, "percentage": 29.6, "elapsed_time": "13:01:02", "remaining_time": "1 day, 6:57:32", "throughput": "17744.02", "total_tokens": 831520768}
{"current_steps": 794, "total_steps": 2679, "loss": 0.8189, "learning_rate": 4.033810448027682e-06, "epoch": 0.8883916083916084, "percentage": 29.64, "elapsed_time": "13:01:56", "remaining_time": "1 day, 6:56:21", "throughput": "17745.83", "total_tokens": 832569344}
{"current_steps": 795, "total_steps": 2679, "loss": 0.8026, "learning_rate": 4.0314724861669e-06, "epoch": 0.8895104895104895, "percentage": 29.68, "elapsed_time": "13:03:01", "remaining_time": "1 day, 6:55:37", "throughput": "17743.57", "total_tokens": 833617920}
{"current_steps": 796, "total_steps": 2679, "loss": 0.6769, "learning_rate": 4.029132378418092e-06, "epoch": 0.8906293706293706, "percentage": 29.71, "elapsed_time": "13:03:53", "remaining_time": "1 day, 6:54:22", "throughput": "17746.05", "total_tokens": 834666496}
{"current_steps": 797, "total_steps": 2679, "loss": 0.8146, "learning_rate": 4.0267901280601985e-06, "epoch": 0.8917482517482518, "percentage": 29.75, "elapsed_time": "13:04:56", "remaining_time": "1 day, 6:53:31", "throughput": "17744.77", "total_tokens": 835715072}
{"current_steps": 798, "total_steps": 2679, "loss": 0.7063, "learning_rate": 4.024445738375164e-06, "epoch": 0.8928671328671328, "percentage": 29.79, "elapsed_time": "13:05:50", "remaining_time": "1 day, 6:52:19", "throughput": "17746.84", "total_tokens": 836763648}
{"current_steps": 799, "total_steps": 2679, "loss": 0.7581, "learning_rate": 4.022099212647933e-06, "epoch": 0.893986013986014, "percentage": 29.82, "elapsed_time": "13:06:43", "remaining_time": "1 day, 6:51:07", "throughput": "17748.81", "total_tokens": 837812224}
{"current_steps": 800, "total_steps": 2679, "loss": 0.7643, "learning_rate": 4.019750554166436e-06, "epoch": 0.8951048951048951, "percentage": 29.86, "elapsed_time": "13:07:49", "remaining_time": "1 day, 6:50:23", "throughput": "17746.49", "total_tokens": 838860800}
{"current_steps": 801, "total_steps": 2679, "loss": 0.6969, "learning_rate": 4.017399766221599e-06, "epoch": 0.8962237762237762, "percentage": 29.9, "elapsed_time": "13:08:43", "remaining_time": "1 day, 6:49:12", "throughput": "17748.41", "total_tokens": 839909376}
{"current_steps": 802, "total_steps": 2679, "loss": 0.7633, "learning_rate": 4.015046852107327e-06, "epoch": 0.8973426573426574, "percentage": 29.94, "elapsed_time": "13:09:44", "remaining_time": "1 day, 6:48:18", "throughput": "17747.51", "total_tokens": 840957952}
{"current_steps": 803, "total_steps": 2679, "loss": 0.7878, "learning_rate": 4.012691815120508e-06, "epoch": 0.8984615384615384, "percentage": 29.97, "elapsed_time": "13:10:44", "remaining_time": "1 day, 6:47:22", "throughput": "17747.08", "total_tokens": 842006528}
{"current_steps": 804, "total_steps": 2679, "loss": 0.7283, "learning_rate": 4.0103346585610015e-06, "epoch": 0.8995804195804196, "percentage": 30.01, "elapsed_time": "13:11:46", "remaining_time": "1 day, 6:46:28", "throughput": "17746.29", "total_tokens": 843055104}
{"current_steps": 805, "total_steps": 2679, "loss": 0.6244, "learning_rate": 4.007975385731637e-06, "epoch": 0.9006993006993007, "percentage": 30.05, "elapsed_time": "13:12:33", "remaining_time": "1 day, 6:45:03", "throughput": "17750.46", "total_tokens": 844103680}
{"current_steps": 806, "total_steps": 2679, "loss": 0.7264, "learning_rate": 4.005613999938211e-06, "epoch": 0.9018181818181819, "percentage": 30.09, "elapsed_time": "13:13:33", "remaining_time": "1 day, 6:44:05", "throughput": "17750.29", "total_tokens": 845152256}
{"current_steps": 807, "total_steps": 2679, "loss": 0.7475, "learning_rate": 4.003250504489481e-06, "epoch": 0.9029370629370629, "percentage": 30.12, "elapsed_time": "13:14:32", "remaining_time": "1 day, 6:43:05", "throughput": "17750.37", "total_tokens": 846200832}
{"current_steps": 808, "total_steps": 2679, "loss": 0.7093, "learning_rate": 4.000884902697159e-06, "epoch": 0.904055944055944, "percentage": 30.16, "elapsed_time": "13:15:32", "remaining_time": "1 day, 6:42:09", "throughput": "17749.94", "total_tokens": 847249408}
{"current_steps": 809, "total_steps": 2679, "loss": 0.6377, "learning_rate": 3.998517197875908e-06, "epoch": 0.9051748251748252, "percentage": 30.2, "elapsed_time": "13:16:20", "remaining_time": "1 day, 6:40:45", "throughput": "17753.97", "total_tokens": 848297984}
{"current_steps": 810, "total_steps": 2679, "loss": 0.619, "learning_rate": 3.996147393343338e-06, "epoch": 0.9062937062937063, "percentage": 30.24, "elapsed_time": "13:17:10", "remaining_time": "1 day, 6:39:24", "throughput": "17757.41", "total_tokens": 849346560}
{"current_steps": 811, "total_steps": 2679, "loss": 0.6694, "learning_rate": 3.993775492420005e-06, "epoch": 0.9074125874125875, "percentage": 30.27, "elapsed_time": "13:18:04", "remaining_time": "1 day, 6:38:13", "throughput": "17759.35", "total_tokens": 850395136}
{"current_steps": 812, "total_steps": 2679, "loss": 0.7513, "learning_rate": 3.9914014984293955e-06, "epoch": 0.9085314685314685, "percentage": 30.31, "elapsed_time": "13:18:58", "remaining_time": "1 day, 6:37:03", "throughput": "17761.08", "total_tokens": 851443712}
{"current_steps": 813, "total_steps": 2679, "loss": 0.648, "learning_rate": 3.989025414697935e-06, "epoch": 0.9096503496503496, "percentage": 30.35, "elapsed_time": "13:19:52", "remaining_time": "1 day, 6:35:53", "throughput": "17762.86", "total_tokens": 852492288}
{"current_steps": 814, "total_steps": 2679, "loss": 0.8291, "learning_rate": 3.986647244554974e-06, "epoch": 0.9107692307692308, "percentage": 30.38, "elapsed_time": "13:20:58", "remaining_time": "1 day, 6:35:08", "throughput": "17760.55", "total_tokens": 853540864}
{"current_steps": 815, "total_steps": 2679, "loss": 0.7154, "learning_rate": 3.984266991332787e-06, "epoch": 0.9118881118881119, "percentage": 30.42, "elapsed_time": "13:21:51", "remaining_time": "1 day, 6:33:56", "throughput": "17762.63", "total_tokens": 854589440}
{"current_steps": 816, "total_steps": 2679, "loss": 0.7503, "learning_rate": 3.981884658366566e-06, "epoch": 0.9130069930069931, "percentage": 30.46, "elapsed_time": "13:22:56", "remaining_time": "1 day, 6:33:11", "throughput": "17760.36", "total_tokens": 855638016}
{"current_steps": 817, "total_steps": 2679, "loss": 0.6959, "learning_rate": 3.9795002489944216e-06, "epoch": 0.9141258741258741, "percentage": 30.5, "elapsed_time": "13:23:51", "remaining_time": "1 day, 6:32:03", "throughput": "17761.94", "total_tokens": 856686592}
{"current_steps": 818, "total_steps": 2679, "loss": 0.6735, "learning_rate": 3.977113766557367e-06, "epoch": 0.9152447552447552, "percentage": 30.53, "elapsed_time": "13:24:39", "remaining_time": "1 day, 6:30:39", "throughput": "17765.97", "total_tokens": 857735168}
{"current_steps": 819, "total_steps": 2679, "loss": 0.7275, "learning_rate": 3.9747252143993265e-06, "epoch": 0.9163636363636364, "percentage": 30.57, "elapsed_time": "13:25:39", "remaining_time": "1 day, 6:29:40", "throughput": "17765.85", "total_tokens": 858783744}
{"current_steps": 820, "total_steps": 2679, "loss": 0.8853, "learning_rate": 3.972334595867122e-06, "epoch": 0.9174825174825175, "percentage": 30.61, "elapsed_time": "13:26:40", "remaining_time": "1 day, 6:28:48", "throughput": "17764.82", "total_tokens": 859832320}
{"current_steps": 821, "total_steps": 2679, "loss": 0.6545, "learning_rate": 3.969941914310469e-06, "epoch": 0.9186013986013986, "percentage": 30.65, "elapsed_time": "13:27:35", "remaining_time": "1 day, 6:27:39", "throughput": "17766.46", "total_tokens": 860880896}
{"current_steps": 822, "total_steps": 2679, "loss": 0.6411, "learning_rate": 3.967547173081976e-06, "epoch": 0.9197202797202797, "percentage": 30.68, "elapsed_time": "13:28:28", "remaining_time": "1 day, 6:26:25", "throughput": "17768.81", "total_tokens": 861929472}
{"current_steps": 823, "total_steps": 2679, "loss": 0.6289, "learning_rate": 3.965150375537137e-06, "epoch": 0.9208391608391608, "percentage": 30.72, "elapsed_time": "13:29:15", "remaining_time": "1 day, 6:25:00", "throughput": "17773.00", "total_tokens": 862978048}
{"current_steps": 824, "total_steps": 2679, "loss": 0.6772, "learning_rate": 3.9627515250343275e-06, "epoch": 0.921958041958042, "percentage": 30.76, "elapsed_time": "13:30:09", "remaining_time": "1 day, 6:23:49", "throughput": "17774.91", "total_tokens": 864026624}
{"current_steps": 825, "total_steps": 2679, "loss": 0.699, "learning_rate": 3.9603506249348e-06, "epoch": 0.9230769230769231, "percentage": 30.8, "elapsed_time": "13:31:03", "remaining_time": "1 day, 6:22:39", "throughput": "17776.76", "total_tokens": 865075200}
{"current_steps": 826, "total_steps": 2679, "loss": 0.7727, "learning_rate": 3.957947678602676e-06, "epoch": 0.9241958041958042, "percentage": 30.83, "elapsed_time": "13:32:04", "remaining_time": "1 day, 6:21:45", "throughput": "17776.00", "total_tokens": 866123776}
{"current_steps": 827, "total_steps": 2679, "loss": 0.7034, "learning_rate": 3.955542689404948e-06, "epoch": 0.9253146853146853, "percentage": 30.87, "elapsed_time": "13:32:58", "remaining_time": "1 day, 6:20:36", "throughput": "17777.66", "total_tokens": 867172352}
{"current_steps": 828, "total_steps": 2679, "loss": 0.6624, "learning_rate": 3.953135660711467e-06, "epoch": 0.9264335664335664, "percentage": 30.91, "elapsed_time": "13:33:51", "remaining_time": "1 day, 6:19:22", "throughput": "17780.06", "total_tokens": 868220928}
{"current_steps": 829, "total_steps": 2679, "loss": 0.6719, "learning_rate": 3.950726595894947e-06, "epoch": 0.9275524475524476, "percentage": 30.94, "elapsed_time": "13:34:44", "remaining_time": "1 day, 6:18:11", "throughput": "17782.08", "total_tokens": 869269504}
{"current_steps": 830, "total_steps": 2679, "loss": 0.6623, "learning_rate": 3.94831549833095e-06, "epoch": 0.9286713286713286, "percentage": 30.98, "elapsed_time": "13:35:38", "remaining_time": "1 day, 6:17:00", "throughput": "17783.94", "total_tokens": 870318080}
{"current_steps": 831, "total_steps": 2679, "loss": 0.7096, "learning_rate": 3.9459023713978895e-06, "epoch": 0.9297902097902098, "percentage": 31.02, "elapsed_time": "13:36:31", "remaining_time": "1 day, 6:15:49", "throughput": "17785.92", "total_tokens": 871366656}
{"current_steps": 832, "total_steps": 2679, "loss": 0.8014, "learning_rate": 3.94348721847702e-06, "epoch": 0.9309090909090909, "percentage": 31.06, "elapsed_time": "13:37:37", "remaining_time": "1 day, 6:15:04", "throughput": "17783.67", "total_tokens": 872415232}
{"current_steps": 833, "total_steps": 2679, "loss": 0.8499, "learning_rate": 3.941070042952437e-06, "epoch": 0.932027972027972, "percentage": 31.09, "elapsed_time": "13:38:39", "remaining_time": "1 day, 6:14:12", "throughput": "17782.44", "total_tokens": 873463808}
{"current_steps": 834, "total_steps": 2679, "loss": 0.7902, "learning_rate": 3.938650848211068e-06, "epoch": 0.9331468531468532, "percentage": 31.13, "elapsed_time": "13:39:36", "remaining_time": "1 day, 6:13:09", "throughput": "17783.19", "total_tokens": 874512384}
{"current_steps": 835, "total_steps": 2679, "loss": 0.8006, "learning_rate": 3.936229637642672e-06, "epoch": 0.9342657342657342, "percentage": 31.17, "elapsed_time": "13:40:41", "remaining_time": "1 day, 6:12:24", "throughput": "17780.96", "total_tokens": 875560960}
{"current_steps": 836, "total_steps": 2679, "loss": 0.8366, "learning_rate": 3.933806414639832e-06, "epoch": 0.9353846153846154, "percentage": 31.21, "elapsed_time": "13:41:47", "remaining_time": "1 day, 6:11:40", "throughput": "17778.46", "total_tokens": 876609536}
{"current_steps": 837, "total_steps": 2679, "loss": 0.6731, "learning_rate": 3.931381182597949e-06, "epoch": 0.9365034965034965, "percentage": 31.24, "elapsed_time": "13:42:40", "remaining_time": "1 day, 6:10:28", "throughput": "17780.53", "total_tokens": 877658112}
{"current_steps": 838, "total_steps": 2679, "loss": 0.8729, "learning_rate": 3.928953944915242e-06, "epoch": 0.9376223776223777, "percentage": 31.28, "elapsed_time": "13:43:45", "remaining_time": "1 day, 6:09:43", "throughput": "17778.24", "total_tokens": 878706688}
{"current_steps": 839, "total_steps": 2679, "loss": 0.7003, "learning_rate": 3.926524704992736e-06, "epoch": 0.9387412587412587, "percentage": 31.32, "elapsed_time": "13:44:39", "remaining_time": "1 day, 6:08:33", "throughput": "17780.17", "total_tokens": 879755264}
{"current_steps": 840, "total_steps": 2679, "loss": 0.7721, "learning_rate": 3.9240934662342665e-06, "epoch": 0.9398601398601398, "percentage": 31.35, "elapsed_time": "13:45:39", "remaining_time": "1 day, 6:07:35", "throughput": "17779.89", "total_tokens": 880803840}
{"current_steps": 841, "total_steps": 2679, "loss": 0.7873, "learning_rate": 3.9216602320464655e-06, "epoch": 0.940979020979021, "percentage": 31.39, "elapsed_time": "13:46:35", "remaining_time": "1 day, 6:06:29", "throughput": "17781.02", "total_tokens": 881852416}
{"current_steps": 842, "total_steps": 2679, "loss": 0.7316, "learning_rate": 3.9192250058387656e-06, "epoch": 0.9420979020979021, "percentage": 31.43, "elapsed_time": "13:47:29", "remaining_time": "1 day, 6:05:20", "throughput": "17782.76", "total_tokens": 882900992}
{"current_steps": 843, "total_steps": 2679, "loss": 0.6785, "learning_rate": 3.916787791023386e-06, "epoch": 0.9432167832167833, "percentage": 31.47, "elapsed_time": "13:48:22", "remaining_time": "1 day, 6:04:09", "throughput": "17784.66", "total_tokens": 883949568}
{"current_steps": 844, "total_steps": 2679, "loss": 0.7708, "learning_rate": 3.914348591015335e-06, "epoch": 0.9443356643356643, "percentage": 31.5, "elapsed_time": "13:49:25", "remaining_time": "1 day, 6:03:18", "throughput": "17783.45", "total_tokens": 884998144}
{"current_steps": 845, "total_steps": 2679, "loss": 0.7246, "learning_rate": 3.911907409232402e-06, "epoch": 0.9454545454545454, "percentage": 31.54, "elapsed_time": "13:50:23", "remaining_time": "1 day, 6:02:18", "throughput": "17783.54", "total_tokens": 886046720}
{"current_steps": 846, "total_steps": 2679, "loss": 0.7081, "learning_rate": 3.90946424909515e-06, "epoch": 0.9465734265734266, "percentage": 31.58, "elapsed_time": "13:51:18", "remaining_time": "1 day, 6:01:09", "throughput": "17785.23", "total_tokens": 887095296}
{"current_steps": 847, "total_steps": 2679, "loss": 0.7986, "learning_rate": 3.907019114026922e-06, "epoch": 0.9476923076923077, "percentage": 31.62, "elapsed_time": "13:52:17", "remaining_time": "1 day, 6:00:11", "throughput": "17785.00", "total_tokens": 888143872}
{"current_steps": 848, "total_steps": 2679, "loss": 0.8037, "learning_rate": 3.90457200745382e-06, "epoch": 0.9488111888111889, "percentage": 31.65, "elapsed_time": "13:53:16", "remaining_time": "1 day, 5:59:12", "throughput": "17785.04", "total_tokens": 889192448}
{"current_steps": 849, "total_steps": 2679, "loss": 0.8904, "learning_rate": 3.902122932804713e-06, "epoch": 0.9499300699300699, "percentage": 31.69, "elapsed_time": "13:54:24", "remaining_time": "1 day, 5:58:33", "throughput": "17781.71", "total_tokens": 890241024}
{"current_steps": 850, "total_steps": 2679, "loss": 0.7235, "learning_rate": 3.899671893511226e-06, "epoch": 0.951048951048951, "percentage": 31.73, "elapsed_time": "13:55:24", "remaining_time": "1 day, 5:57:36", "throughput": "17781.53", "total_tokens": 891289600}
{"current_steps": 851, "total_steps": 2679, "loss": 0.8046, "learning_rate": 3.897218893007737e-06, "epoch": 0.9521678321678322, "percentage": 31.77, "elapsed_time": "13:56:30", "remaining_time": "1 day, 5:56:51", "throughput": "17779.14", "total_tokens": 892338176}
{"current_steps": 852, "total_steps": 2679, "loss": 0.8139, "learning_rate": 3.894763934731373e-06, "epoch": 0.9532867132867133, "percentage": 31.8, "elapsed_time": "13:57:35", "remaining_time": "1 day, 5:56:06", "throughput": "17776.92", "total_tokens": 893386752}
{"current_steps": 853, "total_steps": 2679, "loss": 0.6991, "learning_rate": 3.8923070221220035e-06, "epoch": 0.9544055944055944, "percentage": 31.84, "elapsed_time": "13:58:35", "remaining_time": "1 day, 5:55:08", "throughput": "17776.70", "total_tokens": 894435328}
{"current_steps": 854, "total_steps": 2679, "loss": 0.7715, "learning_rate": 3.889848158622237e-06, "epoch": 0.9555244755244755, "percentage": 31.88, "elapsed_time": "13:59:33", "remaining_time": "1 day, 5:54:08", "throughput": "17776.88", "total_tokens": 895483904}
{"current_steps": 855, "total_steps": 2679, "loss": 0.677, "learning_rate": 3.887387347677413e-06, "epoch": 0.9566433566433566, "percentage": 31.91, "elapsed_time": "14:00:27", "remaining_time": "1 day, 5:52:58", "throughput": "17778.72", "total_tokens": 896532480}
{"current_steps": 856, "total_steps": 2679, "loss": 0.7913, "learning_rate": 3.884924592735604e-06, "epoch": 0.9577622377622378, "percentage": 31.95, "elapsed_time": "14:01:26", "remaining_time": "1 day, 5:51:58", "throughput": "17778.78", "total_tokens": 897581056}
{"current_steps": 857, "total_steps": 2679, "loss": 0.8272, "learning_rate": 3.882459897247603e-06, "epoch": 0.9588811188811189, "percentage": 31.99, "elapsed_time": "14:02:33", "remaining_time": "1 day, 5:51:17", "throughput": "17775.82", "total_tokens": 898629632}
{"current_steps": 858, "total_steps": 2679, "loss": 0.6715, "learning_rate": 3.8799932646669235e-06, "epoch": 0.96, "percentage": 32.03, "elapsed_time": "14:03:28", "remaining_time": "1 day, 5:50:09", "throughput": "17777.35", "total_tokens": 899678208}
{"current_steps": 859, "total_steps": 2679, "loss": 0.7166, "learning_rate": 3.8775246984497924e-06, "epoch": 0.9611188811188811, "percentage": 32.06, "elapsed_time": "14:04:22", "remaining_time": "1 day, 5:49:00", "throughput": "17779.02", "total_tokens": 900726784}
{"current_steps": 860, "total_steps": 2679, "loss": 0.7947, "learning_rate": 3.875054202055148e-06, "epoch": 0.9622377622377623, "percentage": 32.1, "elapsed_time": "14:05:22", "remaining_time": "1 day, 5:48:02", "throughput": "17778.77", "total_tokens": 901775360}
{"current_steps": 861, "total_steps": 2679, "loss": 0.6249, "learning_rate": 3.87258177894463e-06, "epoch": 0.9633566433566434, "percentage": 32.14, "elapsed_time": "14:06:09", "remaining_time": "1 day, 5:46:40", "throughput": "17782.66", "total_tokens": 902823936}
{"current_steps": 862, "total_steps": 2679, "loss": 0.6987, "learning_rate": 3.870107432582582e-06, "epoch": 0.9644755244755244, "percentage": 32.18, "elapsed_time": "14:07:04", "remaining_time": "1 day, 5:45:31", "throughput": "17784.34", "total_tokens": 903872512}
{"current_steps": 863, "total_steps": 2679, "loss": 0.6279, "learning_rate": 3.867631166436038e-06, "epoch": 0.9655944055944056, "percentage": 32.21, "elapsed_time": "14:07:53", "remaining_time": "1 day, 5:44:12", "throughput": "17787.60", "total_tokens": 904921088}
{"current_steps": 864, "total_steps": 2679, "loss": 0.7039, "learning_rate": 3.865152983974724e-06, "epoch": 0.9667132867132867, "percentage": 32.25, "elapsed_time": "14:08:52", "remaining_time": "1 day, 5:43:14", "throughput": "17787.54", "total_tokens": 905969664}
{"current_steps": 865, "total_steps": 2679, "loss": 0.8572, "learning_rate": 3.862672888671051e-06, "epoch": 0.9678321678321679, "percentage": 32.29, "elapsed_time": "14:09:58", "remaining_time": "1 day, 5:42:28", "throughput": "17785.29", "total_tokens": 907018240}
{"current_steps": 866, "total_steps": 2679, "loss": 0.7775, "learning_rate": 3.860190884000112e-06, "epoch": 0.968951048951049, "percentage": 32.33, "elapsed_time": "14:10:52", "remaining_time": "1 day, 5:41:20", "throughput": "17786.92", "total_tokens": 908066816}
{"current_steps": 867, "total_steps": 2679, "loss": 0.7948, "learning_rate": 3.857706973439672e-06, "epoch": 0.97006993006993, "percentage": 32.36, "elapsed_time": "14:11:57", "remaining_time": "1 day, 5:40:34", "throughput": "17784.68", "total_tokens": 909115392}
{"current_steps": 868, "total_steps": 2679, "loss": 0.8362, "learning_rate": 3.85522116047017e-06, "epoch": 0.9711888111888112, "percentage": 32.4, "elapsed_time": "14:12:57", "remaining_time": "1 day, 5:39:37", "throughput": "17784.40", "total_tokens": 910163968}
{"current_steps": 869, "total_steps": 2679, "loss": 0.6686, "learning_rate": 3.852733448574707e-06, "epoch": 0.9723076923076923, "percentage": 32.44, "elapsed_time": "14:13:45", "remaining_time": "1 day, 5:38:14", "throughput": "17788.41", "total_tokens": 911212544}
{"current_steps": 870, "total_steps": 2679, "loss": 0.6899, "learning_rate": 3.850243841239047e-06, "epoch": 0.9734265734265735, "percentage": 32.47, "elapsed_time": "14:14:38", "remaining_time": "1 day, 5:37:04", "throughput": "17790.26", "total_tokens": 912261120}
{"current_steps": 871, "total_steps": 2679, "loss": 0.6706, "learning_rate": 3.8477523419516115e-06, "epoch": 0.9745454545454545, "percentage": 32.51, "elapsed_time": "14:15:26", "remaining_time": "1 day, 5:35:43", "throughput": "17793.98", "total_tokens": 913309696}
{"current_steps": 872, "total_steps": 2679, "loss": 0.6796, "learning_rate": 3.8452589542034686e-06, "epoch": 0.9756643356643356, "percentage": 32.55, "elapsed_time": "14:16:20", "remaining_time": "1 day, 5:34:33", "throughput": "17795.76", "total_tokens": 914358272}
{"current_steps": 873, "total_steps": 2679, "loss": 0.8015, "learning_rate": 3.842763681488337e-06, "epoch": 0.9767832167832168, "percentage": 32.59, "elapsed_time": "14:17:25", "remaining_time": "1 day, 5:33:47", "throughput": "17793.61", "total_tokens": 915406848}
{"current_steps": 874, "total_steps": 2679, "loss": 0.705, "learning_rate": 3.8402665273025726e-06, "epoch": 0.9779020979020979, "percentage": 32.62, "elapsed_time": "14:18:20", "remaining_time": "1 day, 5:32:39", "throughput": "17795.16", "total_tokens": 916455424}
{"current_steps": 875, "total_steps": 2679, "loss": 0.6312, "learning_rate": 3.837767495145171e-06, "epoch": 0.9790209790209791, "percentage": 32.66, "elapsed_time": "14:19:13", "remaining_time": "1 day, 5:31:29", "throughput": "17797.03", "total_tokens": 917504000}
{"current_steps": 876, "total_steps": 2679, "loss": 0.7085, "learning_rate": 3.835266588517757e-06, "epoch": 0.9801398601398601, "percentage": 32.7, "elapsed_time": "14:20:08", "remaining_time": "1 day, 5:30:21", "throughput": "17798.55", "total_tokens": 918552576}
{"current_steps": 877, "total_steps": 2679, "loss": 0.7812, "learning_rate": 3.832763810924583e-06, "epoch": 0.9812587412587412, "percentage": 32.74, "elapsed_time": "14:21:08", "remaining_time": "1 day, 5:29:24", "throughput": "17798.19", "total_tokens": 919601152}
{"current_steps": 878, "total_steps": 2679, "loss": 0.6654, "learning_rate": 3.830259165872523e-06, "epoch": 0.9823776223776224, "percentage": 32.77, "elapsed_time": "14:22:01", "remaining_time": "1 day, 5:28:14", "throughput": "17800.02", "total_tokens": 920649728}
{"current_steps": 879, "total_steps": 2679, "loss": 1.0061, "learning_rate": 3.827752656871067e-06, "epoch": 0.9834965034965035, "percentage": 32.81, "elapsed_time": "14:23:09", "remaining_time": "1 day, 5:27:32", "throughput": "17797.11", "total_tokens": 921698304}
{"current_steps": 880, "total_steps": 2679, "loss": 0.6913, "learning_rate": 3.825244287432316e-06, "epoch": 0.9846153846153847, "percentage": 32.85, "elapsed_time": "14:23:58", "remaining_time": "1 day, 5:26:14", "throughput": "17800.38", "total_tokens": 922746880}
{"current_steps": 881, "total_steps": 2679, "loss": 0.6779, "learning_rate": 3.822734061070979e-06, "epoch": 0.9857342657342657, "percentage": 32.89, "elapsed_time": "14:24:52", "remaining_time": "1 day, 5:25:05", "throughput": "17802.08", "total_tokens": 923795456}
{"current_steps": 882, "total_steps": 2679, "loss": 0.6848, "learning_rate": 3.82022198130437e-06, "epoch": 0.9868531468531468, "percentage": 32.92, "elapsed_time": "14:25:46", "remaining_time": "1 day, 5:23:57", "throughput": "17803.64", "total_tokens": 924844032}
{"current_steps": 883, "total_steps": 2679, "loss": 0.809, "learning_rate": 3.817708051652392e-06, "epoch": 0.987972027972028, "percentage": 32.96, "elapsed_time": "14:26:46", "remaining_time": "1 day, 5:23:00", "throughput": "17803.38", "total_tokens": 925892608}
{"current_steps": 884, "total_steps": 2679, "loss": 0.8708, "learning_rate": 3.8151922756375485e-06, "epoch": 0.9890909090909091, "percentage": 33.0, "elapsed_time": "14:27:52", "remaining_time": "1 day, 5:22:14", "throughput": "17801.12", "total_tokens": 926941184}
{"current_steps": 885, "total_steps": 2679, "loss": 0.7192, "learning_rate": 3.812674656784924e-06, "epoch": 0.9902097902097902, "percentage": 33.03, "elapsed_time": "14:28:46", "remaining_time": "1 day, 5:21:06", "throughput": "17802.74", "total_tokens": 927989760}
{"current_steps": 886, "total_steps": 2679, "loss": 0.6901, "learning_rate": 3.8101551986221896e-06, "epoch": 0.9913286713286713, "percentage": 33.07, "elapsed_time": "14:29:36", "remaining_time": "1 day, 5:19:49", "throughput": "17805.64", "total_tokens": 929038336}
{"current_steps": 887, "total_steps": 2679, "loss": 0.6654, "learning_rate": 3.8076339046795897e-06, "epoch": 0.9924475524475525, "percentage": 33.11, "elapsed_time": "14:30:34", "remaining_time": "1 day, 5:18:49", "throughput": "17805.95", "total_tokens": 930086912}
{"current_steps": 888, "total_steps": 2679, "loss": 0.6568, "learning_rate": 3.8051107784899443e-06, "epoch": 0.9935664335664336, "percentage": 33.15, "elapsed_time": "14:31:27", "remaining_time": "1 day, 5:17:39", "throughput": "17807.82", "total_tokens": 931135488}
{"current_steps": 889, "total_steps": 2679, "loss": 0.6623, "learning_rate": 3.8025858235886394e-06, "epoch": 0.9946853146853147, "percentage": 33.18, "elapsed_time": "14:32:17", "remaining_time": "1 day, 5:16:21", "throughput": "17811.01", "total_tokens": 932184064}
{"current_steps": 890, "total_steps": 2679, "loss": 0.6758, "learning_rate": 3.8000590435136213e-06, "epoch": 0.9958041958041958, "percentage": 33.22, "elapsed_time": "14:33:16", "remaining_time": "1 day, 5:15:23", "throughput": "17810.89", "total_tokens": 933232640}
{"current_steps": 891, "total_steps": 2679, "loss": 0.6512, "learning_rate": 3.7975304418053986e-06, "epoch": 0.9969230769230769, "percentage": 33.26, "elapsed_time": "14:34:10", "remaining_time": "1 day, 5:14:14", "throughput": "17812.62", "total_tokens": 934281216}
{"current_steps": 892, "total_steps": 2679, "loss": 0.7217, "learning_rate": 3.795000022007027e-06, "epoch": 0.998041958041958, "percentage": 33.3, "elapsed_time": "14:35:08", "remaining_time": "1 day, 5:13:13", "throughput": "17813.00", "total_tokens": 935329792}
{"current_steps": 893, "total_steps": 2679, "loss": 0.8638, "learning_rate": 3.7924677876641147e-06, "epoch": 0.9991608391608392, "percentage": 33.33, "elapsed_time": "14:36:13", "remaining_time": "1 day, 5:12:27", "throughput": "17810.70", "total_tokens": 936378368}
{"current_steps": 894, "total_steps": 2679, "loss": 0.7973, "learning_rate": 3.789933742324807e-06, "epoch": 1.0002797202797202, "percentage": 33.37, "elapsed_time": "14:37:08", "remaining_time": "1 day, 5:11:20", "throughput": "17812.14", "total_tokens": 937426944}
{"current_steps": 895, "total_steps": 2679, "loss": 0.7068, "learning_rate": 3.787397889539792e-06, "epoch": 1.0013986013986014, "percentage": 33.41, "elapsed_time": "14:38:03", "remaining_time": "1 day, 5:10:14", "throughput": "17813.42", "total_tokens": 938475520}
{"current_steps": 896, "total_steps": 2679, "loss": 0.6657, "learning_rate": 3.7848602328622864e-06, "epoch": 1.0025174825174825, "percentage": 33.45, "elapsed_time": "14:38:57", "remaining_time": "1 day, 5:09:05", "throughput": "17815.04", "total_tokens": 939524096}
{"current_steps": 897, "total_steps": 2679, "loss": 0.6522, "learning_rate": 3.782320775848038e-06, "epoch": 1.0036363636363637, "percentage": 33.48, "elapsed_time": "14:39:51", "remaining_time": "1 day, 5:07:57", "throughput": "17816.68", "total_tokens": 940572672}
{"current_steps": 898, "total_steps": 2679, "loss": 0.696, "learning_rate": 3.7797795220553136e-06, "epoch": 1.0047552447552448, "percentage": 33.52, "elapsed_time": "14:40:45", "remaining_time": "1 day, 5:06:48", "throughput": "17818.39", "total_tokens": 941621248}
{"current_steps": 899, "total_steps": 2679, "loss": 0.655, "learning_rate": 3.7772364750449002e-06, "epoch": 1.005874125874126, "percentage": 33.56, "elapsed_time": "14:41:39", "remaining_time": "1 day, 5:05:39", "throughput": "17820.10", "total_tokens": 942669824}
{"current_steps": 900, "total_steps": 2679, "loss": 0.6671, "learning_rate": 3.774691638380096e-06, "epoch": 1.006993006993007, "percentage": 33.59, "elapsed_time": "14:42:33", "remaining_time": "1 day, 5:04:31", "throughput": "17821.67", "total_tokens": 943718400}
{"current_steps": 901, "total_steps": 2679, "loss": 0.7109, "learning_rate": 3.772145015626709e-06, "epoch": 1.008111888111888, "percentage": 33.63, "elapsed_time": "14:43:32", "remaining_time": "1 day, 5:03:33", "throughput": "17821.44", "total_tokens": 944766976}
{"current_steps": 902, "total_steps": 2679, "loss": 0.6671, "learning_rate": 3.769596610353047e-06, "epoch": 1.0092307692307692, "percentage": 33.67, "elapsed_time": "14:44:26", "remaining_time": "1 day, 5:02:25", "throughput": "17823.06", "total_tokens": 945815552}
{"current_steps": 903, "total_steps": 2679, "loss": 0.7918, "learning_rate": 3.767046426129917e-06, "epoch": 1.0103496503496503, "percentage": 33.71, "elapsed_time": "14:45:37", "remaining_time": "1 day, 5:01:49", "throughput": "17819.07", "total_tokens": 946864128}
{"current_steps": 904, "total_steps": 2679, "loss": 0.7056, "learning_rate": 3.764494466530618e-06, "epoch": 1.0114685314685314, "percentage": 33.74, "elapsed_time": "14:46:33", "remaining_time": "1 day, 5:00:45", "throughput": "17820.10", "total_tokens": 947912704}
{"current_steps": 905, "total_steps": 2679, "loss": 0.8757, "learning_rate": 3.7619407351309377e-06, "epoch": 1.0125874125874126, "percentage": 33.78, "elapsed_time": "14:47:44", "remaining_time": "1 day, 5:00:10", "throughput": "17816.03", "total_tokens": 948961280}
{"current_steps": 906, "total_steps": 2679, "loss": 0.7706, "learning_rate": 3.7593852355091463e-06, "epoch": 1.0137062937062937, "percentage": 33.82, "elapsed_time": "14:48:49", "remaining_time": "1 day, 4:59:23", "throughput": "17813.90", "total_tokens": 950009856}
{"current_steps": 907, "total_steps": 2679, "loss": 0.676, "learning_rate": 3.7568279712459908e-06, "epoch": 1.0148251748251749, "percentage": 33.86, "elapsed_time": "14:49:49", "remaining_time": "1 day, 4:58:25", "throughput": "17813.72", "total_tokens": 951058432}
{"current_steps": 908, "total_steps": 2679, "loss": 0.6838, "learning_rate": 3.7542689459246907e-06, "epoch": 1.015944055944056, "percentage": 33.89, "elapsed_time": "14:50:48", "remaining_time": "1 day, 4:57:28", "throughput": "17813.49", "total_tokens": 952107008}
{"current_steps": 909, "total_steps": 2679, "loss": 0.6933, "learning_rate": 3.7517081631309336e-06, "epoch": 1.0170629370629372, "percentage": 33.93, "elapsed_time": "14:51:44", "remaining_time": "1 day, 4:56:23", "throughput": "17814.56", "total_tokens": 953155584}
{"current_steps": 910, "total_steps": 2679, "loss": 0.762, "learning_rate": 3.7491456264528703e-06, "epoch": 1.018181818181818, "percentage": 33.97, "elapsed_time": "14:52:49", "remaining_time": "1 day, 4:55:36", "throughput": "17812.53", "total_tokens": 954204160}
{"current_steps": 911, "total_steps": 2679, "loss": 0.6466, "learning_rate": 3.746581339481108e-06, "epoch": 1.0193006993006992, "percentage": 34.01, "elapsed_time": "14:53:43", "remaining_time": "1 day, 4:54:27", "throughput": "17814.16", "total_tokens": 955252736}
{"current_steps": 912, "total_steps": 2679, "loss": 0.6643, "learning_rate": 3.7440153058087064e-06, "epoch": 1.0204195804195804, "percentage": 34.04, "elapsed_time": "14:54:37", "remaining_time": "1 day, 4:53:20", "throughput": "17815.63", "total_tokens": 956301312}
{"current_steps": 913, "total_steps": 2679, "loss": 0.6281, "learning_rate": 3.741447529031173e-06, "epoch": 1.0215384615384615, "percentage": 34.08, "elapsed_time": "14:55:25", "remaining_time": "1 day, 4:52:00", "throughput": "17819.18", "total_tokens": 957349888}
{"current_steps": 914, "total_steps": 2679, "loss": 0.6409, "learning_rate": 3.7388780127464586e-06, "epoch": 1.0226573426573426, "percentage": 34.12, "elapsed_time": "14:56:19", "remaining_time": "1 day, 4:50:53", "throughput": "17820.72", "total_tokens": 958398464}
{"current_steps": 915, "total_steps": 2679, "loss": 0.7178, "learning_rate": 3.7363067605549515e-06, "epoch": 1.0237762237762238, "percentage": 34.15, "elapsed_time": "14:57:21", "remaining_time": "1 day, 4:49:58", "throughput": "17819.91", "total_tokens": 959447040}
{"current_steps": 916, "total_steps": 2679, "loss": 0.661, "learning_rate": 3.733733776059468e-06, "epoch": 1.024895104895105, "percentage": 34.19, "elapsed_time": "14:58:14", "remaining_time": "1 day, 4:48:50", "throughput": "17821.63", "total_tokens": 960495616}
{"current_steps": 917, "total_steps": 2679, "loss": 0.6455, "learning_rate": 3.7311590628652584e-06, "epoch": 1.026013986013986, "percentage": 34.23, "elapsed_time": "14:59:08", "remaining_time": "1 day, 4:47:41", "throughput": "17823.32", "total_tokens": 961544192}
{"current_steps": 918, "total_steps": 2679, "loss": 0.6748, "learning_rate": 3.7285826245799904e-06, "epoch": 1.0271328671328672, "percentage": 34.27, "elapsed_time": "15:00:02", "remaining_time": "1 day, 4:46:33", "throughput": "17824.86", "total_tokens": 962592768}
{"current_steps": 919, "total_steps": 2679, "loss": 0.6433, "learning_rate": 3.726004464813752e-06, "epoch": 1.0282517482517481, "percentage": 34.3, "elapsed_time": "15:00:56", "remaining_time": "1 day, 4:45:25", "throughput": "17826.48", "total_tokens": 963641344}
{"current_steps": 920, "total_steps": 2679, "loss": 0.7869, "learning_rate": 3.723424587179039e-06, "epoch": 1.0293706293706293, "percentage": 34.34, "elapsed_time": "15:01:56", "remaining_time": "1 day, 4:44:28", "throughput": "17826.11", "total_tokens": 964689920}
{"current_steps": 921, "total_steps": 2679, "loss": 0.8307, "learning_rate": 3.72084299529076e-06, "epoch": 1.0304895104895104, "percentage": 34.38, "elapsed_time": "15:03:07", "remaining_time": "1 day, 4:43:53", "throughput": "17822.04", "total_tokens": 965738496}
{"current_steps": 922, "total_steps": 2679, "loss": 0.5991, "learning_rate": 3.718259692766221e-06, "epoch": 1.0316083916083916, "percentage": 34.42, "elapsed_time": "15:03:55", "remaining_time": "1 day, 4:42:34", "throughput": "17825.58", "total_tokens": 966787072}
{"current_steps": 923, "total_steps": 2679, "loss": 0.8823, "learning_rate": 3.7156746832251266e-06, "epoch": 1.0327272727272727, "percentage": 34.45, "elapsed_time": "15:05:01", "remaining_time": "1 day, 4:41:48", "throughput": "17823.29", "total_tokens": 967835648}
{"current_steps": 924, "total_steps": 2679, "loss": 0.8461, "learning_rate": 3.7130879702895733e-06, "epoch": 1.0338461538461539, "percentage": 34.49, "elapsed_time": "15:06:07", "remaining_time": "1 day, 4:41:02", "throughput": "17821.02", "total_tokens": 968884224}
{"current_steps": 925, "total_steps": 2679, "loss": 0.7179, "learning_rate": 3.710499557584045e-06, "epoch": 1.034965034965035, "percentage": 34.53, "elapsed_time": "15:07:06", "remaining_time": "1 day, 4:40:05", "throughput": "17820.86", "total_tokens": 969932800}
{"current_steps": 926, "total_steps": 2679, "loss": 0.785, "learning_rate": 3.7079094487354055e-06, "epoch": 1.0360839160839161, "percentage": 34.57, "elapsed_time": "15:08:06", "remaining_time": "1 day, 4:39:08", "throughput": "17820.54", "total_tokens": 970981376}
{"current_steps": 927, "total_steps": 2679, "loss": 0.7238, "learning_rate": 3.705317647372898e-06, "epoch": 1.0372027972027973, "percentage": 34.6, "elapsed_time": "15:09:12", "remaining_time": "1 day, 4:38:22", "throughput": "17818.31", "total_tokens": 972029952}
{"current_steps": 928, "total_steps": 2679, "loss": 0.6425, "learning_rate": 3.702724157128135e-06, "epoch": 1.0383216783216782, "percentage": 34.64, "elapsed_time": "15:10:05", "remaining_time": "1 day, 4:37:13", "throughput": "17820.05", "total_tokens": 973078528}
{"current_steps": 929, "total_steps": 2679, "loss": 0.6641, "learning_rate": 3.700128981635094e-06, "epoch": 1.0394405594405594, "percentage": 34.68, "elapsed_time": "15:11:00", "remaining_time": "1 day, 4:36:07", "throughput": "17821.29", "total_tokens": 974127104}
{"current_steps": 930, "total_steps": 2679, "loss": 0.7375, "learning_rate": 3.6975321245301183e-06, "epoch": 1.0405594405594405, "percentage": 34.71, "elapsed_time": "15:12:05", "remaining_time": "1 day, 4:35:20", "throughput": "17819.25", "total_tokens": 975175680}
{"current_steps": 931, "total_steps": 2679, "loss": 0.7991, "learning_rate": 3.6949335894519033e-06, "epoch": 1.0416783216783216, "percentage": 34.75, "elapsed_time": "15:13:11", "remaining_time": "1 day, 4:34:33", "throughput": "17817.11", "total_tokens": 976224256}
{"current_steps": 932, "total_steps": 2679, "loss": 0.6662, "learning_rate": 3.6923333800414997e-06, "epoch": 1.0427972027972028, "percentage": 34.79, "elapsed_time": "15:14:16", "remaining_time": "1 day, 4:33:46", "throughput": "17815.15", "total_tokens": 977272832}
{"current_steps": 933, "total_steps": 2679, "loss": 0.6159, "learning_rate": 3.6897314999423e-06, "epoch": 1.043916083916084, "percentage": 34.83, "elapsed_time": "15:15:04", "remaining_time": "1 day, 4:32:26", "throughput": "17818.69", "total_tokens": 978321408}
{"current_steps": 934, "total_steps": 2679, "loss": 0.8542, "learning_rate": 3.68712795280004e-06, "epoch": 1.045034965034965, "percentage": 34.86, "elapsed_time": "15:16:03", "remaining_time": "1 day, 4:31:29", "throughput": "17818.44", "total_tokens": 979369984}
{"current_steps": 935, "total_steps": 2679, "loss": 0.6883, "learning_rate": 3.6845227422627904e-06, "epoch": 1.0461538461538462, "percentage": 34.9, "elapsed_time": "15:17:03", "remaining_time": "1 day, 4:30:32", "throughput": "17818.15", "total_tokens": 980418560}
{"current_steps": 936, "total_steps": 2679, "loss": 0.7343, "learning_rate": 3.681915871980954e-06, "epoch": 1.0472727272727274, "percentage": 34.94, "elapsed_time": "15:18:03", "remaining_time": "1 day, 4:29:34", "throughput": "17817.91", "total_tokens": 981467136}
{"current_steps": 937, "total_steps": 2679, "loss": 0.7385, "learning_rate": 3.679307345607257e-06, "epoch": 1.0483916083916085, "percentage": 34.98, "elapsed_time": "15:19:03", "remaining_time": "1 day, 4:28:38", "throughput": "17817.47", "total_tokens": 982515712}
{"current_steps": 938, "total_steps": 2679, "loss": 0.6768, "learning_rate": 3.676697166796749e-06, "epoch": 1.0495104895104894, "percentage": 35.01, "elapsed_time": "15:19:58", "remaining_time": "1 day, 4:27:32", "throughput": "17818.64", "total_tokens": 983564288}
{"current_steps": 939, "total_steps": 2679, "loss": 0.5939, "learning_rate": 3.6740853392067925e-06, "epoch": 1.0506293706293706, "percentage": 35.05, "elapsed_time": "15:20:47", "remaining_time": "1 day, 4:26:15", "throughput": "17821.87", "total_tokens": 984612864}
{"current_steps": 940, "total_steps": 2679, "loss": 0.6575, "learning_rate": 3.6714718664970624e-06, "epoch": 1.0517482517482517, "percentage": 35.09, "elapsed_time": "15:21:41", "remaining_time": "1 day, 4:25:07", "throughput": "17823.45", "total_tokens": 985661440}
{"current_steps": 941, "total_steps": 2679, "loss": 0.6733, "learning_rate": 3.6688567523295356e-06, "epoch": 1.0528671328671328, "percentage": 35.13, "elapsed_time": "15:22:35", "remaining_time": "1 day, 4:23:59", "throughput": "17825.04", "total_tokens": 986710016}
{"current_steps": 942, "total_steps": 2679, "loss": 0.7445, "learning_rate": 3.6662400003684915e-06, "epoch": 1.053986013986014, "percentage": 35.16, "elapsed_time": "15:23:34", "remaining_time": "1 day, 4:23:01", "throughput": "17824.85", "total_tokens": 987758592}
{"current_steps": 943, "total_steps": 2679, "loss": 0.7437, "learning_rate": 3.663621614280505e-06, "epoch": 1.0551048951048951, "percentage": 35.2, "elapsed_time": "15:24:34", "remaining_time": "1 day, 4:22:04", "throughput": "17824.57", "total_tokens": 988807168}
{"current_steps": 944, "total_steps": 2679, "loss": 0.8643, "learning_rate": 3.661001597734438e-06, "epoch": 1.0562237762237763, "percentage": 35.24, "elapsed_time": "15:25:35", "remaining_time": "1 day, 4:21:09", "throughput": "17823.86", "total_tokens": 989855744}
{"current_steps": 945, "total_steps": 2679, "loss": 0.6972, "learning_rate": 3.6583799544014397e-06, "epoch": 1.0573426573426574, "percentage": 35.27, "elapsed_time": "15:26:36", "remaining_time": "1 day, 4:20:15", "throughput": "17823.16", "total_tokens": 990904320}
{"current_steps": 946, "total_steps": 2679, "loss": 0.7712, "learning_rate": 3.655756687954937e-06, "epoch": 1.0584615384615386, "percentage": 35.31, "elapsed_time": "15:27:41", "remaining_time": "1 day, 4:19:28", "throughput": "17821.08", "total_tokens": 991952896}
{"current_steps": 947, "total_steps": 2679, "loss": 0.7906, "learning_rate": 3.653131802070631e-06, "epoch": 1.0595804195804195, "percentage": 35.35, "elapsed_time": "15:28:42", "remaining_time": "1 day, 4:18:32", "throughput": "17820.53", "total_tokens": 993001472}
{"current_steps": 948, "total_steps": 2679, "loss": 0.6038, "learning_rate": 3.6505053004264936e-06, "epoch": 1.0606993006993006, "percentage": 35.39, "elapsed_time": "15:29:30", "remaining_time": "1 day, 4:17:14", "throughput": "17823.95", "total_tokens": 994050048}
{"current_steps": 949, "total_steps": 2679, "loss": 0.7082, "learning_rate": 3.6478771867027585e-06, "epoch": 1.0618181818181818, "percentage": 35.42, "elapsed_time": "15:30:35", "remaining_time": "1 day, 4:16:26", "throughput": "17821.94", "total_tokens": 995098624}
{"current_steps": 950, "total_steps": 2679, "loss": 0.6744, "learning_rate": 3.64524746458192e-06, "epoch": 1.062937062937063, "percentage": 35.46, "elapsed_time": "15:31:29", "remaining_time": "1 day, 4:15:19", "throughput": "17823.41", "total_tokens": 996147200}
{"current_steps": 951, "total_steps": 2679, "loss": 0.7289, "learning_rate": 3.642616137748727e-06, "epoch": 1.064055944055944, "percentage": 35.5, "elapsed_time": "15:32:29", "remaining_time": "1 day, 4:14:22", "throughput": "17823.10", "total_tokens": 997195776}
{"current_steps": 952, "total_steps": 2679, "loss": 0.7373, "learning_rate": 3.6399832098901726e-06, "epoch": 1.0651748251748252, "percentage": 35.54, "elapsed_time": "15:33:30", "remaining_time": "1 day, 4:13:27", "throughput": "17822.38", "total_tokens": 998244352}
{"current_steps": 953, "total_steps": 2679, "loss": 0.7249, "learning_rate": 3.637348684695498e-06, "epoch": 1.0662937062937063, "percentage": 35.57, "elapsed_time": "15:34:30", "remaining_time": "1 day, 4:12:30", "throughput": "17822.16", "total_tokens": 999292928}
{"current_steps": 954, "total_steps": 2679, "loss": 0.7168, "learning_rate": 3.63471256585618e-06, "epoch": 1.0674125874125875, "percentage": 35.61, "elapsed_time": "15:35:29", "remaining_time": "1 day, 4:11:31", "throughput": "17822.12", "total_tokens": 1000341504}
{"current_steps": 955, "total_steps": 2679, "loss": 0.823, "learning_rate": 3.632074857065928e-06, "epoch": 1.0685314685314686, "percentage": 35.65, "elapsed_time": "15:36:29", "remaining_time": "1 day, 4:10:34", "throughput": "17821.75", "total_tokens": 1001390080}
{"current_steps": 956, "total_steps": 2679, "loss": 0.757, "learning_rate": 3.6294355620206824e-06, "epoch": 1.0696503496503496, "percentage": 35.68, "elapsed_time": "15:37:35", "remaining_time": "1 day, 4:09:48", "throughput": "17819.47", "total_tokens": 1002438656}
{"current_steps": 957, "total_steps": 2679, "loss": 0.6583, "learning_rate": 3.6267946844186023e-06, "epoch": 1.0707692307692307, "percentage": 35.72, "elapsed_time": "15:38:29", "remaining_time": "1 day, 4:08:40", "throughput": "17821.06", "total_tokens": 1003487232}
{"current_steps": 958, "total_steps": 2679, "loss": 0.7008, "learning_rate": 3.6241522279600674e-06, "epoch": 1.0718881118881118, "percentage": 35.76, "elapsed_time": "15:39:29", "remaining_time": "1 day, 4:07:44", "throughput": "17820.69", "total_tokens": 1004535808}
{"current_steps": 959, "total_steps": 2679, "loss": 0.6306, "learning_rate": 3.621508196347667e-06, "epoch": 1.073006993006993, "percentage": 35.8, "elapsed_time": "15:40:22", "remaining_time": "1 day, 4:06:35", "throughput": "17822.37", "total_tokens": 1005584384}
{"current_steps": 960, "total_steps": 2679, "loss": 0.7172, "learning_rate": 3.618862593286199e-06, "epoch": 1.0741258741258741, "percentage": 35.83, "elapsed_time": "15:41:22", "remaining_time": "1 day, 4:05:38", "throughput": "17822.07", "total_tokens": 1006632960}
{"current_steps": 961, "total_steps": 2679, "loss": 0.7115, "learning_rate": 3.6162154224826627e-06, "epoch": 1.0752447552447553, "percentage": 35.87, "elapsed_time": "15:42:21", "remaining_time": "1 day, 4:04:40", "throughput": "17821.95", "total_tokens": 1007681536}
{"current_steps": 962, "total_steps": 2679, "loss": 0.6473, "learning_rate": 3.6135666876462565e-06, "epoch": 1.0763636363636364, "percentage": 35.91, "elapsed_time": "15:43:15", "remaining_time": "1 day, 4:03:32", "throughput": "17823.62", "total_tokens": 1008730112}
{"current_steps": 963, "total_steps": 2679, "loss": 0.6132, "learning_rate": 3.6109163924883668e-06, "epoch": 1.0774825174825176, "percentage": 35.95, "elapsed_time": "15:44:04", "remaining_time": "1 day, 4:02:16", "throughput": "17826.60", "total_tokens": 1009778688}
{"current_steps": 964, "total_steps": 2679, "loss": 0.6439, "learning_rate": 3.6082645407225673e-06, "epoch": 1.0786013986013987, "percentage": 35.98, "elapsed_time": "15:44:57", "remaining_time": "1 day, 4:01:08", "throughput": "17828.27", "total_tokens": 1010827264}
{"current_steps": 965, "total_steps": 2679, "loss": 0.6203, "learning_rate": 3.6056111360646134e-06, "epoch": 1.0797202797202796, "percentage": 36.02, "elapsed_time": "15:45:46", "remaining_time": "1 day, 3:59:50", "throughput": "17831.59", "total_tokens": 1011875840}
{"current_steps": 966, "total_steps": 2679, "loss": 0.7, "learning_rate": 3.602956182232438e-06, "epoch": 1.0808391608391608, "percentage": 36.06, "elapsed_time": "15:46:45", "remaining_time": "1 day, 3:58:52", "throughput": "17831.45", "total_tokens": 1012924416}
{"current_steps": 967, "total_steps": 2679, "loss": 0.6136, "learning_rate": 3.60029968294614e-06, "epoch": 1.081958041958042, "percentage": 36.1, "elapsed_time": "15:47:34", "remaining_time": "1 day, 3:57:35", "throughput": "17834.63", "total_tokens": 1013972992}
{"current_steps": 968, "total_steps": 2679, "loss": 0.8232, "learning_rate": 3.5976416419279892e-06, "epoch": 1.083076923076923, "percentage": 36.13, "elapsed_time": "15:48:34", "remaining_time": "1 day, 3:56:40", "throughput": "17834.07", "total_tokens": 1015021568}
{"current_steps": 969, "total_steps": 2679, "loss": 0.6797, "learning_rate": 3.594982062902412e-06, "epoch": 1.0841958041958042, "percentage": 36.17, "elapsed_time": "15:49:29", "remaining_time": "1 day, 3:55:34", "throughput": "17835.35", "total_tokens": 1016070144}
{"current_steps": 970, "total_steps": 2679, "loss": 0.5934, "learning_rate": 3.5923209495959923e-06, "epoch": 1.0853146853146853, "percentage": 36.21, "elapsed_time": "15:50:17", "remaining_time": "1 day, 3:54:15", "throughput": "17838.83", "total_tokens": 1017118720}
{"current_steps": 971, "total_steps": 2679, "loss": 0.663, "learning_rate": 3.5896583057374607e-06, "epoch": 1.0864335664335665, "percentage": 36.24, "elapsed_time": "15:51:11", "remaining_time": "1 day, 3:53:09", "throughput": "17840.26", "total_tokens": 1018167296}
{"current_steps": 972, "total_steps": 2679, "loss": 0.6741, "learning_rate": 3.5869941350576958e-06, "epoch": 1.0875524475524476, "percentage": 36.28, "elapsed_time": "15:52:08", "remaining_time": "1 day, 3:52:07", "throughput": "17840.68", "total_tokens": 1019215872}
{"current_steps": 973, "total_steps": 2679, "loss": 0.5814, "learning_rate": 3.5843284412897127e-06, "epoch": 1.0886713286713288, "percentage": 36.32, "elapsed_time": "15:52:56", "remaining_time": "1 day, 3:50:50", "throughput": "17844.08", "total_tokens": 1020264448}
{"current_steps": 974, "total_steps": 2679, "loss": 0.8484, "learning_rate": 3.5816612281686636e-06, "epoch": 1.08979020979021, "percentage": 36.36, "elapsed_time": "15:53:57", "remaining_time": "1 day, 3:49:54", "throughput": "17843.51", "total_tokens": 1021313024}
{"current_steps": 975, "total_steps": 2679, "loss": 0.7998, "learning_rate": 3.5789924994318267e-06, "epoch": 1.0909090909090908, "percentage": 36.39, "elapsed_time": "15:55:03", "remaining_time": "1 day, 3:49:08", "throughput": "17841.30", "total_tokens": 1022361600}
{"current_steps": 976, "total_steps": 2679, "loss": 0.6548, "learning_rate": 3.5763222588186053e-06, "epoch": 1.092027972027972, "percentage": 36.43, "elapsed_time": "15:56:03", "remaining_time": "1 day, 3:48:11", "throughput": "17840.93", "total_tokens": 1023410176}
{"current_steps": 977, "total_steps": 2679, "loss": 0.6207, "learning_rate": 3.5736505100705223e-06, "epoch": 1.0931468531468531, "percentage": 36.47, "elapsed_time": "15:56:52", "remaining_time": "1 day, 3:46:56", "throughput": "17843.85", "total_tokens": 1024458752}
{"current_steps": 978, "total_steps": 2679, "loss": 0.703, "learning_rate": 3.5709772569312097e-06, "epoch": 1.0942657342657343, "percentage": 36.51, "elapsed_time": "15:57:43", "remaining_time": "1 day, 3:45:43", "throughput": "17846.31", "total_tokens": 1025507328}
{"current_steps": 979, "total_steps": 2679, "loss": 0.7447, "learning_rate": 3.568302503146413e-06, "epoch": 1.0953846153846154, "percentage": 36.54, "elapsed_time": "15:58:40", "remaining_time": "1 day, 3:44:42", "throughput": "17846.72", "total_tokens": 1026555904}
{"current_steps": 980, "total_steps": 2679, "loss": 0.6949, "learning_rate": 3.565626252463977e-06, "epoch": 1.0965034965034965, "percentage": 36.58, "elapsed_time": "15:59:35", "remaining_time": "1 day, 3:43:37", "throughput": "17847.92", "total_tokens": 1027604480}
{"current_steps": 981, "total_steps": 2679, "loss": 0.8627, "learning_rate": 3.5629485086338432e-06, "epoch": 1.0976223776223777, "percentage": 36.62, "elapsed_time": "16:00:41", "remaining_time": "1 day, 3:42:51", "throughput": "17845.66", "total_tokens": 1028653056}
{"current_steps": 982, "total_steps": 2679, "loss": 0.6752, "learning_rate": 3.560269275408048e-06, "epoch": 1.0987412587412588, "percentage": 36.66, "elapsed_time": "16:01:36", "remaining_time": "1 day, 3:41:45", "throughput": "17846.92", "total_tokens": 1029701632}
{"current_steps": 983, "total_steps": 2679, "loss": 0.6785, "learning_rate": 3.557588556540712e-06, "epoch": 1.09986013986014, "percentage": 36.69, "elapsed_time": "16:02:30", "remaining_time": "1 day, 3:40:39", "throughput": "17848.23", "total_tokens": 1030750208}
{"current_steps": 984, "total_steps": 2679, "loss": 0.6627, "learning_rate": 3.554906355788041e-06, "epoch": 1.100979020979021, "percentage": 36.73, "elapsed_time": "16:03:25", "remaining_time": "1 day, 3:39:32", "throughput": "17849.63", "total_tokens": 1031798784}
{"current_steps": 985, "total_steps": 2679, "loss": 0.7507, "learning_rate": 3.552222676908313e-06, "epoch": 1.102097902097902, "percentage": 36.77, "elapsed_time": "16:04:30", "remaining_time": "1 day, 3:38:45", "throughput": "17847.60", "total_tokens": 1032847360}
{"current_steps": 986, "total_steps": 2679, "loss": 0.819, "learning_rate": 3.5495375236618795e-06, "epoch": 1.1032167832167832, "percentage": 36.8, "elapsed_time": "16:05:35", "remaining_time": "1 day, 3:37:58", "throughput": "17845.49", "total_tokens": 1033895936}
{"current_steps": 987, "total_steps": 2679, "loss": 0.7001, "learning_rate": 3.5468508998111596e-06, "epoch": 1.1043356643356643, "percentage": 36.84, "elapsed_time": "16:06:29", "remaining_time": "1 day, 3:36:51", "throughput": "17846.96", "total_tokens": 1034944512}
{"current_steps": 988, "total_steps": 2679, "loss": 0.6529, "learning_rate": 3.5441628091206276e-06, "epoch": 1.1054545454545455, "percentage": 36.88, "elapsed_time": "16:07:23", "remaining_time": "1 day, 3:35:44", "throughput": "17848.48", "total_tokens": 1035993088}
{"current_steps": 989, "total_steps": 2679, "loss": 0.5991, "learning_rate": 3.5414732553568194e-06, "epoch": 1.1065734265734266, "percentage": 36.92, "elapsed_time": "16:08:13", "remaining_time": "1 day, 3:34:30", "throughput": "17851.10", "total_tokens": 1037041664}
{"current_steps": 990, "total_steps": 2679, "loss": 0.6535, "learning_rate": 3.538782242288316e-06, "epoch": 1.1076923076923078, "percentage": 36.95, "elapsed_time": "16:09:09", "remaining_time": "1 day, 3:33:27", "throughput": "17851.99", "total_tokens": 1038090240}
{"current_steps": 991, "total_steps": 2679, "loss": 0.8031, "learning_rate": 3.5360897736857464e-06, "epoch": 1.108811188811189, "percentage": 36.99, "elapsed_time": "16:10:12", "remaining_time": "1 day, 3:32:34", "throughput": "17850.88", "total_tokens": 1039138816}
{"current_steps": 992, "total_steps": 2679, "loss": 0.7237, "learning_rate": 3.533395853321778e-06, "epoch": 1.10993006993007, "percentage": 37.03, "elapsed_time": "16:11:06", "remaining_time": "1 day, 3:31:28", "throughput": "17852.27", "total_tokens": 1040187392}
{"current_steps": 993, "total_steps": 2679, "loss": 0.5778, "learning_rate": 3.5307004849711114e-06, "epoch": 1.111048951048951, "percentage": 37.07, "elapsed_time": "16:11:54", "remaining_time": "1 day, 3:30:11", "throughput": "17855.55", "total_tokens": 1041235968}
{"current_steps": 994, "total_steps": 2679, "loss": 0.6749, "learning_rate": 3.528003672410477e-06, "epoch": 1.112167832167832, "percentage": 37.1, "elapsed_time": "16:12:48", "remaining_time": "1 day, 3:29:04", "throughput": "17856.94", "total_tokens": 1042284544}
{"current_steps": 995, "total_steps": 2679, "loss": 0.7667, "learning_rate": 3.5253054194186297e-06, "epoch": 1.1132867132867132, "percentage": 37.14, "elapsed_time": "16:13:53", "remaining_time": "1 day, 3:28:17", "throughput": "17854.95", "total_tokens": 1043333120}
{"current_steps": 996, "total_steps": 2679, "loss": 0.6976, "learning_rate": 3.5226057297763407e-06, "epoch": 1.1144055944055944, "percentage": 37.18, "elapsed_time": "16:14:53", "remaining_time": "1 day, 3:27:20", "throughput": "17854.63", "total_tokens": 1044381696}
{"current_steps": 997, "total_steps": 2679, "loss": 0.7189, "learning_rate": 3.5199046072663968e-06, "epoch": 1.1155244755244755, "percentage": 37.22, "elapsed_time": "16:15:58", "remaining_time": "1 day, 3:26:31", "throughput": "17852.72", "total_tokens": 1045430272}
{"current_steps": 998, "total_steps": 2679, "loss": 0.787, "learning_rate": 3.5172020556735897e-06, "epoch": 1.1166433566433567, "percentage": 37.25, "elapsed_time": "16:16:58", "remaining_time": "1 day, 3:25:34", "throughput": "17852.45", "total_tokens": 1046478848}
{"current_steps": 999, "total_steps": 2679, "loss": 0.6373, "learning_rate": 3.5144980787847155e-06, "epoch": 1.1177622377622378, "percentage": 37.29, "elapsed_time": "16:17:51", "remaining_time": "1 day, 3:24:27", "throughput": "17854.02", "total_tokens": 1047527424}
{"current_steps": 1000, "total_steps": 2679, "loss": 0.7507, "learning_rate": 3.511792680388567e-06, "epoch": 1.118881118881119, "percentage": 37.33, "elapsed_time": "16:18:52", "remaining_time": "1 day, 3:23:31", "throughput": "17853.45", "total_tokens": 1048576000}
{"current_steps": 1000, "total_steps": 2679, "eval_loss": 0.7350466847419739, "epoch": 1.118881118881119, "percentage": 37.33, "elapsed_time": "16:22:59", "remaining_time": "1 day, 3:30:26", "throughput": "17778.71", "total_tokens": 1048576000}
{"current_steps": 1001, "total_steps": 2679, "loss": 0.6723, "learning_rate": 3.5090858642759273e-06, "epoch": 1.12, "percentage": 37.36, "elapsed_time": "16:24:19", "remaining_time": "1 day, 3:30:02", "throughput": "17772.41", "total_tokens": 1049624576}
{"current_steps": 1002, "total_steps": 2679, "loss": 0.6434, "learning_rate": 3.5063776342395693e-06, "epoch": 1.121118881118881, "percentage": 37.4, "elapsed_time": "16:25:13", "remaining_time": "1 day, 3:28:55", "throughput": "17773.80", "total_tokens": 1050673152}
{"current_steps": 1003, "total_steps": 2679, "loss": 0.758, "learning_rate": 3.503667994074244e-06, "epoch": 1.1222377622377622, "percentage": 37.44, "elapsed_time": "16:26:20", "remaining_time": "1 day, 3:28:09", "throughput": "17771.57", "total_tokens": 1051721728}
{"current_steps": 1004, "total_steps": 2679, "loss": 0.7988, "learning_rate": 3.5009569475766793e-06, "epoch": 1.1233566433566433, "percentage": 37.48, "elapsed_time": "16:27:20", "remaining_time": "1 day, 3:27:12", "throughput": "17771.18", "total_tokens": 1052770304}
{"current_steps": 1005, "total_steps": 2679, "loss": 0.676, "learning_rate": 3.4982444985455744e-06, "epoch": 1.1244755244755245, "percentage": 37.51, "elapsed_time": "16:28:14", "remaining_time": "1 day, 3:26:05", "throughput": "17772.53", "total_tokens": 1053818880}
{"current_steps": 1006, "total_steps": 2679, "loss": 0.6572, "learning_rate": 3.495530650781591e-06, "epoch": 1.1255944055944056, "percentage": 37.55, "elapsed_time": "16:29:08", "remaining_time": "1 day, 3:24:57", "throughput": "17774.20", "total_tokens": 1054867456}
{"current_steps": 1007, "total_steps": 2679, "loss": 0.6558, "learning_rate": 3.4928154080873556e-06, "epoch": 1.1267132867132867, "percentage": 37.59, "elapsed_time": "16:30:07", "remaining_time": "1 day, 3:23:59", "throughput": "17774.04", "total_tokens": 1055916032}
{"current_steps": 1008, "total_steps": 2679, "loss": 0.8024, "learning_rate": 3.490098774267444e-06, "epoch": 1.1278321678321679, "percentage": 37.63, "elapsed_time": "16:31:13", "remaining_time": "1 day, 3:23:11", "throughput": "17772.09", "total_tokens": 1056964608}
{"current_steps": 1009, "total_steps": 2679, "loss": 0.7818, "learning_rate": 3.487380753128385e-06, "epoch": 1.128951048951049, "percentage": 37.66, "elapsed_time": "16:32:14", "remaining_time": "1 day, 3:22:15", "throughput": "17771.44", "total_tokens": 1058013184}
{"current_steps": 1010, "total_steps": 2679, "loss": 0.8533, "learning_rate": 3.4846613484786497e-06, "epoch": 1.1300699300699302, "percentage": 37.7, "elapsed_time": "16:33:20", "remaining_time": "1 day, 3:21:27", "throughput": "17769.46", "total_tokens": 1059061760}
{"current_steps": 1011, "total_steps": 2679, "loss": 0.6416, "learning_rate": 3.4819405641286476e-06, "epoch": 1.131188811188811, "percentage": 37.74, "elapsed_time": "16:34:14", "remaining_time": "1 day, 3:20:20", "throughput": "17770.96", "total_tokens": 1060110336}
{"current_steps": 1012, "total_steps": 2679, "loss": 0.6771, "learning_rate": 3.4792184038907212e-06, "epoch": 1.1323076923076922, "percentage": 37.78, "elapsed_time": "16:35:13", "remaining_time": "1 day, 3:19:21", "throughput": "17770.95", "total_tokens": 1061158912}
{"current_steps": 1013, "total_steps": 2679, "loss": 0.6969, "learning_rate": 3.4764948715791425e-06, "epoch": 1.1334265734265734, "percentage": 37.81, "elapsed_time": "16:36:12", "remaining_time": "1 day, 3:18:22", "throughput": "17770.93", "total_tokens": 1062207488}
{"current_steps": 1014, "total_steps": 2679, "loss": 0.8173, "learning_rate": 3.473769971010105e-06, "epoch": 1.1345454545454545, "percentage": 37.85, "elapsed_time": "16:37:12", "remaining_time": "1 day, 3:17:26", "throughput": "17770.42", "total_tokens": 1063256064}
{"current_steps": 1015, "total_steps": 2679, "loss": 0.6521, "learning_rate": 3.471043706001719e-06, "epoch": 1.1356643356643357, "percentage": 37.89, "elapsed_time": "16:38:06", "remaining_time": "1 day, 3:16:18", "throughput": "17772.04", "total_tokens": 1064304640}
{"current_steps": 1016, "total_steps": 2679, "loss": 0.6838, "learning_rate": 3.468316080374007e-06, "epoch": 1.1367832167832168, "percentage": 37.92, "elapsed_time": "16:39:05", "remaining_time": "1 day, 3:15:20", "throughput": "17771.89", "total_tokens": 1065353216}
{"current_steps": 1017, "total_steps": 2679, "loss": 0.7206, "learning_rate": 3.465587097948898e-06, "epoch": 1.137902097902098, "percentage": 37.96, "elapsed_time": "16:40:05", "remaining_time": "1 day, 3:14:21", "throughput": "17771.86", "total_tokens": 1066401792}
{"current_steps": 1018, "total_steps": 2679, "loss": 0.6745, "learning_rate": 3.462856762550223e-06, "epoch": 1.139020979020979, "percentage": 38.0, "elapsed_time": "16:40:59", "remaining_time": "1 day, 3:13:15", "throughput": "17773.14", "total_tokens": 1067450368}
{"current_steps": 1019, "total_steps": 2679, "loss": 0.8624, "learning_rate": 3.4601250780037064e-06, "epoch": 1.1401398601398602, "percentage": 38.04, "elapsed_time": "16:42:06", "remaining_time": "1 day, 3:12:29", "throughput": "17770.71", "total_tokens": 1068498944}
{"current_steps": 1020, "total_steps": 2679, "loss": 0.6621, "learning_rate": 3.4573920481369666e-06, "epoch": 1.1412587412587412, "percentage": 38.07, "elapsed_time": "16:43:06", "remaining_time": "1 day, 3:11:30", "throughput": "17770.69", "total_tokens": 1069547520}
{"current_steps": 1021, "total_steps": 2679, "loss": 0.7106, "learning_rate": 3.4546576767795036e-06, "epoch": 1.1423776223776223, "percentage": 38.11, "elapsed_time": "16:44:05", "remaining_time": "1 day, 3:10:33", "throughput": "17770.42", "total_tokens": 1070596096}
{"current_steps": 1022, "total_steps": 2679, "loss": 0.6669, "learning_rate": 3.4519219677626986e-06, "epoch": 1.1434965034965034, "percentage": 38.15, "elapsed_time": "16:45:00", "remaining_time": "1 day, 3:09:27", "throughput": "17771.75", "total_tokens": 1071644672}
{"current_steps": 1023, "total_steps": 2679, "loss": 0.7286, "learning_rate": 3.4491849249198074e-06, "epoch": 1.1446153846153846, "percentage": 38.19, "elapsed_time": "16:45:55", "remaining_time": "1 day, 3:08:21", "throughput": "17772.89", "total_tokens": 1072693248}
{"current_steps": 1024, "total_steps": 2679, "loss": 0.7738, "learning_rate": 3.446446552085954e-06, "epoch": 1.1457342657342657, "percentage": 38.22, "elapsed_time": "16:46:52", "remaining_time": "1 day, 3:07:20", "throughput": "17773.38", "total_tokens": 1073741824}
{"current_steps": 1025, "total_steps": 2679, "loss": 0.7004, "learning_rate": 3.4437068530981266e-06, "epoch": 1.1468531468531469, "percentage": 38.26, "elapsed_time": "16:47:52", "remaining_time": "1 day, 3:06:21", "throughput": "17773.22", "total_tokens": 1074790400}
{"current_steps": 1026, "total_steps": 2679, "loss": 0.7089, "learning_rate": 3.4409658317951717e-06, "epoch": 1.147972027972028, "percentage": 38.3, "elapsed_time": "16:48:57", "remaining_time": "1 day, 3:05:33", "throughput": "17771.33", "total_tokens": 1075838976}
{"current_steps": 1027, "total_steps": 2679, "loss": 0.6312, "learning_rate": 3.438223492017787e-06, "epoch": 1.1490909090909092, "percentage": 38.34, "elapsed_time": "16:49:51", "remaining_time": "1 day, 3:04:25", "throughput": "17772.93", "total_tokens": 1076887552}
{"current_steps": 1028, "total_steps": 2679, "loss": 0.7368, "learning_rate": 3.435479837608521e-06, "epoch": 1.1502097902097903, "percentage": 38.37, "elapsed_time": "16:50:51", "remaining_time": "1 day, 3:03:28", "throughput": "17772.62", "total_tokens": 1077936128}
{"current_steps": 1029, "total_steps": 2679, "loss": 0.7439, "learning_rate": 3.432734872411761e-06, "epoch": 1.1513286713286712, "percentage": 38.41, "elapsed_time": "16:51:57", "remaining_time": "1 day, 3:02:40", "throughput": "17770.53", "total_tokens": 1078984704}
