{"lr": 2.3873462214411236e-05, "data_time": 0.08352530002593994, "grad_norm": 0.9491337060928344, "loss": 8.422504138946532, "time": 1.235518217086792, "epoch": 1, "step": 100}
{"lr": 4.783655536028112e-05, "data_time": 0.06839199066162109, "grad_norm": 1.0252236127853394, "loss": 8.124733829498291, "time": 1.1617755174636841, "epoch": 1, "step": 200}
{"lr": 7.179964850615093e-05, "data_time": 0.0872997760772705, "grad_norm": 1.3762734174728393, "loss": 7.98416543006897, "time": 1.1762173175811768, "epoch": 1, "step": 300}
{"lr": 9.576274165202062e-05, "data_time": 0.07882838249206543, "grad_norm": 1.218595600128174, "loss": 7.8940873622894285, "time": 1.150697612762451, "epoch": 1, "step": 400}
{"lr": 0.00011972583479789059, "data_time": 0.07924716472625733, "grad_norm": 1.7431656420230865, "loss": 7.836980104446411, "time": 1.1606164216995238, "epoch": 1, "step": 500}
{"lr": 0.00014368892794376054, "data_time": 0.08540904521942139, "grad_norm": 0.9570196390151977, "loss": 7.779762077331543, "time": 1.114922547340393, "epoch": 1, "step": 600}
{"lr": 0.00017388242530755644, "data_time": 0.08129973411560058, "grad_norm": 1.1496512234210967, "loss": 7.690268850326538, "time": 1.1390388488769532, "epoch": 2, "step": 726}
{"lr": 0.00019784551845342612, "data_time": 0.07606403827667237, "grad_norm": 1.0596347630023957, "loss": 7.594428539276123, "time": 1.128803014755249, "epoch": 2, "step": 826}
{"lr": 0.0002218086115992958, "data_time": 0.07229931354522705, "grad_norm": 1.6890751600265503, "loss": 7.595923709869385, "time": 1.1515222549438477, "epoch": 2, "step": 926}
{"lr": 0.00024577170474516564, "data_time": 0.08629395961761474, "grad_norm": 1.6247538626194, "loss": 7.565337181091309, "time": 1.1394296169281006, "epoch": 2, "step": 1026}
{"lr": 0.0002697347978910357, "data_time": 0.0834970235824585, "grad_norm": 1.804697149991989, "loss": 7.486994361877441, "time": 1.1214743614196778, "epoch": 2, "step": 1126}
{"lr": 0.0002936978910369062, "data_time": 0.08614318370819092, "grad_norm": 1.773136818408966, "loss": 7.436375761032105, "time": 1.118928861618042, "epoch": 2, "step": 1226}
{"lr": 0.0003238913884007022, "data_time": 0.08525383472442627, "grad_norm": 1.7598792672157288, "loss": 7.390654277801514, "time": 1.1966296911239624, "epoch": 3, "step": 1352}
{"lr": 0.000347854481546573, "data_time": 0.09460792541503907, "grad_norm": 1.6943426132202148, "loss": 7.281207847595215, "time": 1.182296347618103, "epoch": 3, "step": 1452}
{"lr": 0.00037181757469244325, "data_time": 0.07390902042388917, "grad_norm": 1.4433117032051086, "loss": 7.266968059539795, "time": 1.175376009941101, "epoch": 3, "step": 1552}
{"lr": 0.000395780667838314, "data_time": 0.09509601593017578, "grad_norm": 1.1911187648773194, "loss": 7.162929248809815, "time": 1.1650798082351685, "epoch": 3, "step": 1652}
{"lr": 0.0004197437609841844, "data_time": 0.07952842712402344, "grad_norm": 1.2938464105129241, "loss": 7.119700002670288, "time": 1.1464053869247437, "epoch": 3, "step": 1752}
{"lr": 0.00044370685413005464, "data_time": 0.08240573406219483, "grad_norm": 2.1488844990730285, "loss": 7.1791643619537355, "time": 1.1372448682785035, "epoch": 3, "step": 1852}
{"lr": 0.00047390035149385163, "data_time": 0.08197636604309082, "grad_norm": 1.4154476881027223, "loss": 7.071741819381714, "time": 1.2079867839813232, "epoch": 4, "step": 1978}
{"lr": 0.0004978634446397224, "data_time": 0.08393046855926514, "grad_norm": 1.7253454685211183, "loss": 7.006098127365112, "time": 1.1549755334854126, "epoch": 4, "step": 2078}
{"lr": 0.0005218265377855917, "data_time": 0.09145307540893555, "grad_norm": 1.5144005715847015, "loss": 6.9980944156646725, "time": 1.1680310010910033, "epoch": 4, "step": 2178}
{"lr": 0.0005457896309314621, "data_time": 0.09275951385498046, "grad_norm": 0.5779982656240463, "loss": 7.922142362594604, "time": 1.1724093198776244, "epoch": 4, "step": 2278}
{"lr": 0.0005697527240773326, "data_time": 0.07876980304718018, "grad_norm": 1.8099968552589416, "loss": 7.469259452819824, "time": 1.160907793045044, "epoch": 4, "step": 2378}
{"lr": 0.0005937158172232015, "data_time": 0.093389892578125, "grad_norm": 1.207740169763565, "loss": 7.065079689025879, "time": 1.1497637748718261, "epoch": 4, "step": 2478}
{"lr": 0.0006239093145869979, "data_time": 0.10479090213775635, "grad_norm": 1.1574186623096465, "loss": 6.966976356506348, "time": 1.1970284938812257, "epoch": 5, "step": 2604}
{"lr": 0.0006478724077328678, "data_time": 0.08908360004425049, "grad_norm": 1.5745812058448792, "loss": 6.934801959991455, "time": 1.1592916250228882, "epoch": 5, "step": 2704}
{"lr": 0.0006718355008787382, "data_time": 0.0901538610458374, "grad_norm": 2.184846878051758, "loss": 6.916006469726563, "time": 1.160930371284485, "epoch": 5, "step": 2804}
{"lr": 0.0006957985940246077, "data_time": 0.08781893253326416, "grad_norm": 1.492243218421936, "loss": 6.873108625411987, "time": 1.1554414987564088, "epoch": 5, "step": 2904}
{"lr": 0.0007197616871704767, "data_time": 0.08783724308013915, "grad_norm": 1.4368518769741059, "loss": 6.809550046920776, "time": 1.169290018081665, "epoch": 5, "step": 3004}
{"lr": 0.0007437247803163457, "data_time": 0.09047894477844239, "grad_norm": 1.2230656504631043, "loss": 6.77853422164917, "time": 1.1289460897445678, "epoch": 5, "step": 3104}
{"lr": 0.0007739182776801425, "data_time": 0.07091894149780273, "grad_norm": 1.4491536617279053, "loss": 6.75510082244873, "time": 1.1827389478683472, "epoch": 6, "step": 3230}
{"lr": 0.0007978813708260117, "data_time": 0.08156461715698242, "grad_norm": 1.2326496601104737, "loss": 6.686501836776733, "time": 1.1587774515151978, "epoch": 6, "step": 3330}
{"lr": 0.0008218444639718804, "data_time": 0.08585515022277831, "grad_norm": 1.5784627497196198, "loss": 6.683119010925293, "time": 1.1601664543151855, "epoch": 6, "step": 3430}
{"lr": 0.0008458075571177496, "data_time": 0.06818795204162598, "grad_norm": 1.750105482339859, "loss": 6.707097911834717, "time": 1.1424386501312256, "epoch": 6, "step": 3530}
{"lr": 0.0008697706502636198, "data_time": 0.09033935070037842, "grad_norm": 1.651491367816925, "loss": 6.690623569488525, "time": 1.1297106504440309, "epoch": 6, "step": 3630}
{"lr": 0.0008937337434094891, "data_time": 0.06459441184997558, "grad_norm": 1.9274215817451477, "loss": 6.60945143699646, "time": 1.1289054870605468, "epoch": 6, "step": 3730}
{"lr": 0.0009239272407732854, "data_time": 0.1075430154800415, "grad_norm": 1.2923402845859528, "loss": 6.580605220794678, "time": 1.1781140089035034, "epoch": 7, "step": 3856}
{"lr": 0.0009478903339191552, "data_time": 0.08731324672698974, "grad_norm": 1.27638099193573, "loss": 6.608668470382691, "time": 1.1270154953002929, "epoch": 7, "step": 3956}
{"lr": 0.0009718534270650264, "data_time": 0.10752544403076172, "grad_norm": 2.344351029396057, "loss": 6.66406946182251, "time": 1.1375146389007569, "epoch": 7, "step": 4056}
{"lr": 0.0009958165202108966, "data_time": 0.07816410064697266, "grad_norm": 0.4073693215847015, "loss": 7.9008176803588865, "time": 1.12248055934906, "epoch": 7, "step": 4156}
{"lr": 0.0010197796133567668, "data_time": 0.11899840831756592, "grad_norm": 0.6388155609369278, "loss": 7.580292415618897, "time": 1.157643723487854, "epoch": 7, "step": 4256}
{"lr": 0.0010437427065026364, "data_time": 0.10139455795288085, "grad_norm": 1.0905654311180115, "loss": 7.230022811889649, "time": 1.153193974494934, "epoch": 7, "step": 4356}
{"lr": 0.001073936203866432, "data_time": 0.06987209320068359, "grad_norm": 1.1944232642650605, "loss": 6.7262852668762205, "time": 1.1908584117889405, "epoch": 8, "step": 4482}
{"lr": 0.0010978992970123, "data_time": 0.0853029727935791, "grad_norm": 1.2120230555534364, "loss": 6.612608766555786, "time": 1.1928131818771361, "epoch": 8, "step": 4582}
{"lr": 0.0011218623901581698, "data_time": 0.07176795005798339, "grad_norm": 0.8716796576976776, "loss": 6.549582052230835, "time": 1.178091335296631, "epoch": 8, "step": 4682}
{"lr": 0.0011458254833040375, "data_time": 0.08355410099029541, "grad_norm": 0.8115405440330505, "loss": 6.518727493286133, "time": 1.1352470159530639, "epoch": 8, "step": 4782}
{"lr": 0.0011697885764499073, "data_time": 0.07907819747924805, "grad_norm": 0.8864546716213226, "loss": 6.54967041015625, "time": 1.1354489088058473, "epoch": 8, "step": 4882}
{"lr": 0.0011937516695957773, "data_time": 0.09167776107788086, "grad_norm": 0.744516259431839, "loss": 6.476688003540039, "time": 1.1418145179748536, "epoch": 8, "step": 4982}
{"lr": 0.0012239451669595713, "data_time": 0.06934914588928223, "grad_norm": 0.84982950091362, "loss": 6.486936616897583, "time": 1.1931581020355224, "epoch": 9, "step": 5108}
{"lr": 0.0012479082601054388, "data_time": 0.07826032638549804, "grad_norm": 1.0509164452552795, "loss": 6.417401647567749, "time": 1.1394390344619751, "epoch": 9, "step": 5208}
{"lr": 0.0012718713532513077, "data_time": 0.06880099773406982, "grad_norm": 1.2428844571113586, "loss": 6.479604768753052, "time": 1.15473313331604, "epoch": 9, "step": 5308}
{"lr": 0.0012958344463971767, "data_time": 0.09171507358551026, "grad_norm": 0.6748064130544662, "loss": 6.381825685501099, "time": 1.1391008615493774, "epoch": 9, "step": 5408}
{"lr": 0.0013197975395430462, "data_time": 0.0896329402923584, "grad_norm": 0.5672484636306763, "loss": 7.552138996124268, "time": 1.169378662109375, "epoch": 9, "step": 5508}
{"lr": 0.0013437606326889137, "data_time": 0.08891284465789795, "grad_norm": 0.9085614174604416, "loss": 6.718552827835083, "time": 1.1262605667114258, "epoch": 9, "step": 5608}
{"lr": 0.001373954130052708, "data_time": 0.06650779247283936, "grad_norm": 0.7380014926195144, "loss": 6.4685924530029295, "time": 1.1725317001342774, "epoch": 10, "step": 5734}
{"lr": 0.0013979172231985753, "data_time": 0.07251255512237549, "grad_norm": 0.7263779580593109, "loss": 6.488245582580566, "time": 1.1279207229614259, "epoch": 10, "step": 5834}
{"lr": 0.0014218803163444406, "data_time": 0.06630632877349854, "grad_norm": 0.8867326319217682, "loss": 6.435011672973633, "time": 1.1490319490432739, "epoch": 10, "step": 5934}
{"lr": 0.0014458434094903065, "data_time": 0.07067465782165527, "grad_norm": 1.0398584067821504, "loss": 6.453606510162354, "time": 1.122061061859131, "epoch": 10, "step": 6034}
{"lr": 0.0014698065026361755, "data_time": 0.0714648962020874, "grad_norm": 0.8326262831687927, "loss": 6.390722179412842, "time": 1.1302236795425415, "epoch": 10, "step": 6134}
{"lr": 0.001493769595782043, "data_time": 0.07692699432373047, "grad_norm": 0.6954630970954895, "loss": 6.406313753128051, "time": 1.1329723119735717, "epoch": 10, "step": 6234}
{"lr": 0.0014999989066690166, "data_time": 0.09146802425384522, "grad_norm": 0.8332736551761627, "loss": 6.373015737533569, "time": 1.160567307472229, "epoch": 11, "step": 6360}
{"lr": 0.001499995582392903, "data_time": 0.10117838382720948, "grad_norm": 0.9660639464855194, "loss": 6.38512020111084, "time": 1.1170629501342773, "epoch": 11, "step": 6460}
{"lr": 0.0014999900270700323, "data_time": 0.10570683479309081, "grad_norm": 0.5873611718416214, "loss": 6.374302721023559, "time": 1.1624947309494018, "epoch": 11, "step": 6560}
{"lr": 0.0014999822407170394, "data_time": 0.08980019092559814, "grad_norm": 0.5932401299476624, "loss": 6.32376275062561, "time": 1.132708239555359, "epoch": 11, "step": 6660}
{"lr": 0.0014999722233572427, "data_time": 0.09681766033172608, "grad_norm": 0.9979702889919281, "loss": 6.3630458354949955, "time": 1.1404176235198975, "epoch": 11, "step": 6760}
{"lr": 0.0014999599750206418, "data_time": 0.09852590560913085, "grad_norm": 0.9182680368423461, "loss": 7.019189691543579, "time": 1.1296245098114013, "epoch": 11, "step": 6860}
{"lr": 0.0014999413657091737, "data_time": 0.06663353443145752, "grad_norm": 0.7669404149055481, "loss": 6.38889479637146, "time": 1.1587861776351929, "epoch": 12, "step": 6986}
{"lr": 0.0014999240755104548, "data_time": 0.10330924987792969, "grad_norm": 0.8277837365865708, "loss": 6.43386549949646, "time": 1.147247886657715, "epoch": 12, "step": 7086}
{"lr": 0.0014999045544791193, "data_time": 0.08282666206359864, "grad_norm": 0.6773287385702134, "loss": 6.363525915145874, "time": 1.1460523843765258, "epoch": 12, "step": 7186}
{"lr": 0.001499882802673627, "data_time": 0.0856102705001831, "grad_norm": 0.6188061267137528, "loss": 6.33180866241455, "time": 1.1407484769821168, "epoch": 12, "step": 7286}
{"lr": 0.0014998588201591198, "data_time": 0.06689550876617431, "grad_norm": 0.8442979246377945, "loss": 6.28070707321167, "time": 1.1157282829284667, "epoch": 12, "step": 7386}
{"lr": 0.0014998326070074187, "data_time": 0.08107988834381104, "grad_norm": 0.47343507409095764, "loss": 6.300201797485352, "time": 1.1271000623703002, "epoch": 12, "step": 7486}
{"lr": 0.0014997964025771629, "data_time": 0.11007838249206543, "grad_norm": 0.6555793851613998, "loss": 6.286029243469239, "time": 1.1757495403289795, "epoch": 13, "step": 7612}
{"lr": 0.0014997651484849165, "data_time": 0.10198209285736085, "grad_norm": 0.6776884019374847, "loss": 6.262293958663941, "time": 1.1306779861450196, "epoch": 13, "step": 7712}
{"lr": 0.0014997316640359977, "data_time": 0.09965245723724366, "grad_norm": 0.4601239025592804, "loss": 6.305837726593017, "time": 1.1311154127120973, "epoch": 13, "step": 7812}
{"lr": 0.0014996959493306786, "data_time": 0.07860772609710694, "grad_norm": 0.49628026187419894, "loss": 6.323536968231201, "time": 1.1013127565383911, "epoch": 13, "step": 7912}
{"lr": 0.001499658004475916, "data_time": 0.08457729816436768, "grad_norm": 0.8656015217304229, "loss": 6.311141586303711, "time": 1.1168241024017334, "epoch": 13, "step": 8012}
{"lr": 0.001499617829585343, "data_time": 0.0897716760635376, "grad_norm": 0.5697761565446854, "loss": 6.300175189971924, "time": 1.0777748346328735, "epoch": 13, "step": 8112}
{"lr": 0.001499564034285053, "data_time": 0.06569252014160157, "grad_norm": 0.7605577945709229, "loss": 6.306610059738159, "time": 1.1212515830993652, "epoch": 14, "step": 8238}
{"lr": 0.0014995188199670897, "data_time": 0.07652528285980224, "grad_norm": 0.9333890199661254, "loss": 6.264597940444946, "time": 1.127130913734436, "epoch": 14, "step": 8338}
{"lr": 0.001499471376030133, "data_time": 0.07108972072601319, "grad_norm": 0.8565931469202042, "loss": 6.242331504821777, "time": 1.122644853591919, "epoch": 14, "step": 8438}
{"lr": 0.001499421702616265, "data_time": 0.08680391311645508, "grad_norm": 0.5608468681573868, "loss": 6.197241973876953, "time": 1.1517184495925903, "epoch": 14, "step": 8538}
{"lr": 0.0014993697998742438, "data_time": 0.06710593700408936, "grad_norm": 0.581649661064148, "loss": 7.771830463409424, "time": 1.116100287437439, "epoch": 14, "step": 8638}
{"lr": 0.0014993156679595017, "data_time": 0.0699007511138916, "grad_norm": 0.4318750649690628, "loss": 6.463739919662475, "time": 1.1104565620422364, "epoch": 14, "step": 8738}
{"lr": 0.0014992442881022925, "data_time": 0.11259334087371826, "grad_norm": 0.6225933045148849, "loss": 6.370961046218872, "time": 1.1936349630355836, "epoch": 15, "step": 8864}
{"lr": 0.0014991851188647098, "data_time": 0.08941986560821533, "grad_norm": 0.48378168046474457, "loss": 6.235713815689087, "time": 1.1385921716690064, "epoch": 15, "step": 8964}
{"lr": 0.0014991237210074731, "data_time": 0.08316495418548583, "grad_norm": 0.46330522894859316, "loss": 6.2788733959198, "time": 1.1512133598327636, "epoch": 15, "step": 9064}
{"lr": 0.001499060094714452, "data_time": 0.09332852363586426, "grad_norm": 0.5130493104457855, "loss": 6.276291561126709, "time": 1.1643208265304565, "epoch": 15, "step": 9164}
{"lr": 0.0014989942401761863, "data_time": 0.09918458461761474, "grad_norm": 0.7988549023866653, "loss": 6.3163947582244875, "time": 1.1695189714431762, "epoch": 15, "step": 9264}
{"lr": 0.001498926157589896, "data_time": 0.10827322006225586, "grad_norm": 0.6081429287791252, "loss": 6.263222408294678, "time": 1.1914340734481812, "epoch": 15, "step": 9364}
{"lr": 0.0014988372015525126, "data_time": 0.06811702251434326, "grad_norm": 0.49149343371391296, "loss": 6.276322650909424, "time": 1.1658409595489503, "epoch": 16, "step": 9490}
{"lr": 0.0014987640843390825, "data_time": 0.07955915927886963, "grad_norm": 0.4809699594974518, "loss": 6.267184591293335, "time": 1.1554269552230836, "epoch": 16, "step": 9590}
{"lr": 0.0014986887397668756, "data_time": 0.08222134113311767, "grad_norm": 0.56675805747509, "loss": 6.219944906234741, "time": 1.193008542060852, "epoch": 16, "step": 9690}
{"lr": 0.0014986111680615273, "data_time": 0.08803200721740723, "grad_norm": 0.4352340489625931, "loss": 6.203954410552979, "time": 1.1871110200881958, "epoch": 16, "step": 9790}
{"lr": 0.0014985313694553437, "data_time": 0.0662491798400879, "grad_norm": 0.5146336376667022, "loss": 6.230696296691894, "time": 1.1644105911254883, "epoch": 16, "step": 9890}
{"lr": 0.0014984493441872954, "data_time": 0.0920952558517456, "grad_norm": 0.44286521077156066, "loss": 6.220179653167724, "time": 1.2241258382797242, "epoch": 16, "step": 9990}
{"lr": 0.0014983428224091219, "data_time": 0.12128734588623047, "grad_norm": 0.596603399515152, "loss": 6.2141844749450685, "time": 1.1643854141235352, "epoch": 17, "step": 10116}
{"lr": 0.0014982557658004749, "data_time": 0.08451380729675292, "grad_norm": 0.9898445338010788, "loss": 6.231795167922973, "time": 1.136660647392273, "epoch": 17, "step": 10216}
{"lr": 0.0014981664833553184, "data_time": 0.10107331275939942, "grad_norm": 0.5849262058734894, "loss": 6.249002838134766, "time": 1.1523273706436157, "epoch": 17, "step": 10316}
{"lr": 0.0014980749753410274, "data_time": 0.09699087142944336, "grad_norm": 0.7859920799732208, "loss": 6.174241590499878, "time": 1.1548095703125, "epoch": 17, "step": 10416}
{"lr": 0.0014979812420316423, "data_time": 0.09748828411102295, "grad_norm": 0.7911879509687424, "loss": 6.2300276279449465, "time": 1.160532331466675, "epoch": 17, "step": 10516}
{"lr": 0.001497885283707868, "data_time": 0.08352427482604981, "grad_norm": 0.648396971821785, "loss": 6.242800378799439, "time": 1.1280606269836426, "epoch": 17, "step": 10616}
{"lr": 0.001497761208689706, "data_time": 0.07784018516540528, "grad_norm": 0.4547905385494232, "loss": 6.199664211273193, "time": 1.139499044418335, "epoch": 18, "step": 10742}
{"lr": 0.001497660222902321, "data_time": 0.06877262592315674, "grad_norm": 0.6394618898630142, "loss": 6.17699875831604, "time": 1.1022571802139283, "epoch": 18, "step": 10842}
{"lr": 0.0014975570130619075, "data_time": 0.08522975444793701, "grad_norm": 0.5263616293668747, "loss": 6.205042886734009, "time": 1.1141348838806153, "epoch": 18, "step": 10942}
{"lr": 0.0014974515794775457, "data_time": 0.0717963457107544, "grad_norm": 0.46777002811431884, "loss": 6.1868198871612545, "time": 1.1168910264968872, "epoch": 18, "step": 11042}
{"lr": 0.0014973439224649809, "data_time": 0.09166765213012695, "grad_norm": 0.5990155726671219, "loss": 6.196297550201416, "time": 1.1442065477371215, "epoch": 18, "step": 11142}
{"lr": 0.0014972340423466135, "data_time": 0.06580228805541992, "grad_norm": 0.6317641586065292, "loss": 6.241032648086548, "time": 1.1324118614196776, "epoch": 18, "step": 11242}
{"lr": 0.0014970924286492194, "data_time": 0.0978384017944336, "grad_norm": 0.903348171710968, "loss": 6.163993501663208, "time": 1.1663374185562134, "epoch": 19, "step": 11368}
{"lr": 0.0014969775255342273, "data_time": 0.08403115272521973, "grad_norm": 0.7913348972797394, "loss": 6.193544149398804, "time": 1.1454696178436279, "epoch": 19, "step": 11468}
{"lr": 0.0014968604004106866, "data_time": 0.11719892024993897, "grad_norm": 1.026800599694252, "loss": 6.215503931045532, "time": 1.1661480903625487, "epoch": 19, "step": 11568}
{"lr": 0.0014967410536293525, "data_time": 0.07947707176208496, "grad_norm": 0.8300428956747055, "loss": 6.18181414604187, "time": 1.139579701423645, "epoch": 19, "step": 11668}
{"lr": 0.0014966194855476347, "data_time": 0.10365231037139892, "grad_norm": 0.6439772099256516, "loss": 6.187466239929199, "time": 1.1330519914627075, "epoch": 19, "step": 11768}
{"lr": 0.0014964956965295938, "data_time": 0.09457776546478272, "grad_norm": 0.38432839065790175, "loss": 6.139675331115723, "time": 1.1203187704086304, "epoch": 19, "step": 11868}
{"lr": 0.0014963365607719584, "data_time": 0.07452583312988281, "grad_norm": 0.3793606847524643, "loss": 6.177168273925782, "time": 1.1818388938903808, "epoch": 20, "step": 11994}
{"lr": 0.0014962077538137476, "data_time": 0.09340143203735352, "grad_norm": 0.7385234534740448, "loss": 6.168047094345093, "time": 1.1579730749130248, "epoch": 20, "step": 12094}
{"lr": 0.0014960767271522324, "data_time": 0.08462967872619628, "grad_norm": 0.7607008665800095, "loss": 6.192141389846801, "time": 1.18638916015625, "epoch": 20, "step": 12194}
{"lr": 0.0014959434811797983, "data_time": 0.07921977043151855, "grad_norm": 0.5204597979784011, "loss": 6.208904266357422, "time": 1.128187894821167, "epoch": 20, "step": 12294}
{"lr": 0.0014958080162954797, "data_time": 0.064290452003479, "grad_norm": 0.32701108157634734, "loss": 6.181735706329346, "time": 1.1598257303237915, "epoch": 20, "step": 12394}
{"lr": 0.0014956703329049553, "data_time": 0.09758782386779785, "grad_norm": 0.8163347750902176, "loss": 6.207579469680786, "time": 1.128312849998474, "epoch": 20, "step": 12494}
{"lr": 0.001495493693762371, "data_time": 0.08713295459747314, "grad_norm": 0.4588005855679512, "loss": 6.149515390396118, "time": 1.1662253379821776, "epoch": 21, "step": 12620}
{"lr": 0.0014953509980770113, "data_time": 0.08324322700500489, "grad_norm": 0.585504412651062, "loss": 6.179350757598877, "time": 1.1228465795516969, "epoch": 21, "step": 12720}
{"lr": 0.0014952060852540816, "data_time": 0.08201413154602051, "grad_norm": 0.4836617648601532, "loss": 6.167264747619629, "time": 1.1392805099487304, "epoch": 21, "step": 12820}
{"lr": 0.0014950589557275581, "data_time": 0.07330138683319092, "grad_norm": 0.4596558779478073, "loss": 6.156923866271972, "time": 1.1321951866149902, "epoch": 21, "step": 12920}
{"lr": 0.001494909609938047, "data_time": 0.08830389976501465, "grad_norm": 0.4111696928739548, "loss": 6.180008363723755, "time": 1.1466331005096435, "epoch": 21, "step": 13020}
{"lr": 0.0014947580483327964, "data_time": 0.07420890331268311, "grad_norm": 0.6491823256015777, "loss": 6.119488763809204, "time": 1.1177634477615357, "epoch": 21, "step": 13120}
{"lr": 0.0014945639265346463, "data_time": 0.07477478981018067, "grad_norm": 0.5648597061634064, "loss": 6.146530675888061, "time": 1.1680334091186524, "epoch": 22, "step": 13246}
{"lr": 0.0014944073588681061, "data_time": 0.08663275241851806, "grad_norm": 0.4902139902114868, "loss": 6.183145427703858, "time": 1.1600675106048584, "epoch": 22, "step": 13346}
{"lr": 0.001494248576889925, "data_time": 0.08145263195037841, "grad_norm": 0.7846917003393173, "loss": 6.173889446258545, "time": 1.1774061441421508, "epoch": 22, "step": 13446}
{"lr": 0.0014940875810756076, "data_time": 0.08086488246917725, "grad_norm": 0.3863218933343887, "loss": 6.162805366516113, "time": 1.1162629127502441, "epoch": 22, "step": 13546}
{"lr": 0.0014939243719072925, "data_time": 0.07562203407287597, "grad_norm": 0.5077043056488038, "loss": 6.181052303314209, "time": 1.1389439582824707, "epoch": 22, "step": 13646}
{"lr": 0.0014937589498737428, "data_time": 0.07965028285980225, "grad_norm": 0.6589199870824813, "loss": 6.1324381828308105, "time": 1.1279481410980225, "epoch": 22, "step": 13746}
{"lr": 0.0014935473682010805, "data_time": 0.10892491340637207, "grad_norm": 0.5346403151750565, "loss": 6.1291868686676025, "time": 1.1533003807067872, "epoch": 23, "step": 13872}
{"lr": 0.0014933769469272706, "data_time": 0.09201602935791016, "grad_norm": 0.522069451212883, "loss": 6.08466215133667, "time": 1.1462144136428833, "epoch": 23, "step": 13972}
{"lr": 0.0014932043144276067, "data_time": 0.09041423797607422, "grad_norm": 0.46865294575691224, "loss": 6.102357816696167, "time": 1.1385055541992188, "epoch": 23, "step": 14072}
{"lr": 0.0014930294712190725, "data_time": 0.07521750926971435, "grad_norm": 0.29599695205688475, "loss": 6.1437901020050045, "time": 1.1166694402694701, "epoch": 23, "step": 14172}
{"lr": 0.0014928524178252729, "data_time": 0.10121920108795165, "grad_norm": 0.6619588285684586, "loss": 6.0747847080230715, "time": 1.137457799911499, "epoch": 23, "step": 14272}
{"lr": 0.0014926731547764299, "data_time": 0.0766575813293457, "grad_norm": 0.4733997881412506, "loss": 6.137448692321778, "time": 1.117971134185791, "epoch": 23, "step": 14372}
{"lr": 0.0014924441380593035, "data_time": 0.07296874523162841, "grad_norm": 0.9791599750518799, "loss": 6.141348791122437, "time": 1.1589645147323608, "epoch": 24, "step": 14498}
{"lr": 0.0014922598831779168, "data_time": 0.09020373821258545, "grad_norm": 0.44596984684467317, "loss": 6.148757791519165, "time": 1.1322027683258056, "epoch": 24, "step": 14598}
{"lr": 0.0014920734204159592, "data_time": 0.07470028400421143, "grad_norm": 0.5245147794485092, "loss": 6.0847584247589115, "time": 1.1591213941574097, "epoch": 24, "step": 14698}
{"lr": 0.001491884750331832, "data_time": 0.08674097061157227, "grad_norm": 0.6525487154722214, "loss": 6.145807218551636, "time": 1.1263118505477905, "epoch": 24, "step": 14798}
{"lr": 0.001491693873490548, "data_time": 0.06831791400909423, "grad_norm": 0.5651695400476455, "loss": 6.139700222015381, "time": 1.133176851272583, "epoch": 24, "step": 14898}
{"lr": 0.0014915007904637279, "data_time": 0.08161988258361816, "grad_norm": 0.48039994239807127, "loss": 6.100747299194336, "time": 1.122533392906189, "epoch": 24, "step": 14998}
{"lr": 0.0014912543655782636, "data_time": 0.11105420589447021, "grad_norm": 0.3961913675069809, "loss": 6.115591287612915, "time": 1.1792709589004517, "epoch": 25, "step": 15124}
{"lr": 0.0014910562987124293, "data_time": 0.08232665061950684, "grad_norm": 0.4115993857383728, "loss": 6.095423746109009, "time": 1.162469458580017, "epoch": 25, "step": 15224}
{"lr": 0.0014908560275704087, "data_time": 0.09179129600524902, "grad_norm": 0.38940606117248533, "loss": 6.139225006103516, "time": 1.1468290090560913, "epoch": 25, "step": 15324}
{"lr": 0.0014906535527519581, "data_time": 0.08297829627990723, "grad_norm": 0.4618272572755814, "loss": 6.073189735412598, "time": 1.149387526512146, "epoch": 25, "step": 15424}
{"lr": 0.0014904488748634338, "data_time": 0.10364751815795899, "grad_norm": 0.33043105751276014, "loss": 6.116808605194092, "time": 1.1626547813415526, "epoch": 25, "step": 15524}
{"lr": 0.001490241994517785, "data_time": 0.08972351551055908, "grad_norm": 0.411043556034565, "loss": 6.086409282684326, "time": 1.1648491144180297, "epoch": 25, "step": 15624}
{"lr": 0.001489978190383037, "data_time": 0.09162304401397706, "grad_norm": 0.506519815325737, "loss": 6.118365383148193, "time": 1.137656855583191, "epoch": 26, "step": 15750}
{"lr": 0.001489766334776776, "data_time": 0.09048628807067871, "grad_norm": 0.6475946187973023, "loss": 6.113751792907715, "time": 1.1282639503479004, "epoch": 26, "step": 15850}
{"lr": 0.0014895522787574, "data_time": 0.0920987606048584, "grad_norm": 0.6032128751277923, "loss": 6.165223836898804, "time": 1.1359817504882812, "epoch": 26, "step": 15950}
{"lr": 0.0014893360229659481, "data_time": 0.08976423740386963, "grad_norm": 0.6529540330171585, "loss": 6.145842504501343, "time": 1.1316508293151855, "epoch": 26, "step": 16050}
{"lr": 0.0014891175680500413, "data_time": 0.10529177188873291, "grad_norm": 0.40408940613269806, "loss": 6.115272521972656, "time": 1.137672758102417, "epoch": 26, "step": 16150}
{"lr": 0.0014888969146638902, "data_time": 0.0825228214263916, "grad_norm": 0.738173657655716, "loss": 6.136606454849243, "time": 1.137540864944458, "epoch": 26, "step": 16250}
{"lr": 0.0014886157622384423, "data_time": 0.10415506362915039, "grad_norm": 0.608621346950531, "loss": 6.13971962928772, "time": 1.1742061376571655, "epoch": 27, "step": 16376}
{"lr": 0.001488390142753951, "data_time": 0.10819940567016602, "grad_norm": 0.3293007656931877, "loss": 6.112208414077759, "time": 1.1146725416183472, "epoch": 27, "step": 16476}
{"lr": 0.001488162326977644, "data_time": 0.09398629665374755, "grad_norm": 0.4459797561168671, "loss": 6.111444330215454, "time": 1.1380535364151, "epoch": 27, "step": 16576}
{"lr": 0.0014879323155917628, "data_time": 0.09144294261932373, "grad_norm": 0.5585973918437958, "loss": 6.078958559036255, "time": 1.1135217428207398, "epoch": 27, "step": 16676}
{"lr": 0.0014877001092851278, "data_time": 0.11254184246063233, "grad_norm": 0.45779127180576323, "loss": 6.050326728820801, "time": 1.1436564445495605, "epoch": 27, "step": 16776}
{"lr": 0.001487465708753127, "data_time": 0.08175952434539795, "grad_norm": 0.4321734294295311, "loss": 6.101701307296753, "time": 1.110058641433716, "epoch": 27, "step": 16876}
{"lr": 0.0014871672410314656, "data_time": 0.07508985996246338, "grad_norm": 0.3624123454093933, "loss": 6.112020635604859, "time": 1.1410366296768188, "epoch": 28, "step": 17002}
{"lr": 0.0014869278841461889, "data_time": 0.08388056755065917, "grad_norm": 0.5053034156560898, "loss": 6.127461290359497, "time": 1.131105399131775, "epoch": 28, "step": 17102}
{"lr": 0.0014866863353481404, "data_time": 0.07023928165435792, "grad_norm": 0.4168893352150917, "loss": 6.082732248306274, "time": 1.129146409034729, "epoch": 28, "step": 17202}
{"lr": 0.0014864425953606873, "data_time": 0.08972556591033935, "grad_norm": 0.9394630432128906, "loss": 6.637428331375122, "time": 1.1433965921401978, "epoch": 28, "step": 17302}
{"lr": 0.0014861966649137612, "data_time": 0.08352999687194824, "grad_norm": 0.37108329981565474, "loss": 6.185878086090088, "time": 1.158143663406372, "epoch": 28, "step": 17402}
{"lr": 0.001485948544743852, "data_time": 0.08605155944824219, "grad_norm": 0.4544988438487053, "loss": 6.129090976715088, "time": 1.121772336959839, "epoch": 28, "step": 17502}
{"lr": 0.0014856327967524838, "data_time": 0.10690615177154542, "grad_norm": 0.2783099591732025, "loss": 6.105822372436523, "time": 1.153731679916382, "epoch": 29, "step": 17628}
{"lr": 0.0014853797305560178, "data_time": 0.09201564788818359, "grad_norm": 0.40010859072208405, "loss": 6.130463600158691, "time": 1.13615825176239, "epoch": 29, "step": 17728}
{"lr": 0.0014851244770830517, "data_time": 0.10367422103881836, "grad_norm": 0.6202672123908997, "loss": 6.0300538539886475, "time": 1.1481379985809326, "epoch": 29, "step": 17828}
{"lr": 0.0014848670370979965, "data_time": 0.08277318477630616, "grad_norm": 0.4914248138666153, "loss": 6.070322942733765, "time": 1.1182928085327148, "epoch": 29, "step": 17928}
{"lr": 0.0014846074113718088, "data_time": 0.09295060634613037, "grad_norm": 0.3367126554250717, "loss": 6.073860120773316, "time": 1.1351324796676636, "epoch": 29, "step": 18028}
{"lr": 0.0014843456006819937, "data_time": 0.08268930912017822, "grad_norm": 0.5051405265927315, "loss": 6.062775897979736, "time": 1.1131535053253174, "epoch": 29, "step": 18128}
{"lr": 0.001484012609475349, "data_time": 0.09841411113739014, "grad_norm": 0.7575819164514541, "loss": 6.092672872543335, "time": 1.1835420608520508, "epoch": 30, "step": 18254}
{"lr": 0.0014837458636661309, "data_time": 0.06684777736663819, "grad_norm": 0.39526701271533965, "loss": 6.062423467636108, "time": 1.1302995681762695, "epoch": 30, "step": 18354}
{"lr": 0.0014834769354733737, "data_time": 0.09677374362945557, "grad_norm": 0.47108394652605057, "loss": 6.064792633056641, "time": 1.179744791984558, "epoch": 30, "step": 18454}
{"lr": 0.001483205825702436, "data_time": 0.07988593578338624, "grad_norm": 0.5180205494165421, "loss": 6.111719608306885, "time": 1.143750834465027, "epoch": 30, "step": 18554}
{"lr": 0.0014829325351652174, "data_time": 0.08844974040985107, "grad_norm": 0.3161155849695206, "loss": 6.069922685623169, "time": 1.1517072439193725, "epoch": 30, "step": 18654}
{"lr": 0.0014826570646801413, "data_time": 0.08130838871002197, "grad_norm": 0.40873357355594636, "loss": 6.089505672454834, "time": 1.1340928554534913, "epoch": 30, "step": 18754}
{"lr": 0.0014823068693362107, "data_time": 0.09799575805664062, "grad_norm": 0.42539741694927213, "loss": 6.06600866317749, "time": 1.1597592115402222, "epoch": 31, "step": 18880}
{"lr": 0.0014820264752180442, "data_time": 0.08825931549072266, "grad_norm": 0.34834582507610323, "loss": 6.083920288085937, "time": 1.15988028049469, "epoch": 31, "step": 18980}
{"lr": 0.00148174390386541, "data_time": 0.12088091373443603, "grad_norm": 0.6790808573365211, "loss": 6.03754415512085, "time": 1.1740911483764649, "epoch": 31, "step": 19080}
{"lr": 0.0014814591561245277, "data_time": 0.09686169624328614, "grad_norm": 0.41202564984560014, "loss": 6.051353502273559, "time": 1.1312536716461181, "epoch": 31, "step": 19180}
{"lr": 0.0014811722328481327, "data_time": 0.09337136745452881, "grad_norm": 0.6329362690448761, "loss": 6.093054533004761, "time": 1.1447856664657592, "epoch": 31, "step": 19280}
{"lr": 0.001480883134895481, "data_time": 0.10069646835327148, "grad_norm": 0.3367758959531784, "loss": 6.099015617370606, "time": 1.1486722707748414, "epoch": 31, "step": 19380}
{"lr": 0.001480515776511244, "data_time": 0.08479495048522949, "grad_norm": 0.4227337047457695, "loss": 6.038627052307129, "time": 1.1778574228286742, "epoch": 32, "step": 19506}
{"lr": 0.0014802217669896234, "data_time": 0.07217133045196533, "grad_norm": 0.32936125844717024, "loss": 6.087255048751831, "time": 1.1263702869415284, "epoch": 32, "step": 19606}
{"lr": 0.0014799255856381126, "data_time": 0.06654365062713623, "grad_norm": 0.6572806358337402, "loss": 6.07629623413086, "time": 1.1561203479766846, "epoch": 32, "step": 19706}
{"lr": 0.0014796272333436932, "data_time": 0.08254883289337159, "grad_norm": 0.48898551762104037, "loss": 6.099635362625122, "time": 1.1361000061035156, "epoch": 32, "step": 19806}
{"lr": 0.0014793267109998425, "data_time": 0.08318831920623779, "grad_norm": 0.4839837238192558, "loss": 6.056450510025025, "time": 1.179740047454834, "epoch": 32, "step": 19906}
{"lr": 0.0014790240195065388, "data_time": 0.07663378715515137, "grad_norm": 0.8553561687469482, "loss": 6.07158203125, "time": 1.137586283683777, "epoch": 32, "step": 20006}
{"lr": 0.0014786395411931284, "data_time": 0.09348857402801514, "grad_norm": 0.8651165634393692, "loss": 6.1290606498718265, "time": 1.1441839218139649, "epoch": 33, "step": 20132}
{"lr": 0.001478331950771373, "data_time": 0.0932633638381958, "grad_norm": 0.4040623605251312, "loss": 6.063412094116211, "time": 1.1182808876037598, "epoch": 33, "step": 20232}
{"lr": 0.0014780221941791847, "data_time": 0.09661309719085694, "grad_norm": 0.31970351934432983, "loss": 6.080948400497436, "time": 1.1166743278503417, "epoch": 33, "step": 20332}
{"lr": 0.0014777102723441933, "data_time": 0.08008062839508057, "grad_norm": 0.4518062353134155, "loss": 6.073513698577881, "time": 1.1454270839691163, "epoch": 33, "step": 20432}
{"lr": 0.001477396186200517, "data_time": 0.09364974498748779, "grad_norm": 0.47814269214868543, "loss": 6.0636878490448, "time": 1.1240228176116944, "epoch": 33, "step": 20532}
{"lr": 0.001477079936688751, "data_time": 0.08003504276275634, "grad_norm": 0.5359969884157181, "loss": 6.11933479309082, "time": 1.1342924594879151, "epoch": 33, "step": 20632}
{"lr": 0.001476678383566404, "data_time": 0.07389435768127442, "grad_norm": NaN, "loss": 6.05370774269104, "time": 1.1342128753662108, "epoch": 34, "step": 20758}
{"lr": 0.0014763572483416144, "data_time": 0.08121590614318848, "grad_norm": 0.3539979562163353, "loss": 6.097451496124267, "time": 1.1326274633407594, "epoch": 34, "step": 20858}
{"lr": 0.0014760339528600553, "data_time": 0.06675419807434083, "grad_norm": 0.5702305987477303, "loss": 6.118643426895142, "time": 1.1379237651824952, "epoch": 34, "step": 20958}
{"lr": 0.001475708498089903, "data_time": 0.06880896091461182, "grad_norm": 0.4768792510032654, "loss": 6.101509523391724, "time": 1.1376789808273315, "epoch": 34, "step": 21058}
{"lr": 0.0014753808850057994, "data_time": 0.07488315105438233, "grad_norm": 0.365270771086216, "loss": 6.104735088348389, "time": 1.12116219997406, "epoch": 34, "step": 21158}
{"lr": 0.0014750511145888543, "data_time": 0.08475258350372314, "grad_norm": 0.3289398029446602, "loss": 6.069293260574341, "time": 1.1268129110336305, "epoch": 34, "step": 21258}
{"lr": 0.001474632533781609, "data_time": 0.0954970121383667, "grad_norm": 0.5730440706014633, "loss": 6.079661417007446, "time": 1.1499523162841796, "epoch": 35, "step": 21384}
{"lr": 0.0014742978914404278, "data_time": 0.08188292980194092, "grad_norm": 0.6466028153896332, "loss": 6.121696615219117, "time": 1.1371952056884767, "epoch": 35, "step": 21484}
{"lr": 0.0014739610950096555, "data_time": 0.08865721225738525, "grad_norm": 0.39390854090452193, "loss": 6.062417697906494, "time": 1.1517019987106323, "epoch": 35, "step": 21584}
{"lr": 0.0014736221454979017, "data_time": 0.08520641326904296, "grad_norm": 0.3663003444671631, "loss": 6.060164928436279, "time": 1.1333990573883057, "epoch": 35, "step": 21684}
{"lr": 0.001473281043920222, "data_time": 0.0866128921508789, "grad_norm": 0.5582701355218888, "loss": 6.044389343261718, "time": 1.152376079559326, "epoch": 35, "step": 21784}
{"lr": 0.001472937791298116, "data_time": 0.09759688377380371, "grad_norm": 0.33753936141729357, "loss": 6.055581951141358, "time": 1.1366772413253785, "epoch": 35, "step": 21884}
{"lr": 0.0014725022319282838, "data_time": 0.06693775653839111, "grad_norm": 0.3398722425103188, "loss": 6.013634777069091, "time": 1.1506155014038086, "epoch": 36, "step": 22010}
{"lr": 0.0014721541217424727, "data_time": 0.08788728713989258, "grad_norm": 0.3397289589047432, "loss": 6.081168079376221, "time": 1.1317227363586426, "epoch": 36, "step": 22110}
{"lr": 0.0014718038638870429, "data_time": 0.07149684429168701, "grad_norm": 0.5078216612339019, "loss": 6.0394618034362795, "time": 1.1546036720275878, "epoch": 36, "step": 22210}
{"lr": 0.001471451459410915, "data_time": 0.09370253086090088, "grad_norm": 0.4886025801301003, "loss": 6.053720283508301, "time": 1.1330447673797608, "epoch": 36, "step": 22310}
{"lr": 0.0014710969093694405, "data_time": 0.0748244047164917, "grad_norm": 0.30189037024974824, "loss": 6.0554272651672365, "time": 1.1300127029418945, "epoch": 36, "step": 22410}
{"lr": 0.0014707402148243923, "data_time": 0.0931441307067871, "grad_norm": 0.2921936511993408, "loss": 6.07034707069397, "time": 1.1131723403930665, "epoch": 36, "step": 22510}
{"lr": 0.0014702877280068014, "data_time": 0.11777708530426026, "grad_norm": 0.2848963141441345, "loss": 6.039185094833374, "time": 1.1731084823608398, "epoch": 37, "step": 22636}
{"lr": 0.001469926190828633, "data_time": 0.08326067924499511, "grad_norm": 0.3725484013557434, "loss": 6.070005083084107, "time": 1.1404145002365111, "epoch": 37, "step": 22736}
{"lr": 0.0014695625126528579, "data_time": 0.09026727676391602, "grad_norm": 0.6331101536750794, "loss": 6.047107267379761, "time": 1.1489177465438842, "epoch": 37, "step": 22836}
{"lr": 0.0014691966945685865, "data_time": 0.08109776973724366, "grad_norm": 0.44001046419143675, "loss": 6.061321544647217, "time": 1.120911192893982, "epoch": 37, "step": 22936}
{"lr": 0.0014688287376713354, "data_time": 0.09584743976593017, "grad_norm": 0.5387399718165398, "loss": 6.054054498672485, "time": 1.1266669273376464, "epoch": 37, "step": 23036}
{"lr": 0.0014684586430630312, "data_time": 0.09015479087829589, "grad_norm": 0.5028769791126251, "loss": 6.04950590133667, "time": 1.1104104042053222, "epoch": 37, "step": 23136}
{"lr": 0.0014679892818990105, "data_time": 0.09973526000976562, "grad_norm": 0.34717683792114257, "loss": 6.026773166656494, "time": 1.1516788005828857, "epoch": 38, "step": 23262}
{"lr": 0.00146761436015648, "data_time": 0.07060298919677735, "grad_norm": 0.3234023153781891, "loss": 6.031812763214111, "time": 1.1251895427703857, "epoch": 38, "step": 23362}
{"lr": 0.0014672373043396053, "data_time": 0.10026969909667968, "grad_norm": 0.29533483982086184, "loss": 6.056343746185303, "time": 1.139544916152954, "epoch": 38, "step": 23462}
{"lr": 0.0014668581155775613, "data_time": 0.06171436309814453, "grad_norm": 0.5461410462856293, "loss": 6.051615047454834, "time": 1.1154391050338746, "epoch": 38, "step": 23562}
{"lr": 0.0014664767950059083, "data_time": 0.09849774837493896, "grad_norm": 0.3991223216056824, "loss": 6.060608816146851, "time": 1.1502544403076171, "epoch": 38, "step": 23662}
{"lr": 0.0014660933437665921, "data_time": 0.07134184837341309, "grad_norm": 0.4024945110082626, "loss": 6.035777473449707, "time": 1.1187916040420531, "epoch": 38, "step": 23762}
