diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,8711 @@ +{ + "global_steps": 27050, + "loss": { + "train": [ + { + "step": 50, + "logs": { + "loss": 1.2091, + "epoch": 0.018484288354898338, + "learning_rate": 1.0584432680980562e-05 + } + }, + { + "step": 100, + "logs": { + "loss": 1.1315, + "epoch": 0.036968576709796676, + "learning_rate": 1.2459822860720352e-05 + } + }, + { + "step": 150, + "logs": { + "loss": 1.1118, + "epoch": 0.05545286506469501, + "learning_rate": 1.355685579689696e-05 + } + }, + { + "step": 200, + "logs": { + "loss": 1.1091, + "epoch": 0.07393715341959335, + "learning_rate": 1.4335213053988223e-05 + } + }, + { + "step": 250, + "logs": { + "loss": 1.1063, + "epoch": 0.09242144177449169, + "learning_rate": 1.4938953847726847e-05 + } + }, + { + "step": 300, + "logs": { + "loss": 1.1012, + "epoch": 0.11090573012939002, + "learning_rate": 1.5432245994674192e-05 + } + }, + { + "step": 350, + "logs": { + "loss": 1.0844, + "epoch": 0.12939001848428835, + "learning_rate": 1.5849318562425815e-05 + } + }, + { + "step": 400, + "logs": { + "loss": 1.0911, + "epoch": 0.1478743068391867, + "learning_rate": 1.6210603254020138e-05 + } + }, + { + "step": 450, + "logs": { + "loss": 1.0816, + "epoch": 0.16635859519408502, + "learning_rate": 1.652927893686328e-05 + } + }, + { + "step": 500, + "logs": { + "loss": 1.0819, + "epoch": 0.18484288354898337, + "learning_rate": 1.6814344049111566e-05 + } + }, + { + "step": 550, + "logs": { + "loss": 1.0757, + "epoch": 0.2033271719038817, + "learning_rate": 1.70722168104602e-05 + } + }, + { + "step": 600, + "logs": { + "loss": 1.0706, + "epoch": 0.22181146025878004, + "learning_rate": 1.7307636196960783e-05 + } + }, + { + "step": 650, + "logs": { + "loss": 1.0698, + "epoch": 0.24029574861367836, + "learning_rate": 1.7524201039271304e-05 + } + }, + { + "step": 700, + "logs": { + "loss": 1.0632, + "epoch": 0.2587800369685767, + "learning_rate": 1.7724708765356604e-05 + } + }, + { + "step": 750, + "logs": { + "loss": 1.0601, + "epoch": 0.27726432532347506, + "learning_rate": 1.7911376992503148e-05 + } + }, + { + "step": 800, + "logs": { + "loss": 1.0582, + "epoch": 0.2957486136783734, + "learning_rate": 1.808599345743407e-05 + } + }, + { + "step": 850, + "logs": { + "loss": 1.0512, + "epoch": 0.3142329020332717, + "learning_rate": 1.825002041317456e-05 + } + }, + { + "step": 900, + "logs": { + "loss": 1.0544, + "epoch": 0.33271719038817005, + "learning_rate": 1.840466914065299e-05 + } + }, + { + "step": 950, + "logs": { + "loss": 1.043, + "epoch": 0.3512014787430684, + "learning_rate": 1.855095428760744e-05 + } + }, + { + "step": 1000, + "logs": { + "loss": 1.047, + "epoch": 0.36968576709796674, + "learning_rate": 1.8689734253201903e-05 + } + }, + { + "step": 1050, + "logs": { + "loss": 1.0408, + "epoch": 0.38817005545286504, + "learning_rate": 1.8821741709263542e-05 + } + }, + { + "step": 1100, + "logs": { + "loss": 1.0431, + "epoch": 0.4066543438077634, + "learning_rate": 1.89476070147965e-05 + } + }, + { + "step": 1150, + "logs": { + "loss": 1.0431, + "epoch": 0.42513863216266173, + "learning_rate": 1.9067876421445012e-05 + } + }, + { + "step": 1200, + "logs": { + "loss": 1.034, + "epoch": 0.4436229205175601, + "learning_rate": 1.9183026401502057e-05 + } + }, + { + "step": 1250, + "logs": { + "loss": 1.039, + "epoch": 0.46210720887245843, + "learning_rate": 1.9293475049105016e-05 + } + }, + { + "step": 1300, + "logs": { + "loss": 1.0304, + "epoch": 0.4805914972273567, + "learning_rate": 1.9399591243986016e-05 + } + }, + { + "step": 1350, + "logs": { + "loss": 1.0385, + "epoch": 0.49907578558225507, + "learning_rate": 1.950170208484624e-05 + } + }, + { + "step": 1400, + "logs": { + "loss": 1.027, + "epoch": 0.5175600739371534, + "learning_rate": 1.9600098970219974e-05 + } + }, + { + "step": 1450, + "logs": { + "loss": 1.0281, + "epoch": 0.5360443622920518, + "learning_rate": 1.9695042611797287e-05 + } + }, + { + "step": 1500, + "logs": { + "loss": 1.0282, + "epoch": 0.5545286506469501, + "learning_rate": 1.978676719749536e-05 + } + }, + { + "step": 1550, + "logs": { + "loss": 1.0186, + "epoch": 0.5730129390018485, + "learning_rate": 1.9875483871657987e-05 + } + }, + { + "step": 1600, + "logs": { + "loss": 1.0237, + "epoch": 0.5914972273567468, + "learning_rate": 1.9961383662539012e-05 + } + }, + { + "step": 1650, + "logs": { + "loss": 1.0176, + "epoch": 0.609981515711645, + "learning_rate": 1.9978762732528416e-05 + } + }, + { + "step": 1700, + "logs": { + "loss": 1.4539, + "epoch": 0.6284658040665434, + "learning_rate": 1.9939434459432887e-05 + } + }, + { + "step": 1750, + "logs": { + "loss": 1.2593, + "epoch": 0.6469500924214417, + "learning_rate": 1.990010618633736e-05 + } + }, + { + "step": 1800, + "logs": { + "loss": 1.5937, + "epoch": 0.6654343807763401, + "learning_rate": 1.986077791324183e-05 + } + }, + { + "step": 1850, + "logs": { + "loss": 1.0165, + "epoch": 0.6839186691312384, + "learning_rate": 1.9821449640146302e-05 + } + }, + { + "step": 1900, + "logs": { + "loss": 1.4354, + "epoch": 0.7024029574861368, + "learning_rate": 1.9782121367050776e-05 + } + }, + { + "step": 1950, + "logs": { + "loss": 1.021, + "epoch": 0.7208872458410351, + "learning_rate": 1.9742793093955247e-05 + } + }, + { + "step": 2000, + "logs": { + "loss": 1.0078, + "epoch": 0.7393715341959335, + "learning_rate": 1.9703464820859717e-05 + } + }, + { + "step": 2050, + "logs": { + "loss": 1.0057, + "epoch": 0.7578558225508318, + "learning_rate": 1.966413654776419e-05 + } + }, + { + "step": 2100, + "logs": { + "loss": 1.0001, + "epoch": 0.7763401109057301, + "learning_rate": 1.9624808274668662e-05 + } + }, + { + "step": 2150, + "logs": { + "loss": 1.006, + "epoch": 0.7948243992606284, + "learning_rate": 1.9585480001573133e-05 + } + }, + { + "step": 2200, + "logs": { + "loss": 1.0103, + "epoch": 0.8133086876155268, + "learning_rate": 1.9546151728477603e-05 + } + }, + { + "step": 2250, + "logs": { + "loss": 1.0046, + "epoch": 0.8317929759704251, + "learning_rate": 1.9506823455382077e-05 + } + }, + { + "step": 2300, + "logs": { + "loss": 0.9941, + "epoch": 0.8502772643253235, + "learning_rate": 1.9467495182286548e-05 + } + }, + { + "step": 2350, + "logs": { + "loss": 0.9963, + "epoch": 0.8687615526802218, + "learning_rate": 1.942816690919102e-05 + } + }, + { + "step": 2400, + "logs": { + "loss": 0.9926, + "epoch": 0.8872458410351202, + "learning_rate": 1.938883863609549e-05 + } + }, + { + "step": 2450, + "logs": { + "loss": 0.9918, + "epoch": 0.9057301293900185, + "learning_rate": 1.934951036299996e-05 + } + }, + { + "step": 2500, + "logs": { + "loss": 0.9899, + "epoch": 0.9242144177449169, + "learning_rate": 1.9310182089904434e-05 + } + }, + { + "step": 2550, + "logs": { + "loss": 0.9887, + "epoch": 0.9426987060998152, + "learning_rate": 1.9270853816808904e-05 + } + }, + { + "step": 2600, + "logs": { + "loss": 0.9942, + "epoch": 0.9611829944547134, + "learning_rate": 1.9231525543713375e-05 + } + }, + { + "step": 2650, + "logs": { + "loss": 0.9817, + "epoch": 0.9796672828096118, + "learning_rate": 1.919219727061785e-05 + } + }, + { + "step": 2700, + "logs": { + "loss": 0.9811, + "epoch": 0.9981515711645101, + "learning_rate": 1.915286899752232e-05 + } + }, + { + "step": 2750, + "logs": { + "loss": 0.8466, + "epoch": 1.0166358595194085, + "learning_rate": 1.911354072442679e-05 + } + }, + { + "step": 2800, + "logs": { + "loss": 0.8316, + "epoch": 1.0351201478743068, + "learning_rate": 1.9074212451331264e-05 + } + }, + { + "step": 2850, + "logs": { + "loss": 0.8363, + "epoch": 1.0536044362292052, + "learning_rate": 1.9034884178235735e-05 + } + }, + { + "step": 2900, + "logs": { + "loss": 0.8378, + "epoch": 1.0720887245841035, + "learning_rate": 1.8995555905140205e-05 + } + }, + { + "step": 2950, + "logs": { + "loss": 0.8386, + "epoch": 1.0905730129390019, + "learning_rate": 1.895622763204468e-05 + } + }, + { + "step": 3000, + "logs": { + "loss": 0.8465, + "epoch": 1.1090573012939002, + "learning_rate": 1.891689935894915e-05 + } + }, + { + "step": 3050, + "logs": { + "loss": 0.8375, + "epoch": 1.1275415896487986, + "learning_rate": 1.887757108585362e-05 + } + }, + { + "step": 3100, + "logs": { + "loss": 0.8379, + "epoch": 1.146025878003697, + "learning_rate": 1.8838242812758095e-05 + } + }, + { + "step": 3150, + "logs": { + "loss": 0.836, + "epoch": 1.1645101663585953, + "learning_rate": 1.8798914539662565e-05 + } + }, + { + "step": 3200, + "logs": { + "loss": 0.8411, + "epoch": 1.1829944547134936, + "learning_rate": 1.8759586266567036e-05 + } + }, + { + "step": 3250, + "logs": { + "loss": 0.8442, + "epoch": 1.201478743068392, + "learning_rate": 1.872025799347151e-05 + } + }, + { + "step": 3300, + "logs": { + "loss": 0.8397, + "epoch": 1.21996303142329, + "learning_rate": 1.868092972037598e-05 + } + }, + { + "step": 3350, + "logs": { + "loss": 0.8415, + "epoch": 1.2384473197781884, + "learning_rate": 1.864160144728045e-05 + } + }, + { + "step": 3400, + "logs": { + "loss": 0.8397, + "epoch": 1.2569316081330868, + "learning_rate": 1.8602273174184925e-05 + } + }, + { + "step": 3450, + "logs": { + "loss": 0.8396, + "epoch": 1.2754158964879851, + "learning_rate": 1.8562944901089396e-05 + } + }, + { + "step": 3500, + "logs": { + "loss": 0.8452, + "epoch": 1.2939001848428835, + "learning_rate": 1.8523616627993866e-05 + } + }, + { + "step": 3550, + "logs": { + "loss": 0.8362, + "epoch": 1.3123844731977818, + "learning_rate": 1.848428835489834e-05 + } + }, + { + "step": 3600, + "logs": { + "loss": 0.8396, + "epoch": 1.3308687615526802, + "learning_rate": 1.844496008180281e-05 + } + }, + { + "step": 3650, + "logs": { + "loss": 0.8334, + "epoch": 1.3493530499075785, + "learning_rate": 1.840563180870728e-05 + } + }, + { + "step": 3700, + "logs": { + "loss": 0.834, + "epoch": 1.3678373382624769, + "learning_rate": 1.8366303535611752e-05 + } + }, + { + "step": 3750, + "logs": { + "loss": 0.8333, + "epoch": 1.3863216266173752, + "learning_rate": 1.8326975262516223e-05 + } + }, + { + "step": 3800, + "logs": { + "loss": 0.8357, + "epoch": 1.4048059149722736, + "learning_rate": 1.8287646989420697e-05 + } + }, + { + "step": 3850, + "logs": { + "loss": 0.8374, + "epoch": 1.423290203327172, + "learning_rate": 1.8248318716325167e-05 + } + }, + { + "step": 3900, + "logs": { + "loss": 0.837, + "epoch": 1.4417744916820703, + "learning_rate": 1.8208990443229638e-05 + } + }, + { + "step": 3950, + "logs": { + "loss": 0.835, + "epoch": 1.4602587800369686, + "learning_rate": 1.816966217013411e-05 + } + }, + { + "step": 4000, + "logs": { + "loss": 0.8352, + "epoch": 1.478743068391867, + "learning_rate": 1.8130333897038583e-05 + } + }, + { + "step": 4050, + "logs": { + "loss": 0.8475, + "epoch": 1.4972273567467653, + "learning_rate": 1.8091005623943053e-05 + } + }, + { + "step": 4100, + "logs": { + "loss": 0.8339, + "epoch": 1.5157116451016637, + "learning_rate": 1.8051677350847524e-05 + } + }, + { + "step": 4150, + "logs": { + "loss": 0.8382, + "epoch": 1.534195933456562, + "learning_rate": 1.8012349077751998e-05 + } + }, + { + "step": 4200, + "logs": { + "loss": 0.8376, + "epoch": 1.5526802218114601, + "learning_rate": 1.797302080465647e-05 + } + }, + { + "step": 4250, + "logs": { + "loss": 0.8302, + "epoch": 1.5711645101663585, + "learning_rate": 1.793369253156094e-05 + } + }, + { + "step": 4300, + "logs": { + "loss": 0.8312, + "epoch": 1.5896487985212568, + "learning_rate": 1.7894364258465413e-05 + } + }, + { + "step": 4350, + "logs": { + "loss": 0.8361, + "epoch": 1.6081330868761552, + "learning_rate": 1.7855035985369884e-05 + } + }, + { + "step": 4400, + "logs": { + "loss": 0.8321, + "epoch": 1.6266173752310535, + "learning_rate": 1.7815707712274354e-05 + } + }, + { + "step": 4450, + "logs": { + "loss": 0.8374, + "epoch": 1.645101663585952, + "learning_rate": 1.777637943917883e-05 + } + }, + { + "step": 4500, + "logs": { + "loss": 0.8287, + "epoch": 1.6635859519408502, + "learning_rate": 1.77370511660833e-05 + } + }, + { + "step": 4550, + "logs": { + "loss": 0.8367, + "epoch": 1.6820702402957486, + "learning_rate": 1.769772289298777e-05 + } + }, + { + "step": 4600, + "logs": { + "loss": 0.8439, + "epoch": 1.700554528650647, + "learning_rate": 1.7658394619892244e-05 + } + }, + { + "step": 4650, + "logs": { + "loss": 0.829, + "epoch": 1.7190388170055453, + "learning_rate": 1.7619066346796714e-05 + } + }, + { + "step": 4700, + "logs": { + "loss": 0.8325, + "epoch": 1.7375231053604436, + "learning_rate": 1.7579738073701185e-05 + } + }, + { + "step": 4750, + "logs": { + "loss": 0.837, + "epoch": 1.756007393715342, + "learning_rate": 1.754040980060566e-05 + } + }, + { + "step": 4800, + "logs": { + "loss": 0.8371, + "epoch": 1.7744916820702403, + "learning_rate": 1.750108152751013e-05 + } + }, + { + "step": 4850, + "logs": { + "loss": 0.832, + "epoch": 1.7929759704251387, + "learning_rate": 1.74617532544146e-05 + } + }, + { + "step": 4900, + "logs": { + "loss": 0.8342, + "epoch": 1.811460258780037, + "learning_rate": 1.7422424981319074e-05 + } + }, + { + "step": 4950, + "logs": { + "loss": 0.8331, + "epoch": 1.8299445471349354, + "learning_rate": 1.7383096708223545e-05 + } + }, + { + "step": 5000, + "logs": { + "loss": 0.8234, + "epoch": 1.8484288354898337, + "learning_rate": 1.7343768435128015e-05 + } + }, + { + "step": 5050, + "logs": { + "loss": 0.8311, + "epoch": 1.866913123844732, + "learning_rate": 1.7304440162032486e-05 + } + }, + { + "step": 5100, + "logs": { + "loss": 0.8259, + "epoch": 1.8853974121996304, + "learning_rate": 1.726511188893696e-05 + } + }, + { + "step": 5150, + "logs": { + "loss": 0.8253, + "epoch": 1.9038817005545285, + "learning_rate": 1.722578361584143e-05 + } + }, + { + "step": 5200, + "logs": { + "loss": 0.8297, + "epoch": 1.922365988909427, + "learning_rate": 1.71864553427459e-05 + } + }, + { + "step": 5250, + "logs": { + "loss": 0.8264, + "epoch": 1.9408502772643252, + "learning_rate": 1.7147127069650372e-05 + } + }, + { + "step": 5300, + "logs": { + "loss": 0.8283, + "epoch": 1.9593345656192236, + "learning_rate": 1.7107798796554846e-05 + } + }, + { + "step": 5350, + "logs": { + "loss": 0.8262, + "epoch": 1.977818853974122, + "learning_rate": 1.7068470523459316e-05 + } + }, + { + "step": 5400, + "logs": { + "loss": 0.829, + "epoch": 1.9963031423290203, + "learning_rate": 1.7029142250363787e-05 + } + }, + { + "step": 5450, + "logs": { + "loss": 0.6969, + "epoch": 2.014787430683919, + "learning_rate": 1.6989813977268258e-05 + } + }, + { + "step": 5500, + "logs": { + "loss": 0.6715, + "epoch": 2.033271719038817, + "learning_rate": 1.695048570417273e-05 + } + }, + { + "step": 5550, + "logs": { + "loss": 0.6636, + "epoch": 2.0517560073937156, + "learning_rate": 1.6911157431077202e-05 + } + }, + { + "step": 5600, + "logs": { + "loss": 0.6649, + "epoch": 2.0702402957486137, + "learning_rate": 1.6871829157981673e-05 + } + }, + { + "step": 5650, + "logs": { + "loss": 0.665, + "epoch": 2.088724584103512, + "learning_rate": 1.6832500884886147e-05 + } + }, + { + "step": 5700, + "logs": { + "loss": 0.6746, + "epoch": 2.1072088724584104, + "learning_rate": 1.6793172611790617e-05 + } + }, + { + "step": 5750, + "logs": { + "loss": 0.6678, + "epoch": 2.1256931608133085, + "learning_rate": 1.6753844338695088e-05 + } + }, + { + "step": 5800, + "logs": { + "loss": 0.6739, + "epoch": 2.144177449168207, + "learning_rate": 1.6714516065599562e-05 + } + }, + { + "step": 5850, + "logs": { + "loss": 0.6666, + "epoch": 2.162661737523105, + "learning_rate": 1.6675187792504033e-05 + } + }, + { + "step": 5900, + "logs": { + "loss": 0.674, + "epoch": 2.1811460258780038, + "learning_rate": 1.6635859519408503e-05 + } + }, + { + "step": 5950, + "logs": { + "loss": 0.6737, + "epoch": 2.199630314232902, + "learning_rate": 1.6596531246312977e-05 + } + }, + { + "step": 6000, + "logs": { + "loss": 0.675, + "epoch": 2.2181146025878005, + "learning_rate": 1.6557202973217448e-05 + } + }, + { + "step": 6050, + "logs": { + "loss": 0.6743, + "epoch": 2.2365988909426986, + "learning_rate": 1.651787470012192e-05 + } + }, + { + "step": 6100, + "logs": { + "loss": 0.6718, + "epoch": 2.255083179297597, + "learning_rate": 1.647854642702639e-05 + } + }, + { + "step": 6150, + "logs": { + "loss": 0.6735, + "epoch": 2.2735674676524953, + "learning_rate": 1.6439218153930863e-05 + } + }, + { + "step": 6200, + "logs": { + "loss": 0.6798, + "epoch": 2.292051756007394, + "learning_rate": 1.6399889880835334e-05 + } + }, + { + "step": 6250, + "logs": { + "loss": 0.6768, + "epoch": 2.310536044362292, + "learning_rate": 1.6360561607739804e-05 + } + }, + { + "step": 6300, + "logs": { + "loss": 0.6796, + "epoch": 2.3290203327171906, + "learning_rate": 1.632123333464428e-05 + } + }, + { + "step": 6350, + "logs": { + "loss": 0.6813, + "epoch": 2.3475046210720887, + "learning_rate": 1.628190506154875e-05 + } + }, + { + "step": 6400, + "logs": { + "loss": 0.6774, + "epoch": 2.3659889094269873, + "learning_rate": 1.624257678845322e-05 + } + }, + { + "step": 6450, + "logs": { + "loss": 0.6745, + "epoch": 2.3844731977818854, + "learning_rate": 1.6203248515357694e-05 + } + }, + { + "step": 6500, + "logs": { + "loss": 0.6809, + "epoch": 2.402957486136784, + "learning_rate": 1.6163920242262164e-05 + } + }, + { + "step": 6550, + "logs": { + "loss": 0.6831, + "epoch": 2.421441774491682, + "learning_rate": 1.6124591969166635e-05 + } + }, + { + "step": 6600, + "logs": { + "loss": 0.6803, + "epoch": 2.43992606284658, + "learning_rate": 1.608526369607111e-05 + } + }, + { + "step": 6650, + "logs": { + "loss": 0.6791, + "epoch": 2.4584103512014788, + "learning_rate": 1.604593542297558e-05 + } + }, + { + "step": 6700, + "logs": { + "loss": 0.6812, + "epoch": 2.476894639556377, + "learning_rate": 1.600660714988005e-05 + } + }, + { + "step": 6750, + "logs": { + "loss": 0.6866, + "epoch": 2.4953789279112755, + "learning_rate": 1.596727887678452e-05 + } + }, + { + "step": 6800, + "logs": { + "loss": 0.6837, + "epoch": 2.5138632162661736, + "learning_rate": 1.5927950603688995e-05 + } + }, + { + "step": 