{"current_steps": 1030, "total_steps": 2679, "loss": 0.8508, "learning_rate": 3.4299886002737313e-06, "epoch": 1.1524475524475524, "percentage": 38.45, "elapsed_time": "16:53:08", "remaining_time": "1 day, 3:02:00", "throughput": "17767.00", "total_tokens": 1080033280}
{"current_steps": 1031, "total_steps": 2679, "loss": 0.6596, "learning_rate": 3.4272410250424893e-06, "epoch": 1.1535664335664335, "percentage": 38.48, "elapsed_time": "16:54:02", "remaining_time": "1 day, 3:00:53", "throughput": "17768.48", "total_tokens": 1081081856}
{"current_steps": 1032, "total_steps": 2679, "loss": 0.7512, "learning_rate": 3.4244921505679175e-06, "epoch": 1.1546853146853147, "percentage": 38.52, "elapsed_time": "16:55:01", "remaining_time": "1 day, 2:59:54", "throughput": "17768.49", "total_tokens": 1082130432}
{"current_steps": 1033, "total_steps": 2679, "loss": 0.6346, "learning_rate": 3.4217419807017177e-06, "epoch": 1.1558041958041958, "percentage": 38.56, "elapsed_time": "16:55:50", "remaining_time": "1 day, 2:58:39", "throughput": "17771.49", "total_tokens": 1083179008}
{"current_steps": 1034, "total_steps": 2679, "loss": 0.8083, "learning_rate": 3.4189905192974087e-06, "epoch": 1.156923076923077, "percentage": 38.6, "elapsed_time": "16:56:55", "remaining_time": "1 day, 2:57:49", "throughput": "17769.77", "total_tokens": 1084227584}
{"current_steps": 1035, "total_steps": 2679, "loss": 0.6709, "learning_rate": 3.416237770210317e-06, "epoch": 1.158041958041958, "percentage": 38.63, "elapsed_time": "16:57:48", "remaining_time": "1 day, 2:56:42", "throughput": "17771.33", "total_tokens": 1085276160}
{"current_steps": 1036, "total_steps": 2679, "loss": 0.7477, "learning_rate": 3.413483737297576e-06, "epoch": 1.1591608391608392, "percentage": 38.67, "elapsed_time": "16:58:48", "remaining_time": "1 day, 2:55:44", "throughput": "17771.19", "total_tokens": 1086324736}
{"current_steps": 1037, "total_steps": 2679, "loss": 0.6706, "learning_rate": 3.4107284244181154e-06, "epoch": 1.1602797202797204, "percentage": 38.71, "elapsed_time": "16:59:46", "remaining_time": "1 day, 2:54:43", "throughput": "17771.39", "total_tokens": 1087373312}
{"current_steps": 1038, "total_steps": 2679, "loss": 0.8134, "learning_rate": 3.4079718354326583e-06, "epoch": 1.1613986013986013, "percentage": 38.75, "elapsed_time": "17:00:52", "remaining_time": "1 day, 2:53:55", "throughput": "17769.41", "total_tokens": 1088421888}
{"current_steps": 1039, "total_steps": 2679, "loss": 0.5733, "learning_rate": 3.40521397420372e-06, "epoch": 1.1625174825174824, "percentage": 38.78, "elapsed_time": "17:01:40", "remaining_time": "1 day, 2:52:39", "throughput": "17772.48", "total_tokens": 1089470464}
{"current_steps": 1040, "total_steps": 2679, "loss": 0.9083, "learning_rate": 3.402454844595593e-06, "epoch": 1.1636363636363636, "percentage": 38.82, "elapsed_time": "17:02:52", "remaining_time": "1 day, 2:52:00", "throughput": "17768.90", "total_tokens": 1090519040}
{"current_steps": 1041, "total_steps": 2679, "loss": 0.677, "learning_rate": 3.39969445047435e-06, "epoch": 1.1647552447552447, "percentage": 38.86, "elapsed_time": "17:03:51", "remaining_time": "1 day, 2:51:02", "throughput": "17768.75", "total_tokens": 1091567616}
{"current_steps": 1042, "total_steps": 2679, "loss": 0.628, "learning_rate": 3.396932795707836e-06, "epoch": 1.1658741258741259, "percentage": 38.9, "elapsed_time": "17:04:45", "remaining_time": "1 day, 2:49:55", "throughput": "17770.21", "total_tokens": 1092616192}
{"current_steps": 1043, "total_steps": 2679, "loss": 0.6666, "learning_rate": 3.3941698841656594e-06, "epoch": 1.166993006993007, "percentage": 38.93, "elapsed_time": "17:05:41", "remaining_time": "1 day, 2:48:50", "throughput": "17771.19", "total_tokens": 1093664768}
{"current_steps": 1044, "total_steps": 2679, "loss": 0.7819, "learning_rate": 3.3914057197191936e-06, "epoch": 1.1681118881118882, "percentage": 38.97, "elapsed_time": "17:06:47", "remaining_time": "1 day, 2:48:02", "throughput": "17769.25", "total_tokens": 1094713344}
{"current_steps": 1045, "total_steps": 2679, "loss": 0.6897, "learning_rate": 3.3886403062415653e-06, "epoch": 1.1692307692307693, "percentage": 39.01, "elapsed_time": "17:07:47", "remaining_time": "1 day, 2:47:05", "throughput": "17768.89", "total_tokens": 1095761920}
{"current_steps": 1046, "total_steps": 2679, "loss": 0.6811, "learning_rate": 3.3858736476076503e-06, "epoch": 1.1703496503496504, "percentage": 39.04, "elapsed_time": "17:08:46", "remaining_time": "1 day, 2:46:06", "throughput": "17768.92", "total_tokens": 1096810496}
{"current_steps": 1047, "total_steps": 2679, "loss": 0.6543, "learning_rate": 3.3831057476940716e-06, "epoch": 1.1714685314685314, "percentage": 39.08, "elapsed_time": "17:09:40", "remaining_time": "1 day, 2:44:59", "throughput": "17770.36", "total_tokens": 1097859072}
{"current_steps": 1048, "total_steps": 2679, "loss": 0.616, "learning_rate": 3.3803366103791892e-06, "epoch": 1.1725874125874125, "percentage": 39.12, "elapsed_time": "17:10:34", "remaining_time": "1 day, 2:43:52", "throughput": "17771.79", "total_tokens": 1098907648}
{"current_steps": 1049, "total_steps": 2679, "loss": 0.7484, "learning_rate": 3.3775662395431e-06, "epoch": 1.1737062937062936, "percentage": 39.16, "elapsed_time": "17:11:31", "remaining_time": "1 day, 2:42:50", "throughput": "17772.34", "total_tokens": 1099956224}
{"current_steps": 1050, "total_steps": 2679, "loss": 0.6972, "learning_rate": 3.3747946390676246e-06, "epoch": 1.1748251748251748, "percentage": 39.19, "elapsed_time": "17:12:25", "remaining_time": "1 day, 2:41:43", "throughput": "17773.81", "total_tokens": 1101004800}
{"current_steps": 1051, "total_steps": 2679, "loss": 0.8078, "learning_rate": 3.372021812836311e-06, "epoch": 1.175944055944056, "percentage": 39.23, "elapsed_time": "17:13:26", "remaining_time": "1 day, 2:40:48", "throughput": "17773.20", "total_tokens": 1102053376}
{"current_steps": 1052, "total_steps": 2679, "loss": 0.6498, "learning_rate": 3.369247764734424e-06, "epoch": 1.177062937062937, "percentage": 39.27, "elapsed_time": "17:14:20", "remaining_time": "1 day, 2:39:40", "throughput": "17774.75", "total_tokens": 1103101952}
{"current_steps": 1053, "total_steps": 2679, "loss": 0.7215, "learning_rate": 3.3664724986489368e-06, "epoch": 1.1781818181818182, "percentage": 39.31, "elapsed_time": "17:15:19", "remaining_time": "1 day, 2:38:43", "throughput": "17774.48", "total_tokens": 1104150528}
{"current_steps": 1054, "total_steps": 2679, "loss": 0.7566, "learning_rate": 3.363696018468534e-06, "epoch": 1.1793006993006994, "percentage": 39.34, "elapsed_time": "17:16:25", "remaining_time": "1 day, 2:37:54", "throughput": "17772.62", "total_tokens": 1105199104}
{"current_steps": 1055, "total_steps": 2679, "loss": 0.7554, "learning_rate": 3.360918328083598e-06, "epoch": 1.1804195804195805, "percentage": 39.38, "elapsed_time": "17:17:25", "remaining_time": "1 day, 2:36:56", "throughput": "17772.41", "total_tokens": 1106247680}
{"current_steps": 1056, "total_steps": 2679, "loss": 0.7479, "learning_rate": 3.3581394313862094e-06, "epoch": 1.1815384615384614, "percentage": 39.42, "elapsed_time": "17:18:24", "remaining_time": "1 day, 2:35:58", "throughput": "17772.21", "total_tokens": 1107296256}
{"current_steps": 1057, "total_steps": 2679, "loss": 0.7084, "learning_rate": 3.3553593322701374e-06, "epoch": 1.1826573426573426, "percentage": 39.46, "elapsed_time": "17:19:22", "remaining_time": "1 day, 2:34:56", "throughput": "17772.70", "total_tokens": 1108344832}
{"current_steps": 1058, "total_steps": 2679, "loss": 0.6758, "learning_rate": 3.3525780346308354e-06, "epoch": 1.1837762237762237, "percentage": 39.49, "elapsed_time": "17:20:16", "remaining_time": "1 day, 2:33:50", "throughput": "17774.09", "total_tokens": 1109393408}
{"current_steps": 1059, "total_steps": 2679, "loss": 0.7811, "learning_rate": 3.3497955423654395e-06, "epoch": 1.1848951048951049, "percentage": 39.53, "elapsed_time": "17:21:16", "remaining_time": "1 day, 2:32:53", "throughput": "17773.69", "total_tokens": 1110441984}
{"current_steps": 1060, "total_steps": 2679, "loss": 0.6947, "learning_rate": 3.3470118593727557e-06, "epoch": 1.186013986013986, "percentage": 39.57, "elapsed_time": "17:22:11", "remaining_time": "1 day, 2:31:48", "throughput": "17774.76", "total_tokens": 1111490560}
{"current_steps": 1061, "total_steps": 2679, "loss": 0.6182, "learning_rate": 3.3442269895532604e-06, "epoch": 1.1871328671328671, "percentage": 39.6, "elapsed_time": "17:23:01", "remaining_time": "1 day, 2:30:34", "throughput": "17777.57", "total_tokens": 1112539136}
{"current_steps": 1062, "total_steps": 2679, "loss": 0.6655, "learning_rate": 3.3414409368090932e-06, "epoch": 1.1882517482517483, "percentage": 39.64, "elapsed_time": "17:23:55", "remaining_time": "1 day, 2:29:28", "throughput": "17778.83", "total_tokens": 1113587712}
{"current_steps": 1063, "total_steps": 2679, "loss": 0.6562, "learning_rate": 3.338653705044051e-06, "epoch": 1.1893706293706294, "percentage": 39.68, "elapsed_time": "17:24:49", "remaining_time": "1 day, 2:28:22", "throughput": "17780.28", "total_tokens": 1114636288}
{"current_steps": 1064, "total_steps": 2679, "loss": 0.838, "learning_rate": 3.3358652981635826e-06, "epoch": 1.1904895104895106, "percentage": 39.72, "elapsed_time": "17:25:55", "remaining_time": "1 day, 2:27:33", "throughput": "17778.40", "total_tokens": 1115684864}
{"current_steps": 1065, "total_steps": 2679, "loss": 0.7933, "learning_rate": 3.3330757200747828e-06, "epoch": 1.1916083916083915, "percentage": 39.75, "elapsed_time": "17:27:00", "remaining_time": "1 day, 2:26:44", "throughput": "17776.48", "total_tokens": 1116733440}
{"current_steps": 1066, "total_steps": 2679, "loss": 0.5994, "learning_rate": 3.3302849746863873e-06, "epoch": 1.1927272727272726, "percentage": 39.79, "elapsed_time": "17:27:48", "remaining_time": "1 day, 2:25:29", "throughput": "17779.55", "total_tokens": 1117782016}
{"current_steps": 1067, "total_steps": 2679, "loss": 0.6728, "learning_rate": 3.3274930659087694e-06, "epoch": 1.1938461538461538, "percentage": 39.83, "elapsed_time": "17:28:49", "remaining_time": "1 day, 2:24:32", "throughput": "17779.16", "total_tokens": 1118830592}
{"current_steps": 1068, "total_steps": 2679, "loss": 0.7122, "learning_rate": 3.3246999976539315e-06, "epoch": 1.194965034965035, "percentage": 39.87, "elapsed_time": "17:29:44", "remaining_time": "1 day, 2:23:26", "throughput": "17780.37", "total_tokens": 1119879168}
{"current_steps": 1069, "total_steps": 2679, "loss": 0.6045, "learning_rate": 3.321905773835498e-06, "epoch": 1.196083916083916, "percentage": 39.9, "elapsed_time": "17:30:31", "remaining_time": "1 day, 2:22:10", "throughput": "17783.54", "total_tokens": 1120927744}
{"current_steps": 1070, "total_steps": 2679, "loss": 0.564, "learning_rate": 3.319110398368718e-06, "epoch": 1.1972027972027972, "percentage": 39.94, "elapsed_time": "17:31:19", "remaining_time": "1 day, 2:20:55", "throughput": "17786.67", "total_tokens": 1121976320}
{"current_steps": 1071, "total_steps": 2679, "loss": 0.7708, "learning_rate": 3.316313875170449e-06, "epoch": 1.1983216783216784, "percentage": 39.98, "elapsed_time": "17:32:24", "remaining_time": "1 day, 2:20:05", "throughput": "17784.97", "total_tokens": 1123024896}
{"current_steps": 1072, "total_steps": 2679, "loss": 0.7236, "learning_rate": 3.3135162081591592e-06, "epoch": 1.1994405594405595, "percentage": 40.01, "elapsed_time": "17:33:19", "remaining_time": "1 day, 2:18:59", "throughput": "17786.23", "total_tokens": 1124073472}
{"current_steps": 1073, "total_steps": 2679, "loss": 0.8704, "learning_rate": 3.310717401254919e-06, "epoch": 1.2005594405594406, "percentage": 40.05, "elapsed_time": "17:34:23", "remaining_time": "1 day, 2:18:08", "throughput": "17784.75", "total_tokens": 1125122048}
{"current_steps": 1074, "total_steps": 2679, "loss": 0.6604, "learning_rate": 3.307917458379397e-06, "epoch": 1.2016783216783216, "percentage": 40.09, "elapsed_time": "17:35:17", "remaining_time": "1 day, 2:17:02", "throughput": "17786.14", "total_tokens": 1126170624}
{"current_steps": 1075, "total_steps": 2679, "loss": 0.7461, "learning_rate": 3.305116383455852e-06, "epoch": 1.2027972027972027, "percentage": 40.13, "elapsed_time": "17:36:17", "remaining_time": "1 day, 2:16:04", "throughput": "17785.85", "total_tokens": 1127219200}
{"current_steps": 1076, "total_steps": 2679, "loss": 0.709, "learning_rate": 3.3023141804091295e-06, "epoch": 1.2039160839160838, "percentage": 40.16, "elapsed_time": "17:37:16", "remaining_time": "1 day, 2:15:06", "throughput": "17785.67", "total_tokens": 1128267776}
{"current_steps": 1077, "total_steps": 2679, "loss": 0.6079, "learning_rate": 3.2995108531656566e-06, "epoch": 1.205034965034965, "percentage": 40.2, "elapsed_time": "17:38:18", "remaining_time": "1 day, 2:14:11", "throughput": "17784.99", "total_tokens": 1129316352}
{"current_steps": 1078, "total_steps": 2679, "loss": 0.6995, "learning_rate": 3.2967064056534342e-06, "epoch": 1.2061538461538461, "percentage": 40.24, "elapsed_time": "17:39:15", "remaining_time": "1 day, 2:13:10", "throughput": "17785.43", "total_tokens": 1130364928}
{"current_steps": 1079, "total_steps": 2679, "loss": 0.8537, "learning_rate": 3.2939008418020334e-06, "epoch": 1.2072727272727273, "percentage": 40.28, "elapsed_time": "17:40:21", "remaining_time": "1 day, 2:12:21", "throughput": "17783.56", "total_tokens": 1131413504}
{"current_steps": 1080, "total_steps": 2679, "loss": 0.7067, "learning_rate": 3.2910941655425903e-06, "epoch": 1.2083916083916084, "percentage": 40.31, "elapsed_time": "17:41:20", "remaining_time": "1 day, 2:11:23", "throughput": "17783.39", "total_tokens": 1132462080}
{"current_steps": 1081, "total_steps": 2679, "loss": 0.706, "learning_rate": 3.2882863808077993e-06, "epoch": 1.2095104895104896, "percentage": 40.35, "elapsed_time": "17:42:19", "remaining_time": "1 day, 2:10:23", "throughput": "17783.48", "total_tokens": 1133510656}
{"current_steps": 1082, "total_steps": 2679, "loss": 0.7535, "learning_rate": 3.285477491531908e-06, "epoch": 1.2106293706293707, "percentage": 40.39, "elapsed_time": "17:43:13", "remaining_time": "1 day, 2:09:17", "throughput": "17784.76", "total_tokens": 1134559232}
{"current_steps": 1083, "total_steps": 2679, "loss": 0.6834, "learning_rate": 3.2826675016507094e-06, "epoch": 1.2117482517482516, "percentage": 40.43, "elapsed_time": "17:44:13", "remaining_time": "1 day, 2:08:19", "throughput": "17784.72", "total_tokens": 1135607808}
{"current_steps": 1084, "total_steps": 2679, "loss": 0.6554, "learning_rate": 3.279856415101543e-06, "epoch": 1.212867132867133, "percentage": 40.46, "elapsed_time": "17:45:06", "remaining_time": "1 day, 2:07:12", "throughput": "17786.14", "total_tokens": 1136656384}
{"current_steps": 1085, "total_steps": 2679, "loss": 0.7024, "learning_rate": 3.277044235823281e-06, "epoch": 1.213986013986014, "percentage": 40.5, "elapsed_time": "17:46:06", "remaining_time": "1 day, 2:06:14", "throughput": "17785.97", "total_tokens": 1137704960}
{"current_steps": 1086, "total_steps": 2679, "loss": 0.7157, "learning_rate": 3.2742309677563307e-06, "epoch": 1.215104895104895, "percentage": 40.54, "elapsed_time": "17:47:11", "remaining_time": "1 day, 2:05:25", "throughput": "17784.14", "total_tokens": 1138753536}
{"current_steps": 1087, "total_steps": 2679, "loss": 0.7182, "learning_rate": 3.2714166148426204e-06, "epoch": 1.2162237762237762, "percentage": 40.57, "elapsed_time": "17:48:12", "remaining_time": "1 day, 2:04:29", "throughput": "17783.58", "total_tokens": 1139802112}
{"current_steps": 1088, "total_steps": 2679, "loss": 0.6355, "learning_rate": 3.2686011810256023e-06, "epoch": 1.2173426573426573, "percentage": 40.61, "elapsed_time": "17:49:01", "remaining_time": "1 day, 2:03:14", "throughput": "17786.57", "total_tokens": 1140850688}
{"current_steps": 1089, "total_steps": 2679, "loss": 0.6785, "learning_rate": 3.2657846702502404e-06, "epoch": 1.2184615384615385, "percentage": 40.65, "elapsed_time": "17:49:54", "remaining_time": "1 day, 2:02:07", "throughput": "17788.14", "total_tokens": 1141899264}
{"current_steps": 1090, "total_steps": 2679, "loss": 0.7554, "learning_rate": 3.2629670864630104e-06, "epoch": 1.2195804195804196, "percentage": 40.69, "elapsed_time": "17:50:54", "remaining_time": "1 day, 2:01:10", "throughput": "17787.71", "total_tokens": 1142947840}
{"current_steps": 1091, "total_steps": 2679, "loss": 0.6407, "learning_rate": 3.2601484336118887e-06, "epoch": 1.2206993006993008, "percentage": 40.72, "elapsed_time": "17:51:48", "remaining_time": "1 day, 2:00:04", "throughput": "17789.09", "total_tokens": 1143996416}
{"current_steps": 1092, "total_steps": 2679, "loss": 0.6531, "learning_rate": 3.257328715646351e-06, "epoch": 1.221818181818182, "percentage": 40.76, "elapsed_time": "17:52:43", "remaining_time": "1 day, 1:58:58", "throughput": "17790.34", "total_tokens": 1145044992}
{"current_steps": 1093, "total_steps": 2679, "loss": 0.8666, "learning_rate": 3.2545079365173672e-06, "epoch": 1.222937062937063, "percentage": 40.8, "elapsed_time": "17:53:54", "remaining_time": "1 day, 1:58:17", "throughput": "17787.00", "total_tokens": 1146093568}
{"current_steps": 1094, "total_steps": 2679, "loss": 0.6811, "learning_rate": 3.2516861001773904e-06, "epoch": 1.224055944055944, "percentage": 40.84, "elapsed_time": "17:54:48", "remaining_time": "1 day, 1:57:12", "throughput": "17788.25", "total_tokens": 1147142144}
{"current_steps": 1095, "total_steps": 2679, "loss": 0.7283, "learning_rate": 3.248863210580358e-06, "epoch": 1.2251748251748251, "percentage": 40.87, "elapsed_time": "17:55:51", "remaining_time": "1 day, 1:56:19", "throughput": "17787.09", "total_tokens": 1148190720}
{"current_steps": 1096, "total_steps": 2679, "loss": 0.6472, "learning_rate": 3.2460392716816826e-06, "epoch": 1.2262937062937063, "percentage": 40.91, "elapsed_time": "17:56:40", "remaining_time": "1 day, 1:55:05", "throughput": "17789.87", "total_tokens": 1149239296}
{"current_steps": 1097, "total_steps": 2679, "loss": 0.6829, "learning_rate": 3.2432142874382442e-06, "epoch": 1.2274125874125874, "percentage": 40.95, "elapsed_time": "17:57:40", "remaining_time": "1 day, 1:54:07", "throughput": "17789.74", "total_tokens": 1150287872}
{"current_steps": 1098, "total_steps": 2679, "loss": 0.5833, "learning_rate": 3.240388261808394e-06, "epoch": 1.2285314685314686, "percentage": 40.99, "elapsed_time": "17:58:29", "remaining_time": "1 day, 1:52:54", "throughput": "17792.37", "total_tokens": 1151336448}
{"current_steps": 1099, "total_steps": 2679, "loss": 0.6781, "learning_rate": 3.237561198751935e-06, "epoch": 1.2296503496503497, "percentage": 41.02, "elapsed_time": "17:59:29", "remaining_time": "1 day, 1:51:56", "throughput": "17792.18", "total_tokens": 1152385024}
{"current_steps": 1100, "total_steps": 2679, "loss": 0.6834, "learning_rate": 3.2347331022301293e-06, "epoch": 1.2307692307692308, "percentage": 41.06, "elapsed_time": "18:00:22", "remaining_time": "1 day, 1:50:50", "throughput": "17793.59", "total_tokens": 1153433600}
{"current_steps": 1101, "total_steps": 2679, "loss": 0.6747, "learning_rate": 3.231903976205684e-06, "epoch": 1.231888111888112, "percentage": 41.1, "elapsed_time": "18:01:17", "remaining_time": "1 day, 1:49:44", "throughput": "17794.88", "total_tokens": 1154482176}
{"current_steps": 1102, "total_steps": 2679, "loss": 0.6939, "learning_rate": 3.2290738246427494e-06, "epoch": 1.2330069930069931, "percentage": 41.13, "elapsed_time": "18:02:11", "remaining_time": "1 day, 1:48:39", "throughput": "17796.17", "total_tokens": 1155530752}
{"current_steps": 1103, "total_steps": 2679, "loss": 0.6245, "learning_rate": 3.2262426515069144e-06, "epoch": 1.234125874125874, "percentage": 41.17, "elapsed_time": "18:03:02", "remaining_time": "1 day, 1:47:28", "throughput": "17798.39", "total_tokens": 1156579328}
{"current_steps": 1104, "total_steps": 2679, "loss": 0.7372, "learning_rate": 3.223410460765198e-06, "epoch": 1.2352447552447552, "percentage": 41.21, "elapsed_time": "18:04:01", "remaining_time": "1 day, 1:46:30", "throughput": "17798.21", "total_tokens": 1157627904}
{"current_steps": 1105, "total_steps": 2679, "loss": 0.66, "learning_rate": 3.220577256386043e-06, "epoch": 1.2363636363636363, "percentage": 41.25, "elapsed_time": "18:04:52", "remaining_time": "1 day, 1:45:20", "throughput": "17800.34", "total_tokens": 1158676480}
{"current_steps": 1106, "total_steps": 2679, "loss": 0.7805, "learning_rate": 3.217743042339318e-06, "epoch": 1.2374825174825175, "percentage": 41.28, "elapsed_time": "18:05:58", "remaining_time": "1 day, 1:44:31", "throughput": "17798.54", "total_tokens": 1159725056}
{"current_steps": 1107, "total_steps": 2679, "loss": 0.7018, "learning_rate": 3.2149078225963e-06, "epoch": 1.2386013986013986, "percentage": 41.32, "elapsed_time": "18:06:52", "remaining_time": "1 day, 1:43:25", "throughput": "17799.77", "total_tokens": 1160773632}
{"current_steps": 1108, "total_steps": 2679, "loss": 0.6365, "learning_rate": 3.2120716011296794e-06, "epoch": 1.2397202797202798, "percentage": 41.36, "elapsed_time": "18:07:48", "remaining_time": "1 day, 1:42:22", "throughput": "17800.53", "total_tokens": 1161822208}
{"current_steps": 1109, "total_steps": 2679, "loss": 0.7791, "learning_rate": 3.2092343819135485e-06, "epoch": 1.240839160839161, "percentage": 41.4, "elapsed_time": "18:08:48", "remaining_time": "1 day, 1:41:25", "throughput": "17800.24", "total_tokens": 1162870784}
{"current_steps": 1110, "total_steps": 2679, "loss": 0.7503, "learning_rate": 3.206396168923398e-06, "epoch": 1.241958041958042, "percentage": 41.43, "elapsed_time": "18:09:45", "remaining_time": "1 day, 1:40:23", "throughput": "17800.74", "total_tokens": 1163919360}
{"current_steps": 1111, "total_steps": 2679, "loss": 0.6576, "learning_rate": 3.203556966136113e-06, "epoch": 1.2430769230769232, "percentage": 41.47, "elapsed_time": "18:10:39", "remaining_time": "1 day, 1:39:17", "throughput": "17802.14", "total_tokens": 1164967936}
{"current_steps": 1112, "total_steps": 2679, "loss": 0.7734, "learning_rate": 3.2007167775299613e-06, "epoch": 1.2441958041958041, "percentage": 41.51, "elapsed_time": "18:11:34", "remaining_time": "1 day, 1:38:12", "throughput": "17803.30", "total_tokens": 1166016512}
{"current_steps": 1113, "total_steps": 2679, "loss": 0.7638, "learning_rate": 3.197875607084595e-06, "epoch": 1.2453146853146853, "percentage": 41.55, "elapsed_time": "18:12:34", "remaining_time": "1 day, 1:37:15", "throughput": "17803.03", "total_tokens": 1167065088}
{"current_steps": 1114, "total_steps": 2679, "loss": 0.6899, "learning_rate": 3.195033458781042e-06, "epoch": 1.2464335664335664, "percentage": 41.58, "elapsed_time": "18:13:28", "remaining_time": "1 day, 1:36:10", "throughput": "17804.22", "total_tokens": 1168113664}
{"current_steps": 1115, "total_steps": 2679, "loss": 0.7669, "learning_rate": 3.192190336601698e-06, "epoch": 1.2475524475524475, "percentage": 41.62, "elapsed_time": "18:14:34", "remaining_time": "1 day, 1:35:21", "throughput": "17802.37", "total_tokens": 1169162240}
{"current_steps": 1116, "total_steps": 2679, "loss": 0.6825, "learning_rate": 3.189346244530327e-06, "epoch": 1.2486713286713287, "percentage": 41.66, "elapsed_time": "18:15:29", "remaining_time": "1 day, 1:34:16", "throughput": "17803.49", "total_tokens": 1170210816}
{"current_steps": 1117, "total_steps": 2679, "loss": 0.6848, "learning_rate": 3.18650118655205e-06, "epoch": 1.2497902097902098, "percentage": 41.69, "elapsed_time": "18:16:28", "remaining_time": "1 day, 1:33:18", "throughput": "17803.30", "total_tokens": 1171259392}
{"current_steps": 1118, "total_steps": 2679, "loss": 0.5578, "learning_rate": 3.183655166653339e-06, "epoch": 1.250909090909091, "percentage": 41.73, "elapsed_time": "18:17:16", "remaining_time": "1 day, 1:32:03", "throughput": "17806.34", "total_tokens": 1172307968}
{"current_steps": 1119, "total_steps": 2679, "loss": 0.6381, "learning_rate": 3.180808188822019e-06, "epoch": 1.252027972027972, "percentage": 41.77, "elapsed_time": "18:18:13", "remaining_time": "1 day, 1:31:02", "throughput": "17806.91", "total_tokens": 1173356544}
{"current_steps": 1120, "total_steps": 2679, "loss": 0.631, "learning_rate": 3.177960257047252e-06, "epoch": 1.2531468531468533, "percentage": 41.81, "elapsed_time": "18:19:07", "remaining_time": "1 day, 1:29:55", "throughput": "17808.29", "total_tokens": 1174405120}
{"current_steps": 1121, "total_steps": 2679, "loss": 0.7872, "learning_rate": 3.175111375319541e-06, "epoch": 1.2542657342657342, "percentage": 41.84, "elapsed_time": "18:20:15", "remaining_time": "1 day, 1:29:09", "throughput": "17805.83", "total_tokens": 1175453696}
{"current_steps": 1122, "total_steps": 2679, "loss": 0.7283, "learning_rate": 3.1722615476307173e-06, "epoch": 1.2553846153846153, "percentage": 41.88, "elapsed_time": "18:21:14", "remaining_time": "1 day, 1:28:11", "throughput": "17805.62", "total_tokens": 1176502272}
{"current_steps": 1123, "total_steps": 2679, "loss": 0.767, "learning_rate": 3.1694107779739394e-06, "epoch": 1.2565034965034965, "percentage": 41.92, "elapsed_time": "18:22:14", "remaining_time": "1 day, 1:27:14", "throughput": "17805.34", "total_tokens": 1177550848}
{"current_steps": 1124, "total_steps": 2679, "loss": 0.7658, "learning_rate": 3.1665590703436843e-06, "epoch": 1.2576223776223776, "percentage": 41.96, "elapsed_time": "18:23:19", "remaining_time": "1 day, 1:26:24", "throughput": "17803.63", "total_tokens": 1178599424}
{"current_steps": 1125, "total_steps": 2679, "loss": 0.6358, "learning_rate": 3.1637064287357433e-06, "epoch": 1.2587412587412588, "percentage": 41.99, "elapsed_time": "18:24:14", "remaining_time": "1 day, 1:25:19", "throughput": "17804.74", "total_tokens": 1179648000}
{"current_steps": 1126, "total_steps": 2679, "loss": 0.7587, "learning_rate": 3.1608528571472174e-06, "epoch": 1.25986013986014, "percentage": 42.03, "elapsed_time": "18:25:18", "remaining_time": "1 day, 1:24:27", "throughput": "17803.45", "total_tokens": 1180696576}
{"current_steps": 1127, "total_steps": 2679, "loss": 0.6578, "learning_rate": 3.1579983595765107e-06, "epoch": 1.260979020979021, "percentage": 42.07, "elapsed_time": "18:26:12", "remaining_time": "1 day, 1:23:21", "throughput": "17804.87", "total_tokens": 1181745152}
{"current_steps": 1128, "total_steps": 2679, "loss": 0.6376, "learning_rate": 3.1551429400233235e-06, "epoch": 1.2620979020979022, "percentage": 42.11, "elapsed_time": "18:27:05", "remaining_time": "1 day, 1:22:15", "throughput": "17806.33", "total_tokens": 1182793728}
{"current_steps": 1129, "total_steps": 2679, "loss": 0.8811, "learning_rate": 3.1522866024886497e-06, "epoch": 1.2632167832167833, "percentage": 42.14, "elapsed_time": "18:28:11", "remaining_time": "1 day, 1:21:25", "throughput": "17804.51", "total_tokens": 1183842304}
{"current_steps": 1130, "total_steps": 2679, "loss": 0.7592, "learning_rate": 3.149429350974767e-06, "epoch": 1.2643356643356642, "percentage": 42.18, "elapsed_time": "18:29:16", "remaining_time": "1 day, 1:20:35", "throughput": "17802.75", "total_tokens": 1184890880}
{"current_steps": 1131, "total_steps": 2679, "loss": 0.6601, "learning_rate": 3.1465711894852364e-06, "epoch": 1.2654545454545454, "percentage": 42.22, "elapsed_time": "18:30:16", "remaining_time": "1 day, 1:19:38", "throughput": "17802.48", "total_tokens": 1185939456}
{"current_steps": 1132, "total_steps": 2679, "loss": 0.686, "learning_rate": 3.143712122024893e-06, "epoch": 1.2665734265734265, "percentage": 42.25, "elapsed_time": "18:31:14", "remaining_time": "1 day, 1:18:37", "throughput": "17802.83", "total_tokens": 1186988032}
{"current_steps": 1133, "total_steps": 2679, "loss": 0.7381, "learning_rate": 3.1408521525998403e-06, "epoch": 1.2676923076923077, "percentage": 42.29, "elapsed_time": "18:32:14", "remaining_time": "1 day, 1:17:40", "throughput": "17802.49", "total_tokens": 1188036608}
{"current_steps": 1134, "total_steps": 2679, "loss": 0.6579, "learning_rate": 3.1379912852174477e-06, "epoch": 1.2688111888111888, "percentage": 42.33, "elapsed_time": "18:33:07", "remaining_time": "1 day, 1:16:33", "throughput": "17804.03", "total_tokens": 1189085184}
{"current_steps": 1135, "total_steps": 2679, "loss": 0.7443, "learning_rate": 3.135129523886341e-06, "epoch": 1.26993006993007, "percentage": 42.37, "elapsed_time": "18:34:12", "remaining_time": "1 day, 1:15:42", "throughput": "17802.40", "total_tokens": 1190133760}
{"current_steps": 1136, "total_steps": 2679, "loss": 0.697, "learning_rate": 3.1322668726163983e-06, "epoch": 1.271048951048951, "percentage": 42.4, "elapsed_time": "18:35:06", "remaining_time": "1 day, 1:14:37", "throughput": "17803.71", "total_tokens": 1191182336}
{"current_steps": 1137, "total_steps": 2679, "loss": 0.8191, "learning_rate": 3.129403335418747e-06, "epoch": 1.2721678321678322, "percentage": 42.44, "elapsed_time": "18:36:10", "remaining_time": "1 day, 1:13:46", "throughput": "17802.20", "total_tokens": 1192230912}
{"current_steps": 1138, "total_steps": 2679, "loss": 0.8157, "learning_rate": 3.1265389163057537e-06, "epoch": 1.2732867132867134, "percentage": 42.48, "elapsed_time": "18:37:11", "remaining_time": "1 day, 1:12:48", "throughput": "17801.90", "total_tokens": 1193279488}
{"current_steps": 1139, "total_steps": 2679, "loss": 0.7923, "learning_rate": 3.123673619291021e-06, "epoch": 1.2744055944055943, "percentage": 42.52, "elapsed_time": "18:38:16", "remaining_time": "1 day, 1:11:58", "throughput": "17800.14", "total_tokens": 1194328064}
{"current_steps": 1140, "total_steps": 2679, "loss": 0.6936, "learning_rate": 3.1208074483893833e-06, "epoch": 1.2755244755244755, "percentage": 42.55, "elapsed_time": "18:39:16", "remaining_time": "1 day, 1:11:00", "throughput": "17799.95", "total_tokens": 1195376640}
{"current_steps": 1141, "total_steps": 2679, "loss": 0.6132, "learning_rate": 3.1179404076168983e-06, "epoch": 1.2766433566433566, "percentage": 42.59, "elapsed_time": "18:40:10", "remaining_time": "1 day, 1:09:55", "throughput": "17801.26", "total_tokens": 1196425216}
{"current_steps": 1142, "total_steps": 2679, "loss": 0.7471, "learning_rate": 3.115072500990841e-06, "epoch": 1.2777622377622377, "percentage": 42.63, "elapsed_time": "18:41:15", "remaining_time": "1 day, 1:09:05", "throughput": "17799.45", "total_tokens": 1197473792}
{"current_steps": 1143, "total_steps": 2679, "loss": 0.6598, "learning_rate": 3.1122037325297027e-06, "epoch": 1.2788811188811189, "percentage": 42.67, "elapsed_time": "18:42:09", "remaining_time": "1 day, 1:08:00", "throughput": "17800.72", "total_tokens": 1198522368}
{"current_steps": 1144, "total_steps": 2679, "loss": 0.7787, "learning_rate": 3.1093341062531797e-06, "epoch": 1.28, "percentage": 42.7, "elapsed_time": "18:43:14", "remaining_time": "1 day, 1:07:09", "throughput": "17799.12", "total_tokens": 1199570944}
{"current_steps": 1145, "total_steps": 2679, "loss": 0.6321, "learning_rate": 3.1064636261821716e-06, "epoch": 1.2811188811188812, "percentage": 42.74, "elapsed_time": "18:44:07", "remaining_time": "1 day, 1:06:01", "throughput": "17800.87", "total_tokens": 1200619520}
{"current_steps": 1146, "total_steps": 2679, "loss": 0.8964, "learning_rate": 3.103592296338775e-06, "epoch": 1.2822377622377623, "percentage": 42.78, "elapsed_time": "18:45:19", "remaining_time": "1 day, 1:05:20", "throughput": "17797.39", "total_tokens": 1201668096}
{"current_steps": 1147, "total_steps": 2679, "loss": 0.5977, "learning_rate": 3.1007201207462745e-06, "epoch": 1.2833566433566435, "percentage": 42.81, "elapsed_time": "18:46:07", "remaining_time": "1 day, 1:04:06", "throughput": "17800.30", "total_tokens": 1202716672}
{"current_steps": 1148, "total_steps": 2679, "loss": 0.771, "learning_rate": 3.097847103429143e-06, "epoch": 1.2844755244755244, "percentage": 42.85, "elapsed_time": "18:47:07", "remaining_time": "1 day, 1:03:09", "throughput": "17799.98", "total_tokens": 1203765248}
{"current_steps": 1149, "total_steps": 2679, "loss": 0.8276, "learning_rate": 3.09497324841303e-06, "epoch": 1.2855944055944055, "percentage": 42.89, "elapsed_time": "18:48:18", "remaining_time": "1 day, 1:02:27", "throughput": "17796.72", "total_tokens": 1204813824}
{"current_steps": 1150, "total_steps": 2679, "loss": 0.8319, "learning_rate": 3.092098559724761e-06, "epoch": 1.2867132867132867, "percentage": 42.93, "elapsed_time": "18:49:30", "remaining_time": "1 day, 1:01:44", "throughput": "17793.39", "total_tokens": 1205862400}
{"current_steps": 1151, "total_steps": 2679, "loss": 0.7226, "learning_rate": 3.089223041392329e-06, "epoch": 1.2878321678321678, "percentage": 42.96, "elapsed_time": "18:50:30", "remaining_time": "1 day, 1:00:47", "throughput": "17793.09", "total_tokens": 1206910976}
{"current_steps": 1152, "total_steps": 2679, "loss": 0.7661, "learning_rate": 3.086346697444888e-06, "epoch": 1.288951048951049, "percentage": 43.0, "elapsed_time": "18:51:29", "remaining_time": "1 day, 0:59:49", "throughput": "17792.89", "total_tokens": 1207959552}
{"current_steps": 1153, "total_steps": 2679, "loss": 0.665, "learning_rate": 3.0834695319127516e-06, "epoch": 1.29006993006993, "percentage": 43.04, "elapsed_time": "18:52:23", "remaining_time": "1 day, 0:58:43", "throughput": "17794.31", "total_tokens": 1209008128}
{"current_steps": 1154, "total_steps": 2679, "loss": 0.8001, "learning_rate": 3.080591548827382e-06, "epoch": 1.2911888111888112, "percentage": 43.08, "elapsed_time": "18:53:29", "remaining_time": "1 day, 0:57:53", "throughput": "17792.54", "total_tokens": 1210056704}
{"current_steps": 1155, "total_steps": 2679, "loss": 0.823, "learning_rate": 3.077712752221388e-06, "epoch": 1.2923076923076924, "percentage": 43.11, "elapsed_time": "18:54:29", "remaining_time": "1 day, 0:56:56", "throughput": "17792.18", "total_tokens": 1211105280}
{"current_steps": 1156, "total_steps": 2679, "loss": 0.8342, "learning_rate": 3.074833146128519e-06, "epoch": 1.2934265734265735, "percentage": 43.15, "elapsed_time": "18:55:34", "remaining_time": "1 day, 0:56:05", "throughput": "17790.58", "total_tokens": 1212153856}
{"current_steps": 1157, "total_steps": 2679, "loss": 0.6794, "learning_rate": 3.0719527345836568e-06, "epoch": 1.2945454545454544, "percentage": 43.19, "elapsed_time": "18:56:32", "remaining_time": "1 day, 0:55:05", "throughput": "17790.79", "total_tokens": 1213202432}
{"current_steps": 1158, "total_steps": 2679, "loss": 0.7232, "learning_rate": 3.0690715216228143e-06, "epoch": 1.2956643356643356, "percentage": 43.23, "elapsed_time": "18:57:32", "remaining_time": "1 day, 0:54:07", "throughput": "17790.59", "total_tokens": 1214251008}
{"current_steps": 1159, "total_steps": 2679, "loss": 0.6864, "learning_rate": 3.066189511283126e-06, "epoch": 1.2967832167832167, "percentage": 43.26, "elapsed_time": "18:58:25", "remaining_time": "1 day, 0:53:01", "throughput": "17792.04", "total_tokens": 1215299584}
{"current_steps": 1160, "total_steps": 2679, "loss": 0.6201, "learning_rate": 3.063306707602842e-06, "epoch": 1.2979020979020979, "percentage": 43.3, "elapsed_time": "18:59:17", "remaining_time": "1 day, 0:51:53", "throughput": "17793.87", "total_tokens": 1216348160}
{"current_steps": 1161, "total_steps": 2679, "loss": 0.7225, "learning_rate": 3.0604231146213276e-06, "epoch": 1.299020979020979, "percentage": 43.34, "elapsed_time": "19:00:17", "remaining_time": "1 day, 0:50:55", "throughput": "17793.64", "total_tokens": 1217396736}
{"current_steps": 1162, "total_steps": 2679, "loss": 0.6241, "learning_rate": 3.0575387363790505e-06, "epoch": 1.3001398601398602, "percentage": 43.37, "elapsed_time": "19:01:05", "remaining_time": "1 day, 0:49:42", "throughput": "17796.51", "total_tokens": 1218445312}
{"current_steps": 1163, "total_steps": 2679, "loss": 0.611, "learning_rate": 3.054653576917581e-06, "epoch": 1.3012587412587413, "percentage": 43.41, "elapsed_time": "19:01:53", "remaining_time": "1 day, 0:48:29", "throughput": "17799.26", "total_tokens": 1219493888}
{"current_steps": 1164, "total_steps": 2679, "loss": 0.7342, "learning_rate": 3.051767640279585e-06, "epoch": 1.3023776223776224, "percentage": 43.45, "elapsed_time": "19:02:53", "remaining_time": "1 day, 0:47:32", "throughput": "17798.94", "total_tokens": 1220542464}
{"current_steps": 1165, "total_steps": 2679, "loss": 0.5949, "learning_rate": 3.048880930508813e-06, "epoch": 1.3034965034965036, "percentage": 43.49, "elapsed_time": "19:03:42", "remaining_time": "1 day, 0:46:19", "throughput": "17801.62", "total_tokens": 1221591040}
{"current_steps": 1166, "total_steps": 2679, "loss": 0.6798, "learning_rate": 3.0459934516501035e-06, "epoch": 1.3046153846153845, "percentage": 43.52, "elapsed_time": "19:04:36", "remaining_time": "1 day, 0:45:13", "throughput": "17803.00", "total_tokens": 1222639616}
{"current_steps": 1167, "total_steps": 2679, "loss": 0.6861, "learning_rate": 3.0431052077493693e-06, "epoch": 1.3057342657342657, "percentage": 43.56, "elapsed_time": "19:05:33", "remaining_time": "1 day, 0:44:12", "throughput": "17803.43", "total_tokens": 1223688192}
{"current_steps": 1168, "total_steps": 2679, "loss": 0.7506, "learning_rate": 3.0402162028535985e-06, "epoch": 1.3068531468531468, "percentage": 43.6, "elapsed_time": "19:06:35", "remaining_time": "1 day, 0:43:18", "throughput": "17802.54", "total_tokens": 1224736768}
{"current_steps": 1169, "total_steps": 2679, "loss": 0.7873, "learning_rate": 3.0373264410108422e-06, "epoch": 1.307972027972028, "percentage": 43.64, "elapsed_time": "19:07:41", "remaining_time": "1 day, 0:42:28", "throughput": "17800.78", "total_tokens": 1225785344}
{"current_steps": 1170, "total_steps": 2679, "loss": 0.8313, "learning_rate": 3.0344359262702135e-06, "epoch": 1.309090909090909, "percentage": 43.67, "elapsed_time": "19:08:40", "remaining_time": "1 day, 0:41:30", "throughput": "17800.60", "total_tokens": 1226833920}
{"current_steps": 1171, "total_steps": 2679, "loss": 0.7804, "learning_rate": 3.0315446626818816e-06, "epoch": 1.3102097902097902, "percentage": 43.71, "elapsed_time": "19:09:41", "remaining_time": "1 day, 0:40:32", "throughput": "17800.29", "total_tokens": 1227882496}
{"current_steps": 1172, "total_steps": 2679, "loss": 0.6189, "learning_rate": 3.0286526542970624e-06, "epoch": 1.3113286713286714, "percentage": 43.75, "elapsed_time": "19:10:29", "remaining_time": "1 day, 0:39:20", "throughput": "17802.94", "total_tokens": 1228931072}
{"current_steps": 1173, "total_steps": 2679, "loss": 0.6606, "learning_rate": 3.0257599051680175e-06, "epoch": 1.3124475524475525, "percentage": 43.78, "elapsed_time": "19:11:23", "remaining_time": "1 day, 0:38:15", "throughput": "17804.15", "total_tokens": 1229979648}
{"current_steps": 1174, "total_steps": 2679, "loss": 0.7247, "learning_rate": 3.022866419348046e-06, "epoch": 1.3135664335664337, "percentage": 43.82, "elapsed_time": "19:12:18", "remaining_time": "1 day, 0:37:11", "throughput": "17805.28", "total_tokens": 1231028224}
{"current_steps": 1175, "total_steps": 2679, "loss": 0.6983, "learning_rate": 3.0199722008914787e-06, "epoch": 1.3146853146853146, "percentage": 43.86, "elapsed_time": "19:13:19", "remaining_time": "1 day, 0:36:15", "throughput": "17804.77", "total_tokens": 1232076800}
{"current_steps": 1176, "total_steps": 2679, "loss": 0.6892, "learning_rate": 3.0170772538536735e-06, "epoch": 1.315804195804196, "percentage": 43.9, "elapsed_time": "19:14:19", "remaining_time": "1 day, 0:35:17", "throughput": "17804.51", "total_tokens": 1233125376}
{"current_steps": 1177, "total_steps": 2679, "loss": 0.636, "learning_rate": 3.0141815822910094e-06, "epoch": 1.3169230769230769, "percentage": 43.93, "elapsed_time": "19:15:13", "remaining_time": "1 day, 0:34:12", "throughput": "17805.66", "total_tokens": 1234173952}
{"current_steps": 1178, "total_steps": 2679, "loss": 0.8924, "learning_rate": 3.011285190260879e-06, "epoch": 1.318041958041958, "percentage": 43.97, "elapsed_time": "19:16:19", "remaining_time": "1 day, 0:33:23", "throughput": "17803.73", "total_tokens": 1235222528}
{"current_steps": 1179, "total_steps": 2679, "loss": 0.6605, "learning_rate": 3.008388081821687e-06, "epoch": 1.3191608391608391, "percentage": 44.01, "elapsed_time": "19:17:14", "remaining_time": "1 day, 0:32:19", "throughput": "17804.86", "total_tokens": 1236271104}
{"current_steps": 1180, "total_steps": 2679, "loss": 0.6842, "learning_rate": 3.005490261032839e-06, "epoch": 1.3202797202797203, "percentage": 44.05, "elapsed_time": "19:18:08", "remaining_time": "1 day, 0:31:14", "throughput": "17806.04", "total_tokens": 1237319680}
{"current_steps": 1181, "total_steps": 2679, "loss": 0.76, "learning_rate": 3.0025917319547417e-06, "epoch": 1.3213986013986014, "percentage": 44.08, "elapsed_time": "19:19:08", "remaining_time": "1 day, 0:30:16", "throughput": "17805.73", "total_tokens": 1238368256}
{"current_steps": 1182, "total_steps": 2679, "loss": 0.6304, "learning_rate": 2.999692498648792e-06, "epoch": 1.3225174825174826, "percentage": 44.12, "elapsed_time": "19:20:02", "remaining_time": "1 day, 0:29:11", "throughput": "17807.10", "total_tokens": 1239416832}
{"current_steps": 1183, "total_steps": 2679, "loss": 0.5649, "learning_rate": 2.9967925651773745e-06, "epoch": 1.3236363636363637, "percentage": 44.16, "elapsed_time": "19:20:50", "remaining_time": "1 day, 0:27:59", "throughput": "17809.77", "total_tokens": 1240465408}
{"current_steps": 1184, "total_steps": 2679, "loss": 0.774, "learning_rate": 2.9938919356038548e-06, "epoch": 1.3247552447552446, "percentage": 44.2, "elapsed_time": "19:21:56", "remaining_time": "1 day, 0:27:08", "throughput": "17808.07", "total_tokens": 1241513984}
{"current_steps": 1185, "total_steps": 2679, "loss": 0.6848, "learning_rate": 2.990990613992573e-06, "epoch": 1.325874125874126, "percentage": 44.23, "elapsed_time": "19:22:49", "remaining_time": "1 day, 0:26:03", "throughput": "17809.45", "total_tokens": 1242562560}
{"current_steps": 1186, "total_steps": 2679, "loss": 0.7088, "learning_rate": 2.9880886044088416e-06, "epoch": 1.326993006993007, "percentage": 44.27, "elapsed_time": "19:23:44", "remaining_time": "1 day, 0:24:58", "throughput": "17810.53", "total_tokens": 1243611136}
{"current_steps": 1187, "total_steps": 2679, "loss": 0.6251, "learning_rate": 2.9851859109189335e-06, "epoch": 1.328111888111888, "percentage": 44.31, "elapsed_time": "19:24:38", "remaining_time": "1 day, 0:23:53", "throughput": "17811.89", "total_tokens": 1244659712}
{"current_steps": 1188, "total_steps": 2679, "loss": 0.6438, "learning_rate": 2.9822825375900816e-06, "epoch": 1.3292307692307692, "percentage": 44.34, "elapsed_time": "19:25:32", "remaining_time": "1 day, 0:22:48", "throughput": "17813.10", "total_tokens": 1245708288}
{"current_steps": 1189, "total_steps": 2679, "loss": 0.77, "learning_rate": 2.9793784884904733e-06, "epoch": 1.3303496503496504, "percentage": 44.38, "elapsed_time": "19:26:36", "remaining_time": "1 day, 0:21:56", "throughput": "17811.60", "total_tokens": 1246756864}
{"current_steps": 1190, "total_steps": 2679, "loss": 0.7754, "learning_rate": 2.9764737676892375e-06, "epoch": 1.3314685314685315, "percentage": 44.42, "elapsed_time": "19:27:37", "remaining_time": "1 day, 0:21:00", "throughput": "17811.04", "total_tokens": 1247805440}
{"current_steps": 1191, "total_steps": 2679, "loss": 0.6628, "learning_rate": 2.9735683792564506e-06, "epoch": 1.3325874125874126, "percentage": 44.46, "elapsed_time": "19:28:32", "remaining_time": "1 day, 0:19:56", "throughput": "17812.04", "total_tokens": 1248854016}
{"current_steps": 1192, "total_steps": 2679, "loss": 0.7496, "learning_rate": 2.9706623272631206e-06, "epoch": 1.3337062937062938, "percentage": 44.49, "elapsed_time": "19:29:38", "remaining_time": "1 day, 0:19:05", "throughput": "17810.45", "total_tokens": 1249902592}
{"current_steps": 1193, "total_steps": 2679, "loss": 0.6997, "learning_rate": 2.967755615781186e-06, "epoch": 1.3348251748251747, "percentage": 44.53, "elapsed_time": "19:30:32", "remaining_time": "1 day, 0:18:01", "throughput": "17811.62", "total_tokens": 1250951168}
{"current_steps": 1194, "total_steps": 2679, "loss": 0.6166, "learning_rate": 2.96484824888351e-06, "epoch": 1.335944055944056, "percentage": 44.57, "elapsed_time": "19:31:25", "remaining_time": "1 day, 0:16:55", "throughput": "17812.96", "total_tokens": 1251999744}
{"current_steps": 1195, "total_steps": 2679, "loss": 0.6393, "learning_rate": 2.9619402306438738e-06, "epoch": 1.337062937062937, "percentage": 44.61, "elapsed_time": "19:32:25", "remaining_time": "1 day, 0:15:57", "throughput": "17812.88", "total_tokens": 1253048320}
{"current_steps": 1196, "total_steps": 2679, "loss": 0.6455, "learning_rate": 2.959031565136971e-06, "epoch": 1.3381818181818181, "percentage": 44.64, "elapsed_time": "19:33:18", "remaining_time": "1 day, 0:14:51", "throughput": "17814.26", "total_tokens": 1254096896}
{"current_steps": 1197, "total_steps": 2679, "loss": 0.7384, "learning_rate": 2.956122256438403e-06, "epoch": 1.3393006993006993, "percentage": 44.68, "elapsed_time": "19:34:21", "remaining_time": "1 day, 0:13:57", "throughput": "17813.29", "total_tokens": 1255145472}
{"current_steps": 1198, "total_steps": 2679, "loss": 0.6676, "learning_rate": 2.9532123086246704e-06, "epoch": 1.3404195804195804, "percentage": 44.72, "elapsed_time": "19:35:14", "remaining_time": "1 day, 0:12:52", "throughput": "17814.65", "total_tokens": 1256194048}
{"current_steps": 1199, "total_steps": 2679, "loss": 0.7254, "learning_rate": 2.9503017257731727e-06, "epoch": 1.3415384615384616, "percentage": 44.76, "elapsed_time": "19:36:14", "remaining_time": "1 day, 0:11:54", "throughput": "17814.40", "total_tokens": 1257242624}
{"current_steps": 1200, "total_steps": 2679, "loss": 0.7849, "learning_rate": 2.947390511962196e-06, "epoch": 1.3426573426573427, "percentage": 44.79, "elapsed_time": "19:37:20", "remaining_time": "1 day, 0:11:03", "throughput": "17812.69", "total_tokens": 1258291200}
{"current_steps": 1201, "total_steps": 2679, "loss": 0.7132, "learning_rate": 2.9444786712709122e-06, "epoch": 1.3437762237762239, "percentage": 44.83, "elapsed_time": "19:38:14", "remaining_time": "1 day, 0:09:59", "throughput": "17813.84", "total_tokens": 1259339776}
{"current_steps": 1202, "total_steps": 2679, "loss": 0.7177, "learning_rate": 2.9415662077793705e-06, "epoch": 1.3448951048951048, "percentage": 44.87, "elapsed_time": "19:39:23", "remaining_time": "1 day, 0:09:13", "throughput": "17811.24", "total_tokens": 1260388352}
{"current_steps": 1203, "total_steps": 2679, "loss": 0.6699, "learning_rate": 2.9386531255684942e-06, "epoch": 1.3460139860139861, "percentage": 44.9, "elapsed_time": "19:40:20", "remaining_time": "1 day, 0:08:12", "throughput": "17811.74", "total_tokens": 1261436928}
{"current_steps": 1204, "total_steps": 2679, "loss": 0.6733, "learning_rate": 2.9357394287200724e-06, "epoch": 1.347132867132867, "percentage": 44.94, "elapsed_time": "19:41:14", "remaining_time": "1 day, 0:07:07", "throughput": "17812.85", "total_tokens": 1262485504}
{"current_steps": 1205, "total_steps": 2679, "loss": 0.7537, "learning_rate": 2.9328251213167557e-06, "epoch": 1.3482517482517482, "percentage": 44.98, "elapsed_time": "19:42:14", "remaining_time": "1 day, 0:06:10", "throughput": "17812.60", "total_tokens": 1263534080}
{"current_steps": 1206, "total_steps": 2679, "loss": 0.8106, "learning_rate": 2.9299102074420504e-06, "epoch": 1.3493706293706293, "percentage": 45.02, "elapsed_time": "19:43:19", "remaining_time": "1 day, 0:05:18", "throughput": "17811.07", "total_tokens": 1264582656}
{"current_steps": 1207, "total_steps": 2679, "loss": 0.7536, "learning_rate": 2.9269946911803134e-06, "epoch": 1.3504895104895105, "percentage": 45.05, "elapsed_time": "19:44:13", "remaining_time": "1 day, 0:04:13", "throughput": "17812.31", "total_tokens": 1265631232}
{"current_steps": 1208, "total_steps": 2679, "loss": 0.6991, "learning_rate": 2.9240785766167428e-06, "epoch": 1.3516083916083916, "percentage": 45.09, "elapsed_time": "19:45:13", "remaining_time": "1 day, 0:03:15", "throughput": "17812.12", "total_tokens": 1266679808}
{"current_steps": 1209, "total_steps": 2679, "loss": 0.7489, "learning_rate": 2.9211618678373775e-06, "epoch": 1.3527272727272728, "percentage": 45.13, "elapsed_time": "19:46:18", "remaining_time": "1 day, 0:02:25", "throughput": "17810.45", "total_tokens": 1267728384}
{"current_steps": 1210, "total_steps": 2679, "loss": 0.6784, "learning_rate": 2.9182445689290894e-06, "epoch": 1.353846153846154, "percentage": 45.17, "elapsed_time": "19:47:13", "remaining_time": "1 day, 0:01:20", "throughput": "17811.53", "total_tokens": 1268776960}
{"current_steps": 1211, "total_steps": 2679, "loss": 0.6933, "learning_rate": 2.9153266839795756e-06, "epoch": 1.3549650349650348, "percentage": 45.2, "elapsed_time": "19:48:13", "remaining_time": "1 day, 0:00:23", "throughput": "17811.27", "total_tokens": 1269825536}
{"current_steps": 1212, "total_steps": 2679, "loss": 0.6613, "learning_rate": 2.912408217077356e-06, "epoch": 1.3560839160839162, "percentage": 45.24, "elapsed_time": "19:49:07", "remaining_time": "23:59:18", "throughput": "17812.50", "total_tokens": 1270874112}
{"current_steps": 1213, "total_steps": 2679, "loss": 0.7275, "learning_rate": 2.909489172311765e-06, "epoch": 1.3572027972027971, "percentage": 45.28, "elapsed_time": "19:50:07", "remaining_time": "23:58:20", "throughput": "17812.26", "total_tokens": 1271922688}
{"current_steps": 1214, "total_steps": 2679, "loss": 0.6383, "learning_rate": 2.906569553772945e-06, "epoch": 1.3583216783216783, "percentage": 45.32, "elapsed_time": "19:51:00", "remaining_time": "23:57:15", "throughput": "17813.53", "total_tokens": 1272971264}
{"current_steps": 1215, "total_steps": 2679, "loss": 0.715, "learning_rate": 2.9036493655518456e-06, "epoch": 1.3594405594405594, "percentage": 45.35, "elapsed_time": "19:52:00", "remaining_time": "23:56:17", "throughput": "17813.42", "total_tokens": 1274019840}
{"current_steps": 1216, "total_steps": 2679, "loss": 0.6316, "learning_rate": 2.900728611740213e-06, "epoch": 1.3605594405594406, "percentage": 45.39, "elapsed_time": "19:52:48", "remaining_time": "23:55:06", "throughput": "17815.97", "total_tokens": 1275068416}
{"current_steps": 1217, "total_steps": 2679, "loss": 0.787, "learning_rate": 2.8978072964305848e-06, "epoch": 1.3616783216783217, "percentage": 45.43, "elapsed_time": "19:53:54", "remaining_time": "23:54:15", "throughput": "17814.34", "total_tokens": 1276116992}
{"current_steps": 1218, "total_steps": 2679, "loss": 0.799, "learning_rate": 2.894885423716289e-06, "epoch": 1.3627972027972028, "percentage": 45.46, "elapsed_time": "19:55:05", "remaining_time": "23:53:31", "throughput": "17811.27", "total_tokens": 1277165568}
{"current_steps": 1219, "total_steps": 2679, "loss": 0.7139, "learning_rate": 2.89196299769143e-06, "epoch": 1.363916083916084, "percentage": 45.5, "elapsed_time": "19:56:00", "remaining_time": "23:52:27", "throughput": "17812.24", "total_tokens": 1278214144}
{"current_steps": 1220, "total_steps": 2679, "loss": 0.7172, "learning_rate": 2.8890400224508912e-06, "epoch": 1.365034965034965, "percentage": 45.54, "elapsed_time": "19:56:58", "remaining_time": "23:51:28", "throughput": "17812.32", "total_tokens": 1279262720}
{"current_steps": 1221, "total_steps": 2679, "loss": 0.7018, "learning_rate": 2.8861165020903235e-06, "epoch": 1.3661538461538463, "percentage": 45.58, "elapsed_time": "19:57:58", "remaining_time": "23:50:29", "throughput": "17812.27", "total_tokens": 1280311296}
{"current_steps": 1222, "total_steps": 2679, "loss": 0.6721, "learning_rate": 2.883192440706141e-06, "epoch": 1.3672727272727272, "percentage": 45.61, "elapsed_time": "19:58:54", "remaining_time": "23:49:28", "throughput": "17812.76", "total_tokens": 1281359872}
{"current_steps": 1223, "total_steps": 2679, "loss": 0.7519, "learning_rate": 2.88026784239552e-06, "epoch": 1.3683916083916083, "percentage": 45.65, "elapsed_time": "20:00:00", "remaining_time": "23:48:37", "throughput": "17811.08", "total_tokens": 1282408448}
{"current_steps": 1224, "total_steps": 2679, "loss": 0.8246, "learning_rate": 2.8773427112563833e-06, "epoch": 1.3695104895104895, "percentage": 45.69, "elapsed_time": "20:01:00", "remaining_time": "23:47:40", "throughput": "17810.71", "total_tokens": 1283457024}
{"current_steps": 1225, "total_steps": 2679, "loss": 0.7777, "learning_rate": 2.8744170513874054e-06, "epoch": 1.3706293706293706, "percentage": 45.73, "elapsed_time": "20:02:00", "remaining_time": "23:46:43", "throughput": "17810.46", "total_tokens": 1284505600}
{"current_steps": 1226, "total_steps": 2679, "loss": 0.7001, "learning_rate": 2.871490866887998e-06, "epoch": 1.3717482517482518, "percentage": 45.76, "elapsed_time": "20:03:00", "remaining_time": "23:45:45", "throughput": "17810.16", "total_tokens": 1285554176}
{"current_steps": 1227, "total_steps": 2679, "loss": 0.6791, "learning_rate": 2.8685641618583098e-06, "epoch": 1.372867132867133, "percentage": 45.8, "elapsed_time": "20:03:55", "remaining_time": "23:44:41", "throughput": "17811.22", "total_tokens": 1286602752}
{"current_steps": 1228, "total_steps": 2679, "loss": 0.7116, "learning_rate": 2.8656369403992192e-06, "epoch": 1.373986013986014, "percentage": 45.84, "elapsed_time": "20:04:49", "remaining_time": "23:43:36", "throughput": "17812.49", "total_tokens": 1287651328}
{"current_steps": 1229, "total_steps": 2679, "loss": 0.6922, "learning_rate": 2.8627092066123263e-06, "epoch": 1.375104895104895, "percentage": 45.88, "elapsed_time": "20:05:48", "remaining_time": "23:42:38", "throughput": "17812.32", "total_tokens": 1288699904}
{"current_steps": 1230, "total_steps": 2679, "loss": 0.7006, "learning_rate": 2.85978096459995e-06, "epoch": 1.3762237762237763, "percentage": 45.91, "elapsed_time": "20:06:48", "remaining_time": "23:41:40", "throughput": "17812.18", "total_tokens": 1289748480}
{"current_steps": 1231, "total_steps": 2679, "loss": 0.5692, "learning_rate": 2.856852218465124e-06, "epoch": 1.3773426573426573, "percentage": 45.95, "elapsed_time": "20:07:36", "remaining_time": "23:40:28", "throughput": "17814.87", "total_tokens": 1290797056}
{"current_steps": 1232, "total_steps": 2679, "loss": 0.6211, "learning_rate": 2.8539229723115826e-06, "epoch": 1.3784615384615384, "percentage": 45.99, "elapsed_time": "20:08:29", "remaining_time": "23:39:23", "throughput": "17816.15", "total_tokens": 1291845632}
{"current_steps": 1233, "total_steps": 2679, "loss": 0.772, "learning_rate": 2.8509932302437665e-06, "epoch": 1.3795804195804195, "percentage": 46.02, "elapsed_time": "20:09:29", "remaining_time": "23:38:25", "throughput": "17815.96", "total_tokens": 1292894208}
{"current_steps": 1234, "total_steps": 2679, "loss": 0.6958, "learning_rate": 2.8480629963668075e-06, "epoch": 1.3806993006993007, "percentage": 46.06, "elapsed_time": "20:10:29", "remaining_time": "23:37:27", "throughput": "17815.77", "total_tokens": 1293942784}
{"current_steps": 1235, "total_steps": 2679, "loss": 0.6951, "learning_rate": 2.8451322747865286e-06, "epoch": 1.3818181818181818, "percentage": 46.1, "elapsed_time": "20:11:29", "remaining_time": "23:36:31", "throughput": "17815.32", "total_tokens": 1294991360}
{"current_steps": 1236, "total_steps": 2679, "loss": 0.7116, "learning_rate": 2.8422010696094356e-06, "epoch": 1.382937062937063, "percentage": 46.14, "elapsed_time": "20:12:23", "remaining_time": "23:35:26", "throughput": "17816.50", "total_tokens": 1296039936}
{"current_steps": 1237, "total_steps": 2679, "loss": 0.7113, "learning_rate": 2.83926938494271e-06, "epoch": 1.3840559440559441, "percentage": 46.17, "elapsed_time": "20:13:25", "remaining_time": "23:34:31", "throughput": "17815.71", "total_tokens": 1297088512}
{"current_steps": 1238, "total_steps": 2679, "loss": 0.6888, "learning_rate": 2.836337224894209e-06, "epoch": 1.385174825174825, "percentage": 46.21, "elapsed_time": "20:14:20", "remaining_time": "23:33:27", "throughput": "17816.86", "total_tokens": 1298137088}
{"current_steps": 1239, "total_steps": 2679, "loss": 0.682, "learning_rate": 2.833404593572453e-06, "epoch": 1.3862937062937064, "percentage": 46.25, "elapsed_time": "20:15:13", "remaining_time": "23:32:22", "throughput": "17818.16", "total_tokens": 1299185664}
{"current_steps": 1240, "total_steps": 2679, "loss": 0.8484, "learning_rate": 2.8304714950866225e-06, "epoch": 1.3874125874125873, "percentage": 46.29, "elapsed_time": "20:16:19", "remaining_time": "23:31:31", "throughput": "17816.40", "total_tokens": 1300234240}
{"current_steps": 1241, "total_steps": 2679, "loss": 0.6114, "learning_rate": 2.827537933546555e-06, "epoch": 1.3885314685314685, "percentage": 46.32, "elapsed_time": "20:17:08", "remaining_time": "23:30:21", "throughput": "17818.83", "total_tokens": 1301282816}
{"current_steps": 1242, "total_steps": 2679, "loss": 0.7367, "learning_rate": 2.8246039130627356e-06, "epoch": 1.3896503496503496, "percentage": 46.36, "elapsed_time": "20:18:07", "remaining_time": "23:29:22", "throughput": "17818.71", "total_tokens": 1302331392}
{"current_steps": 1243, "total_steps": 2679, "loss": 0.7446, "learning_rate": 2.821669437746291e-06, "epoch": 1.3907692307692308, "percentage": 46.4, "elapsed_time": "20:19:07", "remaining_time": "23:28:24", "throughput": "17818.60", "total_tokens": 1303379968}
{"current_steps": 1244, "total_steps": 2679, "loss": 0.575, "learning_rate": 2.818734511708987e-06, "epoch": 1.391888111888112, "percentage": 46.44, "elapsed_time": "20:19:55", "remaining_time": "23:27:13", "throughput": "17821.10", "total_tokens": 1304428544}
{"current_steps": 1245, "total_steps": 2679, "loss": 0.719, "learning_rate": 2.8157991390632206e-06, "epoch": 1.393006993006993, "percentage": 46.47, "elapsed_time": "20:20:58", "remaining_time": "23:26:19", "throughput": "17820.16", "total_tokens": 1305477120}
{"current_steps": 1246, "total_steps": 2679, "loss": 0.5899, "learning_rate": 2.812863323922015e-06, "epoch": 1.3941258741258742, "percentage": 46.51, "elapsed_time": "20:21:46", "remaining_time": "23:25:08", "throughput": "17822.83", "total_tokens": 1306525696}
{"current_steps": 1247, "total_steps": 2679, "loss": 0.783, "learning_rate": 2.8099270703990124e-06, "epoch": 1.395244755244755, "percentage": 46.55, "elapsed_time": "20:22:51", "remaining_time": "23:24:16", "throughput": "17821.24", "total_tokens": 1307574272}
{"current_steps": 1248, "total_steps": 2679, "loss": 0.8601, "learning_rate": 2.8069903826084695e-06, "epoch": 1.3963636363636365, "percentage": 46.58, "elapsed_time": "20:23:57", "remaining_time": "23:23:25", "throughput": "17819.54", "total_tokens": 1308622848}
{"current_steps": 1249, "total_steps": 2679, "loss": 0.6974, "learning_rate": 2.8040532646652515e-06, "epoch": 1.3974825174825174, "percentage": 46.62, "elapsed_time": "20:24:51", "remaining_time": "23:22:21", "throughput": "17820.67", "total_tokens": 1309671424}
{"current_steps": 1250, "total_steps": 2679, "loss": 0.7827, "learning_rate": 2.8011157206848266e-06, "epoch": 1.3986013986013985, "percentage": 46.66, "elapsed_time": "20:25:56", "remaining_time": "23:21:30", "throughput": "17819.16", "total_tokens": 1310720000}
{"current_steps": 1250, "total_steps": 2679, "eval_loss": 0.7286360263824463, "epoch": 1.3986013986013985, "percentage": 46.66, "elapsed_time": "20:30:03", "remaining_time": "23:26:12", "throughput": "17759.62", "total_tokens": 1310720000}
{"current_steps": 1251, "total_steps": 2679, "loss": 0.6823, "learning_rate": 2.7981777547832604e-06, "epoch": 1.3997202797202797, "percentage": 46.7, "elapsed_time": "20:31:30", "remaining_time": "23:25:45", "throughput": "17752.84", "total_tokens": 1311768576}
{"current_steps": 1252, "total_steps": 2679, "loss": 0.5896, "learning_rate": 2.7952393710772097e-06, "epoch": 1.4008391608391608, "percentage": 46.73, "elapsed_time": "20:32:19", "remaining_time": "23:24:34", "throughput": "17755.38", "total_tokens": 1312817152}
{"current_steps": 1253, "total_steps": 2679, "loss": 0.7552, "learning_rate": 2.792300573683915e-06, "epoch": 1.401958041958042, "percentage": 46.77, "elapsed_time": "20:33:19", "remaining_time": "23:23:36", "throughput": "17755.16", "total_tokens": 1313865728}
{"current_steps": 1254, "total_steps": 2679, "loss": 0.6488, "learning_rate": 2.7893613667211983e-06, "epoch": 1.403076923076923, "percentage": 46.81, "elapsed_time": "20:34:12", "remaining_time": "23:22:30", "throughput": "17756.51", "total_tokens": 1314914304}
{"current_steps": 1255, "total_steps": 2679, "loss": 0.6017, "learning_rate": 2.7864217543074544e-06, "epoch": 1.4041958041958043, "percentage": 46.85, "elapsed_time": "20:35:00", "remaining_time": "23:21:19", "throughput": "17759.11", "total_tokens": 1315962880}
{"current_steps": 1256, "total_steps": 2679, "loss": 0.6918, "learning_rate": 2.7834817405616476e-06, "epoch": 1.4053146853146854, "percentage": 46.88, "elapsed_time": "20:36:00", "remaining_time": "23:20:21", "throughput": "17758.93", "total_tokens": 1317011456}
{"current_steps": 1257, "total_steps": 2679, "loss": 0.6059, "learning_rate": 2.780541329603303e-06, "epoch": 1.4064335664335665, "percentage": 46.92, "elapsed_time": "20:36:48", "remaining_time": "23:19:09", "throughput": "17761.56", "total_tokens": 1318060032}
{"current_steps": 1258, "total_steps": 2679, "loss": 0.6015, "learning_rate": 2.7776005255525022e-06, "epoch": 1.4075524475524475, "percentage": 46.96, "elapsed_time": "20:37:36", "remaining_time": "23:17:58", "throughput": "17764.19", "total_tokens": 1319108608}
{"current_steps": 1259, "total_steps": 2679, "loss": 0.8008, "learning_rate": 2.77465933252988e-06, "epoch": 1.4086713286713286, "percentage": 47.0, "elapsed_time": "20:38:36", "remaining_time": "23:17:00", "throughput": "17763.88", "total_tokens": 1320157184}
{"current_steps": 1260, "total_steps": 2679, "loss": 0.8286, "learning_rate": 2.7717177546566126e-06, "epoch": 1.4097902097902097, "percentage": 47.03, "elapsed_time": "20:39:42", "remaining_time": "23:16:09", "throughput": "17762.28", "total_tokens": 1321205760}
{"current_steps": 1261, "total_steps": 2679, "loss": 0.763, "learning_rate": 2.7687757960544193e-06, "epoch": 1.410909090909091, "percentage": 47.07, "elapsed_time": "20:40:42", "remaining_time": "23:15:11", "throughput": "17762.06", "total_tokens": 1322254336}
{"current_steps": 1262, "total_steps": 2679, "loss": 0.7339, "learning_rate": 2.7658334608455495e-06, "epoch": 1.412027972027972, "percentage": 47.11, "elapsed_time": "20:41:42", "remaining_time": "23:14:12", "throughput": "17761.88", "total_tokens": 1323302912}
{"current_steps": 1263, "total_steps": 2679, "loss": 0.7166, "learning_rate": 2.7628907531527815e-06, "epoch": 1.4131468531468532, "percentage": 47.14, "elapsed_time": "20:42:42", "remaining_time": "23:13:14", "throughput": "17761.74", "total_tokens": 1324351488}
{"current_steps": 1264, "total_steps": 2679, "loss": 0.6082, "learning_rate": 2.759947677099417e-06, "epoch": 1.4142657342657343, "percentage": 47.18, "elapsed_time": "20:43:35", "remaining_time": "23:12:09", "throughput": "17762.95", "total_tokens": 1325400064}
{"current_steps": 1265, "total_steps": 2679, "loss": 0.7741, "learning_rate": 2.7570042368092724e-06, "epoch": 1.4153846153846155, "percentage": 47.22, "elapsed_time": "20:44:33", "remaining_time": "23:11:09", "throughput": "17763.21", "total_tokens": 1326448640}