{"lr": 0.0014656071633377613, "data_time": 0.11050434112548828, "grad_norm": 0.35640538930892945, "loss": 6.008076763153076, "time": 1.191015362739563, "epoch": 39, "step": 23888}
{"lr": 0.0014652189010295937, "data_time": 0.08741235733032227, "grad_norm": 0.5374388337135315, "loss": 6.046767091751098, "time": 1.1062860488891602, "epoch": 39, "step": 23988}
{"lr": 0.0014648285118207909, "data_time": 0.11035013198852539, "grad_norm": 0.723004686832428, "loss": 6.00876030921936, "time": 1.1271605968475342, "epoch": 39, "step": 24088}
{"lr": 0.0014644359968804563, "data_time": 0.08618547916412353, "grad_norm": 0.44285824745893476, "loss": 6.072492408752441, "time": 1.1210344076156615, "epoch": 39, "step": 24188}
{"lr": 0.001464041357384061, "data_time": 0.104408860206604, "grad_norm": 0.5372582584619522, "loss": 6.05531759262085, "time": 1.1338766813278198, "epoch": 39, "step": 24288}
{"lr": 0.001463644594513434, "data_time": 0.08191273212432862, "grad_norm": 0.653644859790802, "loss": 6.05628342628479, "time": 1.0950695276260376, "epoch": 39, "step": 24388}
{"lr": 0.0014631416518752164, "data_time": 0.06892199516296386, "grad_norm": 0.3920275062322617, "loss": 5.999837589263916, "time": 1.2174052238464355, "epoch": 40, "step": 24514}
{"lr": 0.001462740094565716, "data_time": 0.07645132541656494, "grad_norm": 0.5179062962532044, "loss": 5.9840960025787355, "time": 1.1507955312728881, "epoch": 40, "step": 24614}
{"lr": 0.001462336417778891, "data_time": 0.07259945869445801, "grad_norm": 0.33065625876188276, "loss": 6.046910762786865, "time": 1.1517101526260376, "epoch": 40, "step": 24714}
{"lr": 0.0014619306227236387, "data_time": 0.09346675872802734, "grad_norm": 0.35610626488924024, "loss": 6.056421279907227, "time": 1.125280213356018, "epoch": 40, "step": 24814}
{"lr": 0.0014615227106151948, "data_time": 0.08020844459533691, "grad_norm": 0.5107739776372909, "loss": 6.0535240650177, "time": 1.1328427076339722, "epoch": 40, "step": 24914}
{"lr": 0.0014611126826751396, "data_time": 0.09701278209686279, "grad_norm": 0.3726542204618454, "loss": 6.0487274646759035, "time": 1.1591431856155396, "epoch": 40, "step": 25014}
{"lr": 0.0014605930368500866, "data_time": 0.08905642032623291, "grad_norm": 1.029637685418129, "loss": 6.027295351028442, "time": 1.1199942827224731, "epoch": 41, "step": 25140}
{"lr": 0.0014601782316637831, "data_time": 0.08234806060791015, "grad_norm": 0.29560427069664, "loss": 6.055532312393188, "time": 1.1047762632369995, "epoch": 41, "step": 25240}
{"lr": 0.0014597613146721967, "data_time": 0.10422708988189697, "grad_norm": 0.5689835399389267, "loss": 6.0262034893035885, "time": 1.1408944606781006, "epoch": 41, "step": 25340}
{"lr": 0.0014593422871238713, "data_time": 0.07816274166107177, "grad_norm": 0.4457477331161499, "loss": 5.980231380462646, "time": 1.1052053928375245, "epoch": 41, "step": 25440}
{"lr": 0.0014589211502736758, "data_time": 0.09869625568389892, "grad_norm": 0.4519372820854187, "loss": 6.049089765548706, "time": 1.1150448083877564, "epoch": 41, "step": 25540}
{"lr": 0.0014584979053827917, "data_time": 0.07809879779815673, "grad_norm": 0.5973658919334411, "loss": 6.060606050491333, "time": 1.1037164688110352, "epoch": 41, "step": 25640}
{"lr": 0.001457961617353645, "data_time": 0.10117762088775635, "grad_norm": 0.6633815437555313, "loss": 6.017280101776123, "time": 1.1434529781341554, "epoch": 42, "step": 25766}
{"lr": 0.0014575336129697704, "data_time": 0.07008442878723145, "grad_norm": 0.46625675857067106, "loss": 6.05749740600586, "time": 1.1396209478378296, "epoch": 42, "step": 25866}
{"lr": 0.0014571035047004795, "data_time": 0.08314945697784423, "grad_norm": 0.4667463153600693, "loss": 6.064465475082398, "time": 1.135409903526306, "epoch": 42, "step": 25966}
{"lr": 0.0014566712938338252, "data_time": 0.07932825088500976, "grad_norm": 0.3857710540294647, "loss": 6.022931814193726, "time": 1.1414951801300048, "epoch": 42, "step": 26066}
{"lr": 0.0014562369816641526, "data_time": 0.08230078220367432, "grad_norm": 0.4054423078894615, "loss": 6.0754743099212645, "time": 1.1606532335281372, "epoch": 42, "step": 26166}
{"lr": 0.0014558005694921001, "data_time": 0.08385753631591797, "grad_norm": 0.6541201114654541, "loss": 6.004066848754883, "time": 1.125022292137146, "epoch": 42, "step": 26266}
{"lr": 0.001455247702194636, "data_time": 0.09929702281951905, "grad_norm": 0.40083927661180496, "loss": 5.999754333496094, "time": 1.1507221460342407, "epoch": 43, "step": 26392}
{"lr": 0.001454806548841398, "data_time": 0.08231871128082276, "grad_norm": 0.6984338790178299, "loss": 6.0294177532196045, "time": 1.1227852344512939, "epoch": 43, "step": 26492}
{"lr": 0.0014543632997695152, "data_time": 0.10558838844299316, "grad_norm": 0.4025086462497711, "loss": 5.994599962234497, "time": 1.1469764709472656, "epoch": 43, "step": 26592}
{"lr": 0.0014539179563063912, "data_time": 0.07941417694091797, "grad_norm": 0.46262974441051485, "loss": 6.070513391494751, "time": 1.136423373222351, "epoch": 43, "step": 26692}
{"lr": 0.0014534705197856974, "data_time": 0.10309324264526368, "grad_norm": 0.40359847247600555, "loss": 5.96215853691101, "time": 1.1638715505599975, "epoch": 43, "step": 26792}
{"lr": 0.0014530209915473813, "data_time": 0.09812238216400146, "grad_norm": 0.6849942058324814, "loss": 6.06538233757019, "time": 1.1240063667297364, "epoch": 43, "step": 26892}
{"lr": 0.0014524516098630297, "data_time": 0.06613576412200928, "grad_norm": 0.4142729341983795, "loss": 6.013626480102539, "time": 1.1417945623397827, "epoch": 44, "step": 27018}
{"lr": 0.001451997359311731, "data_time": 0.09567625522613525, "grad_norm": 0.45660233199596406, "loss": 6.001144504547119, "time": 1.1695303678512574, "epoch": 44, "step": 27118}
{"lr": 0.0014515410214544994, "data_time": 0.08324618339538574, "grad_norm": 0.3551636874675751, "loss": 6.040364742279053, "time": 1.1117022514343262, "epoch": 44, "step": 27218}
{"lr": 0.0014510825976579355, "data_time": 0.08145067691802979, "grad_norm": 0.3135509595274925, "loss": 6.022354173660278, "time": 1.1127917051315308, "epoch": 44, "step": 27318}
{"lr": 0.001450622089294886, "data_time": 0.07552549839019776, "grad_norm": 0.35019456744194033, "loss": 5.99403223991394, "time": 1.1202528715133666, "epoch": 44, "step": 27418}
{"lr": 0.00145015949774444, "data_time": 0.11457469463348388, "grad_norm": 0.3116399675607681, "loss": 6.036564588546753, "time": 1.1480561017990112, "epoch": 44, "step": 27518}
{"lr": 0.001449573668492664, "data_time": 0.10827476978302002, "grad_norm": 0.6408279627561569, "loss": 6.031566190719604, "time": 1.1569737672805787, "epoch": 45, "step": 27644}
{"lr": 0.001449106374051622, "data_time": 0.09134063720703126, "grad_norm": 0.36394174844026567, "loss": 6.007384586334228, "time": 1.126037859916687, "epoch": 45, "step": 27744}
{"lr": 0.0014486370009623118, "data_time": 0.09511699676513671, "grad_norm": 0.8277514547109603, "loss": 6.014773178100586, "time": 1.1513926982879639, "epoch": 45, "step": 27844}
{"lr": 0.0014481655506303735, "data_time": 0.09094891548156739, "grad_norm": 0.44133155345916747, "loss": 6.058913135528565, "time": 1.1268364429473876, "epoch": 45, "step": 27944}
{"lr": 0.001447692024467663, "data_time": 0.09186773300170899, "grad_norm": 0.40526082217693327, "loss": 5.998563480377197, "time": 1.1204967498779297, "epoch": 45, "step": 28044}
{"lr": 0.0014472164238922547, "data_time": 0.0843095064163208, "grad_norm": 0.3796423256397247, "loss": 5.991895961761474, "time": 1.1454308748245239, "epoch": 45, "step": 28144}
{"lr": 0.0014466142158227186, "data_time": 0.09307832717895508, "grad_norm": 0.387615779042244, "loss": 6.010231685638428, "time": 1.1579883337020873, "epoch": 46, "step": 28270}
{"lr": 0.0014461339323310072, "data_time": 0.06413204669952392, "grad_norm": 0.31936192214488984, "loss": 6.027888441085816, "time": 1.1444724321365356, "epoch": 46, "step": 28370}
{"lr": 0.0014456515790926333, "data_time": 0.08604762554168702, "grad_norm": 0.30745644122362137, "loss": 6.067444896697998, "time": 1.1637009859085083, "epoch": 46, "step": 28470}
{"lr": 0.0014451671575521075, "data_time": 0.0656853199005127, "grad_norm": 0.4428530693054199, "loss": 6.049535846710205, "time": 1.1191229820251465, "epoch": 46, "step": 28570}
{"lr": 0.0014446806691601308, "data_time": 0.08799707889556885, "grad_norm": 0.7005870655179024, "loss": 6.017680215835571, "time": 1.1363576173782348, "epoch": 46, "step": 28670}
{"lr": 0.001444192115373598, "data_time": 0.07302236557006836, "grad_norm": 0.3378881633281708, "loss": 6.004381275177002, "time": 1.1347527980804444, "epoch": 46, "step": 28770}
{"lr": 0.0014435735991580859, "data_time": 0.10032854080200196, "grad_norm": 0.7932251498103142, "loss": 6.039410781860352, "time": 1.189574933052063, "epoch": 47, "step": 28896}
{"lr": 0.001443080382979101, "data_time": 0.08250653743743896, "grad_norm": 0.45375114381313325, "loss": 5.966511631011963, "time": 1.1441587686538697, "epoch": 47, "step": 28996}
{"lr": 0.0014425851061979549, "data_time": 0.1016427755355835, "grad_norm": 0.39026477336883547, "loss": 6.059238386154175, "time": 1.1703198909759522, "epoch": 47, "step": 29096}
{"lr": 0.0014420877702978577, "data_time": 0.10350182056427001, "grad_norm": 0.40949045717716215, "loss": 5.993701505661011, "time": 1.1486580133438111, "epoch": 47, "step": 29196}
{"lr": 0.0014415883767681878, "data_time": 0.09791245460510253, "grad_norm": 0.3603795006871223, "loss": 6.024757766723633, "time": 1.1481386184692384, "epoch": 47, "step": 29296}
{"lr": 0.0014410869271044818, "data_time": 0.08593986034393311, "grad_norm": 0.5634496331214904, "loss": 6.013681125640869, "time": 1.1367560148239135, "epoch": 47, "step": 29396}
{"lr": 0.001440452175328614, "data_time": 0.06566753387451171, "grad_norm": 0.6856780022382736, "loss": 6.012750148773193, "time": 1.1563958883285523, "epoch": 48, "step": 29522}
{"lr": 0.0014399460843434624, "data_time": 0.11948888301849366, "grad_norm": 0.2986711263656616, "loss": 6.015340328216553, "time": 1.1722603559494018, "epoch": 48, "step": 29622}
{"lr": 0.0014394379421424695, "data_time": 0.08095014095306396, "grad_norm": 0.4761431202292442, "loss": 6.031393003463745, "time": 1.1498526573181151, "epoch": 48, "step": 29722}
{"lr": 0.0014389277502473727, "data_time": 0.10330626964569092, "grad_norm": 0.32355292737483976, "loss": 5.995506238937378, "time": 1.1396496772766114, "epoch": 48, "step": 29822}
{"lr": 0.0014384155101860511, "data_time": 0.06477131843566894, "grad_norm": 0.5515806525945663, "loss": 6.029589080810547, "time": 1.1466416835784912, "epoch": 48, "step": 29922}
{"lr": 0.0014379012234925145, "data_time": 0.08587040901184081, "grad_norm": 0.3433693140745163, "loss": 6.030298852920533, "time": 1.130052351951599, "epoch": 48, "step": 30022}
{"lr": 0.0014372503106472247, "data_time": 0.10467758178710937, "grad_norm": 0.4568896532058716, "loss": 6.026084947586059, "time": 1.1698165893554688, "epoch": 49, "step": 30148}
{"lr": 0.0014367314042479278, "data_time": 0.09584577083587646, "grad_norm": 0.41639868915081024, "loss": 5.998816347122192, "time": 1.1713486909866333, "epoch": 49, "step": 30248}
{"lr": 0.0014362104562598275, "data_time": 0.09535706043243408, "grad_norm": 0.41862197816371916, "loss": 5.967977523803711, "time": 1.1566861867904663, "epoch": 49, "step": 30348}
{"lr": 0.0014356874682430127, "data_time": 0.0925980567932129, "grad_norm": 0.35075019896030424, "loss": 6.040096616744995, "time": 1.1223847627639771, "epoch": 49, "step": 30448}
{"lr": 0.0014351624417636825, "data_time": 0.10412468910217285, "grad_norm": 0.3129697635769844, "loss": 5.973993110656738, "time": 1.1512513875961303, "epoch": 49, "step": 30548}
{"lr": 0.001434635378394138, "data_time": 0.0948256492614746, "grad_norm": 0.40622184574604037, "loss": 6.048525094985962, "time": 1.1193556308746337, "epoch": 49, "step": 30648}
{"lr": 0.0014339683808669366, "data_time": 0.09333522319793701, "grad_norm": 0.3050158619880676, "loss": 6.025237560272217, "time": 1.1532304763793946, "epoch": 50, "step": 30774}
{"lr": 0.001433436719949461, "data_time": 0.07706410884857177, "grad_norm": 0.47392724454402924, "loss": 6.044050025939941, "time": 1.1400477409362793, "epoch": 50, "step": 30874}
{"lr": 0.0014329030273098114, "data_time": 0.11243925094604493, "grad_norm": 0.49453950822353365, "loss": 5.991436576843261, "time": 1.1772313356399535, "epoch": 50, "step": 30974}
{"lr": 0.0014323673045462425, "data_time": 0.06681346893310547, "grad_norm": 0.3518908366560936, "loss": 6.001587581634522, "time": 1.1521570920944213, "epoch": 50, "step": 31074}
{"lr": 0.001431829553263089, "data_time": 0.09838654994964599, "grad_norm": 0.4328737407922745, "loss": 6.015827274322509, "time": 1.1324909210205079, "epoch": 50, "step": 31174}
{"lr": 0.0014312897750707607, "data_time": 0.07306790351867676, "grad_norm": 0.5445087075233459, "loss": 5.985088539123535, "time": 1.115360689163208, "epoch": 50, "step": 31274}
{"lr": 0.001430606771136765, "data_time": 0.09833500385284424, "grad_norm": 0.30822718888521194, "loss": 5.992765092849732, "time": 1.1661310195922852, "epoch": 51, "step": 31400}
{"lr": 0.001430062418093876, "data_time": 0.0805279016494751, "grad_norm": 0.7049906015396118, "loss": 5.965604400634765, "time": 1.115795660018921, "epoch": 51, "step": 31500}
{"lr": 0.0014295160434338707, "data_time": 0.10243980884552002, "grad_norm": 0.3799318015575409, "loss": 6.03234896659851, "time": 1.139962863922119, "epoch": 51, "step": 31600}
{"lr": 0.0014289676487929872, "data_time": 0.08381409645080566, "grad_norm": 0.5216901659965515, "loss": 6.0353161811828615, "time": 1.1265788316726684, "epoch": 51, "step": 31700}
{"lr": 0.0014284172358135076, "data_time": 0.1135101318359375, "grad_norm": 0.4640476644039154, "loss": 5.982605218887329, "time": 1.1616614818573, "epoch": 51, "step": 31800}
{"lr": 0.001427864806143758, "data_time": 0.07785592079162598, "grad_norm": 0.5580692023038865, "loss": 6.011713314056396, "time": 1.1073763608932494, "epoch": 51, "step": 31900}
{"lr": 0.0014271658759565096, "data_time": 0.06659834384918213, "grad_norm": 0.5230441153049469, "loss": 5.9859662532806395, "time": 1.169332456588745, "epoch": 52, "step": 32026}
{"lr": 0.0014266088946704504, "data_time": 0.0798379898071289, "grad_norm": 0.41411018520593645, "loss": 5.981087684631348, "time": 1.147585892677307, "epoch": 52, "step": 32126}
{"lr": 0.0014260499021095836, "data_time": 0.07769227027893066, "grad_norm": 0.33807840943336487, "loss": 5.966258764266968, "time": 1.1656152963638307, "epoch": 52, "step": 32226}
{"lr": 0.0014254888999479293, "data_time": 0.10147311687469482, "grad_norm": 0.6328616261482238, "loss": 5.98481593132019, "time": 1.1406234979629517, "epoch": 52, "step": 32326}
{"lr": 0.001424925889865528, "data_time": 0.06675970554351807, "grad_norm": 0.3310208275914192, "loss": 6.016088056564331, "time": 1.147050428390503, "epoch": 52, "step": 32426}
{"lr": 0.0014243608735484293, "data_time": 0.0874927282333374, "grad_norm": 0.3283880725502968, "loss": 5.975950479507446, "time": 1.1272286891937255, "epoch": 52, "step": 32526}
{"lr": 0.0014236460991304827, "data_time": 0.07984764575958252, "grad_norm": 0.32306310087442397, "loss": 5.9907914161682125, "time": 1.173071026802063, "epoch": 53, "step": 32652}
{"lr": 0.0014230765549654758, "data_time": 0.09402213096618653, "grad_norm": 0.42541341930627824, "loss": 6.011547517776489, "time": 1.1171786308288574, "epoch": 53, "step": 32752}
{"lr": 0.0014225050101039985, "data_time": 0.08378548622131347, "grad_norm": 0.3950974866747856, "loss": 5.9392979621887205, "time": 1.1302117347717284, "epoch": 53, "step": 32852}
{"lr": 0.0014219314662576635, "data_time": 0.09609522819519042, "grad_norm": 0.47192666679620743, "loss": 5.999931907653808, "time": 1.1542296171188355, "epoch": 53, "step": 32952}
{"lr": 0.0014213559251440681, "data_time": 0.08546607494354248, "grad_norm": 0.6029416635632515, "loss": 5.996711015701294, "time": 1.139654541015625, "epoch": 53, "step": 33052}
{"lr": 0.0014207783884867924, "data_time": 0.09736192226409912, "grad_norm": 0.4344814375042915, "loss": 5.9971380710601805, "time": 1.1247821569442749, "epoch": 53, "step": 33152}
{"lr": 0.0014200478537200936, "data_time": 0.07685627937316894, "grad_norm": 0.34033711552619933, "loss": 5.993275165557861, "time": 1.1912718057632445, "epoch": 54, "step": 33278}
{"lr": 0.001419465813514674, "data_time": 0.07719278335571289, "grad_norm": 0.5432113349437714, "loss": 5.997642993927002, "time": 1.1244770526885985, "epoch": 54, "step": 33378}
{"lr": 0.001418881783425914, "data_time": 0.07732977867126464, "grad_norm": 0.3131008341908455, "loss": 5.980378341674805, "time": 1.1361600875854492, "epoch": 54, "step": 33478}
{"lr": 0.0014182957652028134, "data_time": 0.07323424816131592, "grad_norm": 0.3524564504623413, "loss": 5.971358060836792, "time": 1.127239465713501, "epoch": 54, "step": 33578}
{"lr": 0.0014177077606003271, "data_time": 0.08947098255157471, "grad_norm": 0.46838692724704745, "loss": 5.982228136062622, "time": 1.1187885522842407, "epoch": 54, "step": 33678}
{"lr": 0.001417117771379359, "data_time": 0.08832387924194336, "grad_norm": 0.3296696349978447, "loss": 5.983973407745362, "time": 1.1296059608459472, "epoch": 54, "step": 33778}
{"lr": 0.0014163715619954042, "data_time": 0.08980760574340821, "grad_norm": 0.5366921573877335, "loss": 5.978169727325439, "time": 1.14601731300354, "epoch": 55, "step": 33904}
{"lr": 0.0014157770940545693, "data_time": 0.0964205026626587, "grad_norm": 0.2627499461174011, "loss": 6.00129919052124, "time": 1.1651602506637573, "epoch": 55, "step": 34004}
{"lr": 0.0014151806472770428, "data_time": 0.08610973358154297, "grad_norm": 0.6247596204280853, "loss": 5.967999649047852, "time": 1.1387426137924195, "epoch": 55, "step": 34104}
{"lr": 0.0014145822234490114, "data_time": 0.08482744693756103, "grad_norm": 0.7401712477207184, "loss": 6.0229198932647705, "time": 1.1424805879592896, "epoch": 55, "step": 34204}
{"lr": 0.0014139818243625805, "data_time": 0.12177002429962158, "grad_norm": 0.6168628677725791, "loss": 6.004356622695923, "time": 1.1623757123947143, "epoch": 55, "step": 34304}
{"lr": 0.001413379451815773, "data_time": 0.08287174701690674, "grad_norm": 0.39792290329933167, "loss": 6.00106897354126, "time": 1.119165015220642, "epoch": 55, "step": 34404}
{"lr": 0.0014126176553855407, "data_time": 0.06883320808410645, "grad_norm": 0.27071732729673387, "loss": 5.979773044586182, "time": 1.181106185913086, "epoch": 56, "step": 34530}
{"lr": 0.0014120108294727411, "data_time": 0.0818450927734375, "grad_norm": 0.33890830874443056, "loss": 5.98643741607666, "time": 1.1237279415130614, "epoch": 56, "step": 34630}
{"lr": 0.001411402036002124, "data_time": 0.0767669677734375, "grad_norm": 0.3852496936917305, "loss": 6.0310876846313475, "time": 1.13551025390625, "epoch": 56, "step": 34730}
{"lr": 0.0014107912767968526, "data_time": 0.0923732042312622, "grad_norm": 0.4181489050388336, "loss": 5.9735260009765625, "time": 1.1424809217453002, "epoch": 56, "step": 34830}
{"lr": 0.0014101785536859732, "data_time": 0.06725802421569824, "grad_norm": 0.6503882259130478, "loss": 6.040816020965576, "time": 1.1759116888046264, "epoch": 56, "step": 34930}
{"lr": 0.001409563868504415, "data_time": 0.09163503646850586, "grad_norm": 0.9869982302188873, "loss": 6.026489782333374, "time": 1.15411696434021, "epoch": 56, "step": 35030}
{"lr": 0.0014087865744280919, "data_time": 0.08713610172271728, "grad_norm": 0.7956333488225937, "loss": 7.023530292510986, "time": 1.1497824668884278, "epoch": 57, "step": 35156}
{"lr": 0.001408167461757036, "data_time": 0.08291044235229492, "grad_norm": 0.4614290490746498, "loss": 6.08555793762207, "time": 1.1189528465270997, "epoch": 57, "step": 35256}
{"lr": 0.0014075463930379405, "data_time": 0.0793088436126709, "grad_norm": 0.4180247738957405, "loss": 6.033245086669922, "time": 1.144051718711853, "epoch": 57, "step": 35356}
{"lr": 0.0014069233701307277, "data_time": 0.09144854545593262, "grad_norm": 0.39588944613933563, "loss": 6.015507984161377, "time": 1.1225213050842284, "epoch": 57, "step": 35456}
{"lr": 0.0014062983949011715, "data_time": 0.09117789268493652, "grad_norm": 0.39303849041461947, "loss": 6.028058671951294, "time": 1.1426988601684571, "epoch": 57, "step": 35556}
{"lr": 0.0014056714692208922, "data_time": 0.08059995174407959, "grad_norm": 0.6971276819705963, "loss": 6.033206129074097, "time": 1.1324095964431762, "epoch": 57, "step": 35656}
{"lr": 0.001404878768717388, "data_time": 0.09957482814788818, "grad_norm": 0.43438581228256223, "loss": 6.006912660598755, "time": 1.120519995689392, "epoch": 58, "step": 35782}
{"lr": 0.0014042474419436936, "data_time": 0.09398267269134522, "grad_norm": 0.5284758284687996, "loss": 5.95556116104126, "time": 1.1254592418670655, "epoch": 58, "step": 35882}
{"lr": 0.001403614170861289, "data_time": 0.08411941528320313, "grad_norm": 0.5702990740537643, "loss": 6.038475084304809, "time": 1.128145980834961, "epoch": 58, "step": 35982}
{"lr": 0.0014029789573666374, "data_time": 0.07447938919067383, "grad_norm": 0.6029028356075287, "loss": 5.962485933303833, "time": 1.1253432273864745, "epoch": 58, "step": 36082}
{"lr": 0.0014023418033620198, "data_time": 0.0952028512954712, "grad_norm": 0.2851359412074089, "loss": 5.938777112960816, "time": 1.1226949453353883, "epoch": 58, "step": 36182}
{"lr": 0.0014017027107555292, "data_time": 0.09479014873504639, "grad_norm": 0.2926585078239441, "loss": 5.961891222000122, "time": 1.1184259891510009, "epoch": 58, "step": 36282}
{"lr": 0.0014008946968517618, "data_time": 0.11273601055145263, "grad_norm": 0.37632625848054885, "loss": 6.0185966968536375, "time": 1.167616605758667, "epoch": 59, "step": 36408}
{"lr": 0.0014002512300644266, "data_time": 0.08460967540740967, "grad_norm": 0.4562999427318573, "loss": 5.981879043579101, "time": 1.1210788011550903, "epoch": 59, "step": 36508}
{"lr": 0.0013996058309358815, "data_time": 0.10823819637298585, "grad_norm": 0.29528965055942535, "loss": 5.946838283538819, "time": 1.1416278600692749, "epoch": 59, "step": 36608}
{"lr": 0.001398958501398911, "data_time": 0.09096965789794922, "grad_norm": 0.5723665922880172, "loss": 5.995713758468628, "time": 1.128970241546631, "epoch": 59, "step": 36708}
{"lr": 0.0013983092433920792, "data_time": 0.11363608837127685, "grad_norm": 0.32934854477643966, "loss": 5.965889930725098, "time": 1.135512399673462, "epoch": 59, "step": 36808}
{"lr": 0.0013976580588597265, "data_time": 0.10161426067352294, "grad_norm": 0.4474482357501984, "loss": 6.017895650863648, "time": 1.129581594467163, "epoch": 59, "step": 36908}
{"lr": 0.0013968348263797086, "data_time": 0.0751798391342163, "grad_norm": 0.5083043783903122, "loss": 5.992754983901977, "time": 1.1750840663909912, "epoch": 60, "step": 37034}
{"lr": 0.0013961792950924127, "data_time": 0.07775292396545411, "grad_norm": 0.45293882489204407, "loss": 5.9794717788696286, "time": 1.1238380193710327, "epoch": 60, "step": 37134}
{"lr": 0.0013955218436581761, "data_time": 0.0860253095626831, "grad_norm": 0.2961050733923912, "loss": 5.979570722579956, "time": 1.1488806009292603, "epoch": 60, "step": 37234}
{"lr": 0.0013948624740458774, "data_time": 0.08912520408630371, "grad_norm": 0.29366644769906997, "loss": 6.024430751800537, "time": 1.1310178518295289, "epoch": 60, "step": 37334}
{"lr": 0.0013942011882301377, "data_time": 0.07939708232879639, "grad_norm": 0.7141527205705642, "loss": 6.02070460319519, "time": 1.153810691833496, "epoch": 60, "step": 37434}
{"lr": 0.0013935379881913179, "data_time": 0.08240230083465576, "grad_norm": 0.362700554728508, "loss": 6.007016229629516, "time": 1.1066113710403442, "epoch": 60, "step": 37534}
{"lr": 0.0013926996337450326, "data_time": 0.08114495277404785, "grad_norm": 0.3910347789525986, "loss": 5.970829439163208, "time": 1.1847310543060303, "epoch": 61, "step": 37660}
{"lr": 0.0013920321148872769, "data_time": 0.07738828659057617, "grad_norm": 0.35015968382358553, "loss": 6.023947763442993, "time": 1.1429302215576171, "epoch": 61, "step": 37760}
{"lr": 0.0013913626883021901, "data_time": 0.08059384822845458, "grad_norm": 0.3129228979349136, "loss": 5.959291172027588, "time": 1.135330033302307, "epoch": 61, "step": 37860}
{"lr": 0.0013906913559945109, "data_time": 0.07648441791534424, "grad_norm": 0.29778705090284346, "loss": 6.0215002536773685, "time": 1.1322206020355225, "epoch": 61, "step": 37960}
{"lr": 0.001390018119974685, "data_time": 0.08141889572143554, "grad_norm": 0.3053443133831024, "loss": 5.997538042068482, "time": 1.1526366233825684, "epoch": 61, "step": 38060}
{"lr": 0.001389342982258861, "data_time": 0.07553737163543701, "grad_norm": 0.5374154508113861, "loss": 6.000007438659668, "time": 1.139865779876709, "epoch": 61, "step": 38160}
{"lr": 0.0013884896042309187, "data_time": 0.0806549072265625, "grad_norm": 0.5304447993636131, "loss": 6.006138467788697, "time": 1.138827133178711, "epoch": 62, "step": 38286}
{"lr": 0.0013878101761390042, "data_time": 0.07732582092285156, "grad_norm": 0.4077939555048943, "loss": 5.996597528457642, "time": 1.1006492137908936, "epoch": 62, "step": 38386}
{"lr": 0.0013871288529632428, "data_time": 0.09635155200958252, "grad_norm": 0.3008129641413689, "loss": 6.019603157043457, "time": 1.136854600906372, "epoch": 62, "step": 38486}
{"lr": 0.0013864456367440054, "data_time": 0.07231030464172364, "grad_norm": 0.4064247578382492, "loss": 6.003834581375122, "time": 1.1251490354537963, "epoch": 62, "step": 38586}
{"lr": 0.0013857605295273263, "data_time": 0.09158482551574706, "grad_norm": 0.2835320830345154, "loss": 5.99259386062622, "time": 1.126497006416321, "epoch": 62, "step": 38686}
{"lr": 0.0013850735333649038, "data_time": 0.09153423309326172, "grad_norm": 0.5461244344711303, "loss": 6.017613124847412, "time": 1.1164268255233765, "epoch": 62, "step": 38786}
{"lr": 0.0013842052319030006, "data_time": 0.12330336570739746, "grad_norm": 0.47352167069911955, "loss": 5.979089546203613, "time": 1.1995705604553222, "epoch": 63, "step": 38912}
{"lr": 0.0013835139743108218, "data_time": 0.08679516315460205, "grad_norm": 0.3714992180466652, "loss": 5.9677716255187985, "time": 1.1648781299591064, "epoch": 63, "step": 39012}
{"lr": 0.0013828208345006852, "data_time": 0.09496254920959472, "grad_norm": 0.37138035893440247, "loss": 5.990806913375854, "time": 1.1488014698028564, "epoch": 63, "step": 39112}
{"lr": 0.0013821258145483423, "data_time": 0.09223675727844238, "grad_norm": 0.4224657565355301, "loss": 5.927015781402588, "time": 1.1663089275360108, "epoch": 63, "step": 39212}
{"lr": 0.0013814289165351783, "data_time": 0.10389342308044433, "grad_norm": 0.25826777666807177, "loss": 5.987163829803467, "time": 1.160094690322876, "epoch": 63, "step": 39312}
{"lr": 0.001380730142548202, "data_time": 0.08162617683410645, "grad_norm": 0.43666909337043763, "loss": 6.001009893417359, "time": 1.1470776796340942, "epoch": 63, "step": 39412}
{"lr": 0.0013798470195513625, "data_time": 0.0711390495300293, "grad_norm": 0.38104124218225477, "loss": 6.015258264541626, "time": 1.181052851676941, "epoch": 64, "step": 39538}
{"lr": 0.001379144013581067, "data_time": 0.08390398025512695, "grad_norm": 0.3916811004281044, "loss": 6.002936172485351, "time": 1.1559218168258667, "epoch": 64, "step": 39638}
{"lr": 0.0013784391384795845, "data_time": 0.06616425514221191, "grad_norm": 0.359374126791954, "loss": 5.986703968048095, "time": 1.135046911239624, "epoch": 64, "step": 39738}
{"lr": 0.0013777323963578096, "data_time": 0.09110970497131347, "grad_norm": 0.31046559661626816, "loss": 6.000963020324707, "time": 1.1510689496994018, "epoch": 64, "step": 39838}
{"lr": 0.0013770237893322324, "data_time": 0.08169667720794678, "grad_norm": 0.5184414178133011, "loss": 5.986391639709472, "time": 1.1235307931900025, "epoch": 64, "step": 39938}
{"lr": 0.0013763133195249266, "data_time": 0.10674488544464111, "grad_norm": 0.48582708835601807, "loss": 5.976307439804077, "time": 1.1238475084304809, "epoch": 64, "step": 40038}
{"lr": 0.0013754154786315513, "data_time": 0.09346070289611816, "grad_norm": 0.36782582849264145, "loss": 6.016955900192261, "time": 1.1460631370544434, "epoch": 65, "step": 40164}
{"lr": 0.001374700806784005, "data_time": 0.08092758655548096, "grad_norm": 0.28055842518806456, "loss": 5.969164180755615, "time": 1.1660041570663453, "epoch": 65, "step": 40264}
{"lr": 0.0013739842791113924, "data_time": 0.0790705680847168, "grad_norm": 0.301402647793293, "loss": 5.975613832473755, "time": 1.1493847370147705, "epoch": 65, "step": 40364}
{"lr": 0.001373265897759506, "data_time": 0.07793021202087402, "grad_norm": 0.3743509709835052, "loss": 6.005718994140625, "time": 1.1350462198257447, "epoch": 65, "step": 40464}
{"lr": 0.0013725456648796927, "data_time": 0.08391799926757812, "grad_norm": 0.49322719871997833, "loss": 5.982674884796142, "time": 1.151710534095764, "epoch": 65, "step": 40564}
{"lr": 0.0013718235826288394, "data_time": 0.07466833591461182, "grad_norm": 0.2515269383788109, "loss": 5.990388059616089, "time": 1.141152596473694, "epoch": 65, "step": 40664}
{"lr": 0.0013709111292045395, "data_time": 0.08227136135101318, "grad_norm": 0.6332836449146271, "loss": 5.991411352157593, "time": 1.1565819263458252, "epoch": 66, "step": 40790}
{"lr": 0.0013701848753496517, "data_time": 0.08002331256866455, "grad_norm": 0.6296037256717681, "loss": 6.016357374191284, "time": 1.1025591850280763, "epoch": 66, "step": 40890}
{"lr": 0.0013694567791936085, "data_time": 0.09843852519989013, "grad_norm": 0.3165943816304207, "loss": 5.967462539672852, "time": 1.1338762998580934, "epoch": 66, "step": 40990}
{"lr": 0.0013687268429168479, "data_time": 0.06946361064910889, "grad_norm": 0.5709905534982681, "loss": 5.987451505661011, "time": 1.1171234369277954, "epoch": 66, "step": 41090}
{"lr": 0.0013679950687053204, "data_time": 0.09123003482818604, "grad_norm": 0.998463585972786, "loss": 5.964568710327148, "time": 1.1316925525665282, "epoch": 66, "step": 41190}
{"lr": 0.0013672614587504754, "data_time": 0.07582030296325684, "grad_norm": 0.2876211151480675, "loss": 6.0006420612335205, "time": 1.1350676536560058, "epoch": 66, "step": 41290}
{"lr": 0.0013663344998757021, "data_time": 0.09347460269927979, "grad_norm": 0.3245444282889366, "loss": 6.001859378814697, "time": 1.152499794960022, "epoch": 67, "step": 41416}
{"lr": 0.0013655967492425835, "data_time": 0.08587934970855712, "grad_norm": 0.546844756603241, "loss": 5.994469404220581, "time": 1.133614730834961, "epoch": 67, "step": 41516}
{"lr": 0.0013648571700484232, "data_time": 0.11468725204467774, "grad_norm": 0.35759367048740387, "loss": 5.950184631347656, "time": 1.1605128288269042, "epoch": 67, "step": 41616}
{"lr": 0.0013641157645080482, "data_time": 0.0904883623123169, "grad_norm": 0.25428699851036074, "loss": 6.000801420211792, "time": 1.1384550094604493, "epoch": 67, "step": 41716}
{"lr": 0.001363372534841753, "data_time": 0.11320061683654785, "grad_norm": 0.5443596303462982, "loss": 5.968864393234253, "time": 1.1742788314819337, "epoch": 67, "step": 41816}
{"lr": 0.0013626274832752975, "data_time": 0.07605447769165039, "grad_norm": 0.47652722150087357, "loss": 5.9691473007202145, "time": 1.1181171894073487, "epoch": 67, "step": 41916}
{"lr": 0.0013616861277327892, "data_time": 0.0693169116973877, "grad_norm": 0.45487292408943175, "loss": 6.01004056930542, "time": 1.1679990530014037, "epoch": 68, "step": 42042}
{"lr": 0.0013609369668997461, "data_time": 0.07382311820983886, "grad_norm": 0.31113094091415405, "loss": 6.008313703536987, "time": 1.134024453163147, "epoch": 68, "step": 42142}
{"lr": 0.001360185991460366, "data_time": 0.07936284542083741, "grad_norm": 0.4017535150051117, "loss": 5.978674745559692, "time": 1.1680842638015747, "epoch": 68, "step": 42242}
{"lr": 0.0013594332036636005, "data_time": 0.08676762580871582, "grad_norm": 0.3517987385392189, "loss": 5.98218731880188, "time": 1.142963695526123, "epoch": 68, "step": 42342}
{"lr": 0.0013586786057638361, "data_time": 0.0726464033126831, "grad_norm": 0.2885983273386955, "loss": 5.977091836929321, "time": 1.1520464658737182, "epoch": 68, "step": 42442}
{"lr": 0.0013579222000208735, "data_time": 0.09549267292022705, "grad_norm": 0.40696834176778796, "loss": 5.95719575881958, "time": 1.1493529796600341, "epoch": 68, "step": 42542}
{"lr": 0.0013569665582828722, "data_time": 0.08318262100219727, "grad_norm": 0.7646633744239807, "loss": 6.008660173416137, "time": 1.1270553588867187, "epoch": 69, "step": 42668}
{"lr": 0.0013562060751672528, "data_time": 0.09161243438720704, "grad_norm": 0.4114830866456032, "loss": 5.9502387046813965, "time": 1.1048765897750854, "epoch": 69, "step": 42768}
{"lr": 0.00135544379161295, "data_time": 0.10276682376861572, "grad_norm": 0.33188958317041395, "loss": 5.979812669754028, "time": 1.1461333513259888, "epoch": 69, "step": 42868}
{"lr": 0.0013546797099027838, "data_time": 0.09487650394439698, "grad_norm": 0.27570581883192063, "loss": 5.993147182464599, "time": 1.129033088684082, "epoch": 69, "step": 42968}
{"lr": 0.0013539138323249594, "data_time": 0.08932240009307861, "grad_norm": 0.32789196670055387, "loss": 5.972379684448242, "time": 1.1261665582656861, "epoch": 69, "step": 43068}
{"lr": 0.0013531461611730575, "data_time": 0.09485704898834228, "grad_norm": 0.6695767760276794, "loss": 5.9603677749633786, "time": 1.109174656867981, "epoch": 69, "step": 43168}
{"lr": 0.0013521763453883616, "data_time": 0.06346125602722168, "grad_norm": 0.30939916968345643, "loss": 5.968118953704834, "time": 1.1803350687026977, "epoch": 70, "step": 43294}
{"lr": 0.0013514046292362287, "data_time": 0.07000596523284912, "grad_norm": 0.44402696192264557, "loss": 5.9708925724029545, "time": 1.1341484546661378, "epoch": 70, "step": 43394}
{"lr": 0.0013506311270243588, "data_time": 0.08025827407836914, "grad_norm": 0.23578677475452423, "loss": 5.957961559295654, "time": 1.1517166376113892, "epoch": 70, "step": 43494}
{"lr": 0.001349855841069166, "data_time": 0.07040891647338868, "grad_norm": 0.5621877372264862, "loss": 5.979478359222412, "time": 1.1416358470916748, "epoch": 70, "step": 43594}
{"lr": 0.0013490787736924093, "data_time": 0.08250949382781983, "grad_norm": 0.5277099564671517, "loss": 5.972079277038574, "time": 1.1606693506240844, "epoch": 70, "step": 43694}
{"lr": 0.0013482999272211808, "data_time": 0.07032337188720703, "grad_norm": 0.417068088054657, "loss": 5.948963832855225, "time": 1.1504414558410645, "epoch": 70, "step": 43794}
{"lr": 0.0013473160512019694, "data_time": 0.10445032119750977, "grad_norm": 0.5080834597349166, "loss": 5.979866552352905, "time": 1.2126867532730103, "epoch": 71, "step": 43920}
{"lr": 0.0013465331925776374, "data_time": 0.08620913028717041, "grad_norm": 0.6211875379085541, "loss": 5.961548805236816, "time": 1.1610340118408202, "epoch": 71, "step": 44020}
{"lr": 0.0013457485624821146, "data_time": 0.10042395591735839, "grad_norm": 0.7336911231279373, "loss": 5.976415634155273, "time": 1.163102674484253, "epoch": 71, "step": 44120}
{"lr": 0.0013449621632651413, "data_time": 0.08531136512756347, "grad_norm": 0.3479576423764229, "loss": 5.953590679168701, "time": 1.164095973968506, "epoch": 71, "step": 44220}
{"lr": 0.001344173997281757, "data_time": 0.09756999015808106, "grad_norm": 0.6038575768470764, "loss": 5.998320150375366, "time": 1.1636054039001464, "epoch": 71, "step": 44320}
{"lr": 0.00134338406689229, "data_time": 0.08638184070587158, "grad_norm": 0.2641226500272751, "loss": 5.950572681427002, "time": 1.1597083568573, "epoch": 71, "step": 44420}
{"lr": 0.0013423862461007805, "data_time": 0.06539030075073242, "grad_norm": 0.45896527618169786, "loss": 5.918967962265015, "time": 1.1902748823165894, "epoch": 72, "step": 44546}
{"lr": 0.0013415923368761782, "data_time": 0.07897067070007324, "grad_norm": 0.4293077617883682, "loss": 5.971351957321167, "time": 1.1686460018157958, "epoch": 72, "step": 44646}
{"lr": 0.0013407966709768238, "data_time": 0.06841657161712647, "grad_norm": 0.2521386116743088, "loss": 5.927708911895752, "time": 1.1947353839874268, "epoch": 72, "step": 44746}
{"lr": 0.0013399992507855043, "data_time": 0.08169500827789307, "grad_norm": 0.5589964717626572, "loss": 5.970107650756836, "time": 1.1581413269042968, "epoch": 72, "step": 44846}
{"lr": 0.0013392000786902614, "data_time": 0.0779184341430664, "grad_norm": 0.380573707818985, "loss": 6.008903312683105, "time": 1.1558081388473511, "epoch": 72, "step": 44946}
{"lr": 0.0013383991570843892, "data_time": 0.08666353225708008, "grad_norm": 0.2705168783664703, "loss": 5.949561977386475, "time": 1.1262359619140625, "epoch": 72, "step": 45046}
{"lr": 0.0013373875086192736, "data_time": 0.06318302154541015, "grad_norm": 0.30344269424676895, "loss": 5.982342767715454, "time": 1.1811081171035767, "epoch": 73, "step": 45172}
{"lr": 0.0013365826419631697, "data_time": 0.08197240829467774, "grad_norm": 0.5693861365318298, "loss": 5.985724878311157, "time": 1.150942063331604, "epoch": 73, "step": 45272}
{"lr": 0.001335776033634903, "data_time": 0.07257988452911376, "grad_norm": 0.29694426208734515, "loss": 5.964947700500488, "time": 1.145991849899292, "epoch": 73, "step": 45372}
{"lr": 0.0013349676860500293, "data_time": 0.0778900146484375, "grad_norm": 0.3633132502436638, "loss": 5.97931056022644, "time": 1.1421821117401123, "epoch": 73, "step": 45472}
{"lr": 0.0013341576016293187, "data_time": 0.08028314113616944, "grad_norm": 0.5176481544971466, "loss": 5.9980034828186035, "time": 1.185021948814392, "epoch": 73, "step": 45572}
{"lr": 0.001333345782798742, "data_time": 0.08327722549438477, "grad_norm": 0.3783198341727257, "loss": 5.981907701492309, "time": 1.135319972038269, "epoch": 73, "step": 45672}
{"lr": 0.001332320425381453, "data_time": 0.0677177906036377, "grad_norm": 0.44177474081516266, "loss": 5.960647487640381, "time": 1.157321286201477, "epoch": 74, "step": 45798}
{"lr": 0.0013315046957485183, "data_time": 0.06960928440093994, "grad_norm": 0.5377579599618911, "loss": 5.9816686630249025, "time": 1.1343549728393554, "epoch": 74, "step": 45898}
{"lr": 0.0013306872396504023, "data_time": 0.07655270099639892, "grad_norm": 0.3958230629563332, "loss": 5.995617389678955, "time": 1.1291419982910156, "epoch": 74, "step": 45998}
{"lr": 0.0013298680595351506, "data_time": 0.06485638618469239, "grad_norm": 0.42154947966337203, "loss": 5.99543514251709, "time": 1.1027517795562745, "epoch": 74, "step": 46098}
{"lr": 0.0013290471578559723, "data_time": 0.09474375247955322, "grad_norm": 0.33261304199695585, "loss": 5.960365343093872, "time": 1.1386883020401002, "epoch": 74, "step": 46198}
{"lr": 0.0013282245370712306, "data_time": 0.06922991275787353, "grad_norm": 0.2504578962922096, "loss": 5.929322385787964, "time": 1.154211735725403, "epoch": 74, "step": 46298}
{"lr": 0.001327185591032007, "data_time": 0.0921781301498413, "grad_norm": 0.3569064199924469, "loss": 5.929420518875122, "time": 1.1786984920501709, "epoch": 75, "step": 46424}
{"lr": 0.0013263590941517292, "data_time": 0.09385347366333008, "grad_norm": 0.4210678070783615, "loss": 5.977832460403443, "time": 1.165161681175232, "epoch": 75, "step": 46524}
{"lr": 0.001325530886215863, "data_time": 0.08983650207519531, "grad_norm": 0.2900852903723717, "loss": 5.954569673538208, "time": 1.1616245985031128, "epoch": 75, "step": 46624}
{"lr": 0.0013247009697046523, "data_time": 0.0817713975906372, "grad_norm": 0.42670906633138656, "loss": 5.958621597290039, "time": 1.1513517141342162, "epoch": 75, "step": 46724}
{"lr": 0.0013238693471034572, "data_time": 0.10925226211547852, "grad_norm": 0.5857397139072418, "loss": 5.9456523895263675, "time": 1.1756668567657471, "epoch": 75, "step": 46824}
{"lr": 0.001323036020902746, "data_time": 0.08008396625518799, "grad_norm": 0.4374226704239845, "loss": 5.983535242080689, "time": 1.1317925691604613, "epoch": 75, "step": 46924}
{"lr": 0.0013219836081665114, "data_time": 0.07952041625976562, "grad_norm": 0.3189860597252846, "loss": 5.9257509231567385, "time": 1.1766815900802612, "epoch": 76, "step": 47050}
{"lr": 0.001321146441031963, "data_time": 0.09318821430206299, "grad_norm": 0.4185600519180298, "loss": 5.936506605148315, "time": 1.145109510421753, "epoch": 76, "step": 47150}
{"lr": 0.0013203075784522181, "data_time": 0.07871100902557374, "grad_norm": 0.35721337497234346, "loss": 5.957849597930908, "time": 1.1760549068450927, "epoch": 76, "step": 47250}
{"lr": 0.0013194670229394272, "data_time": 0.0770498514175415, "grad_norm": 0.36331414580345156, "loss": 5.9724977016448975, "time": 1.1331621408462524, "epoch": 76, "step": 47350}
