mono-gemma-2-27b / trainer_log.jsonl
w3en2g's picture
Upload folder using huggingface_hub
54efee6 verified
{"current_steps": 10, "total_steps": 1773, "loss": 1.0771, "lr": 2.52808988764045e-07, "epoch": 0.01692047377326565, "percentage": 0.56, "elapsed_time": "0:01:40", "remaining_time": "4:56:37"}
{"current_steps": 20, "total_steps": 1773, "loss": 1.0561, "lr": 5.337078651685394e-07, "epoch": 0.0338409475465313, "percentage": 1.13, "elapsed_time": "0:02:38", "remaining_time": "3:51:28"}
{"current_steps": 30, "total_steps": 1773, "loss": 0.9336, "lr": 8.146067415730338e-07, "epoch": 0.050761421319796954, "percentage": 1.69, "elapsed_time": "0:03:41", "remaining_time": "3:34:14"}
{"current_steps": 40, "total_steps": 1773, "loss": 0.8594, "lr": 1.0955056179775282e-06, "epoch": 0.0676818950930626, "percentage": 2.26, "elapsed_time": "0:04:34", "remaining_time": "3:18:03"}
{"current_steps": 50, "total_steps": 1773, "loss": 0.8374, "lr": 1.3764044943820225e-06, "epoch": 0.08460236886632826, "percentage": 2.82, "elapsed_time": "0:05:36", "remaining_time": "3:13:01"}
{"current_steps": 60, "total_steps": 1773, "loss": 0.818, "lr": 1.6573033707865168e-06, "epoch": 0.10152284263959391, "percentage": 3.38, "elapsed_time": "0:06:30", "remaining_time": "3:05:51"}
{"current_steps": 70, "total_steps": 1773, "loss": 0.8089, "lr": 1.9382022471910116e-06, "epoch": 0.11844331641285956, "percentage": 3.95, "elapsed_time": "0:08:29", "remaining_time": "3:26:29"}
{"current_steps": 80, "total_steps": 1773, "loss": 0.7992, "lr": 2.2191011235955057e-06, "epoch": 0.1353637901861252, "percentage": 4.51, "elapsed_time": "0:09:28", "remaining_time": "3:20:34"}
{"current_steps": 90, "total_steps": 1773, "loss": 0.7848, "lr": 2.5e-06, "epoch": 0.15228426395939088, "percentage": 5.08, "elapsed_time": "0:10:26", "remaining_time": "3:15:11"}
{"current_steps": 100, "total_steps": 1773, "loss": 0.7785, "lr": 2.7808988764044947e-06, "epoch": 0.1692047377326565, "percentage": 5.64, "elapsed_time": "0:11:28", "remaining_time": "3:12:02"}
{"current_steps": 110, "total_steps": 1773, "loss": 0.7629, "lr": 3.0617977528089893e-06, "epoch": 0.18612521150592218, "percentage": 6.2, "elapsed_time": "0:12:30", "remaining_time": "3:09:00"}
{"current_steps": 120, "total_steps": 1773, "loss": 0.7764, "lr": 3.3426966292134834e-06, "epoch": 0.20304568527918782, "percentage": 6.77, "elapsed_time": "0:13:32", "remaining_time": "3:06:31"}
{"current_steps": 130, "total_steps": 1773, "loss": 0.7961, "lr": 3.623595505617978e-06, "epoch": 0.21996615905245348, "percentage": 7.33, "elapsed_time": "0:14:30", "remaining_time": "3:03:25"}
{"current_steps": 140, "total_steps": 1773, "loss": 0.7913, "lr": 3.9044943820224724e-06, "epoch": 0.23688663282571912, "percentage": 7.9, "elapsed_time": "0:15:25", "remaining_time": "2:59:59"}
{"current_steps": 150, "total_steps": 1773, "loss": 0.7715, "lr": 4.185393258426967e-06, "epoch": 0.25380710659898476, "percentage": 8.46, "elapsed_time": "0:16:54", "remaining_time": "3:02:54"}
{"current_steps": 160, "total_steps": 1773, "loss": 0.803, "lr": 4.466292134831461e-06, "epoch": 0.2707275803722504, "percentage": 9.02, "elapsed_time": "0:17:57", "remaining_time": "3:00:59"}
{"current_steps": 170, "total_steps": 1773, "loss": 0.7765, "lr": 4.747191011235955e-06, "epoch": 0.2876480541455161, "percentage": 9.59, "elapsed_time": "0:19:28", "remaining_time": "3:03:42"}
{"current_steps": 180, "total_steps": 1773, "loss": 0.7903, "lr": 4.999995150597375e-06, "epoch": 0.30456852791878175, "percentage": 10.15, "elapsed_time": "0:20:55", "remaining_time": "3:05:10"}
{"current_steps": 190, "total_steps": 1773, "loss": 0.8039, "lr": 4.99941324504621e-06, "epoch": 0.32148900169204736, "percentage": 10.72, "elapsed_time": "0:22:17", "remaining_time": "3:05:47"}
{"current_steps": 200, "total_steps": 1773, "loss": 0.7915, "lr": 4.997861717637322e-06, "epoch": 0.338409475465313, "percentage": 11.28, "elapsed_time": "0:23:36", "remaining_time": "3:05:38"}
{"current_steps": 210, "total_steps": 1773, "loss": 0.7833, "lr": 4.995341170269931e-06, "epoch": 0.3553299492385787, "percentage": 11.84, "elapsed_time": "0:24:28", "remaining_time": "3:02:10"}
{"current_steps": 220, "total_steps": 1773, "loss": 0.7846, "lr": 4.991852580764663e-06, "epoch": 0.37225042301184436, "percentage": 12.41, "elapsed_time": "0:25:27", "remaining_time": "2:59:44"}