{"current_steps": 1266, "total_steps": 2679, "loss": 0.7008, "learning_rate": 2.754060436406674e-06, "epoch": 1.4165034965034966, "percentage": 47.26, "elapsed_time": "20:45:33", "remaining_time": "23:10:11", "throughput": "17763.01", "total_tokens": 1327497216}
{"current_steps": 1267, "total_steps": 2679, "loss": 0.7292, "learning_rate": 2.7511162800164536e-06, "epoch": 1.4176223776223775, "percentage": 47.29, "elapsed_time": "20:46:33", "remaining_time": "23:09:13", "throughput": "17762.79", "total_tokens": 1328545792}
{"current_steps": 1268, "total_steps": 2679, "loss": 0.7581, "learning_rate": 2.748171771763941e-06, "epoch": 1.4187412587412587, "percentage": 47.33, "elapsed_time": "20:47:32", "remaining_time": "23:08:14", "throughput": "17762.78", "total_tokens": 1329594368}
{"current_steps": 1269, "total_steps": 2679, "loss": 0.7157, "learning_rate": 2.7452269157749614e-06, "epoch": 1.4198601398601398, "percentage": 47.37, "elapsed_time": "20:48:32", "remaining_time": "23:07:15", "throughput": "17762.70", "total_tokens": 1330642944}
{"current_steps": 1270, "total_steps": 2679, "loss": 0.6362, "learning_rate": 2.7422817161758234e-06, "epoch": 1.420979020979021, "percentage": 47.41, "elapsed_time": "20:49:26", "remaining_time": "23:06:11", "throughput": "17763.90", "total_tokens": 1331691520}
{"current_steps": 1271, "total_steps": 2679, "loss": 0.7119, "learning_rate": 2.7393361770933198e-06, "epoch": 1.422097902097902, "percentage": 47.44, "elapsed_time": "20:50:25", "remaining_time": "23:05:12", "throughput": "17763.74", "total_tokens": 1332740096}
{"current_steps": 1272, "total_steps": 2679, "loss": 0.8825, "learning_rate": 2.7363903026547196e-06, "epoch": 1.4232167832167832, "percentage": 47.48, "elapsed_time": "20:51:32", "remaining_time": "23:04:22", "throughput": "17761.90", "total_tokens": 1333788672}
{"current_steps": 1273, "total_steps": 2679, "loss": 0.6226, "learning_rate": 2.7334440969877584e-06, "epoch": 1.4243356643356644, "percentage": 47.52, "elapsed_time": "20:52:26", "remaining_time": "23:03:17", "throughput": "17763.08", "total_tokens": 1334837248}
{"current_steps": 1274, "total_steps": 2679, "loss": 0.7138, "learning_rate": 2.7304975642206394e-06, "epoch": 1.4254545454545455, "percentage": 47.56, "elapsed_time": "20:53:28", "remaining_time": "23:02:22", "throughput": "17762.37", "total_tokens": 1335885824}
{"current_steps": 1275, "total_steps": 2679, "loss": 0.7926, "learning_rate": 2.7275507084820226e-06, "epoch": 1.4265734265734267, "percentage": 47.59, "elapsed_time": "20:54:34", "remaining_time": "23:01:30", "throughput": "17760.78", "total_tokens": 1336934400}
{"current_steps": 1276, "total_steps": 2679, "loss": 0.7555, "learning_rate": 2.724603533901019e-06, "epoch": 1.4276923076923076, "percentage": 47.63, "elapsed_time": "20:55:39", "remaining_time": "23:00:37", "throughput": "17759.43", "total_tokens": 1337982976}
{"current_steps": 1277, "total_steps": 2679, "loss": 0.7921, "learning_rate": 2.7216560446071904e-06, "epoch": 1.4288111888111887, "percentage": 47.67, "elapsed_time": "20:56:45", "remaining_time": "22:59:46", "throughput": "17757.79", "total_tokens": 1339031552}
{"current_steps": 1278, "total_steps": 2679, "loss": 0.6401, "learning_rate": 2.718708244730537e-06, "epoch": 1.4299300699300699, "percentage": 47.7, "elapsed_time": "20:57:38", "remaining_time": "22:58:41", "throughput": "17759.12", "total_tokens": 1340080128}
{"current_steps": 1279, "total_steps": 2679, "loss": 0.6946, "learning_rate": 2.7157601384014927e-06, "epoch": 1.431048951048951, "percentage": 47.74, "elapsed_time": "20:58:38", "remaining_time": "22:57:42", "throughput": "17758.99", "total_tokens": 1341128704}
{"current_steps": 1280, "total_steps": 2679, "loss": 0.7291, "learning_rate": 2.7128117297509233e-06, "epoch": 1.4321678321678322, "percentage": 47.78, "elapsed_time": "20:59:43", "remaining_time": "22:56:49", "throughput": "17757.63", "total_tokens": 1342177280}
{"current_steps": 1281, "total_steps": 2679, "loss": 0.6842, "learning_rate": 2.7098630229101174e-06, "epoch": 1.4332867132867133, "percentage": 47.82, "elapsed_time": "21:00:36", "remaining_time": "22:55:45", "throughput": "17758.86", "total_tokens": 1343225856}
{"current_steps": 1282, "total_steps": 2679, "loss": 0.6862, "learning_rate": 2.706914022010782e-06, "epoch": 1.4344055944055945, "percentage": 47.85, "elapsed_time": "21:01:31", "remaining_time": "22:54:40", "throughput": "17759.99", "total_tokens": 1344274432}
{"current_steps": 1283, "total_steps": 2679, "loss": 0.6372, "learning_rate": 2.7039647311850347e-06, "epoch": 1.4355244755244756, "percentage": 47.89, "elapsed_time": "21:02:25", "remaining_time": "22:53:36", "throughput": "17761.15", "total_tokens": 1345323008}
{"current_steps": 1284, "total_steps": 2679, "loss": 0.6159, "learning_rate": 2.7010151545654006e-06, "epoch": 1.4366433566433567, "percentage": 47.93, "elapsed_time": "21:03:19", "remaining_time": "22:52:32", "throughput": "17762.32", "total_tokens": 1346371584}
{"current_steps": 1285, "total_steps": 2679, "loss": 0.7926, "learning_rate": 2.6980652962848055e-06, "epoch": 1.4377622377622377, "percentage": 47.97, "elapsed_time": "21:04:22", "remaining_time": "22:51:37", "throughput": "17761.36", "total_tokens": 1347420160}
{"current_steps": 1286, "total_steps": 2679, "loss": 0.8506, "learning_rate": 2.6951151604765668e-06, "epoch": 1.4388811188811188, "percentage": 48.0, "elapsed_time": "21:05:28", "remaining_time": "22:50:45", "throughput": "17759.79", "total_tokens": 1348468736}
{"current_steps": 1287, "total_steps": 2679, "loss": 0.8527, "learning_rate": 2.6921647512743963e-06, "epoch": 1.44, "percentage": 48.04, "elapsed_time": "21:06:39", "remaining_time": "22:49:59", "throughput": "17756.94", "total_tokens": 1349517312}
{"current_steps": 1288, "total_steps": 2679, "loss": 0.6399, "learning_rate": 2.689214072812384e-06, "epoch": 1.441118881118881, "percentage": 48.08, "elapsed_time": "21:07:35", "remaining_time": "22:48:57", "throughput": "17757.69", "total_tokens": 1350565888}
{"current_steps": 1289, "total_steps": 2679, "loss": 0.6748, "learning_rate": 2.686263129224999e-06, "epoch": 1.4422377622377622, "percentage": 48.11, "elapsed_time": "21:08:29", "remaining_time": "22:47:52", "throughput": "17758.93", "total_tokens": 1351614464}
{"current_steps": 1290, "total_steps": 2679, "loss": 0.6529, "learning_rate": 2.683311924647083e-06, "epoch": 1.4433566433566434, "percentage": 48.15, "elapsed_time": "21:09:29", "remaining_time": "22:46:54", "throughput": "17758.67", "total_tokens": 1352663040}
{"current_steps": 1291, "total_steps": 2679, "loss": 0.7679, "learning_rate": 2.6803604632138403e-06, "epoch": 1.4444755244755245, "percentage": 48.19, "elapsed_time": "21:10:40", "remaining_time": "22:46:09", "throughput": "17755.74", "total_tokens": 1353711616}
{"current_steps": 1292, "total_steps": 2679, "loss": 0.7352, "learning_rate": 2.6774087490608384e-06, "epoch": 1.4455944055944057, "percentage": 48.23, "elapsed_time": "21:11:40", "remaining_time": "22:45:11", "throughput": "17755.50", "total_tokens": 1354760192}
{"current_steps": 1293, "total_steps": 2679, "loss": 0.7613, "learning_rate": 2.674456786323998e-06, "epoch": 1.4467132867132868, "percentage": 48.26, "elapsed_time": "21:12:46", "remaining_time": "22:44:18", "throughput": "17754.06", "total_tokens": 1355808768}
{"current_steps": 1294, "total_steps": 2679, "loss": 0.717, "learning_rate": 2.6715045791395855e-06, "epoch": 1.4478321678321677, "percentage": 48.3, "elapsed_time": "21:13:39", "remaining_time": "22:43:14", "throughput": "17755.27", "total_tokens": 1356857344}
{"current_steps": 1295, "total_steps": 2679, "loss": 0.7243, "learning_rate": 2.668552131644214e-06, "epoch": 1.4489510489510489, "percentage": 48.34, "elapsed_time": "21:14:45", "remaining_time": "22:42:21", "throughput": "17753.85", "total_tokens": 1357905920}
{"current_steps": 1296, "total_steps": 2679, "loss": 0.7768, "learning_rate": 2.6655994479748313e-06, "epoch": 1.45006993006993, "percentage": 48.38, "elapsed_time": "21:15:46", "remaining_time": "22:41:24", "throughput": "17753.34", "total_tokens": 1358954496}
{"current_steps": 1297, "total_steps": 2679, "loss": 0.7053, "learning_rate": 2.6626465322687144e-06, "epoch": 1.4511888111888112, "percentage": 48.41, "elapsed_time": "21:16:45", "remaining_time": "22:40:26", "throughput": "17753.24", "total_tokens": 1360003072}
{"current_steps": 1298, "total_steps": 2679, "loss": 0.6195, "learning_rate": 2.659693388663469e-06, "epoch": 1.4523076923076923, "percentage": 48.45, "elapsed_time": "21:17:41", "remaining_time": "22:39:23", "throughput": "17754.02", "total_tokens": 1361051648}
{"current_steps": 1299, "total_steps": 2679, "loss": 0.6968, "learning_rate": 2.656740021297017e-06, "epoch": 1.4534265734265734, "percentage": 48.49, "elapsed_time": "21:18:36", "remaining_time": "22:38:19", "throughput": "17755.03", "total_tokens": 1362100224}
{"current_steps": 1300, "total_steps": 2679, "loss": 0.7104, "learning_rate": 2.653786434307596e-06, "epoch": 1.4545454545454546, "percentage": 48.53, "elapsed_time": "21:19:35", "remaining_time": "22:37:21", "throughput": "17754.97", "total_tokens": 1363148800}
{"current_steps": 1301, "total_steps": 2679, "loss": 0.7466, "learning_rate": 2.6508326318337498e-06, "epoch": 1.4556643356643357, "percentage": 48.56, "elapsed_time": "21:20:40", "remaining_time": "22:36:28", "throughput": "17753.64", "total_tokens": 1364197376}
{"current_steps": 1302, "total_steps": 2679, "loss": 0.7815, "learning_rate": 2.6478786180143253e-06, "epoch": 1.4567832167832169, "percentage": 48.6, "elapsed_time": "21:21:40", "remaining_time": "22:35:30", "throughput": "17753.36", "total_tokens": 1365245952}
{"current_steps": 1303, "total_steps": 2679, "loss": 0.6731, "learning_rate": 2.644924396988465e-06, "epoch": 1.4579020979020978, "percentage": 48.64, "elapsed_time": "21:22:39", "remaining_time": "22:34:31", "throughput": "17753.34", "total_tokens": 1366294528}
{"current_steps": 1304, "total_steps": 2679, "loss": 0.6903, "learning_rate": 2.641969972895601e-06, "epoch": 1.4590209790209792, "percentage": 48.67, "elapsed_time": "21:23:33", "remaining_time": "22:33:26", "throughput": "17754.64", "total_tokens": 1367343104}
{"current_steps": 1305, "total_steps": 2679, "loss": 0.7819, "learning_rate": 2.6390153498754506e-06, "epoch": 1.46013986013986, "percentage": 48.71, "elapsed_time": "21:24:44", "remaining_time": "22:32:40", "throughput": "17751.86", "total_tokens": 1368391680}
{"current_steps": 1306, "total_steps": 2679, "loss": 0.6417, "learning_rate": 2.6360605320680117e-06, "epoch": 1.4612587412587412, "percentage": 48.75, "elapsed_time": "21:25:41", "remaining_time": "22:31:38", "throughput": "17752.35", "total_tokens": 1369440256}
{"current_steps": 1307, "total_steps": 2679, "loss": 0.64, "learning_rate": 2.633105523613551e-06, "epoch": 1.4623776223776224, "percentage": 48.79, "elapsed_time": "21:26:35", "remaining_time": "22:30:34", "throughput": "17753.54", "total_tokens": 1370488832}
{"current_steps": 1308, "total_steps": 2679, "loss": 0.7603, "learning_rate": 2.6301503286526076e-06, "epoch": 1.4634965034965035, "percentage": 48.82, "elapsed_time": "21:27:35", "remaining_time": "22:29:36", "throughput": "17753.24", "total_tokens": 1371537408}
{"current_steps": 1309, "total_steps": 2679, "loss": 0.7417, "learning_rate": 2.6271949513259764e-06, "epoch": 1.4646153846153847, "percentage": 48.86, "elapsed_time": "21:28:37", "remaining_time": "22:28:40", "throughput": "17752.50", "total_tokens": 1372585984}
{"current_steps": 1310, "total_steps": 2679, "loss": 0.8494, "learning_rate": 2.6242393957747112e-06, "epoch": 1.4657342657342658, "percentage": 48.9, "elapsed_time": "21:29:43", "remaining_time": "22:27:49", "throughput": "17750.91", "total_tokens": 1373634560}
{"current_steps": 1311, "total_steps": 2679, "loss": 0.5646, "learning_rate": 2.6212836661401154e-06, "epoch": 1.466853146853147, "percentage": 48.94, "elapsed_time": "21:30:32", "remaining_time": "22:26:39", "throughput": "17753.32", "total_tokens": 1374683136}
{"current_steps": 1312, "total_steps": 2679, "loss": 0.6526, "learning_rate": 2.618327766563735e-06, "epoch": 1.4679720279720279, "percentage": 48.97, "elapsed_time": "21:31:26", "remaining_time": "22:25:34", "throughput": "17754.47", "total_tokens": 1375731712}
{"current_steps": 1313, "total_steps": 2679, "loss": 0.811, "learning_rate": 2.615371701187355e-06, "epoch": 1.4690909090909092, "percentage": 49.01, "elapsed_time": "21:32:26", "remaining_time": "22:24:36", "throughput": "17754.28", "total_tokens": 1376780288}
{"current_steps": 1314, "total_steps": 2679, "loss": 0.6345, "learning_rate": 2.6124154741529934e-06, "epoch": 1.4702097902097901, "percentage": 49.05, "elapsed_time": "21:33:16", "remaining_time": "22:23:27", "throughput": "17756.42", "total_tokens": 1377828864}
{"current_steps": 1315, "total_steps": 2679, "loss": 0.8762, "learning_rate": 2.609459089602892e-06, "epoch": 1.4713286713286713, "percentage": 49.09, "elapsed_time": "21:34:22", "remaining_time": "22:22:36", "throughput": "17754.77", "total_tokens": 1378877440}
{"current_steps": 1316, "total_steps": 2679, "loss": 0.6337, "learning_rate": 2.6065025516795165e-06, "epoch": 1.4724475524475524, "percentage": 49.12, "elapsed_time": "21:35:17", "remaining_time": "22:21:32", "throughput": "17755.72", "total_tokens": 1379926016}
{"current_steps": 1317, "total_steps": 2679, "loss": 0.6049, "learning_rate": 2.6035458645255467e-06, "epoch": 1.4735664335664336, "percentage": 49.16, "elapsed_time": "21:36:05", "remaining_time": "22:20:22", "throughput": "17758.19", "total_tokens": 1380974592}
{"current_steps": 1318, "total_steps": 2679, "loss": 0.6951, "learning_rate": 2.6005890322838697e-06, "epoch": 1.4746853146853147, "percentage": 49.2, "elapsed_time": "21:36:59", "remaining_time": "22:19:18", "throughput": "17759.39", "total_tokens": 1382023168}
{"current_steps": 1319, "total_steps": 2679, "loss": 0.5612, "learning_rate": 2.597632059097577e-06, "epoch": 1.4758041958041959, "percentage": 49.23, "elapsed_time": "21:37:47", "remaining_time": "22:18:07", "throughput": "17761.95", "total_tokens": 1383071744}
{"current_steps": 1320, "total_steps": 2679, "loss": 0.7296, "learning_rate": 2.5946749491099597e-06, "epoch": 1.476923076923077, "percentage": 49.27, "elapsed_time": "21:38:47", "remaining_time": "22:17:09", "throughput": "17761.74", "total_tokens": 1384120320}
{"current_steps": 1321, "total_steps": 2679, "loss": 0.7324, "learning_rate": 2.5917177064644974e-06, "epoch": 1.478041958041958, "percentage": 49.31, "elapsed_time": "21:39:46", "remaining_time": "22:16:10", "throughput": "17761.64", "total_tokens": 1385168896}
{"current_steps": 1322, "total_steps": 2679, "loss": 0.7148, "learning_rate": 2.5887603353048564e-06, "epoch": 1.4791608391608393, "percentage": 49.35, "elapsed_time": "21:40:46", "remaining_time": "22:15:12", "throughput": "17761.45", "total_tokens": 1386217472}
{"current_steps": 1323, "total_steps": 2679, "loss": 0.8008, "learning_rate": 2.585802839774883e-06, "epoch": 1.4802797202797202, "percentage": 49.38, "elapsed_time": "21:41:46", "remaining_time": "22:14:15", "throughput": "17761.11", "total_tokens": 1387266048}
{"current_steps": 1324, "total_steps": 2679, "loss": 0.7134, "learning_rate": 2.5828452240186002e-06, "epoch": 1.4813986013986014, "percentage": 49.42, "elapsed_time": "21:42:41", "remaining_time": "22:13:11", "throughput": "17762.11", "total_tokens": 1388314624}
{"current_steps": 1325, "total_steps": 2679, "loss": 0.5837, "learning_rate": 2.579887492180197e-06, "epoch": 1.4825174825174825, "percentage": 49.46, "elapsed_time": "21:43:29", "remaining_time": "22:12:01", "throughput": "17764.55", "total_tokens": 1389363200}
{"current_steps": 1326, "total_steps": 2679, "loss": 0.6575, "learning_rate": 2.576929648404025e-06, "epoch": 1.4836363636363636, "percentage": 49.5, "elapsed_time": "21:44:25", "remaining_time": "22:10:58", "throughput": "17765.39", "total_tokens": 1390411776}
{"current_steps": 1327, "total_steps": 2679, "loss": 0.6686, "learning_rate": 2.5739716968345922e-06, "epoch": 1.4847552447552448, "percentage": 49.53, "elapsed_time": "21:45:24", "remaining_time": "22:10:00", "throughput": "17765.33", "total_tokens": 1391460352}
{"current_steps": 1328, "total_steps": 2679, "loss": 0.5814, "learning_rate": 2.5710136416165602e-06, "epoch": 1.485874125874126, "percentage": 49.57, "elapsed_time": "21:46:12", "remaining_time": "22:08:50", "throughput": "17767.72", "total_tokens": 1392508928}
{"current_steps": 1329, "total_steps": 2679, "loss": 0.6269, "learning_rate": 2.5680554868947346e-06, "epoch": 1.486993006993007, "percentage": 49.61, "elapsed_time": "21:47:06", "remaining_time": "22:07:46", "throughput": "17768.90", "total_tokens": 1393557504}
{"current_steps": 1330, "total_steps": 2679, "loss": 0.5988, "learning_rate": 2.5650972368140587e-06, "epoch": 1.488111888111888, "percentage": 49.65, "elapsed_time": "21:47:55", "remaining_time": "22:06:36", "throughput": "17771.34", "total_tokens": 1394606080}
{"current_steps": 1331, "total_steps": 2679, "loss": 0.6707, "learning_rate": 2.5621388955196113e-06, "epoch": 1.4892307692307694, "percentage": 49.68, "elapsed_time": "21:48:54", "remaining_time": "22:05:37", "throughput": "17771.15", "total_tokens": 1395654656}
{"current_steps": 1332, "total_steps": 2679, "loss": 0.627, "learning_rate": 2.5591804671566003e-06, "epoch": 1.4903496503496503, "percentage": 49.72, "elapsed_time": "21:49:48", "remaining_time": "22:04:33", "throughput": "17772.26", "total_tokens": 1396703232}
{"current_steps": 1333, "total_steps": 2679, "loss": 0.8331, "learning_rate": 2.5562219558703504e-06, "epoch": 1.4914685314685314, "percentage": 49.76, "elapsed_time": "21:50:54", "remaining_time": "22:03:41", "throughput": "17770.83", "total_tokens": 1397751808}
{"current_steps": 1334, "total_steps": 2679, "loss": 0.6928, "learning_rate": 2.5532633658063095e-06, "epoch": 1.4925874125874126, "percentage": 49.79, "elapsed_time": "21:51:48", "remaining_time": "22:02:37", "throughput": "17771.97", "total_tokens": 1398800384}
{"current_steps": 1335, "total_steps": 2679, "loss": 0.557, "learning_rate": 2.55030470111003e-06, "epoch": 1.4937062937062937, "percentage": 49.83, "elapsed_time": "21:52:36", "remaining_time": "22:01:26", "throughput": "17774.51", "total_tokens": 1399848960}
{"current_steps": 1336, "total_steps": 2679, "loss": 0.8429, "learning_rate": 2.5473459659271715e-06, "epoch": 1.4948251748251749, "percentage": 49.87, "elapsed_time": "21:53:38", "remaining_time": "22:00:31", "throughput": "17773.78", "total_tokens": 1400897536}
{"current_steps": 1337, "total_steps": 2679, "loss": 0.7329, "learning_rate": 2.544387164403493e-06, "epoch": 1.495944055944056, "percentage": 49.91, "elapsed_time": "21:54:37", "remaining_time": "21:59:32", "throughput": "17773.61", "total_tokens": 1401946112}
{"current_steps": 1338, "total_steps": 2679, "loss": 0.7075, "learning_rate": 2.541428300684845e-06, "epoch": 1.4970629370629371, "percentage": 49.94, "elapsed_time": "21:55:37", "remaining_time": "21:58:34", "throughput": "17773.44", "total_tokens": 1402994688}
{"current_steps": 1339, "total_steps": 2679, "loss": 0.6779, "learning_rate": 2.5384693789171656e-06, "epoch": 1.498181818181818, "percentage": 49.98, "elapsed_time": "21:56:32", "remaining_time": "21:57:31", "throughput": "17774.47", "total_tokens": 1404043264}
{"current_steps": 1340, "total_steps": 2679, "loss": 0.7811, "learning_rate": 2.5355104032464746e-06, "epoch": 1.4993006993006994, "percentage": 50.02, "elapsed_time": "21:57:43", "remaining_time": "21:56:44", "throughput": "17771.73", "total_tokens": 1405091840}
{"current_steps": 1341, "total_steps": 2679, "loss": 0.7379, "learning_rate": 2.532551377818866e-06, "epoch": 1.5004195804195803, "percentage": 50.06, "elapsed_time": "21:58:48", "remaining_time": "21:55:51", "throughput": "17770.32", "total_tokens": 1406140416}
{"current_steps": 1342, "total_steps": 2679, "loss": 0.9567, "learning_rate": 2.5295923067805054e-06, "epoch": 1.5015384615384615, "percentage": 50.09, "elapsed_time": "22:00:00", "remaining_time": "21:55:05", "throughput": "17767.47", "total_tokens": 1407188992}
{"current_steps": 1343, "total_steps": 2679, "loss": 0.6989, "learning_rate": 2.526633194277622e-06, "epoch": 1.5026573426573426, "percentage": 50.13, "elapsed_time": "22:00:59", "remaining_time": "21:54:06", "throughput": "17767.35", "total_tokens": 1408237568}
{"current_steps": 1344, "total_steps": 2679, "loss": 0.7396, "learning_rate": 2.5236740444565016e-06, "epoch": 1.5037762237762238, "percentage": 50.17, "elapsed_time": "22:01:56", "remaining_time": "21:53:05", "throughput": "17767.89", "total_tokens": 1409286144}
{"current_steps": 1345, "total_steps": 2679, "loss": 0.6208, "learning_rate": 2.5207148614634836e-06, "epoch": 1.504895104895105, "percentage": 50.21, "elapsed_time": "22:02:48", "remaining_time": "21:51:59", "throughput": "17769.43", "total_tokens": 1410334720}
{"current_steps": 1346, "total_steps": 2679, "loss": 0.6852, "learning_rate": 2.5177556494449534e-06, "epoch": 1.506013986013986, "percentage": 50.24, "elapsed_time": "22:03:49", "remaining_time": "21:51:02", "throughput": "17769.08", "total_tokens": 1411383296}
{"current_steps": 1347, "total_steps": 2679, "loss": 0.7523, "learning_rate": 2.514796412547337e-06, "epoch": 1.5071328671328672, "percentage": 50.28, "elapsed_time": "22:04:49", "remaining_time": "21:50:03", "throughput": "17768.88", "total_tokens": 1412431872}
{"current_steps": 1348, "total_steps": 2679, "loss": 0.6517, "learning_rate": 2.5118371549170967e-06, "epoch": 1.5082517482517481, "percentage": 50.32, "elapsed_time": "22:05:43", "remaining_time": "21:49:00", "throughput": "17769.96", "total_tokens": 1413480448}
{"current_steps": 1349, "total_steps": 2679, "loss": 0.6122, "learning_rate": 2.5088778807007203e-06, "epoch": 1.5093706293706295, "percentage": 50.35, "elapsed_time": "22:06:31", "remaining_time": "21:47:50", "throughput": "17772.34", "total_tokens": 1414529024}
{"current_steps": 1350, "total_steps": 2679, "loss": 0.7379, "learning_rate": 2.505918594044724e-06, "epoch": 1.5104895104895104, "percentage": 50.39, "elapsed_time": "22:07:30", "remaining_time": "21:46:51", "throughput": "17772.26", "total_tokens": 1415577600}
{"current_steps": 1351, "total_steps": 2679, "loss": 0.7122, "learning_rate": 2.502959299095636e-06, "epoch": 1.5116083916083916, "percentage": 50.43, "elapsed_time": "22:08:31", "remaining_time": "21:45:54", "throughput": "17772.00", "total_tokens": 1416626176}
{"current_steps": 1352, "total_steps": 2679, "loss": 0.588, "learning_rate": 2.5e-06, "epoch": 1.5127272727272727, "percentage": 50.47, "elapsed_time": "22:09:21", "remaining_time": "21:44:46", "throughput": "17773.99", "total_tokens": 1417674752}
{"current_steps": 1353, "total_steps": 2679, "loss": 0.6763, "learning_rate": 2.4970407009043646e-06, "epoch": 1.5138461538461538, "percentage": 50.5, "elapsed_time": "22:10:19", "remaining_time": "21:43:46", "throughput": "17774.08", "total_tokens": 1418723328}
{"current_steps": 1354, "total_steps": 2679, "loss": 0.7872, "learning_rate": 2.4940814059552763e-06, "epoch": 1.514965034965035, "percentage": 50.54, "elapsed_time": "22:11:25", "remaining_time": "21:42:54", "throughput": "17772.65", "total_tokens": 1419771904}
{"current_steps": 1355, "total_steps": 2679, "loss": 0.7358, "learning_rate": 2.49112211929928e-06, "epoch": 1.5160839160839161, "percentage": 50.58, "elapsed_time": "22:12:24", "remaining_time": "21:41:55", "throughput": "17772.56", "total_tokens": 1420820480}
{"current_steps": 1356, "total_steps": 2679, "loss": 0.6192, "learning_rate": 2.488162845082904e-06, "epoch": 1.5172027972027973, "percentage": 50.62, "elapsed_time": "22:13:12", "remaining_time": "21:40:46", "throughput": "17774.92", "total_tokens": 1421869056}
{"current_steps": 1357, "total_steps": 2679, "loss": 0.7241, "learning_rate": 2.4852035874526632e-06, "epoch": 1.5183216783216782, "percentage": 50.65, "elapsed_time": "22:14:07", "remaining_time": "21:39:43", "throughput": "17775.89", "total_tokens": 1422917632}
{"current_steps": 1358, "total_steps": 2679, "loss": 0.7823, "learning_rate": 2.4822443505550474e-06, "epoch": 1.5194405594405596, "percentage": 50.69, "elapsed_time": "22:15:12", "remaining_time": "21:38:50", "throughput": "17774.48", "total_tokens": 1423966208}
{"current_steps": 1359, "total_steps": 2679, "loss": 0.6195, "learning_rate": 2.479285138536517e-06, "epoch": 1.5205594405594405, "percentage": 50.73, "elapsed_time": "22:16:06", "remaining_time": "21:37:46", "throughput": "17775.61", "total_tokens": 1425014784}
{"current_steps": 1360, "total_steps": 2679, "loss": 0.551, "learning_rate": 2.4763259555434997e-06, "epoch": 1.5216783216783218, "percentage": 50.77, "elapsed_time": "22:16:55", "remaining_time": "21:36:37", "throughput": "17777.95", "total_tokens": 1426063360}
{"current_steps": 1361, "total_steps": 2679, "loss": 0.7314, "learning_rate": 2.473366805722379e-06, "epoch": 1.5227972027972028, "percentage": 50.8, "elapsed_time": "22:17:56", "remaining_time": "21:35:39", "throughput": "17777.54", "total_tokens": 1427111936}
{"current_steps": 1362, "total_steps": 2679, "loss": 0.6199, "learning_rate": 2.470407693219495e-06, "epoch": 1.523916083916084, "percentage": 50.84, "elapsed_time": "22:18:45", "remaining_time": "21:34:31", "throughput": "17779.63", "total_tokens": 1428160512}
{"current_steps": 1363, "total_steps": 2679, "loss": 0.6151, "learning_rate": 2.4674486221811345e-06, "epoch": 1.525034965034965, "percentage": 50.88, "elapsed_time": "22:19:40", "remaining_time": "21:33:28", "throughput": "17780.57", "total_tokens": 1429209088}
{"current_steps": 1364, "total_steps": 2679, "loss": 0.7297, "learning_rate": 2.4644895967535267e-06, "epoch": 1.5261538461538462, "percentage": 50.91, "elapsed_time": "22:20:40", "remaining_time": "21:32:30", "throughput": "17780.37", "total_tokens": 1430257664}
{"current_steps": 1365, "total_steps": 2679, "loss": 0.6494, "learning_rate": 2.4615306210828357e-06, "epoch": 1.5272727272727273, "percentage": 50.95, "elapsed_time": "22:21:35", "remaining_time": "21:31:27", "throughput": "17781.28", "total_tokens": 1431306240}
{"current_steps": 1366, "total_steps": 2679, "loss": 0.6477, "learning_rate": 2.4585716993151555e-06, "epoch": 1.5283916083916083, "percentage": 50.99, "elapsed_time": "22:22:28", "remaining_time": "21:30:23", "throughput": "17782.47", "total_tokens": 1432354816}
{"current_steps": 1367, "total_steps": 2679, "loss": 0.6864, "learning_rate": 2.4556128355965076e-06, "epoch": 1.5295104895104896, "percentage": 51.03, "elapsed_time": "22:23:22", "remaining_time": "21:29:19", "throughput": "17783.59", "total_tokens": 1433403392}
{"current_steps": 1368, "total_steps": 2679, "loss": 0.7068, "learning_rate": 2.4526540340728285e-06, "epoch": 1.5306293706293705, "percentage": 51.06, "elapsed_time": "22:24:22", "remaining_time": "21:28:21", "throughput": "17783.45", "total_tokens": 1434451968}
{"current_steps": 1369, "total_steps": 2679, "loss": 0.7852, "learning_rate": 2.449695298889971e-06, "epoch": 1.531748251748252, "percentage": 51.1, "elapsed_time": "22:25:27", "remaining_time": "21:27:28", "throughput": "17781.98", "total_tokens": 1435500544}
{"current_steps": 1370, "total_steps": 2679, "loss": 0.6291, "learning_rate": 2.4467366341936922e-06, "epoch": 1.5328671328671328, "percentage": 51.14, "elapsed_time": "22:26:21", "remaining_time": "21:26:24", "throughput": "17783.12", "total_tokens": 1436549120}
{"current_steps": 1371, "total_steps": 2679, "loss": 0.5578, "learning_rate": 2.44377804412965e-06, "epoch": 1.533986013986014, "percentage": 51.18, "elapsed_time": "22:27:09", "remaining_time": "21:25:15", "throughput": "17785.55", "total_tokens": 1437597696}
{"current_steps": 1372, "total_steps": 2679, "loss": 0.6644, "learning_rate": 2.440819532843401e-06, "epoch": 1.5351048951048951, "percentage": 51.21, "elapsed_time": "22:28:03", "remaining_time": "21:24:11", "throughput": "17786.68", "total_tokens": 1438646272}
{"current_steps": 1373, "total_steps": 2679, "loss": 0.7209, "learning_rate": 2.4378611044803887e-06, "epoch": 1.5362237762237763, "percentage": 51.25, "elapsed_time": "22:29:08", "remaining_time": "21:23:18", "throughput": "17785.29", "total_tokens": 1439694848}
{"current_steps": 1374, "total_steps": 2679, "loss": 0.6141, "learning_rate": 2.434902763185942e-06, "epoch": 1.5373426573426574, "percentage": 51.29, "elapsed_time": "22:29:57", "remaining_time": "21:22:09", "throughput": "17787.57", "total_tokens": 1440743424}
{"current_steps": 1375, "total_steps": 2679, "loss": 0.6641, "learning_rate": 2.431944513105266e-06, "epoch": 1.5384615384615383, "percentage": 51.33, "elapsed_time": "22:30:51", "remaining_time": "21:21:06", "throughput": "17788.64", "total_tokens": 1441792000}
{"current_steps": 1376, "total_steps": 2679, "loss": 0.6653, "learning_rate": 2.4289863583834406e-06, "epoch": 1.5395804195804197, "percentage": 51.36, "elapsed_time": "22:31:45", "remaining_time": "21:20:02", "throughput": "17789.62", "total_tokens": 1442840576}
{"current_steps": 1377, "total_steps": 2679, "loss": 0.7575, "learning_rate": 2.426028303165409e-06, "epoch": 1.5406993006993006, "percentage": 51.4, "elapsed_time": "22:32:45", "remaining_time": "21:19:04", "throughput": "17789.48", "total_tokens": 1443889152}
{"current_steps": 1378, "total_steps": 2679, "loss": 0.7988, "learning_rate": 2.4230703515959765e-06, "epoch": 1.541818181818182, "percentage": 51.44, "elapsed_time": "22:33:45", "remaining_time": "21:18:06", "throughput": "17789.29", "total_tokens": 1444937728}
{"current_steps": 1379, "total_steps": 2679, "loss": 0.7933, "learning_rate": 2.420112507819804e-06, "epoch": 1.542937062937063, "percentage": 51.47, "elapsed_time": "22:34:50", "remaining_time": "21:17:13", "throughput": "17787.80", "total_tokens": 1445986304}
{"current_steps": 1380, "total_steps": 2679, "loss": 0.6814, "learning_rate": 2.4171547759813998e-06, "epoch": 1.544055944055944, "percentage": 51.51, "elapsed_time": "22:35:50", "remaining_time": "21:16:15", "throughput": "17787.66", "total_tokens": 1447034880}
{"current_steps": 1381, "total_steps": 2679, "loss": 0.6962, "learning_rate": 2.4141971602251176e-06, "epoch": 1.5451748251748252, "percentage": 51.55, "elapsed_time": "22:36:45", "remaining_time": "21:15:12", "throughput": "17788.62", "total_tokens": 1448083456}
{"current_steps": 1382, "total_steps": 2679, "loss": 0.8119, "learning_rate": 2.411239664695145e-06, "epoch": 1.5462937062937063, "percentage": 51.59, "elapsed_time": "22:37:45", "remaining_time": "21:14:15", "throughput": "17788.25", "total_tokens": 1449132032}
{"current_steps": 1383, "total_steps": 2679, "loss": 0.6718, "learning_rate": 2.4082822935355035e-06, "epoch": 1.5474125874125875, "percentage": 51.62, "elapsed_time": "22:38:47", "remaining_time": "21:13:18", "throughput": "17787.67", "total_tokens": 1450180608}
{"current_steps": 1384, "total_steps": 2679, "loss": 0.6898, "learning_rate": 2.4053250508900416e-06, "epoch": 1.5485314685314684, "percentage": 51.66, "elapsed_time": "22:39:47", "remaining_time": "21:12:21", "throughput": "17787.35", "total_tokens": 1451229184}
{"current_steps": 1385, "total_steps": 2679, "loss": 0.7117, "learning_rate": 2.402367940902423e-06, "epoch": 1.5496503496503498, "percentage": 51.7, "elapsed_time": "22:40:51", "remaining_time": "21:11:26", "throughput": "17786.33", "total_tokens": 1452277760}
{"current_steps": 1386, "total_steps": 2679, "loss": 0.7005, "learning_rate": 2.3994109677161316e-06, "epoch": 1.5507692307692307, "percentage": 51.74, "elapsed_time": "22:41:50", "remaining_time": "21:10:28", "throughput": "17786.20", "total_tokens": 1453326336}
{"current_steps": 1387, "total_steps": 2679, "loss": 0.7743, "learning_rate": 2.396454135474454e-06, "epoch": 1.551888111888112, "percentage": 51.77, "elapsed_time": "22:42:56", "remaining_time": "21:09:34", "throughput": "17784.85", "total_tokens": 1454374912}
{"current_steps": 1388, "total_steps": 2679, "loss": 0.6175, "learning_rate": 2.393497448320484e-06, "epoch": 1.553006993006993, "percentage": 51.81, "elapsed_time": "22:43:49", "remaining_time": "21:08:30", "throughput": "17786.02", "total_tokens": 1455423488}
{"current_steps": 1389, "total_steps": 2679, "loss": 0.8076, "learning_rate": 2.3905409103971096e-06, "epoch": 1.554125874125874, "percentage": 51.85, "elapsed_time": "22:44:54", "remaining_time": "21:07:37", "throughput": "17784.75", "total_tokens": 1456472064}
{"current_steps": 1390, "total_steps": 2679, "loss": 0.5888, "learning_rate": 2.3875845258470074e-06, "epoch": 1.5552447552447553, "percentage": 51.89, "elapsed_time": "22:45:42", "remaining_time": "21:06:28", "throughput": "17787.14", "total_tokens": 1457520640}
{"current_steps": 1391, "total_steps": 2679, "loss": 0.6998, "learning_rate": 2.384628298812646e-06, "epoch": 1.5563636363636364, "percentage": 51.92, "elapsed_time": "22:46:36", "remaining_time": "21:05:24", "throughput": "17788.19", "total_tokens": 1458569216}
{"current_steps": 1392, "total_steps": 2679, "loss": 0.6847, "learning_rate": 2.3816722334362656e-06, "epoch": 1.5574825174825175, "percentage": 51.96, "elapsed_time": "22:47:35", "remaining_time": "21:04:26", "throughput": "17788.09", "total_tokens": 1459617792}
{"current_steps": 1393, "total_steps": 2679, "loss": 0.6591, "learning_rate": 2.3787163338598854e-06, "epoch": 1.5586013986013985, "percentage": 52.0, "elapsed_time": "22:48:35", "remaining_time": "21:03:27", "throughput": "17787.99", "total_tokens": 1460666368}
{"current_steps": 1394, "total_steps": 2679, "loss": 0.6874, "learning_rate": 2.37576060422529e-06, "epoch": 1.5597202797202798, "percentage": 52.03, "elapsed_time": "22:49:34", "remaining_time": "21:02:29", "throughput": "17787.89", "total_tokens": 1461714944}
{"current_steps": 1395, "total_steps": 2679, "loss": 0.7028, "learning_rate": 2.3728050486740244e-06, "epoch": 1.5608391608391607, "percentage": 52.07, "elapsed_time": "22:50:34", "remaining_time": "21:01:30", "throughput": "17787.76", "total_tokens": 1462763520}
{"current_steps": 1396, "total_steps": 2679, "loss": 0.7958, "learning_rate": 2.3698496713473937e-06, "epoch": 1.561958041958042, "percentage": 52.11, "elapsed_time": "22:51:34", "remaining_time": "21:00:33", "throughput": "17787.45", "total_tokens": 1463812096}
{"current_steps": 1397, "total_steps": 2679, "loss": 0.7713, "learning_rate": 2.3668944763864486e-06, "epoch": 1.563076923076923, "percentage": 52.15, "elapsed_time": "22:52:40", "remaining_time": "20:59:40", "throughput": "17785.96", "total_tokens": 1464860672}
{"current_steps": 1398, "total_steps": 2679, "loss": 0.8503, "learning_rate": 2.363939467931989e-06, "epoch": 1.5641958041958042, "percentage": 52.18, "elapsed_time": "22:53:46", "remaining_time": "20:58:48", "throughput": "17784.42", "total_tokens": 1465909248}
{"current_steps": 1399, "total_steps": 2679, "loss": 0.7456, "learning_rate": 2.3609846501245494e-06, "epoch": 1.5653146853146853, "percentage": 52.22, "elapsed_time": "22:54:51", "remaining_time": "20:57:54", "throughput": "17783.11", "total_tokens": 1466957824}
{"current_steps": 1400, "total_steps": 2679, "loss": 0.8466, "learning_rate": 2.3580300271044e-06, "epoch": 1.5664335664335665, "percentage": 52.26, "elapsed_time": "22:55:57", "remaining_time": "20:57:01", "throughput": "17781.72", "total_tokens": 1468006400}
{"current_steps": 1401, "total_steps": 2679, "loss": 0.7182, "learning_rate": 2.3550756030115364e-06, "epoch": 1.5675524475524476, "percentage": 52.3, "elapsed_time": "22:56:56", "remaining_time": "20:56:03", "throughput": "17781.59", "total_tokens": 1469054976}
{"current_steps": 1402, "total_steps": 2679, "loss": 0.625, "learning_rate": 2.3521213819856756e-06, "epoch": 1.5686713286713285, "percentage": 52.33, "elapsed_time": "22:57:50", "remaining_time": "20:54:59", "throughput": "17782.75", "total_tokens": 1470103552}
{"current_steps": 1403, "total_steps": 2679, "loss": 0.7735, "learning_rate": 2.349167368166251e-06, "epoch": 1.56979020979021, "percentage": 52.37, "elapsed_time": "22:58:50", "remaining_time": "20:54:01", "throughput": "17782.54", "total_tokens": 1471152128}
{"current_steps": 1404, "total_steps": 2679, "loss": 0.6306, "learning_rate": 2.3462135656924046e-06, "epoch": 1.5709090909090908, "percentage": 52.41, "elapsed_time": "22:59:43", "remaining_time": "20:52:57", "throughput": "17783.68", "total_tokens": 1472200704}
{"current_steps": 1405, "total_steps": 2679, "loss": 0.62, "learning_rate": 2.343259978702984e-06, "epoch": 1.5720279720279722, "percentage": 52.44, "elapsed_time": "23:00:37", "remaining_time": "20:51:53", "throughput": "17784.82", "total_tokens": 1473249280}
{"current_steps": 1406, "total_steps": 2679, "loss": 0.6094, "learning_rate": 2.3403066113365323e-06, "epoch": 1.573146853146853, "percentage": 52.48, "elapsed_time": "23:01:25", "remaining_time": "20:50:45", "throughput": "17787.12", "total_tokens": 1474297856}
{"current_steps": 1407, "total_steps": 2679, "loss": 0.7058, "learning_rate": 2.337353467731286e-06, "epoch": 1.5742657342657342, "percentage": 52.52, "elapsed_time": "23:02:24", "remaining_time": "20:49:46", "throughput": "17787.05", "total_tokens": 1475346432}
{"current_steps": 1408, "total_steps": 2679, "loss": 0.6864, "learning_rate": 2.33440055202517e-06, "epoch": 1.5753846153846154, "percentage": 52.56, "elapsed_time": "23:03:19", "remaining_time": "20:48:43", "throughput": "17788.00", "total_tokens": 1476395008}
{"current_steps": 1409, "total_steps": 2679, "loss": 0.6795, "learning_rate": 2.3314478683557863e-06, "epoch": 1.5765034965034965, "percentage": 52.59, "elapsed_time": "23:04:15", "remaining_time": "20:47:42", "throughput": "17788.56", "total_tokens": 1477443584}
{"current_steps": 1410, "total_steps": 2679, "loss": 0.7887, "learning_rate": 2.3284954208604154e-06, "epoch": 1.5776223776223777, "percentage": 52.63, "elapsed_time": "23:05:21", "remaining_time": "20:46:49", "throughput": "17787.12", "total_tokens": 1478492160}
{"current_steps": 1411, "total_steps": 2679, "loss": 0.6818, "learning_rate": 2.3255432136760026e-06, "epoch": 1.5787412587412586, "percentage": 52.67, "elapsed_time": "23:06:21", "remaining_time": "20:45:51", "throughput": "17786.89", "total_tokens": 1479540736}
{"current_steps": 1412, "total_steps": 2679, "loss": 0.7175, "learning_rate": 2.322591250939162e-06, "epoch": 1.57986013986014, "percentage": 52.71, "elapsed_time": "23:07:21", "remaining_time": "20:44:53", "throughput": "17786.59", "total_tokens": 1480589312}
{"current_steps": 1413, "total_steps": 2679, "loss": 0.6623, "learning_rate": 2.3196395367861605e-06, "epoch": 1.5809790209790209, "percentage": 52.74, "elapsed_time": "23:08:20", "remaining_time": "20:43:54", "throughput": "17786.72", "total_tokens": 1481637888}
{"current_steps": 1414, "total_steps": 2679, "loss": 0.8412, "learning_rate": 2.316688075352918e-06, "epoch": 1.5820979020979022, "percentage": 52.78, "elapsed_time": "23:09:26", "remaining_time": "20:43:01", "throughput": "17785.19", "total_tokens": 1482686464}
{"current_steps": 1415, "total_steps": 2679, "loss": 0.659, "learning_rate": 2.3137368707750018e-06, "epoch": 1.5832167832167832, "percentage": 52.82, "elapsed_time": "23:10:26", "remaining_time": "20:42:03", "throughput": "17784.91", "total_tokens": 1483735040}
{"current_steps": 1416, "total_steps": 2679, "loss": 0.6678, "learning_rate": 2.310785927187616e-06, "epoch": 1.5843356643356643, "percentage": 52.86, "elapsed_time": "23:11:25", "remaining_time": "20:41:04", "throughput": "17784.91", "total_tokens": 1484783616}
{"current_steps": 1417, "total_steps": 2679, "loss": 0.599, "learning_rate": 2.3078352487256045e-06, "epoch": 1.5854545454545454, "percentage": 52.89, "elapsed_time": "23:12:18", "remaining_time": "20:40:00", "throughput": "17786.25", "total_tokens": 1485832192}
{"current_steps": 1418, "total_steps": 2679, "loss": 0.7434, "learning_rate": 2.3048848395234337e-06, "epoch": 1.5865734265734266, "percentage": 52.93, "elapsed_time": "23:13:18", "remaining_time": "20:39:02", "throughput": "17786.01", "total_tokens": 1486880768}
{"current_steps": 1419, "total_steps": 2679, "loss": 0.6261, "learning_rate": 2.301934703715196e-06, "epoch": 1.5876923076923077, "percentage": 52.97, "elapsed_time": "23:14:09", "remaining_time": "20:37:56", "throughput": "17787.63", "total_tokens": 1487929344}
{"current_steps": 1420, "total_steps": 2679, "loss": 0.5925, "learning_rate": 2.2989848454346007e-06, "epoch": 1.5888111888111887, "percentage": 53.0, "elapsed_time": "23:14:57", "remaining_time": "20:36:47", "throughput": "17790.00", "total_tokens": 1488977920}
{"current_steps": 1421, "total_steps": 2679, "loss": 0.6348, "learning_rate": 2.2960352688149657e-06, "epoch": 1.58993006993007, "percentage": 53.04, "elapsed_time": "23:15:48", "remaining_time": "20:35:41", "throughput": "17791.70", "total_tokens": 1490026496}
{"current_steps": 1422, "total_steps": 2679, "loss": 0.6674, "learning_rate": 2.293085977989219e-06, "epoch": 1.591048951048951, "percentage": 53.08, "elapsed_time": "23:16:47", "remaining_time": "20:34:43", "throughput": "17791.67", "total_tokens": 1491075072}
{"current_steps": 1423, "total_steps": 2679, "loss": 0.6147, "learning_rate": 2.290136977089883e-06, "epoch": 1.5921678321678323, "percentage": 53.12, "elapsed_time": "23:17:41", "remaining_time": "20:33:39", "throughput": "17792.66", "total_tokens": 1492123648}
{"current_steps": 1424, "total_steps": 2679, "loss": 0.733, "learning_rate": 2.287188270249077e-06, "epoch": 1.5932867132867132, "percentage": 53.15, "elapsed_time": "23:18:35", "remaining_time": "20:32:36", "throughput": "17793.67", "total_tokens": 1493172224}
{"current_steps": 1425, "total_steps": 2679, "loss": 0.7475, "learning_rate": 2.2842398615985086e-06, "epoch": 1.5944055944055944, "percentage": 53.19, "elapsed_time": "23:19:38", "remaining_time": "20:31:41", "throughput": "17792.89", "total_tokens": 1494220800}
{"current_steps": 1426, "total_steps": 2679, "loss": 0.5872, "learning_rate": 2.281291755269464e-06, "epoch": 1.5955244755244755, "percentage": 53.23, "elapsed_time": "23:20:26", "remaining_time": "20:30:32", "throughput": "17795.18", "total_tokens": 1495269376}
{"current_steps": 1427, "total_steps": 2679, "loss": 0.7137, "learning_rate": 2.27834395539281e-06, "epoch": 1.5966433566433567, "percentage": 53.27, "elapsed_time": "23:21:26", "remaining_time": "20:29:34", "throughput": "17794.99", "total_tokens": 1496317952}
{"current_steps": 1428, "total_steps": 2679, "loss": 0.6452, "learning_rate": 2.2753964660989813e-06, "epoch": 1.5977622377622378, "percentage": 53.3, "elapsed_time": "23:22:26", "remaining_time": "20:28:36", "throughput": "17794.85", "total_tokens": 1497366528}
{"current_steps": 1429, "total_steps": 2679, "loss": 0.6903, "learning_rate": 2.2724492915179787e-06, "epoch": 1.5988811188811187, "percentage": 53.34, "elapsed_time": "23:23:29", "remaining_time": "20:27:41", "throughput": "17793.87", "total_tokens": 1498415104}
{"current_steps": 1430, "total_steps": 2679, "loss": 0.7619, "learning_rate": 2.269502435779362e-06, "epoch": 1.6, "percentage": 53.38, "elapsed_time": "23:24:32", "remaining_time": "20:26:45", "throughput": "17793.07", "total_tokens": 1499463680}
{"current_steps": 1431, "total_steps": 2679, "loss": 0.745, "learning_rate": 2.2665559030122424e-06, "epoch": 1.601118881118881, "percentage": 53.42, "elapsed_time": "23:25:37", "remaining_time": "20:25:51", "throughput": "17791.86", "total_tokens": 1500512256}
{"current_steps": 1432, "total_steps": 2679, "loss": 0.5718, "learning_rate": 2.2636096973452813e-06, "epoch": 1.6022377622377624, "percentage": 53.45, "elapsed_time": "23:26:24", "remaining_time": "20:24:43", "throughput": "17794.20", "total_tokens": 1501560832}
{"current_steps": 1433, "total_steps": 2679, "loss": 0.7402, "learning_rate": 2.2606638229066802e-06, "epoch": 1.6033566433566433, "percentage": 53.49, "elapsed_time": "23:27:23", "remaining_time": "20:23:43", "throughput": "17794.31", "total_tokens": 1502609408}
{"current_steps": 1434, "total_steps": 2679, "loss": 0.6025, "learning_rate": 2.257718283824177e-06, "epoch": 1.6044755244755244, "percentage": 53.53, "elapsed_time": "23:28:17", "remaining_time": "20:22:40", "throughput": "17795.38", "total_tokens": 1503657984}
{"current_steps": 1435, "total_steps": 2679, "loss": 0.7073, "learning_rate": 2.254773084225039e-06, "epoch": 1.6055944055944056, "percentage": 53.56, "elapsed_time": "23:29:16", "remaining_time": "20:21:42", "throughput": "17795.23", "total_tokens": 1504706560}
{"current_steps": 1436, "total_steps": 2679, "loss": 0.6828, "learning_rate": 2.2518282282360597e-06, "epoch": 1.6067132867132867, "percentage": 53.6, "elapsed_time": "23:30:10", "remaining_time": "20:20:38", "throughput": "17796.30", "total_tokens": 1505755136}
{"current_steps": 1437, "total_steps": 2679, "loss": 0.8252, "learning_rate": 2.2488837199835477e-06, "epoch": 1.6078321678321679, "percentage": 53.64, "elapsed_time": "23:31:21", "remaining_time": "20:19:50", "throughput": "17793.79", "total_tokens": 1506803712}
{"current_steps": 1438, "total_steps": 2679, "loss": 0.6887, "learning_rate": 2.2459395635933267e-06, "epoch": 1.6089510489510488, "percentage": 53.68, "elapsed_time": "23:32:15", "remaining_time": "20:18:47", "throughput": "17794.79", "total_tokens": 1507852288}
{"current_steps": 1439, "total_steps": 2679, "loss": 0.6742, "learning_rate": 2.2429957631907285e-06, "epoch": 1.6100699300699302, "percentage": 53.71, "elapsed_time": "23:33:13", "remaining_time": "20:17:47", "throughput": "17794.95", "total_tokens": 1508900864}
{"current_steps": 1440, "total_steps": 2679, "loss": 0.6127, "learning_rate": 2.240052322900583e-06, "epoch": 1.611188811188811, "percentage": 53.75, "elapsed_time": "23:34:01", "remaining_time": "20:16:39", "throughput": "17797.24", "total_tokens": 1509949440}
{"current_steps": 1441, "total_steps": 2679, "loss": 0.6728, "learning_rate": 2.2371092468472193e-06, "epoch": 1.6123076923076924, "percentage": 53.79, "elapsed_time": "23:34:56", "remaining_time": "20:15:36", "throughput": "17798.17", "total_tokens": 1510998016}
{"current_steps": 1442, "total_steps": 2679, "loss": 0.7416, "learning_rate": 2.2341665391544522e-06, "epoch": 1.6134265734265734, "percentage": 53.83, "elapsed_time": "23:35:50", "remaining_time": "20:14:33", "throughput": "17799.11", "total_tokens": 1512046592}
{"current_steps": 1443, "total_steps": 2679, "loss": 0.6883, "learning_rate": 2.2312242039455816e-06, "epoch": 1.6145454545454545, "percentage": 53.86, "elapsed_time": "23:36:44", "remaining_time": "20:13:30", "throughput": "17800.09", "total_tokens": 1513095168}
{"current_steps": 1444, "total_steps": 2679, "loss": 0.6225, "learning_rate": 2.2282822453433878e-06, "epoch": 1.6156643356643356, "percentage": 53.9, "elapsed_time": "23:37:38", "remaining_time": "20:12:27", "throughput": "17801.17", "total_tokens": 1514143744}
{"current_steps": 1445, "total_steps": 2679, "loss": 0.6233, "learning_rate": 2.2253406674701206e-06, "epoch": 1.6167832167832168, "percentage": 53.94, "elapsed_time": "23:38:32", "remaining_time": "20:11:24", "throughput": "17802.18", "total_tokens": 1515192320}
{"current_steps": 1446, "total_steps": 2679, "loss": 0.7264, "learning_rate": 2.2223994744474986e-06, "epoch": 1.617902097902098, "percentage": 53.98, "elapsed_time": "23:39:32", "remaining_time": "20:10:26", "throughput": "17802.04", "total_tokens": 1516240896}
{"current_steps": 1447, "total_steps": 2679, "loss": 0.7442, "learning_rate": 2.2194586703966976e-06, "epoch": 1.6190209790209789, "percentage": 54.01, "elapsed_time": "23:40:32", "remaining_time": "20:09:28", "throughput": "17801.85", "total_tokens": 1517289472}
{"current_steps": 1448, "total_steps": 2679, "loss": 0.6927, "learning_rate": 2.2165182594383532e-06, "epoch": 1.6201398601398602, "percentage": 54.05, "elapsed_time": "23:41:31", "remaining_time": "20:08:29", "throughput": "17801.73", "total_tokens": 1518338048}
{"current_steps": 1449, "total_steps": 2679, "loss": 0.8128, "learning_rate": 2.213578245692546e-06, "epoch": 1.6212587412587411, "percentage": 54.09, "elapsed_time": "23:42:37", "remaining_time": "20:07:36", "throughput": "17800.29", "total_tokens": 1519386624}
{"current_steps": 1450, "total_steps": 2679, "loss": 0.7315, "learning_rate": 2.210638633278802e-06, "epoch": 1.6223776223776225, "percentage": 54.12, "elapsed_time": "23:43:37", "remaining_time": "20:06:38", "throughput": "17800.07", "total_tokens": 1520435200}
{"current_steps": 1451, "total_steps": 2679, "loss": 0.7408, "learning_rate": 2.2076994263160863e-06, "epoch": 1.6234965034965034, "percentage": 54.16, "elapsed_time": "23:44:37", "remaining_time": "20:05:40", "throughput": "17799.83", "total_tokens": 1521483776}
{"current_steps": 1452, "total_steps": 2679, "loss": 0.6227, "learning_rate": 2.204760628922791e-06, "epoch": 1.6246153846153846, "percentage": 54.2, "elapsed_time": "23:45:32", "remaining_time": "20:04:38", "throughput": "17800.70", "total_tokens": 1522532352}
{"current_steps": 1453, "total_steps": 2679, "loss": 0.7336, "learning_rate": 2.20182224521674e-06, "epoch": 1.6257342657342657, "percentage": 54.24, "elapsed_time": "23:46:37", "remaining_time": "20:03:44", "throughput": "17799.36", "total_tokens": 1523580928}
{"current_steps": 1454, "total_steps": 2679, "loss": 0.6759, "learning_rate": 2.1988842793151743e-06, "epoch": 1.6268531468531469, "percentage": 54.27, "elapsed_time": "23:47:36", "remaining_time": "20:02:46", "throughput": "17799.30", "total_tokens": 1524629504}
{"current_steps": 1455, "total_steps": 2679, "loss": 0.6754, "learning_rate": 2.1959467353347494e-06, "epoch": 1.627972027972028, "percentage": 54.31, "elapsed_time": "23:48:30", "remaining_time": "20:01:43", "throughput": "17800.33", "total_tokens": 1525678080}
{"current_steps": 1456, "total_steps": 2679, "loss": 0.6758, "learning_rate": 2.193009617391532e-06, "epoch": 1.6290909090909091, "percentage": 54.35, "elapsed_time": "23:49:25", "remaining_time": "20:00:40", "throughput": "17801.24", "total_tokens": 1526726656}
{"current_steps": 1457, "total_steps": 2679, "loss": 0.6846, "learning_rate": 2.190072929600989e-06, "epoch": 1.6302097902097903, "percentage": 54.39, "elapsed_time": "23:50:19", "remaining_time": "19:59:37", "throughput": "17802.14", "total_tokens": 1527775232}
{"current_steps": 1458, "total_steps": 2679, "loss": 0.7539, "learning_rate": 2.1871366760779857e-06, "epoch": 1.6313286713286712, "percentage": 54.42, "elapsed_time": "23:51:24", "remaining_time": "19:58:44", "throughput": "17800.83", "total_tokens": 1528823808}
{"current_steps": 1459, "total_steps": 2679, "loss": 0.6996, "learning_rate": 2.1842008609367794e-06, "epoch": 1.6324475524475526, "percentage": 54.46, "elapsed_time": "23:52:24", "remaining_time": "19:57:46", "throughput": "17800.63", "total_tokens": 1529872384}
{"current_steps": 1460, "total_steps": 2679, "loss": 0.6688, "learning_rate": 2.1812654882910134e-06, "epoch": 1.6335664335664335, "percentage": 54.5, "elapsed_time": "23:53:18", "remaining_time": "19:56:43", "throughput": "17801.68", "total_tokens": 1530920960}
{"current_steps": 1461, "total_steps": 2679, "loss": 0.642, "learning_rate": 2.1783305622537106e-06, "epoch": 1.6346853146853146, "percentage": 54.54, "elapsed_time": "23:54:12", "remaining_time": "19:55:39", "throughput": "17802.70", "total_tokens": 1531969536}
{"current_steps": 1462, "total_steps": 2679, "loss": 0.7306, "learning_rate": 2.175396086937265e-06, "epoch": 1.6358041958041958, "percentage": 54.57, "elapsed_time": "23:55:06", "remaining_time": "19:54:36", "throughput": "17803.72", "total_tokens": 1533018112}
{"current_steps": 1463, "total_steps": 2679, "loss": 0.7331, "learning_rate": 2.1724620664534453e-06, "epoch": 1.636923076923077, "percentage": 54.61, "elapsed_time": "23:56:06", "remaining_time": "19:53:38", "throughput": "17803.50", "total_tokens": 1534066688}
{"current_steps": 1464, "total_steps": 2679, "loss": 0.7023, "learning_rate": 2.169528504913378e-06, "epoch": 1.638041958041958, "percentage": 54.65, "elapsed_time": "23:57:06", "remaining_time": "19:52:41", "throughput": "17803.26", "total_tokens": 1535115264}
{"current_steps": 1465, "total_steps": 2679, "loss": 0.6027, "learning_rate": 2.166595406427548e-06, "epoch": 1.6391608391608392, "percentage": 54.68, "elapsed_time": "23:57:54", "remaining_time": "19:51:33", "throughput": "17805.45", "total_tokens": 1536163840}
{"current_steps": 1466, "total_steps": 2679, "loss": 0.8737, "learning_rate": 2.163662775105792e-06, "epoch": 1.6402797202797204, "percentage": 54.72, "elapsed_time": "23:59:06", "remaining_time": "19:50:44", "throughput": "17802.90", "total_tokens": 1537212416}
{"current_steps": 1467, "total_steps": 2679, "loss": 0.69, "learning_rate": 2.1607306150572905e-06, "epoch": 1.6413986013986013, "percentage": 54.76, "elapsed_time": "1 day, 0:00:00", "remaining_time": "19:49:41", "throughput": "17803.90", "total_tokens": 1538260992}
{"current_steps": 1468, "total_steps": 2679, "loss": 0.6028, "learning_rate": 2.1577989303905657e-06, "epoch": 1.6425174825174826, "percentage": 54.8, "elapsed_time": "1 day, 0:00:48", "remaining_time": "19:48:34", "throughput": "17806.03", "total_tokens": 1539309568}
{"current_steps": 1469, "total_steps": 2679, "loss": 0.6556, "learning_rate": 2.154867725213472e-06, "epoch": 1.6436363636363636, "percentage": 54.83, "elapsed_time": "1 day, 0:01:48", "remaining_time": "19:47:36", "throughput": "17805.82", "total_tokens": 1540358144}
{"current_steps": 1470, "total_steps": 2679, "loss": 0.7768, "learning_rate": 2.151937003633193e-06, "epoch": 1.6447552447552447, "percentage": 54.87, "elapsed_time": "1 day, 0:02:51", "remaining_time": "19:46:40", "throughput": "17805.01", "total_tokens": 1541406720}
{"current_steps": 1471, "total_steps": 2679, "loss": 0.774, "learning_rate": 2.149006769756234e-06, "epoch": 1.6458741258741258, "percentage": 54.91, "elapsed_time": "1 day, 0:04:01", "remaining_time": "19:45:50", "throughput": "17802.73", "total_tokens": 1542455296}
{"current_steps": 1472, "total_steps": 2679, "loss": 0.7, "learning_rate": 2.146077027688418e-06, "epoch": 1.646993006993007, "percentage": 54.95, "elapsed_time": "1 day, 0:05:01", "remaining_time": "19:44:52", "throughput": "17802.57", "total_tokens": 1543503872}
{"current_steps": 1473, "total_steps": 2679, "loss": 0.6505, "learning_rate": 2.1431477815348775e-06, "epoch": 1.6481118881118881, "percentage": 54.98, "elapsed_time": "1 day, 0:05:54", "remaining_time": "19:43:49", "throughput": "17803.64", "total_tokens": 1544552448}
{"current_steps": 1474, "total_steps": 2679, "loss": 0.7912, "learning_rate": 2.1402190354000502e-06, "epoch": 1.6492307692307693, "percentage": 55.02, "elapsed_time": "1 day, 0:07:05", "remaining_time": "19:43:00", "throughput": "17801.19", "total_tokens": 1545601024}
{"current_steps": 1475, "total_steps": 2679, "loss": 0.6772, "learning_rate": 2.1372907933876745e-06, "epoch": 1.6503496503496504, "percentage": 55.06, "elapsed_time": "1 day, 0:07:59", "remaining_time": "19:41:57", "throughput": "17802.20", "total_tokens": 1546649600}
{"current_steps": 1476, "total_steps": 2679, "loss": 0.681, "learning_rate": 2.134363059600781e-06, "epoch": 1.6514685314685313, "percentage": 55.1, "elapsed_time": "1 day, 0:08:59", "remaining_time": "19:40:59", "throughput": "17802.05", "total_tokens": 1547698176}
{"current_steps": 1477, "total_steps": 2679, "loss": 0.6805, "learning_rate": 2.1314358381416906e-06, "epoch": 1.6525874125874127, "percentage": 55.13, "elapsed_time": "1 day, 0:09:56", "remaining_time": "19:39:59", "throughput": "17802.32", "total_tokens": 1548746752}
{"current_steps": 1478, "total_steps": 2679, "loss": 0.6777, "learning_rate": 2.1285091331120028e-06, "epoch": 1.6537062937062936, "percentage": 55.17, "elapsed_time": "1 day, 0:10:54", "remaining_time": "19:38:58", "throughput": "17802.68", "total_tokens": 1549795328}
{"current_steps": 1479, "total_steps": 2679, "loss": 0.7613, "learning_rate": 2.125582948612595e-06, "epoch": 1.6548251748251748, "percentage": 55.21, "elapsed_time": "1 day, 0:11:54", "remaining_time": "19:38:01", "throughput": "17802.28", "total_tokens": 1550843904}
{"current_steps": 1480, "total_steps": 2679, "loss": 0.8748, "learning_rate": 2.1226572887436175e-06, "epoch": 1.655944055944056, "percentage": 55.24, "elapsed_time": "1 day, 0:13:01", "remaining_time": "19:37:08", "throughput": "17800.78", "total_tokens": 1551892480}
{"current_steps": 1481, "total_steps": 2679, "loss": 0.8277, "learning_rate": 2.1197321576044803e-06, "epoch": 1.657062937062937, "percentage": 55.28, "elapsed_time": "1 day, 0:14:12", "remaining_time": "19:36:19", "throughput": "17798.26", "total_tokens": 1552941056}
{"current_steps": 1482, "total_steps": 2679, "loss": 0.6905, "learning_rate": 2.1168075592938592e-06, "epoch": 1.6581818181818182, "percentage": 55.32, "elapsed_time": "1 day, 0:15:08", "remaining_time": "19:35:18", "throughput": "17798.79", "total_tokens": 1553989632}
{"current_steps": 1483, "total_steps": 2679, "loss": 0.6963, "learning_rate": 2.1138834979096778e-06, "epoch": 1.6593006993006993, "percentage": 55.36, "elapsed_time": "1 day, 0:16:03", "remaining_time": "19:34:15", "throughput": "17799.73", "total_tokens": 1555038208}
{"current_steps": 1484, "total_steps": 2679, "loss": 0.6557, "learning_rate": 2.1109599775491096e-06, "epoch": 1.6604195804195805, "percentage": 55.39, "elapsed_time": "1 day, 0:17:01", "remaining_time": "19:33:16", "throughput": "17799.89", "total_tokens": 1556086784}
{"current_steps": 1485, "total_steps": 2679, "loss": 0.683, "learning_rate": 2.1080370023085713e-06, "epoch": 1.6615384615384614, "percentage": 55.43, "elapsed_time": "1 day, 0:17:56", "remaining_time": "19:32:14", "throughput": "17800.56", "total_tokens": 1557135360}
{"current_steps": 1486, "total_steps": 2679, "loss": 0.6941, "learning_rate": 2.1051145762837115e-06, "epoch": 1.6626573426573428, "percentage": 55.47, "elapsed_time": "1 day, 0:18:56", "remaining_time": "19:31:16", "throughput": "17800.45", "total_tokens": 1558183936}
{"current_steps": 1487, "total_steps": 2679, "loss": 0.7029, "learning_rate": 2.102192703569416e-06, "epoch": 1.6637762237762237, "percentage": 55.51, "elapsed_time": "1 day, 0:19:55", "remaining_time": "19:30:17", "throughput": "17800.33", "total_tokens": 1559232512}
{"current_steps": 1488, "total_steps": 2679, "loss": 0.5943, "learning_rate": 2.0992713882597883e-06, "epoch": 1.664895104895105, "percentage": 55.54, "elapsed_time": "1 day, 0:20:44", "remaining_time": "19:29:11", "throughput": "17802.31", "total_tokens": 1560281088}
{"current_steps": 1489, "total_steps": 2679, "loss": 0.7091, "learning_rate": 2.0963506344481556e-06, "epoch": 1.666013986013986, "percentage": 55.58, "elapsed_time": "1 day, 0:21:44", "remaining_time": "19:28:12", "throughput": "17802.16", "total_tokens": 1561329664}
{"current_steps": 1490, "total_steps": 2679, "loss": 0.6906, "learning_rate": 2.0934304462270568e-06, "epoch": 1.6671328671328671, "percentage": 55.62, "elapsed_time": "1 day, 0:22:43", "remaining_time": "19:27:14", "throughput": "17802.06", "total_tokens": 1562378240}
{"current_steps": 1491, "total_steps": 2679, "loss": 0.8045, "learning_rate": 2.0905108276882356e-06, "epoch": 1.6682517482517483, "percentage": 55.66, "elapsed_time": "1 day, 0:23:50", "remaining_time": "19:26:21", "throughput": "17800.58", "total_tokens": 1563426816}
{"current_steps": 1492, "total_steps": 2679, "loss": 0.6821, "learning_rate": 2.087591782922645e-06, "epoch": 1.6693706293706294, "percentage": 55.69, "elapsed_time": "1 day, 0:24:49", "remaining_time": "19:25:23", "throughput": "17800.40", "total_tokens": 1564475392}
{"current_steps": 1493, "total_steps": 2679, "loss": 0.7651, "learning_rate": 2.0846733160204244e-06, "epoch": 1.6704895104895106, "percentage": 55.73, "elapsed_time": "1 day, 0:25:50", "remaining_time": "19:24:25", "throughput": "17800.08", "total_tokens": 1565523968}
{"current_steps": 1494, "total_steps": 2679, "loss": 0.6843, "learning_rate": 2.081755431070911e-06, "epoch": 1.6716083916083915, "percentage": 55.77, "elapsed_time": "1 day, 0:26:50", "remaining_time": "19:23:27", "throughput": "17799.92", "total_tokens": 1566572544}
{"current_steps": 1495, "total_steps": 2679, "loss": 0.7936, "learning_rate": 2.0788381321626237e-06, "epoch": 1.6727272727272728, "percentage": 55.8, "elapsed_time": "1 day, 0:27:54", "remaining_time": "19:22:32", "throughput": "17798.84", "total_tokens": 1567621120}
{"current_steps": 1496, "total_steps": 2679, "loss": 0.7214, "learning_rate": 2.075921423383258e-06, "epoch": 1.6738461538461538, "percentage": 55.84, "elapsed_time": "1 day, 0:28:59", "remaining_time": "19:21:38", "throughput": "17797.57", "total_tokens": 1568669696}
{"current_steps": 1497, "total_steps": 2679, "loss": 0.6312, "learning_rate": 2.0730053088196883e-06, "epoch": 1.6749650349650351, "percentage": 55.88, "elapsed_time": "1 day, 0:29:47", "remaining_time": "19:20:30", "throughput": "17799.79", "total_tokens": 1569718272}
{"current_steps": 1498, "total_steps": 2679, "loss": 0.8513, "learning_rate": 2.07008979255795e-06, "epoch": 1.676083916083916, "percentage": 55.92, "elapsed_time": "1 day, 0:30:59", "remaining_time": "19:19:42", "throughput": "17797.23", "total_tokens": 1570766848}
{"current_steps": 1499, "total_steps": 2679, "loss": 0.8586, "learning_rate": 2.0671748786832447e-06, "epoch": 1.6772027972027972, "percentage": 55.95, "elapsed_time": "1 day, 0:32:05", "remaining_time": "19:18:48", "throughput": "17795.80", "total_tokens": 1571815424}
{"current_steps": 1500, "total_steps": 2679, "loss": 0.6795, "learning_rate": 2.064260571279928e-06, "epoch": 1.6783216783216783, "percentage": 55.99, "elapsed_time": "1 day, 0:32:59", "remaining_time": "19:17:46", "throughput": "17796.74", "total_tokens": 1572864000}
{"current_steps": 1500, "total_steps": 2679, "eval_loss": 0.7240810394287109, "epoch": 1.6783216783216783, "percentage": 55.99, "elapsed_time": "1 day, 0:37:06", "remaining_time": "19:21:00", "throughput": "17747.17", "total_tokens": 1572864000}
{"current_steps": 1501, "total_steps": 2679, "loss": 0.7162, "learning_rate": 2.061346874431507e-06, "epoch": 1.6794405594405595, "percentage": 56.03, "elapsed_time": "1 day, 0:38:33", "remaining_time": "19:20:23", "throughput": "17741.52", "total_tokens": 1573912576}
{"current_steps": 1502, "total_steps": 2679, "loss": 0.6341, "learning_rate": 2.0584337922206303e-06, "epoch": 1.6805594405594406, "percentage": 56.07, "elapsed_time": "1 day, 0:39:27", "remaining_time": "19:19:20", "throughput": "17742.50", "total_tokens": 1574961152}
{"current_steps": 1503, "total_steps": 2679, "loss": 0.6942, "learning_rate": 2.0555213287290886e-06, "epoch": 1.6816783216783215, "percentage": 56.1, "elapsed_time": "1 day, 0:40:26", "remaining_time": "19:18:21", "throughput": "17742.54", "total_tokens": 1576009728}
{"current_steps": 1504, "total_steps": 2679, "loss": 0.7038, "learning_rate": 2.052609488037805e-06, "epoch": 1.682797202797203, "percentage": 56.14, "elapsed_time": "1 day, 0:41:26", "remaining_time": "19:17:22", "throughput": "17742.39", "total_tokens": 1577058304}
