| { | |
| "best_metric": 1.1429458856582642, | |
| "best_model_checkpoint": "temp/checkpoint-23670", | |
| "epoch": 10.0, | |
| "eval_steps": 500, | |
| "global_step": 23670, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00044, | |
| "loss": 4.2184, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00094, | |
| "loss": 2.1594, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0009990685859441152, | |
| "loss": 1.7535, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009980101608806097, | |
| "loss": 1.5491, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009969517358171041, | |
| "loss": 1.5243, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009958933107535987, | |
| "loss": 1.4251, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000994834885690093, | |
| "loss": 1.4419, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0009937764606265877, | |
| "loss": 1.3417, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.000992718035563082, | |
| "loss": 1.4644, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009916596104995767, | |
| "loss": 1.2834, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.000990601185436071, | |
| "loss": 1.3385, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009895427603725656, | |
| "loss": 1.6052, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009884843353090602, | |
| "loss": 1.4924, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009874259102455546, | |
| "loss": 1.306, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009863674851820492, | |
| "loss": 1.2142, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009853090601185436, | |
| "loss": 1.4172, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009842929720575784, | |
| "loss": 1.3925, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009832345469940728, | |
| "loss": 1.2059, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009821761219305674, | |
| "loss": 1.2021, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009811176968670618, | |
| "loss": 1.3228, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009800592718035564, | |
| "loss": 1.2756, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.000979000846740051, | |
| "loss": 1.2971, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0009779424216765453, | |
| "loss": 1.2016, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00097688399661304, | |
| "loss": 1.1839, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0009758255715495343, | |
| "loss": 1.2254, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.000974809483488569, | |
| "loss": 1.2166, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0009737510584250635, | |
| "loss": 1.2849, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0009727349703640982, | |
| "loss": 1.3064, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0009716765453005926, | |
| "loss": 1.2186, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0009706181202370872, | |
| "loss": 1.2075, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0009695596951735818, | |
| "loss": 1.2558, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0009685012701100763, | |
| "loss": 1.1401, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0009674428450465708, | |
| "loss": 1.3006, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0009663844199830653, | |
| "loss": 1.4694, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0009653259949195598, | |
| "loss": 1.2387, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0009642675698560543, | |
| "loss": 1.218, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0009632091447925487, | |
| "loss": 1.2104, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0009621507197290432, | |
| "loss": 1.1609, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0009610922946655377, | |
| "loss": 1.3551, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0009600338696020322, | |
| "loss": 1.1045, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0009589754445385267, | |
| "loss": 1.2547, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0009579170194750212, | |
| "loss": 1.1996, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0009568585944115157, | |
| "loss": 1.2455, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0009558001693480101, | |
| "loss": 1.2218, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0009547417442845046, | |
| "loss": 1.2038, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0009536833192209991, | |
| "loss": 1.3026, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0009526248941574936, | |
| "loss": 1.1764, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0009515664690939881, | |
| "loss": 1.2109, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0009505080440304827, | |
| "loss": 1.1735, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0009494496189669772, | |
| "loss": 1.2407, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0009483911939034717, | |
| "loss": 1.1924, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0009473327688399661, | |
| "loss": 1.1716, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0009462743437764606, | |
| "loss": 1.2708, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0009452159187129551, | |
| "loss": 1.228, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0009441574936494496, | |
| "loss": 1.3187, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0009430990685859442, | |
| "loss": 1.1862, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0009420406435224387, | |
| "loss": 1.1531, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0009409822184589332, | |
| "loss": 1.204, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0009399237933954277, | |
| "loss": 1.1805, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0009388653683319221, | |
| "loss": 1.2059, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0009378069432684166, | |
| "loss": 1.1864, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0009367485182049111, | |
| "loss": 1.305, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0009356900931414056, | |
| "loss": 1.1419, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0009346316680779001, | |
| "loss": 1.1619, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.0009335732430143946, | |
| "loss": 1.1292, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.000932514817950889, | |
| "loss": 1.3506, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.0009314563928873836, | |
| "loss": 1.2318, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.0009303979678238781, | |
| "loss": 1.1838, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.0009293395427603726, | |
| "loss": 1.2502, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.0009282811176968671, | |
| "loss": 1.2143, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.0009272226926333616, | |
| "loss": 1.1603, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.0009261642675698561, | |
| "loss": 1.1874, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.0009251058425063506, | |
| "loss": 1.1922, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.000924047417442845, | |
| "loss": 1.1762, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.0009229889923793395, | |
| "loss": 1.2008, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.000921930567315834, | |
| "loss": 1.366, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 0.0009208721422523285, | |
| "loss": 1.2688, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 0.000919813717188823, | |
| "loss": 1.1463, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 0.0009187552921253175, | |
| "loss": 1.2489, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 0.000917696867061812, | |
| "loss": 1.2217, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 0.0009166384419983066, | |
| "loss": 1.1869, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 0.000915580016934801, | |
| "loss": 1.1489, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 0.0009145215918712955, | |
| "loss": 1.1774, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 0.0009134631668077901, | |
| "loss": 1.05, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 0.0009124047417442846, | |
| "loss": 1.1775, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 0.0009113463166807791, | |
| "loss": 1.2268, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 0.0009102878916172736, | |
| "loss": 1.2144, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 0.0009092294665537681, | |
| "loss": 1.2081, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 0.0009081710414902625, | |
| "loss": 1.1241, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 0.000907112616426757, | |
| "loss": 1.157, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 0.0009060541913632515, | |
| "loss": 1.1556, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 0.000904995766299746, | |
| "loss": 1.1942, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 0.0009039373412362405, | |
| "loss": 1.1466, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 0.000902878916172735, | |
| "loss": 1.1809, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 1.3190160989761353, | |
| "eval_runtime": 7617.6135, | |
| "eval_samples_per_second": 0.62, | |
| "eval_steps_per_second": 0.078, | |
| "eval_wer": 21.58414639083438, | |
| "step": 2367 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.0009018204911092295, | |
| "loss": 1.1354, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 0.000900762066045724, | |
| "loss": 1.1331, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 0.0008997036409822184, | |
| "loss": 1.0803, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 0.0008986452159187129, | |
| "loss": 1.0812, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 0.0008975867908552074, | |
| "loss": 1.0893, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 0.0008965283657917019, | |
| "loss": 1.0959, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 0.0008954699407281964, | |
| "loss": 1.0893, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 0.000894411515664691, | |
| "loss": 1.011, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 0.0008933530906011855, | |
| "loss": 1.0836, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 0.0008922946655376799, | |
| "loss": 1.1211, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 0.0008912362404741744, | |
| "loss": 1.0742, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 0.0008901778154106689, | |
| "loss": 1.1619, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 0.0008891193903471635, | |
| "loss": 1.0852, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 0.000888060965283658, | |
| "loss": 2.875, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 0.0008870025402201525, | |
| "loss": 5.1135, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 0.000885944115156647, | |
| "loss": 1.3177, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 0.0008848856900931415, | |
| "loss": 1.1476, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 0.0008838272650296359, | |
| "loss": 1.0364, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 0.0008827688399661304, | |
| "loss": 1.1123, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 0.0008817104149026249, | |
| "loss": 1.1715, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 0.0008806519898391194, | |
| "loss": 1.1435, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 0.0008795935647756139, | |
| "loss": 1.1766, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 0.0008785351397121084, | |
| "loss": 1.1142, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 0.0008774767146486029, | |
| "loss": 1.1495, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 0.0008764182895850974, | |
| "loss": 1.0953, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 0.0008753598645215919, | |
| "loss": 1.0642, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 0.0008743014394580864, | |
| "loss": 1.0634, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 0.0008732430143945809, | |
| "loss": 1.1236, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 0.0008721845893310754, | |
| "loss": 1.1002, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 0.0008711261642675699, | |
| "loss": 1.0747, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 0.0008700677392040644, | |
| "loss": 1.1069, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 0.0008690093141405588, | |
| "loss": 1.11, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 0.0008679508890770533, | |
| "loss": 1.0691, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 0.0008668924640135478, | |
| "loss": 1.0444, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 0.0008658340389500423, | |
| "loss": 1.1357, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 0.0008647756138865368, | |
| "loss": 1.0616, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 0.0008637171888230313, | |
| "loss": 1.1194, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 0.0008626587637595258, | |
| "loss": 1.0896, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 0.0008616003386960204, | |
| "loss": 1.1071, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 0.0008605419136325148, | |
| "loss": 1.0941, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 0.0008594834885690093, | |
| "loss": 0.9956, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 0.0008584250635055038, | |
| "loss": 1.0825, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 0.0008573666384419984, | |
| "loss": 1.1306, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 0.0008563082133784929, | |
| "loss": 1.1743, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 0.0008552497883149874, | |
| "loss": 1.1171, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 0.0008541913632514819, | |
| "loss": 1.1154, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 0.0008531329381879764, | |
| "loss": 1.2423, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 0.0008520745131244708, | |