{"lr": 0.0013186247770108144, "data_time": 0.07117705345153809, "grad_norm": 0.5425533816218376, "loss": 6.002526807785034, "time": 1.1900317668914795, "epoch": 76, "step": 47450}
{"lr": 0.0013177808431886603, "data_time": 0.09009597301483155, "grad_norm": 0.3945063441991806, "loss": 5.953194189071655, "time": 1.1252198934555053, "epoch": 76, "step": 47550}
{"lr": 0.001316715087260717, "data_time": 0.0984968900680542, "grad_norm": 0.6192181691527366, "loss": 5.981994152069092, "time": 1.1769918203353882, "epoch": 77, "step": 47676}
{"lr": 0.0013158673481171764, "data_time": 0.09945685863494873, "grad_norm": 0.6711529493331909, "loss": 5.977766513824463, "time": 1.1857715129852295, "epoch": 77, "step": 47776}
{"lr": 0.0013150179293377978, "data_time": 0.09936935901641845, "grad_norm": 0.29274220019578934, "loss": 5.9932427406311035, "time": 1.1875182867050171, "epoch": 77, "step": 47876}
{"lr": 0.0013141668334663479, "data_time": 0.09012393951416016, "grad_norm": 0.48446747064590456, "loss": 5.955671930313111, "time": 1.1217061519622802, "epoch": 77, "step": 47976}
{"lr": 0.0013133140630516103, "data_time": 0.09684007167816162, "grad_norm": 0.40086347460746763, "loss": 5.923968887329101, "time": 1.1706108093261718, "epoch": 77, "step": 48076}
{"lr": 0.0013124596206473883, "data_time": 0.08650736808776856, "grad_norm": 0.3133845284581184, "loss": 5.946231079101563, "time": 1.1295164346694946, "epoch": 77, "step": 48176}
{"lr": 0.0013113806465989192, "data_time": 0.06966891288757324, "grad_norm": 0.4208630844950676, "loss": 5.9453223705291744, "time": 1.1586623430252074, "epoch": 78, "step": 48302}
{"lr": 0.001310522434932333, "data_time": 0.07603437900543213, "grad_norm": 0.4560974404215813, "loss": 5.938033962249756, "time": 1.150236225128174, "epoch": 78, "step": 48402}
{"lr": 0.0013096625596363811, "data_time": 0.08423516750335694, "grad_norm": 0.3653676822781563, "loss": 5.95796570777893, "time": 1.1361256837844849, "epoch": 78, "step": 48502}
{"lr": 0.001308801023286143, "data_time": 0.06665477752685547, "grad_norm": 0.3999556332826614, "loss": 5.9746551513671875, "time": 1.1525969505310059, "epoch": 78, "step": 48602}
{"lr": 0.0013079378284616707, "data_time": 0.07151811122894287, "grad_norm": 0.4345929056406021, "loss": 5.943682146072388, "time": 1.1518735408782959, "epoch": 78, "step": 48702}
{"lr": 0.0013070729777479844, "data_time": 0.07187526226043701, "grad_norm": 0.5001774251461029, "loss": 5.962345790863037, "time": 1.13692467212677, "epoch": 78, "step": 48802}
{"lr": 0.0013059809122013782, "data_time": 0.10034916400909424, "grad_norm": 0.4264831244945526, "loss": 5.98369345664978, "time": 1.160683274269104, "epoch": 79, "step": 48928}
{"lr": 0.0013051123287266882, "data_time": 0.08054068088531494, "grad_norm": 0.3122409924864769, "loss": 5.987103223800659, "time": 1.1402294635772705, "epoch": 79, "step": 49028}
{"lr": 0.0013042420978243404, "data_time": 0.11165783405303956, "grad_norm": 0.3048932135105133, "loss": 5.954012107849121, "time": 1.18705894947052, "epoch": 79, "step": 49128}
{"lr": 0.0013033702221004256, "data_time": 0.09365956783294678, "grad_norm": 0.5854964792728424, "loss": 5.990843772888184, "time": 1.1356818437576295, "epoch": 79, "step": 49228}
{"lr": 0.001302496704165958, "data_time": 0.08650541305541992, "grad_norm": 0.5494866207242012, "loss": 5.943973636627197, "time": 1.1363715410232544, "epoch": 79, "step": 49328}
{"lr": 0.0013016215466368733, "data_time": 0.10119140148162842, "grad_norm": 0.366167913377285, "loss": 5.952949094772339, "time": 1.1248692512512206, "epoch": 79, "step": 49428}
{"lr": 0.0013005165177508637, "data_time": 0.07009825706481934, "grad_norm": 0.3371738404035568, "loss": 5.967464923858643, "time": 1.158127522468567, "epoch": 80, "step": 49554}
{"lr": 0.0012996376644001943, "data_time": 0.08936285972595215, "grad_norm": 0.3323973208665848, "loss": 5.9353859424591064, "time": 1.1539998054504395, "epoch": 80, "step": 49654}
{"lr": 0.0012987571800169063, "data_time": 0.07381165027618408, "grad_norm": 0.4523161843419075, "loss": 5.956382799148559, "time": 1.1476571798324584, "epoch": 80, "step": 49754}
{"lr": 0.0012978750672377954, "data_time": 0.07694597244262695, "grad_norm": 0.655720554292202, "loss": 5.9618621349334715, "time": 1.1509273052215576, "epoch": 80, "step": 49854}
{"lr": 0.0012969913287045368, "data_time": 0.06326735019683838, "grad_norm": 0.35522390007972715, "loss": 5.927091884613037, "time": 1.1726223945617675, "epoch": 80, "step": 49954}
{"lr": 0.0012961059670636708, "data_time": 0.08721029758453369, "grad_norm": 0.9240565717220306, "loss": 5.984279489517212, "time": 1.1460140943527222, "epoch": 80, "step": 50054}
{"lr": 0.0012949881045182963, "data_time": 0.11842966079711914, "grad_norm": 0.3520996555685997, "loss": 5.959913349151611, "time": 1.2216269969940186, "epoch": 81, "step": 50180}
{"lr": 0.0012940990844289828, "data_time": 0.08638873100280761, "grad_norm": 0.2419215828180313, "loss": 5.943864440917968, "time": 1.1586106061935424, "epoch": 81, "step": 50280}
{"lr": 0.0012932084498934993, "data_time": 0.10175919532775879, "grad_norm": 0.22480016946792603, "loss": 5.956103086471558, "time": 1.13712317943573, "epoch": 81, "step": 50380}
{"lr": 0.0012923162035790392, "data_time": 0.09605097770690918, "grad_norm": 0.3903705194592476, "loss": 5.959099006652832, "time": 1.1197611808776855, "epoch": 81, "step": 50480}
{"lr": 0.001291422348157623, "data_time": 0.08587958812713622, "grad_norm": 0.25745525062084196, "loss": 5.998903894424439, "time": 1.152444076538086, "epoch": 81, "step": 50580}
{"lr": 0.0012905268863060895, "data_time": 0.0921247959136963, "grad_norm": 0.39452024102211, "loss": 5.950999116897583, "time": 1.1297286987304687, "epoch": 81, "step": 50680}
{"lr": 0.0012893963212874764, "data_time": 0.07832574844360352, "grad_norm": 0.5222692936658859, "loss": 5.900654077529907, "time": 1.125641894340515, "epoch": 82, "step": 50806}
{"lr": 0.0012884972387899704, "data_time": 0.07018437385559081, "grad_norm": 0.3836747780442238, "loss": 5.901660823822022, "time": 1.1306558132171631, "epoch": 82, "step": 50906}
{"lr": 0.001287596558622205, "data_time": 0.0891254186630249, "grad_norm": 0.3687071934342384, "loss": 5.972447919845581, "time": 1.147500205039978, "epoch": 82, "step": 51006}
{"lr": 0.0012866942834814572, "data_time": 0.07541189193725586, "grad_norm": 0.3622806712985039, "loss": 5.955408811569214, "time": 1.1297496557235718, "epoch": 82, "step": 51106}
{"lr": 0.00128579041606978, "data_time": 0.08300666809082032, "grad_norm": 0.3932304486632347, "loss": 5.977088737487793, "time": 1.151204562187195, "epoch": 82, "step": 51206}
{"lr": 0.001284884959093997, "data_time": 0.07269773483276368, "grad_norm": 0.5571122959256172, "loss": 5.980148601531982, "time": 1.1325096130371093, "epoch": 82, "step": 51306}
{"lr": 0.0012837418242789631, "data_time": 0.10573346614837646, "grad_norm": 0.39631338119506837, "loss": 5.9786780834197994, "time": 1.180653715133667, "epoch": 83, "step": 51432}
{"lr": 0.0012828327848845797, "data_time": 0.07853355407714843, "grad_norm": 0.3626196011900902, "loss": 5.963466596603394, "time": 1.1474165201187134, "epoch": 83, "step": 51532}
{"lr": 0.00128192216478334, "data_time": 0.09161806106567383, "grad_norm": 0.33108143508434296, "loss": 5.95880970954895, "time": 1.159969115257263, "epoch": 83, "step": 51632}
{"lr": 0.001281009966702292, "data_time": 0.09291329383850097, "grad_norm": 0.7273123681545257, "loss": 6.009570169448852, "time": 1.128585982322693, "epoch": 83, "step": 51732}
{"lr": 0.0012800961933732036, "data_time": 0.10260477066040039, "grad_norm": 0.30600198209285734, "loss": 5.950501441955566, "time": 1.168059754371643, "epoch": 83, "step": 51832}
{"lr": 0.0012791808475325634, "data_time": 0.08143403530120849, "grad_norm": 0.47429298609495163, "loss": 5.9865518569946286, "time": 1.1486830234527587, "epoch": 83, "step": 51932}
{"lr": 0.0012780252770730474, "data_time": 0.06975681781768799, "grad_norm": 0.5340912580490113, "loss": 5.993922472000122, "time": 1.1443936586380006, "epoch": 84, "step": 52058}
{"lr": 0.001277106387461589, "data_time": 0.0991471290588379, "grad_norm": 0.4307296484708786, "loss": 5.935948181152344, "time": 1.1453722953796386, "epoch": 84, "step": 52158}
{"lr": 0.0012761859342921864, "data_time": 0.06703400611877441, "grad_norm": 0.27644978016614913, "loss": 5.970244932174682, "time": 1.1477325916290284, "epoch": 84, "step": 52258}
{"lr": 0.0012752639203213267, "data_time": 0.09277150630950928, "grad_norm": 0.3656679943203926, "loss": 5.93161416053772, "time": 1.1378324985504151, "epoch": 84, "step": 52358}
{"lr": 0.0012743403483101772, "data_time": 0.08832619190216065, "grad_norm": 0.29670659005641936, "loss": 5.917687034606933, "time": 1.1806027173995972, "epoch": 84, "step": 52458}
{"lr": 0.0012734152210245714, "data_time": 0.09340493679046631, "grad_norm": 0.330329130589962, "loss": 6.000988101959228, "time": 1.1243299961090087, "epoch": 84, "step": 52558}
{"lr": 0.0012722473505318934, "data_time": 0.09051783084869384, "grad_norm": 0.5300319492816925, "loss": 5.904342555999756, "time": 1.1718830585479736, "epoch": 85, "step": 52684}
{"lr": 0.0012713187185391332, "data_time": 0.0700761079788208, "grad_norm": 0.3894667848944664, "loss": 5.949025058746338, "time": 1.1661365509033204, "epoch": 85, "step": 52784}
{"lr": 0.0012703885403208244, "data_time": 0.0750899076461792, "grad_norm": 0.5549083590507508, "loss": 5.96131649017334, "time": 1.1554491996765137, "epoch": 85, "step": 52884}
{"lr": 0.001269456818662582, "data_time": 0.07692224979400634, "grad_norm": 0.5269693940877914, "loss": 5.9770252227783205, "time": 1.1097928762435914, "epoch": 85, "step": 52984}
{"lr": 0.0012685235563546426, "data_time": 0.0797872543334961, "grad_norm": 0.4153095901012421, "loss": 5.947478580474853, "time": 1.1355010271072388, "epoch": 85, "step": 53084}
{"lr": 0.00126758875619186, "data_time": 0.0705906629562378, "grad_norm": 0.38182750195264814, "loss": 5.963369083404541, "time": 1.118984317779541, "epoch": 85, "step": 53184}
{"lr": 0.0012664087227207954, "data_time": 0.07994115352630615, "grad_norm": 1.187532413005829, "loss": 6.383737516403198, "time": 1.1763462781906129, "epoch": 86, "step": 53310}
{"lr": 0.0012654704573258136, "data_time": 0.07129302024841308, "grad_norm": 0.5826466321945191, "loss": 5.958202838897705, "time": 1.1366897583007813, "epoch": 86, "step": 53410}
{"lr": 0.0012645306632191355, "data_time": 0.0743830919265747, "grad_norm": 0.5502940207719803, "loss": 5.9677880764007565, "time": 1.1508915185928346, "epoch": 86, "step": 53510}
{"lr": 0.0012635893432151775, "data_time": 0.06648187637329102, "grad_norm": 0.49111542850732803, "loss": 5.954170036315918, "time": 1.1211874961853028, "epoch": 86, "step": 53610}
{"lr": 0.0012626465001329193, "data_time": 0.09032969474792481, "grad_norm": 0.30381157994270325, "loss": 5.957295513153076, "time": 1.1561277389526368, "epoch": 86, "step": 53710}
{"lr": 0.001261702136795902, "data_time": 0.06778826713562011, "grad_norm": 0.5058012172579766, "loss": 5.971249628067016, "time": 1.1443254709243775, "epoch": 86, "step": 53810}
{"lr": 0.0012605100788286138, "data_time": 0.0894665002822876, "grad_norm": 0.37079331576824187, "loss": 5.973957109451294, "time": 1.1543641567230225, "epoch": 87, "step": 53936}
{"lr": 0.0012595622901410153, "data_time": 0.09715797901153564, "grad_norm": 0.3755783036351204, "loss": 5.908332061767578, "time": 1.1596158504486085, "epoch": 87, "step": 54036}
{"lr": 0.0012586129904349783, "data_time": 0.1031721591949463, "grad_norm": 0.49056253731250765, "loss": 5.998127365112305, "time": 1.153687334060669, "epoch": 87, "step": 54136}
{"lr": 0.0012576621825533794, "data_time": 0.09927916526794434, "grad_norm": 0.4236664354801178, "loss": 5.9579987049102785, "time": 1.161246418952942, "epoch": 87, "step": 54236}
{"lr": 0.0012567098693436172, "data_time": 0.0962947130203247, "grad_norm": 0.4026647299528122, "loss": 5.9774411678314205, "time": 1.121614146232605, "epoch": 87, "step": 54336}
{"lr": 0.0012557560536575923, "data_time": 0.0866387128829956, "grad_norm": 0.4466620057821274, "loss": 5.942190217971802, "time": 1.148650312423706, "epoch": 87, "step": 54436}
{"lr": 0.0012545521110873658, "data_time": 0.0772993803024292, "grad_norm": 0.2589650198817253, "loss": 5.966948318481445, "time": 1.1491186141967773, "epoch": 88, "step": 54562}
{"lr": 0.0012535949103343546, "data_time": 0.08316020965576172, "grad_norm": 0.38424911051988603, "loss": 5.938647651672364, "time": 1.1234890222549438, "epoch": 88, "step": 54662}
{"lr": 0.0012526362164334843, "data_time": 0.0685840368270874, "grad_norm": 0.4586631000041962, "loss": 5.937576675415039, "time": 1.136135959625244, "epoch": 88, "step": 54762}
{"lr": 0.0012516760322557662, "data_time": 0.09413762092590332, "grad_norm": 1.0066485166549684, "loss": 6.011474895477295, "time": 1.1532591819763183, "epoch": 88, "step": 54862}
{"lr": 0.0012507143606766773, "data_time": 0.066963791847229, "grad_norm": 0.582021889090538, "loss": 5.912692975997925, "time": 1.1849779844284059, "epoch": 88, "step": 54962}
{"lr": 0.0012497512045761443, "data_time": 0.09143476486206055, "grad_norm": 0.4094421982765198, "loss": 5.937488555908203, "time": 1.1392094850540162, "epoch": 88, "step": 55062}
{"lr": 0.001248535518690982, "data_time": 0.07745134830474854, "grad_norm": 0.3008067712187767, "loss": 5.934429550170899, "time": 1.1495212316513062, "epoch": 89, "step": 55188}
{"lr": 0.00124756901820431, "data_time": 0.08469958305358886, "grad_norm": 0.30516761243343354, "loss": 5.973344612121582, "time": 1.1346296787261962, "epoch": 89, "step": 55288}
{"lr": 0.0012466010426155872, "data_time": 0.08490300178527832, "grad_norm": 0.34148562252521514, "loss": 5.948353815078735, "time": 1.1693363428115844, "epoch": 89, "step": 55388}
{"lr": 0.0012456315948236203, "data_time": 0.07676141262054444, "grad_norm": 0.3418197140097618, "loss": 5.926943445205689, "time": 1.1282455682754517, "epoch": 89, "step": 55488}
{"lr": 0.0012446606777316276, "data_time": 0.07675373554229736, "grad_norm": 0.3049547761678696, "loss": 5.910919141769409, "time": 1.139417004585266, "epoch": 89, "step": 55588}
{"lr": 0.0012436882942472242, "data_time": 0.10988664627075195, "grad_norm": 0.288907516002655, "loss": 5.918744039535523, "time": 1.1366025447845458, "epoch": 89, "step": 55688}
{"lr": 0.0012424610077132504, "data_time": 0.06862919330596924, "grad_norm": 1.0851263344287871, "loss": 5.953103017807007, "time": 1.1372151613235473, "epoch": 90, "step": 55814}
{"lr": 0.001241485320916035, "data_time": 0.08506572246551514, "grad_norm": 0.40514673441648485, "loss": 5.93851900100708, "time": 1.1603929996490479, "epoch": 90, "step": 55914}
{"lr": 0.0012405081772356877, "data_time": 0.07930395603179932, "grad_norm": 0.3941562741994858, "loss": 5.914385890960693, "time": 1.1472847700119018, "epoch": 90, "step": 56014}
{"lr": 0.0012395295795984691, "data_time": 0.07178945541381836, "grad_norm": 0.4833121657371521, "loss": 5.904386806488037, "time": 1.1426273584365845, "epoch": 90, "step": 56114}
{"lr": 0.0012385495309349985, "data_time": 0.06958086490631103, "grad_norm": 0.41884296387434006, "loss": 5.959301328659057, "time": 1.1432337522506715, "epoch": 90, "step": 56214}
{"lr": 0.0012375680341802397, "data_time": 0.08872151374816895, "grad_norm": 0.38679917007684705, "loss": 5.96683554649353, "time": 1.1457978248596192, "epoch": 90, "step": 56314}
{"lr": 0.0012363292910249734, "data_time": 0.1068427324295044, "grad_norm": 0.3714729607105255, "loss": 5.939336442947388, "time": 1.193755578994751, "epoch": 91, "step": 56440}
{"lr": 0.0012353445324183876, "data_time": 0.08630995750427246, "grad_norm": 0.33207579255104064, "loss": 5.948008394241333, "time": 1.138670563697815, "epoch": 91, "step": 56540}
{"lr": 0.0012343583353185568, "data_time": 0.10066075325012207, "grad_norm": 0.436361289024353, "loss": 5.932031488418579, "time": 1.1643973350524903, "epoch": 91, "step": 56640}
{"lr": 0.0012333707026788584, "data_time": 0.07858963012695312, "grad_norm": 0.3728587687015533, "loss": 5.911770248413086, "time": 1.1477263927459718, "epoch": 91, "step": 56740}
{"lr": 0.0012323816374569677, "data_time": 0.10055253505706788, "grad_norm": 0.36261132955551145, "loss": 5.962857341766357, "time": 1.1493150234222411, "epoch": 91, "step": 56840}
{"lr": 0.0012313911426148477, "data_time": 0.09788694381713867, "grad_norm": 0.3505276769399643, "loss": 5.9281806468963625, "time": 1.1222040891647338, "epoch": 91, "step": 56940}
{"lr": 0.00123014108821029, "data_time": 0.0683903694152832, "grad_norm": 0.5869177564978599, "loss": 5.910654973983765, "time": 1.1907387256622315, "epoch": 92, "step": 57066}
{"lr": 0.0012291473733601216, "data_time": 0.08952758312225342, "grad_norm": 0.34330395311117173, "loss": 5.92529935836792, "time": 1.145604419708252, "epoch": 92, "step": 57166}
{"lr": 0.0012281522385754117, "data_time": 0.07752470970153809, "grad_norm": 0.27478490173816683, "loss": 5.933989143371582, "time": 1.148154044151306, "epoch": 92, "step": 57266}
{"lr": 0.001227155686836304, "data_time": 0.08718793392181397, "grad_norm": 0.3144341707229614, "loss": 5.9649796962738035, "time": 1.1296645879745484, "epoch": 92, "step": 57366}
{"lr": 0.0012261577211271802, "data_time": 0.06435489654541016, "grad_norm": 0.3137108415365219, "loss": 5.980357217788696, "time": 1.1563371658325194, "epoch": 92, "step": 57466}
{"lr": 0.0012251583444366607, "data_time": 0.09903547763824463, "grad_norm": 0.4948516756296158, "loss": 5.910208511352539, "time": 1.142432737350464, "epoch": 92, "step": 57566}
{"lr": 0.0012238971254822381, "data_time": 0.09845221042633057, "grad_norm": 0.33817318081855774, "loss": 5.956662225723266, "time": 1.1717318534851073, "epoch": 93, "step": 57692}
{"lr": 0.0012228945710053298, "data_time": 0.09295485019683838, "grad_norm": 0.36547179967164994, "loss": 6.012391757965088, "time": 1.1201168060302735, "epoch": 93, "step": 57792}
{"lr": 0.00122189061531922, "data_time": 0.09575634002685547, "grad_norm": 0.46522265672683716, "loss": 5.984765529632568, "time": 1.1292947053909301, "epoch": 93, "step": 57892}
{"lr": 0.0012208852614304661, "data_time": 0.08260493278503418, "grad_norm": 0.36365339159965515, "loss": 5.950558710098266, "time": 1.1269480228424071, "epoch": 93, "step": 57992}
{"lr": 0.0012198785123498108, "data_time": 0.12538137435913085, "grad_norm": 0.4784631088376045, "loss": 5.952192068099976, "time": 1.1548020362854003, "epoch": 93, "step": 58092}
{"lr": 0.0012188703710921793, "data_time": 0.08876886367797851, "grad_norm": 0.3700883075594902, "loss": 5.9709801197052, "time": 1.1339830875396728, "epoch": 93, "step": 58192}
{"lr": 0.0012175981355975274, "data_time": 0.06608264446258545, "grad_norm": 0.49625315964221955, "loss": 5.967178869247436, "time": 1.1747071266174316, "epoch": 94, "step": 58318}
{"lr": 0.0012165868591480942, "data_time": 0.08741023540496826, "grad_norm": 0.3287467584013939, "loss": 5.968858098983764, "time": 1.137064790725708, "epoch": 94, "step": 58418}
{"lr": 0.0012155742003792346, "data_time": 0.08452904224395752, "grad_norm": 0.44487570226192474, "loss": 5.923342323303222, "time": 1.158511233329773, "epoch": 94, "step": 58518}
{"lr": 0.00121456016232357, "data_time": 0.08827452659606934, "grad_norm": 0.6868809521198272, "loss": 5.9382048606872555, "time": 1.1342451095581054, "epoch": 94, "step": 58618}
{"lr": 0.0012135447480178525, "data_time": 0.08677356243133545, "grad_norm": 0.441635961830616, "loss": 5.926968145370483, "time": 1.1207197904586792, "epoch": 94, "step": 58718}
{"lr": 0.001212527960502954, "data_time": 0.09344799518585205, "grad_norm": 0.5399691551923752, "loss": 5.951030158996582, "time": 1.1096189498901368, "epoch": 94, "step": 58818}
{"lr": 0.0012112448577705518, "data_time": 0.09382936954498292, "grad_norm": 0.48658692836761475, "loss": 6.011189031600952, "time": 1.1465787172317505, "epoch": 95, "step": 58944}
{"lr": 0.0012102249780263678, "data_time": 0.09696419239044189, "grad_norm": 0.4836925983428955, "loss": 5.917660045623779, "time": 1.1442961692810059, "epoch": 95, "step": 59044}
{"lr": 0.0012092037350147555, "data_time": 0.11537055969238282, "grad_norm": 0.3539814054965973, "loss": 5.927972602844238, "time": 1.158270788192749, "epoch": 95, "step": 59144}
{"lr": 0.0012081811317940423, "data_time": 0.10455067157745361, "grad_norm": 0.42812573313713076, "loss": 5.882105016708374, "time": 1.1189505815505982, "epoch": 95, "step": 59244}
{"lr": 0.0012071571714266306, "data_time": 0.10298213958740235, "grad_norm": 0.5906159967184067, "loss": 5.984285593032837, "time": 1.1121259927749634, "epoch": 95, "step": 59344}
{"lr": 0.001206131856978988, "data_time": 0.08372023105621337, "grad_norm": 0.5195813268423081, "loss": 5.949293899536133, "time": 1.127106499671936, "epoch": 95, "step": 59444}
{"lr": 0.0012048380375866271, "data_time": 0.06309893131256103, "grad_norm": 0.5322571843862534, "loss": 5.97384033203125, "time": 1.1484119415283203, "epoch": 96, "step": 59570}
{"lr": 0.0012038096742351015, "data_time": 0.08209271430969238, "grad_norm": 0.36288042962551115, "loss": 5.976075124740601, "time": 1.1319307804107666, "epoch": 96, "step": 59670}
{"lr": 0.0012027799668281312, "data_time": 0.07096762657165527, "grad_norm": 0.5747067421674729, "loss": 5.954782199859619, "time": 1.1606284379959106, "epoch": 96, "step": 59770}
{"lr": 0.0012017489184493939, "data_time": 0.07328112125396728, "grad_norm": 0.33428733646869657, "loss": 5.943312406539917, "time": 1.146904706954956, "epoch": 96, "step": 59870}
{"lr": 0.0012007165321865794, "data_time": 0.06481499671936035, "grad_norm": 0.6059017777442932, "loss": 5.933037805557251, "time": 1.1466857194900513, "epoch": 96, "step": 59970}
{"lr": 0.0011996828111313894, "data_time": 0.08727142810821534, "grad_norm": 0.34080035239458084, "loss": 5.920138835906982, "time": 1.1396230459213257, "epoch": 96, "step": 60070}
{"lr": 0.0011983784269145134, "data_time": 0.06986427307128906, "grad_norm": 0.3276124596595764, "loss": 5.955558443069458, "time": 1.2157765626907349, "epoch": 97, "step": 60196}
{"lr": 0.001197341700638645, "data_time": 0.09611387252807617, "grad_norm": 0.5903342217206955, "loss": 5.9260810852050785, "time": 1.1477031230926513, "epoch": 97, "step": 60296}
{"lr": 0.0011963036496770477, "data_time": 0.08066337108612061, "grad_norm": 0.3825358659029007, "loss": 5.925460910797119, "time": 1.160703682899475, "epoch": 97, "step": 60396}
{"lr": 0.0011952642771383831, "data_time": 0.0908705711364746, "grad_norm": 0.5359808936715126, "loss": 5.950384855270386, "time": 1.1641186952590943, "epoch": 97, "step": 60496}
{"lr": 0.0011942235861352704, "data_time": 0.06645991802215576, "grad_norm": 0.3850712388753891, "loss": 5.9509093284606935, "time": 1.1750927686691284, "epoch": 97, "step": 60596}
{"lr": 0.0011931815797842805, "data_time": 0.08371074199676513, "grad_norm": 0.3013857200741768, "loss": 5.906967878341675, "time": 1.1563400506973267, "epoch": 97, "step": 60696}
{"lr": 0.001191866783818159, "data_time": 0.07413926124572753, "grad_norm": 0.35881339609622953, "loss": 5.941803932189941, "time": 1.1281746625900269, "epoch": 98, "step": 60822}
{"lr": 0.0011908218162823743, "data_time": 0.07260518074035645, "grad_norm": 0.24321231096982956, "loss": 5.925480079650879, "time": 1.1236594438552856, "epoch": 98, "step": 60922}
{"lr": 0.0011897755435860318, "data_time": 0.08430569171905518, "grad_norm": 0.32409948110580444, "loss": 5.923730134963989, "time": 1.1332406759262086, "epoch": 98, "step": 61022}
{"lr": 0.0011887279688624149, "data_time": 0.08803513050079345, "grad_norm": 0.6799983829259872, "loss": 5.974217271804809, "time": 1.1292805671691895, "epoch": 98, "step": 61122}
{"lr": 0.0011876790952487087, "data_time": 0.09229478836059571, "grad_norm": 0.5106385439634323, "loss": 5.9270893096923825, "time": 1.123417901992798, "epoch": 98, "step": 61222}
{"lr": 0.0011866289258859854, "data_time": 0.08196506500244141, "grad_norm": 0.5861621499061584, "loss": 5.9399810314178465, "time": 1.1176900386810302, "epoch": 98, "step": 61322}
{"lr": 0.0011853038724677564, "data_time": 0.1328282356262207, "grad_norm": 0.27014212161302564, "loss": 5.989165639877319, "time": 1.2023267507553101, "epoch": 99, "step": 61448}
{"lr": 0.0011842507863036287, "data_time": 0.0905155897140503, "grad_norm": 0.27371449768543243, "loss": 5.9222630023956295, "time": 1.1762964487075807, "epoch": 99, "step": 61548}
{"lr": 0.001183196414657281, "data_time": 0.09053359031677247, "grad_norm": 0.3536302462220192, "loss": 5.935958290100098, "time": 1.1308477401733399, "epoch": 99, "step": 61648}
{"lr": 0.0011821407606862515, "data_time": 0.09114217758178711, "grad_norm": 0.28453782200813293, "loss": 5.986387825012207, "time": 1.1235327482223512, "epoch": 99, "step": 61748}
{"lr": 0.001181083827551919, "data_time": 0.103725266456604, "grad_norm": 0.3068633183836937, "loss": 5.944028377532959, "time": 1.1584013462066651, "epoch": 99, "step": 61848}
{"lr": 0.0011800256184194956, "data_time": 0.08178777694702148, "grad_norm": 0.4900693118572235, "loss": 6.01353006362915, "time": 1.1372230052947998, "epoch": 99, "step": 61948}
{"lr": 0.0011786904630500506, "data_time": 0.0659231424331665, "grad_norm": 0.486112454533577, "loss": 5.973079490661621, "time": 1.191157841682434, "epoch": 100, "step": 62074}
{"lr": 0.0011776293818419085, "data_time": 0.09397766590118409, "grad_norm": 0.3092531755566597, "loss": 5.926517009735107, "time": 1.1296998262405396, "epoch": 100, "step": 62174}
{"lr": 0.0011765670349807426, "data_time": 0.06774659156799316, "grad_norm": 1.0428422912955284, "loss": 5.923243761062622, "time": 1.1515358924865722, "epoch": 100, "step": 62274}
{"lr": 0.0011755034256479757, "data_time": 0.08905668258666992, "grad_norm": 0.3491449147462845, "loss": 5.943637704849243, "time": 1.1365566968917846, "epoch": 100, "step": 62374}
{"lr": 0.0011744385570288075, "data_time": 0.07215440273284912, "grad_norm": 0.4137667179107666, "loss": 5.902126836776733, "time": 1.1317317008972168, "epoch": 100, "step": 62474}
{"lr": 0.0011733724323122145, "data_time": 0.08134090900421143, "grad_norm": 0.617685715854168, "loss": 5.93717360496521, "time": 1.137228512763977, "epoch": 100, "step": 62574}
{"lr": 0.0011720273316779594, "data_time": 0.11640810966491699, "grad_norm": 0.47946045994758607, "loss": 5.896366786956787, "time": 1.1916053295135498, "epoch": 101, "step": 62700}
{"lr": 0.0011709583799483884, "data_time": 0.08991293907165528, "grad_norm": 0.37610401809215543, "loss": 5.934094619750977, "time": 1.131985116004944, "epoch": 101, "step": 62800}
{"lr": 0.001169888182543519, "data_time": 0.103340482711792, "grad_norm": 0.30453811287879945, "loss": 5.9438393115997314, "time": 1.1542717456817626, "epoch": 101, "step": 62900}
{"lr": 0.0011688167426682802, "data_time": 0.08427503108978271, "grad_norm": 0.4427429109811783, "loss": 5.9463948726654055, "time": 1.122009015083313, "epoch": 101, "step": 63000}
{"lr": 0.0011677440635313296, "data_time": 0.11014225482940673, "grad_norm": 0.3324523091316223, "loss": 5.884866714477539, "time": 1.1739118576049805, "epoch": 101, "step": 63100}
{"lr": 0.0011666701483450297, "data_time": 0.10084853172302247, "grad_norm": 0.679563769698143, "loss": 5.980484580993652, "time": 1.137212872505188, "epoch": 101, "step": 63200}
{"lr": 0.0011653152602994843, "data_time": 0.07912616729736328, "grad_norm": 1.0421788692474365, "loss": 5.93116455078125, "time": 1.1409405946731568, "epoch": 102, "step": 63326}
{"lr": 0.00116423856349471, "data_time": 0.07867755889892578, "grad_norm": 0.39770756363868714, "loss": 5.959849452972412, "time": 1.1460237979888916, "epoch": 102, "step": 63426}
{"lr": 0.0011631606411385486, "data_time": 0.10225114822387696, "grad_norm": 0.520155543088913, "loss": 5.889603233337402, "time": 1.1203138828277588, "epoch": 102, "step": 63526}
{"lr": 0.001162081496459065, "data_time": 0.06754918098449707, "grad_norm": 0.37720531672239305, "loss": 5.899972152709961, "time": 1.1179084062576294, "epoch": 102, "step": 63626}
{"lr": 0.0011610011326879852, "data_time": 0.08966705799102784, "grad_norm": 0.8352929517626763, "loss": 5.932026815414429, "time": 1.1342637538909912, "epoch": 102, "step": 63726}
{"lr": 0.0011599195530606887, "data_time": 0.07062041759490967, "grad_norm": 0.35705478489398956, "loss": 5.92777533531189, "time": 1.1231086730957032, "epoch": 102, "step": 63826}
{"lr": 0.0011585550366059727, "data_time": 0.08638536930084229, "grad_norm": 0.29777794480323794, "loss": 5.934955406188965, "time": 1.1528066158294679, "epoch": 103, "step": 63952}
{"lr": 0.0011574707210811384, "data_time": 0.08074421882629394, "grad_norm": 0.3418032795190811, "loss": 5.903954982757568, "time": 1.1349730014801025, "epoch": 103, "step": 64052}
{"lr": 0.001156385200272651, "data_time": 0.08580865859985351, "grad_norm": 0.43765868842601774, "loss": 5.953861951828003, "time": 1.1294618606567384, "epoch": 103, "step": 64152}
{"lr": 0.001155298477431332, "data_time": 0.07755608558654785, "grad_norm": 1.0225495129823685, "loss": 5.993545961380005, "time": 1.1177762985229491, "epoch": 103, "step": 64252}
{"lr": 0.001154210555811601, "data_time": 0.09856989383697509, "grad_norm": 0.3143718332052231, "loss": 5.955521059036255, "time": 1.1546024322509765, "epoch": 103, "step": 64352}
{"lr": 0.001153121438671471, "data_time": 0.09207324981689453, "grad_norm": 0.33502965420484543, "loss": 5.9154599666595455, "time": 1.1246509075164794, "epoch": 103, "step": 64452}
{"lr": 0.0011517474539396424, "data_time": 0.07591483592987061, "grad_norm": 0.46187633872032163, "loss": 5.991174936294556, "time": 1.22600257396698, "epoch": 104, "step": 64578}
{"lr": 0.0011506556469439823, "data_time": 0.0819322109222412, "grad_norm": 0.44660802781581876, "loss": 5.946576070785523, "time": 1.167148494720459, "epoch": 104, "step": 64678}
{"lr": 0.0011495626550738494, "data_time": 0.08422431945800782, "grad_norm": 0.9392472505569458, "loss": 5.936155223846436, "time": 1.1720288515090942, "epoch": 104, "step": 64778}
{"lr": 0.0011484684816024386, "data_time": 0.08637092113494874, "grad_norm": 0.563396991789341, "loss": 5.945455026626587, "time": 1.1314651727676392, "epoch": 104, "step": 64878}
{"lr": 0.0011473731298064832, "data_time": 0.06630663871765137, "grad_norm": 0.8999257445335388, "loss": 5.998994636535644, "time": 1.1530310392379761, "epoch": 104, "step": 64978}
{"lr": 0.001146276602966247, "data_time": 0.0794447660446167, "grad_norm": 0.48226363509893416, "loss": 5.919787740707397, "time": 1.115070629119873, "epoch": 104, "step": 65078}
{"lr": 0.0011448933112005054, "data_time": 0.1055765151977539, "grad_norm": 0.4421203464269638, "loss": 5.9071215152740475, "time": 1.1567619800567628, "epoch": 105, "step": 65204}
{"lr": 0.0011437941408624136, "data_time": 0.08965897560119629, "grad_norm": 0.3240388408303261, "loss": 5.9175670623779295, "time": 1.143540334701538, "epoch": 105, "step": 65304}
{"lr": 0.0011426938061980779, "data_time": 0.10080814361572266, "grad_norm": 0.2740115612745285, "loss": 5.95161395072937, "time": 1.1411247968673706, "epoch": 105, "step": 65404}
{"lr": 0.0011415923105026818, "data_time": 0.09229564666748047, "grad_norm": 0.9447798401117324, "loss": 5.96579761505127, "time": 1.1404192447662354, "epoch": 105, "step": 65504}
{"lr": 0.001140489657074891, "data_time": 0.09992399215698242, "grad_norm": 0.3327198803424835, "loss": 5.927903032302856, "time": 1.140627908706665, "epoch": 105, "step": 65604}
{"lr": 0.0011393858492168318, "data_time": 0.08334698677062988, "grad_norm": 0.33028471767902373, "loss": 5.906451797485351, "time": 1.1438207864761352, "epoch": 105, "step": 65704}
{"lr": 0.0011379934127525987, "data_time": 0.07873950004577637, "grad_norm": 0.34355160892009734, "loss": 5.954739761352539, "time": 1.154331350326538, "epoch": 106, "step": 65830}
{"lr": 0.0011368870080645842, "data_time": 0.0674081563949585, "grad_norm": 0.3194579750299454, "loss": 5.930437278747559, "time": 1.1422687768936157, "epoch": 106, "step": 65930}
{"lr": 0.001135779459735197, "data_time": 0.07020363807678223, "grad_norm": 0.6195970416069031, "loss": 5.959962987899781, "time": 1.1623843669891358, "epoch": 106, "step": 66030}
{"lr": 0.0011346707710812236, "data_time": 0.07651934623718262, "grad_norm": 0.3200709387660027, "loss": 5.9197124481201175, "time": 1.1374504089355468, "epoch": 106, "step": 66130}
{"lr": 0.0011335609454228665, "data_time": 0.07936155796051025, "grad_norm": 0.4962000086903572, "loss": 5.9278887748718265, "time": 1.1185785293579102, "epoch": 106, "step": 66230}
{"lr": 0.0011324499860837322, "data_time": 0.06809885501861572, "grad_norm": 0.39632711112499236, "loss": 5.945466756820679, "time": 1.120305585861206, "epoch": 106, "step": 66330}
{"lr": 0.0011310485683296006, "data_time": 0.10035517215728759, "grad_norm": 0.4731178179383278, "loss": 5.958231019973755, "time": 1.1432645320892334, "epoch": 107, "step": 66456}
{"lr": 0.0011299350591331614, "data_time": 0.09160208702087402, "grad_norm": 0.6207894772291184, "loss": 5.917717790603637, "time": 1.1182496309280396, "epoch": 107, "step": 66556}
{"lr": 0.0011288204271144289, "data_time": 0.10856382846832276, "grad_norm": 0.5648850455880166, "loss": 5.916716194152832, "time": 1.1596648931503295, "epoch": 107, "step": 66656}
{"lr": 0.0011277046756114054, "data_time": 0.10360019207000733, "grad_norm": 0.4099093824625015, "loss": 5.91952338218689, "time": 1.1286765098571778, "epoch": 107, "step": 66756}
{"lr": 0.001126587807965443, "data_time": 0.1185420036315918, "grad_norm": 0.32574869096279147, "loss": 5.955192136764526, "time": 1.1412009239196776, "epoch": 107, "step": 66856}
{"lr": 0.001125469827521239, "data_time": 0.08330771923065186, "grad_norm": 0.2659356534481049, "loss": 5.942582130432129, "time": 1.1068877458572388, "epoch": 107, "step": 66956}
{"lr": 0.0011240595929397957, "data_time": 0.075644850730896, "grad_norm": 0.6481514573097229, "loss": 5.962624645233154, "time": 1.1878412961959839, "epoch": 108, "step": 67082}
{"lr": 0.0011229391099101735, "data_time": 0.08037779331207276, "grad_norm": 0.5512772917747497, "loss": 5.916602754592896, "time": 1.1195967197418213, "epoch": 108, "step": 67182}
{"lr": 0.001121817525009105, "data_time": 0.0732311725616455, "grad_norm": 0.826190111041069, "loss": 5.945050001144409, "time": 1.152940535545349, "epoch": 108, "step": 67282}
{"lr": 0.0011206948415954175, "data_time": 0.07922444343566895, "grad_norm": 0.3843740448355675, "loss": 5.956867980957031, "time": 1.1117633819580077, "epoch": 108, "step": 67382}
{"lr": 0.0011195710630312198, "data_time": 0.07056884765625, "grad_norm": 0.596770292520523, "loss": 5.932462549209594, "time": 1.1383725881576539, "epoch": 108, "step": 67482}
{"lr": 0.0011184461926819064, "data_time": 0.0956766128540039, "grad_norm": 0.5368087142705917, "loss": 5.98334231376648, "time": 1.1174128532409668, "epoch": 108, "step": 67582}
{"lr": 0.001117027306770433, "data_time": 0.0848850965499878, "grad_norm": 0.518245379626751, "loss": 5.9282036304473875, "time": 1.1653519868850708, "epoch": 109, "step": 67708}
{"lr": 0.0011158999814012771, "data_time": 0.1013568639755249, "grad_norm": 0.34709242880344393, "loss": 5.956662988662719, "time": 1.1279486656188964, "epoch": 109, "step": 67808}
{"lr": 0.0011147715752408344, "data_time": 0.08578023910522461, "grad_norm": 0.6542444229125977, "loss": 5.932026720046997, "time": 1.1345453977584838, "epoch": 109, "step": 67908}
{"lr": 0.0011136420916683567, "data_time": 0.09567832946777344, "grad_norm": 0.3998682230710983, "loss": 5.922366666793823, "time": 1.1116802930831908, "epoch": 109, "step": 68008}
{"lr": 0.001112511534066323, "data_time": 0.08430843353271485, "grad_norm": 0.6208891347050667, "loss": 5.931216192245484, "time": 1.1400345802307128, "epoch": 109, "step": 68108}
{"lr": 0.0011113799058204235, "data_time": 0.09504883289337158, "grad_norm": 0.5064598709344864, "loss": 5.952439498901367, "time": 1.1245888710021972, "epoch": 109, "step": 68208}
{"lr": 0.00110995253509147, "data_time": 0.07779998779296875, "grad_norm": 0.41095715165138247, "loss": 5.876709985733032, "time": 1.167044734954834, "epoch": 110, "step": 68334}
{"lr": 0.0011088184996794128, "data_time": 0.08186616897583007, "grad_norm": 0.31425403207540514, "loss": 5.94204740524292, "time": 1.132136845588684, "epoch": 110, "step": 68434}
{"lr": 0.0011076834046830632, "data_time": 0.07575719356536866, "grad_norm": 0.6017682030797005, "loss": 5.912658262252807, "time": 1.168715810775757, "epoch": 110, "step": 68534}
{"lr": 0.0011065472535017038, "data_time": 0.0989450454711914, "grad_norm": 0.36853061318397523, "loss": 5.931354951858521, "time": 1.115494179725647, "epoch": 110, "step": 68634}
{"lr": 0.0011054100495377782, "data_time": 0.08541350364685059, "grad_norm": 0.29654675871133807, "loss": 5.907352638244629, "time": 1.1470691442489624, "epoch": 110, "step": 68734}
{"lr": 0.0011042717961968868, "data_time": 0.09546935558319092, "grad_norm": 0.6456631273031235, "loss": 5.997237205505371, "time": 1.1280537843704224, "epoch": 110, "step": 68834}
{"lr": 0.0011028361081587394, "data_time": 0.1187819004058838, "grad_norm": 0.4069757580757141, "loss": 5.929867792129516, "time": 1.1526042699813843, "epoch": 111, "step": 68960}
{"lr": 0.0011016954957878632, "data_time": 0.09342923164367675, "grad_norm": 0.4976291760802269, "loss": 5.912707042694092, "time": 1.1678747415542603, "epoch": 111, "step": 69060}
{"lr": 0.0011005538451640355, "data_time": 0.09384193420410156, "grad_norm": 0.40144018083810806, "loss": 5.974095249176026, "time": 1.1827951669692993, "epoch": 111, "step": 69160}
{"lr": 0.0010994111597061713, "data_time": 0.08825132846832276, "grad_norm": 0.4760901272296906, "loss": 5.899662494659424, "time": 1.1568395137786864, "epoch": 111, "step": 69260}
{"lr": 0.0010982674428362822, "data_time": 0.10778000354766845, "grad_norm": 0.4620590090751648, "loss": 5.9247735977172855, "time": 1.1592845916748047, "epoch": 111, "step": 69360}
{"lr": 0.0010971226979794727, "data_time": 0.08474948406219482, "grad_norm": 0.37723133713006973, "loss": 5.865742778778076, "time": 1.1031591653823853, "epoch": 111, "step": 69460}
{"lr": 0.0010956788611164875, "data_time": 0.07220532894134521, "grad_norm": 0.7228259563446044, "loss": 5.941835880279541, "time": 1.181381106376648, "epoch": 112, "step": 69586}
{"lr": 0.0010945318056427113, "data_time": 0.09054021835327149, "grad_norm": 0.5814821600914002, "loss": 5.9186852931976315, "time": 1.1395603656768798, "epoch": 112, "step": 69686}
{"lr": 0.0010933837333691686, "data_time": 0.07264721393585205, "grad_norm": 0.47404374480247496, "loss": 5.898768329620362, "time": 1.141564679145813, "epoch": 112, "step": 69786}
{"lr": 0.0010922346477340048, "data_time": 0.07062246799468994, "grad_norm": 0.4221287727355957, "loss": 5.947577285766601, "time": 1.114712905883789, "epoch": 112, "step": 69886}
{"lr": 0.0010910845521783976, "data_time": 0.0642092227935791, "grad_norm": 0.5186704486608505, "loss": 5.940918588638306, "time": 1.1660966873168945, "epoch": 112, "step": 69986}
{"lr": 0.001089933450146554, "data_time": 0.08814373016357421, "grad_norm": 0.7107293501496315, "loss": 5.968683910369873, "time": 1.1397040367126465, "epoch": 112, "step": 70086}
{"lr": 0.0010884816338993905, "data_time": 0.11234195232391357, "grad_norm": 0.6986543953418731, "loss": 5.911306571960449, "time": 1.1834341526031493, "epoch": 113, "step": 70212}
{"lr": 0.0010873282699347603, "data_time": 0.09905147552490234, "grad_norm": 0.5555307403206825, "loss": 5.909650659561157, "time": 1.158875870704651, "epoch": 113, "step": 70312}
{"lr": 0.0010861739107428747, "data_time": 0.11572475433349609, "grad_norm": 0.5377844288945198, "loss": 5.936724090576172, "time": 1.1942872524261474, "epoch": 113, "step": 70412}
{"lr": 0.0010850185597807037, "data_time": 0.09358692169189453, "grad_norm": 0.8026067614555359, "loss": 5.912623262405395, "time": 1.1455767154693604, "epoch": 113, "step": 70512}
{"lr": 0.0010838622205081932, "data_time": 0.12358827590942383, "grad_norm": 0.4927935808897018, "loss": 5.940057277679443, "time": 1.182131004333496, "epoch": 113, "step": 70612}
{"lr": 0.0010827048963882432, "data_time": 0.09118680953979492, "grad_norm": 0.6230049878358841, "loss": 5.937043237686157, "time": 1.109062671661377, "epoch": 113, "step": 70712}
{"lr": 0.0010812452711339716, "data_time": 0.06573338508605957, "grad_norm": 1.0229796141386032, "loss": 5.904368495941162, "time": 1.210599136352539, "epoch": 114, "step": 70838}
{"lr": 0.0010800857340308616, "data_time": 0.06888911724090577, "grad_norm": 0.39259577095508574, "loss": 5.92167296409607, "time": 1.1625903844833374, "epoch": 114, "step": 70938}
{"lr": 0.0010789252233898011, "data_time": 0.08824703693389893, "grad_norm": 0.5208641603589058, "loss": 5.931085252761841, "time": 1.1568600416183472, "epoch": 114, "step": 71038}