{"current_steps": 230, "total_steps": 1773, "loss": 0.7861, "lr": 4.98739730248421e-06, "epoch": 0.38917089678510997, "percentage": 12.97, "elapsed_time": "0:26:23", "remaining_time": "2:57:03"}
{"current_steps": 240, "total_steps": 1773, "loss": 0.7738, "lr": 4.981977063808307e-06, "epoch": 0.40609137055837563, "percentage": 13.54, "elapsed_time": "0:27:26", "remaining_time": "2:55:18"}
{"current_steps": 250, "total_steps": 1773, "loss": 0.7685, "lr": 4.9755939674632255e-06, "epoch": 0.4230118443316413, "percentage": 14.1, "elapsed_time": "0:29:28", "remaining_time": "2:59:31"}
{"current_steps": 260, "total_steps": 1773, "loss": 0.7888, "lr": 4.968250489706043e-06, "epoch": 0.43993231810490696, "percentage": 14.66, "elapsed_time": "0:30:30", "remaining_time": "2:57:32"}
{"current_steps": 270, "total_steps": 1773, "loss": 0.7833, "lr": 4.959949479364002e-06, "epoch": 0.45685279187817257, "percentage": 15.23, "elapsed_time": "0:32:30", "remaining_time": "3:00:57"}
{"current_steps": 280, "total_steps": 1773, "loss": 0.7692, "lr": 4.95069415672934e-06, "epoch": 0.47377326565143824, "percentage": 15.79, "elapsed_time": "0:33:33", "remaining_time": "2:58:56"}
{"current_steps": 290, "total_steps": 1773, "loss": 0.7781, "lr": 4.940488112310007e-06, "epoch": 0.4906937394247039, "percentage": 16.36, "elapsed_time": "0:35:05", "remaining_time": "2:59:27"}
{"current_steps": 300, "total_steps": 1773, "loss": 0.7659, "lr": 4.929335305436764e-06, "epoch": 0.5076142131979695, "percentage": 16.92, "elapsed_time": "0:36:41", "remaining_time": "3:00:08"}
{"current_steps": 310, "total_steps": 1773, "loss": 0.7971, "lr": 4.9172400627272065e-06, "epoch": 0.5245346869712352, "percentage": 17.48, "elapsed_time": "0:37:37", "remaining_time": "2:57:34"}
{"current_steps": 320, "total_steps": 1773, "loss": 0.7652, "lr": 4.904207076407291e-06, "epoch": 0.5414551607445008, "percentage": 18.05, "elapsed_time": "0:39:07", "remaining_time": "2:57:37"}
{"current_steps": 330, "total_steps": 1773, "loss": 0.7711, "lr": 4.890241402491037e-06, "epoch": 0.5583756345177665, "percentage": 18.61, "elapsed_time": "0:40:35", "remaining_time": "2:57:29"}
{"current_steps": 340, "total_steps": 1773, "loss": 0.7697, "lr": 4.875348458819096e-06, "epoch": 0.5752961082910322, "percentage": 19.18, "elapsed_time": "0:41:44", "remaining_time": "2:55:53"}
{"current_steps": 350, "total_steps": 1773, "loss": 0.7492, "lr": 4.85953402295696e-06, "epoch": 0.5922165820642978, "percentage": 19.74, "elapsed_time": "0:42:37", "remaining_time": "2:53:17"}
{"current_steps": 360, "total_steps": 1773, "loss": 0.7747, "lr": 4.842804229953607e-06, "epoch": 0.6091370558375635, "percentage": 20.3, "elapsed_time": "0:44:00", "remaining_time": "2:52:45"}
{"current_steps": 370, "total_steps": 1773, "loss": 0.8018, "lr": 4.825165569961479e-06, "epoch": 0.626057529610829, "percentage": 20.87, "elapsed_time": "0:44:56", "remaining_time": "2:50:26"}
{"current_steps": 380, "total_steps": 1773, "loss": 0.7741, "lr": 4.806624885718687e-06, "epoch": 0.6429780033840947, "percentage": 21.43, "elapsed_time": "0:46:00", "remaining_time": "2:48:40"}
{"current_steps": 390, "total_steps": 1773, "loss": 0.7383, "lr": 4.7871893698944435e-06, "epoch": 0.6598984771573604, "percentage": 22.0, "elapsed_time": "0:47:08", "remaining_time": "2:47:11"}
{"current_steps": 400, "total_steps": 1773, "loss": 0.7498, "lr": 4.766866562298742e-06, "epoch": 0.676818950930626, "percentage": 22.56, "elapsed_time": "0:48:13", "remaining_time": "2:45:32"}
{"current_steps": 410, "total_steps": 1773, "loss": 0.7513, "lr": 4.745664346957362e-06, "epoch": 0.6937394247038917, "percentage": 23.12, "elapsed_time": "0:49:41", "remaining_time": "2:45:11"}
{"current_steps": 420, "total_steps": 1773, "loss": 0.7757, "lr": 4.723590949053337e-06, "epoch": 0.7106598984771574, "percentage": 23.69, "elapsed_time": "0:50:52", "remaining_time": "2:43:51"}
{"current_steps": 430, "total_steps": 1773, "loss": 0.8053, "lr": 4.700654931736093e-06, "epoch": 0.727580372250423, "percentage": 24.25, "elapsed_time": "0:52:00", "remaining_time": "2:42:26"}
{"current_steps": 440, "total_steps": 1773, "loss": 0.7609, "lr": 4.676865192799443e-06, "epoch": 0.7445008460236887, "percentage": 24.82, "elapsed_time": "0:52:53", "remaining_time": "2:40:14"}