| "loss": 1.0663, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 0.0008510160880609653, | |
| "loss": 1.0785, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 0.0008499576629974598, | |
| "loss": 1.1861, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 0.0008488992379339543, | |
| "loss": 1.1114, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 0.0008478408128704488, | |
| "loss": 1.1205, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 0.0008467823878069433, | |
| "loss": 1.2242, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 0.0008457239627434378, | |
| "loss": 1.1194, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 0.0008446655376799322, | |
| "loss": 1.0022, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 0.0008436071126164267, | |
| "loss": 1.1052, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 0.0008425486875529212, | |
| "loss": 1.0726, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 0.0008414902624894157, | |
| "loss": 1.1301, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 0.0008404318374259102, | |
| "loss": 1.1998, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 0.0008393734123624047, | |
| "loss": 1.0497, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 0.0008383149872988993, | |
| "loss": 1.0563, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 0.0008372565622353937, | |
| "loss": 1.0837, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 0.0008361981371718882, | |
| "loss": 1.0801, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 0.0008351397121083828, | |
| "loss": 1.1004, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 0.0008340812870448773, | |
| "loss": 1.1603, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 0.0008330228619813718, | |
| "loss": 1.1206, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 0.0008319644369178663, | |
| "loss": 1.041, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 0.0008309060118543608, | |
| "loss": 1.0606, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 0.0008298475867908553, | |
| "loss": 1.1693, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 0.0008287891617273497, | |
| "loss": 1.1468, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 0.0008277307366638442, | |
| "loss": 1.1067, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 0.0008266723116003387, | |
| "loss": 1.1099, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 0.0008256138865368332, | |
| "loss": 1.1336, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 0.0008245554614733277, | |
| "loss": 1.0815, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 0.0008234970364098222, | |
| "loss": 1.1248, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 0.0008224386113463167, | |
| "loss": 1.0526, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 0.0008213801862828111, | |
| "loss": 1.117, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 0.0008203217612193057, | |
| "loss": 1.1244, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 0.0008192633361558002, | |
| "loss": 1.0995, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 0.0008182049110922947, | |
| "loss": 1.0677, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 0.0008171464860287892, | |
| "loss": 1.1009, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 0.0008160880609652837, | |
| "loss": 1.0575, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 0.0008150296359017782, | |
| "loss": 1.1141, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 0.0008139712108382727, | |
| "loss": 1.1315, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 0.0008129127857747671, | |
| "loss": 1.1144, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 0.0008118543607112616, | |
| "loss": 1.0226, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 0.0008107959356477561, | |
| "loss": 1.1154, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 0.0008097375105842506, | |
| "loss": 1.2119, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 0.0008086790855207451, | |
| "loss": 1.1489, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 0.0008076206604572397, | |
| "loss": 1.1911, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 0.0008065622353937342, | |
| "loss": 1.0795, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 0.0008055038103302286, | |
| "loss": 1.2184, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 0.0008044453852667231, | |
| "loss": 1.2255, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 0.0008033869602032176, | |
| "loss": 1.2472, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 0.0008023285351397121, | |
| "loss": 1.1805, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 1.2831470966339111, | |
| "eval_runtime": 9835.5257, | |
| "eval_samples_per_second": 0.48, | |
| "eval_steps_per_second": 0.06, | |
| "eval_wer": 72.5011133152504, | |
| "step": 4734 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 0.0008012701100762067, | |
| "loss": 1.0692, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 0.0008002116850127012, | |
| "loss": 1.0087, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 0.0007991532599491957, | |
| "loss": 1.0464, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 0.0007980948348856902, | |
| "loss": 1.0134, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 0.0007970364098221846, | |
| "loss": 1.0308, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 0.0007959779847586791, | |
| "loss": 1.0106, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 0.0007949195596951736, | |
| "loss": 1.0188, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 0.0007938611346316681, | |
| "loss": 1.0394, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 0.0007928027095681626, | |
| "loss": 1.0306, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 0.0007917442845046571, | |
| "loss": 0.9953, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 0.0007906858594411516, | |
| "loss": 0.9624, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 0.000789627434377646, | |
| "loss": 1.0608, | |
| "step": 5025 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 0.0007885690093141405, | |
| "loss": 0.9998, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 0.000787510584250635, | |
| "loss": 1.0115, | |
| "step": 5075 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 0.0007864521591871295, | |
| "loss": 1.1508, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 0.000785393734123624, | |
| "loss": 1.0396, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 0.0007843353090601185, | |
| "loss": 1.0139, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 0.000783276883996613, | |
| "loss": 1.0031, | |
| "step": 5175 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 0.0007822184589331076, | |
| "loss": 1.0304, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 0.000781160033869602, | |
| "loss": 0.9774, | |
| "step": 5225 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 0.0007801016088060966, | |
| "loss": 1.0338, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 0.0007790431837425911, | |
| "loss": 1.0422, | |
| "step": 5275 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 0.0007779847586790856, | |
| "loss": 0.9857, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 0.0007769263336155801, | |
| "loss": 1.0122, | |
| "step": 5325 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 0.0007758679085520746, | |
| "loss": 1.1033, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 0.0007748094834885691, | |
| "loss": 1.0402, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 0.0007737510584250635, | |
| "loss": 1.0065, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 0.000772692633361558, | |
| "loss": 1.0225, | |
| "step": 5425 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 0.0007716342082980525, | |
| "loss": 1.0859, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 0.000770575783234547, | |
| "loss": 1.1253, | |
| "step": 5475 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 0.0007695173581710415, | |
| "loss": 0.9967, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 0.000768458933107536, | |
| "loss": 0.9974, | |
| "step": 5525 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 0.0007674005080440305, | |
| "loss": 0.9759, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 0.000766342082980525, | |
| "loss": 1.0695, | |
| "step": 5575 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 0.0007652836579170194, | |
| "loss": 1.0083, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 0.000764225232853514, | |
| "loss": 1.0405, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 0.0007631668077900085, | |
| "loss": 1.0207, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 0.000762108382726503, | |
| "loss": 0.9772, | |
| "step": 5675 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 0.0007610499576629975, | |
| "loss": 1.0163, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 0.000759991532599492, | |
| "loss": 1.006, | |
| "step": 5725 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 0.0007589331075359865, | |
| "loss": 1.0609, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 0.0007578746824724809, | |
| "loss": 1.0546, | |
| "step": 5775 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 0.0007568162574089754, | |
| "loss": 0.9899, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 0.0007557578323454699, | |
| "loss": 0.9937, | |
| "step": 5825 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 0.0007546994072819644, | |
| "loss": 1.0969, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 0.000753640982218459, | |
| "loss": 1.0273, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 0.0007525825571549535, | |
| "loss": 1.0957, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 0.000751524132091448, | |
| "loss": 1.0264, | |
| "step": 5925 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 0.0007504657070279425, | |
| "loss": 1.0108, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 0.0007494072819644369, | |
| "loss": 1.0734, | |
| "step": 5975 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 0.0007483488569009314, | |
| "loss": 1.0501, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 0.0007472904318374259, | |
| "loss": 1.0607, | |
| "step": 6025 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 0.0007462320067739204, | |
| "loss": 1.0588, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 0.000745173581710415, | |
| "loss": 1.066, | |
| "step": 6075 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 0.0007441151566469095, | |
| "loss": 1.072, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 0.000743056731583404, | |
| "loss": 1.0028, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 0.0007419983065198984, | |
| "loss": 1.0594, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 0.0007409398814563929, | |
| "loss": 1.0995, | |
| "step": 6175 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 0.0007398814563928874, | |
| "loss": 1.0289, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 0.0007388230313293819, | |
| "loss": 1.0261, | |
| "step": 6225 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 0.0007377646062658764, | |
| "loss": 1.0534, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 0.0007367061812023709, | |
| "loss": 1.0424, | |
| "step": 6275 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 0.0007356477561388654, | |
| "loss": 1.0151, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 0.0007345893310753598, | |
| "loss": 1.0553, | |
| "step": 6325 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 0.0007335309060118543, | |
| "loss": 1.0892, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 0.0007324724809483488, | |
| "loss": 0.9578, | |
| "step": 6375 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 0.0007314140558848433, | |
| "loss": 0.9877, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 0.0007303556308213378, | |
| "loss": 0.9604, | |
| "step": 6425 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 0.0007292972057578323, | |
| "loss": 1.0158, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 0.0007282387806943268, | |
| "loss": 1.0495, | |
| "step": 6475 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 0.0007271803556308212, | |
| "loss": 1.0623, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 0.000726121930567316, | |
| "loss": 1.0074, | |
| "step": 6525 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 0.0007250635055038104, | |
| "loss": 1.0418, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 0.0007240050804403049, | |
| "loss": 1.0479, | |
| "step": 6575 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 0.0007229466553767994, | |
| "loss": 1.1352, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 0.0007218882303132939, | |
| "loss": 0.9721, | |
| "step": 6625 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 0.0007208298052497884, | |
| "loss": 1.046, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 0.0007197713801862829, | |
| "loss": 1.0826, | |
| "step": 6675 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 0.0007187129551227773, | |
| "loss": 0.9712, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 0.0007176545300592718, | |
| "loss": 1.0559, | |
| "step": 6725 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 0.0007165961049957663, | |
| "loss": 1.0685, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 0.0007155376799322608, | |
| "loss": 1.0689, | |
| "step": 6775 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 0.0007144792548687553, | |
| "loss": 1.0437, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 0.0007134208298052498, | |
| "loss": 1.0884, | |
| "step": 6825 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 0.0007123624047417443, | |
| "loss": 1.0217, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 0.0007113039796782388, | |
| "loss": 1.1075, | |
| "step": 6875 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 0.0007102455546147332, | |
| "loss": 1.0848, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 0.0007091871295512277, | |