{"lr": 0.0010777637426861828, "data_time": 0.06282715797424317, "grad_norm": 0.5591172426939011, "loss": 5.886385297775268, "time": 1.1340189933776856, "epoch": 114, "step": 71138}
{"lr": 0.0010766012953983073, "data_time": 0.07720248699188233, "grad_norm": 0.44506843835115434, "loss": 5.885833120346069, "time": 1.132682704925537, "epoch": 114, "step": 71238}
{"lr": 0.001075437885007368, "data_time": 0.06894159317016602, "grad_norm": 0.5093083679676056, "loss": 5.928372240066528, "time": 1.115595293045044, "epoch": 114, "step": 71338}
{"lr": 0.0010739706220394853, "data_time": 0.11779541969299316, "grad_norm": 0.431011438369751, "loss": 5.910376644134521, "time": 1.1660981893539428, "epoch": 115, "step": 71464}
{"lr": 0.0010728050478747172, "data_time": 0.08547859191894532, "grad_norm": 0.47829402089118955, "loss": 5.925011348724365, "time": 1.1149681568145753, "epoch": 115, "step": 71564}
{"lr": 0.001071638521975552, "data_time": 0.10596108436584473, "grad_norm": 0.3880864351987839, "loss": 5.924763917922974, "time": 1.145044732093811, "epoch": 115, "step": 71664}
{"lr": 0.001070471047835397, "data_time": 0.0865250825881958, "grad_norm": 0.29533391147851945, "loss": 5.912916803359986, "time": 1.1558835744857787, "epoch": 115, "step": 71764}
{"lr": 0.001069302628950502, "data_time": 0.09481639862060547, "grad_norm": 0.35269360095262525, "loss": 5.931954431533813, "time": 1.1413608551025392, "epoch": 115, "step": 71864}
{"lr": 0.0010681332688199428, "data_time": 0.0797996997833252, "grad_norm": 0.5554694831371307, "loss": 5.960912466049194, "time": 1.159062933921814, "epoch": 115, "step": 71964}
{"lr": 0.0010666585403282637, "data_time": 0.06841435432434081, "grad_norm": 0.28938051164150236, "loss": 5.934980964660644, "time": 1.1556399822235108, "epoch": 116, "step": 72090}
{"lr": 0.001065487065887133, "data_time": 0.08290960788726806, "grad_norm": 0.3236757144331932, "loss": 5.890507316589355, "time": 1.1436251878738404, "epoch": 116, "step": 72190}
{"lr": 0.0010643146616268501, "data_time": 0.07700016498565673, "grad_norm": 0.30623178482055663, "loss": 5.935060214996338, "time": 1.162510919570923, "epoch": 116, "step": 72290}
{"lr": 0.0010631413310584234, "data_time": 0.10060760974884034, "grad_norm": 0.5558295682072639, "loss": 5.945681428909301, "time": 1.1279229879379273, "epoch": 116, "step": 72390}
{"lr": 0.0010619670776956425, "data_time": 0.0863128662109375, "grad_norm": 0.44053888618946074, "loss": 5.947464513778686, "time": 1.1473335027694702, "epoch": 116, "step": 72490}
{"lr": 0.0010607919050550582, "data_time": 0.08858981132507324, "grad_norm": 0.41179691851139066, "loss": 5.926612901687622, "time": 1.1344764947891235, "epoch": 116, "step": 72590}
{"lr": 0.00105930988410551, "data_time": 0.08672680854797363, "grad_norm": 0.5044178321957589, "loss": 5.947739887237549, "time": 1.1589092493057251, "epoch": 117, "step": 72716}
{"lr": 0.0010581326468657357, "data_time": 0.08938012123107911, "grad_norm": 0.28527630865573883, "loss": 5.945118188858032, "time": 1.130141305923462, "epoch": 117, "step": 72816}
{"lr": 0.00105695450183117, "data_time": 0.10365550518035889, "grad_norm": 0.48949515372514724, "loss": 5.9603455543518065, "time": 1.1722837686538696, "epoch": 117, "step": 72916}
{"lr": 0.0010557754525300192, "data_time": 0.08268611431121826, "grad_norm": 0.40186732113361356, "loss": 5.916364049911499, "time": 1.1211796998977661, "epoch": 117, "step": 73016}
{"lr": 0.0010545955024931942, "data_time": 0.09748263359069824, "grad_norm": 0.6185775026679039, "loss": 5.949509811401367, "time": 1.1695311307907104, "epoch": 117, "step": 73116}
{"lr": 0.001053414655254304, "data_time": 0.08871209621429443, "grad_norm": 0.4199476480484009, "loss": 5.888955545425415, "time": 1.1329704761505126, "epoch": 117, "step": 73216}
{"lr": 0.001051925515768617, "data_time": 0.07767341136932374, "grad_norm": 0.4263496816158295, "loss": 5.950114107131958, "time": 1.188562536239624, "epoch": 118, "step": 73342}
{"lr": 0.0010507426538842104, "data_time": 0.06514492034912109, "grad_norm": 0.33460338711738585, "loss": 5.919772672653198, "time": 1.1566632509231567, "epoch": 118, "step": 73442}
{"lr": 0.0010495589063359077, "data_time": 0.07116594314575195, "grad_norm": 0.7597144216299057, "loss": 5.916442012786865, "time": 1.1807637691497803, "epoch": 118, "step": 73542}
{"lr": 0.0010483742766686927, "data_time": 0.09136171340942383, "grad_norm": 0.5352236300706863, "loss": 5.886375474929809, "time": 1.1549029111862184, "epoch": 118, "step": 73642}
{"lr": 0.0010471887684301903, "data_time": 0.08430249691009521, "grad_norm": 0.5026347622275352, "loss": 5.982009935379028, "time": 1.1916983604431153, "epoch": 118, "step": 73742}
{"lr": 0.0010460023851706535, "data_time": 0.08533782958984375, "grad_norm": 0.41074643433094027, "loss": 6.1150397777557375, "time": 1.141736388206482, "epoch": 118, "step": 73842}
{"lr": 0.0010445063019059453, "data_time": 0.08813049793243408, "grad_norm": 0.5038836777210236, "loss": 5.92188868522644, "time": 1.1538963556289672, "epoch": 119, "step": 73968}
{"lr": 0.0010433179541909907, "data_time": 0.08250966072082519, "grad_norm": 0.5439974695444107, "loss": 5.90074200630188, "time": 1.1259888887405396, "epoch": 119, "step": 74068}
{"lr": 0.0010421287430469747, "data_time": 0.10349576473236084, "grad_norm": 0.5909752935171128, "loss": 5.980953788757324, "time": 1.1397988319396972, "epoch": 119, "step": 74168}
{"lr": 0.0010409386720352403, "data_time": 0.09597251415252686, "grad_norm": 0.32846508026123045, "loss": 5.869764375686645, "time": 1.1363802909851075, "epoch": 119, "step": 74268}
{"lr": 0.0010397477447197077, "data_time": 0.08851087093353271, "grad_norm": 0.336596354842186, "loss": 5.882306337356567, "time": 1.1401272058486938, "epoch": 119, "step": 74368}
{"lr": 0.0010385559646668607, "data_time": 0.09541654586791992, "grad_norm": 0.3774871572852135, "loss": 5.932938861846924, "time": 1.1225779056549072, "epoch": 119, "step": 74468}
{"lr": 0.0010370531131951395, "data_time": 0.07743136882781983, "grad_norm": 0.43759627640247345, "loss": 5.907608509063721, "time": 1.1468849658966065, "epoch": 120, "step": 74594}
{"lr": 0.001035859419107512, "data_time": 0.09028072357177734, "grad_norm": 0.38472740054130555, "loss": 5.923024845123291, "time": 1.126123571395874, "epoch": 120, "step": 74694}
{"lr": 0.0010346648839269822, "data_time": 0.06408865451812744, "grad_norm": 0.3952073633670807, "loss": 5.902151012420655, "time": 1.1385752439498902, "epoch": 120, "step": 74794}
{"lr": 0.001033469511230838, "data_time": 0.09496963024139404, "grad_norm": 0.40789293199777604, "loss": 5.900078868865966, "time": 1.1533782720565795, "epoch": 120, "step": 74894}
{"lr": 0.0010322733045988767, "data_time": 0.08610780239105224, "grad_norm": 0.6773358106613159, "loss": 5.915967226028442, "time": 1.1695374727249146, "epoch": 120, "step": 74994}
{"lr": 0.0010310762676133914, "data_time": 0.0823289155960083, "grad_norm": 0.42807742655277253, "loss": 5.8497151851654055, "time": 1.1362483978271485, "epoch": 120, "step": 75094}
{"lr": 0.0010295668243009387, "data_time": 0.11569757461547851, "grad_norm": 0.6286475569009781, "loss": 5.9059008121490475, "time": 1.1857511043548583, "epoch": 121, "step": 75220}
{"lr": 0.001028367923925931, "data_time": 0.09478087425231933, "grad_norm": 0.2920087769627571, "loss": 5.912559700012207, "time": 1.1694184303283692, "epoch": 121, "step": 75320}
{"lr": 0.0010271682048928884, "data_time": 0.11339766979217529, "grad_norm": 0.38752754479646684, "loss": 5.913249778747558, "time": 1.15744206905365, "epoch": 121, "step": 75420}
{"lr": 0.001025967670794622, "data_time": 0.1011385440826416, "grad_norm": 0.39091664254665376, "loss": 5.936518144607544, "time": 1.141672921180725, "epoch": 121, "step": 75520}
{"lr": 0.001024766325226384, "data_time": 0.08670048713684082, "grad_norm": 0.3850811094045639, "loss": 5.937835693359375, "time": 1.1687981843948365, "epoch": 121, "step": 75620}
{"lr": 0.0010235641717858581, "data_time": 0.09482638835906983, "grad_norm": 0.5073703676462173, "loss": 5.953851556777954, "time": 1.1417778253555297, "epoch": 121, "step": 75720}
{"lr": 0.001022048313772541, "data_time": 0.07128317356109619, "grad_norm": 0.43990194648504255, "loss": 5.906078672409057, "time": 1.1696685314178468, "epoch": 122, "step": 75846}
{"lr": 0.001020844347806429, "data_time": 0.07501485347747802, "grad_norm": 0.8374699532985688, "loss": 5.943559408187866, "time": 1.1530627965927125, "epoch": 122, "step": 75946}
{"lr": 0.0010196395857132207, "data_time": 0.08391129970550537, "grad_norm": 0.5804040998220443, "loss": 5.935866832733154, "time": 1.1703419923782348, "epoch": 122, "step": 76046}
{"lr": 0.0010184340311008304, "data_time": 0.07134475708007812, "grad_norm": 0.35195859521627426, "loss": 5.905844354629517, "time": 1.1475171089172362, "epoch": 122, "step": 76146}
{"lr": 0.0010172276875795446, "data_time": 0.08524103164672851, "grad_norm": 0.3625198662281036, "loss": 5.957522439956665, "time": 1.1742786407470702, "epoch": 122, "step": 76246}
{"lr": 0.0010160205587620161, "data_time": 0.07130682468414307, "grad_norm": 0.3603613153100014, "loss": 5.927192306518554, "time": 1.148768949508667, "epoch": 122, "step": 76346}
{"lr": 0.0010144984639404923, "data_time": 0.1046821117401123, "grad_norm": 0.36685008406639097, "loss": 5.951968050003051, "time": 1.1587380409240722, "epoch": 123, "step": 76472}
{"lr": 0.0010132895736740192, "data_time": 0.08231151103973389, "grad_norm": 0.29722344130277634, "loss": 5.879330635070801, "time": 1.123805046081543, "epoch": 123, "step": 76572}
{"lr": 0.0010120799099048183, "data_time": 0.11255803108215331, "grad_norm": 0.47108888924121856, "loss": 5.985146522521973, "time": 1.1571702003479003, "epoch": 123, "step": 76672}
{"lr": 0.0010108694762554834, "data_time": 0.07974417209625244, "grad_norm": 0.7678628265857697, "loss": 5.928579902648925, "time": 1.1181922197341918, "epoch": 123, "step": 76772}
{"lr": 0.0010096582763509106, "data_time": 0.1073720932006836, "grad_norm": 0.5819731369614601, "loss": 5.897527027130127, "time": 1.1466592311859132, "epoch": 123, "step": 76872}
{"lr": 0.0010084463138182974, "data_time": 0.09128053188323974, "grad_norm": 0.3332034945487976, "loss": 5.922654056549073, "time": 1.1137380838394164, "epoch": 123, "step": 76972}
{"lr": 0.0010069181608131433, "data_time": 0.07137889862060547, "grad_norm": 0.3378681629896164, "loss": 5.895267581939697, "time": 1.1815085887908936, "epoch": 124, "step": 77098}
{"lr": 0.0010057044881149441, "data_time": 0.0805596113204956, "grad_norm": 0.46687051057815554, "loss": 5.926430654525757, "time": 1.177817153930664, "epoch": 124, "step": 77198}
{"lr": 0.0010044900646291573, "data_time": 0.0680769920349121, "grad_norm": 0.29645767956972124, "loss": 5.941996145248413, "time": 1.1919397115707397, "epoch": 124, "step": 77298}
{"lr": 0.0010032748939926305, "data_time": 0.09884178638458252, "grad_norm": 0.415161107480526, "loss": 5.888551044464111, "time": 1.1395784854888915, "epoch": 124, "step": 77398}
{"lr": 0.0010020589798444496, "data_time": 0.09685366153717041, "grad_norm": 0.3930174052715302, "loss": 5.882475090026856, "time": 1.1865612268447876, "epoch": 124, "step": 77498}
{"lr": 0.0010008423258259277, "data_time": 0.09936685562133789, "grad_norm": 0.4047203689813614, "loss": 5.958216857910156, "time": 1.1402431726455688, "epoch": 124, "step": 77598}
{"lr": 0.0009993082939726721, "data_time": 0.10712435245513915, "grad_norm": 0.36382413655519485, "loss": 5.913980102539062, "time": 1.1528018236160278, "epoch": 125, "step": 77724}
{"lr": 0.0009980899812726154, "data_time": 0.08746845722198486, "grad_norm": 0.40924117565155027, "loss": 5.968903923034668, "time": 1.1170271635055542, "epoch": 125, "step": 77824}
{"lr": 0.0009968709405882236, "data_time": 0.10065691471099854, "grad_norm": 0.340419040620327, "loss": 5.934813070297241, "time": 1.155091118812561, "epoch": 125, "step": 77924}
{"lr": 0.0009956511755701684, "data_time": 0.07672491073608398, "grad_norm": 0.3976607173681259, "loss": 5.925721216201782, "time": 1.1159748077392577, "epoch": 125, "step": 78024}
{"lr": 0.000994430689871297, "data_time": 0.09409406185150146, "grad_norm": 0.42786200642585753, "loss": 5.915234899520874, "time": 1.128198766708374, "epoch": 125, "step": 78124}
{"lr": 0.000993209487146611, "data_time": 0.08089895248413086, "grad_norm": 0.3751335173845291, "loss": 5.879275608062744, "time": 1.1092721939086914, "epoch": 125, "step": 78224}
{"lr": 0.0009916697564706948, "data_time": 0.06702210903167724, "grad_norm": 0.6537502452731132, "loss": 5.894894456863403, "time": 1.1767783403396606, "epoch": 126, "step": 78350}
{"lr": 0.0009904469467431796, "data_time": 0.06959722042083741, "grad_norm": 0.3927721858024597, "loss": 5.899837970733643, "time": 1.1269218683242799, "epoch": 126, "step": 78450}
{"lr": 0.000989223431920011, "data_time": 0.07438952922821045, "grad_norm": 0.35489792227745054, "loss": 5.917544221878051, "time": 1.1348533630371094, "epoch": 126, "step": 78550}
{"lr": 0.000987999215665263, "data_time": 0.06503973007202149, "grad_norm": 0.41535941064357756, "loss": 5.935659885406494, "time": 1.1058036088943481, "epoch": 126, "step": 78650}
{"lr": 0.0009867743016451094, "data_time": 0.09158263206481934, "grad_norm": 0.6002070307731628, "loss": 5.9436994075775145, "time": 1.1301995754241942, "epoch": 126, "step": 78750}
{"lr": 0.0009855486935278106, "data_time": 0.06414854526519775, "grad_norm": 0.48252182006835936, "loss": 5.971200847625733, "time": 1.1180040836334229, "epoch": 126, "step": 78850}
{"lr": 0.0009840034447234482, "data_time": 0.11547420024871827, "grad_norm": 0.2600523725152016, "loss": 5.9044664859771725, "time": 1.1412708520889283, "epoch": 127, "step": 78976}
{"lr": 0.000982776281470613, "data_time": 0.11187961101531982, "grad_norm": 0.5596750110387803, "loss": 5.866733741760254, "time": 1.1438423156738282, "epoch": 127, "step": 79076}
{"lr": 0.0009815484360935517, "data_time": 0.10778648853302002, "grad_norm": 0.4010975569486618, "loss": 5.919249439239502, "time": 1.1402872800827026, "epoch": 127, "step": 79176}
{"lr": 0.0009803199122693086, "data_time": 0.08831355571746827, "grad_norm": 0.33485729098320005, "loss": 5.865493202209473, "time": 1.134370756149292, "epoch": 127, "step": 79276}
{"lr": 0.0009790907136769582, "data_time": 0.10594844818115234, "grad_norm": 0.34725817292928696, "loss": 5.894652080535889, "time": 1.1557974815368652, "epoch": 127, "step": 79376}
{"lr": 0.0009778608439975941, "data_time": 0.08428094387054444, "grad_norm": 0.45069547146558764, "loss": 5.984815073013306, "time": 1.1213261842727662, "epoch": 127, "step": 79476}
{"lr": 0.0009763102584065912, "data_time": 0.08089110851287842, "grad_norm": 0.6088568359613419, "loss": 5.9327045440673825, "time": 1.1794360399246215, "epoch": 128, "step": 79602}
{"lr": 0.000975078885641495, "data_time": 0.08238942623138427, "grad_norm": 0.9871360659599304, "loss": 5.918376302719116, "time": 1.1160174131393432, "epoch": 128, "step": 79702}
{"lr": 0.000973846853803652, "data_time": 0.06728272438049317, "grad_norm": 0.3988438159227371, "loss": 5.872703552246094, "time": 1.1849466562271118, "epoch": 128, "step": 79802}
{"lr": 0.0009726141665826429, "data_time": 0.10049898624420166, "grad_norm": 1.098926618695259, "loss": 6.106503200531006, "time": 1.1517240047454833, "epoch": 128, "step": 79902}
{"lr": 0.0009713808276700091, "data_time": 0.06365933418273925, "grad_norm": 0.4859109356999397, "loss": 5.928013706207276, "time": 1.1748884677886964, "epoch": 128, "step": 80002}
{"lr": 0.000970146840759246, "data_time": 0.09386835098266602, "grad_norm": 0.4667504459619522, "loss": 5.891679525375366, "time": 1.1275855779647828, "epoch": 128, "step": 80102}
{"lr": 0.0009685911003497035, "data_time": 0.10145399570465088, "grad_norm": 0.5708036571741104, "loss": 5.95439772605896, "time": 1.1462546586990356, "epoch": 129, "step": 80228}
{"lr": 0.0009673556625793934, "data_time": 0.08337202072143554, "grad_norm": 0.45115804970264434, "loss": 5.915782499313354, "time": 1.1272203207015992, "epoch": 129, "step": 80328}
{"lr": 0.0009661195888651617, "data_time": 0.10079185962677002, "grad_norm": 0.3797899648547173, "loss": 5.913655376434326, "time": 1.1427541017532348, "epoch": 129, "step": 80428}
{"lr": 0.0009648828829086924, "data_time": 0.08781552314758301, "grad_norm": 0.25576244592666625, "loss": 5.971766614913941, "time": 1.1357442378997802, "epoch": 129, "step": 80528}
{"lr": 0.0009636455484135617, "data_time": 0.10336227416992187, "grad_norm": 0.526859974861145, "loss": 5.901565361022949, "time": 1.1451031923294068, "epoch": 129, "step": 80628}
{"lr": 0.0009624075890852299, "data_time": 0.08431482315063477, "grad_norm": 0.6472901910543442, "loss": 5.968143701553345, "time": 1.1140771150588988, "epoch": 129, "step": 80728}
{"lr": 0.0009608468764301752, "data_time": 0.09199559688568115, "grad_norm": 0.3257821574807167, "loss": 5.932780838012695, "time": 1.1908706665039062, "epoch": 130, "step": 80854}
{"lr": 0.0009596075186387516, "data_time": 0.07233214378356934, "grad_norm": 0.5264076083898545, "loss": 5.878485012054443, "time": 1.1410110712051391, "epoch": 130, "step": 80954}
{"lr": 0.0009583675481068612, "data_time": 0.10068013668060302, "grad_norm": 0.3421544522047043, "loss": 5.891832590103149, "time": 1.1399464368820191, "epoch": 130, "step": 81054}
{"lr": 0.0009571269685478574, "data_time": 0.07325081825256348, "grad_norm": 0.5612343490123749, "loss": 5.948489046096801, "time": 1.1176229953765868, "epoch": 130, "step": 81154}
{"lr": 0.000955885783676918, "data_time": 0.08176267147064209, "grad_norm": 0.79799744784832, "loss": 5.915041065216064, "time": 1.1190478801727295, "epoch": 130, "step": 81254}
{"lr": 0.0009546439972110343, "data_time": 0.08519284725189209, "grad_norm": 0.6224112987518311, "loss": 5.940141010284424, "time": 1.1155405521392823, "epoch": 130, "step": 81354}
{"lr": 0.0009530784954670178, "data_time": 0.09798192977905273, "grad_norm": 0.6906657487154007, "loss": 5.905481767654419, "time": 1.1965864419937133, "epoch": 131, "step": 81480}
{"lr": 0.000951835363098613, "data_time": 0.0959003210067749, "grad_norm": 0.34832354784011843, "loss": 5.935010242462158, "time": 1.1729482650756835, "epoch": 131, "step": 81580}
{"lr": 0.0009505916412651035, "data_time": 0.09962978363037109, "grad_norm": 0.8498586505651474, "loss": 5.906704044342041, "time": 1.1538807153701782, "epoch": 131, "step": 81680}
{"lr": 0.0009493473336910786, "data_time": 0.08885684013366699, "grad_norm": 0.3136249080300331, "loss": 5.884167289733886, "time": 1.1634082555770875, "epoch": 131, "step": 81780}
{"lr": 0.0009481024441028796, "data_time": 0.1017683506011963, "grad_norm": 0.5993318229913711, "loss": 5.890771436691284, "time": 1.183905053138733, "epoch": 131, "step": 81880}
{"lr": 0.0009468569762285902, "data_time": 0.10167584419250489, "grad_norm": 0.37654411494731904, "loss": 5.925649118423462, "time": 1.1565048456192017, "epoch": 131, "step": 81980}
{"lr": 0.0009452868691141825, "data_time": 0.07720398902893066, "grad_norm": 0.35083127319812774, "loss": 5.880148696899414, "time": 1.1407610177993774, "epoch": 132, "step": 82106}
{"lr": 0.0009440401080558924, "data_time": 0.08158965110778808, "grad_norm": 0.6374706566333771, "loss": 5.887694883346557, "time": 1.1006965637207031, "epoch": 132, "step": 82206}
{"lr": 0.000942792780877037, "data_time": 0.07859611511230469, "grad_norm": 0.38576226085424425, "loss": 5.914789915084839, "time": 1.107600784301758, "epoch": 132, "step": 82306}
{"lr": 0.000941544891313001, "data_time": 0.10334572792053223, "grad_norm": 0.6551582127809524, "loss": 5.961748743057251, "time": 1.1053674936294555, "epoch": 132, "step": 82406}
{"lr": 0.0009402964431008533, "data_time": 0.06621365547180176, "grad_norm": 0.3192993700504303, "loss": 5.887858629226685, "time": 1.1236868381500245, "epoch": 132, "step": 82506}
{"lr": 0.0009390474399793369, "data_time": 0.09310643672943116, "grad_norm": 0.4313513159751892, "loss": 5.860831260681152, "time": 1.1188383102416992, "epoch": 132, "step": 82606}
{"lr": 0.0009374729117535692, "data_time": 0.0966782808303833, "grad_norm": 0.4131302639842033, "loss": 5.932148551940918, "time": 1.1171242237091064, "epoch": 133, "step": 82732}
{"lr": 0.0009362226683183306, "data_time": 0.07892446517944336, "grad_norm": 0.2606526389718056, "loss": 5.916907787322998, "time": 1.118739700317383, "epoch": 133, "step": 82832}
{"lr": 0.0009349718821735024, "data_time": 0.101338791847229, "grad_norm": 0.33395712822675705, "loss": 5.892584180831909, "time": 1.1312951564788818, "epoch": 133, "step": 82932}
{"lr": 0.0009337205570648283, "data_time": 0.08396198749542236, "grad_norm": 0.2988636165857315, "loss": 5.897963142395019, "time": 1.0880598545074462, "epoch": 133, "step": 83032}
{"lr": 0.0009324686967396654, "data_time": 0.09814481735229492, "grad_norm": 0.4460457444190979, "loss": 5.8676214694976805, "time": 1.0994723558425903, "epoch": 133, "step": 83132}
{"lr": 0.0009312163049469753, "data_time": 0.08161699771881104, "grad_norm": 0.324621719121933, "loss": 5.909614038467407, "time": 1.091481375694275, "epoch": 133, "step": 83232}
{"lr": 0.0009296375403877174, "data_time": 0.08231899738311768, "grad_norm": 0.654061472415924, "loss": 5.929091596603394, "time": 1.1908572912216187, "epoch": 134, "step": 83358}
{"lr": 0.0009283839612971411, "data_time": 0.09678409099578858, "grad_norm": 0.29179663211107254, "loss": 5.934166765213012, "time": 1.1821057796478271, "epoch": 134, "step": 83458}
{"lr": 0.0009271298629716402, "data_time": 0.07014009952545167, "grad_norm": 0.3319045230746269, "loss": 5.886166191101074, "time": 1.1737110137939453, "epoch": 134, "step": 83558}
{"lr": 0.000925875249166878, "data_time": 0.0760951042175293, "grad_norm": 0.7889124304056168, "loss": 5.869593763351441, "time": 1.1396313667297364, "epoch": 134, "step": 83658}
{"lr": 0.0009246201236400603, "data_time": 0.06384923458099365, "grad_norm": 0.5497846364974975, "loss": 5.907565689086914, "time": 1.1566388607025146, "epoch": 134, "step": 83758}
{"lr": 0.0009233644901499263, "data_time": 0.07788872718811035, "grad_norm": 0.3737520158290863, "loss": 5.9565176486969, "time": 1.1469618082046509, "epoch": 134, "step": 83858}
{"lr": 0.0009217816745321981, "data_time": 0.11108925342559814, "grad_norm": 0.42153449952602384, "loss": 5.866285896301269, "time": 1.1741820573806763, "epoch": 135, "step": 83984}
{"lr": 0.0009205249068993485, "data_time": 0.08540105819702148, "grad_norm": 0.3330084338784218, "loss": 5.8943640232086185, "time": 1.160816216468811, "epoch": 135, "step": 84084}
{"lr": 0.0009192676435671744, "data_time": 0.09846479892730713, "grad_norm": 0.6366879671812058, "loss": 5.887965154647827, "time": 1.1691054821014404, "epoch": 135, "step": 84184}
{"lr": 0.000918009888300817, "data_time": 0.09824175834655761, "grad_norm": 0.5213485598564148, "loss": 5.869602632522583, "time": 1.1422744989395142, "epoch": 135, "step": 84284}
{"lr": 0.0009167516448668896, "data_time": 0.10478768348693848, "grad_norm": 0.5036669909954071, "loss": 5.878853273391724, "time": 1.1507387638092041, "epoch": 135, "step": 84384}
{"lr": 0.0009154929170334679, "data_time": 0.07974355220794678, "grad_norm": 0.49828323125839236, "loss": 5.914030075073242, "time": 1.1367658138275147, "epoch": 135, "step": 84484}
{"lr": 0.0009139062361077, "data_time": 0.05732958316802979, "grad_norm": 0.29592266082763674, "loss": 5.909069395065307, "time": 1.1801461219787597, "epoch": 136, "step": 84610}
{"lr": 0.0009126464274198308, "data_time": 0.08806865215301514, "grad_norm": 0.6451803565025329, "loss": 5.950252437591553, "time": 1.1282743453979491, "epoch": 136, "step": 84710}
{"lr": 0.0009113861466264092, "data_time": 0.06355361938476563, "grad_norm": 0.39954535067081454, "loss": 5.898391246795654, "time": 1.1603251218795776, "epoch": 136, "step": 84810}
{"lr": 0.0009101253975016132, "data_time": 0.08001816272735596, "grad_norm": 0.3401047602295876, "loss": 5.890069437026978, "time": 1.1421869039535522, "epoch": 136, "step": 84910}
{"lr": 0.000908864183821023, "data_time": 0.06534602642059326, "grad_norm": 0.4293471872806549, "loss": 5.896468448638916, "time": 1.1268984794616699, "epoch": 136, "step": 85010}
{"lr": 0.0009076025093616073, "data_time": 0.09730041027069092, "grad_norm": 0.5392979726195335, "loss": 5.914101457595825, "time": 1.1170035123825073, "epoch": 136, "step": 85110}
{"lr": 0.0009060121493318409, "data_time": 0.09161980152130127, "grad_norm": 0.41061772108078004, "loss": 5.933100652694702, "time": 1.1628484010696412, "epoch": 137, "step": 85236}
{"lr": 0.0009047494474330877, "data_time": 0.09313020706176758, "grad_norm": 0.48455446362495425, "loss": 5.909909105300903, "time": 1.1430241584777832, "epoch": 137, "step": 85336}
{"lr": 0.0009034862970779583, "data_time": 0.08313593864440919, "grad_norm": 0.38851273357868193, "loss": 5.875559139251709, "time": 1.15244562625885, "epoch": 137, "step": 85436}
{"lr": 0.0009022227020492235, "data_time": 0.08611376285552978, "grad_norm": 0.8453392803668975, "loss": 5.93798861503601, "time": 1.133792018890381, "epoch": 137, "step": 85536}
{"lr": 0.0009009586661309851, "data_time": 0.09136483669281006, "grad_norm": 0.43737626522779466, "loss": 5.93598484992981, "time": 1.152688193321228, "epoch": 137, "step": 85636}
{"lr": 0.0008996941931086677, "data_time": 0.09146244525909424, "grad_norm": 0.4558751553297043, "loss": 5.914357423782349, "time": 1.1428255081176757, "epoch": 137, "step": 85736}
{"lr": 0.0008981003406107033, "data_time": 0.07378029823303223, "grad_norm": 0.5562750399112701, "loss": 5.928148651123047, "time": 1.1758991956710816, "epoch": 138, "step": 85862}
{"lr": 0.0008968348936847328, "data_time": 0.08640737533569336, "grad_norm": 0.3099683836102486, "loss": 5.873336744308472, "time": 1.1242082834243774, "epoch": 138, "step": 85962}
{"lr": 0.0008955690220041894, "data_time": 0.0769991159439087, "grad_norm": 0.5143577665090561, "loss": 5.904338645935058, "time": 1.170441699028015, "epoch": 138, "step": 86062}
{"lr": 0.0008943027293599955, "data_time": 0.08838546276092529, "grad_norm": 0.4566510319709778, "loss": 5.8715461730957035, "time": 1.1325636863708497, "epoch": 138, "step": 86162}
{"lr": 0.0008930360195443304, "data_time": 0.0751699686050415, "grad_norm": 0.26201545596122744, "loss": 5.925997924804688, "time": 1.1568641424179078, "epoch": 138, "step": 86262}
{"lr": 0.0008917688963506249, "data_time": 0.08071589469909668, "grad_norm": 0.38898957818746566, "loss": 5.9054888725280765, "time": 1.1270731449127198, "epoch": 138, "step": 86362}
{"lr": 0.0008901717384301211, "data_time": 0.11275279521942139, "grad_norm": 0.40614586174488065, "loss": 5.864133739471436, "time": 1.1950520753860474, "epoch": 139, "step": 86488}
{"lr": 0.0008889036949827394, "data_time": 0.08688347339630127, "grad_norm": 0.520568060874939, "loss": 5.883603811264038, "time": 1.1447613716125489, "epoch": 139, "step": 86588}
{"lr": 0.000887635250532437, "data_time": 0.10171620845794678, "grad_norm": 0.3606201454997063, "loss": 5.942994594573975, "time": 1.1432355642318726, "epoch": 139, "step": 86688}
{"lr": 0.000886366408877839, "data_time": 0.1040921926498413, "grad_norm": 0.25478361248970033, "loss": 5.804792833328247, "time": 1.1269776582717896, "epoch": 139, "step": 86788}
{"lr": 0.00088509717381876, "data_time": 0.09683623313903808, "grad_norm": 0.597423329949379, "loss": 5.911824321746826, "time": 1.176517128944397, "epoch": 139, "step": 86888}
{"lr": 0.0008838275491561914, "data_time": 0.09154608249664306, "grad_norm": 0.33144939243793486, "loss": 5.904957437515259, "time": 1.1353056192398072, "epoch": 139, "step": 86988}
{"lr": 0.0008822272732467114, "data_time": 0.06531836986541747, "grad_norm": 0.6781396925449371, "loss": 5.8748393058776855, "time": 1.1474101781845092, "epoch": 140, "step": 87114}
{"lr": 0.0008809567820884412, "data_time": 0.08258509635925293, "grad_norm": 1.0440288871526717, "loss": 5.878601980209351, "time": 1.112859034538269, "epoch": 140, "step": 87214}
{"lr": 0.0008796859137259603, "data_time": 0.0776484489440918, "grad_norm": 0.2774754285812378, "loss": 5.938889265060425, "time": 1.142921018600464, "epoch": 140, "step": 87314}
{"lr": 0.0008784146719651531, "data_time": 0.09125206470489503, "grad_norm": 0.4268113374710083, "loss": 5.901406908035279, "time": 1.1043914794921874, "epoch": 140, "step": 87414}
{"lr": 0.000877143060613022, "data_time": 0.0772352933883667, "grad_norm": 0.3651388108730316, "loss": 5.88587498664856, "time": 1.1152520418167113, "epoch": 140, "step": 87514}
{"lr": 0.0008758710834776753, "data_time": 0.0830613136291504, "grad_norm": 0.5745868891477585, "loss": 5.912789821624756, "time": 1.097874689102173, "epoch": 140, "step": 87614}
{"lr": 0.0008742678773786915, "data_time": 0.07399544715881348, "grad_norm": 0.6121982961893082, "loss": 5.942958879470825, "time": 1.15999276638031, "epoch": 141, "step": 87740}
{"lr": 0.0008729950876073023, "data_time": 0.08503067493438721, "grad_norm": 0.5025985524058342, "loss": 5.953808212280274, "time": 1.1201706171035766, "epoch": 141, "step": 87840}
{"lr": 0.0008717219444746792, "data_time": 0.08781211376190186, "grad_norm": 0.5402039468288422, "loss": 5.900206422805786, "time": 1.1483599185943603, "epoch": 141, "step": 87940}
{"lr": 0.0008704484517935197, "data_time": 0.08599629402160644, "grad_norm": 0.6685911834239959, "loss": 5.8857090950012205, "time": 1.122741150856018, "epoch": 141, "step": 88040}
{"lr": 0.0008691746133775659, "data_time": 0.07911942005157471, "grad_norm": 0.3210131272673607, "loss": 5.925947284698486, "time": 1.111537480354309, "epoch": 141, "step": 88140}
{"lr": 0.0008679004330415974, "data_time": 0.0840682029724121, "grad_norm": 0.3958491876721382, "loss": 5.94044713973999, "time": 1.1202955007553101, "epoch": 141, "step": 88240}
{"lr": 0.0008662944848964549, "data_time": 0.09801731109619141, "grad_norm": 0.2627044230699539, "loss": 5.925823593139649, "time": 1.117617630958557, "epoch": 142, "step": 88366}
{"lr": 0.0008650195458794697, "data_time": 0.08388664722442626, "grad_norm": 0.3068695396184921, "loss": 5.889929628372192, "time": 1.1450360536575317, "epoch": 142, "step": 88466}
{"lr": 0.0008637442773856963, "data_time": 0.07399301528930664, "grad_norm": 0.3166705906391144, "loss": 5.874868154525757, "time": 1.0999254941940309, "epoch": 142, "step": 88566}
{"lr": 0.0008624686832341965, "data_time": 0.06762943267822266, "grad_norm": 0.5358074992895127, "loss": 5.925902271270752, "time": 1.1249626636505128, "epoch": 142, "step": 88666}
{"lr": 0.0008611927672450067, "data_time": 0.09350795745849609, "grad_norm": 0.4633476033806801, "loss": 5.9232401847839355, "time": 1.1169226169586182, "epoch": 142, "step": 88766}
{"lr": 0.0008599165332391271, "data_time": 0.06871094703674316, "grad_norm": 0.4034370183944702, "loss": 5.913910722732544, "time": 1.1170114278793335, "epoch": 142, "step": 88866}
{"lr": 0.0008583080315129603, "data_time": 0.11684820652008057, "grad_norm": 0.30823927372694016, "loss": 5.8948948860168455, "time": 1.1786498785018922, "epoch": 143, "step": 88992}
{"lr": 0.0008570310928701246, "data_time": 0.07837097644805908, "grad_norm": 0.489545513689518, "loss": 5.8563331127166744, "time": 1.1431438684463502, "epoch": 143, "step": 89092}
{"lr": 0.0008557538486736133, "data_time": 0.11795330047607422, "grad_norm": 0.43461879789829255, "loss": 5.877287149429321, "time": 1.1358987808227539, "epoch": 143, "step": 89192}
{"lr": 0.0008544763027484026, "data_time": 0.08229928016662598, "grad_norm": 0.791984960436821, "loss": 5.886487102508545, "time": 1.1264714241027831, "epoch": 143, "step": 89292}
{"lr": 0.0008531984589203747, "data_time": 0.11178369522094726, "grad_norm": 0.3190664738416672, "loss": 5.897650194168091, "time": 1.1525726079940797, "epoch": 143, "step": 89392}
{"lr": 0.0008519203210163034, "data_time": 0.08049592971801758, "grad_norm": 0.36420334279537203, "loss": 5.902570199966431, "time": 1.1268996715545654, "epoch": 143, "step": 89492}
{"lr": 0.0008503094544739219, "data_time": 0.06284394264221191, "grad_norm": 0.5801370680332184, "loss": 5.904312229156494, "time": 1.1349682807922363, "epoch": 144, "step": 89618}
{"lr": 0.0008490306660596479, "data_time": 0.08499839305877685, "grad_norm": 0.3605352252721786, "loss": 5.889072275161743, "time": 1.1436068534851074, "epoch": 144, "step": 89718}
{"lr": 0.0008477515960506672, "data_time": 0.08438396453857422, "grad_norm": 0.4335325941443443, "loss": 5.889867162704467, "time": 1.1403260707855225, "epoch": 144, "step": 89818}
{"lr": 0.0008464722482774249, "data_time": 0.06972510814666748, "grad_norm": 0.5197192579507828, "loss": 5.883781480789184, "time": 1.127281904220581, "epoch": 144, "step": 89918}
{"lr": 0.000845192626571199, "data_time": 0.06748974323272705, "grad_norm": 0.44976538717746734, "loss": 5.9153436660766605, "time": 1.1657143831253052, "epoch": 144, "step": 90018}
{"lr": 0.0008439127347640868, "data_time": 0.08690493106842041, "grad_norm": 0.40053500831127165, "loss": 5.916880512237549, "time": 1.1562398433685304, "epoch": 144, "step": 90118}
{"lr": 0.0008422996924478203, "data_time": 0.09038527011871338, "grad_norm": 0.4402675122022629, "loss": 5.901292228698731, "time": 1.1112262725830078, "epoch": 145, "step": 90244}
{"lr": 0.0008410192043335987, "data_time": 0.08474798202514648, "grad_norm": 0.35018030405044553, "loss": 5.884569883346558, "time": 1.0912941455841065, "epoch": 145, "step": 90344}
{"lr": 0.000839738458616686, "data_time": 0.08934566974639893, "grad_norm": 0.5537538021802902, "loss": 5.912517642974853, "time": 1.1324899435043334, "epoch": 145, "step": 90444}
{"lr": 0.000838457459132545, "data_time": 0.08680636882781982, "grad_norm": 0.616409295797348, "loss": 5.930874633789062, "time": 1.1222151041030883, "epoch": 145, "step": 90544}
{"lr": 0.0008371762097174002, "data_time": 0.1001274585723877, "grad_norm": 0.38670057952404024, "loss": 5.906041526794434, "time": 1.1111004590988158, "epoch": 145, "step": 90644}
{"lr": 0.0008358947142082242, "data_time": 0.10490083694458008, "grad_norm": 0.4094926193356514, "loss": 5.902013301849365, "time": 1.1366746187210084, "epoch": 145, "step": 90744}
{"lr": 0.0008342796854157419, "data_time": 0.07495934963226318, "grad_norm": 0.40794195532798766, "loss": 5.902117300033569, "time": 1.1795050859451295, "epoch": 146, "step": 90870}
{"lr": 0.0008329976478725301, "data_time": 0.06955149173736572, "grad_norm": 0.4401905983686447, "loss": 5.896510887145996, "time": 1.1688705921173095, "epoch": 146, "step": 90970}
{"lr": 0.0008317153767488735, "data_time": 0.07365617752075196, "grad_norm": 0.4306128680706024, "loss": 5.9480503559112545, "time": 1.1645455837249756, "epoch": 146, "step": 91070}
{"lr": 0.0008304328758848039, "data_time": 0.06885068416595459, "grad_norm": 0.6097143143415451, "loss": 5.9222818374633786, "time": 1.1298475027084351, "epoch": 146, "step": 91170}
{"lr": 0.0008291501491210411, "data_time": 0.08185329437255859, "grad_norm": 0.504659554362297, "loss": 5.911089849472046, "time": 1.1242857456207276, "epoch": 146, "step": 91270}
{"lr": 0.0008278672002989822, "data_time": 0.08566434383392334, "grad_norm": 0.47603140473365785, "loss": 5.855150556564331, "time": 1.1224889039993287, "epoch": 146, "step": 91370}
{"lr": 0.000826250374561074, "data_time": 0.1114351511001587, "grad_norm": 0.46985897421836853, "loss": 5.906177282333374, "time": 1.20157949924469, "epoch": 147, "step": 91496}
{"lr": 0.0008249669380416616, "data_time": 0.08799073696136475, "grad_norm": 0.5384791910648346, "loss": 5.867704629898071, "time": 1.1520915508270264, "epoch": 147, "step": 91596}
{"lr": 0.0008236832919914628, "data_time": 0.11393935680389404, "grad_norm": 0.37772473096847536, "loss": 5.885032367706299, "time": 1.1896771430969237, "epoch": 147, "step": 91696}
{"lr": 0.000822399440254627, "data_time": 0.08590013980865478, "grad_norm": 0.36551922410726545, "loss": 5.8704259395599365, "time": 1.1423407793045044, "epoch": 147, "step": 91796}
{"lr": 0.00082111538667592, "data_time": 0.11364026069641113, "grad_norm": 0.41723415851593015, "loss": 5.860180521011353, "time": 1.1782943248748778, "epoch": 147, "step": 91896}
{"lr": 0.0008198311351007104, "data_time": 0.08575608730316162, "grad_norm": 0.3415259137749672, "loss": 5.904793310165405, "time": 1.139191222190857, "epoch": 147, "step": 91996}
{"lr": 0.0008182127021590441, "data_time": 0.07369890213012695, "grad_norm": 0.41166655719280243, "loss": 5.898732662200928, "time": 1.169845199584961, "epoch": 148, "step": 92122}
{"lr": 0.0008169280172803959, "data_time": 0.08974993228912354, "grad_norm": 0.5158905059099197, "loss": 5.885541534423828, "time": 1.1473474502563477, "epoch": 148, "step": 92222}
{"lr": 0.00081564314694521, "data_time": 0.08370037078857422, "grad_norm": 0.3288813531398773, "loss": 5.9132246494293215, "time": 1.188268232345581, "epoch": 148, "step": 92322}
{"lr": 0.0008143580950013016, "data_time": 0.08184099197387695, "grad_norm": 0.4155571237206459, "loss": 5.898439550399781, "time": 1.149541473388672, "epoch": 148, "step": 92422}
{"lr": 0.0008130728652970297, "data_time": 0.07222528457641601, "grad_norm": 0.6529416292905807, "loss": 5.855322217941284, "time": 1.1738258600234985, "epoch": 148, "step": 92522}
{"lr": 0.0008117874616812862, "data_time": 0.08709149360656739, "grad_norm": 0.3529802069067955, "loss": 5.8945225238800045, "time": 1.130035161972046, "epoch": 148, "step": 92622}
{"lr": 0.0008101676114661439, "data_time": 0.0929133415222168, "grad_norm": 0.5559948235750198, "loss": 5.927177667617798, "time": 1.188599705696106, "epoch": 149, "step": 92748}
{"lr": 0.0008088818289917291, "data_time": 0.08886640071868897, "grad_norm": 0.609112274646759, "loss": 5.897275400161743, "time": 1.1551589488983154, "epoch": 149, "step": 92848}
{"lr": 0.0008075958851567876, "data_time": 0.08853850364685059, "grad_norm": 0.3689275667071342, "loss": 5.857785558700561, "time": 1.1734303951263427, "epoch": 149, "step": 92948}
{"lr": 0.000806309783812349, "data_time": 0.08878648281097412, "grad_norm": 0.9147683471441269, "loss": 5.905228614807129, "time": 1.141190528869629, "epoch": 149, "step": 93048}
{"lr": 0.0008050235288099159, "data_time": 0.09113340377807617, "grad_norm": 0.39652461260557176, "loss": 5.920036315917969, "time": 1.1697908163070678, "epoch": 149, "step": 93148}
{"lr": 0.0008037371240014516, "data_time": 0.09169607162475586, "grad_norm": 0.8994909465312958, "loss": 5.900497245788574, "time": 1.1463976860046388, "epoch": 149, "step": 93248}
{"lr": 0.0008021160466094377, "data_time": 0.0865635633468628, "grad_norm": 0.4196804970502853, "loss": 5.897960805892945, "time": 1.188970136642456, "epoch": 150, "step": 93374}
{"lr": 0.0008008293174315287, "data_time": 0.06397538185119629, "grad_norm": 0.5173692882061005, "loss": 5.917107439041137, "time": 1.181291937828064, "epoch": 150, "step": 93474}
{"lr": 0.0007995424510080418, "data_time": 0.08747155666351318, "grad_norm": 0.39292745292186737, "loss": 5.891694736480713, "time": 1.1662384033203126, "epoch": 150, "step": 93574}
{"lr": 0.0007982554511927679, "data_time": 0.07122688293457032, "grad_norm": 0.5389650136232376, "loss": 5.8912938117980955, "time": 1.131688904762268, "epoch": 150, "step": 93674}
{"lr": 0.0007969683218399021, "data_time": 0.08714864253997803, "grad_norm": 0.29233588129281995, "loss": 5.905285310745239, "time": 1.1402824640274047, "epoch": 150, "step": 93774}
{"lr": 0.0007956810668040242, "data_time": 0.07687363624572754, "grad_norm": 0.6205848306417465, "loss": 5.836964464187622, "time": 1.1317736625671386, "epoch": 150, "step": 93874}
{"lr": 0.0007940589524757581, "data_time": 0.12397146224975586, "grad_norm": 0.3416769042611122, "loss": 5.885630893707275, "time": 1.1475542068481446, "epoch": 151, "step": 94000}
{"lr": 0.0007927714275977296, "data_time": 0.11332559585571289, "grad_norm": 0.6170224756002426, "loss": 5.903843212127685, "time": 1.1280392169952393, "epoch": 151, "step": 94100}
{"lr": 0.0007914837896051782, "data_time": 0.10387425422668457, "grad_norm": 0.40189307034015653, "loss": 5.851130628585816, "time": 1.1451342821121215, "epoch": 151, "step": 94200}
{"lr": 0.0007901960423542055, "data_time": 0.09708867073059083, "grad_norm": 0.44003673195838927, "loss": 5.912279224395752, "time": 1.131386947631836, "epoch": 151, "step": 94300}
{"lr": 0.0007889081897012432, "data_time": 0.1050790786743164, "grad_norm": 0.4032346248626709, "loss": 5.9149562358856205, "time": 1.1475008249282836, "epoch": 151, "step": 94400}
{"lr": 0.0007876202355030387, "data_time": 0.10094385147094727, "grad_norm": 0.35531261563301086, "loss": 5.889270973205567, "time": 1.1306964635849, "epoch": 151, "step": 94500}