{"current_steps": 1505, "total_steps": 2679, "loss": 0.7551, "learning_rate": 2.0496982742268273e-06, "epoch": 1.6839160839160838, "percentage": 56.18, "elapsed_time": "1 day, 0:42:27", "remaining_time": "19:16:24", "throughput": "17742.08", "total_tokens": 1578106880}
{"current_steps": 1506, "total_steps": 2679, "loss": 0.6798, "learning_rate": 2.04678769137533e-06, "epoch": 1.6850349650349652, "percentage": 56.22, "elapsed_time": "1 day, 0:43:26", "remaining_time": "19:15:26", "throughput": "17741.94", "total_tokens": 1579155456}
{"current_steps": 1507, "total_steps": 2679, "loss": 0.7259, "learning_rate": 2.043877743561598e-06, "epoch": 1.6861538461538461, "percentage": 56.25, "elapsed_time": "1 day, 0:44:31", "remaining_time": "19:14:31", "throughput": "17740.81", "total_tokens": 1580204032}
{"current_steps": 1508, "total_steps": 2679, "loss": 0.6198, "learning_rate": 2.0409684348630292e-06, "epoch": 1.6872727272727273, "percentage": 56.29, "elapsed_time": "1 day, 0:45:25", "remaining_time": "19:13:28", "throughput": "17741.87", "total_tokens": 1581252608}
{"current_steps": 1509, "total_steps": 2679, "loss": 0.6017, "learning_rate": 2.038059769356127e-06, "epoch": 1.6883916083916084, "percentage": 56.33, "elapsed_time": "1 day, 0:46:13", "remaining_time": "19:12:20", "throughput": "17744.08", "total_tokens": 1582301184}
{"current_steps": 1510, "total_steps": 2679, "loss": 0.732, "learning_rate": 2.0351517511164903e-06, "epoch": 1.6895104895104895, "percentage": 56.36, "elapsed_time": "1 day, 0:47:13", "remaining_time": "19:11:21", "throughput": "17743.98", "total_tokens": 1583349760}
{"current_steps": 1511, "total_steps": 2679, "loss": 0.6047, "learning_rate": 2.032244384218815e-06, "epoch": 1.6906293706293707, "percentage": 56.4, "elapsed_time": "1 day, 0:48:06", "remaining_time": "19:10:18", "throughput": "17745.11", "total_tokens": 1584398336}
{"current_steps": 1512, "total_steps": 2679, "loss": 0.6204, "learning_rate": 2.0293376727368798e-06, "epoch": 1.6917482517482516, "percentage": 56.44, "elapsed_time": "1 day, 0:49:00", "remaining_time": "19:09:15", "throughput": "17746.17", "total_tokens": 1585446912}
{"current_steps": 1513, "total_steps": 2679, "loss": 0.5819, "learning_rate": 2.02643162074355e-06, "epoch": 1.692867132867133, "percentage": 56.48, "elapsed_time": "1 day, 0:49:48", "remaining_time": "19:08:07", "throughput": "17748.29", "total_tokens": 1586495488}
{"current_steps": 1514, "total_steps": 2679, "loss": 0.6841, "learning_rate": 2.0235262323107633e-06, "epoch": 1.693986013986014, "percentage": 56.51, "elapsed_time": "1 day, 0:50:48", "remaining_time": "19:07:08", "throughput": "17748.22", "total_tokens": 1587544064}
{"current_steps": 1515, "total_steps": 2679, "loss": 0.6528, "learning_rate": 2.020621511509528e-06, "epoch": 1.6951048951048953, "percentage": 56.55, "elapsed_time": "1 day, 0:51:41", "remaining_time": "19:06:05", "throughput": "17749.26", "total_tokens": 1588592640}
{"current_steps": 1516, "total_steps": 2679, "loss": 0.6367, "learning_rate": 2.0177174624099193e-06, "epoch": 1.6962237762237762, "percentage": 56.59, "elapsed_time": "1 day, 0:52:36", "remaining_time": "19:05:03", "throughput": "17750.19", "total_tokens": 1589641216}
{"current_steps": 1517, "total_steps": 2679, "loss": 0.763, "learning_rate": 2.014814089081067e-06, "epoch": 1.6973426573426573, "percentage": 56.63, "elapsed_time": "1 day, 0:53:41", "remaining_time": "19:04:08", "throughput": "17748.94", "total_tokens": 1590689792}
{"current_steps": 1518, "total_steps": 2679, "loss": 0.8083, "learning_rate": 2.0119113955911596e-06, "epoch": 1.6984615384615385, "percentage": 56.66, "elapsed_time": "1 day, 0:54:42", "remaining_time": "19:03:11", "throughput": "17748.63", "total_tokens": 1591738368}
{"current_steps": 1519, "total_steps": 2679, "loss": 0.6938, "learning_rate": 2.0090093860074273e-06, "epoch": 1.6995804195804196, "percentage": 56.7, "elapsed_time": "1 day, 0:55:41", "remaining_time": "19:02:11", "throughput": "17748.65", "total_tokens": 1592786944}
{"current_steps": 1520, "total_steps": 2679, "loss": 0.862, "learning_rate": 2.006108064396146e-06, "epoch": 1.7006993006993008, "percentage": 56.74, "elapsed_time": "1 day, 0:56:51", "remaining_time": "19:01:21", "throughput": "17746.48", "total_tokens": 1593835520}
{"current_steps": 1521, "total_steps": 2679, "loss": 0.7229, "learning_rate": 2.0032074348226268e-06, "epoch": 1.7018181818181817, "percentage": 56.77, "elapsed_time": "1 day, 0:57:51", "remaining_time": "19:00:22", "throughput": "17746.36", "total_tokens": 1594884096}
{"current_steps": 1522, "total_steps": 2679, "loss": 0.7758, "learning_rate": 2.000307501351209e-06, "epoch": 1.702937062937063, "percentage": 56.81, "elapsed_time": "1 day, 0:58:51", "remaining_time": "18:59:24", "throughput": "17746.18", "total_tokens": 1595932672}
{"current_steps": 1523, "total_steps": 2679, "loss": 0.6983, "learning_rate": 1.997408268045259e-06, "epoch": 1.704055944055944, "percentage": 56.85, "elapsed_time": "1 day, 0:59:51", "remaining_time": "18:58:25", "throughput": "17746.01", "total_tokens": 1596981248}
{"current_steps": 1524, "total_steps": 2679, "loss": 0.8414, "learning_rate": 1.994509738967161e-06, "epoch": 1.7051748251748253, "percentage": 56.89, "elapsed_time": "1 day, 1:01:01", "remaining_time": "18:57:35", "throughput": "17743.68", "total_tokens": 1598029824}
{"current_steps": 1525, "total_steps": 2679, "loss": 0.6005, "learning_rate": 1.9916119181783135e-06, "epoch": 1.7062937062937062, "percentage": 56.92, "elapsed_time": "1 day, 1:01:50", "remaining_time": "18:56:28", "throughput": "17745.85", "total_tokens": 1599078400}
{"current_steps": 1526, "total_steps": 2679, "loss": 0.6042, "learning_rate": 1.9887148097391217e-06, "epoch": 1.7074125874125874, "percentage": 56.96, "elapsed_time": "1 day, 1:02:43", "remaining_time": "18:55:24", "throughput": "17746.94", "total_tokens": 1600126976}
{"current_steps": 1527, "total_steps": 2679, "loss": 0.6072, "learning_rate": 1.9858184177089915e-06, "epoch": 1.7085314685314685, "percentage": 57.0, "elapsed_time": "1 day, 1:03:37", "remaining_time": "18:54:21", "throughput": "17748.01", "total_tokens": 1601175552}
{"current_steps": 1528, "total_steps": 2679, "loss": 0.5928, "learning_rate": 1.982922746146327e-06, "epoch": 1.7096503496503497, "percentage": 57.04, "elapsed_time": "1 day, 1:04:25", "remaining_time": "18:53:14", "throughput": "17750.19", "total_tokens": 1602224128}
{"current_steps": 1529, "total_steps": 2679, "loss": 0.6829, "learning_rate": 1.9800277991085217e-06, "epoch": 1.7107692307692308, "percentage": 57.07, "elapsed_time": "1 day, 1:05:24", "remaining_time": "18:52:15", "throughput": "17750.10", "total_tokens": 1603272704}
{"current_steps": 1530, "total_steps": 2679, "loss": 0.784, "learning_rate": 1.9771335806519544e-06, "epoch": 1.7118881118881117, "percentage": 57.11, "elapsed_time": "1 day, 1:06:24", "remaining_time": "18:51:17", "throughput": "17749.86", "total_tokens": 1604321280}
{"current_steps": 1531, "total_steps": 2679, "loss": 0.6659, "learning_rate": 1.9742400948319838e-06, "epoch": 1.713006993006993, "percentage": 57.15, "elapsed_time": "1 day, 1:07:19", "remaining_time": "18:50:14", "throughput": "17750.85", "total_tokens": 1605369856}
{"current_steps": 1532, "total_steps": 2679, "loss": 0.7278, "learning_rate": 1.9713473457029384e-06, "epoch": 1.714125874125874, "percentage": 57.19, "elapsed_time": "1 day, 1:08:19", "remaining_time": "18:49:16", "throughput": "17750.62", "total_tokens": 1606418432}
{"current_steps": 1533, "total_steps": 2679, "loss": 0.7213, "learning_rate": 1.9684553373181197e-06, "epoch": 1.7152447552447554, "percentage": 57.22, "elapsed_time": "1 day, 1:09:18", "remaining_time": "18:48:17", "throughput": "17750.54", "total_tokens": 1607467008}
{"current_steps": 1534, "total_steps": 2679, "loss": 0.6836, "learning_rate": 1.965564073729787e-06, "epoch": 1.7163636363636363, "percentage": 57.26, "elapsed_time": "1 day, 1:10:17", "remaining_time": "18:47:17", "throughput": "17750.67", "total_tokens": 1608515584}
{"current_steps": 1535, "total_steps": 2679, "loss": 0.6943, "learning_rate": 1.962673558989158e-06, "epoch": 1.7174825174825175, "percentage": 57.3, "elapsed_time": "1 day, 1:11:16", "remaining_time": "18:46:18", "throughput": "17750.66", "total_tokens": 1609564160}
{"current_steps": 1536, "total_steps": 2679, "loss": 0.6404, "learning_rate": 1.959783797146402e-06, "epoch": 1.7186013986013986, "percentage": 57.33, "elapsed_time": "1 day, 1:12:10", "remaining_time": "18:45:16", "throughput": "17751.58", "total_tokens": 1610612736}
{"current_steps": 1537, "total_steps": 2679, "loss": 0.6365, "learning_rate": 1.956894792250631e-06, "epoch": 1.7197202797202797, "percentage": 57.37, "elapsed_time": "1 day, 1:13:04", "remaining_time": "18:44:13", "throughput": "17752.64", "total_tokens": 1611661312}
{"current_steps": 1538, "total_steps": 2679, "loss": 0.6816, "learning_rate": 1.9540065483498978e-06, "epoch": 1.7208391608391609, "percentage": 57.41, "elapsed_time": "1 day, 1:14:04", "remaining_time": "18:43:14", "throughput": "17752.47", "total_tokens": 1612709888}
{"current_steps": 1539, "total_steps": 2679, "loss": 0.6292, "learning_rate": 1.9511190694911875e-06, "epoch": 1.7219580419580418, "percentage": 57.45, "elapsed_time": "1 day, 1:14:57", "remaining_time": "18:42:11", "throughput": "17753.57", "total_tokens": 1613758464}
{"current_steps": 1540, "total_steps": 2679, "loss": 0.5989, "learning_rate": 1.948232359720416e-06, "epoch": 1.7230769230769232, "percentage": 57.48, "elapsed_time": "1 day, 1:15:46", "remaining_time": "18:41:04", "throughput": "17755.64", "total_tokens": 1614807040}
{"current_steps": 1541, "total_steps": 2679, "loss": 0.6825, "learning_rate": 1.9453464230824186e-06, "epoch": 1.724195804195804, "percentage": 57.52, "elapsed_time": "1 day, 1:16:45", "remaining_time": "18:40:06", "throughput": "17755.54", "total_tokens": 1615855616}
{"current_steps": 1542, "total_steps": 2679, "loss": 0.7772, "learning_rate": 1.9424612636209503e-06, "epoch": 1.7253146853146855, "percentage": 57.56, "elapsed_time": "1 day, 1:17:50", "remaining_time": "18:39:11", "throughput": "17754.44", "total_tokens": 1616904192}
{"current_steps": 1543, "total_steps": 2679, "loss": 0.6889, "learning_rate": 1.939576885378674e-06, "epoch": 1.7264335664335664, "percentage": 57.6, "elapsed_time": "1 day, 1:18:44", "remaining_time": "18:38:08", "throughput": "17755.37", "total_tokens": 1617952768}
{"current_steps": 1544, "total_steps": 2679, "loss": 0.6516, "learning_rate": 1.9366932923971583e-06, "epoch": 1.7275524475524475, "percentage": 57.63, "elapsed_time": "1 day, 1:19:38", "remaining_time": "18:37:05", "throughput": "17756.33", "total_tokens": 1619001344}
{"current_steps": 1545, "total_steps": 2679, "loss": 0.6828, "learning_rate": 1.9338104887168753e-06, "epoch": 1.7286713286713287, "percentage": 57.67, "elapsed_time": "1 day, 1:20:35", "remaining_time": "18:36:05", "throughput": "17756.80", "total_tokens": 1620049920}
{"current_steps": 1546, "total_steps": 2679, "loss": 0.727, "learning_rate": 1.9309284783771857e-06, "epoch": 1.7297902097902098, "percentage": 57.71, "elapsed_time": "1 day, 1:21:35", "remaining_time": "18:35:06", "throughput": "17756.54", "total_tokens": 1621098496}
{"current_steps": 1547, "total_steps": 2679, "loss": 0.613, "learning_rate": 1.9280472654163436e-06, "epoch": 1.730909090909091, "percentage": 57.75, "elapsed_time": "1 day, 1:22:25", "remaining_time": "18:34:00", "throughput": "17758.43", "total_tokens": 1622147072}
{"current_steps": 1548, "total_steps": 2679, "loss": 0.6623, "learning_rate": 1.9251668538714814e-06, "epoch": 1.7320279720279719, "percentage": 57.78, "elapsed_time": "1 day, 1:23:19", "remaining_time": "18:32:58", "throughput": "17759.42", "total_tokens": 1623195648}
{"current_steps": 1549, "total_steps": 2679, "loss": 0.7716, "learning_rate": 1.9222872477786124e-06, "epoch": 1.7331468531468532, "percentage": 57.82, "elapsed_time": "1 day, 1:24:24", "remaining_time": "18:32:03", "throughput": "17758.11", "total_tokens": 1624244224}
{"current_steps": 1550, "total_steps": 2679, "loss": 0.7813, "learning_rate": 1.919408451172619e-06, "epoch": 1.7342657342657342, "percentage": 57.86, "elapsed_time": "1 day, 1:25:30", "remaining_time": "18:31:09", "throughput": "17756.87", "total_tokens": 1625292800}
{"current_steps": 1551, "total_steps": 2679, "loss": 0.7538, "learning_rate": 1.916530468087249e-06, "epoch": 1.7353846153846155, "percentage": 57.89, "elapsed_time": "1 day, 1:26:26", "remaining_time": "18:30:08", "throughput": "17757.39", "total_tokens": 1626341376}
{"current_steps": 1552, "total_steps": 2679, "loss": 0.7096, "learning_rate": 1.9136533025551126e-06, "epoch": 1.7365034965034964, "percentage": 57.93, "elapsed_time": "1 day, 1:27:23", "remaining_time": "18:29:08", "throughput": "17757.74", "total_tokens": 1627389952}
{"current_steps": 1553, "total_steps": 2679, "loss": 0.6097, "learning_rate": 1.9107769586076716e-06, "epoch": 1.7376223776223776, "percentage": 57.97, "elapsed_time": "1 day, 1:28:13", "remaining_time": "18:28:01", "throughput": "17759.67", "total_tokens": 1628438528}
{"current_steps": 1554, "total_steps": 2679, "loss": 0.683, "learning_rate": 1.9079014402752392e-06, "epoch": 1.7387412587412587, "percentage": 58.01, "elapsed_time": "1 day, 1:29:04", "remaining_time": "18:26:57", "throughput": "17761.17", "total_tokens": 1629487104}
{"current_steps": 1555, "total_steps": 2679, "loss": 0.6598, "learning_rate": 1.9050267515869709e-06, "epoch": 1.7398601398601399, "percentage": 58.04, "elapsed_time": "1 day, 1:29:58", "remaining_time": "18:25:54", "throughput": "17762.11", "total_tokens": 1630535680}
{"current_steps": 1556, "total_steps": 2679, "loss": 0.633, "learning_rate": 1.9021528965708576e-06, "epoch": 1.740979020979021, "percentage": 58.08, "elapsed_time": "1 day, 1:30:52", "remaining_time": "18:24:51", "throughput": "17763.11", "total_tokens": 1631584256}
{"current_steps": 1557, "total_steps": 2679, "loss": 0.6395, "learning_rate": 1.8992798792537265e-06, "epoch": 1.742097902097902, "percentage": 58.12, "elapsed_time": "1 day, 1:31:42", "remaining_time": "18:23:46", "throughput": "17764.91", "total_tokens": 1632632832}
{"current_steps": 1558, "total_steps": 2679, "loss": 0.6078, "learning_rate": 1.8964077036612262e-06, "epoch": 1.7432167832167833, "percentage": 58.16, "elapsed_time": "1 day, 1:32:29", "remaining_time": "18:22:39", "throughput": "17767.06", "total_tokens": 1633681408}
{"current_steps": 1559, "total_steps": 2679, "loss": 0.7763, "learning_rate": 1.8935363738178288e-06, "epoch": 1.7443356643356642, "percentage": 58.19, "elapsed_time": "1 day, 1:33:30", "remaining_time": "18:21:40", "throughput": "17766.84", "total_tokens": 1634729984}
{"current_steps": 1560, "total_steps": 2679, "loss": 0.6203, "learning_rate": 1.8906658937468205e-06, "epoch": 1.7454545454545456, "percentage": 58.23, "elapsed_time": "1 day, 1:34:23", "remaining_time": "18:20:38", "throughput": "17767.85", "total_tokens": 1635778560}
{"current_steps": 1561, "total_steps": 2679, "loss": 0.7323, "learning_rate": 1.8877962674702977e-06, "epoch": 1.7465734265734265, "percentage": 58.27, "elapsed_time": "1 day, 1:35:23", "remaining_time": "18:19:39", "throughput": "17767.71", "total_tokens": 1636827136}
{"current_steps": 1562, "total_steps": 2679, "loss": 0.738, "learning_rate": 1.8849274990091599e-06, "epoch": 1.7476923076923077, "percentage": 58.31, "elapsed_time": "1 day, 1:36:23", "remaining_time": "18:18:41", "throughput": "17767.60", "total_tokens": 1637875712}
{"current_steps": 1563, "total_steps": 2679, "loss": 0.6101, "learning_rate": 1.8820595923831025e-06, "epoch": 1.7488111888111888, "percentage": 58.34, "elapsed_time": "1 day, 1:37:17", "remaining_time": "18:17:38", "throughput": "17768.52", "total_tokens": 1638924288}
{"current_steps": 1564, "total_steps": 2679, "loss": 0.5869, "learning_rate": 1.8791925516106169e-06, "epoch": 1.74993006993007, "percentage": 58.38, "elapsed_time": "1 day, 1:38:05", "remaining_time": "18:16:32", "throughput": "17770.58", "total_tokens": 1639972864}
{"current_steps": 1565, "total_steps": 2679, "loss": 0.76, "learning_rate": 1.876326380708979e-06, "epoch": 1.751048951048951, "percentage": 58.42, "elapsed_time": "1 day, 1:39:05", "remaining_time": "18:15:33", "throughput": "17770.44", "total_tokens": 1641021440}
{"current_steps": 1566, "total_steps": 2679, "loss": 0.607, "learning_rate": 1.8734610836942467e-06, "epoch": 1.752167832167832, "percentage": 58.45, "elapsed_time": "1 day, 1:39:59", "remaining_time": "18:14:30", "throughput": "17771.48", "total_tokens": 1642070016}
{"current_steps": 1567, "total_steps": 2679, "loss": 0.6074, "learning_rate": 1.8705966645812544e-06, "epoch": 1.7532867132867134, "percentage": 58.49, "elapsed_time": "1 day, 1:40:47", "remaining_time": "18:13:23", "throughput": "17773.61", "total_tokens": 1643118592}
{"current_steps": 1568, "total_steps": 2679, "loss": 0.7481, "learning_rate": 1.8677331273836025e-06, "epoch": 1.7544055944055943, "percentage": 58.53, "elapsed_time": "1 day, 1:41:46", "remaining_time": "18:12:24", "throughput": "17773.56", "total_tokens": 1644167168}
{"current_steps": 1569, "total_steps": 2679, "loss": 0.5741, "learning_rate": 1.8648704761136604e-06, "epoch": 1.7555244755244757, "percentage": 58.57, "elapsed_time": "1 day, 1:42:34", "remaining_time": "18:11:18", "throughput": "17775.66", "total_tokens": 1645215744}
{"current_steps": 1570, "total_steps": 2679, "loss": 0.6466, "learning_rate": 1.8620087147825528e-06, "epoch": 1.7566433566433566, "percentage": 58.6, "elapsed_time": "1 day, 1:43:28", "remaining_time": "18:10:16", "throughput": "17776.51", "total_tokens": 1646264320}
{"current_steps": 1571, "total_steps": 2679, "loss": 0.7783, "learning_rate": 1.8591478474001601e-06, "epoch": 1.7577622377622377, "percentage": 58.64, "elapsed_time": "1 day, 1:44:34", "remaining_time": "18:09:21", "throughput": "17775.27", "total_tokens": 1647312896}
{"current_steps": 1572, "total_steps": 2679, "loss": 0.6752, "learning_rate": 1.8562878779751074e-06, "epoch": 1.7588811188811189, "percentage": 58.68, "elapsed_time": "1 day, 1:45:28", "remaining_time": "18:08:19", "throughput": "17776.23", "total_tokens": 1648361472}
{"current_steps": 1573, "total_steps": 2679, "loss": 0.741, "learning_rate": 1.8534288105147644e-06, "epoch": 1.76, "percentage": 58.72, "elapsed_time": "1 day, 1:46:28", "remaining_time": "18:07:20", "throughput": "17776.06", "total_tokens": 1649410048}
{"current_steps": 1574, "total_steps": 2679, "loss": 0.6407, "learning_rate": 1.850570649025234e-06, "epoch": 1.7611188811188812, "percentage": 58.75, "elapsed_time": "1 day, 1:47:21", "remaining_time": "18:06:18", "throughput": "17777.08", "total_tokens": 1650458624}
{"current_steps": 1575, "total_steps": 2679, "loss": 0.5687, "learning_rate": 1.8477133975113516e-06, "epoch": 1.762237762237762, "percentage": 58.79, "elapsed_time": "1 day, 1:48:12", "remaining_time": "18:05:13", "throughput": "17778.70", "total_tokens": 1651507200}
{"current_steps": 1576, "total_steps": 2679, "loss": 0.7221, "learning_rate": 1.8448570599766772e-06, "epoch": 1.7633566433566434, "percentage": 58.83, "elapsed_time": "1 day, 1:49:11", "remaining_time": "18:04:14", "throughput": "17778.65", "total_tokens": 1652555776}
{"current_steps": 1577, "total_steps": 2679, "loss": 0.7117, "learning_rate": 1.8420016404234897e-06, "epoch": 1.7644755244755244, "percentage": 58.87, "elapsed_time": "1 day, 1:50:17", "remaining_time": "18:03:19", "throughput": "17777.43", "total_tokens": 1653604352}
{"current_steps": 1578, "total_steps": 2679, "loss": 0.653, "learning_rate": 1.8391471428527835e-06, "epoch": 1.7655944055944057, "percentage": 58.9, "elapsed_time": "1 day, 1:51:16", "remaining_time": "18:02:20", "throughput": "17777.42", "total_tokens": 1654652928}
{"current_steps": 1579, "total_steps": 2679, "loss": 0.6882, "learning_rate": 1.836293571264258e-06, "epoch": 1.7667132867132866, "percentage": 58.94, "elapsed_time": "1 day, 1:52:15", "remaining_time": "18:01:22", "throughput": "17777.37", "total_tokens": 1655701504}
{"current_steps": 1580, "total_steps": 2679, "loss": 0.624, "learning_rate": 1.8334409296563165e-06, "epoch": 1.7678321678321678, "percentage": 58.98, "elapsed_time": "1 day, 1:53:12", "remaining_time": "18:00:21", "throughput": "17777.76", "total_tokens": 1656750080}
{"current_steps": 1581, "total_steps": 2679, "loss": 0.8613, "learning_rate": 1.830589222026062e-06, "epoch": 1.768951048951049, "percentage": 59.01, "elapsed_time": "1 day, 1:54:18", "remaining_time": "17:59:27", "throughput": "17776.46", "total_tokens": 1657798656}
{"current_steps": 1582, "total_steps": 2679, "loss": 0.7239, "learning_rate": 1.8277384523692827e-06, "epoch": 1.77006993006993, "percentage": 59.05, "elapsed_time": "1 day, 1:55:18", "remaining_time": "17:58:29", "throughput": "17776.21", "total_tokens": 1658847232}
{"current_steps": 1583, "total_steps": 2679, "loss": 0.6033, "learning_rate": 1.8248886246804598e-06, "epoch": 1.7711888111888112, "percentage": 59.09, "elapsed_time": "1 day, 1:56:07", "remaining_time": "17:57:23", "throughput": "17778.18", "total_tokens": 1659895808}
{"current_steps": 1584, "total_steps": 2679, "loss": 0.5767, "learning_rate": 1.8220397429527484e-06, "epoch": 1.7723076923076924, "percentage": 59.13, "elapsed_time": "1 day, 1:56:55", "remaining_time": "17:56:16", "throughput": "17780.22", "total_tokens": 1660944384}
{"current_steps": 1585, "total_steps": 2679, "loss": 0.6316, "learning_rate": 1.819191811177982e-06, "epoch": 1.7734265734265735, "percentage": 59.16, "elapsed_time": "1 day, 1:57:49", "remaining_time": "17:55:14", "throughput": "17781.20", "total_tokens": 1661992960}
{"current_steps": 1586, "total_steps": 2679, "loss": 0.8537, "learning_rate": 1.8163448333466622e-06, "epoch": 1.7745454545454544, "percentage": 59.2, "elapsed_time": "1 day, 1:58:59", "remaining_time": "17:54:23", "throughput": "17778.97", "total_tokens": 1663041536}
{"current_steps": 1587, "total_steps": 2679, "loss": 0.7122, "learning_rate": 1.813498813447951e-06, "epoch": 1.7756643356643358, "percentage": 59.24, "elapsed_time": "1 day, 1:59:59", "remaining_time": "17:53:24", "throughput": "17778.84", "total_tokens": 1664090112}
{"current_steps": 1588, "total_steps": 2679, "loss": 0.649, "learning_rate": 1.8106537554696736e-06, "epoch": 1.7767832167832167, "percentage": 59.28, "elapsed_time": "1 day, 2:00:53", "remaining_time": "17:52:22", "throughput": "17779.81", "total_tokens": 1665138688}
{"current_steps": 1589, "total_steps": 2679, "loss": 0.6713, "learning_rate": 1.8078096633983023e-06, "epoch": 1.7779020979020979, "percentage": 59.31, "elapsed_time": "1 day, 2:01:45", "remaining_time": "17:51:18", "throughput": "17781.17", "total_tokens": 1666187264}
{"current_steps": 1590, "total_steps": 2679, "loss": 0.7523, "learning_rate": 1.804966541218959e-06, "epoch": 1.779020979020979, "percentage": 59.35, "elapsed_time": "1 day, 2:02:44", "remaining_time": "17:50:20", "throughput": "17781.02", "total_tokens": 1667235840}
{"current_steps": 1591, "total_steps": 2679, "loss": 0.769, "learning_rate": 1.8021243929154063e-06, "epoch": 1.7801398601398601, "percentage": 59.39, "elapsed_time": "1 day, 2:03:44", "remaining_time": "17:49:21", "throughput": "17780.85", "total_tokens": 1668284416}
{"current_steps": 1592, "total_steps": 2679, "loss": 0.6585, "learning_rate": 1.7992832224700391e-06, "epoch": 1.7812587412587413, "percentage": 59.43, "elapsed_time": "1 day, 2:04:39", "remaining_time": "17:48:19", "throughput": "17781.72", "total_tokens": 1669332992}
{"current_steps": 1593, "total_steps": 2679, "loss": 0.7053, "learning_rate": 1.7964430338638883e-06, "epoch": 1.7823776223776224, "percentage": 59.46, "elapsed_time": "1 day, 2:05:38", "remaining_time": "17:47:20", "throughput": "17781.65", "total_tokens": 1670381568}
{"current_steps": 1594, "total_steps": 2679, "loss": 0.686, "learning_rate": 1.793603831076602e-06, "epoch": 1.7834965034965036, "percentage": 59.5, "elapsed_time": "1 day, 2:06:38", "remaining_time": "17:46:22", "throughput": "17781.52", "total_tokens": 1671430144}
{"current_steps": 1595, "total_steps": 2679, "loss": 0.6467, "learning_rate": 1.7907656180864519e-06, "epoch": 1.7846153846153845, "percentage": 59.54, "elapsed_time": "1 day, 2:07:32", "remaining_time": "17:45:20", "throughput": "17782.48", "total_tokens": 1672478720}
{"current_steps": 1596, "total_steps": 2679, "loss": 0.6553, "learning_rate": 1.7879283988703223e-06, "epoch": 1.7857342657342659, "percentage": 59.57, "elapsed_time": "1 day, 2:08:25", "remaining_time": "17:44:17", "throughput": "17783.50", "total_tokens": 1673527296}
{"current_steps": 1597, "total_steps": 2679, "loss": 0.5995, "learning_rate": 1.7850921774037012e-06, "epoch": 1.7868531468531468, "percentage": 59.61, "elapsed_time": "1 day, 2:09:13", "remaining_time": "17:43:11", "throughput": "17785.52", "total_tokens": 1674575872}
{"current_steps": 1598, "total_steps": 2679, "loss": 0.7491, "learning_rate": 1.7822569576606833e-06, "epoch": 1.787972027972028, "percentage": 59.65, "elapsed_time": "1 day, 2:10:19", "remaining_time": "17:42:16", "throughput": "17784.30", "total_tokens": 1675624448}
{"current_steps": 1599, "total_steps": 2679, "loss": 0.6929, "learning_rate": 1.7794227436139569e-06, "epoch": 1.789090909090909, "percentage": 59.69, "elapsed_time": "1 day, 2:11:14", "remaining_time": "17:41:14", "throughput": "17785.09", "total_tokens": 1676673024}
{"current_steps": 1600, "total_steps": 2679, "loss": 0.6819, "learning_rate": 1.776589539234803e-06, "epoch": 1.7902097902097902, "percentage": 59.72, "elapsed_time": "1 day, 2:12:13", "remaining_time": "17:40:15", "throughput": "17785.08", "total_tokens": 1677721600}
{"current_steps": 1601, "total_steps": 2679, "loss": 0.7201, "learning_rate": 1.7737573484930853e-06, "epoch": 1.7913286713286714, "percentage": 59.76, "elapsed_time": "1 day, 2:13:10", "remaining_time": "17:39:15", "throughput": "17785.43", "total_tokens": 1678770176}
{"current_steps": 1602, "total_steps": 2679, "loss": 0.7498, "learning_rate": 1.770926175357251e-06, "epoch": 1.7924475524475525, "percentage": 59.8, "elapsed_time": "1 day, 2:14:15", "remaining_time": "17:38:20", "throughput": "17784.23", "total_tokens": 1679818752}
{"current_steps": 1603, "total_steps": 2679, "loss": 0.5936, "learning_rate": 1.7680960237943174e-06, "epoch": 1.7935664335664336, "percentage": 59.84, "elapsed_time": "1 day, 2:15:09", "remaining_time": "17:37:18", "throughput": "17785.16", "total_tokens": 1680867328}
{"current_steps": 1604, "total_steps": 2679, "loss": 1.0174, "learning_rate": 1.7652668977698714e-06, "epoch": 1.7946853146853146, "percentage": 59.87, "elapsed_time": "1 day, 2:16:21", "remaining_time": "17:36:28", "throughput": "17782.68", "total_tokens": 1681915904}
{"current_steps": 1605, "total_steps": 2679, "loss": 0.7075, "learning_rate": 1.7624388012480656e-06, "epoch": 1.795804195804196, "percentage": 59.91, "elapsed_time": "1 day, 2:17:21", "remaining_time": "17:35:30", "throughput": "17782.47", "total_tokens": 1682964480}
{"current_steps": 1606, "total_steps": 2679, "loss": 0.7212, "learning_rate": 1.7596117381916068e-06, "epoch": 1.7969230769230768, "percentage": 59.95, "elapsed_time": "1 day, 2:18:18", "remaining_time": "17:34:29", "throughput": "17782.92", "total_tokens": 1684013056}
{"current_steps": 1607, "total_steps": 2679, "loss": 0.6554, "learning_rate": 1.756785712561756e-06, "epoch": 1.798041958041958, "percentage": 59.99, "elapsed_time": "1 day, 2:19:07", "remaining_time": "17:33:24", "throughput": "17784.72", "total_tokens": 1685061632}
{"current_steps": 1608, "total_steps": 2679, "loss": 0.8153, "learning_rate": 1.7539607283183191e-06, "epoch": 1.7991608391608391, "percentage": 60.02, "elapsed_time": "1 day, 2:20:08", "remaining_time": "17:32:26", "throughput": "17784.44", "total_tokens": 1686110208}
{"current_steps": 1609, "total_steps": 2679, "loss": 0.9455, "learning_rate": 1.7511367894196426e-06, "epoch": 1.8002797202797203, "percentage": 60.06, "elapsed_time": "1 day, 2:21:20", "remaining_time": "17:31:36", "throughput": "17782.01", "total_tokens": 1687158784}
{"current_steps": 1610, "total_steps": 2679, "loss": 0.6293, "learning_rate": 1.7483138998226102e-06, "epoch": 1.8013986013986014, "percentage": 60.1, "elapsed_time": "1 day, 2:22:13", "remaining_time": "17:30:33", "throughput": "17782.99", "total_tokens": 1688207360}
{"current_steps": 1611, "total_steps": 2679, "loss": 0.6063, "learning_rate": 1.7454920634826334e-06, "epoch": 1.8025174825174826, "percentage": 60.13, "elapsed_time": "1 day, 2:23:01", "remaining_time": "17:29:27", "throughput": "17785.03", "total_tokens": 1689255936}
{"current_steps": 1612, "total_steps": 2679, "loss": 0.6952, "learning_rate": 1.7426712843536497e-06, "epoch": 1.8036363636363637, "percentage": 60.17, "elapsed_time": "1 day, 2:23:58", "remaining_time": "17:28:27", "throughput": "17785.42", "total_tokens": 1690304512}
{"current_steps": 1613, "total_steps": 2679, "loss": 0.7721, "learning_rate": 1.7398515663881117e-06, "epoch": 1.8047552447552446, "percentage": 60.21, "elapsed_time": "1 day, 2:25:09", "remaining_time": "17:27:36", "throughput": "17783.18", "total_tokens": 1691353088}
{"current_steps": 1614, "total_steps": 2679, "loss": 0.5806, "learning_rate": 1.7370329135369906e-06, "epoch": 1.805874125874126, "percentage": 60.25, "elapsed_time": "1 day, 2:25:59", "remaining_time": "17:26:30", "throughput": "17784.94", "total_tokens": 1692401664}
{"current_steps": 1615, "total_steps": 2679, "loss": 0.6257, "learning_rate": 1.73421532974976e-06, "epoch": 1.806993006993007, "percentage": 60.28, "elapsed_time": "1 day, 2:26:53", "remaining_time": "17:25:28", "throughput": "17785.83", "total_tokens": 1693450240}
{"current_steps": 1616, "total_steps": 2679, "loss": 0.6538, "learning_rate": 1.731398818974398e-06, "epoch": 1.8081118881118883, "percentage": 60.32, "elapsed_time": "1 day, 2:27:53", "remaining_time": "17:24:30", "throughput": "17785.70", "total_tokens": 1694498816}
{"current_steps": 1617, "total_steps": 2679, "loss": 0.7351, "learning_rate": 1.7285833851573802e-06, "epoch": 1.8092307692307692, "percentage": 60.36, "elapsed_time": "1 day, 2:28:53", "remaining_time": "17:23:32", "throughput": "17785.44", "total_tokens": 1695547392}
{"current_steps": 1618, "total_steps": 2679, "loss": 0.6641, "learning_rate": 1.72576903224367e-06, "epoch": 1.8103496503496503, "percentage": 60.4, "elapsed_time": "1 day, 2:29:52", "remaining_time": "17:22:33", "throughput": "17785.36", "total_tokens": 1696595968}
{"current_steps": 1619, "total_steps": 2679, "loss": 0.6571, "learning_rate": 1.7229557641767191e-06, "epoch": 1.8114685314685315, "percentage": 60.43, "elapsed_time": "1 day, 2:30:46", "remaining_time": "17:21:31", "throughput": "17786.40", "total_tokens": 1697644544}
{"current_steps": 1620, "total_steps": 2679, "loss": 0.7547, "learning_rate": 1.7201435848984582e-06, "epoch": 1.8125874125874126, "percentage": 60.47, "elapsed_time": "1 day, 2:31:49", "remaining_time": "17:20:35", "throughput": "17785.53", "total_tokens": 1698693120}
{"current_steps": 1621, "total_steps": 2679, "loss": 0.7071, "learning_rate": 1.7173324983492912e-06, "epoch": 1.8137062937062938, "percentage": 60.51, "elapsed_time": "1 day, 2:32:50", "remaining_time": "17:19:37", "throughput": "17785.29", "total_tokens": 1699741696}
{"current_steps": 1622, "total_steps": 2679, "loss": 0.7692, "learning_rate": 1.7145225084680939e-06, "epoch": 1.8148251748251747, "percentage": 60.54, "elapsed_time": "1 day, 2:33:50", "remaining_time": "17:18:38", "throughput": "17785.03", "total_tokens": 1700790272}
{"current_steps": 1623, "total_steps": 2679, "loss": 0.7484, "learning_rate": 1.7117136191922013e-06, "epoch": 1.815944055944056, "percentage": 60.58, "elapsed_time": "1 day, 2:34:50", "remaining_time": "17:17:40", "throughput": "17784.86", "total_tokens": 1701838848}
{"current_steps": 1624, "total_steps": 2679, "loss": 0.7389, "learning_rate": 1.70890583445741e-06, "epoch": 1.817062937062937, "percentage": 60.62, "elapsed_time": "1 day, 2:35:52", "remaining_time": "17:16:43", "throughput": "17784.24", "total_tokens": 1702887424}
{"current_steps": 1625, "total_steps": 2679, "loss": 0.7051, "learning_rate": 1.7060991581979668e-06, "epoch": 1.8181818181818183, "percentage": 60.66, "elapsed_time": "1 day, 2:36:48", "remaining_time": "17:15:42", "throughput": "17784.82", "total_tokens": 1703936000}
{"current_steps": 1626, "total_steps": 2679, "loss": 0.6826, "learning_rate": 1.7032935943465664e-06, "epoch": 1.8193006993006993, "percentage": 60.69, "elapsed_time": "1 day, 2:37:47", "remaining_time": "17:14:44", "throughput": "17784.74", "total_tokens": 1704984576}
{"current_steps": 1627, "total_steps": 2679, "loss": 0.6736, "learning_rate": 1.7004891468343445e-06, "epoch": 1.8204195804195804, "percentage": 60.73, "elapsed_time": "1 day, 2:38:41", "remaining_time": "17:13:41", "throughput": "17785.67", "total_tokens": 1706033152}
{"current_steps": 1628, "total_steps": 2679, "loss": 0.6188, "learning_rate": 1.6976858195908707e-06, "epoch": 1.8215384615384616, "percentage": 60.77, "elapsed_time": "1 day, 2:39:35", "remaining_time": "17:12:39", "throughput": "17786.66", "total_tokens": 1707081728}
{"current_steps": 1629, "total_steps": 2679, "loss": 0.7019, "learning_rate": 1.6948836165441487e-06, "epoch": 1.8226573426573427, "percentage": 60.81, "elapsed_time": "1 day, 2:40:35", "remaining_time": "17:11:41", "throughput": "17786.54", "total_tokens": 1708130304}
{"current_steps": 1630, "total_steps": 2679, "loss": 0.8377, "learning_rate": 1.6920825416206032e-06, "epoch": 1.8237762237762238, "percentage": 60.84, "elapsed_time": "1 day, 2:41:36", "remaining_time": "17:10:43", "throughput": "17786.12", "total_tokens": 1709178880}
{"current_steps": 1631, "total_steps": 2679, "loss": 0.6087, "learning_rate": 1.6892825987450811e-06, "epoch": 1.8248951048951048, "percentage": 60.88, "elapsed_time": "1 day, 2:42:30", "remaining_time": "17:09:41", "throughput": "17786.90", "total_tokens": 1710227456}
{"current_steps": 1632, "total_steps": 2679, "loss": 0.6364, "learning_rate": 1.6864837918408422e-06, "epoch": 1.8260139860139861, "percentage": 60.92, "elapsed_time": "1 day, 2:43:25", "remaining_time": "17:08:39", "throughput": "17787.79", "total_tokens": 1711276032}
{"current_steps": 1633, "total_steps": 2679, "loss": 0.6452, "learning_rate": 1.6836861248295522e-06, "epoch": 1.827132867132867, "percentage": 60.96, "elapsed_time": "1 day, 2:44:18", "remaining_time": "17:07:37", "throughput": "17788.83", "total_tokens": 1712324608}
{"current_steps": 1634, "total_steps": 2679, "loss": 0.7781, "learning_rate": 1.6808896016312832e-06, "epoch": 1.8282517482517484, "percentage": 60.99, "elapsed_time": "1 day, 2:45:18", "remaining_time": "17:06:38", "throughput": "17788.68", "total_tokens": 1713373184}
{"current_steps": 1635, "total_steps": 2679, "loss": 0.7269, "learning_rate": 1.6780942261645022e-06, "epoch": 1.8293706293706293, "percentage": 61.03, "elapsed_time": "1 day, 2:46:21", "remaining_time": "17:05:42", "throughput": "17787.83", "total_tokens": 1714421760}
{"current_steps": 1636, "total_steps": 2679, "loss": 0.7237, "learning_rate": 1.6753000023460698e-06, "epoch": 1.8304895104895105, "percentage": 61.07, "elapsed_time": "1 day, 2:47:21", "remaining_time": "17:04:44", "throughput": "17787.60", "total_tokens": 1715470336}
{"current_steps": 1637, "total_steps": 2679, "loss": 0.7058, "learning_rate": 1.6725069340912306e-06, "epoch": 1.8316083916083916, "percentage": 61.1, "elapsed_time": "1 day, 2:48:21", "remaining_time": "17:03:45", "throughput": "17787.52", "total_tokens": 1716518912}
{"current_steps": 1638, "total_steps": 2679, "loss": 0.7408, "learning_rate": 1.6697150253136136e-06, "epoch": 1.8327272727272728, "percentage": 61.14, "elapsed_time": "1 day, 2:49:26", "remaining_time": "17:02:51", "throughput": "17786.37", "total_tokens": 1717567488}
{"current_steps": 1639, "total_steps": 2679, "loss": 0.7289, "learning_rate": 1.666924279925219e-06, "epoch": 1.833846153846154, "percentage": 61.18, "elapsed_time": "1 day, 2:50:27", "remaining_time": "17:01:53", "throughput": "17785.96", "total_tokens": 1718616064}
{"current_steps": 1640, "total_steps": 2679, "loss": 0.6819, "learning_rate": 1.6641347018364182e-06, "epoch": 1.8349650349650348, "percentage": 61.22, "elapsed_time": "1 day, 2:51:27", "remaining_time": "17:00:55", "throughput": "17785.79", "total_tokens": 1719664640}
{"current_steps": 1641, "total_steps": 2679, "loss": 0.5842, "learning_rate": 1.6613462949559494e-06, "epoch": 1.8360839160839162, "percentage": 61.25, "elapsed_time": "1 day, 2:52:15", "remaining_time": "16:59:49", "throughput": "17787.86", "total_tokens": 1720713216}
{"current_steps": 1642, "total_steps": 2679, "loss": 0.5558, "learning_rate": 1.6585590631909072e-06, "epoch": 1.837202797202797, "percentage": 61.29, "elapsed_time": "1 day, 2:53:04", "remaining_time": "16:58:43", "throughput": "17789.65", "total_tokens": 1721761792}
{"current_steps": 1643, "total_steps": 2679, "loss": 0.6536, "learning_rate": 1.6557730104467407e-06, "epoch": 1.8383216783216785, "percentage": 61.33, "elapsed_time": "1 day, 2:53:58", "remaining_time": "16:57:41", "throughput": "17790.56", "total_tokens": 1722810368}
{"current_steps": 1644, "total_steps": 2679, "loss": 0.7932, "learning_rate": 1.6529881406272457e-06, "epoch": 1.8394405594405594, "percentage": 61.37, "elapsed_time": "1 day, 2:54:59", "remaining_time": "16:56:44", "throughput": "17790.23", "total_tokens": 1723858944}
{"current_steps": 1645, "total_steps": 2679, "loss": 0.8077, "learning_rate": 1.6502044576345614e-06, "epoch": 1.8405594405594405, "percentage": 61.4, "elapsed_time": "1 day, 2:56:00", "remaining_time": "16:55:46", "throughput": "17789.76", "total_tokens": 1724907520}
{"current_steps": 1646, "total_steps": 2679, "loss": 0.754, "learning_rate": 1.647421965369165e-06, "epoch": 1.8416783216783217, "percentage": 61.44, "elapsed_time": "1 day, 2:57:11", "remaining_time": "16:54:55", "throughput": "17787.60", "total_tokens": 1725956096}
{"current_steps": 1647, "total_steps": 2679, "loss": 0.7005, "learning_rate": 1.6446406677298632e-06, "epoch": 1.8427972027972028, "percentage": 61.48, "elapsed_time": "1 day, 2:58:11", "remaining_time": "16:53:56", "throughput": "17787.42", "total_tokens": 1727004672}
{"current_steps": 1648, "total_steps": 2679, "loss": 0.6724, "learning_rate": 1.6418605686137914e-06, "epoch": 1.843916083916084, "percentage": 61.52, "elapsed_time": "1 day, 2:59:05", "remaining_time": "16:52:55", "throughput": "17788.22", "total_tokens": 1728053248}
{"current_steps": 1649, "total_steps": 2679, "loss": 0.678, "learning_rate": 1.6390816719164022e-06, "epoch": 1.845034965034965, "percentage": 61.55, "elapsed_time": "1 day, 3:00:05", "remaining_time": "16:51:56", "throughput": "17788.18", "total_tokens": 1729101824}
{"current_steps": 1650, "total_steps": 2679, "loss": 0.8139, "learning_rate": 1.6363039815314668e-06, "epoch": 1.8461538461538463, "percentage": 61.59, "elapsed_time": "1 day, 3:01:15", "remaining_time": "16:51:04", "throughput": "17786.01", "total_tokens": 1730150400}
{"current_steps": 1651, "total_steps": 2679, "loss": 0.6761, "learning_rate": 1.6335275013510638e-06, "epoch": 1.8472727272727272, "percentage": 61.63, "elapsed_time": "1 day, 3:02:15", "remaining_time": "16:50:06", "throughput": "17785.92", "total_tokens": 1731198976}
{"current_steps": 1652, "total_steps": 2679, "loss": 0.719, "learning_rate": 1.630752235265577e-06, "epoch": 1.8483916083916085, "percentage": 61.66, "elapsed_time": "1 day, 3:03:21", "remaining_time": "16:49:11", "throughput": "17784.65", "total_tokens": 1732247552}
{"current_steps": 1653, "total_steps": 2679, "loss": 0.7217, "learning_rate": 1.6279781871636896e-06, "epoch": 1.8495104895104895, "percentage": 61.7, "elapsed_time": "1 day, 3:04:26", "remaining_time": "16:48:16", "throughput": "17783.43", "total_tokens": 1733296128}
{"current_steps": 1654, "total_steps": 2679, "loss": 0.6829, "learning_rate": 1.6252053609323758e-06, "epoch": 1.8506293706293706, "percentage": 61.74, "elapsed_time": "1 day, 3:05:26", "remaining_time": "16:47:17", "throughput": "17783.38", "total_tokens": 1734344704}
{"current_steps": 1655, "total_steps": 2679, "loss": 0.7143, "learning_rate": 1.6224337604569012e-06, "epoch": 1.8517482517482518, "percentage": 61.78, "elapsed_time": "1 day, 3:06:31", "remaining_time": "16:46:22", "throughput": "17782.22", "total_tokens": 1735393280}
{"current_steps": 1656, "total_steps": 2679, "loss": 0.6368, "learning_rate": 1.6196633896208118e-06, "epoch": 1.852867132867133, "percentage": 61.81, "elapsed_time": "1 day, 3:07:25", "remaining_time": "16:45:20", "throughput": "17783.14", "total_tokens": 1736441856}
{"current_steps": 1657, "total_steps": 2679, "loss": 0.6704, "learning_rate": 1.616894252305929e-06, "epoch": 1.853986013986014, "percentage": 61.85, "elapsed_time": "1 day, 3:08:21", "remaining_time": "16:44:20", "throughput": "17783.67", "total_tokens": 1737490432}
{"current_steps": 1658, "total_steps": 2679, "loss": 0.6929, "learning_rate": 1.6141263523923512e-06, "epoch": 1.855104895104895, "percentage": 61.89, "elapsed_time": "1 day, 3:09:15", "remaining_time": "16:43:18", "throughput": "17784.50", "total_tokens": 1738539008}
{"current_steps": 1659, "total_steps": 2679, "loss": 0.7426, "learning_rate": 1.6113596937584358e-06, "epoch": 1.8562237762237763, "percentage": 61.93, "elapsed_time": "1 day, 3:10:21", "remaining_time": "16:42:23", "throughput": "17783.26", "total_tokens": 1739587584}
{"current_steps": 1660, "total_steps": 2679, "loss": 0.6341, "learning_rate": 1.6085942802808068e-06, "epoch": 1.8573426573426572, "percentage": 61.96, "elapsed_time": "1 day, 3:11:15", "remaining_time": "16:41:21", "throughput": "17784.22", "total_tokens": 1740636160}
{"current_steps": 1661, "total_steps": 2679, "loss": 0.7194, "learning_rate": 1.6058301158343408e-06, "epoch": 1.8584615384615386, "percentage": 62.0, "elapsed_time": "1 day, 3:12:18", "remaining_time": "16:40:24", "throughput": "17783.43", "total_tokens": 1741684736}
{"current_steps": 1662, "total_steps": 2679, "loss": 0.6637, "learning_rate": 1.6030672042921647e-06, "epoch": 1.8595804195804195, "percentage": 62.04, "elapsed_time": "1 day, 3:13:17", "remaining_time": "16:39:25", "throughput": "17783.52", "total_tokens": 1742733312}
{"current_steps": 1663, "total_steps": 2679, "loss": 0.7159, "learning_rate": 1.600305549525651e-06, "epoch": 1.8606993006993007, "percentage": 62.08, "elapsed_time": "1 day, 3:14:11", "remaining_time": "16:38:24", "throughput": "17784.33", "total_tokens": 1743781888}
{"current_steps": 1664, "total_steps": 2679, "loss": 0.6029, "learning_rate": 1.5975451554044074e-06, "epoch": 1.8618181818181818, "percentage": 62.11, "elapsed_time": "1 day, 3:15:05", "remaining_time": "16:37:21", "throughput": "17785.30", "total_tokens": 1744830464}
{"current_steps": 1665, "total_steps": 2679, "loss": 0.7696, "learning_rate": 1.5947860257962808e-06, "epoch": 1.862937062937063, "percentage": 62.15, "elapsed_time": "1 day, 3:16:08", "remaining_time": "16:36:25", "throughput": "17784.44", "total_tokens": 1745879040}
{"current_steps": 1666, "total_steps": 2679, "loss": 0.7132, "learning_rate": 1.5920281645673412e-06, "epoch": 1.864055944055944, "percentage": 62.19, "elapsed_time": "1 day, 3:17:14", "remaining_time": "16:35:30", "throughput": "17783.30", "total_tokens": 1746927616}
{"current_steps": 1667, "total_steps": 2679, "loss": 0.6487, "learning_rate": 1.5892715755818855e-06, "epoch": 1.865174825174825, "percentage": 62.22, "elapsed_time": "1 day, 3:18:13", "remaining_time": "16:34:31", "throughput": "17783.22", "total_tokens": 1747976192}
{"current_steps": 1668, "total_steps": 2679, "loss": 0.6523, "learning_rate": 1.586516262702425e-06, "epoch": 1.8662937062937064, "percentage": 62.26, "elapsed_time": "1 day, 3:19:03", "remaining_time": "16:33:27", "throughput": "17784.84", "total_tokens": 1749024768}
{"current_steps": 1669, "total_steps": 2679, "loss": 0.6681, "learning_rate": 1.5837622297896832e-06, "epoch": 1.8674125874125873, "percentage": 62.3, "elapsed_time": "1 day, 3:19:57", "remaining_time": "16:32:25", "throughput": "17785.71", "total_tokens": 1750073344}
{"current_steps": 1670, "total_steps": 2679, "loss": 0.9453, "learning_rate": 1.5810094807025923e-06, "epoch": 1.8685314685314687, "percentage": 62.34, "elapsed_time": "1 day, 3:21:03", "remaining_time": "16:31:31", "throughput": "17784.40", "total_tokens": 1751121920}
{"current_steps": 1671, "total_steps": 2679, "loss": 0.7012, "learning_rate": 1.5782580192982827e-06, "epoch": 1.8696503496503496, "percentage": 62.37, "elapsed_time": "1 day, 3:22:03", "remaining_time": "16:30:32", "throughput": "17784.23", "total_tokens": 1752170496}
{"current_steps": 1672, "total_steps": 2679, "loss": 0.7607, "learning_rate": 1.575507849432083e-06, "epoch": 1.8707692307692307, "percentage": 62.41, "elapsed_time": "1 day, 3:23:09", "remaining_time": "16:29:37", "throughput": "17783.06", "total_tokens": 1753219072}
{"current_steps": 1673, "total_steps": 2679, "loss": 0.6264, "learning_rate": 1.5727589749575107e-06, "epoch": 1.8718881118881119, "percentage": 62.45, "elapsed_time": "1 day, 3:24:04", "remaining_time": "16:28:36", "throughput": "17783.77", "total_tokens": 1754267648}
{"current_steps": 1674, "total_steps": 2679, "loss": 0.6274, "learning_rate": 1.5700113997262695e-06, "epoch": 1.873006993006993, "percentage": 62.49, "elapsed_time": "1 day, 3:24:52", "remaining_time": "16:27:30", "throughput": "17785.74", "total_tokens": 1755316224}
{"current_steps": 1675, "total_steps": 2679, "loss": 0.6287, "learning_rate": 1.56726512758824e-06, "epoch": 1.8741258741258742, "percentage": 62.52, "elapsed_time": "1 day, 3:25:48", "remaining_time": "16:26:29", "throughput": "17786.32", "total_tokens": 1756364800}
{"current_steps": 1676, "total_steps": 2679, "loss": 0.6618, "learning_rate": 1.564520162391479e-06, "epoch": 1.875244755244755, "percentage": 62.56, "elapsed_time": "1 day, 3:26:43", "remaining_time": "16:25:28", "throughput": "17786.99", "total_tokens": 1757413376}
{"current_steps": 1677, "total_steps": 2679, "loss": 0.8474, "learning_rate": 1.5617765079822133e-06, "epoch": 1.8763636363636365, "percentage": 62.6, "elapsed_time": "1 day, 3:27:44", "remaining_time": "16:24:30", "throughput": "17786.67", "total_tokens": 1758461952}
{"current_steps": 1678, "total_steps": 2679, "loss": 0.7726, "learning_rate": 1.5590341682048285e-06, "epoch": 1.8774825174825174, "percentage": 62.64, "elapsed_time": "1 day, 3:28:54", "remaining_time": "16:23:38", "throughput": "17784.61", "total_tokens": 1759510528}
{"current_steps": 1679, "total_steps": 2679, "loss": 0.6259, "learning_rate": 1.5562931469018738e-06, "epoch": 1.8786013986013987, "percentage": 62.67, "elapsed_time": "1 day, 3:29:48", "remaining_time": "16:22:36", "throughput": "17785.50", "total_tokens": 1760559104}
{"current_steps": 1680, "total_steps": 2679, "loss": 0.663, "learning_rate": 1.5535534479140469e-06, "epoch": 1.8797202797202797, "percentage": 62.71, "elapsed_time": "1 day, 3:30:42", "remaining_time": "16:21:34", "throughput": "17786.40", "total_tokens": 1761607680}
{"current_steps": 1681, "total_steps": 2679, "loss": 0.7949, "learning_rate": 1.550815075080193e-06, "epoch": 1.8808391608391608, "percentage": 62.75, "elapsed_time": "1 day, 3:31:48", "remaining_time": "16:20:40", "throughput": "17785.06", "total_tokens": 1762656256}
{"current_steps": 1682, "total_steps": 2679, "loss": 0.5959, "learning_rate": 1.5480780322373026e-06, "epoch": 1.881958041958042, "percentage": 62.78, "elapsed_time": "1 day, 3:32:37", "remaining_time": "16:19:35", "throughput": "17786.98", "total_tokens": 1763704832}
{"current_steps": 1683, "total_steps": 2679, "loss": 0.6469, "learning_rate": 1.5453423232204968e-06, "epoch": 1.883076923076923, "percentage": 62.82, "elapsed_time": "1 day, 3:33:31", "remaining_time": "16:18:33", "throughput": "17787.87", "total_tokens": 1764753408}
{"current_steps": 1684, "total_steps": 2679, "loss": 0.5968, "learning_rate": 1.5426079518630345e-06, "epoch": 1.8841958041958042, "percentage": 62.86, "elapsed_time": "1 day, 3:34:25", "remaining_time": "16:17:31", "throughput": "17788.75", "total_tokens": 1765801984}
{"current_steps": 1685, "total_steps": 2679, "loss": 0.6284, "learning_rate": 1.5398749219962935e-06, "epoch": 1.8853146853146852, "percentage": 62.9, "elapsed_time": "1 day, 3:35:18", "remaining_time": "16:16:29", "throughput": "17789.69", "total_tokens": 1766850560}
{"current_steps": 1686, "total_steps": 2679, "loss": 0.758, "learning_rate": 1.5371432374497778e-06, "epoch": 1.8864335664335665, "percentage": 62.93, "elapsed_time": "1 day, 3:36:20", "remaining_time": "16:15:32", "throughput": "17789.15", "total_tokens": 1767899136}
{"current_steps": 1687, "total_steps": 2679, "loss": 0.7631, "learning_rate": 1.5344129020511029e-06, "epoch": 1.8875524475524474, "percentage": 62.97, "elapsed_time": "1 day, 3:37:20", "remaining_time": "16:14:33", "throughput": "17788.94", "total_tokens": 1768947712}
{"current_steps": 1688, "total_steps": 2679, "loss": 0.832, "learning_rate": 1.5316839196259932e-06, "epoch": 1.8886713286713288, "percentage": 63.01, "elapsed_time": "1 day, 3:38:32", "remaining_time": "16:13:42", "throughput": "17786.74", "total_tokens": 1769996288}
{"current_steps": 1689, "total_steps": 2679, "loss": 0.5984, "learning_rate": 1.5289562939982822e-06, "epoch": 1.8897902097902097, "percentage": 63.05, "elapsed_time": "1 day, 3:39:26", "remaining_time": "16:12:40", "throughput": "17787.63", "total_tokens": 1771044864}
{"current_steps": 1690, "total_steps": 2679, "loss": 0.7892, "learning_rate": 1.5262300289898956e-06, "epoch": 1.8909090909090909, "percentage": 63.08, "elapsed_time": "1 day, 3:40:25", "remaining_time": "16:11:41", "throughput": "17787.52", "total_tokens": 1772093440}
{"current_steps": 1691, "total_steps": 2679, "loss": 0.7893, "learning_rate": 1.523505128420858e-06, "epoch": 1.892027972027972, "percentage": 63.12, "elapsed_time": "1 day, 3:41:36", "remaining_time": "16:10:49", "throughput": "17785.41", "total_tokens": 1773142016}
{"current_steps": 1692, "total_steps": 2679, "loss": 0.8584, "learning_rate": 1.5207815961092798e-06, "epoch": 1.8931468531468532, "percentage": 63.16, "elapsed_time": "1 day, 3:42:42", "remaining_time": "16:09:54", "throughput": "17784.16", "total_tokens": 1774190592}
{"current_steps": 1693, "total_steps": 2679, "loss": 0.6513, "learning_rate": 1.518059435871353e-06, "epoch": 1.8942657342657343, "percentage": 63.2, "elapsed_time": "1 day, 3:43:33", "remaining_time": "16:08:51", "throughput": "17785.57", "total_tokens": 1775239168}
{"current_steps": 1694, "total_steps": 2679, "loss": 0.6159, "learning_rate": 1.5153386515213514e-06, "epoch": 1.8953846153846152, "percentage": 63.23, "elapsed_time": "1 day, 3:44:21", "remaining_time": "16:07:46", "throughput": "17787.46", "total_tokens": 1776287744}
{"current_steps": 1695, "total_steps": 2679, "loss": 0.6859, "learning_rate": 1.5126192468716152e-06, "epoch": 1.8965034965034966, "percentage": 63.27, "elapsed_time": "1 day, 3:45:15", "remaining_time": "16:06:44", "throughput": "17788.33", "total_tokens": 1777336320}
{"current_steps": 1696, "total_steps": 2679, "loss": 0.7051, "learning_rate": 1.5099012257325563e-06, "epoch": 1.8976223776223775, "percentage": 63.31, "elapsed_time": "1 day, 3:46:10", "remaining_time": "16:05:42", "throughput": "17789.13", "total_tokens": 1778384896}
{"current_steps": 1697, "total_steps": 2679, "loss": 0.6118, "learning_rate": 1.5071845919126448e-06, "epoch": 1.8987412587412589, "percentage": 63.34, "elapsed_time": "1 day, 3:46:58", "remaining_time": "16:04:37", "throughput": "17791.07", "total_tokens": 1779433472}
{"current_steps": 1698, "total_steps": 2679, "loss": 0.625, "learning_rate": 1.5044693492184092e-06, "epoch": 1.8998601398601398, "percentage": 63.38, "elapsed_time": "1 day, 3:47:49", "remaining_time": "16:03:33", "throughput": "17792.48", "total_tokens": 1780482048}
{"current_steps": 1699, "total_steps": 2679, "loss": 0.6267, "learning_rate": 1.5017555014544273e-06, "epoch": 1.900979020979021, "percentage": 63.42, "elapsed_time": "1 day, 3:48:41", "remaining_time": "16:02:31", "throughput": "17793.66", "total_tokens": 1781530624}
{"current_steps": 1700, "total_steps": 2679, "loss": 0.9697, "learning_rate": 1.4990430524233213e-06, "epoch": 1.902097902097902, "percentage": 63.46, "elapsed_time": "1 day, 3:49:53", "remaining_time": "16:01:39", "throughput": "17791.35", "total_tokens": 1782579200}
{"current_steps": 1701, "total_steps": 2679, "loss": 0.7214, "learning_rate": 1.4963320059257565e-06, "epoch": 1.9032167832167832, "percentage": 63.49, "elapsed_time": "1 day, 3:50:59", "remaining_time": "16:00:44", "throughput": "17790.14", "total_tokens": 1783627776}
{"current_steps": 1702, "total_steps": 2679, "loss": 0.6651, "learning_rate": 1.493622365760431e-06, "epoch": 1.9043356643356644, "percentage": 63.53, "elapsed_time": "1 day, 3:51:53", "remaining_time": "15:59:42", "throughput": "17791.03", "total_tokens": 1784676352}
{"current_steps": 1703, "total_steps": 2679, "loss": 0.638, "learning_rate": 1.4909141357240731e-06, "epoch": 1.9054545454545453, "percentage": 63.57, "elapsed_time": "1 day, 3:52:47", "remaining_time": "15:58:41", "throughput": "17791.92", "total_tokens": 1785724928}
{"current_steps": 1704, "total_steps": 2679, "loss": 0.6439, "learning_rate": 1.4882073196114343e-06, "epoch": 1.9065734265734267, "percentage": 63.61, "elapsed_time": "1 day, 3:53:37", "remaining_time": "15:57:37", "throughput": "17793.39", "total_tokens": 1786773504}
{"current_steps": 1705, "total_steps": 2679, "loss": 0.7888, "learning_rate": 1.4855019212152852e-06, "epoch": 1.9076923076923076, "percentage": 63.64, "elapsed_time": "1 day, 3:54:43", "remaining_time": "15:56:42", "throughput": "17792.25", "total_tokens": 1787822080}
{"current_steps": 1706, "total_steps": 2679, "loss": 0.6843, "learning_rate": 1.4827979443264113e-06, "epoch": 1.908811188811189, "percentage": 63.68, "elapsed_time": "1 day, 3:55:40", "remaining_time": "15:55:42", "throughput": "17792.53", "total_tokens": 1788870656}
{"current_steps": 1707, "total_steps": 2679, "loss": 0.7045, "learning_rate": 1.4800953927336036e-06, "epoch": 1.9099300699300699, "percentage": 63.72, "elapsed_time": "1 day, 3:56:46", "remaining_time": "15:54:47", "throughput": "17791.32", "total_tokens": 1789919232}
{"current_steps": 1708, "total_steps": 2679, "loss": 0.6767, "learning_rate": 1.47739427022366e-06, "epoch": 1.911048951048951, "percentage": 63.76, "elapsed_time": "1 day, 3:57:40", "remaining_time": "15:53:45", "throughput": "17792.23", "total_tokens": 1790967808}
{"current_steps": 1709, "total_steps": 2679, "loss": 0.7113, "learning_rate": 1.4746945805813707e-06, "epoch": 1.9121678321678321, "percentage": 63.79, "elapsed_time": "1 day, 3:58:34", "remaining_time": "15:52:43", "throughput": "17793.03", "total_tokens": 1792016384}
{"current_steps": 1710, "total_steps": 2679, "loss": 0.6605, "learning_rate": 1.4719963275895239e-06, "epoch": 1.9132867132867133, "percentage": 63.83, "elapsed_time": "1 day, 3:59:34", "remaining_time": "15:51:45", "throughput": "17792.85", "total_tokens": 1793064960}
{"current_steps": 1711, "total_steps": 2679, "loss": 0.7026, "learning_rate": 1.4692995150288896e-06, "epoch": 1.9144055944055944, "percentage": 63.87, "elapsed_time": "1 day, 4:00:37", "remaining_time": "15:50:48", "throughput": "17792.15", "total_tokens": 1794113536}
{"current_steps": 1712, "total_steps": 2679, "loss": 0.8619, "learning_rate": 1.4666041466782227e-06, "epoch": 1.9155244755244756, "percentage": 63.9, "elapsed_time": "1 day, 4:01:43", "remaining_time": "15:49:53", "throughput": "17790.93", "total_tokens": 1795162112}
{"current_steps": 1713, "total_steps": 2679, "loss": 0.7531, "learning_rate": 1.4639102263142546e-06, "epoch": 1.9166433566433567, "percentage": 63.94, "elapsed_time": "1 day, 4:02:48", "remaining_time": "15:48:58", "throughput": "17789.80", "total_tokens": 1796210688}
{"current_steps": 1714, "total_steps": 2679, "loss": 0.6497, "learning_rate": 1.4612177577116843e-06, "epoch": 1.9177622377622376, "percentage": 63.98, "elapsed_time": "1 day, 4:03:42", "remaining_time": "15:47:56", "throughput": "17790.70", "total_tokens": 1797259264}
{"current_steps": 1715, "total_steps": 2679, "loss": 0.6201, "learning_rate": 1.4585267446431817e-06, "epoch": 1.918881118881119, "percentage": 64.02, "elapsed_time": "1 day, 4:04:30", "remaining_time": "15:46:51", "throughput": "17792.61", "total_tokens": 1798307840}
{"current_steps": 1716, "total_steps": 2679, "loss": 0.7372, "learning_rate": 1.4558371908793734e-06, "epoch": 1.92, "percentage": 64.05, "elapsed_time": "1 day, 4:05:35", "remaining_time": "15:45:56", "throughput": "17791.48", "total_tokens": 1799356416}
{"current_steps": 1717, "total_steps": 2679, "loss": 0.779, "learning_rate": 1.4531491001888421e-06, "epoch": 1.921118881118881, "percentage": 64.09, "elapsed_time": "1 day, 4:06:36", "remaining_time": "15:44:58", "throughput": "17791.25", "total_tokens": 1800404992}
{"current_steps": 1718, "total_steps": 2679, "loss": 0.6333, "learning_rate": 1.4504624763381207e-06, "epoch": 1.9222377622377622, "percentage": 64.13, "elapsed_time": "1 day, 4:07:24", "remaining_time": "15:43:53", "throughput": "17793.12", "total_tokens": 1801453568}
{"current_steps": 1719, "total_steps": 2679, "loss": 0.7656, "learning_rate": 1.4477773230916872e-06, "epoch": 1.9233566433566434, "percentage": 64.17, "elapsed_time": "1 day, 4:08:25", "remaining_time": "15:42:55", "throughput": "17792.75", "total_tokens": 1802502144}
{"current_steps": 1720, "total_steps": 2679, "loss": 0.6512, "learning_rate": 1.44509364421196e-06, "epoch": 1.9244755244755245, "percentage": 64.2, "elapsed_time": "1 day, 4:09:18", "remaining_time": "15:41:53", "throughput": "17793.73", "total_tokens": 1803550720}
{"current_steps": 1721, "total_steps": 2679, "loss": 0.5951, "learning_rate": 1.442411443459289e-06, "epoch": 1.9255944055944056, "percentage": 64.24, "elapsed_time": "1 day, 4:10:06", "remaining_time": "15:40:48", "throughput": "17795.65", "total_tokens": 1804599296}
{"current_steps": 1722, "total_steps": 2679, "loss": 0.7829, "learning_rate": 1.4397307245919534e-06, "epoch": 1.9267132867132868, "percentage": 64.28, "elapsed_time": "1 day, 4:11:17", "remaining_time": "15:39:56", "throughput": "17793.54", "total_tokens": 1805647872}
{"current_steps": 1723, "total_steps": 2679, "loss": 0.745, "learning_rate": 1.4370514913661576e-06, "epoch": 1.9278321678321677, "percentage": 64.32, "elapsed_time": "1 day, 4:12:22", "remaining_time": "15:39:00", "throughput": "17792.46", "total_tokens": 1806696448}
{"current_steps": 1724, "total_steps": 2679, "loss": 0.5664, "learning_rate": 1.4343737475360236e-06, "epoch": 1.928951048951049, "percentage": 64.35, "elapsed_time": "1 day, 4:13:11", "remaining_time": "15:37:55", "throughput": "17794.30", "total_tokens": 1807745024}
{"current_steps": 1725, "total_steps": 2679, "loss": 0.653, "learning_rate": 1.4316974968535873e-06, "epoch": 1.93006993006993, "percentage": 64.39, "elapsed_time": "1 day, 4:14:05", "remaining_time": "15:36:54", "throughput": "17795.13", "total_tokens": 1808793600}
{"current_steps": 1726, "total_steps": 2679, "loss": 0.7096, "learning_rate": 1.4290227430687903e-06, "epoch": 1.9311888111888111, "percentage": 64.43, "elapsed_time": "1 day, 4:15:00", "remaining_time": "15:35:53", "throughput": "17795.87", "total_tokens": 1809842176}
{"current_steps": 1727, "total_steps": 2679, "loss": 0.6817, "learning_rate": 1.4263494899294794e-06, "epoch": 1.9323076923076923, "percentage": 64.46, "elapsed_time": "1 day, 4:15:59", "remaining_time": "15:34:54", "throughput": "17795.71", "total_tokens": 1810890752}
{"current_steps": 1728, "total_steps": 2679, "loss": 0.7689, "learning_rate": 1.4236777411813951e-06, "epoch": 1.9334265734265734, "percentage": 64.5, "elapsed_time": "1 day, 4:17:00", "remaining_time": "15:33:56", "throughput": "17795.40", "total_tokens": 1811939328}
{"current_steps": 1729, "total_steps": 2679, "loss": 0.5943, "learning_rate": 1.4210075005681737e-06, "epoch": 1.9345454545454546, "percentage": 64.54, "elapsed_time": "1 day, 4:17:49", "remaining_time": "15:32:52", "throughput": "17797.11", "total_tokens": 1812987904}
{"current_steps": 1730, "total_steps": 2679, "loss": 0.6229, "learning_rate": 1.4183387718313374e-06, "epoch": 1.9356643356643357, "percentage": 64.58, "elapsed_time": "1 day, 4:18:37", "remaining_time": "15:31:47", "throughput": "17799.00", "total_tokens": 1814036480}
{"current_steps": 1731, "total_steps": 2679, "loss": 0.7617, "learning_rate": 1.4156715587102875e-06, "epoch": 1.9367832167832169, "percentage": 64.61, "elapsed_time": "1 day, 4:19:37", "remaining_time": "15:30:49", "throughput": "17798.87", "total_tokens": 1815085056}
{"current_steps": 1732, "total_steps": 2679, "loss": 0.6782, "learning_rate": 1.4130058649423057e-06, "epoch": 1.9379020979020978, "percentage": 64.65, "elapsed_time": "1 day, 4:20:37", "remaining_time": "15:29:50", "throughput": "17798.77", "total_tokens": 1816133632}
{"current_steps": 1733, "total_steps": 2679, "loss": 0.8375, "learning_rate": 1.4103416942625397e-06, "epoch": 1.9390209790209791, "percentage": 64.69, "elapsed_time": "1 day, 4:21:43", "remaining_time": "15:28:55", "throughput": "17797.51", "total_tokens": 1817182208}
{"current_steps": 1734, "total_steps": 2679, "loss": 0.7826, "learning_rate": 1.4076790504040084e-06, "epoch": 1.94013986013986, "percentage": 64.73, "elapsed_time": "1 day, 4:22:49", "remaining_time": "15:28:00", "throughput": "17796.24", "total_tokens": 1818230784}
{"current_steps": 1735, "total_steps": 2679, "loss": 0.6335, "learning_rate": 1.4050179370975886e-06, "epoch": 1.9412587412587412, "percentage": 64.76, "elapsed_time": "1 day, 4:23:43", "remaining_time": "15:26:58", "throughput": "17797.12", "total_tokens": 1819279360}
{"current_steps": 1736, "total_steps": 2679, "loss": 0.7191, "learning_rate": 1.4023583580720112e-06, "epoch": 1.9423776223776223, "percentage": 64.8, "elapsed_time": "1 day, 4:24:42", "remaining_time": "15:26:00", "throughput": "17796.99", "total_tokens": 1820327936}
{"current_steps": 1737, "total_steps": 2679, "loss": 0.7197, "learning_rate": 1.3997003170538608e-06, "epoch": 1.9434965034965035, "percentage": 64.84, "elapsed_time": "1 day, 4:25:45", "remaining_time": "15:25:03", "throughput": "17796.39", "total_tokens": 1821376512}
{"current_steps": 1738, "total_steps": 2679, "loss": 0.7631, "learning_rate": 1.3970438177675632e-06, "epoch": 1.9446153846153846, "percentage": 64.87, "elapsed_time": "1 day, 4:26:48", "remaining_time": "15:24:06", "throughput": "17795.59", "total_tokens": 1822425088}
{"current_steps": 1739, "total_steps": 2679, "loss": 0.6191, "learning_rate": 1.3943888639353866e-06, "epoch": 1.9457342657342658, "percentage": 64.91, "elapsed_time": "1 day, 4:27:42", "remaining_time": "15:23:05", "throughput": "17796.42", "total_tokens": 1823473664}