| "loss": 1.0616, | |
| "step": 6925 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 0.0007081287044877223, | |
| "loss": 1.0098, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 0.0007070702794242168, | |
| "loss": 1.0668, | |
| "step": 6975 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 0.0007060118543607113, | |
| "loss": 1.0463, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 0.0007049534292972058, | |
| "loss": 0.9984, | |
| "step": 7025 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 0.0007038950042337003, | |
| "loss": 1.069, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 0.0007028365791701947, | |
| "loss": 1.0914, | |
| "step": 7075 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 0.0007017781541066892, | |
| "loss": 1.013, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_loss": 1.2100714445114136, | |
| "eval_runtime": 4312.9003, | |
| "eval_samples_per_second": 1.095, | |
| "eval_steps_per_second": 0.137, | |
| "eval_wer": 46.87866887980244, | |
| "step": 7101 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 0.0007007197290431837, | |
| "loss": 1.0658, | |
| "step": 7125 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 0.0006997036409822185, | |
| "loss": 0.9727, | |
| "step": 7150 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 0.000698645215918713, | |
| "loss": 0.9168, | |
| "step": 7175 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 0.0006975867908552075, | |
| "loss": 0.9955, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 0.000696528365791702, | |
| "loss": 0.9705, | |
| "step": 7225 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 0.0006954699407281965, | |
| "loss": 0.9692, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 0.000694411515664691, | |
| "loss": 0.9555, | |
| "step": 7275 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 0.0006933530906011855, | |
| "loss": 0.9421, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 0.00069229466553768, | |
| "loss": 0.9501, | |
| "step": 7325 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 0.0006912362404741744, | |
| "loss": 0.9575, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 0.0006901778154106689, | |
| "loss": 0.9371, | |
| "step": 7375 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 0.0006891193903471634, | |
| "loss": 0.9485, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 0.0006880609652836579, | |
| "loss": 0.9627, | |
| "step": 7425 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 0.0006870025402201524, | |
| "loss": 0.9501, | |
| "step": 7450 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 0.0006859441151566469, | |
| "loss": 0.9377, | |
| "step": 7475 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 0.0006848856900931414, | |
| "loss": 0.9765, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 0.0006838272650296358, | |
| "loss": 0.9471, | |
| "step": 7525 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 0.0006827688399661303, | |
| "loss": 1.0329, | |
| "step": 7550 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 0.0006817104149026248, | |
| "loss": 1.0231, | |
| "step": 7575 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 0.0006806519898391193, | |
| "loss": 0.9933, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 0.0006795935647756139, | |
| "loss": 0.9411, | |
| "step": 7625 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 0.0006785351397121085, | |
| "loss": 0.9513, | |
| "step": 7650 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 0.000677476714648603, | |
| "loss": 0.9823, | |
| "step": 7675 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 0.0006764182895850975, | |
| "loss": 0.9649, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 0.000675359864521592, | |
| "loss": 0.9613, | |
| "step": 7725 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 0.0006743014394580864, | |
| "loss": 1.0052, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 0.0006732430143945809, | |
| "loss": 1.035, | |
| "step": 7775 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 0.0006721845893310754, | |
| "loss": 0.9563, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 0.0006711261642675699, | |
| "loss": 0.9533, | |
| "step": 7825 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 0.0006700677392040644, | |
| "loss": 1.0112, | |
| "step": 7850 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 0.0006690093141405589, | |
| "loss": 0.9479, | |
| "step": 7875 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 0.0006679508890770534, | |
| "loss": 0.9666, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 0.0006668924640135478, | |
| "loss": 0.9874, | |
| "step": 7925 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 0.0006658340389500423, | |
| "loss": 1.0469, | |
| "step": 7950 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 0.0006647756138865368, | |
| "loss": 1.056, | |
| "step": 7975 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 0.0006637171888230313, | |
| "loss": 0.9862, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 0.0006626587637595258, | |
| "loss": 0.9715, | |
| "step": 8025 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 0.0006616003386960203, | |
| "loss": 0.9228, | |
| "step": 8050 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 0.0006605419136325149, | |
| "loss": 0.9851, | |
| "step": 8075 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 0.0006594834885690093, | |
| "loss": 0.9781, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 0.0006584250635055038, | |
| "loss": 0.9534, | |
| "step": 8125 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 0.0006573666384419983, | |
| "loss": 1.061, | |
| "step": 8150 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 0.0006563082133784928, | |
| "loss": 0.931, | |
| "step": 8175 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 0.0006552497883149873, | |
| "loss": 0.9586, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 0.0006541913632514818, | |
| "loss": 1.0125, | |
| "step": 8225 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 0.0006531329381879763, | |
| "loss": 0.9899, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 0.0006520745131244707, | |
| "loss": 0.962, | |
| "step": 8275 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 0.0006510160880609653, | |
| "loss": 0.9799, | |
| "step": 8300 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 0.0006499576629974598, | |
| "loss": 0.9534, | |
| "step": 8325 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 0.0006488992379339543, | |
| "loss": 0.9681, | |
| "step": 8350 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 0.0006478408128704488, | |
| "loss": 0.9591, | |
| "step": 8375 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 0.0006467823878069433, | |
| "loss": 1.021, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 0.0006457239627434378, | |
| "loss": 1.0503, | |
| "step": 8425 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 0.0006446655376799323, | |
| "loss": 0.957, | |
| "step": 8450 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 0.0006436071126164267, | |
| "loss": 1.0219, | |
| "step": 8475 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 0.0006425486875529213, | |
| "loss": 0.968, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 0.0006414902624894158, | |
| "loss": 0.9812, | |
| "step": 8525 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 0.0006404318374259103, | |
| "loss": 0.9924, | |
| "step": 8550 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 0.0006393734123624048, | |
| "loss": 0.9772, | |
| "step": 8575 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 0.0006383149872988993, | |
| "loss": 1.0635, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 0.0006372565622353938, | |
| "loss": 1.05, | |
| "step": 8625 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 0.0006361981371718883, | |
| "loss": 1.0148, | |
| "step": 8650 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 0.0006351397121083827, | |
| "loss": 0.9081, | |
| "step": 8675 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 0.0006340812870448772, | |
| "loss": 0.991, | |
| "step": 8700 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 0.0006330228619813717, | |
| "loss": 0.9306, | |
| "step": 8725 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 0.0006319644369178662, | |
| "loss": 0.9934, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 0.0006309060118543607, | |
| "loss": 0.8858, | |
| "step": 8775 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 0.0006298475867908552, | |
| "loss": 0.9914, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 0.0006287891617273497, | |
| "loss": 0.9792, | |
| "step": 8825 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 0.0006277307366638441, | |
| "loss": 0.9537, | |
| "step": 8850 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 0.0006266723116003386, | |
| "loss": 1.0256, | |
| "step": 8875 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 0.0006256138865368331, | |
| "loss": 1.019, | |
| "step": 8900 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 0.0006245554614733276, | |
| "loss": 0.9613, | |
| "step": 8925 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 0.0006234970364098223, | |
| "loss": 0.9606, | |
| "step": 8950 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 0.0006224386113463168, | |
| "loss": 1.0394, | |
| "step": 8975 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 0.0006213801862828113, | |
| "loss": 1.0117, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 0.0006203217612193058, | |
| "loss": 0.9886, | |
| "step": 9025 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 0.0006192633361558002, | |
| "loss": 1.0083, | |
| "step": 9050 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 0.0006182049110922947, | |
| "loss": 0.9849, | |
| "step": 9075 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 0.0006171464860287892, | |
| "loss": 0.9729, | |
| "step": 9100 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 0.0006160880609652837, | |
| "loss": 0.9799, | |
| "step": 9125 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 0.0006150296359017782, | |
| "loss": 1.0039, | |
| "step": 9150 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 0.0006139712108382727, | |
| "loss": 1.0674, | |
| "step": 9175 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 0.0006129127857747672, | |
| "loss": 0.9812, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 0.0006118543607112616, | |
| "loss": 0.9987, | |
| "step": 9225 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 0.0006107959356477561, | |
| "loss": 0.9258, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 0.0006097375105842506, | |
| "loss": 0.9777, | |
| "step": 9275 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 0.0006086790855207451, | |
| "loss": 0.9997, | |
| "step": 9300 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 0.0006076206604572396, | |
| "loss": 0.9908, | |
| "step": 9325 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 0.0006065622353937341, | |
| "loss": 0.998, | |
| "step": 9350 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 0.0006055038103302287, | |
| "loss": 0.954, | |
| "step": 9375 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 0.0006044453852667232, | |
| "loss": 1.0006, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 0.0006033869602032176, | |
| "loss": 0.986, | |
| "step": 9425 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 0.0006023285351397121, | |
| "loss": 0.9995, | |
| "step": 9450 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_loss": 1.2009096145629883, | |
| "eval_runtime": 4732.0809, | |
| "eval_samples_per_second": 0.998, | |
| "eval_steps_per_second": 0.125, | |
| "eval_wer": 29.031213311201974, | |
| "step": 9468 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 0.0006012701100762066, | |
| "loss": 0.9877, | |
| "step": 9475 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 0.0006002116850127011, | |
| "loss": 0.9236, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 0.0005991532599491956, | |
| "loss": 0.9548, | |
| "step": 9525 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 0.0005980948348856901, | |
| "loss": 0.9139, | |
| "step": 9550 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 0.0005970364098221847, | |
| "loss": 0.898, | |
| "step": 9575 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 0.0005959779847586791, | |
| "loss": 0.8503, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 0.0005949195596951736, | |
| "loss": 0.9089, | |
| "step": 9625 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 0.0005938611346316681, | |
| "loss": 0.9742, | |
| "step": 9650 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 0.0005928027095681626, | |
| "loss": 0.9052, | |
| "step": 9675 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 0.0005917442845046571, | |
| "loss": 0.9267, | |
| "step": 9700 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 0.0005906858594411516, | |
| "loss": 0.8613, | |
| "step": 9725 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 0.0005896274343776461, | |
| "loss": 0.9189, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 4.13, | |
| "learning_rate": 0.0005885690093141405, | |
| "loss": 0.9323, | |
| "step": 9775 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 0.000587510584250635, | |
| "loss": 0.9681, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "learning_rate": 0.0005864521591871296, | |
| "loss": 0.9357, | |
| "step": 9825 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 0.0005853937341236241, | |
| "loss": 0.9255, | |
| "step": 9850 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 0.0005843353090601186, | |
| "loss": 0.8853, | |
| "step": 9875 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 0.0005833192209991532, | |
| "loss": 1.0524, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 4.19, | |
| "learning_rate": 0.0005822607959356477, | |
| "loss": 0.9598, | |