{"lr": 0.0007859972746008295, "data_time": 0.07085556983947754, "grad_norm": 0.6958285510540009, "loss": 5.88708906173706, "time": 1.1663403749465941, "epoch": 152, "step": 94626}
{"lr": 0.0007847091051194344, "data_time": 0.08382225036621094, "grad_norm": 0.3951941251754761, "loss": 5.873230695724487, "time": 1.1530879735946655, "epoch": 152, "step": 94726}
{"lr": 0.0007834208466678437, "data_time": 0.06689765453338622, "grad_norm": 0.5789601504802704, "loss": 5.9113517761230465, "time": 1.1403009653091432, "epoch": 152, "step": 94826}
{"lr": 0.0007821325031040204, "data_time": 0.08254859447479249, "grad_norm": 1.1097056895494462, "loss": 5.866238403320312, "time": 1.1345666170120239, "epoch": 152, "step": 94926}
{"lr": 0.0007808440782861808, "data_time": 0.07385115623474121, "grad_norm": 0.3826011329889297, "loss": 5.874586915969848, "time": 1.159176516532898, "epoch": 152, "step": 95026}
{"lr": 0.0007795555760727848, "data_time": 0.07965126037597656, "grad_norm": 0.5052998274564743, "loss": 5.891992568969727, "time": 1.1307729482650757, "epoch": 152, "step": 95126}
{"lr": 0.0007779319590582908, "data_time": 0.11325089931488037, "grad_norm": 0.4209845572710037, "loss": 5.858003425598144, "time": 1.2002219915390016, "epoch": 153, "step": 95252}
{"lr": 0.0007766432961459296, "data_time": 0.09274013042449951, "grad_norm": 0.4287016525864601, "loss": 5.871334028244019, "time": 1.159742522239685, "epoch": 153, "step": 95352}
{"lr": 0.0007753545684181425, "data_time": 0.0922853946685791, "grad_norm": 0.42903703451156616, "loss": 5.918087577819824, "time": 1.165820264816284, "epoch": 153, "step": 95452}
{"lr": 0.0007740657797342974, "data_time": 0.08780598640441895, "grad_norm": 0.45672300159931184, "loss": 5.87124924659729, "time": 1.1374908208847045, "epoch": 153, "step": 95552}
{"lr": 0.000772776933953945, "data_time": 0.10272133350372314, "grad_norm": 0.44896192848682404, "loss": 5.858778381347657, "time": 1.1486265182495117, "epoch": 153, "step": 95652}
{"lr": 0.0007714880349368057, "data_time": 0.09231624603271485, "grad_norm": 0.5267978608608246, "loss": 5.897035551071167, "time": 1.1535175561904907, "epoch": 153, "step": 95752}
{"lr": 0.00076986395234868, "data_time": 0.06394484043121337, "grad_norm": 0.3071231782436371, "loss": 5.845596218109131, "time": 1.1671384811401366, "epoch": 154, "step": 95878}
{"lr": 0.000768574947235659, "data_time": 0.07613625526428222, "grad_norm": 0.45853247940540315, "loss": 5.890467786788941, "time": 1.1332758903503417, "epoch": 154, "step": 95978}
{"lr": 0.0007672859014695883, "data_time": 0.06938095092773437, "grad_norm": 0.37645198702812194, "loss": 5.851026201248169, "time": 1.1491792678833008, "epoch": 154, "step": 96078}
{"lr": 0.0007659968189107888, "data_time": 0.08432841300964355, "grad_norm": 0.3020784676074982, "loss": 5.871330070495605, "time": 1.1241849899291991, "epoch": 154, "step": 96178}
{"lr": 0.0007647077034196907, "data_time": 0.08410096168518066, "grad_norm": 0.5947802603244782, "loss": 5.887246465682983, "time": 1.1521095037460327, "epoch": 154, "step": 96278}
{"lr": 0.0007634185588568219, "data_time": 0.07155120372772217, "grad_norm": 0.34975470006465914, "loss": 5.892618989944458, "time": 1.1250992298126221, "epoch": 154, "step": 96378}
{"lr": 0.0007617942012883587, "data_time": 0.08668713569641114, "grad_norm": 0.44109864830970763, "loss": 5.851242446899414, "time": 1.1813165426254273, "epoch": 155, "step": 96504}
{"lr": 0.0007605050052451426, "data_time": 0.08471040725708008, "grad_norm": 0.49508627206087114, "loss": 5.922236728668213, "time": 1.1549827575683593, "epoch": 155, "step": 96604}
{"lr": 0.000759215792716024, "data_time": 0.09663877487182618, "grad_norm": 0.6379734963178635, "loss": 5.88875560760498, "time": 1.1629010677337646, "epoch": 155, "step": 96704}
{"lr": 0.0007579265675618229, "data_time": 0.08432693481445312, "grad_norm": 0.6211353883147239, "loss": 5.849482011795044, "time": 1.1366520404815674, "epoch": 155, "step": 96804}
{"lr": 0.0007566373336433972, "data_time": 0.09243710041046142, "grad_norm": 0.513187363743782, "loss": 5.8820513725280765, "time": 1.1635618209838867, "epoch": 155, "step": 96904}
{"lr": 0.0007553480948216295, "data_time": 0.10140843391418457, "grad_norm": 0.34778564274311063, "loss": 5.9228438377380375, "time": 1.149403738975525, "epoch": 155, "step": 97004}
{"lr": 0.0007537236528983913, "data_time": 0.07311322689056396, "grad_norm": 0.511077082157135, "loss": 5.884862232208252, "time": 1.1601097583770752, "epoch": 156, "step": 97130}
{"lr": 0.0007524344172178525, "data_time": 0.09218099117279052, "grad_norm": 0.359463769197464, "loss": 5.898964071273804, "time": 1.1361580610275268, "epoch": 156, "step": 97230}
{"lr": 0.0007511451892204928, "data_time": 0.08356025218963622, "grad_norm": 0.6493773937225342, "loss": 5.874616193771362, "time": 1.1565164089202882, "epoch": 156, "step": 97330}
{"lr": 0.0007498559727671791, "data_time": 0.0848158836364746, "grad_norm": 0.6221873998641968, "loss": 5.866672658920288, "time": 1.145080018043518, "epoch": 156, "step": 97430}
{"lr": 0.0007485667717187426, "data_time": 0.06950089931488038, "grad_norm": 0.5482774987816811, "loss": 5.854631853103638, "time": 1.129721736907959, "epoch": 156, "step": 97530}
{"lr": 0.0007472775899359681, "data_time": 0.08127114772796631, "grad_norm": 0.4527749717235565, "loss": 5.861706113815307, "time": 1.1050116062164306, "epoch": 156, "step": 97630}
{"lr": 0.0007456532542934165, "data_time": 0.08057684898376465, "grad_norm": 0.43088763058185575, "loss": 5.914872169494629, "time": 1.1530169725418091, "epoch": 157, "step": 97756}
{"lr": 0.0007443641302730771, "data_time": 0.0884969711303711, "grad_norm": 0.5260224893689156, "loss": 5.865060615539551, "time": 1.1480329036712646, "epoch": 157, "step": 97856}
{"lr": 0.0007430750381040983, "data_time": 0.09877829551696778, "grad_norm": 0.6511211216449737, "loss": 5.867560863494873, "time": 1.1540220737457276, "epoch": 157, "step": 97956}
{"lr": 0.0007417859816469378, "data_time": 0.08734607696533203, "grad_norm": 0.33504813015460966, "loss": 5.8802735805511475, "time": 1.1624089002609252, "epoch": 157, "step": 98056}
{"lr": 0.000740496964761949, "data_time": 0.0933373212814331, "grad_norm": 0.33389939218759535, "loss": 5.891502761840821, "time": 1.1574238061904907, "epoch": 157, "step": 98156}
{"lr": 0.0007392079913093653, "data_time": 0.08725235462188721, "grad_norm": 0.4498338758945465, "loss": 5.8101396560668945, "time": 1.1407601356506347, "epoch": 157, "step": 98256}
{"lr": 0.0007375839525704906, "data_time": 0.07358067035675049, "grad_norm": 0.5693616598844529, "loss": 5.893869543075562, "time": 1.173857283592224, "epoch": 158, "step": 98382}
{"lr": 0.0007362950914947714, "data_time": 0.08539028167724609, "grad_norm": 0.48495678305625917, "loss": 5.854817008972168, "time": 1.1318801403045655, "epoch": 158, "step": 98482}
{"lr": 0.0007350062864348542, "data_time": 0.07620484828948974, "grad_norm": 0.36233885735273363, "loss": 5.895419645309448, "time": 1.1567935466766357, "epoch": 158, "step": 98582}
{"lr": 0.0007337175412503383, "data_time": 0.08742468357086182, "grad_norm": 0.3125539720058441, "loss": 5.879114294052124, "time": 1.1304647207260132, "epoch": 158, "step": 98682}
{"lr": 0.0007324288598006442, "data_time": 0.08558170795440674, "grad_norm": 0.8475690603256225, "loss": 5.924400758743286, "time": 1.1563384532928467, "epoch": 158, "step": 98782}
{"lr": 0.000731140245945001, "data_time": 0.0798201322555542, "grad_norm": 0.663231936097145, "loss": 5.8922830581665036, "time": 1.1143884897232055, "epoch": 158, "step": 98882}
{"lr": 0.000729516694697949, "data_time": 0.11650478839874268, "grad_norm": 0.3779069572687149, "loss": 5.8780145168304445, "time": 1.1519742965698243, "epoch": 159, "step": 99008}
{"lr": 0.0007282282478204109, "data_time": 0.09855515956878662, "grad_norm": 0.3560379043221474, "loss": 5.8358509063720705, "time": 1.1343944787979126, "epoch": 159, "step": 99108}
{"lr": 0.0007269398811165441, "data_time": 0.12304220199584961, "grad_norm": 0.4167697697877884, "loss": 5.880685091018677, "time": 1.1548065423965455, "epoch": 159, "step": 99208}
{"lr": 0.0007256515984446341, "data_time": 0.08114569187164307, "grad_norm": 0.7397151470184327, "loss": 5.856280994415283, "time": 1.1288419246673584, "epoch": 159, "step": 99308}
{"lr": 0.0007243634036627168, "data_time": 0.10948772430419922, "grad_norm": 0.332868418097496, "loss": 5.9163525104522705, "time": 1.158082127571106, "epoch": 159, "step": 99408}
{"lr": 0.0007230753006285636, "data_time": 0.08294808864593506, "grad_norm": 0.49912774115800856, "loss": 5.894221925735474, "time": 1.1208430051803588, "epoch": 159, "step": 99508}
{"lr": 0.0007214524274042727, "data_time": 0.0826646327972412, "grad_norm": 0.27389969527721403, "loss": 5.838910245895386, "time": 1.2045281887054444, "epoch": 160, "step": 99634}
{"lr": 0.000720164545929869, "data_time": 0.08675339221954345, "grad_norm": 0.914960652589798, "loss": 5.90774941444397, "time": 1.1705343008041382, "epoch": 160, "step": 99734}
{"lr": 0.0007188767687775968, "data_time": 0.08675453662872315, "grad_norm": 0.5760587215423584, "loss": 5.89190149307251, "time": 1.1554473400115968, "epoch": 160, "step": 99834}
{"lr": 0.000717589099803977, "data_time": 0.0766831874847412, "grad_norm": 0.3628993943333626, "loss": 5.855901288986206, "time": 1.144657564163208, "epoch": 160, "step": 99934}
{"lr": 0.0007163015428652059, "data_time": 0.08087148666381835, "grad_norm": 0.43924602270126345, "loss": 5.857441806793213, "time": 1.1632855415344239, "epoch": 160, "step": 100034}
{"lr": 0.000715014101817145, "data_time": 0.08793332576751708, "grad_norm": 0.35774108916521075, "loss": 5.9069726943969725, "time": 1.1235724210739135, "epoch": 160, "step": 100134}
{"lr": 0.0007133920970669827, "data_time": 0.11570022106170655, "grad_norm": 0.4439970925450325, "loss": 5.845342445373535, "time": 1.1549962759017944, "epoch": 161, "step": 100260}
{"lr": 0.000712104932134315, "data_time": 0.07919354438781738, "grad_norm": 0.2787479028105736, "loss": 5.880828857421875, "time": 1.1311481237411498, "epoch": 161, "step": 100360}
{"lr": 0.0007108178956599966, "data_time": 0.09016542434692383, "grad_norm": 0.4979416072368622, "loss": 5.907213497161865, "time": 1.1496177911758423, "epoch": 161, "step": 100460}
{"lr": 0.0007095309914983297, "data_time": 0.08986027240753174, "grad_norm": 0.2885790094733238, "loss": 5.899783945083618, "time": 1.1203002691268922, "epoch": 161, "step": 100560}
{"lr": 0.0007082442235032195, "data_time": 0.08950104713439941, "grad_norm": 0.4110834449529648, "loss": 5.888212490081787, "time": 1.14398136138916, "epoch": 161, "step": 100660}
{"lr": 0.0007069575955281665, "data_time": 0.10102677345275879, "grad_norm": 0.6815917015075683, "loss": 5.896526861190796, "time": 1.138930368423462, "epoch": 161, "step": 100760}
{"lr": 0.000705336649601582, "data_time": 0.07121524810791016, "grad_norm": 0.3664647191762924, "loss": 5.90227861404419, "time": 1.148197841644287, "epoch": 162, "step": 100886}
{"lr": 0.0007040503522651627, "data_time": 0.06992936134338379, "grad_norm": 0.3889915198087692, "loss": 5.8900350570678714, "time": 1.1271373271942138, "epoch": 162, "step": 100986}
{"lr": 0.0007027642075082345, "data_time": 0.08980367183685303, "grad_norm": 0.3493741363286972, "loss": 5.874885702133179, "time": 1.1619338750839234, "epoch": 162, "step": 101086}
{"lr": 0.0007014782191824298, "data_time": 0.07989866733551025, "grad_norm": 0.36902746856212615, "loss": 5.890432214736938, "time": 1.1451083660125732, "epoch": 162, "step": 101186}
{"lr": 0.0007001923911389115, "data_time": 0.08357770442962646, "grad_norm": 0.4131565481424332, "loss": 5.8853826999664305, "time": 1.1668404817581177, "epoch": 162, "step": 101286}
{"lr": 0.0006989067272283633, "data_time": 0.0818213939666748, "grad_norm": 0.5154852628707886, "loss": 5.878996992111206, "time": 1.124135708808899, "epoch": 162, "step": 101386}
{"lr": 0.0006972870303505481, "data_time": 0.08402822017669678, "grad_norm": 1.0253054946660995, "loss": 5.851195478439331, "time": 1.2091378450393677, "epoch": 163, "step": 101512}
{"lr": 0.0006960017515630728, "data_time": 0.10142600536346436, "grad_norm": 0.32522352039813995, "loss": 5.866663932800293, "time": 1.1633235931396484, "epoch": 163, "step": 101612}
{"lr": 0.0006947166494583243, "data_time": 0.10236246585845947, "grad_norm": 0.36014376282691957, "loss": 5.87786602973938, "time": 1.1840185880661012, "epoch": 163, "step": 101712}
{"lr": 0.0006934317278848121, "data_time": 0.08417460918426514, "grad_norm": 0.4334037184715271, "loss": 5.907054281234741, "time": 1.1664806127548217, "epoch": 163, "step": 101812}
{"lr": 0.0006921469906905062, "data_time": 0.10343258380889893, "grad_norm": 0.5749285578727722, "loss": 5.86801815032959, "time": 1.1786268949508667, "epoch": 163, "step": 101912}
{"lr": 0.000690862441722823, "data_time": 0.08406269550323486, "grad_norm": 0.3568458199501038, "loss": 5.867608070373535, "time": 1.1468806743621827, "epoch": 163, "step": 102012}
{"lr": 0.0006892441839723868, "data_time": 0.08306772708892822, "grad_norm": 0.40298030972480775, "loss": 5.87237319946289, "time": 1.1968066453933717, "epoch": 164, "step": 102138}
{"lr": 0.0006879600745670194, "data_time": 0.079372239112854, "grad_norm": 0.5177748262882232, "loss": 5.891255235671997, "time": 1.125289535522461, "epoch": 164, "step": 102238}
{"lr": 0.0006866761659268803, "data_time": 0.07504920959472657, "grad_norm": 0.466918820142746, "loss": 5.874371767044067, "time": 1.1572590112686156, "epoch": 164, "step": 102338}
{"lr": 0.0006853924618969059, "data_time": 0.08233435153961181, "grad_norm": 0.5641774088144302, "loss": 5.886456823348999, "time": 1.1377572774887086, "epoch": 164, "step": 102438}
{"lr": 0.0006841089663214185, "data_time": 0.0649259328842163, "grad_norm": 0.3493661344051361, "loss": 5.880020666122436, "time": 1.1562971591949462, "epoch": 164, "step": 102538}
{"lr": 0.0006828256830441173, "data_time": 0.09000163078308106, "grad_norm": 0.6607194155454635, "loss": 5.857368278503418, "time": 1.1416564226150512, "epoch": 164, "step": 102638}
{"lr": 0.0006812090543307823, "data_time": 0.10878610610961914, "grad_norm": 0.39382045418024064, "loss": 5.836415767669678, "time": 1.169282603263855, "epoch": 165, "step": 102764}
{"lr": 0.0006799262650034547, "data_time": 0.08597526550292969, "grad_norm": 0.4840283542871475, "loss": 5.855339860916137, "time": 1.1129881381988525, "epoch": 165, "step": 102864}
{"lr": 0.000678643700500297, "data_time": 0.11473631858825684, "grad_norm": 0.37019642889499665, "loss": 5.899655532836914, "time": 1.1604430437088014, "epoch": 165, "step": 102964}
{"lr": 0.0006773613646622191, "data_time": 0.10027930736541749, "grad_norm": 0.6722755342721939, "loss": 5.876105403900146, "time": 1.1263354063034057, "epoch": 165, "step": 103064}
{"lr": 0.0006760792613294467, "data_time": 0.10186228752136231, "grad_norm": 0.4282017767429352, "loss": 5.896460151672363, "time": 1.1466156721115113, "epoch": 165, "step": 103164}
{"lr": 0.0006747973943415102, "data_time": 0.07667620182037353, "grad_norm": 0.4435056269168854, "loss": 5.915565443038941, "time": 1.110688066482544, "epoch": 165, "step": 103264}
{"lr": 0.000673182584383824, "data_time": 0.07731711864471436, "grad_norm": 0.6167363226413727, "loss": 5.830650901794433, "time": 1.1484325885772706, "epoch": 166, "step": 103390}
{"lr": 0.0006719012656755529, "data_time": 0.08235280513763428, "grad_norm": 0.2834465354681015, "loss": 5.855179929733277, "time": 1.1185412883758545, "epoch": 166, "step": 103490}
{"lr": 0.0006706201958240076, "data_time": 0.06912057399749756, "grad_norm": 0.4694878160953522, "loss": 5.905605792999268, "time": 1.1123679876327515, "epoch": 166, "step": 103590}
{"lr": 0.0006693393786656236, "data_time": 0.07902677059173584, "grad_norm": 0.4507208362221718, "loss": 5.852618932723999, "time": 1.1211504459381103, "epoch": 166, "step": 103690}
{"lr": 0.0006680588180360776, "data_time": 0.07397019863128662, "grad_norm": 0.4236129879951477, "loss": 5.8642418384552, "time": 1.1150142192840575, "epoch": 166, "step": 103790}
{"lr": 0.0006667785177702801, "data_time": 0.08565044403076172, "grad_norm": 0.49689535796642303, "loss": 5.844017601013183, "time": 1.1184376001358032, "epoch": 166, "step": 103890}
{"lr": 0.0006651657160733505, "data_time": 0.092630934715271, "grad_norm": 0.7625216156244278, "loss": 5.8388608455657955, "time": 1.1795263051986695, "epoch": 167, "step": 104016}
{"lr": 0.0006638860183525659, "data_time": 0.08358800411224365, "grad_norm": 0.3504440009593964, "loss": 5.8764214515686035, "time": 1.1386086702346803, "epoch": 167, "step": 104116}
{"lr": 0.0006626065934918601, "data_time": 0.08915953636169434, "grad_norm": 0.45279273837804795, "loss": 5.906772089004517, "time": 1.154382848739624, "epoch": 167, "step": 104216}
{"lr": 0.0006613274453227411, "data_time": 0.08330249786376953, "grad_norm": 0.4753339529037476, "loss": 5.894889259338379, "time": 1.1034315824508667, "epoch": 167, "step": 104316}
{"lr": 0.0006600485776758888, "data_time": 0.09295432567596436, "grad_norm": 0.7345453947782516, "loss": 5.892112970352173, "time": 1.1488099813461303, "epoch": 167, "step": 104416}
{"lr": 0.0006587699943811429, "data_time": 0.09575827121734619, "grad_norm": 0.5436973750591279, "loss": 5.893951034545898, "time": 1.1274191379547118, "epoch": 167, "step": 104516}
{"lr": 0.0006571593902143999, "data_time": 0.06931395530700683, "grad_norm": 0.36472645998001096, "loss": 5.880357027053833, "time": 1.1788370609283447, "epoch": 168, "step": 104642}
{"lr": 0.0006558814636593034, "data_time": 0.08426718711853028, "grad_norm": 0.3969034180045128, "loss": 5.870892858505249, "time": 1.1575241804122924, "epoch": 168, "step": 104742}
{"lr": 0.0006546038339356173, "data_time": 0.07954626083374024, "grad_norm": 0.3622688174247742, "loss": 5.888128852844238, "time": 1.1662017107009888, "epoch": 168, "step": 104842}
{"lr": 0.0006533265048694717, "data_time": 0.09688777923583984, "grad_norm": 0.4707462951540947, "loss": 5.859631204605103, "time": 1.1465051174163818, "epoch": 168, "step": 104942}
{"lr": 0.0006520494802861017, "data_time": 0.07420468330383301, "grad_norm": 0.9288780063390731, "loss": 5.868044137954712, "time": 1.1369807481765748, "epoch": 168, "step": 105042}
{"lr": 0.0006507727640098266, "data_time": 0.08224592208862305, "grad_norm": 0.3171519339084625, "loss": 5.859716558456421, "time": 1.1433688163757325, "epoch": 168, "step": 105142}
{"lr": 0.0006491645463848099, "data_time": 0.10511324405670167, "grad_norm": 0.3068121403455734, "loss": 5.905925369262695, "time": 1.1533085584640503, "epoch": 169, "step": 105268}
{"lr": 0.000647888540965749, "data_time": 0.074153733253479, "grad_norm": 0.7088098376989365, "loss": 5.832067632675171, "time": 1.1502386808395386, "epoch": 169, "step": 105368}
{"lr": 0.000646612856314595, "data_time": 0.09408609867095948, "grad_norm": 0.38736526668071747, "loss": 5.897410821914673, "time": 1.1358576536178588, "epoch": 169, "step": 105468}
{"lr": 0.0006453374962516546, "data_time": 0.08693492412567139, "grad_norm": 0.40908791720867155, "loss": 5.885287714004517, "time": 1.1534065008163452, "epoch": 169, "step": 105568}
{"lr": 0.0006440624645962632, "data_time": 0.10059399604797363, "grad_norm": 0.6294166684150696, "loss": 5.86735897064209, "time": 1.1506470680236816, "epoch": 169, "step": 105668}
{"lr": 0.0006427877651667737, "data_time": 0.09066028594970703, "grad_norm": 0.40406317114830015, "loss": 5.795055389404297, "time": 1.1135269403457642, "epoch": 169, "step": 105768}
{"lr": 0.0006411821228149521, "data_time": 0.07778284549713135, "grad_norm": 0.519443342089653, "loss": 5.90288143157959, "time": 1.172747802734375, "epoch": 170, "step": 105894}
{"lr": 0.0006399081882768199, "data_time": 0.07297167778015137, "grad_norm": 1.0497499048709868, "loss": 5.828254079818725, "time": 1.1213742733001708, "epoch": 170, "step": 105994}
{"lr": 0.0006386345984054475, "data_time": 0.078094744682312, "grad_norm": 0.37804767340421674, "loss": 5.862012052536011, "time": 1.128357481956482, "epoch": 170, "step": 106094}
{"lr": 0.0006373613570148697, "data_time": 0.07271640300750733, "grad_norm": 0.48465749621391296, "loss": 5.88193678855896, "time": 1.129082703590393, "epoch": 170, "step": 106194}
{"lr": 0.0006360884679180755, "data_time": 0.09367756843566895, "grad_norm": 0.42111496031284334, "loss": 5.818399047851562, "time": 1.149453353881836, "epoch": 170, "step": 106294}
{"lr": 0.0006348159349270012, "data_time": 0.0726020336151123, "grad_norm": 0.31586736142635347, "loss": 5.787298536300659, "time": 1.1446446895599365, "epoch": 170, "step": 106394}
{"lr": 0.0006332130562776244, "data_time": 0.11826810836791993, "grad_norm": 0.3105630248785019, "loss": 5.861040639877319, "time": 1.1478863954544067, "epoch": 171, "step": 106520}
{"lr": 0.000631941342122288, "data_time": 0.07802104949951172, "grad_norm": 0.3612762436270714, "loss": 5.815962028503418, "time": 1.1663947105407715, "epoch": 171, "step": 106620}
{"lr": 0.0006306699964921159, "data_time": 0.10300495624542236, "grad_norm": 0.610898369550705, "loss": 5.850476360321045, "time": 1.1604379177093507, "epoch": 171, "step": 106720}
{"lr": 0.0006293990231944213, "data_time": 0.07931158542633057, "grad_norm": 0.5709618836641311, "loss": 5.919247150421143, "time": 1.1511693716049194, "epoch": 171, "step": 106820}
{"lr": 0.0006281284260354038, "data_time": 0.12213234901428223, "grad_norm": 0.4874806791543961, "loss": 5.880928087234497, "time": 1.1641274213790893, "epoch": 171, "step": 106920}
{"lr": 0.000626858208820134, "data_time": 0.09715392589569091, "grad_norm": 0.5990855306386947, "loss": 5.878335046768188, "time": 1.1406109094619752, "epoch": 171, "step": 107020}
{"lr": 0.0006252582819781184, "data_time": 0.0726309061050415, "grad_norm": 0.6190836906433106, "loss": 5.906735849380493, "time": 1.1535472631454469, "epoch": 172, "step": 107146}
{"lr": 0.0006239889374468717, "data_time": 0.08801307678222656, "grad_norm": 0.28828482031822206, "loss": 5.866574048995972, "time": 1.1416642189025878, "epoch": 172, "step": 107246}
{"lr": 0.0006227199852559465, "data_time": 0.0631178617477417, "grad_norm": 0.3227415531873703, "loss": 5.83726544380188, "time": 1.1329279899597169, "epoch": 172, "step": 107346}
{"lr": 0.0006214514292054882, "data_time": 0.09145147800445556, "grad_norm": 0.319605153799057, "loss": 5.882088613510132, "time": 1.140897274017334, "epoch": 172, "step": 107446}
{"lr": 0.0006201832730944566, "data_time": 0.07335700988769531, "grad_norm": 0.4402287513017654, "loss": 5.87560248374939, "time": 1.128044843673706, "epoch": 172, "step": 107546}
{"lr": 0.0006189155207206133, "data_time": 0.08976867198944091, "grad_norm": 0.5967879265546798, "loss": 5.836408567428589, "time": 1.1212703943252564, "epoch": 172, "step": 107646}
{"lr": 0.0006173187334444675, "data_time": 0.0892723798751831, "grad_norm": 0.5971623867750168, "loss": 5.866678905487061, "time": 1.1337262868881226, "epoch": 173, "step": 107772}
{"lr": 0.0006160519075005198, "data_time": 0.08502392768859864, "grad_norm": 0.39969599843025205, "loss": 5.865939617156982, "time": 1.1087284564971924, "epoch": 173, "step": 107872}
{"lr": 0.0006147854976660073, "data_time": 0.09236867427825927, "grad_norm": 0.5170482814311981, "loss": 5.875676536560059, "time": 1.1435232162475586, "epoch": 173, "step": 107972}
{"lr": 0.000613519507733462, "data_time": 0.07787253856658935, "grad_norm": 0.42615656852722167, "loss": 5.844228506088257, "time": 1.1217651605606078, "epoch": 173, "step": 108072}
{"lr": 0.0006122539414941592, "data_time": 0.09392757415771484, "grad_norm": 0.7733329206705093, "loss": 5.878574085235596, "time": 1.1381622791290282, "epoch": 173, "step": 108172}
{"lr": 0.0006109888027381041, "data_time": 0.08448524475097656, "grad_norm": 0.3788865268230438, "loss": 5.821966886520386, "time": 1.1343350410461426, "epoch": 173, "step": 108272}
{"lr": 0.0006093953424178943, "data_time": 0.0806347370147705, "grad_norm": 0.48401660323143003, "loss": 5.854061841964722, "time": 1.1725953102111817, "epoch": 174, "step": 108398}
{"lr": 0.0006081311837288866, "data_time": 0.06457283496856689, "grad_norm": 0.568607573211193, "loss": 5.901620960235595, "time": 1.1566054105758667, "epoch": 174, "step": 108498}
{"lr": 0.0006068674648695964, "data_time": 0.07786211967468262, "grad_norm": 0.3453993797302246, "loss": 5.850628232955932, "time": 1.1353230953216553, "epoch": 174, "step": 108598}
{"lr": 0.0006056041896244965, "data_time": 0.06856279373168946, "grad_norm": 0.6844847053289413, "loss": 5.8825170516967775, "time": 1.139102029800415, "epoch": 174, "step": 108698}
{"lr": 0.0006043413617767318, "data_time": 0.08438348770141602, "grad_norm": 0.3197690337896347, "loss": 5.857572364807129, "time": 1.134425902366638, "epoch": 174, "step": 108798}
{"lr": 0.0006030789851081078, "data_time": 0.09005753993988037, "grad_norm": 0.5261712372303009, "loss": 5.8428998470306395, "time": 1.147342824935913, "epoch": 174, "step": 108898}
{"lr": 0.0006014890387434675, "data_time": 0.1016085386276245, "grad_norm": 0.35328352749347686, "loss": 5.875074625015259, "time": 1.1670634508132935, "epoch": 175, "step": 109024}
{"lr": 0.0006002276956640274, "data_time": 0.08432037830352783, "grad_norm": 0.4936865419149399, "loss": 5.882696723937988, "time": 1.1339371919631958, "epoch": 175, "step": 109124}
{"lr": 0.0005989668160829694, "data_time": 0.10631539821624755, "grad_norm": 0.3908138260245323, "loss": 5.877822399139404, "time": 1.1765783071517943, "epoch": 175, "step": 109224}
{"lr": 0.0005977064037762651, "data_time": 0.07792441844940186, "grad_norm": 0.4058652579784393, "loss": 5.875142335891724, "time": 1.1607227563858031, "epoch": 175, "step": 109324}
{"lr": 0.0005964464625184851, "data_time": 0.09847960472106934, "grad_norm": 0.634030182659626, "loss": 5.8149150848388675, "time": 1.1580872774124145, "epoch": 175, "step": 109424}
{"lr": 0.0005951869960827906, "data_time": 0.09173157215118408, "grad_norm": 0.3713893800973892, "loss": 5.876027154922485, "time": 1.1339625358581542, "epoch": 175, "step": 109524}
{"lr": 0.0005936007502609755, "data_time": 0.07144122123718262, "grad_norm": 0.6174564301967621, "loss": 5.857844543457031, "time": 1.1510656595230102, "epoch": 176, "step": 109650}
{"lr": 0.000592342370815305, "data_time": 0.09332785606384278, "grad_norm": 0.3743207424879074, "loss": 5.863217449188232, "time": 1.1555144309997558, "epoch": 176, "step": 109750}
{"lr": 0.0005910844784822906, "data_time": 0.08156116008758545, "grad_norm": 0.47474541068077086, "loss": 5.841516733169556, "time": 1.1953946113586427, "epoch": 176, "step": 109850}
{"lr": 0.0005898270770289553, "data_time": 0.0890357494354248, "grad_norm": 0.2832383468747139, "loss": 5.888814783096313, "time": 1.1675886869430543, "epoch": 176, "step": 109950}
{"lr": 0.000588570170220854, "data_time": 0.0628519058227539, "grad_norm": 0.6794520944356919, "loss": 5.815428686141968, "time": 1.1510556936264038, "epoch": 176, "step": 110050}
{"lr": 0.00058731376182206, "data_time": 0.08623244762420654, "grad_norm": 0.8658915847539902, "loss": 5.856323528289795, "time": 1.1456584692001344, "epoch": 176, "step": 110150}
{"lr": 0.0005857314026960508, "data_time": 0.12206037044525146, "grad_norm": 0.32219154238700864, "loss": 5.858745288848877, "time": 1.161293387413025, "epoch": 177, "step": 110276}
{"lr": 0.0005844761345605582, "data_time": 0.08273820877075196, "grad_norm": 0.323838597536087, "loss": 5.878402948379517, "time": 1.126931381225586, "epoch": 177, "step": 110376}
{"lr": 0.0005832213770948277, "data_time": 0.10433008670806884, "grad_norm": 0.28777970373630524, "loss": 5.893116521835327, "time": 1.1586906909942627, "epoch": 177, "step": 110476}
{"lr": 0.0005819671340564972, "data_time": 0.10210471153259278, "grad_norm": 0.3469745934009552, "loss": 5.846255254745484, "time": 1.1364312648773194, "epoch": 177, "step": 110576}
{"lr": 0.0005807134092016601, "data_time": 0.11201410293579102, "grad_norm": 0.3690516486763954, "loss": 5.781593465805054, "time": 1.1418010950088502, "epoch": 177, "step": 110676}
{"lr": 0.0005794602062848631, "data_time": 0.08212687969207763, "grad_norm": 0.23386605829000473, "loss": 5.9021501541137695, "time": 1.1260303497314452, "epoch": 177, "step": 110776}
{"lr": 0.0005778819195515179, "data_time": 0.08559181690216064, "grad_norm": 0.6176223814487457, "loss": 5.881058311462402, "time": 1.155335259437561, "epoch": 178, "step": 110902}
{"lr": 0.0005766299100374839, "data_time": 0.06361830234527588, "grad_norm": 0.4558860003948212, "loss": 5.831100511550903, "time": 1.1091485261917113, "epoch": 178, "step": 111002}
{"lr": 0.0005753784346903897, "data_time": 0.07381243705749511, "grad_norm": 0.36478806287050247, "loss": 5.838795566558838, "time": 1.1244322538375855, "epoch": 178, "step": 111102}
{"lr": 0.0005741274972580448, "data_time": 0.07110729217529296, "grad_norm": 0.3403775244951248, "loss": 5.87858395576477, "time": 1.1155144691467285, "epoch": 178, "step": 111202}
{"lr": 0.0005728771014866437, "data_time": 0.0896188735961914, "grad_norm": 1.0184823036193849, "loss": 5.866007709503174, "time": 1.1237383604049682, "epoch": 178, "step": 111302}
{"lr": 0.0005716272511207615, "data_time": 0.08578007221221924, "grad_norm": 0.7017623543739319, "loss": 5.844645071029663, "time": 1.1491268873214722, "epoch": 178, "step": 111402}
{"lr": 0.000570053221999027, "data_time": 0.09851725101470947, "grad_norm": 0.501512710750103, "loss": 5.8382011413574215, "time": 1.1911350727081298, "epoch": 179, "step": 111528}
{"lr": 0.0005688046180353194, "data_time": 0.08101112842559814, "grad_norm": 0.3422634243965149, "loss": 5.8518319606781, "time": 1.1821027517318725, "epoch": 179, "step": 111628}
{"lr": 0.0005675565716730432, "data_time": 0.09086296558380128, "grad_norm": 0.4465008333325386, "loss": 5.891118478775025, "time": 1.155061936378479, "epoch": 179, "step": 111728}
{"lr": 0.0005663090866497372, "data_time": 0.09041264057159423, "grad_norm": 0.3443212449550629, "loss": 5.90455641746521, "time": 1.1303245067596435, "epoch": 179, "step": 111828}
{"lr": 0.000565062166701259, "data_time": 0.12193796634674073, "grad_norm": 0.3338688641786575, "loss": 5.824947261810303, "time": 1.1946617126464845, "epoch": 179, "step": 111928}
{"lr": 0.0005638158155617738, "data_time": 0.07586150169372559, "grad_norm": 0.33255511075258254, "loss": 5.8702277660369875, "time": 1.1529697179794312, "epoch": 179, "step": 112028}
{"lr": 0.0005622462287709455, "data_time": 0.06370928287506103, "grad_norm": 0.360939122736454, "loss": 5.822474241256714, "time": 1.149590826034546, "epoch": 180, "step": 112154}
{"lr": 0.0005610011768867753, "data_time": 0.11499769687652588, "grad_norm": 0.45904198884963987, "loss": 5.807773399353027, "time": 1.1335485696792602, "epoch": 180, "step": 112254}
{"lr": 0.0005597567059730905, "data_time": 0.06832323074340821, "grad_norm": 0.5325130999088288, "loss": 5.860711479187012, "time": 1.1400354623794555, "epoch": 180, "step": 112354}
{"lr": 0.0005585128197567232, "data_time": 0.09661152362823486, "grad_norm": 0.4761264681816101, "loss": 5.852570533752441, "time": 1.115534234046936, "epoch": 180, "step": 112454}
{"lr": 0.0005572695219627529, "data_time": 0.0780627965927124, "grad_norm": 0.3523791432380676, "loss": 5.860933017730713, "time": 1.136228823661804, "epoch": 180, "step": 112554}
{"lr": 0.0005560268163144986, "data_time": 0.10193920135498047, "grad_norm": 0.6901637822389602, "loss": 5.847500514984131, "time": 1.1155183553695678, "epoch": 180, "step": 112654}
{"lr": 0.0005544618560525461, "data_time": 0.07461793422698974, "grad_norm": 0.45014827847480776, "loss": 5.8438738822937015, "time": 1.1365841388702393, "epoch": 181, "step": 112780}
{"lr": 0.0005532205023602718, "data_time": 0.09651567935943603, "grad_norm": 0.9955015033483505, "loss": 5.831204843521118, "time": 1.1286787748336793, "epoch": 181, "step": 112880}
{"lr": 0.0005519797529393582, "data_time": 0.09147877693176269, "grad_norm": 0.6318849086761474, "loss": 5.8476625919342045, "time": 1.1487455129623414, "epoch": 181, "step": 112980}
{"lr": 0.0005507396115054918, "data_time": 0.08059177398681641, "grad_norm": 0.5133996665477752, "loss": 5.832869195938111, "time": 1.146650791168213, "epoch": 181, "step": 113080}
{"lr": 0.0005495000817725376, "data_time": 0.09670164585113525, "grad_norm": 0.3672635018825531, "loss": 5.814952325820923, "time": 1.1336605072021484, "epoch": 181, "step": 113180}
{"lr": 0.0005482611674525292, "data_time": 0.08211197853088378, "grad_norm": 1.239008466899395, "loss": 5.890477132797241, "time": 1.1043837547302247, "epoch": 181, "step": 113280}
{"lr": 0.0005467010173744783, "data_time": 0.09507532119750976, "grad_norm": 0.4404306888580322, "loss": 5.855711460113525, "time": 1.1593577861785889, "epoch": 182, "step": 113406}
{"lr": 0.0005454635075524607, "data_time": 0.07342936992645263, "grad_norm": 0.5342427879571915, "loss": 5.837563800811767, "time": 1.148940348625183, "epoch": 182, "step": 113506}
{"lr": 0.0005442266252317638, "data_time": 0.08218533992767334, "grad_norm": 0.32908626049757006, "loss": 5.87031397819519, "time": 1.1188303470611571, "epoch": 182, "step": 113606}
{"lr": 0.0005429903741164933, "data_time": 0.08547358512878418, "grad_norm": 0.42078950703144075, "loss": 5.8595668315887455, "time": 1.1496926069259643, "epoch": 182, "step": 113706}
{"lr": 0.0005417547579088656, "data_time": 0.0830453872680664, "grad_norm": 0.6346892654895783, "loss": 5.840695095062256, "time": 1.140052032470703, "epoch": 182, "step": 113806}
{"lr": 0.0005405197803091933, "data_time": 0.07730200290679931, "grad_norm": 0.7960800468921662, "loss": 5.850083541870117, "time": 1.1318960189819336, "epoch": 182, "step": 113906}
{"lr": 0.0005389646235055729, "data_time": 0.11327478885650635, "grad_norm": 0.38047494143247607, "loss": 5.8601624965667725, "time": 1.1630496740341187, "epoch": 183, "step": 114032}
{"lr": 0.0005377311027810757, "data_time": 0.0844506025314331, "grad_norm": 0.35331162959337237, "loss": 5.891927671432495, "time": 1.1255380868911744, "epoch": 183, "step": 114132}
{"lr": 0.0005364982327142198, "data_time": 0.08488118648529053, "grad_norm": 0.42190803587436676, "loss": 5.842668581008911, "time": 1.1597575187683105, "epoch": 183, "step": 114232}
{"lr": 0.000535266016997096, "data_time": 0.0917515754699707, "grad_norm": 0.7349583774805069, "loss": 5.852200603485107, "time": 1.1359161138534546, "epoch": 183, "step": 114332}
{"lr": 0.000534034459319834, "data_time": 0.10959620475769043, "grad_norm": 0.4063972383737564, "loss": 5.842382097244263, "time": 1.1833488702774049, "epoch": 183, "step": 114432}
{"lr": 0.0005328035633705945, "data_time": 0.08516836166381836, "grad_norm": 0.9208964824676513, "loss": 5.841634798049927, "time": 1.1272028923034667, "epoch": 183, "step": 114532}
{"lr": 0.0005312535823459503, "data_time": 0.07824118137359619, "grad_norm": 0.6253752157092094, "loss": 5.863636445999146, "time": 1.1442023277282716, "epoch": 184, "step": 114658}
{"lr": 0.0005300241954780991, "data_time": 0.07785341739654542, "grad_norm": 0.39784484803676606, "loss": 5.837616300582885, "time": 1.1369251489639283, "epoch": 184, "step": 114758}
{"lr": 0.0005287954823478527, "data_time": 0.07282919883728027, "grad_norm": 0.3108549311757088, "loss": 5.826462268829346, "time": 1.1584266901016236, "epoch": 184, "step": 114858}
{"lr": 0.0005275674466348538, "data_time": 0.09743688106536866, "grad_norm": 0.6838562905788421, "loss": 5.821998834609985, "time": 1.1120739936828614, "epoch": 184, "step": 114958}
{"lr": 0.0005263400920167131, "data_time": 0.07321202754974365, "grad_norm": 0.5751105487346649, "loss": 5.854239130020142, "time": 1.1613162994384765, "epoch": 184, "step": 115058}
{"lr": 0.0005251134221690028, "data_time": 0.08209562301635742, "grad_norm": 0.6828768640756607, "loss": 5.841863298416138, "time": 1.10652334690094, "epoch": 184, "step": 115158}
{"lr": 0.000523568798820481, "data_time": 0.08523156642913818, "grad_norm": 0.4321915149688721, "loss": 5.86065707206726, "time": 1.1427444696426392, "epoch": 185, "step": 115284}
{"lr": 0.0005223436900832764, "data_time": 0.08483166694641113, "grad_norm": 0.2724657252430916, "loss": 5.818017864227295, "time": 1.1194697141647338, "epoch": 185, "step": 115384}
{"lr": 0.0005211192780845734, "data_time": 0.08710956573486328, "grad_norm": 0.5406217247247695, "loss": 5.834639930725098, "time": 1.105434489250183, "epoch": 185, "step": 115484}
{"lr": 0.0005198955664911336, "data_time": 0.08263556957244873, "grad_norm": 0.680504533648491, "loss": 5.860195922851562, "time": 1.1027838230133056, "epoch": 185, "step": 115584}
{"lr": 0.0005186725589676197, "data_time": 0.08409976959228516, "grad_norm": 0.7344372689723968, "loss": 5.841193056106567, "time": 1.127027416229248, "epoch": 185, "step": 115684}
{"lr": 0.0005174502591765847, "data_time": 0.08010985851287841, "grad_norm": 0.4691518500447273, "loss": 5.885836029052735, "time": 1.1193053007125855, "epoch": 185, "step": 115784}
{"lr": 0.0005159111747725835, "data_time": 0.08452339172363281, "grad_norm": 0.3575795590877533, "loss": 5.872170448303223, "time": 1.211909294128418, "epoch": 186, "step": 115910}
{"lr": 0.0005146904879379681, "data_time": 0.0795759916305542, "grad_norm": 0.5887443363666535, "loss": 5.8376117706298825, "time": 1.1192171812057494, "epoch": 186, "step": 116010}
{"lr": 0.0005134705207609857, "data_time": 0.07695283889770507, "grad_norm": 0.3894213825464249, "loss": 5.795796871185303, "time": 1.1658950567245483, "epoch": 186, "step": 116110}
{"lr": 0.0005122512768950866, "data_time": 0.0858182430267334, "grad_norm": 0.31887614279985427, "loss": 5.867947673797607, "time": 1.135944414138794, "epoch": 186, "step": 116210}
{"lr": 0.0005110327599915548, "data_time": 0.09874022006988525, "grad_norm": 0.4038084506988525, "loss": 5.855932807922363, "time": 1.1233581066131593, "epoch": 186, "step": 116310}
{"lr": 0.000509814973699495, "data_time": 0.07346916198730469, "grad_norm": 0.2880212277173996, "loss": 5.907099676132202, "time": 1.117252492904663, "epoch": 186, "step": 116410}
{"lr": 0.0005082816088583918, "data_time": 0.10543255805969239, "grad_norm": 0.6567251354455947, "loss": 5.841455316543579, "time": 1.181060266494751, "epoch": 187, "step": 116536}
{"lr": 0.0005070654871793763, "data_time": 0.09571845531463623, "grad_norm": 0.3540728598833084, "loss": 5.8478326320648195, "time": 1.1297202825546264, "epoch": 187, "step": 116636}
{"lr": 0.0005058501079926712, "data_time": 0.1049156665802002, "grad_norm": 0.9656085342168808, "loss": 5.828788614273071, "time": 1.1413957595825195, "epoch": 187, "step": 116736}
{"lr": 0.0005046354749379865, "data_time": 0.09931538105010987, "grad_norm": 0.5675554871559143, "loss": 5.850773906707763, "time": 1.1340513944625854, "epoch": 187, "step": 116836}
{"lr": 0.0005034215916527978, "data_time": 0.10391423702239991, "grad_norm": 0.37169051021337507, "loss": 5.842585897445678, "time": 1.15208842754364, "epoch": 187, "step": 116936}
{"lr": 0.0005022084617723342, "data_time": 0.1004469633102417, "grad_norm": 0.3789556622505188, "loss": 5.841266632080078, "time": 1.1392695665359498, "epoch": 187, "step": 117036}
{"lr": 0.0005006809964412934, "data_time": 0.10234584808349609, "grad_norm": 0.4922573819756508, "loss": 5.838725948333741, "time": 1.1707304000854493, "epoch": 188, "step": 117162}
{"lr": 0.0004994695826351505, "data_time": 0.0917043924331665, "grad_norm": 0.4702362760901451, "loss": 5.845366668701172, "time": 1.1080172538757325, "epoch": 188, "step": 117262}
{"lr": 0.0004982589340688602, "data_time": 0.08239943981170654, "grad_norm": 0.4199694722890854, "loss": 5.890081977844238, "time": 1.1420504570007324, "epoch": 188, "step": 117362}
{"lr": 0.0004970490543679622, "data_time": 0.08279521465301513, "grad_norm": 0.45659683644771576, "loss": 5.815092706680298, "time": 1.132524347305298, "epoch": 188, "step": 117462}
{"lr": 0.0004958399471556997, "data_time": 0.07492284774780274, "grad_norm": 0.6618358194828033, "loss": 5.821709537506104, "time": 1.124010443687439, "epoch": 188, "step": 117562}
{"lr": 0.0004946316160529993, "data_time": 0.08335866928100585, "grad_norm": 1.2355714946985246, "loss": 5.842529201507569, "time": 1.1244735240936279, "epoch": 188, "step": 117662}
{"lr": 0.000493110229486847, "data_time": 0.08152916431427001, "grad_norm": 0.43069205433130264, "loss": 5.818144083023071, "time": 1.169891929626465, "epoch": 189, "step": 117788}
{"lr": 0.0004919036657183552, "data_time": 0.09617609977722168, "grad_norm": 0.4047471135854721, "loss": 5.863590955734253, "time": 1.1445535182952882, "epoch": 189, "step": 117888}