{"current_steps": 450, "total_steps": 1773, "loss": 0.7744, "lr": 4.652230961229798e-06, "epoch": 0.7614213197969543, "percentage": 25.38, "elapsed_time": "0:53:56", "remaining_time": "2:38:34"}
{"current_steps": 460, "total_steps": 1773, "loss": 0.753, "lr": 4.626761793625858e-06, "epoch": 0.7783417935702199, "percentage": 25.94, "elapsed_time": "0:54:54", "remaining_time": "2:36:44"}
{"current_steps": 470, "total_steps": 1773, "loss": 0.7616, "lr": 4.600467570491242e-06, "epoch": 0.7952622673434856, "percentage": 26.51, "elapsed_time": "0:55:51", "remaining_time": "2:34:50"}
{"current_steps": 480, "total_steps": 1773, "loss": 0.784, "lr": 4.573358492401438e-06, "epoch": 0.8121827411167513, "percentage": 27.07, "elapsed_time": "0:56:52", "remaining_time": "2:33:12"}
{"current_steps": 490, "total_steps": 1773, "loss": 0.7779, "lr": 4.545445076046596e-06, "epoch": 0.8291032148900169, "percentage": 27.64, "elapsed_time": "0:57:49", "remaining_time": "2:31:23"}
{"current_steps": 500, "total_steps": 1773, "loss": 0.7643, "lr": 4.516738150151689e-06, "epoch": 0.8460236886632826, "percentage": 28.2, "elapsed_time": "0:59:15", "remaining_time": "2:30:52"}
{"current_steps": 510, "total_steps": 1773, "loss": 0.7784, "lr": 4.487248851275617e-06, "epoch": 0.8629441624365483, "percentage": 28.76, "elapsed_time": "1:00:50", "remaining_time": "2:30:40"}
{"current_steps": 520, "total_steps": 1773, "loss": 0.7658, "lr": 4.45698861949089e-06, "epoch": 0.8798646362098139, "percentage": 29.33, "elapsed_time": "1:01:47", "remaining_time": "2:28:54"}
{"current_steps": 530, "total_steps": 1773, "loss": 0.7739, "lr": 4.425969193945572e-06, "epoch": 0.8967851099830795, "percentage": 29.89, "elapsed_time": "1:03:14", "remaining_time": "2:28:18"}
{"current_steps": 540, "total_steps": 1773, "loss": 0.7561, "lr": 4.3942026083091996e-06, "epoch": 0.9137055837563451, "percentage": 30.46, "elapsed_time": "1:04:07", "remaining_time": "2:26:26"}
{"current_steps": 550, "total_steps": 1773, "loss": 0.7753, "lr": 4.361701186104437e-06, "epoch": 0.9306260575296108, "percentage": 31.02, "elapsed_time": "1:05:40", "remaining_time": "2:26:02"}
{"current_steps": 560, "total_steps": 1773, "loss": 0.7941, "lr": 4.328477535926301e-06, "epoch": 0.9475465313028765, "percentage": 31.58, "elapsed_time": "1:07:36", "remaining_time": "2:26:26"}
{"current_steps": 570, "total_steps": 1773, "loss": 0.7598, "lr": 4.294544546550772e-06, "epoch": 0.9644670050761421, "percentage": 32.15, "elapsed_time": "1:08:29", "remaining_time": "2:24:33"}
{"current_steps": 580, "total_steps": 1773, "loss": 0.775, "lr": 4.2599153819347376e-06, "epoch": 0.9813874788494078, "percentage": 32.71, "elapsed_time": "1:09:52", "remaining_time": "2:23:43"}
{"current_steps": 590, "total_steps": 1773, "loss": 0.7617, "lr": 4.224603476109159e-06, "epoch": 0.9983079526226735, "percentage": 33.28, "elapsed_time": "1:10:54", "remaining_time": "2:22:10"}
{"current_steps": 600, "total_steps": 1773, "loss": 0.5589, "lr": 4.188622527967487e-06, "epoch": 1.015228426395939, "percentage": 33.84, "elapsed_time": "1:14:21", "remaining_time": "2:25:22"}
{"current_steps": 610, "total_steps": 1773, "loss": 0.5167, "lr": 4.151986495951307e-06, "epoch": 1.0321489001692048, "percentage": 34.4, "elapsed_time": "1:15:26", "remaining_time": "2:23:50"}
{"current_steps": 620, "total_steps": 1773, "loss": 0.5409, "lr": 4.114709592635307e-06, "epoch": 1.0490693739424704, "percentage": 34.97, "elapsed_time": "1:16:27", "remaining_time": "2:22:11"}
{"current_steps": 630, "total_steps": 1773, "loss": 0.5173, "lr": 4.076806279213656e-06, "epoch": 1.0659898477157361, "percentage": 35.53, "elapsed_time": "1:17:30", "remaining_time": "2:20:38"}
{"current_steps": 640, "total_steps": 1773, "loss": 0.5477, "lr": 4.038291259889927e-06, "epoch": 1.0829103214890017, "percentage": 36.1, "elapsed_time": "1:18:55", "remaining_time": "2:19:43"}
{"current_steps": 650, "total_steps": 1773, "loss": 0.5309, "lr": 3.999179476172759e-06, "epoch": 1.0998307952622675, "percentage": 36.66, "elapsed_time": "1:20:23", "remaining_time": "2:18:53"}
{"current_steps": 660, "total_steps": 1773, "loss": 0.5207, "lr": 3.9594861010794435e-06, "epoch": 1.116751269035533, "percentage": 37.23, "elapsed_time": "1:21:22", "remaining_time": "2:17:13"}