| "step": 9925 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 0.0005812023708721422, | |
| "loss": 0.9312, | |
| "step": 9950 | |
| }, | |
| { | |
| "epoch": 4.21, | |
| "learning_rate": 0.0005801439458086367, | |
| "loss": 0.9171, | |
| "step": 9975 | |
| }, | |
| { | |
| "epoch": 4.22, | |
| "learning_rate": 0.0005790855207451312, | |
| "loss": 0.9126, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 4.24, | |
| "learning_rate": 0.0005780270956816257, | |
| "loss": 0.8642, | |
| "step": 10025 | |
| }, | |
| { | |
| "epoch": 4.25, | |
| "learning_rate": 0.0005769686706181202, | |
| "loss": 0.9462, | |
| "step": 10050 | |
| }, | |
| { | |
| "epoch": 4.26, | |
| "learning_rate": 0.0005759102455546148, | |
| "loss": 0.8599, | |
| "step": 10075 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 0.0005748518204911093, | |
| "loss": 0.8809, | |
| "step": 10100 | |
| }, | |
| { | |
| "epoch": 4.28, | |
| "learning_rate": 0.0005737933954276038, | |
| "loss": 0.9371, | |
| "step": 10125 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "learning_rate": 0.0005727349703640983, | |
| "loss": 0.8878, | |
| "step": 10150 | |
| }, | |
| { | |
| "epoch": 4.3, | |
| "learning_rate": 0.0005716765453005928, | |
| "loss": 0.8485, | |
| "step": 10175 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 0.0005706181202370873, | |
| "loss": 0.8717, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 0.0005695596951735818, | |
| "loss": 0.9409, | |
| "step": 10225 | |
| }, | |
| { | |
| "epoch": 4.33, | |
| "learning_rate": 0.0005685012701100762, | |
| "loss": 0.8741, | |
| "step": 10250 | |
| }, | |
| { | |
| "epoch": 4.34, | |
| "learning_rate": 0.0005674428450465707, | |
| "loss": 0.9594, | |
| "step": 10275 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 0.0005663844199830652, | |
| "loss": 0.9609, | |
| "step": 10300 | |
| }, | |
| { | |
| "epoch": 4.36, | |
| "learning_rate": 0.0005653259949195597, | |
| "loss": 0.9146, | |
| "step": 10325 | |
| }, | |
| { | |
| "epoch": 4.37, | |
| "learning_rate": 0.0005642675698560542, | |
| "loss": 0.9324, | |
| "step": 10350 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 0.0005632091447925487, | |
| "loss": 0.9157, | |
| "step": 10375 | |
| }, | |
| { | |
| "epoch": 4.39, | |
| "learning_rate": 0.0005621507197290432, | |
| "loss": 0.9275, | |
| "step": 10400 | |
| }, | |
| { | |
| "epoch": 4.4, | |
| "learning_rate": 0.0005610922946655376, | |
| "loss": 0.9113, | |
| "step": 10425 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 0.0005600338696020321, | |
| "loss": 0.9281, | |
| "step": 10450 | |
| }, | |
| { | |
| "epoch": 4.43, | |
| "learning_rate": 0.0005589754445385266, | |
| "loss": 0.9332, | |
| "step": 10475 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 0.0005579170194750212, | |
| "loss": 0.902, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 4.45, | |
| "learning_rate": 0.0005568585944115157, | |
| "loss": 0.9507, | |
| "step": 10525 | |
| }, | |
| { | |
| "epoch": 4.46, | |
| "learning_rate": 0.0005558001693480102, | |
| "loss": 1.015, | |
| "step": 10550 | |
| }, | |
| { | |
| "epoch": 4.47, | |
| "learning_rate": 0.0005547417442845047, | |
| "loss": 0.9087, | |
| "step": 10575 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 0.0005536833192209992, | |
| "loss": 0.9933, | |
| "step": 10600 | |
| }, | |
| { | |
| "epoch": 4.49, | |
| "learning_rate": 0.0005526248941574936, | |
| "loss": 0.9266, | |
| "step": 10625 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 0.0005515664690939881, | |
| "loss": 0.928, | |
| "step": 10650 | |
| }, | |
| { | |
| "epoch": 4.51, | |
| "learning_rate": 0.0005505080440304826, | |
| "loss": 0.973, | |
| "step": 10675 | |
| }, | |
| { | |
| "epoch": 4.52, | |
| "learning_rate": 0.0005494496189669772, | |
| "loss": 0.9384, | |
| "step": 10700 | |
| }, | |
| { | |
| "epoch": 4.53, | |
| "learning_rate": 0.0005483911939034717, | |
| "loss": 0.9409, | |
| "step": 10725 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "learning_rate": 0.0005473327688399662, | |
| "loss": 0.908, | |
| "step": 10750 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 0.0005462743437764607, | |
| "loss": 0.9074, | |
| "step": 10775 | |
| }, | |
| { | |
| "epoch": 4.56, | |
| "learning_rate": 0.0005452159187129551, | |
| "loss": 0.9468, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 4.57, | |
| "learning_rate": 0.0005441574936494496, | |
| "loss": 0.9009, | |
| "step": 10825 | |
| }, | |
| { | |
| "epoch": 4.58, | |
| "learning_rate": 0.0005430990685859441, | |
| "loss": 0.9129, | |
| "step": 10850 | |
| }, | |
| { | |
| "epoch": 4.59, | |
| "learning_rate": 0.0005420406435224386, | |
| "loss": 0.962, | |
| "step": 10875 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 0.0005409822184589331, | |
| "loss": 0.937, | |
| "step": 10900 | |
| }, | |
| { | |
| "epoch": 4.62, | |
| "learning_rate": 0.0005399237933954277, | |
| "loss": 0.8833, | |
| "step": 10925 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 0.0005388653683319222, | |
| "loss": 0.9346, | |
| "step": 10950 | |
| }, | |
| { | |
| "epoch": 4.64, | |
| "learning_rate": 0.0005378069432684167, | |
| "loss": 0.9755, | |
| "step": 10975 | |
| }, | |
| { | |
| "epoch": 4.65, | |
| "learning_rate": 0.0005367485182049111, | |
| "loss": 0.9885, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 0.0005356900931414056, | |
| "loss": 0.9258, | |
| "step": 11025 | |
| }, | |
| { | |
| "epoch": 4.67, | |
| "learning_rate": 0.0005346316680779001, | |
| "loss": 0.9542, | |
| "step": 11050 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "learning_rate": 0.0005335732430143946, | |
| "loss": 0.9243, | |
| "step": 11075 | |
| }, | |
| { | |
| "epoch": 4.69, | |
| "learning_rate": 0.0005325148179508891, | |
| "loss": 0.9488, | |
| "step": 11100 | |
| }, | |
| { | |
| "epoch": 4.7, | |
| "learning_rate": 0.0005314563928873836, | |
| "loss": 0.919, | |
| "step": 11125 | |
| }, | |
| { | |
| "epoch": 4.71, | |
| "learning_rate": 0.0005303979678238781, | |
| "loss": 0.9595, | |
| "step": 11150 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 0.0005293395427603725, | |
| "loss": 0.9857, | |
| "step": 11175 | |
| }, | |
| { | |
| "epoch": 4.73, | |
| "learning_rate": 0.000528281117696867, | |
| "loss": 0.8741, | |
| "step": 11200 | |
| }, | |
| { | |
| "epoch": 4.74, | |
| "learning_rate": 0.0005272226926333615, | |
| "loss": 0.9894, | |
| "step": 11225 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 0.000526164267569856, | |
| "loss": 0.9435, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 4.76, | |
| "learning_rate": 0.0005251058425063505, | |
| "loss": 0.958, | |
| "step": 11275 | |
| }, | |
| { | |
| "epoch": 4.77, | |
| "learning_rate": 0.000524047417442845, | |
| "loss": 0.9052, | |
| "step": 11300 | |
| }, | |
| { | |
| "epoch": 4.78, | |
| "learning_rate": 0.0005229889923793395, | |
| "loss": 0.9745, | |
| "step": 11325 | |
| }, | |
| { | |
| "epoch": 4.8, | |
| "learning_rate": 0.000521930567315834, | |
| "loss": 0.9233, | |
| "step": 11350 | |
| }, | |
| { | |
| "epoch": 4.81, | |
| "learning_rate": 0.0005208721422523286, | |
| "loss": 0.9504, | |
| "step": 11375 | |
| }, | |
| { | |
| "epoch": 4.82, | |
| "learning_rate": 0.0005198137171888231, | |
| "loss": 0.9513, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 4.83, | |
| "learning_rate": 0.0005187552921253176, | |
| "loss": 0.8862, | |
| "step": 11425 | |
| }, | |
| { | |
| "epoch": 4.84, | |
| "learning_rate": 0.0005176968670618121, | |
| "loss": 0.9127, | |
| "step": 11450 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 0.0005166384419983066, | |
| "loss": 0.9122, | |
| "step": 11475 | |
| }, | |
| { | |
| "epoch": 4.86, | |
| "learning_rate": 0.0005155800169348011, | |
| "loss": 0.8923, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 4.87, | |
| "learning_rate": 0.0005145215918712956, | |
| "loss": 0.9599, | |
| "step": 11525 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "learning_rate": 0.00051346316680779, | |
| "loss": 0.908, | |
| "step": 11550 | |
| }, | |
| { | |
| "epoch": 4.89, | |
| "learning_rate": 0.0005124047417442845, | |
| "loss": 0.9405, | |
| "step": 11575 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 0.000511346316680779, | |
| "loss": 0.95, | |
| "step": 11600 | |
| }, | |
| { | |
| "epoch": 4.91, | |
| "learning_rate": 0.0005102878916172735, | |
| "loss": 0.9096, | |
| "step": 11625 | |
| }, | |
| { | |
| "epoch": 4.92, | |
| "learning_rate": 0.000509229466553768, | |
| "loss": 0.8797, | |
| "step": 11650 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 0.0005081710414902625, | |
| "loss": 0.9073, | |
| "step": 11675 | |
| }, | |
| { | |
| "epoch": 4.94, | |
| "learning_rate": 0.000507112616426757, | |
| "loss": 0.9584, | |
| "step": 11700 | |
| }, | |
| { | |
| "epoch": 4.95, | |
| "learning_rate": 0.0005060541913632514, | |
| "loss": 0.9772, | |
| "step": 11725 | |
| }, | |
| { | |
| "epoch": 4.96, | |
| "learning_rate": 0.0005049957662997459, | |
| "loss": 0.8993, | |
| "step": 11750 | |
| }, | |
| { | |
| "epoch": 4.97, | |
| "learning_rate": 0.0005039373412362404, | |
| "loss": 0.9135, | |
| "step": 11775 | |
| }, | |
| { | |
| "epoch": 4.99, | |
| "learning_rate": 0.0005028789161727349, | |
| "loss": 0.922, | |
| "step": 11800 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 0.0005018204911092295, | |
| "loss": 0.9102, | |
| "step": 11825 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_loss": 1.170865774154663, | |
| "eval_runtime": 4761.8546, | |
| "eval_samples_per_second": 0.992, | |
| "eval_steps_per_second": 0.124, | |
| "eval_wer": 27.24788470102425, | |
| "step": 11835 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 0.000500762066045724, | |
| "loss": 0.9302, | |
| "step": 11850 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 0.0004997036409822185, | |
| "loss": 0.8748, | |
| "step": 11875 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 0.000498645215918713, | |
| "loss": 0.8842, | |
| "step": 11900 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 0.0004975867908552074, | |
| "loss": 0.8762, | |
| "step": 11925 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 0.0004965283657917019, | |
| "loss": 0.8903, | |
| "step": 11950 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 0.0004954699407281964, | |
| "loss": 0.8895, | |
| "step": 11975 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 0.0004944115156646909, | |
| "loss": 0.9031, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 0.0004933530906011854, | |
| "loss": 0.8865, | |
| "step": 12025 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 0.00049229466553768, | |
| "loss": 0.853, | |
| "step": 12050 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 0.0004912362404741745, | |
| "loss": 0.8781, | |
| "step": 12075 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 0.000490177815410669, | |
| "loss": 0.8667, | |
| "step": 12100 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 0.0004891193903471634, | |
| "loss": 0.8595, | |
| "step": 12125 | |
| }, | |
| { | |
| "epoch": 5.13, | |
| "learning_rate": 0.000488060965283658, | |
| "loss": 0.8717, | |
| "step": 12150 | |
| }, | |
| { | |
| "epoch": 5.14, | |
| "learning_rate": 0.00048700254022015246, | |
| "loss": 0.8627, | |
| "step": 12175 | |
| }, | |
| { | |
| "epoch": 5.15, | |
| "learning_rate": 0.00048594411515664695, | |
| "loss": 0.8646, | |
| "step": 12200 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "learning_rate": 0.00048488569009314143, | |
| "loss": 0.901, | |
| "step": 12225 | |
| }, | |
| { | |
| "epoch": 5.18, | |
| "learning_rate": 0.0004838272650296359, | |
| "loss": 0.8932, | |
| "step": 12250 | |
| }, | |
| { | |
| "epoch": 5.19, | |
| "learning_rate": 0.0004827688399661304, | |
| "loss": 0.8447, | |
| "step": 12275 | |
| }, | |
| { | |
| "epoch": 5.2, | |
| "learning_rate": 0.0004817104149026249, | |
| "loss": 0.8813, | |
| "step": 12300 | |
| }, | |
| { | |
| "epoch": 5.21, | |
| "learning_rate": 0.0004806519898391194, | |
| "loss": 0.8897, | |
| "step": 12325 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 0.00047959356477561386, | |
| "loss": 0.8765, | |
| "step": 12350 | |
| }, | |
| { | |
| "epoch": 5.23, | |
| "learning_rate": 0.0004785351397121084, | |
| "loss": 0.917, | |
| "step": 12375 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "learning_rate": 0.0004774767146486029, | |
| "loss": 0.8642, | |
| "step": 12400 | |
| }, | |
| { | |
| "epoch": 5.25, | |
| "learning_rate": 0.0004764182895850974, | |
| "loss": 0.8903, | |
| "step": 12425 | |
| }, | |
| { | |
| "epoch": 5.26, | |
| "learning_rate": 0.00047535986452159186, | |
| "loss": 0.8825, | |
| "step": 12450 | |
| }, | |
| { | |
| "epoch": 5.27, | |
| "learning_rate": 0.0004743014394580864, | |
| "loss": 0.864, | |
| "step": 12475 | |
| }, | |
| { | |
| "epoch": 5.28, | |
| "learning_rate": 0.0004732430143945809, | |
| "loss": 0.9053, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 0.00047218458933107537, | |
| "loss": 0.8633, | |
| "step": 12525 | |
| }, | |
| { | |
| "epoch": 5.3, | |
| "learning_rate": 0.00047112616426756986, | |
| "loss": 0.8638, | |
| "step": 12550 | |
| }, | |
| { | |
| "epoch": 5.31, | |
| "learning_rate": 0.00047006773920406434, | |
| "loss": 0.818, | |
| "step": 12575 | |
| }, | |
| { | |
| "epoch": 5.32, | |
| "learning_rate": 0.0004690093141405589, | |
| "loss": 0.8435, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 5.33, | |
| "learning_rate": 0.00046795088907705337, | |
| "loss": 0.8343, | |
| "step": 12625 | |
| }, | |
| { | |
| "epoch": 5.34, | |
| "learning_rate": 0.00046689246401354786, | |
| "loss": 0.8535, | |
| "step": 12650 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 0.00046583403895004234, | |
| "loss": 0.8622, | |
| "step": 12675 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 0.00046477561388653683, | |
| "loss": 0.9058, | |
| "step": 12700 | |
| }, | |
| { | |
| "epoch": 5.38, | |
| "learning_rate": 0.0004637171888230313, | |
| "loss": 0.8637, | |
| "step": 12725 | |