6850, + "logs": { + "loss": 0.6814, + "epoch": 2.532347504621072, + "learning_rate": 1.5888622330593465e-05 + } + }, + { + "step": 6900, + "logs": { + "loss": 0.6845, + "epoch": 2.5508317929759703, + "learning_rate": 1.5849294057497936e-05 + } + }, + { + "step": 6950, + "logs": { + "loss": 0.6832, + "epoch": 2.569316081330869, + "learning_rate": 1.5809965784402407e-05 + } + }, + { + "step": 7000, + "logs": { + "loss": 0.6795, + "epoch": 2.587800369685767, + "learning_rate": 1.5770637511306877e-05 + } + }, + { + "step": 7050, + "logs": { + "loss": 0.679, + "epoch": 2.6062846580406656, + "learning_rate": 1.573130923821135e-05 + } + }, + { + "step": 7100, + "logs": { + "loss": 0.6851, + "epoch": 2.6247689463955637, + "learning_rate": 1.5691980965115822e-05 + } + }, + { + "step": 7150, + "logs": { + "loss": 0.692, + "epoch": 2.6432532347504623, + "learning_rate": 1.5652652692020292e-05 + } + }, + { + "step": 7200, + "logs": { + "loss": 0.6956, + "epoch": 2.6617375231053604, + "learning_rate": 1.5613324418924766e-05 + } + }, + { + "step": 7250, + "logs": { + "loss": 0.6883, + "epoch": 2.680221811460259, + "learning_rate": 1.5573996145829237e-05 + } + }, + { + "step": 7300, + "logs": { + "loss": 0.6806, + "epoch": 2.698706099815157, + "learning_rate": 1.5534667872733708e-05 + } + }, + { + "step": 7350, + "logs": { + "loss": 0.6832, + "epoch": 2.7171903881700556, + "learning_rate": 1.549533959963818e-05 + } + }, + { + "step": 7400, + "logs": { + "loss": 0.6871, + "epoch": 2.7356746765249538, + "learning_rate": 1.5456011326542652e-05 + } + }, + { + "step": 7450, + "logs": { + "loss": 0.6843, + "epoch": 2.7541589648798523, + "learning_rate": 1.5416683053447123e-05 + } + }, + { + "step": 7500, + "logs": { + "loss": 0.6869, + "epoch": 2.7726432532347505, + "learning_rate": 1.5377354780351597e-05 + } + }, + { + "step": 7550, + "logs": { + "loss": 0.689, + "epoch": 2.7911275415896486, + "learning_rate": 1.5338026507256067e-05 + } + }, + { + "step": 7600, + "logs": { + "loss": 0.6864, + "epoch": 2.809611829944547, + "learning_rate": 1.5298698234160538e-05 + } + }, + { + "step": 7650, + "logs": { + "loss": 0.6875, + "epoch": 2.8280961182994453, + "learning_rate": 1.5259369961065012e-05 + } + }, + { + "step": 7700, + "logs": { + "loss": 0.6844, + "epoch": 2.846580406654344, + "learning_rate": 1.5220041687969483e-05 + } + }, + { + "step": 7750, + "logs": { + "loss": 0.6916, + "epoch": 2.865064695009242, + "learning_rate": 1.5180713414873953e-05 + } + }, + { + "step": 7800, + "logs": { + "loss": 0.6849, + "epoch": 2.8835489833641406, + "learning_rate": 1.5141385141778426e-05 + } + }, + { + "step": 7850, + "logs": { + "loss": 0.6835, + "epoch": 2.9020332717190387, + "learning_rate": 1.5102056868682898e-05 + } + }, + { + "step": 7900, + "logs": { + "loss": 0.6866, + "epoch": 2.9205175600739373, + "learning_rate": 1.5062728595587369e-05 + } + }, + { + "step": 7950, + "logs": { + "loss": 0.6863, + "epoch": 2.9390018484288354, + "learning_rate": 1.502340032249184e-05 + } + }, + { + "step": 8000, + "logs": { + "loss": 0.6903, + "epoch": 2.957486136783734, + "learning_rate": 1.4984072049396311e-05 + } + }, + { + "step": 8050, + "logs": { + "loss": 0.686, + "epoch": 2.975970425138632, + "learning_rate": 1.4944743776300784e-05 + } + }, + { + "step": 8100, + "logs": { + "loss": 0.6909, + "epoch": 2.9944547134935307, + "learning_rate": 1.4905415503205256e-05 + } + }, + { + "step": 8150, + "logs": { + "loss": 0.5832, + "epoch": 3.0129390018484288, + "learning_rate": 1.4866087230109727e-05 + } + }, + { + "step": 8200, + "logs": { + "loss": 0.5314, + "epoch": 3.0314232902033273, + "learning_rate": 1.4826758957014197e-05 + } + }, + { + "step": 8250, + "logs": { + "loss": 0.5332, + "epoch": 3.0499075785582255, + "learning_rate": 1.4787430683918671e-05 + } + }, + { + "step": 8300, + "logs": { + "loss": 0.5359, + "epoch": 3.068391866913124, + "learning_rate": 1.4748102410823142e-05 + } + }, + { + "step": 8350, + "logs": { + "loss": 0.5398, + "epoch": 3.086876155268022, + "learning_rate": 1.4708774137727613e-05 + } + }, + { + "step": 8400, + "logs": { + "loss": 0.5433, + "epoch": 3.1053604436229203, + "learning_rate": 1.4669445864632087e-05 + } + }, + { + "step": 8450, + "logs": { + "loss": 0.5368, + "epoch": 3.123844731977819, + "learning_rate": 1.4630117591536557e-05 + } + }, + { + "step": 8500, + "logs": { + "loss": 0.5419, + "epoch": 3.142329020332717, + "learning_rate": 1.4590789318441028e-05 + } + }, + { + "step": 8550, + "logs": { + "loss": 0.5398, + "epoch": 3.1608133086876156, + "learning_rate": 1.45514610453455e-05 + } + }, + { + "step": 8600, + "logs": { + "loss": 0.5442, + "epoch": 3.1792975970425137, + "learning_rate": 1.4512132772249972e-05 + } + }, + { + "step": 8650, + "logs": { + "loss": 0.544, + "epoch": 3.1977818853974123, + "learning_rate": 1.4472804499154443e-05 + } + }, + { + "step": 8700, + "logs": { + "loss": 0.5455, + "epoch": 3.2162661737523104, + "learning_rate": 1.4433476226058915e-05 + } + }, + { + "step": 8750, + "logs": { + "loss": 0.548, + "epoch": 3.234750462107209, + "learning_rate": 1.4394147952963386e-05 + } + }, + { + "step": 8800, + "logs": { + "loss": 0.5453, + "epoch": 3.253234750462107, + "learning_rate": 1.4354819679867858e-05 + } + }, + { + "step": 8850, + "logs": { + "loss": 0.5416, + "epoch": 3.2717190388170057, + "learning_rate": 1.431549140677233e-05 + } + }, + { + "step": 8900, + "logs": { + "loss": 0.5542, + "epoch": 3.290203327171904, + "learning_rate": 1.4276163133676801e-05 + } + }, + { + "step": 8950, + "logs": { + "loss": 0.5469, + "epoch": 3.3086876155268024, + "learning_rate": 1.4236834860581272e-05 + } + }, + { + "step": 9000, + "logs": { + "loss": 0.5545, + "epoch": 3.3271719038817005, + "learning_rate": 1.4197506587485746e-05 + } + }, + { + "step": 9050, + "logs": { + "loss": 0.5538, + "epoch": 3.345656192236599, + "learning_rate": 1.4158178314390216e-05 + } + }, + { + "step": 9100, + "logs": { + "loss": 0.5504, + "epoch": 3.364140480591497, + "learning_rate": 1.4118850041294687e-05 + } + }, + { + "step": 9150, + "logs": { + "loss": 0.5511, + "epoch": 3.3826247689463957, + "learning_rate": 1.4079521768199161e-05 + } + }, + { + "step": 9200, + "logs": { + "loss": 0.554, + "epoch": 3.401109057301294, + "learning_rate": 1.4040193495103632e-05 + } + }, + { + "step": 9250, + "logs": { + "loss": 0.5541, + "epoch": 3.4195933456561924, + "learning_rate": 1.4000865222008102e-05 + } + }, + { + "step": 9300, + "logs": { + "loss": 0.5502, + "epoch": 3.4380776340110906, + "learning_rate": 1.3961536948912575e-05 + } + }, + { + "step": 9350, + "logs": { + "loss": 0.5513, + "epoch": 3.4565619223659887, + "learning_rate": 1.3922208675817047e-05 + } + }, + { + "step": 9400, + "logs": { + "loss": 0.5579, + "epoch": 3.4750462107208873, + "learning_rate": 1.3882880402721517e-05 + } + }, + { + "step": 9450, + "logs": { + "loss": 0.5636, + "epoch": 3.4935304990757854, + "learning_rate": 1.384355212962599e-05 + } + }, + { + "step": 9500, + "logs": { + "loss": 0.5617, + "epoch": 3.512014787430684, + "learning_rate": 1.380422385653046e-05 + } + }, + { + "step": 9550, + "logs": { + "loss": 0.5549, + "epoch": 3.530499075785582, + "learning_rate": 1.3764895583434933e-05 + } + }, + { + "step": 9600, + "logs": { + "loss": 0.5599, + "epoch": 3.5489833641404807, + "learning_rate": 1.3725567310339405e-05 + } + }, + { + "step": 9650, + "logs": { + "loss": 0.5565, + "epoch": 3.567467652495379, + "learning_rate": 1.3686239037243876e-05 + } + }, + { + "step": 9700, + "logs": { + "loss": 0.5565, + "epoch": 3.5859519408502774, + "learning_rate": 1.3646910764148346e-05 + } + }, + { + "step": 9750, + "logs": { + "loss": 0.5558, + "epoch": 3.6044362292051755, + "learning_rate": 1.360758249105282e-05 + } + }, + { + "step": 9800, + "logs": { + "loss": 0.5599, + "epoch": 3.622920517560074, + "learning_rate": 1.356825421795729e-05 + } + }, + { + "step": 9850, + "logs": { + "loss": 0.561, + "epoch": 3.641404805914972, + "learning_rate": 1.3528925944861761e-05 + } + }, + { + "step": 9900, + "logs": { + "loss": 0.5596, + "epoch": 3.6598890942698707, + "learning_rate": 1.3489597671766235e-05 + } + }, + { + "step": 9950, + "logs": { + "loss": 0.5596, + "epoch": 3.678373382624769, + "learning_rate": 1.3450269398670706e-05 + } + }, + { + "step": 10000, + "logs": { + "loss": 0.5579, + "epoch": 3.6968576709796674, + "learning_rate": 1.3410941125575177e-05 + } + }, + { + "step": 10050, + "logs": { + "loss": 0.5611, + "epoch": 3.7153419593345656, + "learning_rate": 1.3371612852479649e-05 + } + }, + { + "step": 10100, + "logs": { + "loss": 0.5627, + "epoch": 3.733826247689464, + "learning_rate": 1.3332284579384121e-05 + } + }, + { + "step": 10150, + "logs": { + "loss": 0.5613, + "epoch": 3.7523105360443623, + "learning_rate": 1.3292956306288592e-05 + } + }, + { + "step": 10200, + "logs": { + "loss": 0.5607, + "epoch": 3.770794824399261, + "learning_rate": 1.3253628033193064e-05 + } + }, + { + "step": 10250, + "logs": { + "loss": 0.5651, + "epoch": 3.789279112754159, + "learning_rate": 1.3214299760097535e-05 + } + }, + { + "step": 10300, + "logs": { + "loss": 0.5604, + "epoch": 3.807763401109057, + "learning_rate": 1.3174971487002005e-05 + } + }, + { + "step": 10350, + "logs": { + "loss": 0.5676, + "epoch": 3.8262476894639557, + "learning_rate": 1.313564321390648e-05 + } + }, + { + "step": 10400, + "logs": { + "loss": 0.5601, + "epoch": 3.844731977818854, + "learning_rate": 1.309631494081095e-05 + } + }, + { + "step": 10450, + "logs": { + "loss": 0.5645, + "epoch": 3.8632162661737524, + "learning_rate": 1.305698666771542e-05 + } + }, + { + "step": 10500, + "logs": { + "loss": 0.5653, + "epoch": 3.8817005545286505, + "learning_rate": 1.3017658394619895e-05 + } + }, + { + "step": 10550, + "logs": { + "loss": 0.5618, + "epoch": 3.900184842883549, + "learning_rate": 1.2978330121524365e-05 + } + }, + { + "step": 10600, + "logs": { + "loss": 0.5624, + "epoch": 3.918669131238447, + "learning_rate": 1.2939001848428836e-05 + } + }, + { + "step": 10650, + "logs": { + "loss": 0.5621, + "epoch": 3.9371534195933457, + "learning_rate": 1.289967357533331e-05 + } + }, + { + "step": 10700, + "logs": { + "loss": 0.5633, + "epoch": 3.955637707948244, + "learning_rate": 1.286034530223778e-05 + } + }, + { + "step": 10750, + "logs": { + "loss": 0.5645, + "epoch": 3.9741219963031424, + "learning_rate": 1.2821017029142251e-05 + } + }, + { + "step": 10800, + "logs": { + "loss": 0.5675, + "epoch": 3.9926062846580406, + "learning_rate": 1.2781688756046723e-05 + } + }, + { + "step": 10850, + "logs": { + "loss": 0.4943, + "epoch": 4.011090573012939, + "learning_rate": 1.2742360482951194e-05 + } + }, + { + "step": 10900, + "logs": { + "loss": 0.4282, + "epoch": 4.029574861367838, + "learning_rate": 1.2703032209855666e-05 + } + }, + { + "step": 10950, + "logs": { + "loss": 0.4285, + "epoch": 4.048059149722736, + "learning_rate": 1.2663703936760139e-05 + } + }, + { + "step": 11000, + "logs": { + "loss": 0.4307, + "epoch": 4.066543438077634, + "learning_rate": 1.262437566366461e-05 + } + }, + { + "step": 11050, + "logs": { + "loss": 0.4372, + "epoch": 4.085027726432532, + "learning_rate": 1.258504739056908e-05 + } + }, + { + "step": 11100, + "logs": { + "loss": 0.4377, + "epoch": 4.103512014787431, + "learning_rate": 1.2545719117473554e-05 + } + }, + { + "step": 11150, + "logs": { + "loss": 0.4296, + "epoch": 4.121996303142329, + "learning_rate": 1.2506390844378025e-05 + } + }, + { + "step": 11200, + "logs": { + "loss": 0.4358, + "epoch": 4.140480591497227, + "learning_rate": 1.2467062571282495e-05 + } + }, + { + "step": 11250, + "logs": { + "loss": 0.4383, + "epoch": 4.1589648798521255, + "learning_rate": 1.2427734298186969e-05 + } + }, + { + "step": 11300, + "logs": { + "loss": 0.4355, + "epoch": 4.177449168207024, + "learning_rate": 1.238840602509144e-05 + } + }, + { + "step": 11350, + "logs": { + "loss": 0.44, + "epoch": 4.195933456561923, + "learning_rate": 1.234907775199591e-05 + } + }, + { + "step": 11400, + "logs": { + "loss": 0.4393, + "epoch": 4.214417744916821, + "learning_rate": 1.2309749478900384e-05 + } + }, + { + "step": 11450, + "logs": { + "loss": 0.4427, + "epoch": 4.232902033271719, + "learning_rate": 1.2270421205804855e-05 + } + }, + { + "step": 11500, + "logs": { + "loss": 0.4399, + "epoch": 4.251386321626617, + "learning_rate": 1.2231092932709326e-05 + } + }, + { + "step": 11550, + "logs": { + "loss": 0.4401, + "epoch": 4.269870609981516, + "learning_rate": 1.2191764659613798e-05 + } + }, + { + "step": 11600, + "logs": { + "loss": 0.4446, + "epoch": 4.288354898336414, + "learning_rate": 1.2152436386518269e-05 + } + }, + { + "step": 11650, + "logs": { + "loss": 0.4436, + "epoch": 4.306839186691312, + "learning_rate": 1.211310811342274e-05 + } + }, + { + "step": 11700, + "logs": { + "loss": 0.4478, + "epoch": 4.32532347504621, + "learning_rate": 1.2073779840327213e-05 + } + }, + { + "step": 11750, + "logs": { + "loss": 0.447, + "epoch": 4.343807763401109, + "learning_rate": 1.2034451567231684e-05 + } + }, + { + "step": 11800, + "logs": { + "loss": 0.4505, + "epoch": 4.3622920517560075, + "learning_rate": 1.1995123294136154e-05 + } + }, + { + "step": 11850, + "logs": { + "loss": 0.4461, + "epoch": 4.380776340110906, + "learning_rate": 1.1955795021040628e-05 + } + }, + { + "step": 11900, + "logs": { + "loss": 0.4511, + "epoch": 4.399260628465804, + "learning_rate": 1.1916466747945099e-05 + } + }, + { + "step": 11950, + "logs": { + "loss": 0.4481, + "epoch": 4.417744916820703, + "learning_rate": 1.187713847484957e-05 + } + }, + { + "step": 12000, + "logs": { + "loss": 0.4457, + "epoch": 4.436229205175601, + "learning_rate": 1.1837810201754044e-05 + } + }, + { + "step": 12050, + "logs": { + "loss": 0.4491, + "epoch": 4.454713493530499, + "learning_rate": 1.1798481928658514e-05 + } + }, + { + "step": 12100, + "logs": { + "loss": 0.453, + "epoch": 4.473197781885397, + "learning_rate": 1.1759153655562985e-05 + } + }, + { + "step": 12150, + "logs": { + "loss": 0.4557, + "epoch": 4.491682070240295, + "learning_rate": 1.1719825382467455e-05 + } + }, + { + "step": 12200, + "logs": { + "loss": 0.4588, + "epoch": 4.510166358595194, + "learning_rate": 1.168049710937193e-05 + } + }, + { + "step": 12250, + "logs": { + "loss": 0.4478, + "epoch": 4.5286506469500925, + "learning_rate": 1.16411688362764e-05 + } + }, + { + "step": 12300, + "logs": { + "loss": 0.4558, + "epoch": 4.547134935304991, + "learning_rate": 1.160184056318087e-05 + } + }, + { + "step": 12350, + "logs": { + "loss": 0.453, + "epoch": 4.565619223659889, + "learning_rate": 1.1562512290085343e-05 + } + }, + { + "step": 12400, + "logs": { + "loss": 0.4566, + "epoch": 4.584103512014788, + "learning_rate": 1.1523184016989815e-05 + } + }, + { + "step": 12450, + "logs": { + "loss": 0.4512, + "epoch": 4.602587800369686, + "learning_rate": 1.1483855743894286e-05 + } + }, + { + "step": 12500, + "logs": { + "loss": 0.4537, + "epoch": 4.621072088724584, + "learning_rate": 1.1444527470798758e-05 + } + }, + { + "step": 12550, + "logs": { + "loss": 0.4596, + "epoch": 4.639556377079482, + "learning_rate": 1.1405199197703229e-05 + } + }, + { + "step": 12600, + "logs": { + "loss": 0.4544, + "epoch": 4.658040665434381, + "learning_rate": 1.1365870924607701e-05 + } + }, + { + "step": 12650, + "logs": { + "loss": 0.4552, + "epoch": 4.676524953789279, + "learning_rate": 1.1326542651512173e-05 + } + }, + { + "step": 12700, + "logs": { + "loss": 0.4561, + "epoch": 4.695009242144177, + "learning_rate": 1.1287214378416644e-05 + } + }, + { + "step": 12750, + "logs": { + "loss": 0.456, + "epoch": 4.7134935304990755, + "learning_rate": 1.1247886105321115e-05 + } + }, + { + "step": 12800, + "logs": { + "loss": 0.455, + "epoch": 4.7319778188539745, + "learning_rate": 1.1208557832225589e-05 + } + }, + { + "step": 12850, + "logs": { + "loss": 0.4579, + "epoch": 4.750462107208873, + "learning_rate": 1.116922955913006e-05 + } + }, + { + "step": 12900, + "logs": { + "loss": 0.4569, + "epoch": 4.768946395563771, + "learning_rate": 1.112990128603453e-05 + } + }, + { + "step": 12950, + "logs": { + "loss": 0.459, + "epoch": 4.787430683918669, + "learning_rate": 1.1090573012939004e-05 + } + }, + { + "step": 13000, + "logs": { + "loss": 0.456, + "epoch": 4.805914972273568, + "learning_rate": 1.1051244739843475e-05 + } + }, + { + "step": 13050, + "logs": { + "loss": 0.4605, + "epoch": 4.824399260628466, + "learning_rate": 1.1011916466747945e-05 + } + }, + { + "step": 13100, + "logs": { + "loss": 0.4565, + "epoch": 4.842883548983364, + "learning_rate": 1.0972588193652417e-05 + } + }, + { + "step": 13150, + "logs": { + "loss": 0.4606, + "epoch": 4.861367837338262, + "learning_rate": 1.093325992055689e-05 + } + }, + { + "step": 13200, + "logs": { + "loss": 0.4586, + "epoch": 4.87985212569316, + "learning_rate": 1.089393164746136e-05 + } + }, + { + "step": 13250, + "logs": { + "loss": 0.4605, + "epoch": 4.898336414048059, + "learning_rate": 1.0854603374365833e-05 + } + }, + { + "step": 13300, + "logs": { + "loss": 0.4565, + "epoch": 4.9168207024029575, + "learning_rate": 1.0815275101270303e-05 + } + }, + { + "step": 13350, + "logs": { + "loss": 0.4585, + "epoch": 4.935304990757856, + "learning_rate": 1.0775946828174776e-05 + } + }, + { + "step": 13400, + "logs": { + "loss": 0.4588, + "epoch": 4.953789279112754, + "learning_rate": 1.0736618555079248e-05 + } + }, + { + "step": 13450, + "logs": { + "loss": 0.4639, + "epoch": 4.972273567467653, + "learning_rate": 1.0697290281983718e-05 + } + }, + { + "step": 13500, + "logs": { + "loss": 0.4666, + "epoch": 4.990757855822551, + "learning_rate": 1.0657962008888189e-05 + } + }, + { + "step": 13550, + "logs": { + "loss": 0.4109, + "epoch": 5.009242144177449, + "learning_rate": 1.0618633735792663e-05 + } + }, + { + "step": 13600, + "logs": { + "loss": 0.3538, + "epoch": 5.027726432532347, + "learning_rate": 1.0579305462697134e-05 + } + }, + { + "step": 13650, + "logs": { + "loss": 0.346, + "epoch": 5.046210720887246, + "learning_rate": 1.0539977189601604e-05 + } + }, + { + "step": 13700, + "logs": { + "loss": 0.3483, + "epoch": 5.064695009242144, + "learning_rate": 1.0500648916506078e-05 + } + }, + { + "step": 13750, + "logs": { + "loss": 0.3548, + "epoch": 5.0831792975970425, + "learning_rate": 1.0461320643410549e-05 + } + }, + { + "step": 13800, + "logs": { + "loss": 0.3567, + "epoch": 5.101663585951941, + "learning_rate": 1.042199237031502e-05 + } + }, + { + "step": 13850, + "logs": { + "loss": 0.3525, + "epoch": 5.12014787430684, + "learning_rate": 1.0382664097219492e-05 + } + }, + { + "step": 13900, + "logs": { + "loss": 0.3563, + "epoch": 5.138632162661738, + "learning_rate": 1.0343335824123964e-05 + } + }, + { + "step": 13950, + "logs": { + "loss": 0.3606, + "epoch": 5.157116451016636, + "learning_rate": 1.0304007551028435e-05 + } + }, + { + "step": 14000, + "logs": { + "loss": 0.3585, + "epoch": 5.175600739371534, + "learning_rate": 1.0264679277932907e-05 + } + }, + { + "step": 14050, + "logs": { + "loss": 0.3547, + "epoch": 5.194085027726432, + "learning_rate": 1.0225351004837378e-05 + } + }, + { + "step": 14100, + "logs": { + "loss": 0.359, + "epoch": 5.212569316081331, + "learning_rate": 1.018602273174185e-05 + } + }, + { + "step": 14150, + "logs": { + "loss": 0.3616, + "epoch": 5.231053604436229, + "learning_rate": 1.0146694458646322e-05 + } + }, + { + "step": 14200, + "logs": { + "loss": 0.3602, + "epoch": 5.249537892791127, + "learning_rate": 1.0107366185550793e-05 + } + }, + { + "step": 14250, + "logs": { + "loss": 0.3607, + "epoch": 5.2680221811460255, + "learning_rate": 1.0068037912455264e-05 + } + }, + { + "step": 14300, + "logs": { + "loss": 0.3642, + "epoch": 5.2865064695009245, + "learning_rate": 1.0028709639359738e-05 + } + }, + { + "step": 14350, + "logs": { + "loss": 0.3616, + "epoch": 5.304990757855823, + "learning_rate": 9.989381366264208e-06 + } + }, + { + "step": 14400, + "logs": { + "loss": 0.364, + "epoch": 5.323475046210721, + "learning_rate": 9.95005309316868e-06 + } + }, + { + "step": 14450, + "logs": { + "loss": 0.3656, + "epoch": 5.341959334565619, + "learning_rate": 9.910724820073151e-06 + } + }, + { + "step": 14500, + "logs": { + "loss": 0.3613, + "epoch": 5.360443622920518, + "learning_rate": 9.871396546977623e-06 + } + }, + { + "step": 14550, + "logs": { + "loss": 0.3655, + "epoch": 5.378927911275416, + "learning_rate": 9.832068273882096e-06 + } + }, + { + "step": 14600, + "logs": { + "loss": 0.3677, + "epoch": 5.397412199630314, + "learning_rate": 9.792740000786566e-06 + } + }, + { + "step": 14650, + "logs": { + "loss": 0.3678, + "epoch": 5.415896487985212, + "learning_rate": 9.753411727691039e-06 + } + }, + { + "step": 14700, + "logs": { + "loss": 0.364, + "epoch": 5.434380776340111, + "learning_rate": 9.71408345459551e-06 + } + }, + { + "step": 14750, + "logs": { + "loss": 0.3654, + "epoch": 5.452865064695009, + "learning_rate": 9.67475518149998e-06 + } + }, + { + "step": 14800, + "logs": { + "loss": 0.3671, + "epoch": 5.4713493530499075, + "learning_rate": 9.635426908404452e-06 + } + }, + { + "step": 14850, + "logs": { + "loss": 0.3724, + "epoch": 5.489833641404806, + "learning_rate": 9.596098635308924e-06 + } + }, + { + "step": 14900, + "logs": { + "loss": 1.4537, + "epoch": 5.508317929759705, + "learning_rate": 9.556770362213395e-06 + } + }, + { + "step": 14950, + "logs": { + "loss": 0.3685, + "epoch": 5.526802218114603, + "learning_rate": 9.517442089117867e-06 + } + }, + { + "step": 15000, + "logs": { + "loss": 0.3688, + "epoch": 5.545286506469501, + "learning_rate": 9.47811381602234e-06 + } + }, + { + "step": 15050, + "logs": { + "loss": 0.3726, + "epoch": 5.563770794824399, + "learning_rate": 9.43878554292681e-06 + } + }, + { + "step": 15100, + "logs": { + "loss": 0.3718, + "epoch": 5.582255083179297, + "learning_rate": 9.399457269831283e-06 + } + }, + { + "step": 15150, + "logs": { + "loss": 0.3691, + "epoch": 5.600739371534196, + "learning_rate": 9.360128996735755e-06 + } + }, + { + "step": 15200, + "logs": { + "loss": 0.3684, + "epoch": 5.619223659889094, + "learning_rate": 9.320800723640226e-06 + } + }, + { + "step": 15250, + "logs": { + "loss": 0.3747, + "epoch": 5.6377079482439925, + "learning_rate": 9.281472450544698e-06 + } + }, + { + "step": 15300, + "logs": { + "loss": 0.3727, + "epoch": 5.656192236598891, + "learning_rate": 9.24214417744917e-06 + } + }, + { + "step": 15350, + "logs": { + "loss": 0.3736, + "epoch": 5.67467652495379, + "learning_rate": 9.20281590435364e-06 + } + }, + { + "step": 15400, + "logs": { + "loss": 0.3725, + "epoch": 5.693160813308688, + "learning_rate": 9.163487631258111e-06 + } + }, + { + "step": 15450, + "logs": { + "loss": 0.3711, + "epoch": 5.711645101663586, + "learning_rate": 9.124159358162584e-06 + } + }, + { + "step": 15500, + "logs": { + "loss": 0.3705, + "epoch": 5.730129390018484, + "learning_rate": 9.084831085067054e-06 + } + }, + { + "step": 15550, + "logs": { + "loss": 0.3698, + "epoch": 5.748613678373383, + "learning_rate": 9.045502811971527e-06 + } + }, + { + "step": 15600, + "logs": { + "loss": 0.3725, + "epoch": 5.767097966728281, + "learning_rate": 9.006174538875999e-06 + } + }, + { + "step": 15650, + "logs": { + "loss": 0.3767, + "epoch": 5.785582255083179, + "learning_rate": 8.96684626578047e-06 + } + }, + { + "step": 15700, + "logs": { + "loss": 0.3696, + "epoch": 5.804066543438077, + "learning_rate": 8.927517992684942e-06 + } + }, + { + "step": 15750, + "logs": { + "loss": 0.3797, + "epoch": 5.822550831792976, + "learning_rate": 8.888189719589414e-06 + } + }, + { + "step": 15800, + "logs": { + "loss": 0.371, + "epoch": 5.8410351201478745, + "learning_rate": 8.848861446493885e-06 + } + }, + { + "step": 15850, + "logs": { + "loss": 0.3769, + "epoch": 5.859519408502773, + "learning_rate": 8.809533173398357e-06 + } + }, + { + "step": 15900, + "logs": { + "loss": 0.3731, + "epoch": 5.878003696857671, + "learning_rate": 8.77020490030283e-06 + } + }, + { + "step": 15950, + "logs": { + "loss": 0.3769, + "epoch": 5.896487985212569, + "learning_rate": 8.7308766272073e-06 + } + }, + { + "step": 16000, + "logs": { + "loss": 0.3711, + "epoch": 5.914972273567468, + "learning_rate": 8.691548354111772e-06 + } + }, + { + "step": 16050, + "logs": { + "loss": 0.3752, + "epoch": 5.933456561922366, + "learning_rate": 8.652220081016243e-06 + } + }, + { + "step": 16100, + "logs": { + "loss": 0.3697, + "epoch": 5.951940850277264, + "learning_rate": 8.612891807920715e-06 + } + }, + { + "step": 16150, + "logs": { + "loss": 0.3751, + "epoch": 5.970425138632162, + "learning_rate": 8.573563534825186e-06 + } + }, + { + "step": 16200, + "logs": { + "loss": 0.3791, + "epoch": 5.988909426987061, + "learning_rate": 8.534235261729658e-06 + } + }, + { + "step": 16250, + "logs": { + "loss": 0.3443, + "epoch": 6.007393715341959, + "learning_rate": 8.494906988634129e-06 + } + }, + { + "step": 16300, + "logs": { + "loss": 0.2903, + "epoch": 6.0258780036968576, + "learning_rate": 8.455578715538601e-06 + } + }, + { + "step": 16350, + "logs": { + "loss": 0.2906, + "epoch": 6.044362292051756, + "learning_rate": 8.416250442443073e-06 + } + }, + { + "step": 16400, + "logs": { + "loss": 0.2884, + "epoch": 6.062846580406655, + "learning_rate": 8.376922169347544e-06 + } + }, + { + "step": 16450, + "logs": { + "loss": 0.2913, + "epoch": 6.081330868761553, + "learning_rate": 8.337593896252016e-06 + } + }, + { + "step": 16500, + "logs": { + "loss": 0.2964, + "epoch": 6.099815157116451, + "learning_rate": 8.298265623156489e-06 + } + }, + { + "step": 16550, + "logs": { + "loss": 0.2877, + "epoch": 6.118299445471349, + "learning_rate": 8.25893735006096e-06 + } + }, + { + "step": 16600, + "logs": { + "loss": 0.2957, + "epoch": 6.136783733826248, + "learning_rate": 8.219609076965432e-06 + } + }, + { + "step": 16650, + "logs": { + "loss": 0.2974, + "epoch": 6.155268022181146, + "learning_rate": 8.180280803869902e-06 + } + }, + { + "step": 16700, + "logs": { + "loss": 0.295, + "epoch": 6.173752310536044, + "learning_rate": 8.140952530774374e-06 + } + }, + { + "step": 16750, + "logs": { + "loss": 0.2947, + "epoch": 6.1922365988909425, + "learning_rate": 8.101624257678847e-06 + } + }, + { + "step": 16800, + "logs": { + "loss": 0.2957, + "epoch": 6.210720887245841, + "learning_rate": 8.062295984583317e-06 + } + }, + { + "step": 16850, + "logs": { + "loss": 0.3001, + "epoch": 6.22920517560074, + "learning_rate": 8.02296771148779e-06 + } + }, + { + "step": 16900, + "logs": { + "loss": 0.2958, + "epoch": 6.247689463955638, + "learning_rate": 7.98363943839226e-06 + } + }, + { + "step": 16950, + "logs": { + "loss": 0.2956, + "epoch": 6.266173752310536, + "learning_rate": 7.944311165296733e-06 + } + }, + { + "step": 17000, + "logs": { + "loss": 0.3, + "epoch": 6.284658040665434, + "learning_rate": 7.904982892201203e-06 + } + }, + { + "step": 17050, + "logs": { + "loss": 0.3, + "epoch": 6.303142329020333, + "learning_rate": 7.865654619105676e-06 + } + }, + { + "step": 17100, + "logs": { + "loss": 0.3003, + "epoch": 6.321626617375231, + "learning_rate": 7.826326346010146e-06 + } + }, + { + "step": 17150, + "logs": { + "loss": 0.3029, + "epoch": 6.340110905730129, + "learning_rate": 7.786998072914618e-06 + } + }, + { + "step": 17200, + "logs": { + "loss": 0.301, + "epoch": 6.358595194085027, + "learning_rate": 7.74766979981909e-06 + } + }, + { + "step": 17250, + "logs": { + "loss": 0.297, + "epoch": 6.377079482439926, + "learning_rate": 7.708341526723561e-06 + } + }, + { + "step": 17300, + "logs": { + "loss": 0.2993, + "epoch": 6.3955637707948245, + "learning_rate": 7.669013253628034e-06 + } + }, + { + "step": 17350, + "logs": { + "loss": 0.3035, + "epoch": 6.414048059149723, + "learning_rate": 7.629684980532506e-06 + } + }, + { + "step": 17400, + "logs": { + "loss": 0.2988, + "epoch": 6.432532347504621, + "learning_rate": 7.590356707436977e-06 + } + }, + { + "step": 17450, + "logs": { + "loss": 0.3001, + "epoch": 6.45101663585952, + "learning_rate": 7.551028434341449e-06 + } + }, + { + "step": 17500, + "logs": { + "loss": 0.3024, + "epoch": 6.469500924214418, + "learning_rate": 7.51170016124592e-06 + } + }, + { + "step": 17550, + "logs": { + "loss": 0.3053, + "epoch": 6.487985212569316, + "learning_rate": 7.472371888150392e-06 + } + }, + { + "step": 17600, + "logs": { + "loss": 0.3081, + "epoch": 6.506469500924214, + "learning_rate": 7.433043615054863e-06 + } + }, + { + "step": 17650, + "logs": { + "loss": 0.3002, + "epoch": 6.524953789279113, + "learning_rate": 7.393715341959336e-06 + } + }, + { + "step": 17700, + "logs": { + "loss": 0.3044, + "epoch": 6.543438077634011, + "learning_rate": 7.354387068863806e-06 + } + }, + { + "step": 17750, + "logs": { + "loss": 0.3016, + "epoch": 6.561922365988909, + "learning_rate": 7.3150587957682786e-06 + } + }, + { + "step": 17800, + "logs": { + "loss": 0.3071, + "epoch": 6.580406654343808, + "learning_rate": 7.27573052267275e-06 + } + }, + { + "step": 17850, + "logs": { + "loss": 0.3002, + "epoch": 6.598890942698706, + "learning_rate": 7.2364022495772215e-06 + } + }, + { + "step": 17900, + "logs": { + "loss": 0.3042, + "epoch": 6.617375231053605, + "learning_rate": 7.197073976481693e-06 + } + }, + { + "step": 17950, + "logs": { + "loss": 0.3042, + "epoch": 6.635859519408503, + "learning_rate": 7.157745703386165e-06 + } + }, + { + "step": 18000, + "logs": { + "loss": 0.3063, + "epoch": 6.654343807763401, + "learning_rate": 7.118417430290636e-06 + } + }, + { + "step": 18050, + "logs": { + "loss": 0.3031, + "epoch": 6.672828096118299, + "learning_rate": 7.079089157195108e-06 + } + }, + { + "step": 18100, + "logs": { + "loss": 0.3081, + "epoch": 6.691312384473198, + "learning_rate": 7.0397608840995805e-06 + } + }, + { + "step": 18150, + "logs": { + "loss": 0.3041, + "epoch": 6.709796672828096, + "learning_rate": 7.000432611004051e-06 + } + }, + { + "step": 18200, + "logs": { + "loss": 0.3036, + "epoch": 6.728280961182994, + "learning_rate": 6.961104337908523e-06 + } + }, + { + "step": 18250, + "logs": { + "loss": 0.3044, + "epoch": 6.7467652495378925, + "learning_rate": 6.921776064812995e-06 + } + }, + { + "step": 18300, + "logs": { + "loss": 0.3057, + "epoch": 6.7652495378927915, + "learning_rate": 6.882447791717466e-06 + } + }, + { + "step": 18350, + "logs": { + "loss": 0.3081, + "epoch": 6.78373382624769, + "learning_rate": 6.843119518621938e-06 + } + }, + { + "step": 18400, + "logs": { + "loss": 0.3058, + "epoch": 6.802218114602588, + "learning_rate": 6.80379124552641e-06 + } + }, + { + "step": 18450, + "logs": { + "loss": 0.3072, + "epoch": 6.820702402957486, + "learning_rate": 6.764462972430881e-06 + } + }, + { + "step": 18500, + "logs": { + "loss": 0.3042, + "epoch": 6.839186691312385, + "learning_rate": 6.725134699335353e-06 + } + }, + { + "step": 18550, + "logs": { + "loss": 0.3091, + "epoch": 6.857670979667283, + "learning_rate": 6.6858064262398245e-06 + } + }, + { + "step": 18600, + "logs": { + "loss": 0.3036, + "epoch": 6.876155268022181, + "learning_rate": 6.646478153144296e-06 + } + }, + { + "step": 18650, + "logs": { + "loss": 0.3081, + "epoch": 6.894639556377079, + "learning_rate": 6.607149880048767e-06 + } + }, + { + "step": 18700, + "logs": { + "loss": 0.3018, + "epoch": 6.913123844731977, + "learning_rate": 6.56782160695324e-06 + } + }, + { + "step": 18750, + "logs": { + "loss": 0.3055, + "epoch": 6.931608133086876, + "learning_rate": 6.52849333385771e-06 + } + }, + { + "step": 18800, + "logs": { + "loss": 0.3046, + "epoch": 6.9500924214417745, + "learning_rate": 6.489165060762183e-06 + } + }, + { + "step": 18850, + "logs": { + "loss": 0.3083, + "epoch": 6.968576709796673, + "learning_rate": 6.449836787666655e-06 + } + }, + { + "step": 18900, + "logs": { + "loss": 0.3118, + "epoch": 6.987060998151571, + "learning_rate": 6.4105085145711256e-06 + } + }, + { + "step": 18950, + "logs": { + "loss": 0.2902, + "epoch": 7.00554528650647, + "learning_rate": 6.371180241475597e-06 + } + }, + { + "step": 19000, + "logs": { + "loss": 0.241, + "epoch": 7.024029574861368, + "learning_rate": 6.331851968380069e-06 + } + }, + { + "step": 19050, + "logs": { + "loss": 0.2432, + "epoch": 7.042513863216266, + "learning_rate": 6.29252369528454e-06 + } + }, + { + "step": 19100, + "logs": { + "loss": 0.2405, + "epoch": 7.060998151571164, + "learning_rate": 6.253195422189012e-06 + } + }, + { + "step": 19150, + "logs": { + "loss": 0.2447, + "epoch": 7.079482439926063, + "learning_rate": 6.2138671490934846e-06 + } + }, + { + "step": 19200, + "logs": { + "loss": 0.2458, + "epoch": 7.097966728280961, + "learning_rate": 6.174538875997955e-06 + } + }, + { + "step": 19250, + "logs": { + "loss": 0.2414, + "epoch": 7.116451016635859, + "learning_rate": 6.1352106029024275e-06 + } + }, + { + "step": 19300, + "logs": { + "loss": 0.2465, + "epoch": 7.134935304990758, + "learning_rate": 6.095882329806899e-06 + } + }, + { + "step": 19350, + "logs": { + "loss": 0.2474, + "epoch": 7.153419593345657, + "learning_rate": 6.05655405671137e-06 + } + }, + { + "step": 19400, + "logs": { + "loss": 0.2464, + "epoch": 7.171903881700555, + "learning_rate": 6.017225783615842e-06 + } + }, + { + "step": 19450, + "logs": { + "loss": 0.2443, + "epoch": 7.190388170055453, + "learning_rate": 5.977897510520314e-06 + } + }, + { + "step": 19500, + "logs": { + "loss": 0.2466, + "epoch": 7.208872458410351, + "learning_rate": 5.938569237424785e-06 + } + }, + { + "step": 19550, + "logs": { + "loss": 0.2497, + "epoch": 7.22735674676525, + "learning_rate": 5.899240964329257e-06 + } + }, + { + "step": 19600, + "logs": { + "loss": 0.2469, + "epoch": 7.245841035120148, + "learning_rate": 5.859912691233728e-06 + } + }, + { + "step": 19650, + "logs": { + "loss": 0.2468, + "epoch": 7.264325323475046, + "learning_rate": 5.8205844181382e-06 + } + }, + { + "step": 19700, + "logs": { + "loss": 0.2495, + "epoch": 7.282809611829944, + "learning_rate": 5.7812561450426715e-06 + } + }, + { + "step": 19750, + "logs": { + "loss": 0.2496, + "epoch": 7.3012939001848425, + "learning_rate": 5.741927871947143e-06 + } + }, + { + "step": 19800, + "logs": { + "loss": 0.2516, + "epoch": 7.3197781885397415, + "learning_rate": 5.702599598851614e-06 + } + }, + { + "step": 19850, + "logs": { + "loss": 0.2516, + "epoch": 7.33826247689464, + "learning_rate": 5.663271325756087e-06 + } + }, + { + "step": 19900, + "logs": { + "loss": 0.2493, + "epoch": 7.356746765249538, + "learning_rate": 5.623943052660557e-06 + } + }, + { + "step": 19950, + "logs": { + "loss": 0.2458, + "epoch": 7.375231053604436, + "learning_rate": 5.58461477956503e-06 + } + }, + { + "step": 20000, + "logs": { + "loss": 0.2494, + "epoch": 7.393715341959335, + "learning_rate": 5.545286506469502e-06 + } + }, + { + "step": 20050, + "logs": { + "loss": 0.2534, + "epoch": 7.412199630314233, + "learning_rate": 5.5059582333739726e-06 + } + }, + { + "step": 20100, + "logs": { + "loss": 0.2499, + "epoch": 7.430683918669131, + "learning_rate": 5.466629960278445e-06 + } + }, + { + "step": 20150, + "logs": { + "loss": 0.249, + "epoch": 7.449168207024029, + "learning_rate": 5.427301687182916e-06 + } + }, + { + "step": 20200, + "logs": { + "loss": 0.2498, + "epoch": 7.467652495378928, + "learning_rate": 5.387973414087388e-06 + } + }, + { + "step": 20250, + "logs": { + "loss": 0.2541, + "epoch": 7.486136783733826, + "learning_rate": 5.348645140991859e-06 + } + }, + { + "step": 20300, + "logs": { + "loss": 0.2579, + "epoch": 7.5046210720887245, + "learning_rate": 5.3093168678963316e-06 + } + }, + { + "step": 20350, + "logs": { + "loss": 0.2476, + "epoch": 7.523105360443623, + "learning_rate": 5.269988594800802e-06 + } + }, + { + "step": 20400, + "logs": { + "loss": 0.2538, + "epoch": 7.541589648798522, + "learning_rate": 5.2306603217052745e-06 + } + }, + { + "step": 20450, + "logs": { + "loss": 0.2537, + "epoch": 7.56007393715342, + "learning_rate": 5.191332048609746e-06 + } + }, + { + "step": 20500, + "logs": { + "loss": 0.2522, + "epoch": 7.578558225508318, + "learning_rate": 5.152003775514217e-06 + } + }, + { + "step": 20550, + "logs": { + "loss": 0.2491, + "epoch": 7.597042513863216, + "learning_rate": 5.112675502418689e-06 + } + }, + { + "step": 20600, + "logs": { + "loss": 0.2509, + "epoch": 7.615526802218114, + "learning_rate": 5.073347229323161e-06 + } + }, + { + "step": 20650, + "logs": { + "loss": 0.2496, + "epoch": 7.634011090573013, + "learning_rate": 5.034018956227632e-06 + } + }, + { + "step": 20700, + "logs": { + "loss": 0.2542, + "epoch": 7.652495378927911, + "learning_rate": 4.994690683132104e-06 + } + }, + { + "step": 20750, + "logs": { + "loss": 0.2517, + "epoch": 7.6709796672828094, + "learning_rate": 4.9553624100365756e-06 + } + }, + { + "step": 20800, + "logs": { + "loss": 0.2544, + "epoch": 7.689463955637708, + "learning_rate": 4.916034136941048e-06 + } + }, + { + "step": 20850, + "logs": { + "loss": 0.2489, + "epoch": 7.707948243992607, + "learning_rate": 4.876705863845519e-06 + } + }, + { + "step": 20900, + "logs": { + "loss": 0.2506, + "epoch": 7.726432532347505, + "learning_rate": 4.83737759074999e-06 + } + }, + { + "step": 20950, + "logs": { + "loss": 0.2535, + "epoch": 7.744916820702403, + "learning_rate": 4.798049317654462e-06 + } + }, + { + "step": 21000, + "logs": { + "loss": 0.2506, + "epoch": 7.763401109057301, + "learning_rate": 4.758721044558934e-06 + } + }, + { + "step": 21050, + "logs": { + "loss": 0.2547, + "epoch": 7.7818853974122, + "learning_rate": 4.719392771463405e-06 + } + }, + { + "step": 21100, + "logs": { + "loss": 0.2525, + "epoch": 7.800369685767098, + "learning_rate": 4.6800644983678775e-06 + } + }, + { + "step": 21150, + "logs": { + "loss": 0.2535, + "epoch": 7.818853974121996, + "learning_rate": 4.640736225272349e-06 + } + }, + { + "step": 21200, + "logs": { + "loss": 0.2508, + "epoch": 7.837338262476894, + "learning_rate": 4.60140795217682e-06 + } + }, + { + "step": 21250, + "logs": { + "loss": 0.254, + "epoch": 7.855822550831793, + "learning_rate": 4.562079679081292e-06 + } + }, + { + "step": 21300, + "logs": { + "loss": 0.251, + "epoch": 7.8743068391866915, + "learning_rate": 4.522751405985763e-06 + } + }, + { + "step": 21350, + "logs": { + "loss": 0.2545, + "epoch": 7.89279112754159, + "learning_rate": 4.483423132890235e-06 + } + }, + { + "step": 21400, + "logs": { + "loss": 0.2497, + "epoch": 7.911275415896488, + "learning_rate": 4.444094859794707e-06 + } + }, + { + "step": 21450, + "logs": { + "loss": 0.253, + "epoch": 7.929759704251387, + "learning_rate": 4.4047665866991786e-06 + } + }, + { + "step": 21500, + "logs": { + "loss": 0.249, + "epoch": 7.948243992606285, + "learning_rate": 4.36543831360365e-06 + } + }, + { + "step": 21550, + "logs": { + "loss": 0.2534, + "epoch": 7.966728280961183, + "learning_rate": 4.3261100405081215e-06 + } + }, + { + "step": 21600, + "logs": { + "loss": 0.2563, + "epoch": 7.985212569316081, + "learning_rate": 4.286781767412593e-06 + } + }, + { + "step": 21650, + "logs": { + "loss": 0.2489, + "epoch": 8.00369685767098, + "learning_rate": 4.247453494317064e-06 + } + }, + { + "step": 21700, + "logs": { + "loss": 0.2079, + "epoch": 8.022181146025877, + "learning_rate": 4.208125221221537e-06 + } + }, + { + "step": 21750, + "logs": { + "loss": 0.2079, + "epoch": 8.040665434380776, + "learning_rate": 4.168796948126008e-06 + } + }, + { + "step": 21800, + "logs": { + "loss": 0.2055, + "epoch": 8.059149722735675, + "learning_rate": 4.12946867503048e-06 + } + }, + { + "step": 21850, + "logs": { + "loss": 0.2115, + "epoch": 8.077634011090574, + "learning_rate": 4.090140401934951e-06 + } + }, + { + "step": 21900, + "logs": { + "loss": 0.2105, + "epoch": 8.096118299445472, + "learning_rate": 4.050812128839423e-06 + } + }, + { + "step": 21950, + "logs": { + "loss": 0.2055, + "epoch": 8.11460258780037, + "learning_rate": 4.011483855743895e-06 + } + }, + { + "step": 22000, + "logs": { + "loss": 0.2086, + "epoch": 8.133086876155268, + "learning_rate": 3.972155582648366e-06 + } + }, + { + "step": 22050, + "logs": { + "loss": 0.2102, + "epoch": 8.151571164510166, + "learning_rate": 3.932827309552838e-06 + } + }, + { + "step": 22100, + "logs": { + "loss": 0.21, + "epoch": 8.170055452865064, + "learning_rate": 3.893499036457309e-06 + } + }, + { + "step": 22150, + "logs": { + "loss": 0.2081, + "epoch": 8.188539741219962, + "learning_rate": 3.854170763361781e-06 + } + }, + { + "step": 22200, + "logs": { + "loss": 0.213, + "epoch": 8.207024029574862, + "learning_rate": 3.814842490266253e-06 + } + }, + { + "step": 22250, + "logs": { + "loss": 0.2114, + "epoch": 8.22550831792976, + "learning_rate": 3.7755142171707245e-06 + } + }, + { + "step": 22300, + "logs": { + "loss": 0.2116, + "epoch": 8.243992606284658, + "learning_rate": 3.736185944075196e-06 + } + }, + { + "step": 22350, + "logs": { + "loss": 0.2106, + "epoch": 8.262476894639557, + "learning_rate": 3.696857670979668e-06 + } + }, + { + "step": 22400, + "logs": { + "loss": 0.2097, + "epoch": 8.280961182994455, + "learning_rate": 3.6575293978841393e-06 + } + }, + { + "step": 22450, + "logs": { + "loss": 0.2119, + "epoch": 8.299445471349353, + "learning_rate": 3.6182011247886107e-06 + } + }, + { + "step": 22500, + "logs": { + "loss": 0.2144, + "epoch": 8.317929759704251, + "learning_rate": 3.5788728516930826e-06 + } + }, + { + "step": 22550, + "logs": { + "loss": 0.2125, + "epoch": 8.336414048059149, + "learning_rate": 3.539544578597554e-06 + } + }, + { + "step": 22600, + "logs": { + "loss": 0.2134, + "epoch": 8.354898336414047, + "learning_rate": 3.5002163055020256e-06 + } + }, + { + "step": 22650, + "logs": { + "loss": 0.2127, + "epoch": 8.373382624768947, + "learning_rate": 3.4608880324064974e-06 + } + }, + { + "step": 22700, + "logs": { + "loss": 0.2111, + "epoch": 8.391866913123845, + "learning_rate": 3.421559759310969e-06 + } + }, + { + "step": 22750, + "logs": { + "loss": 0.213, + "epoch": 8.410351201478743, + "learning_rate": 3.3822314862154404e-06 + } + }, + { + "step": 22800, + "logs": { + "loss": 0.2109, + "epoch": 8.428835489833642, + "learning_rate": 3.3429032131199122e-06 + } + }, + { + "step": 22850, + "logs": { + "loss": 0.2093, + "epoch": 8.44731977818854, + "learning_rate": 3.3035749400243837e-06 + } + }, + { + "step": 22900, + "logs": { + "loss": 0.2126, + "epoch": 8.465804066543438, + "learning_rate": 3.264246666928855e-06 + } + }, + { + "step": 22950, + "logs": { + "loss": 0.2127, + "epoch": 8.484288354898336, + "learning_rate": 3.2249183938333275e-06 + } + }, + { + "step": 23000, + "logs": { + "loss": 0.2184, + "epoch": 8.502772643253234, + "learning_rate": 3.1855901207377985e-06 + } + }, + { + "step": 23050, + "logs": { + "loss": 0.2108, + "epoch": 8.521256931608134, + "learning_rate": 3.14626184764227e-06 + } + }, + { + "step": 23100, + "logs": { + "loss": 0.2123, + "epoch": 8.539741219963032, + "learning_rate": 3.1069335745467423e-06 + } + }, + { + "step": 23150, + "logs": { + "loss": 0.2125, + "epoch": 8.55822550831793, + "learning_rate": 3.0676053014512137e-06 + } + }, + { + "step": 23200, + "logs": { + "loss": 0.2124, + "epoch": 8.576709796672828, + "learning_rate": 3.028277028355685e-06 + } + }, + { + "step": 23250, + "logs": { + "loss": 0.2112, + "epoch": 8.595194085027726, + "learning_rate": 2.988948755260157e-06 + } + }, + { + "step": 23300, + "logs": { + "loss": 0.2103, + "epoch": 8.613678373382625, + "learning_rate": 2.9496204821646285e-06 + } + }, + { + "step": 23350, + "logs": { + "loss": 0.2128, + "epoch": 8.632162661737523, + "learning_rate": 2.9102922090691e-06 + } + }, + { + "step": 23400, + "logs": { + "loss": 0.2125, + "epoch": 8.65064695009242, + "learning_rate": 2.8709639359735715e-06 + } + }, + { + "step": 23450, + "logs": { + "loss": 0.2109, + "epoch": 8.669131238447319, + "learning_rate": 2.8316356628780434e-06 + } + }, + { + "step": 23500, + "logs": { + "loss": 0.2132, + "epoch": 8.687615526802219, + "learning_rate": 2.792307389782515e-06 + } + }, + { + "step": 23550, + "logs": { + "loss": 0.2113, + "epoch": 8.706099815157117, + "learning_rate": 2.7529791166869863e-06 + } + }, + { + "step": 23600, + "logs": { + "loss": 0.2109, + "epoch": 8.724584103512015, + "learning_rate": 2.713650843591458e-06 + } + }, + { + "step": 23650, + "logs": { + "loss": 0.2138, + "epoch": 8.743068391866913, + "learning_rate": 2.6743225704959296e-06 + } + }, + { + "step": 23700, + "logs": { + "loss": 0.2098, + "epoch": 8.761552680221811, + "learning_rate": 2.634994297400401e-06 + } + }, + { + "step": 23750, + "logs": { + "loss": 0.2128, + "epoch": 8.78003696857671, + "learning_rate": 2.595666024304873e-06 + } + }, + { + "step": 23800, + "logs": { + "loss": 0.2143, + "epoch": 8.798521256931608, + "learning_rate": 2.5563377512093444e-06 + } + }, + { + "step": 23850, + "logs": { + "loss": 0.2132, + "epoch": 8.817005545286506, + "learning_rate": 2.517009478113816e-06 + } + }, + { + "step": 23900, + "logs": { + "loss": 0.2099, + "epoch": 8.835489833641406, + "learning_rate": 2.4776812050182878e-06 + } + }, + { + "step": 23950, + "logs": { + "loss": 0.2124, + "epoch": 8.853974121996304, + "learning_rate": 2.4383529319227597e-06 + } + }, + { + "step": 24000, + "logs": { + "loss": 0.2122, + "epoch": 8.872458410351202, + "learning_rate": 2.399024658827231e-06 + } + }, + { + "step": 24050, + "logs": { + "loss": 0.2136, + "epoch": 8.8909426987061, + "learning_rate": 2.3596963857317026e-06 + } + }, + { + "step": 24100, + "logs": { + "loss": 0.2103, + "epoch": 8.909426987060998, + "learning_rate": 2.3203681126361745e-06 + } + }, + { + "step": 24150, + "logs": { + "loss": 0.2132, + "epoch": 8.927911275415896, + "learning_rate": 2.281039839540646e-06 + } + }, + { + "step": 24200, + "logs": { + "loss": 0.2082, + "epoch": 8.946395563770794, + "learning_rate": 2.2417115664451174e-06 + } + }, + { + "step": 24250, + "logs": { + "loss": 0.213, + "epoch": 8.964879852125692, + "learning_rate": 2.2023832933495893e-06 + } + }, + { + "step": 24300, + "logs": { + "loss": 0.2142, + "epoch": 8.98336414048059, + "learning_rate": 2.1630550202540607e-06 + } + }, + { + "step": 24350, + "logs": { + "loss": 0.2123, + "epoch": 9.00184842883549, + "learning_rate": 2.123726747158532e-06 + } + }, + { + "step": 24400, + "logs": { + "loss": 0.1854, + "epoch": 9.020332717190389, + "learning_rate": 2.084398474063004e-06 + } + }, + { + "step": 24450, + "logs": { + "loss": 0.1848, + "epoch": 9.038817005545287, + "learning_rate": 2.0450702009674755e-06 + } + }, + { + "step": 24500, + "logs": { + "loss": 0.1802, + "epoch": 9.057301293900185, + "learning_rate": 2.0057419278719474e-06 + } + }, + { + "step": 24550, + "logs": { + "loss": 0.1859, + "epoch": 9.075785582255083, + "learning_rate": 1.966413654776419e-06 + } + }, + { + "step": 24600, + "logs": { + "loss": 0.1849, + "epoch": 9.094269870609981, + "learning_rate": 1.9270853816808904e-06 + } + }, + { + "step": 24650, + "logs": { + "loss": 0.1841, + "epoch": 9.11275415896488, + "learning_rate": 1.8877571085853622e-06 + } + }, + { + "step": 24700, + "logs": { + "loss": 0.1854, + "epoch": 9.131238447319777, + "learning_rate": 1.848428835489834e-06 + } + }, + { + "step": 24750, + "logs": { + "loss": 0.1847, + "epoch": 9.149722735674677, + "learning_rate": 1.8091005623943054e-06 + } + }, + { + "step": 24800, + "logs": { + "loss": 0.1852, + "epoch": 9.168207024029575, + "learning_rate": 1.769772289298777e-06 + } + }, + { + "step": 24850, + "logs": { + "loss": 0.1846, + "epoch": 9.186691312384474, + "learning_rate": 1.7304440162032487e-06 + } + }, + { + "step": 24900, + "logs": { + "loss": 0.1848, + "epoch": 9.205175600739372, + "learning_rate": 1.6911157431077202e-06 + } + }, + { + "step": 24950, + "logs": { + "loss": 0.1861, + "epoch": 9.22365988909427, + "learning_rate": 1.6517874700121919e-06 + } + }, + { + "step": 25000, + "logs": { + "loss": 0.185, + "epoch": 9.242144177449168, + "learning_rate": 1.6124591969166637e-06 + } + }, + { + "step": 25050, + "logs": { + "loss": 0.1846, + "epoch": 9.260628465804066, + "learning_rate": 1.573130923821135e-06 + } + }, + { + "step": 25100, + "logs": { + "loss": 0.1827, + "epoch": 9.279112754158964, + "learning_rate": 1.5338026507256069e-06 + } + }, + { + "step": 25150, + "logs": { + "loss": 0.1871, + "epoch": 9.297597042513864, + "learning_rate": 1.4944743776300785e-06 + } + }, + { + "step": 25200, + "logs": { + "loss": 0.1865, + "epoch": 9.316081330868762, + "learning_rate": 1.45514610453455e-06 + } + }, + { + "step": 25250, + "logs": { + "loss": 0.1859, + "epoch": 9.33456561922366, + "learning_rate": 1.4158178314390217e-06 + } + }, + { + "step": 25300, + "logs": { + "loss": 0.1874, + "epoch": 9.353049907578558, + "learning_rate": 1.3764895583434931e-06 + } + }, + { + "step": 25350, + "logs": { + "loss": 0.1847, + "epoch": 9.371534195933457, + "learning_rate": 1.3371612852479648e-06 + } + }, + { + "step": 25400, + "logs": { + "loss": 0.1849, + "epoch": 9.390018484288355, + "learning_rate": 1.2978330121524365e-06 + } + }, + { + "step": 25450, + "logs": { + "loss": 0.1895, + "epoch": 9.408502772643253, + "learning_rate": 1.258504739056908e-06 + } + }, + { + "step": 25500, + "logs": { + "loss": 0.1829, + "epoch": 9.426987060998151, + "learning_rate": 1.2191764659613798e-06 + } + }, + { + "step": 25550, + "logs": { + "loss": 0.1837, + "epoch": 9.445471349353049, + "learning_rate": 1.1798481928658513e-06 + } + }, + { + "step": 25600, + "logs": { + "loss": 0.1864, + "epoch": 9.463955637707949, + "learning_rate": 1.140519919770323e-06 + } + }, + { + "step": 25650, + "logs": { + "loss": 0.1862, + "epoch": 9.482439926062847, + "learning_rate": 1.1011916466747946e-06 + } + }, + { + "step": 25700, + "logs": { + "loss": 0.19, + "epoch": 9.500924214417745, + "learning_rate": 1.061863373579266e-06 + } + }, + { + "step": 25750, + "logs": { + "loss": 0.1834, + "epoch": 9.519408502772643, + "learning_rate": 1.0225351004837378e-06 + } + }, + { + "step": 25800, + "logs": { + "loss": 0.1863, + "epoch": 9.537892791127542, + "learning_rate": 9.832068273882094e-07 + } + }, + { + "step": 25850, + "logs": { + "loss": 0.1859, + "epoch": 9.55637707948244, + "learning_rate": 9.438785542926811e-07 + } + }, + { + "step": 25900, + "logs": { + "loss": 0.1829, + "epoch": 9.574861367837338, + "learning_rate": 9.045502811971527e-07 + } + }, + { + "step": 25950, + "logs": { + "loss": 0.1838, + "epoch": 9.593345656192236, + "learning_rate": 8.652220081016244e-07 + } + }, + { + "step": 26000, + "logs": { + "loss": 0.1817, + "epoch": 9.611829944547136, + "learning_rate": 8.258937350060959e-07 + } + }, + { + "step": 26050, + "logs": { + "loss": 0.1834, + "epoch": 9.630314232902034, + "learning_rate": 7.865654619105675e-07 + } + }, + { + "step": 26100, + "logs": { + "loss": 0.1857, + "epoch": 9.648798521256932, + "learning_rate": 7.472371888150393e-07 + } + }, + { + "step": 26150, + "logs": { + "loss": 0.1858, + "epoch": 9.66728280961183, + "learning_rate": 7.079089157195108e-07 + } + }, + { + "step": 26200, + "logs": { + "loss": 0.1858, + "epoch": 9.685767097966728, + "learning_rate": 6.685806426239824e-07 + } + }, + { + "step": 26250, + "logs": { + "loss": 0.1848, + "epoch": 9.704251386321626, + "learning_rate": 6.29252369528454e-07 + } + }, + { + "step": 26300, + "logs": { + "loss": 0.1832, + "epoch": 9.722735674676525, + "learning_rate": 5.899240964329256e-07 + } + }, + { + "step": 26350, + "logs": { + "loss": 0.1842, + "epoch": 9.741219963031423, + "learning_rate": 5.505958233373973e-07 + } + }, + { + "step": 26400, + "logs": { + "loss": 0.1842, + "epoch": 9.75970425138632, + "learning_rate": 5.112675502418689e-07 + } + }, + { + "step": 26450, + "logs": { + "loss": 0.1847, + "epoch": 9.77818853974122, + "learning_rate": 4.7193927714634056e-07 + } + }, + { + "step": 26500, + "logs": { + "loss": 0.185, + "epoch": 9.796672828096119, + "learning_rate": 4.326110040508122e-07 + } + }, + { + "step": 26550, + "logs": { + "loss": 0.1826, + "epoch": 9.815157116451017, + "learning_rate": 3.9328273095528375e-07 + } + }, + { + "step": 26600, + "logs": { + "loss": 0.1833, + "epoch": 9.833641404805915, + "learning_rate": 3.539544578597554e-07 + } + }, + { + "step": 26650, + "logs": { + "loss": 0.1846, + "epoch": 9.852125693160813, + "learning_rate": 3.14626184764227e-07 + } + }, + { + "step": 26700, + "logs": { + "loss": 0.184, + "epoch": 9.870609981515711, + "learning_rate": 2.7529791166869866e-07 + } + }, + { + "step": 26750, + "logs": { + "loss": 0.1864, + "epoch": 9.88909426987061, + "learning_rate": 2.3596963857317028e-07 + } + }, + { + "step": 26800, + "logs": { + "loss": 0.1841, + "epoch": 9.907578558225508, + "learning_rate": 1.9664136547764187e-07 + } + }, + { + "step": 26850, + "logs": { + "loss": 0.1855, + "epoch": 9.926062846580407, + "learning_rate": 1.573130923821135e-07 + } + }, + { + "step": 26900, + "logs": { + "loss": 0.1794, + "epoch": 9.944547134935306, + "learning_rate": 1.1798481928658514e-07 + } + }, + { + "step": 26950, + "logs": { + "loss": 0.1853, + "epoch": 9.963031423290204, + "learning_rate": 7.865654619105675e-08 + } + }, + { + "step": 27000, + "logs": { + "loss": 0.1855, + "epoch": 9.981515711645102, + "learning_rate": 3.9328273095528373e-08 + } + } + ], + "train_wrt_samples": [ + { + "step": 19200, + "logs": { + "loss": 1.2091, + "epoch": 0.018484288354898338, + "learning_rate": 1.0584432680980562e-05 + } + }, + { + "step": 38400, + "logs": { + "loss": 1.1315, + "epoch": 0.036968576709796676, + "learning_rate": 1.2459822860720352e-05 + } + }, + { + "step": 57600, + "logs": { + "loss": 1.1118, + "epoch": 0.05545286506469501, + "learning_rate": 1.355685579689696e-05 + } + }, + { + "step": 76800, + "logs": { + "loss": 1.1091, + "epoch": 0.07393715341959335, + "learning_rate": 1.4335213053988223e-05 + } + }, + { + "step": 96000, + "logs": { + "loss": 1.1063, + "epoch": 0.09242144177449169, + "learning_rate": 1.4938953847726847e-05 + } + }, + { + "step": 115200, + "logs": { + "loss": 1.1012, + "epoch": 0.11090573012939002, + "learning_rate": 1.5432245994674192e-05 + } + }, + { + "step": 134400, + "logs": { + "loss": 1.0844, + "epoch": 0.12939001848428835, + "learning_rate": 1.5849318562425815e-05 + } + }, + { + "step": 153600, + "logs": { + "loss": 1.0911, + "epoch": 0.1478743068391867, + "learning_rate": 1.6210603254020138e-05 + } + }, + { + "step": 172800, + "logs": { + "loss": 1.0816, + "epoch": 0.16635859519408502, + "learning_rate": 1.652927893686328e-05 + } + }, + { + "step": 192000, + "logs": { + "loss": 1.0819, + "epoch": 0.18484288354898337, + "learning_rate": 1.6814344049111566e-05 + } + }, + { + "step": 211200, + "logs": { + "loss": 1.0757, + "epoch": 0.2033271719038817, + "learning_rate": 1.70722168104602e-05 + } + }, + { + "step": 230400, + "logs": { + "loss": 1.0706, + "epoch": 0.22181146025878004, + "learning_rate": 1.7307636196960783e-05 + } + }, + { + "step": 249600, + "logs": { + "loss": 1.0698, + "epoch": 0.24029574861367836, + "learning_rate": 1.7524201039271304e-05 + } + }, + { + "step": 268800, + "logs": { + "loss": 1.0632, + "epoch": 0.2587800369685767, + "learning_rate": 1.7724708765356604e-05 + } + }, + { + "step": 288000, + "logs": { + "loss": 1.0601, + "epoch": 0.27726432532347506, + "learning_rate": 1.7911376992503148e-05 + } + }, + { + "step": 307200, + "logs": { + "loss": 1.0582, + "epoch": 0.2957486136783734, + "learning_rate": 1.808599345743407e-05 + } + }, + { + "step": 326400, + "logs": { + "loss": 1.0512, + "epoch": 0.3142329020332717, + "learning_rate": 1.825002041317456e-05 + } + }, + { + "step": 345600, + "logs": { + "loss": 1.0544, + "epoch": 0.33271719038817005, + "learning_rate": 1.840466914065299e-05 + } + }, + { + "step": 364800, + "logs": { + "loss": 1.043, + "epoch": 0.3512014787430684, + "learning_rate": 1.855095428760744e-05 + } + }, + { + "step": 384000, + "logs": { + "loss": 1.047, + "epoch": 0.36968576709796674, + "learning_rate": 1.8689734253201903e-05 + } + }, + { + "step": 403200, + "logs": { + "loss": 1.0408, + "epoch": 0.38817005545286504, + "learning_rate": 1.8821741709263542e-05 + } + }, + { + "step": 422400, + "logs": { + "loss": 1.0431, + "epoch": 0.4066543438077634, + "learning_rate": 1.89476070147965e-05 + } + }, + { + "step": 441600, + "logs": { + "loss": 1.0431, + "epoch": 0.42513863216266173, + "learning_rate": 1.9067876421445012e-05 + } + }, + { + "step": 460800, + "logs": { + "loss": 1.034, + "epoch": 0.4436229205175601, + "learning_rate": 1.9183026401502057e-05 + } + }, + { + "step": 480000, + "logs": { + "loss": 1.039, + "epoch": 0.46210720887245843, + "learning_rate": 1.9293475049105016e-05 + } + }, + { + "step": 499200, + "logs": { + "loss": 1.0304, + "epoch": 0.4805914972273567, + "learning_rate": 1.9399591243986016e-05 + } + }, + { + "step": 518400, + "logs": { + "loss": 1.0385, + "epoch": 0.49907578558225507, + "learning_rate": 1.950170208484624e-05 + } + }, + { + "step": 537600, + "logs": { + "loss": 1.027, + "epoch": 0.5175600739371534, + "learning_rate": 1.9600098970219974e-05 + } + }, + { + "step": 556800, + "logs": { + "loss": 1.0281, + "epoch": 0.5360443622920518, + "learning_rate": 1.9695042611797287e-05 + } + }, + { + "step": 576000, + "logs": { + "loss": 1.0282, + "epoch": 0.5545286506469501, + "learning_rate": 1.978676719749536e-05 + } + }, + { + "step": 595200, + "logs": { + "loss": 1.0186, + "epoch": 0.5730129390018485, + "learning_rate": 1.9875483871657987e-05 + } + }, + { + "step": 614400, + "logs": { + "loss": 1.0237, + "epoch": 0.5914972273567468, + "learning_rate": 1.9961383662539012e-05 + } + }, + { + "step": 633600, + "logs": { + "loss": 1.0176, + "epoch": 0.609981515711645, + "learning_rate": 1.9978762732528416e-05 + } + }, + { + "step": 652800, + "logs": { + "loss": 1.4539, + "epoch": 0.6284658040665434, + "learning_rate": 1.9939434459432887e-05 + } + }, + { + "step": 672000, + "logs": { + "loss": 1.2593, + "epoch": 0.6469500924214417, + "learning_rate": 1.990010618633736e-05 + } + }, + { + "step": 691200, + "logs": { + "loss": 1.5937, + "epoch": 0.6654343807763401, + "learning_rate": 1.986077791324183e-05 + } + }, + { + "step": 710400, + "logs": { + "loss": 1.0165, + "epoch": 0.6839186691312384, + "learning_rate": 1.9821449640146302e-05 + } + }, + { + "step": 729600, + "logs": { + "loss": 1.4354, + "epoch": 0.7024029574861368, + "learning_rate": 1.9782121367050776e-05 + } + }, + { + "step": 748800, + "logs": { + "loss": 1.021, + "epoch": 0.7208872458410351, + "learning_rate": 1.9742793093955247e-05 + } + }, + { + "step": 768000, + "logs": { + "loss": 1.0078, + "epoch": 0.7393715341959335, + "learning_rate": 1.9703464820859717e-05 + } + }, + { + "step": 787200, + "logs": { + "loss": 1.0057, + "epoch": 0.7578558225508318, + "learning_rate": 1.966413654776419e-05 + } + }, + { + "step": 806400, + "logs": { + "loss": 1.0001, + "epoch": 0.7763401109057301, + "learning_rate": 1.9624808274668662e-05 + } + }, + { + "step": 825600, + "logs": { + "loss": 1.006, + "epoch": 0.7948243992606284, + "learning_rate": 1.9585480001573133e-05 + } + }, + { + "step": 844800, + "logs": { + "loss": 1.0103, + "epoch": 0.8133086876155268, + "learning_rate": 1.9546151728477603e-05 + } + }, + { + "step": 864000, + "logs": { + "loss": 1.0046, + "epoch": 0.8317929759704251, + "learning_rate": 1.9506823455382077e-05 + } + }, + { + "step": 883200, + "logs": { + "loss": 0.9941, + "epoch": 0.8502772643253235, + "learning_rate": 1.9467495182286548e-05 + } + }, + { + "step": 902400, + "logs": { + "loss": 0.9963, + "epoch": 0.8687615526802218, + "learning_rate": 1.942816690919102e-05 + } + }, + { + "step": 921600, + "logs": { + "loss": 0.9926, + "epoch": 0.8872458410351202, + "learning_rate": 1.938883863609549e-05 + } + }, + { + "step": 940800, + "logs": { + "loss": 0.9918, + "epoch": 0.9057301293900185, + "learning_rate": 1.934951036299996e-05 + } + }, + { + "step": 960000, + "logs": { + "loss": 0.9899, + "epoch": 0.9242144177449169, + "learning_rate": 1.9310182089904434e-05 + } + }, + { + "step": 979200, + "logs": { + "loss": 0.9887, + "epoch": 0.9426987060998152, + "learning_rate": 1.9270853816808904e-05 + } + }, + { + "step": 998400, + "logs": { + "loss": 0.9942, + "epoch": 0.9611829944547134, + "learning_rate": 1.9231525543713375e-05 + } + }, + { + "step": 1017600, + "logs": { + "loss": 0.9817, + "epoch": 0.9796672828096118, + "learning_rate": 1.919219727061785e-05 + } + }, + { + "step": 1036800, + "logs": { + "loss": 0.9811, + "epoch": 0.9981515711645101, + "learning_rate": 1.915286899752232e-05 + } + }, + { + "step": 1055744, + "logs": { + "loss": 0.8466, + "epoch": 1.0166358595194085, + "learning_rate": 1.911354072442679e-05 + } + }, + { + "step": 1074944, + "logs": { + "loss": 0.8316, + "epoch": 1.0351201478743068, + "learning_rate": 1.9074212451331264e-05 + } + }, + { + "step": 1094144, + "logs": { + "loss": 0.8363, + "epoch": 1.0536044362292052, + "learning_rate": 1.9034884178235735e-05 + } + }, + { + "step": 1113344, + "logs": { + "loss": 0.8378, + "epoch": 1.0720887245841035, + "learning_rate": 1.8995555905140205e-05 + } + }, + { + "step": 1132544, + "logs": { + "loss": 0.8386, + "epoch": 1.0905730129390019, + "learning_rate": 1.895622763204468e-05 + } + }, + { + "step": 1151744, + "logs": { + "loss": 0.8465, + "epoch": 1.1090573012939002, + "learning_rate": 1.891689935894915e-05 + } + }, + { + "step": 1170944, + "logs": { + "loss": 0.8375, + "epoch": 1.1275415896487986, + "learning_rate": 1.887757108585362e-05 + } + }, + { + "step": 1190144, + "logs": { + "loss": 0.8379, + "epoch": 1.146025878003697, + "learning_rate": 1.8838242812758095e-05 + } + }, + { + "step": 1209344, + "logs": { + "loss": 0.836, + "epoch": 1.1645101663585953, + "learning_rate": 1.8798914539662565e-05 + } + }, + { + "step": 1228544, + "logs": { + "loss": 0.8411, + "epoch": 1.1829944547134936, + "learning_rate": 1.8759586266567036e-05 + } + }, + { + "step": 1247744, + "logs": { + "loss": 0.8442, + "epoch": 1.201478743068392, + "learning_rate": 1.872025799347151e-05 + } + }, + { + "step": 1266944, + "logs": { + "loss": 0.8397, + "epoch": 1.21996303142329, + "learning_rate": 1.868092972037598e-05 + } + }, + { + "step": 1286144, + "logs": { + "loss": 0.8415, + "epoch": 1.2384473197781884, + "learning_rate": 1.864160144728045e-05 + } + }, + { + "step": 1305344, + "logs": { + "loss": 0.8397, + "epoch": 1.2569316081330868, + "learning_rate": 1.8602273174184925e-05 + } + }, + { + "step": 1324544, + "logs": { + "loss": 0.8396, + "epoch": 1.2754158964879851, + "learning_rate": 1.8562944901089396e-05 + } + }, + { + "step": 1343744, + "logs": { + "loss": 0.8452, + "epoch": 1.2939001848428835, + "learning_rate": 1.8523616627993866e-05 + } + }, + { + "step": 1362944, + "logs": { + "loss": 0.8362, + "epoch": 1.3123844731977818, + "learning_rate": 1.848428835489834e-05 + } + }, + { + "step": 1382144, + "logs": { + "loss": 0.8396, + "epoch": 1.3308687615526802, + "learning_rate": 1.844496008180281e-05 + } + }, + { + "step": 1401344, + "logs": { + "loss": 0.8334, + "epoch": 1.3493530499075785, + "learning_rate": 1.840563180870728e-05 + } + }, + { + "step": 1420544, + "logs": { + "loss": 0.834, + "epoch": 1.3678373382624769, + "learning_rate": 1.8366303535611752e-05 + } + }, + { + "step": 1439744, + "logs": { + "loss": 0.8333, + "epoch": 1.3863216266173752, + "learning_rate": 1.8326975262516223e-05 + } + }, + { + "step": 1458944, + "logs": { + "loss": 0.8357, + "epoch": 1.4048059149722736, + "learning_rate": 1.8287646989420697e-05 + } + }, + { + "step": 1478144, + "logs": { + "loss": 0.8374, + "epoch": 1.423290203327172, + "learning_rate": 1.8248318716325167e-05 + } + }, + { + "step": 1497344, + "logs": { + "loss": 0.837, + "epoch": 1.4417744916820703, + "learning_rate": 1.8208990443229638e-05 + } + }, + { + "step": 1516544, + "logs": { + "loss": 0.835, + "epoch": 1.4602587800369686, + "learning_rate": 1.816966217013411e-05 + } + }, + { + "step": 1535744, + "logs": { + "loss": 0.8352, + "epoch": 1.478743068391867, + "learning_rate": 1.8130333897038583e-05 + } + }, + { + "step": 1554944, + "logs": { + "loss": 0.8475, + "epoch": 1.4972273567467653, + "learning_rate": 1.8091005623943053e-05 + } + }, + { + "step": 1574144, + "logs": { + "loss": 0.8339, + "epoch": 1.5157116451016637, + "learning_rate": 1.8051677350847524e-05 + } + }, + { + "step": 1593344, + "logs": { + "loss": 0.8382, + "epoch": 1.534195933456562, + "learning_rate": 1.8012349077751998e-05 + } + }, + { + "step": 1612544, + "logs": { + "loss": 0.8376, + "epoch": 1.5526802218114601, + "learning_rate": 1.797302080465647e-05 + } + }, + { + "step": 1631744, + "logs": { + "loss": 0.8302, + "epoch": 1.5711645101663585, + "learning_rate": 1.793369253156094e-05 + } + }, + { + "step": 1650944, + "logs": { + "loss": 0.8312, + "epoch": 1.5896487985212568, + "learning_rate": 1.7894364258465413e-05 + } + }, + { + "step": 1670144, + "logs": { + "loss": 0.8361, + "epoch": 1.6081330868761552, + "learning_rate": 1.7855035985369884e-05 + } + }, + { + "step": 1689344, + "logs": { + "loss": 0.8321, + "epoch": 1.6266173752310535, + "learning_rate": 1.7815707712274354e-05 + } + }, + { + "step": 1708544, + "logs": { + "loss": 0.8374, + "epoch": 1.645101663585952, + "learning_rate": 1.777637943917883e-05 + } + }, + { + "step": 1727744, + "logs": { + "loss": 0.8287, + "epoch": 1.6635859519408502, + "learning_rate": 1.77370511660833e-05 + } + }, + { + "step": 1746944, + "logs": { + "loss": 0.8367, + "epoch": 1.6820702402957486, + "learning_rate": 1.769772289298777e-05 + } + }, + { + "step": 1766144, + "logs": { + "loss": 0.8439, + "epoch": 1.700554528650647, + "learning_rate": 1.7658394619892244e-05 + } + }, + { + "step": 1785344, + "logs": { + "loss": 0.829, + "epoch": 1.7190388170055453, + "learning_rate": 1.7619066346796714e-05 + } + }, + { + "step": 1804544, + "logs": { + "loss": 0.8325, + "epoch": 1.7375231053604436, + "learning_rate": 1.7579738073701185e-05 + } + }, + { + "step": 1823744, + "logs": { + "loss": 0.837, + "epoch": 1.756007393715342, + "learning_rate": 1.754040980060566e-05 + } + }, + { + "step": 1842944, + "logs": { + "loss": 0.8371, + "epoch": 1.7744916820702403, + "learning_rate": 1.750108152751013e-05 + } + }, + { + "step": 1862144, + "logs": { + "loss": 0.832, + "epoch": 1.7929759704251387, + "learning_rate": 1.74617532544146e-05 + } + }, + { + "step": 1881344, + "logs": { + "loss": 0.8342, + "epoch": 1.811460258780037, + "learning_rate": 1.7422424981319074e-05 + } + }, + { + "step": 1900544, + "logs": { + "loss": 0.8331, + "epoch": 1.8299445471349354, + "learning_rate": 1.7383096708223545e-05 + } + }, + { + "step": 1919744, + "logs": { + "loss": 0.8234, + "epoch": 1.8484288354898337, + "learning_rate": 1.7343768435128015e-05 + } + }, + { + "step": 1938944, + "logs": { + "loss": 0.8311, + "epoch": 1.866913123844732, + "learning_rate": 1.7304440162032486e-05 + } + }, + { + "step": 1958144, + "logs": { + "loss": 0.8259, + "epoch": 1.8853974121996304, + "learning_rate": 1.726511188893696e-05 + } + }, + { + "step": 1977344, + "logs": { + "loss": 0.8253, + "epoch": 1.9038817005545285, + "learning_rate": 1.722578361584143e-05 + } + }, + { + "step": 1996544, + "logs": { + "loss": 0.8297, + "epoch": 1.922365988909427, + "learning_rate": 1.71864553427459e-05 + } + }, + { + "step": 2015744, + "logs": { + "loss": 0.8264, + "epoch": 1.9408502772643252, + "learning_rate": 1.7147127069650372e-05 + } + }, + { + "step": 2034944, + "logs": { + "loss": 0.8283, + "epoch": 1.9593345656192236, + "learning_rate": 1.7107798796554846e-05 + } + }, + { + "step": 2054144, + "logs": { + "loss": 0.8262, + "epoch": 1.977818853974122, + "learning_rate": 1.7068470523459316e-05 + } + }, + { + "step": 2073344, + "logs": { + "loss": 0.829, + "epoch": 1.9963031423290203, + "learning_rate": 1.7029142250363787e-05 + } + }, + { + "step": 2092288, + "logs": { + "loss": 0.6969, + "epoch": 2.014787430683919, + "learning_rate": 1.6989813977268258e-05 + } + }, + { + "step": 2111488, + "logs": { + "loss": 0.6715, + "epoch": 2.033271719038817, + "learning_rate": 1.695048570417273e-05 + } + }, + { + "step": 2130688, + "logs": { + "loss": 0.6636, + "epoch": 2.0517560073937156, + "learning_rate": 1.6911157431077202e-05 + } + }, + { + "step": 2149888, + "logs": { + "loss": 0.6649, + "epoch": 2.0702402957486137, + "learning_rate": 1.6871829157981673e-05 + } + }, + { + "step": 2169088, + "logs": { + "loss": 0.665, + "epoch": 2.088724584103512, + "learning_rate": 1.6832500884886147e-05 + } + }, + { + "step": 2188288, + "logs": { + "loss": 0.6746, + "epoch": 2.1072088724584104, + "learning_rate": 1.6793172611790617e-05 + } + }, + { + "step": 2207488, + "logs": { + "loss": 0.6678, + "epoch": 2.1256931608133085, + "learning_rate": 1.6753844338695088e-05 + } + }, + { + "step": 2226688, + "logs": { + "loss": 0.6739, + "epoch": 2.144177449168207, + "learning_rate": 1.6714516065599562e-05 + } + }, + { + "step": 2245888, + "logs": { + "loss": 0.6666, + "epoch": 2.162661737523105, + "learning_rate": 1.6675187792504033e-05 + } + }, + { + "step": 2265088, + "logs": { + "loss": 0.674, + "epoch": 2.1811460258780038, + "learning_rate": 1.6635859519408503e-05 + } + }, + { + "step": 2284288, + "logs": { + "loss": 0.6737, + "epoch": 2.199630314232902, + "learning_rate": 1.6596531246312977e-05 + } + }, + { + "step": 2303488, + "logs": { + "loss": 0.675, + "epoch": 2.2181146025878005, + "learning_rate": 1.6557202973217448e-05 + } + }, + { + "step": 2322688, + "logs": { + "loss": 0.6743, + "epoch": 2.2365988909426986, + "learning_rate": 1.651787470012192e-05 + } + }, + { + "step": 2341888, + "logs": { + "loss": 0.6718, + "epoch": 2.255083179297597, + "learning_rate": 1.647854642702639e-05 + } + }, + { + "step": 2361088, + "logs": { + "loss": 0.6735, + "epoch": 2.2735674676524953, + "learning_rate": 1.6439218153930863e-05 + } + }, + { + "step": 2380288, + "logs": { + "loss": 0.6798, + "epoch": 2.292051756007394, + "learning_rate": 1.6399889880835334e-05 + } + }, + { + "step": 2399488, + "logs": { + "loss": 0.6768, + "epoch": 2.310536044362292, + "learning_rate": 1.6360561607739804e-05 + } + }, + { + "step": 2418688, + "logs": { + "loss": 0.6796, + "epoch": 2.3290203327171906, + "learning_rate": 1.632123333464428e-05 + } + }, + { + "step": 2437888, + "logs": { + "loss": 0.6813, + "epoch": 2.3475046210720887, + "learning_rate": 1.628190506154875e-05 + } + }, + { + "step": 2457088, + "logs": { + "loss": 0.6774, + "epoch": 2.3659889094269873, + "learning_rate": 1.624257678845322e-05 + } + }, + { + "step": 2476288, + "logs": { + "loss": 0.6745, + "epoch": 2.3844731977818854, + "learning_rate": 1.6203248515357694e-05 + } + }, + { + "step": 2495488, + "logs": { + "loss": 0.6809, + "epoch": 2.402957486136784, + "learning_rate": 1.6163920242262164e-05 + } + }, + { + "step": 2514688, + "logs": { + "loss": 0.6831, + "epoch": 2.421441774491682, + "learning_rate": 1.6124591969166635e-05 + } + }, + { + "step": 2533888, + "logs": { + "loss": 0.6803, + "epoch": 2.43992606284658, + "learning_rate": 1.608526369607111e-05 + } + }, + { + "step": 2553088, + "logs": { + "loss": 0.6791, + "epoch": 2.4584103512014788, + "learning_rate": 1.604593542297558e-05 + } + }, + { + "step": 2572288, + "logs": { + "loss": 0.6812, + "epoch": 2.476894639556377, + "learning_rate": 1.600660714988005e-05 + } + }, + { + "step": 2591488, + "logs": { + "loss": 0.6866, + "epoch": 2.4953789279112755, + "learning_rate": 1.596727887678452e-05 + } + }, + { + "step": 2610688, + "logs": { + "loss": 0.6837, + "epoch": 2.5138632162661736, + "learning_rate": 1.5927950603688995e-05 + } + }, + { + "step": 2629888, + "logs": { + "loss": 0.6814, + "epoch": 2.532347504621072, + "learning_rate": 1.5888622330593465e-05 + } + }, + { + "step": 2649088, + "logs": { + "loss": 0.6845, + "epoch": 2.5508317929759703, + "learning_rate": 1.5849294057497936e-05 + } + }, + { + "step": 2668288, + "logs": { + "loss": 0.6832, + "epoch": 2.569316081330869, + "learning_rate": 1.5809965784402407e-05 + } + }, + { + "step": 2687488, + "logs": { + "loss": 0.6795, + "epoch": 2.587800369685767, + "learning_rate": 1.5770637511306877e-05 + } + }, + { + "step": 2706688, + "logs": { + "loss": 0.679, + "epoch": 2.6062846580406656, + "learning_rate": 1.573130923821135e-05 + } + }, + { + "step": 2725888, + "logs": { + "loss": 0.6851, + "epoch": 2.6247689463955637, + "learning_rate": 1.5691980965115822e-05 + } + }, + { + "step": 2745088, + "logs": { + "loss": 0.692, + "epoch": 2.6432532347504623, + "learning_rate": 1.5652652692020292e-05 + } + }, + { + "step": 2764288, + "logs": { + "loss": 0.6956, + "epoch": 2.6617375231053604, + "learning_rate": 1.5613324418924766e-05 + } + }, + { + "step": 2783488, + "logs": { + "loss": 0.6883, + "epoch": 2.680221811460259, + "learning_rate": 1.5573996145829237e-05 + } + }, + { + "step": 2802688, + "logs": { + "loss": 0.6806, + "epoch": 2.698706099815157, + "learning_rate": 1.5534667872733708e-05 + } + }, + { + "step": 2821888, + "logs": { + "loss": 0.6832, + "epoch": 2.7171903881700556, + "learning_rate": 1.549533959963818e-05 + } + }, + { + "step": 2841088, + "logs": { + "loss": 0.6871, + "epoch": 2.7356746765249538, + "learning_rate": 1.5456011326542652e-05 + } + }, + { + "step": 2860288, + "logs": { + "loss": 0.6843, + "epoch": 2.7541589648798523, + "learning_rate": 1.5416683053447123e-05 + } + }, + { + "step": 2879488, + "logs": { + "loss": 0.6869, + "epoch": 2.7726432532347505, + "learning_rate": 1.5377354780351597e-05 + } + }, + { + "step": 2898688, + "logs": { + "loss": 0.689, + "epoch": 2.7911275415896486, + "learning_rate": 1.5338026507256067e-05 + } + }, + { + "step": 2917888, + "logs": { + "loss": 0.6864, + "epoch": 2.809611829944547, + "learning_rate": 1.5298698234160538e-05 + } + }, + { + "step": 2937088, + "logs": { + "loss": 0.6875, + "epoch": 2.8280961182994453, + "learning_rate": 1.5259369961065012e-05 + } + }, + { + "step": 2956288, + "logs": { + "loss": 0.6844, + "epoch": 2.846580406654344, + "learning_rate": 1.5220041687969483e-05 + } + }, + { + "step": 2975488, + "logs": { + "loss": 0.6916, + "epoch": 2.865064695009242, + "learning_rate": 1.5180713414873953e-05 + } + }, + { + "step": 2994688, + "logs": { + "loss": 0.6849, + "epoch": 2.8835489833641406, + "learning_rate": 1.5141385141778426e-05 + } + }, + { + "step": 3013888, + "logs": { + "loss": 0.6835, + "epoch": 2.9020332717190387, + "learning_rate": 1.5102056868682898e-05 + } + }, + { + "step": 3033088, + "logs": { + "loss": 0.6866, + "epoch": 2.9205175600739373, + "learning_rate": 1.5062728595587369e-05 + } + }, + { + "step": 3052288, + "logs": { + "loss": 0.6863, + "epoch": 2.9390018484288354, + "learning_rate": 1.502340032249184e-05 + } + }, + { + "step": 3071488, + "logs": { + "loss": 0.6903, + "epoch": 2.957486136783734, + "learning_rate": 1.4984072049396311e-05 + } + }, + { + "step": 3090688, + "logs": { + "loss": 0.686, + "epoch": 2.975970425138632, + "learning_rate": 1.4944743776300784e-05 + } + }, + { + "step": 3109888, + "logs": { + "loss": 0.6909, + "epoch": 2.9944547134935307, + "learning_rate": 1.4905415503205256e-05 + } + }, + { + "step": 3128832, + "logs": { + "loss": 0.5832, + "epoch": 3.0129390018484288, + "learning_rate": 1.4866087230109727e-05 + } + }, + { + "step": 3148032, + "logs": { + "loss": 0.5314, + "epoch": 3.0314232902033273, + "learning_rate": 1.4826758957014197e-05 + } + }, + { + "step": 3167232, + "logs": { + "loss": 0.5332, + "epoch": 3.0499075785582255, + "learning_rate": 1.4787430683918671e-05 + } + }, + { + "step": 3186432, + "logs": { + "loss": 0.5359, + "epoch": 3.068391866913124, + "learning_rate": 1.4748102410823142e-05 + } + }, + { + "step": 3205632, + "logs": { + "loss": 0.5398, + "epoch": 3.086876155268022, + "learning_rate": 1.4708774137727613e-05 + } + }, + { + "step": 3224832, + "logs": { + "loss": 0.5433, + "epoch": 3.1053604436229203, + "learning_rate": 1.4669445864632087e-05 + } + }, + { + "step": 3244032, + "logs": { + "loss": 0.5368, + "epoch": 3.123844731977819, + "learning_rate": 1.4630117591536557e-05 + } + }, + { + "step": 3263232, + "logs": { + "loss": 0.5419, + "epoch": 3.142329020332717, + "learning_rate": 1.4590789318441028e-05 + } + }, + { + "step": 3282432, + "logs": { + "loss": 0.5398, + "epoch": 3.1608133086876156, + "learning_rate": 1.45514610453455e-05 + } + }, + { + "step": 3301632, + "logs": { + "loss": 0.5442, + "epoch": 3.1792975970425137, + "learning_rate": 1.4512132772249972e-05 + } + }, + { + "step": 3320832, + "logs": { + "loss": 0.544, + "epoch": 3.1977818853974123, + "learning_rate": 1.4472804499154443e-05 + } + }, + { + "step": 3340032, + "logs": { + "loss": 0.5455, + "epoch": 3.2162661737523104, + "learning_rate": 1.4433476226058915e-05 + } + }, + { + "step": 3359232, + "logs": { + "loss": 0.548, + "epoch": 3.234750462107209, + "learning_rate": 1.4394147952963386e-05 + } + }, + { + "step": 3378432, + "logs": { + "loss": 0.5453, + "epoch": 3.253234750462107, + "learning_rate": 1.4354819679867858e-05 + } + }, + { + "step": 3397632, + "logs": { + "loss": 0.5416, + "epoch": 3.2717190388170057, + "learning_rate": 1.431549140677233e-05 + } + }, + { + "step": 3416832, + "logs": { + "loss": 0.5542, + "epoch": 3.290203327171904, + "learning_rate": 1.4276163133676801e-05 + } + }, + { + "step": 3436032, + "logs": { + "loss": 0.5469, + "epoch": 3.3086876155268024, + "learning_rate": 1.4236834860581272e-05 + } + }, + { + "step": 3455232, + "logs": { + "loss": 0.5545, + "epoch": 3.3271719038817005, + "learning_rate": 1.4197506587485746e-05 + } + }, + { + "step": 3474432, + "logs": { + "loss": 0.5538, + "epoch": 3.345656192236599, + "learning_rate": 1.4158178314390216e-05 + } + }, + { + "step": 3493632, + "logs": { + "loss": 0.5504, + "epoch": 3.364140480591497, + "learning_rate": 1.4118850041294687e-05 + } + }, + { + "step": 3512832, + "logs": { + "loss": 0.5511, + "epoch": 3.3826247689463957, + "learning_rate": 1.4079521768199161e-05 + } + }, + { + "step": 3532032, + "logs": { + "loss": 0.554, + "epoch": 3.401109057301294, + "learning_rate": 1.4040193495103632e-05 + } + }, + { + "step": 3551232, + "logs": { + "loss": 0.5541, + "epoch": 3.4195933456561924, + "learning_rate": 1.4000865222008102e-05 + } + }, + { + "step": 3570432, + "logs": { + "loss": 0.5502, + "epoch": 3.4380776340110906, + "learning_rate": 1.3961536948912575e-05 + } + }, + { + "step": 3589632, + "logs": { + "loss": 0.5513, + "epoch": 3.4565619223659887, + "learning_rate": 1.3922208675817047e-05 + } + }, + { + "step": 3608832, + "logs": { + "loss": 0.5579, + "epoch": 3.4750462107208873, + "learning_rate": 1.3882880402721517e-05 + } + }, + { + "step": 3628032, + "logs": { + "loss": 0.5636, + "epoch": 3.4935304990757854, + "learning_rate": 1.384355212962599e-05 + } + }, + { + "step": 3647232, + "logs": { + "loss": 0.5617, + "epoch": 3.512014787430684, + "learning_rate": 1.380422385653046e-05 + } + }, + { + "step": 3666432, + "logs": { + "loss": 0.5549, + "epoch": 3.530499075785582, + "learning_rate": 1.3764895583434933e-05 + } + }, + { + "step": 3685632, + "logs": { + "loss": 0.5599, + "epoch": 3.5489833641404807, + "learning_rate": 1.3725567310339405e-05 + } + }, + { + "step": 3704832, + "logs": { + "loss": 0.5565, + "epoch": 3.567467652495379, + "learning_rate": 1.3686239037243876e-05 + } + }, + { + "step": 3724032, + "logs": { + "loss": 0.5565, + "epoch": 3.5859519408502774, + "learning_rate": 1.3646910764148346e-05 + } + }, + { + "step": 3743232, + "logs": { + "loss": 0.5558, + "epoch": 3.6044362292051755, + "learning_rate": 1.360758249105282e-05 + } + }, + { + "step": 3762432, + "logs": { + "loss": 0.5599, + "epoch": 3.622920517560074, + "learning_rate": 1.356825421795729e-05 + } + }, + { + "step": 3781632, + "logs": { + "loss": 0.561, + "epoch": 3.641404805914972, + "learning_rate": 1.3528925944861761e-05 + } + }, + { + "step": 3800832, + "logs": { + "loss": 0.5596, + "epoch": 3.6598890942698707, + "learning_rate": 1.3489597671766235e-05 + } + }, + { + "step": 3820032, + "logs": { + "loss": 0.5596, + "epoch": 3.678373382624769, + "learning_rate": 1.3450269398670706e-05 + } + }, + { + "step": 3839232, + "logs": { + "loss": 0.5579, + "epoch": 3.6968576709796674, + "learning_rate": 1.3410941125575177e-05 + } + }, + { + "step": 3858432, + "logs": { + "loss": 0.5611, + "epoch": 3.7153419593345656, + "learning_rate": 1.3371612852479649e-05 + } + }, + { + "step": 3877632, + "logs": { + "loss": 0.5627, + "epoch": 3.733826247689464, + "learning_rate": 1.3332284579384121e-05 + } + }, + { + "step": 3896832, + "logs": { + "loss": 0.5613, + "epoch": 3.7523105360443623, + "learning_rate": 1.3292956306288592e-05 + } + }, + { + "step": 3916032, + "logs": { + "loss": 0.5607, + "epoch": 3.770794824399261, + "learning_rate": 1.3253628033193064e-05 + } + }, + { + "step": 3935232, + "logs": { + "loss": 0.5651, + "epoch": 3.789279112754159, + "learning_rate": 1.3214299760097535e-05 + } + }, + { + "step": 3954432, + "logs": { + "loss": 0.5604, + "epoch": 3.807763401109057, + "learning_rate": 1.3174971487002005e-05 + } + }, + { + "step": 3973632, + "logs": { + "loss": 0.5676, + "epoch": 3.8262476894639557, + "learning_rate": 1.313564321390648e-05 + } + }, + { + "step": 3992832, + "logs": { + "loss": 0.5601, + "epoch": 3.844731977818854, + "learning_rate": 1.309631494081095e-05 + } + }, + { + "step": 4012032, + "logs": { + "loss": 0.5645, + "epoch": 3.8632162661737524, + "learning_rate": 1.305698666771542e-05 + } + }, + { + "step": 4031232, + "logs": { + "loss": 0.5653, + "epoch": 3.8817005545286505, + "learning_rate": 1.3017658394619895e-05 + } + }, + { + "step": 4050432, + "logs": { + "loss": 0.5618, + "epoch": 3.900184842883549, + "learning_rate": 1.2978330121524365e-05 + } + }, + { + "step": 4069632, + "logs": { + "loss": 0.5624, + "epoch": 3.918669131238447, + "learning_rate": 1.2939001848428836e-05 + } + }, + { + "step": 4088832, + "logs": { + "loss": 0.5621, + "epoch": 3.9371534195933457, + "learning_rate": 1.289967357533331e-05 + } + }, + { + "step": 4108032, + "logs": { + "loss": 0.5633, + "epoch": 3.955637707948244, + "learning_rate": 1.286034530223778e-05 + } + }, + { + "step": 4127232, + "logs": { + "loss": 0.5645, + "epoch": 3.9741219963031424, + "learning_rate": 1.2821017029142251e-05 + } + }, + { + "step": 4146432, + "logs": { + "loss": 0.5675, + "epoch": 3.9926062846580406, + "learning_rate": 1.2781688756046723e-05 + } + }, + { + "step": 4165376, + "logs": { + "loss": 0.4943, + "epoch": 4.011090573012939, + "learning_rate": 1.2742360482951194e-05 + } + }, + { + "step": 4184576, + "logs": { + "loss": 0.4282, + "epoch": 4.029574861367838, + "learning_rate": 1.2703032209855666e-05 + } + }, + { + "step": 4203776, + "logs": { + "loss": 0.4285, + "epoch": 4.048059149722736, + "learning_rate": 1.2663703936760139e-05 + } + }, + { + "step": 4222976, + "logs": { + "loss": 0.4307, + "epoch": 4.066543438077634, + "learning_rate": 1.262437566366461e-05 + } + }, + { + "step": 4242176, + "logs": { + "loss": 0.4372, + "epoch": 4.085027726432532, + "learning_rate": 1.258504739056908e-05 + } + }, + { + "step": 4261376, + "logs": { + "loss": 0.4377, + "epoch": 4.103512014787431, + "learning_rate": 1.2545719117473554e-05 + } + }, + { + "step": 4280576, + "logs": { + "loss": 0.4296, + "epoch": 4.121996303142329, + "learning_rate": 1.2506390844378025e-05 + } + }, + { + "step": 4299776, + "logs": { + "loss": 0.4358, + "epoch": 4.140480591497227, + "learning_rate": 1.2467062571282495e-05 + } + }, + { + "step": 4318976, + "logs": { + "loss": 0.4383, + "epoch": 4.1589648798521255, + "learning_rate": 1.2427734298186969e-05 + } + }, + { + "step": 4338176, + "logs": { + "loss": 0.4355, + "epoch": 4.177449168207024, + "learning_rate": 1.238840602509144e-05 + } + }, + { + "step": 4357376, + "logs": { + "loss": 0.44, + "epoch": 4.195933456561923, + "learning_rate": 1.234907775199591e-05 + } + }, + { + "step": 4376576, + "logs": { + "loss": 0.4393, + "epoch": 4.214417744916821, + "learning_rate": 1.2309749478900384e-05 + } + }, + { + "step": 4395776, + "logs": { + "loss": 0.4427, + "epoch": 4.232902033271719, + "learning_rate": 1.2270421205804855e-05 + } + }, + { + "step": 4414976, + "logs": { + "loss": 0.4399, + "epoch": 4.251386321626617, + "learning_rate": 1.2231092932709326e-05 + } + }, + { + "step": 4434176, + "logs": { + "loss": 0.4401, + "epoch": 4.269870609981516, + "learning_rate": 1.2191764659613798e-05 + } + }, + { + "step": 4453376, + "logs": { + "loss": 0.4446, + "epoch": 4.288354898336414, + "learning_rate": 1.2152436386518269e-05 + } + }, + { + "step": 4472576, + "logs": { + "loss": 0.4436, + "epoch": 4.306839186691312, + "learning_rate": 1.211310811342274e-05 + } + }, + { + "step": 4491776, + "logs": { + "loss": 0.4478, + "epoch": 4.32532347504621, + "learning_rate": 1.2073779840327213e-05 + } + }, + { + "step": 4510976, + "logs": { + "loss": 0.447, + "epoch": 4.343807763401109, + "learning_rate": 1.2034451567231684e-05 + } + }, + { + "step": 4530176, + "logs": { + "loss": 0.4505, + "epoch": 4.3622920517560075, + "learning_rate": 1.1995123294136154e-05 + } + }, + { + "step": 4549376, + "logs": { + "loss": 0.4461, + "epoch": 4.380776340110906, + "learning_rate": 1.1955795021040628e-05 + } + }, + { + "step": 4568576, + "logs": { + "loss": 0.4511, + "epoch": 4.399260628465804, + "learning_rate": 1.1916466747945099e-05 + } + }, + { + "step": 4587776, + "logs": { + "loss": 0.4481, + "epoch": 4.417744916820703, + "learning_rate": 1.187713847484957e-05 + } + }, + { + "step": 4606976, + "logs": { + "loss": 0.4457, + "epoch": 4.436229205175601, + "learning_rate": 1.1837810201754044e-05 + } + }, + { + "step": 4626176, + "logs": { + "loss": 0.4491, + "epoch": 4.454713493530499, + "learning_rate": 1.1798481928658514e-05 + } + }, + { + "step": 4645376, + "logs": { + "loss": 0.453, + "epoch": 4.473197781885397, + "learning_rate": 1.1759153655562985e-05 + } + }, + { + "step": 4664576, + "logs": { + "loss": 0.4557, + "epoch": 4.491682070240295, + "learning_rate": 1.1719825382467455e-05 + } + }, + { + "step": 4683776, + "logs": { + "loss": 0.4588, + "epoch": 4.510166358595194, + "learning_rate": 1.168049710937193e-05 + } + }, + { + "step": 4702976, + "logs": { + "loss": 0.4478, + "epoch": 4.5286506469500925, + "learning_rate": 1.16411688362764e-05 + } + }, + { + "step": 4722176, + "logs": { + "loss": 0.4558, + "epoch": 4.547134935304991, + "learning_rate": 1.160184056318087e-05 + } + }, + { + "step": 4741376, + "logs": { + "loss": 0.453, + "epoch": 4.565619223659889, + "learning_rate": 1.1562512290085343e-05 + } + }, + { + "step": 4760576, + "logs": { + "loss": 0.4566, + "epoch": 4.584103512014788, + "learning_rate": 1.1523184016989815e-05 + } + }, + { + "step": 4779776, + "logs": { + "loss": 0.4512, + "epoch": 4.602587800369686, + "learning_rate": 1.1483855743894286e-05 + } + }, + { + "step": 4798976, + "logs": { + "loss": 0.4537, + "epoch": 4.621072088724584, + "learning_rate": 1.1444527470798758e-05 + } + }, + { + "step": 4818176, + "logs": { + "loss": 0.4596, + "epoch": 4.639556377079482, + "learning_rate": 1.1405199197703229e-05 + } + }, + { + "step": 4837376, + "logs": { + "loss": 0.4544, + "epoch": 4.658040665434381, + "learning_rate": 1.1365870924607701e-05 + } + }, + { + "step": 4856576, + "logs": { + "loss": 0.4552, + "epoch": 4.676524953789279, + "learning_rate": 1.1326542651512173e-05 + } + }, + { + "step": 4875776, + "logs": { + "loss": 0.4561, + "epoch": 4.695009242144177, + "learning_rate": 1.1287214378416644e-05 + } + }, + { + "step": 4894976, + "logs": { + "loss": 0.456, + "epoch": 4.7134935304990755, + "learning_rate": 1.1247886105321115e-05 + } + }, + { + "step": 4914176, + "logs": { + "loss": 0.455, + "epoch": 4.7319778188539745, + "learning_rate": 1.1208557832225589e-05 + } + }, + { + "step": 4933376, + "logs": { + "loss": 0.4579, + "epoch": 4.750462107208873, + "learning_rate": 1.116922955913006e-05 + } + }, + { + "step": 4952576, + "logs": { + "loss": 0.4569, + "epoch": 4.768946395563771, + "learning_rate": 1.112990128603453e-05 + } + }, + { + "step": 4971776, + "logs": { + "loss": 0.459, + "epoch": 4.787430683918669, + "learning_rate": 1.1090573012939004e-05 + } + }, + { + "step": 4990976, + "logs": { + "loss": 0.456, + "epoch": 4.805914972273568, + "learning_rate": 1.1051244739843475e-05 + } + }, + { + "step": 5010176, + "logs": { + "loss": 0.4605, + "epoch": 4.824399260628466, + "learning_rate": 1.1011916466747945e-05 + } + }, + { + "step": 5029376, + "logs": { + "loss": 0.4565, + "epoch": 4.842883548983364, + "learning_rate": 1.0972588193652417e-05 + } + }, + { + "step": 5048576, + "logs": { + "loss": 0.4606, + "epoch": 4.861367837338262, + "learning_rate": 1.093325992055689e-05 + } + }, + { + "step": 5067776, + "logs": { + "loss": 0.4586, + "epoch": 4.87985212569316, + "learning_rate": 1.089393164746136e-05 + } + }, + { + "step": 5086976, + "logs": { + "loss": 0.4605, + "epoch": 4.898336414048059, + "learning_rate": 1.0854603374365833e-05 + } + }, + { + "step": 5106176, + "logs": { + "loss": 0.4565, + "epoch": 4.9168207024029575, + "learning_rate": 1.0815275101270303e-05 + } + }, + { + "step": 5125376, + "logs": { + "loss": 0.4585, + "epoch": 4.935304990757856, + "learning_rate": 1.0775946828174776e-05 + } + }, + { + "step": 5144576, + "logs": { + "loss": 0.4588, + "epoch": 4.953789279112754, + "learning_rate": 1.0736618555079248e-05 + } + }, + { + "step": 5163776, + "logs": { + "loss": 0.4639, + "epoch": 4.972273567467653, + "learning_rate": 1.0697290281983718e-05 + } + }, + { + "step": 5182976, + "logs": { + "loss": 0.4666, + "epoch": 4.990757855822551, + "learning_rate": 1.0657962008888189e-05 + } + }, + { + "step": 5201920, + "logs": { + "loss": 0.4109, + "epoch": 5.009242144177449, + "learning_rate": 1.0618633735792663e-05 + } + }, + { + "step": 5221120, + "logs": { + "loss": 0.3538, + "epoch": 5.027726432532347, + "learning_rate": 1.0579305462697134e-05 + } + }, + { + "step": 5240320, + "logs": { + "loss": 0.346, + "epoch": 5.046210720887246, + "learning_rate": 1.0539977189601604e-05 + } + }, + { + "step": 5259520, + "logs": { + "loss": 0.3483, + "epoch": 5.064695009242144, + "learning_rate": 1.0500648916506078e-05 + } + }, + { + "step": 5278720, + "logs": { + "loss": 0.3548, + "epoch": 5.0831792975970425, + "learning_rate": 1.0461320643410549e-05 + } + }, + { + "step": 5297920, + "logs": { + "loss": 0.3567, + "epoch": 5.101663585951941, + "learning_rate": 1.042199237031502e-05 + } + }, + { + "step": 5317120, + "logs": { + "loss": 0.3525, + "epoch": 5.12014787430684, + "learning_rate": 1.0382664097219492e-05 + } + }, + { + "step": 5336320, + "logs": { + "loss": 0.3563, + "epoch": 5.138632162661738, + "learning_rate": 1.0343335824123964e-05 + } + }, + { + "step": 5355520, + "logs": { + "loss": 0.3606, + "epoch": 5.157116451016636, + "learning_rate": 1.0304007551028435e-05 + } + }, + { + "step": 5374720, + "logs": { + "loss": 0.3585, + "epoch": 5.175600739371534, + "learning_rate": 1.0264679277932907e-05 + } + }, + { + "step": 5393920, + "logs": { + "loss": 0.3547, + "epoch": 5.194085027726432, + "learning_rate": 1.0225351004837378e-05 + } + }, + { + "step": 5413120, + "logs": { + "loss": 0.359, + "epoch": 5.212569316081331, + "learning_rate": 1.018602273174185e-05 + } + }, + { + "step": 5432320, + "logs": { + "loss": 0.3616, + "epoch": 5.231053604436229, + "learning_rate": 1.0146694458646322e-05 + } + }, + { + "step": 5451520, + "logs": { + "loss": 0.3602, + "epoch": 5.249537892791127, + "learning_rate": 1.0107366185550793e-05 + } + }, + { + "step": 5470720, + "logs": { + "loss": 0.3607, + "epoch": 5.2680221811460255, + "learning_rate": 1.0068037912455264e-05 + } + }, + { + "step": 5489920, + "logs": { + "loss": 0.3642, + "epoch": 5.2865064695009245, + "learning_rate": 1.0028709639359738e-05 + } + }, + { + "step": 5509120, + "logs": { + "loss": 0.3616, + "epoch": 5.304990757855823, + "learning_rate": 9.989381366264208e-06 + } + }, + { + "step": 5528320, + "logs": { + "loss": 0.364, + "epoch": 5.323475046210721, + "learning_rate": 9.95005309316868e-06 + } + }, + { + "step": 5547520, + "logs": { + "loss": 0.3656, + "epoch": 5.341959334565619, + "learning_rate": 9.910724820073151e-06 + } + }, + { + "step": 5566720, + "logs": { + "loss": 0.3613, + "epoch": 5.360443622920518, + "learning_rate": 9.871396546977623e-06 + } + }, + { + "step": 5585920, + "logs": { + "loss": 0.3655, + "epoch": 5.378927911275416, + "learning_rate": 9.832068273882096e-06 + } + }, + { + "step": 5605120, + "logs": { + "loss": 0.3677, + "epoch": 5.397412199630314, + "learning_rate": 9.792740000786566e-06 + } + }, + { + "step": 5624320, + "logs": { + "loss": 0.3678, + "epoch": 5.415896487985212, + "learning_rate": 9.753411727691039e-06 + } + }, + { + "step": 5643520, + "logs": { + "loss": 0.364, + "epoch": 5.434380776340111, + "learning_rate": 9.71408345459551e-06 + } + }, + { + "step": 5662720, + "logs": { + "loss": 0.3654, + "epoch": 5.452865064695009, + "learning_rate": 9.67475518149998e-06 + } + }, + { + "step": 5681920, + "logs": { + "loss": 0.3671, + "epoch": 5.4713493530499075, + "learning_rate": 9.635426908404452e-06 + } + }, + { + "step": 5701120, + "logs": { + "loss": 0.3724, + "epoch": 5.489833641404806, + "learning_rate": 9.596098635308924e-06 + } + }, + { + "step": 5720320, + "logs": { + "loss": 1.4537, + "epoch": 5.508317929759705, + "learning_rate": 9.556770362213395e-06 + } + }, + { + "step": 5739520, + "logs": { + "loss": 0.3685, + "epoch": 5.526802218114603, + "learning_rate": 9.517442089117867e-06 + } + }, + { + "step": 5758720, + "logs": { + "loss": 0.3688, + "epoch": 5.545286506469501, + "learning_rate": 9.47811381602234e-06 + } + }, + { + "step": 5777920, + "logs": { + "loss": 0.3726, + "epoch": 5.563770794824399, + "learning_rate": 9.43878554292681e-06 + } + }, + { + "step": 5797120, + "logs": { + "loss": 0.3718, + "epoch": 5.582255083179297, + "learning_rate": 9.399457269831283e-06 + } + }, + { + "step": 5816320, + "logs": { + "loss": 0.3691, + "epoch": 5.600739371534196, + "learning_rate": 9.360128996735755e-06 + } + }, + { + "step": 5835520, + "logs": { + "loss": 0.3684, + "epoch": 5.619223659889094, + "learning_rate": 9.320800723640226e-06 + } + }, + { + "step": 5854720, + "logs": { + "loss": 0.3747, + "epoch": 5.6377079482439925, + "learning_rate": 9.281472450544698e-06 + } + }, + { + "step": 5873920, + "logs": { + "loss": 0.3727, + "epoch": 5.656192236598891, + "learning_rate": 9.24214417744917e-06 + } + }, + { + "step": 5893120, + "logs": { + "loss": 0.3736, + "epoch": 5.67467652495379, + "learning_rate": 9.20281590435364e-06 + } + }, + { + "step": 5912320, + "logs": { + "loss": 0.3725, + "epoch": 5.693160813308688, + "learning_rate": 9.163487631258111e-06 + } + }, + { + "step": 5931520, + "logs": { + "loss": 0.3711, + "epoch": 5.711645101663586, + "learning_rate": 9.124159358162584e-06 + } + }, + { + "step": 5950720, + "logs": { + "loss": 0.3705, + "epoch": 5.730129390018484, + "learning_rate": 9.084831085067054e-06 + } + }, + { + "step": 5969920, + "logs": { + "loss": 0.3698, + "epoch": 5.748613678373383, + "learning_rate": 9.045502811971527e-06 + } + }, + { + "step": 5989120, + "logs": { + "loss": 0.3725, + "epoch": 5.767097966728281, + "learning_rate": 9.006174538875999e-06 + } + }, + { + "step": 6008320, + "logs": { + "loss": 0.3767, + "epoch": 5.785582255083179, + "learning_rate": 8.96684626578047e-06 + } + }, + { + "step": 6027520, + "logs": { + "loss": 0.3696, + "epoch": 5.804066543438077, + "learning_rate": 8.927517992684942e-06 + } + }, + { + "step": 6046720, + "logs": { + "loss": 0.3797, + "epoch": 5.822550831792976, + "learning_rate": 8.888189719589414e-06 + } + }, + { + "step": 6065920, + "logs": { + "loss": 0.371, + "epoch": 5.8410351201478745, + "learning_rate": 8.848861446493885e-06 + } + }, + { + "step": 6085120, + "logs": { + "loss": 0.3769, + "epoch": 5.859519408502773, + "learning_rate": 8.809533173398357e-06 + } + }, + { + "step": 6104320, + "logs": { + "loss": 0.3731, + "epoch": 5.878003696857671, + "learning_rate": 8.77020490030283e-06 + } + }, + { + "step": 6123520, + "logs": { + "loss": 0.3769, + "epoch": 5.896487985212569, + "learning_rate": 8.7308766272073e-06 + } + }, + { + "step": 6142720, + "logs": { + "loss": 0.3711, + "epoch": 5.914972273567468, + "learning_rate": 8.691548354111772e-06 + } + }, + { + "step": 6161920, + "logs": { + "loss": 0.3752, + "epoch": 5.933456561922366, + "learning_rate": 8.652220081016243e-06 + } + }, + { + "step": 6181120, + "logs": { + "loss": 0.3697, + "epoch": 5.951940850277264, + "learning_rate": 8.612891807920715e-06 + } + }, + { + "step": 6200320, + "logs": { + "loss": 0.3751, + "epoch": 5.970425138632162, + "learning_rate": 8.573563534825186e-06 + } + }, + { + "step": 6219520, + "logs": { + "loss": 0.3791, + "epoch": 5.988909426987061, + "learning_rate": 8.534235261729658e-06 + } + }, + { + "step": 6238464, + "logs": { + "loss": 0.3443, + "epoch": 6.007393715341959, + "learning_rate": 8.494906988634129e-06 + } + }, + { + "step": 6257664, + "logs": { + "loss": 0.2903, + "epoch": 6.0258780036968576, + "learning_rate": 8.455578715538601e-06 + } + }, + { + "step": 6276864, + "logs": { + "loss": 0.2906, + "epoch": 6.044362292051756, + "learning_rate": 8.416250442443073e-06 + } + }, + { + "step": 6296064, + "logs": { + "loss": 0.2884, + "epoch": 6.062846580406655, + "learning_rate": 8.376922169347544e-06 + } + }, + { + "step": 6315264, + "logs": { + "loss": 0.2913, + "epoch": 6.081330868761553, + "learning_rate": 8.337593896252016e-06 + } + }, + { + "step": 6334464, + "logs": { + "loss": 0.2964, + "epoch": 6.099815157116451, + "learning_rate": 8.298265623156489e-06 + } + }, + { + "step": 6353664, + "logs": { + "loss": 0.2877, + "epoch": 6.118299445471349, + "learning_rate": 8.25893735006096e-06 + } + }, + { + "step": 6372864, + "logs": { + "loss": 0.2957, + "epoch": 6.136783733826248, + "learning_rate": 8.219609076965432e-06 + } + }, + { + "step": 6392064, + "logs": { + "loss": 0.2974, + "epoch": 6.155268022181146, + "learning_rate": 8.180280803869902e-06 + } + }, + { + "step": 6411264, + "logs": { + "loss": 0.295, + "epoch": 6.173752310536044, + "learning_rate": 8.140952530774374e-06 + } + }, + { + "step": 6430464, + "logs": { + "loss": 0.2947, + "epoch": 6.1922365988909425, + "learning_rate": 8.101624257678847e-06 + } + }, + { + "step": 6449664, + "logs": { + "loss": 0.2957, + "epoch": 6.210720887245841, + "learning_rate": 8.062295984583317e-06 + } + }, + { + "step": 6468864, + "logs": { + "loss": 0.3001, + "epoch": 6.22920517560074, + "learning_rate": 8.02296771148779e-06 + } + }, + { + "step": 6488064, + "logs": { + "loss": 0.2958, + "epoch": 6.247689463955638, + "learning_rate": 7.98363943839226e-06 + } + }, + { + "step": 6507264, + "logs": { + "loss": 0.2956, + "epoch": 6.266173752310536, + "learning_rate": 7.944311165296733e-06 + } + }, + { + "step": 6526464, + "logs": { + "loss": 0.3, + "epoch": 6.284658040665434, + "learning_rate": 7.904982892201203e-06 + } + }, + { + "step": 6545664, + "logs": { + "loss": 0.3, + "epoch": 6.303142329020333, + "learning_rate": 7.865654619105676e-06 + } + }, + { + "step": 6564864, + "logs": { + "loss": 0.3003, + "epoch": 6.321626617375231, + "learning_rate": 7.826326346010146e-06 + } + }, + { + "step": 6584064, + "logs": { + "loss": 0.3029, + "epoch": 6.340110905730129, + "learning_rate": 7.786998072914618e-06 + } + }, + { + "step": 6603264, + "logs": { + "loss": 0.301, + "epoch": 6.358595194085027, + "learning_rate": 7.74766979981909e-06 + } + }, + { + "step": 6622464, + "logs": { + "loss": 0.297, + "epoch": 6.377079482439926, + "learning_rate": 7.708341526723561e-06 + } + }, + { + "step": 6641664, + "logs": { + "loss": 0.2993, + "epoch": 6.3955637707948245, + "learning_rate": 7.669013253628034e-06 + } + }, + { + "step": 6660864, + "logs": { + "loss": 0.3035, + "epoch": 6.414048059149723, + "learning_rate": 7.629684980532506e-06 + } + }, + { + "step": 6680064, + "logs": { + "loss": 0.2988, + "epoch": 6.432532347504621, + "learning_rate": 7.590356707436977e-06 + } + }, + { + "step": 6699264, + "logs": { + "loss": 0.3001, + "epoch": 6.45101663585952, + "learning_rate": 7.551028434341449e-06 + } + }, + { + "step": 6718464, + "logs": { + "loss": 0.3024, + "epoch": 6.469500924214418, + "learning_rate": 7.51170016124592e-06 + } + }, + { + "step": 6737664, + "logs": { + "loss": 0.3053, + "epoch": 6.487985212569316, + "learning_rate": 7.472371888150392e-06 + } + }, + { + "step": 6756864, + "logs": { + "loss": 0.3081, + "epoch": 6.506469500924214, + "learning_rate": 7.433043615054863e-06 + } + }, + { + "step": 6776064, + "logs": { + "loss": 0.3002, + "epoch": 6.524953789279113, + "learning_rate": 7.393715341959336e-06 + } + }, + { + "step": 6795264, + "logs": { + "loss": 0.3044, + "epoch": 6.543438077634011, + "learning_rate": 7.354387068863806e-06 + } + }, + { + "step": 6814464, + "logs": { + "loss": 0.3016, + "epoch": 6.561922365988909, + "learning_rate": 7.3150587957682786e-06 + } + }, + { + "step": 6833664, + "logs": { + "loss": 0.3071, + "epoch": 6.580406654343808, + "learning_rate": 7.27573052267275e-06 + } + }, + { + "step": 6852864, + "logs": { + "loss": 0.3002, + "epoch": 6.598890942698706, + "learning_rate": 7.2364022495772215e-06 + } + }, + { + "step": 6872064, + "logs": { + "loss": 0.3042, + "epoch": 6.617375231053605, + "learning_rate": 7.197073976481693e-06 + } + }, + { + "step": 6891264, + "logs": { + "loss": 0.3042, + "epoch": 6.635859519408503, + "learning_rate": 7.157745703386165e-06 + } + }, + { + "step": 6910464, + "logs": { + "loss": 0.3063, + "epoch": 6.654343807763401, + "learning_rate": 7.118417430290636e-06 + } + }, + { + "step": 6929664, + "logs": { + "loss": 0.3031, + "epoch": 6.672828096118299, + "learning_rate": 7.079089157195108e-06 + } + }, + { + "step": 6948864, + "logs": { + "loss": 0.3081, + "epoch": 6.691312384473198, + "learning_rate": 7.0397608840995805e-06 + } + }, + { + "step": 6968064, + "logs": { + "loss": 0.3041, + "epoch": 6.709796672828096, + "learning_rate": 7.000432611004051e-06 + } + }, + { + "step": 6987264, + "logs": { + "loss": 0.3036, + "epoch": 6.728280961182994, + "learning_rate": 6.961104337908523e-06 + } + }, + { + "step": 7006464, + "logs": { + "loss": 0.3044, + "epoch": 6.7467652495378925, + "learning_rate": 6.921776064812995e-06 + } + }, + { + "step": 7025664, + "logs": { + "loss": 0.3057, + "epoch": 6.7652495378927915, + "learning_rate": 6.882447791717466e-06 + } + }, + { + "step": 7044864, + "logs": { + "loss": 0.3081, + "epoch": 6.78373382624769, + "learning_rate": 6.843119518621938e-06 + } + }, + { + "step": 7064064, + "logs": { + "loss": 0.3058, + "epoch": 6.802218114602588, + "learning_rate": 6.80379124552641e-06 + } + }, + { + "step": 7083264, + "logs": { + "loss": 0.3072, + "epoch": 6.820702402957486, + "learning_rate": 6.764462972430881e-06 + } + }, + { + "step": 7102464, + "logs": { + "loss": 0.3042, + "epoch": 6.839186691312385, + "learning_rate": 6.725134699335353e-06 + } + }, + { + "step": 7121664, + "logs": { + "loss": 0.3091, + "epoch": 6.857670979667283, + "learning_rate": 6.6858064262398245e-06 + } + }, + { + "step": 7140864, + "logs": { + "loss": 0.3036, + "epoch": 6.876155268022181, + "learning_rate": 6.646478153144296e-06 + } + }, + { + "step": 7160064, + "logs": { + "loss": 0.3081, + "epoch": 6.894639556377079, + "learning_rate": 6.607149880048767e-06 + } + }, + { + "step": 7179264, + "logs": { + "loss": 0.3018, + "epoch": 6.913123844731977, + "learning_rate": 6.56782160695324e-06 + } + }, + { + "step": 7198464, + "logs": { + "loss": 0.3055, + "epoch": 6.931608133086876, + "learning_rate": 6.52849333385771e-06 + } + }, + { + "step": 7217664, + "logs": { + "loss": 0.3046, + "epoch": 6.9500924214417745, + "learning_rate": 6.489165060762183e-06 + } + }, + { + "step": 7236864, + "logs": { + "loss": 0.3083, + "epoch": 6.968576709796673, + "learning_rate": 6.449836787666655e-06 + } + }, + { + "step": 7256064, + "logs": { + "loss": 0.3118, + "epoch": 6.987060998151571, + "learning_rate": 6.4105085145711256e-06 + } + }, + { + "step": 7275008, + "logs": { + "loss": 0.2902, + "epoch": 7.00554528650647, + "learning_rate": 6.371180241475597e-06 + } + }, + { + "step": 7294208, + "logs": { + "loss": 0.241, + "epoch": 7.024029574861368, + "learning_rate": 6.331851968380069e-06 + } + }, + { + "step": 7313408, + "logs": { + "loss": 0.2432, + "epoch": 7.042513863216266, + "learning_rate": 6.29252369528454e-06 + } + }, + { + "step": 7332608, + "logs": { + "loss": 0.2405, + "epoch": 7.060998151571164, + "learning_rate": 6.253195422189012e-06 + } + }, + { + "step": 7351808, + "logs": { + "loss": 0.2447, + "epoch": 7.079482439926063, + "learning_rate": 6.2138671490934846e-06 + } + }, + { + "step": 7371008, + "logs": { + "loss": 0.2458, + "epoch": 7.097966728280961, + "learning_rate": 6.174538875997955e-06 + } + }, + { + "step": 7390208, + "logs": { + "loss": 0.2414, + "epoch": 7.116451016635859, + "learning_rate": 6.1352106029024275e-06 + } + }, + { + "step": 7409408, + "logs": { + "loss": 0.2465, + "epoch": 7.134935304990758, + "learning_rate": 6.095882329806899e-06 + } + }, + { + "step": 7428608, + "logs": { + "loss": 0.2474, + "epoch": 7.153419593345657, + "learning_rate": 6.05655405671137e-06 + } + }, + { + "step": 7447808, + "logs": { + "loss": 0.2464, + "epoch": 7.171903881700555, + "learning_rate": 6.017225783615842e-06 + } + }, + { + "step": 7467008, + "logs": { + "loss": 0.2443, + "epoch": 7.190388170055453, + "learning_rate": 5.977897510520314e-06 + } + }, + { + "step": 7486208, + "logs": { + "loss": 0.2466, + "epoch": 7.208872458410351, + "learning_rate": 5.938569237424785e-06 + } + }, + { + "step": 7505408, + "logs": { + "loss": 0.2497, + "epoch": 7.22735674676525, + "learning_rate": 5.899240964329257e-06 + } + }, + { + "step": 7524608, + "logs": { + "loss": 0.2469, + "epoch": 7.245841035120148, + "learning_rate": 5.859912691233728e-06 + } + }, + { + "step": 7543808, + "logs": { + "loss": 0.2468, + "epoch": 7.264325323475046, + "learning_rate": 5.8205844181382e-06 + } + }, + { + "step": 7563008, + "logs": { + "loss": 0.2495, + "epoch": 7.282809611829944, + "learning_rate": 5.7812561450426715e-06 + } + }, + { + "step": 7582208, + "logs": { + "loss": 0.2496, + "epoch": 7.3012939001848425, + "learning_rate": 5.741927871947143e-06 + } + }, + { + "step": 7601408, + "logs": { + "loss": 0.2516, + "epoch": 7.3197781885397415, + "learning_rate": 5.702599598851614e-06 + } + }, + { + "step": 7620608, + "logs": { + "loss": 0.2516, + "epoch": 7.33826247689464, + "learning_rate": 5.663271325756087e-06 + } + }, + { + "step": 7639808, + "logs": { + "loss": 0.2493, + "epoch": 7.356746765249538, + "learning_rate": 5.623943052660557e-06 + } + }, + { + "step": 7659008, + "logs": { + "loss": 0.2458, + "epoch": 7.375231053604436, + "learning_rate": 5.58461477956503e-06 + } + }, + { + "step": 7678208, + "logs": { + "loss": 0.2494, + "epoch": 7.393715341959335, + "learning_rate": 5.545286506469502e-06 + } + }, + { + "step": 7697408, + "logs": { + "loss": 0.2534, + "epoch": 7.412199630314233, + "learning_rate": 5.5059582333739726e-06 + } + }, + { + "step": 7716608, + "logs": { + "loss": 0.2499, + "epoch": 7.430683918669131, + "learning_rate": 5.466629960278445e-06 + } + }, + { + "step": 7735808, + "logs": { + "loss": 0.249, + "epoch": 7.449168207024029, + "learning_rate": 5.427301687182916e-06 + } + }, + { + "step": 7755008, + "logs": { + "loss": 0.2498, + "epoch": 7.467652495378928, + "learning_rate": 5.387973414087388e-06 + } + }, + { + "step": 7774208, + "logs": { + "loss": 0.2541, + "epoch": 7.486136783733826, + "learning_rate": 5.348645140991859e-06 + } + }, + { + "step": 7793408, + "logs": { + "loss": 0.2579, + "epoch": 7.5046210720887245, + "learning_rate": 5.3093168678963316e-06 + } + }, + { + "step": 7812608, + "logs": { + "loss": 0.2476, + "epoch": 7.523105360443623, + "learning_rate": 5.269988594800802e-06 + } + }, + { + "step": 7831808, + "logs": { + "loss": 0.2538, + "epoch": 7.541589648798522, + "learning_rate": 5.2306603217052745e-06 + } + }, + { + "step": 7851008, + "logs": { + "loss": 0.2537, + "epoch": 7.56007393715342, + "learning_rate": 5.191332048609746e-06 + } + }, + { + "step": 7870208, + "logs": { + "loss": 0.2522, + "epoch": 7.578558225508318, + "learning_rate": 5.152003775514217e-06 + } + }, + { + "step": 7889408, + "logs": { + "loss": 0.2491, + "epoch": 7.597042513863216, + "learning_rate": 5.112675502418689e-06 + } + }, + { + "step": 7908608, + "logs": { + "loss": 0.2509, + "epoch": 7.615526802218114, + "learning_rate": 5.073347229323161e-06 + } + }, + { + "step": 7927808, + "logs": { + "loss": 0.2496, + "epoch": 7.634011090573013, + "learning_rate": 5.034018956227632e-06 + } + }, + { + "step": 7947008, + "logs": { + "loss": 0.2542, + "epoch": 7.652495378927911, + "learning_rate": 4.994690683132104e-06 + } + }, + { + "step": 7966208, + "logs": { + "loss": 0.2517, + "epoch": 7.6709796672828094, + "learning_rate": 4.9553624100365756e-06 + } + }, + { + "step": 7985408, + "logs": { + "loss": 0.2544, + "epoch": 7.689463955637708, + "learning_rate": 4.916034136941048e-06 + } + }, + { + "step": 8004608, + "logs": { + "loss": 0.2489, + "epoch": 7.707948243992607, + "learning_rate": 4.876705863845519e-06 + } + }, + { + "step": 8023808, + "logs": { + "loss": 0.2506, + "epoch": 7.726432532347505, + "learning_rate": 4.83737759074999e-06 + } + }, + { + "step": 8043008, + "logs": { + "loss": 0.2535, + "epoch": 7.744916820702403, + "learning_rate": 4.798049317654462e-06 + } + }, + { + "step": 8062208, + "logs": { + "loss": 0.2506, + "epoch": 7.763401109057301, + "learning_rate": 4.758721044558934e-06 + } + }, + { + "step": 8081408, + "logs": { + "loss": 0.2547, + "epoch": 7.7818853974122, + "learning_rate": 4.719392771463405e-06 + } + }, + { + "step": 8100608, + "logs": { + "loss": 0.2525, + "epoch": 7.800369685767098, + "learning_rate": 4.6800644983678775e-06 + } + }, + { + "step": 8119808, + "logs": { + "loss": 0.2535, + "epoch": 7.818853974121996, + "learning_rate": 4.640736225272349e-06 + } + }, + { + "step": 8139008, + "logs": { + "loss": 0.2508, + "epoch": 7.837338262476894, + "learning_rate": 4.60140795217682e-06 + } + }, + { + "step": 8158208, + "logs": { + "loss": 0.254, + "epoch": 7.855822550831793, + "learning_rate": 4.562079679081292e-06 + } + }, + { + "step": 8177408, + "logs": { + "loss": 0.251, + "epoch": 7.8743068391866915, + "learning_rate": 4.522751405985763e-06 + } + }, + { + "step": 8196608, + "logs": { + "loss": 0.2545, + "epoch": 7.89279112754159, + "learning_rate": 4.483423132890235e-06 + } + }, + { + "step": 8215808, + "logs": { + "loss": 0.2497, + "epoch": 7.911275415896488, + "learning_rate": 4.444094859794707e-06 + } + }, + { + "step": 8235008, + "logs": { + "loss": 0.253, + "epoch": 7.929759704251387, + "learning_rate": 4.4047665866991786e-06 + } + }, + { + "step": 8254208, + "logs": { + "loss": 0.249, + "epoch": 7.948243992606285, + "learning_rate": 4.36543831360365e-06 + } + }, + { + "step": 8273408, + "logs": { + "loss": 0.2534, + "epoch": 7.966728280961183, + "learning_rate": 4.3261100405081215e-06 + } + }, + { + "step": 8292608, + "logs": { + "loss": 0.2563, + "epoch": 7.985212569316081, + "learning_rate": 4.286781767412593e-06 + } + }, + { + "step": 8311552, + "logs": { + "loss": 0.2489, + "epoch": 8.00369685767098, + "learning_rate": 4.247453494317064e-06 + } + }, + { + "step": 8330752, + "logs": { + "loss": 0.2079, + "epoch": 8.022181146025877, + "learning_rate": 4.208125221221537e-06 + } + }, + { + "step": 8349952, + "logs": { + "loss": 0.2079, + "epoch": 8.040665434380776, + "learning_rate": 4.168796948126008e-06 + } + }, + { + "step": 8369152, + "logs": { + "loss": 0.2055, + "epoch": 8.059149722735675, + "learning_rate": 4.12946867503048e-06 + } + }, + { + "step": 8388352, + "logs": { + "loss": 0.2115, + "epoch": 8.077634011090574, + "learning_rate": 4.090140401934951e-06 + } + }, + { + "step": 8407552, + "logs": { + "loss": 0.2105, + "epoch": 8.096118299445472, + "learning_rate": 4.050812128839423e-06 + } + }, + { + "step": 8426752, + "logs": { + "loss": 0.2055, + "epoch": 8.11460258780037, + "learning_rate": 4.011483855743895e-06 + } + }, + { + "step": 8445952, + "logs": { + "loss": 0.2086, + "epoch": 8.133086876155268, + "learning_rate": 3.972155582648366e-06 + } + }, + { + "step": 8465152, + "logs": { + "loss": 0.2102, + "epoch": 8.151571164510166, + "learning_rate": 3.932827309552838e-06 + } + }, + { + "step": 8484352, + "logs": { + "loss": 0.21, + "epoch": 8.170055452865064, + "learning_rate": 3.893499036457309e-06 + } + }, + { + "step": 8503552, + "logs": { + "loss": 0.2081, + "epoch": 8.188539741219962, + "learning_rate": 3.854170763361781e-06 + } + }, + { + "step": 8522752, + "logs": { + "loss": 0.213, + "epoch": 8.207024029574862, + "learning_rate": 3.814842490266253e-06 + } + }, + { + "step": 8541952, + "logs": { + "loss": 0.2114, + "epoch": 8.22550831792976, + "learning_rate": 3.7755142171707245e-06 + } + }, + { + "step": 8561152, + "logs": { + "loss": 0.2116, + "epoch": 8.243992606284658, + "learning_rate": 3.736185944075196e-06 + } + }, + { + "step": 8580352, + "logs": { + "loss": 0.2106, + "epoch": 8.262476894639557, + "learning_rate": 3.696857670979668e-06 + } + }, + { + "step": 8599552, + "logs": { + "loss": 0.2097, + "epoch": 8.280961182994455, + "learning_rate": 3.6575293978841393e-06 + } + }, + { + "step": 8618752, + "logs": { + "loss": 0.2119, + "epoch": 8.299445471349353, + "learning_rate": 3.6182011247886107e-06 + } + }, + { + "step": 8637952, + "logs": { + "loss": 0.2144, + "epoch": 8.317929759704251, + "learning_rate": 3.5788728516930826e-06 + } + }, + { + "step": 8657152, + "logs": { + "loss": 0.2125, + "epoch": 8.336414048059149, + "learning_rate": 3.539544578597554e-06 + } + }, + { + "step": 8676352, + "logs": { + "loss": 0.2134, + "epoch": 8.354898336414047, + "learning_rate": 3.5002163055020256e-06 + } + }, + { + "step": 8695552, + "logs": { + "loss": 0.2127, + "epoch": 8.373382624768947, + "learning_rate": 3.4608880324064974e-06 + } + }, + { + "step": 8714752, + "logs": { + "loss": 0.2111, + "epoch": 8.391866913123845, + "learning_rate": 3.421559759310969e-06 + } + }, + { + "step": 8733952, + "logs": { + "loss": 0.213, + "epoch": 8.410351201478743, + "learning_rate": 3.3822314862154404e-06 + } + }, + { + "step": 8753152, + "logs": { + "loss": 0.2109, + "epoch": 8.428835489833642, + "learning_rate": 3.3429032131199122e-06 + } + }, + { + "step": 8772352, + "logs": { + "loss": 0.2093, + "epoch": 8.44731977818854, + "learning_rate": 3.3035749400243837e-06 + } + }, + { + "step": 8791552, + "logs": { + "loss": 0.2126, + "epoch": 8.465804066543438, + "learning_rate": 3.264246666928855e-06 + } + }, + { + "step": 8810752, + "logs": { + "loss": 0.2127, + "epoch": 8.484288354898336, + "learning_rate": 3.2249183938333275e-06 + } + }, + { + "step": 8829952, + "logs": { + "loss": 0.2184, + "epoch": 8.502772643253234, + "learning_rate": 3.1855901207377985e-06 + } + }, + { + "step": 8849152, + "logs": { + "loss": 0.2108, + "epoch": 8.521256931608134, + "learning_rate": 3.14626184764227e-06 + } + }, + { + "step": 8868352, + "logs": { + "loss": 0.2123, + "epoch": 8.539741219963032, + "learning_rate": 3.1069335745467423e-06 + } + }, + { + "step": 8887552, + "logs": { + "loss": 0.2125, + "epoch": 8.55822550831793, + "learning_rate": 3.0676053014512137e-06 + } + }, + { + "step": 8906752, + "logs": { + "loss": 0.2124, + "epoch": 8.576709796672828, + "learning_rate": 3.028277028355685e-06 + } + }, + { + "step": 8925952, + "logs": { + "loss": 0.2112, + "epoch": 8.595194085027726, + "learning_rate": 2.988948755260157e-06 + } + }, + { + "step": 8945152, + "logs": { + "loss": 0.2103, + "epoch": 8.613678373382625, + "learning_rate": 2.9496204821646285e-06 + } + }, + { + "step": 8964352, + "logs": { + "loss": 0.2128, + "epoch": 8.632162661737523, + "learning_rate": 2.9102922090691e-06 + } + }, + { + "step": 8983552, + "logs": { + "loss": 0.2125, + "epoch": 8.65064695009242, + "learning_rate": 2.8709639359735715e-06 + } + }, + { + "step": 9002752, + "logs": { + "loss": 0.2109, + "epoch": 8.669131238447319, + "learning_rate": 2.8316356628780434e-06 + } + }, + { + "step": 9021952, + "logs": { + "loss": 0.2132, + "epoch": 8.687615526802219, + "learning_rate": 2.792307389782515e-06 + } + }, + { + "step": 9041152, + "logs": { + "loss": 0.2113, + "epoch": 8.706099815157117, + "learning_rate": 2.7529791166869863e-06 + } + }, + { + "step": 9060352, + "logs": { + "loss": 0.2109, + "epoch": 8.724584103512015, + "learning_rate": 2.713650843591458e-06 + } + }, + { + "step": 9079552, + "logs": { + "loss": 0.2138, + "epoch": 8.743068391866913, + "learning_rate": 2.6743225704959296e-06 + } + }, + { + "step": 9098752, + "logs": { + "loss": 0.2098, + "epoch": 8.761552680221811, + "learning_rate": 2.634994297400401e-06 + } + }, + { + "step": 9117952, + "logs": { + "loss": 0.2128, + "epoch": 8.78003696857671, + "learning_rate": 2.595666024304873e-06 + } + }, + { + "step": 9137152, + "logs": { + "loss": 0.2143, + "epoch": 8.798521256931608, + "learning_rate": 2.5563377512093444e-06 + } + }, + { + "step": 9156352, + "logs": { + "loss": 0.2132, + "epoch": 8.817005545286506, + "learning_rate": 2.517009478113816e-06 + } + }, + { + "step": 9175552, + "logs": { + "loss": 0.2099, + "epoch": 8.835489833641406, + "learning_rate": 2.4776812050182878e-06 + } + }, + { + "step": 9194752, + "logs": { + "loss": 0.2124, + "epoch": 8.853974121996304, + "learning_rate": 2.4383529319227597e-06 + } + }, + { + "step": 9213952, + "logs": { + "loss": 0.2122, + "epoch": 8.872458410351202, + "learning_rate": 2.399024658827231e-06 + } + }, + { + "step": 9233152, + "logs": { + "loss": 0.2136, + "epoch": 8.8909426987061, + "learning_rate": 2.3596963857317026e-06 + } + }, + { + "step": 9252352, + "logs": { + "loss": 0.2103, + "epoch": 8.909426987060998, + "learning_rate": 2.3203681126361745e-06 + } + }, + { + "step": 9271552, + "logs": { + "loss": 0.2132, + "epoch": 8.927911275415896, + "learning_rate": 2.281039839540646e-06 + } + }, + { + "step": 9290752, + "logs": { + "loss": 0.2082, + "epoch": 8.946395563770794, + "learning_rate": 2.2417115664451174e-06 + } + }, + { + "step": 9309952, + "logs": { + "loss": 0.213, + "epoch": 8.964879852125692, + "learning_rate": 2.2023832933495893e-06 + } + }, + { + "step": 9329152, + "logs": { + "loss": 0.2142, + "epoch": 8.98336414048059, + "learning_rate": 2.1630550202540607e-06 + } + }, + { + "step": 9348096, + "logs": { + "loss": 0.2123, + "epoch": 9.00184842883549, + "learning_rate": 2.123726747158532e-06 + } + }, + { + "step": 9367296, + "logs": { + "loss": 0.1854, + "epoch": 9.020332717190389, + "learning_rate": 2.084398474063004e-06 + } + }, + { + "step": 9386496, + "logs": { + "loss": 0.1848, + "epoch": 9.038817005545287, + "learning_rate": 2.0450702009674755e-06 + } + }, + { + "step": 9405696, + "logs": { + "loss": 0.1802, + "epoch": 9.057301293900185, + "learning_rate": 2.0057419278719474e-06 + } + }, + { + "step": 9424896, + "logs": { + "loss": 0.1859, + "epoch": 9.075785582255083, + "learning_rate": 1.966413654776419e-06 + } + }, + { + "step": 9444096, + "logs": { + "loss": 0.1849, + "epoch": 9.094269870609981, + "learning_rate": 1.9270853816808904e-06 + } + }, + { + "step": 9463296, + "logs": { + "loss": 0.1841, + "epoch": 9.11275415896488, + "learning_rate": 1.8877571085853622e-06 + } + }, + { + "step": 9482496, + "logs": { + "loss": 0.1854, + "epoch": 9.131238447319777, + "learning_rate": 1.848428835489834e-06 + } + }, + { + "step": 9501696, + "logs": { + "loss": 0.1847, + "epoch": 9.149722735674677, + "learning_rate": 1.8091005623943054e-06 + } + }, + { + "step": 9520896, + "logs": { + "loss": 0.1852, + "epoch": 9.168207024029575, + "learning_rate": 1.769772289298777e-06 + } + }, + { + "step": 9540096, + "logs": { + "loss": 0.1846, + "epoch": 9.186691312384474, + "learning_rate": 1.7304440162032487e-06 + } + }, + { + "step": 9559296, + "logs": { + "loss": 0.1848, + "epoch": 9.205175600739372, + "learning_rate": 1.6911157431077202e-06 + } + }, + { + "step": 9578496, + "logs": { + "loss": 0.1861, + "epoch": 9.22365988909427, + "learning_rate": 1.6517874700121919e-06 + } + }, + { + "step": 9597696, + "logs": { + "loss": 0.185, + "epoch": 9.242144177449168, + "learning_rate": 1.6124591969166637e-06 + } + }, + { + "step": 9616896, + "logs": { + "loss": 0.1846, + "epoch": 9.260628465804066, + "learning_rate": 1.573130923821135e-06 + } + }, + { + "step": 9636096, + "logs": { + "loss": 0.1827, + "epoch": 9.279112754158964, + "learning_rate": 1.5338026507256069e-06 + } + }, + { + "step": 9655296, + "logs": { + "loss": 0.1871, + "epoch": 9.297597042513864, + "learning_rate": 1.4944743776300785e-06 + } + }, + { + "step": 9674496, + "logs": { + "loss": 0.1865, + "epoch": 9.316081330868762, + "learning_rate": 1.45514610453455e-06 + } + }, + { + "step": 9693696, + "logs": { + "loss": 0.1859, + "epoch": 9.33456561922366, + "learning_rate": 1.4158178314390217e-06 + } + }, + { + "step": 9712896, + "logs": { + "loss": 0.1874, + "epoch": 9.353049907578558, + "learning_rate": 1.3764895583434931e-06 + } + }, + { + "step": 9732096, + "logs": { + "loss": 0.1847, + "epoch": 9.371534195933457, + "learning_rate": 1.3371612852479648e-06 + } + }, + { + "step": 9751296, + "logs": { + "loss": 0.1849, + "epoch": 9.390018484288355, + "learning_rate": 1.2978330121524365e-06 + } + }, + { + "step": 9770496, + "logs": { + "loss": 0.1895, + "epoch": 9.408502772643253, + "learning_rate": 1.258504739056908e-06 + } + }, + { + "step": 9789696, + "logs": { + "loss": 0.1829, + "epoch": 9.426987060998151, + "learning_rate": 1.2191764659613798e-06 + } + }, + { + "step": 9808896, + "logs": { + "loss": 0.1837, + "epoch": 9.445471349353049, + "learning_rate": 1.1798481928658513e-06 + } + }, + { + "step": 9828096, + "logs": { + "loss": 0.1864, + "epoch": 9.463955637707949, + "learning_rate": 1.140519919770323e-06 + } + }, + { + "step": 9847296, + "logs": { + "loss": 0.1862, + "epoch": 9.482439926062847, + "learning_rate": 1.1011916466747946e-06 + } + }, + { + "step": 9866496, + "logs": { + "loss": 0.19, + "epoch": 9.500924214417745, + "learning_rate": 1.061863373579266e-06 + } + }, + { + "step": 9885696, + "logs": { + "loss": 0.1834, + "epoch": 9.519408502772643, + "learning_rate": 1.0225351004837378e-06 + } + }, + { + "step": 9904896, + "logs": { + "loss": 0.1863, + "epoch": 9.537892791127542, + "learning_rate": 9.832068273882094e-07 + } + }, + { + "step": 9924096, + "logs": { + "loss": 0.1859, + "epoch": 9.55637707948244, + "learning_rate": 9.438785542926811e-07 + } + }, + { + "step": 9943296, + "logs": { + "loss": 0.1829, + "epoch": 9.574861367837338, + "learning_rate": 9.045502811971527e-07 + } + }, + { + "step": 9962496, + "logs": { + "loss": 0.1838, + "epoch": 9.593345656192236, + "learning_rate": 8.652220081016244e-07 + } + }, + { + "step": 9981696, + "logs": { + "loss": 0.1817, + "epoch": 9.611829944547136, + "learning_rate": 8.258937350060959e-07 + } + }, + { + "step": 10000896, + "logs": { + "loss": 0.1834, + "epoch": 9.630314232902034, + "learning_rate": 7.865654619105675e-07 + } + }, + { + "step": 10020096, + "logs": { + "loss": 0.1857, + "epoch": 9.648798521256932, + "learning_rate": 7.472371888150393e-07 + } + }, + { + "step": 10039296, + "logs": { + "loss": 0.1858, + "epoch": 9.66728280961183, + "learning_rate": 7.079089157195108e-07 + } + }, + { + "step": 10058496, + "logs": { + "loss": 0.1858, + "epoch": 9.685767097966728, + "learning_rate": 6.685806426239824e-07 + } + }, + { + "step": 10077696, + "logs": { + "loss": 0.1848, + "epoch": 9.704251386321626, + "learning_rate": 6.29252369528454e-07 + } + }, + { + "step": 10096896, + "logs": { + "loss": 0.1832, + "epoch": 9.722735674676525, + "learning_rate": 5.899240964329256e-07 + } + }, + { + "step": 10116096, + "logs": { + "loss": 0.1842, + "epoch": 9.741219963031423, + "learning_rate": 5.505958233373973e-07 + } + }, + { + "step": 10135296, + "logs": { + "loss": 0.1842, + "epoch": 9.75970425138632, + "learning_rate": 5.112675502418689e-07 + } + }, + { + "step": 10154496, + "logs": { + "loss": 0.1847, + "epoch": 9.77818853974122, + "learning_rate": 4.7193927714634056e-07 + } + }, + { + "step": 10173696, + "logs": { + "loss": 0.185, + "epoch": 9.796672828096119, + "learning_rate": 4.326110040508122e-07 + } + }, + { + "step": 10192896, + "logs": { + "loss": 0.1826, + "epoch": 9.815157116451017, + "learning_rate": 3.9328273095528375e-07 + } + }, + { + "step": 10212096, + "logs": { + "loss": 0.1833, + "epoch": 9.833641404805915, + "learning_rate": 3.539544578597554e-07 + } + }, + { + "step": 10231296, + "logs": { + "loss": 0.1846, + "epoch": 9.852125693160813, + "learning_rate": 3.14626184764227e-07 + } + }, + { + "step": 10250496, + "logs": { + "loss": 0.184, + "epoch": 9.870609981515711, + "learning_rate": 2.7529791166869866e-07 + } + }, + { + "step": 10269696, + "logs": { + "loss": 0.1864, + "epoch": 9.88909426987061, + "learning_rate": 2.3596963857317028e-07 + } + }, + { + "step": 10288896, + "logs": { + "loss": 0.1841, + "epoch": 9.907578558225508, + "learning_rate": 1.9664136547764187e-07 + } + }, + { + "step": 10308096, + "logs": { + "loss": 0.1855, + "epoch": 9.926062846580407, + "learning_rate": 1.573130923821135e-07 + } + }, + { + "step": 10327296, + "logs": { + "loss": 0.1794, + "epoch": 9.944547134935306, + "learning_rate": 1.1798481928658514e-07 + } + }, + { + "step": 10346496, + "logs": { + "loss": 0.1853, + "epoch": 9.963031423290204, + "learning_rate": 7.865654619105675e-08 + } + }, + { + "step": 10365696, + "logs": { + "loss": 0.1855, + "epoch": 9.981515711645102, + "learning_rate": 3.9328273095528373e-08 + } + } + ], + "Evaluation": [ + { + "step": 2705, + "logs": { + "eval_loss": 0.9760558605194092 + } + }, + { + "step": 5410, + "logs": { + "eval_loss": 0.9268760681152344 + } + }, + { + "step": 8115, + "logs": { + "eval_loss": 0.9153628945350647 + } + }, + { + "step": 10820, + "logs": { + "eval_loss": 0.9207033514976501 + } + }, + { + "step": 13525, + "logs": { + "eval_loss": 0.9357024431228638 + } + }, + { + "step": 16230, + "logs": { + "eval_loss": 0.948551595211029 + } + }, + { + "step": 18935, + "logs": { + "eval_loss": 0.992328941822052 + } + }, + { + "step": 21640, + "logs": { + "eval_loss": 1.0297634601593018 + } + }, + { + "step": 24345, + "logs": { + "eval_loss": 1.0606719255447388 + } + }, + { + "step": 27050, + "logs": { + "eval_loss": 1.0935026407241821 + } + } + ] + } +} \ No newline at end of file