{"lr": 0.0004906978898474521, "data_time": 0.0883781909942627, "grad_norm": 1.204849624633789, "loss": 5.82728910446167, "time": 1.1575557470321656, "epoch": 189, "step": 117988}
{"lr": 0.0004894929054850897, "data_time": 0.0842137098312378, "grad_norm": 0.6476440995931625, "loss": 5.818372297286987, "time": 1.1274755001068115, "epoch": 189, "step": 118088}
{"lr": 0.0004882887162398482, "data_time": 0.08426249027252197, "grad_norm": 0.34474226236343386, "loss": 5.87538595199585, "time": 1.1219626188278198, "epoch": 189, "step": 118188}
{"lr": 0.0004870853257179291, "data_time": 0.09524941444396973, "grad_norm": 0.41289720237255095, "loss": 5.850769996643066, "time": 1.1143067836761475, "epoch": 189, "step": 118288}
{"lr": 0.0004855701964581298, "data_time": 0.09451191425323487, "grad_norm": 0.2640489757061005, "loss": 5.8412007808685305, "time": 1.1683730363845826, "epoch": 190, "step": 118414}
{"lr": 0.00048436862432290104, "data_time": 0.07677757740020752, "grad_norm": 0.3559255123138428, "loss": 5.837112903594971, "time": 1.1462703227996827, "epoch": 190, "step": 118514}
{"lr": 0.0004831678626505366, "data_time": 0.10866351127624511, "grad_norm": 0.9037373512983322, "loss": 5.868391370773315, "time": 1.14111647605896, "epoch": 190, "step": 118614}
{"lr": 0.0004819679150369705, "data_time": 0.08533961772918701, "grad_norm": 0.33605472296476363, "loss": 5.8061364650726315, "time": 1.1375224113464355, "epoch": 190, "step": 118714}
{"lr": 0.00048076878507569994, "data_time": 0.08949804306030273, "grad_norm": 0.9495092272758484, "loss": 5.835743188858032, "time": 1.137837290763855, "epoch": 190, "step": 118814}
{"lr": 0.0004795704763577726, "data_time": 0.09028818607330322, "grad_norm": 0.4399471715092659, "loss": 5.82120714187622, "time": 1.1532464742660522, "epoch": 190, "step": 118914}
{"lr": 0.00047806178221145956, "data_time": 0.09765892028808594, "grad_norm": 0.35890508592128756, "loss": 5.843024778366089, "time": 1.1757203102111817, "epoch": 191, "step": 119040}
{"lr": 0.0004768653427193093, "data_time": 0.10050816535949707, "grad_norm": 0.48416494727134707, "loss": 5.834229278564453, "time": 1.134543514251709, "epoch": 191, "step": 119140}
{"lr": 0.0004756697361601864, "data_time": 0.08676111698150635, "grad_norm": 0.29018054306507113, "loss": 5.798689603805542, "time": 1.1441323041915894, "epoch": 191, "step": 119240}
{"lr": 0.0004744749661145863, "data_time": 0.079229736328125, "grad_norm": 0.3620026409626007, "loss": 5.843367195129394, "time": 1.1281131505966187, "epoch": 191, "step": 119340}
{"lr": 0.00047328103616050177, "data_time": 0.0796250581741333, "grad_norm": 0.3781736433506012, "loss": 5.816145563125611, "time": 1.168668532371521, "epoch": 191, "step": 119440}
{"lr": 0.0004720879498734079, "data_time": 0.09032895565032958, "grad_norm": 0.587303426861763, "loss": 5.826679420471192, "time": 1.1141980171203614, "epoch": 191, "step": 119540}
{"lr": 0.0004705858678925273, "data_time": 0.0650632619857788, "grad_norm": 0.4626503840088844, "loss": 5.865500354766846, "time": 1.1701918125152588, "epoch": 192, "step": 119666}
{"lr": 0.00046939470145093503, "data_time": 0.10404279232025146, "grad_norm": 0.38052930533885954, "loss": 5.843356037139893, "time": 1.1582432031631469, "epoch": 192, "step": 119766}
{"lr": 0.00046820439031478413, "data_time": 0.06629183292388915, "grad_norm": 0.2960992559790611, "loss": 5.857059574127197, "time": 1.164945697784424, "epoch": 192, "step": 119866}
{"lr": 0.00046701493804871344, "data_time": 0.08501114845275878, "grad_norm": 0.4126649424433708, "loss": 5.843468809127808, "time": 1.1496209621429443, "epoch": 192, "step": 119966}
{"lr": 0.0004658263482147882, "data_time": 0.06916844844818115, "grad_norm": 0.6321693867444992, "loss": 5.825021171569825, "time": 1.152131485939026, "epoch": 192, "step": 120066}
{"lr": 0.00046463862437249237, "data_time": 0.08113784790039062, "grad_norm": 0.3205049902200699, "loss": 5.835408926010132, "time": 1.1046528339385986, "epoch": 192, "step": 120166}
{"lr": 0.00046314333083302624, "data_time": 0.09826672077178955, "grad_norm": 0.4155921310186386, "loss": 5.804748773574829, "time": 1.1250011205673218, "epoch": 193, "step": 120292}
{"lr": 0.0004619575772306553, "data_time": 0.09046616554260253, "grad_norm": 0.49507237076759336, "loss": 5.800561857223511, "time": 1.1222917556762695, "epoch": 193, "step": 120392}
{"lr": 0.000460772701205766, "data_time": 0.0969388246536255, "grad_norm": 0.42824932038784025, "loss": 5.850623941421508, "time": 1.1405163764953614, "epoch": 193, "step": 120492}
{"lr": 0.00045958870630672077, "data_time": 0.09545583724975586, "grad_norm": 0.41070824563503266, "loss": 5.8638382911682125, "time": 1.1380567312240601, "epoch": 193, "step": 120592}
{"lr": 0.00045840559607924233, "data_time": 0.10314273834228516, "grad_norm": 0.34807087779045104, "loss": 5.77863130569458, "time": 1.135060691833496, "epoch": 193, "step": 120692}
{"lr": 0.00045722337406640435, "data_time": 0.08847005367279052, "grad_norm": 0.37861691415309906, "loss": 5.864627885818481, "time": 1.1344290494918823, "epoch": 193, "step": 120792}
{"lr": 0.0004557350444476789, "data_time": 0.06958889961242676, "grad_norm": 0.3937924847006798, "loss": 5.854122066497803, "time": 1.1513432502746581, "epoch": 194, "step": 120918}
{"lr": 0.00045455484283797126, "data_time": 0.08611252307891845, "grad_norm": 0.5263557285070419, "loss": 5.859612989425659, "time": 1.1107974290847777, "epoch": 194, "step": 121018}
{"lr": 0.00045337554097479914, "data_time": 0.08172976970672607, "grad_norm": 0.5690052926540374, "loss": 5.850764036178589, "time": 1.1251104116439818, "epoch": 194, "step": 121118}
{"lr": 0.0004521971423898305, "data_time": 0.07474722862243652, "grad_norm": 0.7625202685594559, "loss": 5.790527200698852, "time": 1.1488943338394164, "epoch": 194, "step": 121218}
{"lr": 0.00045101965061202957, "data_time": 0.07926077842712402, "grad_norm": 0.5387503027915954, "loss": 5.818626403808594, "time": 1.1637078046798706, "epoch": 194, "step": 121318}
{"lr": 0.00044984306916764346, "data_time": 0.09247798919677734, "grad_norm": 0.5103333860635757, "loss": 5.828676414489746, "time": 1.1052474975585938, "epoch": 194, "step": 121418}
{"lr": 0.00044836187813173853, "data_time": 0.09268319606781006, "grad_norm": 0.37516108751296995, "loss": 5.8283275127410885, "time": 1.1944997787475586, "epoch": 195, "step": 121544}
{"lr": 0.00044718736701658616, "data_time": 0.07405786514282227, "grad_norm": 0.5011601895093918, "loss": 5.814549112319947, "time": 1.1766161441802978, "epoch": 195, "step": 121644}
{"lr": 0.00044601377771143295, "data_time": 0.09447824954986572, "grad_norm": 0.34359992146492, "loss": 5.867408275604248, "time": 1.1793496370315553, "epoch": 195, "step": 121744}
{"lr": 0.0004448411137308396, "data_time": 0.08108022212982177, "grad_norm": 0.2865199014544487, "loss": 5.858260631561279, "time": 1.1396662712097168, "epoch": 195, "step": 121844}
{"lr": 0.00044366937858659686, "data_time": 0.09690818786621094, "grad_norm": 0.3476496174931526, "loss": 5.82594747543335, "time": 1.1880145072937012, "epoch": 195, "step": 121944}
{"lr": 0.00044249857578771286, "data_time": 0.09271190166473389, "grad_norm": 0.6972350716590882, "loss": 5.809050273895264, "time": 1.1471171379089355, "epoch": 195, "step": 122044}
{"lr": 0.0004410246971589632, "data_time": 0.06457667350769043, "grad_norm": 0.5751657366752625, "loss": 5.8740602970123295, "time": 1.142466402053833, "epoch": 196, "step": 122170}
{"lr": 0.00043985601437245323, "data_time": 0.09279837608337402, "grad_norm": 0.3510866329073906, "loss": 5.8028052806854244, "time": 1.1229114055633544, "epoch": 196, "step": 122270}
{"lr": 0.00043868827535122697, "data_time": 0.07324779033660889, "grad_norm": 0.48245612978935243, "loss": 5.82285475730896, "time": 1.1495562314987182, "epoch": 196, "step": 122370}
{"lr": 0.00043752148359232673, "data_time": 0.09421136379241943, "grad_norm": 0.4566105917096138, "loss": 5.802973365783691, "time": 1.119209623336792, "epoch": 196, "step": 122470}
{"lr": 0.0004363556425899563, "data_time": 0.08928110599517822, "grad_norm": 0.3626717358827591, "loss": 5.828020143508911, "time": 1.1476603031158448, "epoch": 196, "step": 122570}
{"lr": 0.0004351907558354733, "data_time": 0.08584115505218506, "grad_norm": 0.36653596460819243, "loss": 5.869286775588989, "time": 1.1151878833770752, "epoch": 196, "step": 122670}
{"lr": 0.00043372436258007113, "data_time": 0.10119245052337647, "grad_norm": 0.5000987708568573, "loss": 5.771759176254273, "time": 1.189177680015564, "epoch": 197, "step": 122796}
{"lr": 0.0004325616452723102, "data_time": 0.084175443649292, "grad_norm": 1.0293599188327789, "loss": 5.867826128005982, "time": 1.1260144472122193, "epoch": 197, "step": 122896}
{"lr": 0.00043139989357436365, "data_time": 0.09431872367858887, "grad_norm": 0.6716716885566711, "loss": 5.785914707183838, "time": 1.133377695083618, "epoch": 197, "step": 122996}
{"lr": 0.00043023911096534223, "data_time": 0.08451390266418457, "grad_norm": 0.3424486994743347, "loss": 5.834241437911987, "time": 1.1537242174148559, "epoch": 197, "step": 123096}
{"lr": 0.0004290793009214555, "data_time": 0.09814014434814453, "grad_norm": 0.438232946395874, "loss": 5.848936271667481, "time": 1.1591901540756226, "epoch": 197, "step": 123196}
{"lr": 0.00042792046691599896, "data_time": 0.09459497928619384, "grad_norm": 0.3473838835954666, "loss": 5.845520925521851, "time": 1.144092082977295, "epoch": 197, "step": 123296}
{"lr": 0.00042646173112169427, "data_time": 0.0700235366821289, "grad_norm": 0.4004535585641861, "loss": 5.890472745895385, "time": 1.1235769748687745, "epoch": 198, "step": 123422}
{"lr": 0.0004253051157427135, "data_time": 0.07902781963348389, "grad_norm": 0.33738564401865007, "loss": 5.841191291809082, "time": 1.1263467073440552, "epoch": 198, "step": 123522}
{"lr": 0.00042414948770475865, "data_time": 0.07392692565917969, "grad_norm": 0.47874170243740083, "loss": 5.789399147033691, "time": 1.1317730665206909, "epoch": 198, "step": 123622}
{"lr": 0.00042299485046860244, "data_time": 0.0757929801940918, "grad_norm": 0.34988497048616407, "loss": 5.844764232635498, "time": 1.140676164627075, "epoch": 198, "step": 123722}
{"lr": 0.00042184120749204975, "data_time": 0.07580978870391845, "grad_norm": 0.3663668304681778, "loss": 5.816971302032471, "time": 1.1461077690124513, "epoch": 198, "step": 123822}
{"lr": 0.0004206885622299288, "data_time": 0.06417562961578369, "grad_norm": 0.6963218718767166, "loss": 5.808780717849731, "time": 1.145102334022522, "epoch": 198, "step": 123922}
{"lr": 0.00041923765508583533, "data_time": 0.10405359268188477, "grad_norm": 0.7392590284347534, "loss": 5.84466552734375, "time": 1.1583650588989258, "epoch": 199, "step": 124048}
{"lr": 0.00041808727736957694, "data_time": 0.09476249217987061, "grad_norm": 0.7044235736131668, "loss": 5.81554102897644, "time": 1.158628273010254, "epoch": 199, "step": 124148}
{"lr": 0.00041693790860968807, "data_time": 0.09062759876251221, "grad_norm": 0.3861686199903488, "loss": 5.816455745697022, "time": 1.1414184093475341, "epoch": 199, "step": 124248}
{"lr": 0.00041578955224819575, "data_time": 0.080265212059021, "grad_norm": 0.3784444957971573, "loss": 5.809068298339843, "time": 1.1212925434112548, "epoch": 199, "step": 124348}
{"lr": 0.00041464221172409524, "data_time": 0.10566339492797852, "grad_norm": 0.37563730478286744, "loss": 5.821004104614258, "time": 1.14219229221344, "epoch": 199, "step": 124448}
{"lr": 0.0004134958904733413, "data_time": 0.08143248558044433, "grad_norm": 0.5554909229278564, "loss": 5.845484209060669, "time": 1.1095214128494262, "epoch": 199, "step": 124548}
{"lr": 0.00041205298224984493, "data_time": 0.072674560546875, "grad_norm": 0.5779559969902038, "loss": 5.878154468536377, "time": 1.159222459793091, "epoch": 200, "step": 124674}
{"lr": 0.00041090897719823577, "data_time": 0.08329403400421143, "grad_norm": 0.37322678565979006, "loss": 5.825311374664307, "time": 1.1290606021881104, "epoch": 200, "step": 124774}
{"lr": 0.00040976600259993247, "data_time": 0.06662540435791016, "grad_norm": 0.4041883111000061, "loss": 5.849827575683594, "time": 1.14744713306427, "epoch": 200, "step": 124874}
{"lr": 0.00040862406187781495, "data_time": 0.08140881061553955, "grad_norm": 1.0158457189798356, "loss": 5.824125814437866, "time": 1.1155531406402588, "epoch": 200, "step": 124974}
{"lr": 0.00040748315845166513, "data_time": 0.0729903221130371, "grad_norm": 0.4443429455161095, "loss": 5.775118446350097, "time": 1.1360925912857056, "epoch": 200, "step": 125074}
{"lr": 0.00040634329573815985, "data_time": 0.08763723373413086, "grad_norm": 0.39675251990556715, "loss": 5.791192150115966, "time": 1.13468279838562, "epoch": 200, "step": 125174}
{"lr": 0.0004049085557669383, "data_time": 0.08989934921264649, "grad_norm": 0.3418740093708038, "loss": 5.889250373840332, "time": 1.1421183347702026, "epoch": 201, "step": 125300}
{"lr": 0.0004037710576340424, "data_time": 0.09228324890136719, "grad_norm": 0.9525330990552903, "loss": 5.817905521392822, "time": 1.0996381521224976, "epoch": 201, "step": 125400}
{"lr": 0.00040263461133046175, "data_time": 0.09035916328430176, "grad_norm": 0.43538518249988556, "loss": 5.861793422698975, "time": 1.1121850252151488, "epoch": 201, "step": 125500}
{"lr": 0.0004014992202595247, "data_time": 0.08897714614868164, "grad_norm": 0.6320938467979431, "loss": 5.877745056152344, "time": 1.1204468965530396, "epoch": 201, "step": 125600}
{"lr": 0.0004003648878214007, "data_time": 0.09070878028869629, "grad_norm": 0.3241824507713318, "loss": 5.8310959815979, "time": 1.150154948234558, "epoch": 201, "step": 125700}
{"lr": 0.00039923161741308734, "data_time": 0.08339231014251709, "grad_norm": 0.4810209274291992, "loss": 5.818523502349853, "time": 1.1259645700454712, "epoch": 201, "step": 125800}
{"lr": 0.00039780521406723745, "data_time": 0.07317349910736085, "grad_norm": 0.4085342824459076, "loss": 5.808364725112915, "time": 1.1776256799697875, "epoch": 202, "step": 125926}
{"lr": 0.00039667435634350496, "data_time": 0.07510004043579102, "grad_norm": 0.43449583649635315, "loss": 5.833912515640259, "time": 1.1304749965667724, "epoch": 202, "step": 126026}
{"lr": 0.00039554457170165944, "data_time": 0.09425780773162842, "grad_norm": 0.47715107202529905, "loss": 5.825851964950561, "time": 1.1799324750900269, "epoch": 202, "step": 126126}
{"lr": 0.0003944158635250803, "data_time": 0.08241279125213623, "grad_norm": 0.4305856317281723, "loss": 5.88740816116333, "time": 1.1206579685211182, "epoch": 202, "step": 126226}
{"lr": 0.00039328823519392287, "data_time": 0.0830690622329712, "grad_norm": 0.5023164868354797, "loss": 5.829777193069458, "time": 1.1363489151000976, "epoch": 202, "step": 126326}
{"lr": 0.0003921616900851086, "data_time": 0.08447005748748779, "grad_norm": 0.6281728506088257, "loss": 5.799366855621338, "time": 1.1126367807388307, "epoch": 202, "step": 126426}
{"lr": 0.0003907437907593875, "data_time": 0.10718121528625488, "grad_norm": 0.2950758129358292, "loss": 5.846985006332398, "time": 1.135333490371704, "epoch": 203, "step": 126552}
{"lr": 0.0003896197061559857, "data_time": 0.09116063117980958, "grad_norm": 0.4134052813053131, "loss": 5.832204818725586, "time": 1.11666579246521, "epoch": 203, "step": 126652}
{"lr": 0.0003884967157611123, "data_time": 0.10294899940490723, "grad_norm": 0.48547622561454773, "loss": 5.848833847045898, "time": 1.1333383560180663, "epoch": 203, "step": 126752}
{"lr": 0.0003873748229378, "data_time": 0.08011114597320557, "grad_norm": 0.38026393204927444, "loss": 5.7815680503845215, "time": 1.117108654975891, "epoch": 203, "step": 126852}
{"lr": 0.00038625403104579315, "data_time": 0.09710218906402587, "grad_norm": 0.3294539228081703, "loss": 5.84071192741394, "time": 1.1523387670516967, "epoch": 203, "step": 126952}
{"lr": 0.00038513434344154113, "data_time": 0.08229231834411621, "grad_norm": 0.4775559067726135, "loss": 5.8346474170684814, "time": 1.122135543823242, "epoch": 203, "step": 127052}
{"lr": 0.00038372511453272116, "data_time": 0.07688229084014893, "grad_norm": 0.27643914222717286, "loss": 5.827446413040161, "time": 1.1848448753356933, "epoch": 204, "step": 127178}
{"lr": 0.00038260793496596243, "data_time": 0.09814186096191406, "grad_norm": 0.49073740243911745, "loss": 5.842457962036133, "time": 1.158978271484375, "epoch": 204, "step": 127278}
{"lr": 0.0003814918706059632, "data_time": 0.07413961887359619, "grad_norm": 0.34855978935956955, "loss": 5.792661952972412, "time": 1.1836740493774414, "epoch": 204, "step": 127378}
{"lr": 0.00038037692479501393, "data_time": 0.08433089256286622, "grad_norm": 0.44257702231407164, "loss": 5.794777822494507, "time": 1.1436846494674682, "epoch": 204, "step": 127478}
{"lr": 0.0003792631008720567, "data_time": 0.0881882905960083, "grad_norm": 0.3720664232969284, "loss": 5.858532333374024, "time": 1.1463346242904664, "epoch": 204, "step": 127578}
{"lr": 0.00037815040217267136, "data_time": 0.11616666316986084, "grad_norm": 0.49333955645561217, "loss": 5.839489269256592, "time": 1.1686742544174193, "epoch": 204, "step": 127678}
{"lr": 0.0003767500090600135, "data_time": 0.06352126598358154, "grad_norm": 0.49719840586185454, "loss": 5.845705556869507, "time": 1.183523464202881, "epoch": 205, "step": 127804}
{"lr": 0.00037563986563587297, "data_time": 0.10192885398864746, "grad_norm": 0.5347693473100662, "loss": 5.8097755908966064, "time": 1.1475387334823608, "epoch": 205, "step": 127904}
{"lr": 0.00037453085828584884, "data_time": 0.0681605577468872, "grad_norm": 0.6371447533369065, "loss": 5.839630031585694, "time": 1.165355134010315, "epoch": 205, "step": 128004}
{"lr": 0.0003734229903310975, "data_time": 0.09250879287719727, "grad_norm": 0.43395684361457826, "loss": 5.833372163772583, "time": 1.1252598285675048, "epoch": 205, "step": 128104}
{"lr": 0.00037231626508936477, "data_time": 0.08678543567657471, "grad_norm": 0.3727925568819046, "loss": 5.861469554901123, "time": 1.1527764320373535, "epoch": 205, "step": 128204}
{"lr": 0.0003712106858749727, "data_time": 0.0952303409576416, "grad_norm": 0.39430456459522245, "loss": 5.822108125686645, "time": 1.119288945198059, "epoch": 205, "step": 128304}
{"lr": 0.0003698192929008179, "data_time": 0.0835911750793457, "grad_norm": 0.5985407143831253, "loss": 5.874240732192993, "time": 1.192416524887085, "epoch": 206, "step": 128430}
{"lr": 0.0003687163158995486, "data_time": 0.08049490451812744, "grad_norm": 0.5962876230478287, "loss": 5.855148220062256, "time": 1.1257871866226197, "epoch": 206, "step": 128530}
{"lr": 0.00036761449570642957, "data_time": 0.0742034912109375, "grad_norm": 0.37772982120513915, "loss": 5.807941627502442, "time": 1.1546057224273683, "epoch": 206, "step": 128630}
{"lr": 0.0003665138356210933, "data_time": 0.0858198881149292, "grad_norm": 0.48853477239608767, "loss": 5.841989755630493, "time": 1.128897500038147, "epoch": 206, "step": 128730}
{"lr": 0.00036541433893969965, "data_time": 0.07587220668792724, "grad_norm": 0.49782038629055025, "loss": 5.827784299850464, "time": 1.1311185359954834, "epoch": 206, "step": 128830}
{"lr": 0.000364316008954923, "data_time": 0.08650562763214112, "grad_norm": 0.3874822214245796, "loss": 5.834616470336914, "time": 1.1103909015655518, "epoch": 206, "step": 128930}
{"lr": 0.0003629337794054047, "data_time": 0.11337287425994873, "grad_norm": 0.3521536380052567, "loss": 5.8083714008331295, "time": 1.161410355567932, "epoch": 207, "step": 129056}
{"lr": 0.0003618380982662479, "data_time": 0.08099987506866455, "grad_norm": NaN, "loss": 5.819540786743164, "time": 1.1469537258148192, "epoch": 207, "step": 129156}
{"lr": 0.0003607435945335177, "data_time": 0.11872766017913819, "grad_norm": 0.5754655122756958, "loss": 5.78229513168335, "time": 1.185733437538147, "epoch": 207, "step": 129256}
{"lr": 0.0003596502714849373, "data_time": 0.08266136646270753, "grad_norm": 0.4290273278951645, "loss": 5.817071580886841, "time": 1.144490647315979, "epoch": 207, "step": 129356}
{"lr": 0.0003585581323946923, "data_time": 0.1084284782409668, "grad_norm": 0.7728776335716248, "loss": 5.828006982803345, "time": 1.1609721899032592, "epoch": 207, "step": 129456}
{"lr": 0.00035746718053342473, "data_time": 0.0889371395111084, "grad_norm": 0.3704113349318504, "loss": 5.830186414718628, "time": 1.1376866102218628, "epoch": 207, "step": 129556}
{"lr": 0.00035609427661931014, "data_time": 0.08082678318023681, "grad_norm": 1.3791946113109588, "loss": 5.8365891456604, "time": 1.169404721260071, "epoch": 208, "step": 129682}
{"lr": 0.0003550060199253062, "data_time": 0.08404006958007812, "grad_norm": 0.7247931331396102, "loss": 5.811668586730957, "time": 1.1396244525909425, "epoch": 208, "step": 129782}
{"lr": 0.00035391896109783147, "data_time": 0.06371304988861085, "grad_norm": 0.29859724938869475, "loss": 5.815511226654053, "time": 1.1688552379608155, "epoch": 208, "step": 129882}
{"lr": 0.00035283310339231347, "data_time": 0.08414721488952637, "grad_norm": 0.40630641877651213, "loss": 5.806740427017212, "time": 1.126373791694641, "epoch": 208, "step": 129982}
{"lr": 0.00035174845006058286, "data_time": 0.06746644973754883, "grad_norm": 0.5486225545406341, "loss": 5.813385629653931, "time": 1.1372069835662841, "epoch": 208, "step": 130082}
{"lr": 0.0003506650043508626, "data_time": 0.09427797794342041, "grad_norm": 0.30385656505823133, "loss": 5.823325872421265, "time": 1.1335305452346802, "epoch": 208, "step": 130182}
{"lr": 0.0003493015871885127, "data_time": 0.11436870098114013, "grad_norm": 0.42943995296955106, "loss": 5.873994064331055, "time": 1.1757094860076904, "epoch": 209, "step": 130308}
{"lr": 0.0003482208826514098, "data_time": 0.08382792472839355, "grad_norm": 0.44905997812747955, "loss": 5.844398641586304, "time": 1.105911946296692, "epoch": 209, "step": 130408}
{"lr": 0.0003471413963003634, "data_time": 0.0969583511352539, "grad_norm": 0.4198110714554787, "loss": 5.80714979171753, "time": 1.145637536048889, "epoch": 209, "step": 130508}
{"lr": 0.0003460631313681244, "data_time": 0.07792632579803467, "grad_norm": 0.5525525897741318, "loss": 5.859788799285889, "time": 1.1406264305114746, "epoch": 209, "step": 130608}
{"lr": 0.00034498609108378443, "data_time": 0.09379920959472657, "grad_norm": 0.47980256378650665, "loss": 5.789713191986084, "time": 1.1369050979614257, "epoch": 209, "step": 130708}
{"lr": 0.000343910278672769, "data_time": 0.08049752712249755, "grad_norm": 0.39216458797454834, "loss": 5.876212882995605, "time": 1.1300658702850341, "epoch": 209, "step": 130808}
{"lr": 0.00034255650826523344, "data_time": 0.09161789417266845, "grad_norm": 0.38326638638973237, "loss": 5.796266746520996, "time": 1.1659380197525024, "epoch": 210, "step": 130934}
{"lr": 0.0003414834827104999, "data_time": 0.08970255851745605, "grad_norm": 0.3280979931354523, "loss": 5.842937231063843, "time": 1.1604289531707763, "epoch": 210, "step": 131034}
{"lr": 0.000340411695518392, "data_time": 0.07298572063446045, "grad_norm": 0.32719871401786804, "loss": 5.83006477355957, "time": 1.1308820486068725, "epoch": 210, "step": 131134}
{"lr": 0.00033934114989860205, "data_time": 0.07568347454071045, "grad_norm": 0.3994252622127533, "loss": 5.7966546535491945, "time": 1.1521320343017578, "epoch": 210, "step": 131234}
{"lr": 0.00033827184905710654, "data_time": 0.07654292583465576, "grad_norm": 0.4259240746498108, "loss": 5.825330591201782, "time": 1.1144049882888794, "epoch": 210, "step": 131334}
{"lr": 0.0003372037961961524, "data_time": 0.06764531135559082, "grad_norm": 0.3715969830751419, "loss": 5.8069652080535885, "time": 1.0984671115875244, "epoch": 210, "step": 131434}
{"lr": 0.00033585983141438996, "data_time": 0.09945359230041503, "grad_norm": 0.3513563543558121, "loss": 5.817813873291016, "time": 1.14629065990448, "epoch": 211, "step": 131560}
{"lr": 0.0003347946107663324, "data_time": 0.0844038486480713, "grad_norm": 0.42304641008377075, "loss": 5.819278573989868, "time": 1.1071047067642212, "epoch": 211, "step": 131660}
{"lr": 0.0003337306485121415, "data_time": 0.1017982006072998, "grad_norm": 0.36360932886600494, "loss": 5.784662294387817, "time": 1.1498887300491334, "epoch": 211, "step": 131760}
{"lr": 0.0003326679478380763, "data_time": 0.0872920036315918, "grad_norm": 0.3552371308207512, "loss": 5.782162189483643, "time": 1.1185497999191285, "epoch": 211, "step": 131860}
{"lr": 0.00033160651192661955, "data_time": 0.09282028675079346, "grad_norm": 1.0214543879032134, "loss": 5.800967741012573, "time": 1.1668049335479735, "epoch": 211, "step": 131960}
{"lr": 0.00033054634395646496, "data_time": 0.09251244068145752, "grad_norm": 1.2306238055229186, "loss": 5.840344429016113, "time": 1.1268020391464233, "epoch": 211, "step": 132060}
{"lr": 0.0003292123425207029, "data_time": 0.06232938766479492, "grad_norm": 0.3794278740882874, "loss": 5.8034077167510985, "time": 1.1495325565338135, "epoch": 212, "step": 132186}
{"lr": 0.00032815505178768715, "data_time": 0.08270959854125977, "grad_norm": 0.3337337553501129, "loss": 5.822618675231934, "time": 1.141702651977539, "epoch": 212, "step": 132286}
{"lr": 0.0003270990393321006, "data_time": 0.07330875396728516, "grad_norm": 0.6148483335971833, "loss": 5.83585844039917, "time": 1.1460248947143554, "epoch": 212, "step": 132386}
{"lr": 0.00032604430831639607, "data_time": 0.09398908615112304, "grad_norm": 0.2744159817695618, "loss": 5.82869963645935, "time": 1.1722803831100463, "epoch": 212, "step": 132486}
{"lr": 0.00032499086189918824, "data_time": 0.08731553554534913, "grad_norm": 0.6522115617990494, "loss": 5.879381370544434, "time": 1.1608891487121582, "epoch": 212, "step": 132586}
{"lr": 0.00032393870323524593, "data_time": 0.08069589138031005, "grad_norm": 0.5831417560577392, "loss": 5.815738534927368, "time": 1.120077085494995, "epoch": 212, "step": 132686}
{"lr": 0.00032261482169646585, "data_time": 0.090034818649292, "grad_norm": 0.37398968636989594, "loss": 5.845333814620972, "time": 1.1774984836578368, "epoch": 213, "step": 132812}
{"lr": 0.0003215655849562478, "data_time": 0.08938117027282715, "grad_norm": 0.44823708534240725, "loss": 5.7711930751800535, "time": 1.1459316968917848, "epoch": 213, "step": 132912}
{"lr": 0.00032051764622701036, "data_time": 0.10638594627380371, "grad_norm": 0.5039619922637939, "loss": 5.8244377136230465, "time": 1.1567411184310914, "epoch": 213, "step": 133012}
{"lr": 0.00031947100864702775, "data_time": 0.08286333084106445, "grad_norm": 0.5529278725385666, "loss": 5.845650339126587, "time": 1.1238743305206298, "epoch": 213, "step": 133112}
{"lr": 0.0003184256753506775, "data_time": 0.1437744379043579, "grad_norm": 0.44427303075790403, "loss": 5.801747274398804, "time": 1.2019760608673096, "epoch": 213, "step": 133212}
{"lr": 0.0003173816494684309, "data_time": 0.09797787666320801, "grad_norm": 1.4276378110051156, "loss": 5.859511137008667, "time": 1.158813238143921, "epoch": 213, "step": 133312}
{"lr": 0.00031606804319000205, "data_time": 0.09602465629577636, "grad_norm": 0.55423024892807, "loss": 5.794899034500122, "time": 1.2088537454605102, "epoch": 214, "step": 133438}
{"lr": 0.00031502698357516515, "data_time": 0.07802584171295165, "grad_norm": 0.5928488433361053, "loss": 5.813194131851196, "time": 1.1264053344726563, "epoch": 214, "step": 133538}
{"lr": 0.0003139872415525326, "data_time": 0.08216519355773926, "grad_norm": 0.3705421507358551, "loss": 5.855261135101318, "time": 1.1361729383468628, "epoch": 214, "step": 133638}
{"lr": 0.00031294882023583256, "data_time": 0.09138240814208984, "grad_norm": 0.43864052593708036, "loss": 5.812609243392944, "time": 1.1061997652053832, "epoch": 214, "step": 133738}
{"lr": 0.0003119117227348364, "data_time": 0.07024087905883789, "grad_norm": 0.3768822729587555, "loss": 5.815310096740722, "time": 1.1402101278305055, "epoch": 214, "step": 133838}
{"lr": 0.00031087595215535235, "data_time": 0.07828328609466553, "grad_norm": 0.4107623606920242, "loss": 5.795873498916626, "time": 1.1012017726898193, "epoch": 214, "step": 133938}
{"lr": 0.0003095727752947957, "data_time": 0.10207366943359375, "grad_norm": 0.35366730093955995, "loss": 5.835953330993652, "time": 1.1521449565887452, "epoch": 215, "step": 134064}
{"lr": 0.00030854001497830077, "data_time": 0.09373693466186524, "grad_norm": 0.28636544942855835, "loss": 5.833176994323731, "time": 1.1257678508758544, "epoch": 215, "step": 134164}
{"lr": 0.00030750859168060974, "data_time": 0.09885408878326415, "grad_norm": 0.42187730967998505, "loss": 5.801164245605468, "time": 1.1474151372909547, "epoch": 215, "step": 134264}
{"lr": 0.00030647850849053695, "data_time": 0.08686537742614746, "grad_norm": 0.5963913351297379, "loss": 5.831814336776733, "time": 1.1428432941436768, "epoch": 215, "step": 134364}
{"lr": 0.0003054497684928853, "data_time": 0.10155119895935058, "grad_norm": 0.4641357511281967, "loss": 5.802445888519287, "time": 1.1668628692626952, "epoch": 215, "step": 134464}
{"lr": 0.00030442237476843324, "data_time": 0.09686813354492188, "grad_norm": 0.45978637635707853, "loss": 5.822929096221924, "time": 1.124445915222168, "epoch": 215, "step": 134564}
{"lr": 0.00030312978025933096, "data_time": 0.07535920143127442, "grad_norm": 0.398827052116394, "loss": 5.8006524562835695, "time": 1.159698748588562, "epoch": 216, "step": 134690}
{"lr": 0.00030210544044018123, "data_time": 0.09100518226623536, "grad_norm": 0.4023218333721161, "loss": 5.792924404144287, "time": 1.1332848787307739, "epoch": 216, "step": 134790}
{"lr": 0.00030108245690952827, "data_time": 0.07795629501342774, "grad_norm": 0.5910998910665513, "loss": 5.793186902999878, "time": 1.154535698890686, "epoch": 216, "step": 134890}
{"lr": 0.0003000608327309132, "data_time": 0.0929030179977417, "grad_norm": 0.8629683345556259, "loss": 5.782440423965454, "time": 1.1088308572769165, "epoch": 216, "step": 134990}
{"lr": 0.00029904057096380546, "data_time": 0.0796983003616333, "grad_norm": 0.44485132992267606, "loss": 5.747850131988526, "time": 1.1354900121688842, "epoch": 216, "step": 135090}
{"lr": 0.0002980216746635942, "data_time": 0.09280500411987305, "grad_norm": 0.3301954597234726, "loss": 5.852215385437011, "time": 1.1130648374557495, "epoch": 216, "step": 135190}
{"lr": 0.0002967398141976434, "data_time": 0.10738227367401124, "grad_norm": 0.2674620658159256, "loss": 5.8162449359893795, "time": 1.1594681978225707, "epoch": 217, "step": 135316}
{"lr": 0.0002957240150866574, "data_time": 0.1043593168258667, "grad_norm": 0.6076219648122787, "loss": 5.803328371047973, "time": 1.1550942420959474, "epoch": 217, "step": 135416}
{"lr": 0.0002947095913746959, "data_time": 0.1033411979675293, "grad_norm": 0.332926544547081, "loss": 5.858902215957642, "time": 1.1407412528991698, "epoch": 217, "step": 135516}
{"lr": 0.00029369654609966465, "data_time": 0.08260114192962646, "grad_norm": 0.3468136727809906, "loss": 5.807943630218506, "time": 1.107240867614746, "epoch": 217, "step": 135616}
{"lr": 0.00029268488229534294, "data_time": 0.09910602569580078, "grad_norm": 1.6497383117675781, "loss": 5.818518018722534, "time": 1.1137499809265137, "epoch": 217, "step": 135716}
{"lr": 0.0002916746029913713, "data_time": 0.10110757350921631, "grad_norm": 0.3063407763838768, "loss": 5.8214881896972654, "time": 1.1344970703125, "epoch": 217, "step": 135816}
{"lr": 0.0002904036270005809, "data_time": 0.10274450778961182, "grad_norm": 0.5339775741100311, "loss": 5.820132446289063, "time": 1.1442155838012695, "epoch": 218, "step": 135942}
{"lr": 0.00028939648780628856, "data_time": 0.0822793960571289, "grad_norm": 0.43786306083202364, "loss": 5.813488388061524, "time": 1.1404438495635987, "epoch": 218, "step": 136042}
{"lr": 0.00028839074296013875, "data_time": 0.0879148006439209, "grad_norm": 0.5433317601680756, "loss": 5.7722694873809814, "time": 1.1509143114089966, "epoch": 218, "step": 136142}
{"lr": 0.0002873863954740473, "data_time": 0.07782742977142335, "grad_norm": 0.46459448635578154, "loss": 5.7829368114471436, "time": 1.1458458662033082, "epoch": 218, "step": 136242}
{"lr": 0.000286383448355744, "data_time": 0.08557741641998291, "grad_norm": 0.49367022812366484, "loss": 5.826438760757446, "time": 1.1308542251586915, "epoch": 218, "step": 136342}
{"lr": 0.00028538190460876726, "data_time": 0.06708602905273438, "grad_norm": 0.4189988702535629, "loss": 5.81932053565979, "time": 1.1299654960632324, "epoch": 218, "step": 136442}
{"lr": 0.0002841219622478052, "data_time": 0.08903629779815674, "grad_norm": 0.3988147974014282, "loss": 5.829268074035644, "time": 1.1615935802459716, "epoch": 219, "step": 136568}
{"lr": 0.00028312360116245617, "data_time": 0.08762645721435547, "grad_norm": 0.6686510920524598, "loss": 5.814680910110473, "time": 1.1386348485946656, "epoch": 219, "step": 136668}
{"lr": 0.00028212665321073496, "data_time": 0.10368764400482178, "grad_norm": 0.4524615228176117, "loss": 5.8468116283416744, "time": 1.1776755094528197, "epoch": 219, "step": 136768}
{"lr": 0.00028113112137821304, "data_time": 0.08952529430389404, "grad_norm": 0.3885635077953339, "loss": 5.814153671264648, "time": 1.1383257627487182, "epoch": 219, "step": 136868}
{"lr": 0.00028013700864622004, "data_time": 0.11038167476654052, "grad_norm": 0.6618211656808853, "loss": 5.727328634262085, "time": 1.1600799322128297, "epoch": 219, "step": 136968}
{"lr": 0.0002791443179918375, "data_time": 0.09621174335479736, "grad_norm": 0.5936644375324249, "loss": 5.7767726421356205, "time": 1.1472826957702638, "epoch": 219, "step": 137068}
{"lr": 0.00027789555712052604, "data_time": 0.07299761772155762, "grad_norm": 0.3895461201667786, "loss": 5.805097818374634, "time": 1.15813250541687, "epoch": 220, "step": 137194}
{"lr": 0.000276906091306221, "data_time": 0.08079860210418702, "grad_norm": 0.39064545929431915, "loss": 5.791266250610351, "time": 1.1248209476470947, "epoch": 220, "step": 137294}
{"lr": 0.00027591805724519046, "data_time": 0.08296670913696289, "grad_norm": 0.6532296776771546, "loss": 5.769882726669311, "time": 1.1249594926834106, "epoch": 220, "step": 137394}
{"lr": 0.0002749314578963121, "data_time": 0.07658078670501708, "grad_norm": 0.39297776222229003, "loss": 5.858463907241822, "time": 1.1212374448776246, "epoch": 220, "step": 137494}
{"lr": 0.0002739462962141663, "data_time": 0.08604331016540527, "grad_norm": 0.35648894757032396, "loss": 5.80408411026001, "time": 1.1198574781417847, "epoch": 220, "step": 137594}
{"lr": 0.0002729625751490276, "data_time": 0.08012447357177735, "grad_norm": 1.170739471912384, "loss": 5.849447870254517, "time": 1.1194491863250733, "epoch": 220, "step": 137694}
{"lr": 0.00027172514231499204, "data_time": 0.10004847049713135, "grad_norm": 0.5672762304544449, "loss": 5.802788352966308, "time": 1.1447321414947509, "epoch": 221, "step": 137820}
{"lr": 0.00027074468788993014, "data_time": 0.08823153972625733, "grad_norm": 0.533667242527008, "loss": 5.808526420593262, "time": 1.1313076734542846, "epoch": 221, "step": 137920}
{"lr": 0.0002697656836697688, "data_time": 0.10406033992767334, "grad_norm": 0.4585422337055206, "loss": 5.82483868598938, "time": 1.1536786317825318, "epoch": 221, "step": 138020}
{"lr": 0.000268788132586343, "data_time": 0.08018705844879151, "grad_norm": 0.3996250659227371, "loss": 5.829040145874023, "time": 1.1352843999862672, "epoch": 221, "step": 138120}
{"lr": 0.0002678120375671354, "data_time": 0.09780361652374267, "grad_norm": 1.0848483830690383, "loss": 5.836079406738281, "time": 1.1706261873245238, "epoch": 221, "step": 138220}
{"lr": 0.00026683740153526946, "data_time": 0.09532954692840576, "grad_norm": 0.30604676306247713, "loss": 5.848490238189697, "time": 1.139822292327881, "epoch": 221, "step": 138320}
{"lr": 0.00026561144195673896, "data_time": 0.0760199785232544, "grad_norm": 0.3255427464842796, "loss": 5.776264333724976, "time": 1.1740403652191163, "epoch": 222, "step": 138446}
{"lr": 0.0002646401139815951, "data_time": 0.07114691734313965, "grad_norm": 0.3661819726228714, "loss": 5.81643123626709, "time": 1.131419587135315, "epoch": 222, "step": 138546}
{"lr": 0.000263670254492789, "data_time": 0.09798645973205566, "grad_norm": 0.4920639991760254, "loss": 5.816004991531372, "time": 1.1308319330215455, "epoch": 222, "step": 138646}
{"lr": 0.000262701866394769, "data_time": 0.07188358306884765, "grad_norm": 0.33440726101398466, "loss": 5.801692819595337, "time": 1.1365905284881592, "epoch": 222, "step": 138746}
{"lr": 0.0002617349525875783, "data_time": 0.08149480819702148, "grad_norm": 1.261145108938217, "loss": 5.8198404788970945, "time": 1.1353728771209717, "epoch": 222, "step": 138846}
{"lr": 0.00026076951596684455, "data_time": 0.08334822654724121, "grad_norm": 0.3911236494779587, "loss": 5.825663900375366, "time": 1.1312325716018676, "epoch": 222, "step": 138946}
{"lr": 0.0002595551735156114, "data_time": 0.10023798942565917, "grad_norm": 0.3723272055387497, "loss": 5.800387763977051, "time": 1.17132568359375, "epoch": 223, "step": 139072}
{"lr": 0.0002585930859800295, "data_time": 0.08547985553741455, "grad_norm": 0.38334476947784424, "loss": 5.743080949783325, "time": 1.1359610319137574, "epoch": 223, "step": 139172}
{"lr": 0.0002576324850398886, "data_time": 0.09364886283874511, "grad_norm": 0.31809611320495607, "loss": 5.817375802993775, "time": 1.1588512182235717, "epoch": 223, "step": 139272}
{"lr": 0.00025667337357191165, "data_time": 0.0950402021408081, "grad_norm": 0.47308976054191587, "loss": 5.798387670516968, "time": 1.1667490482330323, "epoch": 223, "step": 139372}
{"lr": 0.00025571575444835985, "data_time": 0.10477294921875, "grad_norm": 0.4743340998888016, "loss": 5.812469100952148, "time": 1.1527668952941894, "epoch": 223, "step": 139472}
{"lr": 0.00025475963053702664, "data_time": 0.09530839920043946, "grad_norm": 0.9359201639890671, "loss": 5.758427143096924, "time": 1.1355227470397948, "epoch": 223, "step": 139572}
{"lr": 0.00025355704772156255, "data_time": 0.0638587474822998, "grad_norm": 0.33950754404067995, "loss": 5.800224542617798, "time": 1.1699861288070679, "epoch": 224, "step": 139698}
{"lr": 0.0002526043135307825, "data_time": 0.07701144218444825, "grad_norm": 0.3901512444019318, "loss": 5.785845804214477, "time": 1.1403908014297486, "epoch": 224, "step": 139798}
{"lr": 0.0002516530838700883, "data_time": 0.07092697620391845, "grad_norm": 1.0099602848291398, "loss": 5.81041522026062, "time": 1.156825876235962, "epoch": 224, "step": 139898}
{"lr": 0.00025070336158813814, "data_time": 0.08772380352020263, "grad_norm": 0.4565738320350647, "loss": 5.787730121612549, "time": 1.169636845588684, "epoch": 224, "step": 139998}
{"lr": 0.00024975514952907667, "data_time": 0.07042956352233887, "grad_norm": 0.43922463059425354, "loss": 5.794416189193726, "time": 1.16071298122406, "epoch": 224, "step": 140098}
{"lr": 0.0002488084505325252, "data_time": 0.09922654628753662, "grad_norm": 0.42340507805347444, "loss": 5.796671962738037, "time": 1.1258607864379884, "epoch": 224, "step": 140198}
{"lr": 0.00024761776848125557, "data_time": 0.10464973449707031, "grad_norm": 0.4498280227184296, "loss": 5.795484733581543, "time": 1.1732100248336792, "epoch": 225, "step": 140324}
{"lr": 0.00024667449944285725, "data_time": 0.08993635177612305, "grad_norm": 0.3226303026080132, "loss": 5.840158367156983, "time": 1.153219747543335, "epoch": 225, "step": 140424}
{"lr": 0.0002457327526926264, "data_time": 0.1070298433303833, "grad_norm": 0.5552108228206635, "loss": 5.809341096878052, "time": 1.1355212688446046, "epoch": 225, "step": 140524}
{"lr": 0.0002447925310508233, "data_time": 0.07921967506408692, "grad_norm": 0.5115152299404144, "loss": 5.8107208728790285, "time": 1.1188408851623535, "epoch": 225, "step": 140624}
{"lr": 0.0002438538373331393, "data_time": 0.09681985378265381, "grad_norm": 1.5213526636362076, "loss": 5.815022134780884, "time": 1.1266449213027954, "epoch": 225, "step": 140724}
{"lr": 0.0002429166743506914, "data_time": 0.09782249927520752, "grad_norm": 0.9599039614200592, "loss": 5.809674882888794, "time": 1.1397344827651978, "epoch": 225, "step": 140824}
{"lr": 0.00024173803279543146, "data_time": 0.07255690097808838, "grad_norm": 0.3468910872936249, "loss": 5.806313467025757, "time": 1.154296636581421, "epoch": 226, "step": 140950}
{"lr": 0.00024080433960621726, "data_time": 0.09680852890014649, "grad_norm": 0.3533232301473618, "loss": 5.816194200515747, "time": 1.1409835338592529, "epoch": 226, "step": 141050}
{"lr": 0.00023987218628460552, "data_time": 0.06625747680664062, "grad_norm": 0.3482687771320343, "loss": 5.849392890930176, "time": 1.158472466468811, "epoch": 226, "step": 141150}