| }, | |
| { | |
| "epoch": 5.39, | |
| "learning_rate": 0.0004626587637595258, | |
| "loss": 0.8698, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 5.4, | |
| "learning_rate": 0.0004616003386960203, | |
| "loss": 0.8424, | |
| "step": 12775 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 0.0004605419136325149, | |
| "loss": 0.8894, | |
| "step": 12800 | |
| }, | |
| { | |
| "epoch": 5.42, | |
| "learning_rate": 0.00045948348856900936, | |
| "loss": 0.8493, | |
| "step": 12825 | |
| }, | |
| { | |
| "epoch": 5.43, | |
| "learning_rate": 0.00045842506350550385, | |
| "loss": 0.8802, | |
| "step": 12850 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 0.00045736663844199834, | |
| "loss": 0.9134, | |
| "step": 12875 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 0.0004563082133784928, | |
| "loss": 0.8727, | |
| "step": 12900 | |
| }, | |
| { | |
| "epoch": 5.46, | |
| "learning_rate": 0.0004552497883149873, | |
| "loss": 0.905, | |
| "step": 12925 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 0.0004541913632514818, | |
| "loss": 0.8687, | |
| "step": 12950 | |
| }, | |
| { | |
| "epoch": 5.48, | |
| "learning_rate": 0.0004531329381879763, | |
| "loss": 0.9142, | |
| "step": 12975 | |
| }, | |
| { | |
| "epoch": 5.49, | |
| "learning_rate": 0.00045207451312447077, | |
| "loss": 0.8446, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 0.0004510160880609653, | |
| "loss": 0.8934, | |
| "step": 13025 | |
| }, | |
| { | |
| "epoch": 5.51, | |
| "learning_rate": 0.0004499576629974598, | |
| "loss": 0.8856, | |
| "step": 13050 | |
| }, | |
| { | |
| "epoch": 5.52, | |
| "learning_rate": 0.0004488992379339543, | |
| "loss": 0.8806, | |
| "step": 13075 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 0.00044784081287044876, | |
| "loss": 0.9213, | |
| "step": 13100 | |
| }, | |
| { | |
| "epoch": 5.54, | |
| "learning_rate": 0.0004467823878069433, | |
| "loss": 0.8352, | |
| "step": 13125 | |
| }, | |
| { | |
| "epoch": 5.56, | |
| "learning_rate": 0.0004457239627434378, | |
| "loss": 0.8827, | |
| "step": 13150 | |
| }, | |
| { | |
| "epoch": 5.57, | |
| "learning_rate": 0.0004446655376799323, | |
| "loss": 0.872, | |
| "step": 13175 | |
| }, | |
| { | |
| "epoch": 5.58, | |
| "learning_rate": 0.00044360711261642676, | |
| "loss": 0.895, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 5.59, | |
| "learning_rate": 0.00044254868755292125, | |
| "loss": 0.936, | |
| "step": 13225 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 0.0004414902624894158, | |
| "loss": 0.8979, | |
| "step": 13250 | |
| }, | |
| { | |
| "epoch": 5.61, | |
| "learning_rate": 0.00044043183742591027, | |
| "loss": 0.8698, | |
| "step": 13275 | |
| }, | |
| { | |
| "epoch": 5.62, | |
| "learning_rate": 0.00043937341236240476, | |
| "loss": 0.8921, | |
| "step": 13300 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "learning_rate": 0.00043831498729889924, | |
| "loss": 0.9028, | |
| "step": 13325 | |
| }, | |
| { | |
| "epoch": 5.64, | |
| "learning_rate": 0.00043725656223539373, | |
| "loss": 0.8962, | |
| "step": 13350 | |
| }, | |
| { | |
| "epoch": 5.65, | |
| "learning_rate": 0.0004361981371718882, | |
| "loss": 0.9641, | |
| "step": 13375 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 0.0004351397121083827, | |
| "loss": 0.8767, | |
| "step": 13400 | |
| }, | |
| { | |
| "epoch": 5.67, | |
| "learning_rate": 0.0004340812870448772, | |
| "loss": 0.9176, | |
| "step": 13425 | |
| }, | |
| { | |
| "epoch": 5.68, | |
| "learning_rate": 0.0004330228619813717, | |
| "loss": 0.8864, | |
| "step": 13450 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 0.00043196443691786627, | |
| "loss": 0.8844, | |
| "step": 13475 | |
| }, | |
| { | |
| "epoch": 5.7, | |
| "learning_rate": 0.00043090601185436075, | |
| "loss": 0.8831, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 5.71, | |
| "learning_rate": 0.00042984758679085524, | |
| "loss": 0.9046, | |
| "step": 13525 | |
| }, | |
| { | |
| "epoch": 5.72, | |
| "learning_rate": 0.0004287891617273497, | |
| "loss": 0.8509, | |
| "step": 13550 | |
| }, | |
| { | |
| "epoch": 5.74, | |
| "learning_rate": 0.0004277307366638442, | |
| "loss": 0.9291, | |
| "step": 13575 | |
| }, | |
| { | |
| "epoch": 5.75, | |
| "learning_rate": 0.0004266723116003387, | |
| "loss": 0.8443, | |
| "step": 13600 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 0.0004256138865368332, | |
| "loss": 0.8935, | |
| "step": 13625 | |
| }, | |
| { | |
| "epoch": 5.77, | |
| "learning_rate": 0.00042455546147332767, | |
| "loss": 0.8912, | |
| "step": 13650 | |
| }, | |
| { | |
| "epoch": 5.78, | |
| "learning_rate": 0.00042349703640982215, | |
| "loss": 0.9162, | |
| "step": 13675 | |
| }, | |
| { | |
| "epoch": 5.79, | |
| "learning_rate": 0.0004224386113463167, | |
| "loss": 0.8835, | |
| "step": 13700 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 0.0004213801862828112, | |
| "loss": 0.8494, | |
| "step": 13725 | |
| }, | |
| { | |
| "epoch": 5.81, | |
| "learning_rate": 0.0004203217612193057, | |
| "loss": 0.8818, | |
| "step": 13750 | |
| }, | |
| { | |
| "epoch": 5.82, | |
| "learning_rate": 0.0004192633361558002, | |
| "loss": 0.9461, | |
| "step": 13775 | |
| }, | |
| { | |
| "epoch": 5.83, | |
| "learning_rate": 0.0004182049110922947, | |
| "loss": 0.8506, | |
| "step": 13800 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 0.0004171464860287892, | |
| "loss": 0.8769, | |
| "step": 13825 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 0.00041608806096528366, | |
| "loss": 0.8792, | |
| "step": 13850 | |
| }, | |
| { | |
| "epoch": 5.86, | |
| "learning_rate": 0.00041502963590177815, | |
| "loss": 0.9077, | |
| "step": 13875 | |
| }, | |
| { | |
| "epoch": 5.87, | |
| "learning_rate": 0.0004139712108382727, | |
| "loss": 0.8636, | |
| "step": 13900 | |
| }, | |
| { | |
| "epoch": 5.88, | |
| "learning_rate": 0.0004129127857747672, | |
| "loss": 0.8776, | |
| "step": 13925 | |
| }, | |
| { | |
| "epoch": 5.89, | |
| "learning_rate": 0.00041185436071126166, | |
| "loss": 0.8983, | |
| "step": 13950 | |
| }, | |
| { | |
| "epoch": 5.9, | |
| "learning_rate": 0.00041079593564775614, | |
| "loss": 0.8978, | |
| "step": 13975 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 0.00040973751058425063, | |
| "loss": 0.875, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 5.93, | |
| "learning_rate": 0.0004086790855207451, | |
| "loss": 0.8698, | |
| "step": 14025 | |
| }, | |
| { | |
| "epoch": 5.94, | |
| "learning_rate": 0.0004076206604572396, | |
| "loss": 0.8782, | |
| "step": 14050 | |
| }, | |
| { | |
| "epoch": 5.95, | |
| "learning_rate": 0.00040656223539373414, | |
| "loss": 0.8637, | |
| "step": 14075 | |
| }, | |
| { | |
| "epoch": 5.96, | |
| "learning_rate": 0.00040550381033022863, | |
| "loss": 0.9123, | |
| "step": 14100 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 0.00040444538526672317, | |
| "loss": 0.8662, | |
| "step": 14125 | |
| }, | |
| { | |
| "epoch": 5.98, | |
| "learning_rate": 0.00040338696020321765, | |
| "loss": 0.9166, | |
| "step": 14150 | |
| }, | |
| { | |
| "epoch": 5.99, | |
| "learning_rate": 0.00040232853513971214, | |
| "loss": 0.8683, | |
| "step": 14175 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 0.0004012701100762066, | |
| "loss": 0.9044, | |
| "step": 14200 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_loss": 1.1590253114700317, | |
| "eval_runtime": 4555.39, | |
| "eval_samples_per_second": 1.037, | |
| "eval_steps_per_second": 0.13, | |
| "eval_wer": 33.9115825270232, | |
| "step": 14202 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 0.0004002116850127011, | |
| "loss": 0.8429, | |
| "step": 14225 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 0.0003991532599491956, | |
| "loss": 0.8537, | |
| "step": 14250 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 0.0003980948348856901, | |
| "loss": 0.8102, | |
| "step": 14275 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 0.00039703640982218457, | |
| "loss": 0.8547, | |
| "step": 14300 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 0.00039597798475867905, | |
| "loss": 0.8256, | |
| "step": 14325 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 0.0003949195596951736, | |
| "loss": 0.8556, | |
| "step": 14350 | |
| }, | |
| { | |
| "epoch": 6.07, | |
| "learning_rate": 0.0003938611346316681, | |
| "loss": 0.8445, | |
| "step": 14375 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 0.0003928027095681626, | |
| "loss": 0.839, | |
| "step": 14400 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 0.0003917442845046571, | |
| "loss": 0.7951, | |
| "step": 14425 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 0.0003906858594411516, | |
| "loss": 0.8409, | |
| "step": 14450 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "learning_rate": 0.0003896274343776461, | |
| "loss": 0.8427, | |
| "step": 14475 | |
| }, | |
| { | |
| "epoch": 6.13, | |
| "learning_rate": 0.00038856900931414056, | |
| "loss": 0.8153, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 6.14, | |
| "learning_rate": 0.00038751058425063505, | |
| "loss": 0.8264, | |
| "step": 14525 | |
| }, | |
| { | |
| "epoch": 6.15, | |
| "learning_rate": 0.00038645215918712953, | |
| "loss": 0.8903, | |
| "step": 14550 | |
| }, | |
| { | |
| "epoch": 6.16, | |
| "learning_rate": 0.0003853937341236241, | |
| "loss": 0.8659, | |
| "step": 14575 | |
| }, | |
| { | |
| "epoch": 6.17, | |
| "learning_rate": 0.00038433530906011856, | |
| "loss": 0.858, | |
| "step": 14600 | |
| }, | |
| { | |
| "epoch": 6.18, | |
| "learning_rate": 0.00038327688399661305, | |
| "loss": 0.8571, | |
| "step": 14625 | |
| }, | |
| { | |
| "epoch": 6.19, | |
| "learning_rate": 0.00038221845893310753, | |
| "loss": 0.847, | |
| "step": 14650 | |
| }, | |
| { | |
| "epoch": 6.2, | |
| "learning_rate": 0.000381160033869602, | |
| "loss": 0.8815, | |
| "step": 14675 | |
| }, | |
| { | |
| "epoch": 6.21, | |
| "learning_rate": 0.0003801016088060965, | |
| "loss": 0.8295, | |
| "step": 14700 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 0.00037904318374259104, | |
| "loss": 0.8391, | |
| "step": 14725 | |
| }, | |
| { | |
| "epoch": 6.23, | |
| "learning_rate": 0.00037798475867908553, | |
| "loss": 0.8106, | |
| "step": 14750 | |
| }, | |
| { | |
| "epoch": 6.24, | |
| "learning_rate": 0.00037692633361558, | |
| "loss": 0.8494, | |
| "step": 14775 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "learning_rate": 0.00037586790855207456, | |
| "loss": 0.7969, | |
| "step": 14800 | |
| }, | |
| { | |
| "epoch": 6.26, | |
| "learning_rate": 0.00037480948348856904, | |
| "loss": 0.8104, | |
| "step": 14825 | |
| }, | |
| { | |
| "epoch": 6.27, | |
| "learning_rate": 0.0003737510584250635, | |
| "loss": 0.8214, | |
| "step": 14850 | |
| }, | |
| { | |
| "epoch": 6.28, | |
| "learning_rate": 0.000372692633361558, | |
| "loss": 0.8566, | |
| "step": 14875 | |
| }, | |
| { | |
| "epoch": 6.29, | |
| "learning_rate": 0.0003716342082980525, | |
| "loss": 0.8315, | |
| "step": 14900 | |
| }, | |
| { | |
| "epoch": 6.31, | |
| "learning_rate": 0.000370575783234547, | |
| "loss": 0.8438, | |
| "step": 14925 | |
| }, | |
| { | |
| "epoch": 6.32, | |
| "learning_rate": 0.00036951735817104147, | |
| "loss": 0.8244, | |
| "step": 14950 | |
| }, | |
| { | |
| "epoch": 6.33, | |
| "learning_rate": 0.00036845893310753596, | |
| "loss": 0.8052, | |
| "step": 14975 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 0.00036740050804403044, | |
| "loss": 0.8215, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 6.35, | |
| "learning_rate": 0.000366342082980525, | |
| "loss": 0.8244, | |
| "step": 15025 | |
| }, | |
| { | |
| "epoch": 6.36, | |
| "learning_rate": 0.0003652836579170195, | |
| "loss": 0.8554, | |
| "step": 15050 | |
| }, | |
| { | |
| "epoch": 6.37, | |
| "learning_rate": 0.000364225232853514, | |
| "loss": 0.837, | |
| "step": 15075 | |
| }, | |
| { | |
| "epoch": 6.38, | |
| "learning_rate": 0.0003631668077900085, | |
| "loss": 0.8483, | |
| "step": 15100 | |
| }, | |
| { | |
| "epoch": 6.39, | |
| "learning_rate": 0.000362108382726503, | |
| "loss": 0.8399, | |
| "step": 15125 | |
| }, | |
| { | |
| "epoch": 6.4, | |
| "learning_rate": 0.00036104995766299747, | |
| "loss": 0.8631, | |
| "step": 15150 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 0.00035999153259949195, | |
| "loss": 0.8093, | |
| "step": 15175 | |
| }, | |
| { | |
| "epoch": 6.42, | |
| "learning_rate": 0.00035893310753598644, | |
| "loss": 0.8194, | |
| "step": 15200 | |
| }, | |
| { | |
| "epoch": 6.43, | |
| "learning_rate": 0.000357874682472481, | |
| "loss": 0.8645, | |
| "step": 15225 | |
| }, | |
| { | |
| "epoch": 6.44, | |
| "learning_rate": 0.00035681625740897546, | |
| "loss": 0.8283, | |
| "step": 15250 | |
| }, | |
| { | |
| "epoch": 6.45, | |
| "learning_rate": 0.00035575783234546995, | |
| "loss": 0.8526, | |
| "step": 15275 | |
| }, | |
| { | |
| "epoch": 6.46, | |
| "learning_rate": 0.00035469940728196443, | |
| "loss": 0.8235, | |
| "step": 15300 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 0.0003536409822184589, | |
| "loss": 0.8818, | |
| "step": 15325 | |
| }, | |
| { | |
| "epoch": 6.49, | |
| "learning_rate": 0.0003525825571549534, | |
| "loss": 0.8597, | |
| "step": 15350 | |
| }, | |
| { | |
| "epoch": 6.5, | |
| "learning_rate": 0.00035152413209144795, | |
| "loss": 0.8344, | |
| "step": 15375 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 0.00035046570702794243, | |
| "loss": 0.8545, | |
| "step": 15400 | |
| }, | |
| { | |
| "epoch": 6.52, | |
| "learning_rate": 0.0003494072819644369, | |
| "loss": 0.8534, | |
| "step": 15425 | |
| }, | |
| { | |
| "epoch": 6.53, | |
| "learning_rate": 0.00034834885690093146, | |
| "loss": 0.8258, | |
| "step": 15450 | |
| }, | |
| { | |
| "epoch": 6.54, | |
| "learning_rate": 0.00034729043183742594, | |
| "loss": 0.8462, | |
| "step": 15475 | |
| }, | |
| { | |
| "epoch": 6.55, | |
| "learning_rate": 0.00034623200677392043, | |
| "loss": 0.9024, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 6.56, | |