{"current_steps": 1740, "total_steps": 2679, "loss": 0.6907, "learning_rate": 1.391735459277434e-06, "epoch": 1.946853146853147, "percentage": 64.95, "elapsed_time": "1 day, 4:28:42", "remaining_time": "15:22:06", "throughput": "17796.30", "total_tokens": 1824522240}
{"current_steps": 1741, "total_steps": 2679, "loss": 0.7179, "learning_rate": 1.3890836075116343e-06, "epoch": 1.9479720279720278, "percentage": 64.99, "elapsed_time": "1 day, 4:29:42", "remaining_time": "15:21:08", "throughput": "17796.20", "total_tokens": 1825570816}
{"current_steps": 1742, "total_steps": 2679, "loss": 0.7048, "learning_rate": 1.3864333123537446e-06, "epoch": 1.9490909090909092, "percentage": 65.02, "elapsed_time": "1 day, 4:30:47", "remaining_time": "15:20:12", "throughput": "17795.09", "total_tokens": 1826619392}
{"current_steps": 1743, "total_steps": 2679, "loss": 0.6714, "learning_rate": 1.3837845775173375e-06, "epoch": 1.9502097902097901, "percentage": 65.06, "elapsed_time": "1 day, 4:31:40", "remaining_time": "15:19:10", "throughput": "17796.06", "total_tokens": 1827667968}
{"current_steps": 1744, "total_steps": 2679, "loss": 0.6153, "learning_rate": 1.3811374067138016e-06, "epoch": 1.9513286713286715, "percentage": 65.1, "elapsed_time": "1 day, 4:32:34", "remaining_time": "15:18:09", "throughput": "17796.88", "total_tokens": 1828716544}
{"current_steps": 1745, "total_steps": 2679, "loss": 0.6374, "learning_rate": 1.3784918036523346e-06, "epoch": 1.9524475524475524, "percentage": 65.14, "elapsed_time": "1 day, 4:33:29", "remaining_time": "15:17:08", "throughput": "17797.55", "total_tokens": 1829765120}
{"current_steps": 1746, "total_steps": 2679, "loss": 0.6838, "learning_rate": 1.3758477720399339e-06, "epoch": 1.9535664335664336, "percentage": 65.17, "elapsed_time": "1 day, 4:34:29", "remaining_time": "15:16:09", "throughput": "17797.42", "total_tokens": 1830813696}
{"current_steps": 1747, "total_steps": 2679, "loss": 0.6767, "learning_rate": 1.3732053155813987e-06, "epoch": 1.9546853146853147, "percentage": 65.21, "elapsed_time": "1 day, 4:35:24", "remaining_time": "15:15:08", "throughput": "17798.15", "total_tokens": 1831862272}
{"current_steps": 1748, "total_steps": 2679, "loss": 0.6895, "learning_rate": 1.3705644379793182e-06, "epoch": 1.9558041958041958, "percentage": 65.25, "elapsed_time": "1 day, 4:36:23", "remaining_time": "15:14:09", "throughput": "17798.09", "total_tokens": 1832910848}
{"current_steps": 1749, "total_steps": 2679, "loss": 0.659, "learning_rate": 1.3679251429340717e-06, "epoch": 1.956923076923077, "percentage": 65.29, "elapsed_time": "1 day, 4:37:23", "remaining_time": "15:13:11", "throughput": "17797.96", "total_tokens": 1833959424}
{"current_steps": 1750, "total_steps": 2679, "loss": 0.6489, "learning_rate": 1.3652874341438203e-06, "epoch": 1.958041958041958, "percentage": 65.32, "elapsed_time": "1 day, 4:38:18", "remaining_time": "15:12:10", "throughput": "17798.58", "total_tokens": 1835008000}
{"current_steps": 1750, "total_steps": 2679, "eval_loss": 0.7199289202690125, "epoch": 1.958041958041958, "percentage": 65.32, "elapsed_time": "1 day, 4:42:25", "remaining_time": "15:14:21", "throughput": "17756.08", "total_tokens": 1835008000}
{"current_steps": 1751, "total_steps": 2679, "loss": 0.7072, "learning_rate": 1.3626513153045024e-06, "epoch": 1.9591608391608393, "percentage": 65.36, "elapsed_time": "1 day, 4:43:50", "remaining_time": "15:13:36", "throughput": "17751.67", "total_tokens": 1836056576}
{"current_steps": 1752, "total_steps": 2679, "loss": 0.7325, "learning_rate": 1.3600167901098282e-06, "epoch": 1.9602797202797202, "percentage": 65.4, "elapsed_time": "1 day, 4:44:50", "remaining_time": "15:12:37", "throughput": "17751.51", "total_tokens": 1837105152}
{"current_steps": 1753, "total_steps": 2679, "loss": 0.6664, "learning_rate": 1.3573838622512743e-06, "epoch": 1.9613986013986016, "percentage": 65.43, "elapsed_time": "1 day, 4:45:44", "remaining_time": "15:11:35", "throughput": "17752.39", "total_tokens": 1838153728}
{"current_steps": 1754, "total_steps": 2679, "loss": 0.6499, "learning_rate": 1.3547525354180796e-06, "epoch": 1.9625174825174825, "percentage": 65.47, "elapsed_time": "1 day, 4:46:37", "remaining_time": "15:10:33", "throughput": "17753.35", "total_tokens": 1839202304}
{"current_steps": 1755, "total_steps": 2679, "loss": 0.7961, "learning_rate": 1.3521228132972414e-06, "epoch": 1.9636363636363636, "percentage": 65.51, "elapsed_time": "1 day, 4:47:43", "remaining_time": "15:09:38", "throughput": "17752.17", "total_tokens": 1840250880}
{"current_steps": 1756, "total_steps": 2679, "loss": 0.717, "learning_rate": 1.3494946995735075e-06, "epoch": 1.9647552447552448, "percentage": 65.55, "elapsed_time": "1 day, 4:48:43", "remaining_time": "15:08:39", "throughput": "17752.03", "total_tokens": 1841299456}
{"current_steps": 1757, "total_steps": 2679, "loss": 0.6987, "learning_rate": 1.3468681979293702e-06, "epoch": 1.965874125874126, "percentage": 65.58, "elapsed_time": "1 day, 4:49:42", "remaining_time": "15:07:40", "throughput": "17751.93", "total_tokens": 1842348032}
{"current_steps": 1758, "total_steps": 2679, "loss": 0.81, "learning_rate": 1.3442433120450642e-06, "epoch": 1.966993006993007, "percentage": 65.62, "elapsed_time": "1 day, 4:50:48", "remaining_time": "15:06:45", "throughput": "17750.83", "total_tokens": 1843396608}
{"current_steps": 1759, "total_steps": 2679, "loss": 0.6939, "learning_rate": 1.3416200455985607e-06, "epoch": 1.968111888111888, "percentage": 65.66, "elapsed_time": "1 day, 4:51:45", "remaining_time": "15:05:44", "throughput": "17751.23", "total_tokens": 1844445184}
{"current_steps": 1760, "total_steps": 2679, "loss": 0.7493, "learning_rate": 1.3389984022655617e-06, "epoch": 1.9692307692307693, "percentage": 65.7, "elapsed_time": "1 day, 4:52:45", "remaining_time": "15:04:46", "throughput": "17751.02", "total_tokens": 1845493760}
{"current_steps": 1761, "total_steps": 2679, "loss": 0.6629, "learning_rate": 1.3363783857194957e-06, "epoch": 1.9703496503496503, "percentage": 65.73, "elapsed_time": "1 day, 4:53:39", "remaining_time": "15:03:44", "throughput": "17751.88", "total_tokens": 1846542336}
{"current_steps": 1762, "total_steps": 2679, "loss": 0.6622, "learning_rate": 1.3337599996315087e-06, "epoch": 1.9714685314685316, "percentage": 65.77, "elapsed_time": "1 day, 4:54:32", "remaining_time": "15:02:42", "throughput": "17752.87", "total_tokens": 1847590912}
{"current_steps": 1763, "total_steps": 2679, "loss": 0.6485, "learning_rate": 1.3311432476704655e-06, "epoch": 1.9725874125874125, "percentage": 65.81, "elapsed_time": "1 day, 4:55:32", "remaining_time": "15:01:43", "throughput": "17752.78", "total_tokens": 1848639488}
{"current_steps": 1764, "total_steps": 2679, "loss": 0.7794, "learning_rate": 1.3285281335029387e-06, "epoch": 1.9737062937062937, "percentage": 65.85, "elapsed_time": "1 day, 4:56:32", "remaining_time": "15:00:45", "throughput": "17752.60", "total_tokens": 1849688064}
{"current_steps": 1765, "total_steps": 2679, "loss": 0.812, "learning_rate": 1.325914660793207e-06, "epoch": 1.9748251748251748, "percentage": 65.88, "elapsed_time": "1 day, 4:57:38", "remaining_time": "14:59:50", "throughput": "17751.40", "total_tokens": 1850736640}
{"current_steps": 1766, "total_steps": 2679, "loss": 0.6568, "learning_rate": 1.3233028332032516e-06, "epoch": 1.975944055944056, "percentage": 65.92, "elapsed_time": "1 day, 4:58:38", "remaining_time": "14:58:51", "throughput": "17751.21", "total_tokens": 1851785216}
{"current_steps": 1767, "total_steps": 2679, "loss": 0.6246, "learning_rate": 1.3206926543927435e-06, "epoch": 1.9770629370629371, "percentage": 65.96, "elapsed_time": "1 day, 4:59:32", "remaining_time": "14:57:49", "throughput": "17752.07", "total_tokens": 1852833792}
{"current_steps": 1768, "total_steps": 2679, "loss": 0.6539, "learning_rate": 1.3180841280190476e-06, "epoch": 1.978181818181818, "percentage": 65.99, "elapsed_time": "1 day, 5:00:26", "remaining_time": "14:56:47", "throughput": "17752.99", "total_tokens": 1853882368}
{"current_steps": 1769, "total_steps": 2679, "loss": 0.6982, "learning_rate": 1.3154772577372104e-06, "epoch": 1.9793006993006994, "percentage": 66.03, "elapsed_time": "1 day, 5:01:25", "remaining_time": "14:55:48", "throughput": "17753.00", "total_tokens": 1854930944}
{"current_steps": 1770, "total_steps": 2679, "loss": 0.6398, "learning_rate": 1.3128720471999606e-06, "epoch": 1.9804195804195803, "percentage": 66.07, "elapsed_time": "1 day, 5:02:18", "remaining_time": "14:54:46", "throughput": "17754.06", "total_tokens": 1855979520}
{"current_steps": 1771, "total_steps": 2679, "loss": 0.6829, "learning_rate": 1.310268500057701e-06, "epoch": 1.9815384615384617, "percentage": 66.11, "elapsed_time": "1 day, 5:03:12", "remaining_time": "14:53:44", "throughput": "17754.93", "total_tokens": 1857028096}
{"current_steps": 1772, "total_steps": 2679, "loss": 0.6814, "learning_rate": 1.307666619958501e-06, "epoch": 1.9826573426573426, "percentage": 66.14, "elapsed_time": "1 day, 5:04:11", "remaining_time": "14:52:46", "throughput": "17754.86", "total_tokens": 1858076672}
{"current_steps": 1773, "total_steps": 2679, "loss": 0.5909, "learning_rate": 1.305066410548097e-06, "epoch": 1.9837762237762238, "percentage": 66.18, "elapsed_time": "1 day, 5:05:00", "remaining_time": "14:51:41", "throughput": "17756.57", "total_tokens": 1859125248}
{"current_steps": 1774, "total_steps": 2679, "loss": 0.6712, "learning_rate": 1.3024678754698827e-06, "epoch": 1.984895104895105, "percentage": 66.22, "elapsed_time": "1 day, 5:05:58", "remaining_time": "14:50:42", "throughput": "17756.83", "total_tokens": 1860173824}
{"current_steps": 1775, "total_steps": 2679, "loss": 0.5639, "learning_rate": 1.2998710183649066e-06, "epoch": 1.986013986013986, "percentage": 66.26, "elapsed_time": "1 day, 5:06:47", "remaining_time": "14:49:37", "throughput": "17758.51", "total_tokens": 1861222400}
{"current_steps": 1776, "total_steps": 2679, "loss": 0.6183, "learning_rate": 1.2972758428718668e-06, "epoch": 1.9871328671328672, "percentage": 66.29, "elapsed_time": "1 day, 5:07:35", "remaining_time": "14:48:33", "throughput": "17760.40", "total_tokens": 1862270976}
{"current_steps": 1777, "total_steps": 2679, "loss": 0.7198, "learning_rate": 1.2946823526271023e-06, "epoch": 1.988251748251748, "percentage": 66.33, "elapsed_time": "1 day, 5:08:40", "remaining_time": "14:47:37", "throughput": "17759.36", "total_tokens": 1863319552}
{"current_steps": 1778, "total_steps": 2679, "loss": 0.6938, "learning_rate": 1.292090551264595e-06, "epoch": 1.9893706293706295, "percentage": 66.37, "elapsed_time": "1 day, 5:09:34", "remaining_time": "14:46:35", "throughput": "17760.21", "total_tokens": 1864368128}
{"current_steps": 1779, "total_steps": 2679, "loss": 0.7327, "learning_rate": 1.2895004424159557e-06, "epoch": 1.9904895104895104, "percentage": 66.41, "elapsed_time": "1 day, 5:10:30", "remaining_time": "14:45:35", "throughput": "17760.63", "total_tokens": 1865416704}
{"current_steps": 1780, "total_steps": 2679, "loss": 0.5803, "learning_rate": 1.286912029710427e-06, "epoch": 1.9916083916083918, "percentage": 66.44, "elapsed_time": "1 day, 5:11:18", "remaining_time": "14:44:30", "throughput": "17762.52", "total_tokens": 1866465280}
{"current_steps": 1781, "total_steps": 2679, "loss": 0.7363, "learning_rate": 1.2843253167748745e-06, "epoch": 1.9927272727272727, "percentage": 66.48, "elapsed_time": "1 day, 5:12:19", "remaining_time": "14:43:32", "throughput": "17762.31", "total_tokens": 1867513856}
{"current_steps": 1782, "total_steps": 2679, "loss": 0.7181, "learning_rate": 1.2817403072337798e-06, "epoch": 1.9938461538461538, "percentage": 66.52, "elapsed_time": "1 day, 5:13:24", "remaining_time": "14:42:36", "throughput": "17761.21", "total_tokens": 1868562432}
{"current_steps": 1783, "total_steps": 2679, "loss": 0.6783, "learning_rate": 1.2791570047092413e-06, "epoch": 1.994965034965035, "percentage": 66.55, "elapsed_time": "1 day, 5:14:23", "remaining_time": "14:41:37", "throughput": "17761.20", "total_tokens": 1869611008}
{"current_steps": 1784, "total_steps": 2679, "loss": 0.6798, "learning_rate": 1.2765754128209614e-06, "epoch": 1.996083916083916, "percentage": 66.59, "elapsed_time": "1 day, 5:15:17", "remaining_time": "14:40:35", "throughput": "17762.08", "total_tokens": 1870659584}
{"current_steps": 1785, "total_steps": 2679, "loss": 0.7397, "learning_rate": 1.2739955351862488e-06, "epoch": 1.9972027972027973, "percentage": 66.63, "elapsed_time": "1 day, 5:16:23", "remaining_time": "14:39:40", "throughput": "17760.98", "total_tokens": 1871708160}
{"current_steps": 1786, "total_steps": 2679, "loss": 0.5976, "learning_rate": 1.2714173754200094e-06, "epoch": 1.9983216783216782, "percentage": 66.67, "elapsed_time": "1 day, 5:17:12", "remaining_time": "14:38:36", "throughput": "17762.66", "total_tokens": 1872756736}
{"current_steps": 1787, "total_steps": 2679, "loss": 0.6467, "learning_rate": 1.2688409371347422e-06, "epoch": 1.9994405594405595, "percentage": 66.7, "elapsed_time": "1 day, 5:18:02", "remaining_time": "14:37:32", "throughput": "17764.18", "total_tokens": 1873805312}
{"current_steps": 1788, "total_steps": 2679, "loss": 0.5426, "learning_rate": 1.266266223940533e-06, "epoch": 2.0005594405594405, "percentage": 66.74, "elapsed_time": "1 day, 5:18:50", "remaining_time": "14:36:28", "throughput": "17765.97", "total_tokens": 1874853888}
{"current_steps": 1789, "total_steps": 2679, "loss": 0.5883, "learning_rate": 1.2636932394450502e-06, "epoch": 2.001678321678322, "percentage": 66.78, "elapsed_time": "1 day, 5:19:44", "remaining_time": "14:35:26", "throughput": "17766.82", "total_tokens": 1875902464}
{"current_steps": 1790, "total_steps": 2679, "loss": 0.7158, "learning_rate": 1.2611219872535412e-06, "epoch": 2.0027972027972027, "percentage": 66.82, "elapsed_time": "1 day, 5:20:44", "remaining_time": "14:34:28", "throughput": "17766.64", "total_tokens": 1876951040}
{"current_steps": 1791, "total_steps": 2679, "loss": 0.6502, "learning_rate": 1.2585524709688268e-06, "epoch": 2.003916083916084, "percentage": 66.85, "elapsed_time": "1 day, 5:21:44", "remaining_time": "14:33:29", "throughput": "17766.58", "total_tokens": 1877999616}
{"current_steps": 1792, "total_steps": 2679, "loss": 0.7371, "learning_rate": 1.2559846941912942e-06, "epoch": 2.005034965034965, "percentage": 66.89, "elapsed_time": "1 day, 5:22:44", "remaining_time": "14:32:31", "throughput": "17766.31", "total_tokens": 1879048192}
{"current_steps": 1793, "total_steps": 2679, "loss": 0.6541, "learning_rate": 1.2534186605188933e-06, "epoch": 2.006153846153846, "percentage": 66.93, "elapsed_time": "1 day, 5:23:43", "remaining_time": "14:31:31", "throughput": "17766.37", "total_tokens": 1880096768}
{"current_steps": 1794, "total_steps": 2679, "loss": 0.7293, "learning_rate": 1.2508543735471305e-06, "epoch": 2.0072727272727273, "percentage": 66.97, "elapsed_time": "1 day, 5:24:43", "remaining_time": "14:30:33", "throughput": "17766.19", "total_tokens": 1881145344}
{"current_steps": 1795, "total_steps": 2679, "loss": 0.7359, "learning_rate": 1.2482918368690666e-06, "epoch": 2.0083916083916082, "percentage": 67.0, "elapsed_time": "1 day, 5:25:48", "remaining_time": "14:29:37", "throughput": "17765.17", "total_tokens": 1882193920}
{"current_steps": 1796, "total_steps": 2679, "loss": 0.5873, "learning_rate": 1.2457310540753093e-06, "epoch": 2.0095104895104896, "percentage": 67.04, "elapsed_time": "1 day, 5:26:42", "remaining_time": "14:28:35", "throughput": "17766.06", "total_tokens": 1883242496}
{"current_steps": 1797, "total_steps": 2679, "loss": 0.6087, "learning_rate": 1.2431720287540097e-06, "epoch": 2.0106293706293705, "percentage": 67.08, "elapsed_time": "1 day, 5:27:38", "remaining_time": "14:27:35", "throughput": "17766.53", "total_tokens": 1884291072}
{"current_steps": 1798, "total_steps": 2679, "loss": 0.7105, "learning_rate": 1.2406147644908537e-06, "epoch": 2.011748251748252, "percentage": 67.11, "elapsed_time": "1 day, 5:28:38", "remaining_time": "14:26:36", "throughput": "17766.38", "total_tokens": 1885339648}
{"current_steps": 1799, "total_steps": 2679, "loss": 0.611, "learning_rate": 1.2380592648690629e-06, "epoch": 2.012867132867133, "percentage": 67.15, "elapsed_time": "1 day, 5:29:32", "remaining_time": "14:25:35", "throughput": "17767.19", "total_tokens": 1886388224}
{"current_steps": 1800, "total_steps": 2679, "loss": 0.8724, "learning_rate": 1.2355055334693828e-06, "epoch": 2.013986013986014, "percentage": 67.19, "elapsed_time": "1 day, 5:30:32", "remaining_time": "14:24:36", "throughput": "17767.02", "total_tokens": 1887436800}
{"current_steps": 1801, "total_steps": 2679, "loss": 0.7183, "learning_rate": 1.2329535738700838e-06, "epoch": 2.015104895104895, "percentage": 67.23, "elapsed_time": "1 day, 5:31:32", "remaining_time": "14:23:38", "throughput": "17766.88", "total_tokens": 1888485376}
{"current_steps": 1802, "total_steps": 2679, "loss": 0.7514, "learning_rate": 1.2304033896469543e-06, "epoch": 2.016223776223776, "percentage": 67.26, "elapsed_time": "1 day, 5:32:33", "remaining_time": "14:22:40", "throughput": "17766.59", "total_tokens": 1889533952}
{"current_steps": 1803, "total_steps": 2679, "loss": 0.5897, "learning_rate": 1.2278549843732915e-06, "epoch": 2.0173426573426574, "percentage": 67.3, "elapsed_time": "1 day, 5:33:26", "remaining_time": "14:21:38", "throughput": "17767.51", "total_tokens": 1890582528}
{"current_steps": 1804, "total_steps": 2679, "loss": 0.6102, "learning_rate": 1.2253083616199045e-06, "epoch": 2.0184615384615383, "percentage": 67.34, "elapsed_time": "1 day, 5:34:21", "remaining_time": "14:20:37", "throughput": "17768.27", "total_tokens": 1891631104}
{"current_steps": 1805, "total_steps": 2679, "loss": 0.6717, "learning_rate": 1.2227635249551014e-06, "epoch": 2.0195804195804197, "percentage": 67.38, "elapsed_time": "1 day, 5:35:15", "remaining_time": "14:19:35", "throughput": "17769.09", "total_tokens": 1892679680}
{"current_steps": 1806, "total_steps": 2679, "loss": 0.6703, "learning_rate": 1.2202204779446866e-06, "epoch": 2.0206993006993006, "percentage": 67.41, "elapsed_time": "1 day, 5:36:09", "remaining_time": "14:18:34", "throughput": "17769.88", "total_tokens": 1893728256}
{"current_steps": 1807, "total_steps": 2679, "loss": 0.5805, "learning_rate": 1.2176792241519628e-06, "epoch": 2.021818181818182, "percentage": 67.45, "elapsed_time": "1 day, 5:36:57", "remaining_time": "14:17:30", "throughput": "17771.64", "total_tokens": 1894776832}
{"current_steps": 1808, "total_steps": 2679, "loss": 0.7194, "learning_rate": 1.2151397671377135e-06, "epoch": 2.022937062937063, "percentage": 67.49, "elapsed_time": "1 day, 5:37:57", "remaining_time": "14:16:31", "throughput": "17771.55", "total_tokens": 1895825408}
{"current_steps": 1809, "total_steps": 2679, "loss": 0.7313, "learning_rate": 1.212602110460209e-06, "epoch": 2.0240559440559442, "percentage": 67.53, "elapsed_time": "1 day, 5:39:02", "remaining_time": "14:15:35", "throughput": "17770.56", "total_tokens": 1896873984}
{"current_steps": 1810, "total_steps": 2679, "loss": 0.6724, "learning_rate": 1.2100662576751932e-06, "epoch": 2.025174825174825, "percentage": 67.56, "elapsed_time": "1 day, 5:40:02", "remaining_time": "14:14:36", "throughput": "17770.48", "total_tokens": 1897922560}
{"current_steps": 1811, "total_steps": 2679, "loss": 0.6227, "learning_rate": 1.2075322123358857e-06, "epoch": 2.026293706293706, "percentage": 67.6, "elapsed_time": "1 day, 5:40:56", "remaining_time": "14:13:35", "throughput": "17771.25", "total_tokens": 1898971136}
{"current_steps": 1812, "total_steps": 2679, "loss": 0.6343, "learning_rate": 1.2049999779929733e-06, "epoch": 2.0274125874125875, "percentage": 67.64, "elapsed_time": "1 day, 5:41:50", "remaining_time": "14:12:34", "throughput": "17772.05", "total_tokens": 1900019712}
{"current_steps": 1813, "total_steps": 2679, "loss": 0.6051, "learning_rate": 1.2024695581946016e-06, "epoch": 2.0285314685314684, "percentage": 67.67, "elapsed_time": "1 day, 5:42:44", "remaining_time": "14:11:32", "throughput": "17772.96", "total_tokens": 1901068288}
{"current_steps": 1814, "total_steps": 2679, "loss": 0.6979, "learning_rate": 1.1999409564863793e-06, "epoch": 2.0296503496503497, "percentage": 67.71, "elapsed_time": "1 day, 5:43:37", "remaining_time": "14:10:31", "throughput": "17773.82", "total_tokens": 1902116864}
{"current_steps": 1815, "total_steps": 2679, "loss": 0.6427, "learning_rate": 1.1974141764113617e-06, "epoch": 2.0307692307692307, "percentage": 67.75, "elapsed_time": "1 day, 5:44:32", "remaining_time": "14:09:29", "throughput": "17774.61", "total_tokens": 1903165440}
{"current_steps": 1816, "total_steps": 2679, "loss": 0.7831, "learning_rate": 1.1948892215100557e-06, "epoch": 2.031888111888112, "percentage": 67.79, "elapsed_time": "1 day, 5:45:34", "remaining_time": "14:08:32", "throughput": "17774.13", "total_tokens": 1904214016}
{"current_steps": 1817, "total_steps": 2679, "loss": 0.7386, "learning_rate": 1.192366095320411e-06, "epoch": 2.033006993006993, "percentage": 67.82, "elapsed_time": "1 day, 5:46:39", "remaining_time": "14:07:36", "throughput": "17773.08", "total_tokens": 1905262592}
{"current_steps": 1818, "total_steps": 2679, "loss": 0.8826, "learning_rate": 1.189844801377811e-06, "epoch": 2.0341258741258743, "percentage": 67.86, "elapsed_time": "1 day, 5:47:50", "remaining_time": "14:06:43", "throughput": "17770.99", "total_tokens": 1906311168}
{"current_steps": 1819, "total_steps": 2679, "loss": 0.7653, "learning_rate": 1.1873253432150769e-06, "epoch": 2.0352447552447552, "percentage": 67.9, "elapsed_time": "1 day, 5:48:55", "remaining_time": "14:05:46", "throughput": "17770.05", "total_tokens": 1907359744}
{"current_steps": 1820, "total_steps": 2679, "loss": 0.6843, "learning_rate": 1.1848077243624525e-06, "epoch": 2.036363636363636, "percentage": 67.94, "elapsed_time": "1 day, 5:49:51", "remaining_time": "14:04:46", "throughput": "17770.54", "total_tokens": 1908408320}
{"current_steps": 1821, "total_steps": 2679, "loss": 0.6073, "learning_rate": 1.1822919483476089e-06, "epoch": 2.0374825174825175, "percentage": 67.97, "elapsed_time": "1 day, 5:50:46", "remaining_time": "14:03:45", "throughput": "17771.30", "total_tokens": 1909456896}
{"current_steps": 1822, "total_steps": 2679, "loss": 0.6871, "learning_rate": 1.1797780186956307e-06, "epoch": 2.0386013986013984, "percentage": 68.01, "elapsed_time": "1 day, 5:51:45", "remaining_time": "14:02:46", "throughput": "17771.26", "total_tokens": 1910505472}
{"current_steps": 1823, "total_steps": 2679, "loss": 0.705, "learning_rate": 1.177265938929021e-06, "epoch": 2.03972027972028, "percentage": 68.05, "elapsed_time": "1 day, 5:52:50", "remaining_time": "14:01:50", "throughput": "17770.27", "total_tokens": 1911554048}
{"current_steps": 1824, "total_steps": 2679, "loss": 0.7486, "learning_rate": 1.1747557125676853e-06, "epoch": 2.0408391608391607, "percentage": 68.09, "elapsed_time": "1 day, 5:53:51", "remaining_time": "14:00:52", "throughput": "17769.99", "total_tokens": 1912602624}
{"current_steps": 1825, "total_steps": 2679, "loss": 0.6472, "learning_rate": 1.1722473431289344e-06, "epoch": 2.041958041958042, "percentage": 68.12, "elapsed_time": "1 day, 5:54:45", "remaining_time": "13:59:50", "throughput": "17770.74", "total_tokens": 1913651200}
{"current_steps": 1826, "total_steps": 2679, "loss": 0.7062, "learning_rate": 1.1697408341274781e-06, "epoch": 2.043076923076923, "percentage": 68.16, "elapsed_time": "1 day, 5:55:51", "remaining_time": "13:58:54", "throughput": "17769.66", "total_tokens": 1914699776}
{"current_steps": 1827, "total_steps": 2679, "loss": 0.6114, "learning_rate": 1.1672361890754165e-06, "epoch": 2.0441958041958044, "percentage": 68.2, "elapsed_time": "1 day, 5:56:44", "remaining_time": "13:57:53", "throughput": "17770.61", "total_tokens": 1915748352}
{"current_steps": 1828, "total_steps": 2679, "loss": 0.8218, "learning_rate": 1.1647334114822434e-06, "epoch": 2.0453146853146853, "percentage": 68.23, "elapsed_time": "1 day, 5:57:50", "remaining_time": "13:56:57", "throughput": "17769.42", "total_tokens": 1916796928}
{"current_steps": 1829, "total_steps": 2679, "loss": 0.6464, "learning_rate": 1.1622325048548303e-06, "epoch": 2.046433566433566, "percentage": 68.27, "elapsed_time": "1 day, 5:58:51", "remaining_time": "13:55:59", "throughput": "17769.17", "total_tokens": 1917845504}
{"current_steps": 1830, "total_steps": 2679, "loss": 0.6844, "learning_rate": 1.159733472697428e-06, "epoch": 2.0475524475524476, "percentage": 68.31, "elapsed_time": "1 day, 5:59:50", "remaining_time": "13:55:00", "throughput": "17769.10", "total_tokens": 1918894080}
{"current_steps": 1831, "total_steps": 2679, "loss": 0.7415, "learning_rate": 1.1572363185116648e-06, "epoch": 2.0486713286713285, "percentage": 68.35, "elapsed_time": "1 day, 6:00:55", "remaining_time": "13:54:04", "throughput": "17768.07", "total_tokens": 1919942656}
{"current_steps": 1832, "total_steps": 2679, "loss": 0.5814, "learning_rate": 1.1547410457965314e-06, "epoch": 2.04979020979021, "percentage": 68.38, "elapsed_time": "1 day, 6:01:43", "remaining_time": "13:53:00", "throughput": "17769.97", "total_tokens": 1920991232}
{"current_steps": 1833, "total_steps": 2679, "loss": 0.7473, "learning_rate": 1.1522476580483893e-06, "epoch": 2.050909090909091, "percentage": 68.42, "elapsed_time": "1 day, 6:02:46", "remaining_time": "13:52:02", "throughput": "17769.34", "total_tokens": 1922039808}
{"current_steps": 1834, "total_steps": 2679, "loss": 0.7224, "learning_rate": 1.149756158760953e-06, "epoch": 2.052027972027972, "percentage": 68.46, "elapsed_time": "1 day, 6:03:51", "remaining_time": "13:51:06", "throughput": "17768.28", "total_tokens": 1923088384}
{"current_steps": 1835, "total_steps": 2679, "loss": 0.5895, "learning_rate": 1.1472665514252943e-06, "epoch": 2.053146853146853, "percentage": 68.5, "elapsed_time": "1 day, 6:04:39", "remaining_time": "13:50:02", "throughput": "17770.05", "total_tokens": 1924136960}
{"current_steps": 1836, "total_steps": 2679, "loss": 0.6445, "learning_rate": 1.1447788395298318e-06, "epoch": 2.0542657342657344, "percentage": 68.53, "elapsed_time": "1 day, 6:05:39", "remaining_time": "13:49:04", "throughput": "17769.94", "total_tokens": 1925185536}
{"current_steps": 1837, "total_steps": 2679, "loss": 0.6205, "learning_rate": 1.142293026560328e-06, "epoch": 2.0553846153846154, "percentage": 68.57, "elapsed_time": "1 day, 6:06:33", "remaining_time": "13:48:02", "throughput": "17770.79", "total_tokens": 1926234112}
{"current_steps": 1838, "total_steps": 2679, "loss": 0.7758, "learning_rate": 1.1398091159998887e-06, "epoch": 2.0565034965034963, "percentage": 68.61, "elapsed_time": "1 day, 6:07:39", "remaining_time": "13:47:06", "throughput": "17769.66", "total_tokens": 1927282688}
{"current_steps": 1839, "total_steps": 2679, "loss": 0.5974, "learning_rate": 1.137327111328949e-06, "epoch": 2.0576223776223777, "percentage": 68.65, "elapsed_time": "1 day, 6:08:41", "remaining_time": "13:46:09", "throughput": "17769.11", "total_tokens": 1928331264}
{"current_steps": 1840, "total_steps": 2679, "loss": 0.6062, "learning_rate": 1.1348470160252772e-06, "epoch": 2.0587412587412586, "percentage": 68.68, "elapsed_time": "1 day, 6:09:35", "remaining_time": "13:45:08", "throughput": "17769.95", "total_tokens": 1929379840}
{"current_steps": 1841, "total_steps": 2679, "loss": 0.6752, "learning_rate": 1.1323688335639637e-06, "epoch": 2.05986013986014, "percentage": 68.72, "elapsed_time": "1 day, 6:10:34", "remaining_time": "13:44:09", "throughput": "17769.88", "total_tokens": 1930428416}
{"current_steps": 1842, "total_steps": 2679, "loss": 0.6068, "learning_rate": 1.1298925674174191e-06, "epoch": 2.060979020979021, "percentage": 68.76, "elapsed_time": "1 day, 6:11:28", "remaining_time": "13:43:07", "throughput": "17770.70", "total_tokens": 1931476992}
{"current_steps": 1843, "total_steps": 2679, "loss": 0.7561, "learning_rate": 1.12741822105537e-06, "epoch": 2.0620979020979022, "percentage": 68.79, "elapsed_time": "1 day, 6:12:34", "remaining_time": "13:42:12", "throughput": "17769.58", "total_tokens": 1932525568}
{"current_steps": 1844, "total_steps": 2679, "loss": 0.6795, "learning_rate": 1.1249457979448521e-06, "epoch": 2.063216783216783, "percentage": 68.83, "elapsed_time": "1 day, 6:13:39", "remaining_time": "13:41:15", "throughput": "17768.59", "total_tokens": 1933574144}
{"current_steps": 1845, "total_steps": 2679, "loss": 0.7066, "learning_rate": 1.122475301550208e-06, "epoch": 2.0643356643356645, "percentage": 68.87, "elapsed_time": "1 day, 6:14:39", "remaining_time": "13:40:17", "throughput": "17768.44", "total_tokens": 1934622720}
{"current_steps": 1846, "total_steps": 2679, "loss": 0.5849, "learning_rate": 1.120006735333078e-06, "epoch": 2.0654545454545454, "percentage": 68.91, "elapsed_time": "1 day, 6:15:33", "remaining_time": "13:39:15", "throughput": "17769.23", "total_tokens": 1935671296}
{"current_steps": 1847, "total_steps": 2679, "loss": 0.6686, "learning_rate": 1.117540102752398e-06, "epoch": 2.0665734265734264, "percentage": 68.94, "elapsed_time": "1 day, 6:16:33", "remaining_time": "13:38:17", "throughput": "17769.13", "total_tokens": 1936719872}
{"current_steps": 1848, "total_steps": 2679, "loss": 0.6698, "learning_rate": 1.1150754072643966e-06, "epoch": 2.0676923076923077, "percentage": 68.98, "elapsed_time": "1 day, 6:17:33", "remaining_time": "13:37:18", "throughput": "17769.03", "total_tokens": 1937768448}
{"current_steps": 1849, "total_steps": 2679, "loss": 0.6424, "learning_rate": 1.1126126523225869e-06, "epoch": 2.0688111888111886, "percentage": 69.02, "elapsed_time": "1 day, 6:18:27", "remaining_time": "13:36:17", "throughput": "17769.78", "total_tokens": 1938817024}
{"current_steps": 1850, "total_steps": 2679, "loss": 0.6023, "learning_rate": 1.110151841377764e-06, "epoch": 2.06993006993007, "percentage": 69.06, "elapsed_time": "1 day, 6:19:27", "remaining_time": "13:35:18", "throughput": "17769.65", "total_tokens": 1939865600}
{"current_steps": 1851, "total_steps": 2679, "loss": 0.6791, "learning_rate": 1.1076929778779965e-06, "epoch": 2.071048951048951, "percentage": 69.09, "elapsed_time": "1 day, 6:20:21", "remaining_time": "13:34:17", "throughput": "17770.45", "total_tokens": 1940914176}
{"current_steps": 1852, "total_steps": 2679, "loss": 0.6703, "learning_rate": 1.1052360652686275e-06, "epoch": 2.0721678321678323, "percentage": 69.13, "elapsed_time": "1 day, 6:21:21", "remaining_time": "13:33:19", "throughput": "17770.29", "total_tokens": 1941962752}
{"current_steps": 1853, "total_steps": 2679, "loss": 0.7182, "learning_rate": 1.1027811069922634e-06, "epoch": 2.073286713286713, "percentage": 69.17, "elapsed_time": "1 day, 6:22:15", "remaining_time": "13:32:17", "throughput": "17771.07", "total_tokens": 1943011328}
{"current_steps": 1854, "total_steps": 2679, "loss": 0.6414, "learning_rate": 1.1003281064887744e-06, "epoch": 2.0744055944055946, "percentage": 69.2, "elapsed_time": "1 day, 6:23:14", "remaining_time": "13:31:18", "throughput": "17771.07", "total_tokens": 1944059904}
{"current_steps": 1855, "total_steps": 2679, "loss": 0.679, "learning_rate": 1.097877067195288e-06, "epoch": 2.0755244755244755, "percentage": 69.24, "elapsed_time": "1 day, 6:24:14", "remaining_time": "13:30:20", "throughput": "17770.96", "total_tokens": 1945108480}
{"current_steps": 1856, "total_steps": 2679, "loss": 0.7014, "learning_rate": 1.0954279925461802e-06, "epoch": 2.0766433566433564, "percentage": 69.28, "elapsed_time": "1 day, 6:25:13", "remaining_time": "13:29:21", "throughput": "17770.89", "total_tokens": 1946157056}
{"current_steps": 1857, "total_steps": 2679, "loss": 0.7407, "learning_rate": 1.092980885973079e-06, "epoch": 2.077762237762238, "percentage": 69.32, "elapsed_time": "1 day, 6:26:13", "remaining_time": "13:28:22", "throughput": "17770.73", "total_tokens": 1947205632}
{"current_steps": 1858, "total_steps": 2679, "loss": 0.6654, "learning_rate": 1.0905357509048498e-06, "epoch": 2.0788811188811187, "percentage": 69.35, "elapsed_time": "1 day, 6:27:13", "remaining_time": "13:27:23", "throughput": "17770.68", "total_tokens": 1948254208}
{"current_steps": 1859, "total_steps": 2679, "loss": 0.6313, "learning_rate": 1.088092590767599e-06, "epoch": 2.08, "percentage": 69.39, "elapsed_time": "1 day, 6:28:06", "remaining_time": "13:26:22", "throughput": "17771.53", "total_tokens": 1949302784}
{"current_steps": 1860, "total_steps": 2679, "loss": 0.6359, "learning_rate": 1.0856514089846656e-06, "epoch": 2.081118881118881, "percentage": 69.43, "elapsed_time": "1 day, 6:28:56", "remaining_time": "13:25:19", "throughput": "17773.08", "total_tokens": 1950351360}
{"current_steps": 1861, "total_steps": 2679, "loss": 0.6465, "learning_rate": 1.0832122089766143e-06, "epoch": 2.0822377622377624, "percentage": 69.47, "elapsed_time": "1 day, 6:29:55", "remaining_time": "13:24:20", "throughput": "17773.02", "total_tokens": 1951399936}
{"current_steps": 1862, "total_steps": 2679, "loss": 0.6765, "learning_rate": 1.080774994161235e-06, "epoch": 2.0833566433566433, "percentage": 69.5, "elapsed_time": "1 day, 6:30:50", "remaining_time": "13:23:19", "throughput": "17773.76", "total_tokens": 1952448512}
{"current_steps": 1863, "total_steps": 2679, "loss": 0.8147, "learning_rate": 1.0783397679535343e-06, "epoch": 2.0844755244755246, "percentage": 69.54, "elapsed_time": "1 day, 6:32:01", "remaining_time": "13:22:25", "throughput": "17771.81", "total_tokens": 1953497088}
{"current_steps": 1864, "total_steps": 2679, "loss": 0.7105, "learning_rate": 1.075906533765734e-06, "epoch": 2.0855944055944056, "percentage": 69.58, "elapsed_time": "1 day, 6:33:01", "remaining_time": "13:21:27", "throughput": "17771.62", "total_tokens": 1954545664}
{"current_steps": 1865, "total_steps": 2679, "loss": 0.69, "learning_rate": 1.073475295007265e-06, "epoch": 2.0867132867132865, "percentage": 69.62, "elapsed_time": "1 day, 6:34:00", "remaining_time": "13:20:28", "throughput": "17771.55", "total_tokens": 1955594240}
{"current_steps": 1866, "total_steps": 2679, "loss": 0.5658, "learning_rate": 1.0710460550847593e-06, "epoch": 2.087832167832168, "percentage": 69.65, "elapsed_time": "1 day, 6:34:49", "remaining_time": "13:19:25", "throughput": "17773.17", "total_tokens": 1956642816}
{"current_steps": 1867, "total_steps": 2679, "loss": 0.6637, "learning_rate": 1.068618817402052e-06, "epoch": 2.0889510489510488, "percentage": 69.69, "elapsed_time": "1 day, 6:35:49", "remaining_time": "13:18:26", "throughput": "17773.06", "total_tokens": 1957691392}
{"current_steps": 1868, "total_steps": 2679, "loss": 0.6657, "learning_rate": 1.0661935853601688e-06, "epoch": 2.09006993006993, "percentage": 69.73, "elapsed_time": "1 day, 6:36:49", "remaining_time": "13:17:28", "throughput": "17772.81", "total_tokens": 1958739968}
{"current_steps": 1869, "total_steps": 2679, "loss": 0.6223, "learning_rate": 1.0637703623573278e-06, "epoch": 2.091188811188811, "percentage": 69.76, "elapsed_time": "1 day, 6:37:45", "remaining_time": "13:16:27", "throughput": "17773.37", "total_tokens": 1959788544}
{"current_steps": 1870, "total_steps": 2679, "loss": 0.6222, "learning_rate": 1.0613491517889326e-06, "epoch": 2.0923076923076924, "percentage": 69.8, "elapsed_time": "1 day, 6:38:39", "remaining_time": "13:15:26", "throughput": "17774.14", "total_tokens": 1960837120}
{"current_steps": 1871, "total_steps": 2679, "loss": 0.5786, "learning_rate": 1.058929957047564e-06, "epoch": 2.0934265734265733, "percentage": 69.84, "elapsed_time": "1 day, 6:39:28", "remaining_time": "13:14:23", "throughput": "17775.80", "total_tokens": 1961885696}
{"current_steps": 1872, "total_steps": 2679, "loss": 0.7348, "learning_rate": 1.0565127815229815e-06, "epoch": 2.0945454545454547, "percentage": 69.88, "elapsed_time": "1 day, 6:40:28", "remaining_time": "13:13:24", "throughput": "17775.65", "total_tokens": 1962934272}
{"current_steps": 1873, "total_steps": 2679, "loss": 0.7973, "learning_rate": 1.0540976286021115e-06, "epoch": 2.0956643356643356, "percentage": 69.91, "elapsed_time": "1 day, 6:41:28", "remaining_time": "13:12:26", "throughput": "17775.38", "total_tokens": 1963982848}
{"current_steps": 1874, "total_steps": 2679, "loss": 0.7, "learning_rate": 1.0516845016690502e-06, "epoch": 2.096783216783217, "percentage": 69.95, "elapsed_time": "1 day, 6:42:28", "remaining_time": "13:11:27", "throughput": "17775.22", "total_tokens": 1965031424}
{"current_steps": 1875, "total_steps": 2679, "loss": 0.6746, "learning_rate": 1.0492734041050532e-06, "epoch": 2.097902097902098, "percentage": 69.99, "elapsed_time": "1 day, 6:43:28", "remaining_time": "13:10:28", "throughput": "17775.19", "total_tokens": 1966080000}
{"current_steps": 1876, "total_steps": 2679, "loss": 0.663, "learning_rate": 1.0468643392885335e-06, "epoch": 2.099020979020979, "percentage": 70.03, "elapsed_time": "1 day, 6:44:27", "remaining_time": "13:09:30", "throughput": "17775.09", "total_tokens": 1967128576}
{"current_steps": 1877, "total_steps": 2679, "loss": 0.6262, "learning_rate": 1.0444573105950543e-06, "epoch": 2.10013986013986, "percentage": 70.06, "elapsed_time": "1 day, 6:45:21", "remaining_time": "13:08:28", "throughput": "17775.87", "total_tokens": 1968177152}
{"current_steps": 1878, "total_steps": 2679, "loss": 0.6432, "learning_rate": 1.0420523213973253e-06, "epoch": 2.101258741258741, "percentage": 70.1, "elapsed_time": "1 day, 6:46:18", "remaining_time": "13:07:28", "throughput": "17776.24", "total_tokens": 1969225728}
{"current_steps": 1879, "total_steps": 2679, "loss": 0.5529, "learning_rate": 1.0396493750652008e-06, "epoch": 2.1023776223776225, "percentage": 70.14, "elapsed_time": "1 day, 6:47:12", "remaining_time": "13:06:27", "throughput": "17777.08", "total_tokens": 1970274304}
{"current_steps": 1880, "total_steps": 2679, "loss": 0.6106, "learning_rate": 1.0372484749656723e-06, "epoch": 2.1034965034965034, "percentage": 70.18, "elapsed_time": "1 day, 6:48:05", "remaining_time": "13:05:26", "throughput": "17777.95", "total_tokens": 1971322880}
{"current_steps": 1881, "total_steps": 2679, "loss": 0.6082, "learning_rate": 1.0348496244628633e-06, "epoch": 2.1046153846153848, "percentage": 70.21, "elapsed_time": "1 day, 6:48:59", "remaining_time": "13:04:25", "throughput": "17778.73", "total_tokens": 1972371456}
{"current_steps": 1882, "total_steps": 2679, "loss": 0.6217, "learning_rate": 1.0324528269180252e-06, "epoch": 2.1057342657342657, "percentage": 70.25, "elapsed_time": "1 day, 6:49:53", "remaining_time": "13:03:24", "throughput": "17779.56", "total_tokens": 1973420032}
{"current_steps": 1883, "total_steps": 2679, "loss": 0.6896, "learning_rate": 1.030058085689532e-06, "epoch": 2.106853146853147, "percentage": 70.29, "elapsed_time": "1 day, 6:50:53", "remaining_time": "13:02:25", "throughput": "17779.48", "total_tokens": 1974468608}
{"current_steps": 1884, "total_steps": 2679, "loss": 0.7581, "learning_rate": 1.0276654041328787e-06, "epoch": 2.107972027972028, "percentage": 70.32, "elapsed_time": "1 day, 6:51:58", "remaining_time": "13:01:29", "throughput": "17778.46", "total_tokens": 1975517184}
{"current_steps": 1885, "total_steps": 2679, "loss": 0.6518, "learning_rate": 1.0252747856006735e-06, "epoch": 2.109090909090909, "percentage": 70.36, "elapsed_time": "1 day, 6:52:52", "remaining_time": "13:00:27", "throughput": "17779.33", "total_tokens": 1976565760}
{"current_steps": 1886, "total_steps": 2679, "loss": 0.688, "learning_rate": 1.0228862334426335e-06, "epoch": 2.1102097902097903, "percentage": 70.4, "elapsed_time": "1 day, 6:53:57", "remaining_time": "12:59:31", "throughput": "17778.35", "total_tokens": 1977614336}
{"current_steps": 1887, "total_steps": 2679, "loss": 0.7086, "learning_rate": 1.0204997510055793e-06, "epoch": 2.111328671328671, "percentage": 70.44, "elapsed_time": "1 day, 6:55:05", "remaining_time": "12:58:36", "throughput": "17776.83", "total_tokens": 1978662912}
{"current_steps": 1888, "total_steps": 2679, "loss": 0.6979, "learning_rate": 1.0181153416334344e-06, "epoch": 2.1124475524475526, "percentage": 70.47, "elapsed_time": "1 day, 6:56:02", "remaining_time": "12:57:36", "throughput": "17777.12", "total_tokens": 1979711488}
{"current_steps": 1889, "total_steps": 2679, "loss": 0.5586, "learning_rate": 1.015733008667214e-06, "epoch": 2.1135664335664335, "percentage": 70.51, "elapsed_time": "1 day, 6:56:50", "remaining_time": "12:56:33", "throughput": "17778.90", "total_tokens": 1980760064}
{"current_steps": 1890, "total_steps": 2679, "loss": 0.5597, "learning_rate": 1.0133527554450262e-06, "epoch": 2.114685314685315, "percentage": 70.55, "elapsed_time": "1 day, 6:57:38", "remaining_time": "12:55:29", "throughput": "17780.65", "total_tokens": 1981808640}
{"current_steps": 1891, "total_steps": 2679, "loss": 0.6179, "learning_rate": 1.0109745853020655e-06, "epoch": 2.1158041958041958, "percentage": 70.59, "elapsed_time": "1 day, 6:58:33", "remaining_time": "12:54:28", "throughput": "17781.38", "total_tokens": 1982857216}
{"current_steps": 1892, "total_steps": 2679, "loss": 0.7409, "learning_rate": 1.0085985015706045e-06, "epoch": 2.116923076923077, "percentage": 70.62, "elapsed_time": "1 day, 6:59:34", "remaining_time": "12:53:30", "throughput": "17781.03", "total_tokens": 1983905792}
{"current_steps": 1893, "total_steps": 2679, "loss": 0.6989, "learning_rate": 1.0062245075799966e-06, "epoch": 2.118041958041958, "percentage": 70.66, "elapsed_time": "1 day, 7:00:32", "remaining_time": "12:52:31", "throughput": "17781.19", "total_tokens": 1984954368}
{"current_steps": 1894, "total_steps": 2679, "loss": 0.6449, "learning_rate": 1.0038526066566624e-06, "epoch": 2.119160839160839, "percentage": 70.7, "elapsed_time": "1 day, 7:01:27", "remaining_time": "12:51:30", "throughput": "17781.73", "total_tokens": 1986002944}
{"current_steps": 1895, "total_steps": 2679, "loss": 0.619, "learning_rate": 1.0014828021240932e-06, "epoch": 2.1202797202797203, "percentage": 70.74, "elapsed_time": "1 day, 7:02:21", "remaining_time": "12:50:29", "throughput": "17782.59", "total_tokens": 1987051520}
{"current_steps": 1896, "total_steps": 2679, "loss": 0.6311, "learning_rate": 9.991150973028428e-07, "epoch": 2.1213986013986013, "percentage": 70.77, "elapsed_time": "1 day, 7:03:15", "remaining_time": "12:49:28", "throughput": "17783.34", "total_tokens": 1988100096}
{"current_steps": 1897, "total_steps": 2679, "loss": 0.7862, "learning_rate": 9.967494955105197e-07, "epoch": 2.1225174825174826, "percentage": 70.81, "elapsed_time": "1 day, 7:04:21", "remaining_time": "12:48:32", "throughput": "17782.25", "total_tokens": 1989148672}
{"current_steps": 1898, "total_steps": 2679, "loss": 0.6566, "learning_rate": 9.9438600006179e-07, "epoch": 2.1236363636363635, "percentage": 70.85, "elapsed_time": "1 day, 7:05:20", "remaining_time": "12:47:33", "throughput": "17782.17", "total_tokens": 1990197248}
{"current_steps": 1899, "total_steps": 2679, "loss": 0.7726, "learning_rate": 9.92024614268364e-07, "epoch": 2.124755244755245, "percentage": 70.88, "elapsed_time": "1 day, 7:06:26", "remaining_time": "12:46:37", "throughput": "17781.17", "total_tokens": 1991245824}
{"current_steps": 1900, "total_steps": 2679, "loss": 0.5821, "learning_rate": 9.896653414389996e-07, "epoch": 2.125874125874126, "percentage": 70.92, "elapsed_time": "1 day, 7:07:20", "remaining_time": "12:45:36", "throughput": "17781.92", "total_tokens": 1992294400}
{"current_steps": 1901, "total_steps": 2679, "loss": 0.6046, "learning_rate": 9.873081848794926e-07, "epoch": 2.126993006993007, "percentage": 70.96, "elapsed_time": "1 day, 7:08:14", "remaining_time": "12:44:35", "throughput": "17782.76", "total_tokens": 1993342976}
{"current_steps": 1902, "total_steps": 2679, "loss": 0.7002, "learning_rate": 9.84953147892673e-07, "epoch": 2.128111888111888, "percentage": 71.0, "elapsed_time": "1 day, 7:09:14", "remaining_time": "12:43:36", "throughput": "17782.57", "total_tokens": 1994391552}
{"current_steps": 1903, "total_steps": 2679, "loss": 0.6932, "learning_rate": 9.82600233778402e-07, "epoch": 2.129230769230769, "percentage": 71.03, "elapsed_time": "1 day, 7:10:10", "remaining_time": "12:42:36", "throughput": "17783.01", "total_tokens": 1995440128}
{"current_steps": 1904, "total_steps": 2679, "loss": 0.6934, "learning_rate": 9.802494458335643e-07, "epoch": 2.1303496503496504, "percentage": 71.07, "elapsed_time": "1 day, 7:11:09", "remaining_time": "12:41:38", "throughput": "17782.95", "total_tokens": 1996488704}
{"current_steps": 1905, "total_steps": 2679, "loss": 0.6303, "learning_rate": 9.77900787352068e-07, "epoch": 2.1314685314685313, "percentage": 71.11, "elapsed_time": "1 day, 7:12:04", "remaining_time": "12:40:37", "throughput": "17783.65", "total_tokens": 1997537280}
{"current_steps": 1906, "total_steps": 2679, "loss": 0.6924, "learning_rate": 9.755542616248361e-07, "epoch": 2.1325874125874127, "percentage": 71.15, "elapsed_time": "1 day, 7:13:04", "remaining_time": "12:39:38", "throughput": "17783.50", "total_tokens": 1998585856}
{"current_steps": 1907, "total_steps": 2679, "loss": 0.6302, "learning_rate": 9.732098719398025e-07, "epoch": 2.1337062937062936, "percentage": 71.18, "elapsed_time": "1 day, 7:13:57", "remaining_time": "12:38:37", "throughput": "17784.34", "total_tokens": 1999634432}
{"current_steps": 1908, "total_steps": 2679, "loss": 0.6511, "learning_rate": 9.708676215819098e-07, "epoch": 2.134825174825175, "percentage": 71.22, "elapsed_time": "1 day, 7:14:51", "remaining_time": "12:37:36", "throughput": "17785.17", "total_tokens": 2000683008}
{"current_steps": 1909, "total_steps": 2679, "loss": 0.6657, "learning_rate": 9.68527513833101e-07, "epoch": 2.135944055944056, "percentage": 71.26, "elapsed_time": "1 day, 7:15:51", "remaining_time": "12:36:37", "throughput": "17785.05", "total_tokens": 2001731584}
{"current_steps": 1910, "total_steps": 2679, "loss": 0.658, "learning_rate": 9.661895519723183e-07, "epoch": 2.1370629370629373, "percentage": 71.3, "elapsed_time": "1 day, 7:16:51", "remaining_time": "12:35:39", "throughput": "17784.80", "total_tokens": 2002780160}
{"current_steps": 1911, "total_steps": 2679, "loss": 0.7099, "learning_rate": 9.638537392754968e-07, "epoch": 2.138181818181818, "percentage": 71.33, "elapsed_time": "1 day, 7:17:51", "remaining_time": "12:34:40", "throughput": "17784.71", "total_tokens": 2003828736}
{"current_steps": 1912, "total_steps": 2679, "loss": 0.7836, "learning_rate": 9.615200790155612e-07, "epoch": 2.139300699300699, "percentage": 71.37, "elapsed_time": "1 day, 7:18:51", "remaining_time": "12:33:42", "throughput": "17784.49", "total_tokens": 2004877312}
{"current_steps": 1913, "total_steps": 2679, "loss": 0.7269, "learning_rate": 9.591885744624183e-07, "epoch": 2.1404195804195805, "percentage": 71.41, "elapsed_time": "1 day, 7:19:52", "remaining_time": "12:32:44", "throughput": "17784.24", "total_tokens": 2005925888}
{"current_steps": 1914, "total_steps": 2679, "loss": 0.7035, "learning_rate": 9.56859228882954e-07, "epoch": 2.1415384615384614, "percentage": 71.44, "elapsed_time": "1 day, 7:20:58", "remaining_time": "12:31:48", "throughput": "17783.06", "total_tokens": 2006974464}
{"current_steps": 1915, "total_steps": 2679, "loss": 0.6048, "learning_rate": 9.54532045541031e-07, "epoch": 2.1426573426573428, "percentage": 71.48, "elapsed_time": "1 day, 7:21:47", "remaining_time": "12:30:45", "throughput": "17784.67", "total_tokens": 2008023040}
{"current_steps": 1916, "total_steps": 2679, "loss": 0.7171, "learning_rate": 9.522070276974823e-07, "epoch": 2.1437762237762237, "percentage": 71.52, "elapsed_time": "1 day, 7:22:47", "remaining_time": "12:29:46", "throughput": "17784.55", "total_tokens": 2009071616}
{"current_steps": 1917, "total_steps": 2679, "loss": 0.7006, "learning_rate": 9.498841786101065e-07, "epoch": 2.144895104895105, "percentage": 71.56, "elapsed_time": "1 day, 7:23:48", "remaining_time": "12:28:48", "throughput": "17784.27", "total_tokens": 2010120192}
{"current_steps": 1918, "total_steps": 2679, "loss": 0.6544, "learning_rate": 9.475635015336612e-07, "epoch": 2.146013986013986, "percentage": 71.59, "elapsed_time": "1 day, 7:24:47", "remaining_time": "12:27:49", "throughput": "17784.13", "total_tokens": 2011168768}
{"current_steps": 1919, "total_steps": 2679, "loss": 0.8538, "learning_rate": 9.45244999719862e-07, "epoch": 2.1471328671328673, "percentage": 71.63, "elapsed_time": "1 day, 7:25:59", "remaining_time": "12:26:55", "throughput": "17782.12", "total_tokens": 2012217344}
{"current_steps": 1920, "total_steps": 2679, "loss": 0.7319, "learning_rate": 9.42928676417377e-07, "epoch": 2.1482517482517482, "percentage": 71.67, "elapsed_time": "1 day, 7:27:03", "remaining_time": "12:25:58", "throughput": "17781.41", "total_tokens": 2013265920}
{"current_steps": 1921, "total_steps": 2679, "loss": 0.7846, "learning_rate": 9.406145348718218e-07, "epoch": 2.149370629370629, "percentage": 71.71, "elapsed_time": "1 day, 7:28:03", "remaining_time": "12:25:00", "throughput": "17781.20", "total_tokens": 2014314496}
{"current_steps": 1922, "total_steps": 2679, "loss": 0.5742, "learning_rate": 9.383025783257554e-07, "epoch": 2.1504895104895105, "percentage": 71.74, "elapsed_time": "1 day, 7:28:57", "remaining_time": "12:23:59", "throughput": "17781.96", "total_tokens": 2015363072}
{"current_steps": 1923, "total_steps": 2679, "loss": 0.5967, "learning_rate": 9.359928100186724e-07, "epoch": 2.1516083916083915, "percentage": 71.78, "elapsed_time": "1 day, 7:29:51", "remaining_time": "12:22:58", "throughput": "17782.76", "total_tokens": 2016411648}
{"current_steps": 1924, "total_steps": 2679, "loss": 0.6586, "learning_rate": 9.336852331870052e-07, "epoch": 2.152727272727273, "percentage": 71.82, "elapsed_time": "1 day, 7:30:51", "remaining_time": "12:21:59", "throughput": "17782.62", "total_tokens": 2017460224}
{"current_steps": 1925, "total_steps": 2679, "loss": 0.68, "learning_rate": 9.313798510641117e-07, "epoch": 2.1538461538461537, "percentage": 71.86, "elapsed_time": "1 day, 7:31:56", "remaining_time": "12:21:03", "throughput": "17781.64", "total_tokens": 2018508800}
{"current_steps": 1926, "total_steps": 2679, "loss": 0.7835, "learning_rate": 9.290766668802773e-07, "epoch": 2.154965034965035, "percentage": 71.89, "elapsed_time": "1 day, 7:33:02", "remaining_time": "12:20:06", "throughput": "17780.57", "total_tokens": 2019557376}
{"current_steps": 1927, "total_steps": 2679, "loss": 0.5909, "learning_rate": 9.267756838627079e-07, "epoch": 2.156083916083916, "percentage": 71.93, "elapsed_time": "1 day, 7:33:56", "remaining_time": "12:19:05", "throughput": "17781.36", "total_tokens": 2020605952}
{"current_steps": 1928, "total_steps": 2679, "loss": 0.7329, "learning_rate": 9.244769052355218e-07, "epoch": 2.1572027972027974, "percentage": 71.97, "elapsed_time": "1 day, 7:34:55", "remaining_time": "12:18:07", "throughput": "17781.26", "total_tokens": 2021654528}
{"current_steps": 1929, "total_steps": 2679, "loss": 0.7475, "learning_rate": 9.22180334219753e-07, "epoch": 2.1583216783216783, "percentage": 72.0, "elapsed_time": "1 day, 7:35:55", "remaining_time": "12:17:08", "throughput": "17781.13", "total_tokens": 2022703104}
{"current_steps": 1930, "total_steps": 2679, "loss": 0.5997, "learning_rate": 9.19885974033338e-07, "epoch": 2.1594405594405592, "percentage": 72.04, "elapsed_time": "1 day, 7:36:49", "remaining_time": "12:16:07", "throughput": "17781.95", "total_tokens": 2023751680}
{"current_steps": 1931, "total_steps": 2679, "loss": 0.7637, "learning_rate": 9.175938278911184e-07, "epoch": 2.1605594405594406, "percentage": 72.08, "elapsed_time": "1 day, 7:37:49", "remaining_time": "12:15:08", "throughput": "17781.75", "total_tokens": 2024800256}
{"current_steps": 1932, "total_steps": 2679, "loss": 0.8228, "learning_rate": 9.153038990048335e-07, "epoch": 2.1616783216783215, "percentage": 72.12, "elapsed_time": "1 day, 7:39:00", "remaining_time": "12:14:14", "throughput": "17779.81", "total_tokens": 2025848832}
{"current_steps": 1933, "total_steps": 2679, "loss": 0.6544, "learning_rate": 9.130161905831131e-07, "epoch": 2.162797202797203, "percentage": 72.15, "elapsed_time": "1 day, 7:39:55", "remaining_time": "12:13:14", "throughput": "17780.52", "total_tokens": 2026897408}
{"current_steps": 1934, "total_steps": 2679, "loss": 0.7327, "learning_rate": 9.107307058314793e-07, "epoch": 2.163916083916084, "percentage": 72.19, "elapsed_time": "1 day, 7:40:50", "remaining_time": "12:12:13", "throughput": "17781.20", "total_tokens": 2027945984}
{"current_steps": 1935, "total_steps": 2679, "loss": 0.6742, "learning_rate": 9.084474479523347e-07, "epoch": 2.165034965034965, "percentage": 72.23, "elapsed_time": "1 day, 7:41:49", "remaining_time": "12:11:14", "throughput": "17781.13", "total_tokens": 2028994560}
{"current_steps": 1936, "total_steps": 2679, "loss": 0.6392, "learning_rate": 9.061664201449643e-07, "epoch": 2.166153846153846, "percentage": 72.27, "elapsed_time": "1 day, 7:42:43", "remaining_time": "12:10:13", "throughput": "17781.92", "total_tokens": 2030043136}
{"current_steps": 1937, "total_steps": 2679, "loss": 0.6315, "learning_rate": 9.038876256055288e-07, "epoch": 2.1672727272727275, "percentage": 72.3, "elapsed_time": "1 day, 7:43:42", "remaining_time": "12:09:14", "throughput": "17781.88", "total_tokens": 2031091712}
{"current_steps": 1938, "total_steps": 2679, "loss": 0.722, "learning_rate": 9.016110675270562e-07, "epoch": 2.1683916083916084, "percentage": 72.34, "elapsed_time": "1 day, 7:44:42", "remaining_time": "12:08:16", "throughput": "17781.73", "total_tokens": 2032140288}
{"current_steps": 1939, "total_steps": 2679, "loss": 0.7279, "learning_rate": 8.993367490994451e-07, "epoch": 2.1695104895104893, "percentage": 72.38, "elapsed_time": "1 day, 7:45:47", "remaining_time": "12:07:19", "throughput": "17780.76", "total_tokens": 2033188864}
{"current_steps": 1940, "total_steps": 2679, "loss": 0.7584, "learning_rate": 8.970646735094521e-07, "epoch": 2.1706293706293707, "percentage": 72.42, "elapsed_time": "1 day, 7:46:49", "remaining_time": "12:06:21", "throughput": "17780.27", "total_tokens": 2034237440}
{"current_steps": 1941, "total_steps": 2679, "loss": 0.64, "learning_rate": 8.947948439406934e-07, "epoch": 2.1717482517482516, "percentage": 72.45, "elapsed_time": "1 day, 7:47:44", "remaining_time": "12:05:21", "throughput": "17780.99", "total_tokens": 2035286016}
{"current_steps": 1942, "total_steps": 2679, "loss": 0.6578, "learning_rate": 8.925272635736387e-07, "epoch": 2.172867132867133, "percentage": 72.49, "elapsed_time": "1 day, 7:48:40", "remaining_time": "12:04:21", "throughput": "17781.38", "total_tokens": 2036334592}
{"current_steps": 1943, "total_steps": 2679, "loss": 0.649, "learning_rate": 8.902619355856032e-07, "epoch": 2.173986013986014, "percentage": 72.53, "elapsed_time": "1 day, 7:49:34", "remaining_time": "12:03:20", "throughput": "17782.19", "total_tokens": 2037383168}
{"current_steps": 1944, "total_steps": 2679, "loss": 0.6007, "learning_rate": 8.879988631507494e-07, "epoch": 2.1751048951048952, "percentage": 72.56, "elapsed_time": "1 day, 7:50:27", "remaining_time": "12:02:19", "throughput": "17783.02", "total_tokens": 2038431744}
{"current_steps": 1945, "total_steps": 2679, "loss": 0.6682, "learning_rate": 8.857380494400764e-07, "epoch": 2.176223776223776, "percentage": 72.6, "elapsed_time": "1 day, 7:51:27", "remaining_time": "12:01:20", "throughput": "17782.90", "total_tokens": 2039480320}
{"current_steps": 1946, "total_steps": 2679, "loss": 0.6925, "learning_rate": 8.834794976214206e-07, "epoch": 2.1773426573426575, "percentage": 72.64, "elapsed_time": "1 day, 7:52:24", "remaining_time": "12:00:20", "throughput": "17783.30", "total_tokens": 2040528896}
{"current_steps": 1947, "total_steps": 2679, "loss": 0.5794, "learning_rate": 8.812232108594482e-07, "epoch": 2.1784615384615384, "percentage": 72.68, "elapsed_time": "1 day, 7:53:11", "remaining_time": "11:59:17", "throughput": "17785.05", "total_tokens": 2041577472}
{"current_steps": 1948, "total_steps": 2679, "loss": 0.695, "learning_rate": 8.789691923156524e-07, "epoch": 2.17958041958042, "percentage": 72.71, "elapsed_time": "1 day, 7:54:10", "remaining_time": "11:58:18", "throughput": "17785.02", "total_tokens": 2042626048}
{"current_steps": 1949, "total_steps": 2679, "loss": 0.5919, "learning_rate": 8.767174451483468e-07, "epoch": 2.1806993006993007, "percentage": 72.75, "elapsed_time": "1 day, 7:55:04", "remaining_time": "11:57:17", "throughput": "17785.78", "total_tokens": 2043674624}
{"current_steps": 1950, "total_steps": 2679, "loss": 0.6265, "learning_rate": 8.744679725126621e-07, "epoch": 2.1818181818181817, "percentage": 72.79, "elapsed_time": "1 day, 7:56:04", "remaining_time": "11:56:19", "throughput": "17785.65", "total_tokens": 2044723200}
{"current_steps": 1951, "total_steps": 2679, "loss": 0.6858, "learning_rate": 8.722207775605437e-07, "epoch": 2.182937062937063, "percentage": 72.83, "elapsed_time": "1 day, 7:57:09", "remaining_time": "11:55:22", "throughput": "17784.71", "total_tokens": 2045771776}
{"current_steps": 1952, "total_steps": 2679, "loss": 0.6281, "learning_rate": 8.699758634407452e-07, "epoch": 2.184055944055944, "percentage": 72.86, "elapsed_time": "1 day, 7:58:03", "remaining_time": "11:54:21", "throughput": "17785.53", "total_tokens": 2046820352}
{"current_steps": 1953, "total_steps": 2679, "loss": 0.7444, "learning_rate": 8.677332332988236e-07, "epoch": 2.1851748251748253, "percentage": 72.9, "elapsed_time": "1 day, 7:59:06", "remaining_time": "11:53:23", "throughput": "17784.94", "total_tokens": 2047868928}
{"current_steps": 1954, "total_steps": 2679, "loss": 0.5745, "learning_rate": 8.654928902771359e-07, "epoch": 2.1862937062937062, "percentage": 72.94, "elapsed_time": "1 day, 7:59:54", "remaining_time": "11:52:20", "throughput": "17786.61", "total_tokens": 2048917504}
{"current_steps": 1955, "total_steps": 2679, "loss": 0.7492, "learning_rate": 8.632548375148333e-07, "epoch": 2.1874125874125876, "percentage": 72.97, "elapsed_time": "1 day, 8:00:56", "remaining_time": "11:51:23", "throughput": "17786.05", "total_tokens": 2049966080}
{"current_steps": 1956, "total_steps": 2679, "loss": 0.6301, "learning_rate": 8.610190781478595e-07, "epoch": 2.1885314685314685, "percentage": 73.01, "elapsed_time": "1 day, 8:01:50", "remaining_time": "11:50:22", "throughput": "17786.88", "total_tokens": 2051014656}
{"current_steps": 1957, "total_steps": 2679, "loss": 0.6863, "learning_rate": 8.587856153089444e-07, "epoch": 2.18965034965035, "percentage": 73.05, "elapsed_time": "1 day, 8:02:52", "remaining_time": "11:49:24", "throughput": "17786.43", "total_tokens": 2052063232}
{"current_steps": 1958, "total_steps": 2679, "loss": 0.6687, "learning_rate": 8.565544521276004e-07, "epoch": 2.190769230769231, "percentage": 73.09, "elapsed_time": "1 day, 8:03:51", "remaining_time": "11:48:25", "throughput": "17786.37", "total_tokens": 2053111808}
{"current_steps": 1959, "total_steps": 2679, "loss": 0.6111, "learning_rate": 8.543255917301163e-07, "epoch": 2.1918881118881117, "percentage": 73.12, "elapsed_time": "1 day, 8:04:45", "remaining_time": "11:47:24", "throughput": "17787.15", "total_tokens": 2054160384}
{"current_steps": 1960, "total_steps": 2679, "loss": 0.6618, "learning_rate": 8.520990372395541e-07, "epoch": 2.193006993006993, "percentage": 73.16, "elapsed_time": "1 day, 8:05:45", "remaining_time": "11:46:26", "throughput": "17787.02", "total_tokens": 2055208960}
{"current_steps": 1961, "total_steps": 2679, "loss": 0.6658, "learning_rate": 8.498747917757464e-07, "epoch": 2.194125874125874, "percentage": 73.2, "elapsed_time": "1 day, 8:06:44", "remaining_time": "11:45:27", "throughput": "17787.00", "total_tokens": 2056257536}
{"current_steps": 1962, "total_steps": 2679, "loss": 0.5486, "learning_rate": 8.476528584552896e-07, "epoch": 2.1952447552447554, "percentage": 73.24, "elapsed_time": "1 day, 8:07:34", "remaining_time": "11:44:25", "throughput": "17788.44", "total_tokens": 2057306112}
{"current_steps": 1963, "total_steps": 2679, "loss": 0.6476, "learning_rate": 8.454332403915416e-07, "epoch": 2.1963636363636363, "percentage": 73.27, "elapsed_time": "1 day, 8:08:27", "remaining_time": "11:43:24", "throughput": "17789.24", "total_tokens": 2058354688}
{"current_steps": 1964, "total_steps": 2679, "loss": 0.675, "learning_rate": 8.432159406946128e-07, "epoch": 2.1974825174825177, "percentage": 73.31, "elapsed_time": "1 day, 8:09:27", "remaining_time": "11:42:25", "throughput": "17789.20", "total_tokens": 2059403264}
{"current_steps": 1965, "total_steps": 2679, "loss": 0.7093, "learning_rate": 8.410009624713691e-07, "epoch": 2.1986013986013986, "percentage": 73.35, "elapsed_time": "1 day, 8:10:27", "remaining_time": "11:41:27", "throughput": "17788.91", "total_tokens": 2060451840}
{"current_steps": 1966, "total_steps": 2679, "loss": 0.6202, "learning_rate": 8.387883088254206e-07, "epoch": 2.19972027972028, "percentage": 73.39, "elapsed_time": "1 day, 8:11:21", "remaining_time": "11:40:26", "throughput": "17789.68", "total_tokens": 2061500416}
{"current_steps": 1967, "total_steps": 2679, "loss": 0.8134, "learning_rate": 8.365779828571214e-07, "epoch": 2.200839160839161, "percentage": 73.42, "elapsed_time": "1 day, 8:12:24", "remaining_time": "11:39:28", "throughput": "17789.06", "total_tokens": 2062548992}
{"current_steps": 1968, "total_steps": 2679, "loss": 0.6977, "learning_rate": 8.343699876635655e-07, "epoch": 2.201958041958042, "percentage": 73.46, "elapsed_time": "1 day, 8:13:29", "remaining_time": "11:38:31", "throughput": "17788.20", "total_tokens": 2063597568}
{"current_steps": 1969, "total_steps": 2679, "loss": 0.695, "learning_rate": 8.321643263385776e-07, "epoch": 2.203076923076923, "percentage": 73.5, "elapsed_time": "1 day, 8:14:34", "remaining_time": "11:37:35", "throughput": "17787.26", "total_tokens": 2064646144}
{"current_steps": 1970, "total_steps": 2679, "loss": 0.7216, "learning_rate": 8.299610019727159e-07, "epoch": 2.204195804195804, "percentage": 73.53, "elapsed_time": "1 day, 8:15:32", "remaining_time": "11:36:35", "throughput": "17787.39", "total_tokens": 2065694720}
{"current_steps": 1971, "total_steps": 2679, "loss": 0.6096, "learning_rate": 8.277600176532608e-07, "epoch": 2.2053146853146854, "percentage": 73.57, "elapsed_time": "1 day, 8:16:26", "remaining_time": "11:35:35", "throughput": "17788.19", "total_tokens": 2066743296}
{"current_steps": 1972, "total_steps": 2679, "loss": 0.6693, "learning_rate": 8.255613764642161e-07, "epoch": 2.2064335664335664, "percentage": 73.61, "elapsed_time": "1 day, 8:17:26", "remaining_time": "11:34:36", "throughput": "17788.02", "total_tokens": 2067791872}
{"current_steps": 1973, "total_steps": 2679, "loss": 0.6091, "learning_rate": 8.233650814863026e-07, "epoch": 2.2075524475524477, "percentage": 73.65, "elapsed_time": "1 day, 8:18:17", "remaining_time": "11:33:34", "throughput": "17789.26", "total_tokens": 2068840448}
{"current_steps": 1974, "total_steps": 2679, "loss": 0.6001, "learning_rate": 8.211711357969513e-07, "epoch": 2.2086713286713286, "percentage": 73.68, "elapsed_time": "1 day, 8:19:10", "remaining_time": "11:32:33", "throughput": "17790.06", "total_tokens": 2069889024}