{"current_steps": 670, "total_steps": 1773, "loss": 0.4987, "lr": 3.919226533249711e-06, "epoch": 1.1336717428087986, "percentage": 37.79, "elapsed_time": "1:22:28", "remaining_time": "2:15:47"}
{"current_steps": 680, "total_steps": 1773, "loss": 0.5181, "lr": 3.878416390971983e-06, "epoch": 1.1505922165820643, "percentage": 38.35, "elapsed_time": "1:23:28", "remaining_time": "2:14:10"}
{"current_steps": 690, "total_steps": 1773, "loss": 0.5108, "lr": 3.837071506124422e-06, "epoch": 1.16751269035533, "percentage": 38.92, "elapsed_time": "1:24:26", "remaining_time": "2:12:32"}
{"current_steps": 700, "total_steps": 1773, "loss": 0.5068, "lr": 3.7952079180331116e-06, "epoch": 1.1844331641285957, "percentage": 39.48, "elapsed_time": "1:25:54", "remaining_time": "2:11:41"}
{"current_steps": 710, "total_steps": 1773, "loss": 0.525, "lr": 3.7528418672497615e-06, "epoch": 1.2013536379018612, "percentage": 40.05, "elapsed_time": "1:26:57", "remaining_time": "2:10:11"}
{"current_steps": 720, "total_steps": 1773, "loss": 0.5232, "lr": 3.7099897892513585e-06, "epoch": 1.218274111675127, "percentage": 40.61, "elapsed_time": "1:27:59", "remaining_time": "2:08:41"}
{"current_steps": 730, "total_steps": 1773, "loss": 0.5175, "lr": 3.6666683080641846e-06, "epoch": 1.2351945854483926, "percentage": 41.17, "elapsed_time": "1:29:39", "remaining_time": "2:08:05"}
{"current_steps": 740, "total_steps": 1773, "loss": 0.518, "lr": 3.6228942298146985e-06, "epoch": 1.252115059221658, "percentage": 41.74, "elapsed_time": "1:30:39", "remaining_time": "2:06:33"}
{"current_steps": 750, "total_steps": 1773, "loss": 0.5322, "lr": 3.5786845362097677e-06, "epoch": 1.2690355329949239, "percentage": 42.3, "elapsed_time": "1:31:39", "remaining_time": "2:05:01"}
{"current_steps": 760, "total_steps": 1773, "loss": 0.5157, "lr": 3.5340563779487953e-06, "epoch": 1.2859560067681894, "percentage": 42.87, "elapsed_time": "1:33:08", "remaining_time": "2:04:09"}
{"current_steps": 770, "total_steps": 1773, "loss": 0.5286, "lr": 3.489027068070271e-06, "epoch": 1.3028764805414552, "percentage": 43.43, "elapsed_time": "1:34:00", "remaining_time": "2:02:27"}
{"current_steps": 780, "total_steps": 1773, "loss": 0.5127, "lr": 3.443614075235364e-06, "epoch": 1.3197969543147208, "percentage": 43.99, "elapsed_time": "1:35:03", "remaining_time": "2:01:00"}
{"current_steps": 790, "total_steps": 1773, "loss": 0.5164, "lr": 3.3978350169511293e-06, "epoch": 1.3367174280879865, "percentage": 44.56, "elapsed_time": "1:36:32", "remaining_time": "2:00:07"}
{"current_steps": 800, "total_steps": 1773, "loss": 0.5094, "lr": 3.3517076527359753e-06, "epoch": 1.353637901861252, "percentage": 45.12, "elapsed_time": "1:37:54", "remaining_time": "1:59:05"}
{"current_steps": 810, "total_steps": 1773, "loss": 0.5041, "lr": 3.305249877230044e-06, "epoch": 1.3705583756345177, "percentage": 45.69, "elapsed_time": "1:38:56", "remaining_time": "1:57:37"}
{"current_steps": 820, "total_steps": 1773, "loss": 0.5099, "lr": 3.258479713253162e-06, "epoch": 1.3874788494077834, "percentage": 46.25, "elapsed_time": "1:39:49", "remaining_time": "1:56:00"}
{"current_steps": 830, "total_steps": 1773, "loss": 0.5061, "lr": 3.2114153048130783e-06, "epoch": 1.404399323181049, "percentage": 46.81, "elapsed_time": "1:40:46", "remaining_time": "1:54:29"}
{"current_steps": 840, "total_steps": 1773, "loss": 0.5189, "lr": 3.1640749100666744e-06, "epoch": 1.4213197969543148, "percentage": 47.38, "elapsed_time": "1:42:20", "remaining_time": "1:53:40"}
{"current_steps": 850, "total_steps": 1773, "loss": 0.5154, "lr": 3.1164768942369058e-06, "epoch": 1.4382402707275803, "percentage": 47.94, "elapsed_time": "1:43:13", "remaining_time": "1:52:05"}
{"current_steps": 860, "total_steps": 1773, "loss": 0.5244, "lr": 3.0686397224881947e-06, "epoch": 1.455160744500846, "percentage": 48.51, "elapsed_time": "1:44:10", "remaining_time": "1:50:35"}
{"current_steps": 870, "total_steps": 1773, "loss": 0.5153, "lr": 3.0205819527630633e-06, "epoch": 1.4720812182741116, "percentage": 49.07, "elapsed_time": "1:45:13", "remaining_time": "1:49:12"}
{"current_steps": 880, "total_steps": 1773, "loss": 0.5141, "lr": 2.9723222285827676e-06, "epoch": 1.4890016920473772, "percentage": 49.63, "elapsed_time": "1:46:20", "remaining_time": "1:47:55"}
{"current_steps": 890, "total_steps": 1773, "loss": 0.5104, "lr": 2.923879271814731e-06, "epoch": 1.505922165820643, "percentage": 50.2, "elapsed_time": "1:47:28", "remaining_time": "1:46:37"}
{"current_steps": 900, "total_steps": 1773, "loss": 0.4919, "lr": 2.87527187540959e-06, "epoch": 1.5228426395939088, "percentage": 50.76, "elapsed_time": "1:48:52", "remaining_time": "1:45:36"}
{"current_steps": 910, "total_steps": 1773, "loss": 0.5242, "lr": 2.8265188961106614e-06, "epoch": 1.5397631133671743, "percentage": 51.33, "elapsed_time": "1:49:50", "remaining_time": "1:44:10"}
{"current_steps": 920, "total_steps": 1773, "loss": 0.5215, "lr": 2.777639247138658e-06, "epoch": 1.5566835871404399, "percentage": 51.89, "elapsed_time": "1:50:52", "remaining_time": "1:42:48"}
{"current_steps": 930, "total_steps": 1773, "loss": 0.49, "lr": 2.7286518908545e-06, "epoch": 1.5736040609137056, "percentage": 52.45, "elapsed_time": "1:51:51", "remaining_time": "1:41:23"}
{"current_steps": 940, "total_steps": 1773, "loss": 0.5052, "lr": 2.6795758314030596e-06, "epoch": 1.5905245346869712, "percentage": 53.02, "elapsed_time": "1:53:18", "remaining_time": "1:40:24"}
{"current_steps": 950, "total_steps": 1773, "loss": 0.5358, "lr": 2.6304301073406977e-06, "epoch": 1.6074450084602367, "percentage": 53.58, "elapsed_time": "1:54:12", "remaining_time": "1:38:56"}
{"current_steps": 960, "total_steps": 1773, "loss": 0.5143, "lr": 2.5812337842494517e-06, "epoch": 1.6243654822335025, "percentage": 54.15, "elapsed_time": "1:55:13", "remaining_time": "1:37:34"}
{"current_steps": 970, "total_steps": 1773, "loss": 0.5098, "lr": 2.5320059473407323e-06, "epoch": 1.6412859560067683, "percentage": 54.71, "elapsed_time": "1:56:47", "remaining_time": "1:36:41"}
{"current_steps": 980, "total_steps": 1773, "loss": 0.5262, "lr": 2.482765694051415e-06, "epoch": 1.6582064297800339, "percentage": 55.27, "elapsed_time": "1:58:16", "remaining_time": "1:35:42"}
{"current_steps": 990, "total_steps": 1773, "loss": 0.5183, "lr": 2.433532126635187e-06, "epoch": 1.6751269035532994, "percentage": 55.84, "elapsed_time": "1:59:41", "remaining_time": "1:34:40"}
{"current_steps": 1000, "total_steps": 1773, "loss": 0.5096, "lr": 2.384324344752014e-06, "epoch": 1.6920473773265652, "percentage": 56.4, "elapsed_time": "2:01:11", "remaining_time": "1:33:41"}
{"current_steps": 1010, "total_steps": 1773, "loss": 0.5054, "lr": 2.3351614380586314e-06, "epoch": 1.708967851099831, "percentage": 56.97, "elapsed_time": "2:02:53", "remaining_time": "1:32:50"}
{"current_steps": 1020, "total_steps": 1773, "loss": 0.5442, "lr": 2.2860624788029013e-06, "epoch": 1.7258883248730963, "percentage": 57.53, "elapsed_time": "2:04:25", "remaining_time": "1:31:51"}
{"current_steps": 1030, "total_steps": 1773, "loss": 0.5035, "lr": 2.2370465144249288e-06, "epoch": 1.742808798646362, "percentage": 58.09, "elapsed_time": "2:05:20", "remaining_time": "1:30:24"}
{"current_steps": 1040, "total_steps": 1773, "loss": 0.5138, "lr": 2.188132560167805e-06, "epoch": 1.7597292724196278, "percentage": 58.66, "elapsed_time": "2:06:13", "remaining_time": "1:28:57"}
{"current_steps": 1050, "total_steps": 1773, "loss": 0.5135, "lr": 2.139339591700835e-06, "epoch": 1.7766497461928934, "percentage": 59.22, "elapsed_time": "2:07:10", "remaining_time": "1:27:33"}
{"current_steps": 1060, "total_steps": 1773, "loss": 0.5233, "lr": 2.090686537758124e-06, "epoch": 1.793570219966159, "percentage": 59.79, "elapsed_time": "2:08:02", "remaining_time": "1:26:07"}
{"current_steps": 1070, "total_steps": 1773, "loss": 0.5207, "lr": 2.0421922727953597e-06, "epoch": 1.8104906937394247, "percentage": 60.35, "elapsed_time": "2:08:57", "remaining_time": "1:24:43"}
{"current_steps": 1080, "total_steps": 1773, "loss": 0.508, "lr": 1.993875609667672e-06, "epoch": 1.8274111675126905, "percentage": 60.91, "elapsed_time": "2:10:24", "remaining_time": "1:23:40"}
{"current_steps": 1090, "total_steps": 1773, "loss": 0.5079, "lr": 1.945755292331368e-06, "epoch": 1.844331641285956, "percentage": 61.48, "elapsed_time": "2:11:17", "remaining_time": "1:22:15"}
{"current_steps": 1100, "total_steps": 1773, "loss": 0.5107, "lr": 1.8978499885724098e-06, "epoch": 1.8612521150592216, "percentage": 62.04, "elapsed_time": "2:12:39", "remaining_time": "1:21:09"}
{"current_steps": 1110, "total_steps": 1773, "loss": 0.5226, "lr": 1.8501782827644365e-06, "epoch": 1.8781725888324874, "percentage": 62.61, "elapsed_time": "2:14:03", "remaining_time": "1:20:04"}
{"current_steps": 1120, "total_steps": 1773, "loss": 0.5309, "lr": 1.8027586686591489e-06, "epoch": 1.895093062605753, "percentage": 63.17, "elapsed_time": "2:15:28", "remaining_time": "1:18:59"}
{"current_steps": 1130, "total_steps": 1773, "loss": 0.5162, "lr": 1.7556095422118403e-06, "epoch": 1.9120135363790185, "percentage": 63.73, "elapsed_time": "2:16:30", "remaining_time": "1:17:40"}
{"current_steps": 1140, "total_steps": 1773, "loss": 0.5159, "lr": 1.708749194444881e-06, "epoch": 1.9289340101522843, "percentage": 64.3, "elapsed_time": "2:17:32", "remaining_time": "1:16:22"}
{"current_steps": 1150, "total_steps": 1773, "loss": 0.512, "lr": 1.6621958043518994e-06, "epoch": 1.94585448392555, "percentage": 64.86, "elapsed_time": "2:18:57", "remaining_time": "1:15:16"}
{"current_steps": 1160, "total_steps": 1773, "loss": 0.515, "lr": 1.6159674318454278e-06, "epoch": 1.9627749576988156, "percentage": 65.43, "elapsed_time": "2:19:56", "remaining_time": "1:13:56"}
{"current_steps": 1170, "total_steps": 1773, "loss": 0.4955, "lr": 1.57008201075074e-06, "epoch": 1.9796954314720812, "percentage": 65.99, "elapsed_time": "2:22:04", "remaining_time": "1:13:13"}
{"current_steps": 1180, "total_steps": 1773, "loss": 0.496, "lr": 1.5245573418486136e-06, "epoch": 1.996615905245347, "percentage": 66.55, "elapsed_time": "2:23:31", "remaining_time": "1:12:07"}
{"current_steps": 1190, "total_steps": 1773, "loss": 0.3522, "lr": 1.4794110859696948e-06, "epoch": 2.0135363790186127, "percentage": 67.12, "elapsed_time": "2:26:56", "remaining_time": "1:11:59"}
{"current_steps": 1200, "total_steps": 1773, "loss": 0.2955, "lr": 1.4346607571431562e-06, "epoch": 2.030456852791878, "percentage": 67.68, "elapsed_time": "2:27:56", "remaining_time": "1:10:38"}
{"current_steps": 1210, "total_steps": 1773, "loss": 0.3094, "lr": 1.3903237158023164e-06, "epoch": 2.047377326565144, "percentage": 68.25, "elapsed_time": "2:28:53", "remaining_time": "1:09:16"}
{"current_steps": 1220, "total_steps": 1773, "loss": 0.2993, "lr": 1.346417162049828e-06, "epoch": 2.0642978003384096, "percentage": 68.81, "elapsed_time": "2:29:49", "remaining_time": "1:07:54"}
{"current_steps": 1230, "total_steps": 1773, "loss": 0.3083, "lr": 1.3029581289850824e-06, "epoch": 2.081218274111675, "percentage": 69.37, "elapsed_time": "2:31:10", "remaining_time": "1:06:44"}
{"current_steps": 1240, "total_steps": 1773, "loss": 0.2902, "lr": 1.259963476096381e-06, "epoch": 2.0981387478849407, "percentage": 69.94, "elapsed_time": "2:32:13", "remaining_time": "1:05:26"}
{"current_steps": 1250, "total_steps": 1773, "loss": 0.3012, "lr": 1.2174498827204831e-06, "epoch": 2.1150592216582065, "percentage": 70.5, "elapsed_time": "2:33:09", "remaining_time": "1:04:04"}
{"current_steps": 1260, "total_steps": 1773, "loss": 0.3034, "lr": 1.1754338415720152e-06, "epoch": 2.1319796954314723, "percentage": 71.07, "elapsed_time": "2:34:33", "remaining_time": "1:02:55"}
{"current_steps": 1270, "total_steps": 1773, "loss": 0.3101, "lr": 1.1339316523452967e-06, "epoch": 2.1489001692047376, "percentage": 71.63, "elapsed_time": "2:35:37", "remaining_time": "1:01:38"}
{"current_steps": 1280, "total_steps": 1773, "loss": 0.3142, "lr": 1.0929594153910391e-06, "epoch": 2.1658206429780034, "percentage": 72.19, "elapsed_time": "2:36:36", "remaining_time": "1:00:19"}
{"current_steps": 1290, "total_steps": 1773, "loss": 0.3062, "lr": 1.052533025470379e-06, "epoch": 2.182741116751269, "percentage": 72.76, "elapsed_time": "2:37:40", "remaining_time": "0:59:02"}
{"current_steps": 1300, "total_steps": 1773, "loss": 0.3091, "lr": 1.0126681655886664e-06, "epoch": 2.199661590524535, "percentage": 73.32, "elapsed_time": "2:38:34", "remaining_time": "0:57:41"}
{"current_steps": 1310, "total_steps": 1773, "loss": 0.3132, "lr": 9.733803009114045e-07, "epoch": 2.2165820642978002, "percentage": 73.89, "elapsed_time": "2:40:02", "remaining_time": "0:56:33"}
{"current_steps": 1320, "total_steps": 1773, "loss": 0.3216, "lr": 9.346846727646983e-07, "epoch": 2.233502538071066, "percentage": 74.45, "elapsed_time": "2:41:05", "remaining_time": "0:55:17"}
{"current_steps": 1330, "total_steps": 1773, "loss": 0.3052, "lr": 8.965962927225328e-07, "epoch": 2.250423011844332, "percentage": 75.01, "elapsed_time": "2:42:00", "remaining_time": "0:53:57"}
{"current_steps": 1340, "total_steps": 1773, "loss": 0.299, "lr": 8.591299367831892e-07, "epoch": 2.267343485617597, "percentage": 75.58, "elapsed_time": "2:42:58", "remaining_time": "0:52:39"}
{"current_steps": 1350, "total_steps": 1773, "loss": 0.3102, "lr": 8.223001396370475e-07, "epoch": 2.284263959390863, "percentage": 76.14, "elapsed_time": "2:44:22", "remaining_time": "0:51:30"}
{"current_steps": 1360, "total_steps": 1773, "loss": 0.301, "lr": 7.861211890280024e-07, "epoch": 2.3011844331641287, "percentage": 76.71, "elapsed_time": "2:45:24", "remaining_time": "0:50:13"}
{"current_steps": 1370, "total_steps": 1773, "loss": 0.2919, "lr": 7.506071202106735e-07, "epoch": 2.3181049069373945, "percentage": 77.27, "elapsed_time": "2:46:20", "remaining_time": "0:48:55"}
{"current_steps": 1380, "total_steps": 1773, "loss": 0.3042, "lr": 7.157717105055797e-07, "epoch": 2.33502538071066, "percentage": 77.83, "elapsed_time": "2:47:50", "remaining_time": "0:47:47"}
{"current_steps": 1390, "total_steps": 1773, "loss": 0.2918, "lr": 6.816284739543624e-07, "epoch": 2.3519458544839256, "percentage": 78.4, "elapsed_time": "2:48:55", "remaining_time": "0:46:32"}
{"current_steps": 1400, "total_steps": 1773, "loss": 0.3089, "lr": 6.481906560771525e-07, "epoch": 2.3688663282571913, "percentage": 78.96, "elapsed_time": "2:50:31", "remaining_time": "0:45:25"}
{"current_steps": 1410, "total_steps": 1773, "loss": 0.305, "lr": 6.154712287341066e-07, "epoch": 2.3857868020304567, "percentage": 79.53, "elapsed_time": "2:51:27", "remaining_time": "0:44:08"}
{"current_steps": 1420, "total_steps": 1773, "loss": 0.2979, "lr": 5.834828850931051e-07, "epoch": 2.4027072758037225, "percentage": 80.09, "elapsed_time": "2:52:31", "remaining_time": "0:42:53"}
{"current_steps": 1430, "total_steps": 1773, "loss": 0.3273, "lr": 5.522380347055717e-07, "epoch": 2.4196277495769882, "percentage": 80.65, "elapsed_time": "2:53:33", "remaining_time": "0:41:37"}
{"current_steps": 1440, "total_steps": 1773, "loss": 0.3031, "lr": 5.217487986923048e-07, "epoch": 2.436548223350254, "percentage": 81.22, "elapsed_time": "2:55:02", "remaining_time": "0:40:28"}
{"current_steps": 1450, "total_steps": 1773, "loss": 0.305, "lr": 4.920270050412246e-07, "epoch": 2.4534686971235193, "percentage": 81.78, "elapsed_time": "2:56:04", "remaining_time": "0:39:13"}
{"current_steps": 1460, "total_steps": 1773, "loss": 0.304, "lr": 4.6308418401881633e-07, "epoch": 2.470389170896785, "percentage": 82.35, "elapsed_time": "2:57:05", "remaining_time": "0:37:57"}
{"current_steps": 1470, "total_steps": 1773, "loss": 0.2901, "lr": 4.3493156369708635e-07, "epoch": 2.487309644670051, "percentage": 82.91, "elapsed_time": "2:58:31", "remaining_time": "0:36:47"}
{"current_steps": 1480, "total_steps": 1773, "loss": 0.2974, "lr": 4.0758006559774297e-07, "epoch": 2.504230118443316, "percentage": 83.47, "elapsed_time": "2:59:27", "remaining_time": "0:35:31"}
{"current_steps": 1490, "total_steps": 1773, "loss": 0.2973, "lr": 3.810403004553073e-07, "epoch": 2.521150592216582, "percentage": 84.04, "elapsed_time": "3:00:31", "remaining_time": "0:34:17"}
{"current_steps": 1500, "total_steps": 1773, "loss": 0.3117, "lr": 3.553225641007832e-07, "epoch": 2.5380710659898478, "percentage": 84.6, "elapsed_time": "3:02:07", "remaining_time": "0:33:08"}
{"current_steps": 1510, "total_steps": 1773, "loss": 0.3079, "lr": 3.304368334674965e-07, "epoch": 2.5549915397631136, "percentage": 85.17, "elapsed_time": "3:03:00", "remaining_time": "0:31:52"}
{"current_steps": 1520, "total_steps": 1773, "loss": 0.3095, "lr": 3.0639276272064615e-07, "epoch": 2.571912013536379, "percentage": 85.73, "elapsed_time": "3:03:55", "remaining_time": "0:30:36"}
{"current_steps": 1530, "total_steps": 1773, "loss": 0.2988, "lr": 2.831996795120687e-07, "epoch": 2.5888324873096447, "percentage": 86.29, "elapsed_time": "3:04:52", "remaining_time": "0:29:21"}
{"current_steps": 1540, "total_steps": 1773, "loss": 0.3034, "lr": 2.608665813616704e-07, "epoch": 2.6057529610829104, "percentage": 86.86, "elapsed_time": "3:06:49", "remaining_time": "0:28:15"}
{"current_steps": 1550, "total_steps": 1773, "loss": 0.3045, "lr": 2.3940213216693155e-07, "epoch": 2.6226734348561758, "percentage": 87.42, "elapsed_time": "3:07:59", "remaining_time": "0:27:02"}
{"current_steps": 1560, "total_steps": 1773, "loss": 0.3137, "lr": 2.18814658841838e-07, "epoch": 2.6395939086294415, "percentage": 87.99, "elapsed_time": "3:08:54", "remaining_time": "0:25:47"}
{"current_steps": 1570, "total_steps": 1773, "loss": 0.3209, "lr": 1.9911214808653572e-07, "epoch": 2.6565143824027073, "percentage": 88.55, "elapsed_time": "3:10:39", "remaining_time": "0:24:39"}
{"current_steps": 1580, "total_steps": 1773, "loss": 0.2724, "lr": 1.8030224328897533e-07, "epoch": 2.673434856175973, "percentage": 89.11, "elapsed_time": "3:12:21", "remaining_time": "0:23:29"}
{"current_steps": 1590, "total_steps": 1773, "loss": 0.3038, "lr": 1.623922415597365e-07, "epoch": 2.6903553299492384, "percentage": 89.68, "elapsed_time": "3:13:18", "remaining_time": "0:22:14"}
{"current_steps": 1600, "total_steps": 1773, "loss": 0.2962, "lr": 1.4538909090118846e-07, "epoch": 2.707275803722504, "percentage": 90.24, "elapsed_time": "3:14:21", "remaining_time": "0:21:00"}
{"current_steps": 1610, "total_steps": 1773, "loss": 0.2925, "lr": 1.2929938751208348e-07, "epoch": 2.72419627749577, "percentage": 90.81, "elapsed_time": "3:15:20", "remaining_time": "0:19:46"}
{"current_steps": 1620, "total_steps": 1773, "loss": 0.3131, "lr": 1.1412937322862971e-07, "epoch": 2.7411167512690353, "percentage": 91.37, "elapsed_time": "3:16:11", "remaining_time": "0:18:31"}
{"current_steps": 1630, "total_steps": 1773, "loss": 0.2831, "lr": 9.988493310303271e-08, "epoch": 2.758037225042301, "percentage": 91.93, "elapsed_time": "3:17:10", "remaining_time": "0:17:17"}
{"current_steps": 1640, "total_steps": 1773, "loss": 0.2841, "lr": 8.657159312045082e-08, "epoch": 2.774957698815567, "percentage": 92.5, "elapsed_time": "3:18:23", "remaining_time": "0:16:05"}
{"current_steps": 1650, "total_steps": 1773, "loss": 0.312, "lr": 7.41945180552453e-08, "epoch": 2.7918781725888326, "percentage": 93.06, "elapsed_time": "3:19:18", "remaining_time": "0:14:51"}
{"current_steps": 1660, "total_steps": 1773, "loss": 0.2822, "lr": 6.275850946735856e-08, "epoch": 2.808798646362098, "percentage": 93.63, "elapsed_time": "3:20:13", "remaining_time": "0:13:37"}
{"current_steps": 1670, "total_steps": 1773, "loss": 0.2849, "lr": 5.226800383959979e-08, "epoch": 2.8257191201353637, "percentage": 94.19, "elapsed_time": "3:22:26", "remaining_time": "0:12:29"}
{"current_steps": 1680, "total_steps": 1773, "loss": 0.2752, "lr": 4.2727070856553034e-08, "epoch": 2.8426395939086295, "percentage": 94.75, "elapsed_time": "3:24:03", "remaining_time": "0:11:17"}
{"current_steps": 1690, "total_steps": 1773, "loss": 0.3062, "lr": 3.4139411825788726e-08, "epoch": 2.859560067681895, "percentage": 95.32, "elapsed_time": "3:25:00", "remaining_time": "0:10:04"}
{"current_steps": 1700, "total_steps": 1773, "loss": 0.2695, "lr": 2.6508358241975896e-08, "epoch": 2.8764805414551606, "percentage": 95.88, "elapsed_time": "3:26:37", "remaining_time": "0:08:52"}
{"current_steps": 1710, "total_steps": 1773, "loss": 0.294, "lr": 1.9836870494464866e-08, "epoch": 2.8934010152284264, "percentage": 96.45, "elapsed_time": "3:27:36", "remaining_time": "0:07:38"}
{"current_steps": 1720, "total_steps": 1773, "loss": 0.3155, "lr": 1.4127536718833434e-08, "epoch": 2.910321489001692, "percentage": 97.01, "elapsed_time": "3:28:57", "remaining_time": "0:06:26"}
{"current_steps": 1730, "total_steps": 1773, "loss": 0.2875, "lr": 9.382571792846962e-09, "epoch": 2.927241962774958, "percentage": 97.57, "elapsed_time": "3:30:13", "remaining_time": "0:05:13"}
{"current_steps": 1740, "total_steps": 1773, "loss": 0.3048, "lr": 5.603816477217638e-09, "epoch": 2.9441624365482233, "percentage": 98.14, "elapsed_time": "3:31:08", "remaining_time": "0:04:00"}
{"current_steps": 1750, "total_steps": 1773, "loss": 0.2872, "lr": 2.792736701502097e-09, "epoch": 2.961082910321489, "percentage": 98.7, "elapsed_time": "3:32:05", "remaining_time": "0:02:47"}
{"current_steps": 1760, "total_steps": 1773, "loss": 0.3175, "lr": 9.50422995407707e-10, "epoch": 2.9780033840947544, "percentage": 99.27, "elapsed_time": "3:33:07", "remaining_time": "0:01:34"}
{"current_steps": 1770, "total_steps": 1773, "loss": 0.2763, "lr": 7.759006573210271e-11, "epoch": 2.99492385786802, "percentage": 99.83, "elapsed_time": "3:34:53", "remaining_time": "0:00:21"}