| "learning_rate": 0.0003451735817104149, | |
| "loss": 0.8388, | |
| "step": 15525 | |
| }, | |
| { | |
| "epoch": 6.57, | |
| "learning_rate": 0.0003441151566469094, | |
| "loss": 0.8628, | |
| "step": 15550 | |
| }, | |
| { | |
| "epoch": 6.58, | |
| "learning_rate": 0.0003430567315834039, | |
| "loss": 0.8322, | |
| "step": 15575 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 0.00034199830651989837, | |
| "loss": 0.85, | |
| "step": 15600 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "learning_rate": 0.00034093988145639286, | |
| "loss": 0.8186, | |
| "step": 15625 | |
| }, | |
| { | |
| "epoch": 6.61, | |
| "learning_rate": 0.00033988145639288734, | |
| "loss": 0.8289, | |
| "step": 15650 | |
| }, | |
| { | |
| "epoch": 6.62, | |
| "learning_rate": 0.00033882303132938194, | |
| "loss": 0.8344, | |
| "step": 15675 | |
| }, | |
| { | |
| "epoch": 6.63, | |
| "learning_rate": 0.0003377646062658764, | |
| "loss": 0.8698, | |
| "step": 15700 | |
| }, | |
| { | |
| "epoch": 6.64, | |
| "learning_rate": 0.0003367061812023709, | |
| "loss": 0.8414, | |
| "step": 15725 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 0.0003356477561388654, | |
| "loss": 0.8704, | |
| "step": 15750 | |
| }, | |
| { | |
| "epoch": 6.66, | |
| "learning_rate": 0.0003345893310753599, | |
| "loss": 0.8711, | |
| "step": 15775 | |
| }, | |
| { | |
| "epoch": 6.68, | |
| "learning_rate": 0.00033353090601185437, | |
| "loss": 0.8546, | |
| "step": 15800 | |
| }, | |
| { | |
| "epoch": 6.69, | |
| "learning_rate": 0.00033247248094834885, | |
| "loss": 0.8396, | |
| "step": 15825 | |
| }, | |
| { | |
| "epoch": 6.7, | |
| "learning_rate": 0.00033141405588484334, | |
| "loss": 0.8935, | |
| "step": 15850 | |
| }, | |
| { | |
| "epoch": 6.71, | |
| "learning_rate": 0.0003303556308213378, | |
| "loss": 0.8649, | |
| "step": 15875 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "learning_rate": 0.00032929720575783236, | |
| "loss": 0.8264, | |
| "step": 15900 | |
| }, | |
| { | |
| "epoch": 6.73, | |
| "learning_rate": 0.00032823878069432685, | |
| "loss": 0.8468, | |
| "step": 15925 | |
| }, | |
| { | |
| "epoch": 6.74, | |
| "learning_rate": 0.00032718035563082134, | |
| "loss": 0.8418, | |
| "step": 15950 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 0.0003261219305673158, | |
| "loss": 0.8974, | |
| "step": 15975 | |
| }, | |
| { | |
| "epoch": 6.76, | |
| "learning_rate": 0.00032506350550381036, | |
| "loss": 0.8239, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 6.77, | |
| "learning_rate": 0.00032400508044030485, | |
| "loss": 0.8184, | |
| "step": 16025 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 0.00032294665537679933, | |
| "loss": 0.8399, | |
| "step": 16050 | |
| }, | |
| { | |
| "epoch": 6.79, | |
| "learning_rate": 0.0003218882303132938, | |
| "loss": 0.8312, | |
| "step": 16075 | |
| }, | |
| { | |
| "epoch": 6.8, | |
| "learning_rate": 0.0003208298052497883, | |
| "loss": 0.8497, | |
| "step": 16100 | |
| }, | |
| { | |
| "epoch": 6.81, | |
| "learning_rate": 0.00031977138018628284, | |
| "loss": 0.8752, | |
| "step": 16125 | |
| }, | |
| { | |
| "epoch": 6.82, | |
| "learning_rate": 0.00031871295512277733, | |
| "loss": 0.8138, | |
| "step": 16150 | |
| }, | |
| { | |
| "epoch": 6.83, | |
| "learning_rate": 0.0003176545300592718, | |
| "loss": 0.8271, | |
| "step": 16175 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 0.0003165961049957663, | |
| "loss": 0.8722, | |
| "step": 16200 | |
| }, | |
| { | |
| "epoch": 6.85, | |
| "learning_rate": 0.0003155376799322608, | |
| "loss": 0.8186, | |
| "step": 16225 | |
| }, | |
| { | |
| "epoch": 6.87, | |
| "learning_rate": 0.0003144792548687553, | |
| "loss": 0.8512, | |
| "step": 16250 | |
| }, | |
| { | |
| "epoch": 6.88, | |
| "learning_rate": 0.00031342082980524976, | |
| "loss": 0.8943, | |
| "step": 16275 | |
| }, | |
| { | |
| "epoch": 6.89, | |
| "learning_rate": 0.00031236240474174425, | |
| "loss": 0.788, | |
| "step": 16300 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 0.00031130397967823884, | |
| "loss": 0.8147, | |
| "step": 16325 | |
| }, | |
| { | |
| "epoch": 6.91, | |
| "learning_rate": 0.0003102455546147333, | |
| "loss": 0.8541, | |
| "step": 16350 | |
| }, | |
| { | |
| "epoch": 6.92, | |
| "learning_rate": 0.0003091871295512278, | |
| "loss": 0.8571, | |
| "step": 16375 | |
| }, | |
| { | |
| "epoch": 6.93, | |
| "learning_rate": 0.0003081287044877223, | |
| "loss": 0.8386, | |
| "step": 16400 | |
| }, | |
| { | |
| "epoch": 6.94, | |
| "learning_rate": 0.0003070702794242168, | |
| "loss": 0.8413, | |
| "step": 16425 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 0.00030601185436071127, | |
| "loss": 0.8362, | |
| "step": 16450 | |
| }, | |
| { | |
| "epoch": 6.96, | |
| "learning_rate": 0.00030495342929720575, | |
| "loss": 0.8559, | |
| "step": 16475 | |
| }, | |
| { | |
| "epoch": 6.97, | |
| "learning_rate": 0.00030389500423370024, | |
| "loss": 0.85, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 6.98, | |
| "learning_rate": 0.0003028365791701947, | |
| "loss": 0.8117, | |
| "step": 16525 | |
| }, | |
| { | |
| "epoch": 6.99, | |
| "learning_rate": 0.00030177815410668927, | |
| "loss": 0.8633, | |
| "step": 16550 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_loss": 1.1495270729064941, | |
| "eval_runtime": 4489.4556, | |
| "eval_samples_per_second": 1.052, | |
| "eval_steps_per_second": 0.132, | |
| "eval_wer": 30.79429982591798, | |
| "step": 16569 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 0.00030071972904318375, | |
| "loss": 0.8121, | |
| "step": 16575 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 0.00029966130397967824, | |
| "loss": 0.8392, | |
| "step": 16600 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 0.0002986028789161727, | |
| "loss": 0.8302, | |
| "step": 16625 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 0.00029754445385266726, | |
| "loss": 0.8181, | |
| "step": 16650 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 0.00029648602878916175, | |
| "loss": 0.7676, | |
| "step": 16675 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 0.00029542760372565623, | |
| "loss": 0.8366, | |
| "step": 16700 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 0.0002943691786621507, | |
| "loss": 0.8093, | |
| "step": 16725 | |
| }, | |
| { | |
| "epoch": 7.08, | |
| "learning_rate": 0.0002933107535986452, | |
| "loss": 0.7931, | |
| "step": 16750 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 0.00029225232853513975, | |
| "loss": 0.7998, | |
| "step": 16775 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 0.00029119390347163423, | |
| "loss": 0.7657, | |
| "step": 16800 | |
| }, | |
| { | |
| "epoch": 7.11, | |
| "learning_rate": 0.0002901354784081287, | |
| "loss": 0.8145, | |
| "step": 16825 | |
| }, | |
| { | |
| "epoch": 7.12, | |
| "learning_rate": 0.0002890770533446232, | |
| "loss": 0.8068, | |
| "step": 16850 | |
| }, | |
| { | |
| "epoch": 7.13, | |
| "learning_rate": 0.0002880186282811177, | |
| "loss": 0.8559, | |
| "step": 16875 | |
| }, | |
| { | |
| "epoch": 7.14, | |
| "learning_rate": 0.0002869602032176122, | |
| "loss": 0.8337, | |
| "step": 16900 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 0.00028590177815410666, | |
| "loss": 0.8231, | |
| "step": 16925 | |
| }, | |
| { | |
| "epoch": 7.16, | |
| "learning_rate": 0.00028484335309060115, | |
| "loss": 0.8278, | |
| "step": 16950 | |
| }, | |
| { | |
| "epoch": 7.17, | |
| "learning_rate": 0.0002837849280270957, | |
| "loss": 0.7838, | |
| "step": 16975 | |
| }, | |
| { | |
| "epoch": 7.18, | |
| "learning_rate": 0.0002827265029635902, | |
| "loss": 0.8531, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 7.19, | |
| "learning_rate": 0.0002816680779000847, | |
| "loss": 0.781, | |
| "step": 17025 | |
| }, | |
| { | |
| "epoch": 7.2, | |
| "learning_rate": 0.0002806096528365792, | |
| "loss": 0.8274, | |
| "step": 17050 | |
| }, | |
| { | |
| "epoch": 7.21, | |
| "learning_rate": 0.0002795512277730737, | |
| "loss": 0.8052, | |
| "step": 17075 | |
| }, | |
| { | |
| "epoch": 7.22, | |
| "learning_rate": 0.00027849280270956817, | |
| "loss": 0.8197, | |
| "step": 17100 | |
| }, | |
| { | |
| "epoch": 7.23, | |
| "learning_rate": 0.00027743437764606266, | |
| "loss": 0.7912, | |
| "step": 17125 | |
| }, | |
| { | |
| "epoch": 7.25, | |
| "learning_rate": 0.00027637595258255714, | |
| "loss": 0.7939, | |
| "step": 17150 | |
| }, | |
| { | |
| "epoch": 7.26, | |
| "learning_rate": 0.00027531752751905163, | |
| "loss": 0.7834, | |
| "step": 17175 | |
| }, | |
| { | |
| "epoch": 7.27, | |
| "learning_rate": 0.0002742591024555461, | |
| "loss": 0.7949, | |
| "step": 17200 | |
| }, | |
| { | |
| "epoch": 7.28, | |
| "learning_rate": 0.00027320067739204065, | |
| "loss": 0.7883, | |
| "step": 17225 | |
| }, | |
| { | |
| "epoch": 7.29, | |
| "learning_rate": 0.00027214225232853514, | |
| "loss": 0.8239, | |
| "step": 17250 | |
| }, | |
| { | |
| "epoch": 7.3, | |
| "learning_rate": 0.0002710838272650296, | |
| "loss": 0.7861, | |
| "step": 17275 | |
| }, | |
| { | |
| "epoch": 7.31, | |
| "learning_rate": 0.00027002540220152416, | |
| "loss": 0.8098, | |
| "step": 17300 | |
| }, | |
| { | |
| "epoch": 7.32, | |
| "learning_rate": 0.00026896697713801865, | |
| "loss": 0.8416, | |
| "step": 17325 | |
| }, | |
| { | |
| "epoch": 7.33, | |
| "learning_rate": 0.00026795088907705333, | |
| "loss": 0.8691, | |
| "step": 17350 | |
| }, | |
| { | |
| "epoch": 7.34, | |
| "learning_rate": 0.0002668924640135478, | |
| "loss": 0.7902, | |
| "step": 17375 | |
| }, | |
| { | |
| "epoch": 7.35, | |
| "learning_rate": 0.0002658340389500423, | |
| "loss": 0.8293, | |
| "step": 17400 | |
| }, | |
| { | |
| "epoch": 7.36, | |
| "learning_rate": 0.0002647756138865368, | |
| "loss": 0.8133, | |
| "step": 17425 | |
| }, | |
| { | |
| "epoch": 7.37, | |
| "learning_rate": 0.0002637171888230314, | |
| "loss": 0.8323, | |
| "step": 17450 | |
| }, | |
| { | |
| "epoch": 7.38, | |
| "learning_rate": 0.00026265876375952587, | |
| "loss": 0.8169, | |
| "step": 17475 | |
| }, | |
| { | |
| "epoch": 7.39, | |
| "learning_rate": 0.00026160033869602036, | |
| "loss": 0.7773, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 0.00026054191363251484, | |
| "loss": 0.7888, | |
| "step": 17525 | |
| }, | |
| { | |
| "epoch": 7.41, | |
| "learning_rate": 0.00025948348856900933, | |
| "loss": 0.828, | |
| "step": 17550 | |
| }, | |
| { | |
| "epoch": 7.43, | |
| "learning_rate": 0.0002584250635055038, | |
| "loss": 0.8018, | |
| "step": 17575 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "learning_rate": 0.0002573666384419983, | |
| "loss": 0.7747, | |
| "step": 17600 | |
| }, | |
| { | |
| "epoch": 7.45, | |
| "learning_rate": 0.0002563082133784928, | |
| "loss": 0.7546, | |
| "step": 17625 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "learning_rate": 0.00025524978831498727, | |
| "loss": 0.8317, | |
| "step": 17650 | |
| }, | |
| { | |
| "epoch": 7.47, | |
| "learning_rate": 0.0002541913632514818, | |
| "loss": 0.8325, | |
| "step": 17675 | |
| }, | |
| { | |
| "epoch": 7.48, | |
| "learning_rate": 0.0002531329381879763, | |
| "loss": 0.8144, | |
| "step": 17700 | |
| }, | |
| { | |
| "epoch": 7.49, | |
| "learning_rate": 0.0002520745131244708, | |
| "loss": 0.8127, | |
| "step": 17725 | |
| }, | |
| { | |
| "epoch": 7.5, | |
| "learning_rate": 0.00025101608806096527, | |
| "loss": 0.823, | |
| "step": 17750 | |
| }, | |
| { | |
| "epoch": 7.51, | |
| "learning_rate": 0.0002499576629974598, | |
| "loss": 0.8121, | |
| "step": 17775 | |
| }, | |
| { | |
| "epoch": 7.52, | |
| "learning_rate": 0.0002488992379339543, | |
| "loss": 0.8345, | |
| "step": 17800 | |
| }, | |
| { | |
| "epoch": 7.53, | |
| "learning_rate": 0.0002478408128704488, | |
| "loss": 0.8262, | |
| "step": 17825 | |
| }, | |
| { | |
| "epoch": 7.54, | |
| "learning_rate": 0.00024678238780694327, | |
| "loss": 0.822, | |
| "step": 17850 | |
| }, | |
| { | |
| "epoch": 7.55, | |
| "learning_rate": 0.0002457239627434378, | |
| "loss": 0.8367, | |
| "step": 17875 | |
| }, | |
| { | |
| "epoch": 7.56, | |
| "learning_rate": 0.0002446655376799323, | |
| "loss": 0.8349, | |
| "step": 17900 | |
| }, | |
| { | |
| "epoch": 7.57, | |
| "learning_rate": 0.00024360711261642675, | |
| "loss": 0.8144, | |
| "step": 17925 | |
| }, | |
| { | |
| "epoch": 7.58, | |
| "learning_rate": 0.00024254868755292126, | |
| "loss": 0.8128, | |
| "step": 17950 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 0.00024149026248941575, | |
| "loss": 0.8355, | |
| "step": 17975 | |
| }, | |
| { | |
| "epoch": 7.6, | |
| "learning_rate": 0.00024043183742591023, | |
| "loss": 0.8113, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 7.62, | |
| "learning_rate": 0.00023937341236240475, | |
| "loss": 0.7984, | |
| "step": 18025 | |
| }, | |
| { | |
| "epoch": 7.63, | |
| "learning_rate": 0.00023831498729889926, | |
| "loss": 0.8436, | |
| "step": 18050 | |
| }, | |
| { | |
| "epoch": 7.64, | |
| "learning_rate": 0.00023725656223539375, | |
| "loss": 0.7881, | |
| "step": 18075 | |
| }, | |
| { | |
| "epoch": 7.65, | |
| "learning_rate": 0.00023619813717188823, | |
| "loss": 0.7962, | |
| "step": 18100 | |
| }, | |
| { | |
| "epoch": 7.66, | |
| "learning_rate": 0.00023513971210838272, | |
| "loss": 0.8086, | |
| "step": 18125 | |
| }, | |
| { | |
| "epoch": 7.67, | |
| "learning_rate": 0.0002340812870448772, | |
| "loss": 0.7818, | |
| "step": 18150 | |
| }, | |
| { | |
| "epoch": 7.68, | |
| "learning_rate": 0.00023302286198137174, | |
| "loss": 0.8205, | |
| "step": 18175 | |
| }, | |
| { | |
| "epoch": 7.69, | |
| "learning_rate": 0.00023196443691786623, | |
| "loss": 0.8163, | |
| "step": 18200 | |
| }, | |
| { | |
| "epoch": 7.7, | |
| "learning_rate": 0.00023090601185436072, | |
| "loss": 0.8084, | |
| "step": 18225 | |
| }, | |
| { | |
| "epoch": 7.71, | |
| "learning_rate": 0.0002298475867908552, | |
| "loss": 0.8133, | |
| "step": 18250 | |
| }, | |
| { | |
| "epoch": 7.72, | |