{"current_steps": 1975, "total_steps": 2679, "loss": 0.658, "learning_rate": 8.18979542470304e-07, "epoch": 2.20979020979021, "percentage": 73.72, "elapsed_time": "1 day, 8:20:10", "remaining_time": "11:31:35", "throughput": "17789.96", "total_tokens": 2070937600}
{"current_steps": 1976, "total_steps": 2679, "loss": 0.7175, "learning_rate": 8.167903045772041e-07, "epoch": 2.210909090909091, "percentage": 73.76, "elapsed_time": "1 day, 8:21:06", "remaining_time": "11:30:35", "throughput": "17790.42", "total_tokens": 2071986176}
{"current_steps": 1977, "total_steps": 2679, "loss": 0.6342, "learning_rate": 8.146034251851959e-07, "epoch": 2.212027972027972, "percentage": 73.8, "elapsed_time": "1 day, 8:22:00", "remaining_time": "11:29:34", "throughput": "17791.18", "total_tokens": 2073034752}
{"current_steps": 1978, "total_steps": 2679, "loss": 0.722, "learning_rate": 8.124189073585204e-07, "epoch": 2.213146853146853, "percentage": 73.83, "elapsed_time": "1 day, 8:23:05", "remaining_time": "11:28:37", "throughput": "17790.23", "total_tokens": 2074083328}
{"current_steps": 1979, "total_steps": 2679, "loss": 0.6942, "learning_rate": 8.102367541581055e-07, "epoch": 2.214265734265734, "percentage": 73.87, "elapsed_time": "1 day, 8:24:10", "remaining_time": "11:27:41", "throughput": "17789.25", "total_tokens": 2075131904}
{"current_steps": 1980, "total_steps": 2679, "loss": 0.5802, "learning_rate": 8.080569686415704e-07, "epoch": 2.2153846153846155, "percentage": 73.91, "elapsed_time": "1 day, 8:25:04", "remaining_time": "11:26:40", "throughput": "17790.03", "total_tokens": 2076180480}
{"current_steps": 1981, "total_steps": 2679, "loss": 0.7118, "learning_rate": 8.05879553863213e-07, "epoch": 2.2165034965034964, "percentage": 73.95, "elapsed_time": "1 day, 8:26:04", "remaining_time": "11:25:41", "throughput": "17789.88", "total_tokens": 2077229056}
{"current_steps": 1982, "total_steps": 2679, "loss": 0.5638, "learning_rate": 8.037045128740115e-07, "epoch": 2.217622377622378, "percentage": 73.98, "elapsed_time": "1 day, 8:26:53", "remaining_time": "11:24:39", "throughput": "17791.42", "total_tokens": 2078277632}
{"current_steps": 1983, "total_steps": 2679, "loss": 0.7202, "learning_rate": 8.015318487216184e-07, "epoch": 2.2187412587412587, "percentage": 74.02, "elapsed_time": "1 day, 8:27:48", "remaining_time": "11:23:38", "throughput": "17792.02", "total_tokens": 2079326208}
{"current_steps": 1984, "total_steps": 2679, "loss": 0.6565, "learning_rate": 7.993615644503531e-07, "epoch": 2.21986013986014, "percentage": 74.06, "elapsed_time": "1 day, 8:28:42", "remaining_time": "11:22:38", "throughput": "17792.76", "total_tokens": 2080374784}
{"current_steps": 1985, "total_steps": 2679, "loss": 0.745, "learning_rate": 7.971936631012033e-07, "epoch": 2.220979020979021, "percentage": 74.09, "elapsed_time": "1 day, 8:29:36", "remaining_time": "11:21:37", "throughput": "17793.48", "total_tokens": 2081423360}
{"current_steps": 1986, "total_steps": 2679, "loss": 0.6211, "learning_rate": 7.950281477118155e-07, "epoch": 2.222097902097902, "percentage": 74.13, "elapsed_time": "1 day, 8:30:32", "remaining_time": "11:20:37", "throughput": "17794.00", "total_tokens": 2082471936}
{"current_steps": 1987, "total_steps": 2679, "loss": 0.5893, "learning_rate": 7.928650213164945e-07, "epoch": 2.2232167832167833, "percentage": 74.17, "elapsed_time": "1 day, 8:31:26", "remaining_time": "11:19:37", "throughput": "17794.70", "total_tokens": 2083520512}
{"current_steps": 1988, "total_steps": 2679, "loss": 0.7493, "learning_rate": 7.90704286946197e-07, "epoch": 2.224335664335664, "percentage": 74.21, "elapsed_time": "1 day, 8:32:31", "remaining_time": "11:18:40", "throughput": "17793.76", "total_tokens": 2084569088}
{"current_steps": 1989, "total_steps": 2679, "loss": 0.6657, "learning_rate": 7.885459476285292e-07, "epoch": 2.2254545454545456, "percentage": 74.24, "elapsed_time": "1 day, 8:33:29", "remaining_time": "11:17:40", "throughput": "17793.94", "total_tokens": 2085617664}
{"current_steps": 1990, "total_steps": 2679, "loss": 0.5458, "learning_rate": 7.863900063877397e-07, "epoch": 2.2265734265734265, "percentage": 74.28, "elapsed_time": "1 day, 8:34:17", "remaining_time": "11:16:38", "throughput": "17795.56", "total_tokens": 2086666240}
{"current_steps": 1991, "total_steps": 2679, "loss": 0.6809, "learning_rate": 7.842364662447161e-07, "epoch": 2.227692307692308, "percentage": 74.32, "elapsed_time": "1 day, 8:35:17", "remaining_time": "11:15:39", "throughput": "17795.38", "total_tokens": 2087714816}
{"current_steps": 1992, "total_steps": 2679, "loss": 0.6604, "learning_rate": 7.820853302169845e-07, "epoch": 2.228811188811189, "percentage": 74.36, "elapsed_time": "1 day, 8:36:12", "remaining_time": "11:14:39", "throughput": "17796.09", "total_tokens": 2088763392}
{"current_steps": 1993, "total_steps": 2679, "loss": 0.6575, "learning_rate": 7.799366013187007e-07, "epoch": 2.22993006993007, "percentage": 74.39, "elapsed_time": "1 day, 8:37:11", "remaining_time": "11:13:40", "throughput": "17795.94", "total_tokens": 2089811968}
{"current_steps": 1994, "total_steps": 2679, "loss": 0.5596, "learning_rate": 7.77790282560649e-07, "epoch": 2.231048951048951, "percentage": 74.43, "elapsed_time": "1 day, 8:38:00", "remaining_time": "11:12:38", "throughput": "17797.55", "total_tokens": 2090860544}
{"current_steps": 1995, "total_steps": 2679, "loss": 0.6729, "learning_rate": 7.75646376950234e-07, "epoch": 2.232167832167832, "percentage": 74.47, "elapsed_time": "1 day, 8:38:54", "remaining_time": "11:11:37", "throughput": "17798.22", "total_tokens": 2091909120}
{"current_steps": 1996, "total_steps": 2679, "loss": 0.6427, "learning_rate": 7.735048874914805e-07, "epoch": 2.2332867132867134, "percentage": 74.51, "elapsed_time": "1 day, 8:39:54", "remaining_time": "11:10:38", "throughput": "17798.16", "total_tokens": 2092957696}
{"current_steps": 1997, "total_steps": 2679, "loss": 0.679, "learning_rate": 7.713658171850289e-07, "epoch": 2.2344055944055943, "percentage": 74.54, "elapsed_time": "1 day, 8:40:53", "remaining_time": "11:09:40", "throughput": "17798.07", "total_tokens": 2094006272}
{"current_steps": 1998, "total_steps": 2679, "loss": 0.6468, "learning_rate": 7.692291690281267e-07, "epoch": 2.2355244755244756, "percentage": 74.58, "elapsed_time": "1 day, 8:41:47", "remaining_time": "11:08:39", "throughput": "17798.88", "total_tokens": 2095054848}
{"current_steps": 1999, "total_steps": 2679, "loss": 0.6207, "learning_rate": 7.670949460146329e-07, "epoch": 2.2366433566433566, "percentage": 74.62, "elapsed_time": "1 day, 8:42:46", "remaining_time": "11:07:40", "throughput": "17798.82", "total_tokens": 2096103424}
{"current_steps": 2000, "total_steps": 2679, "loss": 0.6875, "learning_rate": 7.649631511350025e-07, "epoch": 2.237762237762238, "percentage": 74.65, "elapsed_time": "1 day, 8:43:49", "remaining_time": "11:06:43", "throughput": "17798.14", "total_tokens": 2097152000}
{"current_steps": 2000, "total_steps": 2679, "eval_loss": 0.7205791473388672, "epoch": 2.237762237762238, "percentage": 74.65, "elapsed_time": "1 day, 8:47:56", "remaining_time": "11:08:07", "throughput": "17760.92", "total_tokens": 2097152000}
{"current_steps": 2001, "total_steps": 2679, "loss": 0.7097, "learning_rate": 7.628337873762928e-07, "epoch": 2.238881118881119, "percentage": 74.69, "elapsed_time": "1 day, 8:49:25", "remaining_time": "11:07:18", "throughput": "17756.46", "total_tokens": 2098200576}
{"current_steps": 2002, "total_steps": 2679, "loss": 0.6993, "learning_rate": 7.607068577221516e-07, "epoch": 2.24, "percentage": 74.73, "elapsed_time": "1 day, 8:50:26", "remaining_time": "11:06:19", "throughput": "17756.11", "total_tokens": 2099249152}
{"current_steps": 2003, "total_steps": 2679, "loss": 0.6197, "learning_rate": 7.585823651528157e-07, "epoch": 2.241118881118881, "percentage": 74.77, "elapsed_time": "1 day, 8:51:20", "remaining_time": "11:05:18", "throughput": "17756.97", "total_tokens": 2100297728}
{"current_steps": 2004, "total_steps": 2679, "loss": 0.6454, "learning_rate": 7.564603126451117e-07, "epoch": 2.242237762237762, "percentage": 74.8, "elapsed_time": "1 day, 8:52:14", "remaining_time": "11:04:18", "throughput": "17757.69", "total_tokens": 2101346304}
{"current_steps": 2005, "total_steps": 2679, "loss": 0.6997, "learning_rate": 7.543407031724415e-07, "epoch": 2.2433566433566434, "percentage": 74.84, "elapsed_time": "1 day, 8:53:14", "remaining_time": "11:03:19", "throughput": "17757.57", "total_tokens": 2102394880}
{"current_steps": 2006, "total_steps": 2679, "loss": 0.7814, "learning_rate": 7.522235397047881e-07, "epoch": 2.2444755244755243, "percentage": 74.88, "elapsed_time": "1 day, 8:54:14", "remaining_time": "11:02:20", "throughput": "17757.33", "total_tokens": 2103443456}
{"current_steps": 2007, "total_steps": 2679, "loss": 0.7047, "learning_rate": 7.501088252087046e-07, "epoch": 2.2455944055944057, "percentage": 74.92, "elapsed_time": "1 day, 8:55:09", "remaining_time": "11:01:20", "throughput": "17757.97", "total_tokens": 2104492032}
{"current_steps": 2008, "total_steps": 2679, "loss": 0.5965, "learning_rate": 7.479965626473118e-07, "epoch": 2.2467132867132866, "percentage": 74.95, "elapsed_time": "1 day, 8:56:03", "remaining_time": "11:00:19", "throughput": "17758.80", "total_tokens": 2105540608}
{"current_steps": 2009, "total_steps": 2679, "loss": 0.7117, "learning_rate": 7.458867549802998e-07, "epoch": 2.247832167832168, "percentage": 74.99, "elapsed_time": "1 day, 8:57:05", "remaining_time": "10:59:21", "throughput": "17758.37", "total_tokens": 2106589184}
{"current_steps": 2010, "total_steps": 2679, "loss": 0.7982, "learning_rate": 7.437794051639133e-07, "epoch": 2.248951048951049, "percentage": 75.03, "elapsed_time": "1 day, 8:58:05", "remaining_time": "10:58:22", "throughput": "17758.16", "total_tokens": 2107637760}
{"current_steps": 2011, "total_steps": 2679, "loss": 0.5344, "learning_rate": 7.41674516150957e-07, "epoch": 2.2500699300699303, "percentage": 75.07, "elapsed_time": "1 day, 8:58:53", "remaining_time": "10:57:20", "throughput": "17759.77", "total_tokens": 2108686336}
{"current_steps": 2012, "total_steps": 2679, "loss": 0.6165, "learning_rate": 7.395720908907842e-07, "epoch": 2.251188811188811, "percentage": 75.1, "elapsed_time": "1 day, 8:59:47", "remaining_time": "10:56:19", "throughput": "17760.56", "total_tokens": 2109734912}
{"current_steps": 2013, "total_steps": 2679, "loss": 0.6814, "learning_rate": 7.374721323292985e-07, "epoch": 2.252307692307692, "percentage": 75.14, "elapsed_time": "1 day, 9:00:47", "remaining_time": "10:55:20", "throughput": "17760.50", "total_tokens": 2110783488}
{"current_steps": 2014, "total_steps": 2679, "loss": 0.6214, "learning_rate": 7.353746434089476e-07, "epoch": 2.2534265734265735, "percentage": 75.18, "elapsed_time": "1 day, 9:01:40", "remaining_time": "10:54:19", "throughput": "17761.28", "total_tokens": 2111832064}
{"current_steps": 2015, "total_steps": 2679, "loss": 0.6101, "learning_rate": 7.332796270687159e-07, "epoch": 2.2545454545454544, "percentage": 75.21, "elapsed_time": "1 day, 9:02:35", "remaining_time": "10:53:19", "throughput": "17762.00", "total_tokens": 2112880640}
{"current_steps": 2016, "total_steps": 2679, "loss": 0.7496, "learning_rate": 7.311870862441265e-07, "epoch": 2.2556643356643358, "percentage": 75.25, "elapsed_time": "1 day, 9:03:34", "remaining_time": "10:52:20", "throughput": "17761.88", "total_tokens": 2113929216}
{"current_steps": 2017, "total_steps": 2679, "loss": 0.7339, "learning_rate": 7.290970238672307e-07, "epoch": 2.2567832167832167, "percentage": 75.29, "elapsed_time": "1 day, 9:04:40", "remaining_time": "10:51:23", "throughput": "17760.95", "total_tokens": 2114977792}
{"current_steps": 2018, "total_steps": 2679, "loss": 0.5609, "learning_rate": 7.270094428666099e-07, "epoch": 2.257902097902098, "percentage": 75.33, "elapsed_time": "1 day, 9:05:28", "remaining_time": "10:50:20", "throughput": "17762.58", "total_tokens": 2116026368}
{"current_steps": 2019, "total_steps": 2679, "loss": 0.772, "learning_rate": 7.24924346167366e-07, "epoch": 2.259020979020979, "percentage": 75.36, "elapsed_time": "1 day, 9:06:28", "remaining_time": "10:49:22", "throughput": "17762.40", "total_tokens": 2117074944}
{"current_steps": 2020, "total_steps": 2679, "loss": 0.7736, "learning_rate": 7.228417366911222e-07, "epoch": 2.2601398601398603, "percentage": 75.4, "elapsed_time": "1 day, 9:07:28", "remaining_time": "10:48:23", "throughput": "17762.29", "total_tokens": 2118123520}
{"current_steps": 2021, "total_steps": 2679, "loss": 0.7358, "learning_rate": 7.207616173560158e-07, "epoch": 2.2612587412587413, "percentage": 75.44, "elapsed_time": "1 day, 9:08:33", "remaining_time": "10:47:26", "throughput": "17761.41", "total_tokens": 2119172096}
{"current_steps": 2022, "total_steps": 2679, "loss": 0.674, "learning_rate": 7.18683991076694e-07, "epoch": 2.262377622377622, "percentage": 75.48, "elapsed_time": "1 day, 9:09:32", "remaining_time": "10:46:27", "throughput": "17761.36", "total_tokens": 2120220672}
{"current_steps": 2023, "total_steps": 2679, "loss": 0.618, "learning_rate": 7.166088607643123e-07, "epoch": 2.2634965034965036, "percentage": 75.51, "elapsed_time": "1 day, 9:10:27", "remaining_time": "10:45:26", "throughput": "17762.03", "total_tokens": 2121269248}
{"current_steps": 2024, "total_steps": 2679, "loss": 0.7095, "learning_rate": 7.145362293265265e-07, "epoch": 2.2646153846153845, "percentage": 75.55, "elapsed_time": "1 day, 9:11:27", "remaining_time": "10:44:28", "throughput": "17761.87", "total_tokens": 2122317824}
{"current_steps": 2025, "total_steps": 2679, "loss": 0.637, "learning_rate": 7.124660996674951e-07, "epoch": 2.265734265734266, "percentage": 75.59, "elapsed_time": "1 day, 9:12:21", "remaining_time": "10:43:27", "throughput": "17762.63", "total_tokens": 2123366400}
{"current_steps": 2026, "total_steps": 2679, "loss": 0.5973, "learning_rate": 7.103984746878673e-07, "epoch": 2.2668531468531468, "percentage": 75.63, "elapsed_time": "1 day, 9:13:16", "remaining_time": "10:42:27", "throughput": "17763.25", "total_tokens": 2124414976}
{"current_steps": 2027, "total_steps": 2679, "loss": 0.6508, "learning_rate": 7.083333572847831e-07, "epoch": 2.267972027972028, "percentage": 75.66, "elapsed_time": "1 day, 9:14:15", "remaining_time": "10:41:28", "throughput": "17763.17", "total_tokens": 2125463552}
{"current_steps": 2028, "total_steps": 2679, "loss": 0.6465, "learning_rate": 7.062707503518718e-07, "epoch": 2.269090909090909, "percentage": 75.7, "elapsed_time": "1 day, 9:15:12", "remaining_time": "10:40:28", "throughput": "17763.55", "total_tokens": 2126512128}
{"current_steps": 2029, "total_steps": 2679, "loss": 0.6888, "learning_rate": 7.042106567792406e-07, "epoch": 2.2702097902097904, "percentage": 75.74, "elapsed_time": "1 day, 9:16:12", "remaining_time": "10:39:29", "throughput": "17763.35", "total_tokens": 2127560704}
{"current_steps": 2030, "total_steps": 2679, "loss": 0.5843, "learning_rate": 7.021530794534803e-07, "epoch": 2.2713286713286713, "percentage": 75.77, "elapsed_time": "1 day, 9:17:00", "remaining_time": "10:38:27", "throughput": "17764.96", "total_tokens": 2128609280}
{"current_steps": 2031, "total_steps": 2679, "loss": 0.6222, "learning_rate": 7.000980212576522e-07, "epoch": 2.2724475524475523, "percentage": 75.81, "elapsed_time": "1 day, 9:17:54", "remaining_time": "10:37:26", "throughput": "17765.69", "total_tokens": 2129657856}
{"current_steps": 2032, "total_steps": 2679, "loss": 0.638, "learning_rate": 6.980454850712878e-07, "epoch": 2.2735664335664336, "percentage": 75.85, "elapsed_time": "1 day, 9:18:48", "remaining_time": "10:36:25", "throughput": "17766.44", "total_tokens": 2130706432}
{"current_steps": 2033, "total_steps": 2679, "loss": 0.5787, "learning_rate": 6.959954737703872e-07, "epoch": 2.2746853146853145, "percentage": 75.89, "elapsed_time": "1 day, 9:19:36", "remaining_time": "10:35:23", "throughput": "17768.06", "total_tokens": 2131755008}
{"current_steps": 2034, "total_steps": 2679, "loss": 0.6328, "learning_rate": 6.939479902274088e-07, "epoch": 2.275804195804196, "percentage": 75.92, "elapsed_time": "1 day, 9:20:30", "remaining_time": "10:34:22", "throughput": "17768.80", "total_tokens": 2132803584}
{"current_steps": 2035, "total_steps": 2679, "loss": 0.6582, "learning_rate": 6.919030373112748e-07, "epoch": 2.276923076923077, "percentage": 75.96, "elapsed_time": "1 day, 9:21:30", "remaining_time": "10:33:24", "throughput": "17768.67", "total_tokens": 2133852160}
{"current_steps": 2036, "total_steps": 2679, "loss": 0.7161, "learning_rate": 6.89860617887356e-07, "epoch": 2.278041958041958, "percentage": 76.0, "elapsed_time": "1 day, 9:22:34", "remaining_time": "10:32:26", "throughput": "17767.96", "total_tokens": 2134900736}
{"current_steps": 2037, "total_steps": 2679, "loss": 0.5773, "learning_rate": 6.87820734817477e-07, "epoch": 2.279160839160839, "percentage": 76.04, "elapsed_time": "1 day, 9:23:22", "remaining_time": "10:31:24", "throughput": "17769.58", "total_tokens": 2135949312}
{"current_steps": 2038, "total_steps": 2679, "loss": 0.669, "learning_rate": 6.857833909599063e-07, "epoch": 2.2802797202797205, "percentage": 76.07, "elapsed_time": "1 day, 9:24:22", "remaining_time": "10:30:25", "throughput": "17769.49", "total_tokens": 2136997888}
{"current_steps": 2039, "total_steps": 2679, "loss": 0.7104, "learning_rate": 6.837485891693541e-07, "epoch": 2.2813986013986014, "percentage": 76.11, "elapsed_time": "1 day, 9:25:22", "remaining_time": "10:29:26", "throughput": "17769.33", "total_tokens": 2138046464}
{"current_steps": 2040, "total_steps": 2679, "loss": 0.5769, "learning_rate": 6.817163322969708e-07, "epoch": 2.2825174825174823, "percentage": 76.15, "elapsed_time": "1 day, 9:26:10", "remaining_time": "10:28:24", "throughput": "17770.91", "total_tokens": 2139095040}
{"current_steps": 2041, "total_steps": 2679, "loss": 0.6786, "learning_rate": 6.796866231903402e-07, "epoch": 2.2836363636363637, "percentage": 76.19, "elapsed_time": "1 day, 9:27:09", "remaining_time": "10:27:25", "throughput": "17770.90", "total_tokens": 2140143616}
{"current_steps": 2042, "total_steps": 2679, "loss": 0.5712, "learning_rate": 6.776594646934756e-07, "epoch": 2.2847552447552446, "percentage": 76.22, "elapsed_time": "1 day, 9:28:02", "remaining_time": "10:26:24", "throughput": "17771.82", "total_tokens": 2141192192}
{"current_steps": 2043, "total_steps": 2679, "loss": 0.7821, "learning_rate": 6.756348596468168e-07, "epoch": 2.285874125874126, "percentage": 76.26, "elapsed_time": "1 day, 9:29:13", "remaining_time": "10:25:29", "throughput": "17770.06", "total_tokens": 2142240768}
{"current_steps": 2044, "total_steps": 2679, "loss": 0.5473, "learning_rate": 6.736128108872244e-07, "epoch": 2.286993006993007, "percentage": 76.3, "elapsed_time": "1 day, 9:30:01", "remaining_time": "10:24:26", "throughput": "17771.65", "total_tokens": 2143289344}
{"current_steps": 2045, "total_steps": 2679, "loss": 0.6668, "learning_rate": 6.715933212479791e-07, "epoch": 2.2881118881118883, "percentage": 76.33, "elapsed_time": "1 day, 9:30:55", "remaining_time": "10:23:26", "throughput": "17772.37", "total_tokens": 2144337920}
{"current_steps": 2046, "total_steps": 2679, "loss": 0.6109, "learning_rate": 6.695763935587752e-07, "epoch": 2.289230769230769, "percentage": 76.37, "elapsed_time": "1 day, 9:31:49", "remaining_time": "10:22:25", "throughput": "17773.18", "total_tokens": 2145386496}
{"current_steps": 2047, "total_steps": 2679, "loss": 0.7932, "learning_rate": 6.675620306457172e-07, "epoch": 2.2903496503496505, "percentage": 76.41, "elapsed_time": "1 day, 9:33:00", "remaining_time": "10:21:30", "throughput": "17771.32", "total_tokens": 2146435072}
{"current_steps": 2048, "total_steps": 2679, "loss": 0.7036, "learning_rate": 6.655502353313147e-07, "epoch": 2.2914685314685315, "percentage": 76.45, "elapsed_time": "1 day, 9:34:00", "remaining_time": "10:20:31", "throughput": "17771.16", "total_tokens": 2147483648}
{"current_steps": 2049, "total_steps": 2679, "loss": 0.8149, "learning_rate": 6.635410104344819e-07, "epoch": 2.2925874125874124, "percentage": 76.48, "elapsed_time": "1 day, 9:35:06", "remaining_time": "10:19:34", "throughput": "17770.13", "total_tokens": 2148532224}
{"current_steps": 2050, "total_steps": 2679, "loss": 0.7342, "learning_rate": 6.615343587705284e-07, "epoch": 2.2937062937062938, "percentage": 76.52, "elapsed_time": "1 day, 9:36:07", "remaining_time": "10:18:36", "throughput": "17769.93", "total_tokens": 2149580800}
{"current_steps": 2051, "total_steps": 2679, "loss": 0.721, "learning_rate": 6.595302831511607e-07, "epoch": 2.2948251748251747, "percentage": 76.56, "elapsed_time": "1 day, 9:37:12", "remaining_time": "10:17:39", "throughput": "17769.01", "total_tokens": 2150629376}
{"current_steps": 2052, "total_steps": 2679, "loss": 0.6407, "learning_rate": 6.575287863844753e-07, "epoch": 2.295944055944056, "percentage": 76.6, "elapsed_time": "1 day, 9:38:06", "remaining_time": "10:16:38", "throughput": "17769.75", "total_tokens": 2151677952}
{"current_steps": 2053, "total_steps": 2679, "loss": 0.7026, "learning_rate": 6.555298712749538e-07, "epoch": 2.297062937062937, "percentage": 76.63, "elapsed_time": "1 day, 9:39:12", "remaining_time": "10:15:41", "throughput": "17768.78", "total_tokens": 2152726528}
{"current_steps": 2054, "total_steps": 2679, "loss": 0.5839, "learning_rate": 6.535335406234627e-07, "epoch": 2.2981818181818183, "percentage": 76.67, "elapsed_time": "1 day, 9:40:07", "remaining_time": "10:14:41", "throughput": "17769.30", "total_tokens": 2153775104}
{"current_steps": 2055, "total_steps": 2679, "loss": 0.6603, "learning_rate": 6.515397972272444e-07, "epoch": 2.2993006993006992, "percentage": 76.71, "elapsed_time": "1 day, 9:41:06", "remaining_time": "10:13:42", "throughput": "17769.27", "total_tokens": 2154823680}
{"current_steps": 2056, "total_steps": 2679, "loss": 0.7736, "learning_rate": 6.495486438799181e-07, "epoch": 2.3004195804195806, "percentage": 76.75, "elapsed_time": "1 day, 9:42:12", "remaining_time": "10:12:45", "throughput": "17768.37", "total_tokens": 2155872256}
{"current_steps": 2057, "total_steps": 2679, "loss": 0.6303, "learning_rate": 6.475600833714743e-07, "epoch": 2.3015384615384615, "percentage": 76.78, "elapsed_time": "1 day, 9:43:09", "remaining_time": "10:11:46", "throughput": "17768.57", "total_tokens": 2156920832}
{"current_steps": 2058, "total_steps": 2679, "loss": 0.6659, "learning_rate": 6.455741184882674e-07, "epoch": 2.3026573426573425, "percentage": 76.82, "elapsed_time": "1 day, 9:44:08", "remaining_time": "10:10:47", "throughput": "17768.54", "total_tokens": 2157969408}
{"current_steps": 2059, "total_steps": 2679, "loss": 0.7286, "learning_rate": 6.435907520130191e-07, "epoch": 2.303776223776224, "percentage": 76.86, "elapsed_time": "1 day, 9:45:13", "remaining_time": "10:09:49", "throughput": "17767.67", "total_tokens": 2159017984}
{"current_steps": 2060, "total_steps": 2679, "loss": 0.6201, "learning_rate": 6.41609986724806e-07, "epoch": 2.3048951048951047, "percentage": 76.89, "elapsed_time": "1 day, 9:46:07", "remaining_time": "10:08:49", "throughput": "17768.40", "total_tokens": 2160066560}
{"current_steps": 2061, "total_steps": 2679, "loss": 0.6331, "learning_rate": 6.396318253990628e-07, "epoch": 2.306013986013986, "percentage": 76.93, "elapsed_time": "1 day, 9:47:04", "remaining_time": "10:07:49", "throughput": "17768.73", "total_tokens": 2161115136}
{"current_steps": 2062, "total_steps": 2679, "loss": 0.5273, "learning_rate": 6.376562708075753e-07, "epoch": 2.307132867132867, "percentage": 76.97, "elapsed_time": "1 day, 9:47:52", "remaining_time": "10:06:47", "throughput": "17770.31", "total_tokens": 2162163712}
{"current_steps": 2063, "total_steps": 2679, "loss": 0.736, "learning_rate": 6.356833257184747e-07, "epoch": 2.3082517482517484, "percentage": 77.01, "elapsed_time": "1 day, 9:48:50", "remaining_time": "10:05:48", "throughput": "17770.45", "total_tokens": 2163212288}
{"current_steps": 2064, "total_steps": 2679, "loss": 0.5976, "learning_rate": 6.337129928962385e-07, "epoch": 2.3093706293706293, "percentage": 77.04, "elapsed_time": "1 day, 9:49:44", "remaining_time": "10:04:47", "throughput": "17771.27", "total_tokens": 2164260864}
{"current_steps": 2065, "total_steps": 2679, "loss": 0.7811, "learning_rate": 6.317452751016815e-07, "epoch": 2.3104895104895107, "percentage": 77.08, "elapsed_time": "1 day, 9:50:55", "remaining_time": "10:03:52", "throughput": "17769.52", "total_tokens": 2165309440}
{"current_steps": 2066, "total_steps": 2679, "loss": 0.6262, "learning_rate": 6.297801750919558e-07, "epoch": 2.3116083916083916, "percentage": 77.12, "elapsed_time": "1 day, 9:51:54", "remaining_time": "10:02:53", "throughput": "17769.47", "total_tokens": 2166358016}
{"current_steps": 2067, "total_steps": 2679, "loss": 0.7044, "learning_rate": 6.278176956205462e-07, "epoch": 2.3127272727272725, "percentage": 77.16, "elapsed_time": "1 day, 9:52:49", "remaining_time": "10:01:52", "throughput": "17770.06", "total_tokens": 2167406592}
{"current_steps": 2068, "total_steps": 2679, "loss": 0.6227, "learning_rate": 6.25857839437263e-07, "epoch": 2.313846153846154, "percentage": 77.19, "elapsed_time": "1 day, 9:53:43", "remaining_time": "10:00:52", "throughput": "17770.77", "total_tokens": 2168455168}
{"current_steps": 2069, "total_steps": 2679, "loss": 0.5727, "learning_rate": 6.239006092882438e-07, "epoch": 2.314965034965035, "percentage": 77.23, "elapsed_time": "1 day, 9:54:31", "remaining_time": "9:59:50", "throughput": "17772.36", "total_tokens": 2169503744}
{"current_steps": 2070, "total_steps": 2679, "loss": 0.5844, "learning_rate": 6.219460079159434e-07, "epoch": 2.316083916083916, "percentage": 77.27, "elapsed_time": "1 day, 9:55:25", "remaining_time": "9:58:49", "throughput": "17773.09", "total_tokens": 2170552320}
{"current_steps": 2071, "total_steps": 2679, "loss": 0.7689, "learning_rate": 6.19994038059136e-07, "epoch": 2.317202797202797, "percentage": 77.3, "elapsed_time": "1 day, 9:56:31", "remaining_time": "9:57:52", "throughput": "17772.14", "total_tokens": 2171600896}
{"current_steps": 2072, "total_steps": 2679, "loss": 0.775, "learning_rate": 6.180447024529074e-07, "epoch": 2.3183216783216785, "percentage": 77.34, "elapsed_time": "1 day, 9:57:36", "remaining_time": "9:56:55", "throughput": "17771.18", "total_tokens": 2172649472}
{"current_steps": 2073, "total_steps": 2679, "loss": 0.6933, "learning_rate": 6.160980038286529e-07, "epoch": 2.3194405594405594, "percentage": 77.38, "elapsed_time": "1 day, 9:58:34", "remaining_time": "9:55:56", "throughput": "17771.41", "total_tokens": 2173698048}
{"current_steps": 2074, "total_steps": 2679, "loss": 0.6771, "learning_rate": 6.141539449140718e-07, "epoch": 2.3205594405594407, "percentage": 77.42, "elapsed_time": "1 day, 9:59:28", "remaining_time": "9:54:55", "throughput": "17772.17", "total_tokens": 2174746624}
{"current_steps": 2075, "total_steps": 2679, "loss": 0.602, "learning_rate": 6.122125284331646e-07, "epoch": 2.3216783216783217, "percentage": 77.45, "elapsed_time": "1 day, 10:00:22", "remaining_time": "9:53:55", "throughput": "17772.89", "total_tokens": 2175795200}
{"current_steps": 2076, "total_steps": 2679, "loss": 0.702, "learning_rate": 6.1027375710623e-07, "epoch": 2.3227972027972026, "percentage": 77.49, "elapsed_time": "1 day, 10:01:21", "remaining_time": "9:52:56", "throughput": "17772.78", "total_tokens": 2176843776}
{"current_steps": 2077, "total_steps": 2679, "loss": 0.6474, "learning_rate": 6.083376336498608e-07, "epoch": 2.323916083916084, "percentage": 77.53, "elapsed_time": "1 day, 10:02:15", "remaining_time": "9:51:55", "throughput": "17773.52", "total_tokens": 2177892352}
{"current_steps": 2078, "total_steps": 2679, "loss": 0.645, "learning_rate": 6.064041607769397e-07, "epoch": 2.325034965034965, "percentage": 77.57, "elapsed_time": "1 day, 10:03:16", "remaining_time": "9:50:57", "throughput": "17773.28", "total_tokens": 2178940928}
{"current_steps": 2079, "total_steps": 2679, "loss": 0.6766, "learning_rate": 6.044733411966336e-07, "epoch": 2.3261538461538462, "percentage": 77.6, "elapsed_time": "1 day, 10:04:14", "remaining_time": "9:49:58", "throughput": "17773.47", "total_tokens": 2179989504}
{"current_steps": 2080, "total_steps": 2679, "loss": 0.7105, "learning_rate": 6.025451776143923e-07, "epoch": 2.327272727272727, "percentage": 77.64, "elapsed_time": "1 day, 10:05:14", "remaining_time": "9:48:59", "throughput": "17773.33", "total_tokens": 2181038080}
{"current_steps": 2081, "total_steps": 2679, "loss": 0.8447, "learning_rate": 6.006196727319452e-07, "epoch": 2.3283916083916085, "percentage": 77.68, "elapsed_time": "1 day, 10:06:20", "remaining_time": "9:48:02", "throughput": "17772.26", "total_tokens": 2182086656}
{"current_steps": 2082, "total_steps": 2679, "loss": 0.8273, "learning_rate": 5.986968292472955e-07, "epoch": 2.3295104895104894, "percentage": 77.72, "elapsed_time": "1 day, 10:07:24", "remaining_time": "9:47:04", "throughput": "17771.48", "total_tokens": 2183135232}
{"current_steps": 2083, "total_steps": 2679, "loss": 0.6558, "learning_rate": 5.967766498547181e-07, "epoch": 2.330629370629371, "percentage": 77.75, "elapsed_time": "1 day, 10:08:24", "remaining_time": "9:46:06", "throughput": "17771.33", "total_tokens": 2184183808}
{"current_steps": 2084, "total_steps": 2679, "loss": 0.6443, "learning_rate": 5.94859137244754e-07, "epoch": 2.3317482517482517, "percentage": 77.79, "elapsed_time": "1 day, 10:09:24", "remaining_time": "9:45:07", "throughput": "17771.20", "total_tokens": 2185232384}
{"current_steps": 2085, "total_steps": 2679, "loss": 0.7713, "learning_rate": 5.929442941042066e-07, "epoch": 2.3328671328671327, "percentage": 77.83, "elapsed_time": "1 day, 10:10:35", "remaining_time": "9:44:11", "throughput": "17769.52", "total_tokens": 2186280960}
{"current_steps": 2086, "total_steps": 2679, "loss": 0.6315, "learning_rate": 5.910321231161409e-07, "epoch": 2.333986013986014, "percentage": 77.86, "elapsed_time": "1 day, 10:11:29", "remaining_time": "9:43:11", "throughput": "17770.22", "total_tokens": 2187329536}
{"current_steps": 2087, "total_steps": 2679, "loss": 0.5425, "learning_rate": 5.891226269598768e-07, "epoch": 2.335104895104895, "percentage": 77.9, "elapsed_time": "1 day, 10:12:18", "remaining_time": "9:42:09", "throughput": "17771.75", "total_tokens": 2188378112}
{"current_steps": 2088, "total_steps": 2679, "loss": 0.6496, "learning_rate": 5.872158083109874e-07, "epoch": 2.3362237762237763, "percentage": 77.94, "elapsed_time": "1 day, 10:13:09", "remaining_time": "9:41:08", "throughput": "17772.89", "total_tokens": 2189426688}
{"current_steps": 2089, "total_steps": 2679, "loss": 0.6263, "learning_rate": 5.853116698412913e-07, "epoch": 2.3373426573426572, "percentage": 77.98, "elapsed_time": "1 day, 10:14:02", "remaining_time": "9:40:07", "throughput": "17773.63", "total_tokens": 2190475264}
{"current_steps": 2090, "total_steps": 2679, "loss": 0.6418, "learning_rate": 5.83410214218855e-07, "epoch": 2.3384615384615386, "percentage": 78.01, "elapsed_time": "1 day, 10:14:57", "remaining_time": "9:39:07", "throughput": "17774.33", "total_tokens": 2191523840}
{"current_steps": 2091, "total_steps": 2679, "loss": 0.6581, "learning_rate": 5.815114441079825e-07, "epoch": 2.3395804195804195, "percentage": 78.05, "elapsed_time": "1 day, 10:15:51", "remaining_time": "9:38:07", "throughput": "17774.97", "total_tokens": 2192572416}
{"current_steps": 2092, "total_steps": 2679, "loss": 0.6411, "learning_rate": 5.796153621692174e-07, "epoch": 2.340699300699301, "percentage": 78.09, "elapsed_time": "1 day, 10:16:51", "remaining_time": "9:37:08", "throughput": "17774.84", "total_tokens": 2193620992}
{"current_steps": 2093, "total_steps": 2679, "loss": 0.6982, "learning_rate": 5.777219710593365e-07, "epoch": 2.341818181818182, "percentage": 78.13, "elapsed_time": "1 day, 10:17:57", "remaining_time": "9:36:11", "throughput": "17773.91", "total_tokens": 2194669568}
{"current_steps": 2094, "total_steps": 2679, "loss": 0.5993, "learning_rate": 5.758312734313442e-07, "epoch": 2.3429370629370627, "percentage": 78.16, "elapsed_time": "1 day, 10:18:47", "remaining_time": "9:35:09", "throughput": "17775.08", "total_tokens": 2195718144}
{"current_steps": 2095, "total_steps": 2679, "loss": 0.7793, "learning_rate": 5.739432719344737e-07, "epoch": 2.344055944055944, "percentage": 78.2, "elapsed_time": "1 day, 10:19:57", "remaining_time": "9:34:13", "throughput": "17773.56", "total_tokens": 2196766720}
{"current_steps": 2096, "total_steps": 2679, "loss": 0.7526, "learning_rate": 5.720579692141773e-07, "epoch": 2.345174825174825, "percentage": 78.24, "elapsed_time": "1 day, 10:21:03", "remaining_time": "9:33:16", "throughput": "17772.62", "total_tokens": 2197815296}
{"current_steps": 2097, "total_steps": 2679, "loss": 0.7495, "learning_rate": 5.70175367912128e-07, "epoch": 2.3462937062937064, "percentage": 78.28, "elapsed_time": "1 day, 10:22:08", "remaining_time": "9:32:19", "throughput": "17771.69", "total_tokens": 2198863872}
{"current_steps": 2098, "total_steps": 2679, "loss": 0.7236, "learning_rate": 5.682954706662137e-07, "epoch": 2.3474125874125873, "percentage": 78.31, "elapsed_time": "1 day, 10:23:10", "remaining_time": "9:31:21", "throughput": "17771.23", "total_tokens": 2199912448}
{"current_steps": 2099, "total_steps": 2679, "loss": 0.701, "learning_rate": 5.664182801105314e-07, "epoch": 2.3485314685314687, "percentage": 78.35, "elapsed_time": "1 day, 10:24:12", "remaining_time": "9:30:23", "throughput": "17770.80", "total_tokens": 2200961024}
{"current_steps": 2100, "total_steps": 2679, "loss": 0.6934, "learning_rate": 5.645437988753877e-07, "epoch": 2.3496503496503496, "percentage": 78.39, "elapsed_time": "1 day, 10:25:12", "remaining_time": "9:29:24", "throughput": "17770.69", "total_tokens": 2202009600}
{"current_steps": 2101, "total_steps": 2679, "loss": 0.6752, "learning_rate": 5.626720295872911e-07, "epoch": 2.350769230769231, "percentage": 78.42, "elapsed_time": "1 day, 10:26:09", "remaining_time": "9:28:24", "throughput": "17771.03", "total_tokens": 2203058176}
{"current_steps": 2102, "total_steps": 2679, "loss": 0.6061, "learning_rate": 5.608029748689514e-07, "epoch": 2.351888111888112, "percentage": 78.46, "elapsed_time": "1 day, 10:27:02", "remaining_time": "9:27:24", "throughput": "17771.77", "total_tokens": 2204106752}
{"current_steps": 2103, "total_steps": 2679, "loss": 0.6042, "learning_rate": 5.589366373392754e-07, "epoch": 2.353006993006993, "percentage": 78.5, "elapsed_time": "1 day, 10:27:57", "remaining_time": "9:26:24", "throughput": "17772.45", "total_tokens": 2205155328}
{"current_steps": 2104, "total_steps": 2679, "loss": 0.7077, "learning_rate": 5.570730196133597e-07, "epoch": 2.354125874125874, "percentage": 78.54, "elapsed_time": "1 day, 10:28:53", "remaining_time": "9:25:24", "throughput": "17772.88", "total_tokens": 2206203904}
{"current_steps": 2105, "total_steps": 2679, "loss": 0.5474, "learning_rate": 5.552121243024935e-07, "epoch": 2.355244755244755, "percentage": 78.57, "elapsed_time": "1 day, 10:29:41", "remaining_time": "9:24:22", "throughput": "17774.44", "total_tokens": 2207252480}
{"current_steps": 2106, "total_steps": 2679, "loss": 0.7622, "learning_rate": 5.533539540141483e-07, "epoch": 2.3563636363636364, "percentage": 78.61, "elapsed_time": "1 day, 10:30:47", "remaining_time": "9:23:25", "throughput": "17773.39", "total_tokens": 2208301056}
{"current_steps": 2107, "total_steps": 2679, "loss": 0.5663, "learning_rate": 5.514985113519794e-07, "epoch": 2.3574825174825174, "percentage": 78.65, "elapsed_time": "1 day, 10:31:35", "remaining_time": "9:22:23", "throughput": "17774.98", "total_tokens": 2209349632}
{"current_steps": 2108, "total_steps": 2679, "loss": 0.6412, "learning_rate": 5.4964579891582e-07, "epoch": 2.3586013986013987, "percentage": 78.69, "elapsed_time": "1 day, 10:32:29", "remaining_time": "9:21:22", "throughput": "17775.67", "total_tokens": 2210398208}
{"current_steps": 2109, "total_steps": 2679, "loss": 0.7015, "learning_rate": 5.477958193016758e-07, "epoch": 2.3597202797202796, "percentage": 78.72, "elapsed_time": "1 day, 10:33:24", "remaining_time": "9:20:22", "throughput": "17776.26", "total_tokens": 2211446784}
{"current_steps": 2110, "total_steps": 2679, "loss": 0.712, "learning_rate": 5.459485751017263e-07, "epoch": 2.360839160839161, "percentage": 78.76, "elapsed_time": "1 day, 10:34:25", "remaining_time": "9:19:24", "throughput": "17776.04", "total_tokens": 2212495360}
{"current_steps": 2111, "total_steps": 2679, "loss": 0.6735, "learning_rate": 5.441040689043148e-07, "epoch": 2.361958041958042, "percentage": 78.8, "elapsed_time": "1 day, 10:35:24", "remaining_time": "9:18:25", "throughput": "17775.90", "total_tokens": 2213543936}
{"current_steps": 2112, "total_steps": 2679, "loss": 0.6204, "learning_rate": 5.422623032939508e-07, "epoch": 2.363076923076923, "percentage": 78.84, "elapsed_time": "1 day, 10:36:18", "remaining_time": "9:17:25", "throughput": "17776.65", "total_tokens": 2214592512}
{"current_steps": 2113, "total_steps": 2679, "loss": 0.7015, "learning_rate": 5.404232808513027e-07, "epoch": 2.364195804195804, "percentage": 78.87, "elapsed_time": "1 day, 10:37:18", "remaining_time": "9:16:26", "throughput": "17776.51", "total_tokens": 2215641088}
{"current_steps": 2114, "total_steps": 2679, "loss": 0.6456, "learning_rate": 5.385870041531963e-07, "epoch": 2.365314685314685, "percentage": 78.91, "elapsed_time": "1 day, 10:38:17", "remaining_time": "9:15:27", "throughput": "17776.47", "total_tokens": 2216689664}
{"current_steps": 2115, "total_steps": 2679, "loss": 0.6823, "learning_rate": 5.367534757726079e-07, "epoch": 2.3664335664335665, "percentage": 78.95, "elapsed_time": "1 day, 10:39:17", "remaining_time": "9:14:28", "throughput": "17776.42", "total_tokens": 2217738240}
{"current_steps": 2116, "total_steps": 2679, "loss": 0.6688, "learning_rate": 5.349226982786632e-07, "epoch": 2.3675524475524474, "percentage": 78.98, "elapsed_time": "1 day, 10:40:17", "remaining_time": "9:13:30", "throughput": "17776.18", "total_tokens": 2218786816}
{"current_steps": 2117, "total_steps": 2679, "loss": 0.722, "learning_rate": 5.330946742366356e-07, "epoch": 2.368671328671329, "percentage": 79.02, "elapsed_time": "1 day, 10:41:17", "remaining_time": "9:12:31", "throughput": "17776.08", "total_tokens": 2219835392}
{"current_steps": 2118, "total_steps": 2679, "loss": 0.6293, "learning_rate": 5.312694062079385e-07, "epoch": 2.3697902097902097, "percentage": 79.06, "elapsed_time": "1 day, 10:42:11", "remaining_time": "9:11:30", "throughput": "17776.79", "total_tokens": 2220883968}
{"current_steps": 2119, "total_steps": 2679, "loss": 0.7684, "learning_rate": 5.294468967501248e-07, "epoch": 2.370909090909091, "percentage": 79.1, "elapsed_time": "1 day, 10:43:17", "remaining_time": "9:10:33", "throughput": "17775.85", "total_tokens": 2221932544}
{"current_steps": 2120, "total_steps": 2679, "loss": 0.6777, "learning_rate": 5.276271484168808e-07, "epoch": 2.372027972027972, "percentage": 79.13, "elapsed_time": "1 day, 10:44:13", "remaining_time": "9:09:33", "throughput": "17776.27", "total_tokens": 2222981120}
{"current_steps": 2121, "total_steps": 2679, "loss": 0.6338, "learning_rate": 5.258101637580238e-07, "epoch": 2.373146853146853, "percentage": 79.17, "elapsed_time": "1 day, 10:45:12", "remaining_time": "9:08:35", "throughput": "17776.22", "total_tokens": 2224029696}
{"current_steps": 2122, "total_steps": 2679, "loss": 0.6652, "learning_rate": 5.239959453195001e-07, "epoch": 2.3742657342657343, "percentage": 79.21, "elapsed_time": "1 day, 10:46:06", "remaining_time": "9:07:34", "throughput": "17776.94", "total_tokens": 2225078272}
{"current_steps": 2123, "total_steps": 2679, "loss": 0.6101, "learning_rate": 5.221844956433794e-07, "epoch": 2.375384615384615, "percentage": 79.25, "elapsed_time": "1 day, 10:47:00", "remaining_time": "9:06:34", "throughput": "17777.71", "total_tokens": 2226126848}
{"current_steps": 2124, "total_steps": 2679, "loss": 0.7176, "learning_rate": 5.203758172678522e-07, "epoch": 2.3765034965034966, "percentage": 79.28, "elapsed_time": "1 day, 10:48:01", "remaining_time": "9:05:36", "throughput": "17777.30", "total_tokens": 2227175424}
{"current_steps": 2125, "total_steps": 2679, "loss": 0.6639, "learning_rate": 5.185699127272243e-07, "epoch": 2.3776223776223775, "percentage": 79.32, "elapsed_time": "1 day, 10:48:59", "remaining_time": "9:04:36", "throughput": "17777.52", "total_tokens": 2228224000}
{"current_steps": 2126, "total_steps": 2679, "loss": 0.7719, "learning_rate": 5.167667845519167e-07, "epoch": 2.378741258741259, "percentage": 79.36, "elapsed_time": "1 day, 10:49:59", "remaining_time": "9:03:38", "throughput": "17777.33", "total_tokens": 2229272576}
{"current_steps": 2127, "total_steps": 2679, "loss": 0.7191, "learning_rate": 5.149664352684586e-07, "epoch": 2.37986013986014, "percentage": 79.4, "elapsed_time": "1 day, 10:51:00", "remaining_time": "9:02:39", "throughput": "17777.14", "total_tokens": 2230321152}
{"current_steps": 2128, "total_steps": 2679, "loss": 0.6281, "learning_rate": 5.131688673994867e-07, "epoch": 2.380979020979021, "percentage": 79.43, "elapsed_time": "1 day, 10:51:59", "remaining_time": "9:01:40", "throughput": "17777.01", "total_tokens": 2231369728}
{"current_steps": 2129, "total_steps": 2679, "loss": 0.7414, "learning_rate": 5.113740834637407e-07, "epoch": 2.382097902097902, "percentage": 79.47, "elapsed_time": "1 day, 10:53:00", "remaining_time": "9:00:42", "throughput": "17776.83", "total_tokens": 2232418304}
{"current_steps": 2130, "total_steps": 2679, "loss": 0.6552, "learning_rate": 5.095820859760572e-07, "epoch": 2.383216783216783, "percentage": 79.51, "elapsed_time": "1 day, 10:53:54", "remaining_time": "8:59:41", "throughput": "17777.45", "total_tokens": 2233466880}
{"current_steps": 2131, "total_steps": 2679, "loss": 0.7107, "learning_rate": 5.077928774473714e-07, "epoch": 2.3843356643356644, "percentage": 79.54, "elapsed_time": "1 day, 10:54:54", "remaining_time": "8:58:43", "throughput": "17777.34", "total_tokens": 2234515456}
{"current_steps": 2132, "total_steps": 2679, "loss": 0.6555, "learning_rate": 5.060064603847079e-07, "epoch": 2.3854545454545453, "percentage": 79.58, "elapsed_time": "1 day, 10:55:52", "remaining_time": "8:57:43", "throughput": "17777.56", "total_tokens": 2235564032}
{"current_steps": 2133, "total_steps": 2679, "loss": 0.617, "learning_rate": 5.042228372911815e-07, "epoch": 2.3865734265734266, "percentage": 79.62, "elapsed_time": "1 day, 10:56:51", "remaining_time": "8:56:45", "throughput": "17777.42", "total_tokens": 2236612608}
{"current_steps": 2134, "total_steps": 2679, "loss": 0.6716, "learning_rate": 5.024420106659928e-07, "epoch": 2.3876923076923076, "percentage": 79.66, "elapsed_time": "1 day, 10:57:46", "remaining_time": "8:55:44", "throughput": "17778.10", "total_tokens": 2237661184}
{"current_steps": 2135, "total_steps": 2679, "loss": 0.6374, "learning_rate": 5.006639830044219e-07, "epoch": 2.388811188811189, "percentage": 79.69, "elapsed_time": "1 day, 10:58:45", "remaining_time": "8:54:46", "throughput": "17778.00", "total_tokens": 2238709760}
{"current_steps": 2136, "total_steps": 2679, "loss": 0.7336, "learning_rate": 4.98888756797829e-07, "epoch": 2.38993006993007, "percentage": 79.73, "elapsed_time": "1 day, 10:59:45", "remaining_time": "8:53:47", "throughput": "17777.84", "total_tokens": 2239758336}
{"current_steps": 2137, "total_steps": 2679, "loss": 0.627, "learning_rate": 4.971163345336469e-07, "epoch": 2.391048951048951, "percentage": 79.77, "elapsed_time": "1 day, 11:00:40", "remaining_time": "8:52:47", "throughput": "17778.52", "total_tokens": 2240806912}
{"current_steps": 2138, "total_steps": 2679, "loss": 0.6792, "learning_rate": 4.953467186953814e-07, "epoch": 2.392167832167832, "percentage": 79.81, "elapsed_time": "1 day, 11:01:37", "remaining_time": "8:51:47", "throughput": "17778.78", "total_tokens": 2241855488}
{"current_steps": 2139, "total_steps": 2679, "loss": 0.5442, "learning_rate": 4.935799117626058e-07, "epoch": 2.393286713286713, "percentage": 79.84, "elapsed_time": "1 day, 11:02:26", "remaining_time": "8:50:46", "throughput": "17780.15", "total_tokens": 2242904064}
{"current_steps": 2140, "total_steps": 2679, "loss": 0.7343, "learning_rate": 4.918159162109559e-07, "epoch": 2.3944055944055944, "percentage": 79.88, "elapsed_time": "1 day, 11:03:32", "remaining_time": "8:49:48", "throughput": "17779.19", "total_tokens": 2243952640}
{"current_steps": 2141, "total_steps": 2679, "loss": 0.6794, "learning_rate": 4.900547345121304e-07, "epoch": 2.3955244755244753, "percentage": 79.92, "elapsed_time": "1 day, 11:04:31", "remaining_time": "8:48:50", "throughput": "17779.16", "total_tokens": 2245001216}
{"current_steps": 2142, "total_steps": 2679, "loss": 0.6673, "learning_rate": 4.882963691338832e-07, "epoch": 2.3966433566433567, "percentage": 79.96, "elapsed_time": "1 day, 11:05:31", "remaining_time": "8:47:51", "throughput": "17779.03", "total_tokens": 2246049792}
{"current_steps": 2143, "total_steps": 2679, "loss": 0.7118, "learning_rate": 4.865408225400234e-07, "epoch": 2.3977622377622376, "percentage": 79.99, "elapsed_time": "1 day, 11:06:31", "remaining_time": "8:46:52", "throughput": "17778.92", "total_tokens": 2247098368}
{"current_steps": 2144, "total_steps": 2679, "loss": 0.7726, "learning_rate": 4.847880971904106e-07, "epoch": 2.398881118881119, "percentage": 80.03, "elapsed_time": "1 day, 11:07:36", "remaining_time": "8:45:55", "throughput": "17778.00", "total_tokens": 2248146944}
{"current_steps": 2145, "total_steps": 2679, "loss": 0.7406, "learning_rate": 4.830381955409497e-07, "epoch": 2.4, "percentage": 80.07, "elapsed_time": "1 day, 11:08:36", "remaining_time": "8:44:56", "throughput": "17777.82", "total_tokens": 2249195520}
{"current_steps": 2146, "total_steps": 2679, "loss": 0.7886, "learning_rate": 4.812911200435913e-07, "epoch": 2.4011188811188813, "percentage": 80.1, "elapsed_time": "1 day, 11:09:47", "remaining_time": "8:44:00", "throughput": "17776.18", "total_tokens": 2250244096}
{"current_steps": 2147, "total_steps": 2679, "loss": 0.6341, "learning_rate": 4.795468731463232e-07, "epoch": 2.402237762237762, "percentage": 80.14, "elapsed_time": "1 day, 11:10:47", "remaining_time": "8:43:01", "throughput": "17776.08", "total_tokens": 2251292672}
{"current_steps": 2148, "total_steps": 2679, "loss": 0.5739, "learning_rate": 4.778054572931723e-07, "epoch": 2.403356643356643, "percentage": 80.18, "elapsed_time": "1 day, 11:11:36", "remaining_time": "8:42:00", "throughput": "17777.52", "total_tokens": 2252341248}
{"current_steps": 2149, "total_steps": 2679, "loss": 0.7629, "learning_rate": 4.7606687492419785e-07, "epoch": 2.4044755244755245, "percentage": 80.22, "elapsed_time": "1 day, 11:12:36", "remaining_time": "8:41:01", "throughput": "17777.37", "total_tokens": 2253389824}
{"current_steps": 2150, "total_steps": 2679, "loss": 0.7077, "learning_rate": 4.7433112847548893e-07, "epoch": 2.4055944055944054, "percentage": 80.25, "elapsed_time": "1 day, 11:13:35", "remaining_time": "8:40:02", "throughput": "17777.25", "total_tokens": 2254438400}
{"current_steps": 2151, "total_steps": 2679, "loss": 0.7129, "learning_rate": 4.725982203791607e-07, "epoch": 2.4067132867132868, "percentage": 80.29, "elapsed_time": "1 day, 11:14:35", "remaining_time": "8:39:03", "throughput": "17777.11", "total_tokens": 2255486976}
{"current_steps": 2152, "total_steps": 2679, "loss": 0.7221, "learning_rate": 4.708681530633499e-07, "epoch": 2.4078321678321677, "percentage": 80.33, "elapsed_time": "1 day, 11:15:35", "remaining_time": "8:38:05", "throughput": "17777.00", "total_tokens": 2256535552}
{"current_steps": 2153, "total_steps": 2679, "loss": 0.6069, "learning_rate": 4.691409289522156e-07, "epoch": 2.408951048951049, "percentage": 80.37, "elapsed_time": "1 day, 11:16:30", "remaining_time": "8:37:05", "throughput": "17777.54", "total_tokens": 2257584128}
{"current_steps": 2154, "total_steps": 2679, "loss": 0.5686, "learning_rate": 4.6741655046593087e-07, "epoch": 2.41006993006993, "percentage": 80.4, "elapsed_time": "1 day, 11:17:19", "remaining_time": "8:36:03", "throughput": "17778.93", "total_tokens": 2258632704}
{"current_steps": 2155, "total_steps": 2679, "loss": 0.7829, "learning_rate": 4.6569502002068336e-07, "epoch": 2.4111888111888113, "percentage": 80.44, "elapsed_time": "1 day, 11:18:20", "remaining_time": "8:35:05", "throughput": "17778.75", "total_tokens": 2259681280}
{"current_steps": 2156, "total_steps": 2679, "loss": 0.6782, "learning_rate": 4.639763400286679e-07, "epoch": 2.4123076923076923, "percentage": 80.48, "elapsed_time": "1 day, 11:19:19", "remaining_time": "8:34:06", "throughput": "17778.67", "total_tokens": 2260729856}
{"current_steps": 2157, "total_steps": 2679, "loss": 0.7621, "learning_rate": 4.622605128980862e-07, "epoch": 2.413426573426573, "percentage": 80.52, "elapsed_time": "1 day, 11:20:14", "remaining_time": "8:33:06", "throughput": "17779.28", "total_tokens": 2261778432}
{"current_steps": 2158, "total_steps": 2679, "loss": 0.7352, "learning_rate": 4.6054754103314255e-07, "epoch": 2.4145454545454546, "percentage": 80.55, "elapsed_time": "1 day, 11:21:19", "remaining_time": "8:32:08", "throughput": "17778.34", "total_tokens": 2262827008}
{"current_steps": 2159, "total_steps": 2679, "loss": 0.617, "learning_rate": 4.588374268340412e-07, "epoch": 2.4156643356643355, "percentage": 80.59, "elapsed_time": "1 day, 11:22:13", "remaining_time": "8:31:08", "throughput": "17779.08", "total_tokens": 2263875584}
{"current_steps": 2160, "total_steps": 2679, "loss": 0.8254, "learning_rate": 4.5713017269698207e-07, "epoch": 2.416783216783217, "percentage": 80.63, "elapsed_time": "1 day, 11:23:14", "remaining_time": "8:30:10", "throughput": "17778.84", "total_tokens": 2264924160}
{"current_steps": 2161, "total_steps": 2679, "loss": 0.7123, "learning_rate": 4.5542578101415576e-07, "epoch": 2.4179020979020978, "percentage": 80.66, "elapsed_time": "1 day, 11:24:14", "remaining_time": "8:29:11", "throughput": "17778.66", "total_tokens": 2265972736}
{"current_steps": 2162, "total_steps": 2679, "loss": 0.6617, "learning_rate": 4.5372425417374466e-07, "epoch": 2.419020979020979, "percentage": 80.7, "elapsed_time": "1 day, 11:25:08", "remaining_time": "8:28:11", "throughput": "17779.39", "total_tokens": 2267021312}
{"current_steps": 2163, "total_steps": 2679, "loss": 0.6397, "learning_rate": 4.5202559455991473e-07, "epoch": 2.42013986013986, "percentage": 80.74, "elapsed_time": "1 day, 11:26:08", "remaining_time": "8:27:12", "throughput": "17779.21", "total_tokens": 2268069888}
{"current_steps": 2164, "total_steps": 2679, "loss": 0.5518, "learning_rate": 4.5032980455281596e-07, "epoch": 2.4212587412587414, "percentage": 80.78, "elapsed_time": "1 day, 11:26:56", "remaining_time": "8:26:10", "throughput": "17780.72", "total_tokens": 2269118464}
{"current_steps": 2165, "total_steps": 2679, "loss": 0.7122, "learning_rate": 4.48636886528577e-07, "epoch": 2.4223776223776223, "percentage": 80.81, "elapsed_time": "1 day, 11:28:01", "remaining_time": "8:25:13", "throughput": "17779.90", "total_tokens": 2270167040}
{"current_steps": 2166, "total_steps": 2679, "loss": 0.6848, "learning_rate": 4.469468428593016e-07, "epoch": 2.4234965034965033, "percentage": 80.85, "elapsed_time": "1 day, 11:29:01", "remaining_time": "8:24:14", "throughput": "17779.83", "total_tokens": 2271215616}
{"current_steps": 2167, "total_steps": 2679, "loss": 0.5997, "learning_rate": 4.4525967591306757e-07, "epoch": 2.4246153846153846, "percentage": 80.89, "elapsed_time": "1 day, 11:29:54", "remaining_time": "8:23:14", "throughput": "17780.59", "total_tokens": 2272264192}
{"current_steps": 2168, "total_steps": 2679, "loss": 0.7304, "learning_rate": 4.4357538805391953e-07, "epoch": 2.425734265734266, "percentage": 80.93, "elapsed_time": "1 day, 11:30:54", "remaining_time": "8:22:15", "throughput": "17780.53", "total_tokens": 2273312768}
{"current_steps": 2169, "total_steps": 2679, "loss": 0.6174, "learning_rate": 4.418939816418699e-07, "epoch": 2.426853146853147, "percentage": 80.96, "elapsed_time": "1 day, 11:31:48", "remaining_time": "8:21:15", "throughput": "17781.17", "total_tokens": 2274361344}
{"current_steps": 2170, "total_steps": 2679, "loss": 0.6731, "learning_rate": 4.4021545903289357e-07, "epoch": 2.427972027972028, "percentage": 81.0, "elapsed_time": "1 day, 11:32:44", "remaining_time": "8:20:15", "throughput": "17781.51", "total_tokens": 2275409920}
{"current_steps": 2171, "total_steps": 2679, "loss": 0.6229, "learning_rate": 4.3853982257892335e-07, "epoch": 2.429090909090909, "percentage": 81.04, "elapsed_time": "1 day, 11:33:38", "remaining_time": "8:19:15", "throughput": "17782.20", "total_tokens": 2276458496}
{"current_steps": 2172, "total_steps": 2679, "loss": 0.7666, "learning_rate": 4.3686707462784945e-07, "epoch": 2.43020979020979, "percentage": 81.08, "elapsed_time": "1 day, 11:34:39", "remaining_time": "8:18:17", "throughput": "17781.95", "total_tokens": 2277507072}
{"current_steps": 2173, "total_steps": 2679, "loss": 0.6355, "learning_rate": 4.3519721752351305e-07, "epoch": 2.4313286713286715, "percentage": 81.11, "elapsed_time": "1 day, 11:35:34", "remaining_time": "8:17:17", "throughput": "17782.51", "total_tokens": 2278555648}
{"current_steps": 2174, "total_steps": 2679, "loss": 0.6604, "learning_rate": 4.3353025360570623e-07, "epoch": 2.4324475524475524, "percentage": 81.15, "elapsed_time": "1 day, 11:36:34", "remaining_time": "8:16:18", "throughput": "17782.38", "total_tokens": 2279604224}
{"current_steps": 2175, "total_steps": 2679, "loss": 0.7101, "learning_rate": 4.3186618521016745e-07, "epoch": 2.4335664335664333, "percentage": 81.19, "elapsed_time": "1 day, 11:37:34", "remaining_time": "8:15:19", "throughput": "17782.29", "total_tokens": 2280652800}
{"current_steps": 2176, "total_steps": 2679, "loss": 0.6598, "learning_rate": 4.302050146685757e-07, "epoch": 2.4346853146853147, "percentage": 81.22, "elapsed_time": "1 day, 11:38:32", "remaining_time": "8:14:20", "throughput": "17782.40", "total_tokens": 2281701376}
{"current_steps": 2177, "total_steps": 2679, "loss": 0.5535, "learning_rate": 4.2854674430855224e-07, "epoch": 2.435804195804196, "percentage": 81.26, "elapsed_time": "1 day, 11:39:20", "remaining_time": "8:13:19", "throughput": "17783.83", "total_tokens": 2282749952}
{"current_steps": 2178, "total_steps": 2679, "loss": 0.575, "learning_rate": 4.26891376453652e-07, "epoch": 2.436923076923077, "percentage": 81.3, "elapsed_time": "1 day, 11:40:08", "remaining_time": "8:12:17", "throughput": "17785.37", "total_tokens": 2283798528}
{"current_steps": 2179, "total_steps": 2679, "loss": 0.6521, "learning_rate": 4.2523891342336506e-07, "epoch": 2.438041958041958, "percentage": 81.34, "elapsed_time": "1 day, 11:41:05", "remaining_time": "8:11:17", "throughput": "17785.75", "total_tokens": 2284847104}
{"current_steps": 2180, "total_steps": 2679, "loss": 0.6924, "learning_rate": 4.2358935753311125e-07, "epoch": 2.4391608391608393, "percentage": 81.37, "elapsed_time": "1 day, 11:42:04", "remaining_time": "8:10:19", "throughput": "17785.62", "total_tokens": 2285895680}
{"current_steps": 2181, "total_steps": 2679, "loss": 0.7124, "learning_rate": 4.219427110942348e-07, "epoch": 2.44027972027972, "percentage": 81.41, "elapsed_time": "1 day, 11:43:10", "remaining_time": "8:09:21", "throughput": "17784.77", "total_tokens": 2286944256}
{"current_steps": 2182, "total_steps": 2679, "loss": 0.7705, "learning_rate": 4.2029897641400584e-07, "epoch": 2.4413986013986015, "percentage": 81.45, "elapsed_time": "1 day, 11:44:16", "remaining_time": "8:08:24", "throughput": "17783.80", "total_tokens": 2287992832}
{"current_steps": 2183, "total_steps": 2679, "loss": 0.6574, "learning_rate": 4.186581557956124e-07, "epoch": 2.4425174825174825, "percentage": 81.49, "elapsed_time": "1 day, 11:45:09", "remaining_time": "8:07:24", "throughput": "17784.49", "total_tokens": 2289041408}
{"current_steps": 2184, "total_steps": 2679, "loss": 0.6799, "learning_rate": 4.170202515381605e-07, "epoch": 2.443636363636364, "percentage": 81.52, "elapsed_time": "1 day, 11:46:10", "remaining_time": "8:06:25", "throughput": "17784.34", "total_tokens": 2290089984}
{"current_steps": 2185, "total_steps": 2679, "loss": 0.6479, "learning_rate": 4.153852659366697e-07, "epoch": 2.4447552447552447, "percentage": 81.56, "elapsed_time": "1 day, 11:47:10", "remaining_time": "8:05:26", "throughput": "17784.20", "total_tokens": 2291138560}
{"current_steps": 2186, "total_steps": 2679, "loss": 0.589, "learning_rate": 4.137532012820708e-07, "epoch": 2.445874125874126, "percentage": 81.6, "elapsed_time": "1 day, 11:48:03", "remaining_time": "8:04:26", "throughput": "17784.92", "total_tokens": 2292187136}
{"current_steps": 2187, "total_steps": 2679, "loss": 0.617, "learning_rate": 4.1212405986119975e-07, "epoch": 2.446993006993007, "percentage": 81.63, "elapsed_time": "1 day, 11:48:58", "remaining_time": "8:03:26", "throughput": "17785.50", "total_tokens": 2293235712}
{"current_steps": 2188, "total_steps": 2679, "loss": 0.6502, "learning_rate": 4.104978439567972e-07, "epoch": 2.448111888111888, "percentage": 81.67, "elapsed_time": "1 day, 11:49:53", "remaining_time": "8:02:26", "throughput": "17786.09", "total_tokens": 2294284288}
{"current_steps": 2189, "total_steps": 2679, "loss": 0.716, "learning_rate": 4.0887455584750547e-07, "epoch": 2.4492307692307693, "percentage": 81.71, "elapsed_time": "1 day, 11:50:53", "remaining_time": "8:01:28", "throughput": "17785.93", "total_tokens": 2295332864}
{"current_steps": 2190, "total_steps": 2679, "loss": 0.5442, "learning_rate": 4.072541978078642e-07, "epoch": 2.4503496503496502, "percentage": 81.75, "elapsed_time": "1 day, 11:51:41", "remaining_time": "8:00:26", "throughput": "17787.40", "total_tokens": 2296381440}
{"current_steps": 2191, "total_steps": 2679, "loss": 0.6587, "learning_rate": 4.0563677210830763e-07, "epoch": 2.4514685314685316, "percentage": 81.78, "elapsed_time": "1 day, 11:52:41", "remaining_time": "7:59:27", "throughput": "17787.32", "total_tokens": 2297430016}
{"current_steps": 2192, "total_steps": 2679, "loss": 0.6286, "learning_rate": 4.0402228101516036e-07, "epoch": 2.4525874125874125, "percentage": 81.82, "elapsed_time": "1 day, 11:53:35", "remaining_time": "7:58:27", "throughput": "17787.96", "total_tokens": 2298478592}
{"current_steps": 2193, "total_steps": 2679, "loss": 0.69, "learning_rate": 4.0241072679063437e-07, "epoch": 2.453706293706294, "percentage": 81.86, "elapsed_time": "1 day, 11:54:40", "remaining_time": "7:57:30", "throughput": "17787.18", "total_tokens": 2299527168}
{"current_steps": 2194, "total_steps": 2679, "loss": 0.7809, "learning_rate": 4.0080211169282956e-07, "epoch": 2.454825174825175, "percentage": 81.9, "elapsed_time": "1 day, 11:55:45", "remaining_time": "7:56:32", "throughput": "17786.31", "total_tokens": 2300575744}
{"current_steps": 2195, "total_steps": 2679, "loss": 0.7086, "learning_rate": 3.991964379757232e-07, "epoch": 2.455944055944056, "percentage": 81.93, "elapsed_time": "1 day, 11:56:48", "remaining_time": "7:55:34", "throughput": "17785.67", "total_tokens": 2301624320}
{"current_steps": 2196, "total_steps": 2679, "loss": 0.5642, "learning_rate": 3.975937078891762e-07, "epoch": 2.457062937062937, "percentage": 81.97, "elapsed_time": "1 day, 11:57:37", "remaining_time": "7:54:33", "throughput": "17787.12", "total_tokens": 2302672896}
{"current_steps": 2197, "total_steps": 2679, "loss": 0.5554, "learning_rate": 3.959939236789212e-07, "epoch": 2.458181818181818, "percentage": 82.01, "elapsed_time": "1 day, 11:58:25", "remaining_time": "7:53:32", "throughput": "17788.57", "total_tokens": 2303721472}
{"current_steps": 2198, "total_steps": 2679, "loss": 0.6969, "learning_rate": 3.9439708758656323e-07, "epoch": 2.4593006993006994, "percentage": 82.05, "elapsed_time": "1 day, 11:59:25", "remaining_time": "7:52:33", "throughput": "17788.45", "total_tokens": 2304770048}
{"current_steps": 2199, "total_steps": 2679, "loss": 0.7139, "learning_rate": 3.9280320184957864e-07, "epoch": 2.4604195804195803, "percentage": 82.08, "elapsed_time": "1 day, 12:00:25", "remaining_time": "7:51:34", "throughput": "17788.37", "total_tokens": 2305818624}
{"current_steps": 2200, "total_steps": 2679, "loss": 0.6424, "learning_rate": 3.912122687013065e-07, "epoch": 2.4615384615384617, "percentage": 82.12, "elapsed_time": "1 day, 12:01:23", "remaining_time": "7:50:35", "throughput": "17788.45", "total_tokens": 2306867200}
{"current_steps": 2201, "total_steps": 2679, "loss": 0.6316, "learning_rate": 3.896242903709532e-07, "epoch": 2.4626573426573426, "percentage": 82.16, "elapsed_time": "1 day, 12:02:17", "remaining_time": "7:49:35", "throughput": "17789.16", "total_tokens": 2307915776}
{"current_steps": 2202, "total_steps": 2679, "loss": 0.75, "learning_rate": 3.8803926908358047e-07, "epoch": 2.463776223776224, "percentage": 82.19, "elapsed_time": "1 day, 12:03:19", "remaining_time": "7:48:37", "throughput": "17788.69", "total_tokens": 2308964352}
{"current_steps": 2203, "total_steps": 2679, "loss": 0.6467, "learning_rate": 3.8645720706010997e-07, "epoch": 2.464895104895105, "percentage": 82.23, "elapsed_time": "1 day, 12:04:14", "remaining_time": "7:47:37", "throughput": "17789.31", "total_tokens": 2310012928}
{"current_steps": 2204, "total_steps": 2679, "loss": 0.6384, "learning_rate": 3.8487810651731463e-07, "epoch": 2.4660139860139862, "percentage": 82.27, "elapsed_time": "1 day, 12:05:08", "remaining_time": "7:46:37", "throughput": "17789.98", "total_tokens": 2311061504}
{"current_steps": 2205, "total_steps": 2679, "loss": 0.6291, "learning_rate": 3.8330196966781723e-07, "epoch": 2.467132867132867, "percentage": 82.31, "elapsed_time": "1 day, 12:06:02", "remaining_time": "7:45:37", "throughput": "17790.66", "total_tokens": 2312110080}
{"current_steps": 2206, "total_steps": 2679, "loss": 0.7577, "learning_rate": 3.817287987200918e-07, "epoch": 2.468251748251748, "percentage": 82.34, "elapsed_time": "1 day, 12:06:59", "remaining_time": "7:44:38", "throughput": "17790.88", "total_tokens": 2313158656}
{"current_steps": 2207, "total_steps": 2679, "loss": 0.6575, "learning_rate": 3.8015859587845233e-07, "epoch": 2.4693706293706295, "percentage": 82.38, "elapsed_time": "1 day, 12:07:59", "remaining_time": "7:43:39", "throughput": "17790.77", "total_tokens": 2314207232}
{"current_steps": 2208, "total_steps": 2679, "loss": 0.6444, "learning_rate": 3.78591363343056e-07, "epoch": 2.4704895104895104, "percentage": 82.42, "elapsed_time": "1 day, 12:09:04", "remaining_time": "7:42:41", "throughput": "17789.92", "total_tokens": 2315255808}
{"current_steps": 2209, "total_steps": 2679, "loss": 0.6611, "learning_rate": 3.7702710330989765e-07, "epoch": 2.4716083916083917, "percentage": 82.46, "elapsed_time": "1 day, 12:10:03", "remaining_time": "7:41:42", "throughput": "17789.85", "total_tokens": 2316304384}
{"current_steps": 2210, "total_steps": 2679, "loss": 0.6607, "learning_rate": 3.7546581797080567e-07, "epoch": 2.4727272727272727, "percentage": 82.49, "elapsed_time": "1 day, 12:11:03", "remaining_time": "7:40:44", "throughput": "17789.72", "total_tokens": 2317352960}