{"lr": 0.00023894157562212598, "data_time": 0.0835862398147583, "grad_norm": 0.42843723595142363, "loss": 5.769619083404541, "time": 1.142272448539734, "epoch": 226, "step": 141250}
{"lr": 0.00023801251040568954, "data_time": 0.07075145244598388, "grad_norm": 0.4076755464076996, "loss": 5.800172424316406, "time": 1.1341986894607543, "epoch": 226, "step": 141350}
{"lr": 0.0002370849934175777, "data_time": 0.07986502647399903, "grad_norm": 0.5006009250879287, "loss": 5.79516339302063, "time": 1.1349498748779296, "epoch": 226, "step": 141450}
{"lr": 0.0002359185306771414, "data_time": 0.0873556137084961, "grad_norm": 0.34967695623636247, "loss": 5.847423410415649, "time": 1.1863970756530762, "epoch": 227, "step": 141576}
{"lr": 0.0002349945229101484, "data_time": 0.10057320594787597, "grad_norm": 0.40432921051979065, "loss": 5.8227455615997314, "time": 1.1617863655090332, "epoch": 227, "step": 141676}
{"lr": 0.00023407207240948188, "data_time": 0.10479466915130616, "grad_norm": 0.5911456644535065, "loss": 5.815785932540893, "time": 1.196147608757019, "epoch": 227, "step": 141776}
{"lr": 0.0002331511819376146, "data_time": 0.08761241436004638, "grad_norm": 0.3514255493879318, "loss": 5.805701398849488, "time": 1.157913327217102, "epoch": 227, "step": 141876}
{"lr": 0.00023223185425234828, "data_time": 0.10200648307800293, "grad_norm": 0.36768121272325516, "loss": 5.813676500320435, "time": 1.1648844957351685, "epoch": 227, "step": 141976}
{"lr": 0.00023131409210680364, "data_time": 0.0841524839401245, "grad_norm": 0.7264789521694184, "loss": 5.792337799072266, "time": 1.1340005159378053, "epoch": 227, "step": 142076}
{"lr": 0.00023015994507076588, "data_time": 0.07648651599884033, "grad_norm": 0.35011285841464995, "loss": 5.806136655807495, "time": 1.19527587890625, "epoch": 228, "step": 142202}
{"lr": 0.0002292457311624023, "data_time": 0.08827428817749024, "grad_norm": 0.43871269524097445, "loss": 5.772765111923218, "time": 1.154149866104126, "epoch": 228, "step": 142302}
{"lr": 0.00022833309173633736, "data_time": 0.06775004863739013, "grad_norm": 0.3459246873855591, "loss": 5.829715204238892, "time": 1.1646499633789062, "epoch": 228, "step": 142402}
{"lr": 0.00022742202952566289, "data_time": 0.07788224220275879, "grad_norm": 0.4468257069587708, "loss": 5.769400119781494, "time": 1.151556134223938, "epoch": 228, "step": 142502}
{"lr": 0.0002265125472587469, "data_time": 0.06903631687164306, "grad_norm": 0.46278809010982513, "loss": 5.828677606582642, "time": 1.1431050062179566, "epoch": 228, "step": 142602}
{"lr": 0.00022560464765922672, "data_time": 0.08919227123260498, "grad_norm": 0.40467155277729033, "loss": 5.778173875808716, "time": 1.1422853946685791, "epoch": 228, "step": 142702}
{"lr": 0.00022446295177185882, "data_time": 0.10510785579681396, "grad_norm": 0.4888130247592926, "loss": 5.769316673278809, "time": 1.1819838523864745, "epoch": 229, "step": 142828}
{"lr": 0.00022355863900918018, "data_time": 0.09204623699188233, "grad_norm": 0.33161976337432864, "loss": 5.837370252609253, "time": 1.1248563289642335, "epoch": 229, "step": 142928}
{"lr": 0.00022265591775999955, "data_time": 0.08109357357025146, "grad_norm": 0.4222720980644226, "loss": 5.771837568283081, "time": 1.147058367729187, "epoch": 229, "step": 143028}
{"lr": 0.00022175479072770717, "data_time": 0.0847313404083252, "grad_norm": 0.3993717014789581, "loss": 5.8236864566802975, "time": 1.1413732290267944, "epoch": 229, "step": 143128}
{"lr": 0.0002208552606109181, "data_time": 0.08724734783172608, "grad_norm": 0.36767806112766266, "loss": 5.782551956176758, "time": 1.1487056970596314, "epoch": 229, "step": 143228}
{"lr": 0.0002199573301034653, "data_time": 0.0922703504562378, "grad_norm": 0.3563183218240738, "loss": 5.815572738647461, "time": 1.1271517753601075, "epoch": 229, "step": 143328}
{"lr": 0.00021882821934784207, "data_time": 0.08067491054534912, "grad_norm": 0.32877432405948637, "loss": 5.779047918319702, "time": 1.223728632926941, "epoch": 230, "step": 143454}
{"lr": 0.00021793391385596038, "data_time": 0.06813292503356934, "grad_norm": 0.40954460501670836, "loss": 5.794441795349121, "time": 1.1708744049072266, "epoch": 230, "step": 143554}
{"lr": 0.00021704121672200567, "data_time": 0.08272387981414794, "grad_norm": 0.38414518535137177, "loss": 5.774995756149292, "time": 1.1925501108169556, "epoch": 230, "step": 143654}
{"lr": 0.00021615013061934812, "data_time": 0.082698392868042, "grad_norm": 0.5555909931659698, "loss": 5.7809333324432375, "time": 1.1622838497161865, "epoch": 230, "step": 143754}
{"lr": 0.00021526065821653344, "data_time": 0.07170612812042236, "grad_norm": 0.6253628551959991, "loss": 5.805950164794922, "time": 1.1760618448257447, "epoch": 230, "step": 143854}
{"lr": 0.00021437280217727456, "data_time": 0.06884877681732178, "grad_norm": 0.5565850466489792, "loss": 5.8072281837463375, "time": 1.1522672653198243, "epoch": 230, "step": 143954}
{"lr": 0.00021325640905954925, "data_time": 0.08618979454040528, "grad_norm": 0.3744298994541168, "loss": 5.783818864822388, "time": 1.1791263818740845, "epoch": 231, "step": 144080}
{"lr": 0.0002123722157891794, "data_time": 0.08232612609863281, "grad_norm": 0.3681435823440552, "loss": 5.791044902801514, "time": 1.1413953065872193, "epoch": 231, "step": 144180}
{"lr": 0.00021148964753241718, "data_time": 0.08960566520690919, "grad_norm": 0.5637295812368393, "loss": 5.799612951278687, "time": 1.1580400228500367, "epoch": 231, "step": 144280}
{"lr": 0.0002106087069322994, "data_time": 0.09644021987915039, "grad_norm": 0.4113847345113754, "loss": 5.766573333740235, "time": 1.1559345245361328, "epoch": 231, "step": 144380}
{"lr": 0.00020972939662698941, "data_time": 0.09132015705108643, "grad_norm": 0.3882318288087845, "loss": 5.765857172012329, "time": 1.152491855621338, "epoch": 231, "step": 144480}
{"lr": 0.0002088517192497673, "data_time": 0.08310353755950928, "grad_norm": 0.3417552292346954, "loss": 5.794068193435669, "time": 1.138245987892151, "epoch": 231, "step": 144580}
{"lr": 0.0002077481747836216, "data_time": 0.06520888805389405, "grad_norm": 0.358804939687252, "loss": 5.804124402999878, "time": 1.1590627431869507, "epoch": 232, "step": 144706}
{"lr": 0.00020687419749876434, "data_time": 0.09413104057312012, "grad_norm": 0.4678418666124344, "loss": 5.790010929107666, "time": 1.1208242416381835, "epoch": 232, "step": 144806}
{"lr": 0.0002060018616924923, "data_time": 0.08781716823577881, "grad_norm": 0.41680169105529785, "loss": 5.782616758346558, "time": 1.1921539545059203, "epoch": 232, "step": 144906}
{"lr": 0.00020513116997719977, "data_time": 0.08654594421386719, "grad_norm": 0.38618104457855223, "loss": 5.8313652038574215, "time": 1.1266148328781127, "epoch": 232, "step": 145006}
{"lr": 0.0002042621249603569, "data_time": 0.07010114192962646, "grad_norm": 0.33027598857879636, "loss": 5.777177572250366, "time": 1.156262755393982, "epoch": 232, "step": 145106}
{"lr": 0.00020339472924450335, "data_time": 0.08252909183502197, "grad_norm": 0.5702888250350953, "loss": 5.766435337066651, "time": 1.1271883726119996, "epoch": 232, "step": 145206}
{"lr": 0.00020230416293577362, "data_time": 0.0755835771560669, "grad_norm": 0.6048224925994873, "loss": 5.76473217010498, "time": 1.1516791343688966, "epoch": 233, "step": 145332}
{"lr": 0.00020144050420153816, "data_time": 0.0918926477432251, "grad_norm": 0.3573628753423691, "loss": 5.819333457946778, "time": 1.139470911026001, "epoch": 233, "step": 145432}
{"lr": 0.00020057850321823125, "data_time": 0.07140748500823975, "grad_norm": 0.3609772205352783, "loss": 5.78171672821045, "time": 1.181435799598694, "epoch": 233, "step": 145532}
{"lr": 0.00019971816256729704, "data_time": 0.09486699104309082, "grad_norm": 0.3618034332990646, "loss": 5.782682609558106, "time": 1.1321046352386475, "epoch": 233, "step": 145632}
{"lr": 0.00019885948482520772, "data_time": 0.07034914493560791, "grad_norm": 1.603577771782875, "loss": 5.779383850097656, "time": 1.1630973815917969, "epoch": 233, "step": 145732}
{"lr": 0.00019800247256345505, "data_time": 0.10399363040924073, "grad_norm": 0.3771954596042633, "loss": 5.8133138656616214, "time": 1.1211416959762572, "epoch": 233, "step": 145832}
{"lr": 0.00019692501239493344, "data_time": 0.08359251022338868, "grad_norm": 0.4012915432453156, "loss": 5.828849506378174, "time": 1.1721033811569215, "epoch": 234, "step": 145958}
{"lr": 0.0001960717735655016, "data_time": 0.08645410537719726, "grad_norm": 0.3142498642206192, "loss": 5.763548469543457, "time": 1.1514462947845459, "epoch": 234, "step": 146058}
{"lr": 0.00019522020856479707, "data_time": 0.09182031154632568, "grad_norm": 0.4577339768409729, "loss": 5.764734077453613, "time": 1.1820291042327882, "epoch": 234, "step": 146158}
{"lr": 0.0001943703199430113, "data_time": 0.08217439651489258, "grad_norm": 0.45572789907455447, "loss": 5.7574375629425045, "time": 1.128169345855713, "epoch": 234, "step": 146258}
{"lr": 0.00019352211024531554, "data_time": 0.10031321048736572, "grad_norm": 0.36713957488536836, "loss": 5.84942398071289, "time": 1.153548574447632, "epoch": 234, "step": 146358}
{"lr": 0.00019267558201185316, "data_time": 0.09075031280517579, "grad_norm": 0.35214991867542267, "loss": 5.74454402923584, "time": 1.1219730615615844, "epoch": 234, "step": 146458}
{"lr": 0.00019161135442826715, "data_time": 0.09625890254974365, "grad_norm": 0.36743144094944, "loss": 5.791076278686523, "time": 1.1518132209777832, "epoch": 235, "step": 146584}
{"lr": 0.0001907686356349982, "data_time": 0.10766398906707764, "grad_norm": 0.4015147924423218, "loss": 5.808550548553467, "time": 1.1450820207595824, "epoch": 235, "step": 146684}
{"lr": 0.00018992760655182485, "data_time": 0.08180508613586426, "grad_norm": 0.3755199730396271, "loss": 5.781462383270264, "time": 1.1516366243362426, "epoch": 235, "step": 146784}
{"lr": 0.0001890882696973855, "data_time": 0.08933100700378419, "grad_norm": 0.45036722123622897, "loss": 5.757919836044311, "time": 1.1219177484512328, "epoch": 235, "step": 146884}
{"lr": 0.00018825062758525226, "data_time": 0.07506699562072754, "grad_norm": 0.3480181097984314, "loss": 5.798154497146607, "time": 1.1219653367996216, "epoch": 235, "step": 146984}
{"lr": 0.0001874146827239213, "data_time": 0.09606707096099854, "grad_norm": 0.458583265542984, "loss": 5.765235519409179, "time": 1.1131927728652955, "epoch": 235, "step": 147084}
{"lr": 0.0001863638126170966, "data_time": 0.0771336317062378, "grad_norm": 0.4913405865430832, "loss": 5.847097778320313, "time": 1.1541708946228026, "epoch": 236, "step": 147210}
{"lr": 0.00018553171275677795, "data_time": 0.0829322099685669, "grad_norm": 0.35764620900154115, "loss": 5.763581609725952, "time": 1.1438170194625854, "epoch": 236, "step": 147310}
{"lr": 0.00018470131828962684, "data_time": 0.0762484312057495, "grad_norm": 0.3443481206893921, "loss": 5.793811416625976, "time": 1.1534202098846436, "epoch": 236, "step": 147410}
{"lr": 0.00018387263170243544, "data_time": 0.09933207035064698, "grad_norm": 1.1690261036157608, "loss": 5.775680112838745, "time": 1.1358129024505614, "epoch": 236, "step": 147510}
{"lr": 0.00018304565547688044, "data_time": 0.08591320514678955, "grad_norm": 0.37107275426387787, "loss": 5.779628419876099, "time": 1.1644757747650147, "epoch": 236, "step": 147610}
{"lr": 0.00018222039208951694, "data_time": 0.0977529764175415, "grad_norm": 0.3131284713745117, "loss": 5.730849027633667, "time": 1.1327041864395142, "epoch": 236, "step": 147710}
{"lr": 0.00018118300278372008, "data_time": 0.10953671932220459, "grad_norm": 0.5423652410507203, "loss": 5.788910531997681, "time": 1.1845283031463623, "epoch": 237, "step": 147836}
{"lr": 0.00018036161950695914, "data_time": 0.09417257308959961, "grad_norm": 0.4414084434509277, "loss": 5.823451137542724, "time": 1.1490997552871705, "epoch": 237, "step": 147936}
{"lr": 0.00017954195710630333, "data_time": 0.0950998067855835, "grad_norm": 0.4141547858715057, "loss": 5.8276573657989506, "time": 1.1593731880187987, "epoch": 237, "step": 148036}
{"lr": 0.0001787240180364046, "data_time": 0.08719527721405029, "grad_norm": 0.5191623628139496, "loss": 5.7823401927948, "time": 1.1564382791519165, "epoch": 237, "step": 148136}
{"lr": 0.00017790780474675428, "data_time": 0.09592375755310059, "grad_norm": 0.6501368373632431, "loss": 5.763240671157837, "time": 1.1749902725219727, "epoch": 237, "step": 148236}
{"lr": 0.00017709331968167576, "data_time": 0.08296680450439453, "grad_norm": 0.35734902173280714, "loss": 5.753174114227295, "time": 1.1189055442810059, "epoch": 237, "step": 148336}
{"lr": 0.00017606953291914097, "data_time": 0.0999070405960083, "grad_norm": 0.36596898436546327, "loss": 5.809824657440186, "time": 1.16639187335968, "epoch": 238, "step": 148462}
{"lr": 0.00017525896261890768, "data_time": 0.0719186782836914, "grad_norm": 0.3417811721563339, "loss": 5.778141880035401, "time": 1.1533417224884033, "epoch": 238, "step": 148562}
{"lr": 0.00017445012847576365, "data_time": 0.08189506530761718, "grad_norm": 0.36317709982395174, "loss": 5.810060691833496, "time": 1.1439709424972535, "epoch": 238, "step": 148662}
{"lr": 0.00017364303291193416, "data_time": 0.08509349822998047, "grad_norm": 2.7210209250450133, "loss": 5.819358777999878, "time": 1.1547695875167847, "epoch": 238, "step": 148762}
{"lr": 0.00017283767834443752, "data_time": 0.08572735786437988, "grad_norm": 0.30947440564632417, "loss": 5.801280689239502, "time": 1.1298358917236329, "epoch": 238, "step": 148862}
{"lr": 0.00017203406718507762, "data_time": 0.06949901580810547, "grad_norm": 0.3917178809642792, "loss": 5.727130031585693, "time": 1.1316831350326537, "epoch": 238, "step": 148962}
{"lr": 0.0001710240031117207, "data_time": 0.1131129503250122, "grad_norm": 0.4130794286727905, "loss": 5.7597434520721436, "time": 1.1792989015579223, "epoch": 239, "step": 149088}
{"lr": 0.0001702243409120332, "data_time": 0.08144454956054688, "grad_norm": 0.37161506414413453, "loss": 5.832103109359741, "time": 1.1502427339553833, "epoch": 239, "step": 149188}
{"lr": 0.00016942642994667405, "data_time": 0.08974983692169189, "grad_norm": 0.9260107994079589, "loss": 5.802710247039795, "time": 1.1387399911880494, "epoch": 239, "step": 149288}
{"lr": 0.00016863027260515692, "data_time": 0.08276762962341308, "grad_norm": 0.37333790212869644, "loss": 5.809211349487304, "time": 1.1378005981445312, "epoch": 239, "step": 149388}
{"lr": 0.00016783587127174257, "data_time": 0.10904321670532227, "grad_norm": 0.3625162273645401, "loss": 5.7814719676971436, "time": 1.152052140235901, "epoch": 239, "step": 149488}
{"lr": 0.0001670432283254343, "data_time": 0.08049848079681396, "grad_norm": 0.4807463109493256, "loss": 5.6979625701904295, "time": 1.1284282445907592, "epoch": 239, "step": 149588}
{"lr": 0.00016604700547675287, "data_time": 0.07628943920135497, "grad_norm": 0.3671208530664444, "loss": 5.767177486419678, "time": 1.1620601415634155, "epoch": 240, "step": 149714}
{"lr": 0.000165258345221516, "data_time": 0.08094229698181152, "grad_norm": 0.5390792578458786, "loss": 5.788657236099243, "time": 1.1235010623931885, "epoch": 240, "step": 149814}
{"lr": 0.00016447145107233166, "data_time": 0.07457511425018311, "grad_norm": 0.30763678848743437, "loss": 5.8066448211669925, "time": 1.1241458892822265, "epoch": 240, "step": 149914}
{"lr": 0.0001636863253857205, "data_time": 0.09090821743011475, "grad_norm": 0.5731742739677429, "loss": 5.774758958816529, "time": 1.1171563625335694, "epoch": 240, "step": 150014}
{"lr": 0.0001629029705129075, "data_time": 0.07781178951263427, "grad_norm": 0.6492692291736603, "loss": 5.747259044647217, "time": 1.1506725788116454, "epoch": 240, "step": 150114}
{"lr": 0.0001621213887998145, "data_time": 0.09172019958496094, "grad_norm": 0.3279366433620453, "loss": 5.79118390083313, "time": 1.1228790998458862, "epoch": 240, "step": 150214}
{"lr": 0.0001611391240869765, "data_time": 0.1106421947479248, "grad_norm": 0.3334874302148819, "loss": 5.743478679656983, "time": 1.1755435705184936, "epoch": 241, "step": 150340}
{"lr": 0.00016036155832896835, "data_time": 0.07931547164916992, "grad_norm": 0.39429150223731996, "loss": 5.805735015869141, "time": 1.1131664276123048, "epoch": 241, "step": 150440}
{"lr": 0.0001595857733414756, "data_time": 0.09809069633483887, "grad_norm": 0.328505802154541, "loss": 5.780761861801148, "time": 1.1501173734664918, "epoch": 241, "step": 150540}
{"lr": 0.00015881177144775108, "data_time": 0.08705298900604248, "grad_norm": 0.32038585245609286, "loss": 5.7522687911987305, "time": 1.1377085208892823, "epoch": 241, "step": 150640}
{"lr": 0.0001580395549657066, "data_time": 0.09063706398010254, "grad_norm": 0.38822020292282106, "loss": 5.762095355987549, "time": 1.1354414224624634, "epoch": 241, "step": 150740}
{"lr": 0.00015726912620790825, "data_time": 0.0809023380279541, "grad_norm": 0.45661344230175016, "loss": 5.799965047836304, "time": 1.1245941162109374, "epoch": 241, "step": 150840}
{"lr": 0.00015630093490403307, "data_time": 0.07943029403686523, "grad_norm": 0.38482271134853363, "loss": 5.799509048461914, "time": 1.2050023078918457, "epoch": 242, "step": 150966}
{"lr": 0.00015553455489404202, "data_time": 0.06455955505371094, "grad_norm": 0.5031806647777557, "loss": 5.84288535118103, "time": 1.1492462873458862, "epoch": 242, "step": 151066}
{"lr": 0.00015476997011004942, "data_time": 0.07258658409118653, "grad_norm": 0.41600541472435, "loss": 5.8012628078460695, "time": 1.1737199783325196, "epoch": 242, "step": 151166}
{"lr": 0.00015400718284176525, "data_time": 0.07029788494110108, "grad_norm": 0.7447382032871246, "loss": 5.787922954559326, "time": 1.1462757110595703, "epoch": 242, "step": 151266}
{"lr": 0.00015324619537351777, "data_time": 0.08377208709716796, "grad_norm": 0.34863389283418655, "loss": 5.773152494430542, "time": 1.1581945180892945, "epoch": 242, "step": 151366}
{"lr": 0.00015248700998424464, "data_time": 0.0770871639251709, "grad_norm": 0.48408488035202024, "loss": 5.801655864715576, "time": 1.134101891517639, "epoch": 242, "step": 151466}
{"lr": 0.0001515330057108747, "data_time": 0.10137920379638672, "grad_norm": 0.4507765740156174, "loss": 5.785968065261841, "time": 1.1895692110061646, "epoch": 243, "step": 151592}
{"lr": 0.00015077790138699336, "data_time": 0.09390676021575928, "grad_norm": 0.37775078117847444, "loss": 5.793779468536377, "time": 1.1337115287780761, "epoch": 243, "step": 151692}
{"lr": 0.00015002460653391482, "data_time": 0.1017338514328003, "grad_norm": 0.5181674957275391, "loss": 5.770510244369507, "time": 1.1542938709259034, "epoch": 243, "step": 151792}
{"lr": 0.0001492731234075394, "data_time": 0.07971422672271729, "grad_norm": 0.47068157196044924, "loss": 5.775929403305054, "time": 1.128226137161255, "epoch": 243, "step": 151892}
{"lr": 0.00014852345425834208, "data_time": 0.09147021770477295, "grad_norm": 0.4266134947538376, "loss": 5.80249285697937, "time": 1.1586535453796387, "epoch": 243, "step": 151992}
{"lr": 0.000147775601331366, "data_time": 0.09738395214080811, "grad_norm": 0.4369888216257095, "loss": 5.811405229568481, "time": 1.1465930461883544, "epoch": 243, "step": 152092}
{"lr": 0.00014683589604513232, "data_time": 0.08103981018066406, "grad_norm": 0.3528355538845062, "loss": 5.756355047225952, "time": 1.1320122957229615, "epoch": 244, "step": 152218}
{"lr": 0.00014609215602220125, "data_time": 0.09110639095306397, "grad_norm": 0.3713892698287964, "loss": 5.759419870376587, "time": 1.111910843849182, "epoch": 244, "step": 152318}
{"lr": 0.00014535023950252774, "data_time": 0.06978139877319336, "grad_norm": 0.9166756391525268, "loss": 5.708377265930176, "time": 1.142241644859314, "epoch": 244, "step": 152418}
{"lr": 0.00014461014870793763, "data_time": 0.09434101581573487, "grad_norm": 0.37026721239089966, "loss": 5.764369153976441, "time": 1.1224659204483032, "epoch": 244, "step": 152518}
{"lr": 0.00014387188585478905, "data_time": 0.07963802814483642, "grad_norm": 0.33943328857421873, "loss": 5.798433923721314, "time": 1.1430835962295531, "epoch": 244, "step": 152618}
{"lr": 0.00014313545315396676, "data_time": 0.09445679187774658, "grad_norm": 0.3333191603422165, "loss": 5.77093358039856, "time": 1.116184663772583, "epoch": 244, "step": 152718}
{"lr": 0.00014221015713345152, "data_time": 0.08649988174438476, "grad_norm": 0.44944917857646943, "loss": 5.778038740158081, "time": 1.1489333152770995, "epoch": 245, "step": 152844}
{"lr": 0.00014147786869265984, "data_time": 0.08304765224456787, "grad_norm": 0.3542033612728119, "loss": 5.769199275970459, "time": 1.1139482736587525, "epoch": 245, "step": 152944}
{"lr": 0.00014074741757358426, "data_time": 0.10152523517608643, "grad_norm": 0.36265529990196227, "loss": 5.816446924209595, "time": 1.1553549289703369, "epoch": 245, "step": 153044}
{"lr": 0.0001400188059637156, "data_time": 0.07504205703735352, "grad_norm": 0.4564021110534668, "loss": 5.738273477554321, "time": 1.1465099096298217, "epoch": 245, "step": 153144}
{"lr": 0.00013929203604503505, "data_time": 0.11108849048614503, "grad_norm": 0.48813303709030154, "loss": 5.775686264038086, "time": 1.1538785457611085, "epoch": 245, "step": 153244}
{"lr": 0.00013856710999400861, "data_time": 0.08620798587799072, "grad_norm": 0.6380798667669296, "loss": 5.790521287918091, "time": 1.1081463813781738, "epoch": 245, "step": 153344}
{"lr": 0.0001376563318268023, "data_time": 0.06724293231964111, "grad_norm": NaN, "loss": 5.739969158172608, "time": 1.1573368310928345, "epoch": 246, "step": 153470}
{"lr": 0.00013693558090544511, "data_time": 0.0815669059753418, "grad_norm": 0.47283260226249696, "loss": 5.790373849868774, "time": 1.15425705909729, "epoch": 246, "step": 153570}
{"lr": 0.00013621668090864556, "data_time": 0.07345128059387207, "grad_norm": 0.4080804169178009, "loss": 5.770839405059815, "time": 1.1381874561309815, "epoch": 246, "step": 153670}
{"lr": 0.00013549963398930105, "data_time": 0.07441861629486084, "grad_norm": 0.36995136737823486, "loss": 5.752910995483399, "time": 1.1288056135177613, "epoch": 246, "step": 153770}
{"lr": 0.00013478444229476057, "data_time": 0.08975675106048583, "grad_norm": 0.5224605888128281, "loss": 5.749049043655395, "time": 1.145700454711914, "epoch": 246, "step": 153870}
{"lr": 0.0001340711079668169, "data_time": 0.06965072154998779, "grad_norm": 1.2892817825078964, "loss": 5.750385761260986, "time": 1.122023868560791, "epoch": 246, "step": 153970}
{"lr": 0.00013317495453677518, "data_time": 0.09070565700531005, "grad_norm": 0.4787515699863434, "loss": 5.798700571060181, "time": 1.190072512626648, "epoch": 247, "step": 154096}
{"lr": 0.00013246582571816915, "data_time": 0.09399003982543945, "grad_norm": 0.3887360364198685, "loss": 5.822479200363159, "time": 1.1406051874160767, "epoch": 247, "step": 154196}
{"lr": 0.0001317585612097474, "data_time": 0.09221217632293702, "grad_norm": 0.4481057286262512, "loss": 5.7984518051147464, "time": 1.1713525772094726, "epoch": 247, "step": 154296}
{"lr": 0.00013105316312956325, "data_time": 0.07549548149108887, "grad_norm": 0.5879404455423355, "loss": 5.806184434890747, "time": 1.1314701795578004, "epoch": 247, "step": 154396}
{"lr": 0.00013034963359008007, "data_time": 0.11109106540679932, "grad_norm": 0.298465271294117, "loss": 5.768896865844726, "time": 1.1683011293411254, "epoch": 247, "step": 154496}
{"lr": 0.00012964797469816564, "data_time": 0.0924489974975586, "grad_norm": 0.8938257545232773, "loss": 5.7957014560699465, "time": 1.1753090858459472, "epoch": 247, "step": 154596}
{"lr": 0.00012876655117286814, "data_time": 0.06866276264190674, "grad_norm": 0.4131979435682297, "loss": 5.79362473487854, "time": 1.1631155014038086, "epoch": 248, "step": 154722}
{"lr": 0.0001280691276764246, "data_time": 0.0909661054611206, "grad_norm": 0.3764687955379486, "loss": 5.783654356002808, "time": 1.1308702945709228, "epoch": 248, "step": 154822}
{"lr": 0.00012737358165700675, "data_time": 0.06898238658905029, "grad_norm": 2.0919087797403337, "loss": 5.754807043075561, "time": 1.1282005548477172, "epoch": 248, "step": 154922}
{"lr": 0.00012667991519757445, "data_time": 0.09344813823699952, "grad_norm": 0.3349137991666794, "loss": 5.73399133682251, "time": 1.1236137390136718, "epoch": 248, "step": 155022}
{"lr": 0.00012598813037545842, "data_time": 0.06963768005371093, "grad_norm": 0.3761751264333725, "loss": 5.766470623016358, "time": 1.115975546836853, "epoch": 248, "step": 155122}
{"lr": 0.0001252982292623545, "data_time": 0.08957571983337402, "grad_norm": 0.45869594216346743, "loss": 5.778063106536865, "time": 1.1240641832351685, "epoch": 248, "step": 155222}
{"lr": 0.00012443163908075895, "data_time": 0.09326074123382569, "grad_norm": 0.33385213911533357, "loss": 5.773870801925659, "time": 1.1889944791793823, "epoch": 249, "step": 155348}
{"lr": 0.00012374600275221698, "data_time": 0.08138670921325683, "grad_norm": 0.7816939353942871, "loss": 5.758875989913941, "time": 1.1529706239700317, "epoch": 249, "step": 155448}
{"lr": 0.0001230622568472123, "data_time": 0.09609663486480713, "grad_norm": 0.3344453603029251, "loss": 5.807507753372192, "time": 1.1544638872146606, "epoch": 249, "step": 155548}
{"lr": 0.0001223804034133662, "data_time": 0.08110756874084472, "grad_norm": 0.43534277081489564, "loss": 5.780541801452637, "time": 1.1367111682891846, "epoch": 249, "step": 155648}
{"lr": 0.00012170044449263323, "data_time": 0.09003243446350098, "grad_norm": 0.3695129781961441, "loss": 5.7647310256958, "time": 1.1713979721069336, "epoch": 249, "step": 155748}
{"lr": 0.00012102238212129364, "data_time": 0.0781599998474121, "grad_norm": 0.4333116412162781, "loss": 5.722969818115234, "time": 1.1323616266250611, "epoch": 249, "step": 155848}
{"lr": 0.00012017072698159937, "data_time": 0.07367727756500245, "grad_norm": 0.3177123010158539, "loss": 5.809254932403564, "time": 1.1515620470046997, "epoch": 250, "step": 155974}
{"lr": 0.000119496958283421, "data_time": 0.07028384208679199, "grad_norm": 0.41446852385997773, "loss": 5.760070848464966, "time": 1.1403920650482178, "epoch": 250, "step": 156074}
{"lr": 0.00011882509273344203, "data_time": 0.06708674430847168, "grad_norm": 0.3225682437419891, "loss": 5.75435061454773, "time": 1.1263306856155395, "epoch": 250, "step": 156174}
{"lr": 0.00011815513234370565, "data_time": 0.07720417976379394, "grad_norm": 0.3520232528448105, "loss": 5.773544454574585, "time": 1.1123795747756957, "epoch": 250, "step": 156274}
{"lr": 0.00011748707912054976, "data_time": 0.07104249000549316, "grad_norm": 0.4018701702356339, "loss": 5.752345132827759, "time": 1.1371686935424805, "epoch": 250, "step": 156374}
{"lr": 0.00011682093506460118, "data_time": 0.08147323131561279, "grad_norm": 0.4654542177915573, "loss": 5.774739789962768, "time": 1.1292467355728149, "epoch": 250, "step": 156474}
{"lr": 0.00011598431491231556, "data_time": 0.10375494956970215, "grad_norm": 0.7120180547237396, "loss": 5.787373161315918, "time": 1.1466168642044068, "epoch": 251, "step": 156600}
{"lr": 0.00011532249291424591, "data_time": 0.09200468063354492, "grad_norm": 0.47542492151260374, "loss": 5.766167116165161, "time": 1.1418381214141846, "epoch": 251, "step": 156700}
{"lr": 0.0001146625865656944, "data_time": 0.09016318321228027, "grad_norm": 0.48601248264312746, "loss": 5.764625453948975, "time": 1.1373083353042603, "epoch": 251, "step": 156800}
{"lr": 0.00011400459784289032, "data_time": 0.10061395168304443, "grad_norm": 0.32196661829948425, "loss": 5.773367786407471, "time": 1.1420790672302246, "epoch": 251, "step": 156900}
{"lr": 0.00011334852871631982, "data_time": 0.0990443229675293, "grad_norm": 0.3853831648826599, "loss": 5.730289268493652, "time": 1.1185272216796875, "epoch": 251, "step": 157000}
{"lr": 0.00011269438115072034, "data_time": 0.08962523937225342, "grad_norm": 0.3659596472978592, "loss": 5.776445722579956, "time": 1.124234175682068, "epoch": 251, "step": 157100}
{"lr": 0.00011187289416692769, "data_time": 0.07539474964141846, "grad_norm": 0.37411381006240846, "loss": 5.790820646286011, "time": 1.1719306230545044, "epoch": 252, "step": 157226}
{"lr": 0.00011122309653671253, "data_time": 0.10462405681610107, "grad_norm": 0.43073598146438596, "loss": 5.720920610427856, "time": 1.130401849746704, "epoch": 252, "step": 157326}
{"lr": 0.00011057522683252453, "data_time": 0.08564302921295167, "grad_norm": 0.41190997064113616, "loss": 5.77963809967041, "time": 1.1327893733978271, "epoch": 252, "step": 157426}
{"lr": 0.00010992928699454689, "data_time": 0.09273998737335205, "grad_norm": 0.40972971022129057, "loss": 5.791935443878174, "time": 1.1190600872039795, "epoch": 252, "step": 157526}
{"lr": 0.00010928527895718226, "data_time": 0.08353173732757568, "grad_norm": 0.35179815590381625, "loss": 5.7226794242858885, "time": 1.1511516094207763, "epoch": 252, "step": 157626}
{"lr": 0.00010864320464904928, "data_time": 0.09037444591522217, "grad_norm": 0.39466372728347776, "loss": 5.811177396774292, "time": 1.1267634630203247, "epoch": 252, "step": 157726}
{"lr": 0.00010783694723888961, "data_time": 0.11198704242706299, "grad_norm": 0.40943554043769836, "loss": 5.782953929901123, "time": 1.188768482208252, "epoch": 253, "step": 157852}
{"lr": 0.00010719925023316176, "data_time": 0.08360636234283447, "grad_norm": 0.3921759933233261, "loss": 5.747098684310913, "time": 1.1347854137420654, "epoch": 253, "step": 157952}
{"lr": 0.00010656349320371957, "data_time": 0.10813341140747071, "grad_norm": 0.3615278214216232, "loss": 5.766953372955323, "time": 1.1582031488418578, "epoch": 253, "step": 158052}
{"lr": 0.00010592967805447175, "data_time": 0.095298171043396, "grad_norm": 0.39145264625549314, "loss": 5.75675139427185, "time": 1.1192518711090087, "epoch": 253, "step": 158152}
{"lr": 0.00010529780668351204, "data_time": 0.10357058048248291, "grad_norm": 1.7136290282011033, "loss": 5.771187496185303, "time": 1.1540024280548096, "epoch": 253, "step": 158252}
{"lr": 0.00010466788098311246, "data_time": 0.08472611904144287, "grad_norm": 0.4004722237586975, "loss": 5.753530025482178, "time": 1.1518787145614624, "epoch": 253, "step": 158352}
{"lr": 0.00010387694776446734, "data_time": 0.06708862781524658, "grad_norm": 0.34940231442451475, "loss": 5.770198774337769, "time": 1.1452804803848267, "epoch": 254, "step": 158478}
{"lr": 0.0001032514262197981, "data_time": 0.07272698879241943, "grad_norm": 0.3508256316184998, "loss": 5.7657159805297855, "time": 1.149019432067871, "epoch": 254, "step": 158578}
{"lr": 0.00010262785647400789, "data_time": 0.08272573947906495, "grad_norm": 0.3844974279403687, "loss": 5.746848773956299, "time": 1.153124451637268, "epoch": 254, "step": 158678}
{"lr": 0.0001020062403945084, "data_time": 0.07968752384185791, "grad_norm": 0.37092725932598114, "loss": 5.781133460998535, "time": 1.1562062740325927, "epoch": 254, "step": 158778}
{"lr": 0.00010138657984286, "data_time": 0.09673883914947509, "grad_norm": 0.3939981162548065, "loss": 5.722955846786499, "time": 1.1512654304504395, "epoch": 254, "step": 158878}
{"lr": 0.00010076887667476752, "data_time": 0.06969296932220459, "grad_norm": 0.8019292742013931, "loss": 5.759806871414185, "time": 1.1313606262207032, "epoch": 254, "step": 158978}
{"lr": 9.999336046715733e-05, "data_time": 0.11683566570281982, "grad_norm": 2.0359122067689897, "loss": 5.800741052627563, "time": 1.1569996118545531, "epoch": 255, "step": 159104}
{"lr": 9.938008779127315e-05, "data_time": 0.07912750244140625, "grad_norm": 0.6660475969314575, "loss": 5.747298336029052, "time": 1.1227171182632447, "epoch": 255, "step": 159204}
{"lr": 9.876877850780951e-05, "data_time": 0.10686585903167725, "grad_norm": 0.3824374794960022, "loss": 5.790456962585449, "time": 1.165967011451721, "epoch": 255, "step": 159304}
{"lr": 9.815943444746088e-05, "data_time": 0.08227860927581787, "grad_norm": 0.38513678014278413, "loss": 5.754270696640015, "time": 1.134126567840576, "epoch": 255, "step": 159404}
{"lr": 9.755205743503724e-05, "data_time": 0.09021296501159667, "grad_norm": 0.4126559764146805, "loss": 5.740778064727783, "time": 1.1393055200576783, "epoch": 255, "step": 159504}
{"lr": 9.694664928945768e-05, "data_time": 0.07783045768737792, "grad_norm": 0.4395492017269135, "loss": 5.771453094482422, "time": 1.1348999261856079, "epoch": 255, "step": 159604}
{"lr": 9.618664110314712e-05, "data_time": 0.07816865444183349, "grad_norm": 0.47633311450481414, "loss": 5.722107219696045, "time": 1.178165602684021, "epoch": 256, "step": 159730}
{"lr": 9.558568926631593e-05, "data_time": 0.08323359489440918, "grad_norm": 0.36807602643966675, "loss": 5.742690563201904, "time": 1.1350136041641234, "epoch": 256, "step": 159830}
{"lr": 9.498671218503348e-05, "data_time": 0.06596939563751221, "grad_norm": 0.3733653724193573, "loss": 5.7749128341674805, "time": 1.1675300598144531, "epoch": 256, "step": 159930}
{"lr": 9.438971165306342e-05, "data_time": 0.08949651718139648, "grad_norm": 0.40307962000370023, "loss": 5.727363443374633, "time": 1.1248482704162597, "epoch": 256, "step": 160030}
{"lr": 9.379468945825018e-05, "data_time": 0.07335760593414306, "grad_norm": 0.3413607835769653, "loss": 5.747158002853394, "time": 1.1693635940551759, "epoch": 256, "step": 160130}
{"lr": 9.320164738251308e-05, "data_time": 0.08515655994415283, "grad_norm": 0.37009196579456327, "loss": 5.768011236190796, "time": 1.1351900339126586, "epoch": 256, "step": 160230}
{"lr": 9.245723640783347e-05, "data_time": 0.0829787015914917, "grad_norm": 0.41317311823368075, "loss": 5.76592264175415, "time": 1.1783053874969482, "epoch": 257, "step": 160356}
{"lr": 9.186867593441736e-05, "data_time": 0.09289882183074952, "grad_norm": 0.3757458388805389, "loss": 5.812195873260498, "time": 1.1431585550308228, "epoch": 257, "step": 160456}
{"lr": 9.128210134793239e-05, "data_time": 0.0722114086151123, "grad_norm": 0.6763582348823547, "loss": 5.775519609451294, "time": 1.1449946403503417, "epoch": 257, "step": 160556}
{"lr": 9.069751440500019e-05, "data_time": 0.08835909366607667, "grad_norm": 0.4028261870145798, "loss": 5.767878437042237, "time": 1.1289905071258546, "epoch": 257, "step": 160656}
{"lr": 9.011491685628975e-05, "data_time": 0.07424020767211914, "grad_norm": 0.3566751092672348, "loss": 5.772627592086792, "time": 1.132437539100647, "epoch": 257, "step": 160756}
{"lr": 8.953431044651275e-05, "data_time": 0.08448030948638915, "grad_norm": 0.3833979547023773, "loss": 5.774602890014648, "time": 1.113528800010681, "epoch": 257, "step": 160856}
{"lr": 8.880558404339188e-05, "data_time": 0.08354322910308838, "grad_norm": 0.44475882947444917, "loss": 5.747508764266968, "time": 1.2023401260375977, "epoch": 258, "step": 160982}
{"lr": 8.822948400357233e-05, "data_time": 0.06921601295471191, "grad_norm": 0.3158591270446777, "loss": 5.7865808486938475, "time": 1.1301526784896851, "epoch": 258, "step": 161082}
{"lr": 8.765538074901332e-05, "data_time": 0.09069104194641113, "grad_norm": 0.4123353987932205, "loss": 5.744429445266723, "time": 1.1467918395996093, "epoch": 258, "step": 161182}
{"lr": 8.708327599898828e-05, "data_time": 0.083518385887146, "grad_norm": 0.37334443628787994, "loss": 5.772551965713501, "time": 1.1372550964355468, "epoch": 258, "step": 161282}
{"lr": 8.65131714667863e-05, "data_time": 0.07152225971221923, "grad_norm": 0.3439988136291504, "loss": 5.732161951065064, "time": 1.169676375389099, "epoch": 258, "step": 161382}
{"lr": 8.594506885970543e-05, "data_time": 0.06669211387634277, "grad_norm": 0.37930641770362855, "loss": 5.772709131240845, "time": 1.1492722034454346, "epoch": 258, "step": 161482}
{"lr": 8.523211254741786e-05, "data_time": 0.10249989032745362, "grad_norm": 0.38997062742710115, "loss": 5.743217802047729, "time": 1.1669665336608888, "epoch": 259, "step": 161608}
{"lr": 8.466854054908857e-05, "data_time": 0.09369113445281982, "grad_norm": 0.415522500872612, "loss": 5.748693656921387, "time": 1.1358316659927368, "epoch": 259, "step": 161708}
{"lr": 8.410697600001777e-05, "data_time": 0.08598952293395996, "grad_norm": 0.4096496939659119, "loss": 5.747376728057861, "time": 1.1458932876586914, "epoch": 259, "step": 161808}
{"lr": 8.354742058192956e-05, "data_time": 0.0785353422164917, "grad_norm": 0.448575434088707, "loss": 5.761192035675049, "time": 1.142290425300598, "epoch": 259, "step": 161908}
{"lr": 8.298987597053043e-05, "data_time": 0.11433312892913819, "grad_norm": 0.4225851148366928, "loss": 5.746141004562378, "time": 1.1493013381958008, "epoch": 259, "step": 162008}
{"lr": 8.243434383550581e-05, "data_time": 0.09168627262115478, "grad_norm": 0.3434442013502121, "loss": 5.798742628097534, "time": 1.1536666870117187, "epoch": 259, "step": 162108}
{"lr": 8.173724128263602e-05, "data_time": 0.07136895656585693, "grad_norm": 0.4296407371759415, "loss": 5.745871686935425, "time": 1.1644436597824097, "epoch": 260, "step": 162234}
{"lr": 8.118626346346974e-05, "data_time": 0.08516800403594971, "grad_norm": 0.34349716305732725, "loss": 5.734437751770019, "time": 1.1499226331710815, "epoch": 260, "step": 162334}
{"lr": 8.063730352197701e-05, "data_time": 0.06648187637329102, "grad_norm": 0.44142367541790006, "loss": 5.768960332870483, "time": 1.1753209352493286, "epoch": 260, "step": 162434}
{"lr": 8.009036310213416e-05, "data_time": 0.08738520145416259, "grad_norm": 2.22181396484375, "loss": 5.741491317749023, "time": 1.1751598119735718, "epoch": 260, "step": 162534}
{"lr": 7.954544384187028e-05, "data_time": 0.06704046726226806, "grad_norm": 0.434308385848999, "loss": 5.738249683380127, "time": 1.1778010129928589, "epoch": 260, "step": 162634}
{"lr": 7.900254737306071e-05, "data_time": 0.09029014110565185, "grad_norm": 0.3412525326013565, "loss": 5.742122030258178, "time": 1.155939245223999, "epoch": 260, "step": 162734}
{"lr": 7.832138038768398e-05, "data_time": 0.09460906982421875, "grad_norm": 1.3234595686197281, "loss": 5.799000835418701, "time": 1.2248484134674071, "epoch": 261, "step": 162860}
{"lr": 7.778306140737107e-05, "data_time": 0.09354913234710693, "grad_norm": 0.3468974530696869, "loss": 5.683085680007935, "time": 1.150629472732544, "epoch": 261, "step": 162960}
{"lr": 7.724677049633939e-05, "data_time": 0.10471270084381104, "grad_norm": 0.33620011508464814, "loss": 5.736225128173828, "time": 1.1355162143707276, "epoch": 261, "step": 163060}
{"lr": 7.671250926062567e-05, "data_time": 0.08481671810150146, "grad_norm": 0.40950300097465514, "loss": 5.724966239929199, "time": 1.1300506353378297, "epoch": 261, "step": 163160}
{"lr": 7.618027930018809e-05, "data_time": 0.1275503396987915, "grad_norm": 0.38400216698646544, "loss": 5.726879453659057, "time": 1.1556410789489746, "epoch": 261, "step": 163260}
{"lr": 7.565008220890152e-05, "data_time": 0.08667597770690919, "grad_norm": 0.3512387633323669, "loss": 5.7538519382476805, "time": 1.1341594934463501, "epoch": 261, "step": 163360}
{"lr": 7.49849307289838e-05, "data_time": 0.08741426467895508, "grad_norm": 0.3819189578294754, "loss": 5.786046886444092, "time": 1.1578183889389038, "epoch": 262, "step": 163486}
{"lr": 7.445933376164289e-05, "data_time": 0.08492631912231445, "grad_norm": 0.33211687207221985, "loss": 5.714509630203247, "time": 1.146024513244629, "epoch": 262, "step": 163586}
{"lr": 7.393577481718848e-05, "data_time": 0.09131433963775634, "grad_norm": 0.4145526051521301, "loss": 5.708537149429321, "time": 1.1753360033035278, "epoch": 262, "step": 163686}
{"lr": 7.341425546352844e-05, "data_time": 0.08003466129302979, "grad_norm": 0.36552774012088773, "loss": 5.798552560806274, "time": 1.1389586448669433, "epoch": 262, "step": 163786}
{"lr": 7.289477726246286e-05, "data_time": 0.08422064781188965, "grad_norm": 0.32797254621982574, "loss": 5.741376495361328, "time": 1.1401875972747804, "epoch": 262, "step": 163886}
{"lr": 7.23773417696788e-05, "data_time": 0.08186774253845215, "grad_norm": 0.4510981023311615, "loss": 5.72013897895813, "time": 1.145850944519043, "epoch": 262, "step": 163986}
{"lr": 7.172828385369495e-05, "data_time": 0.09502148628234863, "grad_norm": 0.3915104329586029, "loss": 5.791245603561402, "time": 1.164212131500244, "epoch": 263, "step": 164112}
{"lr": 7.121547058046107e-05, "data_time": 0.09139406681060791, "grad_norm": 0.446964305639267, "loss": 5.732020473480224, "time": 1.137699007987976, "epoch": 263, "step": 164212}
{"lr": 7.070470504454757e-05, "data_time": 0.0897826910018921, "grad_norm": NaN, "loss": 5.734700107574463, "time": 1.160252332687378, "epoch": 263, "step": 164312}
{"lr": 7.019598877555023e-05, "data_time": 0.08606140613555908, "grad_norm": 0.3691542148590088, "loss": 5.673322534561157, "time": 1.1366913795471192, "epoch": 263, "step": 164412}
{"lr": 6.9689323296927e-05, "data_time": 0.10079054832458496, "grad_norm": 0.5199244052171708, "loss": 5.783981513977051, "time": 1.1483565807342528, "epoch": 263, "step": 164512}
{"lr": 6.918471012599496e-05, "data_time": 0.0828899621963501, "grad_norm": 0.3190386176109314, "loss": 5.760017395019531, "time": 1.1210177421569825, "epoch": 263, "step": 164612}
{"lr": 6.855182194377018e-05, "data_time": 0.06831843852996826, "grad_norm": 0.40787574350833894, "loss": 5.755548572540283, "time": 1.1738235712051392, "epoch": 264, "step": 164738}
{"lr": 6.805185254555446e-05, "data_time": 0.09235208034515381, "grad_norm": 0.3983605235815048, "loss": 5.741102933883667, "time": 1.143406343460083, "epoch": 264, "step": 164838}
{"lr": 6.755394035878206e-05, "data_time": 0.07127590179443359, "grad_norm": 0.36363485455513, "loss": 5.726354503631592, "time": 1.1469137907028197, "epoch": 264, "step": 164938}
{"lr": 6.705808687455664e-05, "data_time": 0.09915368556976319, "grad_norm": 0.423150634765625, "loss": 5.79075026512146, "time": 1.1572452545166017, "epoch": 264, "step": 165038}
{"lr": 6.656429357781589e-05, "data_time": 0.06758315563201904, "grad_norm": 0.33415708541870115, "loss": 5.76365032196045, "time": 1.1512429475784303, "epoch": 264, "step": 165138}
{"lr": 6.60725619473284e-05, "data_time": 0.0957291841506958, "grad_norm": 0.35774548053741456, "loss": 5.716566371917724, "time": 1.1241297721862793, "epoch": 264, "step": 165238}
{"lr": 6.545591777109624e-05, "data_time": 0.08569316864013672, "grad_norm": 0.39652336239814756, "loss": 5.731371974945068, "time": 1.14302818775177, "epoch": 265, "step": 165364}
{"lr": 6.49688509215242e-05, "data_time": 0.07107260227203369, "grad_norm": 0.4027964621782303, "loss": 5.736865711212158, "time": 1.141807532310486, "epoch": 265, "step": 165464}
{"lr": 6.448385051609569e-05, "data_time": 0.09172518253326416, "grad_norm": 0.36994641125202177, "loss": 5.741760540008545, "time": 1.1887196063995362, "epoch": 265, "step": 165564}
{"lr": 6.400091800724686e-05, "data_time": 0.07849369049072266, "grad_norm": 0.46001358032226564, "loss": 5.793391990661621, "time": 1.1241961240768432, "epoch": 265, "step": 165664}
{"lr": 6.352005484122133e-05, "data_time": 0.08611242771148682, "grad_norm": 0.44003704786300657, "loss": 5.788587379455566, "time": 1.1519792795181274, "epoch": 265, "step": 165764}
{"lr": 6.304126245806546e-05, "data_time": 0.07155547142028809, "grad_norm": 0.3547033488750458, "loss": 5.750321865081787, "time": 1.1516875743865966, "epoch": 265, "step": 165864}
{"lr": 6.244093465375433e-05, "data_time": 0.07917029857635498, "grad_norm": 0.3704475313425064, "loss": 5.755492115020752, "time": 1.175808882713318, "epoch": 266, "step": 165990}
{"lr": 6.196682751228012e-05, "data_time": 0.06757700443267822, "grad_norm": 0.4095900684595108, "loss": 5.757775783538818, "time": 1.1514817953109742, "epoch": 266, "step": 166090}
{"lr": 6.149479580514341e-05, "data_time": 0.07674849033355713, "grad_norm": 0.37663896977901457, "loss": 5.732877397537232, "time": 1.1344271659851075, "epoch": 266, "step": 166190}
{"lr": 6.1024840945942565e-05, "data_time": 0.08030545711517334, "grad_norm": 0.3460545241832733, "loss": 5.764481067657471, "time": 1.146944260597229, "epoch": 266, "step": 166290}
{"lr": 6.0556964342056486e-05, "data_time": 0.08703625202178955, "grad_norm": 0.35783534646034243, "loss": 5.769899320602417, "time": 1.1447254180908204, "epoch": 266, "step": 166390}
{"lr": 6.009116739464109e-05, "data_time": 0.08420464992523194, "grad_norm": 0.3673717319965363, "loss": 5.76475076675415, "time": 1.1317182064056397, "epoch": 266, "step": 166490}
{"lr": 5.950722641337973e-05, "data_time": 0.1103175163269043, "grad_norm": 0.5140005588531494, "loss": 5.784167957305908, "time": 1.16228129863739, "epoch": 267, "step": 166616}
{"lr": 5.904613461857842e-05, "data_time": 0.08031582832336426, "grad_norm": 0.4476148962974548, "loss": 5.7777063846588135, "time": 1.1179278135299682, "epoch": 267, "step": 166716}
{"lr": 5.858712700474676e-05, "data_time": 0.09396986961364746, "grad_norm": 0.42621271312236786, "loss": 5.720041942596436, "time": 1.108368420600891, "epoch": 267, "step": 166816}
{"lr": 5.813020494647989e-05, "data_time": 0.09167463779449463, "grad_norm": 0.3516966372728348, "loss": 5.759512138366699, "time": 1.1337384939193726, "epoch": 267, "step": 166916}
{"lr": 5.7675369812127663e-05, "data_time": 0.10718913078308105, "grad_norm": 0.3951938897371292, "loss": 5.78127236366272, "time": 1.1448860168457031, "epoch": 267, "step": 167016}
{"lr": 5.7222622963789775e-05, "data_time": 0.09541685581207275, "grad_norm": 0.4044166922569275, "loss": 5.751982116699219, "time": 1.1257947921752929, "epoch": 267, "step": 167116}
{"lr": 5.6655137333642057e-05, "data_time": 0.06854126453399659, "grad_norm": 0.45330590903759005, "loss": 5.752873277664184, "time": 1.178441333770752, "epoch": 268, "step": 167242}
{"lr": 5.6207114996680587e-05, "data_time": 0.08061926364898682, "grad_norm": 0.3998320668935776, "loss": 5.7365587711334225, "time": 1.1590155601501464, "epoch": 268, "step": 167342}
{"lr": 5.576118534273292e-05, "data_time": 0.0671804666519165, "grad_norm": 0.45459010601043703, "loss": 5.776596927642823, "time": 1.1643911361694337, "epoch": 268, "step": 167442}
{"lr": 5.531734970722981e-05, "data_time": 0.08195281028747559, "grad_norm": 0.3312510967254639, "loss": 5.809001779556274, "time": 1.130410647392273, "epoch": 268, "step": 167542}
{"lr": 5.487560941933096e-05, "data_time": 0.0739290714263916, "grad_norm": 0.37237039506435393, "loss": 5.77931661605835, "time": 1.1454358577728272, "epoch": 268, "step": 167642}
{"lr": 5.443596580192088e-05, "data_time": 0.08383164405822754, "grad_norm": 0.37192603051662443, "loss": 5.726451110839844, "time": 1.138811993598938, "epoch": 268, "step": 167742}
{"lr": 5.388500211984165e-05, "data_time": 0.08040363788604736, "grad_norm": 0.338889142870903, "loss": 5.741161870956421, "time": 1.1757803916931153, "epoch": 269, "step": 167868}
{"lr": 5.345010181812731e-05, "data_time": 0.09213807582855224, "grad_norm": 0.37039240896701814, "loss": 5.737962150573731, "time": 1.1269494771957398, "epoch": 269, "step": 167968}
{"lr": 5.301730245588586e-05, "data_time": 0.07590487003326415, "grad_norm": 0.3166171252727509, "loss": 5.757773685455322, "time": 1.143139624595642, "epoch": 269, "step": 168068}
{"lr": 5.258660532922613e-05, "data_time": 0.084446382522583, "grad_norm": 0.3954766899347305, "loss": 5.712533807754516, "time": 1.1129740476608276, "epoch": 269, "step": 168168}
{"lr": 5.21580117279621e-05, "data_time": 0.07421095371246338, "grad_norm": 0.37196207642555235, "loss": 5.74572639465332, "time": 1.1316765546798706, "epoch": 269, "step": 168268}
{"lr": 5.1731522935607567e-05, "data_time": 0.0960543155670166, "grad_norm": 0.371764263510704, "loss": 5.753801155090332, "time": 1.1195156812667846, "epoch": 269, "step": 168368}
{"lr": 5.119714585962759e-05, "data_time": 0.0753291368484497, "grad_norm": 0.33259631097316744, "loss": 5.759149885177612, "time": 1.1948675394058228, "epoch": 270, "step": 168494}
{"lr": 5.0775418630639405e-05, "data_time": 0.0709414005279541, "grad_norm": 0.3927265167236328, "loss": 5.718443393707275, "time": 1.1619228839874267, "epoch": 270, "step": 168594}
{"lr": 5.0355800351027926e-05, "data_time": 0.08024840354919434, "grad_norm": 0.3488045156002045, "loss": 5.739958143234253, "time": 1.1566734075546266, "epoch": 270, "step": 168694}
{"lr": 4.9938292277429197e-05, "data_time": 0.06861062049865722, "grad_norm": 0.4146613210439682, "loss": 5.718755435943604, "time": 1.1508020877838134, "epoch": 270, "step": 168794}
{"lr": 4.952289566015914e-05, "data_time": 0.07746739387512207, "grad_norm": 0.3478323996067047, "loss": 5.725694799423218, "time": 1.1653418064117431, "epoch": 270, "step": 168894}
{"lr": 4.9109611743210836e-05, "data_time": 0.06608359813690186, "grad_norm": 0.4206526666879654, "loss": 5.724931764602661, "time": 1.1399616718292236, "epoch": 270, "step": 168994}
{"lr": 4.8591883984849346e-05, "data_time": 0.10045855045318604, "grad_norm": 0.4350380927324295, "loss": 5.721332454681397, "time": 1.1605103731155395, "epoch": 271, "step": 169120}
{"lr": 4.818337932014737e-05, "data_time": 0.08284657001495362, "grad_norm": 0.3739706575870514, "loss": 5.773172330856323, "time": 1.1386407136917114, "epoch": 271, "step": 169220}
{"lr": 4.777699136723188e-05, "data_time": 0.09672050476074219, "grad_norm": 0.3978067100048065, "loss": 5.738738298416138, "time": 1.604537296295166, "epoch": 271, "step": 169320}
{"lr": 4.7372721343117474e-05, "data_time": 0.09805455207824706, "grad_norm": 0.38797968327999116, "loss": 5.769476270675659, "time": 1.1439598321914672, "epoch": 271, "step": 169420}
{"lr": 4.697057045847646e-05, "data_time": 0.09738669395446778, "grad_norm": 0.3855207055807114, "loss": 5.767810678482055, "time": 1.4169532299041747, "epoch": 271, "step": 169520}
{"lr": 4.65705399176344e-05, "data_time": 0.09796898365020752, "grad_norm": 0.4156062364578247, "loss": 5.729190540313721, "time": 1.1333404779434204, "epoch": 271, "step": 169620}
{"lr": 4.606952223453943e-05, "data_time": 0.07123398780822754, "grad_norm": 0.39511339366436005, "loss": 5.737242603302002, "time": 1.1738526105880738, "epoch": 272, "step": 169746}
{"lr": 4.56742880739571e-05, "data_time": 0.11383700370788574, "grad_norm": 0.3926361471414566, "loss": 5.76159987449646, "time": 1.121191954612732, "epoch": 272, "step": 169846}
{"lr": 4.528117813916553e-05, "data_time": 0.09157941341400147, "grad_norm": 0.44359651803970335, "loss": 5.786869287490845, "time": 1.1731866121292114, "epoch": 272, "step": 169946}
{"lr": 4.489019360741536e-05, "data_time": 0.081229567527771, "grad_norm": 0.3594667881727219, "loss": 5.717581462860108, "time": 1.1262938737869264, "epoch": 272, "step": 170046}
{"lr": 4.450133564959268e-05, "data_time": 0.0764855146408081, "grad_norm": 0.47775813937187195, "loss": 5.721354961395264, "time": 1.1438637971878052, "epoch": 272, "step": 170146}
{"lr": 4.411460543021458e-05, "data_time": 0.08989112377166748, "grad_norm": 0.39799695611000063, "loss": 5.750506734848022, "time": 1.1285364389419557, "epoch": 272, "step": 170246}
{"lr": 4.363035661903268e-05, "data_time": 0.10178380012512207, "grad_norm": 0.37856597304344175, "loss": 5.751418542861939, "time": 1.194056749343872, "epoch": 273, "step": 170372}
{"lr": 4.3248439345051174e-05, "data_time": 0.07798817157745361, "grad_norm": 0.33335305750370026, "loss": 5.774227571487427, "time": 1.155825138092041, "epoch": 273, "step": 170472}
{"lr": 4.2868653561577365e-05, "data_time": 0.10652661323547363, "grad_norm": 0.3856283277273178, "loss": 5.74650936126709, "time": 1.1796772480010986, "epoch": 273, "step": 170572}
{"lr": 4.249100040595984e-05, "data_time": 0.0886077880859375, "grad_norm": 0.3512688338756561, "loss": 5.774381065368653, "time": 1.14072687625885, "epoch": 273, "step": 170672}
{"lr": 4.211548100916082e-05, "data_time": 0.11085195541381836, "grad_norm": 0.35839647948741915, "loss": 5.710602617263794, "time": 1.1745318651199341, "epoch": 273, "step": 170772}
{"lr": 4.174209649575263e-05, "data_time": 0.07780625820159912, "grad_norm": 0.8086986124515534, "loss": 5.752270126342774, "time": 1.1503887176513672, "epoch": 273, "step": 170872}
{"lr": 4.1274673385228686e-05, "data_time": 0.07001962661743164, "grad_norm": 0.3478165239095688, "loss": 5.7660445213317875, "time": 1.16757972240448, "epoch": 274, "step": 170998}
{"lr": 4.090611781753365e-05, "data_time": 0.06551663875579834, "grad_norm": 0.3357541084289551, "loss": 5.712995910644532, "time": 1.1511950492858887, "epoch": 274, "step": 171098}
{"lr": 4.0539700754923035e-05, "data_time": 0.07347559928894043, "grad_norm": 0.3814141571521759, "loss": 5.735586452484131, "time": 1.1945709228515624, "epoch": 274, "step": 171198}
{"lr": 4.0175423294710475e-05, "data_time": 0.07577359676361084, "grad_norm": 0.36984158754348756, "loss": 5.758527183532715, "time": 1.1399245500564574, "epoch": 274, "step": 171298}
{"lr": 3.981328652780162e-05, "data_time": 0.07409696578979492, "grad_norm": 0.35336928367614745, "loss": 5.753588104248047, "time": 1.1602259874343872, "epoch": 274, "step": 171398}
{"lr": 3.945329153869151e-05, "data_time": 0.07030942440032958, "grad_norm": 0.3848176449537277, "loss": 5.76283974647522, "time": 1.1296145439147949, "epoch": 274, "step": 171498}
{"lr": 3.900274898300031e-05, "data_time": 0.09796686172485351, "grad_norm": 0.35643379390239716, "loss": 5.7605822563171385, "time": 1.169931674003601, "epoch": 275, "step": 171624}
{"lr": 3.86475983732216e-05, "data_time": 0.08390860557556153, "grad_norm": 0.3262264370918274, "loss": 5.709920597076416, "time": 1.1214582920074463, "epoch": 275, "step": 171724}
{"lr": 3.829459303214128e-05, "data_time": 0.09698007106781006, "grad_norm": 0.46539921760559083, "loss": 5.739927577972412, "time": 1.145631241798401, "epoch": 275, "step": 171824}
{"lr": 3.79437340169086e-05, "data_time": 0.08331599235534667, "grad_norm": 0.3493007332086563, "loss": 5.678823518753052, "time": 1.1406450033187867, "epoch": 275, "step": 171924}
{"lr": 3.759502237824497e-05, "data_time": 0.11688179969787597, "grad_norm": 0.34975431859493256, "loss": 5.721836805343628, "time": 1.1713880538940429, "epoch": 275, "step": 172024}
{"lr": 3.724845916044107e-05, "data_time": 0.09052770137786866, "grad_norm": 0.361782568693161, "loss": 5.77157506942749, "time": 1.124829316139221, "epoch": 275, "step": 172124}
{"lr": 3.681485003274876e-05, "data_time": 0.07937490940093994, "grad_norm": 0.4112180292606354, "loss": 5.786804819107056, "time": 1.1624911308288575, "epoch": 276, "step": 172250}
{"lr": 3.6473146059385115e-05, "data_time": 0.08872051239013672, "grad_norm": 0.3556621104478836, "loss": 5.7612651824951175, "time": 1.1299624919891358, "epoch": 276, "step": 172350}
{"lr": 3.613359386657703e-05, "data_time": 0.08428854942321777, "grad_norm": 0.37905153930187224, "loss": 5.736531829833984, "time": 1.1815142393112184, "epoch": 276, "step": 172450}
{"lr": 3.5796194471185196e-05, "data_time": 0.08278830051422119, "grad_norm": 0.34796366691589353, "loss": 5.754681015014649, "time": 1.141203808784485, "epoch": 276, "step": 172550}
{"lr": 3.546094888362368e-05, "data_time": 0.07799282073974609, "grad_norm": 0.3973605275154114, "loss": 5.683867645263672, "time": 1.1456838607788087, "epoch": 276, "step": 172650}
{"lr": 3.5127858107855895e-05, "data_time": 0.08390977382659912, "grad_norm": 0.3738328218460083, "loss": 5.749679613113403, "time": 1.1688827991485595, "epoch": 276, "step": 172750}
{"lr": 3.4711233294116224e-05, "data_time": 0.09505445957183838, "grad_norm": 0.3378933995962143, "loss": 5.713853645324707, "time": 1.171869945526123, "epoch": 277, "step": 172876}
{"lr": 3.4383016057643916e-05, "data_time": 0.09397478103637695, "grad_norm": 0.4289078265428543, "loss": 5.733654165267945, "time": 1.1653957605361938, "epoch": 277, "step": 172976}
{"lr": 3.405695686106322e-05, "data_time": 0.08940305709838867, "grad_norm": 0.357159161567688, "loss": 5.7318253993988035, "time": 1.1299182176589966, "epoch": 277, "step": 173076}
{"lr": 3.373305668082746e-05, "data_time": 0.08521087169647217, "grad_norm": 0.38783019185066225, "loss": 5.769573211669922, "time": 1.1344634294509888, "epoch": 277, "step": 173176}
{"lr": 3.341131648692391e-05, "data_time": 0.10475921630859375, "grad_norm": 0.34335191249847413, "loss": 5.7564620018005375, "time": 1.150757360458374, "epoch": 277, "step": 173276}
{"lr": 3.3091737242872035e-05, "data_time": 0.08098952770233155, "grad_norm": 0.36520259380340575, "loss": 5.673382520675659, "time": 1.1117955923080445, "epoch": 277, "step": 173376}
{"lr": 3.2692145635854015e-05, "data_time": 0.08221163749694824, "grad_norm": 0.3495479702949524, "loss": 5.726906442642212, "time": 1.16918625831604, "epoch": 278, "step": 173502}
{"lr": 3.237745365402094e-05, "data_time": 0.06614186763763427, "grad_norm": 0.3898126006126404, "loss": 5.719159269332886, "time": 1.1137770175933839, "epoch": 278, "step": 173602}
{"lr": 3.206492571815999e-05, "data_time": 0.08013935089111328, "grad_norm": 0.4087958484888077, "loss": 5.795549201965332, "time": 1.1143391370773315, "epoch": 278, "step": 173702}
{"lr": 3.175456276420235e-05, "data_time": 0.07898597717285157, "grad_norm": 0.39901227951049806, "loss": 5.7502206325531, "time": 1.1170063495635987, "epoch": 278, "step": 173802}
{"lr": 3.144636572159545e-05, "data_time": 0.08408975601196289, "grad_norm": 0.3915440261363983, "loss": 5.751674365997315, "time": 1.103913164138794, "epoch": 278, "step": 173902}
{"lr": 3.1140335513300245e-05, "data_time": 0.06953706741333007, "grad_norm": 0.3633316904306412, "loss": 5.723173952102661, "time": 27.944438695907593, "epoch": 278, "step": 174002}
{"lr": 3.0757824006850956e-05, "data_time": 0.10294294357299805, "grad_norm": 0.3459452480077744, "loss": 5.788645315170288, "time": 1.1862873077392577, "epoch": 279, "step": 174128}
{"lr": 3.0456694210156348e-05, "data_time": 0.09085869789123535, "grad_norm": 0.3848790168762207, "loss": 5.71556978225708, "time": 1.1416098117828368, "epoch": 279, "step": 174228}
{"lr": 3.015773421155429e-05, "data_time": 0.10226953029632568, "grad_norm": 0.3278244435787201, "loss": 5.733898448944092, "time": 1.13945574760437, "epoch": 279, "step": 174328}
{"lr": 2.9860944906343567e-05, "data_time": 0.07783386707305909, "grad_norm": 0.39774649441242216, "loss": 5.7423145294189455, "time": 1.1247852563858032, "epoch": 279, "step": 174428}
{"lr": 2.9566327183322476e-05, "data_time": 0.08940417766571045, "grad_norm": 0.3398143768310547, "loss": 5.7689735889434814, "time": 1.1440364837646484, "epoch": 279, "step": 174528}
{"lr": 2.9273881924786004e-05, "data_time": 0.08406741619110107, "grad_norm": 0.43730249106884, "loss": 5.780057573318482, "time": 1.1272007703781128, "epoch": 279, "step": 174628}
{"lr": 2.890849540832647e-05, "data_time": 0.06992738246917725, "grad_norm": 0.33491242825984957, "loss": 5.74701886177063, "time": 1.1867772817611695, "epoch": 280, "step": 174754}
{"lr": 2.862096313568728e-05, "data_time": 0.10053274631500245, "grad_norm": 0.372708386182785, "loss": 5.749267435073852, "time": 1.1480721712112427, "epoch": 280, "step": 174854}
{"lr": 2.8335606158625388e-05, "data_time": 0.07587568759918213, "grad_norm": 0.37349733114242556, "loss": 5.782386827468872, "time": 1.152611494064331, "epoch": 280, "step": 174954}
{"lr": 2.8052425331702437e-05, "data_time": 0.09386954307556153, "grad_norm": 0.4681560605764389, "loss": 5.7597190856933596, "time": 1.127829337120056, "epoch": 280, "step": 175054}
{"lr": 2.7771421502963245e-05, "data_time": 0.08512942790985108, "grad_norm": 0.38028375804424286, "loss": 5.709570360183716, "time": 1.1673262596130372, "epoch": 280, "step": 175154}
{"lr": 2.7492595513933264e-05, "data_time": 0.07964572906494141, "grad_norm": 0.36355302631855013, "loss": 5.741687297821045, "time": 1.1128944873809814, "epoch": 280, "step": 175254}
{"lr": 2.7144376867191347e-05, "data_time": 0.09224658012390137, "grad_norm": 0.37851524651050567, "loss": 5.744727945327758, "time": 1.1731966733932495, "epoch": 281, "step": 175380}
{"lr": 2.6870475861794554e-05, "data_time": 0.09288222789764404, "grad_norm": 0.3642707198858261, "loss": 5.684340143203736, "time": 1.1394716024398803, "epoch": 281, "step": 175480}
{"lr": 2.6598755394179186e-05, "data_time": 0.09829959869384766, "grad_norm": 0.399265718460083, "loss": 5.727534198760987, "time": 1.1608144521713257, "epoch": 281, "step": 175580}
{"lr": 2.632921627806955e-05, "data_time": 0.09621937274932861, "grad_norm": 0.3529059052467346, "loss": 5.727055788040161, "time": 1.1334298610687257, "epoch": 281, "step": 175680}
{"lr": 2.6061859320657587e-05, "data_time": 0.0784269094467163, "grad_norm": 0.34868076741695403, "loss": 5.736284399032593, "time": 1.1356116771697997, "epoch": 281, "step": 175780}
{"lr": 2.5796685322600178e-05, "data_time": 0.09264962673187256, "grad_norm": 0.35503140091896057, "loss": 5.708736562728882, "time": 1.119950795173645, "epoch": 281, "step": 175880}
{"lr": 2.5465675410577667e-05, "data_time": 0.0739469051361084, "grad_norm": 0.3843166470527649, "loss": 5.723715686798096, "time": 1.11658718585968, "epoch": 282, "step": 176006}
{"lr": 2.5205437815920745e-05, "data_time": 0.08284649848937989, "grad_norm": 0.3478920102119446, "loss": 5.736210727691651, "time": 1.109895396232605, "epoch": 282, "step": 176106}
{"lr": 2.4947385745353462e-05, "data_time": 0.06919338703155517, "grad_norm": 0.3557295948266983, "loss": 5.742731142044067, "time": 1.1013916969299316, "epoch": 282, "step": 176206}
{"lr": 2.4691519971667467e-05, "data_time": 0.08329012393951415, "grad_norm": 0.3426623433828354, "loss": 5.764427042007446, "time": 1.1199353456497192, "epoch": 282, "step": 176306}
{"lr": 2.4437841261106746e-05, "data_time": 0.0763540267944336, "grad_norm": 0.3491647392511368, "loss": 5.723922824859619, "time": 1.1205650568008423, "epoch": 282, "step": 176406}
{"lr": 2.4186350373366058e-05, "data_time": 0.09092366695404053, "grad_norm": 0.38010830283164976, "loss": 5.72817645072937, "time": 1.1277915716171265, "epoch": 282, "step": 176506}
{"lr": 2.3872588041543845e-05, "data_time": 0.10606520175933838, "grad_norm": 0.35520262718200685, "loss": 5.76618070602417, "time": 1.2071519136428832, "epoch": 283, "step": 176632}
{"lr": 2.36260443976622e-05, "data_time": 0.09825346469879151, "grad_norm": 0.36581000685691833, "loss": 5.733280801773072, "time": 1.1680604696273804, "epoch": 283, "step": 176732}
{"lr": 2.3381691007697766e-05, "data_time": 0.11197667121887207, "grad_norm": 0.3400140106678009, "loss": 5.763313055038452, "time": 1.1764022588729859, "epoch": 283, "step": 176832}
{"lr": 2.313952860341846e-05, "data_time": 0.10669190883636474, "grad_norm": 0.3616050392389297, "loss": 5.791895723342895, "time": 1.183747410774231, "epoch": 283, "step": 176932}
{"lr": 2.2899557910030805e-05, "data_time": 0.11025850772857666, "grad_norm": 0.45299550890922546, "loss": 5.738042831420898, "time": 1.1963876008987426, "epoch": 283, "step": 177032}
{"lr": 2.26617796461777e-05, "data_time": 0.08554096221923828, "grad_norm": 0.35658420622348785, "loss": 5.700463342666626, "time": 1.1453574419021606, "epoch": 283, "step": 177132}
{"lr": 2.2365301715956845e-05, "data_time": 0.07017679214477539, "grad_norm": 0.3502520024776459, "loss": 5.737922859191895, "time": 1.160937762260437, "epoch": 284, "step": 177258}
{"lr": 2.2132480955840222e-05, "data_time": 0.0875617504119873, "grad_norm": 0.36717180609703065, "loss": 5.7531774044036865, "time": 1.119579029083252, "epoch": 284, "step": 177358}
{"lr": 2.1901854922432373e-05, "data_time": 0.07134637832641602, "grad_norm": 0.3550779461860657, "loss": 5.723111248016357, "time": 1.131381106376648, "epoch": 284, "step": 177458}
{"lr": 2.1673424306391787e-05, "data_time": 0.08560900688171387, "grad_norm": 0.3904575526714325, "loss": 5.738759851455688, "time": 1.1418194770812988, "epoch": 284, "step": 177558}
{"lr": 2.1447189791801825e-05, "data_time": 0.07208034992218018, "grad_norm": 0.3859094977378845, "loss": 5.752407360076904, "time": 1.423690915107727, "epoch": 284, "step": 177658}
{"lr": 2.122315205616998e-05, "data_time": 0.07813076972961426, "grad_norm": 0.3380678027868271, "loss": 5.76754732131958, "time": 1.1467663526535035, "epoch": 284, "step": 177758}
{"lr": 2.094399332054989e-05, "data_time": 0.08605329990386963, "grad_norm": 0.3809026062488556, "loss": 5.7567973136901855, "time": 1.1633685111999512, "epoch": 285, "step": 177884}
{"lr": 2.0724922766747e-05, "data_time": 0.07752387523651123, "grad_norm": 0.3634903788566589, "loss": 5.709422492980957, "time": 1.1119197130203247, "epoch": 285, "step": 177984}
{"lr": 2.0508051154883238e-05, "data_time": 0.07278683185577392, "grad_norm": 0.34878631234169005, "loss": 5.726205635070801, "time": 1.1419503927230834, "epoch": 285, "step": 178084}
{"lr": 2.029337913442668e-05, "data_time": 0.08393232822418213, "grad_norm": 0.3790946274995804, "loss": 5.720271587371826, "time": 1.119580912590027, "epoch": 285, "step": 178184}
{"lr": 2.0080907348258123e-05, "data_time": 0.07714900970458985, "grad_norm": 0.373299652338028, "loss": 5.716838026046753, "time": 1.1414368629455567, "epoch": 285, "step": 178284}
{"lr": 1.9870636432668852e-05, "data_time": 0.09466454982757569, "grad_norm": 0.36838139593601227, "loss": 5.717089748382568, "time": 1.1342641353607177, "epoch": 285, "step": 178384}
{"lr": 1.960882965216515e-05, "data_time": 0.0782268762588501, "grad_norm": 0.45961306393146517, "loss": 5.763252258300781, "time": 1.185261869430542, "epoch": 286, "step": 178510}
{"lr": 1.9403535013576648e-05, "data_time": 0.08000171184539795, "grad_norm": 0.37470221519470215, "loss": 5.77378282546997, "time": 1.1586987018585204, "epoch": 286, "step": 178610}
{"lr": 1.9200443274102336e-05, "data_time": 0.08578870296478272, "grad_norm": 0.36295404136180875, "loss": 5.69330325126648, "time": 1.1481769561767579, "epoch": 286, "step": 178710}
{"lr": 1.899955504194334e-05, "data_time": 0.07857670783996581, "grad_norm": 0.3315127998590469, "loss": 5.750452995300293, "time": 1.1321211576461792, "epoch": 286, "step": 178810}
{"lr": 1.8800870918701932e-05, "data_time": 0.10159926414489746, "grad_norm": 0.3533594995737076, "loss": 5.7284423828125, "time": 1.1431978225708008, "epoch": 286, "step": 178910}
{"lr": 1.8604391499379824e-05, "data_time": 0.06418087482452392, "grad_norm": 0.34185250103473663, "loss": 5.754705762863159, "time": 1.1330010652542115, "epoch": 286, "step": 179010}
{"lr": 1.8359967398178788e-05, "data_time": 0.08619177341461182, "grad_norm": 0.38208375573158265, "loss": 5.724708843231201, "time": 1.1131033897399902, "epoch": 287, "step": 179136}
{"lr": 1.8168472767040813e-05, "data_time": 0.09453809261322021, "grad_norm": 0.34674530327320097, "loss": 5.745168018341064, "time": 50.483151006698606, "epoch": 287, "step": 179236}
{"lr": 1.7979184733671884e-05, "data_time": 0.07580382823944092, "grad_norm": 0.324243551492691, "loss": 5.739998626708984, "time": 1.1744712352752686, "epoch": 287, "step": 179336}
{"lr": 1.779210386493523e-05, "data_time": 0.08486957550048828, "grad_norm": 0.39227524399757385, "loss": 5.772956085205078, "time": 1.1184715509414673, "epoch": 287, "step": 179436}
{"lr": 1.7607230721084025e-05, "data_time": 0.08212482929229736, "grad_norm": 0.36918599605560304, "loss": 5.728632307052612, "time": 1.1264405488967895, "epoch": 287, "step": 179536}
{"lr": 1.7424565855759806e-05, "data_time": 0.08330390453338624, "grad_norm": 0.335470575094223, "loss": 5.769360160827636, "time": 1.111212682723999, "epoch": 287, "step": 179636}
{"lr": 1.719755311811363e-05, "data_time": 0.08853647708892823, "grad_norm": 0.32245343923568726, "loss": 5.732523012161255, "time": 1.1765613794326781, "epoch": 288, "step": 179762}
{"lr": 1.7019880967169986e-05, "data_time": 0.08012909889221191, "grad_norm": 0.35586377084255216, "loss": 5.721513557434082, "time": 1.1552485942840576, "epoch": 288, "step": 179862}
{"lr": 1.684441885369656e-05, "data_time": 0.08625462055206298, "grad_norm": 0.3953152418136597, "loss": 5.769218492507934, "time": 1.1641263246536255, "epoch": 288, "step": 179962}
{"lr": 1.6671167303151625e-05, "data_time": 0.08983397483825684, "grad_norm": 0.3510527491569519, "loss": 5.740756034851074, "time": 1.1238786458969117, "epoch": 288, "step": 180062}
{"lr": 1.6500126834373633e-05, "data_time": 0.08557097911834717, "grad_norm": 0.3592973738908768, "loss": 5.737707853317261, "time": 1.1501529216766357, "epoch": 288, "step": 180162}
{"lr": 1.6331297959579356e-05, "data_time": 0.10216846466064453, "grad_norm": 0.35873481929302214, "loss": 5.781558465957642, "time": 1.1272095441818237, "epoch": 288, "step": 180262}
{"lr": 1.6121723226439074e-05, "data_time": 0.09364545345306396, "grad_norm": 0.3594337493181229, "loss": 5.771740198135376, "time": 1.192115306854248, "epoch": 289, "step": 180388}
{"lr": 1.5957894406304648e-05, "data_time": 0.09117169380187988, "grad_norm": 0.35308792293071745, "loss": 5.720946550369263, "time": 1.1360401868820191, "epoch": 289, "step": 180488}
{"lr": 1.57962788039836e-05, "data_time": 0.08356258869171143, "grad_norm": 0.3518590360879898, "loss": 5.728117513656616, "time": 1.2199239492416383, "epoch": 289, "step": 180588}
{"lr": 1.5636876903468036e-05, "data_time": 0.08972904682159424, "grad_norm": 0.3665816247463226, "loss": 5.730434417724609, "time": 1.380129909515381, "epoch": 289, "step": 180688}
{"lr": 1.5479689182120574e-05, "data_time": 0.09806547164916993, "grad_norm": NaN, "loss": 5.717375802993774, "time": 1.1554871320724487, "epoch": 289, "step": 180788}
{"lr": 1.532471611067316e-05, "data_time": 0.08331511020660401, "grad_norm": 0.4019266992807388, "loss": 5.747746324539184, "time": 1.13038911819458, "epoch": 289, "step": 180888}
{"lr": 1.5132603976563028e-05, "data_time": 0.08924210071563721, "grad_norm": 0.36824420988559725, "loss": 5.746180200576783, "time": 1.1711061239242553, "epoch": 290, "step": 181014}
{"lr": 1.4982637713276212e-05, "data_time": 0.08111302852630616, "grad_norm": 0.3516339331865311, "loss": 5.816068077087403, "time": 1.1529174089431762, "epoch": 290, "step": 181114}
{"lr": 1.483488758841507e-05, "data_time": 0.08932325839996338, "grad_norm": 0.3467527687549591, "loss": 5.732984018325806, "time": 1.361546516418457, "epoch": 290, "step": 181214}
{"lr": 1.4689354044448555e-05, "data_time": 0.07578303813934326, "grad_norm": 0.3501555860042572, "loss": 5.707541418075562, "time": 1.1493781566619874, "epoch": 290, "step": 181314}
{"lr": 1.4546037517207808e-05, "data_time": 0.09770612716674805, "grad_norm": 0.3517675012350082, "loss": 5.74357042312622, "time": 1.1654887437820434, "epoch": 290, "step": 181414}
{"lr": 1.4404938435884319e-05, "data_time": 0.07477302551269531, "grad_norm": 0.43846403658390043, "loss": 5.741662073135376, "time": 1.155333113670349, "epoch": 290, "step": 181514}
{"lr": 1.4230311446015073e-05, "data_time": 0.11838400363922119, "grad_norm": 0.36369662880897524, "loss": 5.728892993927002, "time": 1.193715786933899, "epoch": 291, "step": 181640}
{"lr": 1.4094225338781662e-05, "data_time": 0.08333888053894042, "grad_norm": 0.3586647152900696, "loss": 5.7408699035644535, "time": 1.1264527559280395, "epoch": 291, "step": 181740}
{"lr": 1.3960358030512649e-05, "data_time": 0.09881088733673096, "grad_norm": 0.3482006251811981, "loss": 5.746242380142212, "time": 1.16133394241333, "epoch": 291, "step": 181840}
{"lr": 1.3828709922102052e-05, "data_time": 0.09509220123291015, "grad_norm": 0.38824011087417604, "loss": 5.731892776489258, "time": 1.1258938789367676, "epoch": 291, "step": 181940}
{"lr": 1.3699281407797757e-05, "data_time": 0.09945785999298096, "grad_norm": 0.3673950761556625, "loss": 5.6982027053833, "time": 1.1714969635009767, "epoch": 291, "step": 182040}
{"lr": 1.3572072875200948e-05, "data_time": 0.0995819091796875, "grad_norm": 0.37309953570365906, "loss": 5.713322973251342, "time": 1.153254222869873, "epoch": 291, "step": 182140}
{"lr": 1.3414951522823994e-05, "data_time": 0.07070386409759521, "grad_norm": 0.36824172735214233, "loss": 5.7761465549469, "time": 1.1616283416748048, "epoch": 292, "step": 182266}
{"lr": 1.3292761541952224e-05, "data_time": 0.0855605125427246, "grad_norm": 0.39960812628269193, "loss": 5.7381116390228275, "time": 1.138487696647644, "epoch": 292, "step": 182366}
{"lr": 1.3172792760197462e-05, "data_time": 0.0691408395767212, "grad_norm": 0.34519781470298766, "loss": 5.736845922470093, "time": 1.1670047521591187, "epoch": 292, "step": 182466}
{"lr": 1.3055045536831856e-05, "data_time": 0.09923205375671387, "grad_norm": 0.33126620948314667, "loss": 5.743561697006226, "time": 1.1411662340164184, "epoch": 292, "step": 182566}
{"lr": 1.2939520224474194e-05, "data_time": 0.07324609756469727, "grad_norm": 0.3471830993890762, "loss": 5.760906457901001, "time": 1.1569430351257324, "epoch": 292, "step": 182666}
{"lr": 1.28262171690893e-05, "data_time": 0.0864400863647461, "grad_norm": 0.3605998933315277, "loss": 5.780074834823608, "time": 1.1328535318374633, "epoch": 292, "step": 182766}
{"lr": 1.2686619893092226e-05, "data_time": 0.11604804992675781, "grad_norm": 0.3527462244033813, "loss": 5.770840692520141, "time": 1.1657098293304444, "epoch": 293, "step": 182892}
{"lr": 1.257834037811793e-05, "data_time": 0.09477348327636718, "grad_norm": 0.42038992047309875, "loss": 5.727756547927856, "time": 1.1136072397232055, "epoch": 293, "step": 182992}
{"lr": 1.2472284201746042e-05, "data_time": 0.10300617218017578, "grad_norm": 0.3514576435089111, "loss": 5.7104919910430905, "time": 1.1170990467071533, "epoch": 293, "step": 183092}
{"lr": 1.2368451681584117e-05, "data_time": 0.0859731912612915, "grad_norm": 0.3446946918964386, "loss": 5.70970664024353, "time": 1.115463948249817, "epoch": 293, "step": 183192}
{"lr": 1.2266843128580503e-05, "data_time": 0.107726788520813, "grad_norm": 0.3464801996946335, "loss": 5.757800817489624, "time": 1.1726274967193604, "epoch": 293, "step": 183292}
{"lr": 1.2167458847023642e-05, "data_time": 0.09120612144470215, "grad_norm": 0.341851481795311, "loss": 5.747572326660157, "time": 6.131296324729919, "epoch": 293, "step": 183392}
{"lr": 1.2045402029765499e-05, "data_time": 0.07461388111114502, "grad_norm": 0.3571335643529892, "loss": 5.699615859985352, "time": 1.2364864110946656, "epoch": 294, "step": 183518}
{"lr": 1.1951045687769992e-05, "data_time": 0.06855006217956543, "grad_norm": 0.37352963387966154, "loss": 5.738583517074585, "time": 1.2339762926101685, "epoch": 294, "step": 183618}
{"lr": 1.185891456294337e-05, "data_time": 0.0909541368484497, "grad_norm": 0.37067265808582306, "loss": 5.738201904296875, "time": 1.2893973350524903, "epoch": 294, "step": 183718}
{"lr": 1.1769008931191722e-05, "data_time": 0.06548185348510742, "grad_norm": 0.3460902124643326, "loss": 5.731589651107788, "time": 1.2435533046722411, "epoch": 294, "step": 183818}
{"lr": 1.1681329061756564e-05, "data_time": 0.07662231922149658, "grad_norm": 0.40349251627922056, "loss": 5.716646194458008, "time": 1.29113712310791, "epoch": 294, "step": 183918}
{"lr": 1.1595875217213632e-05, "data_time": 0.06635792255401611, "grad_norm": 0.35012447237968447, "loss": 5.738239574432373, "time": 1.172874879837036, "epoch": 294, "step": 184018}
{"lr": 1.1491373182603211e-05, "data_time": 0.07300357818603516, "grad_norm": 0.3404331624507904, "loss": 5.74043607711792, "time": 1.1996308326721192, "epoch": 295, "step": 184144}
{"lr": 1.141095108672155e-05, "data_time": 0.07874691486358643, "grad_norm": 0.36229685246944426, "loss": 5.744084358215332, "time": 1.1345310449600219, "epoch": 295, "step": 184244}
{"lr": 1.133275582543613e-05, "data_time": 0.07248680591583252, "grad_norm": 0.3975618153810501, "loss": 5.697024202346801, "time": 1.136539053916931, "epoch": 295, "step": 184344}
{"lr": 1.125678763291935e-05, "data_time": 0.07657759189605713, "grad_norm": 0.35753586888313293, "loss": 5.702203941345215, "time": 1.155711841583252, "epoch": 295, "step": 184444}
{"lr": 1.1183046736673963e-05, "data_time": 0.07210593223571778, "grad_norm": 0.3655414402484894, "loss": 5.710383415222168, "time": 1.2923620700836183, "epoch": 295, "step": 184544}
{"lr": 1.1111533357532713e-05, "data_time": 0.0899082899093628, "grad_norm": 0.34987816512584685, "loss": 5.745873546600341, "time": 1.132818913459778, "epoch": 295, "step": 184644}
{"lr": 1.1024598369347037e-05, "data_time": 0.06493699550628662, "grad_norm": 0.3579767793416977, "loss": 5.730368566513062, "time": 1.1730557680130005, "epoch": 296, "step": 184770}
{"lr": 1.0958119957468723e-05, "data_time": 0.08810515403747558, "grad_norm": 0.3707918643951416, "loss": 5.746795177459717, "time": 1.1332197904586792, "epoch": 296, "step": 184870}
{"lr": 1.08938697362852e-05, "data_time": 0.07225875854492188, "grad_norm": 0.351994651556015, "loss": 5.738555765151977, "time": 1.163670301437378, "epoch": 296, "step": 184970}
{"lr": 1.0831847898207502e-05, "data_time": 0.09305434226989746, "grad_norm": 0.3349885165691376, "loss": 5.725433349609375, "time": 1.137357497215271, "epoch": 296, "step": 185070}
{"lr": 1.0772054628972993e-05, "data_time": 0.06398954391479492, "grad_norm": 0.3767035245895386, "loss": 5.7169270515441895, "time": 1.144608473777771, "epoch": 296, "step": 185170}
{"lr": 1.0714490107645383e-05, "data_time": 0.08391771316528321, "grad_norm": 0.35894864797592163, "loss": 5.7857341289520265, "time": 1.143824863433838, "epoch": 296, "step": 185270}
{"lr": 1.064513236809082e-05, "data_time": 0.0914414644241333, "grad_norm": 0.34577714204788207, "loss": 5.7694604873657225, "time": 1.1927481651306153, "epoch": 297, "step": 185396}
{"lr": 1.0592605441752245e-05, "data_time": 0.09515435695648193, "grad_norm": 0.3496968448162079, "loss": 5.779800224304199, "time": 1.1511216163635254, "epoch": 297, "step": 185496}
{"lr": 1.0542307800718937e-05, "data_time": 0.10777373313903808, "grad_norm": 0.3466358929872513, "loss": 5.722778272628784, "time": 1.1679192543029786, "epoch": 297, "step": 185596}
{"lr": 1.049423959561802e-05, "data_time": 0.08439290523529053, "grad_norm": 0.3900257468223572, "loss": 5.758104991912842, "time": 1.1411136627197265, "epoch": 297, "step": 185696}
{"lr": 1.0448400970399704e-05, "data_time": 0.10359351634979248, "grad_norm": 0.3504709661006927, "loss": 5.6951078414917, "time": 1.1543539762496948, "epoch": 297, "step": 185796}
{"lr": 1.0404792062337553e-05, "data_time": 0.0844303846359253, "grad_norm": 0.3644426614046097, "loss": 5.726172161102295, "time": 1.1463996410369872, "epoch": 297, "step": 185896}
{"lr": 1.0353019710852329e-05, "data_time": 0.07085556983947754, "grad_norm": 0.3780123323202133, "loss": 5.750947570800781, "time": 1.1542398929595947, "epoch": 298, "step": 186022}
{"lr": 1.0314450434321159e-05, "data_time": 0.07542483806610108, "grad_norm": 0.3634962201118469, "loss": 5.757241678237915, "time": 1.159548807144165, "epoch": 298, "step": 186122}
{"lr": 1.0278111276089231e-05, "data_time": 0.06771912574768066, "grad_norm": 0.3497780919075012, "loss": 5.768102502822876, "time": 1.1621356725692749, "epoch": 298, "step": 186222}
{"lr": 1.024400234498204e-05, "data_time": 0.06749849319458008, "grad_norm": 0.3600718915462494, "loss": 5.731571197509766, "time": 1.13809335231781, "epoch": 298, "step": 186322}
{"lr": 1.0212123743145855e-05, "data_time": 0.07271761894226074, "grad_norm": 0.35358281135559083, "loss": 5.741707706451416, "time": 1.1765593528747558, "epoch": 298, "step": 186422}
{"lr": 1.0182475566047918e-05, "data_time": 0.0695796251296997, "grad_norm": 0.34786203801631926, "loss": 5.750644159317017, "time": 1.1262090921401977, "epoch": 298, "step": 186522}
{"lr": 1.0148294678347008e-05, "data_time": 0.09078547954559327, "grad_norm": 0.3564026474952698, "loss": 5.699419260025024, "time": 1.1318748235702514, "epoch": 299, "step": 186648}
{"lr": 1.012368757789886e-05, "data_time": 0.08145513534545898, "grad_norm": 0.3392304599285126, "loss": 5.731689739227295, "time": 1.1224574804306031, "epoch": 299, "step": 186748}
{"lr": 1.010131116702973e-05, "data_time": 0.089778733253479, "grad_norm": 0.3475814461708069, "loss": 5.73746919631958, "time": 1.1155065298080444, "epoch": 299, "step": 186848}
{"lr": 1.0081165512750512e-05, "data_time": 0.09606294631958008, "grad_norm": 0.3627645939588547, "loss": 5.736703729629516, "time": 1.1516608953475953, "epoch": 299, "step": 186948}
{"lr": 1.0063250675391713e-05, "data_time": 0.1030566930770874, "grad_norm": 0.33437620401382445, "loss": 5.759677314758301, "time": 1.1684059619903564, "epoch": 299, "step": 187048}
{"lr": 1.0047566708602977e-05, "data_time": 0.09172620773315429, "grad_norm": 0.40497554540634156, "loss": 5.7584281921386715, "time": 1.1280571222305298, "epoch": 299, "step": 187148}
{"lr": 1.0030981295965218e-05, "data_time": 0.06776392459869385, "grad_norm": 0.3838208556175232, "loss": 5.732621574401856, "time": 1.2144843101501466, "epoch": 300, "step": 187274}
{"lr": 1.0020339259352401e-05, "data_time": 0.09993472099304199, "grad_norm": 0.38278211653232574, "loss": 5.739838266372681, "time": 1.1694608211517334, "epoch": 300, "step": 187374}
{"lr": 1.0011928221817039e-05, "data_time": 0.07925164699554443, "grad_norm": 0.3483845978975296, "loss": 5.741621589660644, "time": 1.1564635276794433, "epoch": 300, "step": 187474}
{"lr": 1.0005748208547712e-05, "data_time": 0.09263648986816406, "grad_norm": 0.34128336906433104, "loss": 5.728607845306397, "time": 1.1496654272079467, "epoch": 300, "step": 187574}
{"lr": 1.0001799238051798e-05, "data_time": 0.0734743595123291, "grad_norm": 0.3867007553577423, "loss": 5.706835460662842, "time": 1.1449659585952758, "epoch": 300, "step": 187674}
{"lr": 1.0000081322155401e-05, "data_time": 0.0791393756866455, "grad_norm": 0.350445294380188, "loss": 5.6970374584198, "time": 1.1339036703109742, "epoch": 300, "step": 187774}