| "learning_rate": 0.00022878916172734971, | |
| "loss": 0.8366, | |
| "step": 18275 | |
| }, | |
| { | |
| "epoch": 7.73, | |
| "learning_rate": 0.0002277307366638442, | |
| "loss": 0.7847, | |
| "step": 18300 | |
| }, | |
| { | |
| "epoch": 7.74, | |
| "learning_rate": 0.00022667231160033869, | |
| "loss": 0.8318, | |
| "step": 18325 | |
| }, | |
| { | |
| "epoch": 7.75, | |
| "learning_rate": 0.0002256138865368332, | |
| "loss": 0.8503, | |
| "step": 18350 | |
| }, | |
| { | |
| "epoch": 7.76, | |
| "learning_rate": 0.0002245554614733277, | |
| "loss": 0.8147, | |
| "step": 18375 | |
| }, | |
| { | |
| "epoch": 7.77, | |
| "learning_rate": 0.0002234970364098222, | |
| "loss": 0.8033, | |
| "step": 18400 | |
| }, | |
| { | |
| "epoch": 7.78, | |
| "learning_rate": 0.00022243861134631668, | |
| "loss": 0.8463, | |
| "step": 18425 | |
| }, | |
| { | |
| "epoch": 7.79, | |
| "learning_rate": 0.00022138018628281117, | |
| "loss": 0.7797, | |
| "step": 18450 | |
| }, | |
| { | |
| "epoch": 7.81, | |
| "learning_rate": 0.00022032176121930565, | |
| "loss": 0.8037, | |
| "step": 18475 | |
| }, | |
| { | |
| "epoch": 7.82, | |
| "learning_rate": 0.0002192633361558002, | |
| "loss": 0.8347, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 7.83, | |
| "learning_rate": 0.00021820491109229468, | |
| "loss": 0.8004, | |
| "step": 18525 | |
| }, | |
| { | |
| "epoch": 7.84, | |
| "learning_rate": 0.00021714648602878917, | |
| "loss": 0.7951, | |
| "step": 18550 | |
| }, | |
| { | |
| "epoch": 7.85, | |
| "learning_rate": 0.00021608806096528365, | |
| "loss": 0.8264, | |
| "step": 18575 | |
| }, | |
| { | |
| "epoch": 7.86, | |
| "learning_rate": 0.00021502963590177816, | |
| "loss": 0.8092, | |
| "step": 18600 | |
| }, | |
| { | |
| "epoch": 7.87, | |
| "learning_rate": 0.00021397121083827265, | |
| "loss": 0.8175, | |
| "step": 18625 | |
| }, | |
| { | |
| "epoch": 7.88, | |
| "learning_rate": 0.00021291278577476714, | |
| "loss": 0.843, | |
| "step": 18650 | |
| }, | |
| { | |
| "epoch": 7.89, | |
| "learning_rate": 0.00021185436071126165, | |
| "loss": 0.8217, | |
| "step": 18675 | |
| }, | |
| { | |
| "epoch": 7.9, | |
| "learning_rate": 0.00021079593564775614, | |
| "loss": 0.8412, | |
| "step": 18700 | |
| }, | |
| { | |
| "epoch": 7.91, | |
| "learning_rate": 0.00020973751058425065, | |
| "loss": 0.8262, | |
| "step": 18725 | |
| }, | |
| { | |
| "epoch": 7.92, | |
| "learning_rate": 0.00020867908552074513, | |
| "loss": 0.7942, | |
| "step": 18750 | |
| }, | |
| { | |
| "epoch": 7.93, | |
| "learning_rate": 0.00020762066045723962, | |
| "loss": 0.7943, | |
| "step": 18775 | |
| }, | |
| { | |
| "epoch": 7.94, | |
| "learning_rate": 0.0002065622353937341, | |
| "loss": 0.8093, | |
| "step": 18800 | |
| }, | |
| { | |
| "epoch": 7.95, | |
| "learning_rate": 0.00020550381033022865, | |
| "loss": 0.836, | |
| "step": 18825 | |
| }, | |
| { | |
| "epoch": 7.96, | |
| "learning_rate": 0.00020444538526672313, | |
| "loss": 0.8339, | |
| "step": 18850 | |
| }, | |
| { | |
| "epoch": 7.97, | |
| "learning_rate": 0.00020338696020321762, | |
| "loss": 0.8359, | |
| "step": 18875 | |
| }, | |
| { | |
| "epoch": 7.98, | |
| "learning_rate": 0.0002023285351397121, | |
| "loss": 0.8264, | |
| "step": 18900 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 0.00020127011007620662, | |
| "loss": 0.8115, | |
| "step": 18925 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_loss": 1.1491515636444092, | |
| "eval_runtime": 4396.2585, | |
| "eval_samples_per_second": 1.074, | |
| "eval_steps_per_second": 0.134, | |
| "eval_wer": 30.25990850572851, | |
| "step": 18936 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 0.0002002116850127011, | |
| "loss": 0.7965, | |
| "step": 18950 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 0.00019915325994919561, | |
| "loss": 0.7948, | |
| "step": 18975 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 0.0001980948348856901, | |
| "loss": 0.7899, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 0.00019703640982218459, | |
| "loss": 0.8032, | |
| "step": 19025 | |
| }, | |
| { | |
| "epoch": 8.05, | |
| "learning_rate": 0.0001959779847586791, | |
| "loss": 0.8157, | |
| "step": 19050 | |
| }, | |
| { | |
| "epoch": 8.06, | |
| "learning_rate": 0.00019491955969517358, | |
| "loss": 0.8071, | |
| "step": 19075 | |
| }, | |
| { | |
| "epoch": 8.07, | |
| "learning_rate": 0.00019386113463166807, | |
| "loss": 0.8087, | |
| "step": 19100 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 0.00019280270956816256, | |
| "loss": 0.7975, | |
| "step": 19125 | |
| }, | |
| { | |
| "epoch": 8.09, | |
| "learning_rate": 0.0001917866215071973, | |
| "loss": 0.772, | |
| "step": 19150 | |
| }, | |
| { | |
| "epoch": 8.1, | |
| "learning_rate": 0.0001907281964436918, | |
| "loss": 0.8019, | |
| "step": 19175 | |
| }, | |
| { | |
| "epoch": 8.11, | |
| "learning_rate": 0.0001896697713801863, | |
| "loss": 0.7898, | |
| "step": 19200 | |
| }, | |
| { | |
| "epoch": 8.12, | |
| "learning_rate": 0.00018861134631668078, | |
| "loss": 0.7724, | |
| "step": 19225 | |
| }, | |
| { | |
| "epoch": 8.13, | |
| "learning_rate": 0.00018755292125317526, | |
| "loss": 0.7709, | |
| "step": 19250 | |
| }, | |
| { | |
| "epoch": 8.14, | |
| "learning_rate": 0.00018649449618966978, | |
| "loss": 0.7686, | |
| "step": 19275 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 0.0001854360711261643, | |
| "loss": 0.8047, | |
| "step": 19300 | |
| }, | |
| { | |
| "epoch": 8.16, | |
| "learning_rate": 0.00018437764606265877, | |
| "loss": 0.7813, | |
| "step": 19325 | |
| }, | |
| { | |
| "epoch": 8.17, | |
| "learning_rate": 0.00018331922099915326, | |
| "loss": 0.7989, | |
| "step": 19350 | |
| }, | |
| { | |
| "epoch": 8.19, | |
| "learning_rate": 0.00018226079593564777, | |
| "loss": 0.7918, | |
| "step": 19375 | |
| }, | |
| { | |
| "epoch": 8.2, | |
| "learning_rate": 0.00018120237087214226, | |
| "loss": 0.8023, | |
| "step": 19400 | |
| }, | |
| { | |
| "epoch": 8.21, | |
| "learning_rate": 0.00018014394580863675, | |
| "loss": 0.8119, | |
| "step": 19425 | |
| }, | |
| { | |
| "epoch": 8.22, | |
| "learning_rate": 0.00017908552074513123, | |
| "loss": 0.7831, | |
| "step": 19450 | |
| }, | |
| { | |
| "epoch": 8.23, | |
| "learning_rate": 0.00017802709568162574, | |
| "loss": 0.8159, | |
| "step": 19475 | |
| }, | |
| { | |
| "epoch": 8.24, | |
| "learning_rate": 0.00017696867061812026, | |
| "loss": 0.8149, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 8.25, | |
| "learning_rate": 0.00017591024555461474, | |
| "loss": 0.7954, | |
| "step": 19525 | |
| }, | |
| { | |
| "epoch": 8.26, | |
| "learning_rate": 0.00017485182049110923, | |
| "loss": 0.7951, | |
| "step": 19550 | |
| }, | |
| { | |
| "epoch": 8.27, | |
| "learning_rate": 0.00017379339542760371, | |
| "loss": 0.8232, | |
| "step": 19575 | |
| }, | |
| { | |
| "epoch": 8.28, | |
| "learning_rate": 0.00017273497036409823, | |
| "loss": 0.8129, | |
| "step": 19600 | |
| }, | |
| { | |
| "epoch": 8.29, | |
| "learning_rate": 0.00017167654530059274, | |
| "loss": 0.8182, | |
| "step": 19625 | |
| }, | |
| { | |
| "epoch": 8.3, | |
| "learning_rate": 0.00017061812023708723, | |
| "loss": 0.7852, | |
| "step": 19650 | |
| }, | |
| { | |
| "epoch": 8.31, | |
| "learning_rate": 0.0001695596951735817, | |
| "loss": 0.787, | |
| "step": 19675 | |
| }, | |
| { | |
| "epoch": 8.32, | |
| "learning_rate": 0.0001685012701100762, | |
| "loss": 0.8176, | |
| "step": 19700 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 0.0001674428450465707, | |
| "loss": 0.7965, | |
| "step": 19725 | |
| }, | |
| { | |
| "epoch": 8.34, | |
| "learning_rate": 0.0001663844199830652, | |
| "loss": 0.7661, | |
| "step": 19750 | |
| }, | |
| { | |
| "epoch": 8.35, | |
| "learning_rate": 0.00016532599491955968, | |
| "loss": 0.8, | |
| "step": 19775 | |
| }, | |
| { | |
| "epoch": 8.37, | |
| "learning_rate": 0.0001642675698560542, | |
| "loss": 0.7841, | |
| "step": 19800 | |
| }, | |
| { | |
| "epoch": 8.38, | |
| "learning_rate": 0.0001632091447925487, | |
| "loss": 0.7963, | |
| "step": 19825 | |
| }, | |
| { | |
| "epoch": 8.39, | |
| "learning_rate": 0.0001621507197290432, | |
| "loss": 0.7871, | |
| "step": 19850 | |
| }, | |
| { | |
| "epoch": 8.4, | |
| "learning_rate": 0.00016109229466553768, | |
| "loss": 0.7896, | |
| "step": 19875 | |
| }, | |
| { | |
| "epoch": 8.41, | |
| "learning_rate": 0.00016003386960203217, | |
| "loss": 0.785, | |
| "step": 19900 | |
| }, | |
| { | |
| "epoch": 8.42, | |
| "learning_rate": 0.00015897544453852668, | |
| "loss": 0.8418, | |
| "step": 19925 | |
| }, | |
| { | |
| "epoch": 8.43, | |
| "learning_rate": 0.0001579170194750212, | |
| "loss": 0.7504, | |
| "step": 19950 | |
| }, | |
| { | |
| "epoch": 8.44, | |
| "learning_rate": 0.00015685859441151568, | |
| "loss": 0.8043, | |
| "step": 19975 | |
| }, | |
| { | |
| "epoch": 8.45, | |
| "learning_rate": 0.00015580016934801016, | |
| "loss": 0.7874, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "learning_rate": 0.00015474174428450465, | |
| "loss": 0.7888, | |
| "step": 20025 | |
| }, | |
| { | |
| "epoch": 8.47, | |
| "learning_rate": 0.00015368331922099916, | |
| "loss": 0.7702, | |
| "step": 20050 | |
| }, | |
| { | |
| "epoch": 8.48, | |
| "learning_rate": 0.00015262489415749365, | |
| "loss": 0.8029, | |
| "step": 20075 | |
| }, | |
| { | |
| "epoch": 8.49, | |
| "learning_rate": 0.00015156646909398813, | |
| "loss": 0.772, | |
| "step": 20100 | |
| }, | |
| { | |
| "epoch": 8.5, | |
| "learning_rate": 0.00015050804403048265, | |
| "loss": 0.7777, | |
| "step": 20125 | |
| }, | |
| { | |
| "epoch": 8.51, | |
| "learning_rate": 0.00014944961896697716, | |
| "loss": 0.7935, | |
| "step": 20150 | |
| }, | |
| { | |
| "epoch": 8.52, | |
| "learning_rate": 0.00014839119390347164, | |
| "loss": 0.8093, | |
| "step": 20175 | |
| }, | |
| { | |
| "epoch": 8.53, | |
| "learning_rate": 0.00014733276883996613, | |
| "loss": 0.7686, | |
| "step": 20200 | |
| }, | |
| { | |
| "epoch": 8.54, | |
| "learning_rate": 0.00014627434377646062, | |
| "loss": 0.8484, | |
| "step": 20225 | |
| }, | |
| { | |
| "epoch": 8.56, | |
| "learning_rate": 0.0001452159187129551, | |
| "loss": 0.8237, | |
| "step": 20250 | |
| }, | |
| { | |
| "epoch": 8.57, | |
| "learning_rate": 0.00014415749364944964, | |
| "loss": 0.7896, | |
| "step": 20275 | |
| }, | |
| { | |
| "epoch": 8.58, | |
| "learning_rate": 0.00014309906858594413, | |
| "loss": 0.8252, | |
| "step": 20300 | |
| }, | |
| { | |
| "epoch": 8.59, | |
| "learning_rate": 0.0001420406435224386, | |
| "loss": 0.8128, | |
| "step": 20325 | |
| }, | |
| { | |
| "epoch": 8.6, | |
| "learning_rate": 0.0001409822184589331, | |
| "loss": 0.8201, | |
| "step": 20350 | |
| }, | |
| { | |
| "epoch": 8.61, | |
| "learning_rate": 0.0001399237933954276, | |
| "loss": 0.7848, | |
| "step": 20375 | |
| }, | |
| { | |
| "epoch": 8.62, | |
| "learning_rate": 0.0001388653683319221, | |
| "loss": 0.8147, | |
| "step": 20400 | |
| }, | |
| { | |
| "epoch": 8.63, | |
| "learning_rate": 0.0001378069432684166, | |
| "loss": 0.7742, | |
| "step": 20425 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 0.0001367485182049111, | |
| "loss": 0.7973, | |
| "step": 20450 | |
| }, | |
| { | |
| "epoch": 8.65, | |
| "learning_rate": 0.0001356900931414056, | |
| "loss": 0.7897, | |
| "step": 20475 | |
| }, | |
| { | |
| "epoch": 8.66, | |
| "learning_rate": 0.0001346316680779001, | |
| "loss": 0.8092, | |
| "step": 20500 | |
| }, | |
| { | |
| "epoch": 8.67, | |
| "learning_rate": 0.00013357324301439458, | |
| "loss": 0.749, | |
| "step": 20525 | |
| }, | |
| { | |
| "epoch": 8.68, | |
| "learning_rate": 0.00013251481795088907, | |
| "loss": 0.7846, | |
| "step": 20550 | |
| }, | |
| { | |
| "epoch": 8.69, | |
| "learning_rate": 0.00013145639288738355, | |
| "loss": 0.8339, | |
| "step": 20575 | |
| }, | |
| { | |
| "epoch": 8.7, | |
| "learning_rate": 0.0001303979678238781, | |
| "loss": 0.8287, | |
| "step": 20600 | |
| }, | |
| { | |
| "epoch": 8.71, | |
| "learning_rate": 0.00012933954276037258, | |
| "loss": 0.8201, | |
| "step": 20625 | |
| }, | |
| { | |
| "epoch": 8.72, | |
| "learning_rate": 0.00012828111769686706, | |
| "loss": 0.7572, | |
| "step": 20650 | |
| }, | |
| { | |
| "epoch": 8.73, | |
| "learning_rate": 0.00012722269263336155, | |
| "loss": 0.806, | |
| "step": 20675 | |
| }, | |
| { | |
| "epoch": 8.75, | |
| "learning_rate": 0.00012616426756985606, | |
| "loss": 0.7886, | |
| "step": 20700 | |
| }, | |
| { | |
| "epoch": 8.76, | |
| "learning_rate": 0.00012510584250635055, | |
| "loss": 0.8396, | |
| "step": 20725 | |
| }, | |
| { | |
| "epoch": 8.77, | |
| "learning_rate": 0.00012404741744284506, | |
| "loss": 0.7755, | |
| "step": 20750 | |
| }, | |
| { | |
| "epoch": 8.78, | |
| "learning_rate": 0.00012298899237933955, | |
| "loss": 0.8043, | |
| "step": 20775 | |
| }, | |
| { | |
| "epoch": 8.79, | |
| "learning_rate": 0.00012193056731583405, | |
| "loss": 0.7772, | |
| "step": 20800 | |
| }, | |
| { | |
| "epoch": 8.8, | |
| "learning_rate": 0.00012087214225232853, | |
| "loss": 0.7739, | |
| "step": 20825 | |
| }, | |
| { | |
| "epoch": 8.81, | |
| "learning_rate": 0.00011981371718882303, | |
| "loss": 0.7837, | |
| "step": 20850 | |
| }, | |
| { | |
| "epoch": 8.82, | |
| "learning_rate": 0.00011875529212531753, | |
| "loss": 0.8591, | |
| "step": 20875 | |
| }, | |
| { | |
| "epoch": 8.83, | |
| "learning_rate": 0.00011769686706181203, | |
| "loss": 0.817, | |
| "step": 20900 | |
| }, | |
| { | |
| "epoch": 8.84, | |
| "learning_rate": 0.00011663844199830652, | |
| "loss": 0.8072, | |
| "step": 20925 | |
| }, | |
| { | |
| "epoch": 8.85, | |
| "learning_rate": 0.00011558001693480103, | |
| "loss": 0.799, | |
| "step": 20950 | |
| }, | |
| { | |
| "epoch": 8.86, | |
| "learning_rate": 0.00011452159187129551, | |
| "loss": 0.8034, | |
| "step": 20975 | |
| }, | |
| { | |
| "epoch": 8.87, | |
| "learning_rate": 0.00011346316680779001, | |
| "loss": 0.7951, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 8.88, | |