{"current_steps": 2211, "total_steps": 2679, "loss": 0.708, "learning_rate": 3.739075095134437e-07, "epoch": 2.473846153846154, "percentage": 82.53, "elapsed_time": "1 day, 12:12:02", "remaining_time": "7:39:45", "throughput": "17789.66", "total_tokens": 2318401536}
{"current_steps": 2212, "total_steps": 2679, "loss": 0.7428, "learning_rate": 3.7235218012130067e-07, "epoch": 2.474965034965035, "percentage": 82.57, "elapsed_time": "1 day, 12:13:02", "remaining_time": "7:38:46", "throughput": "17789.57", "total_tokens": 2319450112}
{"current_steps": 2213, "total_steps": 2679, "loss": 0.6259, "learning_rate": 3.707998319736936e-07, "epoch": 2.4760839160839163, "percentage": 82.61, "elapsed_time": "1 day, 12:13:56", "remaining_time": "7:37:46", "throughput": "17790.22", "total_tokens": 2320498688}
{"current_steps": 2214, "total_steps": 2679, "loss": 0.6235, "learning_rate": 3.692504672457606e-07, "epoch": 2.4772027972027972, "percentage": 82.64, "elapsed_time": "1 day, 12:14:49", "remaining_time": "7:36:46", "throughput": "17791.08", "total_tokens": 2321547264}
{"current_steps": 2215, "total_steps": 2679, "loss": 0.6792, "learning_rate": 3.677040881084609e-07, "epoch": 2.478321678321678, "percentage": 82.68, "elapsed_time": "1 day, 12:15:43", "remaining_time": "7:35:46", "throughput": "17791.72", "total_tokens": 2322595840}
{"current_steps": 2216, "total_steps": 2679, "loss": 0.6478, "learning_rate": 3.6616069672856885e-07, "epoch": 2.4794405594405595, "percentage": 82.72, "elapsed_time": "1 day, 12:16:37", "remaining_time": "7:34:46", "throughput": "17792.42", "total_tokens": 2323644416}
{"current_steps": 2217, "total_steps": 2679, "loss": 0.6594, "learning_rate": 3.6462029526867335e-07, "epoch": 2.4805594405594404, "percentage": 82.75, "elapsed_time": "1 day, 12:17:37", "remaining_time": "7:33:47", "throughput": "17792.30", "total_tokens": 2324692992}
{"current_steps": 2218, "total_steps": 2679, "loss": 0.574, "learning_rate": 3.6308288588717434e-07, "epoch": 2.481678321678322, "percentage": 82.79, "elapsed_time": "1 day, 12:18:35", "remaining_time": "7:32:48", "throughput": "17792.44", "total_tokens": 2325741568}
{"current_steps": 2219, "total_steps": 2679, "loss": 0.6648, "learning_rate": 3.615484707382777e-07, "epoch": 2.4827972027972027, "percentage": 82.83, "elapsed_time": "1 day, 12:19:34", "remaining_time": "7:31:49", "throughput": "17792.42", "total_tokens": 2326790144}
{"current_steps": 2220, "total_steps": 2679, "loss": 0.601, "learning_rate": 3.600170519719956e-07, "epoch": 2.483916083916084, "percentage": 82.87, "elapsed_time": "1 day, 12:20:27", "remaining_time": "7:30:49", "throughput": "17793.13", "total_tokens": 2327838720}
{"current_steps": 2221, "total_steps": 2679, "loss": 0.6916, "learning_rate": 3.584886317341396e-07, "epoch": 2.485034965034965, "percentage": 82.9, "elapsed_time": "1 day, 12:21:25", "remaining_time": "7:29:50", "throughput": "17793.29", "total_tokens": 2328887296}
{"current_steps": 2222, "total_steps": 2679, "loss": 0.6992, "learning_rate": 3.5696321216632164e-07, "epoch": 2.4861538461538464, "percentage": 82.94, "elapsed_time": "1 day, 12:22:27", "remaining_time": "7:28:52", "throughput": "17792.84", "total_tokens": 2329935872}
{"current_steps": 2223, "total_steps": 2679, "loss": 0.7376, "learning_rate": 3.5544079540594884e-07, "epoch": 2.4872727272727273, "percentage": 82.98, "elapsed_time": "1 day, 12:23:33", "remaining_time": "7:27:54", "throughput": "17791.96", "total_tokens": 2330984448}
{"current_steps": 2224, "total_steps": 2679, "loss": 0.5345, "learning_rate": 3.5392138358621956e-07, "epoch": 2.4883916083916082, "percentage": 83.02, "elapsed_time": "1 day, 12:24:27", "remaining_time": "7:26:54", "throughput": "17792.67", "total_tokens": 2332033024}
{"current_steps": 2225, "total_steps": 2679, "loss": 0.6042, "learning_rate": 3.5240497883612333e-07, "epoch": 2.4895104895104896, "percentage": 83.05, "elapsed_time": "1 day, 12:25:21", "remaining_time": "7:25:54", "throughput": "17793.30", "total_tokens": 2333081600}
{"current_steps": 2226, "total_steps": 2679, "loss": 0.5901, "learning_rate": 3.5089158328043383e-07, "epoch": 2.4906293706293705, "percentage": 83.09, "elapsed_time": "1 day, 12:26:15", "remaining_time": "7:24:54", "throughput": "17793.92", "total_tokens": 2334130176}
{"current_steps": 2227, "total_steps": 2679, "loss": 0.6453, "learning_rate": 3.4938119903971195e-07, "epoch": 2.491748251748252, "percentage": 83.13, "elapsed_time": "1 day, 12:27:09", "remaining_time": "7:23:54", "throughput": "17794.61", "total_tokens": 2335178752}
{"current_steps": 2228, "total_steps": 2679, "loss": 0.8318, "learning_rate": 3.4787382823029626e-07, "epoch": 2.492867132867133, "percentage": 83.17, "elapsed_time": "1 day, 12:28:18", "remaining_time": "7:22:57", "throughput": "17793.21", "total_tokens": 2336227328}
{"current_steps": 2229, "total_steps": 2679, "loss": 0.669, "learning_rate": 3.4636947296430274e-07, "epoch": 2.493986013986014, "percentage": 83.2, "elapsed_time": "1 day, 12:29:18", "remaining_time": "7:21:59", "throughput": "17793.09", "total_tokens": 2337275904}
{"current_steps": 2230, "total_steps": 2679, "loss": 0.8892, "learning_rate": 3.4486813534962444e-07, "epoch": 2.495104895104895, "percentage": 83.24, "elapsed_time": "1 day, 12:30:30", "remaining_time": "7:21:02", "throughput": "17791.35", "total_tokens": 2338324480}
{"current_steps": 2231, "total_steps": 2679, "loss": 0.7369, "learning_rate": 3.43369817489923e-07, "epoch": 2.4962237762237764, "percentage": 83.28, "elapsed_time": "1 day, 12:31:30", "remaining_time": "7:20:04", "throughput": "17791.24", "total_tokens": 2339373056}
{"current_steps": 2232, "total_steps": 2679, "loss": 0.7378, "learning_rate": 3.418745214846314e-07, "epoch": 2.4973426573426574, "percentage": 83.31, "elapsed_time": "1 day, 12:32:29", "remaining_time": "7:19:05", "throughput": "17791.13", "total_tokens": 2340421632}
{"current_steps": 2233, "total_steps": 2679, "loss": 0.6418, "learning_rate": 3.40382249428948e-07, "epoch": 2.4984615384615383, "percentage": 83.35, "elapsed_time": "1 day, 12:33:23", "remaining_time": "7:18:05", "throughput": "17791.79", "total_tokens": 2341470208}
{"current_steps": 2234, "total_steps": 2679, "loss": 0.6464, "learning_rate": 3.388930034138321e-07, "epoch": 2.4995804195804197, "percentage": 83.39, "elapsed_time": "1 day, 12:34:17", "remaining_time": "7:17:05", "throughput": "17792.54", "total_tokens": 2342518784}
{"current_steps": 2235, "total_steps": 2679, "loss": 0.5521, "learning_rate": 3.374067855260055e-07, "epoch": 2.5006993006993006, "percentage": 83.43, "elapsed_time": "1 day, 12:35:06", "remaining_time": "7:16:04", "throughput": "17793.88", "total_tokens": 2343567360}
{"current_steps": 2236, "total_steps": 2679, "loss": 0.6952, "learning_rate": 3.3592359784794497e-07, "epoch": 2.501818181818182, "percentage": 83.46, "elapsed_time": "1 day, 12:36:06", "remaining_time": "7:15:05", "throughput": "17793.74", "total_tokens": 2344615936}
{"current_steps": 2237, "total_steps": 2679, "loss": 0.6042, "learning_rate": 3.344434424578824e-07, "epoch": 2.502937062937063, "percentage": 83.5, "elapsed_time": "1 day, 12:37:00", "remaining_time": "7:14:05", "throughput": "17794.36", "total_tokens": 2345664512}
{"current_steps": 2238, "total_steps": 2679, "loss": 0.6093, "learning_rate": 3.3296632142980097e-07, "epoch": 2.504055944055944, "percentage": 83.54, "elapsed_time": "1 day, 12:37:52", "remaining_time": "7:13:05", "throughput": "17795.36", "total_tokens": 2346713088}
{"current_steps": 2239, "total_steps": 2679, "loss": 0.5873, "learning_rate": 3.314922368334322e-07, "epoch": 2.505174825174825, "percentage": 83.58, "elapsed_time": "1 day, 12:38:46", "remaining_time": "7:12:05", "throughput": "17796.02", "total_tokens": 2347761664}
{"current_steps": 2240, "total_steps": 2679, "loss": 0.806, "learning_rate": 3.300211907342521e-07, "epoch": 2.5062937062937065, "percentage": 83.61, "elapsed_time": "1 day, 12:39:52", "remaining_time": "7:11:08", "throughput": "17795.05", "total_tokens": 2348810240}
{"current_steps": 2241, "total_steps": 2679, "loss": 0.652, "learning_rate": 3.2855318519347924e-07, "epoch": 2.5074125874125874, "percentage": 83.65, "elapsed_time": "1 day, 12:40:52", "remaining_time": "7:10:09", "throughput": "17794.84", "total_tokens": 2349858816}
{"current_steps": 2242, "total_steps": 2679, "loss": 0.6257, "learning_rate": 3.270882222680727e-07, "epoch": 2.5085314685314684, "percentage": 83.69, "elapsed_time": "1 day, 12:41:46", "remaining_time": "7:09:09", "throughput": "17795.55", "total_tokens": 2350907392}
{"current_steps": 2243, "total_steps": 2679, "loss": 0.5849, "learning_rate": 3.2562630401072796e-07, "epoch": 2.5096503496503497, "percentage": 83.73, "elapsed_time": "1 day, 12:42:40", "remaining_time": "7:08:09", "throughput": "17796.24", "total_tokens": 2351955968}
{"current_steps": 2244, "total_steps": 2679, "loss": 0.7776, "learning_rate": 3.241674324698743e-07, "epoch": 2.5107692307692306, "percentage": 83.76, "elapsed_time": "1 day, 12:43:41", "remaining_time": "7:07:11", "throughput": "17795.96", "total_tokens": 2353004544}
{"current_steps": 2245, "total_steps": 2679, "loss": 0.646, "learning_rate": 3.227116096896718e-07, "epoch": 2.511888111888112, "percentage": 83.8, "elapsed_time": "1 day, 12:44:34", "remaining_time": "7:06:11", "throughput": "17796.66", "total_tokens": 2354053120}
{"current_steps": 2246, "total_steps": 2679, "loss": 0.6269, "learning_rate": 3.2125883771000787e-07, "epoch": 2.513006993006993, "percentage": 83.84, "elapsed_time": "1 day, 12:45:28", "remaining_time": "7:05:11", "throughput": "17797.33", "total_tokens": 2355101696}
{"current_steps": 2247, "total_steps": 2679, "loss": 0.6682, "learning_rate": 3.198091185664964e-07, "epoch": 2.514125874125874, "percentage": 83.87, "elapsed_time": "1 day, 12:46:27", "remaining_time": "7:04:12", "throughput": "17797.31", "total_tokens": 2356150272}
{"current_steps": 2248, "total_steps": 2679, "loss": 0.642, "learning_rate": 3.183624542904734e-07, "epoch": 2.515244755244755, "percentage": 83.91, "elapsed_time": "1 day, 12:47:27", "remaining_time": "7:03:13", "throughput": "17797.20", "total_tokens": 2357198848}
{"current_steps": 2249, "total_steps": 2679, "loss": 0.7413, "learning_rate": 3.169188469089945e-07, "epoch": 2.5163636363636366, "percentage": 83.95, "elapsed_time": "1 day, 12:48:33", "remaining_time": "7:02:16", "throughput": "17796.28", "total_tokens": 2358247424}
{"current_steps": 2250, "total_steps": 2679, "loss": 0.7462, "learning_rate": 3.1547829844483125e-07, "epoch": 2.5174825174825175, "percentage": 83.99, "elapsed_time": "1 day, 12:49:38", "remaining_time": "7:01:18", "throughput": "17795.45", "total_tokens": 2359296000}
{"current_steps": 2250, "total_steps": 2679, "eval_loss": 0.7195046544075012, "epoch": 2.5174825174825175, "percentage": 83.99, "elapsed_time": "1 day, 12:53:45", "remaining_time": "7:02:05", "throughput": "17762.38", "total_tokens": 2359296000}
{"current_steps": 2251, "total_steps": 2679, "loss": 0.7276, "learning_rate": 3.1404081091647027e-07, "epoch": 2.5186013986013984, "percentage": 84.02, "elapsed_time": "1 day, 12:55:28", "remaining_time": "7:01:14", "throughput": "17756.55", "total_tokens": 2360344576}
{"current_steps": 2252, "total_steps": 2679, "loss": 0.6048, "learning_rate": 3.126063863381076e-07, "epoch": 2.51972027972028, "percentage": 84.06, "elapsed_time": "1 day, 12:56:23", "remaining_time": "7:00:14", "throughput": "17757.10", "total_tokens": 2361393152}
{"current_steps": 2253, "total_steps": 2679, "loss": 0.6247, "learning_rate": 3.111750267196492e-07, "epoch": 2.5208391608391607, "percentage": 84.1, "elapsed_time": "1 day, 12:57:17", "remaining_time": "6:59:14", "throughput": "17757.74", "total_tokens": 2362441728}
{"current_steps": 2254, "total_steps": 2679, "loss": 0.7673, "learning_rate": 3.097467340667057e-07, "epoch": 2.521958041958042, "percentage": 84.14, "elapsed_time": "1 day, 12:58:23", "remaining_time": "6:58:17", "throughput": "17756.85", "total_tokens": 2363490304}
{"current_steps": 2255, "total_steps": 2679, "loss": 0.7936, "learning_rate": 3.083215103805895e-07, "epoch": 2.523076923076923, "percentage": 84.17, "elapsed_time": "1 day, 12:59:29", "remaining_time": "6:57:19", "throughput": "17755.91", "total_tokens": 2364538880}
{"current_steps": 2256, "total_steps": 2679, "loss": 0.619, "learning_rate": 3.068993576583149e-07, "epoch": 2.5241958041958044, "percentage": 84.21, "elapsed_time": "1 day, 13:00:23", "remaining_time": "6:56:19", "throughput": "17756.48", "total_tokens": 2365587456}
{"current_steps": 2257, "total_steps": 2679, "loss": 0.6772, "learning_rate": 3.0548027789259057e-07, "epoch": 2.5253146853146853, "percentage": 84.25, "elapsed_time": "1 day, 13:01:23", "remaining_time": "6:55:20", "throughput": "17756.40", "total_tokens": 2366636032}
{"current_steps": 2258, "total_steps": 2679, "loss": 0.6342, "learning_rate": 3.04064273071821e-07, "epoch": 2.5264335664335666, "percentage": 84.29, "elapsed_time": "1 day, 13:02:14", "remaining_time": "6:54:20", "throughput": "17757.45", "total_tokens": 2367684608}
{"current_steps": 2259, "total_steps": 2679, "loss": 0.8086, "learning_rate": 3.0265134518010274e-07, "epoch": 2.5275524475524476, "percentage": 84.32, "elapsed_time": "1 day, 13:03:23", "remaining_time": "6:53:22", "throughput": "17756.13", "total_tokens": 2368733184}
{"current_steps": 2260, "total_steps": 2679, "loss": 0.6255, "learning_rate": 3.012414961972185e-07, "epoch": 2.5286713286713285, "percentage": 84.36, "elapsed_time": "1 day, 13:04:17", "remaining_time": "6:52:22", "throughput": "17756.82", "total_tokens": 2369781760}
{"current_steps": 2261, "total_steps": 2679, "loss": 0.6212, "learning_rate": 2.9983472809863996e-07, "epoch": 2.52979020979021, "percentage": 84.4, "elapsed_time": "1 day, 13:05:11", "remaining_time": "6:51:22", "throughput": "17757.51", "total_tokens": 2370830336}
{"current_steps": 2262, "total_steps": 2679, "loss": 0.6406, "learning_rate": 2.9843104285551844e-07, "epoch": 2.5309090909090908, "percentage": 84.43, "elapsed_time": "1 day, 13:06:06", "remaining_time": "6:50:23", "throughput": "17758.04", "total_tokens": 2371878912}
{"current_steps": 2263, "total_steps": 2679, "loss": 0.7093, "learning_rate": 2.970304424346887e-07, "epoch": 2.532027972027972, "percentage": 84.47, "elapsed_time": "1 day, 13:07:06", "remaining_time": "6:49:24", "throughput": "17757.90", "total_tokens": 2372927488}
{"current_steps": 2264, "total_steps": 2679, "loss": 0.6845, "learning_rate": 2.95632928798662e-07, "epoch": 2.533146853146853, "percentage": 84.51, "elapsed_time": "1 day, 13:08:06", "remaining_time": "6:48:25", "throughput": "17757.79", "total_tokens": 2373976064}
{"current_steps": 2265, "total_steps": 2679, "loss": 0.6001, "learning_rate": 2.942385039056231e-07, "epoch": 2.5342657342657344, "percentage": 84.55, "elapsed_time": "1 day, 13:09:00", "remaining_time": "6:47:25", "throughput": "17758.52", "total_tokens": 2375024640}
{"current_steps": 2266, "total_steps": 2679, "loss": 0.6591, "learning_rate": 2.9284716970943094e-07, "epoch": 2.5353846153846153, "percentage": 84.58, "elapsed_time": "1 day, 13:09:59", "remaining_time": "6:46:26", "throughput": "17758.42", "total_tokens": 2376073216}
{"current_steps": 2267, "total_steps": 2679, "loss": 0.6419, "learning_rate": 2.9145892815961194e-07, "epoch": 2.5365034965034967, "percentage": 84.62, "elapsed_time": "1 day, 13:10:53", "remaining_time": "6:45:26", "throughput": "17759.09", "total_tokens": 2377121792}
{"current_steps": 2268, "total_steps": 2679, "loss": 0.7377, "learning_rate": 2.9007378120136044e-07, "epoch": 2.5376223776223776, "percentage": 84.66, "elapsed_time": "1 day, 13:11:58", "remaining_time": "6:44:28", "throughput": "17758.30", "total_tokens": 2378170368}
{"current_steps": 2269, "total_steps": 2679, "loss": 0.5976, "learning_rate": 2.886917307755349e-07, "epoch": 2.5387412587412586, "percentage": 84.7, "elapsed_time": "1 day, 13:12:52", "remaining_time": "6:43:28", "throughput": "17758.98", "total_tokens": 2379218944}
{"current_steps": 2270, "total_steps": 2679, "loss": 0.7133, "learning_rate": 2.8731277881865275e-07, "epoch": 2.53986013986014, "percentage": 84.73, "elapsed_time": "1 day, 13:13:52", "remaining_time": "6:42:29", "throughput": "17758.94", "total_tokens": 2380267520}
{"current_steps": 2271, "total_steps": 2679, "loss": 0.6143, "learning_rate": 2.859369272628928e-07, "epoch": 2.540979020979021, "percentage": 84.77, "elapsed_time": "1 day, 13:14:45", "remaining_time": "6:41:29", "throughput": "17759.64", "total_tokens": 2381316096}
{"current_steps": 2272, "total_steps": 2679, "loss": 0.5998, "learning_rate": 2.845641780360872e-07, "epoch": 2.542097902097902, "percentage": 84.81, "elapsed_time": "1 day, 13:15:39", "remaining_time": "6:40:29", "throughput": "17760.29", "total_tokens": 2382364672}
{"current_steps": 2273, "total_steps": 2679, "loss": 0.6149, "learning_rate": 2.8319453306172225e-07, "epoch": 2.543216783216783, "percentage": 84.85, "elapsed_time": "1 day, 13:16:39", "remaining_time": "6:39:30", "throughput": "17760.21", "total_tokens": 2383413248}
{"current_steps": 2274, "total_steps": 2679, "loss": 0.6118, "learning_rate": 2.818279942589347e-07, "epoch": 2.5443356643356645, "percentage": 84.88, "elapsed_time": "1 day, 13:17:30", "remaining_time": "6:38:30", "throughput": "17761.29", "total_tokens": 2384461824}
{"current_steps": 2275, "total_steps": 2679, "loss": 0.7162, "learning_rate": 2.804645635425091e-07, "epoch": 2.5454545454545454, "percentage": 84.92, "elapsed_time": "1 day, 13:18:30", "remaining_time": "6:37:31", "throughput": "17761.14", "total_tokens": 2385510400}
{"current_steps": 2276, "total_steps": 2679, "loss": 0.6296, "learning_rate": 2.7910424282287407e-07, "epoch": 2.546573426573427, "percentage": 84.96, "elapsed_time": "1 day, 13:19:24", "remaining_time": "6:36:31", "throughput": "17761.82", "total_tokens": 2386558976}
{"current_steps": 2277, "total_steps": 2679, "loss": 0.6946, "learning_rate": 2.7774703400610086e-07, "epoch": 2.5476923076923077, "percentage": 84.99, "elapsed_time": "1 day, 13:20:28", "remaining_time": "6:35:33", "throughput": "17761.20", "total_tokens": 2387607552}
{"current_steps": 2278, "total_steps": 2679, "loss": 0.7511, "learning_rate": 2.763929389939005e-07, "epoch": 2.5488111888111886, "percentage": 85.03, "elapsed_time": "1 day, 13:21:33", "remaining_time": "6:34:35", "throughput": "17760.42", "total_tokens": 2388656128}
{"current_steps": 2279, "total_steps": 2679, "loss": 0.5908, "learning_rate": 2.750419596836215e-07, "epoch": 2.54993006993007, "percentage": 85.07, "elapsed_time": "1 day, 13:22:21", "remaining_time": "6:33:34", "throughput": "17761.80", "total_tokens": 2389704704}
{"current_steps": 2280, "total_steps": 2679, "loss": 0.675, "learning_rate": 2.736940979682462e-07, "epoch": 2.551048951048951, "percentage": 85.11, "elapsed_time": "1 day, 13:23:21", "remaining_time": "6:32:35", "throughput": "17761.76", "total_tokens": 2390753280}
{"current_steps": 2281, "total_steps": 2679, "loss": 0.6393, "learning_rate": 2.723493557363885e-07, "epoch": 2.5521678321678323, "percentage": 85.14, "elapsed_time": "1 day, 13:24:16", "remaining_time": "6:31:35", "throughput": "17762.23", "total_tokens": 2391801856}
{"current_steps": 2282, "total_steps": 2679, "loss": 0.6421, "learning_rate": 2.710077348722906e-07, "epoch": 2.553286713286713, "percentage": 85.18, "elapsed_time": "1 day, 13:25:11", "remaining_time": "6:30:35", "throughput": "17762.84", "total_tokens": 2392850432}
{"current_steps": 2283, "total_steps": 2679, "loss": 0.6638, "learning_rate": 2.696692372558224e-07, "epoch": 2.5544055944055946, "percentage": 85.22, "elapsed_time": "1 day, 13:26:09", "remaining_time": "6:29:36", "throughput": "17762.94", "total_tokens": 2393899008}
{"current_steps": 2284, "total_steps": 2679, "loss": 0.673, "learning_rate": 2.683338647624767e-07, "epoch": 2.5555244755244755, "percentage": 85.26, "elapsed_time": "1 day, 13:27:08", "remaining_time": "6:28:37", "throughput": "17762.90", "total_tokens": 2394947584}
{"current_steps": 2285, "total_steps": 2679, "loss": 0.6883, "learning_rate": 2.670016192633687e-07, "epoch": 2.556643356643357, "percentage": 85.29, "elapsed_time": "1 day, 13:28:08", "remaining_time": "6:27:38", "throughput": "17762.84", "total_tokens": 2395996160}
{"current_steps": 2286, "total_steps": 2679, "loss": 0.6892, "learning_rate": 2.6567250262522937e-07, "epoch": 2.5577622377622378, "percentage": 85.33, "elapsed_time": "1 day, 13:29:02", "remaining_time": "6:26:38", "throughput": "17763.46", "total_tokens": 2397044736}
{"current_steps": 2287, "total_steps": 2679, "loss": 0.7378, "learning_rate": 2.6434651671040894e-07, "epoch": 2.5588811188811187, "percentage": 85.37, "elapsed_time": "1 day, 13:30:07", "remaining_time": "6:25:40", "throughput": "17762.66", "total_tokens": 2398093312}
{"current_steps": 2288, "total_steps": 2679, "loss": 0.6978, "learning_rate": 2.6302366337686765e-07, "epoch": 2.56, "percentage": 85.41, "elapsed_time": "1 day, 13:31:06", "remaining_time": "6:24:41", "throughput": "17762.62", "total_tokens": 2399141888}
{"current_steps": 2289, "total_steps": 2679, "loss": 0.707, "learning_rate": 2.6170394447817824e-07, "epoch": 2.561118881118881, "percentage": 85.44, "elapsed_time": "1 day, 13:32:06", "remaining_time": "6:23:42", "throughput": "17762.59", "total_tokens": 2400190464}
{"current_steps": 2290, "total_steps": 2679, "loss": 0.7546, "learning_rate": 2.603873618635222e-07, "epoch": 2.5622377622377623, "percentage": 85.48, "elapsed_time": "1 day, 13:33:10", "remaining_time": "6:22:44", "throughput": "17761.83", "total_tokens": 2401239040}
{"current_steps": 2291, "total_steps": 2679, "loss": 0.6234, "learning_rate": 2.590739173776841e-07, "epoch": 2.5633566433566433, "percentage": 85.52, "elapsed_time": "1 day, 13:34:10", "remaining_time": "6:21:45", "throughput": "17761.79", "total_tokens": 2402287616}
{"current_steps": 2292, "total_steps": 2679, "loss": 0.7245, "learning_rate": 2.577636128610539e-07, "epoch": 2.5644755244755246, "percentage": 85.55, "elapsed_time": "1 day, 13:35:14", "remaining_time": "6:20:47", "throughput": "17761.05", "total_tokens": 2403336192}
{"current_steps": 2293, "total_steps": 2679, "loss": 0.6064, "learning_rate": 2.5645645014961947e-07, "epoch": 2.5655944055944055, "percentage": 85.59, "elapsed_time": "1 day, 13:36:04", "remaining_time": "6:19:46", "throughput": "17762.36", "total_tokens": 2404384768}
{"current_steps": 2294, "total_steps": 2679, "loss": 0.7154, "learning_rate": 2.551524310749684e-07, "epoch": 2.566713286713287, "percentage": 85.63, "elapsed_time": "1 day, 13:37:03", "remaining_time": "6:18:48", "throughput": "17762.27", "total_tokens": 2405433344}
{"current_steps": 2295, "total_steps": 2679, "loss": 0.808, "learning_rate": 2.53851557464283e-07, "epoch": 2.567832167832168, "percentage": 85.67, "elapsed_time": "1 day, 13:38:12", "remaining_time": "6:17:50", "throughput": "17760.94", "total_tokens": 2406481920}
{"current_steps": 2296, "total_steps": 2679, "loss": 0.743, "learning_rate": 2.525538311403367e-07, "epoch": 2.5689510489510488, "percentage": 85.7, "elapsed_time": "1 day, 13:39:17", "remaining_time": "6:16:52", "throughput": "17760.18", "total_tokens": 2407530496}
{"current_steps": 2297, "total_steps": 2679, "loss": 0.5671, "learning_rate": 2.5125925392149533e-07, "epoch": 2.57006993006993, "percentage": 85.74, "elapsed_time": "1 day, 13:40:06", "remaining_time": "6:15:51", "throughput": "17761.59", "total_tokens": 2408579072}
{"current_steps": 2298, "total_steps": 2679, "loss": 0.6165, "learning_rate": 2.499678276217102e-07, "epoch": 2.571188811188811, "percentage": 85.78, "elapsed_time": "1 day, 13:40:59", "remaining_time": "6:14:51", "throughput": "17762.29", "total_tokens": 2409627648}
{"current_steps": 2299, "total_steps": 2679, "loss": 0.58, "learning_rate": 2.4867955405051826e-07, "epoch": 2.5723076923076924, "percentage": 85.82, "elapsed_time": "1 day, 13:41:48", "remaining_time": "6:13:51", "throughput": "17763.69", "total_tokens": 2410676224}
{"current_steps": 2300, "total_steps": 2679, "loss": 0.6418, "learning_rate": 2.4739443501304e-07, "epoch": 2.5734265734265733, "percentage": 85.85, "elapsed_time": "1 day, 13:42:48", "remaining_time": "6:12:52", "throughput": "17763.56", "total_tokens": 2411724800}
{"current_steps": 2301, "total_steps": 2679, "loss": 0.6953, "learning_rate": 2.4611247230997366e-07, "epoch": 2.5745454545454547, "percentage": 85.89, "elapsed_time": "1 day, 13:43:44", "remaining_time": "6:11:52", "throughput": "17763.96", "total_tokens": 2412773376}
{"current_steps": 2302, "total_steps": 2679, "loss": 0.5972, "learning_rate": 2.4483366773759705e-07, "epoch": 2.5756643356643356, "percentage": 85.93, "elapsed_time": "1 day, 13:44:37", "remaining_time": "6:10:52", "throughput": "17764.65", "total_tokens": 2413821952}
{"current_steps": 2303, "total_steps": 2679, "loss": 0.6312, "learning_rate": 2.4355802308776073e-07, "epoch": 2.576783216783217, "percentage": 85.96, "elapsed_time": "1 day, 13:45:37", "remaining_time": "6:09:53", "throughput": "17764.56", "total_tokens": 2414870528}
{"current_steps": 2304, "total_steps": 2679, "loss": 0.6304, "learning_rate": 2.422855401478891e-07, "epoch": 2.577902097902098, "percentage": 86.0, "elapsed_time": "1 day, 13:46:31", "remaining_time": "6:08:54", "throughput": "17765.21", "total_tokens": 2415919104}
{"current_steps": 2305, "total_steps": 2679, "loss": 0.6455, "learning_rate": 2.410162207009761e-07, "epoch": 2.579020979020979, "percentage": 86.04, "elapsed_time": "1 day, 13:47:30", "remaining_time": "6:07:55", "throughput": "17765.18", "total_tokens": 2416967680}
{"current_steps": 2306, "total_steps": 2679, "loss": 0.6233, "learning_rate": 2.397500665255825e-07, "epoch": 2.58013986013986, "percentage": 86.08, "elapsed_time": "1 day, 13:48:24", "remaining_time": "6:06:55", "throughput": "17765.84", "total_tokens": 2418016256}
{"current_steps": 2307, "total_steps": 2679, "loss": 0.7204, "learning_rate": 2.384870793958349e-07, "epoch": 2.581258741258741, "percentage": 86.11, "elapsed_time": "1 day, 13:49:24", "remaining_time": "6:05:56", "throughput": "17765.77", "total_tokens": 2419064832}
{"current_steps": 2308, "total_steps": 2679, "loss": 0.6586, "learning_rate": 2.372272610814208e-07, "epoch": 2.5823776223776225, "percentage": 86.15, "elapsed_time": "1 day, 13:50:23", "remaining_time": "6:04:57", "throughput": "17765.70", "total_tokens": 2420113408}
{"current_steps": 2309, "total_steps": 2679, "loss": 0.7094, "learning_rate": 2.3597061334758864e-07, "epoch": 2.5834965034965034, "percentage": 86.19, "elapsed_time": "1 day, 13:51:18", "remaining_time": "6:03:57", "throughput": "17766.29", "total_tokens": 2421161984}
{"current_steps": 2310, "total_steps": 2679, "loss": 0.6513, "learning_rate": 2.3471713795514412e-07, "epoch": 2.5846153846153848, "percentage": 86.23, "elapsed_time": "1 day, 13:52:15", "remaining_time": "6:02:58", "throughput": "17766.57", "total_tokens": 2422210560}
{"current_steps": 2311, "total_steps": 2679, "loss": 0.7583, "learning_rate": 2.334668366604481e-07, "epoch": 2.5857342657342657, "percentage": 86.26, "elapsed_time": "1 day, 13:53:20", "remaining_time": "6:02:00", "throughput": "17765.76", "total_tokens": 2423259136}
{"current_steps": 2312, "total_steps": 2679, "loss": 0.638, "learning_rate": 2.3221971121541343e-07, "epoch": 2.586853146853147, "percentage": 86.3, "elapsed_time": "1 day, 13:54:20", "remaining_time": "6:01:01", "throughput": "17765.67", "total_tokens": 2424307712}
{"current_steps": 2313, "total_steps": 2679, "loss": 0.5978, "learning_rate": 2.3097576336750248e-07, "epoch": 2.587972027972028, "percentage": 86.34, "elapsed_time": "1 day, 13:55:14", "remaining_time": "6:00:01", "throughput": "17766.33", "total_tokens": 2425356288}
{"current_steps": 2314, "total_steps": 2679, "loss": 0.6024, "learning_rate": 2.2973499485972634e-07, "epoch": 2.589090909090909, "percentage": 86.38, "elapsed_time": "1 day, 13:56:07", "remaining_time": "5:59:01", "throughput": "17767.02", "total_tokens": 2426404864}
{"current_steps": 2315, "total_steps": 2679, "loss": 0.8446, "learning_rate": 2.2849740743064063e-07, "epoch": 2.5902097902097903, "percentage": 86.41, "elapsed_time": "1 day, 13:57:14", "remaining_time": "5:58:03", "throughput": "17766.07", "total_tokens": 2427453440}
{"current_steps": 2316, "total_steps": 2679, "loss": 0.8431, "learning_rate": 2.2726300281434467e-07, "epoch": 2.591328671328671, "percentage": 86.45, "elapsed_time": "1 day, 13:58:25", "remaining_time": "5:57:06", "throughput": "17764.47", "total_tokens": 2428502016}
{"current_steps": 2317, "total_steps": 2679, "loss": 0.6557, "learning_rate": 2.260317827404762e-07, "epoch": 2.5924475524475525, "percentage": 86.49, "elapsed_time": "1 day, 13:59:20", "remaining_time": "5:56:07", "throughput": "17764.94", "total_tokens": 2429550592}
{"current_steps": 2318, "total_steps": 2679, "loss": 0.6948, "learning_rate": 2.2480374893421142e-07, "epoch": 2.5935664335664335, "percentage": 86.52, "elapsed_time": "1 day, 14:00:24", "remaining_time": "5:55:08", "throughput": "17764.30", "total_tokens": 2430599168}
{"current_steps": 2319, "total_steps": 2679, "loss": 0.5599, "learning_rate": 2.2357890311626328e-07, "epoch": 2.594685314685315, "percentage": 86.56, "elapsed_time": "1 day, 14:01:16", "remaining_time": "5:54:08", "throughput": "17765.28", "total_tokens": 2431647744}
{"current_steps": 2320, "total_steps": 2679, "loss": 0.6915, "learning_rate": 2.2235724700287592e-07, "epoch": 2.5958041958041957, "percentage": 86.6, "elapsed_time": "1 day, 14:02:21", "remaining_time": "5:53:10", "throughput": "17764.48", "total_tokens": 2432696320}
{"current_steps": 2321, "total_steps": 2679, "loss": 0.8178, "learning_rate": 2.2113878230582615e-07, "epoch": 2.596923076923077, "percentage": 86.64, "elapsed_time": "1 day, 14:03:27", "remaining_time": "5:52:12", "throughput": "17763.58", "total_tokens": 2433744896}
{"current_steps": 2322, "total_steps": 2679, "loss": 0.6875, "learning_rate": 2.1992351073241684e-07, "epoch": 2.598041958041958, "percentage": 86.67, "elapsed_time": "1 day, 14:04:27", "remaining_time": "5:51:13", "throughput": "17763.46", "total_tokens": 2434793472}
{"current_steps": 2323, "total_steps": 2679, "loss": 0.617, "learning_rate": 2.1871143398547735e-07, "epoch": 2.599160839160839, "percentage": 86.71, "elapsed_time": "1 day, 14:05:21", "remaining_time": "5:50:13", "throughput": "17764.14", "total_tokens": 2435842048}
{"current_steps": 2324, "total_steps": 2679, "loss": 0.6776, "learning_rate": 2.1750255376336126e-07, "epoch": 2.6002797202797203, "percentage": 86.75, "elapsed_time": "1 day, 14:06:19", "remaining_time": "5:49:14", "throughput": "17764.31", "total_tokens": 2436890624}
{"current_steps": 2325, "total_steps": 2679, "loss": 0.5371, "learning_rate": 2.162968717599423e-07, "epoch": 2.6013986013986012, "percentage": 86.79, "elapsed_time": "1 day, 14:07:07", "remaining_time": "5:48:13", "throughput": "17765.72", "total_tokens": 2437939200}
{"current_steps": 2326, "total_steps": 2679, "loss": 0.6982, "learning_rate": 2.1509438966461433e-07, "epoch": 2.6025174825174826, "percentage": 86.82, "elapsed_time": "1 day, 14:08:06", "remaining_time": "5:47:14", "throughput": "17765.66", "total_tokens": 2438987776}
{"current_steps": 2327, "total_steps": 2679, "loss": 0.7472, "learning_rate": 2.1389510916228513e-07, "epoch": 2.6036363636363635, "percentage": 86.86, "elapsed_time": "1 day, 14:09:17", "remaining_time": "5:46:17", "throughput": "17764.18", "total_tokens": 2440036352}
{"current_steps": 2328, "total_steps": 2679, "loss": 0.6019, "learning_rate": 2.1269903193337853e-07, "epoch": 2.604755244755245, "percentage": 86.9, "elapsed_time": "1 day, 14:10:10", "remaining_time": "5:45:17", "throughput": "17764.86", "total_tokens": 2441084928}
{"current_steps": 2329, "total_steps": 2679, "loss": 0.7517, "learning_rate": 2.115061596538287e-07, "epoch": 2.605874125874126, "percentage": 86.94, "elapsed_time": "1 day, 14:11:12", "remaining_time": "5:44:19", "throughput": "17764.50", "total_tokens": 2442133504}
{"current_steps": 2330, "total_steps": 2679, "loss": 0.7789, "learning_rate": 2.103164939950797e-07, "epoch": 2.606993006993007, "percentage": 86.97, "elapsed_time": "1 day, 14:12:12", "remaining_time": "5:43:20", "throughput": "17764.34", "total_tokens": 2443182080}
{"current_steps": 2331, "total_steps": 2679, "loss": 0.6756, "learning_rate": 2.0913003662408254e-07, "epoch": 2.608111888111888, "percentage": 87.01, "elapsed_time": "1 day, 14:13:12", "remaining_time": "5:42:21", "throughput": "17764.21", "total_tokens": 2444230656}
{"current_steps": 2332, "total_steps": 2679, "loss": 0.6985, "learning_rate": 2.079467892032924e-07, "epoch": 2.609230769230769, "percentage": 87.05, "elapsed_time": "1 day, 14:14:12", "remaining_time": "5:41:22", "throughput": "17764.11", "total_tokens": 2445279232}
{"current_steps": 2333, "total_steps": 2679, "loss": 0.7547, "learning_rate": 2.0676675339066726e-07, "epoch": 2.6103496503496504, "percentage": 87.08, "elapsed_time": "1 day, 14:15:16", "remaining_time": "5:40:24", "throughput": "17763.45", "total_tokens": 2446327808}
{"current_steps": 2334, "total_steps": 2679, "loss": 0.5374, "learning_rate": 2.0558993083966388e-07, "epoch": 2.6114685314685313, "percentage": 87.12, "elapsed_time": "1 day, 14:16:05", "remaining_time": "5:39:23", "throughput": "17764.84", "total_tokens": 2447376384}
{"current_steps": 2335, "total_steps": 2679, "loss": 0.7885, "learning_rate": 2.0441632319923798e-07, "epoch": 2.6125874125874127, "percentage": 87.16, "elapsed_time": "1 day, 14:17:11", "remaining_time": "5:38:25", "throughput": "17763.95", "total_tokens": 2448424960}
{"current_steps": 2336, "total_steps": 2679, "loss": 0.5847, "learning_rate": 2.0324593211384026e-07, "epoch": 2.6137062937062936, "percentage": 87.2, "elapsed_time": "1 day, 14:18:04", "remaining_time": "5:37:25", "throughput": "17764.64", "total_tokens": 2449473536}
{"current_steps": 2337, "total_steps": 2679, "loss": 0.5384, "learning_rate": 2.020787592234133e-07, "epoch": 2.614825174825175, "percentage": 87.23, "elapsed_time": "1 day, 14:18:53", "remaining_time": "5:36:25", "throughput": "17766.03", "total_tokens": 2450522112}
{"current_steps": 2338, "total_steps": 2679, "loss": 0.5595, "learning_rate": 2.0091480616339197e-07, "epoch": 2.615944055944056, "percentage": 87.27, "elapsed_time": "1 day, 14:19:41", "remaining_time": "5:35:24", "throughput": "17767.38", "total_tokens": 2451570688}
{"current_steps": 2339, "total_steps": 2679, "loss": 0.6298, "learning_rate": 1.9975407456469808e-07, "epoch": 2.6170629370629372, "percentage": 87.31, "elapsed_time": "1 day, 14:20:35", "remaining_time": "5:34:24", "throughput": "17768.06", "total_tokens": 2452619264}
{"current_steps": 2340, "total_steps": 2679, "loss": 0.6429, "learning_rate": 1.9859656605374062e-07, "epoch": 2.618181818181818, "percentage": 87.35, "elapsed_time": "1 day, 14:21:28", "remaining_time": "5:33:25", "throughput": "17768.79", "total_tokens": 2453667840}
{"current_steps": 2341, "total_steps": 2679, "loss": 0.7437, "learning_rate": 1.9744228225241248e-07, "epoch": 2.619300699300699, "percentage": 87.38, "elapsed_time": "1 day, 14:22:34", "remaining_time": "5:32:27", "throughput": "17767.97", "total_tokens": 2454716416}
{"current_steps": 2342, "total_steps": 2679, "loss": 0.797, "learning_rate": 1.962912247780868e-07, "epoch": 2.6204195804195805, "percentage": 87.42, "elapsed_time": "1 day, 14:23:40", "remaining_time": "5:31:29", "throughput": "17766.99", "total_tokens": 2455764992}
{"current_steps": 2343, "total_steps": 2679, "loss": 0.6242, "learning_rate": 1.9514339524361742e-07, "epoch": 2.6215384615384614, "percentage": 87.46, "elapsed_time": "1 day, 14:24:34", "remaining_time": "5:30:29", "throughput": "17767.67", "total_tokens": 2456813568}
{"current_steps": 2344, "total_steps": 2679, "loss": 0.6448, "learning_rate": 1.939987952573344e-07, "epoch": 2.6226573426573427, "percentage": 87.5, "elapsed_time": "1 day, 14:25:28", "remaining_time": "5:29:29", "throughput": "17768.31", "total_tokens": 2457862144}
{"current_steps": 2345, "total_steps": 2679, "loss": 0.7, "learning_rate": 1.928574264230429e-07, "epoch": 2.6237762237762237, "percentage": 87.53, "elapsed_time": "1 day, 14:26:33", "remaining_time": "5:28:31", "throughput": "17767.51", "total_tokens": 2458910720}
{"current_steps": 2346, "total_steps": 2679, "loss": 0.6599, "learning_rate": 1.9171929034002112e-07, "epoch": 2.624895104895105, "percentage": 87.57, "elapsed_time": "1 day, 14:27:33", "remaining_time": "5:27:32", "throughput": "17767.44", "total_tokens": 2459959296}
{"current_steps": 2347, "total_steps": 2679, "loss": 0.8065, "learning_rate": 1.9058438860301621e-07, "epoch": 2.626013986013986, "percentage": 87.61, "elapsed_time": "1 day, 14:28:39", "remaining_time": "5:26:34", "throughput": "17766.51", "total_tokens": 2461007872}
{"current_steps": 2348, "total_steps": 2679, "loss": 0.5611, "learning_rate": 1.894527228022447e-07, "epoch": 2.6271328671328673, "percentage": 87.64, "elapsed_time": "1 day, 14:29:28", "remaining_time": "5:25:34", "throughput": "17767.76", "total_tokens": 2462056448}
{"current_steps": 2349, "total_steps": 2679, "loss": 0.628, "learning_rate": 1.883242945233879e-07, "epoch": 2.6282517482517482, "percentage": 87.68, "elapsed_time": "1 day, 14:30:22", "remaining_time": "5:24:34", "throughput": "17768.41", "total_tokens": 2463105024}
{"current_steps": 2350, "total_steps": 2679, "loss": 0.8103, "learning_rate": 1.871991053475916e-07, "epoch": 2.629370629370629, "percentage": 87.72, "elapsed_time": "1 day, 14:31:28", "remaining_time": "5:23:36", "throughput": "17767.52", "total_tokens": 2464153600}
{"current_steps": 2351, "total_steps": 2679, "loss": 0.8288, "learning_rate": 1.8607715685146244e-07, "epoch": 2.6304895104895105, "percentage": 87.76, "elapsed_time": "1 day, 14:32:34", "remaining_time": "5:22:38", "throughput": "17766.65", "total_tokens": 2465202176}
{"current_steps": 2352, "total_steps": 2679, "loss": 0.617, "learning_rate": 1.849584506070673e-07, "epoch": 2.631608391608392, "percentage": 87.79, "elapsed_time": "1 day, 14:33:28", "remaining_time": "5:21:38", "throughput": "17767.32", "total_tokens": 2466250752}
{"current_steps": 2353, "total_steps": 2679, "loss": 0.5503, "learning_rate": 1.8384298818192814e-07, "epoch": 2.632727272727273, "percentage": 87.83, "elapsed_time": "1 day, 14:34:16", "remaining_time": "5:20:38", "throughput": "17768.72", "total_tokens": 2467299328}
{"current_steps": 2354, "total_steps": 2679, "loss": 0.6904, "learning_rate": 1.8273077113902276e-07, "epoch": 2.6338461538461537, "percentage": 87.87, "elapsed_time": "1 day, 14:35:13", "remaining_time": "5:19:38", "throughput": "17768.92", "total_tokens": 2468347904}
{"current_steps": 2355, "total_steps": 2679, "loss": 0.7589, "learning_rate": 1.8162180103678177e-07, "epoch": 2.634965034965035, "percentage": 87.91, "elapsed_time": "1 day, 14:36:13", "remaining_time": "5:18:39", "throughput": "17768.78", "total_tokens": 2469396480}
{"current_steps": 2356, "total_steps": 2679, "loss": 0.6092, "learning_rate": 1.805160794290861e-07, "epoch": 2.636083916083916, "percentage": 87.94, "elapsed_time": "1 day, 14:37:07", "remaining_time": "5:17:40", "throughput": "17769.48", "total_tokens": 2470445056}
{"current_steps": 2357, "total_steps": 2679, "loss": 0.6892, "learning_rate": 1.79413607865265e-07, "epoch": 2.6372027972027974, "percentage": 87.98, "elapsed_time": "1 day, 14:38:07", "remaining_time": "5:16:41", "throughput": "17769.30", "total_tokens": 2471493632}
{"current_steps": 2358, "total_steps": 2679, "loss": 0.6571, "learning_rate": 1.7831438789009337e-07, "epoch": 2.6383216783216783, "percentage": 88.02, "elapsed_time": "1 day, 14:39:07", "remaining_time": "5:15:42", "throughput": "17769.23", "total_tokens": 2472542208}
{"current_steps": 2359, "total_steps": 2679, "loss": 0.6386, "learning_rate": 1.772184210437894e-07, "epoch": 2.639440559440559, "percentage": 88.06, "elapsed_time": "1 day, 14:40:01", "remaining_time": "5:14:42", "throughput": "17769.86", "total_tokens": 2473590784}
{"current_steps": 2360, "total_steps": 2679, "loss": 0.9497, "learning_rate": 1.7612570886201442e-07, "epoch": 2.6405594405594406, "percentage": 88.09, "elapsed_time": "1 day, 14:41:13", "remaining_time": "5:13:45", "throughput": "17768.17", "total_tokens": 2474639360}
{"current_steps": 2361, "total_steps": 2679, "loss": 0.7045, "learning_rate": 1.7503625287586896e-07, "epoch": 2.641678321678322, "percentage": 88.13, "elapsed_time": "1 day, 14:42:14", "remaining_time": "5:12:46", "throughput": "17767.98", "total_tokens": 2475687936}
{"current_steps": 2362, "total_steps": 2679, "loss": 0.6049, "learning_rate": 1.739500546118908e-07, "epoch": 2.642797202797203, "percentage": 88.17, "elapsed_time": "1 day, 14:43:08", "remaining_time": "5:11:47", "throughput": "17768.61", "total_tokens": 2476736512}
{"current_steps": 2363, "total_steps": 2679, "loss": 0.6196, "learning_rate": 1.728671155920525e-07, "epoch": 2.643916083916084, "percentage": 88.2, "elapsed_time": "1 day, 14:44:02", "remaining_time": "5:10:47", "throughput": "17769.26", "total_tokens": 2477785088}
{"current_steps": 2364, "total_steps": 2679, "loss": 0.5948, "learning_rate": 1.717874373337608e-07, "epoch": 2.645034965034965, "percentage": 88.24, "elapsed_time": "1 day, 14:44:56", "remaining_time": "5:09:47", "throughput": "17769.89", "total_tokens": 2478833664}
{"current_steps": 2365, "total_steps": 2679, "loss": 0.6138, "learning_rate": 1.7071102134985224e-07, "epoch": 2.646153846153846, "percentage": 88.28, "elapsed_time": "1 day, 14:45:50", "remaining_time": "5:08:48", "throughput": "17770.55", "total_tokens": 2479882240}
{"current_steps": 2366, "total_steps": 2679, "loss": 0.64, "learning_rate": 1.6963786914859338e-07, "epoch": 2.6472727272727274, "percentage": 88.32, "elapsed_time": "1 day, 14:46:44", "remaining_time": "5:07:48", "throughput": "17771.12", "total_tokens": 2480930816}
{"current_steps": 2367, "total_steps": 2679, "loss": 0.8437, "learning_rate": 1.6856798223367777e-07, "epoch": 2.6483916083916084, "percentage": 88.35, "elapsed_time": "1 day, 14:47:50", "remaining_time": "5:06:50", "throughput": "17770.25", "total_tokens": 2481979392}
{"current_steps": 2368, "total_steps": 2679, "loss": 0.5819, "learning_rate": 1.6750136210422235e-07, "epoch": 2.6495104895104893, "percentage": 88.39, "elapsed_time": "1 day, 14:48:40", "remaining_time": "5:05:50", "throughput": "17771.43", "total_tokens": 2483027968}
{"current_steps": 2369, "total_steps": 2679, "loss": 0.6379, "learning_rate": 1.66438010254768e-07, "epoch": 2.6506293706293707, "percentage": 88.43, "elapsed_time": "1 day, 14:49:33", "remaining_time": "5:04:50", "throughput": "17772.15", "total_tokens": 2484076544}
{"current_steps": 2370, "total_steps": 2679, "loss": 0.6556, "learning_rate": 1.6537792817527543e-07, "epoch": 2.651748251748252, "percentage": 88.47, "elapsed_time": "1 day, 14:50:32", "remaining_time": "5:03:51", "throughput": "17772.10", "total_tokens": 2485125120}
{"current_steps": 2371, "total_steps": 2679, "loss": 0.6788, "learning_rate": 1.6432111735112277e-07, "epoch": 2.652867132867133, "percentage": 88.5, "elapsed_time": "1 day, 14:51:32", "remaining_time": "5:02:52", "throughput": "17771.97", "total_tokens": 2486173696}
{"current_steps": 2372, "total_steps": 2679, "loss": 0.6448, "learning_rate": 1.6326757926310748e-07, "epoch": 2.653986013986014, "percentage": 88.54, "elapsed_time": "1 day, 14:52:32", "remaining_time": "5:01:53", "throughput": "17771.93", "total_tokens": 2487222272}
{"current_steps": 2373, "total_steps": 2679, "loss": 0.6232, "learning_rate": 1.622173153874379e-07, "epoch": 2.6551048951048952, "percentage": 88.58, "elapsed_time": "1 day, 14:53:31", "remaining_time": "5:00:54", "throughput": "17771.85", "total_tokens": 2488270848}
{"current_steps": 2374, "total_steps": 2679, "loss": 0.5747, "learning_rate": 1.611703271957371e-07, "epoch": 2.656223776223776, "percentage": 88.62, "elapsed_time": "1 day, 14:54:19", "remaining_time": "4:59:54", "throughput": "17773.27", "total_tokens": 2489319424}
{"current_steps": 2375, "total_steps": 2679, "loss": 0.8071, "learning_rate": 1.601266161550366e-07, "epoch": 2.6573426573426575, "percentage": 88.65, "elapsed_time": "1 day, 14:55:25", "remaining_time": "4:58:56", "throughput": "17772.45", "total_tokens": 2490368000}
{"current_steps": 2376, "total_steps": 2679, "loss": 0.593, "learning_rate": 1.5908618372777656e-07, "epoch": 2.6584615384615384, "percentage": 88.69, "elapsed_time": "1 day, 14:56:19", "remaining_time": "4:57:56", "throughput": "17773.10", "total_tokens": 2491416576}
{"current_steps": 2377, "total_steps": 2679, "loss": 0.6145, "learning_rate": 1.5804903137180415e-07, "epoch": 2.6595804195804194, "percentage": 88.73, "elapsed_time": "1 day, 14:57:12", "remaining_time": "4:56:56", "throughput": "17773.75", "total_tokens": 2492465152}
{"current_steps": 2378, "total_steps": 2679, "loss": 0.6186, "learning_rate": 1.5701516054036886e-07, "epoch": 2.6606993006993007, "percentage": 88.76, "elapsed_time": "1 day, 14:58:07", "remaining_time": "4:55:57", "throughput": "17774.35", "total_tokens": 2493513728}
{"current_steps": 2379, "total_steps": 2679, "loss": 0.6529, "learning_rate": 1.5598457268212353e-07, "epoch": 2.661818181818182, "percentage": 88.8, "elapsed_time": "1 day, 14:59:06", "remaining_time": "4:54:58", "throughput": "17774.32", "total_tokens": 2494562304}
{"current_steps": 2380, "total_steps": 2679, "loss": 0.6891, "learning_rate": 1.5495726924111942e-07, "epoch": 2.662937062937063, "percentage": 88.84, "elapsed_time": "1 day, 15:00:06", "remaining_time": "4:53:59", "throughput": "17774.18", "total_tokens": 2495610880}
{"current_steps": 2381, "total_steps": 2679, "loss": 0.7731, "learning_rate": 1.5393325165680707e-07, "epoch": 2.664055944055944, "percentage": 88.88, "elapsed_time": "1 day, 15:01:11", "remaining_time": "4:53:01", "throughput": "17773.42", "total_tokens": 2496659456}
{"current_steps": 2382, "total_steps": 2679, "loss": 0.6344, "learning_rate": 1.5291252136403284e-07, "epoch": 2.6651748251748253, "percentage": 88.91, "elapsed_time": "1 day, 15:02:11", "remaining_time": "4:52:02", "throughput": "17773.30", "total_tokens": 2497708032}
{"current_steps": 2383, "total_steps": 2679, "loss": 0.6363, "learning_rate": 1.5189507979303575e-07, "epoch": 2.666293706293706, "percentage": 88.95, "elapsed_time": "1 day, 15:03:10", "remaining_time": "4:51:03", "throughput": "17773.27", "total_tokens": 2498756608}
{"current_steps": 2384, "total_steps": 2679, "loss": 0.8332, "learning_rate": 1.5088092836944844e-07, "epoch": 2.6674125874125876, "percentage": 88.99, "elapsed_time": "1 day, 15:04:16", "remaining_time": "4:50:05", "throughput": "17772.35", "total_tokens": 2499805184}
{"current_steps": 2385, "total_steps": 2679, "loss": 0.6673, "learning_rate": 1.4987006851429147e-07, "epoch": 2.6685314685314685, "percentage": 89.03, "elapsed_time": "1 day, 15:05:13", "remaining_time": "4:49:05", "throughput": "17772.70", "total_tokens": 2500853760}
{"current_steps": 2386, "total_steps": 2679, "loss": 0.6535, "learning_rate": 1.4886250164397458e-07, "epoch": 2.6696503496503494, "percentage": 89.06, "elapsed_time": "1 day, 15:06:12", "remaining_time": "4:48:06", "throughput": "17772.63", "total_tokens": 2501902336}
{"current_steps": 2387, "total_steps": 2679, "loss": 0.6838, "learning_rate": 1.4785822917029318e-07, "epoch": 2.670769230769231, "percentage": 89.1, "elapsed_time": "1 day, 15:07:10", "remaining_time": "4:47:07", "throughput": "17772.81", "total_tokens": 2502950912}
{"current_steps": 2388, "total_steps": 2679, "loss": 0.6759, "learning_rate": 1.4685725250042692e-07, "epoch": 2.671888111888112, "percentage": 89.14, "elapsed_time": "1 day, 15:08:09", "remaining_time": "4:46:08", "throughput": "17772.79", "total_tokens": 2503999488}
{"current_steps": 2389, "total_steps": 2679, "loss": 0.6384, "learning_rate": 1.4585957303693664e-07, "epoch": 2.673006993006993, "percentage": 89.18, "elapsed_time": "1 day, 15:09:09", "remaining_time": "4:45:09", "throughput": "17772.69", "total_tokens": 2505048064}
{"current_steps": 2390, "total_steps": 2679, "loss": 0.7674, "learning_rate": 1.4486519217776273e-07, "epoch": 2.674125874125874, "percentage": 89.21, "elapsed_time": "1 day, 15:10:08", "remaining_time": "4:44:10", "throughput": "17772.62", "total_tokens": 2506096640}
{"current_steps": 2391, "total_steps": 2679, "loss": 0.6736, "learning_rate": 1.4387411131622592e-07, "epoch": 2.6752447552447554, "percentage": 89.25, "elapsed_time": "1 day, 15:11:08", "remaining_time": "4:43:11", "throughput": "17772.54", "total_tokens": 2507145216}
{"current_steps": 2392, "total_steps": 2679, "loss": 0.5869, "learning_rate": 1.4288633184101953e-07, "epoch": 2.6763636363636363, "percentage": 89.29, "elapsed_time": "1 day, 15:11:56", "remaining_time": "4:42:11", "throughput": "17773.95", "total_tokens": 2508193792}
{"current_steps": 2393, "total_steps": 2679, "loss": 0.6028, "learning_rate": 1.4190185513621473e-07, "epoch": 2.6774825174825176, "percentage": 89.32, "elapsed_time": "1 day, 15:12:47", "remaining_time": "4:41:11", "throughput": "17774.94", "total_tokens": 2509242368}
{"current_steps": 2394, "total_steps": 2679, "loss": 0.6303, "learning_rate": 1.409206825812523e-07, "epoch": 2.6786013986013986, "percentage": 89.36, "elapsed_time": "1 day, 15:13:41", "remaining_time": "4:40:12", "throughput": "17775.58", "total_tokens": 2510290944}
{"current_steps": 2395, "total_steps": 2679, "loss": 0.655, "learning_rate": 1.3994281555094386e-07, "epoch": 2.6797202797202795, "percentage": 89.4, "elapsed_time": "1 day, 15:14:37", "remaining_time": "4:39:12", "throughput": "17775.92", "total_tokens": 2511339520}
{"current_steps": 2396, "total_steps": 2679, "loss": 0.6869, "learning_rate": 1.3896825541547003e-07, "epoch": 2.680839160839161, "percentage": 89.44, "elapsed_time": "1 day, 15:15:37", "remaining_time": "4:38:13", "throughput": "17775.79", "total_tokens": 2512388096}
{"current_steps": 2397, "total_steps": 2679, "loss": 0.696, "learning_rate": 1.3799700354037605e-07, "epoch": 2.681958041958042, "percentage": 89.47, "elapsed_time": "1 day, 15:16:43", "remaining_time": "4:37:15", "throughput": "17774.97", "total_tokens": 2513436672}
{"current_steps": 2398, "total_steps": 2679, "loss": 0.6224, "learning_rate": 1.370290612865749e-07, "epoch": 2.683076923076923, "percentage": 89.51, "elapsed_time": "1 day, 15:17:38", "remaining_time": "4:36:16", "throughput": "17775.44", "total_tokens": 2514485248}
{"current_steps": 2399, "total_steps": 2679, "loss": 0.7083, "learning_rate": 1.3606443001033864e-07, "epoch": 2.684195804195804, "percentage": 89.55, "elapsed_time": "1 day, 15:18:38", "remaining_time": "4:35:17", "throughput": "17775.28", "total_tokens": 2515533824}
{"current_steps": 2400, "total_steps": 2679, "loss": 0.5806, "learning_rate": 1.3510311106330247e-07, "epoch": 2.6853146853146854, "percentage": 89.59, "elapsed_time": "1 day, 15:19:33", "remaining_time": "4:34:17", "throughput": "17775.81", "total_tokens": 2516582400}
{"current_steps": 2401, "total_steps": 2679, "loss": 0.6754, "learning_rate": 1.341451057924592e-07, "epoch": 2.6864335664335663, "percentage": 89.62, "elapsed_time": "1 day, 15:20:32", "remaining_time": "4:33:18", "throughput": "17775.78", "total_tokens": 2517630976}
{"current_steps": 2402, "total_steps": 2679, "loss": 0.6539, "learning_rate": 1.3319041554015782e-07, "epoch": 2.6875524475524477, "percentage": 89.66, "elapsed_time": "1 day, 15:21:32", "remaining_time": "4:32:20", "throughput": "17775.71", "total_tokens": 2518679552}
{"current_steps": 2403, "total_steps": 2679, "loss": 0.6658, "learning_rate": 1.3223904164410494e-07, "epoch": 2.6886713286713286, "percentage": 89.7, "elapsed_time": "1 day, 15:22:23", "remaining_time": "4:31:20", "throughput": "17776.70", "total_tokens": 2519728128}
{"current_steps": 2404, "total_steps": 2679, "loss": 0.8401, "learning_rate": 1.3129098543735758e-07, "epoch": 2.6897902097902096, "percentage": 89.73, "elapsed_time": "1 day, 15:23:23", "remaining_time": "4:30:21", "throughput": "17776.50", "total_tokens": 2520776704}
{"current_steps": 2405, "total_steps": 2679, "loss": 0.5324, "learning_rate": 1.303462482483256e-07, "epoch": 2.690909090909091, "percentage": 89.77, "elapsed_time": "1 day, 15:24:11", "remaining_time": "4:29:21", "throughput": "17777.87", "total_tokens": 2521825280}
{"current_steps": 2406, "total_steps": 2679, "loss": 0.5929, "learning_rate": 1.2940483140076788e-07, "epoch": 2.6920279720279723, "percentage": 89.81, "elapsed_time": "1 day, 15:25:00", "remaining_time": "4:28:20", "throughput": "17779.18", "total_tokens": 2522873856}
{"current_steps": 2407, "total_steps": 2679, "loss": 0.5994, "learning_rate": 1.2846673621379035e-07, "epoch": 2.693146853146853, "percentage": 89.85, "elapsed_time": "1 day, 15:25:55", "remaining_time": "4:27:21", "throughput": "17779.71", "total_tokens": 2523922432}
{"current_steps": 2408, "total_steps": 2679, "loss": 0.6117, "learning_rate": 1.2753196400184548e-07, "epoch": 2.694265734265734, "percentage": 89.88, "elapsed_time": "1 day, 15:26:48", "remaining_time": "4:26:21", "throughput": "17780.37", "total_tokens": 2524971008}
{"current_steps": 2409, "total_steps": 2679, "loss": 0.7487, "learning_rate": 1.2660051607472885e-07, "epoch": 2.6953846153846155, "percentage": 89.92, "elapsed_time": "1 day, 15:27:49", "remaining_time": "4:25:23", "throughput": "17780.17", "total_tokens": 2526019584}
{"current_steps": 2410, "total_steps": 2679, "loss": 0.6408, "learning_rate": 1.2567239373757923e-07, "epoch": 2.6965034965034964, "percentage": 89.96, "elapsed_time": "1 day, 15:28:43", "remaining_time": "4:24:23", "throughput": "17780.74", "total_tokens": 2527068160}
{"current_steps": 2411, "total_steps": 2679, "loss": 0.6753, "learning_rate": 1.2474759829087413e-07, "epoch": 2.6976223776223778, "percentage": 90.0, "elapsed_time": "1 day, 15:29:38", "remaining_time": "4:23:24", "throughput": "17781.26", "total_tokens": 2528116736}
{"current_steps": 2412, "total_steps": 2679, "loss": 0.6925, "learning_rate": 1.2382613103043062e-07, "epoch": 2.6987412587412587, "percentage": 90.03, "elapsed_time": "1 day, 15:30:38", "remaining_time": "4:22:25", "throughput": "17781.12", "total_tokens": 2529165312}
{"current_steps": 2413, "total_steps": 2679, "loss": 0.7279, "learning_rate": 1.2290799324740144e-07, "epoch": 2.6998601398601396, "percentage": 90.07, "elapsed_time": "1 day, 15:31:43", "remaining_time": "4:21:27", "throughput": "17780.36", "total_tokens": 2530213888}
{"current_steps": 2414, "total_steps": 2679, "loss": 0.5556, "learning_rate": 1.2199318622827473e-07, "epoch": 2.700979020979021, "percentage": 90.11, "elapsed_time": "1 day, 15:32:32", "remaining_time": "4:20:26", "throughput": "17781.70", "total_tokens": 2531262464}
{"current_steps": 2415, "total_steps": 2679, "loss": 0.5939, "learning_rate": 1.2108171125487177e-07, "epoch": 2.7020979020979023, "percentage": 90.15, "elapsed_time": "1 day, 15:33:25", "remaining_time": "4:19:27", "throughput": "17782.38", "total_tokens": 2532311040}
{"current_steps": 2416, "total_steps": 2679, "loss": 0.7539, "learning_rate": 1.2017356960434406e-07, "epoch": 2.7032167832167833, "percentage": 90.18, "elapsed_time": "1 day, 15:34:28", "remaining_time": "4:18:28", "throughput": "17781.93", "total_tokens": 2533359616}
{"current_steps": 2417, "total_steps": 2679, "loss": 0.6324, "learning_rate": 1.1926876254917314e-07, "epoch": 2.704335664335664, "percentage": 90.22, "elapsed_time": "1 day, 15:35:22", "remaining_time": "4:17:29", "throughput": "17782.55", "total_tokens": 2534408192}
{"current_steps": 2418, "total_steps": 2679, "loss": 0.5899, "learning_rate": 1.1836729135716818e-07, "epoch": 2.7054545454545456, "percentage": 90.26, "elapsed_time": "1 day, 15:36:16", "remaining_time": "4:16:29", "throughput": "17783.19", "total_tokens": 2535456768}
{"current_steps": 2419, "total_steps": 2679, "loss": 0.7363, "learning_rate": 1.174691572914638e-07, "epoch": 2.7065734265734265, "percentage": 90.29, "elapsed_time": "1 day, 15:37:15", "remaining_time": "4:15:30", "throughput": "17783.08", "total_tokens": 2536505344}
{"current_steps": 2420, "total_steps": 2679, "loss": 0.6672, "learning_rate": 1.1657436161051916e-07, "epoch": 2.707692307692308, "percentage": 90.33, "elapsed_time": "1 day, 15:38:15", "remaining_time": "4:14:32", "throughput": "17782.94", "total_tokens": 2537553920}
{"current_steps": 2421, "total_steps": 2679, "loss": 0.7783, "learning_rate": 1.1568290556811495e-07, "epoch": 2.7088111888111888, "percentage": 90.37, "elapsed_time": "1 day, 15:39:26", "remaining_time": "4:13:34", "throughput": "17781.45", "total_tokens": 2538602496}
{"current_steps": 2422, "total_steps": 2679, "loss": 0.6515, "learning_rate": 1.1479479041335368e-07, "epoch": 2.7099300699300697, "percentage": 90.41, "elapsed_time": "1 day, 15:40:26", "remaining_time": "4:12:35", "throughput": "17781.34", "total_tokens": 2539651072}
{"current_steps": 2423, "total_steps": 2679, "loss": 0.745, "learning_rate": 1.1391001739065432e-07, "epoch": 2.711048951048951, "percentage": 90.44, "elapsed_time": "1 day, 15:41:31", "remaining_time": "4:11:37", "throughput": "17780.61", "total_tokens": 2540699648}
{"current_steps": 2424, "total_steps": 2679, "loss": 0.8179, "learning_rate": 1.1302858773975634e-07, "epoch": 2.7121678321678324, "percentage": 90.48, "elapsed_time": "1 day, 15:42:42", "remaining_time": "4:10:39", "throughput": "17779.07", "total_tokens": 2541748224}
{"current_steps": 2425, "total_steps": 2679, "loss": 0.7885, "learning_rate": 1.121505026957112e-07, "epoch": 2.7132867132867133, "percentage": 90.52, "elapsed_time": "1 day, 15:43:54", "remaining_time": "4:09:41", "throughput": "17777.53", "total_tokens": 2542796800}
{"current_steps": 2426, "total_steps": 2679, "loss": 0.6156, "learning_rate": 1.1127576348888502e-07, "epoch": 2.7144055944055943, "percentage": 90.56, "elapsed_time": "1 day, 15:44:47", "remaining_time": "4:08:42", "throughput": "17778.19", "total_tokens": 2543845376}
{"current_steps": 2427, "total_steps": 2679, "loss": 0.6352, "learning_rate": 1.1040437134495708e-07, "epoch": 2.7155244755244756, "percentage": 90.59, "elapsed_time": "1 day, 15:45:41", "remaining_time": "4:07:42", "throughput": "17778.82", "total_tokens": 2544893952}
{"current_steps": 2428, "total_steps": 2679, "loss": 0.8094, "learning_rate": 1.0953632748491455e-07, "epoch": 2.7166433566433565, "percentage": 90.63, "elapsed_time": "1 day, 15:46:43", "remaining_time": "4:06:43", "throughput": "17778.49", "total_tokens": 2545942528}
{"current_steps": 2429, "total_steps": 2679, "loss": 0.6211, "learning_rate": 1.0867163312505452e-07, "epoch": 2.717762237762238, "percentage": 90.67, "elapsed_time": "1 day, 15:47:38", "remaining_time": "4:05:44", "throughput": "17779.05", "total_tokens": 2546991104}
{"current_steps": 2430, "total_steps": 2679, "loss": 0.6925, "learning_rate": 1.0781028947698113e-07, "epoch": 2.718881118881119, "percentage": 90.71, "elapsed_time": "1 day, 15:48:32", "remaining_time": "4:04:45", "throughput": "17779.65", "total_tokens": 2548039680}
{"current_steps": 2431, "total_steps": 2679, "loss": 0.7175, "learning_rate": 1.0695229774760147e-07, "epoch": 2.7199999999999998, "percentage": 90.74, "elapsed_time": "1 day, 15:49:32", "remaining_time": "4:03:46", "throughput": "17779.49", "total_tokens": 2549088256}
{"current_steps": 2432, "total_steps": 2679, "loss": 0.5854, "learning_rate": 1.0609765913912828e-07, "epoch": 2.721118881118881, "percentage": 90.78, "elapsed_time": "1 day, 15:50:20", "remaining_time": "4:02:46", "throughput": "17780.80", "total_tokens": 2550136832}
{"current_steps": 2433, "total_steps": 2679, "loss": 0.6098, "learning_rate": 1.0524637484907424e-07, "epoch": 2.7222377622377625, "percentage": 90.82, "elapsed_time": "1 day, 15:51:12", "remaining_time": "4:01:46", "throughput": "17781.69", "total_tokens": 2551185408}
{"current_steps": 2434, "total_steps": 2679, "loss": 0.8065, "learning_rate": 1.0439844607025324e-07, "epoch": 2.7233566433566434, "percentage": 90.85, "elapsed_time": "1 day, 15:52:19", "remaining_time": "4:00:48", "throughput": "17780.69", "total_tokens": 2552233984}
{"current_steps": 2435, "total_steps": 2679, "loss": 0.6329, "learning_rate": 1.0355387399077627e-07, "epoch": 2.7244755244755243, "percentage": 90.89, "elapsed_time": "1 day, 15:53:13", "remaining_time": "3:59:48", "throughput": "17781.36", "total_tokens": 2553282560}
{"current_steps": 2436, "total_steps": 2679, "loss": 0.5419, "learning_rate": 1.0271265979405254e-07, "epoch": 2.7255944055944057, "percentage": 90.93, "elapsed_time": "1 day, 15:54:01", "remaining_time": "3:58:48", "throughput": "17782.72", "total_tokens": 2554331136}
{"current_steps": 2437, "total_steps": 2679, "loss": 0.5454, "learning_rate": 1.0187480465878418e-07, "epoch": 2.7267132867132866, "percentage": 90.97, "elapsed_time": "1 day, 15:54:48", "remaining_time": "3:57:48", "throughput": "17784.11", "total_tokens": 2555379712}
{"current_steps": 2438, "total_steps": 2679, "loss": 0.8302, "learning_rate": 1.0104030975896794e-07, "epoch": 2.727832167832168, "percentage": 91.0, "elapsed_time": "1 day, 15:55:55", "remaining_time": "3:56:50", "throughput": "17783.18", "total_tokens": 2556428288}
{"current_steps": 2439, "total_steps": 2679, "loss": 0.693, "learning_rate": 1.0020917626389209e-07, "epoch": 2.728951048951049, "percentage": 91.04, "elapsed_time": "1 day, 15:56:55", "remaining_time": "3:55:51", "throughput": "17783.10", "total_tokens": 2557476864}
{"current_steps": 2440, "total_steps": 2679, "loss": 0.5975, "learning_rate": 9.938140533813478e-08, "epoch": 2.73006993006993, "percentage": 91.08, "elapsed_time": "1 day, 15:57:48", "remaining_time": "3:54:52", "throughput": "17783.76", "total_tokens": 2558525440}
{"current_steps": 2441, "total_steps": 2679, "loss": 0.8575, "learning_rate": 9.855699814156266e-08, "epoch": 2.731188811188811, "percentage": 91.12, "elapsed_time": "1 day, 15:58:54", "remaining_time": "3:53:53", "throughput": "17782.92", "total_tokens": 2559574016}
{"current_steps": 2442, "total_steps": 2679, "loss": 0.7426, "learning_rate": 9.773595582932921e-08, "epoch": 2.7323076923076925, "percentage": 91.15, "elapsed_time": "1 day, 16:00:01", "remaining_time": "3:52:55", "throughput": "17781.93", "total_tokens": 2560622592}
{"current_steps": 2443, "total_steps": 2679, "loss": 0.6154, "learning_rate": 9.691827955187222e-08, "epoch": 2.7334265734265735, "percentage": 91.19, "elapsed_time": "1 day, 16:00:55", "remaining_time": "3:51:56", "throughput": "17782.56", "total_tokens": 2561671168}
{"current_steps": 2444, "total_steps": 2679, "loss": 0.6935, "learning_rate": 9.61039704549141e-08, "epoch": 2.7345454545454544, "percentage": 91.23, "elapsed_time": "1 day, 16:01:55", "remaining_time": "3:50:57", "throughput": "17782.46", "total_tokens": 2562719744}
{"current_steps": 2445, "total_steps": 2679, "loss": 0.6179, "learning_rate": 9.52930296794588e-08, "epoch": 2.7356643356643358, "percentage": 91.27, "elapsed_time": "1 day, 16:02:48", "remaining_time": "3:49:57", "throughput": "17783.10", "total_tokens": 2563768320}
{"current_steps": 2446, "total_steps": 2679, "loss": 0.7338, "learning_rate": 9.448545836179102e-08, "epoch": 2.7367832167832167, "percentage": 91.3, "elapsed_time": "1 day, 16:03:48", "remaining_time": "3:48:58", "throughput": "17782.98", "total_tokens": 2564816896}
{"current_steps": 2447, "total_steps": 2679, "loss": 0.6381, "learning_rate": 9.368125763347336e-08, "epoch": 2.737902097902098, "percentage": 91.34, "elapsed_time": "1 day, 16:04:48", "remaining_time": "3:47:59", "throughput": "17782.94", "total_tokens": 2565865472}
{"current_steps": 2448, "total_steps": 2679, "loss": 0.7011, "learning_rate": 9.288042862134556e-08, "epoch": 2.739020979020979, "percentage": 91.38, "elapsed_time": "1 day, 16:05:47", "remaining_time": "3:47:00", "throughput": "17782.93", "total_tokens": 2566914048}
{"current_steps": 2449, "total_steps": 2679, "loss": 0.6146, "learning_rate": 9.208297244752362e-08, "epoch": 2.74013986013986, "percentage": 91.41, "elapsed_time": "1 day, 16:06:41", "remaining_time": "3:46:01", "throughput": "17783.46", "total_tokens": 2567962624}
{"current_steps": 2450, "total_steps": 2679, "loss": 0.6778, "learning_rate": 9.128889022939791e-08, "epoch": 2.7412587412587412, "percentage": 91.45, "elapsed_time": "1 day, 16:07:40", "remaining_time": "3:45:02", "throughput": "17783.42", "total_tokens": 2569011200}
{"current_steps": 2451, "total_steps": 2679, "loss": 0.5959, "learning_rate": 9.049818307963004e-08, "epoch": 2.7423776223776226, "percentage": 91.49, "elapsed_time": "1 day, 16:08:35", "remaining_time": "3:44:03", "throughput": "17784.03", "total_tokens": 2570059776}
{"current_steps": 2452, "total_steps": 2679, "loss": 0.6909, "learning_rate": 8.971085210615321e-08, "epoch": 2.7434965034965035, "percentage": 91.53, "elapsed_time": "1 day, 16:09:34", "remaining_time": "3:43:04", "throughput": "17783.91", "total_tokens": 2571108352}
{"current_steps": 2453, "total_steps": 2679, "loss": 0.5937, "learning_rate": 8.892689841216995e-08, "epoch": 2.7446153846153845, "percentage": 91.56, "elapsed_time": "1 day, 16:10:25", "remaining_time": "3:42:04", "throughput": "17784.89", "total_tokens": 2572156928}
{"current_steps": 2454, "total_steps": 2679, "loss": 0.7223, "learning_rate": 8.81463230961499e-08, "epoch": 2.745734265734266, "percentage": 91.6, "elapsed_time": "1 day, 16:11:25", "remaining_time": "3:41:05", "throughput": "17784.79", "total_tokens": 2573205504}
{"current_steps": 2455, "total_steps": 2679, "loss": 0.6198, "learning_rate": 8.736912725182983e-08, "epoch": 2.7468531468531467, "percentage": 91.64, "elapsed_time": "1 day, 16:12:19", "remaining_time": "3:40:06", "throughput": "17785.43", "total_tokens": 2574254080}
{"current_steps": 2456, "total_steps": 2679, "loss": 0.5295, "learning_rate": 8.659531196821142e-08, "epoch": 2.747972027972028, "percentage": 91.68, "elapsed_time": "1 day, 16:13:07", "remaining_time": "3:39:06", "throughput": "17786.73", "total_tokens": 2575302656}
{"current_steps": 2457, "total_steps": 2679, "loss": 0.654, "learning_rate": 8.582487832955788e-08, "epoch": 2.749090909090909, "percentage": 91.71, "elapsed_time": "1 day, 16:14:01", "remaining_time": "3:38:07", "throughput": "17787.34", "total_tokens": 2576351232}
{"current_steps": 2458, "total_steps": 2679, "loss": 0.6923, "learning_rate": 8.505782741539626e-08, "epoch": 2.75020979020979, "percentage": 91.75, "elapsed_time": "1 day, 16:15:01", "remaining_time": "3:37:08", "throughput": "17787.24", "total_tokens": 2577399808}
{"current_steps": 2459, "total_steps": 2679, "loss": 0.7479, "learning_rate": 8.429416030051179e-08, "epoch": 2.7513286713286713, "percentage": 91.79, "elapsed_time": "1 day, 16:16:07", "remaining_time": "3:36:09", "throughput": "17786.42", "total_tokens": 2578448384}
{"current_steps": 2460, "total_steps": 2679, "loss": 0.6552, "learning_rate": 8.353387805494967e-08, "epoch": 2.7524475524475527, "percentage": 91.83, "elapsed_time": "1 day, 16:17:06", "remaining_time": "3:35:10", "throughput": "17786.35", "total_tokens": 2579496960}
{"current_steps": 2461, "total_steps": 2679, "loss": 0.6799, "learning_rate": 8.277698174401189e-08, "epoch": 2.7535664335664336, "percentage": 91.86, "elapsed_time": "1 day, 16:18:06", "remaining_time": "3:34:12", "throughput": "17786.26", "total_tokens": 2580545536}
{"current_steps": 2462, "total_steps": 2679, "loss": 0.5874, "learning_rate": 8.202347242825565e-08, "epoch": 2.7546853146853145, "percentage": 91.9, "elapsed_time": "1 day, 16:18:54", "remaining_time": "3:33:12", "throughput": "17787.55", "total_tokens": 2581594112}
{"current_steps": 2463, "total_steps": 2679, "loss": 0.5944, "learning_rate": 8.127335116349305e-08, "epoch": 2.755804195804196, "percentage": 91.94, "elapsed_time": "1 day, 16:19:48", "remaining_time": "3:32:12", "throughput": "17788.19", "total_tokens": 2582642688}
{"current_steps": 2464, "total_steps": 2679, "loss": 0.6769, "learning_rate": 8.052661900078835e-08, "epoch": 2.756923076923077, "percentage": 91.97, "elapsed_time": "1 day, 16:20:45", "remaining_time": "3:31:13", "throughput": "17788.43", "total_tokens": 2583691264}
{"current_steps": 2465, "total_steps": 2679, "loss": 0.725, "learning_rate": 7.978327698645705e-08, "epoch": 2.758041958041958, "percentage": 92.01, "elapsed_time": "1 day, 16:21:46", "remaining_time": "3:30:14", "throughput": "17788.22", "total_tokens": 2584739840}
{"current_steps": 2466, "total_steps": 2679, "loss": 0.6167, "learning_rate": 7.90433261620649e-08, "epoch": 2.759160839160839, "percentage": 92.05, "elapsed_time": "1 day, 16:22:41", "remaining_time": "3:29:15", "throughput": "17788.64", "total_tokens": 2585788416}
{"current_steps": 2467, "total_steps": 2679, "loss": 0.5821, "learning_rate": 7.830676756442529e-08, "epoch": 2.76027972027972, "percentage": 92.09, "elapsed_time": "1 day, 16:23:35", "remaining_time": "3:28:16", "throughput": "17789.24", "total_tokens": 2586836992}
{"current_steps": 2468, "total_steps": 2679, "loss": 0.6192, "learning_rate": 7.757360222559878e-08, "epoch": 2.7613986013986014, "percentage": 92.12, "elapsed_time": "1 day, 16:24:29", "remaining_time": "3:27:16", "throughput": "17789.89", "total_tokens": 2587885568}
{"current_steps": 2469, "total_steps": 2679, "loss": 0.715, "learning_rate": 7.684383117289141e-08, "epoch": 2.7625174825174827, "percentage": 92.16, "elapsed_time": "1 day, 16:25:34", "remaining_time": "3:26:18", "throughput": "17789.11", "total_tokens": 2588934144}
{"current_steps": 2470, "total_steps": 2679, "loss": 0.7061, "learning_rate": 7.61174554288524e-08, "epoch": 2.7636363636363637, "percentage": 92.2, "elapsed_time": "1 day, 16:26:34", "remaining_time": "3:25:19", "throughput": "17789.02", "total_tokens": 2589982720}
{"current_steps": 2471, "total_steps": 2679, "loss": 0.7753, "learning_rate": 7.539447601127542e-08, "epoch": 2.7647552447552446, "percentage": 92.24, "elapsed_time": "1 day, 16:27:34", "remaining_time": "3:24:20", "throughput": "17788.84", "total_tokens": 2591031296}
{"current_steps": 2472, "total_steps": 2679, "loss": 0.6054, "learning_rate": 7.46748939331926e-08, "epoch": 2.765874125874126, "percentage": 92.27, "elapsed_time": "1 day, 16:28:28", "remaining_time": "3:23:21", "throughput": "17789.47", "total_tokens": 2592079872}
{"current_steps": 2473, "total_steps": 2679, "loss": 0.7063, "learning_rate": 7.39587102028777e-08, "epoch": 2.766993006993007, "percentage": 92.31, "elapsed_time": "1 day, 16:29:28", "remaining_time": "3:22:22", "throughput": "17789.34", "total_tokens": 2593128448}
{"current_steps": 2474, "total_steps": 2679, "loss": 0.6988, "learning_rate": 7.324592582384215e-08, "epoch": 2.7681118881118882, "percentage": 92.35, "elapsed_time": "1 day, 16:30:28", "remaining_time": "3:21:23", "throughput": "17789.23", "total_tokens": 2594177024}
{"current_steps": 2475, "total_steps": 2679, "loss": 0.553, "learning_rate": 7.2536541794834e-08, "epoch": 2.769230769230769, "percentage": 92.39, "elapsed_time": "1 day, 16:31:17", "remaining_time": "3:20:23", "throughput": "17790.44", "total_tokens": 2595225600}
{"current_steps": 2476, "total_steps": 2679, "loss": 0.6731, "learning_rate": 7.183055910983671e-08, "epoch": 2.77034965034965, "percentage": 92.42, "elapsed_time": "1 day, 16:32:11", "remaining_time": "3:19:24", "throughput": "17791.03", "total_tokens": 2596274176}
{"current_steps": 2477, "total_steps": 2679, "loss": 0.6203, "learning_rate": 7.112797875806904e-08, "epoch": 2.7714685314685314, "percentage": 92.46, "elapsed_time": "1 day, 16:33:06", "remaining_time": "3:18:25", "throughput": "17791.55", "total_tokens": 2597322752}
{"current_steps": 2478, "total_steps": 2679, "loss": 0.6989, "learning_rate": 7.042880172398043e-08, "epoch": 2.772587412587413, "percentage": 92.5, "elapsed_time": "1 day, 16:34:06", "remaining_time": "3:17:26", "throughput": "17791.38", "total_tokens": 2598371328}
{"current_steps": 2479, "total_steps": 2679, "loss": 0.6745, "learning_rate": 6.973302898725303e-08, "epoch": 2.7737062937062937, "percentage": 92.53, "elapsed_time": "1 day, 16:35:06", "remaining_time": "3:16:27", "throughput": "17791.23", "total_tokens": 2599419904}
{"current_steps": 2480, "total_steps": 2679, "loss": 0.7327, "learning_rate": 6.904066152279815e-08, "epoch": 2.7748251748251747, "percentage": 92.57, "elapsed_time": "1 day, 16:36:12", "remaining_time": "3:15:29", "throughput": "17790.45", "total_tokens": 2600468480}
{"current_steps": 2481, "total_steps": 2679, "loss": 0.6567, "learning_rate": 6.835170030075638e-08, "epoch": 2.775944055944056, "percentage": 92.61, "elapsed_time": "1 day, 16:37:06", "remaining_time": "3:14:29", "throughput": "17790.98", "total_tokens": 2601517056}
{"current_steps": 2482, "total_steps": 2679, "loss": 0.6064, "learning_rate": 6.766614628649525e-08, "epoch": 2.777062937062937, "percentage": 92.65, "elapsed_time": "1 day, 16:38:00", "remaining_time": "3:13:30", "throughput": "17791.57", "total_tokens": 2602565632}
{"current_steps": 2483, "total_steps": 2679, "loss": 0.6902, "learning_rate": 6.698400044060777e-08, "epoch": 2.7781818181818183, "percentage": 92.68, "elapsed_time": "1 day, 16:39:05", "remaining_time": "3:12:32", "throughput": "17790.83", "total_tokens": 2603614208}
{"current_steps": 2484, "total_steps": 2679, "loss": 0.7422, "learning_rate": 6.630526371891155e-08, "epoch": 2.7793006993006992, "percentage": 92.72, "elapsed_time": "1 day, 16:40:11", "remaining_time": "3:11:33", "throughput": "17790.04", "total_tokens": 2604662784}
{"current_steps": 2485, "total_steps": 2679, "loss": 0.6695, "learning_rate": 6.56299370724478e-08, "epoch": 2.78041958041958, "percentage": 92.76, "elapsed_time": "1 day, 16:41:11", "remaining_time": "3:10:34", "throughput": "17789.94", "total_tokens": 2605711360}
{"current_steps": 2486, "total_steps": 2679, "loss": 0.7529, "learning_rate": 6.495802144747904e-08, "epoch": 2.7815384615384615, "percentage": 92.8, "elapsed_time": "1 day, 16:42:11", "remaining_time": "3:09:35", "throughput": "17789.75", "total_tokens": 2606759936}
{"current_steps": 2487, "total_steps": 2679, "loss": 0.752, "learning_rate": 6.428951778548881e-08, "epoch": 2.782657342657343, "percentage": 92.83, "elapsed_time": "1 day, 16:43:11", "remaining_time": "3:08:37", "throughput": "17789.65", "total_tokens": 2607808512}
{"current_steps": 2488, "total_steps": 2679, "loss": 0.6004, "learning_rate": 6.362442702317923e-08, "epoch": 2.783776223776224, "percentage": 92.87, "elapsed_time": "1 day, 16:44:05", "remaining_time": "3:07:37", "throughput": "17790.24", "total_tokens": 2608857088}
{"current_steps": 2489, "total_steps": 2679, "loss": 0.6841, "learning_rate": 6.296275009247121e-08, "epoch": 2.7848951048951047, "percentage": 92.91, "elapsed_time": "1 day, 16:45:08", "remaining_time": "3:06:39", "throughput": "17789.74", "total_tokens": 2609905664}
{"current_steps": 2490, "total_steps": 2679, "loss": 0.572, "learning_rate": 6.230448792050065e-08, "epoch": 2.786013986013986, "percentage": 92.95, "elapsed_time": "1 day, 16:45:57", "remaining_time": "3:05:39", "throughput": "17790.94", "total_tokens": 2610954240}
{"current_steps": 2491, "total_steps": 2679, "loss": 0.6711, "learning_rate": 6.164964142962027e-08, "epoch": 2.787132867132867, "percentage": 92.98, "elapsed_time": "1 day, 16:46:51", "remaining_time": "3:04:40", "throughput": "17791.51", "total_tokens": 2612002816}
{"current_steps": 2492, "total_steps": 2679, "loss": 0.6412, "learning_rate": 6.099821153739637e-08, "epoch": 2.7882517482517484, "percentage": 93.02, "elapsed_time": "1 day, 16:47:45", "remaining_time": "3:03:40", "throughput": "17792.15", "total_tokens": 2613051392}
{"current_steps": 2493, "total_steps": 2679, "loss": 0.6598, "learning_rate": 6.035019915660717e-08, "epoch": 2.7893706293706293, "percentage": 93.06, "elapsed_time": "1 day, 16:48:41", "remaining_time": "3:02:41", "throughput": "17792.44", "total_tokens": 2614099968}
{"current_steps": 2494, "total_steps": 2679, "loss": 0.6347, "learning_rate": 5.970560519524327e-08, "epoch": 2.79048951048951, "percentage": 93.09, "elapsed_time": "1 day, 16:49:36", "remaining_time": "3:01:42", "throughput": "17793.01", "total_tokens": 2615148544}
{"current_steps": 2495, "total_steps": 2679, "loss": 0.6806, "learning_rate": 5.906443055650496e-08, "epoch": 2.7916083916083916, "percentage": 93.13, "elapsed_time": "1 day, 16:50:36", "remaining_time": "3:00:43", "throughput": "17792.87", "total_tokens": 2616197120}
{"current_steps": 2496, "total_steps": 2679, "loss": 0.6992, "learning_rate": 5.84266761388011e-08, "epoch": 2.792727272727273, "percentage": 93.17, "elapsed_time": "1 day, 16:51:35", "remaining_time": "2:59:44", "throughput": "17792.77", "total_tokens": 2617245696}
{"current_steps": 2497, "total_steps": 2679, "loss": 0.7976, "learning_rate": 5.779234283574936e-08, "epoch": 2.793846153846154, "percentage": 93.21, "elapsed_time": "1 day, 16:52:35", "remaining_time": "2:58:45", "throughput": "17792.66", "total_tokens": 2618294272}
{"current_steps": 2498, "total_steps": 2679, "loss": 0.6158, "learning_rate": 5.7161431536171816e-08, "epoch": 2.794965034965035, "percentage": 93.24, "elapsed_time": "1 day, 16:53:29", "remaining_time": "2:57:46", "throughput": "17793.24", "total_tokens": 2619342848}
{"current_steps": 2499, "total_steps": 2679, "loss": 0.6014, "learning_rate": 5.653394312409771e-08, "epoch": 2.796083916083916, "percentage": 93.28, "elapsed_time": "1 day, 16:54:26", "remaining_time": "2:56:47", "throughput": "17793.54", "total_tokens": 2620391424}
{"current_steps": 2500, "total_steps": 2679, "loss": 0.7546, "learning_rate": 5.590987847875845e-08, "epoch": 2.797202797202797, "percentage": 93.32, "elapsed_time": "1 day, 16:55:31", "remaining_time": "2:55:48", "throughput": "17792.81", "total_tokens": 2621440000}
{"current_steps": 2500, "total_steps": 2679, "eval_loss": 0.7187779545783997, "epoch": 2.797202797202797, "percentage": 93.32, "elapsed_time": "1 day, 16:59:38", "remaining_time": "2:56:06", "throughput": "17763.04", "total_tokens": 2621440000}
{"current_steps": 2501, "total_steps": 2679, "loss": 0.5569, "learning_rate": 5.528923847458928e-08, "epoch": 2.7983216783216784, "percentage": 93.36, "elapsed_time": "1 day, 17:00:57", "remaining_time": "2:55:09", "throughput": "17760.57", "total_tokens": 2622488576}
{"current_steps": 2502, "total_steps": 2679, "loss": 0.7357, "learning_rate": 5.467202398122651e-08, "epoch": 2.7994405594405594, "percentage": 93.39, "elapsed_time": "1 day, 17:02:03", "remaining_time": "2:54:10", "throughput": "17759.78", "total_tokens": 2623537152}
{"current_steps": 2503, "total_steps": 2679, "loss": 0.7066, "learning_rate": 5.4058235863506116e-08, "epoch": 2.8005594405594403, "percentage": 93.43, "elapsed_time": "1 day, 17:03:03", "remaining_time": "2:53:11", "throughput": "17759.61", "total_tokens": 2624585728}
{"current_steps": 2504, "total_steps": 2679, "loss": 0.8059, "learning_rate": 5.3447874981464034e-08, "epoch": 2.8016783216783216, "percentage": 93.47, "elapsed_time": "1 day, 17:04:07", "remaining_time": "2:52:12", "throughput": "17759.09", "total_tokens": 2625634304}
{"current_steps": 2505, "total_steps": 2679, "loss": 0.6861, "learning_rate": 5.2840942190333086e-08, "epoch": 2.802797202797203, "percentage": 93.51, "elapsed_time": "1 day, 17:05:06", "remaining_time": "2:51:13", "throughput": "17759.04", "total_tokens": 2626682880}
{"current_steps": 2506, "total_steps": 2679, "loss": 0.6513, "learning_rate": 5.223743834054329e-08, "epoch": 2.803916083916084, "percentage": 93.54, "elapsed_time": "1 day, 17:06:06", "remaining_time": "2:50:14", "throughput": "17758.98", "total_tokens": 2627731456}
{"current_steps": 2507, "total_steps": 2679, "loss": 0.5416, "learning_rate": 5.1637364277719595e-08, "epoch": 2.805034965034965, "percentage": 93.58, "elapsed_time": "1 day, 17:06:54", "remaining_time": "2:49:14", "throughput": "17760.29", "total_tokens": 2628780032}
{"current_steps": 2508, "total_steps": 2679, "loss": 0.6172, "learning_rate": 5.104072084268136e-08, "epoch": 2.806153846153846, "percentage": 93.62, "elapsed_time": "1 day, 17:07:53", "remaining_time": "2:48:15", "throughput": "17760.24", "total_tokens": 2629828608}
{"current_steps": 2509, "total_steps": 2679, "loss": 0.5928, "learning_rate": 5.044750887144151e-08, "epoch": 2.807272727272727, "percentage": 93.65, "elapsed_time": "1 day, 17:08:47", "remaining_time": "2:47:16", "throughput": "17760.86", "total_tokens": 2630877184}
{"current_steps": 2510, "total_steps": 2679, "loss": 0.6963, "learning_rate": 4.9857729195203486e-08, "epoch": 2.8083916083916085, "percentage": 93.69, "elapsed_time": "1 day, 17:09:47", "remaining_time": "2:46:17", "throughput": "17760.79", "total_tokens": 2631925760}
{"current_steps": 2511, "total_steps": 2679, "loss": 0.7047, "learning_rate": 4.927138264036291e-08, "epoch": 2.8095104895104894, "percentage": 93.73, "elapsed_time": "1 day, 17:10:46", "remaining_time": "2:45:18", "throughput": "17760.75", "total_tokens": 2632974336}
{"current_steps": 2512, "total_steps": 2679, "loss": 0.6311, "learning_rate": 4.8688470028503966e-08, "epoch": 2.810629370629371, "percentage": 93.77, "elapsed_time": "1 day, 17:11:40", "remaining_time": "2:44:19", "throughput": "17761.39", "total_tokens": 2634022912}
{"current_steps": 2513, "total_steps": 2679, "loss": 0.6982, "learning_rate": 4.810899217639997e-08, "epoch": 2.8117482517482517, "percentage": 93.8, "elapsed_time": "1 day, 17:12:45", "remaining_time": "2:43:20", "throughput": "17760.62", "total_tokens": 2635071488}
{"current_steps": 2514, "total_steps": 2679, "loss": 0.8509, "learning_rate": 4.753294989601032e-08, "epoch": 2.812867132867133, "percentage": 93.84, "elapsed_time": "1 day, 17:13:52", "remaining_time": "2:42:21", "throughput": "17759.76", "total_tokens": 2636120064}
{"current_steps": 2515, "total_steps": 2679, "loss": 0.669, "learning_rate": 4.696034399448185e-08, "epoch": 2.813986013986014, "percentage": 93.88, "elapsed_time": "1 day, 17:14:46", "remaining_time": "2:41:22", "throughput": "17760.34", "total_tokens": 2637168640}
{"current_steps": 2516, "total_steps": 2679, "loss": 0.5719, "learning_rate": 4.639117527414527e-08, "epoch": 2.815104895104895, "percentage": 93.92, "elapsed_time": "1 day, 17:15:35", "remaining_time": "2:40:22", "throughput": "17761.56", "total_tokens": 2638217216}
{"current_steps": 2517, "total_steps": 2679, "loss": 0.6416, "learning_rate": 4.582544453251597e-08, "epoch": 2.8162237762237763, "percentage": 93.95, "elapsed_time": "1 day, 17:16:29", "remaining_time": "2:39:23", "throughput": "17762.18", "total_tokens": 2639265792}
{"current_steps": 2518, "total_steps": 2679, "loss": 0.6419, "learning_rate": 4.52631525622918e-08, "epoch": 2.817342657342657, "percentage": 93.99, "elapsed_time": "1 day, 17:17:28", "remaining_time": "2:38:24", "throughput": "17762.11", "total_tokens": 2640314368}
{"current_steps": 2519, "total_steps": 2679, "loss": 0.7284, "learning_rate": 4.470430015135197e-08, "epoch": 2.8184615384615386, "percentage": 94.03, "elapsed_time": "1 day, 17:18:34", "remaining_time": "2:37:25", "throughput": "17761.36", "total_tokens": 2641362944}
{"current_steps": 2520, "total_steps": 2679, "loss": 0.7285, "learning_rate": 4.414888808275619e-08, "epoch": 2.8195804195804195, "percentage": 94.06, "elapsed_time": "1 day, 17:19:34", "remaining_time": "2:36:26", "throughput": "17761.20", "total_tokens": 2642411520}
{"current_steps": 2521, "total_steps": 2679, "loss": 0.6511, "learning_rate": 4.35969171347439e-08, "epoch": 2.820699300699301, "percentage": 94.1, "elapsed_time": "1 day, 17:20:34", "remaining_time": "2:35:27", "throughput": "17761.12", "total_tokens": 2643460096}
{"current_steps": 2522, "total_steps": 2679, "loss": 0.7493, "learning_rate": 4.304838808073281e-08, "epoch": 2.821818181818182, "percentage": 94.14, "elapsed_time": "1 day, 17:21:39", "remaining_time": "2:34:29", "throughput": "17760.33", "total_tokens": 2644508672}
{"current_steps": 2523, "total_steps": 2679, "loss": 0.6405, "learning_rate": 4.25033016893181e-08, "epoch": 2.822937062937063, "percentage": 94.18, "elapsed_time": "1 day, 17:22:33", "remaining_time": "2:33:29", "throughput": "17760.93", "total_tokens": 2645557248}
{"current_steps": 2524, "total_steps": 2679, "loss": 0.5696, "learning_rate": 4.1961658724270496e-08, "epoch": 2.824055944055944, "percentage": 94.21, "elapsed_time": "1 day, 17:23:27", "remaining_time": "2:32:30", "throughput": "17761.53", "total_tokens": 2646605824}
{"current_steps": 2525, "total_steps": 2679, "loss": 0.6327, "learning_rate": 4.1423459944536224e-08, "epoch": 2.825174825174825, "percentage": 94.25, "elapsed_time": "1 day, 17:24:27", "remaining_time": "2:31:31", "throughput": "17761.47", "total_tokens": 2647654400}
{"current_steps": 2526, "total_steps": 2679, "loss": 0.6895, "learning_rate": 4.088870610423512e-08, "epoch": 2.8262937062937064, "percentage": 94.29, "elapsed_time": "1 day, 17:25:32", "remaining_time": "2:30:32", "throughput": "17760.75", "total_tokens": 2648702976}
{"current_steps": 2527, "total_steps": 2679, "loss": 0.8083, "learning_rate": 4.035739795266086e-08, "epoch": 2.8274125874125873, "percentage": 94.33, "elapsed_time": "1 day, 17:26:38", "remaining_time": "2:29:34", "throughput": "17759.96", "total_tokens": 2649751552}
{"current_steps": 2528, "total_steps": 2679, "loss": 0.6886, "learning_rate": 3.982953623427876e-08, "epoch": 2.8285314685314686, "percentage": 94.36, "elapsed_time": "1 day, 17:27:31", "remaining_time": "2:28:34", "throughput": "17760.59", "total_tokens": 2650800128}
{"current_steps": 2529, "total_steps": 2679, "loss": 0.517, "learning_rate": 3.9305121688723855e-08, "epoch": 2.8296503496503496, "percentage": 94.4, "elapsed_time": "1 day, 17:28:21", "remaining_time": "2:27:35", "throughput": "17761.71", "total_tokens": 2651848704}
{"current_steps": 2530, "total_steps": 2679, "loss": 0.6603, "learning_rate": 3.87841550508028e-08, "epoch": 2.830769230769231, "percentage": 94.44, "elapsed_time": "1 day, 17:29:15", "remaining_time": "2:26:36", "throughput": "17762.29", "total_tokens": 2652897280}
{"current_steps": 2531, "total_steps": 2679, "loss": 0.6326, "learning_rate": 3.8266637050489716e-08, "epoch": 2.831888111888112, "percentage": 94.48, "elapsed_time": "1 day, 17:30:12", "remaining_time": "2:25:36", "throughput": "17762.51", "total_tokens": 2653945856}
{"current_steps": 2532, "total_steps": 2679, "loss": 0.6368, "learning_rate": 3.7752568412927346e-08, "epoch": 2.833006993006993, "percentage": 94.51, "elapsed_time": "1 day, 17:31:06", "remaining_time": "2:24:37", "throughput": "17763.14", "total_tokens": 2654994432}
{"current_steps": 2533, "total_steps": 2679, "loss": 0.599, "learning_rate": 3.7241949858424777e-08, "epoch": 2.834125874125874, "percentage": 94.55, "elapsed_time": "1 day, 17:32:00", "remaining_time": "2:23:38", "throughput": "17763.72", "total_tokens": 2656043008}
{"current_steps": 2534, "total_steps": 2679, "loss": 0.6076, "learning_rate": 3.673478210245718e-08, "epoch": 2.835244755244755, "percentage": 94.59, "elapsed_time": "1 day, 17:32:54", "remaining_time": "2:22:38", "throughput": "17764.30", "total_tokens": 2657091584}
{"current_steps": 2535, "total_steps": 2679, "loss": 0.6969, "learning_rate": 3.623106585566388e-08, "epoch": 2.8363636363636364, "percentage": 94.62, "elapsed_time": "1 day, 17:33:54", "remaining_time": "2:21:39", "throughput": "17764.19", "total_tokens": 2658140160}
{"current_steps": 2536, "total_steps": 2679, "loss": 0.6353, "learning_rate": 3.573080182384864e-08, "epoch": 2.8374825174825173, "percentage": 94.66, "elapsed_time": "1 day, 17:34:55", "remaining_time": "2:20:41", "throughput": "17764.02", "total_tokens": 2659188736}
{"current_steps": 2537, "total_steps": 2679, "loss": 0.7189, "learning_rate": 3.523399070797795e-08, "epoch": 2.8386013986013987, "percentage": 94.7, "elapsed_time": "1 day, 17:36:00", "remaining_time": "2:19:42", "throughput": "17763.32", "total_tokens": 2660237312}
{"current_steps": 2538, "total_steps": 2679, "loss": 0.5916, "learning_rate": 3.47406332041797e-08, "epoch": 2.8397202797202796, "percentage": 94.74, "elapsed_time": "1 day, 17:36:54", "remaining_time": "2:18:43", "throughput": "17763.92", "total_tokens": 2661285888}
{"current_steps": 2539, "total_steps": 2679, "loss": 0.5792, "learning_rate": 3.425073000374257e-08, "epoch": 2.840839160839161, "percentage": 94.77, "elapsed_time": "1 day, 17:37:42", "remaining_time": "2:17:43", "throughput": "17765.22", "total_tokens": 2662334464}
{"current_steps": 2540, "total_steps": 2679, "loss": 0.6197, "learning_rate": 3.3764281793115804e-08, "epoch": 2.841958041958042, "percentage": 94.81, "elapsed_time": "1 day, 17:38:41", "remaining_time": "2:16:44", "throughput": "17765.22", "total_tokens": 2663383040}
{"current_steps": 2541, "total_steps": 2679, "loss": 0.6129, "learning_rate": 3.328128925390667e-08, "epoch": 2.8430769230769233, "percentage": 94.85, "elapsed_time": "1 day, 17:39:34", "remaining_time": "2:15:45", "throughput": "17765.88", "total_tokens": 2664431616}
{"current_steps": 2542, "total_steps": 2679, "loss": 0.6925, "learning_rate": 3.280175306288103e-08, "epoch": 2.844195804195804, "percentage": 94.89, "elapsed_time": "1 day, 17:40:40", "remaining_time": "2:14:46", "throughput": "17765.12", "total_tokens": 2665480192}
{"current_steps": 2543, "total_steps": 2679, "loss": 0.8072, "learning_rate": 3.2325673891961394e-08, "epoch": 2.845314685314685, "percentage": 94.92, "elapsed_time": "1 day, 17:41:45", "remaining_time": "2:13:47", "throughput": "17764.37", "total_tokens": 2666528768}
{"current_steps": 2544, "total_steps": 2679, "loss": 0.6871, "learning_rate": 3.1853052408226395e-08, "epoch": 2.8464335664335665, "percentage": 94.96, "elapsed_time": "1 day, 17:42:44", "remaining_time": "2:12:48", "throughput": "17764.36", "total_tokens": 2667577344}
{"current_steps": 2545, "total_steps": 2679, "loss": 0.5945, "learning_rate": 3.138388927391017e-08, "epoch": 2.8475524475524474, "percentage": 95.0, "elapsed_time": "1 day, 17:43:33", "remaining_time": "2:11:49", "throughput": "17765.56", "total_tokens": 2668625920}
{"current_steps": 2546, "total_steps": 2679, "loss": 0.558, "learning_rate": 3.091818514639994e-08, "epoch": 2.8486713286713288, "percentage": 95.04, "elapsed_time": "1 day, 17:44:21", "remaining_time": "2:10:49", "throughput": "17766.84", "total_tokens": 2669674496}
{"current_steps": 2547, "total_steps": 2679, "loss": 0.5643, "learning_rate": 3.045594067823704e-08, "epoch": 2.8497902097902097, "percentage": 95.07, "elapsed_time": "1 day, 17:45:10", "remaining_time": "2:09:49", "throughput": "17768.09", "total_tokens": 2670723072}
{"current_steps": 2548, "total_steps": 2679, "loss": 0.7352, "learning_rate": 2.99971565171156e-08, "epoch": 2.850909090909091, "percentage": 95.11, "elapsed_time": "1 day, 17:46:09", "remaining_time": "2:08:50", "throughput": "17768.00", "total_tokens": 2671771648}
{"current_steps": 2549, "total_steps": 2679, "loss": 0.6695, "learning_rate": 2.9541833305880287e-08, "epoch": 2.852027972027972, "percentage": 95.15, "elapsed_time": "1 day, 17:47:05", "remaining_time": "2:07:51", "throughput": "17768.45", "total_tokens": 2672820224}
{"current_steps": 2550, "total_steps": 2679, "loss": 0.5442, "learning_rate": 2.9089971682526862e-08, "epoch": 2.8531468531468533, "percentage": 95.18, "elapsed_time": "1 day, 17:47:55", "remaining_time": "2:06:52", "throughput": "17769.45", "total_tokens": 2673868800}
{"current_steps": 2551, "total_steps": 2679, "loss": 0.692, "learning_rate": 2.864157228019998e-08, "epoch": 2.8542657342657343, "percentage": 95.22, "elapsed_time": "1 day, 17:48:55", "remaining_time": "2:05:53", "throughput": "17769.39", "total_tokens": 2674917376}
{"current_steps": 2552, "total_steps": 2679, "loss": 0.7277, "learning_rate": 2.8196635727194276e-08, "epoch": 2.855384615384615, "percentage": 95.26, "elapsed_time": "1 day, 17:49:54", "remaining_time": "2:04:54", "throughput": "17769.29", "total_tokens": 2675965952}
{"current_steps": 2553, "total_steps": 2679, "loss": 0.7013, "learning_rate": 2.7755162646950773e-08, "epoch": 2.8565034965034966, "percentage": 95.3, "elapsed_time": "1 day, 17:50:54", "remaining_time": "2:03:55", "throughput": "17769.18", "total_tokens": 2677014528}
{"current_steps": 2554, "total_steps": 2679, "loss": 0.592, "learning_rate": 2.731715365805937e-08, "epoch": 2.8576223776223775, "percentage": 95.33, "elapsed_time": "1 day, 17:51:43", "remaining_time": "2:02:55", "throughput": "17770.45", "total_tokens": 2678063104}
{"current_steps": 2555, "total_steps": 2679, "loss": 0.8021, "learning_rate": 2.688260937425413e-08, "epoch": 2.858741258741259, "percentage": 95.37, "elapsed_time": "1 day, 17:52:42", "remaining_time": "2:01:56", "throughput": "17770.36", "total_tokens": 2679111680}
{"current_steps": 2556, "total_steps": 2679, "loss": 0.6836, "learning_rate": 2.645153040441578e-08, "epoch": 2.8598601398601398, "percentage": 95.41, "elapsed_time": "1 day, 17:53:42", "remaining_time": "2:00:57", "throughput": "17770.27", "total_tokens": 2680160256}
{"current_steps": 2557, "total_steps": 2679, "loss": 0.618, "learning_rate": 2.6023917352568652e-08, "epoch": 2.860979020979021, "percentage": 95.45, "elapsed_time": "1 day, 17:54:37", "remaining_time": "1:59:58", "throughput": "17770.76", "total_tokens": 2681208832}
{"current_steps": 2558, "total_steps": 2679, "loss": 0.7565, "learning_rate": 2.5599770817881508e-08, "epoch": 2.862097902097902, "percentage": 95.48, "elapsed_time": "1 day, 17:55:43", "remaining_time": "1:59:00", "throughput": "17769.96", "total_tokens": 2682257408}
{"current_steps": 2559, "total_steps": 2679, "loss": 0.7696, "learning_rate": 2.5179091394665346e-08, "epoch": 2.8632167832167834, "percentage": 95.52, "elapsed_time": "1 day, 17:56:46", "remaining_time": "1:58:01", "throughput": "17769.47", "total_tokens": 2683305984}
{"current_steps": 2560, "total_steps": 2679, "loss": 0.6276, "learning_rate": 2.4761879672372535e-08, "epoch": 2.8643356643356643, "percentage": 95.56, "elapsed_time": "1 day, 17:57:40", "remaining_time": "1:57:01", "throughput": "17770.10", "total_tokens": 2684354560}
{"current_steps": 2561, "total_steps": 2679, "loss": 0.7476, "learning_rate": 2.4348136235597398e-08, "epoch": 2.8654545454545453, "percentage": 95.6, "elapsed_time": "1 day, 17:58:50", "remaining_time": "1:56:03", "throughput": "17768.71", "total_tokens": 2685403136}
{"current_steps": 2562, "total_steps": 2679, "loss": 0.5979, "learning_rate": 2.3937861664074523e-08, "epoch": 2.8665734265734266, "percentage": 95.63, "elapsed_time": "1 day, 17:59:44", "remaining_time": "1:55:04", "throughput": "17769.33", "total_tokens": 2686451712}
{"current_steps": 2563, "total_steps": 2679, "loss": 0.7925, "learning_rate": 2.3531056532677122e-08, "epoch": 2.8676923076923075, "percentage": 95.67, "elapsed_time": "1 day, 18:00:55", "remaining_time": "1:54:05", "throughput": "17767.91", "total_tokens": 2687500288}
{"current_steps": 2564, "total_steps": 2679, "loss": 0.6842, "learning_rate": 2.3127721411417836e-08, "epoch": 2.868811188811189, "percentage": 95.71, "elapsed_time": "1 day, 18:01:55", "remaining_time": "1:53:06", "throughput": "17767.82", "total_tokens": 2688548864}
{"current_steps": 2565, "total_steps": 2679, "loss": 0.8109, "learning_rate": 2.272785686544682e-08, "epoch": 2.86993006993007, "percentage": 95.74, "elapsed_time": "1 day, 18:02:56", "remaining_time": "1:52:07", "throughput": "17767.66", "total_tokens": 2689597440}
{"current_steps": 2566, "total_steps": 2679, "loss": 0.6439, "learning_rate": 2.233146345505144e-08, "epoch": 2.871048951048951, "percentage": 95.78, "elapsed_time": "1 day, 18:03:49", "remaining_time": "1:51:08", "throughput": "17768.26", "total_tokens": 2690646016}
{"current_steps": 2567, "total_steps": 2679, "loss": 0.7115, "learning_rate": 2.1938541735655183e-08, "epoch": 2.872167832167832, "percentage": 95.82, "elapsed_time": "1 day, 18:04:55", "remaining_time": "1:50:09", "throughput": "17767.54", "total_tokens": 2691694592}
{"current_steps": 2568, "total_steps": 2679, "loss": 0.7045, "learning_rate": 2.154909225781654e-08, "epoch": 2.8732867132867135, "percentage": 95.86, "elapsed_time": "1 day, 18:05:55", "remaining_time": "1:49:10", "throughput": "17767.39", "total_tokens": 2692743168}
{"current_steps": 2569, "total_steps": 2679, "loss": 0.601, "learning_rate": 2.1163115567230386e-08, "epoch": 2.8744055944055944, "percentage": 95.89, "elapsed_time": "1 day, 18:06:49", "remaining_time": "1:48:11", "throughput": "17767.99", "total_tokens": 2693791744}
{"current_steps": 2570, "total_steps": 2679, "loss": 0.5456, "learning_rate": 2.078061220472355e-08, "epoch": 2.8755244755244753, "percentage": 95.93, "elapsed_time": "1 day, 18:07:37", "remaining_time": "1:47:12", "throughput": "17769.23", "total_tokens": 2694840320}
{"current_steps": 2571, "total_steps": 2679, "loss": 0.733, "learning_rate": 2.0401582706257304e-08, "epoch": 2.8766433566433567, "percentage": 95.97, "elapsed_time": "1 day, 18:08:37", "remaining_time": "1:46:13", "throughput": "17769.11", "total_tokens": 2695888896}
{"current_steps": 2572, "total_steps": 2679, "loss": 0.7031, "learning_rate": 2.0026027602925158e-08, "epoch": 2.8777622377622376, "percentage": 96.01, "elapsed_time": "1 day, 18:09:31", "remaining_time": "1:45:13", "throughput": "17769.68", "total_tokens": 2696937472}
{"current_steps": 2573, "total_steps": 2679, "loss": 0.9375, "learning_rate": 1.9653947420951448e-08, "epoch": 2.878881118881119, "percentage": 96.04, "elapsed_time": "1 day, 18:10:38", "remaining_time": "1:44:15", "throughput": "17768.76", "total_tokens": 2697986048}
{"current_steps": 2574, "total_steps": 2679, "loss": 0.5872, "learning_rate": 1.928534268169302e-08, "epoch": 2.88, "percentage": 96.08, "elapsed_time": "1 day, 18:11:26", "remaining_time": "1:43:15", "throughput": "17770.08", "total_tokens": 2699034624}
{"current_steps": 2575, "total_steps": 2679, "loss": 0.719, "learning_rate": 1.892021390163562e-08, "epoch": 2.8811188811188813, "percentage": 96.12, "elapsed_time": "1 day, 18:12:31", "remaining_time": "1:42:17", "throughput": "17769.34", "total_tokens": 2700083200}
{"current_steps": 2576, "total_steps": 2679, "loss": 0.5889, "learning_rate": 1.8558561592395275e-08, "epoch": 2.882237762237762, "percentage": 96.16, "elapsed_time": "1 day, 18:13:20", "remaining_time": "1:41:17", "throughput": "17770.55", "total_tokens": 2701131776}
{"current_steps": 2577, "total_steps": 2679, "loss": 0.6301, "learning_rate": 1.8200386260716352e-08, "epoch": 2.8833566433566435, "percentage": 96.19, "elapsed_time": "1 day, 18:14:14", "remaining_time": "1:40:18", "throughput": "17771.07", "total_tokens": 2702180352}
{"current_steps": 2578, "total_steps": 2679, "loss": 0.6651, "learning_rate": 1.7845688408471563e-08, "epoch": 2.8844755244755245, "percentage": 96.23, "elapsed_time": "1 day, 18:15:14", "remaining_time": "1:39:19", "throughput": "17770.99", "total_tokens": 2703228928}
{"current_steps": 2579, "total_steps": 2679, "loss": 0.6397, "learning_rate": 1.74944685326614e-08, "epoch": 2.8855944055944054, "percentage": 96.27, "elapsed_time": "1 day, 18:16:13", "remaining_time": "1:38:20", "throughput": "17770.96", "total_tokens": 2704277504}
{"current_steps": 2580, "total_steps": 2679, "loss": 0.5979, "learning_rate": 1.7146727125411655e-08, "epoch": 2.8867132867132868, "percentage": 96.3, "elapsed_time": "1 day, 18:17:08", "remaining_time": "1:37:21", "throughput": "17771.54", "total_tokens": 2705326080}
{"current_steps": 2581, "total_steps": 2679, "loss": 0.7543, "learning_rate": 1.6802464673975893e-08, "epoch": 2.8878321678321677, "percentage": 96.34, "elapsed_time": "1 day, 18:18:13", "remaining_time": "1:36:22", "throughput": "17770.83", "total_tokens": 2706374656}
{"current_steps": 2582, "total_steps": 2679, "loss": 0.7046, "learning_rate": 1.6461681660731865e-08, "epoch": 2.888951048951049, "percentage": 96.38, "elapsed_time": "1 day, 18:19:12", "remaining_time": "1:35:23", "throughput": "17770.74", "total_tokens": 2707423232}
{"current_steps": 2583, "total_steps": 2679, "loss": 0.5703, "learning_rate": 1.6124378563182053e-08, "epoch": 2.89006993006993, "percentage": 96.42, "elapsed_time": "1 day, 18:20:00", "remaining_time": "1:34:24", "throughput": "17772.05", "total_tokens": 2708471808}
{"current_steps": 2584, "total_steps": 2679, "loss": 0.6367, "learning_rate": 1.5790555853953116e-08, "epoch": 2.8911888111888113, "percentage": 96.45, "elapsed_time": "1 day, 18:20:54", "remaining_time": "1:33:24", "throughput": "17772.63", "total_tokens": 2709520384}
{"current_steps": 2585, "total_steps": 2679, "loss": 0.6915, "learning_rate": 1.546021400079506e-08, "epoch": 2.8923076923076922, "percentage": 96.49, "elapsed_time": "1 day, 18:21:54", "remaining_time": "1:32:25", "throughput": "17772.54", "total_tokens": 2710568960}
{"current_steps": 2586, "total_steps": 2679, "loss": 0.6201, "learning_rate": 1.513335346658068e-08, "epoch": 2.8934265734265736, "percentage": 96.53, "elapsed_time": "1 day, 18:22:48", "remaining_time": "1:31:26", "throughput": "17773.08", "total_tokens": 2711617536}
{"current_steps": 2587, "total_steps": 2679, "loss": 0.6935, "learning_rate": 1.4809974709304176e-08, "epoch": 2.8945454545454545, "percentage": 96.57, "elapsed_time": "1 day, 18:23:50", "remaining_time": "1:30:27", "throughput": "17772.81", "total_tokens": 2712666112}
{"current_steps": 2588, "total_steps": 2679, "loss": 0.6633, "learning_rate": 1.4490078182081979e-08, "epoch": 2.8956643356643355, "percentage": 96.6, "elapsed_time": "1 day, 18:24:44", "remaining_time": "1:29:28", "throughput": "17773.37", "total_tokens": 2713714688}
{"current_steps": 2589, "total_steps": 2679, "loss": 0.5979, "learning_rate": 1.4173664333149983e-08, "epoch": 2.896783216783217, "percentage": 96.64, "elapsed_time": "1 day, 18:25:38", "remaining_time": "1:28:29", "throughput": "17773.97", "total_tokens": 2714763264}
{"current_steps": 2590, "total_steps": 2679, "loss": 0.7129, "learning_rate": 1.3860733605865761e-08, "epoch": 2.8979020979020977, "percentage": 96.68, "elapsed_time": "1 day, 18:26:32", "remaining_time": "1:27:30", "throughput": "17774.51", "total_tokens": 2715811840}
{"current_steps": 2591, "total_steps": 2679, "loss": 0.5724, "learning_rate": 1.3551286438705513e-08, "epoch": 2.899020979020979, "percentage": 96.72, "elapsed_time": "1 day, 18:27:21", "remaining_time": "1:26:31", "throughput": "17775.72", "total_tokens": 2716860416}
{"current_steps": 2592, "total_steps": 2679, "loss": 0.6127, "learning_rate": 1.324532326526351e-08, "epoch": 2.90013986013986, "percentage": 96.75, "elapsed_time": "1 day, 18:28:14", "remaining_time": "1:25:31", "throughput": "17776.32", "total_tokens": 2717908992}
{"current_steps": 2593, "total_steps": 2679, "loss": 0.553, "learning_rate": 1.2942844514254038e-08, "epoch": 2.9012587412587414, "percentage": 96.79, "elapsed_time": "1 day, 18:29:02", "remaining_time": "1:24:32", "throughput": "17777.60", "total_tokens": 2718957568}
{"current_steps": 2594, "total_steps": 2679, "loss": 0.8386, "learning_rate": 1.2643850609507512e-08, "epoch": 2.9023776223776223, "percentage": 96.83, "elapsed_time": "1 day, 18:30:08", "remaining_time": "1:23:33", "throughput": "17776.80", "total_tokens": 2720006144}
{"current_steps": 2595, "total_steps": 2679, "loss": 0.5611, "learning_rate": 1.2348341969972143e-08, "epoch": 2.9034965034965037, "percentage": 96.86, "elapsed_time": "1 day, 18:30:56", "remaining_time": "1:22:34", "throughput": "17778.08", "total_tokens": 2721054720}
{"current_steps": 2596, "total_steps": 2679, "loss": 0.71, "learning_rate": 1.2056319009712824e-08, "epoch": 2.9046153846153846, "percentage": 96.9, "elapsed_time": "1 day, 18:32:03", "remaining_time": "1:21:35", "throughput": "17777.22", "total_tokens": 2722103296}
{"current_steps": 2597, "total_steps": 2679, "loss": 0.5938, "learning_rate": 1.1767782137909467e-08, "epoch": 2.9057342657342655, "percentage": 96.94, "elapsed_time": "1 day, 18:32:51", "remaining_time": "1:20:36", "throughput": "17778.50", "total_tokens": 2723151872}
{"current_steps": 2598, "total_steps": 2679, "loss": 0.5475, "learning_rate": 1.148273175885839e-08, "epoch": 2.906853146853147, "percentage": 96.98, "elapsed_time": "1 day, 18:33:43", "remaining_time": "1:19:37", "throughput": "17779.26", "total_tokens": 2724200448}
{"current_steps": 2599, "total_steps": 2679, "loss": 0.6893, "learning_rate": 1.1201168271969266e-08, "epoch": 2.907972027972028, "percentage": 97.01, "elapsed_time": "1 day, 18:34:42", "remaining_time": "1:18:38", "throughput": "17779.22", "total_tokens": 2725249024}
{"current_steps": 2600, "total_steps": 2679, "loss": 0.7421, "learning_rate": 1.0923092071767615e-08, "epoch": 2.909090909090909, "percentage": 97.05, "elapsed_time": "1 day, 18:35:53", "remaining_time": "1:17:39", "throughput": "17777.88", "total_tokens": 2726297600}
{"current_steps": 2601, "total_steps": 2679, "loss": 0.6537, "learning_rate": 1.0648503547891487e-08, "epoch": 2.91020979020979, "percentage": 97.09, "elapsed_time": "1 day, 18:36:52", "remaining_time": "1:16:40", "throughput": "17777.87", "total_tokens": 2727346176}
{"current_steps": 2602, "total_steps": 2679, "loss": 0.5967, "learning_rate": 1.0377403085092275e-08, "epoch": 2.9113286713286715, "percentage": 97.13, "elapsed_time": "1 day, 18:37:46", "remaining_time": "1:15:41", "throughput": "17778.47", "total_tokens": 2728394752}
{"current_steps": 2603, "total_steps": 2679, "loss": 0.6689, "learning_rate": 1.0109791063233898e-08, "epoch": 2.9124475524475524, "percentage": 97.16, "elapsed_time": "1 day, 18:38:50", "remaining_time": "1:14:42", "throughput": "17777.86", "total_tokens": 2729443328}
{"current_steps": 2604, "total_steps": 2679, "loss": 0.5944, "learning_rate": 9.84566785729224e-09, "epoch": 2.9135664335664337, "percentage": 97.2, "elapsed_time": "1 day, 18:39:39", "remaining_time": "1:13:43", "throughput": "17779.07", "total_tokens": 2730491904}
{"current_steps": 2605, "total_steps": 2679, "loss": 0.7026, "learning_rate": 9.585033837355151e-09, "epoch": 2.9146853146853147, "percentage": 97.24, "elapsed_time": "1 day, 18:40:39", "remaining_time": "1:12:44", "throughput": "17778.92", "total_tokens": 2731540480}
{"current_steps": 2606, "total_steps": 2679, "loss": 0.6758, "learning_rate": 9.32788936862078e-09, "epoch": 2.9158041958041956, "percentage": 97.28, "elapsed_time": "1 day, 18:41:39", "remaining_time": "1:11:45", "throughput": "17778.81", "total_tokens": 2732589056}
{"current_steps": 2607, "total_steps": 2679, "loss": 0.7042, "learning_rate": 9.074234811398408e-09, "epoch": 2.916923076923077, "percentage": 97.31, "elapsed_time": "1 day, 18:42:38", "remaining_time": "1:10:46", "throughput": "17778.73", "total_tokens": 2733637632}
{"current_steps": 2608, "total_steps": 2679, "loss": 0.7203, "learning_rate": 8.824070521106787e-09, "epoch": 2.9180419580419583, "percentage": 97.35, "elapsed_time": "1 day, 18:43:45", "remaining_time": "1:09:47", "throughput": "17777.83", "total_tokens": 2734686208}
{"current_steps": 2609, "total_steps": 2679, "loss": 0.7092, "learning_rate": 8.577396848274134e-09, "epoch": 2.9191608391608392, "percentage": 97.39, "elapsed_time": "1 day, 18:44:50", "remaining_time": "1:08:48", "throughput": "17777.12", "total_tokens": 2735734784}
{"current_steps": 2610, "total_steps": 2679, "loss": 0.7815, "learning_rate": 8.334214138538132e-09, "epoch": 2.92027972027972, "percentage": 97.42, "elapsed_time": "1 day, 18:45:56", "remaining_time": "1:07:50", "throughput": "17776.34", "total_tokens": 2736783360}
{"current_steps": 2611, "total_steps": 2679, "loss": 0.6925, "learning_rate": 8.094522732644272e-09, "epoch": 2.9213986013986015, "percentage": 97.46, "elapsed_time": "1 day, 18:46:55", "remaining_time": "1:06:51", "throughput": "17776.31", "total_tokens": 2737831936}
{"current_steps": 2612, "total_steps": 2679, "loss": 0.7153, "learning_rate": 7.858322966446397e-09, "epoch": 2.9225174825174824, "percentage": 97.5, "elapsed_time": "1 day, 18:47:55", "remaining_time": "1:05:52", "throughput": "17776.22", "total_tokens": 2738880512}
{"current_steps": 2613, "total_steps": 2679, "loss": 0.751, "learning_rate": 7.625615170906153e-09, "epoch": 2.923636363636364, "percentage": 97.54, "elapsed_time": "1 day, 18:49:01", "remaining_time": "1:04:53", "throughput": "17775.45", "total_tokens": 2739929088}
{"current_steps": 2614, "total_steps": 2679, "loss": 0.6437, "learning_rate": 7.396399672092158e-09, "epoch": 2.9247552447552447, "percentage": 97.57, "elapsed_time": "1 day, 18:49:55", "remaining_time": "1:03:54", "throughput": "17776.02", "total_tokens": 2740977664}
{"current_steps": 2615, "total_steps": 2679, "loss": 0.6028, "learning_rate": 7.17067679117861e-09, "epoch": 2.9258741258741257, "percentage": 97.61, "elapsed_time": "1 day, 18:50:45", "remaining_time": "1:02:55", "throughput": "17777.01", "total_tokens": 2742026240}
{"current_steps": 2616, "total_steps": 2679, "loss": 0.6755, "learning_rate": 6.948446844447787e-09, "epoch": 2.926993006993007, "percentage": 97.65, "elapsed_time": "1 day, 18:51:43", "remaining_time": "1:01:56", "throughput": "17777.11", "total_tokens": 2743074816}
{"current_steps": 2617, "total_steps": 2679, "loss": 0.6773, "learning_rate": 6.729710143286161e-09, "epoch": 2.9281118881118884, "percentage": 97.69, "elapsed_time": "1 day, 18:52:43", "remaining_time": "1:00:57", "throughput": "17777.08", "total_tokens": 2744123392}
{"current_steps": 2618, "total_steps": 2679, "loss": 0.7518, "learning_rate": 6.514466994185786e-09, "epoch": 2.9292307692307693, "percentage": 97.72, "elapsed_time": "1 day, 18:53:48", "remaining_time": "0:59:58", "throughput": "17776.31", "total_tokens": 2745171968}
{"current_steps": 2619, "total_steps": 2679, "loss": 0.5816, "learning_rate": 6.302717698744298e-09, "epoch": 2.9303496503496502, "percentage": 97.76, "elapsed_time": "1 day, 18:54:42", "remaining_time": "0:58:59", "throughput": "17776.96", "total_tokens": 2746220544}
{"current_steps": 2620, "total_steps": 2679, "loss": 0.5671, "learning_rate": 6.094462553662972e-09, "epoch": 2.9314685314685316, "percentage": 97.8, "elapsed_time": "1 day, 18:55:30", "remaining_time": "0:57:59", "throughput": "17778.15", "total_tokens": 2747269120}
{"current_steps": 2621, "total_steps": 2679, "loss": 0.6945, "learning_rate": 5.889701850747276e-09, "epoch": 2.9325874125874125, "percentage": 97.84, "elapsed_time": "1 day, 18:56:30", "remaining_time": "0:57:00", "throughput": "17778.08", "total_tokens": 2748317696}
{"current_steps": 2622, "total_steps": 2679, "loss": 0.7404, "learning_rate": 5.688435876906873e-09, "epoch": 2.933706293706294, "percentage": 97.87, "elapsed_time": "1 day, 18:57:30", "remaining_time": "0:56:01", "throughput": "17777.98", "total_tokens": 2749366272}
{"current_steps": 2623, "total_steps": 2679, "loss": 0.6878, "learning_rate": 5.490664914153676e-09, "epoch": 2.934825174825175, "percentage": 97.91, "elapsed_time": "1 day, 18:58:29", "remaining_time": "0:55:02", "throughput": "17777.92", "total_tokens": 2750414848}
{"current_steps": 2624, "total_steps": 2679, "loss": 0.6308, "learning_rate": 5.296389239603239e-09, "epoch": 2.9359440559440557, "percentage": 97.95, "elapsed_time": "1 day, 18:59:23", "remaining_time": "0:54:03", "throughput": "17778.50", "total_tokens": 2751463424}
{"current_steps": 2625, "total_steps": 2679, "loss": 0.5787, "learning_rate": 5.10560912547281e-09, "epoch": 2.937062937062937, "percentage": 97.98, "elapsed_time": "1 day, 19:00:11", "remaining_time": "0:53:04", "throughput": "17779.72", "total_tokens": 2752512000}
{"current_steps": 2626, "total_steps": 2679, "loss": 0.6914, "learning_rate": 4.918324839082444e-09, "epoch": 2.9381818181818184, "percentage": 98.02, "elapsed_time": "1 day, 19:01:11", "remaining_time": "0:52:05", "throughput": "17779.70", "total_tokens": 2753560576}
{"current_steps": 2627, "total_steps": 2679, "loss": 0.7354, "learning_rate": 4.734536642853338e-09, "epoch": 2.9393006993006994, "percentage": 98.06, "elapsed_time": "1 day, 19:02:10", "remaining_time": "0:51:06", "throughput": "17779.63", "total_tokens": 2754609152}
{"current_steps": 2628, "total_steps": 2679, "loss": 0.8268, "learning_rate": 4.554244794308382e-09, "epoch": 2.9404195804195803, "percentage": 98.1, "elapsed_time": "1 day, 19:03:22", "remaining_time": "0:50:08", "throughput": "17778.18", "total_tokens": 2755657728}
{"current_steps": 2629, "total_steps": 2679, "loss": 0.6978, "learning_rate": 4.377449546071055e-09, "epoch": 2.9415384615384617, "percentage": 98.13, "elapsed_time": "1 day, 19:04:16", "remaining_time": "0:49:08", "throughput": "17778.76", "total_tokens": 2756706304}
{"current_steps": 2630, "total_steps": 2679, "loss": 0.682, "learning_rate": 4.204151145865421e-09, "epoch": 2.9426573426573426, "percentage": 98.17, "elapsed_time": "1 day, 19:05:10", "remaining_time": "0:48:09", "throughput": "17779.25", "total_tokens": 2757754880}
{"current_steps": 2631, "total_steps": 2679, "loss": 0.6442, "learning_rate": 4.034349836516127e-09, "epoch": 2.943776223776224, "percentage": 98.21, "elapsed_time": "1 day, 19:06:05", "remaining_time": "0:47:10", "throughput": "17779.78", "total_tokens": 2758803456}
{"current_steps": 2632, "total_steps": 2679, "loss": 0.588, "learning_rate": 3.8680458559475775e-09, "epoch": 2.944895104895105, "percentage": 98.25, "elapsed_time": "1 day, 19:06:59", "remaining_time": "0:46:11", "throughput": "17780.37", "total_tokens": 2759852032}
{"current_steps": 2633, "total_steps": 2679, "loss": 0.6812, "learning_rate": 3.705239437183372e-09, "epoch": 2.946013986013986, "percentage": 98.28, "elapsed_time": "1 day, 19:07:58", "remaining_time": "0:45:12", "throughput": "17780.27", "total_tokens": 2760900608}
{"current_steps": 2634, "total_steps": 2679, "loss": 0.6295, "learning_rate": 3.5459308083471422e-09, "epoch": 2.947132867132867, "percentage": 98.32, "elapsed_time": "1 day, 19:08:52", "remaining_time": "0:44:13", "throughput": "17780.87", "total_tokens": 2761949184}
{"current_steps": 2635, "total_steps": 2679, "loss": 0.5187, "learning_rate": 3.3901201926606063e-09, "epoch": 2.9482517482517485, "percentage": 98.36, "elapsed_time": "1 day, 19:09:40", "remaining_time": "0:43:14", "throughput": "17782.12", "total_tokens": 2762997760}
{"current_steps": 2636, "total_steps": 2679, "loss": 0.6132, "learning_rate": 3.237807808444404e-09, "epoch": 2.9493706293706294, "percentage": 98.39, "elapsed_time": "1 day, 19:10:34", "remaining_time": "0:42:15", "throughput": "17782.74", "total_tokens": 2764046336}
{"current_steps": 2637, "total_steps": 2679, "loss": 0.7231, "learning_rate": 3.088993869117818e-09, "epoch": 2.9504895104895104, "percentage": 98.43, "elapsed_time": "1 day, 19:11:39", "remaining_time": "0:41:16", "throughput": "17782.03", "total_tokens": 2765094912}
{"current_steps": 2638, "total_steps": 2679, "loss": 0.6015, "learning_rate": 2.943678583197662e-09, "epoch": 2.9516083916083917, "percentage": 98.47, "elapsed_time": "1 day, 19:12:33", "remaining_time": "0:40:17", "throughput": "17782.61", "total_tokens": 2766143488}
{"current_steps": 2639, "total_steps": 2679, "loss": 0.6697, "learning_rate": 2.8018621542988402e-09, "epoch": 2.9527272727272726, "percentage": 98.51, "elapsed_time": "1 day, 19:13:32", "remaining_time": "0:39:18", "throughput": "17782.55", "total_tokens": 2767192064}
{"current_steps": 2640, "total_steps": 2679, "loss": 0.6365, "learning_rate": 2.6635447811332315e-09, "epoch": 2.953846153846154, "percentage": 98.54, "elapsed_time": "1 day, 19:14:32", "remaining_time": "0:38:19", "throughput": "17782.44", "total_tokens": 2768240640}
{"current_steps": 2641, "total_steps": 2679, "loss": 0.7623, "learning_rate": 2.52872665751025e-09, "epoch": 2.954965034965035, "percentage": 98.58, "elapsed_time": "1 day, 19:15:38", "remaining_time": "0:37:20", "throughput": "17781.71", "total_tokens": 2769289216}
{"current_steps": 2642, "total_steps": 2679, "loss": 0.7513, "learning_rate": 2.397407972336008e-09, "epoch": 2.956083916083916, "percentage": 98.62, "elapsed_time": "1 day, 19:16:38", "remaining_time": "0:36:21", "throughput": "17781.55", "total_tokens": 2770337792}
{"current_steps": 2643, "total_steps": 2679, "loss": 0.6677, "learning_rate": 2.2695889096133184e-09, "epoch": 2.957202797202797, "percentage": 98.66, "elapsed_time": "1 day, 19:17:38", "remaining_time": "0:35:22", "throughput": "17781.44", "total_tokens": 2771386368}
{"current_steps": 2644, "total_steps": 2679, "loss": 0.6647, "learning_rate": 2.145269648441417e-09, "epoch": 2.9583216783216786, "percentage": 98.69, "elapsed_time": "1 day, 19:18:38", "remaining_time": "0:34:23", "throughput": "17781.25", "total_tokens": 2772434944}
{"current_steps": 2645, "total_steps": 2679, "loss": 0.7776, "learning_rate": 2.0244503630154066e-09, "epoch": 2.9594405594405595, "percentage": 98.73, "elapsed_time": "1 day, 19:19:38", "remaining_time": "0:33:25", "throughput": "17781.16", "total_tokens": 2773483520}
{"current_steps": 2646, "total_steps": 2679, "loss": 0.7085, "learning_rate": 1.907131222626535e-09, "epoch": 2.9605594405594404, "percentage": 98.77, "elapsed_time": "1 day, 19:20:39", "remaining_time": "0:32:26", "throughput": "17780.98", "total_tokens": 2774532096}
{"current_steps": 2647, "total_steps": 2679, "loss": 0.6998, "learning_rate": 1.7933123916613614e-09, "epoch": 2.961678321678322, "percentage": 98.81, "elapsed_time": "1 day, 19:21:46", "remaining_time": "0:31:27", "throughput": "17780.05", "total_tokens": 2775580672}
{"current_steps": 2648, "total_steps": 2679, "loss": 0.7513, "learning_rate": 1.6829940296023139e-09, "epoch": 2.9627972027972027, "percentage": 98.84, "elapsed_time": "1 day, 19:22:43", "remaining_time": "0:30:28", "throughput": "17780.27", "total_tokens": 2776629248}
{"current_steps": 2649, "total_steps": 2679, "loss": 0.6613, "learning_rate": 1.5761762910260214e-09, "epoch": 2.963916083916084, "percentage": 98.88, "elapsed_time": "1 day, 19:23:42", "remaining_time": "0:29:29", "throughput": "17780.23", "total_tokens": 2777677824}
{"current_steps": 2650, "total_steps": 2679, "loss": 0.6361, "learning_rate": 1.4728593256055357e-09, "epoch": 2.965034965034965, "percentage": 98.92, "elapsed_time": "1 day, 19:24:36", "remaining_time": "0:28:30", "throughput": "17780.85", "total_tokens": 2778726400}
{"current_steps": 2651, "total_steps": 2679, "loss": 0.8096, "learning_rate": 1.3730432781070002e-09, "epoch": 2.966153846153846, "percentage": 98.95, "elapsed_time": "1 day, 19:25:38", "remaining_time": "0:27:31", "throughput": "17780.51", "total_tokens": 2779774976}
{"current_steps": 2652, "total_steps": 2679, "loss": 0.6983, "learning_rate": 1.2767282883927035e-09, "epoch": 2.9672727272727273, "percentage": 98.99, "elapsed_time": "1 day, 19:26:38", "remaining_time": "0:26:32", "throughput": "17780.37", "total_tokens": 2780823552}
{"current_steps": 2653, "total_steps": 2679, "loss": 0.6842, "learning_rate": 1.1839144914180256e-09, "epoch": 2.9683916083916086, "percentage": 99.03, "elapsed_time": "1 day, 19:27:38", "remaining_time": "0:25:33", "throughput": "17780.30", "total_tokens": 2781872128}
{"current_steps": 2654, "total_steps": 2679, "loss": 0.7869, "learning_rate": 1.094602017233104e-09, "epoch": 2.9695104895104896, "percentage": 99.07, "elapsed_time": "1 day, 19:28:49", "remaining_time": "0:24:34", "throughput": "17778.90", "total_tokens": 2782920704}
{"current_steps": 2655, "total_steps": 2679, "loss": 0.7337, "learning_rate": 1.0087909909817228e-09, "epoch": 2.9706293706293705, "percentage": 99.1, "elapsed_time": "1 day, 19:30:00", "remaining_time": "0:23:35", "throughput": "17777.55", "total_tokens": 2783969280}
{"current_steps": 2656, "total_steps": 2679, "loss": 0.6917, "learning_rate": 9.264815329021459e-10, "epoch": 2.971748251748252, "percentage": 99.14, "elapsed_time": "1 day, 19:31:00", "remaining_time": "0:22:36", "throughput": "17777.46", "total_tokens": 2785017856}
{"current_steps": 2657, "total_steps": 2679, "loss": 0.7033, "learning_rate": 8.476737583251737e-10, "epoch": 2.9728671328671328, "percentage": 99.18, "elapsed_time": "1 day, 19:32:01", "remaining_time": "0:21:37", "throughput": "17777.16", "total_tokens": 2786066432}
{"current_steps": 2658, "total_steps": 2679, "loss": 0.7057, "learning_rate": 7.723677776763639e-10, "epoch": 2.973986013986014, "percentage": 99.22, "elapsed_time": "1 day, 19:32:57", "remaining_time": "0:20:38", "throughput": "17777.55", "total_tokens": 2787115008}
{"current_steps": 2659, "total_steps": 2679, "loss": 0.6595, "learning_rate": 7.005636964732554e-10, "epoch": 2.975104895104895, "percentage": 99.25, "elapsed_time": "1 day, 19:33:56", "remaining_time": "0:19:39", "throughput": "17777.51", "total_tokens": 2788163584}
{"current_steps": 2660, "total_steps": 2679, "loss": 0.6379, "learning_rate": 6.322616153275896e-10, "epoch": 2.976223776223776, "percentage": 99.29, "elapsed_time": "1 day, 19:34:53", "remaining_time": "0:18:40", "throughput": "17777.73", "total_tokens": 2789212160}
{"current_steps": 2661, "total_steps": 2679, "loss": 0.6597, "learning_rate": 5.674616299436441e-10, "epoch": 2.9773426573426574, "percentage": 99.33, "elapsed_time": "1 day, 19:35:48", "remaining_time": "0:17:41", "throughput": "17778.20", "total_tokens": 2790260736}
{"current_steps": 2662, "total_steps": 2679, "loss": 0.6397, "learning_rate": 5.061638311187889e-10, "epoch": 2.9784615384615387, "percentage": 99.37, "elapsed_time": "1 day, 19:36:43", "remaining_time": "0:16:42", "throughput": "17778.62", "total_tokens": 2791309312}
{"current_steps": 2663, "total_steps": 2679, "loss": 0.5986, "learning_rate": 4.4836830474265235e-10, "epoch": 2.9795804195804196, "percentage": 99.4, "elapsed_time": "1 day, 19:37:37", "remaining_time": "0:15:43", "throughput": "17779.24", "total_tokens": 2792357888}
{"current_steps": 2664, "total_steps": 2679, "loss": 0.6381, "learning_rate": 3.9407513179851034e-10, "epoch": 2.9806993006993006, "percentage": 99.44, "elapsed_time": "1 day, 19:38:36", "remaining_time": "0:14:44", "throughput": "17779.19", "total_tokens": 2793406464}
{"current_steps": 2665, "total_steps": 2679, "loss": 0.648, "learning_rate": 3.432843883610648e-10, "epoch": 2.981818181818182, "percentage": 99.48, "elapsed_time": "1 day, 19:39:32", "remaining_time": "0:13:45", "throughput": "17779.49", "total_tokens": 2794455040}
{"current_steps": 2666, "total_steps": 2679, "loss": 0.6502, "learning_rate": 2.9599614559810975e-10, "epoch": 2.982937062937063, "percentage": 99.51, "elapsed_time": "1 day, 19:40:33", "remaining_time": "0:12:46", "throughput": "17779.35", "total_tokens": 2795503616}
{"current_steps": 2667, "total_steps": 2679, "loss": 0.768, "learning_rate": 2.522104697696981e-10, "epoch": 2.984055944055944, "percentage": 99.55, "elapsed_time": "1 day, 19:41:38", "remaining_time": "0:11:47", "throughput": "17778.62", "total_tokens": 2796552192}
{"current_steps": 2668, "total_steps": 2679, "loss": 0.7205, "learning_rate": 2.1192742222786444e-10, "epoch": 2.985174825174825, "percentage": 99.59, "elapsed_time": "1 day, 19:42:32", "remaining_time": "0:10:48", "throughput": "17779.15", "total_tokens": 2797600768}
{"current_steps": 2669, "total_steps": 2679, "loss": 0.7303, "learning_rate": 1.7514705941690247e-10, "epoch": 2.986293706293706, "percentage": 99.63, "elapsed_time": "1 day, 19:43:36", "remaining_time": "0:09:49", "throughput": "17778.64", "total_tokens": 2798649344}
{"current_steps": 2670, "total_steps": 2679, "loss": 0.6629, "learning_rate": 1.418694328730874e-10, "epoch": 2.9874125874125874, "percentage": 99.66, "elapsed_time": "1 day, 19:44:30", "remaining_time": "0:08:50", "throughput": "17779.22", "total_tokens": 2799697920}
{"current_steps": 2671, "total_steps": 2679, "loss": 0.5753, "learning_rate": 1.1209458922495365e-10, "epoch": 2.988531468531469, "percentage": 99.7, "elapsed_time": "1 day, 19:45:18", "remaining_time": "0:07:51", "throughput": "17780.47", "total_tokens": 2800746496}
{"current_steps": 2672, "total_steps": 2679, "loss": 0.6927, "learning_rate": 8.58225701930171e-11, "epoch": 2.9896503496503497, "percentage": 99.74, "elapsed_time": "1 day, 19:46:12", "remaining_time": "0:06:52", "throughput": "17781.03", "total_tokens": 2801795072}
{"current_steps": 2673, "total_steps": 2679, "loss": 0.5823, "learning_rate": 6.30534125889426e-11, "epoch": 2.9907692307692306, "percentage": 99.78, "elapsed_time": "1 day, 19:47:00", "remaining_time": "0:05:53", "throughput": "17782.19", "total_tokens": 2802843648}
{"current_steps": 2674, "total_steps": 2679, "loss": 0.6706, "learning_rate": 4.3787148317209205e-11, "epoch": 2.991888111888112, "percentage": 99.81, "elapsed_time": "1 day, 19:48:00", "remaining_time": "0:04:54", "throughput": "17782.10", "total_tokens": 2803892224}
{"current_steps": 2675, "total_steps": 2679, "loss": 0.7391, "learning_rate": 2.8023804372889762e-11, "epoch": 2.993006993006993, "percentage": 99.85, "elapsed_time": "1 day, 19:49:06", "remaining_time": "0:03:55", "throughput": "17781.37", "total_tokens": 2804940800}
{"current_steps": 2676, "total_steps": 2679, "loss": 0.7509, "learning_rate": 1.57634028441489e-11, "epoch": 2.9941258741258743, "percentage": 99.89, "elapsed_time": "1 day, 19:50:11", "remaining_time": "0:02:56", "throughput": "17780.69", "total_tokens": 2805989376}
{"current_steps": 2677, "total_steps": 2679, "loss": 0.634, "learning_rate": 7.005960910022591e-12, "epoch": 2.995244755244755, "percentage": 99.93, "elapsed_time": "1 day, 19:51:04", "remaining_time": "0:01:57", "throughput": "17781.29", "total_tokens": 2807037952}
{"current_steps": 2678, "total_steps": 2679, "loss": 0.6167, "learning_rate": 1.7514908409732578e-12, "epoch": 2.996363636363636, "percentage": 99.96, "elapsed_time": "1 day, 19:51:56", "remaining_time": "0:00:58", "throughput": "17782.04", "total_tokens": 2808086528}
{"current_steps": 2679, "total_steps": 2679, "loss": 0.5345, "learning_rate": 0.0, "epoch": 2.9974825174825175, "percentage": 100.0, "elapsed_time": "1 day, 19:52:44", "remaining_time": "0:00:00", "throughput": "17783.30", "total_tokens": 2809135104}
{"current_steps": 2679, "total_steps": 2679, "epoch": 2.9974825174825175, "percentage": 100.0, "elapsed_time": "1 day, 19:53:13", "remaining_time": "0:00:00", "throughput": "17780.12", "total_tokens": 2809135104}