| {"current_steps": 5, "total_steps": 510, "loss": 0.8996, "learning_rate": 4.998814299283415e-05, "epoch": 0.0963855421686747, "percentage": 0.98, "elapsed_time": "0:03:10", "remaining_time": "5:21:00", "throughput": "411.79", "total_tokens": 78528} | |
| {"current_steps": 10, "total_steps": 510, "loss": 0.6806, "learning_rate": 4.995258321842611e-05, "epoch": 0.1927710843373494, "percentage": 1.96, "elapsed_time": "0:06:29", "remaining_time": "5:24:19", "throughput": "408.84", "total_tokens": 159120} | |
| {"current_steps": 15, "total_steps": 510, "loss": 0.618, "learning_rate": 4.989335440737586e-05, "epoch": 0.2891566265060241, "percentage": 2.94, "elapsed_time": "0:09:12", "remaining_time": "5:04:08", "throughput": "404.26", "total_tokens": 223552} | |
| {"current_steps": 20, "total_steps": 510, "loss": 0.5594, "learning_rate": 4.98105127417984e-05, "epoch": 0.3855421686746988, "percentage": 3.92, "elapsed_time": "0:12:01", "remaining_time": "4:54:35", "throughput": "403.27", "total_tokens": 290944} | |
| {"current_steps": 25, "total_steps": 510, "loss": 0.5253, "learning_rate": 4.9704136802031485e-05, "epoch": 0.4819277108433735, "percentage": 4.9, "elapsed_time": "0:15:17", "remaining_time": "4:56:36", "throughput": "396.87", "total_tokens": 364064} | |
| {"current_steps": 30, "total_steps": 510, "loss": 0.4794, "learning_rate": 4.957432749209755e-05, "epoch": 0.5783132530120482, "percentage": 5.88, "elapsed_time": "0:18:41", "remaining_time": "4:59:09", "throughput": "392.36", "total_tokens": 440176} | |
| {"current_steps": 35, "total_steps": 510, "loss": 0.4546, "learning_rate": 4.942120794399002e-05, "epoch": 0.6746987951807228, "percentage": 6.86, "elapsed_time": "0:21:52", "remaining_time": "4:56:56", "throughput": "393.95", "total_tokens": 517184} | |
| {"current_steps": 40, "total_steps": 510, "loss": 0.4703, "learning_rate": 4.9244923400875245e-05, "epoch": 0.7710843373493976, "percentage": 7.84, "elapsed_time": "0:24:56", "remaining_time": "4:53:00", "throughput": "395.28", "total_tokens": 591424} | |
| {"current_steps": 45, "total_steps": 510, "loss": 0.4407, "learning_rate": 4.9045641079320484e-05, "epoch": 0.8674698795180723, "percentage": 8.82, "elapsed_time": "0:27:56", "remaining_time": "4:48:45", "throughput": "395.30", "total_tokens": 662784} | |
| {"current_steps": 50, "total_steps": 510, "loss": 0.4425, "learning_rate": 4.882355001067892e-05, "epoch": 0.963855421686747, "percentage": 9.8, "elapsed_time": "0:30:58", "remaining_time": "4:44:58", "throughput": "395.36", "total_tokens": 734784} | |
| {"current_steps": 55, "total_steps": 510, "loss": 0.4081, "learning_rate": 4.857886086178194e-05, "epoch": 1.0602409638554218, "percentage": 10.78, "elapsed_time": "0:34:02", "remaining_time": "4:41:40", "throughput": "395.68", "total_tokens": 808336} | |
| {"current_steps": 60, "total_steps": 510, "loss": 0.4002, "learning_rate": 4.8311805735108894e-05, "epoch": 1.1566265060240963, "percentage": 11.76, "elapsed_time": "0:37:11", "remaining_time": "4:38:53", "throughput": "395.63", "total_tokens": 882672} | |
| {"current_steps": 65, "total_steps": 510, "loss": 0.3587, "learning_rate": 4.802263794862385e-05, "epoch": 1.2530120481927711, "percentage": 12.75, "elapsed_time": "0:39:57", "remaining_time": "4:33:30", "throughput": "395.35", "total_tokens": 947680} | |
| {"current_steps": 70, "total_steps": 510, "loss": 0.356, "learning_rate": 4.7711631795488096e-05, "epoch": 1.3493975903614457, "percentage": 13.73, "elapsed_time": "0:43:01", "remaining_time": "4:30:24", "throughput": "395.98", "total_tokens": 1022112} | |
| {"current_steps": 75, "total_steps": 510, "loss": 0.3639, "learning_rate": 4.7379082283876566e-05, "epoch": 1.4457831325301205, "percentage": 14.71, "elapsed_time": "0:45:55", "remaining_time": "4:26:23", "throughput": "396.16", "total_tokens": 1091744} | |
| {"current_steps": 80, "total_steps": 510, "loss": 0.3288, "learning_rate": 4.702530485714461e-05, "epoch": 1.5421686746987953, "percentage": 15.69, "elapsed_time": "0:48:59", "remaining_time": "4:23:19", "throughput": "395.91", "total_tokens": 1163728} | |
| {"current_steps": 85, "total_steps": 510, "loss": 0.3563, "learning_rate": 4.665063509461097e-05, "epoch": 1.6385542168674698, "percentage": 16.67, "elapsed_time": "0:52:24", "remaining_time": "4:22:03", "throughput": "396.14", "total_tokens": 1245728} | |
| {"current_steps": 90, "total_steps": 510, "loss": 0.3642, "learning_rate": 4.625542839324036e-05, "epoch": 1.7349397590361446, "percentage": 17.65, "elapsed_time": "0:55:20", "remaining_time": "4:18:15", "throughput": "396.03", "total_tokens": 1315056} | |
| {"current_steps": 95, "total_steps": 510, "loss": 0.3407, "learning_rate": 4.584005963052799e-05, "epoch": 1.8313253012048194, "percentage": 18.63, "elapsed_time": "0:58:31", "remaining_time": "4:15:39", "throughput": "396.47", "total_tokens": 1392224} | |
| {"current_steps": 100, "total_steps": 510, "loss": 0.3216, "learning_rate": 4.540492280890555e-05, "epoch": 1.927710843373494, "percentage": 19.61, "elapsed_time": "1:01:49", "remaining_time": "4:13:28", "throughput": "396.56", "total_tokens": 1471008} | |
| {"current_steps": 105, "total_steps": 510, "loss": 0.3197, "learning_rate": 4.4950430682006e-05, "epoch": 2.0240963855421685, "percentage": 20.59, "elapsed_time": "1:05:12", "remaining_time": "4:11:30", "throughput": "395.40", "total_tokens": 1546912} | |
| {"current_steps": 110, "total_steps": 510, "loss": 0.2904, "learning_rate": 4.447701436314176e-05, "epoch": 2.1204819277108435, "percentage": 21.57, "elapsed_time": "1:08:01", "remaining_time": "4:07:21", "throughput": "394.79", "total_tokens": 1611328} | |
| {"current_steps": 115, "total_steps": 510, "loss": 0.2498, "learning_rate": 4.398512291636768e-05, "epoch": 2.216867469879518, "percentage": 22.55, "elapsed_time": "1:11:06", "remaining_time": "4:04:13", "throughput": "394.39", "total_tokens": 1682528} | |
| {"current_steps": 120, "total_steps": 510, "loss": 0.269, "learning_rate": 4.347522293051648e-05, "epoch": 2.3132530120481927, "percentage": 23.53, "elapsed_time": "1:14:06", "remaining_time": "4:00:51", "throughput": "393.99", "total_tokens": 1751856} | |
| {"current_steps": 125, "total_steps": 510, "loss": 0.2838, "learning_rate": 4.294779807661105e-05, "epoch": 2.4096385542168672, "percentage": 24.51, "elapsed_time": "1:17:27", "remaining_time": "3:58:35", "throughput": "393.80", "total_tokens": 1830288} | |
| {"current_steps": 130, "total_steps": 510, "loss": 0.2466, "learning_rate": 4.2403348649073174e-05, "epoch": 2.5060240963855422, "percentage": 25.49, "elapsed_time": "1:20:39", "remaining_time": "3:55:47", "throughput": "393.67", "total_tokens": 1905296} | |
| {"current_steps": 135, "total_steps": 510, "loss": 0.2272, "learning_rate": 4.184239109116393e-05, "epoch": 2.602409638554217, "percentage": 26.47, "elapsed_time": "1:23:42", "remaining_time": "3:52:30", "throughput": "393.15", "total_tokens": 1974464} | |
| {"current_steps": 140, "total_steps": 510, "loss": 0.2484, "learning_rate": 4.126545750510605e-05, "epoch": 2.6987951807228914, "percentage": 27.45, "elapsed_time": "1:27:15", "remaining_time": "3:50:37", "throughput": "393.10", "total_tokens": 2058176} | |
| {"current_steps": 145, "total_steps": 510, "loss": 0.2339, "learning_rate": 4.067309514735267e-05, "epoch": 2.7951807228915664, "percentage": 28.43, "elapsed_time": "1:30:11", "remaining_time": "3:47:03", "throughput": "392.63", "total_tokens": 2124912} | |
| {"current_steps": 150, "total_steps": 510, "loss": 0.2597, "learning_rate": 4.0065865909481417e-05, "epoch": 2.891566265060241, "percentage": 29.41, "elapsed_time": "1:33:54", "remaining_time": "3:45:22", "throughput": "392.84", "total_tokens": 2213456} | |
| {"current_steps": 155, "total_steps": 510, "loss": 0.2525, "learning_rate": 3.9444345785206285e-05, "epoch": 2.9879518072289155, "percentage": 30.39, "elapsed_time": "1:36:49", "remaining_time": "3:41:46", "throughput": "392.73", "total_tokens": 2281680} | |
| {"current_steps": 160, "total_steps": 510, "loss": 0.1832, "learning_rate": 3.880912432401265e-05, "epoch": 3.0843373493975905, "percentage": 31.37, "elapsed_time": "1:39:47", "remaining_time": "3:38:16", "throughput": "392.41", "total_tokens": 2349408} | |
| {"current_steps": 165, "total_steps": 510, "loss": 0.1519, "learning_rate": 3.81608040719339e-05, "epoch": 3.180722891566265, "percentage": 32.35, "elapsed_time": "1:42:57", "remaining_time": "3:35:17", "throughput": "392.60", "total_tokens": 2425456} | |
| {"current_steps": 170, "total_steps": 510, "loss": 0.1707, "learning_rate": 3.7500000000000003e-05, "epoch": 3.2771084337349397, "percentage": 33.33, "elapsed_time": "1:45:53", "remaining_time": "3:31:46", "throughput": "392.58", "total_tokens": 2494064} | |
| {"current_steps": 175, "total_steps": 510, "loss": 0.1603, "learning_rate": 3.6827338920900254e-05, "epoch": 3.3734939759036147, "percentage": 34.31, "elapsed_time": "1:49:12", "remaining_time": "3:29:03", "throughput": "392.77", "total_tokens": 2573616} | |
| {"current_steps": 180, "total_steps": 510, "loss": 0.1683, "learning_rate": 3.6143458894413465e-05, "epoch": 3.4698795180722892, "percentage": 35.29, "elapsed_time": "1:52:39", "remaining_time": "3:26:32", "throughput": "393.19", "total_tokens": 2657744} | |
| {"current_steps": 185, "total_steps": 510, "loss": 0.1734, "learning_rate": 3.544900862216959e-05, "epoch": 3.566265060240964, "percentage": 36.27, "elapsed_time": "1:55:27", "remaining_time": "3:22:50", "throughput": "392.80", "total_tokens": 2721200} | |
| {"current_steps": 190, "total_steps": 510, "loss": 0.1543, "learning_rate": 3.474464683231698e-05, "epoch": 3.662650602409639, "percentage": 37.25, "elapsed_time": "1:58:49", "remaining_time": "3:20:07", "throughput": "392.50", "total_tokens": 2798320} | |
| {"current_steps": 195, "total_steps": 510, "loss": 0.1601, "learning_rate": 3.403104165467883e-05, "epoch": 3.7590361445783134, "percentage": 38.24, "elapsed_time": "2:02:07", "remaining_time": "3:17:16", "throughput": "392.93", "total_tokens": 2879200} | |
| {"current_steps": 200, "total_steps": 510, "loss": 0.1911, "learning_rate": 3.330886998699149e-05, "epoch": 3.855421686746988, "percentage": 39.22, "elapsed_time": "2:04:58", "remaining_time": "3:13:42", "throughput": "393.04", "total_tokens": 2947024} | |
| {"current_steps": 205, "total_steps": 510, "loss": 0.1741, "learning_rate": 3.257881685282609e-05, "epoch": 3.9518072289156625, "percentage": 40.2, "elapsed_time": "2:07:52", "remaining_time": "3:10:14", "throughput": "393.20", "total_tokens": 3016656} | |
| {"current_steps": 210, "total_steps": 510, "loss": 0.1334, "learning_rate": 3.1841574751802076e-05, "epoch": 4.048192771084337, "percentage": 41.18, "elapsed_time": "2:10:41", "remaining_time": "3:06:41", "throughput": "393.35", "total_tokens": 3084416} | |
| {"current_steps": 215, "total_steps": 510, "loss": 0.1027, "learning_rate": 3.109784300270943e-05, "epoch": 4.144578313253012, "percentage": 42.16, "elapsed_time": "2:14:00", "remaining_time": "3:03:52", "throughput": "393.86", "total_tokens": 3166784} | |
| {"current_steps": 220, "total_steps": 510, "loss": 0.0955, "learning_rate": 3.0348327080162435e-05, "epoch": 4.240963855421687, "percentage": 43.14, "elapsed_time": "2:17:00", "remaining_time": "3:00:36", "throughput": "394.08", "total_tokens": 3239584} | |
| {"current_steps": 225, "total_steps": 510, "loss": 0.1006, "learning_rate": 2.9593737945414264e-05, "epoch": 4.337349397590361, "percentage": 44.12, "elapsed_time": "2:20:02", "remaining_time": "2:57:23", "throughput": "394.31", "total_tokens": 3313360} | |
| {"current_steps": 230, "total_steps": 510, "loss": 0.1007, "learning_rate": 2.8834791371967142e-05, "epoch": 4.433734939759036, "percentage": 45.1, "elapsed_time": "2:22:43", "remaining_time": "2:53:45", "throughput": "394.43", "total_tokens": 3377840} | |
| {"current_steps": 235, "total_steps": 510, "loss": 0.1033, "learning_rate": 2.8072207266617855e-05, "epoch": 4.530120481927711, "percentage": 46.08, "elapsed_time": "2:25:54", "remaining_time": "2:50:44", "throughput": "394.76", "total_tokens": 3455904} | |
| {"current_steps": 240, "total_steps": 510, "loss": 0.1023, "learning_rate": 2.7306708986582553e-05, "epoch": 4.626506024096385, "percentage": 47.06, "elapsed_time": "2:28:54", "remaining_time": "2:47:31", "throughput": "395.01", "total_tokens": 3529360} | |
| {"current_steps": 245, "total_steps": 510, "loss": 0.1121, "learning_rate": 2.653902265334858e-05, "epoch": 4.72289156626506, "percentage": 48.04, "elapsed_time": "2:32:00", "remaining_time": "2:44:25", "throughput": "395.28", "total_tokens": 3605344} | |
| {"current_steps": 250, "total_steps": 510, "loss": 0.1028, "learning_rate": 2.5769876463904265e-05, "epoch": 4.8192771084337345, "percentage": 49.02, "elapsed_time": "2:35:01", "remaining_time": "2:41:13", "throughput": "395.48", "total_tokens": 3678352} | |
| {"current_steps": 255, "total_steps": 510, "loss": 0.1055, "learning_rate": 2.5e-05, "epoch": 4.9156626506024095, "percentage": 50.0, "elapsed_time": "2:38:03", "remaining_time": "2:38:03", "throughput": "395.69", "total_tokens": 3752608} | |
| {"current_steps": 260, "total_steps": 510, "loss": 0.1037, "learning_rate": 2.4230123536095748e-05, "epoch": 5.0120481927710845, "percentage": 50.98, "elapsed_time": "2:40:50", "remaining_time": "2:34:39", "throughput": "395.79", "total_tokens": 3819744} | |
| {"current_steps": 265, "total_steps": 510, "loss": 0.0633, "learning_rate": 2.346097734665143e-05, "epoch": 5.108433734939759, "percentage": 51.96, "elapsed_time": "2:43:58", "remaining_time": "2:31:36", "throughput": "396.05", "total_tokens": 3896592} | |
| {"current_steps": 270, "total_steps": 510, "loss": 0.0521, "learning_rate": 2.2693291013417453e-05, "epoch": 5.204819277108434, "percentage": 52.94, "elapsed_time": "2:47:01", "remaining_time": "2:28:28", "throughput": "396.24", "total_tokens": 3970976} | |
| {"current_steps": 275, "total_steps": 510, "loss": 0.0625, "learning_rate": 2.192779273338215e-05, "epoch": 5.301204819277109, "percentage": 53.92, "elapsed_time": "2:50:16", "remaining_time": "2:25:30", "throughput": "396.59", "total_tokens": 4051760} | |
| {"current_steps": 280, "total_steps": 510, "loss": 0.059, "learning_rate": 2.116520862803286e-05, "epoch": 5.397590361445783, "percentage": 54.9, "elapsed_time": "2:53:14", "remaining_time": "2:22:18", "throughput": "396.74", "total_tokens": 4124096} | |
| {"current_steps": 285, "total_steps": 510, "loss": 0.0648, "learning_rate": 2.0406262054585738e-05, "epoch": 5.493975903614458, "percentage": 55.88, "elapsed_time": "2:55:56", "remaining_time": "2:18:54", "throughput": "396.77", "total_tokens": 4188448} | |
| {"current_steps": 290, "total_steps": 510, "loss": 0.0709, "learning_rate": 1.965167291983757e-05, "epoch": 5.590361445783133, "percentage": 56.86, "elapsed_time": "2:58:55", "remaining_time": "2:15:43", "throughput": "396.92", "total_tokens": 4261056} | |
| {"current_steps": 295, "total_steps": 510, "loss": 0.0641, "learning_rate": 1.890215699729057e-05, "epoch": 5.686746987951807, "percentage": 57.84, "elapsed_time": "3:01:45", "remaining_time": "2:12:27", "throughput": "396.97", "total_tokens": 4329024} | |
| {"current_steps": 300, "total_steps": 510, "loss": 0.0689, "learning_rate": 1.815842524819793e-05, "epoch": 5.783132530120482, "percentage": 58.82, "elapsed_time": "3:04:54", "remaining_time": "2:09:26", "throughput": "397.19", "total_tokens": 4406624} | |
| {"current_steps": 305, "total_steps": 510, "loss": 0.055, "learning_rate": 1.7421183147173915e-05, "epoch": 5.879518072289157, "percentage": 59.8, "elapsed_time": "3:07:55", "remaining_time": "2:06:18", "throughput": "397.36", "total_tokens": 4480352} | |
| {"current_steps": 310, "total_steps": 510, "loss": 0.0626, "learning_rate": 1.6691130013008514e-05, "epoch": 5.975903614457831, "percentage": 60.78, "elapsed_time": "3:10:55", "remaining_time": "2:03:10", "throughput": "397.56", "total_tokens": 4554080} | |
| {"current_steps": 315, "total_steps": 510, "loss": 0.0465, "learning_rate": 1.5968958345321178e-05, "epoch": 6.072289156626506, "percentage": 61.76, "elapsed_time": "3:13:56", "remaining_time": "2:00:03", "throughput": "397.77", "total_tokens": 4628576} | |
| {"current_steps": 320, "total_steps": 510, "loss": 0.0421, "learning_rate": 1.5255353167683017e-05, "epoch": 6.168674698795181, "percentage": 62.75, "elapsed_time": "3:17:00", "remaining_time": "1:56:58", "throughput": "397.98", "total_tokens": 4704512} | |
| {"current_steps": 325, "total_steps": 510, "loss": 0.0303, "learning_rate": 1.4550991377830426e-05, "epoch": 6.265060240963855, "percentage": 63.73, "elapsed_time": "3:19:56", "remaining_time": "1:53:49", "throughput": "398.18", "total_tokens": 4776912} | |
| {"current_steps": 330, "total_steps": 510, "loss": 0.0337, "learning_rate": 1.3856541105586545e-05, "epoch": 6.36144578313253, "percentage": 64.71, "elapsed_time": "3:23:06", "remaining_time": "1:50:47", "throughput": "398.43", "total_tokens": 4855600} | |
| {"current_steps": 335, "total_steps": 510, "loss": 0.0333, "learning_rate": 1.3172661079099752e-05, "epoch": 6.457831325301205, "percentage": 65.69, "elapsed_time": "3:26:04", "remaining_time": "1:47:38", "throughput": "398.54", "total_tokens": 4927600} | |
| {"current_steps": 340, "total_steps": 510, "loss": 0.039, "learning_rate": 1.2500000000000006e-05, "epoch": 6.554216867469879, "percentage": 66.67, "elapsed_time": "3:29:07", "remaining_time": "1:44:33", "throughput": "398.74", "total_tokens": 5003184} | |
| {"current_steps": 343, "total_steps": 510, "epoch": 6.612048192771084, "percentage": 67.25, "elapsed_time": "3:31:15", "remaining_time": "1:42:51", "throughput": "398.58", "total_tokens": 5052448} | |