| "learning_rate": 0.00011240474174428451, | |
| "loss": 0.7958, | |
| "step": 21025 | |
| }, | |
| { | |
| "epoch": 8.89, | |
| "learning_rate": 0.000111346316680779, | |
| "loss": 0.7567, | |
| "step": 21050 | |
| }, | |
| { | |
| "epoch": 8.9, | |
| "learning_rate": 0.0001102878916172735, | |
| "loss": 0.7791, | |
| "step": 21075 | |
| }, | |
| { | |
| "epoch": 8.91, | |
| "learning_rate": 0.00010922946655376798, | |
| "loss": 0.8189, | |
| "step": 21100 | |
| }, | |
| { | |
| "epoch": 8.92, | |
| "learning_rate": 0.0001081710414902625, | |
| "loss": 0.8048, | |
| "step": 21125 | |
| }, | |
| { | |
| "epoch": 8.94, | |
| "learning_rate": 0.00010711261642675698, | |
| "loss": 0.7884, | |
| "step": 21150 | |
| }, | |
| { | |
| "epoch": 8.95, | |
| "learning_rate": 0.00010605419136325148, | |
| "loss": 0.7908, | |
| "step": 21175 | |
| }, | |
| { | |
| "epoch": 8.96, | |
| "learning_rate": 0.00010499576629974598, | |
| "loss": 0.768, | |
| "step": 21200 | |
| }, | |
| { | |
| "epoch": 8.97, | |
| "learning_rate": 0.00010393734123624048, | |
| "loss": 0.811, | |
| "step": 21225 | |
| }, | |
| { | |
| "epoch": 8.98, | |
| "learning_rate": 0.00010287891617273497, | |
| "loss": 0.8204, | |
| "step": 21250 | |
| }, | |
| { | |
| "epoch": 8.99, | |
| "learning_rate": 0.00010182049110922948, | |
| "loss": 0.7989, | |
| "step": 21275 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 0.00010076206604572397, | |
| "loss": 0.7884, | |
| "step": 21300 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_loss": 1.1450048685073853, | |
| "eval_runtime": 4394.6356, | |
| "eval_samples_per_second": 1.075, | |
| "eval_steps_per_second": 0.134, | |
| "eval_wer": 30.689040929517024, | |
| "step": 21303 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 9.970364098221845e-05, | |
| "loss": 0.7619, | |
| "step": 21325 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 9.864521591871296e-05, | |
| "loss": 0.7994, | |
| "step": 21350 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 9.758679085520745e-05, | |
| "loss": 0.7808, | |
| "step": 21375 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "learning_rate": 9.652836579170195e-05, | |
| "loss": 0.8, | |
| "step": 21400 | |
| }, | |
| { | |
| "epoch": 9.05, | |
| "learning_rate": 9.546994072819644e-05, | |
| "loss": 0.7678, | |
| "step": 21425 | |
| }, | |
| { | |
| "epoch": 9.06, | |
| "learning_rate": 9.441151566469095e-05, | |
| "loss": 0.7871, | |
| "step": 21450 | |
| }, | |
| { | |
| "epoch": 9.07, | |
| "learning_rate": 9.335309060118543e-05, | |
| "loss": 0.7902, | |
| "step": 21475 | |
| }, | |
| { | |
| "epoch": 9.08, | |
| "learning_rate": 9.229466553767993e-05, | |
| "loss": 0.7675, | |
| "step": 21500 | |
| }, | |
| { | |
| "epoch": 9.09, | |
| "learning_rate": 9.123624047417443e-05, | |
| "loss": 0.7976, | |
| "step": 21525 | |
| }, | |
| { | |
| "epoch": 9.1, | |
| "learning_rate": 9.017781541066893e-05, | |
| "loss": 0.7772, | |
| "step": 21550 | |
| }, | |
| { | |
| "epoch": 9.11, | |
| "learning_rate": 8.911939034716342e-05, | |
| "loss": 0.7614, | |
| "step": 21575 | |
| }, | |
| { | |
| "epoch": 9.13, | |
| "learning_rate": 8.806096528365792e-05, | |
| "loss": 0.8134, | |
| "step": 21600 | |
| }, | |
| { | |
| "epoch": 9.14, | |
| "learning_rate": 8.700254022015242e-05, | |
| "loss": 0.7837, | |
| "step": 21625 | |
| }, | |
| { | |
| "epoch": 9.15, | |
| "learning_rate": 8.59441151566469e-05, | |
| "loss": 0.7952, | |
| "step": 21650 | |
| }, | |
| { | |
| "epoch": 9.16, | |
| "learning_rate": 8.488569009314142e-05, | |
| "loss": 0.8178, | |
| "step": 21675 | |
| }, | |
| { | |
| "epoch": 9.17, | |
| "learning_rate": 8.38272650296359e-05, | |
| "loss": 0.8111, | |
| "step": 21700 | |
| }, | |
| { | |
| "epoch": 9.18, | |
| "learning_rate": 8.27688399661304e-05, | |
| "loss": 0.7577, | |
| "step": 21725 | |
| }, | |
| { | |
| "epoch": 9.19, | |
| "learning_rate": 8.17104149026249e-05, | |
| "loss": 0.7944, | |
| "step": 21750 | |
| }, | |
| { | |
| "epoch": 9.2, | |
| "learning_rate": 8.06519898391194e-05, | |
| "loss": 0.8085, | |
| "step": 21775 | |
| }, | |
| { | |
| "epoch": 9.21, | |
| "learning_rate": 7.959356477561388e-05, | |
| "loss": 0.8087, | |
| "step": 21800 | |
| }, | |
| { | |
| "epoch": 9.22, | |
| "learning_rate": 7.853513971210838e-05, | |
| "loss": 0.787, | |
| "step": 21825 | |
| }, | |
| { | |
| "epoch": 9.23, | |
| "learning_rate": 7.747671464860288e-05, | |
| "loss": 0.7749, | |
| "step": 21850 | |
| }, | |
| { | |
| "epoch": 9.24, | |
| "learning_rate": 7.641828958509737e-05, | |
| "loss": 0.7469, | |
| "step": 21875 | |
| }, | |
| { | |
| "epoch": 9.25, | |
| "learning_rate": 7.535986452159187e-05, | |
| "loss": 0.7828, | |
| "step": 21900 | |
| }, | |
| { | |
| "epoch": 9.26, | |
| "learning_rate": 7.430143945808637e-05, | |
| "loss": 0.78, | |
| "step": 21925 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "learning_rate": 7.324301439458087e-05, | |
| "loss": 0.7895, | |
| "step": 21950 | |
| }, | |
| { | |
| "epoch": 9.28, | |
| "learning_rate": 7.218458933107535e-05, | |
| "loss": 0.8275, | |
| "step": 21975 | |
| }, | |
| { | |
| "epoch": 9.29, | |
| "learning_rate": 7.112616426756987e-05, | |
| "loss": 0.8019, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 9.31, | |
| "learning_rate": 7.006773920406435e-05, | |
| "loss": 0.7995, | |
| "step": 22025 | |
| }, | |
| { | |
| "epoch": 9.32, | |
| "learning_rate": 6.900931414055885e-05, | |
| "loss": 0.7857, | |
| "step": 22050 | |
| }, | |
| { | |
| "epoch": 9.33, | |
| "learning_rate": 6.795088907705335e-05, | |
| "loss": 0.7791, | |
| "step": 22075 | |
| }, | |
| { | |
| "epoch": 9.34, | |
| "learning_rate": 6.689246401354785e-05, | |
| "loss": 0.7538, | |
| "step": 22100 | |
| }, | |
| { | |
| "epoch": 9.35, | |
| "learning_rate": 6.583403895004234e-05, | |
| "loss": 0.7916, | |
| "step": 22125 | |
| }, | |
| { | |
| "epoch": 9.36, | |
| "learning_rate": 6.477561388653684e-05, | |
| "loss": 0.8062, | |
| "step": 22150 | |
| }, | |
| { | |
| "epoch": 9.37, | |
| "learning_rate": 6.371718882303133e-05, | |
| "loss": 0.7551, | |
| "step": 22175 | |
| }, | |
| { | |
| "epoch": 9.38, | |
| "learning_rate": 6.265876375952582e-05, | |
| "loss": 0.7946, | |
| "step": 22200 | |
| }, | |
| { | |
| "epoch": 9.39, | |
| "learning_rate": 6.160033869602032e-05, | |
| "loss": 0.8288, | |
| "step": 22225 | |
| }, | |
| { | |
| "epoch": 9.4, | |
| "learning_rate": 6.054191363251482e-05, | |
| "loss": 0.7838, | |
| "step": 22250 | |
| }, | |
| { | |
| "epoch": 9.41, | |
| "learning_rate": 5.948348856900931e-05, | |
| "loss": 0.7944, | |
| "step": 22275 | |
| }, | |
| { | |
| "epoch": 9.42, | |
| "learning_rate": 5.842506350550381e-05, | |
| "loss": 0.7861, | |
| "step": 22300 | |
| }, | |
| { | |
| "epoch": 9.43, | |
| "learning_rate": 5.7366638441998303e-05, | |
| "loss": 0.8022, | |
| "step": 22325 | |
| }, | |
| { | |
| "epoch": 9.44, | |
| "learning_rate": 5.63082133784928e-05, | |
| "loss": 0.7889, | |
| "step": 22350 | |
| }, | |
| { | |
| "epoch": 9.45, | |
| "learning_rate": 5.52497883149873e-05, | |
| "loss": 0.7619, | |
| "step": 22375 | |
| }, | |
| { | |
| "epoch": 9.46, | |
| "learning_rate": 5.4191363251481795e-05, | |
| "loss": 0.7825, | |
| "step": 22400 | |
| }, | |
| { | |
| "epoch": 9.47, | |
| "learning_rate": 5.3132938187976294e-05, | |
| "loss": 0.7598, | |
| "step": 22425 | |
| }, | |
| { | |
| "epoch": 9.48, | |
| "learning_rate": 5.2074513124470793e-05, | |
| "loss": 0.7659, | |
| "step": 22450 | |
| }, | |
| { | |
| "epoch": 9.5, | |
| "learning_rate": 5.1016088060965286e-05, | |
| "loss": 0.7651, | |
| "step": 22475 | |
| }, | |
| { | |
| "epoch": 9.51, | |
| "learning_rate": 4.9957662997459785e-05, | |
| "loss": 0.7848, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 9.52, | |
| "learning_rate": 4.889923793395427e-05, | |
| "loss": 0.7968, | |
| "step": 22525 | |
| }, | |
| { | |
| "epoch": 9.53, | |
| "learning_rate": 4.784081287044877e-05, | |
| "loss": 0.8155, | |
| "step": 22550 | |
| }, | |
| { | |
| "epoch": 9.54, | |
| "learning_rate": 4.678238780694327e-05, | |
| "loss": 0.8096, | |
| "step": 22575 | |
| }, | |
| { | |
| "epoch": 9.55, | |
| "learning_rate": 4.572396274343776e-05, | |
| "loss": 0.7779, | |
| "step": 22600 | |
| }, | |
| { | |
| "epoch": 9.56, | |
| "learning_rate": 4.466553767993226e-05, | |
| "loss": 0.7743, | |
| "step": 22625 | |
| }, | |
| { | |
| "epoch": 9.57, | |
| "learning_rate": 4.360711261642676e-05, | |
| "loss": 0.8104, | |
| "step": 22650 | |
| }, | |
| { | |
| "epoch": 9.58, | |
| "learning_rate": 4.2548687552921254e-05, | |
| "loss": 0.7797, | |
| "step": 22675 | |
| }, | |
| { | |
| "epoch": 9.59, | |
| "learning_rate": 4.149026248941575e-05, | |
| "loss": 0.7557, | |
| "step": 22700 | |
| }, | |
| { | |
| "epoch": 9.6, | |
| "learning_rate": 4.0431837425910246e-05, | |
| "loss": 0.8151, | |
| "step": 22725 | |
| }, | |
| { | |
| "epoch": 9.61, | |
| "learning_rate": 3.9373412362404745e-05, | |
| "loss": 0.8009, | |
| "step": 22750 | |
| }, | |
| { | |
| "epoch": 9.62, | |
| "learning_rate": 3.8314987298899244e-05, | |
| "loss": 0.7728, | |
| "step": 22775 | |
| }, | |
| { | |
| "epoch": 9.63, | |
| "learning_rate": 3.725656223539373e-05, | |
| "loss": 0.7901, | |
| "step": 22800 | |
| }, | |
| { | |
| "epoch": 9.64, | |
| "learning_rate": 3.619813717188823e-05, | |
| "loss": 0.8149, | |
| "step": 22825 | |
| }, | |
| { | |
| "epoch": 9.65, | |
| "learning_rate": 3.513971210838273e-05, | |
| "loss": 0.7773, | |
| "step": 22850 | |
| }, | |
| { | |
| "epoch": 9.66, | |
| "learning_rate": 3.408128704487722e-05, | |
| "loss": 0.811, | |
| "step": 22875 | |
| }, | |
| { | |
| "epoch": 9.67, | |
| "learning_rate": 3.302286198137172e-05, | |
| "loss": 0.7869, | |
| "step": 22900 | |
| }, | |
| { | |
| "epoch": 9.69, | |
| "learning_rate": 3.196443691786621e-05, | |
| "loss": 0.7734, | |
| "step": 22925 | |
| }, | |
| { | |
| "epoch": 9.7, | |
| "learning_rate": 3.090601185436071e-05, | |
| "loss": 0.7745, | |
| "step": 22950 | |
| }, | |
| { | |
| "epoch": 9.71, | |
| "learning_rate": 2.984758679085521e-05, | |
| "loss": 0.7964, | |
| "step": 22975 | |
| }, | |
| { | |
| "epoch": 9.72, | |
| "learning_rate": 2.87891617273497e-05, | |
| "loss": 0.8091, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 9.73, | |
| "learning_rate": 2.77307366638442e-05, | |
| "loss": 0.7633, | |
| "step": 23025 | |
| }, | |
| { | |
| "epoch": 9.74, | |
| "learning_rate": 2.6672311600338696e-05, | |
| "loss": 0.7806, | |
| "step": 23050 | |
| }, | |
| { | |
| "epoch": 9.75, | |
| "learning_rate": 2.5613886536833192e-05, | |
| "loss": 0.7932, | |
| "step": 23075 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "learning_rate": 2.455546147332769e-05, | |
| "loss": 0.7834, | |
| "step": 23100 | |
| }, | |
| { | |
| "epoch": 9.77, | |
| "learning_rate": 2.3497036409822188e-05, | |
| "loss": 0.7714, | |
| "step": 23125 | |
| }, | |
| { | |
| "epoch": 9.78, | |
| "learning_rate": 2.243861134631668e-05, | |
| "loss": 0.7877, | |
| "step": 23150 | |
| }, | |
| { | |
| "epoch": 9.79, | |
| "learning_rate": 2.1380186282811176e-05, | |
| "loss": 0.7596, | |
| "step": 23175 | |
| }, | |
| { | |
| "epoch": 9.8, | |
| "learning_rate": 2.0321761219305672e-05, | |
| "loss": 0.802, | |
| "step": 23200 | |
| }, | |
| { | |
| "epoch": 9.81, | |
| "learning_rate": 1.926333615580017e-05, | |
| "loss": 0.7799, | |
| "step": 23225 | |
| }, | |
| { | |
| "epoch": 9.82, | |
| "learning_rate": 1.8204911092294667e-05, | |
| "loss": 0.801, | |
| "step": 23250 | |
| }, | |
| { | |
| "epoch": 9.83, | |
| "learning_rate": 1.714648602878916e-05, | |
| "loss": 0.7705, | |
| "step": 23275 | |
| }, | |
| { | |
| "epoch": 9.84, | |
| "learning_rate": 1.6088060965283656e-05, | |
| "loss": 0.7636, | |
| "step": 23300 | |
| }, | |
| { | |
| "epoch": 9.85, | |
| "learning_rate": 1.5029635901778154e-05, | |
| "loss": 0.8273, | |
| "step": 23325 | |
| }, | |
| { | |
| "epoch": 9.86, | |
| "learning_rate": 1.3971210838272651e-05, | |
| "loss": 0.7904, | |
| "step": 23350 | |
| }, | |
| { | |
| "epoch": 9.88, | |
| "learning_rate": 1.2912785774767146e-05, | |
| "loss": 0.7859, | |
| "step": 23375 | |
| }, | |
| { | |
| "epoch": 9.89, | |
| "learning_rate": 1.1854360711261643e-05, | |
| "loss": 0.8122, | |
| "step": 23400 | |
| }, | |
| { | |
| "epoch": 9.9, | |
| "learning_rate": 1.079593564775614e-05, | |
| "loss": 0.7548, | |
| "step": 23425 | |
| }, | |
| { | |
| "epoch": 9.91, | |
| "learning_rate": 9.737510584250635e-06, | |
| "loss": 0.8049, | |
| "step": 23450 | |
| }, | |
| { | |
| "epoch": 9.92, | |
| "learning_rate": 8.679085520745131e-06, | |
| "loss": 0.8015, | |
| "step": 23475 | |
| }, | |
| { | |
| "epoch": 9.93, | |
| "learning_rate": 7.620660457239628e-06, | |
| "loss": 0.8026, | |
| "step": 23500 | |
| }, | |
| { | |
| "epoch": 9.94, | |
| "learning_rate": 6.562235393734124e-06, | |
| "loss": 0.7563, | |
| "step": 23525 | |
| }, | |
| { | |
| "epoch": 9.95, | |
| "learning_rate": 5.50381033022862e-06, | |
| "loss": 0.772, | |
| "step": 23550 | |
| }, | |
| { | |
| "epoch": 9.96, | |
| "learning_rate": 4.445385266723117e-06, | |
| "loss": 0.8002, | |
| "step": 23575 | |
| }, | |
| { | |
| "epoch": 9.97, | |
| "learning_rate": 3.386960203217612e-06, | |
| "loss": 0.7878, | |
| "step": 23600 | |
| }, | |
| { | |
| "epoch": 9.98, | |
| "learning_rate": 2.328535139712108e-06, | |
| "loss": 0.767, | |
| "step": 23625 | |
| }, | |
| { | |
| "epoch": 9.99, | |
| "learning_rate": 1.2701100762066045e-06, | |
| "loss": 0.802, | |
| "step": 23650 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_loss": 1.1429458856582642, | |
| "eval_runtime": 4707.3076, | |
| "eval_samples_per_second": 1.003, | |
| "eval_steps_per_second": 0.126, | |
| "eval_wer": 26.11837577426015, | |
| "step": 23670 | |
| } | |
| ], | |
| "logging_steps": 25, | |
| "max_steps": 23670, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 10, | |
| "save_steps": 500, | |
| "total_flos": 2.034797628063744e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |